Differential and Integral Calculus - N Piskunov

N.
PISKUNOV
DIFFERENTIAL
and
INTEGRAL CALCULUS
Milt PUBLISHERS
Moscow
1969
TRANSLATED FROM THE RUSSIAN
BY G. YANKOVSKY
H. c. nncKyHOB
.IUf<ll<llEPEHUHAJibHOE 11 HHTEfPAJibHOE.
HCI.JHCJIEHH51
Ha aHeAuliCKOM R3bt/CI
CONTENTS
Preface 11
Chapter /. NUMBER. VARIABLE. FUNCTION
1. Real Numbers. Real Numbers as Points on a Number Scale 13

2. The Absolute Value of a Real Number . . . . . . . . . 14
3. Variables and Constants . . . . . . . . . . . . . . . . . 16
4. The Range of a Variable . . . . . . . . . . . . . . . . 16
5. Ordered Variables. Increasing and Decreasing Variables. Bounded
Variables . . . . . . . . . 18
6. Function . . . . . . . . . . . . . . . 19
7. Ways of Representing Functions . . . . . . . 20
8. Basic Elementary Functions. Elementary Functions 22
9. Algebraic Functions . . 26
10. Polar Coordinate System 28
Exercises on Chapter I 30
Chapter 1/. LIMIT. CONTINUITY OF A FUNCTION
1. The Limit of a Variable. An Infinitely Large Variable . 32

2. The Limit of a Function . . . . . . . . . . . . . . 35
3. A Function that Approaches Infinity. Bounded Functions 38
4. Infinitesimals and Their Basic Properties 42
5. Basic Theorems on Limits . . . . . . . 45
6. The Limit of the Function sin x as x-+ 0 50
X
7. The Number e . . . . . 51
8. Natural Logarithms . . . . . . . . . 56
9. Continuity of Functions . . . . . . . 57
10. Certain Properties of Continuous Functions 61
II. Comparing Infinitesimals 63
Exercises on Chapter /1 . . . . . . . 65
Chapter 1//. DERIVATIVE AND DIFFERENTIAL
1. Velocity of Motion . . . . . . . . . . . 69
2. Definition of Derivative . . . . . . . . . . 71
3. Geometric Meaning of the Derivative . 73
4. Differentiability of Functions . . . . . . . . .... 74
5. Finding the Derivatives of Elementary Functions. The Derivative of
the Function y=xn, Where n ls Positive and Integral . . . 76
6. Derivatives of the Functions y=sinx; y=cosx . . . . . . . . . . 78
7. Derivatives of: a Constant, the Product of a Constant by a Function,
a Sum, a Product, and a Quotient . . . . . . . . . . 80
8. The Derivative of a Logarithmic Function . . . . . 84
9. The Derivative of a Composite Function . . . . . 85
10. Derivatives of the Functions y=tanx, y=cotx, y=lnjxJ 83
,.
4 Contents
11. An Implicit Function and Its Differentiation . . . . . . . . . . . 89

12. Derivatives of a Power Function for an Arbitrary Real Exponent, of
an Exponential Function, and a Composite Exponential Function 91
13. An Inverse Function and Its Differentiation . . . . . . . 94
]4. Inverse Trigonometric Functions and Their Differentiation 98
15. Table of Basic Differentiation Formulas . . . . . . 102
16. Parametric Representation of a Function . . . . . . . . 103
17. The Equations of Certain Curves in Parametric Form 105
18. The Derivative of a Function Represented Parametrically 108
19. Hyperbolic Functions . . . . . . . . . 110
20. The Differential . . . . . . . . . . . . . . . . . 113
21. The Geometric Significance of the Differential . 117
22. Derivatives of Different Orders . . . . . 118
23. Differentials of Various Orders . . . . . . . . . . . 121
24. Different-Order Derivatives of Implicit Functions and of Functions
Represented Parametrically . . . . . . . . . . . . . . 122
25. The Mechanical Significance of the Second Derivative . . . . . . . 124
26. The Equations of a Tangent and of a Normal. The Lengths of .the
Subtangent and the Subnormal . . . . . . . . . . . , . . . 126
27. The Geometric Significance of the Derivative of the Radius Vector
with Respect to the Polar Angle . . . . 129
Exercises on Chapter Ill . . . . . . . . . . . . . . . . . . . 130
Chapter IV. SOME THEOREMS ON DIFFERENTIABLE FUNCTIONS
1. A Theorem on the Roots of a Derivative (Rolle's Theorem) . 140

2. A Theorem on Finite Increments (Lagrange's Theorem) . . . . 142
3. A Theorem on the Ratio of the Increments of Two Functions
(Cauchy's Theorem) . . . . . . . 143
4. The Limit of a Ratio of Two Infinitesimals (Evaluation of lndeter-
. '
mmate Forms of the Type
0
0) . . . . . . . . . .. 144
5. The Limit of a Ratio of Two Infinitely Large Quantities (Evaluation
of Indeterminate Forms of the Type : ) . . . . . 147

6. Taylor's Formula . . . . . . . . . . . . . . 152
7. Expansion of the Functions e", sin x, and cos x in a Taylor Series !56
Exercises on Chapter IV . . . . . . . . . . . 159
Chapter V. INVESTIGATING THE BEHAVIOUR OF FUNCTIONS
1. Statement of the Problem . . . . . . . . . 162

2. Increase and Decrease of a Function . . . . . . . . . . . 163
3. Maxima and Minima of Functions . . . . . . . . . . 164
4. Testing a Differentiable Function for Maximum and Minimum with
a First Derivative . . . . . . . . . . . . . . 171
5. Testin~ a Function for Maximum and Minimum with a Second
DerivatIVe . . . . . . . . . . . . . . 174
6. Maxima and Minima of a Function on an Interval .. 178
Contents 5
7. Applying the Theory of Maxima and Minima of Functions to the

Solution of Problems . . . . . . . . . . . . . . . . . . , , 179
8. Testing a Function for Maximum and Minimum by Means of Taylor's
Formula . . . . . . . . . . . . . . . . . . . . . 181
9. Convexity and Concavity of a Curve. Points of Inflection 183
10. Asymptotes . . . . . . . -. . . . . . . . . , . . 189
11. General Plan for Investigating Functions and Constructing Graphs 194
12. Investigating Curves Represented Parametrically . . . . . . 199
Exercises on Chapter V . . . . . . . . . . . . . . . . 203
Chapte,. VI. THE CURVATURE OF A CURVE
I. The Length of an Arc and Its Derivative . . . . ... 208

2. Curvature . . . . . . ,. . . . . . . . . . . . . . . 210
3. Calculation of Curvature . . . . . . . . . . . . . . 212
4. Calculation of the Curvature of a Line Represented Parametrically . . 215
5. Calculation of the Curva lure of a Line Given by an Equation of
Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . 215
6. The Radius and Circle of Curvature. Centre of Curvature. Evolute and
Involute . . . . . . . .... 217
7. The Propert!es of an Eyolute . . . . 221
8. Approximating the Real Roots of an Equation 225
E,"Cercises on Chapter VI . . . 229
Chapte,. Vll. COMPLEX NUMBERS. POLYNOMIALS
I. Complex Numbers. Basic Definitions . . . . . . . 233

2. Basic Operations on Complex Numbers ; . . . . . . 234
3. Powers and Roots of Complex Numbers . . . . . . . 237
4. Exponential Function with Complex Exponent and Its Properties 240
5. Euler's Formula. The Exponential Form of a Complex Number . . 243
6. Factoring a Polynomial . . . . . . . . . . . . . 244
7. The Multiple Roots of a Polynomial . . . . . 247
8. Factorisation of a Polynomial in the Ca5e of Complex Roots . 248
9. Interpolation. Lagrange's Interpolation Formula . . . . . 250
10. On the Best Approximation of Functions by Polynomials. Chebyshev's
Theory . 252
Exercises on Chapter VII 253
Chaptu V/1/. FUNCTIONS OF SEVERAL VARIABLES
l. Definition of a Function of Several Variables ..... 255

2. Geometric Representation of a Function of Two Variables 25&
3. Partial and Total Increment of a Function . . . . . 259
4. Continuity of a Function of Several Variables . . . . . . 260
5. Partial Derivatives of a Function of Several Variables . 263
6. The Geometric Interpretat!on of the Partial Derivatives of a Func-
tion of Two Variables . . . . . . . . . . 264
7. Total Increment and Total Differentials . . . . 265
8. Approximation by Total Differentials . . . . . . 268
9. Error Approximation by Differentials . . . . 270
10. The Derivative of a Composite Function. The Total Derivative 273
11. The Derivative of a Function Defined Implicitly . . . . . 276
12. Partial Derivatives of Different Orders . 279
6 Contents
13. Level Surfaces . . . . . . . . . . . . 283

14. Directional Derivatives . . . . . . . . . 284
15. Gradient . . . . . . . . . . . . . . 286
16. Taylor's Formula for a Function of Two Variables . . 290
17. Maximum and ..Minimum of a Function of Several Variables 292
18. Maximum and Minimum of a Function of Several Variables Related
by Given Equations (Conditional Maxima and Minima) 300
19. Singular Points of a Curve . . . . . . 305
Exerci~es on Chapter VIII . . . . . . . . . 310
Chapter IX. APPLICATIONS OF DIFFERENTIAL CALCULUS TO SOLID GEOMETRY
I. The Equations of a Curve in Space . . . . . . . . . . . . . . 314

2. The Limit and Derivative of the Vector Function of a Scalar
Argument. The Equation of a Tangent to a Curve. The Equation of a
Normal Plane . . . . . . . . . . . . . . . . . . . . 317
3. Rules for Differentiating Vectors (Vector Functions) . . . . . . . . 322
4. The First and Second Derivatives of a Vector with Respect to the
Arc Length. The Curvature of a Curve. The Principal Normal . . . 324
5. Osculating Plane. Binormal. Torsion 331
6. A Tangent Plane and Normal to a Surface 336
Exercises on Chapter IX . . 340
Chapter }(. INDEFINITE INTEGRALS
1. Antiderivative and the Indefinite Integral . . . . 342

2. Table of Integrals . . . . . . . . . . . . . 344
3. Some Properties of an Indefinite Integral . . . 346
4. Integration by Substitution (Change of Variable) . . . 348
5. Integrals of Functions Containing a Quadratic Trinomial 351
6. Integration by Parts . . . . . . . . . . . . . . . . . . . . 354
7. Rational Fractions. Partial Rational Fractions and Their Integration 357
8. Decomposition of a Rational Fraction into Partial Fractions . . . 361
9. lntt>gration of Rational Fractions . . . . . 365
Io. Ostrogradsky's Method . . . . . . . . . . . 368
II. Integrals of Irrational Functions . . . . . . . 371
12. Integrals of the Form ~ R (x, vax 2 + bx+c) dx 372
13. Integration of Binomial Differentials . . . . . . . . 375
14. Integration of Certain Classes of Trigonometric Functions . . . 378
15. Integration of Certain Irrational Functions by Means of Trigonometric
Substitutions . . . . . . . . . . . . . . 383
16. Functions Whose Integrals Cannot Be Expressed in Terms of
Elementary Functions 385
Exercises on Chapter X 386
Chapter XI. THE DEFINITE INTEGRAL
1. Statement of the Problem. The Lower and. Upper Integral Sums . 396
2. The Definite Integral . . . . . . . . . . . . . . . . . . 398
3. Basic Properties of the Definite Integral . . . . . . . . . 404
4. Evaluating a Definite Integral. Newton-Leibniz Formula . 407
5. Changing the Variable in the Definite Integral . 412
Contents 7
6. Integration by Parts . . . . . . 413

7. Improper Integrals . . . . . . . 416
8. Approximating Definite Integrals 424
9. Chebyshev's Formula . . . . . . . 430
10. Integrals Dependent on a Parameter . 435
Exercises on Chapter XI . . . . . . . 438
Chapter XII. GEOMETRIC AND MECHANICAL APPLICATIONS OF THE DEFINITE

INTEGRAL
I. Computing Areas in Rectangular Coordinates . . . . . . . . . . 442

2. The Area of a Curvilinear Sector in Polar Coordinates . . . . . . 445
3. The Arc Length of a Curve . . . . . . . . . . . . . . . . . . . . 44 7
4. Computing the Volume of a Solid from the Areas of Parallel Sections
(Volumes by Slicing) . . . . . . . . . . . 453
5. The Volume of a Solid of Revolution . 455
6. The Surface of a Solid of Revolution . . . . 455
7. Computing Work by the Definite Integral . 457
8. Coordinates of the Centre of Gravity . 459
Exercises on Chapter XII . . . . . . . . . . . 462
Chapter X/1/. DIFFERENTIAL EQUATIONS
I. Statement of the Problem. The Equation of Motion of a Body with

Resistance of the Medium Proportional to the Velocity. The
Equation of a Catenary . . . . . . . . . . . . . . . . . . . 469
2. Definitions . . . . . . . . . . . . . . . . . . . . . . . .. 472
3. First-Order Differential Equations (General Notions) ... 473
4. Equations with Separated and Separable Variables. The Problem of
the Disintegration of Radium . . . . . . . 478
5. Homogeneous First-Order Equations . . . . . 482
6. Equations Reducible to Homogeneous Equations . 484
7. First-Order Linear Equations 487
8. Bernoulli's Equation . . . . . . . . . 490
9. Exact Differential Equations . . . . . . . . . 492
I 0. Integrating Factor . . . . . . . . . . . . . . 495
II. The Envelope of a Family of Curves . . . . . . . . . . 497
12. Singular Solutions of a First-Order Diffrrential Equation 504
13. Clairaut's Equation . . . . . . . . . . . . . 505
14. Lagrange's Equation . . . . . . . . . . . . . 507
15. Orthogonal and Isogonal Trajectories . . . . . 509
16. Higher-Or~er Differential E~~la~ions (Fundamentals) 514
17. An Equat10n of the Form y -f(x) . . . . . . . . . . . . . . . . 516
18. Some Types of Second-Order Differential Equations Reducible to
First-Order Equations . . . . . . . . . . . . . . . . . . . . . . . 518
19. Graphical Method of Integrating Second-Order Differential Equations 527
20. Homogeneous Linear Equations. Definitions and General Properties 528
21. Second-Order Homogeneous Linear Equations with Constant Coeffi-
cients . . . . . . . , . . . . . . . . . . . . . . . . . . . 535
22. Homogeneous Linear Equations of the nth Order with Constant
Coefficients . . , , . . . , . . . . . . . . . . 539
23. Nonhomogeneous Second-Order Linear Equations . . . .... 541
24. Nonhomogeneous Second-Order Linear Equations with Constant
Coefficients , , , . . . . 545
8 Contents
25. Higher-Order Nonhomogeneous Linear Equations . . . 551

25. The Differential_ .Equation of Mechanical Vibrations 555
27. Free Oscillations . . . . . . . . . . . . . . 557
28. Forced Oscillations . . . . . . . . . . . . . 559
29. Systems of Ordinary Differential Equations . . . ..... 563
30. Systems of Linear Differential Equations with Constant Coefficients 569
31. On Lyapunov's Theory of Stability . . . . . . . . . . . 576
32. Euler's Method of Approximate Solution of First-Order Differential
Equations . . . . . . . . . . . . . . . . . . . . , . . . 581
33. A Difference Method for Approximate Solution of Differential Equa-
tions Based on Taylor's Formula. Adams Method . , , . . 584
34. An Approximate Method for Integrating Systems of First-Order
Differential Equations 591
Exercises on Chapter XIII . . . . . . . . 595
Chapter XIV. MULTIPLE INTEGRALS
1. Double Integrals . . . . . . . . . . . . . . . 608

2. Calculating Double Integrals . . . . . . . . . 610
3. Calculating Double Integrals (Continued) . . . . . . . 617
4. Calculating Areas and Volumes by Means of Double Integrals 623
5. The Double Integral in Polar Coordinates . . . . . . . 626
G. Changing Variables in a Double Integral (General Case) . . . 633
7. Computing the Area of a Surface . . . . . . . . . . . . . 638
8. The Density of Distribution of Matter and the Double Integral . . . 642
9. The Moment of Inertia of the Area of a Plane Figure . . . . . . 643
10. The Coordinates of the Centre of Gravity of the Area of a Plane Figure 648
11.Triple Integrals . . . . . . . . . . . . . . . . . . . . . . . . . 650
12. Evaluating a Triple Integral . . . . . . . . . . . . . . . . . 651
13. Change of Variables in a Triple Integral . . . . . . . . . . . . . . 656
14. The Moment of Inertia and the Coordinates of the Centre of Gravity
of a Solid . . . . . . . . . , , , , . . . . . . 660
15. Computing Integrals Dependent on a Parameter . . . . . 662
Exercises on Chapter XIV . . . - . . . . . . 663
Chapter XV. LINE INTEGRALS AND SURFACE INTEGRALS
L Line Integrals . . . . . . . . . . . . . . . . 670

2. Evaluating a Line Integral . . . . . . . . . . 673
3. Green's Formula . . . . . . . . . . . . . . . . . . ... 679
4. Conditions for a Line Integral Being Independent of the Path of
Integration . . . . . . . . . 681
5. Surface Integrals . . . . . . 687
6. Evaluating Surface Integrals . 689
7. Stokes' Formula . . . . . . . 692
8. Ostrogradsky's Formula . . . . . . . . . . . . . . 697
9. The Hamiltonian Operator and Certain Applications of It 700
Exercises on Chapter XV . . . . . . . . . . . . . . . 703
Chapter XV/. SERIES
1. Series. Sum of a Series . . . . . . . 710

2. Necessary Condit ion for Convergence of a Series 713
3. Comparing Series with Positive Terms , . ' 716
Contents 9
4. D' Alembert's Test . . . . . . . . . . . . . . 718

5. Cauchy's Test . . . . . . . . . . . . . . . . 721
6. The Integral Test for Convergence of a Series 723
7. Alternating Series. Leibniz' Theorem . . . . . . . . . . . . 727
8. Pfus-and-Minus Series. Absolute and Contitional Convergence 729
9. Functional Series . . . . . . . . . . . . 733
10. Majorised Series . . . . . . . . . . . . . 734
11. The Continuity of the Sum of a Series . . 736
12. Integration and Differentiation of Series . 739
13. Power Series. Interval of Convergence . 742
14. Differentiation of Power Series . . . . . 747
15. Series in Powers of x- a . . . . . . . 748
16. Taylor's Series and Maclaurin's Series . . . . 750
17. Examples of Expansion of Functions in Series . 751
18. Euler's Formula . . . . . . . . . . . 753
19. The Binomial Series . . . . . . . . . . . . . . . . . . . . . . . 754
20. Expansion of the Function In (I +x) in a Power Series. Computing
Logarithms . . . . . . . . . . . . . . . . . . . . . . . . . . . 756
21. Integration by U~e of Series (Calculating Definite Integrals) . . 758
22. Integrating Differential Equations by Means of Series . . . . . 760
23. Bessel's Equation . . . 763
Exercises on Chapter XV I . 768
Chapter XV/I. FOURIER SERIES

1. Definition. Statement of the Problem . . . . . 776
2. Expansions of Functions in Fourier Serie~ . . 780
3. A Remark on the Expansion of a Periodic Function in a Fourier
Series . . . . -. . . . . . . . . . . . . . . . . . . . 785
4. Fourier Series for Even and Odd Functions . . . . . 787
5. The Fourier Series for a Function with Period 2l . . 789
6. On the Expansion of a Nonperiodic Function in a Fourier Series . . 791
7. Approximation by a Trigonometric Polynomial of a Function
Represented in the Mean . . . . . . . . . . . . . . . . . . . . . 792
8. The Dirichlet Integral . . . . . . . . . . . . . . . . . . . . . . . 798
9. The Convergence of a Fourier Series at a Given Point . . . . . . . 801
10. Certain Sufflcient Conditions for the Convergence of a Fourier Series 802
11. Practical Harmonic Analysis . . . 805
12. Fourier Integral . . . . . . . . . 810
13. The Fourier Integral in Complex Form 810
Exercises on Chapter XV II . . . . . . 812
Chapter XVIII. EQUATIONS OF MATHEMATICAL PHYSICS

1. Basic Types of Equations of Mathematical Physics . . . . . . . . . 815
2. Derivation of the Equation of Oscillations of a String. Formulation
of the Boundary-Value Problem. Derivation of Equations of Electric
Oscillations in Wires . . . . . . . . . . . . . . . . . . . . . 816
3. Solution of the Equation of Oscillations of a String by the Method
of Separation of Variables (The Fourier Method) . . . . . . . . . . 820
4. The Equation for Propagation of Heat in a Rod. Formulation of the
Boundary-Value Problem . . . . . . . . . . . . . . .. 823
5. Heat Propagation in Space . . . . . . . . . . . . . .... 825
6. Solution of the First Boundary-Value Problem for the Heat-
Conductivity Equation by the Method of Finite Differences 829
10 Contents
7. Propagation of Heat in an Unbounded Rod . . . . . . . . . .. 831

8. Problems That Reduce to Investigating Solutions of the Laplace
Equation. Stating Boundary-Value Problems . . . . . . . . . . . . 836
9. The Laplace Equation in Cylindrical Coordinates. Solution of the
Dirichlet Problem for a Ring with Constant Values of the Desired
Function on the Inner and Outer Circumferences . . . . . . . . . . 841
10. The Solution of Dirichlet's Problem for a Circle . . . . . . . . . 843
It. Solution of the Dirichlet Problem by the Method of Finite Differences 847
Exercises on Chapter XV I II . . . . . . . . . . . . . . . . . . . . 850
Chapter X/X. OPERATIONAL CALCULUS AND CERTAIN OF ITS APPLICATIONS
I. The Initial Function and Its Transform . . . . . . . . . . . . . . 854

2. Transforms of the Functions .0' 0 (1), sin I, cost . . . . . . . . . . . 855
3. The Transform of a Function with Changed Scale of the Independent
Variable. Transforms of the Functions sin at, cos at . . . . . . . . . 856
4. The Linearity Property of a Transform . . . . . . . . . , . . . . 857
5. The Shift Theorem . . . . . . . . . . . . . . . . . . . . , 858
6. Transforms of the Functions e-at, sinh at, cosh at, e-t sin at, e-1 cos at 858
7. Differentiation of Transforms . . . . . . . . . . . . . . 860
8. The Transforms of Derivatives . . . . . . . . . . . . . . . . . . 861
9. Table of Transforms . . . . . . . . . . . . . . . . . -. . . . . 862
10. An Auxiliary Equation for a Given Differential Equation . . . . . 864
II. Decomposition Theorem . . . . . . . . . . . . . . . . . . . . . 867
12. Examples of Solutions of Differential Equations and Systems of
Differential Equations by the Operational Method . . . . . . . . . 869
13. The Convolution Theorem . . . . . . . . . . . . . . . . . . . . 871
14. The Differential Equations of Mechanical Oscillations. The Differen-
tial Equations of Electric-Circuit Theory . . . . . 873
15. Solution of the Differential Oscillation Equation . . 874
16. Investigating Free Oscillations . . . . . . . . . . . . . 875
17. Investigating Mechanical and Electrical Oscillations in the Case of a
Periodic External Force . . . . . . . . . . . . . . . . . . 876
18. Solving the Oscillation Equation in the Case of Resonance 878
19. The Delay Theorem . 879
Exercises on Chapter XIX . . . . . . . . . . . . . . 880
Subject Index
PREFACE
This text is designed as a course of mathematics for higher technical

schools. It contains many worked examples that illustrate the theoretical
material and serve as models for solvir.g problems.
The first two chapters "Number. Variable. Function" and "Limit. Cpnti
nuity of a Function" have been made as short as possible. Some of the ques-
tions that are usually discussed in these chapters have been put in the third
and subsequent chapters without loss of continuity. This has made it possible
to take up very early the basic concept of differential calculus-the deriva-
tive-which is required in the study of technical subjects. Experience has
shown this arrangement of the material to be the best and most convenient
for the student.
A large number of problems have been included, many of which illust-
rate the interrelationships of mathematics and other disciplines. The problems
are specially selected (and in sufficient number) for each section of the course
thus helping the student to master the theoretical material. To a large extent,
this makes the use of a separate book of problems unnecessary and extends
the usefulness of this text as a course of mathematics for self-instruction.
N. S. Piskunov
CHAPTER
NUMBER. VARIABLE. FUNCTION
SEC. 1. REAL NUMBERS. REAL NUMBERS AS POINTS ON A

NUMBER SCALE
Number is one of the basic concepts of mathematics. It originated

in ancient times and has undergone expansion and generalisation
over the centuries.
Whole numbers and fractions, both positive and negative, together
with the number zero are called rational numbers. Every rational
number may be represented in the form of a ratio, .!!.... , of two
q
integers p and q; for example,
5 5
7 1.25=4
In particular, the integer p may be regarded as a ratio of the
integersi; for example,
6
6--
-1'
0-~
-1
Rational numbers may be represented in the form of periodic

terminating or nonterminating fractions. Numbers represented by
nonterminating, but nonperiodic, decimal fractions are called
irrational numbers; such are the numbers V2, V3, 5 -V2. etc.
The collection of all rational and irrational numbers makes up
the set of -real numbers. The real numbers are ordered in magnitude;
that is to say, for each pair of real numbers x andy there is one,
and only one, of the following relations:
x<y, x=y, x>y.
R.eal numbers may be depicted as points on a number scale.
A number scale is an infinite straight line on which are chosen:
1) a certain po_int 0 called the origin, 2) a positive direction
indicated by an arrow, and 3) a suitable unit of length. We shall'
usually make the number scale horizontal and take the
positive direction to be from left to right.
If the number X 1 is positive, it is depicted as a point M 1 at
a distance OM 1 =X1 to the right of the origin 0; if the number x.
is negative, it is represented by a point M 1 to the left of 0 at a
14 Number. Variable. Function
distance OM 2 = - X 2 (Fig. 1). The point 0 represents the number

zero. It is obvious that every real number is represented by a
definite point on the number scale. Two different real numbers are
represented by different points on the number scale.
The following assertion is also true: each point on the number
scale represents only one real number (rational or irrational).
To summarise, all real numbers and all points on the number
scale are in one-to-one correspondence: to each number there cor-
responds only one point, and conversely, to each point there cor-
responds only one number. This frequently enables us to regard "the
number x" and "the point x" as, in a certain sense, equivalent
expressions. We shall make wide use
Hz 0 Ht of this circumstance in our course.
---offO I I I o
- - -2 -1 1 2 3 -x We state without proof the follow-
Fig. I. ing important property of the set
of real numbers: both rational and
irrational numbers may be found
between any two arbitrary real numbers. In geometrical terms,
this proposition reads thus: both rational and irrational points may
be found between any two arbitrary points on the number scale.
In conclusion we give the following theorem, which, in a certain
sense, represents a bridge between theory and practice.
Theorem. Every irrational number a. may be expressed, to any
degree of precision, with the aid of rational numbers.
Indeed, let the irrational number a.>O and let it be required
n
to evaluate a. with an accuracy of I ( for example, I , 100
16
I , and so
forth).
No matter what a. is, it lies between two integral numbers N
and N +I. We divide the segment bet_ween N and N 1 into n +
parts; then a. wi II lie somewhere between the rational numbers
N +m n
and N +m+l. Since their difference is equal to_!_, each
n n
of them expresses a. to the given degree of accuracy, the former
being smaller and the latter greater.
Example. The irrational number 2 is expressed by rational numbers:
1.4 and 1.5 to one decimal place,
1.41 and 1.42 to two decimal places,
1.414 and 1.415 to three decimal places, etc.
SEC. 2. THE ABSOLUTE VALUE OF A REAL NUMBER

Let us introduce a concept which we shall need later on: the
absolute value of a real number.
The Absolute Value of a Real Number 15
Definition. The absolute value (or modulus) of a real number x

(written IxI) is a nonnegative real number that satisfies the con-
ditions
IX I= X if X~ 0;
/XI=-X ifx<O.
Examples. 121=2; 1-51=5; 101=0.
From the definition it follows' that the relationship x,;:;;; 1x 1 holds
for any x.
Let us examine some of the properties of absolute values.
1. The absolute value of an algebraic sum of several real numbers
is no greater than the sum of the absolute values of the terms
lx+yl..;;; lxl +Jyl.
Proof. Let x + y;;;:.;. 0, then
lx+yl=x+y:s:;;;/xi+/YI (since x:s:;;;lxl and y:s:;;;/yl).
Let x+y< 0, then
lx+yl=-(x-1-y)= (-x) -1-(-y) :s:;;;lxl +IYI
This completes the proof.
The foregoing proof is readily extended to any number of terms.
Examples.
-1- 2 + 3 I < 1-21 +I 31 = 2 + 3 = 5 or I <~5;
1-3-51=1-31+1-51=3+5~8 or 8=8.
2. The absolute value of a ditJerence is no less than the
ditJerence of the absolute values of the minuend and subtrahend:
lx-yl ~lxi-IYI
Proof. Let x-y=z, then x=y+z and from what has been
proved
IX I= I y + z I..;; I y I +I z I= I y I+ IX - y I.
whence
lxl-/yl:s:;;;/x-yj,
thus completing the proof.
3. The absolute value of a product is equal to the product of
the absolute values of the factors:
Ixyz I = IxI IY I I z I
4. The absolute value of a quotient is equal to the quotient
of the absolute values of the dividend and the divisor:
I; I=:;:
The latter two properties follow directly from the definition of
absolute value.
SEC. 3. VARIABLES AND CONSTANTS
The numerical values of such physical quantities as time, length,

area, volume, mass, velocity, pressure, temperature, etc., are deter-
mined by measurement. Mathematics deals with quantities divested
of any spec i fie content. From now on, when speaking of quanti ties,
we shall have in view their numerical values. In various phenomena,
the numerical values of certain quantities vary, while the numerical
values of others remain fixed. For instance, in uniform motion of
a point, time qnd distance change, while the velocity remains constant.
A variable is a quantity that takes on various numerical values.
A constant is a quantity whose numerical values remain fixed. We
shall use the letters x, y, z, u, ... , etc., to designate variables,
and the letters a, b, c, ... , etc., to designate constants.
Note. In mathematics, a constant is frequently regarded as a
special case of variable whose numerical values are the same.
It should be noted that when considering specific physical pheno-
mena it may happen that one and the same quantity in one pheno-
menon is a constant while in another it is a variable. For example,
the velocity of uniform motion is a constant, while the velocity of
uniformly accelerated motion is a variable. Quantities that have
the same value under all circumstances are called absolute constants.
For example, the ratio of the circumference of a circle to its dia-
meter is an absolute constant: n: = 3.14159.
As we shall see throughout this course, the concept of a variable
quantity is the basic concept of differential and integral calculus.
In "Dialectics of Nature", Friedrich Engels wrote: "The turning
point in mathematics was Descartes' variable magnitude. With
that came motion and hence dialectics in mathematics, and at
once, too, of necessity the differential and integral calculus."
SEC. 4. THE RANGE OF A VARIABLE
A variable takes on a series of numerical values. The collection

of these values may differ depending on the character of the prob-
lem. For example, the temperature of water heated under ordinary
conditions will vary from room temperature (15-18C} to the
boiling point, 100C. The variable quantity x=cosa can take on
all values from -1 to+ 1.
The values of a variable are geometrically depicted as points on
a number scale. For instance, the values of the variable x =cos a
for all possible values of a are depicted as the set of points of an
interval on the number scale, from - 1 to I, including the points
-1 and I (Fig. 2).
The Range of a Variable 17
Definition. The set of all numerical values of a variable quantity

is called the range of the variable.
We shall now define the following ranges of a variable that will
be frequently used later on.
An 9pen interval is the collection of
all numbers x lying between and excluding
the given numbers a and b (a< b); it
is denoted (a, b) or by means of the
inequalities a<x<b.
A closed in! erval is the set of all
numbers x lying between and including
the two given numbers a and b; it is
denoted (a, b] or, by means of inequali- Fig. 2
ties, a .;;;;;; x .;;;;;; b.
If one of the numbers a or b (say, a) belongs to the interval,
while the other does not, we have a partly closed interval, which
may be given by the inequalities
a.;;;;;x<b
and is denoted (a, b). If the number b belongs to the set and a
does not, we have the partly closed interval (a, b], which may be
given by the inequalities
a<x.;;;;;b.
If the variable x assumes all possible values greater than a, such
an interval is denoted (a, oo) and is represented by the conditional
inequalities
a<x<oo.
In the same way we regard the infinite intervals and partly closed infi-
nite intervals represented by the conditional inequalities a .;;;;;x< oo;
- oo < x < c;- oo < x.;;;;;; c;- oo< x < oo.
Example. The range of the variable x=cosa for all possible values of a
is the interval [-I, I] and is defined by the inequalities -I,;;;;;; x,;;;;;; I.
The foregoing definitions may be formulated for a "point" in

place of a "number".
An interval is the set of all points x lying between the given points
a and b (the end points) and is called closed or open accordingly
as it does or does not .include its end points.
The neighbourhood of a given point X 0 is an arbitrary interval
.(a, b) containing this point within it; that is, the .interval (a, b)
whose end points satisfy the condition a< X 0 <b. One often
considers the neighbourhood (a, b)

0
of the point X 0 for which X 0 is the
midpoint. Then X 0 is called the
Fig. 3. centre of the neighbourhood and
the quantity b a, the radius of
2
the neighbourhood. Fig. 3 shows the neighbourhood (x 0 - e, x 0 +e)
of the point X 0 with radius e.
SEC. 5. ORDERED VARIABLES.
INCREASING AND DECREASING VARIABLES. BOUNDED VARIABLES
We shall say that the variable x is an ordered variable quantity
if its range is known and if about each of any two of its values
it may be said which value is the preceding one and which is the
following one. Here, the notions "preceding" and "following" are
not connected with time, but serve as a way to "order" the values
of the variable, i. e., to establish the order of the respective values
of the variable.
Definition 1. A variable is called increasing if each subsequent
value of it is greater than the preceding value. A variable is called
decreasing if each subsequent value is less than the preceding value.
Increasing variable quantities and decreasing variable quantities
are called monotonically varying variables or simply monotonic
quantities.
Example. When the number of sides of a regular polygon i'nscribed in a
circle is doubled, the area s of the polygon is an increasing variable. The
area of a regular polygon circumscribed about a circle, when the number of
sides is doubled, is a decreasing variable. It may be noted that not every
variable quantity is necessar'ily increasing or decreasing. Thus, if a is an
increasing variable over the interval [0, 2Jt], the variable x=sln a is not a
monotonic quantity. It first increases from 0 to I, then decreases from I to
- I, and then increases from - I to 0.
Definition 2. The variable x is called bounded if there exists a
constant M >0 such that all subsequent values of the variable,
after a certain one, satisfy the condition
-M,;;;;x,;;;;M, that is, lxj,;;;;M.
In other words, a variable is called bounded if it is possible to
indicate an interval [-M, M) such that all subsequent.values of
the variable, after a certain one, will belong to this interval.
However, one should not think that the variable will necessarily
assume all values of the interval [- M, M]. For example, the
variable that assumes all possible rational values on the interval
[- 2, 2) is bounded, and nevertheless it does not assume all values
on [- 2, 2], namely, the irrational values.
Function
SEC. 6. FUNCTION
In the study of natural phenomena and the solution of technical
and mathematical problems, one finds it necessary to consider the
variation of one quantity as dependent on the variation of another.
For instance, in studies of motion, the path traversed is (egarded
as a variable which varies with time. Here, the path traversed is
a function of the time.
Let us consider another example. We know that the area of a
circle, in terms of the radius, is Q = nR". If the radius R takes
on a variety of numerical values, the area Q will also assume
various numerical values. Thus, the variation of one variable brings
about a variation in the other. Here, the area of a circle Q is a
function of the radius R. Let us formulate a definition of the con-
cept "function".
Definition 1. If to each value of the variable x (within a certain
range) there corresponds one definite value of another variable y,
then y is a function of x or,'in functional notation, y=f(x), y=cp(x),
and so forth.
The variable x is called the independent variable or argument.
The relation between the variables x and y is called a functional
relation. The letter f in the functional notation y = f (x) indicates
that some kind of operations must be performed on the value of
x in order to obtain the value of y. In place of the notation
y = f (x), u = cp (x), etc., one occasionally finds y = y (x), u = u (x),
etc., the letters y, u designating both the dependent variable and the
symbol of the totality of operations to be performed on x.
The notation y = C, where C is a constant, denotes a function
whose value for any value of x is the same and is equal to C.
Definition 2. The set of values of x for which the values of the
function y are determined by virtue of the rule f (x) is called the
domain of definition of the function.
Example I. The function y =sin x is defined for all values of x. Therefore,
its domain of definition is the infinite interval -oo< x < oo.
Note 1. If we have a functional relation of two variable -quan-

tities x and y = f (x) and if x and y = f (x) are regarded as ordered
variables, then of the two values of the function y* = f (x*) and
y** = f (x**) corresponding to two values of the argument x* and
x**, the subsequent value of the function will be that one which
corresponds to th~ subsequent value of the argument. The following
definition is, therefore, natural.
Definition 3. If the function y = f (x) is such that to a greater
value of the argument_ x there coriespond~ a greater value of the
runction, then the function y = f (x) is called increasing. A decreas-

ing function is similarly defined.
Example 2. The function Q =:rr.R 2 for 0 < R < oo is an increasing function
because to a greatei value of R there corresponds a greater value of Q.
Note 2. The definition of function is sometimes broadened 50

that to each value of x, within a certain range, there corresponds
not one but several values of y or even an infinitude of values
of y. In this case we have a multiple-valued function in contrast
to the one defined above, which is called a single-valued function.
Hencerorward, when speaking of a function, we shall have in view
only single-valued functions. If it becomes necessary to deal with
multiple-valued functions we shall specify this fact.
SEC. 7. WAYS OF REPRESENTING FUNCTIONS
I. Tabular representation of a function

Here, the values of the argument xtt X2 , , xn and the cor-
responding values of the function Ytt y 2 , , Yn are written out
in a definite order.
X x, x.
1-----1 Xn
!I y, IJ2 1--~--1 Yn
Examples are tables of trigonometric functions, tables of

logarithms, and so on.
An experimental study of phenomena can result in tables that
express a runctional relation between the measured quantities. For
example, temperature measurements of the air at a meteorological
station on a definite day yield a table like the following.
The temperature T (in degrees) is de pendent on the time
(in hours).
t 2 3 4 5 6
I7 8 9
I I I
T 0 -I -2 -2 J -o.5J 3 3.5 4
I I
This table defines T as a function ol t.
Ways of Represwting Functions 21
II. Graphical representation of a function

If in a rectangular coordinate system on a plane we have a set
of points M (x. y), and no two points lie on a straight line parallel
to the y-ax is, this set of points defines y
a certain single-valued function y = y=f(xJ
=f (x); the abscissas of the points
are tire values of the argument, the
corresponding ordinates are the values y
of the function (Fig. 4).
The collection of points in the 0 X X.
xy- plane whose abscissas are the Fig. 4.

values of the independent variable
and whose ordinates are the corresponding values of the function
is called a graph of the given function.
- III. Analytical representation of a functioQ

Let us first explain what "analytical expression" means. By ana-
lytical expression we will understand a series of symbols denoting
a totality of known mathematical operations that are performed in
a definite sequence on numbers and letters which designate constant
or variable quantities.
By totality of known mathematical operations we mean not only
the mathematical operations familiar from the course of secondary
school (addition, subtraction, extraction of roots, etc.) but also
those which will be defined as we proceed in this course.
The following are examples of analytical expressions:
X
4_
2 .' Iogx-sinx.
5x'+ 1 ' 2 x_v 5 + 3
X,
elc.
If the functional relation y = f (x) is such that f denotes an
analytical expression, we say that the function y of x is represented
analytically.
Examples of functions repreoented analytically are: 1) y=x 4 -2;
x+ 11 ; 3) y = Vt-x; 4) y =sin x; 5) Q == nR, and so forth.
2) y= x-
Here, the functions are represented analytically by means of a
single formula (a formula is understood to be the equality of two
analytical expressions). In such cases one may speak of the natural
domain of definition of the function.
The set of values of x for which the analytical expression on
the right-hand side has a fully definite value is the natural domain
22 . Number. Variable. Funt.:i:ion
of definition of a function represented analytically. Thus, the natu-

ral domain of definition of the function y = X 4 -2 is the infinite
interval - oo < x < oo, because the function is defined for all
values of x. The function y=~+: is defined for all values of x,
with tht! exception of x = I, because for this value of x the deno
minator vanishes. For the function y= Vl-x 2 , the natural domain
2
of definition is the closed interval-! :s;;:; x :s;;:; I,
. Y y=x and so on.
Note. It is sometimes necessary to consider
only a part of the natural domain of a function,
and not the whole domain. For instance, the
dependence of the area Q of a circle upon the
radius R is defined by the function Q = nW.
The domain of this function, when considering
a given geometrical problem, is the infinite
interval O<R<+oo. But the natural domain
x of this function is the infinite interval- oo <
Fig. 5. <R<+oo.
If the function y = f (x) is represented anp.ly-
tically, it may be shown graphically on a coordinate xy-plane.
Thus, the graph of the function y = X 2 is a parabola as shown in
Fig. 5.
SEC. 8. BASIC ELEMENTARY FUNCTIONS. ELEMENTARY FUNCTIONS
The basic elementary functions are the following analytically

represented functions.
I. Power function: y = x~, where a.. is a rea I number. *)
II. Exponential function: y =aX, where a is a positive number
not equal to unity. .
III. Logarithmic function: y= logax where the base of logarithms
a is a positive number not equal to unity.
IV. Trigonometric functions: y=sinx,. y=cosx, y=tanx,
y=cotx, y=secx, y=cscx.
V. Inverse trigonometric functions:
y =arc sin x, y =arc cos x, !] =arc tan x,
y =arc cot x, y =arc sec x, y =arc esc x.
Let us consider the domains of definition and the graphs of the
basic elementary fund ions.
*) If a is irrational, this function is evaluated by taking logarithms and
antilogarithms: logy =a log x. lt is assumed that x > 0.
Basic Elementary Functions. Elementary Functions 23
Power function y=x~.

I. a is a positive integer. The function is defined in the infi-
nite +
interval - oo < x < oo. In this case, the graphs of the func-
tion for certain values of a have the form shown in Figs. 6
and 7.
y
Fig. 6. Fi;t. 7.
2. a is a negative integer. In this case, the function is defined

for all values of x with the exception of x = 0. The graphs of the
functions for certain values of a y
have the form shown in Figs. 8
and 9.
\e
--1-------=....,-x
Fig. 8.
\ Fig. 9.
Figs. 10, 11, and 12 show graphs of a power function with

fractional rational values of a.
y
, y=xlt'J
Fig. 10.
X
.
Fig. /1.
X
Fig. 12.
X
Exponential function, y =ax, a> 0 and a 7/= 1. Thi.s function is

defined for all values of x. Its graph is shown in Figs. 13 and 14.
Y rtYr-fYMl Y tax 3x zx
!J=(fY y=2)(
Fig. 13. Fig. 14.
Logarithmic function, y = loga x, a> 0 and a =F 1. This function

is defined for x > 0. Its graph is shown in Fig. 15.
Trigonometric functions. In the formulas y =sin x, etc., the
independent variable x is expressed in radians. All the enumerated
y trigonometric functions are periodic ..
Let us give a general definition of a
periodic function.
Definition 1. The function y = f (x)
is called periodic if there exists a con-
stant C, which, when added to (or sub-
x traded from) the argument x, does not
change the value of the function:
f (x +C)= f (x). The least such number
is called the period of the function; it
Fig. 15. will henceforward be designated as 2/.
From the definition it follows directly
that y ~sin x is a periodic function with a period 2n: sin x =
=sin (x+2n). The period of cosx is likewise 2n. The functions
y =tan x and y =cot x have a period equal to n.
The functions Y= sinx, y=cosx are defined for all values of x;
the functions y=tanx and y=secx are defined everywhere except
the points x = (2k ~ 1) ; (k = 0, 1, 2, ... ); the functions y =cot x
and y= cscx are defined for all values of x except the points
x = kn (k = 0, 1, 2, ... ). Graphs of trigonometric functions are
shown in Figs. 16, 17, 18, and 19.
The inverse trigonometric functions will be discussed in more
detail later on.
Basic Elementary Functions. Elementary Functions 25
y
y=sinx
Fig. 16.
Fig. 17.
Fig. 18. Fig. 19.
Let us now introduce the concept of a function of a function.

If y is a function of u, and u (in turn) is dependent on the var-
iable x, then y is also dependent on x. Let
y = F (u)
u = cp (x).
We get y as a function of x
y=F[cp(x)].
This function is called a function of a function or a composite
function.
Example 1. Let y=sin u, u=x2 The function y=sin (x 2 ) is a composite
function of x.
Note. The domain of definition of the function y = F [cp (x)] is

either the entire domain of the function, u = cp (x), or that part
of it in which those values of u are defined that do not go beyond

the domain of the function F (u).
Example 2. The domain of definition of the function y= Yl-x 2 (y= Vu
11 =l-x 2) is the closed interval [-I, I]. because when I x 1> I u < 0 and, conse:
quently, the function Y.u is not defined (although the function u = l-x 2 is
defined for all values of x). The graph of this function is the upper half of
a circle with centre at the origin of the coordinate system and with radius
unity.
The operation "function of a function". may be performed any

number of times. For instance, the function y=ln [sin(x 2 + 1)] is
obtained as a result of the following operations (defining the
following functions):
v=x 2 +l, u=sinv, y=lnu.
Let us now define an elementary function.
yl Definition 2. An elementary function is
a function which may be represented oy
U
a single formula of the type y=f(x),
where the expression on the right-hand
side is made up of basic elementary func-
- - - - x tions and constants by means of a finite
o t 2 number of operations of addition,
Fig. 20. subtraction, multiplication, division and
taking the function of a function.
From the definition it follows that elementary functions are
functions represented analytically.
Examples of elementary functions:
logx+4 v;+2tanx
y= Y! +4 sin 2
x; y= wx-x+ 10
and the like.
Examples of non-elementary functions:
I. The function y = 123 ... n (y=f (n)] is not elementary because the
number of operations that must be performed to obtain y increases with n,
that is to say, it is not bounded.
2. The function given in Fig. 20 is not elementary either because it is
represented by means of two formulas:
f(x)=x, i!O.;;;;x.;;;;l,
f(x)=2x-1, ifl.;;;;x.,;;;;2.
SEC. 9. ALGEBRAIC FUNCTIONS
Algebraic functions include elementary functions of the following

kind:
Algebraic Functions 27
I. The rational integral function, or polynomial

y=aoxn-t-a1xn-1+ .. +an,
where a0 , a., ... , an are constants called coefficients, and n is a
nonnegative integer called the degree of the polynomial. It is
obvious that this function is defined for all values of x, that is,
it is defined in an infinite interval.
Examples: 1. y =ax+ b is a linear- Y a>-0 Y a<O
function. When b =0, the linear function
y =ax expresses y as being directly pro-
portional to x. For a=O, y=b, the
function is a constant.
2. y=axR+bx+c is a quadratic
function. The graph of a quadratic func-
tion is a parabola (Fig. 21 ). These (a)
functions are considered in detail in
analytic geometry. Fig. 21.
N. Fractional rational function. This function is defined as the

ratio of two polynomials:
aoxn+ a1xn-1 + ... +an
y= boxrn+b1xm-1+ ... +bm
For example, the following is a fractional rational function:
it expresses inverse variation. Its graph is shown in Fig. 22. It is

obvious that a fractional rational function is defined for all values
of x with the exception
_)a<Oy of those for which the

denominator
zero.
becomes
Ill. Irrational func-
X tion. If in the formula
y =I (x), operations of
addition, subtraction,
multiplication, division
(a) (b) and raising to a power
with rational non-inte-
Fig. 22. gral exponents are per-
formed on the right-
hand side, the function y = f (x) is called irrational. Examples
of irrational functions are: y= ~~:~:; y= Vx; etc.
Note l. The above-mentioned three types of algebraic function:}

do not exhaust all algebraic functions. An algebraic function is any
function y=f"(x) which satisfies an equation of the form
P 0 (x)yn+P 1 (X)yn-t+ ... +Pn(X)=0, (1)
where P0 (x), P 1 (x), ... , Pn (x) are certain polynomials in x.
It may be proved that each of the enumerated three types of
function satisfies a certain equation of type (1), but not every func-
tion that satisfies an equation like (1) is a function of one of
the three types given above.
Note 2. A function which is not algebraic is called transcendental.
Examples of transcendental functions are:
y= cos x; y= 10xx
and the like.
SEC. 10. POLAR COORDINATE SYSTEM
The position of a point in a plane may be determined by means
of a so-called polar coordinate system.
We choose a point 0 in a plane and call it the pole;. the half-
line issuing from this point is called the polar axis. The position
of the point M in the plane may be specified by two numbers:
the number Q, which expresses the distance of M from the pole,
and the number q:>, which is the angle formed
M by the line segment OM and the polar axis.

The positive direction of the angle q:> is
reckoned counterclockwise. The numbers Q
and q:> are called the polar coordinates of the
0- X point M (Fig. 23).
We will always consider the radius vector
Fig. 23. Q nonnegative. If the polar angle q:> is taken
within the limits 0 ~ q:> < 2n, then to each
point of the plane (with the exception of the pole) there corre-
sponds a definite number pair Q and <p. For the pole, Q=O and q:>
is arbitrary.
Let us now see how the polar and rectangular Cartesian coordi-
nates are related. Let the origin of the rectangular coordinate
system coincide with the pole, and the positive direction of the
x-axis, with the polar axis. We establish a relationship between
the rectangular and polar coordinates of one and the same point.
From Fig._ 24 it follows directly that
x = Q cos q:>, y = Q sin <p
and, conversely, that
Polar Coordinate System 29
Note. To find cp, it is necessary to take into account the quad

rant in which the point is located and then take the correspond-
y
y=9sinrp
xRpco_srp
Fig. 24. Fig. 25.
ing value of cp. The equationQ=F(cp) in polar coordinates defines

a certain line.
Example 1. Equation Q =a, where a= const, defines in polar coordinates
a circle with centre in the pole and with radius a. The equation of this
circle in a rectangular coordinate system situated as shown in Fig. 24 is
Yx 2 +y 2 =a or x 2 +y 2 =a 2
(Fig. 25).
Example 2. Q = a<p, where a= const.
Let us tabulate the values of Q for certain values of <p
l't
2n 3n 4n
4
Q I0 I::::::0. 78a I::::::1.57a I:::::: 2.36a I;,3, 14a I::::::'4.71 a I=:::6.28a I: : : 9.42a I: : : I2.56a
The corresponding curve is shown in Fig. 26. It is called the spiral of Archi-
medes.
Example 3.
Q=2a cos <p.
This is the equation of a circle of radius a, the centre of which is at the
point Q0 =a, <p= 0 (Fig. 27). Let us write the equation of this circle in rect-
Fig. 26. Fig. 27.

angular coordinates. Substituting Q= Yx'+y', coscp= Jl x into the gi

x'+y'
ven equation, we get
Jfx'+y 2 =2a x
Vx+y 2
or
x'+y 2 -2ax =0.
Exercises on Chapter
t. Given the function [(x)=x'+6x-4. Verify the equalities /(1)=3,

{(3)=23.
2. f(x)=x'+I. Evaluate: a) {(4). Ans. 17. b) {(Jf2).Ans.3. c)f(a+I).
Ans. a'+2a+2. d) f(a)+l. Ans. a'+2. e) f(a 2 ). Ans. a4 +1. f) lf(a)] 2
Ans. a+2a'+l. g) f(2a). Ans. 4a 2 +I.
. cp(x)= x-1
3. W t th .
3x+ 5 n e e expressiOns cp x
( I ) an d cp (x).1 A ns. cp ( I ) = x
1-x I 3(x)+5
=3+5x; q>(x)=~
4. \jl (x) = Jl x + 4. Write the expressions \jl (2x) and \jl (0). Ans. \jl (2x) =
+
= 2 V x 1; \jl (0) = 2.
5. f (8) =tan 8. Verify the equality f (28) = ~-=~~~~-))".
6. cp(x)=log~=i=> Verify the equality cp(a)+cp(b)=q>(~::b)
7. f (x) =log x; cp (x) =x 1 Write the expressions: a) f [cp (2)]. Ans. 3 log 2.
b) f [cp (a)]. Ans. 3log a. c) cp (f (a)]. Ans. [log a) 1
8. Find the natural domain of definition of the function y = 2x 2 + 1 Ans.
-oo<x<+oo.
9. Find the natural domains of definition of the functions: a) Jl 1-x.
Ans. -1,.;;;;x..;;;;+I. b) Jf3+x+ V--
7-x. Ans. -3.;;;;;x.;;;;;7. c) x+u- v--
-- Vx-b. Ans. -oo <x<+oo. d) a+x. Ans. x;l: a. e) arc sin 2 x. Ans.
a-x
-l.;;;;;x,;;;;l. f) y=logx. Ans. x>O. g) y=ax(a>O). An.s.-oo<x<+oo.
Construct the graphs of the functions:
I
10. y=-3x+5. 11. y='ix'+I. 12. y=3-2x 2 13. y=x'+2x-1.
1 1
14. y= - - . 15. y=sin 2x. 16. y=cos 3x. 17. y=x 2 -4x+6. 18. y=-- -.
x- 1 1-x.
19. y=sin( x+:). 20. y=cos(x-~).21. y=tan}x. 22.y=cot~x.
25. y=log 2 ..!... 26. y=x 1 +1. 27. y=4-x 8 .28. y=
1
23.y=3x. 24.y=2-x
X
1 1 _!_
~z . 29. y=x. 30. y=x'. 31. y=x 2 . 32. y=x
2
33. y=x 1 34. y= lx I
35. y=log 2 )x[. 36. y=log 2 (1-x). 37. y=3sin(2x+~) 38. y=
Exercises on Chapter I 31
= 4 cos ( x+ ~-) . 39. The function f (x) is defined on the interval L-I, I]
as follows:
f(x)=l+x for-lo;;;;;x<O:
f(x)= l-2x forOo;;;;x <!.
40. The f~;~nction f (x) is defined on the interval [0, 2] as follows:
f(x)=x' forOo;;;;xo;;;;l;
f(x)=x for lo;;;;xo;;;;2.
Plot the curves given by the polar equations: 41. Q=~ (hyperbolic sp!-
<p
Y
ral). 42. Q= a'l' (logarithmic spiral). 43. Q=a cos 2<p (lemniscate). 44. Q =
=a (!-cos <p) (cardioid). 45. Q=a sin 3<p.
CHAP T E.R II
LIMIT. CONTINUITY OF A FUNCTION
SEC. 1. THE LIMIT OF A VARIABLE.

AN INFINITELY LARGE VARIABLE
In this section we shall consider ordered variables that vary in

a special way defined as follows: "the variable approaches a
limit". Throughout the remainder of the 'course, the concept of
limit of a variable will play a fundamental role, for it is intimate-
ly bound up with the basic concepts of mathematical analysis,
such as derivative, integral, etc.
Definition 1. A constant number a is said to be the limit of a
variable x, if for every preassigned arbitrarily small positive num-
ber 8 it is possible to indicate a value of the variable x such
that all subsequent values of the variable will satisfy the inequality
jx-aj<e.
If the number a is the limit of the variable x, one says that x
approaches the limit a; in symbols we have
x-+a or limx=a.
In geometric terms, limit may be defined as follows.
The constant number a is the limit
of the variable x if for any preassigned
0 a e arbitrarily small neighbourhood with
lx-al
centre in the point a and with radius
Fig. 'i8. 8 there is a value of x such that all
points corresponding to subsequent values
of the variable will be within this neighbourhood (Fig. 28). Let us
consider several cases of variables approaching limits.
Example t. The variable x takes on successive values:
l I I
X1 =2; X 2 =1 2; X3=l3; ... ; Xn=l n;
We shall prove that this variable has uni!y as its limit. We have
lxn-11=1 (I+~ )-II=~

* For any e, all subsequent values of the variable begin with n, where
< e, or n > ~ will satisfy the inequality I Xn ...:..1 I < e and the proof is
complete.
The Limit of a Variable 33
It will be noted here that the variable quantity decreases as it approaches

the limit.
I
Example 2. The variable x takes on successive values: x 1 =I - ; x =
2
I 1 1
=1+ .; X3 =1- 23 ; X =1-2'; ... ; Xn=1+(-1)n 211 ;
2
This variable has a limit of unity. Indeed,
IX n- I I= I ( 1 + (-1 )n in) --I I= 21n

For any e, beginning with n, which satisfies the relation
I
2n <e,
from which it follows that

n 1
2 >-
e ,
1
n log2 >log-
e
or
1
log-
e
n > log2 '
all subsequent values of x will satisfy the relation

lxn-ll<e.
It will be noted here that the values of the variable are greater than or
less than the limit, and the variable approaches its limit by "oscillating
about it". -
Note 1. As was pointed out in Sec. 3 (see Ch. 1), a constant

quantity c is frequently regarded as a variable whose values
all coincide: x =c.
Obviously, the limit of a constant is equal to the constant
itself, sincewe always have the inequality jx-c~=lc-cj=O<e
for any e.
Note 2. From the definition of a limit it follows that a vari
able cannot have two limits. Indeed, if limx=a and limx=
=b (a< b), then x must satisfy, at one and the same time, two
inequalities:
lx-aj<e and lx-bj<e
for an arbitrarily small e; but this is impossible if e < b;: a
(Fig. 29).
2-3388
34 Limit. Continuity of a Function
Note 3. One should not think that every variable has a limit.
Let the variable x take on the following successive values:
I I I I
x,= 2 ; x,=l- 4 ; x.= 8 ; x,=l- 16 ; ... ,
I
x,k= 1- 2.k; x,k+1 = 2k+'
(Fig. 30). Fork sufficiently large, the value x,k and all subsequent
values with even labels will differ from unity by as small a
a ' \:17
2e t l 2e
e.<----z-
Fig. 29. Fig. 3U.
number as we please, while the next value x!k+t and all subse-
quent values of x with odd labels will differ from zero by as
small a number as we please. Consequently, the variable x does
not approach a limit.
In the definition of a limit it is stated that if the variable
approaches the limit a, then a is a constant. But the word "appro-
-aches" is used also to describe another type of variation of a
variable, as will be seen from the following definition.
Definition 2. A variable x approaches infinity if for every
preassigned positive number M it is possible to indicate a value
of x such that, beginning with this value, all subsequent values
of the variable will satisfy the inequality IxI> M.
If the variable x approaches infinity, it is called an j_nfinitely
large variable and we write x--..cx>.
Example 3. The variable x takes on the values
X 1 =-l; x 2 =-2; x 3 =-3; ... ; Xn=(-l)nn
This is an infinitely large variable quantity, since for an arbitrary M > 0 all
values of the variable, beginning with a certain one, are, in absolute
magnitude, greater than M.
J The variablex "approaches plus infinity", x -++ oo, if for an

arbitrary M > 0 all subsequent values of the variable, beginning
with a certain one, satisfy the inequality M < x.
An example of a variable quantity approaching plus infinity is the variable
x that takes on the values x 1 = 1, x 2 = 2, ... , Xn = n, ...
The Limit of a Function 35
A variable approaches minus infinity, x-..-oo, if for an arbi-

trary M > 0, all subsequent values of the variable, beginning with
a certain one, satisfy the inequality x<-M.
For example, a variable x that assumes the values x 1 =-l, x2 =-2,
Xn=-n, ... , approaches minus infinity.
... ,
SEC. 2. THE LIMIT OF A FUNCTION
In this section we shall consider certain cases of the variation

of a function when the argument x approaches a certain limit a
or infinity.
Definition l. Let the function y = f(x) be defined in a certain
neighbourhood of the point a or at certain points of this neigh-
bourhood. The function y = f (x) approaches the limit b (y---.. b) as x
approaches a (x---.. a), if for every positive number e, no matter
how small, it is possible to indicate a positive number b such
that for all x, different from a and satisfying the inequality*)
lx-al<6,
we have the inequ.ality
If (x)- b I< e.
If b is the limit_ of the functio.n f (x) as x __,..a, we write
y
limf(x)=b
x-+a
b+e~~""~~""~~~~
or f (x) - + b as x---.. a. b~~~~~~~~~
If f(x)-+b as x-+a, this is p-e~""~~~~~~~a-
illustrated on the graph of the
fu net ion y = f (x) as follows
(Fig. 31).
Since from the inequality
lx-al<b there follows the o X
inequality If (x)- b I< e, this
means that for all points x Fig. 31.
*) Here we mean the values of x that satisfy the inequality

1x -a I < l'J and belong to the domain of definition of the function. We shall
encounter similar circumstances in the future. For instance, when considering
the behaviour of a function as x-+ oo, it may happen that the function 1s
defined only for positive integral values of x. Artd so in this case X-+oo,
assuming only positive integral values. We shall not specify this when it
comes up later on.
2*
that are not more distant from the point a than ll, the points
M of the graph of the function y = f (x) lie within a band of
width 28 bounded by the lines y = b- 8 and y = b 8. +
Note 1. We may define the limit of the function f(x) as x~a
as follows.
Let a variable x assume values such (that is, ordered in such
fashion) that if
lx*- a I> lx**-a I.
then x** is the subsequent value and x* is the preceding value;
but if
1 x*-
a 1= 1 x**-
a 1 and .x:~ < x**,
then x**
is the subsequent value and x*
is the preceding value.
In other words, of two points on a number scale, the subsequent
one is that which is closer to the point a; at equal distances, the
subsequent one is that which is to the right of the point a.
Let a variable quantity x ordered in this fashion approach the
limit a[x--+a or limx=a].
Let us further consider the variable y= f(x). We shall here and
henceforward consider that of the two values of a function, the
subsequent one is that which corresponds to the subsequent value
of the argument.
If, as x.--+a, a variable y thus defined approaches a certain
limit b, we shall write
lim f (x) = b
x-+a
and we shall say that the function y. f (x) approaches the limit
y b as X-+a.
It is easy to prove that both
definitions of the limit of a function
are equivalent.
Note 2. If f (x) approaches the limit
b 1 as x approaches a certain number
a, so that x takes on only values less
than a, we write lim f (x) = b 1 and
x-.a-o
call b 1 the limit of the function f (x)
o X on the left of the point a. If x takes
Fig. 32. on only values greater than a, we
write lim f(x)=b 2 and call b 2 the
x ..... a+o
limit of the function on the right of the point a (Fig. 32).
It can be proved that if the limit on the right and the limit on
the left exist and are equal, that is, b 1 = b 2 = b, then b will be
The Limit of a Function 37
the limit in the sense of the foregoing definition of a limit at th~

point a. And conversely, if there exists a limit b of a function at
the point a, then there exist limits of the function at the point a
.both on the right and on the left and they are equal.
Example 1. Let us prove that lim(3x+l)=7. Indeed, let an arbitrary
X-+2
e >0 be given; for the inequality I (3x +I) -71 < 8 to be fulfilled it is neces
sary to have the following inequalities fulfilled:
' 8 8 8
13x-6l<8, lx-21<3, -3<x-2<3
Thus, given any e, for all values of x satisfying the inequality I x-21 <1-=
= 6, the value of the function 3x +I will differ from 7 by less than e. And
this means that 7 is the limit of the function as x - 2.
Note 3. For a function to have a limit as x-a, it is not ne

cessary that the function be defined at the point x =a. When find-
ing the limit we consider the values of the function in the
neighbourhood of the point a that are different from a; this is
clearly illustrated in the following case.
x 2 -4 x 2 -4
Example 2. We shall prove that lim - - = 4. Here, the function - -
X-+Z X - 2 X- 2
is not defined for x = 2.
It is necessary to prove that for an arbitrary 8, there will be a 6 such
that the following inequality will be fulfilled:
lxz-4_4, <e
x-2
(I)
if l x-21 < 6. But when x :F 2 inequality (I) is equivalent to the ine'quality
(x- 2) (x+ 2)
I x-2
41=1 (x+2)-41 <e
lx-21 <e. (2)

Thus, for an arbitrary e, inequality (1) will be fulfilled if inequality (2)
is fulfilled (here, 6=e), which means that the given function ha3 the
number4 as its limit as X-+2.
Let us now consider certain cases of variation of a function

as X-+ oo.
Definition 2. The function f (x) approaches the limit b aA-+ oo
if for each arbitrarily small positive number e it is possible to
indicate a positive number N such that for all values of x that
satisfy the inequality lxi>N the inequality lf(x)-bl<e will
be fulfilled.
Example 3. To prove that

lim (x+l)=l
X-+ CIO X
or
lim
X-+ CIO
(I + _!._) X
= I.
It is necessary to prove that, for an arbitrary e, the following inequality

will be fulfilled
p1ovided 1x 1 > N, where N is determined by the choice of e. Inequality (3)

is equivalent to the following inequality: I; I< e, which will be fulfilled if
ixi>_!._=N.
e
And this means that lim I) hm.
(I+-= x+l
- - = 1 (Fig. 33).
X-+ a> X X-+a> X
Fig. 33.
Knowin~ the meanings of the symbols x-+ oo and x -+- oo the meanings
of the following expressions are obvious:
uf (x) approaches b as x-+ + oo" and
uf (x) approaches b as x-+- oo", or, in symbols,
lim f (x) =b,
X-+ + QO
lim f (x) =b.

x-.-oo
SEC. 3. A FUNCTION THAT APPROACHES INFINITY.

BOUNDED FUNCTIONS
We have considered cases when the function f (x} approaches a
certain limit b as X-+ a or as X-+ oo.
Let us now take the case when the function y = f (x} approaches
infinity when the argument varies in some way.
A Function that Approaches Infinity. Bounded Functions 39
Definition I. The function f (x) approaches infinity as x-+ a,

i.e., it is an infinitely large quantity as x-+ a, if for each
positive number M, no matter how large, it is possible to find
a 6 > 0 such that for all values of x different from a and satisfying
the condition Jx-al<6, we have the inequality Jf(x)J>M.
If f (x) approaches infinity as x-+ a, we write
lim f (x) = oo
or f (x) --+ oo as x-+ a.
If f (x) approaches infinity as X-+ a and, in the process, assumes
only positive or only negative values, the appropriate notation is
lim f (x) = + oo or lim f (x) = - oo.
1
Example I. We shall prove that lim ( - I
x~ -x
)" =+ oo. Indeed, for any
M >0 we will have

1
(1-x)z> M,
provided
1 1
(l-x) 2 <M' ll-xl< YM =~.
The function (l ~x) assumes only positive values (Fig. 34).
Example 2. We shall prove that lim ( -

X-+0
_!_)
X
= oo. Indeed, for any
M > 0 we will have
provided
I
jxl=lx-01 <M=~.
Here ( - ~) > 0 for x < 0 and ( - +) < 0 for x > 0 (Fig. 35).
If the function t (x) approaches infinity as x--+ oo, we write
lim f (x) = oo,
and we may have the particular cases:

lim /(x)=oo, lim /(X)=oo, lim f(x)=-oo
.r-t-+ao X-+-~ X~+ 4
For example,
Note l. The function y = f (x) as x-+ a or as x-+ oo may not

approach a finite limit or infinity.
y
---o-+-----'--....at------x
Fig. 84. Fig. 35:
Example 3. The function y =sin x defined on the infinite interval

- oo < x < + oo, as x-+ +
oo, does not approach either a fin it~ limit or
infinity (Fig. 36).
Fig. 36.
Example 4. The function y=sin .!._ defined for all values of x, except
X.
x=O, does not approach either a finite limH or infinity as x-+ 0. The
graph of this function is shown in Fig. 37.
y y=s;nf
------ -7\-fll'::t-rltH--( ~
_/!]\0 X
-1
Fig. 37.
Definition 2.- The function y = f (x) is called bounded in a given

range of the argument x if there exists a positive number M such
that for all values of x in the range under consideration the
A Function that Approaches Infinity. Bounded Functions 41
inequality If (x) j.:;;; M will be fulfilled. If there is no such num-

ber M, the function f (x) is called unbounded in the given range.
Example 5. The function y=sln x, defined In the infinite interval
- oo< x < + oo, is bounded, since for all values of x
I sin x I .;;;;; 1 = M.
Definition 3. The function f (x) is called bounded as x -+a if

there exists a neighbourhood with centre at the point a, in which
the given function is bounded.
Definition 4. The function y = f (x) is called bounded as x-+ oo
if there exists a number N > 0 such that for all values of x
satisfying the inequality I x 1 > N, the function f (x) is bounded.
The boundedness of a function approaching a limit is decided
by the following theorem.
Theorem 1. If lim f (x) = b, where b is a finite number, the
X-+Q
function f (x) is bounded as x -+a.
Proof. From the equality lim f(x)=b it follows that for any
x-+a
8 >0
there will be a 6 such that in the neighbourhood
a-6<x<a+<'> the inequality
If (x)- b I< e
or
If (x) I < I b I+ s
will be fulfilled, which means that the function f (x) Is bounded
as x-+a.
Note 2. From the definition of a bounded function f (x) it
follows that if
lim f (x) = oo or lim f (x) ---:- oo,
that is, if f (x) is an infinitely large function, it is unbounded.

The converse is not true: an unbounded function may not be
infinite! y Iarge.
For example, the function y = x sin x as x---+ oo is unbounded
because, for any M > 0, values of x can be found such that
1x sin xI > M. But the function y = x sin x is not infinitely large
because it becomes zero when x = 0, n, 2n, . . . The graph of the
unction y = x sin x is shown in Fig. 38.
Theorem 2. If lim f (x) = b =1= 0, then the function y = f (I ) is a
x-.a X
bounded function as x -+a.
ProQf. From the statement of the theorem it follows that for an
arbitrary 8 > 0 in a certain neighbourhood of the point x =a we
wiiJ have lf(x)-bl<e, or llf(x)J-Ib!J<e, or -e<lf(x)l-

- I b I < e, or I b 1-e< If (x) I \f(x)l>lbl+e
For example, taking e=~lbl, we get
10 1 10
9fbl > mxrr > TiJbT
which means that the function (\) is bounded.
1
SEC. 4. INFINITESIMALS AND THEIR BASIC PROPERTIES
ln this seCtion we shall consider functions approaching zero as

~he argument varies in a certain manner.
Definition. The function a= a (x) is called infinitesimal as x--+ a
or as x-+ oo if lim a (x) = 0 or lim a (x) = 0.
X-+a X__..crJ
From the definition of a limit it follows that if, for example,

lim a (x) = 0, this means that for any preassigned arbitrarily
x-+a
small positive e there will be a 6 > 0 such that for all x satisfying
the condition I x-a I< 6, the condition I a (x) I< e will be satisfied.
Example 1. The function a= (x-1) 2 is an infinitesimal as x-+ 1 because
lim a= lim (x-1) 2 =0 (Fig. 39) .
.t-+1 .t--+1
lnfinitesima/s and Their Basic Properties 43
Example 2. The function a =_I_ is an infinitesimal as x--+ oo (Fig. 40)

X
(see Example 3, Sec. 2).
,y
y
Fig. 89. Fig. 40.
Let us establish a relationship that will be important later on.

Theorem l. If the function y = f (x) is in the form of a sum of
a constant b and an infinitesimal a:
(1)
then
Iimy=b (as x-a or x-oo).
Conversely, if limy=b, we may write y=b+a. where a is an.
infinitesimal.
Proof. From equality (1) it follows that IY-bl=lal. But for
an arbitrary 8, all values of a, from a certain value onwards,
satisfy the rei a tionship Ia I< e; consequently, the inequality
I y- b I< 8 will be fulfilled for all values of y from a certain
value onwards. And this means that limy= b.
Conversely: if. limy= b, then given an arbitrary 8, for all
values of y, from a certain value onwards, we will have 1y- b I< e.
But if we denote y- b =a, then it follows that for all values
of a, from a certain one onwards, we will have Ia I< e; and this
means that a is an infinitesimal.
Example 3. Let a function be given (Fig. 41)
1
y=l+-.
X
then
lim y= 1,
X->"'
and, conversely, if
lim y =1
X-+CO
the variable y may be represented in the

!I form of a sum of the limit 1 and the
infinitesimal a=_.!__; that is (Fig. 41),
X
y=1+a.
' 1
Theorem 2. If a= a (x) approaches
zero as x --..a (or as x ---+ oo) and
1
does not become zero, then y=- a
-:0+-------l!J~---~x approaches infimty.
Proof. For any M > 0, no matter
Fig. 41. 1
how large, the inequality iUi > M will
be fulfilled provided the inequality jal< ~ is fulfilled. The latter
inequality will be fulfilled for all values of a, from a certain
one onwards, since a (x) - + 0.
Theorem 3. The algebraic sum of two, three and, in general, a
definite number of infinitesimals is an infinitesimal function.
Proof. We shall prove the theorem for two terms, since the
proof is similar for any number of terms.
Let u (x) =a (x) + ~ (x), where lim a (x) = 0, lim ~ (x) == 0. We
X-+0 X-+G
shall prove that for any 8 >0, no matter how small, there will
be a 0 > 0 such that when the inequality I x-a 1 < 0 is satisfied,
the inequality 1u I< 8 will be fulfilled. Since a (x) is an infinites-
imal, a 0 will be found such that in a neighbourhood with centre
at the point a and radius OP we will have
Ia (x) I< ; .
Since ~ (x) is an infinitesimal, we will have I~ (x) I< ~ in the
neighbourhood of the point a with radius 02
Let us take 0 equal to the smaller of the _two quantities 01 and
02 , then the inequalities Ia 1< ~ and I~ I< f
will be fulfilled in
the neighbourhood of the point a with radius {), Hence, in this
neighbourhood we will have
IuI = Ia<x> + ~ <x> I ~ 1a(x) I +I ~ (x) I < ~ +~ = 8,
and so Iu I< 8, as required.

The proof is similar for the case when
lim a (x) = 0, lim ~ (x) = 0.
X-+ oo X-+ oo
Basic Theorems on Limits 45
Note. Later on we shall have to consider sums of infinitesimals

such that the number of terms increases with a decrease in each
term. In this case, the theorem may not hold. To take an example,
1
consider u=-+-+
X
1
X
+-
.. X1 where x takes on only positive
'"----.,...---"
x terms
integral values (x = 1, 2, 3, ... , n, ... ). It is obvious that as
x--+ oo each term is an infinitesimal, but the sum u = l is not an
infinitesimal.
Theorem 4. The product of the function of an infinitesimal
a= a (x) by a function bounded by z = z (a), as x--+ a (or x--+ co)
is an infinitesimal quantity (function).
Proof. Let us prove the theorem for the case x--+a. For a
certain M > 0 there will be a neighbo_urhood of the point x =a
in which the inequality jzj<M will be satisfied. For any e>O
there will be a neighbourhood in which the inequality Ia I< ~
will be fulfilled. The following inequality will be fulfilled in the
least of these two neighbourhoods:
Jazj< ~ M =e
which means that az is an infinitesimal. The proof is similar for
the case x--+ oo. Two corollaries follow from this theorem.
Corollary 1. If lim a= 0, lim~= 0, then lim a~= 0 because ~ (x)
is a bounded quantity. This holds for any finite number of factors.
Corollary 2. If lim a= 0 and c = const, then lim ca = 0.
Theorem 5. The quotient ; ~:~ obtained by dividing the infini-
tesimal a (x) by a function whose limit differs from zero is an
infinitesimal.
Proof. Let lim a (x) = 0, lim z (x) = b =1= 0. By Theorem 2, Sec. 3,
it follows that z ~a) is a bounded quantity. For this reason, the
a(x) 1
fractions z (x) =a (x) z (x) are a product of an infinitesimal by a
bounded quantity, that is, an infinitesimal.
SEC. 5. BASIC THEOREMS ON LIMITS

In this section, as in the preceding one, we shall consider sets
of functions that depend on the same argument x, where x--+a
or x--+oo.
We shall carry out the proof for one of these cases, since the
other is proved analogously. Sometimes we will not even write
x--+ a or x--+ oo, but will take them for granted.
Theorem 1. The limit of an algebraic sum of two, three and,

in general, any definite number of variables is equal to the
algebraic sum of the limits of these variables:
lim (u, +u, + ... +uk) = limu,-+ lim u, + ... +lim uk.
Proof. We shall carry out the proof for two terms, since it is
the same for any number of terms. Let lim u, =a,, lim u, =a,.
Then on the basis of Theorem 1, Sec. 4, we can write
u, =a, +a,, u,=a,+a.
where a, and a 2 are infinitesimals. Consequently,
u, + u, =(a,+ a,)+ (a,+ a,).
Since (a, +a,) is a constant and (a, +a.) is an infinitesimal,
again by Theorem 1, Sec. 4; we conclude that
lim (u, + u,) = a 1 +a. =lim u, +lim u2
Example 1.
lim x'~ 2x= lim (t+ ~)=lim I+ lim !=I+ lim !=1+0=1.
X-+-CXI X X-+00 X X-+00 . X-+IX> X X-+-CO X
Theorem 2. The limit of a product of two, three and, in general,

any definite number of variables is eqllal to the product of the
limits of these variables:
lim U 1 u, . . . uk =lim U 1 lim u, . . . lim uk.
Proof. To save space we shall carry out the proof for two
factors. Let lim u 1 =a~' lim u, =a,. Therefore,
U 1 = a 1 +a,, u, =a.+ a,,
u,u. =(a,+ a,) (a,+ a,)= a 1 a 2 + a,a, +a, a,+ a 1 a 2
The product a 1 a, is a constant. By the theorems of Sec. 4, the
quantity a,a,+a,a,+a,a, is an infinitesimal. Hence, limu,u,=
= a, a.= lim u, lim u,.
Corollary. A constant factor may be taken outside the limit
sign. Indeed, if lim u 1 =a" c is a constant and, consequently,
limc=c, then lim(cu 1 )=limclimu 1 =Climu 1 , as required.
Example 2.
lim 5x'=5lim x'=58=40.
X-+2 X-+-2
Theorem 3.. The limit of a quotient of two variables is equal

to the quotient of the limits of these variables if the limit of the
denominator is not zero:
. -u = -lim
11m : v =F O
.--u 1'f 1un
v 11m v
Basic Theorems on Limits 47
Proof. Let lim u =a, lim o = b =1= 0. Then u =a+ u, v = b + ~.

where a and ~ are infinitesimals.
We write the id~ntities
u a +a a ( a+ a a) a ab- ~a
-c= b+~ ="b+ b+~ -"b =b"+ b (b+M
or
u a ab-~a
-c= b+ b(b+~).
The fraction is a constant number, while the fraction ba~b-:~,
1;
is an infinitesimal variable by virtue of Theorems 4 and 5 (Sec. 4),
since a.b-~a is an infinitesimal, while. the denominator b (b ~) +
. t bs =1= 0 Th us, 1'Im-c=b"=umv
h as th e IJmJ u a lim u

Example 3.
Jim (3x+5) 3limx+5
+
lim 3x 5 = x .... a = :...;1:...._---,
,::.x'-;-;.... 31 +5_~-4
41-2- 2- .
x-+a 4x-2 lim (4x-2) 4 lim x-2
x-+ 1 x .... 1
Here, we made use of the already proved theorem for the limit of a fraction
because the limit of the denominator differs from zero as x-+ 1. If the limit
of the denominator is zero, the theorem for the limit of a fraction is not
applicable, and special considerations have to be invoked.
2
4
Example 4. Find lim x - .
x-. 2 x- 2
Here the denominator and numerator approach zero as x--+ 2, and, con-
sequently, Theorem 3 is inapplicable. Perform the following identical
transformation:
x 2 -4 (x-2) (x+2)
--= x+2.
x-2 x-2
This transformation holds for all values of x different from 2. And so,
having in view the definition of a limit, we can write
lim (x+2)=4.
X-+2
1
Example 5. Find lim - - . As x--+ I the denominator approaches zero
x-+-t x- 1
bilt the numerator does not (it approaches unity). Thus, the limit ol the
reciprocal quantity is zero:
Whence, by Theorem 2 of the preceding section, we have

. X
1lffi -~=00.
X-+ 1 X -
Theorem 4. If the inequalities u ~ z ~ v are fulfilled between

the corresponding values of three functions u = u (x), z = z (x),
and v = v (x), where u (x) and v (x), as x--+ a (or as x--+ oo ),
approach one and the same limit b, then z = z (x) as x--+ a (or
as x--+ oo) approaches the 5ame limit.
Proof. For definiteness we shall consider variations of the
functions as x -+a. From the inequalities u ~ z ~ v follow the
inequalities
u-b~z-b~v-b;
it is given that
limu=b, limv=b.
X-+-a x-+-a
Consequently, for any e > 0 there will be a certain neighbourhood

with centre at the point a, in which the inequality Iu-bI< e
will be fulfilled; likewise, there will be a certain neighbourhood
with centre at the point a in which the inequality I v- b 1 < e
will be fulfilled. The following inequalities will be fulfilled in the
smaller of these neighbourhoods:
-e<u..:......b<e and -e<v-b<e,
and thus the inequalities
-e<z-b<e
will be fulfilled; that is,
lim z= b.
x-+a
Theorem 5. If as x--+ a (or as x--+ oo) the function y takes on

nonnegative values y;:;;. 0 and, at the same time, approaches the
limit b, then b is a nonnegative number b;:;;. 0.
Proof. Assume that b < 0, then IY- b I;:;;. b; that is, the
difference modulus ly-b I is greater than the positive number lbl
and, hence, does not approach zero as x....:::...a. But then y does
not approach b as x--+ a; this contradicts the statement of the
theorem. Thus, the assumption that b < 0 leads to a contradiction.
Consequently, b;:;;. 0.
In similar fashion we can prove that if y ~ 0, then limy~ 0.
Theorem 6. If the inequality v;:;;. u is fulfilled -between corre-
sponding values of two functions u = u (x) and v = v (x) which
approach limits as x--+a (or as x--+oo), then limv;;;.limu.
Basic Theorem on Limits 49
Proof. It is given that v-u ~ 0. Hence, by

Theorem 5, lim (v- u) ~ 0 or lim v-lim u~ 0,
and so lim v;;;;.: lim u.
Example 6-. Prove that lim sin x = 0.
X->-0
From Fig. 42 it follows that if OA = 1, x >0, then
AC =sin x, AB =x, sin x < x. Obviously, when x < 0
we will have I sin xI-0 Fig. 42.
Example 7 Prove that !i~ 0 sin; =0. Indeed, Isin ; I -0
Example 8. Prove that lim cos.x =I; no1e that

X ->-0
COSX= J-2sin 2 -f ,
therefore,
lim cosx= lim (t-2sln 2 x )=1-2lim

2
sin 2 ~=1-0=1.
X-+0 X-+0 X-+0 2
In some investigations concerning the limits of variables, one

has to solve two independent problems:
1) to prove that the limit of the variable exists and to
establish the boundaries within which the limit under consideration
exists;
2) to calculate the limit to the necessary degree of accuracy.
The first problem is sometimes solved by means of the following
theorem which will be important later on.
Theorem 7. If a variable v is an increasing variable, that is,
each subsequent value is greater than the preceding value, and if
it is bounded, that is, v < M, then this variable- has the limit
lim v = a, where a ,;;;:;; M.
A similar assertion may be made with respect to a decreasing
bounded variable quantity.
We do not give the proof of this theorem here since it is based
on the theory of real numbers, which we shall not consider in
this text.
In the following two sections we shall derive the limits of two
functions that find wide application in mathematics.
50 Limll. Continuity of a Function
SEC. 6. THE LIMIT OF THE FUNCTION si: x AS x-+ 0
This function is not defined for x = 0 since

the numerator and denominator of the fraction
become zero. Let us find the limit of this
function as x----..0. Let us consider a circle
of radius 1 (Fig. 43); denote the central angle
MOB by x; O<x<;. From Fig. 43 it
follows in straightforward fashion that
area 6 MOA <area of sector MOA <
Fig. 43. area 6 GOA. (1)
The area 6M0A= ~ OA-MB=i1sinx=; sinx.
1 ,-.. 1 1
The area of sector MOA= 2 0AAM= 2 1x= 2 x.
1 1 1
The area of 6 COA= 2 0AAC= 2 1tanx= 2 tanx.
. 1
After cancelling 2 , inequality (1) is rewritten
sin x<x< tanx.

Divide all terms by sin x:
1<2..__<_1_
sin x cos x
or
sin x
1 > ->cosx.
X
We derived this inequality on the assumption that x > 0; noting

that sir;(-;)
-x
=sinx and cos(-x)=cosx, we conclude that it
x
holds for x < 0 as well But lim cos x = 1, lim 1 = 1.
x~o x~o
Hence, the variable sin x lies between two quantities that have
X
the same limit (unity). Thus by Theorem 4 of the preceding
section,
. l lsin-x= l.
II
X .... o X
. sin x
The graph of the functwn y = X- is shown in Fig. 44.
The Number e 51
y f
37T X
Fig. 44.
Examples.
1) lim tan x = lim sin x _1_ = lim sin x lim _I_ = 1. _!_ = 1
x,.. 0 X x,.. 0 X COS X x,.. 0 X x-.o COS X I
2) )1m--= . k-
. sin kx . 11m sin kx k . sin (kx) k k
- = 11 m - - = 1 =
t)
(k=COnS,
x-.o. X x-.o kx X-+O (kx)
(kX-+0)
l-cosx
3) l i m - - -
x-+o X
sin ax lim sin ax

ax a x-.o ax
sin ~x =13 lim sin ~x =
~X X-+0 ~X
a 1 a
=If y= If (a=const, ~=canst).
SEC. 7. THE NUMBER E
Let us consider the variable
(t+ ~ r
where n is an increasing variable that takes on the values 1,
2, 3, ...
Theorem t. The variable ( 1 + ~
between the numbers 2 and 3.
r,
as n----.. oo, has a limit
Proof. By Newton's binomial formula we have

+ _!_ ) n = l + _:: _!_ + n (n - 1) (n - 2) (.!_)a -I-
+ n (n - 123
2
I) ( _!_ )
(1 n 1 n 12 n n
... [n-(n-1)] (.!:.)n
... + n(n-l)(n-2)
1 2 .. . n n (I)
Carrying out the obvious algebraic manipulations in (1), we get
( 1+_!_)n
n
= l-j-1 +-1
12
(1 _ _!__)
n
+- 1
123
(1-_!_)
n
(1-~)
n.
+ ...
+ 12 ~ .. n ( 1 - ~) ( 1- ~) ( 1 -
1
nn ) . (2)
From the latter equality it follows that the variable ( 1+_!_f

n '
is an increasing variable as n increases.
Indeed, when passing from the value n to the value n+ 1, each
term in the latter sum increases,
I(
-1--
12 , n
I) < -11
12
( -n +-
1)1 and so forth,
and another_ term is added. (All terms of the expansion are

positive.)
We shall show that the variable ( 1 + ~ is .bounded. Noting r
that (1-~) <1;
(1-~)(1-~)<1, etc., we obtain from
expression (2) the inequality
(1 -~-~r <l+ 1 +1~2+1.;.3+ ... +12L ... n
Further noting that

I I _I_<]_. (
123 < 22; 1234 21 ' 12 ... n < 2n-t 1
we can write the inequality
(
1 )n
1 +n- <1+1+2+21+ +=~
1 1 1
The grouped terms on the right-hand side of this inequality form

a geometric progression with the common ratio q = ; and the
first term a=1, and so
(1 +n1 )n < 1+ [1 + 21 +21'+

1 ... + 2 1 n-J
1
=
1- ( -1 )n
a-aqn
=1+ 1,-q =1+ 1-]_
2
=1+ 2- 2
[ ( 1 )n-t] <3.
2
The Number e 53
Consequently, for all n we get
(1 +! r < 3.
From equality (2) it follows that
(1 +*r~2.
Thus, we get the inequality
2 ~ ( 1+ ! ) < 3,
n (3)
This proves that the variable ( 1 + ~ r is bounded.

Thus, the variable (1+*r is an increasing and bounded
variable; therefore, by Theorem 7, Sec. 5, it has a limit. This
limit is denoted by the letter e.
Definition. The limit of the variable ( 1 + ~
number e:
as n-H:O is the r
e = lim ( 1 + _!_)n. *)
no+oo n
By Theorem 6, Sec. 5, it follows from inequality (3) that the

number e satisfies the inequality 2 ~ e ~ 3. The "theorem is thus
proved.
The number e is an irrational number. Later on, a method will
be shown that permits calculating e to any degree of accuracy. Its
value to ten significant decimal places is
e = 2. 7182818284 ...
Theorem 2. The function ( 1 + ~ )" approaches the limit e as x

approaches infinity, lim ( 1 +_!_)"=e.
X-+-c:o X
Proof. It has been shown that ( 1 + ~ -+e as n-+ oo, if n r

takes on positive integral values. Now let x approach infinity
while taking on fractional and negative values.
) It may be shown that (I+~ ) "-+ e as n-+ + oo even if n is not an

increasing variable quantity.
1) Let x---++oo. Each of its values lies between two positive

integral numbers,
n~x<n+l.
The following inequalities will be fulfilled:

1 1 1
n~x>n+1'
1 1 1
1 +n-~ 1-1-x-> 1+n+l'
( 1 + ~ )n+l > ( 1 + ~ )" > ( 1 + n~ I )n
If x---+ oo, it is obvious that n--+ oo. Let us find the limits of the
variables between which the variable ( 1 +~ lies: r
lim ( 1 + J...)n+l = lim ( i + J...)n ( 1 + J...) =
n-++oo n n-++oo , . n n
=lim (1-1-J...)n lim (1+J...).=e1=e,
n-++oo n n-++oo n
( 1+-11)n+l
lim ( 1 + - 1
n-++oo
-)n =
n+ 1
lim
n-++oo
n+
1+-1-
-
n+1
1
lim (1+- -)n+l
n-++oo n+1 e
------'-..,..----'--=-"~ = - = e.
1 1
lim (1+--)
n-++ao n+ 1
Hence, by Theorem 4, Sec. 5,
lim
x-..+CD
(I +!)"=e. X
(4)
2) Let X--+-oo. We introduce a new variable t = - (x+ l) or

x=-(1+1). When t--.+oo then X---+-oo. We can write
lim
X-+-oo
(1 +-I)" =
X
lim
1-++ao
(1 - t +11)-t-1 = lim
1-++ao
(t +t 1 )-t-1 =
_
1 )1+1 ( 1 + T1 )I+ I=
= lim ( t+
-t- = lim
1-++ao 1-+ao
= /~U:, +
(1 + Y(1 + +) = e 1 =e.
The Number e 55
The theorem is proved. The graph of the function, y =

is shown in Fig. 45.
(1+ +r
y
y=(t+fY
+1
-1 0 )(
Fig. 45.
If in equality (4) we put ! =a, then as x-+oo we have a-+0

(but a. + 0) and we get
I
lim(l+a)=e.
ao+o
Examples:
I) lim (r+_!_)n+=
R-t-CDn
Urn (l+...!..)n(l+....!...)'=
n R-+CD n
=lim
n_..ao
(1-1-....!...)n.
n
lim (r+....!...)'
n-+ oo n
=e1=e.
2) lim
JC-+CIO ( 1+-XI )ax =lim
JC-+00
(1+-I )"' ( I+-I )"' ( 1+-I )"' =
X X X
=lim
.r_..oo
(r+J..)"'.
X
lim (r+J..)"'.
X-+co X
Jim (I+_!_)"'
X
=eee=e
X-+CIO
3) lim
xo+oo ( lim ( I+-y1 )2y =;e
I+-X2 )"' = IJ-+00 2

1
4) . (x+3)"'+
hm -- = hm. (x-1 +4)"'+ 1 = hm 4 )"'+' = '
( 1+ --
x-+oox-i x-+oo x-1 x-+oo x-1
4 )(X -tr+ 4 ( 1+-4 )Y+4=
=lim
x-+oo ( 1+--
x-l
=lim
y-+oo y
=lim
JI-+<IJ
(1+i)Y.
y
Urn (r+.!)"=e"l=e".
IJ-+00 y
SEC. 8. NATURAL LOGARITHMS
In Sec. 8 of Chapter I we defined the logarithmic function

y = log 4 x. The number a is called the base of the logarithms.
If a= 10, then y is the decimal (common) logarithm of the num-
ber x and is denoted y= Iogx. In school courses of mathematics
we have tables of common logarithms, which are called Briggs'
logarithms after the English mathematician Briggs (1556-1630).
Logarithms to the base e = 2. 71828. . . are called natural or
Napierian logarithms after one of the first inventors of logarithmic
y,=tnx
y,
Yz A2 togto-t
7 X 8 g X
Fig. 46.
tables, the mathematician Napier (1550-1617). *) Therefore, if

eY = x, then y is called the natural logarithm of the number x. In
wnting we have y =In x (after the initial letters of logarithmus
natura/is) in place of y =loge x. Graphs of the function y = ln x
and y =log x are plotted in Fig. 46.
Let us now establish a relationship between decimal and
natural logarithms of one and the same number x.
Let y= logx or X= IOY. We take logarithms of the left and
right sides of the latter equality to the baseeandget lnx=ylnlO.
We determine y = ln\o ln x, or, substituting the value of y, we
I
have logx= In 10 lnx.
Thus, if we know the natural logarithm of a number x, the com-
mon (decimal) logarithm of this number is found by multiplying
by the factor M = ln 10 ~ 0.434294, which factor is independent
1
of x. The number M is the modulus of common logarithms with

_respect to natural logarithms:
logx= M Jnx.
*) The first logarithmic tables were constructed by the Swiss mathemati-

cian Biirgi (1552-1632) to a base close to the number e.
C,ontinuitg of Functions 57
If in this identity we put x = e, we obtain an expression of the

number M in terms of common logarithms:
loge=M(lne= 1).
Natural logarithms are expressed in terms of common logarithms
as follows:
1
lnx = M logx
where
1
M = 2.302585.
SEC. 9. CONTINUITY OF FUNCTIONS
Let the function y=f(x) be defined for some value x 0 and in

some neighbourhood with centre at X 0 Let Yo= f (x 0 ).
If x receives some positive or negative (it is immaterial which)
increment !lx and assumes the value !/
X=X 0 +1lx, then the function y too will
receive an increment !ly. The new in-
creased value of the function will be
+
Yo+ !ly = f (X 0 !lx) (Fig. 47). The ittcre-
ment of the function !ly will be ex pressed
by the formula
!ly = i (X + !lx)- f (X
0 0 ).
Definition 1. The function y = f (x) is 0 Xo x0+Lix X'
called continuous for the value x = X 0 Fig. 47.

(or at the point X 0 ) if it is defined in
some neighbourhood of the point x0 (obviously, at the point X0
as well) and if
(1)
or, which is the same thing,

lim [f (x 0
t;x .... o
+ 8x)-f (X 0 )] = 0. (2)
In descriptive geometrical terms, the continuity of a function at a

given point signifies that the difference of the ordinates of the
graph of the function Y=f(x) at the points X 0 +1lx and X 0 will,
in absolute magnitude, be arbitrarily small, provided I !lx 1 is
sufficiently small.
Example 1. We shall prove that the function y=x is continuous at an

arbitrary point x 0 Indeed,
Yo= x!, Yo+ b,.y = (x 0 + b,.x) 2, b,.y = (x 0 + b..X) 2 -x! = 2x 0 b,.x + b..X 2 ,
lim b,.y = lim (2x 0 b,.x + b,.x 2) = 2x lim b,.x + lim b,.x Urn b,.x = 0
il.X-+0 il.X-+0 il.X-+0 il.X-+0 il.X-+0
for any way that b,.x may approach zero (Figs. 48, a and 48, b).
, y Ax>O, Ay>O
(a) (b)
.Fig. 48.
Example 2. We shall prove that the function y=slnx is con.tinuous at

an arbitrary point x 0 Indeed,
Yo= sin x0 , Yo+ b..y =sin (X 0 + b,.x),
b,.y= sin (x 0 + b,.x)-sln x 0 =2 sin ~x Cos ( x0 + b.;) .

It was shown that lim sin b,.x =0 (Example 7, Sec. 5). The function
2
il.x -+0
c'os (x + b,.x) is bounded. Therefore, llm b..Y = 0.

2 ll.x -+0
In similar fashion, it is possible to prove the following theorem

by considering each basic elementary function and each
elementary function.
Theorem. Every elementary function is continuous at each point
at which it is defined.
The condition of continuity (2) may be written thus:
lim f (x 0 + Llx) = J (x 0 )
ll.x ... o
or
lim f (x) = f (x 0 ),
%-+Xo
but
Continuity of Functions 59
Con seq uentl y,

lim f (x) = f ( lim x). (3)
X-+Xo X-+X 0
In other words, in order to find the limit of a continuous function

as x - X0 it is sufficient to substitute into the expression of the
function the value of the argument, X0 , in place of the argument x.
Example 3. The function y = x 2 is continuous at every point x0 and
therefore
lim x 2 = x~,
X-+X 0
lim x 1 =3 2 =9.
x .....
Example 4. The funct-ion y =sin x is continuous at every point and

therefore
2
rt"!t s1n x=s1"'
n =Y
-- .
x .... -
4 2
&
Example 5. The function y=e" is continuous at every point and therefore

lim e" =e 4
x .... a
I
E xamp Ie 6. lim
In (1 + x) lim ..!..1n (1 +x)= lim In [(1 +x)x1. Since
x .... o X X-+0 X X-+0
lim (1 +x) x =e and the function In z is continuous for z > 0, and,

x .... o
consequently, for z=e,
lim In [(l
X-+0
+ x) x 1=In [ lim X-+0
(I + x) x 1=In e = 1.
Definition 2. If the function y = f (x) is continuous at each point

of a certain interval (a, b), where a< b, then it is said that the
function is continuous in this interval.
If the fu net ion is also defined for x =a and lim f (x) = f (a),
x-+a+o
it is said that f (x) at the point x =a is continuous on the right.
If lim f(x)=f(b), it is said that the function f(x) is continuous
x .-.b-o
on the left of the point x =b.
If the function f (x) is continuous at each point of the interval
(a, b) and is continuous at the end points of the interval, on the
right and left, respectively, it is said that the function f(x) is
continuous over the closed interval [a, b].
Example 7. The function y=x 1 is continuous in any closed interval [a, b}.
'This follows from Example 1.
If at some point x = x., at least one of the conditions of conti-

nuity is not fulfilled for the function y =.f (x), that is, if for x = x.
the function is not defined or there does not exist a limit lim f (x)
X~Xo
or lim f(x) =I= f(x 0 ) in the arbitrary approach of x-x., although

X-+Xo
the expressions on the right and left exist, then at x = x 0 the
function y = f (x) is discontinuous. In this case, the point x = x.
is called the point of discontinuity of the function.
Example 8. The function y=_!_ is discontinuous at x=O. Indeed, the

X
function is not defined at x=O.
I
lim _!_ =+ oo; lim - = - oo
X-+0+0 X X-+0-0 X
(see Fig. 35). It is easy to show that this function is continuous for any
value x-# 0.
Example 9. The function y=2x is discontinuous at x=O. Indeed,

I I
lim 2 x = oo, lim 2x = 0. The function is not defined at x = 0 (Fig. 49).
y
yf(x)
X yf(X) -f
Fig. 49. Fig. 50.
Example 10. Consider the function f(x)= ; At x<O, ; =-1;

1 1 1 1
X
at x>O, TXT=l. Hence,
lim /(x)= lim ~=-1;

X-+0-0 A-+0-0 lXI
1
lim f (x) = lim - - =I;
X-+O+O X-+O+O 1X 1
the function is not defined at x=O. We have thus established the fact that
the functionf(x)= ; is discontinuous at x=O (Fig. 50).
1 1
Certai1t Properties of Contrnuous Functions 61
Example 11. The earlier examined function y =sin_!_ is discontinuous

X
at x=O.
Definition 3. If the function f (x) is such that there exist finite
x~x0 +o
+
limits lim f (x) = f (x 0 0) and lim f (x) = f (X 0 -0), but either
x-..x0 -o
lim f (x) =F lim f (x) or the value of the function f (x) at x = X0
X-+X 0 +0 X-+X 0 -0 -
is not defined, then x = X 0 is called a point of discontinuity of the
first kind. (For example, for the function considered in Example 10,
the point x = 0 is a point of discontinuity of the first kind).
SEC. 10. CERTAIN PROPERTIES OF CONTINUOUS FUNCTIONS
In this section we shall consider a number of properties of

functions that are continuous on an interval. These properties
will be stated in the form of theorems given without proof.
Theorem 1. If a function y = f (x) is continuous on some inter-
val [a, b] (a~ x ~b), there will be, on this interval at least one
point x = X 1 such that the value of the function at this point will
satisfy the relation
f (x 1 ) ~ f (x),
where x is any other point of the interval, and there will be at
least one point X2 such that the value of the function at this point
will satisfy the relation
We !;hall call the value of the function f (x 1 ) the greatest value

of the function y = f (x) on the interval [a, b], and the value of
the- function f (x2 ) the smallest
(least) value of the function on U
the interval [a, b ].
This theorem is briefly stated as
follows:
A function continuous on the M
interval a~ x ~ b attains on this
interval (at least once) a greatest 0 a b X
value M and a smallest value m.
Fig. 51.
The meaning of this theorem is
clearly illustrated in Fig. 51.
Note. The assertion that there exists a greatest value of the
function may prove incorrect if one considers the values of the
function in the interval a<x<b. For instance, if we consider
the function y=x in the interval O<x< 1, there will be no
G2 Limit. Continuity of a Function
~;reatest and no least (smallest) values among them. Indeed, there

is no least value or greatest value of x in the interval. (There
is no extreme left point, since no matter what point x* we take
there will be a point left of it, for instance, the point ~; like-
wise, there is no extreme right point; consequently, there is no
least and no greatest value of the function y = x.)
Theorem 2. Let the function y = f (x) be continuous on the inter-
val [a, b] and at the end point of this interval let it take on
values of different sipn; then bet ween the points a and b there will
be at least one point x = c, at which the function becomes zero:
f(c) = 0, a< c< b.
This theorem has a simple geometrical meaning. The graph of a
continuous function y = f(x) joining the points M 1 [a, f (a)) and
M 2 [b, f(b)J, where f(a)<O and f(b)>O .!/ 1
or f (a)> 0 and f (b)< 0, cuts the x-axis l
at least at one point (Fig. 52).
6
lj
Mtfq.f(a)/
/'
Fig. 52. Fig. 53.
Example. Given the function y=x'-2. Yx= 1 =-l, Y:c= 1 =6. It is conti-
nuous in the interval [1. 2]. Hence, in this interval there is a poit1t where
y = x'-2 becomes zero. Indeed, y=O when x= t/2 Wig. 53).
Theorem 3. Let the function y = f (x) be defined and continuous

in the interval [a, b). If at the end poin.ts of this interval the
function takes on unequal values f (a)= A, f (b)= B, then no mat-
ter what the number ~ between numbers A and B, there will be a
point X=C between a and b such that f(c)=~.
The meaning of this theorem is clearly illustrated in Fig. 54.
In the given case, any straight line y = 1-1- cuts the graph of the
function y = f.(x).
Comparing I nfinitesimals 63
Note. It will be noted that Theorem 2 is a particular case of

this theorem, for if A and B have different signs, then for f.t one
can take 0, and then f.t = 0 will lie between the numbers A and B.
X
X
Fig. 54. Fig. 55.
Corollary of Theorem 3. If a function y = f (x) is continuous ia

some interval and takes on a greatest value and a least value,
then in this interval it takes on, at least once, any value lying
between the greatest and least values.
Indeed, let f (x 1 ) = M, f (x 2 ) = m. Consider the interval [x 1 , x.].
By Theorem 3, in this interval the function y = f (x) takes on
any value f.t lying between M and m. But the interval [x 1 , x 2 ]
lies inside the interval under consideration tn which the function
f (x} is defined (Fig. 55).
SEC. 11. COMPARING INFINITESIMALS
Let several infinitesimal quantities

a, ~. y, ...
be at the same time functions of one and the same argument x
and let them approach zero as x approaches some limit a or
infinity. We shall describe the approach of these variables to zero
when we consider their ratios.*)
We shall, in future, make use of the following definitions.
Definition 1. If the ratio 1. has a finite nonzero limit, that
a
is, if lim ~ =A =1= 0, and. therefore, lim ~ = +
=1=0, the infinit_es
imals -~ and a are called infinitesimals of the same order.
*) We assume that the infinitesimal in the denominator does not vanish

in some neighbourhood of the point a.
Example 1. Let a =X, ~=sin 2x, where x-+ 0. The inflnitesimals a and ~
are of the same order because
. ~ . sin 2x
1lffi - = 1liD --=2.
X-+0 a X-+-0 X
Example 2. When x-+ 0, the infinitesimals x, sln3x, tan 2x, 7ln (1 +x) are
infinitesimals of the same order. The proof is similar to that given in
Example 1.
Definition 2. If the ratio of two infinitesimals 1..

a
approaches
zero, that is, if lim ~ = 0 (and lim ~ = oo),
then. the infinitesi-
mal ~ is called an infinitesimal of higher order than a, and the
infinitesimal a is called, an infinitesimal of lower order than ~-
Example 3. Let a=x, ~=xn, n> l, X-+0. The infinitesimal ~is an
infinitesimal of higher order than the infinitesimal a, since
lim xn = lim xn-t =0.
x-+-0 X X-+-0
Here, the infinitesimal a is an infinitesimal of lower order than ~-
Definition 3. An infinitesimal ~ is called an infinitesimal of the

kth order relative to an infinitesimal a, if~ and ak are infinitesimals
of the same order, that is, if lim ~k = A =I= 0.
a
Example 4. If a =x, ~ = x 3 , then as x-+ 0 the infinitesimal ~ is an
infinitesimal of the third order relative to the infinitesimal a since
. ~
11m-= 1'1m -x= 1
3 1
x-+o a x-+ o (x)
Definition 4. If the ratio of two infinitesimals 1..

a
approaches
unity, that is, if lim ~ = 1, the infinitesimals ~ and a are called
equivalent infinitesimals and we write a....__ ~-
Example 5. Let a=x and ~=slnx, where X-+0. The infinitesimals a
and ~ are equivalent, since
lim ~=1.
X-+0 X
Example 6. Let a =x, ~=In (1 +x), where x-+ 0. The infinitesimals a

and ~ are equivalent, since
lim ln(1+x)
.J-+0 X
Comparing lnfinitesimals 65
Theorem 1. If a and ~ are equivalent infinitesimals, their diUer-

ence a-~ is an infinitesimal of higher order than a and than ~
Proof. Indeed,
lim a- ~=lim ( 1-1_) = 1-lim !_ = 1- 1 = 0.

a a a
Theorem 2. If the diUerence of two infinitesimals a-~ is an

infinitesimal of higher order than a and than ~. then a and ~ are
equivalent infinitesimals.
Proof. Let Jim a-~
a
=0, then lim ( 1-!)
a
= 0, or 1-lim 1=
a
0,
or l=lim ~ , i.e., a~~ If lim a~ ~=0, then lim ( ~ -1) =0,
lim ~ = 1, that is, a~ ~
Example 7. Let a=x, ~=x+xs, where x---+ 0.

The infinitesimals a and ~are equivalent, since their difference ~~a=xs
i> an infinitesimal of higher order than a and than ~ Indeed,
~-a xs
lim - - = lim -=lim x 2 =0,
x~o a x~o X x~o
. a-~ .
lim-~~.-= ltm - +
xs
_3 = ltm I+
. x2
=0.
2
p X~ 0 X X X~ 0 X
Example 8. As X--->- oo the infinitesimalsa=x+2 l and

- X
~~I.areequivalent
X
.111 fiInlt es1ma
. Is, s1nce
. th.
e1r d'IT x+l
A =
1 erence a - p -2- - -l= 2I Is
an 1111n1
. fi 't es1ma
I
X X X
of higher order than a and than ~ The limit of the ratio of a and ~ is
unity:
lim!_= lim +xi= lim x+I= lim (t+...!..)=L

a x~wx x-+CJJ X x~w X
X2
Note. If the ratio of two infinitesimals 1_
a
has no limit and
does not approach infinity, then ~ and a are not comparable in
the above sense.
Example 9. Let a=x, ~=x. sin_!_,
X
where x--->- 0. The infinitesimals
a and ~cannot be compared because their ratio !_=sin...!..

a x
as X--->- 0 does not
approach either a finite limit or infinity (see Example 4, Sec. 3).
3-3388
Exercises on Chapter II
Find the indicated limits:
2. lim [2 sinx-cos x+co! xl. Ans. 2.

:rt
X-+-
2
3. lim x- 2 . Ans. 0. 4. lim (2-_.!_+.!). Ans. 2. 5. lim 4x-52x"+ 1 .

x-+ Y2+x x-+c:o x x x-+c:o 3x -5
Ans. 4
3
. 6. lim x+ 1
X-+CO X
Ans. l. 7.
n-...co n .
lim 1 +2+ ... +n
. Ans.
2
.
8. lim 1"+ 2"+ 3'+ +n'. Ans. _!_.

n-+c:o n 3
Hint. Write the formula (k+1) 3 -k 3 =3k 2 +3k+1 for k=O, 1, 2, ... , n."
P=1;
23 -1'=3 !2+31 + 1;
33 -2 3 =32'+32+ I;
(n + 1) 3 -n 3 =3n'+ 3n + l.
Adding the left and right sides, we get

(n+ 1) 3 =3 (P+ 2'+ ... +n'} +3 (I +2+ ... +n}+(n + 1},
(n+ 1} 3 =3 (12+2'+ ... +n"}-3 n (n: I} +<n+ I},
whence
9. lim x+x-1 Ans. co. 10. lim 3x"- 2x-l . Ans. 0.

x-+c:o 2x+S x-+"' x +4
4x"-2x"+x I 1
11. lim Ans. . 12. lim x"-4. Ans. 4. 13. lim x - .' Ans. 3.
x -..o 3x 2 + 2x 2 x-.. x-2 x-+1 x-1
14. lim x-Sx+6 Ans. 1 15. rtm x 2 +3x-10 . Ans. I.

x-+ x- 12x +i!u 8" x-+ 3x-sx-2
lim u+3u'+2y 2 u"+4u 2 +4u

16.
y-.. -
y-y-6
Ans. --s 17. lim
u-..- (u+2){u-3)
Ans. 0.
18. lim (x+h) 3 -x'. Ans. 3x 2 19. lim [1 1 x -1 3 xJ- Ans. -I.
h-..o h X-+1
20. lim - - . A ns. n (n .IS a pos1'!'1ve .mt eger.

xn-1 ) 21. 1'1m 'VT+X-1 . A ns.-.
1
X .... l X-1 X-+0 X 2
Exercises on Chapter /1
22. lim Jf2x+l-3 22

Ans. -3-. Ans. J.....
x-H Yx-2- V 2 p
24. lim
3 --
)/ x-1
. Ans. 2 25. lim 'V -x- vr-a
~--"--- Ans. 'Vli.
x-+t V x-I 3' x-+a x-a ma
26.
2
lim VI +x+x - l Ans. I 27. lim Vx 2 -3 Ans. l.
X-+0 X 2' x-++"' Vx'+I
28. VXi+T. Ans. I asx-++oo. - I asx-+-oo. 29. lim (Yx"+l~

lim
x-+"'x+ I x-+"'
1
- Y x 2 -l). Ans. 0. 30. lim x ( Vx"+ I-x). Ans. - as x-+- + oo, -oo as
X-+ CD 2
sin 2~
X -+-00. 31. Jim sin x . Ans I. 32. lim sin 4x . Ans. 4. 33. lim __3_ ,
x-+o tanx X-+-0 .t X-+o x 1
Ans. I
. 34. lim x . Ans. 2 35. lim x cot x. Ans. I.
9 x-+o YI-cosx V2' X-+0
36. lim 1-2 cos v . Ans. y3. 37. Jim (1-z) tan nz . Ans. 2
n
u-+ ~ sin ( v - ~ )
z-+t 2
38. lim 2 arc sin x . Ans. 2 39. lim sln(a+x)-sln(a-x). Ans. 2 cos c.
X-+0 3x 3' X-+0 X
40. lim tanx-slnx.

x-+o x3
Ans. I
2'
41. Jim (I+~X r. Ans. e.
r
X-+-CO
lim ( 1 --~-r. -.
42. Ans. 43. lim __ . Ans.
( X
X-+CO X e x-+"' I +x e
44. lim ( I+-

I r+ . Ans. e. 45. lim {n [In (n +I)-In nl}. Ans. I.
n-+ co n. n-+"'
46. lim (I +cos x)' sec". Ans. e. 47. lim In (I +ax) Ans. a.
X-+-
:n: X-+0 X

1
48. .
hm
X-+a>
cx+3r+
--
2x+J
Ans. e. 49. lim (1 +3 tan x)co!' ". '
X-+0
Ans. e.
50. lim (cos.!!._)m. Ans. I. 51. lim In(! +ea). Ans. 1 as a-++ oo, 0 as
m-..CD m a-.co a
a><-J
a-+-oo. 52. lim~. Ans. ~ 53. lim - - ( a > 1). An.s. +.co
x ... o sin ~x x.,.., X
68 Limit. Continuity ol a Function
2.. ] eu-eil-"
as x . . .,. + oo, 0 as x -+- oo. 54. lim n [ a n - I . Ans. In a. 55. lim x
n~oo x-..o
Ans. a-~. 56. lim e~x-e~x Ans. 1.
r-+ o sin ax-sin ~x
Determine the points of discontinuity of the functions:
57. Y= x (x;D(~ 2 _ 4 ). Ans. Discontinuities of second kind for x= -2; -I;
0;2.58.y=tan_!_. Ans. Discontinuities of second kind for x=O and

X
x= n2 ;
2
3:rt ; ;
2
(2n + I) :rt ;
1
59. Find the points of discontinuity of the functions y=1+2x and con-
struct the graph of this function. Ans. Discontinuity of second kind at x=O
(y-...+oo as x--..0+0, y ....... l as X->-0-0).
60. From among the following infinitesimals (as x ....... 0); x 2 , V x (1-x),
sin 3x, 2x cos x V
tan 2 x, xe 2 x, select infinitesimals of the same order as x,
and also of higher and lower order than x. Ans. lnfinitesimals of the same
order are sin 3x and xe 2 x; infinitesimals of higher order, x 1 and 2x cos x ~an2 x, V
infinitesimals of lower order, V x (1-x).
61. Choose from among the same infinitesimals (as x ....... 0) such that are
equivalent to the infinitesimal ;.:: 2sin x, ~tan 2x, x-3x 2 , V2x 1 +xa,
ln(l+x), x3+3x'. Ans. { tan2x, x-3x 2 , ln(l+x).

62. Check to see that as x ....... 1, the inflnitesimals 1-x and 1- Vx are
of the same order of smallness. Are they equivalent? Ans. lim l-x 3;
x-+11-Vx
hence, these infinitesimals are of the same order, but they are not equivalent.
CHAPTER III
DERIVATIVE AND DIFFERENTIAL
SEC. t. VELOCITY MOTION
Let us consider the rectilinear motion of some solid, say a stone,

thrown vertically upwards, or the motion of a piston in the cylin-
der of an engine, etc. Idealising the situation and disregarding
dimensions and shapes, we shall always represent such a body in
the form of a moving point M. The distance s of the
moving point reckoned from some initial position M 0
will depend on the time t; in other words, s will be a
function of time t:
s = f (f). ( 1)
At some instant of time*) t, let the moving point M

be at a distance s from the initial position M 0 , and at
some later instant t + M let the point be at M ~' a
distance s+ !:J.s from the initial position (Fig. 56). Thus, Fig. 56.
during the interval of time M the distance s changed
by the quantity !:J.s. In such cases, one says that during the time
M the quantity s received an increment !:J.s.
Let us consider the ratio ~~; it gives us the average velocity of
motion of the point during the time !:J.t:
!ls
Vav = M (2)
The average velocity cannot in all cases give an exact picture
of the rate of translation of the point M at time t. If, for example,
the body moved very fast at the beginning of the interval M and
very slow at the end, the average velocity obviously cannot reflect
these peculiarities in the motion of the point and give us a correct
idea of the true velocity of motion at time t. In order to express
more precisely this true velocity in terms of the average velocity,
one has to take a small interval of time M. The most complete
description of the rate of motion of the point at time t is given
by the limit which the average velocity approaches as M _,.. 0.
1 Here and henceforward we shall denote the specific value of a variable

and the variable itself by the same letter.
70 Der'iuallue and Difjerenf ial
This limit is called the rate of motion at a given instant:

v = r!ill 6.s
6.t. (3)
6.1-+0
Thus, the rate (velocity) of motion at a given instant is the limit

of the ratio of increment of path /1s to increment of time M, as
the time increment approaches zero.
Let us write equality (3) in full. Since
11s= f (t + M)-f (t),
v = run f(t+M>-tU>
111
(3')
M .... o
This is the velocity of variable motion. It is thus obvious that

the notion of velocity of variable motion is intimately related to
the concept of a limit. It is only with the aid of the limit concept
that we can determine the velocity of variable motion.
From formula (3') it follows that v is independent of the increment
in time M, but depends on the value of t and the type of
function f (t).
Example. Find the velocity of uniformly accelerated motion at an. arbitrary
time t and at 1=2 sec if the relation of the path traversed to the time is
expressed by the formula
I
S=2gfe.
Solution. At time t we have s= ~ gt'; at time t+ M we get
I I
2 g(t +M) ="2 g W+2t 6.t +6.t').
2
s+6.s=
We find 6.s:
I I I
6.s=2: g(t'+2t M +M')-"2 gt'=gt M +2g 6.t.
We form the ratio !: :

I
6.s gt M+2gM' 1
'Ki= 6.t gt+2g M;
by definition we have
u=lim ~st =lim
t.t .... ou t.t .... o
(et+
1
2
gM)=gt.
Thus, the velocity at an arbitrary time t i.st u =gt.

At 1=2 we have (. )j, 2 =g2=9.82= 19.6 mjsec.
Definition of Derivative. 71
SEC. 2. DEFINITION OF DERIVATIVE
Let there be a function

y = f(x) (I)
defined in a certain interval. The function y=f(x) has a definite
value for each value of the argument x in this interval.
Let the argument x receive a certain increment !1x (it is imma-
terial whether it be positive or negative). Then the function y will
receive a certain increment !1y. Thus, with the value of the argu-
ment x we will have y=f(x), with the value of the argument
+ +
x !1x we will have y !1y = f (x !1x). +
Let us find the increment of the function !1y:
!1y = f (x + !1x)- f (x). (2)
Forming the ratio of the increment of the function to the increment
of the argument, we get
llq !(x+llx)-f(x)
llx = llx (3)
We then find the limit of this ratio as !1x-+ 0. If this limit exists,
it is called the derivative of the given function f (x) and is denoted
f' (x). Thus, by definition,
f' (x) =lim
ll.X-
flY
0 /).X
or
f' (x) =lim f (x + /lX)- f (x). (4)
ll.X -o /).X
Consequently, the derivative of a given function y = f (x) with

respect to the argument x is the limit of the ratio of the increment
of the function !1y to the increment of the argument !1x, when
the latter approaches zero in arbitrary fashion.
It will be noted that in the general case, the derivative f' (x)
has a definite value for each value of x, which means that the
derivative is also a function of x.
The designation f' (x) is not the only one used for a derivative.
Alternative symbols are
I I dl/
Y 'Y "' dx'
The specific value of the derivative for x =a is denoted f' (a) or
YIIT=h~operation of finding the derivative of a function f (x) is

called diOerentiation of the function.
72 Derivative and Differential
Example t. Given the function y=x 2 ; find its derivaUve y':

1) at an arbitrary point x,
2) at x=3.
Solution. 1) For the value of the argument x, we have lf=X 2 When the
value of the argument is x + tlx, we have y + llY = (x + tlx) 2
Find the increment of the function:
llY = (x + llx) 2 -x 2 = 2xtlx + (llx) 2
Forming the ratio A!J; we have
tlX
llY = 2xllx + (llx) 2x + M.
2
tlX tlX
Passing to the limit, we get the derivative of the given function:
y' = lim tJ.y =lim (2x
t.x-+ o llX t.x-+ o
+ tlx) = 2x.
Hence, the derivative of the function y=x 2 at an arbitrary point is y' =2x,
2) When x=3 we have
y' \ x=a=23=6.
Example 2. y=..!..; find y'.
X
Solution. Reasoning as before, we get
1 I
lj =X ; y +
llY = X+ llX;
- 1 I X-X-llX /!lX
M= x+llx --x= x (x+f!lx) x (x+llx)'
llY I
tlX = -X (X+ /!lX) ;
y' = t.x-+ollXt.x-+o
lim l!ly lim [-
x(x+f!lx)
1
J= - .!.. .x2
Note. In the preceding section it was established that if the

dependence upon time t of the distance s of a moving point is
S=f(t),
the velocity v at time t is expressed by the formula
v = lim l!ls = lim f (t + M)- I (t)
t.l-+ 0 Ill t.t-+ 0 llt
Hence
v = s~ = f' (t),
or, the velocity is equal to the derivative*) of the distance with
respect to the time.
) When we say "the derivative with respect to x" or "the derivative
with respect to t" we mean that in computrng the derivative we consider
the variable x (or the time t, etc.) the argument (independent variable).
Geometric Meaning of the Derivative 73
SEC. 3. GEOMETRIC MEANING OF THE DERIVATIVE

We approached the notion of a derivative by regarding the
velocity of a moving body (point), that is to say, by proceeding
from mechanical concepts. We shall now give a no less important
geometric interpretation of the derivative. To do this we must
first define a line tangent to a curve at a given point.
We take a curve with a fixed point M 0 on it. Taking a point
M, on the curve we draw the secant M0 M, (Fig .57). If the point
M, approaches the point M 0 without limit, the secant M 0 M, will
occupy various positions M 0 M;,M M:, and so on.
If, in the limitles~ approach of the point M, (along the curve)
to the point M 0 from either side, the secant tends to occupy the
position of a definite straight line M 0 T, this line is called the
tangent to the curve at the point M 0 (the concept "tends to occupy"
will be explained later on).
Let us consider the function f(x) and the corresponding curve
Y= f (x)
in a rectangular coordinate system (Fig. 58). At a certain value of
x the function has the value y = t (x). Corresponding to these values
of x and y on the curve we have the point M 0 (x, y). Let us increase
X X+!JX X
Fig. 58.
the argument x by ().x. Corresponding to the new value of the

argument, x + ().x, we have an increased value of the function,
+ + /).)-).
y ().y = f (x Another corresponding point on the curve will
+ +
be M 1 (x ().x, y ().y). Draw the secant M 0 M, and denote by cp the
angle formed by the secant and the positive direction of thex-axis.
Form the ratio M. From Fig. 58 it follows immediately that
6,X
(l)
Now if !lx approaches zero, the point M 1 will move along the
curve always approaching M 0 The secant M 0 M 1 will turn about
M 0 and the angle cp will change in !lx. If as !lx--+ 0 the angle cp
approaches a certain limit a, the straight
line passing through M 0 and forming an
angle a with the positive direction of
the abscissa axis will be the sought-for
Iine tangent. It is easy to find its slope:
tan a= lim tan cp =lim t::..y = f' (x).
, liX .... o llx .... o t::..x
Hence,
t' (x) =tan a, (2)
Fig. 59. which means that the values of the

derivative f' (x), for a given value of
the argument x, is equal to the tangent of the angle formed with
the positive direction of the x-axis by the line tangent to the graph
of the function f (x) at the corresponding point M 0 (x, y).
Example. Find the tangents of the angles of inclination of the line tangent
to the curve y=x 2 at the points M1 ( ~, !\ M2 (-1,1) (Fig. 59).
Solution. On the basis of Example l, Sec . .J.. we have y' = 2x; hence,
tan a 1 =y' I x c:
1 =I;
2
tan a 2 =y' Ix:::-J
= -2.
SEC. 4. DIFFERENTIABILITY OF FUNCTIONS
Definition. If the function

y=f(x) (I)
has a derivative at the point x = X0 , that is, if there exists

-
1lm t::..y 1"
= lm
f(xo+t::..x)-{(x 0 )
(2)
1u .... o t::..x 1u .... o t::..x
we say that for the given value x = X 0 the function is differentiable

or (which is the same thing) has a derivative.
If a function is differentiable at every point of some interval
Ia, b] or (a, b), we say that it is differentiable over the interval.
Theorem. If a function y = f (x) is differentiable at some point
x = x 0 , it is continuous at this point.
Viflerentlability of- Function~ 75
Indeed, if
then
~~ =f' (X +y, 0)
where y is a quantity that approaches zero as fl.x-+ 0. But then

fl.y = f' (x0 ) fl.x yfl.x; +
whence it follows that fl.y-+ 0 as fl.x _.. 0; and this means that the
function f (x) is continuous at the point X0 (see Sec. 9, Ch. II).
In other words, a function cannot have a derivative at points
of discontinuity. The converse is not true; from the fact that at
some point x=x 0 the function y=f(x) is continuous, it does not
yet follow that it is differentiable at this point: the function f (x)
may not have a derivative at ihe point X 0 To convince ourselves
of this, let us examine several cases.
Example I. A function f (x) is defined in an interval [0, 2) as follows (see
Fig. 60):
f(x)=x when Oo;;;;;xoe:;;;l,
f(x)=2x-l when I <xo;;;;;2.
At x= I this function has no derivative, although it is continuous at this point.
Indeed, when ax> 0 we have
lim f(l+ax)-{(1)= lim [2(1+11x)-IJ-[21-I)= lim 2ax= 2
lix -+ 0 /),X /iX -+ 0 ax lix -+ 0 AX
when ax < 0 we get
lim /(1 +ax)-{(!)= lim [I +axi-l lim f),x = l.
Ax-+ o AX Ax-+ o AX /i;t-+ o /),X
Thus, this limit depends on the sign of ax. and this means that the function
has no derivative") at the point x =I. Geometrically, this is in accord with
the fact that at the pointx=l the given"curve" does not have a definite lin~
tangent.
Now the continuity of the function at the point x= I follows from the
fact that
when ax< 0,
when ax> 0,
and, therefore, in both cases ay ..... 0 as ax ..... 0.
j The definition of a derivative requires that the ratio /lt/ should (as
. ax
/),x ..... 0) approach one and the same limit regardless of the way in which ax
approaches zero.
Examplc 2. A lunct ion y = V-

x , the graph ol which is shown in Fig. 61,
is defined and continuous lor all values of the independent variable.
Let us try to find out whether this function has a derivative at x=O; to
do this, we find the values of the !unction at x=O and at x=D+8x: at
x=O we have y=O, at X=0+8x we have Y+ l\y= v(l\x).
Fig. 60. Fig. 61.
Therefore,
M= V -(8x).
Find the limit of the ratio of the increment of the function to the incre-
ment of the argument:
. 8Y . V-
(8x) . I
hm - = lim - - - = l1m ---=+oo.
dx.-.ol\X 6x.-.o l\X dx.-.oV8x2
Thus, the ratio of the increment of the function to the increment of the argument
at the point x=O approaches infinity as l\X -..0 (hence there is no limit). Consequ-
ently, this function is not differentiable at the point x= 0. The line tangent to the
curve at .this point forms, with the x-axis, an angle ; , which means that it
coincides with the y-axis.
SEC. 5. FINDING THE DERIVATIVES OF ELEMENTARY FUNCTIONS.

THE DERIVATIVE OF THE FUNCTION y=xn, WHERE n IS POSITIVE
AND INTEGRAL
To find the derivative of a given function y = f (x), it is neces-

sary to carry out the following operations (on the basis of the
general definition of a derivative):
l) increase the argument x by !lx, calculate the increased value
of the function:
+
y !ly = f (x !lx); +
Finding the Deriuafiues of Elementary Functions 77
2) find the corresponding increment of the function:

!1y = f (x !1x)- f (x); +
3) form the ratio of the increment of the function to the
increment of the argument:
M f<x+~x)-f(x)
~x= ~x ;
4) find the limit of this ratio as !1x-+ 0:

, . ~Y . f(x+~x)-f(x)
y = 11m -= 11m
fix -+ o ~x fix -+ o ~x
Here and in the following sections, we shall apply this general

method for evaluating the derivatives of certain elementary
functions.
Theorem. The derivative of the function y=x", where n is a
positive integer, is equal to nx"- 1; that is,
if y=x", then y'=nx"- 1 (I)
Proof. We have the function
y=x".
1) If x receives an increment !1x, then
Y+ !1y= (x+ !1x)".
2) Applying N ewton's binomial formula, we find
n
!1y = (x + !1x)"-x"= x" + T x"- 111x+
+ n <~~I) x"- 2
(/1x) 2 + ... + (~xt-x"
or
!1y = nxn-1 !1x + n ~n.;-1) xn- ~ (!:l.x)~ + ... + (!:l.x)".
3) We find the ratio
~~ = nx"-1 + n (7~1) x"-2 !1x + ... + (!:l.x)"-1.
4) Then we find the limit of this ratio
y' = I'1m-=
M
fix-+ o ~X
= lim [nx"- 1 + n <~.-;I) xn-l !1x + ... + (~x)"- 1 ] = nx"- 1 ,

fix -+ o
consequently, y' =nx"- 1 , and thus we have proved the theorem.
Example 1. y=x 5 , y' =5x 5 - 1 =5x4

Example 2. y=x, y'=lx 1 - 1 , y'=l. The latter result has a simple geo-
metric interpretation: the line tangent to the straight line y =X for any value
of x coincides with this line and, consequently, forms with the positive
direction of the x-axis an angle, the tangent of which is I.
Note that formula (1) also holds true when n is fractional or
negative. (This will be proved in Sec. 12).
Example 3. y= Yx.
Let us represent the function in the form of a power:
I
y=x 2;
then by formula (1), taking into consideration what we have just said, we get
I
l --1
y'=-X2
2
or
I
Example 4. y = .r- .
X Y X
Represent y in the form of a power function:
3
y=x -2.
Then
3 _2__. 3 -~ 3
y'= -2x 2 = -2x 2 =-2x2Vx'
SEC. 6. DERIVATIVES OF THE FUNCTIONS _v=sinx .v=cosx
Theorem 1. The derivative of sin x is cos x, or

if y= sin X, then y' =COSX. (II)
Proof. Increase the argument x by the increment Llx; then
1) y+Lly= sin (x+Llx);
2) Lly =sin (x+ Llx)-sin X= 2 sin x+ Ll;-x cos x+Ll;+x =
. !lx -cos ( x+-y
=2 sm !lx) ;
2
Derivatives of the Functions y =sin x; y = cos x 79
. tlX
4) y' = lim llY = lim smT - - . lim cos ( x + 1'12x ) ,
lix-+ o t'lX lix-+ o llX 11x -+ o
2
but since
sin tlx
2
lim - - =1,
/1X-+ o I'! X
2
we get
y, = 1.tm COS ( x+ f'lx)

2 =COSX,
lix-+ o
This latter equality is obtained on the grounds that cos x is a

continuous function.
Theorem 2. The derivative of cos x is -sin x, or
if y=cosx, then y' = -sinx. (II I)
Proof. Increase the argument x by the increment Ax, then

y + Ay = cos (x + Ax);
Ay= cos (x+Ax)-cosx=-2sin x+f'l;-x sin x+f'l;+x =
= - 2 Sin ~X Sin ( X+~X);
t'lX
llY sin 2 . ( tlx)
-=---SID
f'lX f'lX
x+-"
2 '
2
sin t'lx
y' = lim M = - lim
lix - o t'lX lix -+ o
T2 sin(x + !12 x) = - lim
11x - o
sin (x+ ~x);
2
taking into account the fact that sin x is a continuous function,

we finally get
y' = -sinx.
80 Derivative a_nd Differential
SEC. 7. DERIVATIVES OF: A CONSTANT, THE PRODUCT OF A CONSTANT

BY A FUNCTION. A SUM, A PRODUCT, AND A QUOTIENT
Theorem 1. The derivative of a constant is equal to zero: that is,

if y = C, where C = const, then y' = 0. (IV)
Proof. y= C is a function of x such that the values of it are
equal to G for all x.
Hence, for any value of x
y=f(x)=C.
We increase the argument x by an increment f!.x (f!.x =I= 0). Since
the function y retains the value C for all values of the argument,
we have
Y+ lly=f (x+ llx) =C.
Therefore, the increment of the function is
lly= f (x+ llx)-f (x) = 0,
the ratio of the increment of the function to the increment of the
argument
and, consequently,
y = I"tm -
I M= 0 ,
/).X-+- o.iX
that is,
y'=O.
The latter result has a simple geometric interpretation. The
graph of the function y=C is a straight line parallel to the x-axis.
Obviously, the line tangent to the graph at any one of its points
coincides with this straight line and, therefore, forms with the
x-axis an angle whose tangent y' is zero.
Theorem 2. A constant factor may be taken outside the deriva-
tive sign, i.e.,
if y = Cu (x) (C = const), then y' = Cu' (x). (V)
Proof. Reasoning as in the proof of the preceding theorem, we
have
y=Cu (x);
Y+ lly=Cu(x+ llx);
lly= Cu (x+ llx)-Cu (x) = C [u (x+ llx)-u (x)],
Derivatives of: A Constant, the Product of a Constant by a Function 81
f>.y =Cu (x+~x)--u (x)

t,.x f>.x '
, . f>.y C .
y = 11 m - = 11m
u (x+~x)-u (x)
, i. e. y' = Cu' (x).
l!.x ..... o t,.x l!.x ..... o t,.x
I
Example 1. y = 3 Vx .
y'-3 (
1 )'
Yx =3 ( x -~)' ( I ) -~-~
= =3 -2 x 2
3 _...!._
=-2x 2,
or
, 3
y----
- zxv:x
Theorem 3. The derivative of the sum of a finite number of diffe-
rentiable functions is equal to the corresponding sum of the
derivatives of these functions.*)
For the case of three terms, for example, we have
y=u(x)+v(x)+w(x); y'=u'(x)+v'(x)+w'(x). (VI)
Proof. For the values of the argument x
y=u+v+w
(for the sake. of brevity we drop the argument x in denoting the
function).
For the value of the argument x+ ~x we have
y+ ~y= (u+~u)+ (v +~v) + (w+~w),
where ~y. ~u. ~v. and ~w are increments of the functions y, u,
v and w, which correspond to the increment ~x in the argument
x. Hence,
~v
~y=~U-t-Llv+~w , -
f>.x
A
= t,.x
-
f>.y
+ -+-
f>.u
!J.x
f>.w
t,.x '
y' = lim t,.y = lim. t,.u + lim t,.v + lim ~~

l!.x -+ o t,.x 1!1 <-+ o ~X l!.x-+ o f>.x l!.x ..... o t,.x
or
y' = u' (x) + v' (x) + w'(x).
Example 2 y =3x -
4
V1x ,
*)The expression y=u(x)-v(x) is equivalent.to y='=u(x)+(-l)v(x)

and y' = [u (x) + (-I) v (x)j' = u' (x) + [1- v (x)j' = u' (x)- v' (x).
x- - I'
1 )' 1 1
--
y'=3(x 4) ' - ( =34x-( -
3 )x 1
,
and so
y I = 12X '+ 3I I
J/-
Xy X
Theorem 4. The derivative of a product of two dif}erentiable

functions is equal to the product of the derivative of the first fun-
ction by the second function plus the product of the first function
by the derivative of the second function; that is,
if y=uv, then y'=u'v+uv'. (VII)
Proof. R.easoning as in the proof of the preceding theorem, we
get
y=uv,
y + IJ.y= (u + /J.u) (v + /J.v),
!J.y = (u + /J.u) (v + /J.v) -uv = tluv + u!J.v.+ /J.u !J.v,
f).y
/).X
= /).U V
/).X
+ U /).V + /J.u /).V
/).X !).X'
y' = lim M = lim tJ.u v + lim u tJ.u + lim IJ.u tJ.v =

11x-. o /).X 11x-+ o /).X 11x-+ o /).X 11x ..... o /).X
= ( lim
11x -+0 /).X
tJ.u) v +u lim tJ.u + lim flu lim tJ.V
I!..X-+0 /).X l!..x-+0 /1X-+0 /).X
(since tt and v are independent of /J.x).

Let us consider the last term Ol=l the right-hand side:
lim IJ.u lim tJ.v
l!.x -+ o l!.x -+ o AX
Since u (x) is a differentiable function, it is continuous.

Consequently, lim /J.u = 0. Also,
Ax-+ o
/).V 1
I1m - = V =f=oo.
l!.x ... o /).X
Thus, the term under consideration is zero and we finally get

y' = u' v + uv'.
The theorem just proved readily gives us the rule for differentiating
the product of any number of functions.
Thus, if we have a product of three functions
y= uvw,
Derivatives of: A Constant, the Product of a Constant by a Function .83
then by representing the right-hand side as the product of u and

(vw), we get y' = u' (vw) + u (vw)' = u'vw+ u (v'w+ vw') = u'vw+
+uv'w+uvw'.
In this way we can obtain a similar formula for the derivative
of the product of any (finite) number of functions. Namely, if
y = u,u ... un, then
y' = u;u ... Un-IUn + U 1 U; , .. Un_IUn +, .. + U1 U 0 , , , Un_ 1 U~.
Example 3. If y = x 2 sin x, then

y' = (x 2 )' sin x + x 2 (sin x)' = 2x sin x + x cos x.
Example 4. If y = Yx sin x cos x, then

y' =(Yx)' sinxcosx+ Vx (sinx)' cos x+ Vx sinx (cosx)' =
= ~r- sin x cos x + Vx cos x cos x + Vx sin x (- sin x) =

2 r x
I .r- sin 2x .r-
= .r- sin x cos x+ r x (cos x-sin 2 x) = .r- + r x cos 2x.
2rx 4yx
Theorem 5. The derivative of a fraction (that is, the quotient

obtained by the division of two functions) is equal to a fraction
whose denominator is the square of the denominator of the given
fraction, and the numerator is the difference between the product of
the denominator by the derivative of thl' numerator, and the pro-
duct of the numerator by the deriv11tive of the denominator; i. e.,
if y = vu ' then y'-
- u'u-uv'
v (VIII)
Proof. If 11y, 11u, and 11v are increments of the functions y, u,

and v, corresponding to the increment 11x of the argument x. then
/). _ u+~u
Y + Y- v+~v '
/). = u+~u _.!!_= v ~u-u ~v
Y v+~v v v(v+~u) '
v ~u-u ~v ~u ~v
-'U-U-
~y ~X ~X ~X
-=
M u(v+~v) v (v+~v) '
v li~ ~u- u lim ~ v
ax .... o~X ax .... o~X
v lim (v
ax-o
~v) +
Whence, noting that 11 v - 0 as 11x --+-o, *) we get

u'u-uv'
y'= v'
x
Example 5. If y = - - , then
cosx
y'
(x 3 )' cos x-x (cos x)' 3x 2 cos x+x sinx
cos 2 x cos 2 x
Note. If we have a function of the form

u (x)
y=-c-,
where the denominator c is a constant, then when differentiating
this function we do not need to use formula (VI II); it is better
to make use of formula (V):
1 1
1 ) 1 U
Y= -;u
I (
=-;U=--c I
Of course, the same result is obtained if formula (VIII) is applied.

COS X
Example 6. If y = -- , then
7
, (cos x)' sin x
y =-7-=--7-.
SEC. 8. THE DERIVATIVE OF A LOGARITHMIC FUNCTION
Theorem. The derivative of the function loga x is _.!_

X
log 4 e, that is,
if y = log 4 x, then 1
lj = ~ log 4 e. (I X)
Proof. If 11y is an increment of the function y = log 4 x that

corresponds to the increment 11x of the argument x, then
y + 11y = loga (x + 11x);
11y = loga (x + 11x) -log 4
X= log 4 x \ flx loga (I + tlxx) ;
M =-
--=-
flX
I log ( I
flX a
+ -tlx)
X
.
*) lim tlv=O since u (x) is a differentiable and, consequently, continuous
l\.X-+0
function.
Derivative of Composite Function 85
Multiply and divide by x the expression on the right-hand side of

the latter equality:
X
l'lY =_!_.!_log ( 1
/).X X /).X a
+ llx) =_!_log a ( 1 + M)t.x.
X X X
We denote the quantity t:.x

X
in terms of ex.. Obviously, for the
given x, cx.-0 as dx-0. Consequently,
' I
l'lY
/).X
=~log
X a
( 1 +ex. fa
But, as we know from Sec. 7, Ch. II,
1
lim (1 +cx.)a- =e.

U-+0
But if the expression under the sign of the logarithm approaches

the number e, then the logarithm of this expression approaches
loga e (in virtue of the continuity of the logarithmic function).
We therefore finally get
1
y' = lim M =lim _!_ loga (1 + cx.)a = _!_ loga e.

l!.x -+ 0 /).X a -+ 0 X X
1
Noting that loga e = In a , we can rewrite the formula as follows:
I 1 I
y =-xTrili.
The following is an important particular case of this formula:
ii a=e, then lna=lne= 1; that is,
if y= lnx, then y' =..!...
X
(X)
SEC. 9. THE DERIVATIVE OF A COMPOSITE FUNCTION
Given a composite function y= f(x), that is, such that it may

be represented in the following form:
y=F(u), u=cp(x)_
or y = F [cp (x)] (see Ch. I, Sec. 8). In the expression y = F (u),
u is called the intermediate argument.
Let us establish a rule for differentiating composite functions.
Theorem. If a function u = cp (x) has, at some point x. a deriva-
tive u~= cp' (x), and the function y = F (u) has, at the corresponding
value of u, the derivative y;, = F' (u), then the composite function
y=F [<p(x)] at the given point x also has a derivative, which is
equal to
y: = F~ (u) <p' (x),
where in place of u we must substitute the expressi01t u=<p(x).
Briefly,
,
Yx= YuUx.
In other words, the derivative of a composite function is equal to
the product of the derivative of the given function with respect to
the intermediate argument u by the derivative of the intermediate
argument with respect to x.
Proof. For a definite value of x we will have
U=<p(x), y=F(u).
For the increased value of the argument x + !1x,
u + du = 'ljl (x + !1x), y + D..y = F (u -1- !1u).
Thus, to the increment !1x there corresponds an increment !1u,
to which corresponds an increment !1y, whereby /).u ---.... 0 and
!1y --o as !1x---.... 0. It is given that
. tlY ,
hm - = Yu
Llx~o /),.U
From this relation (taking advantage of the definition of a limit.)

we get (for [).u +0)
tJ.y ,
tJ.u = Yu+ a, (l)
where a---+0 as [).u -0. We rewrite (1) as

11y= y~du +a du. (2)
Equality (2) also holds true when du = 0 for an arbitrary a,
since it turns into an identity, 0=0. For du=O we shall assume
a= 0. Divide all terms of (2) by !1x:
(3)
It is given that

ltm. /),.U= Ux,
-
1
J"1m CG= 0
Llx~o /),.X .:lx-+0
Derivative of .Composite Function 87
Passing to the limit as ~x--+ 0 in (3), we get

(4)
which is the required proof.
' Represent the given
Example I. Given a function y =sin (x 2 )o Find Yxo
function as a function of a function as follows:
y=slnu, u=x 2 o
We find
y~=COSU, u:=2x.
Hence, by formula (4),
y~ = y~u~ =cos u 2xo 0
Substituting, in place of u, its expression, we finally get

y~ = 2x cos (x 2 )o
Example 2o Given the function y =(In x) 1 Find Yx o
Represent this function as follows:

y=u 1 , u=lnxo
We find
Hence,
, I 0
I
yX = 3u 1 -=3
X
(In x)'-.
X
II a function y = f (x) is such that it may be represented in the form
y=f(u), U=!p(V), V='i'(x),
the derivative y~ is found by a successive application ol the foregoing theorem.
Applying the proved rule, we have
Yx =yuux
Applying the same theorem to find u~, we have
' I
ux=uvvx.
Sutstituting the expression of u~ into the preceding equality, we get
(5)
or
y~= F~ (u) 'P'v(V) 'I'~ (x).
Example 3. Given the function y =sin [(In x) 1 ). Find Yx Represent the
!unction as follows:
y=slnu, u=v, v=lnx.
88 Deriuatiue and Differential
We then find
, I
!)
x
=-.
X
In this way, by formula (5), we get
, , ., ' I
Yx =YuUvVx =3 (cos u) v2 x,
or finally,
It is to be noted that. the function considered is defined only for x > 0.
SEC. 10. DERIVATIVES OF THE FUNCTIONS y=tanx,

y=cot x, y=ln 1xI
1
Theorem 1. The derivative of the function tanx is - -,
COS 2 X
1
or if y = tan x, then y' = -COS 2-X (XI)
Proof. Since
slnx
Y= cosx '
by the rule of differentiation of a fraction [see formula (VIII),
Sec. 7, Ch. III] we get
, (sin x)' cos x-sinx (cos x)' _cos x cos x-sin x (- sinx) _
Y = cos 2 x - cos 2 x -
+
cos x sin 2 x
cos 2 x = cos 2 x
Theorem 2. The derivative of the function cot x is

1 1
- - .-2 - , or
sm x
if y =cot x, then y' = - -.--.
sm 2 x
(XII)
. cosx
Proof. Stnce y= -.-,
smx
we have
, (cos x)' sin x- cos x (sin x)' -sin x sin x- cos x cos ~
Y = sin 2 x = sin 2 x =
+
sin 2 x cos 2 x
= - sin 2 x = - sin 2 x
Example 1. If y =tan Yx. then
y'= I (Jlx)' =~ I .
cos Yx 2 Y x cos Y x 2
An Implicit Function and Its Dif!~rentiation 89
Example 2. If y=lncotx, then

_l_- (cotx)'=-1-
Y, -_ cot (--~-)=- I __2_
x cot x 2 sin x cos x sin x sin 2x
Theorem 3. The derivative of the function In I x 1 (Fig. 62) is ..!._,

X
or if y=lnlxl, then y'=.!....

X
(XIII)
y
Fig. 62.
Proof. a) If x > 0, then I xI= x, In I xI= In x, and therefore

, 1
y =-x
b) Let x<:o, then lxl=-x. But
lnlxl=ln(-x).
(It will be noted that if x<O, then -x>O:) Let us represent
the function y= In (-x) as a composite function by putting
'y lnu; U=-x.
Then
I
Yx=Y.,Ux=-(-1}=
u -1( - 1 ) = -x1.
-x
And so for negative values of x we also have the equation
1
Yx=-;
Hence, formula (X III) has been proved for any value x+O.
(For x = 0 the function In I xI is not defined.)
SEC. II. AN IMPLICIT FUNCTION AND ITS DIFFERENTIATION

LE't the values of two variables x and y be related by some
equation, which we can symbolise as follows:
F(x, y)=O. (l)
90 Derivative and DiOerent ial
If the function y = f (x), defined on some interval (a, b), is

such that equation (1) becomes an identity in x when the expres~
y y
Fig. 63. Fig. 64.
sion f(x) is substituted into it in place of y, the function y=f(x)

is an implicit function defined by equation (1).
For example, the equation
x"+y"-a 2 =0 (2)
defines implicitly the following elementary functions (Figs. 63
and 64):
y= Yaa-x, (3)
(4)
Indeed, substitution into equation (2) yields the identity
x +(a"-x")-a"= 0.
Expressions (3) and (4) were obtained by solving equation (2)
for y. But not every implicitly defined function may be represented
explicitly, that is, in the form y = f (x), *) where f (x) is an ele-
mentary function.
For instance, functions defined by the equations
y 0 -y-x=o
or
y-x- 41 smy=
. 0
are not expressible in terms of elementary functions; that is, these

equations cannot be solved for y by means of elementary functions.
Note 1. Observe that the terms "explicit function" and "implicit
function" do not characterise the nature of the function but merely
the way it is defined. Every explicit function y="f(x) may also
be represented as an implicit function y-f(x)=O.
*) If a function is defined by an equation of the form y =f (x), one says
that the function is defined explicitly or is explicit,
Derivative.~ of a Power Function for an Arbitrary Real Exponent 91
We shall now give the rule for finding the derivative of an

implicit function without transforming it into an explicit one,
that is, without representing it in the form y=f(x).
Assume the function is defined by the equation
x2 + y2-a2 = 0.
Here, if y is a function of x defined by this equality, then the
equality is an identity.
Differentiating both sides of this identity with respect to x, and
regarding y as a function of x, we get (via the rule of differentiat-
ing a composite function)
1
2x+2yy =0,
whence
I X
y =-y
.Observe that if we were to differentiate the corresponding explicit
function
Y= Va2-x2,
we would obtain
, X
y=- =--XVa2-x y '
which is the same result.
Let us consider another case of an implicit function y of x:
y-y-x 2 =0.
Differentiate with respect to x:
1 1
6y'y -y -2x= 0,
whence
1 2x
y = 6y 5 - l .
Note 2. From the foregoing examples it follows that to find the

value of the derivative of an implicit function for a given value
of the argument x, one also has to know the value of the function y
for a given value of x.
SEC. 12. DERIVATIVES OF A POWER FUNCTION FOR AN ARBITRARY
REAL EXPONENT, OF AN EXPONENTIAL FUNCTION,
AND A COMPOSITE EXPONENTIAL FUNCTION
Theorem 1. The derivative of the function xn, where n is any
real number, is equal to nxn- 1 ; that is,
(I I)
92 Derivative and DitJerential
Proof. Let x > 0. Taking logarithms of this function, we get

iny=ninx.
Differentiate, with respect to x, both sides of the equality obtained,
taking y to be a function of x:
y' 1
-=n-
y X '
Substituting into this equation the value y = xn, we finally get

y'=nxn-l.
It is easy to show that this formula holds true also for x < 0
provided xn is meaningful. *)
Theorem 2. The derivative of the function a"', where a> 0, is
a"' In a; that is,
if y=a"', then y'=a"'lna. (XIV)
Proof. Taking logarithms of the equality y =a"', we get
lny=xina.
Differentiate the equality obtained regarding y as a function of x:
1
-y y' = In a; y' =yIn a
or
y' =a"' In a.
If the base is a=e, then lne= 1 and we have the formula
y=e"', y' =e"'. (XIV')
Example 1. Given the function
IJ=e"''.
Represent it as a composite function by introducing the intermediate argument u:
y=e", u =x 2;
then
and, therefore,
*) This formula was proved in Sec. 5, Ch. III, for the case when n is a
positive integer. Formula (I) has now been proved for the general case (for
any constant number n).
Derivatives of a Power Function for an Arbitrary Real Exponent 93
A composite exponential function is a function in which both

the base and the exponent are functions of x, for instance, (sin x)x 2 ,
xtanx, r, (In x)X, and the !ike; generally, any function of the form
y = [u (x) t' <x) =uTI
is an exponential function (composite exponential function). *)
Theorem 3.
If y=uv. then y' =VUTI-tu' +uTiv' !nu. (XV)
Proof. Taking logarithms of the function y, we have
lny= v lnu.
Differentiating the resultant equation with respect to x, we get
J...y
y
= v J..u u' + v' In u
whence
y' = y ( v u~ + v' In u) .
Substituting into this equation the expression y= uTI, we obtain
y' = vuTI-lu' uTiv' In u.+
Thus, the derivative of an exponential function (composite expo-
nential function) consists of two terms: the first term is obtained
by assuming, when differentiating, that u is a function of x and v
is a constant (that is to say, if we regard uTI as a power function);
the second term is obtained on the assumption that v is a function
of x, and u = const (i. e., if we regard uTI as an exponential
function).
Example 2. If y=xX, then y'=xxx- 1 (x')+x~(x')lnx
or y' =Xx +xx In x =xx (I+ In x).
Example 3. If y =(sin x)x, then

y' = x 2 (sin x)x - 1 (sin x)' +(sin x)x (x 2 )' In sin x =
=x 2 (sin x)x - 1 cos x +(sin x)x 2x In sin x.
The procedure applied in this section for finding derivatives

(first finding the derivative of the logarithm of the given function)
is widely used in differentiating functions. Very often the use of
this method greatly simplifies calculations.
*) In the Russian mathematical literature this function is also called an

exponential-power function or a power-exponential function.
94 Der ivai ive and Differential
Example 4. To find the derivative of the function

(x+l)" Yx=i
Y (x+4)'e"
Solution. Taking logarithms we get
1
In y=21n (x+ 1)+ 1n (x-1)--31n (x+4)-x.
2
Differentiate both sides of this equality:
~ 2 I 3
-y= x+ 1+2(x-l) -x+4-l.
Multiplying by y and substituting, in place of y, the expression
(x+l) 2 VX-T
(x + 4) ~: , we get
,_(x+1)' Yx=i [-2- __1_ _ _3_ _ ]

1
Y- (x+4)"e" . x+ I +2 (x-1) x+4
Note. The expression-~=(\ny)',

y
which is the derivative, with
respect to x, of the natural logarithm of the given function
y = y (x), is called the logarithmic derivative.
SEC. 13. AN INVERSE FUNCTION AND ITS DIFFERENTIATION
Take an increasing or decreasing function (Fig. 65)

y= f (x) (1)
defined in some interval (a, b) (a< b) (see Sec. 6, Ch. 1). Let
f (a)= c, f (b)= d. For definiteness we shall henceforward consider
an increasing function.
Yt Let us consider two different values x,
and x. in the interval. (a, b). From the
definition of an increasing function it
follows that if x, < x, and y, = f (x,),
y1 = f (x.), then y, < y1 Hence, to two
a x2 b x different values x, and x. there correspond
two different values of the function, y,
Fig. 65.
and y.. The converse is also true: if
Y 1 < Ya Y1 = f (x,), and Y. = f (x 1 ), then
from the definition of an increasing function it follows that x, < x .
Thus, a one-to-one correspondence is established between the values
of x and the corresponding values of y.
Regarding these values of y as values of the argument and the
values of x as values of the function, we get x as a function of y:
~=,~~ w
An /nuerse Function and Its Differentiation 95
This function is called the inverse function of y = f (x). It is obvi-

ous too that the function y = f (x) is the inverse of x = cp (y). With
similar reasoning it is possible to prove that a decreasing function
also has an inverse.
Note 1. We state, without proof, that if an increasing (or de-
creasing) function y = f (x) is continuous on the interval [a, b]. where
f (a)= c, f (b)= d, then the inverse function is defined and is
continuous on the interval (c, d).
Example t. Given the function y =x 8 This function is increasing on the
infinite interval - oo < x < oo; it has an inverse function x = Vu
.(Fig. 66).
It will be noted that the inverse function x = q> {y) is found by solving the
equation Y=f(x) for X.
y=xJ
x=W
Fig. 66. Fig. 67.
Example 2. Given the fur.ction y=e". This function is increasing on the

infinite interval - oo < x < oo. It has an inverse x =In II The domain of
definition of the inverse function is 0 < y < oo (Fig 67).
Note 2. If the function y = f (x) is neither increasing nor decreas-

ing on a certain interval, it can have several inverse functions.*)
Example 3. The function y =x is defined on an infinite interval
-oo<x<+oo.It is neither increasing nor decreasing and does not have
an inV('rse function. If we consider the interval 0,..;;;; x < oo, then the function
here. is increasing and x = Jfyis its inverse. But in the interval - oo < x < 0
the function is decreasing and its inverse is x = - Jfy {Fig. 68).
Note 3. If the functions y = f (x) and x = cp (y) are reciprocal,

thejr .graphs are represented by a single curve. But if we again
*) Let it be noted once again that when speaking of y as a function of x we

have in mind that y is a siugle-valued function of x.
denote the argument of the inverse function by x, and the function

by y and then construct them in a single coordinate system, we
y will get two different graphs.
It will readily be seen that the graphs
will be symmetric about the bisector of
the first quadrantal angle.
Example 4. Fig. 67 gives the graphs of the

function y=ex(orx=lny) and its inverse
y=lnx, which are considered in Example 2.
Let us now prove a theorem that per-

mits finding the derivative of a function
Fig. 68. y=f(x) if we know the derivative of
the inverse function.
Theorem. If for the function
Y=f(x) (1)
there exists an inverse function
X=cp (y) (2)
which at the point under consideration y has a nonzero derivative
cp' (y), then at the corresponding point x the function y = f (x) has
a derivative f' (x) equal to cp' \y); that is, the following formula
is true
f' (x) = cp' l(y) . (XVI)
Thus, the derivative of one of two reciprocal functions is equal to
unity divided by the derivative of the second function for corre-
sponding values of x and y. *)
Proof. Differentiate, with respect to x, both sides of equality (2),
taking y as a function of x **):
I ( I
l=<p Y)Yx,
*) When we write f' (x) or y: we regard x as the independent variable

when evaluating the derivative; but when We write cp' (y) or X~ We assume
that y is the independent variable when evaluating the derivative. It should
be noted that after dift'erentiating with respect toy, as indicated on the right
side of formula (XVI), f (x) must be substituted for y.
**) Actually, here we find the derivative of a function of x defined
implicitly by the equation
X-<p(Y)=O
An Inverse Function and Its Differentiation 97
whence
I I
Yx = cp' (y)
Noting that y: = f' (x) we get formula (XVI), which may also be
written as y
I I
Yx= -;.
xu
The result obtained is clearly illustrated
geometrically. Consider the graph of the A t---M
function y = f (x) (Fig. 69). This curve will
also be the graph of the function x = IJl (y),
where x is now regarded as the function
and y as the independent variable. Take some X
point M (x, y) on this curve. Draw a tangent
to the curve at this point. Denote by a and~ Fig. 69.
the angles formed by the given tangent and
the positive directions of the x- and y-axes. On the basis of the
results of Sec. 3 concerning the geometrical meaning of a derivative
we have
f' (x) =tan a, }
(3)
IP' (y) =tan~
From Fig. 69 it follows directly that if a< ~ , then

l't
~=2-a.
l't 3l't
But if a> 2 , then, as is readily seen, ~ =
2 -a. Hence, in
any case
tan~= cot a,
whence
tan a tan ~ = tan a cot a = 1,
or
I
tana=tan ~
Substituting the expressions for tan a and tan~ from formula (3),
we get
[' (x) = IPI l(y)
4-3388
91\ Derivative and Differential
SEC. 14. INVERSE TRIGONOMETRIC FUNCTIONS

AND THEIR DIFFERENTIATION
I) The function y= arcsinx.

Let us consider the function
x=siny (1)
and construct its graph by directing they-axis vertically upwards
(Fig. 70). This function is defined in the infinite interval
- oo < y < + oo. Over the interval
Y - ~ ~ y ~ ~ , the function x = sin y
is increasing and its values fill the in-
terval - I ~ x ~ 1. For this reason, the
function x = sin y has an inverse which is
denoted by
-t f X
y = arc sin x. *)
7T i This function is defined on the inter-

~-~~2 1
val - I .s;;;;; x .s;;;;; 1, and its values fill the
interval - ~ .s;;;;; y,.;;; ~ In Fig. 70, the
graph of y =arc sin x is shown by the
heavy line.
Theorem 1. The derivative of the function arc sin x is equal to
l .
r
r 1-x'
; I. e.'
., .
ty=arcsmx, theny=,r I I
. (XVII)
r l-x
Proof. On the basis of (1) we have
x11' = cosy.
By the rule lor differentiating an inverse function,
' l I
Yx =--;=COSy
XII
but
cosy= VI- sin y= V1-x,
"lit may be noted that thefamiliarequatlony=arcsln.a:ofttigonoriletr'y

is another way of writing (I). Here (for a given x) y denotes the totality of
values of angles whose sine is equal tO' x. ,
Inverse Trigonometric Functions 99
therefore,
, I
Yx= Jfl-~;
the sign in front of the radical is plus because the function

n n
y =arc sin x takes on values in the interval - 2 ~ y ~ 2 , and.
consequently, cosy;;;;;..O.
Example 1. y=arcsine",
, l ( ")' e"
y = Jfl-(e") 2 e = Jfl-e 2"
Example 2.
y=(arcsin ~ y,
y'=2arcsin_!_
X
Yl I
1--
(.!...)'=-2arcsin_!_
X X X
VIx1 - l .
x
2) The function y = arc cos x.

As before, we consider the function
x= cosy (2)
and construct its graph with the y-axis extending
upwards (Fig. 71 ). This function is defined on
the infinite interval- oo < y < +
oo. On the
interval 0 ~ y ~ n, the function x =cosy is
decreasing and has an inverse that we denote -_~-~-r.--x
1 0 1
y= arccosx. XcCOSIJ
This function is defined on the interval Fig. 71.

- 1 ~ x < I. The values of the function fill
the interval n;;;;;..y;;;;;..O. In Fig. 71, the function y=arccosx is
depicted by the heavy line.
1
Theorem 2. The derivative of the function arc cos x is - ,r ;
r 1-x
i. e.,
1
if y =arc cos x, then y' = - V (XV!II)
1-x1
Proof. from (2) we have

x~=- siny.
4*
Hence
, I I I
Yx= x~=- siny = - Yl-cosy"
But cosy= x, and so
, I
Yx = - ,/
r 1-x

V
In sin y = l-cos y the radical is taken with the plus sign,
since the function y =arc cos x is defined on the interval 0..;;; y..;;; n
and, consequently, sin y;;;:;;;:: 0.
Example 3. y=arcros (tanx),
,
y =- y I
(t anx) , = - y .
1-tanx l-tan 2 xcos 2 x
3) The function y=arctanx.

We consider the function
______t~ ~------- X= tany (3)
and construct its graph (Fig. 72).
This function is defined for all
values of y except y = (2k l) ~ +
(k = 0, l, 2, ... ). On the
interval - ~ < y < ~ the function
x =tan y is increasing and has an
inverse:
Y= arctanx.
-------=+!( This function is defined on the

interval - oo < x < oo. The values
of the function fill the interval
Fig. 72.
- ~ <y< ~ .
In Fig. 72; the
graph of the function y=arctanx is shown as a heavy line.
Theorem 3. The derivative of the function arctanx is l~x;
i. e.,
ify=arctanx, theny'= 1 ~x {XI X)
Proof. From {3) we have
, I
X----
v- cos y
In verse Trigonometric Functions 101
Hence
I 1 2
Y.x =---,=cos y
xy
but
I I
cosz Y = sec2 y =I +tan 2 y;
since tan y=x, we get, finally,
I I
Y =I +x 2 '
Example 4. y =(arc tan x) 4 ,
y' = 4 (arc tan x) 1 (arc tan x)' = 4 (arc tan x)"
1
+I xz.
4) The fu net ion y = arc cotx.
Consider the function __......,.._______y ----------
21'(
X= cofy. (4)
This function is defined for all
values of y except Y=-kn(k=O, 1,
2). The graph of this function is
shown in Fig. 73. On the interval
0 < y < n, the function x=coty is
decreasing and has an inverse: ---==------;;t-----=----
Y= arc cot x.
Consequently, this function is de-
fined on the infinite interval - oo <
< x < oo, and its values fill the Fig. 73.
interval n > y > 0.
Theorem 4. The derivative of the function arccotx is- ~x 2 ;
i. e.,
1
if y =arc cot x, theny' = - 1 ~xz. (XX)
ProoL From (4) we have
I
X y = - -2-
si n y
Hence
z I I
Yx = - Sin y = - csc 2 11 = - I +cot 2y
But
coty=x.
Therefore
I I
Yx=-1 +x2"
SEC. 15. TABLE OF BASIC DIFFERENTJATION FORMULAS
Let us now bring together into a single table all the basic for-
mulas and rules of differentiation derived in the preceding sections.
Y= canst, y' = 0.
Power function:
particular instances:
y= Vx, y
- ,= I
.r-;
2 f X
1
Y=-
x'
Trigonometric functions:
y= sin x, y' =cos x,
y= cosx, y' =-sin x,
y= t anx, y ' =coszx

I
y=co t X, y ' =-sinzx

I
Inverse trigonometric functions:

y= arc sm x, y ' = -.r==,
1
r l-x 2
I I
y =arc cos x, y = - -y-.,..,=_=x=z,
1
y=arc t anx, y ' = I

l+x
. ' . 1
y=arccotx, y ~-l+x
Exponential function:
y=a~, y'=a~lna;
Parametric Representation of a Function 103
in particular,
y=e"', y' =e"'.
Logarithmic function:
y = I og 0 X, y ' = X-
1 I oga e;
in particular,
y= I nx, y ' =-.
1
X
General rules for differentiation:

y = Cu (x), y' = Cu' (x) (C = const),
y=u+v-w, y'=u'+v'-w',
y=UV, y' =u'v+uv',
, u'v-uv'
y = v
y=f(u), }
y~ = fu (u) q>~ (x),
U=q>(x),
y=UfJ,
If y=f(x), x=q>(y), where f and q> are reciprocal functions,

then
1
f' (x) = <p' (y), where y = f (x).
SEC. 16. PARAMETRIC REPRESENTATION Of A FUNCTION
Given two equations:

X=q>(t), }
(1)
Y='i'(t),
where t assumes values that lie in the interval [T" T 1 ]. To each
value of. t there correspond values of x and y (the functions q>
and 'lj1 are assumed to be single-valued). If one regards the values
of x and y as coordinates of a point in a coordin11te xy-plane,
then to each value of t there will correspond a definite point in
the plane. And when t var-ies from T 1 toT 1 , this point will de-
scribe a certain curve. Equations (1) are called parametric equations
of this curve, t is the parameter, and parametric is the way the
curve is represented by equations (l).
104 Deriuatlue and Differential
Let us further assume that the function x = cp (t) has an inverse,

t = <D (x). Then, obviously, y is a function of x;
Y='!J[<D(x)]. (2)
Thus, equations (1) define y as a function of x, and it is said
that the function y of x is represented parametrically.
The explicit expression of the dependence of y on x, y = f(x),
is obtained by eliminating the parameter t from equations (1).
Parametric representation of curves is widely used in mechanics.
If in the xy-plane there is a certain material point in motion and
if we know the laws of motion of the projections of this point
on the coordinate axes, then
y
X=cp(t), } ( 1')
Y='IJ(t)
where the parameter t is the time. Then
equations (1 ') are parametric equations of .
the trajectory of the moving point. Eli-
Yo minating from these equations the para-
meter t, we get the equation of the
trajectory in the form y = f (x) or
F (x, y) = 0. By way of illustration, let
0 X us take the following problem.
Fig. 74. Problem. Determine the trajectory and
point of impact of a load dropped from an
airplane moving horizontally with a velocity u0 at an altitude Yo (air
resistance is disregarded).
Solution. Taking a coordinate system as shown in Fig. 74, we assume
that the airplane drops the load at the instant it cuts the y-axis. It is
obvious that the horizontal translation of the load will be uniform and with
constant velocity u0 :
X=U 0 t.
Vertical displacement of the falling load due to the force of gravity will be
expressed by the formu Ia
gt2
S=-y
Hence the distance of the load from the ground at any instant will be
gt2
Y=Yo-y
The two equations

The Equations of Certain Curves in Paraml'tric Form 105
will be the parametric equations of the trajectory. To eliminate the parame-

ter t, we find the value t =!... from the first equation and substitute it into
Vo
the second equation. Then we get the equation of the trajectory in the form
g I
Y=Yo--x
2v:
This is the equation of a parabola with vertex at the point M (0, y 0 ), the
11-axis serving as the axis of symmetry of the parabola.
We determine the length of OC, denote the abscissa of C by X, and note
that the ordinate of this point is y=O. Putting these values into the
preceding formula, we get
whence
SEC. 17. THE EQUATIONS OF CERTAIN CURVES IN PARAMETRIC FORM

Circle. Given a circle with centre at the coordinate origin and with
radius r (Fig. 75).
Denote by t the angle formed by the x-axis and the radius to some point
M (.t, y) of the circle. Then the coordinates of any point on the circle will
be expressed in terms of the. parameter t as follows:
x= r cost,
y=r sin t,
}o..; t..; 2n.
These are the parametric equations of the circle. If we eliminate the para-
meter t from .these equations, we will have an equation of the circle contain-
ing only x and y. Squaring the parametric equations and adding, we get
x 2 + y 2 = r 2 (cos 2 t + sln 2 t)
or
xz+yz=,z.
Ellipse. Given the equation of the ellipse
xz yz
(ii'+/}i'=l. (l)
Set
x=a cost. (2')
Putting this expression into equation (1), we get
y=b sin t. (2")
The equations
x = a cos t, } 0 ..; t ..; 2n (2}
y= b slnt,
are the parametric equa lions of the ellipse.
106 Deriuatiue and DiUerential
Let us find out the geometrical meaning of the parameter t. Draw two
circles with centres at the coordinate origin and with radii a and b (Fig. 76).
Let the point M (x, y) lie on the ellipse, and let 8 be a point of the large
y y
Fig. 75. Fig. 76.
circle with the same abscissa as M. Denote by t the angle formed by the
radius 08 with the x-axis. From the figure it follows directly that
x=OP=acost [this is equation (2')],
CQ = b sin t.
From (2") we conclude that CQ = y; in other words, the straight line CM is
r-arallel to the x-axis.
Consequently, in equations (2) t is an angle formed by the radius 08 and
the axis of abscissas. The angle t is sometimes called an eccentric angle. ,
lj
Fig. 77.
Cycloid. The cycloid is a curve described by a point lying on the circum-

ference of a circle if this circle rolls upon a straight line without sliding
(Fig. 77). Suppose that when motion began the point M of the rolling circle
lay at the origin. Let us determine the coordinates of M after the circle has
turned through an angle t. If a is the radius of the rolling circle,. it will be
seen from Fig. 77 that
x=0P=08-P8,
but since the circle rolls without sliding, .we have
08=MB=at, P8=MK=aslnt.
Hence, x=at-a sin t=a (1-sln 1).
The Equations of Certain Curves in Parametric Form 107
Further,
y=MP=KB=CB-CK =a-a cos i=a (l-cos 1).
The equations
x=a (t-sin t), } 0 ,;;;;;; t .;;;;; 2n (3)
y=a (I-eos t),
are the parametric equations of the cycloid. As I varies between 0 and 2rt,
the point M will describe one arc of the cycloid.
Eliminating the parameter t from the latter equations, we get x as a
function of y directly. In the interval O.;;;;;t.;;;;;:rt, the function y=a(l-cost)
has an inverse:
t =arc cos a-y .
a
Substituting the expression for t into the first of equations (3), we get
x=aarccosa ay-asin(arccosa ay)

or
x=a arc cos a - y - V2ay- y 2 when 0 .;;;;;x .;;;;;na.
a
Examining the figure we note that when na.;;;;;x.;;;;;2na
x=2na-( a arc cosa a Y- Y2ay-y 2 ) .
It will be noted that the function

x=a (t-sin t)
has an inverse, but it is not expressible in terms of elementary functions.

And so the function y = f (x) is not expressible in terms of elementary
functions either.
Note 1. The cycloid clearly shows that in certain cases it is more con-
venient to use the parametric equations for studying functions and curves
'than the direcf relationship of y and x (y as a function of x or x as a
function of y).
Astroid. The astroid is a curve represented by the following parametric

equations:
x=a cos t, } 0 ..;;;; t <; 2n. (4)
y=asln 1 t,
Raising the terms of both equations to the power 2{3 and adding, we get
lOB DerivaUve and Differential
the relationship between x and y

2
or
(5)
A
Later on (Sec. 12, Ch. V) it will be shown
that this curve is of -the form shown in
Fig. 78. It can be obtained as the trajectory
of a certain point on the circumference of
a circle of radius af4 rolling (without
sliding) upon another circle of radius a
(the smaller circle always remains inside
the larger one; see Fig. 78).
Fig. 78. Note 2. It will be noted that equations
(4) and equation (5) define more than one
function y = f (x). They define two continuous functions on the intervai
-a.;;;;;x..;+a. One takes on nonnegative values, the other nonpositive
values.
SEC. 18. THE DERIVATIVE Of A FUNCTION REPRESENTED

PARAMETRICALLY
Let a function y of x be represented by the parametric equations
X= (x), which also has a derivative. Then the
function y = f (x) defined by the parametric equations may be regarded as a
composite function:
y='ljl(t), t=<l> (x),
t being the intermediate argument.
By the ru)e for differentiating a composite function we get
y; = y; t~ = 'ljl;(t) :(x). (2)
From the theorem for the differentiation ol an inverse functiOn, it follows

that
, I
 (x) = - ,- .
X (j)t(f)
Putting this expression into (2), we have
' 'ljl' (I)
Yx = q>'(t)
or
(XXI)
The Derivative of a Function Represented Parametrically 109
The derived formula permits finding the derivative of a u:

function represented parametrically without having to find the
expression of y as a direct function of x.
Example 1. The function y of x is given by the parametric equations
x=ac?st,} (Oe;;;;;t.;;;;;n).
y=a sm t
Find the derivative ~: 1) for any value of t; 2) for t =~,
Solution.
l) , (a sin I)' a co~ t =-cot t;
Yx =(a cost)'= -a sm t
2) ( y') n =-cot~= -I.

X 1"'-
4
4
Example 2. Find the slope of a tangent to the cycloid

x =a (t -sin t),
!I =a (1 -cost)
at an arbitrary point (Oe;;;;;t.;;;;;2n).
Solution. The slope of a tangent at each point is equal to the value of
the derivative u:
at that point; i. e., it is
, Yt
!I
X
=-...
Xt
But
Consequently,
2 sin .!_ cos !.
, a sin t 2 2 t ( n t )
Yx=a(1-cost)= =cot2 =tan 2-2
2 sin 2~
He.nce, the slope of a tangent to a cycloid at every point is equal fo

tan ( i -~) , where t Is the value of the parameter corresponding to this
point. But this means that the angle a of the slope of the tangent to the x-axis is
1t t
equal to
2- 2 (for values of t lying between -n and n)"''
indeed, the slope Is equal to the tangent of the angle of Inclination a

of the tangent .
to the x-ax1s. And so tana=tan (n2 - 2t) and a= 2n- 2t
for those values of t for which ~- ~ lies between 0 and n,
SEC. 19. HYPERBOLIC FUNCTIONS
In many applications of mathematical analysis we encounter

combinations of exponential functions of the form {(ex -e-") and
i (ex+ e-"). These combinations are regarded as new functions
and are designated as follows:
.
smh x= eX-e-x }
2 (1)
eX+ e-;
cosh X=
2
.
The first of these functions is called the hyperbolic sine, the

second, the hyperbolic cosine. These functions may be used to define
. sinh x cosh x
two more functiOns: tanh x=--h-and
COS X
coth x=-.-h-:
Sin X
tanh x-:-~+:==-the hyperbolic tangent, }

ex +e-x (1')
coth X= ex-e_,.-the hyperbolic cotangent.
The functions sinh x, c.osh x, tanh x are obviously defined for

all values of x. But the function coth x is defined everywhere,
except the point x = 0.
The graphs of the hyperbolic functions are given in Figs. 79,
80, 81.
From the definitions of the functions sinh x and cosh x [formu-
las (I)] there follow relationships similar to those between the
appropriate trigonometric functions:
2
cosh" x~sinh x =I, (2)
cosh (a+ b)= cosh a cosh b+sinh a sinh b, (3)
sinh (a+ b)= sinh a cosh b+cosh a sinh b. (3')
Indeed,
cosh 1 x-smh
1
x= (e"'-" -2- +e-") - (e-'"-e-")
1
2
1
_
e"" +2+e-x-ezx +2-e-zx
= 4 =1.
Further, noting that
ea+b + e-a-b
cosh(a+b)=, . 2 ,
Hyperbolic Functions Ill
we get
. . ea+e-aeb+e-b ea-e-aeb-e-b
cosh a cosh b + smh a smh b = - -
2
- - --
2
+- -
2
- - --
2
=
ea+b +e-a+b +ea-b +e-a-b +ea+b_e-a+b -ea-b +e-a-b
4 -
ea+b +e-a-11
= 2
=Cosh (a+b).
The prove is similar for relation

(3').
The name "hyperbolic functions"
comes from the fact that the func-
tions sinh t and cosh t play the
same role in the parametric
representation of the hyper bola,
X
2
-!/ = 1,
IJ
- - - - -1
Fig. 79. 0 X
IJ
-t
Fig. 80. Fig. 81.
as the trigonometric functionssin t and cos t do in the parametric

representation of the cirCle,
Xz +Yz= 1.
Indeed, eliminating the parameter t from the equations

X= cost, !J= sin t,
we get
X2 + !J 2 = cos 2 t + sin 2 t
or
x2 + y 2 = 1 (the equation of the circle).
Similarly, the equations
x= cosh t,
y = sirih t
are the parametric equations of the hyperbola.
Indeed, squaring these equations termwise and subtracting the
second from the first, we get
X -!J = cosh" t- sinh" t.
2 2
Since, on the basis of formula (2), the expression on the right

side is equal to unity, we have
xz-yz =-1,
which is the equation of the hyperbola.
Let us consider a circle with the equation X 2 + !J 2 = l (Fig. 82).
In the equations x=cost, y=sint, the parameter tis numerically
equal to the central angle AOM or to the doubled area S of the sector
AOM, since 1=25.
Fig. 82. Fig. 83.
Let it be noted, without proof, that in the parametric equations

of the hyperbola,
x= cosh t,
y=sinht,
the parameter t is also numerically equal to the doubled area of
the "hyperbolic sector" AOM (Fig. 83).
The Differential 113
The derivatives of the hyperbolic 'functions are defined by the

formulas
'll
1
(sinh x)' = cosh .t, (tanh x)' = - h 2
COS X
1
(cosh x)' =sinh x, (coth x)' = - -,---h
Stn 2 X
J (XXII)
which follow from the very definition of hyperbolic functions; for
eX-e-"
instance, for the function sinh x = 2
we have
.
(smhx) =, (e"-e-")'
2
= eX +e-"
2
=coshx.
SEC. 20. THE DIFFERENTIAL
Let the function y=f(x) be differentiable on the interval [a, b].

The derivative of this function at some point x of [a, b] is
determined by the equality
lim ~Y = f' (x).
L'.x-+ouX
As ~x-0, the ratio !~ approaches a definite number f' (x) and,

consequently, differs from the derivative f' (x) by an infinitesimal:
!~=f' (x) +a,

where a--+ 0 as ~x ___.. 0.
Multiplying all terms of the latter equality by ~x. we get
~y = f' (x) ~x -t a~x. (l)
Since in the general case f' (x) =1= 0, for a constant x and a variable
~x---.. 0, the product f' (x) ~x is an infinitesimal of the first or-
der relative to Ax. But the product a~x is always an infinitesimal
of higher order relative to Ax, because
lim at-..x = lim a= 0.
L'.X-+0 dX L'.X-+0
Thus, the increment ~y of the function consists of two terms, of

+
which the first is (when f' (x) 0] tile so-called principal part of
the increment, and is linear relative to t!x. The product f' (x) t!x
is called the differential of the function and is denoted by dy or
df (x) (read, dy or df of x).
And so if a function y=f(x) has a derivative f' (x) at the point

x, the product of the derivative f' (x) by the increment llx of the
argument is called the diOerential of the function and is denoted
by the symbol dy:
dy = f' (x) llx. (2)
Find the differential of the function y = x; here,
y'=(x)'=1,
and, consequently, dy = dx = llx or dx = llx. Thus, the differential
dx of the independent variable x coincides with its increment llx.
The equality dx= llx might be regarded likewise as a definition
of the differential of an independent variable, and then the fore-
going example would indicate that this does not contradict the de-
finition of the differential of a function. In any case, we can write
formula (2) as
dy = f' (x) dx.
But from this relationship it follows that
f' (x) = dx
dy.
Hence, the derivative /' (x) may be regarded as the ratio of the
differential of a function to the differential of the independent
variable.
Let us return to expression (1), which, taking (2) into account,
may be rewritten thus:
lly = dy + allx. (3)
Thus, the increment of a function differs from the differential of
a function by an infinitesimal of higher order relative to llx. If
f' (x)+O, then allx is an infinitesimal of higher order relative to
dy and
. d-=
I lffi M 1 I'lffi -atJ.x
+ 1 I' a 1
' ( = + lffi f'-()= .
f )A
dX-+0 y ll.X-+0 X oX ll.X-+0 X
For this reason, in approximate calculations one sometimes uses

the approximate equality
lly ~ dy . (4)
or, in expanded form,
f (x + llx)- f (x) ~ f' (x) llx, (5)
thus reducing the volume of computation.
The Differential 115
Example I. Find the differential dy and the increment AY of the function

I/=X 2:
I) for arbitrary values of x and Ax;
2) for x=20, M=O.!.
Solution. I) Ay=(x+Ax) 2 -x 2 =2xAx+Ax 2 ,
dy=(x 2 )' Ax=2xAx.
2) If x=20, Ax=O.I, then M=2200.\ +(0.1)'=4.01,
dy=2200.! =4.00.
Replacing Ay by dy yields an error of 0.01. In many cases, it may be
considered small compared to Ay=4.01 and the-
refore disregarded.
Fig. 84 gives a clear picture of the above
problem.
In approximate calculations, one also
makes use of the following equality, which
is obtained from (5):
+
f (x ~x) ~ f (x)+f' (x) ~x. (6)
Example 2. Let f (x) =sin x, then f' (x) =cos x.
In this case the approximate equality (6) takes
the form
sin (x+ Ax)::::::: sin x+cos x Ax. (7)
Fig. 84.
Let us calculate the approximate value of
sin 46.
Put x=45=_:!_, AX= 10 =180' "
4
mto (7) we get
sln46 =sin + =sln +cos4
0
4
("
180
")
4 " ""
180
or
sin 46 = 2
-- + 2"
-- =0.7071 +0.70710.017=0.7194.
2 2 180
Example 3. If in (7) we put x=O, Ax=a, we get the following approxi-
mate equality:
sina:::::a.
Example 4. If f (x) =tan x, then by (6), we get the following approximate
equality:
I
tan (x +Ax) :::::::tan x +cos' x Ax,
for x=O, Ax=a, we get
tan a :::::a.
Example 5. If I (x) = yx, then (6) yields
Yx+~x::::: rx+ 2 ~x.~x..

Putting x = 1, Llx =a, we get the approximate equality

. r- 1
y 1 +a::::::1 +ya
The problem of finding the differential of a function is equiva-

lent to that of finding the derivative, since, by multiplying the
latter into the differential of the argument we get the differential
of the function. Consequently, most theorems and formulas perta-
ining to derivatives are also valid for differentials. Let us illustrate
this.
The ditferential of the sum of two ditfermtiable functions u and
v is equal to the sum of the ditJerentials of these functions:
d(u+v)=du+du.
The ditferential of the product of two ditferentiable functions tt
and v is determined by the formula
d (uv) = u dv + v du.
By way of illustration, let us prove the latter formula. If y = uv,
then
dy = y' tu = (uv' + vu') ~x = uv' ~x + vu' 8x,
but
v' ~x=dv, u' ~x=du,
therefore
dy = u dv + v du.
Other formulas (for instance, the formula defining the differen-
tial of a quotient) are proved in similar fashion:
. u udu~udv
tf y = -
v , then dy = v2
Let us solve some problems dealing with calculating the diffe-

rential of a function.
I
Example 6. y = tan 2 x, dy =2 tan x - - dx.
COS 1 X
.r--
Example 7. y = y I +In x, dy = .r I
-I dx.
2 r I+ In x x
We find the expression for the differential of a composite function.

Let
y=f(u), u=cp(x), ory=flcp(x)].
The Geometric Significance of the Difjtrential ll7
Then by the rule for the differentiation of a composite function,

dy
dx = f u' (u) <:p ' (x).
Hence,
dy = f~ (u) <:p' (x) dx,
but q>' (x) dx = du, therefore
dy = f' (u) du.
Thus, the differential of a composite function has the same form
as it waul d have if the intermediate argument were the independent
variable. In other words, the form of the differential does not de-
pend on whether the argument of a function is an independent va-
riable or a function of another argument. This important property
of a differential, called invariance of the form of the differential,
will be widely used later on.
Example 8. Given a function y =sin VX: Find dy. Soluiion. Representing
the given function as a composite one:
y=slnu, u= VX.
we find
1
dy =cos u .r- dx;
2 y X
but
2
!r- dx = du, so we can write
y X
dy=cosudu
or
dy = ccs <Yxl d c v-:t).
SEC. 21. THE GEOMETRIC SIGNIFICANCE OF THE DIFFERENTIAL
Let us consider the function

y=f (x)
and the curve it represents (Fig. 85).
On the curve y=f(x), take an arbitrary point M (x, y), draw a
line tangent to the curve at this point and denote by a the angle*)
which the line tangent forms with the positive direction of the
x-axis. Increase the independent variable by Llx; then the function
*) Assuming that the function f (x) has a finite derivative at the point x,
we get a :p 2:rt ,
will change by !!:..y=NM,. To the values x+!!:..x, y+!!:..y on the

+
curve y = f (x) there wi II correspond the point M 1 (x !!:..x, y !!:..y). +
From the triangle MNT we find
NT=MN tan a;
since
tan a= f' (x), MN = !!:..x,
we get
NT = f' (x) !!:..x;
but by the definition of a differential f' (x) !!:..x = dy. Thus,
NT=dy.
The latter equality signifies that the differential of a function f (x),
which corresponds to the given values x and !!:..x, is equal to the
y
X X+!JX X X
Fig. 8.~. Fig. 86.
increment in the ordinate of the line tangent to the curve y = f (x)

at the given point x.
From Fig. 85 it follows directly that
MJ=!!:..y-dy.
MT
By what has already been proved, NT
--+ 0 as !!:..x- 0.
One should not think that the increment !!:..y is always greater
than dy. For instance, in Fig. 86,
!!:..y=M,N, dy=NT, and !!:..y<dy.
SEC. 22. DERIVATIVES OF DIFFERENT ORDERS
Let a function y= f (x) be differentiable on some interval [a, b).

Generally speaking, the values of the derivative f' (x) depend on x,
which is to say that the derivative f' (x) is also a function of x.
Derivatives of Dif}erent Orders 119
Differentiating this function, we obtain the so-called second

derivative of the function f (x).
The derivative of a first derivative is called a derivative of the
second order or the second derivative of the original function and
is denoted by the symbol y" or f" (x):
y" = (y')' = f" (x).
For example, if y = x', then
y' = 5x 4; y'' = (5x')' = 20x 1.
The derivative of the second derivative is called a derivative uf
the third order or the third derivative and is denoted by y"' or
f"' (x).
Generally, a derivative of the nth order of a function f (x) is
called the derivative (first-order) of the derivative of the (n-1) st
order and is denoted by the symbol y<n) or (x): rm
y<n> = (y<n-11)' =fin> (x).
(The order of the derivative is taken in parentheses so as to avoid

confusion with tlie exponent of a power.)
Derivatives of the fourth, fifth, and higher orders are also
denoted by R.oman numerals: y 1v, yv, yv 1, Here, the order of the
derivative may be written without brackets. For instance, if y=x 5 ,
theny' = Sx\ y" = 20x1, y'" =60x1, y 1V =y'' 1= 120x, yV = y151 = 120,
y<GI = yPI =. = 0.
'Example 1. Given a function y=e'"' (k=const). Find the expression of its
derivative of any order n.
Solution. y' = kekX, y = k 1ekx, , y<n> = knekx.
Example 2. y =sin x. Find y<n>.
Solution.
y' =cos x =sin ( x + ~) ,
y =- sin x =sin ( x + 2 ~ ) ,
y'" =-cos x=sin ( x+3 ~),

1
yv =sinx=sin ( x+4 ~),
12Q Derill at i ve and D iflerent ial
In similar fashion we can also derive the formulas for the

derivatives of any order of certain other elementary functions. The
reader himself can find the formulas for derivatives of the nth
order of the functions y = x,., y =cos x, y =In x.
The rules given in theorems 2 and 3, Sec. 7, are readily
generalised to the case of derivatives of any order.
In this case we have obvious formulas:
(u + v)'n' = u'n' + v'n', (Cu)'n' =Cu n'.
1
Let us derive a formula (called the Leibniz rule, or formula)

that will enable us to calculate the nth derivative of the product
of two functions u (x) v (x). To obtain this formula, Jet us first find
several derivatives and then establish the general rule for finding
the derivative of any order:
Y=UV,
y' =u'v+uv',
y" = u"v + u'v' + u'v' + uv" = u"v + 2u'v' t uv",
y"' = u"'v + u"v' +2u"v' + 2u'v" + u'v" + uv"' =
= u'"v + 3u"v' +3u'v" +uv"',
y 1V = u 1vv + 4u"' v' + 6u"v" + 4u' v"' + uv 1v.
The rule for forming derivatives holds for the derivative of any
order and obviously consists in the following.
+
The expression (u v)n is expanded by the binomial theorem,
and in the expansion obtained the exponents of the powers of u
and v are replaced by indices that are the order of the derivati-
ves, and the zero powers (u 0 = V 0 = I) in the end terms of the
expansion are replaced by the functions .themselves (that is,
"derivatives of zero order"):
y'n) = (uv)<n) = u'n'v + nuln-1)v.' + n (~~1) u(n-2)v" + .. + uvln).
This is the Leibniz rule.

A rigorous proof of this formula may be performed by the me-
thod of complete mathematical induction (in other words, to prove
that if this formula holds for the nth order it will hold for the
order n +I).
Example 3. y=e 0 xx 2 Find the derivative ol y 1n1
Solution.
U=eax, !I=Xz,
u' =aeax, u' =2x,

Differentials of Various Orders 121
or
SEC. 23. DiffERENTIALS Of VARIOUS ORDERS
Suppose we have a function y = f (x), where x is the independent

variable. The differential of this function
dy=f' (x)dx
is some function of x, but only the first factor, f' (x), can depend
on x; the second factor, (dx), is an increment of the independent
variable x and is independent of the value of this variable. Since
dy is a function of x we have the right to speak of the differen-
tia I of this function.
The differential of the differential of a function is called the
second differential or the second-order differential of this function
and is denoted by d 2 y:
d (dy) = d1 y.
Let us find the expression for the second differential. By virtue
of the general definition of a 'differential we have
d 1y = [f' (x) dx]' dx.
Since dx is independent of x, dx is taken outside the sign of the
derivative upon differentiation, and we get
d"y = r (x) (dx) 1

When writing the degree of a differential it is common to drop

the brackets; in place of (dxl we write dx" to mean the square
of the expression dx; in place of (dx) 1 we write dx', etc.
The third differential or the third-order differential of a function
is the differential of its second differential:
d 3 y = d (d 2y) = [f" (x) dx 1
]' dx = f"' (x) dx 3
Generally, the nth differential is the first differential of a diffe-
rential of the (n-l)st order:
d"y=d(d"- 1 y)=lf 1"- 11 (x)dx"- 1 ]' dx, d"y=f 1"'(x)dx". (1)
Using differentials of different orders, the derivative of any order

may be represented as a ratio of differentials of the appropriate
order:
f' (x) = :~ ; r (x) = :~ ' ... ' rn) (x) = ::~ . (2)
It should, however, be noted that equalities (1) and (2) (for n> 1)
hold only for the case when x is an independent variable.*)
SEC. 24. DIFFERENT-ORDER DERIVATIVES OF IMPLICIT FUNCTIONS

AND OF FUNCTIONS REPRESENTED PARAMETRICALLY
1. An example will illustrate the finding of derivatives of

different orders of implicit functions.
Let an implicit function y of x be defined by the equality
(I)
Differentiate, with respect to x, all terms of the equation andre-

member that y is a function of x:
2x + 211 dy = 0.'
a2 b 2 dx
from this we get
(2)
Again differentiate this equality with respect to x (having in view
that y is a function of x):
dy
dzy b y-xdx
dxz=- az liz
Substituting, in place of the derivative :~, its expression from

(2), we get
Nevertheless, we shall also write equality (2) when x Is not an lnd'epen-

dz dnl/
dent variable; but In this case, the expression a}a, ... , dxn should be
regarded as symbols of derivatives.
Different-Order Derivatives of Implicit Functions 123
From equation (1) it follows that

a~y~ + b2x2 = a2b2;
therefore the second derivative may be represented as
d2y b4
dx 2 = - a 2y'
d'y
Differentiating the latter equation with respect to x, we find dx',
etc.
2. Let us now consider the problem of finding the derivatives
of higher orders of a function represented parametrically.
Let the function y of x be represented by parametric equation$
X= cp (f), }
y='ljl(t), ! 0 <.t<.T; (3)
the function x = cp (t) has an inverse function t = t (x) on the inter-

val [! 0 , T].
In Sec. 18 it was proved that in this case the derivative ~~ is
defined by the equation
dy
dy dt
dx = dx (4)
df
To find the second derivative, ~~, differentiate (4) with respect

to x, bearing in mind that t is a function of x:
d 2y d
dx 2 = dx
iii
dy)
dx =
d (dy)
dt
Cit dx dx '
dt
(5)
(
dt I dt
but
.!!.
dy) =dTdT dT
dt
dx d (dy). dy d (dx)
-(jf(fl di
dt (dx
dt '
(dx) . dt .
2
: . .; .
Substituting the latter expressions into (5), we get

dx d 2 y dy d'x
d 2y dt dt2- iii (jj2
dx 1 =
This formula may be written in more compact form as follows:

d'y cp' (t) 'ljl" (t)-'ljl' (t) cp" (t)
dx 1 = [cp' (t)] 3
In similar fashion we can find the derivatives
d 1y d 4y
dx 3 ' dx
and so forth.
Example. A function y of x is represented parametrically:
x =a cos t, y = b sin t.
dy d 2 y
Find the derivatives dx, dx
Solution.
dx . d'x
dt=-asmt; dt= -acvs t;
dy d2 y .
dt=bcost; dt=-bsint,
dy b cost =_~cot t;
dx -a slnt a
d 2y (-a sin t) ( -b sin t)-(b cost) (-a cost) b
dx 1 (-a sin /) 1 = - (i2 sin1 t
SEC. 25. THE MECHANICAL SIGNIFICANCE OF THE SECOND DERIVATIVE
Let s be the path covered by a body under translation as a

function of the time; it is expressed as
s=f(t). (I)
As we a! ready know (see Sec. I, Ch. I II), the velocity u of a body
at any time is equal to the first derivative of the path with
respect to time:
ds
(} = dt. (2)
At some time t, let the velocity of the body be v. If the motion
is not uniform, then during an interval of time ll.t that has elapsed
since t the velocity will change by the increment 11v.
The Mechanical Significance of the Second Derivative 125
The average acceleration during time l1t is the ratio of the

increment in velocity llv to the increment in time:
~v
aau = ~t.
Acceleration at a given instant is the limit of the ratio of the

increment in velocity to the increment in time as the latter
approaches zero:
a= lim ~v;
M .... o ~I
in other words, acceleration (at a given instant) is equal to the

derivative of the velocity with respect to time:
dv
a= dt'
.
bu t smce ds consequen tl y,
v = dt,
d (ds) d's
a= dt di = dt 2 '
or the acceleration of linear motion is equal to the second deriva-

tive of the path covered with respect to time. Reverting to equation
(1 }, we get
a=r (f).
Example. Find the velocity v and the acceleration a of a freely falling
body, if the dependence of distance s upon time t is given by the formula
&=; gt 2 + u0 t+so (3)
where g = 9.8 m{sec 2 is the acceleration of gravity, and s0 = s 1, 0 is the value

of s at i=O.
Solution. Differentiating, we find
ds
v=dt=gt +v 0 ; (4)
from this formula It follows that V0 = (v) 1 _ 0

Differentiating again, we find -
de d 2s
a=dt"= dt'=g.
Let It be noted that, conversely, if the acceleration of some motion is con-

stant and equal to g, the velocity will be expressed by equal ion (4); and
the distance by equation (3) provided that (v) 1, 0 =v 0 and (sJt= 0 =s 0
126 Derivative and D iflerential
SEC. 26. THE EQUATIONS OF A TANGENT AND OF A NORMAL.

THE LENGTHS OF THE SUBTANGENT AND THE SUBNORMAL
Let us consider a curve whose equation is

y=f(x).
On this curve take a point M (x , y ) (Fig. 87) and write the
equation of the tangent line to the' gi~en curve at the point M,
assuming that this tangent is not parallel to the axis of ordinates.
The equation of a straight line with slope k passing through the
point M is of the form
y
y-y, =k (x-x,).
For the tangent line (see Sec. 3)
k= f' (x,),
and so the equation of the tangent
is of the form
X
Y-'-Y, =f' (x,) (x-x,).
In addition to the tangent to a
Fzg. 87. curve at a given point, one often
has to consider the normal.
Definition. The normal to a curve at a given point is a straight
line passing through the given point perpendicular to the tangent
at this point.
From the definition of a normal it follows that its slope kn is
connected with the slope k 1 of the tangent by the equation
k = _ _!_
n kt
or
1
kn = - f' (x,)
Hence, the equation of a normal to a curve y = f (x) at a point

M (x,, y,) is of the form
I
y-y, = - f' (x,) (x- x,).
Example 1. Write the equations of a tangent and a normal to the curve
y=x1 at the point M (l, 1).
Solution. Since y' =3x1 , the slope of the tangent is (y'}x= 1 = 3.
Therefore, the equation of the tangent is
y-1 =3 (x-1) or y=3x--:-2.
The Equations of a Tangent and of a Normal 127
The equation of the normal is y

1
y-1 = - (x-1)
3
or
I 4
y=-3 X +3
(see Fig. 88).
The length T of the segment QM X

(Fig. 87) of the tangent between
the point of tangency and the
x-axis is called the length of the
tangent. The projection of this
segment on the x-axis, that is,
QP, is called the subtangent; the
lenght of the subtangent is deno-
Fig 88.
ted by Sr. The length N of the
segment MR is called the length
of the normal, while the projection RP of the segment RM on the
x-axis is called the subnormal; the length of the subnormal is
denoted by SN'
Let us find the quantities T, ST> N, Stv for the curve y = f (x)
and the point M (x" y,).
From Fig. 87 it will be seen that
Qp = y 1 cot a= _!/)_ =--= J!..!..

tan a Y,' '
therefore
Sr=l :; I
T=V y~+ :: 2 =1 :: y y;+ II.
It is .further clear lrom this same figure that
PR=y, tana=y,y;,
and so
Stv=ly,y;i,
N= V y: + (y,y;) = jy, V 1 + y; \.
These formulas are derived on the assumption that y,
However, they hold i:-~ the gen~ral case as well.
> 0, u:> b.
y Example 2. Find the equations of

the tangent and normal, the lengths
of the tangent and the subtangent,
!he lengths of the normal and
+---+---4..6---+----=::::..o..-.x subnormal for the ellipse
Q
x = a cos t, y = b sin t (1)
at the point M (x 1 , g 1) for which
Fig. 89.
t = T (Fig. 89).
Solution. From equations (I) we find
dx dy . dy b
dt=-aslnt; dt=bcost, dx=--acott;
We find the coordinates of the point of tangency of M:

a b
xt = (x) = .r-n2 , y.=(y) = ,r-
t =2!. r "' t=.!!. r 2
4 4
The equation of the tangent is
or
y- h =- : ( X - V2)
bx+ay-ab y'2 =0.
:2 =%(
The equation of the normal is
y- X - V2)
or
(ax-by) V2 -a 2 +b 2 =0.
The lengths of the subtangent and subnormal are
b
V2 a
ST= _.!!_ =y2'
a
SN=I h (--~)I= ;2 a
The lengths of the tangent and the normal are
b
T= ~! V (- :r +I = ; 2 V az + b':
a
N=i :7[ ~I +( -~ rl= a ~ 2 Va 2

+b2,
The Geometric Significance of the Derivative 129
SEC. 27. THE GEOMETRIC SIGNIFICANCE OF THE DERIVATIVE

OF THE RADIUS VECTOR WITH RESPECT TO THE POLAR ANGLE
We have the following equation of a curve in polar coordinates:

Q= f (6). (1)
Let us write the formulas for changing from polar coordinates
to rectangular Cartesian coordinates:
!/
X=QcosO, y=Q sinfJ.
Substituting, in place of Q, its expression
in terms of 0 from equation (I). we get
X= f (0) COS fJ, }
(2)
y = f (0) sin f).
Equations (2) are parametric equations of 0 X

the given' curve, the parameter being the Fig. 90.
polar angle 6 (Fig. 90).
If we denote by q> the angle form.ed by the tangent to the
curve at some point M (ll, 0) with the positive direction of the
x-axis, we will have
dy
dy de
tan cp = dx =dx
de
or
~ sin 8 + Q cos 8
tancp=d (3)
Is cos e- Q sin e
Denote by f.l. the angle between the direction of the radius vector
and the tangent. It is obvious thatf.l.=q>-6,
tan !p- tan 6
tanf.l. =.,.....,......,....:--,.---""'
I +tan !Jl tan 8
Substituting, in place of tan cp, its expression (3) and making
the necessary changes, we get
t _ iQ'sln 0 +Q cos 8) cos.S- (Q' cos 6-Q sin 6) sin 6 _ Q
an f.l.- (Q' cos 0-Q sin 8) cos 8+(Q'sin 8+Q cos6) sine-?
ot
Q; = Q COtf.l.. (4)
5-338R
Thus, the derivative of the radius vector with respect to the

polar angle is equal to the length of the radius vector multiplied
by the cotangent of the angle between the radius vector and the
tangent to ,the curve at the given point.
Example. To show that the tangent to the logarithmic spiral
Q=ea&
intersects the radius vector at a constant angle.

Solution. From the equation of the spiral we get
Q' = aea&,
From formula (4) we have
cot f.l. = fLQ =a; that is, f.l. =arc cot a= const.
Exercises on Chapter Ill
Find the derivatives of functions using the definition of a derivative.

I 1 ,r-
], y=x1 Ans. 3x1 2. y=-. Ans. - 1 . 3. y= r x. Ans. -=.
1
x x ' 2Yx
4.y=;x. Ans.-2x~~. 2
5. y=sln x. Ans. 2slnxcosx. 6. y=2x-x.
Ans. 4x-1.
Determine the tangents of the angles of inclination of tangents to the
curves: 7. y=x 3 a) When x= 1. Ans. 3. b) When x=- 1. Ans. 3. Make a .
drawing. 8. y=+ a) When x=~ Ans. -4. b) When x=l. Ans. -1.
Make a drawing. 9. y= Vx when x=2. Ans. .~-.

2" 2
Find the derivatives of the functions: 10. y= x'+ 3x2 -6. Ans. !/ = 4x~ + 6.~:.
x' x2
11. y=6x 3 -x 2 Ans. y'=18x 2 -2x. 12. y=a+b-a-b-x. Ans. y'=
2x x 3 -x"+ I ,
5x4
= a-+-b- -a--b- 1. 13 Y = ----::-5-'-- Ans. y = 3x"-2x
- -
5
-. 14. y=2ax 3 .....:.
-li+c. Ans. y'=6ax"- 15. y=6x 2 +4x 2 +2x. Ans. y'=2lx 2

+
(x+1) 1 , 3(x+l) 2 (..t-l) x m x2 n

11. y=-,-. Ans. y = , . 1s. Y=m+-x+na+x-2.
x' 2x
Exercises on Chapler Ill 131
2
Ans. y' =_!__m
m x
+~- n".
n x
19. y = v
2
Xi -2Yx +5. Ans. y' =-3 d)-x.-
v
- Yx. 20.
I ax b x
Y= vx-+ X Yx- Yx . Ans.
V- ,5.!..3
y = 3 ax a - 2 bx
5
+
?
-- 1
+6x e.
21. y=(l +4x 8 ) (l+2x 2 ). Ans. y'=4x (I +3x+ 10x 8). 22. y=x (2x-1)(3x+2).
Ans. y' =2(9x 2 +x-1). 23. y=(2x-1) (x 2 -6x+3). Ans. y' =6x 2 -26x+ 12.
2x 4 4x 5 (2b 2 -x 2) a-x A , 2a
24. y=-b-x 2 Ans. y' (b 2 -x 2 ) 2 25 ' Y=a+x' ns. y =-(a+x) 2
t 12 (3+1 2) (s+4) 2
26. f (I)= 1 t. + Ans. f' (t) (I +t). 27. f(s)= 5
,
+ 3 . Ans. f (s) =
(s+2)(s+4) x +I , :c4 -2x 3 -6x 2 -2x+ I
(s +3) 2 28. Y= x- x-2. Ans. y = (x- x-2)"
xP , xP- [(p-m) xm-pam]
1
29, y. xm-am An.~. Y = (x'n-am) . 30. Y=(2x 2 -<3) 2,

Ans. y'=8x(2x 2 -3). 31. y=(x 2 +a 2) 5 Ans. y=l0x(x 2 +a 2) 4 32. Y=
= .~
r x-+a-. Ans. y' = .r x . 33. y= (a + x ) .r--
r a-x. Ans. y' a-3x
.r .
" x+a 2 " a-x
, /1 +x I 2x 2 - l
34. y= V 1-x. Ans. y' (1-x) Yl -x. 35. y= x Yl + x' Ans. y' =
I + 4x 2
)/ , 2x + I
36. y= x +x+l. Ans. y = )/ . 37. Y=
-2 3 (x'+x+l) 2
x (1 +x 2
v-J
)
=0+)/-;) . Ans. y'=(1+ 38. y=yx+Vx+ Jfx. Ans. y'=
=
2
V x+Vx+
1
Vx
[1 + 2
V
1
_
x+Vx
(t + .~..,)].
2 I' x
39. y=sin 2 x. Ans.
Y.'=sln2x. y=2slnx+cos3x. Ans. y'=2cosx-3sln3x. 41. Y=

40.
b) A ns.y=
, a sin x A , I
=ant (ax+. '( +b)' 42. y= 1 + cos x . ns.y= 1 + cos x
cos ax
43. y =sin 2x cos 3x. An~. y' = 2 cos 2x cos 3x-3 sin 2x sin 3x. 44. y = cot 5x.
Ans. y' =-10 cot 5x esc 5x. 45. y=t sin t +cost. Ans. y' = tcos t. 46. y =
=sln 1 tcost. Ans. y'=sln 2 1(3cost-sln 2 t). 47. y=aVcos2x. Ans. y'=
tan ~+cot~
a sin 2x . 3 q> A ' q> q> 2 2
8 1 2 cos . 49.y
Vcos2x.4.r=asm 3 ns. r'l'=asn 3 3 X
s
132 Derivatiue and Differential
,
Ans. y = -
2xcosx+sin'x(tan1-+coti)

(
.50.y=a I-eos -
,x)'.An.,.y, =
X 2 Sin 2 X 2
I
= 2a sln 1 2X cos
X
2
. 51. u=
2
tan x. Anb. y' =tan x sec x. 52. y =In cos x.
Ans. y'=-tanx. 53. y=lntanx. Ans. y'=~

Sin 2 X
. 54. y=ln sln 2 x. Ans.y'=
tanx-1 , yl+sinx
=2cotx. 55. y= . Ans. y =sinx+cosx. 56. y=m .
secx 1
1 -snx
Ans. y' = -- . 57. y= In tan (~ + ~ ). An~. y' = --. 58. y=sin (x+a)x
4 2
1 1
COS X COSX
X cos (x +a). Ans. y' =cos 2 (x +a). 59. I (x) =sin (In x). Ans. {' (x) =
cos (In x) sec (In x)
= X
. 60. I (x) =tan (In x). Ans. r (X)= X
. 61 f (x) =sin (COS x).
An~. f' (x) =- sin x cos (cos x}. 62. r = ~ tan 3 <p- tan (jl + <p. Ans ~~ = tan (jl.
63. /(x)=(xcotx) 2
An.s. {'(x)=2xcotx!cotx-xcscx). 64. y=ln(ax+b)
2x 66. y=
'a y=log(l(x+ 1). A
n.\. '
2
A.nb. y=ax+b" 65. y (x+l)lna
=
I'+
x A , 2 2
1
lnl-x ns. y =1-x 67. y= 1oga(X -snx).
A , 2x- cos x
ns. y =(x'-sinx)ln3.
I + x , 4x . , 2x + I
68. y= In - - -2
1 -x
Ans. y =-1 - -
-x4
69.
.
y= In (x 2 +x). An~. y =-.--.
x +x
a 2x + 5). Ans. y' =x-2x+
70. y= In(x- I y=xlnx. Ans. y'=lnx+l
3x'- 2 . 7.
5
72. y=ln 1 x. Ans. y'=aln'x. 73.y=ln(x+JIT+xf). Anb. ~'=,r 1
x r I +x
1 1
74. y=ln(lnx). Ans.y'=-- - . 75. f(x)=ln .. /ll+x. Ans.f'(x)=-,
x 1nx V -x 1-x
Yx"+ 1-x 2 .r--
76. {(x)=ln .r . Ans. f' (x)=- .r . 77.. y= r a'+x-
r x-1 +x r I + x
a+ Jl a+x' Ya'+x Vx'+a"
-aln . Ans. y' = . 78. y= In <x+ Vx+a')-.;;..--=---
x X X
, V x + a cos x I x , I
Ans. y = x 79. y=- 2 sin' x + 2 In tan 2. Ans. Y = sin x
sin x , I + sln x 1 2
W. y-=---,-. Ans. !I= 81. y=- tan x+ln cos x. Ans. y' =
2 COS X 2 COS 1 X 2
=tan'x. ~2. y=ea". Ans. y'=aea". 83. y=e"'+-~. Ans. y'=4e'""+.
84. (J=a". An.s. 2xa"lna. 85. y=7""+"'. An.s. y'=2(x+l)7""+""1n7,
86. fi&:::Ca'-""'. An~. y' = - 2xca-x In c. 87. y = a/,.x, An~. y' = .~- eVX.
2 r x
Exercises on Chapter III 133
1 8
6 , o .In 8 dr a " 1na lna-1
88. r=a. Ans, r =a Ina. 89. .r=a . Ans. de= e =8 In a.
e:~- I A
90. y=e x ( 1 - x). A ns. y , =e x ( 1 - 2 x-x ) . 91 . y=ex+ , 2ex
. ns. Y =(ex+ I)
1
X X
' I a - --
92. Ans. y=l+e"'' 93. y=-(ea -e a). Ans. y' =
2
I ~ -~
= (ea_+e a). 94. lj=elnx. Ans. y'=e 51"xcosx. 95. y=atannx. Ans. y'=
2
= na 13 " nx sec 2 nx In a. 96. Y= ecosx sin x. Ans. y' = ecos x (cos x -sin 2 x).
97. y=exlnsinx. Ans. y'=ex (cotx+lnsinx). 98. y=xne' 1"-". Ans. y'=
1 -
A ns. y ' =xx(l-lnx)

-x-2- _.
Ans. y' = e-"" (I+ In x) x"'.
104. 1
y=xnx. I
Ans. y'=xnx (sin X
x-+lnxcosx ) .
105.y~(smx)x.Ans.y'=
=(sin x)x (In sin x+x cot x). y= (sin x) 13 " -". An.~. y' =(sin x)_13 " -"x
106.
2
l-ex , evc I
X(l+sec xlnslnx). 107. y=tan l+ex Ans. Y =-(l+ex)z .l-ex
cos 1 +e'~
108. y=sln ,r;----;:;x Yf=2X 2x In 2. 109,.y= 1"-"

r I.-..:--. A ns.y ' = - COS,r u tan x A ns. y' =
2 r 1-2x
-= w tan X In 10. (tan X+ _x_)
cosx
.
Find the derivatives of the functions after first' taking logarithms of these
functions: '
110. - vx(x +1l

Y- (x-1)2 .
2
Ans.
I v x (x
y'=3
I) (I 2x
2
2 )
(x-1) 2 x+x 2 +1+x-l
+
_!x+I)' Y~ A , (:t+l)' V~ (-3-+ 3 -

Ill. y-
(x-3)
ns. yV . V (x-:- 3) x
+ I 4 (x- 2)
(x + I ) +
2
2 ) 112 , (x + I) (5x 2 + 14x 5)
-5(x-3) ' y=(x+2J' :x+3)4 Ans. Y = - (x+2)'(x+3) 5
113. y
Vcx-t) Ans. y' -161x 2 +480x-2?1
V<x- 2)' Vex- 3) 7

114. y
(1-x)
=5x"(a+3x) (a-2x)(a +2ax-l2x 2 ). 116. y=arc

2 2
sin~. A'ns. y'= y'
1
a a-x
. ) . Ans. y , = 2.~.
I 17. y=(Jrcsmx arc sin x 118. y=arcco t (x 2 + 1). Ans. y , =
y 1-x
2x 2x A , 2 ( 2)
1
+(x+ I). 119. y=arc cot 1-x. ns. y = 1 +x. 120. y=arc cos x .
,. -2x arc cos x A , - (x + JfT=Xi arc cos x)
Ans. y=,r . 121.y=---. ns.y= .r .
y 1- x x x r 1 -x
X+ 1 I ,r=;--::; X
122. y=arc sm ,r-. Ans. y'=,r . 123. y=x r a-x+aarc sin- .
r 2 r 1-2x-x 2
a
Ans. y'=2Va-x. 124. Ans. y' =
= -./aa-+x'
V 125. v+a
u=arccotl-av Ans. du 1
dv=l+v 126. Y=
=y-
I
arccot 1_x
x Y3 Ans. y'
x 2 +1 .
x"+x+ 1 . 127. y=xarcsmx. Ans. y'=
3
1
=nrcsinx+y x
1-x
. 128. f(x)=arccos(lnx). Ans. f'(x)=-
x
Y 1-!nx
129. [(x)=arcsinYsinx. Ans. f'(x)= cosx . 130. y=.
2 Ysinx-sin 2 x
.. /I-eos x
=arccot V 1+cosx (Oo;;;;;;x<n). Ans.y'= 21 . 131. y=earcco t x.Ans.y'=
erc cot x ex -e-x 2 1 arc sin x
= I+x . 1.32. y=arccot 2 . Ans. y' e"+e x. 33. y=x .
Ans. y'=xarcstnx(arcsinx+~) 134. y=arcsin(slnx). Ans. y'=

x 1-x
cos x { +I in 1st and 4th quadrants. t 4 sin x A ,
= jcosxj= -1 in 2d and 3d quadrants. 135 .y=arcco 3+5cosx ns.y =
4
136. a
y=arccot-+ln ~x-a ~ . 137.y=
- + . Ans. y , =-.--
3
=s+ 3cos x X X a X -a
I
l+x)4 I x2 3x 1 - l ,r;-;--::-;
=In ( l-x - 2 arctanx. Ans. y'=t-x 138, y=3T+Inr 1+x+
, x1 +1 I x+1 I 2x-l
+arctanx. An~. y =x'+x 139. y=-
3 ln Vx-x+ +,r-arctan ,;-.
1 r 3 r 3
Exercises on Chapter Ill 135
I + x Y2 + x
,
X -
I
A ns. y =-,-- . 140. y=ln.
1 J- X
Y 2 + x 2 +2arc tan -x1--Y2
X
, 4 Y2
-2 Ans. y =+ -.
1 X
x"-1 2n I xI"
141. y =arc cos --uz----- . Ans.
X +1 x(x 2"+ I)
Differentiation of Implicit Functions
dy f dy 2p
Find dx t. 142. Ans. dx=y 143.
dy x . dy
Ans. - = - - . 1<14. b2x"+a 2y 2 =a 2 b 2 Ans. -d
dx y x
tx
ay
=---..
145. y'-3y+2ax=O.
dy 2a _:_ .:._ 2._ dy vr-y )-=-

Ans. dx= 3 (I-y) 146. x +y =a 2 Ans. d~=- x' 147. tt +
+Y' ~ Ans. ddy
-=- =a'.
X
=;_
v- }!_.
X .
148. y 2 -2xy+b"=O. Ans. d!
X
,
dy =-y-.
y-x
dy ay-x 2 dy
149. x'+y'-3axy=0. Ans. -dx =- --. 150. y=cos (x+y). Ans. dx=
y 2 -ax
sin (x + y) . 151. cos (xy) = x. Ans. ddy =- I+ Y_ si~ (~y)
I +sin (x + y) x x sm xy
Find ~~ of functions represented parametrically:

dy b
152. x=acost; y=bslnt. Ans. -=--cott. 153. x=a(t-sint); y=
dx a
=a (1 -cos t) . A ns. dy
dx=co t 2t . 154. x=acos't ; y= b sm
. 't . A ns. dy
dx=
b 3at 3at dy 2t
=--atant. 155. x= 1 + 12 ; y=I+t"' Ans. dx=I-t' 156. u=2lncots;
du
v=tans+cots. Show that dv=tan2s.
Find the tangents of a ogles of the slopes of tangent lines to curves:
157. x= cost, y= sinl at the poi~t x=- ~, Y= ~3 . Make a drawing
Ans.
1
. 158. x=2 cost, y=sin I at the point x= 1: y=- Y23 . Make a
v3
drawing. Ans.
2 ~3 159. x=a (t-sln t), y=a (I-eos t) when t= ~. Make
a drawing.Ans. I. 160. x =a cos t, y =a sin' t when t = : . Make a drawing.

Ans. -1.-161. A body thrown at an angle a to the horizon (in airless space)
described a curve, under the force of gravity, whose equations are: x =
gt
=v 0 cosat, y=v 0 sinat-2(g=9.8 mfsec 2 ). Knowing that a=60, v 0 =
=50 mfsec, determine the direction of motion when: 1) t =2 sec; 2) t =7 sec.
Make a drawing. Ans. 1) tan<p 1 =0.948, <p 1 =4330'; 2) tan<p1 =-1.012,
<p, = + 1347'.
Find the differentials of the following functions:
162. y=(a 2 -x 2) 5 Ans. dy=-!Ox (a 2 -x 2 ) 4 dx. 163. y=
Ans. dy= Yl +x 2
xdx I xlnx
= Y . 164. y=-3 tan x+tanx. Ans. dy=sec xdx. 165. y=-
3 4
-+
I +x' 1 -x
In xdx
+ In (1-x). Ans. dy = (l-x).
Calculate the increments and differentials of the functions:

166. y=2x 2 -x when x=1, t~x=O.Ol. Ans. t~y=0.0302, dy=0.03. 167. Gi
ven y=x 3 +2x. Find flY and dy when x=-1, t~x=0.02.Ans.t~y=0.098808,
dy =0.1. 168. Given y=sin x. Find dy when x=~, t~x=fi. Ans. dy =
=~=0.00873. 169. Knowing thatsin60= ~"3=0.866025; cos60=~,find

the approximate values of sin 603' and sin 6018'. Compare the results with
tabular data. Ans. sin 603':::::: 0.866461; sin 6018' ::::::0.868643. 170. Find the
approximate value of tan 454'30". An.s. 1.00262. 171. Knowing that
1og 10 200=2 30103 find the approximate value of log 10 200.2. Ans. 2.30146.
Derivatives of different orders. 172. -y=3x 1 -2x'+5x-l. Find y".
12
Ans. 18x- 4 . I 73. y= V --;x. F"_tn d y., . Ans. 42 - , 174

125
x y=x.
I F"In d y (I) .
Ans. 6 I. I 75 . y = xn.
C F"tn d y. + I) C . I 76 . y
" A ns. n (nxn+ = .r-,--,
r a -x . F"md y" .
Ans.- a.r . I 77. y = 2 .r-

r x. F"tn d y 141 .. A ns. - IS . 178. y =
.r-7
(a 2 -x 2) r a2 -x 2 8 r X
=ax'+?x+c. Find y"'. Ans. 0. 179. f(x)=ln(x+1). Find f 1v(x).

Ans.- (x ~I)' . 180. y =tan x. Find y"'. Ans. 6 sec 4 x-4 sec x. 181. y =In sin x.
Find. Ans. 2 cot X esc x. 182. f (x) = y sec 2x. Find. (x). Ans.
!I'. r (x) = r
s f (x). 183. y= _x.
= 3 [f (x))- xs Find t<J (x). A ns. (l-x)'.
4I 184 P =
1
.
Fmd d
2
dxy2 .Ans. ay2 .186. y=cosax. Find y 1n1 An~. ancos ( ax+n n)
2 .
187. y=a". Find ylnJ. Ans. (lnan)a". 188. y=ln(l+x). Find ylnJ.
A ns. (- 1) n-t(n-1)1
(l+x)n.
18 _1-x F" d (nJ A 2 -In
9. Y-l+x" tn y . ns_. (
nl
) (l+x)n+t.
190. y=e"x. Find y1n'. Ans. e" (x+n). 191. y=xn- 1 1nx. Find y 1n1,
1 1
Ans. (n x ) 192.y=!in2 x. Find ycnl. Ans. -2.n-t cos (2x + ; n). 193. y =X slnx.
Find y 1m. Ans. xsln(x+-in)-ncos(x+i n). 194.lf y-e"slnx, prove

d 2y 4a 2
that y"-2y' +2y=0. 195. y 2 =4ax. Find -d. Ans.-- 2
. 196. b x'+ay=
X y1
A ns. - 2b
3; 3b"x
-....-s. I 97. x +11 a =r. F'111 d
ay ay
dy
198. y2 -2xy=0. Find dx. Ans. 0. 199. Q=tan (CJ>+Q).
2 (5 + 8Q 2 + 3Q 4 ) . d.Q
Ans. Q8
200. sec q> cos Q =C. Fmd dcp
tan 2 Q-tan 2 q> " y . d 2y (1-e"+Y) (e" -eY)
Ans. t 1 201. e +x=e +Y F1nd -d2 Ans.
(eY + l)a
an Q x
d 2q 2a 1 xy
202. y 1 +x'-3axy=0. Find d;. Ans. (y 2 -ax). 203. x=a (t-sint),
. d 2y I
y =a (I-eos t). Ftnd dx 2 Ans. ( t ) . 204. x=a cos 2t, y = b sin' t.
4
4asln
2
Sh ow th a t dy t F'111 d d'y 3 cost
0 20.
dx 2 =. 5 x=acost, y=asm. dx A ns.- asinst
dan dn+l .
206. Show that dxn (sinh x) =sinh x; d.tzn+t (sinh x) =cosh x.
Equations of a Tangent and Normal.

Lengths of a Subtangent and a Subnormal
207. Write tl~e equations of the tan~ent and normal to the curve y=x-
- 3x 2 -x+5 at the point M (3, 2). Ans. The tangent is Sx-y-22=0; the
normal, x+8y-19=0. 208. Find the equations of the tangent and normal
of the length of the subtangent and subnormal of the circle x'+u'=r at the
point M(x 1 , y 1). Ans. The tangent is xx,+yy 1 =r 2 ; the normal is x 1 y-y,x=0;
.I u: I
sr= -~ ; sN=I-x 1 j.
209. Show that the subtangent of the parabola y 2 =4px at any point is
divided into two by the vertex, and the subnormal is constant and equal to
2p. Make a drawing.
210. Find the equation of a tangent at the point M (x 1 , y1):
. x y xx, YYt l
a) To the ellipse az+b 2 =l. Ans. (12+()2=.
x y 1 A xx 1 yy,
b) To the hyperbola iii-b 2 =l. ns. (i2-/ii=l.
138 Derivative and DiUerential
21 I. Find the equations of the tangent and normal to the Witch of Agnes!
sa
y= 4a+x' at the point where x=2a. Ans. The tangent ts x+2y=4a; the
.
normal is y=2x-3a.
212. Show that the normal to the curve 3y = 6x -5x drawn to the point
(I,+) coor~inate
M passes through the
213. Show that the tangent to the curve (
M (a, b) is -; f=2. +
origin.
~ r fr
+ ( =2 at the poinf
214. Find the equation of that tangent to the parabola, y 2 =20x, which
forms an angle of 45 with the x-axis. Ans. y=x+5 [at the point (5, 10)].
215. Find the equations of those tangents to the circle x 2 +y'=52, which
are parallel to the straight line 2x+3y=6. Ans. 2x+3y26=0.
216. Find the equations of those tangents to the hyperbola 4x 2 ~9y 2 =36,
which are perpendicular to the straight line 2y + 5x = 10. A ns. There are no
such tangents.
217. Show that the segment (lying between the coordinate axes) of the
tangent to the hyperbola xy=m is divided into two by the point of tangency.
218. Prove that the segment (between the coordinate axes) of a tangent

to the asteroid x + y =a is of constant length._
219. At what angle a do the curves y =a" and y = b" intersect? Ans.
In a-In b
tan a
l+lnalnb'
220. Find the lengths of the subtangent, subnormal, tangent and normal
of the cycloid x=a(8-sin8), y=a(l-cos8) at the point at which 8=-i-.
Ans. sr=a; sN=a; T=a Y2; N=a 2.
221. Find the quantities sr, sN, T and N for the hypocycloid x=4acos t,
. t
y=4asin 1 t. Ans. sr::;-4asin 2 tcost; sN=-4asm t; T=4asin't; N=
cos
= 4a sin t tan t.
Miscellaneous Problems
sin x I
Find the derivatives of the following functions: 222. y- - - - - X
-2 cos x 2
x lrHan ( 1t - x ). Ans. y' = - \ - . 223. y =arc sin_.!_. Ans. y' = ~ .

4 2 cos x x Jxl x -1
2
, COSX 2
224. y =arc sin (sin x). Ans. y =I cos xI 225. Y= V a _ b' X
xarctan( -v:~~tan ~) (a>O, b>O). Ans.-y'=a+blcosx' 226.y=JxJ.
Ans. y' =I; I' 227. u=arc sin y 1-x. Ans. y' =-,:I y /_x2.
228. From the formulas for the volume and surface of a sphere,
t1 = ._ :n:r and s = 4:n:r 2

3
it follows that : ; = s. Explain the geometric significance of this result. Find
a similar relationship between the area of .a--circle and the length of the cir
cumference. '-
229. In a triangle ABC, the side a is expressed in terms of the other two
sides b, c and the angle A between them by the formula
a= Y b +c"-2bc cos A.
2
For b and c constant, side a is a function of the angle A. Show that

:~ = ha, where ha is the altitude of the triangle corresponding to the base a.
Interpret this result geometrically.
230. Using the differential concept, determine the origin of the approxi
mate formulas
Ya+b'::::::: a+ 2~, V a'+b:::::::a+.......
3a 2
where I b I is a number small compared with a.

231. The period of oscillation of a pendulum is computed by the formula
T=:n: v-i
In calculating the period T, how will the error be affected by an error of I%
in the measurement of: 1) the length of the pendulum I; 2) the acceleration
of gravity g? Ans. 1) ::::::: 1/2%; 2) :::::::1/2% .
. 232. The tractrix has the property that for any point of it, the segment
of the tangent T remains constant in length. Prove this on the basis of:
1) the equation of the tractrix in the form
-- a a- Y a-y
x=Jfa-y+- In
2
Y (a>O);
a+ a-y
2) the parametric equations of the curve
x=a (In tan f+cos.t),
y=a sin t.
233. Prove that the function y = C,ex + c.e-X satisfies the equation
y + 3y' +2y =0 (here C1 and C2 are constants).
234. Putting y=ex sin x, z=ex cos x prove the equalities y" =2z, z"=- 2y.
235. Prove that the function y=sln(marcslnx) satisfies the equation
+
(1 -x 2 ) y -xy' my = 0.
2
.!!.... d y ( dt ).
236. Prove that if (a+ bx)e x =x, then x dx= x d~ -y .
CHAPTER IV
SOME THEOREMS ON DIFFERENTIABLE FUNCTIONS
SEC. I. A THEOREM ON THE ROOTS OF A DERIVATIVE (ROLLE'S

THEOREM)
Rolle's Theorem. If a function f (x) is continuous on an interval

[a, b] and. is di{jerentiable at all interior points of this interval,
and vanishes [f(a)=f(b)=O] at the end points x=a and x~b,
then inside [a, b] there exists at least one point x=c, a<c<b,
at which the derivative f' (x) vanishes, that is, f' (c)= 0. *) .
Proof. Since the function f (x) is continuous on the interval [a, b],
it has a maximum M and a minimum m on this interval.
If M = m the function f (x) is constant, which means that for
all values of x it has a constant value f (x) = m. But then at any
point of the interval f' (x) = 0, and the theorem is proved.
Suppose M =1= m. Then at least one of these numbers is not equal
to zero.
For the sake of definiteness, let us assume that M > 0 and
that the function takes on its maximum value at x=c, so that
f (c)= M. Let it be noted that, here, c is not equal either to a
or to b, since it is given that f (a)= 0, f(b) = 0. Since f (c) is the
maximum value of the function, f(c+~x)-f(c)~O, both when
~x > 0 and when ~x < 0. Whence it follows that
f(c+tJ.x)-f(c) &0 when ~x>O, (1 ')

tJ.x .....,
f(c+ L'lx)-f (c) ::::;, 0 when ~X< 0. (I")

j.X ,_..-
Since it is given in the theore~ that the derivative at x=c

exists, we get, upon passing to th'e limit as ~x--+ 0,
lim f(c+L'lx)-f(c) =f' (c),;;;O when ~x>O;
t\X-+ 0 tJ.X
lim f (c+ L'lx)-f (c) = f' (c)~ 0 when ~x < 0.

t\x-+ o i'lX
But the relations f' (c),;;; 0 and f' (c)~ 0 are compatible only if
f' {c)= 0.
Consequently, there is a point c inside the interval [a, b]
at which the derivative f' (x) is equal to zero.
I
*) The number c is called the root of the function cp (x) if cp (c)= 0.

A Theorem on the Roots of a Derivative (Rolle's Theorem) 141
The theorem about the roots of a derivative has a simple geo-

metric interpretation: if a. continuous curve, which at each point
has a tangent, intersects the x-axis at points with abscissas a and
b, then on this curve there will be at least one point with abs-
cissa c, a < c < b, at which the tangent is parallel to the x-axis.
y
!J
f{a)=
-=f(b}
0 a c1
Fig. 91. Fig. 92.
Note I. The theorem that has just been proved also holds for
a differentiable function such that cioes not vnnish at the end points
of the interval (a, b), but takes on equal values f(a)=f(b) (Fig. 91).
The proof in this case in exactly the same as before.
Note 2. If the function f(x) is such that the derivative does not
exist at all points within the interval [a, bJ, the assertion of the
theorem may prove erroneous (in this case there might not be a
point c in the interval (a, bJ, at which the derivative f' (x)
vanishes).
For example, the function
y=f(x)= 1- X2 V-
(Fig. 92) is continuous on the interval (-1, I J and vanishes at
the end points of the interval, yet the derivative
, <x>=-
3
v2-
X
within the interval does not vanish. This is because there is a
point x = 0 inside the interval at which the derivative does not
exist (becomes infinite).
The graph shown in Fig. 93 is another y
instance of a function whose derivative
does not vanish in the interval (0, 2].
The conditions of the Rolle theorem are
not fulfilled for this function either,
because at the point x =I the function has 0
no derivative. Fig. 93.
142 Some Theorems on Dif}erentiable Functions
SEC. 2. A THEOREM ON FINITE INCREMENTS (LAGRANGE'S THEOREM)

Lagrange's Theorem. If a function f (x) is continuous on the in-
terval [a, b] and ditferentiable at all interior points of this interval,
there will be, within [a, b], at least one point c, a<c< b,
such that
f(b)-f(a)=f'(c)(b-a). (1)
Proof. Let us denote by Q the number f(b)-f (a)

b-a
Q= f(b)-f(a) (2)
b-a '
and let us consider the auxiliary function F (x) defined by the
equation
F (x) = f (x)-f (a)- (x-a) Q. (3)
What is the geometric signi fica nee of the function F (x)? First
write the equation of the chord AB (Fig. 94), taking into account
that its slope is f(b;=~(a)=Q and that it passes through the
y point (a, f (a)):
y-f (a)= Q (x-a);
whence
y= f (a) +Q (x-a).
f(b)But F(x) = f(x)-[f(a) + Q(x-a)].
Thus, for each value of x, F (x) is
equal to the difference of the ordinates
of the curve y = f (x) and the chord
a c l( . y = f (a)+ Q (x- a) for points with
X the same abscissa.
Fig. 94. It will be readily seen that F (x)
. is continuous on the interval [a, b];
is differ~ntiable within this interval, and vanishes at the
end points of the interval; in other words, F(a)=O, F(b)=O.
Hence, the Rolle theorem is applicable to the function F (x). By
this theorem, there exists within the interval a point x = c such
that
F' (c)= 0.
But
F' (x) = f' (x)-Q.
And so
F' (c)=f' (c)-Q=O,
A Theorem on the Ratio of the Increments of Two Functions 143
when.ce
Q= f' (c).
Substituting the value of Q in (2), we get
ftbl-f(a) = f' (c) (1 ')
b-a ' '
whence follows formula (1) directly. The theorem is thus proved.

See Fig. 94 for an explanation of the geometric significance of
the Lagrange theorem. From the figure it is immediately clear that
the quantity f(b~=~(a) is the tangent of the angle of inclination !X
of the chord passing through the points A and 8 of the graph
with abscissas a and b.
On the other hand, f' (c) is the tangent of the angle of inclination
of the tangent line to the curve at the point with abscissa c. Thus,
the geometric significance of (1 ') or its equivalent (1) consists in
the following: if at all points of the arc AB there is a tangent
line, then there will be, on this arc, a point C between A and 8
at which the tangent is parallel to the chord connecting points A
and B.
Now note the following. Since the value of c satisfies the
condition a<c< b, it follows that c-a < b-a, or
c-a = 8 (b -a),
where 8 is a certain number between 0 and 1, that is,
0<8<1.
But then
c=a+8(b-a),
and formula (1) may be written as follows:
f(b)-f(a)=(b-a)f'[a+8(b-a)], 0<8<1. (1")
SEC. 3. A THEOREM ON THE RATIO OF THE INCREMENTS OF TWO

FUNCTIONS (CAUCHY'S THEOREM)
Cauchy's Theorem. If { (x) and <p (x) are two functions continuous
on the interval [a, b1 and differentiable within it, and <p' (x) does
not vanish anywhere inside the interval, there will be found, in
[a, b], some point x = c, a< c < b, such that
f(b)-f{a) f' (c)
q> (b)-q> (a)= q>' (c)'
(I)
144 Some Theorems on Differentiable Functions
Proof. Let us define the number Q by the equation

Q = f (b) - f (a) ( 2)
q>(b)-q>(a)
It will be noted that cp (b)-cp (a) =1= 0, since otherwise tp (b) would
be equal to tp (a), and then, by the Rolle theorem, the derivative
cp' (x) would vanish in the interval; but this contradicts the state-
ment of the theorem.
Let us construct an auxiliary function
F (x) = f (x)- f (a)- Q [cp (x) -cp (a)].
It is obvious that F (a)= 0 and F (b)= 0 (this follows from the
definition of the function F (x) and the defi_nition of the number Q).
Noting that the function F (x) satisfies all the hypotheses of the
Rolle theorem on the interval [a, b], we conclude that there exists
between a and b a value X=C (a<c<b) such that F'(c)=O.
But F' (x) = f' (x)-- Qcp' (x), hence
F' (c)= f' (c)- Qq>' (c)= 0,
whence
I' (c)
Q = cp' (c).
Substituting the value of Q into (2) we get (1).
Note. The Cauchy theorem cannot be proved (as it might appear
at first glance) by applying the Lagrange theorem to the nume~ator
and denominator of the fraction
{(b)-f(a)
<p(b)-cp (a)
Indeed, in this case we would (after cancelling out b-a) get the
formula
f(b)-{(a) f'(c 1)
cp (b)-cp (a)= cp' (c 2 )
In which a< C1 < b, a< c, <b. But since, generally, C1 + c,, the
result obtained obviously does not yet yield the Cauchy theorem.
SEC. 4. THE LIMIT OF A RATIO OF TWO INFINITESIMALS

(EVALUATION OF INDETERMINATE FORMS OF THE TYPE :)
Let the functions f (x) and cp (x), on a certain interv a! [a, b],
satisfy the Cauchy theorem and vanish at the point x =a of this
interval; f (a)= 0 and cp (a)= 0.
The Limit of a Ratio of, Two Infinitely Large Quantities 145
The ratio ; ~;~ is not defined for x =a, but has a very definite
meaning for the values x=1=a. Hence, we can rais~ the question
of searching lor the limit of this ratio as x - a. Evaluating limits
of this type is usually known as evaluating indeterminate forms
0
of the type 0
We have already encountered such problems, for instance when
considering the limit lim sinx and when finding derivatives of ele-
x-.. o X
mentary fund ions. For x = 0, the expression sin

X
x is meaningless;
the fund ion F (x),; sinX x is not defined for x = 0, but we have seen ,
that the lim it of the expression sin x as x- 0 exists and is equal
X .
to unity.
L'Hospital's Theorem (Rule). Let the functions f (x) and <p (x),
in some inter.val, satisfy the Cauchy theorem and vanish at some
point x =a: f (a)= <p (a)= 0; then, if the ratio ;, ~~) has a limit
as x---.. a, there also exists lim f (x) . and
x_,.a <p(x)'
lim f (x) = lim l', (x)

.t-+a<p(x) x_,.a<p(x)
Proof. On the interval [a, PI take some point x=/=a. Applying

the Cauchy formula we have
f (x)-f, (a) I'm
<p (x)-<p (a)= cp' m
where ~ Iies between a and x. But it is given that f (a)= <p (a)= 0,
and so
f, (x) f' m (I)
cp (x) = cp' (~)
If x---.. a, then ~-+ a also, since ~ Iies between x and a. And
if lim 1: =A, then lim ': ((~)) exists and is equal to A. Whence
((x)
x-+acp X) ;-+all'':> -
it is clear that
lim f (x) = lim
"-+ a cp (x)
(~) = lim t;
(x) = A ,
~ .... a c:p (S)
t;
. .r-+ a c:p (x)
and, finally,
lim l (x) = lim 1: (x)
x-+a c:p(x) K-+a c:p (x)
Note l. The theorem holds also for the case when the functions
f (x) or cp (x) are not defined for x =a, but
lim f (x) = 0, lim cp (x) = 0.
x-+a x-+a
In order to reduce this case to the earlier considered case, we

redefine the functions f (x) and cp (x) at the point x =a so that
they become continuous at the point a. To do this, it is sufficient
to put
f (a)= lim f (x) = 0; cp (a)= lim cp (x) = 0,
x-+a X-+a
since it is obvious that the limit of the ratio f ((x)) as x --+a does
(j) X .
not depend on_ whether the functions f (x) and cp (x) are defined
at x=a. 1
Note 2. If f'(a)=cp'(a)=O and the clerivatives f'(x) and cp'(x)

satisfy the conditions that were imposed by the theorem on the
functions f(x) and cp (x), then applying the L'Hospital rule to the
ra t .to f'--,--(
(x)
) , we arnve a t th e formu 1a 11m f'--;-:--(
(x) 1 f' (x)
) = 1m -;;--() , an
d
cp X X-+acp X x-+acp X
so forth.
Example I.
lim sin 5x = lim (sin 5x)' = lim 5 cos 5x = ~.
X-+0 3X X-+0 (3x)' X-+0 3 3
Example 2.
lim ln(l+x)= lim I+x=_!_=l.

X-+0 X X-+0 I 1
Example 3 .
. e-'"-e-x-2x . e-'"+e--'"-2 . e-'"-e-x . ex+e-x 2
l1m = hm =!1m = hm - - - =-=2.
x-+o x-sinx .t-+o 1-cosx x-+o slnx x-+o cosx I
Here, we had to apply the L'Hospital rule three times because the ratios
ol the first, second and third derivatives at x=O yield the indeterminate
0
form
0.
Note 3. The L'Hospital rule is also applicable if
lim f (x) = 0 and lim cp (x) = 0.
X-+ oo X-> OD
Indeed, putting x=.!..,

z
we see that z--.0 as X-+oo and
therefore
lim f
z -+0
(.!..)
z
=0, lim cp
-+,0
(.!..)
z
= 0
The Limit of a Ratio of Two Infinitely Large Quantities 147
. f(
Applyi[lg- the L'Hospital rule to the ratw
+)) we find
1 1
'P\ z
lim f (x) = lim f ( ~) = lim f' ( ~) ( - ~) =
x-. oo 'P (x) z-. o qJ ( ~) z-. o qJ' ( +) (_ ~)
r(_!_)
=lim
z -. o qJ' (
z
+)
= lim
x _. a>
f',(x) ,
qJ (x)
which is what we wanted to prove.

Example 4.
.
Stn -
k k COS -
k( - -1\)
Urn _ _
x = lim x x lim k cos!!.._= k.
X-+0> I X->"' I X-+"' X
X
SEC. 5. THE LIMIT OF A RATIO OF TWO INFINITELY LARGE QUANTITIES

(EVALUATION OF INDETERMINATE FORMS OF THE TYPE : )
Let us now consider the question of the limit of a ratio of two

functions f (x) and cp (x) approaching infinity as x-+ a (or as x - oo ).
Theorem. Let the functions f (x) and cp (x) be con/ inuous and
ditJerentiable for all x =1= a in the neighbourhopd of the point a:
the derivative cp' (x) does not vanish; furl her, let
lim f (x) = oo, lim cp (x) = oo
x-.a X-+ a
and let there be a limit
. f' (x)
hm -,--()=A. (1)
x-a'P X
Then there is a limit lim I (x) and

x -a (jl (X)
lim f (x) = lim

x-> a 'P (X) JC-+ a
r,'P (x)(x) = A. (2)
Proof. In the given neighbourhood of the point a, take two
points a and x such that a<x<a (or a>x>a). By Cauchy's
theorem we have
f (x)- I (a) f' (c) .
c:p(x}-qJ(a) =(jl'(c)'
(3)
148 Some Theorems on Di(Jerentiable Functions
where a< c < x. We transform the left side of (3) as follows:

_ f(a)
1
f (x)-f (a) f (x) f (x)
q>(x)-q> (a.)= q>(x) _q>(a.)
(4)
1
' (c) = (x) = <p' (c)
1
_t (a)
f (x)
From the condition (1) it follows that for an arbitrarily small
e > 0, a may. be chosen so close to a that for all x = c where
a< c <a, the following inequality will be fulfilled:
I~: ~~~ -A I< 8
or
f' (c)
< -,-< ' + 8.
lj1 c> < A
A- 8 (6)
Let us further consider the fraction
1- (a.)
lim 'P (x) =1
x-+a 1-f(a.)
f(x)
and, consequently, for the earlier chosen 8 >0 (for x sufficiently

close to a) we will have
1- q> (a.)
1- q>(x) <8
_ f (a)
1
f (x)
The Limit of a Ratio of Two Infinitely Large Quantities 149
or
1-<p(a)
1- e < <p (x) < l +e. (7)
_f(a)
1
f (x)
Multiplying together the appropriate terms of inequalities (6) and
(7), we get
1- )<(A + e)"(l +e).
Since e is an arbitrarily small number for x sufficiently close to a:

it follows from the latter inequalities that
lim f (x) =A
x .... a'P(X)
or, by (1},
which is what had to be proved.

Note t. If in premise (l) A= oo, that is,
. f' (x)
hm ---,--() = oo,
x .... a'P X
then equality (2) holds in this case as well. Indeed, from the
preceding expression it follows that
. <p' (x)
I tm - ,( )=0.
x .... a 1 x
Then by the theorem just proved

. <p(x) . <p'(x)
Itm -f()= 1tm -,-) =0,
x-.a X x .... a 1 (x
whence
lim f (x) = oo,
x-. a <p(x)
!50 Some Theorems on Differentiable Functions
Note 2. The theorem just proved is readily extended to the

case where x-..oo. If limf(x)=oo, limcp(x)=oo and lim f'(x)
X -+ rD X -+ <sJ X-+ OD !p' (X)
exists, then
lim f (x) =lim f' (x) (8)
X-+rD (jl(X) X-+OD cp' (X)
The proof is performed by replacing x = _!_,

z
as was done under
similar conditions in the case of the indeterminate form ~ (see
Sec. 4, Note 3).
Example I.
. e" . (e")' . e"
hm -=lim-)' =hm y=oo.
X -+ ao X X -+ ct:1 lX X ...,.. ao
Note 3. Once again note that formulas (2) and (8) hold only if
the limit on the right (finite or infinite) exists. It may happen
that the limit on the left exists while there is. no limit on the
right. To illustrate, let it be required to find
. x+slnx
I tm --x-
X-+ rD
This limit exists and is equal to 1. Indeed,

I .tm - - -x= 1.tm (I
x+sin +slnx)
- = I.
X-+ X X-+oo X
But the ratio of derivatives

(x +sin x)' _ I +cos x _
(x)' - -
I + COSX
1
as X-+ oo does not approach any limit, it oscillates between 0

and 2.
Example 2.
lim ax:
+ b =lim 2ax = !!_ :
x .... oo ex -d x .... ,., 2cx c
Example 3.
. tan x I' cos x I' I cos 3x I' I 23 cos 3x sin 3x

11m - - = 1 m - - = 1 m - - - = 1m--..,.,.---~-
n tan 3x n 3 n 3 cos x n 3 2 cos x sin x
x-+z x .... .- cos3x x-+2 x-+-.
The Limit of a Ratio of Two Infinitely Large Quantities 15!
Example 4.
. X I' I 0
I m
x-+ oo
-x= m -x=.
e x-+c:o e
Generally, for any integral n>O.
. xn I' nxn-1 I' n (n-I) ... I 0
I m -:;;=1m -x-== 1m " = .
x-+-e e X-+w e x-+oo e
The other indeterminate forms reduce to the foregoing cases.

These forms may be written symbolically as follows:
a) Ooo, b) 0, c) oo 0 , d) 100 , e) oo-oo. They have the
following meaning.
a) Let limf(x)=O; limcp(x)=oo; it is required to find
x -+-a x-+a
lim [f (x) rp (x)].

x-+a
This indeterminate form is of the type 0 oo.

If the required expression is rewritten ;:ts follows:
lim [f (x) <:p (x)] = lim f <;>

X-+a x-..a _
q> (x)
or in the form
lim [f (x) <:p (x)] =lim
x-+-a X-+a
q> t>
f (x)
then as x - a we obtain the indeterminate form ~ or :
Example 5.
lnx . 7 . xn
limxnlnx=lim - = h m - = - h m -=0.
x-+o x-+o I x-.o n .t-+o n
7 -X'f~Ti
b) Let
limf(x) =0, limcp(x)=O;
x-+a .x-+a
it is required to find
Jim [f (x)]Cjl <x>
or, as we say, to evaluate the indeterminate form 0.
Putting
take logarithms of both sides of the equality:

lny=cp(x) [lnf(x)].
As x-+ a we obtain (on the right) the indeterminate form 0 oo,
Finding lim lny, it is easy to get limy. Indeed, 'by virtue of the
x~a x~a
continuity of the logarithmic function, lim In y =In limy and if

x~a X-+a
lnlimy=b, it is obvious that limy=eb. If, in particular, b_:. +oo
x~a x~a
or - oo, then we will have limy= + oo or 0, respectively.
Example 6. It is required to find lim xx. Putting y=r we find In limY=

X-+0
=lim In y =lim In (xx) =lim (x In x);
I
llm(xln~)=lim lnlx=lim xi =-limx=O,
X-+O x~o x-.o x-.o
X - x2
consequently, In lim y=O, whence lim y=e 0 = I, or
lim xx =I.
X->0
'The technique is similar for fin.diug 11mits in other cases.
SEC. 6. TAYLOR'S FORMULA
Let us assume that the function y = f (x) has a II the derivatives

up to the (n + l)th order, inclusive, in some interval containing
the point x=a. Let us find a polynomial y=Pn(x) of degree not
above n, the value of which at x =a is equal to the value of the
function f (x) at this point, and the values of its derivatives up
to the nth order at. x =a are equal to the values of the
corresponding derivatives of the function f (x) at this point:
P n (a)= f (a), P~ (a)= f' (a), P~ (a)= f" (a), ... , P<:> (a)= f<n> (a). (I)
It is natural to expect that, in a certain sense, such a polynomial
is "close" to the function f (x).
Let us look for this polynomial in the form of a polynomial in
degrees of (x-a) with undetermined coefficients:
Pn (x) =Co+ C1 (x-a) -t-C 2 (x-a) 2 -t-C, (x-a)' +
+ ... +en (x-a)n. (2)
We difine the undetermined coefficients C 1 , C2 , , Cn so that
they will satisfy conditions (1).
Taylor's Formula !53
Let us first find the derivatives of P,. (x):

p~ (x) =C 1+ 2c. (x-a) + 3C (x-a) + ... +nC,. (x-:~r-t,}
3
P."~x~ . ~c:-:~.2~ 3 ~X.~a! -~: . . ~.n~n.~1!~".(~-~~) . .'. . (3)

p~nl(x)= n(n-1) ... 21C,..
Substituting, into the left and right sides of (2) and (3), the
value of. a in place of x and replacing, by equalities (1), P,.(a)
by I (a), P~ (a)= f' (a), etc., we get
f(a)=Co,
f' (a) =CJ'
f"(a)=21C 2 ,
f"'(a)=321C 3,
f'", (a)= n (n -1)(n -2) ... 2 1C,.,

whence we find
Co=f(a), C 1 =((a), C.=~f"(a),
c --123
1
- 1- I"' (a,) ... , c,.-12
- ...I n f(rl) (
a.) (4)
Substituting into (2) the values 1

, of C c. C,. that have been
found, we. get the required polynomial:
P11 (x) = I (a) +- (x-a) f"' (a)+ ...

2 1
x-af'( a)+ (x-a)
l T / " ( a) +T-2.3
1-

+ 12
(x-a)" f<"l (a)
... n (5)
Designate by R,. (x) the difference of the values of the given
function f (x) and of the constructed polynomial P,. (x) (Fig. 95):
R,. (x) = f (x)- P,. (x),
whence
I (x) = P,. (x) + R,. (x)
or, in expanded form,
I (x) =f (a)+ x~a f' (a)+ (x;a) f" (a)+ .. .
, (x-a)"f<n) ( )
... ,--n-L- a+ R,.x.
( ) (6)
!54 Some Theorems on Differ~ntiable Functions
y Rn (x)
is called the remainder. For
those values of x, for which the
remainder Rn (x) is small, the
polynomial Pn (x) yields an approxi-
mate representation of the function
f (x).
Thus, formula (6) enables one
to replace the function y = f (x) by
the polynomial y=Pn(x) to an
appropriate degree of accuracy
equal to the value of the remainder
0 a X X Rn (x).
Our next problem is to evaluate
Fig. 95.
the quantity Rn (x) for various
values of x.
Let us write the remainder in the form
(x-a)n+t
Rn (x) = (n+ 1)1 Q (x), (7)
where Q (x) is a certain function to be defined, and accordingly
rewrite (6):
f (x) = f (a)+ x 1a f' (a)+ (x -;,a)2 f" (a)+ ...
(x-a}n (nl (x-a)n+1
.. +-nl-f (a)+ (n+1)1 Q(x). (6')
For fixed x and a, the function Q (x) has a definite value; denote
it by Q.
Let us further examine the auxiliary function of t (t lying
between a and x):
F (t) = t (x)-f (t)- X-; t f' (t)- (x-;lt)l r (t)- ...
_(x-t}nf<nl(t) (x-t)n+tQ
" nl . (n+l}l '
where Q has the value defined by the relationship (6'); here we

consider a and x to be definite numbers.
We find the derivative f' (t):
F' (t) = -,t' (t) + f' (t)- X 1 t f" (t) + 2 (x2-;-t) r (t)-
+ (x-W- f'n) (/) +n (x-W- r' (t)-
1 1
f'" (t) +
2
_(x -t}
:.!1 (n-1}1 nl
_(x-tt r+tl(t)+(n+l}(x-t}nQ
nl (n+ 1)1 '
Taylor's Formula !55
or, on cancelling,
F' (t) = - (x-:/)n t<n+t) (f)+ (x n/t Q. (8)
Thus, the function F (f) has a derivative at all points t lying

near the point with abscissa a.
It will further be noted that, on the basis of (6'),
F (x) = 0, F (a)= 0.
Therefore, the Rolle theorem is applicable to the function F(t)
and, consequently, there exists a value t = ~ lying between a and
x such that F' (~) = 0. Whence, on the basis of relation (8), we
get
- (x-l;)n r+'l (~) + (x-st Q = 0
nl nl '
and from this
Substituting this expression into (7), we get

n+t
R (x)== (x-a ) rn+'l(~)
n (n+1)l
This is the so-called Lagrange form of the remainder.
Since ~ lies between x and a, it may be represented in the
form*)
~=a-t-6(x-a)
where 6 is a number lying between 0 and 1, that is, O<b<l;
then the formula of the remainder takes the form
R (x) = (x-a)n+' r+l) [a+ 6 (x-a)]
n (n+ 1)1 '
The following formula
f (x) = f (a)+ x~a f' (a)+ (x-;;a) f" (a)+ ...
2
... + (x:;a)n rn) (a)+<~;;~~~~ f1n+t) [a+ e (x-a)] (9)

is called Taylor's formula of the function f (x).
If in the Taylor formula we put a= 0 we will have
f (x) = f (0) + ~ f' (0) + ~ f" (0) + ...
n n+t
.. +~I f<n) (0) + (nx+ 1)1 r+tl(8x) (10)
where 8 lies between 0 and l. This special case ol the Tayfor
formula is sometimes called Maclaurin's formula.
") See end of Sec. 2 of this chapter.
-
SEC. 7. EXPANSION OF THE FUNCTIONS ex, SIN x, AND COS x

IN A TAYLOR SERIES
1. Expansion of the function f (x) = e".
Finding the successive derivatives of f (x), we have
f(x)=e", f(O)= I,
f' (x) = e", f' (0) = I,
f'm (x) = e", {'" 1 (0) = I.

Substituting the expressions obtained into formula (10), Sec. 6,
we get
" x x1 x1 ex
e = I + T + 21 + 3f + ... + nix" -1- (nxn+t
+ I )I e ' 0 < e< 1.
If 1x 1 ~ I, then, taking n = 8, we obtain an evaluation of the
remainder:
R, < 91I 3;
For x = I we get a formula that permits approximating the number e:
I I
e=I-1-1+ 21 + 31 + ... + 81I i
evaluating to the fifth decimal place, we have
e=2.71827.
Here there are four significant digits, -since the error does not exceed
3
, or 0.00001.
91
Observe that no matter what x is, the remainder
xn+l ax
Rn=(n+!)le --0 as n-+-oo.
Indeed, since e< 1' the guan tity e8x for fixed X is bounded (it
x
Is less than e" for > 0, and less than I for x < 0).
We shall prove that, no matter what the fixed number .t,
xn+t
(n+l)l--.0 as n--oo.
Indeed,
I(nx:+l rl = 1-T ~ ' i ~ ' n~ 1!
1
If x is a fixed number, there will be a positive integer N such

that
Expansion of Functions e", sin x, and cos x in a Taylor Series 157
We, introduce the notation' I~I = q; then, noting that 0 < q < 1~
we can write (for n=N+I. N+2. N+3, etc.):
,=I TI ~ I ~ I I Nx 1 11 ~ I :I ~ I) ~I I<n
X X X X xN-I n- 'V+ 2
<-2'3 ... N-1qq .. _.q=(N-1)1q' I
for the reason that
I~ I= IN: I<
q; 1 q; ln:11<q.
xN-1
But (N- 1) 1 is a constant quantity; that is to say, it is independy
ent of n, while qn-N+ approaches zero as n - oo. And so
. xn+l
hm(
n-+-oo n
+
1) 1 =0. (1)
ij XR+I
Consequently, R n (x) = e "(n + 1) 1 also approaches zero as n approaches
infinity.
From the foregoing it follows that for any x (if a sufficient num
ber of terms is taken) we can evaluate e" to any degree of accuracy.
2. Expansion of the function f (x) =sin x.
We find the successive derivatives of f(x) = sin x:
I (x) =sin x, I <O> = o,
f' (X) =COS X= sin (X + ~ ) , f' (0) = l,
{" (x) = -- sin x = sin ( x + 2 ~ ) , f" (0) = 0,
f"' (~) = - cosx_; sin ( x +3 ~), f'" (0) = -1.
f1v (x) = sin x = sin ( x +4 ; ) , IIV (0) = 0,
r (X)= sin (X + n ; ), rn (0) = sin n ~ ,
f'n+II(X)=Sin(x+(n+l) ~ ) [In+ II()= Sin [ + (n + 1) ~ J o

Substituting the values obtained into (10), Sec. 6, we get an

expansion of the function f (x) =sin x by the Taylor formula:
xn . :rt xn + t [ :rt ]
... + nT stn n 2 + (n + l) 1 sm ~ + (n + 1) 2 .
Since Jsin [~+(n+I)-i] ~~1, ~e have}~~ Rn(x)=O for all
values of x.
Let us apply the formula obtained for an approximate evaluation
of sin 20. Put n = 3, thus restricting ourselves to the first two
terms of the expansion:
. 20 = sm
. 9:rt 1 3
sm ~
:rt
9-31. ( 9:rt )
= 0.343 .
Evaluate the error, which is equal to the remainder:
I Ral =I (~ r -&sin(~ +2n)l ~ ( ~ r-A-=0.0006 < 0.001.

!/ I
I
.J
I
!y=x-Lx3+..Lx5
6 120
!
-1
2 X
7 y=sinx
/
/ y=x-fxJ ~
/ I
/ I
I
I
IJ
Fig. 96.
Hence, the error is less than 0.001, and so sin 20~ = 0.343 to three
places of decimals.
Fig. 96 shows the graphs of the function f (x) =sin x and the
Exercises on Chapter IV 159
first three approximations: S, (X)= x; S 2 (X)= X-~; Sa (X)= X-
-31=51
3. Expansion of the function f (x) =cos x.
Finding the values of the successive derivatives for x = 0 of the
function f (x) =cos x and substituting them into the Maclaurin
formula, we get the expansion:
2
cos x = I - x21 + x xn ( n )
41 - ... + ni cos n 2 +
+(nx:+1,)1cos [+(n+l) ~],
11; I< lxl.
Here again, lim Rn (x) = 0 for all .values of x.
n-+"'
Exercises on Chapter IV
Verify the truth of Rolle's theorem for the functions: 1. y = x 1 -3x + 2 on

the interval (1, 2]. 2. y=x 1 +5x 2 -6x on the interval (0, 11. 3. y=(x-1)
(x-2) (x-3) on the interval (1, 3j. 4. y=sln 2 x on the intetval (0, nj.
5. The function f(x)=4x 1 +x 2 -4x-1 has roots I and -1. Find the root
of the derivative f' (x) mentioned in Rolle's theorem.
6. .Verify that between the roots of the function y =
root of its derivative.
V
x 2 -5x +6 lies the

7. Verify the truth of Rolle's theorem for the function y= cos x on the
Interval [- : , + :] .
8. The function y = 1- V ;< becomes zero at the end points of the inter-
val ( -1, 11. Make it clear that the derivative of this function does not
vanish anywhere in the interval ( -1, 1). Explain why Rolle's theorem is not
applicable here.
9. Form Lagrange's formula for tbe function y=sinx on the interval
(x 1 , x1 j. Ans. sinx 2 -sinx 1 =(x 2 -x1 )cosc, x,<c<x1 .
10. Verify the truth of the Lagrange formula for the function y= 2x-x1
on the interval (0, IJ.
II. At what point is the tangent to the curve y=xn parallel to the chord
from point M, (0, 0) to M 2 (a, an)? Ans. At the point with abscissa
a
c=n-v;;.
12. At what point is the tangent to the curve y=lnx parallel to the chord
linking the points M 1 (1, 0) and M1 (e, I)? Ans. At the point with abscissa
c=e-1.
. Applying the Lagrange theorem, prove the inequalities: 13. e" > 1 x. +
14. In (1 +x) < x (x > 0). 15. bn-an < nbn- (b -2) for b >a. 16. arc
tanx<x.
17. Write the Cauchy formula for the functions f(x)=x, cp(x)=x on the
interval [I, 2) and find c. Ans. c = ~.
Evaluate the following limits: 18. lim x - 1 . Ans. ~. 19. lim e" -e-"
X-+1 xn-1 tl x-+o sin x
. tan x-x x I
Ans. 2. 20. 1tm . Ans. 2. 21. lim e - Ans. -2.
x... o x-slnx :X ... u cosx-1
sinx
22. lim Y . Ans. There is no limit (Y2 as x -++O, - Y2 as
x-+o I-eos x
X-+ -0). 23. lim In sin x . Ans. I 24. lim a"- u" . Ans. In .!:
X-+-
n (n-2x) 1 B X-+0 X b '
2
. x-arc sinx lim sin x-sin a.

25. 1tm . , . Ans. 26 Ans. cos a.
X-+ 0 Sin X x-+a x-a
27. lim eY +siny-1. Ans. 2. 28. lime" sinx-x. Ans I 29. lim 3x-l
y-+0 Jn(l+y) X-+0 3x 1 +x' a X-+oo2x+5.
Ans. 3 30. lim lnnx (where n > 0). Ans. 0. 31. lim
In (I +I_)
x . Ans. I.
2 x-+:J:> x x-+oo arc tan x
In x+l
82. lim _ _x_ Ans. -I. 33. lim ..1L . Ans. 0 for a > 0; oo for a< 0.
X-+""ln ~1 fl-++ooeay
X
84. lim e: + e-:. -

x-+ + "" e -- e-
Ans. l. 35. lim 1~ sin 3x
x-.o n sin x
36 lim In tan 7x
Ans. 1.
x-+o In tan 2x
Ans. l. 37. lim In (x-l)-x. Ans. 0. 38. lim (1 -x) tan ~. Ans. 2
2 1t
x-+l tan~ X-+l
2x
39. lim [
c-+1
2
x1 -i-x-1 ]
1
Ans. - - .
2
40. Urn
x-+1
[-~--~].
lnx In x
Ans. -1.
1
41. lim (sec p-tan <p). Ans. .
n 2
Q)-+2
'I I
43. lim x cot 2 x. Ans. -1 44. lim xeXi. Ans. oo. 45. lim x 1 -x. Ans.
X-+0 2 .t-+o .c~t e
46. lim
1-+oo
{liz. Ans. 1. 47. lim
X-+0
(J.-)
X
tan x. Ans. I. 48. lim (1
.C-+CD
+.!:X )
I n
--x
Ans. ea. 49. lim (cot x)lnx. Ans. 50. lim (cos x) 1
Ans. 1.
X-+;t> e x-..-n
a
Exercises on Chapter IV 161
51. lim sIn.
1
Ans. ~.52. :n:x)tan~
lim ( tan- 2 I
Ans. - .
q>-+o
( <p Ve x-+t 4 e
53. Expand, in powers of x-2, the polynomial x4 -5x 3 +5x 2 +x+2.
Ans. 2-7 (x-2)-(x-2) 2 +3 (x-2) 3 +(x-2) 4
54. Expand, in powers of x+l, the polynomial x5 +2x 4 -x 2 +x+l.
Atzs. (x+ 1) 2 +2 (x + 1)3 -3 (x + 1)4 +(x+ 1)5
55. Write Taylor's formula for the function y= Yx when a= I, 11=3.
,r- x-1 I (x-1) 2 I (x-1) 3 3 (x-1)' 5
Atzs. r x = 1 -.
1
+-
2 - T2. 4 + ---r:2-3 . 8 - -4-1- 16 [I +tlx
X(x-1)] , 0 < 9 < l.
56. Write the Maclaurin fOimula for the function y= Yl +x ~hen n=2.
I I x3
Ans. VT+X= I + 2 x- 8 x+ 5
, 0 < 9 < l.
2
16(1 +9x)
57. Using the results of the preceding exercise, evaluate the error of the
.r- 1 I
approximate equality r l+x~t+ x- x when x=0.2. Ans. Less than
2 8
2 -to
Determine the origin of the approximate equalities for small values of .ll
x 2 x
and evaluate the errors of these .equalities: 58. lncosx~----
2 12.
.. x 3 2x 5 x1 x
+ T5 . 60." arc sin x ~ x +
tan x ~x +
6 . arc tan x ~ x -
59. 61.
3 3
ex +e-x x x ,r-- x
~1+
2 + 24 . 63. ln(x+ r l-x )~x-:-3f 2
62.
2
Usirig. Taylor's formula, compute the limits of the following expressions:
X-Sln X
Ans. l. 65 . lim In' (I +x)-sin 2 x. Ans. 0.
64. lim - - - - - .1
X-+0 X
e - 1 - x -X- x -+O 1-e-x
2
66 . lim 2 (tan x-:ln x)-x 3 r. 67. lim [x-x ln(t+..!..)] AtlS . .!..
2
Ans.
.t-+ o X X-+0 . X 2
68. lim (..!...-cot
X-+0
2
x).
X X
Ans . ..!.:. . 69. lim
3 X-+0
(_!_ ~cot x).
X
2
Ans. 3...
3
6 -3388
CHAPTER V
INVESTIGATING THE BEHAVIOUR OF FUNCTIONS
SEC. I. STATEMENT OF THE PROBLEM
A study of the quantitative aspect of natural phenomena leads

to the establishment and study of functional relations between the
variables involved. If such a functional relationship can be expres-
sed analytically, that is, in the form of one or more formulas, we
are then in a position to investigate it wtth the tools of mathema-
tical analysis. For instance, a study of the flight of a shell in empty
space yields a formula that gives the dependence of the range R
upon the angle of elevation a and the initial velocity v 0 :
v~ sin 2a
R= g
(g is the acreler at ion of gravity).
With this formula we can determine at what angle a the range
R will be greatest, or least, and what the conditions must be for
the range to increa:>e as the angle a is increased, etc.
Let us consider another instance. Studies of oscillations of a load
on a spring (of a tank or automobile) yielded a formula showing
how the deviation y of the load from a position of equilibrium
depends on the time t:
+
y = e-kt (A cos wt B sin wt).
The quantities k, A, B, w that enter into this formula have a very
definite significance for a given oscillatory system (they depend
upon the elasticity of the spring, the load, etc .. but do not change
with time t) and for this reason are considered constant.
On the basis of this formula we can find out at what values of
t the deviation y will increase with increasing t, how the maximum
deviation varies as a function of time, for what values of t we
observe these maximum deviations, for what values of t we obtain
maximum velocities of motion of the load, and a .number of other
things.
All these questions are embraced by the concept "investigating
the behaviour of a function". It is obviously very difficult to de-
termine all these questions by calculating the values of a function
at specific points (like we did in Chapter II). The purpose of this
chapter is to establish more general techniques for investigating
the behaviour of functions.
Increase and Decrease of a -Function 163
SEC. 2. INCREASE AND DECREASE OF A FUNCTION
In Sec. 6 of Ch. I we gave a difinition of an increasing and a

decreasing function. We will now apply the concept of the
derivative to investigate the increase and decrease of a function.
Theorem. If a function f(x), which has a derivative on the in-
terval [a, b], increases on this interval, then its derivative on
[a, b] is not negative, that is, f' (x) ~ 0.
2) If the function f(x) is continuous on the interval [a, b] and
is, ditJerentiable on (a, b), where f' (x) > 0 for a < x < b, then
this function increases on the interval [a, b].
Proof. Let us first prove the first part of the theorem. Let f(x)
increase on the interval [a, b]. Increase the argument x by tu
and consider the relation
f (x+ ~x) - f (x) (1)
~X
Since f (x) is an increasing function,

f (x + ~x) > f (x) for ~x >0
and
f (x + ~x) < f (x) for ~x < 0.
In both cases
(2)
and consequently
lim {(x+M)-f(x) ~O
dx-o ~X
which means f' (x) ~ 0, which is what we set out to prove. [If we
had f' (x) < 0, then for sufficiently small values cif ~x. relation(!)
would be negative, but this would contradict relationship (2).]
Let us now prove the second part of the theorem. Let f' (x) > 0
for all values of x on the interval (a, b).
Let us consider any two values x, and X 2 , x, < X 2 , on the
interval [a, b].
By the Lagrange theorem on finite increments we have
f (x.)- f (x.) = f' () (x 2 -X,), x, < < X2
It is given that f' () > 0, hence f (x 2 ) - f (x,) > 0, and this means
that f (x) is an increasing function.
There is a similar theorem for a decreasing (differentiable)
function as well, namely: ..
If f (x) decreases on an interval [a, b], then f' (x)..;;;;; 0 ori this
interval. If f' (x) < 0 on (a, b), then f (x) . decreases on [a, {J]. [Of
6*
164 lnuestigatirig the Behuuiou; of Functions
course, we again assume that the fUl~dion is continuous at all

points of [a, b] and is differentiable everywhere on (a, b).]
Note. The foregoing theorem expresses the following geometric
fact. If on an interval [a, b) a function f (x) increases, then the
tangent to the curve y = f (x) at each point on this interval forms
y
X 0 X
(a) (b)
Fig. 97.
an acute angle cp with the x-axis or (at .certain points) is horizon-

tal; the tangent of this angle is not negative: f' (x) =tan q>;;;;:. 0
(Fig. 97, a). If the function f(x) decreases on the interval [a, b],
then the angle of inclination of the tangent forms an obtuse angle
(or, at some points, the tangent is horizontal); the tangent of
this angle is not positive (Fig. 97, b). We can illustrate the second
part of the theorem in similar fashion. This theorem permits
judging the increase or decrease of a function by the sign of its
derivative.
Example. Determine the domains of increase and decrease of the function
y=x.
Solution. The derivative is equal to
y' =4x 3 ;
for x >0 we have y' >0 and the function increases;
for. x <0 we have y' <0 and the function decreases (Fig. 98).
SEC. 3. MAXIMA AND MINIMA OF fUNCTIONS
Definition of a maximum. A function f (x) has a maximum at

x, if the value of the function f (x) at the point x, is
th~ 'poi,ht
greater than its values at all points of a certain interval.. contain-
ing the point x 1 In other words, the function f (x) has a maxi-
Maxima and Minima of Functions 165
mum when x = x, if f (x, + ~x) < f (xJ for any ~x (positive and
negative) that are sufficiently small in absolute value.*)
For example, the function y = f (x), whose graph is given in
Fig. 99; has a maximum at x=x,.
Definition of a minimum. A function f (x) has a minimum at
x =X 2 if
f (X 2 + ~x) > f (x,)
for any ~x (positive and negative) that are sufficiently small in
absolute value (Fig. 99).
For instance, the function y=x 4 considered at the end of the
preceding section (see Fig. 98) has a minimum for x = 0, since
y = 0 when x = 0 and y > 0 for all other values of x.
y
y
0 Q Xt Xz Xj X
Fig. 98. Fig. 99.
In connection with the definitions of maximum and minimum,

note the following.
l. A function defined on an interval can reach maximum and
minimum values only for values of x that lie within the given
interval.
2. One should not think that the maximum and minimum of a
function are its respective largest and smallest values over a given
interval: at a point of maximum, a function has the largest value
only in comparison with those values that it has at all points
sufficiently close to the point of maximum, and the smallest value
'") This definition is sometimes formulated as- follows: the function f (x)
has a maximum at x, if it is possible to find a neighbourhood (a, ~) of
x1 (a< x1 < Ill. such that for all points of this. neighbourhood diiTerent !rom
x 1 the inequality f (x) < f (x 1) is fulfilled.
166 Investigating the Behaviour of Functions
only in comparison with those that it has at all points sufficiently

close to the minimum point.
To illustrate, take Fig. 100, which shows a fundion defined on
the interval [a, b], which
at x = x1 and x = x. has a maximum;
at x = x 2 and x = x, has a minimum,
but the minimum of the function at x = x, is greater than the
maximum of the function at x=x 1 At x=b, the value of the
function is greater than any maxi-
mum of the function on the
interval under consideration.
The generic terms for maxima
and minima of. a function are
extremum (pl. extrema) or extreme
values of the function.
To some extent, the extrema of
a function and their positions on
the interval [a, b] characterise the
0 a x1 x variation of the function versus
changes in the argument.
Below we give a method for
finding extrema.
Theorem t. (A necessary condition for the existence of an
extremum). If at the point x=x, a ditferentiable function y=f(x)
has a maximum or minimum, its derivative vanishes at this point:
f'(x,)=O.
Proof. For definiteness, let us assume that at the point x = x.
the !unction has a maximum. Then, for sufficiently small (in
absolute value) increments 6.x (6.x=;I=O} we have
f (X 1 + 6.x) < f (X 1 ),
that is,
f (x, + 6.x)- f (x,) < 0.
But in this case the sign of the ratio
/(x,+M)-/(x,)
/j.X
is deteiminect by the sign of Ax, namely:

/(x,+Ml-/,x,) >0 when 6.x<0
/j.X
1 (x, + M)- 1(x,) < 0 when 6.x > 0,

/j.X
By the definition of a derivative we have

f' (x.) =lim f (x, + Ax)-f (x,)
t>x-+o Ax
If f (x,) has a derivative at x = x,, the limit on the right is
independent of how tu approaches zero (remaining positive or
negative).
But if ~x-- 0 and remains negative, then
f' (x,) ~0.
But if ~x - 0 and remains positive, then
f' (x,) ~ 0.
Since f' (x,) is a definite number that is independent of the way
in which /).x approaches zero, the latter two inequalities arc
compatible only if
f' (x,) = 0.
The proof is similar for the case of a minimum of a function.
Corresponding to this theorem is the following obviou.s geometric
fact: if at points of maximum and minimum, a function f (x) has
a derivative, the tangent line to the curve
y=f(x) at these points is parallel to the x-axis. y
Indeed, from the fact that f' (x,) =tan rp = 0,
where cp is the angle between the tangent 1ine
and the x-axis, it follows that cp = 0 (Fig. 99).
From Theorem I it follows straightway that
if for all considered values of the argument x
the function f(x) has a derivative, then it can
have an extremum (maximum or minimum) only
at those values for which the derivative vanishes.
The converse does not hold: it cannot be said
that there definitely exists a maximum or mini-
mum for every value at which the derivative
vanishes. For instance, in Fig. 99 we have a
function for which the derivative at x = x.
vanishes (the tangent line is horizontal}, yet the Fig. 101.
function at this point is. neither a maximum
nor a minimum.
In exactly the same way, the functiony=x" (Fig. 101) atx=O
has a derivative equal to zero:
(y').t=o = (3r)x=o = 0,
but at this point the function has neither a maximum nor a
minimum. Indeed, no matter how close the point x is to 0, we
168 ln.uestigating the Be_hauiour of' Functions
will always have

x 5 <0 when x<O
and
xs > 0 when x > 0.
We have investigated the case when a function has a derivative
at all points on some closed interval. Now what about those points
at which there is no derivative? The following examples will
show that at these points there can only
be a maximum or a minimum, but there
may not be either one or the other.
Example 1. The . function y =I xI has no
derivative at the point x=O (at this point
1he curve does not have a definite tangent
x line), but the function has a minimum at this
point. y=O when x=O, whereas for any other
Fig. 102. point x different from zero, we have y > 0
(Fig. 102).
. 2
-S/2
Example 2. The function y = (1-x ) has no derivative at x=O, since
I I
-x-
2
!1' = -(1-x'/) becomes inQnite at x=O, but the function has a
maximum at this point: {(0}=1, f(x) < 1 at x different from zero (Fig. 103).
Example 3. The function y = Vx
has no derivative at x = 0 (y' ----. oo
as x__. 0). At this point the function does not have either a maximum or a
minimum: f (0) =0; f (x) < 0 for x < 0; f (x) > 0 for x >
0 (Fig. 104).
y
1
-1
Fig. 103. Fig. 104.
Thus, a function can have an extremum only in two cases:

either at points where the derivative exists and is zero; or at
points where the derivative does not exist.
It must be noted that if the derivative does not exist at some
point (but exists at close-lying points), then at this point the
derivative is discontinuous.
The values of the argument for which the derivative vanishes
or is discontinuous are called cri-tical points or critical values.
From what has been said it follows that not for every critical
value does a function have a maximum or a minimum. However,
if at some point the function attains a maximum or a minimum,
this point is definitely critical. And so to find the extrema of a
function do as follows: find all the critical points, arid then,
investigating separately each critical point, find out whether the
function will have a maximum or a minimum at this point, or
whether there will be neither maximum nor minimum.
Investigations of functions at critical points is based on the
following theorem.
Theorem 2. (Sufficient conditions for the existence of an extre-
mum). Let there be a function f ( x) continuous on some interval
containing a critical point x, and differentiable at all points of
this interval (with the exception, possibly, of the point x, itself).
If in moving from left to right through this point the derivative
changes sign from plus to minus, then at x = x, the function has
a maximum. But if in moving through the point. x, from left to
right the derivative changes sign from minus to plus, the function
has a minimum at this point.
And so
. { f' (x) >0 when x<x,,
tf a)
f' (x) <0 when x>x,
then at x, the function has a maximum;
if b { f' (x) < 0 when x < x,,
) f' (x) >0 when x>x.,
:I .
then at x 1 the function has a minimum. Note here that the con-
ditions a) or b) .must be fulfilled for all values of x that are
sufficiently close to x., that is, at all points of some sufficiently
small neighbourhood of the critical point X 1
Proof. Let us first assume that the dei'ivative changes sign from
plus to minus, . in other words, that for all x sufficiently close to
x, we have
f' (x) > 0 when x < x,,
f' (x)<O when x>x,.
Applying the Lagrange theorem to the differe_nce f (x) -f (x.)
we have
f (x) -f (x.) = f' m(x_- x,)
where ~ is a point lying between x and X 1
I) Let x < x,; then

<x,, f'(s)>O. f'(s)(x-x,)<O
and, consequently,
f (x)- f (x,) < 0,
or
f (x) < f (x,). (1)
2) Let x>x,; then

s>x., f'(S)<O, t'<S)(x-x,)<O
and, consequently,
f(x)-f (x.) <0
or
f (x) < f (x,). (2)
The relations (1) and (2) show that for all values of x suffici-
ently close to x, the values of the function are less than those
at x,. Hence, the function f (x) has a maximum at the point x,.
The second part of the theorem on the sufficient condition for
a minimum is proved in similar fashion.
11
Fig. 105 illustrates the meaning
~ of Theorem 2.
At x = x., let there be f' (x,) = 0
and let the following inequalities
be fulfilled for all x sufficiently
close to x,:
f' (x) >0 when x < x.,
x3 x f' (x)<O when x>x,.
Then when x < x, the tangent
to the curve forms with the
x-axis an acute angle, and the function increases, but when x > x,
the tangent forms with the x-axis an obtuse angle, and the func-
tion decreases; at x = x, the function passes from increasing to
decreasing, which means it has a maximum.
If at X 2 we have f' (x 2 ) =0 and for all values of. x sufficiently
close to x 1 the following inequalities are fulfilled:
't' (x) < 0 when x<x 2,
f' (x) >0 when x >X 2,

Testing a Differentiable Function for Maximum and Minimum 171
then at x < x. the tangent to the curve forms with the x-axis an
obtuse angle, the function decreases, and at x >x. the tangent
to the curve forms an acute angle, and the function increases.
At x = x. the function passes from decreasing to increasing, which
means it has a minimum.
If at x = x. we have f' (x.) = 0 and for all values of x sufficiently
close to x. the following inequalities are fulfilled:
f' (x) >0 when x<x
f' (x) > 0 when x >x .
then the function increases both for x < x. and for x > x . Th_ere-
fore, at x = x, the function has neither a maximum nor a mint
mum. Such is the case with the function y = x at x = 0.
Indeed, the derivative y' = 3x, hence,
(y')x=o = 0,
(y')x<o > 0,
(y')x>o >0,
and this means that at x = 0 the function has neither a maximum
nor a minimum (see above, Fig. 101).
SEC. 4. TESTING A DIFFERENTIABLE FUNCTION

FOR MAXIMUM AND MINIMUM WITH A FIRST DERIVATIVE
The preceding section permits us to formulate a rule for testing

a differentiable function, y=f(x), for maximum and minimum:
1. Find the first derivative of the function, i.e., f' (x).
2. Find the critical values of the argument x; to do this:
a) equate the first derivative to zero and find the real roots of
the equation f' (x) =0 obtained;
b) find the values of x at which the derivative f' (x) become~
discontinuous.
3. Investigate the sign of the derivative on the left and right
of the critical point. Since the sign of the derivative remains
constant on the interval between two critical points, it is sufficient,
for investigating the sign of the derivative on the left and right
of, say, the critical point x. (Fig. 105), to determine the sign of
the derivative at the points a and ~ (X 1 < u < x., x. < ~ < x,,
where X 1 and x, are the closest critical points).
4. Evaluate the function f (x) for every critical value of the
argument.
172 lnv.esiigaring the Behaviour of .Functions
This gives us the following diagram of possible cases:
I Signs ol derivative f' (x) when passing

through cri\ical point x 1 : l Character of critical point
I
x <x, x=x,
I I l x>x, l
+ f'(x 1 )=0 - Maximum point
or is discontinuous
- f'(x 1 )=0 + Minimum point
or is discontinuous
+ f'(x,)=O + Neither maximum nor
or is discontinuous minimum (function in-
- f' (x,) =0 creases)
or is discontinuous - Neither maximum nor
minimum (function de-
creases)
I
Example I. Test the following function lor maxtmum .and minimum:
x 2
y=
3 -2x +3x+ l.
Solution. I) Find the first derivative:
2
y' =X -4x+3.
2) Find the real roots of the derivative:
x 2 -4x+3=0.
Consequently,
x 1 =I, X 2 =3.
The deriv alive is everywhere continuous and so there are. no other critical
~oints.
3) Investigate the critical values and record the results in Fig. 106.
Investigate the first critical point x 1 = l. Since l = (x-1) (x-3),
for x Q,
for x> I we have y'=(+)-(_:_)<0.
Thus, when passing (from left to right) through the value x, =I the deri-
vative changes sign from plus to minus. Hence, at x = 1 the function has a
maximum, namely '
7'
(y)x=1=3
Investigate lhe second critical point x 2 = 31

when x<3_ we have y'=(+)(_:_)<O,
when x> 3 we have y'=l+)(+J>O.
Testing a DiUerentiable Function for "Maximum and Minimum 173
Thus, when passing through the value x = 3 the derivative changes sign
from minus 1.o plus. Therefore, at x=3
the function has a minimum, namely:
(Yh=a = l.
This investigation yields 1.he graph of the function (Fig. 106).
Example 2. Test for maximum and minimum the function
Y=(x-1) vx~.
Solution. 1) Find the tirst derivative;

, _
y - 3~1 +2(x-J)_ 5x-2
X -
3Vx 3Vx
Fig. 106. Fig. 107.
2) Find the critical values of the argument: a) ftnd the points at which
th~ derivative vanishes:
5x-2
y =3 vx-=0.
b) find the points at which the derivative becomes discontinuous (in this
instance, it becomes infinite). Obviously, that point is
X2=0.
(It will be noted that for x 2 =0 the function is defined and continuous.)
There are no other critical points.
3) Investigate the character of the critical points obtained. Investigate
the point x1 = ~.., . Noting that
(y') 2 < 0, (y') .. > 0,
x< 6 x>-i'-
174 Investigating the- Behauiour of Functi'Ons
2
we conclude that at x = S the function has a minimum. The value of the
function at the minimum point is
(Y\=!_ =( i-l) v245=- ~ v~

Investigate the second critical point x =0. Noting that
(Y)x < o > 0, (y')x >o < 0
we conclude that at x=O the function has a maximum, and (y)x=o=O. The
graph of the investigated function is shown in Fig. 107.
SEC. 5. TESTING A FUNCTiON FOR MAXIMUM AND MINIMUM

WITH A SECOND DERIVATIVE
Let the derivative of the function y = f (x) vanish at x = x,; we
t
have f' (x,) = 0. Also, Jet the second derivative (x) exist and be
continuous in some neighbourhood of the point x,. Then the fol-
lowing theorem holds.
Theorem. Let f' ( x,) = 0; then at x = x, the function has a
maximum if f" (x,) < 0, and a minimum if f" (x,) > 0.
Proof. Let us first prove the first part of the theorem. Let
f' (x,) = 0 and f" (x,) < 0.
Since it is given that f" (x) is continuous in some small interval
about the point x = x .. there will obviously be some small closed
interval about the point x = x .. at all points of which the second
derivative f" (x) will be negative.
Since f" (x) is the first derivative of the first derivative, r
(x) =
= (f' (x))', it follows from the condition (t' (x))' < 0 that f' (x)
decreases on the closed interval containing x = x, (Sec. 2, Ch. V).
But f' (x,) = 0, and so on this interval we have f' (x) > 0 when
x < x" and when x > x, we have t' (x) < 0; in other words, the
derivative f' (x) changes sign from plus to minus when passing
through the point x = x,, and this means that at the point x, the
function f (x) has a maximum. The first part of the theorem is
proved.
The second part of the theorem is proved in similar fashion:
r
if (x,) > 0 then f" (x) > 0 at all points of some closed interval
about the point x .. but then on this interval f" (x) = (f' (x))' > 0
and, hence, f' (x) increases. Since f' (x,) = 0 the derivative f' (x)
changes sign from minus to plus when passing through the point
x,, i.e., the function f(x) has a minimum at X= X,.
If at the critical point f"{x,) = 0; then at this point there may
be either a maximum or a minimum or neither maximum nor
Testing a Function for Maximum and Minimum 175
minimum. In this case, investigate by the first method (see Sec. 4,

Ch. V).
The scheme for investigating extrema with a second derivative
is shown in the following table.
Character ol critical point
0 Maximum point
0
0
+
0
Minimum point
Unknown
Example 1. Examine tile following function for maximum and minimum

y= 2 sin x +cos 2x.
Solution. Since the function is periodic with a period of 2n, it is suffi.
cient to investigate the function in the interval [0, 21t].
l) Find the derivative:
y' = 2 cos x-2 sin 2x= 2 (cos x-2 sin x cos x) =2 cosx (l-2 sinx).
2) Find the critical values of the argument:
2 cos x (l-2 sin x) =0,
:rt :t 5:rt 31t
xJ=6; x~=T; x.=6; x,=T
3) Find the second derivative:
y" = -2 sin x-4 cos 2x.
4) Investigate the character of each critical point:
(y") =-2..!..-4 . ..!..~-3 <0.

x=~ 2 2
8
Hence, at the point x 1 = ~ we have a maximum:

I I 3
(y) -2-+---
x=~- ~ 2-2

Further,
And so at the point x1 = ~ we have a minimum:

(H) n = 2 l - 1=I.
A=-

5rc
At x,= we have
6
I . I
(y")x=~ =-~2-42=- 3 < 0.

5
Thus, at x1 = : the function has a maximum:
(y)x,='~ =2~+ ~ = ~.
Finally,
(y") an=-2(-1)-4(-1)=6>0.
x==-

Consequen tl y, at x4 = 3rc we h ave a mm1mum:
. .
2
(y) an=2(-1)-l=-3.
x=-
z
The graph of the function under investigation Is shown In Fig. 108.
2 y y=2sinX+COS2X
Fig. 108.
'.
The following examples will show that if at a certain point x=x1
we have f' (x,)= 0 and f" (x 1 ) = 0, then at this point the function
f (x) can have either a maximum or a minimum or neither.
Example 2. Test 'the following function for maximum and minimum:
y=l-x4
Solution. I) Find the critical points:
y' = - 4x 1 , -4x 1 =0, x=O.
2) Determine the sign of the second derivative at x =01
y" = - 12x2 , (.!i")x=o=O.
Testing a Function for~ Maximum and Minimum 177
It is thus impossible here to determine the character of the critical point

by means of the sign of the second derivative.
3) Investigate the character of the critical point by the first method (re~:
Sec. 4., Ch. V):
(y')x < o > 0, (y')x > o < 0.
Consequently, at x=O the function has a maximum, namely
(Y)x=o= I.
The graph of this function is given in Fig. 109.
lj
y
. -
0
Fig. 109. Fif!.. 110.
Example 3. Test for maximum and minimum the function

y=xo.
Solution. By the second method we find
!) y' =6x 5 ,
y' =6x 6 =0, x=O;
2) y"=30x, (y")x=o=O.
Thus, the second method does not yield anything. Resorting to the first
method we get
(y')x<o < 0, (y')x > o > 0. {/
Therefore, at x=O the function has a minimum
(Fig. 110). y-(x-1)3
Example 4. Test for maximum and minimum
the function
y=(x-J).
Solution. Second method:
y'=3(x-l),' 3(x-1)2=0, x=l; X
y" =6 (x-1), uf)x=l =0.
Thus, the second method does not yield an answer.
By the first method we get
(y')x < 1 > 0 . (!l')x > 1 > 0.
Consequently, at X= 1 the function does not
have either a maximum or a minimum (Fig. Ill). Fig. 111.
SEC. 6. MAXIMA AND MINIMA OF A FUNCTION -ON AN INTERVAL

Let the function y=f(x) be continuous on the interval [a, b).
Then the function on this interval will have a maximum (see Sec. 10,
Ch. II). We will assume that on the given interval the function
f(x) has a finite number of critical points. If the maximum is
reached within the interval [a, b). it is obvious that this value will
be one of the maxima of the function (if there are several maxima),
namely, the greatest maximum. But it may happen that the max-
imum value is reached at one of the end points of the interval.
To summarise, then, on the interval [a, b) the function reaches
its greatest value either at one of the end points of the interval,
or at such an interior point as is the maximum point.
The same may be said about the minimum value of the function:
it is attained either at one of the end points of the. interval or at
y~x3-Jx+J an i_nterior P?int such that the latter is the
mtmmum pomt.
y From the foregoing we get the following
rule: if it is required to find the maximum of
a continuous function on an interval [a, b). do
the following:
I) Find all maxima of the function on the
interval.
2) Determine the values of the function at
the end points of the interval; that is, eval-
.-3 ~ -1 0 uate f (a) and f (b).
3) Of all the values of the function obtained
choose the greatest; it will be the maxi-
mum value of the function on the interval.
The minimum value of a function on an
interval is found in similar fashion.
Example. Determine the maximum and minimum
of the function y=x- 3x+3 on the interval
[- 3, i].
Solution. I) Find the maxima and minima of the
function on the interval [- 3, { J
1
y' = 3x'-3, ~~x'-3=0, x1 =I, x2 = - I,

y" = 6x, (y")xzl = 6 > 0.
Thus, at x=l \here is a minimum:
(!J)x=l =I.
c).'----~-15. Further,
f1g. 112:
Applying the Theol'y of Maxima and Minima 179
And so at X=- 1 we have a maximum:

(Y)x=-1 =5.
2) Determine the value of the function at the end points of the interval!
15
(y)x=.!.. =s (Y)x=-=-15.
2
Thus, the greatest value of this function on the interval [ - 3, {] is

(Y)x=-1 =5,
and the smallest value is
(Y)x=-=- 15.
The graph of the function is shown in Fig. 112.
SEC. 7. APPLYING THE THEORY OF MAXIMA AND MINIMA

OF FUNCTIONS TO THE SOLUTION OF PROBLEMS
The theory of maxima and minima is applied in the solution of

many problems of geometry, mechanics, and so forth. Let us
examine a few.
Problem 1. The range R = OA (Fig. 113) of a shell (in empty
space) fired with an initial velocity v 0 from a gun inclined to the
horizon at an angle cp, is determined
by the formula
u~ sin 2<p
R= g
(g is the acceleration of gravity).

Determine the angle cp at which the
-range R will be a maximum for a Fig. 113.
given initial velocity V 0
Solution. The quantity R is a function of the variable angle cp.
Test this function for a maximum on the interval 0 ~ cp ~ i:
dR 2u~ cos 2<p 2u! cos 2<p
d<p = g g 0; crit ica1 value cp = -T ;
dR = _ 4u! sin 2<p (dR) __ 4u!
d<p' g ' dcp ll -
(j)=-
g < 0.
4
Hence, for the value cp = Tthe function R has a maximum

The values of the function R at the end points of the interval

[ 0, ~] are
(R)'!'=o= 0, (R) n
(ji=-
= 0.
3
Thus, the maximum obtained is the sought-for greatest value of R.

Problem 2. What should the dimensions be of a cylinder so that
for a given volume v its total surface S is a minimum?
Solution. Denoting by r the radius of the base of the cylinder
and by h the altitude, we have
S = 2nr 2 + 2nrh.
Since the volume of the cylinder is given, for a given r the
quantity h is determined by the formula
v=nr 2h,
whence
v
h=-a
n.r
Substituting this expression of h into the formula for S, we have
S = 2nr 2 +2nr -n.rv2
or
S = 2 ( nr 2 + -7-) .
Here, v is given, so . we have represented S as a function of a
single independent variable r.
Find the minimum value of this function on the interval 0 < r < oo:
dr = 2 (2nr-~)
!!!i. r2 '
dZS)
(-d ~ . = 2 (2n + 32v) > 0.
. r r=r 1
. r r=r 1
.
Thus, at the point r = r 1 the function S has a minimum. Notic-

ing that limS=oo and limS=oo; that is, that as r approaches
r-Jo-0 r-+CD
zero or infinity the surface S increases without bound, we arrive
at the conclusion that at r = r 1 the function S has a minimum.
Testing a Function for Maximum and Minlmum 181
But if r= v 2~ then
~/v
h = _.!!.__
:nr = 2 V 2n= 2r.
Therefore, for the total surface S of a cylinder to be a mtmmum
for a given volume v, the altitude of the cylinder must be equal
to its diameter.
SEC. 8. TESTING A FUNCTION FOR MAXIMUM AND MINIMUM

BY MEANS OF TAYLOR'S FORMULA
In Sec. 5, Ch. V, it was noted that if at a certain point x=a

we have f' (a)= 0 and r
(a)= 0, then at this point there may be
either a maximum or a minimum or neither. And it was noted that
in this instance the problem is solved by investigating by the first
method; in other words, by testing the sign the first derivative on
the left and on the right of the point X=a.
Now we will show that it is possible in this case to investigate
by means of Taylor's formula, which was derived in Sec. 6, Ch. IV.
For greater generality, we assume that not only .f" (x), but also
all derivatives up to the nth order inclusive of the functions f (x)
vanish at x =a:
f' (a)= f" (a)= ... = !' 111 (a)= 0 (1}
and
f<n+IJ (a)+O.
Further, assume that {(x) has continuous derivatives up to the

(n+ I) st order inclusive in the neighbourhood of the point x=a.
Write the Taylor formula for {(x), taking account bf equality (1):
(x )"+I
f(x)=f(a)+ -a rn+ll() (2)
(n+l)l '
where s is a number that lies between a and x.
Since f"+ 11 (x) is continuous in the neighbourhood of the point
a and f"+ 11 (a)+ 0, there will be a small positive number h such
that for any x that satisfies the inequality 1x~a 1 < h, there will
be f"+ 1> (x) =1= 0. And if /'"+ 11 (a)> 0, then at all points of the interval
(a- h, a+ h) We Will have flll+IJ (X)> 0; if r+ll (a)< 0, then at all
points of this interval we will have fl"+ 11 (x) < 0.
. Rewrite formula (2) in the form
f(x)-f(a)= (x-a)n+l fl"+l> (~) (2')
(n + 1)1 --
and consider various special cases.
182 Investigating the BehauiouF of Functions
Case t. n is odd.
a) Let f<"+' 1 (a)< 0. Then there will be an interval (a -h, a+ h)
at all points of which the (n+ l)st derivative is negative. If xis
a point of this interval then likewise lies between a-hand a+ h
and, consequently, [<"+' 1 () < 0. Since n+ I is an even number,
(X- ar+ > 0 for X =i=' a, and therefore the right side of formula
(2') is negative.
Thus, for x=foa at all points of the interval (a-h, a+h) we have
f (x)-f (a)< 0,
and this means that at x= a the function has a maximum.
b) Let f<"+'' (a)> 0. Then we have [<"+' 1 (S) > 0 for a sufficiently
small value of h at all points x of the interval (a-h, a+ h). He:1ce,
the right side of formula (2') will be positive; in other words, for
x =fo a we will have the following at all points in the given interval:
f (x) -f (a)> 0,
and this means that at x =a the function has a minimum.
Case 2. n is even.
Then n+ 1 is odd and the quantity (x-at+ has different signs
for x<a and x>a.
If h is sufficiently small in absolute value, then the (n +I) st
derivative retains the same sign as at the point a at all points of
the interval (a-h, a+h). Thus, f(x)-f(a) has different signs for
x <a and x >a. But this means that there is neither maximum
nor minimum at x=a.
It will be noted that if f<"+' 1 (a)> 0 when n is even, then
f (x) < f (a) for x <a and f (x) > f (a) for x >a.
But if f<"+' 1 (a)<0 when n is even, then f(x)>f(a) for x<a
and f(x)<f(a) for x>a.
The results obtained may be formulated as follows.
If at x=a we have
f' (a)= f" (a)= ... = f<"' (a)= 0
and the first nonvanishing derivative r+'l (a) is a derivative of even
order, then at the point a
f (x) has a maximum if f1"+' 1 (a)< 0,
f (x) has a minimum if fl"+ ' 1 (a) > 0.
But if the first nonvanishing derivative [<"+''(a) is a derivative
of odd order, then the function has neither maximum nor minimum
at the point a. Here,
f (x) increases if f'n+' 1 (a)> 0,
f (x) decreases. if rn+l) (a)< 0.
Convexity and Concavity of a Curve 183
Example. Test the following function for maximum and minimum:

f (x) =X4 -4x -f-.6x 2 -4x-l- 1.
Solution. Let us find the critical values of the functiou
f' (x) =4x 3 -12x 2 + 12x-4 = 4 (x'-3x 2 + 3x-l).
From equation
4 (x 3 -3x 2 +3x-l)=0
we obtain the only critical point
x=l
(since this equation has only one real root).
Investigate the character of the critical point x= 1:
t." (x) = 12x2 -24x + 12=0 for x= I,
f'" (x)=24x-24=0 for x= I,
flV (i) = 24 > 0 for any x.
Consequently, for x= I the function f(x) has a minimum.
SEC. 9. CONVEXITY AND CONCAVITY OF A CURVE.

POl NTS OF INFLECTION
Let us consider, in a plane, the curve y=f(x), which is the

graph of a single-valued differentiable function f (x).
Definition 1. We say .that . a curve is convex upwards on the
interval (a, b) if all points of the curve lie below any tangent
to it on this interval.
We say that the curve is convex !I
downwards on the interval (b, c)
if all points of the curve lie above
any tangent to it on this interval.
We shall call a curve convex up,
a convex curve, and a curve convex
down, a concave curve.
Fig. 114 shows a curve convex
C X
on the interval (a, b) and concave
on the interval (b, c). Fig. 114.
An important characteristic of
the shape of a curve is its con-
vexity or concavity. This sect ion will be devoted to establishing
the characteristics by which, when investigating a function y= f (x),
one can judge of the convexity or concavity (direction of bulge)
on various intervals.
We shall prove the following theorem.
Theorem 1. If at all points of an interval (a, b) the second deriv-
ative of the function f (x) is negative, L e., 1" (x) < 0, the curve
y= f (x) on this interval is convex upwards (the curve is cotivex).
Proof. In the interval (a, b) take an arbitrary point x=x 0

(Fig. 114) and draw a tangent to the curve at the point with
abscissa x = X 0 The theorem will be proved provided we establish
that all the points of the curve on the interval (a, b) lie below
this tangent; that is, that the ordinate of any point of the curve
y=f(x) is less than the ordinate y of the tangent line for one and
the same value of x.
The equation of the curve is of the form
y= f (x). (1)
But the equation of the tangent to the curve at this point
x= X 0 is of the form
y- f (x 0
) = f' (X 0 ) (x---: X0 )
or
(2)
From equations (1) and (2) it follows that the difference of the
ordinates of the curve and the tangent for the same value of x is
y- Y= f (x)- f (X 0 )- f' (xo>,(x- X 0 )
Applying the Lagrange theorem totb~ difference f(x)-{(x 0 ),
we get
y-y=f' (c) (X-X )-f' (X 0 0 );(X-X 0 )
(where c lies between x 0 and x) or
y-y= [f' (c)-f' (X 0 )] (X-X 0 ).
We again apply the Lagrange theorem to the expression in the

square brackets; then
y- y = f" (c 1 ) (c- X 0 ) (X- X 0 ) (3)
(where c, lies between X 0 and c).

Let us first examine the case when x > x 0 In this case, x 0 <
<c<x; since
X-X 0 > 0, C-X 0 >0
and since, in addition, it is given that

f"(c,)<O,
it follows from equality (3) that y- y < 0.
Now let us consider the case when x < X 0 In this case x < c <
<C1 <x'0 and x-x 0 <0, c-x 0 < 0; and since il is given that
f" (c 1)<0, then it follows from (3) that

y-y<O.
We have thus proved that every point of the curve Iies below
the tangent to the curve, no matter what values x and X 0 have on
the interval (a, b). And this signifies that the curve is convex.
The theorem is proved.
The following theorem is proved in similar fashion.
Theorem 1'. If at all points of the interval ( b, c), the second
derivative of the function f(x) is positive, that is, f"(x)>O. then
the curve y=f(x) on this interval is convex downwards (the curve
is concave).
Note. The content of Theorems l and l' may be illustrated
geometrically. Consider the curve y = f (x), convex upwards on the
interval (a, b) (Fig. 115). The derivati,e f' (x) is equal to the
y
Fig. 115. Fig. 116.
tangent of the angle of inclination a of the tangent line at the

point with abscissa x, or f' (x) =tan a. For this reason, , f" (x) =
=[tana]~.If{"(x)<Ofor all xon the interval (a,b), this meansthat
tan a decreases with increa~ing x. It is geometncally obvious that if
tan a decreases with increasing x. then the corresponding curve
is convex. Theorem 1 is an analytic proof of this. fact.
Theorem l' is illustrated geometrically in similar .fashion (Fig.
116}.
Example I. Establish the intervals of convexity and concavity of a curve
represented by the equation
y=2-x1
Solution. The second derivative
y"=-2 <0
lo1 all values of x. Hence, the curve is everywhere convex upwards (Fig. 117).
Example 2. The curve is given by t-he equation

y=ex.
Since
y"=ex > 0
for all values of x, the curve is therefore everywhere concave (bulges, or is
convex, downwards) (Fig. 118).
Example 3. A curve is defined by the equation
y=xa.
Since
y"=6x,
y" < 0 for x < 0 and y">O for x > 0. Hence . .for x <0 the curve is convex
upwards, and for x > 0, convex down (Fig. 119).'
y
2
Fig. 117. Fig. 118. Fig. 119.
Definition 2. The point that separates the convex part of a

continuous curve from the concave part is called the point of
inflection of the curve.
In Figs. 119 and 120 the points 0 and B are points of inflection.
It is obvious that at the point of inflection the tangent cuts the
curve, because on one side the curve lies under the tangent and
on the other side, above it.
Let us now establish the sufficient conditions for a given point
of a curve to be a point of inflection.
Theorem 2. Let a curve be defined by the equation y=f(x). If
f" (a)= 0 or f" (a) does not exist and if the derivativP f" (x) changes
sign when passing through x=a, then the point of the curve with
abscissa x=a is the point of inflection.
Proof. I) Let f" (x)<O for x<a and f' (x)>O for x>a.
Then for x<a the curve is convrx up and for x >a, it is convex
down. Hence, the point A of the curve with abscissa x =a is the
point of inflection (Fig. 120).
2) If f'(x)>O lor xb, then for x b, it is convex up. Hence, the
point B of the curve with abscissa x = b is the point of inflection
(see Fig. 121).
y y
0 a x
Fig. 120. Fig. 121.
Example 4. Find the points of ir.ftection and determine the intervals o[

convexity and concavity of the cur\'e
y=e-x (Gaussian curve).
Solution. I) Find the first and second derivatives:
y' = -2xe-x',
lf=2e-x (2x 2-1).
2) The second derivative exists everywhere. Find the values of x for which
y"=O:
2
2e-x (2x2-J)=0,
I I
x.=- Y2 x2= y2
3) Investigate the values obtained:
I
for x <- .r- we have If> 0,
r 2
I
for x >- Y we have y" < 0;
2
the second derivative changes sign when passing through the point x1 Hence1
for x 1 = - ; , there is a point of inflection on the curve; its coordinates
2
are: ( - V2 , e- f),
1
For ;x< Y2 !I <0,
I
for
x> V2 y">U.
Thus, there is also a point of inflection on the curve for x1 = : i Its co-
2
ordinates are ( ;
2
, e- +). Incidentally, the existence of the second point
of inflection follows direct! y from the symmetry of the curve about the
y-axis.
4) From the foregoing it follows that
1
for - oo < x <- V the curve is concave;
2
1 1
for - V <x < V the curve Is convex;
2 2
for ; <x< oo the curve is concave.
2
5) From the expression of the first derivative
y' =-2xe-x
It follows that
for x<O y'>O, the function Increases;
for x > 0 y' < 0, the function decreases;
for x=O y' =0.
At this point the function has a maximum, .namely, Y==r l. Tl)e foregoing
analysis makes it easy to construct a graph of the curve (Fig. 122).
!I
Fig. 122.
Example 5. Test the curve y = x for points of inflection.

Solution. 1) Find the second derivative:
y'!= l2x.
2) Determine the points at which y"=O: 12x~=O; x=:O.
3) Investigate the value x=O obtained:
for x <0 y" > 0, the curve is 'concave;
for x>O y">O, the curve is concave.
Thus, the curve has no points of inflection (Fig. 123). '
Example 6. Investigate the following curve for pointS' of lnflectlom
y=(x-1) .
Asymptotes 189'
Solution. I) Find the first and second derivatives:

I -~ 2 _..!._
y' = (x-1) 3; y" = - g (x-1) s.
3
2) The second derivative does not vanish auywhere, but at x =I it does
not exist (y"= oo).
y
Fig. 123. Fig. 124.
3) Investigate the value x =I:

for x 0, the curve is concave;
for x> I if <0, the curve is convex.
Consequent! y. at x =I there -is a point of inflection (I, 0).
It will be noted that for x =I y' = oo; the curve at this point has a vel'
tical tangent (Fig. 124).
SEC. 10. ASYMPTOTES
Very frequently one has to investigate the shape of a curv!"

y = f (x) and, consequently, the type of variation of the correspond-
ing function in the case of an unlimited increase (in absolute
value) of the abscissa or ordinate of a variable point of the curve.
or of the abscissa and ordinate simultaneously. Here, an important
special case is when the curve under study approaches a given
line without bound as the variable point of the curve recedes to
infinity.*'
Definition. The straight lin~ A is called an asymptote to a curve,
if the distance fJ from the variable point M of the curve to this
straight line approaches zero as the point M recedes to infinity
(Figs. 125 and 126).
*' We say the variable point M moves along a curve to infinity if lhc
distance of the point [rom the origin increases without bound.
In future we shall differentiate between vertical asymptotes (paral-

lel to the axis of ordinates) and inclined asymptotes (not parallel
to the axis of ordinates).
Fig. 125. Fig. 126.
I. Vertical asymptotes.
From the definition of an asymptote it follows that
if lim f (x) = oo or lim f (x) = oo _or lim f (x) = oo,
x-+a+O x ... a-0 x ... a
then the straight line x=a is an asymptote to the curve y = f (x);

y and, conversely, if the. straight
line x =a is an asymptote, then
one of the foregoing equalities is
fulfilled.
Consequently, to find vertical
asymptotes one has to find values
of x =a such that when they are
X
approached by the function
y = f (x) the latter approaches
infinity. Then the straight line
x =a will be a vertical asymptote.
2
Example. l. The curve y = - - has a
x- 5
Fig, 127. vertical asymptote x = 5, since Y-+()
as x-+5 (Fig. 127).
Example 2. The curve y =tan x has an infinite number of v-ertical asymp-
totes:
. n 3:rt 5:~:
x=2; x=T; x=T; ...
This follows !rom the !act that tan x_,. () as x approaches the values
n 3:~: 5:~: :rt 3:~: 5:rt .
2 . 2 , 2 , ... , or - 2 , - , -
2 , . . . (F1g. 128).
2
Asymptotes 191
1 ~
Example 3. The curve y =ex has a vert ica I asymptote x = 0, si nee lime x =
X-++ 0
= oo (Fig. 129).
(
Fig. 128.
II. Inclined asymptotes.

Let the curve. y = f (x) have an inclined asymptote whose
equation is
(1)
Fig. 129. Fig. 130.
Determine the numbers k and b (Fig. 130). Let M (x, y) be a point

lying on the curve and N (x, y), a point lying on the asymptote.
The length of MP is equal to the distance from the point M to
192 lntestigating the Behaviour of Function,,
the asymptote. It is given that

lim MP=O. (2)
X-++ oo
Designating the angle of inclination of the asymptote to the x-axis

by qJ, we find from 6. NMP that
NM = MP.
cos lp
Since rp is a constant angle (not equal to ~), by virtue of tht!

foregoing equation
lim NM =0 (2')
X---++ oo
and, conversely, from (2') we get (2). But

NM = IQM- QN I= IY-Y I= if(x)-(kx+ b) I,
and (2') takes the form
lim [f(x)-kx-b]=O. {3)
X-++ CXI
To summarise: if the straight line (1) is an asymptote, then (3)

is fulfilled; and conversely, if, given constants k and b, equation
(3) is fulfilled, then the straight line y=kx+b is an asymptot~.
Let us now define k and b. Taking x outside the brackets in
(3), we get
lim x [ 1(x) - k - .!!..] = 0.
X-++CO . X X
Since x-.. + oo, the following equation must be fulfilled:

lim
X-++oo
l' (x)-
X
k-!!...J
.X
= 0.
For b constant, lim!:=o. l-Ienee,

X-t oo X
lim ['
.t-++oo
(xl
X
-k] = 0,
or
k= lim [(x). (4)
.t-++oo X
Knowing k, we find b from (3)1

b = lim [{ (x)-kx). (5)
A:-+ +00'
Asymptotes 193
+
Thus, if the straight line y = kx b is an asymptote, then k and
b may be found from (4) and (5). Conversely, if the limits (4) and (5)
exist, then (3) is fulfilled and the straight line y = kx + b is an
asymptote. If even one of the limits (4) or (5) does not exist, then
the curve does not have an y
asy rn ptote.
It should be noted that we
carried out our investigation as
applied to Fig. 130, as x-++ oo,
but all the arguments hold also
for the case X - + - oo.
Example 4. Find the asymptotes
of the curve
x2 -f-2x-l
y= X
Solution. 1) Look for vertical

asymptotes:
when x__..-6 y__,. oo; +
when x___,.-f-0 Y---+--oo. X X
Therefore, the straight line x,;, 0
is a vertical asymptote.
2) Look for inclined asymptotes:
k= lim JL= lim xz+2x-l =
X-+ co X X-+ oo X2
=lim
X-+ co
[t~2-~J
.'C X .
=1,
. Fig 131.
that is,
k=I,
b =lim [y-x)= lim [xz-f-2x-l_x] ~ lim [xz+2x-l-xz] =
X-+ co X-+ co X X-+ "' X
= X-+
lim [2-.!_],=2,
co X
or, finally,
b=2.
Therefore, the straight line
Y=X-f-2
is an inclined asymptote to the given curve.
To investigate the mutual posi lions. of a curve and an asymptote, let us
consider the difference of the ordinates of the curve and the asymptote fol'
one and the same value of x:
2
x -f-2x-l_(x-f- 2) = - _.!_.
X X
This difference is negative for x > 0, and positive for x < 0; and so for x>O
the curve lies below the asymptote, and for x<O, it lies above the asymptote
(Fig. 131).
7- 3388
Example 5. Find the asymptotes of the curve

y=e-" sin x +x.
Solution. I) It is obvious that there are no vertical asymptotes.
2) Look for inclined asymptotes:
k= lim .1!._= lim e-"sinx+x= lim [e-"slnx+l] =I,

X->+ co X X->+oo X X->+oo X
b= lim [e-"sinx+x-x]= lim e-"slnx=O.
X--t-+a:> X-++W
Hence, the straight line

y=x
Is an inclined asymptote as X-++ oo.
The given curve has noasymptote as x->--oo. Indeed, the limit lim .!L
x-.-O'J x
does not exist, since lL= e-x x sin x +I. (Here, the first term increases without
X X
bound as x .... - oo and, therefore it has no limit.)
SEC. 11. GENERAL PLAN FOR INVESTIGATING FUNCTIONS

AND CONSTRUCTING GRAPHS
The term "investigation of a function" usually implies the

finding of:
I) the natural domain of the function;
2) the discontinuities of the function;
3) the intervals of increase and decrease of the function;
4) the maximum point and the minimum point, and also the
maximal and minimal values of the functions;
5) the regions of convexity and concavity of the graph, and
points of inflection;
6) the asymptotes of the graph of the function.
The graph of the function is constructed on the basis of such
an investigation (it is sometimes wise to plot elements of the
graph in the very process of investigation).
Note 1. If the function under investigation y = f (x} is even,
that is, such that upon change of sign of the argument the value
of the function does not change, i. e., if
f (-x) = f (x),
then it is sufficient to investigate the function and construct its
graph for positive values of the argument that lie within the
domain of definition of the function. For negative values of the
argument, the graph of the function is constructed on the. grounds
that the graph of an even function is symmetric about the
ordinate axis.
General Plan for Investigating Functions and Constructing Graphs .195
Example I. The function y=x is even, since ( -x) 2 =x 2 (see Fig. 5).
Example 2. The function y =cos x is even, since cos ( -x) =cos x (see
Fig. 17).
/
Note 2. If the function y = f (x) is odd, that is, such that for
any change in the argument the function changes sign, i.e., if
f (-x) = -f (x),
then it is sufficient to investigate this function in the case of
positive values of the argument. The graph of an odd function is
symmetric about the origin.
Example 3. The function y =X 3 is odd, since ( -x) 1 = - X 1 (see Fig. 7).
Example 4. The function y =sin x is odd, since sin ( -x) =-sin x (see
Fig. 16).
Note 3. Since a knowledge of certain properties of a function

allows us to judge of the other properties, it is sometimes advi-
sable to choose the order of investigation on the basis of the
specific peculiarities of the given function. For example, if we
have found out that the given function is continuous and differen-
tiable and if we have found the maximum p_oint and the mini
mum point of this function, we have thus already determined
also the range of increase and decrease of the function.
Example 5. Investigate the function
X
y= I +x 2
and construct its graph.
Solution. I) The domain of the function is the interval -oo<x<oo. It will
straightway be noted that for x<O we have y<O. and for x>O we have y>O.
2) The function is everywhere continuous. .
3) Test the function for maximum and minimum, from the equation
1-x
t/ =(I +x)=O.
Find the critical points:
x1 = - l , x2 =1.
Investigate the character of the critical points:
for x, <-1 we have y' < 0;
for x 1 >-1 we have y' > 0.
Hence, at x = - 1 the function has a minimum:
!lmin=(Y)x,-t=-1.
And
for x 0;
for x > 1 we have y' < 0.
7
196 I noestigating the Behaviour of Functions
Hence, at x= I the function has a maximum:

Ymax=([l)x=t=l.
4) Determine the domain of increase and decrease of the function:
for - oo < x <-I we have y' < 0, the function decreases;
for - I < x 0, the function increases;
for I < x < oo we have t/ < 0, the function decreases.
5) Determine the domains of convexity and concavity of the curve and
the points of inflection: from the equality
,_2x(x2-3)_0
Y - (I +xz)a -
we get
X1 = - Jf3, X 2 =0, X~= Jf3.
Investigating y" as a function of x we find that
lor - oo < x <- Jl3 y" < 0, the curve is convex;
for - Jl3 < x<0 y" > 0, the curve is concave;
for 0 < x < Jl3 y" < 0, the curve is convex;
for Jl3 < x < co t/ > 0, the curve is .concave.
Thus, the point with coordinates X=- vi y=- ~3 is a point of
inflection; in exactly the same way, the points (0, 0) and ( Jl3, ~ 3 ) are
points of inflection.
6) Determine the asymptotes of the curve:
for x -+ +
oo y -+ 0,
forx-+-oo y-+0.
Consequently, the straight line y = 0 is the only inc! ined asymptote. The
curve has no vertical asymptotes because the function does not approach
infinity for a single finite value of x.
--------~I--L~o t
Fig. 132.
The graph of the curve under study is given in Fig. 132.

Example 6. Investigate the function
y= v~2a-X7
' 2 ---x--;1
and construct its graph.

Gtmeral Platt for Investigating Functiuns and Constructing Graphs 197
Solution. I) The function is defined for all values of x.

2) The function is everywhere continuous.
3) Test the function for maximum and minimum:
, 4ax-3x 2 4a-3x
y = 3 V(2ax 2 - x 3 ) 2 3 Vx (2a-x) 2
There is a derivative everywhere except for the points
x1 = 0 and x 2 = 2a.
Investigate the limiting values of the derivative as x -+~0 and
as x-+ +O:
. 4a-3x lim 4a-3x
1Im = - oo,
x-+-o3 V x V(2a+x) 2 X-++o 3 V x Vt2a + x) 2
for x < 0 y' < 0, and for x > 0 t/ > 0.

Hence, at x=O the function has a minimum. The value of the function
at this point is zero.
Now investigate the function at the other critical point x2 =2a. As x ..... 2a
the derivative also approaches infinity. However, in this case, for all values
of x close to 2a (both on the right and left of 2a), the derivative is nega-
tive. Therefore, at this point the function has neither a maximum nor a
minimum. At and about the point x 2 =2a the function decreases; the tangent
to the curve at this point is vertical.
At x= ~ the derivative vanishes. Let us investigate the character of
this critical point. Examining the expression of the first derivative, we note
that
4a 4a
for x < y' > 0, and for x >3 y' < 0.
3
4a
Thus, at x= the function has a maximum:
3
. 2
Ymax=3a v- 4.
4) On the basis of this study we get the dqmains of increase and

decrease of the function:
for - oo < x < '0 the func.tion decreases;
for 0 <x<~ the function increases;
4a
for
3 <x< oo the functio.n decreases.
5) Determine the domains of convexity and concavity of the curve and

the points of inflection: the second derivative
8a 2
!l' = - --.----5
3
9x. (2a-x)
does not vanish at a single point. Yet there are two points at which the
second derivative is discontinuous: x,=O and x.=2a.
Let us investigate the sign of the second derivative near each of these
points:
for x <0 we have y" <0 and the curve is convex up;
for x >0 we have y" <0 and the curve is convex u_p.
Hence, the point with abscissa x= 0 is not a point of inflection.

For x < 2a we have y" < 0 and the- curve is convex upwards;
for x > 2a we have y" > 0 and the curve is convex down.
2a x
Fig. 133.
Hence, the point (2a, 0) on the curve is a point of inflection.

6) Determine the asymptotes of the curve:
V2ax-x
V -x-1 =-1,
k= lim JL= lim lim ~-
X-+< X X-+< X X-+<
b= lim [V2ax 2 -x+x]=

X-+ ao
Thus, the straight line
Is an inclined asymptote to the curve y= V2ax-x. The graph of this

function is shown in Fig. 133.
Investigating Curves Represented Parametrically 199
SEC. 12. INVESTIGATING CURVES REPRESENTED PARAMETRICALLY
Let a curve be given by the parametric equations

X= q> (!}, ~
(I)
y='iJ(/). J
In this case the investigation and construction of the curve is
carried out just as for the curve given by the equation
y = f (x).
Evaluate the derivatives
dx
df=q> , ( t ) t }
(2)
~~ = 'i'' (t).
For those points of the curve near which it is the graph of a
certain function y = f (x}, evaluate the derivative
dy "'' (/) (3)
dx = tp' (1)
We find the values of the parameter t = ttt t,, ... , tk for which at
least one of the derivatives q>' (t) or 'I'' (t) vanishes or becomes
discontinuous. (We shall call these values of t critical values.)
By formula (3}, in each oftheintervals(/ 1 , t,); (t., t.); .. . ; (tk_J'tk)
and hence, in each of the intervals (x~' x.); (x x.); ... ; (xk_ 1 , xk)
(where xi= q> (ti )}, we determine the sign of ~, in this way determin-
ing the domain of increase and decrease. This likewise enables us
to determine the character of points that correspond to the values
of the parameter t,, t., ... , tk. Next, evaluate
2
d y . "'" (/) ljl' (t)-tp" (/) "'' (t) (4)
dx 2 = [tp' (t)]
From this formula, determine the direction of convexity of the

curve at each point.
To find the asymptotes determine those values of t, upon
approach to which either x or y approaches infinity, and those
values of t upon approach to which both x and y approach in-
finity. Then carry out the investigation in the usual way.
The following examples will serve to illustrate some of the
peculiarities that appear when investigating curves represented
parametrically.
Example 1. Investigate the curve given by the equations

x=a cos t, } (I')
y=a sin t.
Solution. The quantities x and y are defined for all values of t. But since
the functions of cos t and sin t are periodic, with a period 2rc, it is suffi.
cienl lo consider the variation of the parameter t in the range from 0 to 2rc;
here the interval [-a, a] is the range of x and the interval [-a, a) is the
range ol y. Consequently. this curve has no asymptotes. Next, we find
~; = - 3a cos 2
t sin t, }
:r = 3a sin 2 t cost.
(2')
3rc
These derivatives vanish at t=O, ~ , rc,
2 , 2rc. Evaluate
dy 3a sln 2 t cost
-tan t. (3')
dx = -3a cos 2 t sin t
On the basis of (2') and (3') we compile the following table:
Sign I'ofTypeasof a varlatlo~

Range ol
I
rc
I
Corresponding
range of x
I
Corresponding
range of y
l ol dy
dx
y
of X
functIon
(Y=f (X))
0<t < 2 a>x>O O<y<a - Decreases

rc
2
<t<rc 0> x>-a a>y>O + Increases
3n
rc<t<
2 -a<x<O O>y >-a - Decreases
3rc
2
<t<2rc O<x<a -a<IJ<O + Increases
From the table it follows that equations (I') define two continuous functions
of the type y = f (x), for 0 <;tEO;; n y ~ 0 (see first two lines of the table),
for n < t E0;;2rc y < 0 (see two last lines of the table). From (3') it follows
that
lim dy =co
n dx
1-+-
2
and
lim dy =co.
mdx
X-+--
2
At these points the tangent to the curve is vertical. We now find

dyl
dT t =
-0
0- '
dyl t = 0--0' dyl
dt
-0
dt t = 2n -
Investigating Curves Represented Paramelricall.:J 201
At these points the tangent to the curve is

horizontal. We then find y
d'y
dx 2 = 3a cos t sin t
Whence it follows that

dy
for 0 < t < :rt dx > 0 the curve is concave, X
for :rt < t < 2:rt d;. <

d' -
0 the curve is convex. l
On the basis of this investigation we can

construct a curve (Fig. 134), which is called Fig. 134.
an astroid.
Example 2. ~onstruct a curve given by the following equations (folium of
Descartes):
(!")
Solution. Both functions are defined for all values of t except t =-I. and
.
I lffi X= I'lffi -3at
-= + oo, . .
--=
~~at
1 - oo;
I1m y= 1trn
1-+-t-o t-+-1 +ol + t 1--1-o 1-+-t-o I +1
lim
t-+-1+0
x=- oo, lim
t-+-1+o
y= + oo.
Further note that
when f=O X=O. y=O,
when + oo
t:::::: x ---.. 0, y _.., 0,
when t :::::: - co x ---.. 0, y _.., 0.
dx dy
Find ([[ and ([[ :
dx
(.!..-ta)
6a
2 dy 3at (2 - 13 )
([{ = (I+ ta)z ([[= (1+1 1 ) 2
(2")
For t we get the following critical values:
Then we find
dy
dy ([[ t (2-1 3 )
(3")
dx = dx= 2 (.!..-t)'
dt 2
On the basis of formulas (1"), (2"), and (3") we compile the following table:
Range of Corresponding Corresponding I I

SigJ T(Pe of variation
I
I range of x
I range of y of .J!.. o y as a function
dx of x (y=f (x))
-oo<t<-1 O<x<+oo O>y>-oo - Decreases

-1<1<'0 -oo<x<O +oo>y>O - Decreases
O<t x>aV2 a V2<y<aV4
2 - Decreases
v2 < t <co a V2>x>0 a V4>y>0 + Increases
From (3") we find
Thus, the curve cuts the origin twice: with the tangent parallel to the x-axis
and with the tangent parallel to the y-axis. Further,
(:~)t=-'- =co.
V2
(u=aV
x=aV~)-
2
At this point the tangent to the curve is vertical.
(~~)~ =V2=o.
(:::~~)
At this poinf the 'tangent to the curve is horizontal. Let us investigate the
question of the existence of an asymptote:
k= lim JL= lim 3at2(1+t') __ 1
x.,.++ooX t ... -t-o3at(1+t')- '
b= lim (y-kx)= lim [13atlt-(-1) 13+att] =

x_.+oo x... -t+o +.
___ lim [3at (t +I >J _ lim 3at =-a
-1 ... -t-o 1+1 1 -1-+-t-ol-1+1 2 _ '
Exercises on Chapter V 203
Hence, the straight line y=-x-a is an

asymptote to a branch of the curve as
X--++00.
Simi! arly we find
k= lim JL= -I,

X-+-00 X
b= lim (y- kx) =-a.

x~-co
Thus, the straight line is also an asymp

tote to a branch of the curve as x --+- oo.
On the basis of this investigation we
construct the curve (Fig. 135).
Some problems involving investigation of
curves will again be discussed in Chapter
VIII "Singular Points of a Curve". Fig. 135.
Exercises on Chapter V
Find the extremes of the functions: I. y =X 2 -2x +3. Ans. Ymln = 2 at
x1 7
X= 1. 2. y=3-2x 2 +3x+ I. Ans. Ymax= at x= I. 3. y=x 1 -9x 2 + 15x+3.
2
Ans. Ymax= 10 at X= I, Ymin= -22 at x=5. 4. y= -x 4 +2x 2 Ans. Ymax= l
at X= I. Ymtn=O at X=O. 5. y=x-8x~+2. Ans. Ymax=2 at x=O,
Ymtn=-14 at x= 2. 6. y=3x 5 -125x 1 +2160x. Ans. max at x=-4 and
2
1
x=3, min at x=-3 and x=4. 7. y=2-(x-1) Ans. Ymax=2 at X= I.
...!._ x 2 -3x+2
8. y=3-2 (x+ I) Ans. Neither max nor min. 9. y= x2 + ~+ . Ans. min
3 2
1
at x= y2, max at x=- 2. 10. y=(x- ) ~J-x). Ans. max at x=.!_s~.

2
X .
-" A ns.mmax=-2.
ll.y= 2e" +e. . t In 2 I 2. x
y=lnx' A ns. Ymtn=e at
x=e. 13. y=cosx+sinx ( - ~ ..,;;;x..,;;; ~} Ans. Ymax= 2 atX=~
"14. y=sln2x-x ( - ~ ..;x..; ~ ) Ans. max

at x= ~ . min at x=- ~
15. y=x+tanx. Ans. There is neither max nor min. 16. y=e"sinx.
Ans. min at x=2k:t- ~ , max at x=2k:t+! :t. 17. y=x4 -2x 2 +2. Ans.
maxwhen x = 0; two min when X=- I and when x =I. 18. y=(x-2) 1 (2x+ 1).
I I
Ans. Ymln ::::::-8.24 when x=s. 19. y=x+x. Ans. min when x= I; max
a a
when x=-1. 20. y=x 2 (a-x) 2 Ans. Ymax= when x=2:Ymtn=Owhen
16
a bl at
x=O and when x=a. 21. y=-+-- Ans. max when x=--b; min
x a-x a-
at .~
whenx=a+b" 22. y=x+rl-x. Ans.Ymax=l whenx=l:Ymtn=-1
when x=-1. 23.

.r-
y=x r 1-x (x,.;;;;l). Ans. Ymax=3
2 ,/T 2
V 3whenx=3
X
24. y=l+xz Ans. min when x=-1; max when x=l. 25. y=_xlnx. Ans.
I
min when X=_!_. 26. y=xln 2 x. Ans. max when x=e 2 ;minwhen x=l.
e
27. y=lnx-arc tan x. Ans. The function increases. 28. y=sin3x-3sinx.
3t 33t
Ans. min when x=
2 ; max when x=
2 . 29. y=2x+ arc tan x. Ans. No
extrema. 30. y=sinxcos 2 x. Ans. min when _x=~; two max: when
x =arc cos V~ .and when x =arc cos(.- V~). 31. y =arc sin (sin x).
(4m+ I) 1t . (4m+3) n
Ans. max when x= . ; mtn when x= .
2 2
Find the maximum and minimum values of the function on tile indicated
intervals: 32. y=-3x'+6x 2 -1 (-2..o;;x,.;;;;2). Ans. Maximum y=2 at
3
X=::!:l. minimum y=-25 at x=2. 33. y=~-2x 2 +3x+1 (-1..;;;;x.;;;;5) .
. A ns M ax1mum
. 23 13
va I ue y=g a t x= 5 , mm1mum
. . va I ue y= - at. x=- 1.
3
x-1 3
34. y=x+l (O,.;;;;x.;;;;4). Ans. Maximum value u=r; at x=4, minimum
value y= -I at t=O. 35. y=sin2x-x ( - ; ..o;;x..;;;; ~). Ans. Maximum

n t n. . . I :t :t
I
vaue y=T a x=-
2 , m1111mum vauey=-
2 at x=
2 .
36. Using square tin sheet. with a side a, make a topless box of maximum
volume by cutting equal squares at the corners and removin& them and then
bending the tin so as to form the sides of the box. What w11l the length_of
a side of the squares be? Ans. ~ .
37. Prove that ol all rectangles that may be inscribed in a given l;iicle,
the square has the gre.atest area. Also show that the square will have the
maximum perimeter as well. .
38. Show that of all isosceles triangles inscribed in a given circle, an equi
lateral triangle has the largest perimeter.
39. Find a right triangle of maximum area with a hypotenuse h. Ans.
Length of each side, ~.
40. Find the height ol a right cylinder with greatest volume that can be
inscribed in a sphere of radius R. Ans. Height, ~;.
41. Find the height of a right cylinder with greatest lateral surface that
may be inscribed in a given sphere of radius R. Ans. Hei~ht, R V2.
42. Find the height of a right cone with least volume circumscribed about
a given sphere of radius R. Ans. 4R (the volume of the cone is equal to two
volumes of the sphere).
43. A reservoir with a square bottom and open top is to be lined inside
with lead. What are the dimensions of the reservoir (to hold 32 litres) that
Exercises on Chapter V 205
will require the smallest amount of !earl? Ans. Height, 0.2 metre, side of
base, 0.4 metre (the side of the base must be twice the height). .
44. A roofer wants to make an open channel of maximum capacity with
bottom and sides 10 em in width, and with the sides inclined at the same
angle to the bottom What is the width of the channel at the top? Ans.
20 em.
45. Prove that a conic tent of given storage capacity re_quires the least
material when its height is 2
times thE: radiu.s of the base.
46. It is required to make a cylinder, open at the top, the walls and
bottom of which have a given thickness. What should the dimensions of the
cylinder be so that for a given storage capacity it will require the least
material? Ans. If R is the inner radius of the base, u the inner volume of
the cylinder, then
.
R= V u
1t
47. It is required to build a boiler out of a cylinder topped by two
hemispheres and with walls of constant thickness so that for a given volume
u it should have a minimum outer surface. Ans. It should have the shape of
a sphere with inner radius R= V :~ .
48. Construct an isosceles trapezoid, which lor a given area S has a mini-
mum perimeter; the angle at the base of the trapezoid is equal to a. Ans.
The length of one of the nonparallel sides is
V
.. / Is .
sna
49. Inscribe in a given sphere of radius R a regular triangular prism of
maximum volume. Ans. The attitude of the prism Is .~R_
r a
50. It is required to circumscribe about a hemisphere of radius R a cone
of minimum volume; the plane of the base of the cone coincides with that
of the hemisphere; find the altitude of the cone. Ans. The altitude ol the
cone is R 3.
51. About a given cylinder o! radius r circumscribe a right cone of mini-
mum volume; we assume the planes and centres ol the circular bases of the
cylinder and the cone coincide. Ans. The radius of the base of the cone is
3
equal to
2 r.
52. Out of sheet metal, having the shape of a circb of radius R, cut
a sector such that it may be bent into a funnel of maximum storage. capacity.
.
Ans. The central angle of the sector is 2n
.../2
V . 3
53. Of all circular cylinders inscribed in a given cube with side a so that
their axes coincide with the diagonal of the cube and the circumferences of
the bases touch its planes, find the cylinder with maximum volume. Ans. The
altitude of the cylinder Is equal to a Y 3 ; the radius of the base is .~-.
3 r 6
54. Given, in a rectangular coordinate system, a point (x0 , y 0 ) lying in the
first quadrant. Draw a straight line through this point so that it forms
a triangle of least area with the positive directions of the axes. Ans. The
straight line intercepts on the axes the segments 2x 0 and 2y0 ; thus, it has the
.
equa t 10n -X + -y = I.
2Xo 2Yo
55. Given a point on the axis of the parabola y 2 =2px at a distance a

from the vertex, find the abscissa of the point of the curve closest to it.
~Ans. X=a-p.
56. Assuming that the strength of a beam of rectangular cross-section is
directly proportional to the width and to the cube of the altitude, find the
width of a beam of maximum strength that may be cut out of a log of diameter
16 em. Ans. The width is 8 em.
57. A torpedo boat is standing at anchor 9 km from the closest point of
the ~hore; a messenger has to be sent to a camp 15 km lalong the shore)
from the point of. the shore closest to the boat. Where should the messenger
land so as to get to the camp in the shortest .possible time, if he does 5 kmjhr
walking and 4 kmjhr .rowing. Ans. At a point 3 km from -the camp.
58. A point moves over a plane in a medium situated outside the line MN
with velocity v 1 , and along the line M N with velocity v 2 . What path
between A and B, situated on M N, will it cover in the shortest time? The
distance of A from M N is h, the distance of the projection a of A on the
line M N from B is a.. Ans. If ACB is the path of the point, then ~~ = ~:
aB v1 aB v1
I or AB:;;.v; and aC=aB for AB<v;
59. A load w ..is hoisted by a le-ver; a force F is applied to one end, the
point of support is at the other end of the lever. If the load is suspended
!rom a point a centimetres from the fulcrum, and the lever rod weighs v
grams per centimetre of length, what should the length of the rod be for the
force (required to raise the load) to be a minimum? Ans. x= y :w
60. For n measurements of an unknown quantity x the following readings
2
em.
have been obtained: x~> x 2 , , Xn- Show .that the sum of the squares ,of the
errors (x-x1 ) 2 +(x-x2 ) 2 + ... +(x-xn) 2 will be least if for x we take the
+
number x1 x;+ Xn. +
n
61. To reduce the friction of a liquid against -the walls of a channel, the
area in contact with the liquid must be a minimum. Show that the best shape
ol an open rectangular channel with given cross-sectiomil area is that for
which the width of the channel is twice its altitude.
Determine the points of inflection and the intervals of convexity and con-
cavity of the curves.
62. y=x 5 Ans. For x < 0 the curve -is convex; for x > 0 the curve is con-
cave; at x = 0 there is a point of inflection. 63. y =I -x 2 Ans. The curve
+
is everywhere convex. 64. y =x -3x 2 -9x 9. Ans. Point of inflection at
x=l. 65. y=(x-b)1 Ans. Point of inflection at x=b. 66. y=x. Ans. The
curve is everywhere concave. 67. y= x2 ~ 1 . Ans. Point of inflection at
x= ,;.,
y 3
. 68. y=tanx. Ans. Point of inflection at x=mt. 69. y=xe-"'.
. .
Ans. Point of inflection at x=2. 70. y=a-,Vx-b. Ans. Point of inflec-
tion at x= b. 71. y=a- V<x-b) 2 Ans. The curve has no point of inflection.
1
Find the asymptotes of the following curves: 72. y = --1 - Ans. X= I,
X-
I a
y=O. 73. y=(x+ 2)"' Ans. x=-2, y=O. 74. y=c+(x-b) 2 Ans. x=b,
Exercises on Chapter V '207
y=c. 75. y=ex -!. Ans. x=O, y=O. 76. y=lnx. Ans. x'"'O.
4
4
77. y =6x'+x'. Ans. y=x+2. 78. y'=a-x. Ans. y+x,=U.
x' .
79. y=-2.a--x
--.Ans.x=2a.80. y'(x-2a)=x'-a'. Ans. x=2a, u=(x-j-a).
Investigate the following functions and construct their graphs:
1
8a 3 6x
81. y=x4 -2x+!O. 82. y=x+ 4a 83. y=e x. 84. y= I+x,
4+x x x+2 x ,
85. y=-.-. 86. y=-
2 - . 87. y=-,-. 88. y =- - . 89. y 2 =x -x.
x'
90. y= ----..
3 -:x
X
91. y= x"+2. 92. y=x-

X-
v- 1
x'+!. 93. y= --.
X
x+ I
1 +x
v- yx-!
94, y=xe-". 95. y=x'e-"'. 96. y=x-ln(x-f:"l). 97. y=ln (x 2 +1)
::~. y::~:~~inx~9. 1 :;x+ys:;~x .~::~. {Y~-::~~.x. 1:;~. y{=:: 1s:,nx.

X y = 2 t. y = t',
x=a(t-slnt'), { x=ae 1 cost,
106. { y=a(l-cost). 107. t .
y=ae smt.
Additional Exercises
Find the asymptotes of the following lines: 108. y=~:;. Ans. x=-l;
y=x-1. 109: y=x+e-x. Ans. y=x. 110. 2y(x+l)'=x'. Ans. x=-1;
y= ~ x-1. Ill. y'=a 1 -x2 Ans. x+y=O. t.12. y=e-xsinx. Ans. y=O.
113. y=e-xsln2x+x. Ans. y=x. 114. y=xln(e+~).Ans.x=-+;

1 ...!.... 2t t
y=x+e 115. y=xex. Ans. x=O; y=x. 116. x= _ 12 Y= _ z
1 1 1
1 I
Ans. y= 2 x- 2 .
Investigate and graph the following functions: 117 .. y=l.xj. 118. y=lnlxl.
119. y=x'-x. 120. y=(x+l)"(x-2). 121. y=x+ixj. 1-22. y= x-x. V-
123. y=x
--
Yx+i. 124. u= 2~ -lnx. ~
125. y=2lnx. 126. y= e"-l
I
x ~ In x
127. y=-1- . 12i:!. y=x+-. 129. y=xlnx. 130. y=e" -x. 131. u=
nx x
sin x
=lsin3xj. 132. u=--. 133. y=xarctanx. 134. y=x-2 arctanx.
X
135. y=e-Xsin3x. 136. g=jSinxl+x. 137, y=SinX1 138. g=COS 3 x+sin 3 X.
CHAPTER VI
THE CURVATURE OF A CURVE
SEC. 1. THE LENGTH OF AN ARC AND ITS DERIVATIVE
Let the arc of a curve M 0 M (Fig. 136) be the graph of the

function y = f (x) defined on the interval (a, b). Let us determine
the arc length of the curve. On the curve M 0 M take the points
M 0 ,M" M., .. . , M 1_., M 1, .. , Mn-P M. Connecting the points
we get a broken line M 0 M 1 M 2 M 1_ 1 M 1 . Mn_ 1 M inscribed in
the arc M 0 M. Denote the length of this
broken line by Pn.
The length of the arc M~M is the limit
(we denote it by s) approached by the
length of the broken line as the largest
of the lengths of .the segments of the bro-
ken line M 1_ 1 M 1 approaches zero, .if this
limit exists and is independent of any
Fig 136. choice of points of the broken line
MoM 1M2 Mi-1 M i Mn_lM .
It will be noted that this definition of the arc length of an
arbitrary curve is similar to the definition of the- length of a
circumference.
In Ch. XII it will be prov~d that if a function f (x) and its
derivative f' (x) are continuous on an interval [a, b), then the arc
of the curve y=f (x) lying between the points [a, f(a)] and
lb. f (b)] has a definite length; a method will be shown for com-
puting this length. There also, it will be established (as a corollary)
that under the giv~n conditions the ratio of the length of any arc
of this curve to the length of its chor~ approaches unity when the
length of the chord approaches zero, that is,
--.
Jim length M 0 M
1.
M 0 M-> o length M 0 M _
This theorem may be readily proved for the circumference*) of
*) Consider the arc AB, the central angle of which is 2a. (Fig. 137). The
length o[ this arc is 2Ra (R is the radius of the ~ircle), and the length of its
chord is 2R sin a. There lore, lim length AB = lim ~= 1 -
a-+ o length AB a-+ o 2R sin a
The Length of an Arc and Its Derivative 209
a circle; however, in the general case we shall accept it without

proof (Fig. 137).
Let us consider the following question.
On a plane we have a curve given by the equation
y =f (x).
Let M 0 (x 0 , Yo) be some fixed point of the curve and M (x, y),
some variable point of the curve. Denote by s the arc length
MOM (Fig. 138).
0 X0 X X+!JX X
Fig. 137. Fig. 138.
The arc length s will vary with changes in the abscissa x of

the po.int M; in other words, s is a function of x. Find the deri-
vative of s with respect to x.
Increase 'x by !1x. Then the arc s will change by f1s= the
length of MM,. Let MM, be the chord subtending this arc. In
order to find lim Lls do as follows: from !1MM;Q find
6X-> 0 dX - .
MM~ = (!1x) 2 + (!1y) 2

Multiply and divide the left-hand side by !1s 2 :
( :~) z !1sz = (!1x)z +(!1y)z ..

Divide all terms of the equation by f1x 2 :
(M::;r (:;r = 1+ (!~Y

Find the limit of the left and right sides as !1x---? 0. Taking into
account that lim
M M,-> o
MM,
dS
= 1 and that lim LlY =
6x _,. o dX
~ we get
X
210 The Curvature of a Curve
or
ds
dx
= y 1+(dy)". dx (1)
For the differential of the arc we get the following expression:
y' 1 + (~~ )
2
ds = dx (2)
or*)
ds = V dx" + dy". (2')
We have obtained an expression for the differential of arc
length for the cas-e when the curve is given by the equation
y = f (x). However, (2') holds also for the case when the curve is
represented by parametric equations.
If the curve is represented parametrically,
X=IJl (t), y="' (f),
then
dx = <p' (f) dt, dy ='I'' (t) dt,
and expression (2') takes the form
v
d s = =-[.IJl--,-----,(:-:-I):-:--]z +---,--------;--["'-;-,~(t ):-7:] dt.
1
SEC. 2. CURVATURE
One of the elements that characterise the shape of a curve is

the degree _of its bentness, or curvature.
Let there be a curve that does not intersect itself and has
a definite tangent at each point. Draw tangents to the curve at
any two points A and B and denote the angle formed by these
tangents by a [or, more precisely, the angle through which the
tangent turns from A to B' (Fig. 139)]. This angle is called the
angle of contingence of the arc AB. Of two arcs of the same
length, that arc is more curved which has a greater angle of
contingence (Figs. 139 and 140).
On the other hand, when considering arcs of -different length we
cannot evaluate the degree of their curvature solely by the appro-
*) Strictly speaking, (2') holds only for the case when dx > 0. But if
dx < 0, then ds=- Jf dx 2 +dy 2 For this reason, in the general case this
formula is more correctly written as Ids!= Jfdx 1 +dy 2
Curvature 211
priate angles of contingence. Whence it follows that a complete

description of the curvature of a curve is given by the ratio of
the angle of contingence to the length of the corresponding arc,
Fig. 139. Fig. 140.

,-...
Definition 1. The average curvature Kav of an arc AB is the
ra1io of the corresponding angle of contingence a to the length of
the arc:
a
Kav=-;:::::::.
AB
For one and the same curve, the average curvature of its diffe-
rent parts (arcs) may be different; for example, for the curve shown
in Fig. 141, the average curvature of
,-...
-
the arc AB is not equal to the average
curvature of the arc A 1 BP although
the lengths of their arcs are the same.
What is more, at different points the
curvature of the curve differs. To cha-
racterise the degree of curvature of a
given line in the immediate neighbour-
hood of a given point A, we introduce Fig. 141.
the concept of curvature of a curve at
a given point.
Definition 2. The curvature KA of a line at a given point A is
the limit of the average curvature of the arc AB when the length
of this arc approaches zero (that is, when the point B approa-
ches the point A):
KA = lim Kav= lim :._ *).
B-+A AB-+0 AB
*) We assume that the magnitude of the limit does not depend on which
side of the point A we take the variable point B on the curve.
212 The Curvatwe of a Curve
Example. For a circle of radius r: I) determine the average curvature of

the arc AB
subtending the central angle a (Fig. 142); 2) determine the
curvature at the point A.
Solution. 1) Obviously the angle of contingence of the arc AB is a, the
length of the arc is ar. Hence,
a
Kav=-
ar
or
I
Kav=-.
r
2) The curvature at the point A is
K= lim ~=_!_.
a_,.oar r
Fig. 142. Thus, the a\"erage curvature of the arc of a circle

of radius r is independent of the length and po-
l sition of the arc, and for all arcs it is equal
to r. Likewise, the curvature of a circle at any point is independent of the
_!_
choice of this point and is equal to
r
Note. It should _be noted that, generally speaking, for any curve
the curvature at its various points differs (this will be seen later).
SEC. 3. CALCULATION OF CURVATURE
Let us develop a formula for finding the curvature of any line

at any point M (x, y). We shall assume that the curve is represen-
ted in the Cartesian coordinate y
system by an equation of the form
y = f (x) (l)
and that the function f (x) has a
continuous second derivative.
Draw tangents to the curve at the
pomts M and M, with abscissas x
and- x + tu and denote by q> and
q> + ~q> the angles of inclination of
these tangents (Fig. 143). 0 x
We reckon the length of the
.-,. Fig. 143 .
. arc M.M from some fixed point M.
and denote it by s; then ~s = M-:-M,-M~M, and I ~s I= MM,.
As will be seen from Fig. 143, the angle of contingence corres-
Calculation of Curvature 213
ponding to the arc MM 1 is equal to the absolute value*) of the

difference of the angles <p and <p + ~<p. which means it is equal
to [~<p[.
According to the definition of average curvature of a curve, on
the segment MM 1 we have
. Kav = : ~:r= I~; I
To obtain the curvature at the point M, it is necessary to find
the limit of the expression obtained on the condition that the
arc length MM 1 approaches zero:
K= ds~
limo jMAS I
Since the quantities cp and s both depend on x (are functions
of x), cp may thus be considered as a function of s. We may con-
sider that this function is represented parametrically by means
of the parameter x. Then
lim t!QJ = d<p
as~ 0 tJ,s ds
and, consequently,
K=l::l (2)
To calculate d;,
-d
we make use of the formula for dif-
ferentiating a function represented parametrically:
d<p
d<pdX
Ts=Ts
dX
To express the derivative~~ in terms of the function y=f(x), we
note that tan ~:p= ~~ and, therefore,
dy
cp = arc tan dx .
Differentiating the latter equality with respect to x, we get
d2y
dO.
As regards the derivative ::, we found in Sec. 1, Ch. VI, that
ds =
dx
.. /
V l+(dy)". dx
Therefore,
dq>
;tcp = dx
ds ds
dx
or, since K= J :; j. we finally get
. 1~1
K = [' + (~~rJ .,.. (3)
It is thus possible to find the curvature at any point of a

curve where there exists a second derivative :~ and where it is
continuous. Calculations are done wi~h formula (3). It should be
noted that when calculating the curvatttre of a curve only the
arithmetical (positive) value of the root in the denominator should
be taken, since the curvature of a line cannot (by definition) be
negative.
Example I. Determine the curvature of the parabola y"=2px:
a) at an arbitrary point M (x, y);
b) at the point M 1 (0, 0);
c) at the_ point M 2 (f. p).
Solution. Find the first and second derivatives of the function y= f2px:
dy p d~ ~
dx = y 2px ; dx = - (2px)'l
Substituting the expressions obtained into (3), we get

p
a) K= ,
(2px +p 2 ) /, '
_I.
K x=o--
b)
y=o P
c)
Calculation of the Curvature of a Line Represented Parametrically 215
Example 2. Determine the curvature of the straight line y=ax+b at an

arbitrary point (x, y).
Solution.
y'=a, y"=O.
Referring to (3) we get
K=O.
Thus, a straight line is a "line of zero curvature". This very same result Is
readily obtainable directly from the definition of curvature.
SEC. 4. CALCULATION OF THE CURVATURE Of A LINE REPRESENTED

PARAMETRICALLY
Let a curve be represented parametrically:
X= <p (t), y = 'Jl (t).
Then (see Sec. 24, Ch. III).

dy '(I) d 2y "<p' -'<p"
dx = <p' (t)' dx 2 = (<p' )' .
Substituting the expressions obtained into formula (3) of the

preceding section, we get
I "<p' -'<p" I
K = l'"+'I''J'' (l)
Example. Determine the curvature of the cycloid
X=a(t-sint), y=a(l-cost)
at an arbitrary point (x, y).
Solution.
dx d 2x dy d2y
dt=a(l-c~sl), (il!=asint, dt=aslnt, dt 2 =acost.
Substituting the expressions obtained into (3), we get
K = I a (I -cos t) a cos t- a sin t a sin t I I cos t- I 1
1a2 (I -cos !)I+ a2 sin" t I'I 2'1a (I -cos t)'/, =
. I I
= 2'ia(l-cos I) '/ 4a ]sin~!
SEC. 5. CALCULATION OF THE CURVATURE OF A LINE GIVEN BY AN

EQUATION IN POLAR COORDINATES
Given a curve represented by an equation of the form
Q=f(O). (1)
Write the transformation formulas from polar coordinates to

Cartesian coordinates:
X=QCOS0,}
(2)
y= Q sin 0.
If in these formulas we replace Q by its expression in terms

of 6, i.e., f (6), we get ..
X= f (6) COS 0, }
(3)
y = f (0) sin 0.
The latter equations may be regarded as parametric equations

of curve (I), the parameter being 0.
Then
dx dQ
dS = dS cos 0-Q
.
Stn 0,
dy dQ .
df)= dS Stn
O+QCOSv,n
Substituting the latter expressions into (I) of the preceding

section, we get a formula for calculating the curvature of a curve
in polar coordinates:
[( _ I Q2 +2Q''-QQ" I ( )
4
- (ez+ e'">"'
Example. Determine the curvatur-e o[ the spiral of Archimedes Q = aO (a> 0)
at an arbitrary point (Fig. 144).
Solution.
dQ d 2Q
de=a; dsz=O.
Hence
K __ \ azsz + 2az I = _!__ ez +2
-(azez+a)t. a (82+1)t.
lt will be noted that for
- large values of f) we have the
approx1rna 1 1es 82+2
. t e equa l't" --oz=::::
82 + 1
ez-:::::: 1; therefore, rep-
:::::: 1,
<J
Jacinr f)Z + 2 by f)Z and 8 2+ 1

Ftg. 144, by e in the foregoing formula,
The Rqdi !i.S and Circle of Curvature. Euo/ute and I nuolute 217
we get an approximate formula {for large values of 8)

I 82 1
K:::::::---=-
a (82)'1 a8'
Thus, for large values of 8 the spiral of Archimedes has, approximately,
the same curvature as a circle of radius a8 .
. SEC. 6. THE RADIUS AND CIRCLE OF CURVATURE.

CENTRE OF CURVATURE. EVOLUTE AND INVOLUTE
Definition. The quantity R, which is the reciprocal of the cur-

vature K of a line at a given point M, is called the radius of
curvature of the line at the point in question:
I
R =K (1)
or
(2)
Draw a normal, at the point M, to a curve in the direction of

the concavity of the curve, and lay off a segment MC equal to
the radius R of the curvature of the curve at the point M. The
Ftg. 145. Fig. 146.
point C is called the centre of curvature of the given curve at M;

the circle, of radius R, with centre at C (passing through M) is
called the circle of curvature of the given curve at the point M
(Fig. 145).
From the definition of circle of curvature it follows that at a
given point the curvature of a curve and the curvature of a circle
of curvature are the same.
218 The Curvature of a Curue
Let us derive formulas defining the coordinates of the centre of

curvature.
Let a curve be given by the equation
y=f(x). (3)
Take a point M (x, y) on this curve and determine the coordi-
nates a and ~ of the centre of curvature corresponding to this
point (Fig. 146). To do this, write the equation of the normal to
the curve at M:
1
Y-y= --,(X-x).
y
(4)
(Here, X and Y are the moving coordinates of the point of the
normal.)
Since the point C (ex, ~) lies on the normal, its coordinates
must satisfy equation (4):
~~y= _..; (ex--x). (5)
Y,
Further, the point C (a, ~) is separated from M (x, y) by a
distance equal to the radius of curvature R:
+
(ex- x) 2 (~- y) 1 = R 1 (6)
Solving equations (5) and (6) simultaneously, we find a and ~:
(a-x) 1 +-h
y (a-x)
1
= R1 ,
11
(a-x) =
2 -
1 +Y y'"R I
Whence
y'
a=x y 1+g' .R.
(I+ !/)/,
and since R= I u"l
y' (I+ y'") ' l+y'"
a=x I y" I ' ~=Y=FTifT
In order to decide which signs (top or bottom) to take in the
latter formulas, we must examine the case y" > 0 and the case
y" < 0. If y" > 0, then at this point the curve is concave, and,
hence, ~ > y (Fig. 146), and for this reason we take the bottom
signs. Taking into account that in this case Iy" J = y", the formulas
of the coordinates of the centre of curvature will be
g' (I +y'') )
a=x- y"_ ,
+ y''
1
~ = y +----yo- .
f1 (7)
The Radius and Circle of Curvature. volute and involute 219
Similarly, it may be shown that formulas (7) will hold for the
case y" < 0 as well.
If the curve is represented by the parametric equations
X=cp(f), y='IJ(f),
then the coordinates of the centre of curvature are readily obtain-

able from (7) by substituting, in place of y' and y", their
expressions in terms of the parameter
I Yt
y =-.;
Xt
Then
a-x y' (x' + y') }
~ x'!l'-x"y' '
x' (x'' + y'") (7')
~ =Y-1- x'y"-x"y'
Example 1. To determine the coordinates of the centre of curvature of. the
parabola
y 2 =2px:
a) at an arbitrary point M (x, y); b) at the point M0 (0, 0); c) at the

point M 1 ( f , p) .
Solution. Substituting the values :~ and ~ into (7) we get (Fig. 147):
a) a=3x+p, ~=
(2x)t.
YP ; ly
~
b) at x=O we find a=p, ~=0;
c) at x= ~ we have a=T, ~=-p. M

)(
If at M, (x, y) of a given line the cur-
vature differs from zero, then a very de-
finite centre of curvature C, (a, ~) corres-
ponds. to this point. The totality of all
centres of curvature of the given line forms
a certain new line, called the evolute, with Fig. 147.
respect to the first.
Thus, the locus of centres of curvature of a given line is called
the evolute. As related to its evolute, the given line is called the
evolvent or involute.
If a given curve is defined by the equation y = f (x), then equa-
tions (7) may be regarded as the parametric equations of the evo-
220 The Curpature of a Curve
lute with parameter x. Eliminating from these equations the para-

meter x (if this is possible), we get an immediate relationship
between the coordinates of the evolute a and ~- But if the curve
is given by parametric equations x = c:p (t), y = '\jJ (t), then equa-
tions (7') yield the parametric equations of the evolute (since the
quantities x, y, x', y', x", y" are functions of t).
Example 2. Find the equation of the evolute of the parabola
y 2 =2px.
Solution .. On the basis of Example 1 we, have, for any point (x, y)
of the parabola,
a=3x+p,
(2x) t.
~=--
VP'
Eliminating the parameter x from these.
equations, we get
8
~z= 27p (a- p)'.
This is the equation of a semicubical

parabola (Fig. 148).
Example 3. Find the equation of the
evolute of an ellipse represented by the
parametric equations
x=acost, y=bsint.
Solution. Evaluate the derivatives of x
and y with respect to t:
x' =-a sin t, y' = b cost;
Fig. 148.
x" =-a cost, y" =:= -b sin t.
Substituting the expressions of the derivatives into (7'), we get
b cost (a 2 -sin 2 t+b 2 cos 2 t)

a =a cos t - --;-'-;-.....-:c-:--.-'---;;-;---'-
ab sin 2 t +ab cos 2 t
,;a cost-a cost sln 2 t- ~ cos' t =
2
(a~~) cos' t.
Thus,
a= (a-~) cos' t.
Similarly we get
The Properties of an volute 221
Eliminating the parameter t, we get the equation of the evolute of the ell ips~
in the form
( r (! r= (a2a/2)".
~ +
Here, u and ~ are the coordinates of the evolute (Fig. 149).
Example 4. Find the parametric equations of the evolute of the cycloid
x=a(t-sint), y
y=a (1-cosl).
Solution.
x'=a(l-cost); y'=asint;
x" =a sin t; y" =-a cost.
Substituting the expressions obtained
into (7'), we get
a=a (t +sin t),
~=-a(l-cost).
Rearrange the variables, putting

u=~-:rta,
~ =TJ-2a,
t='t'-:rt;
then the equations of the evolute will
take the form
6=a (1"-sin 't'), Fig. 149.
TJ =a (I-eos 't');
they. define, in coordinates s.
TJ, a cycloid with the same generating circle of
radius a. Thus, the evolute of a cycioid is that same cycloid dis placed alon'"'
the x-axis by -:rta and along the y-axis by -2a (Fig. !50). o
Fig. 150.
SEC. 7. THE PROPERTIES OF AN EVOLUTE
Theorem L The normal to a given curve is a tangent to its evolute.

Proof. The slope of the line tangent to an evolute defined
by the parametric equations (7') of the preceding section is
222 The Curuature of a Curue
equal to
d~
d~ dx
da = da
dx
Noting that [by virtue of the same equations (7')]
da 3y""y''-y'y'"-y' 3 y'"' '3y'y''-y"'-y'y"'
dx= = -y (1)
y""
d~ 3!/y'-y'"-y'"y"'
dx=- (2)
we get the relationship
d~ I
da = -y-
But y' is the slope of the line tangent to the curve at the corre-
sponding point; it therefore follows from the relationship obtained
that the tangent to the curve and the tangent to its evolute at
the corresponding point are mutually perpendicular; that is, the
normal to a curve is the tangent to the evolute.
Theorem 2. If, over a certain segment M1 M2 of a curve, the
radius of curvature varies monotonically (i.e., either only increases
or only decreases), then the increment in the arc length of the eva-
lute on this segment of the curve is equal (in absolute value) to
the corresponding increment in the radius of curvature of the given
cur.ve.
Proof. From formula (2'), Sec. 1, Ch. VI, we have
ds = da. + d~
where ds is the differential of the arc length of the evolute;
whence
(::r r+ (:~ r. = ( ::
Substituting, here, the expressions (l) and (2), we get
ds)- ,z (3y'y" -y"'-y''y"') 1 2
(3)
( dx - ( l + Y ) !/'"
Then find (::r Since .
R =(I + y'"l'' + y' 1 3
R2 = (I )
y" , y"'
Differentiating both sides of this equation with respect to x, we
get the following (after appropriate manipulations):
dR 2(1 + y' 1 2
) (3y'y""-y"' -y' 1 y'")
2R dx = {y")s
The Properties of an volute 223
2 (I+ y'')S/2
Dividing both sides of the equation by 2R --'---'--;,i-'-- ,
y
we have
dR (I+ y' 2 ) ' /2 (3y'.r/''- i/"- y''y"')
dx y""
Squaring, we get
dR)' = 1 + y ,, ) (3y'y"'-y'"-y''y"')
2
(4)
( dx ( r/' '
Comparing (3) and (4), we find
dR)' = (ds )'

( dx dx '
whence
dR = =f ds
dx dx
It is given that ~~ does not change sign (R only increases or

ds
only decreases); hence, dx does not change sign either. For the
sake of definiteness, let ~~ .:;;;;; 0, ~: ~ 0 (which corresponds to
. , dR ds
Ftg. 151). Hence., lit=- ax
Let the point M, have abscissa X 1 and M, have abscissa Apply x,.
the Cauchy theorem to the functions s (x) and R (x) on the
interval [x,, x,]:
where ~ is a number lying between x, and x. (x 1 < ~ < x,).

We introduce _the designations (Fig. 151)
s(x,)=s s(xJ=s 11 R(x,)=R,, R(x1 )=R,.
Then Rs,-sR, = -1, or s,-s, = - (R,-R,). But this means that

.- I
js,-s,I=IR,-R,I.
This equality is proved in exactly the same manner if the radius
of curvature increases.
We have proved Theorems I and 2 for the case when the curva
is given by an explicit equation, y=f(x).
If the curve is represented by parametric equations, these

theorems also hold, and their proof is exactly the same.
Note. The following is a simple mechanical method for constructing
a curve (involute) from its evolute.
y
Fig. 151. Fig. 152.
Let a flexible ruler be bent into the shape of an evolute c.c,

(Fig. 152). Suppose one end of an unstretchable string is attached
to the point c. and bends round the ruler. If we hold the string
taut ~md unwind it, the end of the string will describe a curve M,M 0 ,
fJ whi~h is the involute (or evol
vent, the name coming from
this process of "evolving"). Proof
that this curve is indeed an
involute may be carried out by
means of the above-estaolished
properties of the evolute.
It should be noted that to a
single evol ute there correspond
an infinitude of various
involutes (Fig. 152}.
Example. Let there be a circle of.
radius a {Fig. 153). Take the involute
of this circle that passes through the
Ftg. 153. point M 0 (a, 0).
Approximating the Real Roots of an Equatio.1 225
Taking into account that CM =CM 0 =at, it is easy to obtain the equations
of the involute of the circle:
OP = x =a (cost+ t sin t),
PM =y=a (sin t-t cost).
It will be noted that the profile of a tooth of a gear wheel is most often
in l he shape of the involute of a ci rei e.
SEC. 8. APPROXIMATING THE REAL ROOTS Of AN EQUATION
Methods of investigating the behaviour of functions enable us to

approximate the roots of an equation:
f (x) = 0.
If the equation is an algebraic equation*) of the first, second,
third, or fourth degree, there are formulas which permit expressing
the roots of the equation in terms of its coefficients by means of
a finite number of operations of addition, subtraction, multiplica-
tion, division and evolution. Generally speaking, there are no such
formulas for equations above the fourth degree. If the coefficients
of any equation algebraic or nonalgebraic (transcendental) are not
literal but numerical, then the roots of the equation may be cal-
culated approximately to any degree of accuracy. It should be noted
that even when the roots of an algebraic equation are expressed
in terms of radicals, it is sometimes better, practically speaking,
to apply an approximation method of solving the equation. Below
we give some methods of approximating the roots of an equation.
1. Method of chords. Let there be an equation
f(x)=O (1)
where f (x) is a continuous, doubly differentiable function on th~
interval[a, b). Suppose that by investigating the function !J=f(x)
within the interval [a, b) we isolate a subinterval [x,, x.] such that
within this subinterval the function is monotonic (either increas-
ing or decreasing), and at the end points the values of the func-
tion f(x,) and f(x 2 ) are of different signs. For definiteness, we say
that f (x,) < 0, f (x.) > 0 (Fig. 154). Since the function y = f (x)
is continuous on the interval [x,, x.], its graph will cut the x-axis
in some one point between x, and X2
Draw a chord AB connecting the end points of the curve
y = f (x), which correspond to abscissas x, and x.. Then the
) The equation [(x)=O is called algebraic if f(x) is a polynomial (see

Sec. 6, Ch. VII).
8 - 3388
abscissa a 1 of the point of intersection of this chord with the x-axis

will be the approximate value of the root (Fig. 155). In order to
find this approximate value let us write the equation of the straight
Jir.e AB that passes through two given points A [x 11 f (x,)]
and B [x., f(x,)]: y
11
y-f(xd x-x 1
f (x,)-f (x,) = X 2 - x1
A
Fig. 154. Fig. 155.
Since y = 0 at x =a,, it follows that

-f(x 1) a1 -x1
f (x1 )-f (x,) = x1 -x1 '
whence
(2)
To obtain a more exact value of the root, we determine f(a,).

If f (a,)< 0, then repeat the same procedure applying formula (2)
to the interval [a" x.]. If f (a,)> 0, then apply this formula to
the interval [x,, a,]. By repeating this procedure several times we
wi II obviously obtain more and more precise values of the root
a,a 8 , etc.
Example l. Approximate the roots of the equation
f (x)=x 3 -6x+2=0.
Solution. First find the segments where the function f (x) is monotonic.
Evaluating the derivative f' (x) ==; 3x 1 -6, we find that if is positive for
X<- 2, negative for - 2 <X<+ 2 and again positive for X> v2
(Fig. 156). Thus, the function has three segments of monotonicity, on each of
which there is one root.
To make the calculations more convenient, let us narrow these segments
of monotonicity (but in such manner that there should be a corresponding
'
Approximating the Real Roots of an Equation 227
root on each segment). To do this, substitute into expression f (x), at random,

some values of x, then isolate (within each segment of monotonicity) such
shorter intervals that the functions at the end
points will have different signs: y
x. =0, f (0) =2, } 12
x2 =1, f(l)=-3, tt-
X a=- 3, f (- 3) = - 7, } tO
x. = - 2, f (- 2) = 6, 9
X5 =2, f (2) =-2, }
x8 =3, !(3)=11.
Thus, the roots lie within the intervals
(0, I), (-3, -2), (2. 3).
Find the approximate value of the root in the
interval (0, I); from formula (2) we have
(1-0) 2-~-0 4
a.= 0 -3-2 - 5 -
3 X
Since
f (0.4) =0.4 -60.4 +2 = - 0.336, f (0) =2,
3
it follows that the root lies between 0 and 0.4. Again -3

applying (2) to this interval, we get the following
approximation: -4 y-x 3-6x+2
-5
(0.4-0)2 0.8
-0.336-2 =2.336= 0- 342 etc. -6
0----~-'1
Similarly we approximate the roots in the other
intervals. Fig. 156.
2. Method of tangents (Newton's method). Again, let f (x < 0, 1

)
f (x.) > 0; and on the interval [x., x.] the first derivative does not
change sign. Then there is one root of the equation f (x) = 0 in the
interval (x., x2 ). Let us assume that the second derivative does not
change sign in the interval [x., x.] either; this can be achieved by
reducing the length of the interval within which the root lies.
Retention of the sign of the second derivative on the interval
[x., x.] means that the curve is either only convex or only
concave on [x., x.].
Draw a tangent to the curve at the point 8 (Fig. 157). The
abscissa a. of the point of intersection of the tangent with the
x-axis will be an approximate value of the root. To find this
abscissa write the equation of the line tangent at the point 8:
y- f (x.) = f' (x.) (x- x.).
Noting that x =a, at y = 0, we have
f (xa)
al = x . - f' (Xa) (3)
8*
22f:l The Curvature of a Curve
Then, drawing the line tangent at the point B 1 , we analogously

find a mor~ exact value of the root a 2 By repeating this procedure
y B
A
Fig: 157. Fig. 158.
we can calculate the approximate value of the root to any desired

degree of accuracy.
Note the following. If we drew the line tangent to the curve
not at the point B but at A, it might appear that the point of
intersection of the tangent with the x-axis
lies outside the interval (xp x 2 ).
From Figs. 157 and . 158 it follows that
the tangent should be drawn at the end
of the arc at which the signs of the func-
tion and its second derivative coincide.
~H...,--.tr----'?"::?--~x Since it is given that on the interval [x,.
x,] the second derivative retains its sign,
the signs of the function and the second
Fig. 159.
derivative must coincide at one of the end
points. This rule also holds for the case
when f' (x) < 0. If the line tangent is
drawn at the left end point of the interval, then in formula (3)
we must put X1 in place of x 2 :
f (x 1) ( ')
a. =X 1 - f' (x.)' 3
When there is a point of inflection C in the interval (x 1 , x 2 ),
the method of tangents can yield an approximate value of the
root lying without the interval (x 1 , X 2 ) (Fig. 159).
Example 2. Apply formula (3) to finding the root of the equation
f(x)=x 8 -6x+2=0
within the interval (0, I). We have
/(0)=2, {'(0)=(3x2 -6)Jx=o=-6,
Exercises on Chapter VI 229
and so from (3) we get
3. Combined method (Fig. 160). Applying at the same time on

the interval [xJI X 2 ] the method of chords and the method of tan-
gents, we get two points a 1 and !I
a. 1 lying on either side of the B
desired root a, since f (aJ and
f (aJ have different signs. Then,
on the interval [aJI il 1 ] again
apply the method of chords and
the method of tangents. This
yields two numbers: a 2 and a,,
which are still closer to the
value of the root. We continue 0
Xz X
in this manner until the difference
between the approximate val ties
found is less than the required
degree of accuracy.
It will be noted that in the
combined method we approach Fig. 160.
the sought-for root from two
sides simultaneously (i. e., at the same time we approximate
the root with an excess and with a deficit).
To illustrate, in the case we have examined it will be clear that by
substitution we have
f (0.333) > 0, f (0.342) < 0.
Hence, the root is between the approximate values obtained:
0.333 <X< 0.342.
Exercises on Chapter VI
Find the curvature of the curves at the indicated points:
b
1. b 2x 2 +a y
2 2
= a2 b 2 at the points (0, b) and (a, 0). Ans. if at (0, b);
a
b2 at (a, 0).
24
2. xy = 12 at the point (3, 4). Ans.
124
3. y = x at the point (x 1 , y 1). Ans. Gx 1
(I +9x~ )' 12
I
4. 16y 2 = 4x'-x 6 at the point (2, 0). Ans.
2 .
..!_ I .!_
5. x + y =a at an arbitrary point. Ans.
3 (axy)
Find the radius of curvature of the following curves at the indicated
points; draw each curve and construct the appropriate circle of curvature.
. 8o
6. y 2 =x' at the potnt (4, 8). Ans. R = - -- .
10
3
7. x2 = 4ay at the point (0, 0). Ans. R = 2a.
(b 4x 1 + a 4y 1) ' '
1
8. bzx-azy=a 1b2 at the point (x 1 , y 1). Ans. R= ab .
9. y=lnx at the point (1, 0). Ans. R=2 2.
10. y =sin x at the point ( ~ , 1) . Ans. R = I.
x=acos't}
11. l't for t=t 1 Ans. R=3aslntcost.
y=asn
Find the radius of curvature of the indicated curves:
1
X=3l }
12. lf= 3t-t for l=l. Ans. R=6.
13. Circle Q=asin6. Ans. R=-i.

. (Q" + a)./1
14. Spiral of Archimedes Q=a6. Ans. R= Q2 +2a 2 '
15. Cardioid Q=a(1-cos 6). Ans. R=f }"2aQ.

a
16. Lemniscate Q2 =a cos 26. Ans. R=a
Q
6 6
17. Parabola Q=a sec
2 . Ans. R=2a sec'
2 .
18. Q=asin 6
3 . Ans. R={- astn {.
Find the points of curves at which the radius of curvature is a minimum:
19. y=lnx. Ans.
2
(~ ,
-{ ln2).
20. y=e". Ans. ( - ~ ln2, ~).

21. Vx+Y!i=Yi Ans. (~.f)
22. y =a In ( 1-~) . Ans. At the point (0, 0) R =-i.
Find the coordinates of the centre of curvature (a, ~) and the equation of
the evo1ute lor each of the following curves:
x 11 (a 2 + b2 ) x (a 1 + b 2 ) y'
23. Qi-IJz= I. Ans. a= a ; ~ =- b'
I 2 2 I
~=y+3x
1
24. x' +Y' =a'. Ans. a=x+3x' y; y'.
Exercises on Chapter VI 231
.a'+l5y4 _ ay-9y5 .
Ans. a= 6ay ' ~= 2a .
X=3t, 4 3
26. { Ans. a=--t; ~=31 2 --.
y=t 2 -6. 3 2
t
x=k In col
27. {
2 -kcost, (tractrix).
y= k sin t.
x =a (cos t + t sin t),
28. { y=a (sin t-t cost). Ans. a=acost; ~=asint.
{ x =a cos t,
29
y=asint. Ans. a= a cos t + 3a cost sin 2 t;
~ = a sin' t + 3a cos 2 t sin t.
30. Find the roots of the equation x'-4+2=0 to three decimal places.
Ans. x 1 = 1.675, x 2 = 0.539, x 3 = - 2.214.
31. For the equation f (x) =x5 -x-0.2 = 0, approximate the root in .the
interval (I. !.1). Ans. 1.045.
32. Evaluate the roots of the equation x'+ 2x 2 -6x + 2 = 0 to two decimal
places. Ans. 0.38 < x 1 < 0.39; 1.24 < x 2 < 1.25.
33. Solve the equation x'-5=0 approximately. Ans. x 1 ::::::: 1.71,
-!i3
x2 , 3 =l.71 .
2
34. Approximate the root of the equation x-tan x=O lying between 0
3n
and . Ans. 4.4935.
2
35. Evaluate the root of the equation sin x = !-x to three places of deci-
mals. Hint. Reduce the equation to the form f(x) =0. Ans. 0.5110 < <x
< 0.5111.
Miscellaneous Problems
36. Show that at each point of the lemniscate Q2 = a 2 cos 2<p the curvature
is proportional to the radius vector of the point.
37. Find the greatest value of the radius of curvature of the curve
Q=a sm.,<p . Ans. R = 3 a.
3 4
38. Find the coordinates of the centre of curvature of the curve y = x In x
at the point where y' =0. Ans. (e- 1 , 0).
39. Prove that for points of the spiral of Archimedes Q =a<p as <p-+ oo the
magnitude of the difference between the radius vector and the radius of cur-
vature approaches zero.
40. Find the parabola y=ax 2 +bx+c, which has common tangent and cur-
vature with the sine curve y =sin x at the point ( i , I) . Make a drawing.
x2
nx n2
Ans. u = - 2 + 2 + ' - s
41. The function y = f (x) is defined as follows:
f(t)=x in the interval-co<x,.;;;;l,
f(x)=ax 2 +bx+c in the interval 1 <x<+co.
What must a, b and c be for the line Y= f (x) to have continuous curvature
everywhere? Make a drawing. Ans. a= 3, b = - 3, c = 1.
42. Show that the radius of a curvature of a cycloid at any one of its
points is twice the length of the normal at that point.
43. Write the equation of the circle of curvature of the parabola y=x
at the point (I, 1). Ans. (x+4) 1 y
+ ( y- ~ = 1 ~.
the point ( : , I). Ans. ( x- n 4 lOr+ (y- :r
44. Write the equation of the circle of curvature of the curve y=tanx at
2
= \ :.
45. Find the length of the entire evolute of an ellipse whose semi-axes
4 (a1 -b 1 )
are a and b. Ans. ab .
46. Find the approximate value of the roots of the equation xe~=2 to
within O.Dl. Ans. The equation has only one real root, x::::::: 0.84
47. Find the approximate value of the roots of the equation x in x = 0.8
to within 0.01. Ans. The equation has only one real root, x::::::: 1.64.
48. Find the approximate value of the roots of the equation x2 arc tan x = 1
to within 0.001. Ans. The equation has only oRe real root, x::::::: 1.096.
CHAPTER VII
COMPLEX NUMBERS. POLYNOMIALS
SEC. l. COMPLEX NUMBERS. BASIC DEFINITIONS
A complex number is the expression

a+ bi (1)
where a and b are real numbers, i is the so-called imaginary unit,
which is defined by the equalities
i= v
-1 or i. = - l; (2)
a is called the real part, and bi, the imaginary part of the complex
number. Two complex numbers a+ bi and a- bi that differ
only in the sign of the imaginary part are called conjugate.
If a= 0, the number 0 + bi = bi is called a pure imaginary; if
b=O, we get a real number: a+Oi=a.
We agree upon the two following ba:;ic statements:
+ +
l) two complex numbers a 1 b 1 i and a 2 b.i are equal if
a 1 =a 2 , b1 =b.,
that is, if their real parts are equal and their imaginary parts are
equal;
2) a complex number is equal to zero:
a+ bi=O
if and only if a=O, b=O.
1. Geometric representation of complex numbers. Any complex
number a+ bi may be represented in an xy-plane as a point A (a, b)
with coordinates a and b (Fig. 161); and conversely, any point
M (a, b) in an xy-plane may be regarded as the geometric image
of a complex number a+ bi.
But if to each point A (a, b) there corresponds a complex
number a+ bi, then, to take a specific case, to points lying on
the x-axis there correspond real numbers (b = 0). But if a point
lies on the y-axis, it represents a pure imaginary number, since
a= 0. For this reason, when complex numbers are represented in
the plane, the y-axis is called the imaginary axis or axis of
imaginaries, and the x-axis, the real axis (axis of reals).
Joining the point A (a, b) with the origin, we get a vector OA.
In certaini nstances, it is convenient to consider the vector OA as
the geometric representation of the complex number a+ bi.
234 Complex Numbers. Polynomials
2. Trigonometric form of a complex number. Denote by cp and

r (r ~ 0) the polar coordinates of the point A (a, b) and consider
the origin as the pole and the positive direction of the x-axis, the
polar axis. Then (Fig. 161) we have the
y
A(a,b) familiar relationships:
r
a= r cos cp, b = r sin cp,
b
and, hence, the complex number may be
-~-"---;a:----'--~x ' given in the form
0
a+bi=r(cosq;+isincp). (3)
Fig. 161. The expression on the right-hand side is
called the trigonometric form of a complex
number a+ bi. The quantities r and cp are expressed in terms of a
and b, by the formulas
r= Va + b", cp =arctan.!!..
a
and are called: r, the modulus, cp, the argument (amplitude or
phase) of the complex number a~- bi. . .
The amplitude of a complex number, the angle !p, is considered
positive if it is reckoned from the positive x-axis counterclockwise,
and negative, in the opposite sense. The amplitude cp is obviously
not determined uniquely but to within the accuracy of the term
2nk, where k is any integer.
The modulus r of the complex number a+ bi is sometimes
denoted by the symbol I a+ bi J:
r=la+biJ.
It will be noted that the real number A can also be written in
the form (3), namely:
A = I A I(cos 0 + i sin 0) for A > 0,
A~ I A I (cos rt + i sin rt) for A< 0.
The modulus of the complex number 0 is zero: I0 I= 0. Any
angle cp may be taken for amplitude zero. Indeed, for any angle
cp we have the equality
0=0(coscp+i sincp).
SEC. 2. BASIC OPERATIONS ON COMPLEX NUMBERS
1. Addition of complex numbers. The sum of two complex

numbers a1 +bJ and a,+b,i is a complex number defined by
the equality
+
(a 1 b,i) +(a,+ b,i) = (a 1 +a,)+ (b 1 +b.) i. (I)
Basic Operations on Complex Numbers 235
From (1) it follows that the addition of complex numbers given

in vectors is performed by the rule of the addition of vectors.
2. Subtraction of complex numbers. The difference of two com-
plex numbers a 2 + b 2 i and a,+ b,i is a complex number such that
when it is added to a,+ b,i it yields
a,+b.i. Y
It is easy to see that
(a,+b,i)-(a, + b,i)=(a.-a,)-1-(b.-b,)i.
(2)
It will be noted that the modulus of the
difference of two complex numbers x
V(a,-a,)-1-(b,-b.)' is equal to the
distance between the points representing
these numbers in the plane of the com- Fig. 162.
plex variable (Fig. 162).
3. Multiplication of complex numbers. The product of two com-
plex numbers a,+b,i and a.+b.i is a complex number obtained
when these two numbers are multiplied as binomials by the rules
of algebra, provided that
i'=-1; i 1 =(-1)i=-i; i4 = (-i)(i)=-i'=1; i~=li etc.,
and, generally, for integral k,
i 4 k=1; i 4 k+ 1 =i; i 4 k+=-1; i 4 k+3=-i.
From this rule we get
(a,+ b,i) (a,+ b,i) =a, a,+ b,a,i + a,b,i + a,b.i',
or
(a,+ b,i) (a,+ b,i) =(a, a,- b,b,) + (b,a, + a,b,) i. (3)
If the complex numbers are written in trigonometric form, we
have
r 1 (cos cp, + i sin cp,) r (cos cp. + i sin cp.) =
cr.
= r ,r .lcos cp, cos + i sin !p, cos cp. + i cos cp, sin IJl. + i 2 sin IJl, sincp.] =
= r ,r 2 ( (COS cp 1 COS Cjl 2 - sin <p 1 sin cr.) -f- i (sin cp 1 COS cp 2 -f- COSIJl 1 sin!p 2 )] =
= r ,r [cos (cp, + 1Jl 2 ) + i sin (cp, +cr.)].
Thus,
r, (cos cp, + i sin cp,) r. (cos 1Jl 2 + i sin cp 2 ) =
=r,r, [cos(IJl, +cp,)-1-i sin (cr, +cr.)]. (3')
the product of two complex numbers is a complex number, the
modulus of which is . equal to the product of the moduli of the
factors, and the amplitude is equal to the sum of the amplitudes

of the factors.
Note 1. By virtue of (3), the conjugate numbers a+bi and
a- bi satisfy the equality
(a+ ib) (a- ib) = a +b 1
;
the product of conjugate complex numbers is equal to the sum of

the squares of the moduli of each of them.
4. Division of complex numbers. The division of complex
numbers is defined as the inverse operation of multiplication: if
a1 +b 1t
-+b .=X
+ yt.
az zl
(where V a:+ b:=t=O), then x and y must be such as to fulfil the

equality
or
Consequently,
whence we find
(4)
Actually, complex numbers are divided as follows: to divide

a 1 + ib 1 by a 2 + ib 2 , multiply the dividend and divisor by a num-
ber conjugate to the divisor (that is, by a 1 -ib 1 ). Then the divisor
will be a real number; dividing the real and imaginary parts of
the dividend by it, we get the quotient
a 1 + b1i (a 1 + b 1i) (a 2 -b 2 i} (a 1a 2 + b 1 b 2 ) +(a 2 b 1 -a 1 b 2 ) i
a 2 +b 1i=(a 2 +b 2 i)(a 2 -b 2i) = a 2 +b 2 =
a,a,+ b 1b2 a 2 b1 -a 1 b 2
= a2 +b' i a2 +b 2 t.
For the trigonometric form of a complex number we have
r 1 (cosq> 1 +isin2)=r 2 COS q>.-q>z +t SID q>,-q>, '
Basic Operations on Complex Numbers 237
To verify this equality, multiply the divisor by the quotient:

r, (cos q> 2 + i sin q>,) C!.. [cos (q>,- q>,) + i sin (q>,-q> 2 )] =
'
. r 2 C!.. [cos (1p 2
r2
+ q> 1
- q>,) + i sin (!p, + q>,- q>,)] = r, (cos q>, + i sin q>, ).
Thus, the modulus of the quotient of two complex numbers is
equal to the quotient of the moduli of the dividend and the divisor;
the amplitude of the quotient is equal to the difference between
the amplitudes of the dividend and divisor.
Note 2. From the rules of operations involving complex num-
bers it follows that the operations of addition, subtraction, multi-
plication and division of complex numbers yield a complex number.
If the rules of operations on complex numbers are applied to
real numbers, regarding the latter as a special case of complex
numbers, these rules will coincide with the ordinary rules of
arithmetic.
Note 3. Returning to the definitions of a sum, difference, pro-
duct and quotient of complex numbers, it is easy to show that if
each complex number in these expressions is replaced by its con-
jugate, then the results of the aforemen~ioned operations will yield
conjugate numbers. Whence follows (as a particular instance) the
following theorem.
Theorem. If in a polynomial with real coefficients
Aoxn + A,xn-t + ... -1- An
we put the number a+ bi in place of x, and then the conjugate
number a- bi in place of x, the results of these substitutions will
be mutually conjugate.
SEC. 3. POWERS AND ROOTS OF COMPLEX NUMBERS
l. Powers. From formula (3) of the preceding section it follows

that if n is a positive integer. then
[r (cos q> + i sin q>W = rn (cos n q> + i sin n q>). (1)
This formula is called De Moivre's formula. It shows that when
a complex number is raised to a positive integral power the
modulus is raised to this power, while the amplitude is multiplied
by the exponent.
Now consider another application of De Moivre's formula.
Setting r = 1 in this formula, we get
(cosq>+i sin q>)n =cos nq> + i sin n q>.
Expanding the left-hand side in a binomial expansion and
equating the real and imaginary parts, we can express sin nq> and
cos ncp in terms of the powers of sin q> and cos q>. For instance, if
n=3 we have
cos cp+ i 3 cos q> sin q>-3 cos q> sin' q>- i sin' cp =cos 3 cp+ i sin 3cp;
making use of the condition of equality of two complex numbers,
we get:
cos 3cp = cos' q>- 3 cos q> sin q>,
sin 3q>=- sin'q> + 3 cos q> sin q>.
2. Roots. The nth root of a complex number is another complex
number whose nth power is equal to the radicand, or
vr (cos q> + i sin q>) = Q (cos 'I'+ i sin '\j)),
if
rl (cos n 'ljJ +i sin n 'ljJ) =--= r (cos q> + i sin q>).
Since the moduli of equal complex numbers must be equal,
while their amplitudes may differ by a number that is a multiple
ol 2n, we have
Whence we find
n/- h-_IP+2k1t,
Q= V r, "' n
where k is any integer, vr

is the arithmetic (real positive)
value of the root of the positive number r. Therefore,
V r (cos cp+ i sin q>) = ;Yr (cos 1P+n2kn + i sin 1P+n2kn) . (2)
Giving k the values 0, 1, 2, ... , n- 1, we get n different
values of the root. For the other values of k, the amplitudes will
differ from those obtained by a number which is a multiple of 2.n,
and, for this reason, root values will be obtained that coincide
with those considered.
Thus, the nth root of a complex number has n different values.
The nth root of a real nonzero number A also has n values,
since a real number is a special case of a complex number and
may be represented in trigonometric form:
ifA>O, then A=IAI (cosO+isinO);
if A< 0, then A= I A I (cos .1t + isin n).
Example I. Find all the values of t'he cube root of unity.
Solution. We represent unity in trigonomet~ic form;
+
1 = cos 0 i sin 0.
Powers and Roots of Complex Numbers 239
By formula (2) we have

fj
V -I = v
cos 0 + 1 sm 0 =
.. _ 0+2kn: . 0+2kn:
cos--
3
+ 1 sln-3- .
Setting k equal to 0, 1, 2, we find three values
of the root: A .
X
X1 = COS 0 + r Stn. 0 = 1; X 2 = COS
2,;
3
'
+I Sin
21t
3 ;
41t ' . 41t
x, =COS 3 +I Stn 3.
Fig. 163.
Noting that
cos
211: 1
; sin 3
2n 3 .
= -- . cos
4n
3
1
-2; . 4n: V3
3 = -
2 2
= Slfi 3 = - -2- '
we get
x 1 = 1: x 2 =
3
- 21 + l --; x1 = -2 -
1 3 .
l --
2 2
In Fig. 163, the points A, B, C are geometric representations of the roots
obtained.
3. Solution of a binomial equation. An equation o[ the lorm

x"=A
is called a binomial equation. Let us find its roots.
If A is a real positive number, then
x = }YA (cos ~n + i sin ~n:)
2 2
(k=O, 1, 2, ... , n-1).

The expression in the brackets gives all the values of the nth
root of l.
If A is a real negative number, then
x= vn/1-AI ( c o1ts
+-2kn: + t. sm
. 1t-+ 2kn:)
n - n- .
The expression in the brackets gives all the values of the nth
root of -1.
If A is a complex number, then the values of x are found lrom
formula (2}.
Example 2. Solve the equation
x'=l.
Solution.
4/
x = 11 cos 2k n + 1. sm
. 2k 2krr.
n =cosT+ . I 2krr.
1s n T .
Setting k equal to 0, I, 2, 3, WE' get
x 1 = cos 0+ i sin 0 = 1,
2n 2rc
X2 =Cos
4 +tsin ;r=i,
x1 =cos 44n +tsm. 4n
4 =-1,
6n 6n
x4 = cos + r sin = -I.
4 4
SEC. 4. EXPONENTIAL FUNCTION WITH COMPLEX EXPONENT

AND JTS PROPERTIES
Let z = x + iy. If x and y are real variables, then z is called a

complex variable. To each value of the complex variable z in the
xy-plane (the complex plane) there corresponds a definite point
(see Fig. 161).
Definition. If to every value of the complex variable z, out of
a certain range of complex values, there corresponds a definite
value of another complex quantity w, then w is a function of the
complex variable z. The functions of a complex argument are
denoted by w = f (z) or w = w (z).
We introduce the concepts of the limit of a function of a com-
plex variable, of the derivative, of the integral, and so forth.
Here, we consider one function of a complex variable, the
exponential function:
or
The complex values of the function w are defined as follows:*)

eX+iy = e" (cosy+ i sin y), (I)
that is
w (z) = e" (cosy+ i sin y). (2)
Examples:
e +- =e(cos;-+isln:)=e(~2 +i ~2 ).
1 1
I. z=1+: i,
) The advisability of this definition of the exponential function of a

complex variable will also be shown later on, Sec. 21, Ch. XIII, and Sec. 18,
Ch. XVI.
Exponential Function with Complex Exponent 241
o+ ~I
2 . z= 0 + n2 t, e z =eo( cos n + tsn
.1 n ) =1,.
2 2
3. Z= I+ i, e'+'=e0 (cos I+ i sin I)=0.54+ i0.83,
. 4. z=x is a real number.
+
ex+ 0Y =ex (cos 0 i sin 0) =ex is an ordinary exponential function.
Properties of an exponential function.

I. If z, and Z2 are two complex numbers, then
(3)
Proof. Let
then
ez,+z 3 = e(x,+jg1 )+(Xz+iY 3 ) =e(X 1 +X 3 )+1 (y 1 +y 3 ) =
=exex, [cos (y, +Y 2 )+i sin (y 1+ Y2)]. (4)
On the other hand, by the theorem of the product of two complex
numbers in trigonometric form we will have
ez,ez, = ex,+iYeXz+IYz = ex, (cosy, + i sin yl) ex (cos Yz + i sin y2) =
=exex [cos(y,+y 2 )+isin(y 1 +Y 2 )l. (5)
In (4) and (5) the right sides are equa], hence the left sides are
equal too:
2. The following formula is similarly proved:
(6)
3. If m is an integer, then
(ez)m =emz. (7)
For m > 0, this formula is readily obtained from (3); if m < 0,
then it is obtained from formulas (3) and (6).
4. The identity
(8)
holds.
Indeed, from (3) and (1) we get
eZH1tl = eZeJrtl = tf (COS 2Jt + i Sin 2Jt) = ez.
From identity (8) it follows that the exponential function ez is a
periodic function with a period of 2ni.
5. Let us now consider the complex quantity
W= u (x) + iu (x),
where u (x) and v (x) are real functions of the real variable x.
This is the complex function of a real variable.
a) Let there exist the limits
lim u (x) = u (x 0 ), lim v (x) = v (X 0 ).
X-+ Xo X-+ X 0
Then u (x 0 ) + iv (x 0 ) = W 0 is called the limit of the complex variable w.

b) If the derivatives u' (x) and v' (x) exist, then we shall call
the expression
w~ = u' (x) + iu' (x) (9)
the derivative of the complex function of a real variable with
respect to a real argument.
Let us now consider the following exponential function:
W = eax+i~X = e(a+/~) X,
where a and ~ are constant real numbers, and xis a real variable.
This is a complex function of a real variable, which function may
be rewritten, according to (1), as follows:
w=eu [cos ~x+i sin ~xJ
or
w = eu cos ~X+ ieax sin ~x:
Let us find the derivative w~. From (9) we have
w~ = (eax cos ~x)' + i (e"" sin ~x)' =
= e""(acos~x-~siri~x)+ie"" (us in ~x+~ cos~x)=
=a [ e"'' (cos ~x+i sin ~x)] +i~ [eax (cos~x+isin~x) J =
= (a+i~) [eax (cos~x+i sin~x)]=(a+i~)e 1~+i~> ".
To summarise then, if W= ec+ipl" then w' =(a+ i~) e 1 +i~>" or
[e 1 -+;~, ")'=(a+ i~) ec+i~> ". ( 10)
Thus, if k is a complex number (or, in the special case, a real
number) and x is a real number, then
(ekx)' = kek". (9')
We have thus obtained the ordinary formula for differentiation of
an exponential function.
Further,
and for arbitrary n

(ekx)cm = knlx.
We shall need these formulas later on.
Euler's Formula. The Exponential Form of a Complex Number 243
SEC. 5. EULER'S FORMULA. THE EXPONENTIAL FORM

OF A COMPLEX NUMBER
If we put x= 0 in formula (1) of the preceding section, we get

eiy= cosy+ i siny. (I)
This is Euler's formula, which expresses an exponential function
with an imaginary exponent in terms of trigonometric functions.
Replacing y by -y in (1) we get
e-iy =cosy- i sin y. {2)
From (1) and (2) we find cosy and siny:
cosy= eiY +e-iY
2
, }
. eiY_e-iY (3)
Sin y= 2i
These formulas are us.ed, among other things, to express the pow-
ers of cos <p and sin <p and their products in terms of the. sine
and cosine of multiple arcs.
eiY +e-iY)z 1 12. .
Examples: 1. cos 2 y= ( = - (e Y+2+e-'"Y)=
. 2 4
= i [(cos 2y+ i sin 2y) + 2 +(cos 2y-i sin 2y)J=
=-;r1 (2 cos 2y+2)=2"1 (1 +cos 2y).

.
2.. cos2 q> sm1 q>=
(ei'f + e-i'f) (ei'f_e-i'f)
i =
2 2
(ei'f-e-iZ'f)2 1 1
= 44i 2 = -8 cosIJl+g
The exponential form of a complex number. Let us 'represent a
complex number in trigonometric form:
Z::==r(cos<p+isin<p),
where r is the modulus of the complex number and <p is the am-
plitude of the complex number. By Euler's formula,
cos <p + i sin <p = ei'f.
Thus, any complex number may be represented in the so-called
exponential form:
z=rei'f.
Examples. Represent the numbers I, i, -2, - i in
'
the exponential
.. form.
Solution. I =cos 2kn + i sin 2kn = ezbi,

It
'= .:os 2 +' sm 2 = e2 I

1t . 1t
-2=2 (cos n+i sin 1t)=2e~i,

It
1t 1t -2 1
2=
0 0
-t=cos -t sin e
2
SEC. 6. FACTORING A POLYNOMIAL
The function
f (x) = A.xn + A xn-l + ... +An,1
where n is an integer, is known as a polynomial or a rational

intPgral function of x; the number n is called the degree of the
polynomial. Here, the coefficients A 0 , AI' ... , An are real or
complex numbers; the independent Variable x can also take on
both real and complex values. The root of a polynomial is that
value of the variable x at which the polynomial becomes zero.
Theorem 1 (Remainder Theorem). Division of a polynomial f (x)
by x- a yields a remainder equal to f (a).
Proof. The quotient obtained by the division of f (x) by x-a
will be a polynomial { 1 (x) of degree one less than that of f(x),
and the remainder will be a constant R. We can thus write
f (x) = (x-a) !1 (x) + R. (I)
This equality holds for all values of x different from a (division
bv x-a when x=a is meaningless).
Now let x approach a. Then the limit of the left side of (I)
will equal f (a), while the limit of the right side will equal R.
Since the functions f (x) and (x- a) f, (x) R are equal for all +
x4=a, their limits are likewise equal as X-+-a, that is, f(a)=R.
Corollary. If a is a root of the polynomial, that is, if f (a)= 0
then x- a divides f (x) without remainder and, hence, f (x) is
represented in the form of a product
f (x) = (x- a) ! 1 (x)
where f 1 (x) is a polynomial.
Example 1. The polynomial f(x)=.x'-6x 1 +llx-6 becomes zeroforx=l;
thus, f (1)=0, and so x-1 divides this polynomial without remainder:
+
x 3 -6x2 llx-6=(x-l) (x2 -5x+6).
Let us now consider equations in one unknown, x.
Any number (real or complex) which, when substituted into the
equation in place of x, converts the equation into an identity is
. called a root of the equation.
Euler's Formula. The Exponential Form of a Complex Number 245
n 5n 9n
Example 2. The numbers x 1 =
4
; x2 =
4 ; x1 =
4 ; ... , are the roots of
the equation cos x =sin x.
If the equation is of the form P (x) = 0, where P (x) is a poly-

nomial of degree n, it is called an algebraic equation of degree n.
From the definition it follows that the roots of an algebraic equa-
tion P (x) = 0 are the same as are the roots of the polynomial P (x).
Quite naturally the question arises: Does every equation have
roots?
In the case of rtonalgebraic equations, the answer is no: there are
nonalgebraic equations which do not have a single root, either real
or complex; for example, the equation e" = 0. *)
But in the case of an algebraic equation the answer is yes. This
is given by the fundamental theorem of algebra.
Theorem 2 (Fundamental Theorem of Algebra). Every rational
integral function f (x) has at least one root, real or complex.
The proof of this theorem is given in higher algebra. Here we
give it without proof.
With the aid of the fundamental theorem of algebra it is easy
to prove the following theorem.
Theorem 3. Every polynomial of degree n may be factored into
n linear factors of the form x-a and a factor equal to the
coefficient of xn.
Proof.. Let f (x) be a polynomial of degree n:
+
f (x) = A 0 Xn A 1 Xn-t + ... +An.
By virtue of the fundamental theorem, this polynomial has at
least one root; we denote it by a . Then, by the corollary of the
remainder theorem, we can write
f (x) = (x-a.)f. (x)
where f. (x) is a polynomial of degree n- 1; f. (x) also has a root.
We designate it by a . Then
f. (x) = (x-a.) f. (x)
where f. (x) is a polynomial of degree n-2. Similarly,
f. (x) = (x-a,) f. (x).
*) Indeed, if the number. x 1 =a+bi were the root of this equation, we
would have the identity ea+"'=O or(by Euler'sformula)ea(cosb+isinb)=O.
But ea cannot equal zero for any real value of a; neither is cos b + i sin b equal
to zero (because the modulus of this number is V cos b :1- sin 2 b =I for any
b). Hence, the product ea(cosb+isinb)::;!=O, i.e., ea+ 1"t=O; but thismeans
.that the equation e" =0 has no roots.
Continuing this process of factoring out linear factors, we arrive

at the relation
fn-t (x)=(x-an)fn
where fn is a polynomial of degree zero, i. e., some fixed number.
This number is obviously equal to the coefficient of xn; that is,
fn= Ao.
On the basis of the equalities obtained we can write
f (x) = A 0 (x-aJ (x-a.) ... (x-an) (2)
From the expansion (2) it follows that the numbers a,, a 2 , , an
are roots of the polynomial f (x), since upon the substitution X= a,,
x = a 2 , , , x =an the right side, and hence, the left, becomes zero.
Example 3. The polynomial f(x)=x 8 -bx2 +iix-6 becomes zero when
x= I, x=2, x=3.
Therefore,
x 8 -6x2 + llx-6= (x-1) (x-2) (x-3).
No value x =a that is different from at' a 1 , , an can be a
root of the polynomial f (x), since no factor on the right side of
(2) vanishes when x=a. Whence the following proposition.
A polynomial of degree n cannot have more than n distinct roots.
But then the following theorem obtains.
Theorem 4. If the values of two polynomials of degree n, q> 1 (x)
and q> 2 (x), coincide for n+ I distinct values a 0 , a,, a 2 , , an of
thP argument x, then these polynomials are identical.
Proof. Denote the difference of the polynomials by f (x):
f (x) = q> 1 (x) -cp. (x).
It is given that f (x) is a polynomial of degree not higher than
n that becomes zero at the points a,, ... , an. It can therefore be
represented in the form
f (x) = A 0 (x-a,) (x-a.) ... (x-an).
But it is given that f (x) also vanishes at the point a0 Then f (a 0 )=0
and not a single one of the linear factors equals zero. For this
reason, A 0 = 0 and then from (2) it follows that the polynomial
f(x) is identically equal to zero. Consequently, q>,(x)-q> 2 (x)=O
or q> 1 (x) = cp. (x).
Theorem 5. If a polynomial
P(x)=A 0 xn+A,xn-t+ ... +An_ 1 x+An
is identically equal to zero, all its coefficients equal zero.
The Multiple Roots of a Polynomial 247
Proof. Let us write its factorisation using formula (2):

P (x)=.4 0 xn+A,xn-'+ ... +An_,x+An=A 0 (x-a,) ... (x-an) (1 ')
If this polynomial is identically equal to zero, it is also equal to
zero for some value of x different from a,, ... , an. But then none
of the bracketed values x-a 1 , , x-an is equal to zero, and,
hence, A.= 0.
Similarly it is proved that A,= 0, A!= 0, and so forth.
Theorem 6. If two polynomials are identically equal, the coeffi
cients of one polynomial are equal to the corresponding coefficients
of the other.
This follows from the fact that the difference between the
polynomials is a polynomial identically equal to zero. Therefore,
from the preceding theorem all its coefficients are zeros.
Example 4. If the polynomial ax 1 +bx"+cx+d is identically equal tothe
polynomial x-5x, then a=O, b=l, c=-5, and d=O.
SEC. 7. THE MULTIPLE ROOTS OF A POLYNOMIAL
If, in the factorisation of a polynomial of degree n into linear

factors
(1)
certain linear factors turn out the same, they may be combined,
and then factorisation of the polynomial will yield
f(x)=A 0 (x-a,)k(x-a 2 )k ... (x~am)km. (1')
And
k 1 +ka+ ... +km=n.
In this case, the root a, is called a root of multiplicity k,. or a
k,-tuple root, a 1 , a root of multiplicity k 2 , etc.
Example. The polynomial f (x) =x 1 -5x" +Bx-4 may be factored into the
following linear factors:
f (x) = (x-2) (x-2) (x-1).
This factorisation may be written as follows:
f (x)=(x-2) 2 (x-1).
The root a1 = 2 is a double root, a2 = 1 is a simple root.
If a polynomial has a root a of multiplicity k, then we will

consider that the polynomial has k coincident roots. Then from
the theorem of factorisation of a polynomial into linear factors
we get the following theorem.
Every polynomial of degree n has exactly n roots (real or
complex).
Note. All that has been said of the roots of the polynomial
f(x)=A 0 xn+A,xn-t+ ... +An,
may obviously be formulated in terms of the roots of the
algebraic equation
A 0 Xn+A,xn-t+ ... + .. .
An=O.
Let us further prove the following theorem.
Theorem. If, for the polynomial f (x), a, is a root of multiplicity
k,>1, then for the derivative f' (x) this number is a root of
multiplicity k,- 1.
Proof. If a, is a root of multiplicity k,>l, thenitfollowsfrom
formula (l ') that
f (x) = (x- a,)kcp (x)
where cp (x) = (x-- a.)k ... (x- am)km does not become zero at x =a,;
that is, cp (a,) =1= 0. Differentiating, we get
f' (x) = k, (x- a.)k - <p (x) + (x- a,)k11p' (x) =
= (x-a.)k- [k,cp (x) + (x-a,)cp' (x)].
Put
' (x) = k,cp (x) + (x- a,) cp' (x).
Then
f' (x) = (x-a,)k,-' 'ljJ (x)
and here
'ljJ (a,)= k,\fi (a,)+ (a, -a,) cp' (a,)= k,cp (a,) =1= 0.
In other words, x =a, is a root of multiplicity k,-1 of the
polynomial f' (x). From the foregoing proof it follows that if k, = 1,
then a, is not a root of the derivative f' (x).
From the proved theorem it follows that a, is a root of multi-
plicity k -2 for the derivative f"(x), a root of multiplicityk,-3
for the derivative f'" (x) ... , and a root of multiplicity one (simple
root) for the derivative [<k,-,, (x) and is not a root for the deri-
vative f<k,, (x), or
f(a)=O, f'(a,)=O. r<a,)=O, ... , f<k,-t>(a,)=O,
but
SEC. 8. FACTORISATION OF A POLYNOMIAL IN THE CASE

OF COMPLEX ROOTS
In formula (1), Sec. 7, Chapter VII, the roots a,, a2 , , an may
be either real or complex. We have the following theorem.
Theorem. If a polynomial f (x) with real coefficients has a complex
root a+ bi, it also has a conjugate root a- bi.
Factorisation of a Polynomial 249
Proof. Substitute, in the polynomial f (x), a+ ~i in place of x,

raise to' a power and collect separately terms containing i and
those not containing i; we then get
{(a+ bi)= M +Ni,
where M and N are expressions that do not contain i.
Since a+ bi is a root of the polynomial, we have
f(a+ bi) =M +Ni =0
whence
M=O, N=O.
Now substitute the expression a- bi for x in the polynomial.
Then (on the basis of Note 3 at the end of Sec. 2 of this chap-
ter) we get a number that is a conjugate of the number M +Ni, or
{(a-bi)=M-Ni.
Since M = 0 and N = 0, we have f (a-bi) = 0; a- bi is a root of
the polynomial.
Thus, in the factorisation
f (x) = A 0 (x-a }(X-a 1 2
) ... (x-an)
the complex roots enter as conjugate pairs.
Multiplying together the linear factor:s that correspond to a
pair of complex conjugate roots, we get a trinomial of degree two
with real coefficients:
[x- (a+ bi)] [x-.(a- bi)] =
= [(x-a)- bi] [(x-a) + bi] =
= (x-a) +b =X -2::zx +a 2 + b 2 =X 2 +px+ q,
2 2 2
+
where p= -2a, q=a 2 b 1 are re_al numbers.
If the number a+ bi is a root of multiplicity k, the conjugate
number a- bi must be a root of the same multiplicity k, so that
factorisation of the polynomial will yield the same number of
linear factors x-(a+bi) as those of the form x-(a-bi).
Thus, a polynomial with real coefficients may be factored into
real factors of the first and second degree of corresponding
multiplicity; that is,
f (x) = A 0 (x- a.)k (x -a 2
)k ..
, .. (x-ar)kr(x'+p.x+q 1 ) 1 (x'+psx+qsY'
where
SEC. 9. INTERPOLATION. LAGRANGE'S INTERPOLATION FORMULA
Let it be established, in the study of some phenomenon, that

there is a functional relationship between the quantities y and x
which describes the quantitative
y aspect of the phenomenon; the
function y = q> (x) is unknown,
but experiment has established
the values of this function y 0 ,
y" y,, ... , Yn for certain values
of the argument X 0 , x., x., ... ,
xn, in the interval [a, b].
Yo Yt Yn The problem is to find a func-
tion (as simple as possible from
0 a x0 x1 Xn b X the computational standpoint; for
Fig. 164.
example, a polynomial) which
would represent the unknown
function y = q> (x) on the interval
[a, b] either exactly or approximately. In more abstract fashion
the problem may be formulated as follows: given on the interval
[a, b) the values of an unknown function Y=q>(x) at n+ 1
distinct points X0 , Xl' ... , Xn: .
Yo.= q> (xo), y; = q> (xt), ... Yn = q> (xn);
it is required to find a polynomial P (x) of degree ~ n that ap-
proximately expresses the function q> (x).
For such a polynomial, it is natural to take a polynomial whose
values at the points X0 , xt' x,, ... , xn coincide with the corre-
sponding values Yo y" y., ... , Yn of the function q> (x) (Fig. 164).
Then the problem, which is called the "problem of interpolating
a function", is formulated thus: for a given function q> (x) find a
polynomial P (x) of degree ,:;;; n, which, for the given values of
x 0 , x" ... , xn, will take on the values
Yo= q> (xo), Yt = q> (xJ, Yn = IJl (xn)
For the desired polynomial, take a polynomial of degree n of
the form
P(x)=C 0 (X-X 1 )(x-x.) ... (x-xn)+
+ cl (x-xo) (x-x,) ... (x-xn) +
+C. (X-X 0 }(X-X 1 ) (X-X1 ) (x-xn) + ...
. . . +Cn (X-X 0 ) (X-X 1 ) (x-xn_J (1)
and define the coefficients C0 , C1 , , Cn so that the following
Interpolation. Lagrange's Interpolation Formula 251
conditions are fulfilled:

P (X 0 ) = y 0 , P (x,) = yl' ... , P (xn) = Yn (2)
I, {1) put x=x.; then, taking into account equality (2), we get
Yo= C 0 (x 0 -x,) (x 0 - X 2 ) (x 0 -Xn),
whence
C = Yo
o (x0 -x 1 )(x0 -x1 ) ... (x 0 -Xn)
Then, setting x = x,, we get
y,=C,(x,-x.)(x,-x.) ... (x,-xn),
whence
c t
=
(x 1 -x 0) (x, -x 1 )
y,
(x1 -xn)

In the same way we find

C1 = Ya
(x1 -x 0) (x1 -x 1 ) (x1 -x1 ) (x1 -xn)
C = Yn _
n (Xn -x0 ) (Xn-X 1 ) (Xn-X 1 ) (Xn -Xn- 1 )
Substituting these values of the coefficients into (1), we get
p (x) = (x-x,) (x-x1 ) (x-xn) Y +
(x0 -x 1 ) (x 0 -x 2 ) (x 0 -xn) 0
+ -,---(x,_,('-x-x
_x-7)-i-(x_-_x.,>_._.._,(7-x_x"-"n)c..,. y-,-
0
0
)(x -x1 2 ) (x -xn) '
1
1
+ (x-x (x-x 0) 1) (x-x 3 ) (x.-x,.)

(x -x (x -x
2 0
) 2 1) (x 2 -x0 ) (x 2 -Xn) Y. + ''
+ (x-x (x-x 0) 1 ) (xn-Xn-tl
(3)
' ' (xn -x0 ) (xn-X 1 ) (xn-Xn_ 1 ) Yn
This formula is called the Lagrange interpolation formula.

Let it be noted, without proof, that if cp (x) has a derivative of
the (n+ l)st order on the interval [a, b), the error resulting from
replacing the function. q> (x) by the polynomial p (x), i. e., the
quantity R (x) = cp (x)- P (x), satisfies the inequality
IR (x) I< I (x-x.) (x-x,) . (x-xn) I (n ~ l) 1 max I cp 1"+'l (x) /.
Note. From Theorem 4, Sec. 6, Ch. VII, it follows that the
polynomial P (x) which we found is the only one that satisfies
the given conditions.
Example. From experiment we get the values of the function y = qJ (x):

=3 for x 0 =1. 1} 1 =-5 for X 1 =2, 1}2 =4 for x 2 =-4.
1} 0
It is required to represent the function y = qJ (x) approximately by a
polynomial of degree two.
Solution. From (3) we have (for n = 2):
(x-2) (x+4) (x-I) (x+4) (x-I) (x-2)
p (x) =(1-2) (I +4) 3 + (2-1) (2 +4} (- 5)+ (- 4-1) (- 4-2)
4
or
39 2 123 252
P (x)=- 30x - w x + w
SEC. 10. ON THE BEST APPROXIMATION OF FUNCTIONS

BY POLYNOMIALS. CHEBYSHEV'S THEORY
A natural question follows from what has been discussed in the

previous section: If a continuous function (j) (x) is given on the
closed interval [a, b], can this function be represented approxi-
mately in the form of a polynomial P (x) to any preassigned de-
gree of accuracy? In other words, is it possible to choose a poly-
nomial P (x) such that the absolute difference between (j) (x) and
P (x) at all points of the interval [a, b] should be less than any
preassigned positive number e? The following theorem, which we
give without proof, answers this question in the affirmative.*)
Weierstrass' Approximation Theorem. If- a function (j) (x) is con-
tinuous on a closed interval [a, b ], then for every e > 0 there
exists a polynomial P(x) such that lf(x)-P(x)j<e. for every x
in the interval.
The outstanding Soviet mathematician Academician S. N. Bern-
stein gave the following method of direct construction of such
polynomials that are approximately equal to the continuous func-
tion (j) (x) on the given interval.
Let (j) (x) be continuous on the interval [0, l]. We write the
ex pression n
Bn (x) = L (j) ( : ) c~xm (1-x)n-m.

m=o
Here, C'/: are binomial coefficients, (j) ('~) is the value of the
given function at the point X= m The expression Bn (x) is an nth
n
degree polynomial called the Bernstein polynomial.
*) It will be noted that the Lagrange interpolation formula (see (3) Sec. 9]
cannot yet answer this question. Its values are equal to those of the function
at the points x 0 , x 1 , x2 , . . . , Xn, but they may be very far from the values
o! the function at other points of the interval [a, b].
Exercises on Chapter VII 253
If an arbitrary e>O is given, one can choose a Bernstein poly-

nomial (that is, select its degree n) such that for all values of x
on the interval [0, 1), the following inequality will be fulfilled:
IBn (x)-<p (x) I< e.
It should be noted that consideration of the interval [0, 1), and
not an arbitrary interval [a, bj, is not an essential limitation of
generality, since by changing the variable X=a+t (b-a) it is
possible to convert any interval [a, b) into [0, 1 ]. In this case,
the nth degree polynomial will be transformed into a polynomial
of the same degree.
The creator of the theory of best approximation of functions by
polynomials is the brilliant Russian mathematician P. L. Cheby-
shev (1821-1894). In this field, he obtained the most profound
results, which exerted a great influence on the work of later mathe-
maticians. Studies involving the theory of articulated mechanisms,
which are widely used in machines, served as the starting point
of Chebyshev's theory. While studying these mechanisms he arrived
at the problem of finding, among all polynomials of a given
degree with the leading coeffi~ient equal to unity, a polynomial
of least deviation from zero on the given interval. He found these
polynomials, which subsequently became known as the ChebyshPv
polynomials. They possess many remarkable properties, and at
present are a powerful tool of investigation in many problems of
mathematics and engineering.
Exercises on Chapter VII
t. Find (3+5i)(4-1). Ans. 17+17i. 2. Find (6+Ili)(7+3i). Ans. 9+9.5i.

3-i 7 19.
3. Find
4
+ Si. Ans.
41 - 41 1. 4. Find (4 -7i)'. Ans. -524 + 1i.
5. F .d
tn r 1 A ns. )1"
,;-7"" l+i . 6. F"d
tn r - 5 - 121. Ans. (2 - 3"'11 . 7 R e-
,;-
2
duce the following expressions to trigonometric form: a) I +i.
Y2 (cos~+ isin ~ "V2 7
:n 7
:n) .
Ans.
8. Find V"f. Ans. i+ r; ) ; b) 1-i. Ans.
-i; i - {
(cos
4
+ i sin
sions in terms of powers of sin x and cos x: sin 2x, cos 2x, sin 4x, cos 4x,
4
3 . 9. Express the following expres-
sin 5x, cos 5x. to. Express the following in terms of the sine and cosine of
multiple arcs: cos 2 x, cos x, cos' x, cos 8 x; sin2 x, sin' x, sin 4 x, sin5 x. 11. Divide
{(x)=x'-4x 2 +Bx-1 by x+4. Ans. f(x)=(x+4) (x 2 -8x+40)-161, that
is, the quotient is equal to x 2 -8x+40; and the remainder is f(-4)=-161.
12. Divide {(x)=x 4 +12x'+54x"+108x+81 by x+3. Ans. f(x)=
=(x+3)(x'+9x"+27x+27). 13. Divide f(x)=x 1 - l by x-1. Ans. f(x)=
8 8
= (x-1) (x +x +x+x' +x" +x+ 1).
Factor the following polynomials: 14. f (x) =x-I. Ans. {(x) =

=(x-1)(x+1)(x 2 +1). 15. f(x)=x 2 -x-2. Ans. f(x)=(x-2)(x+1).
16. /(x)=x 1 +1. Ans. /(x)=(x+1)(x"-x+l).
17. Experiment yielded the following values of y as a function of x:
y,= 4 for x,=O,
Yz = 6 for x 1 = 1,
Ya=10 for x3 =2.
Represent (approximately) the function by a second-degree polynomial.
Ans. x+x+4.
18. Find a polynomial of degree four that takes on the values 2, 1, -1, 5,
. 3 129 1
0 for x=1, 2, 3, 4, 5, respecbvely. Ans. x-17x+ x -92x+35.
2 2
19. Find a polynomial of the lowest possible degree that takes on the
values 3, 7, 9, 19 for x=2, 4, 5, 10, respectively. Ans. 2x-1.
20. Find the Bernstein polynomials of degree l, 2, 3. and 4 for the func-
Bdx)= 3 r3
tion y =sin nx on the interval [0, 1). Ans. B.(x) =0; 8 2 (x) = 2x (1-x);
x(1-x); 8 (x)=2x(1-x)[(2 y2 -3)x 2 -(2 V2 -3)x+ V2].
4
CHAPTER VIII
FUNCTIONS OF SEVERAL VARIABLES
SEC. 1. DEFINITION OF A FUNCTION OF SEVERAL VARIABLES
When considering a function of one variable we pointed out

that in the study of many phenomena one encounters functions
of two and more independent variables. Some examples follow.
Example 1. The area S of a rectangle with sides of length x and y is
S=xy.
To each pair of values of x and y there corresponds a definite value of
the area S. S is a function of two variables.
Example 2. The volume V of a rectangular parallelepiped with edges of
length x, y, z is expressed by the formula
V=xyz.
Here, V is a function of three variables, x, y, z.
Example 3. The range R of a shell fired with initial velocity v 0 from a
gun, whose barrel is inclined to the horizon at an angle q>, is expressed by
the formula
R= v! sin 2q>
g
(air resistance is disregarded). Here, g is the acceleration of gravity.
For every pair of values of v 0 and q> this formula yields a definite value
of R; in other words, R is a function of two variables, v0 and q>.
Example 4.
u-
x"+ y'+z'+ t
- Yl +x
Here, u is a function of four variables x, y, z, t.
Definition 1. If to each pair (x, y) of values of two independent
variable quantities x and y (from some range D) there corresponds
a definite value of the quantity z, we say that z is a function of
the two independent variables x and y defined in D.
A function of two variables is symbolically given as
z = f (x, y), z = F (x, y) and so forth.
A function of two variables may be represented, for example,
by means of a table or analytically (by a formula) as in the
four examples given above. The formula may be used to construct
2')6 Functions of Several Variable~
a table of values of the function for certain number pairs of the

independent variables. From Example 1 we can build the
following table:
S=xy
;~I 0
I I
I
1.5
I
2
I 3
I 0 I 1.5 2 3
:1 0 2 3 4 6
3 0 3 4.5 6 9
4 0 4 6 8 12
In this table, the intersections of the lines and columns, which

correspond to definite values of x and y, yield the corresponding
values of the function S.
If the functional relation z-:- f (x, y) is obtained as a result of
changes in the quantity z in some experimental study of a phe-
nomenon, we straightway get a table defining z as a function of
two variables. In this case, the function is specified by the table
alone. .
As in the case of a single independent variable, a function of
two variables does not, generally speaking, exist for all values of
x and y.
Definition 2. The collection of pairs (x, y) of values of x and
y, for which the function
z =f(x, y)
is defined, is called the domain of definition of this function.
The domain of a function is apparent when illustrated geomet-
rically. If each number pair x and y is given as a point M (x, y)
in the xy-plane, then the domain of definition of the function
will be a certain collection of points in the plane. We shall also
call this collection of points the domain of definition of the func~
tion. In particular, the entire plane may be the domain. In future
we shall mainly have to do with such domains as are parts of
the plane bounded by lines. The line bounding the given domain
we shall call the boundary of ihe domain. The points of the do-
main not lying on the boundary we shall call interior points of
the domain. A domain consisting solely of interior points is
called an open domain; that which includes the points of the
boundary is called a closed domain.
Definition of a Function of Several Variables 257
Example 5. Determine the natural domain of definition of the function

z=2x-y.
The analitic expression 2x-y is meaningful for all values of x and y.
Therefore, the entire xyplane is the natural domain of the function.
Example 6. z= YJ-x2-y2.
For z to have a real value it is necessary that the radicand be a nonne-
gative number; in other words, x and y must satisfy the inequality
l-x 2 -y 2 ;a. 0, or x2 +y 2
,.;;;; I.
All the points M (x, y) whose coordinates satisfy the given inequality lie
in a circle of radius I with centre at the origin and on the boundary of
this circle.
Example 7. Y
z=ln(x+y).
Since logarithms are defined only for positive
numbers, the following inequality must be
satisfied:
x+y>O or y>-x.
This means _that the natural domain of
'definition of the function z is the half-pi ane
above the straight line y = -x, the line itself
not included (Fig. 165).
Example 8. The area of the triangle S is a
function of the base x and the altitude y:
xy
S=2
Fig. 165.
The domain of this function is x > 0, 11 > 0
(since the base of a triangle and its altitude
cannot be negative or zero). We notice that the domain of this function
does not coincide with the natural domain of definition of the analytic
expression used to define the function, because the natural domain of
the expression x: is obviously the entire xy-plane.
It is easy to generalise the definition of a function of two

variables to the case of three or more variables.
Definition 3. If to every collection of values of the variables
x, y, z, ... , u, t there corresponds a definite value of the vari
able w, we shall then call w the function of the independent vari-
ables x, y, z, ... , u, t and write w=F(x, y, z, ... , u, t) or
w=f(x, y, z, u, t), and so on.
Just as in the case of a function of two variables, we can
speak of the domain of definition of a function of three, four and
more variables.
To take an example, for a function of three variables, the do-
main of definition is a certain collection of number triples (x, y, z).
Let it be noted that each number triple is associat~d with some
point M (x, y, z) in xyz-space. Consequently, the domain of
9 - 3388
258 Functions of Severa(Variables-.
-definition of a function of three variables is some collection of

points in space.
Similarly, one can speak of the domain of definition of a func-
tion of four variables u = f (x, y, z, t) as of a certain collection
of number quadruples (x, y, z, t). However, the dotnain of
definition of a function of four or a larger number of variables no
longer permits of a simple geometric interpretation.
Example 2 gives a function of three variables defined for all
values of x, y, z.
In Example 4 we have a function of four variables.
Example 9.
W= Vl-x2-y2-z2-u2.
Here w is a function of the four variables x, y, z, u defined fot values of
the variables that satisfy the relationship
l-x2-y2-z2-u2:;;;,.0.
SEC. 2. GEOMETRIC REPRESENTATION OF A FUNCTION OF TWO

VARIABLES
We consider the function
z=f(x, y), (I)
defined in the domain G in the xy-plane (as a particular case,
this domain may be the entire. plane), and a system of rec-
z
Fig. 166. Fig. 167.
tangular Cartesian coordinates Oxyz (Fig. 166). At each point (x, y)

erect a perpendicular to the xy-plane and on it lay off a segment
equal to f (x, y).
This gives us a point P in space with coordinates
X, y, Z=f(x, y).
The locus of points P whose coordinates satisfy equation (I)
is the graph of a function of two variables. From the course of
Partial and Total 1ncrement o~ a Function 259
analytic geometry we know that equation (l) pefines a surface

in space. Thus, the graph of a function of two var iahles' is a
surface projected onto the xy-plane in the domain of definition
of the function. Each perpendicular to the xy-plane intersects
the surface z = f (x, y) at not more than one point.
Example. As we know from analytic geometry, the graph of the function
.
z=x 2 +y 2 is a paraboloid of revolution (Fig. 167) .
Note. It is impressible to depict a function of three or more
variables by means of a graph in space.
SEC. 3. PA~TIAL AND TOTAL INC~EMENT OF A FUNCTION
Consider the line of intersection PS of the surface

z = t (x, y)
with the plane y-const parallel to the xz-plane (Fig. 168).
Since in this plane y remains constant, z will vary along the
curve PS depending only on the changes in x. Increase the inde-
pendent variable x by dx; then z will be increased; this increase
is called the partial increment of z with respect to x and it is
denoted by d xz (the segment SS'
in the figure), so that
dxz= f (x+ dx, y)- f(x, y). (I)
Similarly, if x is held constant
and y is increased by dy, then z
is increased, and this increase is
called the partial increment of z
with respect to y (symbolised by
dyz, the segment TT' in the
figure):
dyz=f(x, y+dy)-f(x, y). (2)
The function receives the in- Fig. 168.
crement dyz "along the line" of
intersection of the surface z=f(x, y) with the plane x-const
parallel to the yz-plane.
Final_ly, increasing the argument x by dx, and the argument y
by the tncrement dy, we get for z a new increment dz, which is
called the total increment of the function z and is defined by the
formula
dz=f(x+Ax, y+dy)-f(x, y). (3)
In Fig. 168 dz is shown as the segment QQ'.
9*
260 Functions of Several Variables
It must be noted that, generally speaking, the total increment

is n0t equal to the sum of the partial increments, flz=Pflxz flyz. +
Example. z = xy.
tJ.xz= (x+ tJ.x) y-xy=y tJ.x,
tJ.yZ= X (y+ tJ.Y)-xy= X tJ.Y,
tJ.z=(x+tJ.x) (Y+tJ.y)-xy=y tJ.X+x M+tJ.X tJ.y.
For x=l, y=2, tJ.x=0.2, tJ.y=0.3 we have tJ.xz=0.4, tJ.yz=0.3, tJ.z=0.76.
Similarly we define the partial and total increments of a function
of any number of variables. Thus, for a function of three variables
u=f(x, y, t) we have
6-xu=f(x+flx, y, t)-f(x, y, t),
tlyu=f(x, y+fly, t)-f(x, y, t),
fl 1u =f(x, y, t+M)-f(x, y, t),
flu=f(x+flx, y+tly, t+M)-f(x, y, t).
SEC. 4. CONTINUITY OF A FUNCTION OF SEVERAL VARIABLES

We introduce an important auxiliary concep.t, that of the neigh-
bourhood of a given point.
The neighbourhood, of radius r, of a point M 0 (x 0 , Yo) is the
collection of all points (x, y) that satisfy the inequality
!I
V +
(X-X 0 )" (y-y0 ) 2 < r; that is, the
set of all points that lie inside a circle
of radius r with centre in the point
Mo (Xo, Yo)
If we say that a function f (x, y)
possesses some property "near the point
. (x 0 , Yo)" or "in the neighbourhood of the
point (x 0 , Yo)" we mean that there is a
circle with centre at (X 0 , Yo). at all
points of which circle the given function
0 X possesses the given property.
Fig. 169. Before considering the concept of
continuity of a function of several
variables, let us examine the notion of the limit of a function
of several variables.*) Let there be a function
z= f (x, y)
defined in some domain G of an xy-plane.
Let us consider some definite point M 0 (x 0 , Yo) in G or on its
boundary (Fig. 169).
*) We shall mainly consider functions of two variables, since three and
more variables do not introduce any fundamental changes, but do introduce
additional technical difficulties.
Continuity of a Function of Several Variables 261
Definition 1. The number A is called the limit of the function

as M (x, y) approaches M0 (x 0 , y.) if for every e > 0 there
f (x, y)
is an r > 0 such that for all points M (x, y) for which the
inequality MM.< r is fulfilled we have the inequality
if (x, y)-A J <~::.
If A is the limit of f (x, y) as M (x, y)-+ M 0 (x 0 , y.), then we write
limf(x,y)=A.
X-+Xo
Y-+Yo
Definition 2. Let the point M. (x0 , Y.) belong to the domain of

definition of the function f (x, y). The function z = f (x, y) is
called continuous at the point M 0 (x 0 , Yo) if we have
lim f (x, y) = f (x., Yo), (1)
X_.,Xo
Y-+Yo
and M (x, y) approaches M 0 (x 0 , Yo) in arbitrary fashion all the

while remaining in the domain of the function.
+
Designate x = X 0 ~x, y =Yo+ ~y, then ( 1) may be rewritten as
follows:
lim f (X 0 -!- ~X, Yo+ ~y) = f (X 0 , Yo) (1 ')
llx-+o
/ly-+0
or
lim [f (x 0 +~X, Yo+ ~y)- f (X 0 , Yo)l = 0. (1")
llx-+0
lly-+ 0
We set ~Q= V(~x) 2 + (~Y) 2 (see Fig. 168). As ~x-0 and ~y-0,
~Q-0; and conversely, if ~Q-0, then ~x-+0 and ~y--.0.
Noting further that the expression in the square brackets in (1")
is the total increment of the function ~z, (1") may be rewritten
in the form
lim ~z=O. ( 1'")
A function continuous at each point of some domain is continuous

in the domain.
If at some point N (x 0 , Yo) condition (1) is not fulfilled, then
the point N (x 0 , Yo) is called a point of discontinuity of the function
z=f(x, y). For example, condition (1') may not be fulfilled in the
following cases:
l) z = f (x, y) is defined at all points of a certain neighbourhood
of the point N (x 0 , Yo) with the exception of the point N (x 0 , y 0 )
itself;
!1:62 Functions of Seueral Variables
2) the function z = f (x, y) is defined at all points of a neigh-

bourhood of the point N (x 0 , y 0 ) but there is no limit lim f (x, y);
x-.A'o
V->Vo
3) the function is defined at all points of the neighbourhood of
N (X 0 , y 0 ) and the limit exists: lim f (x, y), but
X-t-Xo
V->Yo
lim f (x, y)=l=f (x 0 , Yo),

X-+Xo
Y->Yo
Example 1. The function

Z=xz+ yz
is continuous for all values of x and y; that is, it is continuous at every
point in the xy-plane.
Indeed, no matter what the numbers x and y, AX and 1\y, we have
L\z=[(x+ L\X) 2 + (y+ M) 2 ]-[x 2 + y') =2x L\x+2y flY+ L\X 2
+ L\Y 2
Consequently,
lim M=O.
L'.X->0
t.y->0
The following is an example of a discontinuous function.

Z= 2xy
xz+y
is dell ned everywhere except at the point x = 0, y = 0 (Figs. 170, 171 ).
z
X d
Fig. 170. Fig. 171.
Let us examine the values of z along the straight line y = kx (k =cons!).

Obviously, along this line
Partial Deriuatiues ol a Function of Seuer.al Variables 263
This means that a function z along any straight line passing through the
origin retains a constant value that depends upon the slope k of the line.
Thus; approaching the origin along different paths we will obtain different
limiting vaiues, and this means that the function f (x, y) has no limit when
the point (x, y) in the xy-plane approaches the origin. Thus, the function is
discontinuous at this point. It is impossible to redefine this function at the
coordinate origin so that it should become continuous. On the other hand, it
is readily seen that the function is continuous at all other points.
SEC. 5. PARTIAL DERIVATIVES OF A FUNCTION OF SEVERAL

VARIABLES
Definition. The partial derivative, with respect to x, of a function

z= f (x, y) is the lim,it of the ratio of the partial increment 6."z,
with respect to x, to the increment 6.x as 6.x approaches zero.
The partial derivative, with respect to x, of the function
z = f (x, y) is denoted by one of the symbols
z:; riC (x, y);

az. ax
ax, at .
Thus, by definition,
iJ~= lit11
0
AxZ = lim f(x+!!x, y)-f(x, y)
X .ax~o fiX 6x~o 8X v
Similarly, the partial derivative, with respect to y, of a function

z = f (x, y) is defined as the limit of the ratio of the partial incre-
ment of the function 6.Yz with respect to y to the increment of 6.y
as 6.y approaches zero. The partial derivative with respect to y
is denoted by one of the following symbols:
z'.
11 1
f.II; dy;
az at
aj;
Thus,
aiJ__:y = lim A.vz = lim t (x, Y + AY)- f (x, 1/)
6.y-+o AY 6.y-+o 11Y
Noting that 6."z is calculated withy held constant, and 6.Y.z with
x held constant, we can formulate the definitions of partial deri-
vatives as follows: the partial derivative of the function z = f (x, y}
with respect to x is the derivative with respect to x calculated on
the assumption that y is constant. The partial derivative of the
function z = f (x, y) with respect to y is the derivative with respect
to y calculated on the assumption that x is constant.
It is clear from this definition that the rules for computing
partial derivatives coincide with the rules given for functions of
one variable, and the only thing to remember is with respect to
which variable the derivative is sought.
Example 1. Given the function z = x 2 sin y; find the partial derivatives

az az
ax and ay
Solution.
az 2 az
iJx= XSiny; iJy=X Cosy.
:
Example 2. z=xY,
Here
az
-=yxY-t
ax
az
ay =xY!nx.
The partial derivatives of a function of any number of variables

are determined similarly. Thus, if we have a function u of four
variables x, y, z, t:
u=f(x,y,z,t)
then
au_ Jim f(x+tJ.x,y, Z, i)-f(x, y, Z, t)
ax- Ll.X-+0 flX t
au= lim f (x, Y+fly. z, t)-f(x, y, z, t), and so forth.

ay Ll.X-+0 flY
Example 3.
u =x2 + y 2 +xtz',
au B au au z au 1
i}x=2x+tz; -ay=2y; i}z= 3xtz; i}t =XZ.
SEC. 6. THE GEOMETIUC INTERPRETATION OF THE PARTIAL

DERIVATIVES OF A FUNCTION OF TWO VARIABLES
Let the equation

z=f(x, y)
be the equation of a surface shown in Fig. 172.
Draw the plane x = const. The intersection of this plane with
the surface yields the line PT. For a given x, let us consider a
certain point M (x, y) in the xy-plane. To the point M there cor-
responds a point P (x, y, z) on the surface z = f (x, y). Holding x
constant, let us increase the variable y by !ly = MN = PT'. Then
the function z will be increased by !lyz = TT' (to the point
N (x, y + !ly) there corresponds a point T (x, y lly, z llyz) on + +
the surface z = f (x, y)].
Total Increment and Total DiUerential 265
The ratio l:lyz is equal to the tan- z

l:ly
gent of the angle formed by the
secant line PT with the positive
y-direction:
/:lyZ
l:ly
=tan rlt'.
Consequently, the limit
. /J.yZ az
1tm - = -
/:;.JJ-+0 tJ.y ay
is equal to the tangent of the angle : x
~ formed by the tangent line PB
to the curve PT at the point P
with the positive y-direction:
az
ay = t an ~ Fig. 172.
Thus, the partial derivative ~~ is numerically equal to the tan-

gent of the angle of inclination of the tangent line to the curve
resulting from the surface z = f (x, y) being cut by the plane
x =const.
Similarly, the partial derivative ~ is numerically equal to the
tangent of the angle of inclination a of the tangent line to the
surface z = f (x, y) cut by the plane y = const.
SEC. 7. TOTAL INC~EMENT AND TOTAL DIFFE~ENTIAL
By the definition of the total increment of the function

z = f (x, y) we have (see Sec. 3, Ch. VIII)
6.z=f(x+6.x, y+6.y)-f(x, y). (1)
Let us suppose that f (x, y) has continuous partial derivatives at
the point (x, y) under consideration.
Express 6.z in terms of partial derivatives. To do this, add to
and subtract from the right side of (1) f(x, y+ 6.y):
- 6.z=[f(x+6.x,y+6.y)-f(x, y+6.y)]+[f(x, y+6.y)-f(x, y)]. (2)
The expression
f(x, y+6.y)-f(x, y)
in the second square brackets may be regarded as the difference
between two values of the function of the variable y alone (the
266 Functions of Several Variabllis
value of x remaining constant). Applying tq this difference the

Lagrange theorem, we get
f(x, Y+ fly)-f(x, y)= 6-y at~~ y), (3)
where y lies between y and y+ fly.

In exactly the same way the expression in the first square
brackets of (2) may be regarded as the difference between two
values of the function of the variable x alone (the second argument
retains the same value y+ fly). Applying the Lagrange theorem to
this difference, we have
f(x+flx, y+fly)-f(x, y+fly)=6.xat(i.h+tw>, (4)
where x lies between x and x 6-x. +

Introducing expressions (3) and (4} into (2) we get
(5)
Since it is assumed that the partiai derivatives are continuous,
lim at (i, Y+ l'lY> = at(x, y) .: }
AX-+0 ax ax
~-0 ~)
lim at (x, y) =at (x, y)
Ax-+o ay ay
Ag-+O
(because andx y
respectively lie between x and x 6-x, and y +
and y+ 6-y, x
and y
approach x and y, respectively, as flx-+0
and 6-y.-+ 0). Equalities (6) may be rewritten in the form
at{'X. Y+ tw> =at (x, y)

ax ax
+
'\'t
}
(6')
at(x, y)_at(x, y)+
ay - ay Yz
where the quantities y, and y 1 approach zero as 6-x and fly
+
approach zero (that is, as 6-Q = V flx 2 fly 2 - + 0).
By virtue of (6')', relation (5) becomes
6-z =at (x, y) flx +at (x,
ax ay
y) fly+ y flx
1
+ y fly.
!
(5')
The sum of the two latter terms of the right side is an infini-
V
tesimal of higher order relative to flQ= 6.x 2 fly 2 Indeed, the +
ratio y,tlx -+0 as AQ-+0, since '\'t is an infinitesimal and tlx
i'lQ AQ
Totq,l" Increment and Total Dif!erential 267
is bounded ( j !~I,;;;; 1) . In similar fashion it is verified that

v.!:ly --+0
6.Q .
The sum of the first two terms is a linear expression in !lx
and lly. For t: (x, y)=/=0 and f~ (x, y)=/=0, this expression is the
principal part of the increment, differing from llz by an
infinitesimal of higher order relative to Q = V t.x + lly".
Definition. The function z = f (x, y) [the total increment (llz) of
which at the given point (x, y) may be represented as a sum of
two terms: a linear expression in llx and lly, and an infinitesimal
of higher order relative to llQ] is called differentiable at the given
point, while the linear part of the increment is known as the
total differential and is denoted by dz or df.
From (5') it follows that if the function f (x, y) has continuous ,
partial derivatives at a given point, it is differentiable at this
point and has a total differential:
dz.= t:<x. y) llx+f~ (x, y) lly.
'I
Equality (5') may be rewritten in the form

llz = dz + y,llx +y.lly,
and, to within infinitesimals of higher order relative to llQ, we
may write the following approximate equality:
/:iz-:::::;dz.
We shall call the increments of the independent variables llx
and lly differentials of the independent variables x and y and we
shall denote them by dx and dy respectively. Then the expression
of the total differential will assume the form
at
dz=ax dx +aydy.
at
Thus, if the function z =f (x, y) has continuous partial derivatives,
it is differentiable at the point (x, y), and its total differential is
equal to the sum of the products of the partial derivatives by the
differentials of the corresponding indepeJ;ldent variables.
Example l. Find the total differential and the total increment of the
function z=xy at the point (2, 3) for {).x=O.l, 6.y=0.2.
Solution.
{).z= (x + 6.X) (y + 6.y) -xy = y {).x +x M+ 6.X {).y,
az az
dz=a;;dx+ ay dy =y dx+x dy=y !:lx+x !:ly.
Consequently,
M=30.1 + 20.2+0.10.2 =0.72;
dz=30.1 +2-0.2=0.7.
Fig. 173 is an illustration of this example.
The foregoing reasoning and definitions are appropriately

generalised to functions of any number of arguments.
If we have a function of any number
~~~~~~IJxiJy of variables
w = f (x, y, z, u, ... , t),
and all partial derivatives~ , :~, .. , ~
[Y yiJx. are continuous at the point (x, y, z, u,
... , t), the expression
of of of of
dw= ox dx+ ay dy+ oz dz+ ... +at dt
is the principal part of the total increment
Fig. 173. of the function and is called the total
differential. Proof of the fact that the
difference flw-dw is an infinitesimal of higher order than
V (flx) 2 + (fly) + ... + (flW is conducted in exactly the same way
as for a function of two variables.
Example 2. Find the total differential of the function u =ex"+Y' sln 2 z of
three variables x, y, z.
Solution. Noting that the partial derivatives
au
---ex +y 2xsmz
.
ax- .
:~ =ex'+Y12y sin2 z,
~ =ex'+Y'2 sin z cos z = ex+Y" sin 2z
are continuous for all values of x, y, z, we find that

ou au
du= xdx+ ydy+ zdz=e +
ou x y (2xsln 2 2
zdx+2ysln zdy+sln2zdz).
0 0 0
SEC. 8. APPROXIMATION BY TOTAL DIFFERENTIALS
Let the function z = f (x, y) be differentiable at the point (x, y).

Find the total increment of this function:
!l.z = f (x + flx, y +fly)- f (x, y),
Approxima(ion by Total Differentials 269
whence
f(x+Ax, y+Ay)=f(x,y)+dz. (I)
We had the approximate formula
dz:::::::.dz, (2)
where
iJf iJf
dz =ax dx+ a-yAy. (3)
Substituting, into formula (1), the expanded expression for dz in

place of dz, we get the approximate formula
f(x+Ax, y+dy):::::::.f(x, y)+ 0'~; y) dx+ 0 '~; Yl dy, (4)
to within infinitesimals of higher order relative to Ax and Ay.

We shall now show how formulas (2) and (4) are used for
approximate calc4lations.
Problem. Calculate the volume of material needed to make a cylindrical
glass of the following dimensions (Fig. 174):
radius of interior cylinder R,
altitude of interior cylinder H,
thickness of walls and bottom of glass k.
Solution. We give two solutions of this problem: exact and approximate.
a) Exact solution. The desired volume u is equal to the difference between
the volumes of the exterior cylinder and interior 1
cylinder. Since the radius of the exterior cylinder is I
equal to R+k, and the altitude is H+k, I
I
I
R~
I
or I
I
(5) I
I
b) Approximate solution. Let us denote by f the

volume of the interior cylinder, then f=nR 2H. This is
a function of two variables R and H. If we. increase R
and H by k, then the function f will increase by t:,J; Fig. 174.
but this will be the sought-for volume u, v = llf.
On the basis of relation (I) we have the approximate equality
tl :::::::.df
-or
But since
270 Functions of Se.ver(ll Variables .
wt> get
(6)
Comparing the results of. (5) and (6), we see that they differ by the quan-
tity n(Hk 2 +2Rk'+k'), which consists of terms of second and third order of
smallness relative to k.
Let us apply these formulas to numerical examples.
Let R=4 em, H=20 em, k=O.I em.
Applying (5), we get, exactly,
v = n (24 -20-0.1 +4'0.1 +20-0.l'+ 2-4 -O.J2+ 0.1') = 17.88ln.
Applying formula (6), we have, approximately,
v:::::: n (2-4 -20-0.1 + 4 2 -0.1) = 17 .6n.
Hence, the approximate formula (6) gives an answer with an error less than
0.3n, which is 100 ~~:~n % . which is less than 2% of the measured
quantity.
SEC. 9. ERROR APPROXIMATION BY DIFFERENTIALS
Let some quantity u be a function of the quantities x, y, z, .. , t

u = f (x, y, z, ... , t)
and Jet .there be errors flx, fly, ... , M made in determining the
values of the quantities x, y, z, ... , t. Then the value of u
computed from the inexact values of the arguments will be
obtained with an error
flu=f(x+~x, y+fly, ... , z+flz, t+M)-f(x, y, z, .. . , t).
Below we shall investigate the evaluation of the error flu, provided
the errors flx, fly, ... , flt are known.
For sufficiently small absolute values of the quantities flx,
fly, ... , M we can replace, approximately, the total increment
by the total differential:
A. iJf A 1 iJt A + +at A(
uUR:! OX uX 1 iJy uy (fi u .
Here, the values of the partial derivatives and the errors of the
arguments may be either positive or negati\'e. Replacing them by
the absolute values, we get the inequality
Iflu I,;;;;; l ax I flx I+ Ja'j

atJ. ay I fly I+ ... + Ja'j
ai I M 1- (I)
If in terms of Jfl*xJ, lfl*yJ, ... , Jfl*ul we denote the

maximum
absolute errors of the corresponding quantities (the boundaries for
the absolute values of the errors), it is obviously possible to take
Jfl*ul=l~\lfl*xJl~Jifl*yJ+ , .. + ]~jlfl*tl. (2)

Error Approxima-tion by Differentials 271
Examples.
I. Let tl=x+y+z. then
I ~u 1=1 ~x 1+1 ~*y 1+1 ~*z 1.
2. Let u=x-y, then
I~ *u I= I~ *xI+ I~ *y I
3. Let u = xy, then
1~ *u I= I x II~ *y I+ IY II~ *xI
4. Let u = .=_
y
, then
I~* ll I = [ ; II ~*X I+ If.\1 ~ . y I =I y II ~*X ~' II ~

I X *y I .
5. The hypotenuse c and the leg a of a right triangle ABC, determined with
maximum absolute errors l~*cj=0.2, l~*al=0.1, are, respectively, c=75,
a=32. Determine the angle A from the formula slnA=~; and determine the
c
maximum absolute error 1~A 1 when calculating the angle A.
Solution. sinA=~. 'A=arcsln~. hence,
c c
oA oA a
iJa vc-a' iJc = c }fc-a'
From formula (2) we get
I~AI Y (75) 21-:-(32) 2 0.1+ 75 Y (75)322 -(32) 2 0.2=0.00275 radian=9'38".
Thus,
A =arc sin~ 9'38".
6. In the right triangle ABC, let the. leg b = 121.56 and the angle
A =2521'40", and the maximum absolute error in determining the leg b is
1 ~ *b 1=0.05 metre, the maximum absolute error in determining the angle A
is I~*A1=12".
Determine the maximum absolute error in calculating the leg a from the
formula a=b tan A.
Solution. From formula (2) we find
l~*al=ltan All~*bi+~I~*Ai.
cos :
Substituting the appropriate values (and remembering that I~* A 1 must be
expressed in radians), we get
I~ a I= tan 25o21 '40" 0.05+ cos1;!~~~'40" 20!2265 =

=0.0237+0.0087=0.0324 metre.
The ratio of the error !u of some quantity to the approximate.

value of x of this quantity is called the relative error of the
quantity. Let us designate it 6x,
6x=t:.x.
X
The maximum relative error of a quantity x is the ratio of the

maximum absolute error to the absolute value of x and is denoted
by \6* X j,
j6*xl=lf;(l. (3)
To evaluate the maximum relative error of a function u, 'divide

all numbers of (2) by Iu I If (x, y, Z, ... ' t) 1:
1
fur 1
= /;f l1 Ll*xl + l~f l1 Ll*yl+ + l:fl, Ll*t 1. (4)
but
at at a!
ax f at a
T=axa In!!; ay a If !;
y=ay!n ... , y,= at In If I
For this reason, (3) may be rewritten as follows:
16* u I= \:x In If I jiLl* xl + jfx In If !\ill *y I+ ... +~~ ln If 1\1 Ll*t 1... ,(5)
or briefly,
l6*ul=lll*lnlfll (6)
From both (3) and (5) it follows that the maximum relative error
of the function is equal to the maximum absolute error of the
loga(ithm of this function.
From (6) follow the rules used in approximate calculations.
1. Let u= xy.
Using the results of Example 3, we get
l 6*u I= I x II t:.*x I+ IY II t:.*y l= I t:.*x I+ I t:.*YI = 16* X J + 16*y I'
I xy I I xy I Ix I IY I '
that is, the maximum relative error of a product is equal to the
sum of the maximum relative errors of the factors.
2. If u = ; , then, using the results of Example 4, we have
16* u I= )6* X I+ 16*y I
The Derivative of a Composite Function 273
Note. From Example 2 it follows that if u =x-y, then

l 6*ui=ld*xl+id*YI.
jx-yl
If x and y are close, it may happen that l6*u I will be very great
compared with the quantity x- y being determined. This should
be taken into account when performing the calculations.
Example 7. The oscillation period of a pendulum is
T=2~-v ~,
where l is the length of the pendulum and g is the acceleration of gravity.
What relative error will be made in determining T when using this for
mula if we take 1t::::::: 3.14 (accurate to 0.005), l =I m (accurate .to 0.01 m),
g= 9.8 mfsec 2 (accurate to 0.02 mfsec 2).
Solution. From (6) the maximum relative error is
I b*T 1=1 d*lnT 1.
But
I 1
In T= In 2+ In n+ In l - 1ng.
2 2
Calculate 1d *InTI . Taking into account that 1t::::::: 3.14, d *1t = 0.005,
1=1 m, d*I=O.Ol m, g=9.8 mfsec 2, d*g=0.02 mfsec 2 , we get
I T= d*:t +d *I+ d*g = 0.005+0.01 + 0.02 =O 0076
d n n 21 2g 3.14 2 29.8
Thus, the maximum relative error is
b*T = 0.0076 = 0. 76%.
SEC. 10. THE DERIVATIVE OF A COMPOSITE FUNCTION.

THE TOTAL DERIVATIVE
Let us assume that in the equation

Z=f(u, v) (1)
u and v are functions of the independent variables x and y:
u = qJ (x, y); v = 'ljl (x, y). (2)
In this case, z is a composite function of the arguments x and y.
Of course, z can be expressed directly in terms of x, y; namely,
z = F (qJ (x, y), 'ljl (x, y)]. (3)
Example 1. Let
then
z = (x2 +y2)a (eX+JI + l)a + (x2+ y2) + l.
274 Functions of Sev'l!ral Variabtes
Now suppose that the functions F (u, v), <p (x, y), ' (x, y) have
continuous partial derivatives with respect to all their arguments;
az az .
and we pose the problem: evaluate ax and ay on the basts of
equations (1) and (2) without having recourse to equation (3).
Increase the argument x by llx, holding the value of y cons-
tant. Then, by virtue of equation (2), u and v will increase by
/)."11 and l::i"v.
But if u and v receive increments llxu and llxv, then the func-
tion z = F (u, v) will receive an increment llz defined by formula
(5), Sec. 7, Ch. VI II:
ap ap
llz = au llxu + av Axv + 'Yt llxu -+- 'Yzllxv.
Divide all terms of this equality by llx:
~_~~xU+~ flxV + flxU+ flxV
flX- au flX av flX 'Yt flX 'Yz flX
If llx- 0, then llxu-+ 0 and llxv _,. 0 (by virtue of the conti-
nuity of the functions u and v). But then y1 and 'Yz also approach
zero. Passing to the limit as llx-0, we get
. l lflZ
1I az . I' flxu au. I' flxV
-=-, Ill-=-, Ill-=-,
av.
dX-+O flX ax dX-+O /).X ax dX-+O flX ax
limy 1 =0; limy 2 =0
dX-+0 dX-+0
and, consequently,
~-~&+M~ (4)
ax - au ax av ax .
If we increased the variable y by fly and held x constant, then
by similar reasoning we would find that
az=aFau+oFav (4')
ay au ay av ay
Example 2.
z=ln(u'+v); u=e"+Y", v=x 2 +y;
az 2u iJz I
au=u+v; av=u 2 +v;
iJU 1 au -_ OV OV
--e"+Y
QX -
---2ye"+Y
iJy- 0 0 -=2x ay
ax -=1.
I
Using formulas (4) and (4') we find

~ = ~ e"+Y 1 + u- 2 +v
ax u 2 +v
1 2
- 2x = - 2 - (ue"+Y" +x)
u +v
az 2u x y
+ +--= - - (2uye"+Y + 1),
I I
-'- = --2ye
ay u+v u+v u +v 2
The Derivative .of a Composite Function 275
Formulas (4) and (4') are. rea.dily generalised to the. case of a

larger number of variables.
For example, if w = F (z, u, v, s) is a function of four argu-
ments z, u, v, s, and each of them depends on x imd y, then
formulas (4) and (4') assume the. form
(5)
If a function is given z = F (x, y, U, v), where .y, u, v in turn

depend on a single independent variable (argument) x:
y=f(x); u=cp(x); v='ljl(x),
then z is actually a function only of the one variable x, and we

may pose the que~tion of finding the derivative :: .
This derivative is calculated from the first of the formulas (5):
~ = az ax+ az ay + az au+ az av .
dx ax ax. ay ax au ax av ax
But since y, u, v are functions of x alone, the partial derivatives

become ordinary derivatives; in addition aax = 1. For this reason,
X
~-!!!. . az dy
dx- ax -r- aydx
+ auaz ~ + ~ dv
dx av dx . (6)
This formula is known as the formula for calculating the total

. t.we dx
derwa dz ('m con t ras t t o th e par t'ta I d enva
. t1ve az)
ax .
Example 3.
z=x 2 + Yy; y=sinx,
az az I dy
ax=2x; ay=2 Yy; dx=COSX.
Formula (6). here, yields the following result:
dz dz iJz d y I . _ I
d-=-a +-a d-=2x+ .r- cosx-2x+ y cosx.
X X Y X 2 y y 2 Sin X
SEC. 11. THE DERIVATIVE OF A FUNCTION DEFINED

IMPLICITLY
Let us begin this discussion with the implicit function of one

variable.*) Let some function y of x be defined by the equation
F(x, y) =0.
We shall prove the following theorem.
Theorem. Let a continuous function y of x be defined impli-
citly by the equation
F(x, y)=O
where F (x, y), F~ (x, y), F~ (x, y) are continuous functions in
some domain D containing the point (x, y) whose coordinates satisfy
equation ( 1); also, at this point F~ (x, y)=/=0. Then the function
y of x has the derivative
F~ (x, y)
Yx =
F~ (x, y)
Proof. Let the value of the function y correspond to some value
of x. Here,
F (x, y) =0.
Increase the independent variable x by !J.x. Then the function y
will receive an increment !J.y; that is, to the value of the argu-
ment x+ !J.x there corresponds the value of the_ function Y+ !J.y.
By virtue of equation F (x, y) = 0 we shall have
F (x + !J.x, y + !J.y) = 0.
Hence
F(x+!J.x, y+!J.y)-F(x, y)=O.
The left member of the latter equality, which is the. total incre-
ment of the function of two variables by formula (5'), Sec. 7, may
be rewritten as follows:
aF aF
F (x + !J.x, y + !J.y) -F (x, y) =ax !J.x + ay !J.y + Y1 !J.x +Yz!J.y,
where y 1 and Yz approach zero as !J.x and !J.y approach zero. Since
the left side of the latter expression is equal to zero, we can
*) In Sec. II, Ch. Ill, we solved the problem of the differentiation of
an imp! ici t function of one variable. We considered individual cases and
did not fwd a general formula that would yield the derivative of an impli-
cit function; likewise we [ailed to clarify the conditions of the existence ol
this derivative.
The Derivative of a Function Defined 1mplicil/y 277
write
aF aF
ax llx + ay lly + Y1 llx + YzllY = 0.
Divide the latter equality by !lx and calculate ~~ r
aF
flY ax +'Yt
/).X=--ar-
ay +'Yz
Let !lx approach zero. Then, taking into account that y 1 and Ya
also approach zero and that ~~=FQ, we have, in the limit,
aF
ax
Yx=- aF. (1)
ay
We have proved the existence of the derivative of a function y;
defined implicitly, and we have found the formula for calcu-
lating it.
Example 1. The equation
xz+yz-1=0
defines y as an implicit function of x. Here,
aF aF
F(x, y)=x 2 +if-1, ax=2x; ay=2y.
Consequently, from (1),

dy 2x x
dx=-2y=--y
It will be noted that the given equation defines two different functions
[since to every value of x in the interval (-1, 1) there correspond two
values of y]; however, the value that we found of y:
holds for both functions.
Example 2. An equation is given that connects x and y:
eY-e" +xy =0.
Here, F(x, y)=eY-e"+xy,
aF x aF
ax=-e +y; ay=eY+x.
Consequently, from formula (1) we get

dy -e"+y ex-y
dx = - eY +x = eY + x'
278 FunctiOns: of'. Several Variables
Let us now consider an equation of the form

F (x, y, z) = 0. {2)
If to each number pair x and yin some domain there correspond
one or several values of z that satisfy equation {2), then this
equation implicitly defines one or several single-valued functions
z of x and y.
For instance, the equation
x+ Y +Z -R
2 2 2
=0
implicitly defines two continuous functions z of x and y, which
functions may be expressed explicitly by solving the equation for
z; in this case we have
z =VR-x-y and Z= -:-V R-x-yz.

Let us find- the partial derivatives ~ and :; of the implicit
function z of x and y defined by equation (2) .
. az . . . . _ ' '
When we seek ax we consider y fixed. And so formula (1) is
applicable, provided x is considered the independent variable and
z the function. Thus,
aF
In the same way we find

BF
. ay
Zy=- tif
az
Similarly, we determine the implicit functions of any number
of variables and find their partial derivatives.
Example 3.
x 2 +y 2 +z 2 -R 2 =0,
Bz 2x X Bz y
ax=-2z=--z. ay=-z-
Differentiating this function as an explicit function (after solving the
equation for z), we would obtain the very same result.
Example 4.
Parttai Deriuatives of Different O~ders 279
Here, F (x, y. z)=ez +x 2 y'4z+5. i!
aF aF iJF
ax =2xy; ay =x; Fz =ez +I;
az 2xy az x
. ax= - e + I ; ay = - e + I '
SEC. 12. PARTIAL DERIVATIVES OF DIFFERENT ORDERS
Let there be given a function of two variables:

z = f(x, y).
az ;:' d az ;:'
The partial derivatives ax= lx (x, y) an ay = ,11 (x, y) are, gene~
rally speaking, functions of the variables x and y. And so from
them we can again find partial derivatives. Thus, there are four
partial derivatives of the second order of a function of two vari-
ables, since each of the functions aaz and aaz may be differentiated
X Y
both with respect to x and with respect to y.
The second partial derivatives are denoted as follows:
:~ = t:x (x,y); here f is differentiated twice successively with
respect to x;
az ,.
axa Y = fxu (x, y); here f is first differentiated with respect to x and
then the result is differentiated with respect to
y;
az ..
ayax = fux (x, y); here f
is differentiated first with respect to y and
then the result is differentiated with respect to x;
az " (x,
ay = fuu y); here the function f is differentiated twice succes-
sively with respect to y.
Derivatives of the second order may again be differentiated
both with respect to x and y. We then get partial derivatives of
the third order. Obviously, there will be eight of them:
az az a z az az az az az
ax ; axay ; axayax ; axay ; ayax ; ayaxay ; ayax ; ay .
Generally speaking, a partial derivative of the nth order is the
first derivative of the derivative of the (n-1) st order. For exam-
ple, ~xP:;~ Pis a derivative of the nth order; here the function
z was first differentiated p times with respect to x, and then

n- p times with respect to y.
For a function of any number of variables, the higher-order
partial derivatives are determined in similar fashion.
Example I. Compute the second-order partial derivatives of the function
f(x, y)=xzy+y'.
Solution. We find successively
:: =2xy; ~y=x2+3yz;
;;! = 2y; a~~~Y =a ~:y) = 2x;
azt
ay ax=
a(x + 3y
2
ax
2
)
2Xi
az
Example 2. Compute axtay and - - i 2 2 az
2 f z=y eX+x y'+l.
ay ax
Solution. We successively find
az
ax2 ay = ye
2 X
+ 6y '
2
ayaz
axz= 2ye"+6y.2
Solution.
au -
ax -Z e
2 X+Y"
'
The natural question that arises is whether the result of differ-

entiating a function of several variables depends on the order of
differentiation with respect to the different variables; in other
words, will, for instance, the following derivatives be identically
equal:
or
ar (x, y, t) and at (x, y, t)
ax ay at at ax ay '
and so forth. It turns out that the following theorem is true.
Theorem. If the function z=f(x, y) and its partial derivatives
f~, f~, fxu and f;x
are defined and continuous at a point M (x, y)
and in some neighbourhood of it, then at this point
azr ar
ax ay= ay ax
Partial Derivatives of Different Orders 281
Proof. Consider the expression

A= [f(x+ ~x, Y+ ~y)-f (x+ ~x, y)]- [f (x, Y+ ~y)-f(x, y)].
If we introduce an auxiliary function cp (x) defined by the
equality
cp(x)=f(x, y+l'ly)-f(x, y),
then A may be written in the form
A = cp (x + l'lx)- cp (x).
Since it is assumed that t:
is defined in_ the neighbourhood of
the point (x, y), it follows that cp (x) is differentiable on the
interval (x, x+ ~x]; but then, applying the Lagrange theorem,
we get
A= ~xcp' (x)
where ; lies between x and x + ~x.
But
cr' <x> = t: <;, u+ ~u>- t: <;. u>.
Since t:Yis defined io the neighbourhood of the point (x, y),
t: is differentiable on the interval (y, y+ ~y]; and so by applying
once again the Lagrange theorem (with respect to the variable y)
to the difference obtained we have
t~ (x, u+ ~u>- t: (x, u) = ~ut:y <x. 0.
where y lies between y and y+ ~y.
Consequently, the original expression of A is
A = l'lxl'lyf:y (x, y). (1)
Changing the places of the middle terms we get
A=[f(x+~x. Y+~y)-f(x, Y+~Y)]-[f(x+l'lx, y)-f(x, y)].
Introducing the auxiliary function
'IJ(y)=f(x+~x. y)-f(x, y),
we have
A = 'I' (y + ~y)- 'I' (y).
Again applying the Lagrange theorem we get
A= ~y1j)'(,y),
where y lies between y and y+l'ly.
282 Functions of Seqeral Variables
But
1jl' (Y)=f~ (x + 11x, Y)- f~ (x, y).
Again applying the Lagrange theorem we get
f~ (x + 11x, y)- f~ (x, y) = 11xf~x (x, y),
where x
lies between x and x+ 11x.
Thus, the original expression of A may be written in the
form
A= !1y !1xf~x (x, y). (2)
The left members 'Of (1) and (2) are equal to A, therefore the
right ones are equal. too; that is,
11x.i::.yt;u (x, Y) = 11y !1xf~x (x, y),
whence
t:u (~ Y> = f~x (x, y).
Passing to the limit in this equality as 11x -+0 and 11y-+ 0,
we get
lim t:u (x, Y) = lim f~x (x, y).
ax -+ 0 l;.x -+ 0 .
liy -+ 0 liy -+ 0
Since the derivatives fxu and fux are continuous at the point
(x, y), we have lim t:u (x, Y) =fxy (x, y)
lix-+ o
liU-+ o _, .;
and lim f~x (x, Y) = f~x (x, y). And finally we gef
lix-+ o
liy -+ 0
f;y (x, y) = f~x (x, y),
as required.
A corollary of this theorem is that if the partial derivatives
anr anr
-.,---'-~ and k k are continuous, then
axk ayn -k ayn- ax
anr - anr
axkayn-k- ayn-k axk"
A similar theorem holds also for a function of any number of
variables.
au au . _ xy I
Example 4. Find . ax ay az and ay az ax I 1 u -e s n z.
Solution.
au au
ax =ye"Y Sin Zi. ax ay =e"Y Sin Z+xye"Y slnz =e"Y (I +xy) slnzj
Level Surfaces 283
as u . au azu
a a a =e"Y (1 +xy) cosz; -a =xe"Y slnz; - - =xe"Y cosz;
X y z y ax az
asu
ay az ax =e"Y COS z +xye"Y cos z =e"Y {1 + xy) cos z.
Hence,
as u au
ax ay az = ay az ax
(also see Examples 1 and 2 of this section).
SEC. 13. LEVEL SURfACES
In a space (x, y, z) let there be a region D in which the

function
U=U(X, y, z) (I)
is defined. In this case we say that a scalar field is defined in
the region D. If, for example, u (x, y, z) denotes the. temperature
at the point M (x, y, z), then we say that a scalar field of tem-
peratures is defined; if D is filled with a liquid or gas and
u (x, y, z) denotes pressure, we have a scalar field of pressures,
etc .
. Consider the points of a region D in which the function
u (x, y, z) has a fixed value c:
u (x, y, z) =c. (2)
The totality of t_~ese points forms a certain surface. If a different
value of c is tak~n. we obtain a different surface. These surfaces
are called level surfaces.
Example t. Let there be given a scalar field
xz yz zZ
u(x, y, z)=T+g+TI;
Here, the level surfaces are
xz yz zZ
4+9+16=c
or ellipsoids with semi-axes 2 V~. 3 Vc, 4 VC.
If the function u is a function of two variables x and y,
U=U(X, y),
then the level "surfaces" are lines on the xy-plane:
u (x, y) =o (2')
which are called level lines.
If we plot values of u on the z-axis:

z = u (x, y),
the level lines in the xy-plane will be projections of Jines obtained
at the intersection of the surface z = u (x, y) with the planes
z~c (Fig. 175). Knowing the
z level lines, it is easy to study
the character of the surface
z = u (x, y).
z
y
!J
X
Fig. 175. Fig. 176.
Example 2. Determine the level lines of the function z = l-x~-y 1 They

are lines with equations l-x 2 -y2 =c, which are (Fig. 176) circles with radius
Yl-c. In particular, when c=O we get the cir~Je x2 +y 2 =1.
''
SEC. 14. DIRECTIONAL DERIVATIVE
In a region D, consider the function u = u (x, y, z) and the

point M (x, y, z). Draw from M a vector S whose direction cosines
are cos a, cos~. cosy (Fig. 177). On the vector S, at a distance
z
s
Fig. 177.
Directional Derivative 285
f).s from its origin, let us consider a point M, (x+ /).x, Y+ f).y, z+ f).z).
Thus,
/).s = V /).xz + f).yz + f).zz.
We shall assume that the function u (x, y, z) is continuous and
has continuous derivatives with respect to their arguments in the
region D.
As in Sec. 7, we will represent the total increment of the
function as follows:
~ ~ ~ ..
f).u = ;;- f).x + ;;- Ay + :o f).z + ~'\ f).x + .8 2 /).Y + eaf).z, (1)
vX vy ~ .
where e., e2 and 81 approach zero as /).s-+ 0. Divide all terms of

(I) by f).s:
t:,.u au t:,.x au t:,.y au t:,.z t:,.x t:,.y t:,.z
LiS= ax t:,.s + ay t:,.s + az t:,.s + e, t:,.s + 8 2 "t:,.s + 8 a t:,.s. (2)
It is obvious that
t:.x t:,.y t:,.z
t:,.s =cos a, t:,.s =cos~. t:,.s =cosy.
Consequently, equation (2) may be rewritten as

f':,.U au au R au R
- = vx
t:,.s
;;- cos a+;;-
vy
cost-'+;;-
vz
cosy+ e, cos a+ 8 2 cost-'+ 83
.
cosy. (3)
The limit of the ratio~: as /).s-+0 is called the derivative of

the function u=u (x, y, z) at the point (x, y, z) along the direction
of the vector S and is denoted by ~~ ; thus
. t:,.u au
11m-=- (4)
&s--+ 0 f':,.s as.
So, passing to the limit in (3), we get

au au au au
as= ax cos a+ ay cos~+ az cosy. (5)
From formula (5) it follows that if we know the partial derivatives

it is easy to find the derivative along any direction S. The
partial derivatives themselves are a particular case of a directional
derivative.
For instan~e, when a= 0, ~ y ~ we get =.; , = ,
au au au 1t au 1t au
ax = ax cos 0 + ay cos 2 + az cos 2 = ax
286: Functwns of Several Variables
Example. Given a function

u=xz+yz+zz.
Find the derivative~~ at the point M (1, I, I): a) along the direction of
$. the vector S,=2l+J+3k; b) along the
2 direction of the vector S1 =l+J+k.
Solution. a) Find. the direction
cosines of the vector S,:
I 3
cos f} = .r-
f 14
cos v= .lOA.
f 14
Hence,
au au .. 2. au I au 3
Fig. 178. as, =ax T4+.~y Vl4 + az VI4.
The partial derivatives at the point M (I. I. I) are
axau =2.t,; ay
au
=2y,
au.
Tz =2z;
(~~ )M= 2 (~~ )M= 2 (aa~ )M= 2

Thus,
1
au 2 2 2 I 2 3 12 '
as, = V14 + Vi!i+ viii= viii
b) Find the direction cosines of the vector S2:
I I I
cos a= v:r . cos ~ = v:r , cos v= y"3 .
Hence,
au I I I
asz = 2. 3 + 2. 3 + 2. 3 = v63 = 2 ,r-3.
f
We note here (and it will be needed later on) that 2 Y3 > .J:__
f 14
(fig. 178).
SEC. 15. GRADIENT
At every point of the region D, in which the function

u = u (x, y, z)
1s given, we determine the vector whose projections
on the coordinate axes are the values of the partial derivatives
Grad lent 287
:~, :; , : of this function at the appropriate point:

au au . au
grad u=a; i + ayJ+ az k. (1)
This vector is called the gradient of the function u (x, y, z). We

say that a vector field of gradients is defined in D. Let us now
prove the following theorem which establishes a relationship
between the gradient and the directional derivative.
Theorem. Given a scalar I field u = u (x, y, z); in this field, let
there be defined a field of gradients
au .
gra d U=ax 1
+ ayJ
au .+ au
az k '
The derivative ~~ along the direction of som~ vector S is equal to
the projection of the vector grad u on the vector S.
Proof. Consider the on it vector S 0 , which corresponds to the
vector. S: . ,, :
S 0 = i cos a +j .cos ~ + k cos.,, 'V.
Find the scalar product of. the vectors grad u and S 0 :
au au .. au .
grad u. so= ax cos a+ ay cos~+ az cos 'V (2)
The expression on fhe right Is a derivative of the functiorr

S. Hence, we can write
u (x, y, z) along the vector
o au
gra d U S =as
If we designate the angle between the vectors grad u and S 0 by
cp (Fig. 179), we can write
au
as
I grad u 1 cos cp = (3}
or
projection S 0 grad u = aaus (4)
. ..
and the theorem is proved.
This theorem gives us a clear picture of the relationship
between the gradient and the derivative, at a given point, along
any direction. Referring to Fig. 180, construct the vector grad u
at some point M (x, y, z). Construct a sphere for which grad u
is the diameter. Draw the vector S from M. Denote by P the
point of intersection of S with the surface of the sphere. It is
then obvious that MP =I grad u I cos cp, if q> is the angle between
the directions of the gradient and the segment MP (here, cp < ~ ) ,
or MP = ~~. Obviously, when the direction of the vector S is
Fig. 179. Fig. 180.

. - ,
reversed the derivative changes sign, while its absolute value

remains unchanged.
Let us establish certain properties of a gradient.
1) The derivative at a given point along the direction of the
vector S has a maximum if the direction of S coincides with that
of the gradient; this maximal value of the derivative is equal to
lgradul.
The truth o.f this assertion follows directly from (3): ~ will be
a maximum when cp = 0, and in this case
au
a~= I grad u I
2) The derivative along a vector that is tangent to a level

surface is zero.
This assertion follows from formula (3). Indeed, in this case,
1t au
cp = 2 , cos cp = 0 and 05 =I grad u I cos cp ~ 0.
Example l. Given the function

u=x"+Y"+z.
a) Determine the gradient at the point M (1, 1, 1). The expression of the
gradient of' this function at an arbitrary point will be
+ +
grad u = 2xl 2yj 2zk.
Hence,
(fi!rad u)M=21+2J+2k, jgrad u IM-2 y3:
b) Determine the derivative of the function u at the poinf M (1, 1, I)
along the direction of the gradient. The direction cosines of the gradient.
Gradient 289
will be
2 1 I
cos a- = - ' cos~= ,r 3 ' cos v= ,r-3 .
- y 2 +2 + 2
2 2 2
3 , y
And so
or
au
iJs =I grad u I
Note. If the function u = u (x, y) is a function of two variables,
then the vector y
au- t"+au
gra d U =OX -J.
iJy
lies in the xy-plane. We shall prove that

grad u is perpendicular to the level line
u (x, y) = c lying in the xy-plane and
passing through the corresponding point.
Indeed, the slope k 1 of the tangent to the 0 X
u'
level line u (x, y) = c will equal k 1 = - ~ Fig. 181.
Uy
u'
(see Sec. 11 ). The slope kz of the gradient is k 2 = ....!',. Obviously,
u.,
k,k 2 = - 1. This proves our assertion Wig. 181). A similar prop-
erty of the gradient of a function of three variables will be
established in Sec. 6 of Chapter I X.
u
X X
gradu
Fig. 182. Fig. 183.
2 2
Example 2. Determine the gradient of the function u= x +Y (Fig. 182)
2 3
at the point M {2, 4).
10-3388
Solution. Here
Hence
8
grad u=2i+yJ.
The equation of the level line (Fig. I83) passing through the given point
will be
SEC. 16. TAYLOR'S FORMULA FOR A FUNCTION

OF TWO VARIABLES.
Let there be a function of two vari abies

z =f (x, y)
which is continuous, together with all its partial derivatives up
lo the (n + 1)st order inclusive, in some neighbourhood of the
point M (a, b). Then, like the case of one variable (see Sec. 6,
Ch. IV), represent the function of two variables in the form of a
sum of an nth degree polynomial in powers of (x-a) and (y- b)
and some remainder. It will be shown below that for the case
of n = 2 this formula has the form
f(x, y)=A 0 +D(x-a)+E(y~b)+
+ 2iI [A (x-a) + 2B
2
.
(x-a) (y-b) + C (y- b) + R 2
] 1, (1)
where the coefficie11ts A 0 , D, E, A, B, C are independent of x

andy, while R2 is the remainder, the structure of which is simi-
lar to the structure of the remainder in the Taylor formula for a
function of one variable.
Let us apply the Taylor formula for a function f (x, y) of one
variable y considering x constant (we shall confine ourselves to
second-order terms):
f(x, Y)=f(x, b)+ 11 I b f~(x, b)+
+ (y~:)~ f~u (x, b)+(~ . 2 ~t f~~Y (x, f] 1 ~, (2)
where TJ 1 = b + 91 (y- b), 0<9 1 <1. Expand the functions f (x, b),
f~ (x, b), f~y (x, b) in a Taylor's series in powers of (x-a), con-
fining yoursell to mixed derivatives up to the third order inclu-
sive:
Taylor's Formula for a Function of Two Variables 291
f (x, b)=
=f(a, b)+x 1 nf:(a, b)+(x 1 .;) f~x(a, b)+~. ~~f~:x(~ 1 , b), (3) 2
where
~.=x+8.(x-a), 0<8.< 1;
f~(x, b)=f~(a, b)+x-;-aJ:x(a, b)+<\. 2a)f~~x( 2 , b), (4)
where
~.=x+ 8.(x-a), 0<8.< 1;
f~u(X, b)= f~u (a, b)+ x 1 a f~~x (~ b), (5)
where
Sa =x+64 (x-a), 0<6 <1. 4
Substituting expressions (3), (4) and (5) into formula (2), we get
x-a ' (x-a) 2 "
f (x, y)=f(a, b)+-1-fx(a, b)+-r:2 fxx(a, b)+
(x-a)1 '" y-b [ ' x-a
+~fxxx(~ 1 b)+-1- fu(a, b)+-1-fux(a, b)+
2 2
x-a "' ]
+--r:r
(x-a)
fuxx (~ b) +-
"' (y-b)]
2-
1 .-
[
fuu (a, b)+ - 1- fuux (~ b) +
(y- b) "'
+~ fuuu(X, TJ).
Arranging the numbers as indicated in formula (1), we get

f(x, y)=f(a, b)+(x-a)f:(a, b)+(y-b)f~(a, b)+
+i, [(x-a) f:x(a, b)+2(x-a)(x-b){x (a, b)+
2
11
+ (y- b) "fuu (a, I [(x-a)' fxxx

b)] +31 "' (~ , b)+
1
+3 (x-a) 2 (x- b) f~;1, (~ 2 , b) +3 (x-a) (y-b) 1 t:~u (6 1 b)+
+(y-b) 1 t:~u(a, TJ)]. (6)
This is Taylor's formula for n = 2. The expression

I "' "'
(~.,b) +3 (x-a) (y-b) fxxy (~ 2 , b)+
2
R 1 =31 [(x-a)fw
+ 3 (x-a)(y- b)" f~~~~ (~ . b) t (y- b)" f~~~~ (a, TJJl
to
is called the remainder. Further, let us denote x-a = Llx, y- b = Lly,

lle=V(Ilx)"+(lly)". Transform R.:
R,=3i1 [L\x"
M" "'
2
L\X L\y
fxxx(~ 1 , b)+3~ xxy(~ 2 , b}-j-
f'"
L\XL\y "' L\y' "'
+3 ~ fxuu (~ b)+ M"
I
fuuu (a, t'J)] ilQ
Since I Llx I< ilQ, ILly 1< L\e and the third derivatives are bounded
(this is given}, the coefficient of Lle" is bounded in the domain
under consideration; let us denote it by a 0
Then we can write
R. = aolle'.
In this notation, Taylor's formula (6) will then, for the case n = 2,
lake the form
f (x, y) = f (a, b)+ Llxf: (a, b)+ Llyf~ (a, b)+
+ ;, [llx t:x (a, b)+ 26.x.!ly t: 11 (a, b)+ Lly fuu (a, b)]+ a 0 ilQ1 {6')
Taylor's formula is of a similar form for arbitrary n.
SEC. 17. MAXIMUM AND MINIMUM OF A FUNCTION

OF SEVERAL VARIABLES ,
Definition t. We say that the function z = f (x, y) has a maxi-

mum at the point M 0 (X 0 , y 0 ) (that is, when X=X 0 and y=y 0 ) if
f (xo, Yo)> f (x, y)
for all points (x, y) sufficiently close to the point (x 0 , y 0 } and
different from it.
Definition 2: Quite analogously we say that a function z = f (x, y)
has a minimum at the point M 0 (X 0 , Yo) if
f (xo, Yo)< f (x, Y)
for all points (x, y) sufficiently close to the point (x 0 , Yo) and
different from it.
The maximum and minimum of a function are called extrema
of the function; we say that a function has an extremum at a
given point if this function has a maximum or minimum at the
given point.
z=(x-1)'+ (y-2) 2 -1
attains a minimum at x = 1, y = 2; i.e., at the point (1, 2). Indeed, f (I, 2) =- 1,
Maximum and Minimum of a Function of Seueral Variabes 25)3
--------------------------------------------~-------
and since (x-1) 2 and (y-2) 2 are always positive for x ;6 1, !J ;6 2, so

(x-1) 2 -t-(y-2) 2 -1> -1,
that is,
f (x, y) > f (1, 2).
The geometric analogy of this case is shown in Fig. 184.
Fig 184.

1
z=
2 -sin (x 2 +y 2
)
for x = 0, y =0 (coordinate origin) attains a maximum (Fig. 185).

Indeed,
1
/(0, 0)=2.
Inside the circle x 2 +y 2

=
6
1t
let us take the point (x, y) different from
the point (0, 0). Then for 0<x 2 +y < 2 ~ ,
sin (x 2 +y )>02
and therefore
or
f (x, y)<f (0, 0).
The definition, given above, of the maximum and minimum of

a function may be rephrased as follows.
Let X=X 0 +Ax; y =Yo+ Ay; then
f(x, y)-f(x., Y.)=f(x.+Ax, Yo+Ay)-f(x., Y0 )=Af.
1) If Af<O for all sufficiently small increments in the independ-
ent variables, then the function f (x, y) reaches a maximum at
the point M (x 0 , y.).
294 Functions of Severai Variables
2) If M>O for all sufficiently small increments in the independ-

ent variables, then the function f(x, y) reaches a minimum at
the point M (x., Y.).
These formulations may be extended, without any change, to
functions of any number of variables.
Theorem 1. (Necessary conditions of an extremum). If a function
z=f(x, y) attains an extremum at X=X 0 , Y=Y. then each first-
order partial derivative with respect to z either vanishes for these
values of the arguments or does not exist.
Indeed, give the variable y a definite value y=y. Then the
function f (x, y.) will be a function of one variable, x. Since at
x=x. it has an extremum (maximum or minimum), it follows
that (~)x=x. is either equal to zero or does not exist. In exactly
y=y.
the same fashion it is possible to prove that (~z)
<iX X=Xo
is either
y=y.
equal to zero or does not exist.
This theorem is not sufficient for investigating the extremal
values of a function, but permits finding these values for cases
in which we are sure of the existence of a maximum or minimum.
Otherwise, more investigation is required.
For instance, the function z=y 2 -x 2 has derivatives ::=-2x; ~;=+2y,

which vanish at x = 0 and y = 0. But for the given values, this function has
neither maximum nor minimum. Indeed, this
function is equal to zero at the origin and
takes on both positive and negative values at
points arbitrarily close to the origin. Hence,
the value zero is neither a maximum nor a
minimum (Fig. 186).
Points at which ~ = 0 (or does not

exist) and ~ = 0 (or does not exist) are
Fig. 186. called critical points of the function
z = f (x, y). If a function reaches an
extremum at some point, then (by virtue of Theorem 1) this can
occur only at a critical point. .
For investigating a function at critical points, let us establish
sufficient conditions for the extremum of a function of two vari
ables:
Theorem 2. Let a function f (x, y) have corttin.uous partia.l deri~
vatives up to order three inclusive in a certain domain containing
the point M 0 (x 0 , y 0 ); in addition, let the point M 0 (x 0 , Yo) be a
Maximum and Minimum of a Function of Several Variables 295
critical point of the function f(x, y); that is,

iJf (Xo, Yo)= 0 iJf (Xo, Yo) _ 0
OX , iJy -
Then for X=X 0 , y=y0 :

1) f (x, y) has a maximum if
at (xo, Yo) at (xo, Yo)_ (at (xo, Yo) ) 2 >0 and aat (xo, Yo) <0
ax ay ax ay . ax ,
2) f (x. y) has a minimum if
at (x 0 , Yo) at (x 0 , Yo) (at (xo, Yo))~>O and at (xo. Yo)> 0
ax . ay ax ay ax ,
3) f (x, y) has neither maximum nor minimum if
at (xo, Yo) 0
at (xo, Yo) (at (xo, Yo) ) 2 < O
ax ay . ax ay ,
4) if at (xo. Yo). at (xo,
Yo) (at (xo, Yo)) = 0 then there may or
. ax ay ax ay .
may not be an extremum (in this case, an additional investigation
is required).
Proof. Let us write the second-order Taylor formula for the
function f (x, y) [Formula (6), Sec. 16]. Assuming
a=X 0 , b=y0 ,. x=x.+-~x. y=Yo+~y.
we will have
f (x 0 + ~x. Yo+ ~y) = f (x., Yo)+ at(~~ Yo) ~X+ at(~~ Yo) ~y +
Yo) ~r+ 2 at(x., Yo) ~x~y+at(xo, Yo) ~y] +a
+ _!_2 [at(xo,
ax ax ay ay 0
(D.n)'
.:
where ~Q = V ~x + ~y and a 0 approaches zero as ~Q--. 0.

It is given that
at (xo, Yo)= 0 at (x., Yo) 0
ax ay .
Hence
(1)
Let us now denote the values of the second partial der ivatlves
at the point M 0 (x 0 ,. Yo) in terms of A, B, C: '
at)
(ax M,.= A.
. .(iJxiJy
at ) Mo = B; (at)
ay Mo =c.
Denote by cp the angle between the direction of the segment M 0 M,

where M is the point M(x.+Ax, Y.+Ay), and the x-axis; then
Ax= Ae cos cp; Ay = Ae .sin cp.
Substituting these expressions into the formula for b.[, we find
M= ; 2
(AQ) 2 [A cos cp + 28 cos cp sin cp + C sin cp + 2a.Ae1. 2
(2)
Sup pose that A + 0.

Dividing and multiplying by A the expression in the paren-
theses, we have
M=; (Ae) 2 [(A cos<p+Bsln<p~+(AC-B )sln 0, A<O. Then in the numerator of the fraction
we have a sum of two nonnegative quantities. They do not vanish
simultaneously because the first term vanishes for tan cp = - ~ ,
while the second vanishes for sin cp = 0.
If A <0, then the fraction is a negative quantity that does not
vanish. Denote it by -m 2 ; then
M = 2I (AQ)
2
[- m 2 + 2a.Ae1.
where m is independent of Ae. a.Ae-+ 0 as Ae-+ 0. Hence, for
sufficiently small Ae we have
b.f<O
or
f(x 0 +Ax, Y.+Ay)-f(x Yo)<O.
But then for all points (x. +Ax, Yo+ Ay) sufficiently close to
the point (x 0 , y 0 ) we have the inequality
f (X 0 +Ax, Yo+ Ay)<f (X 0 , Y0 ),
which means that at the point (x 0 , y 0 ) the function attains a
maximum.
2) Let AC-8 2 >0, A>O. Then, reasoning in the same way,
we get
M = 2I (A e)' [m
2
+ 2ct 0 Ael
or
f (x 0 +Ax, Yo+ Ay)>f (X 0 , y 0 ),
that is, f (x, y) has a minimum at the point (x 0 , y 0 ).
3') Let AC- B 2 <0, A >0. In this case the function has neither
a maximum nor a minimum. The fundion increases when we move
from the point (X 0 , Yo) in certain directions and decreases when
we move in other directions. Indeed, when moving along the ray
0
we have
Llf={(ilQ) 1 [AC;B 2
sin 2 cp 0 +2a 0 ~QJ <0;
when moving along this ray the function decreases.
3") Let AC-W<O, A<O. Here the function again has neither a
maximum nor a minimum. The investigation is conducted in the
same way as for 3'.
3"') Let AC--B 2 <0, A=O. Then B=t=O, and equality (2) may
be rewritten as follows:
Llf= ~ (~Q) 2 [sincp(2Bcoscp+Csiilcp)+2a 0 ~Q].

For sufficiently small values of O and cp<O we can take Q so small that
2a0 will not change the sign of the whole square bracket). Conse-
quently, in this case, too, Llf changes sign for different cp, that
is, for different ~x and l).y; hence, in this case too there is neither
a maximum nor a minimum.
Thus, no matter what the sign of A we always have the follow-
ing situation:
If AC-B"<O at the point (X 0 , y 0 ), then the function has nei-
ther a maximum nor a minimum at this point. In this case, the
surface, which serves as a graph of the function, can, near this
point, have, say, the shape of a saddle (see Fig. 186). The func-
tion at this point is said to have a minimax.
4) Let AC- B 2 = 0. In this case, by formulas (2) and (3), it is
impossible to decide about the sign of Llf. For instance, when
A=i=O we will have
l).f= ~ (I).Q) 2 [(Acos<p18sin<pY+ 2ao/).Q],

when qJ =arc tan ( - ~ ) , the sign of M is determined by the

sign of 2a 0 ; here, a special additional investigation is required
(for example, with the aid of a higher-order Taylor formula or in
some other way). Thus, Theorem 2 is fully proved.
Example 3. Test the lollowing function for maximum and minimum:
z=x'-xy +u'+3x-2y+ l.
Solution. I) Find the critical points
oz oz
ax=2x-y+3; ay= -x+ 2y-2.
Solving the system of equations

2x-y+3=0,}
-x+2y-2=0,
we get
4 1
x=-
3 ; u=
3 .
2) Find the second-order derivatives at the ~ritical point (.:.... ! , ~)
and determine the character of the critical point:
az i}"z az
A=ax=2; B=axay=-1; C=au=2;
AC-8 2 =22-( -1)'=3>0.
Thus, af the point ( - ! , !) the given function has a minimum, namely

4
2 rnin= -3.
Example4. Test for a maximum and minimum the function z=x'+y'-3xy.
Solution. 1) Find the critical points using the necessary conditions of an
extremum:
ax=
oz 3x'-3y = o, }
~~=3y'-3x=0.
Whrnce we get two critical points:
x1 = 1, y1 = 1 and x1 = 0, y1 = 0.
2) Fmd the second-order derivatives:
az
- - - - 3
axay-
3) Investigate the character of the first critical point:
(iJ z) (az)
2 02
A- -6 8- ( z ) --3 c = a----z = 6;
- OX 2 X=l- ' - OXOy X = l - ' Y x=t
Y=l y=t y=l
AC-8 2 =36-9=27>0; A>O.

Hence, at the point (1, I) the given function has a minimum, namely:
Zmin= -I.
4} Investigate the character of the second critical point M 2 (0, 0}:
A =0; 8= -3; C=O;
AC-8'=- 9 < 0.
Henre, at the second critical point the function has neither a maximum nor
a minimum (minimax).
Example 5. Decompose a given positive number a into three positive terms
so that their product is a maximum.
Solution. Denote the first term by x, the second by y; then the third will
be a-x-y. The product of these terms is
U=XY (a-x-y).
It is given that x>O, y>O, a-x-y>O. that is, x+y<a. u>O. Hence, x andy
can as~ume values in the domain bounded by the straight lines x=O, y=O,
x+y=a.
Find the partial derivatives of the function u:
au
ax=y (a-2x-y),
au
iJy =X (a-2y-x}.
Equating the derivatives to zero, we. get a system of equations:

y (a-2x-y)=0; x (a-2y-x)=0.
Solving this system we get the critical points:
XI =0, Yt =0, Mt (0, 0);
x2 =0, y 2 =a, M 2 (0, a};
XJ=a, y 3 =0, M 3 (a, 0};
a a
x,=3 y,=3 M, ( ~ ~).
The first three points lie on the boundary of the region, the last one, inside.
On the boundary of the region, the function u is equal to zero, while inside
it is positive; consequently, at the point ( ~ , ~ ) ,the functi~n u has a max-
imum (since it is the only extremal point inside the triangle). The maximum
value of the product
Investigate the character of the critical points using the sufficiency condi-
tions. Find the second-order partial derivatives of the function u:
ou
2
0 2U 0 2U
i}x2= -2y; OX oy =a-2x-2y; oy'= -2x.
At the point M 1 (0,0) we have

au o2u iJ 1u
A= 0x 2=0;B=axay=a,C= 0 y1 =0,
AC-8 2 = - a 2 <0. Hence, at the point M 1 there is neither a maximum nor
au
a minimum. At the point M 1 (0, a) we have A= x 1 =-2a; B=axay=-a;
au
0
0 1u
c = oy'=O;
Which means that at the point M 2 there is neither a maximum nor a mini
n1um. At the point M 1 (a, 0) we have A=O, B=-a, C=-2a:
AC-8 1 = -a 2 <0.
At M 1 too there is neither a maximum nor a minimum. At the point
~ ~) we have A=- ; ; B ~- ~ ; C = - ~;
2
M4 ( ,
4a 2 a2
AC-B'=g--g-> 0; A< 0.
Hence, at M 4 we have a maximum.
SEC. 18. MAXIMUM AND MINIMUM OF A FUNCTION

OF SEVERAL VARIABLES RELATED BY GIVEN EQUATIONS
(CONDITIONAL MAXIMA AND MINIMA)
In many maximum and minimum problems, one has to find the

extrema of a function of several variables that are not indepen-
dent, but are related to one another. by side conditions (for
example, they must satisfy given equations).
By way of illustration let us consider the following problem.
Using a piece of tin 2a in area it is required to build a closed
box in the form of a parallelepiped of maximum volume.
Denote the length, width and height of the box by x, y, and z.
The problem reduces to finding the maximum of the function
V=XYZ
+ +
provided that 2xy 2xz 2yz = 2a. The prob Iem here deals with
a conditional extremum: the variables x, y, z are restricted by the
+ +
condition that 2xy 2xz 2yz = 2a. In this section we shall con-
sider methods of solving such problems.
Let us first consider the question of the conditional extremum
of a function of' two variables if these variables are restricted by
a single condition.
Let it be required to find the maxima and minima of the func-

tion
u=f(x, y) (1)
with 1.he proviso that x and y are connected by the equation
q> (x, y) = 0. (2)
Given condition (2), of the two variables x and y there will be
only one which is independent (for instance, x) since y is deter
mined from (2) as a function of x. If we solved equation (2) for
y and put into (1) the expression found in place of y, we would
obtain a function of one variable, x, and would reduce the prob
!em to one that would involve finding the maximum and minimum
of a function of one independent variable, x.
But the problem may be solved without solving equation (2) for
x or y. For those values of x at which the function u can have
a maximum or minimum, the derivative of u with respect to x
should vanish.
From (1) we find ~~, remembering that y is a function of x:
du =at+ at dy
dx ax ay dx
Hence, at the points of the extremum

at +~dy _
ax ay dx- 0 (3)
From equation (2) we find
aq> + a<p dy _ 0
ax ay dx- . (4)
This equality is satisfied for all x and y that satisfy equation (2)
(see Sec. 11, Ch. VIII).
Multiplying the terms of (4) by an (as yet) undetermined coef-
ficient f.. and adding them to the corresponding terms of (3),
we have
or
(5)
The latter equality is fulfilled at all extremum points. Choose 'A
such that for the values of x and y which correspond to the extre-
mum of the function u, the second parentheses in (5) should vanish: *)

at+~., acp = o.
ay ay
But then, for these values of x and y, from (5) we have
~~+1..: =0.
It thus turns out that at the extremum points three equations
(with three unknowns x, y, A.) are satisfied:
at+~.,acp=O}
ax ax ,
~+ 1.. acp=O (6)
ay ay '
cp(x, Y)=O.
From these equations determine x, y, and A.; the latter only played
an auxiliary role and will not be needed an more1
From this conclusion it follows that equations (6) are necessary
conditions of a conditional extremum; or equations (6) are satisfied
at the extremum points. But there will not be a conditional extre-
mum for every x and y (and A.) that satisfy equations (6). A sup
plementary investigation of the nature of the critical point is re-
quired. In the solution of concrete problems it is sometimes pos-
sible to establish the character of the critical point from the
statement of the problem. It will be noted that the left-hand sides
of equations (6) are partial derivatives of the function
F (x, y, 1..) = f (x, y) + l..cp (x, y) (7)
with respect to the variables x, y and 1...
Thus, in order to find the values of x and y which satisfy con-
dition (2), for which the function u = f (x, y) can have a condi-
tional maximum or a conditional minimum, one has to construct
an auxiliary function (7), equate to zero its derivatives with re-
spect to x, y, and 1.., and from the three equations (6) thus
obtained determine the sought-for x, y (and the auxiliary factor A.).
The foregoing method can be extended to a study of the condition-
al extremum of a function of any number of variables.
Let it be required to find the maxima and minima of a function
of n variables, u = f (x1 , x .. , xn) provided that the variables
*) For the sake of definiteness, we shall assume that at the critical points
~: ~ o.
x., X2 ; , xn are connected by m (m < n) equations:

<p, (x,, X1 , , Xn): 0, }
<p. (x., x.. .. xn)- 0,
. . . . . . . . . . . . (8)
<Jlm (x" x., ... , xJ = 0.
In order to find the values of x,. x,, .. , xn, for which there
may be conditional maxima and minima, one has to form the
function
F(x,. XI,xn, A.,, .. , A.m>=f(x,, .. , xn)+A.,cp, (x,, ... , xn>+
,
+
A-.<p. (x., . . . xn> + ... +A,
miP rn (x,' ... ' xn>
equate to zero its partial derivatives with respect to x .. X1 , , xn:
~~+A-, ~<p'+ ... +A-m ~ll'm=O, )
vx, ux, ux, I
:~.~.A,::~~.~ ..... ~.A,~~~: .. 0: ~ (9)
~ +A,
OXn , 1
iJ<p, + .. +A,
OXn
O!pm
m OXn
= 0 J
and from the m + n equations (8) and (9) determine x" x., ... , xn
and the auxiliary unknowns A.,, ... , 'A.m. Just as in the case of a
function of two variables, we shall, in the general case, leave
undecided the question of whether the function, for the values
found, will have a maximum or minimum or will have neither.
We will decide this matter on the basis of additional reasoning.
Example l. Let us return to the problem formula ted at the beginning of
this section: to find the maximum or the function
v =xyz
provided that
xy+xz+yz-a=O (x>O, y>O, z>O). (10)
We form the auxiliary function
F(x, y, A.)=xyz+A.(xy+xz+yz-a).
Find its partial derivatives and equate them to zero:
yz +A. (y + z) = 0, }
xz+A.(x+z)=O, (I i)
,xy+A.(x+y)=O.
The problem reduces to solving a system of four equations (10) and (II)
In four unknowns (x, y, z and A.). To solve this system, multiply the first of
equations (11) by x, the second by y, the third by z, and add; taking (10)
into account we find that /., = -
3
~~z. Putting this value of /., into equations ,
(II) we get
yz [1-~ (y+z)J =0,
xz [1-~(x+z)J =0,
xy [1-~(x+u)J =0.
Since it is evident from the statement of the problem that x, y, z are diffe-
rent from zero, we get from the latter equations
3x 3y 3z
2li (y + z) = I, a (x + z) = 1, 2a (x + y) = 1.
2
From the first two equations we find x=y, from the second and third equations,
y=z. But then from equation (10) we get x=y=z= ]-.This is the y
only system of values of x, y, and z, for which there can be a maximum or
minimum.
1t can be proved that the solution obtained yields a maximum. Incidentally,
this is also evident from geometrical reasoning (the stat~ment of the problem
indicates that the volume of the box cannot be big without bound; it is
therefore natural to expect that for some definite values of the sides the
volume will be a maximum).
Thus, for the volume of the box to be a maximum, the box must be a
cube, an edge of which is equal to y: .
Example 2. Determine the maximum value of the nth root of a product
of numbers x 1 , x,, ... , Xn provided that their sum is equal to a given num
ber a. Thus, the problem is stated as follows: it i~ required to find the max-
imum of the function u = JY
x 1 Xn on the condition that
x 1 +x,+ ... +xn-a=O
(12)
(x 1 > 0,
X 2 > 0, ... , Xn > 0).
Form an auxiliary function:

F (x 1, ... , Xn I.)= JY X 1
Find its partial derivatives:

p' _I
x, -n x2x 3
n-1 -n X
Xn +I.- 1 u +1.-0
1
- or
_
U-
t.,
-n x.,
(x.. Xn) n
' 1 u
F =--+1. =0
x. 11 x, or u=-nlvx 2 ,
Singular Points of a Curve 305
From the foregoing equations we find
and from equation (12) we have

a
XI = Xz = = Xn = n
By the meaning of the problem these values yield a maximum of the
function iYx 1 Xn equal to ;i-.
Thus, for any positive numbers x1 , x,, ... , Xn connected by the relation
ship x1 +x2 + ... +xn=a, the inequality
(13)
is fulfilled (since it has already been proved that ~ is the maximum of this
n
function). Now substituting into (13) the value of a obtained from (12), we get
(14)
This inequality holds for all positive numbers xl' x 2 , , Xn The expression
on the left-hand side of (14) is called the geometric mean of these numbers.
Thus, the geometric mean of several positive numbers is not greater than
their arithmetic mean.
SEC. 19. SINGULAR POINTS OF A CURVE
The concept of a partial derivative is used in investigating

curves.
Let a curve be given by the equation
F (x, y) =0.
The slope of the tangent to the curve is determined from the
formula
aF
(see Sec. 11, Ch. VIII).

lf at a given point M (x, y) of the curve under consideration,
at least one of the partial derivatives ~~ and ~~ does not vanish,
then at this point either ~~ or ~: is completely determined. The

curve F(x, y)=O has a very definite line tangent at this point.
In this case, the point M (x, y) is called an ordinary point.
But if at some point M 0 (x 0 , y0 ) we have

( aaF)
X x=x
= 0 and ( aF)
ay x=x,
=O,
0
y=y,
then the slope of the tangent becomes indeterminate.

Definition. If at the point M. (x 0 , y 0 ) of the curve F (x, y) = 0,
both partial derivatives~~ and ~~vanish, then such a point is called
a singular point of the curve. Thus, a singular point of a curve
is defined by the system of equations
ap ap
F=O' -=0
ax ' -=0
ay .
Naturally, not every curve has singular points. For example,
for the ellipse
obviously,
x2 y2 aF 2x aF 2y
F(x, Y)=az+fiZ-l; ax=i2; ay=fi2
the derivatives ~~ and ~: vanish on\~ when x = 0, y = 0, but th~se
values of x and y do not satisfy the equation of the ellipse.
Consequently, the ellipse does not have any singular points.
Without undertaking a detailed investigation of the behaviour
of a curve near a singular point, Jet us examine some examples
of curves that have singular points.
Example I, Investigate the singular points of the curve
y 2 -x(x-a) 2 =0 (a>O).
Solution. Here, F (x, g)= y 2 -x (x-a) 2 and therefore
aF aF
ax =(x-a)(a-3x); ay=2y.
Solving thethree equations simultaneously,
aF aF
F (x, y) =0, ax =0, ay =0,
we ftnd the only system of values of x and y that satisfy them:
Xo=a, Yo=D.
Consequently, the point M 0 (a, 0) is a singular point of the curve.
Let us investigate the behaviour of the curve near a singular point and
then const rud the curve.
Rewrite the equation in the form

u= (x-a) Vx.
From this formula it follows that the curve: 1) is defined only for x ?-0;
2) is symmetrical about the x-axis; 3) cuts the x-axis at the points (0, 0) and
(a, 0). The latter point is singular, as we have pointed out.
Let us first examine that part of the curve which corresponds to the plus
sign:
Find the first and second derivatives of y with respect to x:

, 3x-a 3x+a
Y = 2 Vx ' Y = 4x Vx
For x=O we have y=oo. Thus, the curve touches they-axis at the origin.
a a
For x=a we have y'=O, u">O, which means that for x=y the func-
tion y has a minimum:
u=-~ vf-
On the interval 0 < x <a we have y < 0; for x > 1- y' > 0; as x- oo y- oo.
For x =a we have y' = Va, which means that at the singular point M 0 (a, 0)
the branch of the curve y= +(x-a) Yx has a tangent
u= Ya (x-a).
Since the second branch of the curve y=- (x-a) Yx is symmetrical

with the first about the x-axis, the curve has also a second tangent (to the
second branch) at the singular point
y=- Ya (x-a). !J
The curve passes through the singular point
twice. Such a point is called a nodal point.
The foregoing curve is shown in Fig. 187.
Example 2. Test for singular points the curve
(semicubical parabola)
yt-xa=o.
Solution. The coordinates of the singular
points are determined from the following set of
equations:
y 2 -x 1 =0; 3x 2 =0; 2y=0.
Consequently, M 0 (0, 0) is a singular point.
Let us rewrite the given equation as
y= Vx.
To construct the curve let us first Investigate
the branch to which the plus sign in the equation . Fig. 187.
corresponds, since the branch of the curve corresponding to the minus sign is
symmetric with the first about the x-axis.
The function y is defined only for x ;-;. 0, it is nonnegative and increases
as x increases.
Let us find the first and second derivatives of the function y= Yx 1 :
y'=; Yx;
For x=O we have y=O, y' =0. And so the given branch of the curve has a
tangent y = 0 at the origin. The second branch of the curve y = - YXJ also
passes through the origin and has the same tangent y = 0.
Y Thus, two different branches of the curve meet at the ori-
gin, have the same tangent, and are situated on different
sides of the tangent. This kind of singular point is called
a cusp of the first kind (Fig. 188).
Note. The curve y 2 -x1 =0 may be regarded as a limit-
ing case of the curve y 2 =x (x-a) 2 =0 (considered in Exam-
ple 1) as a->- 0; that is, when the loop of the curve is
contracted into a point.
x Example 3. Investigate the curve
(y-xZ)Z-xi=O.
Solution. The coordinates of the singular points are de-
fined by the following set of equations:
-4x(y-x 2)-5x 4 =0; 2(y-x')=0,
Fig. 186. which has only one solution: x=O, y=O. Hence, the
oTJgln is a singular point.
Rewrite the given equation in the form
Y=x 2 Yx'.
From this equation it follows that x can take on values from 0 to + oo.
Let us determine the first and second derivatives:
y' = 2x } }"7; y = 2 ~ VX.

Investigate, separately, the branches of the curve corresponding to plus and
minus. In both cases, when x=O we have y=O, y'=O, which means that for
both branches the x-axis is a tangent.
Let us first consider the branch
y=x 1 + Yx'.
As x increases from 0 to oo, y increases (rom 0 to oo,
The second branch
y=xz- yXi
cuts the x-axis at the points (0, 0) and (1, 0).
For x = ~ the function y = ~~- YXi has a maximum. If x -+ +oo, then
IJ-+-00,
Thus, in this case the two branches of the curve meet at the ongtn; both
branches have the same tangent and are situated on the same side of the
tangent near the point of tangency. This kind of singular point is _called a
cusp of the second kind. The graph of this
function is shown in Fig. 189. Y
Example 4. Investigate the curve
y 2 -x'+x'= 0.
Solution. The origin is a singular point.
To investigate the curve near this point re-
write the equation of the curve in the form
y= X 2 v1 x2
Since the equation of the curve contains
only even powers of the variables, the curve
is symmetric about the coordinate axes and,
consequentlr, it is sufficient to investigate
that part o the curve which corresponds to Fig. 189.
the positive values of x and y. From the
latter equation it follows that x can vary over the interval from 0 to I, that
is, O..;;;;xo;;;;l.
Let us evaluate the first derivative for that branch of the curve which is
a graph of the function y= +x' V1-x2 :
y' x (2-3x 2)
V1-x 2
For x = 0 we haye 1J = 0, 'y' = 0. Thus, the curve touches the x-axis af the
origin.
For x=1 we have y=O, y'=(Y); consequently, at the point (1, 0) the
tangent is parallel to the y-axis. For x= Yi the function has a maximum
(Fig. 190).
At the origin (at the singular point) the two branches of the curve corre-
sponding to plus and minus in front_ of the radical sign are mutually tangent.
A singular point of this kind is
Y called a point of osculation (also
known as tacnode or_ double cusp).
yt-x"'+xs=o Example 5. Investigate the curve
y 2 -x 2 (x-1)=0.
Solution. Let us wr_ite the sys-
tem of equations defining the sin-
gular points: -
y 2 -x 2 (x-1)=0;
Fig. 190. -3x 2 +2x=0, 2y=0.
This system has the solution
x = 0, y = 0. Therefore, the- point
(0, 0) Is a singular point of the curve. Let us rewrite the given equation in
the form
u= x Vx--=-1.
310 Functions of SetJerar Variables
It is obvious that x can vary from I to oo and also take the value 0 (in
which case y=O). .
Let us investigate the branch of the curve corresponding to the plus sign
in front of the radical. As x increases from I to oo, y increases from 0 to oo.
The derivative
!J
, 3x-2
y = .
2 Yx-1
When x = 1 we have y' = oo; hence, at the

point (I, 0) the tangent is parallel to the y-axis.
The second branch of the curve corresponding
to the minus sign is symmetric with the first about
the x-axis.
The point (0, 0) has coordinates that satisfy the
0 X equation and, consequently, belongs to the curve,
but near it there are no other points of the curve
(Fig. 191). This kind of singular point is called an
isolated. singular point.
Exercises on Chapter VIII
Fig. 191; Find the partial derivatives of the fqllowing

functions:
az az
ay =X2 sin 2y. ax=
=Y2XY
l 1
-;
az
ay=X
yo
-2yJnx. 3. u=eX+Y+Z,
I I I
~ = 2yeX'+Y"+Z"; ~- 2zeX"+Y 1 +ZI,

~ ~- '
, au X az 1/ az
A ns. ax= :-Jfr-x=::=+===:Y":=+=z=;: 5. z =arc tan (xy). Ans. ax= 1 +.xy ;_ ay = +Xx"y .
1
6. z=arc tan 1L az -y . az X 1-"Xz'+Y-x

X Ans. ax=x+y"' ay=x+y"' 1. z=ln Yx'+y'+x.
X f X
Ans. - = -
az 2 az 2x
8. u=e
11
+e
11
Ans.
au 1 y'
-=-
ax . Vx+u"
.
ay= y Y x'+y'

ax y e
au X .:._ Z _:_ au 1 ..!. az
-=--eY--ell.
dy
-=-ell
y y2 ' az y
z= arc sin (x+ y).
9. Ans. ax=
I az z=.arc tan,/ x u".
az y
10. Ans. ax= y~;
= Jfl-(x+ y) 2 JY V x'+y" x x-y
P? _-y
:;-t=l'
oy vx-y .
Find the total diFferentials of the-following functions: 11. z=x'+xy'+slny.
Ans. dz=(2x+y")dx+(2xy+cosy)dy. 12. z=ln(xy). Ans. dz=dx+~.
X y
Exercises on Chapter V Ill 3\l
13. z=ex"+Y". Ans. d2=2ex"+Y"(xdx+ydy). 14. u=tan(3x-y)+6Y+'".

Ans. du ~:x ) + (- cos (;x-y )+6Y+Z In 6) dy+6Y+Z !n 6 d2. 15. w =
cos 2 x-y
x y dx-xdy
=arc sin-. Ans. dw
Y I y I Vu-x.
16. Evaluate f~(2, 3) and {~(2, 3) if f(x, y)=x'+y'. Ans. f~(2, 3)=4,
~~ (2, 3)=27.
17. Evaluate df(x,y) forx=1,y=O;dx={,dy=~ iff(x,y)=Vx+y.
1
Ans.
2 .
18. Form a formula which, for small absolute values of the quantities x,
y and z, yields an approximate expression for y(l +lui(~ +z). Ans. 1 +

+2I (x-y-2).
, / 1 +x 1
19. Do the same for V 1
+u+ . Ans. 1 +
2 2 (x-y-z).
. i'Jz
20. F !nd l.fX and
02
ay 'tf Z=u+o; o=ln (x+y).
& 1 & . 1
Ans. a-=2x+2o-+
X X y ; vy;;-=Cosy+2o-+ X y .
02 a2. , /I+ u i'Jz
21. Find i'Jx and oy tf z = V o=cosx. Ans. ;rx=
1 + 0 ; u = -cos x;
1 02
=---X ; Oy =0.
2 cos
2
22. Find ~ and~ if z=eu-u, u=sin x, o=.r'+!l Ans. !;=
02
=eU-11/D (COS X-6X 2 ), i)y ==ieU-11/D (0-2 2y) = -'-4yea-HJ.
23.Find the total derivatives of the given functions: z =arc sin (u + o);
oz . n n
U=Sinxcosa; ti=COSXSina. ~ns. i'Jx=1 If 2kn-2<x+a<2kn+2 1
oz . n n e"x (y-z)
;rx=-ltf2kn+ 2 <x+a<(2k+1)n+ 2 . 24. u=
a+
1
; y=asinx;
ou , r::;-:-;; i'Jz
Z=COSX. Ans. ax=eax sin x. 25. Z=ln (1-x'); X= f sm6; aa= -2 tan 6.
Find the derivatives of implicit functions of x given by the following equa-
x y dy b2 x x y 2 dy
tions: 26. a2 +-b.-1=0. Ans. d-=- x 2a - -.-b 2 =1. Ans. -d
y . 27. a~ x
=
2
b X x dy yxY-l-yX iny
= % - 28. y =xY.Ans. -d = x Yl . 29. sin(xy)-exY-xy=O.
a y x xy - 1 '-X n x
dy y [cos (xy)-exY-2x] x 2 y 2 z2 oz o?
Ans. -dx = x [ x +xy
e -cos (xy )]' 30. .a2 +b.+ 2c =1; fmd ;;-and;;-
vX vy
oz c2x. oz c2y . . ow ow iJw
Ans. ;;- = - 2 , ;;- =- b2 . 31. u-v tan aw=O, fmd ;;-and;;-. Ans. ;;- =
vX a Z vy Z vU vV vU
= cos 2 aw ; ow=_ sin 2aw . 32 22 + ~ = y yz_ 22 show that xz!!! + _!_oz =..!. .
av OV 2av X ' OX y oy Z
33. !__ = F show that x ~z + y ~z =z, no matter what the differentiable
(.!!...);
X Z vX vy
function F.
Compute the second-order partial derivatives:
ozz ozz
34. Z=X 3 -4x 2y+5y 2 . Ans. oxz=6x-By; iJyox -Bx; =
azz x sin y
35. z=eXiny+sinyinx. Ans. sz=e lny --2-'
vX. X
azz ex .
oyz=- y' -smy In x.
o o
.
36. Prove that tf U=v xz +I yz + zZ 'thena-z+;;z+-a
x
iJZu
vy z
2
u
z=O.
2
u
. xzyz aaz ozz oz

37. Prove that tf Z=-+
X y ' then
' n+Y -Xa
X
vX a y =2a-.
X
. f I ( z z) t h ozz ozz 0
38. P rove th a t I Z= n X + Y , en axz + oy =
azz ozz
39. Prove that if z=<p(y+ax)+'IJ(y-ax), then a 2 y 1 - x =0 for any
0 0 1
doubly differentiable <p and .
40. Find the derivative of the function z=3x'-xy+y 1 at the point M~ 2)
in the direction thatmakes an angle of 60with !hex-axis. Ans. 5+~.
41. Find the derivative of the function z=5x 1 -3x-y-l at the point
M (2, I) in the direction from this poi~! to the point N (5, 5). Ans. ~ =9.4.
42. Find the derivative of the function f (x, y) in the direction of: I) the
bisector of the quadrantal angle Oxy. Ans. ~ (:~ + :~); 2) the negative
. . ox.
semt-axts of
Ans. -ox
43. f(x, y)=x 3 +3x 2 +4xy+y 2 Show that at the point M ( ~ , -~)the
derivative in any direction is equal to zero (the "function is stationary").
44. or
all triangles with the same. perimeter 2p, determine the triangle
with greatest area. Ans. Equilateral triangle.
45. Find a rectangular parallelepiped of greatest volume for a given total
surface S. Ans. A cube with edge V~ .
Exercises on Chapter VIII 313
46. Find the distance between two straight lines in space whose equations
x-1 y z x y z V2
are -1-=2=T' y=y=T Ans. T
Test for maximum and minimum the functions:
47. z=x 3 y 2 (a-x-y). Ans. Maximum z at x=]-;y=;.
48. z=x+xy+y++++. Ans. Minimum z at x=y= V3 .

49. z=sinx+siny+sin(x+y) (O,;;;;;x,;;;;; ~; O,;;;;;y,;;;;; ~). Ans. Maximum
1t
Z at X=Y=3
50. z=slnxsinysln(x+y)(O,;;;;;x,;;;;;1t; O,;;;;;y,;;;;;1t). Ans. Maximum z at
1t
x=y =s.
Find the singular points of the following curves, investigate their character
and form equations of the tangents at these points:
51. x'+y'-3axy=O. Ans .. M 0 (0,0) is a node; x=O, y=O are the
equations of the tangents.
52. a'y 1 =x' (a 2 -x 2). Ans. A double cusp at the origin; the double tangent
y=o.
a
53. y 2 = ~. Ans. M 0 (0, 0) is a cusp of the first kind; y 2 =0 is a
2 a-x
tangent.
54. y 2 =x 2 (9-x 2 ). Ans. M 0 (0, 0) is a node; y= 3x are the equations of
the tangents.
55. x'-2ax 2y-axy 2 +a 2.~ 2 =;=0. Ans. M 0 (0, 0) is a cusp of the second kind;
y 2 = 0 is a double tangent.
56. y 2 (a 2 +x 2)=x 2 (a2 -x2). Ans. M 0 (0, 0) is. a node; Y=x are the
equations of the tangents.
57. b 2x 2 +a2y 2 =x 2y 1 Ans. M 0 (0, 0) is an isolated point,
58. Show that the curve y=x In x has an end point at the coordinate
origin and a tangent which is the y-axis.
59. Show that the curve y = _x_l has a nodal point at the origin and
1 +ex
that the tangents at this point are: on the right y=O, on the left y=x.
CHAPTER IX
APPLICATIONS OF DIFFERENTIAL CALCULUS TO SOLID

GEOMETRY
SEC. 1. THE EQUATIONS OF A CURVE IN SPACE
Let us consider the vector OA = r whose origin is coincident

with the coordinate origin and whose terminus is a certain point
A (x, y, z) (Fig. 192). A vector of this kind is called a radius
vector.
Let us express this vector in terms of the project ions on the
coordinate axes:
r=xi+yj+zk. (1)
Let the projections of the vector r be
functions of some parameter t:
x=<p(t),}
y y = 'ljJ (t), . (2)
X
z=x(t).
Then formula (1) may be rewritten as follows:
Fig. 192.
r = <p (t) i +'$ (t)j+ x (t) k (1')
cr, in abbreviated form,
r=r(t).
As t varies, x, y, and z vary; and the point A (the term in us of
the vector r) will trace out a l'ine in space that is called the
hodograph of the vector r=r(t). Equation (1') or (l") is called
the vector equation of the I ine in space. Equations (2) are known
as the parametric equations of the line in space. With the aid of
these equations, the coordinates x, y, z of the correspond'ing point
of the curve are determined for each value of t.
Note. A curve in space can also be defined as the locus of
points of the intersection of two surfaces. It can therefore be
given by two equations of two surfaces:
1> 1 (x, y, z) = 0, } (3}
1> 2 (x, y, z) = 0.
Thus, for example, the equations
x2+y2+z2=4, Z= 1
The Equations of a Curve in Space 315
are the equations of a circle obtained at the intersection of a

sphere and a plane (Fig. 193).
Thus, a curve in space may be represented either by paramet-
ric equations (2) or by two equations of surfaces (3).
If we eliminate the parameter t from equations (2) and get two
equations connecting x, y, z; we will thus make the transition
from the parametric method of representing a line to the surface
z
z-v4-xLy2
X
Fig. 193.
method. And conversely, if we put x=cp(t), where cp(t) is an ar-

bitrary function, and find y and z as functions oft from equations
<D. [cp (t), y, z] = 0, (!>, [cp (t), y, z] = 0,
we will then make the transition from representation of a line by
means of surfaces to its parametric representation.
Example I. The equations
x=4t-!,. y=3t, z=t+2
are parametric equations of a straight line. Eliminating the parameter t we
get two equations, each of which is an equation of a plane. For instance, if
from the first equation we subtract, termwise, the second and third, we get
x-:-y-z=-3. But subtractii:Jg (from the first) four times the second we get,
x-4z= -9. Thus. the given straight line is the line of intersection of the
planes x- y-z+3=0 and x-4z +9=0.
Example 2. Let us consider ;J right circular cylinder of radius a, whose
axis coincides with the z-axis (Fig. 194). Onto this cylinder we wind a right
triangle C 1 AC so that the vertex 'A of the triangle lies at the point of inter-
section of the generator of the cylinder with the x-axis, while the leg AC 1 is
wound onto the circular section of the cylinder lying in the xy-plane. Then
the hypotenuse -will generate on the cylinder a line that is called a helix.
316 Applications of Differential Calculus to Solid Geometry
Let us write the equation of the helix, denoting by x, y, and z the coor-
dinates of its variable point M and by t the angle AOP (see Fig. 194). Then
x=acost, y=asint, Z=PM=AP tan e,
where e denotes the acute angle of the triangle C,AC. Noting that iP =at,
since AP is an arc of the circle of radius a corresponding to the central angle
e
t, and designating tan in terms of m, we get the parametric equations of
the helix in the form
x=a cost, y=a sin t, z=amt
X C,
Fig. 194.
(here t is the parameter), or in the vector form:

r=ia cost+ ja sin t +kamt.
lt is not difficult to e!iminatr the parameter t from the parametric equa-
tions of the helix: square the first two equations and add. We find x+y 2 =a.
This is the equation of the C)llinder on which the helix lies. Then, dividing
termwise the second equation by the first and substituting into the obtained
equation the value of t found from the third equation, we find the equation
of another surface en which the helix lies:
.!L=tan.!.....
x am
Tllis is the so-called helicoid. It is generated as the trace of a half-line
parallel to the xy-plane if the end point of this half-line lies on the z-axis and
if the half-line itself rotates about the z-axis at a constant angular velocity,
and rises with constant velocity so that its extremity is translated along the
z-axis. The helix is the line of intersection of these two surfaces, and so can
be represented by two equations:
x 2 +y'=a2 , .!L=tan.!.....
x am
The Limit and Derivative of the Vector Function 317
SEC. 2. THE LIMIT AND DERIVATIVE OF THE VECTOR

FUNCTION OF A SCALAR ARGUMENT. THE EQUATION
OF A TANGENT TO A CURVE. THE EQUATION OF A NORMAL
PLANE
Reverting to the formulas (1') and (1") of the preceding section,
we have
r='P (t) i + 'iJ (t)j+ x (t)k
or
r= r (t).
When t varies, the vector r varies in the general case both in
magnitude and direction. We say that r is a vector function of
the scalar argument t. Let us suppose that
lim 'P (t) = <p,,
1-+1,
lim \jJ (t) = \jl 0 ,
1-+1 0
lim X (t) = Xo
1-+t,
Then we say that the vector r, ='Poi+ 'i'oi+

+ Xok is the limit of the vector r = r (t) and
we write (Fig. 195) Fig. 195.
lim r (t) = r 0
1-+1,
From the latter equation follow the obvious equations

lim Ir (t)-r0 I= lim V [IJl (t)-<p 0 ] 2 +['I' (t)- \jl 0 ] 2 +[X (t)- X0 ] 2 = 0
1-+1 0 1-+1 0
and
lim lr(t)l=lrol
1-+lo
Let us now take up the question of the derivative of the

vector function of a scalar argument,
r (t) = IJl (t) i + \jJ (t) j+ X (t) k, (I)
assuming that the origin of the vector r (t) lies at the coordinate
origin. We know that the latter equation is the equation of some
space curve.
Let us take some fixed value t corresponding to a definite point
M on the curve, and let us change t by the increment M; we
then get the vector
r (t + M) = 1P (t + M) i + 'i' (t + M)j+ x (t + M) k,
318 Applications of Differential Calculus to Solid Oeometr.u
which defines a certain point M, on the curve (Fig. 196). Let us

find the increment of the vector
D.r=r(t+ M)-r(t) =
= [cp (t + M)-cp (t)] i +
+['I> (t + M)- 'iJ (t)]j+
+ (x (t + M)-x (t)] k. j
r(t)
In Fig. 196, where OM =r(t), OM,=
=r (t + M), this increment is shown by
the vector MM, = D.r(t).
Let us consider the ratio t..:~t) of the
increment of a vector function to the
Flg. 196. increment oi a scalar argument; this is
obviously a vector collinear with the
vector D.r (t), since it is obtained from the latter by multiplication
with the scalar factor ~~. We can write this vector as follows:
M {I)_ <p U+ M)-<p (t)
M- M t
+ 1jJ (I+ M)-'i'
M
(I) +X (I+ M)~x (t) k
J. M'.
If the functions cp (t), 'iJ (t), x(t) have derivatives for the chosen
value of t, the factors of i, j, k will in the limit become the de-
rivatives tp' (t), 'I>' (t), x' (t) as M -0. Therefore, in this case the
limit of!~ asM-Oexistsand isequaltothevectorcp' (t)i+'i''(t)j+
+X' (t)k:
lim !~ = cp' (t) i +'I>' (t) j + x' (t) k.
lil-+0
The vector defined by the latter equation is called the deriva-

tive of the vector r(t) with respect to the scalar argument t. The
derivative is denoted by the symbol :~ or r'.
Thus,
:; =r' =cp' (t) i +'I'' (t)j+x' (t)k (2)
or
dr dx dy dz
dt =& 1+&J+([fk. (2')
Let us determine the direction of the vector '::;

Since as M - 0 the point M, approaches M, the direction of
the secant MM, yields, in the limit, the direction of the tangent.
The Limit and Derivative of the Vector Functior& 319
Hence, the vector of the derivative ~~ lies along the tangent to

dr
the curve at M. The length of the vector dt is defined by the
formula*)
~~~/=V[ql'(t)]'+l'll' U)]'+[x' U>J'. (3)
From the results obtained it is easy to write the equation ofihe

tangent to the curve
r=xi+ yj+ zk
at the point M (x, y, z), bearing in mind that in the equation of
the curve x = <:p (t), y -:-- 'iJ (t), z =X(/).
The equation of the straight line passing through the point
M (x, y, z) is of the form
X-x Y-y Z-z
--;;- = -n-= -P-'
where X, Y, Z are the coordinates of the variable point of the

straight line, while m, n, and p are quantities proportional to the
direction cosines of this straight line (that is to say, to the pro-
jections of the directional vector of the straight line).
On the other hand, we have established that the vector
dr =~i-1-dy .+dzk
dl dl dt J dl
is directed along the tangent. For this reason, the projections of
. this vector are numbers that are proportional to the direction co-
sines of the tangent, hence also to the numbers m, n, p. Thus,
the equation of the tangent will be of the form
(4)
Example 1. Write the equation of a tangent to the helix

x =a cos t, y = a sin t, z =ami
lor an arbitrary value of t and. lor t ~1- .

Solution.
dx dy dz
dT =-a sin t . dT =a cost, ;:n.= am.
*) '!J~. shall assume ttiat: at the points under consideration I:~ l

;i: O. '
From formula (4) we have

X -acos t Y -a sin t Z-amt
- a sin t a cos t
=--
am
In particular, for t =~ we get
X _a 2 y-a2
2 2 .l-am_::
4
--=-
a2- aY2 am
--2- --2-
Just as in the case of a plane ~urve, a straight line perpendi-

cular to a tangent and passmg through the point of
tangency is called a normal to the space curve at the given point.
Obviously, one can draw an infinitude of normals to a given space
curve at a given point. They all lie in the plane perpendicular
to the tangent line. This plane is the normal plane.
From the condition of perpendicularity of a normal plane to a
tangent (4}, we get the equation of the normal plane:
dx
df(X-x) + dy dz
dt (Y-y) + dt(Z-z)=O. (5)
Example 2. Write the equation of a normal plane to a helix at a point
for which t= T.
Solution. From Example 1 and formula (5) we get
_ ~2( X- a ~2) + ~2 (y _a ~2) +m( l-am~ )=o.

Let us now derive the equation of a tangent line and the nor-
mal plane of a space curve for the case when this curve is given
by the equations
<1>, (x, y, z) = 0, <D, (x, y, z) = 0. (6)
Let us express the coordinates x, y, z of this curve as functions
of some parameter t:
x=cp(t), y='ljJ(t}, z=x(t). (7}
We shall assume that cp (t), 'lj; (t}, x (t) are differentiable functions
of t.
Substituting into equations (6), in place of x, y, z, their
values for the points of the curve expressed in terms of t, we get
two identities in t:
<1>, [cp (t}, 'ljJ (t}, X (t)] = 0, (Ba)
<D 2 [cp (t), '1jJ (t), X (t)] = 0. (8b)
Differentiating the identities (8a) and (8b) with respect to t, we
The Limit and Derivative of the Vector Function 321
get
a<D, d:r: + a<D, +
dy a<D, dz =O }
ax dt ay dt (}z dt
a<Da dx + a<D. dy + a<D. dz = 0. (9)
ax dt ay dt az dt
From these equations it follows that

dx a<D, a<D. a<D, a<D. dy
dt Ty -az-az---ay dt
dz = a<D, a<D, a<D, a<D. ; dz = a<D, a<Da a<D, a<D
(10)
Tt -ax-ay-ay-ax: dT -ax--ay-a;;ax
. a<D a<D a<D a<D
Here, we naturally assume that the expression ax' ay" -a;J ax"=/=
=/= 0; however, it may be proved that the final formulas (II) and
(I2) (see below) hold also for the case when this expression is
equal to zero, provided that at least one of the determinants in
the final formulas differs from zero.
From equations (10) we have
dx dy dz
([[ ([[ CIT
a<D, a<D, a<D, a<D. = a<D, a<D. a<D, a<D, = a<D, a<D, a<D, a<D,
ay-&-az-a;; -azax-ax:Tz axa;;--ayax:
Consequently, from formula (4) the equation of the tangent line
will-have the form
X-x Y-y Z-z
a<D, il<Da a<D, a<D. a<D, d<D, a<D, a<D.
a<D, arD. a<D, a <D.
a;;az--azau az-ax-Txaz axay--ay-ax:
or, using determinants,
X-x Y-y Z-z
a<D, a<D, arD, a<D, iJ<D, a<D, (II)
CIYTz Tzax ax au
a<D. a<D. a<D.a<D. il<Da o<D,
ay Tz azax ax au
The normal plane is represented by the equation
(12)
a<D. a<D.
ax au
These formulas are meaningful only when at least one of the
determinants mvolved is different from zero. But if at some point
11-3388
3~2 A pp/ications of Differential Calculus to Solid Geometru
ol the curve all three determinants

o, o, a, o, o, o(Jl,
7fY az azax ax au
a. a. a. o, o(Jl, a 2
auaz azax axay
vanish, this point is called a singular point of the space curve.
At this point the curve may not have a tangent at all, as was the
case with singular points in plane curves (see Sec. 19, Ch. VIII).
Example 3. Find the equations of a tangent line and a normal plane to
the line of intersection of the sphere x 2 +y 2 +z 2 =4r 2 and the cylinder
z X 2 Y"=2ry at the point M (r, r, r
Jf 2 ) (Fig. 197).
Solution.
, (x, y, z)=x"+u"+z"-4r",
,(x, y, z)=x 2 +Y"-2ry,
o, =2x, a, o, =2z

ay-=2y, az
; dx
a. = 2x a.=2y-2r
oy
a-o
oz-.
OX '
The values of the derivatives at the given

point M will be
a,=2r
ax
a 1 =2r
ay
a~z'=2r'r'2:'
v r ..:.
Fig. 197. ~~ 2 =2r, a.

{) y -
-o . a,_o
az- .
For this reason the equation of the tangent line has the form
X-r Y-r Z-r Jl2
-o-= Jf2= -I
The equation of the normal plane is
Jf2 (Y -r)-(Z-r Jf2)=0.
SEC. 3. RULES FOR DIFFERENTIATING VECTORS

(VECTOR FUNCTIONS)
-
As we have seen, the derivative of a vector
r (t) = q> (t) i + \jJ (t) j+ x(t) k, (1)
is, by definition, equal to
r' (t) = q/ (t) i + 'i'' (t)J+ x' (t) k. (2)
Rules for DifJerentialing Vectors (Vector Functions) 323
Whence it straightway follows that the basic rules for differen-

tiating functions hold for vectors as well. Here, we shall derive
the formulas for differentiating a sum and a scalar product of
vectors; the other formulas we shall write down and leave their
derivation for the student.
I. The derivative of a sum of vectors is equal to the sum of the
derivatives of the vectors.
Indeed, let there be two vectors:
r 1 (t) = <p 1 (t) i + \jl 1 (t) j + x. (t) k, } ( )
3
r. (t) = <p 2 (t) i + '~' (t) }+ x. (t) k;
their sum is
r, (t) + r. (t) = [<p 1 (t) + <p. (t) 1i + l'll. (t) + '1'. (t) ]} + lx. (t) + x. (t) 1k.
By the definition of a derivative of a variable vector, we have
d lr. (t~ir (t)] = [<p 1 (t) + <r. (t) ]' l + [\jl 1 (t) + '~' (t)]'j + lx. (t) + x.(t) J' k
or
d lr. u~;r.(t)J = [<p; (t) +<r;(t)1l+ ['IJ; (t) +'IJ;{t)l}+ lx:(t) +x: (t)lk =
=<p; (t) l + \jl; (t)}+ x; (t) k+ <p: (t) l + \jl~ (t)}+ x~ (t) k=r; +r~.
Hence,
d [r1 (t) +r 2 (t)] = dr1 + dr 2
(I)
dt dt dt
I I. The derivative of a scalar product of vectors is expressed by
the formula
(II)
Indeed, if r 1 (f), r. (t) are defined by formulas (3), then, as we

know, the scalar product of these vectors is equal to
r 1 (!) r, (t) = <r. <r. + '1' 1 '1'. + x. x.
For this reason
d (r1r 2 ) I ~ I I I I
d / = <p.<p. + <r.<p. + 'I' 'I' +'I'. 'I' + X X.+ X. X =

I I I I 1 1
+ 'll'llz + X X.)+ (<p.<pz +'I'. 'llz + XzXz) =

= (<p.<p.
= (<r;i+'I';J+x:k)(<p.i +'I'.J+x.k) +<<r.t +'ll.J+x.k) (<p~i +'I'J+x;k) =
dr1
=drr
+ rdf
dr
1

From formula (II) we have the following important corollary.
Corollary. If the vector e is a unit vector, that is, I e I= 1, then
its derivative is a vector perpendicular to it.
11" I.
Proof. It e is a unit vector, then

ee= 1.
Let us take the derivative, with respect to t, of both sides of
the latter equation:
or
de
2edl =0,
that is, the scalar product
de
ecn =0,
and this means that the vector~ is perpendicular to the vector e.

II I. The constant numerical factor may be taken outside the
sign of the derivative:
dar (t) _ dr (t) _ , (!)
dt -a dt -ar (II I)
IV. The derivative of a vector product of vectors r 1 and r. is

determined by the formula
d (r1 xr,] = dr1 X _L X dr.
dt dt r ' r1 dt
(IV)
SEC. 4. THE FIRST AND SECOND DERIVATIVES

OF A VECTOR WITH RESPECT TO THE ARC LENGTH.
THE CURVATURE OF A CURVE. THE PRINCIPAL NORMAL
-
The arc length*) of a space curve M 0 A = s (Fig. 198) is deter-
mined just as in the case of curves in a plane. When a variable
point A (x, y, z) moves along a curve, the arc length s varies;
conversely, when s varies, the coordinates x, y, z of a variable
point A lying on the curve also vary. Therefore, the coordinates
x, y, z of a variable point A of the curve may be regarded as
functions of the arc length s:
x = <p (s),
y = 'll (s),
z = x (s).
*) The arc length of a space curve is defined in exactly the same way as
\he arc length of a plane curve (see Sec. I, Ch. VI and Sec. 3, Ch. XII).
First and Second Derivatives of Vector with Respect to Arc Length 325
In. these parametric equations of the curve, the arc length s is

the parameter. The vector OA=r is, accordingly, expressed as
r = <p (s) l + 1jJ (s) j+ 'X (s) k
or
r= r (s). (1)
Thus the vector r is a function of the arc length s.
Fig. 198. Fig. 199.
Let us find out the geometrical meaning of the derivative ~.

As is evident from Fig. 198, we have the following equations:
M 0 A=s, AB=~s, Mfi=s+~s.

OA=r(s), OB=r(s+~s),
AB = ~r= r (s + ~s)-r(s),
M AB
6,s = AB .
We have already seen in Sec. 2 that the vector ~r = lim t:.r is

5 t.s ..... o tJ.s
in the direction of the tangent to the curve at the point A towards
increasing s. On the other hand, we have the equality lim/ ~ I= 1
[the limit of the ratio of the chord length to the arc length*)].
*) In Sec. I. Ch. VI, we mentioned this relation for a plane curve. lti
also holds lor a space curve: r(t)=qJ(t)i+'(t)j+X(t)k if the functions
1p (t), 'ljl (t) and X (t) have continuous derivatives that do not vanish simulta
neously.
326 Applications of Differential Calculus' to Solid Geometry
Hence, s
is a unit vector in the direction of the tangent; let us
denote it by 0':
dr
as=a. (2)
If the vector r is represented by the projections
r=xi+yJ+zk,
then
O'=~i+~
ds dsJ
'+~k
ds '
(3)
and
-v (~r + (~r + (~r=

Let us now examine the second derivative of the vector func-
1.
tion ~7., that is, the derivative with respect to s, and deter-
mine its geometric significance.
From formula (2) it follows that
d
2
r d
ds 2 = ds
[d']
ds =
da
ds'
Consequently, we have to find lim t:.a.

li.S-+ 0 /J.S
From Fig. 199 we have AB=~s. AL=a, BK=O'+~O'. Draw

lrom the point B the vector BL 1 = 0'. From the triangle BKL 1
we find
or
O'+~O'=O'+L 1 K.
Thus, L1 K = ~0'. Since, by what has been proved, the length of

the vector a does not change, jO' j =I a+ ~0' I; hence, the triangle
BKL 1 is an isosceles triangle. .
The angle L\<p at the vertex of the triangle is the angle through
which the tangent to the curve turns from the point A to the
point B; in other words, it .corresponds to the increment in the
arc length ~s. From the triangle BKL 1 we find
L1 K = I ~a I= 2! a 11 sin !J.2
2_ l- sin__1\q>2_1 Ml
1\0,_
11\S -
2 __
I 1\S - 1\q> 1\S '
2
Let us now pass to the limit on both sides of the latter equation
as ds--.. 0. On the left side we have
1,:\ol/dol '
I.I l l - - -
6.s .... 0 1\S - ds
Then
sin 1\q>
2
lim
1\q>
= 1,
2
since in this case we consider curves such that there exists a limit
li~ 1q> and, consequently, d .... 0 s
to the limit we have
Idol= lim JM\

ds 6.s .... 0 1\S
(4)
The ratio of the angle of turn d<p of the tangent, when the point
A goes to the point B, to the length Lls of the arc AB (in abso-
lute value) is called (just as it is in the case of a plane curve)
the average curvature of the given line on the segment AB:
average curvature= I~; I
The limit of the average curvature as lls --+0 is called the curva-
ture of the line at the point A and is denoted by K:
K= 6.s-+
lim o IMI1\S
do
But then from (4) it follows that ds = K; which means that the
length of the derivative of a unit vector*) of a tangent with res-
pect to the arc length is equal to the curvature of the line at the
given point. Since the vector o is a unit vector, its derivative
dd~ is perpendicular to it (see Sec. 3, Ch. IX, Corollary).
*) It should be remembered that the derivative of a vector is a vector

and for this reason we can speak of the length of the derivative.
Thus, the vector dd~ is equal, in length, to the curvature of the

curve, and, in direction, is perpendicular to the vector of the
tangent.
Definition. The straight line that has the same direction as the
vector dd: and passes through the corresponding point of the curve
is called the principal normal of the curve at the given point.
We denote by n the unit vector of this direction.
Since the length of the vector dd: is equal to K, which is the
curvature of the curve, we have
du
ds=Kn.
The reciprocal of the curvature is called the radius of curva-
ture of the Iine at the given point and is denoted by R, ~ = R.
So we can write
(5)
From this formula it follows that
1 (d'dsr)z
R2 = 2
(6)
But
Hence,
_!_ _ -. /
R - V
(ddsx)' + (dds'y)' + (d'z)'
2
2
2
ds' (6')
This formula enables us to compute the curvature of a line at

any point provided that this line is represented by parametric
equations in which the parameter is the arc length s (in other
words, if the radius vector of the variable point of the given line
is expressed as a function of the arc length).
Let us consider the case when the radius vector r is expressed
as a function of an arbitrary parameter t:
r =r(t).
In this case the arc length s will be regarded as a function of
the parameter t. Then the curvature is computed as follows:
dr drds
dt =ds di. (7)
First and Second Derivatives of Vector with Respect to Arc Length 329
Since
I ~,.=
ds
1 *)
'
we have
( ~)
dt = (~)
dt (8)
Differentiating the right and left sides of (8) and reducing by

two, we get
drd 2r ds d 2s
dt dt 2 = dt dt. (9)
Further, from formula (7) it follows that

dr dr I
as=cndi-
dt
Differentiate, with respect to s, both sides of this equation:

d's
d 2r d 2r I dr dt 2
ds= dt' (~y-en(~)'
Substituting into formula (6) the expression obtained for :7z we get
(dtd'r)' (~)~- 2 d rdr~ ds (~) (d s)=

2 2 2 2
2
dt dt dt dt dt + dt dt2
- (~)'
) This equation follows from the fact that J ddr

s
I= lim
tls-+o
I I
M
~s
But M
M
is a chord subtending an arc of length ~s. Therefore - approaches 1 as
~s
~s-+ 0,
330 Applications of Differential Calculus to ~olid Geometry
. ds d 2s
Expressmg dt and dt' by formulas (8) and (9) in terms of the
derivatives of r (f), we get*)
(d'r)' (~) (d'r

2 2
!!:.!._ \
_
1 dt 2 d/
dt 2 dt )
(10)
R = {(~rr
Formula (10) may be rewritten as follows: **)
(II)
We have obtained a formula that enables us to calculate the

curvature of a given line at any point for an arbitrary paramet-
nc representation of this curve.
If in a particular case the curve is a plane curve and lies in
the xy-plane, then its parametric equations have the form
X=<p(t),
Y= 'I' (t),
z=O.
Putting these expressions of x, y, z into formula (II), we get the
earlier derived (in Ch. VI) formula that yields the curvature of a
plane curve represented parametrically:
K- I cp' (I)""" (t)-11'' (I) cp" (t) I
- {lcp' (t)]" + w (/))"} .,, .
Example. Compute the curvature of the helix
r= i a cos t +j a sin t + k amt
at an arbitrary point.
*) We transform the denominator as follows: (dtds)' = {(ds)'

dt }s =
= {( dtdr)'}' . Here (dr)
di . By (dr)'
we cannot write dt we mean the scalar
dr by {(dr)'}'
square of the vector dt; iii , the third power of (dr)'
ill . The ex-
. (dr).
pressiOn di 1s meanmg . 1ess.
*"') We uhlised the identity a 2b 2 -(ab) 2 =(axb? whose validity is readily
recognisable if one rewrites the identity as foUows: a 2 b 2 -(ab cos <p) 2 = (ab sin <p) 1
Osculating Plane. Binormal. Torsion 331
Solution.
~~ = - i a sin t +1 a cos I+ k am,
d 2r
dt=- i a cos t-1a sin I,
i 1 k
dr d 2r
ill x dt= -a sin t a cost am = ia 2 m sin t -1 a 2m cost+ ka 2,
-a cost -a sint 0
Consequently,
1 a 4 (m"+l)
R= [a (I+ m2 ) ] 1 = a (1 +m 2) ' '
whence
+
R=a (1 m2 ) =const.
Thus, the helix has a ronstant radius of curvature.
Note. If a curve lies in a plane, then without violating genera-

lity, we can assume that it lies in the xy-plane (this can always
be achieved by transforming the coordinates). Now if the curve
lies in the xy-plane, then z = 0; but then ~:~ = 0 also and, conse-
quently, the vector n likewise lies in the xy-plane. We thus con-
clude that if a. curve lies in a plane then its principal normal
lies in the same plane.
SEC. 5. OSCULATING PLANE. BINORMAL. TORSION
Definition 1. The plane passing through the tangent line and the
principal normal to a given curve at the point A is called an
osculating plane at the point A.
For the plane of a curve, the osculating plane coincides with
the plane of the curve. But if the curve is not a plane curve,
and if we take two points on it, P and PI' we get two different
osculating planes that form a dihedral angle f.t. The bigger the
angle f.t, the more the curve differs in shape from a plane curve.
To make this more precise, let us introduce another definition.
Definition 2. The normal (to a curve) perpendicular to an oscu-
lating plane is called a binormal:
On the binormal let us take. a unit vector b and make its
direction such that the vectors a, n, b form a triple with the
332 Applications ot Differential Calculus to Solid Geometry
same orientation as the unit vecturs i, j, k lying on the coordi-

nate axes (Figs. 200, 201).
_b
Fig. 200. Fig. 201.
By virtue of the definition of a vector and scalar product of

vectors we have
b =ox n; bb = 1. (1)
We find the derivative of ~~ . By formula (IV), Sec. 3,

db = d (aXn) = da X n + fJ X dn (2)
ds ds ds ds
da
But ds = Rn (see Sec. 4), therefore
d(J I
ds xn=R nxn=O,
and formula (2) takes the form
db _ 0 xdn (3)
ds - ds
From this it follows (by the definition of a vector product)
that :: is a vector perpendicular to the vector of the tangent CJ.
On the other hand, since b is a unit vector, :: is perpendicular
to b (see Sec. 3, Corollary).
This means that the vector :: is perpendicular both to C1 and
to b; that is, it is coli inear with the vector n.
db f
Let us denote the length of the vector ds by T ; we put
1:: I= i;
then
db I
ds =yn (4)
The quantity ~ is the torsion of the given curve.

The dihedral angle 1-1. between the osculating planes that corre-
spond to two points of the curve is equal to the angle between
the binormals. By analogy with formula (4), Sec. 4, Ch. IX, one
can write
l ~j=
ds
lim _j.l_.
6.s-+o ills I
To summarise, then, the torsion of a curve at a point_ A is
equal, in absolute value, to the limit which is approached
(as dS-+0), by the ratio of the angle 1-1. between the osculating
planes at the point A and the neighbouring point B to the length
I dsj of the arc AB.
If the curve is plane then the osculating plane does not change
its direction and, consequently, the torsion is equal to zero.
From the definition of torsion it is clear that it is a measure
of the deviation of a space curve from a plane curve.
The quantity T is called the radius of torsion of the curve.
Let us find a formula for computing torsion. From (3) and (4)
it follows that
Multiplying scalarly both sides by n, we get
On the right side of this equation we have the so-called mixed

(or triple) product of three vectors n, a and ~; . In a product
of this kind the factors, as we know, may be circularly permuted.
In addition, taking into consideration that nn = 1, we rewrite the
latter equation in the following form:
or
..!....=-a[nx.!!:!!..l
T ds (5)
d 2r
But since n = R CfS2, we have
dn d 3r dR d 2r
ds = R ds 3 +ds ds 2
334 Applications of Diffuential Calculus to Solid Geometry
and
dn
[ n x Ts =
J d 2r J d'r dR d r }
+ Ys
R "'ifSi" x l R ""liSi
CiS" =
2
[d'r d'r] dR[d"r d"r]

= R ds X dSS + R dS _ds X
2 2- ds 2
But since the vector product of a vector into itself is equal to

zero,
Thus,
dn]
[_n X ds = R [d~
ds'
d~]
X ds'
Noting that o = s and revertlng to (5)~ we get

I 2 drld"r d'r]
T = - R ds ds 2 X ds 1 {6)
If the factor r is expressed as a function of an arbitrary param-

eter t, it may be shown,*) much like was done in the preceding
*) Indeed,
dr drds
dt=dSdt'
Differentiating this equality once again with respect to t, we get
d2 r d (dr) ds ds dr d 2s d 2r ( ds)' dr d 2s
dt =ds ds dtdt+dsdt= ds 2 dt, +dS dt
Differentiate it once more with respect to t:
2 2 2
d'r d ( d r) ds ( ds)" d r 2 d s d s d (dr) ds d 2s dr d's
dt'=ds ds dt dt +ds" dtdt 2 +ds ds dtdt 2 +dsdt 1 =
d'r (ds)' d 2r ds d 2s drd's
= dSS dt +3 ds 2 dt dt 2 + ds dt'
Let us now form a triple product:
1
dr ( d"r d r)
dt dt 2 X (jj8 =
2
drds { [ d r (ds) dr
=CIS 'iii ds 2 'iii +ds d/ 2
ds] X [ddsr(dsds )' + 3 ddsr iflds d's
2 1
1 1
dr d'r J}
dt +ds dt
2
2 2 1
Opening the brackets of this product by the rule of multiplying polyno-

mials, and disregarding those terms that contain even two identical vector
factors (since the tri pie product of three factors where at least twl) are equal
is zero), we get
~( d r2 d r )=~( d"r2 d r) (~)'
2 1 1
dt d/ X dt 1 ds ds X ds 1 dt '
section, that
dr[ d'r
3
dr J
2
dr[ d r X d r
1
J= di ([l!x(W
{ (~r}'
3
ds ds ds
Putting this expression into formula (6) and replacing R' by

its expression from formula (11 ), Sec. 4, we finally get
1
dr [ d'r
dt ([[2xdf3
3
dr J
-r=- '[~x:;~r . (7)
This formula makes it possible to compute the torsion of the

curve at any point if the curve is represented by parametric
equations with an arbitrary parameter t. .
Concluding this section, we note that the formulas which
express the derivatives of the vectors a, b, n are called Serret
Frenet formulas:
da n db n dn a b
ds =R' Ts=r ds = - 7 [ - r
The last one of them is obtained as follows:

n=bxa,
dn = d (bxa) =db xa +bx da = ..!.xa+ bX.!!_=
~ ~ ~ ~ T R
I I
=-rnxa +Rbxn;
but
nxa=-b; bxn=-a,
therefore
Finally, noting that
( ~)
dt
= (t!!.)2
dl '
or
we obtain the required equality.

336 Applications of Difjerential Calculus to Solid Geometry
Example. Compute the torsion of the helix

r=ia cost +ja sin t+kamt.
Solution.
dr [ d'r
dt dt'
dar
X ([[I
J= ~- a sin t a c~s t am
-a c?s t -a sm t 0 =a'm,
I
a sm t - a cos t 0
dr dzr
[ dt X dt 2
J = a (1 + m
2 4 2
) (see Example, Sec. 4).
Consequently,
SEC. 6. A TANGENT PLANE AND NORMAL TO A SURFACE
Let there be a .surface given by an equation of the form

F(x, y, z)=O. (1)
We introduce the following definition.
Definition 1. A straight line is a tangent to a surface at some
point P (x, y, z) if it is a tangent to some curve lying on the
surface and passing through P.
Since an infinitude of different curves lying on the surface pass
through the point P, then, generally speaking, there will also be
an infinitude of tangents to the surface passing through this
point.
We introduce the concept of singular and ordinary points of
a surface F (x, y, z) = 0.
If at the point M (x, y, z) all three derivatives ::, :: , ~~are
equal to zero or at least one of these derivatives does not exist,
,,
then M is called a singular point of the surface. If at M (x, y, z)
all three derivatives ~= ~: ~~ exist and are continuous, and at
least one of them differs from zero, then M is an ordinary point
of the surface.
We can now formulate the following theorem.
Theorem. All tangent lines to a given surface (1) at an ordinary
point of it P lie in one plane.
Proof. Let us consider, on a surface, a certain line L, (Fig. 202)
passing through a given point P of the surface. Let this curve be
represented by parametric equations:
X= (jJ (t); y = 'iJ (t); z =X (t). (2)
A Tangent Plane and Normal to a Surface 337
A tangent to the curve will be a tangent to the surface. The

equations of this tangent have the form
X-x Y-y Z-z
dX -----;ry = """"ifZ
dt dt dt
lf we put expressions (2) inlo equation (1), the latter will be-
come an identity in t, since the curve (2) lies on the surface (1).
Differentiating it with respect to t, we get*)
aF dx
ax dt
+ aF dy + aF dz = 0
ay dt az dt .
(3)
dr
Let us further examine the vectors N and dt
that pass through P:
N -ap.+ ap '+ aF k (4)

-ax' ay J az .
The projections of this vector ~=, ~=, az Fig. 202.

depend on x, y, z, which are the coordinates
of P; it will be noted that since P is an ordinary point, these
projections at the point P do not simultaneously vanish and
therefore
The vector
~= :; i :r
+ J+ :: k. (5)
is tangent. to the curve passing through the point P and lying on

the surface. The projections of this vector are computed from
equations (2) with the value of the parameter t corresponding to
the point P. Let us compute the scalar product of the vectors N
and ~~, which product is equal to the sum of -the products of
Iike projections:
~ Here we apply the rule for differentiating a composite function of three

variables. This rule Is applicable here since all the partial. derivatives ~:.r.
::, ~~ are, as stated, continuous.
338 Applicatinns of Dif!erential Calwlus to Solid Geometry
On the basis of (3), the expression on the right is equal to

zero; hence
From the latter equality it follows that the vector N and the
tangent vector ~; to the curve (2) at the point P are perpendicu-
lar. The foregoing reasoning holds for any curve (2) passing
through the point P and lying on the surface. Therefore, every
tangent to the surface at the point P is perpendicular to one and
the same vector N and for this reason
all these tangents lie in a single plane
that is perpendicular to the vector
N. The theorem is prov-ed.
Definition 2. The plane in which
lie all the tangent lines to the lines
on the surface passing through the
given point P is called the tangent
plane to the surface at the point P
(Fig. 203).
Fig. 203. It should be noted that there
may not exist a tangent plane at
the singular points of the surface. At such points, the tangent
lines to the surface may not lie in one plane. For instance, the
vertex of a conical surface is a singular point. The tangents to
the conical surface at this point do not lie in one plane (they
themselves form a conical surface).
Let us write the equation of a tangent plane to a surface (I)
at an ordinary point. Since this plane is perpendicular to the
vector (4), its equation has the form
aF ap aF
-(X-x)+-
ax az (Z-z)=O.
ay (Y-y)+- (6)
If the equation of a surface is given in the form
z=f(x, y), or z-f(x, y)=O,
then
ap at - aF at ap
ax = - ax ' ay = - ay ' az = 1'
and the equation of the tangent plane is then of the form
Z-z =~(X -x) + ~~ (Y -y). (6')

A Tangent Plane and Normal to a Surface 339
Note. If in formula (6') we put X -x = dx; Y -y = dy, then

this formula will take the form
iJf iJf
Z-z =ax dx--j- iJy dy;
its right side is the total differential of the function z = f (x, y).
Therefore, Z- z = dz. Thus, the total differential of a function of
two variables at the point M (x, y), which corresponds to the
increments dx and dy of the independent variables x and y, is
equal to the corresponding increment on the z-axis of the tangent
plane to the surface which is a graph of the given function.
Definition 3. The straight line drawn through the point P ( x, y, z)
of surface (1) perpendicular to the tangent plane is called the nor-
mal to the surface (Fig. 203).
Let us write the equations of the normal. Since its direction
coincides with that of the vector N, its equations will have the
form
X-x Y-y Z-z
7fF = --ar = aJi" . (7)
iJx iJy iJz
If the equation of the surface is given in the form z = f (x, y), or

z-f(x, y)=O,
then the equations of the normal have the form
X-x Y-y Z-z
ar = --a;=-~-
~ iJx - iJy
Note. Let the surface F (x, y, z) = 0 be the level surface for

some function of three variables u = u (x, y, z); that is,
F(x, y, z)=u(x, y, z)-C=O.
Obviously, the vector N defined by formula (4) and in the
direction of the normal to the level surface F = u (x, y, z)-C= 0,
will be
that is,
N=grad u.
We have thus proved that the gradient of the function u (x, y, z)
ts tn the direction of the normal to the level surface pas:;ing through
the given point.
Example. Write the equation of the tangent plane and the equations of
!he norn;tal to the surface of the sphere x + y + z = 14 at the point P (1, 2, 3).
Solution.
aF aF aF
F(x, y, Z)=x+y 2 +z"-14=0; ax =2x; ay =2y; az =2z;
for x=l, y=2, z=3 we have
axF = 2; aF = 4 aF = 6.
a ay ' az
Therefore, the equation of the tangent plane will be
2 (x-1) +4 (y-2) +6 (z-3)=0 or x+2y +3z-14=0.
The equations of the normal are
x-1 y-2 z-3
-2-= -4---6-
or
x-! y-2 z-3
-,-=-2-=-3-.
Exercises on Chapter IX
Find the derivatives of the vectors: 1. r=lcott+jarctan t.
Ans. r' = - si~ 1 l + 1 ~ 1.J. 2. r=le- 1+ j2t +kIn I. Ans. r' =-le-t+
.
+~+y3.r=
2 k tJ j k
-y+p A 2l j
ns. r' =I+Ji-iT"
2k
4. Find the vector of a tangent, the equations of the tangent and the
equations of the normal plane to the curve r= tl + tj+ t 1k at the point
x-3 y-9 z-27
(3, 9, 27). Ans. r'=l+6J+27k; tangent: --=--=~; normal
1 6
plane: x+6y+27z=786.
5. Find the vector of a tangent, the equations of the tangent and the
equation of the normal plane to the curve r= l cos ~ +} j sin t +k: sin ~ .
Ans. r' = - ; l sin t +~}cos t +; k cos {r; the equation of the tangent
1
X -cos Y- I II Z It
2 2 sn -sn
2
-sin cos 1 I the equation of the normal plane:
1 1
cos 2
t t
+X sin t-Y cost -Z cos = - x sin t + y cost +z cos
2 2 where x, y, .z are the
coordinates of that point of the curve at which the normal plane is drawn
(that is, x=cos 2 ~, y= ~ sin t, z.= sin~). .
6. Find the equations of the tangenttothe curve x=t-slnt, y=l-cost,
z = 1 sin ~ and the cosines of the angles that it makes with the coordinate
Exercises orl Chapter I X 341
X-X Y-Y Z-Z I

axes. Ans. ---=--
0
1-=--1- .
0 0
cosa=sin 2 ~ ; cos~ =
2 slnt 0 ;
sin .!..rL cos ...!... cot ....!!....
2 2 2
to
cosy=cos 2.
7. Find the equation of the normal plane to the curve z=x 2 -y 2 , y=x
at the origin. Hint. Write the equations of the curve in parametric form.
Ans. x+y=O.
1[
8. Find a, n, b at the point t = T for the curve r= i (cost+ iln 2 I)+
+}sin t (I -cos t)-k cost. Ans. U=-~-(-i+J+k); n= -5i-4j-k :-

b i-2J+3k
3 Y42
YI4
9. Find the equations of the principal normal and the binormal to the
f4 t3 t2 x-x
curve x=-;r; Y=3 i Z=2 at the point (x0 , y 0 , z0). Ans. - - -0=
1:+21 0
= y-y 0 z-z 0 x-x 0 y-y 0 z-z 0
1- t 4 -2t 3 -to' -~-=24=-12 -.
0 0 0
10. Find the equation of the osculating plane to the curve y 2 =x; x 2 =z
at the point M (1, 1, 1). Ans. 6x-8y-z+3=0.
II. Find the radius of curvature for a curve represented by the equations
x2 +y 2 +z 2 -4=0, x+y-z=O. Ans. R=2.
et-e-t
12. Find the radius of torsion of the curve: r= i cost+ j sin t + k - -- .
2
(el-e-t)a
Ans. T
2(e 1-e- 1)
13. Find the radius of curvature and the torsion for the curve r= t 2i + 2t 3j.
2 .,
Ans. R=:rf(l+9t 2) , T=oo.
14. Prove that the curve r=(a 1 t 2 +b 1 t+c1 )i+(a2 t 1 +b 2 t+c2 )j+
+ (a1 t 1 + b3 t +c1 ) k is plane. Ans. r'" sa 0; therefore the torsion is equal to zero.
15. Find the curvature and torsion of the curve x=e1, u=e- 1, z=t J(2.
A ns. Th e curva t ure IS . Y 2 th e t ors10n
(x + y)z;
. . -
IS (x _ y).
Y2
16. Find the curvature and torsion of the curve x=e- 1 sin t, g=e- 1 cost<
z=e 1 Ans. The curvature is
2 1
~
e ; the torsion is ; e1
. ~ ~
17. Find the equation of the tangent plane to the hyperboloid
7-v-
z2 x.x YtY z.z
- =I at the po1nt (x1 , y 1 , z1 ). Ans. - 1 - -b
- 2 1 - - . = 1.
c . a c
18. Find the equation of the normal to the surface x 2 -4y 2 +2z 2 =6 at
the point (2, 2, 3). Ans. y+4x=l0; 3x-z=3.
19. Find the equation of the tangent plane to the surface z=2x 2 +4y 2 3t
the point M (2, 1, 12). Ans. Bx+Bu-z= 12.
20. Draw to the surface x2 +2u"+z 1 = I a tangent plane parallel to the
plane x-y+2z=0. Ans. x-y+2z= (V.
CHAPTER X
INDEFINITE INTEGRALS
SEC. t. ANTIDERIVATIVE AND THE INDEFINITE INTEGRAL
In Chapter III we considered a problem like the following:

Given a function F (x), find its derivative, that is, the function
f (x) = F' (x).
In this chapter we shall consider the reverse problem: Given
the function f (x), it is required to find a function F (x) such that
its derivative is equal to f (x), that is,
F' (x) = f (x).
Definition 1. The function F (x) is called the antiderivative of
the function f (x) on the interval [a, b] if at all points of this
interval the equality F' (x) = f (x) is fulfilled.
Example. Find the antiderivative of the function f (x) = x 2
From the definition of an antiderivative it follows that the function
F (x) =~ is an antiderivative, since ( ~ )' =x 2

It is easy to see that if for the given function f (x) there exists
an antideri vative, then this anti derivative is not the only one.
In the foregoing example, we could take the following functions
as antiderivatives: F(x)=~+l; F(x)=~-7 or, generally,
F(x)=x; +C (where C is an arbitrary constant), since
(x; +C)' = x.
On the other hand, it may be proved that functions of the form
~ + C exhaust all anti derivatives of the function x. This follows
from the following theorem.
Theorem. If F, (x) and F 2 (x) are two antiderivatives of the
function f (x) on the interval [a, b ], then the difference between
them is a constant.
Proof. By virtue of the definition of an antiderivative we have
F; (x) = f (x), } (1)
F, (x) = f (x)
for any value of x on the interval [a, b].
Antiderivative and the Indefinite Integral 343
Let us put
(2)
Then by (1) we have
F; (x)- F~ (x) = f (x)- f (x) = 0
or
<p' (x) = [F 1 (x)- F. (x) ]' =0
for any value of x on the interval [a, b]. But from cp' (x) = 0 it
follows that <p (x) is a constant.
Indeed, let us apply the Lagrange theorem (see Sec. 2, Ch. IV)
to the function <p (x), which, obviously, is continuous and differen-
tiable on the interval la, bl.
No matter what the point x on the interval [a, b]; we have,
by virtue of the Lagrange theorem,
<p (x)-<p (a)= (x-a) <p' (6),
where a<6<x.
Since <p' (s) = 0,
<p (x)-<p (a)= 0
or
w(x) = <p (a). (3)
Thus, the function <p (x) at any point x of the interval [a, b]
retains the value cp (a), and this means that the function 'P (x) is
constant on [a, b]. Denoting the constant <p(a) by C, we get,
from (2) and (3),
F 1 (x)-F. (x) =C.
From the proved theorem it follows that if for a given function
f (x)some one antiderivative F (x) is found, then any other anti-
derivative of f (x) has the form F (x) + C, where C = const.
Definition 2. If the function F (x) is an anti derivative of f (x),
then the expression F (x) + C is the indefinite integral of the
function f (x) and is denoted by the symbol ~ f (X) dx. Thus, by
definition
~ f (x) dx = F (x) + C,
if
F' (x) = f (x).
Here, the function f (x) is called the integrand, f (x) dx is the element
of integration (the expression under the integral sign), and ~ is
the integral sign. .
Thus, an indefinite integral is a family of functions Y.= F (x) +C.
344 Indefinite I nlegra/s
From the geometrical point of view, an indefinite integral is an

assemblage (family) of curves, each of which is obtained by trans-
lating one of the curves parallel to itself upwards or downwards
(that is, along the y-axis).
A natural question arises: do antiderivatives (and, hence, an
indefinite integral) exist for every function f (x)? The answer is no.
Let us note, however, without proof, that if a function f(x) is
continuous on the interval (a, b], then there is an antiderivative
of this function (and, hence, there is also an indefinite integral).
This chapter is devoted to working out methods by means of
which we can find antiderivati\"es (and indefinite integrals) of
certain classes of elementary functions.
The finding of an antiderivative of a given function f(x) is
called integration of the function f (x).
Note the following: it the derivative of an elementary function
is always an elementary function, then the antiderivative of the
elementary function may not prove to be representable by a finite
number of elementary functions. We shall return to this question
at the end of the chapter.
From Definition 2 it follows that:
1. The derivative of an indefinite integral is equal to the in-
tegrand, that is, if F' (x) = f (x), then also
(~ f(x)dx)' =(F(x)+C)'=f (x). (4)
This equation should be understood in the sense that the deriva-
tive of any antiderivative is equal to the integrand.
2. The differential of an indefinite integral is equal to the
expression under the integral sign:
d0f(x)dx)=f(x)dx. (5)
This results from formula (4).
3. The indefinite integral of the differential of some function is
equal to this function plus an arbitrary constant:
~ dF(x)= F(x)+C.
The truth of this equation may easily be checked by differentia-
tion [the differentials of both sides are equal to dF (x)].
SEC. 2. TABLE Of INTEGRALS

Before starting on methods of integration, we give the following
table of integrals of the simplest functions.
The table of integrals follows directly from Definition 2, Sec. 1.
Ch. X, and from the table of derivatives (Sec. 15, Ch. III).
Table of Integrals 345
(The truth of the equations can easily be checked by differentia-

tion: to establish that the derivative of 1.he right side Is equal to
1.he integrand).
x~+l
1. xadx=a+l +C(a*=-1). (Here and in the formulas that
S
follow, C stands for an arbitrary constant.)
2. sd:=lnlxi+C.
3. ~ sin x dx = -cos x +C.
4. ~cosxdx=sinx-f-C.
5. S d~
COS X
=tanx+C.
6. S ~~
Sin X
= -cotx+C.
7. ~ tan x dx = -In I cos xI+ C.
8. ~ cotxdx=lnl sinxj+C.
9. ~ e"' dx=e"'+C.
10. Sa"' dx = 1 ~xa +C.
11. S ~x=arctanx+C.
1
11 ,. sa
z---+ dx I
.x =-a arc tan-+
2 a C.
X
12 s~=_!_lnla+xi+C.
a-x 2a a-x
13 Svdx =arcsinx+C.
1-x
13' Sy- dx
a-x =arc sin!....a +C.
14. S y dx =In I x+ Vx"aJ +C.
x 2 a2
Note. The table of derivatives (Sec. 15, Ch. III) does not have
formulas corresponding to formulas 7, 8, 11', 12, 13' and 14.
However, differentiation will readily prove the truth of these as
well.
In the case of formula 7 we have
(-In I cosx I)'=- -slnx =tanx,
COS X
consequently, ~tan x dx =-In I cosx J +C.

346 Indefinite Integrals
In the case of formula 8

1 COSt
(In I sm xI ) =-.-=cot
Stn X
x,
consequently,~ cotx=lnl sinx/+C.
In the case of formula I2,
(~lnj~:;j)' =~[inla+xl-lnla-xl)'=
=~ [a~x+a~x] =a
21
x 1 '
therefore,
5 a~=_.!_
2
-x 2a
lnja+xi+C.
a-x
2
It should be noted that the latter formula will also follow from
the general results of Sec. 9, Ch. X.
In the case of formula 14,
(Inlx+Vx"a /)'=
1
x+ ; x a2 2
(I+ yx xa )=vx a
1 2
1
2 2
,
hence,
This formula likewise will follow from the general results of Sec. II.
Formulas ll' and 13' may be verified in similar fashion. These
formulas will later be derived from formulas II and I3 (see Sec.4,
Examples 3 and 4).
SEC. 3. SOME PROPERTIES OF AN INDEFINITE INTEGRAL
Theorem 1. The indefinite integral of an algebraic sum of two

or several functions is equal to the sum of their integrals
~ [f, (x) + fz (x)] dx = ~ f, (x) dx + ~ f1 (x)dx. (l)
For proof, let us find the derivatives of the left and right sides
of this equation. On the basis of (4) of the preceding section we
have
+
(~ [f, (x) fz (x)] dx)' = f, (x) fz (x), +
0 f, (x) dx+ ~ f (x) dx)' =
1
= (~ f, (x) dx)' + 0 { (x) dx)' =f, (x)+f

1 1 (X).
Some Properties of an Indefinite Integral 347
Thus, the derivatives of the left and right sides of (1) are equal;
in other words, the derivative of any antiderivative on the left-
hand side is equal to the derivative of any function on the right-
hand side of the equation. Therefore, by the theorem of Sec. 1,
Ch. X, any function on the left of (1) differs from any function
on the right of (1) by a constant term .. That is how we should
understand (1).
Theorem 2. The constant factor may be taken outside the integral
sign; that is, if a=const, then
~ af (x) dx =a ~ f (x) dx. (2)

To prove (2), let us find the derivatives of the left and right
sides:
n af {X) dx )' = af (X),
(a~ f (x) dx) =a ( ~ f (x) dx) = af (x).
1 1
The derivatives of the right and left sides are equal, therefore,
as in (1 ), the difference of any two functions on the left and
right is a constant. That is:how we should understand equation (2).
When evaluating indefinite integrals it is useful to bear in mind
the following rules.
I. If
~ f (x) dx = F (x) + C,
then
Sf (ax) dx=~ F (ax) +C. (3)
Indeed, differentiating the left and right sides of (3), we get

( ~ f (ax) dx )' = f (ax),
1(ax) a= F 1(ax)= f (ax).
( . a.!. F (ax)) I=...!..a (F(axj )~ =...!..a F
The derivatives of the right and left sides are equal, which is
what we set out to prove.
II. If
~ f (x) dx = F (x) + C,
then
(4)
Ill. If
~ f (x) dx=F (x) + ~
then
St(ax+b)dx=! F(ax+b)+C. (5)
Equations (4) and (5) are proved by differentiation of the right
and left sides.
Example I.
~ (2x'-3slnx+5Jfx)dx= ~ 2x'dx-~ 3slnxdx+) 5 Vxdx=
= 2 ~ x dx- 3 ~ ~x
2
3
sin x dx + 5 dx =
1
-+1
2
x'+ 1 X I . 10
=2
31 -3 (-cosx)+S-1- - +C= x+3cosx+ x
2 3 Yx+C.
+ 2+1
Example 2.
Example 3.
dx
Sx+ 3 =In I x+31 +C.
Example 4.
Scos 7x dx=; sin 7x+C.
Example 5.
Ssin (2x-6) dx= -~cos (2x- 6>+C.
SEC. 4. INTEGRATION BY SUBSTITUTION (CHANGE
OF VARIABLE)
Let it be required to find the integral

~ f(x) dx;
we cannot directly select the anti derivative of f (x) but we know
that it exists.
Integration by Substitution 349
Let us change the variable in the expression under the integral

sign, putting
X= <p (f), (1)
where <p (t) is a continuous function with continuous derivative
having an inverse function. Then dx = cp' (t) dt; we shall prove
that in this case we have the following equation:
~ f (x) dx = ~ f [cp (t) 1cp' (t) dt. (2)
Here we assume that after integration we substitute, on the right
side, the expression of t in terms of x on the basis of (1).
To establish that the expressions to the right and left are the
same in the sense indicated above, it is necessary to prove
that their derivatives with respect to x are equal. Find the deri-
vative of the left side:
( ~ f (x) dx )~ = f (x).
We differentiate the right side of (2) with respect to x as a com-
posite function, where t is the intermediate argument. The de-
pendence of t on x is expressed by (I); here, ~~ = cp' (t) and by
the rule of differentiating an inverse function,
dt 1
dx = lp 1 (t)
We thus have
(5 r [(j) (!) 1 (j)' u> dt ): = (S r r dt ); ~ =

= f [cp (t)1 cp' (t) IJl' ~t) = f [<p (t) 1=I (x).
Therefore, the derivatives, with respect to x, of the right and
left side of (2) are equal, as required.
The function x = qJ (t) should be chosen so that one can evaluate
the indefinite integral on the right side of (2).
Note. When integrating, it is sometimes better to choose a
change of the variable in the form of t = 'IJ' (x) and not x = <p (t).
By way of illustration, let it be required to calculate an integral
of the form
1jl' (x) dx
S 1jJ (x)
Here it is convenient to put

'ljl (x) = t
then
'ljl' (x) dx= dt,
'ljl' (x) dx
S 'ljl (x) =
sdt
T =In It I+ C =In I 'ljJ (x) I+ C.
The following are a number of instances of integration by
substitution.
Exa~ple 1.. ~ Jfsinx cosxdx=? Wr; mal<e the substitution t=slnx; then
v~inxcosxdx= s
2
dl=COS xdx and, consequently, s Ytdt= s t dt=
2tf. 2
=T+C= 3 sin ''x+C.
x dx s xdx
Example 2. l+x =?We putt=l+x 2; thendt=2xdx and l+x=
S
= 2I sdt I
t+C= 21 In(I+x')+C.
T='iln
Example 3. 5a/+x x 2 =~5

a I+
t) :_
a
2 We put l=~a; then dx=adt,
dx 1 s a dt 1 s dt 1 1 x
S a'+x=(l! I+t=a I+l 2 =a arc tant+C=aarc tana+C.
Example 4. S dx
,r 2 2 = -
y a -x
1
a
dx
. ' 1- ( -x ) 2
sV x
We put 1=-; then
a
.
a
dx=adt, r dx =-
1 s
.radt =
s .r-=arcsm
dt . t+C=
S 2
y a - x2 a y 1-1 2 y 2 l-1
=arc sin!_+ C (if is assumed that a>O).
a
The formulas 11' and 13' given in the Table of Integrals (see above, Sec. 2)
are derived in Examples 3 and 4.
ExampleS. s(lnx>'=? Put t=lnx; then dt=~x. S(lnx>'7=

t 1
=
St'dt= 4 +C= 4 (1nx)+C.
x dx =?Putl=x;thendl=2xdx, sl+x
x dx 2
= 21 s ~=
dt
ExampleS.
I
Sl+x
I
=:= arc tan t +C= arc tan x 2 +C.
2 2
The method of substitution is one of the basic methods for calculat-
ing indefinite integrals. Even when we integrate by some other
Integrals of Functions Containing a Quadratic Trinomial 351
method, we often resort to substitution in the intermediate stages

of calculation. The success of integration depends largely on how
appropriate the substitution is for simplifying the given integral.
Essentially, the study of methods of integration reduces to fi,nding
out what kind of substitution has to be performed for a given ele-
ment of integration. Most of this chapter is devoted to this problem.
SEC. 5. INTEGRALS OF FUNCTIONS CONTAINING

A QUADRATIC TRINOMIAL
I. Let us consider the integral

I-S ax +bx+c
1 - 2
dx
Let us first transform the trinomial in the denominator by rep-

resenting it in the form of a sum or difference of squares:
ax 2 +bx+c=a [x 2 +~x+ ~] =
=a [x 2 +2~x+ (~Y +%-(~YJ =
where
(
=a[(x+ 2~r+ ~ - 4b;~)J =a[(x+ ~Yk 1] ,,
c b2 2
-
a - -4a-2 -
+-k
The plus or minus is taken depending on whether the expression

on the left is positive or negative, that is, on whether the roots
of the trinomial ax 2 +bx+c are complex or real.
Thus, the integral I 1 will take the form
I.= Iaxz:~x+c= ~I [(x+ ~rkz]

In the latter integral let us change the variable:
x + 2tib = t, dx = dt.
We then get
I.=aI s rlt
t2k2'
These are tabular integrals (see formulas 11' and 12).
Example 1. Calculate the integral
S2x +:+20
2
Solution.
1
= 5 dx
2x"+Bx+20 =2
I s dx
x'+4x+ 10 =
=2
Is x2 +4x+4+ I0-4
~
2
Is ~
(x+2)'+6
Let us make the substitution x+2=t, dx=dt. Putting It into the integral
we get the tabular integral
1= 2
I st+
dt
6= 2
I I
Vtfarctan
t
yij+C.
Substituting in place of t its expression in terms of x, we finally get
I x+2
I= .r-are tan .r- +C.
2 f 6 f 6
II. Let us consider an integral of a more general form:

Ax+B
1 =
Sax+ bx+c dx.
Perform an identical transformation of the integrand:
Represent the latter integral in the form of a sum of two inte-

grals. Taking the constant factors outside the integral sign, we
get
I2 _
-
A
2a
sax+ +bx+c dx + (B -A sax +bx +c'
2ax b b)
2a 2
dx
The latter integral is- the integral It> which we are able to
evaluate. In the first _integral make a substitution:
ax"+bx+c=t, (2ax+b)dx=dt.
Thus,
S~~~~~!)~:= S =In It I +C= In I ax"+ bx+cJC.
And we finally get
1, = :a In Iax + bx + c I+ ( B- ~:) I 1
Example 2. Evaluate the integral

x+3
I=
S X- 2X - 5 dx.
Integrals of Functions Containing a Quadratic Trinomial 353
Applying the foregoing technique we have
-J
I-
x+ 3
z.- 2
x x - 5-
-~{(2x-2>+(3+}2) dx =
xz -2x - 5
s
- _!_ (2x-2) dx 4
- 2 x2 -2.t-5 +
dx
x2 -2x-5
s
=}lnJx 2 -2x-51+4 S(x-~;._ 6 =
.
=-
I 2 I /J/6-(x-1)1
Y
2 lnJ..:

-2x-5J+4.r_ln
r 6
+C.
6+(x-l)
III. Let us consider the integral
dx
SYax+bx+c
Bv means of transformations considered in Item I, this integral
reduces (depending on the sign of a) to tabular integrals of the
form
which have already been examined in the Table ol Integrals (see

formulas 13' anci 14).
IV. An integral of the form
Ax+B d
SYax +bx+c 2 X
is evaluated by means of .the following transformations, which

are similar to those considered in Item II:
A (2ax+ b)+ ( B-2ci
~ 2a" A b)
~
Ax+B
--:r-==:;::::::::::;:==
2
dx = dx =
Jfax +bx +c Jfax 2 + bx+c
= ~
2a
s 2ax + b
Jfax 2 +b~+c
. dx + ( B _A b )
2a
s dx
Jfax 2 +bx+c"
Applying substitution to the first of the integrals obtained,

ax 2 +bx+c=t, (2ax+b)dx=dt,
we get
(2ax+ = sb)dx dt__ = 2VT + C = 2 Vax + bx+c +C.
SJfax +bx+c
2
Yl .
The second integral was considered in Item Ill of this section.
12-3388
Example 3.
5
Sx+ 3 )"2(2x+4)+(3-IO)
) -;r:=:;=:;~=;=;:;;: dx = dx =
Y x2 +4x+ 10 + 10
= ~
2
\
J Yx
2x + 4
2
+4x+ 10
dx -7 s
Yx"+4x
dx
V(x+2) 2 + 6
=
=5 Vx"+4x+ 10-71n 1x+2+ Y<x+2) 2 +61 +C=

=5Yx'+4x+ 10-71n I x+2+ Yx 2 +4x+ 10 I+C.
SEC;- 6. INTEGRATION BY PARTS
Let u and v be two differentiable functions of x. Then the

differential of the product uv is found from the following formula:
d (uv) = u dv = v du.
Whence, by integration, we have
uv = ~ u dv + ~ v du
or
~udv==uv-~vdu. (I)
This formula is called the formula of integration by. parts. It is
most frequently used in the integration of expressions that may
be represented in the form of a product of two factors u and dv
in such a way that the finding of the function v from its differen-
tial dv, and the evalua.tion of the integral ~ v du should, taken
together, be a simpler problem than the direct evaluation of the
integral ~ u dv. T'o become adept at breaking up a given element
of integration into the factors u and dv, orie has to solve
problems; we shall show how this is done in a number of cases.
Example 1. ~ x sin x dx =? . We let

u=x, dv=slnxdx;
then.
du =dx, t1 =-cos x.
Hence,
~ x slnx dx= -x cosx + ~ cosx dx= -x cos x+slnx +C.
Note. When determining the function v from the differential
dv we can take any arbitrary constant, since it does not enter
into the final result [this can be seen by putting the expression
'I
Integration by Parts 355
v +C into (1) in place of o]. It is therefore convcnl~nt to con

sider this constant equal to zero.
The rule for integration by parts is widely used. For example,
integrals of the form
~ xk sin ax dx, ~ xk cos ax dx,
~ xkea"dx, ~ xk In x dx,
and certain integrals containing inverse trigonometric functions
are evaluated by means of integration by parts.
Example 2. It is required to evaluate~ arctanxdx. LettinJ U=arctan.r,

dx
dv ='d~, we have du =
1
+ x', v =x, Thus,
Sarc tan x dx=x arc tanx- S1x::,=xacc tanx-; In II +x'I+C.

Example 3. It is required to evaluate ) x'e" dx. Let us put u =X",
dv =e"dx; then du =2x dx, v +ex,
~ x'e" dx = xe"- 2 ~ xe" dx.
The last integral we again Integrate by parts, letting
u 1 =x, du 1 =dx,
dVJ =e" QX, V1 =e".
Then
Finally we get
~ x'e" dx =x'e" -2 (xe" -e") +C=x'e" ~2xe".+2e" +C =e" (x'-2x+2>+C.
Example 4. It is required to evaluate ~ (x 2 + 7x-5) cos 2x dx. We let
u=x'+ 7x-5; dv =Cos 2x dx; then
sln2x
du=(2x+1)dx, v = -- ,
2
S (x'+7x-5) cos2x dX=!x'+7x -5) -~2x

2
-- s
(2x+7l -~2x
2
- dx.
. 2x+7
Apply Integration by parts to the latter integral, lethng u 1 = -- , dv 1 =
2
=sin 2x dx; then
12*
3f6 Indefinite Integrals
Therefore, we _finally get

2 2
sin 2x cos 2x sin 2x
S(x +7x-5) cos 2x dx=(x +7x-5) - 2 -+(2x+7) - 4- - -4-+C.
Example 5. I=~ Va-xdx=?
Perform Identical transformations. Multiply and divide the integrand by

Jf a-x:
Jfa-xdx=J ;-x dx=aJ dx
S a-x V a-x
=;=a
2 x
arc sin--
a
s v
X
xdx
a-x

Integrate the latter integral by parts, letting
du=dx,
v =- V a-x;
1hen
r X
2
J Va'-x'=
dx s X
X dx
va-x
-xYa'-x'+ S Va 2 -x 2 dx.
Putting the last result In the earlier obtained expression of the given integral,
we will have
SVa -x dx=a arc ~In~ +x Va -x SVa -x dx.

2 2 2 2 2
-
2 1
Transposing the integral from right to left and performing elementary trans-
formations, we finally get
-- a X X -
2 -
SVa2-x2dx=2arc sin a-+"2 Jf a -x'+C.
Example 6. Evaluate the integrals
/1 = ~ e4 " cos bx dx and 12 = ~ e4 " sin bx dx.

Applying integration by parts to the first Integral, we get
u=e=, du=aeax,
dv = cos.bx dx, ,,=_!_sinbx
b
Se 4
" cos bx dx = +e4 " sin bx- : Se 4
" sin bx dx.
Again apply the method of integration by parts to the last integral;
u=eax, du=ae 4 ",
1
dv =sin b~ dx, v=--cosbx
b .
Se 4
" sin bx dx = --1;- e 4
" cos bx +~ Se 4
" cos bx dx.
Partial Rational Fractions and Their Integration 357
Putting into the preceding equation the expression obtained gives us
From this equation let us find / 1 :
(I+~:) Seax cos bx dx=eax (; sin bx+ :~cos bx),

whence
ax b d eax (b sin bx+a cos bx)+C
I1=
5e COS X X= al+b 2
Similarly we find
_
I a- 5 ax . b dx- eax (a sin bx-b cos bx) +C
e sm x - a~+ b' .
SEC. 7. RATIONAL FRACTIONS. PARTIAL RATIONAL

FRACTIONS AND THEIR INTEGRATION
As we shall see below, not every elementary function by far

has an integral expressed in elementary functions. For this reason,
it is very important to separate out those classes of functions
whose integrals are expressed in terms of elementary functions.
The simplest of these classes is the class of rational functions.
Every rational function may be represented in the form of a
rational fraction, that is to say, as a ratio of two polynomials:
Q (x) 8 0 xm+B,xm- 1 + ... + Bm
f(x) = A 0 xn+A 1 xn 1 + ... +An
Without restricting the generality of our reasoning, we shall
assume that these polynomials do not have common roots.
If the degree of the numerator is lower than that of the denomina-
tor, then the fraction is called proper, otherwise the fraction is
called improper.
If the fraction is an improper one, then by dividing the nu-
merator by the denominator (by the rule of division of polyno-
mials), it is possible to represent the fraction as the sum of a
polynomial and a proper fraction:
Q (x) = M (x)
f(x)
F (x)
f (x) '
+
here M (x) is a polynomial, and ~ ~;~ is a proper fraction.
Example 1. Given an improper rational !raction
x4-~
x 2 +2x +1
358 !_ndefinite I nlegrals
Dividing the numerator by the denominator (by the rule of. division of
polynomials), we get
x'-3 4x-6
x'+2x +I =x'-2x+3-x'+2x +I
Since integration of polynomials does not present any difficul-
ties, the basic barrier when integrating rational fractions is the
integration of proper rational fractions.
Definition. Proper rational fractions of the form:
A
I. x=a
II. (x~ a)k (k is a positive integer~ 2),
III. Ax+B (the roots of the denominator are complex, that
x'+ px+q
. p )
IS, 4-q<O ,
Ax+B
IV. (x+px+q)k (k is a positive integer ~ 2; the roots of the
denominator are complex) are called partial fractions of types I,
II, Ill, and IV.
It will be proved below (see Sec .. 8) that every rational fraction
may be represented as a sum of partial fractions. We shall there"
fore first consider integrals of partial fractions.
The integration of partial fractions of types I, II and III does
not present any particular difficulties so we shall perform their
integration without any remarks:
I. Sx~adx=:'Alnlx-ai+C.
II.
A
S--"dx= A
. (x- a)
s "
(x-a)- dx= A
(x-a)-k+
-k+l
.
+C= .
A .
- ",+C.
(1-k)(x-a) -
At+B ) ~ (2x+Pl+ ( B- ~p)

I I I ) x + px + q dx == x + px + q dx =
-~S 2x+p dx+(B-Ap)S dx =
- 2 x'+px+q 2 x'+px+q
= ~ lnix'+px+ql+(B-A/) \ ( P/x ( p')
J x+2 + q-4
= A lnlx'+px+q!+
2
~-Ap1 arctan V2x+p 2 +C
4q-p 4q-p
(see Sec. 5).
Partial Rational Fractions and Their Integration 359
The integration of partial fractions of type IV requires more

involved computations. Suppose we have an integral of this type:
IV. s Ax+B
(x 2 +px +q)k
dx.
Perform the transformations:
\ Ax+B dx= r4-(2x+p)+ ( B-~) dx=

J~+~+~ J ~+~+#
- _A_S 2x + p dx + ( B - A p) S dx
- 2 (x+px+q)k . 2 (x'+px+q)k"
The first integral is taken by substitution, x+px+ q = t;

(2x+ p) dx = dt:
2x+p dx=s~=st-kdt=t-k+l+C=
S(x'+px+q)k . tk .1-k
- . l . +C.
(l-k) (x'+px+q)k-l
We write the second integral (let us denote it by I k) in the form
I,~ i (x'+::H~ ~ i [(x+ f)': (q-~) r'- i (l';'m'J'

assuming
x+ ~ =t, dx=dt,
(it is assumed that the roots of the denominator are complex,
and hence, q-f > 0). We then do as follows:
, dt l s(t +m 2 2
)-t 2
I k-- dt-
J(t'+m)k-- -m (t'+m)k -
= _!_ s
m u+m>k-1
dt - _!__ s
m W+m'>k .
t dt (1)
Transform the last integral:

t2 dl
SW+m')k =
sW+mjk
ltdt
_ _!,_S'dw+m'l__: _ _._l_Std( 1 )
- 2 (t"+m')k - 2(k-l) W+m')k- 1
Integrating by parts we get

t'dt
S(t'+m')k =
I -s
Putting this expression into (1), we have
k-
dt
(t'+m')k m
I
JI (t'+ dtm)k- +
+ ~'2(k -l> [ua+~>k-- S(l~+d~>k-] =
1
= t + 2k-3
2m 2 (k-l) (t'+m 2 )k-
dt
2m 2 (k:.._ I)
s UZ+ m)k-

On the right side is an integral of the same type as I"' but. the
exponent of the .denominator of the integrand is less by unity
(k-1); we have thus expressed I k in terms of I k-
Continuing in the same manner we will arrive at the familiar
integrai
dt 1 t
I.= 5t+m=m arctan-;n+C.
Then substituting everywhere in place of t and m their values, we
get the expression of integral IV in terms of .x and the given
numbers A, B, p, q.
Example 2.
~ 5 ~
_!_ (2x + 2) + (-1 -I)
x-1 2
(x 2 +2x+3) 2 dx= (x 2 +2x+3) 2 dx=
_ _!_ 2x+2 dx- 25 dx

- 2 (x 2 +2x+3) 2 (x'+2x+3) 1
__ _!_
-
I
2 (x 2 +2x+3)
2 5 dx
(x 2 +2x+3)2
.
We apply the substitution x+ 1 =I to the last integral:

dx dx _
5 dt ..!..
5W+2>-t' '=
5(x 2 +2x+3} 2 5l(x+ 1} 2 +2) 2 - (t 2 + 2) 2 2 (1 1 +2) 2 d
5 dt I 5 t
I
=2 1 +2-2 (1 +2) dl=
2 2
2
I
='2
I t 1
y2"arctan y2'-2
s+ (t 2
t dt
2 )a'
Decomposition of a Rational Fraction into Partial Fractions 361
Let us consider the last integral:
5
2
t dt I
(12+2)2=2
5 td (t' + 2)
(!2+2)2 = -2
1 5( td f2
1 )
+ 2 =
=
1
-2 12
t 5
+2+2 t +2=
1
2
dt
t 1 t
= - 2 (t2 + 2) + 2 Jf"2 arc tany 2
(we do not yet write the arbitrary constant but will take it into account In
the final result).
Consequently,
dx 1 x+ 1
5
(x2+ 2x+3)z 2 V2arc tan Y2-
-
1[
2
x+1
2 (x2 + 2x + 3) + 2
1
Y2 arc tan
x+1]
Jf2
Finally we get
x-1 x+2
5 (x 2 +2x+3) 2 dx=- 2 (x 2 +2x+3)
SEC 8. DECOMPOSITION OF A RATIONAL FRACTION

INTO PARTIAL FRACTIONS
We shall now show that every proper rational fraction may be

decomposed into a sum of partial fractions.
Suppose we have a proper rational fraction
F (x)
f (x)
We shall assume that the coefficients of the polynomials are real
numbers and that the given fraction is nonreducible (this means
that the numerator and denominator do not have common roots).
Theorem 1. Let x =a be a root of the. denominator of multi-
plicity k; that is f(x)=(x-a)k{,(x} where f,(a)=I=O (see Sec. 6,
Ch. VII). Then the given proper fraction f(~) may be represented
in the form of a sum of two other proper fractions as follows:
F (x)=_A_+ F 1 (x)
f (x) (x-a)k (x-a)k-tf, (x)' (1)
where A is a con,stant not equal to zero, and F, (x) is a polyno-
mial whose degree is less than the degree of the denominator
(x-a)k-t f, (x).
Proof. Let us write the identity
F (x) = - A - + F (x)-Af, (x)
f(x) (x-a)k (x-a)kfdx) (2)
(which is true for every A) and let us define the constant A so that
the polynomial F (x)- Af, (x) can be divided by x_:a. For this,
by the remainder theorem, it is necessary and sufficient that the
following equality be fulfilled:
F (a)- Af, (a)= 0.
Since f, (a) =1= 0, F (a) =1= 0, A is uniquely defined by
A-F(a)
-/,(a).
For such an A we shall have

F (x)- Af, (x) = (x-~) F, (x),
where F, (x) is a polynomi-al of degree less than that of the

polynomial (x-a)k-'f, (x). Cancelling (x-a) from the fraction in
formula (2), we get (1) ..
Corollary. Similar reasoning may be applied to the proper ra-
tional fraction
(x-a)k-l {,.(x) '
in equation (1). Thus, if the denominator has a root x =a of

multiplicity k, one can write
F(x)=_A_+ A, + +Ak_,+Fk(x)
/(x) (x-a)k (x-a)k-I x-a /,(x)'
where Fk(x) is a proper nonreducible fraction. To it we can apply

!, (x)
the theorem that has just been proved, provided f, (x) has other
real roots.
Let us 'now consider the case of complex roots of the denomi-
nator. Recall that the complex roots of a polynomial with real
coefficients are always conjugate in pairs (see Sec. 8, Ch. VII).
When factoring a polynomial into real factors, to each pair of
complex roots of the polynomial there corresponds an expression
of the form x 1 +px+q. But if the complex roots are of multi-
plicity !J., they correspond to the expression (x 1 +px+q)l'.
Theorem2. lff(x)=(x 1 +px+q)l'<p,(x), where the polynomial
<p, (x) is not divisible by x 1 + px+ q, then the proper rational
F(x)
fraction (x) may be represented as a sum of two other proper
1
fractions in the following manner:
F (x) Mx+N ID 1 (x) 3)
f (x) = + + + (x + px + q)~'-- 1 q>
(x 2 px q)l" 1
1 (x) (
Decomposition of a Rational Fraction into Partial Fractions_ 363
where , (x) is a polynomial of degree less than that of the poly
nomial (x+px+q)""-'<p,(x).
Proof. Let us write the identity
F (x) F (x) Mx+N F(x)-(Mx+N) q>.(x)
f (x) = (x+ px + q)~'-q>, (x) = (x 2 + px + q)l'- + (x 2 + px+ q)l'- q>.(x) (
4)
which is true for all M and N, and let us define M and N so

that the polynomial F (x)- (Mx + N) <p, (x) is divisible by
x+px+q. To do this, it is necessary and sufficient that the
equation
F (x)-(Mx+ N) <p, (x) = 0
have the same roots d.i~ as the polynomial x + px + q. Thus,
F (a+ i~)-lM (a+ i~) + Nl <p, (ex+ i~) = 0
or
F!a+i~)
M( a; + tp'R)+N= q>,(a+i~)'
B t F(a+iM is a definite complex number which may be written
u q>, (a+iM
in the form K + iL, where K and L are certain real numbers.
Thus,
M (a+ i~) + N = K + iL;
whence
Ma+N=K. M~=L
or
M=1-'
L
N --1(~-La . ~
With these values of the_ coefficients M and N the polynomial

F (x)- (Mx_+ N) <p,(x) has the number a.+ i~ for a root, and,
hence, also the conjugate number cx-i~. But theri the polynomial
can be divided, without any remainder, by- the differences
x-(a+i~) and x-(a.~i~). and, therefore, by their product,
which is x + px + q. Denoting the quotient of this division by
<D, (x), we get
F (x)-(Mx+ N) <p, (x) = (x'+ px+ q)a>, (x).
Cancelling x'+px+q from the last fraction in (4), we get (3),
and it is clear that the degree of <D, (x) is less than that of the
denominator, which is what we set out to prove.
Now applying to the proper fraction ~~;~ the results of Theorems
1 and 2, we can obtain,- successively, all the partial fractions
:'l64 Indefinite Integrals
corresponding to all the roots of the denominator f (x). Thus,

from the foregoing there follows the result that
If
f (x) = (x-ar (x- b)~ ... (x + px + q)"' .. . (x+ lx+s)',
then the fraction f~~~ can be represented as follows:
F(x)
f(x) =
A
(x~ a)a
+ (x-at
A,
_,
+ + X=ti
A_, +
8 8, 8~- 1
+ (x-b)~+ (x-b)~- + + x-b +
. . ~~~~ . . . ~~~~~~ . . . . . M~_-,x+N~_,. 1 (5)
+ (x 2 +px+q)~'- +(x 2 +px+q)l'--+ x 2 +px+q+ +
0 0 0 0 0
+ Px+Q + P,x+Q 1 + + P,_ x+Q,_ 1 1

(x"+lx+sf (x 2 +lx+s)'- 1 x'+lx+s
The coefficients A, A, ... B, BJ' . . . may be determined by
the following reasoning. This equality is an identity; and for
this reason, by reducing the fractions to a common denominator
we get identical polynomials in the numerators on the right and
left. Equating the coefficients of the same degrees of x, we get
a system of equations to determine the unknown coefficients A,
A .. ... , B, 8 1 ,
In addition, to determine the coefficients we can take advan-
tage of the following: since the polynomials obtained on the
right and left sides of the equality must be identically equal after
reducing to a common denominator, their values are equal for all
particular values of x. Assigning particular values to x, we get
equations for determining the coefficients.
We thus see that every proper rational fraction may be repre-
sented in the form of a sum of partial rational fractions.
Example. Let it be required to decompose the fraction (x+~~~;_ 2) into

partial fractions. From (5) we have
x"+2 A A1 A, 8
(x + 1) 8 (x-2) (x + 1) 8 + (x + 1) 2 + x+ 1 + x-2
Reducing to a common ctenominator and equating the numerators, we
have
x 2 +2= A (x-2)+ A 1 (x+ I) (x-2)+A 2 (x + 1) 1 (x-2) + B (x + 1)1 , (6)
or
x"+2= (A,+ B) x"+(A 1 + 38) x'+
+<A -A 1 -3A 2 +3B) x+(- 2A-2A,-2A 2 + 8>-
Integration of Ratr:onal Fractions 365
Equating the coefficients of x 3 , x 2 , x1 , x 0 (absolute term), we get a system

of equations for determining the coefficients:
0=A 2 +B.
l=A,+3B,
O=A -A,-3A 2 +3B,
2=- 2A -2A, -2A 2 +B.
Solving this system we find
I 2 2
A=-!; A,= 3 : A2 =--g: B=-g
It might also be possible to determine some of the coefficients of the
equations that result for some particular values of x from equality (6),
which is an identity in x.
Thus, setting x=-1 we have 3=-3A or A=- I; setting x=2, v:e
2
have 6=278; B=-g
If to these two equations we add two .equations that result from equating
the coefficients of the same powers of x, we get four equations for deter-
mining the four unknown coefficients. As a result, we have the decomposition
x 2 +2 I I 2 2
(x+l) 3 (x-2)=-Ct"+ 1) 3 +3(x+l) 2 9 (x+l)+9(x-2)
SEC. 9. INTEGRATION Of RATIONAL FRACTIONS
Let it be required to evaluate the integral of a rational fraction

f<~~l; that is. the integral
Sf<~~) dx.
If the given fraction is improper, we represent it as the sum
of a polynomial M (x) and the proper rational fraction ~(~)
(see Sec. 7). This latter we represent, applying formula (5), Sec. 8,
as a sum of partial fractions. Thus, the integration of a rational
fraction reduces to the integration of a polynomial and several
partial fractions.
From the results of Sec. 8 it follows that the form of partial
fractions is determined by the roots of the denominator f (x).
Here, the following cases are possible.
Case I. The roots of the denominator are real and distinct,
that is
f(x)'= (x-a) (x-b) ... (x-d).
Here, the fraction ft? is decomposable into partial fractions
of type 1:
366 Indefinite lnll'grals
and then
Sf~~~ dx = Sx~ a dx + Sx~ b dx + " + Sx D d dx =
=Ainjx-ai+Blnlx-bl+ ... +Dlnlx-di+C.
Case II. The roots o/ the denominator are real, and some of them
are multiple:
f (x) = (x-at (x- b)~ ... (x-.:. d) 6
In this case the fraction ~~:~ is decomposable into partial frac-

tions of types I and II.
Example I. (see example in Sec. 8, Ch. X).
x + 2 . . s dx
S(x+l) (x-2) dx=- <x+
1
I
1) 3 +3
s (x+.ll"-9
'
dx ' 2 s dx
x+l+
+g
25 x-2=2
dx I I I 2 2
<x+IJ'~ 3 (x+ I)---gin lx+ll+g-lnlx-2.1+C=
2x- I 2 -~x-21
=-6(x+l) 2 +-gln x+l +C.
Case Ill. Among the roots o/ the denofrtinator there- are . complex
nonrepeating (that is, distinct) roots:
f(x) = (x" + px + q) (x+ lx + s) .. (x-at ... (x- d)0
In this case the fraction n;~ is decomposable into partial frac-
tions of types I, II, and III.
Example 2. Evaluate the integral

x dx
S(x + I) (x :- I) '
Decompose the fraction under the integral sign into partial fractions [see (5),
Sec. 8, Ch. X]
x Ax+ B C
(x 2 + I) (x-1) = x1 + I + x-1 ~
Consequently,
x=(Ax+B) (x-l)+C (x 2 + I),
Setting x=l, we get I= 2C, C= ~;setting x=O, we_ get 0=-.B+C,

I
8=2
Integration cf.Rational Fractions 367
Equating the coefficients. of x, we get" 0 =-A + C, whence. A,=-~. Thus,
5 (x + ~)~:- U = - ~ 5; + \ dx + ~ 5x~ 1 =
=-2
I 5xdx I
x"+1 +2
5 dx 1
x+ 1+2
5 dx
x-1=
1 I 1
=-
4 lnJx'+II+ 2 arctanx+ 2 lnJx-IJ+C.
Case IV. Among the roots of the denominator there are complex
multiple toots
f (x) = (X 2 + px+ q)"" (x" + lx+ sr .. .(x--at .. . (x-d)~.
In this ease, decomposition of the fraction f~~~ will also contain
partial fractions of type IV.
Example 3. It is required to evaluate the integral
x 4 +4x'+ llx 2 + 12x+8 d
5 (x 2 +2x+3) 2 (x+ I) x.
Solution. Decompose the fraction Into partial fractions:
x 4 +4x'.f-1Jx"+l2x+B Ax+B . Cx+D .E
(x 2 +2x+3) 2 (x+ l) (x 2 + 2x+3) 2 + (x 2 +2x +3) + x+ 1'
whence
x 4 +4x1 + llx"+ 12x+8=
=(Ax+ B) (x +I) +(Cx+ D) (x"+2x+ 3) (x+ I)+ E (x 2 +2x+3) 1 .
Combining the above-indicated methods of' determining coefficients, we find
A= l, B = - l, C = 0, D = 0, E = l.
Thus, we get
5 x+4x'+11x"+12x+8
rx+2x+3) 2 (x+l) dx=
5(x +2x+3)
2
x-1 5 dx
dx-{- x+1 =
2
x+2
=- 2(x"+ 2x+ 3)
V2 x+1
- 4-arctan ]12 +lnJx+1l+C.
The first integral on the right was considered in Example 2, Sec. 7, Ch. X.
The second integral is taken directly.
From the foregoing it follows that the integral of any rational

function may be expressed in terms of elementary functions in
fmal form, namely, in terms of:
l) logarithms in the case of partial fractions of type I;
2) rational functions in the case of partial fractions of type II;
3) logarithms and arc tangents in the case of partial fractions

of type III;
4) rational functions and arc tangents in the case of partial
fractions of type IV.
SEC. tO. OSTROGRADSKY'S METHOD
In the case of multiple roots in the denominator, the integral

of a rational function may be evaluated by a different method
that leads to simpler computations. This method permits separat-
ing out the rational part of the integral without decomposing
the fraction. into partial fractions, and then integrating the rational
fraction whose denominator has only simple roots. It is easy to
integrate such a fraction since it is decompo~able into partial
fractions of types I and III. This method belongs to the noted
Russian mathematician M. V. Ostrogradsky (1801-1862) and is
based on the following reasoning.
Let it be required to integrate the proper rational fraction
F (x)
f (x) , where
f (x) = (x-a)" (x- b)~ . .. (x1 + px + q)".
Here, on the basis of (5), Sec. 8, everything is reduced to inte-
grating proper rational fractions of four types (see Sec. 7). Here,
I) the integral of a fraction of the form - A(
x-a
>" is a fraction. of
A*
the form (x-a )" 1 ,
2) the integral of the fraction Mx+ N is a sum of frac-
(x2+px+q)"'
M*x+N* *
tions of the form (xz+px+q)~'-* where fJ. .;;;;;fJ.-1, and of an in-
tegral of the form
We will not yet integrate fractions of types I and III.

Combining the rational fractiOns obtained after integrating
fractions of types II and IV, we get a proper fraction of the
form ~ ~;i, where the polynomial Q (x) is equal to
Q (x) = (x-a)"- 1 (x-b)~-~ ... (r + px+ q)"'- 1 (r + lx+s)'- 1
Y (x) is a polynomial of degree one less than that of the poly-
nomial Q.
Ostrogradsky's Method 369
Combining the integrals of all the fractions of types I and III

(including those integrals of the form x px q
S ,_;**+
dx which are
obtained by integration of fractions of type IV), we get an integral
of a proper fraction of the form ~~;~, where the polynomial
P (x) is
P (x) = (x- a)(x- b) .. (x 2 + px + q) .. -. (x 2 + tx + s).
We thus find that
F (x) dx _ Y (x)
S l (x) - Q (x)
+ s
X (x) dx
P (x)
(1)
Here X (x) is a polynomial of degree one less than that of the

polynomial P (x).
No.w let us determine the polynomials X (x) and Y (x) in the
numerators. To do this, differentiate both sides of (1):
F(x) = QY'-Q'Y +~
l (x) Q2 P
or
F( X)
= f(x)Y' __ f(x)Q'Y +f,(x)X
(2)
Q Q2 p
We shall show that the expression on the right is a polynomial.

Noting that f (x) = PQ we can rewrite (2) in the form
F(x) = PY'- p~y + QX. (2')
What remains now is to prove that the expression - P~Y is a

polynomial or that PQ' is divisible by Q. We note that
~ =[In Q ]'=[(a- 1) In (x-a) +(~~I) In (x-b) + ...
. . . + (f.t-1) In (x + px+ q) +
2
... +(v-1) In (x 2 + lx+s>l' =
=a-1+~-1+ +(J1-I)(2x+p)+ +(v-1)(2x+l)
x- a x- b x2 + px + q x2 + lx + s
The polynomial P is the common denominator of the fractions
on the right side. In the numerator there will be a certain poly-
nomial of degree less than that of P. Let us denote it by T. Then,
Q' T
Q=p
Hence, the expression
Q' T
P QY=P p-Y=TY
370 : Indefinite Integrals ;
is a polynomial.. Equation. (2') takes. the form ,

F (x) =PY' -TY + 'QX. (3)
Comparing ,the coefficients of the same powers of the variable
in (3), we get a system of equations from which we find the
unknown coefficients of the polynomials X and Y.
Example. Evaluate
S(x ~ I ) d.
2
Solution. In this case, .

f (x)= (x-1) 2 (x 2 +x+ 1)2,
P (x) = (x-1) (x 2 -f-x + l)=X 1 -"-l,
Q (x)= =x-t.
Equation (1) has the form
~_Ax +Bx+C +S Ex +Fx+O dx
2 2
(4)
1)2 -
S (x1 - x1 - l x-L
Differentiating both sides of (4) we get
1 (x 1 -l) (2Ax+ B)-(Ax 2 +Bx+C) 3x 1 Ex 2 +Fx+O
(x 1 -l) 2 = . (x 1 -l) 2 + x1 - l
Clearing fractions, we have
1=(x1 -l) (2Ax+ B)-(Ax 1 + Bx+C) 3x 2 -l-(x 1 -l) (Ex'+Fx-1-0).
Equating the coefficients of identical powers of x on different sides of the
equation, we get a system of six equations for determining the coefficients,
A, B, C, E, F. 0.
O=E,
0=-A+F.
0= -28+0.
0=3C-,
0= -2A-F,
1= -B-0.
Solving this system we find
I
E=O, A =0, C=O, 8=-:r F=O,
Putting the values of the coefficients thus found into (4), we get
r dx
J(x'-1) 2 =
-!X + Jr -!
x-1 x- 1 dx.
The denominator of the latter integral has only simple roots, thus making it
easy to compute the integral. We finally obtain
r' dx -X r[-g2 gX-1-g

2 4]
J (x -1) =3 (x -l) + J x-L + x+x+ 1
3 2 1 dx=
x 2 -V3
92 In (x-'- 11 +g1 In (x t +x+.1)-1--. t 2x-l- 1 -1-C .
3 (x 2 _ 1) 9- arc an V3'
Integrals of Irrational Functions 371
SEC. II. INTEGRALS OF IRRATIONAL FUNCTIONS
It is impossible to express in terms of elementary functions the

integral of every irrational function. In this and the following
sections we shall consider irrational functions whose integrals are
reduced (by means of substitution) to integrals of rational functions
and, consequently, are integrated to the end.
m '
l. We consider the integral ~ R ( x, x 11 , ... , x5 ) dx where R
is a rational function of its arguments.*)
Let k be a common denominator of the fracfions!!!.. n
, . , .!....
s
We
make the substitution
x=t", dx=ktll- 1 dt.
Then each fractional power of x will be expres~ed in terms of
an integral power oft and the integrand will thus be transformed
into a rational function of t,
Example 1. It is required to compute the integral
r x;:,.
J x
Solution. The common denominator of the fractions { . f Is 4: and so we

substitute: x=t\ dx=4t 1 dt; then
~ ~ 2
x dx
~=4 5 1
t+i t dt=4 5
1 t at=4 5
t+i ( t- .+
r ) dt=
1 1
x +I
=45t'dt-4 5~dl=4~-i.lnjl
I +I 3 3
1 +1l +C=
=~ [x7-lnjx~+ll]+c.
1 The notation R (x, x.!fi- ... , x-f)
m
indicates that only rational operations
r
are performed on the quantities x, x n , ... , x s.
This is precisely the way that the following notations are henceforwan1 to
m
be understood: R(x, (ax+b) r ax + bx +c), R (sin x, cos x),

ex+ d n .. .. ) , R (x. .r
etc. For instance, the notation R (slnx, cosx) indtca~es that rational opera
t ions are to be performed on sin x and cos x.
372 I ndefinile Integrals
II. Now consider an integral of the form

m
SR [x.(::t!)n ..... (:;t!rldx.

This integral reduces to the integral of a rational function by
means of substitution:
ax+b =tk
cx+d '
r
where k is the common denominator of the fractions : , .... ,s
s Vxx+4 dx.
Solution. We make the substitution x+4=t 2 , x=t'-4; dx=2tdt; then
SVx+4
- x - dx=2 s t 2 _t
2
4 dt =2
J( I+ 12 _4
4
) dt =2 s s
dt +B dt
11 _ 4 =
=2t+21n 1t-2~
I+ 2 +C=2 .
rr -
x+4+2ln Vx+4-21
Vx+ + +C.
4 2
I
SEC. 12. INTEGRALS OF THE FORM ~R(x, Vax"+bx+c)dx
Let us consider the integral

~R (x, Vax 2 +bx+c)dx. (1)
An Integral of this kind reduces to the integral of a rational

function of a new variable by means of the following Euler sub-
stitutions.
I. First Euler substitution. If a>O, then we put
Vax 2 + bx+c= Vax+t.
For the sake of definiteness we take the plus sign in front of Va.
Then
ax'+ bx+c=ax1 +2 Vaxt +t,
\Vhenc<! x is determined as a rational function of t:
tz-c
x----=:-
- x-2 Vat
Integrals of the Form~ R ( x, Yax+ bx+c) dx 373
(thus, dx will also be expressed rationally in terms of t). Therefore,
Vax 1 + bx+c= Vax+t = va b-2tt-V a +t.

+ +
and Vax 2 bx c is a rational function of t.
+ +
Since Vax bx c, x and dx are expressed rationally in terms
of t, the given integral (1) is transformed into an integral of a
rational function of t.
Example 1. It Is required to compute the integral
.fv;~c
Solution. Since here a= 1 > 0, we put Y x + C = -x + t;
2
then
whence
t-c
X=----cg-
Consequently,
t+t
dx =2t"J dt,
t-c =t 2-+C
Yx +C=-x+t=---+t
2
2t 2t-
Returning to the initial integral, we have
t'+C
dx 2f2 dt 5T=lnlt!+C
dt -- .
SYxz+c= S~ =
2
=1nlx+Yx +CI+CL
1
2t
[see formula 14 in the Table of Integrals).
2. Second Euler substitution. If c>O, we put

V ax 2 + bx+c=xtVC;
then
ax 2 +bx+c=x 2t 2 +2xt Vc+c.
(For the sake of definiteness we took the plus sign in front of
the radical.) Then x is determined as a rational function of t:
2 Vct-b
X= a-t 2
Since dx and V-ax.....,::-+-:----:b:-x-+-,--c are also expressed rationally in terms
+
of t, by substituting the values of x, Vax bx -+- c and dx into
374 lndefinlte integral~
the integral~ R (x, Vax + bx +c) di, we reduce it to an integral

of a rational function of t.
Example 2. It is required to corn pule the integral
2
(1- Y1 +x+x 2) dx.
5 x VI +x+x'
Solution. We set VI +x+x'-xl +I; then
21-1
dx= 21 (1-1
2 -21+2
I +x+x"=x 212 +2xl + J x=--
. 1-1 2 ' 2) 1
dt
t'-t+l
Vl+x+x'=xt+1= 1
_
1
, ;
1- r l+x+x 2 = -2t"+t
.r _ ,
1 1
Putting the expressions obtained Into the original integral, we find
+x+x)' dx =S(-2t'+WO-t"l"(l-l"l (21-21 + 2> d _
('(1- J11
J x'Yl+x+x' 11-12 ) 1 (21-1) 2 (1 2 -1+1)(1-/ 1 ) 2 X-
= +2 5bdt+C=-2t+ln ~~~+C=
2(Vl +x+x'-l) +In lx+ J/1 +x+x'-II+C=
x x- Vr +x+x+ I
2
( J/l +x+x'-l) +In J2x+2 VI +x+x+ 11 +C.
X .
3. Third Euler substitutia,n. Let a and ~ be the rea! roots of

the trinomial ax + bx +c. We put
Vax + bx + C= (x-a) t.
Since ax +bx+c=a(x-a)(x-~).
1
we have
Va (x-a) (x-~)= (x-a) t,
a (x-a) (x- ~) = (x-a)' t,
a (x-M = (x-a) t.
Whence we find x as a rational function of t:
a~-a/ 1
X= a-t
Since dx and Vax + bx + c also rationally depend upon t, the

given integral is transformed into an integral of a rational function
of t.
Integration of Binomial Differentials ~75
Note t. The third Euler substitution is applicable not only for

a<O, but also for a>O, provided the polynomial ax + bx +c has
two real roots.
dx
SVx + 3x-4
2
Solution. Since x 2 +3x-4=(x+4) (x-1), we put

V(x+4) (x-1)=(x+4) t;
then
(x+4) (x-1)=(x+4) 2 t 2 , x-1 =(x+4) t 1,
1 +4t 1 lOt
x=_T=fF dx=( 1-t 2) 2 dt,
.. [1 +4t 2 ]' 5t
V(x+4)(x-1)= l-t 2 +4 t=i-t
Returning to the original integral, we have

2
dx sl0t(i-t ) s 2 ll+tl
S Jfxt+3x-4= (J-t2)Z5t dt= 1-tzdt=ln 1-t +C=
=In I+ Y~ +C=In, Vx+4.:_

_ , / x-1
VX+4+ Jfx=TI+C.
Vx-1
1
V x+4
Note 2. It will be noted that to reduce integral (1) to an integral
of a rational function, the first and third Euler substitutions are
+
sufficient. Let us consider the trinomial ax 2 bx +c. If b 1 -4ac > 0,
then the roots of the trinomial are real, and, hence, the third
Euler substitution is applicable. If b1 -4ac,.;;;; 0, then in this case
I
ax 2 + bx +c= 4a [(2ax+ b)'+ (4ac- b1 )]
and therefore the trinomial has the same sign as that of a. For
Vax 1 + bx + c to be real it is necessary that the tnnomial be posi-
tive, and we must have a>O. In this case, the first substitution
is applicable.
SEC. 13. INTEGRATION OF BINOMIAL DIFFERENTIALS
An expression of the form

xm (a+ bxn)p dx,
where m, n, p, a, b are constants 1s called a binomial di(jerential.
376 Indefinite lntl?grals
Theorem. The integral of a binomial differential

~ xm (a+ bxn)P dx
if m, n, p are rational numbers, is reduced to an integral of a ra-

tional function and thus is expressed in terms of elementary func-
tions in the following three cases:
1. p is an integer (positive, negative or zero);
1
2. m+ is an integer (positive, negative or zero);
n
1
3. m+
n
+ p is an integer (positive, negative or zero).
Proof. Transform the given integral by substitution:
I I
- I --1
X=zn, dx=-
n
zn dz.
Then
5xm (a+ bxn)P dx = :

1
5zmn+~_ 1\a+ bz)P dz = ! 5zq (a+ bz)P dz, (I)
where
q=m+l_l.
n
1. Let p be an integer. Since q is a rational number, we denote
it by .!_. Integral (1) is then of the form
s
,
~ R (zs, z) dx.
As was pointed out in Sec. II, Ch. X, it reduces to an integral of a
rational function by the substitution z = ts.
2. Let m+l be an integer. Then q=m+l_I is also an integer.
n n
The number p is rational, p=!::__. Here the integral (I) is of the
J.L
form
~ R [xq, (a+ bz)~] dx. "
This integral was considered in Sec. II, Ch. X. It reduces to an
integral of a rational function by the substitution a+ bz = tiJ..
1
3. Let m+l+p be an integer. But then m+ -1-tp=q+p is
n n
an integer. We transform integral (1):
5zq (a+ bz)P dz 5zQ+P (a~bzydz,

=
Integration of Binomial Dif!erentia/s 377
where q + p is an integer and p = ~ is a rational number. The

latter integral belongs to the class of integrals
k
SR [z.(a~bz)T] dz.
This integral was considered in Sec. 11, Ch. X. It reduces to an
integral of a rational function by the substitution a+ bz = t 1
' z
Let us examine examples of integration in all the three cases.
S
dx
Example I. Vx 2 (l+Vx 2 )=
s x
2
-~
(l+x) 3
1 -1
dx. Here p=-1 (integer).
2.
Now make the substitution z 2 =t. Then z=l 1 , dz=21dl and
Example 2. sV xa
1-x 2
dx = s
x (1 -x
3 2
)-
1
2 dx.
Here,
1
m=3, n=2,
I
p=-;, m~l=2(integerl). Wesubstitutex =z;thenx=z-2 ,

2
dx=-}z- 2 dz
and
2.
In order to make the second parenthesis rational we put (l-z) 2 = t; then
l-z=l ; z=t -l; dz=2tdt. Hence,
1 2
~
Example 3. jx
n
2
V dx
(I +x 2
)'
=

x- 2(1 +x')
_..!_
dx. Here, m=-2, n=2, p=.-'--
2

3
1
and m +
n
+p= -2 (integer). We reduce the expression in the parentheses to
a linear function: 1 1
- I --
x 2 =z; x = z 2 ; dx = - z 2 dz;
2
1
5x- (1 +x2 2) --,:dx= 5z-'(1 +z)-2{-z -~dz=

=} 5z --i-(1 +z) -~dz=; 5z-C ~zt+dz.
he first factor is a rational function. "In order to make 'the second factor
rational as well, we make the substitution:
1
c~zt =t;
then
I +z _ z. _ I _ 21 dt
--:-z-t, z- 1._ 1 , dz--(t'-l)''
Thus,
1 1 1 1
= -(1 +z)!
z '
-(-z )2 + C=-(l+x')2 -(___::._)! +C=
+z x 1 +x 2
1 2
_ Yl+x 2 _ x, +C.
x Yl +x'
Note. The noted Russian mathematician P. L. Chebyshev proved
that only in the above three cases in an integral of binomial
differentials with rational exponents expressed in terms of elemen-
tary functions (provided, of course, that a+.O and b=FO). But if
neither p, nor m + 1 , nor m + + p are integers, then the integrai
1
n n
cannot be expressed in terms of elementary functions.
SEC. 14. INTEGRATION OF CERTAIN CLASSES

OF TRIGONOMETRIC FUNCTIONS
Up to now we have made a systematic study only of the integ-

rals of algebraic functions (rationa 1 and irrational). In this section
we shall consider integrals of certain Classes of nonalgebraic
Integration of. Certain Trigonometric Functions 379
functions, primarily trigonometric. Let us consider an integral .of

the. form
~ R (sin x, cos x) dx. (1)
We shall show that this integral, by the substitution

X
tan 2 = t (2)
always reduces to an integral of a rational function. Let us express
sin x and cos x in terms. of tan ~ , and hence, in terms of t:
2sin~cos~
X X X
2sln cos 2 2tan 2 21
. 2 2 2
SIOX= I
sln1 ~ -i-cos 2 ~ - I +lani = +t''
1
cos 2 ~-sln
2
2~
2 cos 2-s1n
2 x 1 x
2 1- t an z 2 x I -t
COS X= ---.,--1- - - . X X =
2
X = I+ t'
cos + stn I+ tan
2 2 2
And
2dt
x=2arctant, dx=l+t
In this way, sinx, cosx and dx are expressed rationally in
terms of t. Since a rational function of rational functions is a
rational function, by substituting the expressions obtained into the
integral (1) we get an integral of a rational function:
21 1-t]
SR [l+t''
2dt
SR(sinx, cosx)dx= l+t' 1_ 1,.
Example I. Consider the integral
Ss~:x
On the basis ~r the foregoing formulas we have
Ss~:x=S tt.'
12
= S=In I IJ+C=In Itan i I+C.
. I +t
This substitution enabies us to integrate any function of the
form R(cosx, sinx). For this reason it is sometimes calleu
a "universal trigonometric substitution". However, in practice It
[requentl.y leads to extremely complex rational Junctions .. It i&
therefore convenient to know some other substitutions (in addi-

tion to the "universal" one) that sometimes lead more quickly to
the desired end.
I) If an integral is of the form ~ R (sin x) cos x dx the substitu-
tion sin~=t, cosxdx=dt reduces this integral to the form
~ R (t)dt.
2) II the integral has the form ~ R (cos x) sin x dx, it is. reduced
to an integral of a rational function by the substitution cos x = t,
sinxdx=-dt.
3) If the integrand is dependent only on tan x, then the
substitution tan x=t, x=arc tan t, dx= ~ , reduces this inte- 1 1
gral to an integral of a rational function:
SR (tan x) dx = SR (t) 1 ~ 1,.
4) If the integrand has the form R (sin x; cos x), but sin X and
cos x are involved only in even powers, then the same substitu-
tion is applied:
tanx= t, (2')
2
because sin x and cos x are expressed rationally in terms of tan x:
1 1
cos X= 1 +tan 2 x= l+t 1 '
2 tan' x
510 X= l +tan 2 x
dl
dx= l +tz
After the substitution we obtain an integral of a rational
function.
Example 2. Compute the mtegral .
+sin' x dx.
2 cosx
s
Solution. This integral is readily reduced to the form ~ R (cos x) sin x dx.
Indeed,
2
1
SJn X dX= SSillr XSillXdX sl-COS 1 X. d
S2 + COS X :l + COS X 2+
SlnX X.
COS X
We make the substitutiom cosx=z. Then sinxdx=-dzl
s2~nco: s~+:'
z 2
x dx=(-dz)= :z+; dz=
2
cos x
z-2+ z!2) dz= s S(
= -2z+3In (z+2>+C=---2 cosx+3ln (cosx+2)+C.
2 2
Integration of Certain Trigonometric Functions 381
dx
5
Example 3. Compute 2 -sln"x.
Make the substitution tan x = t:
~ 2-~~nx ~ ( 2 _~) (l + t) = -~ 2 : 1 J
.= 2 arc tan~~+
+C= J2 arc tan c;-n;) +C.
5) Now let us consider one more integral of the form

~ R (sin x, cosx) dx, namely an integr.al under the sign of which
is the product sinm x cos" x dx (where m and n are integers). Here
we shall have to consider three cases.
a) ~ sinm x cosn x dx, where m and n are such that at least one
of them is odd. For definiteness let us assume that n is odd. Put
n = 2p + 1 and transform the integral:
r sinm X cos'P+I X dX= ~ sinm X COS 2 p X COS X dx =
= ~ sinm x (1- sin 2 x)P cos x dx.
Change the variable
sin x= t, cos xdx= dt.
Putting the new variable into the given integral, we get
~ sinm x cosn x dx = ~ tm (1- t")P dt,
which is an integral of a rational function of t.
Example 4.
cos x dx=5 cos x cos x dx (l-sln 2 x) cos x dx
5 sin 4 x sln 4 x 5 sin l
Denoting slnx=t, cosxdx=dt, we get

2
cosxdx=5(!-t )dt 5dt 5dt I I
5 sin4 x t y - """iT=-3t+T+C=
I 1
=-3sin 1 x+ sin.t +C.
b) ~ sinni x cosn xdx, where m and n are nonnegative and even

numbers.
Put m = 2p, n = 2q. Write the familiar trigonometric formulas:
sm z X= 21 - 21 cos 2x, cos 2 x=-z1+1

0 2
2 cos x. (3)
382 I ndefinlle Integrals
Putting them into the integral we get

ssin 1
PX cosq X dx= S( }-{cos 2x r r
(~+!.cos 2x dx.
Powering and opening brackets, we get terms containing cos 2x
in odd and even powers. The terms with odd powers are integra-
ted as indicated in Case (a). We again reduce the even exponents
by formulas (3). Continuing in this manner we arrive at terms of
the form ~ cos kx dx, which can easily be integrated.
Example 5.
Jsln'xdx=~ S S(1-2 cos 2x+cos 2x) dx=
(1-cos2x) 2 dx=+ 2
=-!-. [x-sln 2x+~ S(I +cos 4x) dx l =} [-} x-sln 2x+ sln84x] +C.
c) If both exponents are even, and at least one of them is
negative, then the prec;eding technique does not give the d~sired
result. Here, one should m;tke the substttuti_b!J . tan =t' (or x
cotx= t).
Example 6.
S
2
sin xdx_sstn"x(stn"x+cosx)d
cos'x - cos'x
2
x- . an x
(l+t. z )ld
an x x.
-St
' dt .
Put tan x = t; then x =ate tan t, dx = + t and we get
Ssin!
cos x
dx=S t
X
2
(l +t 2 ) 2 ~=s
_tan x1
1
I + t2
t 2 (1 +t") dt =e:__+~+C
1
tan x+C
3 5
=
- 3 + 5 .
6) l n conclusion let us consider integrals of the form

~ cos mx cos nx dx, ~sin mx cos nx dx, ~ sin mx sin nx dx;
They are taken by means of the following*) formulas (m =1= n):
cos mx cos nx = ~ tcos (m + n) x +cos (m-n) x],
") These formulas are easily derivea as follows:

cos (m + n) x =cos mx cos nx ~sin mx sin nx;
cos (m -n) x =cos mx cos nx +sin mx sin nx.
Combining these equations termwise and dividing them In half, we get the
first of the three formulas. Subtracting termwise and dividing in half, we get
the third formula. The second formula is similarlv derived if we write analo
gous equations for sin (m + n)x and sin (m- n).x and then combine them
termwise.
Integration of Certain Irrational Functions ~83
sin mx cos nx={ [sin (m +n)x+ sin (m-11) x],
sirl mx sin nx= i [-cos (m + n) x+ cos (m-n)x].

Substituting and integrating, we get
S cosmxcosnxdx=; S[cos(m+n)x+cos(m-n)x]dx=
_sin(m +n) x + sin(m-n)x +C
- 2(m+n) 2(m-n)
The other two integrals are evaluated similarly.
1 s
Example 7.
sin Bx sin 2x
S,sin 5xsin3x dx= 2 [-cos Bx+cos 2x] dx= - 16 + -4-+C.
SEC. 15. INTEGRATION. OF CERTAIN l'RRATIONAL FUNCTIONS
BY MEANS OF TRIGONOMETRIC SUBSTITUTIONS
Let us return tothe integral considered in Sec. 12, Ch. X:

S R (x, Vax 2 +bx+c)dx. (1}
Here we shall give a method of transforming this integral 1nto
one of the form
~ R (sin z, cos z) dz, (2)
which was considered in the preceding section.
Transform the trinomial under the radical sign:
ax 2 + bx+c=a (x+ ~a)z -t ( c -~).
Change the variable, putting
~ + 2iib = t, dx = dt.
Then
Vax+ bx+c= yat 2 + (c--~)
Let us consider all possible cases.
l. Let a> 0, ,_c- :; > 0. We introd~ce the designations: a= mt,
b 1 .
c-2a=n. In this case w.e h~ve
Vax 2 +bx+c= Vmt 2+n.

b
2. Let a>O, c- 4a<O. Then
b"
c-4a= -n.
Thus,
V ax + bx +c= Vmt-n.
b
3. Let a< 0, c- 4a >0. Then
a= -mZ,
Hence,
Vax + bx+c= V n-mt.
b' ~~--~--
4. Let a< 0, c- 4a < 0. In this case V ax + bx-+- c is a com-
plex number for every val'ue of x.
In this way, integral (1) is reduced to one ol the following
types of integrals:
I. SR (t, V mt + n) dt. (3. 1)
II. SR(t, Vmt-n)dt. (3.2)
Ill. fR (t, Vn- mt) dt. (3.3)
Obviously, integral (3.1) is reduced to an integral of the form

(2) by the substitution
t = .!!:..._tan
m
z.
Integral (3.2) is reduced to the form (2) by the substitution
t =_!!__sec z.
m
Integral (3.3) is reduced to (2) by the substitution
t=_!!:_sint.
m
Example._ Compute the integral
SY(a:~x 2) 1
Solution. This is an integral of type Ill. Make the substitution x=aslnz,

then
dx =a cos z dz,
Integrals not Expressed in Terms ot Elementary Functions 385
SEC. 16. FUNCTIONS WHOSE INTEGRALS CANNOT

BE EXPRESSED IN TERMS OF ELEMENTARY FUNCTIONS
In Sec. I, Ch. X, we pointed out (without proof) that any
function f (x) continuous on the interval (a, b) has an antideriva-
tive on this interval; in other words, there exists a function F(x)
such that F' (x) = f (x). However, not every antiderivative, even
when it exists, is expressible, in final form, in terms of elemen-
tary functions.
For instance, we have already pointed out that the antideriva-
tives of binomial differentials that do not belong to the three
examined types cannot be expressed in terms of elementary func-
tions in final form (Chebyshev's theorem). Such are the antideriva-
.
t 1vec; expresse d by t he mt egra 5
Is e-xz dx, sin x
7 dx,
5 cos
-x- 5
x dx,
5V l-k sin 2 2
x dx, J1~xx and many others.
In all such cases, the antiderivative is obviously some new
function which does not reduce to a combination of a finite
number of elementary -functions.
For example, that one of the antiderivatives
~ e-x 3 dx+C,
which vanishes for x=O is called the Gauss function and is deno-
ted by <l>(x). Thus,
if
<D (0) =0.
This function has been studied in detail. Tables of its values
for various values of x have been compiled. We shall see how
this i~ done in Sec. 21, Ch. 2XVI. Figs. 204 and 205 show the
graph of the integrand y =e-x and the graph of the Gauss func-
tion y = <D (x). That one of the anti derivatives
~ Vl-k 2 sin 2 xdx+C (k< 1),
which vanishes for x=O is called an "elliptic integral" and is
13-3388
denoted by E (x),
E(x)= SV1-k 2
sin 2 xdx+C.,
if
E (0) =0.
y
-------------
-4s:-. 0
Fig. 204.
X
Fig. 205.
X
Tables of the values of this function have also been compiled

for various values of x.
Exercises on Chapter X
I. Compute the integrals: 1. Sx 5

dx. Ans. ~~+C. 2. S(x+ YXi dx. ,
Ans. ~+2x Vx+C . 3. s ( ~x-x ~x) dx. Ans. 6 Yx-.
2 3
_ _!_x 2 Y-x+C. 4. sx.:.:_. Ans. !x 2 Yx+C. 5. J(.!..+ .~-+2) dx.

10
1
X

8
Ans. - - - ,,- +2x+C.
r X
r x 5
6. sdx
X
x x r x
Ans. -
3
x'+C. v- . 4v-
z+ )z z
S(
7 X Vx1 d x
x. A ns.5+4x3 v2 ar;: C
x+ 3 vx+.
Integration by substitution: 8. Se 5
dx. Ans. ! e5" +C. 9. Scos 5x dx.
r:
"
Ans. -~~
-+C. s smaxdx. Ans. ~u
--a-+C.
s
10. 11. X dx.
5
Ans. 1 ln 2 x +C. 12. s~

sinz 3x. Ans. _cot3x +C . 13. dx?x .
cos
2 3
Ans. -tan17x
-+C. 14. s
3
dx . Ans.
x_ 7
1
3 Jnj3x-71+C. 15. s 1
dx
_x
Ans. -In 11-x I +C. 16. S dx x. Ans.- ~ In 15- 2x I +C.

5 2
17. Stan 2x dx.
Ans. - ; lnlcos2xi+C. 18. Scot(5x-7)dx. Ans. ~lnlsin(5x-7)1+C.

19. S co~~y . Ans. - ! In I cos 3y I+ C. 20. Scot 4- dx. Ans. 3 In sinI 4-l +C.
21. Stan cp -sec cp d cp. Ans. 22. S(cot ex) ex dx.
Ans. !nlsinex/+C. 23. 5(tan4S-cot~ )ds. Ans.

1
- 4' In I cos 4S 1-
I T
I
-41n sin S +C. 24. 5sm. x cos x dx.
2
3
- 3 -x +C. 25. S cos 3x sin x dx.
Ans. sin
Ans. -cos
4
--x +C. 26. s .r
r x + 1 x dx. Ans. +Y (x'+I)'+C.
27.
4
jf' ./dx . Ans. _!_y2x 2 +3+C. 28. s x'dx . Ans. -2r ,x'r+-1 +C.
29
5
r2x 2 +3
cos x dx
sin 2 x Ans.
2
_ _1_ C 30
sin x + '
s Yx'+1
sin x dx
cos' x
.4
3
_I_ C
ns. 2 cos 2 x +
31.
S
tanx dx.
cos x Ans.
tanx +C
--z- . 32.
s sinz x x. Ans. _co~x +C.
cotx d
33. s dx
cosx Y tanx-1
Ans. 2 Ytan x-1 +C. 34. sln(x+1) dx.
2
Ans. ln (x+1)+C. 35. cosxdx s
x+1
Ans. Y 2sinx+ 1 +C.
36.
J
r
2
sin 2x dx
(I+cos2x) 2
Ans.
Y2sinx+1
1
2(I+cos2x)
+C. 37 sin 2x dx
YI +sinx'
s
A ns. 2 .r1+sin
r 2 x+C. 38. Y tanx+
cos x
1d A S2
x. ns. 3 .r(tanx+1)'+C.
r
39 cos 2x dx
S(2+3sin2x)3
. 1
Ans. - - 1
12(2+3sin2x) 2
+C. 40.
s -===-
sin 3x dx
--;;3
Vcos4 3x
Ans. 1
Vcos3x +C.
41.
J
r ln 2
X dx.
x
Ans. In X+ c. 42.
3
s arc sin X dx
Y1-x 2
s

Ans.
arc sin X+ c 43
2
2
.
5 arc tan X dx
1 +x' .
A
ns.
arc tan X+ c 44
2 .
arccos X d
YI-x x.
2
Ans.
arc cos' x
3 + C. 45.
1
5
arc cot x d
+x' x. Ans.
_arc cot' x C
2 +.
s
46.
X dx
x+ Ans. 1 2
2 In(x+1)+C. 47.
X+ 1
x'+2x+ 3 dx.
5
1
1
Ans. 2 1n (x 2 +2x+3)+C. 48.
cos x dx 1 s
2 sinx+ 3 . Ans. 2 In (2 sinx + 3)+C.
49. r\ -dx
- . Ans. lnlnx+C.
.; x 1nx
50. 2x(x 2 +I) 4 dx. Ans. (x- +
2
s )5 C .
- I=
5
tan x dx 5
. .5
4
51. tan xdx. Ans. - 3--tanx+x+C. 52. ( 1 +x)arctanx'
Ans. In I arc tan xI+ C. 53.

5 ~
cos x (3 tan x + l). Ans.
I
3 1n (3 tan x+ 1) +0.
13*
54. stan' x dx.

cos 2 x
Ans. tanx +C.
4
55.
s dx
Y 1-x 2 arc sin x
Ans. lnlarcsinxJ+C. 56. s cos 2x d
2+3 sin 2x x.
Ans. i In I 2 + 3 sin 2x 1 +C.
dx
57.
Scos (In x) --;: Ans. sin (In x) +C. 58. S cos (a+ bx) dx.
X
Ans. +sin (a+!Jx) +C. 59. Sezx dx. Ans. +ezx+C. 60. Se> dx.
X
Ans. 3e 3
+C. 61. ~ e51 " x ~os x dx. Ans. e51 " x +C. 62.
x
Ans. 2~na+C. 63.
s X
eli dx. Ans. aea +C. 64.
X s (e2X) 2 dx. Ans.
1
Te
..~+C.
65. s3XeXdx. Ans. ln3;~1+C. 66. Se..;.'xdx. Ans. -!e-'X+C.
~(esx+ 1~ :+c).
5
67. S(e 5 x+a 5x)dx. Ans. 68. Sex+X+ 3 (x+2)dx.
Ans.
I
2 ex"+x+'+C. 69.
s(ax-bx)z
axbx dx. Ans.
(:
Ina-In b
r-( ~ r 2x+C.
ex dx I s ezx dx I
70. + ex. Ans. 4 1n (3+4ex)+C. 71. 2 +e'x. Ans. 2 1n (2+e9+C.
S3 4
72.
S
dx I .r-
+ x Ans. y] arc tan ( r 2x) +C.
1 2 2
73.
" dx
y 1 _ 3xz . J
A ns. I arc sin ( .r-
.r-
r 3
r 3x) +C. 74. sY dx
16-9x 2
Ans
3
. -3x+C .
-I arc sm
4
dx x dx I x
75.
Sr 9-xr 2
. Ans. arc s1n -+C.
3
76.
f
--.
4 +x 2
Ans. -arc tan -+C .
2 2
77. dx 1 3x
Sgxz+ 4 . Ans. 6 arc tan 2 +C. 78.
s4
_gxz. Ans. ~In I~+~~ +C.
dx
79. s Y :,x+9. Ans. In I x + Yx 2

+ 91 +C. 80.
5Y b~:-a'
I
Ans. - In I ax+
St. SYb::a'x'' a
+Y b'+a'x'\+C. 82. S
ax
2
-c2
~x Ans.~cln /:+~I+C.
x 2 dx
5-x 83.5
4ns. 1
.r-ln
6r5
x'+ y
x- 5
l vsr
+C. 84.
' xdx
J y 1-x' . Ans.
1
-arc slnx 2 +C.
2
x dx 1 x2 s ex dx .
85.
5- - . Ans. -arc
x+a 2a
tan-+ C. 86.
a' Yi-e 2x
Ans. arc sin ex+ C.
Exercises on Cllapter X 389
~ x +C. cos xdx

Ans. ,)-arc sin -.. /
, 5 v 3
88.
S u +sin'~
2
Ans. 2..arctan(sinx)+C. 89.

a a
s x._
ydx
1-lnx
. Ans. arcsln(lnxi+C,
90.
MCCOSX-X
y d
x. Ans.
I
--(arc cosx) 2 + r l-x 2 +C.
,r--
S 1-x 2
91. 5x-arc tan x d

l+x x. Ans.
I 2 I
2 1n (I +x ) - 2 (arc tan x) +C.
2
~ ~lnx
~
2 VI+Yx
92. Yl dx. Ans. - Yll+lnx)'+C. 93. dJC.
3 Vx
Ans. : Jl (I+ Y x)'+C. 94. dx J
. Ans. 4 Vl+ V x+G.
95.
5 e"+ dxe :x. Ans. arc tan eX +C.
VxVt+Vx
96. sVCOS X dx . Ans. 3 v-- slnx+C.
97. .r
1
2
Sr 1 + 3 cos x sin 2x dx.

2 .r
Ans. - - r (I+ 3cos 2 x) 3 +C. 98.
9
sin 2 x
sV sin 2x dx
I +cosx
,
99.
Vtanx 3
100;
S COS 2 X
dx. Ans. - Vtan'x+C.
5
101. 5 dx
2 sin"x+3 cosx
Ans

) - arc tan ( --./- tan x) +C.
r 6 V 3
2
Integrals of the form 5ax~~ 't ~c dx:
102.
5x+dxx+S' 2 Ans. 2
I x+I
arctan-r+C. 103.
53x-dx2x+ 4
I 3x-l
Ans. ,r-arctan .r- +C. 104.
dx I
"+ +I" Ans.,r_ln V 5
r2x+3- V51
I+C.
r II r II x 3x r 5 2x+3+ 5
dx ,x-5,-
105.
5
x 2 -6x +5
Ans. I
41n x-1 +C. 106
2 +I 52z ~~'
Ans. arctan(2z-IJ+C. 107.
I
5
. 3x-l
3x-d;x+ 2 . Ans. .r- arc tan ,,r- + C. , 5 , ~
(6x-7)dx , (3x-2) dx
108.
S 3xz-?x+ll. Ans.
1nl 3x -7x+lli+C. 109.
55x 2 -:- 3x + 2 '
~ II - IOx-3 3x-l
Ana. - In
10
(5x 2
-3x+2l- .r-
5 , 31
arc tan .r-
+C.
, 31
JIO.
Sx'-x+l dx.
. 3 I 2x- t 7i+l
Ans. -2 In (x 2 -x + 11+ ,,r-are
3
tan ,/'-+C.
, 3
111.
56x"+ x-1 dx.
390 I ndefin'ite I nfegrals
2 I 2x-1
Ans. 112.
3 1n (3x-l)+T In (2x+ I)+C. S5x 2- x+2 dx.
1
Ans. - In (5x 2 -x + 2)
5
8
.r-
5 r 39
+ lOx -I
arc tan .r- +C. 113.
r 39
6x 4 -5x'+4x 2
2x2 -x + 1 dx.
s
x 1 I 2 I I t 4x- I C
A ns. x - 2 + 4 nj 2x -x+ 1+ y""forc an V'f + .
2
dx 2 2 tan x +I
114.
S 2 cos x +sin x cos x + sin x 2 Ans. Y'f arc tan Y
7
+ C.
Integrals of the form s Y Ax+

ax'+ bx+C
8
dx:
115.
S
dx
y 2 _ x- x 2 Ans.
3 4
I
2 arcsln
8x+3
+C. 116. y4T s dx
Yl+x+x'
1
Ans. lnJx+- +Yx+x+lJ+c. 117. Sv dS 2 .
2
Ans. lniS+a+
2aS+S
+ Y2aS+S 2
I+C. 118. s Y dx
5-7x-3x'
. Ans . .)_arcsin~~+?+
r 3 r 109
C.
119. Svx1:X+s) Ans. ; Jnj6x+5+Jfl2x(3x+5)1+C.

3
120.
JI y 2-3x-x
dx . Ans. arc sin
2
:r~ +C.
r 17
121. sy 5x 2 -x-l
dx ,
2
Ans . .)_ ln(!Ox-i+Jf20(5x"-x-I)+C. 122. l ax+b dx.
r 5 .) V ax" + bx + C
Ans. ! Y 4x +4x+3+ 2
(x-3)dx
I .r s
124.
S
Jf3+66x-llx 1 '
Ans. - - r 3+66x-llx"+C.125. Y (x+3) dx . Ans.- 1 .r.
r 3+4x-4x"+
4
+- 7
4
11
2x-l
arcsln--+C. 126.
2
s Y
3x+5
3+4x-4x"
x(2x-l)
dx. Ans.
3 . r -- 23
-2 r 2x 2 -x+ .r-X
4 r 2
X In (4x-l + Jf8 (2x 2 -x) )+C.
II. Integration by parts:

127. ~xe~dx. Ans. ex(x-l)+C. 128. ~xlnxdx. Ans. ;xx
x(lnx-{)+c. 129. S xslnxdx. Ans. slnx-xcosx+C. 130. Slnxdx.
Ans. x (lnx-1)+ C. 131. ~arc slnx dx. Ans. x arc sinx+ Y1-x 2 +C
132. ~ In (1-x) dx. Ans. -x-(1-x)ln(l-x)+C. 133. ~xnlnxdx.
Ans. ~n:l ( lnx-n~l)+c. 134. S xarctanxdx. Ans. {l(x2 +1)X
xarctanx-x]+C. 135. Sxarcsinxdx. Ans. +[(2x 2 -l)arcsinx+
+x Yl-x'J+C. 136. Sln(x'+l)dx. Ans. xln(x+l)-2x+2arctanx+C.
137. S arc tan Vx dx. Ans. (x +I) arc tan Vx- Vx +C.
138.
J
arc sin
Y x
Yx dx. Ans. 2r,,-xarcsinr,,-x+2r,,-
l-x+C.139.
s .V/-x
arcsin x+l d,
Ans. xarc sin y x~ 1

- Vx +arc tan Vx +C. 140.
x I
Ans. -+-xsin2x+- I cos2x+C. 141. sxarcsinx
,, dx. Ans. x-r.~ 1-xx
4 4 8 r 1-x
x arc tan x d A x I t
sS ,,--
xarc sinx+ C. 142. (x'+ I)' x. ns. 4 (I +x') +4 arc an x-
1 arctanx+C . 143. xarctan r x"-l dx. Ans. -I x 2 arctan , ,-
r x'-l-
--
2 l+x2 2
- 2I ,r ,x"-I+C.
--
144. sarc
-x- sin x
2-dx. Arts. In 11- VI
2
x -x 1-x-arcsinx+O.
I
145. Sln(x+YI+x')dx. Ans. xlnlx+YI+x"I-YI+x"+C
146.
S arc smx
. V (Ixdx
-x)
. A
ns. arc sin x +..!..In ~1-x
Yl-x 2 I +x
I
Use trigonometric substitutions in the following examples:
Va-x Va-x x
147.
S -dx. Ans. -
..:._----..
X 2
X I ,,-- I
X
,,--
arc sin -+C.
a
Ans. 2arc sin
2 - 2 x y 4-x+ 4 x r 4 -x"+C.
Ans. -
Yl+x +C.
X
150.
s Vx-a dx.
X
Ans. ,
r, --
x-a-aarc

a
cos- +C.
X
dx x I
151. V . Ans. 2 Y
S ("a'+x') 1 a a+x'
+C.
Integration of rational fractions:

I (x+3)' s x+x-8 x x
Ans. 8 1n (x+S)'(x+ I). 154. x- 4x dx. Ans. T+T+4x +
+In
x 2 (x-2)'1
l(x + 2>. +C. 155.
s xdx
(x'-l)(x+ 21 . Ans.
x
2 -2x+crlnx
1
(x-1) 16
X<x+Il"+ 3 1n(x+2l+C. 156.
s dx
(x- 1)'(x- 2). Ans.
1
x- 1 +
x-2
+ 1n - + C.
X- 1
157. s 1 -
x-8 +
4X 2 4X
dx. Ans. __2_2+ln (x 22)'+ C.
158.. Sx(x+t>dx.
3x + 2
X
4x + 3 x
Ans. 2 (x+ w+ln (x+I)'+C.
2
X-
1ii9.
s x dx
X
(x+ 2l'(x+ 4l".

5x + I 2 n ( x + 4)
2
A n~. _ C s dx A x C
x'+6x+B+ 1 x+2 + . 160. x(x+1)' ns. 1nYx'+I+ .
2
2x 2 -3x-3 2
t x-! +C
Ans. In (x -2x+5) + I
161.
S (x-I) (x 2 -2x+5) dx. x- 1 2 nrc anT .
x'-6 x +4
2
3 x 3 x
162. x'+ 6x'+B dx. Ans. In Yx'+ + 2 arctan 2 - V_ arctan Y +C.
2 2 2
dx I (x + I )2 1 2x -I
163. x + 1 . 6 In x 2 - x + 1 + y 3 arc tan y"3 +C.
S Ans.
164.
S
3x-7 2
x +4 I x
x+x'+ 4x+ 4 . Ans. ln(x+l)+'2arctan 2 +C. 165.
4dx
x'+l.
s
Ans. I
.r-ln
,.. 2
x'+x Y2 +I
x-x
Y
2 +I
.r- x Y2
+r2arctan---.+C.
1 -x
166.
s x
x-tdx.
I
1)I +
A ns. 1 [x + In (x- c.
. 67. sx+x-!
(x'+ ) dx. Ans. 2-x + 2 2
3 4 (x'+ 2) .In (x +2) -
I x
- 4 Y2arctan V'2+C. 168.
s 2
(4x 2 -8x) dx
(x-l)'(x 2 +I)'. Ans.
3x 2 - l
(x- 1)(x+ 1l +
(x- 1) 2
+In--:--+
X 1
+arc tan x+C. 169. (X 2 - X
dx
)( 2
X -X
+
s l)l' Ans. In-
X
x-1
--
10 2x-I 2x-1
- .r-are tan .r- - 3 ( 2 )+C.
3 I' 3 I' 3 X -X-- 11
Integration of irrational functions:
V v-x
170.
~ x +I
dx. Ans.
171.
~
vxa-VxV - dx. Ans. 272 V4-.x -
X 212 1"i
13 vx +C. 172.
~;
V.v-x+J
V dx.
6 x1 + x
6 12
Atis. -,v-;+;y;+2Inx-241n ('2/- )
V x+! -f-C.
175. Sy~ +:d: . Ans. 2arctan y'ri-x

1- +x +In
Vi +x_:_ Vi-x
V l+x+ Y 1-x +C.
;/-; + vx- [14/- , 1 14/-, 1
J V
1/- 7 /-;:
176. 1 14 dx. Ans. 14 v x - 2 v x +3' v x - 4 v x +
15
{/ x'+ X
+s-vx I 14Fsj
+. C 177. f ,V/2+3x d
x- 3 x.
A
ns. .r 2
7 - 6 + 11
r 3x - x V 2 3
x
xln( x-~+ yx-~x-2 )+c.

Integrals of the form ~ R (x,
178. s Yx~.:_x+3.
x Ans.
179 f y dx . Ans.
.)x 2+x-x2
180. dx . Ans.
Sx Y x'+ 4x-4
Ans. Vx+2x+1nlx +I+ vx+2xi+C. 182. s dx
Y(2x-x 2 ) 1
s V2x-x dx.

Ans. x- I +C. 183. 2

Ans. -1 [(x-1)
. V2x-x 2 +
Y2x-x 2
+arcsln(x-l)J+C. 184. Sx- Vx'-1

dx

Ans. x
2
+~2 Yx 2 1-
1
- - 1n)x+rx 1 -1i+C.
2
.r--
185. j' (I +x)
ydx
I+ x+x
.
Ans. In
I x + VI + x + x
2 +x+ Yl+x+x +C.
I 186.
s (X+ I)
(2x+x) Y2x+x dx.
Ans. _ .r 1 +C .18'7. \i-VI+x+xd

.r x. Ans. 1n /2+x-2VI+x+xj+c
r 2x +x' x r I+ x+x x2
V x + 4x
y8
188.
S x
z ax. Ans. -
x+ x 2 +4x
+Inlx+2+ Vx 2 +4x!+C.

Integration of binomial differentials:
1 2 1
189. ~ x3 (2 +xa)4 dx.
Ans. 10x'l~l6(2+x-iy~- +C. 191. s dx Ans. X

Vt +x'
+C.
(I+ xZ) 2
192. S dx .!...Ans.-(l+x)-+(2x+~)+C.193JVC+x+)'dx.
2
x 2
(1 +xZ)
7
Ans. ~(7 Vx -4)(1+ JI'X) 4 +C. 194.
Ans.
2 (4 + 3 V -x) (2-
5
v- x)
195.
5
Ans.
5
x:; 3 (I+ x 0 1
)
Integration of trigonometric functions!
196. S stnx dx. Ans .. ~ cosx-cosx+C. 197. Ssln xdx. Ans.--cosx+

5
2
+acosx--
5
x
cos -+C.
5
198. s cosxsln 1 x. Ans. - 51 cos 5 x +1 cos7 x +
7
c.
cosxd
199
S sin x x.
Ans. esc x-
3
I
esc x+C. 200. Scos x dx.
X I 3 sin 2x +sin 4x + C.
Ans.
2 + 4 sin2x+C. 201.
SsnI x dx.
4 A ns.
8 x - -4- 32
202. S cos x dx. Ans. j] . 2x -sin
I ( 5X + 4 Stn -- 2x+ 3 SIO
. 4X ) +C ,
3 4
203. Sstnxcosxdx. Ans.
1 ;8 (3x-sln4x+si~Bx)+c. 204. Stan 3 xdx.
Ans. tan
- 2- x +In I cos xI +C. 205.
s cot x dx. Ans. - I cot x+ I cot' x +
4 2
cot x
+In I sin x I+C. Ans. - ---ln I slnx I +C.
206.
Scot' x dx. 2
7
207. sec 8 x dx. Ans. tan x+3tanx+t an 3 x +t an x +C .
S 7 5
Exercise.~ on Chapter X 395
208. Stan x sec x dx. Ans.

7
ta~ x +tan; x +C. 209. s dx
cos x
Ans. tanx+
I 3
3 tan x+C. 210. s COS X
sin' x
d
x. Ans. C-csc x.
5 1
211.
sin 3 x d!C
. Ans.
3
cos
s x+3 cos
-a +C. 212. Ssin x sin 3x d:c.
SV cos x 5
Ans.
sin 4x
- -- + --+C.
8 4
sin 2x 213.
s cos4xcos7xdx. Ans. sin 11x+sin3x+a
22 6
cos 6x cos 2x 1 3
214.
Scos2xsin4xdx. Ans. -----rr-- 4
-+C. 215.
Ssin 4 x cos 4 xd:c.
Ans.
cosx 1
- -2-+ cos 2 x+C. 216.
s dx
4 _ 5 sinx. Ans.
1 tan2- 2
3 tn 2tan ~-1 +C.
X
2 L
217.
dx
S5-3 cos x Ans.
1
- arc tan 12 tan !!_ 1+C.
2 2
218. Sl+sinx
sin x d:c ,
2 cos 1C dx x
Ans.
X
+x+C. 219.
S1+ COS X
.Ans. x-tan - +C.
. 2
1 +tan
2
S(l +~~s x) 2
'
Ssln 2 x~tan'x Ans.

1
-{ [cotx+
sln x
223.
S1 + dx. Ans.
COS 2 X
V2 arc tan X
CHAPTER J{l
THE DEFINITE INTEGRAL

\
SEC. 1. STATEMENT OF THE PROBLEM. THE LOWER

AND UPPER INTEGRAL SUMS
The definite integral is one of the basic concepts of mathemat-

ical analysis and is a powerful research tool in mathematics,
physics, mechanics, and other disciplines. Calculation of areas
bounded by curves, of arc lengths, volumes, work, velocity, path
length, moments of inertia, and so forth reduce to the evaluation-
of a definite integral.
Fig. 206. Fig. 207.
Let a continuous function y = f (x) be given on the interval

[a, b] (Figs. 206 and 207). Denote by m and M its smallest and
largest values on this interval. Divide the interval [a, b] into n
subintervals by points of division:
so that
x 0 <x, <x 2 < .. .<xn,
and put .
.x 1 -X 0 = 6-x,; X2 -X 1 = 6.x 2 , , Xn-xn_ 1 = 6-xn.
Then denote the smallest and greatest values of the tunction f (x)
on the interval [x 0 , x,] by m, and M,
on the interval [xP X2 ] by m 2 and M 2
Form the sums

(I)
Statement of the Problem. The Lower and Upper Integral Sums 397
- n
s, = M,~x, + M 2 ~X 2 + ... + Mn~Xn = ~ Mi~Xi. (2)
1=1
The s urn sn is called the lower (integral) sum, and the sum s,.
is called the-upper (integral) sum.
If f (x) ~ o, then the lower sum is numerically equal to the area
of an "inscribed step-like figure" AC 0 N,C,N 2 . : . C,_,NnBA bound-
ed by an "inscribed" broken line, the upper sum is equal numer-
ically to the area of an "circumscribed step-like figure"
AK.C,K, ... Cn_,Kn-,CnBA bounded by an "circumscribed" bro-
ken line.
The following are some properties of upper and lower sums.
a) Since mi ~ Mi for any i (i,;, 1, 2, ... ,. n), by formulas (1)
and (2) we have
(The equal sign occurs only when f (x) = const.)

b) Since
m, ~m. m.~m ... , m,~m.
where m is the smallest value of f (x) on [a, b]. we have

S11 = m, ~X 1 + tn 2 ~X 2 + ... + m,.~xn ~ m~x, + m!!:.x + ... i m~xn =
2
- =m (~x, + ~X 2 + ... + ~x,.) =m (b-a).

Thus,
c) Since sn~m(b-a).
M 1 ~M, M 2 ~M, ... , Mn~M,
where M is the greatest value of f (x) on [a, b), we have
S,. M,~x, +M.~x.+ ... + Mn~Xn ~M~x, + M~X 2 + ...
. . . + ~~Xn= M (~X 1 + ~X 2 + ... + ~xn)= M (b--a).
Thus,
5,; ~M (b-an).
Combining the inequalities
obtetined,' we have
m (b-a) ~sn ~sn ~M (b-a).
If f (x) ~ ((then the latter in-
equality has a simple geometric
meaning (Fig. 208), because the
products m (b -a) and M (b -a) m
are, respectively, numerically
equal to the areasofthe"inscribed"
rectangle AL,L.B and the "cir-
cumscribed" rectangle AL,L~B. Fig. 208.
398 The Definite Integral
SEC. 2. THE DEfiNITE INTEGRAL
Let us continue examining the question of the preceding sec-

tion. In each of the intervals [x., x,]. [x 11 x.], ... , [xn_ 1 , xn]
s
take a point and denote them by 11 ., ... , Sn (Fig. 209):
x.<s,<x" x'~<s.<x., ... , xn_, <sn<xn.
At each of these points find the value of the function f (,),
f (.), ... , f (sn) Form a sum:
n
sn = f (,) L'lx, + f (.)Ax.+ ... +f (sn) flxn = ~ f (;) L'lx;.
&.:.1
(1)
This sum is called the integral sum of the function f (x) on the
interval [a, b ]. Since for an arbitrary s; belonging to the interval
[x;_,, xJ we will have
m; .;;;;.f (s;),;;;;;, M1
and all Axi >0, it follows that
m;AX; .;;;;.f (~;)Ax;,;;;;;, M; Ax;,
and consequently
n n n
tt~ mi Axi .;;;;,i~~(si) Ax; 4~MiAxi,
or
(2)
The geometric meaning of the latter inequality for f(x);;;;;:O con-
sists in the fact that the figure whose area is equal to sn is
y:,.r(x) bounded by a broken line
y '-1----='~n lying between the "inscribed
broken line and the "circum-
scribed broken line.
The sum sn depends upon
the way in which the interval
[a, b] is divided into the sub-
--,0;t---'-::-t---'--!--'----4--.l....:-- x inte rv a Is [xi_ 11 xi] and a Iso
~2 X2 Xn-t ~n Xn-b upon the choice of points 1
Fig. 209. inside the resulting subinter-
vals.
Let us now denote by max [x;_ 1 , x;] the largest of the lengths
of subintervals [x0 , x,], [x 1 , x.], ... , [xn-P xn]. Let us consider
different partitions of the interval (a, b] into subintervals
[x;-p x;] such that max {xi-1' x;] -0- Obviously, the number
of subintervals n approaches infinity here. Choosing the appro-
The Definite Integral 393
priate values of Si it is possible, for each partition, to form the

integral sum
We can thus speak of a sequence of partitions and a correspond-

ing sequence of integral sums. Let this sum*, approach the
limit I for some chosen sequence of partitions when max
11x; ---+- 0.
If for any partitions of the interval la, b] such that max
n
11xi-+0 and for any choice of points 1 the sum ~f(si) 11x1
approaches the same limit I, we say that the function f (x) is
integrable on the interval [a, b]; the limit I is called the defi-
nite integral of' the function f (x) on the interval [a, b]. It is
b
denoted by Sf (x) dx and we write
a
b
max
lim
~x,
~ f(si) 11x1 =Sf
.... o =I a
(x) dx.
The number a is called the lower limit of the integral, b is the

upper limit. The interval [a, b] is called the interval of inte-
gration, the letter x is the variable of integration.
Let it be stated without proof that if a function y = f (x)
is continuous on the interval [a, b), then it is integrable on this
interval.
It is obvious that if for some sequence of partitions such that
max 11xi-+ 0 we consider the sequence of lower integral sums sn
and of upper integral sums 8,; for a continuous function f (x)."
then these sums will" tend towards the same limit I -the defi-
nite integral of the function f (x):
b
lim ~: mi 11xi = S{(x) dx,

max ~x 1 .... o f:11 a
max~x,
lim ~
.... o=l
M 1 11x; =Sf (x) dx.
a
Among discontinuous functions there are both integrable func-

tions and nonintegrable ones.
*) In this case the sum Is an ordered variable quantity.

If we construct the graph of the integrand y=J (x), then in

the case of f (x);;:;.: 0 the integral
b
5f (x) dx
a
will be numerically equal to the area of a so-called curvilinear

trapezoid bounded by the given curve, the straight lines x =a and
x= b, and the x-axis (Fig. 210).
For this reason, if it is required to compute the area of a cur-
vilinear trapezoid bounded by the curve y == f (x), the straight
lines x =a and x = b, and the x-axis, this
area Q is computed by means of the inte-
gral
b
Q= 5f(x) dx. (3)

a
X Note 1. It will be noted that the definite

Fig. 210. integral depends only on the form of the
function f (x) and the limits of integration,
and not on the variable of integration, which may be denoted by
any letter. Thus, without changing the magnitude of a definite
integral it is possible to replace the latter x by any other letter:
b b b
5f
a
(x) dx 5
= f (t) dt = ... = f
a
5
a
(z) dz.
When introducing the concept of the definite integral 5f

a
(x) dx
we assumed that a< b. In the case where b <a. we will, by

definition, have
b a
5f (x) dx = - Sf (x) dx. (4)
a b
Thus, for instance,

0 5
Sxdx = - 5 1(
2
dx.
0
Finally, in the case of a= b we assume, by definition, that for

. The Definite Integral 401
any function f (x) we have

a
5f(x) dx=O. (5}
a
This is natural also from the geometric standpoint. Indeed,

the base of a curvi 1in ear trapezoid has a length equal to zero;
consequently, its area is zero too.
y
b
Ex!lmple I. Compute the integral 5kx dx (b >a).
a
Solution. Geometrically, the problem is equiva-
lent to computing the area Q of a trapezoid
'bounded by the lines y=kx, x=a, x=b, y=O
(Fig. 211).
The function y = kx under the integral sign is
continuous. Therefore, in order to compute the
definite integral we have the right, as was stated X.
above, to divide the interval [a, b] in any way
and choose arbitrary intermediate points ~k- The Fig. 21 I.
result of computing a definite integral is independ-
ent of the wa.y in which the integral sum is formed, provided that the
subinterval approaches zero.
Divide the interval [a, b) into n equal subintervals.
The length .1,x of each subinterval is .1,x = b -a; this number is t)le
n
subinterval (partition unit). The division points have coordinates:
a=x 0 , x1 =a+.1,x,
x1 =a+2.1,x, . , Xn=a+n.1,x.
For the points Sk take the left end points of each subinterval:
1 =a, s =a+.1.x, s =a+2.1,x,
2 3 .... sn=a+(n-1).1-x.
Form the integral sum (1). Since f(~il=k6io we have
.ks2.1.x + ... + ksn.1.x =
sn = ks 1 .1.x +
=ka M+[k (a+M)J M+ ... +~k )a+(n-1) .1-xl} M=
=k ~a+(a+.1.x)+(a+2.1.x)+ ... +[a+(n-1) .1-xl} .1-X=
=k ~na+[.1.x+2.1.x+ ... +<n-1) ~xl} .1,x=
=k {na+)l +2+ ... +<n-1)] .1-x~ .1-x,
where .1,x=b-a. Taking into account that
n
n(n-1)
1+2+---+(n-1)
2
(as the sum of an arithmetic progression),
Sn=k [ na+
n(n-I)b-a] b-a
2 -n- --n-=k
l n-Ib-a]
u+-n---- (b-a).
2
Since lim n - l = I, we have

n-+OD n
lim sn = Q = k [ a+ b- a] (b- a) = k b'- a'
n-. oo 2 2
Thus,
b
b'- a'
5
a
kxdx=k-- .
2
The area of ABba (Fig. 211) is readily computed by the methods of ele-
mentary geometry. The result will be the same.
b
Example 2. Evaluate Sx' dx.
0
Solution. The given integral Is equal to the area Q of a
curvilinear trapezoid bounded by a parabola y=x', the
ordinate x=b, and the straight line y=O (Fig. 212).
Divide the interval [a, b) into n equal parts by the
points
b
I x 0 =0, x 1 =~x. x 2 =2~x ... ,
For the \;{ points take the right extremities of each subin-
terval.
Xn=b=n~x. ~x=-.
n
I Form the integral sum
0 XfX2 X sn =x:M+x~ M+ ... +x~ M= [(~x)' ~x+

Fig. 212. + (2M) 2 ~x+ +(nM)' ~x]=(~x)'[I'+2'+ +n'].
As we know,
P+ 2,+ 3'+ ... +n'= n (n+ I~ (2n+ 1),
therefore
5
n
=!Cn(n+1).(2n+l)=~(
n' 6 6
!+.!...)
n
( 2+..!..)
n '
h
lim Sn= Q=Sx'dx=

n-. oo
~
0
b
Example 3. Evaluate S m dx (m = const).
a
Solution.
=m lim L ~x;=m(b-a).
max D.xt-. o l=l
The Definite Integral 403
n
Here, ~ ;:..x; is the sum of the lengths of the subintervals into which
l==l
the interval [a, b) was divided. No matter what the method of partition,
the sum is equal to the length of the segment b-a.
b
Example 4. Evaluate ~ e"dx.

a
Solution .. Again divide the interval [a, b] into n equal parts!
b-a
x 0 =a, x 1 = a+;:..x, ... , Xn=a+ntJ.x; tJ.x= -n-.
Take the left extremities as the points Si Then form the sum
Sn=ea;:..x+ea+t.x;:..x+ +ea+(n-t)t.xf:..x=
=ea (1 +et.x +ezt.x + ... + e<n-t) t.x) ;:..x.
The expression hi the brackets is a geometric pr.ogression with .common
ratio .et.x and first term I; therefore
ent.x_l tJ.x
sn '= ea ;:..x = e0 (ent.x- I) - - -
et.x_1 et.x_l
Then we have
;:..x
ntJ.x=b-a; l1m -,..-x--=1.
t.x ..... o e -1
z I
(By L'~{ospital's rule lim - 3 - = llmz-=1.) Thus,
z ..... oe-1 z ..... oe
lim Sn= Q =e0 (eb- 0 -1)1 =e0 -e0 ,
n ..... c:n
that is,
Note 2. The foregoing examples show that the direct evaluaa

tion of definite integrals as the limits of integral sums involves
great difficulties. Even when the integrands are very simple (kx,
x. e"), this method involves cumbersome computations. The find
ing of definite integrals of more complicated functions leads to
still greater difficulties. The natural problem that arises is to
find some practically convenient way of evaluating definite inte-
grflls. This method, which was discovered by Newton and Leibrtiz,
utilises the profound relationship that exists between integration
and differentiation. The following sections of this chapter are
devoted to the exposition and substantiation of this method.
404 The Definite Integra(
SEC. 3. BASIC PROPERTIES OF THE DEFINITE INTEGRAL
Property 1. The constant factor may be taken outside the

sign of the definite integral: if A = cons t, then
b b
~ Af (x) dx =A) f (x) dx. (l)
a a
Proof.
b
~ Af (x) dx ~ lim ~ Af (~;) l'lx; =
a max .1-x-+o i=l
b
=A lim
max~x-+-o
t
i=J
f (~;) l'lx; =A ~
a
f (x) dx.
Property 2. The definite integral of an algebraic 5Um of several

functions is equal to the algebraic sum of the integrals of the sum-
mands. Thus, in the case of two terms
b b b
~ [{ 1 (x)+f2 (x)] dx= J{1(x)dx + ~ f 2 (X)dx. (2)

a a a
Proof.
b
~ l/ 1 (x) +/ 1 (x)] dx = lim ~ [f (~;)+f.(~;)] l'lxi =

1
u max.1.X-+O i=l
n 11
lim
max.1.x-+o i=l
l ~ f1 (1) l'lx; + ~ f2 (~;) l'lx;l =
i=t
n n
= lim ~f~(~;)l'lx;+ lim ~f 2 (~;)1'lx;=
max.1.x-+O i=l max,1.x-.o i=l
b b
= ~ {1 (x) dx +) { (x) dx. 2

a u
The proof is similar for any number of terms.

Properties I and 2, though proved on) y for the case a< b, hold
also for a;;;;;a. b.
However, the following property holds only for a< b:
Property 3. If on the interval la, b) (a< b), the functions f(x)
and q> (x) satisfy the condition f (x) ~ q> (x), then
b b
~ f (x) dx ~ ~ q> (x) dx. (3)

a a
Basic Properties .of the Definile Integral 405
Proof. Let us consider the difference

b b b
a
~ cp (x) dx- ~ f (x) dx =
a
J[cp (x)-/ (x)l dx =
a
n
lim ~ (cp (s;)-f (;))~xi.
max6.x-+o j;t
Here, each difference cp(s;)-f(;);;:;;;.0, ~x;;;;;;.O. Thus, each term

of the sum is nonnegative, the entire sum is nonnegative, and its
limit is nonnegative; that is,
b
~liP (x)-/ (x)] dx;;;;;;. 0

a
or
b b
) cp (x) dx- Sf (x) dx;;;;;;. 0,

a a
Fig. 213.
whence follows inequality (3).
If /(x)>O and cp(x)>O, then this property is nicely illustrated
geometrically (Fig. 213). Since cp (x);;;;;;. f (x), the area of the curvi-
linear trapezoid aA 1 B 1 b does not exceed the area of the curvilinear
trapezoid aA 2 B 1 b.
Property 4. If m and M are the smallest and greatest values
of the function f(x) on the interval [a, b] and a~ b, then
b
m (b -a)~) f (x) dx ~ M (b-a), (4)

a
Proof. It is given that
m~f(x)~M.
On the basis of property (3) we have

'b b b
~ m dx ~ Sf (x) dx ~ S M di. (4')

a a a
But
b b
Smdx=m(b-a), )Mdx=M(b-a)
a a
(see Example 3, Sec. 2, Ch. XI). Putting these expressions into

inequality (4'), we get inequality (4).
406 TIre Definite Integral
If f (x) ~ 0, this property is clearly

illustrated geometrically (Fig. 214). The
area of the curvilinear trapezoid aABb
lies between the areas of the rectangles
aA,B,b and aA.B.b.
Property 5. (Mean-value theorem).
If a function f (x) is continuous on the
interval [a, b), then there is a point
on this interval such that the following
equality holds:
:o X b
Fig. 214. ~ f (x) dx = (b-a) f (). (5)

a
Proof. For definiteness let a< b. If m
and M are, respectively, the smallest and greatest values of f (x)
on [a, b], then by virtue of (4)
b
m~b~a St(x)dx~M.
a
Whence
b
b
1
a 5f (x) dx = 11 where m ~ 11 ~ M.
a
Since f (x) is continuous, it takes on all intermediate values

between m and M. Therefore, for some value (a~~ b) we will
have 11 = f (), or
b
) f (x) dx = f () (b -a).
a
Property 6, For any three numbers a, b, c the equality

b c b
~ f (x) dx = ~ f (x) dx + ~ f (x) dx, (6)

a ti
is true, provided all these three integrals exist.

Proof. First suppose that a< c <b, and form the integral sum
of the function f (x) on the interval [a, b].
Since the limit of the integral sum is independent of the way
in which the interval [a, b] is divided into subintervals, we shall
divide [a, b] into subintervals such that the point c is the division
b
point. Then we partition the sum ~. which corresponds to the
a
Evaluating a Definite Integral 407
c
interval [a, b], into two sums: ~. which corresponds to [a, c], and
a
b
~. which corresponds to [c, b]. Then
c
Now, passing to the limit as max ~x;-+ 0, we get relation (6).

If a< b < c, then on the basis of what has been proved we can
write
c b c b c c
~f (x) dx = ) f (x) dx + ~I (x) dx or ) f (x) dx = ~ f (x) dx- Sf (x) dx;
a a b a .a b
but by formula (4), Sec. 2, we have

c b
~f(x)dx=- ~ f(x)dx.
b c
Therefore,
b c b
~ f (x) dx = ~ f (x) dx + ) f (x) dx.

a a c
This property is' similarly proved for

any other arrangement of points a, b,
and c.
Fig. 215 illustrates, geometrically, Fig. 215.
Property 6 fpr the case when f (x) > 0
and a< c < b: the area of the trapezoid aABb is equal to the
sum of the areas of the trapezoids aACc and- cCBb.
SEC. 4. EVALUATING A DEFINITE INTEGRAL.

NEWTON-LEIBNIZ FORMULA
In a definite integral
b
St(x)dx
a
let the lower limit a be fixed, and let the upper limit b vary.
Then the value of the integral will vary as welll that is, the
integral is a function of the upper limit.
So as to retain customary notations, we shall denote the upper
limit by x, and to avoid confusion we shall denote the variable
of integration by t. (This change in notation does not change t.he

X
value of the integral.) We get the integral ) f (t) dt. For constant a,
a
this integral will be a function of the upper limit x. We denote
this function by <D (x):
X
<D (x) = ~ f (t) dt. (l)

a
If f (t) is a nonnegative function, the quantity a> (x) is numeri-

cally equal to the area of the curvilinear trapezoid aAXx (Fig. 216).
It is obvious that this area varies with x.
Let us find the derivative of <D (x) with respect to x, or the
derivative of the definite integral (l) with respect to the upper
limit.
Theorem l. If f (x) is a continuous
X
function and <D (x)--: ~ f (t) dt, then we

a
have the equality
<D' (x) = f (x).
In other words, the derivative of a
definite integral with respect to the
upper limit is equal to the integrand,
in which the value of the upper limit replaces the variable of
integration (provided that the integrand is continuous).
Proof. Let us give the argument x a positive or negative incre-
ment ~x; then (taking into account Property 6 of a definite integral)
we get
x+&x x x+&x
<D(x+~x)=) f(t)dt=~f(t)dt+) f(t)dt.
a a x
The increment of the function <D (x) is equal to

x x+&x x
~<D = (x + ~x) - (x) = ~ f (t) dt + ~ f (t) dt- ~ f (t) dt;
a x u
that is,
x+&x
~<D= ~ f (t) dt.
X
Apply to the latter integral the mean-value theorem (Property 5

of a definite integral):
~= f m(x +~X-X)= f ()Ax,
where s
lies between x and x+Ax.
Find the ratio of the increment of the function to the increment
of the argument:
Hence,
<D' (x) = lim ~~ = lim f ().
6K-+0 ~X .lx-+0
But since -x as Ax-0, we have

lim f (') = lim f (),
llx -+O ~-+-X
and due to the continuity of the function f (x),

lim f () = f (x).
~-x
Thus, <D' (x) = f (x), and the theorem is proved.

The geometric illustration of this theorem (Fig. 216) is simplel
the increment MD= f () ~x is equal to the area of a curvilinear
trapezoid with base ~x. and the derivative <D' (x) = f (x) is equal
to the length of the interval xX.
Note. One consequence of the theorem that has been proved 1s
that every continuous function has an antiderivative. Indeed, if
the function f (t) is continuous on the interval [a, x], then as was
.
pointed out in Sec. 2, Ch. XI, in this case the definite integral
~ f (t) dt exists, which is to say that the following function exists:
a
X
<1> (x) = ~ f (t) dt.

a
But from what has already been proved, it is the antiderivative

of f (x). .
Theorem 2 If F (x) is some antiderivative of the continuous
function f (.x.), then the formula
b
~ f (x) dx = F (b)-F (a) (2)

a
holds,
This formula is known as the Newton-Leibniz formula.*)

Proof. Let F (x-) be some antiderivative of the function f (x). By
X
Theorem I, the function ) f (t) dt is also an antiderivative of f (x).

a
But any two antiderivatives of a given function differ by the con-
stant C*. And so we can write
X
) f (t) dt = F (x) + C*. (3)

a
Within an appropriate choice of C* this equality holds for all

values of x, that is, it is an identity. To determine the constant C*
put x =a in the identity; then
a
)f(t)dt=F(a)+C*,
a
or
O=F(a)+C*,
whence
C*=-F(a).
Hence,
X
~ f (t) dt = F(x)-F (a).

a
Putting X= b, we obtain the Newton-Leibniz formula:

b .
~ f(t)dt=F(b)-F(a),
a
or, replacing the notation of the variable of integration by x,

b
Jf (x) dx=F(b)-F (a).

a
It wil\ be noted that the difference F (b)-F (a) is independent

of the choice of antiderivative F, since all antiderivatives differ by
a constant quantity, which disappears upon subtraction anyway.
*) It is necessary to point out that the name of formula (2) Is not exact,
since neither Newton nor Leibniz had any such formula. The important thing,
however, is that namely Leibniz and Newton were the first to establish a re-
lationship between integration and differentiation, thus making possible the
rule for evaluating definite. integrals.
If we introduce the notation*)

F (b)-F (a)= F (x) 1~.
then formula (2) may be rewritten as follows:
b
~ f(x) dx=F (x) I~=' F (b)-F (a).

a
The Newton-Leibniz formula yields a practical and convenient

method for computing definite integrals in cases where the antide-
rivative of the integrand is known. Only when this formula was
discovered did the definite integral acquire its present significance
in mathematics. Although the ancients (Archimedes) were familiar
with a process similar to the computation of a definite integral as
the limit of an integral sum, the applications of this method were
confined to the very simple cases when the limit of the sum could
be computed directly. The Newton-Leibniz formula greatly expanded
the field of application of the definite integral, because mathemat cs
obtained a general method for solving various problems of a par-
ticular type and so could considerably extend the range of appli-
cations of the definite integral to technology, mechanics, astronomy,
and so on.
Example l.
b
2 2
5xdx=-x 1b = --a
a
2 a
b
-. 2
Example 2.
b
5x2dx=:rx'lba
a
b'-a'
=-3-.
Example 3.
*) The expression ~~ is called the sign of double substitution. In the lite-

rature we find two notations:
F (b)-F (a)= [f (x)]~
or
F (b)-F (a)= F (x) I~
We shall use both notations.
412 The Definite[ n.tegral
Example 4.
b
~ e" dx = e" I~ = e0 - e0
a
Example 5.
2:rt
) slnx dx =-cos xl~n =-(cos 2n:-cos 0) =0.

0
Example 6.
SEC. 5. CHANGING THE VARIABLE IN THE DEFINITE

INTEGRAL
Theorem. Let there be an integral

b
~ f (x)dx,
a
where the function f (x) is continuous on the interval [a, b].
Introduce a new variable t using the formula
X= (jl (t).
If
1) cp(a)=a, cp(~)=b,
2) cp (t) and cp' (t) are continuous on [a, ~].
3) f [cp (t)] is defined and is continuous on [a, ~]. then
b f:l
~ f (x) dx =) f[cp (t)) cp' {t) dt. (1)
a a
Proof. If F (x) is an antiderivative of the function {(x), we can
write the following equations:
~ f (x) dx = F (x) + C, (2)
~ f [cp (t)l cp' (t) dt = F [cp (t)]+C. (3)
. . .
The truth of the latter equation is checked by differentiation of
both sides with respect to t. [It likewise follows from formula (2),
Sec. 4,_Ch. X.) From (2) we have
b . b .
~ f (x) dx = F (x) Ia = F (b) -F (a).
a
Integration by Parts 413
From (3) we have y

II
~,liP(/)] q>' (t) dt = F [cp (t)]
a
r:
=FF [cp(~)]-F (cp(a)l =
= F (b) -F (a).
The right sides of the latter expressions
are equal, and so the left sides are equal
as well, thus proving the theorem.
Note. It will be noted that when com-
puting the definite integral from formula (I) Fig. 217.
we do not return to the old variable. If we
compute the second of the definite integrals of (1), we get a cer-
tain number; the first in~egral is also equal to this number.
Example. Compute the integral
r
~ Yr 2
-x 2 dx.
0
Solution. Change the variable:

x=r sin t, dx =r cost dt.
Determine the new limits:
x=O for 1=0,
x=r for I=~
2.
Consequently,
r z z
~ Yr 2 -x 2 dx= ~ V r 2 - r 2 sin 2 t r cost dt =r' ~ Vl-sin 2 t cos I dt =
0 0 0
~ ~
= r' Scos t dt = r s(;+-} cos 21) dt = r' [ { +sl~ 21] : = n;z.

2 2
0 0
Geometrically, the computed integral is the area of ! of the circle bounded

by the circle x 2 +y'=r 2 (Fig. 217).
SEC. 6. INTEGRATION BY PARTS
Let u and u be dilferentiable functions of x. Then

(uu)' = u'u + uu',
Integrating both sides of the identity from a to b, we have

b b b
~ (uv)' dx = ~ u'v dx + ~ uv' dx. (1)

a a a
b
Since~ (uv)' dx=uv.+C, we have ~ (uv)' dx=uv

a
1:; for this reason,
the equation can be written in the form
b b
uv 1: = ~ v du+ ~ udv,
a a
or, finally,
b b
~ u dv = uv 1: - ~ u du.
a a
!!.
a
Example. Evaluate the integral In=~ sinn x dx.
0
n n: n:
2 2
In=~ sinnxdx= ~ sinn- 1 xslnxdx=- ~s~~=

o o o u du
=-sl~n-tx
-n:
cosx/: +<n-1)
.
n:
~ sinn-xcosxcosx.dx=
0
n:

=(n-1) ~ sinn-xcos 2 xdx=
0
n

=(n-1) ~ slnn-x(1-sln 2 x)dx=
0
n: n

=(n-1) ~ sinn-xdx-(n-1) ~ slnn x dx.
0 0
In the notation chosen we can write the latter equation as

In= (n-1) I n- 2 -(n-1) I n
Integration bu Part~ 415
whence we find
n-1
ln=-n-ln- (2}
Using the same technique we find
and so
n-l n-3
In=-- --2/n-4
n n-
Confinuing in the same way, we arrive at I 0 or I 1 depending on whether
the number n is even or odd.
Let us consider two cases:
1) n is even, _n=2m:
2m-1 2m-3 3 1
1m = 2 m ' 2m-2 4 ' 2 10;
2) n is odd, n=2m+l:
2m 2m-2 4 2
1
m+~=2m+t'2m-I s'3 11 '
but since
2 2
I0= 5sln x dx = Sdx = ~ ,

0 0
~
2
11 = ~ sinxdx=l,
0
we have
n
2m-! 2m-3 5 3 1 n
2m
Ism= Ssin x dx = ~ 2m-2 ... 6 4 2 2
0
~
2
2m+ 1 d 2m 2m-2 6 4 2
I am+t=
SsnI
0
x x=2m+I '2m-l ... 7'5'3
From these formulas there follows the Wallis formula, which expresses the
number ~ in the form of an infinite product.
Indeed, from the latter two equations we find, by means of termwise dt.
vision,
n ( 246 ... 2m ) I 12 ,
(3)
2= 35 .. (2m-l) 2m+l lsm+l.
4!6 The Definite Integral
We shall now prove that

lim ~=1.
m-+ 00 lzm+t
For all x of the interval ( 0, ~) the inequalities
sinzm-t x > sin 2rn x > sin 2 m+t x

hold.
n
Integrating from 0 to -
2 , we get
whence
(41
From (2) it follows that
Hence,
lim lzm-t=lim 2m+I=l.
m-+r:~>l 2 m+Jm-+<~> 2m
From inequality (4) we have
lim ~=1.
m-+ "'12m+t
Passing to the limit in formula (3), we get Wallis' formula (Wallis' product) for
n r [( 246 ... 2m ) 2 1
2=m~"' 35 ... (2rn-l) 2m+I
J
This formula may be written in the form
n . ( 2 2 4 4 6 2m-2 2m 2m )
2=ml~"' T'3'3'5'5"'2m-1 '2m-l '2m+l'
SEC. 7. IMPROPER INTEGRALS
t. Integrals with infinite limits. Let the function f (x) be defined

and continuous for all values of x such that a E:;;; x < oo, +
Consider the integral
b
!(b)=~ f(x)dx.
a
This integral is meaningful for any b >a. The integral varies with
b and is a continuous function of b (see Sec. 4, Ch. XI). Let us
consider the behaviour of this integral when b oo (Fig. 218). -+
Improper Integrals 417
Definition. If there exists a finite limit

b
lim ~ f (x) dx,

b-++aoa
then this limit is called the improper integral of the function f (x)
in the interval (a, + oo) and is denoted by the symbol
+<XI
~ f (x) dx.
a
Thus, by definition, we have

+<XI b
~ f (x) dx = lim ~ f (x) dx.

a b-++aoa
+ao
In this case it is said that the improper integral ~ f (x) dx exists
u
b
or converges. If ~ f (x) dx as b -+ + oo does not have a finite limit,

a
+GO
one says that ~ f (x) dx does not exist y
a
c;>r diverges.
It is easy to see the geometric
meaning of an improper integral
for the case when f (x) ;;;;;=. 0: if the
b
integral ~ f (x) dx expresses the area Fig. 2I8.

a
of a region bounded by the curve y = f (x), the x-axis and
the ordinates x =a, x = b, it is natural to consider that the im
+oo
proper integral ~ l (x) dx expresses the area of an unbounded (in-

a
finite) region lying between the lines y = f (x), x =a, and the axis
of abscissas.
We similarly define the improper integrals of other infinite in
tervals:
a a
S f (x) dx ~ lim Sf (,\:) dx,

-a:~ ' a-+-cou
+ao c +oo
S f (x) dx = S f (x) dx + S f (x) dx.

-a:~ -O"J c
14-3388
The Iat ter equation should be understood as follows: if each of

the improper integrals on the right exists, then, by definition,
the integral on the left also exists (converges).
+ex>
Example 1. Evaluate the integral S 1 ~x (see Figs. 219 and 220).
0
Fig. 219.
Solution. By the definition of an improper integral we find

+"" b
S dx =
l+ X 2
lim
b->+ex>
S I+dx X
2 = lim arctanxlb=
b->+oo o
lim arctanb=l't
b..,.+ao 2
0 0
This integral expresses the area of an infinite curvilinear trapezoid cross.

hatched in Fig. 220.
Example 2. Find out at which values of a (Fig. 221) the integral
+ex>
(' dx
J xa
1.
converges and at which it diverges.

Solution. Since (when a =F 1)
y b
Sd;x =-~-x~-"lb
I
1-a
=-~-w-~~-11,
1-a 1
we have
+ex>
S d;=
Y=-f xa lim _I_(IJI-11_1).
x b .... +oo 1-a
1
Consequently,
+ex>
dx 1
>
0
If a 1, then
S7 =a _ 1 , and the
Fig. 221. integral converges!

Improper Integrals 419
+ao
if a< 1, then S ~: = oo, and the integral diverge$.
I
+oo
Whe,n a= 1, s d: =In X roo= oo, the integral diverges.
+oo
Example 3. Evaluate S 1 :.t2
Solution.
+cc 0 +e
s
-~
I ~x 2 = s
-oo
I ~x 2 + s I !\ 0
2 '
The second integral is equal to ~ (see Example 1). Compute the first InTegral:
0
S 1+x
-oo
dx
--2 =
.
1rm
a-+-ao
a
s0
-dx- = .
1rm
1+x 2 a-+-oo
0
arctanx j =
a
. n
= a-+-ao
1rm (arc tan 0-arc tan a)=-.
2
Therefore,
In many cases it is sufficient to determine whether the given

integral converges or diverges, and to estimate its value. The fol-
lowing theorems, which we give without proof, may be useful in
this respect. We shall illustrate their application in a few cases.
Theorem t. If for all x (x;;;;;:. a) the inequality
0 < f (x) < !p (x)
+< +ao
is fulfilled and if ~ !p (x) dx converges, then ~ f(x) dx 'li-SO
a a
converges, and
+co +c:o
~ f (x) dx ~ ~ !p (x) dx.
a a
Example 4. Investigate the integral
for convergence.
14*
Solution. It will be noted that when 1,.,;;;;; x,

I I
--;;--;.,.--;----;::-:-
xz (I +ex) < xz. -
And
+oo
SI I
2 dx=--
X X
I+"' =I.
1
I
Consequently,
+lXI
dx
S
I
x 2 (1 +e"')
converges, and its value is less than I.
Theorem 2. if for all x(x;;;;;:.a) the inequality O:e;;;;cp(x):e;;;;f(.t)

+< +oo
is fulfilled, and ~ cp (x) dx diverges, then the integral ~ I (x) dx
a a
also diuerges-:--
Example 5. Find out whether the following integral converges:
We notice that
x+ I x I
-->--=--
Vx yxa vx
But
+"'
dx .r-~b
lim 2 r x =+ oo.
S .r-=
1
rx b-++"' 1
Consequently, the given integral also converges.
In the last two theorems we considered improper integrals of

nonnegative functions. For the case of a function f (x) which changes
its sign in an infinite interval we have the following theorem.
+co
Theorem 3. if the integral ~ If (x) I dx converges, then the in-
a
+"'
tegral ) f (x) dx also converges.
a
rn this case, the latter integral is called an absolutely conver-
gent integral.
Improper Integral.~ 421.
Example 6. Investigate the convergence of the integral

+oo
5 x' x x.
I
sin d .
Solution. Here, the integrand Is an alternating function. We note that
.
Therefore, the integral sI
+>
I
7
slnxl dx converges. Whence it follows that the
given integral also converges.
- 2. The integral of a discontinuous function. Let the function

f (x) be defined and continuous when a~ x < c, and for x = c let
the function be either not defined or let it be discontinuous. In
c
thi~ case, one cannot speak of the integral ~ f (x) dx as of the li
a
mit of integral sums, because f(x) is not continuous on the inter
val [a, c), and for this reason the limit may not exist.
c
The integral ~ f(x)dx of the function f(x) discontinuous at the
a.
point c is defined as follows:
c b
~ f (x) dx = lim ~ f (x) dx.

a b .... c-oa
If the limit on the right exists, the integral is called an impro-

per convergent integral, otherwise it is divergent. _
If the function f (x) is discontinuous at the left ,extremity the ot
interval [a, c) (that is, for x =a), then by definition
a c
~f(x)dx= lim ~f(.t)dx.
a b-+-a-t-ob
If the function f (x) is discontinuous at some point x = x 0 inside

the interval [a, c). we put
c x, c
~ f (x) dx = ~ f (x) dx + ~ f (x) dx,
a u ~
If both Improper integrals on the nght side of the equation exist.

422 The. Definite Integral
Example 7. Evaluate
I
dx
5 Vl-x
Solution.
b
5- - - = -
1
dx- = 11m. dx . 2 .y r1---~b =
5
0
--
Y i-x b __,. 1 - o 0
Y I -x
11m
b-+ 1 - o
X
0
= - lim 2 [YT=b"-11=2.
b-+1-0
I
dx.
Example 8. Evaluate, the integral
5Xi ..
-I
Solution. Since inside the interval of integration there exists a point x=O
where the integrand is discontinuous, the integral must be represented as the
sum of two terms:
1 8t t
5x ~= lim
e, -+ - o
5xt!!.+ 2 lim
e, -+ + o
5x dx
2
'
-1 -1 e2
Calculate each limit separately:
e,
.
IJm
e -+- o
5---.=-
x
I'm -I le, = - 1.Jm ( -
dx
e -+- o x -I
1 --
e,-+- o e.
1 ) =00.
-I
1 -I
1
Thus, the integral diverges on the interval [-I, 0]:
1
lim 5d~=- lim (1-_!_)=oo.

e., -+ + o X e, -+ + o e,
e,
And this means that the integral also diverges on the interval [0, I].
Hence, the given integral diverges on the entire interval [-1. 1].
It should be noted that if we had begun to evaluate the given integral
without paying attention to the discontinuity of the integrand at the point
x = 0, the result would have been wrong.
Y Indeed,
I
5 x = - _!_1
dx
x
=-(_!_I __I ) = - 2,
-1 -1
y=fr -I
which is impossible (Fig. 222).
Note. If the function f (x), defined

on the interval [a, b] 1 has, within this
interval, a finite number of points of
X discontinuity a., a,, ... , an, then the
' Fig. 222. integral of the function f (x) on the
1 mproper I nlegrals 423
interval [a, b} is defined as follows:

b a1 a, b
~ f (x) dx = ~ f (x) dx + ~ f (x) dx + ... + ~ f (x) dx,

a a a, an
if each of the improper integrals on the right side of the equation

converges. But if even one of these integrals diverges, then
b
~ f (x) dx is called divergent as well.

a
For determining the convergence of improper integrals of dis-
continuous functions and for estimating their values, one can
frequently make use of theorems similar to those used to estimate
integrals with infinite limits.
Theorem I'; If on the interval [a, c] the functions f (x) and <p (x)
ar1 discontinuous at the point c, and at all po~nts of this interval
the inequalities <p (x) ~ f (x) ~ 0 are fulfilled and ~ <p (x) dx converges,
a
c
then ~ f (x) dx also converges.
a
Theorem 2', If on the interval [a, c] the functions l (x) and
tp (x) 'are discontinuous at the point c, and a( all points of this
c
interval the inequalities f(x)~<p(x)~O are fulfilledand ~<p(x)dx
a
c
diverges, then ~ f (x).dx also diverges.
a . .
Theorem 3'. If f (x) is an alternating function on the interval
[a, c] and discontinuous only at the point c, and the improper
c
integral ~ If (x) 1 dx of the absolute value of this function converges,
a
c
then the integral ~ f (x) dx of the function itself also converges.
a
1
Use is frequently made of -
e-x
( )" as functions with which it is
convenient to compare the functions under the sign of the improper
c
1
integral, It is easy to verify that S(c x)" dx convhges for a< 1,
a
and diverges for a~l.
c
The same applies also to the integrals S (x
1
a)" dx.
a
I
Example 9. Does the integral SV x +I 4x 1

dx converge?
0
Solution. The integrand is discontinuous at the left extremity of the in-
.terval [0, 1]. Comparing it with the function ~x , we have
I I
Vx +4x' < Vx
I
dx
The improper integral
S
0
- , - exists. Consequently, the improper integral
X 12 .
of a lesser !unction, that is, SV x +I 4x' dx, also exists.

0
SEC. 8. APPROXIMATING DEFINITE INTEGRALS
At the end of Chapter X it was pointed out that not for every
continuous function is its antiderivative expressible in terms of
elementary functions. In these cases, computation of definite in-
tegrals by the Newton-Leibniz formula is involved, and various
methods of approximation are used to evaluate the definite inte-
grals. The following are several methods of approximate integration
based on the concept of a definite integral as the limit of a sum.
I. Rectangular formula. Let a continuous function y = f (x) be
given on an interval [a, b]. It is required to evaluate the definite
integral
b
~ f (x) dx.
a
Divide the interval[a, bl by the points a=X 0 , xi, X1 , , Xn=b

into n equal parts of length ~x:
b-a
~X=--,
n
Then denote by Yo Yt~ y 1 , , Yn-l Yn the values of the func-

tion f (x) at the points X0 , X11 X 2 , , xn; that is,
Yo= f (xo); Y1 = f (xJ; ; Yn = f (xn)
Approximating Definite Integrals 425
Form the sums:

y.tJ.x + Y1 !J.X + ... + Yn- 1 !J.X,
Y, !J.x + Ys!J.x + + Ynl\x.
Each of these sums is an integral sum for f (x) on the interval
la. b) and for this reason approximately expresses the integral
b
b-a
Sf (x) dx::::::: -n- (Yo+ Y + Yz + + Yn-t),
a
1
(1)
b
b-a
5f
a
(x) dx::::::: -n- (y 1 + Yz + + Yn). ( 1,)
This is the rectangular formula. From Fig. 223 it is evident

that if f (x) is a positive and increasing function, then formula (I)
expresses the area of the step-like figure composed of "inside"
rectangles, while formula ( 1') yields the area of the step-like figure
composed of "outside" rectangles
y y
Yn-1 Yn
Fig. 223.
The error made when calculating integrals by the rectangular

formula diminishes with increasing n (that is, the smaller the
. . .
d lVISIOnS A b-a)
uX=-n- .
II. The trapezoidal rule. It is natural to expect that we will
obtain a more exact value of the definite integral if we replace
the curve Y=f (x) not by a stepped line, as in th_e rectangular
formula, but by an inscribed broken line (Fig. 224). Then the
area of the curvilinear trapezoid aABb will be replaced by the sum
of the areas of _the rectilinear trapezoids bounded from above by
the chords AA,, A, A,, .. , An_ 1 B. Since the area of the first of
these trapezoids is Yo+Y.

2 11x , the area of the second is Y+Yf1x
2 ,
and so forth, so
s
a
b
f(x) dx ~ (Yo!Y 11x+ x,!y"l1x + ... +Yn-2+Yn 11x)

or
b
St(x)dx~ b~a(y!Yn+Y,+Y.++Yn- 1 )-' (2)

a
This is the trapezoidal formula (trapezoidal rule).

The choice of n is arbitrary. The greater this number, the smaller
will be the division (subinterval) 11x= b-a and the greater will
n
be the accuracy with which the sum, written on the right side of
the approximate equality (2), yields the value of the integral.
Ill. Parabolic formula (Simpson's rule). Divide the interval
[a, b] into an even number of parts n =2m. Replace th.e .ar.ea of
the curvilinear trapezoid, corresponding to the first two subinter-
vals [x 0 , x,] and [X 1 , X 2 ] and bounded by the given curve y =.f (x),
by the area of a curvilinear trapezoid such that is bounded by a
quadratic parabola passing through three points:
M (x., Y.); M 1 (xl' Y1 ); M. (x., Y.),
and with ail axis parallel to the y-axis (Fig. 225). We shall call
this kind of curvilinear trapezoid a parabolic trapezoid.
The equation of a parabola with an axis parallel to the y-axis
is of the form
y=Ax" + Bx+C.
The coefficients A, B and C are uniquely determined from the
condition that the parabola passes through three specified points.
Analogous parabolas are constructed for other pairs of intervals as
well. The sum of the areas of the parabolic trapezoids will yield
the approximate value of the integral.
Let us first compute the areas of one parabolic trapezoid.
Lemma. If a curvilinear trapezoid is bounded by the parabola
y=Ar+Bx+C,
the x-axis and two ordinates separated by a distance 2h, then its
area is
(3)
where Yo and y, are the extreme ordinates and Y1 is the ordinate

of the curve at the midpoint of the interval.
Proof. Arrange an auxiliary coordinate system as shown in
Fig. 226. y
X
Fig. 225. Fig. 226.
The coefficients in the equation of the parabola y =Ax'+ Bx +

+C are determined from the following equations:
if X0 =-h, then Y.=Ah'-Bh+C; }
if x 1 =0, then Y1 = C; (4)
if .-i,=h, then Y,=Ah'+Bh+C.
Considering the coefficients' A, B, C known, we determine the
area of the parabolic trapezoid with the aid of a definite integral:
h
S= s(Ax'+ Bx+
-h
C) dx= [At+ ;' +Cx] ~h =
8
~ (2Ah" + 6C).
But from equalities (4) it follows that
Yo+ 4y. + y, = 2Ah' + 6C.
Hence,
h
S=a<Yo+4y.+y,),
which is what had to be proved.
Let us come back to our basic problem (see Fig. 225). Using
formula (3) we can write the following approximate equalities
(h= Ax):
Xso ~X
f(x) dx ~ 3 (y 0 + 4y 1 + Y1 ),
x,
Sf(x) dx ~ t-.; (Y 2 +4Y 1 +y,),
x,
Adding the left and right sides, we get (on the left) the sought-
for integral and (on the right) its approximate value_:
b
Sf (x) dx ~ t-.; (y 0 + 4y + 2y
1 2 7" 4y + ..
1 o
a
(5)
or
+ 4 (Yt + Ya + + Y.m-t)J.
o
y This is Simpson's formula (rule).

Here, the number of division
points 2m is arbitrary; but the
more of them there are, the
more accurately the sum on the
right side of (5) yields the value
of the integral. *)
Example. Evaluate approximately
0 X 2
dx
'227.
ln2=
Sx'
I
Solution. Divide the interval (I, 2] into 10 equal parts (Fig. 227). Assuming
2-1
/',X= ""T(i"'" = 0 .J,
*) To find out how many division points are needed to compute an inte-
gral to the desired number of decimal places, one can make use of formulas
lor estimating the error resulting !rom approximating the integral. We do not
give these estimates here. The re,ader will find them in more advanced courses
ol analysis; see, for example, Fikhtengolts, "Course of Differential and
Integral Calculus", 1959, .Vol. II, Ch. IX, Sec. 5. (R usdan edition).
we make a table of .the values of the integrand:
I I
X y=- X y=-
X X
-
X 0 = 1.0 y 0 = I 00000 x 6 = 1.6 y.=0.62500
XI=).) y 1 =0.90909 X1 =!.7 y 7 =0.58824
x 2 =1.2 y 2 =0.83333 x 8 = 1.8 y 8 =0.55556
x,= 1.3 y,=0.76923 x. = 1.9 Yo =0.52632
x4 = 1.4 y.=0.71429 x 10 =2.0 y 10 = 0.50000
x. = 1.5 y,=0.66667
Ely the fir~t rectangular formula (!). we get

2
dx
Sx
I
::::::::0.1 (y 0 -f-Y 1 + ... -j-y 9 ) =0.17.18773 =0.71877.
By the second rectangular formula (I') we get

2
Sd: ::::::::0.1 (y 1 + Y+ ... + Y

2 10 ) =0.16.68773 =0.66877.
I
It follows directly from Fig. 227 that in this case the first formula yields
the value of the integral with an excess, the second, with a defect.
II. By the trapezoidal rule (2), we have
2
Sdxx ::::::::0.1 ( 1 +2 5 -j-6.18773) =0.69377.

I
Ill. By Simpson's rule (5), we have

.sd:
I
::: 3 01
[Yo-f-Yio-f-2(y.+y.+yo-f-Ya)-f-4(yl-j-y,-t-ys-f-Y1-l-Yo)]=
-
0 1
= 3 (I+ 0.5.+22.72818 -j- 4J.45955) =0.69315.
Actually, In 2= s 2
dx
--x-=0.6931472 (to seven places of decimals).
I
Thus, .when dividin~ the. interval (0, I) i~to 10 parts by Simpson's rule, we
get five s1gmficant dec1mals; by the trapezoidal rule, only three; and by the
rectangular formula, we are sure only of the first decimal.
SEC. 9. CHEBYSHEV'S FORMULA

In engineering computations, use is frequently made of Cheby-
shev's formula of approximate integration.
b
Once again, let it be required to compute ~ f (x) dx.

a
Replace the integrand by the Lagrange interpolation polynomial
P (x) (Sec. 9, Ch. VII) and take certain n values of the function
on the interval [a, b]:f(xJ, f(x 2 ), , f(xn) wbere x,, X 2 , , xn
are any points of the interval [a, b]:
p (x) = (x-x 2 ) (x-x,) ... (X-Xn) f (X)+
(x 2- x 1)(x 1-x1) .. (x 1-xn)
+ (x-x 1) (x-x,) ... (X-Xn) f(x )+

(x 2 - X 1)(x 2 -x 3) (x 2 -xn) z
(1)
We get the following approximate formula of integration:

b b ' .
~ f (x) dx ~ ~ P (x) dx; (2)
a a
after some computation it takes the form

b
~ f(x) dx ~ e, f(x,) +e 1 {(x 2 ) + ... +en f(xn), (3)

a
where the coefficients e1 are calculated by the formulas
(4)
Formula (3) is cumbersome and inconvenient for computation

because the coefficients e1 are expressed by complex fractions.
Chebyshev posed the inverse p'roblem: specify not the abscissas
x.. XI, ' Xn but the coefficient~ e,, el, ... , en and determine
the abscissas x,, X 1 , , xn.
Chebysheu's Formula 431
The coefficients C1 are specified so that formula (3) should be

as simple as possible for computation. This will obviously occur
when all the coefficients C1 are equal:
C1 =C.= ... =Cn.
If we denote the total value of the coefficients C,, c... , Cn

by en, formula (3) will take the form
b
~ f (x) dx ~ c n [ f (x,) + f (x.) + ... + f (xn)]. (5)

a
Formula (5) is, generally speaking, an approximate equality, but

if f (x) is a polynomial of degree not higher than n-1, then the
eql.)a!ity will be exact. This circumstance is what permits determin-
ing the. quantities en, x,, x ... ' xn.
To obtain a formula that is convenient for any interval of in-
tegration,. let us transform the interval of integration [a, b 1 into
the interval [ -1, lJ. To do this, put
X =a+b +b-at
2 2 '
then for t=-1 we will have x=a, for t=l, x=b.

Hence,
b I I
sf (x) dx = b 2 a sf (at b + b 2 a t ) dt = b 2 a s q> (t) dt'

a -1 -1
where q> (t) denotes the function of t under the integral sign.
Thus, the problem of integrating the given function f (x) on the
interval [a, bj can always be reduced to integrating some other
function q> (x) on the interval [- I, 1].
To summarise, then, the problem has reduced to choosing, in
the formula
1
~f(x)dx=Cn[f(X 1 )+f(x.)+ ... +f(xn)l, (6)

-1
the numbers en, x,, x . , xn so that this formula will be exact

for any function f (x) of the form
(7)
It will be noted that

I I
~ f (x) dx = ~ (a 0 + a,x + a 2 X 2 + a,_,x"-') dx =

_, -I
On the other hand, the sum on the right side of (6) will, on the
basis of (7), be equal to
C,. [na 0 +a, (x 1 + X 2 + ... + x,.) + a 2 (x~ + x: + ... + x~,) + ...
(9)
Equating expressions (8) and (9), we get an equation that should

hold for all a0 , al' a1 , , a,_,:
2 ( ao + t + ~ + ~ + . ) =
= c~ [nao +a, (x, + x2 + ... + x,.) +

+a. (x~ +x~ + ....+ x~) + .. .a,_ 1 (x~-~ + x~-'+ ... +x~- )].1
Equate the coefficients of a0 , a,, a., a3 , , a,_, on the left

and right sides of the equation:
I
2
2 =C n nor Cn =-
n '
X1 + X -t- ... +X,.= 0;
2
. + + 2
x, + x. . . . x, = 3C,.
n .
= "3 ' (10)
x: + x: -t- .. , -t- X~ = 0; 2 n
X~ f X~+ ... + X~= 5-C-,. =s;
..............
From the_ latter .n equations we find the abscissas x,~ x 1 , . , x,..
These solutions were found by Chebyshev for various. values of n.
Chebyshev's Formula 433
The following solutions are those that he found for cases when the
number of intermediate points n is equal to 3, 4, 5, 6, 7, 9:
Number' of
ordinates n
Coefllcten t
c.,
I Values of abscissas
x1 x2 ...., Xn
I
2 x1 = - x 3 =0.707107
3
3 X 2 =0
I x, = - x 4 = 0. 794654
4
2 X2 = - X 3 = 0.187592
2 x, = - x, = 0.832498
5 X 2 = - x4 =0.374541
5 x 3 =0
I x1 =-x,=0.866247
6 x2 =-X5 =0.422519
3 X3 = - X 4 =0.266635
x, = - x, = 0.883862
2 X2 = - X8 = 0.529657
7
7 x3 = - x 5 =0.323912
. X4 =0
x, = - x 9 = 0.911589
2 x2 =-x 8 =0.601019
9 y x. = - x, = 0.528762
x, = - x. =0.167906
x,=O
Thus, on the interval [ -1, 1], an integral can be approximated

by the following Chebyshev formula:
I
5f (x) dx =! [f (X 1) +f(x 2 ) + ... +f(xn)J,

-I
where n is one of the numbers 3, 4, 5, 6, 7 or 9, and xt> .. , x,.

are the numbers g.iven in the table. Here, n cannot be 8 or any
number exceeding 9; for then the system of equations (10) .. yields
imaginary roots.- '.
When the given integral has limits of integration a and b, the

Chebyshev formula takes on the form
b
~f(x)dx=b / [f(X 1)-l--f(X 1 )+ ... +f(Xn)],

a
where X 1 =b+a
-2
b-a
-+- . d
2 -x1 (t=l, 2, ... , n) an x 1 have the values
given in the table.
The following example illustrates the use of Chebyshev's approx-
imation formula for calculating an integral.
!
Example. Evaluate Sdxx (=ln2).

I
Solution. First, by changing variables, transform this integral into a new
one with limits of integration -1 and 1:
x-1 +2 +2-1 t-~+..L- 3+t
- 2 2 -2 2- 2 '
dt
dx=
2 .
Then
5 s
I 1
dx
X
= __!!!.__
3+t '
I -1
Compute the latter integral, taking n=3, by Chebyshev's formula:
-1
s
1
f (t) dt = ~ [f (0.707107) + f (0) + f ( -0.707107)}.
Since
1(0.707107) 3+0.7~7107 = 3.7~7107 0.269752,
I
f (0) = 3 +O =0.333333,
I 1
l< -0.707107) = 3-0.707107 = 2.292893 =0.436130,
we have
s3~1 = i
1
(0.269752+0.333333+0.436130)=
2
=:rl.039215 =0.692810::::::: 0.693.
Integrals Dependent on a Parameter 435
Comparing this result with the results of computation using the rectan-
gular formulas, the trapezoidal rule, and Simpson's rule (see the example
in the preceding section), we note that the result given by Chebyshev's
formula (with three intermediate points) is in better agreement with the
true value of the integral than the result obtained by the trapezoidal rule
(with nine intermediate points).
The theory of approximating integrals was further developed in

the works of Academician A. N. Krylov (1863-1945).
SEC. 10. INTEGRALS DEPENDENT ON A PARAMETER
Differentiating integrals dependent on a parameter. Let there be

an integral
b
I (a)= ~ f (x, a) dx, (I)

a
in which the integrand is dependent upon some parameter a. If

the parameter a varies, then the value of the definite integral will
also vary. And the definite integral is a function of a; we can
therefore denote it l?y I (a).
1. Suppose that f (x, a) and f~ (x, a) are continuous functions
when
co;;;;;ao;;;;;d and ao;;;;;xo;;;;;b. (2)
Find the derivative of the integral with respect to the parameo
ter a:
. /(a+M)-/(a) _ . ()
I tm - 1a a.
Aa-+o .1-a
In finding this derivative we note that

b
I (a+ 6:a) = ~ f (x, a+ L1a) dx
a
and, consequently,
1 (a+ L1a)-l
a
b
a
b
(a)=~ f (x, a+ L1a) dx- ~ f (x, a) dx= \
b
= ~ [f(x, a+L1a)-f(x, a)]dx;

a
I (a+ Mt)-/ (a)=

.1-a
s b
1 (x, a+ ,1.a)-f (x, a) d
.1-a X.
g
Applying the Lagrange theorem to the integrand we have

{(x,a+M)-f(x,a)
,ila
=f'(
a X, a
+eA)
ua ,
where 0 < e < 1.
Since f~ (x, a) is continuous in the closed domain {2), we have
f~(x, a+9da)=f~(x, a)+e,
where the quantity 8, which depends on x, a, da, approaches
zero as da--+0.
Thus,
b b b
1
S
(a+Ll:~-l (a)= [f~ (x, a)+ c:] dx= Sf~ (x, -:x) dx+ S dx. 8
a a a
Passing to the limit as da --0, we have*)
b
. '<a+M>-I<a>
l tm
ll.a-+o
A
..,a
-I<a>-St'
- a- < a X, a )dX
a
or
b b
Uf (X, a) dx ]~ = ~ fa (x, a) dx.
a a
This formula is called the Leibniz formula.

2. Now suppose that in the integral (1) the limits of integration
a and b are functions of a:
b(a)
I (a)= (a, a (a), b (a)J = ~ f (x, a) dx. (1 ')

a(a)
Win, a (a), b (a)] is a composite function of a, and a and b are

intermediate arguments. To find the derivative of I (a), apply the
rule for differentiating a composite function of several variables
(see Sec. 10, Ch. VIII):
I' (a)-aCll
- aa
+ aCll~+aCll
aa da
db
ab da
( 3)
b
*) The integrand in the integral ~ e da approaches zero as ,ila~ O.. From
. a .
the fact that the integrand approaches zero it does not always follow that
b
lhe integral also appro_aches zero. Howev,er, in the given case, ~ e dx
. I Q I
approaches zero .a~ ,ila .:...:...a. We accept .this fact without proof.
Integrals Dependent on a Parameter 437
By the theorem for the differentiation of a definite integral

with respect to the variable upper limit [see formula (1), Sec. 5]
we get
b
acD a Sf(x,
7fb=ab a.)dx=flb (a), a].
a
b a
~~=a~ St(x,a)dx=- 0~ St(x,a)dx=-f[a(a),aJ.

b b
Finally, to evaluate ~: use the above-derived Leibniz formula:

b
~~ = S f~(x, a)dx.
a
Substituting into (3) the expressions obtained for the derivatives,
we have
s
la(a)=
b(a)
fa(X, a)dx+f[b (a), u) db

da-f[a(a). a] da
da. (4)
a(a)
Using the Leibniz formula it is possible to compute some defi-

nite integrals.
Example. Evaluate the integral
OJ
se-x sl:ax dx.

0
Solution. First note that it is impossible to compute the integral directly,

because the antiderivative of the function e-x sinax is not expressible in
X
terms of elementary functions. To compute this integral we shall consider it
as a function of the parameter a:
"' sin ux
I {a)=
Se-x-x_ dx.
0
Then its derivative with respect to a is found from the above-!jerived Leibniz
formula*):
I' (a)=
00
S[ sin ax]' dx =
e-x -x-
0
s
OJ
e-x cos ax dx.

0 0
*) Leibniz' formula was derived on the assumption that fhe limits of inte.
gration a and b are finite. However, in this case Leibniz' formula also holds,
even though one of the limits of integration is equal .to inflnity.
But the latter integral is readily evaluated by means of elementary functions;

1
it is equal to - Therefore,
1 +a2
I' (a)= 1 ~ a 2
Integrating the identity obtained, we find I (a):

I (a)=arc tan a+ C. (5)
We have C to determine now. To do this, we note thaf
en
I (0)= - e-" -
slnOx s
S x - dx = 0 dx=O.
oo
0 0
What is more, arc tan 0=0.

Substituting into (5) a=O, we get
I (O)=arc ian O+C.
whence C =0. Hence, for any value of a we have the equality
I (a)= arc tan a;
that is,
00
sinax
Se-" - x - dx =arc tan a.
0
Exercise~ on Chapter xt
1. Forming the integral sum sn and passing to the limlf, compute the
definite integrals
b
~ x 2 dx.
0
Hint. Divide the interval (a,b] intonpartsbythe pointsx;=aql(l=O,l,

2, .. ,n),
b
where q=
V ab . b 1 -a1
Ans. - - 3
-
2. Sdxx where 0 <a< b. Ans, In !!._.

a
a
Hint. Divide ihe interval [a, h) in the same way as in the preceding
example.
b
3. SYx dx. 2 'I-a'I).

Ans. 3(b
a
Hint. See Example 2.
b
4. Ssin x dx. Ans. _cos a-cos b.
a
Exercises on Chapter XI 439
Hint. First establish the following identity:

sin a+sin (a+h)+sin(a+2h)+ ... +sin [a+ (n-1) h)=
cos (a+ 4) -cos [a + nh-%]

. h
2 smy
h
To do this, multiply and divide all the terms of the left side by sin and
2
replace the product of sines by the difference of cosines.
b
5. Scos x dx. Ans. sin b -sin a.

a
Using the Newton-Leibniz formula, compute the definite integrals:

n
1 1 2
6. Sx dx. Ans. ~ . 7. Se" dx. Ans. e-1. B. Ssin x dx. An,s. 1.

0 0 0
n
I 2 3
9.
S
dx
I+xz
:1:
ns. 4 .10.
0
sY ~x 1 . Ans. : . 11. s tan x dx. Ans. In 2.
8 X X X
12. S~: . Ans. 1 . . 13. Sd: .An~.lnx. 14. Ssin x dx. Ans. 2sin 2
15_. S x'd>J.
va
;
1 1
n
z 2
x 3 -a
Ans. - - .
3
16. s 2xdx 1 . Ans. In (2z-l). 17. S cos x dx. Ans. .:!.
4
I
n
2
18. Ssin x dx. 2

Ans. T.
0
Evaluate the following integrals applying the indicated substitufionsl

n
n
s
2
19.
2
slnxcos xdx,cosx=t.Ans.! .20. S3 +:;osx, tan ~=t.Ans.
0
Vs"
4 I
xdx 3 Y2
21.
SY 2+4-17 , 2+4x=t
l
2
Ans. - 2- . 22. S(! ~~)' x=tant.
-l
23. sr:-5 --
1
dx, x-1c=l 2 Ans. 2 (2- arc tan 2).
11
24 s3
z Yz'+
dz
I
, Z=-. Am. In
1
X
3
2 .
25

s
2
cos q> dq> I

6-5slnq>+sin 2 q>' snq>=t.
3 0
4
Ans In
3 1 1
Prove that 26. S xm (1 -x)n dx = S.~n (l-x)m dx (m > 0, n> 0).

b b a a
27. Sf(x)dx=Sf<a+b-x)dx. 28. St(x 2 )dx=~ St(x 2 )dx.
a a -a
Evaluate the following improper integrals: 9

2.
s.y .
I
x dx
1-x 2
Am. I.
0
"'
30. Se-x dx. Ans. I. 31. s"' dx
a'+x 1
1t
Ans. 2a (a> 0). 32. sJIT=Xl"
.1-xz
I
dx Ans.
2
1t
33. s~~.
CD
Ans. +. I
34. SIn x dx. Ans. - l . 35. S x sin x dx. Ans. The inte-
a>
"' +a>
gral diverges. 36. S:-:

1
Ans. The integral diverges. 37.
_,
Sxz+~:+ 2 . Aru. :t.
s s sy
I 2 a>
3
31$. vdx . Ans. . 39. d: . Ans. The integral diverges. 40. dx .
X 2 X X ~~1
I) 0 1'
I a>
Ans. ~ . 41. S~:. Ans. The integral diverges. 42. Se-ax sin !lxdx (a> 0).
-1 0
.,
Ans. az ~ b'. 43. Se-ax cos bx dx (a> 0). Ans. a'~ bz.
0
I
Evaluate the following integrals approximately: 44. ln5=S ~x by the

1
trapezoidal rule and by Simpson's rule (n = 12). Ans. 1.6182 (by the trapezoidal
11
rule); 1.6098 (by Simpson's rule). 45. Sx 1

d.t by the trapezold1l rule and by
1
Exercises on Chapter XI 441
Simpson's rule (n=IO). Ans. 3690; 3660. 46. SJfl-x 3

dxbythetrapezoidal
0
a
rule (n=6). Ans. 0.8109. 47. J xdx
2 1
by Simpson's rule (n=4). Ans: 0.8111.
10
48. Slog nx dx
1 by the trapezoidal rule. and by Simpson's rule (n= 10).
A ns. 6 .0656 ; 6 .0896 . 49. E va Iua t e n f rom th e rea .

I t Jon 4n = sI
1
dx app Iymg
+x 2
.
n
2
Sx
slnx .
Simpson's rule (n = 10). Ans. 3.14159. 50. dx by S1mpson's rule (n = 10).
"'
Ans. 1.371. 51. EvaluateS e-x xn dx for integral n > 0 by proceeding from the
0
CID
equality Se-x dr=~ where a> 0. An~. nl 52. Proceeding from the equality
0 -
"' d.t
Sx+a= 2 y 0 ,evauae
n I t th . t
emegra
I s"' dx A n 135 ... (2n-1)
(xz+l)n+t' ns. 2 2nnl ,
0
"' 1-e-x
53. Evaluate the integral
S xe" dx.Ans.ln(l+a)(a>-1).54.Utilising
I I
the equality Sxn . J dx= ~ , compute the integral S xn-t (In x) dx.11
0
k kl
An.s. (-1) .-k-.
n +
CHAPTER XII
GEOMETRIC AND MECHANICAL APPLICATIONS

OF THE DEFINITE INTEGRAL
SEC. t. COMPUTING AREAS IN RECTANGULAR COORDINATES
If on the interval [a, b] the function f (x)-;;=: 0, then, as we

know from Sec. 2, Ch. XI, the area of a curvilinear trapezoid
bounded by the curve y = f (x), the x-axis, and the straight lines
x=a andx=b (Fig. 210) is
b
Q= S{(x) dx. (I)

a
b
Iff (x) ~ 0 on [a, b], then the definit~ i.ntegral \ f (x) dx is also~ 0:
"
a
It is equal, in absolute value, to the area Q corresponding to the
curvilinear trapezoid:
b
-Q~ ~ f(x)dx.
a
If f(x) changes sign on the interval [a, b] a finite number of

times, then we break up the integral throughout [a, bJ into the
!/
sum of integrals of the subintervals.
The integral will be positive on those
subintervals where f (x)-;;=: 0, and nega-
tive where f (x) ~ 0. The integral over
the entire interval will yield the differ-
ence of the areas above and below the
x x-axis (Fig. 228). To find the sur:n of the
Fig. 228. areas in the ordinary sense, one has
to find the sum of the absolute values
of the integrals over the above-indicated subintervals or compute
the integral
b
Q= ~ It (t> I dx.
a
Example t. Compute the area Q bounded by the sine curve y =sin x ahd
the x-axis, for 0 ~ x,.;:;; 2n (fig. 229).
Computing Areas in Rectangular Coordinates 443
Solution. Since sinx;;:;;.O when O..:;;;x,.;:;;;n and sin xo;;;;;O whenn <x.;;;;;2n,
we have
lt 2lt 2lt
Q= ~
0
sin x dx + I~ sin x d x I= ~ I sin x I dx,
lt 0
lt
~ sinx dx =-cos x j: =-(cos n-cos 0) = -( -l-1)=2,

0
2n
~ sinxdx=-cosxl:n =-(cos2n-cosn)=-2.
lt
Consequently, Q=2+1-21=4.
X
Fig. 229. Fig. 230.
If one needs to compute the area bounded by the curves y =f. (x),
1 y = fz (x) and the ordinates x =a, x = b, then provided f. (x);;::. f 2 (x)
we will obviously have (Fig. 230)
b b b
Q =~f. (x) dx- ~f. (x) dx = ~ (f. (x) -f 2 (x)] dx. (2)
a a a
Example 2. Compute the area bounded by the curves (Fig. 231)

u= Vx and u=x 2

Solution. Find the points of intersection of the curves:

Yx =x 2 , X=X\ whence x 1 =0, x 2 = l.
Therefore,
Now let us compute the area of the curvilinear trapezoid bounded

by a curve represented by equations in parametric form (Fig. 232):
X= q> (t) Y = 1jJ {t), (3)
where
444 Geometric and Mechanical Applications of the Definite Integral
and
(J)(a)=a, (J)(~)=b.
Let equations (3) define some function y=f(x) on the interval

[a, b] and, consequently, the area of the curvilinear trapezoid
y may be computed from the for-
mula
b b
Q = ~ f (x) dx = ~ y dx.
a a
0\ 1 X
'' '
' .................
Fig. 231. Ftg. 232.
Change the variable in this integral:

X=(j)(t); dX=(j)'(t)dt~
From (3) we have

y= f (x) = f[(J) (t)] = 'iJ (t).
Consequently,
li
Q = ~ "'(t) (j) (i) dt.
1
(4)
(l
This is the formula for computing the area of a curvilinear

trapezoid bounded by a curve represented parametrically.
Example 3. Compute the area of a region bouqded by the ell ipse
x=a cost, y= b sin t.
Solution. Compute the area of the upper half of the ellipse and double it.
Here, x varies from -a to +a, and so t varies between n and 0.
0 0 ll
Q= 2) (bslnt) (-asln tdt) =-2ab ) sin' tdt = 2ab ) sin' tdt =

n n o
n
= 2abSl-cos2t
2
dt-2ab [ t' .stn2']"
- 2--4- o =Jta
b
0
The Area of a Curuilinear Sector in Polar Coordinates 445
Example 4. Compute the area bounded by the x-a:ds and an arc of the
cycloid
x=a (t-sln t), y=a (1-cos f)
Solution. The variation of x from 0 to 2n: a corresponds to the variation
ol I !rom 0 to 2n:.
From (4) we have
zn zn
Q=~ !l(1-cost)a(l-cost)dt=a2 ~ (1-cost)'dt=
0
=a
zn zn
~ dt-2~ costdt+~ cos 2 tdt;
zn
0
J
[
0 0 0
231 Ut 2Tt 21t
+~
2 1 0521
S dt=2n; S cos tdt=O; S cos tdt=S dt=n.
fl 0 0 0
We finally get
Q =a (2n: + n)=3na 2

SEC. 2. THE AREA OF A CURVILINEAR SECTOR

IN POLAR COORDINATES
Suppose in a polar coordinate system we have a curve given

by the equation
Q={(O),
where {(0) is a continuous function when a~6~~-
Let us determine the area of the sector 0 AB bounded by the
curve Q={(IJ) and by the radius vectors O=a and 0=~.
Divide the given area by radius vectors 00 =a, 0 = OP ... , 6n = ~
into n parts. Denote by_ M,, M 2 , , Mn the angles between
the radius vectors that we have drawn (Fig. 233).
Denote by~ the length of a radius vector corresponding to some
angle 61 between 61 _ 1 and 61
Let us consider the circular sector with radius Q; and central angle
801 Its area will be
12
~Qj = 2 Q/ ~01.
The sum
n n
Qn = ~ ~ Q: M; = { ~ [f (01W Ml
. t'='l t .. l
will yield the area of the "step-like" sector.

4-16 Geometric and Mechanical Applications of the Definite Integral
Since this sum is an integral sum of the function Q1 = [f (0)] 1

on the interval a,..;;;O~~. its limit, as max 6.61 -..0, is the defi-
nite integral
It is not dependent on which radius vector Q; we take inside the
Fig. 233. Fig. 234.
angle LlO;. It is natural to consider this limit the sought-for area

of the figure*).
Thus, the area of the sector 0 AB is
+S
Q==
a
li
2
Q dO (I)
=+
or li
Q S [f (OW dO. (1')
a
Ex amp Ie. Compute the area bounded by the lemniscate
Q=a Vcos 26.
(Fig. 234).
Solution. The radius vector will describe a fourth of the sought-for area
il e varies._,between 0 and ' : :
Hence
Q=a.
*) It might be shown that this determination of the area does not contradict
that given earlier. In other words, if one computes the area of a curvilinear
sector by means of curvilinear trapezoids, the result will be the same.
The Arc Length of a Cur.ue 447
SEC. 3. THE ARC LENGTH OF A CURVE
1. The arc length of a curve in rectangular coordinates.

Let a curve be given by the equation y = f (x) in rectangular
coordinates in a plane.
Let us find the length of the arc AB of this curve between
the vertical straight lines x=a and x=b (Fig. 235).
The definition of the length of an arc was given in Chapter VI,
Sec. I. Let us recall that definition. On an arc AB take points A, M,,
M 2 , , M 1 , 8 with abscissas X 0 =a, x,, x 1 , , x 1, , b = xn
and draw the chords AM,, M,M 2 , ,
Mn- B whose lengths we shall denote by !J
ASP AS 2 , , AS,., respectively. This
gives the broken line AM,M1 M 11 _,B
inscribed in the arc AB. The length of
the broken line is
0 OX, X;-f X; X
1=1 Xz
The length, s, of the arc AB is the Fig. 235.
limit which the length of the inscribed
broken line approaches when the length of its greatest segment
approaches zero:
s= lim L" As
max .&s~: .... o t-=-J
1 ( 1)
We shall now prove that if on the interval a so;;. x so;;. b the func
tion f(x) and its derivative f' (x) are continuous, then this limit
exists. At the same time we shall specify a technique for computing
the length of the arc.
Let us introduce the notation
Ay1 = f (x1) - f (x1 _ 1 ).
-
Then
As1 = V (Ax1) 2
+(Ay1) =
2
y"t + ( !~:Y Ax 1
By Lagrange's theorem we have

Mt = f (x,)-f (xt_,) =
/lX! x1-x 1_ 1
f' (sI ),
where
I
Hence,
l!:.s; = V1 + [t' (s;W l!:.x;.
Thus, the length of an inscribed . broken line is
n
i=t
It is given that f' (x) is continuous; hence, the function V 1 + [f'(x)l"

is also continuous. Therefore, this integral sum has a limit that
is equal to a definite integral:
n b
s= lim _L.V1 + lf' (;W l!:.x; = SV1 + [t' (xW dx.

max 6x, """i"O i=t a
We thus have a formula for computing the arc length:

b . b
s= SV 1 + [f' (x)]"
a
dx =
a
SV 1 + (:~)".dx. (2)
Note. Using this formula, it is possible to obtain the derivative

of the arc length with respect to the abscissa. If we consider the
upper limit of integration as variable and denote it by x (we
shall not change the variable of integration), then the arc length
s will be a function of x:
s <x) = S-,11 + ( :~)" dx.

a
Differentiating this integral with respect to the upper limit, we
obtain
~= .. /1+(dy)2 (3)
dx V dx
This formula was derived in ~ec. 1, Ch. VI, on certain other
assumptions.
Example 1. Determine the circumference of \he circle
' x +y=r
2 2 . . '
Solution. First compute the length of a fourth part of the circumference

lying in the first quadrant. Then the equation of the arc AB will be
y= Y r-x,
whence
The Arc Length of a Curve 449
Consequently,
r + ~ dx = sr v r sin~r 1o, = r
1
s = sr .. 11 dx = r arc :rt .
4 r - X rZ _ x2 2
0 0
The length of the circumference is s = 2rcr.
Let us now find the arc length of a curve when the equation
of the curve is represented in parametric form:
X= <p (t), y = 'ljJ (t) (a.:;;:; t.:;;:; ~), (4)
where <p (t) and 'ljJ (t) are continuous functions with continuous de-
rivatives, and <p (t) does not vanish in the given interval. In this
case, equations (4) define a function y=f(x) which is continuous
and has a continuous derivative:
dy 'lj!' (/)
dx = <p' (/)
Let a= <p (a), b =If(~). Then substituting in the integral (2)
X= <p (t),
dx=<p' (t)dt,
we have
1:1,
S=
5.V. ~ 1 + ['lj!' (/)]
a
<p' (t)
2 1
<p (t) dt,
or, finally,
II
s= )V<p' (t) + 'IIJ' (t)
2 2
dt. (5)
a
Note 2. It may be proved that formula (5) holds also for curves
that are crossed by vertical lines in more than one point (in
particular, for closed curves), provided that both derivatives <p' (t)
and 'ljl' (t) are continuous at all points of the curve.
Example 2. Compute the length of the hypocycloid (astroid):
x=acost, y=asint.
Solution. Since the curve is symmetric about both coordinate axes, we shall
first compute the length of a fourth part oL it located in the first quadrant.
We find
:t = - 3a cos t sin t,
'!it= 3a sin t cos t.

2
15-3388
450 Geometr-ic and Mechanlca-1 Applicatlons of the Definite Integral
The parameter t will vary from. 0 to ~ . Hence

n n
I I
! S
s=
0
2
Y9a cos t sin t+9a"sin t cos t dt=3a 2
SY cos t sin
0
1
t dt =
= 3a
2
S . sin t
sw t cos t dt = 3a - 2-
2
Iz = n
Ja
; s = 6a.
0
2
0
Note 3. If a space curve is represented by the parametric equations

x~cp(t), y=\jl(t), z=x(t) (6)
where <i~t~~ (see Sec. I, Ch. IX), then the length of its arc
is defined (in the same way as for a plane arc) as the limit which
the length of an inscribed broken line approaches when the length
of the greatest segment approaches zero. If the functions cp (t),
'ljl (t), and X (t) are continuous and have continuous derivatives on
the interval [a, ~]. then the curve has a definite length (that is,
it . has the above-mentioned limit) which is computed from the
formula
ll .
s= SV [cp' (tW +['I'' (tW +(X' (t)y dt. (7)
a_
This result we accept without proof~
Example 3. Compute the arc length of the helix
x=acost, y=asint, z=amt
as t varies from 0 to 2:rt.
Solution. From the given equations we have
dx=--.,.a sin t dt, dy =a cost dt, dz =am dt.
Substituting into formula (7), we have
:n n
s = ~ Va sin 1 t+a 1 cos 2 t +a'm' dt=a ~ Y! +m' dt =2:rta Y I +m 2

0 0
2. The arc length of a curve in polar coordinates. Given (in

polar coordinates) the equation of the curve
Q={(S) (8)
where Q is the radius vector and e is the vectorial (polar) angle.
The Arc Length of a Curve 451
Let us write the formulas -for passing

from polar coordinates to Cartesian
coordinates:
X=QcosO, Y=QsinO.
If in place of Q we put its expression X
(8) in terms of 0, we get the equations
x=f(O)cosO, y=f(O)sinO.
These equations may be regarded as
the parametric equations of the curve Fig. 236.
and we can apply formula (5) for com
puting the arc length. To do this, find the derivatives of x and y
with respect to the parameter 0:
:~ = f' (0) cos e-f (0) sin 0;
:~ = f' (0) sin 0 +f (0) cos9.

Then
Hence,
~
S=) VQ' 2
+Q" dO.
&.
Example 4. Find the length of the cardioid

Q = a (l +cos 8)
(Fig. 236).
Varying the vectorial angle e from 0 to n:. we gef half the soughf-for
length. Here, Q' =-a sin 8. Hence,
n
s=2 ~ Va(! +cosll) 2 +a"sin 2 1ld8=
0
n
=2a) V2+2 cos s de=
0
" e e n
=4a ) cos 2 dO= 8a sin 2 jo =Ba.
:.: 0
15*
Example 5. Compute the length of the ellipse

x=a co.s t, } 0.;;;;; t..;;;; 2n:,
y= b Sin l,
assuming that a> b.
Solution. We take advantage of formula (5), first computing { the arc
len~th; that is, the length of the arc that corresponds to a variation of the
parameter from t = 0 to t =!:... :
2
n
2
~ =S Ya 2
sin 2 t+b 2 cos 2 tdl=
n n
2 . I
= S Va (I-eos t)+b 2
cos t dt = S
0 0
n
0
v....,.a~b 2 cos 2 tdt=a s
0
where k = y~ < I. -Hence,

a
n
2
s=4a ~ Yi-k 2 cos 2tdt.

0
The only thing that remains is to compute the last integral. But we know
that it is not expressible by elementary functions (see Sec. 16, Ch. X). This
integral can be computed only by approximation methods (by Simpson's rule,
for example).
For instance, if the semi-major axis of an ellipse is equal to 5 and the
semi-minor axis is 4, then k = ~ ! and the circumference of the ellipse is
n
Computing this integral by Simpson's rule (by dividing the interval

into lour parts) we get an approximate value of the integral;
[o. ~ J
s vi-i
2
2
COli t dt::::::: 1.298,
0
and so the. leng'th of the arc of the entire ellipse, is approximately equal to
.s:::::::25.96 units of length.
Computing the Volume of a Solid from the Areas of Parallel Sections -453
SEC. 4. COMPUTING THE VOLUME Of A SOLID FROM THE AREAS

OF PARALLEL SECTIONS (VOLUMES BY SLICING)
Q = Q (x).
We assume that Q (x) is a continuous
function of x and calculate the volume
of the body .
. Draw the planes x =a, x = x., x = x,,
... , X=Xn=b. Fig. 237._
These_ planes will cut the solid up
into layers (slices).
In each subinterval x 1_ 1 .:;;:; x.:;;:; x 1 we choose an arbitrary point
~~ and for each value i =I, 2, ... , n we construct a cylindrical
bo.dy, the generatrix of which is parallel to the x-axis, while the.
directrix is the boundary of the slice of the solid T made by the
plane X=~~
The volume of such an elementary cylinder, the area of the
base of which is
and t~e altitude !J.x1, is
The volume of all the cylinders will be

n
un =
l
L Q (~;) ~x 1
=I
The limit of this sum as max 11x 1 - + 0 (if it exists) is the

volume of the given solid:
n
v= lim L Q (~ 1) 11x
max t!.x1 .... o i = 1
1
Since vn is obviously the integral sum of the continuous function

Q (x) on the interval a.:;;:; x.:;;:; b, the indicated limit exists and is
expressed by the definite integral
b
u= ~ Q (x) dx. (I)

a
454 Geometric and Mechanical Applications of the .Definite Integral
Example. Compute the volume of the triaxial ellipsoid (Fi"g. 238).
y.
Fig. 238.
Solution. In a section of the ellipsoid made by a plane parallel to the

yz-plane and at a distance x from it, we have the ellipse
y2 z2 x
fji" +Ci"= 1-ii"
or
with semi-axes
But the area of such an ellipse is :tb 1c1 (see Example 3, Sec. 1).
Therefore,
Q (x)=:tbc ( 1- ;: ) .
The volume of the ellipsoid will be

a
v =:tbc S(1- ~:) dx =:tbc ( x-::.) I~a=~ :tabc.

-a
In the particular case, a= b =c, the ellipsoid turns into a sph'ere, and we
have
The Volume of a Solid of Revolution 455
SEC. S. THE VOLUME OF A SOLID OF REVOLUTION
Let us consider a solid generated by the revolution, about the

x-axis, of a curvilinear trapezoid a.4Bb bounded by the curve
y = f (x), the x-axis, and the y
lines x=a, x=b.
In this case, an arbitrary
section of the solid made by a
plane perpendicular to the x-axis
is a circle of area
Q = 1tY1 = 1t [f (x) ]2
Applying the general formula X
for computing volume [{1), Sec.
4], we get a formula for calcu
lating the volume of a solid of
re:volution:
b b
v= 1t ~ y 1 dx = 1t ~ [f (xW dx. Fig. 239.

a a
Example. Find the volume of a solid generated by the revolution of the
catenary
y= ~ ( e~ +e-: )
about the x-axis on the interval from x =0 to x = b (Fig. 239).
Solution.
b X X b !X 2t
v=n ~
2
5(ea- +e--ar dx=nt 5(e-a +2+e --u) dx=

0 0
zx x 2b b
na[a-2 ea+2x-2a e -a-]b =s
=4 na (e-a-e--a)+"a b
-2-. 0
3 2
SEC. 6. THE SURFACE OF A SOLID OF REVOLUTION
Suppose we have a surface generated by the revolution of a

curve y = f (x) about the x-axis. Let us determine the area of this
surface on the interval ao;;;x.s;;;.b. We take the function {(x) to
be continuous and to have a continuous derivative at all points
of the interval [a, b].
As in Sec. 3, draw the chords AM 1 , M1 M1 , , Mn_ 1 B, whose
lengths are denoted by 8s 1 , 8s 1 , , 8sn (Fig. 240).
456 Geometric and Mechanical Applicatlons of the Definlte Integral
Each chord of length A.s 1 (i =I, 2, ... , n) describes (in -the

process of revolution) a truncated . cone whose surface A.P 1 is
A.P I= 2n Yi-12+ Yi A.s,.
But
A.s1=-( A.x~+ A.y;= y I+ (!~:r A.x{.

Applying Lagrange's theorem, we get
M; = f(x;)-f(x;_,) === f' (s)
/:1Xj X;-X(-l I
where
hence,
!1st= VI + f'z (st) A.x i

Fig. 240.
A.P{ = 2n Yi-12+ YiVI + rz (s;) A.x,.
The surface described by the broken line will be equal to the sum
n
p n = 2n L Yi-t2+Yr lfi + f'z (s;) A.x;
i= J
or the sum
n
Pn= n~ [f(x;_J+
i=l
(I)
extended to all segments of the broken. line. The limit of this
sum, when the largest segment A.s 1 approaches zero is called the
area of the surface of revolution under consideration. The sum (l)
is not the integral sum of the [unction
2nf (x) JII + f' (.tV, (2)
because the term corresponding to the interval [x 1_" X;) involves
several points of this interval X;_" x 1, S; But it is possible to
prove that the limit of the sum (I) is equal to the limit of the
Computing Work by the Definite Integral 457
integral sum of function (2); that is,

n
P= limn ~ [f(xi_ 1 )+f(x,)]VI+f'(sj) 1 LlX1 =
max <lXj-+ o i =1
n
= limn
max <lXl-+ o i=t
~ 2f (~ 1 ) VI +f' (61) 2
ilx;
or
IJ
P = 2n ~ f (x) VI+ {' 2

(x) dx. (3)
a
Example. Determine the .surface of a paraboloid generated by revolution

about the x-axis of an arc of the parabola y 2 =2px, which corresponds to the
variation of x from x =0 to x =a:
r2- , Y2p Jfl +u'z= .. /1 + 2p= .. /2x+p

y= r px' y =2 yx ' V- 4x V 2x
Solution. By (3) we have
P=21t s a
Jf2px V x~p 2
dx=21t riJSY2x+p d
a
x=
I=
0 0
=2n .r-
r p 2 (2x+p> 'I I a 2n YiJ [(2a+p)'I -p 'I ).
3 2 0 3
SEC. 7. COMPUTING WORK BY THE DEFINITE INTEGRAL
Suppose a material point M is moving in a straight line Os

under a force F, and the direction of the force coincides with
the direction of motion. It is required to find the work performed
by the force F as the point M is moved from s =a to s =b.
I) If the force F is constant, tben the work A is expressed
by the product of the force F by the path length:
A=F(b-a).
2) Let us assume that the force F is constantly varying, depend-
ing on the position of the material point; that is to say, it is
a function F (s) continuous on the interval a,;;;;; s ~b.
Divide the interval [a, b] into n arbitrary parts of length
Lls 1 , Lls 1 , , Llsn,
then in each subinterval [s;-p s;] choose an arbitrary point ;i
and replace the work of the force F (s) along the path
4'58 Geometric' and Mechanical A'pp/'icattons of the Definite Integral
/).si (i= I, 2, ... , n) by the product

F(i)f).si.
This means that within the limits of each subinterval we take
the force F to be constant: we assume F = F (i ). Here, the ex-
pression F (i) /).si will yield an -approximate value of the work
done by the force F over the path /).si (for a sufficiently small
/).si), and the sum
n
An= ~ F (i) /).si
i=l
will be the approximate expression of the work of the force F

over the interval [a, b].
Obviously, An is an integral .sum of the function F = F (s} on
the interval [a, b]. The limit of this sum as max (f).si)-+ 0 exists
and expresses the . work of the force F (s)
-.--~---F-t--,-.- over the path frem s= a to s= b:
I b
.~ 1; A = ~ F (s) ds. (I)

~'ti a
~ ~ Example 1. The compression S of a helical
~~
~.!::; spring is proportional to the applied force F.
Compute the work of the force F when the spring
.~ is compressed 5 em, if a force of one kilogram
~~~~ is required to compress it 1 em (Fig. 241).
Fig. 241. Solution. 11 is given that the force F and the
distance covered S are connected by the relation
F = kS, where k is a constant.
Let us express S in metres and F in kilograms. When 8=0.01, F= 1,
that is, 1 =k0.01, whence k= 100, F= 1008.
By (1) we have
0.05
szl o.os =0.125 kilogram-metre .
A=
5 lOOS dS= 1002
0
Example 2. The force F with which an electric charge e1 repulses another

charge e2 (of the same sign) at a distance of r is expressed by the formula
F=ke,ez
r ,
I
where k is a constant.
Determine the work done by a force F in moving the charge e1 from the
point A, (at a distance of r1 fr.om e1 ) to A 1 (at a distance of r2 from e1)
assuming that e1 is located at the point A 0 as the origin.
Solution. From formula (1) we have
Coordinates of the Centre of Gravity .
When r 2 = oo, we have
A= ske~ dr=
"'
r
ke,e 2
r.

'
When e2 =!, A= k ~. This quantity is called the potential of the field
r
generated by the charge e1
SEC. 8. COORDINATES OF THE CENTRE OF GRAVITY
Suppose on an xy-plane we have a system of material points

P 1 (x .. Y1 ); P 2 (X 2 , Y2 ), .. ,Pn(xn, Yn)
with masses m 1 , m 2 , , mn-
The products X;m; and Y;m; are called the static moments of
the mass m; relative to the y- and x-axes.
We denote by xc and Yc the coordinates of the centre of gravity
of the given system. Then, as we know from mechanics, the
coordinates of the centre of gravity of this material system will be
defined by the formulas
(1)
(2)
We shall use these formulas in finding the centres of gravity of

various figures and solids.
1. The centre of gravity of a plane line. Let there be a curve
AB given by the equation y=f(x), a:o;;;;;,x:o;;;;;,b, and let this curve
be a material line.
Let the linear density*) of such a material curve be 'Y Divide
the line into n parts of length /).s., /).s 2 , , /).sn. The masses of
these parts will be equal to the product of their lengths by the
(constant) density: /).m; = yf).s;. On each part of the arc /).si take
*) Linear density is the mass of unit length of a given line. We assume

that the linear density is the same in all portions of the curve.
an arbitrary point with abscissa ~i Now representing each part

of the arc lls; by ihe material point P; [~;. f (~;)] with mass ylls;
and substituting into (1) and (2) ~i in place of X;, f (~;) in place
of Y;. and the value of ylls; (the mass of the parts lls;) in place
of m;, we obtain approximate formulas for determining the centre
of gravity of ihe arc:
If the function y = f (x) is continuous and has a continuous deri-

vative, the sums in the numerator and denominator of each frac-
tion have, as max lls;-+ 0, limits equal to the limits of the cor-
responding integral sums. Thus, the coordinates of the centre of
gravity of. the arc are expressed by definite integrals:
b b
~xds ~xVi+t' 2 (x)dx
a a
XC= -b- =~b----- (I')
~ ds ~ Vi +I'' (x) dx
a a
b b
~ f (x) ds ~ /(x) Vi+f''(x)dx

a a
Yc=-b:--- (2')
~ ds JVI +f''(x) dx
b '
a a
Example I. Find the coordinates of the centre of gravity of the semi-circle

x + y= a situated above the x-axis.
Solution. Determine the abscissa of the centre of gravity:
a
x dx
\
~
a ,"r 2 --
~a r a -x -a Va'-x'l':a 0
X- = =-=0.
Ia
a s
c- a . X na
dx aarcsm-
a -a
Va-x
-a
Coordinates of the Centre of Grauity 461
Find the ordinate of .the centre of gravity:

a
SJfa-x y a-x
-a
a dx
Yc= na
2. The centre of gravity of a plane figure. Given a figure
bounded by the lines y = f 1 (x), y =f. (x), x =a, x = b, which is a
material plane figure. We con-
sider constant the surface Y
density, which is the mass
of unit area of the surface.
It is equal to 6 for all parts
of the figure.
Divide the given figure by.
straight lines_x=a, x=xl' .. ,
X=Xn=b into strips of width
t!xl' flx ... , flxn. The mass
of each strip will be equal to
the product of its area by
the density 6. If each strip
is replaced by a rectangle Fig. 242.
(Fig. 242) with base flx;
and altitude f. (~;)-f 1 (~;), where ~i = x;-2
+xi, then the mass of a
strip will be approximately equal to
8mi=6[f.(~i)-f 1 (~i)]/1xi (i=l, 2, ... ,n).
The centre of gravity of this strip will be situated approxi-
mately in the centre of the appropriate rectangle:
(xi)c=~i; (yi)c=fdsil~fdsil.
Now replacing each strip by a material point, whose mass is
equal to the mass of the corresponding strip and is concentrated
at the centre of gravity of this strip, we find the approximate
value of. the coordinates of the centre of gravity of the entire
figure [by formulas (I) and (2)]:
x Rl ~ sic5 !f. (sil-tt (silJ11x.-
c ~ c5 <s;)-tdsi)J/1x; u.
2
I
L. u. (s1l + t~ <1)] c5 tf. <s;l- t~ (s;)J/',x;
y Rl ,.., .
C ~c5[f 1 (E;)-f 1 (E;))/1Xj
Passing to the limit as L\xi->-0, we obtain the exact coordinates

of the centre of gravity of the given figure:
b b
s X (f2 (X)-/ 1 (X)] dx

a _ _ _ _ __
XC= =,.b
; Slf. (x)+ f 1
(x)) If. (x)-/dx)) dx
Yc =---'a=----:-.:---------
.b
S[/ 2 (x)-/1 (x)l dx S

If. (x)-/1 (x)] dx
a a
. These formulas hold for any homo-

!J2=aJ{ geneous (that is, having constant
density at all points) plane figure.
We see that the coordinates of the
centre of gravity are independent of
--=~~~7.!:---~-x the density ~ of the figure (~ was
cancelled out in the process of com-
putation).
Example 2. Determine the coordinates of
the centre of gravity of a segment of the
Fig. 243. parabola. y=ax cut off by the straight line
. . x=a (Ftg. 243).
Solu !ion. In this case f 2 (x) = Jf ax, / 1 (x) = - Vax; therefore
a
2 Sx )!'aX dx -22 .t-t

r ax a I 4
0 5 - --5-a- - a
3
Xc=
2 sa )faX dx
-
- 2 v- 2 .f,,a- !_ - 5 ,
a 3 x o 3 a
0
Yc=O (since the segment is symmetric about the x-axis).
Exercises on Chapter XII

Computing Areas
1. Find the area of a figure bounded by the lines y 2 = 9x, y = 3x. Ans. i.
2. Find the area of a figure bounded by the equilateral hyperbola xy = a 2 ,
!hex-axis, and the lines x=a, b=2a. Ans. a In 2.
3. Find the area of a figure lying between the curve y=4-x 2 and the
x-axis. Ans. 10{ .
2 2 2
4. Find the area of a figure bounded by the hypocycloid x 3 +uT=aT.
3
An~. B na.
Exercises on Chapter XII 463
X X
. 5. Find the area of a figure bounded by the catenary y= ; (e a +e -a),

2
the x-axis, the y-axis, and the straight line x=a. Ans. ~ (e 2 -l).
6. Find the area of a figure bounded by the curve y=x 1 , the line y=8,
and the y-axis. Ans. 12.
7. Find the area of a region bounded by one loop of a sine wave and the
x-axis. Ans. 2.
8. Find the area. of a region lying between the parabolas y 2 = 2px, x 2 = 2py.
4 2
A ns. aP.
9. Find the total area of a figure boundedbytbelines y=x, y=2x,y=x.
3
Ans. .
2
10. Find the area of a region bounded by one arc of the eye! oid x =a (t- sin t),
y =a (I-eos t) and the x-axis. Ans. 3na 2
II. Find the area of a figure bounded by the hypocycloid x =a cos 1 t, y =
3
=a sin 1 t. Ans. B na 2
12. Find the area of the entire region bounded by the lemniscate Q2 =a 2 cos 2cp.
Ans. a 2 .
13: Compute the area of a region bounded by one loop of the curve Q=a sin 2cp.
>1 .
Ans. ''F na 2
14. Compute the total area of a region bounded by the cardioid Q=a (1- cos cp).
A ns. 3 na,2
2 I
15. Find the area of the region bounded by the curve. Q =a cos cp. Ans. n: .
16. Find the area of the region bounded by the curve Q =a cos 2cp.
na 2
Ans.
2 .
17. Find the area of the region bounded by the curve Q =cos 3cp. Ans. : .
na 1
18. Find the area of the region bounded by the curve Q=a cos 4cp. Ans.
2 .
Computing Volumes
x y'
\9. The ellipse iii' +p= I revolves about the x-axis. Find the volume of
the solid of revolution. Ans. ~ nab 2

20. The segment of a line connecting the origin with the point (a, b) re-
volves about the y-axis. Find the volume of the resulting cone. Ans. { na 1 b.
21. Find the volume of a torus generated by the revolulion of the circle
x 2 +(y-b) 1 =a 1 about the x-axis (it Is assumed that b ;;;;.a). Ans. 2n 2a 1 b.
22. The area bounded by the lines y=2px and x=a revolves about the
x-axis. Find the volume of the solid of revolution. Ans. npa 1
464 Geometric and Mechanical Applications of- the Definite 1ntegral
2 2 2
23. A figure bounded by the hypocycloid x T +Y3 =aT is revolved about
3
the x-axis. Find the volume of the solid of revolution. Ans. ~;;'.
24. A figure bounded by one arc of the sine wave y =sin x and the x-axis
is revolved about the x-axis, Find the volume of the solid of revolution.
:rt2
Arts.
2 .
25. A figure bounded by the parabola y 2 =4x and the straight line x=4 is
revolved about the x-axis. Find the volume of the solid of revolution. Ans. 32:rt.
26. A figure bounded by the curve y = xex and the straight I ines y =0, x =I,
is revolved about the x-axis. Find the volume of the solid of revolution.
Ans. ~ (e 2 -l).
27. A figure bounded by one arc of a cycloid x =a (t -sin t), y =a (I-eos I)
and the x-axis is revolved about the x-axis. Find the volume of the solid of
revolution. Ans. 5:rt 2a 3
28. The same figure ~s in Problem 27 is revolved about the y-.axis. Find
the volume of the solid of revolution. Ans. 6:rt 3a 8
29. The same figure as in Problem 27 is revolved about a straight line that
is parallel to the y-axis and passes through the vertex of a cycloid. Find the
.. ' " :rt 3
volume of the solid of revolution. Arts. : (9:rt 2 -16).
30. The same figure as in Problem 27 is revolved about a straight line pa-
rallel to the .x-'axis and passing through the vertex of a cycloid. Find the vo-
lume of the solid of revolution. Arts. 7:rt 2a 1
31. A cylinder of radius R is cut by a plane that passes through the dia-
meter of the base at an .angle u to the plane of the base. Find the volume of
2
the cut-off part. Ans.
3 R" tan u.
32. Find a volume that is common to the two cylinders: x 2 +y 2 =R 1 , g 1 +
16 8
+z 2 =R 2 Ans. R
3
33. The point of intersection of the diagonals of a square Is in motion along
the diameter of a circle of radius a; the plane in which the square lies remains
perpendicular to the plane of the circle, while the two opposite vertices of the
square move along the circle (as a result of this motion, the size of the
square obviously varies). Find the volume of the solid genrrated by this moving
square. A ns. 8 a3 .
3
34. Compute the volume of a segment cut off the elliptical paraboloid
yz zz .~
p+2q=x by the plane x=a. Ans. :rta 2 r pq.
2
35. Compute the volume of a solid bounded by the planes z=O, y=O, the
cylindrical surfaces x 2 =2py and z 2 =2px and the plane x,;,a. Ans. a

~~
7r p
(in first octant). .
36. A straight line is in motion parallel to the yz-plane, and cuts two el-
xz yz x z2 .
lipses az+
b':=l . az+C"=l lying in the xy- andxz-planes. Compute the vo-
8
lume of the solid thus obtained. Arts. abc.
3
Exercises on Chapter XII 4~.
Computing. Arc Lengths

.
2 2 2
37. Find the entire length of the hypocycloid x 3 + .113 ~aT. A ns. 6a.
38. Compute the arc length of the semicubical parabola ay 2 =x3 from the
335
t o a pomt
ongm WI'th a bsc1ssa
. x = 5a. Ans. 2f a.
X X
a - --
39. Find the arc length of the catenary y = (e a +e a) from the origin
2
X X
a - -- --
to the point (x, y). Ans. (ea -e a)= Vu-a.
2
40. Find the length of one arc of the cycloid x =a (!-sin t), y =a (1-cos t).
Ans. 8a.
41. Find the length of an arc of the curve y =In x within the limits from
,r- ,r-
x = r 3 to x = r 8. Ans. I
I
+23
In 2 .
42. Find the arc length of the curve y=i-lncosx between x=O and
1t 31t
x=-:r. Ans. In tan .
8
43. Find the length of the spiral of Archimedes Q =aq> from the pole to the
end of the first loop. Ans. 1ta Yi +4n 2
+ ~ In (21t + Yl -f-41t J. 2
44. Find the length of the spiral Q =ea'f' from the pole to the point (Q,cp).
Ans. ~
a
ea'l'=_g_ Yl -f-a
a
2
'q>
45. Find the entire length of the curve Q =a ~in . Ans. 23 na.
3 2
46. Find the length of the evolute of the ellipse x =-ac cos' t, c2
y =b sin' t.
4 (a'- b')
Ans. ab .
47. Find the length of the cardioid Q=a.(l+cosq>). Ans. Ba.
48. Find the arc length of the involute of the circle x =a (cos q> -f-q> sin <p),
y =a (sin q> -q> cos q>) from q>= 0 to (jl =q> 1 . Ans. ~ aq>~.
Compu ling Areas of Surfaces of Solids of Revolution
49. Find the area of a surface obtained by revolving the parabola y 2 =4ax
56 2
about the x-axis, from the origin 0 to a point with abscissa x = 3a. Ans. na
3
50. Find the area of the surface of a cone generated by the revolution of
a line segment y=2x from x=O to x=2: a) About the x-axis. Ans:-'81t VS:
b) About the y-axis. Ans. 4n V5.
51. Find the area of the surface of a torus obtained by revolving the circle
x-t-(y-b) 2 =a 2 about the x-axis. Ans. 4n 2 ab.
52. Find the area of the surface of a solid generated by revolving a car-
dioid about the x-axis. The cardioid is represented by the parametric equations
1 8
x=a (2 cos p-cos 2q>), y=a(2 sinq>-sln 2q>). Ans. ~ na 2
466 Geometric and Mechlinlcal Applications of the Definite Integral
53. Find the area of the surface of a solid obtained by revolving one arc
64na .
of a cycloid x=a(t-sint), y=a(1-cost) about the x-axis. Ans. - - .
3
54. The arc of a cycloid (see Problem 53) is revolved about the y-axis.
Find the surface of the solid of revolution. Ans. 16rc a
1 1
55. The arc of a cycloid (see Problem 53) is revolved about a tangent line
parallel to the x-axis and passing through the vertex. Find the surface of the
32
so I 1'd of revo I u t'ton. A ns. - na- .
3
56. The astroid x=a sin' t, y=a cos' t is revolved about the x-axis. Find
12
the surface of the solid of revolution. Ans. ~a'.
57. An arc of the sine wave y=sinx fromx=Otox=2:t is revolved about
the x-axis. Find the surface of the solid of revolution. Ans. 4n [Jf2+
+In (Jf2+1)].
58. The ellipse:.+
!
tz =
2
1 {a> b) revolves about the x-axis. Find the sur-
. . arcsine
face of the solid of revoluhon. Ans. 2rcb 2 +2nab-- , wheree=
Va 2
-b 1
.
c a
Various Applications of the Definite Integral

q9. Fine! the centre. of gravity of the area of one-fourth of the ellipse.
1
x'
ar + yb= 1 (x;;;;,: 0, 4a 4b
y;;;;,: 0). Ans. 3n' 3n.
60. Find the centre of gravity of the area of a figure bounded by the pa-
rabola x 2 +4y-16=0 and the x-axis. Ans. ( 0, : l.
61. Find the centre of gravity of the volume of a hemisphere. Ans. On the
axis of symmetry at a distance : R from the base.
62. Find the centre of gravity of the surface of a hemisphere. Ans. On the
axis of symmetry at a distance ~ from the base.
63. Find the centre of gravity of the surface of a circular right cone, the
radius of the base of which is R and the altitude h. Ans. On the axis of sym-
metry at a distance ~ from the base.
64. The figure is bounded by the lines y= sin x (0..;;; x..;;; n), y = 0. Find the
centre of gravity of the area of this figure. Ans. ( ~ , ~ ) .
65. Find the centre of gravity of the area of a figure bounded by ll'e pa-
rabolas y 1 =20x, x2 =20y. Ans. (9, 9). ,_
66. Find the centre of gr'avity of the area of a circular sector with central
angle .2a .and radius R. Ans. On the axis of symmetry at a distance ~ R sl:a
from the vertex of the sector.
67. Find the pressure of water on a rectangle vertically submerged In wa-
ter at a depth of 5m if It is known that the base is 8 metres, the altitude, 12
metres, and the upper base .is paralle.l to the free surface of the water. Ans.
1,056 m. '
Exercises on Chapter XU 467
, 68. The upper. edge of a canal lock has the shape of a square with a side
of 8 m lying on the surface of the water. Determine the pressure on each part
of the lock formed by dividing the square by one of its diagonals.
Ans. 85,333.33 kg, 170,666.67 kg.
69. Compute the work needed to pump the water out of a hemispherical
vessel of diameter 20 metres. Ans. 2.5x 106n kg-m.
70. A body is in rec-tilinear motion according to the law X=cl 3, where X
is the path length traversed in time 't, c=const. The resistance of the medium
is proportional to the square of the velocity, and k is the constant of pro-
portionality. Find the work done by the resistance when the body moves from
27 3/-
the point x=O to the point x=a. Ans. k y c2a7
7
71. Compute the work that has to be done in order to pump a liquid of
density y from a reservoir having the shape of a cone with vertex pointing
down, altitude H and radius of base R. Ans. - - - .
nyR'H'
12
72. A wooden float of cylindrical shape whose basal areaS= 4,000 em and
altitude H =50 em is floating on the surface of the water. What work must be
done to pull the float up to the surface? (Specific weight of the wood, 0.8).
Ans. - -
vHs = 32
kg-m.
2
73. Compute the force with which the water presses on a dam in the form
of an equilateral trapezoid (upper base a= 6.4 m, lower base b = 4.2 m, alti-
tude H=3 m). Ans. 22.2 m.
74: Find the axial component P kg of total pressure of steam on the sphe-
rical bottom of a boiler. The diameter of the cylindrical part of the boiler is
D mm, the pressure of the steam in the boiler is P kgfcm. Ans. P = n:o~
75. The end of a vertical shaft of radius r is supported by a flat thrus~
bearing. The weight of the shaft P is distributed equally over the entire sur-
face of the support. Compute the total work of friction in one rotation of the
~haft. Coefficient of friction is fl. Ans. ! 1tJLPr.
76. A vertical shaft ends in a thrust pin having the shape of a truncated
cone. The specific pressure of the pin on the thrust bearing is canst ant and
equal to P. The upper diameter of the pin is D, the lower, d, and the angle
at the vertex of the cone is 2a. Coefficient of friction, fl.
Find the work of friction for one rotation of the shaft. Ans.
77. A prismatic rod of length I is slowly extended by
6 stn a (D -d
n ~fl. 3 0
a force increasing
).
from 0 to P so that at each moment the tensile force is balanced by the for-
ces of elasticity of the rod. Compute the work A expended by the force on
tension, assuming that the tension occurred within the limits of elasticity. F
is ,the cross-sectional area of the rod, and E is the modulus of elasticity of
the material.
Hint. If x is the elongation of the rod and f is the corresponding force,
then f= F7 x. The elongation due to the force P is equal to 111 = :~ .
pIll P 21
Af1~ A=y=2EF'
78. A prismatic beam Is suspended vertically and a tensile force P is ap-
plied to its lower end. Compute the elongation of the beam due to the force
of its weight and to the force P if it is given that the original length. of the
468 Geometric and Mechanical Applications of the Definite I nlegral
beam is l, the cross-sectional area F, the weight Q and the modulus of elas-
ticity of the material E. Ans. M = (Q t2J)
1
.
79. Determine the time during which a liquid will flow out of a prismatic
vessel filled to a height H. The cross-sectional area of the vessel is F. the
area of the aperture f, the exit velocity is computed from the formula
v= 11 Y2gh, where J.t is the coefficient of viscosity, g is the acceleration of
gravity, and h i3 the distance from the aperture to the level of the liquid.
Ans. -T= 2FH _!__ --. /2H.
1-1-f y2gH 1-1-f V g
80. Determine the discharge Q (the quantity of water flowing in unit time)
over a spillway of rectangular cross section. Height of spillway, h, width, !J.
Ans. Q =
2 .r-
!J-bh r 2gh.
3
81. Determine the discharge of water Q flowing from a side rectangular
opening of height a and width IJ, if the height of the open surface of the wa-
y-
2
3 3
ter above the lower side of the opening is H. Ans. Q = 2bJ.t g [H 2 -(H -a)2].
3
CHAPTER X/II
DIFFERENTIAL EQUATIONS
SEC. 1. STATEMENT OF THE PROBLEM.

THE EQUATION OF MOTION OF A BODY WITH RESISTANC
OF THE MEDIUM PROPORTIONAL TO THE VELOCITY. THE EQUATION
OF A CATENARY
Let 1.he function y = f (x) reflect the quantitative aspect of some

phenomenon. Frequently, it is not possible to establish directly the-
1.ype of dependence of .y on x, but it is possible to give the rela-
tionship between x and .y and the derivatives of y with respect to
x: y', y", .. , y<n>. That is, we are able to write a differential
equation.
From the relationship established between the variable x, y and
the derivatives it is required to determine the direct dependence
of y on x; that is, to find y = f (x) or, as we say, to integrate the
differential equation.
Let us consider two examples.
Example l. A body of mass m is dropped from some height. It is required
to establish that law according to which the velocity v will vary as the body
falls, if, in addition to the force of gravity, the body is acted upon by the
decelerating force of the ai-r, which is proportional to the velocity (with con-
stant of proportionality k); in other words, it is required to find v = f (t).
Solution. By Newton's second law
dv
mdt=F
where~~ is the acceleration of a moving body (the derivative of the velocity

with respect to time) and F is the force acting on the body in the direction
of motion. This force is the resultant of two forces: the force of
gravity mg and the force of air resistance, -kv, which has the minus sign be-
cause it is in the opposite direction to that of the velocity. And so we have
dv
mdt=mg-kv. (1)
This relation connects the unknown function v and its derivative ~~, which
is a differential equation in the unknown function v. To solve the differen
tial equation is to find a function v=f (t) such that identically satisfies the
given differential equation. There is an infinitude of such functions. The stu-
dent can easily verify that any function of the form
k
--1 ma
v= Ce m +-f (2)
satisfies equation (I) no matter what the constant C is. Which one of these
470 Differential Equations
functions yields the sought-for dependence of v on t? To find it we take ad-

vantage of a supplementary condition: when the body was dropped it was im-
parted an initial velocity v 0 (which may be zero as a particular case); we as-
sume this initial velocity to be known. But then the unknown function v =
= f (I) must be such that when t = 0 (when motion begins) the condition v = v0
is fulfilled. Substituting 1=0, V=v 0 into formula (2), we find
mg
v 0 =C+k,
whence
mg
C=vo-k
Thus, the constant C is found, and the sought-for dependence of v on I is
kl
_ ( _ mg) e-;;; + mg (2')
V- Vo k k
It will be noted that if k=O (the air resistance is absent or negligibly
small so that we can disregard it), then we have a result familiar from phys-
ics*):
y (2")
T
This function satisfies the differential equalion
(I) and the initial condition: v==v 0 when 1=0.
Example 2. A flexible homogeneous thread
is suspended at two ends. Find the equation
of the curve that it describes under~own
weight (it is the same as any suspended ropes,
wires, chains, as for instance the cater pi liar track
of a tank between two supporting rollers).
Solution. Let M 0 (0, b) be the lowest point
of the thread, and M an arbitrary point
(Fig. 244). Let us consider a part of the thread,
M 0 M. This part is in equilibrium, the resultant
of three forces:
0- X I) the tension T, acting along the tangent
Fig. 244.
to the point M and forming an angle cp with
the x-axis;
2) the tension H at M0 acting horizontally;
3) the weight of the thread ys acting vertically downwards, where sis the
length of the arc M 0 M and 'Y is the linear specific weight of the thread.
Breaking up the tension T into horizontal and vertical components, we get
the equations of equilibrium:
Tcoscp=H, Tslncp=ys.
Dividing the terms of the second equation by the corresponding terms of
the first, we obtain
'Y s.
tan cp= H (3)
) Formula (2") can be obtained from (2') by passing to the limill

kl
kl~ [( v-:g)e-;;;+:g] =v +gt,
0 0
Statement of the Problem 471
Now suppose that the equation of the sought-for curve may be written in
the form y = f (x). Here, f (x) is an unknown function that has to be found.
It will be noted that
tan !p= {' (x) = :~.
Hence,
dy 1
dx=as (4)
where the ratio .!!__ is denoted in terms of a.

Differentiate b~th sides of (4) with respect to x:
d'y 1 ds
dx 2 =a dx (5)
But, as we know (see Sec. 1, Ch. VI),
ds= , / 1+(dy)".
dx Jl dx
Substituting this expression into equation (5), we get the differential equa-
tion of the sought-for curve:
2
d y =_!_ , I I+ (dy)'. (6)
dx 2 a Jl dx
It expresses the relationship between the first and 'second derivatives of
the unknown function y.
Without going into the methods of solving the equations, we shall note
that any function of the form
(7)
satisfies equation (6) for any values that C 1 and C2 may assume. This is evi-
dent if we put the first and second derivatives of the given function into (6).
We shall indicate, without proof, that these functions (for different C 1 and
C2 ) exhaust all possible solutions of equation (6).
The graphs of all -the functions thus obtained are called catenaries.
Let us now find out how one should choose the constants C 1 and C2 so as
to obtain precisely that catenary whose lowest point M has coordinates (0, b).
Since for x=O the point of the catenary occupies the lowest possible position,
the tangent here is horizontal, ~~ =0. Also, it is given that at this point
the ordinate is equal to b, y= b. /
From (7) we find /
y'='[
I ( ~+C, -e - (~+c))
ea a .
Putting x=O here, we obtain O=~(ec'-e-c'). Hence, C 1 =0.

II the ordinate of the point M 0 is b, ihen y= b when x=O.
~72 . Di{jerential Equations
From equation (7) we get b= ~ (l+l)+C1 , assuming x=O and C 1 =0,

whence C2 =b-a. Finally we have
y=; (e~ +e-~)+b-a.

Equation (7) assumes a very simple form if we take the ordinate of M 0 equal
to a. Then the equation of the ca~enary is
y=i (e~ +e-~).

SEC. 2. DEFINiTIONS
Definition J. A di{jerential equation is one which connects an
independent variable, x, an unknown function, y = f (x), and its
derivatives y', y", ... , yn.
Symbolically,
.
a differential equation
' .
may be written as follows:
F(x, y, y', y", .. , y'm)=O
or
dy d 2y dny)
F Y' ( X' ' ' dxn = O
dx ' dx 2
If the sought-for function y = f (x) is a function of one indepen-
dent variable, then the differential equation is called ordinartJ.
We shall deal only with ordinary differential equations*).
Definition 2. The order of a differential equation is the order
of the highest derivative which appears.
For example, the equation
y'- 2xy 2 + 5 = 0
is an equation of the first order.
*) In addition to ordinary differential equations, mathematical analysis
makes a .study of partial di{jerential equations. Such an equation is a relation
between an unknown function z (that is, dependent upon two or several
variables x, y, ... ), these variables x, y, .. , and the partial derivatives
az
of z: ax , ayaz ' d'z
ax.
etc.
The following is an example of a partial differential equation with
unknown function z (x, y):
az az
X a;=Y dy.
It is easy to verify that this equa lion is satisfied by the function z = X 2 1/2
(and also by a multitude of other functions).
In this course we shall l1ave little to do with partial differential equa.
tions.
First-Order Differential Equations 473
The equation
y" + ky'- by-sinx= 0
is an equation of the second order, etc.
The equation considered in the preceding section in Example 1
is an equation of the first order, in Example 2, one of the second
md~.
Definition 3. The solution or integral of a differential equation
is any function y = f (x), which, when put into the equation,
converts it into an identity.
Example 1. Let there be an equation
d'y
dx'+y=O.
The functions y=sinx, y=2cosx, y=3slnx-cosx and, in general,
functions of the form y = C, sin x, y = C2 cos x
or
y=C 1 sinx+C 2 cosx
are solutions of the given equaHon for any choice of constants C1 and C2 ;
this i~ evident if we put these functions into the equation.
Example 2. Let us consider the equation
y'x-x 2 -y=0.
Its solutions are air" functions of the form
y=x 2 +Cx
where C is any constant. Indeed, differentiating the functions y=x'+Cx, we
find
y'=2x+C.
Putting the expressions for y and y' into the initial equation, we get the
identity
Each of the equations considered in Examples I and 2 has an infinitude of

solutions.
SEC. 3. FIRST-ORDER DIFFERENTIAL EQUATIONS

(GENERAL NOTIONS)
1. A differential equation of the first order is of the form

F (x, y, y') = 0. (I)
If this equation can be solved fo~ y', it can be written in the form
y' =f(x, y). (I')
In this case we say that the differential equation is solved for
the derivative. For such an equation the following theorem, called
the theorem of the unique existence of solution of a differential
equation, holds.
Theorem. If in the equation

y' =f (x, y)
the function f (x, y) and its partial derivative with respect to y,
~~
vY
, are continuous in some region D . in an xy-plane containing
some point (x y.). then there is only one. solution to this equation
y = <p (x) which satisfies the condition x = x., y =Yo The geometric
meaning of the theorem consists in the fact that there exists one
and only one such function y = <p (x), the graph of which passes
through the point (x 0 , Y.).
It follows from this theorem that equation (1 ') has an infinitude
of various ~olutions [for example, a solution the graph of which
passes through (x 0 , Y.); another solution whose graph passes through
(x 0 , yJ; through (x y1 ), etc., provided these points lie in the
region D].
The condition that for x=x. the function y must be equal to
the given number Yo is called the initial condition. It is frequent-
ly written in the form
Yl x=xo = Yo
Definition 1. The general solution of a first-order differential
equation is the function
y=<p(x, C), (2)
which depends on a single arbitrary constant C and satisfies the
following conditions:
a) It satisfies the differential equation for any specific value of
the constant C.
b) No matter what the initial condition y=y 0 for x=x., that
is, (Y)x=x.=IJ 0 , it is possible to find a value C=C 0 such that the
function y=<p(x, C.) satisfies the given initial condition. It is
assumed here that the values x 0 and Yo belong to the range of
the variables x and y in which the conditions of the existence
theorem are fulfilled.
2. In searching for the general solution of a differential equation
we often arrive at a relation like
<D(x, y, C)= 0, (2')
which Is not solved for y. Solving this relationship for y, we get
the general solution. However, it is not always possible to express
, y from (2') in terms of elementary functions; in such cases, th~
general solution is left in implicit form. "
First-Order Differential Equations 475
An equation of the form ID(x, y, C)=O which gives an implicit

general solution is called the complete integral of the differential
equation.
Definition 2. A particular solution is any function y = <p (x, C.)
which is obtained from the general solution y = <p (x, C), if in the
latter we assign to 1.he arbitrary constant C a definite value C =c .
In this case, the relation <D (x, y, C 0 ) = 0 is called a particular
integral of the equation. -
Example 1. For the first-order equation
dy y
dx=--;
the general solution is a family of functions y=.E_; this can be checked by
X
simple substitution in the equation.
Let us find a particular solution that will satisfy the following initial
condition: y0 =l when X 0 =2.
Putting these values into the formula y=: , we have 1 = ~ or C=2.
Consequently, the function y=~

X
will be the part-icular solution we are
seeking.
From the geometric viewpoint, the general solution (complete

integral) is a family of curves in a coordinate plane, which family
depends on a single arbitrary constant C (or, as it is common
to say, on a single parameter C). These curves are called integral
curves of the given differential equation. A particular integral is
associated with one curve of this family that passes through a
certain given point of the plane.
Thus, in the latter example, the complete integral is geometri-
cally depicted by a family of hyperbolas y=.E_ x
while the partic-
ular integral defined by the given initial condition is depicted
by one of these hyperbolas passing through the point M. (2, I).
Fig. 245 shows the curves of a family that are associated with
certain values of the parameter: C =}, =
C 1, C 2, C = =-
1, etc.
To make the reasoning still more pictorial, we shall from now
.on say that not only the function y = <p (x, C 0 ) that satisfies the
equation but also the associated integral curve is a solution of
the equation. We will therefore speak of a solution passing through
the point (x., y.).
Note. The equation ddy = _}!_ has no solution passing through a
X X
point lying on the y-axis (see Fig. 245). This is because the right
side of the equation is not defined for x = 0 and consequently is

not continuous. . .
To solve (or as we frequently say, to integrate) a differential
equation means:
a) to find its general solution or complete integral (if the initial
conditions are not specified) or
b) to find a particular so'lution of the equation tqat will satisfy
the given initial conditions (if such exist).
fig. 245.
3. Let us now give a geometric interpretation of a first-order

differential equation.
Let there be a differential equation solved for the derivative
dy
dx=f(x, y) (I')
and let y = <p (x, C) be the general solution of this equation. This
general solution determines the family of integral curves in the
xy-plane.
For each point M with coordinates x and y, . equation (I')
defines the value of the derivative ~, or the slope of the tangent
line to the integral curve passing through this point. Thus, the
differential equation (I') yields a collection of directions or, as
we say, defines a direction-fieLd in the xy-plane.
First-Order Differential Equatlons 477
Consequently, from the geometric point of view, the problem

of integrating a differential equation consists in finding the curves,
Fig. 246.
the direction of the tangents to which coincides with the direc-

tion-field at the corresponding points.
Fig. 246 shows a direction-field defined by the differential
equation
dy y
ax=--x
4. Let us now consider the following problem.
Let there be given a family of functions that depends on a
single parameter C:
y=cp(x, C), (2)
and let only one curve of this family pass through each point of
the plane (or some region in the plane).
For what differential equation is this family of functions a com
plete integral?
From relation (2), differentiating with respect to x, we find
dy
dx = cp_.' (x,
c ). (3)
Since only one curve of the famtly passes through each point
of the plane, for every number pair x and y, a unique value of
Differential Equations
C is determined from equation (2). Putting this value of C into

(3) we find ~~ as a function of x and y. This is what yields
y the differential equation that is
satisfied by every function of
the family (2).
Hence, to establish a rela-
tionship between .x, y and dfx ,_
that is, to write a differential
equation whose general solution
(complete integral) is given by
X formula (2), one has to eli-
minate C from relations (2)
and (3).
Example 2. Find the differential
Fig. 247. equation of the family of parabolas
y = Cx (Fig. 247).
Differentiating the equation of the family with respect to x, we get
dy
dx=2Cx.
Putting the value C=~

x
into this ~quation from the equation of the fam-
ily, we obtain a differentiable equation of the given family:
dy 2y
dx=x
This dillerential equation is meaningful when x;CO; which Is to say, in
any region not containing points on the y-axis.
SEC. 4. EQUATIONS WITH SEPARATED AND SEPARABLE

VARIABLES. THE PROBLEM OF THE DlSINTEGRATION
OF RADIUM
Let us consider a differential equation of the form

dy
dx = f 1 (x) fa (y), (l)
where the right side is a product of a function dependent only

on x by a function dependent only on y. We transform it in the
following manner assuming that f. (y) =f= 0:
t.~>dy={1 (x)dx. (1')
Considering y a known function of x, equation (1') may be, regard-
ed as the equality of two differentials, while the indefinite
Equations wlth Separated and Separable Variables 479
integrals of them will differ by a constant term. Integrating the

left side with respect to y and the right with respect to x, we
obtain y
Sfz ~y) dy= S{ (x) dx+C
1
which is a relationship connecting

the solution of y, the independent
variable x, and an arbitrary con- X
stant C; we have thus obtained a
general solution (complete integral)
of equation (1).
I. A type (1') differential equa-
tion
Fig. 248.
M (x) dx+N (y) dy= 0 (2)
is called an equation with separated variables. From what has
been p~oved, its complete integral is
~ M (x)dx+ ~ N(y)dy==C.
Example I. Given an equation with separated variablesl
xdx+y dy=O.
Integrating we get the general solution:
x y
2+T=Ct.
Since the left side of this equation is nonnegative, the right side Is also
nonnegative. Denoting 2C 1 in terms of c, we will have
x2 +y2 =C 2
This is the equation of a family of concentric circles (Fig. 248) with
centre at the coordinate origin and radius C.
2. An equation of the form

M1 (x) N1 (y) dx +M 1 (x) N. (y) dy = 0 (3)
is called an equation with variables separable. It can be reduced*)
to an equation with separated variables by dividing both sides
by the expression N 1 (y) M 1 (x):
M 1 (x) N 1 (y) dx +M 1 (x) N 1 (1/) d _ O
Ndy)M 1 (x) N 1 (y)M 1 (x) Y-
'*) These transformations are permissible only In a region where neither

N1 (y). nor M1 (x) vanish.
or
M dx) d
M 2 (x) X
+ NN (y)
2
1
d =
(y) y
O
'
that is, to an equation like (2).
Example 2. Given the equation
dy y
dx=-x-
Separating variables, we have
dy dx
-
y
=--
X
Integrating we find
s sd: = - dxx +C,

which is
In 1y I= -In 1xI+ In I C I*) or In I Y I= In I~ I;

whence we get the general solution: y=.
X
(I +x) ydx+(l-y)x dy=O.
Separating variables we have
(I;x)dx+I yydy=O; (~+I) dx+(;-I)dy=O.
Integrating we obtain
In!xl+x+lnlyl-y=C or lnlxul+x-y~C.
This relation is the complete integral of the given equation.
Example 4. It is known that the decay rate of radium is directly propor-
tional to its quantity at each given instant. Find the law of variation of a
mass of radium as a function of the time if at t=O the mass of radium was m9
The decay rate is determined as follows. Let there be mass m at time r,
+ +
and mass m !!,.m at time t M. During M mass !!,.m decays. The ratio
!7' is the mean rate of decay. The limit of this ratio as M ---+0
. !!,.m dm
Il i D - = -
lJ.t-+oM dt
is the rate of decay of radium at time t.
) Having in view subsequent transformations, we denoted the arbitrary

constant by ln 1C 1. which is permissible since In I C I (when C;CO) can take
on any value from -co to +co.
Equations with Separaied and Separable Variables 481
It is given that
dm
dt=-km, (4)
where k is the constant of proportionality (k>O). We use the minus sign

because the mass of radium diminishes with increasing time and therefore
dm
H~- m
Equation (4) is an equation with
variables separable. Let us separate the
variables:
dm
-=-kdt.
m
Solving the equation we obtain
In m=- kt-lnC Fig. 249.
whence
m .
Jnc=-kt,
m=Ce -kt. (5)

Since at t=O the mass of.,. radium
' .
was .m.0 , C must satisfy the relationship
-k
. . m 0 ~Ce . =C.
Putting the value of C into (5) we get the desired mass of radium as a fun-
ction of time (Fig. 249):
-kt
m=m~ (~
The constant k is determined from observations as follows. During time t 0
Jet a% of the original mass of radium decay. Hence, the following relation
ship is fulfilled:
whence
-kt0 =ln (1-~
100)
\
or
k =-_!__In
t0 ( !-~)
100
Thus, it has been determined that for radium k = 0.00044 (the unif of
measure of time is on.e year).
Putting this value of k into (6) we obtain
-o.ooout
m=m 0e
Let us find the radium half-life, which is the interval of time during
which ltalf of the original mass of radium decays. Putting ~ 0 in place of m
16- 3388
in the Jat~er formula, we get an equation for determining the half-life T1 .
whence
-0.00044T=-In 2
or
In 2
T=o.ooo 44 =1.590 years.
Note. The simplest differential equation with separated variables is one

of the form
~~=f(x) or dy=f(x)dx.
Its com p Jete integral is of the form
y:: ~ f (x) dx+C.

We dealt with the solution of equations of this kind in Ch. X.
SEC. 5. HOMOGENEOUS FIRST-ORDER EQUATIONS
Definition 1. The function f (x, y) is called a homogeneous

function of degree n in the variables x and y, if for any A. the
following identity is true:
f(A.x, A.y)=A.nf(x, y).
Example t. The function f(x, y)= V x +y
3 3 is a homogeneous function
of degree one, since
f (Ax, "-Y) = V (A.x) + (Ay) ="-V x + y = "-! (x,
3 3 3 3
y).
Example 2. f (x, y) =xy- y2 is a homogeneous function of degree two,

since (A.x) (A.y)-(A.y) 2 =A. 2 [xy-y"].
x-y
Example 3. f (x, y) = - - is a homogeneous function of zero degree,
xy
. (Ax) 2 -(Ay) 2 x-y
stnce (A.x) (A.y) xy , that is, f (Ax, "-Yl = f (x, y) or f (AX, Ay) =
="- f (x,
0
y).
Definition 2. An equation of the first order

~~=f (x, y) (1)
is called homogeneous in x and y if the function f (x, y) is a ho-

mogeneous function of zero degree in x and y.
Homogeneous First-Order Equations 483
Solution of a homogeneous equation. It is given that f (A.x, A.y) =

= f (x, y). Putting A.=__!_ in this identity, we have
X
f (x, y) = f ( 1, f) .
Thus, a homogeneous function of zero degree is dependent only
on the ratio of the arguments.
In this case, equation (1) takes the form
:~ =f ( 1' ~ ) . (1 ')
Making the substitution
y
U= x' or y=UX,
we get
dy du
dx= u + dxx.
Putting this expression of the derivative into equation (1 '), we
obtain
dx= f (1, u).
u+x du
This is an equation with variables separable:
du du dx
xdx=f(l, u)-u or f(I,u)-u=x
Integrating we find
du =S dx +C
S /(1, u)-u x
Putting the ratio !L

X
in place of u after integration, we get
the integral of equation (1').
dy xy
dx= x 2 -y 2
On the right is a zero-degree homogeneous function, which means that wo

have a homogeneous equation. Making the substitution .!L=u
X
we have
dy du
y=ux; dx=u+xdx;
du u du u1
u + x dx = l-u 2 ; x dx = 1- u2 '
16'"
Separating variables we obtain

(l-u')du
11 1
dx
x
_!__
( u3 u
.!_) du= dx.
x '
Whence, integrating, we find
I I
- 211 ,-lnluJ=InJxJ+lnJCJ or - 2 u=lnJuxCJ.
Substituting u= 1L,
X
we get the general solution of the original equation:
.
x
- 2y =In ICy I
It is impossible here to get y as an explicit function of x in terms of _ele-
mentary functions. Incidentally, it is very easy to express x in terms of y:
x=yY -2C In I CyJ.
Note. An equation of the type

M (x, y) dx + N (x, y) dy = 0
will be homogeneous if, and only if, M (x, y) and N(x, y)are
homogeneous functions of the same degree. This follows from the
fact that the ratio of two homogeneous functions of the same de-
gree is a homogeneous function of degree zero.
Example 5. The equations
(Zx+3y) dx+(x-2y) dy=O,
(x'+ y') dx-2xy dy=O
are homogeneous.
SEC. 6. EQUATIONS REDUCIBLE TO HOMOGENEOUS

EQUATIONS
Equations of the following type are reducible to homogeneous

equations:
dy ax+ by+c
(1)
dx = a 1x + b y +c
1 1
If C1 = c = 0, then equation (1) ~s obviously homogeneous. Now let

c and c1 (or one of them) be dtfferent from zero. Change the va-
riables:
Then
dy dy, (2)
d.~= dx1
Equations Reducible to Homogeneous Equations 485
Putting into (2) the expressions x, y, and~, we obtain

dy 1 ax,+ by 1 +(1h +bk+c
(3)
dx 1 = a1x 1 + b,y, +a1 h+ b1 k+c 1
Choose h and k so that the following equalities are fulfilled:

ah+bk+c=O,} (4)
a,h+ b1 k +c1 =0.
In other words, define h and k as solutions of a system of equa
tions (4). Equation (3) then becomes homogeneous:
dy 1 ax 1 + by 1
dx 1 = a 1x 1 +b,y,
Solving this equation and passing once again to x and y by
formulas (2), we obtain the solution of equation (1).
The system (4) has no solution if
J~. t,J=o,
i. e., ab,=a,b. But if : = !'="-, that
is, a, ="-a, b,-;M, and,
hence, equation (1) may be transformed to
dy (ax+ by) +c
dX='A.(ax+by)+c, (5)
Then by substitution
z =ax+ by (6)
and the equation is reduced to one with variables &ep~raple.
Indeed,
whence
dy 1 dz a
dx=bdX-b (7)
Putting into (5) expressions (6) and (7), we get
I dz a z+c
bd:c-b=A.z+c,'
which is an equation with variables separable.
The device applied to integrating equation (1) is also applied
to the integration of the equation
dy f( ax+by+c )
dx = + b1y +c, '
a 1:c
where f is an arbitrary continuous function.
Example 1, -Given the equation

dy x+y-3
dx =x-y-1'
To convert it into a homogeneous equation, make the substitution x=x 1 +h;
y=y1 +k. Then
dy, x, y, +h +k-3 +
dx, x1 -y 1 +h-k-1
Solving the set of two equations
h+k-3=0; h-k-1=0,
we find
h=2, k=l.
As a result we get the homogeneous equation
dy, x, +y,
dx =xt-Yt'
which we solve by substitutiom
1b=u;
x.
!hen
dy 1 , du
y1 =ux1 ; -=u+x, - .
dx 1 dx 1
du l+u
u+x, d-=- -,
x 1 1-u
and we gef an equation with variables separablel
X
du
----
1+u
1 dx,-1-u '
Separating the variables, we have

1-u du=dx 1
I+u x, '
Integrating we find
I
arc tan u-
2 In (I +u 2)= In x, +InC,
arc tan u =In ( VT+'ilix, C)
or
Cx, JI'T+iii=earc tan u
Putting lL!. in place of u, we obtain
x,
./- arc tan~
Cy x~+u:=e x,,
Passing to the variables x and y, we finally get

arc tan y- 1
CJf(x-2) 2 +{y-1)2=e x-z,
First-Order Linear Equations 487

, 2x+ y-1
y = 4x+2y+5
cannot be solved by the substitution x=x1 +h, y=y 1 +k, since in this case
the set of equations that serves to determine h and k is insolvable (here, the
determinant I~ ~I of the coefficients of the variables is equal to zero).
This equation may be reduced to one with variables separable by the
substitution
2x+y=z.
Then y' = z'- 2 and the equation is reduced to the form
z-1
z'- 2 = 2z+5
or
, 5z+9
z =2z+5
Solving if we find
2 7
5z +
25 tnj5z+91~x+C.
Since z=2x+y, we obtain the final solution of the initial equation in the
form
2 7
5(2x+ul+~ In ll0x+5y+91 =x+C
or
!Oy-5x+7ln ll0x+Su+9l =C 1,
that is, as an implicit function y of x.
SEC. 7. FIRST-ORDER LINEAR EQUATIONS
Definition. A first-order linear equation is an equation that

is linear in the unknown function and its derivative. It is of the
form
~~ + P (x) y = Q (x), (I)
where P (x) and Q (x) are given continuous functions of x (or are
constants).
Solution of linear equation (t ). Let us seek the solution of
equation (I) in the form of a product of two functions of x:
y=u(x)v(x). (2)
One of these functions may be arbitrary, while the other will
be determined from equation (I).
Differentiating both sides of (2), we find
dy = U dv + v du.
dx dx dx
488 Differentia-l Equa~ions
Putting the expression obtained of the derivative into (1), we

have
dv du
udx+dxu+Puv=Q
or
u (:~+Pv )+u :~=Q. (3)
Let us choose the function u such that
dv
dx+Pv=O. (4)
Separating the variables in this differential equation in the func-
tion v, we find
dv
-=-P dX.
v
Integrating we obtain
-lnC 1 +lnv=-~ Pdx
or
Since for us it is sufficient to have some nonzero solution of

equation (4), we take,. as the function u (x),
v (x) =e- JP~x. (5)
where ~ P dx is some antiderivative. Obviously, v (x) ::po.

Putting the value of u (x) which we have found into (3), we
dv
get (noting that dx+Pv=O):
du
v (x) dx = Q (~),
or
whence
u= J~ (~l dx +C.
Substituting- into formula (2), we finally get
y = v (x) [J ~ g~ dx + C]
First-Order Linear Equations 489
or
y= v (x) S~ (~] dx+Cv (x). (6)
Note. It is obvious that expression (6) will not change if in
place of the function v (x) defined by (5) we take some function
v 1 (x)=Cv(x). Indeed, putting v, (x) in (6) in place of v(x), we
get
y = Cv (x) S!!
(x) dx = CCv (x).
Cv (x)
The C's in the first term cancel out; in the second term the product
CC is an arbitrary constant, which we shall denote by C, and we
-again arrive at expression (6). If we denote S~(~) dx = q> (x), then
expression (6) will take the form
y = v (x) q> (x) + Cv (x). (6')
H is obvious that this is a complete integral, since C may be
chosen in such manner that the initial condition will be fulfilled:
when X=X 0 , y=y 0
The value of C is determined from the equation
+
Yo v (X 0 ) q> (X 0 ) Cv (x0 ).
Example. Solve the equation
dy
- - -2- y=(x+ 1)s,
dx x-1-1
Solution. Putting
y=UV
we have
dy =Udv + du v
dx dx dx
Putting the expression :~ i~to the original equation, we obtain

dv du 2
u-+-v---uv=(x-1-1) 1
dx dx x+ l '
u(:~ - x! 1 v) + v~~ = (x + 1)'.

To determine v we get the equation
dv 2
dx-x+1 v=O,
that is,
dv 2dx
v=x+l'
whence
tnv=21n(x+l) or. V=(x+l) 2
Pulling the expression of the function v into equation (7), we get the
following equation for u1
(x + 1)2 du
dx = (x + 1)
1
or
du
dx = (x+ 1),
whence
U= (x+ I)" +C.
2
Thus, the complete integral of the given equation will be of the form
y=(x11)' +C(x+l)".
The family obtained is the general solution. No matter what the initial
condition (x 0 , y 0 ), where x 0 #- I, it is always possible to choose C so that
the corresponding particular solution should satisfy the given initial condi-
tion. For example, the particular solution that satisfies the condition y 0 =3
when x 0 = 0 is found as follows:
3 = (O t I)'+ C (0 + I )2 ;
Consequently, the desired particular solution is
y = (x 11 )' + } (x + I)".
However, if the initial condition (x 0 , y 0) is chosen so that x 0 = - I, we will

not find the particular solution that satisfies this condition. This is due to
the fact that when x 0 =-l the function P(x)=-x~l is discontinuous
and, hence, the conditions of the theorem of the existence of a solution are
not observed.
SEC. 8. BERNOULLI'S EQUATION
We consider an equation of the form *1

~~ + P (x) y = Q (x) yn, (1)
*1 This equation results from the problem of the motion of a body

provided the resistance of medium F depends on the velocity: F=1.. 1 v+1..1 v'.
The equation of motion will then assume the form m M = ...,.- 1.. 10 -1..1 vn or
dv + mA 1 V= -~ vn,
dt m
Bernoulli's Equation 491
where P (x) and Q (x) are continuous functions of x (or constants),

and n =F,-0 and n =1= 1 (otherwise we would have a linear equation).
This equation is called Bernoulli's equation and reduces to
a linear equation by the following transformation,
Dividing all terms of the equation by yn, we get
y-n ~~ + Py-n+J = Q, (2)
Making the substitution
we have
:;=(-n+l)y-n:;.
Substituting into (2), we get
dz
dx+<-n+ l)Pz=(-n+l)Q,
This is a linear equation.
Finding its complete integral and substituting the expression
y-n+J for z, we get the complete integral of th~ Bernoulli equation.
:~ +xy=xsys. (3)
Solution. Dividing all terms by y 3 , we have

y-y' +xy-2=xs. (4)
Introducing the new function
we get
dz __ y_ 3 dy
dx- 2 dx'
Substituting into equation (4), we obtain
dz
--2xz=-2x3 , (5)
dx
This is a linear equation.
Let us find its complete integral:
dz dv du
z=uv; dx=u dx + dx v.
Put expressions z and : : into (5):
dv du
u dx + dx v-2xuv=- 2x1
or
u (:~-2xv) +v ~~~- 2xa

Equate to zero the expression in the brackets:
dv dv
-d -2xv=0; -=2x dx;
X V
In v=x; v=ex,
For u we get the equation

du=- 2e-x x 3 dx, u=- 2 ~ e-x x 8 dx+C.
I nlegra ting by parts, we find
u =xe-x +e-x +C:
Z=Uv=x"+I+Ce-x.
Consequently, the complete integral of the given equation is
I
Y x"+ I +Ce-x'
Note. Just as was done for linear equations, it may be shown
that the solution of the Bernoulli equation may be sought in the
form of a product of two functions:
y= u (x) v (x),
where v (x) is some nonzero function that satisfies the equation
v' + Pv=O.
SEC. 9. EXACT Dl FFERENTIAL EQUATIONS
Definition. The equation

M (x, y) dx+ N (x, y) dy=O (1)
is called an exact differential equation if M (x, y) and N (x, y) are
continuous differentiable functions for which the following rela-
tionship is fulfilled
iJM iJN
(2)
iJM iJN
and -i)y and -ax are continuous in some region.
Integrating exact differential equations. We shall prove that if
the left side of equation (1) is an exact differenti-al, then condi-
Exact Differential Equations 493
tion (2) is fulfilled, and, conversely, if condition (2) is fulfilled

the left side of equation (1) is an exact differential of some fun-
ction u (x, y). That is, equation (1) is an equation of the form
du (x, y) = 0 (3)
and, consequently, its complete integral is

u(x, y)=C.
Let us first assume that the left side of (1). is an exact diffe-
rential of some function u (x, y); that is,
au au
~ dx +~ dy;
M (x, y) dx+N (x, y)dy=du = vX vy
then
M=~xu;
v
N-au
-ay (4)
Differentiating the first relationship with respect to y, and the

sec.ond with respect to x, we obtain
oM o2 u oN o2 u
ay = ax oy ; ax = ay ax
Assuming continuity of the second derivatives, we have

aM oN
ay=ax'
that is, (2) is a necessary condition for the left side of (I) to be
an exact differential of some function u (x, y). We 'shall show that
this condition is also sufficient: if (2) is fulfilled then the left
side of (I) is an exact differential of some function u (x, y).
From the relation
au
ax =M (x, y)
we find
X
u = ~ M (x, y) dx + cp (y),
Xo
where X0 Is the abscissa of any point of the domain of existence

of the solution.
When integrating with respect to x we consider y constant, a'nd
therefore the arbitrary constant of integration may be dependent
on y .. Let us choo~e a function cp (y) so that the second of the
relations (4) is fulfilled. To do this, we differentiate*) both sides

of the latter equation with respect to y and equate the result to
N (x, y):
iJu
iJy = s
X:
iJM
ay-dx+ q> , (y) = N (x, y);
X:o
but since a;:=~~ , we can write

X:
5~~ dx+ q>' (y) = N;

X:o
+
that is, N (x, y) 1~o q>' (y) = N (x, y)
or
N (x, y)-N (X 0 , y) + q>' (y) = N (x, y).
Hence,
q>' (y) = N (X 0 , y)
or
y
q>(y)= ~ N(x 0 , y)dy+C .

Yo
Thus, the function u (x, y) will have the form
X: y
U= ~ M(x, y)dx+ ~ N(x 0, y)dy+C

Xo Yo
Here P (x 0 , Yo) is a point in the neighbourhood of which there

is a solution of the differential equation (1).
Equating this expression to an arbitrary constant C, we get the
complete integral of equation (1 ):
X: y
~ M(x, y)dx+ ~ N(X 0 J y)dy=C. (5)

Yo
*) Jhe. integral ~ M (x, y) dx is dependent on y. To find the derivative of

X:o
this integral with respect to y, differentiate the integrand with respect to y:
X X
aya SM (x, y) dx= SoMiJy dx. This follows from . , t heorem for. d'ff
Lei bmz 1 eren-
xo Xo
tiating a definite inlegral with respect to a parameter (see Sec. 10; C:h. XI),
Integrating Factor 495
Example. Given the equation

2 2
2x d +y -3x d =O
y3 X y y
Let us check to see whether this is an exact differential equation.
Denoting
we have
For !I'# 0, condition (2) is fulfitled. Hence, the left side of this equation is
an exact differential of some unknown function u (x, y). Let us find this
function.
. au. 2x
S1nce a-=
X 3y , it fotlows that
2x x2
u=
5ya dx + cp (y) = a+ cp (y),
y
where cp (y) fs an as yet undefined function of y.
Differentiating this relation with respect toy and noting that
au y 2 -3x 2
au=N=-u-.-,
we find
3 2 3
x+ '() y - x .
-Y' cp Y =-u--,
hence
cp'(y)=; cp(y)=-~+C,,
x1 I
u (x, y)=-a--+C,.
y y
Thus the complete integral of the initial equation is
x I
---=C
ya y
SEC. 10. INTEGRATING FACTOR
Let the left side of the equation

M (x, y) dx + N (x, y) dy = 0 (1)
not be an exact differential. If is sometimes possible to choose
a function f.L.(x, y) such that after multiplying all terms of the
equation by it the left side of the equation is converted into an
exact differential. The general solution of the equation thus ob-
tained coincides with the general solution of the original equation;
the function !J (x, y) is called the integrating factor of equation (1).
In order to find the integrating factor f.t, do as follows. Mul-

tiply both sides of the given equation by the as yet unknown
integrating factor 11:
~-tM dx + ~-tN dy = 0.
For this equation to be an exact differential equation, it is neces-
sary and sufficient that the following relationship be fulfilled:
Ci(ftM) __ o(ftN).
-ay-(JX
that is,
oM
f1 ay + M Oft_ oN N Oft
ay - f1 ax + ax '
or
Maft_Naft_ (oN_oM)
oy ax - f1 ax oy .
After dividing both sides of the latter equation by f.t, we get
M o!n ft -N o!n ft =oN_ oM (2)
ay ax ax oy
It is obvious that any function 11 (x, y) that satisfies this equa-
tion is the integrating factor of equation (1). Equation (2) is
a partial differential equation in the unknown function 11 depen-
dent on the two variables x and y. It can be proved that under -
definite conditions it has an infinitude of solutions and that, con-
sequently, equation (I) has an integrating factor. But in the gene-
ral case, the problem of finding f1 (x, y) from equation (2) is har-
der than the original problem of integrating equation (1). Only
in certain particular cases does one manage to find the function
f1 (x, y).
For instance let equation (1) admit an integrating factor depen-
dent only on y. Then -
olnft_
iJx - 0
and to find 11 we obtain an ordinary differential equation

oN oM
oln ll. ax-a;
ay= M
from which we determine (by a single quadrature) ln f.t, and, hence,
f.l as well. It is clear that this may be done only if the expression
. aN aM
ax:Ty is not dependent on x.
'\ The Envelope of a Family of Curves 497
aN aM
Similarly, if the expression Tx~ay is not dependent on y but
only on x, then it is easy to find an integrating factor that
depends only on x.
(y +xy 2 ) dx-x dy=O.
Solution. Here, M=y+xy 2 ; N=-x;
aM aN aM aN
-a
y =l+2xy; -a
X
=-1; -a
y -a
X
.
Thus, the left side of the equation is not an exact differential. Let us see
whether this equation allows for an integrating factor dependent only on y or
not. Noting that
aN aM
-l-l-2xy 2
y+xy y
we conclude that the equation permits of an integrating !actor dependent
only on y. We find it:
whence
1
In J.1 = - 2ln y, i. e., J.1 = y.
After multiplying through by the integrating factor J.l., we obtain the equation
(.~Y +x) dx-!...u. dy=O

as an exact differential equation
.
(daM=
y
aaM
X
= - -;),
y
Solving this equation 1
we find its complete integral:
; +~+C=O,
or
2x
y=- x'+2C
SEC. 11. THE ENVELOPE OF A FAMILY OF CURVES
Let there be an equation of the form

ll>(x, y, C)=O, (I)
where x and y are variable Cartesian coordinates and C is a para
meter that can take on a variety of fixed values.
498 Di(Jereniial Equations
For each given value of the parameter C, equation (1) defines

some curve in the xy-plane. Assigning to C all possible values,
we obtain a family of curves dependent on a single parameter,
or using the more common term, a one-
y parameter family of curves. Thus,
equation (1) is the equation of a one
parameter family of curves (becaus.e it
contains only one arbitrary constant).
X
Fig. 250. Fig. 251.
Definition. The line L is called the envelope of a one-parameter

family of lines if at each point it touches some line of the family,
and different lines of the given family touch the line L at differ-
ent points (Fig. 250).
Example I. Consider the family of lines
(x-C)"+y"=R 1 ,
where R is a constant and C is a parameter.
This is a family of circles of radius R with centres on the x-axis. This
family will obviously have as envelopes the straight lines y=R and y=- R
(Fig. 251).
Finding the equation of the envelope of a given family. Let
there be given a family of curves,
<D(x,y,C)=O, (1)
that depend on the parameter C.
Let us assume that this family has an envelope whose equation
may be written in the form y = cp (x'), where cp (x) is a continuous
and differentiable function of x. We consider some point M (x, y)
lying on the envelope. This point also lies on some curve of the
family (1). To this curve there corresponds a definite value of the
parameter C, which value is determined from equation (1), for
given (x; y): C =;= C (x, y). Thus, for all points of the envelope the
following equality is fulfilled:
<D(x, y, C(x, y))=O. (2)
Suppose that C'(x~ y) is a differentiable function that is not con-
stant in any interval of the values of x andy under qmsideration.
The Envelope of a Family of Curves 499
From equation (2) of the envelope we find the slope of the tan-
gent to the envelope at the point M (x, y). Differentiate (2} with
respect to x considering that y is a function of x:
aa> + aa> ac + laa> + aa> ac J , = 0
ax ac ax ay ac ay Y
or
. , [acax+ acay Y'] = 0
<D.. + <DuY +<De (3)
The slope of the tangent to the curve of the family (1) at the
point M (x, y) is found from
<D: + <D~y' = 0 (4)
(on this curve, C is constant).
We assume that <l>'y =1= 0, otherwise we would consider x as the
function and y as the argument. Since the slope k of the envelope
is equal to the slope k of the curve of the family, from (3) and
(4) we obtain
ac + ac
<D'e [ Tx ay Y'] = 0
But since on the envelope C (x, y) =1= const, it follows that
~+!!}_t'=I=O
ax ay Y
and so for its points the following equation holds:
<D~ (x, y, C)= 0. (5)
Thus, the following two equations serve to determine the envelope:
<D(x, y, C)=O, }
'
<De (x, y, C)= 0.
(6)
Conversely, if, by eliminating C from these equations, we get an
equation y = cp (x), where cp (x) is a differentiable function, and
C =1= const on this curve, then y = cp (x) is the equation of the
envelope.
Note 1. If for the family (1) a certain function y=cp (x) is the
equation of the locus of singular points, that is, of points where
<D~ = 0 and <D~ = 0, then the coordinates of these points also satisfy
equations (6}. .
Indeed, the coordinates of singular points may be expressed
in terms of the parameter C that enters into equation (1):
X= A (C), y = fl {C). (7)
500 Di(Jereiitial Equations
If these expressions are substituted in equation (I), we get an

identity in C:
<D[f..(C), J.L(C), C]=O.
Differentiating this identity with respect to C, we obtain
' d'}.. I df.l. ' .
ll>x dC -+ <Dude+ <De = 0.
Since for any points the equalities := 0, <l>~= 0, are fulfilled,
it follows that for them the equality ~= 0 is also fulfilled.
We have thus proved that the coordinate of singular points
satisfy equations (6).
Summarising, equations (6) define either the envelope or the
locus of singular points of the curves of the family (1), or a
combination of both. Thus, after obtaining a curve that satisfies
equations (6), one has further to find out whether it is an envelope
or the locus of singular points.
Example 2. Find the envelope of the family of circles
(x-C) 2 +y 1 -R1 =0,
\hat are dependent on the single parameter C.
Solution. Differentiating the equation of the family with respect to C,
we gel
2 (x-C)=O.
Eliminating C from these two equations, we obtain the equation
y 2 -R 2 =0 or y= R.
It is clear, by geometric reasoning, that the pair of straight lines is the
envelope (and not the locus of singular points, since the circles of a family
do not have singular points).
Example 3. Find the envelope of the family of straight lines
xcosa+ysina-p=O (a)
where a is a parameter.
Solution. Differentiating the given equation of the family with respect
to a, we have
-xslna+ycosa=O. (b)
To eliminate the parameter a from equations (a) and (b), multiply the
terms of the first by cos a, and of the second, by sin a, and then subtract
the second from the 'first; we wiH then have
x=p cos a.
Putting this expression into (b), we find
y=p sin a.
Squaring the terms of the two latter equations and adding termwise, we get
x~+yz=pz.
/
This is a circle. It is the envelope of the family (and not the' locus of singu-
lar points, since straight lines do not have singular points) (Fig. 252).
Example 4. Find the envelope of the trajectories of shells fired from a gun
Y with velocity v0 at different angles of incli-
nation of the barrel to the horizon. We shall
consider that the gun is located at the
Fig. 252. Fig. 253.
coordinate origin and that the trajectories of the sheJls He in the xy.plane
(air resistance is disregarded).
Solution. First find the equation of the trajectory of the shell for the case
when the barrel makes an angle a with the positive x-axis. In flight, the
shell participates simultaneously in two motions: a uniform motion with
velocity v0 in the direction of the barrel and a falling motion due to the
force of gravity. Therefore, at each instant of timet th'e position of the shell M
(Fig. 253) will be defined by the equations
X=v 0 tcosa,
gt2
y=v 0 t sina-y.
These are parametric equations of the trajectory (th.e parameter is the time t).
Eliminating t, we get the equation of the trajectory in .the form
'\;
Finally, introducing the notation tan a=k,

2~ 2 =a, we get
y=kx-ax 2 (1 +kZ). (8)
This equation defines a parabola with vertical axis passing through the origin
and with branches downwards. We obtain a variety of traject.ories for the
differertt values of k. Consequently, equation (8) is the e,q(lation of a one-
parameter family of parabolas, which arc the tnl.jeCt()rles. of a shell for
different angles a and for a given initial velocity v 0 (Fig. 254).
Let us find the envelope of this family of parabolas. Differentiating with
respect to. k both sides of (8), we have
x-2akx 2 =0. (9)
Eliminating k from equations (8) and (9), we get
i
1 2
y=.4a- ax'
502 Differential Equal ior;s
This is the equation of a parabola with vertex at the point ( 0,

axis of which coincides with the y-axis. It is not a locus of singular points
4~), the
[since parabolas (8) do n_ot have singular points]. Thus, the parabola
1
y=--ax2
4a
is the envelope of the family of trajectories. It is called a safety parabola
because no point outside it is in reach of a shell fired from a given gun with
a given initial velocity u 0
g
X
Fig. 254.
Example 5. Find the envelope of a family of semicubical parabolas

y 1 -(x-C} 2 =0.
Solution. Differentiate the given equation of the family with respect to
ihe parameter C:
2(x-C)=O.
Eliminating the parameter C from the two equations, we gef
y=O.
The x-axis is a locus of singular points-a cusp of the first kind (Fig. 255).
Indeed, let us find the singular points of the curve
y -(x-C)
1 2
=0
for a fixed value of C. Differentiating with respect to x and y, we find
F~=- 2(x-C)=0;
F'g =3y 2 =0.
Solving the three foregoing equations simultaneously, we find the coordi-
nates of th~ singular point: x=C, y=O; thus, each curve of the given
Y ~~~~I, has a singular point on the
For continuous variation of the
parameter C, the singular points will
fill the entire x-axis.
Example 6. Find the envelope and
~ locus of singular points of the family
Locus Qf smgu/ar points x 2
Fig. 255, (y-C)2-3(x-C)'=O. (10)
Solution. Differentiating both sides of (10) with respect to C, we find

-2(y-C)+i- 3 (x-C} 2 =0
or
(11)
Now eliminate the parameter C from (11) and from the equation (10) of the
family:
y-C=(x-C} 2
Putting the expression y-C into the equation of the family, we get
2
(x-C} 4 - (x-C)'=O
3
or
(x-C)' [<x-C)- ;J =0,
whence we obtain two possible values of C and two solutions of the problem
corresponding to them.
First Solution: Second Solution:
2
C=x, C=x--
3
and so from (11) we find and so from (11) we find
y-x-(x-x} 2 =0 y-x+ 32 - [ x-x+ 2
3
J =0
2
or or
2
y=x. y=x--g.
We have obtained two straight lines: y=x and y=x-f. The first is a locus
of singular points . the second is an envelope (Fig. 256).
Fig. 256. Fig. 257.

Note 2. In Sec. 7, Ch. VI, it was proved that the normal to a

curve serves as a tangent to its evolute. Hence, the family of
normals to a given curve is at the same time a family of tangents
to its evolute. Thus, the evolute of the curve is the envelope of
the family of normals of this curve (Fig. 257).
This remark enables us to point out another method for finding
evolutes: to obtain the equation of an evolute, first find the family
of all normals of the given curve and then find the envelope of
this family.
SEC. 12. SINGULAR SOLUTIONS OF A FIRST-ORDER

DIFFERENTIAL EQUATION
Let the differential equation
F(x,y,_:~)=o (1)
have a complete integral
<D (x, y, C)= 0. (2)
Let us assume that the family of integral curves that corresponds
to equation (2) has an envelope. We shall prove that this envelope
is also an integral curve of the differential equation (1).
Indeed, at each point the envelope touches some curve of the
family; that is, it has a common tangent with it. Thus, at each
common point the envelope and the curve of the family have the
same values of x, y, y'.
But for a curve of the family, the numbers x, y, and y' satisfy
equation (1 ). Consequently, the very same equation is satisfied by
the abscissa, the ordinate and the slope of each point of the
envelope. But this means that the envelope is an integral curve
and its equation is a solution . of the given differential equation.
Since, generally speaking, the envelope is not the curve of the
family, its equation cannot be obtained from the complete inte-
gral (2) for any particular value of C. The solution of the differential
equation which is not obtained from the complete integral for any
value of C and which has as its graph the envelope of a family
of integral curves entering into the general solution, is called a
singular solution of the differential equation.
Let the compl~te integral be_ known:
cD (x, __y, C)= 0;
eliminating C from this equation and from the equ~tio_n <D~ (x; y, C)= 0
we get 1jJ (x, y) = 0. If this function satisfies the differential equation
land does not belong to the family (2)], then it is a singular Integral.
Clairaut's Equation 505
It should be noted that at least two integral curves pass through

each point of the curve that describes a sin'gular solution; that is,
uniqueness of solution is violated at ear:h point of a singular
solution.
Example. Find a singular solution of the equation
y'(I +u')=R.
Solution. Let us find its complete integral. We solve the equation for y':
dy
""df= VRCI/
y .. <*l
Separating var_iables, we obtain
ydy
.r -dx.
r R 2 -Y 2
Whence, integrating, we find the complete integral:
(x-C) 2 +Y 2 =R.
It is easy to see that the family of integral lines is a family of circles of
radius R with centres on the x-axis. The pair of straight lines y = R will
be the envelope of the family of curves.
The functions y = R satisfy the differential .equation (1). This, conse
qu~ntly, is a singular integral.
SEC. 13. CLAIRAUT'S EQUATION
Let us consider the so-called Clairaut equation:
y= X ~~ + ~ ( ~~ ) . (1)
It is integrated by introducing an auxiliary parameter. Put ~~ = p;

then equation (1) will take the form
y=xp+~(p). (1')
Differentiate, with respect to x, all the terms of this ~quation,
bearing in mind that p = ~~ is a function of x:
dp I dp
p=xdX+P+~ (p) dX
or
[x+~'(p)) :: =0.
Equating each factor to zero, we get
~=0
dx (2)
506 Difjerential Equations
and
X -1- \jl' (p) = 0. (3)
1) Integrating (2) we obtain p = C (C = const). Putting this
value of p into (1 '), we find its complete integral:
y=xC +'I' (C), (4)
which, geometrically, is a family of straight lines.
2) If from (3) we find p as a function of x and put it into
(1 '), we obtain the function
y=xp(x)-1-'P[p(x}], (1")
which may be readily shown to be the solution of equation (1).
Indeed, by virtue of (3) we have
~~ =p-t-[x-t-\jl'(p)] ~~ =p.
And so, by substituting the function (1") into equation (l) we get
the identity
xp + \jl (p) = xp + \jl (p).
The solution of (1") is not obtained from the complete integral (4)
for any value of C. This is a singular solution; it is obtained by
elimination of the parameter p from the equations
y=xp-t-\jl(p),}
x+'P' (p) =0,
or, which is the same thing, by eliminating C from the equations
y=xC +'I' (C),
x+'P~(C)=O.
Thus, the singular solution of Clairaut's equation defines the
envelope of a family of straight lines represented by the complete
integral (4).
Example. Find the general and singular solutions of the equation
dy
y 1+( :~ r
dy adx
y=xdx+
Solution. The general solution is obtained by substituti.ng C f~r :~ 1
aC
y=xC+-----=- ===
y' 1 +ca
Lagrange's Equa( ion 507
To obtain the singular solution, differentiate y

the latter equation with respect to C:
x+ a =0 .
..!.
2
(1 +C')
The singular solution (the equation of the
envelope)" is obtained in parametric form (where
"' P'"m1r ':~~(I+:~:,

a X
1 aC 3
l !1= (1 + C')+
Eliminating C, we get a direct relationship Fig. 258.
between x and y. Raising both sides of each
equation to the power ~ and adding the resultant equations termwise, \'lC
get the singular solution in the following form:

~ 2
xs +Ya =as.
This is an astroid. However,. the envelope of the family (and, hence, the
singular solution) is not the entire astroid, but only its left half (since it is
evident from the parametric equations that x..;;;; 0) (Fig. 258).
SEC. 14. LAGRANGE'S EQUATION
The Lagrange equation is an equation of the form

y = X<p (y') + 'iJ (y') (I)
where <p and 'iJ are known functions of :~ .

This equation is linear in y and x. Clairaut's equation, which
was considered in the preceding section, is a particular case of
the Lagrange equation when <p (y') =y', The Lagrange equation,
like Clairaut's, is integrated by means of introducing an auxiliary
parameter p. Put
y'=p;
then the initial equation is written in the form
y = X<p (p) + 'iJ (p). (I')
Differentiating with respect to x, we obtain
p = <p (p) + [x<p' (p) +'ll' (p)] ~~
or
p-ep (p) = [xcp' (p) +'I'' (p)] :~. (1 ")
From this equation we can straightway find certain solutions:
namely, it becomes an identity for any constant value p = p0
that satisfies the condition
Po -cp (P 0 ) = 0.
Indeed, for a constant value p the derivative ~: :::=0, and both
sides of equation (1") vanish.
The solution corresponding to each value p = p 0 , that is, :~ = p0
is a linear function of x (since the derivative ~~ is constant only
in the case of linear functions). To find. this function it is suf-
ficient to put into (1 ') the value p = p0 :
Y = xcp (Po)+ 'I' (Po).
If it turns out that this solution is not obtainable from the gener-
al solution for any value of the arbitrary constant, it will be a
singular solution.
Let us now find the general solution. Write (1") in the form
dx c:p' (p) 'IJ' (p)
([ji-x p-c:p(p) p-c:p(p)
and regard x as a function of p. Then the equation obtained

will be a linear differential equation in the function x of p.
Solving it, we find
X=(!) (p, C). (2)
Eliminating the parameter p from equations (1 ') and (2), we
get the complete integral (1) in the form <D (x, y, C)= 0.
Example. Given the equation
y=xy'z+y'a. (l)
Putting y' =P we have
(I')
Differentiating with respect to x, we get
dp
p=p 2 +[2xp+2p) dx" (l")
Let us find the singular solutions. Since p=p 2 for p 0 =0 and p 1 =1. the
solutions will be linear functions (see (1')):
y=x0 2 +0 2 , that is, y=O,
Orthogonal and Isogonal Trajectories 509
and
y=x+l.
When we find the complete integral, we will see whether these functions arc
particular or singular solutions. To find it, write equation (I") in the form
dx 2p 2
--x--2=--
dp p-p 1-p
and we shall regard x as a function of the independent variable p. Integrating
this linear (in x) equation, we find
C2
x=-l+(p-l) 2 (II)
Eliminating p from equations (I') and (II), we get the complete integral
y=(C+ Yx+l)'.
The singular integral of the initial equation is
y=O
since this solution is not obtainable from the general solution for any- value
of C.
However, the function y=x+ 1 is not a singular but a particular solution;
it is obtained from the general solution when C=O.
SEC. 15. ORTHOGONAL AND ISOGONAL TRAJECTORIES
Suppose we have a one-parameter family of curves

fl>(x, y, C)=O. (1)
Lines intersecting all the curves of the given family (1) at a
constant angle are called isogonal trajectories. If this angle is
a right angle, they are orthogonal trajectories.
Orthogonal trajectories. Let us find the equation of orthogonal
trajectories. Write the differential equation of the given family of
curves, eliminating the parameter C from the equations
<D(x, y, C)=O
and
a<D + acr> d y _ 0
ax ay dx- .
Let this differential equation be

F (x. y, :~)=0. (1 ')
. Here, :: is the slope of the tangent to some member of the

family at the point M (x, y). Since an orthogonal trajectory pass-
ing through the point M (x, y) is perpendicular to the correspond-
. d ..
ing curve of the family, the slope of the tangent to it, J:, is
5!0 Differential Equations
connected with :~ by the relationship (Fig. 259)

dy 1
dx =-=- dyr
di

(2)
Putting this expression into equation (1') and dropping the

subscript T, we get a relationship between the coordinates of an
arbitrary point (x, y) and the slope of the orthogonal trajectory
at this point, that is, a differential
Y equation of orthogonal trajectories:
F (x, y, _.:._ :!~) = 0. (3)
The complete integral of this equa-

tion
<1> 1 (x, y, C)=O
yields a family of orthogonal trajec-
~~~Q/.,.-======1===-~xK' tori es.
A consideration of the plane flow
of a fluid involves orthogonal trajec-
tories.
Fig. 259.
Let us suppose that the fluid flow
in a plane takes place in such man-
ner that at each point of- the xy-plane the velocity vector,
v (x, y), of motion is defined. If this vector depends solely on
the po~ition of the point in the plane, but is. independent of the
time, the motion is called stationary or steady-state. We shall
consider such motion. In addition, we shall assume that there
exists a potential of velocities, that is, a function u (x, y) such
that the projections of the vector v (x, y) on the coordinate axis
vx (x, y) and vy (x, y) are its partial derivatives with respect to x
andy:
(4)
The lines of the family
u(x, y)=C (5)
are called equipotential lines (Jines of equal potential).
The lines, the tangents to which at all points coincide with the
vector v (x, y) in direction, are called flow lines and yield the
trajectories of moving points.
Orthogonal and Isogonal Trajectories 511
We shall show that the flow lines are the orthogonal trajectories
of a family of equipotential lines (Fig. 260).
Let <p be an angle formed by the velocity vector 'V with the
x-axis. Then by relation (4)
au (x, y) =\'VI cos cp au (x, y) =I 'VI sin <p,
ax ' ay
whence we find the slope of the tangent to the flow line
au (x, y)
tan<p=a-(
U. X, y
-au
)' (G)
----ax-
We obtain the siope of the tangent to the equipotential line by
differentiating, with respect to x, relation IJ
(5):
whence
au
dy ax
dx= -au. (7)
ay
Thus, in magnitude and sign, the slope 0 x
of the tangent to the equipotential line is Fig. 260.
the inverse of the slope of the tangent to
the flow line. Whence it follows that equipotential lines and flow
lines are mutually orthogonal.
In the case of an electric or magnetic field, the lines of force
of the field serve as the orthogonal trajectories of the family of
equipotential lines.
Example 1. Find tile orthogonal trajectories of the family of parabolas
y=Cx.
Solution. Write the differential equation of the family
y' ~2Cx.
Eliminating C, we get
y' 2
-y=-x
Substituting -..!,
.y
for y', we obtain a differential equation of the family of
orthogonal trajectories
or
xdx
ydy=--z
Its complete integral is
x2+y2=C2
4 2 .
Hence, the orthogonal tr-ajectories of the given family of parabolas will be
represented by a certain family of ellipses with semi-axes a=2C, b=C 2
(Fig. 261).
Fig. 261.
Isogonal trajectories. Let the trajectories cut the curves of a

given family at an angle a, where tana=k.
y The slope ~~ =tan <p (Fig. 262) of the tan-
gent to a member of the family and the slope
dJ: =tan 'ljl to the isogonal trajectory are con-
nected by the relationship
X
'tan 'ljl-tan a
Fig. 262. tan <p =tan ('ljl-a) = 1 +tan a tan 'ljl;
Orthogonal and. 1sogonal Trajectories 513
that is,
dyT_ k
d11 dx
d~ = -k--,-dy_T_+_l (2')
dx
Substituting this expression into
equation (1 ') and dropping the sub-
script T, we obtain the differential
equation of isogonal trajectories.
Example 2. F.ind the isogonal trajec-
tories of a family of straight lines,
y=Cx, (8)
that cut the lines of the given family
at an angle a, the tangent of which Fig. 263.
equals k: tan a= k..
Solution. Let us write the differential equation of the given family. Diffe-
rentiating equation (8) with respect to x, we find
dy_c.
dx-
On the other hand, from the same equation we have
C=.JL.
X
Consequently, the differential equation of the given family is of the form

dy y
dx=x
Utilising relationship (2') we get the differential equation o! isogonal
trajectories
dyT_k
dx y
kdYT +1 X
dx
Whence, dropping the subscript T, we find
dy
dx
1-k.JL
X
Integrating this homogeneous equation, we get the complete integral:
In
. r-- I y
r x"+y'=karctan x-+lnC, (9)
which defines the family of isogonal tralectories. To find out precisely which
17-3388
514 DiUerential .Equations
sur.ves enter Into this family, let us change to polar coordinates:,

,.. }!_= tancp;
X
Yx +y 2 2
=Q
Substituting these expressions into (9) we obtain

1
In Q = k q> + In C
or
(j)
-cT
Q-e
Consequently, the family of isogonal trajectories Is a family of logarithmic
spirals (Fig. 263).
SEC. 16. HIGHER-ORDER DIFFERENTIAL EQUATIONS

(FUNDAMENTALS)
As has already been indicated above (see Sec. 2), a differential

equation of the nth order may be writte~ symbolically in the form
F (X, y, y ' I y"I ~I y C"l) = 0 (1)
; .
or, if it can be solved for the nth derivative, '!
y cnJ =' f (x,y,y,y,

' n ,yen- 1 l). . (I')
In this chapter we shall consider only such equations of higher
order that may be solved for a higher derivative. For these
equations we have a theorem on the existence and uniqueness of
a solution, similar to the corresponding theorem on the solution
of first-order equations.
Theorem. If in the equation
yi"J=f(x, y, y', ... , y<"-'l)
the function f(x, y, y', .. ,y1"- 11 ) and its partial deriuatiues with
respect to the arguments y. y', . , y 1"- 11 are continuous in some
region containing the values x = X0 , y = Yo y' = y:, ~ .. ,
y'"-' 1= y~n-l), then there is one and only one solution, y = y (x), of
the equation that satisfies the conditions
Yx=xo = Yo
y;,
:,, I
Y:=xo =
.y<n-1)
... .
.t=X0
= y<n-J)
o
} (2)
The~e co.ndi,tions ,are call~p initial conditions .. Th~. proof is .bey,oiid

the scope of this book.
Higher-Order Differential Equations 515
If we consider a second-order equation y = f (x, y, y'), then the

initial conditions for the solution, when X=X 0 , will be
y=y y'=y:
where X 0 , y 0 , y~ are given numbers, which have the following
geometric meaning: only one curve passes through a given point
of the plane (x 0 , y 0 ) with given tangent of the angle of inclination
of the tangent line y:.
From this it follows that if we want to
assign different values of y:
for constant x 0 and y 0 , we get an
infinitude of integral curves with different angles of inclination
passing through the given point.
We now introduce the concept of a general solution of an equa-
tion of the nth order.
'Definition. The general solution of a differential equation of the
nth order is the function
y=cp(x,epe 2 , ,en),
which is dependent on n arbitrary constants e~' e
2
, , en and
such that:
a) it satisfies the equation for any values of the constants
e,, e 2, ,en;
b) for specified initial conditions
Yx=xo=
, Yo,
Yx=x.=Yo
the constants e,. e2 , , en may be chosen so that the func-

tion y = cp (x, e,, e 2 , , en) will satisfy these conditions (on the
assumption that the initial values X 0, y., y~ ... , y~n-IJ belong
to the region where the conditions of the existence of a solution
are fulfilled).
A relationship of the form a> (x, y, e" e 2 , , en)= 0, which
implicit! y defines the general solution, is called the complete
integral of the differential equation.
Any function obtained from the general solution for specific
values of the constants e1 , e1 , , en
is called a particular
solution. The graph of a particular solution is called an integral
curve of the given differential equation.
To solve. (integrate) a differential equation of the. nth order
means: .J
17*
51'6 Differential Equations
I) to find its general solution (if the initial conditions are not
given) or
2) to find a particular solution of the equation that satisfies
the given initial conditions (if there are such).
In the following sections we shall present methods of solving
various equations of the nth order.
SEC. 17. AN EQUATION OF THE FORM y<n> =f(x)
The simplest type of equation of the nth order is of the form

y'n> = f (x). (I)
Let us find the complete integral of this equation.
Integrating the left and right sides with respect to x, and
taking into account that y'n 1=(y<n- 11 )', we obtain
X
y'n-l> = ~ f (x).dx+C 1 ,
where x 0 is any fixed value of x, and C 1 is the constant of

integration.
Integrating once more we get
X X
!/n- I)=~ ( ~ f (x) dx) dx+ cl (x-xo) + c,.

X0 Xc1
Continuing, we finally get (after n integrations) the expression of

the complete integral:
In order to find a parti.cular solution satisfying the initial condi-

tions
Yx=xo =Yo; Y:=x, = y;; 0
...
(n-J) = y<n-1)
, Yx=x o '
it is sufficient to put
Cn=Yo, Cn-1=y;, C~=y~-~ ..
Example t. Find the complete Integral of the equation
y =sln(kx)
and a particular solution satisfying the initial conditions
Yx=o=O, Y~=o= 1.
An Equ,at_ion of the. form y(n) = f (x) 5)7
Solution...
X
!1
,= Ssin kx dx + C = - I
cos kx-
k
1 + .C I
0
Y=-
S
x ( cos kx - I )
k dx +J
r C, dx+C 2
0 0
or
sin kx x
y=- ~+k+C,x+C2
This is the complete integral. To find a particular solution satisfying the

given initial conditions, it is sufficient to determine the corresponding values
of C1 and C1
From the condiUon Yx=o=O, we find C1 =0.
Fromthe.condition u:~ 0 =l, we find C1 =0.
Thus, the desired particular solution is of the form
sin kx ( I )
y=-~+x k+l .
Differential equations of this kind a~e encountered in the theory of the
bending of girders. . .
Example 2. let us consider an elastic prismatic girder bending under the
action of external forces both continuou,sly distribu.ted (\,Veigh1, load~, and
concentrated. Let the x-axis be horizontal -
along the axis of the girder in its underformed L-x
state and let the y-axis be directed vertically x
downwards (Fig. 264). X
Each force acting on the girder (the load
of th~. girder, .and the reaction of the supports,
for instance) has a moment, relative to some
cross section of the girder, equal to the prod-
uct of the force by the distance of thr pojnt p
of application of the force from the given y
cross section. The sum, M (x), of the moments
Fig. 264.
of all the forces applied to that part of the
girder situated to one side of the given cross
section with abscissa x is called the bending moment of the girder relative
to the given cross section. In courses of strength of materials, it is proved that
the bending moment of the girder is
EJ
R'
where E is the so-called modulus of elasticity which depends on the material
of the girder, J is the moment of inertia of the cross-sectional area of the
girder relative to the horizontal line passing through the centre of gravity of
the cross-sectional area, and R is the radius of curvature of the axi:s of the
bent girder, which radius is expressed by the formula (Sec. 6, Ch. VI)
(I+ y'z)t.
R= y" .
518 . Differential Equations
Thus, the differential equation of the bent ais of a girder has the form
y" M (x)
(2)
(I+ y'") 'I =:/.
II we consider that the deformations are small and lhat the tangents to
lhe axis of the girder, when bent, form a small angle with the x-axis, we can
disregard the square of the small quantity y'' and consider
1
R=!/.
Then the differential equation of the bent girder will have the form
M (x) (2')
u=ET
but lhis equation is of the form of (1).
Example 3. A girder is fixed .in place at the extremity 0 and is .subjected
to the action of a concentrated vertical force P applied to the end of the
girder L at a distance l from Q (Fig. 264). The weight of the girder is ignored:
We consider a cross section at the point N (x). The bending .moment rela-
tive to section N is, in the given case, equal to
M (x)=(l-x) P.
The differential equation (2') has the" r'orm
Y* l )
p <.-x.
= EJ
The initial conditions are: for x=O the deflection y is equal to zero and ~h.e
tangent to the bent axis of the girder coincides with the x-axis; that is,
Yx=o=O, u:=O =0.
Integrating the equation, we find
p s" (l-x) dx=-

y'=-
EJ
p ( lx--
EJ 2 '
x')
0 '
p (
y=2EJ. Ix'-3 x) (3)
In particular, from formula (3) we determine the deflection h at the extre-

mity of the girder L:
SEC. 18. SOME TYPES OF SECOND-ORDER DIFFERENTIAL

EQUATIONS REDUCIBLE TO FIRST-ORDER EQUATIONS
I. An equation of the type
. ,: ,.
:~=f(x.~~) (1)
. '
does not .explicitly contain the unknown function y.

Some Types of- Second-Order Dilferenlil:ll Equations 519
Solution. Let us denote the derivative ~; in terms of p, that

. . t dy d'y . dp
IS, we se dx = p. Then dx' = dx
Putting these expressions of the derivatives into equation (1),
we get a first-order equation,
:~=f(x, p),
in the unknown function p of x. Integrating this equation, we
find its general solution:
p=p(x, C1 ),
and then from the relation :; = p we get the complete integral
of equation ( 1}:
y= ~p(x, C,}dx+C.
Example l. Let us consider the differential equation of a ca~enary (see
Sec. 1):
Set
dy
dx=p;
then
d 2y dp
dx= dx'
and we get a first-order differential equation in the auxiliary function p of .xJ
dp l .r +-2
dx=a-r 1 P
dp
-dx
a
whence
In (P + V 1 +P = !_+C,,
2
a
)
p
1 (
='2 e
1J+c,
-e
- (: -c,))
But since p = :; , the latter relation is a differential equation in !he sought-
lor function y. Integrating it, we obtain the equation of a catenary (se'e
520 DlOerentlal Equations ;.
Sec. 1):
a ( !..+c, - ( ~ +C,))
Y=2 e a +11 +C 1.
Let us find the particular solution that satisfies the following initial con-
ditions:
Yx=o=a,
'I
Yx=o==O.
.The first condition yields C1 =0, t~e second, C1 =;:0.
We finally obtain
Note. We can similarly integrate the equation

y<nl = f (x, y<~,-ll).
Setting y<n- '> = p, we get for a ! P.etermtnation of p the first-

order equation dp : .,. .
-=f(x,
dX p)..
. ,.
From here we get p as a function or x, and from the relation

y<n -> = p we find y (see Sec. 17).
II. An equation of the type.
(2)
does not contain the independent variable x explicitly. To solve
it, we again set
dy
dx=p, (3)
but now we shall consider p as a function of y (and not of x,
as before). Then
Putting i nio (2) the expressions =~ and we get a first-

order equation in the auxiliary function p:
dp
p dy = f (y' p). (4)
Integrating it, we find p as a function of y and the arbitrary
constant C,:
Some Types of Second-Order Dt(j"ereiJ.lial Equations 52f.
Substituting this value in (3), we get a first-order differential

equation for the function y of x:
dy
dx=p(y, C,).
dy -dx
p(y, C,)- .
Integrating this equation, we get the complete integral of the
initial equation:
Example 2. Find the complete integral of the equation
3y"=y. a
Solution. Put p =ddy and consider p as a function o y. Then y = p ,ddp

X ; y
and we get a first-order equation for the auxiliary functi~n p:
3p dp =Y
dy ' ....
Integrating this equation, we find
p= Vc, -y-t.
. dy
But p = dx; consequently, for a determination of y we get the equation
, .. i
dx,
whence .
To compute the latter integral we make the substitution

c,y't-1 =t.
Then
I
-I
y'/a=(t +I) a C'i;
1
1
dy=3t (t 2 +1)'1 - - dt.
c"Ja
1
Differential-Equations
Finally we get
x+C.= _!._ V c,y'l-1 (C,y''+2).
c~
Example 3. Let a point move along the x-axis under the action of a force
that depends solely on the position of the point. The differential equation
of motion will be '
d 1x
m dt=F (x).
dx
At t=O let X=X 0, (i[=v 0
Multiplying both sides of the equation by :; dt and integrating from 0

to t, we have
1 ( -dx) --mv
-m 1 = sx F(x)dx
2 dt 2 0 .
or
(:;y + [-SF J
X
; m (x) dx = ~mv:=const.
x.
The first term of this equation is the kinetic energy, the second term,
the potential energy of the moving point. From this equation if follows that
y the sum of the kinetic and potential energy re-
mains constant throughout the time of motion.
The problem of a simple pendulum. Let there
be a material point of mass m, which is in
motion (by the force of gravity)along the Circle L
lying in the vertical plane. Let us find the equa-
tion of motion of the point neglecting resistance
forces (friction, air resistance, etc.).
Putting the origin at the lowest point of the
circle, we put the x-axis along the tangent to
the circle (Fig. 265).
Denote by l the radius of the circle, by s the
arc length from the origin 0 to the variable point
M where the mass m is located; this length is
taken with the appropriate sign (s > 0, if the
point M is on the right of 0; s < 0 if M is on
the left of 0).
Our problem consists in establishing s as a
function of the time t.
Let us de.compose the force of gravity mg into
Fig. 265, tangential and normal components. The former,
Some Types of Second-Order ,Dif!erential. Equations 523
equal to -mg sin q>, produces motion, the latter is cancelled by the reaction
of the curve along which the mass m is moving.
Thus, the equation of motion is of the form
d 2s
m dtz= -mg sin<p.
s
Since the angle q>=y for a circle, we get the equation
d 2.~ s
dtz=-gsln-y.
This i~ a Type q differential equation (since it does not contain the lnde
pendent variable t explicitly).
Let us integrate it in the appropriate fashion:
Hence,
. dp' . '. s
pd~=-gsm-y
or
s
pdp= -g sin T ds,
whence
p 1 =2glcos ; +C 1
Let us denote by s0 the greatest arc length to which the point M swings.
For s='=s~ the 'velocity of the point is zero:
dsl
dt s
This enables us to determine C11
=--pi =- .
s
-0
s
0=2gl cosf+ C,,
whence
Therefore,
= 2gl ( cos--
p'= ( -ds)' s s )
cos..!..
. dt I l
or, applying to the latter expression the formula for the difference of cosines,
d.~)z
.... :( dl
s+.~. s -s
0
=4gl.sln 21 sln 2 l ' (5)
524 . Dif}eren tial Equatiotts
.r- ~I. s+s0

ds
dt = 2 f g1 v Slfi
s -s
--w- Slfl --w-

. 0
(6)
This is an equation with variables separable. Separating the variables, we get

ds .r-
----:r==:===== = 2 f g/ dt. (7)
~ I . S + S0
v --w- --w-
Sin
S0 -
Sin
S .
We shall assume, for the time. being, . that s =F s0 , then the denominator of
the fraction is different from zero. If we consider that s=O for 1=0, then
from (7) we get
r
J ~Vf . s +S
ds
I -s
2 Jlgk (8)
0 sm 2 l s n
0
--w-
S0
'
This is the equation that yields s as a function of t. The integral on the
left cannot be expressed in terms of elementary functions; neither can
the function s of t. Let us consider this problem approximately. We shall
assume that the angles Ts s
and -{ are ~mall. The angles
s +s
T and T
s -s
will not exceed sl In (6) let us replace, approximately, the sines of !he
angles by the angles
ds = 2 Vl ~ Is+ s0 s0 -s
dt g 21 21 v
or
~ ~ /( ")
ds
dt= Vlg
TV so-s (6')
Ceparating variables, we get (assuming, for the time being. that s-/:. s
-v f
0)
ds (7')
dt.
}I s~-s
Again we consider that s = 0 when t = 0. Integrating the latter equation, we get
(8')
or
arc sin f-= V f t,

0
"
*) We put the plus sign in front of the root. From the note at the end of
the solution it follows that there is no need to consider the case with the
minus sign.
Some Types of Second-Order Diflerenlial Equations 525
whence
(9)
Note. When solving, we assumed that s ::f. s0 But it is clear, by direct

substitution, that the function (9) is the solution of equation (6') for any
value of t.
Let it be recalled that the solution (9) is an approximate solution of
equation (5), since equation (6) was replaced by the approximate equation (6').
Equation (9) shows that the point M (which may be regarded as lhe
extremity of the pendulum) performs harmonic oscillations with a period
T=2n y; . This period is independent of the amplitude s0
Example 4. Escape-velocity problem.
Determine the smallest velocity with which a body must be thrown ver-
tically upwards so that it will not return to the earth. Air resistance is
neglected.
Solution. Denote the mass of the earth and the mass of the body by M
and m respectively. By Newton's law of gravitation, the force of attraction f
acting on the body m is
where r Is the distance between the ce'ntre of the earth and the centre of
gravity of the body, and .k is the gravitational constant.
The differential equation of motion of this body with mass m will be
d2r M m
m-=-k--
dt2 r
or
(10)
The minus sign indicates that the acceleration is negative. The differen-
tial equation (10) is an equation of type (2). We shall solve it for the fol-
lowing initial conditions:
for t =0
Here, R Is the radius of the earth and o 0 Is the launching velocity. We deneb
dr d'r do du dr du
dt=U, dt 2=dt=dr' Cfl=U dr'
where o is the velocity of motion. Putting this Into (10), we get
do M
0 - = - kr.-1
dr
Separating variables, we obtain
d,.
odo = -kM r.
o I
'2= +kM ;-+C1 (11)
526 Differentidl Equations
From the condition that V=v 0 at the earth's surface {for r=R), we deter-
mine C1 :
v~ I
2=+kM J[+C 1
or
kM v!
C,=-R+T
We put the value of C1 into (11): .\
v 1 kM v!
2=+kM-,--y+2
or
(12)
It is given that the body should move so that the velocity Is always positive;
hence, ~ > 0. Since for a boundless increase of , the quantity k~ becomes
arbitrarily small, the

condition ;> 0 will be fulfilled for any ,. only for
the case
(13)
or
Hence, the lowest velocity will be determined by the equation
V0 = y 2 ~M' (14)
where
k=6.66I0- 8 cm 3 /gmsec 1 ,
R = 63107 em.
At the earth's surface, for r = R. the acceleration of gravity is g (g = 981 cmjsec~.
For this reason, from (10) we obtain
M
g=k R'
or
Putting this value of M Into (14) we obtain

, -- em .km
v1 = r1 2gR= V29816310':::::11.2101 -=1L2-,
sec sec
Graphical Method of Integration 52.1
SEC. 19. GRAPHICAL METHOD OF INTEGRATING

SECOND-ORDER DIFFERENTIAL EQUATIONS
Let us find out the geometric meaning of a second-order differ-

ential equation. Suppose we have an equation
y"=f(x,y,y'). (l)
Denote by IP the angle formed by the positive x-axis and the
tangent to a curve; then
dy
dx =tan IP (2)
To find the geometric significance of the second derivative,

recall the formula that determines the radius of curvature of a
curve at a given point*)
Whe'nce
But
y~ =tan qJ; 1 + y' = 1 + tan 2 IP = sec 2 q;; (1 + y' 2
)"' =
I . I I
= Isec <p = 1cos <p I '
therefore
y" = R 1 c~s cp I . (3)
Now putting into (1) the expressions obtained for y lind y", we
have
R Ieos <pI
f (x, y, tan IJJ)
or
1
R =I cos <plf (x,y, tan <p) (4)
It is thus evident that a second-order differential equation deter-
mines the magnitude of the radius of curvature of an integral
curve if the coordinates of the point and the direction of the
tangent to this point are specified.
*) Up till now we have always considered the radius of curvature positive;
in this section we shall consider it a number that can take on both positive
and negative values: if the curve is convex (y" < 0), we consider . the -rai!Tus
of curvature negative (R<O); if the curve is_ concave (y">O), it is positive
(R>O). '
528 . DiUerential Equations
From the foregoing there follows a method of approximate con-

struction of an i.ntegral curve by means of a smooth curve com-
posed of arcs of circles. *)
To illustrate, let it be required to find the solution of equation
{I) that satisfies the following initial conditions:
Through the point J\10 (x., y.) draw a ray M.T. with slope y' =
= tan <p = y~ (Fig. 266). From equation (4) we find the magnitude
of R = k. Lay off a segment M.c., equal to R., perpendic~lar
to 111 T , and from the point c. (as centre) strike an arc M.Md
with radius R.. It should be note
NJ 7f Tz that if R.<O, then the segment .M.c.
To must be drawn in that direction so
that the arc of the circle is convex
Yil upwards, and for R.>O, convex down
(see footnote on page 527).
Then let X 1 , y 1 be the coordinates
of the , point 111 1 which lies on the
constructed arc and is sufficiently
close to the point M. while tan <p; is
0 Xo x the slope of the tangent M 1 T, to the
Fig. 266. circle drawn at M,. From equation {4)
we find the value of R =R, that cor-
responds to M,. Draw the segment M.G., perpendicular to M,T,,
equal to R., and from C, (as centre) strike an arc M')i. with
radius R . Then on this arc take a point M. (x., y.) close to M,
and continue construction as before until we get a sufficiently
large piece of the curve consisting of the arcs of circles. From
the foregoing it is clear that this curve is approximately an integral
curve that passes throught the point M 0
Obviously, the smaller the arcs 111:111., M~M., ... , the closer
the constructed curve will be to the integral curve.
SEC. 20. HOMOGENEOUS LINEAR EQUATIONS.

DEFINITIONS AND GENERAL PROPERTIES
Definition I. An nth-order differential equation is called linear

if it is of the first degree in the unknown function y and its
*) A curve is called smooth if it has tangents at all points and the angle
of inclination of the tangent is a continuous function of the arc length s.
Homogeneous Linear qu11t ions 529
derivatives y', .. : ,yin-'', y(nl; that is, if it is of the lorm

a 0 y(n! +a,y(n-t>+ .. . +any=f(x), . (1)
where a 0 , at' a,, ... , an and f (x) are given functions of x or
constants, and a 0 =f=0 for all values of x from the domain in which
we consider equation (1). From now on we shall presume that
the functions a 0 , a" ... , an and f (x) are continuous for all values
of x and that the coefficient a 0 = 1 (if it is not equal to I we
can divide all terms of the equation by it). The function f(x)
on the right side of the equation is called the right-hand member
of the equation.
If f (x);t:O, then the equation is called nonhomogeneous linear
or an equation with a right-hand member. But if f(x) =:0 then
the equation has the form
y(n>+a 1 y(n-tl+ ... +any=0 (2)
and is called homogeneous linear or an equation without- a right-
hrtnd member (the left side of this equation is a homogeneous
function of the first degree in y, y', y", ... , y(n>).
Let us determine some of the basic properties of homogeneous
linear equations,confining our proof to second-order equations.
Theorem t. If y,. and y. are two particular solutions of a homo-
geneous linear equation of the second order
y" +a,y' +a.y=O, (3)
then y, + y. is also a solution of this equation.
Proof. Since y, and Y. are solutions of the equation, we have
y; +a,y; +a.y 1 =0}
and (4)
y: +a 1 y~ +a.y. =0.
Putting into equation (3) the sum y, + Y. and taking into account
the identities (4), we will have
(y, + y.)'' +a, (y, + y.)' +a, (y, + Y.) =
= (y; + a,y; + a.y,) + (y: + a 1 y: + a.y.) = 0 +0 = 0.
Thus, y, + y. is a solution of the equation.

Theorem 2. If y, is a solution of equation (3) and C is a con-
stant, then Cy, is also a solution of (3 ).
Proof. Substituting into (3) the expression Cy" we get
(CyJ' +a, (CyJ' +a. (Cy,) =C [y; + a y: + a,y,] = CO =0;
1
and the theorem is thus proved.

530. DiD,eren(i.al Equatio.ns
Definition 2. The two solutions of equation (3}, Y1 and y,,

are ca!'led linearly independent on an interval [a, b] if their ratio
on this interval is not a constant; that is, if
!b. =f= canst.
u.
Otherwise the solutions are called linearly dependent. In other
words, two solutions, y 1 and y,, are called linearly dependent on
an interval [a, b] if there exists a constant number A. such that
!b.= A. when a,;;;:;; x,;;;:;; b. In this case, Y1 = A.y,.
Yz
Example 1. Let there be an equation y" -y =0. It is easy to verify that
the functions e", e-", 3e", 5e-" are solutions of this equation. , Here, the
functions e" and e-" are linearly independent on any interval because the
ratio " = e2" does not remain constant as x vafiies. But the functions e"
!__"
e-
3
and 3e" are linearly dependent, since : =3=const.
e
Definition 3. If y 1 and y. are functio!Js of x, the determinant
w(yl'y.) = j ~; ~: j = Y~Y~- y;y.

is called the Wronskian of the given functions.
Theorem 3. If the functions Y1 and y, are linearly dependent on
an interval [a, b], then the Wronskian on this interval is identi-
cally zero.
Indeed, if y, = A.y 1 where A.= canst, then y~ = A.y; and
W(yi.Y.>=I~; ~:l=lt; ~t:I=A.I~; ~;j=o.

Theorem 4. If the Wronskian W (y 1 ,y,), formed for the solutions
Y1 and y, of the homogeneous linear equation ( 3 ), is not zero for
some value x = X 0 on an interval [a, b] where the coefficients of
the equation are continuous, then it does not vanish for any value
of x whatso_ever on this interval.
Proof. Since y 1 and y, are two solutions of equation (3), we
have
.. y; + a 1 y; + a2 y 1 = 0, y; + a 1 y~ + a,y, = 0.
Multiplying the terms of the first equation by y 1 , the terms of
the second equation by -y~' and adding, we get
(Y:Y.-YtY:)+at (y~y~~Y 1 Yz).=0. (~)
Homogeneous Linear Equations 531
The difference in the second brackets is the Wronskian- W (y" y.).

The expression in the first brackets is a derivative of the Wrons-
kian W' (y" y,):
W' (y,y,) = (y;y.-y,y~)' = Y:Y. + y;y~ -y;y~- Y,Y: = y;y. -y,y:.
Thus, equation (5) assumes the form
W'=-a,W. (6)
Separating variables (for W =;l=O), we obtain
W'
w=-a,.
Integrating, we find
JC
ln W = - 5a, dx +In C
~.
or
w
ln-=
C -5 a dx X
I '
"
whence
-
"
~ a dx 1
W=Ce " (7)

it is given that
0
Wx=xo = Ce = C =;i=O.
But then from (7) it follows that W =;i=O for any values of x,
because the exponential function does not vanish for any finite
value of the argument.
Note t. If the Wronskian is zero for some value x=x 0 , then
it is also zero for any value x in the interval under consideration.
This follows directly from (7):. if W = 0 when x = x 0 , then
(W)x=xo = C = 0;
consequently, W ==0, no matter what the value of the upper limit
of x in formula (7).
Theorem 5. If the solutions y, and y, of equation ( 3) are linearly
independent on an interval [a, b ], then the Wronskian W, formed
for these solutions, does not vanish at any point of the given
interval.
We shall hint at the proof of this theorem without giving it
cdrnpletely.
O.iffe.rentlal EquallotiS
Suppose . that W = 0 at some point of the interval; then, by

Theorem 3, the Wronskian will be zero at all. points of [a, tr]:
\V=O
or
Y.Y:-Y:Yz =0.
Let us first consider those subintervals in la, bj where y 1 +0.
Then
or
Consequently, on each of these subintervals ~ Is a constant

u.
Ya =A.= const.
.u.
Taking advantage of the existence and uniqueness theorem, it -
may be shown that Ya = A.y 1 for all points of the interval [a, b]
including those where y. =0; but this is impossible since it is
given that Ya and y 1 are linearly independent. Thus, the W rons-
kian does not vanish for any single point of [a, b].
Theorem 6. If y 1 and Ya are two linearly independent solutions
of equation (3), then
(8)
where c. and C 2 are arbitrary constants, is its general solution.
Proof. From Theorems 1 and 2 it follows that the function
c.y. +CzYz
is a solution of equation (3) for any values of C 1 and C 1
We shall now prove that no matter what the initial conditions
Yx=x. = Yo Y~=xO = y~, it is possible to choose the values of the arbit-
rary constants C, and C% so that the corresponding particular
solution C,y, + C 1 y. should satisfy the given initial conditions.
Substituting the initial conditions into (8), we have
(9)
where we put
(y,h=xo=Yto; (yz)x=xo=Yzo; (y;).<=.ta=Y;o; (y~)x=xo=Yzo
Homogeneous Linear E qwations 533
From the system (9) we can determine C1 and G_ since the deter-
minant of this system
Y,,o Y~o
Y1o Yo I I=
Y10 Y:o -y;oY2o
is the Wronskian for x = X 0 and, hence, is not equal to 0 (by

virtue of. the linear independence of the solutiqns y, and y,).
The particular solution obtained from the family (8) for the
found values of C, and c. satisfies the given initial conditions.
Thus, the theorem is proved.
I , I O
y +x-Y -X"y=,
whose coefficients a,=__!_ and a2 = ~ are continuous on any interval that does
X X
not cont<lin.~he point x=O, permits of the particular, solutions
I . . ,
y, =:=x. u.=-x .
(this is readily verified by substitution). Hence, its general solution is of
t.he form
Note 2. There are no general methods for finding (in finite

form) the general solution of a linear equation with variable coef
ficients. However, such a method exists. for an equation with
constant coefficients. It will be given in the, next section. For
the case of equations with variable coefficients, certain devices
will be given in Chapter XVI (Series) that will enable us to find
approximate solutions satisfying definite initial. co.ndit'i'ons. .
Here we shall prove a theorem that will enable us to find the
general solution of a second-order differential equation with variable
coefficients if one of its particular solutions is known. Since it
is sometimes possible to find or guess one particular' solution
directly, this theorem will prove useful in ma'ny cases.
Theorem 7. If we know one particular solution of a second-or-der
homogeneous linear equation, the finding of the gener-al solutiotZ
reduces to integrating the functions.
Proof. Let Y1 be some known particular solution of the equation
y" +a,y' +a.y=O.
We find another particular solution of the given eqJJdion so that
y, and y 2 are linearly independent. Then the g~ne.r.al solution
will be expressed by the formula y = C,y, C.y., where C, and +
c. are arbitrary constants. By virtue of formula (7) (see proof of
5_34 D(fferential Equations
Theorem 4}, we can write

., , - ~a,dx
Y2Yt-YsY 1 = Ce .
Thus, for a determination of y 2 we obtain a first-order linear
equation. Integrate it as follows. Divide all terms by y::
y~y 1 -y 2y:
----=-Ce
1 - a,dx J
u: u:
or
whence
-J a,dx
ft.!=Sce
y, u: dx+c
' ..
Since we are seeking a particular solution, we get (by putting
C 1 =0 and C= 1)
-J a,dx.
Ys=Y, STdx. (10)
It is 'obvious 'that y 1 and Ya are linearly independent solutions

since ~ =Fc~nst.
y,
Thus, the general solution of the initial equation is of the form
y=C 1y, +C 2Y1 se--.-

-J
y,
a,dx
dx. {11)
Example 3. Find the general sol uti on of the equation

(1-x") y -2xy' 2y=0. +
Solution. It is evident, by direct verification, that this equation has a
particular solution y 1 =x. I.,et us find the second particular solution y 2 , so
that y 1 and y 2 should be linearly independent.
-2x
Noting that In our case a1 = _ x, we have, by (10),
1
. r udx
eJt-x se-ln(t-JC) s dx
y=x
S- - dx=x
X1
- dx=x '(l
X2 X - X 2)
=X S(~+2(1 x) +2.(1 ~x))d7=x [-++~In I! :;I].

1
Comsequently, the .general solution is of the form
y=C1x+~. ( {-xln I! +;j-1 )

Second-Order Homogeneous J:.ineaT Equations 535
SEC. 21. SECOND-ORDER HOMOGENEOUS LINEAR

EQUATIONS WITH CONSTANT COEFFICIENTS
We have a second-order homogeneous linear equation

+
y" py' +qy = 0, (1)
where p and q are real constants. To find the complete integral
of this equation, it is sufficient (as has already been proved) to
find two linearly independent particular solutions.
Let us look for the particular solutions in the form
y=e"", where k = const; (2)
then
y' =kek"; y"=k 2 e"".
Substituting the expressions of the derivatives into equation (1),
we find
ekx (k
2
+
pk +q) =0.
Si nee e"x =1=0, it means that
k"+pk+q=O. (3)
Thus, if k satisfies equation (3), then ekJI wiH be a selufion
of (1). Equation (3) is called an auxiliary equation with respect
to equation (1).
The auxiliary equation is a quadratic equation with two roots;
let us denote them by k, and k 2 Then
p -./p-.- p -./pz-
k,=-2+ v 4-q; k.=-2- y 4-q.
The following cases are possible:
I. k, and k 2 are real numbers and not equal (k, =l=k,);
I I. k, and k, are complex numbers;
III. k, and k. are real and equal numbers (k, = k 2 ).
Let us consider each case separately. .
I. The roots of the auxiliary equation are real and distinct,
k 1 :=k2. Here, the particular solutions are the functions
y, =e""; Yz = e"x.
These solutions are linearly independent because
Y e""
~=....,...-x=ec k .-k,)x:=Fconst.
y, e.
Hence, the complete integral has the form
y = C ,e"~ + C.e"".
536. Di(Jerentia.Z Equations . ~,.: .
Example I. Given the equation

y" +u' -2y=O.
The auxiliary equation is of the form
k 2 +k-2=0.
We find the roots of the auxiliary equation:
I ,/I
k,,.= -2 v
-:r+2;
k 1 =I, k 2 =-2.
The complete integral is
y = C,e" + C e-
1
2
".
II. The roots of the auxiliary equation are compl~x. Since complex
roots are conjugate in pairs, we write
k 1 =a+ i~; k 1 = a-,.i~,
where
The particular solutions may be written in the form

y 1 =e<a+illl.t; y 1 ::::e(a-tll>x. (4)
These are complex functions of a real argument that satisfy the
differential equation (1) (see Sec. 4, Ch. VII).
It is obvious that if some complex function of a real argument
y=u(x)+iu(x) (5)
satisfies (1), then this equation is satisfied by the functions u(x)
and u (x).
Indeed, putting expression (5) into {1), we have
[u (x) + iu (x)]" + p [u (x) + io (x) ]' + q Ju (x) + iu (x)] == 0
or
(u" + pu' + qu) + i (u" + pv' + qv) == 0.
But a complex function is equal to zero if, and only if, the real
part and the imaginary part are equal to zero; that is,
u" + pu' + qu =0,
o" + pu' + qo = 0.
Thus we have proved that u (K) and o (x) are 'solutions of the
equation.
Let us rewrite the complex solutions (4) in the form o~ a stun
of the real part and the imaginary part:
y, =&cos ~x + ietu.sin ~x.
Ya = ea"cos ~x -iaax sin ~x.
SeaJitd .Order Homogeneous Linear E qtialions 537
From what has been proved, the particular solutions of ('l) are
the real functions
i/1 = e~x cos ~x. (6')
y.
= eu sin ~x. (6")
The functions y, and Yz are linearly independent, since
y e~X COS ~X
--d- = ~x . ~ =cot ~x+const.
Yz e sm x
Consequently, the general solution of equation (1) in the case of
complex roots of the auxiliary equation is of the form
y= Ay + By
1 1
= Aeax cos ~x + Beu sin Px
or
y=e~x (A cos ~x+ B sin ~x), .(7)
where A and B are arbitrary constants.
y" +2y' +5y=0.
Find the complete integral and a particular solution ~hat satisfies the initial
conditions Yx=o=O, y~=o= 1. Construct the graph.
Solution. 1) We write the auxiliary equation
k 2 +2k +5=0
and find its roots:
k,=-1+2t, k.=-l-2i.
Thus, the complete integral is
y=e-x (A cos 2x+ Bsln 2x).
2) We find a particular solution that satisfies the given initial conditions
and determine the corresponding values o! A and 8.
From the first condition we find
O=e- 0 (A cos 20+ B sin 20), whence A =0.
Noting that
y' = e -x28 cos 2x ..:.....e-x 8 sin 2x
we obtain !rom the second condition
I
1 = 28. so 8=-
2.
Thus, the desired particular solution is
y= ~e-x sin 2x.

lis graph is shown in Fig. 267.
Ill The roots of the auxiliary equation are real and equal.
Here, k 1 = k1
538 Different_iaL Equations
One particular solution, y, = ekx, is obtained from' earlier

reasoning. We must find the second particular solution, which is
y
------ X
Fig. 267.
linearly independent of the first (the function ek,x is identically

equal to eV and therefore cannot be regarded as the second par-
ticular solution).
We shall seek the second particular solution in the form
y2 = u (x) ek,x
where u (x) is the unknown function to be determined.
Differentiating, we find
y~ = u'ek,x +k,uek,x = ek,x (u' +k,u),
y: = u"ek,x + 2k,u' ek,x + k:uek,x = ek,x (u" + 2k,u' + k:).
Putting the expressions of the derivatives into (1), we obtain
ek,x [u" + (2k, + p) u' + (k~ + pk, + q) u] = 0.
Since k, is a multiple root of the auxiliary equation, we have
k: + pk, + q = 0.
In addition, k 1 = k 2 = - i or 2k, = -p, 2k, + p = 0.
Hence, in order to find u (x) we must solve the equation
e11 XU"=0 or u"=O. Integrating, we get u=Ax+ B. In particular,
we can set A= 1 and 8 =0; then
U=X,
Homogeneous Linear Equations of the nth Order 539
Thus, for. the second particular solution we can take

Y.=xek".
This solution is linearly independent of the first, since ~=x~const.
Yt
Therefore, the following function is the complete integral:
y = C,ek,x + c.xek,x = ek,x (C.+ c.x).
y"-4y'+4y=0.
Write the auxiliary equation. k 2 -4k+4=0. Find its roots: k 1 =k 2 =2.
The complete integral is then
y= C,e 2" + C2xezx.
SEC. 22. HOMOGENEOUS LINEAR EQUATIONS OF THE

NTH ORDER WITH CONSTANT COEFFICIENTS
Let us consider a homogeneohs linear equation of the nth order:

~/n>.--ta.ycn-t)+ :::. +any=O. (1)
We, shalt assume that a 1, a., ; .. ; an are constants. Before giving
a method for solving equation (1 ), we introduce a definition that
will be needed later on.
Definition t. If for all x of the interval [a, b) we have the
equality
where A., A 2 , , An are constants, not all equal to zero, then

we say that q>n (x) is expressed linearly in terms of the functions
q> 1 (x), q> 2 (x), ... , q>n-t (x).
Definition 2. n functions q> 1 (x), q> 2 (x), ... , q>n-1 (x), q>n (x) are
called linearly independent if not one of the functions is expressed
linearly in terms of the rest.
Note l. From the definitions it follows that if the functions
q> 1 (x), cp. (x), ... q>n (x) are linearly dependent, there will be
found constants C., c.... , Cn, not all equal to zero, such that
for all x of the interval [a, b) the following identity will be ful-
filled:
+
C1 cp 1 (x) C.cp. (x) + ... +
Cnq>n (x) 0. ==
Ex.amples:.. . . . . . . . .. . . ,.
1. The functions y 1 = e", y 1 =e."",. y 3 = 3e" are linearly dependent, .since for
C1 =i,. c.~o. C1 =-; we have the ide~lity
C 1 e~ +C e"" +C 3e" s:: 0.
1 3
DiUerential .quation.s, ...
2. The functions y 1 =I, y 2 =x, y 3 =X 2 are linearly irJderef!dent, since the

expression
C1 l+C 2x+C 3 x2
will not be identically zero for any C1 , C2 , C3 that are not simultaneously
equal to zero.
3. The [unctions y 1 =ek,x, y 2 =ekx, ... , y,==eknx ... , where k 1 , k 2 , , k 11 ,
are different numbers which are linearly independent. (This assertion is given
without proof.)
..
Let us now solve equation (1). For this equation, the following
theorem holds.
Theorem. If the functions y., y 2 , , y, are linearly independent
solutions of equation (I), then its general solution is
(2)
where cl' ... , c, are arbitrary constants,. . . .
If the coefficints of equation (I) are constant, the general so-
lution is found in the same way as in the case of second-order
equations.
1) We form the auxiliary equation
k"+ a 1 k"- 1 +a.k":-~+ ~ .,: +ti,.,=::.O.
2) We find the roots of the auxiliary equation'
kl, k2, .. , k,..
3) From the character of the roots we write out the particular
linearly independent solutions, taking note of the fact that:
a) to every real root k of order one there corresponds a parti-
cular solution ekx;
b) to every pair of complex conjugate roots k< ,=a+i~ and 1
k' 2 , =a- i~ there correspond two particular solutions eu cos ~x

and ex sin ~x;
c) to every real root k of multiplicity r there correspond r
linearly independent particular solutions
ekx, xekx, ... , xr-lekx;
d) to each pair of complex conjugate roots k< ,=a+i~, 1
k<"l =a- i~ of multiplicity l"' there correspond ~ll particular so-

lutions:
ex cos ~x, xex cos ~x. . , xll-Jex cos ~x.
ex sin px, xex sin ~x, ... , xll-lex sin ~x.
The number of these particular solutio~~ i.s~~~actly equal to the
degree o.f the auxiliary equation (that is, to the orper;,;>f the given
linear differential equation). It may be proved that there solutions
are linearly independent. '
Nonhomogeneous Second-Order Linear Equatiol!s 541.
4) After finding n linearly independent particular. solutions

yl' y., ... , Yn we construct the general solution of the given
linear equation:
y= c.y. +C!y.+ ... + CnYn
where c.. c..... ' en are arbitrary constants.
Example 4. Find the general solution of the equation
YIV -y=O.
Solution. Form the auxiliary equation
k'-1 =0.
Find the roots of the auxiliary equation:
kl=l, k.=-1, k,=t, k,=-l.
Write the complete integral
y=C~e" +C!e-" +A cos x+ 8 slnx,
where C., c. A, B are arbitrary constants.
Note 2. From the foregoing it follows that the whole difficulty

in solving homogeneous linear differential equations with constant
coefficients lies in the solution of the auxiliary equation.
5EC. 23. NONHOMOGENEOUS SECOND-ORDER LINEAR EQUATIONS
Let there be a nonhomogeneous second-order linear equation

y" +a1 y' + a.y = f (x). (I)
The structure of the general solution of such an equation is
determined by the following theorem.
Theorem 1. The general solution of the nonhomogeneous equation
(1) is represented as the sum of some particular solution of the
equation y* and the general solution y
of the corresponding homo
geneous equation
!l+a.y' +a.y=O. (2)
Proof. We need to prove that the sum
y=y + y* (3)
is the general solution of equation (1 ). Let us first prove that the
function (3) is a solution of (I).
y
Substituting the sum + y* into (1) in place of y, we get
<!i+ y*)" +a. Cu + y*)' +a.(y +y*) =f(x)
542 Di(Jereittial Equations
or
(i;" + aJj' + a,y) + (y*" + a 1 y*' + a,y*) = f (x). (4)
Since -y is a solution of (2), the expression in the first brackets
is identically zero. Since y* is a solution of ( 1), the expression
in the second brackets is equal to f (x). Consequently, (4) is an
identity. Thus, the first part of the theorem is proved.
We shall now prove that expression ( 3) is the general solution
of equation (1); in other words, we shall prove that the arbitrary
constants that enter into the expression may be chosen so that
the following initial conditions are satisfied:
Yx=xo
, = Yo' } (5)
Yx=xo=Yo,
no matter what the numbers X 0 , Yo and [provided that x 0 is y;
taken from the region where the functions a 1 , a. and f (x) are
continuous].
Noting that y
may be given in the form
y=C y +C.y., 1 1
where y 1 and y. are linearly independent solutions of equation (2),

and C1 and c. are arbitrary constants, we can rewrite (3) in the
form
y=Clyl +C.y.+y*. (3')
Then, by the conditions (5), we wil'l have*)
ClyiO + c.Yzo + u: = Yo
C1 Y;o + C.u~o + u:' = y~.
From this system of equations we have to determine C 1 and c.
Rewriting the system in the form
(6)
we note that the determinant of this system is the Wronskian

for the functions y 1 and Y. at the point x = X 0 Since it is given
that these functions are linearly independent, the Wronskian is
not zero; consequently, system (6) has a definite solution. !C 1_
*) Here, y10 , y10 , y:, y; 0

, y: 0 , y: denote the numerical values ,ol the
functions y 1 , y1 , y, y;, y:, y' whim -x=x0
Non homogeneous Second -Order Linear Equations 543
and c.; in other words, there exist values C, and c. such that
formula (3) defines the solution of equation (I) which satisfies
tlie given initial conditions. The theorem is completely proved.
Thus,. if we know the general solution y of the homogeneous
equation (2), the basic difficulty, when integrating the nonhomo-
geneous equation(!), )jes in finding some particular solution y*.
We shall give a general method for finding the particular so-
lutions of .a nonhomogeneous equation.
The method of variation of arbitrary constants (parameters).
We write the general solution of the homogeneous equation (2):
y=C,y, +C.y.. (7)
We shall seek a particular solution of the nonhomogeneous
equation (1) in the form (7), considering C, and c. as some (as
yet) undetermined functions of x.
Diffe~e'ntiate' (7):
.. . y'=C,y;+C.it~+C;y,+C~y .
Now choose the needed functi9ns C, and c. so that the following
equation is fulfilled:
(8)
If we 'take note of this additional condition, the first derivative y'
will take the form
y' = c ,y; + c.y;.
Differentiating this expression, we find y":
y" = C,y; +C.y; +C;y; + C~y:.
Putting y, y' and y" into (1), we get
C,y; +C.y:+c;y; + C~y~+ a, (C,y; + C.y;) +
+a. (C,y, +C.y,)=f(x)
or
C, (y; + a,y; + a.y,) +C. (y: + a,y~ + a.y.) + c; y; + C~y; = f (x).
The expressions in the first two brackets vanish, since y, and y.
are solutions of the homogeneous equation. Hence, the latter equa-
tion takes on the form
c;y; + c;y; = f (x). (9)
Thu~ . the function (7) wili be a soi~tion of the .nonhomogeneous
equation (I) provided the functions C, and C1 satisfy the system
. c;~, +
of equations (8) and (9); that _is, if
C~y. = o, c;y; +c;y: = f (x).
544 . Differential Equations
Since the determinant of this system is the Wronskian for the

linearly independent functions y, and y,, it is not equal. to zero.
Hence, in solving the system we will find C~ and C~ as definite
functions of x:
c:
= q>, (x), c~ = q>. (x).
Integrating, we obtain
C, = ~ q>, (x) dx+ C,; C, = ~ q> 2 (x) dx + C,,
where C, and c.
are constants of integration.
Substituting the expressions obtained of C, and C2 into (7), we
find an integral that is dependent on the two arbitrary constants
E, and C,; that is, we find the general solution of the nonhomo-
geneous equation*).
Example. Find the general solution of the equation
y'
y --x=x.
Solution. Let us find the general solution of the homogeneous equation
y_Y' =0.
X
Since
we have In y' =In x+ InC; y' =Cx;
and so
y=C,x+C,.
For the latter expression to be a solution of the given equation, we have
1o define C, and C1 as functions of x from the system
c;x+C~l=O, 2C:x+C~O=x.
Solving this system, we find
c;=;, c:=- }x,
whence, after integration, we get
x - x -
C1 ="2+C 1 , C,=- B+C,.
Putting the functions obtained into the formula y=C,x 2 +Cu we get the
general solution of the nonhomogeneous equation
- - x3 x
y=C,x+c.+ 2 - 6
- - x
or y=C 1x 2 +C,+a, where C, and C2 are arbitrary constants .
.,.) 11 we put C1 = C2 = 0, we get a particular solution of equation (1).

Nonhomogeneous Second-Order Linear Equations 545
When seeking particular solutions, it is useful the take advan-

tage of the results of the following theorem.
Theorem 2. Let the nonhomogeneous equation
y + a 1 y' + a 2 y = f1 (x) + f2 (x) (1 0)
be such that the right side is a sum of two functions, f1 (x) and f. (x).
If Y1 is a particular solution of the equation
y"+a 1 y'+a.y=f1 (X), (11)
and Y2 is a particular solution of the equation
y+a 1 y' +a.y=f2 (X), (12)
+
then Y1 Y. is a particular solution*) of equation (10).
Proof. Substituting the expression Y1 + Y. into (10), we get
(Yt + Y.)" +at (Yt + Y.)' + az (yl +Yz) = ft (x) +f. (x)
or
(y; +a 1y; +a.y 1 ) + (y: + a 1 y~ +a.y =f2) 1 (x) +f. (x). (13)
From equations (11) and (12) it follows that equality (13) is an
identity. And the theorem is proved.
SEC. 24. NONHOMOGENEOUS SECOND-ORDER LINEAR

EQUATIONS WITH CON~TANT COEFFICIENTS
Suppose we have the equation

y" +py' +qy=f(x) (1)
where p and q are real numbers.
A general method for finding the solution of a nonhomogeneous
equation was given in the preceding section. In the case of an
equation with constant coefficients, it_ is sometimes easier to find
a particular solution without resorting to integration. Let us consi-
der several such possibilities for equation (1 ).
I. Let the right side of (1) be the product of an exponential
function by a polynomial; that is, of the form
f (x) = P n (x) ex, (2)
where Pn (x) is a polynomial of degree n. Then the following par-
ticular cases are possible:
*) Obviously, the appropriate theorem remains true lor any number of

terms on the right side.
18-3388
a) The number a is not a root of the auxiliary equation

k 2 +pk-f-q=0.
In this case, the particular solution must be sought for in {he
forrn
y* = (A 0 Xn + A,xn-t + ... +An) e"" = Qn (x} e"". (3)
Indeed, substituting y* into equation (1) and canceJJing e""out
of all terms, we will have
Q: (x} + (2a + p) Q~ (x) + (a + pa+ q) Qn (x) = Pn (x).
2
(4)
Qn (x) is a polynomial of degree n, Q~ (x) is a polynomial of de-
gree n-1, and Q: (x) is a polynomial of degree n-2. Thus,
n-degree polynomials are found on the left and right of the equa-
lity' sign. Eqwiting the coefficients of the same degrees of x (the
+
number of unknown coefficients is n 1), we get a system of n 1 +
equations for determining the unknown coefficients A 0 , A,,
AI, ' An. .
b) The number a is a simple (single) root of the auxiliaryequa-
tion.
If in this case we should seek the particular solution in the
form (3), then on the left side of (4) we would have a polynomial
of degree n-1, since the coefficient of Qn(x), that is, a2 +pa+q
is equal . to zero, and the polynomials Q~ (x) and Q:(x) have deg-
rees less than n. Hence, (4) would not be an identity, no matter
what the A 0 , A 1 , , An. For this reason, the particular solution
in this case has to be taken in the form of a polynomial of degree
n+ 1, but without the absolute term (since the absolute term of
this polynomial vanishes upon differentiation)*):
y* = xQn (x) e"".
c) The number a is a double root of the auxiliary equation.
Then, as a result of the substitution of the function Qn (x) e"" into
the differential equation, the degree of the polynomial is diminished
by two units. Indeed, if a is the root of the auxiliary equation,
then a 2 +pa+q=0; moreover, since a is a double root, it follows
that 2a = - p (since by a familiar theorem of elementary algebra,
the sum of the roots of a reduced quadratic equation is equal to
the coefficient of the unknown in the first degree with sign rever-
+
sed). And so 2a p = 0.
*) We remark that all the results given above also hold for the case
when a is a complex number (this follows from the rules of differentiation of
the function em", where m is any complex number; see Sec. 4, Ch. VII).
Non homogeneous Second-Order Linear Equations 547
Consequently, on the left side of (4) there remains Q~ (x), that

is, a polynomial of degree n- 2. To obtain a polynomial of
degree n as a result of substitution, one should seek the particular
solution in the form of a product of e"x by the (n -l- 2) nd degree
polynomial. Then the absolute term of this polynomial and the
first-degree term will vanish upon differentiation; for this reason,
they need not be included in the particular solution.
Thus, when a is a double root of the auxiliary equation, the
particular solution may be taken in the form
y* = x"Qn (x) ex.
Example l. Find the general solution of the equation
y" +4y' +3y=x.
Solution. The general solution of the corresponding homogeneous equation is
y=C 1e-X+C2e-u.
Since the right-hand side of the given nonhomogeneous equation is of the
form xe 0x (that is, of the form P 1 (x) e0 x), and 0 is not a root of the auxiliary
+
equation k 2 +4k 3= 0, it follows that we should seek the particular solution
in the form y* = Q1 (x) e0 x; in other words, we put
y*=A 0 x+A 1
Substituting this expression into the given equation, we will have
4A 0 +3 (AoX+ A 1)=x.
Equating the coefficients of identical degrees of x, we ge~
3A 0 =l, 4A 0 +3A 1 =0,
whence
Consequently,
1 4
Y =3x--g.
The general solution of y = !i + y* will be
Y =CI e-x+C ze-ax+..!._

3
x-..!9
Example 2. Find the general solution of the equation
yn + 9y = (x'+ I) e3 x.
Solution. The general solution of the homogeneous equation is rea~ily
found:
y=C 1 cos 3x+C2 sln3x.
The .right side of ~he given equation (x 2 + l) e x 3 has .the form
P 2 (x)e 3 x,
18*
Since the coefficient 3 in the exponent is not a root of the auxiliary equa-
tion, we seek the particular solution in the form
y = Q2 (x) e'"' or y = (Ax 2 + Bx= C) e8"'.
Substituting this expression in the differential equation, we will have
(9 (Ax 2 + Bx+C>+ 6 (2Ax+ B)+2A +9 (Ax'+ Bx+ C)] e8"' = (x 2 + l)e1 "'.
Cancelling out e8"' and equating the coefficients of identical degrees of x, we
obtain
18A=1, 12A+188=0, 2A+6B-{-18C=1,
1 1 5
whence A= ; B = - 2't; C = 8I. Consequently, the particular solution is
18
-(118
y - x2- 1
27 x
+5)
8I eIX
and the general solution is
y=C 1 cos 3x+Ca sin 3x+ (t18 x 2 -bx+~) e'"'
Example 3. To solve the equation
+
y- 7 y' 6y = (x- 2) e"'.
Solution. Here, the right side is of the form P 1 (x) e1"' and the coefficient 1
in the exponent is a simple root of the auxiliar polynomial. Hence, we seek
the particular solution in the form y* = xQdx) e or
y =X (Ax+ B)e"';
puttii:Ig this expression in the equation, we get
[(Ax'+Bx)+ (4Ax+2B)+2A-7 (Ax 2 + Bx)-7 (2Ax+B)+
+
6 (Ax'+ Bx)] e"' = (x-2) ex
or
(-lOAx-58 +2A) e"' =(x-2)e"'.
Equating the coefficients of Identical degrees of x, -we gel
-lOA=l. -5B+2A=-2,
1
whence A=- , 8=: .. Consequently, the particular solution is
10 5
y* =X ( - 110 X-{-~) e"'
and the general solution Is
y= C1eex 2
I 259) ex .
+ C ex +x ( -rox+
II. Let the right side have the form
+
1 (x) = P (x) eu cos ~x Q (x) e~"' sin ~x, (5)
where P (x) and Q (x) are polynomials. , . , ,
This case may be considered by_ the technique used in the pre-
ceding case, if we pass from trigonometric functions to exponential
Nonhomogeneous Second-Order Linear Eq11ations 549
functions. Replacing cos ~x and sin ~x by exponential functions

using Euler's formulas (see Sec. 5, Ch. VII), we obtain
f (x) = p (x) eox ei~x ~ e-i~ + Q (x) ex ei~x -;ie-i~x
or
f (x) = (; P (x) + ~~ Q (x) J e<+i~> x + (; P (x)-~ Q (x)J e<a-i~>x. (6)
Here, the square brackets contain polynomials whose degrees are

equal to the highest degree of the polynomials P (x) and Q (x).
We have thus obtained the right side of the form considered in
Case I.
It is proved (we omit the proof) that it is possible to find p-ar-
ticular solutions which do not contain complex numbers.
Thus, if the right side of equation (1) is of the form
f (x) = P (x) eox cos ~X+ Q (x) eax siri ~x, (7)
where P (x) and Q (x) are polynomials in x, then the form of the
particular solution is determined as follows:
a) if the number a+ i~ is not a root of the auxiliary equation,
then the particular solution of equation (1) should be sough tin the
form
y* = U (x) eax cos jix + V (x) eox sin ~X, (8)
where U (x)_ and V (x) are polynomials of degree equal to the high-
est degree of the polynomials P (x) and Q (x);
b) if the number a+ i~ is a root of the auxiliary equation,
we then write the particular solution in the form
y* = x [U (x) eax cos ~X+ V (x) eax sin ~x]. (9)
Here, in order to avoid mistakes we must note that these forms
of particular solutions, (8) and (9), are obviously retained when
one of the polynomials F(x) and Q (x) on the right side of equa-
tion (I) is identically zero; that is, when the right side is of the
form
P (x) ex cos ~x or Q (x) eu sin ~x.
Let us further consider an important special case. Let the right
side of a second-order linear equation have the form
f(x)=Mcos~x+Nsin.~x, (7')
where M and N are constants.
a) If ~i is not a root of the auxiliary equat~on, the p~rticula~
solution should be sought in the form -
y* =.A co:;> ~x + B sin ~x. (8')
550 .. .Dif}erel).tial Equations
b) If ~i is a root of the auxiliary equation, then the particular

solution should be sought in the form
y*=x(Acos~x+Bsin~x). (9')
We remark that the function (7') is a special case of the func-
tion (7) [P (x) = M, (Q) x = N, a= 0]; the functions (8') and (9')
are special cases of the functions (8) and (9).
Example 4. Find the complete integral of the nonhomogeneous linear equa-
tion
y" + 2r/ + 5y = 2 cos x.
Solution. The auxiliary equation k2 +2k+5=0 has roots k 1 =-1 +2i;
k 2 = - l-2i. Therefore, the complete integral of the corresponding homoge-
neous equation is
y=e-"(C 1 cos2x+Ctsln2x).
We seek the !>articular solution of the nonhomogeneous equation in the form
y*=A cosx+8 sinx,
where A and 8 are constant coefficients to be determined.
Putting y* into the given equation, we will have
-A cos x- 8 sln x + 2 (-A sin x + 8 cos x) + 5 (A cos x + 8 sin x) = 2 cos x.
Equating the coefficients of cos x and sin x, we get two equations for de-
termining A and 8:
-A+28+5A=2; -8-2A+58=0,
I I
whence A=
5: 8=w.
The general solution of the given equation is y=y+y*, that is,
. I I
y=e-" (C 1 cos 2x+ C2 sin 2x)+ cosx+m slnx.
5
y"+4y=cos2x.
Solution. The auxiliary equation has roots k 1 =2i, k 2 =-2i; therefore,
the general solution of the homogeneous equation is of the form
y=C 1 cos 2x+C 2 sin 2x.
We seek the particular solution of the nonhomogeneous equation in ihe form
y* =X (A cos 2x + 8 sin 2x).
Then
y*' = 2x (- A sin 2x + 8 cos 2x) +(A cos 2x + 8 sin 2x),
y*" = - 4x(- A cos 2x-8 sin 2x)+4 (-A sin 2x+8 cos 2x).
Putting these expressions of the derivatives into the given equation and
equating the coefficients of cos 2x and sin 2x, we get a system of equations
for determining A and 8:
48=1: -4A=0,
High<rOrder Nonhomogeneous Linear Equations 551
whence A= 0 and 8 = !. Thus, the complete integral of the given equation is
y=C, cos 2x+C 2 sin2x +~x sin 2x.

y" -y=3eu: cosx.
Solution. The right side of the equation has the form
f (x) =e2" (M cosx +N sip x),
and M=3, N=O. Theauxiliaryequation k 2 -1=0has roots k,=l, k 2 =-1.
The general solution of the homogeneous equation is
y=C,e" +C e-". 2
Since the number a+i~=2+il is not a root of the auxiliary equation, we

seek the particular solution in the form
y* =e 2" (A cos x + 8 sin x).
Putting this expression into the equation, we get (after collecting like terms)
+ +
(2A 48) e2" COS X+ (- 4A 28) e2" sin X= 3eu: COS X.
Equating the coefficients of cosx and sinx, we obtain
2A+48=3, ,...;..4A+28=0.
and 8 = ~ . Consequently, the particular solution is
3
Whence A=
10
y* = e2
" ( ~ cos x+ ~ sin x) ,
y=C,e"+C 2e-"+eu: (~ cosx+: slnx).
SEC. 25. HIGHER-ORDER NONHOMOGENEOUS LINEAR EQUATIONS
Let us consider the equation

+
y<"J a,y<"-' 1+ ... +any =f (x), (1)
where a,, a 2 , , an, f (x) are continuous functions of x (or con-
stants).
Suppose we know the general solution
y=C,y, +CzYz+ +CnYn (2)
of the corresponding homogeneous equation
y<"l + a,y<n-1) + a!y{Tl-2) + ... +any= 0. (3)
As in the case of a second-order equation; the following asser-
tion holds for equation (1).
552 DiUerenUal Equations
Theorem. If y is the general solution of. the homogeneous equa-

tion (3) and y* is a particular solution of the nonhomogeneous
equation ( 1 ), then
Y=y+y*
is the general solution of the nonhomogeneous equation.
Thus, the problem of integrating equation (1), as in the case
of a second-order equation, reduces to finding a particular solution
of the nonhomogeneous equation.
As in the case of a second-order equation, the particular solution
of equation (1) may be found by the method of variation of para-
meters, considering C,, c., ... , Cn in e~pression (2) as functions
of x.
We form the system of equations (cf. Sec. 23):
l
C1 Y
(n-)
' (n-1) ' (n-1)

C,y,. +C.y.
+ C., (~--)+ .
. Y
nYn
' (n-1)
++CnYn
- ,
+-C, (,;_;) ~0 .
=f(x).
ll (4)
Th'is system of equations with the unknown functions c;, C~ ... ,

C~ has very definite solutions. (The determinant of the coef-
ficients of c;, C~ ... , C~ is the Wronskian formed for the parti-
cular solutions y., y,, ... , Yn of a homogeneous equation, and
since these particular solutions are, by definition, linearly inde-
pendent, the Wronskian is not zero.)
Thus, the system (4) may be solved for the functions c;, C~, ... ,
C~. Finding them and integrating, we obtain
C1 = ~ c; dx+C.; c.=~ c~ dx+c.; ; Cn= ~ C~ dx+Cn,
where c" c...... c n are the constants of integration.
We shall prove that in- such_ a_ case the expression
y*=C,y, +C,y.+,. .. +CnYn (5)
is the general solution of the nonhomogeneous equation (1).
Differentiate expression (5) n times, each time taking into account
equations (4); this yields
y* =C 1y 1+C.y.+C,y,+ ... +CnYn
y*' =C,y~ +C,y~ +C,y;:+ ... +CnY~ .
............ . ..........
y*<n-1) = C 1 y~n--1) +C,y~n-1) + ~, . +CnY<~_-1>,
y*<"l = C yln) + C:!i~n) + .. >+C~y~n) +f (x),
1
Higher-Order Nonhomogeneous Linear Equations 553
Multiplying the terms of the first, second, ... and, finally, second
to the last equation by an, a,;_,, ... , aP respectively, and adding,
we get
+
y*<n) a,y'*<n-1) + ... +any*= f (x),
since y 11 Yz ... , Yn are particular solutions of the homogeneous
equation; for this reason, the sums of the terms obtained in adding
vertical columns are equal to zero.
Hence, the function y*=C,y,+ ... +CnYn [whereC" ... ,Cn
are functions of x determined from equations (4)) is a solution of
the nonhomogeneous equation (1), and since this solution depends
on the n arbitrary constants CP C1 , , Cn it is the general
solution.
The proposition is thus proved.
For the case of a higher-order nonhomogeneous equation with
constant coefficients (cf. Sec. 24), the particular solutions are found
more easily, namely:
I. Let there be a function on the right side of the differential
equation: f(x)=P(x)eu, where P(x) is a polynomial in x; then
we have to distinguish two cases:
a) if a is not a root of the auxiliary equation, then the parti-
cular solution may be sought in the form
y*=Q(x)eu,
. where Q (x) is a polynomial of the same degree as P (x), but with
undetermined coefficients;
b) if a is a root of multiplicity ~t of the auxiliary equation,
then the particular solution of the nonhomogeneous equation may
be sought in the form
y* = x"Q (x) eu,
where Q (x) is a polynomial of the same degree as P (x).
II. Let the right side of the equation have the form
f(x) = M cos Px+ N sin px,
where M and N are constants. Then the form of the particular
solution will be determined as follows:
. a) if the number .pi is not a root of the auxiliary .equation, then
the particular solution has the form
y* =A cos Px+B sin px,
where A and B are constant undetermined coefficients;
b) H the. number ~~ is a root of the auxiliary equation of mul-
tiplicity I' then .
+
y* = x" (A cos ~x B sin px).
554 .DifJereMwl Bqu<Jtions
III. Let
f (x) = P (x) e"" cos~x + Q (x) e"" sin ~x,
where P (x) and Q (x) are polynomials in x. Then:
a) if the number a+ ~i is not a root of the auxiliary polynomial,
then we seek the particular solution in the form
y* = U (x) e"" cos ~x+ V (x) e"" sin ~x,
where U (x) and V (x) ar~ polynomials of degree equal to the high-
est degree of the polynomials P (x) and Q (x);
b) if the number a+ ~i is a root of multiplicity Jl of the auxiliary
polynomial, then we seek. the particular solution in the form
y* = x"' [U (x) e"" cos ~x + V (x) eux sin ~x],
where U (x) and V (x) have the same meaning as 'in Case a:
General remarks on Cases II and Ill. Even when the right side
of the equation contains an expression with only cos ~x or. only
sin ~x, we must seek the solution m the form indicated, that is,
with sine and cosine. In other words, from the fact that the right
side does not contain cos ~x or sin ~x. it does not in the least
follow that the particular solution of the equation does not contain
these functions. This was evident when we considered Examples -4,
5, 6 of the preceding section, and also Example 2 of the present
section.
Example I. Find the general solution of the equation
YIV -y=xa+ 1.
Solution. The auxiliary equation k'-1=0 has the roots

k,=l, k.=-1, k.=i, k,=-i.
We find the general solution of the homogeneous equation (see Example 4,
Sec: 22):
y= C,e" +C e-" +Ca cosx+C, sinx.
2
We seek the particular solution of the nonhomot:eneous equation in the form

y* = A 0 x 3 + A,x'+ A x+ A
2 0_
Differentiating y* lour times and substituting the expressions obtained

into ihe given equation, we get
- A0 x- A,x- A2 x- A 0 =x 1 +I.
Equating the coefficients of identical degrees. of x, we have
-A 0 =1; -A 1 =0; -A 2 =0; -A 0 =1.
l-ienee
The Ditferenttal Equation of Mechanical Vibrations '555
The complete integral of the nonhomogeneous equation is found from tho

formula y=y+y*:
y=C,e" +C.e-" +C, cosx-t-C, slnx-x'- l.
Example 2. To .solve the equation
y 1V -y=5 cosx.
Solution. The auxiliary equation k'-1 = 0 has the roots k 1 =I, k = - t,
k 3 =i, k,=- i. Hence, the general solution of the corresponding homogeneous
equation is
y = C,e" +C.e-" +C, cos x+C, slnx.
Further, the right side of the given nonhomogeneous equation has the form
f(x)=M cosx+N slnx,
where 111=5 and N=O.
Since i is a simple root of the auxiliary equation, we seek the particular
solution in the form
y*=x(A cosx+8sinx).
Putting this expression into the equation, we find
4A sinx-48 cosx=5 cosx,
whence
4A=0, -48=5
5
or A= 0, 8 = - . Consequently. the particular sol uti~n of the differential
4
equation is
y * = - 5 xsmx
.
4
y=C,e"+C1e-"+C,cosx-t-C,sinx-! xsinx.
SEC. 26. THE DIFFERENTIAL EQUATION OF MECHANICAL

VIBRATIONS
In this and the following sections we shall consider a problem

in applfed mechanics, and investigate and solve it by means of
linear differential equations.
Let a load of mass Q be at rest on an elastic spring (Fig. 268).
We denote by y. the deviation of the load from the equilibrium
position. We shall consider deviation downwards as positive,
upwards as negative. In the equilibrium position, the force of the
weight is balanced by the elasticity of the spring. Let us suppose
that the force that tends to return the load to equilibrium (the
so-called r~storing force) is proportional to the deflection, that is,
equal to ky, where k is some constant for the given spring (the
so-called "spring rigidity")*).
*) Springs whose restoring force is proportional to the deflection are called
springs with a "linear characteristic".
Let us suppose that the motion of the load Q is restricted by

a resistance force operating in a direction opposite to that of
motion and proportional to the velocity of motion of the load
relative to the lower point of the spring; that is, a force -A.v=
=-A.~r. where A.=const>O (shock absorber). Write the dif-
ferential equation of the motion of the load on the spring. By
Newton's second Jaw we have
(1)
(here, k and A. are positive numbers). We thus have a homoge-

neous linear differential equation of the second order with con-
stant coefficients.
0
postrlfri I
EjjliilibrlfJ!P.. Equilibrium position
- --=1#---------
Fig. 268.
I Fig. 269.
t
This equation may be rewritten as follows:

d 2y dy
([t2+ Pdf+ qy = 0, (1 ')
where
A. k
p=Q; q=([
Let it further be assumed that the lower point of the spring A
executes vertical motions 1:1nder the Jaw z = <p (/). This will occur,
for instance, if the lower end of the spring is attached to a rol-
ler, which moves over an uneven spot tog~the~ with the spring
and the load (Fig. 269).
In this case the restoring force will be equal not to -ky, but
to -k [y+<p(t)], the force of resistance wi.ll. be -A. [y' +<p' (t)],
and in place of equation (1) we will have the equation
d 2y dy '. . ;'' ..
Qdt' +A.dt +ky=-k<p(t.)-.Aifl!(f): (2)
Free Oscillations 557
or
d2y
dt 2 + p dtdy + qy = f (t), (2')
where
f (t) = - k<p (t) + A.<p' (t)
Q
We thus have a nonhomogeneous second-order differential
equation.
Equation (l') is called an equation of free oscillations, equation(2')
is an equation of forced oscillations.
SEC. 27. FREE OSCILLATIONS
Let us first consider the equation of free oscillations

y~' +py' +qy=O.
We write the corresponding auxiliary equation
k"+pk+q=O
and find its roots:
p "'fp-- p "'j-pz-
k,=-2+ V 4-q; .kz=-2- V T-q.
1) Let ~~ > q. Then the roots k 1 and k 2 are real negative num-
bers. The general solution is expressed in terms of exponential
functions:
(l)
From this formula it follows that the deviation of y for any
initial conditions approaches zero asymptotically if t _,. oc. In the
given case, there will be no oscillations, since the forces of resist-
ance are great compared to the coefficient of rigidity of the
spring k.
2) Let ~~ = q; then the roots k 1 and k. are equal (and are
also equal to the negative number - ~ ) . Therefore, the general"
solution will be
-~ t . -~t _~!!_
y . C,e +C.te 1
. (C 1 +C 2 t)e (2)
Here the deviation also approaches zero as t-+ oc, but not so
rapidly as in the preceding 'case (due to the factor C1 +Cat).
558 DifJerenlial. Equations
3) Let p = 0 (no resistance). The auxiliary equation is of -the

form
1
k +q=O,
and its roots are k, = ~i; k 2 = - ~i, where ~ = VQ". The general
solution is
(3)
In the latter formula, we replace the arbitrary constants C 1
and C1 with others. We introduce the constants A and q> 0 , which
are connected with C, and C1 by the relations
C,=A sinq>o, Cz=ACOSIJ>o
A and q> 0 are defined as follows in terms of C, and C 2 :
A= y C~ + C~, 'Po = arc tan 2.
I
Substituting the values of C, and C 2 into formula (3), we get

y =A sin q> 0 cos~~+ A cos q> 0 sin ~t
or
y= A sin {~t-1-q> 0 ). (3')
These oscillations are called harmonic. The integral curves are
sine curves. The time interval T, during which the argument of
the sine varies by 2n, is
lj called the period of oscil-
2
lation; here, T = ; The
t frequency is the number
of oscillations during time
2n; here, the frequency is
Fig. 270. ~; A is the greatest de-
viation from equilibrium
and is called the amplitude; 'Po is the initial phase. The graph
of the function (3') is shown in Fig. 270~
4) Let p =1= 0 and ~ <q.
In this case, the roots of the auxiliary equation are complex
numbers:
k 1 =a+ i~, k 2 = a - i~,
a=-f<O, ~= yq-~
where 2
The complete integral has the form

y = e" 1 {C 1 cos~~+ C2 sin t3t} (4)
Forced Oscillations 559
or
y = Ae~ 1 sin (~t + rp 0). (4')
Here, for the amplitude we have to consider the quantity Aeat
which depends on the time. Since a< 0, it approaches zero as
g
----- t
,. ,.
/
/
/
/
Fig. 271.
t __.... oo, which means that here we are dealing with damped oscil-
lations. The graph of damped oscillations is shown in Fig. 271.
SEC. 28. FORCED OSCILLATIONS

The equation of forced oscillations has the form
y" + py' + qy = f (t).
Let us consider an important practical case when the disturb-
ing external force is periodic and varies under the law
f(t).:.._a sin rot;
then the equation will have the form
y"+py'+qy=asinrot. (1)
1) Let us first presume that p =1= 0 and that is, the ~~ < q,
roots of the auxiliary equation are the complex numbers a i~.
In this case [see formulas (4) and (4'), Sec. 27], the general
solution of the homogeneous equation has the form
y
= Ae111 sin (~t tp 0 ). +
(2)
We seek a particular solution of the nonhomogeneous equation
in the form
y* = M cos rot N sin rot. + (3)
560 DiUerential quat ion~
Putting this expression of y* into the original differential equa-

tion, we find the values of M and N:
M_ -pwa . N _ (q-w")a
- (q-w 2 ) 2 +p 2 w2 ' - (q-w 2 ) 2 +p 2w
Before putting these values of M and N into (3), let us introduce
the' new constants A* and <p*, setting
M = A* sin <p*, N = A* cos cp*,
that is
A* = V M + N = f(q-wa
2
2 2
) +p 2 w2
, tan <p* = MN. ,
Then .the particular solution of the nonhomogeneous equation may

be written in the form
y* =A* sin <p* cos wt +A* cos <p* sin wt =A* sin (wt + <p*),
or, finally,
y* = a2 2 sin (rot+ <p*).
V (q- w + p )
2
w2
The complete integral of equation (1) i!) y=y+y* or
y = Ae-~ 1 sin (~t + <p 0 ) + V (q- w)z
a
+ p'wz sin (rot + <p*).
The first term of the sum on the right side (the solution of the
homogeneous equation) represents damped oscillations; it dimini-
shes with increasing t and, consequently, after some interval of
time the second term (which determines the forced oscillations)
will acquire prime importance. The frequency ro of these oscilla-
tions is equal to the frequency of the external force f (t); the
amplitude of the for<;ed vibrations is the greater, the less p and
the closer ro2 is to q.
Let us investigate more closely the dependence of the amplitude
of forced vibrations on the frequency w for various values of p.
For this,- we denote the amplitude of forced vibrations by D (ro):
D(w)= a
Y(q-w)+pw
Putting q . ~: (for p . 0, ~~ would be eqiJal t9 its natural fr~quen

cy), we have
a
Forced Oscillations 561
Introducing the notation

(J) p
j3;-=A.; r=Y,
1'1 .
where A. is the ratio of the frequency of the disturbing force to

the frequency of free oscillations of the system, and the constant y
is independent of the disturbing force, we find that the magnitude
of the amplitude will be expressed by the formula
D (A.)= a (4)
p~ Y(l-A,2)Z+yA,2
Let us find the maximum of this function. It will obviously be

for that value of A. for which the square of the denominator ha.;
a minimum. But the minimum of the function
(5)
is reached when
and is equal to
Hence, the maximum amplitude is equal to

a
The graphs of the function D (A.) for various values of y are shown
in Fig. 272 (in constructing the graphs we put a= 1, ~. = 1 for
the sake of definiteness). These curves are called resonance curves.
From formula (5) it follows that for small y the maximum
value of amplitude is attained for values of .A close to unity, that
is, when the frequency of the external force is close to the fre-
quency of free oscillations. If y=O (thus, p=O), that is, if there
is no resistance to motion, the amplitude of forced vibraUons
increases without bound as A-+ 1 or as w-.. ~~ = Jlq:
lim D (A.)= ..oo.
~.....
('\''"'0),
At ci/= q we have resonance,

562" Diflerentlat Equations
2) Now let us suppose that p = 0; that is, we consider the

equation of elastic oscillations without resistance but with a
periodic external force:
y" + qy =a sin wt.
0().)
5.0~-.---,---.-.rr.--.--.---.---~~
A, r
4.5t----1----+---+-+lf-H+-+---+--+---l-----l-.
,,
4.0t----+-+--+-+6~1-t+-+-+--l---+--+
I..
<i.5~-t----+---+-t+--l--++--+----+----4---1---+
~ot-----1~--+---+~,~N'~~I----1---+--~--4
.{
'l ~
2.5'r-4--+-~h~~~\~~~+-~-+-4
zort---r-~--~~~~~--~--4---+---+
~~~ 1\~,\
Uit----t~--izj~~"-~~~[---+-+--1-~
f.O~=t------+--+---+---+~~~~1---+-+---+
0.5+--t--+-+--t--+--+-~.....~-.......;::---+--+
~
0 0.25 0.5 Q75 1.0 1.25 f.!i t75 2.0 2.25' A.
Fig. 272.
The general solution of the homogeneous equation is

y=C 1 cos~t+C.sin~t (~"=q).
If ~ =1= w, that is, if the frequency of the external force is not
equal to the natural frequency, then the particular solution of
the nonhomogeneous equation will have the form
y* = M cos wt + N sin wt. (6)
Putting this expression into the original e_quation, we find
a
M=O, N=--. q-ro
The general solution is
+
y =A sin (~t + cp 0 ) q a ro sin wt.
Systems of Ordinary DlDerential Equations 563
Thus, motion results from the superposition of a. natural Qscilla-

tion with frequency ~ and a forced vibration with frequency w.
If ~ = w, that is, the natural frequency coincides with the
frequency of the external force, then function (3) is not a solu-
tion of equation t6). In this case,
in accord with the results of Sec.
24, we have to seek the particular /
a /
solution in the form y---
2f3
t cos j3t /
/
+
y* = t (M cos wt N sin wt). (7) a=const
Substituting this expression into J3=const //
the equation, we find M and N: /
M=- 2: ; N=O.
Consequently,
y* = - 2~ t cos wt.
The general solution will have
the form '
y =A sin (~t + cp 0 ) - 2: t cos ~t.
''
''
The second term on the right '
side shows that in this case the Fig. 273.
amplitude increases without bound
with the time t. This phenomenon, which occurs when the
natural frequency of the system coincides with the frequency of
the external force, is called resonance.
The graph of the function y* is shown in Fig. 273.
SEC. 29. SYSTEMS OF ORDINARY DIFFERENTIAL EQUATidNS
In the solution of many problems it is required to find the
functions y, = Y, (x), Y2 = Y. (x), ... , Yn = Yn (x), which satisfy a
system of differential equations containing the argument x, the
unknown functions yP y 2 , , Yn and their derivatives.
Consider the following system of first-order equations:
dy,
dx
=f (
' x, y,, Yz .. I Yn), 1
dy.
dx =
f z (X, Yp Yz Yn), ~ (I)
:~n=fn.(x: ~''. y~, .. , Yn), J

564 Di(Jerential Equations
where yl' Ys ... , y,.. are unknown functions and x is the argu-
ment.
A system of this kind, where the left sides of the equations
contain first-order derivatives, while the right sides do not
contain derivatives, is called normal.
To integrate the system means to determine the functions
y,. y 2 , , y,., which satisfy the system of equations (l) and the
given initial conditions:
{yl)x=xo = Yto (Yz)x=xu = Yzo (Y,.)x=xo = Yno (2)
Integration of a system like (1) is performed as follows. Differen-
tiute the first equation of ( 1) with respect to x:
dy,_ at,--!- at, dy,
dx -ax . ay, dx
+ . . . +~ dy,.
ay,. dx .
.
R ep Iacmg th e d enva . t'tves dy, dy. , ... , dX
dx , dx dy,. 'th th etr
WI
. ex-
pressions {,, {2 , , f,. from equations (l ), we get the equation
d"y,_
dx - F (x, y,, ... , y,.).
Differentiating this equation and then doing as bef.:)re,. we obtain
d'y,
dx = F' (x, y,, Yz y,.).
Continuing in the same fashion, we finally get the equation
d"y,
dx" = F,. (x, y,, ... , y,.).
We thus get the following system:
1
dy,
dx = f 1 (X Yt
dy,_ '
. Yn)

~X~. ~F.! (~, .y'.' . ... y,.),

d"y,
dx" = F,. (x, YP . , !1,.).
I .
(3)
From the first n-l equations we determine (if this is possible)

y1, y,, ... , y,. and express them in terms of x, y, and the
. .' dy, d'y, d"-'y,
denvattves iiX, ifX!, ... , dx"-'
Y. = IPz (x, y" y~ ... , y~n -ll),
Ya = IPa (x, y,, y:, ... ' Y\n-1)),
Yn =<p,. (x, Yu y~, y~n -ll).

} (4)
Systems of o;dinary Differential Equations 565
Putting these expressions into the last of the equations (3), we

get an nth-order equation for determining .Y.:
dnq, <D (X 'I y .1 ' Y,(n -.1 l) (5)
dxn = '"' '
Solving this equation, we find y 1 :

y, =', (x, C" C2 , , Cn). (6)
Differentiating the latter expression n-1 times, we find the
ttves dx
.d ettva dy, dy, dn-'y, f t' f C C C
, dx , ... , d~n-l as unc tons o x, 1, 2 , , n
Substituting these functions into equations (4), we determine
Yz Ya Yn:
y.='.(x, C., C 2 , , Cn),}
(7)
Yn='i'n(X, C., C,, ... , Cn)
For this solution to satisfy the given initial conditions (2), it
remains for us to find [from equations (6) and (7)] the appro-
priate values of the constants CJ' c., ... , Cn (like we did in the
case of a single differential equation).
Note l. If the system (I) is linear in the unknown functions,
then equation (5) is also linear.
Example I. Integrate the system
dy =y+z+x d.z =-4y-3z+2x (a)
dx ' dx
with the initial conditions
(b)
Solution. 1) Differentiating the first equation with respect to x, we have
d y_dy+~+1
2
dx- dx dx
Putting the expressions :~ and ~; from equations (a) into this equation,
we get
d2
d:.=(y+z+x) +<- 4y-3z +2x) + 1
or
dy
dx 2 =-3y-2z+3x+1. (c)
2) From the first equation of system (a) we find

ay
z,= dx -.y- (d)
566 Differenfial Equations
and put it into the equation just obtained; we get
d"y=-3y-2 (dy_y-x)+3x+l
dx dx
or
d 2y dy
dx 2 + 2 dx +Y= 5x+!. (e)
The general solution of this equation is

y=(C 1 +C.x) e-x +5x-9
and from (d) we have
(g)
Choosing the constants C1 and C2 so that the initial conditions (b) are
satisfied,
(Y)x=o= l. (Z)x=o=O,
we get, from equations (f) and (g),
I =C 1 -9, 0=C1 -2C 1 + 14,
whence C, = 10 and C2 =6.
Thus, the solution that satisfies the given initial conditions {b) has the
form
y=(10+6x)e-x+5x-9, z=(-14-12x)e-x-6x+14.
Note 2. In the foregoing we assumed that from the first n-1
equations of the system (3) it is possible to determine the
functions y.. Y ... , Yn It may happen that the variables
y ... , Yn are eliminated not from n, but from a smaller number
of equations. Then to determine Y1 we will have an equation
of order less than n.
Example 2. Integrate the system
dx dy dz
dt=y+z; dt =x+z; d{=x+y.
Solution. Differentiating the first equation with respect to t, we find

d 2x dy dz.
dtz=dt +di=(x+z)+(x+y),
dx
dt 2 =2x+y+z.
Eliminating the variables y and z from the equations

dx d"x
df=y+z; dt =2x+y+z,
we_ get a second-order equation in x:

d1x dx .
dt 1 !_dt -2x=O.
Systems of Ordinary Dl:fJeretitial Equations 567
-Integrating this equation, we obtain its general solution:

x=C~e-I+Ctezt, (a)
Whence we find
1e- + C1e + 2C etl -z.

dx
df=-C t 2 2l dx C e- t
and y=df-z=- 1 2 (~)
Putting into the third of the given equations the expressions that have been
found for x and y, we get an equation for determining z:
dz+ z= 3C 2ezt .
{it

z=C 8e- 1 +C2e2t, ('\')
But then, from equation <M.we get
y=- (C 1 +C2 ) e- 1 +C 2e2 t,
Equations (a), (~). and ('\') give the general solution of the given system.
The differential equations of the system can contain higher-order

derivatives. This then yields a system of differential equations of
higher order.
For instance, the problem of the motion of a material point under the
action of a force F reduces to a system of three second-order differential
equations. Let F"' F.Y' Fz be the projections of the force F on the coordinate
axes. The position or the point at any instant of time t is determined by
its coor.dinates x, y, and z. Hence, x, y, z are functions of t. The projections
of the velocity vector of the point on the axes will be :; ,
Suppose that the force F and, hence, its projections Fx, FY' F ~ depend
, .
~r ~:
on the time t, the position x, y, z of the point, and on the velocity of
mo t'1on of th e pom
t , th a t 1s, dx , dy
on dt dz .
dt , {it
In this problem the following three functions are the sought-for functions;
X=x (1), y=y (t), 2=2 (I).
These functions are determined from equations of dynamics (Newton's law):

dtx (
md 1,=Fx l,x,y, Z,
dx dy dz)
dt'dt'dt'
a,, (I, x, y, dx dy dz)
m dti =Fy
2
m ddtzz =Fz (
2,_
dt > dt
dx
dy dz)
t, x, y, z, dt'dt'dt.
I dt I
1
L
(8)
We thus have a system of three second-order differential equations. In the

case of plane motion, that is, moti()n in which the trajectory is a plane
568 DifJerential Equations
curve (lying, for example, in the xy-plane), we get a system of two equations
for determining the functions x (t) and y (t):
m:;~=Fx
2
(t, X, y, :; , :n, (9)
d y ( dx dy) (10)
mdt=Fy t, x, y, iH' dT .
It is possible to solve a system of differential equations. of higher order
by reducing it to a system of first-order equations. Using equations (9) and
(10) as examples, we shall show how this is done. We introduce the notation
dx dy
dt=U, g=V.
Then
d'x du
dF=dt'
The system of two secondorder equations (9) and (10) with two unknown
functions x (t) and 1J (t) is replaced by a system of four first-order equations
with four unknown functions x, y, u, v:
dx _
dt=U,
dy
dt = .v,
du
md[=Fx(t, x, y, u, v.),
dv
mdt=fy(t, x, y, u, v).
We remark in conclusion that the generaf method that we have considered

of solving the system may, in certain specific cases, be replaced by some
artificial technique that gets the result faster.
Example 3. To find the general solution of the following system of
differential equations:
d2 z
dx=y.
Solution. Differentiate, with respect to x, both sides of the first equation
twice:
d"z
But dx=g, and so we get a fourth-order equation:
dy
dx=y. ;. .. '
Integrating this equation, we obtain lts general solution (see Sec. 22,
Example 4):
Systems of Linear Differential Equations 569
Finding d:.
d' .
from this equation and putting it into the first equation, we
find z:
SEC. 30. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS

WITH CONSTANT COEFFICIENTS
Suppose we have the following system of differential equations:
~. 11 X 1 +a 12X 2+...
dt=a + a ,.Xm 1 l
dx = az.x. + auxz + ... + az,.x,.,
Tt 2
~ (1)
a I 0 I I 0 o
J
where the coefficients ail are constants. Here, t is the argument,
and X1 (t), x, (t), ... , x,. (t) are the unknown functions. The
system (I) is a system of homogeneous linear differential equations
with constant coefficients.
As was pointed out in the preceding section, this system may
be solved by reducing it to a single equation of order n, which
in the given instance will be linear (this was indicated in Note
I of the preceding section). But system (1) rriay be solved in
another way, without reducing it to an equation of the nth order.
This method makes it possible to analyse the character of the
solutions more clearly.
We seek a particular solution of the system in the following
form:
(2)
It is required to determine the constants a., a 1 , , a, and k

in such a way that the functions a.ekt, a 2 ekt, ... , a,.ekt should
satisfy the system of equations (1). Putting them into syste;m (1),
we get:
111
ka.e = (a 11 a 1 -1- a 11 a 2 + ... + a.,a,) ekt, }
ka..ekt = (all a.. +anal + ... + azna.,) ekl,
o o 0 I o
.
I o o o
. I I I o
~

: :
I I
(2)
ka.,.ekt = (an,at +anzaz + ... +a,.,.a.")ekt.
570 DiOerential Equations
Cancel out ekt. Transposing all terrris to one side and collecting
coefficients of a,, a 2 , , an, we get a system of equations:
(a 11 - k) a,+ a 12 a 2 + ... :a, nan= 0, }
a2lal +(au-k) a!+ ... ,a:nan -0,
. . . . . . . . . . . . . (3)
an,a, + an:a1 + ... +(ann- k) an= 0.
Choose a,, a 2 , , an and k such that will satisfy the system (3).
This is a system of linear algebraic equations in a,, a 2 , , an.
Let us form the determinant of the system (3):
a 11 -k a 12 ... a,n
an au-k .. atn
1!:. (k) = (4)
an, an2 (ann- k)
If k is such that the determinant 1!:. is different from zero, then
the system (3) has only trivial solutions u, = a 2 - ; =an= 0
and, hence, formulas (2) yield only trivial solutions:
X1 (t) =X1 (t) =, .. =Xn (t) =0.
Thus, we obtain nontrivial solutions (2) only for k such that
the determinant ( 4) vanishes. We arrive at . an equation of order
n for determining k:
a 11 -k au ... a,n
a11 a 12 -k ... a 2 n
=0. (5)
This equation is called the characteristic equation of the system (1 ),

and its roots are the roots of the characteristic equation.
Let us consider a few cases.
I. The roots of the characteristic equation are real and distinct.
Denote by k 11 k 1 , , kn the roots of the characteristic equation.
For each root ki write the system (3) and determine the
coefficients
It may be shown that one of them is arbitrary; it may be

considered equal to unity. Thus we obtain:
for the root k, the foliowing solution of the system (l)
x<l ~ a('.Jek,t
x(ll = a(l lek,t
1 . 1 , : z n = a<lek,t,
, , x(l n '
Systems of Linear Dif}erential Equations 571
for the root k 2 the solution of the system (l)
for the root kn the solution of the system (1)

x~n) = a~n)eknl' x~n) = a~n)eknl' ' x~n) = a~n)eknl.
where C 1 , C 2 , , Cn are arbitrary constants, is likewise a

solution of the system of differential equations (1 ). This is the
general solution of system (1). It may readily be shown that one
can find values of the constants such that the solution will
satisfy the given initial conditions.
Example t. Find the general solution of the system of equations
dx 1 dx 2
dt= 2X 1 + 2X 2 , dt=x 1 +3x2
Solution. Form the characteristic equation
2-k 2 I
1 I 3-k =O
or k2 -5k+4=0. Find its roots:
kl=l, k.=4.
Seek the solution. of the system in the form
x\tJ = a\tJ et' x~tJ = a~tJ i
and
x\ 2 ) = a\J e41 , x~J = a~> e'1
Form the system (3) for the root k 1 =I and determine ap> and a~l)r
(2-l)ap>+2a~ 1 l=o,}
la~l) + (3-1) a~ l =01
or
a! 1 >+2a~ 1 >=o,
a\ 1 J+2al1 >=o,
572 Differen-tial Equa,t{ons
1
whence a<l)=
2
_ _!_a<l).
2 I Putting a(!>=
I
I. we get a(l>=-
2
. Thus, we obtain
2
the solution of the system:
1
<J>=--et
X2 2 .
Now form the system (3) for the root k 2 =4 and determine a~l and a~>:
- 2a\> + 2a~> = 0,
a\ 21-2a~> =0,
whence a\ 2 >=a~> and a:>=l, a~l=J. We obtain the second solution of
the system:
The general solution of the sy5tem will be [see (6)]

x1 = C 1e1 + C2e41 ,
x, = - 21 c et + c e41
1 2
II. The roots of the characteristic equation are distinct, but

include complex roots. Among the roots of the characteristic equa-
tion let there be two complex conjugate roots:
k 1 =ex+ i~, k 2 =ex- i~.
To these roots will correspond the solutions
X<,.t> = ""
,../<.1) e<+i~>t (.J= 1, 2 ' ... ,n'
) (7)
X <J.> = ""
,..~~> e<-i~>t (.J=' I 2 , ... ,n.
) '(8)
The coefficients ex? 1 and exj" 1 are determined fro~ the system of
equations (3). :.
Just as in Sec. 21, it may be shown tpat the real and imagi-
nary parts of the complex solution are also solutions. We thus
obtain two particular solutions:
~}'' = e~ (~}'>cos ~x+ ~} sin ~x), }
21
(9)
x}" 1 = et (A.}' 1 sin ~x + A.} 21 cos ~x),
where A.J 11 , A.} 21 , W', X}" 1 are real numbers determined in terms
of ex? 1 and ex~>. . .
Appropriate combinations of functions (9) will enter into the
general solution of the system.
Systems of Linear DiOerential Equations 573
Example 2. Find the general solution of the system

dx 1
dt==-1JC, +x~,
dx.
([[=- 2x,- 5x
2
Solution. Form the characteristic equation
1
-'--7-k
-2
I
-5-k =O
I
or- k" + 12k + 37 = 0 and find its roots:
k,=-6+i. k.=-6-l.
Substituting k 1 = - 6 + i into the system (3), we find
a\''=1, a~''=I+i.
We write the solution (7):
x\') = le<-+ill, x~'l =(I+ i) e<-+ilt. (7')
Putting k 2 = - 6-i into system (3), we find
.a~> = I, a~> = 1-l.
We gtf a second system of solutions (8):
x~>=e<--ht, xi'' =(1-i) e<-&-i)t. (8')
Rewrite the solution (7'):
x~ll =e-at (cost+ i sin I),
xi' 1
= (1 + i) e- 81 (cost + i sin t)
or
x~'l = e-1 cost+ ie-1 sin t,
x~ l = e- 81 (cost -sin t) + ie- "1 (cost+ sin t).
1
Rewrite the solution (8'):

x~"'==:e- 81 cos t-te- 1 sin t,
A"1= e- 81 (cos 1-sln t) -le- 1 (cost+ sin 1).
For systems of particular solutions we can take the real parts and the imagi-
nary parts separately:
x~t) =e- 81 cos t, x~l) =e-1 (cos 1-sln t), }
(9')
x~l = e- sin t,. xi"' =e-
1 1
(cos I +sin 1),
The general solution of the system is

x1 = C,e_,, cost+ C,.e- 1sin t,
Xz = c,e -I (cos I -sin I)+ c.e -I (cost+ sin 1).
By a similar method it is possible to find the solution of a system

of linear differential equations of higher order with constant coef-
ficients.
For instance, in mechanics and electric-circuit theory a study
is made of the solution of a system of second-order differential
equations:
(IO)
Again we seek the solution in the form

x = a.ekt, y= ~ekt.
Putting these expressions into system (10) and .cancelling out ekt
we get a system of equations for determining a:, ~ and k:
(a 11 -k )a:+a!~~=0,}
1
(II)
a21 a+(a 11 - k )~=0.
Nonzero a and ~ are determined only when the determinant of

the system is equal to zero:
(I2)
This is the characteristic equation of system (1 0); it is a fourth-

order equation in k. Let k,, k,, k,, and k, be its roots (we assume
that the roots are distinct). For each root k; of system (II) we
find the values of a and ~-The general solution, like '(6), will
have the form
+ + +
x = C1 a:(l'ek,t C2a!"'ek,t C 3a:( 11 eA,t C4. a('ekt
y= C,~(l'ek,t + C.W"'lt + C .Wa'ekt + C,W''eRt.
If there are complex roots, then to each pair of complex roots

in the general solution there will correspond expressions of the
form (9).
Example 3. Find Jhe general solution of the following system of
differential equations
d 2x
dt~=x-4y,
d"y
dta = - x+y.
Systems of Linear Differential-Equations 575
Soluti.on. Write the characteristic equation (12) and find its roots:
l-k 2
l -I
k.=i,
We shall seek the solution in the form
x<J=a<Jeit, y(I) =~(I) eit,
x<J =a<J e-it, y<l = ~<1 e-it,
x<J=a<l eV>t, y(l) =~(I) eVo t,
x<''=a<J e-Vat, y<'l = ~('le -VII,
From system (11) we find a<ll and ~<h:
a<l=i, jt(l) _ _!_

t' - 2.
a<l=I, lt(2)-
t' -2.
_!_
I
a< 1l= I, ~(1)_=-2
I
a<''= I, ~(''=-2.
We wri[e out 'the complex solutions:
xU I =e-i1=cost+ i sin t, y<J = ; (cost+ i sin t),
x<J =e-il=cos t- i sin t, y<J = ; (cos t - i sin t).
The real and imaginary parts separately form the solutiom
x<l=cos t, -(l) = _!_cos t

Y 2
- 1
xUl =sin t, y<l = sin t.
2
We can now write the general solution:
x=C 1 cos.t+C 2 sint+C.ev>t +C,e-Y>t,

1 I C . 1 Vat 1 -VIi
U= C.2cost+2 1 smt-C 1 2e -C 4 2e .
Note. In this section we did not consider the case of multiple

roots of the characteristic equation. This question is dealt with in
detail in "Lectures on the Theory of Ordinary Differential Equations"
by I. G. Petrovsky.
SEC. 31. ON LYAPUNOV'S THEORY OF STABILITY
Since the solutions of most differential equations and systems

of equations are not expressible in terms of elementary functions
or quadratures, use is made (in these cases when solving concrete
differential equations) of approximate methods of integration. The
elements of these methods were given in Sec. 3; in addition, some
of these methods will be considered in Sees. 32 through 34 and in
Chapter XVI.
The drawback of these methods lies in the fact that they yield
only one particular solution; to obtain other particular solutions,
one has to carry out all the calculations again. Knowing one par-
ticular solution does not permit us to draw conclusions about the
character of the other solutions.
In many problems of mechanics and engineering it is sometimes
important to know not the specific values of a solution for some
concrete value of the argument, but the type of behaviour for
changes in the argument and, in particular, for a boundless increase
of the argument. For example, it is sometimes important to know
whether the solutions that satisfy the given initial conditions are
periodic, whether they approach some known function asymptoti
cally, etc. These are the questions with which the qualitative theory
of differential equations deals.
One of the basic problems of the qualitative theory is that of
the stability of the solution or of the stability of motion; this
problem was investigated in detail by the noted Russian mathe-
matician A. M. Lyapunov (1857-1918).
Let there be given a system of differential equations:
::=ff (t,x,y),}
dy
1
(1)
(.[[ = 2 (t, X, y). .
Let x = x (t) and y = y (!) be the solutions of this system that

satisfy the initial conditions
(1 ')
Further, Jet x=x(t) and y=y(t) be the solutions of equation

(1) that satisfy the initial conditions
(1")
On Lyapunov's Theory of Stability 577
Definition. The solutions x = x (t) and y = y (t) that satisfy the

equations (I) and the initial conditions (1') are called Lyapunov's
stable as t--+ oo if for every arbitrarily small e > 0 there is a{) >0
such that for all values t > 0 the following inequalities will be
fulfilled:
y
jx(t)-x(t)/<e, }
(2)
IY <n-y (t) I< e.
if the initial data satisfy the in- y=t
equalities
ixo-xo / < 6, } (3) X
IYo-Yo/<{), Fig. 274.
Let us figure out the meaning
of this definition. From inequali- ,
ties (2) and (3) it follows that for small variations in the initial
conditions, the corresponding solutions differ but little for all positive
values of t. If the system of differential equations is a system that
describes some motion, then in the case of stability of solutions,
the nature of the motions changes but slightly for small changes
in the initial data.
Let us analyse an example of a first-order equation.
Let there be given a differential equation:
dy
ar=-u+L (a)
The general solution of this equation is the function
y=Ce- 1 + I. (b)
Find- a particular solution that satisfies the initial condition
Yt=o= I. (c)
It is obvious that this solution y= 1 results when C=O (Ftg. 274). Then find
the particular solution that satisfies the initial condition
Yt"'o=Yo
Find the value of C from equation (b):
Yo=C + 1,
whence
C=uo-1.
Putting this value of C into equation (b), we get
y=(Yo-l)e-t+l.
The solution y = 1 is obviously stable.
19- 3388
578 DifferenUal Equations
Indeed,
+
y-y= !(i0 -l)e- 1 1]-1 = (Yo-1) e- 1 .... 0
when t-+ co.
Hence, inequality (3) will be fulfilled for an arbitrary e if the following
inequality is fulfilled:
Let us further consider the system of equations
~; =cx+gy, }
dy (4)
'ill= ax+ by,
assuming that the coefficients a, b, c, g are constant and g=I=O.
Let us find out what conditions the coefficients must satisfy
so that the solution x = 0, y = 0 of system (4} should be stable.
Differentiating the first equation and eliminating y, we get a
second-order eq'uation:
d'x dx
dt =c dt + g dtdy =c dtdx +g(ax+ by)=c iH
dx
+ agx+ b (dx
iif-cx
)
or
d~ ' ~ '
--(b+c)
dt 2 -
'--(ag-
dt bc)x= 0 (5)
Its auxiliary equation is of the form

f..'-(b +c)A.-(ag-bc)=O. (6)
Let us denote the, roots of t,he auxiliary equation by /..1 and A.,.
The following cases are poss'ible.
1. The roots of the auxiliary equation are real, negative and
distinct:
Then
x=C I e"'+C 2e'1'
y=[C 1 (A 1 -C)eKI+C,(A.,-c)etJ!,
The solution that satisfies the initial conditions

I
X t=o = Xo, Y I t=o = Yo
will be
On Lyapunov's Theory of Stability 57!}
From the latter formulas it follows that for any 8 > 0 it is

possible to choose X 0 and Yo so small that for all t >0 we will have
jx(t)j<8, jy(t)j<8 since e" 1 <1 and e' 1 <1.
Hence, in this case the solution x = 0, y = 0 is stable.
2. Let f..,= 0, t... < 0. Then
x = C, + C 2e't,
y =: {C 2 (A.-c) e'
1
-cC,J,
and the solution, as in the preceding case, proves stable.
3. Let t.., =A..< 0. Then
x= (C, + C,t) e\1,
1
y= J._e'
g
!C, (1..,-c) + C2 (I+ J..,t-ct)l.
Since
tekt ~ 0 and e'1--+0 when t--+ oo,
it follows that for sufficiently small C, and C 2 {that is, for suffi-
ciently small X 0 and Yo) we will have Ix (l) I< 8 and IY (t) I< e
for any t > 0. The solution is stable.
4. Let A.,= A2 = 0. Then
x=C, +C2 t.
1
g [-cC I +C2 -cC 2 tJ "
y=-
We see that for an arbitrarily small C 2 =fo 0 both x and y approach
infinity (as t---. oo), which means that the solution in this case
is unstable.
5. Let at least one of the roots A., and A., be positive; for
instance, f..,> 0.
From formula {7) it follows that no matter how small x 0 and
Yo if
ex.+ gy 0 - X 0 A2 =fo 0,
that is, if C, =I= 0, then Ix (t) 1- oo as t ~ oo.
Hence, in this case too the solution is unstable.
6. The roots of the auxiliary equation are complex with nega-
tive real part:
19*
580 DiUeren!ial Equations
In this case,
X= Ceot sin (~t + ~), }
y= ~ ce~ 1 ((a-c) sin (~t + 6) +~cos (~t + ~)). .(8)
It is obvious that for any 8 > 0 it is possible to choose X 0 and Yo
such that we will have 1C I< 8 and 1a-c I+ I~ I < s and, conse-
1 dl
quently,
/x{t)/<8 and ly(t)1<8.
The solution is stable.
7. The roots of the auxiliary equation are pure imaginaries:
AI=~~, A2 = -~t.
In this case,
x=C sin (~t+6),
y= g C [~cos (~t +6)-c sin (~t. +6)1
_!_
which means that x (t) and y (t) are periodic functions of t. As

in the preceding case, we verify the solution and find it stable.
8. The roots of the auxiliary equation are complex with
positive real part (a.> 0).
From formulas (8) it follows that here for arbitrarily small X 0
and Yo (that is, for arbitrarily small C+O) and for increasing t
the quantities I x (I) I and I y (t) I can take on arbitrarily large
values, since e.::!.oo as t--oo. The solution is unstable.
To give a general criterion of the stability of solution of the
system (I), we do as follows.
We write the roots of the auxiliary equation in the torm of com-
plex numbers:
AI = A; +
iA.;,
AI= A:+ iA.;
(in the case of real roots, A.;= 0 and A.;= 0).

Let us take the plane of a complex variable A.* A.** and display
the roots of the auxiliary equation by points in this plane. Then.
on the basis of the eight cases that have been considered, the
condition of stability of solution of the system {4) may be for-
mulated as follows. ,
If not a single one of the roots A.l' A2 of the auxiliary equation
(6) lies to the right of the axis of imaginaries, and at least one
root is nonzero, then the solutwn is stable; if at least one root
Euler's Method of Solution of Differential Equations 581
lies to the right of the axis of imaginaries, or both roots are equal
to zero, then the solution is unstable.
Let us now consider a more general system of equations:
:~=cx+gy+P(x, y), l (4')

~~=ax+by+Q(x, y). J
But for exceptional cases, the solution of this system is not expres-
sible in terms of elementary functions and quadratures.
To establish whether the solutions of this system are stable or
unstable, the system is compared with the solutions of a linear
system. Suppose that for x ~o and y ~o. the functions P (x, y)
and Q (x, y). also approach zero and approach it faster than Q,
where Q = V + X
1
y 1 ; in other words,
lim p (x, y) = 0; lim Q (x, y) = 0.
IHO Q Q-+0 Q
Then it may be proved that, save for the exceptional case, the
solution of the system (4') will be stable when the solution of
the system
:;=cx+gy,}
dy (4)
dt=ax+by,
is stable, and unstable when the solution of the system ( 4) is
unstable. The exception is that case when both roots of the auxil-
iary equation lie on the axis of imaginaries; in this case, the
question of the stability or instability of solution of the system
(4') is considerably more involved.
Lyapunov *) investigated the question of the stability of solutions
of systems of equations for rather general assumptions concerning
the form of these equations.
SEC. 32. EULER'S METHOD OF APPROXIMATE SOLUTION

OF FIRST-ORDER DIFFERENTIAL EQUATIONS
We shall consider two methods of numerical solution of a first-

order difierential equation. In this section. we consider Euler's
method.
") A. M. Lyapunov, The General Problem of Stability of Motion, ONTI,

1935 (Russian edition).
582 Difjerential Equations
Find (approximately) the solution of the equation

~~ = f (x, y) (I)
on the interval [x 0 , b] that satisfies the initial condition at x = x 0
Y=Yo Divide the interval [x 0 , b] by the points X 0 , xl' x., .. . , xn= b
into n equal parts (here Xo < x, < x. < ... < xn). Denote x, -Xo =
=x.-x, = ... = b-xn_ 1 = Llx=h; hence,
h=b-xo.
n
Let Y= cp (x) be some approximate solution of equation (I) and
Yo= CJl (xo), Y, = CJl (.tt), Yn = CJl (xn).
Denote
LlYo=Y,-Yo Lly,=y.-y,, LlYn-,=Yn-Yn-t
At each of the points X 0 , x" ... , xn in equation (1) we. replace
the derivative with the ratio of finite differences:
6.Y =
!lx f (x, y), (2)
Lly= f(x, y)'Llr.. (2~)
When x = X0 we have
~~0 = f (xo !J 0 ), LlYo = f (xo, Yo) Llx
or
Y,-Yo-:-f(xo, Yo)h.
In this equation, X0 , y 0 , h are known; thus we find
Y,=Yo+f(xo, Yo)h.
When x = x,, equation (2') takes the form
Lly, =f(x1 , Y1 )h
or
Y.-y,=f(x" y,)h,
Y. = y, + f (xl' yJ h.
Here, x" y,, h are known and y2 is determined.
Similarly, we find
Y.=Y.+f(x., Y.)h,
Euler's Method of Solution of Differential Equations 583
We have thus found the approximate values of the solution at

the points x., xi' ... , xn. Connecting, in a coordinate plane, the
points (X 0 , Y0 ), (xi' yJ ... , (xn, Yn) by y
straight-line segments, we get a broken
line- an approximate integral curve (Fig.
275). This broken line is called Euler's
broken line.
Note. We denote by y = cph (x) an approx-
imate solution of equation (1), which (Xq,!Jo}
corresponds to Euler's broken line when
/).x=h. It may be proved*) that if there
exists a unique solution y = cp* (x) of equa- o x0 x, x2 x3 x
tion (1) that satisfies the initial condi- Fig. 275.
tions and is defined on the interval [x.,
b), then limjcph(x)-cp*(x)I=O for any x of the interval [x 0 , b).
h .... o
Example. Find the approximate value (for x= 1) of the solution of the
equation
y'=y+x
that satisfies the initial condition y 0 =1 for x 0 =0.
Solution. Divide the interval [0, I) into 10 parts by the points x 0 =0,
0.1, 0.2, ... , 1.0. Hence, h =0.1. We seek the values Yu y 2 , . . . , Yn by for
mula (2'):
Mk=(Yk+xk) h
or Yk+t =IJk+(yk+xk) h.
We thus get y 1 = 1 +(I +0)0.1 = 1+0.1 = 1.1,
Yt= 1.1 +(1.1 +0.1)0.1 = 1.22,
Tabulating the results as we solve, we get:
... I Ilk
I Yh+X
I ll.yk=(Yh+x,,) h
XQ =0 1.000 1.000 0.100

x, =0.1 1.100 1.200 0.120
x. =0.2 1.220 1.420 0.142
x 8 =0.3 1.362 1.620 0.162
x. =0.4 1.524 1.924 0.1924
x5 =0.5 1.7164 2.2164 0.2216
x. =0.6 1.9380 2.5380 0.2538
x, =0. 7 2.1918 2.8918 0.2812
x. =0.8 2.4730 3.2730 0.3273
x. =0.9 2.8003 3.7003 0.3700
x 10 = 1.0 3.1703
*) For the proof see, for example, I. G. Petrovsky's "Lectures on the

Theory of Ordinary Differential Equations".
584 DitJerential Equations
We have found the approximate value yl.~= 1 =3.1703. The exact solution
of this equation that satisfies the indicated initial conditions is
y=2ex-x-l.
Hence,
Ylx=t = 2 (e -I)= 3.4365.
The absolute error is 0.2662; the relative error is~:!~~= 0.077:::::::8%.
SEC. 33. A DIFFERENCE METHOD FOR APPROXIMATE

SOLUTION OF DIFFERENTIAL EQUATIONS BASED
ON TAYLOR'S FORMULA. ADAMS METHOD
We once again seek the solution of the equation

y' =f (x, y) (1)
on the interval [x 0 , b], which solution satisfies the initial condi-
tion y =Yo when x = X 0 We introduce notation that will be needed
later on. The approximate values of the solution at the points
will be
Yo Y.. Yz , Yn
The first differences, or differences of. the first order, are
tlyo=Y,-Yo tly,=yz-Yl' llYn-=Yn-Yn-J'
The second differences, or differences of the second order, are
2
ll Y0 = tly,- tlyo = Yz-2Y, + Yo
tlzy, = tlyz- tly, = Y3- 2y2 + y,,
2
ll Yn-z = tlYn-1- tlYn-2 = Yn-2Yn- + Yn-z
Differences of the second differences are called differences of the
third order, and so forth. We denote by y~. y;, ... , y~ the approx-
imate values of the derivatives, and by y:.
y';, ... , y~ the
approximate values of the second derivatives, etc. Similarly we
determine the first differences of the derivatives:
tly~=y;-y~. tly',=y:-y;, ... , tly~_,=y~-Y~-"
the second differences of the derivatives:
tl 2 y~=tly;-tly~, tl 2 y:=tly:-tly;, ... ,tl 2 Y~-z=tly~-1-tly~-2>
and so on.
A Difjerencc Method for Approximate Solution of Difjerential Equations 585
Write Taylor's formula for solving an equation in the neigh-

bourhood of the point x = x. [Ch. IV, Sec. 6, formula (6)]:
x-x 0 ' (x-x 0) 2 " (x-x 0 )m (m)
Y=Yo+-1-Yo+ 12 Yo+ + 1. 2... mY +Rm. (2)
In this formula Yo is known, and the values of i/o, ... of the

derivatives are found from equation (1) as follows. Putting the
y:,
initial values x 0 and Yo into the right side of equation (1), we
find y~:
y: = f (xo, Yo)
Differentiating the terms of (I) with respect to x, we get
Yt"-~+~
-ax ayY I (3)
Substituting into the right side the values X0 , Yo y we find
Y:=(?J+~y')
OX Oy x=x 0 y=y0 , y' =y
..
0
Once more differentiating (3) with respect to x and substituting

the values x 0 , Yo y~. y:, we find y~". Continuing in this fashion.*.
we can find the values of the derivatives of any order for x = X 0 )
All terms are known, except the remainder Rm on the right side
of (2). Thus, neglecting the remainder, we can obtain an approx-
imation of the solution for any value of x; their accuracy will
depend upon the quantity Ix-x 0 I and the number of terms in
the expansion.
In the method given below-, we determine by formula (2) only
the first few values of y when 1 x-x. 1 is small. We determine
the values y, and y. for x, = X 0 + h and for x. = x. + 2h, taking
four terms of the expansion (y. is known from the initial data):
h ' h 2 ., h" "'
y,=Yo+TYo+~Y-I-3JYo, (4)
211 ' (2h) 2 (2/z) 3
Y. =Yo+ T Yo +n Yo+3TYo
" '"
(4')
We thus consi ~ter known three values**) of the function: Y.

y" y.. On the basis of these values and using equation (! ),
) From now on we shall assume that the function f (x, !I) is differentiable
with respect to x and y as many times as is required by the reason in~.
u) If we were to seek the solution with greater accuracy, we would have
to compute more than the first three values of !I This is dealt with in
detail by Ya. S. Bezikovich in "Approximate Calculations" (Gostekhizdat,
1949) (Russ:an edition).
' 586 Differential Equations
we find
Y~=f(x., Y.). y;=f(x,, y.), y~=f(x Y.).
Knowing y~. y;, y;, it is possible to determine f).y~. f).y;, !).y~.
Tabulate the results of the computations:
X
I y
I y'
I IJ.y'
I IJ.'y'
x.
I Yo
I Yo I I
I I I &y. I
x 1 =x 0 +h
I Yt
I y,
I I &u~
I I I tJ.y,
I
x 2 =x0 +2h
I Yz I IJz I I
... I ... I ... I I ... . ..
X.t- 2 =x 0 +(k-2) h
I Yk-z
I Yk-z I . I
I I I I i!J.Yk -
X.t- 1 =x 0 +(k-1) h
I Yk-t I '
!lk-t I I &y;_.
'
I I I: i!J.Yk-t
I
xk=x0 +kh
I Y.t
I Yk' I r
Now suppose that we know the values of the solution
Yo y,, Y Yk
Prom these values we can compute [using equation (1)] the values
of the derivatives I I I I
Y y,, Y2o . , Yk
and, hence,
and
1 1
A! A! A2 '
u y., u y,, ... , u Yk-.
Let us determine the value of Yk+t from Taylor's formula (see
Ch. V, Sec. 6), setting a=xk, X=Xk+J =xk+h:
Yk+=Yk+TYk+J.2Yk
h , h
2
+123yk
h
+ ... +.mtYk
3
h
"'
+ R m m (m)
A Difference Method for Approximate Solution of Difjerential Equations '037
In our case we shall confine ourselves to four terms of the

expansion:
(5)
The unknowns in this formula are y~ and y~'', which we shall

try to determine by using the known first-order and second-order
differences.
First, represent y~_ 1 in Taylor's formula, putting a= xk,
x-a=-h:
2
' ' (-h) "-! (-h) "'
Yk-1 =Y,,+-1- Yk -IT Yk (6)
and y~_ 2 , putting a=xk, x-a=-2h:
' - ' + (-
Yk--Yk
2h) .. -1 (- 2h)
-J-yk --J.-2-yk
2
,,,
(7)
From (6) we find
2
h " h 111
Yk- yk-1 = u.yk-1 = T Yk -n Y.z
' A '
(8)
Subtracting the terms of (7) from those of (6), we get
, , A , h 3h 2 , r
Yk-1-yk_.=u.Yk-=Tyk-TYk (9)
From (8) and- (9) we ~btain
Ay~_ 1 - !1y~- = !1 y~- = hyt
or
", I A. ' (10)
Yk = h' u Yk-1
Putting the expression Yt into (8), we get
(11)
Thus, y: and y~'' have been found. Putting expressions (10)

and (11) into the expansion (5), we obtain
Yk+1 = Yk + ~y~ + ~ !1y~_ 1 + ~ !1"y~_.. (12)

This is the so-called Adams formula with four terms. Formula
(12) enables one to compute Yk+t when Yk Yk-,. Yk- are known.
Thus, knowing y 0 , Y1 andy, we can find Y. and, further, y., Yw ..
Note 1. We state without proof that if there exists a unique
solution of equation (1) on the interval [X 0 , b), which solution
satisfies the initial conditions, then the error of the approximate
588 Ditlerential Equations
values determined from formula (12) do not exceed, in absolute

value, Mh 4 , where M is a constant dependent on the length of
the interval and the form of the function f (x, y) and independent
of the magnitude of h.
Note 2. If we want to obtain greater accuracy in our computa-
tions, we must take more terms than in expansion (5), and for-
mula (12) will change accordingly. For instance, if in place of
formula (5) we take a formula containing five terms to the right,
that is, if we complete it with a term of order h 4 , then in place
of formula (12) we, in similar fashion, get the formula
Yk+a=Yk+~Y~+ ~ ~Y~-~+~~~y~-~+~~y~-a
Here, yk+a is determined by means of the. values yk, Yk-t Yk-z
and Yk-a Thus, in order to begin computation using this formula
we must know the first four values of the solution: Yo, y" y 1 , y .
When calculating these values from formulas of type (4), one
should take five terms of the expansion.
Example I. Approximate the solution of the equation
y'=y+x
that satisfies the initial condition
Yo= 1 when X 0 =0.
Determine the values of the solution lor x = 0.1, 0.2, 0.3, 0.4.
Solution. First we find !11 and Yz using formulas (4) and (4'). From the
uquation and the initial data we get
Y~ = (y -f-x)x=o=Yo-f-0= 1 +O= 1.
Differentiating the given equation, we have
y"=y' +I.
Hence,
u:=(y' + l)x=o= 1+ 1=2.
Differentiating once again, we get
y'''=yh.
Hence,
Substituting into (4) the values y0 , y~, u; and h=0.1, we get
-1 +~.I +(O.I)' 2-f-(0. 1)" 2-1 1103

Yt- I 12 123 - . .
Similarly, for h=0.2 we have
-1
Yz- + 0.2
I '
1+(0.2)' 2+(0.2)1 2-12426
I 2 ' I 23 - . .
A Difference Method for Approximate Solution of Differential Equations 589
Knowing y 0 , y 1 , y., we find (on the basis of the equation)

y~=Ya+O=I,
y: =y 1 +O.l = 1.1103+0.1 = 1.2103,
y~=-Y2 +0.2= 1.2426+0.2= 1.4426.
M~=0.2103,
,:\y; =0.2323,
L\ 2y~ = 0.0220.
Tabulating the values obtained, we have
I
I
u
I
y'
I O.tl
I 0.'y
X 0 =0 I Yo= 1.0000 y~= I

I I I
I I I .M~ =0.2103 I
I L\ y~ =0.0220
x 1 =0.1 2
y, = 1.11031 y; = 1.2103
I I
M:=0.2323
I r I I
I Yz= 1.2426
I
2
X 1 =0.2 y: = 1.4426
I I L\ Y; = 0.0228
I I I
lly: =0.2551
I
x 3 =0.3 y,= 1.39771 y; = 1.6977
I I I
=0.4 Y4 = 1.58121
X4
I I I
From formula (12) we find y 3 :
2
y 3 =1. 426+--1.44
0.1 26 +0.1
2 0. 2323 +5-(0.1) 2
-- -0.020=1.3977.
1 12
We then find the values of y~. L\y:. L\ 2y;. Again using formula (12) we find y 4 ;
0
y,= 1.3977+
0 1
j !.6977+ ~/ 0.2551 + 5
12
.0.10.0228= 1.5812.
The exact expression of the solution of the given equation is
y=2ex-x-I.
Hence, Yx=o.4 =2e 0 4 -0.4-1 = 1.5836. The absolute error is 0.0024; the rela-
tive error, ~:~~:=0.0015~0.15%. (In Euler's method, the absolute error
of y. is 0.06, the relative error, 0.038~ 3.8/ .)
Example 2. Approximate the solution of the equation
y' =u'+x'
that satisfies the initial condition y 0 =0 for x 0 =0. Determine the values of
the solution for x=0.1, 0.2, 0.3, 0.4.
Solution. We find
u:=O'+O'=O,
u:=o=(2yy' +2x)x=o=0,
u:~o = (2y'" + 2yy" + 2)x=o = 2.
By formulas (4) and (4') we have
(0.1)' (0. 2) 1
y 1 = 31 2 = 0.0003, Ya=-aJ 2=0.0026.
From the equation we find
y~=O, y; =0.0100, y~=0.0400.
Using these data, we construct the first rows of the table, and then deter-
mine the values of y1 and y4 from formula (12).
X
I v
I II
I
bog'
I &v
I y~=O I
x 0 =0 Yo=O
I I
M~=O.OIOO
I I I I
I u: =0.0100 ~y~=0.0200
X 1 =0.1 y,= 0.0003
I I I
I I I Ay; =0.0300 I
I ~ u; = 0.0201
x2 =0.2 /j 1 =0.0026 u:=0.0400
I I I
M~=0.0501
I I I I
x1 =0.3 1J1 =0.0089!1:=0.0901
I I I I
x4 =0.4
I
IJ.=0.0204
I I I
An Approximate Method for Integrating Differential Equations 391
Thus,
01 0
y3 =0.0026+ j 0.0400+ ;} 0.0300+ ; .0.10.0200=0.0089,
2
0 5 0
u.=O.OOB9+ / 0.0901 + ;} 0.0501 + 0.10.0201 =0.0204.
12
We note that the first four correct decimals in y, are y,=0.0213. (This
may be obtained by other, more accurate, methods with error evaluation.)
SEC. 34. AN APPROXIMATE METHOD FOR INTEGRATING

SYSTEMS OF FIRST-ORDER DIFFERENTIAL EQUATIONS
The methods of approximate integration of differential equa-

tions considered in Sees. 32 and 33 are also applicable for solving
systems of first-order differential equations. Here, we consider the
difference method for solving systems of equations. Our reasoning
will deal with systems of two equations in two unknown functions.
It is required to find the solutions of a system of equations
dy
dx=f,(x,y,z), (l)
dz
dx =
f (x, y, z) (2)
that satisfy the initial conditions y=y 0 , Z=Z 0 when X=X 0
We determine the values of the function y and z for values of
the argument. X0 , x., x , xk, xk+l' .. , xn. Once more, let
xk+~.....:xk~l1x=h(k--.:O, l, 2, ... , n....:..l). (3)
We denote the approximate values of the function as
yo, y,, :., Yk Yk+t> Yn
and
Z0 , Zt> , Zk,.zk+1' . , Zn.
Write the recurrence formulas of type (12), Sec. 33:
h h, 5
= Yk +1 Yk + 2 AYk-t + 12 hA
21
Yk+t Yk-~t (4)
h I h I 5 I I
zk+,=zk+Tzk+ 2 Azk_,+ 12 hA Zk- (5)
To begin computations using these formulas we must know y" y 2 ;
z , z. in addition to Yo and Z 0 ; we find these values from formu-
l~s of type (4) and (4 Sec. 32: 1
),
h I h' " h3 Ill
Y, =Yo +TYo +2Yo+ 3JYo

2h I (2h) 2 " (2h)' Ill
Ya=Yo+y-Yo+21Yo+3f""Yo
h ' h2 " It' ,,

Z 1 =Z 0 +Tz+2zo+ 31 zo ,
2h (2/t) 2 (2h) 3
z,=z 0 + T zo+-2- zo+ar Zo
I " Iff
I " Ill I " I II

To apply these formu1as one has to know y0 , y., Yo , Z0 , z0 , Z0 ,
which we shall now determine. From (1) and (2) we find
y~ =f) (x., Yo z.),
z~ =f. (X 0 , Yo Z 0 ).
Differentiating (1) and (2) and substituting the values of X0 , y0 , Z0 ,
y~ and z~. we find
(y") (of, +of, ~+of, 1)
Yo = x=xo = iJx ily Y i)z Z X=Xo '
=(ilfz+ilfzyl+ofszl)
z=(z")
0
ilx ily ilz x=x."X=Xo
Differentiating once again, we find y;" and Knowingy" y 1 , z;".
zl' z,, we find from the given equations (1) and (2),
I I I I I ' / I I I I ! I
Y~> Ya. z"

lly., lly" ll Yo. llzo, llz" ll Zo,
Zz.
after which we can fill in the first five rows of the table:
. I ,I I( I IJ.!I
I
IJ.21(
II 2 I z'
I IJ.z'
I ar
I I
z.
x.
I Yo
I Yo
I I II
Zo
I I I
I I'
I I I
llY 0
I II I I
/lZo
I
I I I
x, y, y, !l'y' z, z, /l 2 Z0
I I I I II I I )
I I
Az,
I I I fly,
I II I I I
I I
I
I
ll y, z. z. ,1 1z;
XI
I Yz I Y. I I I I I
I
M~
I I I I I I I /lZ 2
I
I I
z, z,
x.
I
Ya
I Ya I I II I I I
An Approximate Method for Integrating Differential Equations 593
From formulas (4) and (5) we find Ys and Z3 , and from equations
(1) and (2) we find y; and x:.
Computing Ay~, tJ.y;, Az~, tJ.z;,
we find Y4 and y 5 , etc., by applying formulas (4) and (5) once
again.
Example. Approximate the solutions of the system
y' =z, z' =y
with _initial conditions Yt=O and z0 = I for x=O. Compute the values of the
solutions for x=O, 0.1, u.2, 0.3, 0.4.
Solution. From the given equations, we find
y~=Zx=o=i,
z~ = Yx=o = 0.
Differentiating the given equations, we find
u: = (y")x=o = (z')x=o = 0,
z: = (z")x=~ = (y') x=o = I,
Yo'" =y
( "')x=o=(z") x=o=l,
Z0,, =(Z ''') X=o=(y ") X=o=O.
Using formulas of type (4} and (5), we find
=0+~-1 +(O.l) 0+(0.l)s 1=0 1002

2
y, l 12 31 .
0.2 (0.2)' (0.2) 3
Y2 = 0 +--I+IT0+-- -1=0.2016,
1 31
z,= I+ j O+ (~.'r I+
0 1
(O;r 0= 1.0050,
2 3
O+ (Oz~) (03~)
02
Z2 = I+ j i+ 0= 1.0200.
Using the given equations, we find
u: = 1.0050, y~= 1.0200,

z; =0.1002, Z~=0.20J6,
M/~=0.0050, ilZ~ = 0.1002,
L\y: = 0.0150, M; =0.1014,

Ll 2y~=0.0100, z:
Ll 2 = 0.0012.
Filling in the first five rows of the table, we have
u
I I II
I l'>.y"
I 6.'11
X
I y~=l
x =0 Yo=O
0
I I I
M~=0.0050
I I I I
x,=O.I I y, =0.1002 u: = 1.0050 ~~y~=O.OIOO
I I I
I I I M: =0.0150 I
X2 =0.21 y 2 = 0.2016
I u; = 1.0200
I I
~ 2y; =0.0109
I I I M~=0.0259 I
Xs=0.31 Ys = 0.3049
I
I . y~ = 1.0459
I I
x,=0.4
I
y,=0.4117
.I I I
I X 2
I 2'
I liz'
I ll 2 z~
I Z~=O
x =0 z0 = I
0
I I I
I I I. M~=0.10021
x,=O.l I z,= 1.0050
I
z; =0.1002
I I~~z;=0.0012
I I I ~z; =0.1014 I
X2 =0.21 21 == 1.0200
I
Z~=0.2016
Illlz;=0.1033 I ~~z; =0.0019
I I I
x,=0.31 z,= 1.0459
I z;=0.3049 I I
x,=0.41 z,= 1.0817
I I I
Exercises on Chapter XI II 595
From formulas (4) and (5) we find

0.1 0.1 5 .
u.= o. 20 I6+--l.0200+
1 21 .o.ol50+ 12 .o.r.o.OI00=0.3043,
z3 = 1.0200-1- / .J.2016+ 2 0.1014+~0.10.0012= 1.0459
0 0 1
and similarly
0
y,=0.3049+ / 1.0459 +
01
2
0.0259+ ~ 0.10.0109=0.4117,
0 5
z,= 1.0459+ 0 / 0.3049+ ;} 0.1033+ .0.10.0019= 1.0817.
12
It is obvious that the exact solutions of the system of equations (the so
lutions satisfying the initial conditions) will be
1 I
y=2(eX-e-X), Z=2(eX+e-X).
And so, solutions correct to the fourth decimal place are

1 1
y,= (e 0'-e- 0')=0.4107, z,= (e0'+e- 0')= 1.0811.
2 2
Note. Since equations of higher order and systems of equations of higher
order in many cases reduce to a system of first-order equations, the method
given above is applicable to the solution of such problems.
Exercises on Chapter XIII

Show that the indicated functions, which depend on arbitrary constants, satis-
fy the corresponding differential equations:
Functions Differential Equations
dy 1
I. y=slnx-1 +Ce-lnx. dx +u cosx = 2 sin 2x.
dy' 2 dy dy
2. y=Cx+C-C 2 ( dx) - dx -x dx + Y = 0
u (:~r +2x:~ -u=o.
xy [ 1-(:~)] =(x 2 -y 2 -a 2):~.
d 3y 3 d 2y
dx+x dx=O.
d 2y dy 2
+
dx-2k dx k u=ex .
2
d y dy
(1-x) d~ 2 -x dx -ay=O.
8.
c, ca
u=-x+ d'y-1-~dy=O
dx 2 x dx
Integrate the differential equations with variables separable

9. ydx-xdy=O. Ans.y=Cx. 10.(l+u)udu+(1-v)udv=0. Ans. lnuu+
+u-v=C. 11. (l+y)dx-(l-x)dy=O. Ans. (l+y)(l-x)=C
dx t+x x
12. (12-xt')dT+x'+tx=O. Ans. tx-+ln y=C. 13. (y-a)dx+x'dy=O.
~ t-a ~
Ans. (y-a)=Ce" .14. zdt-(tf-a 2 )dz=0. Ans. za=Ct+a 15. dy=
I +x' y+C .r- .r-
=
1
+u'. Ans. x= i-Cy' 16. (1 +s') dt- I' t ds=O. Ans. 2 r t-
-arctans=C. 17. dQ+Qtan8d8=0. Ans. Q=Ccos8. 18. sin8cos<pd8-
-cos 8 sin<p d<p=O. Ans. cos <p=C cos 8. 19. sec atan <p d8+sec 2 <p tan 8d<p=0.
Ans. tan 8 tan <p=C. 20. sec 2 8 tan <p d<p+sec 2 <p tan 8 d8=0. Ans. sln 2 8+
2
+sin <p=C. 21. (1+x')dy-Vi-y dx=0. 2
Ans. arcslny-arctanx=C.
22. VI-xdy- Vl-y 2 dx=0. Ans. yVl-x 2 -x Vl-y 2 =C. 23. 3eX tan yx
xdx+(l-e")sec 2 ydy=0. Ans. tany=C(l-tr) 3 24. (x-y 2x)dx+
+(y-x 2y)dy=0. Ans. x'+y 2 =xy+C.
Problems in Forming Differential Equations

25. Prove that a curve having the slope of the tangent to any point pro-
portional to the abscissa of the point of tangency is a parabola. Ans.
y=ax'+C.
26. Find a curve passing through the point (0,- 2) such that the slope of
the tangent at any point of it is equal to the ordinate of this point increased
by three units. Ans. y=e"-3.
21. Find a curve passing through the point (1, 1) so that the slope of the
tangent to the curve at any point is proportional to the square of the ordi
nate of this point. Ans. k(x-l)y-y+l=O.
28. Find a curve for which the slope of the tangent at any point is n
times the slope of a straight line connecting this point with the origin. Ans.
y=Cx".
29. Through the point (2, I) draw a curve for which the tangent at any
point coincides with the direction of the radius vector drawn from the origin
to the same point. An.s. u={x.
30. In polar coordinates, find the equation of a curve at each point of
which the tangent of the angle between the radius vector and the tangent
line is equal to the reciprocal of the radius vector with sign reversed. Ans.
r(S+C)= I.
31. In polar coordinates, find the equation of a curve at each point of
which the tangent of the angle formed by the radius vector and the tangent
line is equal to the square of the radius vector. Ans. r 2 =2 (S+C).
32. Prove that a curve with the property that all its normals pass through
a constant point is a circle.
33. Find a curve such that at each point of it the length of the subtan
gent is equal to the doubled abscissa. Ans. y = C Vx.
34. Find a curve for which the radius vector is equal to the length of the
tangent between the point of tangency and the x-axis.
Solution. By hypothesis, Y, Vi+y''= vx+y 2 , whence
dy= dx. In
y y X
tegrating, we get two families of curves: y=Cx and y=.E..
X
Exercises on Chapter XIII 597
35. By Newton's law, the rate of cooling of some body in a1r 1s propor.
tiona! to the difference between the temperature of the body and the tempe-
rature of the air. If the temperature of the air is 20 C and the body cools
lor 20 minutes from 100 to 60 C, how long will it take lor its temperature
to drop to 30 C?
Solution. The differential equation of the problem is : ; =k(T-20). In-
tegrating we find: T-20=Cek 1; T=iOO when t=O; T=60 when t=20; there-
fore, C=80; 40=Ceok, ek= ( { );o; consequently, T=20+80 ( {)fa. As

suming T = 30, we find t = 60 min.
36. During what time T will the water flow out of an opening 0.5 em at
the bottom of a conic funnel 10 em high with the vertex angle d = 60?
Solution. In two ways we calculate the volume of water that will flow out
+
during the time between the instants t and t tJ,t. Given a constant rate t',
during 1 sec a cylinder of water with base 0.5 cm 2 and altitude h flows out,
and during time M the outflow is the volume of water du equal to
-du=-0.5udt=-0.3 V2ghdt.)
On the other hand, due to the outflow, the height of the water receives a
negative "increment" dh, and the differential of the volume of water outflow
is
Thus,
whence
Setting h= 0, we get the time of outflow T = 12.5 sec.

37. The retarding action of friction on a disk rotating in a liquid is p.ro
portional to the angular velocity of rotation w. Find the dependence of this
angular velocity on the time if it is known that the disk begins rotating at
100 revolutions per minute and, after the elapseof one minute, rotates at 60
revolutions per minute. Ans. w = 100 (f)' rpm.
38. Suppose that in a vertical column of air their pressure at each level
is due to the pressure of the above-lying layers. Find the dependence of the
pressure on the height if it is known that at sea level this pressure is I kg
per cm 2 , while at 500 m above sea level, 0.92 kg per cm 2
Hint. Take advantage of the BoyleMariotte law, by virtue of which the
density of the gas is proportional to the pressure. The differential equation of
the problem is dp = - kp dh, whence p = e- 0 00017 ". Ans. p =e-o.ooonh.
*) The rate of outflow u of water from an opening a distance h from the

free surface is given by the formula u=0.6 V2gh, where g is the acceleration
of gravity.
598 DiUerenlial Equations
Integrate the following homogeneous differential equations:

39. (y-x)dx+(y+x)dy=O. Ans. y'+2xy-x 2 =C. 40. (x+y)dx+xdy=O.
Ans. x'+2xy=C. 41. (x+y) dx+(y-x) dy=O. Ans. In {x 2 +y 1 ) ' / , _
-arctanlL=C. 42. xdy-ydx=Yx'+y'dx. Ans. 1+2Cy-Cx=o.
X
43.(8y+ lOx) dx+(5y+7x~y=O. Ans. (x+y) 2 (2x+y) 3 =C. 44. (2 Vst-s)X
,/ s c
xdt+tds=O.Ans.teV 1 =C ors=tln'T 45. (t-s)dt+tds=O. Ans.
V- -
-s C
te1=Cors=t!n-. 46. xy'dy=(x 3 +y 3 )dx. Ans. y=x 3lnCx.
t
47. xcoslL(ydx+xdy)=ysin Y (xdy-ydx). Ans. xycos !I =C.
X X X
Integrate the differential eq4ations that lead to homogeneous equations:
48. (3y-7x+7)dx-(3x-7y-3)dy=0. Ans. (x+y-l) 5 (x-y-1) 2 =C.
49. (x+2u+ 1) dx-(2x+4y+3) dy=O. Ans. In (4x+8y+5)+8y-4x=C.
. . ' 4 +5
50. (x+Zu+ 1) dx-(2x-3) dy=O. Ans. In (2x-3)- ~_ =C.
2 3
51. Determine the curve whose subnormal is the arithmetical mean between
the abscissa and the ordinate. Ans. (x-y) 1 (x+2y)=C.
52. Determine the curve in which the ratio of the segment cut off by a
tangent on th~ y-axis to the radius vector is equal to a constant.
dy
Solution. By hypothesis,
y-xdx
.!".
( x
m. whence -C
)m - (-C )m =-.
2y
y X2 +Y 2 . X X
53. Determine the curve in which the ratio of the segment cut off by the
normal on the x-ax is to the radius vector is equal to a constant.
dy
x+ydx
Solution. It is given that .r . ;,m, whence x'+u'=m 2 (x-C) 1
r x+u'
54. Determine the curve in wbich the .segment cu! off by a tangent on the
y-axis is equal to a sec 8, where 8 is the angle between the radius vector and
the x-axis.
Solution. Since tan 8= fL and by hypothesis
X
dtj
y-x d.x =a sec 8,
we ohtain
dy ~
y-x dx=a x '
whence
-~ [~+b
Y- 2 e -e
-(:+b)] .
55. Determine the curve for which the segment cut off on the y-axis by
a normal drawn to wme point of the curve is equal to the distance of this
point from the origin.
Solution. The segment cut off by the normal on they-axis is y + ..;; there-
y
fore, by hypothesis, we have
whence
x 2 =C(2y+C).
56. Find the shape of a mirror such that all rays emerging from a single
point 0 would be reflected parallel to the given direction. .
Solution. For the x-axis we take the given direction, and 0 as the origin.
Let OM be the incident ray, MP the reflected ray, and MQ the normal io
the desired curve.
a=P; OM=OQ, NM=y,
NQ=NO+OQ= -x+ x+y 2 =ycot P=Y~~

whence
ydy=(-x+ J!x+y') dx;
integrating, we have
Integrate the following linear differential equations:

2
57. y'- +Y =(x+ 1)'. Ans. 2y=(x+ l)'+C (x+ 1) 2 58. y' -alL =x+ 1
X 1 X X
X J 1 2
Ans. y=Cxa+-- - - . 59. (x-x ) y' +(2x -1)y-ax'=O. Ans. y=
1-a a
.~ ds ds
=ax+Cx r 1-x. 60. dtcost+ssint=l. Ans. s=slnt+Ccost. 61. dt+
+scost=; sln2t. Ans. s=slnt-l+Ce-slnt. 62. y'-: y=e"xn. Ans.

n a I
y=xn(eX+C). 63. y'+-Y=-;;. Ans. xny=ax+C. 64. y'+Y=-x Ans.
X X e
1
X ! y-1 =0. Ans. y=x ( 1 + cex ) .
exy=x+C. 65. y' +l-2x
Integrate the Bernoulli equations:

66. y'+xy=xy. Ans. y 2 (x 2 +l+Ce"')=l. 67. (1-x 2 )y'-xy-axy2 =0.
Ans. (C Jfl-x -a)y=l. 68. 3y y'-ay-x-1=0.
2 2
Ans. ay=ceax_
.!..y. ] ..!.. y'
-a(x+l)-1. 69. y'(x'u'+xy)=l. Ans. x [ (2-y 2)e 2 +C =e 2
70. (ylnx-2)ydx=xdy. Ans.y(Cx+lnx+i)=l. 71. y-y'cosx=ycosxx
_tan x +sec x
X(l~slnx). A ns. Y- sinx+C
Integrate the following exact differential equations
72. (x 2 + y) dx + (x-2y) dy =0. Ans. [ x~ + yx-y 2 = C J. 73. (y-3x 2 ) dx-

-(4y-x)dy=O. Ans. 2y 2 -xy+x'=C. 74. (Y'-x)y'=y. Ans. y4 =4xy+C.
75

[-y--_!_Jx dx+ [_!_y __
2
(x-y) 2 (x-y)"
2
x_- ] dy=O

Ans.
x x-y
In _!_.EL._ =C.
76. 2 (3xy 2 +2x 3 ) dx+ 3 (2x 2 y +y 2 ) dy=O. Ans. x 4 + 3x 2y 2 + y' =C.
2
77. xdx+( x+y) dy 0. Ans. In (x+y)- +x =C. 78. (-;+- 3Y-,2 ) dx= 2-Y-~_Y
(X+ y) 2 X y X X X '
2
A n8. X 2 +Y 2 = C'X.
79 x dy-yzdx
(
x-y )Z = 0 . A ns. -xy= C. 80. X dx+ydy=
x-y
y dx-x dy A .. z+ 2 2 t x C
= x 2 +yz . ns. X y - arc any= .
81. Determine the curve that has the property that the product of the
square of the distance of any point of it from the origin into the segment
rut off on the x-axis by the normal at this point is equal to the cube of the
abscissa of this point.
Ans. y 2 (2x 2 +y 2)=C.
82. Find the envelope of the following families of lines: a) y = Cx + C 2 .
X X y
Ans. x 2 +4y=O. b) y=c-f-C 2 Ans. 27x 2 =4y1 c) C-C, =2. Ans.
27y=x'. d) C2x+Cy-1=0. Ans. y 2 +4x=O. e) (x-C) 1 +(y-C) 2 =C 2 Ans.
x=O; y=O. f) (x-C) 2 +y'=4C. Ans. y 2 =4x+4. g) (x-C)'+(y-C) 2 =4.
Ans. (x-y) 2 =8. h) Cx 2 +C 2 y=l. Ans. x 4 +4y=0.
83. A straight line is in motion so that the sum of the segments it cuts
off on the axes is a constant a. Form tr.e equation of the envelope of all
positions of the straight line. Ans. x'l + y'l =a'' (parabola).
84. Find the envelope of a family of straight lines on which the coordi
2 2 2
nate axes cut off a segment of constant length a. Ans. x 1 + y' =a 1 .

85. Find the envelope of a family of circles whose diameters are the
doubled ordina~es of the parabola y2 =2px. Ans. y 2 =2p ( x+ ~).
86. Find the envelope of a family of circles whose centres lie on the pa-
rabola y 2 =2px; all the circles of the family pass through the vertex of this
parabola. Ans. The cissoid x'+y 2 (x+2p)=0.
87. Find the envelope of a family of circles whose diameters are chords
2
of the ellipse b 2x 2 +a 2y 2 =a 2 b2 perpendicular to the x-axis. Ans. az~ bz+
y2
+b'=l.
88. Find the evolute of the ellipse b 2x 2 +a 2y 2 =a 2b2 as the envelope of its
2 2 2
3 3
normals. Ans. (ax) +<by)' =(a 2 -b 2 )
Integrate the following equations (Lagrange equations):
2
, ,a A C 2 2C- p' .
90 .y=xy
,a + ,, A
89.y=xy+y. ns.x= p,- 3 p; y=~ y. ns.
3
y=(Yx+I+C). Singular solution: y=O. 91. y=x(l+y')+(y') 2 Ans.
x=Ce-P-2p+2; y=C (P+ I) c-P-p 2 +2. 92. y=yy'"+2xy'. Ans.
4Cx=4C 2 -y 2 93. Find a curve with constant normal. Ans. (x-C)"+Y"=a 2

Singular solution: y = a.
Integrate the given Clai raut equations:
94. y=xy' +!/ -y''. Ans. y=Cx+C-C 2 Singular solution: 4y=(x+ 1)1 .
95. y=xy'+ Y1-y' 1 Ans. y=Cx+ Jfl-C 2 Singular solution: y 2 -x 2 =
= l. 96. y=xy'+y'. Ans. y=Cx+C. 97. y=xy'+~,. Ans. y=Cx+~.
Singular solution: q2 =4x. 98.1f=xy'-y~- An~. y=Cx-~ 2 Singular
solution: y'=-~x 2
99. The area of a triangle formed by the tangent to the sought-for curve
and the coordinate axes is a constant. Find the curve. Ans. The equilateral
hyperbola 4xy = a 2 Also, any straight line of the family y = Cx a Y C.
100. Find a curve such that the segment of its tan~ent between the coor-
dinate axes is of constant length a. Ans. y = Cx Ya . Singular solution:
l+C 2
xt. + y'la = at.
101. Find a curve the tangents to which form, on the axes, segments
2
whose sum is 2a. Ans. y=Cx- aCC. Singular solution: (y-x-2a) 2 =8ax.
1
102. Find curves for which the product of the distance of any tangent line
to two given points is constant. Ans. Ellipses and hyperbolas. (Orthogonal
and isogonal trajectories.)
103. Find the orthogonal trajectories of the family of curves y=axn.
Ans. x 2 +ny 2 =C. ~ .
104. Find the orthogonal trajectories of the family of parabolas y 2 = 2p (x-a)
X
(a is the parameter of the family). Ans. y = Ce P .

105. Find the orthogonal trajectories of the family of curves x 2 --y 2 =a
c .
la is the parameter). Ans. y=x.
106. Find the orthogonal trajectories of the family of circles x'+y 2 =2ax.
Ans. Circles: y=C(x 2 +y 2).
107. Find the orthogonal trajectories of equal parabolas tangent at the
vertex of the given straight line. Ans. If 2p is the parameter of the parabo-
las. and the given straight line is on the y-axis, then the equation of the
,;- s
trajectory will be y+C=~ Jl %x
2
x'
108. Find the orthogonal trajectories of the cissoids y 2 = a-x. Ans.
2
(x"+y 2 ) 2 =C (y 2 +2x 2 ).
109. Find the orthogonal trajectories of the lemn iscates (x 2 +y 2 2
) = (x 2 -y 2) a=.
Ans. (x 2 +y 2 ) 2 =Cxy. _
I 10. Find the isogonal trajectories of the family of curves: x 2 = 2a (y -x Y 3 ),
where a is a variable parameter if the constant angle formed by the trajec-
tories and the lines of the family is ro=60.
Solution. We find the differential equation of the family y' ='2:.!!.-
X
Y3
and for y' substitute the expression q=t~~t;annroro. lf w=60, then
602 Di{jerential Equations
q=
1 +
V
y' 3
3 y'
and we get the differential equation
y'- V3
1 + y'
V
3
"y
~ - V3.
X
The complete integral y 2 =C (x-y 3) yields the desired family of trajec-
tories.
ttl. Find the isogonal trajectories of the family of parabolas y 2 =4Cx
_!__ arc tan II-X
when ro=45. Ans. y 2 -xy+2x 2 =CeY1 x Y?.
112. Find the isogonal trajectories of the family of straight lines y= Cx for
2 Ys arc tan }!_
x 2 +y 2 =e x.
the case ro=30, 45. Ans. The logarithmic spirals
{ 2 arc tan L
x 2 +y 2 =e x .
ll3. Y=C 1ex+c,e-x. Eliminate C1 and C,. Ans. y"-y=O.
114. Write the differential equation of all circles lying in one plane. Ans.
{1 +y'") -3y'y"1 =0.
115. \vrite the differential equation of all second-order central curves
whose principal axes coincide with the x- and y-axes. Ans. x (yy" + y' 1 )-y' y = 0.
ll6. Given the differential equation y'"-2y"-y'+2y=0 and its general
solution y= C,ex +C 2 e-x +C,ezx.
It is required to: 1) verify that the given family of curves is indeed the
general solution; 2) find a particular solution if for x = 0 we have y =I,
y'=O, y"=-1. Ans. y=i (9ex+e-x-4ezx).
ll7. Given the differential equation y= 2~, and its general solution
I
y= a2 {x+C,)- 1
+C1
It is required to: 1) verify that the given family of curves is indeed the
"!eneral solution; 2) find the integral curve passing through the point (1, 2)
if the tangent at this point forms with the positive x-direction an angle of
45. Ans. y~f J!Xi+~
Integrate some of the simpler types of differential equations of the second
order that lead to first-order equations.
118. xy"' =2. Ans. y=x'ln x+C,x 2 +C.x+C 1 ; pick out a particular solu-
tion that satisfies the following initial conditions: x=l; y=1; y'=1;y"=3.
(n) m ml xm+n n-1 " 2
119. y =X. Ans. y=(m+n)l +C,x + ... +Cn_ 1x+Cn. 120. y =ay.
Ans. ax=ln (ay+ Vay+C,)+C1 or y=C1e0 x+c,e-ax. 121. y"=!!.,. Ans.
. y
(C 1x + C2 ) 1 = C,y-a.
In Nos. 122-125 pick out a particular solution that satisfies the following
initial conditions: x=O, y=- 1; y' =0. 122. xy" -y' =xex. Ans. y=
=ex(x-1)+C 1x 2 +C. Particular solution: y=ex(x-1). 123. yy.-(y')'+
+(y')1 =0. Ans. y+C,Iny=x+C,. Particular solution: Y=-1.
124. y"+y'!anx=sln2x. Ans. y=C 2 +C,sinx-x-~sln2x. Particular
solution: y=2slnx-slnxcosx-x-1. 125. (y") 2 +(y') 2 =a 2 An.S. y=C 2 -
~acos(x+C1). Particular solutions! y=a-1-acosx; y=acosx-(a+I).
Exerdses on Chapter XIII 603
(Hint. Parametric form: if=acost, y'=aslnt). 126. y"= 2 ~,. Ans. y=

= {<x+C 1)'/+C1 127. y"' =y"". Ans. y=(C 1 -x) [In (C1 -x)-l]-I-C 2x -1-
+ C,. 128. y'y'"-3y""=O. Ans. x=C 1 y -I-C 1 y+C.
2
Integrate the following linear differential equations with constant coeffi-

cients:
129. 1!"=9y. Ans. y=C 1 e"'+C 2 e-x. 130. y'+y=O. Ans. y=Acosx+
+ 8 sinx. 131. y"-y' =0. Ans. y=C 1 +C 2ex. 132. y' 12y=7y'. Ans. +
y=C 1e"x+C 2 ex. 133. y"-4y'+4y=0. Ans. y=(C1 -I-C 2 x)e 2x. 134. y"-l-
+2y'-l-10y=0. Ans. y=ex(Acos3x+Bsin3x). 135. y"+3y'-2y=0. Ans.
-s+ Vi7
- --x - - Vi.? x
y=C1e 2 +C1e 136. 4y"-12y'+9y=O. Ans. y=
Ans.
I
y=e- 2 xx[Acos(~ 3
-
x)+
+ B sin ( ~3 x) J.
138. Two identical loads are suspended from the end of a spring. Find
the motion imparted to one load if the other breaks loose. Ans. x=
=a cos ( yf t ). where a is the increase in length of the spring under
the action of one load at rest.
139. A material point of mass m is attracted by each of two centres with
a force proportional to the distance. The factor of proportionality is k. The
distance between the centres is 2c. At the initial instant the point lies on
the line connecting the centres at a distance a from the middle. The initial
velocity is zero. Find the law of motion of the point. Ans. x=a cos ( y-~ t) .
140. y 1V-5if+4y=O. Ans. y=C 1ex+c.e-x+c,ezx+c,e-u. 141. y"-
-2y'-y'-l-2y=0. Ans. y=C 1e2x-I-C 2 ex+c.e-x. 142. y"-3ay"+3a 2 y'-
-ay=0. Ans. y=(C 1 -I-C 2x+C 3x 2 )eax. 143. yv -4y'"=O. Ans. y=C 1 +
+C2 x+C,x'+C,ezx+c.e-zx. 144. y 1v +2y"+9y=0. Ans. Y=(C 1 cos V2x+
+C 2 sln V2X)e-x+(C 3 cos V2x+C,sin Jf2x)ex. 145. y 1v-8y"+l6y=0.
1
Ans. y = C 1eox + C2 e-x + C3xex + C,xe-ox. 146. y v + y = 0. Ans. y =
X X
ev \.
fc 1 cos x
V2+ c1 sm
. Vx ) - v2(c cos V"2
x c
+ , sin V
x )
=
2
+ e
2 .
1
147. y v -a'y=O. Find the general solution and pick out a particular
solution that satisfies the initial conditions forx 0 =0, y=l. y'=O, y"=-a,
+
y"' =0. Ans. General solution: y = C,eax C2e-ax + C 3 cos ax+ C, sin ax. Par-
ticular solution: y0 =cosax.
Integrate the following nonhomogeneous linear differential equations (find
the general solution):
148. y" -7y'+l2y=x. Ans. y= cx ,e cx
2e++~ 12x+7 1 9"
4. s -as=t+l.
2
. t+ 1
Ans. s=C 1ea 1 +C,.e-at - 2 - . 150. y"-1-y' -2y=8sin 2x. Ans. y=C,eX+
a
604 DiUerential Equations
+ C2e-2X - ~ (6 sin 2x + 2 cos 2x). 151. y" -y = 5x + 2. Ans. y =C 1r + C,e-x-

et
-5x-2. 152. s"-2as'+a's=e 1 (a:/= 1). Ans. s=C 1ea1 +C,tea 1 +(a-l)'"
153. !I +6y' +5y=e2X. Ans. y=C 1e-x +C,e-sx +; e". 154. y" +9y=6e'x.
1
1
Ans. y=C 1 cos 3x+C 2 sin3x+ eax. 155. !l-3y' =2-6x. Ans. y=C1 +
3
+C,e'x+x'. 156. y"-2y'+3y=e-xcosx. Ans. y=ex(Acos'V2x+
- e-x
+Bsin Y2x)+ 4!(5cosx-4sinx). 157. if+4y=2sin2x. Ans. y=
X
=Asin2x+Bcos2x_-
2 cos2x. 158. y"'-4y"+5y'-2y=2x+3. Ans. y=
=(C 1 +C 2 x) ex +C,eZX-x-4. 159. y 1v -a4 y=5a4e= sin ax. Ans. y=(C 1 -
- sin ax) eax + C,e-ax +Ca cos ax+ C4 sin ax. 160. y 1V + 2a 2y" + a 4 y=8 cos ax.
x'
Ans. y=(C 1 +C 2x) cos ax+(C,+C4 x) slnax- 2 cosax.
a
161. Find the integral curve of the equation y + k"y =0 that passes through
the point M(x 0 , y 0 ) and is tangent at the point of the curve y=ax.
a
An~. y =Yo cos k (x-x 0 ) + k sink (x-x 0 ).
162. Find a solution of the equation y" +2hy' +n 2y=0 that satisfies the
conditions y=a, y' =C when x=O. Ans. For h < n y=e-hx (a cos Y n2 -h 2 x+
+~+ah sinJfn'-h'x);for h=n y=e-hx[(C+ah)x+aJ; for h>n
nz-h
C+a(!t+Jf~ -(h-Vh-n)x C+a(h-Yiii=fil) -(h+Yh-n)x
y- ~ - e
- 2 Jfi! 2 -n 2 2 Yh'-n'
163. Find solutions of the equation y" +n'y=h slnpx (P :/= n) that satisfy
the conditions: y=a, y'=C for x=O. Ans. y=acosnx+
2 2 2
+ C (nn (-p )-hp
n -p2 sinnx+-
')
h
n -p 2- -2

smpx.
164. A load weighing 4 kg is suspended from a spring and increases its
length by I em. Find the law of motion of this load if we assume that the
upper end of the spring performs harmonic oscillations under the law
y =sin Y 100 gt, where y is measured vertically.
Solution. Denoting by x the vertical coordinate of the load reckoned from
the position of rest, we have
4 d 2x
g dt'=-k(x-y-l),
where l Is the length of the spring in the free state and k =400, as Is evi-
dent from the initial conditions. Whence~::+ IOOgx=IOOg sin JflOOgt + 100 lg.
We must seek the particular integral of this equation In the form
t(C 1 cos V100gt+C 2 sin VlOOgt)+g,
since the first term on the right enters into the solution of the homogeneous
equation.
165. In Prob Iem 139, the initial velocity is u0 and the direction is per-
pendicular to the straight line connecting the centres. Find the trajectorie.c;.
Solution. If for the origin we take the mid-point between the centres, the
differential equations of motion will-be m:;:=k(C-x)-k(C+x>= -2kx,
d"y
md .=-2ky. The initial data for 1=0 are
1
dx dy
x=a; d/=0; y=O; dt=t1 0
Integrating, we find
x =a cos ( y ~ t) , y = t1 0 y~ sin ( y~ t) .
x y'2k
Whence + - - = 1 (ellipse).
1
a mu~
166. A horizontal tube is in rotation about a vertical axis with constan~
angular velocity ro. A sphere inside the tube slides along it without friction.
Find the law of motion of the sphere if at the initial instant it lies on the
axis of rotation and has velocity u0 (along the tube).
d'r
Hint. The differential equation of motion ic; dl 2 =ro 2r. The initial data
dr
are: r = 0, dt = t1 0 for t =0. Integrating, we find
r=~~ [ew1 +e-w1].
-Applying the method of variation of parameters, Integrate the following

differential equations:
167 . y - 7.
y + 6y =smx.
. A ns. y=- C ,e"+C eex + 5slnx+7cosx . 168 Y +Y=
2
74
=sec x'. Ans. I/=C 1 cosx+C 2 sinx+x sinx+cosx In cosx. 169. y" +Y =
1
y . . r -
Ans. y=C, cos x+C 2 slnx- r cos 2x
-
cos 2x cos 2x
integrate the following systems of equations:
170. !~=y+l. :r=x+l. Pick out the particular solutions that
satisfy the initial conditions x = - 2_, y =0 for t =0. Ans. !I= C1 cos!+ C2 sin t,
X= (C.+ c.) cos t+(C.-C,) sin t- Particular solution: x* =cost -sin I. y*=cos t.
171. !~=x-2y, ~~=x-y. Pick out the particular solutions that
satisfy the initial conditions: x=l. y=l for 1=0. Ans_ y=C 1 cost+C 2 sint,
x=(C,+C 2 ) cos/ +(C2 -C 1 ) sin/. Particular solution: x* =cos t-sint,
y* =COSt.
dx dy , Ans. x=C 1e-'+C 2e-at,
172. f 4 dtdt + 3x =sin t. y =C,e- 1+ 3C 2 e- 11 +cost.
l U+ !f=COS t.
06 Differential Equations .
' d'y Ans. x=C 1 e1 +C.e- 1+C 3 cos t+C4 sln t,

dt 2 =X, y = C,e 1 +C 2e- 1 -C3 cos t-C 4 sin t.
173. 2
{ dx
(i!i=Y
174. J dx
d 2x dy 1
dt 2 + di + x = e ' Ans. x=C 1 +C.t +Cat- i t +e
8 1
,
l d2 q
dt+dt"=l. y=C4 -(C 1 +2C 8) t- ~ (C 2 -1) t-
- 31 C8 ts + 24I t -e.I
dy Ans. y=(C 1 +C 2x) e-zx,
dX=z-y,
175.
{ dz
dx=-y-3z.
z= (C 2 -C 1 -C2x) e-zx.
dy Ans. y=C 1evc +C 2e-zx,

z = -2 (C1ezx -C2e- zx).
176.
{ dx +z=O,
dz
dx+4y=0.
Ans. y=C1 +C2 x+2 slnx,

:~ +2y +z=slnx,
{
z= -2C1 -C2 (2x+ 1)-3 slnx-2 cosx.
177.
dz
--4y-2z=COSX.
dx
dx
dt=y+z.
r
~
dy
178. dt=x+z,
dz
I\ di=x+y.
dy I
dx= l-z-'
179. J
dz
l dX=y-x
y=x+- e-C,x.
c,c.
I
.
dy X
dx=yz Ans. -=--=C1,
180.
{ dz X
dx=Yi
3
y
zy- .t==C2
2
Integrate the following different types of equations:
A =__!___[ C,(x-C,J+ -c.cx-C,)l 182 xdy-y 2dx
181. yy"=y' +1. ns. y 2C, e e . . (x-y)' 0.
Exercises on Chapter X Ill 607
Ans . ..!JL=C. 183. y=xy''+ y''. Ans. y=(Vx+ I +C) 2 Singular solutions:
x-y
y=O; x+J=O. 184. y"+y=secx. Ans. y=C1 cosx+C 2 slnx+xsinx+
+cosxlncosx. 185. O+x 2 )y'-xy-a=O. Ans. y=ax+CJ"T"+?.
dt sin .lL
186. xcos!LdY=ycos!L-x. Am. xe =C. x187. y"-4y=e2Xsin2x.
X X X
e2X
Ans. 1J=C1e-2X+C 2e2X- (sin2x+2cos2x). 188. xy'+y-y 2 lnx=0.
20
Atts. (In x+ 1 +Cx) y= I. 189. (2x+2y-1) dx+(x+y-2) dy=O. Ans. 2x+
+y-3ln(r.+y+1)=C. 190. 3extanydx+(l-ex)sec 2ydy=0. Ans. tany=
= C (!-ex).
Investigate and determine whether the solution x=O, y=O is stable for
the following systems of differential equations:
dx
dT=2x-3y,
191. J Ans. Unstable.
l dq
dt=5x+6y.
dx
-=-4x-10y,
dl
192. { dy Ans. Stable.
dt=x-2y.
193.
Jl. ~~=12x+l~y. Ans. Unstable.
dy = - 8x-12y.
dt
194. Approximate the solution of the equation y' = y 2 +x that satisfies
the initial condition y= 1 when x=O. Find the values of the solution for x
equal to 0.1, 0.2, 0.3, 0.4, 0.5. Ans. Yx=o.s = 2.114.
195. Approximate the value of Yx=t. 4 of a solution of the equation
y'+_!_y=ex that satisfies the initial conditions y=l when x=l. Compare
X
the result obtained with the exact solution. .
196. Find the approximate values of xt=t. and Yt .. 1 . 4 of the solutions of a
a system of equations ~~=y-x, ~~=-x-3y that satisfy the initial con-
ditions x = 0, y = 1 when t = 1. Compare the values obtained with the exact
values.
CHAPTER XIV
MULTIPLE INTEGRALS
SEC. 1. DOUBLE INTEGRALS
In an xy-plane we consider a closed*) region D bounded by a

line L.
In this region D let there be given a continuous function
z=f(x, y).
Using arbitrary lines we divide the region D into n parts
!ls 11 !ls2 , lls 1 , , !lsn
(Fig. 276) which we shall call subregions. So as not to introduce
new symbols we will denote by !ls11 , !lsn both the subregions
and their areas. In each subregion !ls; (it is immaterial whether
in the interior or on the boundary) take a point Pi; we will then
have n points:
We denote by f (PJ, f (P 2 ), , f (P n) the values of the func-

tions at the chosen points and then form the sum of the products
f(P;)!J.si:
n
Vn = f (P 1 )!ls 1 + f (P 2 ) lls 2 + ... + f (P n)!lsn= ~f (Pi)!ls;. (1)
i=l
This is the integral sum of the function f (x, y) in the region D.

If f ~ 0 in D, then each term f (Pi) lls; may be represented
geometrically as the volume of a small cylinder with base lls;
and altitude f (P; ).
The sum Vn is the sum of the volumes of the indicated ele-
mentary cylinders, that is, the volume of a certain "step-like"
solid (Fig. 277).
Consider an arbitra1y sequence of integral sums formed by
means of the function f (x, y) for the given region D,
vn' v ' ... , v ' ...

l n2 nlc
(2)
") A region D is called closed if if is bounded by a closed line, and the

points lying on the boundary are considered as belonging io the region D.
Dott ble Integrals 609
for different ways of partitioning D into subregions lls;. We shall

assume that the maximum diameter of the subregions !ls 1 ap-
proaches zero as nk ~ oo, and the following y
proposition, which we give without proof,
holds true.
Theorem t. If a function f (x, y) is cant inu-
ous in a closed region D, then there is a
limit of the sequence (2) of integral sums (I) if
the maximum diameter of the subregions dsi
approaches zero as n ~ oo. This limit is the
same for any sequmce of type (2), that is,
il is independent either of the way the region
D is partitioned into subregions !ls 1 or of o x
the choice of the point Pi inside the subre- F tg.
276 .
gion !lsi.
This limit is called the double intfgral of the fu net ion f (x, y)
over the region D and is denoted by
~ ~ f (P) ds or ~ ~ f (x, y) dx dy,
D D
that is,
lim ~ f (Pi) !lsi= ~ ~ f (x, y) dx dy.
d1am &sl .... o i=l 0
This region D is called the domain (region) of integration.

If f (x, y);;;;;:.: 0, then the double integral of f (x, y) over D is
equal to the volume of the solid Q bounded by a surface z=f(x, y),
Fig. 277. Fig. 278.
the plane z =0, and a cylindrical surface whose generators are

parallel to the z-axis, while the directrix is the boundary of the
region D (Fig. 278).
20- 3388
610 Multiple Integrals
Now consider the following theorems about the double integral.

Theorem 2. The double integral of a sum of two functions
<p (x, y) +'II' (x, y) over the region D is equal to the sum of the
double integrals over D of each of the functions taken separately:
B[<p{x, y)+'l'(x, y)]ds=~~<p(x, y)ds+~~'l'(x, y)ds.
D D D
Theorem 3. A constant factor may be taken outside the double

integral sign:
if a-const, then
~ ~ a<p (x, y) ds =a ~ ~ <p (x, y) ds.
D D
The proof of both theorems is exactly the same as that of the
corresponding theorems for the definite integral (see. Sec. 3, Ch. XI).
Theorem 4. If a region D is divided into two regions D 1 and D 2
without common interior points, and the function f (x, y) is
continuous at all points of D, then
~~ f(x, y)dxdy= ~~ f(x, y)dxdy+ ~~ f(x, y)dxdy. (3)
D 0 1 D,
Proof. The integral sum over D may be given in the form

(Fig. 279)
(4)
where the first sum contains terms that correspond to the subre-
gions of DP the second, those corresponding to the subregions of
D,. Indeed, since the double integral does not depend on the
manner of partition, we divide the region D so that the common
boundary of the regions D 1 and D 2 is a boundary of the subre-
gions ilsi. Passing to the limit in (4) as ilsi -0, we get (3).
This theorem is obviously true for any number of terms.
SEC. 2. CALCULATING DOUBLE INTEGRALS
Let a region D lying in the xy-plane be such that any straight

line parallel to one of the coordinate axes (for example, the
y axis) and passing through an interior*) point of the region,
cuts the boundary of the region at two points N 1 and N, (Fig. 280).
*) An interior point of a region is one .that does not lie on the boun
dary of the region.
Calculating Double Integrals 611
In this case we assume that the region D is bounded by the

lines: y=cp,(x), y=cp,(x), x=a, x=b and that
q>, (x) .;;;;; cp. (x), a, (x) and q>, (x) are continuous on the interval
[a, b]. We shall call such a region regular in they-direction. The
definition is similar for a region regular in the x-direction.
!I lj
0 X 0 X
Fig. 279. Ftg. 280.
A region that is regular in both x- and y-directions we shall

simply call a regular region. In Fig. 280 we have a regular
r.egion D.
Let the function f (x, y) be continuous in D.
Consider the expression
b op 2 (X)
ID = ~ ( ~ f (x, y) dy) dx
a QJ 1 (x)
which we shall call an iterated integral of f(x, y) over D. fn this

expression we first calculate the integral in the parentheses (the
integration is performed with respect to y) while x is considered
to be constant. The integration yields a continuous*) function of x:
'I'> (X)
<D (x) = ~ f (x, y) dy.

op, (x)
We integrate this function with respect to x from a to b;

b
lv= ~ <l>(x) dx.

a
This yields a certain constant.
*) We do not here prove that the !unction <1> (xl is. c'onfin'uous.
20*
Example l. To calculate the iterated integral
1 x
In=~
0
0
0
(x2 +y 2
) dy ) dx.
Solution. First calculate the inner integral (in brackets):
Integrating the function obtained from 0 to 1, we find
Determine the region D. Here, D is considered the region bounded by the

lines (Fig. 281)
It may happen that the region D is such that one of the func-
tions y = q:> 1 (x), y = cp 2 (x) cannot be represented by a single
y y
0 X
o a c
Fig. 281. Ftg. 282.
analytic expression over the entire range of x (from x =a to x =b).

For example, let a< c < b, and
cp 1 (x) = 'ljl (x) on the interval [a, c],
cp 1 (x) = x (x) on the interval lc, b].
where 'ljl (x) and x(x) are analytic functions (Fig. 282). Then the
iterated integral wi II be written as follows:

b lj)o (X)
~ [ ~ f(x, y)dy Jdx=

a 1Jl 1 (x)
c ql,(x) b ql 2 (x)
= ~ [ ~ f (x, y) 1y dx + ~
a <p 1 (x)
J c
[ ~ f (x, y) dy Jdx =
cp 1 (x)
c <p 2 (x) b cp,(x)
= ~
a
[ ~ f (x,
'il(x)
y) dy dx J +~ [
c X(x)
J
~ f (x, y) dy dx.
The first of these equations is written on the basis of a familiar

property of the definite integral, the second, due to the fact that
on the interva 1 [a, cj we have !p, (x) = ' (x), and on the interval
[c, b) we have !p, (x) _ x. (x).
We would also have a similar notation for the iterated integral
if the function !p 2 (x) were defined by different analytic expres-
sions on different subintervals of the interval [a, b).
Let us establish some properties of an iterated integral.
Property t. If a regular y-direction region D is divided into two
r_egions D, and D 2 by a straight line parallel to the y-axis or the
x-axis, then the iterated integral I 0 over D will be equal to the
sum of such integrals over D, and D 2 ; that is,
(I)
Proo(. a) Let the straight line x = c (a< c <b) divide the

region D into. two regular y-direction regions*) D, and D 2 Then
.
b cp 2 (x) b c b
/ 0 =~ ( ~ f(x, y) dy)dx=~<D(x)dx=S<D(x)dx+~<D(x) dx=
a . cp 1 (xl a a c
c cp 2 ( x) b <p,(x)
=~( ~ f(x, y)dy)dx+~( ~ f(x, y)dy)dx=lo,+fo,.

a cp,(x) c cp 1(xJ
*) The fact that a part of the boundary of the region 0 1 (and of 0 2) is a

portion of the vertical straight line does not stop this region from being reg-
ular in the y-direction: for a region to be regular, it is only necessary thaf
any vertical straight line passing through an interior point of the region
should have no more than two common points with the boundary (see foot-
note on page 610).
b) Let the straight line y = h divide the region D into

two regular y-direction regions D 1 and D. as shown in Fig. 283.
Denote by M 1 and M, the points of intersection of the straight
line y=h with the boundary L of D. Denote the abscissas of these
points by a. and b 1
y The region D 1 is bounded by con-
tinuous lines:
1) y = cp 1 (x);
2) the curve A 1 M 1 M 2 B, whose
equation we shall conditionally write
in the form
y = cp; (x),
having in view that cp; (x) = cp, (x)
when a~x~a. and when b 1 ~x~b
o a a, btbx and that
Fig. 283.
cp; (x)=h when a, ~x~b,;
3) by the straight lines x=a, x=b.

The region D. is bounded by the lines
y=cp;(x), y=cp.(x), where a 1 ~x~b 1
We write the identity by applying to the inner integral the

theorem for partitioning the interval of integration:
b ,tx)
, b
.
q>J(X) cp, (X)
= ~ [ ~ f (x, y) dy + J
~ f (x, y) dy dx =
a q> 1 ex) q>:(x)
b
.
ql I (X) b cp, (X)
= ~ ( ~ f (x, y) dy) d x + ~ ( ~ f (x, y) dy) dx.

a cp, (X) a q> :(x)
We break up the latter integral into three integrals and apply

to the outer integral the theorem for dividing the interval of
integration:
b Ql 2 (x) a1 Ql 2 (x)
~ ( ~ f (x, y) dy) dx = S( S f (x, y) dy) dx +

a Ql (X) a Ql (X)
I I
b, Ql2 (X) b Ql2 (X)
+~ ( ~ f (x, y) dy) dx +~ ( ~ f (x, y) dy) dx;

a, Ql: (X) b, Ql: (X)
since <p: (x) = <p 2 (x) on the interval (a, a,] and on [9" b], it
follows that the first and third integrals are identically zero.
Therefore,
b

Ql' (x) b1 Ql 2 (x)
In~( ~f(x,y)dy)dx+~( ~f(x, y)dy)dx.

a Ql 1 (x) a, Ql: (x)
Here, the first integral is an iterated integral over D" the second,
over D . Consequently,
ln= ID, + ID,
The proof will be similar for any position of the cutting straight
line M,M 2 If M,M. divides D into three or a larger number of
regions, we get a relation similar to (1), in the first part of
which we will have the appropriate
lj
number of terms.
Corollary. We can again divide l'!t.
each of the regions obtained (using
a straight line parallel to the y-axis '0:
or x-axis) into regular y-direction ....
regions, and we can apply to them
equation (2). Thus, D may be divided
by straight Jines parallel to the
coordinate axes into any number of
regular regions 0 a b X
D" 0 2 , D,, ... , Di, Fig. 284.
and the assertion that the iterated

tntegral over D is equal to the sum of iterated integrals over
subregions holds; that is (Fig. 284),
ID = I D, + I D, + I D, + ... + I Dt (2)
Pl'Qpt;rty 2 . (Evaluation of. an iterated integral). Let m and M
be the least and greatest values of tlie function f (x, y) in the
region D. Denote by S the area of D. Then we have (he relation

b IP2 (X)
mS =eo;; ~ ( ~ f (x, y) dy) dx =eo;; MS. (3)

a Q> 1 !X)
Proof. Evaluate the inner integral denoting it by <l>(x):

"' (xl q> 2 (x)
CD(x)= ~ f(x, y)dy=e;;;; ~ Mdy=M [!Jl 1 (x)-!p 1 (x)].
Q> 1 (X) q> 1 (X)
We then have
b Q>, (Xl b
1 0 =~( ~ {(x, y)dy)dxoe;;;;~Ml!Jl 2 (X)-!p 1 (x)]dx=MS,
a q> 1 (X) a
that is,
10 oe;;;;MS. (3')
Similarly
q>, (Xl IP IX)
<D (x) = ~ f (x, y)dy;;;;a. ~ mdx=mf!Jl.(t)-!p, (x)],

Q>, (X) (j)l (X)
b b
I0 = ~ <1> (x) dx ;;;a.~ m [<p, (x) -!p 1(x)j dx =mS,
a a
that is,
I0 ;;;;;. mS.. (3")
From the inequalities (3') and (3") follows the relation (3):
mS=e;;;; 10 oe;;;;MS.
In the next section we will determine the geometric meaning of
this theorem.
Property 3. (Mean-Value Theorem). An iterated integral 10 of
a continuous {un::tion f ( x, y) over a region D with areaS is equal
to the product of the area S by the value of the function at some
point P in the region D; that is,
b '1'2 (X)
~ ( ~ f(x, y)dy) dx=f(P)S. (4)

a q> 1 (Xl
Proof. From (3) we obtain

I
moe;;;; J0 oe;;;;M.
5
Calculating Double fntegrals (continued) 617
The number ~ 10 lies between the greatest and least values of

f (x, y) in D. Due to the continuity of the function f (x, y), at
some point P of D it takes on a value equal to the number ~ 10 ;
that is,
~I 0 =f(P),
whence
10 = f (P)S. (5)
SEC. 3. CALCULATING DOUBLE INTEGRALS

(CON Tl NU ED)
Theorem. The double integral of a continuous function f(x, y)

over a regular region D is equal to the iterated integral of this
junction over D; that is.*)
~ q>, (XI
~ ~ f (X, y) dx dy = ~ ( ~ f (x. y) dy) dx.

0 a q> 1 (~
Proof. Partition the region D with straight lines parallel to the

coordinate axes into n regular (rectangular) subregions:
t1s;, l1s 2 , , t1sn.
By Property 1 (formula (2)] of the preceding section we have
I 0 =I !1&, +I !1s, + ... + " !1Si
I !1!>y, = ~I ( l)
i=
Each of the terms on the right we transform by the mean-
value theorem for an iterated integral:
It1s1 = f (P J L1s 1
Then ( 1) takes the form
n
10 = f(P,) L'1s, + f (P 2 ) dS 2 + ... + f(P~) dsn = ~ f (P 1) L'1s 1, (2)
- l=t
where P 1 is some point of the subregion L'1s 1 On the right is the

integral sum of the function f(x, y) over the region D. From the
existence theorem of a double integral it follows that the lim it
of this sum, as tt--+ oo and as the greatest diameter of the sub-
regions L1s 1 approach zero, exists and is equal to the double
integral of f(x, y) over D. The value of the double integral I 0 on
*) Here, we again assume that the region D is regular in the ydirection
and bounded by the lines y=cpdx), Y=cp2 (x), x=a, x=b.
the right side of (2) does not depend on n. Thus, passing to the
limit in (2), we obtain
lv= lim ~f(P;)L1s;= lr f(x, y)dxdy
dlam <ls;-+o J J
or
Hf(x, y) dxdy= lv. (3)
D
Writing out in full the expression of the iterated integral I D

we finally get
b q>, (X)
~~f(x, y)dxdy=~[ ~ f(x, y)dy]dx. (4)
D a q> 1 (X)
Note t. For the case when f(x, y);;;;. 0, formula (4) has a pic-
torial geometric interpretation. Consider a solid bounded by the
surface z = f (x, y), the plane z = 0, and a cylindrical surface
whose generators are parallel to the z-axis and the directrix of
which .is the boundary of the
region D (Fig. 285). Calculate
z the volume of this solid V.
It has already been shown
that the volume of this solid
is equal to tne double integral
of the function f (x, y) over
the region D:
V=~~f(x, y)dxdy. (5)
D
Now let us calculate the
x volume of this solid using
Fig. 285.
the results of Sec. 4, Ch.
XI I, on the evaluation of
the volume of a solid from
the areas of parallel sections (slices). Draw the plane x= const
(a<x<b) that cuts the solid. Calculate the area S(x) of the
figure obtained by cutting x = const. This figure. is a curvilinear
trapezoid bounded by the lines z = f (x, y) (x . canst), z = 0, y = qJ,(x},
u= qJ2 (x). Hence, this area can be expressed by the integral
<J>2 (X)
S(x)= ~ f(x, y)dy. (6)

cp, (x) ., '
Knowing the areas of parallel sections, it i~ ~asy td find the

Calculating Double Integrals (continued) 61S
volume of the solid:

b
V = ~ S(x) dx;
a
or, substituting expression (6), we get for the area S(x)

b cp 2 (X)
V =~[ ~ f (x, y) dy] dx. (7)
a cp 1 (x)
In formulas (5) and (7) the left sides are equal; and so the right
sides are equal too:
b cp, (X)
~~f(x, y)dxdy=~ [ ~ f(x, y)dy]dx.
D a q.., (X)
It is now easy to figure out the geometric meaning of the evalu-

ation theorem of an iterated integral (Property 2, Sec. 2): the
volume V of a solid bounded by the surf ace z = f (x, y), the
, ''
''
----\ ' I
I
--;
,
~
!J'=-X---
0 ,,
2 y
X
X x=vy
Fig 286. Fig. 287.
plane z=O, and a cylindrical surface whose directrix is the

boundary of the region D, exceeds the volume of a cylinder with
base area S and altitude m, but is less than the volume of a
cylinder with base area S and altitude M [where m and M are
the least and greatest values of the function z = f (x, y) in the
region D (Fig. 286)]. This follows from the fact that iterated in-
fegral I D is equal to the volume v
of this solid.
6~0 Multiple Integrals
Example 1. Evaluate the double integral ~ ~ (4-x 2 - y') dx dy if the region

D
D is bounded liy the straight lines x=O, x= !, y=O, and y= ~ .
Solution. By the formula
'/2 I 8(2
~= S[S (4 -x 2
- y
2
) J
dx dy= S [ 4x-yx-:-i J: dx=
0 0
.,. 0
.,,
= s(4-y-~) dy=( 4y-Y; -~y) I =~5.
0 0
Example 2. Evaluate the double integral of the function f(x, y)=l+x+y

over a regipn bounded by the lines y=-x, x= y=2, Z=O (Fig. 287). VY.
Solution.
2 vy
v= S[S (l+x+y)d:c] dy= S[x+xy+x;] ~dy=
0 - y 0
2
= S [( VY"+Y Vi!+~)-( -y- y'+y2)] dy~

0
2
=S
0
[Vi/ +~+Y VY' -~] dy=
2 +!
+3y'+~ _fL]~ 44 .r-
1
_ [2y
- 3 4 5 6 o . 15 r 3
Note 2. Let a regular x-direction region D. be bounded by. the

lines
X= 'i', (y), X= 'i'z (y), y = C, y = d,
and let 'P, (y)...;;;; 'P. (y) (Fig. 288).
In this case, obviously,
d 1J), IYl
~ ~ (x, y) dx dy = ~ ( ~ f (x, y) dx) dy. (8)

D ~~~
To evaluate the double integral we must represent it as an

iterated integral. As we have already seen, this may be done in
two different ways: either by formula (4) or by formula (8).
Depending upon the type of the region D or the integrand in each
specific case, we choose one of the formulas to calculate the
double integral.
Calculating Double Integrals (continued) 621
Example 3. Change the order of integration in the integral

1 v:x
I=H ~ {(IC, y)dy)ux.
0 X
Solution. The region of integration is bounded by the straight line y=x

and the parabola y = (Fig. 289).VX:
Every straight line parallel to the x-axis cuts the boundary ol the region
at no more than two points; hence, we can compute the integral by formula
(8), setting
then
I 1/
I=~ ( ~ f (x, y) dx) dy.

0 lj"J
Exam;Jie 4. Evaluate ~ ~ e'X ds if the region D is a triangle bounded by

0
the straight lines y=x, y=O, and x=l (Fig. :l90).
lj
lj
~ ~
~ ~

)oc; ~ I
c
X
0 X 0 X
Fig. 288. Fig. 289. Fig. 290.
Solution. Replace this double integral by an iterated Integral using for-

mula (4). [If we used formula (8), we would have to integrate the function
..!!..
e' with respect to x; but this integral is not expressible in terms of elemen
tary functions]:
y I X Y I k
~ ~ ex dx = ~ [ ~ e-;:d y] dx = ~ [ xe
7
] : dx =
0 0 0 0
I 1
= 5
0
x(e-l)dx=(e-1) ~ 1=e
2
0
2
1
=0,859
---
Note 3. If the region D is not regular either in the x-direction

or the y-direction (that is, there exist vertical and horizontal
straight lines which, while passing through interior points of the
region, rut the boundary of the region at more than two points),
then we cannot represent the double integral over this region in
y
y
;_-2 X
0 X
Fig. 291. Fig. 292.
the form of an iterated integral. If we manage to partition the

irregular region D into a finite number of regular x-direction or
y-direction regions D~' D 1 , , Dn, then, by evaluating the double
integral over each of these subregions by means of the iterated
integral and adding the results obtained, we get the sought-for
integral over D.
Fig. 291 is an example of how an irregular region D may be
divided into two regular subregions D 1 and: D 2
Example 5. Evaluate the double integral
over the region D which lies between two squares with centre at the origin
and with sides parallel to the axes of coordinates, if each side of the inner
square is equal to 2 and that of the outer square is 4 (Fig. 292).
Solution. The region D is irregular. However, the straight lines x= - I
and x= I divide it into four regular subregions D~> D 2 , D 3 , D 4 Therefore,
~~
D
eX+Y dx~ ~ ~
D,
eX+Y ds+ E
D2
eX+Y d.s+ ~ ~ e~+f da+
D, '
H
D,
eX+Y ds.
Calculating Areas and Volumes 623
Representing each of these integrals in the form of an iterated integral,

we find
-t t 1 2
He"+Y ds= ~ [ ~ e"+Y dy] dx+ ~ [ ~ e"+Y dy Jdx +

D - - -1 I
1 -t 2 l
+ ~ [ ~ e"+Y dy
-1 -2
J H~dx+
1 -2
e"+Y dy Jdx=
= (e 2 -e- 2 } (e--e- 2 ) +(e 2 -e) (e-e- 1} +(e- 1 -e- 1 ) (e-e- 1) +
+
(e-e- 2 ) (e-e) =(e-e-')(e -e- 1 ) =4 sinh 3 sinh 1.
Note 4. From now on, when writing the iterated integral

b q> 2 (x)
I0 = ~ [ ~ f (x, y) dy Jdx,
a q> 1 (x)
we will drop the brackets containing the inner integral and will
write b q> 2 (xl
I0 = ~ ~ f (x, y) dy dx.
a q> 1 (xl
Here, just as in the case when we have brackets, we will consider
that the first integration is performed with respect to the variable
whose differential is written first, and then with respect to the variable
whose differential is written second. [We note, however, that this
is not the generally accepted practice; in some books the reverse
is done: integration is performed first with respect to the variable
whose differential is last.*>]
SEC. 4. CALCULATING AREAS AND VOLUMES

BY MEANS OF DOUBLE INTEGRALS
1. Volume. As we saw in Sec. I, the volume V of a solid

bounded by the surface z = f (x, y), where f(x, y) is a nonne-
gative function, by the plane z = 0 and by a cylindrical surface
whose directrix is the boundary of the region D and the generators
are parallel to the z-axis, is equal to the double integral of the
function f (x, y) over the region D:
V= Hf (x, y) ds.
D
> The following notation is also sometimes used:

b "' b "'
l D= ~ [ ~ f (x~ y) dy J dx = ~ dx ~ f (x, y) dy.
a IP1 a <1'1
Example 1. Calculate the volume of a solid bounded by the surfaces x=O,

y=O, x+y+z= 1, z=O (Fig. 293).
Solution.
V= ~ ~ (1-x-y)dydx,
D
where D is (in Fig. 293) the shaded triangular region in the xy-plane bounded
by the straight lines x=O, y=O, and x+y = 1. Putting the limits in the
double integral, we calculate the volume:
V=SJ . (l-x-y)dy dx= S[ o-xlu-~J:- . dx=S ~ o-x>'dx= ~.

0 0 0 0
1
Thus. V = cubic units.
6
Note t. If a solid, the volume of which is being sought, is
bounded above by the surface z = <D, (x, y).;;;;= 0, and below by the
surface z = <D, (x, y) ;;;;;= 0, and the region D
is the projection of both surfaces on the z
xy-plane, then the volume V of this solid
z
1 y y
X
Fig. 293. Fig. 294.
is equal to the difference between the volumes of the two "cy\indrical"

bodies; the first of these cylindrical bodies has the region D as
its lower base, and the surface z = <D 1 (x, y) for its upper base;
the second body also has D as its lower base, and the surface
z = <D, (x, y) for its upper base (Fig. 294).
Therefore, the volume V is equal to the difference between the
two double integrals
V= H<D
D
1 (x, y) ds- ~ ~ <D, (x, y) ds,
D
or
V = ~ ~ [<D 2 (x, y)-<D, (x, y)) ds. (1)
D
Calculating Areas and Volumes 625
Further, it is easy to prove that formula (1) holds true not

only for the case when <D 1 (x, y) and <D 2 (x, y) are nonnegative, but
also when <D 1 (x, y) and <D 2 (x, y) are any continuous functions
that satisfy the relationship
lll 3 (x, y);;;;;dl> 1 (x, y).
Note 2. If in the region D the function f (x, y) changes sign,
then we divide the region into two parts: 1) the subregion D 1
where f (x, y);;;;;;;.: 0; 2) the subregion D2 where f (x, y) .=;;;; 0. Suppose
the subregions D 1 and D. are such that the double integrals over
them exist. Then the integral over D 1 will be positive and equal
to the volume of the solid lying above the xy-p!ane. The integral
over 0 2 will be negative and equal, in absolute value, to the volume of
the solid lying below the xy-plane. Thus, the integral over D will
be expressed as the difference between the corresponding volumes.
2. Calculating the area of a plane region. If we form the inte-
gral sum of the function f (x, y) ==I over the region D, then this
sum will be equal to the area S,
n
S= ~ 1-!lsi,
i=l
for any method of partition. Passing to the limit on the right

side of the equation, we get
S= ~ ~ dxdy.
D
If D is regular (see, for instance, Fig. 280), then the area will
be expressed by the double integral
b cp 2 (X)
S = ~ [ ~ dy dx.
a <p, (.<)
J
Performing the Integration in the brackets, we obviously have
b
S = ~ [cp. (x)-cp 1 (x)] dx

a
(cf. Sec. 1, Ch. XII).
Example 2. Calculate the area of a region bounded by the c-urves
y=2-x, y=x.
Solution. Determine the points of intersection of the given curves lFig. 295).
At the point of intersection the ordinates are equal; that is,
x=2-x,
whence
x 2 +x-2=0,
x 1 =-2
x, = 1.
We get two points of intersection: M 1 ( - 2, - 2), M2 (I. l). Hence, the
required area is
t 2-X1 I
5( 5 )
2 1
(2-x 2 -x)dx= [ 2 xx
- - -x- ]
s=
-2 X
dy dx=
5
-2
3 2 -
=-.
27
6
SEC. 5. THE DOUBLE INTEGRAL IN POLAR COORDINATES
Suppose that in a polar coordinate system e, Q, a region D is

given such that each ray*) passing through an interior point of
the region cuts the boundary of D at no more than two points.
y
x,
-2
Fig. 295. Fig. 296.
Suppose that the region D is bounded by the curves Q = <1> 1 (9),

Q = <D, (8) and the rays e =a and e = ~. where <D, (8) ~<D. (9) and
a<~ (Fig. 296). Again we shall call such a region regular.
In the region D let there be given a continuous function of the
coordinates e and e:
Z=F(9, Q).
We divide D in some way into subregions ds,, ds ... , dsn.
) A ray is any half-line issuing from the coordinate origln, that is, from
the pole P.
The Double Integral in Polar Coordinates 627
Form the integral sum

n
Vn= ~ F(Pk) !:!.sk, (1)
k=t
where P k is some point in the subregion !:!.sk.
From the existence theorem of a double integral it follows that
as the greatest diameter of the subregion !:!.sk approaches zero,
there exists a limit V of the integral sum (1). By definition,
this limit V is the double integral of the function F (6, Q) over
the region D:
V = ~ ~ F (e, Q) ds. (2)
D
Let us now evaluate this double integral.
Since the limit of the sum is independent of the manner of
partitioning D into subregions !:!.sk, we can divide the region in
a way that is most convenient. This most convenient (for purposes
of calculation) manner will be to partition the region by means
of the rays e =eo, e = e, 0 =e ..... e =en (where eo= a, en=~.
9~<0,. <9 1 < ... <9n) and the concentric circles Q=Q 0 , Q=Q,,
, Q= Qm l where Q0 is equal to the least value of the function
cD, (9), and Qm, to the greatest value of the function <D, (9) in
the interval a..;;;:;; e..;;;:;;~. Qo < Q, < ... < Qml
Denote by !:!.sik the subregion bounded by the lines Q = Q; _,.
Q=Q;. 9=9k-t 9=9k.
The subregions !:!.s;k will be of three kinds:
I) those that are not cut by the boundary and lie in D;
2) those that are not cut by the boundary and lie outside D;
3) those that are cut by the boundary of D.
The sum of the terms corresponding to the cut subregions have
zero as their limit when !:!.Ok ----rO and I:!Q;-+ 0 and for this reason
these terms will be disregarded. The subregions !:!.sik that lie
outside D do not interest us since they do not enter into the sum.
Thus, the integral sum may be written as follows:
n
Vn = ~ l~ F (P;k) !:!.s;k],
k=I {
where P1k is an arbitrary point of the subregion !:!.s;k
The double summation sign here should be understood as
meaning that we first perform the summation with respect to
the index i, holding k fast (that is, we pick out all terms tha.t
correspond to the subregions lying between two adjacent rays *).
*) We note that in summing over the index i this index will not run
through all values from I . to m, because not all of the subregions lying
between the rays ll=Elk and El=Elk+t belong to D.
The outer summation sign signifies that we take together all the
sums obtained in the first summation (that is, we sum with
respect to the index k).
Let us find the expression of the area of the subregion 11s,.k
that is not cut by the boundary of the region. It will be equal
to the difference of the areas of the two sectors:
11s;k={(Q; + 11Q,.) 11ek-; Q:11ek= ( Q;

2
+ ~i,.) 11Q,.I1ek
or 11s,. k = Q; 11Q;I18 k where Qi < 11; < Qi + 11Qi.
Thus, the integral sum will have the form*)
n
Vn = ~ [~ F (e;, Q;) Q;f1Q;I1ek],
k=l j
where P (8~, Q;) is a point of the subregion 11sik

Now take the factor 11ek outside the sign of the inner sum
(this is permissible since it is a common factor for all the terms
of this sum):
Suppose that 11Qi .,..........Q and /1Gk remains constant. Then the
expression in the brackets wiJI tend to the integral
, ( a;)
~ F (e;, Q) Q dg.
, ( 8~)
Now, assuming that 11ek -0, we finally get**)

I} , (8)
V = ~ ( ~ F (8, Q) Q dQ) d9. (3)

a lei>,~)
l We can consider the Integral sum in this form because the limit of the
sum does not depend on the position of the point inside the subregion.
**) Our derivation of formula (:'l) is not rigorous; in deriving this formula
we first let ~Q; approach zero, leaving ~ek constant, and only then made ~ek
approach zero. This does not exactly correspond to the definition of a double
integral, which we regard as the limit of an integral sum as the diameters
of the subregions approach zero (i. e., in the simultaneous approach to zero
ol ~ek .and ~Q;). However, though the proof lacks rigour, the result is true
fi. e., formula (3) is true). This formula could be rigorously derived by the
same method used when considering the double integral in rectangular
coordinates. We also note that thi!> formula will be derived once again in
Sec. 6 with different reasoning (as a particular case of the more general
formula lor transforming coordinates in the double integral).
The Double Integral in Polar Coordinates 629
Formula (3) is used to compute double integrals in polar

coordinates.
If the first integration is performed over 9 and the second one
over Q, then we get the formula (Fig. 297)
o. ro, (Q)
V = ~ ( ~ F (9, Q) d9) Q dQ. (3')
Q, ro, (Q)
Let it be required to compute the double

integral of a function f (x, y) over a region
D given in rectangular coordinates:
Hf(x,
D
y)dxdy.
If D is regular in the polar coordinates 9, Fig. 297.

Q then the computation of the given integral
can be reduced to computing the iterated integral in polar
coordinates.
Indeed, since
X= Q COS 8, y = Q sin 8,
f(x, y) = f[Q cos 8, Q sin 8) = F (8, Q),
it follows that
ll >. (0)
~~f(x, y)dxdy=~ ( ~ f[Qcos8, QSin8]QdQ)d8. (4)
D a >, (0)
Example 1. Compute the volume V of a solid bounded by the spherical

surface
an. the cy Iinder

t 2 +y'-2ay=O.
Solution. For the region of integration here we can take the base of the
cylinder x 2 +y 2 -2ay=0, that is, a circle with centre at (0, a) and radius a.
The equation of this circle may be written in the form x 2 +(y-a) 1 =a2
(Fig. 298).
We calculate ! of the required volume V, namely that part which is
situated in the first octant. Then for the region of integration we will have
to take the semicircle whose boundaries are defined by the equations
X=!p 1 (y)=O, x=cp2 (y)= Y,..2ay-y 2 ,
y=O, y=2a.
The integrand is
z =I (x, y) = Y 4a 2
-x 2 - y 1
630 Multiple/ ntegrals
Consequently,
ta Ytay-y
! V= S( S 0 0
Y4a 2 -x 2 -y 2 dx) dy.
Transform the integral obtained to the polar coordinates e, Q:

X=Q cos e, Y=Q sin e.
Determine the limits of integration. To do so, write the equation of the
given circle in polar coordinates;
since
x"+u"=Q,
Y=Q sine,
Fig. 298. Fig. 299.
it follows that
or
Q=2astne.
He11ce, in polar coordinates (Fig. 299), the boundaries of the region are
defined by the e~uations
1t
~=
Q=Cll 1 (e)=0, Q=Cll,(e)=2asine, a=O,
2,
and the integrand has the form
F (e, Q)= V4a 2 -Q1
Thus, we have
n
ta sin a
~ = S( S V4a
2
-Q Q
2
dQ)de= S[
0 0 0

_=- ~ S((4a"-4a 1
stne)'1 - (4a 2)''I dll =
0
n
:= 3sa s . 4
(1-cose) de =-g a (3n-4).
0
The Double lntegr.al in Polar Coordinates 631
Example 2. Evaluate the Poisson integral
Solution. First evaluate the integral I R= ~ ~ e-x -Y' dx dy, where the region
D
of integration D is the circle
(Fig. 300).
Passing to the polar coordinates e, Q, we obtain
mR m R
IR= 5(5 e-P'Q dQ) de=- { 5e-P"I de =lt (1-e -R\_

0 0 0 0
Now, if we increase the radius R without bound (that is, if we expand

Without limit the region of integration), we get the so-called improper
iterated integral:
2lt "' Ut R
5(5 e-P'Q dQ) de= R!~"' 5(5 e-P'Q dQ )-de= l:"'
o a o o
1C (1-e -R') =lt.
We shall show that the integral ~ ~ e-x'-Y' dx dy approaches the limit :rt
D
if the region D' of arbitrary form t!Xpands in such manner that finally any
point of the plane gets into D' and remains there (we shall conditionally
indicate such an expansion of D' by the relationship D'---+- oo).
Fig. 300. Fig 301.
Let R 1 and R, be the least and greatest distances of the boundary of D'
from the origin (Fig. 301 ).
Since the function e-x-y is everywhere greater than zero, the following
inequalities hold:
IR,,;;;;, s
~ e-x~-Y' dx dy ~IR,.
D'
632 Multiple lnLegrals
or
R.' R..
n(1-e- ).;;;;~~e-x'-Y'dxdy.;;;;n(l-e- ).
0'
Since for D' ____,. oo it is obvious that R1 - oo and R ____,. oo, it follows
that the extreme parts of the inequality tend to one and the same limit n.
Hence, the median term also approaches this limit; that is,
pm ~ ~ e-x-y dx dy =n. (5)
0 -+a> 0'
As a particular instance, let D' be a square with side 2a and centre at

the origin; then
a a
~ ~ e-x-Ydx dy= ~ ~ e-x-Y'dxdy=
0' -a -a
a a a a
=~ ~ e-xe-Y'dx dy= ~ [ ~ e-xe-Y' dx ]dy ..
-a -a -a -a
Now take the factor e-Y" outside the sign of the inner integral (this is per-
missible since e-Y' does not depend on the variable of integration x). Then
a a
~ ~ e-x-y dx dy= ~ e-Y [ ~ e-x'dx ]dy.
o -a -a
a
Set ~ e-x dx = Ba Thi~. is a constant (dependent only on a); therefore,
-a
a a
~ ~ e-x-Y dx dy= ~ e-YBa dy= 8 4 ~ e-Y' dy.
0 -a -a
a
But the latter integral Is likewise equal to 8 4 {because ~ e-x ax=
-a
a
= ~ e-Y' dy);
-a
thus,
~ ~ e-x-v dx dy= BaBa = 8~.
0'
We pass to the limit in this equation, by making a approach infinity (in the
process, D' expands without limit):
a +"'
Jim \\ e-X'-Y' dx dy= lim 8~= Jim [ \ e-X' dx] = [ r
2
e-X' dx]".
D' -+ oo J J a -+ co a -+- oo J J
0' -a -a>
Changing Variables in a Double Integral (General Case) 633
But, as has been proved lsee (5)],
,lim ~ ~ e-x-y dx dy=rr..

0 ....... 0'
Hence,
"'
[~ e-...'dxr=rr..
or
_..,
This integral is frequently encountered in probability theory and in statistics.
We remark that we would not be able to compute this integral directly (by
means of an indefinite integral) because the derivative of e-x is not e:xpres-
sible in terms of elementary functions.
SEC. 6. CHANGING VARIABLES IN A DOUBLE INTEGRAL

(GENERAL CASE)
In the xy-plane let there be a region D bounded by the
line L. -Suppose that the _coordinates x and y are functions of
new variables u and v:
x=qJ(u, v), y='ljJ(u, o); (1)
let the functions qJ (u, v) and 'ljJ (u, v) be single-valued and con-
tinuous, and let them have continuous derivatives in some region D',
which will be defined later on. Then by formulas (1) to ea~;h
pair of values u and v there corresponds a unique pair of values
v
L'
V+LlV .I ~ IP:' ),_
~
v I ~v J
I lQ' R Llu P;
\
I' /
0. U U+LlU u
Fig. 302. Fig. 303.
x and y. Further, suppose that the functions qJ and 'ljJ are such
that if we give x and y definite values in D, then by formulas (I)
we will find definite values of u and v.
Consider a rectangular coordinate system Ouv (Fig. 302). From
the foregoing it follows that with each point P (x, y) in the
634 Mul(iple Integrals
xy-plane (Fig. 303) there is uniquely associated a point P' (u, (,')
in the uv-plane with coordinates u, v, which are determined by
formulas (1 ). The numbers u and v are called curvilinear coordi-
nates of the point P.
If in the xy-plane a point describes a closed line L bounding
the region D, then in the uv-plane a corresponding point will
trace out a closed line L' bounding a certain region D'; and to
each point of D' there will correspond a point of D.
Thus, the formulas (1) establish a one-to-one correspondence
bet~:een the points of the regions D and D', or, the mapping, by
formulas (1 ), of the region D onto region D' is said to be one-to-one.
In the region D' let us consider a line u = const. By formulas (1)
we find that in the xy-plane there will, generally speaking,. be a
certain curve that will correspond to it. In exactly the same way,
to each straight line v = const of the uv-plane there will correspond
some line in the xy-plane.
Let us divide the region D' (using the straight lines tt = const
and v = const) into rectangular subregions (we shall disregard
subregions that overlap the boundary of the region D'). Using
suitable curved lines, divide D into certain curvilinear quadran-
gles (Fig. 303).
Consider, in the uv-plane, the rectangular subr('gion !ls' bounded
by the straight lines u=const, u+flu=const, v=const, v+!lv=
=const, and consider also the curvilinear subregion !ls corresponding
to it in the xy-plane. We denote the areas of these subregions
by !ls' and !ls, respectively. Then, obviously,
!ls' =flu!lv.
Generally speaking, the areas !ls and !1s' are different.
In the region D, let there be a continuous function
z = f (x, y).
To each value of the function z = f (x, y) in the region D there
corresponds the very same value of the function z=F (u, v) in
the region D', where
F(u, v)=fl<p(u, v), '(u, v)].
Consider the integral sums of the function z over D. It is obvious
that we have the following equation:
~~ (x, y) !ls = ~ F (u,: v) !ls. (2)
. Let us. compute !ls, which is the area of the curvilinear quad-
rangle f\P 2 P 3 P 4 in the xy-plane (see Fig: 303).'
We determine the coordinates of its vertices:
When computing the area of the curvilinear quadrangle PI' P.,

P,, P. we shall consider the lines P,P., P.P . P,P., P.P, as
parallel in pairs; we shall also replace the increments of the functions
by corresponding differentials. We shall thus ignore infinitesi-
ma:ls of order higher than the infinitesimals l!.u, l!.v. Then formu-
las (3) will have the form
x, = illjl
x. = ilq> 0 0
x, = 
x.=<p(u, v)+ ov l!.v, . Y. = ' (u,
With these assumptions, the curvilinear quadrangle P,P.P,P~

may be ragarded as a parallelogram. Its area l!.s is approximately
equal to the doubled area of the triangle P,P.P, and is found by
the following formula of analytic geometry:
l!.s ~I (x,- x,) (y,-y.)- (x, -x.) (y, -y,) I=
tluL\v.* 1
. "I The doubled .Jines in the determinant indicate that the absolute 'value
1
of the determinant is taken.
636 Multiple integrals
We introduce the notation

acp acp
au au I
a"'a"' =
au au
Thus,
~s~III~s'. (4)
The determinant I is called the functional determinant of the
functions cp (u, v) and 'ljl (u, v). It is also called the Jacobian after
the German mathematician Jacobi.
The equality (4) is only approximate, because in the process
of computing the area of ~s we neglected infinitesimals of higher
order. However, the smaller the dimensions of the subregions ~s
and ~s', the more exact will this equality be. And it becomes
absolutely exact in the limit, when the diameters of the subregions
~s and ~s' approach zero:
II I= lim ,:ls,
dlamlls-+o ,:ls
Let us now apply the equation obtained to an evaluation of

the double integral. From (2) we can write
~ f (x, y) ~s ~ ~ F (u, v) I I I ~s'
(the integral sum on the right is extended over the region D~).
Passing to the limit as diam ~s' -o, we get the exact equation
~ ~ f (x, y) dx dy = ~ ~F (u, v) I I Idu dv. (5)
D 0'
This is the formula for transformations of coordinates in a doiJble

integral. It permits reducing the evaluation of a double integral
over a region D to the computation of a double integral over a
region D', which may simplify the problem. A rigorous proof of
this formula was first given by the noted R.ussian mathematician
M. V. Ostrogradsky.
Note. The transformation from rectangular coordinates to polar
coordinates considered in the preceding section is a special case
of change of variables in a double integral. Here, u = 8, v = e:
X=QCOS e, Y=Q sin e.
The curve AB (Q=Q,) in the xy-plane (Fig. 304) is transformed

into the straight line A' B' in the Oe-plane (Fig. 305). The
curve DC (Q = Q2 ) in the xy-plane is transformed into the straight
line D'C' in the Seplane.
The straight lines AD and BC in the xy-plane are transformed

into the straight lines A'D' and B'C' in the flQ-plane. The curves
L 1 and L1 are transformed into the curves L, and L~.
lj
c
9
c' . L2 D'
,(
D
I
N ~ ' N'
s L( A'
pq
X 0 "Q' 6 8
Ftg 304. Fig 305.
Let us calculate the Jacobian of transformation of the Cartesian

coordinates x and y into the polar coordinates e and Q:
ax ax
. ae a _ Q sin e cos eI =-Qsin9-Qcos9=-Q. 2 2
I= aya~ = .
_ __ 1 Q cos e sm e
ae aQ
Hence, Ill= Q and therefore
II >, (Q)
~~ f(x, y)dxdy= ~ ( ~ F(9, Q)QdQ) dfl.

D a >, (~)
This was the formula that we derived in the preceding section.

Example. Let it be required to compute the double integral
~ ~ (y-x) dxdy
D
over the region D in the xy-plane bounded by the straight lines

I 7 I
y=x+l. y=x-3, y=-
3 x+g y = -
3 x+5.
It would be difficult to compute this double integral directly; however, a
simple change or variables permits reducini: this integral to one over a rect-
angle whose sides are parallel to the coordinate axes.
Set
(G)
v Then the straight lines y=x+l. y=x-3 will

u--3 u-1 be transformed, respectively, into the straight
lines u =I, u = -3 in the uv-plane; and the
---1""-""""'~~r--v-~ straight lines y =-i-x+ ~
will be transformed into the
, y =-; x+5
straight lines
7 .
v=
3 , v=5-
-1!"'""'"""'1""4----v=% Consequent! y, the given region Dis transformed
into the rectangular region D' shown in
Fig. 306. It rem a ins to compute the Jacobian
of transformation. To do this, express x and y
---1-----+--l------i- in terms of u and v. Solving the system of
u equations (6}, we obtain
3 3 \ 3
Fig. 306. x=--;ru+-;rv; y=Tu+ 4 v.
Consequent! y,
OX ox
1-
--
au ov
oy oy
ou ov
and the absolute value of the Jaco.bian is Ill= !. Therefore,
SS<y-x)dxdy=SS [( +}u+t-o )-( -:

D D'
u+! v )] ! dudv=.
1
= SS +ududv = S S! ududv=-!8.
D' 7 -3
3
SEC. 7. COMPUTING THE AREA OF A SU RP ACE
Let it be required to compute the area of a surface bounded by

the line r (Fig. 307); the surface is defined by the equation
z = f (x, y), where the function f (x, y) is continuous and has con-
tinuous partial derivatives.
Denote the projection of the .line. r on the xy-plane by t.
Denote by D the region on the xy-plane bounded by the line L.
In arbitrary fashion, divide D into n elementary subregions
l\s,, l\s,, ... , l\sn. In each subregion l\s1 take a point P1 (s 1TJ1).
To the point P1 there will correspond; on the surface, a point
M1 [;. TJ;, f(;. fl;)}." .
Computing the Area of a Surface 639
Through M 1 draw a tangent plane to the surface. Its equation

is of the form
Z- Z; = f~ a,;, 11;) (X-~;)+ f~ (;, 11; ){y- 11;) (1)
(see Sec. 6, Ch. I X). In this plane, pick out a subregion llcr1
which is projected onto the xy-plane in the form of a subregion
11s1 Consider the sum of all the subregions llcr1 :
Fig. 307. Fig. 308.
We shall call the limit cr of this sum, when the greatest of the
diameters of the subregions !lcr1 approaches zero, the area of the
surface; that is, by definition we set
n
cr= lim ~ !lcr1. (2)
dlamt.o,_,.o i=l
Now let us calculate the area of the surface. Denote by 'Y; the
angle between the tangent plane and the xy-plane. Using a fami-
liar formula of analytic geometry we can write (Fig. 308)
11s; = licr 1 cos 'Y;
or
... ' /lSJ
f1rJ; = cos yl. (3)
The angle y 1 is at the sa.me .time the angle between the z-axis
and the perpendicular to the plane (I). Therefore, by equation
(1) and the formula of analytic geometry we have

I
cosY;= V , .
I +fr 2
(~;. TJ;) +fr !;;, TJj)
Hence,
t!J.a; = y I + t: (~;. lJ;) + t: (~;. lJ;) 6.s;.
Putting this expression into formula (2), we get
(J = lim
dlam dSj-+0 1
.= -{I + r~ (~;.
1
lJ;) + t~ C;. lJ;) 6.s;.
Since the limit of the integral sum on the right side of the last
equation is, by definition, the double integral
5) v 1+ (::r + (;;r dxdy, we finally get

a= 55 v + (~Y + (~;r dxdy.
D
1 (4)
D
This is the formula used to compute the area of the surface

z= f (x, y).
If the equation of the surface is given in the form
x = f.l. (y, z) or in the form y = x. (x, z),
then the corresponding formulas for calculating the surface are of
the form
a=5}, V1+(~)"+(~)"dydz, (3')
a= Uy 1 + (~~) + (:;) dz, 2

dx ( 3")
where D' and D" are the regions in the xy-plane and the xz-piane
in which the given surface is projected.
Example I. Compute the surface cr of the sphere
x+u'+z'= R' ..
Solution. Compute the surface of the upper half of the sphere;
Z= Y R -x1 2 -y 1
(Fig. 309). In this case

iJz x
ax= y R'-x-yl;
Computing the Area of a Surface 641.
i)z y
iJy = - VR-x-y
Hence,
V. +(:;r +(~;r = vR-~:-y.= v R::.x-y:

1
The region of integration is defined by the condition

x2 +y 2
EO;; R 2
Thus, by formula (4) we will have
R YRi=Xi"
1
2
cr= s( s
-R
__
-YR-x
y R -x
R -y
2 2 2
dy)d.YJ.
To compute the double integral obtained let us make the transformation

to polar coordinates. In polar coordinates the boundary of the region of
integration is determined by the equation Q = R. Hence,
zn R 2n
cr=2
0
s(sy R~Q2 Q
0
dQ) d6= 2R s[- YR-Q]~
0 .
d6=
n
=2R) R d6=4n R 2
0
Example 2. Find the area of that part of the surface of the cylinder
x 2 +y2 =a2
which is cut out by the cylinder
x+z=a.
Solution. Fig. 310 shows 1/Sth of the desired surface. The equation of the
surface has the form y = Va-x; z
I
!J
Fig. 309.
21-3388
therefore,
The region of integration is a quarter circle, that is, it is determined by

the conditions
Consequently,
a Ya-x a Y~ a
~a= S( S a
Y az-x' dz)dx-a
- ,}
r Y a2-x2
z 1 dx=a5dx=a 2
0 0 0 0 0
0=8a 1
SEC. 8. THE DENSITY OF DISTRIBUTION OF MATTER

AND THE DOUBLE INTEGRAL
In a region D, let a certain substance be distributed in such

manner that there is a definite amount of this substance per unit area
of D. We shall henceforward speak of the distnbution of mass,
although our reasoning will hold also for the case when speaking
of the distribution of electric charge, of quantity of heat, and so forth.
We consider an arbitrary subregion 11s of the region D. Let the
mass of substance associated with this given subregion be 11m.
Then the ratio !:
is called the mean surface density of the sub-
stance in the subregion 11s.
Now let the subregion As decrease and contract to the point
P (x, y). Consider the limit lim ~m. If this limit exists then
6s ..... o uS ' '
generally speaking, it will depend on the position of the point P,
that is, upon its coordinates x and y, and will be some function
f(P) of the point P. We shall call this limit the surface density
of the substance at the point P:
lim !m=f(P)=f(x, y). (I)

6.s-+ o s
Thus, the surface density is a function f (x, y). of the coordi-
nates of the point of the region .
. Conversely, let there be given, in a region D, the surface den-
sity of some substance as some continuous function f (P) = f (x, y)
The Moment of Inertia of the Area of a Plane Figure 643
and let it be required to determine the total quantity of substance

M contained in the region D. Divide D into subregions As; (i =
= 1, 2, ... , n) and in each subregion take a point Pi; then f(P;)
is the surface density in the point P;.
To within higher-order infinitesimals, the product f(P 1) As1
gives us the quantity of substance contained in the subregion As1,
and the sum
n
~ f(P 1) As1
/=1
expresses approximately the total quantity of substance distribu-

ted in the region D. But this is the integral sum of the function
f (P) in the region D. The exact value is obtained. in the limit
as As1 - 0.
Thus,*)
n
M = lim ~ f (Pi)As1 =
/l.Sj-+8 /=1
Hf (P)ds = Hf(x, y)dxdy,
D D
(2)
or the total quantity of substance in the region D is equal to the

double integral (over D) of the density f(P)=f(x, y) of this sub-
stance.
Example. Determine the mass of a circular plate of radius R if the sur~
face density f (x, y) of. the material of the plate at each point P (x, y) is pro.
portional to the distance of the point (x, y) from the centre of the circle, that
is. if
f(:c, y)=k Vx 2 +y 2
Solution. By formula (2) we have
M= ~ ~ k Y x2 + y 2 dx dy,
D
where the region of integration D is the circle x 2 y 2 ~ Rz. +

Passing to polar coordinates, we obtain .
Ut R R.
M=k S(SQQdQ )d8=k2n~ I=-}knR~.
0 0 0
SEC. 9. THE MOMENT OF INERTIA OF THE AREA

OF A PLANE FIGURE
The moment of inertia I of a material point M of mass m re-

lative to some point 0 is the product of the mass m by the
"') The relationship l'l.s1 -+ 0 is to be understood in the sense that the dia-
meter of the subregion f'lsl approaches zero. .
21*
square of its disi.ance r from the point 0:

I =tnr 2
The moment of inertia of a system of material points m 1 , tn 2 ,
.. , mn relative to 0 is the sum of moments of inertia of the
individual points of the system:
y
n
I=~ mir~.
i=l
_'1/.i Let us determine the moment of inertia
of a material plane figure D.
Let D be located in an xy-coordinate
~i x plane. Let us determine the moment of
. inertia of this figure relative to the origin,
Ftg. 311 assuming that the surf ace density is
everywhere equal to unity.
Divide the region D into elementary subregions t::..si- (i =I,
2, ... , n) (Fig. 311 ). In each subregion take a point P 1 with
coordinates ~i 'I'Ji Let us call the product of the mass of the
subregion t::.s1 by the square of the distance r~z = ~~I + 1'J2, an ele-
mentary moment of inertia t::.Ji of the subregion t::..si:
+
t::../ 1 = (~; 'I'J~) t::.s1,
and let us form the sum of such moments:
n
~ (~~+ 'I'J~) t::..si.
i= 1
This is the integral sum of the function f (x, y) = x 2 y over the +

region D.
We define the moment of inertia of the figure D as the limit
of this integral sum when the diameter of each elementary subre-
gion t::..si approaches zero:
n
10 = lim ~ (~:+1'J~)t::..s1
dlam.!l.sj-+0 i = 1
But the double integral H(x + y

D
2 2
) dx dy is the limit of this sum.
Thus, the moment of inertia of the figure D relative to the
origin is
I 0 = ~ ~ (x 2 + y ) dx dy,
2
(I)
D
where D is a region which c.oincides with the given plane figure

The Moment of Inertia of the Area of a Plane Figur~ 645
-.,----~--------------------
The integrals
(2)
fyy= Hx"dxdy
D
(3)
are called, respectively, the moments of inertia of the figure D

relative to the x-axis and y-axis.
Example I. Compute the moment of inertia of the area of a circle D of
radius R relative to the centre 0.
Solution. By formula (l) we have
10 = ~) (x"+y') dx dy.
D
To evaluate this integral we transform to the polar coordinates 8, Q The
equation of the circle in polar coordinates is Q= R.
Therefore
Note. If the surface density y is not equal to unity, but is soma

function of x and y, i. e., y = y (x, y), then the mass of the sub-
region fl.Si, will, to within infinitesimals of higher order, be equal to
y (~i TJi) fl.si and, for this reason, the moment of inertia of the
plane figure relative to the origin will be
10 = Hy(x, y)(x +Y )dxdy.
D
2 2
(1')
Example 2. Compute the moment of inertia of a plane material figure D

bounded by the lines y 2 =1-x; x=O, y=Orelative to the yaxis if the sur-
lace density at each point is equal to y (Fig. 312).
Solution.
Ellipse of inertia. Let us determine the moment of inertia of

the area of a plane figure D relative to some axis OL that passes
through the point 0, which we shall take as the coordinate ori-
gin. Denote by cp the angle formed by the straight line OL with
the positive x-axis (Fig. 313).
The normal equation of OL is
x sin cp-- y cos cp = 0.
The distance r of some point M (x, y) from this line is

r =I x sin cp-y cos cp j.
Fig. 312. Fig. 313.
The moment of inertia I of the area of D relative to OL is

expressed, by definition, by the integral
I= ~ ~ r 2 dx dy = ~ ~ (x sin cp- y cos cp) 2 dx dy =

D D
=sin 2 cp ~~ x dxdy-2sincpcoscp ~~ xydxdy+cos cp ~~ y dxdy.
2 2 2
D D D
Therefore
I= I YY sin 2 cp-21 xy sin cp cos cp +I xx cos 2 cp; (4)
here, I YY = ~ ~ X
2
dx dy is the moment of inertia of the figure
D
relative to they-axis, I xx = ~ ~ Y 2 dx dy is the moment of inertia
D
relative to the x-axis, and I xy = Hxy dx dy. Dividing all terms
D
of the latter equation by I, we get
l =I
XX
(cos_!)
yI
2
_ 21 ~y (sin.!)
y I (cos
Yl
qJ) +I (slnqJ)
yy YI
1
(5)
On the line OL take a point A (X, Y) such that
I
OA=y/'
To the various directions of ihe OL-axis, that is, to various values
The Moment of Inertia orthe Area of a Plane Figure 641
of the angle <p, there correspond different values I and different

points A. Let us find the locus of the points A. Obviously,
1 y 1 .
X = VI cos <p, = yT Stn <p.
By virtue of (5), the quantities X and Y are connected by the

relationship
1 = fxxX 2 -2fxyXY /YYY 2 (6) +
Thus, the locus of points A (X, Y) is a second-degree curve (6).
We shall prove that this curve is an ellipse.
The following inequality established by the Russian mathema-
tician Bunyakovsky *) holds true:
(H xy dx ely)"<(~~ x dxdy )0 ~ y 2
dxdy)
D D D
or
*) To prove Bunyakovsky's (also spelt Buniakowski) inequality, we con-

sider the following obvious inequality:
H [f(x, y)-A.<p(x, y)]dxdy;;:;,O,
D
where A. is a constant. The equality sign is possible only when f (x, y)-
-A.<p(x, y).,Q; that is, if f(x, y)=A.<p(x, y). If we assume that f((x, y))f:
O.
D D D
Consider the expression on the left as a function of A.. This is a second-
degree polynomial that never vanishes; hence, its roots are complex, and this
will occur when the discriminant formed of the coefficients of the quadratic
polynomial is negative, that is, '
n~ D
fo<Jl dx dy r~ ~ ~ /D
2
dx dy ~~
D
2
<p dx dy <0
or
(H D
{<pdxdy )"< ~~
D
f 2 dxdy HD
2
<p dxdy.
This is Bunyakovsky's inequality.

X
f: canst.
In our case, f(x, y)=x, <p(x, y)=y, -
y
Bunyakovsky's inequality is widely used in various fields of mathema-
tics. In many textbooks it is incorrectly called Schwarz' inequality. Bunya-
kovsky published it (among other important inequalities) in 1859. Schwarz
published his work 16 years later, in 1875.
648 Multiple lntrgtals
Thus, the discriminant of the curve (6) is positive and, con-

sequently, the curve is an ellipse (Fig. 314). This ellipse is called
the ellipse of inertia. The notion of an
lj
ellipse of inertia is very important in me-
chanics.
We note that the lengths of the axes of
the ellipse of inertia and its position in
the plane depend on the shape of the
given plane figure. Since the distance from
x the origin to some point A of the ellipse
is equal to Jy,
where I is the moment
of inertia of the figure relative to the
OA-axis, it follows that, after constructing
Fig. 314. the ellipse, we can readily calculate the
moment of inertia of the figure D relative
to some straight line passing through the coordinate origin. In
partiCular, it is easy to see that the moment of inertia of the figure
will be least relative to the major axis of the ellipse of inertia
and greatest relative to the minor axis of this ellipse.
SEC. 10. THE COORDINATES OF THE CENTRE OF GRAVITY

OF THE AREA OF A PLANE FIGURE
In Sec. 8, Ch. XII, it was stated that the coordinates of the

centre of gravity of a system of material points PI' P 2 , . . . , Pn
with masses m 1 , m1 , ... , mn are defined by the formulas
(1)
Let us now determine the coordinates of the centre of gravity of

a plane figure D. Divide this figure into very small elementary
subregions !lSi. If the surface density is taken as equal to unity,
then the mass of the subregion will be equal to its area. If it is
approximately considered that the entire mass of an elementary
subregion !lSi is concentrated in some point of it P 1 (s 1, 111), the
figure D may be regarded as a system of material points. Then,
by formulas (1), the coordinates of the centre of gravity of this figure
will be approximately determined by the equations
i=n t=n
~ ~~s1 ~ 'I'Ji~s~
/=1 /=1
Xc ~ i-tt Yc ~ ::-1=-':n--
~~S, ~~s~
/=1 1=1
The Coordinates of Centre of Gravity of a Plane Figure 649
In the limit, as !:lSi-+ 0, the integral sums in the numerators

and denominators of the fractions will pass into double integrals,
and we obtain exact formulas for compu- !J
ting the coordinates of the centre of gra
vity of a plane figure:
~~ x dx dy ~~ y dx dy
D D
X= y=-;;:-;;:--- (2)
c ~~ dx dy ' c ~~ dx dy
D. D a
These formulas, which have been derived Fig. 315.
for a plane figure with surface density I,
obviously, hold true also for a figure with any other density y
constant at all points.
If, however, the surface density is variable,
y=y(x, y),
then the corresponding formulas will have the form
By
D
(x, y) x dx dy Hy
D
(x, y) y dx dy
X = y = -=--=------
c ~~ y (X, y) d:c dy c ~ ~ y (X, y) dx dy
D D
The expressions Mv = ~ ~ y (x, y)xdxdy and Mx= Hv(x,y)

D D
y dx dy are called static moments of the plane figure D relative
to the y-axis and x-axis.
The integral ~ ~. y (x, y) dx dy expresses the quantity of mass
of the figure in question.
Example. Determine the coordinates of the centre of gravity of a
quarter of the ellipse (Fig. 315)
x' y'
12+--=-i =I,
assuming that the surface density at all points is equal to I.
Solution. By formulas (2) we have
t(~VT ~).,
a
~s va 2
x'x dx
0 4a
,, t(~VT } 1
-:nab
4
1
-:nab
4 '
=a:n'
SEC. 11. TRIPLE INTEGRALS
Let there be given, in space, a certain region ~' bounded by

a closed surface S. Let some continuous function f (x, y, z), where
x, y, z are the rectangular coordinates of a point of the region,
be given in the region V and on its boundary. For clarity, if
f (x, y, z) ~ 0, we can regard this function as the density of dis-
tribution of some substance in the region V.
Divide V, in arbitrary fashion, into subregions dV;; the sym-
bol dv; will denote not only the region itself, but its volume as
well. Within the limits of each subregion dv;, choose an arbitrary
point P; and denote by f (P;) the value of the function f at this
point. Form an integral sum of the type
~f(P;)dv; (1)
and increase without bound the number of subregions dv; so that
the largest diameter of dv; should approach zero.*> If the function
f (x, y, z) is continuous, there will be a limit of the integral
sums of type (1), where the limit of integral sums is to be un-
derstood in the same sense as for the definition of the double in-
tegral.**) This limit is not dependent either on the manner of par-
titioning the region V or on the choice of points P;; it is desig-
nated by the symbol ~ ~ ~ f (P) dv and is called a triple integral.
v
Thus, by definition,
. lim ~f(P;)dV;=~~~f(P)dv
d1am t.ui-+0 v
or
H~ f(P)dv= ~~~f(x, y, z)dxdydz. (2)
v v
*) The diameter of a subregio_n /'iV; is the maximum distance between
points lying on the boundary of the subregion.
**) This theorem of the existence of a limit of integral sums (that is, of
the existence of a triple integral) for any function continuous in a closed
region V (including the boundary) is accepted without proof.
Evaluating a Triple Integral 651
If f (x, y, z) is considered the volume density of distribution

of a substance over the region V, then the integral (2) yields the
mass of the entire substance contained in V.
SEC. 12. EVALUATING A TRIPLE INTEGRAL
Suppose that the spatial (three-dimensional) region V bounded

by the closed surface S possesses the following properties:
l) every straight line parallel to the z-axis and drawn through
an interior (that is, not lying on
the boundary S) point of the z
region V cuts the surface S at two
points;
z
Fig. 316. Fig. 317.
2) the entire region V is projected on the xy-plane into a

regular (two-dimensional) region D;
3) any part of the region V cut off by a plane parallel to any
one of the coordinate planes (Oxy, Oxz, Oyz) likewise possesses
Properties I and 2.
We shall call the region V that possesses the indicated proper-
ties a regular three-dimensional region.
To illustrate, an ellipsoid, a rectangular parallelepiped, a tet-
rahedron, and so on are examples of regular three-dimensional
regions. An instance of an irregular three-dimensional region is given
in Fig. 316. In this section we will consider only regular regions.
Let the surfa.ce bounding the region V below have the equa-
tion z = x (x, y), and the surface bounding this region above, the
equation z = 'ljJ (x, y) (Fig. 317).
We introduce the concept of a threefold iterated integral I v
over the region V, of a function of three variables f(x, y, z)
defined and contipuous in V. Suppose that the region D is the
projection of the region V onto the xy-plane bounded by the
lines
y= 2 (x) 1jl (x, y)
I v= ~ [ -) { ~ f (x, y, z) dz} dy J dx. (I)

a q> 1 (X) )(.(X, y)
We note that as a result of integration with respect to z and

substitution of limits in the braces (inner brackets) we get a func-
y tion of x and y. We then compute the
double integral of this function over the
region D as has already been done.
The following .is an example of the evalua-
tion of a threefold iterated integral.
Example I. Compute tlie iterated integral of
0 the function f (x, y, z) =xyz over the region V
bounded by the planes
0 y-o x=O, u=O, z=O, x+y+z=l.
Solution. This region is regular, it is bounded
Fig. 318. above and below by the planes. z=O and z=
= 1 - x - y and is projected on the xy-plane
into a regular plane region D, which is a triangle bounded by the straight
lines x =0, y = 0, y = l-x (Fig. 318). Therefore, the threefold iterated
integral lv is computed as follows:
lv= ~ ~ [t-~-Y xyzdz] da.

/) 0 -
Setting up the limits in the twofold iterated integral over the region D, we
obtain
= 51[1-X5xy(l-x-y) dy ] dx= 5I~(l-x)4 dx= 3!0 ,

0 ' 0
2
Let us now consider some of the properties of a threefold iterated

integral.
Property l. If a region V is divided into two regions V1 and
V2 by a plane parallel to some of the coordinate planes, then the
ihreefold iterated integral over V is equal to the sum of the three-
fold iterated integrals over the regions V 1 and V11
Evaluating a Trtple Integral
The proof of this property is exactly the same as that for

twofold iterated integrals. We shall not repeat it.
Corollary. For any kind of partition of the region V into a
finite number of subregions V P . . , Vn by planes parallel to the
coordinate planes, we have the equality
Iv=Iv+Iv++Iv.
~ tn 2
Property. 2 (Theorem of the evaluation of a threefold iterated

integral). If m and M are, respectively, the smallest and largest
values of the function f(x, y, z) in the region V, we have the
inequality
mV~Iv~MV,
where V is the volume of the given region and I v is a threefold
iterated integral of the function f (x, y, z) over the region V.
int~::'~ vL~t ~~,[~\r .~;;~~:. t:le~n]s;cr~ integral in the iterated
D X (x, y)
ljJ (x, y) ljl(x, y) ljl(x, y) ,P(x, y) ,
~ f(x, y, z)dz~ ~ Mdz-M ~ dz= Mz j =

X(x,y) x(x,y) r.(x,y) X(x,y)
= M [1jl (x, y)- X(x, y)].

Thus, the inside integral does not exceed the expression
M [', (x, y) -X (x, y)]. Therefore, by virtue of the theorem of
Sec. I for double integrals, we get (denoting by D the projection
of the region V on the xy-plane)
I v= ~ ~ ["' <1 uJf (x, y, z) dz] dcr ~ ~ ~ M ['IJ(x, y) -x (x, y)] dcr =
D X (x, y) D
=M~~ [1jl(x, y)-x(x, y)]da.

D
But the latter iterated integral is equal to the double integral of

the function 1jJ (x, y) -x (x,' y) and, consequently, is equal to the
volume of the region which lies between the surface z = x (x, y)
and z = 1jJ (x, y), that is, to the volume of the region V. There-
fore,
fv~MV.
It is similarly proved that I v~ mV. Property 2 is thus proved.

Property 3 (Mean-Value Theorem). The threefold iterated integ-
ral I v of a continuous function f (x, y, z) over a region V is equal
to the product of its volume V by the value of the function at
some point P of V; that is,

[cp (x) {
lv= ~
a
b 2
~
cp (x)
1
.P (x, y)
~ f (x,
1(. (x, y)
y, z)dz} dy J dx=f(P) V. (2)
The proof of this property is carried out in the same way as that
for a twofold iterated integral [see Sec. 2, Property 3, formula (4)].
We can now prove the theorem for evaluating a triple integral.
Theorem. The triple integral of a function f (x, y, z) over a
regular region V is equal to a threefold iterated integral over the
same region; that is,
~~~ {(X, y, Z) dV = { [cpr {.p (1 Y) f (X, y, Z) dz} dy] dx,

V a cp (x) (x, y)
1 1(. .
. Proof. Divide the region V by planes parallel to the coordinate

planes into n regular subregions:
~V 1 + ~v. + ... +~vn ..
As done above,: denote by I v the threefold iterated integral of
the function f(x, y, z) over the region f, and by hv1 the three-
fold iterated integral of this function over the subregion llvi' Then
by the corollary of Property I we can write the equation
lv=fav,+fav.++I~v" (3)
We transform .each of the terms .on the right by formula (2):
I v= f(P 1 ) llv 1 + f (P 1 ) ~v. + .. + f <Pn) ~vn, (4)
where P1 is some point of the subregion ~vi.
On the right side of this equation is an integral sum. It is
assumed that the function f (x, y, z) is continuous in the region V;
and for this reason the limit of this sum, as the largest diameter
of ~v 1 approaches zero, exists and is equal to the triple integral
of the function f (x, y, z) over V. Thus, passing to the limit in
(4), as diam ~vi -o, we get
Iv= ~~~ f(x, y, z)dv,

v
or, finally, interchanging the expressions on the right and left,
~ Hf (x,
V
y, z) dv = f(cp.r)
a cp1
{lj! T
(x)
Y)f (x, y, z) dz} dy] dx.
:X. (x, y)
Thus, the theorem is proved.

Evaluating a Triple Integral 655
Here, z=x(x, y) and Z='ljJ(x, y) are the equations of thesur-

faces bounding the regular region V below and above. The lines
y = cp, (x), y = cp 2 (x), x =a, x = b bound the region D, which is
the projection of V onto the xy-plane.
Note. Like in the case of the double integral, we can form a
threefold iterated integral with a different order of integration
with respect to the variables and with other limits, if, of course,
the shape of the region V permits this.
Computing the volume of a solid by means of a threefold
iterated integral. If the integrand f (x, y, z) = I, then the triple
integral over the region V expresses
the volume of the region V:
V= ~ ~ ~ dx dy dz. (5)
v
Example 2. Compute the volume of
the ellipsoid
x2 y2 z2
-.
a +-b +----z-=1.
c z-cVt-1rfzi
Solution. The ellipsoid (Fig. 319) Fig. 319.
is bounded belQW by the surface
z=-c
V 1 -x- - -
a
y , and above by,the surface z=c
b2
1 -x2- - y
a
-.
b'
Y
,!~e projection of this ellipsoid on the xy-plane (region D) is an ellipse,
~2 + ~ 2 ~ 1. Hence, reducing to a threefold iterated integral, we obtain
-(---x>"
-a- ( c -( -x
a b a- -u
- ) 2 2 b1 l
V=~ r ~ _ ~ dz dyJdx=
-aL-b-( a3x - c-( ---- x y-- a2 b2
=2c~
a --
by-x> a
~-- y1- ~:- ::
J
l -a
-b
-( x
-a
-
When computing the inside integral, x is held constant Make the substitu-
dy dx.
V-x. -v-x
tion:
y =b 1- -
a
sm t, dy = b 1- -
a
cos t dt.
Tbe variable y varies from- b , I 1- x to b , I 1- x ; therefore t

Jf a J! a
656 Multiple flltegrals
varies from - ~ to ~ Putting new limits in ihe integral, we get
v~2ct[L V(1-f, )-( 1- f.}'t V1-f, oodd}-
Hence,
~ 2cb!. [( ~)
1- t ooo' }x ~ ':: x:(a' -x~ dx ~ <~bo ,
I dt
4
V= n;abc.
3
If a= b =c, we gel the volume of :the sphere:
4
V ="3 n:a1
'
SEC. 13. CHANGE OF VARIABL~~ IN A TPIPLE INTEGRAL
1. Triple integral in cylindrical coordinates. In the case of cylin-

drical coordinates, the position of a point P in space is determined
by the three numbers e, Q, z, where 8 and Q are polar coordinates
of the projection of the point P on the xy-plane and z is the
z-coordinate of P, that is, the distance of the point to the xy-
plane-with the plus sign if the point lies above the xy-plane,
and with the minus sign if below the xy-plane (Fig. 320).
In this case, we divide the given three-dimensional region V
into elementary volumes by the coordinate surfaces a= ai, Q = ep
z = zk (half-planes adjoining the z-axis, circular cylinders whose
axis coincides with the z-axis, planes perpendicular to the z-axis).
The curvilinear "prism" shown in Fig. 321 will be a volume ele-
ment. The base area of this prism is equal, to within infinitesi-
mals of higher order, to Q~S~Q. the altitude is ~z (to simplify
notation we drop the indices i, j, k). Thus, ~v = Q~88Q~Z. Hence,
the triple integral of the function F (8, Q, -z) over the region
V has the form
l=H) F(S, Q, z)Qd8dedz. (1)
v
The limits of integration are determined by the shape of the
region V.
Change of Variables in a Triple Integral 657
.
X
fig. 320. Fig. 321.
If a triple integral of the function f(x, y, z) is given in

rectangular coordinates, it can readily be changed to a triple in-
tegral in cylindrical coordinates. Indeed, noting that
X=QCos8; y=Q sinS; z=z,
we have
H~f(x, y, z)dxdydz= SHF(S, Q, z)Qd8dQdz,
v v
where
f (Q cos 8, Q sin 8, z) =F (8, Q, z).
Example. Defermine .the mass M of a hemisphere of radius R with cenfre
at the origin, if lhe density F of its substance at each point (x, y, z) is pro
portional to fhe distance of this point from the base, that is, F = kz.
Solution. The equation of ,the upper part of the hemisphere
z= Y Rz-xz-yz
in cylindrical coordinates has lhe form
Z= y R 2
-Q
2

mcr.,
Hence,
M~ IJ I" """d'~ dz )dJd~

=s[sk~zJI"RrQQdQ]dfl= s[s ~ (R 2
-Q 2)QdQ]dfl=
0 0 0 0 0
s[i'- ~~]
2l't
=f dfl= ~ ~ 2n= k~R,

0
2. A triple integral in spherical coordinates. In spherical coor-

dinates, the position of a point P in space is determined by three
numbers, 0, r, cp, where r is
the distance of the point from z
the origin, the so-called radius
vector of the point, cp is the
z
X
Fig. 322. Fig. 323.
angle between the radius vector and the i-axis, 9 is the angle between
the projection of the radius vector on the xy-plane and the x-axis
reckoned from this axis in a positive sense (counterclockwise)
(Fig. 322). For any point of space we have
O<r<oo, O<cp<n; 0<9 <2n.
Divide this region V into volume elements du by the coordi-
nate surfaces r=const (sphere), <p=const (conic surfaces with vertices
at origin), 9=const (half-planes passing through the z-axis). To
within infinitesimals of higher order, the volume element du may
be considered a parallelepiped with edges of length Ar, rdcp,
r sin cp~9. Then the volume element is equal (see Fig. 323) to
du = r2 sin cp dr d9 dcp.
The triple integral of a function F (9, r, cp) over the region V
has the form
1= ~ ~ ~ F(9, r, cp)r 2 sincpdrd9dcp. (I')
v
The limits of integration are determined by the shape of
the region V. From Fig. 322 it is easy to est.ablish the expressi-
ons of Cartesian coordinates in terms of spherical coordinates:
x=rsincpcos9,
y = r sin cp sin 9,
z = r cos cp.
Change of Variables in_ a Triple Integral 659
For this reason, the formula for transforming the triple integral
from Cartesian coordinates to spherical coordinates has the form
~ ~ ~ f (x, y, z) dx dy dz =
v
= ~ Hf(r sin rp cos 8, r sin q; sin 0, r cos rp) r
v
2
sin cp dr d6 drp.
3. General change of variables in the triple integral.

Transformations from Cartesian coordinates to cylindrical and
spherical coordinates in the triple integral represent special cases
of the general transformation of coordinates in space.
Let the functions
x = rp (u, t, w),
y= 'ljJ (u, t, w),
z = x(u, t, w)
map, in one-to-one manner, the region V in Cartesian coordi-
nates x, y, z onto the region V' in curvilinear coordinates u, t, w.
Let the volume element Llv of the _.region V be carried over to the
volume element ilv' of V' and let
lim M, =
Av'-+0 /::,.fJ
Ill.
Then
~Hf(x, y, z)dxdydz=
v
=~ Hf[cp (u,
V'
t, w), 'ljJ (u, t, w), X (u, t, w)] III du dt dw.
As in the case of the double integral, I is called the Jacobian;
and as in the case of double integrals, it may be proved that
the, Jacobian is numerically equal to a determinant of order three:
ax axax
au at aw
ayayay
I= auaiaw
az az az
au at aw
Thus, in the case of cylindrical coordinates we have
X=QcosO, y=Q sinO, z . z (Q=U, 6=t, z=w);
I
l
cos6 -Q sin 6 0
I= sin 6 Q cos 6 0 = Q.
0 0 1
In the case of spherical coordinates:

X=rsincpcosO, y=rsincpsinO, z=rcoscp (r=u, cp=t, O=w);
sin cp cosO r coscp cosO - r sin cp sinO
I= sin cp sin 6 r cos cp sin 0 r sin cp cos 0 = r 2 sin cp.
cos cp - r sin cp 0
SEC. 14. THE MOMENT OF INERTIA AND THf. COORDINATES

OF THE CENTRE OF GRAVITY OF A SOLID
t. The moment of inertia of a solid. The moments of inertia

of a point M (x, y, z) of mass m relative to the coordinate axes
Ox, Oy, and Oz (Fig. 324) are expressed,
respectively, by the formulas
lxx=(Yz +zz)m,
I YY = (x + Z 2 ) m, I zz = (x + y ) m.
2 2 2
z
h
Fig. 324. Fig. -125.
The moments of inertia of a solid are expressed by the corre-

sponding integrals. For instance, the moment of inertia of a solid
relative to the z-axis is expressed by the integral lzz =
= ~ ~ ~ (x 2 + y 2 ) y (x, y, z) dx dy dz, where y (x, y, z) is the den-
v
sity of the substance.
Example I. Compute the moment of inertia of a right circular cylinder
of altitude 2h and radius R relative to the diameter of its median section,
considering the density constant and equal to 'Yo
Solution. Choose a coordinate system as follows: direct the z-axis along
the axis of the cylinder, and put the origin of coordinates at its centre of
symmetry (Fig. 325).
Moment of Inertia and Coordinates of Centre of Gravity of a Solid 661
Then the problem reduces to computing the moment of inertia of the

cylinder relative to the x-axis:
fxx= ~~~ (r;l+z 2) Yo dx dy dz.

v
Changing to cylindrical coordinates, we obtain
2. The coordinates of the centre of gravity of a solid. Like

what we had in Sec. 8, Ch. XII for plane figures, the coordinates
of the centre of gravity of a solid are expressed by the formulas
~~lxy(x, y, z)dxdydz ~~~ yy(x, y, z)dxdydz
x _ v , y _ ~v-::.'--------
c- Hv~ y (x, y, z) dx dy dz ' c - ~ ~ ~ y (x, y, z) dx dy dz
. v
~ ~ ~ zy (x, y, z) dx dy dz
v
Zc = E~ y (x, !J, z) dx dy dz
1
v
where 'Y (x, y, z) is the density.
. Example 2. Determine the coordinates of the centre of gravity of the
up per half of a sphere of radius R with centre at the origin, considering the
density Yo constant.
Solution. The hemisphere is bounded by the surfaces
Z= YR -x"-.r/,
2
z=O.
The z-coordinate of its centre of gravity is given by the formula
~ ~ ~zy 0 dx dydz
Zc = -;;-;;v-;;-----
~ ~ ~ y0 dx dy dz
v
Changing to spherical coordinates, we get
,. HfG, ,. .,. ,. d}l"

:rt
R' I
2l't 42 3
ec= :rt =-4--=sR.
2:rt 2 R -rtR'
Vo ~ [
0
~ ( ~ r 2 sin 
0 0
JdEl 6
Obviously, by virtue of the symmetry of the hemisphere, Xc=Yc=O.
SEC. 15. COMPUTING INTEGRALS DEPENDENT

ON A PARAMETER
Consider the integral dependent on the parameter a.

b
l(a)= Sf(x, a)dx.

a
(We examined such integrals in Sec. 10, Ch. XI.) We state with-
out proof that if a function f (x, a) is continuous with respect
to x over the interval [a, b] and with respect to a over the in-
terval [a,, a.], then the function
b
I (a)= Sf (x, a) dx
a
is a continuous function on [a,, a.]. Consequently, the function
I (a) may be integrated with respect to a on the interval [a 1 , a.]:
a1 a2 b
Sl(a)da= S[St(x, a) dx] da.

a1 a1 a
The expression on the right is an iterated integral of the func-

tion f(x, a) with respect to a rectangle situated in the plane xOa.
We can change the order of integration in this integral:
a, b b a1
S [Sf (x,
a1 a
a) dx] da = S f (x,
a
lS
a1
a) da] dx.
This formula shows that for integration of an integral depen-

dent on a parameter a, it is sufficient to integrate the element
of integration with . respect to the parameter a. This formula is
also useful when computing definite integrals.
Exercises on Chapter XIV 663
Example. Compute the integral

(1J
e-ax_e-bx
5- - - - d x .
0
X
This integral is not expressible in terms ol elementary functions. To evaluate

it, we consider another integral that may be readily computed:
(1J
5
0
e-"xdx = ~(a> 0).
Integrating this equation between the limits a= a and a= b, we get

b (1J b
5[5
a o
e-x dx Jda =5 ~a= In
a
: .
Changing the order of integration in the first integral, we rewrite this equa-
tion in the following form:
(1J b
5[5
o a
e-axda] dx=ln~,
whence, computing the inner integral, we get
(1J
e-ax -e-bx b
5
0
----dx=ln-,
x a
Exercises on Chapter XIV
1 2 4 2
Evaluate the integrals*: t. 55

o 1
(x 2 +Y 2) dxdy. Ans. ~ . 2. 55
a 1
dydx
(x+y)z
xVI n a
Ans. In ~ 3. 55
1 X
xy dx dy. Ans.
1
: . 4. 55
o a sln6
rdrd8. Ans.
N L
*) If the integral is written as 55f (x, y) dx dy then, as has already
M K
been stated, we can consider that the first integration is performed with
respect to the variable whose differential occupies the first place; that is,
N L N L
55 f
MK
(x, y) dx dy = 5-( 5f (x, y) dx) dy.
M K
a x a ay
5, (' Sxdydx
.) xa+ya'
Ans. lt a-
4
a arc tan -
a 6.
Sv Sxy dx dy. Ans. '24 .
1la4
0 X o -a
a
:n
bl
7. ss
b 0
Q dO dQ. A ns. 3
16 n
b2.
Determine the limits of integration for the Integral SSf (x, y) dx dy where
D
the region of integration is bounded by the lines: 8. x=2, x=3, y= -1,
a a 1 1-x
y=5. Ans. SS f(x, y)dydx. 9. y=0,y=l-x 2 Ans. SS f(x,y)dydx.

2 -1 -1 0
a Y a-x
Ans. SS f (x, y) dy dx. 11. Y= 1 :xz, Ans.
-a -Ya-x
2
I 'i+X> av+za
s sf
-1 x2
(X, y)dydx.l2.y=0,y=a,y=x,y=x-2a.Ans. S S f(x,y)dxdy.
0 v
2"
Change the order of integration in the integrals: 13. SSf (x, y) dy dx. Ans.
1 3
3/-
4 2 1Yx 1 v v
SS f (x, y) dx dy. 14. SSf (::, U) dy dx. Ans. S S f (x, y) dx dy.
3 1 0 X' o u
aYau-v a a 1lf1-x2
15. S S f (x, y) dx dy. Ans. S Lt(x, y) dy dx. 16. SS f (x, y) dy dx.
o o a-Ya-~ -1 o
1Y1-y2 11-y oYI='Xi
Ans. ~ ~ f (x, y) dxdy. 17. ~ ~ f (w, y) dxdy. Ans. ~ ~ f (x, y) dy dx +
o -V1-y o -Y1-y -1 o
I 1-X
+~ ~ f (x, y) dy dx.
0 0
Compute the following Integrals by <!hanging to polar coordinates!
aYa-x 2 a
18. S S Ya 2 - X 2 -y 2 dy dx. 1Ans. SSVa 2 2
Q QdQdEI= ~ a
3

Q 0 Q
Exercises on Chapter XIV _665
3t
a'Va-y a
19.
s s (x'+ y 2 ) dx dy. Ans. ss Q' dQ
4
de= 8na . 20. S"'"'Se-<x'+Y>dy dx.
0 0 0 0
3t 3t
2 2Q v.ax-x 2 2a cos 9
Ans. ss
0 0
e-P" Q dQ dO=~ . 21.
s s dy dx. Ans.
ss QdQ dO=
na
=-y
Transform the double integrals by introducing new variables u and u con-
e ~x
nected with x and y by the formulas x=u-uu, y=uu: 22. ) ) f (x, y) dy dx.
o ax
~ e
1+~ t=V c b
Ans. ) ) f (u-uu, uu) u du du. 23. ) ) f (x, y) dy dx.
a u-o 0 0
t+a
b c b
b+c t=V 1 v
Ans. ) ~ f(u-uu, ~u)ududu+ ~ )t(u-uu, uu)ududu.
b 0
b+c
Calculating Surfaces by Means of Double Integrals
24. Compute the area of a figure bounded by the parabola y 2 =2x and the
straight line .11 =X. Ans. ; .
25. Compute the area of a figure bounded by the lines y 2 = 4ax, x + y = 3a,
lO 2
y=O. A ns. a.
3
1 1 1
26. Compute the area of a figure bounded by the lines x 2 +Y2 =a-;:
a
x+y=a. Ans. . 3
27. Compute the area of a figure bounded by the lines y=sinx, y=cosx,
x=o: Ans. 2-1.
na
28. Compute the area of a loop of the curve Q=a sin 28. Ans. B.
29. Compute the entire area bounded by the lemniscate Q2 =a 2 cos 2<p.
Ans. a. 2 2
30. Compute the area of a loop of the curve ( i2 y ) ' = 7.
x +bT 2xy
Hint. Change to new variables x=Qa cos 8 and Y=Qb sinS. Ans.
a2 b 2
c.-.
Calculating Volumes
31. Compute the volumes of solids bounded by the following surfaces:

x y z abc
a+b+c=l. x=O, y=O, z=O. Ans. . 32. z=O, x2 +y 2 =1, x+u+
6
+z=3. Ans. 3n:. 33. (x-1) 2 +(y-1) 2 =1, xy=z, z=O. Ans. n:. 34. x1 +y 2 -
3
-2ax=0, z=O, x 2 +y 2 =z 2 Ans. : a 1 35. y=x 2 , x=y 2 , z=O, z= 12+
+u-x.2 A ns. 549

140
.
36. Compute the volumes of solids bound10d by the coordinate planes, the
plane 2x+3y-12=0 and the cylinder z= ~ y 2 Ans. 16.
37. Compute the volumes of solids bounded by a circular cylinder of
radius a, whose a.xis coincides with the z-axis, the coordinate planes and the
I
pane X
a+a= z 1. A ns. a s(l14
- .1) . 3
38. Compute the volumes of solids bounded by the cylinders x + y 2 = a,
1
x'+z 2 =a2 Ans. : a1 39, y2 +z 2 =x, x=y, z=O. Ans. . 40. x+u+ i4
+z2=a2, x+y 2 =R 2 , a> R. Ans. 4 .r--
n:[a 1 -(r a 2 -R 2) 1 ]. 41. az=.t 2 +y 2 ,
3
. 3
z=O, x'+y 2 =2ax. Ans. 2 n:a 1 42. Q2 =a 2 cos 28, x 2 +u'+z 2 =a 2 , z=O.
(Compute the volume that is interior with respect to the cylinder.) Ans.
I
g-a' .r-
(3n:+20-16 r 2).
Calculating Surface Areas
43. Compute the area of that part of the surface of the cone x + y 2 = z
which is cut out by the cylinder x+y 2 =2ax. Ans, 2n:a J(2.
+
44. Compute the area of that part of the plane x y + z = 2a, which lies
J
in the first octant and is bounded by _the cylinder x'+y 2 = a. Ans. n:: 3.
45. Compute the surface area of a spherical segment (minor) if the radius
of .the sphere is a, while the radius of the base of the segment is b.
Ans. 2n: (a 2 -a Va- b 2 ).
46. Find the area of that part of the surface of the sphere x+y 2 +z 2 =a 2
2 2
which is cut out by the surface of the cylinder ~ 2 +p= 1 (a> b). Ans.
4n:a'-8a 2 -arc sin

Ya 2
-b 2
.
a
47, Find the surf ace area of a solid that is the common part of two
cylinders x 2 +y 2 =a 2 , y2 +z 1 =a 2. Ans. 16a 2
48. Compute the area of that part of the surface of the cylinder
x 2 +y 2 =2ax, which lies between the plane z=O and the cone x2 +Y 1 =z 2
Ans. Ba 2
49. Compute the area of that part of the surface of the cylinder x+ y= a
which lies between the plane z = mx and the plane z = 0, Ans. 2ma.
Exercises on Chapter XIV 667
50. Compute the area of that part of the surface of the paraboloid
y'+z'=2ax, which lies between the parabolic cylinder y 2 =ax and the plane
x =a. Ans. 1 na (3 .r-
r 3-1).
3
Computing the Mass, the Coordinates of the Centre

of Gravity, and the Moment of Inertia of Plane Solids
(In Problems 5164 we consider the surface density constant and equal
to unity.)
51. Determine the mass of a ~lab in the shape of a circle of radius a, if
the density at any point P is inversely proportional to the distance of P
from the axis o'f the cylinder (the proportionality factor is K). Ans. naK.
52. Compute the coordinates of the centre of gravity of an equilateral
triangle if we take its altitude for the x-axis and the vertex; of the triangle
f or t he coor d.tna t e ongw.
- A ns. x= a- Jl3
- ; Y= 0 .
3
53. Find the coordinates of the centre of gravity of a circular sector of
radius a, taking the bisector of its angle as the x-axis. The angle of spread
. 2aslna
of the sector ts 2a. Ans. Xc = - - , Yc=O.
3a .
54. Find the coordinates of the centre of gravity of the upper half of the
4a.
circle x 2 +Y 2 ':"'a 2 Ans. Xc=O; Yc=- rr.
3
. 55. Find the coordinates of the centre of gravity of the area of one arc
5
of the cycloid x=a(t-slnt), y=a(l-cost). Ans. Xc=an, Yc= :.
56. Find the coordinates of the centre of gravity of the area bounded by
a loop of the curve Q2 =a cos 20. Ans.
na2
Xc=--- , Yc=O.
8
57. Find the coordinates of the centre of gravity of the area of the car
dioid Q=a(l+cosO). Ans. x~= 5:, Yc=O.
58. Compute the moment of inertia of the area of a rectangle bounded by
the straight lines x=O, x=a, y=O, y=b relative to the origin. Ans.
ab (a'+ b')
3
2 2
59. Compute the moment of inertia of the ellipse :. + ~. = ll a) relative
to the y-axis; b) relative to the origin. Ans. a) n~b ; b) n:b (a 2 + b).

60. Compute the moment of inertia of the area of the circle Q= 2a cos 0
relative to the pole. Ans. ~ na~.
61. Compute the moment of inertia of the area of the cardioid
35na
Q=a (I -cos 0) relative to the pole. Ans. ! 6 .
62. Compute the moment of inertia. of the area of the circle (x-a) 2 +
+{y-b) 2 =2a 2 relative to they-axis. Ans. 3na.
63. The density at any point of a square slab with side a is a proportion
al to the distance of this point from one of the vertices of the square.
Compute the moment of inertia of the slab relative to the side passing
1
through this vertex. Ans. ka 5 [7 V2+31nCV2+1)] where k is the
40
proportionality factor.
64. Compute the moment of inertia of the area of a figure, bounded by
the parabola y=ax and the straight line x=a, relative to the straight line
u=-a. Ans.
8 sa.
Triple Integrals
65. :compute SSS (x::~z~J) 3 if the region of

integration is bounded
In 2 5
by the coordinate planes and the plane x+u+z= 1. Ans. 2:-
16
a x y
66. EvaluateS [S ( Sxyz dz) dyJ dx. Ans. :; ,

0 0 0
67. Compute the volume of a solid bounded by the sphere x+u'+z=4
- 19
and the surface of the paraboloid x+ y 2 = 3z. Ans. n.
6
68. *) Compute the coordinates of the centre of gravity and the moments
of inertia of a pyramid bounded by the planes x=O, y=O, z=O; : + t+
z a b c 3
a bc 3
b ac 2
c ab
+c-=1. Ans. Xc=Jf Yc=-;r Zc=-;r; fx=-r;o, fy= 00 , lz=-go,
I 0 -_abc( a'+b+ c2 ) .
60
69. Compute the moment of inertia of a circular right cone relative to its
1
axis. Ans. nhr where h is the altitude and z; is the radius of the base of
10
the cone.
70. Compute the volume of a solid bounded by a surfac with equation
1
(x 2 +y 2 +z 2 ) 2 =a 8x. Ans. 8
3 na
71. Compute the moment of inertia of a circular cone relative to the
nhr 2
diameter of the base. Ans. 6o (2h 2 + 3r 2 ).
- 72. Compute the coordinates of the centre of gravity of a solid lying
between a sphere of radius a and a conic surface with angle at the vertex 2a,
if the vertex of the cone coincidEs with the centre of the sphere. Ans. Xc=O,
3 -
Yc=O, Zc=s a(l+cosa) (the z-axis is the axis of the cone, and the ver-
tex lies at the origin).
*) In Problems 68, 69 and 71 to 73 we consider th~ density constant and

equal to unity.
Exercises on Chapter X IV 669
73. Compute the coordinates of the centre of gravity of a solid bounded

:by a sphere of radius a and by two planes passing through the centre of the
sphere and forming an angle of 60. Ans. Q= a, 6=0, <p= ~ (the line
9
16
of intersection of the planes is taken for the z-axis, th:: centre of the
sphere for the origin; Q. e,
 0) compute the
integrals JcoV~x ssiV~x . y ~ ; ~ ~ .

QO
0
and
"'
0
Ans.
CHAPTER XV
LINE INTEGRALS AND SURFACE INTEGRALS
SEC. 1. LINE INTEGRALS
Let the point P (x, y) be in motion along some plane line L

from the point M to the point N. To P is applied a force F
which varies in magnitude and
direction with the motion of P;
it is thus some function of the
!Ji+l!.y; coordinates of P:
F=F(P).
Let us compute the work A
of the force F as the point P is
translated from M to N (Fig. 326).
To do this, we divide the curve MN
into n arbitrary parts by the points
M 0 =M, M~' M 2 , , Mn=N in
0 the direction from M toN and we
Fig. 326. denote by !lsi the vector MiMi+
We denote by Fi the magnitude of
the force F at the point Mi. Then the scalar product F/J.si may
be regarded as an approximate expression of the work of the
..-
force F along the arc M1Mi+:
A1 F~:J F/!!Si.
Let
F,=X(x, y)l+Y(x, y)j
where X (x, y) and Y (x, y) are the projections of the vector F on
the x- and y-axes. Denoting by !1x1 and 11y1 the increments of the
coordinates x 1 and y1 when changing from the point Mi to the
point M 1+v we get
Hence,
Fil1si =X (xi, Yt) !lxt + Y (xi, Yi) 11Y;
The approximate value of the work A of the force F over the
.entire curve MN will be
n n
A F~:J ~ F111si = ~[X (xt, y1) 11xi + Y (x1, y1)11y;]. (1)
/=1 /=1
Line Integrals 671
Without making any precise statements, we shall say that if

there exists a limit of the expression on the right as tl.si ___,. 0
(here, obviously, tl.xi ~ 0 and tl.yi---+ 0}, then this limit expresses
the work of the force F over the curve L from the point M to
the point N:
n
A= lim ~ [X (xi, Yi) tl.xi + Y (xi, Yi) tl.y,.]. (2}

lixt ~ o i=t
/!.y; ..... 0
The limit*) on the right is called the line integral of X (x, y)

and Y (x, y) over the curve L and is denoted by
A=~ X (x, y)dx+ Y (x, y)dy (3)
L
or
(N)
A= ~ X (x, y)dx+ Y (x, y)dy. (3')
(M)
Limits of sums of type (2) frequently occur in mathematics and

mechanics; here, X (x, y) and Y (x, y) are regarded as functions
of two variables in some region D.
The letters M and N, which take the place of the limits of
integration, are in brackets to signify that they are not numbers
but symbols of the end points of the line over which the line
integral is taken. The direction of the curve L from M to N is
called the sense of integration.
If the. curve L is a space curve, then the line integral of three
functions X (x, y, z), Y (x, y, z), Z (x, y, z) is defined similarly:
~ X(x, y, z)dx+Y(x, y, z)dy+Z(x, y, z)dz=

L .
n
= lim ~ X (xk, yk, zk) tl.xk + Y (xk, Yk zk) tl.yk + Z (xk, yk, zk) tl.zk.
f!.Xk-+ 0 k=l
I!.Yk-+ o
/!.Zk _:.. 0
The letter L under the integral sign indicates that the integration
is performed along the curve L.
We note two properties of a line integral.
Property 1. A line integral is determined by the element of
integration, the form of the curve of integration, and the sense
of integration.
) Here, the limit of the integral sum is to be understood in the same

sense as in the case of the definite integral, see Sec. 2, Ch. XI.
672 Line Integrals and Surface Integrals
- A line integral changes sign when the sense of integration is

reversed, since in that case the vector !l.s, and hence its proje-
ctions !l.x and !l.y, changes sign.
....-._ ....-._ -
Property 2. Divide the curve L by the point K into pieces L 1
+
and L 2 so that MN = MK KN (Fig. 327). Then, from formula (1)
it follows directly that
(N) (I() (N)
~X~+Y~=~X~+Y~+~X~+Y~
(M) (M) (I()
This relationship holds for any number of terms.

It will further be noted that the definition of a line integral
holds true also for the case when the curve L is closed.
In this case, the initial and terminal points of the curv.e coin-
cide. Therefore, in the case of a closed curve we cannot write
(N)
~X dx+ Y dy, but only ~X dx+ Y dy; and we

(M) L
have to indicate the direction of circulation
(sense of description) over the closed curve L.
The line integral over a closed con tour L is
Fig. 327.
frequently denoted also by the symbol p
X dx+
L
+Ydy.
Note. We arrived at the concept of a line integral while consi-
dering the problem of the work of a force F on a curved path L.
Here, at all points of the curve L the force F was given as
a vector function F of the coordinates of the point of applica-
tion (x, y); the projections of the variable vector F on the coor-
dinate axes are equal to the scalar (numerical, that is) functions
X (x, y) and Y (x, y). For this reason, line integral of the form
j X dx + Y dy may be regarded as an integral of the vector
L
function F given by the projections X and Y.
The integral of a vector function F over the curve L is deno-
ted by the symbol
~ Fds.
L
If the vector F is defined by its projections X, Y, Z then this

integral is equal to the line integral
~X dx+ Y dy+Z dz.
L
Evaluating a Line Integral 613
As a particular instance, if the vector F lies in the xy-plane,

then the integral of this vector is equal to
~ Xdx+Ydy.
L
When the line integral of a vector function F is taken along
a closed curve L, this line integral is also called a circulation of
the vector F over the closed contour L.
SEC. 2 EVALUATING A LINE INTEGRAL
In this section we shall make more precise the concept of the

Iimit of the sum ( 1) of Sec. 1 and in this connection we shall
make more precise the concept of y
the line integral and indicate a N
method for calculating it.
Let a curve L be represented by tJyt===:::::;;;~C
equations in parametric form:
X= cp (t), y = 'Jl (t).
Consider the arc of the curve MN
(Fig. 328). Let the points M and N 0 !Jx X
correspond to the values of the para- Fig. 328.
IJleter a and ~ Divide the arc MN
into subarcs 11si by the points M, (x,, y,), M 1 (x2 , y 1 ), ,
Mn(Xno Yn}, and put Xi=cp(ti), Yi='iJl(ti).
Consider the line integral
~ X (x, y) dx + Y (x, y) dy (l)
L
defined in the preceding section. We give without proof the exist-
ence theorem of a line integral. If the functions cp(t) and ~(t)
are continuous and have continuous derivatives cp' (t) and '111' {t),
and also continuous are the functions X [cp (t), 'Jl (t)] and Y [cp (t), 'll (t)]
as functions of t on the interval [a, ~]. then the following limits
exist:
lim
t.x1 .... o i=l
X (xi, Yi) 11xi = SX (x, y) dx, }
(2)
a~~~ o ~~ Y (xi, y;) 11yi = ~ Y (x, y) dy,
where xi and Yi are the coordinates of some point lying on the
arc !1si. These limits do not depend on way the arc L is divided
22 - 3388
into subarcs 11s1, provided that !1st-+ 0 and do not depend on the
choice of the point M 1 (x1 , y1) on the subarc f1s.; they are called
line integrals and are denoted as '
n
lim ~ X (x1 , y;) 11x1 = SX (x, y) dx,
lixj-+ o /=1 L
n
lim ~ Y (x1, y;) 11y1 = S Y (x, y) dy.
dYI-+ 0 /=1 L
Note. From this theorem it follows that the sums defined in

the preceding section, where the points M;(xi, fit) arethe extremi-
ties of the subarc 11s1 and the manner of partition of the arc L
into subarcs 11s; is arbitrary, approach the same limit-the line
integral.
This theorem makes it possible to develop a method for comput-
ing a line integral.
Thus, by definition, we have
IN) n
S X (x, y) dx = lim ~ X (x 1, y1) 11x1, (3)
(M) AXj-+ 0 /=1
where
f1x1 =X; -x1_ 1 = qJ (t;)-cp (t;- 1 ).
Transform this latter difference by the Lagrange formula
11X; = qJ (t; )-cp (i;-1) = cp' ('t'l)(t;-t;-l) = cp' ('t';) M;,
where 't'; is some value of t that lies between the values t 1 - I
and t1 Since the point x1, y1 on the subarc 11s1 may be chosen
at pleasure, we shall choose it so that its coordinates correspond
to the value of the parameter 't';:
X;=Cfl('t';). Yi='ijl('t';)
Substituting into (3) the values of x1, y1 and 11x1 that we have
found, we get
(N) n
S X(x,y)dx= lim ~X[cp('t';)'iJl('t';)]cp'('t';)M 1
1M) At1-+ o i=1 .
I
On the right is the limit of the integral sum for the continuous
function of a single variable X [cp (t), 'ljl (t)] cp' (t) on the interval
[a, ~].
Evaluating a Line Integral 675
Hence, this limit is equal to the definite integral of this

function:
(N) ~
~ X (x, y) dx =~X [cp (t), ' (t)] cp' (t) dt.

(M) a
In analogous fashion we get the formula

(N) ~
~ Y (x, y) dy = ~ Y [ cp (t), ' {t)] 'B' (t) dt.

(M) a
Adding these equations term by term, we obtain

(N) ~
S X (x, y) dx + Y (x, y) dy = ~ {X [cp (t) '. (f)] +

(M) a
+ y [cp (t), '!? (t)] 'B' (f)} dt. (4)
This is the desired formula for computing a line integral.

In similar manner we compute the line integral
~ X dx + Y dy + Z dz
over the space curve defined ?Y the equations x = cp (t), y = ~ (t),
z =X (t).
Example I. Compute the line integral of three functions: x 3 , 3zy 2 , - xy
(or, which is the same thing, of the vector function x"i+3zy 2j-x 2yk) along
a segment of a straight line issuing from the point M (3, 2, 1) to the point
N (0, 0, 0) (Fig. 329).
Solution. To find the parametric equations of the line M N, along which
the integration is to b~ performed, we write the equation of the straight line
that passes through the given two points:
X y Z
a=2=1:
and denote all these relations by a single letter t; then we get the equations
of the straight line in parametric form:
x=31, y=2t, z=t.
Here, obviously, to the origin of the segment MN corresponds the value of
the parameter t = 1, and to the terminus of the segment, the value t =0. The
derivatives of x, y, z with respect to the parameter t (which will be needed
for evaluating the line integral) are easily found:
22*
Now the desired line integral may be computed by formula (4):

~) 0
~ x' dx+3zy 2 dy-x 2y dz = ~[(3t)~3+3t (2()22-(3t) 2 2ll]dt =

(M) 1
0
87
= S87t 3 dt= -4.
1
Example 2. Evaluate the line integral of a pair of functions: 6x 2y, 10xy 2

along a plane curve y=x from the point M (1, I) to the point N (2, 8)
(Fig. 330).
Solution. To compute the required y
integral +----<>t{
tN)
~ 6x 2 y dx+ 10xy 2 dy
(M)
we must have the parametric equations of

the given curve. However, the explicitly
defined equation of the curve y=x 3 is a
special case of the parametric equation:
x.
Fig. 329. Fig. 330.
here, the abscissa x of the point of the curve serves as the parameter, and
the parametric equations of the curve are
X=X, Y=X~.
The parameter x varies from x 1 = 1 to x 2 = 2. The derivatives with respect
to the parameter are readily evaluated:
Hence,
<=I. Y~=3xz.
(N) 2
~ 6x 2 y dx+ 10xy2 dy= ~ [6x 2x 1 1 + 10xx 3x 8 2

] dx=
(M) 1
2
= ~ (6x 5 + 30x 9) dx = (x 8 + 3x 10
]: = 1084.
1
Evaluating a Line l11tegral 677
We now indicate certain applica IJ

tions of a line integral.
1. The expression of the area of a
region bounded by a curve in terms
of a line integral. In an xy-plane let
there be given a region D (bounded 11
by the contour L) such thai:
any straight line parallel to one of
the coordinate axes and passing
through an interior point of the reg-
ion cuts the boundary L of the re- 0 q
gion in no more than two points
(which means that the region D is Fig. 331.
regular) (Fig. 331 ).
Suppose that the region D is projected on the x-axis in the
interval [a, b], and it is bounded below by the curve (1 1 ):
!f=Yt (x),
and above by the curve (l 2 ):
!f= Yz (x),
[Yt (x).;;;;;; Yz (x)].
Then the area of the region D is
b b
S= ~ y 2 (x) dx- ~ y 1 (x) dx.

a a
-
But the first integral is a line integral over the curve 12 (MPN),
since y = Y2 (x) is the equation of this curve; hence,
b
~Y 2 (x)dx= ~ ydx.
a MPN
The second integral is a line integral over. the curve 11 (MQN),
that is,
b
-
~ !:/ (x) dx =
1 ~ y dx.
a MQN
By Property 1 of the line integral we have
~ ydx=- ~ y dx.
MPN NPM
Hence,
S = - ~ y dx- ~ y dx = - ~ g dx. (5)
NPM MQN L
Here, the curve L is traced in a counterclockwise direction.

If part of the boundary L is the segment M M, parallel to the
(M) t
y -axis, then ~ y dx = 0, and equation, (5) holds true in this

(M,)
case as well (Fig. 332).
Similarly, it may be shown that
S= ~ xdy. (6)
L
Adding (5} and (6) term by term and dividing by 2, we get

another formula for computing the area S:
S=i Sxdy-ydx. (7)
L
Example 3. Compute the area of the ellipse

X=a cost, y= b sin t.
Solution. By formula (7) we find
21(
S =-} S[a cos t b cos t- b sin t (- a sin t)] dt =nab.

0
We note that formula (7) and formulas (5) and (6) as well hold
true also for areas whose boundaries are cut by coordinate lines
in more than two points (Fig. 333).
To prove this, we divide the y
given region (Fig. 333) into two
regular regions by the line l*.
y y-Y2{x)
X
0 X
Fig. 332. Fig. 333.
Formula (7) holds for each of these regions. Adding the left and
right sides, we get (on the left) the area of the given region, on
the right, a line integral (with coefficient '/ 1 ) taken over the entire
boundary, since the line integral over the division line l* is taken
twice: in the direct and reverse senses; hence, it is equal to zero.
Green's Formula 679
2. Computing the work of a variable force F on some curved

path L. As was shown at the beginning of Sec. I, the work done
by a force F=X(x,y,z)l+Y(x,y,z)j+Z(x,y,z)k along a line
L = MN is equal to the line integral
(N)
A=~ X(x,y,z)dx+Y(x,y,z)dy+Z(x,y,z)dz.
(M)
Let us consider an instance that

shows how to calculate the work of
the force in concrete cases.
Example 4. Determine the work A of the
force of gravity F when the mass m is tran-
slated from the point M1 (a 1 , b1 , c1 ) to the
point M 2 (a 2 , b2 , c2 ) along an arbitrary path L
(Fig. 334).
Solution. The projections of the force of y
gravity f on the coordinate axes are
X
X=O, Y=O, Z=-mg. Fig. 334.
Hence, the desired work is
(M 2 ) c2
A= ~ X dx+Y dy+Z dz= ~ (-mg) dz=mg(c 1 -c2 ).
(M 1) c,
Consequently, rin this case the line integral is independent of the path of
integration and dependent only on the initial and terminal points. More pre-
cisely, the work of the force of gravity is dependent only on the difference
between the heights of the terminal and initial points of the path.
SEC. 3. GREEN'S FORMULA

Let us establish a connection between a double integral over
some plane region D and the line integral around the boundary L
of this region.
In an xy-plane, let there be given a region D, which is regu-
lar both in fhe direction of the x-axis and the y-axis, bounded
by a closed contour L. Let this region be bounded below by the
curve y=y 1 (x), and above by the curve y=y 2 (X), Y 1 (X)~y 2 (X)
(a~x~b) (Fig. 331).
Together, both these curves represent the closed contour L. Let
there be given, in the region D, continuous functions X (x, y) and
Y (x, y) that have continuous partial derivatives. We consider the
integral
iJX(x,y)d d
Ss
D
iJy X y.
Representing it in the form of an iterated integral, we find
SS:: dx dy = S[ S
b Y2 (X)
~: dy] dx =
b
SX (x, y)
r (X)
dx =
D a Yt (X) a Yt (X)
b
=) [X (x, Y (x))-X (x, y 2 1 (x))] dx. (I)
a
We note that the integral
b
) X (x, y 2 (x)) dx
a
is numerically equal to the line integral
) X (x, y) dx
<MPN)
taken along the curve MPN, whose equations, in parametric

form, are
X=x, y . y 2(x),
where x is a parameter.
Thus
b
) X (x, Ya (x)) dx = ~ X (x, y) dx. (2)

a MPN
Similarly, the Integral

b
) X (x, Y1 (x)) dx
a
is numerically equal to the line integral along the arc MQN:

b
) X (x, y 1 (x)) dx = ) X (x, y) dx. (3)
a (MQNJ
Substituting expressions (2) and (3) Into formula (1), we obtain
SS~: dx dy= S 'X (x, y) dx- S X (x, y) dx. (4)

D MPN MQN
But
) X (x, y) dx = - S X (x, y) dx
MQN NQM
Conditions for a Line Integral Being Independent of the Path 681
(see Sec. 1, Property 1). And so formula (4) may be written thus:
SS~: dxdy S X (x, y)dx+ S X (x, y) dx.
D MPN MQN
But the sum of the line integrals on the right is equal to the line
integral taken along the entire closed curve L in the clockwise
direction. Hence, the last equation can be reduced to the form
SS~: dxdy= S X (x, y) dx. (5)
D L (In the clockwise sense)
If part of the boundary is the segment l 1 parallel to they-axis,

then ~X (x, y) dx=O, and equation (5) holds true in this case as
I,
well.
Analogously, we find
sD
s~: dxdy=- s
L (In the clockwise sense!
Y(x,y)dy. (6)
Subtracting (6) from (5}, we obtain
SS(~:-~~)dx dy=
D
S
L (In the clockwise sense)
X dx+ Y dy.
If the contour is traversed in the counterclockwise sense, then *)
SS(~~- ~:)dxdy= SX dx+ Y dy.

D L
This is Green's formula, named after the English physicist and
mathematician D. Green (1793-1841) **).
We assumed that the region D is regular. But, as in the area
problem (see Sec. 2), it may be shown that this formula holds
true for any region that may be divided into regular regions.
SEC. 4. CONDITIONS FOR A LINE INTEGRAL BEING

INDEPENDENT OF THE PATH OF INTEGRATION
Consider the line integral
(N)
~X dx+ Y dy,
(M)
*) If in a line integral along a closed contour the direction of circulation

is not indicated, it is assumed that it is in the counterclockwise sense. If the
direction of circulation is cleickwise, this must be specified.
**) This formula is a special case of a more general formula discovered by
the Russian mathematician M. V. Ostrogradsky.
taken around some plane curve L connecting the points M and N.

We assume that the functions X (x, y) and Y (x, y) have conti-
nuous partial derivatives in the region D
Q N under consideration. Let us find out under
M what conditions the line integral above is
P independent of the shape of the curve L and
p is dependent only on the position of the ini-
tg 335. tial and terminal points M and N.
Consider two arbitrary curves MPN and
MQN lying in the given region D and connecting the points M
and N (Fig. 335). Let
~ Xdx+Ydy= ~ Xdx+Ydy, (1)
MPN MQN
that is,
~ X dx + Y dy- ~ X dx+ Y dy = 0.
MPN MQN
Then, on the basis of Properties 1 and 2 of line integrals (Sec. 1),

we have
~ Xdx+Ydy+ ~ Xdx+Ydy=O,
MPN NQM
which is a line integral around the closed contour L:
~X dx+ Y dy=O.
L
In this formula, the line integral is taken around the closed con-
tour I:., which is made up of the curves MPN and NQM. This
contour L may obviously be considered arbitrary.
Thus, from the condition that for any two points M and N the
line integral is independent of the shape of the curve connecting
them and is dependent only on the position of these points, it
follows that the line integral along any closed contour is equal io
zero . .
The converse conclusion is also true: if a line integral around
any closed contour is equal to zero, then this line integral is inde-
pendent of the shape of the curve connecting the two points, and
depends only upon the position of these puints. Indeed, equation (1)
follows from equation (2). .
In Example 4 of Sec. 2, the line integral is independent of the
path of integration; in Example 3 the line integral depends on the
path of integration because here the integral around the closed
contour is not equal to zero, but yields an are.a bounded by the
contour in question; in Examples I and 2 the line integrals are

likewise dependent on the path of integration.
The natural question arises: what conditions must the functions
X (x, y) and Y (x, y) satisfy in order that the line integral
~X dx+ Y dy along any closed contour be equal to zero. The
answer is given by the following theorem.
Theorem. At all points of some region D, let the functions X (x, y),
Y (x, y), together with their partial derivatives axa~ y) and aY ~: y)
be continuous. Then, for the line integral along any closed contour L
lying in this region to be zero, that is, for
~ X(x, y)dx+Y(x, y)dy=O, (2)
L
it is necessary and sufficient to fulfil the equation

ax av
au= ax (3)
at all points of the region D.

Proof. Consider an arbitrary closed contour L in a region D
and write Green's formula for it:
5DS(~~- ~~ ) dx dy = SL X dx + Y dy.
If condition (3) is fulfilled, then the double. integral on the left
is identically zero and, hence,
~X dx+ Y dy=O.
L
This proves the sufficiency of condition (3).

Now we prove the necessity of this condition; that is, we prove
that if (2) is fulfilled for any closed curve L in the region D,
then condition (3) is also fulfilled at each point of this region.
Let us assume, on the contrary, that equation (2) is fulfilled,
that is,
~X dx+ Y dy=O,
L
and that condition (3) is not fulfilled;

av _ ax=l= 0
. ax au
at least in one point. For example, at some point P (x0 , y 0 ) let
there be the inequality

iW _ax >O
ox iJy
Since there is a continuous function on the left, it will be positive

and greater than some number {) > 0 at all points of some sufficiently
small region D' containing the point P (x 0 , Y0 ). Take the double
integral of the difference ~Y-
ux
~x over this region. It will have a
uy
positive value. Indeed,
tS (::-::) dx dy> ~S r, dx dy= r, tS dx dy= r,v' > o.

But by Green's formula the left side of the last inequality is
equal to a line integral along the boundary L' of the region D',
which, by assumption, is zero. Hence, the last inequality contra-
dicts condition (2) and therefore the assumption that :~-~: is
different from zero in at least one point is not correct. Whence it
follows that
av _ax= 0
ox iJg
at all points of the given region D.
The theorem is thus proved completely.
In Sec. 9, Ch. XI II, it was proved that fulfillment of the con-
dition
iJY (x, y) iJX (x, y)
ox - iJy
is tantamount to the fact that the expression X dx+ Y dy is. an

exact differential of some function u (x, y), or
X dx + Y dy = du (x, y)
and
au au
X (x, y) =ox , Y (x, y) = iJy
But in this case the vector

y . au i + au
F = X z+ J= ax iJy J
is the gradient of the function u (x, y); the function u (x, y), the
gradient of which is equal to the vector Xi+ Y j, is called the
potential of this vector.
(NJ
We shall prove that in this case the line integral I= ) X dx+ Y dy
. (Ml
along any curve L connecting the points M and N is equal to the
difference between the values of the function u at these points:
(N) (Nl
~ Xdx+Ydy= ~ du(x, y)=u(N)-u(M).

(M) (M)
Proof. If X dx+ Y dy is the exact differential of the function

au au d
u (x, y), then X= ax ; Y = ay an the line integral takes on the
form
(N)
au au
I =
S ax dx + ay
(M)
dy.
To evaluate this integral we write the parametric equations of

the curve L connecting the poinfs M and N:
X= (jl (f), y = 1J? (f).
We shall consider that to, the value of the parameter t = t 0
there corresponds the point M, and to t = T, the point N. Then
the line integral reduces to the following definite i~tegral;
.-:
I =S [auax ax+
T
at au
au ay]
at dt.
t.
The expression in the brackets is a function of t, and this func-

tion is the total derivative of the function u [cp (t), "B (t)] with
respect to t. Therefore
T
I= S~; dt = u [cp (t), 'IJ (t)] I[.= u [cp (t), 'I' (t)]
to
-u [cp (t 0 ), 'I' {t 0 )] = U (N)-'-U (M).
As we see, the line integral of an exact differential is independent

of the shape of the curve along which the integration is performed.
We have a similar assertion for a line integral over a space
curve (see below, Sec. 7).
Note. It is sometimes necessary to consider line integrals of

some function X (x, y) along the length of an arc L:
n
~X (x,
L
y) ds =lim
il.Sj-Oi=l
L. X (xi, Yi) Asi, (4)
where ds is the differential of the arc. Such integrals are evaluat-

ed in s1milar fashion to the line integrals considered above. Let
the curve L be represented by the parametric equations
X= cp (t), y = '\j> (!),
where cp (t), 'lj> (t}, cp' (t}, 'lj>' (t) are continuous functions of t.
Let a and P be. values of the parameter t corresponding to the
origin and terminus of the arc L.
Since
+
ds = V cp' (t)" 'lj>' (t) dt,
we get a formula for evaluating integral (4)~
13
~X (x, y) ds = ~X [cp (t), 'iJ (t)] V cp' (t) + 'lj>' (f)" dt.
L a
We can consider the line integral along the arc of the space
curve x=cp(t), y='ljl(t), z=x,(t):
13
~X (x, y, z)ds= ~X [cp(t), 'lj>(t), x.(t)]Vcp' (W+~' (W+x.'(t)1 dt.
L a
By the use of line integrals along an arc we can determine, for

example, the coordinates of the centre of gravity of lines.
Reasoning as in Sec. 8, Ch. XII, we obtain a formula for
evaluating the coordinates of the centre of gravity of a space
curve.
~ x ds ~ y ds
/ L L ,
Xc=v. Yc=-
1
-, (5)
J ds
L L
Example. Find the coordinates of the centre of. gravity of one turn of the
helix
X= a cost, y =a sin t, z = bt (0 ....;; t < 2n),
... 'l
ii its. linear density is consta~t.

Surface Integrals 687
Solution. Applying formula (5), we find

2lt
~ a cos t V a sin t + a cos t + b' dt

Xc ~0~---------------------
2lt
~ V a sin t + a cos t + b' dt

2
0
Zlt
~a cost Va'+b'dt
o a V a+ b 2 0
--m - -0
-2n Va 2 +b 2 -
~ Va 2 +b"dt
0
Similarly, Yc=O,
n
~ bt l.ra sin t + a cos t + b' dt
0
Zc = ;;_------2-n-Y:-;.=a==+=b=.-----
Thus, the coordinates of the centre of gravity of one turn of the helix are
Xc=O, Yc=O. Zc=nb.
SEC. 5. SURFACE INTEGRALS
Let a region V be given in an xyz-coordinate system. Let a

surface a bounded by a certain space line A. be given in V.
With respect to the surface a we shall assume that at each
point P of it the positive direction of the normal is determined
by the unit vector n (P), the direction cosines of which are con-
tinuous functions of the coordinates of the surface points.
At each point of the surface let there be defined a vector;
F~X(x, y. z)l+Y(x, y, z)j+Z(x, y, z)k,
where X, Y, Z are continuous functions of the coordinates.

Divide the surface in some way into subregions 11cri. In each
subregion take an arbitrary point P1 and consider the sum
~(F(P 1 )n{P 1 ))11ai, (1)
I
where F (Pi) is the value of the vector F at the point Pi of the
subregion 11a1; n (Pi) is the unit normal vector at this point and
Fn is the scalar product of these vectors.
The limit of the sum (1) extended over all subregions 11a1 as
the diameters of all such subregions approach zero is called tile
688 - Line Integrals and Sur face Integrals
surface integral and is denoted by the symbol

~ ~ Fnda.
(J
Thus, by definition*)
lim
dlam 6.01-+o
L F n fla = ~ ~ Fn da.
1 1 1
0
(2)
Each term of the sum (1)
F 1n1flal = F1flai cos (ni, F 1) (3)
may be interpreted mechanically as follows: this product is equal
to the volume of ~cylinder with base fla 1 and altitude F1 cos (ni, F1).
If the vector F IS the rate of flow of a liquid through the sur-
face a, then the product (3) is equal to the quantity of liquid
flowing through the subregion !J.ai in unit time in the direction
of the vector n1 (Fig. 336).
Fig. 336. Fig. 337.
The expression ~ ~ Fn da yields the total quantity of liquid

(J
flowing in unit time through the surface a in the positive direc-

tion if by the vector F we assume the flow-rate vector of the
liquid at the given point. Therefore, the surface integral (2) is
called the flux of the vector field F through the surface a.
From the definition of a surface integral it follows that if the
surface a is divided into the parts <T 1 , <T 2 , , ak, then
~~ Fnda= ~~ Fnda+
0 ~
HFnda + .. + HFnda.
~ ~
*)If the surface a is such that at each point of it there exists a tangent
plane that constantly varies as the point P is translated over the surface,
and if the vector function F is continuous on this surface, then this limit
exists (we accept this existence theorem of a surface integral without proof).
Evaluating Surface Integrals 689
Let us express the unit vectQr n in terms of its projections on

the coordinate axes:
n=cos(n, x)i+cos(n, y)J+cos(n, z)k.
Substituting into the integral (2) the expressions of the vectors
F and n in terms of their projections, we get
~ ~ Fndo =
(J
H[X cos (n,
(J
x) + Y cos (n, y) + Z cos (n, z)] do. (2')
The product ilGcos (n, z) is the projection of subregion Llo on

the xy-plane (Fig. 337); an analogous assertion holds true for the
following products as well:
Llo cos (n, x) = Lloyz, Llo cos (n, y) = Lloxz. Llo cos (n, z) = Lloxy. (4)
where ilOyz 1 ilOxz 1 ilOxy are the projections Of the subregion ilu
on the appropriate coordinate planes.
On this basis, integral (2') can also be written in the form
~~Fndo=~~ [Xcos(n, x)+Ycos(n, y)+Zcos(n, z)]do=

C1 C1
=~~X dy dz+ Y dzdx+Z dxdy. (2")

a
SEC. 6. EVALUATING SURFACE INTEGRALS
Computing the integral over a curved surface reduces to eva-

luating a double integral over a plane region.
To illustrate, the following is a method of computing the
integral
~ ~ Z cos(n, z) do.
C1
Let the surf ace o be such that any straight line parallel to the
z-axis cuts it in one point. Then the equation of the surface
may be written in the form
z=f(x, y).
Denoting by D the projection of the surface o on the xy-plane,
we get (by the definition of a surface integral)
n
~ ~ Z (x, y, z) cos (n, z) do=
0
lim L Z (xi,
dlam tJ.<J;-+ o l=l
Yi zi) cos (ni, z) Lloi.
Noting, further, the last of formulas (4), Sec. 5, we obtain

n
~~Zcos(n, z)da= lim ~Z(xi, Yi f(xi, yJ)(!l.axy)i=
C1 dlam lla.,y-o i=t
n
= lim ~Z(xi, Yi f(xi, Yi))l!l.ax.uli;

dlam lla- o i=t
the last expression is the integral sum for a double integral of

the .function Z (x, y, f(x, y)) over the region D. Therefore, .
~ ~ Z cos (n, z) da = + ~ ~ Z (~. y, f (x, y)) dx dy.
C1 D
The plus sign in front of the double integral is taken if cos (n, z);;;;;. 0,
the minus sign, if cos (n, z)...;;;;; 0. .
If the surf ace a does not satisfy the condition indicated at the
beginning of this section, then it is divided into parts that satisfy
this condition, and the integral is computed over each part
separately.
The following integrals are computed in similar fashion:
~~X cos (n, x) da; ~ ~ Y cos (n, y) da.
The foregoing proof justifies the notation of a surface integral

in the form of (2"), Sec. 5.
Here, the right side of (2") may be regarded as the sum of
double integrals over the appropriate projections of the region a
and the signs of these double integrals (or, otherwise stated,
the signs of the products dy dz, dx dz, dx dy) are taken in
accord with the foregoing rule.
Example 1. Let a closed surface a be such that any straight line parallel
to the z-axis cuts it in no more than two points.
~ ~ z cos (n, z) da,
C1
We shall call the outer normal the positive direction of the normal.
In this case, the surface may be divided into two parts: lower and upper;
their equations are, respectively,
z=ft<x, y) and z=fs(x, y).
Denote by D the projection a on the xy-plane (Fig. 338); then
~~
a
z cos (it, z) da = HD
fa (x, y) dx dy-:- ~~
D
/ 1 (x, y) dx dy.
Evaluating Surface Integrals 691
The minus sign in the second integral is taken because in a surface integ-
ral the sign of dx dy on a surface z = / 1 (x, y) must be taken negative, since
for it cos (n, z) Is negative.
~i-~
zr~n
.,..,------
Fig. 338. Fig. 339.
But the difference between the integrals on the right in the last .formula
yields a volume bounded by the surface a. This means that the volume of
the solid bounded by the closed surface a is equal to the following integral
over the surface
v = ~ ~ z cos (n, z) dcr.
a
Example 2. A positive electric charge e placed at the coordinate origin
creates a vector field such that at each point of space ihe vector F is defined.
by the Coulomb law as
e
F=k 2 r,
r
where r is the distance of the given point from the origin, r is the unit
vector directed along the radius vector of the gtven point (Fig. 339); and k
is a constant coefficient.
Determine the vector-field flux through a sphere of radius R with centre
at the origin of coordinates.
Solution. Taking into account that r=R=const, we will have
SSa k : 2 rn da= ~~ SS rn da.

a
But the last integral is equal to the area of the surface a. Indeed, by the
definition of an integral (noting that rn = 1), we obtain
~~ rnda= lim ~rknkMk,;., lim ~Llak=a.

a liak --. o o
lias --.
Hence, the flux' is;~ a=~~ 4nR 2 =4:nke.

SEC. 7. STOKES' FORMULA

Let there be a surface a such that any straight line parallel
to the z-axis cuts it in one point. Denote by 'A. the boundary of
the surface cr. Take the positive direction of the normal n so
that it forms an acute angle with the positive z-axis (Fig. 340).
Let the equation of the surface be Z=f(x, y). The direction
cosines <'~f the normal are expressed by the for'mulas (see Sec. 6,
Ch. I X):
cos (n,
x) = , /1 (at )2 (at )2;
V + ax + ay
at
cos (n,
Y)= (;y
Y + ax + (~)';
oy j
( 1)
Y+(~)'+(U)"
cos (n,
Z)=
Fig. 340. We shall assume that the surface a lies

entirely in sonie region V. Let there be
a function X (x, y, z) given in .V that is continuous together with
first-order partial derivatives. Consider the line integral along
the curve 'A.:
i X (x,
A
y, z)dx.
On the line 'A., z = f (x, y), where x, y are the coordinates of

the points of the line L, which is a projection of the line A. on
the xy-plane (Fig. 340). Thus, we can write the equation
) X (x, y, z) dx =
'A L
i X (x, y, f (x, y)) dx. (2)
The last integral is a line integral along L. Transform this integ-

ral by Green's formula, putting
X (x, y, f(x, _y)) =X (x, y), 0= Y (x, y).
Substituting into Green's formula the expressions of X and Y';
we obtain
.-tS ax (x, Ya/ (x, y)) dx dy = J X (x, y, f (x, y)) dx, (3)
Stokes' Formula 693
where the region D is bounded by the line L. On the basis of

the derivative of the composite function X (x, y, f (x, y)), where
y enters both directly and in terms of the function z = f (x, y),
we find
ax (x, y, f (x, y)) -ax (x, y, z) +ax (x, y, z) of (x, y) (4)
ay - au az oy
Substituting expression (4) into the left side of (3), we obtain
-ss [ax (x,~y, z) +ax (x,~y, z). df (x,~ y)] dx dy =
D
=~X (x, y, f (x, y)) dx.

L
Taking into account (2), the last equation may be rewritten as

ax dx dy- 55 ax
~X
of dxdy.
(x, y, z) dx=-
Ss
D
oy
D
az oy (5)
The last two integrals can be transformed into surface integrals.

Indeed, from formula (2"), Sec. 5, it follows that if we have
some function A (x, y, z), the following equation is true:
H
a
A (x, y, z) cos (n, z) da= ~~A dxdy.
D
On the basis of this equation, the integrals on the right side
of (5) are transformed as follows:
SSD ~: dx dy= SSa ~:cos(~. z) da, }
ax iJf SS ax at (6) ,
SsD oydxdy= a azaycos(n, z)da.
i)z
Transform the last integral using formulas (1) of this section:

dividing the second of these equations by the third termwise,
we find
cos(n, y) of
cos (n, z) = - oy
or
of
oy cos (n, z) = -cos (n, y).
Hence,
Ss ozax oyof dx dy= - 55a ax cos (n, y) da.

D
i}z (7)
Substituting expressions (6) and (7) into equation (5), we get
5 X(x, y, z)dx=-s5::cos(n, z)dcr+Ss~;cos(n, y)dcr. (8)

11. 0 0
The direction of circulation of the contour A. must agree with

the chosen direction of the positive normal n. Namely, if an
observer looks from the end of the normal, he sees the circula-
tion along the curve A. as being counterclockwise.
Formula (8) holds true for any surface if this surface can be
divided into parts whose equations have the form z = f (x, y).
Similarly, we can write the formulas
SY (x, y, z) dy =55 [-;;cos (n, x) +.~~cos (n, z) Jdcr, (8')

11. 0
~ Z (x, y, z) dz = 5J [-;;cos (n, y) +;!cos (n, x)Jdcr. (8")
Adding the left and right sides of (8), (8'), and (8"), we get
the formula
~X dx+ Y dy+Zdz= SJ [(:~-::)cos (n, z)+
+(;:-~~)cos(n, x)+(::-:;)cos(n, y)J dcr. (9)
This formula is called Stokes' formula after the English physicist

and mathematician D. Stokes (1819-1903). It establishes a rela-
tionship between the integral over the surface cr and the line
integral along the boundary A. of this surface, the circulation
about the curve A. being performed according to the same rule as
that given earlier.
The vector B, defined by the projections
B _az_av. B _ ax_ az . _ av _ax
8 =- OX
X - {)y OZ ' y- {)z ax ' ay '
is called the curl or rotation of the vector function F =Xi+ Yj+ Zk
and is denoted by the symbol rot F.
Thus, in vector notation, formula (9) will have the form
~ F ds = ~ ~ n rot Fda, (9')

l. 0
and Stokes' . theorem is formulated thus:

Stokes' Formula 695
The circulation of a vector around the contour of some surface

is equal to the flux of the curl through this surface.
Note. If the surface cr is a piece of plane parallel to the
xy-plane, then f).z = 0, and we get Green's formula as a special
case of Stokes' formula.
From formula (9) it follows that if
ay _ax= 0 az_aY _ 0 ax_dz= 0 ( 10)
ax ay 1
a!J OZ - 1
OZ dX 1
then the line integral along any closed space curve A. is zero:
~ X dx + Y dy + Z dz = 0. ( 11)
A.
Whence it follows 'that the line integral is independent of the

shape of the curve of integration.
As in the case of a plane curve, it may be shown that the
indicated conditions are not only sufficient but also necessary.
In the fulfillment of these conditions, the expression under the
integral sign is an exact differential of some function u (x, y, z):
X dx+ Y dy+Z dz= du (x, y, z)
and, consequently,
(N) (N)
~ Xdx+Ydy+Zdz= ~ du=u(N)-u(M).
(M) (M)
This is proved exactly like the corresponding formula for a

function of two variables (see Sec. 4).
Example 1. Write the basic equations of the dynamics of a material
point:
Here, m. is the mass of the point, X, Y, Z are the projections of a force,

acting on the point, onto the coordinate axes; Vx=~~ Vy=~~ Vz=~~ are
the projections of velocity v on the axes. .
Multiply the left and right sides of these equations by the expressions
Vxdt=d~, vvdt=dy, vzdt=dz.
Adding the given equations term by term, we obtain
m (vx dvx+vy dvv +vz dvz)=X dx+ Y dy+Z dz;
1
m d (;:,!+ v~+ v~) =X dx+ Y dy +Z dz,
2
Since v!+v~+v!=v', we can write
d (; mv 2 ) =X:dx+ Y dy+Z dz.
Take the integral along the trajector-y connecting the points M 1 and M2 :
(M,)
; mv~- ~ mv~= S'X dx+Y dy+Z dz,

(M,)
v2 are the velocities at the points M 1 and M 1
This last equation expresses the theorem
of live forces: the increase in kinetic energy
when passing from one point to another is
equal to the work of .the force acting on the
mass m.
Example 2. Determine the work of the force
of Newtonian attraction to a fixed centre of
mass m in the translation of unit mass from
M1 (a1 , b1 , c1) to M2 (a2 , b2 , c2).
y Solution. Let the origin be in the fixed
centre of attraction. Denote by r the radius
vector of the point M (Fig. 341) corresponding
Fig. 341. to an arbitrary position of unit mass, and by
t.ll the unit vector directed along the vector r.
Then F=- k~t r0 , where k is the constant of gravitation. The projections of
r
the force F on the coordinate axes will be
X=-km _!_~ Y= -km..!..L

,z
r: r' r: '
I z
Z=-km--.
r' t?
Then the work of the force F over the path M 1 M 2 is
A--km
.
s
(M,)
,.
xdx+ydy+zdz=
(M,)
!M 2 ) (M,)
=- km S ',~' = km S d ( : )
(M 1) (M 1)
+y
(since r 2 =x 2 2
+z 2, rdr=xdx+ydy+zdz). If we denote by r1 and rz
the lengths of the radius vectors of the points M 1 and M 1 , then
A =km (__!,.-..!..)
r r 1
.
Thus, here again the line integral does not depend on the shape of the
curve of integration, but only on the position of the initial and terminal
Points. The function u = km
t?
is called the potential of the gravitational field
Ostrogradsky's Formula 697
generated by the mass m. In the given case,

au au au
X= ox ' y = iJy ' Z = iJz '
A =u (M2)-u (M 1).
That is, the work done in moving unit mass is equal to the difference be-
tween the values of the potential at the terminal and initial points.
SEC. 8. OSTROGRADSKY'S FORMULA
Let there be given, in space, a regular three-dimensional region

V bounded by a C'losed surface o and projected on an xy-plane
into a regular two-dimensional region D. We shall assume that
the surface o may be divided into three parts 0 1 , 0 2 , and o 3
such that the equations of the first two have the form
z = { 1 (x, y) and z = { 2 (x, y),
where { 1 (x, y) and {2 (x, y) are functions continuous is the region
D and the third part o3 is a cylindrical surface with generator
parallel to the z-axis.
I= SSS az (xat z) dxdydz.
v
First perform the integration with respect to z:
fo (X, Y) )
I= SS( S ~; dz dx dy =
D f1 (x, V)
= ~ ~ Z (x, y, f. (x, y)) dx dy- ~ ~ Z (x, y, f1 (x, y)) dx dy. (1)
D D
On the normal to the surface, choose a definite direction, name-
ly that which coincides with the direction of the outer normal
to the surface o. Then cos (n, z) will be positive on the surface
a. and negative on the surface o 1 ; on the surface 0 3 it will be
zero.
The double integrals on the right of (1) are equal to the cor-
responding surface integrals:
HZ (x,
D
y, f 2 (x, y)) dx dy = ~ ~ Z (x,
~
y, z) cos (n, z) do, (2')
~ ~ Z (x, y, { 1 (x, y)) dx dy = ~ ~ Z (x, y, z) (-cos (n, z)) do.

D ~
In the last integral we wrote [-cos (n, z)] because the elements
of surface cr, and cr! and the element of area l!..s of the region D
are connected by the relation l!..s =I!.. a [-cos (n, z)], since the
angle (n, z) is obtuse.
Thus,
~~Z(x,y, f,(x,y))dxdy=-~~Z(x, y, f,(x, y))cos(n, z)dcr.(2")

o a,
Substituting (2') and (2") into (I), we obtain
55f'~z (~/' z) dxdydz=.

v
= ~ ~ Z (x, y, z) cos (n, z) dcr+ ~ ~ Z (x, y, z) cos (n, z) dcr.

a2 a.
For the sake of convenience in subsequent formulas, we sh:;LII rewrite

the last equation as foilows [adding ~ ~ Z (x, y, z) cos (n, z) dcr = 0,
a, .
since the equation cos (n, z) = 0 is fulfilled on the surface cr,] :
555 az (xaz y. z) dxdydz =.

v
= ~ ~ Z cos (n, z)rlcr+ ~ ~ Z cos (n, z) dcr+ ~ ~ Z cos (n, z) dcr.
~ ~ ~
But the sum of integrals on the right of this equation is an in

tegral over the entire closed surface cr; therefore,
555~;dxdydz= 55Z(x, y, z)cos(n, z)dcr.

v a
Analogously, we can obtain the relations
555:~dxdydz= 55Y(x, y. z)cos(n, y)dcr,

v a
555a:X dxdydz= 55 J((x, y. i).cos(~. x)dq~: '

v a '
Ostrogradsky's Formula 699
Adding together the last three equations term by term, we get

Ostrogradsky's formula *1:
Sss( ax
iJx
+ iJy
av + iJz
az)dx dy dz =
v
= ~~ (X cos (n, x) + Y cos (n, y) + Z cos (n, z)) dcr. (2)
(1
.
Th e expression ax+aY+iJZ
ax iJy iJz .IS ca II e d th e d.tvergence o f th e vec-
tor (or the divergence of the vector function):
F=Xi+Yj+Zk
and is denoted by the symbol div F:
.
d tv F- ax+av+az
-ax iJy iJz
We note that this formula holds good for any region which
may be divided into subregions that satisfy the conditions indi-
cated at the beginning of this section.
Let us examine a hydromechanical interpretation of this
formula.
+
Let the vector F= Xi+ Yj Zk be the velocity vector of a
liquid flowing through the region V. Then the surface integral in
formula (2) is an integral of the projection of the vector F on
the outer normal n; it yields the quantity of liquid flowing out
of the region V through the surface cr in unit time (or flowing
into V if this integral is negative). This quantity is expressed
in terms of the triple integral of div F.
If div F=-0, then the double integral over any closed surface
is equal to zero, that is, the quantity of liquid flowing out of
(or into) something through any closed surface cr will be zero
(no sources). More precisely, the quantity of liquid flowing into a
region is equal to the quantity of liquid flowing out of this
region.
In vector notation, Ostrogradsky's formula has the form
~~~divFdv= ~~Fnds (l ')

v a
*) This formula (sometimes called the Ostrogradsky-Gauss formula) was

discovered by the noted Russian mathematician M. V. Ostrogradsky (1801-1861)
and published in 1828 in an article entitled "A Note on the Theory of Heat".
and is read: the integral of the divergence of a vector field F

extended over some volume is equal to the vector flux through the
surface bounding the given volume.
SEC. 9. THE HAMILTONIAN OPERATOR AND CERTAIN

APPLICATIONS OF IT
Suppose we have a function u = u (x, y, z). At each point of

the region in which the function u (x, y, z) is defined and diffe-
rentiable, the following gradient is determined:
.au +Jay
gra d U=lax
.au + k az.
au (1)
The gradient of the function u (x, y,. z) is sometimes denoted as

follows:
- . au+ . au+ k au (2)
vu-'ax Jay az.
The symbol v is read "del".
I) It is convenient to write equation (2) symbolically as
VU=--=(i :x+j :y+k :z)u (2')
and to consider the symbol

-i!_+
v- ax 1i.+ki_
ay az (3)
as a "symbolic vector". This symbolic vector is called the Hamil-

tonian operator or del operator (v-operator). From formulas (2)
and (2') it follows that "multiplication" of the symbolic vector v
by the scalar function u gives the gradient of this function:
vu =grad u. (4)
2) We can form the scalar product of the symbolic vector v by
+ +
the vector F= iX jY kZ:
VF = ( i :x + j :y + k ~) (iX + jY + kZ) =
=!!... X + !._ Y + i. z =ax+ av + az = d iv F
ax ay az ax ay az
(see Sec. 8). Thus,
vF=- div F. (5)
The Hamiltonian Operator and Certain Applications of It 701
3) Form the vector product of the symbolic vector v by the

vector F = iX + jY + kZ:
V xF= (i :"' +j ~+ k :z) x(iX +jY +kZ) =

ijk !..~ ~!._ !...~
a a a . ay az . ax az ax ay
ax ay az = 1 Y Z -1 X Z +k XY =
X Y Z
=i (az_av)- .(az_i;X)+k (<'JY _IJX) =

ay az J ax {)z ax {}y
= i (az
ay _ av)
az +j (ax_ ax +k (ay
az az) ax _ax)=
ay rotF
(see Sec. 7). Thus,

vxF=rotF. (6)
From the foregoing it follows that vector operations may be
greatly condensed by the use of the symbolic vector v. Let us
consider several more formulas.
+
4) The vector field F (x, y, z) = lX jY + kZ is called a poten-
tial vector field if the vector F is the gradient of some scalar
function u (x, y, z):
F=grad u
or
-.~+ .au+kau
F -'ax Jay az.
In this case the projections of the vector F will be
au
X=ax' Y=:;, Z=:~.
From these equations it follows (see Ch. VIII, Sec. 12) that
ax av av az ax az
or
ax_av _ 0 av _az= 0 ax~az= 0
ay ax - ' {Jz ay ' az ax
Hence, for the vector F under consideration,
rotF=O.
Thus, we get
rot (grad u) = 0. (7)
Applying the del operator v, we can write (7) as follows [on the
basis of (4) and (5)]:
(V X Vtt) = 0. (7')
Taking advantage of the property that for multiplication of a
vector product by a scalar it is sufficient to multiply this scalar
by one of the factors, we write
(V XV) u = 0. (7")
Here, the del operator again has the properties of an ordinary
vector; the vector product of a vector into itself is zero.
The vector field F (x, y, z), for which rot F = 0, is called irro-
tational. From (7) it follows that every potential field is irrota-
tional.
The converse also holds: if some vector field F is irrotational,
then it is potential. The truth of this statement follows from
reasoning given at the end of Sec. 7.
5) A vector field F (x, y, z) for which
div F=O,
that is, a vector field in which there are no sources (see Sec. 8)
is called solenoidal. We shall prove that
div (rot F)= 0 (8)
or that the rotational field is free of sources.
Indeed, if F=iX+iY+kZ, then
rot F = i(az az + j (ax_
ay _ av) az az)
ax +k (av
ax _ax)
ay
and therefore
div(rotF)=~(az_av)+~ o.
ax ay az ay (ax_az)+2..(av
.Oz ax az ax _ax)
ay
Using the del operator, we can write equation (8) as
v(vxF)=O. (8')
The left side of this equation may be regarded as a vector-scalar
(mixed) product of three vectors: v, v. F, of which two are the
same. This product is obviously equal to zero .
. 6) Let there be a scalar field u = u (x, y, z). Determine the
gradient field:
.au+ .au
gra d u ='ax 1 ay + kauaz
The Hamiltonian Operator and Certain Applications of It 703
Then find
(gra d u)= ax
d tv a (au)
ax + aya (auay ) + aza ( au)
az
or
. au au au
dtv (grad u) = ax+ ay+ az. (9)
The right side of this expression is called the Laplacian ope-

rator of the function u and is denoted by
au au au
tlu= ax+ ay+ az. <10>
Hence, (9) may be written as
div (grad u) = tlu. (11)
Using the del operator v we can write (11) as
<vvu)= !1u. ( 11 ')
We note that the equation
au au au
ox+ ay+ az= 0 (12)
or
(12')
is called Laplace's equation. The function that satisfies the Lap-
lace equation is called a harmonic function.
Exercises on Chapter XV
Compute the following line integrals:
1.5 y'dx+2xydy over the circumference x=acost, y=aslnt.
Ans. 0.
2. ~ ydx-xdy over an arc of the ellipse x=acost, y=bsint.
Ans. -21tab.
3. S(x~y dx- x'~ydy) over a circle with centre at the origin.
Ans. 0.
4. S(y d~+x,dy) over a segment of the. straight line y=x from x= I to
x +u .
x=2. Ans. In 2.
5. ~ yzdx+xzdy+xydz over an arc of the helix x=acost, y=slnt,
z = kt as t varies from 0 to 2n. Ans. 0.
6. ~xdy-ydx over an arc of the hypocycloid x=acos 1 t, y=asln't.

Ans. ~ na (the double area of the hypocycloid).
7. s
3 at
.
x dy- y dx over the loop of the folium of Descartes x = - -
3
1 +t' '
3~
y= I +t'' Ans. 2 a (the double area of the region bounded by the indicated
loop).
8. ~ x dy-y dx over the curve x=a(t-sin t), y=a(l-cos t)(O..;;; t E;;;2n).
Ans. -6na 2 (the double area of the region bounded by one arc of a cycloid
and the x-axis).
Prove that!
9. grad (ccp) =C grad cp where c is a constant.
10. grad (c1cp+c.1Jl)=c1grad cp+c.grad 1Jl where cis a constant.
11. grad (cp'IJl) = cp grad 1Jl +1Jl grad cp,
12. Find grad r, grad r, grad ..!..,
(i
grad f (r) where fi = Y x + y 2+ z. Ans.
r , 2r, - rf'()r
-
r 1r , r-.
r
13. Prove that div(A+B)=divA+divB.
14. Compute divr, where r=xl+uJ+zk.
Ans. 3.
15. Compute div (A cp), where A is a vector function and cp is a scalar
function. Ans. cpdivA+(gradcpA).
16. Compute div (r-c), where cis a constant vector. Ans. (cr)
r
17. Compute div 8 (rA). Ans. AB.
Prove that:
18. rot (c1 A1 + c2 A2) =C1 rot A1 +c1 rot A1 where C1 and c1 are constants.
19. rot(Ac)=grad Axe where o is a constant vector.
20. rot rotA=graddivA-VA.
21. Axrotcp=rot(cpA).
Surface Integrals
22. Prove that~~ cos(n,z)da=O if a is a closed surface and n Is a nor-

mal to it.
23. Find the 1moment of inertia of the surface of a segment ol a sphere
with equation x"+y+z 1 =R" cut off by the plane z=H relative to the
. 2nR
z-ax1s. Ans.-- (2R'-3R 1 H +H').
3
24. Find the moment of inertia of the surface of the paraboloid of revo-
lution x 2 +y 1 =2cz cut off by the plane z=o relative to the z-axis. Ans.
55+9Y3 c~]
[ 65 .
25. Compute the coordinates of the centre of gravity of a part of the sur-
R2 2
face of the cone x"+u"=H,z cut offbytheplanez=H. Ans. 0, 0,3H.
Exercises on Chapter XV 705
26. Compute the coordinates of the centre of gravity of a segment of the surface
ofthespherex 2 +y 2 +z 2 =R 2 cutoff by the plane z=H. Ans. (o. 0, RtH).
27. Find ~ ~ [x cos (nx) + y cos (ny) +z cos (nz)] da, where a is a
0
closed surface. Ans. 3V, where V is the volume of the solid bounded by the
surface a.
28. Find~~ zdxdy where Sis the external side of a sphere x"+u"+z"=
s
= R 2 Ans. ~ nr'.
29. Find~~ x 2 dydz+y 2 dzdx+z 1 dxdy where SIs the external side of
s
the surface of a sphere x+u+z 2 =R 1 Ans. nR'.
30. Find ~ ~ Yx 2 +y 1 ds where S is the lateral surface of a cone
s
x y 2 z 2na Y a+ b 2
a+az-p=O, O.e;;;;z.e;;;;b. Ans. 3
31. Using the Stokes formula, transform the integral ~ y dx +z dy +x dz.
L
Ans. - ~ ~ (cos a+ cos ~+cosy) ds.
s
Find the line integrals, applying the Stokes formula and directly: 32.
~ (y+z)dx+<z+x)dy+(x+y)dz where L is the circle x 2 +y"+z=
L .
=a, x+y+z=O. Ans. 0. ~3. ~ xyd.t+dy+zdz where L is the circle
L
nR'
x+y 1 =R 2 , z=O. Ans. -
8 .
Applying the Ostrogradsky formula, transform the surface Integrals into
volume integrals: 34.~ ~ (x cos a+ y cos~+ z cosy) ds. Ans. ~ S~ 3 dx dy dz.
s v
35. s~ (x +y."+z")(dydz+dxdz+dxdy). Arts. ss~ (x+y+z)dxdydz.
2
s v
36. SS xydxdy+yzdydz+zxdzdx. Ans. 0. 37. SS ~~dydz+~ dxdz+
+~~xdy. Ans. SSS(~;+~+::~)dxdydz. s
v
Using the Ostrogradsl<y formula compute the following integrals: 38.
s~ (xcosa+ycos~+zcosy)ds where s is the surface of the ellipsoid
s
23 - 3388
x y z'
iii+b'+?=l. Ans. 4:rtabc. 39. 55 (x 3 cosa+y 1 cos~+z'cosy)ds wher<! S
s
is the surface of the sphere x+y 2 +z'=R 2 Ans. ~:rtR 5 40. 55 x'dydz+
s
y' z2 x
+y'dzdx+z 2 dxdy where Sis the surface of the cone 2a +--.--=0
b c
(O~.;;;;;b). Ans. :rca;b. 41. 55 xdydz+ydxdz+zdxdy where S is the
s
surface of the cylinder x 2 +y'=a', -H.;;;;;z,;;;;;;.H. Ans. 3:rta2H.
42. Prove the identity 55(~;+:~) dxdy= 5~ds, where Cis a con-
D C
tour bounding the region D, and ~ is the directional derivative of the outer
normal.
Solution.
\5(~;+~~) dx dy=5-Ydx+Xdy=5 [~Ycos(s, x)+Xsin(s, x)]ds,
~ c c
where (s, x) is the angle between the tangent line to the contour C and the
x-axis. If we denote by (n, x) the angle between the normal and the x-axis,
then sin(s, x)=cos(n, x), cos(s, x)=-sin(n, x). Hence,
55(~;+~~) dx dy= 5[Xcos(n, x)+Ysin(n, x)}ds.

D C
au
Setting X=~, y = ay. we get
5J(~;+::~) dx dy= ~[~ cos(n, x)+~ sin(n, x)]ds

or
55[~;+~] dx dy= 5~ ds.
D C
. au au
The expression ax+ oy' is called the Laplacian operator.
43. Prove the identity (called Green's formula)
555 (vh-uAv) dx dy dz= 55( v ~-u ~)da,

V a
where u and v are continuous functions with continuous derivatives to the
second order in the region D.
The symbols Au and Av denote
au au a1u azv av a'v
Au= ax + ayz + az' A0 = ax + ay + az'
These expressions are called Laplacian operators in space.
Solution. In the formula

a_x+av+az)
iJx ay az dx dy dz = j('('
j [X cos (n, x) + Y cos (n, y) + Z cos (n, z)] dcr
Sss(
V cr
we put
X=vu:-uv:,
Y=vu~-uv~,
Z=vu~-uv~.
Then
ax av az " " " " "
ax+ ay + az =v (uxx+uuu+uzz)-u(vxx+vuu+vzz>=vL\u-uL\v,
X cos (n, x) + Y cos (n, y) +Z cos (n, z) =
= v (u~ cos nx + u~ cos ny + u~ cos nz) -u (v; cos nx + v~ cos ny + v~ cos nz) =
au av
=van-uaii.
Hence,
SJS (vL\u-uL\ v) dx dy dz= ~S( v ~ -u g~)da.
44. Prove the identity
SSS L\udxdydz=SS~~da,
V cr
azu asu azu .
where L\u= axz+ ayz+ aza (Laplactan).
Solution. In Green's formula, which was derived in the preceding section,
put v=l. Then L\v=O, and we get the desired identity.
45. If u (x, y, z) is a harmonic function in some region, that is, a func-
tion which at every point of this region satisfies the Laplace equation
asu iJ2u a~u
axz + dyz + azz = 0,
then
where cr is a closed surface.

Solution. This follows directly from the formula of Problem 44.
46. Let u (x, y, z) be a harmonic function in some region V and let there
be, in V, a sphere awith centre at the point M (x1 , y 1 , z1) and with radius
R. Prove that
u (xl, YI> zl) = 4n.IRz ssu da,
a
23*
708 Line ftttegt als and Surface Integrals
Solution. Consider the region Q bunded by two spheres a, of radius R a

and Q (Q < R) with centres at the point M (x 1 , y 1 , z 1). Apply-Green's formula
(found in Problem 43) to this region, taking for u the above-indicated function,
and for the function v,
1 1
v = r= Y(x-x.)2+(y- y.)a +(z-z.)2.
8 2u 8u
2
By direct differentiation and substitution we are convinced that axz + ayz +
a2u
+ azz =0. Consequently,
a +a
or
On the surfaces a and~ the quantity }. is constant ( ~ and i) and so can

be taken outside the integral sign. By virtue. of the result obtained in Prob-
lem 45, we have
.R155 anau da=O;
(J
Q 55 au
_ an
1 da=O.
(J
Hence,
but
Therefore,
or
~z 55 u da = ~ 2 55 u da. (I)
(J 0
Apply the theorem of the mean to the integral on the right;
_!___55 u da = u (~. '11. ~) 55 da

(2)
Q2 Q2
a -a
where u (~. '11 ~) is a point on the surface of a sphere of radius Q with
centre at the point M (x 1 , Yu Zt).
We make Q approach zero; then u (~. fJ, ~) -+ u (x 1 , y1 , z1 );
:2 ss da = 4~~z = 4n.
0
Hence, as Q-+ 0 we gef
1
Q2 iJ u da -+ u (x1 , y 1 , z1) 41t.
a
Further, since the left side of (1) is independent of Q, it follows that a
Q-+ 0 we finally get
; 2 S5 u da = 41tu (x 1, Yu z1)
a.
or
u (x 1, y1 , z1) = 4;Ra iS a
u da.
CHAPTER XVI
SERIES
SEC. 1. SERIES. SUM OF A SERIES
Definition t. Let there be given an infinite sequence of num-

bers *)
The expression
U1 +u.+u,+ ... +un+ (1)
is called a numerical series. Here, the numbers ut' u 1 , , un,
are called the terms of the series.
Definition 2. The sum of a finite number of terms (the first n
terms) of a series is called the nth partial sum of the series:
Sn=U 1 +u.+ . +un.
Consider the partial sums
S1 =U 1 ,
S2 =U 1 +u 2,
s3 = u 1 + u. + u 3,
Sn = U1 + U + U + ... + Un-
2 3
If there exists a finite limit

S= lim Sn,
n-+<X>
it is called the sum of the series (1) and we say that the series
converges.
If lim sn does not exist (for example, Sn-+oo as n-+oo),
n-+<X>
then we say that the series (1) diverges and has no sum.
Example. Consider the series
a+aq+aq'+ ... +aqn- + ... 1
(2)
This is a geometric progression with first term a and ratio .q (a .p 0).
) A sequence is considered specified if we know he law by which it is
possible to determine any ierm Un for a given n.
Series. Sum of a Series 711
The sum of the first n terms of the geometric progression is (when q = l)

a-aq"
s,.=t:=q
or
a aq"
s,. = 1 -q- 1-q .
1) If 1 q I < 1, then q"-+ 0 as n-+ oo and, consequently,
lim s,.= lim (-a__ aq") =-a-
n-+oo n-.oo 1-q 1-q 1-q
Hence, in the case of I q I< 1, the series (2) converges and its sum is
a aq"
2) If I q I> 1, then 1q" I-+ oo as n-+ oo and then - -+ oo as n-+ oo,
1 -q
that is, lim s,. does not exist. Thus, when Iq 1 > 1, the series (2) diverges.
n-.oo
3) If q =I, then the series (2) has the form
a+a+a+
In this case
s,.=na, lim s,.= oo,
n-+oo
and the serie~ diverges.
4) If q=- 1, then the series (2) has the form
a-a+a-a+ ...
In this case
0 when n is even,
5
" = { a when n is odd.
Thus, s,. has no limit and the series diverges.
Thus, a geometric progression (with first term different from zero) conver-
ges only when the ratio of the progression is less than unity in absolute
value.
Theorem l. If a series obtained from a given series (1) by sup-

pression of some of its terms converges, then the given series itself
converges.
Conversely, if a given series converges, then a series obtained
from the given series by suppression of several terms also converges.
In other words, the convergence of a series is not affected by the
suppression of a finite number of its terms .
. Proof. Let s,. be the sum of the first n terms of the series (1),
c,., the sum of k suppressed terms (we note that for a sufficiently
large n, all suppressed terms are contained in the sum s,.), and
crn-k is the sum of the terms of the series that enter into the
712 Series
sum sn but do not enter into ck. Then we have

sn = ck + (Jn-k
where ck is a constant that is independent of n.
From the last relationship it follows that if lim an-k exists,
n-+ao
then lim sn exists as well; if lim sn exists, then lim an-k also
n-..ao n-+-co n-+-oo
exists; which proves the theorem.
We conclude this section with two simple properties of series.
Theorem 2. If a series
a,+a.+ ... (3)
converges and Us sum is s, then the series
ca, +ca.+ ... (4)
where c is some fixed number, also converges, and its sum is cs.
Proof. Denote the nth partial sum of the series (3) by sn, and
that of the series (4), by an. Then
an=ca, + ... +can= c (a,+ ... +an) =CSn.
Whence it is clear that the limit of the nth partial sum of the
series (4) exists, since
lim (Jn = lim (csn) = c lim sn = cs.
n-..co n-+-< n-+oo
Thus, the series (4) converges and its sum is equal to cs.
Theorem 3. If the series
a, +a.+ ... (5)
and
(6)
converge and their sums, respectively, are sand S: then the series
(a, +b,)+(a,+b,)+... (7)
and
(8)
also converge and their sums are s+s and s-S, respectively.
Proof. We prove the convergence of the series (7). Denoting
its nth partial sum by an and the nth partial sums of the series (5)
and (6) by sn and ';n, respectively, we get
crn=(a,+b,)+ ... +(an+bn)=
=(al + ... + an)+(b, + ... +bn)= sn+~n
Necessarg Condi lion for Conoergence of a Series 713
Passing to the limit in this equation as n-+ oo, we get
lim (J n = lim (sn +~n> = lim lim ~n = s+";.

sn + n-+-rD
n-+CI) n ... CJ) n-+oo
Thus, the series (7) converges and its sum is + s s.

It is analogously pro~ed that the series (8) also converges and
its sum is equal to s-s.
Of the series (7) and (8) it is said that they were obtained by
means of termwise addition or, respectively, termwise subtraction
of the series (5) and (6).
SEC. 2. NECESSARY CONDiTION FOR CONVERGENCE

OF A SERIES
One of the basic questions, when investigating series, is that

of whether the given series converges or diverges. We shall
establish sufficient conditions for one to decide this question. We
shall also examine the necessary condition for convergence of a
series; in other words, we shall establish a condition for which
the series will diverge if it is not fulfilled.
Theorem. if a series converges, its nth term approaches zero as n
becomes infinite.
Proof. Let the series
U1 +u +u,+ ... +un+ ...
2
converge; that is, let us have the equality

lim Sn=S,
n-+>
where s is the sum of the series (a finite fixed number). But then
we also have the equation
lim sn-l = s,
n-+<XI
since (n-1) also tends to infinity as n-+ oo. Subtracting the

second equation from the first termwise, we obtain
lim Sn- lim Sn-t = 0
n-+oo n-+r:J
or
lim (sn-sn_ 1 ) = 0.
n-+"'
But
714 Series
Hence,
lim Un= 0,
n-+o
which is what was to be proved.

Corollary. If the nth term of a series does not tend to zero as
r. ~ oo, then the series diverges.
Example. The series
1 2 3 n
3+5+7+" +2n+ 1+ ...
diverges, since
lim
n-+>
Un= lim
n-+>
(-n -)
2n+1 =
1
2
:0.
We stress the fact that this condition is only a necessary con-

dition, but not a sufficient condition; in other words, from the
fact that the nth term approaches zero, it does not follow that the
series converges, for the series may diverge.
For example, the so-called harmonic series
1 1 I 1
1 +2+3+-:r+ ... +-;+ ...
diverges, although
lim un = lim _!_ = 0.
n-+-al n_,..co n
To prove this, write the harmonic series in more detail:
l I I I 1 1 1
1 +2+s+-:r+s+6+1+s+
'-v-' _...
1 1 1 1 1 1 1 1 1
+g-+ w + rr+ 12+ 13+ 14 + 15+ 16+ 11+ (I)
~----------------------~
We also write the auxiliary series
16 terms
The series (2) is constructed as follows: its first term is equal

to unity, its second is 1 / 2 , its third and fourth are 1 /,, the fifth
to the eighth terms are equal to 1 / 8 , the terms 9 to 16 are equal
to 1 / 16 , the terms 17 to 32 are equal to 1 /w etc.
Necessary Condition for Con vergence of a Series 715
Denote by s~1 > the sum of the first n terms of the harmonic
series ( 1) and by s~2 > the sum of the first n terms of the series (2).
Since each term of the series (1) is greater than the correspond-
ing term of the series (2) or equal to it, then for n > 2
(3)
We compute the partial sums of the series (2) for values of n

equal to 2\ 22 , 21 , 24 , 2 1 :
I 3
S2= 1+2=2
1 (I4+41) = 1 +2+2=
s,= 1 +2+ 1 1 1 + 2 '2
1
s. = 1 + ++ ( ! + i-) + ( ~ + ~ + i + ~ ) = + 1 3. ~ '
4 terms 8 terms
4 terms 8 terms
in the same way we find that S2 = 1 +6 ~ , s., = 1 + 7. ~ and,

I
generally, Bi.= I +k']'".
Thus, for sufficiently large k, the partial sums of the series (2)
can be made greater than any positive number; that is,
lim s~> = oo,
n-+"'
but then from the relation (3) it also follows that

lim s~1 > = oo
n-+"'
which means that the harmonic series (1) diverges.

716 Series
SEC. 3. COMPARING SERIES WITH POSITIVE TERMS
Suppose we have two series with positive terms:

U1 +U 2 +u,+ ... +un+ .. . , (1)
V1 + V 2 + v, + ... + vn +... (2)
For them the following assertions hold true.
Theorem 1. If the terms of the series (1) are not greater than
the corresponding terms of the series (2); that is,
Un.;;::;; vn (n = 1, 2, ... ) (3)
and the series (2) converges, then the series (l) also converges.
Proof. Denote by sn and an, respectively, the partial sums of
the first and second series:
From the condition (3) it follows that

~.;;::;;~. ~
Since the series (2) converges, its partial sum has a limit a:
lim an=a.
From the fact that the terms of the series ( 1) and (2) are posi-
tive, it foJiows that an< a, and then by virtue of (4)
sn<a.
We have thus proved that the partial sums sn are bounded.
We note that as n increases, the partial sum sn increases, and
from the fact that the sequence of partial sums is bounded and
increases, it follows that it has a limit*)
lim sn=s,
n .... oo
and it is obvious that

s.s;;;a.
Using Theorem 1, we can judge of the convergence of certain series.
) To convince ourselves that the variable sn has a limit, let us recall a

condition for the existence of a limit of a sequence (see Ch. II): "if a vari-
able is bounded and increases, it has a limit. Here, the sequence of sums sn
is bounded and increases. Hence it has a limit, i. e., the series converges.
Comparing Series with Positive Terms 717
Example I. The series

I 1 1 1
1 + 2 + 3 + 4 + ... + r." + ...
2 3 4
I
converges because its terms are smaller than the corresponding terms of the
series
But the last series converges because its terms, beginning with the second,
form a geometric progression with common ratio ; . The sum of this series
.is equal to 1 ~ . Hence, by virtue of Theorem 1, the given series also con-
I
verges, and the sum does not exceed I .
2
Theorem 2. If thP terms of the series (1) are not smaller than
the corresponding terms of the series (2); that is,
(5)
and the series (2) diverges, then the series (1) also diverges.
Proof. From condition (5) it follows that
(6)
Since the terms of the series (2) are positive, its partial sum a,.
increases with increasing n, and since it diverges, it follows that
lim a,.= oo.
But then, by virtue of (6),

lim s,.= oo,
the series ( 1) diverges.
Example 2. The series
l I 1
1
+ 2 + 3 + + VIi+
diverges because its terms (from the second on) are greater than the corre-
sponding terms of the harmonic series
I 1 I
1
+2+3++11+
which, as we know, diverges.
Note. Both the conditions that we have proved (Theorems 1

and 2) hold only for series with positive terms. They also hold
718 Series
true when some of the terms of the first or second series are zero.
But these conditions do not hold if some of the terms of the
series are negative numbers.
SEC. 4. D'ALEMBEIU'S TEST
Theorem (d' Alembert's Test). If in a series with positive terms

U1 + U + U + ... + Un + . . .
2 3 (1)
+
the ratio of the (n 1) sf term to the nth term, as n - oo, has a
(finite) limit l, that is,
I l. i D Un+l l
--=' (2)
n .... .., Un
then:
I) the series converges for l < 1,
2) the series diverges for l > 1.
(For l = 1, the theorem cannot determine the convergence or
divergence of the series.)
Proof. 1) Let l < 1. Consider a number q that satisfies the
relationship l < q < 1 (Fig. 342).
From the definition of a limit and relation (2) it follows that
for all values of n after a certain integer N, that is, for n ~ N,
we will have the inequality
(2')
Indeed, since the quantity un+J tends to the limit l, the dif-
un
ference between the quantity 11 n+l an~ the number l may (after a
Un .
certain N) be made less (in absolute value) than any positive
number, in particular Jess than q- L; that is,
j u~:' -tj < q -l.
Inequality (2') follows from this last inequality. Writing this

inequality for various values of n, from N onwards, we get
UN+1 < QUN, }

UN+ < quN+t < quN,
(3)
UN+I < QUN+2 < quN,
I I I o I 0 I I I I
D' A lembert' s Test 719
Now consider the two series

u. +u2+us+ ... +uN+uN+ +uNH + ... , (1)
uN+ quN+q2uN+... (1')
The series (1 ') is a geometric progression with positive common
ratio q < I. Hence, this series converges. The terms of the
q-1 q-1 1-t 1-t
,o
,c:=:->c:=;;->t
,101!l.q
1
t
I
,c:;;:-.,~, .
0 t !!JM I
Un Un
Fig. 342. Fig. 343.
series (1), after uN+t are Jess than the terms of the series (I').
By Theorem I, Sec. 3, and Theorem I, Sec. I, it follows that
the series ( 1) converges.
2) Let l >I.
Then from the equation Jim un+t =I (where I> I) it follows
n-+ co Un
that, after a certain N, that is for n;;;;. N, we will have the
inequality
(Fig. 343), or un+ > un for all n;;;;. N. But this means that the
terms of the series increase after the term N +I, and for this
reason the general term of the series does not tend to zero. Hence,
the series diverges.
Note 1. The series will also diverge when lim un+t = oo. This
n-+ ao Un
follows from the fact that if Jim un+t = oo, then after a certain
n-+ c:o Un
n = N we will have the inequality un+ L::> 1, or un+t > un.

Un
Example t. Test the following series for convergence!

1 1 1
1+n+ 1.2.3+ + 123 ... n +
Solution. Here,
1
Un==12 n nt' (n+ 1)1'
720 Series
Hence,
The series converges.

Example 2. Test for convergence the series
2 22 21 2n
1+2+3++n+
Solution. Here,
2n+1
Un+t=n+l:
lim un+t = lim 2 _n_=2> 1.
n ... a> Un n ... a> n+l
The series diverges and its general term un approaches infinity,
Note 2. D'Alembert's test tells us whether a given positive

series converges; but it does so only when lim un+t exists and is
rt-+ ex~ Un
different from 1. But if this limit does not exist or if it does
exist and lim un+t = 1, then d'Alembert's test does not enable us
n ~CD Un
to tell whether the series converges or diverges, because in this
case the series may prove to be both convergent and divergent.
Some other test is needed to determine the convergence of such
series.
It will be noted, however, that if lim Un+t =I, but the ratio
n ..... ao un
for all n (after a certain one) is greater than unity, the se-
ries diverges. This follows from the fact that if un+t >I then
Un '
un+t > un and the general term does not approach zero as n--..oo,
To illustrate, let us examine some examples.
1 2 3 n
2+3+4+ ... + n+l + ...
Solution. Here,
In this case the series diverges because Un+t

Un.
> 1 for all m
Un+l n 2 +2n+ I
-un
-= n2+2n > 1.
Cauchy's Test 721
Example 4. Using the d' AI em bert test, examine the harmonic series
1 1 1
1 +2+a++r1+
1 1
We note that u 11 =n, u 11 + 1= n+
1
and, consequently,
lim u11+1= lim _n_=l.

11-+a> U11 11-+"'n+i
Thus, d' Alembert's test does not allow us to determine the convergence or
divergence of the given series. But we earlier found out by a different
expedient that a harmonic series diverges.
1 I I 1
1.2+2.3+3.4 + + n (n+ 1) +
Solution. Here,
1 1
u,.=n(n+1)' u 11 + 1 (n+1)(n+2)'
lim Un+l= lim n(n+1) = lim _n_=l
11-+"' u11 11-+"' (n+ I) (n+2) 11-+"' n +2
D' Alembert's test does not permit us to infer that the series converges;
but by other reasoning we can establish the fact that this series converges.
Noting that
n (n+i)
we can write the given series in the form
The partial sum of the first n terms, after removing brackets and cancell-
ing, is
1
Sn= 1- n +1.
Hence,
lim sn= lim
1-)=1.
(1--
11 ..... "' 11-+> n+1
That is, the series converges and its sum is I.
SEC. 5. CAUCHY'S TEST
Theorem (Cauchy's Test). If for a series with positive terms

U1 + + U + ... + U 11 + ...
U2 3 (1)
the quantity ).1U11 has a finite limit l as n-+ oo, that is,
lim fYU:=l,
n-+>
722 Series
then: 1) for l < 1, the series converges;

2) for l > 1, the series diverges.
Proof. 1) Let l < 1. Consider the number q that satisfies the
relation l < q < 1.
After some n=N we will have the relation
IJ1Un-ll < q-l;
whence it follows that
JYU:<q
or
for all n;:;:.N.

Now consider two series:
u, + U2 + U3 + ... + uN+uN+ +uN+ + ... , (1)
qN + qN+l + qN+2+... (1')
The series (1 ') converges sirice its terms form a decreasing
geometric progression. The terms of the series (1 ), after uN, are
less than the terms of the series (1'). Consequently, the series (1)
converges.
2) Let l > 1. Then, after some n = N, we will have
;Yun>1
or
un> I.
But if all the terms of this series, after uN, exceed 1, then the
series diverges, since its general term does not tend to zero.
Example. Test for convergence the series
3+
1
5 + 7 + ... +
( 2 )' ( 3 )
3
. ( n
2n+l
)n +
Solution. Apply the Cauchy test:
lim yn/-
n-+a>
un= lim
n-+<D
V'( _n_ )n = lim - -
\2n+l
n =-
n ..... .., 2n+!
I
2
< 1.
The series converges.
Note. As in the d'Aiembert test, the case

Jim V'u:=l=1
requires further investigation. Among the series that satisfy this

condition are convergent and divergent series. Thus, for the har-
The Integral test (or Convergence of a Series 723
monic series (which is known to be divergent)
lim
n--+IXI
JYii: = lim
n--+CD
V..!.. =I.
n
To be sure, we shall prove that lim In v~ = 0. Indeed,
noo+co
. In
I 1m VT . -=lim
n
-Inn
--.
n--+-aJ n
Here, the numerator and denominator of the fraction approach
infinity. Applying !'Hospital's rule, we find
. In
I1m
n-+ct:~
VT -= I"tm--=
n
-Inn
n-+oo
-n 0
I"tm--=.
n n-+-oo 1
1
Thus, In n --..0, but then v~--+ 1, i. e.,
For the series

lim
n -i>- a:~ VT -= 1.
n
l +I+ 1 + . 1 .
T 22 32 -t-n2+
we also have the equality
lim
n-+a>
vn~
un= lim
n-+-co
VT -.= lim
n n-+ctJ
VTVT -
n
-=I.
n
but this series converges, since if we suppress the first term, the
terms of the remaining series will be less than the corresponding
terms of the converging series
1 1 l
1.2+2.3+ ' + n(n+l)+ ''
SEC. 6. THE INTEGRAL TEST FOR CONVERGENCE

OF A SERIES
Theorem. Let the terms of the series

U1 +u +u,+ ... +un + ...
2 (1)
be positive and not increasing, that is,
724 Series
and let f (x) be a continuous nonincreasing function such that

f(I)=u,; f(2)=U 2 ; ; f(n)=un. (2)
Then the following assertions hold true.
I) if the improper integral
CIO
~ f (x) dx
1
converges (see Sec. 7, Ch. XI), then the series (I) converges too;
2) if the given integral diverges, then the series (I) diverges
as well.
Proof. Depict the terms of the series geometrically by plotting
on the x-axis the numbers I, 2, 3, ... , n, n + 1, ... of the terms
of the series, and on they-axis, the corresponding values of the terms
of the series u" U 2 , , un, ... (Fig. 344).
In the same coordinate system plot the graph of the continuous
nonincreasing function
y= f (x)
which satisfies condition (2). .
An examination of Fig. 344 shows that the first of the construct-
ed rectangles has base equal to I and altitude f (I)= u,. The
area of this rectangle is thus u,. The area of the second one is U 2 ,
and so on; finally, the area of the last (nth) of the constructed
rectangles is un. The sum of the areas of the constructed rectangles
is equal to the sum sn of the first n terms of the series.
On the other hand, the step-like figure formed by these rectangles
embraces a region bounded by the curve y = f (x) and the straight
lines x=I, x=n+I, y=O; the area of this region is equal to
n+t
~ I (x) dx. Hence,
n+t
Sn > ~ f (x) dx. (3)
Let us now consider Fig. 345. Here the first of the constructed
rectangles on the left has altitude u,; and so its area is u,.
The area of the second rectangle is U 2 , and so forth. The area
of the last of the constructed rectangles is u.,+ Hence, the sum
of the areas of all constructed rectangles is equal to the sum of
all terms of the series beginning from the second to the (n I) st +
The Integral Test for Convergence of a Series 725
or sn+ 1 -u 1 On the other hand, it is readily seen that the step-

like figure formed by these rectangles is contained within the
y y
0 .-
0 f 2 3 n n+! x
Pig. 344. Pig. 345.
curvilinear figure bounded by the curve y= f (x) and the straight

lines x= 1, x=n+ I, y=O. The area of this curvilinear figure
n+l
is equal to S f (x) dx. Hence,
1
n+l
Sn+ 1 - U 1 ~ ~ f(x)dx,
whence
11+1
Sn+ 1 < 5 {(x)dx+u,.

I
(4)
Let us now consider both cases.

"'
l. We assume that the integral ~ f(x) dx converges, that is 1
has a finite value.
Since
11+1 ao
~ f (x) dx
I
< 5f (x) dx,
I
it follows, by virtue of inequality (4), that

00
Sn < Sn+t < 5f 1

(X) dx+u,.
Thus, the partial sum sn remains bounded for all values of n.

But it increases with insreasing n, since all the terms un are
726 Series
positive. Consequently, sn (as n--+ oo) has the finite limit

lim Sn=S
n,...,
and the series converges.
"'
2. Assume, further, that ~ f (x) dx = oo. This means that
n+l
~ f (x) dx increases without bound as n increases. But then, by
I
virtue of inequality (3), sn likewise increases indefinitely with n;

The theorem is thus proved completely.
Example. Test for convergence the series
Solution. Apply the integral test, putting

1
f(x) = xP.
This function satisfies all the conditions of the theorem. Consider the
integral
5:: = 1
1 _P x -P
1 IN =
1
I 1
1 _p (N -P-1) when p ;e 1,
1
{
In x !i" =In N _ when p = 1.
Allow N to approach infinity and determine whether the improper

integral converges in various cases.
It will then be possible to judge about the convergence or divergence of
the series for various values of p.
(X)
F or p > 1,
S dx
xP = p-1 1 , th e in t egra I is fimite and, hence, the series
converges;
"'
for p < 1, s~;=oo, the integral is infinite, and the series diverges;
I
..,
for p = 1, S dxx =co, the integral is infinite; and the series diverges.
1
Alternating Series. Lei bniz' T!t'eonem 727
We note that neither the d'Alembert test nor tlie Cauchy test, which
were considered earlier, decide whether the series is convergent or not, since
u
Urn ....!!...!.= lim ( n )P =1,
-+
n-+co Un n-+(X) n 1
lim Vn/-
n-+"'
Un= lim
n .... "'
VT- = lim (VT)P
nP
- =lP =1.
n .... "' n
SEC. 7. ALTERNATING SERIES. LEIBNIZ' THEOREM
So far we have been considering series whose terms are all

positive. In this section we consider series whose terms have
alternating signs, that is, series of the form
ul-u2+us-u4 (I) + ... ,
where u 1 , u,, ... , un, ... are positive.
Leibniz' Theorem. if in the alternating series
U1 - U2 +U 3 -U4 + ... (un > 0) (I)
the terms are such that
(2)
and
lim Un=O, (3)
then the series (I) converges, its sum is positive and does not
exceed the first term.
Proof. Consider the sum of the first n =2m terms of the
series (1):
s.m = (ul- u.) +(us- u4) + ... + (u2m-t-u2m>
From condition (2) it follows that the expression in each of
the brackets is positive. Hence, the sum s 2 m is positive,
s.m>O.
and increases with increasing m.
Now write this sum as follows:
S2m= U 1- (u 1- U 8}- (U 4- u 5)- .. - (u 2m_ 2- U 2m_ 1)-u 2;;..
By virtue of condition (2}, each of the parentheses is positive.
Therefore, subtracting these parentheses from u 1 we get a number
Jess than U 1 , or
728 Series
We have thus established that S 2 m increases with increasing m

and is bounded above. Whence it follows that Szm has the limit s:
lim S 1 m =S,
m-+ao
and
O<s<u 1
However, we have not yet proved the convergence of the series;

we have only proved that a sequence of "even" partial sums has
as its limit the number s. We now prove that "odd" partial sums
also approach the limit s.
Consider the sum of the first n =2m+ 1 terms of the series (1 ):
Szm+t = Szm + Uzrn+t
Since, by condition (3), lim um+t = 0, it follows that
m-+ao
lim
m-+r
S 2 m+t = lim
m-+e
S21n + lim
m-+rn
Uzrn+t = Jim
m-+oo
Szm = S.
We have thus proved that lim sn = s both for even n and

n-+""
for odd n. Hence, the series (1) converges.
Note l. The Leibniz theorem may be illustrated geometrically
as follows. Plot the following partial sums on a number line
(Fig. 346):
S 1 = U1 , s, = U1 -u, = s,-u 2 , s, = s, +u,, S 4 =s,-u~, s, =S +u,,
6
etc.
The points corresponding to partial sums will approach a
certain point s, which depicts the sum of the series. Here, the
points corresponding to the
~-----u2
u3
us-auau6+ ~
:1 I
even partial sums lie on
the left of s, and those
~~rrt~~~7~~~goios~dd sums,
Note 2. If an alternating
---~>---.t>--~---c>---.1>-----0---o.---" series satisfies the statement
S2 8+ s6 s S5 S3 s1-u1 of the Lei bniz theorem,
Fig. 346. then it is easy to evaluate
the error that results if
we replace its sum, s, by the partial sum sn. In this substi-
tution we suppress all terms after un+t' But these numbers form
by themselves an alternating series, whose sum (in absolute value)
is less than the first term of this series (that is, less than un+ 1 ).
Thus, the error obtained when replacing s by sn does not exceed
(in absolute value) the first of the suppressed terms.
Plus-and-Minus Series. Absolute and Conditional Convergence 729
Example I. The series
converges, since
1 I
1) 1 > 2> a-> ... :
2) lim Un = lim _!_ = 0.
n-..~ n-+r.:n 11
The sum of the first n terms of this series

I 1 I n l
Sn=J-2+a---;r++(-J) n
differs from the sum s of the series by a quantity less than n+I 1
Example 2. The series
converges by virtue of the Leibniz theorem.
SEC. 8. PLUS-AND-MINUS SERIES. ABSOLUTE

AND CONDITIONAL CONVERGENCE
We give the name plus-and-minus series to a series that has

both positive and negative terms. .
Obviously, the alternating series considered in Sec. 7 is a
special case of plus-and-minus series*).
We shall consider some properties of alternating series.
In contrast to the agreement made in the preceding section we
will now assume that the numbers U 1 , u 1 , , un . . . can be
both positive and negative.
First, let us give an important sufficient condition for the
convergence of an alternating series.
Theorem 1. 1f the alternating series
u.+uz++un+ (I)
is such that a series made up of the absolute values of its terms,
Iu. I+ I u.l + ... + I un I+ .... (2)
converges, then the givm alternating series also converges.
Proof. Let sn and an be the sums of the first n terms of the
series (1) and (2).
) In this English edition we shall use the term alternating series for
both types.- Tt.
730 Series
Also, let s~ be the sum of all the positive terms, and s~. the
sum of the absolute values of all the negative terms of the first
n terms of the given series; then
Sn = s'n -s:; an= s~ +s:.
By hypothesis, an has the limit a; s~ and s: are positive in-
creasing quantities less than a. Consequently, they have the
limits s' and s". From the relationship [sn=s~-s: it follows that
sn also has a limit and that this limit is equal to s' -s", which
means that the alternating series (I) converges.
The above-proved theorem enables one to judge about the
convergence of some alternating series. In this case, the test for
convergence of the alternating series reduces to investigating a
series with positive terms.
Consider two examples.
sin a +sin 2a +sin 3a + +sin na + (3)
la 22 32 ... n2 ... '
where a is any number.

Solution. Also consider the series
I si ~ a I+ I si~;a I+ I si;;a I+ ... +I

1 sl:~a I+ .. (4)
and
(5)
The series (5) converges (see Sec. 6). The terms of the series (4) are not
greater than the corresponding terms of the series (5); hence, the series (4)
also converges. But then, in virtue of the theorem just proved, the given
series (3) likewise converges.
:rt :rt :rt :rt
cos cos3 coss cos(2n-l)T
4 4 4
-3-+--32-+ 32 ++ 3n + (6)
Solution. In addition to this series, consider the series

l I l l
a+32 +aa+ ... +3n+ ... (7)
This series converges because if is a decreasing geometric progression with

ratio ~ . But then the given series (6) converges, since the absolute values
of its terms are less than those of the corresponding terms of the series (7).
Plus-and-Minus Series. Absolute and Conditional Conuergence 731
We note that the convergence condition that was proved earlier

is only a sufficient condition for convergence of an alternatin~
series, but not a necessary condition: there are alternating series
which converge, but series formed from the absolute values of
their terms diverge. In this connection, it is useful to introduce
the concepts of absolute and conditional convergence of an
alternating series and, on the basis of these concepts, to classify
alternating series.
Definition. The alternating series
U1 + U + u, + ... + un +...
2 (1)
is called absolutely convergent if a series made up of the absolute
values of its terms converges:
IU I+ IU +I u, I+ +I Un I+
1 2 1 (2)
If the alternating series (1) converges, while the series (2)
composed of the absolute values of its terms diverges, then the
given alternating series (1) is called a conditionally convergent
series.
Example 3. The alternating series
1 1 1
1 -2+3-4+
Is conditionally convergent, since a series composed of the absolute values

of its terms is a harmonic series,
I I I
1 +2+3+4+ ....
which diverges. The series itself converges (this can be readily verified by
Leibniz' test).
I 1 1
I-2T+3T-4f+ ...
is absolutely convergent, since a series made up of the absolute values of
its terms,
converges, as established in Sec. 4.
Theorem 1 is frequently stated (with the help of the concept

of absolute convergence) as follows: every absolutely convergent
series is a convergent series.
In conclusion, we note (without proof) the following properties
of absolutely convergent and conditionally convergent series.
732 Series
Theorem 2. If a series converges absolutely, it remains abso-

lutely convergent for any rearrangement of its terms. The sum of
the series is independent of the order of its terms.
This property does not hold for conditionally convergent series.
Theorem 3. If a series converges conditionally, then no matter
what number A is given, the terms of this series can be rearranged
in such manner that its sum is exactly Pqual to A. What is more,
it is possible so to rearrange the terms of a conditionally conver-
gent series that the series resulting after the rearrangement is
divergent.
The proofs of these theorems are beyond the scope of this course.
To illustrate the fact that the sum of a conditionally convergent
series can change upon rearrangement of its terms, consider the
following example.
I I I
1-2+3-4+ ... (8)
converges conditionally. Denote its sum by So It is obvious that s > 00

Rearrange the terms of the series (8) so that two negative terms follow one
positive term:
11I11 I I I
I-2-4+3-6-s+o+2k-1-4k-2-4k+... <9>
'-..-' '-..-' '---...----
We shall prove that the resultant series converges, but that its sum s' is
half the sum of the series (8): ~ so Denote by sn and s~ the partial sums of
the series (8) and (9)o Consider the sum of 3k terms of the series (9):
1 1
8
ak = ( 1- ~ - ~ ) + ( ; - ~ - ! )+ + ( 2k 1- 4k 2- 4~) =
= ( ~ - {-) + ( ~ - i) + ... + ( 4/ 2- ~) =
l( 1
=} 1- ~ ) + ( ~ - {-) + + ( 2k 1- 2k)] =
1
0

0
1( 1 1 I 1 1) 1
=2 1-2+3-4+ooo+2k-l-2k =2s2ko
Consequently,
0
1tm '
s,k =
10
tm -1 s2k = -
1
s.
k .... "' k .... "' 2 2
Further,
Functi011al Series 733
And we obtain
. s,.'
Iun = s
' = 21 s.
n-+ co
Thus, in this case the sum of the series changed after its terms were
rearranged (it diminished by a factor of 2).
SEC. 9. FUNCTIONAL SERIES
The series U 1 + U 2 +: ... + u,. + ... is called a functional series

if its terms are functions of x.
Consider the functional series
(I)
Assigning to x definite numerical values, we get different

numerical series, which may prove to be convergent or divergent.
The set of all those values of x for which the functional series
converges is called the domain of convergence of the series.
Obviously, in the domain of convergence of a series its sum is
some function of x. Therefore, the sum of a functional series is
denoted by s (x).
Example. Consider the functional series
1 +x+x 2 + ... +x"+ ...
This series converges for all values of x, in the interval (-1, 1), that is,
for all x that satisfy the condition I xI < 1. For each value of x in the
1
interval (-1, 1)! the sum of the series is equal to - - - (the sum of a
1-X
decreasing geometric progression with ratio x). Thus, in the interval (- 1, 1)
the given series defines the function
1
s(x)=1-x
which is the sum of the series; that is,

1
- - = 1 +x+x 2 +x 3 + ...
1-X
Denote by s,. (x} the sum of the first n terms of the series ( 1).
If this series converges and its sum is equal to s (x), then
s (x) =s,. (x) +r,. (x),
where r,. (x) is the sum of the series u,.+ 1 (x) + u,.+s (x) + ... , i.e.,
r,. (x) = u,.+ 1 (x) + u,.+s (x) + ...
734 Series
Here, the quantity rn (x) is called the remainder of the series (1).
For all values of x in the domain of convergence of the series we
have the relation lim sn (x) = s (x); therefore,
lim rn (x) = lim [s (x) -sn (x)] = 0,

n-+-co n-+-co
which means that the remainder rn (x) of a convergent series ap-

proaches zero as n ___.. oo.
SEC. 10. DOMINATED SERIES
Definition. The functional series

U 1 (x) + U 2 (x} u, (x) + + ... + un (x) + .-;-r (1)
is called dominated in some range of x if there exists a conver-
gent numerical series
a1 +a,+ a.+ ... +an+ . . . (2)
with positive terms such that for all values of x from this range
the following relations are fulfilled:
IU 1 (x) / ,;:;;;; a 1 , Iu, (x) I ,;:; ; a 2 , , Iun (x} / ,;:;;;; an, . (3)
In other words, a series is called dominated if each of its terms
does not exceed, in absolute value, the corresponding term of some
convergent numerical series with positive terms.
For example, the series
cos x + cos 2x + cos 3x + + cos nx +
1 22 32 n
is a series majorised on the entire x-axis. Indeed, for all values
of x, the relation
cos nx
I n I.,;: n
_!_ ( _ l 2 ... ),
--= 2 n- ' '
is fulfilled and the series
1 1 1
1+22+s.+ ... ,
as we know, converges.
From the definition it follows straightway that a series domina-
ted in some range converges absolutely at all points of this range
(see Sec. 8). Also, a dominated series has the following important
property.
Theorem. Let the functi-onal series
U1 (x) +U 1 (x) + ... + un (x) + ...
Dominated Series 735
be dominated on the interval [a, b]. Let s (x) be the sum of this
series and sn (x) the sum of the first n terms of this series. Then
for each arbitrarily small number e > 0 there will be a positive
integer N such that for all n ~ N the following inequality will be
fulfilled,
Is (x)-sn (x) I< e,
no matter what the x uf the interval [a, b].
Proof. Denote by a the sum of the series (2):
a= a1 + a2 +a~ + ... + an + an+ + ... , 1
then
where an is the sum of the first n terms of the series (2), and en
is the sum of the remaining terms of this series; that is,
f,n = an+ 1 +an+ 2 + ...
Since this series converges, it follows that
lim an= a
and, consequently,
lim en= 0.
Let us now represent the sum of the functional series (1) in the
form
S (x) = Sn (x) +
r n (x),
where
Sn (x) = U 1 (x) + ... +
Un (x),
+
r n (x) = un+ 1 (x) + Un+z (x) ttn+z (x) + ...
From condition (3) it follows that

I Un~-~ (x) I,;;;;; an+1 IUn+z (x) I,;;;;; an+ a
and therefore
Ir n (x) I ,;;;;; en
for all x of the range under consideration.
Thus,
Is (x)- Sn (x) I <en
for all x of the interval [a, b], and Bn-+0 as n-+oo,
Note t. This result may be represented geometrically as follows.
Consider the graph of the function y = s (x). About this curve
construct a band of width 2en; in other words, construct the
736 Series
curves y = s (x) +en and y = s (x)- en (Fig. 347). Then for any en the
graph of the function sn (x) will lie completely in the band under
consideration. The graphs of all successive partial sums will like-
y wise lie within this band.
Note 2. Not every function-
/'' .... -- en al series convergent on the
__
', ,/I interval [a, b1 has the pro-
perty indicated in the forego-
ing theorem. However, there
are nondomi nated series such
that possess this property.
A series that possesses this
-.jf---L--------+___,~x property is called a uniform-
a a b ly convergent series on the
. interval [a, b].
Ftg. 347. Thus, the functional series
U1 (x)+ua (x) + ... + un (x)+
+. . . is called a uniformly convergent series on the interval [a, b]
if for any arbitrarily small e > 0 there is an integer N such that
for all n;;;;;;. N the inequality
Is (x)-sn (x) I< e
will be fulfilled for any x of the interval [a, b].

From the theorem that has been proved it follows that a domi-
nated series is a series that uniformly converges.
SEC. 11. THE CONTINUITY OF THE SUM OF A SERIES
Let there be a series made up of continuous functions

U1 (x) +U 1 (x) + ... + un (x) + ... ,
convergent on some interval [a, b].
In Chapter II we proved a theorem which stated that the sum
of a finite number of continuous functions is a continuous func-
tion. This property does not hold for the sum of a series (consist-
ing of an infinite number of terms). Some functional series with
continuous terms have for the sum a continuous function, while
in the case of other functional series with continuous terms, the
sum is a discontinuous function.
Example. Consider the series
1 1 I I 1 1
The Continuity of the Sum of a Series 737
The terms of this series (each term is bracketed) are continuous functions for
all values of x. We shall prove that this series converges and that its sum
is a discontinuous function.
We find the sum of the first n terms of the series:
Find the sum of the series:

if x > 0, then
s= lim s11 (x)= lim (x 2 "+'-x)=l-x,

n-i" m "...,. oo
if x < 0, then
s= lim. s,. (x)= lim (-1 x 12 "+'-x) = - I -x,

n-+co n-+-oo
if x=O, then s,.=O, and so s= lim s,.=O. Thus, we have
s (x) =I -x for x > 0,

s (x) = - 1 -x for x < 0,
s(x)=O for x=O.
And so the sum of the given series Is a discontinuous function. Its graph is
shown in Fig. 348 along with the graphs of the partial sums s, (x), s2 (x),
and s 1 (x).
The following theorem holds true for dominated series.
Theorem. The sum of a series of continuous functions dominated
on some interval [a, b] is a function continuous on this interval.
y
f-X fOt J(~O

S(x)= 0 for x-o
{
-t-x-for x<O
Fig. 348.
24 -3388
738 Series
Proof. Let there be a series of continuous functions dominated

on the interval [a, b]:
U1 (x)+u 2 (x)+u.(x)+... (1)
Let us represent its sum in the form
+
S (x) = Sn (x) rn (x),
where
sn(x)=ul (x)+ ... +un(x)
and
rn (x) = Un+t (x) + un+2 (x) + .. ;
On the interval [a, b] take an arbitrary value of the argument
x and give it an increase !l.x such that the point x+ !l.x should
also lie on the interval [a, b].
We introduce the notations
!l.s = s (x + !l.x) -s (x);
!l.sn = sn (x + !l.x)-sn (x);
then
!l.s = !l.sn + r n(x+ !l.x)-r n (x),
from which we have
I!l.s I~ I!l.snl +I rn(x+ !l.x)l +I rn (x) J. (2)
This inequality is true for any integer n.
To prove the continuity of s (x), we have to show that for any
preassigned and arbitrarily small 8 > 0 there will be a number
6 > 0 such that for all I !l.x I< 6 we will have I !l.s I< e.
Since the given series (I) is dominated, it follows that for any
preassigned 8 > 0 there will be found an integer N such that for
all n ~ N (and as a particular case, n = N) the inequality
JrN(x)J<; (3)
will be fulfilled for any x of the interval [a, b]. The value x+ !l.x
lies on the interval [a, b] and therefore the following inequality
is fulfilled:
(3')
Further, for the chosen N the partial sum SN (x) is a continuous
function (the sum of a finite number of continuous functions) and,
consequently, a positive number 6 may be chosen such that for
every !l.x that satisfies the condition I!l.x J < 6 the following
inequality is fulfilled:
(4)
Integration and Differentiation of Series 739
By inequalities (2), (3), (3'), and (4), we have

lt:lsl<: +i+i=e,
that is,
lt:lsl<e for lt:lxl<l'l,
which means that s (x) is a continuous function at the point x
(and, consequently, at any point of the interval [a, b]).
Note. From this theorem it follows that if the sum of a series
is discontinuous on some interval [a, b], 'then the series is not
dominated on this interval. In particular, the series given in the
example is not dominated (on any interval containing the point
x=O, that is to say, a point of discontinuity of the sum of the
series).
We note, finally, that the converse statement is not true: there
are series, not dominated on an interval, which, however, converge
on this interval to a continuous function. For instance, every
series uniformly convergent on the interval [a, b] (even if it is not
dominated) has a continuous function for its sum (if, of course,
all terms of the series are continuous).
SEC. 12. INTEGR.ATION AND OIFFER.ENTIATION Of SER.IES
Theorem 1. Let there be a series of continuous functions

U 1 (x)+!.l 1 (x)+ ... +un(x)+ . . , (I)
dominated on the interval [a, b) and let s (x) be the sum of this
series. Then the integral of s (x) between the limits from a to x,
which limits belong to the interval [a, b], is equal to the sum of
such integrals of the terms of the given series; that is,
X X X X
~ s (x) dx = ~ u, (x) dx + ~ u. (x) dx + ... + ~ un (x) dx + ...

a a a a,
Proof. The function s (x) may be represented in the form
s (x) = sn (x) + rn (x)
or
s (x) = U1 (x) +U 1 (x) + .. + Un (x) + r n (x).
Then
X X l;.
~ s (x) dx = ~ U 1 (x) dx + ~ U 1 (x) + ... +

a a a
~ I'.
+ ~ u n (x) dx + ~ rn (x) dx (2)

a a
24*
740 Series
(the integral of the sum of a finite number of terms is equal to

the sum of the integrals of these terms).
Since the original series (l) is dominated, it follows that for
every x we have I rn (x) I< c,., where en -+0 as n-+ oo. Therefore,
X X X
I~ rn (x) dx I~~ I rn (x) I dx < ~ e,. dx = e,. (x-a) ~en (b -a).

a a a
Since e,. ___,. 0, it follows that
X
lim ~ r,. (x) dx=O.

n-+ ctJ a.
But from equation (2) we have
X X X X
~ rn (x) dx= ~ s (x) dx- [~ U 1 (x) dx + ... + ~ un (x) dx].

a a a a
Hence
n-+ctJ a
X
lim { ~ S (x) dx- na

X
U1 (X) dx+ ...

X
+ a~ Un (x) dx ]} = 0,
or
X X X
n-+oo
lim U
a
U1 (x) dx + ... + a~ u,. (x) dx J= ~ s (x) dx.
a
(3)
The sum in the brackets is a partial sum of the series

X X
~ U 1 (x) dx + ... + ~ u,. (x) dx + ... (4)

a a
Since the partial sums of this series have a limit, this series
converges and its sum, by virtue of equation (3), is equal to
X
~ s (x) dx, i. e.,

a
X X X X
~ s(x) dx= ~ U 1 (x)dx+ ~ U 2 (x) dx+ ... + ~ u,. (x) dx+ ... ,
a a a <t
this is the equation that had to be proved.

Note 1. If a series is not dominated, term-by-term integration
of it is not always possible. This is to be understood in the sense
X
that the integral ~ s (x) dx of the sum of the series (l) is not always
a
Integration. and Differentiation. of Series 741
equal to the sum of the integrals of its terms [that is, to the sum
of the series (4)].
Theorem 2. If a series,
U1 (x) + U 2 (x) + ... + un (x) + ... , (5)
made up of functions having continuous derivatives on the interval
[a, b] converges (on this interval) to the sum s (x) and the series
d.(x) + u: (x) + ... + Un (x) + ... (6)
made up of the derivatives of its terms is dominated on the same
interval, then the sum of the series of derivatives is equal to the
derivative of the sum of the original series; that is,
,
s (x) = U1
~
(x) +
,
U2 (x) +
~
Ua (x) + ... + Un
~
(x) + ...
Proof. Denote by F (x) the sum of the series (6):
F (x) = u: (x) + u:(x) + ... + u~ (x) + ... ,
and prove that
F (x) = s' (x).
Since the series (6) is dominated, it follows, by the preceding
theorem, that
X X X X
~ F (x) dx = ~ u; (x) dx + ~ u; (x) dx + ... + ~ u~ (x) dx + ...

a a a a
Performing the integration, we get
X
~ F (x) dx = [u 1 (x)-u 1 (a}]+

a
+ [u 2 (X) - U 2 (ex.)]+ ... + [un (x)-un (u)] + ...
But, by hypothesis,
s(x)=u 1 (x)+u.(x)+ +un(x)+ .. ,
s(u)=u,(a)+u.(a}+ ... +un(u)+ ... ,
no matter what the numbers x and u on the interval [a, b].
Therefore,
X
) F (x)dx = s(x)-s(a).
(I
Differentiating both sides of this equation with respect to x,

we obtain
F (x) = s' (x).
742 Series
We have thus proved that when the conditions of the theorem

are fulfilled, the derivative of the sum of the series is equal to
the sum of the derivatives of the terms of the series.
Note 2. The requirement of dominance (majorisation) of a series
of derivatives is extremely essential, and if not fulfilled it can
make term-by-term differentiation of the series impossible. This is
illustrated by a dominated series that does not admit term
by-term differentiation.
Consider the series
sin l'x +sin 2'x +sin 3'x + +sin nx +
1" 22 32 n
This series converges to a continuous function because it is
dominated, Indeed, for every x its terms are (in absolute value)
less than the terms of the numerical convergent series with
positive terms
I I l I
1+22+w+ .. +-n.+ ..
Write a series composed of the derivatives of the terms of the
original series:
cosx-!-2 1 cos2'x+ ... +ncosn'x+ ...
This series diverges. Thus, for instance, for x = 0 it turns into
the series
1 +2 2 +3 2 + ... +n'+ ...
(It may be shown that it diverges not only for x = 0.)
SEC. 13. POWER SERIES. INTERVAL OF CONVERGENCE
Definition 1. A power series is a functional series of the form

a0 + a1 x a,x + + ...
+ anx" + ... , (I)
where a 0 , a 1 , a., ... , a,, ... are constants called coefficients of
the series.
The domain of convergence of a power series is always some
interval, which, in a particular case, can degenerate into a point.
To convince ourselves of this, let us first prove the following theorem,
which is very important for the whole theory of power series.
Theorem l (Abel's Theorem). 1) If a power series converges for
some nonzero value X0 , then it converges absolulely for any value
of x, for which
2) if a series diverges for some value x~, then it diverges for every
x for which
Power Series. Tnterval of Convergence 743
Proof. I) Since, by assumption, the numerical series

a + a X + a 2 X~ + ... + anx~ + . . .
0 1 0 (I')
converges, it follows that its common term anx~'~ 0 as n ~ oo,
and this means that there exists a positive number M such that
all the terms of the series are less than M in absolute value.
Rewrite the serie~ (I) in the form
ao +a X (f.) +a 2 x~ (~r + ... +anx~ (~r +...

1 0 (Ia)
and consider a series of the absolute values of its terms:
laol+latxo1]f.]+la2x!l~f.[a + +lanx~lj ~r +... (2)

The terms of this series are less than the corresponding terms
of the series
M+M I!...J+M
Xo l!.../2+
Xo +MI !_In+.
Xo (3)
For IxI< IX0 I the latter series is a geometric progression with
ratio l;o J <I and, consequently, converges. Since the terms of the
series (2) are less than the corresponding terms of the series (3),
the series (2) also converges, and this means _that the series (Ia)
or (I) converges absolutely.
2) It is now easy to prove the second part of the theorem: let
the series (I) diverge at some point x:.
Then it will diverge at
any point x that satisfies the condition IxI> Jx: 1. Indeed, if at
some point x that satisfies this condition the series converged, then
by virtue of the first part (just proved) of the theorem, it shou"ld
x:
converge at the point x: as well, since I I< I xI But this con
tradicts the condition that at the point x:
Hence the series diverges at the point x as well. The theorem is
thus completely proved.
Abel's theorem makes it possible to judge the position of the
points of convergence and divergence of a power series. Indeed,
if X 0 is a point of convergence, then the entire interval (-I x 0 1.
lxo 1) is filled with points of absolute convergence. If x~ is a point
of divergence, then the whole infinite half-line to the right of the
point I x~ I and the whole half-line to the left of the point -I x~l
consist of points of divergence.
From this it may be concluded that there exists a number R
such that for IxI< R we have points of absolute convergence and
for IxI> R, points of divergence.
744 Series
We thus have the following theorem on the structure of the

domain of convergence of a power series:
Theorem 2. The domain of convergence of a power series is an
interval with centre at the coordinate origin.
Definition 2. The interval of convergence of a power series is an
interval from -R to +
R such that for any point x lying inside
series converges
---~
-v::= I
0.
~
~---
)I
Ser(es diverges . Series diverges

Fig. 349.
this interval, the series converges and converges absolutely, while

for points x lying outside it, the series diverges (Fig. 349). The
number R is called the radius of convergence of the. power series.
At the end points of the interval (at x = R and at x = -R) the
question of the convergence or divergence of a given series is
decided separately for each specific series.
We note that in some series the interval of convergence dege-
nerates into a point (R = 0), while in others it encompasses the
entire x-axis (R = oo).
We give a method for determining the radius of convergence of
a power series.
Let there be a series
ao + a,x + a2X 1 + ... + anxn +... (1)
Consider a series made up of the absolute values of its terms:
Iao I+ I af II X I+ I aa II X lz +I a, II X I'+
+I a, I X 1 + +I an II X In+
4
(4)
To determine the convergence of this series (with positive terms!),
apply the d' Alembert test.
Let us assume that there exists a limit:
lim
n-+tD
Un+
Un
=lim I
n-+ao
an+x:+l
anX
I= lim I
n-+CID
an+
an
II X I= L I X I
Then by the d' Alembert test the series (4) converges, if L IxI< 1;
that is, if lxl<i, and diverges ifLixl>1,thatis, if lxl>i.
Consequently, series (1} converges absolutely when lxl<i. But
if Jxl> l, then lim un+'=lxiL> 1 and series (4)
n-+"' Un
diverges, and
Power Series. /nterual of Convergence 745
its general term does not tend to zero.*) But then neither does
the general term of the given power series (I) tend to zero, and
this means that (on the basis of the necessary condition of con-
vergence) this power series diverges (when IxI>~ ) .
From the foregoing it follows that the interval ( - ~ , ~ ) is
the interval of convergence of the power series (1):
1"tm __
R =L-I = 11-+oo a n_
all+l
I I
.
Similarly, to determine the interval of convergence we can make

use of the Cauchy test, and then
R= .
lim
n-+ oo
JY Iani
Example l. To determine the interval of convergence of the series
1 +x+x +x'+ ... +xn+ ...
2
Solution. Applying d'Aiembert's test directly, we get

1
lim
11-+00
lxn:
X
1=1x1.
Thus, the series converges when 1 x 1 < 1 and diverges when lx 1 > l. At
the extremities of the interval ( -1, 1) it is impossible to investigate the
series by means of d' Alembert's test. However, it is immediately apparent
that when X= -1 and when X= 1 the series diverges.
Example 2. Determine the interval of convergence of the series
2x _ (2x) 2 + (2x)' _
r 2 3
Solution. We apply the d' Alembert test:
(2x)n+
lim
n-+oo
n+T 11
00
( 2x)n = I:_m I +n I
n 1 12x I= 12x I
n
1
The series converges if J2x 1 < I. that is, if I xI < ; ; when x=
2 the series
converges; when X=-; the series diverges.
*) It will be recalled that in proving d' Alembert's test (see Sec. 4) we

found that if lim Un+t > 1, then the general term of the series increases and,
Un
consequently, does not tend to zero.
7-16 Series
Example 3. Determine the interval of convergence of the series

x 1 x' xn
x+2i+3i + ... +iii+ ...
Solution. Applying the d' AI em bert test we get
lim
n-+oo
I Un+tl= lim
Un n-+oo
I
X
:n+'nl
(n+l)l
]= n-+oo
lim 1-x-1=0 < 1
n+l '
Since the limit is independent of x and is less than unity, the series con-
verges for all values of x.
Example 4. The series l+x+(2x)'+(3x)'+ ... +(nx)n+ ... diverges for
all values of x except x =0 because (nx)n-+ CIO as n-+ CIO no matter what
the x, as long as it is different from zero.
Theorem 3. The power series

a0 +a,x+a,x'+ ... +anxn+ .. , (1)
is dominated on any interval [ -Q, e] that lies completely inside
the interval of convergence.
Interval of convergence
---- =~
(fltervat of majorisatiOIT;
Fig. 350.
Proof. It is given that e< R (Fig. 350) and therefore the num-
ber series (with positive terms)
laol+la,le+la.IQ'++IaniQn (5)
converge~. But when IxI< Q, the terms of the series
(1) do not
exceed, in absolute value, the corresponding terms of series (5).
Hence, series (1) is dominated on the interval [-e. e].
. 0 --~ )' l
-.R C( 0 J3 R
Fig. 351.
Corollary 1. On every interval lying entirely within tli.e interval

of convergence, the sum of a power series is a continuous function.
Indeed, the series on this interval is majorised, and its terms nre
continuous fund ions of x. Consequently, on the basis of Theorem 1,
Sec. 11, the sum of this series is a continuous function.
Differentiation of Power Series 747
Corollary 2. If the limits of integration a,~ lie within the interval of

convergence of a power series, then the intearal of the sum of the series
is equal to the sum of the integrals of the terms of the series, be-
cause the region of integration may be taken in the interval
[ -Q, Q], where the series is dominated (Fig. 351) (see Theorem 2,
Sec. 12, on the possibility of term-by-term integration of a domi
nated series).
SEC. 14. DIFFERENTIATION OF POWER SERIES
Theorem 1. If a power series

s(x) = ao +a,x+ azX
1
+ aax' + a,x'+ ... + anx" +... (1)
has an interval of convergence ( -R, R), then the series
cp(x) =a,+ 2a 2 x + 3a 1X 2 + ... + nanx"-' + ... , (2)
obtained by termwise differentiation of the series ( 1) has the same
interval of convergence ( -R, R); here,
cp(x)=s'(x), if Jxl<R,
i.e., inside the interval of convergence. the derivative of the sum
of the power series (1) is equal to the sum of the series obtained by
termwise differentiation of the series ( 1).
-R -p 0 X p ~ R X2 Xf .
Fig. 352.
Proof. We shall prove that the series (2) is majorised on any

interval [ -Q, Q] that lies completely within the interval of con-
vergence.
Take a point s
such that Q < < R (Fig. s 352). The series (1)
converges at this point, hence lim ansn = 0; it is therefore possible
n-+o>
to indicate a constant number M such that
If Jxl ,;:;;;;Q, then
where
748 Series
Thus, in absolute value, the terms of the series (2), when

xI.,;;; Q, are less than the terms of a positive number series with
constant terms:
~(1 +2q+3q 2 + ... +nqn-l+ .. . ).
But this latter series converges, as will be evident if we apply
the d' Alembert test:
. nqn-1
Inn ( -I) qn
n-+> n
2 =q< 1.
Hence, the series (2) is majorised on the interval [ -Q, Q], and by
Theorem 2, Sec. 12, its sum is a derivative of the sum of the
given series on the interval [ -Q, Q], i.e.,
<p (x) = s' (x).
Since every interior point of the interval ( -R, R) may be
included in some interval [ -Q, e], it follows that the series (2)
converges at every interior point of the interval ( -R, R).
We shall prove that outside the interval (-R, R) the series (2)
diverges. Assume that the series (2) converges when X1 > R.
Integrating it termwise in the interval (0, X 2 ), where R < X 2 < X1 ,
we would find that the series (1) converges at the point X 1 , but
this contradicts the hypotheses of the theorem. Thus, the interval
(- R, R) is the interval of convergence of series (2). And the
theorem is proved completely.
Series (2) may again be differentiated term by term, and this
may be continued as many times as one pleases. We thus have
the conclusion:
Theorem 2. If a power series converges in an interval ( -R, R),
its sum is a function which has, inside the interval of convergence,
derivatives of any order, each of which is the sum of a series re-
sulting from term-by-term differentiation of the given series an
appropriate number of times; here, the interval of convergence of
each series obtained by differentiation is the same interval
(-R, R).
SEC. 15. SERIES IN POWERS OF x-a
Also called a power series is a functional series of the form

a 0 +a. (x-a)+a 2 (x-a) 2 + ... +an (x-a)n+ .. . , (1)
where the constants a 0 , a 1 , , an, ... are likewise termed coeffi-
cients of the series. This is a power series arranged in powers of
the binomial x-a.
Series in Powers of x-a 749
When a= 0, we have a power series in powers of x, which,

consequently, is a special case of series (1).
To determine the region of convergence of series (1), substitute
the variable
x-a=X.
Series ( l) then takes on the form
a.+a 1 X +a 2 X 2 + ... +anXn+ .. . , (2)
we thus have a power series in powers of X.
Let the interval -R <X< R be the interval of convergence
of the series (2) (Fig. 353, a). It thus follows that series (l) will
converge for values of x that satisfy the inequality -R < x-a < R
or a-R <x<a+R. Since series (2) diverges for I X I >R the
series (l) will diverge for lx-ai>R, that is, it will diverge
outside the interval a-R <x<a+R (Fig. 353, ~).
-R 0 R X
~~--+--+--~-------
________..
.
a-R ---a+R
m+-~~~~~--~
a
x ..X
Fig. 353. Fig. 354.
And so the interval (a- R, a+ R) with centre at the point a

will be the interval of convergence of series (1). All the properties
of a series in powers of x inside the interval of convergence
( -R, + R) are retained completely for a series in powers of
x-a inside the interval of convergence (a-R, a+R). For example,
after term-by-term integration of the power series (1}, if the limits
of integration lie within the interval of convergence (a-R, a+R),
we get a series whose sum is equal to the corresponding integral
of the sum of the given series (1). In the case of termwise diffe-
rentiation of the power series (1), for all x lying inside the inter-
val of convergence (a-R, a+R) we obtain a series whose sum
is equal to the derivative of the sum of the given series (1).
Example. Find the region of convergence of the series
(x-2) +(x-2) 2 +(x-2) 3 + ... +(x-2)n + ...
Solution. Putting x-2=X, we get the series
X+X 2 +X 3 + ... +Xn+ ...
This series converges when - I <X <+I. Hence, the given series converges
for all x that satisfy the inequality -1 < x-2 < 1, that is, when 1 < x < 3
(Fig. 354).
750 Series
SEC. 16. TAYLOR'S SERIES AND MACLAURIN'S SERIES
In Sec. 6, Ch. IV, it was shown that for a function f (x) that
has all derivatives up to the (n +I
)st order inclusive, Taylor's
formula holds in the neighbourhood of the point x =a (that is,
in some interval containing the point x =a):
f (x) = f (a)+ x 1
a f' (a)+
+ (x-;: ;)2 f" (a) + ... + (x -;:la)n [<n> (a)+ Rn (x), (1)
where the so-called remainder term Rn (x) is computed from the

formula
X a)n+
Rn = . ( (n+l)l
- t<n+J) [a+6 (x-a)]
'
0 <6 < 1.
If the function f (x) has derivatives of all orders in the neigh-
bourhood of the point x =a, then in Taylor's formula the number
n may be taken as large as we please. Suppose that in the
neighbourhood under consideration the remainder term Rn tends
to zero as n--+ oo:
Then, passing to the limit in formula (I) as n--+ oo, we get an

infinite series on the right which is called the Taylor series:
f (x) = f() +- 1
a x-af'()
- a -n-- + .. .
+(x-a)nflnl()a 1
(2) + ...
This equation is valid only when Rn (x) --+0 as n--+ oo. Then the
series on the right converges and its sum is equal to the given
function f (x). Let us prove that this is indeed the case:
f (x) = Pn (x) + Rn (x),
where
P ) = '()
n (x
x-af'()+
a + -1-l a
+(x-a)nf1;.1 (a)
-n-1-
Since it is given that lim Rn = 0, we have
f(x) = lim Pn (x).

no+ao
But Pn (x) is the nth partial sum of the series (2); its limit is
equal to the sum of the series on the right sid~ of (2). Hence,
(2) is true:
f (X)= f (a)+ X l a f' (a)+ (X 2 a) f' (a)+ o o + (X-;:Ia)" f'"l (a)+ o, o
Examples of Expansion of Functions in Series 751
From the foregoing it follows that the Taylor series is a given

function f (x) only when limRn=O. If limRn=FO, then the series
is not the given function, although it may converge (to a different
function).
If in the Taylor series we put a= 0, we get a special case of
this series known as Maclaurin's series:
f (x) = f (0) +-f f' (0) +fr f" (0) + ... + :~ rn) (0) +... (3)
If for some function we have a formally written Taylor's
series, then in order to prove that this series is a given function
it is either necessary to prove that the remainder term approaches
zero, or to be convinced in some way that this series converges
to the given function.
We note that for each of the elementary functions defined in
Sec. 8, Ch. I, there exists an a and an R such that in the inter-
val (a-R, a+R) it may be expanded into a Taylor's series or
(if a= 0) into a Maclaurin's series.
SEC. 17. EXAMPLES OF EXPANSION OF FUNCTIONS

IN SERIES
1. Expanding the function f (x) =sin x in a Maclaurin's series.
In Sec. 7, Ch. IV, we obtained the formula
x' x 1
x2n-t
sinx=x- 31 + 51 + ... +<-1r+ (
2n-l)r+R.n.
Since it was proved that lim R2 n = 0, it follows, by what has been
n-+"'
said in the preceding section, that we get an expansion of sin X
in a Maclaurin's series:
x' x' x2n-I
sinx=x- 31 + 51 + ... +(-1)n+J (2n- 1) 1+... (1)
Since the remainder term approaches zero for any x, the given
series converges and, for its sum, has the function sin x for
any x.
Fig. 355 shows the graphs of the function sin x and of the
first three partial sums of the series (1).
This series is used to compute the values of sin x for different
values of x.
To illustrate, Jet us c0mpute sin 10 to the fifth decimal place.
Since 10=~8 =0.174533, we have
7

sm n
10 o =w-31 n )' +51
I ( I8 n )" -71
1 ( 18 n ) + ...
1 ( 18
752 Series
Confining ourselves to the first two terms, we get the following

approximate equality:
:rt :rt l ( :rt )'
smTB:::::IS-6 T8 ;
here, we are inerror by 6, which in absolute value is less than
the first of the suppressed terms; that is,
6<~ (rsY< ~~o<0.2)5 <410-e.
r
I
''\ I
\ I
\
\
\
\
\
\
\ )c
____,,
I"
,
, ,, "
,
,,
I , '"
I ,,
I ,, I'
I
Fig. 355.
If each term in the expression for sin ~8 is computed to six

decimal places, we get
sin ~=0.173647.
We can be sure of the first four decimals.
2. Expanding the function f (x)= e~ in a Maclaurin's series.
On the basis of Sec. 7, Ch. IV, we have
(2)
Euler's Formula 753
since it was proved that lim Rn (x) =0 for any x. Hence, the
n-oo
series converges for all values of x and is the function e".
3. Expanding the function f(x)=cosx in a Maclaurin's series.
From Sec. 7, Ch. IV, we have
x2 x x
cosx = 1-21 + 41-61 + ... ; (3)
for all values of x the series converges and represents the function
cosx.
SEC. 18. EULER'S FORMULA
Up till now we have considered only series with real terms and
have not dealt with series with complex terms. We shall not give
the complete theory of series with complex terms, for this goes
beyond the scope of this text. We shall consider only one important
example in this field. .
In Chapter VII we defined the function e"+'Y by the equation
. ex+iY=e"(cosy-f-isiny).
When x=O, we get Euler's formula:
e1Y = cos y + i sin y.
If we determine the exponential function e1Y with imaginary
exponent by means of formula (2), Sec. 17, which represents the
function e" in the form of a power series, we will get the very
same Euler equation. Indeed, determine e1Y by putting the expres-
sion iy in place of x in equation (2), Sec. 17:
e
iy -1 + iJ!.+(iy)z + (iy)' + + (iy)n + (I)
- 11 21 31 nI
Taking into account that i 2 =-1, i 1 =-i, i 4 =1, i 5 =i, i"=-1,

and so forth, we transform formula ( 1) to the form
i iy y2 iy' y iy 5
eY = 1 +TI-2i-3f -f-41-f-51- ...
Separating in this series the reaJs from the imaginaries, we find
e'Y= ( 1-~+f,- ... ) +i (fl-fr+~- ... ).
The parentheses contain power series whose sums are equal to
cosy and siny, respectively [see formulas (3) and (1) of the pre-
ceding section]. Consequently,
e1Y =cosy+ i sin y.
Thus, we have again arrived at Euler's formula.
754 Series
SEC. 19. THE BINOMIAL SERIES
1. Let us expand the following function in a Maclaurin's series:

f(x) = (1 +x)m,
where m is an arbitrary constant number.
Here the evaluation of the remainder term presents certain dif-
ficulties and so we shall approach the series expansion of this
function somewhat differently.
+
Noting that the function f (x) = (1 x)m satisfies the differential
equation
+
(1 x) f' (x) = mf (x) (1)
and the condition
t (0) = 1,
we find a power series whose sum s (x) satisfies equation (1) and
the condition s (0) = 1:
s(x) = 1 +a1x+a 2 X 2 + ... +anX'+ ... *). (2)
Putting this series into equation (1), we get
+ +
(1 x) (a 1 2a 2 X 3a 1 X 2+ + ... +
nanxn-l + ... )=
=m (1 -t-a 1 x+a.x 2 + ... +anxn+ .. . ).
Equating the coefficients of identical powers of x in different parts
of the equation, we find
al =m; al +2az=ma1; ... ; nan +<n+ l)an+J =man; ...
Whence for the coefficients of the series we get the expressions
a1 (m-1) m (m-1)
ao = 1; a,= m; az = 2 2
az(m-2) _m(m-1)(m-2). .
a,= 3 - 23 1
'
m (m-1) ... [m-n+ 1].

an= l~ .. n I
These are binomial coefficients.

Putting them into formula (2) 1 we obtain
m(m-1) z
s(x)= 1 +mx+ + ..
1. 2 x
... + m (m-1)i _. 2 ~~:-(n-1)) xn + ... (3)
*) We took the absolute term equal to unity by virtue of the initial con
dition s (0)= 1.
The Binomial Series 755
If m is a positive integer, then beginning with the term con-

taining xm+l all coefficients are equal to zero, and the series is
converted into a polynomial. For m fractional or a negative
integer, we have an infinite series.
Let us determine the radius of convergence of series (3):
m (m-1) .. . [m-n+ I) n
Un+t = nl X'
m (m-1) ... [m-n+21 n-t

Un= (n-1)1 X '
lim Un+t I
I= lim m (m-1) .. . (m-n+ I) (n-1)1 X I=
m (m-1) ... (m-n+2) nr
I Un n-+>
= limlm-n+IIJxl=lxJ.
n-+ex> n
Thus, series (3) converges for IxI< 1.
In the interval ( -1, 1), series (3) is a function s (x) that
satisfies the differential equation (1) and the condition
s(O) = 1.
Since the differential equation (1) and the condition s (0) = 1
are satisfied by a unique function, it follows that the sum of the
series (3) is identically equal to the function (1 + x)m, and we
obtain the expansion
(1 + X)
m_
- 1 + mx +m(m-1)
12 X
2 +m(m-1)(m-2)
123 X
'+ .. (3 ')
For the particular case m = -1, we have
1 1-x+x 2-x I+ .. , (4)
1 +x =
(5)
1
For m= - 2 we have
I
Yl+x = 1-21 x+2.4x
13 2 135 3 1357 4
-2+6x +2468x- (6)
2. We apply the binomial expansion to the expansion of other

functions. Expand the following function in a Maclaurin's series:
f (x) =arc sin x.
756 Series
Putting into equation (6) the expression -x 2 in place of x, we

get
1
= 1 +_!_x 2 +~x+
Y1-x' 2 24
+
135 6 + +135 ... (2n-l) zn+
246x 24-6 ... 2n x '
By the theorem of integration of power series we have, for
lxl<1:
X
dx . I x' 13x5 l-35x 7
S
0
Jf1-x2=arcsmx=x+23+2745+2467+ .. .
135 ... (2n-1) xzn+t
' '+ 246 ... 2n 2n+ I + ' ' '
This series converges in the interval ( -1, 1). One could prove
that the series converges for x = 1 as well as that for these
values the sum of the series is likewise equal to arc sin x. Then,
setting x =I, we get a formula for computing n:
1t 1 1 13 1 13G 1
arcsm I =2= I +2 3+2.4 5+2-46' 1+
SEC. 20. EXPANSION OF THE FUNCTION In (t+x)

IN A POWER SERIES. COMPUTING LOGARITHMS
Integrating equation (4), Sec. I9, from 0 to x (when 1x 1< 1),

we obtain
X X
51 ~x= 5(I-x+x -x'+ ... )dx

0 0
1
or
x2 x' x n+ 1 x"
ln(1 +x)=x-2+a-4+ ... +(-1) n+... (1)
This equation holds true in the interval (-1, 1).
If in this formula x is replaced by -x, then we get the series
x2 x' x
ln (1-x) = - x - 2 - 3 - 4 - ... , (2)
which converges in the interval (-1, 1).
Using the series (1) and (2) we can compute the logarithms of
numbers lying between zero and two. We note, without proof,
that for x = 1 the expansion (1) also holds true.
We will now derive a formula for computing the natural loga-
rithms of all integers.
Expansion of the Function In (I +x) in a Power Series 757
Since in the term-by-term subtraction of two convergent series

we get a convergent series (see Sec. 1, Theorem 3), then by sub-
tracting equation (2) from equation (1) term by term, we find
In (1 +x)-ln (1-x) =In 11+x
_x=2 [ x +a+s+
x' x
J
1+x n+1. 1
Now put 1_x = -n-, then X= 2n+ 1 .
For any n > 0 we have 0 < x < 1; therefore
1 +x n+ l [ 1 I
lnl-x= In-n-= 2 2n+ 1 + 3(2n+ I)'+ 5(2n+ 1) 5 + '
1 J
whence
ln.(n + 1)-ln n = 2 bn1+ I+ 3 (2n1+ I)'+ 5 (2nl+ I)'+ ... J. (3)
For n =I we then obtain
l n 2 = 2 [I~ 3 + / 3" + 5 13' + ... J.

To compute In 2 to a given degree of accuracy ~. one has to
compute the partial sum sP, choosing the number p of its terms
such that the sum of the suppressed terms (that is, the error RP
committed when replacing s by sp) is less than the admissible
error ~. To do this, let us evaluate the error RP:
RP = 2 [(2p + ~~ 3P+ 1 + (2p + 3~ 3P+ 3 + (2p + 5\ 32P+ + ]

Since the numbers 2p + 3, 2p + 5, ... are greater than 2p + 1, it
follows that by replacing them by 2p +I we increase each fraction.
Therefore,
Rp < 2 [(2p+ I~ 32P+ 1 + (2p+ I~ 32P+ 3 + (2p+ 1~ 32P+ 5 + ... J,
or
1 [ 1
Rp< 2p+ 1 I J
1 32P+l + 32P++ 321'+ 5 + ... ~
The series . in the brackets is a geometric progression with ratio

!. Computing the sum of this progression we find
1
2 32J'+l 1
Rp< 2p +I - - 1 = (2p+ I) 32P-4. <4>
1-9
If we now want to compute In 2 to, for example, seven decimal
places, we must choose p such that RP < 0.0000001. This can be
done by selecting p so that the right side of inequality (4) is less
758 Series
than 0.0000001. By direct choice we find that it is sufficient to

take p = 8. To seven-decimal accuracy we have
-[1 1 1 1 1
ln2~sa= 2 ~+ 3.3+ 53+ 737 + 93"+ 113"+
I
+ 13~3,.+ 15~3..] = 0.6931471.
Thus, In 2 = 0.6931471. These seven digits are significant digits.
Assuming n = 2 in formula (3), we obtain
In 3 =In 2+ 2 [ { + 3 ~5.+ / J
5, + ... = 1.098612, and so forth.
In this way we obtain the natural logarithms of any integer.
To get the common logarithms of numbers, use the following
relationship (see Sec. 8, Ch. II)
logN=MlnN,
where M = 0.434294. Then, for example, we get In 2 = G.6931472,
log 2 = 0.30103.
SEC. 21. INTEGRATION BY USE OF SERIES

(CALCULATING DEFINITE INTEGRALS)
In Chapters X and XI it was noted that there exist definite

integrals, which, as functions of the superior limit, are not, in
final form, expressible in terms of elementary functions. It is
~ometimes convenient to compute such integrals by means of
series.
Let us consider several examples.
1. Let it be required to compute the integral
a
) e-" 2 dx.
0
Here, the antiderivative of e-x' is not an elementary function.

To evaluate this integral we expand the integrand in a series,
replacing x by -x' in the expansion of e" [see formula (2),
Sec. 17]:
z x1 x4 x x 1n
e-x =1-rr+ 21 - 31 + ... +(-1tnr+ ...
Integrating both sides of this equality from 0 to a, we obtain
Se-x' dx= ( ~ -t:3 +2~~-:r'7+ ) /: =

0
a a a5 a7
= T-m+ 215-317+
Integration by Use of Series 759
Using this equation, we can calculate the given integral to any

degree of accuracy for any a.
2. It is required to evaluate the integral
a
5sl~x dx.
0
Expand the integrand in a series: from the equation
we get
2 4 8
sin x x x x
x-= 1-ST+sr-71+
the latter series converges for all values of x. Integrating term
by term, we obtain
The sum of the series is readily computed to any degree of

accuracy for any a.
3. Evaluate the elliptic integral
n
I
) Vl-k 2 sin 2 <pd<p (k<1).
Expand the integrand in a binomial series, putting m = l ,

2
x=-k 1 sin'<p [see formula (5), Sec. 19]:
,v/1 - kl sm
. I <p= 1 - 1 kl sm
. I <p- I 1 k4 sm
&<p- 1 1 3 kB sm
8<p-...
2 24 246
This series converges for all values of <p and admits term-by-term
integration because it majorises on any interval. Therefore,
 = 2 we have
. nd 13... (2n-l)n
2
2
Ssm q> q> = 24 ... 2n 2

0
(see Sec. 6, Ch. XI) and, hence,

n
2
5v1-k (I ) k - (13) k (135)k

2 2 4 8
2
sm
2 n [ 1- 2
q>dq>=2
2
2.4 3- 246 5 - .... ]
SEC. 22. INTEGRATING DIFFERENTIAL EQUATIONS

BY MEANS OF SERIES
If the integration of a differential equation does not -reduce to

quadratures, one resorts to approximate methods of integrating
the equation. One of these methods is representing the equation
in a Taylor's series; the sum of a finite number of terms of this
series will be approximately equal to the desired particular
solution.
To take an example, let it be required to find the solution of
a second-order differential equation,
y" = F (x, y, y'), (1)
that satisfies the initial conditions
(2)
Suppose that the solution y =f (x) exists and may be given in
the form of a Taylor's series (we will not discuss the conditions
under which this occurs):
2
12 ) f'( Xo } + ...
0 f'( ) (x-X 0
X= f (Xo ) +-1-
X-X
y= f() Xo + (3)
We have to find f (x 0 ), f' (x 0 ), f" (x 0 ), , i.e., the values of the
derivatives of the particular solution when x = X0 But this can
be done by means of equation (1) and conditions (2).
Indeed, from conditions (2) it follows that
f (xo) = Yo f' (x 0) = y:;
from equation (1) we have
f" (xo) = (y")x=xo = F (xo, Yo Y~).
Integrating Di(Jerential Equations by Means of Serie5 761
Differentiating both sides of (1) with respect to x, we get

y '" = F'( y X, y, y ') y '+F'y' (X, y, y ')"
x X, y, y ')+F'( y ; (4)
and substituting the value X=X 0 into the right side, we find
f'" (x.) = (y''');"=xo
Differentiating the relationship (4) once again, we find
flV (Xo) = (yiV)X=Xo
and so on.
We put these values of the derivatives into (3). For those
values of x for which this series converges, this series represents
the solution of the equation.
Example I. Find the solution of the equation
y" = - yx2,
which satisfies the initial conditions
(y)x== 1, (y')x=o=O.
Solution. We have
f (0) =uo =I; f' (0) =0. =u;
From the given equation we find (y")x= = f' (0) = 0; further,
y'" = - y'x 2 -2.r:y, (y'")x=o=f'" (0)=0,
ylV=-xzif-4xy'-2y, (yiV)x=o=-2
and, generally, differentiating k times both sides of the equation by the
Leibniz formula, we find (Sec. 22, Ch. Ill)
y<k+z'=- y<kl x2-2ky<k-u x-k (k-l) y<k-z,,
Putting x=O, we have
or, setting k+2=n,

Y~=- (n-3) (n-2) y~n->.
Whence
u!v =-12, "~ 81 =-56u!v =(-1) 2 (12)(56),
y~l%) = - 910y~8 ) = (- 1)1 (12) (56) (910),
u!k = <- 1lk (1 2) (5 6) (910) ... [(4k -3) (4k -2)1.

In addition,
y~5) =0, y~') =0, ... , y~k+t) =0,
uio) =0, u!> =0, ... , y~k+z)=O,

y~7) = 0, y~ll) = o. y~k+l) = 0.
Thus, only those derivatives whose order is a multiple of four do not

become zero.
762 Series
Putting the values of the derivatives that we have found into a Macla
urin's series, we- get the solution of the equation
~ ~ xz
y=l-4112+8, (12)(56)-12i(l2)(56)(910)+ ..
k x4k
... +(-1) (4k)l (12) (56) ... [(4k-3) (4k-2)] + ...
By means of d' Alembert's test we can verify that this series converges for
all values of x; hence, it is the solution of the equation,
If the equation is linear, it is more convenient to seek the

coefficients of expansion of the particular solution by the method
of undetermined coefficients. To do this, we put the series
y=a 0 +a.x+a2x2 + ... +a,.x"+ ...
into the differential equation and equate the coefficients of
identical powers of x on different sides of the equation.
Example 2. Find the solution of the equation
y"=2xy' +4y
that satisfies the initial conditions
(Y)x=O = Q, (y')x=o = 1.
Solution. We set
y=ao+a 1x+a 2x2 +a,x1 + ... + a,.x" + .. ,
On the basis of the initial conditions we find
a 0 =0, a 1 =1.
Hence,
y =x+a1x2 +a1x'+ ... +a,.x"+ .
y' = 1 +2a 2 x+3a,x 2 + ... +na,.x"- 1 + ..
y =2a 2 +32a 3x+ ... +n (n-1) a,.x"- 2 + ...
Putting these expressions into the given equation and equating the
coefficients of identical powers of x, we obtain
2a 2 =0, whence a 2 =0;
32a,=2+4, whence a,= 1;
43a4 =4a 2 +4a 2 , whence a 4 =0;
2
n (n-1) a,.=(n-2) 2a,._ 2 +4an-z whence an= a 11 - 2

n- 1
Consequently,
Bessel's Equation 763
1
2Vi=T)i 1
a,k+, = 2k kl 1
a,=O; a 1 =0; a,k=O.
Substituting the coefficients which we have found, we get the desired
solution:
The series thus obtained converges for all values of x.

It will be noted that this particular solution may be expressed in terms
ol the elementary functions: taking x outside the brackets we get (inside the
brackets) an expansion of the function e"". Hence,
y=xe"".
SEC. 23. BESSEL'S EQUATION
Bessel's equation is a differential equation of the form

x'y"+xy'+(x'-p 1 )y=0 (p=const). (1)
The solution of this equation (as also of certain other equations
with variable coefficients) should be sought not in the form of
a power series, but in the form of a product of some power of
x by a power series:
(2)
The coefficient a 0 may be considered nonzero due to the

indefiniteness of the exponent r.
We rewrite the expression (2) in the form
and find its derivatives:

"'
y' = ~(r+ k) akxr+k-;
k=o
y" = "'
~(r + k)(r+ k-1) akxr+k-z.
kco
Put these expressions into equation ( 1):
+k) (r + k-1) akxr+k.:..z +

00
x ~(r
k=o
00 00
+x ~(r+k)a~r+k-'+(x'-p') ~akxr+k=O.
k=o k=o
764 Series
Equating to zero the coefficients of x to the. powers r, r + 1,

r + 2, ... , r + k, we get a system of equations:
[r(r-l)+r-p]a 0 =0 or [r 2 -p"]a 0 =0,)
[(r+l)r+(r+l)-p"]a 1 =0 or [(r+l)"-p"]a,=O,
[(r+ 2) (r+ l)+(r+2)-p 2 ] a.+a 0 =0 or [(r+2)"-p"Ja.+a 0 =0, (3)
I
((r+.k)(~+k~l)-t-<~-i-k)~;.J~k+~k~. 'o' ~r . . . . . . . - I
[(r+kY-p"]ak+ak-=0. J
Let us consider the latter equation:
+
[(r k)"- p"] ak +ak- = 0. (3')
It may be rewritten as follows:
[(r+k-p) (r +k + p)] ak +ak- = 0.
It is given that a0 =I= 0; hence,
r-p =0,
therefore, r1 =P or r.=-p.
Let us first consider the solution for r 1 = p > 0.
From the system of equations (3) we determine all the coeffi-
cients aJI a., ... in succession; a 0 remains arbitrary. For instance,
put a0 = 1. Then
ak-
ak=- k (2p+k).
Assigning various values to k, we find

a1 =0, a3 =0 and, generally, am+ 1 =0; }
1 1
a. =-2(2p+2); a"=24(2p+2)(2p+4); .. ; (4)
1)'+' 1
an= ( - 246 ... 2v(2p+2)(2p+4) ... (2p+2v)
Putting the coefficients found into (2), we obtain
Yt =xP [ 1 2(2;~2) +2.4 (2p ;;) (2p+4)

246 (2p + 2) ~;p +4)(2p+6) + ... J. (5)
All the coefficients an will be determined, since for every k the
coefficient of ak in (3),
(r + k) 2
-p\
will be different from zero.
Thus, y. is a particular solution of equation (I).

Let us further establish the conditions under which all the coef-
ficients ak will be determined for the second root r. = - p as well.
This will occur if for any even integral positive k the following
inequalities are fulfilled:
(6)
or
But p=r.; hence,

rz+k=t=r .
Thus, condition (6) is in this case equivalent to the following
r 1 - r 2 =1= k,
where k is a positive even integer. But
r1 =p, r2 =-p,
hence
r 1 - r 2 =2p.
Thus, if p is not equal to an integer, it is possible to write
a second particular solution that is obtained from expression (5)
by substituting -p for p:
_ -p [ x x4
Pz-X I 2(-2p+2i+2.4(-2p+2)(-2p+4)
246(-2p+2)(~ 2p+4)(-2p+6)+ ...

6
J. (5')
. The power series (5) and (5') converge for all values of x; this
is readily found by d' Alembert's test. It is likewise obvious that
y 1 and Yz are linearly independent.*)
The solution y 1 multiplied by a certain constant is called
a Bessel function of the first kind of order p and is designated
by the symbol JP' The solution Yz is denoted by the symbol J -p
*) The linear independence of functions is verified as follows. Consider tha

relation
1 xz x
Y2 =x-ZP -2(-2p+2)+24(-2p+2)(-2p+4)
~ x x
1
- 2(2p+2) +24 (2p+2) (2p+4)
This relation is not constant, since for x -+0 it approaches infinity. Hence
the functions y 1 and y 2 are linearly independent.
766 Series
Thus, for p not equal to an integer, the general solution of

equation (1) has the form
y= C1 Jp+ c.J -p
For instance, when p= ~ the series (5) will have the form
xT [1-E_ + _x_'-
23 24 35
+ .. J = xs
2 46,35 7
=
l [
Vx x x x J
x-3T+sr-7T+ ..
7
This solution multiplied by the constant factor ~ is called y

Bessel's function J 1: we note that the brackets contain a series
2
whose sum is equal to sin x. Hence,
2
-
J 1 (x) = ,
2
vI 1tX
sin x.
In exactly the same way, using formula (5'), we obtain

2
J
-- (x) = ,vI
1
2
1tX
cos x.
The general integral of (1) for p=} is

y = CJ!.... (x) + C J _.!_ (x).
2
2 2
Now let p be an integer which we shall denote by n (n;;:;;.. 0).

The solution of (5) will in this case be meaningful and is the
first particular solution of (1 ).
But the solution of (5') will not be meaningful because one of
the factors of the denominator will become zero upon expansion.
For positive integral p = n the Bessel function J n is determined
by the series (5) multiplied into the constant factor 2}nl (when
n=O we multiply by 1):
Jn(X)= 2nnl
xn[ l-2(2n+2)+24(2n+2)(2n+4)-
xl x'
246(2n+2)~;n+4)(2n+6)+ ]
or
~
Jn(x)= ~vl(n+v)l
(- I)' ( ~
2
)n+n (7)
v=o
It may be shown that the second particular solution should in

this case be sought in the form
ao
K n(x) = J n(x) In x+ x-n ~ b,.x 11
k=o
Putting this expression into (1), we determine the coefficients bk.
The function K n (x), with the coefficients thus determined, mul-
tiplied by a certain constant is called Be~sel's function of the
second kind of order n.
This is the second solution of (1), which with the first one
forms a linearly independent system.
The general integral will be of the form
y =CJ n (x) + C:Kn (x). (8)
We note that
lim Kn (x) = oo,
X -+0
Hence, if we want to consider the final solutions for x = 0, then

we must put C1 = 0 into formula (8).
Example. Find the solution of Bessel's equation, for p =0,
u" +..!..
X
u' +u=O
that satisfies the initial conditions: for x=O,
u=2, u' =0.
Solution. From (7) we find one particular solutiom
~ (- t) (X2 )10 = 1 -
Jo(x)= ...... (vt)
1
(11) 2
(X2 )2+ (21)I (X2 ) -
2
I
(31) 2
(X)'
2 +
V=O
Using this solution, we can write a solution that satisfies the given
initial conditions, namely:
y=2J 0 (x).
Note. If we had to find the general integral of this given equation we
would seek the second particular solution in the form
"'
K 0 (X)=J 0 1nx+ ~bkxk.
k=o
Without giving all the computations, we indicate that the second
particular solution, which we denote by K0 (x), is of the form
K0 (x)=J 0 (x) lnx+~:-( 2 :) 2 ~

( r (I+ ~) +(3:)' ( i r (I+ ~ +~ ) - ...
This function multiplied by some constant factor is called Bessel's function
of the second kind of order zero.
768 Series
. Exercises on Chapter XVI

Write the first several terms of the series according to the given general
term:
(nl)
3. Un = (2n)l .
1 2 3 n
Test the following series for convergence: 6.
2 ;- 22 + 23+ ... + 2n+ ...
I I I I
Ans. Converges. 7. ,r-+ ,r-+ ,r-+ ... + ,r-+... Ans. Diverges.
r 10 r 20 r 30 r !On
3 4 n+ I . I 1
8. 2;- 2 ;- 3 ;- ... ;--n-;- ... Ans. Diverges. 9. vr+vs++
+ V~+6 ;- ... Ans. Diverges. 10.; +( ~ )"+(! r+ ... + (n~1r;- ...

I 2 3 4 n
Ans. Converges. 11. 2 +s + 10 + 17 + ... + n 2 + +... Ans. Diverges.
1
I I I I 1
12. 2+5 + 10 +17+ ... +I ;-na+ ... Ans. Converges.
Test for convergence the following series with given general terms:
1
13. u,.= 113 . Ans. Converges. 14. Un= VIii.
I .
Ans. Diverges. 15. Un= 511 + 1 .
2
Ans. Diverges. 16. Un=~!~ Ans. Diverges. 17. Un= 112 +~n+ 3 . Ans.
18. Un-t = n ~ n. Ans. Diverges. 19.
Converges.
1 Prove the Inequality
I I 1 1 1 I
1 +2+3+. +-;>In (n+ I)> 2+3+ ... + n+l.
20. Is the Leibniz theorem applicable to the series

I I I I 1 I
2-1- 2+ 1 + Vs- I - 3+ 1+ + V~t-1- Vn+ 1+ ?
Ans. It is not applicable because the terms of the series do not decrease
monotonically in absolute value. The series diverges.
How many first terms must be taken in the series so that their sum should
not differ by more than w- of the sum of the corresponding series:
II II 1 I Ill
21. 2- 22 ;- .-~;- ... ;- n - n+ 1 ; - . . . Ans. n=20. 22.
2 2 2 2 - 3 ;- 4 -
1 I I _ a 1 I I 1 I
- ;- ... +n-n+l ;- ... Ans. n-10. 23. 22 - 32 ;- 42 - 52 ;- ... ;- 11 , -
5
1 1 1 1 1 I
-(n+l);- ... Ans. n=IO'. 24 2-2-3+234-2345;- ... ;-ni-
1
- (n+ i)i+ ... Ans. n= 10.
Exercises on Chapter XV/ 769
Find out which of the following series converges absolutely:

l I I I I n+t l A c
25. 1- 3,+ 5, - 7,+ 9,+---+(-) (2n-l)'+ ns. onvergesabso-
lutely. 26. {--{- ~++ f,- ... -1-(-l)n+ ~ ;n+ ... Ans. Converges
I I I I n 1 ,
absolutely. 27. ln 2 -ii13+fii4-fi15+<-1) lnn+ Ans. Converges
conditionally. 28. -I+~- V3+ V4+ .. (~on Vn- + ... Ans. Con
verges conditionally.
Find the sum of the series1 29. 1 .~. 3 + 2 .~. 4 + ... + n (n + 1 ~ (n +2) + "
1
Ans.
4 .
For
what values of x do the following series converge1
? x2 x xn x 2 x x4 '
vO. l-1-2+4+ ... + 2n+ ... Ans. -2<x<2. 31. x- 2,-1-Ji-42"+
+... (- l)n+t x:+ ... Ans. -1 ~X...;; 1. 32. 3x+:3 x 4 4

+3'x"+ ... +3n"xn"+ ...
n
2
I 100x 10,000x I,OOO,OOOx'
Ans. lxl < 3 . 33. l+w+ _ _ +
13 5
. _ +-- Ans.-oo <x < oo.
13 5 7
34. sin X+ 2 sin ~ -1- 4 sin ~ -1- ... -1- zn Sin ;n -1- .. , Ans. -oo<x<co.
X X2 Xn
35. ,r-.+-
1-1-r1
,r-+ ... +
2 +r2 n+
Y n +-- Ans. -1-;;;;;x<l.
6 2k z + 3k a A 7 21 z
3 , X+ 2f X
nk n
3f X -1- .. , -1- ilf X -1- .. . ns, - 00 <X < QO. 3 , X -1- 2i X +
+ 331, X a+
+nl
fi!' X n+ , , , Ans. - e <X <e. 38 X +2'
4f X z-1-(12:-1)2 X"+
61
(nl) 2 n
+ ... + (Zn)l x +... Ans. - 4 < x < 4. 39. Find the sum of the series
x+2x'+ ... +nxn+ . .. (I xI< 1).
Hint. Write the series in the form
X
Ans. (I - x)2.
Determine which of the following series Is majorlsed on the indicated

x x2 xn
intervals: 40. l+p+ 2.+ .. +;J2+-- (0-;;;;;x.;o;;l). Ans. Majorised.
x x x xn
41. I +T+2"+a+ ... +n+ ... (O.;;;;;x.;;;;; 1). Ans. Not majorised,
42.- sinx +sin ,2x+sin ,~+ . . +slnnx+

nz .. LO , 2n). Ans. Majorised.
12 2 3
25- 3388
770 Series
Expanding Functions in Series
43. Expand
10
~x in powers of x and determine the interval of conver-
gence. Ans. The series converges for - 10 < x < 10.
44. Expand cosx in powers of ( x- ~ ) . Ans. ~ 2 - ) 2 ( x- :) -
~2 (X- : ) + 6 ~2 (X- ~ ) + ',

2 3
- 2
x x 3
45. Expand e-x in powers of x. Ans. 1-x+2f+3f+
e 2
46. Expand eX in powers of (x-2). Ans. e'+e'(x-2)+
21 (x-2) +
e
+ 3T (x- 2) + ,,, 1
_47. Expand x3 -2x'+5x-7 in powers of (x-1). Ans. -3+4(x-1) +

+ (x-1) 2 +(x-1) 3
48. Expand the polynomial x10 +2x 8 -3x1 -6x 8 +3x'+6x 3 -x-2 in a Tay-
lor's series in powers of (x-1); check to see that this polynomial has the
number I for a triple root. Ans, f(x)=IH (x-1) 3 +270(x-1) 4 +342(x-1) 5 +
+ 330 (x-1) 8 + 186 (x-1) 1 +63 (x-1) 8 + 12 (x-1) 8 +(x-1) 10
49. Expand cos(x+a) in powers of x. Ans. cosa-xsina-~ cosa+
' .
x3 x4
+ 3f sm a+ 4T cos a- ...
50. Expand lnx in powers of (x-1). Ans, ~x-1)- ~ (x-1) 2 +; (x-1) 3 -
- { (x-1) 4 + ...
co
51. Expand ex in a series of powers of (x+2). Ans. e- [1 +~ (x:12>n].

52. Expand cos x in a series of powers of ( x- ~).
I i1 n 4n-t (x--1- rn-t .
A ns. "2+""-1(-1) (2n- 1)! (lxl<oo).
n;t
53. Expand~ in a series of powers of(x+l). Ans._~(n+1)(x+l)n

n;o
(- 2 <x < 0).
54. Expand tan x in a series of powers of ( x- : ) .
+2(x-~)"+ ...
Write the first four term~ of the series expansion, in powers of x,
x 1 2x 5 17x1
of the following functions: 55. tan x. Ans. x+3+15+ 315 + ...
Exercises on Chapter XV I 771
0
x 4x' 3Jx
56. ecos x. Ans. e ( 1 - .21 + 41- 720 - . . . )
arc tan x x X 8 1x'
57. e Ans. 1-1-x+z-+6-24+
x x 1x'
58. ln(1+ex). Ans. Jn2+-z-s+ + ...
384
x2 x
59. estnx. Ans. 1-1-x+-z-s+
60. (1 +x)x. Ans. 1 +x'-~ +{ x- ...

x 5x
61. secx. Ans. 1-1-z-+ + ...
24
x x x
1!2. In cosx. Ans. - - - - 2 12 45
. . (kx) 3 (kx) 5 (kx)7
63. Expand s1n kx tn powers of x. Ans. kx----ar +51-71+ ...
64. Expand sin' x in powers of x and determine the interval of conver
2x 2ax 2x 2n-Ixtn -
gence, Ans.
21 - 41 + 61 - ... +(-1)n-I~+ ... The series conver~
ges for all values of x.
65, Expand ~x 2 in a series In powers of x. Ans. 1-x"+x'-x'+ ...
1
66, Expand arc tan x in a series in powers of x.
X
Hint. Take advantag~ of the formula arc tan X= S1 ~x. Ans. x- ~ +

0
+5x - 1x + ...
7
(-l.,.;;x..:;l).
67. Expand {I~ x)' In a series of powers of x. Ans. 1-2x + 3x2 -4x 8 + ...
(..;..l<x<l).
Using the formulas for expansion of the functions eX, sin x, cos x, In (I +x)
and (I +x)m into power series and applying various procedures, expanq the
following functions in power series and determine the intervals of convergence!
~ ~ ~
68. sinhx. Ans. x+3f+5f+ .. (-oo<x<oo).69.coshx. Ans.i-1-2(+
x I ~"' (-l)n (2x) 2 n

+ 41 + ... (-oo<x<oo). 70. cosx. Ans. 1+2.t-.' (2n)l
n=I
~"' xn
(-oo<x<oo). 71. (1-1-x) In (1-1-x). Ans. x+.t-.<-lln(n-J)n (lxJ.o;;;;;l).
n=z
"' 1
72. (i-1-x)e-x. Ans. I+}.:<-l)n-Innll xn (-oo <x<oo). 73. x
4
ll=&
25*
Series
~ "' x4" ,r- . ex-1 . x x x"- 1

Ans . ..:.,. 4n+ (lxj<r 2). 74 x Ans. 1+2f+3f++nr+
n=o
n~ (n+l)x"
0
1
(-oo<x<oo). 75. (I x)'' Ans. (lxl<l). 76. eXsinx.
3 5
2x 4x nn x"
Ans. x+x 2 +
31 - 51 + ... + Y2-;;
xsm --;rnr+... (- oo < x < oo).
5
_..!._ x +~x +
3
77 ' X + ,r + .2
y 1 X'
A
ns. X 2 3 24 5 '''
+(-!)"+ 1 13 ... (2n-l)
2"nl X
X
x"+
x2n+l+ ... (-l.;;;;x.;;;; 1). 78. sIn(~+x) dx. Ans. :E (-1)"+1 :.
"' n
0 n=l
IX I,.;;;; I). 79. s X
arc :an x dx. Ans. L

CXl
x"+
(-I)" (2n+ 1) 2 (-1o;;;;xo;;;;l).
.,
,
0 n=o
80. s co; x dx. Ans. C +In [xI+

n=t
"'
L (-I)" ( 2n~ (2n)l (- oo < x < 0 and
X "'
dx
0 < x < oo). 81. S l-x . Ans. L 9xn-s
n_ 8 . 82. Prove the equations
o n=J
sin (a +x) =sin a cos x +cos a sinx,
cos (a+x) =cos a cos x:-sin a sin x
by expanding the left sides in powers of x.
Utilising appropriate series, compute: 83. cos 10 to four decimals.
Ans. 0.9848. 84. sin 1 to four decimals. Ans. 0.0175.
85. sin 18 to three decimals. Ans. 0.309. 86. sin : to four decimals.
1 .
Ans. 0.7071. 87. arctan to four decimals. Ans. 0.1973. 88. In 5 to three
5
decimals. Ans. 1.609. 89. Iog10 5 to three decimals. Ans. 0.699. 90. arc sin 1 to
within 0.0001. Ans. 1.5708. 91. Veto
within 0.0001. Ans. 1.6487. 92. loge to
within 0.00001. Ans. 0.43429. 93. cos 1 to within 0.00001. Ans. 0.5403.
Using a Maclaurin series expansion of the function f (x) = a" +x, J1
compute to within 0.001: 94. V30. Ans. 3.107. 95. y'7o, Ans. 4.121.
96. V-
500. Ans. 7 .937. 97.
Ans. 1.2598.
v-
250. Ans. 3.017. 98. ,r-
r 84. Ans. 9.165. 99. 2. v-
Expanding the integrand in a series, compute the integralsi
1 1
100. Ssi~ x dx to five decimal places. Ans. 0.94608. 101. Se -x dx to four

0 0
n
..
decimals. Ans. 0.7468, 102. S sin (x 2
) dx to four decimals, Ans. 0.1571,
Exercises on Chapter XVI 773
z so.s arc tan x

103. ~ eVxdx to two decimals. Ans .. 0.81. 104. x dx to three deci-
o
1
mal places. Ans. 0.487. 105. ~cos Vxdx to within 0.001. Ans. 0.764.
0
4 1 ~
106. ~In (1 + y'X) dx to within 0.001. Ans. 0.071. 107. Se -4dx to within
0
1
slnx
0.0001. Ans. 0.9226. 108. SV!-x dx to within 0.0001. Ans. 0.0214.
0
o.s 1
dx In (1 +x) n2
109.
S
0
1
+x1 to within 0.001. Ans. 0.494. 110.
S
0
x dx. Ans. 12
Note. When solving this exercise and the two following ones it is well to
~ ~ ~
1 nz (-l)n-1 nz 1 nz
bear in mind the ~quations: 2 =6 nz Ln 12 2 ,L.
1)z=s L( n-
n=1 n= n=l
which will be established in Sec. 2, Ch. XVII.
1
ln(l-x) n2
Hl.
S
0
1
x dx. Ans. 6
1 +x dx n2
112.
S In 1 _x x. Ans. 4.
0
Integrating Differential Equations by Means of Series

113. Find the solution of the equation y" =XY that satisfies the j.nitial
conditions for x=O, y= I, y' =0.
x
Hint. Look for the solution in the form of a series. Ans. l +w +
x8 xk
+ 2356+ ... +2356 ... (3k-1) 3k + ...
114. Find the solution of the equation y" +xy' + y=O that satisfies the
x x5
x~3+ . .
initial conditions for x=O, y=O, y'=1. Ans.
1 3 5-
(-l)n+xzn-
'135 ... (2n-1)
115. Find the general solution of the equation
xV +xy' + ( x 1 -!) y=O.
Hint. Seek the solution in the form
y=xP (A 0 +Al't+ A2x 2 + ... ).
774 Series
initial conditions for x=O, y=l. y'=O.
-
+ ... {- 1)k --k+
xzk
(k!} 2
...
2 2
Hint. The two latter differential equations are particular cases of the
Bessel equation
1
for n= and n=O.
2
117. Find the general solution of the equation
4xy" +2y' +u=O.
Hint. Seek the solution in the form o_f a series xP(a0 +a,x+a2x 2 + ... ).
Ans. c, cos Vx+c. sln yX.
118. Find the solution of the equation (1-x 2) y" -xy' =0 that satisfies
1x3 13x5
the initial conditions y'=1 when x=O and y=O. Ans. x+23+245+
1 3 5 x7
+2467+
119. Find the solution of the equation (1 +x") y" +2xy' =0 that satisfies
the initial conditions y' = 1 when x=O and y=O. Ans. x-~ +~-~+ ...
120. Find the solution of the equation y"=xyy' that satisfies the initial
x 3 2x" 3x5
conditions y'=1 when x=O and y=l. Ans. 1+x+3f+4T+5f+
121. Find the solution of the equation {1-x) y' = 1 + x-y that satisfies
the initial conditions y=O when x=O, and indicate the interval of conver
x2 x x
gence of the series obtained. Ans. x+T:2+z:a + . + .. . (-1-=;;;x.;;;:; 1).
3 4
122. Find the solution of the equation xy" +u=O that satisfies the initial
conditions y' = 1 when x = 0 and y= 0, and indicate the interval of conver-
x x3 x
gtnce. Ans. x-( 11)2 2 +( 2!)' 3 -{3 !) 4 + ... (-oo<x<oo).
123. Find the solution of the equation y" +~ y' +y=O that satisfies the
- X
initial conditions y'=l when x=O, y=l. Ans. sinx.

X
124. Find the solution of the equation y" +_!_ y' +y=O that satisfies the
X
initial conditions y'=O when x=O and y=l, and indicate the interval of
. . x2 x x
convergence of the senes obtamed, Ans. 1-21+ 42 - 22 . 42 . 62 +
2
(I xI< oo).
Exercises on Chapter XV I 775
Find the first three terms of the expansion in a power series of the solu-
tions of the following differential equations for the given initial conditions
4
~ +
1
125. y'=x"+y"; for x=O, y=l. Ans. 1+x+x"+ 1.26. y=

ex x , . . .
= eY +x; for x=O, y= 1, y' =0. Ans. 1 +T +
6 + ... 127. y =smy-smx,
x x 3
for x=O, y=O. Ans. -2- 6 -
Find several terms of the series expansion of solutions of differential equa-
tions under the indicated initial conditions: 128. !I'= yy' -x when x =0,
x 2x Bx' 14x5
y=O and y' =0. Ans. I +x+2f+3!+4r+5r+... 129. y' =y"+x
I A 1 I I 1 I 9 + 130
h
Wen X= 0 an d y=<j ns. z+-:rx+BX +f6x +32x ... Y =
I
=X -y 2 h
wen x= 0 A ns. 31 x I - 1. x7+ 2 . x11 - ...
an d Y= 0 .
7 9 7 11 27
x1 x r
131. y'=x!y'-1 when x=O and y=l. Ans. l-x+3-2+5-
x 2x1 11x
132. y'=eY+xy when x=O and y=O, Ans. x+2+3+ . . + ...
2 3 4
C H A PTE R XVII
FOURIER SERIES
SEC. t. DEFINITION. STATEMENT Of THE PROBLEM
A functional series of the form
~+a. cosx+ b 1 sin x-j-a 2 cos 2x+ b 1 sin 2x+ ... ,

or, more compactly, a series of the form
~+ L"' (an cos nx + b n sin nx),

n=t
( 1)
is called a trigonometric series. The constants a 0 , an and bn

(n = 1, 2, ... ) are called coefficients of the" trigonometric series.
If series (1) converges, then its sum is a periodic function f (x)
with a period 2n, since sin nx and cos nx are periodic functions
with period 2n.
Thus,
f(x)= f (x+ 2n).
Let us pose the following problem.
Given a function f (x) which is periodic and has a period 2n.
Under what conditions for f(x) is it possible to find a trigonomet-
ric series convergent to the giuen function?
That is the problem that we shall solve in this chapter.
Determining the coefficients of a series from Fourier's formulas.
Let the periodic function f (x) with period 2n be such that it may
be represented as a trigonometric series convergent to a given
function in the interval (- n, n); i. e., that it is the sum of this
series:
"'
f (x) = ~ + L (an cos nx+ bn sin nx). (2)
n==l
Suppose that the integral of the function on the left-hand side
of this equation is equal to the sum of the integrals of the terms
of the series (2). This will be the case, for example, if we assume
that the numerical series made up of the coefficients of the given
Definition. Statement of the Problem 777
trigonometric series converges absolutely; that is, that the follow-

ing positive number series converges:
I~ I+ I al I+ I b I+ Ia2l + I b:I+ +I an I+ I b n I+
1 (3)
Then series (1) is majorised and, consequently, it may be inte-
grated termwise in the interval from - n to n. Let us take ad van-.
tage of this for computing the coefficient a 0
Integrate both sides of (2) from -n to +n:
1t 1t "' 1t 1t
St<x)dx= s~dx+.E (Sancosnxdx+ 5bnsinnxdx).

-n -n n=l -n -n
Evaluate separately each integral on the right side:
n
5 ~ dx= na 0 ;
-n
n n
5
an cos nx dx =an 5cos nx dx =an sin nx In-n =
11
0;
-n -n
Sbn sin nx dx= bn Ssin nxdx= bn co: nx~~n =0.

-n -n
Consequently,
n
~ f(x) dx=na 0 ,
-.rc
whence
1t
ao= ~ 5 f(x)dx. (4)

-.rc
To calculate the other coefficients of the series we shall need
certain definite integrals, which we will consider first.
If n and k are integers, then we have the following equations:
+
H n k, then
n )'
~cos nx cos kxdx=O;
-n
n
~ cos nx sin kx dx = 0; (I)
-n
n
~ sin nx sin kx dx = 0;
-n
778 Fourier Series
but if n = k, then
n
~cos' kx dx ~ n; l
)
-~sin kx cos kxdx = 0; J (II)
~ sin 2 kxdx=n.
-n
To take an .example, evaluate the first integral of group (I).
Since
cos nx cos kx-:;= 2 [cos (n + k) x+ cos (n-k) x],
I
it follows that
n :n :n
Scosnxcoskxdx={- Scos(n+k)xdx+{ 5cos(n-k)xdx=O.
-n. -n -n
The other formulas of (I)*) are obtained in similar fashion. The
integrals of group (I I) are computed directly (see Ch. X).
Now we can compute the coefficients ak and bk of series (2).
To find the coefficient ak for some definite value k =I= 0, mul-
tiply both sides of (2) by coskx:
f (x) cos kx~ ~cos kx + L (an cos nx cos kx + bncos nx cos kx). (2')
n=l
The resulting series on the right may be majorised, since its terms
do not exceed (in absolute value) the terms of the convergent
positive series (3). We can therefore integrate it termwise on any
interval.
Integrate (2') from - n to n:
:n n
Sf (x) cos kx dx = a2 Scos kx dx + _
-:n -:n
"' :n :n
+L(an -:n~ cosnxcoskxdx+bn -:n~ sinnxcoskxdx).
n=t
*} By means of the formulas

cos nx sin kx= 1 / 1 [sin (n+ k) x-sin (n-k) x],
sin nx sin kx= 1/ 1 [-cos (n+k)x+cos (n-k) x].
Definition. Statement of the Problem 779
Taking into account formulas (II) and (1), we see that all the
integrals on the right are equal to zero, with the exception of the
integral with coefficient ak. Hence,
n n
~ f (x) cos kx dx = ak ~ cos 2 kx dx = akrr.,
-n -n
whence
n
ak = ~ 5f (x) cos kx dx. (5)
-n
Multiplying both sides of {2) by sin kx and again integrating
from - rr. to rr., we find
n n
~ f (x) sin kx dx = b k ~ sin 2 kx dx = bkrr.,
-n -n
whence
n
bk=-k- Sf (x) sin kx dx. (6)

-n
The coefficients determined from formulas (4), (5) and (6) are
called Fourier coefficients of the function f (x), and the trigono-
metric series (1) with such coefficients is called a Fourier series
of the function f (x).
Let us now revert to the question posed at the beginning of
this section: What properties must a function have so that the
Fourier series constructed ~or it should converge and so that the
sum of the constructed Fourier series should equal the values of
the given function at corresponding points? We shall here state
a theorem that will yield sufficient conditions for representing
a function f (x) by a Fourier series.
Definition. A function f (x) is called piecewise monotonic on the
interval [a, b] if this interval may be divided by a finite number
'of points xl' X 2 , , xn_, into subintervals (a, x,), (x,, x,), ... ,
(xn-P b) such that the function is monotonic (that is, either nonin-
creasing or nondecreasing) on each of the subintervals.
From the definition it follows that if the function f (x) is piece-
wise monotonic and bounded on the interval [a, b], then it can
have only discontinuities of the first kind. Indeed, if X=C is
a point of discontinuity of the function f (x), then by virtue of the
monotonicity of the functiort there exist the limits
lim f(x)=f(c-0), lim f(x)=f(c+O),
X-+C-0 X-+G+O
i. e., the point c is a discontinuity of the first kind (Fig. 356).

780 Fourier Series
We now state the following theorem.

Theorem. If a periodic function f (x) with period 211: is piecewise
monotonic and bounded on the interval [ -11:, 11:], then the Fourier
series constructed for this function converges at all points. The sum
of the resultant series . s (x) is equal to the value of f (x) at the
y discontinuities of the function. At the
discontinuities of f (x), the sum of the series
is equal to the arithmetical mean of the
limits of f(x) on the right and on the left;
f(c+O) that is, if x = c is a discontinuity of the
f(c-0) function f(x), then
_ f (c-O)+ f (c+O)
s (X )X=C- 2
0 c X From this theorem it follows that the
Fig. 356. class of functions that may be represented
by Fourier series is rather broad. That is why
Fourier series have found extensive applications in various divi-
sions of mathematics. Particularly effective use is made of Fourier
series in mathematical physics and its applications to specific
problems of mechanics and physics (see Ch. XVIII).
We give this theorem without proof. In Sees. 8-10 we will
prove another sufficient condition for the expandability of a func-
tion in a Fourier series, which condition in a certain sense deals
with a narrower class of functions.
SEC 2. EXPANSIONS OF FUNCTIONS IN FOURIER SERIES

The following are some instances of the expansion of functions in Fourier
~eries.
Example 1. A periodic function f (x) with period 2n is defined as follows:
f(x)=x, -n<x~l't.
This function is piecewise monotonic and bounded (Fig. 357). Hence, it
admits expansion in a Fourier series.
By formula (4), Sec. l, we find
:rl
a0 =_!_
l't
S xdx=..!_::_2'j:n-:n =0.
l't
-n
..
~H ~Vlo V2H ~ X
Fig. 357.
Expansions of Functions in Fourier Series 731
Applying formula (5), Sec. 1,. and integrating by parts, we find
I
ak=n Jt xcoskxdx=n1 [ x-k-
sin kx In-n - k
1 s:rt smkxdx
. ] =0.
-n -n
By formula (6), Sec. 1, we have
bk=n I sn xslnkxdx=-n - kkx

I [ - x cos - 'n-n +k
I sn coskxdx ] =(-l)k+l k"
2
-n -n
Thus, we get the series
f( X ) = 2 [sinx_sin2x+sln3x_
1 2 3
(-l)k+l sinkx +
k
J
This equation occurs at all points except points of discontinuity. At each
discontinuity, the sum of the series is equal to the arithmetical mean of its
limits on the right and left, which is zero.
Example 2. A periodic function f (x) with period 2n is dt>fined as follows
f (x)= -x when -l"t ,.;;;;x,.;;;;O,
f(x)=x when O<x,.;;;;n
[or f(x)=l xiJ (Flg. 358). This function is also piecewise monotonic and bound
ed on the interval -n,.;;;;x,.;;;;n.
y
Fig. 358.
Let us determine its Fourier coefficients:
"o~! /{x)dx~! [f.(-x)dx+ ~ xdx]~,

ak = -k-[ S(-x) cos kx dx +fx coskxdx] =
-n o
=n-I [
x sin kx o
---k-1-n +k
I
S
0 x sin kx n I
sin kx dx+--k-io -k 5sinkxdx
:rt ]
=
-n o
= nk
J_ [ _ cos kx 'o +cos kx\n] =
k -n k o
2 { 0 for k even,
=k1 (coskn-1)= 4 f .
1t - nk 1 or k odd,
782 Fourier Series
bk=-k-[ S( -x) slnkxdx+ Sxslnkx dx] =0.

-n o
We thus obtain the series
f( X) =~-_i_
2- l't
[~+cos3x+cos5x"
12 32 52 " " " '
+cos(2p+l)x+
(2p+l)' ....
J
This series converges at all points, and its sum Is equal to the given function.
Example 3. A periodic function f (XI with period 2n is defined as follows!
f(x)= - I for -n<x<O,
I (x) = l for 0 oe;;;; x..,;;; n.
This function (Fig. 359) Is piecewise monotonic and bounded on the interval
-n...;;x..o;;n.
y
Fig. 359.
Let us compute Its Fourier coefficients!
~ Sf(x) dx = -ft [S(-1) dx + 5dx J= 0;

a0 =
-n n o
ak=-;t
- 1 o n J sin kx sin kx n o
[ S (-l)coskxdx+Scoskxdx =-1-k-1-n+-k-1.=0;
-n o
. J=n-I
bk=nI [ so (-l)sinkxdx+ sn slnkx.dx [-cos
-k-.kx ,.-n - -~X In] =
cos
k-
0
-n o
2 { 0 for k even,
= nk [I-eos n1 = :k for k odd.
Consequently, lor the function at hand the Fourier series has the form
_ _i_ [sinx+sin3x+sln5x+ +sln(2p+l)x+ ]
f (X) - n I 3 5 '.' 2p + I ... .
This equation holds at ali points with the exception of discontinuities.
Fig. 360 illustrates how the partial sums Sn of the series represent more
and more accurately the function f(x) as n--+oo.
Example 4. A periodic function f (x) with period 2n is defined as follows:
f (x) = x', -n.;;;;; x,.;;;; n (Fig. 361).
Expansions of Functions in Fourier Series 783
Fig. 360.
Determine its Fouri'er coefficients.
ak =n-1 -nn Xs 2 I [ x sin kx 2 n

2
cos kx dx=n --k-- -n- k -nx sin kxdx =

In s J
2 [
=- nk X cos kx In I sn
- - - k - -n + k -n cos kx dx
J= 4
nk 2 [:rt cos k:rtj =
= { k: 2 for k even,
- k2 for k odd;
1 Sn
bk=n -n x sin kxdx= +-;t
2 1 [
- x
2
cos kx I:TI
k
2 Sn
-n +k -n x cos kx dx
J
=
2
= nk
[x-sinkkx I:TI-n- k1 ,S:rt sin kx dx
J = 0.
-:rt
'184 Fourie Series
Thus, the Fourier series of the given function has the form
X
2 = .n' _
3 4 (cosI x _ cos2
22
x + cos3 3x _
2 .
)
.
Since the function is piecewise monotonic, bounded anct continuous, this equa-
tion is fulfilled at all points.
g
Ftg. 361.
Putting x = .1t in the equality obfained, we ge~
Example 5. A periodic function f (x) with period' 2.n Is defined as follows!

f (x) =0 for -.n <; x <:0,
f (x) =Ai lor O<x <: .1t (Fig. 362).
-!/rr -4rr -3rr -2rr -rr
Fig .'162.
Determine the Fourier coefficients
ak=-;t ~ xcoskxdAi=n
1 It I [ x sin kx n
--k-lo-k
1
Jslnkxdx
It ]
=
= ~k cosk kx In= {- .n~' for k odd,

u 0 for k evem
A Remark on the Expansion of a Periodic Function in a Fourier Series 7 ES
1
bk=n
sn xsmkxdx=n
. 1 [ kkx
cos
- -X - - In.+k1 sn coskxdx ] =
0 0
..!.. for k odd,

=- :k cos krr. =
{ - k ~
for k even.
The Fourier series will thus have the form
f< X ) =~-2
4 1t
(~+cos3x+cos5x+
12 32 52
)+(sinx_sln2x
I 2 + sin3x_
3
)
. ..
At the discontinuities of the function f (x), the sum of the series is equal to
the arithmetical mean of its limits on the right and left (in this case, to the
number ~),
Putting x=O in the equality obtained, we get
rr.' ~
.. I
s= n=J
~ <2n-1) 2

SEC. 3. REMARK ON THE EXPANSION OF A PERIODIC

FUNCTION IN A FOURIER SERIES
We note the following property .of a periodic function ' (x}

with period 2n:
n A+ln
) ' (x) dx = ) ' (x) dx,
-n i\
no matter what the number 'A..

Indeed, since
'(-2n)='P()
it follows that, putting x= -2n, we can write (for all c and d)l
d d+an d+zn d+zn
) ' (x) dx = . S ' (~- 2n) d~ = S ' (~) d~ = S ' (x) dx.
c c+zn c+an c+an
In particular, taking c= -n, d='J.., we get

A A+zn
~ '(x) dx= ~. '(x) dx,
-n n
786 Fourier Series
therefore,
~+tn -n n ~+m
~ 'iJ (x) dx = ~ 'iJ (x) dx + S 'iJ Cx) dx + ~ 1Jl (x) dx =

~ ~ -n n
-n n ~ n
= S 1Jl (x) dx +
~
51Jl (x) dx + S 1Jl (x) dx = S 1Jl (x) dx.
-n -n -n
This property means that the integral of a periodic function 1Jl (x)
over any interval whose length is equal to . the period always has
the same value. This fact is readily illustrated geometrically: the
cross-hatched areas in Fig. 363 are equal.
Fig. 363.
From the property that has been proved it follows that when
computing Fourier coefficients we can replace the interval of in-
tegration (-n, n) by the interval of integration (A., A.+2n), that
is, we can put
~+m ~+n \
ao = ! S f(x) dx,
A
a=-
n 3t
I
~ /(x)cos nxdx, f
A+2n (1)
bn = ! S f(x) sin nx dx,
A J
where A. is any number.
This follows from the fact that the function f(x) is, by hypothe-
sis, periodic with period 2n; hence, both the functions f (x) cos nx
and f (x) sin nx are periodic functions with period 2n. We now
illustrate how this property simplifies the process of finding coef-
ficients in certain cases.
Example. Let it be required lo expand in a Fourier series the function
f (x) with period 211:, which is given on the interval 0 ..-.;;; x ..-.;;; 211: by the equation
f(x)=x.
The graph of f (x) is shown in Fig. 364. On the interval [ -11:, 11:] this func-
tion is represented by two formulas: /(x)=x+23t. on the interval [-11:, Oj
and f (x) =x on the interval [0, 11:]. Yet, on [0, 2 11:] it is far more simply
Fourier Series for Even and Odd Functions 787
represented by a single formula f (x) = x. Therefore, to expand ~his function

in a Fourier series it is better to make use of formulas (1), settmg A-=0:
Ut 211:
a0 = ~ 5f (x) dx = ~ Sx dx = 2n;
0 0
an=...!._
1t
5f
211:
0
(X) COS nx dx = __!_
1t
s
211:
0
X COS nX dx = __!_ [X sin nx
1t n
+COS
n2
nX] 21t = Q
o '
1
bn=-
n
s2!1
1
{(x)sinnxdx=-
1t
s.
231
xsmnxdx=-
1t
x cos nx sin nx]
1 [ ----+-
n
-
n2 o
2n 2,
=--
n
0 0
Consequently,
f (x)=n-2 sinx- 22 sin 2x- 2 sin 3x- 2 sm4x-
. 2
sin 5x- ...
3 4 5
y
3Tr 71( 8K X
Fig. 364.
This series yields the given function at all points with the exception of points
of discontinuity (i.e., except t_he points x=O, 2n, 4n, ... ). At these points the
sum of the series is equal to the half sum of the limiting values of the
function f (x) on the right and on the left (to the number n, in this case).
SEC. 4. FOURIER SERIES FOR EVEN AND ODD FUNCTIONS
From the definition of an even and odd function it follows that

if 'ljJ (x) is an even function, then
n n
~ 'ljl(x)dx=2 ~ 'ljl(x)dx.
-n o
Indeed,
1t 0 1t 1t 1t
~ 'ljJ (x) dx = ~ 'ljJ (x) dx + ~ 'ljJ (x) dx = ~ 'ljJ ( -x) dx + ~ 'ljJ (x) dx =
-n -n o o o
n n n
= ~ 'ljJ (x) dx + ~ 'ljJ (x) dx = 2 ~ 'ljJ (x) dx.
0 0 0
since by the definition of an even functfon 'ljJ ( -x) = 'ljJ (x).

788 FQurier Series
It may similarly be proved that if c:p (x) is an odd function,

then
:n; :n; :rt :rt :rt
~ c:p(x) dx= ~ cp(-x)dx + ~ cp(x)dx=- ~ c:p(x)dx+ ~ c:p(x)dx=O.

-n. 0 0 0 0
If an odd function f (x) is expanded in a Fourier series, then

the product f (x) cos kx is also an odd function, while f (x) sin kx
. is an even function; hence,
:rt )
~
lI
ao= S f(x)dx=O;
-:rt
:rt
ak = ! S f (x) cos kx dx = 0,
-:rt
(1)
bk= ~ _l
:rt
f(x) sin kxdx= ~ ~ f(x) sin kxdx.
:rt
)
Thus the Fourier series or an odd function contains "only sines"

If an even function is expanded in a Fourier series, the pro-
duct f (x) sin kx is an odd function, while f (x) cos kx is an even
function and, hence,
:rt )
~
lI
ao = Sf (x) dx,
0
:rt
ak = ~ Sf (x)cos kx dx, (2)

0
:rt
bk=! S f(x)sinkxdx=O.
-:rt ]
Thus, the Fourier series of an even function contains "only cosines"
The formulas obtained permit simplifying computations when
seeking Fourier coefficients in cases when the given function is
even or odd. It is obvious that not every periodic function is
even or odd (see Example 5, Sec. 2).
Example. Let it be required to expand in a Fourier series the even
function f (x) which has a period of 2l't and on the interval [0, l't) is given by
the equation
y=x,
The Fourier Series for a Function with Period 21 789
We have already expanded this function in a Fourier series in Example 2,

Sec. 2. (Fig. 358). Let us again compute the Fourier series of this function,
.taking advantage of the fact that the given function is even.
By virtue of formulas (2) bk = 0 for any k;
n n
a0 =: Sx dx= n, ak=: Sx cos kxdx=
0 0
2 [ k ] n 2 { 0 for k even,
= - x.slnkx +~ = - [(- 1)k_ 11 = 4
n k k2 o nk 2 - n;k' fork odd.
We obtained the same coefficients as in Example 2, Sec. 2, but this time by

a short cut.
SEC. 5. THE FOURIER SERIES FOR A FUNCTION

WITH PERIOD 21
Let f(x) be a periodic function with period 21, generally

speaking, different from 2n. Expand it in a Fourier series.
Make a substitution by the formula
X=.!_t,
1t
Then the function f ( ~ t) will be a periodic function of t with

period 2n.
It may be expanded in a Fourier series on the interval
-n~x~n:
f (~ t) == a2 + L"" (ak cos kt + b

k=l
k sin kt), (1)
where
n n
ao= ~ S!(~ t)dt,ak= ~ S t(~ t)cosktdt,
-n -n
.bk=-;:1 J~ f (In t.) sin ktdt.

-n
Now let us return to the original variable x:

790 Fourier Series
We will then have

I I \
i,
a,~+ f(x) dx, a,~+/ (x) cos k 'f x dx, f
(2)
bk = +Sf
-I
1
(x) sin k Tx dx. J

Formula (1) takes the form
(3)
where the coefficients a 0 , ak, b k are computed from formulas (2).

This is the Fourier series for a periodic function with period 2!.
We note that all the theorems that hold for Fourier series of
periodic functions with period 2n hold also for Fourier series of
. 'XI
Fig. 365.
periodic functions with some other period 2/. In particular, the

sufficient condition for expansion of a function in a Fourier series
(see end of Sec. 1) holds true, as do also the remark on the possibi-
lity of computing coefficients o-f the series by integrating over any
interval whose length is equal to the period (see Sec. 3), and the
remark on the possibility of simplifying computation of coefficients
of the series if the function is even or odd (Sec. 4).
Example. Expand in a Fourier series the periodic function f (x) with period
21 which on the interval [ -l, l] is given by the equation f (x) =I xI (Fig. 365).
Solution. Since the function at hand is even, it follows that
I
bk=O; a0 = ~ Sxdx=l;
0
I :rt { 0 for k even,

~lk 1
2
ak =y Sx cos -k1tx1- dx =n-,;S
21
x cos kx dx = _ for k odd.
0 0 1t
On the Expansion of a Nonperiodic Function in a Fourier Series 791
Hence, the expansion is of the form
l't
COS - X
31t
COS - X COS
(2p + 1) l't X ]
l 41 l l l
lxJ=2-n [ --~-+--3.-++ (2p+l) 2 + '
SEC. 6. ON THE EXPANSION OF A NONPERIODIC FUNCTION

IN A FOURIER SERIES 1
Let there be given, on some interval [a, b] a piecewise mono-

tonic function f (x) (Fig. 366). We shall show that this function
f (x) may be represented in the form of a sum of a Fourier series
at the points of its discontinuity. To do this, let us consider an
arbitrary periodic piecewise monotonic function {1 (x) with period
2J..t ~I b- a J, which coincides with the function f (x) on the inter-
val [a, b]. [We have redefined the function f (x).]
Y
' >..-2p 'A+4ji X
Fig. 366.
Expand {1 (x) in a Fourier series. At all points of the interval

[a, b] (with the exception of points of discontinuity) the sum of
this series coincides with the given function f (x); in other words,
we expanded the function f (x) in a Fourier series on the interval
[a, b].
Let us now consider the following important case. Let a func-
tion f (x) be given on the interval lO, l]. Redefining this function
in arbitrary fashion on the interval [- l, 0] (retaining piecewise
monotonicity), we can expand it in a Fourier series. In particula.r,
if we redefine this function so that when - l ~ x < 0, f (x) = f (- x},
we will get an even function (Fig. 367). [In this case we say that
the function f (x) is "continued in even fashion".] This function
is expanded in a Fourier series that contains only cosines. Thus,
we have expanded in cosines the function f (x) given on the inter-
val [0, /].
792 Fourier Ser les
But if we redefine the function f (x) when - l ~ x< 0 as follows:

f(x)= -f(-x), then we get an odd function which may be expan-
ded in sines (Fig. 368). [The function f (x) is "continued in odd
[ashion".]
{/
j',
II ' ,, _
I
I
0 X
Fig. 367. Fig. 368.
Thus, if on the interval [0, l] there is given some piecewise

monotonic function f (x), it may be expanded in a Fourier series
both in cosines and in sines.
Example I. Let it be required :to expand ~he function f (x) = x in a series
in sines on the interval [0, n].
Solution. Continuing this function in odd fashion (Fig. 357), we get the
series
x = 2 [sin x _sin 2x +sin 3x _ ]
1 2 3 ...
Example 2. Expand the function f (x)=x in a series In cosines on the
interval [0, n).
Solution. Continuing this function In even fashion, we get
f(x)=lxl, -n<x..;;;n
(Fig. 358). Expanding it in a series we lind
f {X) = ~2 _ .!_
TC
[cos x + cos 3x + cos 5x +
1 32 52
]
(see Example 2, Sec. 2). And so on the interval [0, n] we have the equa-
tion
= _:: _ _! [~+cos23x+cos 25x+ ]
X 2 TC 1 3 5 '''
SEC. 7. MEAN APPROXIMATION OF A GIVEN FUNCTION

BY A TRIGONOMETRIC POLYNOMIAL
Representing a function by an infinite series (Fourier's, Tay-

lor's and so forth) has the following meaning in practice: the
finite sum obtained in terminating the series with the nth term
Mean Approximation of a Given Function 793
is an approximate expression of the function being expanded.

This approximate expression may be made as accurate as desired
by choosing a sufficiently large value of n. However, the charac-
ter of the approximate representation may differ.
For instance, the sum of the first n terms of a Taylor's series
s,. coincides with the function at hand at one point, and at this
point has derivatives up to the nth order that coincide with the
derivatives of the function under consideration. An nth degree
Lagrange polynomial (see Sec. 9, Ch. VII) coincides with the
function under consideration at n+ 1 points.
Let us see what the character is of an approximate represen-
tation of a periodic function f (x) by trigonometric polynomials
of the form
s,. (x) = ~ + L" ak cos kx + bksin kx,

k=l
where a0 , a 1 , b11 a 1 , b 2 , , a,., b,. are Fourier coefficients; that

is, by the sum of the first n terms of a Fourier series. We first
make several remarks.
Suppose we regard some func- y
tion y=f(x)on the interval [a, b]
and want to evaluate the error
when replacing this function by
another function cp (x). For the
measure of error we can, for in-
stance, take max If (x)- cp (x) I ~(]'~.0:+-----r-+----t-::t: ..x
on the interval [a, b], which is 'J
the so-called maximum devia- Fig. 369.
tion of cp (x) from f (x). But it is
sometimes more natural to take for the measure of error the
so-called root mean square deviation 6, which is defined by the
equation
b
c5z = (b I a)
a
s [f (x)-cp (xW dx.
Fig. 369 illustrates the difference between the root mean square
deviation and the maximum deviation.
Let the solid line depict the function y= f (x), the dashed lines
the approximations cp 1 (x) and cp 2 (x). The maximum deviation of
the curve y = cp 1 (x) is Jess than of the curve y = cp 2 (x), but the
root mean square deviation of the first curve is greater than the
second because the curve y = cp 2 (x) is considerably different from
794 Fourier Series
the curve y = f (x) on! y on a narrow sect ion and for this reason
characterises the curve y = f (x) better than the first.
Now let us return to our problem.
Let there be given a periodic function f ( x) with period 2 n.
From among all the trigonometr-ic polynomials of order n
n
~ + L, (ak cos kx + ~k sin kx)
k=t
it is required to find (by choice of the coefficients ak and ~k) that

polynomial for which the root mean square deviation defined by
the equation
~! ~ i.J.[t (x)-%'- t (a,w kx+ ~. in kx)]'dx,

has the smallest value.
The problem reduces to finding the minimum of the function
2n+ 1 of the variables a., a,, ... , an, ~ .. ~ ... , ~n
Expanding the square under the integral sign and integrating
termwise, we get
6~ = 2~ tt{r (x)- it (x{ a2 + t (ak cos kx + ~k sin kx)] +
n
+ [ ~ + t (ak cos kx +
n
~k sin kx)J}~x =
n n
= inS f" (x) dx- ~~ S f (x) dx- ~ .La~ S f (x) cos kx dx +
-n -n k=t -n
n n a n n n
+~L~ks
1
f(x)sinkxdx+in-i-5 dx+ 2n:La:S coskxdx+
k=t -n -n k=t -n
n n n 1t
+;n.L~:s sinkxdx+ina.:LakS coskxdx+

k=t -n k=t -n
n n n n n
+ina.L~ 1~S sinkxdx+! LLaka1
k=t -n k=t /=
S coskxcosjxdx+
-n
k1=J
n n n n n n
+! a 0 L ~ ak~ 1 5cos kx sin jx dx + ! L L ~k~ 5sin kx sinjxdx. 1
k=t !=1 -1'1 k=t /= -1'1
. k:F-l
We note that
n n
~
s
-n
f(x)dx=a 0 ; "it
1
s
-n
f(x)coskxdx=ak;
n
~ S f(x) sin kxdx= bk
-n
are the Fourier coefficients of the function f (x).
Further, by formulas (I) and (II), Sec. I, we have: for k = j
n :t
S cos 2
kx dx= n, S sin 2
kx dx= n,
-:t -n
:t
S sinkx cos jx dx = 0;
-n
for k=l=j
:t n
5 cos kx cos jx dx = 0, 5 -n
sin kx sin jx dx = 0.
-n
Thus, we obtain
n n a n
fl~= dn S
-n
r <x)dx- a;~- k=
I:. <akak+ ~kbk> +-i + ~ L: (a;+~;>.
. k=
Adding and subtracting the sum
a n
4 o + 2I~
~ (ak
2
+ b"k ),
k=l
we will have
n ~ n
6~ =
1
2n S f" (x) dx--i-~ k=
-n
L (a:+ bt) +{ (a -a + 0 0)
2
n
+{ L [(ak-ak).+(~k- bk) 2
] (1)
k=l
The first three terms of this sum are independent of the choice
of coefficients a., a" ... , an, ~.. . . . , ~n The remaining terms
n
-} (a0 -a~) +~L. [(ak-ak) 2 +(~k-bk) 2 ]

2
k=
796 Fourier Series
are nonnegative. Their sum reaches the least value (equal to zero)
if We. pUt U 0 = a 0 , U 1 =aJf . t Un=an, ~I= b 1 , t ~n= bn
With this choice of coefficients a 0 , U 1 , , an, ~ 1 , , ~n the
trigonometric polynomial
n
~+ L. (ak cos kx+ ~k sin kx)
k=l -
will least of all differ from the function f (x) in the sense that in
such a choice of coefficients the square deviation IS~ will be least.
We have thus proved the theorem:
Of all trigonometric polynomials of order n, that polynomial has
the least root mean square deviation from the function f (x), the
coefficients of which polynomial are the Fourier coefficients of the
function f (x) ..
The least square deviation is
n az n
6; = ins rex) dx- 4-; L. (a:+ b:).
-:rt k=l
(2)
Since 6~;;:::. 0, it follows that for any n we have

z n
sr
:rt
1
21t (x) dx;;:::. :
0
+ ; L. (a:+ b%).
-:rt k=l
Hence, the series on the right converges (when n-+ oo ), and we

can write
~ sr
n
-:rt
az
(x) dx;;:::. 2
>
+L. (a:+ b%).

k=l
(3)
This relation is called Bessel's inequality.

We note without proof that for any bounded and piecewise
monotonic function the root mean square deviation obtained ttpon
replacing the given function by the nth partial sum of the Fourier
series tends to zero as n - oo, that is, IS~- 0 as n.-+ oo. But
then from formula (2) there follows the equation
z
sr
<X> :rt
:
0
+L. (at+ b~) = ~ (x) dx, (3')
k=t -n
which is called the Lyapunov equation. (We note that A. M.
Lyapunov proved this equation even for a broader class of func-
tion than that which we here consider.)
From what has been proved it follows that for a function

which satisfies the Lyapunov equation (in particular, for any
bounded piecewise monotonic function), the corresponding Fourier
series yields a root mean square deviation equal to zero.
Note. Let us establish a property of Fourier coefficients that
will be needed in the future. We first introduce a definition.
A function f(x) is called piecewise continuous on the interval
[a, b] if it has a definite number of discontinuities of the first
kind on this interval (or is everywhere continuous).
We shall prove the following proposition.
If a function f ( x) is piecewise continuous on the interval
[-n:, n], then its Fourier coefficients approach zero as n----+oo;
that is,
lim an= 0, (4)
Proof. If the function f (x) is piecewise continuous on the in-
r
terval [ -n:, n:], then the Junction (x) too is piecewise conti
l1
nuous on this interval. Then 5r (x) dx exists and is a finite

-TI
number*). In this case, from the Bessel inequality (3) it follows
"'
that the series ~(a~ + b~) converges. But if the series conver-
n=
ges then its general term approaches zero; in this case,
lim (a~+ b~) = 0. Whence we get equations (4) directly. Thus, the
n-+o
following equations are valid for a piecewise continuous and
bounded function:
l1
lim
n-+a>
5 f (x) cos nx dx = 0,
-TI
l1
lim r f (x) sin nxdx= 0.
n-+trJJ
-TI
If a function f (x) is periodic with period 2n, then the latter

equations may be written as follows (for any a):
a+ITI a+an
1im ~ f (x) cos nx dx = 0; lim S f (x) sin nx dx= 0.
n-+a'l a n~oo a
This integral may be presented as the sum of definite integrals of coo-

t inuous functions over the subintervals into which the interval [ -n n] is
subdivided. '
798 Fourier Series
We note that these equations continue to hold if in the integrals

we take any arbitrary interval of integration [a, b], which is to
say that the integrals
b b
~ f (x) cos nx dx and ~ f (x) sin nx dx
a a
approach zero when n increases without bou_nd if f (x) is a bound-
ed and piecewise continuous function.
Indeed, taking b-a<2~ for definiteness, we consider the auxi-
liary function q> (x) with period 2n defined as follows:
cp (x) = f (x) when a ,;;::;;; x ,;;::;;; b
cp(x)=O when b<x,;;::;;;a+2n,
Then
b a+!n
~ f (x) cos nx dx = ~ cp (x) cos nx dx,
a a
b a+2n
~ f (x) sin nx dx = -~ cp (x) sin nx dx.
a a
Since <p (x) is a bounded and piecewise continuous function, the
integrals on the right approach zero as n--+ oo. Hence, the in-
tegrals on the left approach zero as well. Thus, the proposition is
proved; that is,
b b
lim ~ f (x) cos nx dx = 0; lim ~ f (x) sin nx dx = 0 (5)
n-+CI) a n~oo a
for any numbers a an~ b and any piecewise continuous function

f (x) bounded on [a, b].
SEC. 8. THE DIRICHLET INTEGRAL
In this section we shall derive a formula that expresses the

nth partial sum of a Fourier series in terms of a certain integral.
This formula will be needed in the subsequent sections,
Consider the nth partial sum of a Fourier series for the peri-
odic function f (x) with period 2n:
sn (x) = ~ + ,L (a,. cos kx+ b,. sin kx),
where
It n
a,.=! S -l't
f(t)cosktdt, b,.=! St(t)sinktdt.
-n
The Dirichlet Integral 79i:l
Putting these expressions into the formula for sn (x), we obtain

n
sn (x) = 2~ S f (t) dt +
-n
tJ n - n
+L [co~ kx S f (f) cos kt dt + sl~kx S f (t) sin kt dt J,
k=t -n -n
or bringing cos kx and sin kx under the integral sign (which is
possible since cos kx and sin kx are independent of the variable
of integration and, hence, can be regarded as constants), we get
n
sn (x) = Sf (t) dt +
2~ -n
n n n
+~ ~ r
k-1
sf
-n.
(t) cos kx cos kt dt + sf
-n
(t) sin kx sin kt dt J.
Now taking ~ outside the brackets and replacing the sum of in
tegrals by the integral of the sum, we obtain
sn(x) = ~ L{' ~t) +t (f(t) cos kx cos kt + f (t) sin kx sin kt]} dt,
t
or
sn (x) = ~ Lt {t) [-} + (cos kt cos kx + sin kt sin kx)] dt =
= ~ Lf(t)[ -}+~cosk(t-x)}t. (1)
Transform the expression in the brackets. Let

1
an (z) = 2 +cos z+ cos2z+ ... +cos nz;
then
2crn (z) cos z =cos z + 2 cosz cosz + 2 cos z cos 2z +. :-;
... + 2 cos z co.s nz = cosz +(I+ cos 2z) + (cosz +cos 3z) +
+(cos 2z +cos 4z) + ... +[cos (n-l)z +cos (n + 1) z] =
= 1 + 2 cos z + 2 cos 2z + .. + 2 cos (n- 1) z + cos tlZ + cos (n + 1) z
81)() Fourier Series
or
2on (z) cos z = 2on (z)- cos nz +cos (n + I) z,
cosnz-cos(n+l)z
0 () =
n Z 2(1-cos z)
But
cos nz- cos (n + 1) z = 2 sin (2n + 1) ]- sin ~ ,
. zz
1-COSZ= 2 Sin
2
Hence,
.
sin (2n + 1) 2z
on(Z)= z
2sin
2
Thus, equation (1) may be rewritten as
sn(x)=-nj f(t)
1 r sin(2n+l)2
t-x
t-x
dt.
:-n 2 sin--
2
Since the integrand is periodic (with period 2n), it follows
that the integral retains its value on. any interval of integration
of length 2n. We can the.refore write
X+n sin (2n + 1 ) -
t-x
sn (x) = ! x--n
Sf (t) t -x
2 sin - 2 -
2
dt.
Introducing a new variable a, we put

t-x=a, t=x+a.
Then we get the formula
n sin (2n + I) 2a
sn (x) = ! S t (x +a)
_ 11 2sin ~
da. (2)
The integral on the right is Dirichlet's integral.

In this formula put f (x) ==
1; 'then a 0 = 2, ak = 0, b k = 0 when
k>O; hence, sn (x) = 1 for any n and we get the identity
1 sn sin(2n+l)~
1 =n a da, (3)
-n 2sln
2
which we will need later on.
The Convergence of a Fourier Series at a Given Point 801
SEC. 9. THE CONVERGENCE OF A FOURIER SERIES

AT A GIVEN POINT
Assume that the function f (x) is piecewise continuous on the
interval [ -n, n].
Multiplying both sides of the identity (3) of the preceding
section by f (x) and bringing f (x) under the integral sign, we get
the equation
a
1 511: sin (2n + 1) 2
f(x) =-
n
f(x) . a
da.
-n 2 stn 2
Subtract the terms of the latter equation from the corresponding
terms of (2) of the preced_ing section; we get
1 511: sln (2n + 1) 2a
sn (x)- f (x) =-
n [f (x +a)- f (x)] . a
da.
-n 2 stn 2
lfhus, the convergence of a Fourier series to the value of a func-
tion f (x) at a given point depends on whether the integral on
the right approaches zero as n--+- oo.
Let us break up this integral into two integrals:
a
5 1
n
cos 2
sn(x)-f(x) =-;t (f(x+a)-f(x)]--a sinnada+
-n 2sin
2
n
+{ ~ 5 (f(x+a)-f (x)] cos na da,
-l1
taking advantage of the fact that sin (2n + 1) .~ =sin na cos i+

+cos na sin ~ . Break up the first of the integrals on the right
of the latter equation into three integrals:
sn(x)-f(x) =n-
1 5o (f(x +a)-f(x)] -cos- s2a i n na da+
. a
-0 2 Stn2
+-;t
1 5(f(x+a)-f(x)] -.-a
-6
cos2
+sin nada+
a
-n 2sm2
+-;ts
n a n
l
[f(x+a)-f(x)]--a sinnuda+-;t5[f(x+a)-f(x)hcosnada.
cos 2 1 I
2sin -n
2
26~ 3388
802 Fourier Series
Put<D 1 (a)=f(x+i-f(x). Since f(x) is a bounded piecewise con-

tinuous function, it follows that <1> 1 (a) is also a bounded and
piecewise continuous periodic function of a. Hence, the latter
integral approaches zero as n-+ oo, since it is a Fourier coeffi-
cient of this function. The function
a
cos 2
<D 2 (a)=[f(x+a)-f(x)]-_-a
2sin
2
is bounded when - n-=:;;; a<- <'l and <'l ~a~ n and
I<D 2 (a) I~ [M +M]-

1
-,
e
2sin
2
where M is the upper limit of the quantity If (x) I . Also, the
function <1> 2 (a) is likewise piecewise continuous. Hence, by for-
mulas (5) of Sec. 7, the second and third integrals approach zero
as n-+ oo.
We can thus write
~ cos~
Iim[sn(x)-f(x)]=Iim n
n-+"' n-+ ao
J..S rf(x+a)-f(x)] -
-~
- sinnada.
2 stn a .
2
(1)
2
In the expression on the right, the integration is performed
over the interval -~~a~ <'l; consequently, the integral is depen-
dent on the values of the function f (x) only in the interval from
x-<'l to x+<'l. An important proposition thus follows from the
latter equation: the convergence of a Fourier series at a given
point x depends only on. the behaviour of the function f (x) in an
arbitrarily small neighbourhood of this point.
Therein lies the so-called principle of localisation in the study
of Fourier series. If two functions {1 (x) and f 2 (x) coincide in the
neighbourhood of some point x, then their Fourier series simulta-
neously either converge or diverge at this point.
SEC. 10. CERTAIN SUFFICIENT CONDITIONS FOR -THE

CONVERGENCE OF A FOURIER SERIES
In the preceding section it was shown that if the function f (x)

is piecewise continuous in the interval [-n, n], then the conver-
gence of a Fourier series at the given point x~ to a value of the
function f-(X 0 ) depends on the behaviour 01 the function in a
Certain Sufficient Conditions for the Convergence of a Fouriefl Series 803
certain arbitrary small neighbourhood [x0 - B, x 0 + 6] with centre

at the point x 0
Let us now prove that if in the neighbourhood of the point x0
the function f (x) is such that there exist finite limits
lim f(x.+a)-f(x 0 )=k ( l) '
a-+--o
a 1'
lim f(x.+a)-f(xo)=k (2)

a~+ - a t>
while the function is continuous at the very point x 0 (Fig. 370),

then the Fourier series converges at this point to a correspond-
ing value of the function f (x) *).
Proof. Let us consider the fu ric y
tion <1> 2 (a) defined in the preced-
ing section:
a
cos 2
 2 (a)=lf(x0 +a)-f(x.)] - - a ; X
2sin
2
Fig. 370.
since the function f(x) is piece-
wise continuous on the interval
[ -n, n] and is continuous at the point X0 , it is therefore con-
tinuous in some neighbourhood [x0 --6, x.+61 of the point X0 ,.
For this reason, the function <1> 2 (a) is continuous at all points
where a=I=O and Ia j..;;; 6. When a= 0 the function <1> 1 (a) is not
defined.
Let us find the limits Jim 2 (u) and Iim 1 (a), making us~
a~o-o a~o+o
of conditions (l) and (2):

(l
cos-
2
lim .(a)= Jim [f(x.+a)-f(x0 )1---=
a~o-o a~o.:...o 2 sin ~
2
a
. /(x 0 +a)-f(xo) 2 a
= I tm --cos-=
a~o-o a sin ~ 2
2
a
_ limf(x.+a)-f(x.) lim 2
lim cos!!:..._=k 1-l=k.
a~o-o sin~ a~o-o
1 1
a-+o-o a 2
2
*) If conditions (I) and (2) are fulfilled, then we say that the function
f (x) has, at the point x, a derivative on the right and a derivative on the
left. Fia. 370 shows a function where k 1 =tan1Jl 10 k 2 =tan<p 2 , k1 :;z!:k 2 If
k, = k2 , that is, if the derivatives on the right and left are equal, tht:n the
function will be difierentiable at the given point.
26*
804 Fourier Series.
Thus, if we redefine the function <D, (a) by putting <D1'(0) = k.,

then it will be continuous on the interval [ -1'>, 0], and, hence,
bounded as well. Similarly we prove that
lim <D 2 (a)= k 1
a-+O+O
Consequently, the function <D. (a) is bounded and continuous

on the interval [0, 6]. Thus, on the interval [ -6, 6] the func-
tion <D 1 (a) is bounded and piecewise continuous. Now let us
return to equation (1), Sec. 9 (denoting x in terms of x 0 ),
.
~
lim lsn (x 0 ) -f (x 0 )] =11m -

1 s a
cos 2 .,
[f (x 0 +a) -f (X 0 }1 - - sm na da
n-+-.., n-+"' :n: -a 2 sin ;
or
~
lim [sn (xo) -f (xo)l = lim_!_

n-+"' n-+a>
s<Dl (a)
:n: ~
sin na da.
From formulas (5) of Sec. 7 we conclude that the limit on the

right is equal to zero, and therefore
lim [sn (X 0 ) - f (x 0 )] = 0 .
or
lim sn(xo)=f(xo)
n-+<~>

This theorem differs from the theorem stated in Sec. 1 in that
.in the latter case it was required, for convergence of the Fou-
rier series at a point X 0 to the value of the function f (x 0 ), that
the point X0 should be a point of continuity on the interval
[ -n, n], whereas the function should be piecewise monotonic;
here, however, it is required that the function at the point X 0
should be a point of continuity and that the conditions (1) and
(2) be fulfilled, while throughout the interval l-n, n] the func-
tion should be piecewise continuous and bounded. It is obvious
that these conditions are different.
Note 1. If a piecewise continuous function is differentiable at
the point X0 , it is obvious that conditions (1) and (2) are ful-
filled. Here, k 1 =k 1 Hence, at points where the function f(x)
is differentiable, the Fourier series converges to a value of the
function at the corresponding point.
Note 2:. a) The function considered in Example 2, Sec. 2
(Fig. 358), satisfies conditions (1) and (2) at the points 0, 2n,
4n, ... At all the other points it is differentiable. Consequent-
Practical Harmonic Analysis 805
ly, a Fourier series constructed for it converges to the value of

this function at each point.
b) The function considered in Example 4, Sec. 2 (Fig. 361),
satisfies conditions (1) and (2) at the points n, 3n, 5n.
It is differentiable at all points. It is represented by a _Fourier
series at each point.
c) The function considered in Example 1, Sec. 2 (Fig. 357),
is discontinuous at the points n, 3n, Sn. At all other
points it is differentiable. Hence, at all points, with the excep-
tion of points of discontinuity, the Fourier series corresponding
to it converges to the value of the function at the corresponding
points. At the discontinuities, the sum of the Fourier series is
equal to the arithmetical mean limit of the function on the right
and on the left (in this case, zero).
SEC. 11. PRACTICAL HARMONIC ANALYSIS
The theory of expanding functions in Fourier series is called

harmonic analysis. We shall now make several remarks about
approximate computation of the coefficients of a Fourier series,
that is to say, about practical harmonic analysis.
As we know, the Fourier coefficients of a function f (x) with
period 2n are defined by the formulas
n n
ao = ~ S f (x) dx; ak = ~ S f (x) cos kx dx;
-.n: -.n:
n
bk = ~ Sf (x) sin kx dx.
-n
In many practical cases, the function f (x) is represented either
in tabular form (when the functional relation is obtained by
experiment) or in the form of a curve which is plotted by some
kind of instrument. In these cases the Fourier coefficients are
calculated by means of approximate methods of integration (see
Sec. 8, Ch. XI).
Let us consider the interval -n ~ x ~ n of length 2n. This
can always be done by proper choice of scale on the x-axis.
Divide the interval [ -n, n] into n equal parts by the points
Then the subinterval will be

211:
llx=-n
806 Fourier Series
We denote the values of the function f (x) at the points X0 , x.,

, Xn (respectively) in terms of
X2 ,
Yo Yt> Y2 , Yn
These values are determined either from a table or from the
graph of the given function (by measuring the corresponding
ordinates).
Then, taking advantage, for example, of the rectangular for-
mula [see formula (1), Sec. 8, Ch. XI], we determine the Fourier
coefficients:
n n n
ao= ~LY;. ak= ~ LY; cos kx;, bk = ~ LY; sin kx;.

i=J 1=1 1=1
Diagrams have been devised that simplify computation of Fou-

rier coefficients (see, for instance, V. I. Smirnov, "Course of
Higher Mathematics", Vol. II; A. M. Lopshits, "Models for Har-
monic Analysis"). We cannot deal here with the details but we
can note that there are instruments (harmonic analysers) which
permit approximating the values of Fourier coefficients from the
graph of the function.
SEC. 12. FOURIER INTEGRAL
Let a function f (x) be defined in an infinite interval ( -- oo,

oo) and absolutely integrable over it; that is, there exists an
integral
QO
5 if(x)jdx=Q.
-<t>
(1)
Further, Jet the function f(x) be such that it is expandable

into a Fourier series in any interval (-1, +1):
(fj
f(x) = ~+ L ak cos kt' x+ bk sin kt x, (2)

k=~
where
ak=z1 sl
k:n tdt,.
f(t)cosT
l
bk=+S f(t)sink1:n+dt. (3)

-l -I
Fourier Integral 807
Putting into series (2) the expressions of the coefficients ak and

bk from formulas (3), we can write
I "' I
f(x)=~S f(t)dt+i .L(S f(t)coskt'tdt)cosk/'x+
-1 k=l -1
I
+ (5 /(t) sink; t dt )sin ~n x=

~
l "' t
=~S f(t)dt++~S f(t) [coskt'tcoskt'x+sink/'tsinkt'x] dt
-1 k- 1 -1
or
I "' I
f(x)=~S f(t)dt+i .LS f(t)coskn(tl-x)dt. (4)
-1 k=-t
Let us investigate what form expansion (4) will take when

passing to the limit as l-+ oo.
We introduce the following notation:
1t 21t k1t 1t
a,= l as= T' ... ' ak = T' and flak= T. (5)
Substituting into (4), we get

I "' l
f (x) = ft 5f (t) dt +
-l
~ .L(S f (t)
k=l -1
cos ak (t -x) dt) flak. (6)
As l-+ oo, the first term on the right approaches zero.

Indeed,
l l "'
I~I sf (t) dt I~~ sIf (t) Idt < 2~ sIf (t) Idt = il Q-.0.
-l -l -a>
For any fixed l, the expression in the parentheses is a function

of ak [see formula (5)], which takes on values from ~to oo. We
will show, without proof, that if the function f (x) is piecewise
monotonic on every finite interval, is bounded on an infinite inter-
val and satisfies condition (1), then as oo formula (6) takes 1--+
the form
.. 00
f (x) = ~ S( Sf (t) cos a (t- x) dt) d a. (7)

0 - QD
808 Fourter Series
The expression on the right is known as the Fourier integral of

the function f (x). Equation (7) occurs for all points where the
function is continuous. At points of discontinuity we have the
equation
"' "' 1(x+O) + f (x -0)
~ S ( S f(t)cosa(t-x)dx) = 2 (7')
o -oo
Let us transform the integral on the right of (7) by expanding

cosa(t-x):
cos a (t-x) =cos at cos ax+ sin at sin ax.
Putting this expession into formula (7) and taking cos ax and
sin ax outside the integral signs, where the integration is performed
with respect to the variable t, we get
"' "'
f (x) = ~ S ( Sf (t) cos at dt) cosaxda+
o -oo
"' "'
+~ S( S f (t) sin at dt) sin ax da. (8)
0 -"'
Each of the integrals in brackets with respect to t exists, since

the function f (t) is absolutely integrable in the interval ( -oo, oo),
and therefore the functions f (t) cos at and f (t) sin at are also abso-
lutely integrable.
Let us consider particular cases of formula (8).
1. Let f (x) be even. Then f (t) cos at is an even function, while
f (t) sin at is odd and we have
CD ~ CO
~ f(t)cosatdt=2) f(t)cosatdt,
-"' 0
"'
~ f (t) sin at dt = 0.
-co
Formula (8) in this case takes the form

"' "'
f(x)=! S(St(t)cosatdt)cosaxda. (9)
0 0
Fourier l ntegral 8GJ
2. Let f (x) be odd. Analysing the character of the integrals in

formula (8) in this case, we obtain
"' "'
f(x):= ~~ (S f(t) sinatdt) sinaxda. (10)
0 0
If f (x) is defined only in the interval (0, oo ), then for x > 0

it may be represented by either formula (9) or (10). In the first
case we redefine it in the interval (- oo, 0) in even fashion; in
the latter case, in odd fashion.
Let it be noted once again that at the points of discontinuity
we should write the following expression in place of f (x) in the
left-hand members of (9) and ( 10):
f (x+O)+ f (x- 0)
2
Let us return to formula (8). The integrals in brackets are func-
tions of a. We introduce the following notation:
"'
A (a)= ~ 5 f (t) cos at dt,
"'
B (a)= ~ 5f (t) sin atdt.
Then formula (8) may be rewritten as follows:

"'
f(x)= ~ [A(a)cosax+B(a) sinax]da. (11)
We say the formula (11) yields an expansion of the function f (x)

into harmonics with a frequency a that continuously varies from
0 to oo. The law of distribution of amplitudes and initial phases
as dependent upon the frequency a is expressed in terms of the
functions A (a) and B (a).
Let us return to formula (9). We set
_co
F (a)= y ~ Sf (t) cos at dt; (12)
then formula (9) takes the form

_co
f (x) = Y ~ SF (a) cos ax da. (13)

810 Fourier Series
The function F (a) is called the Fourier cosine transform of the

function f (x).
If in (12) we consider F (a) as given and f (t) as the unknown
function, then it is an integral equation of the function 'f (t).
Formula (13) gives the solution of this equation.
On the basis of formula (10) we can write the following equations:
Y~S
_ttl
<D(a)= f(t) sinatdt, (14)
_>
f (x) =- Y! S <D (x) sin axda. (15)
The function <D (a) is called the Fourier sine transform.

Example. Let
f(x)=e-llX (~>0, x~O).
From (12) we determine ,the Fourier cosine transform:
F(a)= y! Se-~tcosatdt_= y! ~!a

0
From (14) we determine the Fourier sine transform:
ID(a}= y ~ S e-~1 slnaldt= Y-~ ~ 2 ~-a

0
From formulas (13) and (15) we find the reciprocal relationships!
(x::;;:, 0),
-2 sa---a-e...--
a>
sin ax d __ ,.,
(x> 0).
n ~+a
0
_ SEC. 13. THE FOURIER INTEGRAL IN COMPLEX FORM
In the Fourier integral [formula (7), Sec. 12), the brackets con-
tain an even function of a; hence, it is defined for negative values
of a as well. On the basis of the foregoing, formula (7) can be
rewritten as follows:
a> a>
f (x) = 2~ S( Sf(t) cos a (t-x) dt) da. (1)

-oo -co
The Fourier Integral in Complex Form 811
Let us now consider the following expression, which is identically

equal to zero:
M "'
5( 5f(t)sina(t-x)dt)da=O.
-M -CIO
The expression on the left is identically equal to zero because

the function of a in the brackets \s an odd function, and an in-
tegral. of an odd function from -M to + M is equal to zero. It
is obvious that
M "'
lim ~ ( ~ f(t) sina(t-x)dt)da=O
M-+a:> -M -CIO
or
"' (10
~ ( ~ f(t) sina(t-x)dt) da=O. (2)

-<XI -oo
Note. It is necessary to point to the following. A convergent

integral with infinite limits is defined as follows:
"' c "'
~ cp(a)da = ~ cp(a)da + ~ cp(a)da=
-~ -oo c
c M
= lim ~ cp (a) da + lim ~ cp (a) da (*)
M-+"'-M M-+<t>c
! provided that each of the limits to the right exists (see Sec. 7,
Ch. XI). But in equation (2) we wrote
"' M
~ cp (a) da = lim ~ cp (a) da. (**)
-CIO M-+CIO -M
Obviously, it may happen that the limit (**) exists, while the
limits on the nght side of equation (*) do not exist. The expres-
sion on the right of (**) is called the principal value of the in-
tegral. Thus, in equation (2) we consider the principal value of
the improper (outer) integral. The subsequent integrals of this section
will be written in this sense.
Let us multiply the terms of (2) by irr. and add them to the
2
corresponding terms of (1); we then get
(10 (10
1
f(x)= 2 rr. S [S f(t)(cosa(t-x)+isina(t-x))dt] da
-co -w
812 Fourier Series
or
f(x)= 2~ S"' [S"' f(t)eia(l-x>dt] da. (3)

-ao -CD
This is the Fourier integral in complex form. Formula (3) may be

rewritten as follows:
On the basis of this latter equation we can write
F* (a.)= -/2n s"' -<D

f(t)eialdt, (4)
"'
f (x) = -/2n S
_.,
F* (a) e-iaxda. (5)
The function F* (a) defined by formula (4) is called the Fourier

transform of the function f (t). The function f (x) defined by for-
mula (5) is called the Fourier inverse transform of the funCtion
F* (a.) (the transforms differ in the sign in front of i).
Exercises on Chapter XV II
1. Expand the following function in a Fourier series in the interval( -n, n);
f(x)=2x for Ote;;;;x<;;n,
f(x)=x for -n<x..;;;O.
_!_ _ _! (~+cos3x+cos5x+
Ans. 4n n 12 3Z sa . . . )+ 3 (slnx_sln2x+
1 2
sln3x_ )
+ 3 ....
2. Taking advantage of the expansion of 'the function f (x) = 1 In I he infer-
val (0, n) in the sines of multiple arcs, calculate the sum of the series
1 1 1 n
1-a-+s-T+... Ans. T.
3. Utilising the expansion of the function I (x) =x 2 in 'a Fourier series,
. 1 1 1 1 n
compute the sum of the senes
12 - 2 32 - 41 ,+ +... Ans. 12.
4. Expand the function f (x) = ~;- ~~ in a Fourier series in the interval
cos 2x -1 cos 3x cos 4x +
( -n, n). Ans. cos x---"'22 -
32 ---:.p-- ...
Exercises on Chapter XV II 813
5. Expand the following function in a Fourier series in the interval ( -n, l't)
f (x) = - (n t x) for -l't .;;;;; x <0,

1
f (x)="2 (l't~x) for 0 .;;;;x < l't.
Ans. slnx+ ~ sin2x+ ~ sln3x+ ...

6. Expand in a Fourier series, in lhe interval ( -l't, l't), the function
f (x) = -x for -n <x.;;;;; 0,
/(x)=O for O<x..;n.
L
(I) (I)
Ans . ..:!__2_
4 n m=o
cos(2n+1)x
+
(2n 1)2
L (-1)n-t sl:nx.
n=l
7. Ex.pand in a Fourier series, in the interval ( -n, n), the function
/(x)=1 for -n < x.;;;;o,
/(x)= -2 for 0 < xo;;;;n.
(I)
_..!__~~sin (2n+ l)x

Ans. 2
n=o
2n 1n ~ +
8. Expand the function f (x) =x 2 , in the interval (0, n), in a series of sines.
Ans. ! ~(-l)n+'{~~+:.[(-l)n-l]}slnnt.
9. Expand the function y =cos 2x in a series of sines in the interval (0, zt).
_..!_ [ slnx +3sln 3x+5sin 5x+ ]
Ans. n 21 -1 21 -31
22 -5 2 '
10. Ex land the function y =sin x in a series of cosines in the interval (0, n)
..!_ ..!_ + cos 2x + cos 4x + ]
Ans. n 2 1 -22 1 -4 2 " - .
11. Expand the function y =ex in a Fourier series in the interval ( -l, l),
n . nnx
l -l "' ( - l) stn--
Ans. ~+ l (el-e-l)
21
Ln=l l
2
1
+ n2n 1 +
12. Expand the function /(x)=2x in a series of sines in the interval (0,1}
Ans. 1_2
n
L... sin n2nnx.-
n=l
814 Fourier Series
13. Expimd the function f (x) =x in a series of smes 1n the interval tO, I).
"' sin~
Ans. ~ ~ ( -l)n+ - -1- .
n ~ n
n=J
14. Expand -the function
x for 0 < x..,;;;; I,
(x)= { 2-x for 1 <x<2
In 1he Interval (0, 2): a) In a series of sin:s; b) in a series of cosines.
.., (2n +I) n~
8 s1n 2 "'
b) _!__.i_ ~cos
Ans. a) n L (-l)n
n=o
(2n + 1) I , 2 :t 2 ~
n=o
(2n +I) :rtx
(2n+l) 2
C H A P T E R XVIII
EQUATIONS OF MATHEMATICAL PHYSICS
SEC. 1. BASIC TYPES OF EQUATIONS OF MATHEMATICAL PHYSICS
The basic equations of mathematical- physics (for the case of

functions of two independent variables) are the following second-
order partial differential equations.
I. Wave Equation:
(1)
This equation is invoked in the study of processes of transversal

vibrations of a string, the longitudinal vibrations of rods, electric
oscillations in conductors, the torsional oscillations of shafts, gas
vibrations, and so forth. This equation is the simplest of the._class
of hyperbolic equations.
II. Fourier Equation for Heat Conduction:
(2)
This equation is invoked in the study of processes of the propa-

gation of heat, the filtration of liquids and gases in a porous
medium (for example, the filtration of oil and gas in subterranean
sandstones), some problems in probability theory, etc. This equation
is the simplest of the class of pdrabol ic equation.
III. Laplace's Equation:
(3)
This equation is invoked in the study of problems dealing with

electric and magnetic fields, stationary thermal states, problems
in hydrodynamics, diffusion, and so on. This equation is the simplest
in the class of elliptic equations.
In equations (1), (2), and (3), the unknown function u depends
on two variables. Also considered are appropriate equations of
functions with a larger number of variables. Thus, the wave
816 Equations of Mathematical Physics
equation in three independent variables is of the form

au 2 ( au au) (I')
at =a ax+ ay '
the beat-conduction equation in three independent variables is of
the form
au au au)
ax + ay '
2 (
7ff =a (2" J
the Laplace equation in three independent variables has the form
(3' J
SEC. 2. DERIVATION OF THE EQUATION OF OSCILLATION OF A STRING.

FORMULATION OF THE BOUNDARY-VALUE PROBLEM.
DERIVATION OF EQUATIONS OF ELECTRIC OSCILLATIONS IN WIRES
In mathematical physics a string is understood to be a flexible

and elastic thread. The tensions that arise in a string at any
instant of time are directed along a tangent to its profile. Let a
string of length l be, at the initial instant, directed along a seg-
ment of the x-axis from 0 to l. Assume that the ends of the
string are fixed at the points x = 0
u and x = l. If the string is deflected
from its original position and then
let loose; or if without deflecting the
string we impart to its points a cer-
0;:;-F---o----'-~~:--~x tain velocity at the initial time, or
if we deflect the string and impart
Fig. 371. a velocity to its points, then the
points of the string will perform
certain motions; we say that the string is set into oscillation.
The problem is to determine the shape of the string at any instant
of time and to determine the law of motion of every point of the
string as a function of time.
Let us consider small deflections of the points of the string from
the initial position. We may suppose that the motion of the points
of the string is perpendicular to the x-axis and in a single plane.
On this assumption, the process of oscillation of the string is
described by a single function u (x, t), which yields the amount
that a point of the string with abscissa x has moved at time t
(Fig. 371).
Since we consider small deflections of the string in the (x, u)-
plane, we shall assume that the length of an. element of string
Derivation of the FqJJ.ation of Oscillations of a String 817
M M is equal to its projection on the x-axis, that is,*) M 1 M 2 =

=\ ..:_x . We also assume that the tension of the string at all
poirtts i~ the same; we denote it
by T.
Consider an element of the string
MM' (Fig. 372). Forces Tact at the
ends of this element along tangents
to the string. Let the tangents form I X
with the x-axis angles qJ and qJ + llqJ.
Then the projection on the u-axis of Fig. 372.
forces acting on the element MM'
wi II be equal to T sin (qJ + llqJ)- T sin qJ. Since the angle qJ is small,
we can put tan qJ-::::: sin qJ, and we will have
T sin (qJ + llqJ)- T sin qJ=::::
+
=:::: Tt an ( qJ
A
uqJ )
- Tt anqJ = T [ou(x+M.
ox t) ou(x,
---axt)] =
- T o2u (x + e~x. t) A ~T ou (x, y) A
-. ax uX-. ox uX,
0<6<1
(here, we applied the Lagrange theorem to the expression in the
squar~ brackets). .
In order to obtain the equation of motion, we must equate to
the force of inertia the external forces applied to the element.
Let Q be the linear density of. the string. Then the mass of the
element of the string will be Qllx. The acceleration of the element
.JS ou
at. Hence, by d'Alembert's principle we will have
ou au
QllX at = T ox llx.
Cancel1ing out llx and denoting !!_ = a, we get the equation of
~ Q
motion:
(1)
This is the wave equation, the equation of vibrations of a string.

Equation (1) by itself is not sufficient for a complete definition
*) This assumption is equivalent to neglecting u: as compared with 1.

Indeed,
M1 M2 =
xsv-, .
!+ux'dx=
xS.( 1+2ux....,...
1 ,.
...
)
dx=::::
xS. dx=x
2 - x1
~ ~ ~
of the motion of a string. The desired function u (x, t) must also

satisfy boundary conditions that indicate what occurs at the ends
of the string (x = 0 and x = l) and initial conditions, which
describe the state of the string at the initial time (t = 0). The bound-
ary and initial conditions are referred to collectively as boundary-
value conditions.
For example, as we assumed, let the ends of the string at x = 0
and X= l be fixed. Then for any t the following equalities must
hold:
u(O, t)=O, (2')
u(l, t)=O. (2")
These equations are the boundary conditions for our problem.
At t = 0 the string has a definite shape, that which we gave it.
Let this shape be defined by a function f (x). We should then
have
u(x, O)=Uit=o=f(x). (3')
Further, at the initial instant the velocity at each point of the
string must be given; it is defined by the function <p (x). Thus,
we should have
iJul (3")
iJt t= = <p (x).
The conditions (3') and (3") are the initial conditions.

Note. For a special case we may have f(x):=O or <p(x):=O.
But if f (x) := 0 and <p (x) := 0, then the string will be in a state
of rest; hence, u (x, t) := 0.
As has already been pointed out, the problem of electric oscil-
lations in wires likewise leads to equation (1). Let us show this
to be the case. The electric current in a wire is characterised by
the current flow i (x, t) and the voltage v (x, t), which are depen-
dent on the coordinate x of the point of the wire and on the
time t. Regarding an element of wire 11x, we can wr_ite that the
+
voltage drop on the element 11x is equal to v (x, t)- v (x 11x, t)::::::
::::::- ~~ 11x. This voltage drop consists of the ohmic drop, equal
to iRI1x, and the inductive drop, equal to :: Ll1x. Thus,
iJv iJi
- iJx 11x = tRI1x +at Ll1x, (4)
where R and L are the resistance and the coefficient of self-induc-

tion reckoned per unit length of wire. The minus sign indicates
Derivation of the Equation of Oscillations of a String 819
that the current flow is in a direction opposite to the build-up

of v. Cancelling out !:J.x, we get the equation
ax tR+L ~-0
av+ at- . (5)
Further, the difference between the current leaving element !:J.x

and entering it during time M will be
i (x, t)-i(x+ !:J.x, t)=:::- ~x !:J.xM.
It is taken up in charging the element (this is equal to C!:J.xg~ M)

and in leakage through the lateral surface of the wire due to
imperfect insulation, equal to Av!:J.xM (here A is the leak coeffi-
cient). Equating these expressions and cancelling out !:J.xM, we
get the equation
(6)
Equations (5) and (6) are generally called telegraph equations.
From the system of equations (5) and (6) we can obtain an
equation that contains only the desired function i (x, t}, and an
equation containing only the desired function v (x, t). Differentiate
the terms of equation (6) with respect to x; differentiate the terms
of (5) with respect to t and multiply them by C. Subtracting,
we get
azt2 +A av -CR ai -CL azi2 = 0.
ax OX ax ot
Substituting into the latter equation the expression ~: from (5),
we get
021
axz +A(- tR-L Ol~)-CR OX~-CL azi -0
iJtZ -
or
Similarly, we obtain an equation for determining v (x, t):

ov
2
02 V av
axz = CL atz + (CR + AL) at+ ARv. (8)
If we neglect the leakage through the insulation (A= 0) and

the resistance (R =0), then equations (7) and (8) pass into the
820 Equations of Mathematical Physics-,
wave equations
I i)2{ i)Zl I iJ!V iJIV
a iJx2 = i)[Z I a iJxl = iJ[Z I
where a = 2
dL.
The physical conditions dictate the formulation
of the boundary and initial conditions of the problem.
SEC. 3. SOLUTION OF THE EQUATION OF OSCILLATIONS

OF A STRING BY THE METHOD OF SEPARATION OF VARIABLES
(THE FOURIER METHOD)
The method of separation of variables (or the Fourier method),

which we shall now discuss, is typical of the solution of mat:ty
problems of mathematical physics. Let it be required to find the
solution of the equatiOn
iJ 2u z iJ2u
i)[Z =a i)xZ (1)
which satisfies the boundary-value conditions
u(O, t)=O, (2)
u(l, t)=O, (3)
u (x, 0) = f (x), (4)
I
:~ t~o = q> (x). (5)
We shall seek a particular solution (not identically equal to zero)
of equation (1) that satisfies the boundary conditions (2) and (3),
in the form of a product of two functions X (x) and T (t), of
which the former is dependent only on x, and the latter, only
on t:
u (x, t) =X (x) T (t). (6)
Substituting into equation (1), we get X(x)T"(t)=a 2 X"(x)T(t),
and dividing the terms of the equation by a'XT,
T" X"
aT=X' (7)
The left member of this equation is a function that does not
depend on x, the right member is a function that does not depend
on t. Equation (7) is possible only when the left and right mem-
bers are not dependent either on x or on t, that is, are equal to
a constant number. We denote it by -'A., where 'A.>O (later on
we will consider the case 'A.<O). Thus,
T" X"
a2T=x= -'A..
Solution of the Equation of Oscillations of a String 821
From these equations we get two equations:

X" +'AX= 0, (8)
T" + a2'AT = 0. (9)
The general solutions of these equations are (see Ch. XIII, Sec. 21)
X (x) =A cos V~+B sin V~x. (10)
T (x)=CcosaV~t+D sinaV~t. (11)
where A, B, C, and D are arbitrary constants.
Substituting the expressions X (x) and T (t) into (6), we get
u (x, t) =(A cos V~x + B sin V~x) (C cos a"J/~t + D sin aV~).
Now choose the constants A and B so that the conditions (2) and
(3) are satisfied. Since T (t);t:O (otherwise we would have u (x, t)=O,
which contradicts the hypothesis), the function X (x) must satisfy
the conditions {2) and (3); that is, we must have X (0) = 0, X (l) = 0.
Putting the values x=O and x=l .into (10), we obtain, on the
bas is of (2) and (3),
0= A cosV~l +B sin"V1l =0.

From the first equation we find A= 0. From the second it follows
that
B sinVD=O.
B=fO, since otherwise we would have X==O and u==O, which
contradicts the hypothesis. Consequently, we must have
sinV~l=O,
whence
V'i\
"' = T (n = l , 2, ... )
mt
(12)
(we do not take the value n = 0, si nee then we would have X= 0

=
and u 0). And so we have
. nn
X = B smyx. (13)
These values of 'A are called eigenvalues of the given boundary-

value problem. The functions X (x) corresponding to them are
called eigenfunctions.
Note. If in place of -'A we took the expression +I.= k\ then
equation (8) would take the form
X"-k 1 X =0.
The genera1 solution of this equation is

X=Aek"+Be-k".
A nonzero solution in this form cannot satisfy the boundary
conditions (2) and (3).
Knowing V~ we ran [utilising (11)] write
ann . ann
T(t)=ecos- 1
t+Dsm-1
t (n=1, 2, ... ). (14)
For each value of n, hence for every 'A, we put the expressions
(13) and (14) into (6) and obtain a solution of equation (1) that
satisfies the boundary conditions (2) and (3). We denote this so-
lution by un (x, t):
. nn ' ( ann . ann )
un(x, t)=smTx encos-1
-t+Dnsm- -t. (15)
1
For each value of n we can take the constants e
and D and thus
write en and Dn (the constant B is included in en and Dn).
Since equation (1) is linear and homogeneous, the sum of the
solutions is also a solution, and therefore the function represent-
ed ~y the series
rD
u (x, t) = ~ un (x, t)
n=t -
or
~ ( ann . ann ) . nn
u(x, t)=,f;t, encos-
1 t+Dnsm-1 t smTx (16)
will likewise be a solution of the differential equation (1), which

will satisfy the boundary conditions (2) and (3). Series (16) will
obviously be a solution of equation (1) only if the coefficients
en and Dn are such that this series converges and that the series
resulting from a double term-by-term differentiation with respect
to x and to t converge as well.
The solution (16) should also satisfy the initial conditions (4)
and (5). We shall try to do this by choosing the constants en
and Dn. Substituting into (16) t = 0, we get [see condition (4)]:
"'
f (x) = Len sin n~ X. (17)
n=t
If the function f (x) is such that in the interval (0, l) it may be

expanded in a Fourier series (see Sec. 1, Ch. XVII), the condition
(17) will be fulfilled if we put
I
en= ; sf (X) sin n~ X dx. (18)

The Equation for Propagation of Heat in a Rod 823
We then differentiate the terms of (16) with respect to t and

substitute t = 0. From condition (5) we get the equality
00
~ ann . nn
<p (x} = ....... Dn - 1- sm Tx.
n=t
We define the Fourier coefficients of this series:

I
anl't 2 (' . nn
Dn - 1-=-y J <p (x} smTxdx
0
or
I
Dn = a!l'tS <p (x) sin nt x dx. (19)

0
Thus, we have proved that the series (16), where the coefficients
Cn and Dn are defined by formulas (18) and (19) [if it -admits
double termwise differentiation], is a function u (x, t), which is
the solution of equation (l) and satisfies the boundary and initial
conditions (2) to (5).
Note. Solving the problem at hand for the wave equation by
a different method, we can prove that the series (16) is a solution
even when it does not admit termwise differentiation. In this case
the function f (x} must be twice differentiable and <p (x) must be
once differentiable*).
SEC. 4. THE EQUATION FOR PROPAGATION OF HEAT IN A ROD.

FORMULATION OF THE BOUNDARY-VALUE PROBLEM
Let us consider a homogeneous rod of length l. We assume

that the lateral surface of the rod is impenetrable to heat transfer
and that the temperature is the same
at all points of any cross-sectional area
of the rod. Let us study the process of o
propagation of heat in the rod. Fig. 373.
We place the x-axis so that one
end of the rod coincides with the
point x=O, the other with the point x=l (Fig. 373). Let
u (x, t) be the temperature in the cross section of the rod with
abscissa x at time t. Experiment tells us that the rate of prop a-
These conditions are dealt with in detail in "Equations of Mathematical

Physics", A. N. Tikhonov and A. A. ~amarsky, Gostekhizdat, 1954 (Russian
edition).
gation of heat (that is, the quantity of heat passing through a

cross section with abscissa x in unit time) is given by the formula
q= - kau
-S
ax ( 1)
where S is the cross-sectional area of the rod and k is the coef-

ficient of heat conduction*).
Let us examine an element of rod contained between cross
sections with abscissas X 1 and X 2 (x 2 - X 1 = ~x). The quantity of
heat passing through the cross section with abscissa x1 during
time M will be equal to
~Q 1 = -ki!!!:l
ax x=x, SM ' (2)
and the same for the cross section with abscissa X2 !
~Q~ = -k au I
a-X x=x,SM. (3)
The influx of heat ~Q 1 - ~Q 2 into the rod element during time

M will be
~Q 1 - ~Q z = [-k au/
ax x=x,SM]- [-k au'
ax x=x.SM]::::::
azu
::::::k axz ~xSM (4)
(we applied the Lagrange theorem to the difference :; /x=x,-

- :; /x=x.).This influx of heat during time M was spent in
raising the temperature of the rod element by ~u:
~Q 1 -~Q 2 = CQ~XS~u
or (5)
where c is the thermal capacity of the substance of the rod and

Q is the density of the substance (Q~XS is the mass of an element
of rod).
") The rate of propagation of heat, or the rate of the thermal flux, is
determined by
. /!.Q
q= I 'mar
v1here I!,.Q is the quantity of heat that has passed through a cross section S
during a time M.
Heat Propagation in Space 825
Equating expressions (4) and (5) of one and the same quantity
of heat llQ 1 - llQ 2 , we get
k
azu s au
axz !lxSM = CQllX at M
or
au k azu
at =ce ax=
Denoting !:.._ =a 2 , we finally get
CQ
(6)
This is the equation for the propagation of heat (the equation of

heat conduction) in a homogeneous rod.
For the solution of equation (6) to be definite, the function u (x, t)
must satisfy the boundary-value conditions corresponding to the
physical conditions of the problem. For the solution of equation (6),
the boundary-value conditions may differ. The conditions which
correspond to the so-called first boundary-value problem for O:s=;;;t :s;;;;T
are as follows:
u (x, 0) = q> (x), (7)
u (0, t) = 'i't (f), (8)
u(l, t)='!' 2 (t). (9)
Physically, condition (7) (the initial condition) corresponds to
the fact that for t = 0 a temperature is given in various cross
sections of the rod equal to q> (x). Conditions (8) and (9) (the
boundary conditions) correspond to the fact that at the ends of
the rod, x=O and x=l, a temperature is maintained equal to
'1' 1 (t) and 'i'z (t), respectively.
It is proved that the equation (6) has only one solution in the
region O:s;;;;x:s;;;;l, O:s;;;;t:s;;;;T, which satisfies the conditions (7), (8),
and (9).
SEC. 5. HEAT PROPAGATION IN SPACE
Let us further consider the process of propagation of heat in

three-dimensional space. Let u (x, y, z, t) be the temperature at
a point with coordinates (x, y, z) at time t. Experiment states
that the rate of heat passage through an area !ls, that is, the
quantity of heat passing through in unit time is governed by the
formula [similar to formula q) of the preceding section]
!l Q = - k an
au !ls ' (1)
where k is the coefficient of heat conductivity of the medi urn

under consideration, which we regard as homogeneous and isotro-
pic, n is the unit vector directed normally to the area t::.s in the
direction of motion of the heat. Taking advantage of Sec. 14,
Ch. VIII, we can write
au au au au
an= ax cos a+ ay cos~ +az cosy,
where cos a, cos~. cosy are the direction cosines of the vector n, or
au
Fti = ~tgradu.
Substituting the expression ~ into formula (1), we get

!::.Q =-kngrad u t::.s.
The quantity of heat passing in time M through the elementary
area !::.s wi 11 be
!::.QM = - kngrad u M !::.s.
Now let us return to the problem posed at the beginning of
the section. In the medium at hand we pick out a small volume V
bounded by the surface S. The quantity of heat passing through
the surface S will be
Q=-M ~~ kngraduds, (2)
s
where n is the unit vector directed along the external normal to
the surface S. It is obvious that formula (2) yields the quantity
of heat entering the volume V (or leaving the volume V) during
time M. The quantity of heat entering V is spent in raising the
temperature of the substance of this volume.
Let us consider an elementary volume t::.v. Let its temperature
rise by t::.u in time M. Obviously, the quantity of heat expended
on raising the temperature of the element t::.u will be
au
c!::.uQ Au:::::c!::.uQ at M,
where c is the heat capacity of the substance and Q is the density.

The total quantity of heat consumed in raising the temperature
in the volume V during time At will be
At sss
v
CQ ~~ du.
Heat p;opagation in Space 827
But this is the heat that has entered the volume V during the
time At; it is defined by formula (2). Thus, we have the equality
M 55 kngraduds=M 555 CQ~~ dv.

s v
Cancelling out !J..t, we get
55s kngraduds= 555

v
CQ~~ dv. (3)
The surface integral on the left-hand side of this equation we

transform by the Ostrogradsky formula (see Sec. 8, Ch. XV),
assuming F = k grad u:
~ ~ (k grad u) n ds = ~
s
Hdiv (k grad u)dv.
v
Replacing the double integral on the left of (3) by a triple inte-
gral, we get
555 div(kgradu)dv 555 CQaautdv
V
=
V
or
55v 5[div (k grad u)-CQ ~~] dv = 0. (4)
Applying the mean-value theorem to the triple integral on the

left (see Sec. 12, Ch. X IV), we get
[ div(k grad U)-CQ a~] x=x,, y=y,, z=z, = 0, (5)

where the point P (x, y, z) is some point of the volume V.
Since we can pick out an arbitrary volume V in three-dimen-
sional space where propagation of heat is taking place, and since
we assume that the integrand in (4) is continuous, equality (5)
will be fulfilled at each point of the space. Thus,
CQ ~~ = div (k grad u). (6)
But
au k au au
kgra d tt=kaxl+ ayJ+k azk
(see Sec. 14, Ch. VIII) and
div (k grad u) = :x (k ~~) +:U (k ~~) +! (k ~~)

(see Sec. 9, Ch. XV). Substituting into (6), we obtain

cQat -ax (k au)+~
au-~ ax ay (k ay az (k au)
au)+~ az . (7)'
If k is a constant, then
a2u aau o2u)
div (k grad u) = k div (grad u) = k ( axz+ ayz+ aza
and equation (6) then yields
au (a u a u a u)
CQ at = k ax + ayz + dz
2
2
2 2
au (a u a u a u)
2 2 2
ax + ay + az
2
(jj =a 2 2 2
(8)
Equation (8) is briefly written

au 2A
at= a utt,
azu aau a2u
where !:J.u =ax:+ aya+ aza is the Laplace operator. Equation (8)
is the equation of heat conduction . in space. To find its unique
solution that corresponds to the problem posed here, it is necessary
to specify the boundary-value conditions.
Let there be a body Q with a surface cr. In this body we con-
sider the process of propagation of heat. At the initial time the
temperature of the body is specified, which means that the solu-
tion is known for t = 0 (the initial condition):
u (x, y, z, 0) = qJ (x, y, z). (9
In addition to that we must know the temperature at any point M
of the surface cr of the body at any time t (the boundary condi-
tion):
u (M, t) = 'iJ (M, t). (10)
(Other boundary conditions are possible too.)
If the desired function u (x, y, z, t) is independent of z, which
corresponds to the temperature being independent of z, we obtain
the equation
( 11)
which is the equation of heat propagation in a plane.

If we consider heat propagation in a flat regi.on D with bound-
ary C, then the boundary conditions, like (9) and {10), are
The First Boundary-Value Problem for: the Heat-Conductivity Equation 829
formulated as follows:
u (x, y, 0) = cp (x, y),
u (M, t) = 'Jl (M, t),
where cp and 'Jl are specified functions and M is a point on the
boundary C.
But if the function u does not depend either on z or on y,
then we get the equation
iJu 2 au
at= a iJx 2
which is the equation of heat propagation in a rod.
SEC. 6. SOLUTION OF THE FIRST BOUNDARY-VALUE

PROBLEM FOR THE HEAT-CONDUCTIVITY EQUATION
BY THE METHOD OF FINITE DIFFERENCES
'
When we solve partial differential equations by the method of
finite differences, the derivatives, as in the case of ordinary
differential equations, are replaced by appropriate differences
(see Fig. 374):
iJu(x, t) u(x+h. t)-u(x, t)
ox ;::::: h '
( 1)
iJ 2u(x, t)~_!_ [u(x+h, t)-u(x, t)_u(x, t)-u(x-h,
ox - h h h
t)]
or
iJ 1u(x, t)_u(x+h. t)-2u(x, t)+u(x-h, t).
a~ - ~ ' (2)
similarly,
iJu (x, t) u (x, t + l) -u (x, t)
at t (3)
The first boundary-value problem for the heat-conductivity equa-
. tion is stated (see Sec. 4) as follows. It is required to find the
solution of the equation
iJu 2 azu
at =a iJJc 2 (4)
that satisfies the boundary-value conditions

u(x, O)=cp(x), O~x~L. (5)
u(O, !)='Jl1 (t), O~t~T, (6)
u(l, t)='IJ 2 (t), O~t~T. (7)
that is, we have to find the solution u (x, t) in a rectangle boun-
ded by the straight I ines t = 0, x = 0, x = L, t ~ T, if the values
of the desired function are given on three of its sides: t = 0,

x = 0, x = L (Fig. 375). We cover our region with a grid formed
by the straight lines
X=ih, i= I, 2, ... ,
f=kl, k=l, 2, ... ,
and approximate the values at the nodes of the grid, that is, at
the points of intersection of these lines. Introducing the notation
t
t
r~.t~~ 'i,M
I (i-f.k) i k) {t-f.k.
(x-h,t) (x,t) (x+h,t)
X 0 l.x
Fig. 374. Flg. 375.
u (ih, kl) = U;
,., we write lin place of equation (4)] a corresponding
difference equation for the point (ih, kl). In accord with (3)
and (2), we get
Uf, k+t-Uf, k 1 Uf+t, k-2u;, k +ui-t, k
l =a ha (8)
We determine u1, Ht:
U;, k+t = ( 1- 2~:1) U;, k + aa ~ (u; +t, k + U;-,, ,.). (9)
From (9) it follows that if we know three values in the kth

row: u1, ,., u1+,, ,., u 1_,, ,., we can determine the value u1, k+t in
+
the (k 1) st row. We know all the values on the straight line
t = 0 [see formula (5)). By formula (9) determine the values at
all the interior points of the segment t = l. We know the values
of the end points of this segment by virtue of (6) and (7). In this
way, row by row, we determine the values of the desired solution
at all nodes of the grid.
It is proved that from formula (9) we can obtain an approxi-
mate value of the solution not for an arbitrary relationship between
~ 2~ 2
2
the steps h and l, but only if l Formula (9) is greatly sim-

plified if the step l along the t-axis is chosen so that
l-2a21 =0
. h'
Propagation of Heat in an Unbounded Ro-t 831
or
h
l = 2a 2 '
In this case, (9) takes the form

1
ul, k+t =2(ui+t. k+ui-t k). (10)
This formula is particularly convenient for computations (Fig. 376).

This method gives the solution at the nodes of the grid. Solutions
between the nodes may be obtained, for exam-
EHJ
ple, by extrapolation, by drawing a plane
through every three points in the space (x, t, u).
Let us denote by uh (x, t) a solution obtained
by formula (10) and this extrapolation. It is
proved that (i-f,k) 7' (i+f,k)
lim uh (x, t) = u (x, t), Fig. 376.
h-+0
where u (x, t) is the solution of our problem. It is also proved*)

that
Iuh (x, t)-u (x, t) I<Mh\
where M is a constant independent of h.
SEC. 7. PROPAGATION OF HEAT IN AN UNBOUNDED ROD
Let the temperature be given at various sections of an unboun-

ded rod at an initial instant of time. It is required to determine
the temperature distribution in the rod at subsequent instants of
time. (Physical problems reduce to that of heat propagation in
an unbounded rod when the rod is so long that the temperature
in the interior points of the rod at the instants of time under
consideration are but slightly dependent on the conditions at
the ends of the rod.)
If the rod coincides with the x-axis, the problem is stated
mathematically as follows. Find the solution to the equation
(1}
*) This question is deaU with in more detail in D. Yu. Panov's "Ref-

erence on Numerical Solution of Partial Differential Equations'\ Gostekhizda~,
!951; Lothar Collatz, "Numerische Behandlung von Differentialgluchungen'~,
!955.
832 Eq;Jations of Mathematical Physics
in the region - oo < x < oo, 0 < t which satisfies the initial
condition
u (x, 0) = c:p (x). (2)
To find the solution, we apply the method of separation of
variables (see Sec. 3); that is, we shall seek a particular solution
of equation (1) in the form of a product of two functions:
u (x, t) =X (x) T (t). (3)
Putting this into equation (1) we have X (x) T' (t) =a 2 X" (x) T (t)
or
T'
- - - X'
a2T- -
X- - A I. (4)
Neither of these relations can be dependent eiUier on x or

on t; therefore, we equate them to a constant, *)-'A. 2 From (4)
we get two equations:
T' +a 2 'A. 2 T =0, (5)
X" +'A. 2 X =0. (6)
Solving them we find
T = Ce-a).t,
X= A cos'A.x+ B sin'A.x.
Substituting into (3), we obtain
u, (x, t) = e-az>.'zt [A ('A.) cos 'A.x + B ('A.) sin 'A.x] (7)
[the constant C is included in A ('A.) and in B ('A.)].
For each value of 'A. we obtain a solution of the form (7).
For each value of 'A. the arbitrary constants A and B have defi-
nite values. We can therefore consider A and B functions of 'A..
The sum of the solutions of form (7) is likewise a solution
[since equation (1) is linear]:
~ e-az>.zt [A ('A.) cos 'A.x + B ('A.) sin 'A.x].
Integrating expression (7) with respect to the parameter 'A. between 0

and oo, we also get a solution
""
u (x, t) = ~ e-a>.t [A ('A.) cos 'A.x + B ('A.) sin 'A.x} d'A., (8)
*) Since from the meaning of the problem T (t) musl be bounded for
any t, if <p (x) is bounded, it follows that ~ must be negative. And so we
write -'A'.
Propagation of Heat in an Unbounded Rod 833
if A (A.) and B (A.) are such that this integral, its derivative with
respect to t and the second derivative with respect to x exist
and are obtained by differentiation of the integral with respect
to t and x. We choose A (A.) and B (A.) such that the solution u (x, t)
satisfies the condition (2). Putting t = 0 in (8), we get [on the
basis of condition (2)):
00
u (x, 0) = cp (x) = ~ [A (A.) cos AX+ B (A.) sin AX) dA.. (9)
Suppose that the function cp (x) is such that it may be represented

by the Fourier integral (see Sec. 12, Ch. XVII):
00 "'
cp(x)=~ 5( S cp(a)cosA.(a-x)da)dA.
0 -00
or
00 "'
cp (x) = ~S [( S cp (u) cos A.a da) cos A.x +

0 -00
"'
+ ( S cp (a) sin Aada) sin Ax] dl.. (10)
-ao
Comparing the right sides of (9) and (10), we get

""
A(A)=* -~ ~(o)cosl.ado, }
(11)
B (A.)=~ 5cp (a) sin Au da.
-ao
Putting the expressions thus found of A (f.) and B (A.) into (8),
we obtain
00 "'
u (x, t) = ~ Se-az)..zt [( S cp (a) cos A.u da) cos l.x +

0 -00
+ao
+( ~ J
cp (a) sin A.a da) sin 'Ax dl. =
-ao
=~1'e-a~i.zt [~ cp(a)(cos"Aacos"Ax+sin"AasinA.x)da] dl.=

0 -00
00 00 .
=~ Se-a S cp(a) cosf.(a-x)da) dl.

2 21
" (
o -oo
27- 3388
or, changing the order of integration, we finally get
u (x, t) = ~ S"' [IJl(a) (S"' e-a ~ 2 21

J
cos 'A. (a-x) b'A.) da.. (12)
-co o
This Is the solution of the problem.

Let us transform formula (12). Compute the integral in the
parentheses:
J
0
"'
e-az).zt cos')... (a-x) d'A. =a ~t
"'
S
0
e-z= cos ~z dz. (13)
The integral is transformed by substitution:
a"'~ 11.
Y t=z,
~-A
a Vt-...,. (14)
We denote
""
K (~) = ~ e-z cos ~z dz. (15)
0
Differentiating, *) we get
""
K' (~) = - ~ e-zz z sin ~z dz.
0
Integrating by parts, we find

""
K' (~) =} [e-z sin ~z( -{ Se-z cos ~z dz
0
or
K' (~)= - { K(~).
Integrating this differential equation, we obtain
p
K (~)=Ce- 4 (16)
Determine the constant C. From (15) it follows that
J
"'
K (0) = e-z= dz = ~n
0
) Differentiation here is easily ]usUfied.

Propagation of Heat in an Unbounded Rod 835
(see Sec. 5, Ch. XIV). Hence, in (16) we must have

C = Vii
2
And so
v-
K(~)= 2ne--.:.
fl
(17)
Put the value (17) of the integral (15) into (13):
In place of ~ we substitute its expression (14) and finally get the

value of the integral (13):
(18)
Putting this expression of the integral into the solution (12), we

finally get
l s"' _(a-x)'
u(x,t)= ,r- rp(a)e a't da. (19)
2a r nt
This formula, called the Poisson integral, is the solution to

the problem of heat propagation in an unbounded rod.
Note. It may be proved that the function u (x, t), defined by
integral (19), is a solution of equation (l) and satisfies condition
(2) if the function rp (x) is bounded on an infinite interval
(-oo, oo).
Let us, establish the physical meaning of formula (19). We
consider the function
0 for -oo < x<x 0 ,
rp* (:c)= rp (x) for X 0 ~ x,.;;; x 0 + !lx, (20)
{ 0 for :C 0 + !lx<x<oo.
Then the function
u* (x, t) = ~ l
2a nt Jl
s --
00
-<1>
rp* (a) e
(a-x)'
a't da (21)
is the solution to equation (1), which solution takes on the value

27*
rp* (x) when t =0. Taking (20) into consideration, we can write
Xo+ll.x (a-x) 2
u*(x, t)=~
2a Y nt
5 rp(a)e-4li'tda.
Xo
Applying the mean-value theorem to the latter integral, we get
<;-x)
u*(x t)=cp()!1xe- 4 a't x 0 <t<x 0 +!J.x (22)
' 2a Y nt ' "'
Formula (22) gives the value of temperature at a point in the

rod at any time if for t = 0 the temperature in the rod is every-
where u* = 0, with the exception of the interval [X 0 , x 0 L\x], +
where it is rp (x). The sum of temperatures of form (22) is what
yields the solution of (19). It will be noted that if Q is the linear
density of the rod, c the heat capacity of the material, then the
quantity of heat in the element [X 0 , X 0 +!J.x] for t=O will be
!J.Q ~rp () !J.x QC. (23)
Let us now consider the function
('e,-xP
_l_e-4a't (24)
2a V nt '
Comparing it with the right side of (22) and taking into ac-
count (23), we may say that it yields the temperature at any
point of the rod at any instant of time t if for t = 0 there was
an instantaneous heat source with quantity of heat Q =CQ in the
cross section (the Iimiti ng case as !J.x-+ 0).
SEC. 8. PROBLEMS THAT REDUCE TO INVESTIGATING

SOLUTIONS OF THE LAPLACE EQUATION. STATING
BOUNDARY-VALUE PROBLEMS
In this section we shall consider certain problems that reduce

to the solution of the Laplace equation:
azu au au
ax+ay+az=O. (l)
As already pointed out, the left side of equation (1),
au au au
ax + ay + az = !J.u
is called the Laplacian operator. The functions u which satisfy the
Laplace equation are called harmonic functions.
I. A stationary (steady-state) distribution of temperature in a
homogeneous body. Let there be a homogeneous body Q bounded
The Laplace Equal ion 837
by a surface a. In Sec. 7 it was shown that the temperature at

various points of the body statisfies equation (8):
au au au au)
ox + oy + OZ
2 (
iJt = a 2
If the process is steady-state, that is, if the temperature is not

dependent on the time, but only on the coordinates of the points
of the body, then ~~ = 0 and, consequently, the temperature
satisfies the Laplace equation
iJ u iJ u iJ u
2 2 2
ox+ay-1-az=O. (I)
To determine the temperature in the body uniquely from this
equation, one has to know the temperature of the surface cr. Thus,
for equation (1), the boundary-value problem is formulated as
follows.
To find the function u (x, y, z) that satisfies equation (I) inside
the volume Q and that takes on specified values at each point M
of the surface cr:
u I.= 'i' (M). (2)
This problem is called the Dirichlet problem or the first boundary-
value problem of equation (1).
If the temperature on the surface of the body is not known,
but the heat flux at every point of the surface is, which is pro-
portional to ~~ (see Sec. 5), then in place of the boundary-value
condition (2) on the surface a we will have the condition
~~I.= 'I'* (M). (3)

The problem of finding the solution to (1) that satisfies the boun-
dary-value condition (3) is called the Neumann problem or the
second boundary-ualue problem. .
If we con.sider the temperature distribution in a two-dimensi-
onal region D bounded by a contour C, then the function u will
depend on two variables x and y and will satisfy the equation
(4)
wh!ch is called the Laplace equation in a plane. The boundary-

value conditions (2) and (3) must be fulfilled on the contour C.
II. The potential flow of a fluid. Equation- of continuity. Let
there be a flow of liquid inside a volume Q bounded by a sur-
face a (in a particular case, Q may also be unbounded). Let Q
be the density of the liquid. We denote the velocity of the liquid by

v = Vxi + vy}+ vzk, (5)
where vx, vy, Vz are the projections of the vector v on the coor-
dinate axes. In the body Q pick out a small volume w, bounded
by the surfaceS. The following quantity of liquid will pass through
each element !ls of the surface S in a time M:
llQ =Vn llsQ M,
where n is the unit vector directed along the outer normal to
the surfaceS. The total quantity of liquid Q entering the volume w
(or flowing out of the volume w) is expressed by the integral
Q=M~~Qvnds (6)
s
(see Sees. 5 and 6, Ch. XV). The quantity of liquid in the
volume (i) at time t was
During time M the quantity of liquid will change (due to

changes in density) by the amount
Q= sss llQ!lwR:JM sss~~dw. (7)
w w
Assuming that there are no sources in the volume w, we con-
clude that this change is brought about by an influx of liquid to
an amount that is determined by equation (6). Equating the right
sides of (6) and (7) and cancelling out M, we get
- SS Q'Vn ds = + SSS~~ dw. (8)
s w
We transform the iterated integra:! on the left by Ostrogradsky's
formula (Sec. 8, Ch. XV). Then (8) will assume the form
:-SSS div (Q'V)dw= SSS ~ dw
w w
or
SSS(~~ + div (Q'V)) dw = 0.
w
Since the volume w is arbitrary and the integrand is continuous
we obtain
~~ + div (Qv) ~ 0 (9)
The Laplace Equation 839
or
(9')
This is the equation of continuous flow of a compressible liquid.

Note. In certain problems, for instance when considering the
movement of oil or gas in a subterranean porous medium to a
well, it may be taken that
k
'V = -Q grad p,
where p is the pressure and k is the coefficient of permeability

and
A.=const. Substituting into the continuity equation (9), we get
A. a: -div (k grad p) = 0
or
A.
at =~-(k
op ax op) ay (k ap)
ax +i az (k ap)
ay +~ az (10)
If k is a constant, then this equation takes on the form

op k (ap ap ap)
at= T ax + ay + az (11)
and we arrive at Fourier's equation.

Let us return to equation (9). If the liquid is noncompres-
sible, then Q = const, ~; = 0, and (9) becomes
div ('D)= 0. (12)
If the motion is potential, that is, if the vector 'V is a gradient
of some function cp:
'V= grad cp,
then equation (12) takes the form

div (grad cp) =0
or
(13)
that is, the potential function of the velocity cp must satisfy the
Laplace equation.
In many problems, as, for example, those dealing with filtra-

tion, we can put
'V= - k 1 grad p,
where p is the pressure and k 1 is a constant; we then get the

Laplace equation for the determination of the pressure:
(13')
The boundary-value conditions for equation (13) or (13') may

be the following:
I. On the surface a are specified the values of the desired
function p-pressure [condition (2)1. This is the Dirichlet problem.
2. On the surface a are specified the values of the normal
derivative~~; the flow through the surface is specified [condition (3)].
This is the Neumann problem.
3. On parts of the surface u are specified the values of the
desired function p-pressure, and on parts of the surface are
specified the values of the normal derivative ~~-the flow through
the surface. This is the Dirichlet-Neumann problem.
If the motion is two-dimensional-parallel- that. is, the func-
tion cp (or p) does not depend on z- then we get the Laplace
equation in a two-dimensional region D with boundary C:
(14)
Boundary-value conditions of type (2), the Dirichlet problem,

or of type (3), the Neumann problem, are specified on the con-
tour C.
Ill. The potential of a steady-state electric current. Let a ho-
mogeneous medium fill some volume V, and let an electric cur-
rent pass through it whose density at each point is given by the
vector J(x, y, z)=Jxi+Jyj+Jzk. Suppose that the current den-
sity is independent of the time t. Further assume that there are
no current sources in the volume under consideration. Thus, the
flux or a vector J through any closed surface S lying inside the
volume V will be equal to zero:
ns
Jt~ds=O,
where n is a unit vector directed along the outer normal to the

surface.
The Laplace Equation in Cylindrical Coordinates 84l
From Ostrogradsky's formula we conclude that

divJ=O. (15)
The electric force E in the conducting medium at hand is, on the
basis of Ohm's generalised law,
J
E=r (16)
or
J='AE,
where 'A is the conductivity of the medium, which we shall con-
sider constant.
From the general electromagnetic-field equations it follows
that if the process is stationary, then the vector field E is irro-
tational, that is, rot E ~-= 0. Then, like the case we had when
considering the velocity field of a liquid, the vector field is po-
tential (see Sec. 9, Ch. XV). There is a function <p such that
E =grad <p. (17)
From ( 16) we get
J ='A grad <p. (18)
From (15) and (18) we have
'A div (grad <p) = 0
or
a~p a~p a~p
ax+ ayz+ az =0. (19)
We get the Laplace equation.
Solving this equation for appropriate boundary-value conditions,
we find the function <p, and from formulas (18) and (17) we find
the current J and the electric force E.
SEC. 9. THE LAPLACE EQUATION IN CYUNDRICAL COOIWINATES.

SOLUTION OF THE DIRICHLET PROBLEM FOR A RING WITH
'CONSTANT VALUES OF THE DESIRED FUNCTION ON THE INNER
AND OUTER CIRCUMFERENCES
Let u (x, y, z) be a h.armonic function of three variables. Then

au au au 0 (1)
ax+ ay+ az =
We introduce the cylindrical coordinates (r, cp, z):
x=rcos<p, x=rsill<p, Z=Z,
whence
r= V x + yZ, , and z,

we arrive at the function u *:
u(x, y, z)=u*(r, q>, z).
Let us find the equation that will be satisfied by u * (r, q>, z)
as a function of the arguments r, q>, and z; we have
i!!!- au i)r + i)u* i)cp
ax - i)r ax i)cp ax
auau (a' ) 2 au ar 2 i)r i)cp ar ax
iJ2u* i)cp au* (i)cp )' au ac:p
ax - ar ax + ar + ax ax + Oq> 2 ax + a(p ax (3)
similarly,
au au (a')' au ar au or iJq> i)2u* (i)q> ) 2 au iJ22 i)y + acp ay (4}
besides,
(5)_
We find the expressions for

ar iJr ar iJ2r iJ<p i)cp acp acp
ax . i)y .ax t ay ax i)y ax ay
from equations (2). Adding the right sides of (3), (4) and (5),
and equating the sum to zero [since the sum of the left-hand
sides of these equations are zero by virtue of (1)], we get
(6)
This is the Laplace equation in cylindrical coordinates.

If the function u is independent of z and is dependent on x
and y, then the function u*, dependent only on r and q>, satisfies
the equation
iJ2u 1 iJu* I a2 u*
ar +r i)r + "';:z iJ<p2 = 0, (7)
where r and q> are polar coordinates in a plane.
Now let us find the solution to Laplace's equation in the region
D (ring) bounded by thecirclesC 1 :x'+y"=R: and C,.:x'+y"=R!
with the following boundary values imposed:
uiC 1 =U 1 , (8)
uiC. =U 2,
(9)
where U1 and u1 are constants.
The Solution of Dirichlet's Problem for a Circle 843
We will solve the problem in polar coordinates. Obviously, it

is desirable to seek a solution that is independent of q>. Equation
(7) in this case takes the form
au +_!__au- 0
ar r ar- .
Integrating this equation we find
u=C 1 lnr+C. (10)
We determine C 1 and c. from conditions (8) and (9):
u 1 =C 1 lnR 1 +C.,
u.=C 1 lnR.+C .
: Whence we find
R1
C =u 2 -u 1 C2 =tl 1 - (U 2 -u 1 ) In
~.
R '
In---..! In~
R. R.
Substituting the values of C 1 and c. thus found into (10), we
finally get
In!....
U=u 1 + :(u 2 -u 1 ). (11)
In---..!
R.
Note. We have actually solved the following problem. To find
the function u that satisfies the Laplace equation in a tegion
bounded by the surfaces (in cylindrical coordinates)
r=Rt> r=R., z=O, z=H,
and that satisfies the following boundary conditions :
UJr=R,=Ut> UJr=R,=tl 2 ,
~~
8z z=o =0 1 au'
oz z=H =0
(the Dirichlet-Neumann problem). It is obvious that the desired
solution does not depend either on z or on q> and is given by
formula (11).
SEC. 10. THE SOLUTION OF DIR.ICHLET'S PR.OBLEM

FOR. A CIR.CLE
In an xy-plane, let there be a circle of radius R with centre

at the origin and let there be a certain function f (q>), where q>
is the polar angle, be given on its circumference. It is required
to find the function u (r, q>) continuous in the circle (including
the boundary) and satisfying (inside the circle) the Laplace equa-
tion 2
cPu o u
-
ax+ -oy-
-0 (1)
and, on the circumference, assuming the specified values
U lr=R = f (cp). (2)
We shall solve the problem in polar coordinates. R.ewrite equation
(1) in these coordinates:
au + 1 au+ 1 au - 0
or r or ,.. OqJ 2-
or
( 1')
We shall seek the solution by the method of separation of variables,

placing
u=<D(cp)R(r). (3)
Substituting into equation (1 '), we get
r" (cp) R" (r) + r (cp) R' (r) + <D" (cp) R (r) = 0
or
"(<p) r 2 R"(r)+rR'(r) 1
 (qJ) = . R (r) = -k ' (4)
Since the left side of this equation is independent of r and the
right is independent of cp, it follows that they are equal to a
constant which we denote by -k 2 Thus, equation (4) yields two
equations:
<D" (cp) + k 2 <l> (cp) = 0, (5)
rR" + rR' -kR = 0. (5')
The complete integral of (5) will be
<l>=A coskcp+B sin kcp. (6)
We seek the solution of (5') in the form R =r"'. Substituting
R = r"' into (5'), we get ,
+
rm (m-1) r"'- rmrm- 1 - k 2 r"' = 0
or
m2 -k 2 =0.
We can write two particular linearly independent solutions rk and
,-k. The general solution of equation (5') is
R = Crk+ Dr-k. (7)
We substitute expressions (6) and (7) into (3):
+
uk= (AI! cos kcp Bk sin kcp) (Ckrk+ Dkr-k). (8)
The Solution of Dirichlet's Problem for a Circle 845
Function (8) will be the solution of (1') for any value of k different
from zero. If k = 0, then equations (5) and (5') take the form
"= 0, rR" +R' = 0,
and, consequently,
U 0 = (A 0 +B 0 <p) (C 0 +Do lnr). (8')
The solution must be a periodic function of <p, since for one and
+
the same value of r for <p and <p 2n we must have the same
solution, because one and the same point of the circle is considered.
It is therefore obvious that in formula (8') we must have B 0 = 0.
To continue, we seek a solution that is continuous and finite in
the circle. Hence, in the centre of the circle the solution must
be final for r= 0, and for that reason we must have D 0 =.Pin (8') and
Dk=O in (8).
Thus, the right side of (8') becomes the product A 0 C 0 , which
we denote by A0 /2. Thus,
A
Uo=2
0 (S")
We shall form the.solution to our problem as a sum of solutions
of the form (8), since a sum of solutions is a solution. The sum
must be a periodic function of <p. This will be the case if each
term is a periodic function of <p. For this, k must take on integral
values. [We note that if we equated the sides of (4) to the number
+ k\ we would not obtain a periodic so!u tion.] We shall confine
ourselves only to positive values:
k = 1, 2, ... , n, ...
because the constants A, B, C, D are arbitrary and therefore the
negative values of k do not yield new particular solutions.
Thus, ,
u (r, <p) = ~o + L."' (A,. cos n<p + B,. sin n<p) r"
n=l
(9)
(the constant C,. is included in A,. and B,.). Let us now choose
arbitrary constants A,. and B,. so as to satisfy the boundary-value
condition (2). Putting into (9) r==;:R, we get, from condition (2),
""
f (<p) = ~o + L, (A,. cos n<p + B,. sin n<p) R". (10)
n =t
For us to have equality (10), it is necessary that the function
should be expandable in a Fourier series in the interval ( -n, n)
and that A,.R" and B,.R" should be its Fourier coefficients. Hence,
An and Bn must be defjned by the formulas

1t
A. ~ n~" ~/(I) cos nt dt, }

(11)
Bn = n~n 5 f (t) sin nt dt.
-n
Thus, the series (9) with coefficients defined by formulas (11)
will be the solution of our problem if it admits termwise iterated
differentiation with respect to r and IP (but we have not proved
this). Let us transform formula .(9). Putting, in place of An and
JJ~, their expressions (11) and performing the trigonometric trans-
formations, we get
1t "' 1t
1
u(r, IP)=2 n5t(t)dt+~ L. St(t)cosn(t-lj))dt(~r=
-n n=-n
- 2~ 1 f (t) [ 1+ 2 ~. ( i r cos n (t-lj))] dt. (12)
Let us transform the expression in the square brackets: *

1+ 2
00
L. (i
n=l
r cos n (t -lj)) = 1 +
00
L. ( ~ r[ein
n=t
(1-rp) + e-in (l-rp)1=
.!....el<t-rp> .!....e-l<t-rp>
= 1 +_R_ _ _ _ + R =
1-...c__etU-'P> 1-....C e-1(1-rp)
R R
1-(..C..Ri\" _ R2 - r2
----------~~'--~-.~ (13)
= r
l-2Rcos(t-cp)+
2
( r )
R - R2 -2Rr cos (t -cp) +r 2 '
*) In the derivation we determine the sum of an. infinite geometric prog-

ression whose ratio is a complex number the modulus of which is less than
unity. This formula of the sum of a geometric progression is derived in the
same way as in the case of real numbers. It is also necessary to take into
account the definition vf the limit of the complex function of a real argu-
ment. Here, the argument is n (see Sec. 4, Ch. VII).
Solution of Dirichlet's Problem by Method of Finite DifferencJs 847
Replacing the expression in square brackets in (12) by expres

sion (13), we get
(14)
Formula (14) is called Poisson's integral. By an analysis of this

formula it is possible to prove that if the function f (cp) is con
tinuous, then the function u (r, cp) defined by the integral (14)
also satisfies equation (I') and u(r, cp)-f(cp) as r-R.. That is,
it is a solution of the Dirichlet problem for a circle.
SEC. 11. SOLUTION OF THE DIRICHLET PROBLEM

BY THE METHOD OF FINITE DIFFERENCES
In an xyplane, let there be given a region D bounded by

a contour C. Let there be given a continuous function f on the
contour C. It is required to find an approximate solution to
the Laplace equation
(1)
that satisfies the boundary condition

uic=f. (2)
We draw two families of straight lines:
x=ih and y=kh, (3)
where h is the given number, and i and k assume successive
integral values. We shall say that the region D is covered with
a grid. We call the points of intersection of the straight lines
nodes of the grid.
We denote by U;, k the approximate value of the desired
function at the point x=ih, y=kh; that is, u(ih, kh)=U;,k
We approximate the region D by the grid region D*, which
consists of all the squares that lie completely in D and of some
that are crossed by the boundary C (these may be disregarded).
Here, the contour C is approximated by the contour C *, which
consists of segments of straight lines of type (3). In each node
lying on the contour C* we specify the value f*, which is equal
to the value of the function f at the closest point of the con
tour C (Fig. 377).
The values of the desired function will be considered only at
the nodes of the grid. As has already been pointed out in Sec. 6,
the derivatives in this approximate method are replaced by finite

differences:
au
2
ax
I 2
- Ui+t, ~-2Uf, k+ui-1, k
h2
I -
X=ih, y=kh- '
au2
Ui,k+t-2ul,k+ui,k-l
ay 2
X=ih, Y=kh- h2
The differential equation (1) is replaced by a difference equation

(after cancelling out h 2 ):
ui+!, k-2ui, k+ui-,, k+ui, k+,-2ui,k+ui, k-t =0
or (Fig. 378)
l
+ +
ui, k = 4 (ui+1. k ui, k+t ui-t, k + ui, k-t). (4)
For each node of the grid lying inside D* (and not lying on the
boundary C*), we form an equation (4). If the point (x= ih,.
y=kh) is adjacent to the point of the contour C*, then the
right side of (4) will contain known values of f*. Thus, we
obtain a nonhomogeneous system of N equations in N unknowns,
where N is the number of nodes
u. c c of the grid lying inside the
I
I
-region D*.
We shall prove that the sys-
tem (4) has one, and only one,
(i,k+fJ
v
II:
i-f,k) (i,k} i+f.k

!"' ~
i,k-1)
0 X 0 X
Fig. 377. Fig. 378.
solution. This is a system of N linear equations inN unknowns.

It has a unique solution if the determinant of the system is not
zero. The determinant of the system is nonzero if the homoge-
neous system has only a trivial solution. The system will be
homogeneous if f* = 0 at the nodes on the boundary of the
contour C*. We shall prove that in this case all the values u1 , k
at all interior nodes of the grid are equal to zero. Inside the
region, let there be u 1, k different from zero. For the sake of
definiteness, we suppose that the greatest of them is positive.
Let us designate it by 1, k > 0. u
Solution of Dirichlet's Problem by Method of Finite Differences 849
By (4) we write
U;, k = ! (ui+t. k + U;, k+t + ui-t. k + U;, k-1). (4')
This equation is possible only if all the values of u on the

right are equal to the greatest u;,
k' We now have five points
at which the values of the desired function are k If none of u;,
these points is a boundary point, then, taking one of them and
writing for it the equation (4), we will prove that at certain
other points tl)e value of the desired function will be equal to
u;,k Continuing in this fashion, we will reach the boundary and
will prove that at the boundary point the value of the function
will be equal to u;,
k which is contrary to the fact that f* = 0
at boundary points.
Assuming that inside the region there is a least negative
value, we will prove that on the boundary the value of the
function is negative, which contradicts the hypothesis.
And so system (4) has a solution which is unique.
The values u;, k defined from the system ( 4) are approximate
values of the solution of the Dirichlet problem formulated above.
It was proved that if the solution of the Dirichlet problem for
a given region D and a given function f exists [we denote it by
u (x, y)] and if u;, k is the solution of (4), then we have the relation
Iu (x, y)- u;, k I< Ah 2 (5)

where A is a constant independent of h.
Note. It is sometimes justified (though this has not been
rigorously proved) to use the following procedure for evaluating
the error of the approximate solution. Let u~~hk be an approximate
solution for a step 2h, u~~~ an approximate solution for a step h,
and let Eh (x, y) be the error of the solution u}~t Then we have
the approximate equality
E h (x, y) ~
1 (
3 u,. k - u,, k
(2h) (h) )
in the common nodes of the grids. Thus, in order to determine

the error of the approximate solution for a step h, it is necessary
to find the solution for a step 2h. One third of the difference
of these approximate solutions is the error evaluation of the
solution for a step (mesh-length) of h. This remark also refers
to the solution of the heat-conduction equation by the finite-
difference method.
Exercises on Chapter XV Ill

1. Derive an equation of torsional oscillations of a homogeneous cylin
drical rod.
Hint. The torque in a cross section of !he rod with abscissa xis determined
by lhe formula M = Gl ~~, where 8 (x, t) Is lhe angle of torque of a cross
section with abscissa x at time t, G is the shear modulus, and I is lhe polar
moment of inertia of a cross section of lhe rod.
Ans. ~;~
=a 2 ~:~ , where a 2= ~/ , and k is the moment of inertia of unif
length of ;the rod.
. d a sou
2 F m aze =a2 ax
,, e equatwn at
I t'1on of 'h aze 'h
, a. sal'1s fi1es 'h
i< e con
ditions 8(0, 1)=0, 8(/, t)=O, S(x,O)=<p(x),

ae
-a-(x. O)
-=0, where
1
28 0 x
)
<p (X=-~-
28 x f l
<p(x)=--0- + 28 0 or
1 2 o;;;;;xo;;;;;l.
Give a mechanical interpretation of fhe problem.
88 0 ~ (-l)k (2k+l)nx (2k+l)nat

Ans. 8 (x, t) =-;t2'~( 1
2k+l)'sn cos
1 1
k=o
3. Derive an equation of longitudinal oscillations of a homogeneous cylln
drical rod. '
Hint. If u (x, I) is the Iranslaflon of a cross section of ro1 with abscissa x
at lime t, then the lensile stress T in a cross section x Is defined by the
formula T = ES ~, whe r~ E is the elasticity modulus of ~he material and S
is the cross-sectional area of lhe rod.
Ans. ~~~
=a 2 ~:~ , where a 2 = ~ , and Q is lhe density of lhe rod maferial.
4. A homogeneous rod of length 21 was shorted by 2A. under [he action
of forces applied to its ends. At 1=0 it is free of forces acting externally.
Determine ,the displacement u (x, t) of a cross section of the rod with abscis .
sax at iime t (the mid-point of lhe axis of lhe rod has abscissa x=O).
"' (-l)k+l (2k+l)nx (2k+I)nat
SA."'
Ans. u (x, t) =iii~ (2k +
l)' sin 21 cos 21 .
k=o
5. One end of a rod of length 1 is fixed, lhe other end is acted upon by
a \ensile force P. Find the longitudinal oscillations of the rod if the force P
. "'
BPI"" (-l)n (2n+l)nx (2n+l)nat
does not operate when t =0. Ans. Esilt ~ (2n+ l)' sin 21 cos
21
11=0
(E and S as in Problem 3).
Exercises on Chapter XV Ill 851
aau a a2u .
6. Find a solution lo the equation at 2 =a axa thaf satisfies the con
ditions
u (0, t) =0, u (1, t) =A sin cot,
au (x, 0) -0
u (x, 0) =0, at - .
sin~l
a
is the solution tha' satisfies the conditions
v(O, t)=O, v(l, t)=O
(x, 0)
v (x, 0) = - w (x, 0), -.-a-t-=-
av
ow (x, 0)
at
(It is assumed thaf sin ~ l '/: 0.)
au
7. Find a solution to the equation 7i[ =a
2 ott
2
7fX! that satisfies the con
ditions
u (0. t) = 0, u (l, t) = 0, t > 0,
x when O.;;;;;x-;;;;; l
2
u(x,O)= l
{ l-x when
2 < x < l.
oo
_ .!!_ ~ (_ l)n -
(an+ 1 ) 2 n 2 a 2
12 t . (2n + I) n
Ans. h (x, t) -- n 2 ~ (
2n+ l) 2 e sm l
n=o
Hint. Solve the problem by the method of separation of variables.
8. Find a solution to the equation ~~ =a 2 ::~ that satisfies the con-
ditions
x (l-x)
u(O, t)=u(O, l)=O, u(x, 0)=
12
8 I (2n+t) 2n'a'l
8 ~ - l' I (2n+ I) rt..\'1
Ahs. u(x, t)=iti~ (2n+l)a e sn
1
n=o
9. Find a solution to lhe equation ~~ =a ~:~ ~ha~ saiisfies the con-

ditions
-;;----
au
vX
Ix=o
= 0, u (l, l) = u0 , u (x, 0) = cp (x).
Point out lhe physical meaning of lhe problem.

.
Ans. u (x, l) = u0 + L Ane -aA.~t cos (2n+I):n:
"'
21
x,
where An=T 2 s I
cp(x)cos (2n+ I) :rex dx

21
(- l)n4u 0
:n: (2n+ I)
Hint. Seek lhe solution in lhe form u=u 0 +v (x, t).

I t'1on t o ll1e equa t'wn Tt=a
. d a sou
10 . F 1n au ox lhat sa t1s fi1es .th e con-
2 au
ditions
u(O, 1)=0, ~u
vX
I x=L
=-Hul
x=l
, u(x, O)=cp(x).
Point out the physical meaning of lhe problem.

"' z ~~~a'l'
~ A P + lln
A ns. u (x, I)=.(....,. n e- -1-,- sln-
J.lnX
-,
11=1 p (p + 1) + ).l~ 1
I
where An=i Scp (x)sin J.l,t dx, p=Hl, ).l 1 , ).l 2 , . . . , fln are positive roots
0
of the equation tan f.!=- J:..

p
Hint. At the end of the rod (when x= l) a heat exchange occurs with the
environment, which has a temperature of zero.
11. Find [by formula (10), Sec. 6, putting h=0.2] an approximate solution
.
to e equa t wn
th au = 2 ox
Tt au "
h a,. sa t'IS fites lh e con d't'
1 1ons
U. (X, 0) =X ( ~ -X ) , U (0, t) = 0, U (1, t) = ; , 0,.;;; t,.;;; 4l,
au
12. Find a solution lo ~he Laplace equation ax + au .
oy =0, in a strip
0:;::;;; x:;::;;; a, 0 .so;;; y < oo that satisfies lhe conditions
u (0, y)=O, u (a, y)=O, u (x: O)=A ( 1-~), u (x, co)=O.
2A "' 1 -~u nnx

Ans. u (x, t)=- ~- e a sin-.
:n: .(..... n a
n=l
Hint. Use the method of the separation of variables.
Exercises on Chapter XV Ill 853
13. Find a solution lo the Laplace equation : ; + ~;~ = 0 in the rec

tangle 0 ~ x ~a, 0..;;;;;, y ~ b :that satisfies the conditions
u(x, 0)=0, u(x, b)=O, u(O, y)=Ay(b-y), u(a, y)=O.
"" h(2n+l)n(a~x) (2n+l)ny
BAb' s1n b s1n b
Ans. u(x, t)=---na ~ (2n+l)a . h(2n+l)na'
n= . mn b
a2 az .
14. Find a solution to lhe equation ax~+ ay~ =0 inside a ring bounded
by the circles x 2 +y 2 =R~. x 2 +y 2 =R= :that satisfies the conditions
au
ar
Ir=R 1
= + _Q_
1..2nR
1
u I
r=R
= u2.
Give a hydrodynamic inferpretation of the problem.

Hint. Solve the problem in polar coordinates.
Ans. u=u 2 -
2~~.n
~ ln..&.
"
15. The function u (x, y) = e-Y sin x is a solution of the equalion
~:~ + ~:~ =0 in the square Ote;;;;x~l, O~y~llhat satisfies :the conditions
u(O, y)=O, u(l, y)=e-Ysinl, u(x, O)~slnx, u(x, l)=e- 1 sinx.
In Problems 12-15 solve lhe Laplace equations for given boundary condi
lions by lhe finite-difference method for h=0.25. Compare ~he approximate
solution with the exact solution.
CHAPTER XIX
OPERATIONAL CALCULUS AND

CERTAIN OF ITS APPLICATIONS
Opera'tional calculus is an important branch of mathematical analysis.

The methods of operational calculus are used in physics, mechanics, electrical
engineering and elsewhere. Operational calculus finds especially broad appli-
cations in automation and telemechanics. In !his chapter we give (on the
basis of .the foregoing material of this text) .the fundamental concepts of
operational calculus and operational methods of solving ordinary differential
equations.
SEC. 1. THE INITIAL FUNCTION AND ITS Tf{ANSFORM
Let there be given the function of a real variable t defined for t ~0 [we
shall sometimes consider that the function f (t) is defined on an infinite
interval -oo < t < oo, but /(1)=0 when t <0). We shall assume that the.
function f (I) is piecewise continuous, that is, such that in any finite interval
it has a finite number of discontinuities of the first kind (~ee Sec. 9, Ch. II).
To ensure the existence of certain integrals in the infinite interval 0,.;;;;; t < oo
we impose an additional restriction on the function f (t): namely, we suppose
~hat there exist constant positive numbers M and s 0 such that
If (I) I< Mes,t (1)
for any value t in the interval 0,.;;;;; t < oo.
Let us consider the product of the function f (I) by the complex function
e-Pt of a real variable*) t, where p=a+ib is some complex number:
e-P 1f (t). (2)
Function (2) is also a complex function of a real variable t:
e-Ptf (t) =e-<a+ib) 1{ (t) =e-a 1f (t) e-ibt =e-atf (t) cos bt -ie-atf (t) sin bt.
Let us further consider the improper integral
"' 00 00
~ e-P 1f (t) dt =~e-at! (t) cos bt dt-i ~ e-atf (t) sin bt dt. (3)
0
We shall show that if !he function f (t) satisfies condition (1) and a> s0 ,
then the integrals on the right of (3) exist and the convergence of the inte-
grals is absolute. Let us begin by evaluating the first of these integrals:
00 00
I~ e-at[ (I) cos bt dt I,.;;;;;~ Ie-~ 1 / (I) cos I

bt dl <
0 0
*) See Sec. 4, Ch. VII, concerning complex functions of a real variable.

Transforms of the Functions cJ 0 (t), Sin t, Cost 855
In similar fashion we evaluate the second integral. Thus, the integraF

00
~ e-P1 f (t) dt exists. It defines a certain function of p, which we denote*)

0
by F (p):
00
F (p) = ~ e-P
1
f (t) dt. (4)
The function F (p) is called the Laplace transform, or the L-transform, or

simply the transform of the function f (t). The function f (I) is known as the
initial function, or the original. If F (p) is the transform of f (1), then we
write
F (p) ..;.. I (f), (5)
or
{(t)~F(p), (6)
or
L {f (t)} = F (p). (7)
As we shall presently see, the meaning of transforms consisfs in the fact
lha( with their help it is possible to simplify the solution of many problems,
for instance, to reduce the solution of differential equations to simple algeb-
raic operation~ in finding a transform. Knowing the transform one can find
the original either from specially prepared "original-transform" tables or by
methods that will be given below. Certain natural questions arise.
Let there be given a certain function F (p). Does there exist a function
f (t} for which F (p) is a transform? If there does, then is this function the
only one? The answer is yes to both questions, given certain definite assump-
tions with respect to F(p) and f (t). For example, tne following theorem,
which we give without proof, establishes that the transform is unique:
Uniqueness Theorem. If two continuous functions <p (t) and 'ljl (t) have one
and the same L-transform F (p), then these functions are identically equal.
This theorem will play an important role throughout the subsequent text.
Indeed, if in the solution of some practical problem we have determined, in
some way, the transform of a desired function, and from the transform the
original function, then on the basis of the foregoing theorem we conclude
that the function we have found is the solution of the given problem and
that no other solutions exist.
SEC. 2. TRANSFORMS OF THE FUNCTIONS a0 (t), SIN t, COSt

l. The function f (t), defined as
I (t) =I for t?: 0,
f(t)=O for t <0,
is called the Heaviside unit /unction and is denoted by <T0 (t). The graph of
this function is given in Fig. 379. Let us find the L-transform of the Heavi-
side function:
"'
L{a0 (t)}=r
j
e-P 1d t =e-Pt
--
p
I"', =1p- .
0
0
*) The function F (p), for p :1: 0, is the function of a complex variable

(for example, see V. I. Smirnov's "Course of Higher Mathematics", Vol. III,
Part 2) (Russian edition).
856 Operational Calculus and Certain of Its Applications
Thus, *)
~~ 1 ..:. .!_
p
(8)
---+---r Fig. 379.

or, more precisely,
O'o (t) + i
In some books on operational calculus lhe following expression is called
the transform of the function f (t):
00
F* (p) = p ~ e -P 1f (t) dt.

0
With this definition we have cr 0 (t) +1 and, consequently, C + C, more

exactly, Ca 0 (t) +C.
II. Let f (t) =sin t; then
. t}-1 -pi. tdt=e -pi ( -psmx-cosx

00 . ) I00
= -1-
L{ sm - Je sm pz + 1 0 +
pz 1
0
And so
. t .
sm -<;- p+1 1 . (9)
II I. Le~ f (t) =cos t; then

00 pi
L{ cos t} -_\J e -pi cos tdt=e- (tsint-pcost) loo _ _P_
pz + 1 - pz +
1
0
0
And so
. 'p
cos t -<;- p2+ 1. (10)
SEC. 3. THE TRANSFORM OF A FUNCTION WITH CHANGED

SCALE Of THE INDEPENDENT VAlUABLE. TRANSFORMS
OF THE FUNCTIONS SINat, COS at
Lef us consider the transform of the function f (at), where a> 0:
00
L {f (at)}=~ e-Pif (at) dt.

0
We change the variable in the latter integral, putting z =at; hence, dz =a dt;
then we get
p
5
00
2
L {f (at)}=! e --;;- / (z) dz
0
00
"') In computing lhe integral ~e-P 1dt one mighf represent if as lhe sum o
0
integrals of real functions; the same result would be obtained. This also holds
for the two subsequent integrals.
The Linearity Property of a Transform 857
or
L {f (at)} = ~ F ( %) .
Thus, if
F (p) -1-- I (t)
then
~ F(~) -1-- f (at). (11)
Example I. From (9), by (11), we straightway get

. t . 1 1
sma -<;-- ( ) 2
a ...!!...
a
+1
or
. t . a (12)
sma -<;- Pz+a2.
Example 2. From (10), by (II), we obtain
p
'""' + _!_a ( ...!!... +1

a
y
or
. p
cos at -<;- Pz+az (13)
SEC. 4. THE LINEARITY PROPERTY OF A TRANSFORM
Theorem. The transform of a sum of several functions multtplied by constants

is equal to the sum of the tr(lnsforms of these functions multiplied by the
corresponding constants, that is, if
n
f (t) = ~ C;f; (t) (14)
i=l
(C; are constants) and
F (p) -1-- f (t), F; (p) -1-- f; (t),
then
n
F (p) = ~ C;F; (p). (14')
i=l
Proof. Multiplying all lhe terms of (14) by e-PL and integrating with res-
pect to t from 0 to oo (taking the factors C1 outside the integral sign), we
get (14').
Example 1. Find the transform of lhe function
f (t) = 3 sin 4t-2 cos St.
Solution. Applying formulas (12), (13), and (15), we have
4 p 12 2p
L {f (t) } = 3 pz+ 16- 2 pz+25= p 2 + 16- p2 +25'
Example 2. Find .the original function whose transform is expressed by

the formula
5 20p
F(p)= pz+4-p2+9'
Solution. We represent F (p) as
5 2 p
F (p) =2 p2 + (2)2 +20 p2+ (3)2
Hence, by (12), (13), and (14') we have

5
f (I)= 2 sin 21 + 20 cos 31.
From the uniqueness fheorem, Sec. I, it follows that this Is lhe only original
function !hat corresponds .to the given F (p).
SEC, 5. THE SHIFT THEOREM
Theorem. If F(p) is the transform of the /unction f (t), then F (p+a) ts

the transform of the function e-t f (t), that is,
ifF (p) 7- f (t) }
then F (p +a) -;.. e-tf (t). (IS)
[If is assumed here that Re(p+a)>s0 .]
Proof. Find .the transform of the function e-at f (t)1
"' "'
L {e- 1! (t)} = ~ e-Pt-atf (I) dt = ~ e-<P+> 1f (t) dt.
0 0
Thus,
L 1
{e-" f (t)} = F (p +a).
This theorem makes it possible to expand considerably !he class of trans
forms for which it is easy to find the original functions.
SEC. 6. TRANSFORMS OF THE FUNCTIONS

e-t, SINH at, COSH at, e-t SIN at, e-at COS at
From (8), on the basis of (15), we straightway geti
_1_ -+e-t. (16)

p+a'
Similarly,
- -1- + eat. (16')
p-a
Subsfracting from lhe terms of (16') the corresponding terms of (16) and divld
ing the results by two, we get
..!. (-1___1_).:.. .!..(et-e-"')

2 p-a p+a 2
Transforms of the Functions e-l, Sin at, Cos at, e-nl Sin at, e-l Cos at 859
or
p-a h a t.
a 2 7 sm (17)
Similarly, by adding (16) and (16'), we obtain
p !:.._ a ...;.. cosh at. (18)
From (12), by (15), we have

a . at I t (19)
.(P+a)+a2 7 e- s na.
Using formula (15) we get from (13)

p +a . -~t t
(20)
(p+al"+a 2 7e cosa.
Example l.. Find lhe original function whose transform is given by the
formula
7
F (p)= p 2 !Op+20' +
Solution. Transform F (p) to the form of expression on the left-hand side
cf (19):
7 7
Thus
7 4
F(p)=4 (p+5)2+4'
Hence, by formula (19) we will have
F (p)...;.. { e- 51 sin 4t.
Example 2. Find lhe original function whose lransform .is given by the
formula
F (p)
Solution. Transform ~he function F (p):

P+l 2
(p + I) + 3 + (p + I) + 3
2 2 2 2
p+I 2 3
= (p + 1) 2
+32 +3 (p+ 1)'+320
using formulas (19) and (20) we find lhe original function:
F (p).;. e- 1 cos 3t + ~ e- 1 sin 3t,

860 Operational Calculus and Certain o[ Its Applications
SEC. 7. DIFFERENTIATION OF TRANSFORMS
Theorem. IfF (p) -;.. f (t), then

(-l)n dd;n F (p) -1- tnf (t). (21)
Proof. We first prove that if f (t) satisfies condition (1), then the integral
"'
~ e-Pt (-t)n f (t) dt (22)
exisfs.
By hypothesis lf(t)I<Mest-, p=a+ib, a>s 0 ; and a>O,' s0 >0. Obvi
ously, there will be an e > 0 such that the inequality a< s 0 +e will be ful-
filled. As in Sec. 1, it is proved that the following integral-exists:
"'
~ e-<P-)t If (t) I dt.
0
We then evaluate the integral (22)

"'
~I e-Pttnf (t)
0
Idt
"'
=~I e-<P-lte-1tn f (t) / dt.
0
Since the function e-ttn is bounded and, in absolute value, is less than
some number N for any value t > 0, we can write
"'
~I e-Pttnf
0
I
(t) dt
"'
< N ~I e-<P-ltf (t)
0
I
dt = N
"'
0
~ e-<P-)t If (t) Idt < oo.
If is .thus proved that the integral (22) exists. But this integral may be
regarded as an nth-order derivative with respect to the parameter*) p of ihe
integral
"'
~ e-P 1 f (t) dt.
0
And so, from formula

"'
F (p) = ~ e-P 1 f (t) dt
0
we ge~ lhe formula
S"' e-P1( -t)n f (t) dt = d~: S"' e-P1f (t) dt.

0 0
*) Earlier we found a formula for differentiating a definite integral with

respect to a real parameter (see Sec. 10, Ch. XI). Here, the parameter p is a\
complex number, but the differentiation formula holds true..
The Transforms of Derivatives 861
From lhese two equations we have
which is formula (21).

Lef us use (22) to find the transform of a power function. We wrife the
formula (8):
_!_ -+ 1. p
Using formula (21), from this formula we get
(-1) ddp ( ~ ) +t
or
l . t
p2 7 .
Similarly
For any n we have

n! . tn (23)
pn+17
Example I. From the formula [see (12)]
by differentiating the leff and righf sides with respect to the paramefer p,
we get
2pa . t . t
(pz + az)z 7 sm a . (24)
Example 2. From (13), on the basis of (21), we have

az-pz
(p2 + a 2) 2 7 t cos at. (25)
Example 3. From (16), by (12), we have
1 : t -t
(26)
(p+a)z 7 e .
SEC. B. THE TI{ANSFOI{MS OF DEI{IV ATIVES

Theorem. If F(p) +f(t), then
pF (p) -f (0) +f' (t). (27)
Proof. From .the definition of a transform we can write
"'
L {f' (t)} = ~ e-P 1 f' (t) dt. (28)
We shall assume that all the derivatives f' (t), f" (t), , t<n> (t) which we
encounter satisfy the condition (1), and, consequently, the integral (28) and
similar integrals for subsequent derivatives exist. Computing by parts the in
~egral on the right of (28), we find -
m m
L {f' (t)} = ~ e-P1f' (t) dt = e-pt f (t) + p ~
0
1: 0
e-Pt f (t) dt.
But by condition (1)

lim e-pt f (t) = 0
1-'> m
and
m
~ e-Pt f (t) dt = F (p).
Therefore
L {f' (t)} = - f (0) + pF (p).
The theorem is proved. Let us now consider the transforms of derivatives
of any order. Substituting into (27) the expression pF (p)-f (0) in place of
F (p) and the expression f' (t) in place of f (t), we geti
'P [pF (p)- f (0) ) - f' (0) -1- f" (t)
or, removing brackets,
p2F (p)-pf (0)-f' (0) -1- f' (t). (29)
The transform for a derivative of order n will be
pnF (p)- [pn-lf (0) + pn-2[' (0) + ... + pf<n-2) (0) +r-ll (0)1 -1- t<n, (t). (30)
Note. Formulas (27), (29), and (30) are simplified if f (0) =I' (0) = ..
= ... =t<n-1) (0)=0. In this case we gef
F (p) -1- f (t),
pF (p) -1- f' (I),
pn F (p) -1- f(n) (t).
SEC. 9. TABLE OF TRANSFORMS

For convenience, the lransforms which we obtained are here given in the
form of a table.
Note. Formulas 13 and 15 of .this table will be derived later on.
Note. If for the transform of the function f (t) we take
"'
F* (p) = p) e-Pt f (t) dt,
0
[hen in lhe formulas 1-13 of the fable the expressions In Ihe first column
must be multiplied by p, and formulas 14 and 15 will take on the following
form. Since F* (p) = pF (p), it follows !hat by substituting into the left side
Table of Transforms 863
Table I
a>
No. F (p) = j e-P t 1

(f) dl f (t)
0
1 -1p 1
2
a
sin at ;
p'+a'
p
3 cos at
P"+a'
4
1 e-t
p+a
5
a
sinh at
p-a
6
p
cosh at
p-a
7 a e-t sin at
(p+a)'+a"
8 p+a e-t cos at
(p+a)'+a 2
9 nl
p"+l t"
10 2pa
t sin at
(p'+ a')'
11
- a- P'.
(p' + a)
t cos at
12 1 te- 1
(p +a)'
1
13
(p" + a) 2~. (sin at- at cos at)
14
dn
(-1)" dp"F(p) t" f (t)
t
15 Fdp) F, (p) ~ ft ('t') 12 (1-'t') d't'
0
of 14 !he expression p (p) In place of F (p) and multiplying by p, we geti

p
14'. (-1)" p :;,. ( PP(p)) -t- t" f (t).

Substituting into the left side of 15
F ( )= F; (p)
1 p p
and multiplying this product by p, we have
t
15'. p1 p,(p)f2(p)7-
.. 5
/,('t)/2(!-'t)d-r.
0
SEC. 10. AN AUXILIAIU EQUATION FOR A GIVEN

DIFFERENTIAL EQUATION
Suppose we have a linear differential equation of order n with constant

coefficients a 0 , a,, ... , an- 1 , an:
dnx dn-x dx
ao dtn+a, dtn- 1 + ... +an-aT+anx(t)=f(t). (31)
It is required to find a solution of this equation x=x (t) for t;;;;. 0 thaf sati-
sfies the initial conditions
X (0)=X0 , x' (O)=x~ .. , x<n-II (O)=x~n-l). (32)
Before, we used lo solve this problem as follows: we found the general solu-
~ion of equation (31) containing A arbitrary constants; then we determined .the
constants so tha~ .they should satisfy the initial conditions (32).
Here we give a simpler method of solving this problem using operalional
calculus. We seek the L-transform of the solution x (t) of (31) satisfying the
x x
conditions (32). We designate this L-transform by (p); thus, (p) 7- x (t).
Let us suppose .that .there exist transforms of .the solution of. (31) and of
its derivatives of order n (after finding the solution we can test the .truth of
~his assumption). We multiply all terms of (3l)bye-Pt,where p=a+ib,and
integrate with respect to t from 0 to oo l
On the left-hand side of the equation are the L-transforms of the function x (t)
and its derivatives, on the right, the L-transform of the function f (t), which
we denote by F (p). Hence, equation (33) may be rewritten as
a0 L { ~:~} +a 1L { ~::~~} + ... +anL {x (I)}= L {f (t)}.

Substituting into this equation the expressions (27), (29), and (30) in place of
~he .transforms of the function and of its derivatives, we get
aa {pn:x (p)-[pn-xo+ pn-2x~ + pn-a x'~ + ... +x~n-1)1} +

+a, {pn-1 x (p)-[pn-2Xo + pn-x~ + .. +x~n-2) J} +
(34)
An Auxillary Equation 865
Equation (34) Is known as the auxiliary equation, or the transform equation.

The unknown in this equation is the transform x (p), which is determined
from it. Transform it leaving on the left the terms that contain x (p)l
X(p) [a 0pn+a 1pn- 1+ ...+an-tP+anl=
=ao [pn-1Xo +Pn-2 x:+ ... +x~n-t)J+
+ a1 [pn-2 Xo + pn-a J!o + ... + x~n-2)] +
+an- 2 [PX 0 +x~] +

+a,;- 1 [x 0] + F (p). (34')
x
The coefficient of (p) on the left of (34') Is an nth-degree polynomial in p,
which results when in place of the derivatives we put the corresponding
degrees of p into the left-hand member of equation (31). We denote the poly
nomial by QJn (p):
QJn (~) =aoPn+a1pn-t + +an+tP+an. (35)
The right-hand side of (34') is formed as follows:
the coefficient an-t is multiplied by x 0 ,
the coefficient an-z is multiplied by px 0 +x~,
..................
:the coefficient a 1 is multiplied by pn- 2x 0 +pn-x:+ ... +x~n- 1 >,
the coefficient a 0 is multiplied by pn- 1 x 0 +pn- 1 x~+- +x!n-tJ.
All these products are combined. To this Is also added the transform of the
right side of the differential equation F (p). All terms of the right side of (34'),
with the exception of F (p), form, after collecting like terms, a polynomia)
in p of degree n-1 with known coefficients. We denote it by 'i'n-t (p). And
so equation (34') can be written as follows:
x(p) QJn (p) ='IJln-1 (p) + F (p).
From this equation we determine x(p):
x(p) = 'i'n- (p) + F (p) (36)
'Pn (p) 'Pn (p)
x
Determined in this way, (p) is the transform of the solution x (t) of the
equation (31), which solution satisfies the initial conditions (32). If we now
fmd the function x (t) whose transform is the function (p) determined byx
equation (36), then by the uniqueness theorem formulated in Sec. 1 U will
follow that x (t) is the solution of equation (31) that satisfies the conditions
(32), that is,
x (t) = x (t).
If we seek the solution of (31) for zero initial conditions: X 0 =x;=x: ... =
x ~n-t) =0, then In (36) we will have '1Jln- 1 (p) = 0 and the equation will take
the form
"i"(p) = F (p)
'Pn (p)
28- 3388
or
- F(p)
X (p) = aoPn+alpn-1 + ... +an (36')
Example I. Find lhe solution of the equation

dx
dt+x=l
satisfying the conditions x=O for t=O.
Solution. Form the auxiliary equation
- 1 - 1
x(p) (p+I)-O=p or x(p)= (p+l)p
Decomposing the fraction on lhe right into partial fractions, we get
- I I
x(p)=p-p+1.
Using formulas I and 4 of Table I, we find the solution:
x(t)=l-e- 1
d 1x
d t+9x= 1
that satisfies the initial conditions: X0 =x;=o for t=O.
Solution. Write lhe auxiliary equation (34')
- 1 - 1
x(p) (p"+9)=-p or x(p)=p(p+g)
Decomposing this fraction Into partial fractions, we geti
1 1
- -g-P 9
X (p) = P"+9 +p.
Using formulas 1 and 3 of Table 1 we find the solution:
1 1
X (t) = - g COS 3t g +
Example 3. Find lhe solution of the equation
d'x dx
dt +3 dt +2x=t
that salisfies the Initial conditions x0 =x;=o for t=O.

Solution. Write !he auxiliary equation (34')
- I
x(p) (p1 +3p+2)= 1
p
or
- I I
" (p) = pz --c(--cp.~+'3'p-:-+--n2')
Decomposition Theorem' 867
Decomposing this fraction into partial fractions by the method of undetermined

coefficients, we obtain
- II 31 I I
x(p)=2 p!--;rp+p+I- 4{p+2)
From formulas 9, \,and 4 of Table I we find the solution:

1 3 t I
x (t)= t- 2 -:r+e- -
2t
4 e- .
d 2x dx
dt 2 +2 ;n+5x=sint
satisfying the conditions x0 =I, x~ = 2 for t = 0.

Solution. Write the auxiliary equation (34'):
x(p} (p 2
+ 2p + 5} = p I + 2 + 2 I + L { sin t }
or
whence we find x(p):

- p+4 1
X {p)= p 2 +2p +5+ (p 2 + I) (p 2 +2p+5)
Decomposing the latter fraction on the right into partial fractions, we can
write
or
Applying formulas 8, 7, 3, and 2 of Table I, we get the solution
x (t}= II e- 1 cos 2t+ 29 e- t sm

. I 1
2t- cos t+ slnt
10 20 10 5
or, finally,
x (t} =e- 1 ( M 2t +~sin 2t) -~cost+-}
cos sin t.
SEC. 11. DECOMPOSITION THEOREM
From formula (36) of the previous section if follows thaf the transform of
~he solution of a linear differential equation consists of two terms: the first
~erm is a proper rational fraction in p, ihe second term is a fraction whose
numerator is ihe transform of the right side of the equation F (p), while the
denominator Is lhe polynomial CJln (p}. ! If F (p) is a rational fraction, then
the second term will a !so be a rational fraction. It is thus necessary to be
able io find the original function whose transform is the proper rational
28*
858 Operational Calculus and Certain of. Its Applications
fraction. We shall deal with this question in the present section. Let the
L-transform of some function be a proper rational fraction in p:
'l'n-t (p).
cpn lP)
It is required to find the original function. In Sec. 7, Ch. X, it was shown
that any proper rational fraction may be represented in the form of a sum
of elementary fractions of four types:
A
I. p-a '
A
II. (p-a)k'
Ill. A p+B , where the roofs in the denominator are complex, that
p 2 + atp+a2
a~
is, -a2 <0.
4
IV. Ap+B where k~2. the rootsofthe.denominator are complex.
(p2+a,p+as)k
Let us find the original functions for these elementary fractions. For
fraction type 1 we get (on the basis of formula 4 of Table 1)
_A_.:...Aeat.
p-a .
For a type II fraction, by formulas 9 and 4 of Table 1, we have
A . A 1 tk-t ~~
(p-a)" :+ (k-1) l e (37)
Let us now consider the type Ill fraction. We perform identical transforma-
tions:
Denoting the first and second terms by M and N respectively, we get

(from formulas 8 and 7 of Table 1)
Examples of s'ozu'tio~s of Differential Equations 869
(38)
We shall nof consider the case of elementary fraction IV, since H would in
volve considerable calculations. We shall consider certain special cases below.
SEC, 12. EXAMPLES OF SOLUTIONS OF DIFFERENTIAL

EQUATIONS AND SYSTEMS OF DIFFERENTIAL EQUATIONS
BY THE OPERATIONAL METHOD

d2 x
d t +4x=sln 3x
lhat satisfies the initial conditionsx0 =0, x~=O when 1=0.
Solution. Form the auxiliary equation (34')
- 3 - 3
x(p)(p2 +4)=p0+0+..2+ , x(p)
" 9 (p2 9)( o 4) + +
3 3
- -5 5 1 3 3 2
x(p)= p+ g+ p+4 = -5 p1 +9+ 10 p'+ 4'
whence we ge~ .the solution
x (t) =w3 sin 2t - I
5 sin3t.

d1x
dt'+x=O
that .satisfies lhe iniflal condlflons x0 =l,x~=3, x:=B when 1=0.
Solution. Form the auxiliary equation (34')
x(p) (p1 + 1) =p 1 +p3+B,
1
we find
Decomposing lhe rational fraction obtained into partial fractions, we get

p2 +3p+S 2 -p+6
(p+ l}(p2-p+l) = P+ I + P2 -P+ 1 =
I Y3
1 p-2 11 -2-
=2P+T-(p- ~Y+( v!Y +v3 (p-~Y+(~Y

Using Table 1, we write lhe solution:
(t)=2e- 1+e
..!_
2 I (
-cos - - t
y3 + Y3sin
11 Jl3 )
- - t
2 2
Example 3. Find ihe solution of the equation
d"x
dt +X= f COS 2/
that satisfies lhe initial conditions x=O, x:=o

when 1=0.
Solution. Write ~he auxiliary equation (34')
- I 8
X (p) (p"+ 1) = p+ 4- (p"+4)2'
whence
- 51 51 8 I
X (p) = - g p"+
1 +g +4+3 (P 2+4)l" p
Consequently,
5 sin 2/ + 1(1.
x (t) = - g5 sin t + Tii sm 2/ - t cos 2/ ) .
3 2
Obviously, lhe operational method may also be used to solve systems of
ainear differential equations. The following is an illustrati<>n.
Example 4. Find the solutions of the set of equations
dx dy
3
dt 2x dt = 1 + +
dx dy
at+4at+3y=0
lhat salisfy [he initial conditions x=O, y=O when t=O.
Solution. We denote x (t) ......;.:x (p), y (t) ......;- y (p) and write lhe system of
auxiliary equations:
(3p + 2) -X + p -y (p) = -p1
(p) 0
ix+(4p+3> -y =o.
Solving !his syslem, we find
- 4p+3 1 1 ]
X (p)= p(p+l)(llp+6)=2p-5(p+l)-JO (llp+6)'
- I 1 ( I 11 )
y (p)=- (llp+6)(p+J)=5 p+l -llp+6 .
The Convolution Theorem 871
From the transforms we find the original functions -the sought-for solutions
of the system:
I 1 3 -!....t
X (t) = 2 -Se-t -lO e 11
1 (
y(1)=5 e- 1-e
_!..t)
11

\
Linear systems of higher orders are solved in similar fashion.
SEC 13. THE CONVOLUTION THEOREM

The following convolution theorem is frequently useful when solving
differential equations by the operational method.
Convolution Theorem. If F 1 (p) and F 2 (p} are the transforms of the functions
f 1 (t} and f 2 (t), that is,
F1 (p) -':-+/ 1 (t) and F 2 (p) -':-+fa {t),
then F 1 (p) F 2 (p) is the transform of the function
t
~ f1 (-r:}fa(t --r:) d1:,
0
that is,
t
F1 (p) F 2 (p) -;..... ~ f 1 ( -r:) f 2 (t - -r:) d1:. (39)
Proof. We find the transform of the function

t
~ f1 (-r:) fs (t--r:} d1i
0
from the definition of a transform:

t ~ t
L { ~ f1 (-r:)/2(1--r:) d-r:
0
} =
0
~ e-pt [ ~ f1('t)f2(t-'t) d't
0
J
dt.
The Integral on the right is a double Integral of the form ~ ~ Ill (1:, t) dt d't,
D
which is taken over a region bounded by the straight lines 't=O, -r:=t
(Fig. 380). Changing the order of integration in this double integral, we geU
t ~ ~
b { ~ f1 ('t) fa (t -'t) d't} = ~ [ f (-r:) ~ e-P 1f (t -'t) dt d1:.

0 0
1
,;
2 J
Changing lhe variable t-'t=Z in the inner Integral, we obtain
~ ~ ~
~ e-Pt f 1 (t -1:) dt = ~ e-P <z+l ! 2 (z) dz =e-P' ~ e-Pz f 2 (z) dz =e-P' Fa (p).
1: 0 0
Hence,
t 110 110
L { ~ f 1 ('t') f 2 (l-'t') d't} = ~ / 1 ('t') e-pt F 2 (p) d't=F 1 (p) ~ e-pt / 1 ('t') d't'=
0 0 0
= f2 (p)F 1 (p).
And so
t
~ f, ('t) {2 (t-'t') d't' +- F 1 (p) F 2 (p).
0
This is formula 15 in Table I.

t
Note 1. I he expression S{1 ('t') {2 (t -'t') d't is called the conuolution
'( 0
(Faltung, resultant) of two functions {1 (t) and {2 (t).
The operation of obtaining it is also known as the
conuolution of two functions; here
t t
~ f, ('t) fz (t-'t') d't'= ~ ft (t-'t) fz ('t') d't'.
t t 0 0
Fig. 380 That this equation is true is evident if we change

the variable t-'t=Z in the right-hand integral.
Example. Find lhe solution to the equation
d2x
dt +x=f (t)
that satisfies the initial conditions x 0 =x: =0 for t =0.

Solution. Write the auxiliary equation (34')
X (p) (P'+ l)=F (p),
- I
where F (p) is the transform of the function f (t). Hence, x (p) = pz +1 F (p) ,
but p'~ 1 -;...sin t and F (p)-;... f (t). Applying the convolution formula (39)
I
and denoting P'+l=f 2 (p), F(p)=F 1 (p),
we get
t
x (t) = ~ f ('t') sin (t-'t') d-e. (40)
Note 2. On the basis of the convolution theorem if is easy to find the

lransform of the integral of the given function if we know the transform of
this function; namely, if F (p)-;... f (t), !hen
t
t F (p) :+- sf ('t) d't. (41)
The Differentia'! Equations cf Mechallical Oscillations 873
Indeed, if we denote
I
/1(1)=/(t), / 2 (1)=1, then F 1 (p)=F(p), F 2 (p)=-.
p
Putting these functions into (39j, we get formula (41).
SEC. 14. THE DIFFERENTIAL EQUATIONS OF MECHANICAL

OSCILLATIONS. THE DIFFERENTIAL EQUATIONS
OF ELECTRIC-CIRCUIT THEORY
From mechanics we know that the oscillations of a material point of mass

m are described by the equation*)
d2 x A. dx k 1
dt+-;n dt+-;nx=-;nf.(t), (42)
where x is the deflection of the point from a certain position and k is the
rigidity of the elastic system, for instance, a spring (a car spring}, the force
of resistance to motion is proportional (the proportionality constant is A.)
to the ftrst power of the velocity, and {1 (t) is the C L
outer (or disturbing) force.
Equations of type (42) describe small vibrations
of other mechanical systems with one degree of free-
dom, for example, the torsional oscillations of a fly-
wheel on an elastic shaft, if x is the angle of rotation E. R
of the flywheel, m is the moment of inertia of the
flywheel, k is the torsional rigidity of the shaft, and
mf1 lt) is the moment of the outer forces relative to
the axis of rotation. Equations of type (42) describe
not only mechanical vibrations but also phenomena
1-!!J
Flg. 381.
that occur in electric circuits.
Suppose we have an electric circuit consisting of an inductance L, a
resistance R and a capacitance C, to which is applied an e.m.f. E (Fig. 381).
We denote by i the current in the circuit, by Q the charge of the capacitor;
then, as we know from electrical engineering, i and Q satisfy the following
equations:
di . Q
+
L dt+Rt c=E, (43)
dQ .
(44)
cu='
From (44) we get
(44')
Substituting (44) and (44') info (43), we gef for Q an equation of type (42):
d 2Q dQ I
L([i%+Rdt +cQ=E. (45)
*) See, for example, Ch. XIII, Sec. 26, where such an equation is derived
in considering the oscillation of a weight on a car spring.
Differentiating both sides of (43) and utilising (44), we obtain an equation

for determining the current i:
d"i di 1 . dE
Ld 1.+R&+c'=& (46)
Equations (45) and (46) are .type (42) equations.
SEC. 15. SOLUTION OF THE DIFFERENTIAL OSCILLATION

EQUATION
Let us write the oscillation equation in the form
d2 x dx
dt +al dt +a.x=f (t), (47)
where the mechanical and physical meaning of the desired function x, of the
coefficients a 1 , a 2 , and of the function f (t) is readily established by comparing
this equation with equations (42), (45), (46). Let us find the solution to
equation (47) .that satisfies the initial conditions x=x 0 , x' =x: when t =0.
We form .the auxiliary equation for equation (47):
x (p) (p"+a p+a )=xoP+x:+a x +F (p),
1 2 1 0 (48)
where F (p) is the :transform of the function f (t). From (48) we find
_ x0 p+x: +a1x0 F (p)
x (p) = p+a~p+a. + P"+a~p+a. . (49)
Thus, for a solution Q (I) of equation (45) that satisfies the initial conditions
Q=Q 0 , Q'=Q~ when 1=0, the .transform will have the form
_Q ( ) _ L (Q 0 p+Q:) +RQo E (p)

p-
2
I + 1 '
Lp +RP+ C Lp'+Rp +c
The 1ype of solution is significantly dependent on whether the roots of the
trinomial p"+a 1 p+a are complex, or real and distinct, or real and equal.
.that is, when ( r- a

Let us examine in defail the case when ihe roots of .the trinomial are complex,
a2 2 < 0. The other cases are considered in similar fashion.
Since the transform of a sum of two functions is equal to the sum of their
transforms, it follows from formula (38) that the original function for the
first fraction on the right of (49) will have the form
(50)
lnvestigcl.iing Free Oscillations sis
Let us then find the original function corresponding to the fraction

F (p)
p +a 1p +a1
Here. we take advantage of the convolution theorem, firs noting that
Hence, from (39) we get
p
+'(p)
a 1p+a 2
~ v 1
a1 -To
a
I
~~ f ('t')e -~<t-Tl sln(l-'t'),V/ a 1
- a: d't'.
4
(51)
And so, from (49), faking into account (50) and (51), we geC
x(l) ~'
-~-lt
'
[
'"'
, f
t V o, -T+
a
v Xo + t
o,- ~ in
, ~--a-
tV n,--f
l
+
r ;--a-.
v -+
-~{t-'t')
+ V.- ~ /(''. .1
ln(l-)
..
(5~
If the external force f (I)= 0, which means that if we have free mechanical
or electrical oscillations, then the solution is given by the first term on the
righthand side of expression (52). If the initial data are equal fo zero, i.e.,
if X 0 =x: =0, then the solution is given by the second term on the right side
of (52). Let us consider these cases in more detail.
SEC. 16. INVESTIGATING FREE OSCILLATIONS
Let equation (47) describe free oscillations, that is, f (I)""' 0. For con-
venience in writing we introduce the notation a 1 = 2n, a 2 = k 1 , k: = k" - n.
Then (47) will have the form
d 2x dx
dt +2n df+k 2X=0. (53)
The solution of {his equation xfr that satisfies the lniflal conditions x =x0 ,
x' =X~ for t=O is given by fhe formula (50) or by the first Ierm of (52):
(54)
x~ +x0 n
We denote x 0 =a, k b. It is obvious that for any a and b we can
1
~elect M and 6 such that the following equalities will be fulfilledl
a=Msin6, b=Mcos6,
here,
a
M2 =a2 +b 2 , tan 6=/}.
We rewrite formula (54) as
x1r=e-nt [M cos k,t sin 6 + M sin k 1 t cos 6],
or, in final form, .the solution may be written thus:
Xtr= Ya +te-n1 sin(k 1t+6). (55)
Solution (55) corresponds to damped oscillations.
If 2n = a 1 = 0, that is, if there is no internal friction, lhen the solution
will be of 1he form
Xtr= Ya+b 2 sin(k 1 t+6).
In this case harmonic oscillations occur. (In Ch. XIII, Sec. 27, Figs. 270 and
271 give graphs of harmonic and damped oscillations.) -
SEC. 17. INVESTIGATING MECHANICAL AND ELECTR.ICAL

OSCILLATIONS IN THE CASE OF A PER.IODIC EXTER.NAL FOR.CE
When studying elastic oscillations of mechanical systems and, in particular,

when studying electrical oscillations, one has to consider different types of
external force f (t). Let us consider in detail the case of a periodic external
force. Let equation (47) have the form
d2x dx
dt 1 + 2n dt + k 2x = A sin wt. (56)
To determine the nature of the motion it is sufficient to consider lhe case
when x0 =X~ =0. One could obtain the solution of the equation by formula
(52), but pedagogically speaking, it is more convenient to obtain the solution
by carrying out all the intermediate calculations.
Let us write the transform equation
- (I)
X (p) (p'+2np+k 2 )=A~+
p (I)
2,
from which we get

- Aw
x (p) = (p 2 + 2np + k1 ) (p2 +w 2 )
(57)
We consider the case when 2n t= 0 (n1 < k'). Decompose the fraction on
the right into partial fractions:
Aw Np+B + Cp+D (SS)
(p 2 + 2np+ k2 ) (p 2 +w 2 ) p 2 + 2np + k 2 p+w
We determine the constants B, C, D by the method of undetermined
coefficients. Using formula (38), we find the original function from its
Investigating Mechanical and Electrical Oscillations 877
L-transform (57):
x(t)= (kz-wz~+ 4 nzwz 2

{ (k 2 -w )sinwt-2nwcoswt+
+e-nt [ (2n 2 -k 2 +w 2) :
1
sin k1 1 +2nw cos k 11 J}; (59)
here again, k 1 = Jfk 2 -n2 This is the solution of equation (56) that satisfies
the initial conditions x 0 =x:=o when 1=0.
Let us consider a special case when 2n=0. This corresponds to a mecha-
nical system with no internal resistance, or to an electric circuit where R = 0
(no internal resistance in the circuit). Equation (56) then lakes the form
d 2x
iflZ+k 2 x =A sin wl, (60)
and we get. the solution of this equation satisfying the conditions x 0 = x~ = 0

for 1=0 if in (59) we put n=O:
x (I)= (kZ !wz) k [ -w sin kl + k sin wlj. (61)
Here we have the sum of two harmonic oscilla- t

tions: natural oscillations with frequency k:
A w
Xnatt = - kz-w k sinkt,
and forced oscillations with frequency w:
t
A
Xt0r(t)= k-w 2 slnwt.
The type of oscillations for the case k ~ w is

shown in Fig. 382.
Let us again return to formula (59).
If 2n > 0 (which occurs in the mechanical and
0 t
electrical forces under consideration), then the
term containing the factor e-nt, which represents Fig. 382.
damped natural oscillations for increasing I,
rapidly decreases. For t sufficiently large, the character of the oscillations
will be determined by the term that does not contain the factor e-nt; that is,
by the term
x (t) = (k-wz~+ 4 nzw {(k 2

-W
2
)sinwl-2nwcoswt}. (62)
We introduce the notations
(63)
where
M- A
- Y (k 2 -w2 ) 2 + 4n 2w2
The solution (62) may be rewritten as follows:

A
x ~t) =
k
2
-./
V ( I - ~: + 4n ~: r 2
sin (wt + ll). (64)
From formula (64) if follows that the frequency of forced oscillations coi nciC.:es
with that of the external force. If the internal resistance, characterised by
~he number n, is small and the frequency of the external force w is not very
different from that of the natural oscillations k, then the amplitude of oscil-
lations may be made as great as one pleases, since the denominator may be
arbitrarily small. For n=O, ro 2 =k2 , the solution is not expressed by for-
mula (64).
SEC. tS._SOLVING THE OSCILLATION EQUATION

IN THE CASE OF RESONANCE
Le~ us consider the special case when a 1 =2n=0, that is, when there is
no resistance and the frequency of the external force coincides with that of
the natural oscillations k=w. The equation then takes the form
(65)
We shall seek the solution that satisfies the initial conditions x0 = 0, x~ =0

for t = 0. The auxiliary equation will be
x <P'+k">=A p~k,

whence
(66)
We have a proper rational fraction of type IV, which we have no( considered
in the general form. To find the original function for the transform of (66),
we take advantage of the following procedure. We write the identity (formula 2
of Table 1)
"'
p ~ k =Se-Pt sin kt dt. (67)
0
We differentiate both sides of this equation with respecf fo k (the integral on

the right may be represented in the form of a sum of two integr~ls of a real
variable, each of which depends on the parameter k):
"'
Se-Pt t cos kt dt.
0
Utilising (67) we can rewrife this equation as
-(p'!:) =Se-Pt
"'
[t
0
cos kt- ~sin kt] dt.
The Delay Theorem 879
Whence it follows directly that
Ak . A ( 1 )
(p 2 + k 2) 2 :---+- 2k k sin kt- t cos kt
(from 'this formula we have formula 13, Table 1). Thus, the solution of equa-
lion (65) satisfying the initial conditions x0 =X~=0 fort=~ will be
x (t) = :k ( ~ sin kt- t cos kt) . (6B)
Let us study the second term of this equation:

A
X2 (t) = - k
2
t cos kt; (68')
~his quantity is not bounded as t increases. The amplitude of oscillations thati

correspond to formula (6B') increases without bound as t increases withouti
bound. Hence, the amplitude of oscillations corresponding to formula (6B) also
increases without bound. This is resonance; it occurs when the frequency of
lhe natural oscillations coincides with ihat of the external force (see also
Ch. X Ill, Sec. 29, Fig. 273).
SEC. 19. THE DELAY THEOREM
Lef lhe function f (t), for t < 0, be identically equal Io zero (Fig. 3B3, a).
Then the function f (t - t0 ) will be identically zero for t < t 0 (Fig. 383, b).
We shall prove a theorem which is known as the delay theorem.
0 t 0
r
I
I
to t
(q) (b)
Fig. 383.
Theorem. If F (p) is the transform of the function f (t), then e-Pto F (p) is
the transform of the function f (t-t 0 ); that is, iff (t) F (p), then +-
f (t - t 0 ) +- e-P F (p).
1
Proof. By the definition of a transform we have

"' to "'
L{f(t-t 0 )}=~ e-Ptf(t-t 0 )dt=~ e-P 1 f(t-t 0 )dt+~ e-P1f(t-t 0 )dt,
0 0 lo
880 Operational Calculu!l and Certain of .Ns. Applications
The first integral on the right of the equation is zero since f(t-t 0 )=0 for
t< 10 In the last integral we change the variable, putting 1-t0 =zl
~ ~
L {I (t -t.>} = ~ e-P (z+lo) r (z) dz = e-Pio ~ e-PZ f (z) dz = e-Pio F (p).

0 0
Thus, f (t-1 0) ~ e-Ptop (p).
Example. In Sec. 2 it was established for the

Heaviside unit function that
(J0 (f-h)
a0 (I)~ ~
It follows, from the theorem that has just been

0 h t proved, that for the function o0 (t-h) depicted
in Fig. 384, the L-transform is
Fig. 348, ..!..e-Ph
p I
lhaf is,
Exercises on Chapter XIX

Find solutions to the following equations for the indicated initial condi
lions
1. :;: +3 :: +2x=0, x= I, x' =2 for 1=0. Ans. x=4e- 1 -3e-t.
2 d'x d'x 0 2 ' 0 " I x=l-t+e 1

dt 1 - dt 2 = ' X= ' X = ' X= for 1=0. Ans.
1
d x 2 dx ( +b)
3'dt _
2- adt+ a X -0' for 1=0. Ans. x=
I
4 dJC dx
dl 2 - 3 dl + 2x-- e'"' ' x=l 1 x'=2 for 1=0. Ans. X=-esr
12
+
x , =X0, f or 1=0. Ans. x = - a- x

2 m -n1
X
X (cos nl -cos ml) +x0 cos nt +....!cos mt,
m
d2x dJC I
6. dt- dl=l 2 1 x=O~ x'=O for 1=0, Ans. x=3e 1 - t 1 -t 2 -2t-3,
3
Exe'rtises -1}n Chapter XfX- . . '\'
7. d'x
dt' +x= I te 1, x=x'=x"=O for 1=0. Ans. x= I ( 12 -31+
2 4
3)
+2 e1 - 1e- -3I {cos (I2 ,r 31
24
t 1 -)
- ,r-;;
r 3sln (I2 ,r 31 1 -)}
e +t
Ans.
d4x d2 x '
9. dl' -2 dt +x=sln I, X0 =X0 = X0 =x0 =0 for 1=0. Ans. x=
=sI (3-1 2
)
3
sin t-8 t cost.
10. Find solutions to the system of differential equations

d'x d2 y
dt 2 +y= I, dt 1 +x=O,
that satisfy lhe initial conditions X0 =y0 =X~=Y~=0 for 1=0. Ans. x(l) =
t
= - I cost+ I et I e-,
2 4 +4 y(t)=-cost+e t +e- t -1.
INDEX
A Auxiliary equation, 535, 865

Average acceleration, 125
Abel's theorem, 742, 743 Average curvature, 211, 327
Absolute constants, 16
Axis
Absolute value, 15
imaginary, 233
Absolutely convergent integral, 420. of imaginaries, 233
Absolutely convergent series, 731
of reals, 233
Acceleration
polar, 28
average, 125
real, 233
at a given instant, 125
of linear motion, 125
B
Adam's formula, 587
Algebra Bern()ulli's equation, 480-492
fundamental theorem of, 245 Bernstein, S. N., 252
Algebraic equation, 225, 245 Bernstein's polynomial, 252
Algebraic functions, 26, 28 Bessel function of the first kind, 765
Alternating series, 727 Bessel function of the second kind, 767
Amplitude (of a complex number), 234 Bessel's equation, 763, 764
Amplitude (of oscillation), 558 Bessel's inequality, 796
Analysis Binomial differential, 375
harmonic, 805 Binomial series, 754-756
Analytical expression, 21 Binormal, 331
Angle of contingence (of an arc), 210 Boundary conditions, 818, 825, 828
Antiderivative, 342 Boundary of a domain, 256
Arc length of a curve, 447-452 Boundary-value conditions, 818
Archimedes Boundary-value problem
spiral of, 29 first, 825, 837
Argument, 19 second, 837
of a complex number, 234 Bounded function, 40, 41
intermediate, 85 Bounded variable, 18
Astroid, 107 Briggs, 56
Asymptote, 189 Broken line
inclined, 191 Euler's, 583
vertical, 190 Bunyakovsky, 647
Subject Index 883
Bunyakovsky's inequality, 647 powers of, 237

Biirgi, 56 roots of, 238
subtraction of, 235
c trigonometric form of, 234
Complex plane, 240
Calculus Complex roots, 248, 249
operational, 854 Complex variable, 240
Catenary, 471 Composite exponential function, 93
Cauchy's test, 721, 722 Composite function, 25
Cauchy's theorem, 143 Concave curve, 183
Centre of curvature, 217 Cancavity (of a curve), 183
Centre of a neighbourhood, 18 Conditions
Change of variable, 348 boundary, 818, 825, 828
Characteristic equation, 570 boundary-vat ue, 818
Chebyshev, P. L., 253 initial, 474, 514, 818, 825, 828
Chebyshev polynomials, 253 Conditional extremum, 300
Chebyshev's formula, 430435 Conditionally convergent series, 731
Circle of curvature, 217 Conjugate complex numbers, 233
Circulation (of a vector), 673 Conjugate pairs (of complex roots), 249
Clairaut's equation, 505-507 Constant, 16
Closed contour, 672 absolute, 16
Closed domain, 256 Continuous function, 57, 58
Closed interval, 17 Contour
Closed region, 608 closed, 672
Coefficient, 27 Convergence of a series
Coefficients necessary condition for, 713
Fourier, 779 Convergent integral, 421
of a trigonometric series, 776 Convergent series
Combined method, 229 absolutely, 731
Common logarithms, 758 conditionally, 731
Complete integral (of a differential Convex curve, 183
equation), 475, 515 Convex down (downwards), 183
Complex function of a real variable, Convex up (upwards), 183
242 Convexity (of a curve), I 83
Complex number Convolution, 872
imaginary part of, 233 Convolution formula, 872
real part of, 233 Convolution theorem, 871
Complex numbers, 233 Coordinate
addition of, 234 polar, 28
conjugate, 233 Coordinate system
division of, 236 polar, 28
exponential form of, 243 Correspondance
geometric representation . of, 233 one-to-one, 634
multiplication of, 235 Critical points (values), 168, 294
884 Subject Index
Curl (of a vector function), 694 of a function defined implicitly,

Curve 276, 277
concave, 183 logarithmic, 94
convex, 183 of a logarithmic function, 84
convex downwards (upwards) 183 Derivative
Gaussian, 187 of nth order, 119
smooth, 528 partial, 263-265
space, 450 of a product, 82
Curves second, 119
integral, 475, 515 of second order, 119
resonance, 561 of a sum, 81
Curvature, 211, 212, 215, 216, 327 symbols of, 71
average, 211, 327 third, 119
centre of, 217 total, 275
circle of, 217 Determinant
at a point, 211, 212 functional, 636
radius of, 217 Deviation
Curvilinear trapezoid, 400 maximum, 793
Cusp root-mean-square, 793
double, 309 Diameter of a subregion, 650
Cusp of the first kind, 308 Differentiable function, 74
Cusp of the second kind, 309 Differentiable at a point, 267
Cycloid, 106, 107 Differential, 113, 114, 116, 117, 118,
267
D of an arc, 210
D 'Alembert's test, 718 binomial, 375
Decomposition (of a rational fraction nth, 121
into partial fractions), 361 second (second-order), 121
Decomposition theorem, 867 third (third-order), 121
Decreasing function, 20 total, 267
Decreasing variable, 18 Differential equation, 469, 472
Definite integral, 396, 398, 399 exact, 492
Degree of a polynomial, 27, 244, 246 first-order, 4 73-478
Del operator, 700 higher-order, 514-516
Delay theorem, 879, 880 linear, 528, 529
De Moivre's formula, 237 ordinary, 472
Density Differentials
linear, 459 error approximation by, 270
surface, 642 Differentiation, 71
Derivative, 71, 72, 78, 79, 80, 114 Direction of circulation, 672
of a composite function, 85, 86 Direction-field, 476, 477
directional, 284-286 Directional derivative, 284-286
discontinuous, 168 Dirichlet-Neumann problem, 840, 843
of a fraction, 83 Dirichlet problem, 837
Subject Index 885
Dirichlet's integral, 800 Fourier (for heat conduction), 815

Discontinuity (see point of d.) heat-conduction, 815, 816, 825, 828
Discontinuous derivative, 168 Equation (cont.)
Discontinuous function, 60 of heat propagation
Divergence (of a vector, or of a vector in a plane, 828
function), 699 higher-order differential, 514-516
Divergent integral, 421 homogeneous, 482
Domain homogeneous linear, 529
closed, 256 hyperbolic, 815
of convergence (of a series) 733 Lagrange's, 507-509
of definition (of a function), 19, 256 Laplace's 703, 815, 836
natural, 21, 22 in cylindrical coordinates, 842
open, 256 linear, 487
Dominated series, 734-736 linear differential, 528, 529
Double cusp, 309 Lyapunov's, 796
Double integral, 609 nonhomogeneous linear, 529
Double root, 546 of a normal, 126
ordinary differential, 472
E parabolic, 815
partial differential, 472
Eigenfunctions, 821 parabolic, 815
partial differential, 472
Eigenvalues, 821 of a tangent, 126
Element of integration, 343 transform, 865
Elementary function, 26 vector, 314
Ellipse of inertia, 645-648 wave, 815, 817
with a right-hand member, 529
Elliptic equations, 815
with separated variables, 479
Elliptic integral, 385 with variables separable, 479
End points of an interval, 17
without a right-hand member, 529
Envelope (of a family of lines), 498,
501 Equations
Equation parametric, 103, 104, 314
algebraic, 225, 245 telegraph, 819
auxiliary, 535, 865 Equipotential lines, 510
Bernoulli's 490-492 Equivalent infinitesimals, 64, 65
Bessel's, 763, 764 Error
characteristic, 570 maximum absolute, 270
Clairaut's 505-507 maximum relative, 272
of continuity, 837 relative, 272
of continuous flow of a compressible Euler substitution
liquid, 839 _ first, 372
differential, 469, 472 second, 373
elliptic, 815 third, 374, 375
exact differential, 492 Euler's broken line, 583
first-order linear, 487 Euler's formula, 243, 753
886 Stlb feet Index
Euler's method (of approximate Formula (cont.)

solution of first-order differential Green's, 679-681
equations), 581-584 of integration by parts, 354
Evolute, 219 Lagrange's interpolation, 250, 251
Evolvent, 219 Leibniz', 120, 436
Exact differential equation, 492 Maclaurin's, 155
Existence theorem of a line integral, Newton-Leibniz, 410, 411
673 Ostrogradsky's, 697-700
Expansion (of a function), 156-159 In parabolic, 426
a Taylor's series, 156 rectangular, 424, 425
Explicit function, 90, 91 Simpson's, 428
Exponential form (of a complex num- Stokes', 692-697
ber), 243 Taylor's, 152, 155
Exponential function, 22, 24, 93, 102 for transformations of coordinates
properties of, 241 in a double integral, 636
Exponential-power function, 93 trapezoidal, 426
Expression Wallis', 415, 416
analytical, 21 Formulas
Extreme values (of a function), 166 Serret-Frenet, 335
Extremum (extrema) (of a function), Fou.rier coefficients, 779
166, 292 Fourier cosine transform, 810
conditional, 300 Fourier equation for heat cond.uction,
815
F Fourier integral, 806-808
in complex form, 810-812
Factor Fourier inverse transform, 812
integrating, 495-497 Fourier series, 776-812
Faltung, 872 definition of, 779
Family of curves, 475
one-parameter, 498 Fourier sine transform, 810
Family of functions, 343 Fourier transform, 812
Family of orthogonal trajectories, 510 Fraction
Field improper, 357
scalar, 283 partial, 358
vector, of gradients, 287
First Euler substitution, 372 proper, 357
Flow lines, 510 Fractional rational function, 27
Flux (of a vector field through a sur- Free oscillations, 557, 875-876
face), 688 Frenet (see Serret-Frenet formulas,
Forced oscillations, 557, 559-563 335)
Formula Frequency, 558
Adams', 587 Function, 19
Chebyshev's, 430-435 algebraic, 26, 28
convolution, 872 analytical representation of, 21
De Moivre's, 237 basic elementary, 22
Euler's, 243, 753 Bessel, of the first kind, 765
Subject Index 887
Bessel, of the second kind, 767 Function (cont.)

Bounded, 40, 41 piecewise continuous, 797
composite, 25 piecewise monotonic, 779
composite exponential, 93 power, 22, 23, 93, 102
continued in even fashion, 791 power-exponential, 93
continued in odd fashion, 792 quadratic, 27
continuous, 57, 58 rational integral, 27, 244
continuous in a domain, 261 represented parametrically, 104, 123
continuous over an interval, 59 of several variables, 255
continuous on the left, 59 single-valued, 20
continuous at a point, 261 tabular representation of, 20
continuous on the right, 59 transcendental, 28
decrease of, 163 trigonometric, 22, 24, 102
decreasing, 20 unbounded, 41
differentiable, 74 Functional determinant, 636
Function (cont.) Functional relation, 19
differentiabl~ at a point, 267 Functional series, 733
discontinuous, 60 Functions
elementary, 26 hyperbolic, 110, Ill
explicit, 90, 91 linearly dependent, 539
explicitly defined, 90 linearly independent, 539
exponential, 22, 24, 93, 102 rational, 357
exponential-power, 93 Fundamental theorem of algebra, 245
fractional rational, 27
of a function, 25 G
Gauss, 385
Gauss function, 385
graphical representation of, 21
Gaussian curve, 187
harmonic, 703, 836
General solution (of a differential
Heaviside unit, 855
equation), 474, 475, 515
homogenous, 482
Geometric mean, 305
implicit, 90, 91, 122
Ceometric progression, 710
increase of, 163
Gradient, 286, 287
increasing, 20
Graph, 21
infinitesimal, 42, 44, 45 Greatest value (of a function), 61
initial, 855 Green, D., 681
inverse, 95 Green's formula, 679-681
inverse trigonometric, 22, 102
investigation of, 194-198 H
irrational, 27
Hamiltonian operator, 700
linear, 27 Harmonic analysis, 805
logarithmic, 22, 24, 103 Harmonic function, 703, 836
multiple-valued, 20 Harmonic oscillations, 558
periodic, 24 Harmonic series, 714, 715
888 Subject Index
Heat-conduction equation, 815, 816, Infinitesimal of lower order, 64

825, 828 Infinitesimal quantity, 45
Heaviside unit function, 855 Infinitesimals
Helicoid, 316 equivalent, 64, 65
Helix, 315, 316 of same order, 63
Hodograph, 314 Inflection (point of inflection), 186
Homogeneous equation, 482 Initial condition, 474, 514
Homogeneous function, 482 Initial conditions, 8, 18, 825, 828
Homogeneous linear equation, 529 Initial function, 855
Hyperbolic equations, 815 Initial phase, 558
Hyperbolic functions (sine, cosine, Integrable (said of a function), 399
tangent, cotangent), 110, 111 Integral, 473
Hypocycloid, 449 absolutely convergent, 420
complete, 4 75, 515
convergent, 421
Identity, 364 definite, 396, 398, 399
Imaginary Dirichlet's, 800
pure, 233 divergent, 421
Imaginary axis, 233 double, 609
Imaginary part of complex number, elliptic, 385
233 Fourier, 806-80.8
Implicit function, 90, 91, 122 improper, 416, 417
Improper fraction, 357 improper iterated, 631
Improper integral, 416, 417 indefinite, 343
Improper iterated integral, 631 iterated, 611
Inclined asymptotes, 191 line, 671, 674
Increasing function, 20 particular, 475
Increasing variable, 18 Poisson's, 835, 846
Increment three-fold iterated, 651-655
partial (of a function), 259 triple, 650, 654, 656, 658, 659
total (of a function), 259, 265 Integral curves, 475, 515
Indefinite integral, 343 Integral sign, 343
Independent variable, 19 Integral sum, 398, 608
Indeterminate forms, 144-147, 150-152 Integral test (for convergence), 723-726
Inequality Integrals
Bessel's, 796 lable of, 345
Bunyakovsky's, 647 Integrals of irrational functions, 371,
Schwarz', 647 383
Infinitely large quantity, 39 Integrand, 343
Infinitely large variable, 34 Integrate (a differential equation), 476
Infinitesimal, 42-45 Integrating factor, 495-497
Infinitesimal function, 42, 44, 45 Integration (of a function), 344
Infinitesimal of higher order, 64 Integration of binomial differentials,
Infinitesimal of kth order, 64 375
Subject Index 889
Integration by parts, 354-356, 413-416 Laplace transform, 855

Integration of rational fractions, 365 Laplace's equation, 507-509, 703, 815,
Integration by substitution, 348-351 836
Integration of trigonometric functions, Laplacian operator, 703, 836
378-383 Least value (of a function), 61
Interior point (of a region), 610 Leibniz (see Newton-Leibniz formula.
Interior points (of a domain), 256 410, 411)
Intermediate argument, 85 Leibniz' formula, 436
Interpolation, 250 Leibniz' rule (formula), 120
Interval, 17 Leibniz' theorem, 727, 728
closed, 17 Length of
of integration, 399 an arc, 208
open, 17 a normal, 127
Invariance (of form of differential), 117 a subnormal, 127
Inverse function, 95 a subtangent, 127
Inverse trigonometric function, 22, 102 a tangent, 127
Investigation of a function, 194-198 Level lines, 283
Involute, 219 Level surf aces, 283
Irrational function, 27 L'Hospital's theorem (rule), 145
Irrational numbers, 13 Limit
Irrotational vector field, 702 lower (of an infegral), 399
Isogonal trajectories, 509, 512-514 upper (of an integral), 399
Isolated singular pol nf, 310 Limit of
Iterated integral, 611 an algebraic sum of variables, 46
evaluation of, 615, 616 a function, 35, 261
improper, 631 a product, 46
three-fold, 651-655 a quotienf, 46
a variable, 32
J Line
Jacobi, 636 secant, 265
Jacobian, 636, 659 Line integral, 671, 674
Line tangenf, 73
K Linear density, 459
Linear differenlial equation, 528, 529
K:rylov, A. N., 435
Linear equation, 487
Linear function, 27
L
Linearity property (of a transform),
L-transform, 855 857
Lagrange form of remainder, !55 Linearly dependen~ functions, 539
Lagrange's interpolation formula, Linearly dependen~ solutions, 530
250, 251 Linearly independen~ functions, 539
Langrange's lheorem, 142 Linearly independent solutions, 530
Laplace equation in cylindrical coor- Lines
dinates, 842 flow, 510
890
level, 283 Minimum (of a function), 165, 169,

equipotential, 510 178, 292, 297
Logarithm Modulus, 15
common, 258 of a complex number, 234
Napierian, 56 of logarithms, 56
natural, 56, 758 Moments
Logarithmic derivative, 94 static, 649
Logarithmic function, 22, 24, 103 Monotonicity, 226, 227
Lopshifs, A. M., 806 Multiple roots (of a polynomial),
Lower limif (of an integral), 399 247
Lower (integral) sum, 397 Multiple-value function, 20
Lyapunov, A. M., 576, 581 Multiplicity (of roots), 247-249
Lyapunov stable (abou~ solutions,
conditions), 577 N
Lyapunov equation, 796
Lyapunov's theory of stability, 576 Nth partial sum of a series, 710
Napier, 56
M Napierian logarithms, 56
Maclaurin's formula, !55 Natural logarithms, 56, 758
Maclaurin's series, 751-753 Necessary condition (for existence of
Mapping extremum), 166
one-to-one, 634 Necessary conditions of an extremum,
Maxima (see maximum) 294 .
Maximum (of a function), 164, 169, Neumann problem, 837
178, 292, 297 Newfon-Leibniz formula, 410, 411
Maximum absolute error, 270 Newton's method, 227
Maximum deviation, 793 Neighbourhood (of a point), 17, 260
Maximum relative error, 272 centre of, 18
Mean radius of, 18
geometric, 305 Nodal point, 307
Mean-value theorem, 406, 616, 653 Nonhomogeneous linear equation, 529
Member Normal, 221, 320
right-hand (of an equation), 529 principal (of a curve), 328
Method Normal to a curve, 126
of chords, 225 Normal to a surface, 339
combined, 229 Normal plane, 320
Euler's, 581-584 Normal system of equations, 564
Newton's, 227 Number
Ostrogradsky's, 368 complex, 233
of tangents, 227 e, 51, 53
of variation of arbitrary irrational, 13
constants (parameters), 543 rational, 13
Minima (see Minimum) Number (cont.)
Minimax, 297, 299 real, 13
Subject Index 891
Number pair, 256 Parameter, 103

Number quadruple, 258 Parametric, 103
Number scale, 13 equations, 103, 104, 314
Number ~riple, 257 Part
Numerical series, 710 principal (of an increment), 113
Partial derivative, 263-265
0 Partial derivatives
Operator of different orders, 279-283
v-operator, 700 Partial differential equations, 472
del, 700 Partial fractions, 358
Hamiltonian, 700 Partial increment (of a function), 259
Laplacian, 703, 836 Particular integral, 475
One-to-one correspondence (mapping), Particular solution, 475, 515
634 Partition unit, 401
One-parameter family of curves, 498 Period, 24
Open domain, 256 of oscillation, 558
Open interval, 17 Periodic function, 24
Operational calculus, 854 Piecewise continuous function, 797
order of a differential equation, 472 Piecewise monotonic function, 779
Ordered variable quantity, 18 Phase
Ordinary differential equation, 472 of a complex number, 234
Ordinary point, 305, 336 initial, 558
Origin (of a vector), 314 ' Plane
Original, 855 complex, 240
Original-transform tables, 855 normal, 320
Orthogonal lrajectories, 509-512 osculating, 331
Oscillations hngent, 338
forced, 557, 559-563 Plus-and-minus series, 72 9
free, 557, 875-876 Point
Oscillations (cont.) critical, 168, 294
harmonic, 558 of discontinuity, 60, 75
Osculating plane, 331 of inflection (of a curve). 186
Osculation (see Point of osculation'309) interior (of a region), 610
Ostrogradsky, M. V., 368, 636, 681, isolated singular, 310
699 nodal, 307
Ostrogradsky's formula, 697-700 ordinary, 305, 336
Ostrogradsky's method, 368 of osculation, 309
singular, 306, 322, 336
p Points
Parabola interior (of a domain), 256
safety, 502 Poisson's integral, 835, 847
Parabolic equations, 815 Polar axis, 28
Parabolic formula, 426 Polar coordinate system, 28
Parabolic trapezoid, 426 Polar coordinates, 28
892 Subject Index
Pole, 28 Radius of torsion (of a curve), 333

Polynomial, 27, 244 Radius vector, 314
Bernstein's, 252 Range of a variable, 17
Chebyshev, 253 Rate of motion, 70
Potential of a field, 459 Ratio (of a geomrtric progression), 710
Potential of a gravitational field, 696 Rational functions, 357
Potential of a vector, 684 Rational integral function, 27, 244
Potential vector field, 701 Rational ,,umbers, 13
Power-exponential function, 93 Ray, 626
Power function, 22, 23, 93, 102 Real axis, 233
Power series, 742 Real number, 13
Principal normal (of a curve), 328 Real part of complex number, 233
Principal part (of an increment), 113 Rectangular formula, 424, 425
Principal value (of an integral), 811 Region
Principle of localisation, 802 closed, 608
Problem of integration, 609
Dirichlet, 837 regular, 64, 611, 626
Dirichlet-Neumann 840, 843 regular in the x-direction, 611
Firs~ boundary-value, 825, 837 regular in the y-direction, 611
of interpolating a function, 250 Relative error, 272
Neumann, 837
Relation
second boundary-value, 837
functional, 19
of a simple pendulum, 522-525
Remainder, 154
Produd
Lagrange form of, 155
Wallis', 416
Remainder theorem, 244
Progression
Resonance, 563, 879
geometric, 710
Resonance curves, 561
Proper fraction, 357
Resultant, 872
Property
Right-hand member (of an equation),
linearity (of a transform), 1157
529
Pure imaginary, 233
Rolle's theorem, 140
Q Root
Quadratic function, 27 double, 546
Quadratic trinomial, 351 of an equation, 244
Quantity k 1 -tuple, 247
infinitely large, 39 of multiplicity k, 247-249
infinitesimal, 45 of a polynomial, 244
monotonic, 18 simple (single), 546
ordered variable, 18 Rool-mean-square deviation, 793
R Roots
complex, 248, 249
Radius of convergence, 744 multiple (of a polynomial), 247
Radius of curvature, 217, 328
Radius of a neighbourhood, 18 Rotation (of a vector function), 694
Subject Index 893
Rule Single-valued function, 20

Leibniz, 120 Singular point, 306, 322, 336
L'Hospital's, 145 isolated, 310
Simpson's, 426, 428 Singular solution (of differential equa-
trapezoidal, 425, 426 tion), 504
Smallest value (of a function), 61
s Smirnov, V. I., 806
Smooth curve, 528
Safety parabola, 502 Solenoidal vector field, 702
Scalar field, 283 Solid of revolution, 455
Scale Sol uti on
number, 13 of a differential equation, 473
Schwarz' inequality, 647 general, 474, 475, 515
Secant line, 265 particular, 475, 515
Second derivative singular, 504
mechanical significance of, 124 stable, 577, 579, 580
Second Euler substitution, 373 unstable, 579, 580
Sense of description, 672 Solutions
Sense of integration, 671 linearly dependenf, 530
Separated variables, 479 linearly independent, 530
Series Solve (a differential equation), 476
absolutely convergent, 731 Space curve, 450
alternating, 727 Spiral of Archimedes, 29
binomial, 754-756 Stable
conditionally convergent, 731 Lyapunov (about solutions, condi-
dominated, 734-736 tions), 577
Fourier, 776-812 Stable solution, 577, 579, 580
definition of, 779 Static moments, 649
functional, 733 Stokes, D., 694
harmonic, 714, 715 Stokes' formula, 692-697
Maclaurin's, 751-753 Stokes' theorem, 694-695
numerical, 710 Subinterval, 401
plus-and-minus, 729 Subnormal, 127
power, 742 Subregions, 608
Taylor's, 750, 751 Substitution
trigonometric, 776 Euler, 372-375
Serret-Frenet formulas, 335 universal trigonomefric, 379
Shift theorem, 858 Subtangent, 127
Sign Sufficien~ conditions (for existence
of double substitution, 411 of an extremum), 169
integral, 343 Sum
Simple (single) root, 546 integral, 398, 608
Simpson's formula, 428 lower (integral), 397
Simpson's rule, 426, 428 upper (integral), 397
Single (simple) root, 546 Sum of a series, 710
Subject Index
nth partial, 710 Stokes', 694, 695

Surface density, 642 uniqueness, 855
Surfaces Weierstrass' approximation, 252
level, 283 Theory of stability
Symbolic vector, 700 Lyapunov's, 576
Third Euler substitution, 374, 375
T Threefold iterated integral, 651-655
Table of integrals, 345 Torsion (of a curve), 333
Table of transforms, 862, 863 radius of, 333
Tables Total derivative, 275
original-transform, 855 Total differential, 267
Tacnode, 309 Total differentials
Tangent, 73, 336 approximation by, 268, 269
line, 73 Total increment (of a function).259,265
Tangent plane, 338 Trajectories
Taylor's formula, I 52, I 55 isogonal, 509, 512-514
for a function of two variables, 290 orthogonal, 509-512
Taylor's series, 750, 751 Transcendental function, 28
Telegraph equations, 819 Transform (L-transform), 855
Terminus (of a vector), 314 Transform, 855, 856
Terms of a series, 710 Fourier, 812
Test Fourier cosine, 810
Cauchy's, 721, 722 FouriPr inverse, 812
d' Alembert's, 718 Fourier sine, 810
integral (for convergence), 723-726 Laplace, 855
Theorem Transform equation, 865
Abel's, 742, 743 Transforms
Cauchy's, 143 of derivatives, 861, 862
convolution, 871 differentiation of, 860, 861
decomposition, 867 Trapezoid
delay, 879, 880 curvilinear, 400
existence (of a line integral), 673 parabolic, 426
Theorem (cont.) Trapezoidal formula, 426
on finite increments, 142 Trapezoidal rule, 425, 426
fundamental (of algebra), 245 Trigonometric function, 22, 24, 102
l'Hospital's, 145 Trigonometric series, 776
Lagrange's, 142 Trinomial
Leibniz', 727, 728 quadratic, 351
mean-value, 406, 616, 653 Triple integral, 650, 654, 656, 658, 659
on ratio !>f increments of lwo fun- Triple product (of vectors), 333, 334
ctions, 143
remainder, 244 u
Rolle's, 140 Unbounded function, 41
shift, 858 Uniqueness theorem, 855
Sttbject Index 895
Universal trigonometric substitution, of infegration, 399

379 monotonically varying, 18
Unstable solution, 679, 580 Variables separable, 479
Upper limif (of an integral), 399 Variables
Upper (integral) sum, 397 separated, 479
Vector, 233
v symbolic, 700
Value Vector equation, 314
absolute, 15 Vector field
critical, 168 of gradients, 287
greatest (of a function), 61 irrofational, 702
leasf (of a function), 61 potential, 701
principal (of an integral), 811 solenoidal, 702
smallesf (of a function), 61 Velocity of motion, 70
Values Vertical asymptotes, 190
e-xtreme (of a function), 166
Variable, 16 w
bounded, 18 Wallis'formula, 415, 416
complex, 240 Wallis'product, 416
decreasing, 18 Wave equation, 815, 817
increasing, 18 Weierstrass' approximation theorem,
independenT, 19 252
infinitely large, 34 Wronkskian, 530-533, 542, 544, E5l.

Differential and Integral Calculus - N Piskunov

Uploaded by

Copyright:

Available Formats

Differential and Integral Calculus - N Piskunov

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Differential and Integral Calculus - N Piskunov

Uploaded by

Copyright:

Available Formats

N.

Chapter /. NUMBER. VARIABLE. FUNCTION

1. Real Numbers. Real Numbers as Points on a Number Scale 13

Chapter 1/. LIMIT. CONTINUITY OF A FUNCTION

1. The Limit of a Variable. An Infinitely Large Variable . 32

11. An Implicit Function and Its Differentiation . . . . . . . . . . . 89

Chapter IV. SOME THEOREMS ON DIFFERENTIABLE FUNCTIONS

1. A Theorem on the Roots of a Derivative (Rolle's Theorem) . 140

5. The Limit of a Ratio of Two Infinitely Large Quantities (Evaluation

of Indeterminate Forms of the Type : ) . . . . . 147

Chapter V. INVESTIGATING THE BEHAVIOUR OF FUNCTIONS

1. Statement of the Problem . . . . . . . . . 162

7. Applying the Theory of Maxima and Minima of Functions to the

Chapte,. VI. THE CURVATURE OF A CURVE

I. The Length of an Arc and Its Derivative . . . . ... 208

Chapte,. Vll. COMPLEX NUMBERS. POLYNOMIALS

I. Complex Numbers. Basic Definitions . . . . . . . 233

Chaptu V/1/. FUNCTIONS OF SEVERAL VARIABLES

l. Definition of a Function of Several Variables ..... 255

13. Level Surfaces . . . . . . . . . . . . 283

Chapter IX. APPLICATIONS OF DIFFERENTIAL CALCULUS TO SOLID GEOMETRY

I. The Equations of a Curve in Space . . . . . . . . . . . . . . 314

Chapter }(. INDEFINITE INTEGRALS

1. Antiderivative and the Indefinite Integral . . . . 342

Chapter XI. THE DEFINITE INTEGRAL

6. Integration by Parts . . . . . . 413

Chapter XII. GEOMETRIC AND MECHANICAL APPLICATIONS OF THE DEFINITE

I. Computing Areas in Rectangular Coordinates . . . . . . . . . . 442

Chapter X/1/. DIFFERENTIAL EQUATIONS

I. Statement of the Problem. The Equation of Motion of a Body with

25. Higher-Order Nonhomogeneous Linear Equations . . . 551

Chapter XIV. MULTIPLE INTEGRALS

1. Double Integrals . . . . . . . . . . . . . . . 608

L Line Integrals . . . . . . . . . . . . . . . . 670

Chapter XV/. SERIES

1. Series. Sum of a Series . . . . . . . 710

4. D' Alembert's Test . . . . . . . . . . . . . . 718

Chapter XV/I. FOURIER SERIES

Chapter XVIII. EQUATIONS OF MATHEMATICAL PHYSICS

7. Propagation of Heat in an Unbounded Rod . . . . . . . . . .. 831

I. The Initial Function and Its Transform . . . . . . . . . . . . . . 854

This text is designed as a course of mathematics for higher technical

NUMBER. VARIABLE. FUNCTION

SEC. 1. REAL NUMBERS. REAL NUMBERS AS POINTS ON A

Number is one of the basic concepts of mathematics. It originated

Rational numbers may be represented in the form of periodic

distance OM 2 = - X 2 (Fig. 1). The point 0 represents the number

SEC. 2. THE ABSOLUTE VALUE OF A REAL NUMBER

Definition. The absolute value (or modulus) of a real number x

SEC. 3. VARIABLES AND CONSTANTS

The numerical values of such physical quantities as time, length,

SEC. 4. THE RANGE OF A VARIABLE

A variable takes on a series of numerical values. The collection

Definition. The set of all numerical values of a variable quantity

The foregoing definitions may be formulated for a "point" in

considers the neighbourhood (a, b)

Note 1. If we have a functional relation of two variable -quan-

runction, then the function y = f (x) is called increasing. A decreas-

Note 2. The definition of function is sometimes broadened 50

SEC. 7. WAYS OF REPRESENTING FUNCTIONS