MATH 235, by T.

Lakoba, University of Vermont 60

7 Pendulum. Part II: More complicated situations

In this Lecture, we will pursue two main goals. First, we will take a glimpse at a method of
Classical Mechanics that allows one to obtain equations of motion in a much easier way than the
Second Law of Newton does. We will apply this method to derive equations of motion in three
cases where using the Second Law of Newton would be rather difficult. You will further practice
this method in homework problems. However, we will not try to justify this powerful method
this is usually done in graduate courses of Classical Mechanics or Variational Calculus.
Second, we will continue to practice using linearization of equations near equilibrium states,
where we will used the information obtained to analyze stability of these equilibria.
In one of the examples that we will consider, we will enounter a new form of a perturbation
expansion, which will be conceptually different from the one we used in Lecture 3. Again, we
will use this new form without any rigorous justification; such a justification is usually provided
in a graduate course on the perturbation theory.

7.1 Motivation

Consider a pendulum shown on the left, where the

mass is allowed to slide along the weightless rod that
swings in a vertical plane. The mass is attached to the
pivot point of the rod by a spring, so that the motion
of the mass along the rod is restricted by what the
spring allows the mass to do.
As you may suspect, it would be rather difficult to set
up the Second Law of Newton in this case while keep-
ing track of all the forces that act here. Fortunately,
there is an easier method to obtain the equations of
motion.

7.2 Euler-Lagrange equations

The Lagrangian of a system is defined as its kinetic energy minus the potential energy:

L = T V. (7.1)

Both the kinetic and the potential energy may depend on the coordinates that describe the
system. We will denote those coordinates by q1 , q2 , . . .. These are not necessarily the Cartesian
coordinates, as we will see in the examples below.
The kinetic energy also depends on the velocities, and the potential energy usually does
not. In this lecture we will use the notation q = dq dt
, where t is the actual (dimensional) time.
Thus, T = T (q1 , q2 , . . . ; q1 , q2 , . . .), V = V (q1 , q2 , . . .), so that

L = L(q1 , q2 , . . . ; q1 , q2 , . . .).

Then the equations of motion for the coordinates, called Euler-Lagrange equations, are:

L d L
= 0, i = 1, 2, . . . (7.2)
qi dt qi
MATH 235, by T. Lakoba, University of Vermont 61

where L/qi is the partial derivative of L with respect to qi . Recall that when taking the
partial derivative with respect to qi , all the other variables (including qi !) must be treated as
constants.
Verify (yes, do it) that the dimensional units of both terms in (7.2) are the same. (Otherwise,
the equation would not make sense.)
We will now apply Eq. (7.2) to three different modifications of the simple pendulum problem
considered in the previous lecture.

7.3 Pendulum on a spring

Let us return to the problem described in Sec. 7.1. Let l be the natural length of the spring
and be the amount by which the mass compresses or stretches the spring. Thus, (l + )
is the total length of the spring. Also, let be the angle by which the rod deviates from the
vertical. Finally, let x and y be the Cartesian coordinates of the mass, where the Cartesian
origin is at the pivot point (see the figure in Sec 7.1).
The kinetic energy of the mass equals:
mx2 my 2
T = + . (7.3a)
2 2
The potential energy is the sum of the gravitational potential energy and the spring potential
energy:
k2
V = mgy + . (7.3b)
2
Thus, using (7.1), one has that
m(x2 + y 2 ) k2
L= mgy . (7.3c)
2 2
We will now express x and y via l and . From the aforementioned figure, we obtain:

x = (l + ) sin , y = (l + ) cos , (7.4a)

sin ,
x = (l + ) cos + cos .
y = (l + ) sin (7.4b)
Then:

x2 + y 2 = (l + )2 2 cos2 + 2(l + ) cos sin +

2 sin2 +
cos +
(l + )2 2 sin2 2(l + ) sin 2 cos2
= (l + )2 2 + 2.

Substituting the last expression and y from (7.4a) into (7.3c), we obtain:
m 2
2 + mg(l + ) cos k .
L= (l + )2 2 + (7.5)
2 2
The coordinates q1 , q2 in this Lagrangian are:

q1 = , q2 = .

Substituting (7.5) into (7.2) yields:

L d L
=0
dt
MATH 235, by T. Lakoba, University of Vermont 62

d m 2

mg(l + ) sin (l + ) 2 = 0
dt 2
+ (l + )2 ] = 0
g(l + ) sin [2(l + )
+ (l + ) = 0.
g sin + 2 (7.6a)
(Recall that when taking /, the variables , , are treated as constants, and similarly for
. )
/ , /, /
The other Euler-Lagrange equation is:
L d L
=0

dt
hm i d hm i
2
2(l + ) + mg cos k
2 = 0
2 dt 2
k = 0.
(l + ) 2 + g cos (7.6b)
m
The coupled equations (7.6a) and (7.6b) describe the motion of the pendulum on a spring.
Before we proceed with their analysis, we need to check some limiting case(s) where we
know the result. Such a case is that of a very tight spring (i.e., k ), where we expect
these two equations to reduce to the equation of a simple pendulum for , and to = 0. The
latter equation indeed follows from (7.6b) rewritten as:
mh i
0 as
= (l + )2 + g cos k .
k
Then, Eq. (7.6a) with = 0 reduces to Eq. (6.2) of Lecture 6 (verify).
The coupled nonlinear second-order equations (7.6) probably cannot be solved analytically.
However, we can analyze them near the equilibrium states of the system by means of lineariza-
tion. Recall that at an equilibrium, all time derivatives vanish identically:
= =
= = 0.

Using this condition in Eqs. (7.6) yields:

k
sin = 0 and g cos = , (7.7a)
m
which implies two different equilibria:
mg mg
E = 0, E = and E = , E = . (7.7b)
k k
What is the physical meaning of these formulae (i.e., can you see how they follow from elemen-
tary considerations)?
We now linearize Eqs. (7.6) near each equilibrium, to which end we take
= E + , 1;
= E + , E .

We substitute (7.8) into (7.6) and keep only terms O() and O(); all higher-order terms are
neglected. For the first equilibrium in (7.7b), one has:

sin = sin ,
MATH 235, by T. Lakoba, University of Vermont 63

cos = cos = 1 + O(2 ),

= O(),
= O().
Then:
(7.6a) g + O( ) + (l + E + ) = 0

g
= , (7.9a)
l + E
k
(7.6b) O( 2 ) + g (E + ) = 0
m
k
= . (7.9b)
m
In obtaining the final equations in both (7.9a) and (7.9b), we have omitted second-order small
terms. Moreover, in deriving the second line of (7.9b) from the previous line, we used the
expression for E from (7.7b).
Let us repeat the last sentence, as the corresponding step is very important and will occur
every time we derive linearized equations. So, when obtaining first-order equations for small
deviations about an equilibrium, one always uses the zeroth-order equation(s) satisfied by the
system variables in this equilibrium. For example, in deriving the linearizations equations (7.9),
we used the equilibrium equations stated in the first parentheses in (7.7b).
Let us now continue with the analysis of Eqs. (7.9). Equation (7.9a) is precisely the equation
of small oscillations of a pendulum of length (l + E ); see Eqs. (6.2) and (6.4) in Lecture 6.
Equation (7.9b) is precisely the equation of (small) oscillations of mass m on a spring (as
derived in a physics course). From these statements we make two conclusions:

(i) Small oscillations of the pendulum on a spring decouple into two independent modes of
motion: the oscillations of a rigid pendulum and the oscillations of a mass on a non-
swinging spring.
It should be noted that such a decoupling of complex oscillations into basic modes does
not occur always; it is even true that in general, it would not occur.

(ii) Since both Eqs. (7.9a) and (7.9b) have the form of Eq. (6.4) of Lecture 6, which describes
stable oscillations, then the small oscillations of the pendulum on a spring near the first
equilibrium in (7.7b) (i.e., the down position), are stable.

Now consider the small deviation of the pendulum from the second equilibrium (7.7b). Then

sin = sin( + ) = sin

cos = cos( + ) = cos 1

= O() ,
= O().
Similarly to (7.9a) and (7.9b), one obtains (verify):

g + (l + E ) = 0 (7.10a)

k
= 0. (7.10b)
m
In obtaining (7.10b) we have used the definition of E for the second equilibrium in (7.7b).
MATH 235, by T. Lakoba, University of Vermont 64

Conclusion (i) above holds in this case also: the rotational motion of the pendulum and
the vibrations of the mass on the spring are decoupled. But while the latter motion remains
stable (note that (7.10b) is equivalent to (7.9b)), the rotational motion is now unstable. Indeed,
(7.10a) is an analog of Eq. (6.26) of Lecture 6 it describes the unstable motion of an inverted
pendulum. This is precisely what one would intuitively expect in this case.

7.4 Pendulum on a rotating platform

A problem that we will solve in this section is as follows.

A gate is attached to a platform that rotates

uniformly with angular velocity = about
a vertical axis. A pendulum is attached to
the gate so that it can move in a plane per-
pendicular to that of the gate, and its pivot
point is on the axis of rotation of the plat-
form. The mass of the bob is m and the
length of the rod is l. Find the equation of
motion of the pendulum and investigate the
stability of its solution near the equilibrium
state(s).

The location of the bob is now characterized

by three Cartesian coordinates x, y, z:

z = |OM | cos
x = |OA| cos
y = |OA| sin .
Note that |OM |, , and are just the spher-
ical coordinates of point M . Also,

|OA| = |OM | sin .

Finally, using the given information that |OM | = l and = t, one has:

x = l sin cos(t)
y = l sin sin(t) (7.10)
z = l cos .

The kinetic energy is

m 2
T = (x + y 2 + z 2 ) , (7.12a)
2
and the potential energy is simply
V = mgz . (7.12b)
MATH 235, by T. Lakoba, University of Vermont 65

To compute T , we first compute the time derivatives of the Cartesian coordinates (7.10):

x = l[ cos cos(t) sin sin(t)]

y = l[ cos sin(t) + sin cos(t)] (7.12)
z = l sin .

Substituting formulae (7.12) into (7.12a), one finds (verify):

ml2 2 2
T = ( sin + 2 ).
2
Combining this expression with (7.12b), the last line of (7.10), and (7.1) one has:

ml2 2 2
L= ( sin + 2 ) + mgl cos . (7.13)
2
This Lagrangian has only one coordinate: . Then the Euler-Lagrange equation is:
L d L
=0
dt

ml2 2 d ml2
2 sin cos mgl sin 2 = 0
2 dt 2
g
= 2 sin cos sin .
l
Introducing the notations
g
02 = , R= , (7.14)
l 0
the previous equation can be rewritten as

= 02 sin (R2 cos 1). (7.15)

This is the equation for the pendulum on a rotating platform. We now find its equilibria and
then linearize about them.
The equilibria of (7.15) occur where = 0, and hence the r.h.s. of (7.15) vanishes. This
occurs in three distinct cases:
a) sin = 0 E = 0;
b) sin = 0 E = ;
c) R cos 1 = 0 E = arccos R12 .
2

Note that the two states in case c) exist only for R > 1. From the symmetry considerations,
these states are physically equivalent, since one can be obtained from the other by rotating the
platform by 180 . Therefore, below we will consider only one of these states, say,

1
E = + arccos .
R2

As in Eq. (7.8), we take

= E + , 1.
In case a), we have:
sin = sin , cos = cos 1 .
MATH 235, by T. Lakoba, University of Vermont 66

Then the linearized version of (7.15) becomes

= 02 (R2 1). (7.17a)
Thus the down position of the pendulum is stable when R2 1 and unstable otherwise.
In case b), we have:
sin = sin( + )
cos = cos( + ) 1 ,
whence
= 02 (R2 1) = 02 (R2 + 1). (7.17b)
Thus, the up position is always unstable.
In case c), we use the first two terms of the Taylor series for sin(E + ) and cos(E + ):
sin = sin(E + ) sin E + cos E ,
cos = cos(E + ) cos E sin E ,
whence:
= 02 (sin E + cos E )(R2 [cos E sin E ] 1) .
Now, recall the step that we emphasized after deriving Eqs. (7.9). Accordingly, we use the
equation satisfied by the equilibrium angle E in case c): (R2 cos E 1) = 0. Then we see
that the underlined terms in the equation above cancel out. Finally, neglecting terms O(2 ),
as before, we obtain (verify):
= 02 R2 sin2 E . (7.17c)

Since the coefficient in front of on the

r.h.s. is always negative, we conclude that
this tilted equilibrium is stable whenever
it exists, i.e. for R > 1.
In other words, this tilted equilibrium ex-
ists, and is stable, when > 0 (see (7.14)),
i.e. the angular speed of the platform exceeds
the natural frequency of a simple pendulum.

Using the expression sin2 E = 1 cos2 E = 1 R14 and R2 = (2 /02 ), Eq. (7.17c) can be
rewritten as (verify):
2 04
= 2 . (7.17c0 )

To conclude the consideration of this phys-

ical system, we summarize its equilibrium
states and their stability in a so-called bi-
furcation diagram, shown on the left. The
solid (dashed) lines depict the stable (unsta-
ble) equilibria as functions of the bifurcation
parameter R.
MATH 235, by T. Lakoba, University of Vermont 67

7.5 Pendulum with the pivot point rapidly oscillating in the vertical
direction
In this example, we will find that a pendulum described in the title of this section can be stable
in the up position! This physical system was first analysed in 1951 (yes, so recently!) by
a great Russian physicist Pyotr Kapitza (Kapitsa). It is a particular case of a more general
situation where a system whose natural time scale is of order one (i.e. is neither too fast
or too slow) is affected by a fast periodic force. Some links to applications of such systems in
physics (and not just in mechanics) are pointed out on the course web page.

Denote the coordinate of the pivot point

y0 (t) = a cos t

(the minus sign is just for the convenience later

on). Thus, a and are the amplitude and fre-
quency of the pivots vibrations. Clearly,

x = l sin , (7.18a)

y = l cos +y0 (t) = (l cos +a cos t). (7.18b)

The kinetic energy is

m 2
T = (x + y 2 ).
2
From (7.18), we find:
x = l cos (7.19a)
y = l sin + a sin t . (7.19b)
The potential energy is
V = mgy.
Combining these equations and Eq. (7.1), we obtain the Lagrangian (verify):
m 2 2
L= ( l + 2 a l sin sin t + 2 a2 sin2 t ) + mgl cos + mga cos t . (7.19)
2
The only coordinate in this Lagrangian is the angle . The Euler-Lagrange equation is:

L d L
=0
dt

m d ml2 m
2 a l cos sin t mgl sin 2 + 2 a l sin sin t = 0 .
2 dt 2 2
(Note that the two terms in the Lagrangian that depend only on time but not on and , do
not contribute to the equation of motion.)
Simplifying the above equation, we have:

a l cos sin t gl sin ( l2 + a l cos sin t + a l 2 sin cos t ) = 0

g sin = l + a 2 sin cos t ,

g a l
= sin 1 + 2 cos t .
l l g
MATH 235, by T. Lakoba, University of Vermont 68

Using notations (7.14) and also introducing a new notation

a
= , (7.20)
l
the last equation is rewritten as:
= 02 sin (1 + R2 cos t) . (7.22a)
Since the analysis of this problem will be longer than the analysis of the two previous problems,
we introduce one more notation: as in Lecture 6, we define the nondimensional time
= 0 t

(so that t = t
0 0
= R ). With this change of variables, (7.22a) becomes (verify):
d2
2
= sin (1 + R2 cos R ) . (7.22b)
d
In what follows we will analyze Eq. (7.22b). Let us stress that the method we will use is
mathematically incorrect; it is a physicists method based mostly on intuition (originally on
that of the great physicist Kapitza). Both establishing a mathematically correct method for
this problem and explaining why the method we will use is incorrect, belong to the material of a
graduate course on perturbation theory. (E.g., solving (7.22b) in a mathematically correct way
is assigned as Problem 6 for Sec. 6.4 in M.H. Holmes, Introduction to perturbation methods,
Springer, New York, 1998.)
The following analysis is based on two assumptions:
(i) The amplitude of the pivots vibrations is small compared to the length of the pendulum:
a
1; (7.23a)
l
(ii) The frequency of the pivots vibrations is large compared to the natural frequency of the
pendulum:

R 1. (7.23b)
0
We do not specify any relation between and R at this moment; it will become apparent later.
Now, since the amplitude of the pivots vibrations is small, Kapitza sought the solution of
(7.22b) in the form:
= (0)
|{z} + (1)
|{z} . (7.23)
solution varying on the fast and small oscillations
time scale of the natural period superimposed on the
of the pendulum, T = 2/0 slower solution (0)

The fast oscillations can be assumed to be small be-

cause the vibrations of the pivot are small.
Note that unlike in Lecture 3, here the slower
solution (0) is not just the solution of a pen-
dulum with a fixed pivot. All that we assume
about (0) is that it does not have any fast com-
ponents.
MATH 235, by T. Lakoba, University of Vermont 69

Preparing to substitute (7.23) into (7.22b), we write:

sin = sin((0) + (1) ) sin (0) + (1) cos (0) .

This approximation comes from the same Taylor series as used in Case c) in Sec. 7.4, where
now the role of 1 is played by (1) 1 (see (7.23a) and (7.23)). Substituting the last
equation and Eq. (7.23) into (7.22b), we have:

d2 (0) d2 (1)
+ = (sin (0) + (1) cos (0) )(1 + R2 cos R ) . (7.24)
d 2 d 2
In (7.24) let us separate the fast terms from the slower ones. Note that this, again, is different
from the perturbation theory of Lecture 3, where we separated terms based on their size and
not on their time scale.
Fast terms varying as cos R :

d2 (1)
= sin (0) R2 cos R (1) cos (0) , (7.26a)
d 2
Slower terms:
d2 (0)
2
= sin (0) (1) cos (0) R2 cos R . (7.26b)
d
Why is the underlined term in (7.26b), which is the product of two fast terms, (1) and cos(R ),
included into the equation for the slower terms?? The answer will come soon.
Consider first the fast equation (7.26a). This is a linear differential equation with the
forcing term. Moreover, on the time scale of the fast oscillations, the slower function (0) changes
very little (usually, much less than schematically shown in the figure above). Therefore, in the
fast equation, the slower variable (0) can be considered as a constant.
Now, instead of solving (7.26a) exactly, we employ the following hand-waving argument5 :
Let us disregard the second term on the r.h.s., because it is O(1), while the first term is
O(R2 ) 1 (see (7.23b)). Then we simply have:

d2 (1)
2
= sin (0) R2 cos R ,
d
whose solution is
(1) = sin (0) cos R . (7.26)
When verifying (7.26), recall that in the fast equation (7.26a), (0) is to be treated as a
constant on the fast time scale.
Substituting (7.26) into the slower equation (7.26b), we find:

d2 (0)
= sin (0) 2 R2 cos (0) sin (0) cos2 R . (7.27)
d 2
We now recall the trigonometric identity

1 + cos(2R )
cos2 (R ) =
2
5
Again, recall that we are handling this problem using physical intuition rather than mathematical rigor.
MATH 235, by T. Lakoba, University of Vermont 70

and discard the fast term cos(2R ) since by design, Eq. (7.27) is supposed not to have fast
terms. Then (7.27) becomes:

d2 (0) (0) (R)2 (0)
= sin 1+ cos . (7.28)
d 2 2

Note that (7.28) contains only the slower variable (0) . Also note that this equation is different
from the equation of a pendulum with a fixed pivot point, Eq. (6.3) of Lecture 6, as we have
already announced after Eq. (7.23). The difference, i.e. the second term on the r.h.s. of (7.28),
comes from the coupling of the fast part of the solution, (1) , with the external force cos(R );
see Eq. (7.26b).
We now look for the equilibrium states of this equation and then will consider their stability.
As earlier for Eq. (7.15), for Eq. (7.28) we also have three different cases:

(0)
a) sin (0) = 0 E = 0;
(0)
b) sin (0) = 0 E = ;
(R)2 (0) (0) 2
c) 1 + 2
cos =0 E = arccos .
(R)2

As explained in the corresponding place in

Sec. 7.4, we can consider only one of the two
equilibria in case c), say,

(0) 2
E = arccos .
(R)2

It corresponds to the upper point of intersec-

tion between the dashed line and the circle
in the figure on the left.

A study of the stability of these equilibria proceeds as in Sec. 7.4. Below I will present
only the final equations (followed by their interpretations); you are responsible for verification
of the details.
In case a),

d2 (R)2
= 1 + . (7.30a)
d 2 2
Thus, the down position of this pendulum is stable for all amplitudes and frequencies of the
pivots vibrations (at least as long as assumptions (7.23) hold).
MATH 235, by T. Lakoba, University of Vermont 71

In case b):

d2 (R)2
= 1 . (7.30b)
d 2 2
Thus, the up position of this pendulum is unstable when (R)2 < 2 and stable when (R)2 > 2.
That is, sufficiently, fast vibrations, with
l
> 2 ,
0 a
would stabilize the pendulum in its inverted position!
In case c):

d2 (0) (0)
(R)2 (0) (R)2 (R)2
(0) (0)
= (sin E + cos E ) 1 + cos E sin E = sin2 E +O(2 ) .
d 2 | 2{z } 2 2
=0
(7.30c)
(The terms with the underbracket cancel out by the argument emphasized after Eq. (7.9b).)
Thus, the tilted equilibria of this pendulum are always (i.e. for any vibration frequency)
unstable.
Finally, we summarize the above conclusions in a bifurcation diagram for this physical
system: