ME340A Elasticity Spring 2010 Wei Cai Stanford University

11
4/23/06 8:13 PM

C:\Documents and Settings\Wei Cai\My Documents\Courses...\S522a.m

%adpated from maple file S522 from J. R. Barber, Elasticity

% http://www-personal.engin.umich.edu/~jbarber/elasticity/maple/S522
%
% adapted by Wei Cai, [email protected], for ME 340 Elasticity
% Spring 2006, Stanford University
%
%This file gives the solution of the simply-supported beam problem,
%following the strategy of Section 5.2.2.
%
clear all
syms C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 C16 C17 C18
syms x y a b p
%stress function
phi = C1*x^2+C2*x*y+C3*y^2+C4*x^3+C5*x^2*y+C6*x*y^2+C7*y^3+C8*x^4+ ...
C9*x^3*y+C10*x^2*y^2+C11*x*y^3+C12*y^4+C13*x^5+C14*x^4*y+ ...
C15*x^3*y^2+C16*x^2*y^3+C17*x*y^4+C18*y
%stress field
sxx = diff(diff(phi,y),y)
syy = diff(diff(phi,x),x)
sxy = -diff(diff(phi,x),y)
%traction force on y=b
%Ty on y=b
t1 = subs(syy,y,b)
%Tx on y=b
t2 = subs(sxy,y,b)
%traction force on y=-b
%Ty on y=-b
t3 = subs(syy,y,-b)
%Tx on y=-b
t4 = subs(sxy,y,-b)
%traction force on x=a
%Tx on x=a
t5 = subs(sxx,x,a)
%Ty on x=a
t6 = subs(sxy,x,a)
%traction force on x=-a
%Tx on x=-a
t7 = subs(sxx,x,-a)
%Ty on x=-a
t8 = subs(sxy,x,-a)
%find coefficients of polynomials t1,t2,t3,t4
s1 = subs(diff(t1,x,3),x,0)/factorial(3)
s2 = subs(diff(t1,x,2),x,0)/factorial(2)
s3 = subs(diff(t1,x,1),x,0)/factorial(1)
s4 = subs(t1,x,0)
s5
s6
s7
s8

=
=
=
=

subs(diff(t2,x,3),x,0)/factorial(3)
subs(diff(t2,x,2),x,0)/factorial(2)
subs(diff(t2,x,1),x,0)/factorial(1)
subs(t2,x,0)

1 of 3

12
4/23/06 8:13 PM

C:\Documents and Settings\Wei Cai\My Documents\Courses...\S522a.m

s9
s10
s11
s12

=
=
=
=

subs(diff(t3,x,3),x,0)/factorial(3)
subs(diff(t3,x,2),x,0)/factorial(2)
subs(diff(t3,x,1),x,0)/factorial(1)
subs(t3,x,0)

s13
s14
s15
s16

=
=
=
=

subs(diff(t4,x,3),x,0)/factorial(3)
subs(diff(t4,x,2),x,0)/factorial(2)
subs(diff(t4,x,1),x,0)/factorial(1)
subs(t4,x,0)

%The biharmonic equation is 4th order, so applying it to a 5th order polynomial

%generates a first order polynomial.
%
biharm = diff(phi,x,4)+diff(phi,y,4)+2*diff(diff(phi,x,2),y,2)
%coefficients of biharm
b1 = subs(diff(biharm,x,1),{x,y},{0,0})
b2 = subs(diff(biharm,y,1),{x,y},{0,0})
b3 = subs(biharm,{x,y},{0,0})
%integrated force and torque on x=a
Fx1 = simplify(int(t5, y, -b, b))
Fy1 = simplify(int(t6, y, -b, b))
M1 = simplify(int(t5*y, y, -b, b))
%
%integrated force and torque on x=-a
Fx2 = simplify(int(t7, y, -b, b))
Fy2 = simplify(int(t8, y, -b, b))
M2 = simplify(int(t7*y, y, -b, b))
%Solve all these equations together
s = solve(s1, s2, s3, s4+p, s5, s6, s7, s8, ...
s9, s10,s11,s12, s13,s14,s15,s16, ...
b1, b2, b3,
Fx1,M1, Fy1-p*a, ...
'C1','C2','C3','C4','C5','C6','C7','C8','C9',...
'C10','C11','C12','C13','C14','C15','C16','C17','C18')
%construct cell arrays for future use
coeffs = {C1,C2,C3,C4,C5,C6,C7,C8,C9,C10,C11,C12,C13,C14,C15,C16,C17,C18}
solution = {s.C1,s.C2,s.C3,s.C4,s.C5,s.C6,s.C7,s.C8,s.C9, ...
s.C10,s.C11,s.C12,s.C13,s.C14,s.C15,s.C16,s.C17,s.C18}
%stress function and stress field solution
phi2 = simplify(subs(phi, coeffs, solution))
sxx2 =
syy2 =
sxy2 =

simplify(subs(sxx, coeffs, solution))

simplify(subs(syy, coeffs, solution))
simplify(subs(sxy, coeffs, solution))

%print out results

disp('phi=') pretty(phi2)
disp('sxx=') pretty(sxx2)

2 of 3

13
4/23/06 8:13 PM
disp('syy=')
disp('sxy=')

C:\Documents and Settings\Wei Cai\My Documents\Courses...\S522a.m

pretty(syy2)
pretty(sxy2)

%plot stress fields

bn =
pn =
an =
xn = [-1:0.05:1]*an
yn = [-1:0.1:1]*bn
xx = ones(size(yn'))*xn
yy = yn'*ones(size(xn))
phin = subs(phi2,{a,b,p},{an,bn,pn})
sxxn = subs(sxx2,{a,b,p},{an,bn,pn})
syyn = subs(syy2,{a,b,p},{an,bn,pn})
sxyn = subs(sxy2,{a,b,p},{an,bn,pn})

phin
sxxn
syyn
sxyn

=
=
=
=

subs(phin,{x,y},{xx,yy})
subs(sxxn,{x,y},{xx,yy})
subs(syyn,y,yy)
subs(sxyn,{x,y},{xx,yy})

figure(1)
subplot(2,2,1) mesh(xn,yn,phin)
title('\phi') xlabel('x') ylabel('y')
subplot(2,2,2) mesh(xn,yn,sxxn)
title('\sigma_{xx}') xlabel('x') ylabel('y')
subplot(2,2,3) mesh(xn,yn,syyn)
title('\sigma_{yy}') xlabel('x') ylabel('y')
subplot(2,2,4) mesh(xn,yn,sxyn)
title('\sigma_{xy}') xlabel('x') ylabel('y')
% results
%phi=
%
%
2 3
2
2
3 2
3 2
2 3
5
%
p (10 x b + 15 x y b - 2 y b + 5 y a - 5 x y + y )
%
- 1/40 -----------------------------------------------------------%
3
%
b
%sxx=
%
%
2
2
2
2
%
p y (-6 b + 15 a - 15 x + 10 y )
%
- 1/20 ----------------------------------%
3
%
b
%syy=
%
%
3
2
3
%
p (-2 b - 3 y b + y )
%
1/4 ----------------------%
3
%
b
%sxy=
%
%
2
2
%
p x (-b + y )
%
- 3/4 -------------%
3
%
b

3 of 3

xx

5
1

5
1

2
0
1

2
0

0
y

14

0
y

xy

yy

2
0
0

0.5
1
1

2
0

0
y

1 2

2
1
2

0
y

0
1

Appendix

Potential Field of a Uniformly Charged Ellipsoid

Wei Cai
Department of Mechanical Engineering, Stanford University, CA 94305-4040
May 28, 2007

Contents
1 Problem Statement

2 Orthogonal Curvilinear Coordinates

3 Elliptic Coordinates

4 Alternative Definition of Elliptic Coordinates

5 Ellipsoidal Coordinates

6 Ellipsoidal Conductor

7 Ellipsoidal Shell

8 Uniformly Charged Ellipsoid

11

9 Special Cases

12

10 Application to Hertz Contact Problem

14

A Matlab Files for Analytic Derivation

16

Problem Statement

The purpose of this document is to discuss the derivation of a very useful result, which states that
the potential field of a uniformly charged ellipsoid is a quadratic function inside the
ellipsoid [1]. Specifically, consider an ellipsoid with uniform charge density inside the following
region,
x2 y 2 z 2
+ 2 + 2 =1
a2
b
c

(1)

Let V0 specify the volume of the ellipsoid. The potential field is defined as
Z

(x)
dV (x0 )
0|
|x

x
V0
1

(2)

where x = (x, y, z) and x0 = (x0 , y 0 , z 0 ).

If point x is inside the ellipsoid [2],

Z 
x2
ds
y2
z2
p
1 2
(x, y, z) = abc
2
2
a +s b +s c +s
(s)
0

(3)

where
(s) (a2 + s)(b2 + s)(c2 + s)

(4)

If point x is outside the ellipsoid [1],


Z 
x2
ds
y2
z2
p
1 2
(x, y, z) = abc
2
2
a +s b +s c +s
(s)

(5)

where is the greatest root of the equation f (s) = 0, where

f (s)

x2
y2
z2
+
+
1
a2 + s b2 + s c2 + s

(6)

The physical significance of function f (s) is that, for each s > 0, the equation f (s) = 0 defines
an ellipsoid (larger than the original ellipsoid), which is an isosurface of the potential field (x)
generated by the original ellipsoid (with uniform charge density). Therefore, all the value of s
[0, ) corresponds to a family of ellipsoids, called the confocal family; the potential field is a
constant on each ellipsoid in the family.
Physical significance: As a consequence of the above result, the second spatial derivatives of the
potential field is a uniform second order tensor inside the ellipsoid. This is closely analogous to
Eshelbys Theorem, which states that the stress field inside an ellipsoidal inclusion is uniform [3].
Landau and Lifshitz also used the above result to show that normal stress distribution inside the
contact area between two smooth elastic media (Hertzian contact problem) has the ellipsoidal
shape [2].

Orthogonal Curvilinear Coordinates

The proof of this result is best discussed using ellipsoidal coordinates, which is an orthogonal curvilinear coordinate system. In this section, we first summarize the major results concerning orthogonal curvilinear coordinates. We will then introduce the elliptic coordinates (2D) and ellipsoidal
coordinates (3D) in the following sections.
Let the Cartesian coordinates be specified by (x1 , x2 , x3 ) = (x, y, z). An arbitrary differential
length in space ds is specified by (ds)2 = (x1 )2 + (x2 )2 + (x3 )2 . Now consider a general orthogonal
curvilinear coordinate system, (q1 , q2 , q3 ), which are related to the Cartesian coordinates by
qm = qm (x1 , x2 , x3 ),

xm = xm (q1 , q2 , q3 )

(7)

An arbitrary differential length in space can be expressed by

(ds)2 = (h1 dq1 )2 + (h2 dq2 )2 + (h3 dq3 )2

(8)

x = const
y = const
= const

= const

x = -a

x=a

Figure 1: Contour lines in Cartesian and elliptic coordinates [5].

where hi are called scale factors and have the following expressions [4].
(h1 )2 =
(h2 )2 =
(h3 )2 =

xk xk
q1 q1
xk xk
q2 q2
xk xk
q3 q3

(9)

The index notation is used here and k is a dummy index that is summed from 1 to 3. Let ek be
k be the basis vectors of the
the basis vectors of the Cartesian coordinates (x1 , x2 , x3 ) and let e
curvilinear coordinates (q1 , q2 , q3 ). The gradient of a scalar field (q1 , q2 , q3 ) is,
1
= e

1
1
1
2
3
+e
+e
h1 q1
h2 q2
h3 q3

(10)

The Laplacian of a scalar field (q1 , q2 , q3 ) is,









h2 h3

h3 h1

h1 h2
1
2
+
+
=
h1 h2 h3 q1
h1 q1
q2
h2 q2
q3
h3 q3
In 2-dimension, the Laplacian of a scalar field (q1 , q2 ) reduces to the following.






h2

h1
1
2 =
+
h1 h2 q1 h1 q1
q2 h2 q2

Elliptic Coordinates

The most common definition of elliptic coordinate (, ) is [5],

x = a cosh cos
y = a sinh sin
3

(11)

(12)

With this definition, we can show that

x2
x2
+
a2 cosh2 a2 sinh2
x2
x2

a2 cos2
a2 sin2

= cos2 + sin2 = 1
= cosh2 sinh2 = 1

(13)

Therefore, the contour lines of = const form a set of ellipses, and the contour lines of = const
form a set of hyperbolas, as shown in Fig. 1. This figure also shows that in the limit of a 0,
or when the distance from the origin is much greater than a, the elliptic coordinates becomes very
close to polar coordinates.
The Jacobian matrix between Cartesian coordinates (x, y) and elliptic coordinates (, ) is
"
# 

x
x
a sinh cos a cosh sin

J x x =
a cosh sin
a sinh cos

(14)

The elliptic coordinates is an orthogonal coordinate system because the two columns of matrix J
are orthogonal to each other. The scale factors are
s 
 2
x 2
y
h =
+

s 


x 2
y 2
h =
+
(15)

It is easy to show that

q
p
h = h = a sinh2 + sin2 = det(J)

(16)

Therefore, the Laplacian of a scalar field is

 2


2
2
+
(x, y)
=
x2 y 2
 2

1
2

=
+
(, )
a2 (sinh2 + sin2 ) 2 2

(17)

Alternative Definition of Elliptic Coordinates

An alternative definition of elliptic coordinates makes it more natural to generalize the concept to
ellipsoidal coordinates in 3D. Consider an ellipse
x2 y 2
+ 2 =1
a2
b

(18)

We will assume a > b without loss of generality. Now consider a family of curves defined by
f (s)

x2
y2
+
1=0
a2 + s b2 + s

(19)

When s > b2 , it defines an ellipse. When b2 > s > a2 , it defines a hyperbola.

4

For a given (x, y), let (, ) be the two largest roots of the equation f (s) = 0. There is a one-to-one
correspondence between (x, y) and (, ), if we assume x > 0, y > 0. Specifically,
x2 =
y2 =

(a2 + )(a2 + )
a2 b2
2
(b + )(b2 + )
b2 a2

(20)

This relationship can be verified by plugging it into the definition of f (s),

x2
y2
a2 +
b2 +
+

1
=
+
1=0
a2 + b2 +
a2 b2 b2 a2
b2 +
x2
y2
a2 +
+
1=0
f () =
+

1
=
a2 +
b2 +
a2 b2 b2 a2
The four components of the Jacobian matrix can be obtained.

1/2

1/2
x
1
a2 +
x
1
a2 +
=
=

2 (a2 + )(a2 b2 )

2 (a2 + )(a2 b2 )



1/2
1/2
1
b2 +
y
1
b2 +
y
=
=

2 (b2 + )(b2 a2 )

2 (b2 + )(b2 a2 )
f () =

(21)

(22)

The (, ) coordinate system is orthogonal because

x x y y
+
=0

(23)

Define function (s) (a2 + s)(b2 + s). The scale factors can be expressed as
r
1
h =
2 ()
r
1
h =
2 ()
Therefore, the Laplacian of a scalar field (, ) is





1
h
h
2
=
+
h h h
h
s
s
"
!
!#
p
4 ()()

()
()
=
+

()

()





p
p
4
p

=
()
()
+ ()
()

(24)

(25)

For derivation details see elliptic coord.m.

Ellipsoidal Coordinates

Generalizing the elliptic coordinates defined above, we obtain the ellipsoidal coordinates [6]. Consider an ellipsoid, Consider an ellipse
x2 y 2 z 2
+ 2 + 2 =1
a2
b
c

(26)
5

 = const
= const

= const

Figure 2: Isosurfaces of ellipsoidal coordinates [7].

We will assume a > b > c without loss of generality. Now consider a family of curves defined by
f (s)

x2
y2
z2
+
+
1=0
a2 + s b2 + s c2 + s

(27)

For > c2 , f () = 0 defines an ellipsoid. When c2 > > b2 , f () = 0 defines a one-sheet

hyperbola. When b2 > > a2 , f () = 0 defines a two-sheet hyperbola, as shown in Fig. 2.
For a given (x, y, z), let (, , ) be the three largest roots of equation f (s) = 0. There is a
one-to-one correspondence between (x, y, z) and (, , ), if we assume x > 0, y > 0, z > 0.
x2 =
y2 =
z2 =

(a2 + )(a2 + )(a2 + )

(a2 b2 )(a2 c2 )
(b2 + )(b2 + )(b2 + )
(b2 a2 )(b2 c2 )
2
(c + )(c2 + )(c2 + )
(c2 a2 )(c2 b2 )

(28)
(29)

The following limit applies,

> c2 > > b2 > > a2

(30)

The nine components of the Jacobian matrix can be obtained.


1/2
(a2 + )(a2 + )
1
2 (a2 + )(a2 b2 )(a2 c2 )

1/2
1
(a2 + )(a2 + )
2 (a2 + )(a2 b2 )(a2 c2 )

1/2
(a2 + )(a2 + )
1
2 (a2 + )(a2 b2 )(a2 c2 )

(31)


1/2
(b2 + )(b2 + )
1
2 (b2 + )(b2 a2 )(b2 c2 )

1/2
(b2 + )(b2 + )
1
2 (b2 + )(b2 a2 )(b2 c2 )

1/2
(b2 + )(b2 + )
1
2 (b2 + )(b2 a2 )(b2 c2 )

(32)


1/2
1
(c2 + )(c2 + )
2 (c2 + )(c2 a2 )(c2 b2 )

1/2
1
(c2 + )(c2 + )
2 (c2 + )(c2 a2 )(c2 b2 )

1/2
(c2 + )(c2 + )
1
2 (c2 + )(c2 a2 )(c2 b2 )

(33)

The (, , ) coordinate system is orthogonal because

x x y y
z z
+
+
= 0

x x y y
z z
+
+
= 0



z z
x x y y
+
+
= 0



Define function (s) (a2 + s)(b2 + s)(c2 + s). The scale factors can be expressed as
s
1 ( )( )
h =
2
()
s
1 ( )( )
h =
2
()
s
1 ( )( )
h =
2
()
The Laplacian of a scalar field (, , ) is,







1
h h
h h
h h
2 =
+
+
h h h
h

h
p
p




4 ()
4 ()
p

=
()
+
()
( )( )

( )( )

p


4 ()
p

+
()
( )( )

(34)

(35)

(36)

For derivation details see ellipsoidal coord.m.

Ellipsoidal Conductor

The discussion from here on follows Kelloggs book closely [1], with some variables renamed to follow
the notation here. Suppose we would like to solve for the potential function in space produced by
an ellipsoidal conductor that contains surface charges [1]. For a perfect conductor, the potential
on its surface (as well as the interior) is a constant. Therefore, we are trying to solve the Poissons
equation,
2 (x) = 0

(37)

subject to the boundary condition that (x) = 0 when point x is on the ellipsoidal surface,
x2 y 2 z 2
+ 2 + 2 =1
a2
b
c

(38)

and that (x) = 0 as |x| .

Introducing the ellipsoidal coordinates (, , ) as defined in the previous section. The surface
of the (original) ellipsoid is simply the isosurface of = 0. The limit of |x| corresponds to
the limit of . Therefore, when is expressed in term of the ellipsoidal coordinates, i.e.
(, , ), the boundary condition is very simple,
( = 0, , ) = 0
( , , ) = 0

(39)

Notice that the Laplacian in the Possions equation (2 = 0) in the elliptical coordinates is defined
in Eq. (36). A natural trial solution is a function () that only depends on , but not on or .
In this case,
p


4 ()
p

2
=
=0
(40)
()
( )( )

()

E
2

E
= p
2 ()
Z
E ds
p
() =
2 (s)

(41)

where E is a constant. The potential field inside the conductor is a constant and equals to the
potential on the surface ( = 0), which is,
Z
E ds
p
0 =
(42)
2 (s)
0
The surface charge of the conductor (x) can be obtained from the following relationship.
(x)
= 4(x)
n+

(43)
8

where

n+

is the gradient along the surface normal.

1 (x)
4 n+


1 ()
1
=
4 h
=0
p
Notice that at = 0, h = /()/2. Therefore,
!
p
1 2 () E
E
=
=

4
2()
4
=

On the surface of the ellipsoid, = 0,

 2

y2 z2
2 2 2 x
+ 4 + 4
= a b c
a4
b
c

(44)

(45)

(46)

The equation of the plane tangent to ellipsoid at point (x, y, z) is

(X x)

x
y
z
+ (Y y) 2 + (Z z) 2 = 0
2
a
b
c

(47)

The surface normal of the tangent plane is

r
x
y
z
x2 y 2 z 2
n=
, 2, 2 /
+ 4 + 4
2
a
b
c
a4
b
c

(48)

The shortest distance from the origin to this plane is

x2

p = qa

x2
a4

y2
b2

y2
b4

z2
c2

z2
c4

=q

1
x2
a4

y2
b4

z2
c4

abc
=

(49)

Therefore, the surface charge density is

=

E
E
q
p=
4 abc
4 abc x2
a4

1
+

y2
b4

(50)
+

z2
c4

This result is related to the problem of a uniformly charged ellipsoid. As will be shown in the
following section, the above expression of is exactly the amount of charge contained in a thin
shell between two similar ellipsoids, in the limit of shell thickness going to zero.

Ellipsoidal Shell

Consider a set of similar ellipsoids,

x2
y2
z2
x2 y 2 z 2
+
+
=
1
,
equivalently
+ 2 + 2 = u2
(au)2 (bu)2 (cu)2
a2
b
c

(51)

whose semi-axes are au, bu and cu. They are simply the original ellipsoid scaled by a factor u in
all three directions. Notice that this family of ellipsoids are different from the family of ellipsoids
defined by f () = 0 (whose shapes are not similar to each other). For 0 < u < 1, these ellipsoids
9

are smaller than the original ellipsoid (while for 0 < < the ellipsoids defined by f () = 0 are
all larger than the original ellipsoid).
Consider an ellipsoidal shell contained between two ellipsoids defined by u1 and u2 = u1 + u.
Let the volume density of the charge distribution inside the shell to be a constant . In the limit
of u 0, the shell reduces to a surface with a surface charge density . Obviously, the surface
charge density is proportional to the local thickness of the shell, h, i.e.,
= h

(52)

Let (ux, uy, uz) be a point on the surface of the ellipsoid defined by u. Let p(x, y, z, u) be the
shortest distance from the origin to the plane tangent to the ellipsoid at point (ux, uy, uz). Then,
u
p(x, y, z, u) = q
2
2
2
x
+ yb4 + zc4
a4
h(x, y, z) =

u
q

x2
a4

x2
a4

y2
b4

z2
c4

z2
c4

u
+

y2
b4

(53)

Compare this with Eq. (50), we can conclude that the surface charge of the shell is the same as the
equilibrium surface charge of a ellipsoidal conductor. The correspondence is made complete if we
set u1 = 1, u2 = 1 + u, and
E
= u
4abc
E = 4 abc u
(54)
This means that the potential field produced by this ellipsoidal shell (u1 = 1, u2 = 1 + u) is
Z
ds
p
(x) = () = 2 abc u
(55)
(s)

The potential field inside the ellipsoidal shell is a constant and equals to the potential on the surface
( = 0), which is,
Z
ds
p
0 = 2 abc u
(56)
(s)
0
In summary, Eq. (55) describes the potential generated by an ellipsoidal shell with uniform density
, whose boundary is the original ellipsoid and a similar ellipsoid scaled by a factor (1 + u). This
result can be generalized to an ellipsoidal shell between u1 = u and u2 = u + u for arbitrary u.
The potential field at a point x = (x, y, z) outside this shell is,
Z
ds
2
p
(57)
(x) = 2 abc u u
(u, s)
(u)
where (u) is the greatest root of equation f (u, s) = 0 for given (x, y, z) and u. f (u, s) and (u, s)
are generalization of the previously defined functions f (s) and (s).
x2
y2
z2
+
+
1
a2 u2 + b2 u2 + c2 u2 +
(u, s) (a2 u2 + s)(b2 u2 + s)(c2 u2 + s)
f (u, s)

u2

(58)
(59)

The factor
in Eq. (57) accounts for the fact that surface area of the scaled shell and hence its
total charge content is u2 times those of the original shell (u = 1).
10

Uniformly Charged Ellipsoid

A uniformly charged ellipsoid can be considered as a collection of many layers of ellipsoid shells
considered above. For a point x outside the ellipsoid, its potential value should be an integral of u
from 0 to 1,
Z 1 Z
ds
p
u2
Ue (x) = 2 abc
du
(60)
(u, s)
(u)
0
Define new variables v (u)/u2 and t s/u2 . Then (u, s) = u2 (t).
Z 1 Z
dt
p
u
Ue (x) = 2 abc
du
(t)
v
0
R
Perform integration by parts on du,
" Z
#1
Z 1 Z
Z
dt
u2 dt
1 1 2 1 dv
p
p
u
u p
du =
du
+
2 v
2 0
(t)
(t)
(v) du
0
v

(61)

(62)

Notice that v is the greatest root of equation,

x2
y2
z2
+
+
= u2
a2 + v b2 + v c2 + v

(63)

For u = 1, v = , while for u 0, v . Hence

#1
" Z
Z
1 dt
u2 dt
p
p
=
2 v
2
(t)
(t)

(64)

1
2

Z
0

dv
u p
du =
(v) du
2

1
2

1
=
2

u2 p

1
(v)

dv

x2
y2
z2
+
+
a2 + v b2 + v c2 + v

Therefore, the potential outside the ellipsoid is


Z 
y2
z2
x2
1
p
Ue (x) = abc
1 2
+ 2
+ 2
dv
a +v b +v c +v
(v)

1
p

(v)

dv

(65)

(66)

where is the greatest root of equation

x2
y2
z2
+
+
=1
a2 + b2 + c2 +

(67)

as previously defined.
Let us now consider the potential field at a point x = (x, y, z) inside the ellipsoid. Here we have to
cut the ellipsoid into two parts. Point x is on the outside of part 1, but is on the inside of part 2.
Let u0 correspond to the ellipsoid that pass through point x, i.e.
x2 y 2 z 2
+ 2 + 2 = u20
a2
b
c

(68)
11

Therefore, part 1 contains the ellipsoidal shells with 0 < u < u0 and part 2 contains the ellipsoidal
shells with u0 < u < 1. The potential at point x from an ellipsoidal shell in part 1 has the same
expression as the above. But potential at point x from an ellipsoidal shell in part 2 equals to
the potential value at the shell surface (because point x is inside the shell). Therefore, the total
potential at an interior point x is,
"Z
#
Z 1 Z
u0 Z
dt
dt
p
p
Ui (x) = 2 abc
u
u
du +
du
(69)
(t)
(t)
v
0
0
u0
Perform integration by parts on the first integral,
" Z
# u0
Z
Z u0 Z
1 u0 2 1 dv
u2 dt
dt
p
p
+
du =
u p
u
du
2 v
2 0
(t)
(t)
(v) du
v
0

(70)

Notice that when u = 0, v = , and when u = u0 , v = 0. Hence

Z u0 Z
Z
Z
dt
u20 dt
1 2 1
p
p
u
du =

dv
u p
2 0
(t)
(t) 2 0
(v)
0
v

Z
Z 
u20 dt
1
x2
y2
z2
1
p
p
=

+ 2
+ 2
dv
2
2 0
a +v b +v c +v
(t) 2 0
(v)
The second integral in Eq. (69) can be carried out because the inner integral is a constant.
Z

u
u0

dt
1 u20
p
du =
2
(t)

dt
p
(t)

(71)

Therefore, the total potential at an interior point x is,


Z 
y2
z2
1
x2
p
+ 2
+ 2
Ui (x) = abc
1 2
dv
a +v b +v c +v
(v)
0

(72)

Notice that the only difference between Eq. (66) and Eq. (72) is in the lower limit of integration
( versus 0). Because the range of integration is constant, the potential field inside the ellipsoid,
Ui (x), is simply a quadratic function of space.

Special Cases

9.1

Uniformly Charged Sphere

The situation of a = b = c describes a sphere with radius a. When the sphere has a uniform charge
density , the potential distribution inside the sphere is,
(x, y, z) = a3 (A B x2 B y 2 B z 2 ) = a3 (A B r2 )

(73)

where
Z

A =
Z0
B =
0

2
ds
=
3/2
a
+ s)
ds
2
= 3
5/2
2
3a
(a + s)

(74)

(a2

(75)
12

Therefore
2
(3a2 r2 )
3

(r) =

(76)

The potential value on the sphere surface is

(r = a) =

2
4 2
(3a2 a2 ) =
a
3
3

(77)

Notice that Q = 4 a3 /3 is the total charge contained in the sphere. Hence

(r = a) =

Q
a

(78)

This is equivalent to the potential produced by a point charge at origin. This confirms Newtons
Theorem, which states that the potential field outside a uniformly charged sphere is equivalent to
that produced by a point charge located at the center of the sphere. The potential field outside the
sphere (r > a) is
Q
r

(r) =

9.2

(79)

Charged Metal Disc

Consider the case of a = b and c 0. In this limit, the region inside the ellipsoid reduces to a
circle of radius a in the x-y plane (z = 0). The potential distribution inside this area is,
ZZZ
dx0 dy 0 dz 0
p
(x, y) =
(x x0 )2 + (y y 0 )2 + z 02
p
Because the integration limit for z 0 is c 1 (x0 /a)2 (y 0 /a)2 ,
s
 0  2  0 2
ZZ
0
0
dx dy
x
y
p
(x, y) = 2 c
1

(80)
0
2
0
2
a
a
02
02
2
(x x ) + (y y )
x +y a
Using the results obtained above, the potential must be a quadratic function inside the area of
radius a,
(x, y) = a2 c (A B x2 B y 2 )
Z
ds

=
A =
2 + s) s
a
(a
Z0
ds

= 3
B =
2
2
2a
(a + s) s
0

(81)
(82)

In other words, we have obtained the following relationship,

s
 0 2  0  2
ZZ
dx0 dy 0
x
y
p
1

=
a
a
(x x0 )2 + (y y 0 )2
x02 +y 02 a2
=
where r =

2
a (A B x2 B y 2 )
2
2
(2a2 r2 )
4a

(83)

x2 + y 2 . The potential value is maximum at the circumference of the circle, r = a,

(r = a) = 2 c

2 2 Q
a =
4a
3a

(84)
13

where Q = 34 a2 c is the total charge in the ellipsoid.

The results obtained here can be used to model a very thin circular metal (conductor) plate with
radius a. The total charge in the metal plate is Q, whereas the equilibrium charge distribution on
the surface is
s
 0 2  0 2
3Q
x
y
(x, y) =
1

(85)
2
2 a
a
a
The potential on the metal surface is a constant
0 =

10

Q
3a

(86)

Application to Hertz Contact Problem

For simplicity, let us consider a rigid sphere of radius R indenting an elastic half space. The
discussion here follows closely that in Landau and Lifshitz [2]. Choose the coordinate system such
that the elastic half space occupies the z 0 domain and the z = 0 plane is the surface of the half
space. The Boussinesq solution tells us about the surface displacement of the elastic half space in
response to a point force of magnitude F acting at the origin in the z direction.
uz =

F (1 2 ) 1
E
r

(87)

where E = 2(1 + ) is the Youngs modulus of the elastic half space.

The shape of the indentor can be described by function
u0 (x, y) =

x2
y2
+
2R 2R

(88)

Let d be the indentation depth. Hence inside the region of contact (S), the surface displacement
of the half space is
uz (x, y) = d + u0 (x, y) = d +

x2
y2
+
2R 2R

(89)

Let pz (x, y) be the normal stress on the surface of the half space. Inside the region of contact,
pz (x, y) < 0; outside the region of contact, pz (x, y) = 0. The total indenting force F is,
ZZ
F =
pz (x, y) dx dy
(90)
S

Using the Boussinesq solution, the surface stress pz (x, y) and the surface displacement uz (x, y) are
related by,
ZZ
pz (x0 , y 0 )
1 2
p
uz (x, y) =
dx0 dy 0
(91)
0
2
0
2
E
(x x ) + (y y )
S
Therefore, our task is to find a function pz (x, y) that satisfies the following condition,


ZZ
pz (x0 , y 0 )
E
x2
y2
0
0
p
dx dy =
d +
+
1 2
2R 2R
(x x0 )2 + (y y 0 )2
S
14

(92)

By symmetry, we expect the contact area to be a circle, and let a be its radius. Motivated by
Eq. (83), we postulate the following form for pz (x, y),
s
 0 2  0 2
y
x
pz (x, y) = p0 1

(93)
a
a
The constant p0 is related with F by
s
 0 2  0 2
ZZ
y
x
F = p0

dx dy
1
a
a
S
2 a2
3
3F
2 a2

= p0
p0 =

(94)

Plug the expression of pz (x, y) into Eq. (92), and using Eq. (83), we have


2
E
r2
p0
(2a2 r2 ) =
d
+
4a
1 2
2R


2
E
r2
3F
2
2
(2a r ) =
d +

2 a2 4 a
1 2
2R
2
2
3 (1 ) F
r
(2a2 + r2 ) = d +
3
8E a
2R

(95)

Therefore,
3 (1 2 ) F
4 E a3
1/3

3 (1 2 ) F R
a =
4E

1
R

(96)

The indentation depth is

d =


2/3 2/3
3 (1 2 ) F 2 3 (1 2 ) F
3 (1 2 )
F
2a =
=
3
8E a
4E a
4E
R1/3

(97)

Acknowledgement
I want to thank Prof. David Barnett for useful discussions and lending me Kelloggs book to read.

15

Matlab Files for Analytic Derivation

% File: elliptic_coord.m
% Purpose: analytic derivation of the properties of elliptic coordinates
syms x y a b mu nu
x = sqrt( (a^2+mu)*(a^2+nu)/(a^2-b^2) );
y = sqrt( (b^2+mu)*(b^2+nu)/(b^2-a^2) );
dxdmu
dxdnu
dydmu
dydnu

=
=
=
=

simplify(diff(x,mu));
simplify(diff(x,nu));
simplify(diff(y,mu));
simplify(diff(y,nu));

disp(dxdmu=);
disp(dxdnu=);
disp(dydmu=);
disp(dydnu=);

pretty(dxdmu);
pretty(dxdnu);
pretty(dydmu);
pretty(dydnu);

%check orthogonality
simplify(dxdmu*dxdnu + dydmu*dydnu)
h_mu = simplify( sqrt( (dxdmu)^2 + (dydmu)^2 ) );
h_nu = simplify( sqrt( (dxdnu)^2 + (dydnu)^2 ) );
disp(h_mu=); pretty(h_mu);
disp(h_nu=); pretty(h_nu);

16

% File: ellipsoidal_coord.m
% Purpose: analytic derivation of the properties of ellipsoidal coordinates
syms x y z a b c lm mu nu
x = sqrt( (a^2+lm)*(a^2+mu)*(a^2+nu)/(a^2-b^2)/(a^2-c^2) );
y = sqrt( (b^2+lm)*(b^2+mu)*(b^2+nu)/(b^2-a^2)/(b^2-c^2) );
z = sqrt( (c^2+lm)*(c^2+mu)*(c^2+nu)/(c^2-a^2)/(c^2-b^2) );
dxdlm
dxdmu
dxdnu
dydlm
dydmu
dydnu
dzdlm
dzdmu
dzdnu

=
=
=
=
=
=
=
=
=

simplify(diff(x,lm));
simplify(diff(x,mu));
simplify(diff(x,nu));
simplify(diff(y,lm));
simplify(diff(y,mu));
simplify(diff(y,nu));
simplify(diff(z,lm));
simplify(diff(z,mu));
simplify(diff(z,nu));

disp(dxdlm=);
disp(dxdmu=);
disp(dxdnu=);
disp(dydlm=);
disp(dydmu=);
disp(dydnu=);
disp(dzdlm=);
disp(dzdmu=);
disp(dzdnu=);

pretty(dxdlm);
pretty(dxdmu);
pretty(dxdnu);
pretty(dydlm);
pretty(dydmu);
pretty(dydnu);
pretty(dzdlm);
pretty(dzdmu);
pretty(dzdnu);

%check orthogonality
[ simplify(dxdlm*dxdmu + dydlm*dydmu + dzdlm*dzdmu)
simplify(dxdmu*dxdnu + dydmu*dydnu + dzdmu*dzdnu)
simplify(dxdlm*dxdnu + dydlm*dydnu + dzdlm*dzdnu)
]
h_lm = simplify( sqrt( (dxdlm)^2 + (dydlm)^2 + (dzdlm)^2 ) );
h_mu = simplify( sqrt( (dxdmu)^2 + (dydmu)^2 + (dzdmu)^2 ) );
h_nu = simplify( sqrt( (dxdnu)^2 + (dydnu)^2 + (dzdnu)^2 ) );
disp(h_lm=); pretty(h_lm);
disp(h_mu=); pretty(h_mu);
disp(h_nu=); pretty(h_nu);

17

References
[1] O. D. Kellogg, Foundations of Potential Theory, (Dover, New York, 1953).
[2] L. D. Landau and E. M. Lifshitz, Mechanics, (Permagon, New York, 1959).
[3] C. R. Weinberger, W. Cai, D.
http://micro.stanford.edu/me340b

M.

Barnett,

Micromechanics

lecture

notes,

[4] M. H. Sadd, Elasticity: Theory, Applications and Numerics, (Elsevier, New York, 2005).
[5] http://en.wikipedia.org/wiki/Elliptic coordinates
[6] http://en.wikipedia.org/wiki/Ellipsoidal coordinates
[7] http://eom.springer.de/E/e035420.htm

18