boundaryvalue problems in electrostatics iphysics.gmu.edu/~joe/phys685/topic2.pdf · boundaryvalue...

35
Boundary-Value Problems in Electrostatics I Reading:  Jackson 1.10, 2.1 through 2.10 We seek methods for solving Poisson's eqn with boundary conditions.  The mathematical techniques that we will develop have much broader utility in physics. Consider a grounded conducting plane at z = 0 and a point charge q at q z = 0 Dirichlet boundary conditions:  (z = 0) = 0 and  0 as r  in upper half-space What is for z > 0? It's not just since charge is induced on the conductor. Consider, for a moment, a completely different problem:  no conducting plane and 2 point charges: 1

Upload: dangbao

Post on 04-Jun-2018

234 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Boundary­Value Problems in Electrostatics IReading:  Jackson 1.10, 2.1 through 2.10

We seek methods for solving Poisson's eqn with boundary conditions.  The mathematical techniques that we will develop have much broader utility in physics.

Consider a grounded conducting plane at z = 0 and a point charge q at

q

z = 0

Dirichlet boundary conditions:  (z = 0) = 0 and  0 as r ∞ in upper half­space

What is for z > 0?

It's not just since charge is induced on the conductor.

Consider, for a moment, a completely different problem:  no conducting 

plane and 2 point charges:

1

Page 2: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

For this,

For z = 0,  ( x, y, z) = 0.   And,   0  as  r  ∞.

The 1st term satisfies Poisson's eqn in upper half­space:  it's the potential ofa point charge at

Explicitly:

The 2nd term satisfies Laplace's eqn in upper half­space:

= 0  in upper half­space (since does not lie in upper                half­space).

Thus, in upper half­space and  satisfies the boundaryconditions for the original problem, with the conducting plane at z = 0.

=>  This  is the unique soln to the original problem.

2

Page 3: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

This is called the “method of images”, since the “image charge” is placedat the location of the mirror image of q (for this simple geometry).

The attractive force on q,

The work done in bringing charge q from ∞ to z = z' is half the work done in bringing q and ­q to distance  2 z' :

Explicit calculation:

Suppose, instead of a single point charge, there were a charge distributionin the upper half­space.  Then there is an image charge for each part of the distribution (works for both discrete and continuous dists).

3

Page 4: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Suppose the potential on the plane z = 0 is more complicated.  For example,  = V  inside a circle of radius a centered on the origin and   = 0 outside the circle.  In this case, we can use a powerful method developed by Green.

Introduce Green functions which satisfy

Recall Green's Thm:

=>

4

Page 5: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

For the Dirichlet problem, choose such that

for ANY Dirichlet boundary conditions  [i.e., any  on the surface]!

Returning to the image problem with the conducting plane:  We have already found the Green function

5

Page 6: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Thus, the physical meaning of the Dirichlet Green function is:due to (1) a point charge at

plus (2) whatever external charges are needed to make  = 0 onthe boundary (e.g.—charges induced on the surfaces, if they are conductors). 

points AWAY from the region of interest)

6

Page 7: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

On the boundary z' = 0:

If  = V inside a circle of radius a (centered on the origin) and  = 0 otherwise, then we should adopt cylindrical coords (r, , z).

when there is no charge in the upper half­plane.  Otherwise, add

to this.

7

Page 8: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Along the circle axis, r = 0:

How to treat Neumann boundary conditions?    Recall:

It is tempting to choose

Also:

8

(unless surface area S is infinite)

Page 9: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Instead, take on surface, where S = total                surf. area of boundary

= the average of  over the surface.

If the region of interest is infinite and   0 as r  ∞, then

To find for the infinite plane z' = 0, use the “anti­image”:q

q (same sign)

Ez = 0

(the last equality holds when z' = 0)

9

for z' = 0

For the infinite plane, S = ∞, so we want to enforce  GN/n = 0  everywhere.

Page 10: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Let's check that for this example.

First, replace the upper half plane with a hemispherical volume, with radius R,and place on the z­axis.  We already know that for z' = 0, so the flat part of the hemisphere does not contribute to the surface integral.

To evaluate the derivative at r' = R, express GN in spherical coords.

This holds for all R > z  and the infinite plane corresponds to  R  ∞.

10

Page 11: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

To explicitly verify that again consider a hemisphere with radius R.

On the plane z' = 0, 

as  R  ∞.

For the curved part of the hemisphere, as  R  ∞.

Total surface area = 3R2

as  R  ∞.

11

Page 12: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

within a circle of radius a centered on the originand is zero outside.  If there is no charge in the upper half­plane, then

On the circle axis (r = 0):

Check:

Symmetry of Green functions:

Green's Thm:

12

Page 13: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

=>

For Dirichlet boundary conditions,

“Green's Reciprocity Theorem”

plus external (.e.g., induced) charge needed to satisfy boundary conditions.  Reciprocity Thm

are interchangeable!Obviously true for an isolated charge with no boundaries except at ∞.Remarkably, it remains true in the presence of conductors with fixed .

13

Page 14: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

For Neumann boundary conditions,

Redefine

So, if the original GN satisfied and

then the new GN will, too.  Thus, we can choose G

N

14

Page 15: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Note:  The freedom to add a constant to GN arises because only

is specified on the boundary.

Another classic image problem:  a charge q outside a grounded, conductingsphere of radius a

Suppose q is located at Can we find an image charge inside the spheresuch that   = 0 on the sphere's surface By symmetry, it must lieon the x­axis.  So, take an image charge q located at

15

Page 16: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Must be true for all   =>

Image charge must be inside sphere  =>  b < d  =>  2 < 1  => must choose                                                                                                                neg sign

(image charge must have opp sign as q for  = 0 on surface)

Dirichlet Green fcn:

with ; ;

16

Caution:  I'm followingJackson's notation, wherex' is mag of vector, not x'­component!

(Note:   and b can easily be recalled    with dimensional reasoning)

Page 17: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

In spherical coords (with  the polar angle):

Law of Cosines:

17

Page 18: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

(normal points away from region of interest,   i.e.,  from exterior to interior of sphere)

18

Page 19: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Soln of Laplace eqn is

Because of the complicated dependence of  cos   on  ,  ',  ,  ',  the general case does not yield analytic results.

19

Page 20: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

On the z­axis:

Suppose  = V  (const) on the sphere.

, which checks.

Symmetry  =>  z­axis is equivalent to any other axis  =>

20

Page 21: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Suppose  = +V on “Northern” hemisphere

and  = ­V  on “Southern” hemisphere

Splitting integral into 2 pieces:

(check:  = V when z = a)

Note:  Given the solns for  = const and  = V in the Northern/ Southern hemispheres, superposition yields soln for any 2 different potentials in the 2 hemispheres.

21

SP 2.1—2.3 

Page 22: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Orthogonal Functions and Expansions;  Fourier Series

This mathematical interlude is preparation for the next method of solvingelectrostatic boundary value problems—separation of variables.

Consider a set of functions  Un()   (n = 1, 2, 3, ...)

They are orthogonal on interval    (a, b)  if

* denotes complex conjugation:

Normalize so that the integral with n = m is unity.  Then the functions are orthonormal:

The orthonormal functions Un() are “complete” if any well­behaved fcn 

f () can be expressed with arbitrarily small error as a series of them.

22

Page 23: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

In this case,

The coefficients an can be found using this trick:

So,

Thus,

(“completeness relation”)

23

Page 24: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Straightforward generalization to multiple dimensions.  For example, for variables

with

Sines and cosines form a complete set of orthogonal functions.  When a fcnis expressed in terms of them, it's called a “Fourier series”.

Useful trig identities:

Consider the interval

If n ≠ m,

24

Page 25: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Similarly for integrals of

Normalization:

25

Page 26: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Thus, the Fourier series expansion for a fcn  f (x)  is:

with

The ½ is included in front of A0 so that the above formula for A

m will

also work with m = 0.  The interval (­a/2, a/2) may be shifted by anyconstant a

0:  (a

0 ­ a/2, a

0 + a/2).  e.g.:  (0, a).

Note:  The constant term is needed for completeness.  The proof ofcompleteness can be extremely difficult, depending on how well­behaved we require the fcn  f (x) to be; we will omit it.  (This is true of all thecomplete sets of orthogonal fcns that we will consider.)

26

Page 27: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

The following alternative Fourier series are also complete:

1) with

on interval (0, a).      NB:  Not orthogonal on (­a/2, a/2)!

“Fourier sine series”

2) with

on interval (0, a).      NB:  Not orthogonal on (­a/2, a/2)!

3) with

Example:

27

Page 28: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

=>  none of the cosines contribute.  As expected, sinceare even functions and  f (x) = x  is odd.

So,

28

Page 29: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

29

Page 30: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

30

Page 31: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

31

SP 2.4, 2.5

Page 32: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Separation of Variables

Some solns of the Laplace eqn are of the form

For example:

[But not all solns are of this form; e.g.,   = A (x + y + z)]

Each of the 3 terms is a fcn of just 1 variable—x, y, or z.

=>  Each term must be a constant.

Otherwise, if the sum = 0 for one value of x, then it won't for a differentvalue of x, since only the first term would have changed.  But Laplace eqnmust hold for all x.  Likewise for y and z.

32

Page 33: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

So:

The problem got vastly easier, since a PDE was replaced with 3 ODEs!The solutions are sines, cosines, and exponentials.  When the boundary surfaces are planes, a sum of “separated” solns may satisfy the boundary conditions.

There are 11 known coord systems for which the Laplace eqn is separable (i.e., for which  can be expressed as a product of 3 functions of a single coordinate with the Laplace eqn reducing to an ODE for each function).We'll discuss spherical and cylindrical coords later.

Example in Cartesian coords:  Rectangular box with lengths a, b, c inthe x, y, z directions and one corner at the origin.

x

y

z z=c

y=b

x=a

 = 0 on all the faces except z = c, where  = V (x,y).Find  everywhere inside the box.

33

Page 34: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Since  = 0 at both ends in x and y,  X and Y need to be sines; nodifference of exponentials can vanish for 2 different values of x.

=>  C1 and C

2 are negative

Similarly for Y:

Solns are

34

Page 35: BoundaryValue Problems in Electrostatics Iphysics.gmu.edu/~joe/PHYS685/Topic2.pdf · BoundaryValue Problems in Electrostatics I Reading: Jackson 1.10, 2.1 through 2.10 We seek methods

Thus, separated solns that satisfy 5 of the 6 boundary conditions have the

form

Since the Laplace eqn is linear, linear combinations of solns are alsosolns.  Can we find a linear combination of solns of the above separatedform that will satisfy the 6th boundary condition?  We require

This is a 2D Fourier sine series.  Any V (x,y) can be expanded this way.

35