0% found this document useful (0 votes)
77 views7 pages

Example Sheet 6 Solutions

This document contains solutions to example problems involving partial differential equations and Green's functions. Problem 1 shows the derivation of an identity involving the gradient operator. Problem 2 uses integration by parts and the divergence theorem to solve an integral equation. Problem 3 sets up the separation of variables method to solve the eigenvalue problem for the rectangular membrane. Problem 4 defines the 2D free-space Green's function.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
77 views7 pages

Example Sheet 6 Solutions

This document contains solutions to example problems involving partial differential equations and Green's functions. Problem 1 shows the derivation of an identity involving the gradient operator. Problem 2 uses integration by parts and the divergence theorem to solve an integral equation. Problem 3 sets up the separation of variables method to solve the eigenvalue problem for the rectangular membrane. Problem 4 defines the 2D free-space Green's function.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Example Sheet 6 Solutions

1. We need to show that


(f g) = f g + f 2 g.
By expanding the gradient operator and computing the dot product for Cartesian
coordinates in two dimensions we have
   
g g
(f g) = i +j if + jf ,
x y x y
   
g g
= f + f ,
x x y y
f g 2 g f g 2g
= +f 2 + + f 2,
x x x y y y
2
= f g + f g,
as required.
2. (i) Let
Z
I= v (pu) u (pv) dx.
D | {z }
=(pv)

Then using the expression from Qu 1, we have


Z
 
I= v p u + p2 u u p v + p2 v dx.
D

We can also write the Laplacian terms as


pv2 u = (pvu) (pv) u,
pu2 v = (puv) (pu) v.
Therefore
Z
I= vp u + (pvu) (pv) u . . .
D
up v (puv) + (pu) v dx,

then by product rule we have


Z
I=   pv 
 u + (pvu) 
vp
  u 
vp  u . . .
D
  
  v (puv) + 
up 
 v + 
up 
 v dx,
pu 
Z
= (pvu puv) dx,
ZD
= p (vu uv) n ds
D

by the divergence theorem.

1
(ii) We require L such that

hv, Lui = hL , ui

with

L = (p) + q.

This is hard directly, but note that since


Z
I= v (pu) u (pv) dx,
Z D

= v (pu) u (pv) + vqu uqv dx,


D | {z }
add zero
Z
= v (pu) u (pv) + vqu u(qv) dx,
ZD
= v ( (pu) + qu) u( (pu) + qu) dx,
D

but also from (i)


Z
I= p (vu uv) n ds.
D

Therefore
Z Z
vLu uL v dx = p (vu uv) n ds, ()
D D

where

L = (p) + q,

with

B = {(x)u + (x)n u = 0, x D} .

We now must choose B (in terms of v) such that the RHS of () is zero.
Z Z
p (vu uv) n ds = p (vn u un v) ds,
D D
Z    
u
= p v un v ds,
D
Z
p
= u (v + n v) ds,
D
Z
p 
= u v + n v ds,
D

so we choose

B = v + n v = 0, x D .

2
(iii) For L to be Self-adjoint we require p, q, , to be real functions.

3. Because the domain is rectangular, it is sensible to use Cartesian coordinates. The


eigenvalue problem is the following then: Solve
2 2
+ 2 + = 0
x2 y
subject to the BCs which for simplicity here we shall take as zero temperature (heat)
or pinned (membrane), i.e.

(x, 0) = 0, (x, H) = 0,
(L, y) = 0, (L, y) = 0

for x [0, L] and y [0, H]. This itself is a linear homogeneous BVP which suggest
we can use separation of variables (again, since we have already done this in time).
As such pose the solution

(x, y) = X(x)Y (y)

Substituting this form into the ODE and dividing by X(x)Y (y) yields the form
X (x) Y (y)
= =
X(x) Y (y)
where we choose the form of the RHS so as to give oscillatory solutions for > 0
(see sheet). We thus have two separate eigenvalue problems to solve, now in one
dimension:

X (x) + X(x) = 0, X(0) = 0, X(L) = 0,


Y (y) + ( )Y (y) = 0, Y (0) = 0, Y (H) = 0. ()

Furthermore we have already dealt with this case. The eigenvalues are
 n 2
n = , n = 1, 2, 3, ...
L

noting that for this problem zero is not an eigenvalue. The associated eigenfunctions
are
 nx 
Xn (x) = sin .
L
Each value of n gives a separate eigenvalue problem (). We therefore obtain a
infinite set of eigenvalues associated with each n . Therefore we use a double
subscript, giving
 m 2
mn n = , m = 1, 2, 3, ...
H
with associated eigenfunction
 my 
Ym (y) = sin .
H

3
Thus the two dimensional eigenfunctions are
 nx   my 
mn (x, y) = sin sin
L H
with associated eigenvalues
 m 2  n 2
mn = +
H L
with m, n = 1, 2, 3, ....
This allows us to write down the solution in the time domain in the following way.
For the heat equation, we know that we can write
X
X
u(x, y, t) = amn mn (x, y)emn t
m=1 n=1

and the constants amn would be determined by application of the initial condition
u(x, y, 0) = u0 (x, y), using orthogonality of the eigenfunctions. For the wave equation
we know that we can write
X
X
u(x, y, t) = mn (x, y)(amn ei mn t
+ bmn ei mn t
)
m=1 n=1

and the constants amn and bmn would, as in the heat equation case be determined
by applying the initial conditions in the time dependent problem.

4. For 2-D, the free-space Greens function is defined as


1
G2 (x, x0 ) = ln |x x0 |,
2
1 
= ln (x x0 )2 + (y y0 )2 .
4
For x 6= x0 we have

G2 1 (x x0 )
= ,
x 2 ((x x0 )2 + (y y0 )2 )
2
G2 1 (x x0 )2
= ,
x2 2 ((x x0 )2 + (y y0 )2 ) ((x x0 )2 + (y y0 )2 )2
G2 1 (y y0 )
= ,
y 2 ((x x0 )2 + (y y0 )2 )
2
G2 1 (y y0 )2
= ,
y 2 2 ((x x0 )2 + (y y0 )2 ) ((x x0 )2 + (y y0 )2 )2

so

1 (x x0 )2 + (y y0 )2
2x G2 = =0
((x x0 )2 + (y y0 )2 ) ((x x0 )2 + (y y0 )2 )2

4
as required.
For 3-D, the free-space Greens function is given by
1
G3 (x, x0 ) = ,
4|x x0 |
1
= p .
4 (x x0 ) + (y y0 )2 + (z z0 )2
2

Again, differentiating gives

G3 (x x0 )
= 3 ,
x 4 ((x x0 )2 + (y y0 )2 + (z z0 )2 ) 2
2 G3 1 3(x x0 )2
= 3 5 ,
x2 4 (. . .) 2 4 (. . .) 2

and similarly for y and z. Therefore summing these terms gives

2x G3 = 0

as required.

5. The integral at infinity is


Z
I= (G(x, x0 )x0 u(x0 ) u(x0 )x0 G(x0 , x)) n ds.
D

The integral is a parameterization of the curve D in x0 space. Let

x0 = x + r,
(
r(cos , sin ), (2D)
r=
r(cos sin , sin sin , cos ), (3D)

Since we expect to push the boundary out to infinity , WLOG let us take our origin
at x0 = x (i.e. work in (r, , ) coordinates) and also take D to be a circle of radius
R .
For the 2D case, n = er , ds = r d and
G G2 1
n x0 G = = = .
r r 2r
Hence
Z 2  
1 u 1
I lim ln r u r d = 0,
R 0 2 r 2r r=R

for zero contribution, hence


 
u
lim r ln r u =0
R r r=R

5
as required.
For the 3D case, n = er , ds = r 2 sin d and
G G3 1
n x0 G = = = .
r r 4r 2
Hence
Z 2 Z  
1 u 1 2

I lim u r sin d d,
R 0 0 4r r 4r 2
r=R
Z 2  
1 u
= lim r + u d = 0,
R 0 2 r r=R

for zero contribution, hence


 
u
lim r + u =0
R r r=R

as required.

6. Using method of images, try a solution of the form


A
G(x, x0 ) = G2 (x, x0 ) + ln |x x1 |,
2
for x1 = (x1 , y1 ). By applying the given boundary condition on x = 0 we have
1  A 
G x=0 = ln x20 + (y y0 )2 + ln x21 + (y y1 )2 .
4 4
Therefore choose y1 = y0 , A = 1 and x1 = x0 (we cannot have x1 = x0 since this
just eliminates the existing source x). Therefore
 
1 (x x0 )2 + (y y0 )2
G(x, x0 ) = ln .
4 (x + x0 )2 + (y y0 )2

7. Using method of images, try a solution of the form


A
G(x, x0 ) = G2 (x, x0 ) + ln |x x1 |.
2
Differentiating and applying the boundary condition at y = 0 gives

G 1 (y y0 ) A (y y1 )
= + ,
y y=0 2 |x x0 |2 2 |x x1 |2 y=0
y0 Ay1
= .
2 ((x x0 ) + y0 ) 2 ((x x1 )2 + y12)
2 2

We require x1 = x0 , A = 1 and y1 = y0 (again cannot have y1 = y0 since this


eliminates the existing source). Therefore
1   
G(x, x0 ) = ln (x x0 )2 + (y y0 )2 (x x0 )2 + (y + y0 )2 .
4

6
8. By symmetry, we expect to require 3 image sources at

x1 = (x0 , y0 ), x2 = (x0 , y0 ), x3 = (x0 , y0 ).

Therefore, G will be of the form


1  A1 
G(x, x0 ) = ln (x x0 )2 + (y y0 )2 + ln (x x0 )2 + (y + y0 )2 . . .
4 4
A2  A3 
+ ln (x + x0 ) + (y + y0 )2 +
2
ln (x + x0 )2 + (y y0 )2 .
4 4
Applying the boundary condition G = 0 at x = 0 gives
1    
ln x20 + (y y0 )2 + A3 ln x20 + (y y0 )2 + (A1 + A2 ) ln x20 + (y + y0 )2 = 0.
4
Therefore A3 = 1 and A2 = A1 . Applying the boundary condition at y = 0 gives
1   
ln (x x0 )2 + y02 ln (x + x0 )2 + y02 . . .
4  
+A1 ln (x x0 )2 + y02 A1 ln (x + x0 )2 + y02 = 0.

Therefore A1 = 1 and
 
1 [(x x0 )2 + (y y0 )2 ] [(x + x0 )2 + (y + y0 )2 ]
G(x, x0 ) = ln .
4 [(x x0 )2 + (y + y0 )2 ] [(x + x0 )2 + (y y0 )2 ]

You might also like