PARTIAL DIFFERENTIAL EQUATIONS

1
1. Introduction (partial differential equation)

ex 1) Laplace equation
2 2 2
 u  x, y , z   0  2 u  2 u  2 u  0
2

x y z
: gravitational potential, electrostatic potential, steady-state
temperature with no source

ex 2) Poisson’s equation: u  f  x, y, z 
2

: with sources (=f(x,y,z))

ex 3) Diffusion or heat flow equation

1 
 2u  x , y , z , t   u  x, y, z, t  ( : diffusivity)
 2 t

2
 1  2u
ex 4) Wave equation  u 2 2
2

v t

ex 5) Helmholtz equation 2F  k 2F  0

: space part of the solution of either the diffusion or the wave equation

3
 2. Laplace’s equation: steady-state temperature in a rectangular
plate (2D)
In case of no heat source

 2T  2T
 T  0 or
2
 2 0
x 2
y

To solve this equation, we try a solution of the form,

 X ( x) : a function only of x,
T  x, y   X  x  Y  y  , 
 Y ( y ) : a function only of y.


"Separation of variables"

4
  2T  2T  2 XY  2 XY
 0  0
x 2 y 2 x 2 y 2
d2X d 2Y  XY 1 d 2 X 1 d 2Y 1 d2X 1 d 2Y
Y 2
 X 2  0  2
 2
 0, 2

dx dy X dx Y dy X dx Y dy 2

To satisfy the above equation with the variable - separated sides,

1 d2X 1 d 2Y
   const .   k 2
(separation constant k  0)
X dx 2 Y dy 2
 X    k 2 X and Y   k 2Y .
e ky sin kx

sin kx, e ky , e  ky sin kx
X  Y   T  XY  
cos kx, e ,
 ky ky
e cos kx
  ky
e cos kx

5
 e ky sin kx

e  ky sin kx
General solution : T  XY  
ky
e cos kx
  ky
e cos kx

1) In the current problem, boundary conditions are

i ) T  0 as y  .  discard e ky
ii ) T  0 when x  0.  discard cos kx
iii ) T  0 when x  10. sin 10k  0  k  n / 10
iv ) T  10 when y  0.

nx 
nx
T  exp[ny / 10] sin  T   bn sin  100.
10 n 1 10

6
 nx 
nx
T  exp[ny / 10] sin  T   bn sin  100.
10 n 1 10

Using the Fourier series,

10
2 l nx 2 10 nx 10  nx 
bn   f ( x) sin dx   100 sin dx  20    cos 
l 0 l 10 0 10 n  10  0

 400

200
n
  
  1 n  1   n
, odd n,
 0, even n.

400  x 1  3y / 10 3x 
T  exp[ny / 10] sin  e sin  .
  10 3 10 

ex. At x  5 and y  5, T  26.1

7
 2) How about changing the boundary condition? Let us consider a finite
plate of height 30 cm with the top edge at T=0.
T=0 at 30 cm

In this case, e ky can not be discarded.

e  ky  ae ky  be ky
T  0 at y  30  12 e k  30  y   12 e  k  30  y  ( that is, a  12 e30k , b   12 e  30k )  sinh k  30  y 
n
 30  y  sin nx

T   Bn sinh
n 1 10 10

nx  nx
Ty  0  100   Bn sinh 3n sin   bn sin , where bn  Bn sinh 3n .
n 1 10 n 1 10
n
T  
 400 
 
1
 sinh  30  y   sin nx
odd n  n  sinh 3n 10 10

8
 - To be considered I
e kx sin ky
  kx
e sin ky
We could have another general solution, T  XY   , for  k 2 .
kx
e cos ky
  kx
e cos ky

This is correct, but makes the problem more complicated.

(Please check the boundary condition.)

- To be considered II
n case that the two adjacent sides are held at 100 (ex. C=D=100),
he solution can be the combination of C=100 solutions (A, B, D: 0)
and D=100 (A, B, C: 0) solutions.

9
. Summary of separation of variables.

) A solution is a product of functions of the independent variables.

) Separate partial equation into several independent ordinary equation.
) Solve the ordinary differential eq.
) Linear combination of these basic solutions
) Boundary condition (boundary value problem)

10
3. Diffusion or heat flow equation; heat flow in a bar or slab

1 u
- Heat flow :  2u  .
 2 t
u  F  x, y , z  T  t  "Separation of variables"
1 dT 1 2 1 1 dT
T 2 F  F   F   k 2 .
 2
dt F  T dt
2

1 2 1 1 dT
 F  k 2 ,   k 2

F  2 T dt

  2 F  k 2 F ,
dT
dt
 2 2
 k 2 2T T  e  k  t 

cf. “Why do we need to choose –k^2, not +k^2?”

11
Let’s take a look at one example.

At t=0, T=0 for x=0 and T=100 x=l.

From t=0 on, T=0 for x=l.

In case of this 1 - D problem,

1 2 d 2F  sin kx, 
 F  k  2  k 2 F  0  F  
2


F dx  cos kx 
 e  k  t sin kx
2 2

u   2 2
 k  t
e cos kx

For T(x=0)=0 and T(x=l)=100 at t=0, the initial steady-state

temperature distribution:
d 2u 0
 u0  0 (no heat source) 
2
2
0
dx
100
u0  ax  b  u0  x
l 12
 1) u  0 when x  0.  discard cos kx
- Using Boundary condition
2)u  0 when x  l. sin kl  0  kl  n

nx 
nx
  n / l  2 t
 u   bn e   n / l  t sin
2
ue sin
l n 1 l
At t  0, u  u0

nx 100
u   bn sin  u0  x.
n 1 l l

  1 n 1  200   1
n 1
100 2l 1
bn 
l  n  n
200     / l  2 t x 1   2 / l  2 t 2x 1   3 / l  2 t 3x 
u e sin  e sin  e sin  .
  l 2 l 3 l

13
or some variation, when T0,
we need to consider uf.as the final state, maybe a linear functio
this case, we can write down the solution simply like this.

nx
u   bn e   n / l  t sin
2
 uf .
n 1 l

nx 
nx
u0   bn sin  u f  u0  u f   bn sin .
n 1 l n 1 l

14
 4. Wave equation; vibrating string
node
Under the assumption that the string is not stretched,

2 y 1 2 y

x 2 v 2 t 2
x=0 x=l
y  X  x T  t 

1 d2X 1 1 d 2T
2
  2 2
 const.   k 2
 X   k 2
X  0 and T   k v T  0.
2 2

X dx v T dt

  frequency (sec-1 ),   2  angular frequency (radians)

2 2 
  wavelength k    wave number.
 v v

15
 sin kx sin t

sin kx, sin kvt, sin kx cos t
X  T   y  XT   , where   kv.
cos kx, cos kvt, cos kx sin t
cos kx cos t


Boundary condition : y  0 at x  0, x  l.
 nx nvt
 sin sin
l l
y
sin nx cos nvt .
 l l

16
  nx nvt  nx nvt
 sin sin  sin cos
l l l l
y  y  
sin nx cos nvt sin nx sin nvt
 l l  l l

1) case 1

For y (t  0)  f ( x) and y '  0, sin  nvt / l  should be discarded.


nx nvt
 y   bb sin cos .
n 1 l l

nx
y0   bn sin  f  x .
n 1 l
After finding the Fourier series coefficients,
8h  x vt 1 3x 3vt 
y  sin cos  sin cos   
2  l l 9
l l 

17
  nx nvt
 sin sin
l l
y
sin nx cos nvt .
 l l

2) case 2
For y (t  0)  0 and y0 '  0, cos nvt / l  should be discarded.

nx nvt
 y   Bn sin sin .
n 1 l l
 y  
nv nx  nx
    Bn sin   bn sin  V  x  (Use Fourier series expansion.)
 t t  0 n 1 l l n 1 l

18
3) Eigenfunctions
nx nvt
y  sin sin : a characteristic function or eigenfunction
l l
n  nv / l  2 n : characteristic frequency
A vibration with a pure freuquencyis called the normal mode of vibration.

first harmonic, fundamental second harmonic

third fourth

19
Mathematical methods in the physical sciences 2nd edition Mary L. Boas

Chapter 13 Partial differential equations

Lecture 14 Using Bessel equation

20
5. Steady-state temperature in a
cylinder

For this problem, cylindrical coordinate (r, , z) is more useful

1   u  1  2u  2u
 u
2
r   2  2 0
r r  r  r  2
z

u  R  r     Z  z 
Putting this into the equation, and dividing with u  RZ ,
1 1 d  dR  1 1 d 2 1 d 2 Z
r   0
R r dr  dr   r d
2 2
Z dz 2

21
 1 1 d  dR  1 1 d 2 1 d 2 Z
r   0
R r dr  dr   r d
2 2
Z dz 2

In order to say that a term is constant,

1) function of only one variable
2) variable does not elsewhere in the equation.

- 1st step

1 1 d  dR  1 1 d 2 1 d 2Z
r    k 2
R r dr  dr   r d
2 2
Z dz 2

1 d 2Z  e kz

 k 2, Z   .
Z dz 2  e  kz

Because u ~ 0 at z  , u ~ e  kz .

22
- 2nd step
1 1 d  dR  1 1 d 2 r 2 r d  dR  1 d 2
  k  0    r 2k 2  0
2
r r
R r dr  dr   r d
2 2
R dr  dr   d 2

Here,
1 d 2  sin n ,
 n ,   
2

 d 2
 cos n .

- 3rd step
r d  dR  2
r
R dr  dr 
  n  k r  0  r r
2 2 d  dR 
dr  dr 
2 2 2
 2 2 2 2

  k r  n R  0  r R  rR  k r  n R  0 

  
" Bessel equation": x xy  x 2  p 2 y  x 2 y  xy  x 2  p 2 y  0
 Solution of the Bessel equations : J n  kr  (Bessel), N n  kr  (Neumann or Weber) .

Here, we take only R ~ J n (kr ), because N n  kr    at r  0.

- Boundary condition : Rr 1  J n  k   0.
 J n  kr  sin ne  kr
u  RZ   , where k is a zero of J n .
 J  kr  cos ne  kr 23
 n
In azimuthal symmetry, n  0  J 0  k m   0 ( J n : oscillating function)

 u   cm J 0  k m r  e  k m z .
n 1


Boundary condition : u z  0   cm J 0  km r   100.
n 1

Here, using the orthogonality of the Bessel function,


rJ 0  k  r   cm J 0  km r dr   100  rJ 0  k  r  dr
1 1
 0
n 1
0

Only if   m, we have the non - zero term in the left side.

 cm  r  J 0  km r   dr   100rJ 0  km r  dr.
1 2 1

0 0

left term :  r  J 0  km r   dr  12 J12  km .

1 2
0

J p ( a)  0  J p (b)

 0, if a  b
0 xJ p  ax  J p  bx  dx   1 J p2 1  a   1 J p2 1  a   1 J p21  a  , if a  b
1

2 2 2
24
cm  r  J 0  km r   dr   100rJ 0  km r  dr.
1 2 1

0 0

d
For the right term :  xJ1  x    xJ 0  x  (recusion relation)
dx
1 d
 For x  km r ,  km rJ1  km r    km rJ 0  km r  ,
km dr
1 d
 rJ 0  km r    rJ1  km r  
km dr
1
1 1
    J1  k m .
1
0 0 m
rJ k r dr 
km
rJ 1 mk r 
k m
0

Combining two results,

100 J1  km 
0
2 1

0

cm  r  J 0  km r   dr   100rJ 0  km r  dr  cm 12 J12  km  
1
 km
100 J1  km  2 200
 cm   2 
km J1  k m  k m J1  k m 

25
If the given temperature of the base of the cylinderis f  r ,  ,   1, but an cos n  bn sin n .
 
 u   J n  kmnr  amn cos n  bmn sin n  e  k mn z .
m 1 n  0
 
u z  0   J n  kmnr  amn cos n  bmn sin n   f  r , .
m 1 n  0

Because of the orthogonality,

f  r ,  J  k  r  cosrdrd  a  J  k  r  cos2 rdrd

1 2 1 2
 
2

0 0 0 0

 a  12 J21  k     .

f  r ,  J  k  r  sinrdrd .
2 1 2
Similarly, b 
J 2
 1 k   

0 0

26
- Bessel’s equation Bessel’s equation 1

1) Equation and solution

- named equation which have been studied extensively.
- “Bessel function”: solution of a special differential equation.

   
x xy  x 2  p 2 y  x 2 y  xy  x 2  p 2 y  0

- being something like damped sines and cosines.

- many applications.
ex) problems involving cylindrical symmetry (cf. cylinder function); motion
of pendulum whose length increases steadily; small oscillations of a flexible
chain; railway transition curves; stability of a vertical wire or beam;
Fresnel integral in optics; current distribution in a conductor; Fourier series
for the arc of a circle.

27
 Bessel’s equation 2


  
x xy  x 2  p 2 y  x 2 y  xy  x 2  p 2 y  0

- Solution :

J p  x  
  1
n
x
2n  p

 

e dx   p  1!,

p 1  x
  , where  p  x p  0.
n  0   n  1   n  p  1  2 
0

cos p  J p  x   J  p  x 
N p  x  (Neumann or Weber)
sin p
 general solution y  AJ p  x   BN p  x 

- Graph

28
 Bessel’s equation 3
2) Recursion relations

1)
d p
dx
 
x J p  x   x p J p 1  x 

2)
d p
dx
 
x J p  x    x  p J p 1  x 

2p
3) J p 1  x   J p 1  x   J p  x
x
4) J p 1  x   J p 1  x   2 J p  x 
p p
5) J p  x    J p  x   J p 1  x   J p  x   J p 1  x 
x x

29
 Bessel’s equation 4
3) Orthogonality

cf. Comparison
sin x, cos x J p  x, N p  x
consider just sin x consider just J p  x  for one value of p
x  nx, sin x  0 x  a, b,, J p  x   0
x  1, sin nx  0 x  1, J p  ax   0, J p  bx   0,
For y  sin nx, y   n  y  0
2 
 
For J p  ax  , x xy  a 2 x 2  p 2 y  0

 xJ  ax  J  bx  dx  0,
1 1
 sin nx sin mxdx  0,
0
for n  m,
0
p p for a  b

30
 Bessel’s equation 5


  
x xy  x 2  p 2 y  x 2 y  xy  x 2  p 2 y  0 


 x xy  a 2 x 2  p 2 y  0 
J p (a)  0  J p (b)
 0, if a  b
0 xJ p  ax  J p  bx  dx   12 J p21  a   12 J p21  a   12 J p21  a  , if a  b
1

31
6. Vibration of a circular membrane (just like drum)
1 2 z
 z 2 2
2

v t
z  F  x, y T  t    2 F  k 2 F  0 and T  k 2v 2T  0
1   F  1  2 F
In polar coordinate, r  2  k 2 F  0.
r r  r  r  2

F  R r   . Putting this form into the above equation, and then dividing with F ,
1 1 d  dR  1 1 d 2
r  2  k2  0
r R dr  dr  r  d 2

1 d 2 d 2
  n 2
  n 2
  0 (simple harmonic equation)
 d 2
d 2

r
d  dR 
r
dr  dr 
2 2 2

   n  k r R  0 (Bessel's equation)

sin n sin kvt


 sin n  sin kvt  sin n cos kvt
 z  R r   T  t   J n  kr      J n  kr  
 cos n  cos kvt  cos n sin kvt
cos n cos kvt
 32
 sin n sin kvt

sin n cos kvt
z  R r   T  t   J n  kr  
cos n sin kvt
cos n cos kvt


Boundary condition : z (r  1)  0  J n  k   0.
 kmn : zeros of J n (double series)

 Characteristic freuqency of the normal modes :  mn  kmnv /  2 

cf. They are not integral multiples of the fundamental as is true for the
string (characteristics of the bessel function). This is why a drum is less
musical than a violin.

33
ex) z  J n  k mn r  cos n cos k mn vt

34
(m,n)=(1,0) (2,0) (1,1)

American Journal of Physics, 35, 1029 (1967)

35
Mathematical methods in the physical sciences 2nd edition Mary L. Boas

Chapter 13 Partial differential equations

Lecture 15 Using Legendre equation

36
 7. Steady-state temperature in a
sphere
Sphere of radius 1 where the surface of upper half is 100, the other is 0 degree.

1   2 u  1   u  1  2u
 u  2 r  2  sin   2 2 0
2

r r  r  r sin      r sin   2

1 d  2 dR  1 1 d  d  1 1 d 2 
For u  R r        , r   sin   0
R dr  dr   sin  d  d   sin 2  d 2
1 d 2 sin m
 m ,   
2

 d 2 cos m
1 d  2 dR  1 1 d  d  m2
r   sin    0.
R dr  dr   sin  d  d  sin 2 

1 d  2 dR  r l
r k
 R  
R dr  dr  1 / r l 1
1 d  d  m2
 sin    2   k  0 (Associated Legendre equation k  l  l  1 )
sin  d  d  sin 
   Pl m  cos  (Legendre polynomial).

37
1) For the interior of the sphere, no divergence. Therefore, discard 1 / r l 1.
sin m
u  r Pl  cos  
l m

cos m

 u   cl r l Pl  cos 
2) Due to the azimuthal symmetry, m  0 
l 0

 100, 0     / 2
3) ur 1   cl Pl  cos   
l 0 0,  /2  
 0,  1  x  0,
or ur 1   cl Pl  cos   100 f  x  , where f  x   
l 0 1, 0  x  1.
In this case, we can expand this in a series of Legendre polynominals
f  x   12 P0  x   34 P1  x   167 P3  x   11 P  x    (Here, x  cos )
32 5


 c P  cos   100
l 0
l l
1
2
P0  cos   34 P1  cos   167 P3  cos   11 P  cos   
32 5

1 3 7
c0  100  , c1  100  , c2  100   ,.
2 4 16

 u  x   100 12 P0  cos   34 rP1  cos   167 r 3 P3  cos   11
32

r 5 P5  cos   
38
 Legendre’s equation 1
- Legendre’s equation
1) Equation and solution
i. .m  0 (Legendre equation)
l

 
1  x 2 y  2 xy  l  l  1 y  0  Solution : Pl  x  
1 d  2 l
  x 1 .
2!l!  dx 
 

ii. m  0 (Associated Legendre equation)

 m2 
 2

1  x y  2 xy  l  l  1 
1  x 2
y0
 
m

 Solution : Pl  x   1  x
m
 
2 m/2 d 
  Pl  x    l  m  l .
 dx 
 For m  l , Pl m  0 
1 d  d  m2
cf .  sin     l  l  1   0 (Associated Legendre equation).
sin  d  d  sin 2 

39
 Legendre’s equation 2

1 d  d  m2
cf .  sin    l  l  1   0 (Associated Legendre equation)
sin  d  d  sin 2 
1 d d cos  d d cos  m2
 sin     l  l  1   0
sin  d cos x d  d cos d  sin 2 
d  2 d  m2
 sin     l  l  1   0
d cos x  d cos  sin 
2

d 
d 

 1  cos 
d cos x 
2
 
m2
d cos  1  cos2 
  l  l  1   0

d 2 d

1  cos  2

d cos x 2

 2 cos 
m2
d cos 1  cos 2 
  l  l  1   0

d 2 d

1 x 2

dx 2
 2x 
m2
dx 1  x 2
  l  l  1   0

40
 Legendre’s equation 3

- Legendre polynomials - Associated Legendre polynomials

P0  x   1 Pl  m  x     1
m  l  m ! P m  x 
 l  m ! l
P1  x   x
1
P11  x    P11  x 
P2  x  
1 2
2

3x  1  2
P10  x   x
1
2

P3  x   5 x 3  3x  
P11  x    1  x 2  1/ 2

1 2
1

P4  x   35 x 4  30 x 2  3
8
 P2 2  x  
24
P2  x 

1
P21  x    P21  x 
1

P5  x   63x 5  70 x 3  15 x
8
 6

P6  x  
1

231x 6  315x 4  105x 2  5 
1
P20  x   3x 2  1
2
 
16
P21  x   3x 1  x 2  1/ 2

1

P22  x   3 1  x 2 
cf . x   3P1  x   2 P3  x  
3

5
41
 Legendre’s equation 4
2) Orthogonality

2
   

   
1
1 0  sin d 
2 2
Pl x dx  Pl cos .
2l  1
2  l  m !
 P  x  2 dx  0  

Pl m  cos  sin d 
1 2
m
.
2l  1  l  m !
l
1

cf .x  cos  dx  sin d

42
8. Poisson’s equation
  F  0 (F : conservative)  F  V
1) For gravitational force,   F  4G   2V  4G : Poisson equation
cf.   0   2V  0 : Laplace equation
2) For electrostatic force,   E  4   2V  4 (in Gaussian unit).
cf.   0   2V  0

In general,  2u  x, y, z   f  x, y, z 
1 f  x, y, z
u
4     x  x    y  y     z  z  
2 2 2
dxdydz.

 2 w  0 (Laplace equation)
 2  u  w   2u   2 w  f  u  w : solution of Poisson equation

43
Example 1
 2V  4
1  4  x, y, z q
V  x, y , z        x  x    y  y     z  z   dxdydz  .
4 x  y   z  a
2 2 2 2 2 2

In spherical coordinate,
q
Vq  .
r  2ar cos  a
2 2

r l  m sin m 
Basic solution of Laplace equation   Pl  cos     r  l 1Pl  cos 
r  l 1  cos m 
1) zero at infinite  discard r l 
 
 2) symmetric about z - axis  solution is indep. of  .  m  0, cos m  1
 

V  Vq   cl r  l 1Pl  cos 
l
grounded sphere

44
Boundary condition : V  0 for r  R.
q
Vr  R    cl R  l 1Pl  cos   0.
R 2  2aR cos  a 2 l

q R l Pl  cos 
Using  q l 1
,
R  2aR cos  a
2 2
l a
R l Pl  cos 
q l 1
  cl R  l 1Pl  cos 
l a l

 l 1 qR l qR 2l 1
From this relation, cl R   l 1 or cl   l 1 .
a a
q R 2l 1r  l 1Pl  cos 
V   q
r  2ar cos  a
2 2
l a l 1


q
 q( R / a )
 R / a  P  cos 
2 l
l
‘Method of the images’
l 1
r  2ar cos  a
2 2
l r


q

 R / a q .
r  2ar cos  a
2 2
  2
r 2  R 2 / a  2r R 2 / a cos 
In the second term, charge :  R / a  q, position : 0,0, R 2 / a  
45
 cf. Electric multipoles
monopole dipole quadrupole octopole

2) Expansion for the potential of an arbitrary localized charge distribution

1 1
V  r     r dr 
4 0 r  r
  r  
2
 r  
r  r  r   r   2rr cos  r 1     2  cos 
2
 2
  2
 2

  r  r 

 r   r  
r  r  r 1   , where      2 cos   
 r  r 
1 1 1 1 3 5 
 1   
1/ 2
 1     2   3   
r  r r r 2 8 16 
46
 1  1  r   r  
2 2 3 3
1  3  r    r   5  r    r  
 1     2 cos         2 cos         2 cos     
r r  r  2  r  r  8 r   r  16  r   r  

1   r  
2 3
r  

  r 
 
 1    cos     3 cos   1 / 2    5 cos   3 cos / 2  
r   r 
 2
 3
  
r r 
n
1   r 
    Pn  cos  Legendre polynominal
r n 0  r 


1 1
V r   r n 1 
   Pn  cos    r d   multipole expansion 
n
r
4 0 n 0

- n = 0 : monopole contribution
- n = 1 : dipole
- n = 2 : quadrupole
- n = 3 : octopole

47
Sekian Semoga
Dipahami