s39 57 PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

CVD EXERCISES

34

Solving PDEs by Separation of Variables


Part hand-in for
Week 8

39 (E) For each of the PDEs and boundary/initial conditions in the


problems in Q.5, determine how many are homogeneous and how
many are inhomogeneous.
[Ans: (a) 4, 1; (b) 4, 1]
SOLUTION. The inhomogenous ones are (a) u/t = cos x, (b)
u(x, ) = sin x sinh .

Parts (a) and (c)


are part hand-in
for Week 8

40 (E) To which of the following differential equations does the Principle of Superposition apply? Where it does apply, prove it. Where
it does not apply, find a solution u1 such that 2u1 is not a solution.
(Hint: in such cases try a linear function of x, t.)
(a) ut = uxx
(b) ut = (xux )x
(c) ut = uxx + 1
(d) ut = (uux )x
[Ans: (a) yes,(b) yes,(c) u1 = t,(d) u1,2 = t x]
SOLUTION.
(a) If uit = uixx for i = 1, 2, and C1,2 are constants, let u =
C1 u1 + C2 u2 . Then
ut = C1 u1t + C2 u2t
= C1 u1xx + C2 u2xx
= uxx
(b) Like (a)
(c) Though the equation is linear, it is not satisfied by u = 0. Try
u = ax + bt + c. This is a solution b = 1. u = 2t is not
a solution.
(d) The equation is clearly nonlinear. ux = a, and we have a
solution b = a2 . u = t + x is a solution, but u = 2t + 2x
is not.
41 (E) Find all separable solutions of
(a) ux + uy = 0
(b) ux + uy = u
(c) uxx + uyy = 0
(d) uxx + uyy = u
(Assume the separation constant is real.)
[Ans: (a) C exp((xy)); (b) C exp(y+(xy)); (c) (A cosh px+
B sinh px)(C cos py+D sin py), (A cos px+B sin px)(C cosh py+D sinh py), (A+
bx)(C + Dy) where p > 0; (d) (A cos px + B sin px)(C cosh qy +
D sinh qy), q 2 = 1 + p2 , p > 0, (A + Bx)(C cosh y + D sinh y),
(A cosh px+B sinh px)(C cosh qy+D sinh qy), p2 +q 2 = 1, 0 < p < 1,
(A cosh x + B sinh x)(C + Dy), (A cosh px + B sinh px)(C cos qy +
D sin qy), p2 q 2 = 1, p > 1]

CVD EXERCISES

35

SOLUTION.
(a)
X 0 Y + XY 0 = 0
Y0
X0
= .

X
Y
Since the left side is independent of y and the right side is
independent of x, both sides must be constant, say . Hence
u = X(x)Y (y)

X 0 = X
Y 0 = Y

and so X = C1 exp(x), Y = C2 exp(y), and so u = C1 C2 exp((x


y)). Since C1,2 are arbitrary constants, they can be combined
into one.
(b) Now
X 0 Y + XY 0 = XY
Y0
X0
= + 1.

X
Y
By the same argument, we get
u = X(x)Y (y)

X 0 = X
Y 0 = (1 )Y

(c)

Hence X = C1 exp(x), Y = C2 exp((1 )y).


X 00 Y + XY 00 = 0
Y 00
X 00
= .

X
Y
By the usual argument we get
u = X(x)Y (y)

X 00 = X
Y 00 = Y.

(i) If < 0 let = p2 . Then

X = A cos px + B sin px
Y = C cosh py + D sinh py.

(ii) If > 0, let = p2 and proceed similarly.


(iii) If = 0 then the result is obvious.
(d)
X 00 Y + XY 00 = XY
X 00
Y 00

=
+ 1.
X
Y
By the usual argument we get
u = X(x)Y (y)

X 00 = X
Y 00 = (1 )Y.

CVD EXERCISES

36

(i) If < 0, let = p2 and q =


X 00 = p2 X
Y 00 = q 2 Y.

p
1 + p2 . Then

(ii) If = 0 we get
X 00 = 0
Y 00 = Y.
(iii) If 0 < < 1, put = p2 , q =
and so on.

p
1 p2 .

42 (P)
(a) Find a separable solution of the Hamilton-Jacobi equation
for the simple harmonic oscillator
 2
u
1 u
1
= 0,
+ x2 +
2 x
2
t

in the sense that u(x, t) = X(x) + T (t).


(b) Try to use separation of variables (in the more usual sense of
u(x,y) = X(x)Y(y)) on the biharmonic equation 2 (2 u) = 0.
1
x 1
1
[Ans.(a) a2 sin1 + x a2 x2 + C a2 t]
2
a 2
2

SOLUTION.
(a) We get
1 02 1 2
X + x + T0 = 0
2
2
1 02 1 2
X + x = T 0 .
2
2
The left side is independent of t and the right side is independent of x, and so both are constant, say a2 /2. (Note that the
left side is positive.) Hence T = C a2 t/2. Also

X0 =
a2 x2 (if we select the positive square root)
Z
X =
a2 x2 dx
Z
2
= a
cos2 d (where x = a sin )
Z
a2
(1 + cos 2)d
=
2


a2
1
=
+ sin 2
2
2
2
a
=
( + sin cos )
2
!
r
2
x
x
x
a2
sin1 +
1 2 .
=
2
a a
a

CVD EXERCISES

37

(b) We get 2 u = X 00 Y + XY 00 . Hence 2 (2 u) = X iv Y +


2X 00 Y 00 + XY iv = 0. Division by XY makes the first and last
terms independent of one variable, but not the middle term.
There appears to be no way to proceed along this route.
43 (R) Find all eigenfunctions f of d2 /dx2 on the interval [0, ] with
the boundary conditions
(a) f (0) = 0, f 0 () = 0;
(b) f 0 (0) = 0, f () = 0;
(c) f 0 (0) = 0, f 0 () = 0;
[Ans: sin(n + 1/2)x, n = 0, 1, 2, . . .; cos(n + 1/2)x, n = 0, 1, 2, . . .;
cos nx, n = 0, 1, 2, . . ..]
SOLUTION.
(a) Seek f (x) such that f 00 = f and f (0) = f 0 () = 0.
(i) If > 0 set = p2 , p > 0. Then f = A cos px + B sin px.
Since f (0) = 0, A = 0, and so f = B sin px. Hence
f 0 = Bp cos px and we require Bp cos p = 0. Since
p > 0 options are (i) B = 0, giving trivial solution, or
(ii) cos p = 0, and so p = n + 1/2 (n a non-negative
integer).
(ii) If < 0 set = p2 , p > 0. Then f = A cosh px +
B sinh px. Since f (0) = 0, A = 0, and so f = B sinh px.
Hence f 0 = Bp cosh px and we require Bp cosh p = 0.
Since p cosh p > 0 the only possibility is B = 0, giving
trivial solution.
(iii) If = 0 then f = A + Bx and the boundary conditions
give A = B = 0 (trivial solution).
(b) similar
(c) The first two cases (for the sign of ) are similar. If = 0 we
have f = A + Bx, and f 0 = 0 at 0 and if and only if B = 0.
Hence f = A, corresponding to eigenfunction 1 = cos 0x.
44 (S) For each of the three sets of boundary conditions in the previous question, show that the eigenfunctions are orthogonal on the
interval [0, ].

SOLUTION.
(a) If n 6= m then

Z
1
1
1
sin(n + )x sin(m + )xdx =
(cos(n m)x cos(m + n + 1)x)dx
2
2
2 0


1 sin(n m)x sin(n + m + 1)x
=

2
nm
n+m+1
0
= 0
1
(b) Similar, using cos A cos B = (cos(A + B) + cos(A B)).
2
(c) Similar

CVD EXERCISES

38

45 (R) Find the half-range Fourier sine and cosine series of each of the
following functions defined on [0, ]:
(a) f (x) = 1
(b) f (x) = x
(c) f (x) = 1+ x
1 if x < /2
(d) f (x) =
1
otherwise

X
4
sin(2m + 1)x
[Ans: (a)
, 1;
0
2m + 1

4 X cos(2m + 1)x
(1)n
(b) 2
sin nx,
;
2
n
2

(2m
+
1)
0
1

2 X 1 (1 + )(1)n

4 X cos(2m + 1)x
(c)
sin nx, 1 +
;
1
n
2 0
(2m + 1)2

4 X (1)m cos(2m + 1)x


4 X sin(4m + 2)x
,
.]
(d)
0
2m + 1
0
2m + 1

Z
2
f (x) sin nxdx.
SOLUTION. Sine: f (x) =
bn sin nx, bn =
0
1
Z

a0 X
2
Cosine: f (x) =
+
an cos nx, an =
f (x) cos nxdx.
2
0
1 (
0
if n even
2 1
4
(a) bn = [ cos nx]0 =
if n odd, i.e. n = 2m + 1, m 0
n
n
No calculation
required forZthe cosine series.


2
x
1

(b) bn =
[ cos nx]0 +
cos nxdx (integration by parts)

n
n 0
1
2
2 x
= [ cos nx + 2 sin nx]0 = cos n. If n > 0, an =
n
n
n
2 x
1
2

[ sin nx + 2 cos nx]0 =


((1)n 1), a0 = .
n
n
n2
(c) Add theresults for (a) and (b).

2 1
1
4
n
n
/2

(d) bn =
[ cos nx]0 [ cos nx]/2 =
cos
(1 cos
)=0
n
n
n
2
2
unless n is twice an odd non-negative integer, i.e. n = 4m +
2, m = 0, 1, 2, .... Or else draw the odd 2-periodic extension of f , and compare with (a). a0 = 0; if n > 0, an =
o
n
4
2 n
/2
sin
= 0 unless n is
[ sin nx]0 + [sin nx]/2 =
n
n
2
odd. Or else sketch the even 2-periodic extension and compare with (a).

CVD EXERCISES

39

46 (S)
(a) (Try one or two of these, and leave the rest for revision.) Solve
the heat equation
ut = uxx for 0 < x < , t > 0

(a)(i) is hand-in
for Week 9

with the boundary conditions u(0, t) = u(, t) = 0 and each of


the following initial conditions:
(i) u(x, 0) = T
(ii) u(x, 0) = T if 0 < x < /2, u(x, 0) = T if /2 < x <
(iii) u(x, 0) = T x if 0 < x < /2, u(x, 0) = T ( x) if
/2 < x <
where T is a given constant.
(b) Repeat this question with the boundary conditions replaced
by ux (0, t) = ux (, t) = 0.

1
4T X
2
e(2m+1) t sin(2m + 1)x,
[Ans: (a)(i) u(x, t) =
m=0 2m + 1

4T X
1
2
e(4m+2) t sin(4m + 2)x,
m=0 2m + 1
2

4T X (1)m e(2m+1) t sin(2m + 1)x


(iii)
;
m=0
(2m + 1)2
2

4T X (1)m e(2m+1) t cos(2m + 1)x


(b)(i) T ; (ii)
;
m=0
(2m + 1)
2

T
2T X e(4m+2) t cos(4m + 2)x
(iii)

]
4
m=0
(2m + 1)2

(ii)

SOLUTION.
(a) From lectures, the solution has the form
Z

X
2
n2 t
u(x, t) =
bn e
sin nx where bn =
u(x, 0) sin nx dx.
0
n=1
(In an examination question it would be necessary to run
through the derivation of this form of solution.) Computing
the integral,
(
4T
2T 1 (1)n
, n odd, = 2m + 1;
=
(i) bn =
n

n
0,
n even. )
(Z
Z
/2
2T
(ii) bn =
sin nxdx
sin nxdx

0
/2
(
8T
o
n
2T n
if n is of the form 4m + 2
2 cos
1 cos n =
=
n
n
2
0
otherwise
Note that this is just a rescaled version of the solution
to (a).

CVD EXERCISES

40

)
(Z
Z
/2
2T
( x) sin nxdx
x sin nxdx +
(iii) bn =

/2
0


2T
1
x
1
x
/2

=
cos nx 2 sin nx]/2
[ cos nx + 2 sin nx]0 + [

n
n
n
n
(
4T
4T
n
(1)m if n = 2m + 1
2
=
sin
=
n
n2
2
0
otherwise
(b) Separating variables, we seek solutions u(x, t) = X(x)T (t) satisfying X 0 (0) = X 0 () = 0. Hence T 0 (t)X(x) = T (t)X 00 (x),
and so
X 00
T0
=
.
T
X
By the usual argument, each side is constant, say , and so we
have
T 0 = T
X 00 = X.
The eigenfunctions X are given in Q.43(c), i.e. 1 and {cos nx, n =
1, 2, . . .}, and the corresponding eigenvalues are = 0 and
n2 . Hence the corresponding functions T are 1 and exp(n2 t)
(or arbitrary constant multiplesPof these). By superposition,
2
the general solution is u = a0 +
n=1 an exp(n t) cos nx. For
the initial condition u(x, 0) = f (x) we require
f (x) = a0 +

an cos nx,

(15)

n=1

the coefficients are given by the usual Euler formulae, i.e.


Z
1
f (x)dx
a0 =
0
Z
2
an =
f (x) cos nxdx.
0

(i) Here eq.(15) is obviously satisfied if a0 = T, an = 0.


Hence result.
(ii) Here a0 = 0 obviously, while for n 1 we have
(Z
)
Z
/2
2
an =
T cos nxdx
T cos nxdx

0
/2
(
/2 
 )
2T
1
1
sin nx
sin nx
=

n
n
0
/2
2T
{sin n/2 + sin n/2}
n

if n is even
0
4T
=
(1)m if n = 2m + 1 is odd

(2m + 1)

CVD EXERCISES

41

(iii)
(Z

1
=

a0

/2

T xdx +
0

(

/2

T ( x)dx

 )

1 2
+ pix x
2
0
/2
 2 



2
2 2
T
2
+

=
8
2
2
8
2
T
.
=
4
T
=

an

2
=

(Z

1 2
x
2

/2

/2

T x cos nxdx +
0

/2

T ( x) cos nxdx

/2 
 )
1
1
1
1
+ ( x) sin nx 2 cos nx
x sin nx + 2 cos nx
n
n
n
n
0
/2
 


2T
n
1
n
1

n
1
n

1
=
sin
+ 2 cos
2 + 2 cos n
sin
+ 2 cos

2n
2
n
2
n
n
2n
2
n
2


n
2T
1 cos n
2 cos
=
2
n
2

0
if n is odd or a multiple of 4

2T
=
(4) otherwise, when n = 4m + 2 for some m 0,

(4m + 2)2
2T
=

(

which last expression reduces to

2T
.
(2m + 2)2

47 (S) (Try one of these and leave the other for revision.) Find the
solution of the wave equation utt = uxx for 0 < x < , t > 0 which
also satisfies the BC
u(0, t) = u(, t) = 0 (t > 0)
for each of the following sets of initial conditions:
(a) u(x, 0) = sin x, ut (x, 0) = 0 (0 < x < ), Here is a
positive constant. What happens if is an integer?
(b) (The plucked string)

x/a
if 0 < x a,
u(x, 0) =
1 (x a)/( a) if a x <
ut (x, 0) = 0, 0 < x < .

(The complicated formula just describes the triangular function which joins (0, 0) to (a, 1) to (, 0). Here 0 < a < .)

CVD EXERCISES

42

2 sin X (1)n1 n
sin nx cos nt if is not an inte2 2

n
1

X
sin na
2
sin nx cos nt]
ger, sin nx cos nt if = n; (b)
a( a) 1
n2
[Ans: (a)

SOLUTION.
(a) We have homogeneous Dirichlet BC here, so the general separable solution will be u(x, t) = (a cos nt + b sin nt) sin nx as
in lectures. b = 0 because ut (x, 0) = 0. By superposition, we

X
have the general series solution u(x, t) =
an cos nt sin nx
1

To determine the an we set t = 0 and equate the series to the


given IC,
R sin x =

an sin nx.

n=1

As usual, the Fourier coefficients are


2
an =

sin x sin nx dx.


0

This integral can be done using the trig identity


2 sin x sin nx = cos(n )x cos(n + )x
Clearly we have to distinguish the case that is an integer. If
= m, then we get
1
am =

(1 cos 2mx) dx = 1, an = 0 for n 6= m.

(Only the m-th harmonic is excited.)


If is not an integer, we get
1
an =



1 sin(n )x sin(n + )x
(cos(n)xcos(n+)x) dx =

n
n+
0
= 2n

(1)n1 sin
.
(n2 2 )

Notice as a check, that if m, then an 0 for n 6= m and


to 1 if n = m.

CVD EXERCISES

an

43

(b) Here the Fourier coefficients are given by


 
Z a
Z 
xa
x
2
1
sin nxdx +
sin nx dx
=

a
a
0 a
 


a
2
x
1
1
1
=
cos nx
2 sin nx

a
n
a
n
0
 
 



1
1
1
xa
cos nx
2 sin nx
+
1
a
n
a
n

 a
1
2
1
1
1
cos na + 2 sin na + cos na +
=
sin na

n
an
n
( a)n2
2 sin na
=
2
n a( a)
48 (S) (Try either (a) or (b) and leave the other for revision. (c) is a
challenge)
(a) Solve the heat equation ut = uxx for 0 < x < , t > 0 with
the inhomogeneous Dirichlet BC u(0, t) = T0 , u(, t) = T1 and
IC u(x, 0) = 0 for 0 < x < . Here T0 and T1 are constants.
[Hint: u0 can be taken to be a simple function of x only.]
What is the equilibrium state, that is, the limiting function
limt+ u(x, t)?
(b) Find the solution of the wave equation utt = uxx for 0 < x <
, t > 0 which also satisfies the IC and BC:

u(x, 0) = 0, ut (x, 0) = 0 for 0 < x < , u(0, t) = 0, u(, t) = R sin t,


where R > 0 and > 0 are constants, with not an integer.
(This represents a string with one end fixed and the other
being shaken vigorously.) [Hint: look for a special solution of
the form f (x) sin t to absorb the inhomogeneous BC.]
(c) Same problem, but now = m for some integer m.

x 2 X T0 (1)n T1
[Ans: (a) T0 + (T1 T0 )
exp(n2 t) sin nx,
1
n

sin t sin x 2R X (1)n

sin nx sin nt,


(b) R
sin
1 2 n2
R (2 mx sinm t cosm x + 2t m cosm t sin(m x) sinm t sinm x) (1)m
(c)
2
m

2Rm X (1)n

sin nx sin nt.]


1,n6=m m2 n2
SOLUTION.
(a) Since the BC are independent of t, we try u0 a t-independent
solution of the heat equation. Thus u0 (x) = A + Bx for some
constants A and B. Putting in the given BC, these must be
chosen so that u0 (x) = T0 + (T1 T0 )(x/). Then if v(x, t) =

CVD EXERCISES

44

u(x, t) u0 (x), we must have

v(0, t) = v(, t) = 0, v(x, 0) = 0 u0 (x),

and of course v satisfies the heat equation as well. So we can


find v in the form

X
2
v(x, t) =
An en t sin nx
n=1

where the An are the Fourier coefficients of u0 (x),


Z
2
An =
(T0 + (T1 T0 )(x/)) sin nx dx.
0

Doing the integral, I got

2(T0 (1)n T1 )
.
n
So the final solution is

X
2
u(x, t) = (T0 + (T1 T0 )(x/) +
An en t sin nx
An =

n=1

with the An as above. The limiting temperature distribution


is given by u0 (x)all other terms go to zero as t because
of the exponential factors.
(b) We will find u = v + f (x) sin t, where we want v to satisfy the
wave equation and homogeneous BC. Therefore, the particular
integral f (x) sin t must satisfy the wave equation and the
given inhomogenous BC. The wave equation gives
2 f (x) = f 00 (x), so f (x) = A cos x + B sin x

for arbitrary constants A and B. To satisfy the BC, f (0) = 0,


f () = R, so A = 0 and B sin = R. Therefore, if is not
an integer, we can solve for B and our solution will look like
sin t sin x
+ v(x, t)
u(x, t) = R
sin
By construction, v(x, t) will satisfy the wave equation, homogeneous BC and if u(x, 0) = 0, ut (x, 0) = 0 then we also need
sin x
.
sin
Then v will be given by a Fourier series as in the Q.(47), but
this time a will all be zero and the general series solution is

X
v(x, t) =
bn sin nx sin nt
v(x, 0) = 0, vt (x, 0) = R

n=1

and

Z
sin x
2
sin nx dx.
R
nbn =
0
sin
This integral is done in exactly the same way as in Q.(47).

CVD EXERCISES

45

(c) The Ansatz f (x) sin mt doesnt work. You could try a particular solution of the form =(a(x)eimt + b(x)teimt ) instead. You
will need b() = 0 to get the t-dependence right.
An alternative to pulling the solution out of a hat is to take a
limit m in (b). Though the particular integral is singular,
you can combine it with the singular term (n = m) in the
series. Maple gives the following:
>

assume(m, integer);

>
>
>
>

series(subs(omega=m+epsilon,
sin(omega*t)*sin(omega*x)/sin(omega*Pi) 2*omega*(-1)^m*sin(m*x)*sin(m*t)/(Pi*(omega^2-m^2))),
epsilon,1);

sin(m t) sin(m x) (1)m sin(m x) sin(m t) 1

) +
(1)m

sin(m t) cos(m x) x + cos(m t) t sin(m x) 1 (1)m sin(m x) sin(m t)


(

)+
(1)m
2
m
O()

>

simplify(%);

1 (2 m sin(m t) cos(m x) x + 2 m cos(m t) t sin(m x) sin(m t) sin(m x)) (1)m


2
m
+O()
The other terms of the series are uncomplicated.
49 (S) Let D be the domain given by the inequalities 0 x, y
. Find all separated solutions u(x, y) = X(x)Y (y) of Laplaces
equation which satisfy the BCs
u(0, y) = u(, y) = 0 for 0 < y <

(16)

u(x, 0) = 0 for 0 < x < .

(17)

and

Verify that these separated solutions satisfy the maximum principle, i.e. that for 0 < x, y < , u(x, y) is bounded above by the
maximum value of u on the boundary D of D.
Explain how you could use these solutions to find the function
u which satisfies u = 0 in D, the boundary conditions (16) and
(17), and
u(x, ) = h(x) for 0 < x < ,
where h is a given function. Apply your method in the case h(x) =
1, 0 < x < .

4 X sin(2m + 1)x sinh(2m + 1)y


.]
[Ans. sin nx sinh ny,
m=0 (2m + 1) sinh(2m + 1)

CVD EXERCISES

46

SOLUTION. If u(x, y) = X(x)Y (y), then


uxx + uyy

Y 00 (y)
X 00 (x)
=
.
= X (x)Y (y) + X(x)Y (y) = 0 iff
X(x)
Y (y)
00

00

Hence each side is a (possibly complex) constant 2 , and


X 00 (x) + 2 X(x) = 0 Y 00 (y) 2 Y (y) = 0.

(18)

Moreover, the BC give


X(0) = X() = 0, Y (0) = 0.

(19)

If = 0, then X(x) = A + Bx, Y (y) = C + Dy, but the BC 19 are


only satisfied if A = B = 0, so = 0 only yields the zero solution.
If 6= 0, then we get
X(x) = A cos x + B sin x, Y (y) = C cosh y + D sinh y.
The boundary conditions now force A = 0, = n (positive integer)
and C = 0. Hence the separated solutions are proportional to
un (x, y) = sin nx sinh ny.
Consider the separated solution
u(x, y) = sin nx sinh ny
We have sin nx 1 and, for 0 y , 0 sinh ny sinh n.
Hence u sinh n. The only part of the boundary where u can be
non-zero is on y = , where u = sin nx sinh n sinh n, and the
maximum is attained at x = /(2n) (say).
Notice that
un (x, ) = sinh n sin nx
so if
u(x, y) =

X
n=1

we have

u(x, ) =

An sin nx sinh n( y)

An sinh n sin nx.

n=1

Hence to solve the given boundary value problem, we just have to


put

X
h(x) =
An sinh n sin nx.
n=1

Thus the number An sinh n will be the Fourier coefficient of h(x),


Z
2
An =
h(x) sin nx dx.
sinh n 0

CVD EXERCISES

47

If h(x) = 1 we have
An

Q50(a) is hand-in
for week 10

Z
2
=
sin nx dx
sinh n 0
2
[cos nx]0 , dx
=
n sinh n
0
if n is even
4
=
if n = 2m + 1 is odd
n sinh n

50 (S) In each of the following cases, find a bounded single-valued


solution of Laplaces equation in the unit disk r < 1 which satisfies
the stated boundary condition:
(a) u(1, ) = sin3 , 0 < 2
(b) u = x4 on x2 + y 2 = 1
(c) u(1, ) = 2 2 ,
(d) u(1, ) = cos(/2),
[Ans. (a) (3r sin r 3 sin 3)/4,(b) (3 + 4r 2 cos 2 + cos 4)/8,

X
2
(1)n n
2
4 X (1)n n
(c) 2 4
r
cos
n,
(d)

r cos n]
2
2 1
3
n

4n
n=1
n=1

SOLUTION. Use the solution u =


where
an
bn

a0 X n
+
r (an cos n + bn sin n),
2
n=1

Z
1 2
=
u(1, ) cos nd
0
Z
1 2
u(1, ) sin nd,
=
0

though these integral formulae are not needed in the first two cases.
Also the limits can be any appropriate range of length 2. (You
are expected to be able to obtain this solution (bookwork).)
(a) u(1, ) = 3 sin /4 sin 3/4, and so by inspection we take
b1 = 3/4, b3 = 1/4, and all other coefficients vanish.
(b) u(1, ) = cos4 = (3 + 4 cos 2 + cos 4)/8. Proceed similarly.
(c) We have
Z
1 2
a0 =
( 2 )d



1 3
1 2

=

2 3 1 3
( ),
=

CVD EXERCISES

an

48

using the fact that the integrated function is odd. Continuing


for n > 0,
Z
1
=
( 2 2 ) cos nd






1
1
1
2
2 1
=
( ) sin n (2) 2 cos n + (2) 3 sin n

n
n
n



2
1
=
(0 + 2 2 cos n + 0).

n
The integrand for bn is an odd function.

(d)
an

Z
1
=
cos(/2) cos nd


Z 
1
1
1
cos(n + ) + cos(n
d
=
2
2
2


1
1
1
sin(n + 1/2) +
sin(n 1/2)
=
2 n + 1/2
n 1/2



n
n
(1)
(1)
2

=
2 n + 1/2 n 1/2
1 (1)n
= 2
.
n 1/4
Again bn = 0.

51 (S) Find all harmonic functions which (a) are separable in plane
polars, (b) are bounded in the half-disc 0 < r < 1, 0 < < ,
and (c) vanish on the straight part of its boundary. Hence find
the solution which also satisfies the boundary condition u(1, ) =
1, 0 < < .


y
2
y
1
1
tan
(P:) Verify that the function u(x, y) =
+ tan

1+x
1x
satisfies all the conditions of the above problem. Show that the level
sets of u are arcs of circles passing through the points P = (1, 0),
and give a geometrical interpretation of u.

1
4X
r 2m+1 sin(2m + 1);
[Ans. r n sin n, n = 1, 2, 3, . . .;
m=0 2m + 1
2 2P\
P P+ / where P = (x, y)]
SOLUTION. As in class, separation of variables in plane polars
leads to
r 2 R00 + rR0 R = 0
00 = .

(a) For > 0, set = p2 . Then = A cos p + B sin p. We


require (0) = () = 0, and so A = 0 and p = n = 1, 2, 3, . . ..
Then the DE for R(r) becomes r 2 R00 +rR0 p2 R = 0, which (as

CVD EXERCISES

49

in class) has general solution R = Cr n + Dr n . For finiteness


(as r 0) we need D = 0, giving harmonic functions u =
r n sin n.
(b) For = 0, = A + B, and boundary conditions (0) =
() = 0 A = B = 0.
(c) Likewise, < 0 gives P
no non-trivial solutions.
n
By superposition, u =
n=1 bn r sin n is also a solution satisfying the same boundary conditions (assuming satisfactory convergence of thePseries), where {bn } is a sequence of constants. We
require 1 =
n=1 bn sin n, which is a half-range sine series. Thus
Z
2
bn =
sin nd
0
2
= [cos n]0
n
2
= (cos n 1)
n
0
if n is even
4
=
if n is odd, say n = 2m + 1.
n
Hence result.
y
If v = tan1 then v = , which is obviously harmonic (using
x
2 u 1 u
1 2u
the expression 2 u =
+
+
). Since the Laplacian
r 2
r r r 2 2
y
operator is invariant under translation, the functions tan1
1x
are also harmonic, and sois the stated u, by superposition. Clearly
u(x, 0) = 0, while if y = 1 x2 (i.e. on
! the arc of the semi-disk),
r
r
2
1x
1+x
u=
tan1
= 1.
+ tan1

1+x
1x
y
y
+
2y
. Hence level
Now tan(u/2) = 1 + x y 1 y x =
1 x2 y 2
1
1+x1x
sets are parts of the curves Cy = 1 x2 y 2 . P lie on these. If
y
P = (x, y), P\
PO = tan1
, and so u/2 = P\
P P+ .
1x

CVD EXERCISES

50

52 (a) (E) By treating the nth term as the real part of a complex
number, show that

X
r(cos r)
r n cos n =
1 2r cos + r 2
1

if |r| < 1.
(b) (H) For Laplaces equation on the unit disk, find the series
solution which is finite at the origin and satisfies the boundary
condition u(1, ) = h(), where h is a given function of the
polar angle , and show that it can be written in the form
Z 2
Z

1
1 X n 2
0
0
r
u(r, ) =
h( )d +
h( 0 ) cos n( 0 )d 0 .
2 0
1
0
Assuming that the order of summation and integration may
be reversed, deduce that
Z
h( 0 )d 0
1 r 2 2
.
u(r, ) =
2
1 2r cos( 0 ) + r 2
0

SOLUTION.
(a) Let z = r exp(i). Then <z n = <(r n exp(in)) = r n cos n.
Hence the series is

X
z
<
zn = <
if |z| < 1
1z
1
z(1 z)
(1 z)(1 z)
r cos r 2
=
.
1 r(exp i + exp(i)) + r 2

= <

Hence result.
(b) As in Q.50, we have

where

a0 X n
r (an cos n + bn sin n),
+
u=
2
n=1

Z
1 2
an =
h( 0 ) cos n 0 d 0
0
Z
1 2
bn =
h( 0 ) sin n 0 d 0 ,
0
except that we have introduced a dummy variable of integration 0 . Substituting into eq.(52b) we get
Z 2
Z 2

X
1
0
0 1
n
u(r, ) =
h( )d +
r
h( 0 ) {cos n 0 cos n + sin n 0 sin n} d 0 .
2 0
1
0
Hence the first result.

CVD EXERCISES

51

Inverting the order of summation and integration we deduce


that
)
(
Z

X
1 2
1
u(r, ) =
+
r n cos n( 0 ) d 0
h( 0 )
0
2
1


Z 2
r(cos( 0 ) r)
1
0
h( ) 1 + 2
=
d 0
2 0
1 2r cos( 0 ) + r 2
Z 2
1
1 2r cos( 0 ) + r 2 + 2r cos( 0 ) 2r 2 0
d .
=
h( 0 )
2 0
1 2r cos( 0 ) + r 2
Hence result.
53 (S) The spherically symmetric 3-dimensional wave equation is
2 u 2 u
2u
+
=
. Find all separable solutions which (i) are finite
r 2
r r
t2
at r = 0 and (ii) vanish at r = a, where a > 0 is constant. (Hint:
to solve for the radial factor, either use Maple or the fact that the
1 2
left side of the equation may be written as
(ru).)
r r 2
Hence find the solution satisfying the above conditions and the
initial conditions u(r, 0) = 1, ut (r, 0) = 0, 0 < r < a.


nr
nt
nt
1
A cos
+ B sin
,
SOLUTION. [Ans: sin
r
a
a
a

nr
nt
2a X (1)n

sin
cos
]
r 1
n
a
a
Setting u = R(r)T (t) we get R00 T + (2/r)R0 T = RT 00 R00 /R +
(2/r)(R0 /R) = T 00 /T . By the usual argument both sides are constant, say . Thus R00 + (2/r)R0 R = 0, T 00 T = 0. First
take = p2 < 0 (which gives oscillatory solutions for T ). For
the equation for R, Maple gives the solution R = C1 sin(pr)/r +
C2 cos(pr)/r. Alternatively, using the hint, the equation to be
1 d2
solved is
(rR) + p2 R = 0, i.e. (rR)00 + p2 (rR). Hence rR(r) =
r dr 2
C1 sin(pr)+C2 cos(pr), and the Maple solution follows immediately.
For finiteness at r = 0 we require C2 = 0. To make R(a) = 0 we
require sin pa = 0, to avoid the trivial solution, and so pa = n, n =
1, 2, . . .. Since the equation for T gives T = A cos pt + B sin pt as
usual, we obtain the solution in the answer given above.
For = p2 > 0 we similarly obtain R = C1 sinh(pr)/r+C2 cosh(pr)/r.
Finiteness again requires C2 = 0, but the BC at r = a cannot be
satisfied. For = 0 the solution is R = C1 + C2 /r, which again
leads to the trivial solution if we apply the BCs.
The initial-value problem requires ut = 0 at t = 0. This is linear
and homogeneous and can be applied to each separable solution.
Thus we require T 0 (0) = 0, which requires B = 0. Hence the
most general separable solution satisfying the homogeneous BCs
nr
nt
1
cos
. By superposition, these conditions are
is u = sin
r
a
a

CVD EXERCISES

52

1X
nr
nt
bn sin
cos
, assuming satisfactory
r 1
a
a

nr
1X
bn sin
, which is not
convergence. Finally we require 1 =
r 1
a

X
nr
quite a sine-series. But r =
bn sin
, and so
a
1
Z a
2
nr
bn =
r sin
dr
a 0
a
 

a
a
a2
nr 
nr
2
(r)
(1) 2 2 sin
cos
=
a
n
a
n
a
0
2
2a
(1)n .
=
a n
also satisfied by u =

Hence result.
54 (S) The axisymmetric 2-dimensional heat equation is
2 u 1 u
u
+
=
. Find all separable solutions which (i) are finite
r 2
r r
t
at r = 0, (ii) vanish at r = a, where a > 0 is constant, and (iii)
decrease ast increases.

jn r
exp(jn2 t/a2 )]
[Ans: J0
a
SOLUTION. If u(r, t) = R(r)T (t) then R00 T + (1/r)R0 T = RT 0 ,
and so R00 /R + (1/r)R0 /R = T 0 /T = , say, where is constant
(by the usual argument). Then T = C exp(t). By (iii) we take
< 0, and so write = p2 , p > 0. Thus R00 + (1/r)R0 + p2 R = 0.
As in class, for a solution finite at r = 0 we take R = BJ0 (pr), and
R(a) = 0 pa is a positive zero of J0 , i.e. pa = jn , n = 1, 2, . . ..
Hence result.
55 (P) In plane polar coordinates the 2-dimensional wave equation is
2 u 1 u
1 2u
2u
+
+
=
. If we seek a solution in the form u =
r 2
r r r 2 2
t2
R(r)T (t) sin , show that
1
1
R00 + R0 2 R = R
r
r
T 00 = T

(20)
(21)

for some constant . Taking = p2 , p > 0, to give solutions which


are oscillatory in t, use Maple to find the most general solution of
eq.(20) which is finite at r = 0. Show graphically that there are
infinitely many of such solutions which vanish at r = a, where a >
0 is constant. Illustrate the first two of these as Maple animations.
[Ans.J1 (pr)(A cos pt + B sin pt)]

CVD EXERCISES

53

SOLUTION. Substituting the form of the solution into the partial differential equation, we get R00 T sin +(1/r)R0 T sin (1/r 2 )RT sin =
RT 00 sin . Dividing by RT sin we get R00 /R+(1/r)R0 /R(1/r 2 ) =
T 00 /T . By the usual argument, each side is constant, say .
Maple solves eq.(20) as follows:
>
dsolve(diff(R(r),r,r) + (1/r)*diff(R(r),r) +
>
(p^2-1/r^2)*R(r)=0,R(r));
R(r) = C1 BesselJ(1, p r) + C2 BesselY(1, p r)
Plotting the two functions (for p = 1) shows that the second is
singular at r = 0:
>
plot({BesselJ(1,r),BesselY(1,r)},r=0..1,y=-10..1);
















Confirming the limit requires a one-sided limit:


>
limit(BesselY(1,r),r=0,right);

Thus we take C2 = 0.
What Maple calls BesselJ(1,x) is conventionally written J1 (x),
and so we have R(r) = AJ1 (pr), where A is an arbitrary constant.
Plotting J1 over a larger domain gives the following:

CVD EXERCISES

54






 

 





 


 

Evidently (!) J1 has infinitely many zeros, say {jn }, n = 1, 2, . . .,


and the first two of these are given by the Maple command
>
p:=evalf(BesselJZeros(1,1..2));
p := 3.831705970, 7.015586670
The following produces the required animation:
>
addcoords(z_cylindrical,[z,r,theta],\
>
[r*cos(theta),r*sin(theta),z]);

>
>
>

n:=1:animate(plot3d,\
[BesselJ(1,p[n]*r)*cos(p[n]*t)*sin(theta),r=0..1,
theta=0..2*Pi,coords=z_cylindrical],t=0..2*Pi/p[n]);

CVD EXERCISES

55

n = 2 gives the second solution (not shown here).


56 (P) Bessels equation of order 0 is xy 00 + y 0 + xy = 0, and is satisfied
by y = J0 (x), which has infinitely many positive zeros jn , n =
1, 2, 3, .... Show that the equation can be written in the form (xy 0 )0 +
xy = 0. Use the differential equation, and integration by parts
whereZ appropriate, to show that
jn

(a)

Z0 j n

xJ0 (x)dx = jn J00 (jn ), and

1
xJ02 (x)dx = jn2 J002 (jn )
2
0
Using the fact that the functions {J0 (jn x)} are orthogonal on
(0, 1) with weight function x, find the orthogonal expansion of the
function f (x) = 1.

X
1
[Ans. 2
J0 (jn x).]
0
j
J
(j
)
n
n
0
1
(b)

SOLUTION. Clearly (xy 0 )0 = y 00 + y 0 .


(a) y = J0 (x) satisfies the DE, and so (xJ00 )0 +xJ0 = 0. Integrating
from 0 to jn gives the required result.

CVD EXERCISES

56

(b) Similarly
j

Z jn
Z jn
1 2 2 n
2

x2 J0 J00 dx
x J0
xJ0 dx =
2
0
0
0
Z jn
xJ00 (xJ00 )0 dx since J0 (jn ) = 0,
= 0
0

and using the second form of the DE


jn
1
=
(xJ00 )2 0 .
2

X
The orthogonal expansion is
an J0 (jn r), where
1

h1, J0 (jn x)i


hJ0 (jn x), J0 (jn x)i
R1
J0 (jn x)xdx
= R 01
J 2 (j x)xdx
0 0 n
R jn
J0 (t)tdt
where t = jn x.
= R 0jn
2
J
(t)tdt
0
0

an =

Hence result.

Power Series Solutions


dy
57 (E) Solve the problem
= y, y(0) = 1 (a) by constructing its
dx
Taylor series about x = 0, and (b) by substituting a power series
in x.

X
xn
[Ans.
]
n!
0
SOLUTION.
(a) differentiating the DE n times, y (n+1) = y (n) , n 0, where the
superscript denotes the order of differentiation, not a power;

X
y (n) (0) n
(n)
x .
hence y (0) = y(0) = 1. Substitute in
n!
0
P
P
P
P
n
n1
n
n
(b) yP=
=
0 an x
0 nan x
0 an x
0 (n+1)an+1 x =

n
0 an x (replacing n by n + 1 on the left hand side; note that
the limits of summation are then 1 to , but the first term
vanishes, and so we can take the limits as shown.) Equating
like powers of x, an+1 = an /(n + 1), n 0. y(0) = a0 = 1,
whence a1 = a0 /1 = 1, a2 = a1 /2 = 1/2, a3 = a2 /3 = 1/3!, . . .
and in general an = 1/n!.

You might also like