Advanced Analytic Methods in Applied Science and Engineering Hung Cheng
Advanced Analytic Methods in Applied Science and Engineering Hung Cheng
d2 y
y = 2 + x2 + ex .
1.Find the general solution of
dx2
Solution:
m2 = 1 m = 1. Thus the complementary solution is
Y = Aex + Bex .
We also have
1
1
2 = 2 ,
x2 = (1 + D2 + )x2 = x2 2.
2
1
D 1
1
1
1
ex = ex
1 = ex 2
1
D2 1
(D + 1)2 1
D + 2D
1
1
xex
= ex
1 = ex
1=
.
D(D + 2)
D(2)
2
Therefore, the general solution of the differential equation is
xex
y = Aex + Bex 4 x2 +
.
2
D2
e2x
1
1 x
+
e 2x
5
(D 1) 4
e2x
1 1
e2x
ex x
=
+ ex
2x =
+
2x.
5
D4
5
4
Therefore,
xex
e2x
+
2x.
y = aex + bex + c cos x + d sin x +
5
4
=
d3 Y
Y = 0. Thus
dx3
mx
3
2in/3
Y = e , m = 1 mn = e
.
sin x + cos x
y = Aex + ex [B cos(x 3/2) + C sin(x 3/2)] 2
.
2
The complementary solution satisfies
x + ax + x = b cos t, x
, a > 0.
dt
which describes a harmonic oscillator under a linear damping force as well
as an external harmonic force of the same frequency. Find the general solution
of this equation.
Solution:
p
m2 +am + 1 = 0 m1 = a/2 + a2 /4 1,
m2 = a/2
a2 /4 1.
p
If a < 2, the two roots above are a/2 i, where = 1 a2 /4, and the
complementary solution X is given by
X = eat/2 (c cos t + d sin t), where c and d are constants.
We note the motion is a damped oscillation with the angular fequency ,
which is smaller than unitythe angular frequency when a = 0.
p
If a > 2, the two roots are a/2 a2 /4 1. The complementry solution
is
2
2
X = eat/2 (ce a /41t + de a /41t ).
1
1
.
(mn m1 ) (mn mn1 ) D mn
Since
R x mx0
1
1
f (x) = emx [emx f (x)] =emx
e
f (x0 )dx0 ,
Dm
D
we have
yp (x) =
Rx
em1 (xx )
em2 (xx )
[
+
+
(m1 m2 )(m1 m3 ) (m1 mn ) (m2 m1 )(m2 m3 ) (m2 mn )
0
emn (xx )
] f (x0 )dx0 .
(mn m1 )(mn m2 ) (mn mn1 )
Chapter 2
1.As we know, the sum of two complex numbers can be represented as that of
two vectors in a two dimensional space. How about the product of two complex
numbers?
Solution:
Let
a1 + ia2 , b1 + ib2 .
These two complex numbers can be reprsented by the vectors
v a1 e1 + a2 e2 , u b1 e1 + b2 e2 .
where e1 and e2 are unit vectors in the direction of the x-axis and the y-axis,
respectively.
We have
= (a1 ia2 )(b1 + ib2 ) = (a1 b1 + a2 b2 ) + i(a1 b2 a2 b1 ).
The real part of is the scalar product of v and u ; the imaginary part
Solution:
uxx = 6ax , uyy = 0 .
Thus 2 u = 0 only if a = 0, in which case
u = x.
The conjugate of x, denoted as v, satisfies
vy = ux = 1 and vx = uy = 0.
We get
v = y + c,
and
u + iv = z + ic,
where c is a constant.
4. Find the roots of (z + 1)8 = (z 2 1)8 .
Solution:
The equation above gives
(z 2 1) = ei2n/8 (z + 1), n = 0, 1, 2 , 7 ,
or
(z + 1)(z 1 ei2n/8 ) = 0, n = 0, 1, 2 , 7.
Thus the roots are
z = 1
and
z = 1 + ei2n/8 , n = 0, 1, 2 , 7 .
R sin2 x
dx,
1 + x2
c.
1 + 3x2
dx.
(1 + x2 )(4 + x2 )
Solutions:
a. We get
R (1 + eix )dx
J
=0
(1 x2 )
convergent integral.) Thus
I = Re J = 0.
b. I =
Let
R sin2 x
1 R e2ix 2 + e2ix
dx
=
dx.
1 + x2
4
1 + x2
R e2ix 1
dx.
1 + x2
By clsoing contour upstairs, we get
e2 1
= (1 e2 ).
J = 2i
2i
Thus
I = Re J/2 = (1 e2 )/2.
J
I = 2i Residue(z2 ) =
= .
(3 5) 3
5
b. Closing the contour upstairs, we get
I = 2iRes(i).
Since the integrand has a third-order pole at i, we get
6
1
6
3
1
d2
=
=
lim 2
.
2 zi dz (z + i)3
(2i)5
16i
3
We have I =
.
8
Res(i) =
R sin2 x
dx,
x2
Solution:
1 R
dx
.
2 (9 + x2 )(4 + x2 )
Closing the contour upstairs, we get
1
1
+
] = /60.
I = i[
(6i)(5) (5)(4i)
a. I =
b. Since the integrand is an entire function, we deform the contour and get
1 R e2iz 2 + e2iz
I= c
dz,
4
z2
where c is a contour in the upper half-plane joining and . The first
two terms in the integrand give zero. We get
2i
de2iz
I=
lim
= .
4 z0 dz
9. Evaluate the following integrals:
R
sin cx
a.
dx,
(x + a)2 + b2
where a is a real number while b and c are both real and positive.
R
x2
b. 4
dx.
x +1
1/z
R e
c. C
dz, where c is a circle around the origin in the counterclockwise
z
direction.
7
Solution:
eicx
dx = Im[2iRes(a + ib)]
(x + a)2 + b2
eiac ebc
ebc sin ac
= Im[2i
]=
.
2ib
b
b. Closing the contour upstairs, we get
I = (2i)[Res(ei/4 )+Res(ei/4 )].
Since
z2
z2
Res(ei/4 )= 3 ]z= exp(i/4) ,
Res(ei/4 )= 3 ]z=exp(i/4) ,
4z
4z
1
1
2i sin(/4)
I = (2i)[ i/4 +
] = (2i)
= 2/2.
i/4
4
4(e
) 4(e
)
a. I = Im
c. We have
R 1
R e1/z
dz = C (1 + z + z 2 /2 + )dz.
I= C
z
z
By the Cauchy residue theorem, we get
I = 2i,
10. Evaluate the integral I
R
0
x
dx.
1 + xn
Solution: R
Let J C
z
dz,
1 + zn
where C is the contour of a pie with two of its boundaries the infinite ray of
= 0 and that of = 2/n. Then
J = (1 e4i/n )I .
By the Cauchy residue theorem, we have
2i
J = i(n2)/n .
ne
Thus
I=
.
n sin(2/n)
11. Find the branch points of the function f (z) = (z 4 1)1/4 . Draw a set of
branch cuts for this function to make this function single-valued in the complex
plane.
Solution:
The branch points of this function are at
z = 1, i, 1, i .
As z , f (z) z(1 z 4 /4 + ).Thus z is not a branch point of the
function. We draw straight lines joining 1 to i , i to 1 , and 1 to i , and
the function is single-valued in the cut-plane.
1 + z 1/3 1
)
in the ex1z
1 + z2
tended z-plane. Draw branch cut (or cuts) to make this function single-valued.
12. (a) Locate the singularities of the function (
Therefore,
J = (ei/6 e2i/3 ei/6 ).
We get
ei/3 ei/6 ei/3 ei/6
ei/6 e2i/3 ei/6
=
I=
1 e2i/3
ei/3 ei/3
sin(/6)
=
= .
sin(/3)
3
13. (a) Locate and classify the singularities of the function (z + 2)1 (1
z )
in the extended z-plane. Draw branch cut (or cuts) to make the function
single-valued.
(b) Evalaute the integral
R1
dx
.
I = 1
(x + 2)(1 x2 )1/2
2 1/2
Solution:
(a) The function (z + 2)1 (1 z 2 )1/2 has a simple pole at 2 and branch
points at 1. This function is single-valued in the plane cut from 1 to 1.
(b) Let
I
dz
J=
,
2 1/2
c (z + 2)(1 z )
where c is the closed contour wrapping around the branch cut in the clockwise
direction. We note that
J = 2I.
We may view this contour as the one enclosing the cut-plane in the counterclockwise direction. Since the integrand vanishes like z 2 as z , the only
singularity enclosed by the contour is at z = 2. Therefore, the Cauchy residue
theorem gives
2
2i
2i
J = 2i Res(2) =
= = .
(1 4)1/2
i 3
3
Thus
I= .
3
14. Let f (z) be analytic at z0 and f (z0 ) 6= 0. Prove log f (z) does not have
a branch point at z0 .
Solution:
If f (z0 ) 6= 0, f (z) has no zero in a sufficiently small neighborhood of z0 .
Therefore, the argument of f (z) does not change as we go around a closed
contour in this neighborhood. As a result, the value of log f (z) does not change
as we go around this closed contour. Therefore, z0 is not a branch point of f (z).
15. Evaluate the integral
.
.
R
0
dx
, where 0 < Re a < 1.
xa (1 + x)
10
Solution:
The function f (z) =
1
has a branch point at z = 0. We choose the
z a (1 + z)
branch cut of this function to be the positive x-axis.
Let
R
dz
J= c a
,
z (1 + z)
where c wraps around the positive x-axis in the clockwise direction. The
contour goes around the singularity at the origin via a circle of radius .
The integral of f (z) over a circle of radius R is of the order of
R
,
(Ra )(R)
which vanishes as R . ( This is because a > 0.) Therefore, we may close
the contour by adding an infinite circle to c. By the Cauchy residue theorem,
1
J = 2iRes(-1)= 2i ia .
e
Now the contour c had three sub-contours:
1. Sub-contour c1 , the positive x-axis. The integral over c1 is I.
2. Sub-contour c2 , the circle of radius around the origin. The integral over
c2 is proportional to
,
a
which vanishes as 0. (This is because a < 1.)
3. Sub-contour c3 , which is the x-axis below the branch cut. The integrand
on c3 is equal to
1
1
= e2ia a
.
(e2i x)a (1 + x)
x (1 + x)
Thus the integral over c3 is equal to
e2ia I.
The minus sign above us due to the direction of the contour eing clockwise.
Thus we have
J = (1 e2ia )I.
We get
I=
.
sin a
16. Find
Z the value of
Z
ln x
1
I=
dx. Reminder:
dx =
.
5
5
5 sin(/5)
0 1+x
0 1+x
Solution:
Let
I
ln z
J=
dz,
5
c1 + z
where c is the closed pie-shape contour with angle 2/5 which encloses the
simple pole at ei/5 . By the Cauchy residue theorem, we get
i/5
2 2 ei/5
2 2 e4i/5
J = 2i 4i/5 =
=
.
25
25
5e
To extract I from J, we have
11
J =I e
2i/5
Thus we have
(1e2i/5 )I =
2i 2 e2i/5
ln r + 2i/5
2i/5
dr
=
(1
e
)I
.
1 + r5
25 sin(/5)
2 2 i/5
ei/5 i
2 2 i/5
i cos(/5) sin /5
e
(1+
)=
e
(1+
)=
25
sin(/5)
25
sin(/5)
2 2 i/5 i cos /5
e
,
25
sin(/5)
or
2 cos(/5)
.
I=
25 sin2 (/5)
17. Evaluate I =
x
dx.
sinh x
Solution:
We note sinh z = (ez ez )/2 is finite if the real part of z is finite, regardless
z
of the vaue of the imaginary part of z. As a result,
does not vanish at
sinh z
all points of the upper infinite half circle. Consequently, we cannot close the
contour upstairs. For the same reason, we cannot close the contour downstairs.
LetsIclose the contour in a differently way. We define
z
dz,
J=
sinh
z
c
where c is the boundary of the ( infinite) rectangle the horizontal sides of
which are the x-axis and the line z = x + i.
On the line z = x + i, the integrand is equal to
x + i
.
sinh(x + i)
Since
sinh(x + i) = sinh x,
the integrand on this line is equal to
x + i
,
sinh x
which blows up at x = 0.
To avoid the singularity at z = i, we make the indentation of an infinitesimal half circle below i.
We shall denote the x-axis as c1 , the line z = x + i with |x| > as c2 , and
the indented half circle as c3 .
Since c encloses no singularity of the integrand,
J = 0.
Now the integral over c1 is I.
The integral over c2 is
R x + i
R
x
P
dx = P
dx = I,
sinh x
sinh x
as
R idx
P
= 0.
sinh x
z
The residue of
at i is
sinh z
12
i
.
1
Thus the integral over c3 is equal to
i(i) = 2 .
Finally, the integral over the two vertical boundary lines vanishes.
Thus we have
J = 2I 2 .
We get
I = 2 /2.
2
18. Find the Fourier transform of the function f (x) = ex , < x < .
Solution:
2
The
transform of eRx is
R Fourier
2
2
2
dx eikxx = ek /4 dxe(x+ik/2) .
Solution:
f (x) = (x2 + 1)1 ,
R
eikx
f (k) = dx
.
1 + x2
If k > 0, we close contour downstairs and get
ek
f (k) = 2i
= ek , k > 0.
2i
If k < 0, we close contour upstairs and get
ek
= ek , k < 0.
2i
f (k) = 2i
a < x < a,
|x|> a,
13
(a) F (k) =
Ra
a
eikx dx =
eika eika
2 sin(ka)
=
.
ik
k
ik
2
|x| > a,
= 1,
|x| < a.
eika eika
Since
is an even function of k, the integral above is an even
ik
function of x. Thus we only need to show that the above is true when x is
positive, or
R ikx eika eika dk
e
= 0,
x > a,
(I)
ik
2
= 1,
0 < x < a.
(II)
We deform the contour to the upper half plane, avoiding the origin k = 0.
If x > a, both (x + a) and (x a) are positive. Thus we may close the
contour upstairs and get
R ikx eika eika dk
e
= 0,
ik
2
and (I) is proven.
If 0 < x < a, (x + a) is positive while (x a) is negative. Closing the contour
upstairs, we get
R ikx eika dk
= 0.
e
ik 2
We also get, by closing the contour downstairs,
R ikx eika dk
1
1
e
= (2i)( )( ) = 1.
ik 2
i 2
Thus (II) is proven.
21. Find the Fourier transform of
1
, < x < .
f (x) =
1 + x4
Solution:
The Fourier transform of
1
is
1 + x4
R eikx
dx
1 + x4
The poles are located at
ei/4 , e3i/4 .
For k > 0 , we close the contour downstairs; and for k < 0 , we close the
contour upstairs. We then obtain the Fourier
transform of the function as
|k|(1+i)/ 2+i/4
e
+ c.c. = e|k|/ 2 cos(k/ 2 /4) .
2
22. Find the Fourier coefficient cn of the function f () = e , << . By
evauating the Fourier series for f () at = and at = , prove
14
e + e
1
1
=
.
2
e
1
+
n
2
e
2
n=1
Solution:
R
d
ein e
e e
cn = ein e
=
| = (1)n
.
2
2(1 in)
(1 in)
Thus
e e X (1)n ein
.
e =
2
1 in
n=
At = and = , the Fourier seris has the same value
e e X
2
1
e e
2
2
++
+ )
=
(1 +
+
2
2
1
in
2
1
+
1
1
+
2
1
+
n2
n=
X
1
e e
=
(1 + 2
).
2
1
+
n2
n=1
Since f () = e , f () = e , the Fourier seris at = is equal to the
average of f () and f (). Thus
X
e + e
e e
1
=
(1 + 2
),
2
2
1
+
n2
n=1
or
P
1
e + e
1
=
.
2
2 e e
2
n=1 1 + n
23. Find Fourier coefficient an for
1
, with a and b positive and b < a.
f () =
a + b cos
Solution:
We have
1 R
ein d
.
an =
2 a + b cos
Note that
an = an .
Let z Iei , then we have
I
1
z n1 dz
1
z n dz
an =
=
.
1
i c [2a + b(z + z )]
bi c [2az/b + (z 2 + 1)]
where c is the unit circle and n is non-negative. The integrand has simple
poles at
p
z = a/b a2 /b2 1.
p
The pole inside the unit circle is located at a/b+ a2 /b2 1.Using Cauchys
residue theorem, we get,
for n > 0,
(1)n (a a2 b2 )n
an =
.
bn a2 b2
Since an is real, we have
(1)n (a a2 b2 )n
, n 0.
an = an =
bn a2 b2
15
d2 y
+ y = x, x > 0,
2
dx
with the initial condition
y(0) = 1, y 0 (0) = 0.
Solution:
We have
s + (s2 + 1)Y (s) = (3/2)/s3/2 ,
where Y (s) is the Laplace transform of y(x). Thus
(3/2)
s
+ 3/2 2
.
Y (s) = 2
s + 1 s (s + 1)
By performing Ran inverse Laplace transform, we get
x
y(x) = cos x + 0 x01/2 sin(x x0 )dx0 .
The last integral is obtained by using the convolution theorem.
26. Let the Laplace equation 2 u(r, ) = 0 be satisfied inside the disk r < 5,
where r and are the polar coordinates,and let the boundary condition be
u(5, ) = sin2 .
Find u(r, ).
Solution:
(ei ei )2
1 cos 2
sin2 =
=
.
4
2
Thus
1 r2 cos 2/25
u(r, ) =
.
2
16
Chapter 3
1. Find the solution u(x, y) of
ux + (x2 + y)uy = 0,
u(0, y) =
1
.
1 + y2
Solution:
The equation for the characteristic curves is given by
dx
dy
= 2
,
1
x +y
or
dy
= y + x2 .
dx
The complementary solution of the equation above is
Y = cex ,
and a particular solution is
1
yp =
x2 = (1 + D + D2 + )x2 = (x2 + 2x + 2).
1D
Thus the characteristic curves are given by
y = cex (x2 + 2x + 2),
or
(y + x2 + 2x + 2)ex = c.
Therefore, the general solution of the PDE is
u(x, y) = f ((y + x2 + 2x + 2)ex ).
The initial condition gives
1
= f (y + 2),
1 + y2
or
1
.
f (y) =
1 + (y 2)2
Therefore,
u(x, y) =
1 + [(y +
x2
1
.
+ 2x + 2)ex 2)]2
Solution:
The equation for the characteristic curves is
dx
dy
= 2x
,
1
e +y
or y 0 = y + e2x y = cex + e2x (y e2x )ex = c.
Therefore, the general solution for thie PDE is
u(x, y) = f (yex ex ).
From the initial condition, we get
17
ey = f (y 1) f (y) = ey+1 .
Therefore we have
x
x
u(x, y) = eye e +1 .
3. Solve
ut + xux = 0
2
with the initial condition u(x, 0) = ex .
Solution:
The equation for the characteristic curves are
dx
dt
=
.
1
x
Thus the characteristic curves are given by
t = ln x + c0 ,
or
xet = c.
Therefore, the general solution of the equation is
u(x, t) = f (xet ).
To determine f , we set t = 0 and find that
2
f (x) = ex .
Thus
2 2t
u(x, t) = ex e .
Chapters 4 and 5
2u 2u
+ 2 = u holds in the rectangle with vertices (0, 0),
x2
y
y
(a, 0), (a, b), (0, b). At the right vertical side, u(a, y) = sin( ), 0 < y < b. The
b
function u vanishes at the other three sides of the rectangle. Find u.
1. The equation
Solution:
Since u vanishes at the two horizontal sides, we express u as a Fourier sine
series of y :
P
ny
u(x, y) =
An (x) sin(
).
b
n=1
Substituting u into the partial differential equation, we get
d2 An (x) n2 2 + b2
An (x) = 0.
dx2
b2
2
An (x) = cn cosh(x n 2 + b2 /b) + dn sinh(x n2 2 + b2 /b).
Thus
18
ny
[cn cosh(x n2 2 + b2 /b) + dn sinh(x n2 2 + b2 /b)] sin(
).
b
n=1
The boundary condition u(0, y) = 0 gives
P
ny
0=
cn sin(
),
b
n=1
or cn = 0.
y
That u(a, y) = sin( ) gives
b
P
ny
y
2
2
dn sinh(a n + b2 /b)] sin(
) = sin( ),
b
b
n=1
or
dn = 0, n 6= 1,
d1 = 1/(sinh(a 2 + b2 /b).
We have
y
u(x, y) = sinh(x 2 + b2 /b)] sin( )/(sinh(a 2 + b2 /b).
b
u(x, y) =
2u
2u
= 2 , < x < , t > 0, with the
2
x
t
initial conditions u(x, 0) = f (x), ut (x, 0) = 0.
a. Find the
R equation satisfied by L(x, s), where
L(x, s) 0 dt est u(x, t).
b. Assuming that both f (x) and L(x, s) have Fourier transforms, find L(x, s)
in the form of a Fourier integral. (You are allowed to differentiate a Fourier
integral by differentiating its integrand.)
c. Find u(x, t).
Note: the Laplace transform of u00 (t) is equal to s2 L(s)u0 (0)su(0), where
L(s) is the Laplace trandform of u(t).
2. Consider the wave equation
Solution:
a. Let the Laplace transform of u(x, t) be L(s, x).We have, by multiplying
the equation above with est and integrate with respect to t from 0 to ,
d2 L(x, s)
s2 L(x, s) = sf (x).
dx2
b. Let
R dk ikx
e l(k, s)
L(x, s) =
2
and substitute it into the equation for L(x, s) above, we have
(k 2 s2 )l(k, s) = sF (k),
where F (k) is the Fourier transform of f (x).Therefore,
R dk ikx sF (k)
sF (k)
l(k, s) = 2
L(x, s) =
e
.
2
k +s
2
k 2 + s2
c. We have
R ds st R dk ikx sF (k)
u(x, t) = c
e
e
,
2i
2
k 2 + s2
where c is a vertical contour to the right of all singularities of L(x, s). By
closing contour in the s-plane to the left, we get
1 R dk ikx ikt
f (x + t) + f (x t)
u(x, t) =
e (e + eikt )F (k) =
, t > 0.
2
2
2
19
Note: It is easier to find the solution of this problem by making the Fourier
transform with respect to x first, which we have demonstrated in the textbook.
3. Consider the Laplace equation
2 u(r, ) = 0
which holds inside a disk of radius 3 with the center at the origin. The
boundary condition is
sin n
P
.
u(3, ) =
n=1 n!
Find u(r, ) in a closed form.
Hint: Make use of ex = 1 + x + x2 /2! + + xn /n! + .
Solution:
sin n r
ein ein r
P
1 P
u(r, ) =
( )n =
( )n
3
2i n=0
n!
3
n=1 n!
i r n
i r n
)
(e
)
(e
1 P
3
3 = 1 [erei /3 erei /3 ]
=
2i n=0
n!
2i
1 r cos /3 ir sin /3
ir sin /3
= e
[e
e
] = er cos /3 sin(r sin /3).
2i
Note: u(x, y) is the imaginary part of the analytic function ez/3 .
= 2,
= (x, t),
i
t
x
which holds for all values of x and all positive values of t. Let the initial
value of (x, t) be
(x, 0) = f (x)
and let the boundary conditions for be
(, t) = 0.
(k, t)
i
= k 2 (k, t).
t
The solution of the equation above is
(k, t) = a(k)eik t .
Thus
R dk ikxik2 t
(x, t) =
e
a(k).
2
b. To satisfy the initial condition, we have
20
a(k) = f (k).
Therefore
R dk ikxik2 t
f (k).
(x, t) =
e
2
The answer above can be expressed as an integral over x0 :
R dk ikxik2 t R
0
(x, t) =
e
dx0 eikx f (x0 )
2
r
R
1
(x x0 )2 i
= dx0
exp[
] f (x0 ).
4it
4t
5. Find the eigenvalues and the eigenfunctions of the Sturm-LIouviile problem
d2 u
= u, 0 < x < L; u0 (0) = 0 and u(L) = 0.
dx2
Solution:
The general solution
of thedifferential equation above is
a cos( L) = 0.
Since a 6= 0 for a non-trivial solution, the eigenvalues of the non-trivial
solutions
are
n L = (n + 1/2), n = 0, 1, 2,
or
(n + 1/2)2 2
n =
, n = 0, 1, 2, .
L2
The eigenfunction
corresponding to the eigenvalue above is
r
(n + 1/2)x
2
cos[
].
un (x) =
L
L
These eigenfunctions are orthonormal:
RL
un (x)um (x)dx = nm .
0
6. Find the solution of uxx + uyy = 0, > y > 0 and 0 < x < 1,
with ux (0, y) = u(L, y) = 0, u(x, 0) = (x 1/2) and u(x, ) = 0.
Solution:
We express u by the Fourier cosine series
P
u(x, y) =
An (y) cos[(n + 1/2)x],
n=0
21
P
u(x, y) =
bn e(n+1/2)y cos[(n + 1/2)x].
n=0
P
(x 1/2 ) =
bn cos[(n + 1/2]x].
n=0
Thus we have
R1
1
bn = 0 (x 1/2) cos[(n + 1/2)x]dx = cos[(n + 1/2)/2],
2
or
bn = 2 cos[(n + 1/2)/2].
Therefore we have
P
(n + 1/2)
u(x, y) = 2
e(n+1/2)y cos[
] cos[(n + 1/2)x].
2
n=0
7. Consider the vibration of the surface of a drum of unit radious. The
equation of motion is
2u
1 u
1 2u
=
(r
)
+
, r < 1, 0 2,
t2
r r r
r2 2
where u(r, , t) is the displacement of the drum surface, and r and are the
polar coordinates. The boundary condition is
u
.
ur (1, , t) = 0, where ur
r
The initial conditions are
u(r, , 0) = cos ,
ut (r, , 0) = 0.
Find u(r, , t).
Solution:
Substituting u = T (t)R(r)() into the PDE, we get
T
1 d
=
(rR0 ) + 2 = c.
T
rR dr
r
d dR
n2
Thus T = cT, = n2 ,
(r
)
R = crR.
dr dr
r
Since is required to satisfy ( + 2) = (), n is an integer.
The solution
of the equation for R is
R = Jn ( cr),
which satisfies the boundary condition of R(0) being finite, where Jn is the
Bessel function of order n. Let rm,n be the mth root of Jn0 , i.e., Jn0 (rm,n ) = 0,
m = 1, 2,
22
Then
cos(rm,n t)Jn (rm,n r)(Am,n cos n+Bm,n sin n) satifies the boudary
n,m
n,m
Thus all Fourier sine coefficients Bm.n = 0, while the only non-vanishing
Fourier
P cosine coefficients are Am,1 . These coefficients are determined from
J1 (rm,1 r)Am,1 = 1.
m
We get R
1
J1 (rm,1 r)rdr
Am,1 = R 10
[J1 (rm,1 r)]2 rdr
0
and
P
u(r, , t) =
Am,1 cos(rm,1 t)Jn (rm,1 r) cos .
m=1
1
2
1
[ 2 r2
+ 2 2
+ 2
sin ]u = V (r, , )u,
r r r r sin 2
r sin
where
V3 ()
V2 ()
.
V (r, , ) = V1 (r) + 2 2 +
r2
r sin
Perform separation of variables and reduce the equation into three ordinary
differential equations.
Solution:
Substituting
u = R(r)()()
into the PDE and multiplying the resulting equation by r2 , we get
1 d 2 dR
1 d2
1
1 1
d
d
[
[
(r
) r2 V1 (r)] +
V2 ()] + [
( sin
)
2
2
R dr
dr
sin d
d
sin d
V3 ()] = 0.
Note that the first bracket above is a function of r only, while the second
and the third brackets are functions of and only. Thus we have
1
1 d2
1 1 d
d
1 d 2 dR
(r
) r2 V1 (r) = 2 [
V2 ()] [
(sin
)
R dr
dr
sin d
d
sin d2
V3 ()] = a,
where a is a constant. The equation above contains two ordinary differential
equations:
d 2 dR
(r
) [r2 V1 (r) + a]R = 0,
dr
dr
and
1 1 d
1
1 d2
d
2 [ d2 V2 ()] + [ sin d (sin d ) V3 ()] + a = 0.
sin
23
d
d
We see that the first bracket above is a function of only, while the second
bracket above is a function of only. Thus we have
1
d
d
1 d2
V2 () = sin (sin
) + sin2 V3 () a sin2 = b,
d2
d
d
where b is a constant. Therefore we have the ordinary differential equations
d2
[b + V2 ()] = 0
d2
and
d
d
sin [sin
] + [b + a sin2 sin2 V3 ()] = 0.
d
d
V3 ()
V2 ()
(Indeed, it is not difficult to prove that V = V1 (r) + 2 2 + 2 is the
r
r sin
most general form of V for the equation in this problem to be separable.)
d2 u
6.
= u, 0 < x < L; u0 (0) = 0 and u(L) = 0.
dx2
The general solution
of thedifferential equation is
a cos( L) = 0.
Since
a 6= 0, the eigenvalues of the non-trivial solutions are
n L = (n + 1/2), n = 0, 1, 2,
or
(n + 1/2)2 2
n =
, n = 0, 1, 2, .
L2
The eigenfunction
corresponding to the eigenvalue above is
r
2
(n + 1/2)x
un (x) =
cos[
].
L
L
These eigenfunctions are orthonormal:
RL
un (x)um (x)dx = nm .
0
3. uxx + uyy = 0, > y > 0 and 0 < x < 1;
ux (0, y) = 0, u(L, y) = 0
and
u(x, 0) = (x 1/2) and u(x, ) = 0.
Let us express u by the Fourier cosine series
24
u(x, y) =
n=0
P
u(x, y) =
bn e(n+1/2)y cos[(n + 1/2)x].
n=0
bn cos[(n + 1/2]x].
n=0
Thus we have
R1
1
bn = 0 (x 1/2) cos[(n + 1/2)x]dx = cos[(n + 1/2)/2],
2
or
bn = 2 cos[(n + 1/2)/2].
Therefore,
P
(n + 1/2)
] cos[(n + 1/2)x].
u(x, y) = 2
e(n+1/2)y cos[
2
n=0
Chapter 6
1. Find the two independent solutions of
d2 y
dy
+x
+y =0
dx2
dx
in the form of Maclaurin series. One of the series can be summed into a
closed form. Find this closed form.
Solution:
Since x = 0 is an ordinary point of the equation, we put y =
P
d2 y
=
An n(n 1)xn2 ,
dx2
P
P
dy
+y =
An (n + 1)xn = An2 (n 1)xn2 .
x
dx
We have
An2
An =
.
n
25
An xn . Thus
Note: We may take advantage of the fact that the Wronskian between y1
and y2 is a known function, and express y2 as an integral.
2. Find the Frobenius solutions of
3
dy
d2 y
+ y = 0,
x(1 x) 2 +
dx
dx 4
the truncated series of which are good approximations of the solutions near
x = 0.
Solution:
point x = 0 is a regular singular point. Thus we try
y =
P a. The
An xn+s .
We have
P
P
d2 y dy
x 2+
=
An [(n+s)(n+s1)+(n+s)]xn+s1 =
An (n+s)2 xn+s1 ,
dx
dx
P
d2 y
3
x2 2 + y = An [(n + s)(n + s 1) 3/4]xn+s
dx
4
P
P
= An (n + s 3/2)(n + s + 1/2)xn+s = An1 (n + s 5/2)(n + s
n+s1
1/2)x
.
Thus we have
(n + s)2 An = (n + s 5/2)(n + s 1/2)An1 .
The indicial equation is obtained by setting n in the equation above to zero.
We get
s2 = 0 = s1 = s2 = 0.
(n + s 5/2)(n + s 1/2)
An =
An1
(n + s)2
[(n + s 5/2) (s 3/2)][(n + s 1/2) (s + 1/2)
= An =
(n + s)2 (1 + s)2
26
[(s + 1)]2
(n + s 3/2)(n + s + 1/2)
,
[(n + s + 1)]2
(s 3/2)(s + 1/2)
where we have set A0 to unity. Thus one of the solutions is obtained by
setting s = 0 :
P (n 3/2)(n + 1/2) n
1
x .
y1 (x) =
(3/2)(1/2)
[n!]2
The second solution is
X
[(n 5/2) (3/2)][(n 1/2) (1/2)]
= (ln x)y1 (x) +
(n!)2
n=0
1
1
1
1
1
1
[(
++
)+(
++
) 2( + + )]xn
n 5/2
3/2
n 1/2
1/2
n
1
=
P
xy 0 = an1 (n + s 1)xn+s1
and P
xy 00 =
an (n + s)(n + s 1)xn+s1
We then have
an (n + s)(n + s 1/2) = an1 (n + s 1 + b)
Setting n = 0 in the equation above, we get
s = 0 or 1/2 .
For n 6= 0, we have
(n + s 1 + b)
an1
an =
(n + s)(n + s 1/2)
(n + s + b)(s + 1)(s + 1/2)
a0 .
=
(s + b)(n + s + 1)(n + s + 1/2)
Setting s = 0 , we find that one of the solutions is
P
(n + b)
y1 =
xn .
n=0 (n + 1/2)n!
Setting s = 1/2 , we find the other solution is
(n + 1/2 + b)
P
y2 =
xn+1/2 .
n=0 (n + 3/2)n!
4. Consider the differential equation
y + x2 y = 0 .
a.Find and classify the singular points of this equation in the extended complex plane.
b.Find one (any non-trivial one) of its solutions in the form of a series expanded around the origin.
Solution:
a.There are no singular points in the finite plane.
1
Let x = ,
t
then
d dy
d2 y
dy
y = t2 t2
= t4 2 + 2t3
,
dt dt
dt
dt
and hence
d2 y 2 dy
1
+
+ 6y = 0 .
dt2
t dt
t
Thus the infinity is an irregular singular point of the differential equation.
b. Let
P
y = an xn
with the understanding that
a1 = a2 = = 0 .
28
We have
P
y = an n(n 1)xn2
and P
P
x2 y =
an xn+2 =
an4 xn2 .
Thus we have
an n(n 1) = an4 ,
or
an4
.
an =
n(n 1)
Thus
a4(m1)
1
a0 (3/4)
a4m =
= ( )m
.
16m(m 1/4)
16 m!(m + 3/4)
Thus a solution is
X
x4m
1
.
( )m
16 m!(m + 3/4)
m=0
7.a. Find with the Frobenius method one of the independent solutions in
the form of a series (expanded around the origin) for
d2 y
dy
3
x(1 x) 2 +
+ y = 0.
dx
dx 4
b. Find the second independence solution of the equation.
29
30