Problem Set 2

PHYS 112 Mathematical Physics II

1st Sem 2017-18

Problem 1. (Butkov No. 7. 2) Find the Fourier transforms of the following

functions:
2
(a) f (x) = ex cos x ( > 0)
cos x
(b) f (x) =
a4 + x4

cos k0 x (|x| < N /k0 )
(c) f (x) =
0 (|x| > N /k0 ), N Z+

Solution. In what follows, we use the equation for Fourier transformation

given by

1
F (k) = dx eikx f (x) (1)
2

Part (a). We substitute the expression for f (x) given in the problem state-
ment in the equation for the Fourier transformation.

1 2
F (k) = dx eikx ex cos x,
2

1 2
dx ex +ikx eix + eix ,


=
2 2

1 h 2 2
i
= dx ex +i(k+)x + ex +i(k)x ,
2 2

1 h 2 k+ 2 k
i
= dx e(x i x) + e(x i x) ,
2 2
 
1 2 (k+)2 k 2 (k)2
(xi k+ ) (xi )
= dx e 2 e 4 + e 2 e 4 ,
2 2
r r 
1 (k+)2 (k)2
= e 4 + e 4 ,
2 2
2 2
e(k + )/4  k k

= e 2 + e 2 ,
2 2
2 2
e(k + )/4 k
F (k) = cosh . [10 pts] (2)
2 2

1
The last equality above is our sought-for equation for the Fourier transform.

Part (b). Here, we have a more complicated expression for f (x) due to its
poles at four values of x. We have

x4 + a4 = 0

x4 = a4 = ei a4
3 5 7
x = aei 4 , aei 4 , aei 4 , aei 4 . [4 pts] (3)
3
Note that the first two poles at x = aei 4 , aei 4 are located above the real
5 7
axis while the other two poles at x = aei 4 , aei 4 are located below the real
axis in the complex plane.
For simplicity, we assume that > 0 and a > 0 in the Fourier trans-
formation; the results for other combinations of the signs of and a follow
accordingly. We have

1 eikx cos x
F (k) = dx 4
2 a + x4

1 eikx ix ix


= dx 4 e + e
2 2 a + x4

1 ei(k+)x 1 ei(k)x
F (k) = dx 4 + dx (4)
2 2 a + x4 2 2 a4 + x4

We evaluate the two integrals on the right hand side using the method
of residues. In so doing, note that (ignoring the trivial case) there are three
possible combinations for the signs of k + and k in the argument of
the exponential function above. We have

(i) k + > 0 and k > 0

(ii) k + > 0 and k < 0

(iii) k + < 0 and k < 0

(iv) k + < 0 and k > 0 (not possible)

Consider case (i) and work on the complex plane. Let

1 ei(k+)z 1 ei(k)z
I1 = dz 4 and I2 = dz . (5)
2 2 C a + z4 2 2 C a4 + z 4
where the contour C shown in Fig. 1 is composed of a semicircle CR of
radius R and a horizontal line of total length 2R joining the two ends of

2
 Im

CR

Re

R R

Figure 1: Contour C for complex integration.

CR . By virtue of Jordans lemma, the integral over CR in I1 vanishes in the

limit R . As such,
R
1 ei(k+)x 1 ei(k+)z
lim I1 = lim dx 4 + lim dz ,
R R 2 2 R a + x4 R 2 2 CR a4 + z 4

1 ei(k+)x
lim I1 = dx 4 . (6)
R 2 2 a + x4
Following a similar logic, we find
R
1 ei(k)x 1 ei(k)z
lim I2 = lim dx 4 + lim dz ,
R R 2 2 R a + x4 R 2 2 CR a4 + z 4

1 ei(k)x
lim I2 = dx 4 . (7)
R 2 2 a + x4
It follows that
F (k) = lim (I1 + I2 ) . (8)
R

To find I1 we use the residue theorem. Note that the contour C in Fig.
1 in the limit of very large R encloses only two poles of the four poles,
3
namely, those at x = aei 4 , aei 4 . Then
ei(k+)z  ei(k+)z 
   
2i
lim I1 = Res 4 + Res
R 2 2 a + z 4 z=aei 4 a4 + z 4 z=aei 3
4
r  i(k+)z 
ei(k+)z 
 
e 
=i  +
2 4z 3 z=aei 4 4z 3 z=aei 3
4

3
 r  ia(k+)( 1 + i ) 1
ia(k+)( + i ) 
i e 2 2 e 2 2
lim I1 = 3i + 9i
R 4 2 a3 e 4 a3 e 4
r h
i ia(k+)( 1 + i ) 3i ia(k+)( 1
+ i ) i
i
lim I1 = 3 e 2 2 e 4 +e 2 2 e 4 (9)
R 4a 2
For brevity, denote

a(k + )
1 = . (10)
2
Then
r     
i i1 (1+i) 1 i i1 (1+i) 1 i
lim I1 = 3 e +e
R 4a 2 2 2 2 2
ie1
r     
1 i 1 i
= (cos 1 + i sin 1 ) + (cos 1 i sin 1 )
4a3 2 2 2 2 2
ie1 i cos 1 i sin 1 i cos 1 i sin 1
r  
=
4a3 2 2 2 2 2
e1
lim I1 = (cos 1 + sin 1 ) . (11)
R 4a3

To find I2 we likewise use the residue theorem. Since k > 0 (we are
covering case (i)), we can use the same contour as that for I1 . The same
calculation then leads to the last of (9) with k + being replaced by k .
r h
i ia(k)( 1 + i ) 3i 1
ia(k)( + i ) i
i
lim I2 = 3 e 2 2 e 4 + e 2 2 e 4 . (12)
R 4a 2
Denoting

a(k )
2 = . (13)
2
we find the last of (11) with I1 I2 and 1 2 .

e2
lim I2 = (cos 2 + sin 2 ) (14)
R 4a3

Combining our two results above for I1 and I2 , we find upon using (8)

 1
(cos 1 + sin 1 ) + e2 (cos 2 + sin 2 ) ,

F (k) = 3
e [10 pts]
4a

for k + > 0 and k > 0.

4
 For other cases, the calculations are essentially the same. We have the
following results:


F (k) = 3 e1 (cos 1 + sin 1 ) + e2 (cos 2 sin 2 ) ,

(15)
4a
for case (ii) namely, k + > 0 and k < 0,

 1
e (cos 1 sin 1 ) + e2 (cos 2 sin 2 ) ,

F (k) = 3
(16)
4a
for case (iii) namely, k + < 0 and k < 0.

Part (c). Here we are given


cos k0 x (|x| < N /k0 )
f (x) = (17)
0 (|x| > N /k0 ), N Z+
In the equation for Fourier transformation, we only need to perform the
integration over x such that |x| N /k0 since f (x) vanishes otherwise.

1
F (k) = dx eikx f (x)
2
N /k0
1
= dx eikx cos k0 x
2 N /k0
N /k0
1
= dx eikx (eik0 x + eik0 x )
2 2 N /k0
N /k0
1
dx ei(k+k0 )x + ei(kk0 )x
 
=
2 2 N /k0

ei(kk0 )x N /k0
 i(k+k0 )x 
1 e
= +
2i 2 k + k0 k k0 N /k0

iN kk iN kk iN k
iN k
eiN eiN k0 eiN k0 eiN
 
1 e 0 e 0 e e
= +
2i 2 k + k0 k + k0 k k0 k k0
 iN kk iN kk iN kk iN kk 
(1)N e 0 e 0 e 0 e 0
= +
2i 2 k + k0 k + k0 k k0 k k0
(1)N
 
2i k 2i k
= sin N + sin N
2i 2 k + k0 k0 k k0 k0

(1)N 2k k
F (k) = 2 2 sin N . [10 pts] (18)
2 k k0 k0

5
Problem 2. (Arfken 15.4.3) The one-dimensional neutron diffusion equa-
tion with a (plane) source is

d2 (x)
D + K 2 D(x) = Q(x), (19)
dx2
where (x) is the neutron flux, Q(x) is the plane source at x = 0, and
D and K 2 are constants. Apply Fourier transform. Solve the equation in
transform space. Transform back into x-space.

Solution. Based on the lecture, the Fourier transform of the derivative

term in the differential equation above is given by

F {00 (x)} = (ik)2 F {(x)} (20)

It follows that the differential equation can be written in k-space as

2 2 Q
k DF {(x)} + K DF {(x)} = dx eikx (x)
2
Q
k 2 DF {(x)} + K 2 DF {(x)} =
2
Q
F {(x)} = (21)
2D(K 2 + k 2 )

In x-space we have

Q eikx
(x) = dk . (22)
2D k2 + K 2

If x < 0, we can use the contour shown in Fig. 1 where the real axis and
imaginary axis are now real -axis and imaginary -axis, to define

Q eix
I= d 2 . (23)
2D C + K2
In this integral, the real part of is simply k. We can decompose it as
R
Q eikx Q eix
I= dk 2 + d (24)
2D R k + K 2 2D CR 2 + K 2

where the first integral on the right hand side is over the horizontal contour
on the real axis from k = R to k = R and the second integral is over CR .
By Jordans lemma, the integral over the semicircle CR vanishes in the limit
R . We find

lim I = (x), (x < 0). (25)

R

6
 The integral on the left hand side can be computed using the residue
theorem. There are two poles to consider namely, = iK. Without loss
of generality, we assume K > 0. Then for x < 0,

eix 

Q
lim I = (2i)Res 2
R 2D + K 2 =iK

iQ eix 

=
D 2 =iK
Q Kx
(x) = e , (x < 0) [10 pts] (26)
2DK

When x > 0 we can follow the same approach as that when x < 0 to
find (x). However, using the same contour shown in Fig. 1 is impractical;
the contour integral over CR is no longer vanishing in the limit R . (To
have a good insight into this, try taking the limit of eix for = i|| as the
magnitude || increases without bounds. For x < 0, it should tend to zero,
but for x > 0, it will blow up.) To remedy this problem, we use a slightly
different contour shown in Fig. 2 to define the integral

Q eix
I = d 2 . (27)
2D C + K2
Note that it is the same as (23) but with the reverse arrow; we now evaluate
it in a clockwise manner.
Decomposing the integral yields
R
Q eikx Q eix
I = dk 2 + d (28)
2D R k + K 2 2D CR 2 + K 2

The second integral now vanishes as R and we find

lim I = (x), (x > 0). (29)

R

Using the residue theorem, we get

ix 

Q e
lim I = (2i)Res 2 
R 2D + K 2 =iK

iQ eix 

=
D 2 =iK
Q Kx
(x) = e , (x > 0). [10 pts] (30)
2DK
Observe in the second line that there is a negative sign in front of the
exponential function reminiscent of the clockwise integration. Also, we have

7
 Im

R R Re

CR

Figure 2: Contour C for complex integration.

chosen the pole at = iK because it is the pole enclosed by our contour.

When x = 0 the integral for (x) given by (22) becomes

Q dk
(0) =
2D k + K 2
2

k 

Q
= atan 
2DK K
 
Q
=
2DK 2 2
Q
(0) = . [5 pts] (31)
2DK

Combining all our results for (x), we conclude

Q Kx


e x>0


2DK


Q Kx Q K|x|
(x) = e x<0 or simply (x) = e (32)
2DK
2DK

Q


x=0
2DK