Вы находитесь на странице: 1из 7

APPM 4360: Homework 4 Solutions 3/11/13

1. Obtain the radius of convergence of the following series.


a)

n=1
_
5 + 12i
3 + 4i
z
_
n
Use the ratio test:

b
n+1
b
n

5 + 12i
3 + 4i

|z| =
13
5
|z|
This is smaller than 1 if |z| <
5
13
= R.
b)

n=3
n!z
n
2

b
n+1
b
n

= (n + 1)|z|
2n+1
LHopitals rule shows that this limits to zero if |z| < 1. It clearly blows up if |z| > 1. So
R = 1.
Figure 1: Left picture shows contours of u (red) and v (black) for rst ve terms of the series for 1(b). Clearly
something strange is happening at the unit circle. The right gure shows the rst ten terms for 1(d). This series
converges for |z| < e, but the series becomes complicated near the boundary.
c)

n=1
z
n
n
n

b
n+1
b
n

=
1
n + 1
_
1 +
1
n
_
n
|z|
Since (1 + 1/n)
n
e
1
as n , this expression limits to 0 for any nite |z|. So R = .
d)

n=1
n!z
n
n
n

b
n+1
b
n

=
n + 1
n + 1
_
1 +
1
n
_
n
|z|
|z|
e
So R = e.
1
APPM 4360: Homework 4 Solutions 3/11/13
2. Obtain MacLaurin series for the following functions.
a)
z
1 + z
2
=
z
1 w
= z

n=0
w
n
=

n=0
(1)
n
z
2n+1
Radius of convergence is R = 1, since the rst singularity of 1/(1 + z
2
) is at z = i, and a
Taylor series is guaranteed to converge in a disk about a point of analyticity, out to the rst
singularity (VERY IMPORTANT THEOREM!!!)
b)
cosh(z + 1) =
1
2
(ee
z
+
1
e
e
z
) =
1
2

n=0
(e +
1
e
(1)
n
)
z
n
n!
Alternatively, we can use the sum formula for cosh to get (the same answer...)
cosh(z + 1) = cosh(1)

n=0
z
2n
(2n)!
+ sinh(1)

n=0
z
2n+1
(2n + 1)!
Radius of convergence is R = , since cosh is entire.
c)
(2 + 3z
2
)e
z
2
= (2 + 3z
2
)

n=0
z
2n
n!
= 2 +

n=0
(
2
(n + 1)!
+
3
n!
)z
2n+2
= 2 +

n=0
3n + 5
(n + 1)!
z
2n+2
=

m=0
3m + 2
m!
z
2m
Radius of convergence is R = since both (2 + 3z
2
) and e
z
2
are entire.
d) Using partial fractions (and the fact that we can add two uniformly convergent series,
term-by-term) gives
1
(1 + z)(2 z)
=
1
1 + z

1
2 z
=
_

n=0
z
n

1
2

n=0
z
n
2
n
_
=

n=0
_
(1)
n
+
1
2
n+1
_
z
n
This series converges out to the rst singularity of f(z), which is at z = 1, so R = 1.
e) Since cos(t
2
) is entire, its series converges uniformly for all t, and so we can do the inte-
gration term-by-term (VERY IMPORTANT THEOREM!!!). Thus
C(z) =
_
2

_
z
0
cos(t
2
)dt =
_
2

n=0
_
z
0
(1)
n
t
4n
(2n)!
dt =
_
2

n=0
(1)
n
(4n + 1)(2n)!
z
4n+1
This series is for an entire function, so R = .
3. Problem 3.2.3. Note that cosh(z) is entire, so 1/ cosh(z) is analytic except where cosh vanishes,
namely at z = i

2
+ ni for integer n. So the series, expanded around z = 0 will converge
2
APPM 4360: Homework 4 Solutions 3/11/13
for |z| < R =

2
it converges in a disk out to the rst singularity. Note that we CANNOT
gure this out by any other means, since an analytical formula for the E
n
is not available.
Compute the series by using the cosh series and a geometric series
1
cosh z
=
1
1 + z
2
/2 + z
4
/4! + . . .
=
1
1 + w
where w = z
2
/2 + z
4
/4! . . .. So
1
cosh z
=

n=0
(w)
n
= 1 w + w
2
+ . . . = 1 (z
2
/2 + z
4
/4! + . . .) + (z
2
/2 + z
4
/4! + . . .)
2
+ . . .
= 1
z
2
2
+ (
1
4

1
4!
)z
4
+ . . . = 1
1
2
z
2
+
5
4!
z
4
+ . . .
Thus we see that E
0
= 1, E
1
= 0, (indeed all odd E
n
= 0), E
2
= 1, E
4
= 5. The
computation becomes harder for n = 2m, since we have to multiply out each of the terms w
k
for k up to n to get all of the order z
2m
terms.
4. Find the Laurent series expansions.
a) This is really a Taylor series. We just have to shift the origin.
1
z
=
1
2 + (z 2)
=
1
2

n=0
(1)
n
_
z 2
2
_
n
=

n=0
(1)
n
2
n+1
(z 2)
n
b) Pull out a factor of z 2 from the denominator
1
z
=
1
z 2
1
1 +
2
z2
=
1
z 2

n=0
(1)
n
_
2
z 2
_
n
=

n=0
(2)
n
(z 2)
n1
c) This is the derivative of the previous result, and we can dierentiate term-by-term, since
the function is analytic in the domain of the series (VERY IMPORTANT THEOREM!!!)
1
z
2
=
d
dz
1
z
=

n=0
(2)
n
d
dz
(z 2)
n1
=

n=0
(n + 1)(2)
n
(z 2)
n2
=
2

m=
(m + 1)(2)
m
(z 2)
m
d) We can use the sin series and multiple through by the 1/z
2
term-by-term since the series
converges unformly
1
z
2
sin(z) =

n=0
(1)
n
n!
z
n2
=

m=2
(1)
m
(m + 2)!
z
m
5. Problem 3.3.4
a) For this we could just use Cauchys formula 2.6.5, to see that the result is
2i
2!
d
2
dz
2
e
z
|
z=0
.
But lets use series instead:
I =
_
C
e
z
z
3
dz =

n=0
1
n!
_
C
z
n3
dz
3
APPM 4360: Homework 4 Solutions 3/11/13
Thus the integrand has a triple pole at the origin. The integral
_
C
z
k
dz = 0 unless k = 1,
when it is 2i (VERY IMPORTANT THEOREM!!!). This is the term n = 2 in the sum
above, so I =
2i
2!
= i.
b) Now Cauchys formula does not work, and series is more fun anyway:
I =
_
C
1
z
2
sin(z)
dz =
_
C
1
z
3
1
1 z
2
/3! + z
4
/5! . . .
dz
Now the series will converge out to the rst singularity of 1/ sin(z), which is at z = .
So since the contour is inside of this, we can expand, using geometric series, and integrate
term-by-term. Isnt complex analysis wonderful!
I =
_
C
1
z
3
(1 + (z
2
/3! z
4
/5! . . .) + (z
2
/3! z
4
/5! . . .)
2
+ . . .)dz
Of all of the terms, only the z
1
power gives a nonzero integral. So we just have to pick out
this coecient, which comes from the z
2
/3! term. So I = 2i/3! =

3
i.
c) Note that tanh z is analytic inside the unit circle (its rst singularity is at z =

2
i. So
I = 0 by Cauchys theorem 2.5.2.
6. Problem 3.5.1
b)
f(z) =
e
2z
1
z
2
dz =
1
z
2
(2z + (2z)
2
/2! + (2z)
3
/3! . . .) =
2
z
+ 2 +
4
3
z + . . .
Thus f has a simple pole at the origin of strength 2. Now set z = 1/t to look near :
f(1/t) = t
2
(e
1/t
1) = t
2
+

n=0
1
n!
t
n+2
=

n=1
1
n!
t
2n
This has innitely many negative terms, so f has an essential singularity at .
c) e
z
is entire, and tan(z) is analytic except at z = z
n
=

2
+ n. So these are possible
singularities of our function too. Set z = + z
n
and expand for small . Note that
tan( + z
n
) =
sin( + z
n
)
cos( + z
n
)
=
(1)
n
cos()
(1)
n
sin()
=
1
2
/2! + . . .

3
/3! + . . .
=
1

(1
2
/2! + . . .)(1 +
2
/3! + . . .) =
1

+ . . .
Therefore near each z
n
, the function looks like e
1/
: it has an essential singularity at each
z
n
. Note that is a limit of these essential singularities. It too must be a very bad singular
point, an essential singularity.
d) We factor the denominator to obtain
f(z) =
z
3
z
2
+ z + 1
=
z
3
(z z
+
)(z z

)
=

+
(z)
z z
+
=

(z)
z z

where z

=
1

3
2
. Thus f(z) has simple poles at z

. The stengths are simply

+
(z
+
) =
z
3
+
z
+
z

=
i

3
, and

(z

) =
z
3

z
+
=
i

3
4
APPM 4360: Homework 4 Solutions 3/11/13
h) This series converges uniformly for |z| < 1, as can easily be seen from ratio test:

b
n+1
b
n

=
|z|
n(n!)
(n + 1)
0 if |z| < 1
So we suspect there are singularities somewhere on the unit circle (since otherwise the series
would converge in a larger disk). To see where, try to compute the derivative:
f

(z) =

n=1
z
n!1
This series also converges uniformly (as it must) when |z| < 1. However at z = 1 it is formally
innite, since this series becomes

1 when z = 1. Thus the derivative blows up at z = 1,


and we know that the series is valid when |z| < 1, so for our function we must have
lim
z1
f

(z) = .
In fact, the derivative is innite for every rational angle on the unit circle. For example,
suppose z = e
2ij/k!
for some integers j, k. Then every term in the series for f

beyond the
k
th
is equal to 1/z, because
_
e
2ij
k!
_
n!
= 1
since n!/k! is an integer whenever n k. So for such zs the series has an innite number
of terms that are identical and nonzero. So the derivative is innite at all of these rational
values of such points are dense on the unit circle. This is a natural barriera dense set of
singularities. (It is possible that the series converges on the unit circle for some irrational
relative to 2, but this is a very delicate question!).
Figure 2: First four terms in the series for f(z) and f

(z) in 6h. The computer cannot draw the contours when 5


or more terms are summed!
7. Problem 3.5.2
a) This is of the form of Cauchys integral. So
_
C
g(z)
z w
dz =
_

_
0 |w| > 1
2ig(w) |w| < 1
5
APPM 4360: Homework 4 Solutions 3/11/13
The value of the integral for |w| = 1 cannot be determined without more information on g.
For example if g(z) = z 1, then the integral converges when w = 1, but if g(z) = 1, then
the integral diverges for all points on the unit circle.
b) Here if |w| < 1 we have two simple poles, at z = w inside the contour. We can deform
the contour to be two small circles about the singularities, and then we have two forms of the
standard Cauchy integral:
_
C
z
z
2
w
2
dz =
_
C

(w)
z/(z + w)
z w
dz +
_
C

(w)
z/(z w)
z + w
dz
= 2i
_
w
2w
+
w
2w
_
= 4i
Note that I = 0 if |w| > 1, and it does not exist if |w| = 1.
c) For this integral, it is best to expand in a Laurent series
I =
_
C
ze
1/z
2
dz =

n=0
_
C
1
n!
z
z
2n
=
_
C
1
1!
1
z
= 2i
since, as usual only the term with z
1
has a nonzero integral.
8. Grad Student and Extra Credit problems.
3.2.10 Expanding using the binomial series gives
I =
_
C
(z + 1/z)
2n
dz
z
=
_
C
2n

k=0
_
2n
k
_
z
2n2k1
dz = 2i
_
2n
n
_
since only the term with k = n has a nonzero integral. Substitution of z = e
i
into the
integral and noting that z + 1/z = 2 cos(), gives
I = i
_
2
0
(2 cos())
2n
d = 4
n
i
_
2
0
(cos())
2n
d
which, combined with the previous answer, immediately simplies to the result.
3.3.5 Note that J
n
(t) = c
n
, the n
th
Laurent series coecient of the given function. Thus
if we integrate around the essential singularity at z = 0, we get
J
n
(t) =
1
2i
_
C
e
t
2
(z1/z)
z
n+1
dz
Now, let the contour be the unit circle, z = e
i
. Note that the exponent becomes it sin(),
and the denominator e
i(n+1)
, so we get
J
n
(t) =
1
2
_
2
0
e
it sin()
e
in
d
Since everything in the integrand is periodic as a function of , we can just as well
integrate from to . Then note that the exponential can be expanded into real and
imaginary parts, e
iw
= cos w i sin w, with w = n t sin(). Since w is an odd
function of , the sin w term is odd too, and the integral of that term vanishes. Thus
J
n
(t) =
1
2
_

cos(n t sin())d
But this is even, so integration over half the integral and multiplication by two gives the
same answer.
6
APPM 4360: Homework 4 Solutions 3/11/13
3.5.4b The ratio test for the series gives

b
n+1
b
n

=
n + 1
n + 2
|z|
n
|z| < 1
so that the radius of convergence is R = 1. Since the series converges uniformly when
|z| < 1, we can dierentiate term by term to get the series for its derivative
f

(z) =

n=0
z
n
=
1
1 z
So f

has a unique analytical continuation to 1/(1 z), having only a simple pole at
z = 1. Integrating, we have
f(z) f(0) =
_
z
0
f

()d =
_
z
0
d
1
= log(z 1) + log(1)
Note that f(0) = 0 from the original series. To get a branch that agrees with f when
|z| < 1, we have to put the branch cut somewhere outside the unit circle. Thus if we
take the principle branch of the log-function, letting z = 1 + re
i
, we get
f(z) = log(z 1) + log(1) = ln(r) i + i, 0 < 2.
Note that this function has a branch point at z = 1, and a branch cut along the ray
corresponding to = 0, which is {z = 1 +t : 0 < t < }. So this function is an analytic
continuation of f. (One could also choose a dierent branch cut, just as long as it stays
outside the unit circle).
7

Вам также может понравиться