c15.

qxd 7/12/11 3:02 PM Page 279

CHAPTER 15 Power Series, Taylor Series

Power series and, in particular, Taylor series, play a much more fundamental role in
complex analysis than they do in calculus. The student may do well to review what has
been presented about power series in calculus but should become aware that many new
ideas appear in complex, mainly owing to the use of complex integration.

SECTION 15.1. Sequences, Series, Convergence Tests, page 671

Purpose. The beginnings on sequences and series in complex is similar to that in
calculus (differences between real and complex appear only later). Hence this section
can almost be regarded as a review from calculus plus a presentation of convergence
tests for later use.
Main Content, Important Concepts
Sequences, series, convergence, divergence
Comparison test (Theorem 5)
Ratio test (Theorem 8)
Root test (Theorem 10)

SOLUTIONS TO PROBLEM SET 15.1, page 679

n
2. Converges to 0, hence bounded, since n! grows faster than ƒ 1 ⫹ 2i ƒ n ⫽ (15) .
Convergence implies that there are no limit points other than the limit itself.
Note that z n is the nth term of the Maclaurin series of e1⫹2i.
1
3. z n ⫽ ⫺ pi(1>1 ⫹ 2>4ni) by algebra; convergent to ⫺pi>4
4
4. Unbounded, hence divergent, ƒ z ƒ ⫽ 5n>2.
5. Bounded, divergent, ⫾1 ⫹ 5i.
6. Unbounded since cos 2npi ⫽ cosh 2np ⬎ n hence divergent.
7. Unbounded, hence divergent.
8. Divergent; all terms have absolute value 1, and a graph suggests that every point on
the unit circle is a limit point.
9. Convergent to 0, hence bounded.
10. Bounded, divergent, ⫾1> 12 ⫾ i, 0, 1, ⫺2
12. For any P ⬎ 0 there is an N ⫽ N(P) such that
ƒ z n ⫺ l ƒ ⬍ P>3, ƒ z n* ⫺ l* ƒ ⬍ P>3.
Hence by the triangle inequality, for all n ⬎ N we have
ƒ (z n ⫹ z n*) ⫺ (l ⫹ l*) ƒ ⫽ ƒ (z n ⫺ l) ⫹ (z n
* ⫺ l*) ƒ
⬉ ƒ z n ⫺ l ƒ ⫹ ƒ z n* ⫺ l* ƒ
⬉ P>3 ⫹ P>3
⬍ P.

279
c15.qxd 7/12/11 3:02 PM Page 280

280 Instructor’s Manual

This proof is typical of many similar ones.

Much less important is termwise multiplication of sequences, but a similar theorem
holds true for this case. Namely, under the assumptions just made on the convergence
of the two sequences, it follows that the sequence z 1z *1 , z 2z *2 , Á is convergent and
has the limit ll*. The proof is more complicated, as follows.
The two sequences are bounded, ƒ z n ƒ ⬍ K, ƒ z n* ƒ ⬍ K. Since they converge,
for an P ⬎ 0 there is an N ⫽ N(P) (such that ƒ z n ⫺ l ƒ ⬍ P>(3K), ƒ z n* ⫺ l* ƒ ⬍
P>(3 ƒ l ƒ ) (l ⫽ 0; the case l ⫽ 0 is rather trivial), hence

ƒ z n z n* ⫺ ll* ƒ ⫽ ƒ (z n ⫺ l )z n* ⫹ (z n* ⫺ l*)l ƒ

⬉ ƒ zn ⫺ l ƒ ƒ zn
* ƒ ⫹ ƒ zn
* ⫺ l* ƒ ƒ l ƒ

⬍ P>3 ⫹ P>3 ⬍ P (n ⬎ N ).

16. Convergent. In connection with MacLaurin series in Sec. 15.4 the sum will turn
out to be e20⫹30i, of absolute value 0.5 # 109.
18. Divergent
20. Divergent because
1
n⫹i 1 ⫹ i>n
` 2 ` ⫽ ` ` ⬎
1 3

3n ⫹ 2i 3n 1 ⫹ 2i>(3n 2) n

and the harmonic series diverges.

22. Divergent; 1> 12n ⫽ 1> 12 ⴢ 1> 1n; Series with 1> 1n is divergent, since 1> 1n ⬎ 1>n
for n ⫽ 2, 3, Á and the harmonic series diverges.
24. Divergent by the ratio test because

(3i)n⫹1(n ⫹ 1)!>(n ⫹ 1)n⫹1

` ` ⫽ ` `
z n⫹1
zn (3i)nn!>n n

nn
⫽ 3(n ⫹ 1)
(n ⫹ 1)n⫹1
n
⫽3a b
n
n⫹1

3
⫽ n
n⫹1
a n b
3 3
⫽ n : e ⬎ 1.
a1 ⫹ n b
1

25. Divergent
26. It is essential that, from some n on, the test ratio does not become greater than a fixed
q ⬍ 1 instead of coming arbitrarily close to 1, as is the case, for instance, for the
harmonic series, which diverges.
c15.qxd 7/12/11 3:02 PM Page 281

Instructor’s Manual 281

30. The form of the estimate of Rn suggests we use the fact that the ratio test is a
comparison test based on the geometric series. This gives

Rn ⫽ z n⫹1 ⫹ z n⫹2 ⫹ Á ⫽ z n⫹1 a1 ⫹ z ⫹ Áb,

z n⫹2 z n⫹3
⫹z
n⫹1 n⫹1

`z ` ⬉ q, `z ` ⫽ `z ` ⬉ q 2,
z n⫹2 z n⫹3 z n⫹3 z n⫹2
etc.,
n⫹1 n⫹1 n⫹2 z n⫹1

ƒ z n⫹1 ƒ
ƒ Rn ƒ ⬉ ƒ z n⫹1 ƒ (1 ⫹ q ⫹ q 2 ⫹ Á ) ⫽ .
1⫺q
For the given series we obtain the test ratio
1 n⫹1⫹i (n ⫹ 1)2 ⫹ 1
` ` ⫽
n n
ⴢ
2 n⫹1 n⫹i 2(n ⫹ 1) B n 2 ⫹ 1

1 n 4 ⫹ 2n 3 ⫹ 2n 2 1
⫽ ⬍ ;
2 B n ⫹ 2n 3 ⫹ 2n 2 ⫹ 2n ⫹ 1
4
2

from this with q ⫽ 12 we have

ƒ z n⫹1 ƒ ƒn ⫹ 1 ⫹ iƒ 2(n ⫹ 1)2 ⫹ 1
ƒ Rn ƒ ⬉ ⫽ ⫽ ⬍ 0.05.
1⫺q 2n(n ⫹ 1) 2n(n ⫹ 1)

Hence n ⫽ 5 (by computation), and

31 661
s⫽ ⫹ i ⫽ 0.96875 ⫹ 0.688542i.
32 960
Exact to 6 digits is 1 ⫹ 0.693147i.

SECTION 15.2. Power Series, page 680

Purpose. To discuss the convergence behavior of power series, which will be basic to
our further work (and which is simpler than that of series having arbitrary complex
functions as terms).
Comment. Most complex power series appearing in practical work and applications have
real coefficients because most of the complex functions of practical interest are obtained
from calculus by replacing the real variable x with the complex variable z ⫽ x ⫹ iy,
retaining the real coefficients. Accordingly, in the problem set we consider primarily power
series with real coefficients, also because complex coefficients would neither provide
additional difficulties nor contribute new ideas.
Proof of the Assertions in Example 6
苲 苲 苲 苲
R ⫽ 1>L follows from R ⫽ 1> l by noting that in the case of convergence, L ⫽ l (the
苲
only limit point). l exists by the Bolzano–Weierstrass theorem, assuming boundedness
n n
of {2 ƒ an ƒ }. Otherwise, 2 ƒ an ƒ ⬎ K for infinitely many n and any given K. Fix z ⫽ z 0
and take K ⫽ 1> ƒ z ⫺ z 0 ƒ to get
n
2 ƒ an(z ⫺ z 0)n ƒ ⬎ K ƒ z ⫺ z 0 ƒ ⫽ 1

and divergence for every z ⫽ z 0 by Theorem 9, Sec. 15.1.

c15.qxd 7/20/11 2:44 PM Page 282

282 Instructor’s Manual

Now, by the definition of a limit point, for a given P ⬎ 0 we have, for infinitely many n,
苲 n 苲
l ⫺ P ⬍ 2ƒ an ƒ ⬍ l ⫹ P;

hence for all z ⫽ z 0 and those n,

苲 n 苲
(*) (l ⫺ P) ƒ z ⫺ z 0 ƒ ⬍ 2 ƒ an(z ⫺ z 0)n ƒ ⬍ (l ⫹ P) ƒ z ⫺ z 0 ƒ .

The right inequality holds even for all n ⬎ N (N sufficiently large), by the definition of
a greatest limit point.
苲 n
Let l ⫽ 0 . Since 2 ƒ an ƒ ⭌ 0, we then have convergence to 0. Fix any z ⫽ z 1 ⫽ z 0.
n
Then for P ⫽ 1>(2 ƒ z 1 ⫺ z 0 ƒ ) ⬎ 0 there is an N such that 2 ƒ an ƒ ⬍ P for all n ⬎ N; hence
1
ƒ an(z 1 ⫺ z 0)n ƒ ⬎ Pn ƒ z 1 ⫺ z 0 ƒ n ⫽ ,
2n
and convergence for all z1 follows by the comparison test.
苲 苲
Let l ⬎ 0. We establish 1>l as the radius of convergence of (1) by proving
苲
convergence of the series (1) if ƒ z ⫺ z 0 ƒ ⬍ 1>l ,
苲
divergence of the series (1) if ƒ z ⫺ z 0 ƒ ⬎ 1>l .
苲 苲
Let ƒ z ⫺ z 0 ƒ ⬍ 1>l . Then, say, ƒ z ⫺ z 0 ƒ l ⫽ 1 ⫺ b ⬍ 1. With this and P ⫽ b>(2 ƒ z ⫺ z 0 ƒ ) ⬎
0 in (*), for all n ⬎ N,
n 苲
2 ƒ an(z ⫺ z 0)n ƒ ⬍ l ƒ z ⫺ z 0 ƒ ⫹ P ƒ z ⫺ z 0 ƒ ⫽ 1 ⫺ b ⫹ 12 b ⬍ 1.
Convergence now follows from Theorem 9, Sec. 15.1.
苲 苲
Let ƒ z ⫺ z 0 ƒ ⬎ 1>l . Then ƒ z ⫺ z 0 ƒ l ⫽ 1 ⫹ c ⬎ 1. With this and P ⫽ c>(2 ƒ z ⫺ z 0 ƒ ) ⬎ 0
in (*), for infinitely many n,
n 苲
2 ƒ an(z ⫺ z 0)n ƒ ⬎ l ƒ z ⫺ z 0 ƒ ⫺ P ƒ z ⫺ z 0 ƒ ⫽ 1 ⫹ c ⫺ 12 c ⬎ 1,
and divergence follows.

SOLUTIONS TO PROBLEM SET 15.2, page 684

ⴥ
1
4. a n (z ⫺ 4 ⫹ 3pi)n
n⫽0
6
1
6. Center ⫺1, radius of convergence because in (6)
2
2n 1
n⫹1
⫽ .
2 2

2n
7. p>4, ⬁
1
n⫹1
⫽
2 2
8. The center is pi. In (6) we obtain
n n>n! n n(n ⫹ 1) 1 1 1
⫽ ⫽ ⫽
(n ⫹ 1)n⫹1>(n ⫹ 1)!
: .
n⫹1 n
a n b a1 ⫹ n b
(n ⫹ 1)n⫹1 1 n e
c15.qxd 7/12/11 3:02 PM Page 283

Instructor’s Manual 283

9. ⫺i, 12
10. Center 2i, radius of convergence ⬁ because
n n
1>n n n⫹1
n⫹1 ⫽ (n ⫹ 1) a b ⫽ (n ⫹ 1) a1 ⫹ b
1
: ⬁.
1>(n ⫹ 1) n n

11. 0, 29
B 10
12. Center 0, radius of convergence 8
14. Center 0, radius of convergence ⬁ because by (6),

(⫺1)n
42n(n!)2
: ⫺16(n ⫹ 1)2 : ⬁.
(⫺1)n⫹1
42 n⫹2 ((n ⫹ 1)!)2

From Sec. 5.5 you see that this is the complex analog of the MacLaurin series of the
Bessel function J0 (x>2), so that is the MacLaurin series of J0 (z>2) for complex z, as
will follow in Sec. 15.4.
16. Center 0. From (6) we obtain the radius of convergence 2>27 because
(3n)!>(2n(n!)3) (n ⫹ 1)3 2
3 ⫽ 2 : .
(3n ⫹ 3)!>(2 n⫹1
((n ⫹ 1)!) ) (3n ⫹ 3)(3n ⫹ 2)(3n ⫹ 1) 27
17. 1> 13
18. Center 0, radius of convergence ⬁ . We mention that this is the series (36) of the error
function in App. 3.1, extended to a complex variable z. Formula (6) gives

(2n ⫹ 3)(n ⫹ 1)! (2n ⫹ 3)(n ⫹ 1)

⫽ : ⬁.
(2n ⫹ 1)n! 2n ⫹ 1
20. Team Project. (a) The faster the coefficients go to zero, the larger ƒ an>an⫹1 ƒ
becomes.
(b) (i) Nothing. (ii) R is multiplied by 1>k. (iii) The new series has radius of
convergence 1>R.
(c) In Example 6 we took the first term of one series, then the first term of the other,
and so on alternately. We could have taken, for instance, the first three terms of one
series, then the first five terms of the other, then again three terms and five terms,
and so on; or we could have mixed three or more series term by term.
(d) No, because ƒ 30 ⫹ 10i ƒ ⬎ ƒ 31 ⫺ 6i ƒ .

SECTION 15.3. Functions Given by Power Series, page 685

Purpose. To show what operations on power series are mathematically justified and to
prove the basic fact that power series represent analytic functions.
Main Content
Termwise addition, subtraction, and multiplication of power series
Termwise differentiation and integration (Theorems 3, 4)
Analytic functions and derivatives (Theorem 5)
c15.qxd 7/12/11 3:02 PM Page 284

284 Instructor’s Manual

Comment on Content
That a power series is the Taylor series of its sum will be shown in the next section.

SOLUTIONS TO PROBLEM SET 15.3, page 689

5. 4
6. ƒ z>(2p) ƒ 2 ⬍ 1 by integrating the geometric series. Thus ƒ z ƒ ⬍ R ⫽ 2p.
7. 5
1
8. 3 , where 1>n(n ⫹ 1) can be produced by two integrations of the geometric series.
9. 1> 13
10. The binomial coefficient
n(n ⫺ 1) Á (n ⫺ k ⫹ 1)
a b⫽
n
k k!
consists of the fixed k!, which has no effect on R, and factors
n(n ⫺ 1) Á (n ⫺ k ⫹ 1)
as obtained by differentiation. Since g(z>2)n has R ⫽ 2, the answer is 2.

11. 5
B2
12. ⬁, because 2n(2n ⫺ 1) results from differentiation, and for the coefficients without
these factors we have in the Cauchy–Hadamard formula
n
1>n n n⫹1
⫽a b (n ⫹ 1) : ⬁ as n : ⬁.
1>(n ⫹ 1)n⫹1 n

14. 1, by applying Theorem 3 to gz n⫹m.

3
15.
5
16. This follows from
f (z) ⫽ a0 ⫹ a1z ⫹ a2 z 2 ⫹ a3 z 3 ⫹ Á
⫽ f (⫺z) ⫽ a0 ⫺ a1z ⫹ a2 z 2 ⫺ a3 z 3 ⫹ ⫺ Á .
18. This is a useful formula for binomial coefficients. It follows from
p q

(1 ⫹ z)p(1 ⫹ z)q ⫽ a a b z n a a b z m
p q
n⫽0
n m⫽0
m
p⫹q
p⫹q
⫽ (1 ⫹ z) p⫹q
⫽ a a b zr
r⫽0
r

by equating the coefficients of z r on both sides. To get z nz m ⫽ z r on the left, we

must have n ⫹ m ⫽ r; thus m ⫽ r ⫺ n, and this gives the formula in the problem.
20. Team Project. (a) Division of the recursion relation by an gives
an⫹1 anⴚ1
an ⫽ 1 ⫹ an .
c15.qxd 7/12/11 3:02 PM Page 285

Instructor’s Manual 285

Take the limit on both sides, denoting it by L:

1
L⫽1⫹ .
L
Thus L2 ⫺ L ⫺ 1 ⫽ 0, L ⫽ (1 ⫹ 15)>2 ⫽ 1.618, an approximate value reached
after just ten terms.
(b) The list is
1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233.
In the recursion, an is the number of pairs of rabbits present and anⴚ1 is the number
of pairs of offsprings from the pairs of rabbits present at the end of the preceding month.
(c) Using the hint, we calculate
ⴥ ⴥ
(1 ⫺ z ⫺ z 2) a anz n ⫽ a (an ⫺ anⴚ1 ⫺ anⴚ2)z n ⫽ 1
n⫽0 n⫽0

where aⴚ1 ⫽ aⴚ2 ⫽ 0, and Theorem 2 gives a0 ⫽ 1, a1 ⫺ a0 ⫽ 0, an ⫺ anⴚ1 ⫺

anⴚ2 ⫽ 0 for n ⫽ 2, 3, Á . The converse follows from the uniqueness of a power
series representation (see Theorem 2).

SECTION 15.4. Taylor and Maclaurin Series, page 690

Purpose. To derive and explain Taylor series, which include those for real functions
known from calculus as special cases.
Main Content
Taylor series (1), integral formula (2) for the coefficients
Singularity, radius of convergence
Maclaurin series for ez, cos z, sin z, cosh z, sinh z, Ln (1 ⫹ z)
Theorem 2 connecting Taylor series to the last section
Comment
The series just mentioned, with z ⫽ x, are familiar from calculus.

SOLUTIONS TO PROBLEM SET 15.4, page 697

1 2 1 6
3. z ⫺ z ⫹ Á,R⫽ ⬁
2 48
ⴥ
4. (2 ⫹ z) a z 2n ⫽ 2 ⫹ z ⫹ 2z 2 ⫹ z 3 ⫹ 2z 4 ⫹ Á , R⫽1
n⫽0

1 1 4 1 8 1 12 4
5. ⫺ z ⫹ z ⫺ z ⫹ Á , R ⫽ 18
8 64 512 4096

1 1
6. ⫽ ⫽ 1 ⫺ 2iz ⫺ 4 z 2 ⫹ 8 iz 3 ⫹ 16 z 4 ⫺ 32 iz 5 ⫹ Á , R ⫽ 1>2
1 ⫹ 2iz 1 ⫺ (⫺2iz)

1 2 1 4 1 6
7. 1 ⫺ cos z ⫽ z ⫺ z ⫹ z ⫺ Á,R⫽ ⬁
2 24 720
c15.qxd 7/12/11 3:02 PM Page 286

286 Instructor’s Manual

1
8. 2 ⫺ 12 cos 2z ⫽ z 2 ⫺ 13 z 4 ⫹ 45
2 6 1
z ⫺ 315 z8 ⫹ ⫺ Á , R⫽⬁
z

9. 冮 a1 ⫺ t
0
2
⫹
2
t ⫹ Á b dt ⫽ z ⫺ 1>3 z 3 ⫹ 1>10 z 5, R ⫽ ⬁
1 4

10. The series is

2 3 22 23
f⫽z⫹ z ⫹ z5 ⫹ z7 ⫹ Á , R ⫽ ⬁.
1ⴢ3 1ⴢ3ⴢ5 1ⴢ3ⴢ5ⴢ7

It can be obtained in several ways. (a) Integrate the Maclaurin series2 of the integrand
termwise and form the Cauchy product with the series of ez . (b) f satisfies
the differential equation f r ⫽ 2z f ⫹ 1. Use this, its derivatives f s ⫽ 2( f ⫹ z f r ),
etc., f (0) ⫽ 0, f r (0) ⫽ 1, etc., and the coefficient formulas in (1). (c) Substitute
ⴥ ⴥ
f ⫽ a anz n and f r ⫽ a nanz nⴚ1 into the differential equation and compare
n⫽0 n⫽0

coefficients; that is, apply the power series method (Sec. 5.1).
z5 z9 z 13
12. z ⫺ ⫹ ⫺ ⫹⫺Á; R⫽ ⬁
2!5 4!9 6!13
z3 z5 z7
14. z ⫺ ⫹ ⫺ ⫹⫺Á; R⫽ ⬁
3!3 5!5 7!7
16. First of all, since sin (w ⫹ 2p) ⫽ sin w and sin (p ⫺ w) ⫽ sin w, we obtain all values
of sin w by letting w vary in a suitable vertical strip of width p, for example, in the
strip ⫺p>2 ⬉ u ⬉ p>2. Now since

sin a ⫺ iyb ⫽ sin a

p p
⫹ iyb ⫽ cosh y
2 2
and

sin a⫺ ⫺ iyb ⫽ sin a⫺

p p
⫹ iyb ⫽ ⫺cosh y,
2 2

we have to exclude a part of the boundary of that strip, so we exclude the boundary
in the lower half-plane. To solve our problem we have to show that the value of the
series lies in that strip. This follows from ƒ z ƒ ⬍ 1 and
3 3 3
` Re az ⫹ ⫹ Á b ` ⬉ `z ⫹ ⫹ Á ` ⬉ ƒzƒ ⫹
1 z 1 z 1 ƒzƒ
⫹Á
2 3 2 3 2 3
p
⫽ arcsin ƒ z ƒ ⬍ .
2
18. We obtain from the sum of the geometric series
ⴥ
1
⫽ ⫺i a i n(z ⫺ i)n
i(1 ⫺ i(z ⫺ i)) n⫽0

⫽ ⫺i ⫺ i ⴢ i(z ⫺ i) ⫺ i(⫺1)(z ⫺ i)2 ⫺ Á , R ⫽ 1.

⫹ 1>2i ⫺ 1>2 i(z ⫹ i) ⫹ a⫺ ⫹ 1>4 ib (z ⫹ i)2 ⫹ (z ⫹ i)3 ⫹ Á , R ⫽ 12

1 1 1
19.
2 4 4
c15.qxd 7/12/11 3:02 PM Page 287

Instructor’s Manual 287

20. We obtain

1 1
cos2 z ⫽ ⫹ cos 2z
2 2
1 1
⫽ ⫺ cos (2z ⫺ p)
2 2
2 4
42
⫽ c az ⫺ p b ⫺ az ⫺ p b ⫹ ⫺ Á d ,
1 4 1 1
R ⫽ ⬁.
2 2! 2 4! 2
1 1
21. ⫺1 ⫹ (z ⫺ p)2 ⫺ (z ⫺ p)4 ⫹ Á , R ⫽ ⬁
2 24
1 1 1
22. 1 ⫹ (z ⫺ pi)2 ⫹ (z ⫺ pi)4 ⫹ (z ⫺ pi)6 ⫹ Á , R⫽⬁
2! 4! 6!
1 3 5 3
23. ⫺ ⫹ 1>4 i(z ⫹ i) ⫹ (z ⫹ i)2 ⫺ 1>8 i(z ⫹ i)3 ⫺ (z ⫹ i)4 ⫹ i(z ⫹ i)5
4 16 64 64
⫹ Á,R⫽2
24. We obtain
ez(zⴚ2) ⫽ e(zⴚ1⫹1)(zⴚ1ⴚ1)
ⴚ1
2
⫽ e(zⴚ1)
1 ⴥ (z ⫺ 1)
2n
⫽e a n!
n⫽0

⫽ e a1 ⫹ (z ⫺ 1)2 ⫹ 2! (z ⫺ 1)4 ⫹ 3! (z ⫺ 1)6 ⫹ Á b ,

1 1 1
R ⫽ ⬁.

SECTION 15.5. Uniform Convergence. Optional, page 698

Purpose. To explain the concept of uniform convergence. To show that power series have
the advantage that they converge uniformly (exact formulation in Theorem 1). To discuss
properties of general uniformly convergent series.
Main Content
Uniform convergence of power series (Theorem 1)
Continuous sum (Theorem 2)
Termwise integration (Theorem 3) and differentiation (Theorem 4)
Weierstrass test for uniform convergence (Theorem 5)
The test in Theorem 5 is very simple, conceptually and technically in its application.

SOLUTIONS TO PROBLEM SET 15.5, page 704

2. R ⫽ 7, uniform convergence for ƒ z ƒ ⬉ 7 ⫺ d, d ⬎ 0

3. ƒ z ⫹ i ƒ ⱕ 15 ⫺d, d ⬎ 0
4. R ⫽ ⬁, uniform convergence on any bounded set
6. ƒ tanh n 2 ƒ ⫽ 1. Convergence for ƒ z 2 ƒ ⬍ 12 . Uniform convergence for ƒ z ƒ ⫽ 1> 12 ⫺ d,
d⬎0
7. Nowhere
c15.qxd 7/12/11 3:02 PM Page 288

288 Instructor’s Manual

8. R ⫽ 1> 13, uniform convergence for ƒ z ⫺ 1 ƒ ⬉ 1> 13 ⫺ d, d ⬎ 0

9. ƒ z ⫺ 2i ƒ ⱕ 3 ⫺d, d ⬎ 0
10. The MacLaurin series of sinh z converges for all z. Use Theorem 1.
12. ƒ z ƒ ⱕ 1, 1>(n3 sinh ƒ z ƒ ) ⱕ 1>n3 ⬍ 1>n2 and g 1>n 2 converges. Use the Weierstrass
M-test.
zn
` 2n ` ⬉ n ⬉ n and g 2ⴚn converges. Use the Weierstrass M-test.
1 1
14.
ƒzƒ ⫹ 1 ƒzƒ 2
16. ƒ tanh ƒ z ƒ ƒ ⬉ 1 (see App. 3.1), 1>(n(n ⫹ 1)) ⬍ 1>n 2 and g 1>n 2 converges. Use the
Weierstrass M-test.
17. R ⫽ 1> p ⬎ 0.25; Use Theorem 1.
18. Team Project. (a) Convergence follows from the comparison test (Sec. 15.1). Let
Rn (z) and Rn * be the remainders of (1) and (5), respectively. Since (5) converges,
for given P ⬎ 0 we can find an N (P) such that Rn * ⬍ P for all n ⬎ N (P). Since
ƒ fm(z) ƒ ⬉ M m for all z in the region G, we also have ƒ Rn(z) ƒ ⬉ Rn * and therefore
ƒ Rn(z) ƒ ⬍ P for all n ⬎ N (P) and all z in the region G. This proves that the conver-
gence of (1) in G is uniform.
(b) Since f0r ⫹ f1r ⫹ Á converges uniformly, we may integrate term by term, and
the resulting series has the sum F(z), the integral of the sum of that series. Therefore,
the latter sum must be F r(z).
(c) The converse is not true.
(d) Noting that this is a geometric series in powers of q ⫽ (1 ⫹ z 2)ⴚ1, we have
q ⫽ ƒ 1 ⫹ z 2 ƒ ⴚ1 ⬍ 1, 1 ⬍ ƒ 1 ⫹ z 2 ƒ 2 ⫽ (1 ⫹ x 2 ⫺ y 2)2 ⫹ 4x 2y 2, the exterior of a
lemniscate. The series converges also at z ⫽ 0.
(e) We obtain (add and subtract 1)
ⴥ ⴥ
⫽ x 2 a⫺1 ⫹ a b
1 1
x2 a
m⫽1 (1 ⫹ x ) 2 m
m⫽0 (1 ⫹ x 2)m
2
x
⫽ ⫺x 2 ⫹ ⫽ ⫺x 2 ⫹ 1 ⫹ x 2 ⫽ 1.
1
1⫺
1 ⫹ x2
20. We obtain
L

ƒ Bn ƒ ⫽
2
L
` 冮
0
f (x) sin
npx
L
dx ` ⬍ ML
2
L

where M is such that ƒ f (x) ƒ ⬍ M on the interval of integration. Thus

ƒ Bn ƒ ⬍ K (⫽ 2M).
Now when t ⭌ t 0 ⬎ 0,

ƒ u n ƒ ⫽ ` Bn sin ` ⬍ Keⴚlnt0
npx ⴚln2t 2
e
L

because ` sin ` ⬉ 1 and the exponential function decreases in a monotone fashion

npx
L
as t increases. From this,

` ` ⫽ ƒ ⫺l2nu n ƒ ⫽ l2n ƒ u n ƒ ⬍ l2nKe⫺lnt0

0u n 2
when t ⭌ t 0.
0t
c15.qxd 7/12/11 3:02 PM Page 289

Instructor’s Manual 289

Consider
ⴥ
2 ⴚlnt02

a lnKe .
n⫽1

cnp
Since ln ⫽ , for the test ratio we have
L
2 2
l2n⫹1K exp (⫺ln⫹1
2
n⫹1
⫽a b exp c ⫺(2n ⫹ 1) a b t0 d
t 0) cp
2 2 : 0
lnK exp (⫺lnt 0) n L

as n : ⬁, and the series converges. From this and the Weierstrass test it follows that
0u n
g
0u
converges uniformly and, by Theorem 4, has the sum , etc.
0t 0t

SOLUTIONS TO CHAPTER 15 REVIEW QUESTIONS AND PROBLEMS,

page 706

11. 1.
12. R ⫽ 14
13. 4
14. R⫽⬁
1
15. .
3
16. R ⫽ 1, ⫺ln (1 ⫺ z)
17. ⬁, exp(⫺2z).
18. R ⫽ ⬁, sin pz
19. ⬁, sinh 1z
ⴚ1
R ⫽ 5, a1 ⫺ b
z
20.
3 ⫹ 4i

⬁ (z 2)2 n
21. a n⫽0 ,R⫽ ⬁
(2 n)!

1 ⴥ
a n(n ⫺ 1)z R⫽1
nⴚ2
22. ,
2 n⫽2

⬁ (⫺1)n(z 2)2 n
23. a n⫽0
(2 n)!
ⴥ
24. a (⫺pz)n, R ⫽ 1> p
n⫽0

⬁ (z 2n⫺2)
25. a n⫽0 ⫺1
n!

26. [(z ⫺ i) ⫹ i]5 ⫽ (i ⫹ 5 (z ⫺ i) ⫺ 10i(z ⫺ i)2 ⫺ 10(z ⫺ i)3 ⫹ 5i(z ⫺ i)4 ⫹ (z ⫺ i)5),
a binomial expansion readily obtainable from Taylor’s theorem.
c15.qxd 7/12/11 3:02 PM Page 290

290 Instructor’s Manual

1
27. ⫺z ⫹ p ⫹ 1>6 (z ⫺ p)3 ⫺ (z ⫺ p)5 ⫹ Á
120
28. We obtain
1 1
⫽
z ⫺ 2i ⫹ 2i z ⫺ 2i
2i a1 ⫹ b
2i
ⴥ (⫺1)n
⫽ a n⫹1
(z ⫺ 2i)n
n⫽0 (2i)

⫽ ⫺ 12 i ⫹ 14 (z ⫺ 2i) ⫹ 18 i(z ⫺ 2i)2 ⫺ Á , R ⫽ 2.

ⴥ (z ⫺ pi)n
30. ezⴚpi⫹pi ⫽ ⫺ a , R ⫽ ⬁ ; here we used eⴚpi ⫽ ⫺1.
n⫽0
n!