Professional Documents
Culture Documents
0 take NV to be an integer
satisfying N > 1/e!/?. Then for n > N we have nP > NP > ie, and
1/n® < ¢._ To prove (1.2) formally, we would take NV to be an integer
satisfying N > (loge)/(log|s|)- To prove (1.8) formally, we resort to a
trick. Let tp = (ni — 1. We estimate t, from the binomial expansion
n(n
n= (14tq)" = 1+ nty + +>
2
th
‘This yields ty < 2/(n—1), Thus |tn| = | /7—1| <= whenever 2/(n—1) <
©, that is, for n > 142/e?, For the formal definition we can take N to be
any integer satisfying N > 1+ 2/e?
‘We give some definitions and state some theorems, without proofs, that
‘we will be using.
A sequence of complex numbers {sq} is said to be bounded if there is
some number R > 0 such that |sq| < R for all n. In other words, the
sequence is bounded if itis contained in some disk.
Theorem. A convergent sequence is bounded. Further, if {sq} and {tn}
are sequences of complex numbers such that sn —+ 8 and ty —t, then
(a) 8 + th 8 +t,
(b) ata st,
(6) Sn/tn — s/t, provided that t #0.
1. Review of Basie Analysis 35
‘Thus the limit of a sum is the sum of the limits, the limit of a product,
is the product of the limits, and the limit of a quotient is the quotient of
the limits, provided that the denominator is not 0.
Example. We can use these rules to evaluate the Ii
pression of the form
it of @ rational ex-
tim 3024 2n=1
nt Sn? —dn +8 5"
‘As a preliminary trick, we divide numerator and denominator by the lead-
Ing power, and rewrite the expression as
3+ (2/n) — 1/n?
B= Un) Bin
Since 1/m +0 and 1/n? —+0, the sum and product statements show that
the numerator converges to 3 and the denominator converges to 5. By the
qiotient statement, the quotient then converges to §
‘The most useful criteria for convergence of sequences of real and complex
numbers are gathered in the next several theorems. ‘The first criterion is
sometimes called the in-between theorem.
‘Theorem. If rq < sq Sty, and ifr,» L and ty —» L, then 5, — E.
A sequence of real numbers {sq} is said to be monotone increasing
if sa41 > mq for all n, monotone decreasing if sa41 0, we have ay.m — ay a8 m — oo.
7. The Zeros of an Analytic Function
Let f(z) be analytic at zo, and suppose that f(z9) = 0 but f(z) is not
identically zero. We say that (2) has a zero of order N at 29 if f(2o
$20) = = $"M (a9) = 0, while f""(z9) # 0. In view of the for-
‘mula (4.2) for the power series coefficients, this occurs if and only if the
power series expansion of f(z) has the form
S(2) = an (2 ~ 20)" + angale — 20) too
where ay #0. We can factor out the term (2— 29)" from the power series
and write
(7a) F(z) = (2-20) A(z),
where h(z) is analytic at zp and A(z) = ay # 0. Conversely, if there is
1 factorization (7.1) where h(2) is analytic at 20 and h(20) # 0, then the
7. The Zeros of an Analytic Function 155
leading term in the power series for f(z) is h(z9)(z ~ 20)", and f(z) has a
zero of order N at 2.
‘A zero of order one is called a simple zero, and a zero of order two is
called a double zero.
Example. The zeros of sinz are at the points nt, 00 0 such that f(z) is
analytic for |2| > R. Thus f(2) has an isolated singularity at oo if and
only if g(w) = f(1/w) has an isolated singularity at w = 0. We classify
the isolated singularity of f(2) at 00 according to the isolated singularity
of g(w) at w =0. Suppose that f(2) has a Laurent series expansion
Hz) = YO be2*, | >
‘The singularity of f(z) at co is removable if by = 0 for all k > 0, in which
case f(z) is analytic at oo. The singularity of f(z) at oo is essential if
by #0 for infinitely many k > 0. For fixed N’ > 1, f(z) has a pole of
order N at oo if by #0 while by = 0 for k > N.
Suppose f(z) has a pole of order N at oc. The Laurent series expansion
of f(z) becomes
Fle) = bu bye thet byt ht, (eID R,
where by # 0. We define the principal part of f(2) at oo to be the
polynomial
Plz) = bya® tye be thet by
‘The inclusion ¢f the constant bp in the principal part is a matter of conve-
nience. It guarantees that f(2) ~ P(2) is not only analytic at co but also
vanishes there.
Example. Any polynomial of degree N > 1 has a pole of order 1V at oo.
‘The principal part of the polynomial coincides with the polynomial itself.
Example. The function e = 1+2+22/21+--- has an essential singularity
at co.
4. Partial Fractions Decomposition 179
Exercises for VIS
1. Consider the functions in Exe:cise 1 of Section 2 above. Determine
‘which have isolated singularities at oo, and classify them.
2, Suppose that f(2) is an entire function that is not a polynomial.
What kind of singularity can f(2) have at oo?
3. Show that if f(z) is a noncorstant entire function, then e/(2) has
an essential singularity at z = co.
4. Show that each branch of the following functions is meromorphic
at co, and obtain the series expansion for each branch at oo.
(a) (2? =)? (b) (5 = 1) (Oy
4. Partial Fractions Decomposition
Proceeding in analogy with our earlier definition, we say that a functi
f(z) is meromorphic on a domain D in the extended complex plane C* if
(2) is analytic on D except possibly at isolated singularities, each of which
is @ pole. Again, sums and products of meromorphic functions are mero-
‘morphic. Quotients of meromorphic functions are meromorphic, provided
that the denominator is not identically zero.
‘Any rational function is meromorphic on the extended complex plane C*,
including at oo. We aim to establish the converse.
‘Theorem. A meromorphic function on the extended complex plane C* is
rational.
‘To see this, note frst that a meromorphie function f(2) on the extended
complex plane can have only a finite number of poles. Otherwise, they
would accumulate at a point thet would not be an isolated singularity of
f(z). If f(z) is analytic at oo, we define P,,(z) to be the constant function
Foo). Otherwise, #(2) has a pole at oo and we define Pu(z) to be the
Principal part of f(z) at co. In any event, Pa(2) is a polynomial, ond
f(z) — Pao(z) + 0 as 2 = 00. Let 21,... ,2m be the poles of f(z) in the
finive complex plane C, and let F(z} be the principal part of f(2) at ze
Te has the form
AG) = at Gap tt Goa
and in particular, P,(z) is analytic at oo and vanishes there. Consider the
function
9(2) = f(z) — Prol2)— > Py(2).
jt180 VI_ Laurent Series and Isolated Singularities
Since f(z) ~ P,(2) is analytic at ze, and each P,(z) is analytic at zx for
3 # k, g(2) is analytic at each 24. Hence g(z) is an entire function, and
Further, g(z) — 0 as 2 + co. By Liouville’s theorem, g(+) is identically
zero. Thus
(4) F(z) = Pool2)+ Do Pi(2),
which shows in particular that f(2) is a rational function.
The decomposition (4-1) is called the partial fractions decomposition
Of the rational function f(z). As a byproduct of the proof we obtain the
following,
‘Theorem. Every rational function has a partial fractions decomposition,
expressing it as the sum of a polynomial in = and its principal parts at each
of its poles in the finite complex plane.
Suppose that p(2) and q(2) are polynomials. If the degree of q(2) is
strictly less than the degree of p(2), then p(z)/q(2) ~ 0 as 2 —+ oo. Thus
(2) = p(2)/a(2) is analytic at co and vanishes there, and the principal part
Pao(2) is zero. Formula (4.1) expresses f(2) as the sum of the principal
parts at each ofits finite poles.
Example. The function 1/(2? ~ 1) is analytic at oo and vanishes there,
‘and it has poles at +1. The partial fractions decomposition is
1 riooid
Qe
and these summands are the principal parts at +1 and —1, respectively.
For arbitrary polynomials p(=) and q(z), we can use the division algo-
rithm to find the principal part P,.(z) of p(2)/q(2) at co. The division
algorithm is a procedure that produces in a finite number of steps polyno-
mials P,,(z) and r(2) that satis
(42) plz) € Palz)ale) +r(2), deg r(2) < deg a(2)
It proceeds as follows. We assume that 4(2) is 8 monic polynomial of degree
rm, #0 that q(z) — 2" +--+. We start with a polynomial p(2) of degree
12> m, say plz) = e92"-+-- For the frst step we kill the top coefficient of
(=) by defining py(z) = pz) —co2"-"q(z). Thus py[2) has degree m 0 such that |f(z)| > 4 on the circle |z — zo| = p. Since fe()
‘converges uniformly to f(z) for |z—zo| < p, for k large we have | fx(2)| > 6/2
for |2 —29| = p, and further, £,(2)/fu(z) converges uniformly to f'(2)/F(2)
for |z — zo| = p. Hence the integrals converge,
1 Hy 1 fie)
Bi Siem sstay 2) BS. ste FE)
Now, the integral on the let is the number Nj of zeros of fe(z) in the disk
{\2— 20] 0, so the zeros of f(z) must accumulate at zo.
de232 VILL The Logarithmic Integral
An alternative proof of Hurwita’s theorem can be based on Rouché's
theorem, applied to the “little” function fi(2)— f(z) and the “big” function
f(z). See Exercise 2.
‘We say that a function is univalent on a domain D if it is analytic and
‘one-to-one on D. In other words, the univalent functions on D are the
‘conformal maps of D to other domains. If we apply the preceding theorem
to univalent functions, where NV = 1, we obtain the following, which is
sometimes also referred to as Hurwitz’s theorem,
‘Theorem. Suppose {fx(2)} is a sequence of univalent functions on a do-
‘main D that converges normally on D to a function f(z). Then either f(z)
is univalent or f(z) is constant.
Indeed, suppose the limit function f(z) is not constant. Suppose =
and Go satisfy f(co) = F(Co) = uo. Then zo and Gp are zeros of finite order
for {(2) ~ wo. By the preceding theorem, there are sequences 2 —> 29 and
Ge Go stich that f(z) ~ wo = 0, and fe(Ge) ~ wo = 0. Since fy(z) is
tunivalent, we have 24 = Ck, and in the limit 9 = Go. Thus f(2) is univalent.
Example. For each k > 1, the fonction f,(z) = 2/K is univalent on the
complex plane. The sequence {f,(2)} converges normally to the constant
function f(z) = 0. Thus a normal limit of univalent functions need not be
univalent, and both alternatives of the theorem occur.
Exercises for VIILS
1. Let {fx(z)} be a sequence of analytic functions on D that converges
normally to f(z), and suppose that f(=) has a zero of order Nat
39 € D. Use Rouché's theorem to show that there exists p > 0 such
that for k large, f,(2) has exactly N zeros counting multiplicity on
the disk {lz ~ 291 = 0, and f(2) attains the value 20
'm times at 29 if 2 is a pole of f(z) of order m.
‘The number of times that f(2) attains a value w» on D is obtained by
adding the number of times it attains the value wp at the various points
of D.
Example. The polynomial 2 + 1 attains the value w
2 =0, and it attains the value w = co m times at z= 00.
sm times at
Example. A polynomial f(z) of degree m > 1 attains each value w € C* m
times on C*, always counting multiplicity. In fact, for fixed finite w, f(2)~w
is a polynomial of degree mn, which has m zeros counting multiplicity. Since
f(z) has a pole of order m at oo, it attains the value w = oe m times at
We can use the logarithmic integral to study the dependence on w of
the numberof times that an analytic or meromorphic function attains
the value w. ‘The technique is the same as that used in the preceding
section, except that instead of a sequence of functions we treat functions
that depend continuously on a parameter. The basic idea is as follows
Let f(z) be a nonconstant analytic function on a domain D. Let zy € D.
‘wo = f (20), and assume that f(z) ~ wo has a zero of order m at zo. Since
the 2eros of f(z)w are isolated, we can choose p > Oso that f(2)—wo #0
for 0 < |z — z9| < p. Let 6 > 0 satisfy |f(z) ~ wo| > 6 for |z — zo| = p. The
integral
v 1 f(z) 1
4a) Nw) = LO, w—wsl cd
(1) NC) = FE I nten Taye wml <
is then define, and it depends analytically oa w. Since N(u) i the umber
‘of zeros of f(z) — w in the disk {|= ~ 29| 0 be chosen such that |f(z) — wo| > 6 for
|z—29| =p. Then for each w such that jw —wol <6, there is a unique z
satisfying |2 ~ 2|
where ,... 29 ar the zeros and poles of f(2), and m, isthe order
of f(z) at
2. Let f(z) be a meromorphic function on the complex plane that is
doubly periodic. Suppose that the zeros and poles of f(z) are at
the points 2,... ,2, and at their translates by periods of f(2), and
Exercises 235
suppose no z; is a translate by a period of another 2. Let my be
the order of f(z) at 2). Show that >mjz, is a period of f(z). Hint.
Integrate zf'(z)/f(2) around the boundary of the fundamental par-
allelogram P constructed in Section VIS.
3. Let { fx(z)} be a sequence of analytic functions on a domain D that
converges normally to f(z). Suppose that f(z) attains each value w
‘at most m times (counting multiplicity) in D. Show that either f(2)
is constant, or f(z) attains each value w at most m times in D.
4, Let f(z) be an analytic function on the open unit disk D = {|2| <1}.
Suppose there is an annulus / = {r < |2| < 1} such that the
restriction of f(z) to U is one-to-one. Show that f(z) is one-to-one
on D.
5. Let f(z) = p(z)/q(z) be a rational function, where p(z) and q(2)
are polynomials that are relatively prime (no common zeros). We
define the degree of f(z) to be the larger of the degrees of p(2)
and q(2). Denote the degree of f(z) by d. (a) Show that each value
w eC, w# f(co), is assumed d times by f(z) on C. (b) Show that
J (2) attains each value w € C* d times on C* (as always, counting
multiplicity).
6. Let f(z) be a meromorphic function on the complex plane, and
suppose there is an integer m such that f~"(w) has at most m
points for all w € C. Show that f(2) is a rational function,
7. Let F(z,w) be a continuous function of z and w that depends an-
alytically on z for each fixed w, and let F,(z,w) denote the de-
tivative of F(z,w) with respect to z. Suppose F(z, wo) = 0, and
Fi(z0,10) # 0. Choose p such that F(2, 109) # 0 for 0 < |2~z9) < p.
(a) Show that there exists 6 > 0 such that if |w —wo| < 6, there is
a unique 2 = g(w) satisfying |2 ~ z9| < p and F(z,w) = 0.
(b) Show that
te) = CFG.) a
0) = ef Fa, tw wol <8
(€) Suppose further that F(z,.) is analytic in w for each fixed =,
and let Fa(z,w) denote the derivative of F(z, w) with respect
to w. Show that g(w) is analytic, and
gw) = —Fa(g(w),w)/Fi(g(w),w).
(€) Derive the inverse function theorem given in this section, to-
gether with the usual formula for the derivative of the inverse
function, asa corollary of (a), (b), and (¢).236 VIII The Logarithmic Integral
Remark. This is the implicit function theorem for analytic
functions. Note that a specific formula is given for the function
o(w) defined implicitly by F(g(w),
8, Let D be a bounded domain, and let f(2) be a continuous function
on DUD that is analytic on D. Show that (f(D)) © f(OD),
that is, the boundary of the open set. f(D) is contained in the image
under f(2) of the boundary of D.
5. Critical Points
Tet f(z) be a nonconstant analytic function on a domain D. A point =o is
called a critical point of f(2) if f"(s9) — 0. The value (20) ~ wo is called
acritical value of f(2). We define the order of the critical point zo to be
the order of 2er0 of #"(z) at 29. Since the eritical points are the zeros of the
nonconstant analytic function f"(z), critical points are isolated. Our aim is
to understand the behavior of an analytic function near a critical point, and
to understand the behavior of the inverse function near the corresponding
critical value.
Example. The function f(z) = 2" +c has a critical point of order m—1
at 2 = 0, with critical value c. ‘The inverse function for f(z) is given by
the m branches of (w ~ ¢)!/".
Suppose zo is a critical point of f(2) of order m —1 with critical value
‘wo. Then f(z) — wo has a zero of order m at 29, and we can proceed as
before by considering the logarithmic integral (4.1). This time the constant
value of N(w) is m. Thus for p > 0 small and for w near wo, there
are m points 21(tt),.-. .2m(w), repeated according to multiplicity, such
that f(z) attains the value w in the disk {\z ~ 20] < p} precisely at the
points 2(w),... ,2m(t)- We wish to see how the points z;(w) depend
on w. To do this, we will use a simple procedure, based only on the
existence of an analytic inverse for an analytic function at a noneritical
point, The procedure is to make a change of variable in order to show that
the behavior of f(2) near 29 is effectively the same as the behavior of the
analytic function ¢” at the critical point ¢ = 0. We proceed as follows.
ae an aL
Since f(z) uy has order m at zo, we can factor f(z)—w = (2~20)"A(=)-
where A(2) is analytic at zo, and (zo) #0. Near zo we define an analytic
5. Critical Points 2ar
branch of A(z)", and then for o(=) = (2 — zo)h(z)!/" we have
Sl) wo = 92)", Sle) = 9(2)" + wo
Moreover, (:) analyte at so and has simple zero there, 0 (24) #0
tnd 2) inonetoone nea ip, Thus J) ls reprecnted asthe compan
of three functions, the unvaleatfneiong() followed bythe map ¢>>
flowed by a translation € "> € sy. Since we understand the power
function ¢ = (andthe trensation w= € + wy very well and since (2)
isunivalet near 2p, we can dra from this aca pctue ofthe beevor
of fe). For instance, the st where i teal consist of stright ine
Segments passing through 0, dividing the plane near ¢ = Onto 2m sectors
of aperture =/m, on which Im(¢") is alternately positive and negative. The
txt hore f(2) “uy seal athe mage under 4 (¢) of thee ne egrets
tnd ence const of canals curves through. They dive the
plane ner» ito 2m pleslice domains with vertex at zo and aperture
z/m,on which m(f(2)~ ty) i aernately postive and negative
A point w near wo, w 4 wo, has m distinet preimages 21(w),...,2m(W).
They are the m branches of (w ~ wo)"/" composed with g"i(¢). If we
make a branch cut and consider the principal branch (w — wo)!/"" on the
slit disk {lw — wo] < 5}\(wo ~ 6, wo}, the other branches are of the form
29/™(y —~ ug)", and the preimages of w are given by the composition
of g-'(¢) and these branches, -
200) = 97* (e297 —w)"), jem
In particular, for w # wo, the preimages =;(w) are distinct, and each =)(w)
depends analytically on w. The values of 24(t) at the “top edge” of the
coincide with the values af z;+1(w) at the “bottom edge” of the sli,
s0 that 2(w) is continued analytically to =,.,(w) when w follows a circle
around 9 in the positive direction. In the same way that we constructed &
Riemann surface for the inverse function of 6", we can construct a Riemann
surface for the inverse function 2(w) of w = f(z). We create m copies of
the slit disk, and we define z(w) to be =,(w) on the jth copy. Then we glue238 VILL ‘The Logarithmic Integral
Sti >
together the edges of the branch cuts so that the function (w) becomes
continuous on the surface.
Now consider the function
(2 21(w)) ++ (= zal) = 2" + am aw)2"* + ay (w)2 + a0(w),
where
ag(w) = (—1)"21(t) + 2m(w),
am-s(te) = ~ > ay(u)
If we continue ax(w) around a circle centered at arp, the values of the
(ws are permuted and we return to the same value for ay(w). ‘Thus
u(t) is analytic in the punctured disk 0 < r—wol < ¢. Evidently, ox(t)
is bounded as w — wo. By Riemann’s theorem on removable singularities,
ag(w) is analytic at wy. In fact, the value ay(up) is just the coefficient
of 2* in the expansion of (z — zo)” in powers of z, Thus we have found a
polynomial equation
A) Fam alw)™ 1 tan(w)z baa(w) = 0,
with analytic coefficients ap(w),... ,am-i(0}, that reduces to the equa-
tion (z ~ Z9)!" = 0 at w = wp and that has solutions given precisely by
ai(w),--+ .2m(w),
‘We summarize our results as follows.
Theorem. Let f(z) be analytic at 2p. Suppose 2 is a critical point of
order m1 for f(z), with cttical value (29) = up. Let p > 0 satisfy
F(2) # wo for 0 < |7 ~ 70] 0 satisfy |f(z) — wol > 6 for
>. Then for each w such that 0 < |w ~ wol < 6, the equation
F(>) = w has exactly m distinct solutions =1(w),... ,2m(te) in the disk
{ls ~ 20] < p). The functions zi(w).-.. .2m(t) can be chosen to depend
analytically on w in the slit disk {lw — wol < 8}\(up ~ 5,wo). They form
the m branches of an analytic function = 2(w) on an m-sheeted Riemann
le ~ 201
Exercises 239
surface over the punctured disk {0 < |w ~ wol < 6}. The z,(w)'s are the
solutions of a polynomial equation (5.1) with coefficients a,(w) that are
analytic for |w ~ wo| < 6,
We have dealt with the solutions of the equation f(=) — w = 0 near a
solution point (29,9). A similar state of affairs holds for the solutions of
fan equation F(2,w) = 0, where F(z,w) depends analytically on both 2
and w in a neighborhood of a solution point (29,1), though in this case
the solution set is a finite union of surfaces passing through (29,1). This
is the simplest example of an analytic variety, which is defined to be the
set of common zeros of a family of analytic functions on some domain in
complex n-space.
Exercises for VIS
1, Find the critical points and critical values of f(z) = 2-+1/z. Sketch
the curves where f(2) is real. Sketch the regions where Im f(z) > 0
and where Im f(z) <0.
2, Suppose g(z) is analytic at z = 0, with power series 9(2) = 2+i=4+
O(z*).. Sketch and label the curves passing through = = 0 where
Reg(z) = 2 and where Im (z) = 0.
3, Find the critical points and critical values of f(z) = =? +1. Sketch
the set of points such that |f(2)] <1, and locate the critical points
of f(z) on the sketch.
4. Suppose that f(z) is analytic at zo. Show that if the set of =
such that Ref(z) = Ref(zo) consists of just one curve passing
through zo, then J"(20) #0. Show also that ifthe set of = such that
L/(2)| = |f(20)| consists of just one curve passing through zo, then
S'(2) £0.
5. How many critical points, counting multiplicity, does a polynomial
of degree m have in the complex plane? Justify your answer.
6, Find and plot the eritical points and critical values of f(z) = 2? +1
and of its iterates f(f(z2)) = (2?+1)?+1 and f(f(f(2)))- Suggestion.
Use the chain rule.
7. Let f(z) be a polynomial of degree m. How many (finite) critical
points does the V-fold iterate fo---0 f (N times) have? Describe
them in terms of the critical points of (2).
8. We define a pole 29 of f(z) to be a critical point of f(z) of order k
if sy is a critical point of 1/f(2) of order k. We define z — 00 to be
a critical point of /(=) of order & if w ~ 0 is a critical point of
g(w) = f(1/w) of order k. Show that with this definition, a point240
10.
nL.
12
13,
ua,
VIII The Logarithmic Integral
2 € C" is a critical point of order k for a meromorphie function (=)
if and only if there are open sets U containing 29 and V containing
wp = f(z) such that each w € V, w # wo, has exactly k +1
preimages in U. Remark. We say that f(z) is a (k + 1)-sheeted
covering of f-"(V\{wo}) VU over V\{wo}
Show that a polynomial of degree m, regarded as a meromorphic
function on C*, has a eritieal point of order m ~ 1 at 29 = co,
Locate the critical points and critical values in the extended complex
plane of the polynomial f(2) = 24 — 22%. Determine the order of
each critical point. Sketch the set of points > such that Im f(z) > 0.
Show that if f is a rational function, and if g is a fractional linear
transformation, then f and go f have the same critical points in the
extended complex plane C*, What can be said about the critical
values of g of? What can be said about the critical points and
critical values of f 0.9?
Let f(2) = p(2)/a(2) be a rational function of degree d, so that p(2)
and 4(2) are relatively prime, and d is the larger of the degrees of,
p(2) and g(2). (See Exercise 4.5.) Show that f(z) has 2d—2 critical
points, counting multiplicity, in the extended complex plane C*
Hint. If deg # deg. then the numberof critical points of (2) in
the finite plane C is deg(qp! — 4'p) = degp + degq ~ 1, while the
‘order of the critical point at oo is | deg p ~ deg — 1.
Show that the set of solution points (w,2) of the equation 2? —
2(costw)2+1 = 0 consists of the graphs of two entire functions = (w)
and 22(w) of w. Specify the entire functions, and determine where
their graphs meet. Remark. The solution set forms a reducible
‘one-dimensional analytic variety in C2
Let ao(w),--. .am—1(t) be analytic in a neighborhood of w = 0
and vanish at w = 0. Consider the monic polynomial in z whose
coefficients are analytic functions in w,
Pl2,w) = 2 +am-a(w)2""1 +---+ag(t), rol <6.
Suppose that for each fixed w, 0 < |u| < 6, there are m distinct
solutions of P(z, ) = 0.
(a) Show that the m roots of the equation P(:,w) = 0 deter-
mine analytic functions =(w),...,%(w) in the slit disk
{lvl < 6\(-6.0]. Hint. Use the implicit function theorem
(Exercise 4.7).
(b) Glue together branch cuts to form an m-sheeted Riemann sur-
face over the punctured disk on which the branches =,(w) de-
termine a continuous (single-valued, analytic) function.
Exercises 2a
(6) Suppose that the indices are arranged so that for some fixed k,
1 k 0 and 0 < 9 <1, there are constants C(a, 9) > 0
with the following property. If ¥ is a family of analytic functions on
the open unit disk that omits the values 0 and 1, and if |f(0)| 1. Show that {f,(2)} is a normal family on the
punctured disk if and only if the singularity of f(z) at 0 is removable
or a pole,
5. Let Eo,£1,£a be three disjoint compact subsets of the Riemann
sphere, and let F be a family of meromorphic functions on a do-
main D such that each f € F omits at least one point of each of
the three sets. Show that F is a normal family.
6. Let G be the family of univalent analytic functions on a fixed domain
D. (a) Show that ¢ is not normal. (b) Show that the family of
functions in G that omit 0 is normal. (c) Show that the family of
t for infinitely many indices n. It is easy to see that
any such S is unique. The existence of such an S can be deduced from the
preceding theorem. In fact, the existence is equivalent: to the completeness
axiom of the real numbers.36 TI Analytic Functions
‘A “lower limit” of the sequence {sq}, denoted by liminf sq, is defined
similarly. It satisfies
liminf s, =~ limsup(—sn).
‘The sequence {s,,} converges if and only if its lim sup and lim inf are finite
and equal.
Example. The sequence {(-1)"}2o = {+1,
converge. Its upper and lower limits are
limsup (~1)" = +1, limit (—1)"
so Iy...} does not
For complex sequences, the following simple criterion is used very often
‘Theorem. A sequence {sx} of complex numbers converges if and only if
the corresponding sequences of real and imaginary parts of the s,'s con-
verge.
We define a sequence of complex numbers (s,} to be a Cauchy se-
quence if the differences — sm tend to 0 a8 m and m tend to oo. In
the language of formal mathematics, this means that for any © > 0, there
exists V > 1 such that |&n ~ Sm| << if m,n > N. The following theorem
is an equivalent form of the completeness axiom. Itis important because it
provides a means of determining whether a sequence is convergent without
producing explicitly the limit of the sequence.
‘Theorem. A sequence of complex numbers coaverges if and only if itis a
Cauchy sequence.
We say that a complex-valued function f(z) has limit L as 2 tends
to 2 if the values f(2) are near L whenever 2 is near zp, 2 # z. The
formal definition is that f(z) has limit L as z tends to zo if for any € > 0,
there is 6 > 0 such that [f(2) — Z| < © whenever = in the domain of f(2)
satisfies 0 < [2 ~ zo] <6. In this case we write
an 10) = 1
cx (0) Lae + tis imply unerdod that there ae pints
in the domain of f(z) that are arbitrarily close to 29 and different from zo.
‘The definition can be rephrased in terms of convergent sequences.
Lemma. The complex-valued function f(2) hes limit L as 2 —» zy if and
only if f(z) — L for any sequence {2} in the domain of f(z) such that
sn #20 and 29 — Zo
From the theorem on limits, we obtain easily the following.
1, Review of Basie Analysis 37
Theorem. If a function has a limit at 29, then the function is bounded
near zo. Further, if f(2) + L and (2) > M as 2 — 2, then as 2 + 2 we
have
(a) f(2) + 92) + b+ M,
(b) f(z)g(z) > LM,
(c) f(z)/g(z) + L/M, provided that M #0.
We say that f(z) is continuous at 29 if f(z) > f(z0) as 2 + 2.
A continuous function is a function that is continuous at each point
of its domain. The preceding theorem shows that sums and products of,
‘continuous functions are continuous, and so are quotients, provided that the
denominator is not zero. Further, the composition of continuous functions
Js continuous.
Example. Any constant function is continuous. The coordinate function
f(2) = 7 is continuous. ‘Thus any polynomial function p(2) = 492" +
“+412 + a9 is continuous. Any rational function p(2)/4(2) is continuous
wherever the denominator q(2) is not zero.
‘A useful strategy for showing that (2) is continuous at 2o is to obtain
an estimate of the form |f(z) ~ f(zo)| < Olz ~ 20| for z near zo. This
guarantees that |f(2) — f(20)| <€ whenever |2~ z9| < €/C, so that we can
take 5 = 0, So that itis an open set. Other examples of open sets38 TT Analytic Functions
described by strict inequalities are the open sector {8 < arg z < 61}, the
‘open horizontal strip {1 < Im < 11}, the open annulus {r < |s| < #},
and the punctured disk {0 < |2| 0
(b) zy + 3x7y—y? (d)e*—¥" cos(2ey) (F) x/(a? + 9°)
2. Show that if v is a harmonic conjugate for u, then —u is a harmonic
conjugate for v.
3, Define u(z) = Im(1/22) for z #0, and set u(0) = 0.
(a) Show that all partial derivatives of u with respect to x exist at
all points of the plane C, as do all partial derivative of u with
respect to y.
Pu Pu
(b) Show that 28 + =,
(€) Show that 1 is not harmonic on .
Pu
(d) Show that: Bxdy
4. Show that if h(2) Ba complex-valued harmonic funetion (solution
of Laplace's equation) such that 2h() i alo harmonic, then (2)
is analytic
does not exist at (0,0)
5. Show that Laplace's equation in polar coordinates is
a 100, 1 0u
a Or * aR
0.58 II Analytic Functions
6. Show using Laplace's equation in polar coordinates that log|2| is
harmonic on the punctured plane C\{0}.
7. Show that log|z| has no conjugate harmonic function on the pune-
tured plane C\{0}, though it does have a conjugate harmonie func-
tion on the slit plane C\(—20, 0).
8. Show using Laplace's equation in polar coordinates that u(re'®) =
Glogr is harmonic. Use the polar form of the Cauchy-Riemann
equations (Exercise 3.8) to find a harmonic conjugate v for u. What
is the analytic function u-+ iv?
6. Conformal Mappings
Let (t) = z(t) + iy(t), 0 < t <1, be a smooth parameterized curve
terminating at 29 = (0). We refer to
ue) = (0)
YO) = jim
as the tangent vector to the curve 7 at zp. It is the complex representation
of the usual tangent vector. We define the angle between two curves
at 29 to be the angle between their tangent vectors at 29,
= 2 +iy(0)
Theorem. If7(t),0
0.
If 2] N
Then lax| < slaxss| for k > N, and
lavls™ < Jassals** < laysals*? <2 V_ Power Series
Hence the terms ax2* do not converge to 0 for |2| > s, so that the series
does not converge, and s > R. Since s > L is arbitrary, we also have
1 R. We conclude that L
Example. For the series 5 k2*, the ratio test gives
k
a
oe
a1
ane
Hence the radius of convergence is R
Example, For the series 524, the ratio test gives
| _ (e+pt
Ost) AE
Hence the radius of convergence is R= +00.
= k+l $00,
‘The second formula is based on the root test.
‘Theorem. If {/[az) has a limit as k + 00, either finite or +00, then the
radius of convergence of Ya,2* is given by
(3.3) R=
Ir > 1/lim Yfar], then ¢/fag[r > 1 eventually, so that laxir* > 1
‘eventually, the terms of the series 3 ax2* do not converge to 0 for |2| = r
and r > R. On the other hand, if» < 1/lim ¢/fay| then {/ayir <1 even-
tually, so that laglr* < 1 eventually, the sequence [ay|r* <1 is bounded,
and from the definition of R we have r < R. It follows that (3.3) holds.
Example. For the series > kz*, the root test gives
R= 1/lim Vk = 1.
‘There is a more general form of the formula (3.3), called the Cauchy-
Hadamard formula, that gives the radius of convergence for any power
series in terms of a lim sup. Recall (Section II.1) that the lim sup of a
sequence {sq} is characterized as the number S, —oo < S < +00, with
the property that if ¢ > S, then only finitely many terms of the sequence
satisfy s, > t, while if ¢ < S, then infinitely many terms of the sequence
satisfy 5, > t. If the sequence sq has a limit, then the lim sup of the
sequence coincides with the limit. However, every sequence has a lim sup.
The Cauchy-Hadamard formula is obtained simply by replacing the limit
in (3.3) by a lim sup,
1
G4) imp Vind
Bxercises 143
"The proof is identical to the proof given above, except that the character-
izing property of the lim sup is used.
Example, We return to the seres (3.1) treated earlier. This ean be ex:
pressed as a power series )>axz*, where ay = (—1)*/?/2*/? if k is even,
and ay = 0 if kis odd. ‘Ths ¢/far] is 1/2 if kis even and 0 if kis odd
‘The lim sup of this sequence is 1/V2. By the Cauchy-Hadamard formula,
the radius of convergence of the series is V2, as before
Exercises for V.3
1, Find the radius of convergence of the following power series:
ole Olaae © Lime
ms &
a = t,t =
writ — LAE mL toswtret
& & &
= = a = ast
©) ek © ae OD
2. Determine for which z the following series converge.
@ Ye-pF © YL aMe-I™ | Ye
o a =
we wre" Hy Fe-
i 7 a
So wat at
3"
the radius of convergence of the following series.
sade a eet,
0) haste tebe aet eet
4. Show that the function defined by f(z) = > 2" is analytic on
the open unit disk {Jz| <1}. Show that [f(P2)| — +00 as 7 1
whenever 2 is a root of unity. Remark. Thus f(2) does not extend
‘analytically to any larger open set than the open unit disk.
'5. What functions are represented by the following power series?
(@) Skt (b) DE, Reema V_ Power Series
6. Show the series 2 acz*, the differentiated series ST kazz*~!, and
the integrated series > p24, = al have te sme radi of eon-
vergence.
7. Consider the series
Y e+ yy
Use the Cauchy-Hadamard formula to find the radius of convergence
ofthe series. What happens when the ratio testis applied? Evaluate
explicitly the sum of the series
8. Write out a proof of the Cauchy-Hadamard formula (3.4)
4. Power Series Expansion of an Analytic Function
We have seen that power series expansions 3 ay(2— 29)" are analytic inside
the disk of convergence {2 29| < R}. It is an important and far-reaching
fact that conversely, any function analytic on a disk can be expanded in a
power series that converges on the disk.
‘Theorem. Suppose that f(z) is analytic for |z ~ 20] 0. Suppose that there are A,C > 0
such that [fiz + iy)| < Ce*™ for y > 0. Show that
Fl) = YO ane,
oan
where the series converges uniformly in each half-plane {y ><}, for
fixed © > 0.
5. Suppose that +1 are periods of a nonzero doubly periodie function
f(z), and suppose that there are no periods w of f(2) satisfying
0 < |u| < 1. How many periods of f(z) lie on the unit circle?186 VI_ Laurent Series and Isolated Singularities
Describe the possibilities, and sketch the set of periods for each
possibility.
6. We say that w and wz generate the periods of a doubly periodic
function if the periods of the funetion are precisely the complex
numbers of the form mu; + nw, where m and n are integers. Show
that if ws and wp generate the periods of a doubly periodic function
f(z), and if Ar and Az are complex numbers, then 4; and dz generate
the periods of f(z) if and only if there is 2x2 matrix A with integer
entries and with determinant 1 such that A(w, 02) = (1,32).
7. Let and w» be two complex numbers that do not lie on the samme
line through 0. Let k > 3. Show that the series
= 1
rng = (i + a)
converges uniformly on any bounded subset of the complex plane
to a doubly periodic meromorphic function f(z), whose periods are.
generated by w and w. Strategy. Show that the number of periods
in any annulus {N < |s| < N+ 1} is bounded by CN for some
constant C.
6. Fourier Series
A complex Fourier series is a two-tailed series of the form
(6.1) SO cnet? = «+ te ne te pe Hey tere” + ene? +-
Laurent expansions are intimately related to Fourier series. If the Laurent
series
4s the Fourier series expansion of f(re'®), regarded as a function of 8. The
Foutled coefficients of the expansion are the coefficients cy = ayr
‘Suppose the series (6.1) converges uniformly to a function f (¢
fle) = Yo ge®.
6. Fourier Series 187
‘We can capture the coefficients of the series by multiplying by the exponen-
tial function e~*? and integrating with respect to the probability measure
49/2. The orthogonality relations for the exponential functions,
* ngunedd fe 5=k
(62) f on {i jth,
then yield
[serene = 3 og [tet = a.
This leads us to define the Fourier coefficients of any piecewise contin-
uous function (or any integrable function) f(e") to be
(63) a= [seen
De
and we associate to f(e
"e noo ck 1. For [2| large, the term 2” domi-
nates. We take f(z) = 2" and A(z) = ay12"-" +--+ a9. If Ris large,
we have [f(2)| < [f(2)| for |2| = R. Consequently, p(z) = f(z) + f(z) has
the same number of zeros in the disk |2| < Ras f(2) = 2", which is m.
Exercises for VITL2
1. Show that 22° + 6: — 1 has one root in the interval 0< x < 1 and
four roots in the annulus {1 < |2| <2}.
2. How many roots does 2° +25 ~ 8:9 4.22 +1 have between the circles
{l2| = 1} and {|2| = 2}?
3. Show that if m and n are positive integers, then the polynomial
we) = 1t24 54-4204
2 mi
has exactly n zeros in the unit disk,
4, Fix a complex number such that |A| <1, For n> 1, show that
(2—1)"eF — 2 has n zeros satisfying |2 — 1] < 1 and no other zeros
in the right half-plane. Determine the multiplicity of the zeros,
For a fixed A satisfying |A| < 1, show that (z ~ 1)"et + M2 +1)"
hhas n zeros in the right half-plane, which are all simple if A # 0.
6. Let p(z) = 2° +924 + 29 +22 +4 be the polynomial treated in
the example in this section. (a) Determine which quadrants contain
the four zeros of p(2) that lie inside the unit circle. (b) Determine
which quadrants contain the two zer0s of p(2) that lie outside the
unit circle. (e) Show that the two zeros of p(2) that lie outside the
unit circle satisfy {2 + 3i] < 1/10}
7. Let f(z) and g(2) be analytic functions on the bounded domain D
that extend continuously to OD and satisfy |f(2) + 9(2)| < [f(2)] +
lg(z)| on AD. Show that f(z) and g(2) have the same number of
3. Hurwitz’s Theorem 231
zeros in D, counting multiplicity. Remark, ‘This is a variant of
Rouché's theorem, in which the hypotheses are symmetric in f(2)
and g(z). Rouché's theorem is obtained by setting A(z) = ~ f(z) —
.9(2). For the solution of the exercise, see Exercise 9 in the preceding
section.
8, Let D bea bounded domain, and let f(2) and h(2) be meromorphic
functions on D that extend to be analytic on D. Suppose that
|h(2)] <|4(2)| on AD. Show by example that f(z) and f(2)+ h(2)
can have different numbers of zeros on D. What can be said about
J(z) and f(2) + h(2)? Prove your assertion.
9. Let f(z) be a continuously differentiable function on a domain D.
Suppose that for all complex constants a and b, the increase in the
argument of f(2)+az+6 around any small circle in D on whieh
J(2)+az +640 is nonnegative. Show that f(2) is analytic.
3. Hurwitz’s Theorem
‘The argument principle (logarithmic integral) provides a tool for studying,
the behavior of the zeros of a convergent sequence of analytic funetions.
‘The zeros of the funetions in the sequence converge to the zeros of the limit
function, in a sense made precise by the following theorem,
Theorem (Hurwits’s Theorem). Suppose {fx(z}} isa sequence of an-
alvtie functions on a domain D that converges normally on D to f(2), and
suppose that f(z) has a zero of order N at z. Then there exists p > 0
such that for k large, f(z) has exactly N zeros in the disk {|z — zo] < p},
counting multiplicity, and these zeros converge to 29 as k ~ oo.
For the proof, let p > 0 be sufficiently small so that the closed disk
{|2 — 20] r-_ Show further that,
fol6) = F-(6) and fil¢) = —F, (6). Remark. The formula f(z) =
fol2) + fu(z) reflects the jump theorem for the Cauchy integral of
(2) around circles |2| =r.
5. Let 7 be a piecewise smooth curve, and let F(¢) be the Cauchy
integral (7.1) of a continuous function f(z) on 7. Show that if 9(2)
is a smooth function on the complex plane that is zero off some
bounded set, then
[ocrteaas = 2 ff Brea ay
Hint, Recall Pompeiu’s formula (Section IV.8).
6. Determine whether the point = lies inside or outside. Explain.
7. Asimple arc I’ in C is the image of a continuous one-to-one func-
tion 7(f) from a closed interval {a,6] to the complex plane. Show
that a simple are Tin C has @ connected complement, that is, C\T is
connected. You may use the Tietze extension theorem, that a con-
‘tinuous real-valued function on a closed subset of the complex plane
can be extended to a continuous real-valued funccion on the entire
complex plane. Hint. Suppose zo belongs to a bounded component:
of C\P. Find a continuous determination h(s) of log(= — 30) on P,
extend A(z) to @ continuous function on C*, and define f(2) = 20
‘on the component of C\P containing 29, and f(:) = eM) on the
remainder of C*\P. Consider the increase in the argument of f(=)
around circles centered at zo
8. Prove the Jordan curve theorem fora simple closed curve by fling
in the following proof outline.
(a) Show that each component of C\I has boundary [. Hint. For
70 = 7(to) € T, apply the preceding exercise to the simple arc
T(r), where I is a small open parameter Interval conealn-
ing to.
(b) Prove the Jordan curve theorem in the ease where contains
a straight line segment.252 VIN The Logarithmic Integral
(c) Show that for any 29 = vito) € T, any small disk Do con-
taining 2, and any component U of C\P, there are points
2, = (ty) and =2 = 9(t2) such that the image of the parameter
segment between fy and t2 is contained in Dp and such that 21
and 27 can be joined by a broken line segment in U1 Do.
(a) With notation as in (b), let be the simple closed curve ob-
tained by replacing the segment of y in Dy between zo and =; by
the broken line segment in U' Do between them, and let 7 be
the simple close! curve in Dy obtained by following the segment
of 7 in Dp from 29 to 21 and returning to 9 along the broken
line segment. Show that W(r,¢) = 0 and W(7,¢) = W(o,<)
for CE CAL, ¢ ¢ Do.
{e) Using (b) and (4), show that C\ has at least two components
and that W(y,¢) = 41 for ¢ in each bounded component of
ev.
(£) By taking U in (c) to be @ bounded component of C\P, show
that W(7,¢) =0 for ¢ in any other component of C\T
8, Simply Connected Domains
A domain in the plane is ‘simply connected” if it has no “holes.” Disks
and rectangles are simply connected. More generally, star-shaped domains
are simply connected. Annuli, punctured disks, and the punctured plane
have “holes” and are not simply connected. Our aim in this section is to
make more precise the notion of “simple connectedness.”
Let (t), a B}. The estimate
[tein sani < © [eereras
shows that (Ch)(z) is bounded on the half-plane {Rez > B+ <} for any
£>0, and (£h)(2) + 0 as Rec + 00.
Example. Suppose h(s) = s* for some integer k > 0. Since ste~** —+ 0 as
8 — 00 (poiynomials grow more slowly than exponentials), there is for any
= > 0 an estimate of the form [h(s)| < C.e**. Hence the Laplace transform
(Cs*)(z) converges absolutely for all z satisfying Rez > 0. We compute
that.
(cetya) = [ersten ger [eta 0
‘The integral on the right is P(k-+1) = kt, so (Cs*)(=) coincides with kl/2**1
when z= x > 0. By the uniqueness principle it coincides with kt/2**! on366 XIV_ Some Special Functions
the right half-plane,
(22) (esha) = Ao
sr Rez>0, £20
Lemma. Suppose that |h(s)| < Cs%e* for some integer N > 0 and
constants B,C. Then
23) (cance +in)) = 0 (yr ). as 210.
‘This follows from the obvious estimates
© ate BiegM m 1
[cirri < ¢ [ee Pads = 0s = o(=t)
Theorem. Suppose that \h(s)| < Ce®* for s > 0. Fix N 1. IFA(s) has
the Taylor polynomial approximation 1
(24) fs) = bo bret Bea 4 PA BL O(N)
(s) = 18+ oF WH)
as6 04, then i
(2.5) (£h)(2)
as z= Rezo.
‘To see this, we let R(s) be the remainder term in the Taylor approxima-
tion, so that
bw
hfs) = by +s +o SY ERs), 520
(v-
Since |R(s)| < Cos for s > 0 near 0, there is C, > 0 such that
IR(s)| < Cys, OS 800.
Since the Laplace transform operates linearly,
(che) = m(EIN(2) +b (ee) H+ He
‘The result now follows from (22) and (23)
We wish to use the asymptotic approximation theorem to obtain asymp-
totie approximations to T(z). We begin by expressing the derivative of
T'(z)/T(2) as a Laplace transform, From (Cs)(z) = 1/2” we obtain
ie es\(z+k) = [fe Heeds = [re
Di (£s%-*) (2) + (ERY).
=(e*)fsds.
2. Laplace Transforms 367
Summing over k from 0 to N, and recognizing the geometric series, we
obtain
Loe [er Qeertoue- fi
‘he exponentials e~*( +2) decrease to 0, and passage to the limit in the
right-hand side is justified. In view of (1.6), we obtain in the limit as
Noe that
a0) _ [Pps
an [fe pists, Rezo.
‘The integral on the right is the Laplace transform of the function
sven
sds
gs) =
‘This function g(s) is a meromorphic function of the complex variable s,
with poles at s = 2rni for n = +1,42..... The function is analytic at
0, since 1 ~e~* has a simple zero there. The power series expansion
of g(s) about s = 0 is readily calculated to be
1
2
Since g(s)/s + 1 as s+ +00, an estimate of the form 0 < g(s) < Ce’,
8 2 0, holds for some constant C > 0. The theorem then yields the
asymptotic approximation
am) 1,14 1
ete ~ ptastasto(s): *
‘To obtain an asymptotic approximation to T(z), we must integrate twice.
Except for a constant of integration, the antiderivative for any error term
O(/z"), N > 2, is dominated by
[lear = [8 = wate =
Hence
g{s) = 1+
ss) < 2m.
Rez +00,
ot Log 40-5 ~ Ge +0 (4),
where ais @ constant of integration. If we integrate aga, we obtain
1 Lo o(2
logt(2) = 2 Log2—2+a2~} togs+ a+b +0(5
as) = loge— tas 5 Loge+ 8+ 7. +0(%).
where 4 is another constant of integration. If we substitute the asymp-
totic expansions for log? (2 +1) and logT(z) into the relation logl (2 +
1) = logz + logT'(z), we obtain a = 0. To evaluate 8, we compare68 XIV Some Special Functions
T(2n) with P(m), which can be done with the elementary relation in Exer-
cise 1.3(¢),
riot (ve) = enema
If we take logarithms, substitute the approximations, and do some book-
keeping, we obtain 3 = log V2r. We have proved the following,
‘Theorem. ‘The gamma function F(2) satisfies
1
(26) ogt(e) = ( :) logs 2 + on viF + I +0 (5)
Rez 00,
‘The asymptotic approximation to P(=) is obtained by exponentiating,
o()). = Res-+s
Substituting n! = nI'(n), we obtain aaa formula,
ny 1
("40+ a +0(@))
It is natural to consider the asymptotic behavior of Laplace transforms
in the right half-plane as 2 = Rez — co. However, for some classes of
functions it is more appropriate to consider asymptotic behavior ina sector
fs |2| oe rather than in a half-plane. Let f(2) be defined and analytic
in a sector $= {0 < argz < Oj}. We say that 7%» @,/2? is an
asymptotic series for f(2) in the sector S if for each N'>'0,
1-4 = °(=e)
fas 2 —+ co in the sector. If .a,/2? is an asymptotic series for f(). then
ae = dim 28 [4(2) ~ ag ~~. SH]
Hence the coefficients in the asymptotic series of f(2) are uniquely deter-
mined. We refer to 57 a,/2? as the asymptotic expansion of (2) in the
sector.
If f(z) is analytic a¢ co, then its Laurent series is an asymptotic series for
‘J(2) in any sector, However, asymptotic series generally do not converge.
(See the Exercises.) Further, it can happen that different functions have
the same asymptotic expansion. For example, since 2"¢-* —» 0 as z — 00
jn any sector {Jag z| < (7/2) ~<). the function e-* has an asymptotic
im the sector that is identically zero.
Exercises a
‘The asymptotic approsimation theorem shows that if h(s) has at most
exponential growth, ond if f(s) is infinitely differentiable at's = 0, then the
Laplace transform (£h)(2) has the asymptotic expansion J> A'8}(0)/z4+1
in any sector 5, = {larg 2| < (7/2) — =}. Im any such sector there is an
estimate of the form ¢,|z| < Rez < |z|: consequently, |2| —+ 00 if and
only if x = Rez ~ co. In general, the series need not be an asymptotic
expansion for (Ch)(z) in the entire right half-plane, regarded as a sector.
However, in the case ofthe function log (2), its. Infact it can be shown
that the series (2.6) is an asymptotic series for LogI'(2) in any sector
{arg 2| < xe}, for any ¢ >0.
Exercises for XTV.2
1. Show that if Re A > —1, then the Laplace transform of s* converges
absolutely in the right half-plane, and
pees)
(£3)(2) Rez >0.
2. Show that if g(s) = e**h(s), then (Cg)(z) = (£h)(z — a).
3. Suppose h(s) = 0 for s <0. Show that if@ > 0 and g(s) = h(s—a),
then (£9)(2) = e~*(Ch)(2),
4. Show tht ifs) is diferentible, then the Laplace transform of ts
derivative h'(s) is (Ch’)(z) = 2(£h)(z) — h(0),
5. Compure (Z)" V2 and (2)" Vor (1+) for n= 2, 8
au 4 and compare the eeu with ni
6. Define the Bernoulli numbers B,, Bz. Bs, ... by
So? = Bir, Baa, Bio,
23 tg a
{a) Show that B, = 1/6 and By = 1/30. (b) Show that
s 1, Big Boa, Bas
Lt jet Spe - Feat + Beat —
(c) Show that log (2) has asymptotic series
Bol | Byi
Dryas eer
(€) Show that as = — 00 through the sector | arg] < x/2~¢,
Me, 1, 1 1e 1
VEC: i maa mes +°(cR))
How many terms can be specified here using only By and Bs?
log (2) ~ (2 = 1/2)log2 ~ 2+ log VIR + BE
ry370 XIV_ Some Special Functions
Express the function
Fe =f
fas a Laplace transform, and show that It extends analytically to the
right half-plane. Obtain the asymptotic expansion of the function,
and show that the asymptotic expansion diverges at every point.
z>0,
8. Let h(s), 8 > 0, be a continuous function such that [h(s)] < Ce®*
for some constants B,C > 0. Suppose that h(s) extends to be
‘analytic ina neighborhood of the origin, and suppase that the power
series expansion f(s) = S7f2q Al) (0)st/kI has only a finite radius
of convergence. Show thitt the asymptotic series "jg h(®)(0)/2#*?
of the Laplace transform (Ch)(=) diverges at every point =,
9. Show that if f(2) is analytic at oo, then its power series expansion
in descending powers of = is an asymptotic series for f(z) in any
sector.
10. Show that if f(2) and g(z) have asymptotic expansions in a sector,
then so does their product f(2)g(z).
11. Show that if f(2) has an asymptotic series S72 ax/2* in a sector S,
beginning with the 1/2? term, then any antiderivative F(z) of f(2)
has an asymptotic series in S obtained by integrating term by term.
12, Show that if f(z) has an asymptotic series S7{2pax/2* in a see-
tor 5 = {4 < arg < 6}, then f'(2) has the asymptotic series
X(-Ja;)/2", obtained by differentiating term by term, in any
smaller sector Se = {0 +© You might also like