Frontiers in Mathematics
Advisory Editorial Board Leonid Bunimovich (Georgia Institute of Technology, Atlanta) Benoît ...

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Frontiers in Mathematics

Advisory Editorial Board Leonid Bunimovich (Georgia Institute of Technology, Atlanta) Benoît Perthame (Université Pierre et Marie Curie, Paris) Laurent Saloff-Coste (Cornell University, Ithaca) Igor Shparlinski (Macquarie University, New South Wales) Wolfgang Sprössig (TU Bergakademie Freiberg) Cédric Vilani (Ecole Normale Supérieure, Lyon)

Farit G. Avkhadiev Karl-Joachim Wirths

Schwarz-Pick Type

Inequalities

Birkhäuser Verlag Basel . Boston . Berlin

$XWKRUV )DULW*$YNKDGLHY Chebotarev Research Institute .D]DQ6WDWH8QLYHUVLW\ .D]DQ Russia HPDLO)DULW$YNKDGLHY#NVXUX

.DUO-RDFKLP:LUWKV ,QVWLWXWIU$QDO\VLVXQG$OJHEUD 78%UDXQVFKZHLJ %UDXQVFKZHLJ *HUPDQ\ HPDLONMZLUWKV#WXEVGH

0DWKHPDWLFDO6XEMHFW&ODVVL½FDWLRQ$&&&' )0

/LEUDU\RI&RQJUHVV&RQWURO1XPEHU

%LEOLRJUDSKLFLQIRUPDWLRQSXEOLVKHGE\'LH'HXWVFKH%LEOLRWKHN 'LH'HXWVFKH%LEOLRWKHNOLVWVWKLVSXEOLFDWLRQLQWKH'HXWVFKH1DWLRQDOELEOLRJUD½H GHWDLOHGELEOLRJUDSKLFGDWDLVDYDLODEOHLQWKH,QWHUQHWDWKWWSGQEGGEGH!

,6%1%LUNKlXVHU9HUODJ$*%DVHOÀ%RVWRQÀ%HUOLQ 7KLVZRUNLVVXEMHFWWRFRS\ULJKW$OOULJKWVDUHUHVHUYHGZKHWKHUWKHZKROHRUSDUWRIWKH PDWHULDOLVFRQFHUQHGVSHFL½FDOO\WKHULJKWVRIWUDQVODWLRQUHSULQWLQJUHXVHRILOOXVWUD WLRQVUHFLWDWLRQEURDGFDVWLQJUHSURGXFWLRQRQPLFUR½OPVRULQRWKHUZD\VDQGVWRUDJHLQ GDWDEDQNV)RUDQ\NLQGRIXVHSHUPLVVLRQRIWKHFRS\ULJKWRZQHUPXVWEHREWDLQHG %LUNKlXVHU9HUODJ$* %DVHOÀ%RVWRQÀ%HUOLQ 32%R[&+%DVHO6ZLW]HUODQG Part of Springer Science+Business Media &RYHUGHVLJQ%LUJLW%ORKPDQQ=ULFK6ZLW]HUODQG 3ULQWHGRQDFLGIUHHSDSHUSURGXFHGIURPFKORULQHIUHHSXOS7&) 3ULQWHGLQ*HUPDQ\ ,6%1

H,6%1

ZZZELUNKDXVHUFK

Dedicated to our families

Contents 1

2

Introduction 1.1 Historical remarks . . . . . . . . . . 1.2 On inequalities for higher derivatives 1.3 On methods . . . . . . . . . . . . . . 1.4 Survey of the contents . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

1 1 3 5 6

Basic coeﬃcient inequalities 2.1 Subordinate functions . . . . . . . . . . . 2.2 Bieberbach’s conjecture by de Branges . . 2.3 Theorems of Jenkins and Sheil-Small . . . 2.4 Inverse coeﬃcients . . . . . . . . . . . . . 2.5 Domains with bounded boundary rotation

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

7 7 11 15 18 23

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

27 27 30 33 37 40 44

Basic Schwarz-Pick type inequalities 4.1 Two classical inequalities . . . . . . . . . . 4.2 Theorems of Ruscheweyh and Yamashita . 4.3 Pairs of simply connected domains . . . . . 4.4 Holomorphic mappings into convex domains 4.5 Punishing factors for convex pairs . . . . . 4.6 Case n = 2 for all domains . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

49 50 52 55 59 63 66

Punishing factors for special cases 5.1 Solution of the Chua conjecture . . . . . . . . . . . . . . . . . . . . 5.2 Punishing factors for angles . . . . . . . . . . . . . . . . . . . . . .

69 69 72

3 The 3.1 3.2 3.3 3.4 3.5 3.6 4

5

. . . .

. . . .

Poincar´e metric Background . . . . . . . . . . . . . . . . . The Schwarz-Pick inequality . . . . . . . . Estimates using the Euclidean distance . . An application of Teichm¨ uller’s theorem . Domains with uniformly perfect boundary Derivatives of the conformal radius . . . .

viii

Contents 5.3 5.4 5.5

Sharp lower bounds for punishing factors . . . . . . . . . . . . . . Domains in the extended complex plane . . . . . . . . . . . . . . . Maps from convex into concave domains . . . . . . . . . . . . . . .

78 84 90

6

Multiply connected domains 97 6.1 Finitely connected domains . . . . . . . . . . . . . . . . . . . . . . 97 6.2 Pairs of arbitrary domains . . . . . . . . . . . . . . . . . . . . . . . 104 6.3 Some examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

7

Related results 113 7.1 Inequalities for schlicht functions . . . . . . . . . . . . . . . . . . . 113 7.2 Derivatives of α-invariant functions . . . . . . . . . . . . . . . . . . 117 7.3 A characterization of convex domains . . . . . . . . . . . . . . . . 124

8

Some open problems 8.1 The Krzy˙z conjecture . . . . . . . . 8.2 The angle conjecture . . . . . . . . . 8.3 The generalized Goodman conjecture 8.4 Bloch and several variable problems 8.5 On sums of inverse coeﬃcients . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

127 127 128 131 139 140

Bibliography

143

Index

155

Chapter 1

Introduction The aim of the present book is a uniﬁed representation of some recent results in geometric function theory together with a consideration of their historical sources. These results are concerned with functions f , holomorphic or meromorphic in a domain Ω in the extended complex plane C. The only additional condition we impose on these functions is the condition that the range f (Ω) is contained in a given domain Π ⊂ C. This fact will be denoted by f ∈ A(Ω, Π). We shall describe how one may get estimates for the derivatives |f (n) (z0 )|, n ∈ N, f ∈ A(Ω, Π), dependent on the position of z0 in Ω and f (z0 ) in Π.

1.1

Historical remarks

The beginning of this program may be found in the famous article [125] of G. Pick. There, he discusses estimates for the MacLaurin coeﬃcients of functions with positive real part in the unit disc found by C. Carath´eodory in [52]. Pick tells his readers that he wants to generalize Carath´eodory’s estimates such that the special role of the expansion point at the origin is no longer important. For the convenience of our readers we quote this sentence in the original language: Durch lineare Transformation von z oder, wie man sagen darf, durch kreisgeometrische Verallgemeinerung, kann man die Sonderstellung des Wertes z = 0 wegschaﬀen, so daß sich Relationen f¨ ur die Diﬀerentialquotienten von w an beliebiger Stelle ergeben. The ﬁrst great success of this program was G. Pick’s theorem, as it is called by Carath´eodory himself, compare [54], vol II, §286–289. If z0 ∈ Δ = {z | |z| < 1}, f ∈ A(Δ, Δ), and f (z0 ) = w0 , then the inequality w0 − f (z) ≤ z0 − z , z ∈ Δ, (1.1) 1 − z0 z 1 − w0 f (z) is valid.

2

Chapter 1. Introduction

This theorem follows immediately from Schwarz’s Lemma using holomorphic automorphisms of the unit disc. A direct consequence of Pick’s theorem is the inequality 1 − |f (z0 )|2 |f (z0 )| ≤ , f ∈ A(Δ, Δ), z0 ∈ Δ. (1.2) 1 − |z0 |2 To be more complete about the history of the inequality (1.2) we have to mention that O. Sz´ asz (see the footnote in [160], p.308) attributes it to E. Lindel¨ of [105] and indicates that (we quote once more): Neuere Beweise dieser Relation gaben Carath´eodory und Jensen. Herr Jensen zeigte, wie der Satz aus dem schon von Herrn Landau 1906 bewiesenen Spezialfall: |c1 | ≤ 1 − |c0 |2 leicht folgt. As an oft-quoted proverb says “Success has many fathers, but failure is an orphan”. For us, at this place the second part of this proverb hints of unsolved mathematical problems. It is clear that Schwarz’s lemma and its generalizations became widely known in the period of a systematic study of results which are closely connected with the proofs of the Riemann theorem on conformal mappings. By the way, the original version of the lemma may be found in [149], where H.A. Schwarz discussed the Riemann mapping theorem. Now, it is well known that one may generalize the inequalities (1.1) and (1.2) to f ∈ A(Ω, Π), z0 ∈ Ω, and f (z0 ) = w0 , where Ω and Π are domains that have at least three boundary points. This generalization is known as the principle of hyperbolic metric (see for instance R. Nevanlinna [120] and G. M. Goluzin [70]). An important case is presented by simply connected domains. Let Ω and Π be simply connected proper subdomains of C and let ΦΩ,z0 be the unique conformal map of Δ onto Ω such that ΦΩ,z0 (0) = z0 and ΦΩ,z0 (0) > 0. The existence of this map is proved by the Riemann mapping theorem.We deﬁne ΦΠ,f (z0 ) analogously. Then the function w = Φ−1 (1.3) Π,f (z0 ) ◦ f ◦ ΦΩ,z0 belongs to the family A(Δ, Δ) and satisﬁes w(0) = 0. Hence, Schwarz’s Lemma implies ΦΠ,f (z0 ) (0) |f (z0 )| ≤ , ΦΩ,z0 (0) the generalized Schwarz-Pick inequality. The quantities R(z0 , Ω) := ΦΩ,z0 (0)

and R(f (z0 ), Π) := ΦΠ,f (z0 ) (0)

are called the conformal radius of Ω at the point z0 and of Π at w0 = f (z0 ), respectively, and will be used here to describe the positions of the points z0 in Ω and f (z0 ) in Π.

1.2. On inequalities for higher derivatives

1.2

3

On inequalities for higher derivatives

We ask for inequalities of the form |f (n) (z0 )| R(f (z0 ), Π) , ≤ Mn (z0 , Ω, Π) n! (R(z0 , Ω))n

n ∈ N,

(1.4)

where f and z0 are as above. Concerning the history of inequalities for higher derivatives, we should mention here that the interest of researchers in geometric function theory was concentrated for a long time after 1920s on the famous Bieberbach conjecture, i.e., the conjecture that |f (n) (0)| (1.5) ≤ n|f (0)| n! for functions f holomorphic and injective on Δ. This may be seen as a special case of the above inequality, where f maps Δ conformally onto the simply connected domain f (Δ). Since it seemed very diﬃcult to prove this conjecture, less attention was attracted by the generalized Bieberbach conjecture or Rogosinski conjecture, which in the above formulation means that |f (n) (0)| ≤ nR(f (0), Π) n!

(1.6)

for any simply connected proper subdomain Π of C and any f ∈ A(Δ, Π). An equivalent formulation of the Rogosinski conjecture is the following. Let g be holomorphic and injective on Δ and let there exist a function w : Δ → Δ such that w(0) = 0 and f = g ◦ w. Then (1.6) is valid. As usual, we will abbreviate this relation between f and g by f ≺ g and say that under these circumstances f is subordinated to g. It was a great surprise when de Branges succeeded in proving not only (1.5) but also (1.6) in 1985. Soon afterwards, it was recognized, especially by Yamashita, that one could use (1.5) to go further steps in Pick’s program for functions injective on Δ. In fact, it had been seen earlier by Landau and Jakubowski that the validity of the Bieberbach conjecture would imply sharp bounds for |f (n) (z0 )|, z0 ∈ Δ, and f injective on Δ. We will give an outline of these results in the present book. Further, we will speak on the ideas of Chua [56], who indicated that (1.5) could be used to derive similar bounds for functions f injective on simply connected domains Ω. In this paper, he considered in addition the case that Ω is convex and published the following conjectures. Let f be holomorphic and injective on a proper convex subdomain Ω of C. Then for any z0 ∈ Ω, and any n ≥ 2 the inequality |f (n) (z0 )| |f (z0 )| ≤ (n + 1)2n−2 n! (R(z0 , Ω))n−1

(1.7)

4

Chapter 1. Introduction

is valid. If in addition f (Ω) is convex, then |f (z0 )| |f (n) (z0 )| ≤ 2n−1 n! (R(z0 , Ω))n−1

(1.8)

is valid. Motivated by the work of Ruscheweyh and Yamashita, who had proved estimates of the type (1.4) and similar ones, the authors of the present book concentrated research in the last years on formulas like (1.4). The results of this research together with known theorems in this direction form the content of this book. Especially, we shall show that Chua’s conjectures are true. We will show even more, namely that Mn (z0 , Ω, Π) ≤ (n + 1)2n−2 for f ∈ A(Ω, Π), Ω convex and Π simply connected and that Mn (z0 , Ω, Π) ≤ 2n−1 for f ∈ A(Ω, Π), Ω and Π convex. These results will be completed by several considerations for special cases of Ω and Π as well as by computations of the dependences of Mn (z0 , Ω, Π) of the variable z0 . One further implication of the validity of the Bieberbach conjecture is concerned with meromorphic functions univalent on Δ. In 1962, J. Jenkins considered functions f that in addition to the above properties satisfy the condition that the pole of f lies at a point p ∈ (0, 1). He proved that for such functions the Bieberbach conjecture would imply n−1 |f (n) (0)| |f (0)| 2k p . ≤ n−1 n! p

(1.9)

k=0

This result motivated us to generalize the above considerations to proper subdomains of C. It is natural that here the positions of z0 in Ω and f (z0 ) in Π relative to the point at inﬁnity enter the picture. We use the hyperbolic distances to characterize those items. We will prove results of the type (1.4) using a generalization of the conformal radius. These results are satisfying in the following cases: 1) Ω and Π are simply connected with respect to C. 2) Ω is simply connected with respect to C and Π is convex. 3) Ω is convex and C \ Π is compact and convex. In the important case that Ω = Δ or Ω convex and Π is simply connected with respect to C we can present only partial results. One of the chapters is devoted to the most diﬃcult case, the case of multiply connected domains. Here, we deﬁne appropriate questions and give some answers that in most cases are far from the sharpness we achieved in many of the above mentioned theorems. Hence this chapter is more or less an impetus for further research.

1.3. On methods

5

1.3 On methods What can be said about methods? Essentially, the proofs of Schwarz-Pick type inequalities are based on relationships of certain hyperbolic characteristics of domains with the following results in geometric function theory: 1. Littlewood’s ideas on subordinate functions developed by Rogosinski, Goluzin, Clunie, Robertson, and Sheil-Small. 2. Coeﬃcient estimates using L¨owner’s theory on parametric representation of univalent functions. Especially, L¨ owner’s theorem on inverse coeﬃcients and de Branges’ proof of the Bieberbach conjecture. 3. Explicit representation of convex hulls for several families of analytic functions by the Herglotz formula and its generalizations. Usually, coeﬃcient estimates concern univalent or subordinate functions holomorphic in the unit disk. We prove and use several new versions of the known theorems for subordinate or quasi-subordinate functions which are holomorphic in a neighbourhood of the origin. Also, to prove Schwarz-Pick type inequalities for higher derivatives one needs certain hyperbolic characteristics of plane domains. For instance, let Qn (z, w, Ω, Π) be deﬁned as the smallest possible value such that the inequality (n) f (z) R(w, Π) ≤ Qn (z, w, Ω, Π) n! R(z, Ω)n holds for all f ∈ A(Ω, Π), f (z) = w, and Cn (Ω, Π) = sup{Qn (z, w, Ω, Π) | (z, w) ∈ Ω × Π}. The principle of the hyperbolic metric implies that Q1 (z, w, Ω, Π) = 1, and, in turn, that C1 (Ω, Π) = 1 for any pair (Ω, Π) equipped with the Poincar´e metric. In the case n ≥ 2 the quantity Qn depends on the hyperbolic characteristics Rk (z, Ω)

∂ k log R(z, Ω) ∂ k log R(w, Π) and Rk (w, Π) , 1 ≤ k ≤ n − 1. k ∂z ∂wk

In particular, Q2 (z, w, Ω, Π) depends on z and w via the quantities p = |∇ R(z, Ω)| and q = |∇ R(w, Π)|, only. Consequently, explicit estimates of punishing factors Cn (Ω, Π) are closely connected with the behaviour of the above characteristics, and, roughly speaking, estimating of these quantities is related to certain coeﬃcient problems of geometric function theory. Finally, we have to attract the reader’s attention to the following fact which deserves to be widely known: The quantity reciprocal to the conformal or hyperbolic radius is exactly the density of the Poincar´e metric. More precisely, the equation λΩ (z) :=

1 , R(z, Ω)

z ∈ Ω,

6

Chapter 1. Introduction

deﬁnes the density of the hyperbolic metric in the domain Ω with Gaussian curvature K = −4.

1.4

Survey of the contents

In this section, we would like to give a short survey of the contents of the present book. Chapters 2 and 3 have a preparatory character. There we will gather the materials from the work of many researchers in geometric function theory that we need for our results. Therefore we do not give all proofs in detail in these two chapters and we shortened and simpliﬁed old proofs. During these eﬀorts we found “new looks through old holes” now and then. Chapter 2 is dedicated to the most famous coeﬃcient theorems of the last century, namely the Bieberbach conjecture and the conjecture for the coeﬃcients of functions with bounded boundary rotation. The third chapter discusses the second theme that is important for our generalizations of the Schwarz-Pick lemma; the Poincar´e metric, its historical background, and well-known theorems concerning this metric in diﬀerent circumstances. Among them are the famous theorems of Landau, Teichm¨ uller, Beardon and Pommerenke. Most of the material presented in the following chapters has been developed by us in the last ten years, some of it has been published, other results appear in this book for the ﬁrst time. Chapter 4 contains the most prominent members of the family of punishing factors, those for pairs of simply connected domains and those for pairs of convex domains. Moreover, these two are in some sense the most beautiful, namely 4n−1 and 2n−1 . We discuss the work of Ruscheweyh, Chua, and Yamashita on these questions. Further, we present the complete solution for the case n = 2 for any pair of domains. The ﬁfth chapter is devoted to more special results. The most prominent of them may be the proof of a far reaching generalization of Chua’s conjecture. We add the determination of punishing factors for pairs of simply connected domains in the extended plane, and we prove sharp lower bounds in some general circumstances. Chapter 6 is concerned with some generalizations to multiply connected domains for arbitrary n > 2 and it has at some places a tentative character, since we are not sure that we have found the “best” way of generalization. In the last two chapters, we present some material in the neighbourhood of our results. This is meant as the basis for further research on the many questions that are natural to pose here. At this point, we want to thank the Deutsche Forschungsgemeinschaft for continued support of our research. Their many grants for F. G. Avkhadiev enabled us to do all the scientiﬁc work presented in this book.

Chapter 2

Basic coeﬃcient inequalities There are many books that systematically present coeﬃcient problems in geometric function theory. We refer the reader to the excellent monographs by Goluzin [70], Goodman [73], Hayman [78], Pommerenke [128], and Duren [60]. In this chapter we only mention a few classical results on coeﬃcients which are closely connected with the topic of this book. Also, we give several new facts with short proofs that have until now been presented only in original papers.

2.1

Subordinate functions

Let the functions Φ and Ψ be meromorphic in the unit disc Δ. We will say that Φ is subordinate to Ψ and write Φ ≺ Ψ or Φ(z) ≺ Ψ(z), whenever there exists a function ω holomorphic in Δ with properties ω(0) = 0, |ω(z)| < 1, and such that Φ(z) = Ψ(ω(z)). Clearly, |ω(z)| ≤ |z| in Δ by the Schwarz lemma. In [106], Littlewood proved the following result. Theorem 2.1. Let Φ and Ψ be holomorphic in the unit disc Δ, and let Φ(0) = Ψ(0) = 0. If Φ(z) ≺ Ψ(z) and p ∈ (0, ∞), then for any r ∈ (0, 1), 0

2π

|Φ(reiθ )|p dθ ≤

0

2π

|Ψ(reiθ )|p dθ.

We reproduce a short proof of Theorem 2.1 for the case p ∈ N, only.

(2.1)

8

Chapter 2. Basic coeﬃcient inequalities

Proof of Theorem 2.1 in the case p ∈ N. By Poisson’s formula, 2π ζ +z 1 p p dθ, ζ = reiθ , |z| < r < 1. Ψ (ζ)Re Ψ (z) = 2π 0 ζ −z For |ω(z)| ≤ |z| < r we can write |Φ(z)|p = |Ψ(ω(z))|p ≤

1 2π

2π

|Ψ(ζ)|p Re

0

ζ + ω(z) ζ − ω(z)

dθ,

ζ = reiθ .

Integrating the latter inequality over the circle {z = ρeit | 0 ≤ t ≤ 2π}, ρ ∈ (0, r), using 2π 1 ζ + ω(ρeit ) ζ + ω(0) Re dt = Re = 1, 2π 0 ζ − ω(ρeit ) ζ − ω(0) one gets

0

2π

|Φ(ρeit )|p dt ≤

2π

0

|Ψ(reiθ )|p dθ,

0 < ρ < r < 1.

Letting ρ → r gives inequality (2.1) in the case p ∈ N.

Remark 2.2. Clearly, a direct use of Theorem 2.1 with p = 2 and Parseval’s formula for the holomorphic functions Φ(z) =

∞

An z n

and

Ψ(z) =

n=0

leads to the inequality

∞

Bn z n ,

|z| < 1,

n=0 ∞

|An |2 ≤

n=0

∞

|Bn |2 .

n=0

Using Theorem 2.1 with p = 2 in an original way, Rogosinski proved the following assertion on coeﬃcients of these two functions. Theorem 2.3 (Rogosinski, [139]). Let Φ and Ψ be holomorphic in the unit disc Δ. If Φ(z) ≺ Ψ(z), then for all n ∈ N ∪ {0}, n

|Ak |2 ≤

k=0

n

|Bk |2 .

k=0

Simple counterexamples show that the inequality |An | ≤ |Bn | for n ≥ 2 does not hold, for instance, for the functions Φ(z) = z n and Ψ(z) = z. Also, the assertion of Theorem 2.3 with p = 2, i.e., the inequality n k=0

|Ak |p ≤

n k=0

|Bk |p

2.1. Subordinate functions

9

is not true in general (see [139]). In [137], Robertson generalized Theorem 2.3 to the case when the functions Φ and Ψ are holomorphic in the unit disc Δ and Φ(z) = ϕ(z)Ψ(ω(z)),

|z| < 1,

where |ϕ(z)| ≤ 1 for z ∈ Δ and ω is as above. In this case Φ is said to be quasisubordinate to the function Ψ (see also [57] and [128]). Robertson’s theorem can be generalized to meromorphic functions or, more generally, to functions F and G that are holomorphic only in a neighbourhood of the origin (see [19], [30] and [20]). Theorem 2.4 (see [20]). Let the functions F and G be holomorphic in a neighbourhood of the origin, where they have expansions F (z) =

∞

An z n

and

G(z) =

n=0

∞

Bn z n .

n=0

If there exist two functions ϕ and ω holomorphic in the unit disc Δ with |ϕ(z)| ≤ 1 and |ω(z)| ≤ |z| for z ∈ Δ such that the identity F (z) = ϕ(z)G(ω(z))

(2.2)

is satisﬁed in a neighbourhood of the origin, then for all n ∈ N ∪ {0} the following inequalities are valid: n n 2 |Ak | ≤ |Bk |2 . (2.3) k=0

k=0

Proof. The proof diﬀers from the classical one only in some details. We ﬁx n ∈ N ∪ {0}. From (2.2) we conclude that, in a neighbourhood of the origin, ∞

Ak z k − ϕ(z)

k=n+1

∞

Bk ω(z)k = ϕ(z)

k=n+1

n

Bk ω(z)k −

k=0

n

Ak z k .

k=0

If we expand the diﬀerence on the left side of this equation in a Taylor series in a neighbourhood of the origin, we see that the coeﬃcients of z k , 0 ≤ k ≤ n, in this series are zero. Therefore, we get, if we denote this function by H, an expansion H(z) =

∞

Ck z k

k=n+1

in a neighbourhood of the origin. On the other hand, it is immediately clear that the diﬀerence on the right side is holomorphic in the whole unit disc. This implies that the identity n k=0

Ak z k +

∞ k=n+1

Ck z k = ϕ(z)

n k=0

Bk ω(z)k

10

Chapter 2. Basic coeﬃcient inequalities

is valid for |z| < 1, too. Now, we observe that the Littlewood theorem is true for quasi-subordinate holomorphic functions, too. Consequently, we may proceed as in the classical proof of Theorem 2.3 to get n

|Ak |2 ≤

k=0

n

∞

|Ak |2 +

k=0

|Ck |2 ≤

k=n+1

n

|Bk |2

k=0

by use of Parseval’s formula in Theorem 2.1 with p = 2 for the quasi-subordinate holomorphic functions Φ(z) =

n

∞

Ak z k +

k=0

Ck z k ,

Ψ(z) =

k=n+1

n

Bk z k ,

z ∈ Δ.

k=0

This completes the proof of Theorem 2.4.

Following the idea of Goluzin (see [70] and compare also [57] and [137], Theorem 6.3) one may obtain a generalization of Theorem 2.4 as follows. Consider inequality (2.3) for n = 0, 1, . . . , m, where m ∈ N. Let λ0 ≥ λ1 ≥ · · · ≥ λm ≥ λm+1 = 0. We multiply the inequality (2.3) for n = j by λj − λj+1 ≥ 0, j = 0, 1, . . . , m. Summing up the results over j one easily has m

λj |Aj |2 ≤

j=0

m

λj |Bj |2 .

j=0

Since m is arbitrary, this gives the following theorem, Goluzin’s version of the inequalities (2.3). Theorem 2.5. Let n ∈ N ∪ {0} and let F and G be as in Theorem 2.4. If λ0 ≥ λ1 ≥ . . . λn ≥ 0, then the following inequality is valid: n

λk |Ak |2 ≤

k=0

n

λk |Bk |2 .

(2.4)

k=0

The following assertion generalizes a known idea due to Clunie [57]. Corollary 2.6. Let n ∈ N ∪ {0} and let f and g be functions meromorphic in the unit disc Δ with expansions of the form f (z) =

∞ k=0

Ak z k−m

and

g(z) =

∞

Bk z k−m

k=0

in a neighbourhood of the origin with some m ∈ N. If λ0 ≥ λ1 ≥ . . . λn ≥ 0 and |f (z)| ≤ |g(z)| in Δ, then the inequality (2.4) is valid.

2.2. Bieberbach’s conjecture by de Branges

11

To obtain Corollary 2.6 it is suﬃcient to apply Theorem 2.5 to functions F and G deﬁned by F (z) = z m f (z), G(z) = z m g(z), ω(z) = z, ϕ(z) = f (z)/g(z).

2.2

Bieberbach’s conjecture by de Branges

Consider the normal family S of all functions f that are holomorphic and univalent in Δ and have a Taylor expansion of the form f (z) = z +

∞

an z n ,

|z| < 1.

n=2

Before de Branges’ proof of the Bieberbach conjecture in [43] via the Milin conjecture, the following seven conjectures in their full generality were open problems. 1. Bieberbach Conjecture (1916). ([40]) For any f ∈ S the inequality |an | ≤ n holds for all n ≥ 2. The equality occurs if and only if f (z) is the Koebe function kd (z) = z(1 + dz)−2 , |d| = 1. 2. Littlewood Conjecture (1925). ([106]) If f ∈ S and f (z) = w for any z ∈ Δ, then |an | ≤ 4|w|n holds for all n ≥ 2. 3. Robertson Conjecture (1936). ([136]) For any odd function h(z) = z + c3 z 3 + c5 z 5 + · · · in S, the inequality 1 + |c3 |2 + · · · + |c2n−1 |2 ≤ n, is true for all n ≥ 2. 4. Rogosinski (Generalized Bieberbach) Conjecture (1943). ([139]) Let g(z) = b1 z + · · · + bn z n + · · · be a holomorphic function in Δ. If g(Δ) ⊂ f (Δ)and f ∈ S, then the inequality |bn | ≤ n holds for all n ≥ 2. 5. Asymptotic Bieberbach Conjecture (1955), connected with Hayman’s regularity theorem (see [78]). If An = max |an |, f ∈S

then lim

n→∞

An = 1. n

6. Milin Conjecture (1967). ([116]) For any f ∈ S, let γn be deﬁned by log

∞ f (z) γn z n . =2 z n=1

12

Chapter 2. Basic coeﬃcient inequalities Then the inequality

m n m=1 k=1

1 k|γk | − k 2

≤0

holds for all n ≥ 1. 7. Sheil-Small Conjecture (1973). ([152]) For any f ∈ S and any polynomial P (z) = b0 + b1 z + · · · + bn z n the convolution (= Hadamard product) deﬁned by (P ∗ f )(z) = a1 z + a2 b2 z 2 + · · · + an bn z n satisﬁes the inequality max |(P ∗ f )(z)| ≤ n max |P (z)|,

|z|≤1

|z|≤1

for all n ≥ 2. We refer the reader to the nice book [71] concerning details of the logical non-trivial relationship between these seven conjectures. In general, there are the following implications: =⇒ =⇒ =⇒

Milin Conjecture =⇒ Robertson Conjecture Sheil-Small Conjecture =⇒ Rogosinski Conjecture Bieberbach Conjecture Asymptotic Bieberbach Conjecture =⇒ Littlewood Conjecture.

Thus, in [43] de Branges settled all these conjectures by proving the Milin conjecture although he considered only the implications Milin Conjecture ⇒ Robertson Conjecture ⇒ Bieberbach Conjecture. In connection with these seven conjectures, in the next section we shall examine with proof two facts which deserve to be better known. The ﬁrst one concerns a conjecture by Goodman (see[72] and [73]), which says that the Taylor coeﬃcients ∞ of the function fp (z) = z + n=2 an (p)z n , |z| < p, satisfy the sharp inequality |an (p)| ≤ 1 + p2 + · · · + p2n−2 /pn−1 , whenever the function f is meromorphic and univalent in Δ and f has a simple pole at a point peit ∈ Δ, 0 < p < 1. Letting p → 1 this implies the inequality |an | ≤ n conjectured by Bieberbach. In fact, the Bieberbach conjecture is equivalent to the Goodman conjecture (1956) by an elegant proof of Jenkins [86]. Secondly, we shall consider the Sheil-Small conjecture in its full generality (see [152]), which deals with functions subordinate to f ∈ S. This little nuance becomes important in applications to the Schwarz-Pick type inequalities. Moreover, this consideration will contain the whole path from the Robertson conjecture to the Rogosinski conjecture as a special case. For the convenience of the reader who is not familiar with these inequalities, we will explain without technical details the relationships between the diﬀerent conjectures. The easiest step is the implication Robertson conjecture =⇒ Bieberbach conjecture.

2.2. Bieberbach’s conjecture by de Branges ˜ For f ∈ S let h(z) =

13

f (z 2 )/z 2 . Then the odd function

˜ h(z) = z h(z) =z+

∞

c2n−1 z 2n−1

n=2

˜ is an even function belongs to the family S√as well. If we take into account that h ˜ z) is holomorphic in Δ and that f (z) = z(s(z))2 . This we see that s(z) = h( identity implies that, for n ≥ 2, an =

n−1

c2k+1 c2(n−k)−1 ,

k=0

where c1 = 1. According to the Cauchy inequality, this yields |an | ≤

n−1

|c2k+1 |2 .

k=0

Hence, the above implication is obvious. In fact, in the present book, we will need a more general implication from the truth of the Robertson conjecture, namely, the theorem of Sheil-Small, see below. The implication Milin conjecture =⇒ Robertson conjecture follows from a theorem that is concerned with the exponentiation of holomorphic functions, the so-called Second Lebedev-Milin Inequality (see [99] and [116]). Let φ(z) =

∞

αk z k

k=1

be holomorphic in a neighbourhood of the origin and eφ(z) =

∞

βk z k .

k=0

Then for n ≥ 2 the inequalities n−1 1 |βk |2 ≤ exp n k=0

n−1 m 1 1 k|αk |2 − n m=1 k k=1

are valid. If one applies this theorem to the function φ(z) = log(s(z)) =

∞ k=1

γk z k ,

14

Chapter 2. Basic coeﬃcient inequalities

one sees that to prove the Robertson conjecture it is suﬃcient to prove that for n ∈ N the inequalities m n n 1 1 2 2 = (n − k + 1) ≤ 0 k|γk | − k|γk | − k k m=1 k=1

k=1

are valid. It has become customary to formulate this inequality in terms of the so-called logarithmic coeﬃcients of f . Since log(f (z)/z) = 2 log(s(z)) =

∞

2γk z k ,

k=1

many people know this conjecture in the form n 4 2 (n − k + 1) ≤ 0. k|2γk | − k

(2.5)

k=1

From this formulation de Branges found his way to prove the Milin conjecture (see [43], [64], and [71]). The ﬁrst item in this proof is the L¨ owner theory assuring ﬁrstly that, in problems like the above, it is suﬃcient to consider univalent functions that map the unit disc onto the complex plane minus a slit. The second ingredient from L¨owner’s theory is the fact that for any such function f there exists a chain of functions ∞ f (z, t) = et z + an (t)z n , z ∈ Δ, t ∈ [0, ∞), n=2

such that f (z, 0) = f (z) and the L¨ owner diﬀerential equation ∂ f (z, t) 1 + κ(t) z∂ f (z, t) = ∂t 1 − κ(t) ∂z is satisﬁed, where |κ(t)| = 1 and κ continuous on [0, ∞). Naturally, this diﬀerential equation results in diﬀerential equations for the logarithmic coeﬃcients cn (t) deﬁned by ∞ f (z, t) = cn (t)z n log et z n=1 with cn (0) = 2γn . The genial idea of de Branges was to look at (2.5) as an initial value problem. He constructed the so-called special function system of de Branges τn,k (t), 1 ≤ k ≤ n, n ∈ N, t ∈ [0, ∞), such that τn,k (0) = n − k + 1 considering the functions n 4 τn,k (t). k|ck (t)|2 − ϕn (t) = k k=1

2.3. Theorems of Jenkins and Sheil-Small

15

Using the L¨ owner diﬀerential equation he showed that for his system of functions the identity n τn,k (t) ϕn (t) = − |bk−1 (t) + bk (t) + 2|2 k k=1

is valid, where b0 (t) = 0, bk (t) =

k

jcj (t)κ(t)−j ,

k ∈ N.

j=1

The ingredient of the theory of special functions in de Branges’ proof was the proof that τn,k (t) ≤ 0 and limt→∞ τn,k (t) = 0. Hence, limt→∞ ϕn (t) = 0 and ϕn (t) ≥ 0. This in turn implies ϕn (0) ≤ 0 which is equivalent to (2.5).

2.3

Theorems of Jenkins and Sheil-Small

Let p =∈ (0, 1]. We need the expansion κp (z) =

∞ z

=z+ cn (p)z n , z (1 − pz) 1 − p n=2

|z| < p.

(2.6)

It is known that (see [86]) n−1 cn (p) =

j=0 p pn−1

2j

=

pn − p−

1 pn 1 p

.

(2.7)

In 1962, when the proof of the Bieberbach conjecture was a far-oﬀ dream, Jenkins proved the following theorem. Theorem 2.7 (Jenkins [86]). Let fp be a function meromorphic and univalent in the unit disc Δ with a simple pole at the point peit , t ∈ R, and 0 < p < 1, and let fp have the expansion ∞ an (p)z n (2.8) fp (z) = z + n=2

in a neighbourhood of the origin. If the Bieberbach conjecture is true for all coefﬁcients of schlicht functions, then |an (p)| ≤ cn (p) for any n ≥ 2. Equality occurs for the function eit κp (e−it z). Proof. Without loss of generality we suppose that t = 0. Let Δ(p) be the domain obtained from the unit disc by deleting the segment [p, 1]. Consider the class S(p) of all functions holomorphic and univalent in Δ(p) with expansion of the form

16

Chapter 2. Basic coeﬃcient inequalities

(2.8) in a neighbourhood of the origin. It is obvious that S(1) = S. To ﬁnd a conformal map of Δ(p) onto Δ we proceed as follows. It is clear that the function κ ˜ p (z) =

(1 + p)2 κp (z) 4p

maps Δ(p) conformally onto C \ (−∞, −1/4]. Now, we use the inverse K−1 of the Koebe function k−1 (z) = z/(1 − z)2 to map this domain onto Δ. Therefore, the function ϕ = K−1 ◦ κ ˜ p has the desired property. If we consider the expansion ϕ(z) =

∞ (1 + p)2 cn z n , z + 4p n=2

it is important to recognize that cn > 0 for all n. To see this, we compute ϕ explicitly by the above procedure to get ϕ(z) = 1 +

2p(1 − z/p)(1 − zp) 2p(1 + z)(1 − z/p)1/2 (1 − zp)1/2 − . 2 (1 + p) z (1 + p)2 z

It is clear that cn , n ≥ 2, is the sum of two consecutive coeﬃcients of the function (1 − z/p)1/2 (1 − zp)1/2 multiplied by the factor −2p/(1 + p)2 . The fact that these coeﬃcients themselves are negative for p ∈ (0, 1) is easily seen using the generalized binomial expansion and the Cauchy product for the product (1−z/p)1/2 (1−zp)1/2 . The inverse to the function ϕ is frequently used in extremal problems for bounded functions holomorphic and univalent in the unit disc (see for instance [133]). Clearly, any function fp ∈ S(p) admits a representation of the form fp (z) =

4p f (ϕ(z)), (1 + p)2

where f (z) = z +

∞

an z n ,

z ∈ Δ(p),

z ∈ Δ,

n=2

is a function in S. The function fp (z) = 4p(1 + p)−2 f (ϕ(z)) has the expansion of the form (2.8) with n an (p) = λ1 (n) + λj (n)aj , j=2

where λj (n) are polynomials in cm = ϕ(m) (0)/m! , m = 2, . . . , n, with non-negative coeﬃcients, and thus they are themselves non-negative. In particular, if aj = j, i.e. if f (z) = κ1 (z), then an (p) = cn (p).

2.3. Theorems of Jenkins and Sheil-Small

17

The validity of the Bieberbach conjecture implies that |an (p)| ≤ λ1 (n) + ≤ λ1 (n) +

n j=2 n

λj (n)|aj | jλj (n) = cn (p).

j=2

This completes the proof of Theorem 2.7.

Remark 2.8. As is shown in [18], the domain of variablity of coeﬃcients an (p) is only a proper subset of the set {w | |w| ≤ cn (p)}, if the pole at p lies near to the origin. The formulation of the Sheil-Small result in its full generality that we have in mind and which can be found, at least implicitly, in [152] and [71] is the following. Theorem 2.9 (Sheil-Small [152] (1973)). Let g be subordinated to a function f ∈ S and P a polynomial of degree less than or equal to n. If the Robertson conjecture is true, then max |(P ∗ g)(z)| ≤ n max |P (z) =: n M (P ). (2.9) |z|≤1

|z|=1

First we consider the following lemma. Lemma 2.10 (Sheil-Small [152]). Let U (z) =

∞

uk z k

and

V (z) =

k=0

∞

vk z k

k=0

be holomorphic in Δ and wi : Δ → Δ, i = 1, 2, 3, be holomorphic in Δ. If P is a polynomial of degree less than or equal to n and ˜ h(z) = zw1 (z)U (zw2 (z))V (zw3 (z)),

z ∈ Δ,

then we have iθ ˜ )| ≤ r max |P (z)| |(P ∗ h)(re

n−1

|z|=1

k=0

2 2k

|uk | r

1/2 n−1

1/2 2 2k

|vk | r

k=0

for r ∈ (0, 1), θ ∈ [0, 2π]. Proof. Let the abbreviations be as above. We use the representations iθ ˜ (P ∗ h)(re ) =

1 2π

0

2π

˜ i(θ+ϕ) ) d ϕ. P (ei(θ−ϕ) )h(re

(2.10)

18

Chapter 2. Basic coeﬃcient inequalities

and ˜ (P ∗ h)(z) = P (z) ∗

zw1 (z)

n−1

k

uk (zw2 (z))

k=0

n−1

k

vk (zw3 (z))

.

k=0

From this, we obtain iθ ˜ )| |(P ∗ h)(re

n−1 n−1 iϕ iϕ k iϕ iϕ k |w1 (re )| uk (re w2 (re )) vk (re w3 (re )) d ϕ 0 k=0 k=0 ⎞ 1/2 2 2π n−1 1 iϕ iϕ k ≤ r M (P ) ⎝ uk (re w2 (re )) d ϕ⎠ 2π 0

1 ≤ r M (P ) 2π ⎛

⎛

2π

iϕ

k=0

⎞1/2 2 2π n−1 1 ·⎝ vk (reiϕ w2 (reiϕ ))k d ϕ⎠ . 2π 0 k=0

Now, the rest of the proof is an immediate consequence of Littlewood’s Theorem 2.1 and Rogosinski’s Theorem 2.3. Proof of Theorem 2.9 of Sheil-Small. We take g ≺ f ∈ S, g(z) = f (zω(z)), and s as above. Then we may write g(z) = zω(z)s(zω(z))2 . Taking wi = ω, i = 1, 2, 3, U = V = s in Lemma 2.10 and using the maximum principle we see that the Robertson conjecture is the decisive step needed in the proof of Theorem 2.9. Remark 2.11. It is evident that the Rogosinski conjecture follows from the SheilSmall inequality (2.9) for P (z) = z n and that the Bieberbach conjecture is the case ω ≡ 1 of the Rogosinski conjecture. Remark 2.12. The requirement of Lemma 2.10 about the functions U and V can be considerably relaxed. Namely, it is suﬃcient to suppose that the functions U and V are holomorphic in a neighbourhood of the origin, only. For such a case the proof is the same except the ﬁnal step: Instead of Rogosinski’s Theorem 2.3 we can apply Theorem 2.4.

2.4

Inverse coeﬃcients

Let Δ be the unit disc {z| |z| < 1} in the complex plane C and let S be the usual class of functions holomorphic and univalent in Δ with expansion about the origin, f (z) = z +

∞ n=2

an z n .

2.4. Inverse coeﬃcients

19

We will frequently use the following classical theorem. Theorem 2.13 (K. L¨ owner [110]). If F is the inverse of a function in S and has the expansion ∞ F (w) = w + An wn n=2

in a neighbourhood of the origin, then (2n)! 1 |An | ≤ = n!(n + 1)! n

2n n−1

(2.11)

with equality only for the inverses of the Koebe functions kd (z) =

z , |d| = 1. (1 + dz)2

(2.12)

By mathematical induction and the Stirling formula one easily gets that 1 4n (2n − 1)!! 4n 2n = ≤ n−1 n n + 1 (2n)!! (n + 1)3/2 and that 4n (2n − 1)!! 4n = n + 1 (2n)!! (n + 1)3/2

1 √ + O(1/n) π

as n → ∞.

Here, for n ∈ N we use the abbreviations (2n − 1)!! =

(2n − 1)! 2n−1 ((n − 1)!)

and (2n)!! =

(2n)! . (2n − 1)!!

The classical proof of this L¨ owner theorem can be found in [78], [70], [128] and [147]. Here we present two generalizations. Each of them implies the theorem. Let K1 be the inverse of the Koebe function k1 . We have K1 (w) = w + and

∞

(2n)! wn , n!(n + 1)! n=2

∞ 1 S(w, K1 ) := (K1 /K1 ) − (K1 /K1 )2 = 4n 6(n + 1)wn , 2 n=0

and log K1 (w) =

∞ n=1

bn w n ,

20

Chapter 2. Basic coeﬃcient inequalities

where 1 bn = n

2n n

2n−1

+2

4n = n

1 (2n − 1)!! + 2 (2n)!!

4n . n

≤

(2.13)

Moreover, one easily gets bn =

√ 4n−1/2 1 + O(1/ n) n

as n → ∞.

Since the method to ﬁnd the expression (2.13) is behind many of the coeﬃcient results of this book, we will give a short proof for (2.13). We use the Cauchy integral formula in the following way. Let ∞ K1 (w) (n + 1)bn+1 wn , = K1 (w) n=0

then (n + 1)bn+1 = =

1 2πi 1 2πi

1 = 2πi

K1 (w) 1 dw K1 (w) wn+1

k1 (∂Δr ) K1 (k1 (z)) ∂Δr K1 (k1 (z))

∂Δr

1 k1 (z) dz = − n+1 2πi k1 (z) (4 − 2z)(1 + z)2n+1 dz, (1 − z)z n+1

∂Δr

k1 (z) 1 dz k1 (z) k1n+1 (z)

where r ∈ (0, 1). Hence, we have to ﬁnd the nth Taylor coeﬃcient of ∞ ∞ 2n+1 k 2n+1 k 4+2 (4 − 2z)(1 + z) z = (1 + z) z . k=0

k=1

We get from this n−1 2n + 1 2n + 1 (n + 1)bn+1 = 4 +2 k n k=0 2n + 1 =2 + 22n . n

This implies bn =

1 n

1 2n − 1 2n 2 + 22n−1 = + 22n−1 . n−1 n n

Consider the following expansions for F , the inverse of a function in S, ∞ 1 S(w, F ) := (F /F ) − (F /F )2 = Cn wn 2 n=0

2.4. Inverse coeﬃcients

21

and log F (w) =

∞

Bn wn .

n=1

Theorem 2.14 (Klouth and Wirths [91]). If F is the inverse of a function in S, then |Bn | ≤ bn for all n ≥ 1 and |Cn | ≤ 4n 6(n + 1) for all n ≥ 0. Equality for n ≥ 2 occurs only for the functions Kd (w) = kd−1 (z), |d| = 1. Since bn are positive and each An is a polynomial with positive coeﬃcients in the Bn , Theorem 2.14 implies Theorem 2.13. Proof. The proof follows the same line as the classical proof of the Theorem 2.13. We only give here the crucial steps. A function f in S can be embedded into a subordination chain (for details see [128]). It results that the inverse function F has a representation F (w) = lim Φ(e−t w, t), t→∞

∂Φ(w, t)/∂t = w(∂Φ(w, t)/∂w)p(w, t)

where Φ(w, 0) = 0, Re p(w, t) > 0 and p(w, t) = 1 +

∞

pn (t)wn

n=1

for w ∈ Δ and t ≥ 0. Using these and setting L(w, t) := log

∞ ∂Φ(w, t) Bn (t)wn , = ∂w n=0

M (w, t) := S(w, Φ(w, t)) =

∞

Cn (t)wn ,

n=0

we get ∂ 3 (wp) ∂L ∂L ∂(wp) ∂S ∂S ∂(wp) + , = wp + , = wp + 2S w, ∂t ∂w ∂w ∂t ∂w ∂w ∂w3 and, for n ≥ 1, B0 (t) = t, Bn (t) = Cn (t) =

0

⎛ t

n−1

en(t−τ ) ⎝

0

⎞ jBj (τ )pn−j (τ ) + (n + 1)pn (τ )⎠ dτ,

j=1

⎛

t

⎞ (n + 3)! e(n−2)(t−τ ) ⎝ jCj (τ )pn−j (τ )(2n − j + 2) + pn+2 (τ )⎠ dτ. n! j=1 n−1

These formulas show that Re Bn (t), respectively Re Cn (t), is maximal for a ﬁxed t if and only if we choose Bn (t), j = 1, . . . , n − 1, respectively Cn (t), j = 0, . . . , n − 1,

22

Chapter 2. Basic coeﬃcient inequalities

for τ ∈ [0, t] real and maximal and pj (τ ) = 2 in [0, t] for any index j involved in the formulas in question. Since Bn = lim e−nt Bn (t), Cn = lim e−(n+2)t Cn (t), t→∞

t→∞

we get that the maximum of Re Bn , respectively Re Cn , is attained if and only if p(w, t) = (1 + w)/(1 − w). Clearly, the assertion of the theorem for n ≥ 1 follows from the known fact that the problems of ﬁnding the maximum of the real part and the maximum of the modulus for the given coeﬃcient are equivalent. To complete the proof we remark that the desired inequality for C0 = −6(a3 − a22 ) is given by the classical inequality |a3 − a22 | ≤ 1. For a ﬁxed p ∈ (0, 1), let Sp denote the class of all meromorphic univalent functions fp in the unit disc Δ with the normalization fp (0) = fp (0) − 1 = 0 and fp (p) = ∞. If Fp is the inverse of a function in Sp , then it admits an expansion of the form ∞ Fp (w) = w + An (p)wn n=2

in a neighbourhood of the origin. In Sp the function z

κp (z) = (1 − pz) 1 − pz plays the role of the Koebe function in S. Near w = 0 the function Kp (w) = kp−1 (w) has the expansion (see, for instance, [33]) Kp (w) = w +

∞

An (p, Kp )wn

n=2

with An (p, Kp ) =

n−1 (−1)n−1 n n p2j−n+1 . j j+1 n j=0

Theorem 2.15 (Baernstein and Schober [33])). The coeﬃcients of the function Fp satisfy the sharp inequalities |An (p)| ≤ An (p, Kp ). Equality for a single coeﬃcient holds only if Fp = Kp . Clearly, for p = 1 Theorem 2.15 gives the L¨ owner theorem. The proof of Theorem 2.15 (see [33]) is based on the following integral inequality of Baernstein [32]: for any fp ∈ Sp , 2π 2π 1 1 iθ α |fp (re )| dθ ≤ |κp (reiθ )|α dθ, 2π 0 2π 0 whenever 0 < r < 1 and −∞ < α < ∞.

2.5. Domains with bounded boundary rotation

2.5

23

Domains with bounded boundary rotation

Plane domains with bounded boundary rotation were ﬁrst studied by Paatero ([122], see also [10]). In Radon’s paper [134] one can ﬁnd general properties and applications of domains whose boundaries are rectiﬁable curves with bounded rotation. In the following we need the usual abbreviation: If the functions f and g are analytic in a neighbourhood of the origin, then f (z) =

∞

ak z k g(z) =

k=0

∞

bk z k

k=0

if and only if for all k ∈ N ∪ {0} the inequalities |ak | ≤ |bk | are valid. In [100], Lehto considered the family Vk of functions f ∈ S such that the boundary rotation of f (Δ) is at most kπ ≥ 2π. It is well known that the function k/2 ∞ 1+z 1 fk (z) = −1 =z+ An (k)z n k 1−z n=2 belongs to Vk . ∞ For any f ∈ Vk with expansion f (z) = z + n=2 an z n , it is proved that the inequality |an | ≤ An (k) is valid for all n ≥ 2 (see [45], [2] and [44] and [146] ). The crucial facts are gathered in the following two theorems. Theorem 2.16 ([45]). Let ϕ be a function holomorphic in Δ. If α ≥ 1 and ϕ(z) ≺

1 + cz 1−z

for some constant c ∈ Δ, then there exists a probability measure μ : [0, 2π] → R such that α 2π 1 + c eit z α ϕ(z) = dμ(t), z ∈ Δ. 1 − eit z 0 Theorem 2.17. For any c ∈ Δ and α ≥ 1 α α 1 + cz 1+z . 1−z 1−z Consequently, if α ≥ 1 and ϕ(z) ≺

1 + cz 1−z

24

Chapter 2. Basic coeﬃcient inequalities

for some constant c ∈ Δ, then ϕ(z) α

1+z 1−z

α .

Short proofs of these facts are found in [44] and [146]. The original proofs are given in [45] and [2]. In fact, it was shown in [45], that Theorem 2.17 implies the coeﬃcient estimate for the larger class of close-to-convex functions of order k/2 − 1 ≥ 0 introduced by Ch. Pommerenke [127]. Using Theorem 2.17 and the cited proofs in [44] and [146] we get Theorem 2.18 ([17]). If c ∈ Δ \ {−1} and α ≥ 1, then 1 c+1

1 + cz 1−z

α

−1

1 2

1+z 1−z

α

−1 .

Proof. For c ∈ Δ \ {−1} let Tc (z) = (1 + cz)/(1 − z). Since Tc (Δ) is a halfplane whose boundary cuts the real axis in a point of the interval [0, 1) and 1 ∈ Tc (Δ), there exists, for any c ∈ Δ \ {−1} and α ≥ 1, a c1 ∈ Δ such that 1

1

ϕ1 (z) := Tc (z)1− α T1 (z) α ≺

1 + c1 z . 1−z

According to Theorem 2.17 this implies ϕ1 (z)α T1 (z)α . Using this and the nonnegativity of the Taylor coeﬃcients of the functions T1 (z)α and 1/(1 − z 2 ) we get

1 + cz 1−z

α−1

1 1 = ϕ1 (z)α 2 (1 − z) 1 − z2

1+z 1−z

By integration we obtain the assertion of Theorem 2.18.

α

1 . 1 − z2

Now, we prove a subordination theorem which we need for the applications. We are concerned with the class of angular domains Πα = aHα + b (a = 0) with opening angle απ, 1 ≤ α ≤ 2, which means that there exists a linear transformation T (z) = az + b such that Πα = T (Hα ), where απ Hα = z | | arg z| < . 2 Clearly, the assertion of the following theorem is a generalization of the Carath´eodory inequality for Taylor coeﬃcients of holomorphic functions with positive real part (see [51], [52], compare also [42], p. 365).

2.5. Domains with bounded boundary rotation

25

Theorem 2.19 ([17]). For any angular domain Πα , 1 ≤ α ≤ 2, and any function g holomorphic in Δ with g(Δ) ⊂ Πα , the assertion α 1+ζ 1 (g(ζ) − g(0))λΠα (g(0)) − 1 , ζ ∈ Δ, 2α 1−ζ is valid. Proof. Let

Gα (ζ) := ΦΠα ,g(0) (ζ) − g(0) λΠα (g(0)),

ζ ∈ Δ.

The function Gα belongs to the class S and maps Δ univalently onto an angular domain Πα . This yields the existence of a complex number c ∈ Δ \ {−1} such that α 1 + cζ 1 Gα (ζ) = − 1 , ζ ∈ Δ. (c + 1)α 1−ζ Therefore, it follows from our earlier assumption that (g(ζ) − g(0))λΠα (g(0)) ≺ Gα (ζ),

ζ ∈ Δ,

and in turn that 1

φ(ζ) := (1 + (c + 1)α(g(ζ) − g(0))λΠα (g(0))) α ≺

1 + cζ , 1−ζ

ζ ∈ Δ.

According to Theorem 2.16 this implies that α 2π 1 + c eit ζ 1 − 1 dμ(t), (g(ζ) − g(0))λΠα (g(0) = (c + 1)α 0 1 − eit ζ This together with Theorem 2.18 yields the result of Theorem 2.19. Theorem 2.20 ([17]). For 1 ≤ α ≤ 2 let α 1−z 1 1− hα (z) := , 2α 1+z

ζ ∈ Δ.

z ∈ Δ.

Let further h ∈ S and h(Δ) be an angular domain Πα . We denote by h−1 the −1 function inverse to h and by h−1 (w) and α the function inverse to hα . If we let h −1 −1 −1 hα (w) represent the Taylor expansions of h and hα in a neighbourhood of the origin, then h−1 (w) h−1 α (w). Proof. Since there exists a complex number c ∈ Δ \ {−1} such that any function h of the above type may be written in the form α 1 1−z h(z) = 1− , z ∈ Δ, (c + 1)α 1 + cz

26

Chapter 2. Basic coeﬃcient inequalities

we get by a straightforward computation that 1

1 − (1 − (c + 1)α w) α

h−1 (w) =

1

1 + c(1 − (c + 1)α w) α

.

Now, we use the expansion (1 − (c + 1)α w)

1 α

∞

=

k

(−α)

1 α

(c + 1)k wk

k

k=0

and the fact that, for k ≥ 1 and α ≥ 1, Dk (α) := −

(−α)k (c + 1)k

1 α

k

≥ 0.

Hence, h

−1

(w) =

∞

k−1

Dk (α)(c + 1)

k=1

∞

w

k

k−1

Dk (α)2

w

k

1−

k=1

This completes the proof of Theorem 2.20.

1−c

∞

−1 k−1

Dk (α)(c + 1)

k=1 ∞

w

k

−1 k−1

Dk (α)2

w

k

= h−1 α (w).

k=1

Chapter 3

The Poincar´e metric For more than two thousand years, mathematicians and other people believed in the “truth” of the Euclidean parallel axiom, before Lobachevsky and Bolyai about 1825–1830 found that it is possible to get a new geometry by substituting this axiom with the following. Through a point in the plane not lying on the given straight line one can draw more than one straight line not intersecting the given line. At the beginning, the existence of such a geometry was not accepted by mathematicians, except for C. F. Gauss. In 1868 Beltrami interpreted the Lobachevsky (or hyperbolic) geometry as the natural geometry on a surface of constant negative curvature and thus proved that there exists a model for such a geometry. During this work, he was able to use the theorems of Gauss (1827) and Milding (1840) on surface geometry. Some years later, in 1871, Klein made this geometry “visible” to geometers by an interpretation in projective geometry. In accordance with his model, Klein introduced the term “hyperbolic geometry”. In 1882 Poincar´e discovered the following interpretation: the unit disc Δ equipped with the conformally invariant metric λΔ (z)|dz| :=

|dz| 1 − |z|2

can be regarded as the hyperbolic plane.

3.1

Background

Since any conformal automorphism of the unit disc Δ has the form z = T (ζ) := eiα

ζ −a , 1 − aζ

a ∈ Δ, α ∈ R,

28

Chapter 3. The Poincar´e metric

by straightforward computations one gets that the Poincar´e metric is conformally invariant in Δ, that is |dζ| |dz| = (3.1) 2 1 − |z| 1 − |ζ|2 for all z = T (ζ) in Δ. The hyperbolic distance between two points z1 , z2 ∈ Δ is deﬁned by λΔ (z) |d z|, DΔ (z1 , z2 ) = inf

(3.2)

Γ

where the inﬁmum is taken over all piecewise smooth curves Γ ⊂ Δ joining z1 with z2 . The geodesics (or paths of shortest distance) consist of the images of diameters of Δ under conformal automorphisms T . These are the diameters of Δ and the circular arcs in Δ orthogonal to its boundary ∂Δ. If these arcs are called “straightlines”, one has a model of the hyperbolic plane. Using the “straightlines”, one can ﬁnd that the hyperbolic distance for z1 , z2 ∈ Δ is given in the form z1 − z2 1 1+ρ . (3.3) , where ρ = DΔ (z1 , z2 ) = log 2 1−ρ 1 − z1 z2 An alternative interpretation of the hyperbolic geometry is given by the half plane model of Poincar´e. In the half plane H1 = {z ∈ C| Re z > 0} the element of hyperbolic arc length is λH1 (z)|dz| =

|dz| , z = x + iy ∈ H1 . 2x

In this case geodesics are horizontal lines and circles orthogonal to the imaginary axis. Using the conformal map ζ = (1 − z)/(1 + z) of the unit disc onto the half plane, it is easily seen that these two models are equivalent. The equivalence is given by the simple equation λH1 (ζ)|dζ| = λΔ (z)|dz|,

ζ=

1−z . 1+z

Now, let Ω denote a domain in C with three or more boundary points in C. According to the Riemann mapping theorem, if Ω is simply connected and z0 ∈ Ω \ {∞} is ﬁxed, then there exists a unique conformal map f0 of Δ onto Ω such that f0 (0) = z0 and f0 (0) > 0. This quantity f0 (0) is called the conformal radius of Ω at the point z0 and will be denoted in the following by R(z0 , Ω). If Ω is as above but not simply connected, the generalization of Riemann’s mapping theorem due to Poincar´e (see, for instance, [3] and [70], p. 255) asserts that there exists a unique universal covering map f0 of

3.1. Background

29

Δ onto Ω which has the same normalization as the above conformal map. In this case the quantity f0 (0) is called the hyperbolic radius (see, for instance, [34]). The Poincar´e (or hyperbolic) metric in Ω is deﬁned by the equation λΩ (z)|dz| :=

|dζ| , 1 − |ζ|2

z = f0 (ζ) ∈ Ω.

(3.4)

Concerning the universal covering map f0 for a multiply connected domain, we have to mention that f0 is a holomorphic or meromorphic function, while its inverse f0−1 (z) is multivalued. But it is well known that any conformal or universal covering map f of Δ onto Ω is given by iα ζ − a , a ∈ Δ, α ∈ R. f (ζ) = f0 (T (ζ)) = f0 e 1 − aζ This together with (3.1) imply that the density λΩ (z) is well-deﬁned by equation (3.4), i.e., it depends neither on the choice of the mapping function nor on the choice of the branch of f −1 . Further, the metric λΩ (z)|dz| is conformally invariant. From the above deﬁnitions it follows that R(z, Ω) =

1 = |f (ζ)|(1 − |ζ|2 ), λΩ (z)

(3.5)

where f is a conformal or universal covering map of Δ onto Ω and z = f (ζ), ζ ∈ Δ, z ∈ Ω. Taking partial derivatives of R by using (3.5) and the Wirtinger calculus 2

2

∂ R(z, Ω) ∂ R(z, Ω) ∂ R(z, Ω) = −i , z = x + iy ∈ Ω, ∂z ∂x ∂y

∂ R(z, Ω) ∂ R(z, Ω) ∂ R(z, Ω) = +i =: ∇R(z, Ω), z = x + iy ∈ Ω, ∂z ∂x ∂y

leads to the formulas ∂R(z, Ω) |f (ζ)| = ∂z f (ζ) and R(z, Ω)

1 − |ζ|2 f (ζ) −ζ 2 f (ζ)

(3.6)

2 ∂ 2 R(z, Ω) ∂ R(z, Ω) = − 1. ∂z∂z ∂z

Hence, the conformal (or hyperbolic) radius R = R(z, Ω) satisﬁes Liouville’s equation RR = |∇R|2 − 4, (3.7) where R = R(z, Ω), R denotes the Laplacian of R and ∇R its gradient.

30

Chapter 3. The Poincar´e metric

The most known forms of the equation (3.7) are the following nonlinear elliptic equation u = −4e−2u , where u = u(x, y) := − log λΩ (z) = log R(z, Ω) for z = x + iy ∈ Ω, and an equivalent formula K = e2u u = −4 that deﬁnes the Gaussian curvature of the metric. Thus, the Liouville equation for R is equivalent to the fact that the hyperbolic metric deﬁned as above has the Gaussian curvature K = −4 (see, for instance, [3], [34] and [70]). Remark 3.1. Clearly, to deal with the hyperbolic metric with Gaussian curvature K = −1, one has to choose the density of the metric as 2λΩ (z), where λΩ (z) is deﬁned as above.

3.2

The Schwarz-Pick inequality

We begin by the classic Schwarz lemma. Theorem 3.2 (Schwarz’s lemma). Let f be holomorphic in the unit disc Δ. If f (0) = 0 and |f (z)| is bounded by 1 in Δ, then |f (z)| ≤ |z|,

z ∈ Δ,

(3.8)

and |f (0)| ≤ 1

(3.9)

with equality in (3.8) for some z = 0 or in (3.9) occurring if and only if f (z) = cz for some unimodular complex constant c. As is indicated in [3], p. 21, the Schwarz lemma and its classic proof are due to Carath´eodory [53]; Schwarz proved it only for univalent mappings [149]. In [124], [125] Pick realized the invariant character of Schwarz’s lemma. Theorem 3.3 (Pick [125], see also Lindel¨ of [105]). A holomorphic mapping f of the unit disc Δ into itself satisﬁes the inequalities f (z ) − f (z ) z − z 1 2 1 2 (3.10) , z1 , z2 ∈ Δ, ≤ 1 − f (z1 )f (z2 ) 1 − z1 z2 and

|f (z)| 1 ≤ , 1 − |f (z)|2 1 − |z|2

z ∈ Δ.

(3.11)

Nontrivial equality holds if and only if f is a conformal automorphism of Δ, i.e., f (z) = eiα (z − a)/(1 − az), a ∈ Δ, α ∈ R.

3.2. The Schwarz-Pick inequality

31

It is evident that inequalities (3.10) and (3.11) correspond to Schwarz’s inequalities (3.8) and (3.9), respectively. Geometrically, inequality (3.10) means that any holomorphic mapping f of the unit disc Δ into itself decreases the hyperbolic distance between two points. Also, (3.11) implies that f decreases the hyperbolic lengths of an arc and the hyperbolic area of a set. The following assertion is called the hyperbolic metric principle (see, for instance, [70] and [120]). Theorem 3.4. Let Ω and Π be domains in the extended complex plane such that each of them has at least three boundary points. If f ∈ A(Ω, Π), then (i) L(f (γ)) ≤ L(γ) for any rectiﬁable curve γ ⊂ Ω and its image f (γ) ⊂ Π with length L(γ) and L(f (γ)) in the hyperbolic metric of Ω and Π, respectively; equality holds if and only if f (z) = f2 (f1−1 (z)), where f1 and f2 are conformal (universal covering) maps of the unit disc onto Ω and Π, respectively. (ii) the diﬀerential length elements at z ∈ Ω and at its image w = f (z) ∈ Π satisfy the inequality λΠ (w)|dw| ≤ λΩ (z)|dz| (3.12) with the same condition for equality. We shall call the Schwarz-Pick inequality the equation (3.12) written in the form |f (z)| ≤

λΩ (z) . λΠ (f (z))

(3.13)

The Schwarz-Pick inequality applied to the function f ∈ A(Ω , Ω) deﬁned by ζ = f (z) = z, z ∈ Ω , immediately gives the following comparison of densities. Theorem 3.5. If Ω ⊂ Ω, then λΩ (z) ≥ λΩ (z) for any point z ∈ Ω . Explicit formulas for the Poincar´e density are known for several domains. We shall present three of them. They will be useful in applications of the comparison theorem 3.5. 1. For angular domains Hα = {z ∈ C \ {0} | |arg z| < easily gets 1/λHα (z) = R(z, Hα ) = 2 α r cos

απ 2 },

α ∈ (0, 2], one

θ , z = reiθ ∈ Hα , α

and

4 πy cos , z = x + iy, |y| < 1 π 2 for the strip H0 = {z ∈ C | |Im z| < 1}. 1/λH0 (z) = R(z, H0 ) =

2. For the punctured unit disc Δ = {z ∈ C| 0 < |z| < 1}, 1/λΔ (z) = R(z, Δ ) = 2|z| log

1 , z ∈ Δ . |z|

32

Chapter 3. The Poincar´e metric

3. It is a much harder problem to ﬁnd an explicit formula for the density of the domain C \ {0, 1} = C \ {0, 1, ∞}. The following formula is due to Agard [1]: |z||z − 1| 1 dudv = , w = u + iv. λC\{0,1} (z) π C |w||w − 1||w − z| The classical representation formula for λC\{0,1} is given by λC\{0,1} (z) =

1 |g (z)| = , 2|λ (τ )| Im τ 2 Img(z)

where λ(τ ) is the elliptic modular function, g is its inverse, τ ∈ Λ := {τ | Imτ > 0} and z ∈ C \ {0, 1}. The modular function λ : Λ → C \ {0, 1} is a universal covering map which maps the triangle {τ ∈ Λ | 0 < Reτ < 1, |τ − 1/2| > 1/2} onto Λ in such a way that λ(0) = 1, λ(1) = ∞, λ(∞) = 0, and 8 ∞ 1 + q 2n λ(τ ) = 16 q , q = eπiτ (3.14) 2n−1 1 + q n=1 (see, for instance, [107]). The Liouville equation for the hyperbolic density λC\{0,1} (z) and formula (3.14) are used in a proof of some monotonicity properties of the metric. Theorem 3.6 (see A. Bermant [39], J. A. Hempel [80] and compare also A. Yu. Solynin and M. Vuorinen [154] for further results). Let z = x + iy = reiθ . The Poincar´e density λC\{0,1} (z) has the following properties: y

∂λC\{0,1} (z) < 0, ∂y

y = 0,

(3.15)

∂λC\{0,1} (z) < 0, 0 < |θ| < π, (3.16) ∂θ ∂λC\{0,1} (z) λC\{0,1} (z) log r < 0, r = 1. (3.17) + ∂r r There are many generalizations and applications of the classical Theorems 3.2 and 3.3. Especially, there are close relations between Theorems 3.2 and 3.3 in the theory of unimodular bounded holomorphic functions due to Carath´eodory, Denjoy, Julia and Wolﬀ. We cite the following statements, which sometimes are quoted as the Grand Iteration Theorem (see [150], p. 78 and compare also [131], p. 82, and [55]). θ

Theorem 3.7. Let ϕ be a holomorphic self-map of Δ that is not a conformal automorphism of Δ with a ﬁxed point in Δ. Then there is a unique point ω ∈ Δ such that the iterates ϕn (z) = (ϕ ◦ · · · ◦ ϕ)(z) converge to ω as n → ∞ uniformly on any compact subset of Δ and

3.3. Estimates using the Euclidean distance

33

(C1 ) if ω ∈ Δ, then ϕ has no ﬁxed point in Δ \ {ω}, and ϕ(ω) = ω, 0 ≤ |ϕ (ω)| < 1, (C2 ) if ω ∈ ∂Δ (Denjoy-Wolﬀ point of ϕ), then ϕ has no ﬁxed point in Δ and ϕ and ϕ have angular limits at ω such that ϕ(ω) = ω, 0 < ϕ (ω) ≤ 1, where

ω+z ϕ (ω) = sup Re ω−z z∈Δ

−1 ω + ϕ(z) Re . ω − ϕ(z)

For further deep results on the boundary behavior of bounded holomorphic functions we refer the reader to the books by L. V. Ahlfors [3], by J. B. Garnett [66] and by Ch. Pommerenke [131] (see also R. B. Burckel [48] and G. M. Goluzin [70] for other developments and applications of Theorems 3.2 and 3.3.)

3.3

Estimates using the Euclidean distance

Choosing Ω as the disc with center z and radius δ(z) = dist(z, ∂Ω) in Theorem 3.5, one gets Theorem 3.8. If δ(z) denotes the distance from z ∈ Ω to the boundary ∂Ω of Ω, then λΩ (z) ≤ 1/δ(z) for any z ∈ Ω. The next assertion is known as the Koebe 1/4-theorem (for a proof see [60], [70], [78] or [128]). Theorem 3.9. If Ω is a simply connected domain in C, then δ(z) λΩ (z) ≥ any z ∈ Ω.

1 4

for

We will need Landau’s theorem on holomorphic functions that omit two ﬁxed values. More precisely, the Landau theorem concerns functions f (z) =

∞

an z n

n=0

holomorphic in the unit disc Δ and omitting the values 0 and 1 in Δ. By our notation, f ∈ A(Δ, C \ {0, 1}). In the sequel, the known constant 1 2λC\{0,1} (−1)

= iλ (1 + i) =

Γ(1/4)4 = 4.3768796 . . . 4π 2

is used. J. A. Hempel and J. A. Jenkins established an explicit sharp bound in the Landau theorem.

34

Chapter 3. The Poincar´e metric

Theorem 3.10 (Landau; see J. A. Hempel [80] and J. A. Jenkins [87] for proofs). If the function f is holomorphic and omits 0 and 1 in Δ, then Γ(1/4)4 |a1 | ≤ 2|a0 | |log |a0 || + . (3.18) 4π 2 Equality holds only for the universal covering map f : Δ → C\{0, 1} with a0 = −1. Proof by J. A. Hempel [80]. By the hyperbolic metric principle we have the sharp inequality |a1 | = |f (0)| ≤ 2/ρ(a0 ), ρ(z) := 2λC\{0,1} (z). Thus, Theorem 3.10 states that 1

2 = ≤ 2|z| λC\{0,1} (z) ρ(z)

1 |log |z|| + ρ(−1)

.

(3.19)

According to formula (3.15) of Theorem 3.6, to prove inequality (3.19) it is suﬃcient to consider real values of z = x + iy = reiθ lying in the interval (−∞, 0). Let us introduce the real function w(σ) := u(eσ+iπ ) + σ,

σ := log r, u := log ρ.

Equation (3.19) is equivalent to e−w(σ) ≤ |σ| + e−w(0) .

(3.20)

From (3.17) of Theorem 3.6 it follows that σw (σ) < 0 for σ = 0. Again from (3.16) of Theorem 3.6 we deduce that ∂ 2 u/∂θ2 ≥ 0 for θ = π. This together with Liouville’s equation ∂2u ∂2u + 2 = e2(σ+u) ∂σ 2 ∂θ give ∂ 2 u/∂σ 2 ≤ e2(σ+u) , θ = π, which is equivalent to the inequality w ≤ e2w . Integrating this inequality in (−∞, 0] and [0, ∞) using the local behavior w → 0 and e2w = r2 ρ2 (z) → 0 as σ → ∞ (see [80] for details), we have (3.20). This completes the proof of Theorem 3.10. Remark 3.11. Since the function 1 − f (z) = 1 − a0 − a1 z −

∞ n=2

an z n

3.3. Estimates using the Euclidean distance

35

also omits 0 and 1, we can take 1 − a0 in (3.18) instead of a0 . Consequently, Theorem 3.10 implies Γ(1/4)4 |a1 | ≤ . min 2|ζ| |log |ζ|| + 4π 2 ζ∈{a0 ,1−a0 } We shall consider an important application of Landau’s theorem in the theory of uniformly perfect sets. Let Ω ⊂ C be an open set with more than one boundary point in C. The annulus An(r, s; a) = {z | r < |z − a| < s} ⊂ Ω with center a is said to separate the compact set E =C\Ω in the Riemann sphere, whenever both components of C \ An(r, s; a) have nonempty intersections with E. Since in the next section we are concerned with conformal images of such annuli, called conformal annuli, we will characterise annuli of the above form as genuine annuli. Following Ch. Pommerenke (see [129] and also [36], [37], [55], [65], [130]), we deﬁne the maximum modulus M0 (Ω) as the supremum of the moduli of all annuli An ⊂ Ω such that An separates E and the center a of An belongs to ∂Ω. As usual the modulus of an annulus An is deﬁned by M (An(r, s; a)) =

1 s log . 2π r

We take M0 (Ω) = 0, if Ω contains no genuine annulus centered at a boundary point and separating E. Theorem 3.12 (Beardon and Pommerenke, compare [37] and [129]). Let Ω ⊂ C be an open set with more than one boundary point in C. If λΩ (z) is the density of the Poincar´e metric deﬁned in each component of Ω with curvature K = −4 and M0 (Ω) is ﬁnite, then for any z ∈ Ω, 1 (Γ(1/4))4 . ≤ π M0 (Ω) + 2λΩ (z) dist(z, ∂Ω) 4π 2

(3.21)

Proof. Consider the function z − a | |z − a| = dist(z, ∂Ω), a ∈ ∂Ω, b ∈ ∂Ω . βΩ (z) := min log b − a We ﬁx z in Ω and choose points a and b in ∂Ω such that z − a . |z − a| = dist(z, ∂Ω), βΩ (z) = log b − a

36

Chapter 3. The Poincar´e metric

Applying the comparison theorem 3.5 and using the function g(w) = (w − a)/(b − a), w ∈ Ω, one obtains λC\{0,1} (g(z)) ≤ λg(Ω) (g(z)) = |b − a|λΩ (z) which is equivalent to 1 1 ≤ . 2λΩ (z) dist(z, ∂Ω) 2|g(z)|λC\{0,1} (g(z)) This together with Landau’s inequality (3.19) give 1 1 ≤ βΩ (z) + . 2λΩ (z) dist(z, ∂Ω) ρ(−1)

(3.22)

Consider now any point a ∈ Ω such that |z − a | = δ, where δ = dist(z, ∂Ω). Clearly, if the circle {w | |w − a | = δ} meets ∂Ω, then βΩ (z) = 0. Otherwise there exists a maximal non-empty annulus of the form An = {w | δe−m < |w − a | < δem }, such that

a ∈ ∂Ω,

An ⊂ Ω,

m/π ≤ M0 (Ω).

As An is maximal, there is a point b ∈ ∂Ω ∩ ∂An, consequently z − a = m ≤ πM0 (Ω). βΩ (z) ≤ log b − a This together with inequality (3.22) give inequality (3.21). The proof of Theorem 3.12 is complete.

Theorems 3.10 and 3.12 are connected with applications of the Poincar´e metric in problems of function theory and of mathematical physics. As an example we mention a result on Hardy type inequalities. Let C0∞ (Ω) be the usual family of smooth functions with compact support in Ω. For p ∈ [1, ∞) we shall consider the Hardy constant cp (Ω) deﬁned by | f ∈ C0∞ (Ω), (∇f )/δ 2/p−1 p =1 , cp (Ω) = sup f /δ 2/p p L (Ω)

L (Ω)

where δ = dist(x + iy, ∂Ω). Theorem 3.13 (Avkhadiev [12], see also [13]). Let Ω ⊂ C be an open set with more than one boundary point in C. For any p ∈ [1, ∞),

(Γ(1/4))4 min{2, p}M0 (Ω) ≤ cp (Ω) ≤ 2p π M0 (Ω) + 4π 2

2 .

(3.23)

3.4. An application of Teichm¨ uller’s theorem

37

The proof of the upper bound in (3.23) uses the inequality (3.21). The key role in this proof is played by the following lemma. Lemma 3.14 (see [12], p. 10 and [13], Theorem 3). Let Ω ⊂ C be an open set of the Riemann sphere with more than two boundary points in C. If 1 ≤ p < ∞, then for any f ∈ C0∞ (Ω),

p p |∇f |p |f |p λ2Ω dx dy ≤ (λΩ δ)2−2p dx dy, (3.24) 2−p 2 Ω Ω δ where δ = dist(x + iy, ∂Ω) and λΩ = λΩ (x + iy). Remark 3.15. In [13], the formula (4) contains a misprint. Namely, the power 2p − 2 has to be replaced by 2 − 2p. The inequality (3.24) is a generalization of a known fact for simply or doubly connected domains. Namely, if any component of Ω can be mapped conformally onto either the unit disc Δ or an annulus of the form {z | q < |z| < 1}, then for any f ∈ C0∞ (Ω),

p p p 2 |f | λΩ dx dy ≤ |∇f |p λ2−p dx dy. (3.25) Ω 2 Ω Ω The inequality (3.25) can be derived using the original Hardy inequality and the conformal invariance of the hyperbolic metric (see, for instance, [4], [11], [12], [13], [63]). Unfortunately, in the general case, where no constraints on the components of Ω are imposed, inequality (3.25) does not hold even for (p/2)p replaced by any other ﬁnite constant. For instance, an inequality of the form (3.25) does not hold for Ω = C \ {0, 1}.

3.4

An application of Teichm¨ uller’s theorem

As is observed in the book [55] of Carleson and Gamelin (see p. 64): A simple scaling and normal families argument shows that conformal annuli of large modulus contain genuine annuli of large modulus. As usual, the modulus M (D) of a conformal annulus, or a doubly connected domain, equals the modulus of the genuine annulus that is conformally equivalent to D (see for instance [3] and [97]). Let Ω ⊂ C be an open set with more than one boundary point in C. A compact set E = C \ Ω is said to be uniformly perfect if M (Ω) = sup{M (D) | D ⊂ Ω doubly connected and separating E} is ﬁnite. The most known quantity describing the geometry of uniformly perfect sets was deﬁned by Pommerenke in [129]: A compact set E on the Riemann sphere

38

Chapter 3. The Poincar´e metric

C that contains the point at inﬁnity is uniformly perfect if there exists a constant c ∈ (0, 1) such that, for every z0 ∈ E \ {∞} and every r ∈ (0, ∞), the set E ∩ An(cr, r; z0 ) is not empty. Let C0 (E) be the supremum of the admissible constants c ∈ (0, 1) for a uniformly perfect set. It is evident that C0 (E) = e−2π M0 (Ω) ,

Ω = C \ E,

where M0 (Ω) is the quantity deﬁned in Section 3.3 as the supremum of the moduli of all genuine annuli An ⊂ Ω such that An separates E and the center of An belongs to ∂Ω (see [129] and also [36], [37], [55], [65], [130]). We take M (Ω) = 0 for open sets that have simply connected components only and M0 (Ω) = 0, if Ω contains no genuine annulus centered at a point of ∂Ω and separating E. Since any genuine annulus is a conformal annulus, the inequality M0 (Ω) ≤ M (Ω)

(3.26)

is trivial and the above remark of Carleson and Gamelin implies M (Ω) = ∞

=⇒

M0 (Ω) = ∞.

(3.27)

On the other hand, using his theorem on extremal moduli, Teichm¨ uller proved in [162] that any doubly connected domain D with M (D) > 1/2 contains a circle which separates the components of its complement. Moreover, the center of this circle may be chosen at E1 , the bounded component of C \ D, and the radius of the circle equals the diameter of E1 . Solynin (see [153]) and Herron, Liu, and Minda (see [82]) proved that a separating circle exists for doubly connected domains D with M (D) > 1/4, if one does not restrict the position of the center of the circle. In [82] it is also proved that such D contain separating annuli An of modulus M (An) = M (D) − c (c ≈ 0.46), where again no restriction is imposed on the center of the √ annulus. In the same paper, they prove that M (An) ≥ M (D) − c˜ (˜ c = π1 log(2 2 + 2) ≈ 0.501) if one ﬁxes the center of An on ∂D. In the following we will use the result of Teichm¨ uller to derive a new quantitative version of the Carleson-Gamelin remark, where we ﬁnd 1/2 to be the best possible constant in a similar inequality, not only choosing the center of An on ∂D but ﬁxing it as Teichm¨ uller did. We will apply this result to diﬀerent characterizations of uniformly perfect sets and to estimates of the second derivative of functions analytic on open sets with uniformly perfect boundaries. In addition to (3.26) and (3.27) we prove that M (Ω) ≤ M0 (Ω) +

1 2

(3.28)

for any open set Ω ⊂ C. The proof is based on the above cited theorem of Teichm¨ uller and on a related formula of Ahlfors in [3].

3.4. An application of Teichm¨ uller’s theorem

39

We are concerned with the following theorem of Teichm¨ uller (see [162] and also [3]). Theorem 3.16. Of all doubly connected domains that separate the pair {−1, 0} from a pair {w0 , ∞} with |w0 | = R, the one with the biggest modulus is the complement of the union of the segments [−1, 0] and [R, ∞]. Clearly, to prove the inequality (3.28) for any open set Ω it is suﬃcient to consider the special case where Ω is a doubly connected domain. Hence, the inequality (3.28) is a corollary of the following extension of Teichm¨ uller’s result on the existence of a separating circle. Theorem 3.17 ([27]). Any doubly connected domain D ⊂ C with M (D) > 1/2 contains an annulus An = An(r, s; z0 ) such that M (An) = M (D) − 1/2, the center z0 belongs to the bounded component E1 of C \ D and diam E1 = r. The constant 1/2 is sharp. Proof. Let E = C \ D = E1 ∪ E2 , where E1 and E2 are the connected components of E. Without loss of generality we may assume that the diameter of E1 equals 1 and that the points z = 0 and z = −1 belong to E1 . Now, we consider the annulus An(1, m; 0), where m = exp(2π(M (D) − 1/2)). It has the following properties: An(1, m; 0) ∩ E1 = ∅, the center of An(1, m; 0) belongs to E1 , and M (An(1, m; 0)) = M (D) − 1/2. If An(1, m; 0) ⊂ D, then An(1, m; 0) separates E and there is nothing to prove. Let us assume that this is not true, i.e., that there exists a point w0 ∈ E2 such that 1 < |w0 | < m. Taking |w0 | = R we consider the Teichm¨ uller annulus D(R) = C \ ([−1, 0] ∩ [R, ∞)) with the modulus Λ(R) (compare [3]). According to Theorem 3.16 the inequality M (D) ≤ Λ(R)

(3.29)

is valid. On the other hand, we will prove that the assumed condition 1 < R < m implies that 1 1 Λ(R) < log R + < M (D). (3.30) 2π 2 Clearly, (3.30) contradicts (3.29). Hence, to complete the proof of Theorem 3.17 it is suﬃcient to show that the ﬁrst inequality in (3.30) holds for any R > 1. The function Λ(R) is implicitly deﬁned by the following formula due to Ahlfors (see [3], p. 75, also compare formula (3.14)): 8 ∞ 1 1 − q 2n−1 R = , 16 q n=1 1 + q 2n

q = e−2πΛ(R) .

(3.31)

It is known that Λ(1) = 1/2 and that Λ(R) → ∞ as R → ∞. If one considers the ﬁrst equation in (3.31) as the deﬁnition of a function R(q), it is evident that qR(q)

40

Chapter 3. The Poincar´e metric

is a decreasing function of q for 0 < q < e−π = e−2πΛ(1) . The function qR(q) decreases from 1/16 to exp(−π), when q increases from 0 to exp(−π). In particular, 16

1, and (3.30) follows. The sharpness of the constant 1/2 is a consequence of Teichm¨ uller’s considerations, since in the case M (D) = 1/2 there does not exist a separating annulus with the above properties. This completes the proof of Theorem 3.17. In contrast to the conformal characteristic M (Ω) the quantity M0 (Ω) is not a conformal invariant. Nevertheless, it may be useful to remark the following consequence of Theorem 3.17. Corollary 3.18. Let Ω1 and Ω2 be two conformally equivalent domains in C. If M0 (Ω1 ) is ﬁnite, then 1 |M0 (Ω1 ) − M0 (Ω2 )| ≤ . 2 Remark 3.19. For any genuine annulus An, the inequality M0 (An) < M (An) holds, since we consider only those separating genuine annuli in An, which have a center lying on ∂An. Geometrically it is evident that a genuine annulus An has such a separating annulus, if and only if 1 1 log 3 < M (An) ≤ M0 (An) + log 3. 2π 2π

3.5

Domains with uniformly perfect boundary

There are about twenty characterizations of domains with uniformly perfect boundary via the Hayman-Wu condition, domains with strong barrier, local behavior of harmonic measure or logarithmic capacity, etc. (see, for instance, [9], [36], [37], [55], [62], [65], [67], [76], [77], [79], [129], [130]). In this section we continue to use the Pommerenke characteristic C0 (E) = exp(−2π M0 (Ω)), Ω = C \ E. Our aim is to consider the problem of comparing the Euclidian geometry characteristic M0 (Ω) of open sets with uniformly perfect boundary with the following characteristics of the hyperbolic geometry on Ω: α(Ω) = inf{λΩ (z)dist(z, ∂Ω) | z ∈ Ω}

(3.32)

3.5. Domains with uniformly perfect boundary and

41

γ(Ω) = sup ∇λ−1 Ω (z) | z ∈ Ω ,

(3.33)

where λΩ is the density of the Poincar´e metric with curvature −4 deﬁned on the components of Ω. We will prove that 1 sup{M0 (Ω)α(Ω)} = (3.34) 4 and that 1 M0 (Ω) = , sup (3.35) γ(Ω) 4 where the supremum is taken with respect to all open sets Ω ⊂ C that have more than one boundary point in C. In the proof of (3.34) and (3.35) we show that (γ(An))2 1 = + (M (An))2 16 4 for any genuine annulus An and that 4M (An) ∼

1 α(An)

for genuine annuli if M (An) → ∞. We must confess that the above considerations have their origin in the central question of this book, compare section 1.2. Let Ω and Π be open sets in C or in C and let A(Ω, Π) be the set of functions f : Ω → Π locally holomorphic or meromorphic and in general multivalued. What can be said about the inﬂuence of the geometric properties of Ω and Π on the quantities f (n) (z) λΠ (f (z)) Cn (Ω, Π) = sup n | z ∈ Ω, f ∈ A(Ω, Π) ? n! (λΩ (z)) The above considerations give us the possibility to determine bounds for C2 (Ω, Π) in terms of the quantities M0 (Ω) and M0 (Π) that are “visible” in Euclidean geometry. During these proofs (see Chapter 6) we get a sharp form of the OsgoodJørgensen inequality (see [88] and [121]), namely sup{sup{|∇ log λΩ (z)|dist(z, ∂Ω) | z ∈ Ω}} = 2,

(3.36)

where the ﬁrst supremum is taken with respect to all hyperbolic domains Ω ⊂ C. Firstly, we cite some known results. In [121], Osgood proved that 1 2 ≤ γ(Ω) ≤ , α(Ω) α(Ω)

Ω ⊂ C.

From Theorem 3.12 of Beardon and Pommerenke, it follows that (Γ(1/4))4 1 , ≤ 2π M0 (Ω) + α(Ω) 2π 2

Ω ⊂ C.

(3.37)

42

Chapter 3. The Poincar´e metric

In [37], the ﬁrst paper on uniformly perfect sets, it is also proved that M0 (Ω) ≤

1 , 2α(Ω)

Ω ⊂ C.

(3.38)

The following theorem assures that the equations (3.34) and (3.35) hold, where (3.34) is the sharp form of (3.38). Theorem 3.20 ([27]). Let Ω ⊂ C be an open set with more than one boundary point in C. If α(Ω) > 0, then the inequalities M0 (Ω)

0. Denoting by M the modulus of such an annulus, i.e., M := M (An) = we get 1 = 4 M |z| sin λAn (z)

1 log(1/), 2π

log(1/|z|) 2M

,

z ∈ An,

(see, for instance, [37]). Hence, d (1/λAn (z)) =: |s(t)|, |∇(1/λAn (z))| = d |z| where s(t) = 4 M sin t − 2 sin t,

with t =

log(1/|z|) ∈ (0, π). 2M

By a straightforward calculation we derive

γ(An) = max{|s(t)| | t ∈ [0, π]} = |s(t0 )| = 2 1 + 4M 2 , tan t0 = −2M,

which completes the proof of Lemma 3.21. The formulas (3.39), (3.42) and Theorem 3.17 immediately imply lim

→0+

M0 (An(, 1; 0)) 1 = . γ(An(, 1; 0)) 4

(3.43)

From the second inequality of (3.39) we see that lim→0+ M0 (An(, 1; 0))α(An(, 1; 0)) ≤

1 . 4

44

Chapter 3. The Poincar´e metric

Now, let us assume that the second equation in (3.40) is not true, i.e., lim→0+ M0 (An(, 1; 0))α(An(, 1; 0))

0, f (ζ)

ζ ∈ Δ,

which is the classical condition for a conformal map f : Δ → Ω to have a convex image. These give the following theorem. Theorem 3.23 (L¨ owner). A proper subdomain Ω of the plane C is convex if and only if sup |∇R(z, Ω)| ≤ 2. z∈Ω

We have proved the following analog of Theorem 3.23. Theorem 3.24 ([15], [20]). Let Ω ⊂ C be a simply connected domain with more than one boundary point. The set E = C \ Ω is convex if and only if inf |∇R(z, Ω)| ≥ 2.

z∈Ω

The second motivation is given by applications of the gradients to SchwarzPick type inequalities (see Chapter 4, Section 6, below) and by a number of theorems and their applications on the geometry of the surface SΩ = {(w, h) | w ∈ Ω, h = R(w, Ω)} (see [49], [74], [76], [89], [95], [119], [168], [169]). As an example we mention Theorem 3.25. A domain Ω ⊂ C is convex if and only if R(·, Ω) is a concave function on Ω. In the paper [95], Kovalev studied an analog of Theorem 3.25 for unbounded simply connected domains Ω ⊂ C. He proved that C \ Ω is a convex set if and only if the function R(·, Ω) is a locally convex function. In [20] we developed these facts using the following observation. Since the Jacobian of the gradient of R is J(z, Ω) =

∂ 2 R(z, Ω) ∂ 2 R(z, Ω) − ∂x2 ∂y 2

∂ 2 R(z, Ω) ∂x∂y

2 , z = x + iy ∈ Ω,

and a condition necessary for a real-analytic function to be convex or concave on Ω is the inequality J(z, Ω) ≥ 0, z ∈ Ω, we observe that Theorem 3.25 and its generalizations are related to ∇R. Moreover, we consider the image of the domain Ω by the gradient (see [20] for details). We present here the case only when Ω is a polygonal domain.

46

Chapter 3. The Poincar´e metric

Using the Wirtinger calculus, the gradient can be written as a complex variable function ∇R(z, Ω) =

∂ R(z, Ω) ∂ R(z, Ω) ∂ R(w, Ω) +i =2 , z = x + iy ∈ Ω. ∂x ∂y ∂z

We begin with a simple example. Let z = f (ζ) = ζ + 1/ζ, ζ ∈ Δ. One has Ω := f (Δ) = C \ [−2, 2] and ∇R(·, Ω)) is a diﬀeomorphism. Therefore, to ﬁnd the gradient image it is suﬃcient to ﬁnd its boundary. We have 3 1 − ζζ 1 g(ζ) = ∇R ζ + , Ω = 2 |1 − ζ 2 | 2 , ζ ∈ Δ. ζ ζ(1 − ζ )2 Hence,

lim g(ζ) =

ζ→eiθ

− 2i for any θ ∈ (0, π), 2i for any θ ∈ (π, 2π).

Moreover, it is clear that g(ζ) has no limit as ζ → ±1. Straightforward computations show that the set of all limit values of g(ζ) as ζ → 1, ζ ∈ Δ, is a curve γ1 given by the parametric equation t t it , t ∈ (−π, π). 2 cos − i sin w1 (t) = 2 e 2 2 This is one branch between the two contact points 2i and −2i of an epicycloid that is described by a point on a circle of radius 2 rolling on another circle of radius 2. Since g(ζ) = −g(−ζ), the gradient image of Ω is the set of all points outside the two branches of the above epicycloid, where the second branch is described by w2 (t) = −w1 (π + t), t ∈ (−π, π). One may observe that γ1 coincides with the set of values of ∇R for an angular domain with opening angle 2π, which is a branch of an epicycloid. This observation can be extended: If Ω is a polygonal domain, then the gradient image is bounded by branches of epicycloids or hypocycloids. To avoid confusion we want to mention that these epicycloids diﬀer from those occurring in the Ptolemaic system. Theorem 3.26 ([20]). Let Ω be a simply connected domain in C or in C. If the boundary of Ω is a polygon with inner angles παk , vertices zk and sides (zk , zk+1 ), k = 1, . . . , n, zn+1 = z1 , then (a) ∇R(·, Ω) is a real-analytic function on Ω \ {z1 , . . . , zn , ∞}, (b) the gradient image of a side (zk , zk+1 ) is a point wk such that |wk | = 2, (c) the set of all limit values of ∇R as z → zk is a curve γk that joins wk−1 with wk and is deﬁned by a parametric equation π π wk (t) = 2 ei(ck +αk t) (αk cos t − i sin t), t ∈ [− , ], 2 2 where ck is a real constant.

3.6. Derivatives of the conformal radius

47

Using Theorem 2.14 we shall obtain sharp estimates for higher-order derivatives of the conformal radius.. Let us introduce the functions μk (z, Ω) :=

∂ k log R(z, Ω)−2 1 , R(z, Ω)k (k − 1)! ∂z k

k = 1, . . . , n,

z ∈ Ω.

Also, for a ﬁxed a ∈ Ω we consider μk = μk (a, Ω) and the quantities τn,n−1 (α), τn,n−2 (α), . . . , τn,0 (α)), deﬁned by the following recurrent formulas: τk,k (α) = 1 (0 ≤ k ≤ n), τk,0 (α) =

τm,k (α) =

m−k+1 s=1

k−1 α μk−s τs,0 (α) , 1 ≤ k ≤ n, k s=0

1 τs−1,0 (1)τm−s,k−1 (α), s

2 ≤ k ≤ m ≤ n.

(3.45)

(3.46)

In the case 0 ≤ k ≤ n − 1, τn,k (α) depends on n, k, α, a and Ω. For instance, if Ω = Δ and a ∈ Δ, then n + 2α − 1 τn,k (α) = an−k . n−k Theorem 3.27 ([14]). If Ω is a simply connected proper subdomain of C, then 1 sup sup |μk (a, Ω)| = μk (a, C \ [ , ∞)) 4 Ω a∈Ω

(3.47)

and sup sup |τn,k (α)| = Ω a∈Ω

n−k s=0

2n + 3α − 1 s

n−k−s+α−2 n−k−s

.

(3.48)

Proof. Let ψ be the conformal mapping of Ω onto Δ, ψ(a) = 0, ψ (a) = λΩ (a) > 0. By deﬁnition, λΩ (z) = |ψ (z)|(1 − |ψ(z)|2 )−1 , z ∈ Ω. This yields 2ψ(z) ψ (z) ∂ log λ2Ω (z) ψ (z). = − ∂z ψ (z) 1 − ψ(z)ψ(z) Hence λkΩ (a)μk (a, Ω) =

1 1 ∂ k log λ2Ω (a) = k (k − 1)! ∂a (k − 1)!

ψ (z) ψ (z)

(k−1) |z=a

(3.49)

48 and

Chapter 3. The Poincar´e metric ∞

ψ (z) k+1 λΩ (a)μk+1 (a, Ω)(z − a)k = ψ (z)

(3.50)

k=0

in some neighbourhood of a. According to Theorem 2.14 by Klouth and Wirths and the equation (3.50), 1 |μk (a, Ω)| ≤ μk (0, C \ ( , ∞)), 4

k = 1, 2, . . . .

Equality for k ∈ N occurs if and only if {a, Ω} is a linear transformation of {0, C\[ 14 , ∞)}. The conformal mapping Φ0 : Δ → C\[ 14 , ∞), Φ0 (0) = 0, Φ0 (0) > 0, is given by the Koebe function Φ0 (ζ) = ζ(1 + ζ)−2 . It is not diﬃcult to verify by direct calculations that 0 τn,k (α) =

n−k s=0

2n + 3α − 1 s

n−k−s+α−2 n−k−s

for the Koebe domain C \ [ 14 , ∞) at the point z = 0. From (3.47) we obtain (3.48) since τn,k (α) is a polynomial with positive coeﬃcients in the μ1 , . . . , μn . For n = 1, we get |τ1,0 (α)| = α|grad λ−1 Ω (a)|. Hence, |τ1,0 (α)| ≤ 4α by the classical Koebe constant 1/4.

Chapter 4

Basic Schwarz-Pick type inequalities Let Ω ⊂ C and Π ⊂ C be two domains equipped by the Poincar´e metric. We are concerned with the set A(Ω, Π) = {f : Ω → Π} of functions locally holomorphic or meromorphic in Ω and, in general, multivalued. Let λΩ (z), z ∈ Ω, and λΠ (w), w ∈ Π, denote the density of the Poincar´e metric at z ∈ Ω and w ∈ Π, respectively. Consider the functional Ln (f, z, Ω, Π) deﬁned by (n) n f (z) (λΩ (z)) = Ln (f, z, Ω, Π) , z ∈ Ω, f ∈ A(Ω, Π), n ∈ N. n! λΠ (f (z)) Many problems in geometric function theory are devoted to the problem of determining Mn (z, Ω, Π) := sup {Ln (f, z, Ω, Π) | f ∈ A(Ω, Π)}, and, respectively, Cn (Ω, Π) := sup {Mn (z, Ω, Π) | z ∈ Ω}. It is clear that Cn (Ω, Π) is not dependent on f and z ∈ Ω and represents the smallest number possible in the inequality (n) n f (z) (λΩ (z)) ≤ Cn (Ω, Π) , z ∈ Ω, f ∈ A(Ω, Π). n! λΠ (f (z)) The classical Schwarz-Pick lemma says that M1 (z, Δ, Δ) = C1 (Δ, Δ) = 1 and in turn M1 (z, Ω, Π) = C1 (Ω, Π) = 1 for any pair (Ω, Π) of hyperbolic domains in the extended complex plane, as we have discussed above.

50

Chapter 4. Basic Schwarz-Pick type inequalities

We shall consider the problem of determining Cn (Ω, Π) for all n ≥ 2. In the proofs we will frequently use the fact that the functions Mn and Cn are invariant under linear transformation of domains. This means that the equations Mn (z, Ω, Π) = Mn (az + b, aΩ + b, cΠ + d) and Cn (Ω, Π) = Cn (aΩ + b, cΠ + d) are valid, where aΩ + b = {az + b | z ∈ Ω} and cΠ + d = {cw + d | w ∈ Π} for some a, c ∈ C \ {0}, b, d ∈ C. Also, the following fact deserves reader’s attention. For technical reasons we explain a simple case, when Π is a bounded domain and z ∈ Ω ⊂ C. Normal family arguments show that there exists a sequence fk ∈ A(Ω, Π) such that Mn (z, Ω, Π) = lim Ln (fk , z, Ω, Π), k→∞

and fk converges to a holomorphic function f0 uniformly in the interior of the domain Ω. Clearly, there are two possible cases to distinguish: The limit function f0 belongs to the family A(Ω, Π) or f0 (z) ≡ const. ∈ ∂Π, and consequently, f0 ∈ A(Ω, Π). The second case is typical in our problems. If n ≥ 2, then there is no extremal function f ∈ A(Ω, Π) such that Mn (z, Ω, Π) = Ln (f, z, Ω, Π) except in some very special cases.

4.1

Two classical inequalities

Let f ∈ A(Δ, Δ). A consequence of |f (z0 )| ≤

1 − |f (z0 )|2 , 1 − |z0 |2

z0 ∈ Δ,

is the sharp inequality |f (z0 )| ≤

1 , 1 − |z0 |2

z0 ∈ Δ.

Equality is attained for conformal automorphisms f of Δ such that f (z0 ) = 0. In 1920, Sz´ asz extended the latter inequality to higher-order derivatives. Theorem 4.1 (O. Sz´asz [160]). For any f ∈ A(Δ, Δ), z ∈ Δ, and m ∈ N, the sharp inequality 2 m (2m + 1)! m (2m+1) (z) ≤ |z|2k f 2m+1 k (1 − |z|2 ) k=0 is valid. Equality occurs only for the functions m+1 ζ −z iγ m , f (ζ) = e ζ 1 − zζ

γ ∈ R.

4.1. Two classical inequalities

51

Proof. We consider the function g ∈ A(Δ, Δ) deﬁned by ∞ ζ +z ak ζ k , ζ ∈ Δ. = g(ζ) = f 1 + zζ k=0

According to the Cauchy formula, integration along a circle Γ around the origin lying in its neighbourhood and the use of the variable substitution ξ =

ζ +z = S(ζ) 1 + zζ

results in the following chain of equations: 1 f (ξ) dξ g(ζ)(1 + zζ)n−1 dζ 1 f (n) (z) = . = n! 2πi S(Γ) (ξ − z)n+1 2πi Γ ζ n+1 (1 − |z|2 )n For n = 2m + 1 using |g(ζ)| ≤ 1 we obtain 2π 1 1 g(ζ)(1 + zζ)2m dζ ≤ |1 + zeiθ |2m dθ. 2π 2π Γ ζ 2m+2 0

(4.1)

(4.2)

Clearly, formulas (4.1) and (4.2) imply the desired estimate by the binomial and Parseval formulas. A little examination of the inequality (4.2) gives that equality can occur only for the function m ζ +z iγ m+1 g(ζ) = e ζ , γ ∈ R. 1 + zζ This completes the proof of Theorem 4.1.

Remark 4.2. In [160], during the proof of the theorem Sz´ asz gave the formula n n n−1 1 − |z|2 ak z n−k (4.3) f (n) (z) = n−k n! k=1

as a consequence of equation (4.1). Also, in [160] by a similar proof Sz´ asz obtained an explicit sharp bound for |f (z)|. But the problem of ﬁnding an explicit formula for {max |f (2m) (z)| | f ∈ A(Δ, Δ)}, z ∈ Δ, in the case m ≥ 2 is still open (2008). Concerning Sz´asz’s majorant for n = 2m + 1, one easily obtains that 2π 2π 1 (2m − 1)!! 1 |1 + zeiθ |2m dθ = |1 + eiθ |2m dθ = 22m sup . 2π 0 (2m)!! z∈Δ 2π 0 Moreover, for n = 2m the proof gives the estimate 2m 1 − |z|2 22m (2m − 2)!! |f (2m) (z)| ≤ sup (2m)! π (2m − 1)!! z∈Δ which is not sharp, at least for n = 2.

52

Chapter 4. Basic Schwarz-Pick type inequalities

E. Landau remarked in [98] that a special case of the following theorem, namely, Theorem 7.1 (see below), is a consequence of the validity of the Bieberbach conjecture. Theorem 4.3. Let Π be a simply connected proper subdomain of C and z ∈ Δ. Then Mn (z, Δ, Π) ≤ Mn (z, Δ, H2 ) = (n + |z|)(1 + |z|)n−2 , where H2 = C \ (−∞, 0]. Proof. We use the fact that de Branges’ proof of the Bieberbach conjecture implies a proof of the generalized Bieberbach or Rogosinski conjecture. This means that the Taylor coeﬃcients of a function subordinate to a schlicht function are dominated by the Taylor coeﬃcients of the Koebe function. Therefore, for the Taylor coeﬃcients of a function f ∈ A(Δ, Π), the inequalities λΠ (a0 ) |ak | ≤ k are valid. Using the latter inequalities and formula (4.3), we get Mn (z, Δ, Π) ≤

n n−1 k|z|n−k = (n + |z|)(1 + |z|)n−2 . n−k

k=1

Straightforward computations show that Ln (g0 , z, Δ, H2 ) = Mn (z, Δ, H2 ) = (n + |z|)(1 + |z|)n−2 , where 1 g0 (ζ) = 4

1 + ζ |z|/z 1 − ζ |z|/z

2 −1

=

∞

k(ζ |z|/z)k .

k=1

This completes the proof of Theorem 4.3.

Corollary 4.4. If Π is a simply connected proper subdomain of C, then Cn (Δ, Π) ≤ Cn (Δ, H2 ) = 2n−2 (n + 1).

4.2

Theorems of Ruscheweyh and Yamashita

According to our notation Cn (Ω, Π), the identity Cn (Δ, Π) = 2n−1 has been proved by St. Ruscheweyh (see [142] and [143]) in two basic cases, when Π is a half plane or a disc. Here we present the original versions of his theorems.

4.2. Theorems of Ruscheweyh and Yamashita

53

Theorem 4.5 (St. Ruscheweyh [143] (1985)). Let Δ = {ζ | |ζ| < 1}. For any f ∈ A(Δ, Δ), z ∈ Δ, and n ∈ N the sharp inequality 1 (n) 1 − |f (z)|2 f (z) ≤ n n! (1 − |z|) (1 + |z|) is valid. This inequality was conjectured by Ruscheweyh in 1974 (see [142]). Proof. Let the function g(z) =

∞

ak z k

k=0

be holomorphic and |g(z)| ≤ 1 in the unit disc. It is well known that the coeﬃcients of such a function satisfy the inequalities |a0 | ≤ 1,

|ak | ≤ 1 − |a0 |2

(4.4)

for any k ≥ 1. The inequalities (4.4) imply the assertion of the theorem immediately for z = 0. We now suppose that z = 0 and deﬁne g(ζ) as in the proof of Theorem 4.1. Using the Sz´ asz formula (4.3) and (4.4) together with equations f (z) = h(0) = b0 , one easily gets n n 1 − |z|2 (n) n−1 n−1 |z|n−k = (1 − |b0 |2 ) (1 + |z|) f (z) ≤ (1 − |b0 |2 ) n−k n! k=1

which is the inequality to prove. We have to show that the inequality of Theorem 4.5 is sharp for n ≥ 2 and any z ∈ Δ, z = 0. To this end we choose a sequence ak in Δ such that ak → z/|z| as k → ∞ and consider unimodular bounded holomorphic functions z − ak fk (z) = . 1 − ak z Straightforward computations give (n)

|fk (z)|

2

1 − |fk (z)|

=

n!|ak |n−1 n! → n−1 2 n−1 |1 − ak z| (1 − |z| ) (1 − |z|) (1 − |z|2 )

as k → ∞. This completes the proof of Theorem 4.5.

Theorem 4.6 (St. Ruscheweyh [142]). Let f be holomorphic in Δ, n ∈ N, and ρ(f (z)) denote the minimal distance from f (z) to the boundary of the closed convex hull of f (Δ). Then the sharp inequality 2 ρ(f (z)) 1 (n) f (z) ≤ n! (1 − |z|)n (1 + |z|) is valid. Equality occurs for conformal maps of the unit disc onto a halfplane.

54

Chapter 4. Basic Schwarz-Pick type inequalities

Remark 4.7. For H1 = {z | Re z > 0} this implies that for any f ∈ A(Δ, H1 ), z ∈ Δ and n ∈ N the inequality n 1 (n) (λΔ (z)) f (z) ≤ (1 + |z|)n−1 n! λH1 (f (z))

holds. Proof of Theorem 4.6. (As St. Ruscheweyh indicated in [142], he owes the idea of the following proof to T. Sheil-Small.) Without loss of generality we can suppose that Re f (ζ) > 0 in Δ, f (0) = 1 and ρ(f (z)) = Re f (z) for a given point z ∈ Δ.Using the Herglotz formula

2π

1 + ζe−it dμ(t), 1 − ζe−it

f (ζ) = 0

where μ(t) is a nondecreasing function with μ(2π) − μ(0) = 1, we get f (n) (ζ) = 2n!

2π

0

e−int dμ(t). (1 − ζe−it )n+1

Consequently, by simple estimates and the Poisson formula |f (n) (z)| ≤

2n! (1 − |z|)n−1 (1 − |z|2 )

2π

0

2n! Ref (z) 1 − |z|2 dμ(t) = . |1 − ze−it |2 (1 − |z|)n−1 (1 − |z|2 )

Since 1 − |z| = |1 − ze−it | is true only for e−it = e−it0 = z/|z|, it is easily seen that equality can occur only for a piece-wise constant function μ(t) such that μ([0, 2π]) = {0, 1}, and, consequently, the corresponding function f0 has the form f0 (ζ) =

1 + ζe−it0 , 1 − ζe−it0

eit0 = z/|z|,

and f0 maps Δ onto H1 . This completes the proof of Theorem 4.6.

Let Ω be a hyperbolic domain in C, i.e., Ω has more than two boundary points. For a universal covering map ϕ from Δ onto Ω and given z ∈ Ω, let ρΩ (z) denote the greatest r ∈ (0, 1] such that ϕ is univalent in the non-Euclidean disc ζ −w 0}.

4.3. Pairs of simply connected domains

55

Theorem 4.8 (S. Yamashita [170]). Let Ω be a hyperbolic domain in C, let z ∈ Ω, and let n ≥ 2. For any holomorphic function f : Ω → C with positive real part in Ω the sharp inequality n λΩ (z) 1 (n) 2n − 1 Re (f (z)) f (z) ≤ 2 n n! ρΩ (z) is valid. Also, in [170] Yamashita established the case of equality in Theorem 4.8. In particular, if Ω = H2 and f (ζ) = ζ −1/2 , then the equality is attained at the points ζ = z = x > 0. Taking into account the classical equality 1/λH1 (w) = 2Re w and the invariance of Cn under linear transformation of domains, one obtains the equality 2n − 1 (a = 0). Cn (aH2 + b, H1 ) = n Clearly, if Ω is a simply connected domain, then ρΩ (z) = 1 for any point z ∈ Ω. Therefore, this particular case of Yamashita’s theorem can be presented as follows. Theorem 4.9. If Ω is a simply connected proper subdomain of C, then 2n − 1 Cn (Ω, H1 ) ≤ Cn (H2 , H1 ) = . n In fact, Theorem 4.9 is equivalent to Theorem 4.8. Actually, suppose that Ω is a multiply connected domain in C and ϕ as above. The covering map ϕ1 : Δ → Ω deﬁned by ζ +w ϕ1 (ζ) = ϕ 1 + wζ is univalent in the disc Δρ = {ζ | |ζ| < ρΩ (z)}. The domain Ωρ := ϕ(Δρ ) is a simply connected proper subdomain of C and λΩ (z) = λΩρ (z). ρΩ (z) Applying Theorem 4.9 to the function f | Ωρ in the simply connected domain Ωρ one immediately obtains Theorem 4.8. The original proof of Yamasita of these theorems is based on coeﬃcient estimates of univalent functions by Chua (see [56] and [170]). We will obtain Theorem 4.9 as a special case of our theorem proved in Section 4 of the present chapter.

4.3

Pairs of simply connected domains

We again will need the domain H2 = C \ (−∞, 0]. In this section we will discuss the following theorem.

56

Chapter 4. Basic Schwarz-Pick type inequalities

Theorem 4.10 ([16]). If Ω and Π are simply connected proper subdomains of C, then the sharp estimate Cn (Ω, Π) ≤ Cn (H2 , H2 ) = 4n−1 is valid. Proof. For f ∈ A(Ω, Π), z0 ∈ Ω, we consider the functions s(ζ) := (ΦΩ,z0 (ζ) − z0 ) λΩ (z0 ), and

ζ ∈ Δ,

t(ζ) := ΦΠ,f (z0 ) (ζ) − f (z0 ) λΠ (f (z0 )),

ζ ∈ Δ.

Both of them belong to the class S of functions univalent in Δ and normalized in the origin as usual. The fact that f (Ω) is a subset of Π may be expressed in terms of the function u(ζ) := (f (ΦΩ,z0 (ζ)) − f (z0 )) λΠ (f (z0 )),

ζ ∈ Δ.

It means that u(ζ) is subordinate to t(ζ). Using the Taylor expansion u(ζ) =

∞

ak λΠ (f (z0 ))ζ k

k=1

and the function ΨΩ,z0 inverse to ΦΩ,z0 we get f (z) = f (z0 ) +

∞

k

ak (ΨΩ,z0 (z))

k=1

and therefore

n (n) 1 f (n) (z0 ) k ak . = (ΨΩ,z0 (z)) n! n! z=z0 k=1

−1

If we denote by s

(w) the function inverse to s(ζ), deﬁne

∞ k s−1 (w) = Am,k (z0 )wm m=k

and use we see that

s−1 (λΩ (z0 )(z − z0 )) = ΨΩ,z0 (z), (n) 1 k n = An,k (z0 ) (λΩ (z0 )) . (ΨΩ,z0 (z)) n! z=z0

4.3. Pairs of simply connected domains

57

Hence, we get the formula f (n) (z0 ) n ak An,k (z0 ) (λΩ (z0 )) , = n! n

(4.5)

k=1

which will be central in what follows. Let K1 be the inverse of the Koebe function k1 (z) =

z . (1 + z)2

We have K1 (w) = w +

∞

Am (K1 ) wm ,

Am (K1 ) =

m=2

(2m)! , m!(m + 1)!

and, by the L¨ owner theorem 2.13, the sharp estimates |Am,1 | ≤ Am (K1 ) =

(2m)! m!(m + 1)!

are valid. Let us deﬁne Am,k (K1 ) by expansions k

(K1 (w)) =

∞

Am,k (K1 ) wm

m=k

in some neighbourhood of the origin. It is evident that An,k (K1 ) is a polynomial with positive coeﬃcients in Am (K1 ), 2 ≤ m ≤ n, and An,k (z0 ) is the same polynomial in Am,1 (z0 ), 2 ≤ m ≤ n. Accordingly, we have |An,k (z0 )| ≤ An,k (K1 ),

1 ≤ k ≤ n.

For those quantities An,k (K1 ) we get 2k(2n − 1)! k 2n = An,k (K1 ) = . n − k n (n − k)!(n + k)!

(4.6)

(4.7)

This is an immediate consequence of the Cauchy formula according to 1 1 (K1 (w))k (1 + ζ)2n−1 (1 − ζ) dw = dζ, An,k (K1 ) = 2πi k1 (∂Δr ) wn+1 2πi ∂Δr ζ n−k+1 where r ∈ (0, 1) and ∂Δ = {ζ | |ζ| = r} (for (4.6) and (4.7) compare [56]). Further, the function u is subordinate to the function t univalent in Δ and normalized as usual. According to the Rogosinski conjecture settled by de Branges’

58

Chapter 4. Basic Schwarz-Pick type inequalities

proof of the Bieberbach conjecture, the Taylor coeﬃcients of the function u satisfy the inequalities |ak | λΠ (f (z0 )) ≤ k, k ∈ N. Using these estimates and the basic formula (4.5), we obtain |f (n) (z0 )| λΠ (f (z0 )) k An,k (K1 ), n ≤ n! (λΩ (z0 )) n

k=1

where An,k (K1 ) are given by formula (4.7). Using again the Cauchy formula to compute the latter sum, we get n k2 2n = 4n−1 n−k n

k=1

(for the latter formula see also [56]). To complete the proof we have to show that Cn (H2 , H2 ) = 4n−1 , where H2 = C\(−∞, 0]. To this end we consider f0 (z) = 1/z, z ∈ H2 . For any z = x > 0 using λH2 (x) = 1/(4x) and λH2 (1/x) = x/4 one easily gets (n) |f0 (x)| λH2 (1/x)) n−1 . n = 4 n! (λH2 (x)) This completes the proof of Theorem 4.10.

We will see below that it is possible to ﬁnd lower bounds for punishing factors Cn (Δ, Π) using the properties of hyperbolic metrics if the boundaries of the domains in question are “nice” (see Section 5.5). This was the reason why we tried to ﬁnd lower bounds for punishing factors for pairs of simply connected domains by considering the limiting processes going to the boundaries. Our result is the following theorem. Theorem 4.11 (see [16]). Let Ω and Π be two simply connected proper subdomains of C whose boundaries contain sectorial accessible analytic arcs. Then, for any n ≥ 2, the assertion 2n−1 ≤ Cn (Δ, Π) ≤ Cn (Ω, Π) holds. Equality occurs if Ω = Δ and Π is convex. As usual, a boundary arc is said to be sectorial accessible if any point on this arc is the vertex of an open triangle contained in the domain in question (see for instance [131]). The proof for this result is very technical. Therefore we omit the details here. We hope that one of our readers will be able to give a proof for the following conjecture. Conjecture (see [16]). Given n ≥ 3, then Cn (Ω, Π) ≥ 2n−1 for all simply connected domains Ω and Π in C.

4.4. Holomorphic mappings into convex domains

4.4

59

Holomorphic mappings into convex domains

We shall consider a holomorphic function f : Ω → Π, where Π is a proper convex subdomain of the plane and Ω is the unit disc or a simply connected domain. The following theorem is a generalization of the Carath´eodory inequality ([51], [52], see also [42], p. 365, [48], p. 213) for Taylor coeﬃcients of holomorphic functions with positive real part. Also, in [83] Herzig gives a description of extremal functions. Theorem 4.12 (see Rogosinski [139]). Let g and g0 be holomorphic functions with expansions g(z) =

∞

an z n ,

g0 (z) = z +

n=1

∞

bn z n ,

z ∈ Δ.

n=2

Suppose that g0 is univalent in Δ and Π := g0 (Δ) is a convex domain. If g ≺ g0 , then |an | ≤ 1 for any n ≥ 1. Equalities |ak | = 1 for all k = 1, 2, . . . , n with some n ≥ 2 occur if and only if g is a conformal map of Δ onto a half plane. Proof. For n ∈ N we consider the function gn (z) :=

n

∞

k=1

k=2

1 g(ζ k z 1/n ) = an z + ank z k , n

z ∈ Δ,

where ζ = e2πi/n . Since wk = g(ζ k z) ∈ Π for any z ∈ Δ, from the convexity of Π it follows that (w1 + · · · + wn )/n ∈ Π. Hence, the function gn is subordinate to g0 . Applying the Schwarz lemma to the function g0−1 (gn (z)), one immediately gets the desired inequality |an | ≤ 1 for any n ≥ 1. If |a1 | = 1, then Schwarz’s lemma implies that g(z) = g0 (cz) , where |c| = 1. If |a1 | = |a2 | = 1 , then g(z) = g0 (cz) and |b2 | = 1. Consequently, g is a conformal map of the unit disc onto a half plane according to the classical L¨ owner theorem on coeﬃcients of convex univalent functions. The proof is complete. The assertion of the following theorem immediately implies that Mn (Δ, Π) = 2n−1 for any convex domain Π. Theorem 4.13 ([16]). Let Π be a convex proper subdomain of C and let n ∈ N. Then for any z ∈ Δ the equation Mn (z, Δ, Π) = (1 + |z|)n−1

(4.8)

is valid. In the case z = 0 and n ≥ 2, there exist extremal functions for which Ln (f, z, Δ, Π) = (1 + |z|)n−1 if and only if Π is a half plane.

60

Chapter 4. Basic Schwarz-Pick type inequalities

Proof. We ﬁx z ∈ Δ and consider f ∈ A(Δ, Π). Introduce the function f1 ∈ A(Δ, Π) deﬁned by ∞ ζ +z = ak ζ k , ζ ∈ Δ. f1 ζ) = f 1 + zζ k=0

Using the formula (4.3) we obtain n n−1 n−k Mn (z, Δ, Π) ≤ sup |z| |ak |λΠ (a0 ) | g ∈ A(Δ, Π) . n−k

(4.9)

k=1

From Theorem 4.12 for the function g(ζ) = λΠ (a0 )f1 (ζ) it follows that λΠ (a0 )|ak | ≤ 1,

k ∈ N.

This together with inequality (4.9) give n n−1 |z|n−k = (1 + |z|)n−1 . Mn (z, Δ, Π) ≤ n−k k=1

To prove Mn (z, Δ, Π) ≥ (1 + |z|)n−1 we ﬁrst consider the case z = 0. If Φ denotes the conformal map of the unit disc Δ onto Π with Φ(0) = w0 and Φ (0) = 1/λΠ (w0 ) > 0, it is obvious that the function f deﬁned by f (ζ) = Φ(ζ n ) has the desired property: f (n) (0)λΠ (w0 )/n! = 1. The inequalities for coeﬃcients in Theorem 4.12 occur for all k = 1, 2, . . . , n with n ≥ 2 if and only if g is a conformal map of the unit disc onto a half plane. Therefore the existence of a function f for which Ln (f, z, Δ, Π) = (1 + |z|)n−1 in the case z = 0 is possible only if Π is a half plane and this is the case in Theorem 4.5 of Ruscheweyh. In the general case we only know that there is a maximising sequence (fk ) ⊂ A(Ω, Π) such that Mn (z, Δ, Π) = lim Ln (fk , z, Δ, Π). k→∞

Without loss of generality, we can suppose that the sequence (fk ) converges uniformly in the interior of the unit disc. Let f0 (z) = lim fk (z), k→∞

z ∈ Δ.

If z = 0, n ≥ 2, and Π is not a half plane, then the proof gives that f0 ∈ A(Δ, Π). Hence, f0 (z) ≡ const. ∈ ∂Π. To prove, nevertheless, the sharpness of (4.8) for any convex domain Π, we present such a sequence explicitly using the convexity of Π and the linear invariance of Mn (z, Δ, Π). Clearly, there exist a point w1 ∈ ∂Π, a disc Δ1 and a half plane H such that Δ1 ⊂ Π ⊂ H and w1 ∈ ∂Δ1 ∩∂Π∩∂H. Without loss of generality we may suppose that Δ1 = {w | |w − 1| < 1} ⊂ Π ⊂ H1 .

4.4. Holomorphic mappings into convex domains

61

The origin belongs to the boundaries of Δ1 , Π and H1 . Now we use a reﬁnement of an idea presented in the proof of Theorem 4.5 by Ruscheweyh. Namely, for ﬁxed z ∈ Δ \ {0}, we consider the sequence z 1 αk = 1 − , k ∈ N, k + 1 |z| of complex numbers and the sequence of fk , k ∈ N, of conformal maps of Δ onto Δ1 deﬁned by αk ζ − αk fk (ζ) = + 1, ζ ∈ Δ. |αk | 1 − αk ζ A straightforward computation using 1 = 2 Re w − |w|2 , λΔ1 (w) yields

w ∈ Δ1 ,

(1 − |z|2 )n−1 = (1 + |z|)n−1 . k→∞ |1 − αk z|n−1

lim Ln (fk , z, Δ, Δ1 ) = lim

k→∞

Theorem 3.5 assures that 1− Since

|w|2 λH1 (w) λΠ (w) = ≤ ≤ 1, 2 Re w λΔ1 (w) λΔ1 (w)

w ∈ Δ1 .

|fk (z)|2 = 0, k→∞ 2 Re (fk (z)) lim

it is evident that lim Ln (fk , z, Δ, Π) = lim Ln (fk , z, Δ, Δ1 )

k→∞

k→∞

λΠ (fk (z)) = (1 + |z|)n−1 . λΔ1 (fk (z))

This completes the proof of Theorem 4.13.

Finally, we consider pairs (Ω, Π) with simply connected Ω and convex Π. We will need the constant 2n − 1 Cn (H2 , Λ) = n from Theorem 4.9 of Yamashita. By mathematical induction and Stirling’s formula it can be shown that 1 (2n − 1)!! 4n−1/2 2n − 1 √ + O(1/n) as n → ∞, = 4n−1/2 = √ n (2n)!! n π and that

(2n − 1)!! 4n−1/2 (n ≥ 2). < √ (2n)!! n+1 Theorem 4.9 is a special case of the following assertion. 2n−1 < 4n−1/2

62

Chapter 4. Basic Schwarz-Pick type inequalities

Theorem 4.14 ([17]). Let Ω and Π be simply connected proper subdomains of C. If Π is a convex domain, then Cn (Ω, Π) ≤ Cn (H2 , Π) = 4n−1/2

(2n − 1)!! . (2n)!!

Proof. We follow the proof of our Theorem 4.10 with the same notation but some little changes. Since Π is a convex domain and u ≺ t, we can use Theorem 4.12 of Rogosinski to obtain that |ak | λΠ (f (z0 )) ≤ 1. This together with formula (4.5) give |f (n) (z0 )| λΠ (f (z0 )) |An,k (z0 )|. n ≤ n! (λΩ (z0 )) n

k=1

As in the proof of Theorem 4.10 we get |f (n) (z0 )| λΠ (f (z0 )) An,k (K1 ). n ≤ n! (λΩ (z0 )) n

k=1

Summing up gives n k=1

An,k (K1 ) =

1 2πi

∂Δr

(1 + ζ)2n−1 (1 − ζ)n dζ = ζn

2n − 1 n

,

which proves the desired inequality (2n − 1)!! 2n − 1 = 4n−1/2 . Cn (Ω, Π) ≤ n (2n)!! For Π = H1 the lower estimate Cn (H2 , Π) ≥ 4n−1/2

(2n − 1)!! (2n)!!

is given by the example of f (ζ) = ζ −1/2 at the points ζ = z = x > 0. To obtain the lower estimate for any convex domain Π, we proceed by approximation as in the proof of Theorem 4.13. Namely, we consider Δ1 = {w | |w − 1| < 1} ⊂ Π ⊂ H1 and the function f ∈ A(H2 , Δ1 ) ⊂ A(H2 , H1 ) deﬁned by f (ζ) = √

2 , ζ +1

√

1 = 1.

Straightforward computations using the comparison theorem and the asymptotic behavior of the nth derivative of this function at the point z = x > 0, as x → +∞, xn+1/2 |f (n) (x)| (2n − 1)!! =2 + o(1) n! (2n)!!

4.5. Punishing factors for convex pairs

63

gives that |f (n) (x)|λΠ (f (x)) (2n − 1)!! = 4n−1/2 . x→+∞ n! (λH2 (x))n (2n)!! lim

This completes the proof of Theorem 4.14.

4.5

Punishing factors for convex pairs

The main aim of the present section is to prove that Cn (Ω, Π) = 2n−1 for any pair (Ω, Π) of convex domains ( see [24], see also Chua [56], Li [102], [103], and Yamashita [169] for special cases). Theorem 4.15 ([24]). Let Ω and Π be two convex proper subdomains of C and let f ∈ A(Ω, Π), n ∈ N. Then for any z0 ∈ Ω the inequality |f (n) (z0 )| (λΩ (z0 ))n ≤ 2n−1 n! λΠ (f (z0 ))

(4.10)

is valid. The constant 2n−1 can not be replaced by a smaller one independent of f ∈ A(Ω, Π) and z0 ∈ Ω for any pair (Ω, Π) of convex domains. Proof. In the following we consider only the cases n ≥ 2, since the case n = 1 is given by the Schwarz-Pick lemma. We ﬁrst prove that Cn (Ω, Π) ≤ 2n−1 . According to the central formula (4.5) we have to prove that n ck An,k ≤ 2n−1 . k=1

In the present theorem, the functions s and t belong to the class K of functions univalent in Δ that map Δ onto a convex domain and are normalized as usual, e.g., t(0) = 0 and t (0) = 1. Moreover, ck are the Taylor coeﬃcients of a holomorphic function subordinate to the function t, and An,k is the n-th Taylor coeﬃcient of the power F k (w), where F is the inverse of the function s. Since the extreme points of the closed convex hull of the set {u | u ≺ t for some t ∈ K} are the functions

xz , 1 − yz

z ∈ Δ,

for ﬁxed (x, y) ∈ ∂Δ × ∂Δ (see [75], Theorem 5.21), it remains to prove that n n k k xy An,k ≤ y An,k ≤ 2n−1 , y ∈ ∂Δ. Re k=1

k=1

64

Chapter 4. Basic Schwarz-Pick type inequalities

Let an−k,n be deﬁned by the Taylor expansion

z s(z)

n =

∞

aν,n z ν .

ν=0

Then the Schur-Jabotinsky theorem (compare for example [81], Theorem 1.9.a) implies that for 1 ≤ k ≤ n the identities An,k =

k an−k,n n

are valid. Hence, we have to prove that n−1 n − l n−l al,n ≤ 2n−1 , y n

y ∈ ∂Δ.

(4.11)

l=0

The tool for the proof of (4.11) is the resulting representation

z s(z)

n = (1 + zω(z))n = 1 +

n n z σ (ω(z))σ , σ

z ∈ Δ,

σ=1

where |ω(z)| ≤ 1 for any z ∈ Δ by the Marx-Strohh¨ acker theorem, that indicates Re(s(z)/z) > 1/2 in Δ (see [115] and [155], see also Section 7.3 below). If we deﬁne (ω(z))σ =

∞

dj,σ z j ,

z ∈ Δ,

j=0

we get the following formula for the sum appearing in (4.11): n−1 l=0

n−1 1 n n−1 n − l n−l al,n = y n + (n − j)y n−j dj−σ,σ . y σ n n σ=1 j=σ

(4.12)

To have the desired inequality it is suﬃcient to prove that n−1 n−j (n − j)y d j−σ,σ ≤ n − σ. j=σ Now we observe that the functions ω σ also map the disc Δ into Δ. Therefore, we may replace the coeﬃcients dj−σ,σ by the coeﬃcients dj−σ of a unimodular bounded function when we consider the latter estimate. Taking p = n − σ gives us an equivalent inequality p−1 p−τ dτ ≤ p. (p − τ ) y τ =0

4.5. Punishing factors for convex pairs

65

This inequality follows directly from Fej´er’s inequality p−1 (p − τ )dτ ≤ p, τ =0

which has long been known to be valid (see [61] and [159]). This concludes the proof of the inequality (4.11). Now, we shall prove that the constant 2n−1 is best possible in any of the cases in question. Lemma 4.16. Let Ω and Π be two convex proper subdomains of C. Then for any n ≥ 2 the inequality Cn (Ω, Π) ≥ 2n−1 is valid. Proof. We know that the constant Cn (Ω, Π) is invariant under linear transformations of Ω and Π. Hence, without restriction of generality, we may assume that Δ1 = {z | |z − 1| < 1} ⊂ Ω ⊂ H1 = {z | Re z > 0}

(4.13)

Δ 1 ⊂ Π ⊂ H1 .

(4.14)

and Let α ∈ (0, 1) and ξ ∈ (0, 1) and consider the function fα (z) = α

z+2 , z+α

z ∈ Ω.

Obviously, fα ∈ A(Ω, Π). By the comparison theorem applied to the inclusion relations (4.13) and (4.14) we get lim λΩ (β) 2β = lim λΠ (β) 2β = 1.

β→0+

β→0+

Now, by use of these asymptotic equalities, we prove Lemma 4.16 with the following chain of inequalities and equations. (n) fα (ξ) λΠ (fα (ξ)) Cn (Ω, Π) ≥ lim lim n ξ→0+ α→0+ n! (λΩ (ξ)) (n) fα (ξ) (2ξ)n = lim lim ξ+2 ξ→0+ α→0+ n! 2α ξ+α α(2 − α) (2ξ)n 2n = lim = 2n−1 . ξ→0+ α→0+ (ξ + α)n+1 2α ξ+2 ξ→0+ 2 + ξ ξ+α

= lim

lim

This concludes the proof of Theorem 4.15.

66

Chapter 4. Basic Schwarz-Pick type inequalities

Remark 4.17. The last part of the proof shows that the constant 2n−1 is approached for any pair of convex domains, when z0 and f (z0 ) approach the boundaries of Ω and Π at certain points. But there are simple special cases where the constant is attained at inner points. For instance, this happens if Ω and Π are half planes. Actually, if Ω = Π = H1 and f0 (z) = 1/z, then, at any point z0 = x > 0, (n)

n

2n−1 (λΩ (x)) f0 (x) 1 = n+1 = , n! x λΠ (1/x) since 1/λH1 (z) = 2 Re z.

4.6 Case n = 2 for all domains Let Ω and Π be hyperbolic domains on the Riemann sphere C that are equipped with the Poincar´e metric of curvature −4. According to Poincar´e’s generalization of Riemann’s mapping theorem, this means that the boundaries of Ω and Π contain at least three points in C and that the density of this metric in the unit disc Δ = {z | |z| < 1} is deﬁned as λΔ (z) =

1 , 1 − |z|2

z ∈ Δ.

We have considered the functionals Ln , Mn and Cn for several special cases. According to [16], in the general case these functionals depend on the hyperbolic characteristics −k

(λΩ (z))

k ∂ k log (λΩ (z)) −k ∂ log (λΠ (w)) and (λ (w)) , 1 ≤ k ≤ n − 1. Π ∂z k ∂wk

Everything is clear in the case n = 2. In particular, the following theorem describes C2 (Ω, Π) in terms of the gradients −1

∂ log (λΩ (z)) ∂z

−1

∂ log (λΠ (w)) . ∂w

∇ (1/λΩ (z)) = −2 (λΩ (z)) and ∇ (1/λΠ (w)) = −2 (λΠ (w))

Theorem 4.18 ([27]). For all hyperbolic domains Ω ⊂ C and Π ⊂ C, 1 sup |∇ (1/λΩ (z))| + sup |∇ (1/λΠ (w))| . C2 (Ω, Π) = C2 (Π, Ω) = 2 z∈Ω w∈Π Proof. Fix (z, w) ∈ Ω×Π, z = ∞ and w = ∞. We consider a function f ∈ A(Ω, Π) such that f (z) = w. Let Φ : Δ → Ω and Ψ : Δ → Π be universal covering

4.6. Case n = 2 for all domains

67

maps such that Φ(0) = z and Ψ(0) = w. We consider the holomorphic function g : Δ → Δ deﬁned by g(ζ) = Ψ−1 (f (Φ(ζ))) =

∞

cn ζ n ,

ζ ∈ Δ.

n=1

By direct computations, one gets the identity 2

f (z) (Φ (0)) 2 Ψ (0) and therefore

= c2 +

Ψ (0) 2 Φ (0) − c c1 , 2Ψ (0) 1 2Φ (0)

y 2 x |f (z)| λΠ (w) = + c + c1 , c 2 2 2 2 1 2 (λΩ (z))

where x =−

Φ (0) Φ (0)

and y =

(4.15)

Ψ (0) . Ψ (0)

If g ∈ A(Δ, Δ) and g(0) = 0, then the function f = Ψ ◦ g ◦ Φ−1 belongs to the set A(Ω, Π) and f (z) = w. Therefore, we have to ﬁnd y x (4.16) S(x, y) = max c2 + c21 + c1 | g ∈ A(Δ, Δ) with g(0) = 0 . 2 2 Using classical results on the Taylor coeﬃcients of unimodular bounded functions (see [148]) we get y x S(x, y) = max c2 + c21 + c1 | |c1 | ≤ 1, |c2 | ≤ 1 − |c1 |2 2 2 and, by a little analysis, |y| 2 |x| 1 t + t | t ∈ [0, 1] = F (|x|, |y|), S(x, y)) = max 1 − t2 + 2 2 2 where the function F is deﬁned as p + q, F (p, q) := p2 , 2 + 8−4q

if p + 2q ≥ 4, if p + 2q < 4,

for p ≥ 0 and q ≥ 0. Combining this with (4.15) and (4.16) one gets max{|f (z)| | f ∈ A(Ω, Π), ﬁxed w = f (z)} = F (|x|, |y|)

2

(λΩ (z)) , λΠ (w)

where |x| = p = |∇ (1/λΩ (z))| and |y| = q = |∇ (1/λΠ (w))| .

(4.17)

68

Chapter 4. Basic Schwarz-Pick type inequalities For a domain Ω, by formula (3.6) in the form 2 Φ (ζ) |∇ (1/λΩ (z))| = (1 − |ζ| ) − 2ζ Φ (ζ)

and by Theorem 3.23 it is known that sup |∇ (1/λΩ (z))| = 2

z∈Ω

if and only if Ω is convex and that sup |∇ (1/λΩ (z))| > 2

z∈Ω

in all other cases. Accordingly, for all hyperbolic domains Ω and Π, sup |∇ (1/λΩ (z))| ≥ 2,

z∈Ω

sup |∇ (1/λΠ (w))| ≥ 2.

w∈Π

Moreover, we remark that, for a ≥ 2 and b ≥ 2, one easily gets the identity max{F (p, q) | p ∈ [0, a], and q ∈ [0, b]} = a + b, since the condition p + 2q < 4 implies the chain of inequalities 2+

p p2 < 2+ < a + b. 8 − 4q 2

These together with formula (4.17) imply the assertion of Theorem 4.18.

Using Theorem 3.23 and Theorem 4.18 we obtain Corollary 4.19. Suppose that Ω and Π are hyperbolic domains in C. Then the equation C2 (Ω, Π) = 2 is valid if and only if Ω and Π are convex domains in C. Using formulas (3.6) and (3.42) to compute the above gradients, one easily gets the following corollaries of Theorem 4.18. Corollary 4.20. Let α ∈ [1, 2], and let g(ζ) = (ζ + 1)α , g(0) = 1. If Ωα := g(Δ), then C2 (Ωα , Ωα ) = 2 α. Observe that (Ω2 , Ω2 ) is an extremal pair in Theorem 4.10 and compare Section 5.5. Corollary 4.21. Let A1 and A2 be annuli with moduli M1 and M2 , respectively. Then C2 (A1 , A2 ) = 1 + 4M12 + 1 + 4M22 . Using Theorem 3.12, equation (3.37) and Theorem 4.18 we immediately obtain the following assertion. Corollary 4.22. Suppose that Ω and Π are hyperbolic domains in C. The constant C2 (Ω, Π) is ﬁnite if and only if Ω and Π both are domains with uniformly perfect boundary. If ∞ ∈ Ω or ∞ ∈ Π and n ≥ 2, then it is easy to verify that Cn (Ω, Π) = ∞ .

Chapter 5

Punishing factors for special cases After a colloquium talk of the second author on estimates of the form (λΩ (z))n |f (n) (z)| ≤ Cn (Ω, Π) , n! λΠ (f (z))

f ∈ A(Ω, Π), z ∈ Ω,

(5.1)

for simply connected domains Ω and Π in C, Ch. Pommerenke ([132]) proposed to look at (5.1) in the following way. The quotient (λΩ (z))n /λΠ (f (z)) reﬂects the inﬂuence of the positions of the points z and f (z) in Ω and Π on the nth derivative f (n) (z), whereas the quantities Cn (Ω, Π) are factors punishing bad behaviour of Ω or Π at the boundary. This motivates the title of the present chapter as well as the titles of some our papers.

5.1

Solution of the Chua conjecture

In [56] Chua published the following conjecture among others. Chua’s conjecture. Let Ω be a convex proper subdomain of C and let f be holomorphic and injective on Ω. Then for any z ∈ Ω and any n ≥ 2 the inequality (n) f (z) n−1 n−2 (λΩ (z)) (5.2) f (z) n! ≤ (n + 1) 2 holds true. In [56], Chua settled this conjecture for n = 2, 3, 4 (see [102] and [103] for the cases n = 5, 6, 7, 8). Also, taking the limit z → 1, z ∈ (0, 1), in (5.2) for the Koebe function shows that the constant (n + 1)2n−2 on the right side of (5.2) can not be replaced by a smaller one.

70

Chapter 5. Punishing factors for special cases

In the paper [24] we proved that for Ω convex, Π linearly accessible, and n ≥ 2, the inequality Cn (Ω, Π) ≤ (n + 1)2n−2 (5.3) is valid. The equation λΩ (z) , z ∈ Ω, λΠ (f (z)) holds for functions f injective on Ω. Hence, the inequality (5.3) implies the validity of Chua’s conjecture for f that map Ω conformally onto a linearly accessible domain Π. But, in fact more is true. |f (z)| =

Theorem 5.1 ([25]). Let Ω be a convex proper subdomain of C, and Π be a simply connected proper subdomain of C. Let further n ≥ 2. Then the inequality (5.3) is valid. Proof. Because of the central formula (4.5), for n ≥ 2 it is evident that (5.3) will follow from the inequality n ck An,k ≤ (n + 1)2n−2 , (5.4) k=1

where An,k are the coeﬃcients connected with the inverse of a function s ∈ K, and ck are described by the condition that the sum of the series ∞

ck z k =: g1 (z),

z ∈ Δ,

(5.5)

k=1

is subordinate to a member of the family S. Because of the Schur-Jabotinsky theorem and the Marx-Strohh¨ acker inequality we have to prove that n−1 n − l cn−l al,n n l=0 n−1 1 n n−1 = cn + (n − j)cn−j dj−σ,σ ≤ (n + 1)2n−2 . (5.6) σ n σ=1

j=σ

We may replace the coeﬃcients dj−σ,σ by the coeﬃcients dj−σ of a unimodular bounded function when we estimate the modulus of the inner sum in (5.6). Using 2 n σ n = (n + 1)2n−2 , σ n σ=1 it is easily seen that the desired inequality follows from the inequality n−1 (n − j)cn−j dj−σ,σ ≤ (n − σ)2 . j=σ

5.1. Solution of the Chua conjecture

71

This is a consequence of the following lemma with p = n − σ.

Lemma 5.2. Let ω ˜ (z) =

∞

dτ z τ

τ =0

be holomorphic in the unit disc and such that ω ˜ (Δ) ⊂ Δ and let g1 be the function deﬁned by the equation (5.5) and subordinate to a function from S. Then for p ∈ N the inequality p−1 (5.7) (p − τ ) cp−τ dτ ≤ p2 τ =0

is valid. Proof. We shall use Sheil-Small’s theorem 2.9 which says that If g1 is subordinated to a function g ∈ S and P is a polynomial of degree ≤ p, then for z ∈ Δ the inequality |(P ∗ g1 )(z)| ≤ p max{|P (z)| , |z| = 1}

(5.8)

is valid. To prove (5.7) with the help of (5.8) we consider the polynomial P (z) =

p−1

dτ (p − τ )z p−τ .

τ =0

Because of the identity (P ∗ g1 )(1) =

p−1

(p − τ ) cp−τ dτ ,

τ =0

it is suﬃcient for the proof of (5.7) to show that −iθ P e ≤ p, θ ∈ [0, 2π]. Since the family of functions ω ˜ is invariant against rotations of the unit disc, it remains to prove the inequality p−1 (p − τ )dτ ≤ p. τ =0

This exactly is Fej´er’s inequality, used above. This concludes the proof of Lemma 5.2. Theorem 5.1 follows immediately.

72

Chapter 5. Punishing factors for special cases

5.2

Punishing factors for angles

We again consider the quantities Ln (f, z, Ω, Π), Mn (z, Ω, Π) and Cn (Ω, Π) deﬁned by n 1 (n) (λΩ (z)) , n ∈ N, f ∈ A(Ω, Π), z ∈ Ω, f (z) = Ln (f, z, Ω, Π) n! λΠ (f (z)) Mn (z, Ω, Π) := sup{Ln (f, z, Ω, Π) | f ∈ A(Ω, Π)}, and Cn (Ω, Π) := sup{Mn (z, Ω, Π) | z ∈ Ω}. In this section we are concerned with the above quantities where one or both of the domains in question belong to the class of angular domains aHα + b with opening angle απ, 1 ≤ α ≤ 2, which means that there exists a linear transformation T (z) = az + b such that aHα + b := T (Hα ), where απ Hα = z | | arg z| < . 2 Theorem 5.3 ([17]). Let 1 ≤ α ≤ 2, n ∈ N, and z0 ∈ Δ. Then Mn (z0 , Δ, Hα ) =

n−1

(1 + |z0 |)k

k=0

where

α ν

=

n−1 k

α n−k

2n−k−1 , α

ν−1 1 (α − μ). ν! μ=0

If we let |z0 | → 1, then we immediately obtain the following result. Corollary 5.4. Let 1 ≤ α ≤ 2 and n ∈ N. Then 2n−1 n+α−1 . Cn (Δ, Hα ) = n α Proof of Theorem 5.3. The formula (4.3) indicates that we may insert into (4.5) for Ω = Δ the identities n−1 An,k (z0 ) = z0 n−k . n−k Next, we set in Theorem 2.19, g(ζ) = f

ζ + z0 1 + z0 ζ

,

5.2. Punishing factors for angles

73

and deﬁne the Taylor coeﬃcients dk (α) by α ∞ 1+ζ 1 −1 = dk (α)ζ k . 2α 1−ζ k=1

Since the coeﬃcients dk (α) are nonnegative in our cases, we see that in (4.5) we may use the inequalities |ak |λHα (f (z0 )) ≤ dk (α) to get

n n−1 |z0 |n−k dk (α). Mn (z0 , Δ, Hα ) ≤ n−k

(5.9)

k=1

It is easily seen that in this inequality the upper bound is attained for z0 = r0 eiθ , r0 ≥ 0, if we choose f to be given by the identity α ζ + z0 eiθ 1 + e−iθ ζ − 1 . f − f (z0 ) λHα (f (z0 )) = 1 + z0 ζ 2α 1 − e−iθ ζ It remains to show that the upper bound for Mn (z0 , Δ, Hα ) in (5.9) equals the upper bound given in Theorem 5.3. To this end we remark that the left side of the inequality (5.9) is the nth Taylor coeﬃcient of the function α (1 + |z0 |z)n−1 1 + z . 2α 1−z Use of the binomial theorem for n−1

(1 + |z0 |z)n−1 = ((1 + |z0 |)z + 1 − z)

reveals that the said nth Taylor coeﬃcient equals the nth Taylor coeﬃcient of the product n 1 n−1 α z k−1 (1 + |z0 |)k−1 (1 − z)n−k−α . (1 + z) n−k 2α k=1

Hence, we have to sum up the (n − k + 1)th Taylor coeﬃcients of the functions 1 n−1 α (1 + |z0 |)k−1 (1 − z)n−k−α , k = 1, . . . , n. (1 + z) n−k 2α These coeﬃcients may be written in the form n−k+1 α (−1)n−k+1−q n − k − α n−1 (1 + |z0 |)k−1 n−k+1−q q n−k 2α q=0 n−k 2 n−1 α = (1 + |z0 |)k−1 , k = 1, . . . , n. n−k n−k+1 α This proves the result of Theorem 5.3 if we replace k by k − 1 and sum up from k = 0 to k = n − 1.

74

Chapter 5. Punishing factors for special cases

Using theorems on inverse coeﬃcients of conformal mappings we generalize Corollary 5.4 and Yamashita’s results in [170] (see Theorem 4.9) as follows. Theorem 5.5 ([17]).

a) Let 1 ≤ α, β ≤ 2 and n ∈ N. Then α (2β)n +n−1 β Cn (Hβ , Hα ) = . n 2α

b) Let 1 ≤ β ≤ 2, n ∈ N, and Π be a convex proper subdomain of C . Then Cn (Hβ , Π) = Cn (Hβ , H1 ). c) Let 1 ≤ β ≤ 2, n ∈ N, and Π be a simply connected proper subdomain of C. Then Cn (Hβ , Π) ≤ Cn (Hβ , H2 ). d) Let 1 ≤ α ≤ 2, n ∈ N, and Ω be a simply connected proper subdomain of C. Then Cn (Ω, Hα ) ≤ Cn (H2 , Hα ). Remark 5.6. For all α, β ∈ [1, 2], one has that C1 (Hβ , Hα ) = 1 in accordance with the Schwarz-Pick inequality. In the case n ≥ 2, n−1 2n Cn (Hβ , Hα ) = (α + kβ). n! k=1

We see that C2 (Hβ , Hα ) = C2 (Hα , Hβ ) = α + β in accordance with Theorem 4.18, and Cn (Hα , Hβ ) > Cn (Hβ , Hα ) > 2n−1 , whenever n ≥ 3 and α > β ≥ 1. Proof of Theorem 5.5. a) Firstly, we notice that the nth Taylor coeﬃcient of −1 k (h−1 β (w)) is nonnegative, since the Taylor coeﬃcients of hβ (w) are all nonnegative. This fact and Theorem 2.20 imply that we now may use as an upper bound for the quantity |An,k (z0 )|, 1 ≤ k ≤ n, in formula (4.5) this nth Taylor coeﬃcient k of (h−1 β (w)) . For |ak | we use the same inequality as in the proof of Theorem 5.3. To prove that the resulting inequality is sharp and to compute the explicit formula given in the assertion, we use the conformal map fα,β deﬁned by

(z) , (5.10) fα,β (z) := gα h−1 β where gα (ζ) :=

1 2α

1+ζ 1−ζ

α

−1 ,

ζ ∈ Δ.

5.2. Punishing factors for angles

75

One one hand, fα,β maps the special angular domain 1 β π < Hβ = z arg z − 2β 2 conformally onto the special angular domain 1 α π < Hα = z arg z + 2α 2 (n)

such that fα,β (0) = 0. Now, we compute fα,β (0) starting with formula (5.10) in the same way as we obtained (4.5). If we use that in our circumstances λHβ (0) = λHα (0) = 1, we observe that the resulting sum is just the sum that we derived as an upper bound in the beginning of this proof. On the other hand, a direct computation of fα,β shows that fα,β (z) = Hence,

α 1 (1 − 2β z)− β − 1 . 2α

(n)

fα,β (0) n!

=

(2β)n 2α

α β

+n−1 n

.

This proves the assertion in question. b) The inequality Cn (Hβ , Π) ≤ Cn (Hβ , H1 ) for a convex domain Π follows in our case from Rogosinski’s theorem 4.12 on convex subordination according to which we may apply |ak |λΠ (z0 ) ≤ 1 in (4.5) in a way analogous to that of the proof of the case a). The proof of Cn (Hβ , Π) ≥ Cn (Hβ , H1 ) for a convex domain follows the proof of Mn (z, Δ, Π) = (1 + |z|)n−1 for convex Π in Chapter 4. Therefore, we only describe the crucial steps here. Without loss of generality, we may assume that Δ1 = {w | |w − 1| < 1} ⊂ Π ⊂ H1 = {w | Re w > 0}.

76

Chapter 5. Punishing factors for special cases

For the special angular domain Hβ chosen as in the proof of the case a) we consider the holomorphic functions fk (z) =

2 1

1 + k(1 − 2β z) β

,

k ∈ N,

which map Hβ conformally onto Δ1 such that fk (0) = 2/(k + 1) =: wk . We compute (k + 1)2 λHβ (0) = 1, λΔ1 (wk ) = . 4k Since the sequences (kfk )k∈N are uniformly convergent on a neighbourhood of the origin we see that 1 dn (n) −β lim kfk (0) = 2 (1 − 2β z) = 4 (n!)Cn (Hβ , H1 ). k→∞ dz n z=0 The last equation follows from the proof of the case a). Using the comparison principle for densities of the Poincar`e metric (see Theoren 3.5) as in the proof of the corresponding relation in Chapter 4, it is easily seen that lim

k→∞

λΠ (wk ) = 1. λΔ1 (wk )

Gathering together the above relations we get the equation f (n) (0)λΔ1 (wk ) λΠ (wk ) n = Cn (Hβ , H1 ), lim Ln (fk , 0, Hβ , Π) = lim k→∞ k→∞ λΔ1 (wk ) n! λHβ (0) which immediately implies the assertion. c) In this proof we only have to use again the validity of the Rogosinski or generalized Bieberbach conjecture together with the case a) of Theorem 5.5. d) Here we use in the application of (4.5) that, according to L¨ owner’s Theorem 2.13, the Taylor coeﬃcients of the inverses of schlicht functions and their kth powers, k ∈ N, are dominated by the related (positive) coeﬃcients of the inverse of the Koebe function z/(1+z)2 and its kth powers. This results in the assertion. In our next result we are concerned with proper subdomains Ω of C simply connected with respect to C and holomorphic functions f : Ω → Hα . In the sequel we use the quantity p ∈ (0, 1] such that p := p(z0 ) = tanh DΩ (z0 , ∞). By deﬁnition, p = 1, if ∞ ∈ Ω. The following theorem forms a bridge between the results for Π convex and Π simply connected with respect to C.

5.2. Punishing factors for angles

77

Theorem 5.7 ([17]). Let Ω be as above, z0 ∈ Ω \ {∞}, 1 ≤ α ≤ 2, and n ≥ 2. Then for any holomorphic function f : Ω → Hα , the inequality α n−k n 1 (n) (λΩ (z0 ))n 4k 1 n−1 2 p(z0 ) + |f (z0 )| ≤ +2 n−k k n! λHα (f (z0 )) 2α p(z0 ) k=1

is valid. For each n ≥ 2, each α ∈ [1, 2], and each p ∈ (0, 1] there exist Ω, Hα , z0 ∈ Ω \ {∞}, and f as above such that equality is attained in the above inequality. Proof. We again start with the central equality (4.5). Let p = p(z0 ),

1 a = p+ , p

ζ (1 + p ζ)(1 +

κp (ζ) =

ζ p)

, ζ ∈ Δ,

and Kp the function inverse to κp having an expansion Kp (z) = z +

∞

Bν z ν

ν=2

valid in some neighbourhood of the origin. Considering the class Sp of functions meromorphic and univalent in Δ which have a pole in a point b, |b| = p, and an expansion ∞ m(ζ) = ζ + Aν ζ ν , |ζ| < p, ν=2

Baernstein and Schober showed in [33] (see Theorem 2.15) that |Aν | ≤ Bν , ν ≥ 2. This implies that the nth Taylor coeﬃcients of the kth powers of m ∈ Sp are dominated by the nth Taylor coeﬃcients Bn,k of the kth powers of the function Kp . By a procedure analogous to that, which led us to formula (4.5), we get in our case for the meromorphic functions f in question the inequality |f (n) (z0 )| (λΩ (z0 )) Bn,k dk (α). ≤ n! λHα (f (z0 )) n

n

(5.11)

k=1

To show that this inequality is sharp and that the right-hand side has the form given in the assertion, we proceed in principle as in the proof of the case a) of Theorem 5.5. To this end we consider in the present case the function fα,p (z) := gα (Kp (z)) which maps the domain Ω0 = C \ [(a + 2)−1 , (a − 2)−1 ]

78

Chapter 5. Punishing factors for special cases

conformally onto the special angular domain Hα chosen as in the proof . We see that in this case fα,p (0) = 0 and λHα (0) = λΩ0 (0) = 1 and we deduce the sharpness (n) of (5.11) considering fα,p (0) as above. To get the explicit form of the upper bound we consider α ∞ 1 − (a − 2)z 2 1 fα,p (z) = − 1 =: Cn (α, p)z n . 2α 1 − (a + 2)z n=1 This implies 1+

∞

α2 k−1 k

4(a + 2)

z

= 1 + 2α

∞

Cn (α, p)z n .

n=1

k=1

With the abbreviation Z := z(a + 2) this is equivalent to the equation k k ∞ α ∞ 4 Z Cn (α, p) n 2 = 1 + 2α Z . 1+ k a+2 1−Z (a + 2)n n=1 k=1

If we then expand Z k /(1 − Z)k in powers of Z and compare the coeﬃcients on both sides of this equation, we then establish the assertion of Theorem 5.7 on the sharpness of the punishing factor.

5.3

Sharp lower bounds for punishing factors

We shall consider domains with a special local property at a boundary point, which will be a metrical characterization of the “bad” behavior of the boundary. Let Ω be a domain in C. We denote aΩ + b := {z = aζ + b | ζ ∈ Ω}, and

1 := Ω

z=

a, b ∈ C, a = 0,

1 |ζ∈Ω . ζ

For α ∈ [1, 2] we deﬁne α Δ+ α := {z = (ζ + 1/2) | |ζ| < 1/2},

Δ− 0 := [−1, 0] and Δ− 2−α := {z = −

w |w= w+1

1−ζ 1+ζ

2−α , |ζ| ≤ 1}

for α ∈ [1, 2),

where the branch of a power is ﬁxed by 1α = 1. Observe that z = 0 is a boundary − point for all sets Δ+ α and Δ2−α .

5.3. Sharp lower bounds for punishing factors

79

Deﬁnition 5.8. Let Ω be a domain in C, let α ∈ [1, 2], and let z0 ∈ (∂Ω) \ {∞}. If there exist ε > 0 and t ∈ R such that it − εeit Δ+ α + z0 ⊂ Ω and εe Δ2−α + z0 ⊂ C \ Ω,

(5.12)

then we say that z0 is an angular point on ∂Ω of order α. Clearly, the condition (5.12) is equivalent to the inclusions − Δ+ α ⊂ aΩ + b ⊂ C \ Δ2−α ,

a=

e−it , ε

b = −z0

e−it . ε

We extend this deﬁnition to the boundary point at inﬁnity. Deﬁnition 5.9. Let α ∈ [1, 2] and let Ω be a domain in C such that ∞ ∈ ∂Ω. If there exist complex numbers a = 0 and b such that Δ+ α ⊂

1 ⊂ C \ Δ− 2−α , aΩ + b

then we say that the point at inﬁnity is an angular point on ∂Ω of order α. According to Deﬁnitions 5.8 and 5.9 the domain απ Hα := z ∈ C | |argz| < 2 has two boundary angular points of order α that are z0 = 0 and the point at inﬁnity. A non-trivial example is ∞

Ω1 = H1 \

Dk , k=1

where {z ∈ C |

Dk =

|z − 1/k − it| ≤ (t + 2k)−2 }.

0≤t 0. Obviously, this function has an expansion ∞ k (5.18) ΨΩ (z) = λΩ (z0 ) z − z0 + Ak (z − z0 ) , k=2

valid in some neighbourhood of the point z0 . Let b = ΨΩ (∞) ∈ Δ. Because of the conformal invariance of the hyperbolic distance, DΔ (0, b) equals DΩ (z0 , ∞). According to (5.16) we obtain |b| = p = p(z0 ). Let the function κ(ζ) =

ζ

, (1 + p ζ) 1 + pζ

ζ ∈ Δ,

(5.19)

which is univalent in Δ, have an inverse function K = κ−1 deﬁned on κ(Δ). Obviously ∞ ∈ κ(Δ) and 0 ∈ κ(Δ). Using the expansion K(z) = z +

∞

Bk z k ,

k=2

valid in some neighbourhood of the origin, Baernstein and Schober (see Chapter 2, Theorem 2.15) showed that |Ak | ≤ Bk ,

k ∈ N \ {1}.

(5.20)

Consider now for each f ∈ A(Ω, Π) the function ∞

g(ζ) := f (ΦΩ (ζ)) =

ak ζ k ,

ζ ∈ Δ,

(5.21)

k=0

where ΦΩ is inverse to ΨΩ . It is known by formula (4.5) that 1 f (n) (z0 ) An,k (z0 , Ω) ak , n = n! (λΩ (z0 )) n

(5.22)

k=1

where (λΩ (z0 , Ω)) An,k (z0 , Ω) = n!

−n

(n) (ΨΩ (z)) k

.

(5.23)

z=z0

The formulas (5.18) and (5.23) imply that An,n (z0 , Ω) ≡ 1 and that the An,k (z0 , Ω), k = 1, . . . , n, are polynomials with positive coeﬃcients in A1 , . . . , An−1 . Hence, we get as a consequence of this fact and of the formulas (5.18)–(5.23) the inequalities (n) 1 k |An,k (z0 , Ω)| ≤ tn,k := . (5.24) (K(z)) n! z=0

86

Chapter 5. Punishing factors for special cases

Thus,

1 |f (n) (z0 )| tn,k |ak |, n ≤ n! (λΩ (z0 )) n

(5.25)

k=1

where the positive quantities tn,k may be computed, according to the formulas (5.19), (5.24), and the Cauchy formula, as 1 1 (1 + aζ + ζ 2 )n−1 (1 − ζ 2 ) tn,k = d ζ, a = + p, r ∈ (0, 1). (5.26) 2π i |ζ|=r ζ n−k+1 p Let q = tanh DΠ (f (z0 ), ∞) ∈ (0, 1]. We need the expansion ∞ ζ

=ζ+ ck (q)ζ k , ζ (1 − qζ) 1 − q k=2

|ζ| < q.

It is known that (see Chapter 2) k−1

j=0 q q k−1

ck (q) =

2j

=

qk − q−

1 qk 1 q

.

The properties tn,k tn,k+1 ≥ > 0, ck (q) ck+1 (q) and

n

ck (q) tn,k =

k=1

1 a+q+ q

k = 1, . . . , n − 1, n−1 ,

a=

(5.27)

1 + p, p

(5.28)

of the quantities tn,k will be proved below. Let us suppose for the moment that they are true. Applying the Cauchy inequality to (5.25) we get 1 |f (n) (z0 )| ≤ n! (λΩ (z0 ))n

n

12 ck (q) tn,k

k=1

n tn,k |ak |2 ck (q)

12 .

(5.29)

k=1

Let ΦΠ be the conformal map of Δ onto Π with ΦΠ (0) = a0 = f (z0 ) and ΦΠ (0) = (λΠ (f (z0 )))−1 > 0. The function g deﬁned in (5.21) is subordinate to the univalent function ∞ −1 k ΦΠ (ζ) = a0 + (λΠ (f (z0 ))) ζ+ . ck ζ k=2

The starting point for the rest of the proof is Theorem 2.5 which is a generalized version of the Rogosinski-Goluzin theorem. The second ingredient at this point is Jenkins’ theorem 2.7. We easily get 2

(λΠ (f (z0 )))

n n n tn,k tn,k ck (q) tn,k . |ak |2 ≤ |ck |2 ≤ ck (q) ck (q)

k=1

k=1

k=1

(5.30)

5.4. Domains in the extended complex plane

87

From (5.28), (5.29) and (5.30) we obtain the desired inequality (5.17). To complete the proof of (5.17) we have to prove (5.27) and (5.28). An immediate consequence of (5.26) for r ∈ (0, q) is n

ck (q) tn,k =

k=1

1 2π i

1 = 2π i

|ζ|=r

n (1 + aζ + ζ 2 )n−1 (1 − ζ 2 ) ck (q) ζ k−1 d ζ ζn k=1

I(ζ) d ζ = Res(I(ζ), ζ = 0) |ζ|=r

= −Res(I(ζ), ζ = q) − Res(I(ζ, ζ = where I(ζ) =

1 ) − Res(I(ζ), ζ = ∞), q

(1 − ζ 2 ) (1 + aζ + ζ 2 )n−1

. ζn (q − ζ) 1q − ζ

We observe that Res(I(ζ), ζ = 0) = Res(I(ζ), ζ = ∞) and Res(I(ζ), ζ = q) = Res(I(ζ), ζ = 1q ). Therefore n

ck (q) tn,k = −Res(I(ζ), ζ = q) =

k=1

n−1 1 q+ +a . q

To prove (5.27) we consider the polynomial 2 n

Pn (ζ) = (1 + aζ + ζ ) =

2n

bm ζ m ,

(5.31)

m=0

where, as before, a = p + 1/p > 2. The identity Pn (ζ) = (1 − ζ 2 )Pn−1 (ζ) + ζ Pn−1 (ζ) P1 (ζ) implies Pn (ζ) −

ζ P (ζ) = (1 − ζ 2 )Pn−1 (ζ). n n

Using (5.26) we identify tn,k as the coeﬃcient of ζ n−k in the polynomial on the right side of this equation. On the other hand, (5.31) shows that bn−k − n−k n bn−k is the coeﬃcient of ζ n−k in the polynomial on the left side. Therefore, we obtain n

tn,k = bn−k , k

k = 1, . . . , n.

This implies that the inequalities tn,k+1 tn,k ≥ > 0, k k+1

k = 1, . . . , n − 1,

(5.32)

88

Chapter 5. Punishing factors for special cases

are equivalent to 1 ≤ b1 ≤ b2 ≤ · · · ≤ bn−1 , which may be proved by mathematical induction on n considering the coeﬃcients of the polynomials (1 − ζ) Pn (ζ). Moreover, (5.32) implies the desired inequalities (5.27) because of the simple relations ck+1 (q) k+1 ≥ , ck (q) k

k = 1, . . . , n − 1.

Now, it remains to prove that (5.17) is sharp. To this end we consider the following example. For any n ∈ N and any a = p + 1/p > 2, let Ω0 = C \ [(a + 2)−1 , (a − 2)−1 ], f0 (z) =

z

1− a+q+

1 q

z

and Π0 = f0 (Ω0 ). It is an easy task to derive that λΩ0 (0) = λΠ0 (0) = 1, that tanh DΩ0 (0, ∞) equals p, tanh DΠ0 (0, ∞) equals q and that (n)

f0 (0) = n!

p+

1 1 +q+ p q

n−1 .

This completes the proof of Theorem 5.15. Consider now an improvement of (5.17) for convex domains.

If the domain Π is convex, then the coeﬃcients ak , k ∈ N, of the functions g deﬁned in (5.21) satisfy the inequalities of Rogosinski (see Theorem 4.12) λΠ (a0 )|ak | ≤ 1,

k ∈ N.

(5.33)

Therefore (5.25) and (5.33) imply n (λ (z ))n 1 (n) Ω 0 tn,k , f (z0 ) ≤ n! λΠ (f (z0 ))

(5.34)

k=1

where we used the same deﬁnitions as in the proof of Theorem 5.15. A straightforward computation using (5.26) shows that n k=1

tn,k =

1 π

π

p+

0

1 + 2 cos θ p

n−1 (1 + cos θ) d θ =: Cn (p),

(5.35)

5.4. Domains in the extended complex plane

89

if ∞ ∈ Ω and therefore p ∈ (0, 1), respectively n

tn,k =

k=1

2n − 1 n

,

(5.36)

if ∞ ∈ Ω. The formulas (5.34), (5.35) and (5.36) immediately yield Theorem 5.16 ([19]). Suppose that Ω is a hyperbolic simply connected proper subdomain of C and that Π is a convex proper subdomain of C, K = −4. Then the following estimates are valid: If ∞ ∈ Ω, then for any f ∈ A(Ω, Π) and any z ∈ Ω with hyperbolic distance DΩ (z, ∞) = Arctanh p, the inequality n

(λΩ (z)) 1 (n) |f (z)| ≤ Cn (p) n! λΠ (f (z))

(5.37)

holds for any n ∈ N. Remark 5.17. The sharpness of (5.37) can be shown using Ω0 , H1 and the function f1 deﬁned by ! 2f1 (z) :=

1 + K(z) = 1 − K(z)

1 − (a − 2)z . 1 − (a + 2)z

(5.38)

Indeed, the function f1 maps Ω0 conformally onto the half plane H1 , K(0) = 0 and λΩ0 (0) = λΠ1 (1/2) = 1. Now, we compare two expansions ∞

f1 (z) =

∞

1 1 (K(z))k = + dn z n , + 2 2 n=1

(5.39)

k=1

which are valid in some neighbourhood of the origin. The formulas (5.38), (5.39), and (5.24) immediately yield dn =

n n (n) 1 = tn,k = Cn (p). (K(z))k n! z=0 k=1

Thus,

(5.40)

k=1

n (λΩ0 (0)) 1 (n) . f1 (0) = Cn (p) n! λΠ1 (f1 (0))

Moreover, using (5.39), (5.40) and the Taylor series of (f1 )2 in some neighbourhood of the origin, we get

∞

1 Ck (p) z k + 2 k=1

2

∞

=

1 (a + 2)n−1 z n , + 4 n=1

90

Chapter 5. Punishing factors for special cases

which implies C1 (p) = 1, Cn (p) =

C2 (p) = a + 1 = p +

1 + 1, and p

n−1 n−2 1 p+ +2 − Ck (p)Cn−k (p), p

n ≥ 3.

k=1

These formulas show the diﬀerence between the bounds in (5.17) and (5.37). Remark 5.18. It may be worthwhile to notice that the direct determination of the integral (5.35) and the computation of the Taylor coeﬃcients of the function (5.38) with the help of binomial series can be exploited to prove relations between binomial coeﬃcients that seem to be not known. We were not able to ﬁnd another method to prove these identities than the geometrical method above. For the convenience of the interested reader we mention such a formula. For m = 0, . . . , n− 1, n ∈ N, one may get in this way the identities n−1 "m # (−1)n m m 2 ⎛ ⎞ n 1/2 −1/2 ⎝ k n−k m j ⎠, (−1) =2 k n−k j m+1−j j

k=0

where j varies in the second sum in the range determined by the binomial coeﬃcients.

5.5

Maps from convex into concave domains

We shall consider the case when ∞ ∈ Π and the punishing factor is a function of the hyperbolic distance between f (z) and ∞. Moreover, we suppose that Π = C \ Π1 , where Π1 is a compact convex subset of C containing more than one point. Domains Π of this type will be called concave domains. In the sequel we use the quantity p ∈ (0, 1) such that DΔ (0, p) =

1 1+p log 2 1−p

equals DΠ (f (z0 ), ∞). This means that p := p(f (z0 )) = tanh DΠ (f (z0 ), ∞),

(5.41)

where, as usual, tanh x = (ex − e−x )/(ex + e−x ). A central part in the proofs is played by functions h satisfying the following conditions:

5.5. Maps from convex into concave domains

91

(i) The function h is meromorphic and injective on Δ and h has its pole at a point p ∈ (0, 1). (ii) The set C \ h(Δ) is convex. (iii) h(0) = h (0) − 1 = 0. We called functions with the properties (i)–(iii) concave univalent functions with pole p and for the family of those functions we used the abbreviation Co(p). For older results on such functions compare [123], [117] and [108] and for newer ones [30], [31], [15], [20], [23],[166] and [167]. Theorem 5.19 ([28]). Let Π be a concave domain. Then for all n ∈ N, f ∈ A(Δ, Π), z0 ∈ Δ, f (z0 ) ﬁnite, and p as in (5.41) the inequalities n−1 1 |f (n) (z0 )| (1 − |z0 |2 )n 1 2 n−1 |z0 | + (5.42) − p (|z0 | + p) ≤ n! RΠ (f (z0 )) 1 − p2 p are valid. Equality in (5.42) is attained if z0 = r ∈ [0, 1), and f = E ◦ T , E(z) =

z (1 − zp)(1 − z/p)

and

T (z) =

z−r , 1 − zr

z ∈ Δ.

These functions f are conformal maps of Δ onto the domain % $ −p −p . C\ , (1 − p)2 (1 + p)2 Corollary 5.20 (compare [31]). Let Π be a concave domain. Then for all n ≥ 2, f ∈ A(Δ, Π), z0 ∈ Δ, f (z0 ) ﬁnite, and p as in (5.41) the inequalities n |f (n) (z0 )| (1 − |z0 |2 )n (1 + p)n−2 k p ≤ n! RΠ (f (z0 )) pn−1

(5.43)

k=0

are valid. The constant on the right side cannot be replaced by a smaller one independent of Π, f, z0 , and f (z0 ). Proof of Theorem 5.19. We consider the function g ∈ A(Δ, Π) deﬁned by ζ + z0 g(ζ) = f , ζ ∈ Δ. 1 + z0 ζ Since f (z0 ) is ﬁnite, the function g has a Taylor expansion g(ζ) = f (z0 ) +

∞

ak ζ k

k=1

valid on some neighbourhood of the origin. It is clear that, under our circumstances, we may also use the identity (4.3), which has been proved for functions holomorphic in the unit disc.

92

Chapter 5. Punishing factors for special cases

On the other hand, we use that g(0) = f (z0 ) and g(Δ) ⊂ Π. If we let, for this proof, Φ := ΦΠ,f (z0 ) be deﬁned as above, we see that there exists a function ω1 : Δ → Δ holomorphic on Δ such that g(ζ) = Φ(ζω1 (ζ)),

ζ ∈ Δ.

(5.44)

Further, we conclude from the deﬁnition p = tanh DΠ (f (z0 ), ∞) that there exists ϕ ∈ [0, 2π] such that Φ(peiϕ ) = ∞. Now, we consider the function h deﬁned on Δ by e−iϕ h(ζ) = (5.45) (Φ(eiϕ ζ) − f (z0 )), ζ ∈ Δ. RΠ (f (z0 )) This function h belongs to the class Co(p) and from (5.44) and (5.45) we see that there exists a function h ∈ Co(p) and a function ω = e−iϕ ω1 : Δ → Δ such that g(ζ) = eiϕ h(ζω(ζ))RΠ (f (z0 )) + f (z0 ),

ζ ∈ Δ.

(5.46)

In Theorem 8.4 below (compare [26]) we will consider h ∈ Co(p), ω as above such that the Taylor expansion h(ζω(ζ)) =

∞

αk ζ k

(5.47)

k=1

is valid on some neighbourhood of the origin. We will prove there that for k ∈ N the inequalities 1 1 − p2k 1 2 k−1 |αk | ≤ (5.48) = −p p (1 − p2 )pk−1 1 − p2 pk−1 are valid. After these preparations (5.42) is an immediate consequence of (4.3), (5.46), (5.47), and (5.48). The second assertion of Theorem 5.19 becomes clear from the above proof and the expansion E(z) =

∞ k=1

1 − p2k zk , (1 − p2 )pk−1

|z| < p.

This concludes the proof of Theorem 5.19.

Theorem 5.21 ([28]). Let Ω ⊂ C be a convex proper subdomain of C. Let further Π be a concave subdomain of C and f : Ω → Π be meromorphic on Ω. Then for all n ≥ 2, z0 ∈ Ω, and f (z0 ) ﬁnite, and p as in (5.41) the inequalities n n |f (n) (z0 )| (RΩ (z0 )) (1 + p)n−2 k p ≤ n! RΠ (f (z0 )) pn−1 k=0

(5.49)

5.5. Maps from convex into concave domains

93

are valid. Equality in (5.49) is attained for $ Ω = {z | Rez < 1/2}

and

Π=C\

% −p −p , , (1 − p)2 (1 + p)2

z0 = f (z0 ) = 0, where the function z(1 − z) z = f (z) = E 1−z (1 − z(1 + p))(1 − z(1 + p)/p) maps Ω onto Π conformally. Proof. Let n ≥ 2. According to the central formula (4.5) we have to prove that n n (1 + p)n−2 ck An,k ≤ pk , pn−1 k=1

k=0

where ck are the Taylor coeﬃcients of a function h ∈ Co(p), and, as usual, An,k is the n-th coeﬃcient in the Taylor expansion of F k for the function F inverse to an arbitrary function s ∈ K. Because of the Schur-Jabotinsky theorem (compare the proof of Theorem 4.15), we have to prove that n−1 n n − l (1 + p)n−2 pk . (5.50) cn−l al,n ≤ n pn−1 l=0

k=0

Using the Marx-Strohh¨ acker inequality as above, we get the following formula for the sum appearing in (5.50): n−1 l=0

n−1 1 n n−1 n−l (n − j)cn−j dj−σ,σ . cn−l al,n = cn + σ n n σ=1 j=σ

(5.51)

Now it is easily veriﬁed that it is suﬃcient to obtain the inequality n−1 (n − σ)(1 − p2(n−σ) ) ≤ (n − j)c d , n−j j−σ pn−σ−1 (1 − p2 ) j=σ where dj−σ are the coeﬃcients of a unimodular bounded holomorphic function in the unit disc as in the proof of Theorem 4.15. The latter inequality is equivalent to q−1 q(1 − p2q ) . (5.52) (q − τ ) cq−τ dτ ≤ q−1 p (1 − p2 ) τ =0

As is proved in [20] and [167], for any h ∈ Co(p) there exists a function v1 : Δ → Δ such that v1 (z)pz z 1+ , z ∈ Δ. h(z) = 1 − z/p 1 − zp

94

Chapter 5. Punishing factors for special cases

Using this and the generalized version of Sheill-Small’s Lemma 2.10 (see Remark 2.12) in the limiting case r → 1 , i.e., the inequality ˜ iθ )| ≤ max |P (z)| |P ∗ h(e

q−1

|z|=1

2

|uk |

1/2 q−1

k=0

1/2 2

|vk |

k=0

for the polynomial P (z) =

q−1

dτ (q − τ )z q−τ

τ =0

and the functions deﬁned by wi (z) = ω(z), i = 1, 2, 3,

z ∈ Δ,

and U (z) =

q−1 k z k=0

p

∞

and V (z) = 1 +

v1 (z)pz vk z k , = 1 − zp

z ∈ Δ,

k=0

we obtain the inequality q−1 1/2 q−1 1/2 q−1 p−2k |vk |2 . (q − τ ) cq−τ dτ ei(q−τ )θ ≤ max |P (z)| |z|=1 τ =0

k=0

k=0

To estimate the last factor of the right side in this inequality we use that the function V (z) − 1 is quasi-subordinated to the function pz/(1 − zp). According to Theorem 2.4 we get q−1 q−1 |vk |2 ≤ p2k . k=0

k=0

Hence, q−1 1/2 q−1 1/2 q−1 −2k 2k p p (q − τ ) cq−τ dτ ≤ max |P (z)| |z|=1 τ =0

k=0

k=0

1 − p2q = max |P (z)| q−1 . p (1 − p2 ) |z|=1 Fej´er’s inequality (see [61] and [159] or apply the theory of linear functionals on H ∞ in chapter 8 of [59]) q−1 (q − τ )dτ ≤ q τ =0

shows that

−iθ P e ≤ q,

θ ∈ [0, 2π].

5.5. Maps from convex into concave domains

95

This concludes the proof of the desired inequality. The extremal property of the function f mentioned in the second assertion of Theorem 5.21 follows directly from the computation of the Taylor coeﬃcients of its expansion with expansion point at the origin. The proof of Theorem 5.21 is complete.

Chapter 6

Multiply connected domains In the preceding chapters we considered punishing factors for simply connected domains, except the case C2 (Ω, Π). Namely, in Section 4.6 it was proved that for all hyperbolic domains Ω ⊂ C and Π ⊂ C, 1 sup |∇ (1/λΩ (z))| + sup |∇ (1/λΠ (w))| . C2 (Ω, Π) = C2 (Π, Ω) = 2 z∈Ω w∈Π In particular, it was proved that C2 (A1 , A2 ) =

1 + 4M12 + 1 + 4M22 ,

where A1 and A2 are annuli with moduli M1 and M2 , respectively. The main aim of this chapter is to describe multiply connected domains by properties that guarantee existence of ﬁnite punishing factors for all n. Also we will deﬁne ﬁnite modiﬁed punishing factors and consider some examples.

6.1

Finitely connected domains

The theorems of Chapters 4 and 5 show that one cannot expect that Cn (Ω, Π) := sup{Mn (z, Ω, Π) | z ∈ Ω}

(6.1)

is always ﬁnite. The central existence theorem for ﬁnitely connected domains is the following assertion. Theorem 6.1 ([21]). Let Ω and Π be ﬁnitely connected hyperbolic domains in C. Then the punishing factors Cn (Ω, Π) are ﬁnite for all n ∈ N if and only if both ∂Ω and ∂Π do not contain isolated points.

98

Chapter 6. Multiply connected domains

Firstly, we prove some necessary lemmas and propositions. We shall use the hyperbolic radius R(z, Π) which is reciprocal to the density of hyperbolic metric with Gaussian curvature K = −4. Let Π be an m-connected hyperbolic domain in C where m ≥ 2, i.e., ∂Π consists of m disjoint connected sets Γk , k = 1, . . . , m. We shall consider m associated simply connected domains Πk ⊂ C, k = 1, . . . , m, deﬁned by the relations ∂Πk = Γk

and Π ⊂ Πk ,

k = 1, . . . , m.

It is clear that Γj ⊂ Πk for j = k and that Π =

m &

Πk .

k=1

Naturally, the connected sets Γk are either points or continua. It is decisive for our proofs that the behaviour of the hyperbolic radius of Π and Πk is the same at the common boundary. If Γk is a point, this is known (see [3] and [55]). We will use the formula for this case later on (compare (6.7) and 6.8)). In the case that Γk is a continuum, this is the content of the following lemma. Lemma 6.2. Let Π and Πk be as above. If Γk = ∂Πk is not a point, then for z0 ∈ Γk , R(z, Π) lim = 1. (6.2) z→z0 , z∈Π R(z, Πk ) Proof. Since Γk is a continuum, Πk is a simply connected domain in C. Riemann’s mapping theorem implies that there exists a conformal map Φ of Δ onto Πk . The inverse function Φ−1 maps Π onto a set Φ−1 (Π) ⊂ Δ. For all suﬃciently small positive numbers the relations D() := {w | e−π < |w| < 1} ⊂ Φ−1 (Π) ⊂ Δ are valid. According to Theorem 3.5 they imply R(w, D()) ≤ R(w, Φ−1 (Π)) ≤ R(w, Δ),

w ∈ D().

(6.3)

Combining (6.3) and the conformal invariance of the hyperbolic metric, |d z| |d w| = , R(z, Π) R(w, Φ−1 (Π))

and

|d z| |d w| = , R(z, Πk ) R(w, Δ)

we get, for any w = Φ−1 (z) ∈ D(), 1 ≥

R(z, Π) R(w, Φ−1 (Π)) R(w, D()) = ≥ . R(z, Πk ) R(w, Δ) R(w, Δ)

The use of R(w, Δ) = 1 − |w|2

and R(w, D()) = 2 |w| sin

1 1 log |w|

(6.4)

6.1. Finitely connected domains

99

and (6.4) yields R(w, Φ−1 (Π)) = R(w, Δ) |w|→1− lim

R(w, D()) = 1. |w|→1− 1 − |w|2 lim

(6.5)

Since |Φ−1 (z)| → 1− as z → z0 , the assertion (6.2) of Lemma 6.2 follows from (6.4) and (6.5). The second preparation for the proof of Theorem 6.1 is concerned with the Taylor coeﬃcients an (f ) of the local expansion f (z) =

∞

an (f )z n

n=0

for a function f ∈ A(Δ, Π) that is holomorphic at the origin. We wish to describe now all ﬁnitely connected hyperbolic domains for which the quantities An (Π) :=

|an (f )| R(a 0 (f ), Π) f ∈A(Δ,Π) sup

are ﬁnite for all n ∈ N. One might expect that boundedness of Π is the decisive condition, but the proof of the following proposition reveals that this is not the case. Lemma 6.3. Let Π be a ﬁnitely connected hyperbolic domain in C. Then the following statements are equivalent. (a) ∂Π does not contain isolated points. (b) An (Π) is ﬁnite for any n ∈ N. (c) Cn (Δ, Π) is ﬁnite for any n ∈ N. (d) There exists a constant KΠ such that, for all w ∈ Π, dist(w, ∂Π) ≤ R(w, Π) ≤ KΠ dist(w, ∂Π). Proof. If Π is a simply connected domain, then Lemma 6.3 is a consequence of known facts (see [55]). It is also known that (a)⇔(d) is valid (see [130] and [77]). Suppose now that Π is m-connected with m ≥ 2. We will prove that (a) ⇒ (b) ⇒ (c) ⇒ (d) ⇒ (a). (a)⇒(b). The condition (a) means that the associated domains Πk , k = 1, . . . , m, are hyperbolic. Since Π ⊂ C, we may suppose, without loss of generality, that Π1 ⊂ C, and ∞ ∈ Πk ⊂ C, k = 2, . . . , m.

100

Chapter 6. Multiply connected domains

Since f ∈ A(Δ, Π) ⊂ A(Δ, Πk ), we can apply theorems of Chapter 5 to get An (Π) ≤ sup{A(w, n) | w ∈ Π}, where n ∈ N and A(w, n) :=

min{σk (w) R(w, Πk ) | k = 1, . . . , m} R(w, Π)

(6.6)

with σ1 (w) ≡ n and σk (w) = 4n−1 (1 − exp(−4δk (w)))1−n ,

k = 2, . . . , m.

In the latter formula the quantity δk (w) = δ(w, ∞, Πk ) means the hyperbolic distance from w to ∞ in the domain Πk . The function A(·, n) is positive and continuous on Π, since hyperbolic radii and the functions σk are positive and real-analytic functions on Π. Suppose that An (Π) = ∞ for one number n ∈ N. Then there exists a sequence of points wj ∈ Π, j ∈ N, such that wj → w0 ∈ ∂Π,

and A(wj , n) → ∞ as

j → ∞.

But this is impossible. Actually, if w0 ∈ ∂Π1 , then lim A(wj , n) ≤ n lim

j→∞

j→∞

R(wj , Π1 ) =n R(wj , Π)

according to (6.6) and Lemma 6.2. If w0 ∈ ∂Πk for some k ∈ {2, . . . , m}, then lim A(wj , n) ≤ 4n−1 lim (1 − exp(−4δk (wj )))1−n

j→∞

j→∞

R(wj , Πk ) = 4n−1 R(wj , Π)

according to (6.6), Lemma 6.2 and the equation lim δk (w) = δ(w, ∞, Πk ) = ∞.

w→w0

Thus, our assumption leads to a contradiction and therefore (b) holds. (b)⇒(c). Let f ∈ A(Δ, Π). Then, for any ﬁxed z ∈ Δ, the function ζ +z , g : Δ → Π, ζ → g(ζ) = f 1 − zζ also belongs to A(Δ, Π). Using the formula (4.3) f (n) (z)(1 − |z|2 )n = n! n

k=1

n−1 n−k

z n−k ak (g)

6.1. Finitely connected domains

101

and the equation f (z) = a0 (g), we obtain Cn (Δ, Π) ≤ 2n−1 max Ak (Π). 1≤k≤n

This implies the desired implication. (c)⇒(d). Let Φ := ΦΠ be a covering map of Δ onto Π. On the one hand, Φ ∈ A(Δ, Π). Hence, the validity of (c) implies that for all z ∈ Δ, (1 − |z|2 )2 |Φ (z)| ≤ 2 C2 (Δ, Π) < ∞. R(Φ(z), Π) On the other hand, R(w, Π) = |Φ (z)|(1 − |z|2 ) for w = Φ(z), z ∈ Δ. Straightforward computations yield that ∂ R(w, Π) = (1 − |z|2 ) Φ (z) − 2 z |∇R(w, Π)| = 2 ∂w Φ (z) (1 − |z|2 )2 |Φ (z)| |Φ (z)| = − 2 z . R(Φ(z), Π) Φ (z) Therefore |∇R(w, Π)| ≤ 2 + 2 C2 (Δ, Π),

w ∈ Π.

Since R(w, Π) = 0 for w ∈ ∂Π, we immediately get R(w, Π) ≤ (2 + 2 C2 (Δ, Π)) dist(w, ∂Π),

w ∈ Π.

This proves the right inequality. The left inequality in (d) is true for any hyperbolic domain (see [3] and [55]). This completes the proof of (c)⇒(d). The implication (d)⇒(a) is a simple consequence of known facts. Actually, if ∂Π contains an isolated point w0 , then (see [3] and [55]) R(w, Π) = (2 + o(1))|w − w0 | log

1 , |w − w0 |

w → w0 ∈ C,

(6.7)

and R(w, Π) = (2 + o(1))|w| log |w|,

w → w0 = ∞.

(6.8)

Hence, (d) cannot hold if ∂Π contains an isolated point. This completes the proof of Lemma 6.3.

Proof of Theorem 6.1. Suppose that ∂Ω and ∂Π do not contain isolated points and that f ∈ A(Ω, Π). For any ﬁxed z ∈ Ω we consider the function g : Δ → Π, ζ → g(ζ) = f (z + ρζ),

102

Chapter 6. Multiply connected domains

where ρ = dist(z, ∂Ω). It it evident that g ∈ A(Δ, Ω) and that a0 (g) = g(0) = f (z)

and an (g) = ρn f (n) (z), n ∈ N.

Since ∂Π does not contain isolated points, Lemma 6.3 implies that the quantities An (Π) are ﬁnite for all n ∈ N. Hence, (dist(z, ∂Ω))n |f (n) (z)| |an (g)| = ≤ An (Π), R(f (z), Π) R(a0 (g), Π)

(6.9)

for all n ∈ N. Using the equivalence (a)⇔(d) of Lemma 6.3 with respect to Ω, we get R(z, Ω) ≤ KΩ dist(z, ∂Ω), z ∈ Ω, for a positive constant KΩ . This inequality and (6.9) imply that |f (n) (z)| ≤ An (Π)(KΩ )n

R(f (z), Π) , (R(z, Ω))n

z ∈ Ω.

Thus,

An (Π)(KΩ )n < ∞, n! for all n ∈ N. This completes the proof of one direction of the assertion. Suppose now that Ω and Π are ﬁnitely connected hyperbolic domains in C and that C2 (Ω, Π) is ﬁnite. Firstly, we use some counterexamples to show that ∂Π cannot have isolated points. Suppose on the contrary that ∂Π has at least one isolated point w0 . Since Cn (Ω, Π) is invariant under linear transformations of Ω and Π, we may restrict ourselves to consideration of the following two cases without loss of generality. Either Cn (Ω, Π) ≤

w0 = 0 ∈ ∂Π

and

Δ = Δ \ {0} ⊂ Π

(6.10)

or w0 = ∞ ∈ ∂Π

and D∞ = {w | 1 < |w| < ∞} ⊂ Π.

(6.11)

In the case (6.10), we consider the functions Ψt := ft ◦ Φ−1 Ω ∈ A(Ω, Δ ) ⊂ A(Ω, Π),

t ∈ (0, ∞),

where Φ−1 Ω is the inverse of a covering map ΦΩ of Δ onto Ω and ft is deﬁned by 1+ζ . ft : Δ → Δ , ζ → ft (ζ) = exp −t 1−ζ On the one hand, at z0 = ΦΩ (0) the condition C2 (Ω, Π) < ∞ implies |Ψt (z0 )| (R(z0 , Ω))2 ≤ R(Ψt (z0 ), Π) = R(e−t , Π) = (2 + o(1))e−t t 2 C2 (Ω, Π)

6.1. Finitely connected domains

103

as t → ∞. On the other hand, straightforward computations show that this is impossible, since ΦΩ (0) 2 −t 2 . (6.12) |Ψt (z0 )| (R(z0 , Ω)) = 4 e t − t + t Φ (0) Ω

In the case (6.11) we examine the functions ˜ t := f˜t ◦ Φ−1 ∈ A(Ω, D∞ ) ⊂ A(Ω, Π), Ψ Ω

t ∈ (0, ∞),

where Φ−1 Ω is deﬁned as above and

1+ζ . f˜t : Δ → D∞ , ζ → f˜t (ζ) = exp t 1−ζ

In this case, we get for z0 = ΦΩ (0), ˜ Ψt (z0 ) (R(z0 , Ω))2 = et t2 (4 + o(1)), as t → ∞. This contradicts the condition C2 (Ω, Π) < ∞ which implies ˜ Ψt (z0 ) (R(z0 , Ω))2 ≤ R(et , Π) = (2 + o(1))et t, 2 C2 (Ω, Π) as t → ∞. Thus, ∂Π does not contain isolated points. Now, we prove that the analogous assertion holds true for Ω. Firstly, we have just proved that there exists a positive constant KΠ such that |∇R(w, Π)| ≤ KΠ ,

w ∈ Π.

Now, let h be a covering map of Ω onto Π deﬁned by h = ΦΠ ◦ Φ−1 Ω ∈ A(Ω, Π). Computations using the equality R(h(z), Π) = |h (z)|R(z, Ω) show that h (z)(R(z, Ω))2 (h (z))2 ∂R(w, Π) h (z) ∂R(z, Ω) = − , 2 2 R(w, Π) |h (z)| ∂w |h (z)| ∂z where w = h(z), z ∈ Ω. It follows that |∇R(z, Ω)| ≤ |∇R(w, Π)| +

|h (z)|(R(z, Ω))2 ≤ KΠ + C2 (Ω, Π) 2 R(w, Π)

for any z ∈ Ω. The boundedness of |∇R(·, Ω)| on Ω implies that ∂Ω has no isolated point. This completes the proof of Theorem 6.1.

104

6.2

Chapter 6. Multiply connected domains

Pairs of arbitrary domains

We shall show that it is possible to deﬁne ﬁnite modiﬁed punishing factors even if Ω contains the point at inﬁnity and ∂Ω contains isolated points. To this end, we use the Euclidean distance, dist(w, ∂Ω), from a point to the boundary of a domain. As a modiﬁcation of Theorem 6.1 we get Theorem 6.4 ([21]). Let Ω be an arbitrary hyperbolic domain in C and Π a ﬁnitely connected hyperbolic domain in C. Then the modiﬁed punishing factors C˜n (Ω, Π) :=

sup

sup

f ∈A(Ω,Π) z∈Ω

|f (n) (z)|(dist(z, ∂Ω))n n! dist(f (z), ∂Π)

are ﬁnite for all n ∈ N if and only if ∂Π does not contain isolated points. Proof. Let f ∈ A(Ω, Π). If Π is a ﬁnitely connected hyperbolic domain in C and ∂Π does not contain isolated points, then the inequality (6.9) holds for any n ∈ N. Moreover, by the statement (d) of Lemma 6.3, there exists a positive constant KΠ such that R(f (z), Π) ≤ KΠ dist(f (z), ∂Π), z ∈ Ω. This inequality and (6.9) imply that n! C˜n (Ω, Π) ≤ An (Π) KΠ ,

n ∈ N.

This proves one direction of the assertion of Theorem 6.4. We prove the other ˜ t from the proof of Theorem direction considering again the functions Ψt and Ψ 6.1. Actually, suppose that Π is a ﬁnitely connected hyperbolic domain in C, C˜2 (Ω, Π) < ∞, and (6.10) holds. Then |Ψt (z0 )| dist(z0 , ∂Ω) ≤ dist(e−t , ∂Π) = e−t . 2 C˜2 (Ω, Π) If we let t → ∞, we see that this contradicts (6.12), since the quotient dist(z0 , ∂Ω)/R(z0 , Ω) does not depend on t. A similar consideration in the case (6.11) reveals an analoguous contradiction. The next theorem gives a further modiﬁcation of Schwarz-Pick type inequalities for any pair of hyperbolic domains. Theorem 6.5 ([21]). Let Ω and Π be arbitrary hyperbolic domains in C. Then for any f ∈ A(Ω, Π), any n ∈ N, and any z ∈ Ω \ {∞} the estimate n R(f (z), Π) 4n n! (n) |f (z)| ≤ (dist(f (z), ∂Π))n−1 dist(z, ∂Ω) is valid, if f (z) = ∞.

6.2. Pairs of arbitrary domains

105

Proof. By the proof of Theorem 6.1, for f ∈ A(Ω, Π) and any ﬁxed z ∈ Ω (z = ∞, f (z) = ∞) we have (dist(z, ∂Ω))n |f (n) (z)| = n! |an (g)|,

n ∈ N,

(6.13)

where g ∈ A(Δ, Π) is deﬁned as above and a0 (g) = g(0) = f (z). Now, let w0 = f (z) and let ΦΠ be a covering map from Δ to Π, such that ΦΠ is a locally univalent function on Δ and ΦΠ (0) = R(w0 , Π). Moreover, for any simply connected domain Π ⊂ Π, any branch Φ−1 Π Π is a univalent function. We consider the branch Ψ = Φ−1 D(w0 , ρ) = {w | |w − w0 | < ρ = dist(w0 , ∂Π)}, Π D(w ,ρ) , 0

satisfying the condition Ψ(w0 ) = 0. The Koebe one-quarter theorem implies that the function R(w0 , Π) h : Δ → C, ζ → h(ζ) = Ψ(w0 + ρζ) ρ attains all values in the disc D1/4 = {z | |z| < 1/4}. Hence, Da =

z | |z| < a =

ρ 4 R(w0 , Π)

⊂ Ψ(D(w0 , ρ)).

(6.14)

Thus, ΦΠ is univalent on Da . As a consequence of the monodromy theorem, the function ω deﬁned by ω : Δ → Δ, ζ → ω(ζ) = Φ−1 Π ◦ g(ζ),

ω(0) = 0,

is a holomorphic self-map of Δ. Hence, |ω(ζ)| ≤ |ζ| for all ζ ∈ Δ. This implies g(Da ) ⊂ ΦΠ (Da ) ⊂ D(w0 , ρ), since g(ζ) = ΦΠ (ω(ζ)). Applying Cauchy’s estimates for the Taylor coeﬃcients of the function g − w0 in the disc Da , we get an |an (g)| ≤ ρ,

n ∈ N.

(6.15)

From (6.13), (6.14), and (6.15) it follows that (dist(z, ∂Ω))n |f (n) (z)| ≤

n! ρ 4n n! (R(w0 , Π))n ≤ , n a ρn−1

which is the desired inequality of Theorem 6.5.

106

Chapter 6. Multiply connected domains Next, we shall examine some facts connected with Theorem 4.18.

Let Ω and Π be open sets on the Riemann sphere C that are equipped with the Poincar´e metric of curvature −4. According to Poincar´e’s generalization of Riemann’s mapping theorem, this means that the boundaries of Ω and Π contain at least three points in C and that the density of this metric in the unit disc Δ = {z | |z| < 1} is deﬁned as λΔ (z) =

1 , 1 − |z|2

z ∈ Δ.

As usual, we consider the set A(Ω, Π) of functions f : Ω → Π, which are locally holomorphic or meromorphic and in general multivalued. Let n ∈ N and Qn (z, w, Ω, Π) be deﬁned as the smallest possible value such that the inequality (n) n f (z) (λΩ (z)) ≤ Qn (z, w, Ω, Π) n! λΠ (w) holds for all f ∈ A(Ω, Π), f (z) = w. For any pair (z, w) ∈ Ω × Π and any n ∈ N, normal family arguments give that there exists a function f ∈ A(Ω, Π) such that f (z) = w and (n) n f (z) (λΩ (z)) = Qn (z, f (z), Ω, Π) . n! λΠ (f (z)) Moreover, it is easily seen that Cn (Ω, Π) = sup{Qn (z, w, Ω, Π) | (z, w) ∈ Ω × Π}. Here we shall need the function F deﬁned in the proof of Theorem 4.18 by equation p + q, if p + 2q ≥ 4, (6.16) F (p, q) := 2 + p2 /(8 − 4q), if p + 2q < 4, for p ≥ 0 and q ≥ 0. Theorem 6.6 ([27]). Let Ω and Π be open sets on the Riemann sphere that are equipped with the Poincar´e metric of curvature −4. Then for any f ∈ A(Ω, Π) the estimate 2 (λΩ (z)) |f (z)| ≤ F (p, q) (6.17) λΠ (w) is valid for w = f (z), z ∈ Ω = C ∩ Ω, w ∈ Π = C ∩ Π, p = |∇ (1/λΩ (z))|, and q = |∇ (1/λΠ (w))|. The inequality (6.17) is sharp for any (z, w) ∈ Ω × Π . Proof. We ﬁx (z, w) ∈ Ω × Π and consider a function f ∈ A(Ω, Π), f (z) = w. Without loss of generality, we assume that Ω and Π equal their components that contain z and w, respectively. For this case the assertion follows from the proof of Theorem 4.18.

6.2. Pairs of arbitrary domains

107

As in Chapter 5 we denote γ(Ω) = sup |∇ (1/λΩ (z))| , z∈Ω

γ(Π) = sup |∇ (1/λΠ (w))| w∈Π

and we want to give some bounds for C2 (Ω, Π) dependent on the Euclidean geometry and on the conformal geometry of these sets. From Chapter 3 and Theorem 6.6 we immediately have the following corollaries. Corollary 6.7. Let Ω and Π be open sets in C equipped with the usual Poincar´e metric. Then C2 (Ω, Π) is ﬁnite if and only if C\Ω and C\Π are uniformly perfect. Corollary 6.8. Let Ω and Π be open sets in C equipped with the usual Poincar´e metric. Then the following assertions are valid. (i) C2 (Ω, Π) =

1 2 (γ(Ω)

+ γ(Π)),

(ii) 2(M0 (Ω) + M0 (Π)) ≤ C2 (Ω, Π) ≤ 2π(M0 (Ω) + M0 (Π)) + Γ(1/4)4 /π 2 , (iii) 2(M (Ω) + M (Π)) − 2 ≤ C2 (Ω, Π) ≤ 2π(M (Ω) + M (Π)) + Γ(1/4)4 /π 2 , (iv) C2 (Ω, Π) ≥ 2, where equality is attained if and only if all components of Ω and Π are convex. Together with the results of Chapter 4, where we proved that Cn (Ω, Π) = 2n−1 for convex domains Ω and Π, Corollary 6.8 (iv) delivers the proof of a weak form of a conjecture published in [16]. We want to formulate the remaining parts of a stronger conjecture as Conjecture. Let n ∈ N \ {1, 2} and let Ω and Π be open sets in C equipped with the usual Poincar´e metric. Then Cn (Ω, Π) = 2n−1 implies that all components of Ω and Π are convex domains. Now, we turn to the cases where the point at inﬁnity belongs to Ω or Π and we prepare these considerations with a lemma which may deserve some interest of its own. Lemma 6.9 ([27]). Let Ω be an open set in C equipped with the usual Poincar´e metric and let ∞ ∈ Ω. Then the equations (i) lim |∇ (1/λΩ (z))| = ∞, z→∞

(ii) lim |∇ log λΩ (z)|dist(z, ∂Ω) = 2, z→∞

are valid. Proof. Without restriction of generality we assume Ω to be a hyperbolic domain. To prove (i) we consider a universal covering map f of Δ onto Ω such that f (0) = ∞. Let f have the expansion a z = f (ζ) = + ϕ(ζ), ζ ∈ Δ, ζ

108

Chapter 6. Multiply connected domains

where a = 0 and ϕ is a function meromorphic in Δ and holomorphic in a neighbourhood of the origin. Direct computations using λΩ (z) =

1 |f (ζ)|(1

imply |∇ (1/λΩ (z))| =

− |ζ|2 )

,

z = f (ζ),

1 2a + ζ 3 ϕ (ζ) 2 + 2 |ζ| . 2 |ζ| a − ζ ϕ (ζ)

(6.18)

(6.19)

If we take the limit z → ∞ of the left side of equation (6.19), which is equivalent to taking the limit ζ → 0 of the right side, we see that (i) is valid. To prove (ii) we multiply the left sides and the right sides of (6.18) and (6.19) with one another. If we multiply these products with |z| and |a/ζ|, we get a (6.20) lim |∇ log λΩ (z)||z| = lim |∇ log λΩ (f (ζ))| = 2. z→∞ ζ→0 ζ Since ∂Ω is bounded, lim

z→∞

dist(z, ∂Ω) = 1. |z|

(6.21)

The formulae (6.20) and (6.21) imply (ii). This completes the proof of Lemma 6.9. Lemma 6.9 (ii) gives rise to a sharp inﬁnity version of the Osgood-Jørgensen theorem. For hyperbolic domains Ω ⊂ C the fundamental inequality |∇ log λΩ (z)|dist(z, ∂Ω) ≤ 2,

z ∈ Ω,

(6.22)

was discovered by Osgood (see [121], and compare also [111]). The proof uses certain results of Jørgensen (see [88]). One may observe that this proof in [121] and the related results in [88] do not use the fact that ∞ ∈ / Ω. Moreover, in [88] Jørgensen considers the case when ∞ ∈ Ω as a basic one. This observation together with Lemma 6.9 (ii) leads us to the following theorem which proves (3.36). Theorem 6.10 ([21]). Let Ω be an open set in C equipped with the usual Poincar´e metric. For any z ∈ Ω ∩ C the inequality (6.22) is valid. The constant 2 in (6.22) cannot be replaced by a smaller one if ∞ ∈ Ω.

6.3 Some examples Let Π be a hyperbolic domain in the extended complex plane C. For ﬁxed w ∈ Π \ {∞} we consider the local Taylor expansions f (z) = w +

∞ n=1

an (w, f )z n

6.3. Some examples

109

valid in a neighbourhood of the origin. We will discuss some examples showing the behaviour of the quantities An (w, Π) = sup{|an (w, f )| | f ∈ A(Δ, Π), f (0) = w} regarded as a function of w (see [29]). Example 1. We consider the doubly connected domain Π = D∞ := {z ∈ C | |z| > 1} and we prove that for n ∈ N, An (w, D∞ ) = Fn (− log |w|), where −t

Fn (t) = − 2 t e

n−1 k=0

n−1 k

(−2t)k , (k + 1)!

(6.23)

n ∈ N ∪ {0},

is Bateman’s function of order n (compare [35] and [94]). The upper bound is attained if and only if there exist ϕ1 , ϕ2 ∈ R such that f (z) = eiϕ1 ΦD∞ ,w eiϕ2 z . Further, 1 An (w, D∞ )(dist(w, ∂D∞ ))n−1 = w→∞ (R(w, D∞ ))n n! lim

and

An (w, D∞ )(dist(w, ∂D∞ ))n−1 = 21−n . (R(w, D∞ ))n |w|→1 lim

To prove these assertions we use that, for w = |w|eiθ , t =: log |w| > 0, the function ΦD∞ ,w is deﬁned by 1 + e−iθ z iθ ΦD∞ ,w (z) = e exp t 1 − e−iθ z and that for any f ∈ A(Δ, D∞ ), f (0) = w, there exists a uniquely deﬁned function p holomorphic in Δ such that Re p(z) > 0 for z ∈ Δ, p(0) = 1, and f (z) = eiθ exp(t p(z)). The ﬁrst equation yields R(w, D∞ ) = 2|w| log |w| and the second by diﬀerentiation f (z) = t p (z)f (z).

(6.24)

110

Chapter 6. Multiply connected domains

Using the Taylor coeﬃcients of the expansion p(z) = 1 +

∞

pk z k ,

k=1

we see that a1 (w, f ) = t w p1 ,

n−1

n an (w, f ) = t w n pn + ak (w, f )(n − k)pn−k , n ≥ 2. k=1

To prove (6.23) and the assertion on the extremal function we remember that the inequalities |pk | ≤ 2 are valid for k = 1, . . . , n, with equality if and only if p(z) = (1 + cz)/(1 − cz), |c| = 1, and the triangle inequality. Further we use the identity n−1 nFn (t) = − 2 t (n − k)Fk (t) k=0

which we were not able to ﬁnd anywhere, but which may be easily proved by mathematical induction and the diﬀerence equation (n − 1)(Fn (t) − Fn−1 (t)) + (n + 1)(Fn (t) − Fn+1 (t)) = 2tFn (t) (compare [35] and [94]). Since lim

t→∞

Fn (−t) 1 = , (2t)n et n!

the ﬁrst assertion on the asymptotics of An (w, D∞ ) is an immediate consequence of (6.23) and (6.24). The second one follows likewise. This example shows that there are domains where the universal covering functions deliver the extrema for the moduli of all local Taylor coeﬃcients. Example 2. For the doubly connected domain Δ = {z | 0 < |z| < 1}, w = |w|eiθ , t = − log |w| > 0, the universal covering function is given by 1 + e−iθ z iθ ΦΔ ,w = e exp −t 1 − e−iθ z and f ∈ A(Δ, Δ ), f (0) = w, may be written as f (z) = eiθ exp(−t p(z)), where p is as in Example 1. It is easily seen by the same method as in Example 1 that |an (w, ΦD0 ,w )| = |Fn (t)|.

6.3. Some examples

111

This has been proved in [94]. Notwithstanding the fact that the determination of An (w, Δ ) is very diﬃcult the methods of Example 1 may be applied to get the right asymptotics near the origin. An analogous reasoning shows that for f ∈ A(Δ, Δ ), f (0) = w, |p1 |n |an (w, f )|(dist(w, ∂Δ ))n−1 = (R(w, Δ ))n 2n n! |w|→ 0 lim

if we ﬁx p in the above representation of f . Since |p1 | ≤ 2 where equality is attained if f equals a universal covering function, we get An (w, Δ )(dist(w, ∂Δ ))n−1 1 = . (R(w, Δ ))n n! |w|→ 0 lim

In a similar way, we get |an (w, f )|(dist(w, ∂Δ ))n−1 |pn | = n (R(w, Δ ))n 2 |w|→ 1 lim

and conclude

An (w, Δ )(dist(w, ∂Δ ))n−1 = 21−n . (R(w, Δ ))n |w|→ 1 lim

Example 3. Let δ > 1 and the annulus Aδ be deﬁned by Aδ = {z | 1 < |z| < δ}. It is known that the hyperbolic radius in this case has the form log δ log |w| . R (w, Aδ ) = 2 |w| sin π π log δ To get the asymptotics near the boundary we use the following representation of f ∈ A(Δ, Aδ ), f (0) = w. Let τ > 0 and t ∈ (−π/2, π/2) be chosen such that π log δ log τ log δ t+ exp − i . w = exp π 2 π Then there exists a function p that is holomorphic and has positive real part in Δ, p(0) = 1, such that log δ log (i τ (cos t p(z) + i sin t)) . f (z) = exp − i π Diﬀerentiation yields (cos t p(z) + i sin t)f (z) = − i cos t

log δ p (z) f (z). π

112

Chapter 6. Multiply connected domains

The consideration of the Taylor coeﬃcients in this formula yields a1 (w, f ) = − i p1 w

log δ cos t π

and n an (w, f )eit + cos t

n−1

k ak pn−k

k=1

n−1 log δ = −i ak (n − k)pn−k . cos t n pn w + π k=1

By a reasoning analogous to that of the second example we get lim

w→∂Aδ

An (w, Aδ )(dist(w, ∂Aδ ))n−1 = 21−n . (R(w, Aδ ))n

To get a global estimate in this case we use A(Δ, Aδ ) ⊂ A(Δ, Δ) for w ∈ Aδ and |f (z)| < δ for z ∈ Δ. These facts yield immediately δ 2 − |w|2 . An (w, Aδ ) ≤ min Fn (− log |w|), δ The diﬀerence between the ﬁrst two examples and the third one lies in the fact that the quantity R(w, Ω)/dist(w∂Ω) is unbounded in the ﬁrst two cases and bounded in the last one.

Chapter 7

Related results 7.1

Inequalities for schlicht functions

First, we will give an outline of the ideas and results that led to the conjectures of Chua. To our knowledge, E. Landau was the ﬁrst who considered the possibility to follow G. Pick’s program as indicated in the introduction for the higher derivatives of schlicht functions. He proved the following theorem (compare Landau [98], Gong [71]). Theorem 7.1. If the Bieberbach conjecture is valid for n ≥ 2, then for these n, any z0 ∈ Δ, and any function f holomorphic and injective on Δ, the inequality |f (n) (z0 )| n + |z0 | ≤ |f (0)| n! (1 − |z0 |)n+2

(7.1)

is valid. It is clear that, on the other hand, the validity of the Bieberbach conjecture follows from (7.1). The proof uses automorphisms of the unit disc in the same way as the formula of Sz´ asz mentioned above. Since the same method was used to prove a similar formula of Jakubowski that originated in the ﬁrst half of the last century, we give a uniﬁed proof for these two theorems after the presentation of Jakubowski’s Theorem (see [85]). Theorem 7.2. If the Bieberbach conjecture is valid for n ≥ 2, then for these n, any z0 ∈ Δ, and any function f holomorphic and injective on Δ, the inequality |f (n) (z0 )| (n + |z0 |)(1 + |z0 |)n−2 ≤ |f (z0 )| n! (1 − |z0 |2 )n−1 is valid.

(7.2)

114

Chapter 7. Related results

Proof of Theorems 7.1 and 7.2. Let f be injective on Δ. The function

ζ+z0 f 1−z 0ζ g(ζ) = (1 − |z0 |2 )f (z0 ) is injective on Δ and g (0) = 1. Using Szasz’s formula and de Branges’ theorem for ∞ g(ζ) = g(0) + z + ak ζ k k=2

we get immediately, as in Chapter 4, |f (n) (z0 )|(1 − |z0 |2 )n ≤ 2 (1 − |z0 | )|f (z0 )| n! n

k=1

n−1 n−k

|z0 |n−k k = (n + |z0 |)(1 + |z0 |)n−2 .

Apparently, this is Jakubowski’s theorem. Since Landau was interested in an upper bound dependent on |z0 | only, he used in his proof the distortion theorem |f (z0 )| 1 + |z0 | ≤ |f (0)| (1 − |z0 |)3 for functions injective on Δ. Combining this with Jakubowski’s bound delivers Landau’s theorem. The case of equality in Jakubowski’s results has been discussed by Yamashita in [169] in detail. Further, Jakubowski recognized that by the same method, using L¨ owner’s theorem for the MacLaurin coeﬃcients of functions f convex and injective on Δ, one gets the bound |f (n) (z0 )| (1 + |z0 |)n−1 . ≤ |f (z0 )| n! (1 − |z0 |2 )n−1 Here, again the case of equality is due to Yamashita (see [169]). A natural simplifying of these bounds is achieved, if one replaces in the numerator |z0 | by 1. This results in the factors (n + 1)2n−2 in the case of injective functions and 2n−1 in the case of convex functions. These constants play an important part in the present book, Chapter 5. Possibly, the resulting formula, the identity R(z0 , Δ) = 1 − |z0 |2 and naturally de Branges’ theorem motivated Chua in 1996 to consider functions f injective or convex on simply connected proper subdomains Ω of C (see [56]).

7.1. Inequalities for schlicht functions

115

He proceeded as we did in Chapter 4 with the exception that for him there was no need to consider subordination. Therefore he arrived at a formula analogous to our formula (4.5). The ﬁrst diﬀerence is that in his formula there appear the coeﬃcients of schlicht or convex functions and not those of subordinates to them. The second diﬀerence is that he could use the identity R(f (z0 ), f (Ω)) = |f (z0 )| R(z0 , Ω). Concerning a formula equivalent to the formula of Sz´ asz that we used, he found a formula proved by Todorov in [163]. Naturally, he had to consider the coeﬃcients of powers of inverses of schlicht functions, as we do. He knew the proofs of L¨ owner (see [110] and of Schober (see [147]) for the coeﬃcients of the inverses themselves, but he was not aware of the fact that the extremality of the Koebe function for powers, where the exponent is a natural number, is a simple consequence of L¨owner’s theorem. Therefore, he proved this fact using Baernstein’s integral mean theorem (see [32]) and the Schur-Jabotinski theorem (see [81], Thm 1.9.a). By these theorems, he could prove that for f injective on Ω, z0 ∈ Ω, the inequality 4n−1 |f (n) (z0 )| , ≤ |f (z0 )| n! (R(z0 , Ω))n−1 is valid , whereas for f convex and injective he got 2n − 1 n |f (n) (z0 )| , ≤ |f (z0 )| n! (R(z0 , Ω))n−1

n ≥ 2,

n ≥ 2.

Further, Chua recognized that, for the consideration of similar problems for functions f injective or convex on convex proper subdomains of C that led him to his conjectures, one needs bounds for the inverses of convex functions. Using a theorem due to Trimble (see [164]) he suceeded in proving the conjectures up to n = 4. The results of Li (see [102] and [103]), who used theorems of Libera and Zlotkiewicz (see [104]) on the coeﬃcients of inverses of convex functions, imply that these conjectures are valid for n ≤ 8. As we have seen in Chapter 4, the computation of punishing factors is possible, if one has a good knowledge of the inverse coeﬃcients of certain functions on one hand and good results on the coeﬃcients of subordinated functions on the other hand. If one reduces the interest to the conformal maps themselves, one can replace the subordination results by theorems on the bounds of coeﬃcients of certain injective functions. We want to cite two examples where the proofs of the details are the same as in Chapter 4 and Chapter 5. In addition we need the interplay between a geometric property, the accessibility of order β and an analytic property, the close-to-convexity of order β.

116

Chapter 7. Related results

Deﬁnition 7.3. A domain Π is called (angularly) accessible of order β, β ∈ [0, 1], if it is the complement of a union of rays that are pairwise disjoint except that the origin of one ray may lie on another one of the rays, and such that every ray is the bisector of a sector of angle (1 − β)π that wholly lies in the complement of Π. We use the following characterization of domains accessible of order β. Theorem 7.4 (see for instance [127] and [93] and compare [41]). Let f be injective and holomorphic in the unit disc and f (Δ) accessible of order β. Then there exist α ∈ [0, 2π], and functions g and p holomorphic in Δ such that the following conditions hold: (i) g(Δ) is convex, (ii) Re(eiα p(z)) > 0 for z ∈ Δ and p(0) = 1, (iii) f (z) = (p(z))β g (z) for z ∈ Δ. Functions satisfying the proprties (i)–(iii) are called (nonnormalized) functions close-to-convex of order β and we owe to Ahoronov and Friedland the proof (see [2] and compare [146]) that for such functions the following inequalities hold. Theorem 7.5. Let f be close-to-convex of order β. Then 1+β |f (n) (0)| 1+z 1 dn , ≤ |f (0)| 2(β + 1) (dz)n 1 − z

n ∈ N.

z=0

These theorems and the considerations and computations of Chapter 5 immediately imply the following theorems. Theorem 7.6. Let f be holomorphic and injective on a simply connected proper subdomain of C and such that f (Δ) is linearly accessible of order β, β ∈ [0, 1]. Then for z0 ∈ Ω and n ∈ N the inequalities β+1 |f (n) (z0 )| 4n−1 1 2 +n−1 ≤ n |f (z0 )| (R(z0 , Ω))n−1 β + 1 are valid. Theorem 7.7. Let f be holomorphic and injective on a simply connected proper subdomain Ω of C, ∞ ∈ Ω, and such that f (Δ) is linearly accessible of order β, β ∈ [0, 1]. Then for z0 ∈ Ω \ {∞} and n ∈ N and p(z0 ) deﬁned as above, the inequalities 4k |f ( n)(z0 )| 1 ≤ n−1 |f (z0 )| (R(z0 , Ω)) 2(β + 1) n

k=1

are valid.

n−1 n−k

β+1 2

k

p(z0 ) +

n−k 1 +2 p(z0 )

7.2. Derivatives of α-invariant functions

7.2

117

Derivatives of α-invariant functions

Let Ω be a simply connected domain in C with the density λΩ of the Poincar´e metric. Our aim is to ﬁnd bounds for the function Mn (a, Uα (Ω)) =

1 max{|f (n) (a)| : f ∈ Uα (Ω)}, n!

a ∈ Ω,

where α is a parameter and Uα (Ω) is a family of holomorphic functions with a given property. For instance, Uα (Ω) is the unit ball f ≤ 1 in one of the usual spaces of holomorphic functions. To ﬁnd Mn we will use the functions (see Chapter 3) μk = μk (a, Ω) =

1 ∂ k log λ2Ω (a) , λ−k Ω (a) (k − 1)! ∂ak

k = 1, . . . , n.

We start with a simple remark. For all conformal automorphisms ϕ : Ω → Ω, the condition f ∈ A(Ω, Π) implies that the functions f ◦ϕ are members of A(Ω, Π), too. With this observation in mind, we shall consider the following generalization of linear invariant families by Ch.Pommerenke [126] (compare also [8]). The ﬁrst deﬁnition deals with functions holomorphic in the unit disc Δ. Deﬁnition 7.8. Let α = const ≥ 0, and let Uα be a set of functions g such that: (i) g is holomorphic in Δ, (ii) for all conformal automorphisms ϕ : Δ → Δ, g ∈ Uα ⇒ gα, ϕ ∈ Uα , where gα,ϕ (ζ) := g(ϕ(ζ))ϕα (ζ), (iii) g(ζ) are uniformly bounded in the interior of Δ and (0, . . . , 0) is an interior point of the coeﬃcient region {(a0 , . . . , an ) : g(ζ) = a0 + a1 ζ + · · · ∈ Uα }. The following deﬁnition describes α- invariant families of functions holomorphic in a simply connected domain Ω. Deﬁnition 7.9. Given Uα , by Uα (Ω) we denote the set of all functions f such that f is holomorphic in Ω and fα, Φ ∈ Uα for all conformal mappings Φ of Δ onto Ω, where fα, Φ (ζ) = f (Φ(ζ))Φα (ζ). If ∞ ∈ Ω, then we suppose that 2α ∈ N ∪ {0}. Remark 7.10. The set A(Ω, Π) is a 0-invariant family. Let Pw (ζ) be the polynomial 1 +w1 ζ + · · · + wn ζ n . For g(ζ) = a0 + a1 ζ + · · · ∈ Uα , denote gn (ζ) = an +an−1 ζ+· · ·+a0 ζ n . We will consider the Hadamard product (gn ∗ Pw )(ζ) = an + an−1 w1 ζ + · · · + a0 wn ζ n .

118

Chapter 7. Related results

Theorem 7.11 ([14]). If a ∈ Ω and f (z) ∈ Uα (Ω), then Mn (a, Uα (Ω)) = λn+α (a) max max |(gn ∗ Pw )(ζ)|, Ω

(7.3)

g∈Uα |ζ|=1

where w = (τn,n−1 (α), τn,n−2 (α), . . . , τn,0 (α)), τn,k (α) are deﬁned by the following recurrent formulas: τk,k (α) = 1, 0 ≤ k ≤ n,

τm,k (α) =

m−k+1 s=1

τk,0 (α) =

k−1 α μk−s τs,0 (α), 1 ≤ k ≤ n, k s=0

1 τs−1,0 (1)τm−s,k−1 (α), 2 ≤ k ≤ m ≤ n. s

(7.4)

(7.5)

Proof. Let ψ be the conformal mapping of Ω to Δ, ψ(a) = 0, ψ (a) = λΩ (a) > 0. As in Chapter 3, we have 2ψ(z) ∂ log λ2Ω (z) ψ (z) ψ (z). = − ∂z ψ (z) 1 − ψ(z)ψ(z) Hence λkΩ (a)μk (a, Ω) and

1 1 ∂ k log λ2Ω (a) = = (k − 1)! ∂ak (k − 1)!

ψ (z) ψ (z)

(k−1)

(7.6) z=a

∞

ψ (z) k+1 λΩ (a)μk+1 (a, Ω)(z − a)k = ψ (z)

(7.7)

k=0

in some neighbourhood of a. Consider the function g(ζ) = f (Φ(ζ))Φα (ζ) = a0 + a1 ζ + . . . ,

ζ ∈ Δ,

(7.8)

where z = Φ(ζ) is the inverse of the function ζ = ψ(z). It follows from (7.8) that ∞ 1 (m) 1 k α (m) ak (ψ (z)ψ (z)) f (a) = m! m! k=0

=

m

z=a

ak tm,k (α)

k=0

with m! tm,k (α) = (ψ k (z)ψ α (z))(m) |z=a . We have to prove that tm,k (α)/λm+α (a) = τm,k (α) Ω

7.2. Derivatives of α-invariant functions

119

satisfy (7.4) and (7.5) for m = 1, 2, . . . , n. From ψ(z) = λΩ (a)(z − a)[1 + O(z − a)], (a) in accordance with τk,k (α) = 1. Using (7.6) we have directly tk,k (α) = λk+α Ω and the Leibniz formula, we have (k−1) 1 α α ψ (z) (k) α tk,0 (α) = (ψ (z)) ψ (z) = k! k! ψ (z) z=a z=a k−1 (k−1−s) (z) α ψ k−1 (ψ α (z))(s) = · s k! s=0 ψ (z) z=a

z=a

=

α k

k−1

λk−s Ω (a)μk−s ts,0 (α).

s=0

We also have 1 tm,k (α) = m! s=0 m

m s

(ψ k−1 (z)ψ α (z))(m−s)

(s)

(ψ(z))

z=a

m−k+1 1 (s−1) = tm−s,k−1 (α)(ψ (z)) s! s=1

=

m−k+1

z=a

s=1

z=a

1 tm−s,k−1 (α)ts−1,0 (1). s

Mathematical induction gives (7.4) and (7.5) for τn,k (α) = tn,k (α)/λn+α (a). Thus, Ω 1 (n) (a)(gn ∗ Pw )(1) f (a) = λn+α Ω n! for w = (τn,n−1 (α), τn,n−2 (α), . . . , τn,0 (α)), g(ζ) = f (Φ(ζ))Φα (ζ). Taking into account the property g(ζ) ∈ Uα ⇒ g(e−iθ ζ) ∈ Uα , we obtain 1 max{|f (n) (a)| : f ∈ Uα (Ω)} n! = λn+α (a) max |(gn ∗ Pw )(1)| = λn+α (a) max |(gn ∗ Pw )(eiθ )| Ω Ω

Mn (a, Uα (Ω)) =

g∈Uα

=

n+α λΩ (a)

g∈Uα

max max |(gn ∗ Pw )(ζ)|.

g∈Uα |ζ|=1

This completes the proof of Theorem 7.11.

Using the well-known properties of max{|F (ζ)| : |ζ| = r} (0 < r < ∞) for the analytic function F (ζ) = (gn ∗ Pw )(ζ) (see [59]), we obtain the following corollary .

120

Chapter 7. Related results

Corollary 7.12. Let w = (τn,n−1 (α), τn,n−2 (α), . . . , τn,0 (α)). The function (a) Kn (w, Uα ) = Mn (a, Uα (Ω))/λn+α Ω has the following properties: (i) Kn (w, Uα ) depends on a and Ω via w only, (ii) Kn (w, Uα ) is a convex function from Cn (= R2n ) to (0, ∞), (iii) there exist two positive constants c(n, α) and C(n, α) such that

c(n, α) ≤ Kn (w, Uα )/ 1 + w2 ≤ C(n, α),

w ∈ Cn ,

(iv) K(w1 , . . . , wn , Uα ) = K(eiθ w1 , . . . , einθ wn , Uα ) for any θ ∈ [0, 2π], (v) for w ∈ Cn , the function u(r) = Kn (rw1 , r2 w2 , . . . , rn wn , Uα ) is nondecreasing on 0 ≤ r ≤ ∞ and log u(r) is convex with respect to log r. In the case 0 ≤ k ≤ n − 1, τn,k (α) depends on n, k, α, a and Ω. For example, if Ω = Δ and a ∈ Δ, then n + 2α − 1 τn,k (α) = an−k . n−k Note that τn,k (α) does not depend on a ∈ Ω if and only if Ω is a half-plane. In the general case, for domains Ω ⊂ C we proved that τn,k (α) are bounded. Namely, by Theorem 3.27 we have the following assertion. If Ω ⊂ C, then sup sup |τn,k (α)| = Ω a∈Ω

n−k s=0

2n + 3α − 1 s

n−k−s+α−2 n−k−s

.

(7.9)

From Theorem 7.11 and Theorem 3.27 one immediately obtains the following corollary. Corollary 7.13. There exists an increasing convex function K1 (., Uα ) : [0, ∞) → (0, ∞) such that M1 (a, Uα (Ω)) = λ1+α Ω (a)K1 (t, Uα ),

t = α|∇ λ−1 Ω (a)|.

(7.10)

If Ω ⊂ C, then M1 (a, Uα (Ω)) = K1 (4α, Uα ). λ1+α a∈Ω Ω (a)

sup sup Ω

(7.11)

7.2. Derivatives of α-invariant functions

121

Consider now Hardy, Bergman and Bloch spaces (see for instance [59], [7]). We use the following notation: f p = sup

1/p |f (z)| |dz|

for f ∈ Ep (Ω), p > 0,

p

0 0, γ > 1,

for f ∈ Ap,γ (Ω); 1/p

f ∗p = sup |f (z)/λΩ (z)| for f ∈ Bp∗ (Ω), p > 0. z∈Ω

Theorem 7.14 ([14]). If t = |∇ λ−1 Ω (a)|, then for any a ∈ Ω, t f p , p ≥ 1, (a) 1 + p t 1+1/p f p , 0 < p ≤ 1, (a) (2 − p)1/p−1/2 21−1/p p−1/2 + |f (a)| ≤ λΩ p t 1+1/p f ∗p , p > 0, (a) (2 + p)1/p+1/2 2−1/p p−1/2 + |f (a)| ≤ λΩ p t 2Γ(γ + 1/2) 1+1/(γp) √ f p,γ , p ≥ 1, γ > 1, + (a) |f (a)| ≤ λΩ γp πΓ(γ) 1+1/p

|f (a)| ≤ λΩ

(7.12) (7.13) (7.14) (7.15)

where Γ denotes Euler’s gamma function. The estimates (7.12), (7.13), (7.14), (7.15) are asymptotically sharp as t → +∞, (7.12), (7.13) and (7.14) are also sharp at t = 0. Note that if U1/p is the unit ball in the Hardy space Hp , then the function K1 (t, U1/p ) is known by [114] in the cases p ≥ 1 + t/2 and 1 ≤ p ≤ max{2, 1 + t/2}. For Bloch space Bp∗ (Ω) we obtain from Corollary 7.13 and [165], 1+1/p

max |f (z)| = (2 + p)1/p+1/2 2−1/p p−1/2 λΩ

f ∗ p ≤1

(a)kp (|grad λ−1 Ω (a)|),

where kp (t) = max0≤x≤√1+2/p [1 + tx/p − (1 + 2/p)x2 ](1 − x2 )1/p . Proof of Theorem 7.14. From Theorem 7.11 we obtain M1 (a, Uα (Ω)) = sup (|a1 | + |τ1,0 (α)||a0 |)λ1+α Ω (a), g∈Uα

(7.16)

122

Chapter 7. Related results

where a0 = g(0), a1 = g (0), |τ1,0 (α)| = α|∇ λ−1 Ω (a)|. If α = 1/p, p ≥ 1 and g ∈ Hp , then |a0 | ≤ gp and |a1 | ≤ gp (see [59]). Taking into account Corollary 7.13, we have (7.12). If α = 1/p, p > 0 and g ∈ Bp∗ (Δ), then the region {(|a0 |, |a1 |) : g∗p ≤ 1} is given in [165], and we get (7.16) and (7.14). Consider the case g ∈ Ap,γ (Δ), p ≥ 1. By Hardy’s theorem, 2π p 2π|g(0)| ≤ |g(reiθ )|p dθ, 0 < r < 1. 0

This yields gp,γ ≥ |a0 |

γ−1 π

1

0

2r(1 − r2 )γ−2 dr

1/p = |a0 |.

We also have 1 r2 (1 − r2 )γ−2 dr gp,γ ≥ g1,γ ≥ 2|a1 |(γ − 1) 0 √ = |a1 | πΓ(γ)/(2Γ(γ + 1/2)). Hence, (7.15) holds. If g(ζ) ∈ Hp (0 < p < 1), then |a0 | ≤ gp (see [59]). Using the estimates from [157], we have |a1 | ≤ (2 − p)1/p−1/2 21−1/p p−1/2 . Thus, (7.13) holds. Theorem 7.15 ([14]). If f ∈ A2,γ (Ω), then for any a ∈ Ω and any γ ≥ 1, n 2 1/2 1 n+ 2γ 1 1 (n) M k+γ−1 ) τn,k (a) ( . (7.17) max |f (a)| = λΩ k n! f 2,γ ≤1 2γ k=0

Proof. By direct calculation, we obtain 1/2 ∞ k + γ − 1 −1 2 |ak | . g2,γ = k k=0

By the Cauchy-Schwarz inequality |

n

ak τn,k (α)| ≤ g2,γ

k=0

1/2 n k+γ−1 |τn,k (α)|2 . k

k=0

Equality holds for g(ζ) = a0 + a1 ζ + · · · + an ζ n if and only if k+γ−1 k = 0, 1, . . . , n, |τn,k (α)| = ak const, k which completes the proof of (7.17).

7.2. Derivatives of α-invariant functions

123

Theorem 7.16 ([14]). If f ∈ Ep (Ω), then for all a ∈ Ω, c(n, p) ≤

1 |f (n) (a)| ≤ C(n, p) max n+1/p n n! f p ≤1 λ (a)( |τn,k ( 1 )|2 )1/2 Ω

k=1

p

with the constants c(n, p) and C(n, p) such that C(n, p) = const n1/p−1/2 for 0 < p ≤ 2

c(n, p) = 1, and

c(n, p) = n1/2−1/p ,

C(n, p) = 1 for 2 ≤ p ≤ ∞.

Remark 7.17. For p = 2, Theorem 7.15 and Theorem 7.16 are well known (see for instance [114] ). Proof of Theorem 7.16. It is obvious that c(n, 2) = C(n, 2) = 1. Hence, c(n, p) ≥ 1 for 0 < p ≤ 2 and C(n, p) ≤ 1 for 2 ≤ p ≤ +∞, because Hp ⊂ Hx for p > x. Let p > 2. Hence, g2 ≤ gp . By the Hausdorﬀ-Young inequality, we have gn p ≤

n

1/q

|ak |

q

k=0

1 1 + =1 . p q

Consequently, by the H¨ older inequality, 1/2−1/p

gn p ≤ (n + 1)

n

1/2 2

|ak |

.

k=0

Thus, c(n, p) ≥ n1/2−1/p for p > 2. To complete the proof of Theorem 7.16, we have to prove that C(n, p) = O(n1/p−1/2 ) for 0 < p < 2. Let 0 < p ≤ 1. If g(ζ) = a0 +a1 ζ +· · · ∈ Hp and gp ≤ 1, then by the HardyLittlewood theorem (see Theorem 6.5 in [59]) an = o(n1/p−1 ). Consequently, C(n, p) = max ( gp ≤1

n

|ak |2 )1/2 = O(n1/p−1/2 ).

k=0

Let 1 < p < 2. If gp ≤ 1, then by the Hausdorﬀ-Young inequality,

∞

k=0

1/q |ak |

q

≤ gp

1 1 + =1 . p q

Using the H¨ older inequality, we obtain C(n, p) = O(n1/p−1/2 ).

124

7.3

Chapter 7. Related results

A characterization of convex domains

Everybody who reads Chapter 4, Section 5 very carefully will recognize that, for the proof of the 2n−1 conjecture, we only need one of the Marx-Strohh¨ acker theorems. This inequality characterises the functions of the closed convex hull of the convex functions, but not the convex functions themselves. On the other hand, we know from Chapter 4, Section 6 that convex pairs are the only ones for which C2 (Ω, Π) = 2. At ﬁrst glance, this seems to be a contradiction. Thinking of this problem, we found that an invariant formulation of the mentioned MarxStrohh¨ acker inequality in fact characterises convexity. The proof of this fact forms the content of this section and it seems to explain the geometrical background of Theorem 4.15. Let g0 be a function holomorphic and univalent in the unit disc Δ. We suppose that g0 (0) = 0, g0 (0) = 1 and Π := g0 (Δ) is a convex domain. To characterize Π we shall use the fact that for such a function g0 one of the well-known MarxStrohh¨ acker inequalities, namely Re

g0 (z) z

>

1 , 2

z ∈ Δ,

(7.18)

holds (see [115] and [155]). For uniﬁed proofs of this and many related inequalities one should consult [118]. The formula (7.18) is equivalent to the existence of a bounded holomorphic function ω : Δ → Δ such that g0 (z) 1 = , z 1 + zω(z)

z ∈ Δ.

(7.19)

We will consider the following question. Let h be a function holomorphic on acker Δ. Suppose that h (ζ) = 0 for any ζ ∈ Δ and that h(Δ) has the Marx-Strohh¨ property for any point z0 = h(t) ∈ h(Δ), i.e., the function g0 deﬁned by the Koebe transform

w+t − h(t) h 1+tw g0 (w) = h (t)(1 − |t|2 ) satisﬁes inequality (7.18) for any t ∈ Δ. This is equivalent to the inequality Re

h(z) − h(t) 1 − tz h (t) z−t

>

1 − |t|2 , 2

z ∈ Δ, t ∈ Δ.

(7.20)

What can be said about Ω = h(Δ)? We ﬁnd that h(Δ) is a convex domain, so that an assertion inverse to the Marx-Strohh¨ acker theorem is valid.

7.3. A characterization of convex domains

125

Theorem 7.18 ([24]). Let h be a function holomorphic in Δ, such that h (ζ) = 0, ζ ∈ Δ, and such that condition (7.20) is satisﬁed. Then (i) the function h is injective on Δ and h(Δ) = Ω is a convex domain, (ii) for any n ≥ 2 and any z ∈ Δ the following sharp estimate is valid: (n) h (z) n! h(n−1) (z) nz ≤ . h (z) − 1 − |z|2 n−1 h (z) (1 − |z|) (1 + |z|)

(7.21)

Proof. The condition (7.20) immediately implies h(t) = h(z) for z ∈ Δ, t ∈ Δ, t = z, and therefore the injectivity of the function h on Δ. Now, we ﬁx t ∈ Δ and we consider the function ϕ(z) = 2

h(z) − h(t) 1 − tz − 1, h (t) (z − t)(1 − |t|2 )

z ∈ Δ.

It is evident that ϕ(t) = 1 and that Re ϕ(z) > 0 for any z ∈ Δ. In a neighbourhood of the point t we have the Taylor expansion ∞ (n) h (t) t h(n−1) (t) (z − t)n−1 . ϕ(z) = 1 + 2 − (t)n! 2 )(n − 1)! (t) h (1 − |t| h n=2 Since 1/λΛ (ϕ(t)) = 2 Re ϕ(t) = 2, using the Schwarz-Pick lemma one easily gets h (t) 2 2t |ϕ (t)| = , t ∈ Δ, ≤ − 2 h (t) 1 − |t| 1 − |t|2 which is equivalent to the inequality 2 w − 1 + |t| ≤ 2|t| , 1 − |t|2 1 − |t|2 where w = 1+t

t ∈ Δ,

(7.22)

h (t) . h (t)

The condition (7.22) implies Re w > 0. Therefore (see for instance [128] and [146]), h is injective on Δ and Ω = h(Δ) is a convex domain. To get (ii) for n ≥ 3 we apply Ruscheweyh’s theorem 4.6 to get sharp estimates for the derivatives (n) ϕ(n−1) (t) h(n−1) (t) h (t) nt , t ∈ Δ, = 2 − (n − 1)! nh (t) 1 − |t|2 h (t) indicated in (ii). Equality in (7.21) at the point z = z0 ∈ Δ occurs if h(z) =

z z0 z = . 1 − z0 z/z0 z0 − z0 z

This completes the proof of Theorem 7.18.

126

Chapter 7. Related results

Remark 7.19. According to the above, the condition (7.20) is a new necessary and suﬃcient condition for h(Δ) to be convex that does not use the second derivative of h. It may be worthwhile to mention the conditions of this type that have been proved before. To our knowledge the ﬁrst one was zh (z) Re > 0, |t| < |z| < 1, h(z) − h(t) proved by Brickman in [46]. Sheil-Small ([151]) and Suﬀridge ([156]) proved the characterization 1 zh (z) t > , z ∈ Δ, t ∈ Δ, Re − h(z) − h(t) z−t 2 of convex funtions h. The third condition we want to cite seems to be the most famous. It has been proved by Ruscheweyh in [141] and it was used by Ruscheweyh and Sheil-Small in [144] to prove the P´ olya-Schoenberg conjecture. This one is as follows. z ζ − t h(z) − h(t) 1 ζ Re > , z ∈ Δ, t ∈ Δ, ζ ∈ Δ. − z − ζ z − t h(ζ) − h(t) z−ζ 2 One curious diﬀerence between these characterizations of convexity of h(Δ) and (7.20) seems to be that (7.20) contains nonanalytic terms.

Chapter 8

Some open problems 8.1

The Krzy˙z conjecture

As we have seen in Chapter 6 it is a very diﬃcult task to ﬁnd sharp punishing factors or substitutes for them in cases where multiply connected domains are involved. From this point of view, it seems natural that the diﬃculties become nearly insuperable, if one allows the points z0 ∈ Ω or f (z0 ) ∈ Π, or both to vary, and asks for the maximum. Nevertheless, there exists one problem of this type that has attracted researchers for many years because of the conjectured simple solution. This is the so-called Krzy˙z conjecture. It is concerned with functions f ∈ A(Δ, Δ ), where Δ = {z | 0 < |z| < 1} and the conjecture is that max

2 |f (n) (0)| | f ∈ A(Δ, Δ ) = n! e

for any n ∈ N (see [96]). It is obvious that the conjectured bound is attained for the functions 1 + zn f (z) = exp . 1 − zn The story of this conjecture began with a problem posed in [101], where the case n = 1 of the above is considered. It is remarkable that in the solutions found in 1934 the universal covering function of Δ plays the decisive role. After the formulation of the conjecture by Krzy˙z, there was a lot of eﬀort to solve this attractive problem. The cases n = 2 and n = 3 were solved in [84], for n = 4 see [161], [47], and [158]. The case n = 5 was solved recently in [145]. We cannot deny that our considerations for the case of multiply connected domains began with the hope to get a simple dependence of the upper bound of

128

Chapter 8. Some open problems

f (0) in this diﬃcult problem. We were able with the above theorems to get the simple cases n = 1 and n = 2, but not more. We may be allowed to conclude this section with a little observation which may prevent others from attacking this problem. In the above cited article, Szapiel poses the following problem: Do the estimates n (8.1) cj Aj (F ) ≤ n, j=1 hold for all schlicht functions F (z) = z +

∞

An (F )z n ,

n=2

whenever

n j−1 c ζ j ≤ 1 for all |ζ| ≤ 1? j=1

(8.2)

We ﬁnd that the answer is positive and that more is true. Theorem 8.1. Let cj satisfy the condition (8.2). Then the inequality (8.1) holds for any function F subordinate to a schlicht function F0 ∈ S. Proof. If we deﬁne P (z) =

n

cj ζ j ,

j=1

we see that the Schwarz lemma indicates that the condition (8.2) is equivalent to P (Δ) ⊂ Δ. Therefore the Sheil-Small theorem (see Chapter 2 and [152]) imply that the inequality |(P ∗ F )(eiθ )| ≤ n holds for all θ ∈ [0, 2π] and not only for all F ∈ S, but even for the case that F is subordinated to a schlicht function. This proves the theorem, a far-reaching generalization of (8.1).

8.2

The angle conjecture

As we have seen in Chapter 5, for f ∈ A(Ω, Hα ), α ∈ [1, 2], the inequalities n (λΩ (z0 )) 2n−1 |f (n) (z0 )| n+α−1 ≤ , n ≥ 2, z0 ∈ Ω, n n! α λΠ (f (z0 )) are valid as well for Ω = Δ as for Ω = H1 . It seems reasonable that these inequalities are true for any convex domain Ω. Concerning this conjecture, we will show now that it is true, if f (z0 ) lies on the line bisecting the angle Hα .

8.2. The angle conjecture

129

Theorem 8.2. Let Ω be a convex proper subdomain of C and Hα an angular domain with opening angle απ, α ∈ [1, 2]. Let further f ∈ A(Ω, Hα ), z0 ∈ Ω, and f (z0 ) lying on the bisector of Hα . Then for n ≥ 2 the inequalities n (λΩ (z0 )) |f (n) (z0 )| 2n−1 n+α−1 ≤ n n! α λΠ (f (z0 )) are valid. Proof. Let v be the vertex of the angular domain Hα and f (z0 ) − v = |f (z0 ) − v|eiϕ0 . Then the Riemann mapping function ΦHα ,f (z0 ) is given by ΦHα ,f (z0 ) (ζ) = v + (f (z0 ) − v)

1 + ze−iϕ0 1 − ze−iϕ0

α .

Hence, λHα (f (z0 )) =

1 . 2α|f (z0 ) − v|

Choosing an appropriate function ω : Δ → Δ such that ˜ Φ(ζ) = ΦHα ,f (z0 ) (ζω(ζ)) ˜ has the property Φ(Δ) = f (Ω), we get ˜ ◦ ΨΩ,z . f =Φ 0 If we now proceed as in the above proofs of the generalizations of the two Chua conjectures, we see that for the proof of the present assertion we have to show that n 2n−1 n+α−1 , (8.3) ck (α)An,k ≤ n α k=1

where the function represented by the series ∞

ck (α)ζ k

k=1

is subordinated to the function α ∞ 1+ζ 1 −1 = hk (α)ζ k 2α 1−ζ

(8.4)

k=1

and the An,k represent the coeﬃcients of the powers of inverses of convex functions as above.

130

Chapter 8. Some open problems

Analogous to the proofs mentioned above we want to show now that for functions ∞ ω(z) = dτ ζ k , τ =0

mapping the unit disc into the closed unit disc, the inequalities p−1 (p − τ )cp−τ (α)dτ ≤ p hp (α), p ∈ N,

(8.5)

τ =0

are valid. To this end, we consider, as in the proof of the 2n−1 conjecture, the region of variability of the linear functional Lω (g) =

p−1

(p − τ )cp−τ (α)dτ ,

τ =0

where ω is ﬁxed and g is varying in the family of functions subordinated under (8.4). According to [44], [45], and [146] these functions g have a representation α 2π 1 1 + eit ζ g(ζ) = − 1 d μ(t), 2α 0 1 − eit ζ with a probability measure μ : [0, 2π] → R. Hence, it is suﬃcient to prove the inequality (8.5) for cτ (α) = eiτ hτ (α). Further, the rotational invariance of the family of unimodular bounded functions indicates that we only need to show that p−1 (8.6) (p − τ )hp−τ (α)dτ ≤ p hp (α), p ∈ N. τ =0

We want to thank at this point St. Ruscheweyh for pointing out to us the idea of this proof. His idea was to consider the polynomial Q(ζ) =

p−1 (p − τ )hp−τ (α) τ =0

p hp (α)

ζk

and to show that the function Q ∗ ω is a unimodular bounded function. According to [152] (see especially the theorems (2.6) and (3.2) of this article), this can be reduced to the proof of Re(Q(ζ)) >

1 , 2

ζ ∈ Δ.

By the application of a theorem due to Rogosinski in [138], to achieve this aim it is suﬃcient to prove that the coeﬃcients of Q form a monotonic decreasing convex sequence. This means that (k + 1)hk+1 (α) − k hk (α) ≥ 0,

k = 1, . . . , p − 1,

8.3. The generalized Goodman conjecture

131

(k + 2)hk+2 (α) − 2(k + 1)hk+1 (α) + k hk (α) ≥ 0,

k = 1, . . . , p − 2,

and 2h2 (α) − 2h1 (α) ≥ 0. To prove these inequalities we have to insure that the coeﬃcients of 1 d 2α dζ

1+ζ 1−ζ

α =

(1 + ζ)α−1 (1 − ζ)α+1

form a monotonic increasing convex sequence. This is easily seen by recognizing that 1+ζ (1 + ζ)α−1 (1 − ζ)2 = e(α−1) log( 1−ζ ) α+1 (1 − ζ) has nonnegative coeﬃcients. A comparison with the proofs of the generalizations of the two conjectures of Chua above reveals that we can get the inequality (8.3), setting again (ω(ζ))σ =

∞

dj,σ ζ j ,

ζ∈Δ

j=0

and using (8.6) in the chain n n−1 1 n n−1 ck (α)An,k ≤ |cn (α)| + (n − j)cn−j (α)dj−σ,σ σ n σ=1 j=σ k=1 n−1 1 n (n − σ)hn−σ (α) ≤ hn (α) + σ n σ=1 n 2n−1 n+α−1 n−1 . hk (α) = = n n−k α k=1

The last identity has been shown in Chapter 5.

8.3

The generalized Goodman conjecture

In many of the results of the present book it could be proved that the bounds for the moduli of the Taylor coeﬃcients of the functions in a certain family of functions holomorphic in Δ are the right bounds for the related coeﬃcients of the functions subordinated to the above ones. Therefore, we dare to formulate a challenging conjecture that generalizes as well the Rogosinski conjecture, which naturally is a theorem nowadays, as Jenkins’ theorem 2.7 on the Goodman conjecture.

132

Chapter 8. Some open problems

The generalized Goodman conjecture. Let f be injective and meromorphic in Δ. Further let f be normalized by f (0) = f (0) − 1 = 0 and have the pole at the point p ∈ (0, 1). If ω : Δ :→ Δ, then we conjecture that for the expansion g(z) = f (zω(z)) =

∞

an (g)z n ,

n=1

valid in some neighbourhood of the origin, the inequalities |an (g)| ≤

n−1

1 pn−1

p2k ,

n ≥ 2,

k=0

are valid. This relation between f and g will be denoted by g ≺ f as in the holomorphic case. To support this conjecture we prove that it is valid in some special cases. The ﬁrst one deals with little values of p. Theorem 8.3. The generalized Goodman conjecture is valid for n = 2, p ∈ (0, 1), for n = 3, p ∈ (0, 0.7), and for n ≥ 4, p ∈ (0, 1/(2n − 2)). ∞ an z n , |z| < p meromorphic and univalent in D with Proof. For f (z) = z + n=2 ∞ pole p ∈ (0, 1) and ω(z) = n=0 cn z n , ω : D → D, we have to ﬁnd the least upper bound for |c1 + a2 (c0 )2 | in the case n = 2 and for |c2 + 2c0 c1 a2 + (c0 )3 a3 | in the case n = 3. In the ﬁrst the triangle inequality, Jenkins’ inequality (see [86]), and the Schur algorithm (see[148]) deliver as an upper bound 1 − |c0 |2 + |c0 |2

1 + p2 1 + p2 ≤ . p p

In the second case, we use c2 |c1 |2 c0 (c1 )2 ≤ 1− + 1 − |c0 |2 2 2 (1 − |c0 | ) (1 − |c0 |2 )2 (see[148]). Using the triangle inequality and the abbreviations x = |c0 |, we get

y=

|c1 | (1 − |c0 |2 )

|c2 | ≤ (1 − x2 )(xy 2 + 1 − x2 ).

8.3. The generalized Goodman conjecture

133

Therefore, we have to maximize (1 − x2 )(xy 2 + 1 − x2 ) + 2xy(1 − x2 )

1 + p2 + p 4 1 + p2 = F (x, y; p) + x3 p p2

for (x, y) ∈ [0, 1] × [0, 1]. Since ∂F 1 + p2 2 ≥ (≤) 0 = 2(1 − x ) y(x − 1) + x ∂y p for y ≤ (≥) and

x 1 + p2 , 1−x p

x 1 + p2 ≤1 1−x p

for

p , 1 + p + p2 in these cases we have to check the local maximum of F in x≤

y=

x 1 + p2 . 1−x p

Inserting this delivers a monotonic increasing function of x. So we have to check that this function at p x= 1 + p + p2 is less than (1 + p2 + p4 )/p2 . This is the case for p ∈ (0, 1). For p x≥ 1 + p + p2 we have to consider F (x, 1, p) = x

1 − 2p − 2p3 + p4 2 + p + 2p2 + x3 . p p

For those p for which the derivative of this function with respect to x is positive for x ≤ 1, the proof is complete. This is the case for ! √ 2 + 19 p≤ = 0.7088023 . . . . 3 For n ≥ 4 we consider (zω(z))k =

∞ j=k

cj,k z j

134

Chapter 8. Some open problems

and we see that ck,k = (c0 )k and therefore |cj,k | ≤ 1 − |c0 |2k ,

j ≥ k.

Hence, ⎛ ⎞ n−1 n k−1 n−1 p2j p2j ⎝ ⎠ + |c0 |n |an (g)| = cn,k ak ≤ . 1 − |c0 |2k pk−1 pn−1 k=1

j=0

k=1

j=0

n−1 We want to prove that this is less than or equal to j=0 (p2j /pn−1 ). This inequality is equivalent to ⎛ ⎞ k−1 n−1 n−1 n−1 2k−1 p2j p2j j ⎝ ⎠≤ |c0 |j |c | . 0 k−1 p pn−1 j=0 j=0 j= j=0 k=1

We use the rough estimate 2k−1

|c0 |j ≤ 2

j=0

n−1

|c0 |j ,

1 ≤ k ≤ n − 1,

j=0

and we see by some elementary calculations that 2

n−1 k−1 k=1 j=0

n−1 p2j p2j ≤ pk−1 pn−1 j=0

for p ∈ (0, 1/(2n − 2)).

The second theorem is concerned with the condition that f (Δ) is concave, i.e., C\f (Δ) is a convex compact set. The family of functions with these properties will be denoted by Co(p) as in Chapter 5, Section 1, and the family of functions ω : Δ → Δ by B. Theorem 8.4 (see [26]). Let p ∈ (0, 1). If f ∈ Co(p) and g ≺ f, then |an (g)| ≤

1 pn−1

n−1

p2k ,

n ∈ N.

(8.7)

k=0

For the proof of Theorem 8.4 we use a representation formula for the functions in the class Co(p) that was derived in [167] and [23] as a simple consequence of Theorem 4 in [117]. Theorem 8.5. Let p ∈ (0, 1). For any f ∈ Co(p) there exists a function w1 ∈ B such that p 1 − 1+p 2 (1 + w1 (z))z z f (z) = , z ∈ Δ. (8.8) z 1− p 1 − zp

8.3. The generalized Goodman conjecture

135

The ﬁrst step in the proof of Theorem 8.4 concerns the ﬁrst factor in the representation (8.8) that we will denote by f1 , f1 (z) =

z , 1 − pz

z ∈ Δ.

We prove that, for this member of Co(p), one may replace the sum in Theorem 8.4 by 1. Theorem 8.6. Let p ∈ (0, 1). If g ≺ f1 , then |an (g)| ≤

1 pn−1 ,

n ∈ N.

Proof. Let w ∈ B such that g(z) = f1 (zw(z)), z ∈ Δ. This implies zg(z) , z ∈ Δ. g(z) = w(z) z + p

(8.9)

Now, to prove Theorem 8.6, we use a method due to Clunie (see [57]) and a generalization of a theorem of Robertson (see [137]) proved in [30], Lemma 2.1. This is the meromorphic version of the Rogosinski lemma (see our Theorem 2.4). The application of this lemma to the identity (8.9) yields that for any n ∈ N, the inequality n n 2 |ak−1 (g)| 2 |ak (g)| ≤ 1 + p2 k=1

k=2

is valid. The assertion of Theorem 8.6 follows by mathematical induction using 1/p > 1. Next, we consider the second factor in the representation (8.8) that we will denote by f2 and we prove the following inequalities. Theorem 8.7. Let p ∈ (0, 1) and for w ∈ B, g2 (z) = f2 (zw(z)) = 1 +

∞

Bk z k ,

z ∈ D.

k=1

Then for any n ∈ N the inequality n

2

|Bk | ≤

k=1

n

p2k

(8.10)

k=1

is valid. Proof. Obviously, f2 (z) = 1 + Since

p2 −w1 (z) 1+p2

pz

1 − zp

2 p − w1 (z) 1 + p2 ≤ 1,

,

z ∈ Δ.

z ∈ Δ,

136

Chapter 8. Some open problems

there exists a function w2 ∈ B such that w2 (z) p z , 1 − zp

f2 (z) = 1 +

z ∈ Δ.

Let w3 (z) = w2 (zw(z)), z ∈ Δ. Then w3 ∈ B. Hence, the function g2 − 1, g2 (z) − 1 =

w3 (z) p zw(z) , 1 − zw(z)p

z ∈ Δ,

is quasi-subordinate to the function h, h(z) =

pz , 1 − zp

z ∈ Δ.

The notation of quasi-subordination was deﬁned by Robertson (see [137]). He proved that the above relation implies in our case the inequality (7.9) (compare also [128], Theorem 2.2). This completes the proof of Theorem 8.5. After these preparations, it is easy to prove Theorem 8.4 as follows. Proof of Theorem 8.4. If g ≺ f and f ∈ Co(p), we conclude from Theorem 8.5 that there exists a function w ∈ B such that g(z) = f1 (zw(z)) f2 (zw(z)). Let f1 (zw(z)) =

∞

(8.11)

Ak z k ,

k=1

where this expansion is valid in some neighbourhood of the origin. From (8.11) we get for n ∈ N, n−1 Ak Bn−k . an (g) = An + k=1

The Cauchy-Schwarz inequality shows that, for any n ∈ N, the inequality |an (g)| ≤

n

1/2 2

|Ak |

1+

k=1

n−1

1/2 2

|Bk |

k=1

is valid. If we use Theorems 8.6 and 8.7 to estimate the right-hand side of this inequality, we get |an (g)| ≤

n−1 k=0

p−2k

1/2 n−1 k=0

This completes the proof of Theorem 8.4.

1/2 p2k

=

1 pn−1

n−1

p2k .

k=0

8.3. The generalized Goodman conjecture

137

It is obvious that the estimates in Theorems 8.4 and 8.6 are sharp. In Theorem 8.6, equality is attained if g(z) =

cz , 1 − cz p

|c| = 1.

In Theorem 8.4 equality occurs for g(z) =

cz , 1 − cz (1 − zp) p

|c| = 1.

These are conformal maps of Δ onto % $ −p −p , C\ (1 − p)2 (1 + p)2 (compare [15] and [166]). The third theorem proves the generalized Goodman conjecture for univalent functions with real coeﬃcients. They belong to the class of functionss typicallyreal and meromorphic, introduced by Goodman in [72]. The deﬁning relation for this class is, as in the holomorphic case, (f (z) (z) ≥ 0,

z ∈ Δ.

It is easily seen that this implies that in our case the residuum of f at the point p is negative. Goodman proved a representation theorem that implies the following application to univalent meromorphic and typically-real functions normalized as usual. Theorem 8.8 (see [72] Theorem 9). Let f be a function univalent meromorphic and typically-real in Δ having the residuum −m < 0 at the point p. There exists a function t holomorphic and typically-real in Δ and a nonnegative constant M such that 1 M +m −1 =1 p and

f (z) = M t(z) + m

1 −1 p

z , (1 − zp)(1 − z/p)

z ∈ Δ.

Using this representation it is easy to prove (8.7) for univalent meromorphic and typically-real functions. Theorem 8.9. Let f be univalent meromorphic and typically-real in Δ. If f has its pole at the point p ∈ (0, 1) and g ≺ f , then (8.7) holds.

138

Chapter 8. Some open problems

Proof. According to a theorem proved by Robertson in [135], for any function typically-real and holomorphic in Δ there exists a probability measure μ on [0, π] such that π z dμ(θ) t(z) = , z ∈ Δ. 1 − 2z cos θ + z 2 0 Since the kernel functions belong to the class S, the truth of the Rogosinski conjecture for univalent functions and a convex hull argument imply that for g1 ≺ t the inequalities |an (g1 )| ≤ n, n ≥ 2 hold. From Theorem 8.4 above we know that for g2 (z) ≺

z , (1 − zp)(1 − z/p)

the generalized Goodman conjecture holds. Hence, |an (g2 )| ≤

1 pn−1

n−1

p2k .

k=0

From the identity 1 pn−1

n−1

p2k =

k=0

⎧ u ⎨

1 + p2k−1 u 1 k=1 p2k

k=1

⎩ 1+

we conclude that

n−1

1 pn−1

p2k−1 , if n = 2u, + p2k , if n = 2u + 1

p2k ≥ n.

k=0

If we write a function g ≺ f in the form 1 g(z) = M g1 (z) + m − 1 g2 (z), p

z ∈ Δ,

we see that |an (g)| ≤ M |an (g1 )| + m

n−1 1 2k 1 p . − 1 |an (g2 )| ≤ n−1 p p k=0

This proves Theorem 8.9.

Using the bound found by Jenkins and well-known subordination techniques, it is not diﬃcult to ﬁnd upper bounds for |an (g)|. We give two examples ⎛ ⎞ k−1 n 1 ⎝ p2j ⎠ |an (g)| ≤ pk−1 j=0 k=1

8.4. Bloch and several variable problems and

139

⎛ ⎞2 ⎞1/2 k−1 n ⎟ ⎜ ⎝ 1 p2j ⎠ ⎠ . |an (g)| ≤ ⎝ pk−1 j=0 ⎛

k=1

It seems that these estimates reﬂect approximately the right asymptotics of the least upper bounds for p → 0, but it is evident that they are bad for p → 1. An upper bound that ﬁts better for the neighbourhood of p = 1 can be found if one applies the validity of the Rogosinski conjecture to the function fp (z) = f (pz)/p that belongs to the class S. By this procedure we get |an (g)| ≤

n . pn−1

We add a last inequality for the coeﬃcients an (g) that follows immediately from the generalized Goluzin-Rogosinski theorem 2.5 using Jenkins’ theorem 2.7. Taking λk = we get

n k=1

pn−1 k−1 2j j=0 p

|ak (g)|2 λk ≤

2

n

,

k ∈ N,

|ak |2 λk ≤ n.

k=1

Obviously, this inequality is sharp for the function for which Jenkins’ theorem is sharp.

8.4

Bloch and several variable problems

We want to avoid the impression that the above considerations are the only possibility to generalize the Schwarz-Pick lemma to higher derivatives. Therefore we mention some questions that arose when we studied the work of colleagues on generalized Schwarz-Pick type estimates (see [38], [112], and [113]). For example, it has been proved in [112] that for f ∈ A(Δ, Δ), α, β > 0, the implications |f (z)|(1 − |z|2 )β < ∞ 2 α z∈Δ (1 − |f (z)| ) sup

|f (n) (z)|(1 − |z|2 )β+n−1 < ∞, (1 − |f (z)|2 )α z∈Δ

=⇒ sup

n ≥ 2,

are valid. On one hand it seems natural to ask whether one can compute an explicit relation between these two suprema. On the other hand this result together with

140

Chapter 8. Some open problems

our theorems on punishing factors suggests the question for which pairs (Ω, Π) of domains and α, β > 0 implications of the form |f (z)|R(Ω, z)β < ∞ α z∈Ω R(Π, f (z)) sup

|f (n) (z)|R(Ω, z)β+n−1 < ∞, R(Π, f (z))α z∈Ω

=⇒ sup

n ≥ 2,

are valid. The results of MacCluer, Stroethoﬀ, and Zhao in [113] may give rise to analogous problems even for functions of several variables. Since it would be lengthy to cite their deﬁnitions and theorems, we prefer to conclude with open questions that are simpler to formulate and that originated in the following result of B´en´eteau, Dahlner, and Khavinson (see [38]). They proved that for an analytic function f : Δn → Δ,

(z1 , . . . , zn ) → f (z1 , . . . , zn )

and any multiindex α = (α1 , . . . , αn ) ∈ (N ∪ {0})n the inequality sup (z1 ,...,zn )∈Δn

α ,n |Dα f (z1 , . . . , zn )| k=1 1 − |zk |2 k < ∞ 1 − |f (z)|2

holds, where Dα f (z1 , . . . , zn ) is the usual abbreviation for the derivative of order α. Again, one may ask for estimates for this supremum or consider generalizations of this result to other domains instead of Δ. For new results on similar generalizations of Schwarz-Pick estimates compare [5], [6], [68] and [92].

8.5

On sums of inverse coeﬃcients

As we have seen frequently above, a decisive part in the computation of punishing factors is played by inverse coeﬃcients An,k of the kth powers of functions injective in the unit disc. L¨ owner’s Theorem 2.7 and its generalizations provided us with satisfactory solutions for these problems. If subclasses of S are considered, the situation is far less nice. The most famous of such problems is the question for bounds in the convex case. For the function z f (z) = , |c| = 1, (8.12) 1 − cz the identities n−1 |An,k | = k−1

8.5. On sums of inverse coeﬃcients

141

hold for 1 ≤ k ≤ n ∈ N. R. L. Libera and E. J. Zlotkiewicz proved in [104] that |An,1 | ≤ 1,

n = 2, 3, 4, 5, 6, 7

for the inverse coeﬃcients of convex functions. This result implies the validity of the estimate n−1 |An,k | ≤ (8.13) k−1 for n between 2 and 7. On the other hand W. E. Kirwan and G. Schober got by explicit computations that M10 = max {|A10,1 |} > 1.2, where the maximum is taken over all convex functions (see [90]). A detailed discussion of Mn = max {|An,1 |} may be found in [50] and [58]. J.T.P. Campschroer showed in [50] that Mn = O 2n n−3 . It may be interesting that for k big compared with n the inequalities (8.13) hold true. Theorem 8.10. For the inverse coeﬃcients An,k of convex functions, n ≥ 4, and n/2 − 1 ≤ k ≤ n, the inequalities (8.13) are valid. Proof. As above, we use the Schur-Jabotinsky theorem and, in addition, the fact that a convex function f is starlike of order 1/2. This implies that the function h deﬁned by 2 f (z) h(z) = z z is a starlike function. Hence, we may consider the Taylor coeﬃcients of n/2 ∞ z = bl,n/2 (h)z n h(z) n=0

where h is starlike. But from [140] we know that the modulus bl,n/2 (h) of the Taylor coeﬃcients of these functions is bounded from above by bl,n/2 (k), where k is the Koebe function or one of its rotations and l ≤ n/2 + 1. In most cases, the Koebe functions are the only extremal functions. This proves our theorem. But in our computations of punishing factors, weighted sums of the An,k are the most important elements. If one looks at punishing factors for convex pairs (see section 4.5), one may recognize that the proof of Theorem 4.12 would be very short, if the following conjecture (compare [22]) could be proved:

142

Chapter 8. Some open problems

Conjecture. For the inverse coeﬃcients of convex functions and for any n ≥ 2, the inequality n |An,k | ≤ 2n−1 k=1

is valid. The validity of this conjecture for 2 ≤ n ≤ 7 follows immediately from the above. A sharp estimate for the sum of squares can be derived from section 4.5 and an application of Theorem 2.5 on subordinate functions. Actually, one easily gets n

2

|An,k | =

n−1 μ=0

k=1

≤

2 n − μ a μ,n n

n−1 μ=0

n−μ n

2

n μ

2

=

2(n − 1) n−1

.

Here, equality is attained for the functions (8.12). If one uses this inequality to estimate the sum of moduli of the inverse coefﬁcients by the Cauchy inequality and the Stirling formula, one gets immediately n k=1

|An,k | ≤

√

n

2(n − 1) n−1

1/2

≤ 2n−1 n1/4 .

If one wants to compute punishing factors for functions holomorphic on βaccessible domains, it would be necessary to know something of the inverse coeﬃcients of functions close-to-convex of order β or of weighted sums of the moduli of these inverse coeﬃcients (compare section 7.1 and section 5.6 on lower bounds). The problem of such inverse coeﬃcients is addressed in [90], but those results do concern only early coeﬃcients. It seems natural to us to conjecture inequalities for the sums mentioned above to get new conjectures on punishing factors.

Bibliography [1] S. Agard, Distortion theorems for quasiconformal mappings. Ann. Acad. Sci. Fenn. Ser. AI 413 (1968), 12p. [2] D. Aharonov and S. Friedland, On an inequality connected with the coeﬃcient conjecture for functions of bounded boundary rotation. Ann. Acad. Sci. Fenn. Ser. AI 524 (1972), 14p. [3] L. V. Ahlfors, Conformal invariants, Topics in Geometric Function Theory. McGraw-Hill, New York, 1973. [4] A. Ancona, On strong barriers and an inequality of Hardy for domains in Rn . J. London Math. Soc.(2) 34 (1986), no.2, 274–290. [5] J. M. Anderson, M. A. Dritschel, and J. Rovnyak, Schwarz-Pick inequalities for the Schur-Agler class on the polydisk and unit ball. Comp. Methods and Function Theory 8 (2008), 339–361. [6] J. M. Anderson and J. Rovnyak, On generalized Schwarz-Pick estimates. Mathematika 53 (2006), 161–168. [7] J. M. Anderson, J. G. Clunie, and Ch. Pommerenke, On Bloch functions and normal functions. J. Reine Angew. Math. 270 (1974), 12–37. [8] J. Arazy, S. Fisher, and J. Peetre, M¨ obius Invariant Spaces of Analytic Functions. In: Lecture Notes in Math., Springer-Verlag, Nr. 1275 (1987), 10–22. [9] K. Astala, J. L. Fern´ andez, and S. Rohde, Quasilines and the Hayman-Wu theorem. Indiana Univ. Math. J. 42 (1993), no.4, 1077–1100. [10] F. G. Avhadiev and L. A. Aksent’ev, The main results on suﬃcient condition for an analytic function to be schlicht. Russ. Math. Surv, 30 (1975), no.4, 1–63 (translated from Usp. Mat. Nauk, 1975, t. 30, no. 4(184), 3–63). [11] F. G. Avkhadiev, Solution of the Generalized Saint Venant Problem. Sb. Math. 189 (1998), 1739–1748 (translated from Mat. Sb. 189(12) (1998), 3– 12). [12] F. G. Avkhadiev, Hardy Type Inequalities in Higher Dimensions with Explicit Estimate of Constants. Lobachevskii J. Math. 21 (2006), 3–31 (electronic).

144

Bibliography

[13] F. G. Avkhadiev, Hardy-type inequalities on planar and spatial open sets. Proceedings of the Steklov Institute of Mathematics 255 (2006), no.1, 2–12 (translated from Trudy Matem. Inst. V.A. Steklova, 2006, v.255, 8–18). [14] F. G. Avkhadiev and K.-J. Wirths, Estimates of derivatives in an αinvariant set of analytic functions. Analysis 19 (1999),19-27. [15] F. G. Avkhadiev and K.-J. Wirths, Convex holes produce lower bounds for coeﬃcients. Complex Variables 47 (2002), 553–563. [16] F. G. Avkhadiev and K.-J. Wirths, Schwarz-Pick inequalities for derivatives of arbitrary order. Constr. Approx. 19 (2003), 265–277. [17] F. G. Avkhadiev and K.-J. Wirths, Punishing factors for angles. Comp. Methods and Function Theory 3 (2003), 127–141. [18] F. G. Avkhadiev and K.-J. Wirths, Poles near the origin produce lower bounds for coeﬃcients of meromorphic univalent functions. Michigan Math. J. 52 (2004), 119–130. [19] F. G. Avkhadiev and K.-J. Wirths, Schwarz-Pick inequalities for hyperbolic domains in the extended plane. Geom. Dedicata 106 (2004), 1–10. [20] F. G. Avkhadiev and K.-J. Wirths, The conformal radius as a function and its gradient image. Isr. J. Math. 145 (2005), 349–374. [21] F. G. Avkhadiev and K.-J. Wirths, Punishing factors for ﬁnitely connected domains. Monatshefte f. Math. 147 (2006), 103–115. [22] F. G. Avkhadiev and K.-J. Wirths, Sharp bounds for sums of coeﬃcients of inverses of convex functions. Comp. Methods and Function Theory 7 (2007), 105–109. [23] F. G. Avkhadiev and K.-J. Wirths, A proof of the Livingston conjecture. Forum Math. 19 (2007), 149–157. [24] F. G. Avkhadiev and K.-J. Wirths, The punishing factors for convex pairs are 2n−1 . Revista Mat. Iberoamericana 23 (2007), 847–860. [25] F. G. Avkhadiev and K.-J. Wirths, Punishing factors and Chua’s conjecture. Bull. Belg. Math. Soc., Simon Stevin 14 (2007), 333–340. [26] F. G. Avkhadiev and K.-J. Wirths, Subordination under concave univalent functions. Complex Variables and Elliptic Equations 52 (2007), 299–305. [27] F. G. Avkhadiev and K.-J. Wirths, A Theorem of Teichm¨ uller, Uniformly Perfect Sets and Punishing factors. Preprint (Tech. Univ. Braunschweig, 2005) http://www.iaa.tu-bs.de/wirths/preprints/teich.ps [28] F. G. Avkhadiev and K.-J. Wirths, Estimates of the derivatives of meromorphic maps from convex domains into concave domains. Comp. Methods and Function Theory 8 (2008), 107–119.

Bibliography

145

[29] F. G. Avkhadiev and K.-J. Wirths, Local Taylor coeﬃcients and subordination. Preprint (Tech. Univ. Braunschweig, 2005) http://www.iaa.tu-bs.de/wirths/preprints/local.ps [30] F. G. Avkhadiev, Ch. Pommerenke, and K.-J. Wirths, On the coeﬃcients of concave univalent functions. Math. Nachr. 271 (2004), 3–9. [31] F. G. Avkhadiev, Ch. Pommerenke, and K.-J. Wirths, Sharp inequalities for the coeﬃcients of concave schlicht functions. Comment. Math. Helv. 81 (2006), 801–807. [32] A. Baernstein, II, Integral means, univalent functions and circular symmerization. Acta Math. 133 (1974), 139–169. [33] A. Baernstein, II and G. Schober, Estimates for inverse coeﬃcients of univalent functions from integral means. Israel J. Math. 36 (1980), 75-82. [34] C. Bandle and M. Flucher, Harmonic radius and concentration of energy, n+2 hyperbolic radius and Liouville’s equation U = eU and U = U n−2 . SIAM Review 38 (1996), 191–238. [35] H. Bateman, The k-function, a particular case of the conﬂuent hypergeometric function. Trans. Amer. Math. Soc. 33 (1931), 817–831. [36] A. F. Beardon and F. W. Gehring, Schwarzian derivatives, the Poincar´e metric and the kernel function. Comment. Math. Helv. 55 (1980), 50–64. [37] A. F. Beardon and Ch. Pommerenke, The Poincar´e metric of plane domains. J. London Math. Soc. (2) 18 (1978), 475–483. [38] C. B´en´eteau, A. Dahlner, and D. Khavinson, Remarks on the Bohr phenomenon. Comp. Methods and Function Theory 4 (2004), 1–19. [39] A. Bermant, Dilatation d’ une fonction modulaire et probl`emes de recouvrement (in Russian, French summary). Matematicheskii Sbornik 15(57) (1944), No. 2, 285–318. ¨ [40] L. Bieberbach, Uber die Koeﬃzienten derjenigen Potenzreihen, welche eine schlichte Abbildung des Einheitskreises vermitteln. Sitz. Ber. Preuss. Akad. Wiss., (1916), 940–955. [41] M. Biernacki, Sur la repr´esentation conforme des domaines lin´earement accessibles. Prace Mat. Fiz. 44 (1936), 293–314. ´ Borel, Sur les z´eros des fonctions enti`eres. Acta Math. 20 (1896 - 97), [42] E. 357-396. [43] L. de Branges, A proof of the Bieberbach conjecture. Acta Math. 154 (1985), 137-152. [44] D. A. Brannan, On coeﬃcient problems for certain power series. Proc. of the Symposium on Complex Analysis (Univ. Kent, Canterbury,1973, London Mat. Soc. Lecture Note Series, Cambridge Univ. Press) 12 (1974), 17–27.

146

Bibliography

[45] D. A. Brannan, J. G. Clunie, and W. E. Kirwan, On the coeﬃcient problem for functions of bounded boundary rotation. Ann. Acad. Sci. Fenn. AI 523 (1973), 18p. [46] L. Brickman, Subordinate families of analytic functions. Illinois J. Math. 15 (1971), 241–248. [47] J. E. Brown, Iterations of functions subordinate to schlicht functions. Complex Variables 9 (1987), 143–152. [48] R. B. Burckel, An Introduction to Classical Complex Analysis, Vol.1. Birkh¨ auser Verlag, Basel und Stuttgart, 1979. [49] L. A. Caﬀarelli and A. Friedman, Convexity of solutions of semilinear elliptic equations. Duke Math. J. 52 (1985), 15–44. [50] J. T. P. Campschroer, Growth of the coeﬃcients of the inverse of a convex function. Indag. Math. 46 (1984), 1–6. ¨ [51] C. Carath´eodory, Uber den Variabilit¨ atsbereich der Koeﬃzienten von Potenzreihen, die gegebene Werte nicht annehmen. Math. Ann. 64 (1907), 95–115. ¨ [52] C. Carath´eodory, Uber den Variabilit¨ atsbereich der Fourierschen Konstanten von positiven harmonischen Funktionen. Rend. Cir. Mat. Palermo 32 (1911), 193–217. [53] C. Carath´eodory, Untersuchungen u ¨ber die konformen Abbildungen von festen und ver¨ anderlichen Gebieten. Mat. Ann. 52:1 (1912), 107–144. [54] C. Carath´eodory, Theory of functions of a complex variable. Chelsea Publ. Comp., New York, 1960. [55] L. Carleson and T. W. Gamelin, Complex dynamics. Springer, New York, 1993. [56] K. S. Chua, Derivatives of univalent functions and the hyperbolic metric. Rocky Mountain J. Math. 26 (1996), 63–75. [57] J. Clunie, On meromorphic schlicht functions. J. London Math. Soc. 34 (1958), 215–216. [58] J. G. Clunie, Inverse coeﬃcients for convex univalent functions. Journal d’An. Math. 36 (1979), 31–35. [59] P. L. Duren, Theory of H p spaces. Academic Press, New York, 1970. [60] P. L. Duren, Univalent functions. Springer, New York, 1980. ¨ [61] L. Fej´er, Uber gewisse durch die Fouriersche und Laplacesche Reihe deﬁnierten Mittelkurven und Mittelﬂ¨ achen. Palermo Rend. 38 (1914), 79– 97.

Bibliography

147

[62] J. L. Fern´ andez, Domains with strong barrier. Rev. Math. Iberoamericana 5 (1989), no.1-2, 47–65. [63] J. L. Fern´ andez and J. M. Rodr´ıguez, The Exponent of Convergence of Riemann Surfaces. Bass Riemann Surfaces. Ann. Acad. Sci. Fenn., Ser. A.I: Math. 15 (1990), 165–182. [64] C. H. FitzGerald and Ch. Pommerenke, The de Branges theorem on univalent functions. Trans. Amer. Math. Soc. 290 (1985) 683–690. [65] F. P. Gardiner and N. Lakic, Comparing Poincar´e densities. Ann. Math. 154 (2001), 115–127. [66] J. B. Garnett, Bounded analytic functions. Academic Press, Inc. New York, London, 1981. [67] J. B. Garnett and D. E. Marshall, Harmonic measure. Cambridge Univ. Press, Cambridge, 2005. [68] P. Ghatage and D. Zheng, Hyperbolic derivatives and generalized SchwarzPick estimates. Proc. Amer. Math. Soc. 132 (2004), 3309–3318. [69] G. M. Goluzin, On majorant of subordinate analytic functions I. Mat. Sb. 29(71) (1951), 209–224. [70] G. M. Goluzin, Geometric theory of functions of a complex variable. AMS, Providence, 1969. [71] S. Gong, The Bieberbach Conjecture. AMS/IP studies in advanced mathematics 12, AMS, Internatinal Press, 1999. [72] A. W. Goodman, Functions typically-real and meromorphic in the unit circle. Trans. Amer. Math. Soc. 81 (1956), 92–105. [73] A. W. Goodman, Univalent functions. Mariner Publ. Comp., Tampa, FL, 1983. [74] B. Gustafsson, On the convexity of a solution of Liouville’s equation. Duke Math. J. 60 (1990), 303–311. [75] D. J. Hallenbeck and T. H. MacGregor, Linear Problems and Convexity Techniques in Geometric Function Theory. Pitman Press, Boston, 1984. [76] R. Harmelin, Hyperbolic metric, curvature of geodesics and hyperbolic discs in hyperbolic domains. Israel J. Math. 70 (1990), 111–126. [77] R. Harmelin and D. Minda , Quasi-invariant domain constants. Israel J. Math 77 (1992), 115–127. [78] W. K. Hayman, Multivalent functions. Cambridge Univ. Press, Cambridge, 1958.

148

Bibliography

[79] W. K. Hayman and J. M. G. Wu, Level sets of univalent functions. Comment. Math. Helv. 56 (1981), no. 3, 366–403. [80] J. A. Hempel, The Poincar´e metric on the twice punctured plane and the theorems of Landau and Schottky. J. London Math. Soc. (2) 20(1979), 435– 445. [81] P. Henrici, Applied Computational Complex Analysis. Vol.1, Wiley, New York, 1974. [82] D. A. Herron, X. Liu, and D. Minda, Ring domains with separating circles or separating annuli. J. Analyse Math. 53 (1989), 233–252. [83] A. Herzig, Die Winkelderivierte und das Poisson-Stieltjes-Integral. Math. Z. 46 (1940), 129–156. [84] J. A. Hummel, St. Scheinberg, and L. Zalcman, A coeﬃcient problem for bounded nonvanishing functions. J. Anal. Math. 31 (1977), 169–190. [85] Z. J. Jakubowski, On the upper bound of the functional f (n) (z) (n = 2, 3, . . .) in some classes of univalent functions. Comment. Math. Prace Mat. 17 (1973), 65–69. [86] J. A. Jenkins, On a conjecture of Goodman concerning meromorphic univalent functions. Michigan Math. J. 9 (1962), 25-27. [87] J. A. Jenkins, On explicit bounds in Landau’s theorem II. Can. J. Math. 33 (1981), 559–562. [88] V. Jørgensen, On an inequality for the hyperbolic measure and its application in the theory of functions. Math. Scand. 4 (1956), 113–124. [89] S.-A. Kim and C. D. Minda, The hyperbolic and quasihyperbolic metrics in convex regions. J. Anal. 1 (1993), 109–118. [90] W. E. Kirwan and G. Schober, Inverse coeﬃcients for functions of bounded boundary rotation. Journal d’An. Math. 36 (1979), 167–178. [91] R. Klouth and K.-J. Wirths, Two new extremal properties of the Koebefunction. Proc. Amer. Math. Soc. 80 (1980), 594–596. [92] G. Knese, A Schwarz lemma on the polydisk. Proc. Amer. Math. Soc. 135 (2007), 2759–2768. [93] W. Koepf, On close-to-convex functions and linearly accessible domains. Complex Var. 11 (1989), 269–279. [94] W. Koepf and D. Schmersau, Bounded nonvanishing functions and Bateman functions. Compl. Var. 25 (1994), 237–259. [95] L. V. Kovalev, Domains with convex hyperbolic radius. Acta Math. Univ. Comenianae 70 (2001), 207–213.

Bibliography

149

[96] J. G. Krzy˙z, Problem 1, posed in : fourth conference on analytic functions. Ann. Polon. Math. 20 (1967-1968), 314. [97] R. K¨ uhnau, The conformal modul of quadrilaterals and of rings. Handbook in complex analysis: Geometric Function Theory 2 (2005), 99–129, Elsevier, Amsterdam. [98] E. Landau, Einige Bemerkungen u ¨ber schlichte Abbildung. J.ber. Deutsche Math. Verein. 34 (1925/26), 239–243. [99] N.A. Lebedev and I. M. Milin, An inequality. Vestnik Leningrad Univ. 20 (1965), 157–158 (in Russian). [100] O. Lehto, On the distortion of conformal mappings with bounded boundary rotation. Ann. Acad. Sci. Fenn. Ser. AI Math. Phys. 124 (1952), 14p. [101] V. Levin, Aufgabe 163. J.ber. Deutsche Math. Verein. 43 (1933), 133; L¨ osung, ibid. 44 (1934), 80–83 (solutions by W. Fenchel and E. Reissner). [102] J.-L. Li, Estimates for derivatives of holomorphic functions in a hyperbolic domain. J. Math. Anal. Appl. 329 (2007), 581–591. [103] J.-L. Li, Schwarz-Pick inequalities for convex domains. Kodai Math. J. 30 (2007), 252-262. [104] R. J. Libera and E. J. Zlotkiewicz, Early coeﬃcients of the inverse of a regular convex function. Proc. Amer. Math. Soc. 85 (1982), 225–230. [105] E. Lindel¨ of, M´emoire sur certaines in´egalit´es dans la th´eorie des fonctions monog`enes et sur quelques propri´et´es nouvelles de ces fonctions dans le voisinage d’un point singulier essentiel. Acta Societatis Scientiarum Fennicae XXXV (1909), Nr.7 (1908), 35 pp. [106] J. E. Littlewood, On inequalities in the theory of functions. Proc. London Math. Soc. 23 (1925), 481–519. [107] J. E. Littlewood, Lecture on the theory of functions. Oxford University Press, Oxford, 1944. [108] A. E. Livingston, Convex meromorphic mappings. Ann. Pol. Math. 59.3 (1994), 275–291. [109] K. L¨ owner, Untersuchungen u ¨ber die Verzerrung bei konformen Abbildungen des Einheitskreises |z| < 1, die durch Funktionen mit nicht verschwindender Ableitung geliefert werden. Leipziger Berichte 69 (1917), 89–106. [110] K. L¨ owner, Untersuchungen u ¨ber schlichte konforme Abbildungen des Einheitskreises I. Math. Ann. 89 (1923), 103–121. [111] W. Ma and D. Minda, Behavior of domain constants under conformal mappings. Israel J. Math. 91 (1995), 157–171.

150

Bibliography

[112] B. MacCluer, K. Stroethoﬀ, and R. Zhao , Generalized Schwarz-Pick estimates. Proc. Amer. Math. Soc. 131 (2002), 593–599. [113] B. MacCluer, K. Stroethoﬀ, and R. Zhao, Schwarz-Pick estimates. Complex Variables 48 (2003) 711–730. [114] A. J. Macintyre and W. W. Rogosinski, Extremum problems in the theory of analytic functions. Acta Math. 82 (1950), 275–325 . [115] A. Marx, Untersuchungen u ¨ber schlichte Abbildungen. Math. Ann. 107 (1932/33), 40–65. [116] I. M. Milin, On the coeﬃcients of univalent functions. Soviet Math. Doklady 8 (1967), 1255–1258. [117] J. Miller, Convex and starlike meromorphic functions. Proc. Amer. Math. Soc. 80 (1980), 607–613. [118] S. S. Miller and P. T. Mocanu, Diﬀerential subordination, Theory and applications. Marcel Dekker Inc., New York, 2000. [119] C. D. Minda and D. J. Wright, Univalence criteria and the hyperbolic metric in convex regions. Rocky Mtn. J. Math. 12 (1982), 471–479. [120] R. Nevanlinna, Eindeutige Analytische Funktionen. Springer-Verlag, Berlin, 1953. [121] B. G. Osgood, Some properties of f /f and the Poincar´e metric. Indiana Univ. Math. J. 31 (1982), 449–461. ¨ [122] V. Paatero, Uber die konforme Abbildung von Gebieten deren R¨ ander von beschr¨ ankter Drehung sind. Ann. Acad. Sci. Fenn. Ser. A 9 (1931), 77 pp. [123] J. Pfaltzgraﬀ and B. Pinchuk, A variational method for classes of meromorphic functions. J. Analyse Math. 24 (1971), 101–150. ¨ [124] G. Pick, Uber allgemeine Konvergenzs¨ atze der komplexen Funktionentheorie. Mat. Ann. 77:1 (1916), 1–6. ¨ [125] G. Pick, Uber die Beschr¨ ankungen analytischer Funktionen, welche durch vorgeschriebene Funktionswerte bewirkt werden. Mat. Ann. 77 (1916), 7–23. [126] Ch. Pommerenke, Linear-invariante Familien analytischer Funktionen. I. Math. Ann. 155 (1964), 108–154. [127] Ch. Pommerenke, On close-to-convex analytic functions. Trans. Amer. Math. Soc. 114 (1965), 176–186. [128] Ch. Pommerenke, Univalent functions. Vandenhoeck and Ruprecht, G¨ ottingen, 1975. [129] Ch. Pommerenke, Uniformly perfect sets and the Poincar´e metric. Arch. Math. 32 (1979), 192–199.

Bibliography

151

[130] Ch. Pommerenke, On uniformly perfect sets and Fuchsian groups. Analysis 4 (1984), 299–321. [131] Ch. Pommerenke, Boundary behaviour of conformal maps. Springer, New York, 1992. [132] Ch. Pommerenke, Personal Communication. 3. 12. 2002. [133] D. V. Prokhorov, Bounded univalent functions. Handbook in complex analysis: Geometric Function Theory 1 (2005), 207–228, North-Holland, Amsterdam. ¨ [134] J. Radon, Uber die Randwertaufgaben beim logarithmischen Potential. Sitzungsberichte der Oesterreichischen Akademie der Wissenschaften 128 (1920), 1123–1167. [135] M. S. Robertson, On the coeﬃcients of typically-real functions. Bull. Amer. Math. Soc. 41 (1935), 565–572. [136] M. S. Robertson, A remark on odd schlicht functions. Bull. Amer. Math. Soc.42 (1936), 366–370. [137] M. S. Robertson, Quasi-subordination and coeﬃcient conjectures. Bull. Amer. Math. Soc. 76 (1970), 1–9. ¨ [138] W. Rogosinski, Uber positive harmonische Entwicklungen und typisch reelle Potenzreihen. Math. Z. 35 (1932), 93–121. [139] W. Rogosinski, On the coeﬃcients of subordinate functions. Proc. London Math. Soc. 48 (1943), 48–82. [140] O. Roth and K.-J. Wirths, Taylor coeﬃcients of negative powers of schlicht functions. Comp. Methods and Function Theory 1 (2001), 521–533. ¨ [141] St. Ruscheweyh, Uber die Faltung schlichter Funktionen. Math. Z. 128 (1972), 85–92. ¨ [142] St. Ruscheweyh, Uber einige Klassen in Einheitskreis holomorpher Funktionen. Ber. Math.-Stat. Sektion Forschungszentrum Graz 7 (1974), 12p. [143] St. Ruscheweyh, Two remarks on bounded analytic functions. Serdica, Bulg. math. publ. 11 (1985), 200–202. [144] St. Ruscheweyh and T. Sheil-Small, Hadamard products of schlicht functions and the P´ olya-Schoenberg conjecture. Comment. Math. Helv. 48 (1973), 119– 135. [145] N. Samaris, A proof of Krzy˙z’s conjecture for the ﬁfth coeﬃcient. Complex Variables 48 (2003), 753–766. [146] G. Schober, Univalent Functions - Selected Topics. Lecture Notes in Math. 478, Springer, New York, 1975.

152

Bibliography

[147] G. Schober, Coeﬃcient estimates for inverses of schlicht functions. In: Aspects of contemporary complex analysis, Academic Press, New York, 1980. ¨ [148] I. Schur, Uber Potenzreihen, die im Innern des Einheitskreises be-schr¨ ankt sind. J. Reine Angew. Math. 147 (1917), 205–232. [149] H. A. Schwarz, Gesammelte Abhandlungen. Vol.II, Springer, Berlin, 1890. [150] J. H. Shapiro, Composition operators and classical function theory. Springer, New York 1993. [151] T. Sheil-Small, On convex univalent functions. J. London Math. Soc.(2)1 (1969), 483–492. [152] T. Sheil-Small, On the convolution of analytic functions. J. Reine Angew. Math. 258 (1973), 137–152. [153] A. Yu. Solynin, On separation of continua by circles (Russian). Zap. Nauchn. Semin. Leningr. Otd. Mat. Inst. Steklova 168 (1988), 154–157. [154] A. Yu. Solynin and M. Vuorinen, Estimates for the hyperbolic metric of the punctured plane and applications. Isr. J. Math. 124 (2001), 29–60. [155] E. Strohh¨ acker, Beitr¨ age zur Theorie der schlichten Funktionen. Math. Z. 37 (1933), 356–380. [156] T. J. Suﬀridge, Some remarks on convex maps of the unit disc. Duke Math. J. 37 (1970), 775–777. [157] T. J. Suﬀridge, Problems for Non-vanishing H p Functions. In: Lecture Notes in Math.(1990) Nr. 1435, 177–190. [158] W. Szapiel, A new approach to the Krzy˙z conjecture. Ann. Univ. Mariae Curie-Sklodowska Sect. A 48 (1994), 169–192. [159] O. Sz´ asz, Ungleichungen f¨ ur die Koeﬃzienten einer Potenzreihe. Math. Z. 1 (1918), 163–183. [160] O. Sz´ asz, Ungleicheitsbeziehungen f¨ ur die Ableitungen einer Potenzreihe, die eine im Einheitskreis beschr¨ ankte Funktion darstellt. Math. Z. 8 (1920), 303– 309. [161] D. L. Tan, Estimates of coeﬃcients of bounded nonvanishing analytic functions. (Chinese), Chinese Ann. Math., Ser. A 4 (1983), 97–104. English summary in Chinese Ann. Math., Ser. B 4 (1983), 131-132. [162] O. Teichm¨ uller, Untersuchungen u ¨ber konforme und quasikonforme Abbildungen. Deutsche Math. 3 (1938), 621–678. [163] P. Todorov, New explicit formulas for the nth derivative of composite functions. Paciﬁc J. Math. 92 (1981), 217–236.

Bibliography

153

[164] S. Y. Trimble, A coeﬃcient inequality for convex univalent functions. Proc. Amer. Math. Soc. 48 (1975), 266–267. ¨ [165] K.-J. Wirths, Uber holomorphe Funktionen, die einer Wachstumsbeschr¨ ankung unterliegen. Arch. Math. 30 (1978), 606–612. [166] K.-J. Wirths, The Koebe domain for concave univalent functions. Serdica Math. J. 29 (2003), 355-360. [167] K.-J. Wirths, On the residuum of concave univalent functions. Serdica Math. J. 32 (2006), 209–214. [168] S. Yamashita, La d´eriv´ee d’une fonction univalente dans une domaine hyperbolique. C. R. Acad. Sci. Paris Ser I. Math. 314 (1992), 45–48. [169] S. Yamashita, Localization of the coeﬃcient Theorem. Kodai Math. J. 22 (1999), 384–401. [170] S. Yamashita, Higher derivatives of holomorphic functions with positive real part. Hokkaido Math. J. 29 (2000), 23-36.

Index Agard’s formula, 32 Ahlfors’ formula, 39 angle conjecture, 128 angular domains, 24 angular point of order α, 79 angularly accessible domains, 116 asymptotic conjecture, 11 Avkhadiev’s theorem, 36 Baernstein and Schober’s theorem, 22 Bateman functions, 109 Beardon and Pommerenke inequalities, 41 Beardon and Pommerenke’s theorem, 35 Bermant, Hempel theorem, 32 Bieberbach conjecture, 3, 11 Bieberbach theorem, 44 boundary rotation, 23 Brannan, Clunie and Kirwan’s theorem, 23 Carleson and Gamelin’s remark, 37 central existence theorem, 97 Chua conjecture, 69 comparison of densities, 31 concave domains, 90 concave univalent functions, 91 conformal radius, 2 conformal radius as a function its connection with coeﬃcients, 44 its gradient image, 46 its derivatives, 47

conformally invariant families, 117 convex domains, 63 de Branges’ theorem, 14 Denjoy-Wolﬀ point, 33 epicycloids or hypocycloids, 46 factor Cn (Hβ , Hα ), 73 Fej´er’s inequality, 65 Gaussian curvature, 30 generalized Goodman conjecture, 131 generalized punishing factors, 84 genuine annuli, 35 Goluzin type theorem, 10 Goodman conjecture, 12 Grand Iteration Theorem, 32 Hadamard product, 117 Hardy type inequality, 36 hyperbolic distance, 28 hyperbolic metric principle, 31 Jakubowski’s theorem, 113 Jenkins’ theorem, 15 Klouth and Wirths’ theorem, 21 Koebe 1/4-theorem , 33 Koebe transform, 44 Kovalev’s theorem, 45 Krzy˙z conjecture, 127 L¨owner diﬀerential equation, 14 L¨owner theorem, 45

156 L¨owner theorem on inverse coeﬃcients, 19 Landau theorem on holomorphic functions omitting 0 and 1, 33 Landau’s estimate for derivatives, 113 Lebedev-Milin inequality, 13 Liouville equation, 29 Littlewood conjecture, 11 Littlewood’s theorem, 7 Lobachevsky and Bolyai geometry, 27 lower bound conjecture, 58 Marx-Strohh¨ acker theorem, 64 Milin conjecture, 11 modiﬁed punishing factors, 104 modular function, 32 Osgood inequality, 41 Osgood-Jørgensen inequality, 41 Pick, Lindel¨ of theorem, 30 Poincar´e metric, 27 Pommerenke’s characteristic, 40 Ptolemaic system, 46 punishing factor C2 (Ω, Π), 68 quasi-subordinate functions, 9 Radon’s curves, 23 Robertson conjecture, 11 Rogosinski conjecture, 11 Rogosinski’s theorem, 8 Ruscheweyh’s theorems, 52 Schur-Jabotinsky theorem, 64 Schwarz’s lemma, 30 sharp lower bounds, 79 Sheil-Small conjecture, 12 Sheil-Small’s theorem, 17 subordinate functions, 7 Sz´asz inequalities, 50 Szapiel’s problem, 128 Teichm¨ uller’s theorem, 39 theorem on 4n−1 , 55

Index uniformly perfect sets, 37 universal covering map, 28 Wirtinger calculus, 29 Yamashita’s theorem, 55

Advisory Editorial Board Leonid Bunimovich (Georgia Institute of Technology, Atlanta) Benoît Perthame (Université Pierre et Marie Curie, Paris) Laurent Saloff-Coste (Cornell University, Ithaca) Igor Shparlinski (Macquarie University, New South Wales) Wolfgang Sprössig (TU Bergakademie Freiberg) Cédric Vilani (Ecole Normale Supérieure, Lyon)

Farit G. Avkhadiev Karl-Joachim Wirths

Schwarz-Pick Type

Inequalities

Birkhäuser Verlag Basel . Boston . Berlin

$XWKRUV )DULW*$YNKDGLHY Chebotarev Research Institute .D]DQ6WDWH8QLYHUVLW\ .D]DQ Russia HPDLO)DULW$YNKDGLHY#NVXUX

.DUO-RDFKLP:LUWKV ,QVWLWXWIU$QDO\VLVXQG$OJHEUD 78%UDXQVFKZHLJ %UDXQVFKZHLJ *HUPDQ\ HPDLONMZLUWKV#WXEVGH

0DWKHPDWLFDO6XEMHFW&ODVVL½FDWLRQ$&&&' )0

/LEUDU\RI&RQJUHVV&RQWURO1XPEHU

%LEOLRJUDSKLFLQIRUPDWLRQSXEOLVKHGE\'LH'HXWVFKH%LEOLRWKHN 'LH'HXWVFKH%LEOLRWKHNOLVWVWKLVSXEOLFDWLRQLQWKH'HXWVFKH1DWLRQDOELEOLRJUD½H GHWDLOHGELEOLRJUDSKLFGDWDLVDYDLODEOHLQWKH,QWHUQHWDWKWWSGQEGGEGH!

,6%1%LUNKlXVHU9HUODJ$*%DVHOÀ%RVWRQÀ%HUOLQ 7KLVZRUNLVVXEMHFWWRFRS\ULJKW$OOULJKWVDUHUHVHUYHGZKHWKHUWKHZKROHRUSDUWRIWKH PDWHULDOLVFRQFHUQHGVSHFL½FDOO\WKHULJKWVRIWUDQVODWLRQUHSULQWLQJUHXVHRILOOXVWUD WLRQVUHFLWDWLRQEURDGFDVWLQJUHSURGXFWLRQRQPLFUR½OPVRULQRWKHUZD\VDQGVWRUDJHLQ GDWDEDQNV)RUDQ\NLQGRIXVHSHUPLVVLRQRIWKHFRS\ULJKWRZQHUPXVWEHREWDLQHG %LUNKlXVHU9HUODJ$* %DVHOÀ%RVWRQÀ%HUOLQ 32%R[&+%DVHO6ZLW]HUODQG Part of Springer Science+Business Media &RYHUGHVLJQ%LUJLW%ORKPDQQ=ULFK6ZLW]HUODQG 3ULQWHGRQDFLGIUHHSDSHUSURGXFHGIURPFKORULQHIUHHSXOS7&) 3ULQWHGLQ*HUPDQ\ ,6%1

H,6%1

ZZZELUNKDXVHUFK

Dedicated to our families

Contents 1

2

Introduction 1.1 Historical remarks . . . . . . . . . . 1.2 On inequalities for higher derivatives 1.3 On methods . . . . . . . . . . . . . . 1.4 Survey of the contents . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

1 1 3 5 6

Basic coeﬃcient inequalities 2.1 Subordinate functions . . . . . . . . . . . 2.2 Bieberbach’s conjecture by de Branges . . 2.3 Theorems of Jenkins and Sheil-Small . . . 2.4 Inverse coeﬃcients . . . . . . . . . . . . . 2.5 Domains with bounded boundary rotation

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

7 7 11 15 18 23

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

27 27 30 33 37 40 44

Basic Schwarz-Pick type inequalities 4.1 Two classical inequalities . . . . . . . . . . 4.2 Theorems of Ruscheweyh and Yamashita . 4.3 Pairs of simply connected domains . . . . . 4.4 Holomorphic mappings into convex domains 4.5 Punishing factors for convex pairs . . . . . 4.6 Case n = 2 for all domains . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

49 50 52 55 59 63 66

Punishing factors for special cases 5.1 Solution of the Chua conjecture . . . . . . . . . . . . . . . . . . . . 5.2 Punishing factors for angles . . . . . . . . . . . . . . . . . . . . . .

69 69 72

3 The 3.1 3.2 3.3 3.4 3.5 3.6 4

5

. . . .

. . . .

Poincar´e metric Background . . . . . . . . . . . . . . . . . The Schwarz-Pick inequality . . . . . . . . Estimates using the Euclidean distance . . An application of Teichm¨ uller’s theorem . Domains with uniformly perfect boundary Derivatives of the conformal radius . . . .

viii

Contents 5.3 5.4 5.5

Sharp lower bounds for punishing factors . . . . . . . . . . . . . . Domains in the extended complex plane . . . . . . . . . . . . . . . Maps from convex into concave domains . . . . . . . . . . . . . . .

78 84 90

6

Multiply connected domains 97 6.1 Finitely connected domains . . . . . . . . . . . . . . . . . . . . . . 97 6.2 Pairs of arbitrary domains . . . . . . . . . . . . . . . . . . . . . . . 104 6.3 Some examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

7

Related results 113 7.1 Inequalities for schlicht functions . . . . . . . . . . . . . . . . . . . 113 7.2 Derivatives of α-invariant functions . . . . . . . . . . . . . . . . . . 117 7.3 A characterization of convex domains . . . . . . . . . . . . . . . . 124

8

Some open problems 8.1 The Krzy˙z conjecture . . . . . . . . 8.2 The angle conjecture . . . . . . . . . 8.3 The generalized Goodman conjecture 8.4 Bloch and several variable problems 8.5 On sums of inverse coeﬃcients . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

127 127 128 131 139 140

Bibliography

143

Index

155

Chapter 1

Introduction The aim of the present book is a uniﬁed representation of some recent results in geometric function theory together with a consideration of their historical sources. These results are concerned with functions f , holomorphic or meromorphic in a domain Ω in the extended complex plane C. The only additional condition we impose on these functions is the condition that the range f (Ω) is contained in a given domain Π ⊂ C. This fact will be denoted by f ∈ A(Ω, Π). We shall describe how one may get estimates for the derivatives |f (n) (z0 )|, n ∈ N, f ∈ A(Ω, Π), dependent on the position of z0 in Ω and f (z0 ) in Π.

1.1

Historical remarks

The beginning of this program may be found in the famous article [125] of G. Pick. There, he discusses estimates for the MacLaurin coeﬃcients of functions with positive real part in the unit disc found by C. Carath´eodory in [52]. Pick tells his readers that he wants to generalize Carath´eodory’s estimates such that the special role of the expansion point at the origin is no longer important. For the convenience of our readers we quote this sentence in the original language: Durch lineare Transformation von z oder, wie man sagen darf, durch kreisgeometrische Verallgemeinerung, kann man die Sonderstellung des Wertes z = 0 wegschaﬀen, so daß sich Relationen f¨ ur die Diﬀerentialquotienten von w an beliebiger Stelle ergeben. The ﬁrst great success of this program was G. Pick’s theorem, as it is called by Carath´eodory himself, compare [54], vol II, §286–289. If z0 ∈ Δ = {z | |z| < 1}, f ∈ A(Δ, Δ), and f (z0 ) = w0 , then the inequality w0 − f (z) ≤ z0 − z , z ∈ Δ, (1.1) 1 − z0 z 1 − w0 f (z) is valid.

2

Chapter 1. Introduction

This theorem follows immediately from Schwarz’s Lemma using holomorphic automorphisms of the unit disc. A direct consequence of Pick’s theorem is the inequality 1 − |f (z0 )|2 |f (z0 )| ≤ , f ∈ A(Δ, Δ), z0 ∈ Δ. (1.2) 1 − |z0 |2 To be more complete about the history of the inequality (1.2) we have to mention that O. Sz´ asz (see the footnote in [160], p.308) attributes it to E. Lindel¨ of [105] and indicates that (we quote once more): Neuere Beweise dieser Relation gaben Carath´eodory und Jensen. Herr Jensen zeigte, wie der Satz aus dem schon von Herrn Landau 1906 bewiesenen Spezialfall: |c1 | ≤ 1 − |c0 |2 leicht folgt. As an oft-quoted proverb says “Success has many fathers, but failure is an orphan”. For us, at this place the second part of this proverb hints of unsolved mathematical problems. It is clear that Schwarz’s lemma and its generalizations became widely known in the period of a systematic study of results which are closely connected with the proofs of the Riemann theorem on conformal mappings. By the way, the original version of the lemma may be found in [149], where H.A. Schwarz discussed the Riemann mapping theorem. Now, it is well known that one may generalize the inequalities (1.1) and (1.2) to f ∈ A(Ω, Π), z0 ∈ Ω, and f (z0 ) = w0 , where Ω and Π are domains that have at least three boundary points. This generalization is known as the principle of hyperbolic metric (see for instance R. Nevanlinna [120] and G. M. Goluzin [70]). An important case is presented by simply connected domains. Let Ω and Π be simply connected proper subdomains of C and let ΦΩ,z0 be the unique conformal map of Δ onto Ω such that ΦΩ,z0 (0) = z0 and ΦΩ,z0 (0) > 0. The existence of this map is proved by the Riemann mapping theorem.We deﬁne ΦΠ,f (z0 ) analogously. Then the function w = Φ−1 (1.3) Π,f (z0 ) ◦ f ◦ ΦΩ,z0 belongs to the family A(Δ, Δ) and satisﬁes w(0) = 0. Hence, Schwarz’s Lemma implies ΦΠ,f (z0 ) (0) |f (z0 )| ≤ , ΦΩ,z0 (0) the generalized Schwarz-Pick inequality. The quantities R(z0 , Ω) := ΦΩ,z0 (0)

and R(f (z0 ), Π) := ΦΠ,f (z0 ) (0)

are called the conformal radius of Ω at the point z0 and of Π at w0 = f (z0 ), respectively, and will be used here to describe the positions of the points z0 in Ω and f (z0 ) in Π.

1.2. On inequalities for higher derivatives

1.2

3

On inequalities for higher derivatives

We ask for inequalities of the form |f (n) (z0 )| R(f (z0 ), Π) , ≤ Mn (z0 , Ω, Π) n! (R(z0 , Ω))n

n ∈ N,

(1.4)

where f and z0 are as above. Concerning the history of inequalities for higher derivatives, we should mention here that the interest of researchers in geometric function theory was concentrated for a long time after 1920s on the famous Bieberbach conjecture, i.e., the conjecture that |f (n) (0)| (1.5) ≤ n|f (0)| n! for functions f holomorphic and injective on Δ. This may be seen as a special case of the above inequality, where f maps Δ conformally onto the simply connected domain f (Δ). Since it seemed very diﬃcult to prove this conjecture, less attention was attracted by the generalized Bieberbach conjecture or Rogosinski conjecture, which in the above formulation means that |f (n) (0)| ≤ nR(f (0), Π) n!

(1.6)

for any simply connected proper subdomain Π of C and any f ∈ A(Δ, Π). An equivalent formulation of the Rogosinski conjecture is the following. Let g be holomorphic and injective on Δ and let there exist a function w : Δ → Δ such that w(0) = 0 and f = g ◦ w. Then (1.6) is valid. As usual, we will abbreviate this relation between f and g by f ≺ g and say that under these circumstances f is subordinated to g. It was a great surprise when de Branges succeeded in proving not only (1.5) but also (1.6) in 1985. Soon afterwards, it was recognized, especially by Yamashita, that one could use (1.5) to go further steps in Pick’s program for functions injective on Δ. In fact, it had been seen earlier by Landau and Jakubowski that the validity of the Bieberbach conjecture would imply sharp bounds for |f (n) (z0 )|, z0 ∈ Δ, and f injective on Δ. We will give an outline of these results in the present book. Further, we will speak on the ideas of Chua [56], who indicated that (1.5) could be used to derive similar bounds for functions f injective on simply connected domains Ω. In this paper, he considered in addition the case that Ω is convex and published the following conjectures. Let f be holomorphic and injective on a proper convex subdomain Ω of C. Then for any z0 ∈ Ω, and any n ≥ 2 the inequality |f (n) (z0 )| |f (z0 )| ≤ (n + 1)2n−2 n! (R(z0 , Ω))n−1

(1.7)

4

Chapter 1. Introduction

is valid. If in addition f (Ω) is convex, then |f (z0 )| |f (n) (z0 )| ≤ 2n−1 n! (R(z0 , Ω))n−1

(1.8)

is valid. Motivated by the work of Ruscheweyh and Yamashita, who had proved estimates of the type (1.4) and similar ones, the authors of the present book concentrated research in the last years on formulas like (1.4). The results of this research together with known theorems in this direction form the content of this book. Especially, we shall show that Chua’s conjectures are true. We will show even more, namely that Mn (z0 , Ω, Π) ≤ (n + 1)2n−2 for f ∈ A(Ω, Π), Ω convex and Π simply connected and that Mn (z0 , Ω, Π) ≤ 2n−1 for f ∈ A(Ω, Π), Ω and Π convex. These results will be completed by several considerations for special cases of Ω and Π as well as by computations of the dependences of Mn (z0 , Ω, Π) of the variable z0 . One further implication of the validity of the Bieberbach conjecture is concerned with meromorphic functions univalent on Δ. In 1962, J. Jenkins considered functions f that in addition to the above properties satisfy the condition that the pole of f lies at a point p ∈ (0, 1). He proved that for such functions the Bieberbach conjecture would imply n−1 |f (n) (0)| |f (0)| 2k p . ≤ n−1 n! p

(1.9)

k=0

This result motivated us to generalize the above considerations to proper subdomains of C. It is natural that here the positions of z0 in Ω and f (z0 ) in Π relative to the point at inﬁnity enter the picture. We use the hyperbolic distances to characterize those items. We will prove results of the type (1.4) using a generalization of the conformal radius. These results are satisfying in the following cases: 1) Ω and Π are simply connected with respect to C. 2) Ω is simply connected with respect to C and Π is convex. 3) Ω is convex and C \ Π is compact and convex. In the important case that Ω = Δ or Ω convex and Π is simply connected with respect to C we can present only partial results. One of the chapters is devoted to the most diﬃcult case, the case of multiply connected domains. Here, we deﬁne appropriate questions and give some answers that in most cases are far from the sharpness we achieved in many of the above mentioned theorems. Hence this chapter is more or less an impetus for further research.

1.3. On methods

5

1.3 On methods What can be said about methods? Essentially, the proofs of Schwarz-Pick type inequalities are based on relationships of certain hyperbolic characteristics of domains with the following results in geometric function theory: 1. Littlewood’s ideas on subordinate functions developed by Rogosinski, Goluzin, Clunie, Robertson, and Sheil-Small. 2. Coeﬃcient estimates using L¨owner’s theory on parametric representation of univalent functions. Especially, L¨ owner’s theorem on inverse coeﬃcients and de Branges’ proof of the Bieberbach conjecture. 3. Explicit representation of convex hulls for several families of analytic functions by the Herglotz formula and its generalizations. Usually, coeﬃcient estimates concern univalent or subordinate functions holomorphic in the unit disk. We prove and use several new versions of the known theorems for subordinate or quasi-subordinate functions which are holomorphic in a neighbourhood of the origin. Also, to prove Schwarz-Pick type inequalities for higher derivatives one needs certain hyperbolic characteristics of plane domains. For instance, let Qn (z, w, Ω, Π) be deﬁned as the smallest possible value such that the inequality (n) f (z) R(w, Π) ≤ Qn (z, w, Ω, Π) n! R(z, Ω)n holds for all f ∈ A(Ω, Π), f (z) = w, and Cn (Ω, Π) = sup{Qn (z, w, Ω, Π) | (z, w) ∈ Ω × Π}. The principle of the hyperbolic metric implies that Q1 (z, w, Ω, Π) = 1, and, in turn, that C1 (Ω, Π) = 1 for any pair (Ω, Π) equipped with the Poincar´e metric. In the case n ≥ 2 the quantity Qn depends on the hyperbolic characteristics Rk (z, Ω)

∂ k log R(z, Ω) ∂ k log R(w, Π) and Rk (w, Π) , 1 ≤ k ≤ n − 1. k ∂z ∂wk

In particular, Q2 (z, w, Ω, Π) depends on z and w via the quantities p = |∇ R(z, Ω)| and q = |∇ R(w, Π)|, only. Consequently, explicit estimates of punishing factors Cn (Ω, Π) are closely connected with the behaviour of the above characteristics, and, roughly speaking, estimating of these quantities is related to certain coeﬃcient problems of geometric function theory. Finally, we have to attract the reader’s attention to the following fact which deserves to be widely known: The quantity reciprocal to the conformal or hyperbolic radius is exactly the density of the Poincar´e metric. More precisely, the equation λΩ (z) :=

1 , R(z, Ω)

z ∈ Ω,

6

Chapter 1. Introduction

deﬁnes the density of the hyperbolic metric in the domain Ω with Gaussian curvature K = −4.

1.4

Survey of the contents

In this section, we would like to give a short survey of the contents of the present book. Chapters 2 and 3 have a preparatory character. There we will gather the materials from the work of many researchers in geometric function theory that we need for our results. Therefore we do not give all proofs in detail in these two chapters and we shortened and simpliﬁed old proofs. During these eﬀorts we found “new looks through old holes” now and then. Chapter 2 is dedicated to the most famous coeﬃcient theorems of the last century, namely the Bieberbach conjecture and the conjecture for the coeﬃcients of functions with bounded boundary rotation. The third chapter discusses the second theme that is important for our generalizations of the Schwarz-Pick lemma; the Poincar´e metric, its historical background, and well-known theorems concerning this metric in diﬀerent circumstances. Among them are the famous theorems of Landau, Teichm¨ uller, Beardon and Pommerenke. Most of the material presented in the following chapters has been developed by us in the last ten years, some of it has been published, other results appear in this book for the ﬁrst time. Chapter 4 contains the most prominent members of the family of punishing factors, those for pairs of simply connected domains and those for pairs of convex domains. Moreover, these two are in some sense the most beautiful, namely 4n−1 and 2n−1 . We discuss the work of Ruscheweyh, Chua, and Yamashita on these questions. Further, we present the complete solution for the case n = 2 for any pair of domains. The ﬁfth chapter is devoted to more special results. The most prominent of them may be the proof of a far reaching generalization of Chua’s conjecture. We add the determination of punishing factors for pairs of simply connected domains in the extended plane, and we prove sharp lower bounds in some general circumstances. Chapter 6 is concerned with some generalizations to multiply connected domains for arbitrary n > 2 and it has at some places a tentative character, since we are not sure that we have found the “best” way of generalization. In the last two chapters, we present some material in the neighbourhood of our results. This is meant as the basis for further research on the many questions that are natural to pose here. At this point, we want to thank the Deutsche Forschungsgemeinschaft for continued support of our research. Their many grants for F. G. Avkhadiev enabled us to do all the scientiﬁc work presented in this book.

Chapter 2

Basic coeﬃcient inequalities There are many books that systematically present coeﬃcient problems in geometric function theory. We refer the reader to the excellent monographs by Goluzin [70], Goodman [73], Hayman [78], Pommerenke [128], and Duren [60]. In this chapter we only mention a few classical results on coeﬃcients which are closely connected with the topic of this book. Also, we give several new facts with short proofs that have until now been presented only in original papers.

2.1

Subordinate functions

Let the functions Φ and Ψ be meromorphic in the unit disc Δ. We will say that Φ is subordinate to Ψ and write Φ ≺ Ψ or Φ(z) ≺ Ψ(z), whenever there exists a function ω holomorphic in Δ with properties ω(0) = 0, |ω(z)| < 1, and such that Φ(z) = Ψ(ω(z)). Clearly, |ω(z)| ≤ |z| in Δ by the Schwarz lemma. In [106], Littlewood proved the following result. Theorem 2.1. Let Φ and Ψ be holomorphic in the unit disc Δ, and let Φ(0) = Ψ(0) = 0. If Φ(z) ≺ Ψ(z) and p ∈ (0, ∞), then for any r ∈ (0, 1), 0

2π

|Φ(reiθ )|p dθ ≤

0

2π

|Ψ(reiθ )|p dθ.

We reproduce a short proof of Theorem 2.1 for the case p ∈ N, only.

(2.1)

8

Chapter 2. Basic coeﬃcient inequalities

Proof of Theorem 2.1 in the case p ∈ N. By Poisson’s formula, 2π ζ +z 1 p p dθ, ζ = reiθ , |z| < r < 1. Ψ (ζ)Re Ψ (z) = 2π 0 ζ −z For |ω(z)| ≤ |z| < r we can write |Φ(z)|p = |Ψ(ω(z))|p ≤

1 2π

2π

|Ψ(ζ)|p Re

0

ζ + ω(z) ζ − ω(z)

dθ,

ζ = reiθ .

Integrating the latter inequality over the circle {z = ρeit | 0 ≤ t ≤ 2π}, ρ ∈ (0, r), using 2π 1 ζ + ω(ρeit ) ζ + ω(0) Re dt = Re = 1, 2π 0 ζ − ω(ρeit ) ζ − ω(0) one gets

0

2π

|Φ(ρeit )|p dt ≤

2π

0

|Ψ(reiθ )|p dθ,

0 < ρ < r < 1.

Letting ρ → r gives inequality (2.1) in the case p ∈ N.

Remark 2.2. Clearly, a direct use of Theorem 2.1 with p = 2 and Parseval’s formula for the holomorphic functions Φ(z) =

∞

An z n

and

Ψ(z) =

n=0

leads to the inequality

∞

Bn z n ,

|z| < 1,

n=0 ∞

|An |2 ≤

n=0

∞

|Bn |2 .

n=0

Using Theorem 2.1 with p = 2 in an original way, Rogosinski proved the following assertion on coeﬃcients of these two functions. Theorem 2.3 (Rogosinski, [139]). Let Φ and Ψ be holomorphic in the unit disc Δ. If Φ(z) ≺ Ψ(z), then for all n ∈ N ∪ {0}, n

|Ak |2 ≤

k=0

n

|Bk |2 .

k=0

Simple counterexamples show that the inequality |An | ≤ |Bn | for n ≥ 2 does not hold, for instance, for the functions Φ(z) = z n and Ψ(z) = z. Also, the assertion of Theorem 2.3 with p = 2, i.e., the inequality n k=0

|Ak |p ≤

n k=0

|Bk |p

2.1. Subordinate functions

9

is not true in general (see [139]). In [137], Robertson generalized Theorem 2.3 to the case when the functions Φ and Ψ are holomorphic in the unit disc Δ and Φ(z) = ϕ(z)Ψ(ω(z)),

|z| < 1,

where |ϕ(z)| ≤ 1 for z ∈ Δ and ω is as above. In this case Φ is said to be quasisubordinate to the function Ψ (see also [57] and [128]). Robertson’s theorem can be generalized to meromorphic functions or, more generally, to functions F and G that are holomorphic only in a neighbourhood of the origin (see [19], [30] and [20]). Theorem 2.4 (see [20]). Let the functions F and G be holomorphic in a neighbourhood of the origin, where they have expansions F (z) =

∞

An z n

and

G(z) =

n=0

∞

Bn z n .

n=0

If there exist two functions ϕ and ω holomorphic in the unit disc Δ with |ϕ(z)| ≤ 1 and |ω(z)| ≤ |z| for z ∈ Δ such that the identity F (z) = ϕ(z)G(ω(z))

(2.2)

is satisﬁed in a neighbourhood of the origin, then for all n ∈ N ∪ {0} the following inequalities are valid: n n 2 |Ak | ≤ |Bk |2 . (2.3) k=0

k=0

Proof. The proof diﬀers from the classical one only in some details. We ﬁx n ∈ N ∪ {0}. From (2.2) we conclude that, in a neighbourhood of the origin, ∞

Ak z k − ϕ(z)

k=n+1

∞

Bk ω(z)k = ϕ(z)

k=n+1

n

Bk ω(z)k −

k=0

n

Ak z k .

k=0

If we expand the diﬀerence on the left side of this equation in a Taylor series in a neighbourhood of the origin, we see that the coeﬃcients of z k , 0 ≤ k ≤ n, in this series are zero. Therefore, we get, if we denote this function by H, an expansion H(z) =

∞

Ck z k

k=n+1

in a neighbourhood of the origin. On the other hand, it is immediately clear that the diﬀerence on the right side is holomorphic in the whole unit disc. This implies that the identity n k=0

Ak z k +

∞ k=n+1

Ck z k = ϕ(z)

n k=0

Bk ω(z)k

10

Chapter 2. Basic coeﬃcient inequalities

is valid for |z| < 1, too. Now, we observe that the Littlewood theorem is true for quasi-subordinate holomorphic functions, too. Consequently, we may proceed as in the classical proof of Theorem 2.3 to get n

|Ak |2 ≤

k=0

n

∞

|Ak |2 +

k=0

|Ck |2 ≤

k=n+1

n

|Bk |2

k=0

by use of Parseval’s formula in Theorem 2.1 with p = 2 for the quasi-subordinate holomorphic functions Φ(z) =

n

∞

Ak z k +

k=0

Ck z k ,

Ψ(z) =

k=n+1

n

Bk z k ,

z ∈ Δ.

k=0

This completes the proof of Theorem 2.4.

Following the idea of Goluzin (see [70] and compare also [57] and [137], Theorem 6.3) one may obtain a generalization of Theorem 2.4 as follows. Consider inequality (2.3) for n = 0, 1, . . . , m, where m ∈ N. Let λ0 ≥ λ1 ≥ · · · ≥ λm ≥ λm+1 = 0. We multiply the inequality (2.3) for n = j by λj − λj+1 ≥ 0, j = 0, 1, . . . , m. Summing up the results over j one easily has m

λj |Aj |2 ≤

j=0

m

λj |Bj |2 .

j=0

Since m is arbitrary, this gives the following theorem, Goluzin’s version of the inequalities (2.3). Theorem 2.5. Let n ∈ N ∪ {0} and let F and G be as in Theorem 2.4. If λ0 ≥ λ1 ≥ . . . λn ≥ 0, then the following inequality is valid: n

λk |Ak |2 ≤

k=0

n

λk |Bk |2 .

(2.4)

k=0

The following assertion generalizes a known idea due to Clunie [57]. Corollary 2.6. Let n ∈ N ∪ {0} and let f and g be functions meromorphic in the unit disc Δ with expansions of the form f (z) =

∞ k=0

Ak z k−m

and

g(z) =

∞

Bk z k−m

k=0

in a neighbourhood of the origin with some m ∈ N. If λ0 ≥ λ1 ≥ . . . λn ≥ 0 and |f (z)| ≤ |g(z)| in Δ, then the inequality (2.4) is valid.

2.2. Bieberbach’s conjecture by de Branges

11

To obtain Corollary 2.6 it is suﬃcient to apply Theorem 2.5 to functions F and G deﬁned by F (z) = z m f (z), G(z) = z m g(z), ω(z) = z, ϕ(z) = f (z)/g(z).

2.2

Bieberbach’s conjecture by de Branges

Consider the normal family S of all functions f that are holomorphic and univalent in Δ and have a Taylor expansion of the form f (z) = z +

∞

an z n ,

|z| < 1.

n=2

Before de Branges’ proof of the Bieberbach conjecture in [43] via the Milin conjecture, the following seven conjectures in their full generality were open problems. 1. Bieberbach Conjecture (1916). ([40]) For any f ∈ S the inequality |an | ≤ n holds for all n ≥ 2. The equality occurs if and only if f (z) is the Koebe function kd (z) = z(1 + dz)−2 , |d| = 1. 2. Littlewood Conjecture (1925). ([106]) If f ∈ S and f (z) = w for any z ∈ Δ, then |an | ≤ 4|w|n holds for all n ≥ 2. 3. Robertson Conjecture (1936). ([136]) For any odd function h(z) = z + c3 z 3 + c5 z 5 + · · · in S, the inequality 1 + |c3 |2 + · · · + |c2n−1 |2 ≤ n, is true for all n ≥ 2. 4. Rogosinski (Generalized Bieberbach) Conjecture (1943). ([139]) Let g(z) = b1 z + · · · + bn z n + · · · be a holomorphic function in Δ. If g(Δ) ⊂ f (Δ)and f ∈ S, then the inequality |bn | ≤ n holds for all n ≥ 2. 5. Asymptotic Bieberbach Conjecture (1955), connected with Hayman’s regularity theorem (see [78]). If An = max |an |, f ∈S

then lim

n→∞

An = 1. n

6. Milin Conjecture (1967). ([116]) For any f ∈ S, let γn be deﬁned by log

∞ f (z) γn z n . =2 z n=1

12

Chapter 2. Basic coeﬃcient inequalities Then the inequality

m n m=1 k=1

1 k|γk | − k 2

≤0

holds for all n ≥ 1. 7. Sheil-Small Conjecture (1973). ([152]) For any f ∈ S and any polynomial P (z) = b0 + b1 z + · · · + bn z n the convolution (= Hadamard product) deﬁned by (P ∗ f )(z) = a1 z + a2 b2 z 2 + · · · + an bn z n satisﬁes the inequality max |(P ∗ f )(z)| ≤ n max |P (z)|,

|z|≤1

|z|≤1

for all n ≥ 2. We refer the reader to the nice book [71] concerning details of the logical non-trivial relationship between these seven conjectures. In general, there are the following implications: =⇒ =⇒ =⇒

Milin Conjecture =⇒ Robertson Conjecture Sheil-Small Conjecture =⇒ Rogosinski Conjecture Bieberbach Conjecture Asymptotic Bieberbach Conjecture =⇒ Littlewood Conjecture.

Thus, in [43] de Branges settled all these conjectures by proving the Milin conjecture although he considered only the implications Milin Conjecture ⇒ Robertson Conjecture ⇒ Bieberbach Conjecture. In connection with these seven conjectures, in the next section we shall examine with proof two facts which deserve to be better known. The ﬁrst one concerns a conjecture by Goodman (see[72] and [73]), which says that the Taylor coeﬃcients ∞ of the function fp (z) = z + n=2 an (p)z n , |z| < p, satisfy the sharp inequality |an (p)| ≤ 1 + p2 + · · · + p2n−2 /pn−1 , whenever the function f is meromorphic and univalent in Δ and f has a simple pole at a point peit ∈ Δ, 0 < p < 1. Letting p → 1 this implies the inequality |an | ≤ n conjectured by Bieberbach. In fact, the Bieberbach conjecture is equivalent to the Goodman conjecture (1956) by an elegant proof of Jenkins [86]. Secondly, we shall consider the Sheil-Small conjecture in its full generality (see [152]), which deals with functions subordinate to f ∈ S. This little nuance becomes important in applications to the Schwarz-Pick type inequalities. Moreover, this consideration will contain the whole path from the Robertson conjecture to the Rogosinski conjecture as a special case. For the convenience of the reader who is not familiar with these inequalities, we will explain without technical details the relationships between the diﬀerent conjectures. The easiest step is the implication Robertson conjecture =⇒ Bieberbach conjecture.

2.2. Bieberbach’s conjecture by de Branges ˜ For f ∈ S let h(z) =

13

f (z 2 )/z 2 . Then the odd function

˜ h(z) = z h(z) =z+

∞

c2n−1 z 2n−1

n=2

˜ is an even function belongs to the family S√as well. If we take into account that h ˜ z) is holomorphic in Δ and that f (z) = z(s(z))2 . This we see that s(z) = h( identity implies that, for n ≥ 2, an =

n−1

c2k+1 c2(n−k)−1 ,

k=0

where c1 = 1. According to the Cauchy inequality, this yields |an | ≤

n−1

|c2k+1 |2 .

k=0

Hence, the above implication is obvious. In fact, in the present book, we will need a more general implication from the truth of the Robertson conjecture, namely, the theorem of Sheil-Small, see below. The implication Milin conjecture =⇒ Robertson conjecture follows from a theorem that is concerned with the exponentiation of holomorphic functions, the so-called Second Lebedev-Milin Inequality (see [99] and [116]). Let φ(z) =

∞

αk z k

k=1

be holomorphic in a neighbourhood of the origin and eφ(z) =

∞

βk z k .

k=0

Then for n ≥ 2 the inequalities n−1 1 |βk |2 ≤ exp n k=0

n−1 m 1 1 k|αk |2 − n m=1 k k=1

are valid. If one applies this theorem to the function φ(z) = log(s(z)) =

∞ k=1

γk z k ,

14

Chapter 2. Basic coeﬃcient inequalities

one sees that to prove the Robertson conjecture it is suﬃcient to prove that for n ∈ N the inequalities m n n 1 1 2 2 = (n − k + 1) ≤ 0 k|γk | − k|γk | − k k m=1 k=1

k=1

are valid. It has become customary to formulate this inequality in terms of the so-called logarithmic coeﬃcients of f . Since log(f (z)/z) = 2 log(s(z)) =

∞

2γk z k ,

k=1

many people know this conjecture in the form n 4 2 (n − k + 1) ≤ 0. k|2γk | − k

(2.5)

k=1

From this formulation de Branges found his way to prove the Milin conjecture (see [43], [64], and [71]). The ﬁrst item in this proof is the L¨ owner theory assuring ﬁrstly that, in problems like the above, it is suﬃcient to consider univalent functions that map the unit disc onto the complex plane minus a slit. The second ingredient from L¨owner’s theory is the fact that for any such function f there exists a chain of functions ∞ f (z, t) = et z + an (t)z n , z ∈ Δ, t ∈ [0, ∞), n=2

such that f (z, 0) = f (z) and the L¨ owner diﬀerential equation ∂ f (z, t) 1 + κ(t) z∂ f (z, t) = ∂t 1 − κ(t) ∂z is satisﬁed, where |κ(t)| = 1 and κ continuous on [0, ∞). Naturally, this diﬀerential equation results in diﬀerential equations for the logarithmic coeﬃcients cn (t) deﬁned by ∞ f (z, t) = cn (t)z n log et z n=1 with cn (0) = 2γn . The genial idea of de Branges was to look at (2.5) as an initial value problem. He constructed the so-called special function system of de Branges τn,k (t), 1 ≤ k ≤ n, n ∈ N, t ∈ [0, ∞), such that τn,k (0) = n − k + 1 considering the functions n 4 τn,k (t). k|ck (t)|2 − ϕn (t) = k k=1

2.3. Theorems of Jenkins and Sheil-Small

15

Using the L¨ owner diﬀerential equation he showed that for his system of functions the identity n τn,k (t) ϕn (t) = − |bk−1 (t) + bk (t) + 2|2 k k=1

is valid, where b0 (t) = 0, bk (t) =

k

jcj (t)κ(t)−j ,

k ∈ N.

j=1

The ingredient of the theory of special functions in de Branges’ proof was the proof that τn,k (t) ≤ 0 and limt→∞ τn,k (t) = 0. Hence, limt→∞ ϕn (t) = 0 and ϕn (t) ≥ 0. This in turn implies ϕn (0) ≤ 0 which is equivalent to (2.5).

2.3

Theorems of Jenkins and Sheil-Small

Let p =∈ (0, 1]. We need the expansion κp (z) =

∞ z

=z+ cn (p)z n , z (1 − pz) 1 − p n=2

|z| < p.

(2.6)

It is known that (see [86]) n−1 cn (p) =

j=0 p pn−1

2j

=

pn − p−

1 pn 1 p

.

(2.7)

In 1962, when the proof of the Bieberbach conjecture was a far-oﬀ dream, Jenkins proved the following theorem. Theorem 2.7 (Jenkins [86]). Let fp be a function meromorphic and univalent in the unit disc Δ with a simple pole at the point peit , t ∈ R, and 0 < p < 1, and let fp have the expansion ∞ an (p)z n (2.8) fp (z) = z + n=2

in a neighbourhood of the origin. If the Bieberbach conjecture is true for all coefﬁcients of schlicht functions, then |an (p)| ≤ cn (p) for any n ≥ 2. Equality occurs for the function eit κp (e−it z). Proof. Without loss of generality we suppose that t = 0. Let Δ(p) be the domain obtained from the unit disc by deleting the segment [p, 1]. Consider the class S(p) of all functions holomorphic and univalent in Δ(p) with expansion of the form

16

Chapter 2. Basic coeﬃcient inequalities

(2.8) in a neighbourhood of the origin. It is obvious that S(1) = S. To ﬁnd a conformal map of Δ(p) onto Δ we proceed as follows. It is clear that the function κ ˜ p (z) =

(1 + p)2 κp (z) 4p

maps Δ(p) conformally onto C \ (−∞, −1/4]. Now, we use the inverse K−1 of the Koebe function k−1 (z) = z/(1 − z)2 to map this domain onto Δ. Therefore, the function ϕ = K−1 ◦ κ ˜ p has the desired property. If we consider the expansion ϕ(z) =

∞ (1 + p)2 cn z n , z + 4p n=2

it is important to recognize that cn > 0 for all n. To see this, we compute ϕ explicitly by the above procedure to get ϕ(z) = 1 +

2p(1 − z/p)(1 − zp) 2p(1 + z)(1 − z/p)1/2 (1 − zp)1/2 − . 2 (1 + p) z (1 + p)2 z

It is clear that cn , n ≥ 2, is the sum of two consecutive coeﬃcients of the function (1 − z/p)1/2 (1 − zp)1/2 multiplied by the factor −2p/(1 + p)2 . The fact that these coeﬃcients themselves are negative for p ∈ (0, 1) is easily seen using the generalized binomial expansion and the Cauchy product for the product (1−z/p)1/2 (1−zp)1/2 . The inverse to the function ϕ is frequently used in extremal problems for bounded functions holomorphic and univalent in the unit disc (see for instance [133]). Clearly, any function fp ∈ S(p) admits a representation of the form fp (z) =

4p f (ϕ(z)), (1 + p)2

where f (z) = z +

∞

an z n ,

z ∈ Δ(p),

z ∈ Δ,

n=2

is a function in S. The function fp (z) = 4p(1 + p)−2 f (ϕ(z)) has the expansion of the form (2.8) with n an (p) = λ1 (n) + λj (n)aj , j=2

where λj (n) are polynomials in cm = ϕ(m) (0)/m! , m = 2, . . . , n, with non-negative coeﬃcients, and thus they are themselves non-negative. In particular, if aj = j, i.e. if f (z) = κ1 (z), then an (p) = cn (p).

2.3. Theorems of Jenkins and Sheil-Small

17

The validity of the Bieberbach conjecture implies that |an (p)| ≤ λ1 (n) + ≤ λ1 (n) +

n j=2 n

λj (n)|aj | jλj (n) = cn (p).

j=2

This completes the proof of Theorem 2.7.

Remark 2.8. As is shown in [18], the domain of variablity of coeﬃcients an (p) is only a proper subset of the set {w | |w| ≤ cn (p)}, if the pole at p lies near to the origin. The formulation of the Sheil-Small result in its full generality that we have in mind and which can be found, at least implicitly, in [152] and [71] is the following. Theorem 2.9 (Sheil-Small [152] (1973)). Let g be subordinated to a function f ∈ S and P a polynomial of degree less than or equal to n. If the Robertson conjecture is true, then max |(P ∗ g)(z)| ≤ n max |P (z) =: n M (P ). (2.9) |z|≤1

|z|=1

First we consider the following lemma. Lemma 2.10 (Sheil-Small [152]). Let U (z) =

∞

uk z k

and

V (z) =

k=0

∞

vk z k

k=0

be holomorphic in Δ and wi : Δ → Δ, i = 1, 2, 3, be holomorphic in Δ. If P is a polynomial of degree less than or equal to n and ˜ h(z) = zw1 (z)U (zw2 (z))V (zw3 (z)),

z ∈ Δ,

then we have iθ ˜ )| ≤ r max |P (z)| |(P ∗ h)(re

n−1

|z|=1

k=0

2 2k

|uk | r

1/2 n−1

1/2 2 2k

|vk | r

k=0

for r ∈ (0, 1), θ ∈ [0, 2π]. Proof. Let the abbreviations be as above. We use the representations iθ ˜ (P ∗ h)(re ) =

1 2π

0

2π

˜ i(θ+ϕ) ) d ϕ. P (ei(θ−ϕ) )h(re

(2.10)

18

Chapter 2. Basic coeﬃcient inequalities

and ˜ (P ∗ h)(z) = P (z) ∗

zw1 (z)

n−1

k

uk (zw2 (z))

k=0

n−1

k

vk (zw3 (z))

.

k=0

From this, we obtain iθ ˜ )| |(P ∗ h)(re

n−1 n−1 iϕ iϕ k iϕ iϕ k |w1 (re )| uk (re w2 (re )) vk (re w3 (re )) d ϕ 0 k=0 k=0 ⎞ 1/2 2 2π n−1 1 iϕ iϕ k ≤ r M (P ) ⎝ uk (re w2 (re )) d ϕ⎠ 2π 0

1 ≤ r M (P ) 2π ⎛

⎛

2π

iϕ

k=0

⎞1/2 2 2π n−1 1 ·⎝ vk (reiϕ w2 (reiϕ ))k d ϕ⎠ . 2π 0 k=0

Now, the rest of the proof is an immediate consequence of Littlewood’s Theorem 2.1 and Rogosinski’s Theorem 2.3. Proof of Theorem 2.9 of Sheil-Small. We take g ≺ f ∈ S, g(z) = f (zω(z)), and s as above. Then we may write g(z) = zω(z)s(zω(z))2 . Taking wi = ω, i = 1, 2, 3, U = V = s in Lemma 2.10 and using the maximum principle we see that the Robertson conjecture is the decisive step needed in the proof of Theorem 2.9. Remark 2.11. It is evident that the Rogosinski conjecture follows from the SheilSmall inequality (2.9) for P (z) = z n and that the Bieberbach conjecture is the case ω ≡ 1 of the Rogosinski conjecture. Remark 2.12. The requirement of Lemma 2.10 about the functions U and V can be considerably relaxed. Namely, it is suﬃcient to suppose that the functions U and V are holomorphic in a neighbourhood of the origin, only. For such a case the proof is the same except the ﬁnal step: Instead of Rogosinski’s Theorem 2.3 we can apply Theorem 2.4.

2.4

Inverse coeﬃcients

Let Δ be the unit disc {z| |z| < 1} in the complex plane C and let S be the usual class of functions holomorphic and univalent in Δ with expansion about the origin, f (z) = z +

∞ n=2

an z n .

2.4. Inverse coeﬃcients

19

We will frequently use the following classical theorem. Theorem 2.13 (K. L¨ owner [110]). If F is the inverse of a function in S and has the expansion ∞ F (w) = w + An wn n=2

in a neighbourhood of the origin, then (2n)! 1 |An | ≤ = n!(n + 1)! n

2n n−1

(2.11)

with equality only for the inverses of the Koebe functions kd (z) =

z , |d| = 1. (1 + dz)2

(2.12)

By mathematical induction and the Stirling formula one easily gets that 1 4n (2n − 1)!! 4n 2n = ≤ n−1 n n + 1 (2n)!! (n + 1)3/2 and that 4n (2n − 1)!! 4n = n + 1 (2n)!! (n + 1)3/2

1 √ + O(1/n) π

as n → ∞.

Here, for n ∈ N we use the abbreviations (2n − 1)!! =

(2n − 1)! 2n−1 ((n − 1)!)

and (2n)!! =

(2n)! . (2n − 1)!!

The classical proof of this L¨ owner theorem can be found in [78], [70], [128] and [147]. Here we present two generalizations. Each of them implies the theorem. Let K1 be the inverse of the Koebe function k1 . We have K1 (w) = w + and

∞

(2n)! wn , n!(n + 1)! n=2

∞ 1 S(w, K1 ) := (K1 /K1 ) − (K1 /K1 )2 = 4n 6(n + 1)wn , 2 n=0

and log K1 (w) =

∞ n=1

bn w n ,

20

Chapter 2. Basic coeﬃcient inequalities

where 1 bn = n

2n n

2n−1

+2

4n = n

1 (2n − 1)!! + 2 (2n)!!

4n . n

≤

(2.13)

Moreover, one easily gets bn =

√ 4n−1/2 1 + O(1/ n) n

as n → ∞.

Since the method to ﬁnd the expression (2.13) is behind many of the coeﬃcient results of this book, we will give a short proof for (2.13). We use the Cauchy integral formula in the following way. Let ∞ K1 (w) (n + 1)bn+1 wn , = K1 (w) n=0

then (n + 1)bn+1 = =

1 2πi 1 2πi

1 = 2πi

K1 (w) 1 dw K1 (w) wn+1

k1 (∂Δr ) K1 (k1 (z)) ∂Δr K1 (k1 (z))

∂Δr

1 k1 (z) dz = − n+1 2πi k1 (z) (4 − 2z)(1 + z)2n+1 dz, (1 − z)z n+1

∂Δr

k1 (z) 1 dz k1 (z) k1n+1 (z)

where r ∈ (0, 1). Hence, we have to ﬁnd the nth Taylor coeﬃcient of ∞ ∞ 2n+1 k 2n+1 k 4+2 (4 − 2z)(1 + z) z = (1 + z) z . k=0

k=1

We get from this n−1 2n + 1 2n + 1 (n + 1)bn+1 = 4 +2 k n k=0 2n + 1 =2 + 22n . n

This implies bn =

1 n

1 2n − 1 2n 2 + 22n−1 = + 22n−1 . n−1 n n

Consider the following expansions for F , the inverse of a function in S, ∞ 1 S(w, F ) := (F /F ) − (F /F )2 = Cn wn 2 n=0

2.4. Inverse coeﬃcients

21

and log F (w) =

∞

Bn wn .

n=1

Theorem 2.14 (Klouth and Wirths [91]). If F is the inverse of a function in S, then |Bn | ≤ bn for all n ≥ 1 and |Cn | ≤ 4n 6(n + 1) for all n ≥ 0. Equality for n ≥ 2 occurs only for the functions Kd (w) = kd−1 (z), |d| = 1. Since bn are positive and each An is a polynomial with positive coeﬃcients in the Bn , Theorem 2.14 implies Theorem 2.13. Proof. The proof follows the same line as the classical proof of the Theorem 2.13. We only give here the crucial steps. A function f in S can be embedded into a subordination chain (for details see [128]). It results that the inverse function F has a representation F (w) = lim Φ(e−t w, t), t→∞

∂Φ(w, t)/∂t = w(∂Φ(w, t)/∂w)p(w, t)

where Φ(w, 0) = 0, Re p(w, t) > 0 and p(w, t) = 1 +

∞

pn (t)wn

n=1

for w ∈ Δ and t ≥ 0. Using these and setting L(w, t) := log

∞ ∂Φ(w, t) Bn (t)wn , = ∂w n=0

M (w, t) := S(w, Φ(w, t)) =

∞

Cn (t)wn ,

n=0

we get ∂ 3 (wp) ∂L ∂L ∂(wp) ∂S ∂S ∂(wp) + , = wp + , = wp + 2S w, ∂t ∂w ∂w ∂t ∂w ∂w ∂w3 and, for n ≥ 1, B0 (t) = t, Bn (t) = Cn (t) =

0

⎛ t

n−1

en(t−τ ) ⎝

0

⎞ jBj (τ )pn−j (τ ) + (n + 1)pn (τ )⎠ dτ,

j=1

⎛

t

⎞ (n + 3)! e(n−2)(t−τ ) ⎝ jCj (τ )pn−j (τ )(2n − j + 2) + pn+2 (τ )⎠ dτ. n! j=1 n−1

These formulas show that Re Bn (t), respectively Re Cn (t), is maximal for a ﬁxed t if and only if we choose Bn (t), j = 1, . . . , n − 1, respectively Cn (t), j = 0, . . . , n − 1,

22

Chapter 2. Basic coeﬃcient inequalities

for τ ∈ [0, t] real and maximal and pj (τ ) = 2 in [0, t] for any index j involved in the formulas in question. Since Bn = lim e−nt Bn (t), Cn = lim e−(n+2)t Cn (t), t→∞

t→∞

we get that the maximum of Re Bn , respectively Re Cn , is attained if and only if p(w, t) = (1 + w)/(1 − w). Clearly, the assertion of the theorem for n ≥ 1 follows from the known fact that the problems of ﬁnding the maximum of the real part and the maximum of the modulus for the given coeﬃcient are equivalent. To complete the proof we remark that the desired inequality for C0 = −6(a3 − a22 ) is given by the classical inequality |a3 − a22 | ≤ 1. For a ﬁxed p ∈ (0, 1), let Sp denote the class of all meromorphic univalent functions fp in the unit disc Δ with the normalization fp (0) = fp (0) − 1 = 0 and fp (p) = ∞. If Fp is the inverse of a function in Sp , then it admits an expansion of the form ∞ Fp (w) = w + An (p)wn n=2

in a neighbourhood of the origin. In Sp the function z

κp (z) = (1 − pz) 1 − pz plays the role of the Koebe function in S. Near w = 0 the function Kp (w) = kp−1 (w) has the expansion (see, for instance, [33]) Kp (w) = w +

∞

An (p, Kp )wn

n=2

with An (p, Kp ) =

n−1 (−1)n−1 n n p2j−n+1 . j j+1 n j=0

Theorem 2.15 (Baernstein and Schober [33])). The coeﬃcients of the function Fp satisfy the sharp inequalities |An (p)| ≤ An (p, Kp ). Equality for a single coeﬃcient holds only if Fp = Kp . Clearly, for p = 1 Theorem 2.15 gives the L¨ owner theorem. The proof of Theorem 2.15 (see [33]) is based on the following integral inequality of Baernstein [32]: for any fp ∈ Sp , 2π 2π 1 1 iθ α |fp (re )| dθ ≤ |κp (reiθ )|α dθ, 2π 0 2π 0 whenever 0 < r < 1 and −∞ < α < ∞.

2.5. Domains with bounded boundary rotation

2.5

23

Domains with bounded boundary rotation

Plane domains with bounded boundary rotation were ﬁrst studied by Paatero ([122], see also [10]). In Radon’s paper [134] one can ﬁnd general properties and applications of domains whose boundaries are rectiﬁable curves with bounded rotation. In the following we need the usual abbreviation: If the functions f and g are analytic in a neighbourhood of the origin, then f (z) =

∞

ak z k g(z) =

k=0

∞

bk z k

k=0

if and only if for all k ∈ N ∪ {0} the inequalities |ak | ≤ |bk | are valid. In [100], Lehto considered the family Vk of functions f ∈ S such that the boundary rotation of f (Δ) is at most kπ ≥ 2π. It is well known that the function k/2 ∞ 1+z 1 fk (z) = −1 =z+ An (k)z n k 1−z n=2 belongs to Vk . ∞ For any f ∈ Vk with expansion f (z) = z + n=2 an z n , it is proved that the inequality |an | ≤ An (k) is valid for all n ≥ 2 (see [45], [2] and [44] and [146] ). The crucial facts are gathered in the following two theorems. Theorem 2.16 ([45]). Let ϕ be a function holomorphic in Δ. If α ≥ 1 and ϕ(z) ≺

1 + cz 1−z

for some constant c ∈ Δ, then there exists a probability measure μ : [0, 2π] → R such that α 2π 1 + c eit z α ϕ(z) = dμ(t), z ∈ Δ. 1 − eit z 0 Theorem 2.17. For any c ∈ Δ and α ≥ 1 α α 1 + cz 1+z . 1−z 1−z Consequently, if α ≥ 1 and ϕ(z) ≺

1 + cz 1−z

24

Chapter 2. Basic coeﬃcient inequalities

for some constant c ∈ Δ, then ϕ(z) α

1+z 1−z

α .

Short proofs of these facts are found in [44] and [146]. The original proofs are given in [45] and [2]. In fact, it was shown in [45], that Theorem 2.17 implies the coeﬃcient estimate for the larger class of close-to-convex functions of order k/2 − 1 ≥ 0 introduced by Ch. Pommerenke [127]. Using Theorem 2.17 and the cited proofs in [44] and [146] we get Theorem 2.18 ([17]). If c ∈ Δ \ {−1} and α ≥ 1, then 1 c+1

1 + cz 1−z

α

−1

1 2

1+z 1−z

α

−1 .

Proof. For c ∈ Δ \ {−1} let Tc (z) = (1 + cz)/(1 − z). Since Tc (Δ) is a halfplane whose boundary cuts the real axis in a point of the interval [0, 1) and 1 ∈ Tc (Δ), there exists, for any c ∈ Δ \ {−1} and α ≥ 1, a c1 ∈ Δ such that 1

1

ϕ1 (z) := Tc (z)1− α T1 (z) α ≺

1 + c1 z . 1−z

According to Theorem 2.17 this implies ϕ1 (z)α T1 (z)α . Using this and the nonnegativity of the Taylor coeﬃcients of the functions T1 (z)α and 1/(1 − z 2 ) we get

1 + cz 1−z

α−1

1 1 = ϕ1 (z)α 2 (1 − z) 1 − z2

1+z 1−z

By integration we obtain the assertion of Theorem 2.18.

α

1 . 1 − z2

Now, we prove a subordination theorem which we need for the applications. We are concerned with the class of angular domains Πα = aHα + b (a = 0) with opening angle απ, 1 ≤ α ≤ 2, which means that there exists a linear transformation T (z) = az + b such that Πα = T (Hα ), where απ Hα = z | | arg z| < . 2 Clearly, the assertion of the following theorem is a generalization of the Carath´eodory inequality for Taylor coeﬃcients of holomorphic functions with positive real part (see [51], [52], compare also [42], p. 365).

2.5. Domains with bounded boundary rotation

25

Theorem 2.19 ([17]). For any angular domain Πα , 1 ≤ α ≤ 2, and any function g holomorphic in Δ with g(Δ) ⊂ Πα , the assertion α 1+ζ 1 (g(ζ) − g(0))λΠα (g(0)) − 1 , ζ ∈ Δ, 2α 1−ζ is valid. Proof. Let

Gα (ζ) := ΦΠα ,g(0) (ζ) − g(0) λΠα (g(0)),

ζ ∈ Δ.

The function Gα belongs to the class S and maps Δ univalently onto an angular domain Πα . This yields the existence of a complex number c ∈ Δ \ {−1} such that α 1 + cζ 1 Gα (ζ) = − 1 , ζ ∈ Δ. (c + 1)α 1−ζ Therefore, it follows from our earlier assumption that (g(ζ) − g(0))λΠα (g(0)) ≺ Gα (ζ),

ζ ∈ Δ,

and in turn that 1

φ(ζ) := (1 + (c + 1)α(g(ζ) − g(0))λΠα (g(0))) α ≺

1 + cζ , 1−ζ

ζ ∈ Δ.

According to Theorem 2.16 this implies that α 2π 1 + c eit ζ 1 − 1 dμ(t), (g(ζ) − g(0))λΠα (g(0) = (c + 1)α 0 1 − eit ζ This together with Theorem 2.18 yields the result of Theorem 2.19. Theorem 2.20 ([17]). For 1 ≤ α ≤ 2 let α 1−z 1 1− hα (z) := , 2α 1+z

ζ ∈ Δ.

z ∈ Δ.

Let further h ∈ S and h(Δ) be an angular domain Πα . We denote by h−1 the −1 function inverse to h and by h−1 (w) and α the function inverse to hα . If we let h −1 −1 −1 hα (w) represent the Taylor expansions of h and hα in a neighbourhood of the origin, then h−1 (w) h−1 α (w). Proof. Since there exists a complex number c ∈ Δ \ {−1} such that any function h of the above type may be written in the form α 1 1−z h(z) = 1− , z ∈ Δ, (c + 1)α 1 + cz

26

Chapter 2. Basic coeﬃcient inequalities

we get by a straightforward computation that 1

1 − (1 − (c + 1)α w) α

h−1 (w) =

1

1 + c(1 − (c + 1)α w) α

.

Now, we use the expansion (1 − (c + 1)α w)

1 α

∞

=

k

(−α)

1 α

(c + 1)k wk

k

k=0

and the fact that, for k ≥ 1 and α ≥ 1, Dk (α) := −

(−α)k (c + 1)k

1 α

k

≥ 0.

Hence, h

−1

(w) =

∞

k−1

Dk (α)(c + 1)

k=1

∞

w

k

k−1

Dk (α)2

w

k

1−

k=1

This completes the proof of Theorem 2.20.

1−c

∞

−1 k−1

Dk (α)(c + 1)

k=1 ∞

w

k

−1 k−1

Dk (α)2

w

k

= h−1 α (w).

k=1

Chapter 3

The Poincar´e metric For more than two thousand years, mathematicians and other people believed in the “truth” of the Euclidean parallel axiom, before Lobachevsky and Bolyai about 1825–1830 found that it is possible to get a new geometry by substituting this axiom with the following. Through a point in the plane not lying on the given straight line one can draw more than one straight line not intersecting the given line. At the beginning, the existence of such a geometry was not accepted by mathematicians, except for C. F. Gauss. In 1868 Beltrami interpreted the Lobachevsky (or hyperbolic) geometry as the natural geometry on a surface of constant negative curvature and thus proved that there exists a model for such a geometry. During this work, he was able to use the theorems of Gauss (1827) and Milding (1840) on surface geometry. Some years later, in 1871, Klein made this geometry “visible” to geometers by an interpretation in projective geometry. In accordance with his model, Klein introduced the term “hyperbolic geometry”. In 1882 Poincar´e discovered the following interpretation: the unit disc Δ equipped with the conformally invariant metric λΔ (z)|dz| :=

|dz| 1 − |z|2

can be regarded as the hyperbolic plane.

3.1

Background

Since any conformal automorphism of the unit disc Δ has the form z = T (ζ) := eiα

ζ −a , 1 − aζ

a ∈ Δ, α ∈ R,

28

Chapter 3. The Poincar´e metric

by straightforward computations one gets that the Poincar´e metric is conformally invariant in Δ, that is |dζ| |dz| = (3.1) 2 1 − |z| 1 − |ζ|2 for all z = T (ζ) in Δ. The hyperbolic distance between two points z1 , z2 ∈ Δ is deﬁned by λΔ (z) |d z|, DΔ (z1 , z2 ) = inf

(3.2)

Γ

where the inﬁmum is taken over all piecewise smooth curves Γ ⊂ Δ joining z1 with z2 . The geodesics (or paths of shortest distance) consist of the images of diameters of Δ under conformal automorphisms T . These are the diameters of Δ and the circular arcs in Δ orthogonal to its boundary ∂Δ. If these arcs are called “straightlines”, one has a model of the hyperbolic plane. Using the “straightlines”, one can ﬁnd that the hyperbolic distance for z1 , z2 ∈ Δ is given in the form z1 − z2 1 1+ρ . (3.3) , where ρ = DΔ (z1 , z2 ) = log 2 1−ρ 1 − z1 z2 An alternative interpretation of the hyperbolic geometry is given by the half plane model of Poincar´e. In the half plane H1 = {z ∈ C| Re z > 0} the element of hyperbolic arc length is λH1 (z)|dz| =

|dz| , z = x + iy ∈ H1 . 2x

In this case geodesics are horizontal lines and circles orthogonal to the imaginary axis. Using the conformal map ζ = (1 − z)/(1 + z) of the unit disc onto the half plane, it is easily seen that these two models are equivalent. The equivalence is given by the simple equation λH1 (ζ)|dζ| = λΔ (z)|dz|,

ζ=

1−z . 1+z

Now, let Ω denote a domain in C with three or more boundary points in C. According to the Riemann mapping theorem, if Ω is simply connected and z0 ∈ Ω \ {∞} is ﬁxed, then there exists a unique conformal map f0 of Δ onto Ω such that f0 (0) = z0 and f0 (0) > 0. This quantity f0 (0) is called the conformal radius of Ω at the point z0 and will be denoted in the following by R(z0 , Ω). If Ω is as above but not simply connected, the generalization of Riemann’s mapping theorem due to Poincar´e (see, for instance, [3] and [70], p. 255) asserts that there exists a unique universal covering map f0 of

3.1. Background

29

Δ onto Ω which has the same normalization as the above conformal map. In this case the quantity f0 (0) is called the hyperbolic radius (see, for instance, [34]). The Poincar´e (or hyperbolic) metric in Ω is deﬁned by the equation λΩ (z)|dz| :=

|dζ| , 1 − |ζ|2

z = f0 (ζ) ∈ Ω.

(3.4)

Concerning the universal covering map f0 for a multiply connected domain, we have to mention that f0 is a holomorphic or meromorphic function, while its inverse f0−1 (z) is multivalued. But it is well known that any conformal or universal covering map f of Δ onto Ω is given by iα ζ − a , a ∈ Δ, α ∈ R. f (ζ) = f0 (T (ζ)) = f0 e 1 − aζ This together with (3.1) imply that the density λΩ (z) is well-deﬁned by equation (3.4), i.e., it depends neither on the choice of the mapping function nor on the choice of the branch of f −1 . Further, the metric λΩ (z)|dz| is conformally invariant. From the above deﬁnitions it follows that R(z, Ω) =

1 = |f (ζ)|(1 − |ζ|2 ), λΩ (z)

(3.5)

where f is a conformal or universal covering map of Δ onto Ω and z = f (ζ), ζ ∈ Δ, z ∈ Ω. Taking partial derivatives of R by using (3.5) and the Wirtinger calculus 2

2

∂ R(z, Ω) ∂ R(z, Ω) ∂ R(z, Ω) = −i , z = x + iy ∈ Ω, ∂z ∂x ∂y

∂ R(z, Ω) ∂ R(z, Ω) ∂ R(z, Ω) = +i =: ∇R(z, Ω), z = x + iy ∈ Ω, ∂z ∂x ∂y

leads to the formulas ∂R(z, Ω) |f (ζ)| = ∂z f (ζ) and R(z, Ω)

1 − |ζ|2 f (ζ) −ζ 2 f (ζ)

(3.6)

2 ∂ 2 R(z, Ω) ∂ R(z, Ω) = − 1. ∂z∂z ∂z

Hence, the conformal (or hyperbolic) radius R = R(z, Ω) satisﬁes Liouville’s equation RR = |∇R|2 − 4, (3.7) where R = R(z, Ω), R denotes the Laplacian of R and ∇R its gradient.

30

Chapter 3. The Poincar´e metric

The most known forms of the equation (3.7) are the following nonlinear elliptic equation u = −4e−2u , where u = u(x, y) := − log λΩ (z) = log R(z, Ω) for z = x + iy ∈ Ω, and an equivalent formula K = e2u u = −4 that deﬁnes the Gaussian curvature of the metric. Thus, the Liouville equation for R is equivalent to the fact that the hyperbolic metric deﬁned as above has the Gaussian curvature K = −4 (see, for instance, [3], [34] and [70]). Remark 3.1. Clearly, to deal with the hyperbolic metric with Gaussian curvature K = −1, one has to choose the density of the metric as 2λΩ (z), where λΩ (z) is deﬁned as above.

3.2

The Schwarz-Pick inequality

We begin by the classic Schwarz lemma. Theorem 3.2 (Schwarz’s lemma). Let f be holomorphic in the unit disc Δ. If f (0) = 0 and |f (z)| is bounded by 1 in Δ, then |f (z)| ≤ |z|,

z ∈ Δ,

(3.8)

and |f (0)| ≤ 1

(3.9)

with equality in (3.8) for some z = 0 or in (3.9) occurring if and only if f (z) = cz for some unimodular complex constant c. As is indicated in [3], p. 21, the Schwarz lemma and its classic proof are due to Carath´eodory [53]; Schwarz proved it only for univalent mappings [149]. In [124], [125] Pick realized the invariant character of Schwarz’s lemma. Theorem 3.3 (Pick [125], see also Lindel¨ of [105]). A holomorphic mapping f of the unit disc Δ into itself satisﬁes the inequalities f (z ) − f (z ) z − z 1 2 1 2 (3.10) , z1 , z2 ∈ Δ, ≤ 1 − f (z1 )f (z2 ) 1 − z1 z2 and

|f (z)| 1 ≤ , 1 − |f (z)|2 1 − |z|2

z ∈ Δ.

(3.11)

Nontrivial equality holds if and only if f is a conformal automorphism of Δ, i.e., f (z) = eiα (z − a)/(1 − az), a ∈ Δ, α ∈ R.

3.2. The Schwarz-Pick inequality

31

It is evident that inequalities (3.10) and (3.11) correspond to Schwarz’s inequalities (3.8) and (3.9), respectively. Geometrically, inequality (3.10) means that any holomorphic mapping f of the unit disc Δ into itself decreases the hyperbolic distance between two points. Also, (3.11) implies that f decreases the hyperbolic lengths of an arc and the hyperbolic area of a set. The following assertion is called the hyperbolic metric principle (see, for instance, [70] and [120]). Theorem 3.4. Let Ω and Π be domains in the extended complex plane such that each of them has at least three boundary points. If f ∈ A(Ω, Π), then (i) L(f (γ)) ≤ L(γ) for any rectiﬁable curve γ ⊂ Ω and its image f (γ) ⊂ Π with length L(γ) and L(f (γ)) in the hyperbolic metric of Ω and Π, respectively; equality holds if and only if f (z) = f2 (f1−1 (z)), where f1 and f2 are conformal (universal covering) maps of the unit disc onto Ω and Π, respectively. (ii) the diﬀerential length elements at z ∈ Ω and at its image w = f (z) ∈ Π satisfy the inequality λΠ (w)|dw| ≤ λΩ (z)|dz| (3.12) with the same condition for equality. We shall call the Schwarz-Pick inequality the equation (3.12) written in the form |f (z)| ≤

λΩ (z) . λΠ (f (z))

(3.13)

The Schwarz-Pick inequality applied to the function f ∈ A(Ω , Ω) deﬁned by ζ = f (z) = z, z ∈ Ω , immediately gives the following comparison of densities. Theorem 3.5. If Ω ⊂ Ω, then λΩ (z) ≥ λΩ (z) for any point z ∈ Ω . Explicit formulas for the Poincar´e density are known for several domains. We shall present three of them. They will be useful in applications of the comparison theorem 3.5. 1. For angular domains Hα = {z ∈ C \ {0} | |arg z| < easily gets 1/λHα (z) = R(z, Hα ) = 2 α r cos

απ 2 },

α ∈ (0, 2], one

θ , z = reiθ ∈ Hα , α

and

4 πy cos , z = x + iy, |y| < 1 π 2 for the strip H0 = {z ∈ C | |Im z| < 1}. 1/λH0 (z) = R(z, H0 ) =

2. For the punctured unit disc Δ = {z ∈ C| 0 < |z| < 1}, 1/λΔ (z) = R(z, Δ ) = 2|z| log

1 , z ∈ Δ . |z|

32

Chapter 3. The Poincar´e metric

3. It is a much harder problem to ﬁnd an explicit formula for the density of the domain C \ {0, 1} = C \ {0, 1, ∞}. The following formula is due to Agard [1]: |z||z − 1| 1 dudv = , w = u + iv. λC\{0,1} (z) π C |w||w − 1||w − z| The classical representation formula for λC\{0,1} is given by λC\{0,1} (z) =

1 |g (z)| = , 2|λ (τ )| Im τ 2 Img(z)

where λ(τ ) is the elliptic modular function, g is its inverse, τ ∈ Λ := {τ | Imτ > 0} and z ∈ C \ {0, 1}. The modular function λ : Λ → C \ {0, 1} is a universal covering map which maps the triangle {τ ∈ Λ | 0 < Reτ < 1, |τ − 1/2| > 1/2} onto Λ in such a way that λ(0) = 1, λ(1) = ∞, λ(∞) = 0, and 8 ∞ 1 + q 2n λ(τ ) = 16 q , q = eπiτ (3.14) 2n−1 1 + q n=1 (see, for instance, [107]). The Liouville equation for the hyperbolic density λC\{0,1} (z) and formula (3.14) are used in a proof of some monotonicity properties of the metric. Theorem 3.6 (see A. Bermant [39], J. A. Hempel [80] and compare also A. Yu. Solynin and M. Vuorinen [154] for further results). Let z = x + iy = reiθ . The Poincar´e density λC\{0,1} (z) has the following properties: y

∂λC\{0,1} (z) < 0, ∂y

y = 0,

(3.15)

∂λC\{0,1} (z) < 0, 0 < |θ| < π, (3.16) ∂θ ∂λC\{0,1} (z) λC\{0,1} (z) log r < 0, r = 1. (3.17) + ∂r r There are many generalizations and applications of the classical Theorems 3.2 and 3.3. Especially, there are close relations between Theorems 3.2 and 3.3 in the theory of unimodular bounded holomorphic functions due to Carath´eodory, Denjoy, Julia and Wolﬀ. We cite the following statements, which sometimes are quoted as the Grand Iteration Theorem (see [150], p. 78 and compare also [131], p. 82, and [55]). θ

Theorem 3.7. Let ϕ be a holomorphic self-map of Δ that is not a conformal automorphism of Δ with a ﬁxed point in Δ. Then there is a unique point ω ∈ Δ such that the iterates ϕn (z) = (ϕ ◦ · · · ◦ ϕ)(z) converge to ω as n → ∞ uniformly on any compact subset of Δ and

3.3. Estimates using the Euclidean distance

33

(C1 ) if ω ∈ Δ, then ϕ has no ﬁxed point in Δ \ {ω}, and ϕ(ω) = ω, 0 ≤ |ϕ (ω)| < 1, (C2 ) if ω ∈ ∂Δ (Denjoy-Wolﬀ point of ϕ), then ϕ has no ﬁxed point in Δ and ϕ and ϕ have angular limits at ω such that ϕ(ω) = ω, 0 < ϕ (ω) ≤ 1, where

ω+z ϕ (ω) = sup Re ω−z z∈Δ

−1 ω + ϕ(z) Re . ω − ϕ(z)

For further deep results on the boundary behavior of bounded holomorphic functions we refer the reader to the books by L. V. Ahlfors [3], by J. B. Garnett [66] and by Ch. Pommerenke [131] (see also R. B. Burckel [48] and G. M. Goluzin [70] for other developments and applications of Theorems 3.2 and 3.3.)

3.3

Estimates using the Euclidean distance

Choosing Ω as the disc with center z and radius δ(z) = dist(z, ∂Ω) in Theorem 3.5, one gets Theorem 3.8. If δ(z) denotes the distance from z ∈ Ω to the boundary ∂Ω of Ω, then λΩ (z) ≤ 1/δ(z) for any z ∈ Ω. The next assertion is known as the Koebe 1/4-theorem (for a proof see [60], [70], [78] or [128]). Theorem 3.9. If Ω is a simply connected domain in C, then δ(z) λΩ (z) ≥ any z ∈ Ω.

1 4

for

We will need Landau’s theorem on holomorphic functions that omit two ﬁxed values. More precisely, the Landau theorem concerns functions f (z) =

∞

an z n

n=0

holomorphic in the unit disc Δ and omitting the values 0 and 1 in Δ. By our notation, f ∈ A(Δ, C \ {0, 1}). In the sequel, the known constant 1 2λC\{0,1} (−1)

= iλ (1 + i) =

Γ(1/4)4 = 4.3768796 . . . 4π 2

is used. J. A. Hempel and J. A. Jenkins established an explicit sharp bound in the Landau theorem.

34

Chapter 3. The Poincar´e metric

Theorem 3.10 (Landau; see J. A. Hempel [80] and J. A. Jenkins [87] for proofs). If the function f is holomorphic and omits 0 and 1 in Δ, then Γ(1/4)4 |a1 | ≤ 2|a0 | |log |a0 || + . (3.18) 4π 2 Equality holds only for the universal covering map f : Δ → C\{0, 1} with a0 = −1. Proof by J. A. Hempel [80]. By the hyperbolic metric principle we have the sharp inequality |a1 | = |f (0)| ≤ 2/ρ(a0 ), ρ(z) := 2λC\{0,1} (z). Thus, Theorem 3.10 states that 1

2 = ≤ 2|z| λC\{0,1} (z) ρ(z)

1 |log |z|| + ρ(−1)

.

(3.19)

According to formula (3.15) of Theorem 3.6, to prove inequality (3.19) it is suﬃcient to consider real values of z = x + iy = reiθ lying in the interval (−∞, 0). Let us introduce the real function w(σ) := u(eσ+iπ ) + σ,

σ := log r, u := log ρ.

Equation (3.19) is equivalent to e−w(σ) ≤ |σ| + e−w(0) .

(3.20)

From (3.17) of Theorem 3.6 it follows that σw (σ) < 0 for σ = 0. Again from (3.16) of Theorem 3.6 we deduce that ∂ 2 u/∂θ2 ≥ 0 for θ = π. This together with Liouville’s equation ∂2u ∂2u + 2 = e2(σ+u) ∂σ 2 ∂θ give ∂ 2 u/∂σ 2 ≤ e2(σ+u) , θ = π, which is equivalent to the inequality w ≤ e2w . Integrating this inequality in (−∞, 0] and [0, ∞) using the local behavior w → 0 and e2w = r2 ρ2 (z) → 0 as σ → ∞ (see [80] for details), we have (3.20). This completes the proof of Theorem 3.10. Remark 3.11. Since the function 1 − f (z) = 1 − a0 − a1 z −

∞ n=2

an z n

3.3. Estimates using the Euclidean distance

35

also omits 0 and 1, we can take 1 − a0 in (3.18) instead of a0 . Consequently, Theorem 3.10 implies Γ(1/4)4 |a1 | ≤ . min 2|ζ| |log |ζ|| + 4π 2 ζ∈{a0 ,1−a0 } We shall consider an important application of Landau’s theorem in the theory of uniformly perfect sets. Let Ω ⊂ C be an open set with more than one boundary point in C. The annulus An(r, s; a) = {z | r < |z − a| < s} ⊂ Ω with center a is said to separate the compact set E =C\Ω in the Riemann sphere, whenever both components of C \ An(r, s; a) have nonempty intersections with E. Since in the next section we are concerned with conformal images of such annuli, called conformal annuli, we will characterise annuli of the above form as genuine annuli. Following Ch. Pommerenke (see [129] and also [36], [37], [55], [65], [130]), we deﬁne the maximum modulus M0 (Ω) as the supremum of the moduli of all annuli An ⊂ Ω such that An separates E and the center a of An belongs to ∂Ω. As usual the modulus of an annulus An is deﬁned by M (An(r, s; a)) =

1 s log . 2π r

We take M0 (Ω) = 0, if Ω contains no genuine annulus centered at a boundary point and separating E. Theorem 3.12 (Beardon and Pommerenke, compare [37] and [129]). Let Ω ⊂ C be an open set with more than one boundary point in C. If λΩ (z) is the density of the Poincar´e metric deﬁned in each component of Ω with curvature K = −4 and M0 (Ω) is ﬁnite, then for any z ∈ Ω, 1 (Γ(1/4))4 . ≤ π M0 (Ω) + 2λΩ (z) dist(z, ∂Ω) 4π 2

(3.21)

Proof. Consider the function z − a | |z − a| = dist(z, ∂Ω), a ∈ ∂Ω, b ∈ ∂Ω . βΩ (z) := min log b − a We ﬁx z in Ω and choose points a and b in ∂Ω such that z − a . |z − a| = dist(z, ∂Ω), βΩ (z) = log b − a

36

Chapter 3. The Poincar´e metric

Applying the comparison theorem 3.5 and using the function g(w) = (w − a)/(b − a), w ∈ Ω, one obtains λC\{0,1} (g(z)) ≤ λg(Ω) (g(z)) = |b − a|λΩ (z) which is equivalent to 1 1 ≤ . 2λΩ (z) dist(z, ∂Ω) 2|g(z)|λC\{0,1} (g(z)) This together with Landau’s inequality (3.19) give 1 1 ≤ βΩ (z) + . 2λΩ (z) dist(z, ∂Ω) ρ(−1)

(3.22)

Consider now any point a ∈ Ω such that |z − a | = δ, where δ = dist(z, ∂Ω). Clearly, if the circle {w | |w − a | = δ} meets ∂Ω, then βΩ (z) = 0. Otherwise there exists a maximal non-empty annulus of the form An = {w | δe−m < |w − a | < δem }, such that

a ∈ ∂Ω,

An ⊂ Ω,

m/π ≤ M0 (Ω).

As An is maximal, there is a point b ∈ ∂Ω ∩ ∂An, consequently z − a = m ≤ πM0 (Ω). βΩ (z) ≤ log b − a This together with inequality (3.22) give inequality (3.21). The proof of Theorem 3.12 is complete.

Theorems 3.10 and 3.12 are connected with applications of the Poincar´e metric in problems of function theory and of mathematical physics. As an example we mention a result on Hardy type inequalities. Let C0∞ (Ω) be the usual family of smooth functions with compact support in Ω. For p ∈ [1, ∞) we shall consider the Hardy constant cp (Ω) deﬁned by | f ∈ C0∞ (Ω), (∇f )/δ 2/p−1 p =1 , cp (Ω) = sup f /δ 2/p p L (Ω)

L (Ω)

where δ = dist(x + iy, ∂Ω). Theorem 3.13 (Avkhadiev [12], see also [13]). Let Ω ⊂ C be an open set with more than one boundary point in C. For any p ∈ [1, ∞),

(Γ(1/4))4 min{2, p}M0 (Ω) ≤ cp (Ω) ≤ 2p π M0 (Ω) + 4π 2

2 .

(3.23)

3.4. An application of Teichm¨ uller’s theorem

37

The proof of the upper bound in (3.23) uses the inequality (3.21). The key role in this proof is played by the following lemma. Lemma 3.14 (see [12], p. 10 and [13], Theorem 3). Let Ω ⊂ C be an open set of the Riemann sphere with more than two boundary points in C. If 1 ≤ p < ∞, then for any f ∈ C0∞ (Ω),

p p |∇f |p |f |p λ2Ω dx dy ≤ (λΩ δ)2−2p dx dy, (3.24) 2−p 2 Ω Ω δ where δ = dist(x + iy, ∂Ω) and λΩ = λΩ (x + iy). Remark 3.15. In [13], the formula (4) contains a misprint. Namely, the power 2p − 2 has to be replaced by 2 − 2p. The inequality (3.24) is a generalization of a known fact for simply or doubly connected domains. Namely, if any component of Ω can be mapped conformally onto either the unit disc Δ or an annulus of the form {z | q < |z| < 1}, then for any f ∈ C0∞ (Ω),

p p p 2 |f | λΩ dx dy ≤ |∇f |p λ2−p dx dy. (3.25) Ω 2 Ω Ω The inequality (3.25) can be derived using the original Hardy inequality and the conformal invariance of the hyperbolic metric (see, for instance, [4], [11], [12], [13], [63]). Unfortunately, in the general case, where no constraints on the components of Ω are imposed, inequality (3.25) does not hold even for (p/2)p replaced by any other ﬁnite constant. For instance, an inequality of the form (3.25) does not hold for Ω = C \ {0, 1}.

3.4

An application of Teichm¨ uller’s theorem

As is observed in the book [55] of Carleson and Gamelin (see p. 64): A simple scaling and normal families argument shows that conformal annuli of large modulus contain genuine annuli of large modulus. As usual, the modulus M (D) of a conformal annulus, or a doubly connected domain, equals the modulus of the genuine annulus that is conformally equivalent to D (see for instance [3] and [97]). Let Ω ⊂ C be an open set with more than one boundary point in C. A compact set E = C \ Ω is said to be uniformly perfect if M (Ω) = sup{M (D) | D ⊂ Ω doubly connected and separating E} is ﬁnite. The most known quantity describing the geometry of uniformly perfect sets was deﬁned by Pommerenke in [129]: A compact set E on the Riemann sphere

38

Chapter 3. The Poincar´e metric

C that contains the point at inﬁnity is uniformly perfect if there exists a constant c ∈ (0, 1) such that, for every z0 ∈ E \ {∞} and every r ∈ (0, ∞), the set E ∩ An(cr, r; z0 ) is not empty. Let C0 (E) be the supremum of the admissible constants c ∈ (0, 1) for a uniformly perfect set. It is evident that C0 (E) = e−2π M0 (Ω) ,

Ω = C \ E,

where M0 (Ω) is the quantity deﬁned in Section 3.3 as the supremum of the moduli of all genuine annuli An ⊂ Ω such that An separates E and the center of An belongs to ∂Ω (see [129] and also [36], [37], [55], [65], [130]). We take M (Ω) = 0 for open sets that have simply connected components only and M0 (Ω) = 0, if Ω contains no genuine annulus centered at a point of ∂Ω and separating E. Since any genuine annulus is a conformal annulus, the inequality M0 (Ω) ≤ M (Ω)

(3.26)

is trivial and the above remark of Carleson and Gamelin implies M (Ω) = ∞

=⇒

M0 (Ω) = ∞.

(3.27)

On the other hand, using his theorem on extremal moduli, Teichm¨ uller proved in [162] that any doubly connected domain D with M (D) > 1/2 contains a circle which separates the components of its complement. Moreover, the center of this circle may be chosen at E1 , the bounded component of C \ D, and the radius of the circle equals the diameter of E1 . Solynin (see [153]) and Herron, Liu, and Minda (see [82]) proved that a separating circle exists for doubly connected domains D with M (D) > 1/4, if one does not restrict the position of the center of the circle. In [82] it is also proved that such D contain separating annuli An of modulus M (An) = M (D) − c (c ≈ 0.46), where again no restriction is imposed on the center of the √ annulus. In the same paper, they prove that M (An) ≥ M (D) − c˜ (˜ c = π1 log(2 2 + 2) ≈ 0.501) if one ﬁxes the center of An on ∂D. In the following we will use the result of Teichm¨ uller to derive a new quantitative version of the Carleson-Gamelin remark, where we ﬁnd 1/2 to be the best possible constant in a similar inequality, not only choosing the center of An on ∂D but ﬁxing it as Teichm¨ uller did. We will apply this result to diﬀerent characterizations of uniformly perfect sets and to estimates of the second derivative of functions analytic on open sets with uniformly perfect boundaries. In addition to (3.26) and (3.27) we prove that M (Ω) ≤ M0 (Ω) +

1 2

(3.28)

for any open set Ω ⊂ C. The proof is based on the above cited theorem of Teichm¨ uller and on a related formula of Ahlfors in [3].

3.4. An application of Teichm¨ uller’s theorem

39

We are concerned with the following theorem of Teichm¨ uller (see [162] and also [3]). Theorem 3.16. Of all doubly connected domains that separate the pair {−1, 0} from a pair {w0 , ∞} with |w0 | = R, the one with the biggest modulus is the complement of the union of the segments [−1, 0] and [R, ∞]. Clearly, to prove the inequality (3.28) for any open set Ω it is suﬃcient to consider the special case where Ω is a doubly connected domain. Hence, the inequality (3.28) is a corollary of the following extension of Teichm¨ uller’s result on the existence of a separating circle. Theorem 3.17 ([27]). Any doubly connected domain D ⊂ C with M (D) > 1/2 contains an annulus An = An(r, s; z0 ) such that M (An) = M (D) − 1/2, the center z0 belongs to the bounded component E1 of C \ D and diam E1 = r. The constant 1/2 is sharp. Proof. Let E = C \ D = E1 ∪ E2 , where E1 and E2 are the connected components of E. Without loss of generality we may assume that the diameter of E1 equals 1 and that the points z = 0 and z = −1 belong to E1 . Now, we consider the annulus An(1, m; 0), where m = exp(2π(M (D) − 1/2)). It has the following properties: An(1, m; 0) ∩ E1 = ∅, the center of An(1, m; 0) belongs to E1 , and M (An(1, m; 0)) = M (D) − 1/2. If An(1, m; 0) ⊂ D, then An(1, m; 0) separates E and there is nothing to prove. Let us assume that this is not true, i.e., that there exists a point w0 ∈ E2 such that 1 < |w0 | < m. Taking |w0 | = R we consider the Teichm¨ uller annulus D(R) = C \ ([−1, 0] ∩ [R, ∞)) with the modulus Λ(R) (compare [3]). According to Theorem 3.16 the inequality M (D) ≤ Λ(R)

(3.29)

is valid. On the other hand, we will prove that the assumed condition 1 < R < m implies that 1 1 Λ(R) < log R + < M (D). (3.30) 2π 2 Clearly, (3.30) contradicts (3.29). Hence, to complete the proof of Theorem 3.17 it is suﬃcient to show that the ﬁrst inequality in (3.30) holds for any R > 1. The function Λ(R) is implicitly deﬁned by the following formula due to Ahlfors (see [3], p. 75, also compare formula (3.14)): 8 ∞ 1 1 − q 2n−1 R = , 16 q n=1 1 + q 2n

q = e−2πΛ(R) .

(3.31)

It is known that Λ(1) = 1/2 and that Λ(R) → ∞ as R → ∞. If one considers the ﬁrst equation in (3.31) as the deﬁnition of a function R(q), it is evident that qR(q)

40

Chapter 3. The Poincar´e metric

is a decreasing function of q for 0 < q < e−π = e−2πΛ(1) . The function qR(q) decreases from 1/16 to exp(−π), when q increases from 0 to exp(−π). In particular, 16

1, and (3.30) follows. The sharpness of the constant 1/2 is a consequence of Teichm¨ uller’s considerations, since in the case M (D) = 1/2 there does not exist a separating annulus with the above properties. This completes the proof of Theorem 3.17. In contrast to the conformal characteristic M (Ω) the quantity M0 (Ω) is not a conformal invariant. Nevertheless, it may be useful to remark the following consequence of Theorem 3.17. Corollary 3.18. Let Ω1 and Ω2 be two conformally equivalent domains in C. If M0 (Ω1 ) is ﬁnite, then 1 |M0 (Ω1 ) − M0 (Ω2 )| ≤ . 2 Remark 3.19. For any genuine annulus An, the inequality M0 (An) < M (An) holds, since we consider only those separating genuine annuli in An, which have a center lying on ∂An. Geometrically it is evident that a genuine annulus An has such a separating annulus, if and only if 1 1 log 3 < M (An) ≤ M0 (An) + log 3. 2π 2π

3.5

Domains with uniformly perfect boundary

There are about twenty characterizations of domains with uniformly perfect boundary via the Hayman-Wu condition, domains with strong barrier, local behavior of harmonic measure or logarithmic capacity, etc. (see, for instance, [9], [36], [37], [55], [62], [65], [67], [76], [77], [79], [129], [130]). In this section we continue to use the Pommerenke characteristic C0 (E) = exp(−2π M0 (Ω)), Ω = C \ E. Our aim is to consider the problem of comparing the Euclidian geometry characteristic M0 (Ω) of open sets with uniformly perfect boundary with the following characteristics of the hyperbolic geometry on Ω: α(Ω) = inf{λΩ (z)dist(z, ∂Ω) | z ∈ Ω}

(3.32)

3.5. Domains with uniformly perfect boundary and

41

γ(Ω) = sup ∇λ−1 Ω (z) | z ∈ Ω ,

(3.33)

where λΩ is the density of the Poincar´e metric with curvature −4 deﬁned on the components of Ω. We will prove that 1 sup{M0 (Ω)α(Ω)} = (3.34) 4 and that 1 M0 (Ω) = , sup (3.35) γ(Ω) 4 where the supremum is taken with respect to all open sets Ω ⊂ C that have more than one boundary point in C. In the proof of (3.34) and (3.35) we show that (γ(An))2 1 = + (M (An))2 16 4 for any genuine annulus An and that 4M (An) ∼

1 α(An)

for genuine annuli if M (An) → ∞. We must confess that the above considerations have their origin in the central question of this book, compare section 1.2. Let Ω and Π be open sets in C or in C and let A(Ω, Π) be the set of functions f : Ω → Π locally holomorphic or meromorphic and in general multivalued. What can be said about the inﬂuence of the geometric properties of Ω and Π on the quantities f (n) (z) λΠ (f (z)) Cn (Ω, Π) = sup n | z ∈ Ω, f ∈ A(Ω, Π) ? n! (λΩ (z)) The above considerations give us the possibility to determine bounds for C2 (Ω, Π) in terms of the quantities M0 (Ω) and M0 (Π) that are “visible” in Euclidean geometry. During these proofs (see Chapter 6) we get a sharp form of the OsgoodJørgensen inequality (see [88] and [121]), namely sup{sup{|∇ log λΩ (z)|dist(z, ∂Ω) | z ∈ Ω}} = 2,

(3.36)

where the ﬁrst supremum is taken with respect to all hyperbolic domains Ω ⊂ C. Firstly, we cite some known results. In [121], Osgood proved that 1 2 ≤ γ(Ω) ≤ , α(Ω) α(Ω)

Ω ⊂ C.

From Theorem 3.12 of Beardon and Pommerenke, it follows that (Γ(1/4))4 1 , ≤ 2π M0 (Ω) + α(Ω) 2π 2

Ω ⊂ C.

(3.37)

42

Chapter 3. The Poincar´e metric

In [37], the ﬁrst paper on uniformly perfect sets, it is also proved that M0 (Ω) ≤

1 , 2α(Ω)

Ω ⊂ C.

(3.38)

The following theorem assures that the equations (3.34) and (3.35) hold, where (3.34) is the sharp form of (3.38). Theorem 3.20 ([27]). Let Ω ⊂ C be an open set with more than one boundary point in C. If α(Ω) > 0, then the inequalities M0 (Ω)

0. Denoting by M the modulus of such an annulus, i.e., M := M (An) = we get 1 = 4 M |z| sin λAn (z)

1 log(1/), 2π

log(1/|z|) 2M

,

z ∈ An,

(see, for instance, [37]). Hence, d (1/λAn (z)) =: |s(t)|, |∇(1/λAn (z))| = d |z| where s(t) = 4 M sin t − 2 sin t,

with t =

log(1/|z|) ∈ (0, π). 2M

By a straightforward calculation we derive

γ(An) = max{|s(t)| | t ∈ [0, π]} = |s(t0 )| = 2 1 + 4M 2 , tan t0 = −2M,

which completes the proof of Lemma 3.21. The formulas (3.39), (3.42) and Theorem 3.17 immediately imply lim

→0+

M0 (An(, 1; 0)) 1 = . γ(An(, 1; 0)) 4

(3.43)

From the second inequality of (3.39) we see that lim→0+ M0 (An(, 1; 0))α(An(, 1; 0)) ≤

1 . 4

44

Chapter 3. The Poincar´e metric

Now, let us assume that the second equation in (3.40) is not true, i.e., lim→0+ M0 (An(, 1; 0))α(An(, 1; 0))

0, f (ζ)

ζ ∈ Δ,

which is the classical condition for a conformal map f : Δ → Ω to have a convex image. These give the following theorem. Theorem 3.23 (L¨ owner). A proper subdomain Ω of the plane C is convex if and only if sup |∇R(z, Ω)| ≤ 2. z∈Ω

We have proved the following analog of Theorem 3.23. Theorem 3.24 ([15], [20]). Let Ω ⊂ C be a simply connected domain with more than one boundary point. The set E = C \ Ω is convex if and only if inf |∇R(z, Ω)| ≥ 2.

z∈Ω

The second motivation is given by applications of the gradients to SchwarzPick type inequalities (see Chapter 4, Section 6, below) and by a number of theorems and their applications on the geometry of the surface SΩ = {(w, h) | w ∈ Ω, h = R(w, Ω)} (see [49], [74], [76], [89], [95], [119], [168], [169]). As an example we mention Theorem 3.25. A domain Ω ⊂ C is convex if and only if R(·, Ω) is a concave function on Ω. In the paper [95], Kovalev studied an analog of Theorem 3.25 for unbounded simply connected domains Ω ⊂ C. He proved that C \ Ω is a convex set if and only if the function R(·, Ω) is a locally convex function. In [20] we developed these facts using the following observation. Since the Jacobian of the gradient of R is J(z, Ω) =

∂ 2 R(z, Ω) ∂ 2 R(z, Ω) − ∂x2 ∂y 2

∂ 2 R(z, Ω) ∂x∂y

2 , z = x + iy ∈ Ω,

and a condition necessary for a real-analytic function to be convex or concave on Ω is the inequality J(z, Ω) ≥ 0, z ∈ Ω, we observe that Theorem 3.25 and its generalizations are related to ∇R. Moreover, we consider the image of the domain Ω by the gradient (see [20] for details). We present here the case only when Ω is a polygonal domain.

46

Chapter 3. The Poincar´e metric

Using the Wirtinger calculus, the gradient can be written as a complex variable function ∇R(z, Ω) =

∂ R(z, Ω) ∂ R(z, Ω) ∂ R(w, Ω) +i =2 , z = x + iy ∈ Ω. ∂x ∂y ∂z

We begin with a simple example. Let z = f (ζ) = ζ + 1/ζ, ζ ∈ Δ. One has Ω := f (Δ) = C \ [−2, 2] and ∇R(·, Ω)) is a diﬀeomorphism. Therefore, to ﬁnd the gradient image it is suﬃcient to ﬁnd its boundary. We have 3 1 − ζζ 1 g(ζ) = ∇R ζ + , Ω = 2 |1 − ζ 2 | 2 , ζ ∈ Δ. ζ ζ(1 − ζ )2 Hence,

lim g(ζ) =

ζ→eiθ

− 2i for any θ ∈ (0, π), 2i for any θ ∈ (π, 2π).

Moreover, it is clear that g(ζ) has no limit as ζ → ±1. Straightforward computations show that the set of all limit values of g(ζ) as ζ → 1, ζ ∈ Δ, is a curve γ1 given by the parametric equation t t it , t ∈ (−π, π). 2 cos − i sin w1 (t) = 2 e 2 2 This is one branch between the two contact points 2i and −2i of an epicycloid that is described by a point on a circle of radius 2 rolling on another circle of radius 2. Since g(ζ) = −g(−ζ), the gradient image of Ω is the set of all points outside the two branches of the above epicycloid, where the second branch is described by w2 (t) = −w1 (π + t), t ∈ (−π, π). One may observe that γ1 coincides with the set of values of ∇R for an angular domain with opening angle 2π, which is a branch of an epicycloid. This observation can be extended: If Ω is a polygonal domain, then the gradient image is bounded by branches of epicycloids or hypocycloids. To avoid confusion we want to mention that these epicycloids diﬀer from those occurring in the Ptolemaic system. Theorem 3.26 ([20]). Let Ω be a simply connected domain in C or in C. If the boundary of Ω is a polygon with inner angles παk , vertices zk and sides (zk , zk+1 ), k = 1, . . . , n, zn+1 = z1 , then (a) ∇R(·, Ω) is a real-analytic function on Ω \ {z1 , . . . , zn , ∞}, (b) the gradient image of a side (zk , zk+1 ) is a point wk such that |wk | = 2, (c) the set of all limit values of ∇R as z → zk is a curve γk that joins wk−1 with wk and is deﬁned by a parametric equation π π wk (t) = 2 ei(ck +αk t) (αk cos t − i sin t), t ∈ [− , ], 2 2 where ck is a real constant.

3.6. Derivatives of the conformal radius

47

Using Theorem 2.14 we shall obtain sharp estimates for higher-order derivatives of the conformal radius.. Let us introduce the functions μk (z, Ω) :=

∂ k log R(z, Ω)−2 1 , R(z, Ω)k (k − 1)! ∂z k

k = 1, . . . , n,

z ∈ Ω.

Also, for a ﬁxed a ∈ Ω we consider μk = μk (a, Ω) and the quantities τn,n−1 (α), τn,n−2 (α), . . . , τn,0 (α)), deﬁned by the following recurrent formulas: τk,k (α) = 1 (0 ≤ k ≤ n), τk,0 (α) =

τm,k (α) =

m−k+1 s=1

k−1 α μk−s τs,0 (α) , 1 ≤ k ≤ n, k s=0

1 τs−1,0 (1)τm−s,k−1 (α), s

2 ≤ k ≤ m ≤ n.

(3.45)

(3.46)

In the case 0 ≤ k ≤ n − 1, τn,k (α) depends on n, k, α, a and Ω. For instance, if Ω = Δ and a ∈ Δ, then n + 2α − 1 τn,k (α) = an−k . n−k Theorem 3.27 ([14]). If Ω is a simply connected proper subdomain of C, then 1 sup sup |μk (a, Ω)| = μk (a, C \ [ , ∞)) 4 Ω a∈Ω

(3.47)

and sup sup |τn,k (α)| = Ω a∈Ω

n−k s=0

2n + 3α − 1 s

n−k−s+α−2 n−k−s

.

(3.48)

Proof. Let ψ be the conformal mapping of Ω onto Δ, ψ(a) = 0, ψ (a) = λΩ (a) > 0. By deﬁnition, λΩ (z) = |ψ (z)|(1 − |ψ(z)|2 )−1 , z ∈ Ω. This yields 2ψ(z) ψ (z) ∂ log λ2Ω (z) ψ (z). = − ∂z ψ (z) 1 − ψ(z)ψ(z) Hence λkΩ (a)μk (a, Ω) =

1 1 ∂ k log λ2Ω (a) = k (k − 1)! ∂a (k − 1)!

ψ (z) ψ (z)

(k−1) |z=a

(3.49)

48 and

Chapter 3. The Poincar´e metric ∞

ψ (z) k+1 λΩ (a)μk+1 (a, Ω)(z − a)k = ψ (z)

(3.50)

k=0

in some neighbourhood of a. According to Theorem 2.14 by Klouth and Wirths and the equation (3.50), 1 |μk (a, Ω)| ≤ μk (0, C \ ( , ∞)), 4

k = 1, 2, . . . .

Equality for k ∈ N occurs if and only if {a, Ω} is a linear transformation of {0, C\[ 14 , ∞)}. The conformal mapping Φ0 : Δ → C\[ 14 , ∞), Φ0 (0) = 0, Φ0 (0) > 0, is given by the Koebe function Φ0 (ζ) = ζ(1 + ζ)−2 . It is not diﬃcult to verify by direct calculations that 0 τn,k (α) =

n−k s=0

2n + 3α − 1 s

n−k−s+α−2 n−k−s

for the Koebe domain C \ [ 14 , ∞) at the point z = 0. From (3.47) we obtain (3.48) since τn,k (α) is a polynomial with positive coeﬃcients in the μ1 , . . . , μn . For n = 1, we get |τ1,0 (α)| = α|grad λ−1 Ω (a)|. Hence, |τ1,0 (α)| ≤ 4α by the classical Koebe constant 1/4.

Chapter 4

Basic Schwarz-Pick type inequalities Let Ω ⊂ C and Π ⊂ C be two domains equipped by the Poincar´e metric. We are concerned with the set A(Ω, Π) = {f : Ω → Π} of functions locally holomorphic or meromorphic in Ω and, in general, multivalued. Let λΩ (z), z ∈ Ω, and λΠ (w), w ∈ Π, denote the density of the Poincar´e metric at z ∈ Ω and w ∈ Π, respectively. Consider the functional Ln (f, z, Ω, Π) deﬁned by (n) n f (z) (λΩ (z)) = Ln (f, z, Ω, Π) , z ∈ Ω, f ∈ A(Ω, Π), n ∈ N. n! λΠ (f (z)) Many problems in geometric function theory are devoted to the problem of determining Mn (z, Ω, Π) := sup {Ln (f, z, Ω, Π) | f ∈ A(Ω, Π)}, and, respectively, Cn (Ω, Π) := sup {Mn (z, Ω, Π) | z ∈ Ω}. It is clear that Cn (Ω, Π) is not dependent on f and z ∈ Ω and represents the smallest number possible in the inequality (n) n f (z) (λΩ (z)) ≤ Cn (Ω, Π) , z ∈ Ω, f ∈ A(Ω, Π). n! λΠ (f (z)) The classical Schwarz-Pick lemma says that M1 (z, Δ, Δ) = C1 (Δ, Δ) = 1 and in turn M1 (z, Ω, Π) = C1 (Ω, Π) = 1 for any pair (Ω, Π) of hyperbolic domains in the extended complex plane, as we have discussed above.

50

Chapter 4. Basic Schwarz-Pick type inequalities

We shall consider the problem of determining Cn (Ω, Π) for all n ≥ 2. In the proofs we will frequently use the fact that the functions Mn and Cn are invariant under linear transformation of domains. This means that the equations Mn (z, Ω, Π) = Mn (az + b, aΩ + b, cΠ + d) and Cn (Ω, Π) = Cn (aΩ + b, cΠ + d) are valid, where aΩ + b = {az + b | z ∈ Ω} and cΠ + d = {cw + d | w ∈ Π} for some a, c ∈ C \ {0}, b, d ∈ C. Also, the following fact deserves reader’s attention. For technical reasons we explain a simple case, when Π is a bounded domain and z ∈ Ω ⊂ C. Normal family arguments show that there exists a sequence fk ∈ A(Ω, Π) such that Mn (z, Ω, Π) = lim Ln (fk , z, Ω, Π), k→∞

and fk converges to a holomorphic function f0 uniformly in the interior of the domain Ω. Clearly, there are two possible cases to distinguish: The limit function f0 belongs to the family A(Ω, Π) or f0 (z) ≡ const. ∈ ∂Π, and consequently, f0 ∈ A(Ω, Π). The second case is typical in our problems. If n ≥ 2, then there is no extremal function f ∈ A(Ω, Π) such that Mn (z, Ω, Π) = Ln (f, z, Ω, Π) except in some very special cases.

4.1

Two classical inequalities

Let f ∈ A(Δ, Δ). A consequence of |f (z0 )| ≤

1 − |f (z0 )|2 , 1 − |z0 |2

z0 ∈ Δ,

is the sharp inequality |f (z0 )| ≤

1 , 1 − |z0 |2

z0 ∈ Δ.

Equality is attained for conformal automorphisms f of Δ such that f (z0 ) = 0. In 1920, Sz´ asz extended the latter inequality to higher-order derivatives. Theorem 4.1 (O. Sz´asz [160]). For any f ∈ A(Δ, Δ), z ∈ Δ, and m ∈ N, the sharp inequality 2 m (2m + 1)! m (2m+1) (z) ≤ |z|2k f 2m+1 k (1 − |z|2 ) k=0 is valid. Equality occurs only for the functions m+1 ζ −z iγ m , f (ζ) = e ζ 1 − zζ

γ ∈ R.

4.1. Two classical inequalities

51

Proof. We consider the function g ∈ A(Δ, Δ) deﬁned by ∞ ζ +z ak ζ k , ζ ∈ Δ. = g(ζ) = f 1 + zζ k=0

According to the Cauchy formula, integration along a circle Γ around the origin lying in its neighbourhood and the use of the variable substitution ξ =

ζ +z = S(ζ) 1 + zζ

results in the following chain of equations: 1 f (ξ) dξ g(ζ)(1 + zζ)n−1 dζ 1 f (n) (z) = . = n! 2πi S(Γ) (ξ − z)n+1 2πi Γ ζ n+1 (1 − |z|2 )n For n = 2m + 1 using |g(ζ)| ≤ 1 we obtain 2π 1 1 g(ζ)(1 + zζ)2m dζ ≤ |1 + zeiθ |2m dθ. 2π 2π Γ ζ 2m+2 0

(4.1)

(4.2)

Clearly, formulas (4.1) and (4.2) imply the desired estimate by the binomial and Parseval formulas. A little examination of the inequality (4.2) gives that equality can occur only for the function m ζ +z iγ m+1 g(ζ) = e ζ , γ ∈ R. 1 + zζ This completes the proof of Theorem 4.1.

Remark 4.2. In [160], during the proof of the theorem Sz´ asz gave the formula n n n−1 1 − |z|2 ak z n−k (4.3) f (n) (z) = n−k n! k=1

as a consequence of equation (4.1). Also, in [160] by a similar proof Sz´ asz obtained an explicit sharp bound for |f (z)|. But the problem of ﬁnding an explicit formula for {max |f (2m) (z)| | f ∈ A(Δ, Δ)}, z ∈ Δ, in the case m ≥ 2 is still open (2008). Concerning Sz´asz’s majorant for n = 2m + 1, one easily obtains that 2π 2π 1 (2m − 1)!! 1 |1 + zeiθ |2m dθ = |1 + eiθ |2m dθ = 22m sup . 2π 0 (2m)!! z∈Δ 2π 0 Moreover, for n = 2m the proof gives the estimate 2m 1 − |z|2 22m (2m − 2)!! |f (2m) (z)| ≤ sup (2m)! π (2m − 1)!! z∈Δ which is not sharp, at least for n = 2.

52

Chapter 4. Basic Schwarz-Pick type inequalities

E. Landau remarked in [98] that a special case of the following theorem, namely, Theorem 7.1 (see below), is a consequence of the validity of the Bieberbach conjecture. Theorem 4.3. Let Π be a simply connected proper subdomain of C and z ∈ Δ. Then Mn (z, Δ, Π) ≤ Mn (z, Δ, H2 ) = (n + |z|)(1 + |z|)n−2 , where H2 = C \ (−∞, 0]. Proof. We use the fact that de Branges’ proof of the Bieberbach conjecture implies a proof of the generalized Bieberbach or Rogosinski conjecture. This means that the Taylor coeﬃcients of a function subordinate to a schlicht function are dominated by the Taylor coeﬃcients of the Koebe function. Therefore, for the Taylor coeﬃcients of a function f ∈ A(Δ, Π), the inequalities λΠ (a0 ) |ak | ≤ k are valid. Using the latter inequalities and formula (4.3), we get Mn (z, Δ, Π) ≤

n n−1 k|z|n−k = (n + |z|)(1 + |z|)n−2 . n−k

k=1

Straightforward computations show that Ln (g0 , z, Δ, H2 ) = Mn (z, Δ, H2 ) = (n + |z|)(1 + |z|)n−2 , where 1 g0 (ζ) = 4

1 + ζ |z|/z 1 − ζ |z|/z

2 −1

=

∞

k(ζ |z|/z)k .

k=1

This completes the proof of Theorem 4.3.

Corollary 4.4. If Π is a simply connected proper subdomain of C, then Cn (Δ, Π) ≤ Cn (Δ, H2 ) = 2n−2 (n + 1).

4.2

Theorems of Ruscheweyh and Yamashita

According to our notation Cn (Ω, Π), the identity Cn (Δ, Π) = 2n−1 has been proved by St. Ruscheweyh (see [142] and [143]) in two basic cases, when Π is a half plane or a disc. Here we present the original versions of his theorems.

4.2. Theorems of Ruscheweyh and Yamashita

53

Theorem 4.5 (St. Ruscheweyh [143] (1985)). Let Δ = {ζ | |ζ| < 1}. For any f ∈ A(Δ, Δ), z ∈ Δ, and n ∈ N the sharp inequality 1 (n) 1 − |f (z)|2 f (z) ≤ n n! (1 − |z|) (1 + |z|) is valid. This inequality was conjectured by Ruscheweyh in 1974 (see [142]). Proof. Let the function g(z) =

∞

ak z k

k=0

be holomorphic and |g(z)| ≤ 1 in the unit disc. It is well known that the coeﬃcients of such a function satisfy the inequalities |a0 | ≤ 1,

|ak | ≤ 1 − |a0 |2

(4.4)

for any k ≥ 1. The inequalities (4.4) imply the assertion of the theorem immediately for z = 0. We now suppose that z = 0 and deﬁne g(ζ) as in the proof of Theorem 4.1. Using the Sz´ asz formula (4.3) and (4.4) together with equations f (z) = h(0) = b0 , one easily gets n n 1 − |z|2 (n) n−1 n−1 |z|n−k = (1 − |b0 |2 ) (1 + |z|) f (z) ≤ (1 − |b0 |2 ) n−k n! k=1

which is the inequality to prove. We have to show that the inequality of Theorem 4.5 is sharp for n ≥ 2 and any z ∈ Δ, z = 0. To this end we choose a sequence ak in Δ such that ak → z/|z| as k → ∞ and consider unimodular bounded holomorphic functions z − ak fk (z) = . 1 − ak z Straightforward computations give (n)

|fk (z)|

2

1 − |fk (z)|

=

n!|ak |n−1 n! → n−1 2 n−1 |1 − ak z| (1 − |z| ) (1 − |z|) (1 − |z|2 )

as k → ∞. This completes the proof of Theorem 4.5.

Theorem 4.6 (St. Ruscheweyh [142]). Let f be holomorphic in Δ, n ∈ N, and ρ(f (z)) denote the minimal distance from f (z) to the boundary of the closed convex hull of f (Δ). Then the sharp inequality 2 ρ(f (z)) 1 (n) f (z) ≤ n! (1 − |z|)n (1 + |z|) is valid. Equality occurs for conformal maps of the unit disc onto a halfplane.

54

Chapter 4. Basic Schwarz-Pick type inequalities

Remark 4.7. For H1 = {z | Re z > 0} this implies that for any f ∈ A(Δ, H1 ), z ∈ Δ and n ∈ N the inequality n 1 (n) (λΔ (z)) f (z) ≤ (1 + |z|)n−1 n! λH1 (f (z))

holds. Proof of Theorem 4.6. (As St. Ruscheweyh indicated in [142], he owes the idea of the following proof to T. Sheil-Small.) Without loss of generality we can suppose that Re f (ζ) > 0 in Δ, f (0) = 1 and ρ(f (z)) = Re f (z) for a given point z ∈ Δ.Using the Herglotz formula

2π

1 + ζe−it dμ(t), 1 − ζe−it

f (ζ) = 0

where μ(t) is a nondecreasing function with μ(2π) − μ(0) = 1, we get f (n) (ζ) = 2n!

2π

0

e−int dμ(t). (1 − ζe−it )n+1

Consequently, by simple estimates and the Poisson formula |f (n) (z)| ≤

2n! (1 − |z|)n−1 (1 − |z|2 )

2π

0

2n! Ref (z) 1 − |z|2 dμ(t) = . |1 − ze−it |2 (1 − |z|)n−1 (1 − |z|2 )

Since 1 − |z| = |1 − ze−it | is true only for e−it = e−it0 = z/|z|, it is easily seen that equality can occur only for a piece-wise constant function μ(t) such that μ([0, 2π]) = {0, 1}, and, consequently, the corresponding function f0 has the form f0 (ζ) =

1 + ζe−it0 , 1 − ζe−it0

eit0 = z/|z|,

and f0 maps Δ onto H1 . This completes the proof of Theorem 4.6.

Let Ω be a hyperbolic domain in C, i.e., Ω has more than two boundary points. For a universal covering map ϕ from Δ onto Ω and given z ∈ Ω, let ρΩ (z) denote the greatest r ∈ (0, 1] such that ϕ is univalent in the non-Euclidean disc ζ −w 0}.

4.3. Pairs of simply connected domains

55

Theorem 4.8 (S. Yamashita [170]). Let Ω be a hyperbolic domain in C, let z ∈ Ω, and let n ≥ 2. For any holomorphic function f : Ω → C with positive real part in Ω the sharp inequality n λΩ (z) 1 (n) 2n − 1 Re (f (z)) f (z) ≤ 2 n n! ρΩ (z) is valid. Also, in [170] Yamashita established the case of equality in Theorem 4.8. In particular, if Ω = H2 and f (ζ) = ζ −1/2 , then the equality is attained at the points ζ = z = x > 0. Taking into account the classical equality 1/λH1 (w) = 2Re w and the invariance of Cn under linear transformation of domains, one obtains the equality 2n − 1 (a = 0). Cn (aH2 + b, H1 ) = n Clearly, if Ω is a simply connected domain, then ρΩ (z) = 1 for any point z ∈ Ω. Therefore, this particular case of Yamashita’s theorem can be presented as follows. Theorem 4.9. If Ω is a simply connected proper subdomain of C, then 2n − 1 Cn (Ω, H1 ) ≤ Cn (H2 , H1 ) = . n In fact, Theorem 4.9 is equivalent to Theorem 4.8. Actually, suppose that Ω is a multiply connected domain in C and ϕ as above. The covering map ϕ1 : Δ → Ω deﬁned by ζ +w ϕ1 (ζ) = ϕ 1 + wζ is univalent in the disc Δρ = {ζ | |ζ| < ρΩ (z)}. The domain Ωρ := ϕ(Δρ ) is a simply connected proper subdomain of C and λΩ (z) = λΩρ (z). ρΩ (z) Applying Theorem 4.9 to the function f | Ωρ in the simply connected domain Ωρ one immediately obtains Theorem 4.8. The original proof of Yamasita of these theorems is based on coeﬃcient estimates of univalent functions by Chua (see [56] and [170]). We will obtain Theorem 4.9 as a special case of our theorem proved in Section 4 of the present chapter.

4.3

Pairs of simply connected domains

We again will need the domain H2 = C \ (−∞, 0]. In this section we will discuss the following theorem.

56

Chapter 4. Basic Schwarz-Pick type inequalities

Theorem 4.10 ([16]). If Ω and Π are simply connected proper subdomains of C, then the sharp estimate Cn (Ω, Π) ≤ Cn (H2 , H2 ) = 4n−1 is valid. Proof. For f ∈ A(Ω, Π), z0 ∈ Ω, we consider the functions s(ζ) := (ΦΩ,z0 (ζ) − z0 ) λΩ (z0 ), and

ζ ∈ Δ,

t(ζ) := ΦΠ,f (z0 ) (ζ) − f (z0 ) λΠ (f (z0 )),

ζ ∈ Δ.

Both of them belong to the class S of functions univalent in Δ and normalized in the origin as usual. The fact that f (Ω) is a subset of Π may be expressed in terms of the function u(ζ) := (f (ΦΩ,z0 (ζ)) − f (z0 )) λΠ (f (z0 )),

ζ ∈ Δ.

It means that u(ζ) is subordinate to t(ζ). Using the Taylor expansion u(ζ) =

∞

ak λΠ (f (z0 ))ζ k

k=1

and the function ΨΩ,z0 inverse to ΦΩ,z0 we get f (z) = f (z0 ) +

∞

k

ak (ΨΩ,z0 (z))

k=1

and therefore

n (n) 1 f (n) (z0 ) k ak . = (ΨΩ,z0 (z)) n! n! z=z0 k=1

−1

If we denote by s

(w) the function inverse to s(ζ), deﬁne

∞ k s−1 (w) = Am,k (z0 )wm m=k

and use we see that

s−1 (λΩ (z0 )(z − z0 )) = ΨΩ,z0 (z), (n) 1 k n = An,k (z0 ) (λΩ (z0 )) . (ΨΩ,z0 (z)) n! z=z0

4.3. Pairs of simply connected domains

57

Hence, we get the formula f (n) (z0 ) n ak An,k (z0 ) (λΩ (z0 )) , = n! n

(4.5)

k=1

which will be central in what follows. Let K1 be the inverse of the Koebe function k1 (z) =

z . (1 + z)2

We have K1 (w) = w +

∞

Am (K1 ) wm ,

Am (K1 ) =

m=2

(2m)! , m!(m + 1)!

and, by the L¨ owner theorem 2.13, the sharp estimates |Am,1 | ≤ Am (K1 ) =

(2m)! m!(m + 1)!

are valid. Let us deﬁne Am,k (K1 ) by expansions k

(K1 (w)) =

∞

Am,k (K1 ) wm

m=k

in some neighbourhood of the origin. It is evident that An,k (K1 ) is a polynomial with positive coeﬃcients in Am (K1 ), 2 ≤ m ≤ n, and An,k (z0 ) is the same polynomial in Am,1 (z0 ), 2 ≤ m ≤ n. Accordingly, we have |An,k (z0 )| ≤ An,k (K1 ),

1 ≤ k ≤ n.

For those quantities An,k (K1 ) we get 2k(2n − 1)! k 2n = An,k (K1 ) = . n − k n (n − k)!(n + k)!

(4.6)

(4.7)

This is an immediate consequence of the Cauchy formula according to 1 1 (K1 (w))k (1 + ζ)2n−1 (1 − ζ) dw = dζ, An,k (K1 ) = 2πi k1 (∂Δr ) wn+1 2πi ∂Δr ζ n−k+1 where r ∈ (0, 1) and ∂Δ = {ζ | |ζ| = r} (for (4.6) and (4.7) compare [56]). Further, the function u is subordinate to the function t univalent in Δ and normalized as usual. According to the Rogosinski conjecture settled by de Branges’

58

Chapter 4. Basic Schwarz-Pick type inequalities

proof of the Bieberbach conjecture, the Taylor coeﬃcients of the function u satisfy the inequalities |ak | λΠ (f (z0 )) ≤ k, k ∈ N. Using these estimates and the basic formula (4.5), we obtain |f (n) (z0 )| λΠ (f (z0 )) k An,k (K1 ), n ≤ n! (λΩ (z0 )) n

k=1

where An,k (K1 ) are given by formula (4.7). Using again the Cauchy formula to compute the latter sum, we get n k2 2n = 4n−1 n−k n

k=1

(for the latter formula see also [56]). To complete the proof we have to show that Cn (H2 , H2 ) = 4n−1 , where H2 = C\(−∞, 0]. To this end we consider f0 (z) = 1/z, z ∈ H2 . For any z = x > 0 using λH2 (x) = 1/(4x) and λH2 (1/x) = x/4 one easily gets (n) |f0 (x)| λH2 (1/x)) n−1 . n = 4 n! (λH2 (x)) This completes the proof of Theorem 4.10.

We will see below that it is possible to ﬁnd lower bounds for punishing factors Cn (Δ, Π) using the properties of hyperbolic metrics if the boundaries of the domains in question are “nice” (see Section 5.5). This was the reason why we tried to ﬁnd lower bounds for punishing factors for pairs of simply connected domains by considering the limiting processes going to the boundaries. Our result is the following theorem. Theorem 4.11 (see [16]). Let Ω and Π be two simply connected proper subdomains of C whose boundaries contain sectorial accessible analytic arcs. Then, for any n ≥ 2, the assertion 2n−1 ≤ Cn (Δ, Π) ≤ Cn (Ω, Π) holds. Equality occurs if Ω = Δ and Π is convex. As usual, a boundary arc is said to be sectorial accessible if any point on this arc is the vertex of an open triangle contained in the domain in question (see for instance [131]). The proof for this result is very technical. Therefore we omit the details here. We hope that one of our readers will be able to give a proof for the following conjecture. Conjecture (see [16]). Given n ≥ 3, then Cn (Ω, Π) ≥ 2n−1 for all simply connected domains Ω and Π in C.

4.4. Holomorphic mappings into convex domains

4.4

59

Holomorphic mappings into convex domains

We shall consider a holomorphic function f : Ω → Π, where Π is a proper convex subdomain of the plane and Ω is the unit disc or a simply connected domain. The following theorem is a generalization of the Carath´eodory inequality ([51], [52], see also [42], p. 365, [48], p. 213) for Taylor coeﬃcients of holomorphic functions with positive real part. Also, in [83] Herzig gives a description of extremal functions. Theorem 4.12 (see Rogosinski [139]). Let g and g0 be holomorphic functions with expansions g(z) =

∞

an z n ,

g0 (z) = z +

n=1

∞

bn z n ,

z ∈ Δ.

n=2

Suppose that g0 is univalent in Δ and Π := g0 (Δ) is a convex domain. If g ≺ g0 , then |an | ≤ 1 for any n ≥ 1. Equalities |ak | = 1 for all k = 1, 2, . . . , n with some n ≥ 2 occur if and only if g is a conformal map of Δ onto a half plane. Proof. For n ∈ N we consider the function gn (z) :=

n

∞

k=1

k=2

1 g(ζ k z 1/n ) = an z + ank z k , n

z ∈ Δ,

where ζ = e2πi/n . Since wk = g(ζ k z) ∈ Π for any z ∈ Δ, from the convexity of Π it follows that (w1 + · · · + wn )/n ∈ Π. Hence, the function gn is subordinate to g0 . Applying the Schwarz lemma to the function g0−1 (gn (z)), one immediately gets the desired inequality |an | ≤ 1 for any n ≥ 1. If |a1 | = 1, then Schwarz’s lemma implies that g(z) = g0 (cz) , where |c| = 1. If |a1 | = |a2 | = 1 , then g(z) = g0 (cz) and |b2 | = 1. Consequently, g is a conformal map of the unit disc onto a half plane according to the classical L¨ owner theorem on coeﬃcients of convex univalent functions. The proof is complete. The assertion of the following theorem immediately implies that Mn (Δ, Π) = 2n−1 for any convex domain Π. Theorem 4.13 ([16]). Let Π be a convex proper subdomain of C and let n ∈ N. Then for any z ∈ Δ the equation Mn (z, Δ, Π) = (1 + |z|)n−1

(4.8)

is valid. In the case z = 0 and n ≥ 2, there exist extremal functions for which Ln (f, z, Δ, Π) = (1 + |z|)n−1 if and only if Π is a half plane.

60

Chapter 4. Basic Schwarz-Pick type inequalities

Proof. We ﬁx z ∈ Δ and consider f ∈ A(Δ, Π). Introduce the function f1 ∈ A(Δ, Π) deﬁned by ∞ ζ +z = ak ζ k , ζ ∈ Δ. f1 ζ) = f 1 + zζ k=0

Using the formula (4.3) we obtain n n−1 n−k Mn (z, Δ, Π) ≤ sup |z| |ak |λΠ (a0 ) | g ∈ A(Δ, Π) . n−k

(4.9)

k=1

From Theorem 4.12 for the function g(ζ) = λΠ (a0 )f1 (ζ) it follows that λΠ (a0 )|ak | ≤ 1,

k ∈ N.

This together with inequality (4.9) give n n−1 |z|n−k = (1 + |z|)n−1 . Mn (z, Δ, Π) ≤ n−k k=1

To prove Mn (z, Δ, Π) ≥ (1 + |z|)n−1 we ﬁrst consider the case z = 0. If Φ denotes the conformal map of the unit disc Δ onto Π with Φ(0) = w0 and Φ (0) = 1/λΠ (w0 ) > 0, it is obvious that the function f deﬁned by f (ζ) = Φ(ζ n ) has the desired property: f (n) (0)λΠ (w0 )/n! = 1. The inequalities for coeﬃcients in Theorem 4.12 occur for all k = 1, 2, . . . , n with n ≥ 2 if and only if g is a conformal map of the unit disc onto a half plane. Therefore the existence of a function f for which Ln (f, z, Δ, Π) = (1 + |z|)n−1 in the case z = 0 is possible only if Π is a half plane and this is the case in Theorem 4.5 of Ruscheweyh. In the general case we only know that there is a maximising sequence (fk ) ⊂ A(Ω, Π) such that Mn (z, Δ, Π) = lim Ln (fk , z, Δ, Π). k→∞

Without loss of generality, we can suppose that the sequence (fk ) converges uniformly in the interior of the unit disc. Let f0 (z) = lim fk (z), k→∞

z ∈ Δ.

If z = 0, n ≥ 2, and Π is not a half plane, then the proof gives that f0 ∈ A(Δ, Π). Hence, f0 (z) ≡ const. ∈ ∂Π. To prove, nevertheless, the sharpness of (4.8) for any convex domain Π, we present such a sequence explicitly using the convexity of Π and the linear invariance of Mn (z, Δ, Π). Clearly, there exist a point w1 ∈ ∂Π, a disc Δ1 and a half plane H such that Δ1 ⊂ Π ⊂ H and w1 ∈ ∂Δ1 ∩∂Π∩∂H. Without loss of generality we may suppose that Δ1 = {w | |w − 1| < 1} ⊂ Π ⊂ H1 .

4.4. Holomorphic mappings into convex domains

61

The origin belongs to the boundaries of Δ1 , Π and H1 . Now we use a reﬁnement of an idea presented in the proof of Theorem 4.5 by Ruscheweyh. Namely, for ﬁxed z ∈ Δ \ {0}, we consider the sequence z 1 αk = 1 − , k ∈ N, k + 1 |z| of complex numbers and the sequence of fk , k ∈ N, of conformal maps of Δ onto Δ1 deﬁned by αk ζ − αk fk (ζ) = + 1, ζ ∈ Δ. |αk | 1 − αk ζ A straightforward computation using 1 = 2 Re w − |w|2 , λΔ1 (w) yields

w ∈ Δ1 ,

(1 − |z|2 )n−1 = (1 + |z|)n−1 . k→∞ |1 − αk z|n−1

lim Ln (fk , z, Δ, Δ1 ) = lim

k→∞

Theorem 3.5 assures that 1− Since

|w|2 λH1 (w) λΠ (w) = ≤ ≤ 1, 2 Re w λΔ1 (w) λΔ1 (w)

w ∈ Δ1 .

|fk (z)|2 = 0, k→∞ 2 Re (fk (z)) lim

it is evident that lim Ln (fk , z, Δ, Π) = lim Ln (fk , z, Δ, Δ1 )

k→∞

k→∞

λΠ (fk (z)) = (1 + |z|)n−1 . λΔ1 (fk (z))

This completes the proof of Theorem 4.13.

Finally, we consider pairs (Ω, Π) with simply connected Ω and convex Π. We will need the constant 2n − 1 Cn (H2 , Λ) = n from Theorem 4.9 of Yamashita. By mathematical induction and Stirling’s formula it can be shown that 1 (2n − 1)!! 4n−1/2 2n − 1 √ + O(1/n) as n → ∞, = 4n−1/2 = √ n (2n)!! n π and that

(2n − 1)!! 4n−1/2 (n ≥ 2). < √ (2n)!! n+1 Theorem 4.9 is a special case of the following assertion. 2n−1 < 4n−1/2

62

Chapter 4. Basic Schwarz-Pick type inequalities

Theorem 4.14 ([17]). Let Ω and Π be simply connected proper subdomains of C. If Π is a convex domain, then Cn (Ω, Π) ≤ Cn (H2 , Π) = 4n−1/2

(2n − 1)!! . (2n)!!

Proof. We follow the proof of our Theorem 4.10 with the same notation but some little changes. Since Π is a convex domain and u ≺ t, we can use Theorem 4.12 of Rogosinski to obtain that |ak | λΠ (f (z0 )) ≤ 1. This together with formula (4.5) give |f (n) (z0 )| λΠ (f (z0 )) |An,k (z0 )|. n ≤ n! (λΩ (z0 )) n

k=1

As in the proof of Theorem 4.10 we get |f (n) (z0 )| λΠ (f (z0 )) An,k (K1 ). n ≤ n! (λΩ (z0 )) n

k=1

Summing up gives n k=1

An,k (K1 ) =

1 2πi

∂Δr

(1 + ζ)2n−1 (1 − ζ)n dζ = ζn

2n − 1 n

,

which proves the desired inequality (2n − 1)!! 2n − 1 = 4n−1/2 . Cn (Ω, Π) ≤ n (2n)!! For Π = H1 the lower estimate Cn (H2 , Π) ≥ 4n−1/2

(2n − 1)!! (2n)!!

is given by the example of f (ζ) = ζ −1/2 at the points ζ = z = x > 0. To obtain the lower estimate for any convex domain Π, we proceed by approximation as in the proof of Theorem 4.13. Namely, we consider Δ1 = {w | |w − 1| < 1} ⊂ Π ⊂ H1 and the function f ∈ A(H2 , Δ1 ) ⊂ A(H2 , H1 ) deﬁned by f (ζ) = √

2 , ζ +1

√

1 = 1.

Straightforward computations using the comparison theorem and the asymptotic behavior of the nth derivative of this function at the point z = x > 0, as x → +∞, xn+1/2 |f (n) (x)| (2n − 1)!! =2 + o(1) n! (2n)!!

4.5. Punishing factors for convex pairs

63

gives that |f (n) (x)|λΠ (f (x)) (2n − 1)!! = 4n−1/2 . x→+∞ n! (λH2 (x))n (2n)!! lim

This completes the proof of Theorem 4.14.

4.5

Punishing factors for convex pairs

The main aim of the present section is to prove that Cn (Ω, Π) = 2n−1 for any pair (Ω, Π) of convex domains ( see [24], see also Chua [56], Li [102], [103], and Yamashita [169] for special cases). Theorem 4.15 ([24]). Let Ω and Π be two convex proper subdomains of C and let f ∈ A(Ω, Π), n ∈ N. Then for any z0 ∈ Ω the inequality |f (n) (z0 )| (λΩ (z0 ))n ≤ 2n−1 n! λΠ (f (z0 ))

(4.10)

is valid. The constant 2n−1 can not be replaced by a smaller one independent of f ∈ A(Ω, Π) and z0 ∈ Ω for any pair (Ω, Π) of convex domains. Proof. In the following we consider only the cases n ≥ 2, since the case n = 1 is given by the Schwarz-Pick lemma. We ﬁrst prove that Cn (Ω, Π) ≤ 2n−1 . According to the central formula (4.5) we have to prove that n ck An,k ≤ 2n−1 . k=1

In the present theorem, the functions s and t belong to the class K of functions univalent in Δ that map Δ onto a convex domain and are normalized as usual, e.g., t(0) = 0 and t (0) = 1. Moreover, ck are the Taylor coeﬃcients of a holomorphic function subordinate to the function t, and An,k is the n-th Taylor coeﬃcient of the power F k (w), where F is the inverse of the function s. Since the extreme points of the closed convex hull of the set {u | u ≺ t for some t ∈ K} are the functions

xz , 1 − yz

z ∈ Δ,

for ﬁxed (x, y) ∈ ∂Δ × ∂Δ (see [75], Theorem 5.21), it remains to prove that n n k k xy An,k ≤ y An,k ≤ 2n−1 , y ∈ ∂Δ. Re k=1

k=1

64

Chapter 4. Basic Schwarz-Pick type inequalities

Let an−k,n be deﬁned by the Taylor expansion

z s(z)

n =

∞

aν,n z ν .

ν=0

Then the Schur-Jabotinsky theorem (compare for example [81], Theorem 1.9.a) implies that for 1 ≤ k ≤ n the identities An,k =

k an−k,n n

are valid. Hence, we have to prove that n−1 n − l n−l al,n ≤ 2n−1 , y n

y ∈ ∂Δ.

(4.11)

l=0

The tool for the proof of (4.11) is the resulting representation

z s(z)

n = (1 + zω(z))n = 1 +

n n z σ (ω(z))σ , σ

z ∈ Δ,

σ=1

where |ω(z)| ≤ 1 for any z ∈ Δ by the Marx-Strohh¨ acker theorem, that indicates Re(s(z)/z) > 1/2 in Δ (see [115] and [155], see also Section 7.3 below). If we deﬁne (ω(z))σ =

∞

dj,σ z j ,

z ∈ Δ,

j=0

we get the following formula for the sum appearing in (4.11): n−1 l=0

n−1 1 n n−1 n − l n−l al,n = y n + (n − j)y n−j dj−σ,σ . y σ n n σ=1 j=σ

(4.12)

To have the desired inequality it is suﬃcient to prove that n−1 n−j (n − j)y d j−σ,σ ≤ n − σ. j=σ Now we observe that the functions ω σ also map the disc Δ into Δ. Therefore, we may replace the coeﬃcients dj−σ,σ by the coeﬃcients dj−σ of a unimodular bounded function when we consider the latter estimate. Taking p = n − σ gives us an equivalent inequality p−1 p−τ dτ ≤ p. (p − τ ) y τ =0

4.5. Punishing factors for convex pairs

65

This inequality follows directly from Fej´er’s inequality p−1 (p − τ )dτ ≤ p, τ =0

which has long been known to be valid (see [61] and [159]). This concludes the proof of the inequality (4.11). Now, we shall prove that the constant 2n−1 is best possible in any of the cases in question. Lemma 4.16. Let Ω and Π be two convex proper subdomains of C. Then for any n ≥ 2 the inequality Cn (Ω, Π) ≥ 2n−1 is valid. Proof. We know that the constant Cn (Ω, Π) is invariant under linear transformations of Ω and Π. Hence, without restriction of generality, we may assume that Δ1 = {z | |z − 1| < 1} ⊂ Ω ⊂ H1 = {z | Re z > 0}

(4.13)

Δ 1 ⊂ Π ⊂ H1 .

(4.14)

and Let α ∈ (0, 1) and ξ ∈ (0, 1) and consider the function fα (z) = α

z+2 , z+α

z ∈ Ω.

Obviously, fα ∈ A(Ω, Π). By the comparison theorem applied to the inclusion relations (4.13) and (4.14) we get lim λΩ (β) 2β = lim λΠ (β) 2β = 1.

β→0+

β→0+

Now, by use of these asymptotic equalities, we prove Lemma 4.16 with the following chain of inequalities and equations. (n) fα (ξ) λΠ (fα (ξ)) Cn (Ω, Π) ≥ lim lim n ξ→0+ α→0+ n! (λΩ (ξ)) (n) fα (ξ) (2ξ)n = lim lim ξ+2 ξ→0+ α→0+ n! 2α ξ+α α(2 − α) (2ξ)n 2n = lim = 2n−1 . ξ→0+ α→0+ (ξ + α)n+1 2α ξ+2 ξ→0+ 2 + ξ ξ+α

= lim

lim

This concludes the proof of Theorem 4.15.

66

Chapter 4. Basic Schwarz-Pick type inequalities

Remark 4.17. The last part of the proof shows that the constant 2n−1 is approached for any pair of convex domains, when z0 and f (z0 ) approach the boundaries of Ω and Π at certain points. But there are simple special cases where the constant is attained at inner points. For instance, this happens if Ω and Π are half planes. Actually, if Ω = Π = H1 and f0 (z) = 1/z, then, at any point z0 = x > 0, (n)

n

2n−1 (λΩ (x)) f0 (x) 1 = n+1 = , n! x λΠ (1/x) since 1/λH1 (z) = 2 Re z.

4.6 Case n = 2 for all domains Let Ω and Π be hyperbolic domains on the Riemann sphere C that are equipped with the Poincar´e metric of curvature −4. According to Poincar´e’s generalization of Riemann’s mapping theorem, this means that the boundaries of Ω and Π contain at least three points in C and that the density of this metric in the unit disc Δ = {z | |z| < 1} is deﬁned as λΔ (z) =

1 , 1 − |z|2

z ∈ Δ.

We have considered the functionals Ln , Mn and Cn for several special cases. According to [16], in the general case these functionals depend on the hyperbolic characteristics −k

(λΩ (z))

k ∂ k log (λΩ (z)) −k ∂ log (λΠ (w)) and (λ (w)) , 1 ≤ k ≤ n − 1. Π ∂z k ∂wk

Everything is clear in the case n = 2. In particular, the following theorem describes C2 (Ω, Π) in terms of the gradients −1

∂ log (λΩ (z)) ∂z

−1

∂ log (λΠ (w)) . ∂w

∇ (1/λΩ (z)) = −2 (λΩ (z)) and ∇ (1/λΠ (w)) = −2 (λΠ (w))

Theorem 4.18 ([27]). For all hyperbolic domains Ω ⊂ C and Π ⊂ C, 1 sup |∇ (1/λΩ (z))| + sup |∇ (1/λΠ (w))| . C2 (Ω, Π) = C2 (Π, Ω) = 2 z∈Ω w∈Π Proof. Fix (z, w) ∈ Ω×Π, z = ∞ and w = ∞. We consider a function f ∈ A(Ω, Π) such that f (z) = w. Let Φ : Δ → Ω and Ψ : Δ → Π be universal covering

4.6. Case n = 2 for all domains

67

maps such that Φ(0) = z and Ψ(0) = w. We consider the holomorphic function g : Δ → Δ deﬁned by g(ζ) = Ψ−1 (f (Φ(ζ))) =

∞

cn ζ n ,

ζ ∈ Δ.

n=1

By direct computations, one gets the identity 2

f (z) (Φ (0)) 2 Ψ (0) and therefore

= c2 +

Ψ (0) 2 Φ (0) − c c1 , 2Ψ (0) 1 2Φ (0)

y 2 x |f (z)| λΠ (w) = + c + c1 , c 2 2 2 2 1 2 (λΩ (z))

where x =−

Φ (0) Φ (0)

and y =

(4.15)

Ψ (0) . Ψ (0)

If g ∈ A(Δ, Δ) and g(0) = 0, then the function f = Ψ ◦ g ◦ Φ−1 belongs to the set A(Ω, Π) and f (z) = w. Therefore, we have to ﬁnd y x (4.16) S(x, y) = max c2 + c21 + c1 | g ∈ A(Δ, Δ) with g(0) = 0 . 2 2 Using classical results on the Taylor coeﬃcients of unimodular bounded functions (see [148]) we get y x S(x, y) = max c2 + c21 + c1 | |c1 | ≤ 1, |c2 | ≤ 1 − |c1 |2 2 2 and, by a little analysis, |y| 2 |x| 1 t + t | t ∈ [0, 1] = F (|x|, |y|), S(x, y)) = max 1 − t2 + 2 2 2 where the function F is deﬁned as p + q, F (p, q) := p2 , 2 + 8−4q

if p + 2q ≥ 4, if p + 2q < 4,

for p ≥ 0 and q ≥ 0. Combining this with (4.15) and (4.16) one gets max{|f (z)| | f ∈ A(Ω, Π), ﬁxed w = f (z)} = F (|x|, |y|)

2

(λΩ (z)) , λΠ (w)

where |x| = p = |∇ (1/λΩ (z))| and |y| = q = |∇ (1/λΠ (w))| .

(4.17)

68

Chapter 4. Basic Schwarz-Pick type inequalities For a domain Ω, by formula (3.6) in the form 2 Φ (ζ) |∇ (1/λΩ (z))| = (1 − |ζ| ) − 2ζ Φ (ζ)

and by Theorem 3.23 it is known that sup |∇ (1/λΩ (z))| = 2

z∈Ω

if and only if Ω is convex and that sup |∇ (1/λΩ (z))| > 2

z∈Ω

in all other cases. Accordingly, for all hyperbolic domains Ω and Π, sup |∇ (1/λΩ (z))| ≥ 2,

z∈Ω

sup |∇ (1/λΠ (w))| ≥ 2.

w∈Π

Moreover, we remark that, for a ≥ 2 and b ≥ 2, one easily gets the identity max{F (p, q) | p ∈ [0, a], and q ∈ [0, b]} = a + b, since the condition p + 2q < 4 implies the chain of inequalities 2+

p p2 < 2+ < a + b. 8 − 4q 2

These together with formula (4.17) imply the assertion of Theorem 4.18.

Using Theorem 3.23 and Theorem 4.18 we obtain Corollary 4.19. Suppose that Ω and Π are hyperbolic domains in C. Then the equation C2 (Ω, Π) = 2 is valid if and only if Ω and Π are convex domains in C. Using formulas (3.6) and (3.42) to compute the above gradients, one easily gets the following corollaries of Theorem 4.18. Corollary 4.20. Let α ∈ [1, 2], and let g(ζ) = (ζ + 1)α , g(0) = 1. If Ωα := g(Δ), then C2 (Ωα , Ωα ) = 2 α. Observe that (Ω2 , Ω2 ) is an extremal pair in Theorem 4.10 and compare Section 5.5. Corollary 4.21. Let A1 and A2 be annuli with moduli M1 and M2 , respectively. Then C2 (A1 , A2 ) = 1 + 4M12 + 1 + 4M22 . Using Theorem 3.12, equation (3.37) and Theorem 4.18 we immediately obtain the following assertion. Corollary 4.22. Suppose that Ω and Π are hyperbolic domains in C. The constant C2 (Ω, Π) is ﬁnite if and only if Ω and Π both are domains with uniformly perfect boundary. If ∞ ∈ Ω or ∞ ∈ Π and n ≥ 2, then it is easy to verify that Cn (Ω, Π) = ∞ .

Chapter 5

Punishing factors for special cases After a colloquium talk of the second author on estimates of the form (λΩ (z))n |f (n) (z)| ≤ Cn (Ω, Π) , n! λΠ (f (z))

f ∈ A(Ω, Π), z ∈ Ω,

(5.1)

for simply connected domains Ω and Π in C, Ch. Pommerenke ([132]) proposed to look at (5.1) in the following way. The quotient (λΩ (z))n /λΠ (f (z)) reﬂects the inﬂuence of the positions of the points z and f (z) in Ω and Π on the nth derivative f (n) (z), whereas the quantities Cn (Ω, Π) are factors punishing bad behaviour of Ω or Π at the boundary. This motivates the title of the present chapter as well as the titles of some our papers.

5.1

Solution of the Chua conjecture

In [56] Chua published the following conjecture among others. Chua’s conjecture. Let Ω be a convex proper subdomain of C and let f be holomorphic and injective on Ω. Then for any z ∈ Ω and any n ≥ 2 the inequality (n) f (z) n−1 n−2 (λΩ (z)) (5.2) f (z) n! ≤ (n + 1) 2 holds true. In [56], Chua settled this conjecture for n = 2, 3, 4 (see [102] and [103] for the cases n = 5, 6, 7, 8). Also, taking the limit z → 1, z ∈ (0, 1), in (5.2) for the Koebe function shows that the constant (n + 1)2n−2 on the right side of (5.2) can not be replaced by a smaller one.

70

Chapter 5. Punishing factors for special cases

In the paper [24] we proved that for Ω convex, Π linearly accessible, and n ≥ 2, the inequality Cn (Ω, Π) ≤ (n + 1)2n−2 (5.3) is valid. The equation λΩ (z) , z ∈ Ω, λΠ (f (z)) holds for functions f injective on Ω. Hence, the inequality (5.3) implies the validity of Chua’s conjecture for f that map Ω conformally onto a linearly accessible domain Π. But, in fact more is true. |f (z)| =

Theorem 5.1 ([25]). Let Ω be a convex proper subdomain of C, and Π be a simply connected proper subdomain of C. Let further n ≥ 2. Then the inequality (5.3) is valid. Proof. Because of the central formula (4.5), for n ≥ 2 it is evident that (5.3) will follow from the inequality n ck An,k ≤ (n + 1)2n−2 , (5.4) k=1

where An,k are the coeﬃcients connected with the inverse of a function s ∈ K, and ck are described by the condition that the sum of the series ∞

ck z k =: g1 (z),

z ∈ Δ,

(5.5)

k=1

is subordinate to a member of the family S. Because of the Schur-Jabotinsky theorem and the Marx-Strohh¨ acker inequality we have to prove that n−1 n − l cn−l al,n n l=0 n−1 1 n n−1 = cn + (n − j)cn−j dj−σ,σ ≤ (n + 1)2n−2 . (5.6) σ n σ=1

j=σ

We may replace the coeﬃcients dj−σ,σ by the coeﬃcients dj−σ of a unimodular bounded function when we estimate the modulus of the inner sum in (5.6). Using 2 n σ n = (n + 1)2n−2 , σ n σ=1 it is easily seen that the desired inequality follows from the inequality n−1 (n − j)cn−j dj−σ,σ ≤ (n − σ)2 . j=σ

5.1. Solution of the Chua conjecture

71

This is a consequence of the following lemma with p = n − σ.

Lemma 5.2. Let ω ˜ (z) =

∞

dτ z τ

τ =0

be holomorphic in the unit disc and such that ω ˜ (Δ) ⊂ Δ and let g1 be the function deﬁned by the equation (5.5) and subordinate to a function from S. Then for p ∈ N the inequality p−1 (5.7) (p − τ ) cp−τ dτ ≤ p2 τ =0

is valid. Proof. We shall use Sheil-Small’s theorem 2.9 which says that If g1 is subordinated to a function g ∈ S and P is a polynomial of degree ≤ p, then for z ∈ Δ the inequality |(P ∗ g1 )(z)| ≤ p max{|P (z)| , |z| = 1}

(5.8)

is valid. To prove (5.7) with the help of (5.8) we consider the polynomial P (z) =

p−1

dτ (p − τ )z p−τ .

τ =0

Because of the identity (P ∗ g1 )(1) =

p−1

(p − τ ) cp−τ dτ ,

τ =0

it is suﬃcient for the proof of (5.7) to show that −iθ P e ≤ p, θ ∈ [0, 2π]. Since the family of functions ω ˜ is invariant against rotations of the unit disc, it remains to prove the inequality p−1 (p − τ )dτ ≤ p. τ =0

This exactly is Fej´er’s inequality, used above. This concludes the proof of Lemma 5.2. Theorem 5.1 follows immediately.

72

Chapter 5. Punishing factors for special cases

5.2

Punishing factors for angles

We again consider the quantities Ln (f, z, Ω, Π), Mn (z, Ω, Π) and Cn (Ω, Π) deﬁned by n 1 (n) (λΩ (z)) , n ∈ N, f ∈ A(Ω, Π), z ∈ Ω, f (z) = Ln (f, z, Ω, Π) n! λΠ (f (z)) Mn (z, Ω, Π) := sup{Ln (f, z, Ω, Π) | f ∈ A(Ω, Π)}, and Cn (Ω, Π) := sup{Mn (z, Ω, Π) | z ∈ Ω}. In this section we are concerned with the above quantities where one or both of the domains in question belong to the class of angular domains aHα + b with opening angle απ, 1 ≤ α ≤ 2, which means that there exists a linear transformation T (z) = az + b such that aHα + b := T (Hα ), where απ Hα = z | | arg z| < . 2 Theorem 5.3 ([17]). Let 1 ≤ α ≤ 2, n ∈ N, and z0 ∈ Δ. Then Mn (z0 , Δ, Hα ) =

n−1

(1 + |z0 |)k

k=0

where

α ν

=

n−1 k

α n−k

2n−k−1 , α

ν−1 1 (α − μ). ν! μ=0

If we let |z0 | → 1, then we immediately obtain the following result. Corollary 5.4. Let 1 ≤ α ≤ 2 and n ∈ N. Then 2n−1 n+α−1 . Cn (Δ, Hα ) = n α Proof of Theorem 5.3. The formula (4.3) indicates that we may insert into (4.5) for Ω = Δ the identities n−1 An,k (z0 ) = z0 n−k . n−k Next, we set in Theorem 2.19, g(ζ) = f

ζ + z0 1 + z0 ζ

,

5.2. Punishing factors for angles

73

and deﬁne the Taylor coeﬃcients dk (α) by α ∞ 1+ζ 1 −1 = dk (α)ζ k . 2α 1−ζ k=1

Since the coeﬃcients dk (α) are nonnegative in our cases, we see that in (4.5) we may use the inequalities |ak |λHα (f (z0 )) ≤ dk (α) to get

n n−1 |z0 |n−k dk (α). Mn (z0 , Δ, Hα ) ≤ n−k

(5.9)

k=1

It is easily seen that in this inequality the upper bound is attained for z0 = r0 eiθ , r0 ≥ 0, if we choose f to be given by the identity α ζ + z0 eiθ 1 + e−iθ ζ − 1 . f − f (z0 ) λHα (f (z0 )) = 1 + z0 ζ 2α 1 − e−iθ ζ It remains to show that the upper bound for Mn (z0 , Δ, Hα ) in (5.9) equals the upper bound given in Theorem 5.3. To this end we remark that the left side of the inequality (5.9) is the nth Taylor coeﬃcient of the function α (1 + |z0 |z)n−1 1 + z . 2α 1−z Use of the binomial theorem for n−1

(1 + |z0 |z)n−1 = ((1 + |z0 |)z + 1 − z)

reveals that the said nth Taylor coeﬃcient equals the nth Taylor coeﬃcient of the product n 1 n−1 α z k−1 (1 + |z0 |)k−1 (1 − z)n−k−α . (1 + z) n−k 2α k=1

Hence, we have to sum up the (n − k + 1)th Taylor coeﬃcients of the functions 1 n−1 α (1 + |z0 |)k−1 (1 − z)n−k−α , k = 1, . . . , n. (1 + z) n−k 2α These coeﬃcients may be written in the form n−k+1 α (−1)n−k+1−q n − k − α n−1 (1 + |z0 |)k−1 n−k+1−q q n−k 2α q=0 n−k 2 n−1 α = (1 + |z0 |)k−1 , k = 1, . . . , n. n−k n−k+1 α This proves the result of Theorem 5.3 if we replace k by k − 1 and sum up from k = 0 to k = n − 1.

74

Chapter 5. Punishing factors for special cases

Using theorems on inverse coeﬃcients of conformal mappings we generalize Corollary 5.4 and Yamashita’s results in [170] (see Theorem 4.9) as follows. Theorem 5.5 ([17]).

a) Let 1 ≤ α, β ≤ 2 and n ∈ N. Then α (2β)n +n−1 β Cn (Hβ , Hα ) = . n 2α

b) Let 1 ≤ β ≤ 2, n ∈ N, and Π be a convex proper subdomain of C . Then Cn (Hβ , Π) = Cn (Hβ , H1 ). c) Let 1 ≤ β ≤ 2, n ∈ N, and Π be a simply connected proper subdomain of C. Then Cn (Hβ , Π) ≤ Cn (Hβ , H2 ). d) Let 1 ≤ α ≤ 2, n ∈ N, and Ω be a simply connected proper subdomain of C. Then Cn (Ω, Hα ) ≤ Cn (H2 , Hα ). Remark 5.6. For all α, β ∈ [1, 2], one has that C1 (Hβ , Hα ) = 1 in accordance with the Schwarz-Pick inequality. In the case n ≥ 2, n−1 2n Cn (Hβ , Hα ) = (α + kβ). n! k=1

We see that C2 (Hβ , Hα ) = C2 (Hα , Hβ ) = α + β in accordance with Theorem 4.18, and Cn (Hα , Hβ ) > Cn (Hβ , Hα ) > 2n−1 , whenever n ≥ 3 and α > β ≥ 1. Proof of Theorem 5.5. a) Firstly, we notice that the nth Taylor coeﬃcient of −1 k (h−1 β (w)) is nonnegative, since the Taylor coeﬃcients of hβ (w) are all nonnegative. This fact and Theorem 2.20 imply that we now may use as an upper bound for the quantity |An,k (z0 )|, 1 ≤ k ≤ n, in formula (4.5) this nth Taylor coeﬃcient k of (h−1 β (w)) . For |ak | we use the same inequality as in the proof of Theorem 5.3. To prove that the resulting inequality is sharp and to compute the explicit formula given in the assertion, we use the conformal map fα,β deﬁned by

(z) , (5.10) fα,β (z) := gα h−1 β where gα (ζ) :=

1 2α

1+ζ 1−ζ

α

−1 ,

ζ ∈ Δ.

5.2. Punishing factors for angles

75

One one hand, fα,β maps the special angular domain 1 β π < Hβ = z arg z − 2β 2 conformally onto the special angular domain 1 α π < Hα = z arg z + 2α 2 (n)

such that fα,β (0) = 0. Now, we compute fα,β (0) starting with formula (5.10) in the same way as we obtained (4.5). If we use that in our circumstances λHβ (0) = λHα (0) = 1, we observe that the resulting sum is just the sum that we derived as an upper bound in the beginning of this proof. On the other hand, a direct computation of fα,β shows that fα,β (z) = Hence,

α 1 (1 − 2β z)− β − 1 . 2α

(n)

fα,β (0) n!

=

(2β)n 2α

α β

+n−1 n

.

This proves the assertion in question. b) The inequality Cn (Hβ , Π) ≤ Cn (Hβ , H1 ) for a convex domain Π follows in our case from Rogosinski’s theorem 4.12 on convex subordination according to which we may apply |ak |λΠ (z0 ) ≤ 1 in (4.5) in a way analogous to that of the proof of the case a). The proof of Cn (Hβ , Π) ≥ Cn (Hβ , H1 ) for a convex domain follows the proof of Mn (z, Δ, Π) = (1 + |z|)n−1 for convex Π in Chapter 4. Therefore, we only describe the crucial steps here. Without loss of generality, we may assume that Δ1 = {w | |w − 1| < 1} ⊂ Π ⊂ H1 = {w | Re w > 0}.

76

Chapter 5. Punishing factors for special cases

For the special angular domain Hβ chosen as in the proof of the case a) we consider the holomorphic functions fk (z) =

2 1

1 + k(1 − 2β z) β

,

k ∈ N,

which map Hβ conformally onto Δ1 such that fk (0) = 2/(k + 1) =: wk . We compute (k + 1)2 λHβ (0) = 1, λΔ1 (wk ) = . 4k Since the sequences (kfk )k∈N are uniformly convergent on a neighbourhood of the origin we see that 1 dn (n) −β lim kfk (0) = 2 (1 − 2β z) = 4 (n!)Cn (Hβ , H1 ). k→∞ dz n z=0 The last equation follows from the proof of the case a). Using the comparison principle for densities of the Poincar`e metric (see Theoren 3.5) as in the proof of the corresponding relation in Chapter 4, it is easily seen that lim

k→∞

λΠ (wk ) = 1. λΔ1 (wk )

Gathering together the above relations we get the equation f (n) (0)λΔ1 (wk ) λΠ (wk ) n = Cn (Hβ , H1 ), lim Ln (fk , 0, Hβ , Π) = lim k→∞ k→∞ λΔ1 (wk ) n! λHβ (0) which immediately implies the assertion. c) In this proof we only have to use again the validity of the Rogosinski or generalized Bieberbach conjecture together with the case a) of Theorem 5.5. d) Here we use in the application of (4.5) that, according to L¨ owner’s Theorem 2.13, the Taylor coeﬃcients of the inverses of schlicht functions and their kth powers, k ∈ N, are dominated by the related (positive) coeﬃcients of the inverse of the Koebe function z/(1+z)2 and its kth powers. This results in the assertion. In our next result we are concerned with proper subdomains Ω of C simply connected with respect to C and holomorphic functions f : Ω → Hα . In the sequel we use the quantity p ∈ (0, 1] such that p := p(z0 ) = tanh DΩ (z0 , ∞). By deﬁnition, p = 1, if ∞ ∈ Ω. The following theorem forms a bridge between the results for Π convex and Π simply connected with respect to C.

5.2. Punishing factors for angles

77

Theorem 5.7 ([17]). Let Ω be as above, z0 ∈ Ω \ {∞}, 1 ≤ α ≤ 2, and n ≥ 2. Then for any holomorphic function f : Ω → Hα , the inequality α n−k n 1 (n) (λΩ (z0 ))n 4k 1 n−1 2 p(z0 ) + |f (z0 )| ≤ +2 n−k k n! λHα (f (z0 )) 2α p(z0 ) k=1

is valid. For each n ≥ 2, each α ∈ [1, 2], and each p ∈ (0, 1] there exist Ω, Hα , z0 ∈ Ω \ {∞}, and f as above such that equality is attained in the above inequality. Proof. We again start with the central equality (4.5). Let p = p(z0 ),

1 a = p+ , p

ζ (1 + p ζ)(1 +

κp (ζ) =

ζ p)

, ζ ∈ Δ,

and Kp the function inverse to κp having an expansion Kp (z) = z +

∞

Bν z ν

ν=2

valid in some neighbourhood of the origin. Considering the class Sp of functions meromorphic and univalent in Δ which have a pole in a point b, |b| = p, and an expansion ∞ m(ζ) = ζ + Aν ζ ν , |ζ| < p, ν=2

Baernstein and Schober showed in [33] (see Theorem 2.15) that |Aν | ≤ Bν , ν ≥ 2. This implies that the nth Taylor coeﬃcients of the kth powers of m ∈ Sp are dominated by the nth Taylor coeﬃcients Bn,k of the kth powers of the function Kp . By a procedure analogous to that, which led us to formula (4.5), we get in our case for the meromorphic functions f in question the inequality |f (n) (z0 )| (λΩ (z0 )) Bn,k dk (α). ≤ n! λHα (f (z0 )) n

n

(5.11)

k=1

To show that this inequality is sharp and that the right-hand side has the form given in the assertion, we proceed in principle as in the proof of the case a) of Theorem 5.5. To this end we consider in the present case the function fα,p (z) := gα (Kp (z)) which maps the domain Ω0 = C \ [(a + 2)−1 , (a − 2)−1 ]

78

Chapter 5. Punishing factors for special cases

conformally onto the special angular domain Hα chosen as in the proof . We see that in this case fα,p (0) = 0 and λHα (0) = λΩ0 (0) = 1 and we deduce the sharpness (n) of (5.11) considering fα,p (0) as above. To get the explicit form of the upper bound we consider α ∞ 1 − (a − 2)z 2 1 fα,p (z) = − 1 =: Cn (α, p)z n . 2α 1 − (a + 2)z n=1 This implies 1+

∞

α2 k−1 k

4(a + 2)

z

= 1 + 2α

∞

Cn (α, p)z n .

n=1

k=1

With the abbreviation Z := z(a + 2) this is equivalent to the equation k k ∞ α ∞ 4 Z Cn (α, p) n 2 = 1 + 2α Z . 1+ k a+2 1−Z (a + 2)n n=1 k=1

If we then expand Z k /(1 − Z)k in powers of Z and compare the coeﬃcients on both sides of this equation, we then establish the assertion of Theorem 5.7 on the sharpness of the punishing factor.

5.3

Sharp lower bounds for punishing factors

We shall consider domains with a special local property at a boundary point, which will be a metrical characterization of the “bad” behavior of the boundary. Let Ω be a domain in C. We denote aΩ + b := {z = aζ + b | ζ ∈ Ω}, and

1 := Ω

z=

a, b ∈ C, a = 0,

1 |ζ∈Ω . ζ

For α ∈ [1, 2] we deﬁne α Δ+ α := {z = (ζ + 1/2) | |ζ| < 1/2},

Δ− 0 := [−1, 0] and Δ− 2−α := {z = −

w |w= w+1

1−ζ 1+ζ

2−α , |ζ| ≤ 1}

for α ∈ [1, 2),

where the branch of a power is ﬁxed by 1α = 1. Observe that z = 0 is a boundary − point for all sets Δ+ α and Δ2−α .

5.3. Sharp lower bounds for punishing factors

79

Deﬁnition 5.8. Let Ω be a domain in C, let α ∈ [1, 2], and let z0 ∈ (∂Ω) \ {∞}. If there exist ε > 0 and t ∈ R such that it − εeit Δ+ α + z0 ⊂ Ω and εe Δ2−α + z0 ⊂ C \ Ω,

(5.12)

then we say that z0 is an angular point on ∂Ω of order α. Clearly, the condition (5.12) is equivalent to the inclusions − Δ+ α ⊂ aΩ + b ⊂ C \ Δ2−α ,

a=

e−it , ε

b = −z0

e−it . ε

We extend this deﬁnition to the boundary point at inﬁnity. Deﬁnition 5.9. Let α ∈ [1, 2] and let Ω be a domain in C such that ∞ ∈ ∂Ω. If there exist complex numbers a = 0 and b such that Δ+ α ⊂

1 ⊂ C \ Δ− 2−α , aΩ + b

then we say that the point at inﬁnity is an angular point on ∂Ω of order α. According to Deﬁnitions 5.8 and 5.9 the domain απ Hα := z ∈ C | |argz| < 2 has two boundary angular points of order α that are z0 = 0 and the point at inﬁnity. A non-trivial example is ∞

Ω1 = H1 \

Dk , k=1

where {z ∈ C |

Dk =

|z − 1/k − it| ≤ (t + 2k)−2 }.

0≤t 0. Obviously, this function has an expansion ∞ k (5.18) ΨΩ (z) = λΩ (z0 ) z − z0 + Ak (z − z0 ) , k=2

valid in some neighbourhood of the point z0 . Let b = ΨΩ (∞) ∈ Δ. Because of the conformal invariance of the hyperbolic distance, DΔ (0, b) equals DΩ (z0 , ∞). According to (5.16) we obtain |b| = p = p(z0 ). Let the function κ(ζ) =

ζ

, (1 + p ζ) 1 + pζ

ζ ∈ Δ,

(5.19)

which is univalent in Δ, have an inverse function K = κ−1 deﬁned on κ(Δ). Obviously ∞ ∈ κ(Δ) and 0 ∈ κ(Δ). Using the expansion K(z) = z +

∞

Bk z k ,

k=2

valid in some neighbourhood of the origin, Baernstein and Schober (see Chapter 2, Theorem 2.15) showed that |Ak | ≤ Bk ,

k ∈ N \ {1}.

(5.20)

Consider now for each f ∈ A(Ω, Π) the function ∞

g(ζ) := f (ΦΩ (ζ)) =

ak ζ k ,

ζ ∈ Δ,

(5.21)

k=0

where ΦΩ is inverse to ΨΩ . It is known by formula (4.5) that 1 f (n) (z0 ) An,k (z0 , Ω) ak , n = n! (λΩ (z0 )) n

(5.22)

k=1

where (λΩ (z0 , Ω)) An,k (z0 , Ω) = n!

−n

(n) (ΨΩ (z)) k

.

(5.23)

z=z0

The formulas (5.18) and (5.23) imply that An,n (z0 , Ω) ≡ 1 and that the An,k (z0 , Ω), k = 1, . . . , n, are polynomials with positive coeﬃcients in A1 , . . . , An−1 . Hence, we get as a consequence of this fact and of the formulas (5.18)–(5.23) the inequalities (n) 1 k |An,k (z0 , Ω)| ≤ tn,k := . (5.24) (K(z)) n! z=0

86

Chapter 5. Punishing factors for special cases

Thus,

1 |f (n) (z0 )| tn,k |ak |, n ≤ n! (λΩ (z0 )) n

(5.25)

k=1

where the positive quantities tn,k may be computed, according to the formulas (5.19), (5.24), and the Cauchy formula, as 1 1 (1 + aζ + ζ 2 )n−1 (1 − ζ 2 ) tn,k = d ζ, a = + p, r ∈ (0, 1). (5.26) 2π i |ζ|=r ζ n−k+1 p Let q = tanh DΠ (f (z0 ), ∞) ∈ (0, 1]. We need the expansion ∞ ζ

=ζ+ ck (q)ζ k , ζ (1 − qζ) 1 − q k=2

|ζ| < q.

It is known that (see Chapter 2) k−1

j=0 q q k−1

ck (q) =

2j

=

qk − q−

1 qk 1 q

.

The properties tn,k tn,k+1 ≥ > 0, ck (q) ck+1 (q) and

n

ck (q) tn,k =

k=1

1 a+q+ q

k = 1, . . . , n − 1, n−1 ,

a=

(5.27)

1 + p, p

(5.28)

of the quantities tn,k will be proved below. Let us suppose for the moment that they are true. Applying the Cauchy inequality to (5.25) we get 1 |f (n) (z0 )| ≤ n! (λΩ (z0 ))n

n

12 ck (q) tn,k

k=1

n tn,k |ak |2 ck (q)

12 .

(5.29)

k=1

Let ΦΠ be the conformal map of Δ onto Π with ΦΠ (0) = a0 = f (z0 ) and ΦΠ (0) = (λΠ (f (z0 )))−1 > 0. The function g deﬁned in (5.21) is subordinate to the univalent function ∞ −1 k ΦΠ (ζ) = a0 + (λΠ (f (z0 ))) ζ+ . ck ζ k=2

The starting point for the rest of the proof is Theorem 2.5 which is a generalized version of the Rogosinski-Goluzin theorem. The second ingredient at this point is Jenkins’ theorem 2.7. We easily get 2

(λΠ (f (z0 )))

n n n tn,k tn,k ck (q) tn,k . |ak |2 ≤ |ck |2 ≤ ck (q) ck (q)

k=1

k=1

k=1

(5.30)

5.4. Domains in the extended complex plane

87

From (5.28), (5.29) and (5.30) we obtain the desired inequality (5.17). To complete the proof of (5.17) we have to prove (5.27) and (5.28). An immediate consequence of (5.26) for r ∈ (0, q) is n

ck (q) tn,k =

k=1

1 2π i

1 = 2π i

|ζ|=r

n (1 + aζ + ζ 2 )n−1 (1 − ζ 2 ) ck (q) ζ k−1 d ζ ζn k=1

I(ζ) d ζ = Res(I(ζ), ζ = 0) |ζ|=r

= −Res(I(ζ), ζ = q) − Res(I(ζ, ζ = where I(ζ) =

1 ) − Res(I(ζ), ζ = ∞), q

(1 − ζ 2 ) (1 + aζ + ζ 2 )n−1

. ζn (q − ζ) 1q − ζ

We observe that Res(I(ζ), ζ = 0) = Res(I(ζ), ζ = ∞) and Res(I(ζ), ζ = q) = Res(I(ζ), ζ = 1q ). Therefore n

ck (q) tn,k = −Res(I(ζ), ζ = q) =

k=1

n−1 1 q+ +a . q

To prove (5.27) we consider the polynomial 2 n

Pn (ζ) = (1 + aζ + ζ ) =

2n

bm ζ m ,

(5.31)

m=0

where, as before, a = p + 1/p > 2. The identity Pn (ζ) = (1 − ζ 2 )Pn−1 (ζ) + ζ Pn−1 (ζ) P1 (ζ) implies Pn (ζ) −

ζ P (ζ) = (1 − ζ 2 )Pn−1 (ζ). n n

Using (5.26) we identify tn,k as the coeﬃcient of ζ n−k in the polynomial on the right side of this equation. On the other hand, (5.31) shows that bn−k − n−k n bn−k is the coeﬃcient of ζ n−k in the polynomial on the left side. Therefore, we obtain n

tn,k = bn−k , k

k = 1, . . . , n.

This implies that the inequalities tn,k+1 tn,k ≥ > 0, k k+1

k = 1, . . . , n − 1,

(5.32)

88

Chapter 5. Punishing factors for special cases

are equivalent to 1 ≤ b1 ≤ b2 ≤ · · · ≤ bn−1 , which may be proved by mathematical induction on n considering the coeﬃcients of the polynomials (1 − ζ) Pn (ζ). Moreover, (5.32) implies the desired inequalities (5.27) because of the simple relations ck+1 (q) k+1 ≥ , ck (q) k

k = 1, . . . , n − 1.

Now, it remains to prove that (5.17) is sharp. To this end we consider the following example. For any n ∈ N and any a = p + 1/p > 2, let Ω0 = C \ [(a + 2)−1 , (a − 2)−1 ], f0 (z) =

z

1− a+q+

1 q

z

and Π0 = f0 (Ω0 ). It is an easy task to derive that λΩ0 (0) = λΠ0 (0) = 1, that tanh DΩ0 (0, ∞) equals p, tanh DΠ0 (0, ∞) equals q and that (n)

f0 (0) = n!

p+

1 1 +q+ p q

n−1 .

This completes the proof of Theorem 5.15. Consider now an improvement of (5.17) for convex domains.

If the domain Π is convex, then the coeﬃcients ak , k ∈ N, of the functions g deﬁned in (5.21) satisfy the inequalities of Rogosinski (see Theorem 4.12) λΠ (a0 )|ak | ≤ 1,

k ∈ N.

(5.33)

Therefore (5.25) and (5.33) imply n (λ (z ))n 1 (n) Ω 0 tn,k , f (z0 ) ≤ n! λΠ (f (z0 ))

(5.34)

k=1

where we used the same deﬁnitions as in the proof of Theorem 5.15. A straightforward computation using (5.26) shows that n k=1

tn,k =

1 π

π

p+

0

1 + 2 cos θ p

n−1 (1 + cos θ) d θ =: Cn (p),

(5.35)

5.4. Domains in the extended complex plane

89

if ∞ ∈ Ω and therefore p ∈ (0, 1), respectively n

tn,k =

k=1

2n − 1 n

,

(5.36)

if ∞ ∈ Ω. The formulas (5.34), (5.35) and (5.36) immediately yield Theorem 5.16 ([19]). Suppose that Ω is a hyperbolic simply connected proper subdomain of C and that Π is a convex proper subdomain of C, K = −4. Then the following estimates are valid: If ∞ ∈ Ω, then for any f ∈ A(Ω, Π) and any z ∈ Ω with hyperbolic distance DΩ (z, ∞) = Arctanh p, the inequality n

(λΩ (z)) 1 (n) |f (z)| ≤ Cn (p) n! λΠ (f (z))

(5.37)

holds for any n ∈ N. Remark 5.17. The sharpness of (5.37) can be shown using Ω0 , H1 and the function f1 deﬁned by ! 2f1 (z) :=

1 + K(z) = 1 − K(z)

1 − (a − 2)z . 1 − (a + 2)z

(5.38)

Indeed, the function f1 maps Ω0 conformally onto the half plane H1 , K(0) = 0 and λΩ0 (0) = λΠ1 (1/2) = 1. Now, we compare two expansions ∞

f1 (z) =

∞

1 1 (K(z))k = + dn z n , + 2 2 n=1

(5.39)

k=1

which are valid in some neighbourhood of the origin. The formulas (5.38), (5.39), and (5.24) immediately yield dn =

n n (n) 1 = tn,k = Cn (p). (K(z))k n! z=0 k=1

Thus,

(5.40)

k=1

n (λΩ0 (0)) 1 (n) . f1 (0) = Cn (p) n! λΠ1 (f1 (0))

Moreover, using (5.39), (5.40) and the Taylor series of (f1 )2 in some neighbourhood of the origin, we get

∞

1 Ck (p) z k + 2 k=1

2

∞

=

1 (a + 2)n−1 z n , + 4 n=1

90

Chapter 5. Punishing factors for special cases

which implies C1 (p) = 1, Cn (p) =

C2 (p) = a + 1 = p +

1 + 1, and p

n−1 n−2 1 p+ +2 − Ck (p)Cn−k (p), p

n ≥ 3.

k=1

These formulas show the diﬀerence between the bounds in (5.17) and (5.37). Remark 5.18. It may be worthwhile to notice that the direct determination of the integral (5.35) and the computation of the Taylor coeﬃcients of the function (5.38) with the help of binomial series can be exploited to prove relations between binomial coeﬃcients that seem to be not known. We were not able to ﬁnd another method to prove these identities than the geometrical method above. For the convenience of the interested reader we mention such a formula. For m = 0, . . . , n− 1, n ∈ N, one may get in this way the identities n−1 "m # (−1)n m m 2 ⎛ ⎞ n 1/2 −1/2 ⎝ k n−k m j ⎠, (−1) =2 k n−k j m+1−j j

k=0

where j varies in the second sum in the range determined by the binomial coeﬃcients.

5.5

Maps from convex into concave domains

We shall consider the case when ∞ ∈ Π and the punishing factor is a function of the hyperbolic distance between f (z) and ∞. Moreover, we suppose that Π = C \ Π1 , where Π1 is a compact convex subset of C containing more than one point. Domains Π of this type will be called concave domains. In the sequel we use the quantity p ∈ (0, 1) such that DΔ (0, p) =

1 1+p log 2 1−p

equals DΠ (f (z0 ), ∞). This means that p := p(f (z0 )) = tanh DΠ (f (z0 ), ∞),

(5.41)

where, as usual, tanh x = (ex − e−x )/(ex + e−x ). A central part in the proofs is played by functions h satisfying the following conditions:

5.5. Maps from convex into concave domains

91

(i) The function h is meromorphic and injective on Δ and h has its pole at a point p ∈ (0, 1). (ii) The set C \ h(Δ) is convex. (iii) h(0) = h (0) − 1 = 0. We called functions with the properties (i)–(iii) concave univalent functions with pole p and for the family of those functions we used the abbreviation Co(p). For older results on such functions compare [123], [117] and [108] and for newer ones [30], [31], [15], [20], [23],[166] and [167]. Theorem 5.19 ([28]). Let Π be a concave domain. Then for all n ∈ N, f ∈ A(Δ, Π), z0 ∈ Δ, f (z0 ) ﬁnite, and p as in (5.41) the inequalities n−1 1 |f (n) (z0 )| (1 − |z0 |2 )n 1 2 n−1 |z0 | + (5.42) − p (|z0 | + p) ≤ n! RΠ (f (z0 )) 1 − p2 p are valid. Equality in (5.42) is attained if z0 = r ∈ [0, 1), and f = E ◦ T , E(z) =

z (1 − zp)(1 − z/p)

and

T (z) =

z−r , 1 − zr

z ∈ Δ.

These functions f are conformal maps of Δ onto the domain % $ −p −p . C\ , (1 − p)2 (1 + p)2 Corollary 5.20 (compare [31]). Let Π be a concave domain. Then for all n ≥ 2, f ∈ A(Δ, Π), z0 ∈ Δ, f (z0 ) ﬁnite, and p as in (5.41) the inequalities n |f (n) (z0 )| (1 − |z0 |2 )n (1 + p)n−2 k p ≤ n! RΠ (f (z0 )) pn−1

(5.43)

k=0

are valid. The constant on the right side cannot be replaced by a smaller one independent of Π, f, z0 , and f (z0 ). Proof of Theorem 5.19. We consider the function g ∈ A(Δ, Π) deﬁned by ζ + z0 g(ζ) = f , ζ ∈ Δ. 1 + z0 ζ Since f (z0 ) is ﬁnite, the function g has a Taylor expansion g(ζ) = f (z0 ) +

∞

ak ζ k

k=1

valid on some neighbourhood of the origin. It is clear that, under our circumstances, we may also use the identity (4.3), which has been proved for functions holomorphic in the unit disc.

92

Chapter 5. Punishing factors for special cases

On the other hand, we use that g(0) = f (z0 ) and g(Δ) ⊂ Π. If we let, for this proof, Φ := ΦΠ,f (z0 ) be deﬁned as above, we see that there exists a function ω1 : Δ → Δ holomorphic on Δ such that g(ζ) = Φ(ζω1 (ζ)),

ζ ∈ Δ.

(5.44)

Further, we conclude from the deﬁnition p = tanh DΠ (f (z0 ), ∞) that there exists ϕ ∈ [0, 2π] such that Φ(peiϕ ) = ∞. Now, we consider the function h deﬁned on Δ by e−iϕ h(ζ) = (5.45) (Φ(eiϕ ζ) − f (z0 )), ζ ∈ Δ. RΠ (f (z0 )) This function h belongs to the class Co(p) and from (5.44) and (5.45) we see that there exists a function h ∈ Co(p) and a function ω = e−iϕ ω1 : Δ → Δ such that g(ζ) = eiϕ h(ζω(ζ))RΠ (f (z0 )) + f (z0 ),

ζ ∈ Δ.

(5.46)

In Theorem 8.4 below (compare [26]) we will consider h ∈ Co(p), ω as above such that the Taylor expansion h(ζω(ζ)) =

∞

αk ζ k

(5.47)

k=1

is valid on some neighbourhood of the origin. We will prove there that for k ∈ N the inequalities 1 1 − p2k 1 2 k−1 |αk | ≤ (5.48) = −p p (1 − p2 )pk−1 1 − p2 pk−1 are valid. After these preparations (5.42) is an immediate consequence of (4.3), (5.46), (5.47), and (5.48). The second assertion of Theorem 5.19 becomes clear from the above proof and the expansion E(z) =

∞ k=1

1 − p2k zk , (1 − p2 )pk−1

|z| < p.

This concludes the proof of Theorem 5.19.

Theorem 5.21 ([28]). Let Ω ⊂ C be a convex proper subdomain of C. Let further Π be a concave subdomain of C and f : Ω → Π be meromorphic on Ω. Then for all n ≥ 2, z0 ∈ Ω, and f (z0 ) ﬁnite, and p as in (5.41) the inequalities n n |f (n) (z0 )| (RΩ (z0 )) (1 + p)n−2 k p ≤ n! RΠ (f (z0 )) pn−1 k=0

(5.49)

5.5. Maps from convex into concave domains

93

are valid. Equality in (5.49) is attained for $ Ω = {z | Rez < 1/2}

and

Π=C\

% −p −p , , (1 − p)2 (1 + p)2

z0 = f (z0 ) = 0, where the function z(1 − z) z = f (z) = E 1−z (1 − z(1 + p))(1 − z(1 + p)/p) maps Ω onto Π conformally. Proof. Let n ≥ 2. According to the central formula (4.5) we have to prove that n n (1 + p)n−2 ck An,k ≤ pk , pn−1 k=1

k=0

where ck are the Taylor coeﬃcients of a function h ∈ Co(p), and, as usual, An,k is the n-th coeﬃcient in the Taylor expansion of F k for the function F inverse to an arbitrary function s ∈ K. Because of the Schur-Jabotinsky theorem (compare the proof of Theorem 4.15), we have to prove that n−1 n n − l (1 + p)n−2 pk . (5.50) cn−l al,n ≤ n pn−1 l=0

k=0

Using the Marx-Strohh¨ acker inequality as above, we get the following formula for the sum appearing in (5.50): n−1 l=0

n−1 1 n n−1 n−l (n − j)cn−j dj−σ,σ . cn−l al,n = cn + σ n n σ=1 j=σ

(5.51)

Now it is easily veriﬁed that it is suﬃcient to obtain the inequality n−1 (n − σ)(1 − p2(n−σ) ) ≤ (n − j)c d , n−j j−σ pn−σ−1 (1 − p2 ) j=σ where dj−σ are the coeﬃcients of a unimodular bounded holomorphic function in the unit disc as in the proof of Theorem 4.15. The latter inequality is equivalent to q−1 q(1 − p2q ) . (5.52) (q − τ ) cq−τ dτ ≤ q−1 p (1 − p2 ) τ =0

As is proved in [20] and [167], for any h ∈ Co(p) there exists a function v1 : Δ → Δ such that v1 (z)pz z 1+ , z ∈ Δ. h(z) = 1 − z/p 1 − zp

94

Chapter 5. Punishing factors for special cases

Using this and the generalized version of Sheill-Small’s Lemma 2.10 (see Remark 2.12) in the limiting case r → 1 , i.e., the inequality ˜ iθ )| ≤ max |P (z)| |P ∗ h(e

q−1

|z|=1

2

|uk |

1/2 q−1

k=0

1/2 2

|vk |

k=0

for the polynomial P (z) =

q−1

dτ (q − τ )z q−τ

τ =0

and the functions deﬁned by wi (z) = ω(z), i = 1, 2, 3,

z ∈ Δ,

and U (z) =

q−1 k z k=0

p

∞

and V (z) = 1 +

v1 (z)pz vk z k , = 1 − zp

z ∈ Δ,

k=0

we obtain the inequality q−1 1/2 q−1 1/2 q−1 p−2k |vk |2 . (q − τ ) cq−τ dτ ei(q−τ )θ ≤ max |P (z)| |z|=1 τ =0

k=0

k=0

To estimate the last factor of the right side in this inequality we use that the function V (z) − 1 is quasi-subordinated to the function pz/(1 − zp). According to Theorem 2.4 we get q−1 q−1 |vk |2 ≤ p2k . k=0

k=0

Hence, q−1 1/2 q−1 1/2 q−1 −2k 2k p p (q − τ ) cq−τ dτ ≤ max |P (z)| |z|=1 τ =0

k=0

k=0

1 − p2q = max |P (z)| q−1 . p (1 − p2 ) |z|=1 Fej´er’s inequality (see [61] and [159] or apply the theory of linear functionals on H ∞ in chapter 8 of [59]) q−1 (q − τ )dτ ≤ q τ =0

shows that

−iθ P e ≤ q,

θ ∈ [0, 2π].

5.5. Maps from convex into concave domains

95

This concludes the proof of the desired inequality. The extremal property of the function f mentioned in the second assertion of Theorem 5.21 follows directly from the computation of the Taylor coeﬃcients of its expansion with expansion point at the origin. The proof of Theorem 5.21 is complete.

Chapter 6

Multiply connected domains In the preceding chapters we considered punishing factors for simply connected domains, except the case C2 (Ω, Π). Namely, in Section 4.6 it was proved that for all hyperbolic domains Ω ⊂ C and Π ⊂ C, 1 sup |∇ (1/λΩ (z))| + sup |∇ (1/λΠ (w))| . C2 (Ω, Π) = C2 (Π, Ω) = 2 z∈Ω w∈Π In particular, it was proved that C2 (A1 , A2 ) =

1 + 4M12 + 1 + 4M22 ,

where A1 and A2 are annuli with moduli M1 and M2 , respectively. The main aim of this chapter is to describe multiply connected domains by properties that guarantee existence of ﬁnite punishing factors for all n. Also we will deﬁne ﬁnite modiﬁed punishing factors and consider some examples.

6.1

Finitely connected domains

The theorems of Chapters 4 and 5 show that one cannot expect that Cn (Ω, Π) := sup{Mn (z, Ω, Π) | z ∈ Ω}

(6.1)

is always ﬁnite. The central existence theorem for ﬁnitely connected domains is the following assertion. Theorem 6.1 ([21]). Let Ω and Π be ﬁnitely connected hyperbolic domains in C. Then the punishing factors Cn (Ω, Π) are ﬁnite for all n ∈ N if and only if both ∂Ω and ∂Π do not contain isolated points.

98

Chapter 6. Multiply connected domains

Firstly, we prove some necessary lemmas and propositions. We shall use the hyperbolic radius R(z, Π) which is reciprocal to the density of hyperbolic metric with Gaussian curvature K = −4. Let Π be an m-connected hyperbolic domain in C where m ≥ 2, i.e., ∂Π consists of m disjoint connected sets Γk , k = 1, . . . , m. We shall consider m associated simply connected domains Πk ⊂ C, k = 1, . . . , m, deﬁned by the relations ∂Πk = Γk

and Π ⊂ Πk ,

k = 1, . . . , m.

It is clear that Γj ⊂ Πk for j = k and that Π =

m &

Πk .

k=1

Naturally, the connected sets Γk are either points or continua. It is decisive for our proofs that the behaviour of the hyperbolic radius of Π and Πk is the same at the common boundary. If Γk is a point, this is known (see [3] and [55]). We will use the formula for this case later on (compare (6.7) and 6.8)). In the case that Γk is a continuum, this is the content of the following lemma. Lemma 6.2. Let Π and Πk be as above. If Γk = ∂Πk is not a point, then for z0 ∈ Γk , R(z, Π) lim = 1. (6.2) z→z0 , z∈Π R(z, Πk ) Proof. Since Γk is a continuum, Πk is a simply connected domain in C. Riemann’s mapping theorem implies that there exists a conformal map Φ of Δ onto Πk . The inverse function Φ−1 maps Π onto a set Φ−1 (Π) ⊂ Δ. For all suﬃciently small positive numbers the relations D() := {w | e−π < |w| < 1} ⊂ Φ−1 (Π) ⊂ Δ are valid. According to Theorem 3.5 they imply R(w, D()) ≤ R(w, Φ−1 (Π)) ≤ R(w, Δ),

w ∈ D().

(6.3)

Combining (6.3) and the conformal invariance of the hyperbolic metric, |d z| |d w| = , R(z, Π) R(w, Φ−1 (Π))

and

|d z| |d w| = , R(z, Πk ) R(w, Δ)

we get, for any w = Φ−1 (z) ∈ D(), 1 ≥

R(z, Π) R(w, Φ−1 (Π)) R(w, D()) = ≥ . R(z, Πk ) R(w, Δ) R(w, Δ)

The use of R(w, Δ) = 1 − |w|2

and R(w, D()) = 2 |w| sin

1 1 log |w|

(6.4)

6.1. Finitely connected domains

99

and (6.4) yields R(w, Φ−1 (Π)) = R(w, Δ) |w|→1− lim

R(w, D()) = 1. |w|→1− 1 − |w|2 lim

(6.5)

Since |Φ−1 (z)| → 1− as z → z0 , the assertion (6.2) of Lemma 6.2 follows from (6.4) and (6.5). The second preparation for the proof of Theorem 6.1 is concerned with the Taylor coeﬃcients an (f ) of the local expansion f (z) =

∞

an (f )z n

n=0

for a function f ∈ A(Δ, Π) that is holomorphic at the origin. We wish to describe now all ﬁnitely connected hyperbolic domains for which the quantities An (Π) :=

|an (f )| R(a 0 (f ), Π) f ∈A(Δ,Π) sup

are ﬁnite for all n ∈ N. One might expect that boundedness of Π is the decisive condition, but the proof of the following proposition reveals that this is not the case. Lemma 6.3. Let Π be a ﬁnitely connected hyperbolic domain in C. Then the following statements are equivalent. (a) ∂Π does not contain isolated points. (b) An (Π) is ﬁnite for any n ∈ N. (c) Cn (Δ, Π) is ﬁnite for any n ∈ N. (d) There exists a constant KΠ such that, for all w ∈ Π, dist(w, ∂Π) ≤ R(w, Π) ≤ KΠ dist(w, ∂Π). Proof. If Π is a simply connected domain, then Lemma 6.3 is a consequence of known facts (see [55]). It is also known that (a)⇔(d) is valid (see [130] and [77]). Suppose now that Π is m-connected with m ≥ 2. We will prove that (a) ⇒ (b) ⇒ (c) ⇒ (d) ⇒ (a). (a)⇒(b). The condition (a) means that the associated domains Πk , k = 1, . . . , m, are hyperbolic. Since Π ⊂ C, we may suppose, without loss of generality, that Π1 ⊂ C, and ∞ ∈ Πk ⊂ C, k = 2, . . . , m.

100

Chapter 6. Multiply connected domains

Since f ∈ A(Δ, Π) ⊂ A(Δ, Πk ), we can apply theorems of Chapter 5 to get An (Π) ≤ sup{A(w, n) | w ∈ Π}, where n ∈ N and A(w, n) :=

min{σk (w) R(w, Πk ) | k = 1, . . . , m} R(w, Π)

(6.6)

with σ1 (w) ≡ n and σk (w) = 4n−1 (1 − exp(−4δk (w)))1−n ,

k = 2, . . . , m.

In the latter formula the quantity δk (w) = δ(w, ∞, Πk ) means the hyperbolic distance from w to ∞ in the domain Πk . The function A(·, n) is positive and continuous on Π, since hyperbolic radii and the functions σk are positive and real-analytic functions on Π. Suppose that An (Π) = ∞ for one number n ∈ N. Then there exists a sequence of points wj ∈ Π, j ∈ N, such that wj → w0 ∈ ∂Π,

and A(wj , n) → ∞ as

j → ∞.

But this is impossible. Actually, if w0 ∈ ∂Π1 , then lim A(wj , n) ≤ n lim

j→∞

j→∞

R(wj , Π1 ) =n R(wj , Π)

according to (6.6) and Lemma 6.2. If w0 ∈ ∂Πk for some k ∈ {2, . . . , m}, then lim A(wj , n) ≤ 4n−1 lim (1 − exp(−4δk (wj )))1−n

j→∞

j→∞

R(wj , Πk ) = 4n−1 R(wj , Π)

according to (6.6), Lemma 6.2 and the equation lim δk (w) = δ(w, ∞, Πk ) = ∞.

w→w0

Thus, our assumption leads to a contradiction and therefore (b) holds. (b)⇒(c). Let f ∈ A(Δ, Π). Then, for any ﬁxed z ∈ Δ, the function ζ +z , g : Δ → Π, ζ → g(ζ) = f 1 − zζ also belongs to A(Δ, Π). Using the formula (4.3) f (n) (z)(1 − |z|2 )n = n! n

k=1

n−1 n−k

z n−k ak (g)

6.1. Finitely connected domains

101

and the equation f (z) = a0 (g), we obtain Cn (Δ, Π) ≤ 2n−1 max Ak (Π). 1≤k≤n

This implies the desired implication. (c)⇒(d). Let Φ := ΦΠ be a covering map of Δ onto Π. On the one hand, Φ ∈ A(Δ, Π). Hence, the validity of (c) implies that for all z ∈ Δ, (1 − |z|2 )2 |Φ (z)| ≤ 2 C2 (Δ, Π) < ∞. R(Φ(z), Π) On the other hand, R(w, Π) = |Φ (z)|(1 − |z|2 ) for w = Φ(z), z ∈ Δ. Straightforward computations yield that ∂ R(w, Π) = (1 − |z|2 ) Φ (z) − 2 z |∇R(w, Π)| = 2 ∂w Φ (z) (1 − |z|2 )2 |Φ (z)| |Φ (z)| = − 2 z . R(Φ(z), Π) Φ (z) Therefore |∇R(w, Π)| ≤ 2 + 2 C2 (Δ, Π),

w ∈ Π.

Since R(w, Π) = 0 for w ∈ ∂Π, we immediately get R(w, Π) ≤ (2 + 2 C2 (Δ, Π)) dist(w, ∂Π),

w ∈ Π.

This proves the right inequality. The left inequality in (d) is true for any hyperbolic domain (see [3] and [55]). This completes the proof of (c)⇒(d). The implication (d)⇒(a) is a simple consequence of known facts. Actually, if ∂Π contains an isolated point w0 , then (see [3] and [55]) R(w, Π) = (2 + o(1))|w − w0 | log

1 , |w − w0 |

w → w0 ∈ C,

(6.7)

and R(w, Π) = (2 + o(1))|w| log |w|,

w → w0 = ∞.

(6.8)

Hence, (d) cannot hold if ∂Π contains an isolated point. This completes the proof of Lemma 6.3.

Proof of Theorem 6.1. Suppose that ∂Ω and ∂Π do not contain isolated points and that f ∈ A(Ω, Π). For any ﬁxed z ∈ Ω we consider the function g : Δ → Π, ζ → g(ζ) = f (z + ρζ),

102

Chapter 6. Multiply connected domains

where ρ = dist(z, ∂Ω). It it evident that g ∈ A(Δ, Ω) and that a0 (g) = g(0) = f (z)

and an (g) = ρn f (n) (z), n ∈ N.

Since ∂Π does not contain isolated points, Lemma 6.3 implies that the quantities An (Π) are ﬁnite for all n ∈ N. Hence, (dist(z, ∂Ω))n |f (n) (z)| |an (g)| = ≤ An (Π), R(f (z), Π) R(a0 (g), Π)

(6.9)

for all n ∈ N. Using the equivalence (a)⇔(d) of Lemma 6.3 with respect to Ω, we get R(z, Ω) ≤ KΩ dist(z, ∂Ω), z ∈ Ω, for a positive constant KΩ . This inequality and (6.9) imply that |f (n) (z)| ≤ An (Π)(KΩ )n

R(f (z), Π) , (R(z, Ω))n

z ∈ Ω.

Thus,

An (Π)(KΩ )n < ∞, n! for all n ∈ N. This completes the proof of one direction of the assertion. Suppose now that Ω and Π are ﬁnitely connected hyperbolic domains in C and that C2 (Ω, Π) is ﬁnite. Firstly, we use some counterexamples to show that ∂Π cannot have isolated points. Suppose on the contrary that ∂Π has at least one isolated point w0 . Since Cn (Ω, Π) is invariant under linear transformations of Ω and Π, we may restrict ourselves to consideration of the following two cases without loss of generality. Either Cn (Ω, Π) ≤

w0 = 0 ∈ ∂Π

and

Δ = Δ \ {0} ⊂ Π

(6.10)

or w0 = ∞ ∈ ∂Π

and D∞ = {w | 1 < |w| < ∞} ⊂ Π.

(6.11)

In the case (6.10), we consider the functions Ψt := ft ◦ Φ−1 Ω ∈ A(Ω, Δ ) ⊂ A(Ω, Π),

t ∈ (0, ∞),

where Φ−1 Ω is the inverse of a covering map ΦΩ of Δ onto Ω and ft is deﬁned by 1+ζ . ft : Δ → Δ , ζ → ft (ζ) = exp −t 1−ζ On the one hand, at z0 = ΦΩ (0) the condition C2 (Ω, Π) < ∞ implies |Ψt (z0 )| (R(z0 , Ω))2 ≤ R(Ψt (z0 ), Π) = R(e−t , Π) = (2 + o(1))e−t t 2 C2 (Ω, Π)

6.1. Finitely connected domains

103

as t → ∞. On the other hand, straightforward computations show that this is impossible, since ΦΩ (0) 2 −t 2 . (6.12) |Ψt (z0 )| (R(z0 , Ω)) = 4 e t − t + t Φ (0) Ω

In the case (6.11) we examine the functions ˜ t := f˜t ◦ Φ−1 ∈ A(Ω, D∞ ) ⊂ A(Ω, Π), Ψ Ω

t ∈ (0, ∞),

where Φ−1 Ω is deﬁned as above and

1+ζ . f˜t : Δ → D∞ , ζ → f˜t (ζ) = exp t 1−ζ

In this case, we get for z0 = ΦΩ (0), ˜ Ψt (z0 ) (R(z0 , Ω))2 = et t2 (4 + o(1)), as t → ∞. This contradicts the condition C2 (Ω, Π) < ∞ which implies ˜ Ψt (z0 ) (R(z0 , Ω))2 ≤ R(et , Π) = (2 + o(1))et t, 2 C2 (Ω, Π) as t → ∞. Thus, ∂Π does not contain isolated points. Now, we prove that the analogous assertion holds true for Ω. Firstly, we have just proved that there exists a positive constant KΠ such that |∇R(w, Π)| ≤ KΠ ,

w ∈ Π.

Now, let h be a covering map of Ω onto Π deﬁned by h = ΦΠ ◦ Φ−1 Ω ∈ A(Ω, Π). Computations using the equality R(h(z), Π) = |h (z)|R(z, Ω) show that h (z)(R(z, Ω))2 (h (z))2 ∂R(w, Π) h (z) ∂R(z, Ω) = − , 2 2 R(w, Π) |h (z)| ∂w |h (z)| ∂z where w = h(z), z ∈ Ω. It follows that |∇R(z, Ω)| ≤ |∇R(w, Π)| +

|h (z)|(R(z, Ω))2 ≤ KΠ + C2 (Ω, Π) 2 R(w, Π)

for any z ∈ Ω. The boundedness of |∇R(·, Ω)| on Ω implies that ∂Ω has no isolated point. This completes the proof of Theorem 6.1.

104

6.2

Chapter 6. Multiply connected domains

Pairs of arbitrary domains

We shall show that it is possible to deﬁne ﬁnite modiﬁed punishing factors even if Ω contains the point at inﬁnity and ∂Ω contains isolated points. To this end, we use the Euclidean distance, dist(w, ∂Ω), from a point to the boundary of a domain. As a modiﬁcation of Theorem 6.1 we get Theorem 6.4 ([21]). Let Ω be an arbitrary hyperbolic domain in C and Π a ﬁnitely connected hyperbolic domain in C. Then the modiﬁed punishing factors C˜n (Ω, Π) :=

sup

sup

f ∈A(Ω,Π) z∈Ω

|f (n) (z)|(dist(z, ∂Ω))n n! dist(f (z), ∂Π)

are ﬁnite for all n ∈ N if and only if ∂Π does not contain isolated points. Proof. Let f ∈ A(Ω, Π). If Π is a ﬁnitely connected hyperbolic domain in C and ∂Π does not contain isolated points, then the inequality (6.9) holds for any n ∈ N. Moreover, by the statement (d) of Lemma 6.3, there exists a positive constant KΠ such that R(f (z), Π) ≤ KΠ dist(f (z), ∂Π), z ∈ Ω. This inequality and (6.9) imply that n! C˜n (Ω, Π) ≤ An (Π) KΠ ,

n ∈ N.

This proves one direction of the assertion of Theorem 6.4. We prove the other ˜ t from the proof of Theorem direction considering again the functions Ψt and Ψ 6.1. Actually, suppose that Π is a ﬁnitely connected hyperbolic domain in C, C˜2 (Ω, Π) < ∞, and (6.10) holds. Then |Ψt (z0 )| dist(z0 , ∂Ω) ≤ dist(e−t , ∂Π) = e−t . 2 C˜2 (Ω, Π) If we let t → ∞, we see that this contradicts (6.12), since the quotient dist(z0 , ∂Ω)/R(z0 , Ω) does not depend on t. A similar consideration in the case (6.11) reveals an analoguous contradiction. The next theorem gives a further modiﬁcation of Schwarz-Pick type inequalities for any pair of hyperbolic domains. Theorem 6.5 ([21]). Let Ω and Π be arbitrary hyperbolic domains in C. Then for any f ∈ A(Ω, Π), any n ∈ N, and any z ∈ Ω \ {∞} the estimate n R(f (z), Π) 4n n! (n) |f (z)| ≤ (dist(f (z), ∂Π))n−1 dist(z, ∂Ω) is valid, if f (z) = ∞.

6.2. Pairs of arbitrary domains

105

Proof. By the proof of Theorem 6.1, for f ∈ A(Ω, Π) and any ﬁxed z ∈ Ω (z = ∞, f (z) = ∞) we have (dist(z, ∂Ω))n |f (n) (z)| = n! |an (g)|,

n ∈ N,

(6.13)

where g ∈ A(Δ, Π) is deﬁned as above and a0 (g) = g(0) = f (z). Now, let w0 = f (z) and let ΦΠ be a covering map from Δ to Π, such that ΦΠ is a locally univalent function on Δ and ΦΠ (0) = R(w0 , Π). Moreover, for any simply connected domain Π ⊂ Π, any branch Φ−1 Π Π is a univalent function. We consider the branch Ψ = Φ−1 D(w0 , ρ) = {w | |w − w0 | < ρ = dist(w0 , ∂Π)}, Π D(w ,ρ) , 0

satisfying the condition Ψ(w0 ) = 0. The Koebe one-quarter theorem implies that the function R(w0 , Π) h : Δ → C, ζ → h(ζ) = Ψ(w0 + ρζ) ρ attains all values in the disc D1/4 = {z | |z| < 1/4}. Hence, Da =

z | |z| < a =

ρ 4 R(w0 , Π)

⊂ Ψ(D(w0 , ρ)).

(6.14)

Thus, ΦΠ is univalent on Da . As a consequence of the monodromy theorem, the function ω deﬁned by ω : Δ → Δ, ζ → ω(ζ) = Φ−1 Π ◦ g(ζ),

ω(0) = 0,

is a holomorphic self-map of Δ. Hence, |ω(ζ)| ≤ |ζ| for all ζ ∈ Δ. This implies g(Da ) ⊂ ΦΠ (Da ) ⊂ D(w0 , ρ), since g(ζ) = ΦΠ (ω(ζ)). Applying Cauchy’s estimates for the Taylor coeﬃcients of the function g − w0 in the disc Da , we get an |an (g)| ≤ ρ,

n ∈ N.

(6.15)

From (6.13), (6.14), and (6.15) it follows that (dist(z, ∂Ω))n |f (n) (z)| ≤

n! ρ 4n n! (R(w0 , Π))n ≤ , n a ρn−1

which is the desired inequality of Theorem 6.5.

106

Chapter 6. Multiply connected domains Next, we shall examine some facts connected with Theorem 4.18.

Let Ω and Π be open sets on the Riemann sphere C that are equipped with the Poincar´e metric of curvature −4. According to Poincar´e’s generalization of Riemann’s mapping theorem, this means that the boundaries of Ω and Π contain at least three points in C and that the density of this metric in the unit disc Δ = {z | |z| < 1} is deﬁned as λΔ (z) =

1 , 1 − |z|2

z ∈ Δ.

As usual, we consider the set A(Ω, Π) of functions f : Ω → Π, which are locally holomorphic or meromorphic and in general multivalued. Let n ∈ N and Qn (z, w, Ω, Π) be deﬁned as the smallest possible value such that the inequality (n) n f (z) (λΩ (z)) ≤ Qn (z, w, Ω, Π) n! λΠ (w) holds for all f ∈ A(Ω, Π), f (z) = w. For any pair (z, w) ∈ Ω × Π and any n ∈ N, normal family arguments give that there exists a function f ∈ A(Ω, Π) such that f (z) = w and (n) n f (z) (λΩ (z)) = Qn (z, f (z), Ω, Π) . n! λΠ (f (z)) Moreover, it is easily seen that Cn (Ω, Π) = sup{Qn (z, w, Ω, Π) | (z, w) ∈ Ω × Π}. Here we shall need the function F deﬁned in the proof of Theorem 4.18 by equation p + q, if p + 2q ≥ 4, (6.16) F (p, q) := 2 + p2 /(8 − 4q), if p + 2q < 4, for p ≥ 0 and q ≥ 0. Theorem 6.6 ([27]). Let Ω and Π be open sets on the Riemann sphere that are equipped with the Poincar´e metric of curvature −4. Then for any f ∈ A(Ω, Π) the estimate 2 (λΩ (z)) |f (z)| ≤ F (p, q) (6.17) λΠ (w) is valid for w = f (z), z ∈ Ω = C ∩ Ω, w ∈ Π = C ∩ Π, p = |∇ (1/λΩ (z))|, and q = |∇ (1/λΠ (w))|. The inequality (6.17) is sharp for any (z, w) ∈ Ω × Π . Proof. We ﬁx (z, w) ∈ Ω × Π and consider a function f ∈ A(Ω, Π), f (z) = w. Without loss of generality, we assume that Ω and Π equal their components that contain z and w, respectively. For this case the assertion follows from the proof of Theorem 4.18.

6.2. Pairs of arbitrary domains

107

As in Chapter 5 we denote γ(Ω) = sup |∇ (1/λΩ (z))| , z∈Ω

γ(Π) = sup |∇ (1/λΠ (w))| w∈Π

and we want to give some bounds for C2 (Ω, Π) dependent on the Euclidean geometry and on the conformal geometry of these sets. From Chapter 3 and Theorem 6.6 we immediately have the following corollaries. Corollary 6.7. Let Ω and Π be open sets in C equipped with the usual Poincar´e metric. Then C2 (Ω, Π) is ﬁnite if and only if C\Ω and C\Π are uniformly perfect. Corollary 6.8. Let Ω and Π be open sets in C equipped with the usual Poincar´e metric. Then the following assertions are valid. (i) C2 (Ω, Π) =

1 2 (γ(Ω)

+ γ(Π)),

(ii) 2(M0 (Ω) + M0 (Π)) ≤ C2 (Ω, Π) ≤ 2π(M0 (Ω) + M0 (Π)) + Γ(1/4)4 /π 2 , (iii) 2(M (Ω) + M (Π)) − 2 ≤ C2 (Ω, Π) ≤ 2π(M (Ω) + M (Π)) + Γ(1/4)4 /π 2 , (iv) C2 (Ω, Π) ≥ 2, where equality is attained if and only if all components of Ω and Π are convex. Together with the results of Chapter 4, where we proved that Cn (Ω, Π) = 2n−1 for convex domains Ω and Π, Corollary 6.8 (iv) delivers the proof of a weak form of a conjecture published in [16]. We want to formulate the remaining parts of a stronger conjecture as Conjecture. Let n ∈ N \ {1, 2} and let Ω and Π be open sets in C equipped with the usual Poincar´e metric. Then Cn (Ω, Π) = 2n−1 implies that all components of Ω and Π are convex domains. Now, we turn to the cases where the point at inﬁnity belongs to Ω or Π and we prepare these considerations with a lemma which may deserve some interest of its own. Lemma 6.9 ([27]). Let Ω be an open set in C equipped with the usual Poincar´e metric and let ∞ ∈ Ω. Then the equations (i) lim |∇ (1/λΩ (z))| = ∞, z→∞

(ii) lim |∇ log λΩ (z)|dist(z, ∂Ω) = 2, z→∞

are valid. Proof. Without restriction of generality we assume Ω to be a hyperbolic domain. To prove (i) we consider a universal covering map f of Δ onto Ω such that f (0) = ∞. Let f have the expansion a z = f (ζ) = + ϕ(ζ), ζ ∈ Δ, ζ

108

Chapter 6. Multiply connected domains

where a = 0 and ϕ is a function meromorphic in Δ and holomorphic in a neighbourhood of the origin. Direct computations using λΩ (z) =

1 |f (ζ)|(1

imply |∇ (1/λΩ (z))| =

− |ζ|2 )

,

z = f (ζ),

1 2a + ζ 3 ϕ (ζ) 2 + 2 |ζ| . 2 |ζ| a − ζ ϕ (ζ)

(6.18)

(6.19)

If we take the limit z → ∞ of the left side of equation (6.19), which is equivalent to taking the limit ζ → 0 of the right side, we see that (i) is valid. To prove (ii) we multiply the left sides and the right sides of (6.18) and (6.19) with one another. If we multiply these products with |z| and |a/ζ|, we get a (6.20) lim |∇ log λΩ (z)||z| = lim |∇ log λΩ (f (ζ))| = 2. z→∞ ζ→0 ζ Since ∂Ω is bounded, lim

z→∞

dist(z, ∂Ω) = 1. |z|

(6.21)

The formulae (6.20) and (6.21) imply (ii). This completes the proof of Lemma 6.9. Lemma 6.9 (ii) gives rise to a sharp inﬁnity version of the Osgood-Jørgensen theorem. For hyperbolic domains Ω ⊂ C the fundamental inequality |∇ log λΩ (z)|dist(z, ∂Ω) ≤ 2,

z ∈ Ω,

(6.22)

was discovered by Osgood (see [121], and compare also [111]). The proof uses certain results of Jørgensen (see [88]). One may observe that this proof in [121] and the related results in [88] do not use the fact that ∞ ∈ / Ω. Moreover, in [88] Jørgensen considers the case when ∞ ∈ Ω as a basic one. This observation together with Lemma 6.9 (ii) leads us to the following theorem which proves (3.36). Theorem 6.10 ([21]). Let Ω be an open set in C equipped with the usual Poincar´e metric. For any z ∈ Ω ∩ C the inequality (6.22) is valid. The constant 2 in (6.22) cannot be replaced by a smaller one if ∞ ∈ Ω.

6.3 Some examples Let Π be a hyperbolic domain in the extended complex plane C. For ﬁxed w ∈ Π \ {∞} we consider the local Taylor expansions f (z) = w +

∞ n=1

an (w, f )z n

6.3. Some examples

109

valid in a neighbourhood of the origin. We will discuss some examples showing the behaviour of the quantities An (w, Π) = sup{|an (w, f )| | f ∈ A(Δ, Π), f (0) = w} regarded as a function of w (see [29]). Example 1. We consider the doubly connected domain Π = D∞ := {z ∈ C | |z| > 1} and we prove that for n ∈ N, An (w, D∞ ) = Fn (− log |w|), where −t

Fn (t) = − 2 t e

n−1 k=0

n−1 k

(−2t)k , (k + 1)!

(6.23)

n ∈ N ∪ {0},

is Bateman’s function of order n (compare [35] and [94]). The upper bound is attained if and only if there exist ϕ1 , ϕ2 ∈ R such that f (z) = eiϕ1 ΦD∞ ,w eiϕ2 z . Further, 1 An (w, D∞ )(dist(w, ∂D∞ ))n−1 = w→∞ (R(w, D∞ ))n n! lim

and

An (w, D∞ )(dist(w, ∂D∞ ))n−1 = 21−n . (R(w, D∞ ))n |w|→1 lim

To prove these assertions we use that, for w = |w|eiθ , t =: log |w| > 0, the function ΦD∞ ,w is deﬁned by 1 + e−iθ z iθ ΦD∞ ,w (z) = e exp t 1 − e−iθ z and that for any f ∈ A(Δ, D∞ ), f (0) = w, there exists a uniquely deﬁned function p holomorphic in Δ such that Re p(z) > 0 for z ∈ Δ, p(0) = 1, and f (z) = eiθ exp(t p(z)). The ﬁrst equation yields R(w, D∞ ) = 2|w| log |w| and the second by diﬀerentiation f (z) = t p (z)f (z).

(6.24)

110

Chapter 6. Multiply connected domains

Using the Taylor coeﬃcients of the expansion p(z) = 1 +

∞

pk z k ,

k=1

we see that a1 (w, f ) = t w p1 ,

n−1

n an (w, f ) = t w n pn + ak (w, f )(n − k)pn−k , n ≥ 2. k=1

To prove (6.23) and the assertion on the extremal function we remember that the inequalities |pk | ≤ 2 are valid for k = 1, . . . , n, with equality if and only if p(z) = (1 + cz)/(1 − cz), |c| = 1, and the triangle inequality. Further we use the identity n−1 nFn (t) = − 2 t (n − k)Fk (t) k=0

which we were not able to ﬁnd anywhere, but which may be easily proved by mathematical induction and the diﬀerence equation (n − 1)(Fn (t) − Fn−1 (t)) + (n + 1)(Fn (t) − Fn+1 (t)) = 2tFn (t) (compare [35] and [94]). Since lim

t→∞

Fn (−t) 1 = , (2t)n et n!

the ﬁrst assertion on the asymptotics of An (w, D∞ ) is an immediate consequence of (6.23) and (6.24). The second one follows likewise. This example shows that there are domains where the universal covering functions deliver the extrema for the moduli of all local Taylor coeﬃcients. Example 2. For the doubly connected domain Δ = {z | 0 < |z| < 1}, w = |w|eiθ , t = − log |w| > 0, the universal covering function is given by 1 + e−iθ z iθ ΦΔ ,w = e exp −t 1 − e−iθ z and f ∈ A(Δ, Δ ), f (0) = w, may be written as f (z) = eiθ exp(−t p(z)), where p is as in Example 1. It is easily seen by the same method as in Example 1 that |an (w, ΦD0 ,w )| = |Fn (t)|.

6.3. Some examples

111

This has been proved in [94]. Notwithstanding the fact that the determination of An (w, Δ ) is very diﬃcult the methods of Example 1 may be applied to get the right asymptotics near the origin. An analogous reasoning shows that for f ∈ A(Δ, Δ ), f (0) = w, |p1 |n |an (w, f )|(dist(w, ∂Δ ))n−1 = (R(w, Δ ))n 2n n! |w|→ 0 lim

if we ﬁx p in the above representation of f . Since |p1 | ≤ 2 where equality is attained if f equals a universal covering function, we get An (w, Δ )(dist(w, ∂Δ ))n−1 1 = . (R(w, Δ ))n n! |w|→ 0 lim

In a similar way, we get |an (w, f )|(dist(w, ∂Δ ))n−1 |pn | = n (R(w, Δ ))n 2 |w|→ 1 lim

and conclude

An (w, Δ )(dist(w, ∂Δ ))n−1 = 21−n . (R(w, Δ ))n |w|→ 1 lim

Example 3. Let δ > 1 and the annulus Aδ be deﬁned by Aδ = {z | 1 < |z| < δ}. It is known that the hyperbolic radius in this case has the form log δ log |w| . R (w, Aδ ) = 2 |w| sin π π log δ To get the asymptotics near the boundary we use the following representation of f ∈ A(Δ, Aδ ), f (0) = w. Let τ > 0 and t ∈ (−π/2, π/2) be chosen such that π log δ log τ log δ t+ exp − i . w = exp π 2 π Then there exists a function p that is holomorphic and has positive real part in Δ, p(0) = 1, such that log δ log (i τ (cos t p(z) + i sin t)) . f (z) = exp − i π Diﬀerentiation yields (cos t p(z) + i sin t)f (z) = − i cos t

log δ p (z) f (z). π

112

Chapter 6. Multiply connected domains

The consideration of the Taylor coeﬃcients in this formula yields a1 (w, f ) = − i p1 w

log δ cos t π

and n an (w, f )eit + cos t

n−1

k ak pn−k

k=1

n−1 log δ = −i ak (n − k)pn−k . cos t n pn w + π k=1

By a reasoning analogous to that of the second example we get lim

w→∂Aδ

An (w, Aδ )(dist(w, ∂Aδ ))n−1 = 21−n . (R(w, Aδ ))n

To get a global estimate in this case we use A(Δ, Aδ ) ⊂ A(Δ, Δ) for w ∈ Aδ and |f (z)| < δ for z ∈ Δ. These facts yield immediately δ 2 − |w|2 . An (w, Aδ ) ≤ min Fn (− log |w|), δ The diﬀerence between the ﬁrst two examples and the third one lies in the fact that the quantity R(w, Ω)/dist(w∂Ω) is unbounded in the ﬁrst two cases and bounded in the last one.

Chapter 7

Related results 7.1

Inequalities for schlicht functions

First, we will give an outline of the ideas and results that led to the conjectures of Chua. To our knowledge, E. Landau was the ﬁrst who considered the possibility to follow G. Pick’s program as indicated in the introduction for the higher derivatives of schlicht functions. He proved the following theorem (compare Landau [98], Gong [71]). Theorem 7.1. If the Bieberbach conjecture is valid for n ≥ 2, then for these n, any z0 ∈ Δ, and any function f holomorphic and injective on Δ, the inequality |f (n) (z0 )| n + |z0 | ≤ |f (0)| n! (1 − |z0 |)n+2

(7.1)

is valid. It is clear that, on the other hand, the validity of the Bieberbach conjecture follows from (7.1). The proof uses automorphisms of the unit disc in the same way as the formula of Sz´ asz mentioned above. Since the same method was used to prove a similar formula of Jakubowski that originated in the ﬁrst half of the last century, we give a uniﬁed proof for these two theorems after the presentation of Jakubowski’s Theorem (see [85]). Theorem 7.2. If the Bieberbach conjecture is valid for n ≥ 2, then for these n, any z0 ∈ Δ, and any function f holomorphic and injective on Δ, the inequality |f (n) (z0 )| (n + |z0 |)(1 + |z0 |)n−2 ≤ |f (z0 )| n! (1 − |z0 |2 )n−1 is valid.

(7.2)

114

Chapter 7. Related results

Proof of Theorems 7.1 and 7.2. Let f be injective on Δ. The function

ζ+z0 f 1−z 0ζ g(ζ) = (1 − |z0 |2 )f (z0 ) is injective on Δ and g (0) = 1. Using Szasz’s formula and de Branges’ theorem for ∞ g(ζ) = g(0) + z + ak ζ k k=2

we get immediately, as in Chapter 4, |f (n) (z0 )|(1 − |z0 |2 )n ≤ 2 (1 − |z0 | )|f (z0 )| n! n

k=1

n−1 n−k

|z0 |n−k k = (n + |z0 |)(1 + |z0 |)n−2 .

Apparently, this is Jakubowski’s theorem. Since Landau was interested in an upper bound dependent on |z0 | only, he used in his proof the distortion theorem |f (z0 )| 1 + |z0 | ≤ |f (0)| (1 − |z0 |)3 for functions injective on Δ. Combining this with Jakubowski’s bound delivers Landau’s theorem. The case of equality in Jakubowski’s results has been discussed by Yamashita in [169] in detail. Further, Jakubowski recognized that by the same method, using L¨ owner’s theorem for the MacLaurin coeﬃcients of functions f convex and injective on Δ, one gets the bound |f (n) (z0 )| (1 + |z0 |)n−1 . ≤ |f (z0 )| n! (1 − |z0 |2 )n−1 Here, again the case of equality is due to Yamashita (see [169]). A natural simplifying of these bounds is achieved, if one replaces in the numerator |z0 | by 1. This results in the factors (n + 1)2n−2 in the case of injective functions and 2n−1 in the case of convex functions. These constants play an important part in the present book, Chapter 5. Possibly, the resulting formula, the identity R(z0 , Δ) = 1 − |z0 |2 and naturally de Branges’ theorem motivated Chua in 1996 to consider functions f injective or convex on simply connected proper subdomains Ω of C (see [56]).

7.1. Inequalities for schlicht functions

115

He proceeded as we did in Chapter 4 with the exception that for him there was no need to consider subordination. Therefore he arrived at a formula analogous to our formula (4.5). The ﬁrst diﬀerence is that in his formula there appear the coeﬃcients of schlicht or convex functions and not those of subordinates to them. The second diﬀerence is that he could use the identity R(f (z0 ), f (Ω)) = |f (z0 )| R(z0 , Ω). Concerning a formula equivalent to the formula of Sz´ asz that we used, he found a formula proved by Todorov in [163]. Naturally, he had to consider the coeﬃcients of powers of inverses of schlicht functions, as we do. He knew the proofs of L¨ owner (see [110] and of Schober (see [147]) for the coeﬃcients of the inverses themselves, but he was not aware of the fact that the extremality of the Koebe function for powers, where the exponent is a natural number, is a simple consequence of L¨owner’s theorem. Therefore, he proved this fact using Baernstein’s integral mean theorem (see [32]) and the Schur-Jabotinski theorem (see [81], Thm 1.9.a). By these theorems, he could prove that for f injective on Ω, z0 ∈ Ω, the inequality 4n−1 |f (n) (z0 )| , ≤ |f (z0 )| n! (R(z0 , Ω))n−1 is valid , whereas for f convex and injective he got 2n − 1 n |f (n) (z0 )| , ≤ |f (z0 )| n! (R(z0 , Ω))n−1

n ≥ 2,

n ≥ 2.

Further, Chua recognized that, for the consideration of similar problems for functions f injective or convex on convex proper subdomains of C that led him to his conjectures, one needs bounds for the inverses of convex functions. Using a theorem due to Trimble (see [164]) he suceeded in proving the conjectures up to n = 4. The results of Li (see [102] and [103]), who used theorems of Libera and Zlotkiewicz (see [104]) on the coeﬃcients of inverses of convex functions, imply that these conjectures are valid for n ≤ 8. As we have seen in Chapter 4, the computation of punishing factors is possible, if one has a good knowledge of the inverse coeﬃcients of certain functions on one hand and good results on the coeﬃcients of subordinated functions on the other hand. If one reduces the interest to the conformal maps themselves, one can replace the subordination results by theorems on the bounds of coeﬃcients of certain injective functions. We want to cite two examples where the proofs of the details are the same as in Chapter 4 and Chapter 5. In addition we need the interplay between a geometric property, the accessibility of order β and an analytic property, the close-to-convexity of order β.

116

Chapter 7. Related results

Deﬁnition 7.3. A domain Π is called (angularly) accessible of order β, β ∈ [0, 1], if it is the complement of a union of rays that are pairwise disjoint except that the origin of one ray may lie on another one of the rays, and such that every ray is the bisector of a sector of angle (1 − β)π that wholly lies in the complement of Π. We use the following characterization of domains accessible of order β. Theorem 7.4 (see for instance [127] and [93] and compare [41]). Let f be injective and holomorphic in the unit disc and f (Δ) accessible of order β. Then there exist α ∈ [0, 2π], and functions g and p holomorphic in Δ such that the following conditions hold: (i) g(Δ) is convex, (ii) Re(eiα p(z)) > 0 for z ∈ Δ and p(0) = 1, (iii) f (z) = (p(z))β g (z) for z ∈ Δ. Functions satisfying the proprties (i)–(iii) are called (nonnormalized) functions close-to-convex of order β and we owe to Ahoronov and Friedland the proof (see [2] and compare [146]) that for such functions the following inequalities hold. Theorem 7.5. Let f be close-to-convex of order β. Then 1+β |f (n) (0)| 1+z 1 dn , ≤ |f (0)| 2(β + 1) (dz)n 1 − z

n ∈ N.

z=0

These theorems and the considerations and computations of Chapter 5 immediately imply the following theorems. Theorem 7.6. Let f be holomorphic and injective on a simply connected proper subdomain of C and such that f (Δ) is linearly accessible of order β, β ∈ [0, 1]. Then for z0 ∈ Ω and n ∈ N the inequalities β+1 |f (n) (z0 )| 4n−1 1 2 +n−1 ≤ n |f (z0 )| (R(z0 , Ω))n−1 β + 1 are valid. Theorem 7.7. Let f be holomorphic and injective on a simply connected proper subdomain Ω of C, ∞ ∈ Ω, and such that f (Δ) is linearly accessible of order β, β ∈ [0, 1]. Then for z0 ∈ Ω \ {∞} and n ∈ N and p(z0 ) deﬁned as above, the inequalities 4k |f ( n)(z0 )| 1 ≤ n−1 |f (z0 )| (R(z0 , Ω)) 2(β + 1) n

k=1

are valid.

n−1 n−k

β+1 2

k

p(z0 ) +

n−k 1 +2 p(z0 )

7.2. Derivatives of α-invariant functions

7.2

117

Derivatives of α-invariant functions

Let Ω be a simply connected domain in C with the density λΩ of the Poincar´e metric. Our aim is to ﬁnd bounds for the function Mn (a, Uα (Ω)) =

1 max{|f (n) (a)| : f ∈ Uα (Ω)}, n!

a ∈ Ω,

where α is a parameter and Uα (Ω) is a family of holomorphic functions with a given property. For instance, Uα (Ω) is the unit ball f ≤ 1 in one of the usual spaces of holomorphic functions. To ﬁnd Mn we will use the functions (see Chapter 3) μk = μk (a, Ω) =

1 ∂ k log λ2Ω (a) , λ−k Ω (a) (k − 1)! ∂ak

k = 1, . . . , n.

We start with a simple remark. For all conformal automorphisms ϕ : Ω → Ω, the condition f ∈ A(Ω, Π) implies that the functions f ◦ϕ are members of A(Ω, Π), too. With this observation in mind, we shall consider the following generalization of linear invariant families by Ch.Pommerenke [126] (compare also [8]). The ﬁrst deﬁnition deals with functions holomorphic in the unit disc Δ. Deﬁnition 7.8. Let α = const ≥ 0, and let Uα be a set of functions g such that: (i) g is holomorphic in Δ, (ii) for all conformal automorphisms ϕ : Δ → Δ, g ∈ Uα ⇒ gα, ϕ ∈ Uα , where gα,ϕ (ζ) := g(ϕ(ζ))ϕα (ζ), (iii) g(ζ) are uniformly bounded in the interior of Δ and (0, . . . , 0) is an interior point of the coeﬃcient region {(a0 , . . . , an ) : g(ζ) = a0 + a1 ζ + · · · ∈ Uα }. The following deﬁnition describes α- invariant families of functions holomorphic in a simply connected domain Ω. Deﬁnition 7.9. Given Uα , by Uα (Ω) we denote the set of all functions f such that f is holomorphic in Ω and fα, Φ ∈ Uα for all conformal mappings Φ of Δ onto Ω, where fα, Φ (ζ) = f (Φ(ζ))Φα (ζ). If ∞ ∈ Ω, then we suppose that 2α ∈ N ∪ {0}. Remark 7.10. The set A(Ω, Π) is a 0-invariant family. Let Pw (ζ) be the polynomial 1 +w1 ζ + · · · + wn ζ n . For g(ζ) = a0 + a1 ζ + · · · ∈ Uα , denote gn (ζ) = an +an−1 ζ+· · ·+a0 ζ n . We will consider the Hadamard product (gn ∗ Pw )(ζ) = an + an−1 w1 ζ + · · · + a0 wn ζ n .

118

Chapter 7. Related results

Theorem 7.11 ([14]). If a ∈ Ω and f (z) ∈ Uα (Ω), then Mn (a, Uα (Ω)) = λn+α (a) max max |(gn ∗ Pw )(ζ)|, Ω

(7.3)

g∈Uα |ζ|=1

where w = (τn,n−1 (α), τn,n−2 (α), . . . , τn,0 (α)), τn,k (α) are deﬁned by the following recurrent formulas: τk,k (α) = 1, 0 ≤ k ≤ n,

τm,k (α) =

m−k+1 s=1

τk,0 (α) =

k−1 α μk−s τs,0 (α), 1 ≤ k ≤ n, k s=0

1 τs−1,0 (1)τm−s,k−1 (α), 2 ≤ k ≤ m ≤ n. s

(7.4)

(7.5)

Proof. Let ψ be the conformal mapping of Ω to Δ, ψ(a) = 0, ψ (a) = λΩ (a) > 0. As in Chapter 3, we have 2ψ(z) ∂ log λ2Ω (z) ψ (z) ψ (z). = − ∂z ψ (z) 1 − ψ(z)ψ(z) Hence λkΩ (a)μk (a, Ω) and

1 1 ∂ k log λ2Ω (a) = = (k − 1)! ∂ak (k − 1)!

ψ (z) ψ (z)

(k−1)

(7.6) z=a

∞

ψ (z) k+1 λΩ (a)μk+1 (a, Ω)(z − a)k = ψ (z)

(7.7)

k=0

in some neighbourhood of a. Consider the function g(ζ) = f (Φ(ζ))Φα (ζ) = a0 + a1 ζ + . . . ,

ζ ∈ Δ,

(7.8)

where z = Φ(ζ) is the inverse of the function ζ = ψ(z). It follows from (7.8) that ∞ 1 (m) 1 k α (m) ak (ψ (z)ψ (z)) f (a) = m! m! k=0

=

m

z=a

ak tm,k (α)

k=0

with m! tm,k (α) = (ψ k (z)ψ α (z))(m) |z=a . We have to prove that tm,k (α)/λm+α (a) = τm,k (α) Ω

7.2. Derivatives of α-invariant functions

119

satisfy (7.4) and (7.5) for m = 1, 2, . . . , n. From ψ(z) = λΩ (a)(z − a)[1 + O(z − a)], (a) in accordance with τk,k (α) = 1. Using (7.6) we have directly tk,k (α) = λk+α Ω and the Leibniz formula, we have (k−1) 1 α α ψ (z) (k) α tk,0 (α) = (ψ (z)) ψ (z) = k! k! ψ (z) z=a z=a k−1 (k−1−s) (z) α ψ k−1 (ψ α (z))(s) = · s k! s=0 ψ (z) z=a

z=a

=

α k

k−1

λk−s Ω (a)μk−s ts,0 (α).

s=0

We also have 1 tm,k (α) = m! s=0 m

m s

(ψ k−1 (z)ψ α (z))(m−s)

(s)

(ψ(z))

z=a

m−k+1 1 (s−1) = tm−s,k−1 (α)(ψ (z)) s! s=1

=

m−k+1

z=a

s=1

z=a

1 tm−s,k−1 (α)ts−1,0 (1). s

Mathematical induction gives (7.4) and (7.5) for τn,k (α) = tn,k (α)/λn+α (a). Thus, Ω 1 (n) (a)(gn ∗ Pw )(1) f (a) = λn+α Ω n! for w = (τn,n−1 (α), τn,n−2 (α), . . . , τn,0 (α)), g(ζ) = f (Φ(ζ))Φα (ζ). Taking into account the property g(ζ) ∈ Uα ⇒ g(e−iθ ζ) ∈ Uα , we obtain 1 max{|f (n) (a)| : f ∈ Uα (Ω)} n! = λn+α (a) max |(gn ∗ Pw )(1)| = λn+α (a) max |(gn ∗ Pw )(eiθ )| Ω Ω

Mn (a, Uα (Ω)) =

g∈Uα

=

n+α λΩ (a)

g∈Uα

max max |(gn ∗ Pw )(ζ)|.

g∈Uα |ζ|=1

This completes the proof of Theorem 7.11.

Using the well-known properties of max{|F (ζ)| : |ζ| = r} (0 < r < ∞) for the analytic function F (ζ) = (gn ∗ Pw )(ζ) (see [59]), we obtain the following corollary .

120

Chapter 7. Related results

Corollary 7.12. Let w = (τn,n−1 (α), τn,n−2 (α), . . . , τn,0 (α)). The function (a) Kn (w, Uα ) = Mn (a, Uα (Ω))/λn+α Ω has the following properties: (i) Kn (w, Uα ) depends on a and Ω via w only, (ii) Kn (w, Uα ) is a convex function from Cn (= R2n ) to (0, ∞), (iii) there exist two positive constants c(n, α) and C(n, α) such that

c(n, α) ≤ Kn (w, Uα )/ 1 + w2 ≤ C(n, α),

w ∈ Cn ,

(iv) K(w1 , . . . , wn , Uα ) = K(eiθ w1 , . . . , einθ wn , Uα ) for any θ ∈ [0, 2π], (v) for w ∈ Cn , the function u(r) = Kn (rw1 , r2 w2 , . . . , rn wn , Uα ) is nondecreasing on 0 ≤ r ≤ ∞ and log u(r) is convex with respect to log r. In the case 0 ≤ k ≤ n − 1, τn,k (α) depends on n, k, α, a and Ω. For example, if Ω = Δ and a ∈ Δ, then n + 2α − 1 τn,k (α) = an−k . n−k Note that τn,k (α) does not depend on a ∈ Ω if and only if Ω is a half-plane. In the general case, for domains Ω ⊂ C we proved that τn,k (α) are bounded. Namely, by Theorem 3.27 we have the following assertion. If Ω ⊂ C, then sup sup |τn,k (α)| = Ω a∈Ω

n−k s=0

2n + 3α − 1 s

n−k−s+α−2 n−k−s

.

(7.9)

From Theorem 7.11 and Theorem 3.27 one immediately obtains the following corollary. Corollary 7.13. There exists an increasing convex function K1 (., Uα ) : [0, ∞) → (0, ∞) such that M1 (a, Uα (Ω)) = λ1+α Ω (a)K1 (t, Uα ),

t = α|∇ λ−1 Ω (a)|.

(7.10)

If Ω ⊂ C, then M1 (a, Uα (Ω)) = K1 (4α, Uα ). λ1+α a∈Ω Ω (a)

sup sup Ω

(7.11)

7.2. Derivatives of α-invariant functions

121

Consider now Hardy, Bergman and Bloch spaces (see for instance [59], [7]). We use the following notation: f p = sup

1/p |f (z)| |dz|

for f ∈ Ep (Ω), p > 0,

p

0 0, γ > 1,

for f ∈ Ap,γ (Ω); 1/p

f ∗p = sup |f (z)/λΩ (z)| for f ∈ Bp∗ (Ω), p > 0. z∈Ω

Theorem 7.14 ([14]). If t = |∇ λ−1 Ω (a)|, then for any a ∈ Ω, t f p , p ≥ 1, (a) 1 + p t 1+1/p f p , 0 < p ≤ 1, (a) (2 − p)1/p−1/2 21−1/p p−1/2 + |f (a)| ≤ λΩ p t 1+1/p f ∗p , p > 0, (a) (2 + p)1/p+1/2 2−1/p p−1/2 + |f (a)| ≤ λΩ p t 2Γ(γ + 1/2) 1+1/(γp) √ f p,γ , p ≥ 1, γ > 1, + (a) |f (a)| ≤ λΩ γp πΓ(γ) 1+1/p

|f (a)| ≤ λΩ

(7.12) (7.13) (7.14) (7.15)

where Γ denotes Euler’s gamma function. The estimates (7.12), (7.13), (7.14), (7.15) are asymptotically sharp as t → +∞, (7.12), (7.13) and (7.14) are also sharp at t = 0. Note that if U1/p is the unit ball in the Hardy space Hp , then the function K1 (t, U1/p ) is known by [114] in the cases p ≥ 1 + t/2 and 1 ≤ p ≤ max{2, 1 + t/2}. For Bloch space Bp∗ (Ω) we obtain from Corollary 7.13 and [165], 1+1/p

max |f (z)| = (2 + p)1/p+1/2 2−1/p p−1/2 λΩ

f ∗ p ≤1

(a)kp (|grad λ−1 Ω (a)|),

where kp (t) = max0≤x≤√1+2/p [1 + tx/p − (1 + 2/p)x2 ](1 − x2 )1/p . Proof of Theorem 7.14. From Theorem 7.11 we obtain M1 (a, Uα (Ω)) = sup (|a1 | + |τ1,0 (α)||a0 |)λ1+α Ω (a), g∈Uα

(7.16)

122

Chapter 7. Related results

where a0 = g(0), a1 = g (0), |τ1,0 (α)| = α|∇ λ−1 Ω (a)|. If α = 1/p, p ≥ 1 and g ∈ Hp , then |a0 | ≤ gp and |a1 | ≤ gp (see [59]). Taking into account Corollary 7.13, we have (7.12). If α = 1/p, p > 0 and g ∈ Bp∗ (Δ), then the region {(|a0 |, |a1 |) : g∗p ≤ 1} is given in [165], and we get (7.16) and (7.14). Consider the case g ∈ Ap,γ (Δ), p ≥ 1. By Hardy’s theorem, 2π p 2π|g(0)| ≤ |g(reiθ )|p dθ, 0 < r < 1. 0

This yields gp,γ ≥ |a0 |

γ−1 π

1

0

2r(1 − r2 )γ−2 dr

1/p = |a0 |.

We also have 1 r2 (1 − r2 )γ−2 dr gp,γ ≥ g1,γ ≥ 2|a1 |(γ − 1) 0 √ = |a1 | πΓ(γ)/(2Γ(γ + 1/2)). Hence, (7.15) holds. If g(ζ) ∈ Hp (0 < p < 1), then |a0 | ≤ gp (see [59]). Using the estimates from [157], we have |a1 | ≤ (2 − p)1/p−1/2 21−1/p p−1/2 . Thus, (7.13) holds. Theorem 7.15 ([14]). If f ∈ A2,γ (Ω), then for any a ∈ Ω and any γ ≥ 1, n 2 1/2 1 n+ 2γ 1 1 (n) M k+γ−1 ) τn,k (a) ( . (7.17) max |f (a)| = λΩ k n! f 2,γ ≤1 2γ k=0

Proof. By direct calculation, we obtain 1/2 ∞ k + γ − 1 −1 2 |ak | . g2,γ = k k=0

By the Cauchy-Schwarz inequality |

n

ak τn,k (α)| ≤ g2,γ

k=0

1/2 n k+γ−1 |τn,k (α)|2 . k

k=0

Equality holds for g(ζ) = a0 + a1 ζ + · · · + an ζ n if and only if k+γ−1 k = 0, 1, . . . , n, |τn,k (α)| = ak const, k which completes the proof of (7.17).

7.2. Derivatives of α-invariant functions

123

Theorem 7.16 ([14]). If f ∈ Ep (Ω), then for all a ∈ Ω, c(n, p) ≤

1 |f (n) (a)| ≤ C(n, p) max n+1/p n n! f p ≤1 λ (a)( |τn,k ( 1 )|2 )1/2 Ω

k=1

p

with the constants c(n, p) and C(n, p) such that C(n, p) = const n1/p−1/2 for 0 < p ≤ 2

c(n, p) = 1, and

c(n, p) = n1/2−1/p ,

C(n, p) = 1 for 2 ≤ p ≤ ∞.

Remark 7.17. For p = 2, Theorem 7.15 and Theorem 7.16 are well known (see for instance [114] ). Proof of Theorem 7.16. It is obvious that c(n, 2) = C(n, 2) = 1. Hence, c(n, p) ≥ 1 for 0 < p ≤ 2 and C(n, p) ≤ 1 for 2 ≤ p ≤ +∞, because Hp ⊂ Hx for p > x. Let p > 2. Hence, g2 ≤ gp . By the Hausdorﬀ-Young inequality, we have gn p ≤

n

1/q

|ak |

q

k=0

1 1 + =1 . p q

Consequently, by the H¨ older inequality, 1/2−1/p

gn p ≤ (n + 1)

n

1/2 2

|ak |

.

k=0

Thus, c(n, p) ≥ n1/2−1/p for p > 2. To complete the proof of Theorem 7.16, we have to prove that C(n, p) = O(n1/p−1/2 ) for 0 < p < 2. Let 0 < p ≤ 1. If g(ζ) = a0 +a1 ζ +· · · ∈ Hp and gp ≤ 1, then by the HardyLittlewood theorem (see Theorem 6.5 in [59]) an = o(n1/p−1 ). Consequently, C(n, p) = max ( gp ≤1

n

|ak |2 )1/2 = O(n1/p−1/2 ).

k=0

Let 1 < p < 2. If gp ≤ 1, then by the Hausdorﬀ-Young inequality,

∞

k=0

1/q |ak |

q

≤ gp

1 1 + =1 . p q

Using the H¨ older inequality, we obtain C(n, p) = O(n1/p−1/2 ).

124

7.3

Chapter 7. Related results

A characterization of convex domains

Everybody who reads Chapter 4, Section 5 very carefully will recognize that, for the proof of the 2n−1 conjecture, we only need one of the Marx-Strohh¨ acker theorems. This inequality characterises the functions of the closed convex hull of the convex functions, but not the convex functions themselves. On the other hand, we know from Chapter 4, Section 6 that convex pairs are the only ones for which C2 (Ω, Π) = 2. At ﬁrst glance, this seems to be a contradiction. Thinking of this problem, we found that an invariant formulation of the mentioned MarxStrohh¨ acker inequality in fact characterises convexity. The proof of this fact forms the content of this section and it seems to explain the geometrical background of Theorem 4.15. Let g0 be a function holomorphic and univalent in the unit disc Δ. We suppose that g0 (0) = 0, g0 (0) = 1 and Π := g0 (Δ) is a convex domain. To characterize Π we shall use the fact that for such a function g0 one of the well-known MarxStrohh¨ acker inequalities, namely Re

g0 (z) z

>

1 , 2

z ∈ Δ,

(7.18)

holds (see [115] and [155]). For uniﬁed proofs of this and many related inequalities one should consult [118]. The formula (7.18) is equivalent to the existence of a bounded holomorphic function ω : Δ → Δ such that g0 (z) 1 = , z 1 + zω(z)

z ∈ Δ.

(7.19)

We will consider the following question. Let h be a function holomorphic on acker Δ. Suppose that h (ζ) = 0 for any ζ ∈ Δ and that h(Δ) has the Marx-Strohh¨ property for any point z0 = h(t) ∈ h(Δ), i.e., the function g0 deﬁned by the Koebe transform

w+t − h(t) h 1+tw g0 (w) = h (t)(1 − |t|2 ) satisﬁes inequality (7.18) for any t ∈ Δ. This is equivalent to the inequality Re

h(z) − h(t) 1 − tz h (t) z−t

>

1 − |t|2 , 2

z ∈ Δ, t ∈ Δ.

(7.20)

What can be said about Ω = h(Δ)? We ﬁnd that h(Δ) is a convex domain, so that an assertion inverse to the Marx-Strohh¨ acker theorem is valid.

7.3. A characterization of convex domains

125

Theorem 7.18 ([24]). Let h be a function holomorphic in Δ, such that h (ζ) = 0, ζ ∈ Δ, and such that condition (7.20) is satisﬁed. Then (i) the function h is injective on Δ and h(Δ) = Ω is a convex domain, (ii) for any n ≥ 2 and any z ∈ Δ the following sharp estimate is valid: (n) h (z) n! h(n−1) (z) nz ≤ . h (z) − 1 − |z|2 n−1 h (z) (1 − |z|) (1 + |z|)

(7.21)

Proof. The condition (7.20) immediately implies h(t) = h(z) for z ∈ Δ, t ∈ Δ, t = z, and therefore the injectivity of the function h on Δ. Now, we ﬁx t ∈ Δ and we consider the function ϕ(z) = 2

h(z) − h(t) 1 − tz − 1, h (t) (z − t)(1 − |t|2 )

z ∈ Δ.

It is evident that ϕ(t) = 1 and that Re ϕ(z) > 0 for any z ∈ Δ. In a neighbourhood of the point t we have the Taylor expansion ∞ (n) h (t) t h(n−1) (t) (z − t)n−1 . ϕ(z) = 1 + 2 − (t)n! 2 )(n − 1)! (t) h (1 − |t| h n=2 Since 1/λΛ (ϕ(t)) = 2 Re ϕ(t) = 2, using the Schwarz-Pick lemma one easily gets h (t) 2 2t |ϕ (t)| = , t ∈ Δ, ≤ − 2 h (t) 1 − |t| 1 − |t|2 which is equivalent to the inequality 2 w − 1 + |t| ≤ 2|t| , 1 − |t|2 1 − |t|2 where w = 1+t

t ∈ Δ,

(7.22)

h (t) . h (t)

The condition (7.22) implies Re w > 0. Therefore (see for instance [128] and [146]), h is injective on Δ and Ω = h(Δ) is a convex domain. To get (ii) for n ≥ 3 we apply Ruscheweyh’s theorem 4.6 to get sharp estimates for the derivatives (n) ϕ(n−1) (t) h(n−1) (t) h (t) nt , t ∈ Δ, = 2 − (n − 1)! nh (t) 1 − |t|2 h (t) indicated in (ii). Equality in (7.21) at the point z = z0 ∈ Δ occurs if h(z) =

z z0 z = . 1 − z0 z/z0 z0 − z0 z

This completes the proof of Theorem 7.18.

126

Chapter 7. Related results

Remark 7.19. According to the above, the condition (7.20) is a new necessary and suﬃcient condition for h(Δ) to be convex that does not use the second derivative of h. It may be worthwhile to mention the conditions of this type that have been proved before. To our knowledge the ﬁrst one was zh (z) Re > 0, |t| < |z| < 1, h(z) − h(t) proved by Brickman in [46]. Sheil-Small ([151]) and Suﬀridge ([156]) proved the characterization 1 zh (z) t > , z ∈ Δ, t ∈ Δ, Re − h(z) − h(t) z−t 2 of convex funtions h. The third condition we want to cite seems to be the most famous. It has been proved by Ruscheweyh in [141] and it was used by Ruscheweyh and Sheil-Small in [144] to prove the P´ olya-Schoenberg conjecture. This one is as follows. z ζ − t h(z) − h(t) 1 ζ Re > , z ∈ Δ, t ∈ Δ, ζ ∈ Δ. − z − ζ z − t h(ζ) − h(t) z−ζ 2 One curious diﬀerence between these characterizations of convexity of h(Δ) and (7.20) seems to be that (7.20) contains nonanalytic terms.

Chapter 8

Some open problems 8.1

The Krzy˙z conjecture

As we have seen in Chapter 6 it is a very diﬃcult task to ﬁnd sharp punishing factors or substitutes for them in cases where multiply connected domains are involved. From this point of view, it seems natural that the diﬃculties become nearly insuperable, if one allows the points z0 ∈ Ω or f (z0 ) ∈ Π, or both to vary, and asks for the maximum. Nevertheless, there exists one problem of this type that has attracted researchers for many years because of the conjectured simple solution. This is the so-called Krzy˙z conjecture. It is concerned with functions f ∈ A(Δ, Δ ), where Δ = {z | 0 < |z| < 1} and the conjecture is that max

2 |f (n) (0)| | f ∈ A(Δ, Δ ) = n! e

for any n ∈ N (see [96]). It is obvious that the conjectured bound is attained for the functions 1 + zn f (z) = exp . 1 − zn The story of this conjecture began with a problem posed in [101], where the case n = 1 of the above is considered. It is remarkable that in the solutions found in 1934 the universal covering function of Δ plays the decisive role. After the formulation of the conjecture by Krzy˙z, there was a lot of eﬀort to solve this attractive problem. The cases n = 2 and n = 3 were solved in [84], for n = 4 see [161], [47], and [158]. The case n = 5 was solved recently in [145]. We cannot deny that our considerations for the case of multiply connected domains began with the hope to get a simple dependence of the upper bound of

128

Chapter 8. Some open problems

f (0) in this diﬃcult problem. We were able with the above theorems to get the simple cases n = 1 and n = 2, but not more. We may be allowed to conclude this section with a little observation which may prevent others from attacking this problem. In the above cited article, Szapiel poses the following problem: Do the estimates n (8.1) cj Aj (F ) ≤ n, j=1 hold for all schlicht functions F (z) = z +

∞

An (F )z n ,

n=2

whenever

n j−1 c ζ j ≤ 1 for all |ζ| ≤ 1? j=1

(8.2)

We ﬁnd that the answer is positive and that more is true. Theorem 8.1. Let cj satisfy the condition (8.2). Then the inequality (8.1) holds for any function F subordinate to a schlicht function F0 ∈ S. Proof. If we deﬁne P (z) =

n

cj ζ j ,

j=1

we see that the Schwarz lemma indicates that the condition (8.2) is equivalent to P (Δ) ⊂ Δ. Therefore the Sheil-Small theorem (see Chapter 2 and [152]) imply that the inequality |(P ∗ F )(eiθ )| ≤ n holds for all θ ∈ [0, 2π] and not only for all F ∈ S, but even for the case that F is subordinated to a schlicht function. This proves the theorem, a far-reaching generalization of (8.1).

8.2

The angle conjecture

As we have seen in Chapter 5, for f ∈ A(Ω, Hα ), α ∈ [1, 2], the inequalities n (λΩ (z0 )) 2n−1 |f (n) (z0 )| n+α−1 ≤ , n ≥ 2, z0 ∈ Ω, n n! α λΠ (f (z0 )) are valid as well for Ω = Δ as for Ω = H1 . It seems reasonable that these inequalities are true for any convex domain Ω. Concerning this conjecture, we will show now that it is true, if f (z0 ) lies on the line bisecting the angle Hα .

8.2. The angle conjecture

129

Theorem 8.2. Let Ω be a convex proper subdomain of C and Hα an angular domain with opening angle απ, α ∈ [1, 2]. Let further f ∈ A(Ω, Hα ), z0 ∈ Ω, and f (z0 ) lying on the bisector of Hα . Then for n ≥ 2 the inequalities n (λΩ (z0 )) |f (n) (z0 )| 2n−1 n+α−1 ≤ n n! α λΠ (f (z0 )) are valid. Proof. Let v be the vertex of the angular domain Hα and f (z0 ) − v = |f (z0 ) − v|eiϕ0 . Then the Riemann mapping function ΦHα ,f (z0 ) is given by ΦHα ,f (z0 ) (ζ) = v + (f (z0 ) − v)

1 + ze−iϕ0 1 − ze−iϕ0

α .

Hence, λHα (f (z0 )) =

1 . 2α|f (z0 ) − v|

Choosing an appropriate function ω : Δ → Δ such that ˜ Φ(ζ) = ΦHα ,f (z0 ) (ζω(ζ)) ˜ has the property Φ(Δ) = f (Ω), we get ˜ ◦ ΨΩ,z . f =Φ 0 If we now proceed as in the above proofs of the generalizations of the two Chua conjectures, we see that for the proof of the present assertion we have to show that n 2n−1 n+α−1 , (8.3) ck (α)An,k ≤ n α k=1

where the function represented by the series ∞

ck (α)ζ k

k=1

is subordinated to the function α ∞ 1+ζ 1 −1 = hk (α)ζ k 2α 1−ζ

(8.4)

k=1

and the An,k represent the coeﬃcients of the powers of inverses of convex functions as above.

130

Chapter 8. Some open problems

Analogous to the proofs mentioned above we want to show now that for functions ∞ ω(z) = dτ ζ k , τ =0

mapping the unit disc into the closed unit disc, the inequalities p−1 (p − τ )cp−τ (α)dτ ≤ p hp (α), p ∈ N,

(8.5)

τ =0

are valid. To this end, we consider, as in the proof of the 2n−1 conjecture, the region of variability of the linear functional Lω (g) =

p−1

(p − τ )cp−τ (α)dτ ,

τ =0

where ω is ﬁxed and g is varying in the family of functions subordinated under (8.4). According to [44], [45], and [146] these functions g have a representation α 2π 1 1 + eit ζ g(ζ) = − 1 d μ(t), 2α 0 1 − eit ζ with a probability measure μ : [0, 2π] → R. Hence, it is suﬃcient to prove the inequality (8.5) for cτ (α) = eiτ hτ (α). Further, the rotational invariance of the family of unimodular bounded functions indicates that we only need to show that p−1 (8.6) (p − τ )hp−τ (α)dτ ≤ p hp (α), p ∈ N. τ =0

We want to thank at this point St. Ruscheweyh for pointing out to us the idea of this proof. His idea was to consider the polynomial Q(ζ) =

p−1 (p − τ )hp−τ (α) τ =0

p hp (α)

ζk

and to show that the function Q ∗ ω is a unimodular bounded function. According to [152] (see especially the theorems (2.6) and (3.2) of this article), this can be reduced to the proof of Re(Q(ζ)) >

1 , 2

ζ ∈ Δ.

By the application of a theorem due to Rogosinski in [138], to achieve this aim it is suﬃcient to prove that the coeﬃcients of Q form a monotonic decreasing convex sequence. This means that (k + 1)hk+1 (α) − k hk (α) ≥ 0,

k = 1, . . . , p − 1,

8.3. The generalized Goodman conjecture

131

(k + 2)hk+2 (α) − 2(k + 1)hk+1 (α) + k hk (α) ≥ 0,

k = 1, . . . , p − 2,

and 2h2 (α) − 2h1 (α) ≥ 0. To prove these inequalities we have to insure that the coeﬃcients of 1 d 2α dζ

1+ζ 1−ζ

α =

(1 + ζ)α−1 (1 − ζ)α+1

form a monotonic increasing convex sequence. This is easily seen by recognizing that 1+ζ (1 + ζ)α−1 (1 − ζ)2 = e(α−1) log( 1−ζ ) α+1 (1 − ζ) has nonnegative coeﬃcients. A comparison with the proofs of the generalizations of the two conjectures of Chua above reveals that we can get the inequality (8.3), setting again (ω(ζ))σ =

∞

dj,σ ζ j ,

ζ∈Δ

j=0

and using (8.6) in the chain n n−1 1 n n−1 ck (α)An,k ≤ |cn (α)| + (n − j)cn−j (α)dj−σ,σ σ n σ=1 j=σ k=1 n−1 1 n (n − σ)hn−σ (α) ≤ hn (α) + σ n σ=1 n 2n−1 n+α−1 n−1 . hk (α) = = n n−k α k=1

The last identity has been shown in Chapter 5.

8.3

The generalized Goodman conjecture

In many of the results of the present book it could be proved that the bounds for the moduli of the Taylor coeﬃcients of the functions in a certain family of functions holomorphic in Δ are the right bounds for the related coeﬃcients of the functions subordinated to the above ones. Therefore, we dare to formulate a challenging conjecture that generalizes as well the Rogosinski conjecture, which naturally is a theorem nowadays, as Jenkins’ theorem 2.7 on the Goodman conjecture.

132

Chapter 8. Some open problems

The generalized Goodman conjecture. Let f be injective and meromorphic in Δ. Further let f be normalized by f (0) = f (0) − 1 = 0 and have the pole at the point p ∈ (0, 1). If ω : Δ :→ Δ, then we conjecture that for the expansion g(z) = f (zω(z)) =

∞

an (g)z n ,

n=1

valid in some neighbourhood of the origin, the inequalities |an (g)| ≤

n−1

1 pn−1

p2k ,

n ≥ 2,

k=0

are valid. This relation between f and g will be denoted by g ≺ f as in the holomorphic case. To support this conjecture we prove that it is valid in some special cases. The ﬁrst one deals with little values of p. Theorem 8.3. The generalized Goodman conjecture is valid for n = 2, p ∈ (0, 1), for n = 3, p ∈ (0, 0.7), and for n ≥ 4, p ∈ (0, 1/(2n − 2)). ∞ an z n , |z| < p meromorphic and univalent in D with Proof. For f (z) = z + n=2 ∞ pole p ∈ (0, 1) and ω(z) = n=0 cn z n , ω : D → D, we have to ﬁnd the least upper bound for |c1 + a2 (c0 )2 | in the case n = 2 and for |c2 + 2c0 c1 a2 + (c0 )3 a3 | in the case n = 3. In the ﬁrst the triangle inequality, Jenkins’ inequality (see [86]), and the Schur algorithm (see[148]) deliver as an upper bound 1 − |c0 |2 + |c0 |2

1 + p2 1 + p2 ≤ . p p

In the second case, we use c2 |c1 |2 c0 (c1 )2 ≤ 1− + 1 − |c0 |2 2 2 (1 − |c0 | ) (1 − |c0 |2 )2 (see[148]). Using the triangle inequality and the abbreviations x = |c0 |, we get

y=

|c1 | (1 − |c0 |2 )

|c2 | ≤ (1 − x2 )(xy 2 + 1 − x2 ).

8.3. The generalized Goodman conjecture

133

Therefore, we have to maximize (1 − x2 )(xy 2 + 1 − x2 ) + 2xy(1 − x2 )

1 + p2 + p 4 1 + p2 = F (x, y; p) + x3 p p2

for (x, y) ∈ [0, 1] × [0, 1]. Since ∂F 1 + p2 2 ≥ (≤) 0 = 2(1 − x ) y(x − 1) + x ∂y p for y ≤ (≥) and

x 1 + p2 , 1−x p

x 1 + p2 ≤1 1−x p

for

p , 1 + p + p2 in these cases we have to check the local maximum of F in x≤

y=

x 1 + p2 . 1−x p

Inserting this delivers a monotonic increasing function of x. So we have to check that this function at p x= 1 + p + p2 is less than (1 + p2 + p4 )/p2 . This is the case for p ∈ (0, 1). For p x≥ 1 + p + p2 we have to consider F (x, 1, p) = x

1 − 2p − 2p3 + p4 2 + p + 2p2 + x3 . p p

For those p for which the derivative of this function with respect to x is positive for x ≤ 1, the proof is complete. This is the case for ! √ 2 + 19 p≤ = 0.7088023 . . . . 3 For n ≥ 4 we consider (zω(z))k =

∞ j=k

cj,k z j

134

Chapter 8. Some open problems

and we see that ck,k = (c0 )k and therefore |cj,k | ≤ 1 − |c0 |2k ,

j ≥ k.

Hence, ⎛ ⎞ n−1 n k−1 n−1 p2j p2j ⎝ ⎠ + |c0 |n |an (g)| = cn,k ak ≤ . 1 − |c0 |2k pk−1 pn−1 k=1

j=0

k=1

j=0

n−1 We want to prove that this is less than or equal to j=0 (p2j /pn−1 ). This inequality is equivalent to ⎛ ⎞ k−1 n−1 n−1 n−1 2k−1 p2j p2j j ⎝ ⎠≤ |c0 |j |c | . 0 k−1 p pn−1 j=0 j=0 j= j=0 k=1

We use the rough estimate 2k−1

|c0 |j ≤ 2

j=0

n−1

|c0 |j ,

1 ≤ k ≤ n − 1,

j=0

and we see by some elementary calculations that 2

n−1 k−1 k=1 j=0

n−1 p2j p2j ≤ pk−1 pn−1 j=0

for p ∈ (0, 1/(2n − 2)).

The second theorem is concerned with the condition that f (Δ) is concave, i.e., C\f (Δ) is a convex compact set. The family of functions with these properties will be denoted by Co(p) as in Chapter 5, Section 1, and the family of functions ω : Δ → Δ by B. Theorem 8.4 (see [26]). Let p ∈ (0, 1). If f ∈ Co(p) and g ≺ f, then |an (g)| ≤

1 pn−1

n−1

p2k ,

n ∈ N.

(8.7)

k=0

For the proof of Theorem 8.4 we use a representation formula for the functions in the class Co(p) that was derived in [167] and [23] as a simple consequence of Theorem 4 in [117]. Theorem 8.5. Let p ∈ (0, 1). For any f ∈ Co(p) there exists a function w1 ∈ B such that p 1 − 1+p 2 (1 + w1 (z))z z f (z) = , z ∈ Δ. (8.8) z 1− p 1 − zp

8.3. The generalized Goodman conjecture

135

The ﬁrst step in the proof of Theorem 8.4 concerns the ﬁrst factor in the representation (8.8) that we will denote by f1 , f1 (z) =

z , 1 − pz

z ∈ Δ.

We prove that, for this member of Co(p), one may replace the sum in Theorem 8.4 by 1. Theorem 8.6. Let p ∈ (0, 1). If g ≺ f1 , then |an (g)| ≤

1 pn−1 ,

n ∈ N.

Proof. Let w ∈ B such that g(z) = f1 (zw(z)), z ∈ Δ. This implies zg(z) , z ∈ Δ. g(z) = w(z) z + p

(8.9)

Now, to prove Theorem 8.6, we use a method due to Clunie (see [57]) and a generalization of a theorem of Robertson (see [137]) proved in [30], Lemma 2.1. This is the meromorphic version of the Rogosinski lemma (see our Theorem 2.4). The application of this lemma to the identity (8.9) yields that for any n ∈ N, the inequality n n 2 |ak−1 (g)| 2 |ak (g)| ≤ 1 + p2 k=1

k=2

is valid. The assertion of Theorem 8.6 follows by mathematical induction using 1/p > 1. Next, we consider the second factor in the representation (8.8) that we will denote by f2 and we prove the following inequalities. Theorem 8.7. Let p ∈ (0, 1) and for w ∈ B, g2 (z) = f2 (zw(z)) = 1 +

∞

Bk z k ,

z ∈ D.

k=1

Then for any n ∈ N the inequality n

2

|Bk | ≤

k=1

n

p2k

(8.10)

k=1

is valid. Proof. Obviously, f2 (z) = 1 + Since

p2 −w1 (z) 1+p2

pz

1 − zp

2 p − w1 (z) 1 + p2 ≤ 1,

,

z ∈ Δ.

z ∈ Δ,

136

Chapter 8. Some open problems

there exists a function w2 ∈ B such that w2 (z) p z , 1 − zp

f2 (z) = 1 +

z ∈ Δ.

Let w3 (z) = w2 (zw(z)), z ∈ Δ. Then w3 ∈ B. Hence, the function g2 − 1, g2 (z) − 1 =

w3 (z) p zw(z) , 1 − zw(z)p

z ∈ Δ,

is quasi-subordinate to the function h, h(z) =

pz , 1 − zp

z ∈ Δ.

The notation of quasi-subordination was deﬁned by Robertson (see [137]). He proved that the above relation implies in our case the inequality (7.9) (compare also [128], Theorem 2.2). This completes the proof of Theorem 8.5. After these preparations, it is easy to prove Theorem 8.4 as follows. Proof of Theorem 8.4. If g ≺ f and f ∈ Co(p), we conclude from Theorem 8.5 that there exists a function w ∈ B such that g(z) = f1 (zw(z)) f2 (zw(z)). Let f1 (zw(z)) =

∞

(8.11)

Ak z k ,

k=1

where this expansion is valid in some neighbourhood of the origin. From (8.11) we get for n ∈ N, n−1 Ak Bn−k . an (g) = An + k=1

The Cauchy-Schwarz inequality shows that, for any n ∈ N, the inequality |an (g)| ≤

n

1/2 2

|Ak |

1+

k=1

n−1

1/2 2

|Bk |

k=1

is valid. If we use Theorems 8.6 and 8.7 to estimate the right-hand side of this inequality, we get |an (g)| ≤

n−1 k=0

p−2k

1/2 n−1 k=0

This completes the proof of Theorem 8.4.

1/2 p2k

=

1 pn−1

n−1

p2k .

k=0

8.3. The generalized Goodman conjecture

137

It is obvious that the estimates in Theorems 8.4 and 8.6 are sharp. In Theorem 8.6, equality is attained if g(z) =

cz , 1 − cz p

|c| = 1.

In Theorem 8.4 equality occurs for g(z) =

cz , 1 − cz (1 − zp) p

|c| = 1.

These are conformal maps of Δ onto % $ −p −p , C\ (1 − p)2 (1 + p)2 (compare [15] and [166]). The third theorem proves the generalized Goodman conjecture for univalent functions with real coeﬃcients. They belong to the class of functionss typicallyreal and meromorphic, introduced by Goodman in [72]. The deﬁning relation for this class is, as in the holomorphic case, (f (z) (z) ≥ 0,

z ∈ Δ.

It is easily seen that this implies that in our case the residuum of f at the point p is negative. Goodman proved a representation theorem that implies the following application to univalent meromorphic and typically-real functions normalized as usual. Theorem 8.8 (see [72] Theorem 9). Let f be a function univalent meromorphic and typically-real in Δ having the residuum −m < 0 at the point p. There exists a function t holomorphic and typically-real in Δ and a nonnegative constant M such that 1 M +m −1 =1 p and

f (z) = M t(z) + m

1 −1 p

z , (1 − zp)(1 − z/p)

z ∈ Δ.

Using this representation it is easy to prove (8.7) for univalent meromorphic and typically-real functions. Theorem 8.9. Let f be univalent meromorphic and typically-real in Δ. If f has its pole at the point p ∈ (0, 1) and g ≺ f , then (8.7) holds.

138

Chapter 8. Some open problems

Proof. According to a theorem proved by Robertson in [135], for any function typically-real and holomorphic in Δ there exists a probability measure μ on [0, π] such that π z dμ(θ) t(z) = , z ∈ Δ. 1 − 2z cos θ + z 2 0 Since the kernel functions belong to the class S, the truth of the Rogosinski conjecture for univalent functions and a convex hull argument imply that for g1 ≺ t the inequalities |an (g1 )| ≤ n, n ≥ 2 hold. From Theorem 8.4 above we know that for g2 (z) ≺

z , (1 − zp)(1 − z/p)

the generalized Goodman conjecture holds. Hence, |an (g2 )| ≤

1 pn−1

n−1

p2k .

k=0

From the identity 1 pn−1

n−1

p2k =

k=0

⎧ u ⎨

1 + p2k−1 u 1 k=1 p2k

k=1

⎩ 1+

we conclude that

n−1

1 pn−1

p2k−1 , if n = 2u, + p2k , if n = 2u + 1

p2k ≥ n.

k=0

If we write a function g ≺ f in the form 1 g(z) = M g1 (z) + m − 1 g2 (z), p

z ∈ Δ,

we see that |an (g)| ≤ M |an (g1 )| + m

n−1 1 2k 1 p . − 1 |an (g2 )| ≤ n−1 p p k=0

This proves Theorem 8.9.

Using the bound found by Jenkins and well-known subordination techniques, it is not diﬃcult to ﬁnd upper bounds for |an (g)|. We give two examples ⎛ ⎞ k−1 n 1 ⎝ p2j ⎠ |an (g)| ≤ pk−1 j=0 k=1

8.4. Bloch and several variable problems and

139

⎛ ⎞2 ⎞1/2 k−1 n ⎟ ⎜ ⎝ 1 p2j ⎠ ⎠ . |an (g)| ≤ ⎝ pk−1 j=0 ⎛

k=1

It seems that these estimates reﬂect approximately the right asymptotics of the least upper bounds for p → 0, but it is evident that they are bad for p → 1. An upper bound that ﬁts better for the neighbourhood of p = 1 can be found if one applies the validity of the Rogosinski conjecture to the function fp (z) = f (pz)/p that belongs to the class S. By this procedure we get |an (g)| ≤

n . pn−1

We add a last inequality for the coeﬃcients an (g) that follows immediately from the generalized Goluzin-Rogosinski theorem 2.5 using Jenkins’ theorem 2.7. Taking λk = we get

n k=1

pn−1 k−1 2j j=0 p

|ak (g)|2 λk ≤

2

n

,

k ∈ N,

|ak |2 λk ≤ n.

k=1

Obviously, this inequality is sharp for the function for which Jenkins’ theorem is sharp.

8.4

Bloch and several variable problems

We want to avoid the impression that the above considerations are the only possibility to generalize the Schwarz-Pick lemma to higher derivatives. Therefore we mention some questions that arose when we studied the work of colleagues on generalized Schwarz-Pick type estimates (see [38], [112], and [113]). For example, it has been proved in [112] that for f ∈ A(Δ, Δ), α, β > 0, the implications |f (z)|(1 − |z|2 )β < ∞ 2 α z∈Δ (1 − |f (z)| ) sup

|f (n) (z)|(1 − |z|2 )β+n−1 < ∞, (1 − |f (z)|2 )α z∈Δ

=⇒ sup

n ≥ 2,

are valid. On one hand it seems natural to ask whether one can compute an explicit relation between these two suprema. On the other hand this result together with

140

Chapter 8. Some open problems

our theorems on punishing factors suggests the question for which pairs (Ω, Π) of domains and α, β > 0 implications of the form |f (z)|R(Ω, z)β < ∞ α z∈Ω R(Π, f (z)) sup

|f (n) (z)|R(Ω, z)β+n−1 < ∞, R(Π, f (z))α z∈Ω

=⇒ sup

n ≥ 2,

are valid. The results of MacCluer, Stroethoﬀ, and Zhao in [113] may give rise to analogous problems even for functions of several variables. Since it would be lengthy to cite their deﬁnitions and theorems, we prefer to conclude with open questions that are simpler to formulate and that originated in the following result of B´en´eteau, Dahlner, and Khavinson (see [38]). They proved that for an analytic function f : Δn → Δ,

(z1 , . . . , zn ) → f (z1 , . . . , zn )

and any multiindex α = (α1 , . . . , αn ) ∈ (N ∪ {0})n the inequality sup (z1 ,...,zn )∈Δn

α ,n |Dα f (z1 , . . . , zn )| k=1 1 − |zk |2 k < ∞ 1 − |f (z)|2

holds, where Dα f (z1 , . . . , zn ) is the usual abbreviation for the derivative of order α. Again, one may ask for estimates for this supremum or consider generalizations of this result to other domains instead of Δ. For new results on similar generalizations of Schwarz-Pick estimates compare [5], [6], [68] and [92].

8.5

On sums of inverse coeﬃcients

As we have seen frequently above, a decisive part in the computation of punishing factors is played by inverse coeﬃcients An,k of the kth powers of functions injective in the unit disc. L¨ owner’s Theorem 2.7 and its generalizations provided us with satisfactory solutions for these problems. If subclasses of S are considered, the situation is far less nice. The most famous of such problems is the question for bounds in the convex case. For the function z f (z) = , |c| = 1, (8.12) 1 − cz the identities n−1 |An,k | = k−1

8.5. On sums of inverse coeﬃcients

141

hold for 1 ≤ k ≤ n ∈ N. R. L. Libera and E. J. Zlotkiewicz proved in [104] that |An,1 | ≤ 1,

n = 2, 3, 4, 5, 6, 7

for the inverse coeﬃcients of convex functions. This result implies the validity of the estimate n−1 |An,k | ≤ (8.13) k−1 for n between 2 and 7. On the other hand W. E. Kirwan and G. Schober got by explicit computations that M10 = max {|A10,1 |} > 1.2, where the maximum is taken over all convex functions (see [90]). A detailed discussion of Mn = max {|An,1 |} may be found in [50] and [58]. J.T.P. Campschroer showed in [50] that Mn = O 2n n−3 . It may be interesting that for k big compared with n the inequalities (8.13) hold true. Theorem 8.10. For the inverse coeﬃcients An,k of convex functions, n ≥ 4, and n/2 − 1 ≤ k ≤ n, the inequalities (8.13) are valid. Proof. As above, we use the Schur-Jabotinsky theorem and, in addition, the fact that a convex function f is starlike of order 1/2. This implies that the function h deﬁned by 2 f (z) h(z) = z z is a starlike function. Hence, we may consider the Taylor coeﬃcients of n/2 ∞ z = bl,n/2 (h)z n h(z) n=0

where h is starlike. But from [140] we know that the modulus bl,n/2 (h) of the Taylor coeﬃcients of these functions is bounded from above by bl,n/2 (k), where k is the Koebe function or one of its rotations and l ≤ n/2 + 1. In most cases, the Koebe functions are the only extremal functions. This proves our theorem. But in our computations of punishing factors, weighted sums of the An,k are the most important elements. If one looks at punishing factors for convex pairs (see section 4.5), one may recognize that the proof of Theorem 4.12 would be very short, if the following conjecture (compare [22]) could be proved:

142

Chapter 8. Some open problems

Conjecture. For the inverse coeﬃcients of convex functions and for any n ≥ 2, the inequality n |An,k | ≤ 2n−1 k=1

is valid. The validity of this conjecture for 2 ≤ n ≤ 7 follows immediately from the above. A sharp estimate for the sum of squares can be derived from section 4.5 and an application of Theorem 2.5 on subordinate functions. Actually, one easily gets n

2

|An,k | =

n−1 μ=0

k=1

≤

2 n − μ a μ,n n

n−1 μ=0

n−μ n

2

n μ

2

=

2(n − 1) n−1

.

Here, equality is attained for the functions (8.12). If one uses this inequality to estimate the sum of moduli of the inverse coefﬁcients by the Cauchy inequality and the Stirling formula, one gets immediately n k=1

|An,k | ≤

√

n

2(n − 1) n−1

1/2

≤ 2n−1 n1/4 .

If one wants to compute punishing factors for functions holomorphic on βaccessible domains, it would be necessary to know something of the inverse coeﬃcients of functions close-to-convex of order β or of weighted sums of the moduli of these inverse coeﬃcients (compare section 7.1 and section 5.6 on lower bounds). The problem of such inverse coeﬃcients is addressed in [90], but those results do concern only early coeﬃcients. It seems natural to us to conjecture inequalities for the sums mentioned above to get new conjectures on punishing factors.

Bibliography [1] S. Agard, Distortion theorems for quasiconformal mappings. Ann. Acad. Sci. Fenn. Ser. AI 413 (1968), 12p. [2] D. Aharonov and S. Friedland, On an inequality connected with the coeﬃcient conjecture for functions of bounded boundary rotation. Ann. Acad. Sci. Fenn. Ser. AI 524 (1972), 14p. [3] L. V. Ahlfors, Conformal invariants, Topics in Geometric Function Theory. McGraw-Hill, New York, 1973. [4] A. Ancona, On strong barriers and an inequality of Hardy for domains in Rn . J. London Math. Soc.(2) 34 (1986), no.2, 274–290. [5] J. M. Anderson, M. A. Dritschel, and J. Rovnyak, Schwarz-Pick inequalities for the Schur-Agler class on the polydisk and unit ball. Comp. Methods and Function Theory 8 (2008), 339–361. [6] J. M. Anderson and J. Rovnyak, On generalized Schwarz-Pick estimates. Mathematika 53 (2006), 161–168. [7] J. M. Anderson, J. G. Clunie, and Ch. Pommerenke, On Bloch functions and normal functions. J. Reine Angew. Math. 270 (1974), 12–37. [8] J. Arazy, S. Fisher, and J. Peetre, M¨ obius Invariant Spaces of Analytic Functions. In: Lecture Notes in Math., Springer-Verlag, Nr. 1275 (1987), 10–22. [9] K. Astala, J. L. Fern´ andez, and S. Rohde, Quasilines and the Hayman-Wu theorem. Indiana Univ. Math. J. 42 (1993), no.4, 1077–1100. [10] F. G. Avhadiev and L. A. Aksent’ev, The main results on suﬃcient condition for an analytic function to be schlicht. Russ. Math. Surv, 30 (1975), no.4, 1–63 (translated from Usp. Mat. Nauk, 1975, t. 30, no. 4(184), 3–63). [11] F. G. Avkhadiev, Solution of the Generalized Saint Venant Problem. Sb. Math. 189 (1998), 1739–1748 (translated from Mat. Sb. 189(12) (1998), 3– 12). [12] F. G. Avkhadiev, Hardy Type Inequalities in Higher Dimensions with Explicit Estimate of Constants. Lobachevskii J. Math. 21 (2006), 3–31 (electronic).

144

Bibliography

[13] F. G. Avkhadiev, Hardy-type inequalities on planar and spatial open sets. Proceedings of the Steklov Institute of Mathematics 255 (2006), no.1, 2–12 (translated from Trudy Matem. Inst. V.A. Steklova, 2006, v.255, 8–18). [14] F. G. Avkhadiev and K.-J. Wirths, Estimates of derivatives in an αinvariant set of analytic functions. Analysis 19 (1999),19-27. [15] F. G. Avkhadiev and K.-J. Wirths, Convex holes produce lower bounds for coeﬃcients. Complex Variables 47 (2002), 553–563. [16] F. G. Avkhadiev and K.-J. Wirths, Schwarz-Pick inequalities for derivatives of arbitrary order. Constr. Approx. 19 (2003), 265–277. [17] F. G. Avkhadiev and K.-J. Wirths, Punishing factors for angles. Comp. Methods and Function Theory 3 (2003), 127–141. [18] F. G. Avkhadiev and K.-J. Wirths, Poles near the origin produce lower bounds for coeﬃcients of meromorphic univalent functions. Michigan Math. J. 52 (2004), 119–130. [19] F. G. Avkhadiev and K.-J. Wirths, Schwarz-Pick inequalities for hyperbolic domains in the extended plane. Geom. Dedicata 106 (2004), 1–10. [20] F. G. Avkhadiev and K.-J. Wirths, The conformal radius as a function and its gradient image. Isr. J. Math. 145 (2005), 349–374. [21] F. G. Avkhadiev and K.-J. Wirths, Punishing factors for ﬁnitely connected domains. Monatshefte f. Math. 147 (2006), 103–115. [22] F. G. Avkhadiev and K.-J. Wirths, Sharp bounds for sums of coeﬃcients of inverses of convex functions. Comp. Methods and Function Theory 7 (2007), 105–109. [23] F. G. Avkhadiev and K.-J. Wirths, A proof of the Livingston conjecture. Forum Math. 19 (2007), 149–157. [24] F. G. Avkhadiev and K.-J. Wirths, The punishing factors for convex pairs are 2n−1 . Revista Mat. Iberoamericana 23 (2007), 847–860. [25] F. G. Avkhadiev and K.-J. Wirths, Punishing factors and Chua’s conjecture. Bull. Belg. Math. Soc., Simon Stevin 14 (2007), 333–340. [26] F. G. Avkhadiev and K.-J. Wirths, Subordination under concave univalent functions. Complex Variables and Elliptic Equations 52 (2007), 299–305. [27] F. G. Avkhadiev and K.-J. Wirths, A Theorem of Teichm¨ uller, Uniformly Perfect Sets and Punishing factors. Preprint (Tech. Univ. Braunschweig, 2005) http://www.iaa.tu-bs.de/wirths/preprints/teich.ps [28] F. G. Avkhadiev and K.-J. Wirths, Estimates of the derivatives of meromorphic maps from convex domains into concave domains. Comp. Methods and Function Theory 8 (2008), 107–119.

Bibliography

145

[29] F. G. Avkhadiev and K.-J. Wirths, Local Taylor coeﬃcients and subordination. Preprint (Tech. Univ. Braunschweig, 2005) http://www.iaa.tu-bs.de/wirths/preprints/local.ps [30] F. G. Avkhadiev, Ch. Pommerenke, and K.-J. Wirths, On the coeﬃcients of concave univalent functions. Math. Nachr. 271 (2004), 3–9. [31] F. G. Avkhadiev, Ch. Pommerenke, and K.-J. Wirths, Sharp inequalities for the coeﬃcients of concave schlicht functions. Comment. Math. Helv. 81 (2006), 801–807. [32] A. Baernstein, II, Integral means, univalent functions and circular symmerization. Acta Math. 133 (1974), 139–169. [33] A. Baernstein, II and G. Schober, Estimates for inverse coeﬃcients of univalent functions from integral means. Israel J. Math. 36 (1980), 75-82. [34] C. Bandle and M. Flucher, Harmonic radius and concentration of energy, n+2 hyperbolic radius and Liouville’s equation U = eU and U = U n−2 . SIAM Review 38 (1996), 191–238. [35] H. Bateman, The k-function, a particular case of the conﬂuent hypergeometric function. Trans. Amer. Math. Soc. 33 (1931), 817–831. [36] A. F. Beardon and F. W. Gehring, Schwarzian derivatives, the Poincar´e metric and the kernel function. Comment. Math. Helv. 55 (1980), 50–64. [37] A. F. Beardon and Ch. Pommerenke, The Poincar´e metric of plane domains. J. London Math. Soc. (2) 18 (1978), 475–483. [38] C. B´en´eteau, A. Dahlner, and D. Khavinson, Remarks on the Bohr phenomenon. Comp. Methods and Function Theory 4 (2004), 1–19. [39] A. Bermant, Dilatation d’ une fonction modulaire et probl`emes de recouvrement (in Russian, French summary). Matematicheskii Sbornik 15(57) (1944), No. 2, 285–318. ¨ [40] L. Bieberbach, Uber die Koeﬃzienten derjenigen Potenzreihen, welche eine schlichte Abbildung des Einheitskreises vermitteln. Sitz. Ber. Preuss. Akad. Wiss., (1916), 940–955. [41] M. Biernacki, Sur la repr´esentation conforme des domaines lin´earement accessibles. Prace Mat. Fiz. 44 (1936), 293–314. ´ Borel, Sur les z´eros des fonctions enti`eres. Acta Math. 20 (1896 - 97), [42] E. 357-396. [43] L. de Branges, A proof of the Bieberbach conjecture. Acta Math. 154 (1985), 137-152. [44] D. A. Brannan, On coeﬃcient problems for certain power series. Proc. of the Symposium on Complex Analysis (Univ. Kent, Canterbury,1973, London Mat. Soc. Lecture Note Series, Cambridge Univ. Press) 12 (1974), 17–27.

146

Bibliography

[45] D. A. Brannan, J. G. Clunie, and W. E. Kirwan, On the coeﬃcient problem for functions of bounded boundary rotation. Ann. Acad. Sci. Fenn. AI 523 (1973), 18p. [46] L. Brickman, Subordinate families of analytic functions. Illinois J. Math. 15 (1971), 241–248. [47] J. E. Brown, Iterations of functions subordinate to schlicht functions. Complex Variables 9 (1987), 143–152. [48] R. B. Burckel, An Introduction to Classical Complex Analysis, Vol.1. Birkh¨ auser Verlag, Basel und Stuttgart, 1979. [49] L. A. Caﬀarelli and A. Friedman, Convexity of solutions of semilinear elliptic equations. Duke Math. J. 52 (1985), 15–44. [50] J. T. P. Campschroer, Growth of the coeﬃcients of the inverse of a convex function. Indag. Math. 46 (1984), 1–6. ¨ [51] C. Carath´eodory, Uber den Variabilit¨ atsbereich der Koeﬃzienten von Potenzreihen, die gegebene Werte nicht annehmen. Math. Ann. 64 (1907), 95–115. ¨ [52] C. Carath´eodory, Uber den Variabilit¨ atsbereich der Fourierschen Konstanten von positiven harmonischen Funktionen. Rend. Cir. Mat. Palermo 32 (1911), 193–217. [53] C. Carath´eodory, Untersuchungen u ¨ber die konformen Abbildungen von festen und ver¨ anderlichen Gebieten. Mat. Ann. 52:1 (1912), 107–144. [54] C. Carath´eodory, Theory of functions of a complex variable. Chelsea Publ. Comp., New York, 1960. [55] L. Carleson and T. W. Gamelin, Complex dynamics. Springer, New York, 1993. [56] K. S. Chua, Derivatives of univalent functions and the hyperbolic metric. Rocky Mountain J. Math. 26 (1996), 63–75. [57] J. Clunie, On meromorphic schlicht functions. J. London Math. Soc. 34 (1958), 215–216. [58] J. G. Clunie, Inverse coeﬃcients for convex univalent functions. Journal d’An. Math. 36 (1979), 31–35. [59] P. L. Duren, Theory of H p spaces. Academic Press, New York, 1970. [60] P. L. Duren, Univalent functions. Springer, New York, 1980. ¨ [61] L. Fej´er, Uber gewisse durch die Fouriersche und Laplacesche Reihe deﬁnierten Mittelkurven und Mittelﬂ¨ achen. Palermo Rend. 38 (1914), 79– 97.

Bibliography

147

[62] J. L. Fern´ andez, Domains with strong barrier. Rev. Math. Iberoamericana 5 (1989), no.1-2, 47–65. [63] J. L. Fern´ andez and J. M. Rodr´ıguez, The Exponent of Convergence of Riemann Surfaces. Bass Riemann Surfaces. Ann. Acad. Sci. Fenn., Ser. A.I: Math. 15 (1990), 165–182. [64] C. H. FitzGerald and Ch. Pommerenke, The de Branges theorem on univalent functions. Trans. Amer. Math. Soc. 290 (1985) 683–690. [65] F. P. Gardiner and N. Lakic, Comparing Poincar´e densities. Ann. Math. 154 (2001), 115–127. [66] J. B. Garnett, Bounded analytic functions. Academic Press, Inc. New York, London, 1981. [67] J. B. Garnett and D. E. Marshall, Harmonic measure. Cambridge Univ. Press, Cambridge, 2005. [68] P. Ghatage and D. Zheng, Hyperbolic derivatives and generalized SchwarzPick estimates. Proc. Amer. Math. Soc. 132 (2004), 3309–3318. [69] G. M. Goluzin, On majorant of subordinate analytic functions I. Mat. Sb. 29(71) (1951), 209–224. [70] G. M. Goluzin, Geometric theory of functions of a complex variable. AMS, Providence, 1969. [71] S. Gong, The Bieberbach Conjecture. AMS/IP studies in advanced mathematics 12, AMS, Internatinal Press, 1999. [72] A. W. Goodman, Functions typically-real and meromorphic in the unit circle. Trans. Amer. Math. Soc. 81 (1956), 92–105. [73] A. W. Goodman, Univalent functions. Mariner Publ. Comp., Tampa, FL, 1983. [74] B. Gustafsson, On the convexity of a solution of Liouville’s equation. Duke Math. J. 60 (1990), 303–311. [75] D. J. Hallenbeck and T. H. MacGregor, Linear Problems and Convexity Techniques in Geometric Function Theory. Pitman Press, Boston, 1984. [76] R. Harmelin, Hyperbolic metric, curvature of geodesics and hyperbolic discs in hyperbolic domains. Israel J. Math. 70 (1990), 111–126. [77] R. Harmelin and D. Minda , Quasi-invariant domain constants. Israel J. Math 77 (1992), 115–127. [78] W. K. Hayman, Multivalent functions. Cambridge Univ. Press, Cambridge, 1958.

148

Bibliography

[79] W. K. Hayman and J. M. G. Wu, Level sets of univalent functions. Comment. Math. Helv. 56 (1981), no. 3, 366–403. [80] J. A. Hempel, The Poincar´e metric on the twice punctured plane and the theorems of Landau and Schottky. J. London Math. Soc. (2) 20(1979), 435– 445. [81] P. Henrici, Applied Computational Complex Analysis. Vol.1, Wiley, New York, 1974. [82] D. A. Herron, X. Liu, and D. Minda, Ring domains with separating circles or separating annuli. J. Analyse Math. 53 (1989), 233–252. [83] A. Herzig, Die Winkelderivierte und das Poisson-Stieltjes-Integral. Math. Z. 46 (1940), 129–156. [84] J. A. Hummel, St. Scheinberg, and L. Zalcman, A coeﬃcient problem for bounded nonvanishing functions. J. Anal. Math. 31 (1977), 169–190. [85] Z. J. Jakubowski, On the upper bound of the functional f (n) (z) (n = 2, 3, . . .) in some classes of univalent functions. Comment. Math. Prace Mat. 17 (1973), 65–69. [86] J. A. Jenkins, On a conjecture of Goodman concerning meromorphic univalent functions. Michigan Math. J. 9 (1962), 25-27. [87] J. A. Jenkins, On explicit bounds in Landau’s theorem II. Can. J. Math. 33 (1981), 559–562. [88] V. Jørgensen, On an inequality for the hyperbolic measure and its application in the theory of functions. Math. Scand. 4 (1956), 113–124. [89] S.-A. Kim and C. D. Minda, The hyperbolic and quasihyperbolic metrics in convex regions. J. Anal. 1 (1993), 109–118. [90] W. E. Kirwan and G. Schober, Inverse coeﬃcients for functions of bounded boundary rotation. Journal d’An. Math. 36 (1979), 167–178. [91] R. Klouth and K.-J. Wirths, Two new extremal properties of the Koebefunction. Proc. Amer. Math. Soc. 80 (1980), 594–596. [92] G. Knese, A Schwarz lemma on the polydisk. Proc. Amer. Math. Soc. 135 (2007), 2759–2768. [93] W. Koepf, On close-to-convex functions and linearly accessible domains. Complex Var. 11 (1989), 269–279. [94] W. Koepf and D. Schmersau, Bounded nonvanishing functions and Bateman functions. Compl. Var. 25 (1994), 237–259. [95] L. V. Kovalev, Domains with convex hyperbolic radius. Acta Math. Univ. Comenianae 70 (2001), 207–213.

Bibliography

149

[96] J. G. Krzy˙z, Problem 1, posed in : fourth conference on analytic functions. Ann. Polon. Math. 20 (1967-1968), 314. [97] R. K¨ uhnau, The conformal modul of quadrilaterals and of rings. Handbook in complex analysis: Geometric Function Theory 2 (2005), 99–129, Elsevier, Amsterdam. [98] E. Landau, Einige Bemerkungen u ¨ber schlichte Abbildung. J.ber. Deutsche Math. Verein. 34 (1925/26), 239–243. [99] N.A. Lebedev and I. M. Milin, An inequality. Vestnik Leningrad Univ. 20 (1965), 157–158 (in Russian). [100] O. Lehto, On the distortion of conformal mappings with bounded boundary rotation. Ann. Acad. Sci. Fenn. Ser. AI Math. Phys. 124 (1952), 14p. [101] V. Levin, Aufgabe 163. J.ber. Deutsche Math. Verein. 43 (1933), 133; L¨ osung, ibid. 44 (1934), 80–83 (solutions by W. Fenchel and E. Reissner). [102] J.-L. Li, Estimates for derivatives of holomorphic functions in a hyperbolic domain. J. Math. Anal. Appl. 329 (2007), 581–591. [103] J.-L. Li, Schwarz-Pick inequalities for convex domains. Kodai Math. J. 30 (2007), 252-262. [104] R. J. Libera and E. J. Zlotkiewicz, Early coeﬃcients of the inverse of a regular convex function. Proc. Amer. Math. Soc. 85 (1982), 225–230. [105] E. Lindel¨ of, M´emoire sur certaines in´egalit´es dans la th´eorie des fonctions monog`enes et sur quelques propri´et´es nouvelles de ces fonctions dans le voisinage d’un point singulier essentiel. Acta Societatis Scientiarum Fennicae XXXV (1909), Nr.7 (1908), 35 pp. [106] J. E. Littlewood, On inequalities in the theory of functions. Proc. London Math. Soc. 23 (1925), 481–519. [107] J. E. Littlewood, Lecture on the theory of functions. Oxford University Press, Oxford, 1944. [108] A. E. Livingston, Convex meromorphic mappings. Ann. Pol. Math. 59.3 (1994), 275–291. [109] K. L¨ owner, Untersuchungen u ¨ber die Verzerrung bei konformen Abbildungen des Einheitskreises |z| < 1, die durch Funktionen mit nicht verschwindender Ableitung geliefert werden. Leipziger Berichte 69 (1917), 89–106. [110] K. L¨ owner, Untersuchungen u ¨ber schlichte konforme Abbildungen des Einheitskreises I. Math. Ann. 89 (1923), 103–121. [111] W. Ma and D. Minda, Behavior of domain constants under conformal mappings. Israel J. Math. 91 (1995), 157–171.

150

Bibliography

[112] B. MacCluer, K. Stroethoﬀ, and R. Zhao , Generalized Schwarz-Pick estimates. Proc. Amer. Math. Soc. 131 (2002), 593–599. [113] B. MacCluer, K. Stroethoﬀ, and R. Zhao, Schwarz-Pick estimates. Complex Variables 48 (2003) 711–730. [114] A. J. Macintyre and W. W. Rogosinski, Extremum problems in the theory of analytic functions. Acta Math. 82 (1950), 275–325 . [115] A. Marx, Untersuchungen u ¨ber schlichte Abbildungen. Math. Ann. 107 (1932/33), 40–65. [116] I. M. Milin, On the coeﬃcients of univalent functions. Soviet Math. Doklady 8 (1967), 1255–1258. [117] J. Miller, Convex and starlike meromorphic functions. Proc. Amer. Math. Soc. 80 (1980), 607–613. [118] S. S. Miller and P. T. Mocanu, Diﬀerential subordination, Theory and applications. Marcel Dekker Inc., New York, 2000. [119] C. D. Minda and D. J. Wright, Univalence criteria and the hyperbolic metric in convex regions. Rocky Mtn. J. Math. 12 (1982), 471–479. [120] R. Nevanlinna, Eindeutige Analytische Funktionen. Springer-Verlag, Berlin, 1953. [121] B. G. Osgood, Some properties of f /f and the Poincar´e metric. Indiana Univ. Math. J. 31 (1982), 449–461. ¨ [122] V. Paatero, Uber die konforme Abbildung von Gebieten deren R¨ ander von beschr¨ ankter Drehung sind. Ann. Acad. Sci. Fenn. Ser. A 9 (1931), 77 pp. [123] J. Pfaltzgraﬀ and B. Pinchuk, A variational method for classes of meromorphic functions. J. Analyse Math. 24 (1971), 101–150. ¨ [124] G. Pick, Uber allgemeine Konvergenzs¨ atze der komplexen Funktionentheorie. Mat. Ann. 77:1 (1916), 1–6. ¨ [125] G. Pick, Uber die Beschr¨ ankungen analytischer Funktionen, welche durch vorgeschriebene Funktionswerte bewirkt werden. Mat. Ann. 77 (1916), 7–23. [126] Ch. Pommerenke, Linear-invariante Familien analytischer Funktionen. I. Math. Ann. 155 (1964), 108–154. [127] Ch. Pommerenke, On close-to-convex analytic functions. Trans. Amer. Math. Soc. 114 (1965), 176–186. [128] Ch. Pommerenke, Univalent functions. Vandenhoeck and Ruprecht, G¨ ottingen, 1975. [129] Ch. Pommerenke, Uniformly perfect sets and the Poincar´e metric. Arch. Math. 32 (1979), 192–199.

Bibliography

151

[130] Ch. Pommerenke, On uniformly perfect sets and Fuchsian groups. Analysis 4 (1984), 299–321. [131] Ch. Pommerenke, Boundary behaviour of conformal maps. Springer, New York, 1992. [132] Ch. Pommerenke, Personal Communication. 3. 12. 2002. [133] D. V. Prokhorov, Bounded univalent functions. Handbook in complex analysis: Geometric Function Theory 1 (2005), 207–228, North-Holland, Amsterdam. ¨ [134] J. Radon, Uber die Randwertaufgaben beim logarithmischen Potential. Sitzungsberichte der Oesterreichischen Akademie der Wissenschaften 128 (1920), 1123–1167. [135] M. S. Robertson, On the coeﬃcients of typically-real functions. Bull. Amer. Math. Soc. 41 (1935), 565–572. [136] M. S. Robertson, A remark on odd schlicht functions. Bull. Amer. Math. Soc.42 (1936), 366–370. [137] M. S. Robertson, Quasi-subordination and coeﬃcient conjectures. Bull. Amer. Math. Soc. 76 (1970), 1–9. ¨ [138] W. Rogosinski, Uber positive harmonische Entwicklungen und typisch reelle Potenzreihen. Math. Z. 35 (1932), 93–121. [139] W. Rogosinski, On the coeﬃcients of subordinate functions. Proc. London Math. Soc. 48 (1943), 48–82. [140] O. Roth and K.-J. Wirths, Taylor coeﬃcients of negative powers of schlicht functions. Comp. Methods and Function Theory 1 (2001), 521–533. ¨ [141] St. Ruscheweyh, Uber die Faltung schlichter Funktionen. Math. Z. 128 (1972), 85–92. ¨ [142] St. Ruscheweyh, Uber einige Klassen in Einheitskreis holomorpher Funktionen. Ber. Math.-Stat. Sektion Forschungszentrum Graz 7 (1974), 12p. [143] St. Ruscheweyh, Two remarks on bounded analytic functions. Serdica, Bulg. math. publ. 11 (1985), 200–202. [144] St. Ruscheweyh and T. Sheil-Small, Hadamard products of schlicht functions and the P´ olya-Schoenberg conjecture. Comment. Math. Helv. 48 (1973), 119– 135. [145] N. Samaris, A proof of Krzy˙z’s conjecture for the ﬁfth coeﬃcient. Complex Variables 48 (2003), 753–766. [146] G. Schober, Univalent Functions - Selected Topics. Lecture Notes in Math. 478, Springer, New York, 1975.

152

Bibliography

[147] G. Schober, Coeﬃcient estimates for inverses of schlicht functions. In: Aspects of contemporary complex analysis, Academic Press, New York, 1980. ¨ [148] I. Schur, Uber Potenzreihen, die im Innern des Einheitskreises be-schr¨ ankt sind. J. Reine Angew. Math. 147 (1917), 205–232. [149] H. A. Schwarz, Gesammelte Abhandlungen. Vol.II, Springer, Berlin, 1890. [150] J. H. Shapiro, Composition operators and classical function theory. Springer, New York 1993. [151] T. Sheil-Small, On convex univalent functions. J. London Math. Soc.(2)1 (1969), 483–492. [152] T. Sheil-Small, On the convolution of analytic functions. J. Reine Angew. Math. 258 (1973), 137–152. [153] A. Yu. Solynin, On separation of continua by circles (Russian). Zap. Nauchn. Semin. Leningr. Otd. Mat. Inst. Steklova 168 (1988), 154–157. [154] A. Yu. Solynin and M. Vuorinen, Estimates for the hyperbolic metric of the punctured plane and applications. Isr. J. Math. 124 (2001), 29–60. [155] E. Strohh¨ acker, Beitr¨ age zur Theorie der schlichten Funktionen. Math. Z. 37 (1933), 356–380. [156] T. J. Suﬀridge, Some remarks on convex maps of the unit disc. Duke Math. J. 37 (1970), 775–777. [157] T. J. Suﬀridge, Problems for Non-vanishing H p Functions. In: Lecture Notes in Math.(1990) Nr. 1435, 177–190. [158] W. Szapiel, A new approach to the Krzy˙z conjecture. Ann. Univ. Mariae Curie-Sklodowska Sect. A 48 (1994), 169–192. [159] O. Sz´ asz, Ungleichungen f¨ ur die Koeﬃzienten einer Potenzreihe. Math. Z. 1 (1918), 163–183. [160] O. Sz´ asz, Ungleicheitsbeziehungen f¨ ur die Ableitungen einer Potenzreihe, die eine im Einheitskreis beschr¨ ankte Funktion darstellt. Math. Z. 8 (1920), 303– 309. [161] D. L. Tan, Estimates of coeﬃcients of bounded nonvanishing analytic functions. (Chinese), Chinese Ann. Math., Ser. A 4 (1983), 97–104. English summary in Chinese Ann. Math., Ser. B 4 (1983), 131-132. [162] O. Teichm¨ uller, Untersuchungen u ¨ber konforme und quasikonforme Abbildungen. Deutsche Math. 3 (1938), 621–678. [163] P. Todorov, New explicit formulas for the nth derivative of composite functions. Paciﬁc J. Math. 92 (1981), 217–236.

Bibliography

153

[164] S. Y. Trimble, A coeﬃcient inequality for convex univalent functions. Proc. Amer. Math. Soc. 48 (1975), 266–267. ¨ [165] K.-J. Wirths, Uber holomorphe Funktionen, die einer Wachstumsbeschr¨ ankung unterliegen. Arch. Math. 30 (1978), 606–612. [166] K.-J. Wirths, The Koebe domain for concave univalent functions. Serdica Math. J. 29 (2003), 355-360. [167] K.-J. Wirths, On the residuum of concave univalent functions. Serdica Math. J. 32 (2006), 209–214. [168] S. Yamashita, La d´eriv´ee d’une fonction univalente dans une domaine hyperbolique. C. R. Acad. Sci. Paris Ser I. Math. 314 (1992), 45–48. [169] S. Yamashita, Localization of the coeﬃcient Theorem. Kodai Math. J. 22 (1999), 384–401. [170] S. Yamashita, Higher derivatives of holomorphic functions with positive real part. Hokkaido Math. J. 29 (2000), 23-36.

Index Agard’s formula, 32 Ahlfors’ formula, 39 angle conjecture, 128 angular domains, 24 angular point of order α, 79 angularly accessible domains, 116 asymptotic conjecture, 11 Avkhadiev’s theorem, 36 Baernstein and Schober’s theorem, 22 Bateman functions, 109 Beardon and Pommerenke inequalities, 41 Beardon and Pommerenke’s theorem, 35 Bermant, Hempel theorem, 32 Bieberbach conjecture, 3, 11 Bieberbach theorem, 44 boundary rotation, 23 Brannan, Clunie and Kirwan’s theorem, 23 Carleson and Gamelin’s remark, 37 central existence theorem, 97 Chua conjecture, 69 comparison of densities, 31 concave domains, 90 concave univalent functions, 91 conformal radius, 2 conformal radius as a function its connection with coeﬃcients, 44 its gradient image, 46 its derivatives, 47

conformally invariant families, 117 convex domains, 63 de Branges’ theorem, 14 Denjoy-Wolﬀ point, 33 epicycloids or hypocycloids, 46 factor Cn (Hβ , Hα ), 73 Fej´er’s inequality, 65 Gaussian curvature, 30 generalized Goodman conjecture, 131 generalized punishing factors, 84 genuine annuli, 35 Goluzin type theorem, 10 Goodman conjecture, 12 Grand Iteration Theorem, 32 Hadamard product, 117 Hardy type inequality, 36 hyperbolic distance, 28 hyperbolic metric principle, 31 Jakubowski’s theorem, 113 Jenkins’ theorem, 15 Klouth and Wirths’ theorem, 21 Koebe 1/4-theorem , 33 Koebe transform, 44 Kovalev’s theorem, 45 Krzy˙z conjecture, 127 L¨owner diﬀerential equation, 14 L¨owner theorem, 45

156 L¨owner theorem on inverse coeﬃcients, 19 Landau theorem on holomorphic functions omitting 0 and 1, 33 Landau’s estimate for derivatives, 113 Lebedev-Milin inequality, 13 Liouville equation, 29 Littlewood conjecture, 11 Littlewood’s theorem, 7 Lobachevsky and Bolyai geometry, 27 lower bound conjecture, 58 Marx-Strohh¨ acker theorem, 64 Milin conjecture, 11 modiﬁed punishing factors, 104 modular function, 32 Osgood inequality, 41 Osgood-Jørgensen inequality, 41 Pick, Lindel¨ of theorem, 30 Poincar´e metric, 27 Pommerenke’s characteristic, 40 Ptolemaic system, 46 punishing factor C2 (Ω, Π), 68 quasi-subordinate functions, 9 Radon’s curves, 23 Robertson conjecture, 11 Rogosinski conjecture, 11 Rogosinski’s theorem, 8 Ruscheweyh’s theorems, 52 Schur-Jabotinsky theorem, 64 Schwarz’s lemma, 30 sharp lower bounds, 79 Sheil-Small conjecture, 12 Sheil-Small’s theorem, 17 subordinate functions, 7 Sz´asz inequalities, 50 Szapiel’s problem, 128 Teichm¨ uller’s theorem, 39 theorem on 4n−1 , 55

Index uniformly perfect sets, 37 universal covering map, 28 Wirtinger calculus, 29 Yamashita’s theorem, 55

Our partners will collect data and use cookies for ad personalization and measurement. Learn how we and our ad partner Google, collect and use data. Agree & close