This volume aims to keep a record of the 150-year-old heritage starting from Riemann on zeta-functions, which are ubiquitous in all mathematical sciences, wherever there is a notion of the norm. It provides almost all possible equivalent relations to the zeta-functions without requiring a reader's deep knowledge on their definitions. This can be an ideal reference book for those studying zeta-functions.

© All Rights Reserved

1 tayangan

This volume aims to keep a record of the 150-year-old heritage starting from Riemann on zeta-functions, which are ubiquitous in all mathematical sciences, wherever there is a notion of the norm. It provides almost all possible equivalent relations to the zeta-functions without requiring a reader's deep knowledge on their definitions. This can be an ideal reference book for those studying zeta-functions.

© All Rights Reserved

- 02 - Rational Inequalities
- MATH 242 Syllabus
- CAPE Pure Mathematics 2017 U2 P032
- ElectricalEngineering-2011-12
- The Binomial Expansion - Solutions
- Mathematical Methods of Engineers
- FEM Syllabus
- Double Angle Identities Worksheet With Graphysc Calculator
- 096c05s05ns
- 2. Composition of Functions
- Determinations of Analogues of Gauss Sums and Other Trigonometric Sums
- ALG 10TH
- New Termwork4.1
- Programming Camp Syl Lab Programming
- Syllabus.pdf
- fea model -iset i.docx
- Limit Piece Function
- 140 Average Rate of Change
- HW2 FEM 2012
- STPM Maths T Sem 1 Trial 2014 P1 Port Dickson Answer

Anda di halaman 1dari 316

Series on Number Theory and Its Applications ISSN 1793-3161

V. N. Chubarikov (Moscow State University, Russian Federation)

Christopher Deninger (Universität Münster, Germany)

Chaohua Jia (Chinese Academy of Sciences, PR China)

Jianya Liu (Shangdong University, PR China)

H. Niederreiter (National University of Singapore, Singapore)

Advisory Board:

A. Schinzel (Polish Academy of Sciences, Poland)

M. Waldschmidt (Université Pierre et Marie Curie, France)

Published

Vol. 1 Arithmetic Geometry and Number Theory

edited by Lin Weng & Iku Nakamura

Vol. 2 Number Theory: Sailing on the Sea of Number Theory

edited by S. Kanemitsu & J.-Y. Liu

Vol. 4 Problems and Solutions in Real Analysis

by Masayoshi Hata

Vol. 5 Algebraic Geometry and Its Applications

edited by J. Chaumine, J. Hirschfeld & R. Rolland

Vol. 6 Number Theory: Dreaming in Dreams

edited by T. Aoki, S. Kanemitsu & J.-Y. Liu

Vol. 7 Geometry and Analysis of Automorphic Forms of Several Variables

Proceedings of the International Symposium in Honor of Takayuki Oda

on the Occasion of His 60th Birthday

edited by Yoshinori Hamahata, Takashi Ichikawa, Atsushi Murase &

Takashi Sugano

Vol. 8 Number Theory: Arithmetic in Shangri-La

Proceedings of the 6th China–Japan Seminar

edited by S. Kanemitsu, H.-Z. Li & J.-Y. Liu

Vol. 9 Neurons: A Mathematical Ignition

by Masayoshi Hata

Vol. 10 Contributions to the Theory of Zeta-Functions: The Modular

Relation Supremacy

by S. Kanemitsu & H. Tsukada

World Scientific

Published by

World Scientific Publishing Co. Pte. Ltd.

5 Toh Tuck Link, Singapore 596224

USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601

UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

Kanemitsu, Shigeru.

Contributions to the theory of zeta-functions : the modular relation supremacy / Shigeru

Kanemitsu (Kinki University, Japan) and Haruo Tsukada (Kinki University, Japan).

pages cm. -- (Series on number theory and its applications ; volume 10)

Includes bibliographical references and index.

ISBN 978-9814449618 (hardcover : alk. paper)

1. Functions, Zeta. 2. Functional equations. 3. Mathematics. I. Tsukada, Haruo, 1961–

II. Title.

QA351.K33 2015

515'.56--dc23

2014041588

A catalogue record for this book is available from the British Library.

All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,

electronic or mechanical, including photocopying, recording or any information storage and retrieval

system now known or to be invented, without written permission from the publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance

Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy

is not required from the publisher.

Printed in Singapore

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page v

v

May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

November 19, 2014 14:27 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page vii

Preface

equations. There is an eternal conflict between the viewpoints on whether

one assumes the Euler product or not. Without the Euler product, the class

is wider and with it, the class gives rise to some more delicate information.

Here we take the former viewpoint and do not assume the Euler product,

whence no zero-free region is assumed. We try to exhaust a.a. identities

that are equivalent to the functional equation, thus providing the reader or

browser with a rather handy manual for a quick search. We try to make

this as useful as the famous series of source books on special functions by

Erdélyi [ErdH], [ErdT]. Our collection is in that spirit in thoroughness

and unifiedness as well as in Hamburger [Ham2], Chandrasekharan and

Narasimhan [ChN1], [ChN2], [ChN3] or the more recent [BK, Chapter 8]

in diversity. That is, we shall try to locate the genesis of those identities

which appeared hitherto in mathematical or scientific literature in the last

150 years. The most influential works which led us to this point are [Bo1],

[ChN2], [ChN3], [HeckW], [MKn2], [KoshI]-[KoshIII] etc., which will be

referred to as the story goes on.

Although the contents can be seen from the chapter headings, we give a

brief description here. Chapter 1 contains pre-functional equations, i.e.

those material which are related to the functional equation at least in im-

plicit form. The reader is advised to read the Conventions carefully. Chap-

ter 2 assembles material on special functions, which can be read together

with [Vista], [Vi2]. This can be consulted when need occurs. Chapter 3

collects equivalent relations to the functional equation without the process-

ing gamma factor. Chapter 4 is devoted to the presentation of the Fourier-

Bessel expansion, i.e. the results on perturbed Dirichlet series. §4.3 contains

the concrete version of Theorem 4.3, of the main theorem, Theorem 7.1.

vii

November 19, 2014 14:27 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page viii

sion and a fortiori, incomplete gamma expansion. Chapter 6 is devoted

to the Riesz sums, some of which are also given in other chapters. Since

the summability has been traced for a long time and therefore we cannot

be exhaustive. Instead we incorporate integrals and integrated modular

relations. Chapter 7 gives the main theorem, Theorem 7.1, on modular re-

lations with proofs. This chapter needs to be consulted from time to time,

although in most of the chapters we state the main formulas that are used

in the beginning. Chapter 8 is a database of the Hecke type zeta-functions.

A similar database for the Riemann type zeta-functions can be compiled

and we have given several of them in §6.1. Chapter 9 is about the prod-

uct of zeta-functions, where the results are still partial but we are able to

elucidate the hitherto most far-reaching Wilton’s Riesz sum. Chapter 10

collects miscellaneous future projects.

The first author would like to thank many of his colleagues who have

given him explicit and implicit influence in the last forty years toward the

completion of the present book. Thanks are above all due to S. Akiyama,

K. Alladi, T. Aoki, R. Balasubramanian, B. C. Berndt, K. Chakraborty,

H 2 Chan, V. N. Chubarikov, C. Deninger, S. Egami, K. Györy, M. Hata,

Y. Hironaka, C.-H. Jia, J. Jimenes-Urroz, M. Kaneko, T. Kano, I. Kátai,

A. Katsurada, K. Kawada, Y. Kitaoka, the late Professor J. Knopmacher,

I. Kuchi, G. Lachaud, A. Lauričkas, C. Levesque, J.-Y. Liu, K. Miyake,

C. Matsumoto, L. Murata, M. Nakahara, Y.-N. Nakai, R Okazaki, F. Pap-

palardi, Y. Tanigawa, the late Professor K. Ramachandra, A. Schinzel,

the late Professor K. Shiratani my supervisor, I. Wakabayashi, M. Wald-

schmidt, M. Yoshida and many others whose publications are cited, where

H 2 , A, C are shorthand for Heng Huat, Augustinus, Cozy, respectively.

The first author also would like to express his special thanks to Professor

K. K. Phua, the chairman of the World Scientific for his constant and gener-

ous encouragement which has been a great impetus to continue publishing.

Last but not the least, the authors would like to thank Ms. Lai Fun Kwong

for her patience and encouragement which helped them to complete this

heavy-going task.

The Authors

October, 2014

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page ix

Contents

Preface vii

1. Prelude 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Eternal return or every 50 years . . . . . . . . . . . . . . 12

1.2.1 Poincaré recurrence theorem . . . . . . . . . . . . 12

1.2.2 Lerch-Chowla-Selberg formula . . . . . . . . . . . 12

1.2.3 Knopp-Hasse-Sondow formula . . . . . . . . . . . 15

1.3 The theta-transformation formula . . . . . . . . . . . . . 20

1.4 Summation formulas . . . . . . . . . . . . . . . . . . . . 23

1.4.1 Poisson summation formula . . . . . . . . . . . . 23

1.4.2 Generalization of the Plana summation formula . 25

2.1 Formulas for the gamma function and their use . . . . . 29

2.2 Zeta-functions . . . . . . . . . . . . . . . . . . . . . . . . 33

2.3 Bessel functions . . . . . . . . . . . . . . . . . . . . . . . 35

2.4 Ψ-functions . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.5 Generalized hypergeometric functions . . . . . . . . . . . 39

2.6 Fox H-functions . . . . . . . . . . . . . . . . . . . . . . . 49

2.7 Formulas for Fox and Meijer functions . . . . . . . . . . 51

2.8 Special cases of G-functions . . . . . . . . . . . . . . . . 56

M,0 Ñ ,0

3.1 The H0,M ↔ H0,Ñ

formula . . . . . . . . . . . . . . . . . 61

ix

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page x

1,0 1,0

3.1.1 The H0,1 ↔ H0,1 formula . . . . . . . . . . . . . 66

2,0 2,0

3.1.2 The G0,2 ↔ G0,2 formula . . . . . . . . . . . . . . 69

3.2 Dedekind zeta-function I . . . . . . . . . . . . . . . . . . 70

3.3 Transformation formulas for Lambert series . . . . . . . 72

3.3.1 Lambert series . . . . . . . . . . . . . . . . . . . . 73

3.3.2 Lambert series and short character sums . . . . . 78

3.3.3 Ramanujan’s formula leading to the eta-

transformation formula . . . . . . . . . . . . . . . 80

3.3.4 A brief account of modular forms . . . . . . . . . 84

3.3.5 The Ramanujan-Guinand formula . . . . . . . . . 86

3.3.6 The reciprocity law for Dedekind sums . . . . . . 93

3.4 Koshlyakov’s method [KoshI] . . . . . . . . . . . . . . . . 95

3.4.1 Dedekind zeta-function II . . . . . . . . . . . . . 96

3.5 Koshlyakov’s functions . . . . . . . . . . . . . . . . . . . 97

3.5.1 Koshlyakov’s X-functions . . . . . . . . . . . . . 97

3.5.2 Koshlyakov’s Y -function . . . . . . . . . . . . . . 101

1,1 2,0

4. Fourier-Bessel Expansion H1,1 ↔ H0,2 107

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 107

4.2 Stark’s method . . . . . . . . . . . . . . . . . . . . . . . 108

4.2.1 Murty-Sinha theorem . . . . . . . . . . . . . . . . 108

4.2.2 Stark’s method . . . . . . . . . . . . . . . . . . . 111

4.3 The main formula for modular relations . . . . . . . . . . 115

4.3.1 Specification of Theorem 4.4 . . . . . . . . . . . . 119

4.4 Dedekind zeta-function III . . . . . . . . . . . . . . . . . 122

4.5 Elucidation of Koshlyakov’s result in the real quadratic

case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

4.6 Koshlyakov’s K-series . . . . . . . . . . . . . . . . . . . . 128

4.7 The Fourier-Bessel expansion G1,1 2,0

1,1 ↔ G0,2 . . . . . . . . 133

4.8 Bochner-Chandrasekharan and Narasimhan formula . . . 142

5.1 Ewald expansion for zeta-functions with a single

gamma factor . . . . . . . . . . . . . . . . . . . . . . . . 145

5.1.1 Confluent hypergeometric series imply

incomplete gamma series, Ewald expansions . . . 148

5.1.2 Bochner-Chandrasekharan formula as

2,0 1,1

H1,2 ↔ H1,2 . . . . . . . . . . . . . . . . . . . . . 150

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page xi

Contents xi

5.2.1 Landau’s exposition . . . . . . . . . . . . . . . . . 156

5.2.2 Screened Coulomb potential . . . . . . . . . . . . 158

6.1 Various modular relations . . . . . . . . . . . . . . . . . 169

6.1.1 Riesz sums . . . . . . . . . . . . . . . . . . . . . . 169

6.1.2 Improper modular relations as Riesz sums . . . . 172

1,1 2,0

6.1.3 The H2,1 ↔ H1,3 formula . . . . . . . . . . . . . 173

1,1 2,0

6.1.4 The H2,2 ↔ H1,3 formula . . . . . . . . . . . . . 174

6.1.5 Katsurada’s formula combined . . . . . . . . . . . 175

6.1.6 Linearized product of two zeta-functions . . . . . 178

6.2 Modular relations in integral form . . . . . . . . . . . . . 185

6.2.1 Integration in the parameter . . . . . . . . . . . . 185

6.2.2 Generalization of Ramanujan’s integral formula . 188

6.3 Integrated modular relations . . . . . . . . . . . . . . . . 191

6.3.1 The Hardy-Littlewood sum . . . . . . . . . . . . 192

1,1 2,0

6.3.2 The H2,1 ↔ H0,3 formula . . . . . . . . . . . . . 194

6.3.3 Arithmetical Fourier series . . . . . . . . . . . . . 196

6.3.4 Riemann’s legacy . . . . . . . . . . . . . . . . . . 206

7.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . 211

7.2 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . 213

7.3 Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

7.4 The Main Formula (basic version) . . . . . . . . . . . . . 220

8.1 Statement of the formula . . . . . . . . . . . . . . . . . . 225

8.1.1 The bilateral form . . . . . . . . . . . . . . . . . . 227

8.1.2 The Bochner modular relation: G1,0 1,0

0,1 ↔ G0,1 . . . 229

8.2 The Riesz sums or the first J-Bessel expansion:

G1,0 1,0

1,1 ↔ G0,2 . . . . . . . . . . . . . . . . . . . . . . . . . 229

8.3 The partial sum formula: G2,0 1,1

2,2 ↔ G1,3 . . . . . . . . . . 230

8.4 The Fourier-Bessel expansion: G1,1 2,0

1,1 ↔ G0,2 . . . . . . . . 231

8.5 The Ewald expansion: G2,0 1,1

1,2 ↔ G1,2 . . . . . . . . . . . . 231

2,0 1,1

8.6 The Bochner-Chandrasekharan formula: H1,2 ↔ H1,2 . . 232

8.7 The G2,1 2,1

1,2 ↔ G1,2 formula . . . . . . . . . . . . . . . . . . 232

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page xii

2,2 ↔ G1,3 . . . . . . . 232

2,1 2,1

8.9 The H2,2 ↔ H1,3 formula . . . . . . . . . . . . . . . . . 234

8.10 The second K-Bessel expansion: G3,0 1,2

1,3 ↔ G2,2 . . . . . . 235

8.11 The G3,1 2,2

1,3 ↔ G2,2 formula . . . . . . . . . . . . . . . . . . 236

8.12 The G2,3 ↔ G2,1

2,1

2,3 formula . . . . . . . . . . . . . . . . . . 236

8.13 The G2,3 ↔ G1,2

3,0

2,3 formula . . . . . . . . . . . . . . . . . . 237

8.14 The G2,3 ↔ G2,2

3,1

2,3 formula . . . . . . . . . . . . . . . . . . 237

8.15 The G3,2

2,3 ↔ G 3,2

2,3 formula . . . . . . . . . . . . . . . . . . 238

8.16 The G1,p+1 ↔ G1,p

p+1,0

p,2 formula . . . . . . . . . . . . . . . . 239

9.1 The product of zeta-functions . . . . . . . . . . . . . . . 241

9.1.1 Statement of the Main Formula . . . . . . . . . . 241

9.1.2 Wilton’s Riesz sum: G2,2 4,0

4,4 ↔ G2,6 . . . . . . . . . 243

9.2 Powers of zeta-functions . . . . . . . . . . . . . . . . . . 253

9.2.1 Statement of the Main Formula . . . . . . . . . . 253

9.2.2 The GN,0 N,0

N,N ↔ G0,2N formula . . . . . . . . . . . . 257

9.2.3 The Gq+1,q+1 ↔ GN,q−N

q+1,0 +1

q−N +1,q+N +1 formula . . . . . 263

10.1 Future projects . . . . . . . . . . . . . . . . . . . . . . . 267

10.1.1 Rankin-Selberg convolution . . . . . . . . . . . . 267

10.1.2 Maass forms . . . . . . . . . . . . . . . . . . . . . 269

10.1.3 G-functions of two variables . . . . . . . . . . . . 270

10.1.4 Plausible general form . . . . . . . . . . . . . . . 273

10.2 Quellenangaben . . . . . . . . . . . . . . . . . . . . . . . 275

10.2.1 Berndt-Knopp book and Berndt’s series of papers 275

10.2.2 Corrections to “Number Theory and its

Applications” . . . . . . . . . . . . . . . . . . . . 276

Bibliography 279

Index 301

November 19, 2014 14:27 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 1

Chapter 1

Prelude

1.1 Introduction

The year 1859 is a remarkable one in the history of science (or homo sapiens)

because of two important works published in that year. One is Darwin’s

“The origin of species” which gave a tremendous impact on the paradigm

of human beings in the succeeding several decades and there are still new

discoveries occurring surrounding the evolution theory, cf. [ChR]. In math-

ematics in general and in number theory in particular, this is the year when

Riemann published his epoch-making paper [Rie2], which is his unique pa-

per on the distribution of primes, with a statement of the most important

Riemann hypothesis. We are still struggling through inheriting these

magnificent legacies. Since Riemann introduced the Riemann zeta-function

in that paper, there is an enormous amount of papers and books related to

it and generalizations thereof. It is preferable that there be some accessible

surveys which make clear the history and relations of all these zeta-functions

from a certain higher standpoint: “vom höheren Standpunkte aus”—from

part of the title of Klein’s book, or with some ever-lasting mnemonics, so

that this tradition will be inherited by the coming generations.

Thus, in this trilogy, we will be the bards with a bird’s eye of view.

We take the zeta-symmetry—functional equation—as our guiding prin-

ciple and will try to make up a chart of the sea of zeta-functions for the

coming young navigators to continue the tradition. To take the distribution

1

November 19, 2014 14:27 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 2

associated to a certain set of prime elements) into view, we need to in-

troduce the notion of zero-free region, which is furnished by assuming the

Euler product. However, we do not assume the Euler product here because

it restricts the class of zeta-functions to the one connected with number

theory. What we would like to present is a chart to which any sincere re-

searcher in research areas where the appearance of the zeta-symmetry can

make a recourse. A chart which shows what you need at a browse, without

too much pain.

In our previous “Vista” books [Vista] and [Vi2], we tried to give a

fundamental source of special functions and their applications, which, to

our great joy, received rather favorable reviews and reactions. In the present

case, we would like to present a source book which will be an enlightening

and enjoyable handbook of zeta-functions and a.a. equivalent formulas to

their symmetry. As you can see on reading through the present book,

the history of zeta-functions is, so to say, that of equivalent formulas to

functional equations. Here the “every 50 years” phenomena appear to the

eﬀect that there appears a repetition of what was obtained 50 years ago,

which will be illustrated later. It is like a grandfather telling the theory of

zeta-functions to his grandson. This phenomenon being very common in

all fields of human activities, mathematics in general, number theory or the

theory of zeta-functions in particular, is no exception. The diﬃculty of such

meta-physical subjects is that they are mostly too high-brow and entangled

for the public to feel like following. Thus here is a need for bards who can

write sagas which are accessible to the public. In the present book, we are

not so ambitious to do so and will try to make this a source book useful to

researchers, not necessarily mathematicians. In one of the reviews on the

Vista books, there is a passage indicating that the books are available for

non-mathematicians, too. This is what we would like to achieve, providing

something good to scientists and engineers. We are anyway comrades,

tovarishch, making a mad pursuit.

We are the weavers who weave a tapestry which can be used as a magic

carpet. You sit on it and say the magic words “a là peanut butter sand-

wiches”. Then you would fly provided that you do not think about a

monkey with three golden hairs when you say the magic words. To make

full use of the book, you need to remember some notations on the H- and

G-functions and then the magic will work and you can find what you ex-

pect. A tapestry is to have many patterns. In this book they are provided

by a number of worked-out examples.

November 19, 2014 14:27 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 3

Prelude 3

In this regard, we can claim that the reader will benefit, even by brows-

ing through these examples.

This book is just a starter, the first volume of the planned trilogy—

contribution to the theory of zeta-functions. Although we succeeded in

making a thorough treatment of the cases of zeta-functions satisfying the

functional equation with two gamma factors, we were able to make the

slightest touch of the case of more gamma factors, which we have to post-

pone in later volumes. The planned volumes are

and

In Volume II, we are going to treat what should come after the func-

tional equation—post symmetry, i.e. approximate functional equations and

mean values, asymptotic formulas for the coeﬃcient sums, Hamburger type

theorems, i.e. characterization of the zeta-function through their functional

equations, (Voronoĭ) summation formulas, the Riemann-Siegel formula, ex-

plicit formulas–explicit expressions of relevant functions in terms of zeros

of the zeta-functions, a generalization of the modular relation in the case

where the generating functions have infinitely many poles, etc. The last a

few items are intimately related to the Euler product.

It will turn out that our theory of modular relations without Euler prod-

uct covers a wider class of zeta-functions (e.g. the Hurwitz zeta-functions

or Epstein zeta functions) but lacks some delicate details that the class of

zeta-functions with the Euler product possesses, notably, those which the

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 4

Kaczorowski and Perelli [KP7] and references therein referred to.

Thus in the coming Volume III, we will include the influence of the Euler

product in due course and might be able to step into the forbidden critical

strip.

We started thinking of writing this book some 10 years ago. The first

author used the G-functions in his earlier apprenticeship paper [Ka] but

never thought such fancy goods could be of any use in number theoretic

environment in view of the way they are used in statistics and related

fields, e.g. [MaSa]. More recently there appeared the application of the G-

functions to transcendence theory by Nesterenko, Zudilin [Zu] et al, which

came to our attention after we started writing the draft. They seem to

be centered around the case of unit argument. Then after a time lapse of

30 years, the second author made a colossal contribution to the theory of

zeta-functions by making the eﬀective use of the H-functions. Meanwhile

there were many sprouts of this flourishing and we were motivated by a

vast amount of literature, and we can mention just part of them. Certainly

Lavrik [Lav5] gave a decisive influence because he is the first who mentioned

the extraneous parameter, which is the variable τ in the upper half-plane

(3.100).

Hecker-Bochner correspondence the correspondence between the au-

tomorphic forms, dwellers in the upper half-plane H (3.100) and the zeta-

functions, dwellers in the right half-plane Re z > 0 under the rotation

τ = iz : upper half-plane ↔ right half-plane. (1.1)

tion holding true and the positive real axis usually prevails. Its multiplica-

tive group structure is in shorthand referred to as the reciprocal relation

1

x× =1 x > 0, (1.2)

x

which appears in disguised form subsequently. For example, in (3.82) or in

§6.2.1.

In this series of books, we mean by the modular relation the relation

for the key-function X(z, s) defined by (3.2) under the zeta-symmetry s ↔

r − s and the automorphy z ↔ z1 and it is the one for z s X(z, s) that gives

the most information in two aspects.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 5

Prelude 5

other zeta-functions satisfying the Riemann functional equation

(s) ( )

− 2s r−s

π − 2 ψ(r − s)

1−s

Γ π φ(s) = Γ (1.3)

2 2

(r )

Γ (s) π −s φ(2s) = Γ − s π − 2 +s ψ(r − 2s).

1

(1.4)

2

Cf. (8.1).

The standpoint of viewing them as Hecke type has been taken by many

authors including Bellman [Bel4, (10)], Berndt [BeI, p.356], Berndt [BeII,

p.370], [BK, p.121], [KTY7], etc. We note an essential diﬀerence between

the fundamental sequence in (1.3) and (1.4), i.e. in (1.4), the sequence

{λk } (for notation cf. e.g. (1.12)) is squared: {λ2k }, i.e., we consider only

a subsequence. It would be that in doing so, we may lose some essential

ingredients intrinsic to the original sequence. E.g. in the case of

the Riemann zeta-function, the original sequence is {n ∈ N}, while as a

Hecke type zeta, it is {n2 ∈ N}. This treatment corresponds to viewing the

Riemann zeta as an Epstein zeta of the unary quadratic form. However, in

most of the formulas that we will obtain, those for the Riemann zeta are

those for the Hecke type with the replacement of the variable s by 2s (and

replacement of the fundamental sequences by their squares is not very much

perceivable). The following Exercises 1.1 and 1.2 below help to exhibit this

diﬀerence.

We introduce the Riemann zeta-function as a special case of the

Dedekind zeta-function in §3.4.1. It is defined as the Dirichlet series which

is absolutely convergent in the half-plane σ := Re s > 1 which is then

continued meromorphically over the whole s-plane with a simple pole at

s = 1 with reside 1. It satisfies the celebrated Riemann functional

equation (1.3):

(s) ( )

− 2s − 1−s 1−s

π Γ ζ(s) = π 2 Γ ζ(1 − s). (1.5)

2 2

normalized Gauss sum:

∑

q

a

τ (χ) = χ(a)e2πi q . (1.6)

a=1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 6

∑∞

χ(n)

L(s, χ) = , σ > 1. (1.7)

n=1

ns

form for χ a primitive character:

( πi )

−s

L(1 − s, χ) = q s−1 (2π) Γ(s) e− 2 s + χ(−1)e 2 s τ (χ)L(s, χ̄), (1.8)

πi

which is usually stated in the form (see e.g. [Dav, Chapter 9] or [NTA,

Chapter 5])

√

ia q

ξ(1 − s, χ) = ξ(s, χ̄), (1.9)

τ (χ)

where

( )

− s+a s+a

ξ(s, χ) = π 2 Γ L(s, χ), (1.10)

2

and where a is 0 or 1 according as χ(−1) = 1 or χ(−1) = −1 (χ(−1) =

a

(−1) ):

{

1 − χ(−1) 0 χ even

a = a(χ) = = . (1.11)

2 1 χ odd

Conventions

We make some conventions which will be applied throughout in what

follows.

(i) Whenever we refer to a functional equation, we concentrate on the

gamma factors. In this respect we incorporate the factors of either of the

sequences, in the basic sequences or the coeﬃcients, where in the very def-

inition of a Dirichlet series

∞

∑ αk

φ(s) = (1.12)

λk s

k=1

sequence of positive reals {λk } the basic sequence and the sequence {αk }

of complex numbers the coeﬃcients. With another Dirichlet series

∞

∑ βk

ψ(s) = , (1.13)

µsk

k=1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 7

Prelude 7

various forms of the functional equation and we treat the two Dirichlet

series φ and ψ as given once and for all in the forms (1.12) and

(1.13), referring to “given two Dirichlet series”.

For example, in (1.5), we write

∞

∑

− 2s 1

φ(s) = π ζ(s) = √ s (1.14)

k=1

( πk)

to

(s) ( )

1−s

Γ φ(s) = Γ φ(1 − s). (1.15)

2 2

However, it is more far-reaching to consider

(s) ( )

r−s

Γ φ(s) = Γ ψ(r − s). (1.16)

2 2

(1.3) and (1.4) are unified in the single gamma factor case (5.3) or in

a slightly more general (4.22). We refer to (1.16) as the Riemann type

functional equation.

In the cases of (1.9) and (1.10), we put

∞

∑

− s+a αk

αk = π − 2 χ(k)

a

φ(s) = π 2 L(s, χ) = √ s, (1.17)

k=1

( πk)

and

∞

∑ √

βk ia q − a

ψ(s) = π −

s+a

2 L(s, χ) = √ s, βk = π 2 χ̄(k). (1.18)

( πk) τ (χ)

k=1

( ) ( )

s+a 1−s+a

Γ φ(s) = Γ ψ(1 − s). (1.19)

2 2

In this way, we seal up the additional factors in the Dirchlet series them-

selves and we will not give a special mention to them but the corresponding

results are to be readily read oﬀ by specialization of the sequences.

(ii) By abuse of language, by the functional equation for the Dedekind zeta-

function (to be developed in §3.2, §3.4.1, §4.4) we mean several objects that

have the same form of the functional equation. That is, when we speak of

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 8

the Epstein zeta-function of a positive definite quadratic form or the one

associated to a modular form (which is known as the Hecke correspondence)

and when we speak of the Dedekind zeta-function of a real quadratic field

we also refer to the Epstein zeta-function of an indefinite quadratic form,

the Maass wave zeta-function (§7.1.2) or the one associated to the Dirichlet

divisor problem. Here the divisor problem refers to an asymptotic formula

for the summatory function of the divisor function and the divisor function

is the special case σ0 (n) of the sum-of-divisors function in (3.86).

(iii) When we state the modular relations, we often say that they are equiv-

alent to the functional equation to mean that they are all consequences of

the functional equation in question and some of them imply the functional

equation.

(iv) We usually refer the sum of the residues as the residual function and

denote it by P(x), where P refers to capital ρ rather than P as in the

tradition of G. H. Hardy [HarAO], where he means by P(x) the error term

of the Gauss circle problem and since the number of integers in the circle

√

of radius x is denoted by R(x), his P(x) is to mean the Greek capital.

Following Bochner [Bo1], we define the residual function

∫

1

P(x) = χ(s)x−s ds,

2πi C

function to be defined later. Cf. also [Ahm]. We often express the residual

function explicitly as the sum of residues but sometimes we just denote it

as P(x).

familiar n or m so as to avoid the confusion of the indexes in the H-

m,n

functions Hp,q . The prime on the summation sign means that the last

term is to be halved if the summation variable equals to the bound. This

appears especially when we apply the Riesz sum and the partial sum as its

special case. Cf. Chapter 6.

σ almost always indicates the real part of the complex variable s save

for Koshlyakov’s σ-function.

(vi) Although we restate the following in §2.6, we fix the notation and stick

to it once and for all.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 9

Prelude 9

( )

{(1 − aj , Aj )}nj=1 ; {(aj , Aj )}pj=n+1

∆= q (1.20)

{(bj , Bj )}m

j=1 ;{(1 − bj , Bj )}j=m+1

Γ(s | ∆) (1.21)

( )

{(1 − aj , Aj )}nj=1 ; {(aj , Aj )}pj=n+1

= Γ s q

{(bj , Bj )}m

j=1 ;{(1 − bj , Bj )}j=m+1

∏

m ∏n

Γ(bj + Bj s) Γ(aj − Aj s)

j=1 j=1

= (Aj , Bj > 0)

∏

p ∏

q

Γ(aj + Aj s) Γ(bj − Bj s)

j=n+1 j=m+1

H(z | ∆)

( )

(1 − a1 , A1 ), . . . , (1 − an , An ), (an+1 , An+1 ), . . . , (ap , Ap )

= m,n

Hp,q

z

(b1 , B1 ), . . . , (bm , Bm ), (1 − bm+1 , Bm+1 ), . . . , (1 − bq , Bq )

∫

1

= Γ(s | ∆) z −s ds

2πi L

∏

m ∏

n

∫ Γ(bj + Bj s) Γ(aj − Aj s)

1 j=1 j=1

= z −s ds, (1.22)

2πi L ∏ p ∏q

Γ(aj + Aj s) Γ(bj − Bj s)

j=n+1 m+1

We note

Mnemonics

+ −

m n

p q

Here m is an abbreviation for the B-group (bj , Bj )’s with positive Bj ’s.

Similarly, n is an abbreviation for A-group (aj , Aj )’s with negative Aj ’s.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 10

Exercise 1.1 Prove that (1.5) and the following are equivalent.

1 ( −1 )

∞

∑ ∞

∑ πn2

e−πn = x− 2 e−

2 1

x x + x 2 − 1 , Re(x) > 0, (1.23)

n=1 n=1

2

(s) ∑∞

1 (s )

π− 2 Γ ζ(s) = π − 2

s s

2

Γ , πn w (1.24)

2 n=1

ns 2

∑∞ ( ) s−1 s

1 1 − s πn2 w 2 w2

+ π−

1−s

2 Γ , + − , Re(w) > 0,

n=1

n1−s 2 w s−1 s

∞

∑ ∑∞

1

π −s Γ(s) 2 −s

1 1

2 2 s

= 2α ns− 2 Ks− 12 (2παn) (1.25)

n=1

(n + α ) n=1

( )

1 −s+ 1 1−2s 1 1 −s −2s

+ π 2α Γ s− − π α Γ(s),

2 2 2

where Γ(s.z) and Kν (z) indicate the incomplete gamma function (2.12) and

the modified K-Bessel function (2.31), resp.

equivalence of (1.5), (1.23), (1.24), while (1.25) is not mentioned. Note

that (1.23) is the celebrated theta transformation formula (1.72). In-

deed, Riemann [Rie2] deduced the functional equation (1.5) from the theta-

transformation formula (1.23). Cf. §2.2 for more details.

The following Exercise 1.2 is [KTY7, Example 1.1, p.164].

Exercise 1.2 Prove that (1.23) and the following are equivalent

( )

1

π −s Γ(s)ζ(2s) = π −( 2 −s) Γ

1

− s ζ(1 − 2s), (1.26)

2

∑∞

−s 1 −s

π Γ(s)ζ(2s) = π 2s

Γ(s, πn2 w) (1.27)

n=1

n

∑∞ ( ) 1

s− 12 1 1 πn2 ws− 2 ws

+π Γ − s, + − , Re(w) > 0,

n=1

n1−2s 2 w 2s − 1 2s

∞

∑ ∑∞

−s 1

2 −s

1 1

π Γ(s) 2 + α2 )s

= 2α |n|s− 2 Ks− 12 (2πα|n|) , (1.28)

n=−∞

(n n=−∞

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 11

Prelude 11

understood to mean

( )

s− 12 1 1

−s 1

lim u Ks− 12 (2παu) = (πα) 2

Γ s− , (1.29)

u→0 2 2

which will be restated as (2.35) in §2.3 below.

Solution The equivalence of (1.5) and (1.23) ((1.26) and (1.23)) is the

most celebrated modular relation (cf. §2.2 on zeta-functions).

The equivalence of (1.5) and (1.24) ((1.26) and (1.27)) is the Ewald

expansion, which will be developed in Chapter 5.

Similarly, the equivalence of (1.5) and (1.25) ((1.26) and (1.28)) is the

Fourier-Bessel expansion, which will be developed in Chapter 4. We only

indicate the deduction of (1.25) and (1.28) from a general modular relation.

In accordance with λn = n2 , we write α2 for α. Then φ(s, α2 ) in (4.84)

∑∞ 1

becomes n=1 (n2 +α 2 )s . Since

∫ ∞ (u) ( )

1 1 1

A−s e−α u us−1 P − α−2s Γ(s),

2

du = π 1/2−s α1−2s Γ s −

0 π 2 2 2

we obtain (1.25) from (4.82).

Multiplying (1.25) by 2, moving the last term on the right of (1.25)

to the left and adopting the interpretation (1.29) completes the proof of

(1.28).

Remark 1.1 (1.28) or (1.25) are known as Watson’s formula [Wa].

Cf. (4.32) below. It is stated as [BK, Theorem 8.4, p.121] and is a starting

point for proving [BK, Theorem 4.8] below.

The special case s = 1 of (1.25) gives the partial fraction expansion

for the hyperbolic cotangent function (due to Euler) [Vista, p.100]:

∑ 1 π( 2 ) π

= + 1 = coth πw. (1.30)

2

k +w 2 w e 2πw −1 w

k∈Z

the Hamburger theorem (Remark 3.1, (ii)).

As will be seen steadily, one of the main ingredients that make a dis-

tinction between these two types of zeta-functions is the reduction formula

(2.93), which necessitates the sequence to be squared (in general raised to

the power C1 if we have C for 12 ).

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 12

The eternal return used to be once a dominating pessimistic view of the

18th century physics and it is remotely but somewhat relevant to what we

are going to state, so that we find it appropriate to state and give a proof

of the Poincaré recurrence theorem, following Tipler [Tip, pp.418-419].

bounded region in the phase space and let f be a continuous, volume pre-

serving 1 : 1 map from V to V : f : V → f (V ) = V . Then for any point

and in any of its neighborhood U , there exists a point x ∈ V which returns

to U , i.e. there exists an n ∈ N such that f n x ∈ U , where we write f x to

mean f (x).

U, f U, f 2 U, . . . .

the same finite volume. If they never intersect, then, all the terms being

contained in V , the volume of V would be infinite, a contradiction. Hence

there must be i, j ∈ N, i > j such that f i U ∩ f j U ̸= ∅, or f n U ∩ U ̸= ∅, on

setting n = i − j. Hence there exists a point x such that f n x also belongs

to U , whence the theorem.

This phenomenon of every 50 years revival of fashion, intrinsic to the human

society, has been noticed first by Kurokawa [Kur] in mathematics as far as

we know. There must be similar observations made earlier many times,

every 50 years or so. More often 60 years period would appear since this

is the span of two generations. You and your grandson may not share the

same level of intelligence because of the inevitable generation gap. Thus it

would look natural that the knowledge obtained two generations ago may

be forgotten and not inherited by the oﬀsprings.

Prior to [Kur] by two years, F. Sato [Sa] (there are many Sato’s and

we are very specific to this Professor Fumihiro Sato), in expounding the

theory of pre-homogeneous vector spaces, mentioned Eisenstein’s proof of

the functional equation, which has been referred to on [Vista, pp.70-72].

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 13

Prelude 13

paper in the Chowla-Selberg exposition on [Vi2, pp.122-123], and we shall

now provide what could have appeared there.

Matyas Lerch [Ler1], starting from the work of Malmstén [Ma], estab-

lished Lerch’s formula

the gamma function and the automorphic function (elliptic function), which

has been known as the Chowla-Selberg formula. Cf. [WeE, Chapter X].

The most enlightening explanations on historic as well as mathematical

part can be found in Landau [LanCS] and Deninger [Den], in the latter

which the Chowla-Selberg formula for a real quadratic field is also deduced

using the R-function.

Theorem 1.2 (Lerch-Chowla-Selberg formula)

|∆|−1 ( ) ( ) [√ ]

w ∑ |∆| n ∑ 2|∆|π

log Γ = log η(ω1 )η(ω2 ) , (1.32)

2 n=1 n |∆| c

(a,b,c)

where

√ ∆ < 0 is the discriminant of the imaginary quadratic field Ω =

Q( √∆), w indicates the

√ order of the group of roots of unity in Ω, ω1 =

−b+i |∆| −b−i |∆| 2

2c , ω2 = 2c (whence η(ω1 )η(ω2 ) = |η(ω1 )| ) and on the

right (a, b, c) runs through the complete set of representatives of classes

of quadratic forms of discriminant −∆.

For the notation, cf. Section 3.4.

In the times of Kronecker, after the tradition of Gauss, the theory of

binary quadratic forms was the main issue of research than the theory of

quadratic fields, both being equivalent. For classification of this cf. Daven-

port( [Dav) ] which in turn depends on Landau [LanZT].

|∆|

n = χΩ (n) on the left of (1.32) indicates the Kronecker-Jacobi

symbol associated to Ω and is one of primitive odd characters modulo

|∆|. Berger [Berg] and Lerch [Ler1] independently expressed the values

of L′ (1, χΩ ) in terms of the gamma values. Both authors appealed to Kum-

mer’s Fourier series for the log Γ to deduce the results. Lerch [Ler1] then

used the Kronecker limit formula to deduce Theorem 1.2.

Corollary 1.1 The Lerch-Chowla-Selberg formula (1.32) is a conse-

quence of the functional equations for the Dedekind zeta-functions of the

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 14

Proof. The Kronecker limit formula in Corollary 4.3 is a consequence

of the Chowla-Selberg type formula (Corollary 4.2), which is equivalent to

the functional equation for the Dedekind zeta-functions of the imaginary

quadratic field. (4.93) must be divided by the number w of roots of unity

and summed over the h classes. Hence the right-hand side of (1.32) is a

consequence of the functional equation.

On the other hand, the left-hand side follows from the decomposition

∑∞

log n sin 2πnx

(1.34)

n=1

n

which is simply log Γ(x) − log Γ(1 − x) and thus the Kummer Fourier series

for the loggamma function (1.35) given below gives the result. Hence the

left-hand side is a consequence of the functional equation for the Dedekind

zeta-function for the rational field.

Proposition 1.1 ([Vi2, Chapter 7]) For 0 < x < 1 we have

∞ ( )

Γ(x) ∑ 1 γ + log(2πn)

log √ = cos(2πnx) + sin(2πnx) , (1.35)

2π n=1

2n πn

Riemann zeta-function.

The decomposition formula (1.33) holds true for an abelian field over

the rationals, i.e. for the mth cyclotomic field Q(ζ), where ζ = ζm is a

primitive mth root of 1.

Gut [Gut] was the first who expressed the constant term of the Dedekind

zeta-function for the Q(ζ) in terms of the R-function.

Definition 1.1 Let α > 0. Then the Deninger R = Rα -function is

defined as the principal solution to the diﬀerence equation

′′ 2

ζ (0, x) = ∂∂2 s ζ(s, x)|s=0 , where ζ(s, x) indicates the Hurwitz zeta-function

defined by (2.22).

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 15

Prelude 15

higher-derivatives of the L-function at s = 1 in [CKK].

The contents of this subsection is mostly due to [XHW]. There are many

known integral expressions for the Riemann zeta-function which provides us

with a preliminary stage for analytic continuation. Notably, [Vi2, Lemma

2.3, p. 43] ([SC, (30), p. 100]) reads

∫ ∞ ∫ ∞

1 dt e−t dt

Γ(s)ζ(s) = ts t = ts (1.37)

0 e −1 t 0 1 − e−t t

which is true for σ > 1.

We may view (1.37) as the gamma transform of the theta function ϑ1 (t)

in the following manner.

Let κ > 0 and let ϑκ (t) denote the theta series of order κ > 0:

∞

∑ κ

ϑκ (t) = e−n t

(1.38)

n=−∞

∞

∑ κ

ϑκ (t) = e−n t

(1.39)

n=0

1

ϑ1 (t) = (1.40)

et − 1

and ϑ2 (t) is the elliptic theta-function, which satisfies the theta-

transformation formula (1.23), (1.72) (cf. Section 1.3).

A possible application of higher-order theta-functions has been already

noted much earlier. In [HarDP, p.7] he mentioned possible applications of

the higher-order Lambert series

∑∞

d(n) −snα

Fα,β (s) = e , (1.41)

n=1

nβ

where d(n) indicates the number of divisors of n, called the divisor func-

tion, which is the special case σ0 (n) of the sum-of-divisors function in

(3.86). He also refers to the original papers of Mellin [Mel1], [Mel2] as the

main source for a large amount of resources.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 16

The gamma transform [Vi2, p.25] reads for λ > 0 and σ > 0

∫ ∞

−s dt

Γ(s)λ = ts e−λt (1.42)

0 t

whence in particular,

∫ ∞

dt

Γ(s)ζ(κs) = ts ϑκ (t) (1.43)

0 t

for κσ > 1.

In this context, ϑ2 (t) along with its transformation formula, has been

used extensively in literature as a source of the functional equation for the

Riemann zeta-function.

Goss [Gs] uses ϑ1 (t) to provide an analytic continuation of the Riemann

zeta-function whose method consists in removing the singularity at t = 0

of the integrand et1−1 or writing z = e−t , that of the fraction 1−z

z

at z = 1.

Hence Goss’ method can be generalized in the following setting. With a

function analytic at z = 1 such that f (1) = 1, we form the diﬀerence

z f (z) f (z) − z

− = , (1.44)

1−z 1−z z−1

whence it follows that the diﬀerence is analytic at z = 1.

mz m

Goss uses f (z) = 1+z+···+z m−1 , i.e. the diﬀerence in (1.44) is ϕm (z) =

mz m

1−z − 1−z m .

z

Our aim is to find a class of such functions f which give rise to analytic

continuation of the relevant zeta-function.

and bounded on the boundary and that f (0) = 0 and f (1) = 1. Then

∫ ∞

s f (z) dt

F (s) : = Γ(s)ζ(s) − t (1.45)

0 1 − z −t

t

z=e

∫ ∞

s f (z) − z dt

= t

0 z − 1 z=e−t t

expressed as a Taylor series at z = 1, which is absolutely and uniformly

convergent, so that all operations, conducted below, of changing order and

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 17

Prelude 17

obtain

( ∞

)

f (z) − z 1 ∑ f (n) (1)

′

= 1 + f (1)(z − 1) + (z − 1) − z .

n

z−1 z−1 n=2

n!

∑∞ ( )

f (n+1) (1) ∑

n

f (z) − z ′ n−k n

= −1 + f (1) + (−1) e−kt .

z−1 n=1

(n + 1)! k

k=0

∑∞

f (n+1) (1)

LHS = −1 + f ′ (1) + (−1)n

n=1

(n + 1)!

∑∞ ( )

(1) ∑

(n+1) n

f n −kt

+ (−1)n−k e

n=1

(n + 1)! k

k=1

∑∞ ( )

f (n+1) (1) ∑

n

f (z) − z n −kt

= (−1)n−k e . (1.46)

z−1 n=1

(n + 1)! k

k=1

∑∞ ( )

f (n+1) (1) ∑

n

n −s

F (s) = (−1)n−k k (1.47)

n=1

(n + 1)! k

k=1

∞

∑ ∞

(−1)k ∑ f (n+1) (1)

F (s) = (−1)n . (1.48)

ks (n + 1)!

k=1 n=k

By Cauchy’s inequality,

( )

f (n+1) (1) 1

=O , (1.49)

(n + 1)! rn+1

∞ ∞

(∞ )

∑ 1 ∑ 1 ∑ 1

=O = O(1) (1.50)

kσ rn k σ rk+1

k=1 n=k+1 k=1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 18

for all s. Then the Weierstrass Majorant test implies that the series (1.48)

is (absolutely and) uniformly convergent over the whole plane, implying the

analyticity of F (s).

Note that the inner sum in (1.48) is

∑

k−1

f (n+1) (1) ∑k

f (n) (1)

f (0) − (−1)n = (−1)n . (1.51)

n=1

(n + 1)! n=2

n!

mz m (z − 1) z m+1 − z m

f (z) = =m . (1.52)

(1 + z + · · · + z m−1 )(z − 1) zm − 1

By the Euclidean division, we have

(z m − 1)(z − 1) + z − 1 z−1

f (z) = m = m(z −1)+m m = m(z −1)+g(z),

zm − 1 z −1

say, where g(z) = m zz−1

m −1 . Let ζ = e

2πi/m

be the piervot’ny mth root of 1.

Then

∏

m−1

z m−1 + · · · + z + 1 = (z − ζ k )

k=1

and

∏

m−1

(1 − ζ k ) = m.

k=1

1

g(z) = m

1 + z + · · · + z m−1

∑

m−1 (1.53)

1 aj

= m ∏m−1 = m ,

k=1 (z − ζ k ) j=1

z − ζj

say. Since

1 1 − ζj

aj = lim (z − ζ j )g(z) = ∏ = , (1.54)

k̸=j (1 − ζ )

z→1 k m

we have

∑

m−1

1 − ζj

f (z) = m(z − 1) + . (1.55)

j=1

z − ζj

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 19

Prelude 19

Hence for ℓ ∈ N,

ℓ

∑

m−1

1 − ζj

f (ℓ) (z) = δℓ,1 m + (−1) ℓ! ℓ+1

, (1.56)

j=1 (z − ζ j )

ℓ

∑

m−1

1

f (ℓ) (1) = δℓ,1 m + (−1) ℓ! ℓ

. (1.57)

j=1 (1 − ζ j )

1 − e 2πi

m > r(≥ 1),

r = 2 − δ (δ > 0) and the conditions in Theorem 1.3 are satisfied.

But in this case, the function F (s) reduces to the familiar (1 − 21−s )ζ(s)

for which the most famous procedure is that of Euler transformation. [Has],

[So]. Cf. Corollary 1.3 below.

√

We may also take m = 4, in which case r = √2 − δ (δ > 0). The case

√

m = 6 does not work since then |1 − e(2πi)/6 | = 2 − 2 < 1. For bigger

m one has to make some trick and in Goss’s case, it is integration by parts.

Corollary 1.3 The Knopp-Hasse-Sondow diﬀerence operator expression

∑∞ ( ) ∞

1 ∑ ∑

n

k n −s ∆n 1−s

(1 − 2 )ζ(s) =

1−s

(−1) (k + 1) = (1.58)

n=0

2n+1 k n=0

2n+1

k=0

the sum in the mid term (cf. (6.3)).

Proof. We consider the case m = 2 of Example 1.1. Then ζ = −1 and

(1.57) reads

ℓ

(−1) ℓ!

f (ℓ) (1) = (1.59)

2ℓ

for ℓ ≥ 2. Hence (1.47) reads

∑∞ n+1 ∑

n ( )

(−1) n−k n

(1 − 21−s )ζ(s) = F (s) = n+1

(−1) k −s (1.60)

n=1

2 k

k=1

∞

∑ 1

= n+1

S(n),

n=1

2

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 20

say, where

∑

n−1 ( )

k n −s

S(n) := (−1) (k + 1) . (1.61)

k+1

k=0

(n+1) (n) ( n

)

Since k+1 = k + k+1 , it follows that

∑n ( )

k n+1 −s n −s

S(n) = (−1) (k + 1) − (−1) (n + 1) (1.62)

k+1

k=0

∑

n−1 ( )

k n −s

− (−1) (k + 1)

k

k=0

∑n ( )

n −s

= S(n + 1) − (−1)k (k + 1) ,

k

k=0

∑

n ( )

n −s

S(n + 1) − S(n) = (−1)k (k + 1) . (1.63)

k

k=0

Since

∞

∑ ∑∞

1 1

n+1

S(n + 1) = 2 n+2

S(n + 1)

n=1

2 n=1

2

(∞ )

∑ 1 1 1

=2 n+1

S(n) − S(1) = 2(1 − 21−s )ζ(s) − (1.64)

n=1

2 4 2

∑∞ ( )

1 ∑

n

k n −s

n+1

(−1) (k + 1) (1.65)

n=1

2 k

k=0

∞

∑ 1 1

= (S(n + 1) − S(n)) = (1 − 21−s )ζ(s) − ,

n=1

2n+1 2

transformation formula via contour integrals. We denote the elliptic theta

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 21

Prelude 21

∞

∑

e−πn

2

t

ϑ(t) = ϑ2 (t) = (1.66)

n=−∞

for t > 0 (eventually we may take Re t > 0; (1.75)). Let α > 0 and N > 0

be a large integer. Let C = CN denote the rectangle with vertices at

( ) −πz 2 t

± N + 12 ± αi. Then integrate the function f (z) = ee2πiz −1 around CN

and letting N → ∞ to prove that

∫ ∞−αi ∫ ∞+αi

e−πz t e−πz t

2 2

ϑ(t) = dz + dz = I1 + I2 , (1.67)

−∞−αi e2πiz −1 −∞+αi 1 − e2πiz

series

∞

∑

1 ( )−1

= −e−2πiz 1 − e−2πiz = −e−2πiz e−2πikz ,

e2πiz − 1

k=0

∞ ∫

∑ ∞−αi ∞ ∫

∑ ∞−αi

I1 = −e−2πiz e−πz e−πz

2 2

t−2πikz t−2πikz

dz = dz

k=0 −∞−αi k=1 −∞−αi

by termwise integration.

By Cauchy’s theorem,

∫ ∞ we may prove that the inner integral is equal to

the one with α = 0, −∞ e−πx t−2πikx dx.

2

Similarly,

∞ ∫

∑ ∞+αi ∫ ∞

−πz 2 t+2πikz

e−πx

2

t+2πikx

I2 = e dz = dx.

k=0 −∞+αi −∞

∑∞ ∫ ∞ ∫ ∞

−πx2 t

( −2πikx )

e−πx t dx

2

2πikx

ϑ(t) = e e +e dx + (1.68)

k=1 −∞ −∞

∑∞ ∫ ∞ ∫ ∞

e−πx t cos 2πkx dx + e−πx t dx

2 2

=2

k=1 −∞ −∞

∞

∑ ∫ ∞

e−πx t cos 2πkx dx.

2

=

k=−∞ −∞

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 22

The next step is to apply Example 3 [WW, p.114] which asserts that

for λ > 0,

∫ ∞ ∫ ∞

e−λx cos 2λax dx = 2λ− 2 e−λa e−x dx.

2 1 2 2

(1.69)

−∞ −∞

Substituting (1.69) in (1.68), we obtain

( )∫ ∞

1 1

e−x dx.

2

ϑ(t) = √ ϑ (1.70)

πt t −∞

∫ ∞

√

e−x dx = π.

2

(1.71)

−∞

formula

( )

1 1

ϑ(t) = √ ϑ (1.72)

t t

valid for Re t > 0.

This method is stated in [MitK] as one of the methods to evaluate the

probability integral.

If we do not reduce the integral to the one along the real axis, then for

(1.68), we have

∑∞ ∫ ∞ ( )

ϑ(t) = eπα −2πkα e−πx t e2πi(tα−k)x + e−2πi(tα−k)x dx

2 2

(1.73)

k=1 −∞

∫ ∞+αi

e−πz t dx

2

+

−∞+αi

∞ ∫

∑ ∞ ∫ ∞+αi

πα2 −2πkα −πx2 t

e−πz t dz.

2

= 2e e cos 2π(tα − k)x dx +

k=1 −∞ −∞+αi

orem three times. We may clarify the situation and why three times appli-

cations lead to (1.72) by our theory of modular relation which originates

from Riemann (succeeded in a remarkably successful way by Hecke and then

by Bochner for a more extensive class of zeta-functions). According to this

Riemann-Hecke-Bochner correspondence, the theta-transformation formula

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 23

Prelude 23

which is the Riemann zeta-function in this case. Since the functional equa-

tion exhibits the symmetry with respect to the critical line σ = 12 , the above

argument may be thought of as the result of applying the symmetry three

times, resulting one time application, i.e. amounting to the equivalence to

the functional equation.

To emphasize the ideas of Hecke, it is more appropriate to work with

the theta series

∞

∑ 2

Θ(τ ) = eπin τ

(1.74)

−∞

with Re z > 0. Θ(τ ) and ϑ(z) shift to each other under the rotation (1.1)

For Θ(τ ), (1.72) reads

( ) ( )1

1 τ 2

Θ − = Θ (τ ) , (1.76)

τ i

the result of action of the Spiegelung (3.102).

We shall refer to just three of them, the Poisson summation formula, the

Lipschitz summation formula and (a generalization of) the Plana summa-

tion formula. The Lipschitz summation formula is almost equivocal to the

functional equation and is expounded in [NTA, Theorem 5.2, p.128]. Cf.

also Theorem 4.6 below and [PP] for the latest reference. For the most rele-

vant Voronoĭ summation formula, we refer to [BeV] and [AAW4], and many

references given therein. We hope to return to the study on the summation

formulas in subsequent volumes including truncated ones.

The Poisson summation formula has been playing a prominent role in

establishing a huge class of identities in scientific disciplines. We state the

multi-dimensional version according to [Bak, pp.11-113].

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 24

fˆ(t) is defined by

∫ ∞

ˆ

f (t) = f (t)e(−t · x) dx, (1.77)

−∞

written fˆ ∈ C(Rn ). If fˆ(t) ∈ L1 (Rn ), then we have the inversion formula

∫ ∞

f (t) = fˆ(t)e(t · x) dx (1.78)

−∞

n

dimensional unit interval U n = [0, 1) , then we view f as a function on the

n-dimensional torus T = R/Z and then the Fourier transform (1.77) takes

the form

∫

ˆ

f (t) = f (t)e(−t · x) dx. (1.79)

Tn

∑

F(f )(t) = fˆ(n)e(t · x) (1.80)

n∈Zn

∑ ˆ

Lemma 1.1 (Poisson summation formula) If n∈Zn |f (n)| < ∞ and

f ∈ C(R ), then

n

(i)

∑

f (t) = F(f )(t) = fˆ(n)e(t · x). (1.81)

n∈Zn

(ii) If

∑

sup |f (n + t)| < ∞ (1.82)

t∈U n

n∈Zn

and

∑

|fˆ(m)| < ∞ (1.83)

m∈Zn

∑ ∑

fˆ(n) = f (m). (1.84)

n∈Zn m∈Zn

November 19, 2014 14:27 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 25

Prelude 25

The following lemma gives a possibility of continuing the resulting integral

in Theorem 1.4 to the left.

Lemma 1.2 Suppose that f (x) is defined and analytic in the half-plane

Re x ≥ 0. Then the two representations for F(s) (inversion of (6.61)) and

∫ ∞

1 f (ix) − f (−ix) s dx

F(s) = − π x (1.85)

sin 2 s 0 2i x

for 0 < σ < 2, are equivalent. Further, in the case where

Proof. By rotating the positive real axis by π2 , we obtain

∫ ∞ ∫ ∞

s−1 πi

s

F (s) = f (ix) (ix) idx = e 2 f (ix) xs−1 dx

0 0

and similarly

∫ ∞

− πi

2 s

F (s) = e f (−ix) xs−1 dx.

0

Hence

∫ ∞

− e−

πi πi

πs e 2 s 2 s f (ix) − f (−ix) s−1

sin F(s) = F (s) = − x dx,

2 2i 0 2i

which is (1.85).

The following theorem is a slight generalization of the Plana summa-

tion formula [KoshII, p. 217] which is the case z = 1.

We refer to the following growth condition of Stirling type:

( π )

F(s) = O e− 2 |t| |t|

aσ−b

(1.87)

a

+ b ≥ 1. (1.88)

2

Theorem 1.4 (Generalized Plana summation formula) Suppose that the

Mellin transform F(s) of the function f (x) (x > 0) satisfies the Stirling

type condition (1.87). Suppose f (x) is defined and analytic in the half-plane

σ ≥ 0 in Lemma 1.2, that (1.86) holds, and that F(s) is regular in the strip

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 26

we have Plana’s summation formula as a modular relation (Re z > 0)

∞

∑

F (n)f (nz) = ζΩ (0) Res F(s) + F(1)z −1 (1.89)

s=0

n=1

∫

2 ∞

f (ix) − f (−ix) ( x )

− σ dx,

z 0 2i z

provided that the infinite series and the integral in (1.89) are (absolutely)

convergent. The first term on the right may be expressed as

∫∞ζΩ (0)f (0) under

some growth condition on F(s) and the second term is ρzΩ 0 f (x)dx, where

(r)

2r+1 π r2 ζΩ (s)

ρΩ = − √ indicates the residue of the Dedekind zeta-function at

|∆|

s = 1 given in (3.152).

Proof of Theorem 1.4. We move the line of integration to σ = −α by the

Cauchy residue theorem.

( Then) we have for Re z > 0 and α > 1, using the

κ(1−σ)

estimates ζΩ (s) = O |t| and (1.87)

∞

∑ ∫

1

F (n)f (nz) = F(s)ζΩ (s)z −s ds (1.90)

n=1

2πi (α)

∫

1

= R0 + R1 + F(1 − s)ζΩ (1 − s)z s−1 ds,

2πi (1+α)

integral on the right, and where Rj indicates the residue at s = j, j = 0, 1

of the integrand, so that

s=0

ζΩ (1 − s) = A1−2s G(1 − s)ζΩ (s) , (1.91)

so that the integrand of the integral, say Iα , on the right of (1.90) reduces

to

∫ ∑∞ ( 2 )−s

A A n

Iα = F(1 − s)G(1 − s) F (n) ds. (1.92)

2πiz (α) n=1

z

∫ ∞

1 f (ix) − f (−ix) −s

F(1 − s) = − x dx (1.93)

cos π2 s 0 2i

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 27

Prelude 27

absolute convergence as is proved in [KoshII]), we obtain

∫ ∞

2A ∞

f (ix) − f (−ix) ∑ dx

Iα = σ1 (x, n) , (1.94)

zπ 0 2i n=1

x

where

∫

1 π −s

σ1 (x, n) = ( π ) G(1−s)(A2 nx/z) ds = K(A2 nx/z), (1.95)

2πi (c) 2 cos 2 s

A

∑∞ (x)

whence π n=1 σ1 (x, n) = σ z and (1.90) follows, completing the proof.

Corollary 1.4 ([SC, p. 90, (9)]). Under the conditions of Theorem 1.4,

we have Plana’s summation formula

∞

∑ ∫ ∞

1

f (n) = − f (0) + f (x)dx

n=1

2 0

∫ ∞ (1.96)

f (ix) − f (−ix) 1

−2 dx,

0 2i e 2πx −1

π π

x = |x|eiθ , − <θ<

2 2

and f (x) is analytic there, and that the infinite series and the integral in

(1.96) are (absolutely) convergent.

Our expression in the following Theorem 1.5 for the Hurwitz-Lerch zeta

function Φ(z, s, a) ((5.38)) has its genesis in the confluent hypergeometric

function and Plana’s summation formula, which in the long run, amounts

to the functional equation for the Riemann zeta-function.

∫ ∞

1 −s zx

Φ (z, s, a) = a + dx (1.97)

2 (x + a)s

∫ ∞ 0 (

( 2 )− s x) dx

−2 x + a2 2 sin x log z − s arctan 2πx − 1

.

0 a e

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 28

have

1 −s a1−s

ζ (s, a) = a +

2 s−1

∫ ∞ ( x)

(1.98)

( 2 )− s dx

+2 a + x2 2 sin s arctan 2πx − 1

.

0 a e

Corollary 1.6 (Binet’s second expression [SC, p. 17, (30)]). For Re a > 0

we have

( ) ∫ ∞

Γ (a) 1 arctan xa

log √ = a− log a + 2 dx. (1.99)

2π 2 0 e2πx − 1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 29

Chapter 2

is introduced most commonly as the principal solution to the diﬀerence

equation

Γ(1) = 1. (2.2)

The following two well-known formulas for the gamma function are often

used subsequently without further mentioning.

π

The reciprocity relation Γ(s)Γ(1 − s) = , (2.3)

sin(πs)

( )

1

Γ(2s) = π − 2 22s−1 Γ(s) Γ s +

1

The duplication formula . (2.4)

2

function [ErdT, I,p.4]: For integer m ≥ 2

∏

m−1 ( r)

− m−1

Γ(ms) = (2π) 2

mms−1/2 Γ s+ . (2.5)

r=0

m

[ ]

( nThis is) used e.g. by Bochner Bo2 to reduce the gamma factor

Γ m s + a in terms of Γ (s + b). As will be seen, this procedure is the

reduction of the H-function to the G-function.

29

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 30

( )ρ

1 d 2ρ Γ(z − ν + ρ + 1) 2ν−2z−2ρ

− s2ν−2z = s . (2.6)

s ds Γ(z − ν)

( )ρ

1 d Γ(2z + 2ρ + 1)Γ(z + 1) −2z−2ρ−1

− s−2z−1 = 2−ρ s . (2.7)

s ds Γ(z + ρ + 1)Γ(2z + 1)

Below we will see that (2.3), resp. (2.4) will be magnified to (2.101),

resp. (2.97). At present we shall see that (2.3) is used in the derivation of

(2.8) in the following context.

Example 2.1 We presuppose basics for G- and allied functions and prove

( )

a

3,1

G1,3 z (2.8)

b, b + 21 , a

√ ( √ 2b ( √ )

= 2−b π Γ(1 − 2a + 2b) (−2i z) U 2b − 2a + 1, 2b − 2a + 1, −2 z i

√ 2b ( √ ))

+ (2i z) U 2b − 2a + 1, 2b − 2a + 1, 2 z i ,

which entails

( )

a

G3,1 z (2.9)

1,3 0, 1 , a

2

√ ( √ ( √ ) √ ( √ ))

= π Γ(1 − 2a) e−aπi+2 z i Γ 2a, 2 z i + eaπi−2 z i Γ 2a, −2 z i .

mon factors. Extract the sines and cosines by the reciprocity

relation of the gamma function, therewith shifting the gamma

factors in the denominator to the numerator. Use Euler’s iden-

tity. Reduce to the G-functions.

By definition

∫ ( )

1 1

LHS = Γ(b + w)Γ b + w Γ(a + w)Γ(1 − a − w) z −w dw.

2πi L 2

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 31

cos(π(a + w))

22−b−w π 3/2 Γ(2b + 2w)Γ(2a + 2w)Γ(1 − 2a − 2w)

π

√

= 21−b π Γ(2b + 2w)Γ(2a + 2w)Γ(1 − 2a − 2w)

√ −2w √ −2w

× (eπia (−2i z) + e−πia (2i z) ).

( ( ) ( ))

√ √ 2a √ 2a

LHS = 2 −b 2,1

π G1,2 −2i z 2,1

+ G1,2 2i z

2b, 2a 2b, 2a

√ ( √ √

2b

= 2−b π Γ(1 − 2a + 2b) (−2i z) U (2b − 2a + 1, 2b − 2a + 1, −2i z)

√ 2b √ )

+ (2i z) U (2b − 2a + 1, 2b − 2a + 1, 2i z) ,

( )

(1 − s, 1)

H1,3 z 1−s 1

2,1

(2.10)

( 2 , 2 ), (0, 1), (1 − 2s , 12 )

Γ(s) ( 1−s )

= 1−s √ e 2 πi−2zi Γ(1 − s, −2zi) + e− 2 πi+2zi Γ(1 − s, 2zi)

1−s

2 π

and

( )

(1 − 2s, 2)

2,1

H1,3 z (2.11)

( 12 − s, 1), (0, 2), (1 − s, 1)

Γ(2s) ( 1−2s √ √

= 2−2s √ e− 2 πi+2 z i Γ(1 − s, 2 z i)

2 π

1−2s √ √ )

+e 2 πi−2 z i

Γ(1 − s, −2 z i) .

( )

(1 − (1 + 2s ), 12 ), (1 − s, 1)

2,2

H2,4 z s 1

(− 2 , 2 ), (0, 1), (1 − 2s , 1), (1 − s+1 1

2 , 2)

( )

1 2

−s

= 1−s G3,1 1,3 z

2 .

2 π − s , 0, 1

2 2

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 32

( 1−s )

1 3,1

2

G 2

1−s 1,3 z 1−s 1 ,

2 π 2 , 0, 2

whence the result follows by (2.8). The second formula amounts to the first

by (2.93).

Applying (2.4) twice, we find that the LHS of the third equality amounts

to

( )

1 (− 2s , 12 )

H 2,1

z s 1

22−s π 1,3 (− 2 , 2 ), (0, 12 ), ( 21 , 12 )

central role in Ewald expansion (Chapter 5) is defined by

∫ ∞

dt

Γ(s, z) = ts e−t , σ > 0. (2.12)

z t

( )

1 √ √

Γ(1, z) = e−z , Γ ,z = π erfc( z), (2.13)

2

(ii) Prove the formula for the Laplace transform

[ ( a )] √

(p) = 2a 2 p 2 −1 Kν (2 ap),

ν ν

L Γ ν, (2.14)

t

whence deduce

[ ( )] 1 √

L erfc at−1/2 (p) = e−2a p (2.15)

p

( )

Γ(ν) 1

L[Γ(ν, at)](p) = 1− . (2.16)

p (1 + ap )ν

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 33

∫ ∞

C(z, a) = ta−1 cos t dt (2.17)

z

1 πi a ( ) 1 πi ( )

= e 2 Γ a, −iz + e− 2 a Γ a, iz

2 2

and its counterpart

∫ ∞

S(z, a) = ta−1 sin t dt (2.18)

z

1 πi a ( ) 1 ( )

e 2 Γ a, −iz − e− 2 a Γ a, iz

πi

=

2i 2i

for Re a > 0.

(iv) Check the formulas [ErdH, II, p.143]

∫ ∞

E1 (z) = −Ei(−z) = e−t t−1 dt (2.19)

z

= Γ(0, z) = e−z Ψ(1, 1, z)

and

∫ x

1

li(x) = Ei(log x) = dt = −E1 (− log x). (2.20)

0 log t

Remark 2.1 The limit of the integral in (2.19) is called the logarithm

integral and constitutes the main term of the prime number theorem:

∑

π(x) := 1 = Li(x) + E(x), (2.21)

p≤x

p:prime

∫∞

where Li(x) = 0 log1 t dt. Some equivalent forms of (2.21) appear in Chap-

ter 6, (6.121) and Theorem 6.13.

2.2 Zeta-functions

[SpH] is a very comprehensive account of the theory of zeta-functions, re-

stricted mainly to the Riemann and allied zeta-functions, although [SpH,

Vol. II] contains a more extensive class of zeta-functions. [Kaw] is quite a

complete survey on zeta-functions up to 70’s in Japanese, though. [Joy] is,

its title notwithstanding, quite an excellent survey of various zeta-functions

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 34

including modern ones. Also [WeIII] contains important material for the

theory of zeta-functions, esp. [We1], [We3], [We4]. [We2], [Weds] and

other papers in [SpH] will be discussed in our coming Volume II. There are

many pther classes of zeta-functions whose complete theory have not fully

appeared in book form. There exists a vast literature on zeta-functions

arising from representation theory cf. e.g. [Bu1] or from the theory of pre-

homogeneous vector spaces (p.v.), cf. [MS] and many others. Hopefully, a

comprehensive book will appear on zeta-functions in due course.

Of course, regarding a specific zeta-function, there are more literature

specified to that zeta-function. [Titr] is a masterpiece work on the Riemann

zeta-function. [Pra] is a rather extensive survey on Dirichlet L-functions.

So is Suetuna [Sue] which is also devoted to analytic number theory in

algebraic number fields. For a modern authoritative account of analytic

number theory, we refer to [IwKo]. The recent [AIK] contains interesting

expositions on the Bernoulli numbers and zeta-functions.

We follow [Sa] to give a table of zeta-functions whose functional equa-

tions follow from the theta-transformation formula. Although his presenta-

tion is from the view point of prehomogeneous vector spaces and his table

contains the p.v. and the relevant invariant, we refer to the zeta-function

part only. For recent results, we refer e.g. to Ohno et al [Oh], [OTW].

zeta-functions generalizations

Riemann [Rie2] Hecke L-functions with Grössen characters1

Dedekind Hecke L-functions with Grössen characters

Epstein (positive definite) Epstein zeta-function with harmonic polynomials

Siegal (indefinite) [Sie3] Maass [Maa3], Hejhal

Hey (Simple algebra) [Hey] Godement-Jacquet [GJ]

Koecher [Koc] Maass [Maa1], [Maa2], [Maa4]

Eisenstein series of SLn (Z) E-series of SLn (Z) with representation of SOn

Shintani ?

1 means [Heck1], [Heck2].

eralized Hurwitz-Lerch zeta-function in §5.2.1. It reduces to the Hurwitz

zeta-function and the Lipschitz-Lerch transcendent, which then reduce to

the Hurwitz zeta-function

∞

∑ 1

ζ(s, x) = s, σ > 1. (2.22)

n=0

(n + w)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 35

Landau-Walfisz Bochner-Chandrasekharan-Narasimhan

∑ ∑

Z(s) = ∞ αk λ−s

k , σZ ≥ r φ(s) = ∞ α λ−s

k , σφ

∑ k=1

−s ∑k=1 k −s

W (s) = ∞ k=1 k µk , σW ≥ r

β ψ(s) = ∞ k=1 βk µk , σψ

W is regular at s = r, r > 0 r∈R

β1 , · · · , βM > 0, δ1 , · · · , δN > 0 α1 , · · · , α M > 0

α1 , · · · , αM ∈ R, γ1 , · · · , γN ∈ R β1 , · · · , βM ∈ C

∏N

∆2 (r−s) ν=1 Γ(γν −δν s) ∆(r−s) ∏ Γ(αν r+βν −αν s)

G(s) = ∆1 (s)

= ∏N G(s) = ∆(s) = N ν=1 Γ(β +α s)

ν=1 Γ(αµ +βµ s) ν ν

G(0) = 0, has a pole at s = r -

αµ + rβµ > 0, 1 ≤ µ ≤ M -

∑N ∑ ∑N

βν = N δν = H ν=1 αν = A

2η ∑N ν=1 ∑N ν=1 M −N 2

H = r , η := ν=1 γν − µ=1 αµ + 2 η = Ar

1

2

< η < 32 , 52 − η < H -

Z(s) = G(s)W (r − s) φ(s) = G(s)ψ(s)

α

Z(s) = O(eγ|t| ) χ(s) = O(ec|t| ), c > 0, α > 0

Here the last row specifies the growth conditions in the strip σ1 ≤ σ ≤ σ2

and the constants γ = γ(σ1 , σ2 ).

Since the functional equation appears in the form of φ(s) = G(s)ψ(s)

in Table 2.2, we keep this notation G(s) to indicate the gamma quotient.

Some aspects of this form of functional equations have been developed in

[KTT2] and we refer to it for details. The gamma quotient typically appears

in §4.6 and (3.177) prior to it; cf. the remark at the end of §4.6.

functions, which will be touched upon again in the context of H-or G-

functions. We note a priori that the added provisos to the complex integral

representations (Mellin-Barnes integrals) come from the separation of poles

requirement of the relevant gamma factors (cf. definition (1.22) of the H-

function).

The J-Bessel function Js (z) is a solution to the Bessel diﬀerential equa-

tion

( )

d2 w dw d dw

x2 2

+x + (x2 − ν 2 )w = x x + (x2 − ν 2 )w = 0 (2.23)

dx dx dx dx

and is given as a power series (2.46) ([ErdH, III, p.4]) below. What is

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 36

∫

1 1 Γ( 12 ν + 12 z) ( x )−z

Jν (x) = dz (2.24)

2 2πi (c) Γ(1 + 12 ν − 21 z) 2

( )

1 1,0 x −(

= H0,2 ( ) )

2 2 ν , 1 , −ν , 1

2 2 2 2

for − Re ν < c < 1 (x > 0, which may read Re x > 0). We note that on

the left-side of [ErdH, III,(34) p.21] should read 4πiJν (x) and that without

the restriction − Re ν < c < 1, (2.24) still makes sense, but it need not

represent the Bessel function.

Exercise 2.5 Check that (2.24) reduces to [PBM, 8.4.19.1, p. 651]

( )

√ x −

Jν (2 x) = G1,0 , (2.25)

0,2

2 ν2 , − ν2

∫ ( x )ν+2z

1 Γ(−z)

Jν (x) = dz (2.26)

2πi (0) Γ(1 + ν + z) 2

( )

1 x 1, 1 + ν

= G0,1

2 2,0 2 −

as well as (2.110).

Solution All are immediate consequences of formulas in Proposition 2.3.

We note that in general the Bessel functions, say Fs reduce in the case

of s being a half-integer ν + 21 , to an elementary function ([ErdT, II]); cf.

(2.27), (2.32). Js (z) reduces to

1 ( d )ν sin z

ν

Jν+1/2 (z) = (−1) √ π z ν+1 (2.27)

2z

zdz z

( ( )

1 ( π )∑

[n2]

m n + 1/2 −2m

= √π sin z − n (−1) (2z)

2 z 2 m=0

2m

( ) )

( π ) ∑

[ n−1

2 ]

m n + 1/2 −2m−1

+ cos z − n (−1) (2z) ,

2 m=0

2m + 1

where [x] indicates the integer part of x, i.e. the greatest integer not ex-

ceeding x.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 37

Iν (z) = e− 2 iν Jν (ze− 2 iν )

π π

(2.28)

d2 w dw

x2 2

+x − (x2 + ν 2 )w = 0 (2.29)

dx dx

and is given by the power series (2.45).

∫ ( x )s−2z

1

−πYs (x) = Γ(z)Γ(z − s) cos π(z − s) dz, (2.30)

2πi (c) 2

The K-Bessel function Ks (z) may be defined by ([Vi2, (6.14), p.107])

∫

1 ( z )ν ∞ −t− z2 −ν−1

Kν (z) = e 4t t dt,

2 2 0 (2.31)

1 π

Re(ν) > − , | arg z| < .

2 4

Ks (z) reduces ([ErdH, (40), p.10]) to

( π )1/2 ∑ν

Γ(ν + m + 1)

Kν+ 12 (z) = e−z (2z)−m . (2.32)

2z m=0

m! Γ(ν + 1 + m)

√

π −z

K± 12 (z) = e . (2.33)

2z

The following diﬀerential operator ([Wat, p.79]) is often used

( )ρ

1 d { ν }

− s Kν (s) = sν−ρ Kν−ρ (s), (2.34)

s ds

e.g. cf. Exercises 2.1, 2.2.

We recall the following limit relation ([Vi2, (6.22),p. 110]) which is

already referred to as (1.29) as well as the symmetry property which will

turn out to be relevant to the functional equation:

1 −ν

lim z ν Kν (2az) = a Γ(ν), Re ν ≥ 0, ν ̸= 0. (2.35)

z→0 2

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 38

2.4 Ψ-functions

( )

(a1 , A1 ), . . . , (ap , Ap )

p Ψq ;z

(b1 , B1 ), . . . , (bq , Bq )

∏

p

(2.37)

∞ Γ(aj + Aj k)

∑ j=1 zk

=

∏

q

k!

k=0 Γ(bj + Bj k)

j=1

( )

*

(a1 , A1 ), . . . , (ap , Ap )

p Ψq ;z

(b1 , B1 ), . . . , (bq , Bq )

∏

q

(2.38)

Γ(bj ) ( )

j=1 (a1 , A1 ), . . . , (ap , Ap )

= p Ψq ;z

∏p

(b1 , B1 ), . . . , (bq , Bq )

Γ(aj )

j=1

the Ψ-series (the Ψ-function). By the very definition (2.42) in the next

section, the Ψ-functions are generalizations of hypergeometric functions.

The following relation between H and Ψ-function is well known. If

either

∑

p ∑

q

Aj < Bj

j=1 j=1

or

∑

p ∑

q

B1B1 · · · Bq q

B

Aj = Bj and |z| < A

,

1 · · · Ap

AA 1 p

j=1 j=1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 39

then

( )

(1 − a1 , A1 ), . . . , (1 − ap , Ap )

1,p

Hp,q+1 z

(0, 1), (1 − b1 , B1 ), . . . , (1 − bq , Bq )

( ) (2.39)

(a1 , A1 ), . . . , (ap , Ap )

= p Ψq ; −z .

(b1 , B1 ), . . . , (bq , Bq )

[MaSa, p.11, (1.7.8)])

( ) ( )

(1 − a, A) (a + Ac, A)

1,1

H1,2 z = z c 1 Ψ1 ; −z , (2.40)

(c, 1), (1 − b, B) (b + Bc, B)

( ) ∞

∑

(a, A) Γ(a + Ak) z k

1 Ψ1 ;z = .

(b, B) Γ(b + Bk) k !

k=0

symbol ([Vi2, p.27])

k

∏

k−1 z }| { Γ(a + k)

(a)k = (a + j) = a(a + 1) · · · (a + k − 1) = (2.41)

j=0

Γ(a)

( )

a1 , a2 , . . . , ap

F

p q ; z

b1 , b2 , . . . , bq

( )

∗ (a1 , 1), . . . , (ap , 1)

= p Ψq ;z

(b1 , 1), . . . , (bq , 1)

∑∞

(a1 )k (a2 )k · · · (ap )k z k

= , (2.42)

(b1 )k (b2 )k · · · (bq )k k!

k=0

the second equality being due to (2.38), are called the generalized hyper-

geometric series (as functions, they are called the generalized hyperge-

ometric functions).

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 40

dius of convergence is

(a1 )k (a2 )k · · · (ap )k (b1 )k+1 (b2 )k+1 · · · (bq )k+1 (k + 1)!

lim

k→∞ (b1 )k (b2 )k · · · (bq )k (a1 )k+1 (a2 )k+1 · · · (ap )k+1 k!

(k + b1 )(k + b2 ) · · · (k + bq )

= lim (k + 1)

k→∞ (k + a1 )(k + a2 ) · · · (k + ap )

∞, (p < q + 1)

= 1, (p = q + 1) .

0, (p > q + 1)

Example 2.2

( ) ∞

∑

− zk

0 F0 ;z = = ez , (2.43)

− k!

k=0

( ) ∑∞

a zk

1 F0 ;z = (a)k = (1 − z)−a , (2.44)

− k!

k=0

( ) ∑∞

− 1 zk Γ(b) ( √ )

0 F1 ;z = = √ b−1 Ib−1 2 z , (2.45)

b (b)k k! z

k=0

( ) ∑ ∞

− 1 (−z) k

Γ(b) ( √ )

0 F1 ; −z = = √ b−1 Jb−1 2 z , (2.46)

b (b)k k! z

k=0

ing notation ([ErdH, p.248])

( )

a

1 F1 ;z ≡ Φ(a, c; z), (2.47)

c

U = Ψ (sometimes referred to as the Tricomi function) is defined by ([ErdH,

(2), p.255], [Tem, p.175])

U (a, c; z) ≡ Ψ(a, c; z)

∫ ∞

1 (2.48)

e−zt ta−1 (1 + t)

c−1−a

= dt, Re a > 0.

Γ(a) 0

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 41

These functions are the solutions (the latter in Re x > 0) of the confluent

hypergeometric equation

d2 y dy

x 2

+ (c − x) − ay = 0 (2.49)

dx dx

and are referred to as the confluent hypergeometric function of the first

and the second kind, respectively. In the case of the confluent hypergeomet-

ric function of the second kind, we shall use the U notation predominantly

over Ψ notation so as to avoid confusion with the Ψ-function while for the

first kind, we use both notation Φ and 1 F1 interchangeably. The following

relation holds true:

Γ(1 − c) Γ(c − 1) 1−c

= Φ(a, c; z) + x Φ(1 − c + a, 2 − c; z),

Γ(1 − c + a) Γ(a)

function [PBM, p.799].

We shall therefore use several symbols to denote the confluent hyperge-

ometric functions of the first kind

( ) ( ) ( )

a ∗ (a, 1) Γ(c) (a, 1)

Φ(a, c; z) = 1 F1 ; z = 1 Ψ1 ;z = 1 Ψ1 ; z . (2.51)

c (c, 1) Γ(a) (c, 1)

of two independent solutions to the Whittaker diﬀerential equation (cf.

[ErdH, p.248, (4)]).

( )

d2 w 1 κ 1

−µ

2

+ − + + 4

w = 0. (2.52)

dx 4 x x2

( )

a1 , a2 , . . . , ap , ap+1 Γ(bq+1 )

p+1 Fq+1 ;z = (2.53)

b1 , b2 , . . . , bq , bq+1 Γ(ap+1 )Γ(bq+1 − ap+1 )

∫ 1 ( )

a , a , . . . , ap

× tap+1 −1 (1 − t)bq+1 −ap+1 −1 p Fq 1 2 ; tz dt.

0 b1 , b2 , . . . , bq

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 42

∫ 1 ( ∞ )

∑

t (1 − t)

b−1 c−b−1 k

ak (tz) dt

0 k=0

function

∫ 1 (absolutely convergent in the unit circle). Then the resulting in-

tegral 0 tb+k−1 (1 − t)c−b−1 dt being the beta-function, we deduce that

∞ ∫ 1 ( ∞ )

∑ (b)k k Γ(c) ∑

ak z = t (1 − t)

b−1 c−b−1 k

ak (tz) dt. (2.54)

(c)k Γ(b)Γ(c − b) 0

k=0 k=0

Viewing

( )

a1 , a2 , . . . , ap , ap+1

p+1 Fq+1 ;z

b1 , b2 , . . . , bq , bq+1

∑∞ (b)k (a1 )k (a2 )k ···(ap )k (b)k

as the power series k=0 ak (c)k z k with ak = (b1 )k (b2 )k ···(bq )k and (c)k =

(ap+1 )k

(bq+1 )k , we apply (2.54) ro conclude (2.53).

Example 2.5 (i) Using (2.43) and (2.44) in (2.53), respectively, we have

( )

a

Φ(a, b; z) = 1 F1 ;z

b

∫ 1

Γ(b)

= ta−1 (1 − t)b−a−1 etz dt, Re a > 0 (2.55)

Γ(a)Γ(b − a) 0

( ) ∫ 1

a, b Γ(c)

2 F1 ;z = tb−1 (1 − t)c−b−1 (1 − tz)−a dt. (2.56)

c Γ(b)Γ(c − b) 0

[KTT3]. Rewrite (2.56) in the form of [GR, p.314, 3.197.1] (with µ = ρ = s):

∫ ∞

xν−1 (x + α)−s (x + β)−s dx (2.57)

0

( )

−s ν−s s, ν α

=β α B(ν, 2s − ν)2 F1 ;1 − ,

2s β

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 43

(note that the argument is taken in [−π, π)), though in [GR] the condition

on the argument of α and β is stated wrongly as

∫ ∞

xν−1 (x + α)−s (x + β)−s dx (2.59)

0

( )

s, ν α

= α−s β ν−s B(ν, 2s − ν)2 F1 ;1 −

2s β

and confirm the consistency of (2.57) and (2.59) by Kummer’s relation

([ErdH, p.105, (1), (3)]).

Solution E.g. to prove (2.57) we rotate the line of integration from R+ to

R+ ei arg α and apply [ErdH, p.115, (5)].

Exercise 2.7 Deduce the Mellin inversion formulas

∫ ( )

α−s β −s s, ν α

= αν B(ν, 2s − ν)2 F1 ;1 − x−ν dν

2πi (c) 2s β

and

∫ ( )

α−s β −s s, ν β

= β B(ν, 2s − ν)2 F1

ν

;1 − x−ν dν,

2πi (c) 2s α

where 1 < c < 2σ.

Remark 2.2 We may refer to (2.60) and (2.61) as the hypergeometric

transform in the same context as with the beta transform (4.2) and the

gamma transform ( (1.42), [Vi2, p.25]). To apply them care must be taken,

see Proposition 2.2.

Exercise 2.8 Recall the following formula for the Gauss hypergeometric

function:

( ) ( )

a, b −a Γ(c)Γ(b − a) a, c − b 1

2 F1 ; z = (1 − z) 2 F1 ; (2.62)

c Γ(c − a)Γ(b) 1+a−b 1−z

( )

Γ(c)Γ(a − b) c − a, b 1

+ (1 − z)−b F

2 1 ;

Γ(c − b)Γ(a) 1−a+b 1−z

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 44

valid for | arg(−z)| < π, | arg(1 − z)| < π, a − b ̸∈ Z ([ErdH, p.109 (3)]).

Start from the initial domain of α and β lying in the first and the fourth

quadrant, respectively, and continue (2.60) into another region

π π

≤ arg α < π, −π < arg β ≤ − . (2.63)

2 2

Proposition 2.2 (Hypergeometric transform) To apply the hypergeomet-

ric transforms (2.60), (2.61) to the series whose terms satisfy (2.58), we

are to use

−2s ∫

( )

β Γ(2s − ν)Γ(ν − s) s, 2s − ν α

= βν 2 F1 ; x−ν dν

2πi (c) Γ(s) 1+s−ν β

∫ ( )

α−s β −s ν Γ(ν)Γ(s − ν) s, ν α

+ α 2 F1 ; x−ν dν

2πi (c) Γ(s) 1−s+ν β

and

∫ ( )

α−2s Γ(2s − ν)Γ(ν − s) s, 2s − ν β

= αν F

2 1 ;

2πi (c) Γ(s) 1+s−ν α

−s −s ∫ ( )

α β ν Γ(ν)Γ(s − ν) s, ν β

+ β 2 F1 ; ,

2πi (c) Γ(s) 1−s+ν α

B(ν, 2s − ν)2 F1 (s, ν; 1 + s − ν; 1 − αβ ) (α and β are interchanged) into the

region (2.63) is given by

( )

α

B(ν, 2s − ν)2 F1 s, ν; 1 + s − ν; 1 − (2.66)

β

( )

Γ(2s − ν)Γ(ν − s) β

= α−s β s 2 F1 s, 2s − ν; 1 + s − ν;

Γ(s) α

( )

Γ(ν)Γ(s − ν) β

+ α−ν β ν 2 F1 s, ν; 1 − s + ν; .

Γ(s) α

Let Z(s) denote the Epstein zeta-function of the positive definite binary

quadratic form given in (4.88). In [KTT2] the integral formula is proved.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 45

√

2 22s as−1 π Γ(s − 12 ) 2{ }

Z(s) = s ζ(2s)+ 1 ζ(2s−1)+ s T1 (s)+T2 (s) , (2.67)

a s−

∆ 2 Γ(s) a

where T1 (s) and T2 (s) are given by

∫ ( ) Γ(2s − ν)Γ(ν − s)

ω̄ −2s

T1 (s) = 1 + eπi(ν−2s) ω̄ ν (2.68)

2πi (−ε) Γ(s)

( ω)

× 2 F1 s, 2s − ν; 1 + s − ν; ζ(2s − ν)ζ(ν)dν

ω̄

and

∫ ( ) Γ(ν)Γ(s − ν)

ω̄ −s ω −s

T2 (s) = 1 + e−πiν) ω ν (2.69)

2πi (−ε) Γ(s)

( ω )

× 2 F1 s, ν; 1 − s + ν; ζ(2s − ν)ζ(ν)dν.

ω̄

The Gauss hypergeometric function is often used in other number-

theoretic settings including Example 5.1 in Chapter 5, too. In [Sar], it

is used with (2.62). The following exercise will be used in solving Exer-

cise 5.2.

Exercise 2.9

( ) ∫ z

ν, s

F

2 1 ; −z = xs−1 (x + 1)−ν dx. (2.70)

s+1 0

([ErdT, p.310]).

∑L

(ii) Let P(x) = k=1 ρk x−sk be the residual function. Transform the

integral of P(x) as follows.

∫ 1 ∑

L ∫ 1/δ

xs−1 P(x + δ) dx = ρk δ s−sk xs−1 (x + 1)−sk dx

0 k=1 0

( ) (2.71)

∑

L

sk , s 1

−1 −sk

=s ρk δ 2 F1 ;− .

s+1 δ

k=1

( ) ( )

0, b a, b

F

2 1 ; z = lim F

2 1 ; z = 1. (2.72)

c z→0 c

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 46

( )

ν, s

2 F1 : −z = s(−z)−s B−s (s, 1 − ν), (2.73)

s+1

where

∫ z

β−1

Bz (α, β) = tα−1 (1 − t) dt

0

metric functions. We suppose that all the poles of

∏

m ∏

n

Γ(bj + s) Γ(aj − s)

j=1 j=1

( )

1 − a1 , . . . , 1 − an , an+1 , . . . , ap

Gm,n z (2.74)

p,q b1 , . . . , bm , 1 − bm+1 , . . . , 1 − bq

∏

m ∏n

Γ(bj − bk ) Γ(aj + bk )

∑ j̸=k j=1 j=1

m

z bk

∏p ∏q

Γ(aj − bk ) Γ(bj + bk )

k=1

j=n+1 j=m+1

{aj + bk }j=1 , {1 − aj + bk }j=n+1

n p

×p Fq−1

; (−1)p+m+n z ,

{1 − b + b } m

, {b + b } q

j k j=1 j k j=m+1

j̸=k

if p < q or (p = q, |z| < 1)

= ∏n ∏

m

Γ(aj − ak )

Γ(bj + ak )

∑n j=1 j=1

1

j̸=k

∏q ∏p

Γ(bj − ak )

z ak

k=1 Γ(aj + ak )

j=m+1 j=n+1

q

{b + a } m

, {1 − b + a }

×q Fp−1

j k j=1 j k j=m+1 1

; (−1)q+m+n ,

{1 − }n

{a } p

a j + a k j=1 , j + a k j=n+1 z

j̸=k

if p > q or (p = q, |z| > 1).

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 47

In particular, we have

( )

1 − a1 , . . . , 1 − an , an+1 , . . . , ap

1,n

Gp,q+1 z (2.75)

0, 1 − b1 , . . . , 1 − bq

∏

n

Γ(aj ) ( )

{aj }nj=1 , {1 − aj }pj=n+1

j=1

; −(−1) p+n

∏p ∏

q p Fq z ,

{b } q

Γ(aj ) Γ(bj ) j j=1

j=n+1 j=1

if p < q + 1 or (p = q + 1, |z| < 1)

∏n

Γ(aj − ak )

Γ(ak )

= ∑ n j=1

j̸=k 1

∏ ∏

q p

z ak

k=1 Γ(bj − ak ) Γ(aj + ak )

j=1 j=n+1

a , {1 − b + a } q

k j k j=1 1

×q+1 Fp−1

p ; (−1)q+n ,

{1 − a j + a k }n

, {a j + ak } z

j=1 j=n+1

j̸=k

if p > q + 1 or (p = q + 1, |z| > 1)

and

( )

1 − a1 , . . . , 1 − ap

G1,p z (2.76)

p,q+1

0, 1 − b1 , . . . , 1 − bq

p

∏

Γ(aj ) ( )

a1 , . . . , ap

j=1

; −z ,

∏

q p Fq

b1 , . . . , bq

Γ(b j )

j=1

if p < q + 1 or (p = q + 1, |z| < 1)

∏p

Γ(ak ) Γ(aj − ak )

= ∑ p j=1

j̸=k 1

∏

q

z ak

k=1 Γ(bj − ak )

j=1

a , 1 − b + a , . . . , 1 − b + a 1

×q+1 Fp−1

k 1 k q k

; (−1)p+q ,

k

z

1 − a + a , . ∨

. ., 1 − a + a

1 k p k

if p > q + 1 or (p = q + 1, |z| > 1)

k

where the symbol ∨ means that the k-th term should be omitted.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 48

identity gives an analytic continuation to |z| > 1 of the left-hand side.

( )

a1 , . . . , aq+1

q+1 Fq ;z

b1 , . . . , bq

∏

q

Γ(bj ) ( ) (2.77)

1 − a1 , . . . , 1 − aq+1

j=1

= q+1 G1,q+1 −z (z ∈

̸ [0, ∞]),

∏ q+1,q+1 0, 1 − b1 , . . . , 1 − bq

Γ(aj )

j=1

and we have

( )

a1 , . . . , aq+1

q+1 Fq ; −z

b1 , . . . , b q

∏

q+1

∏

q Γ(ak ) Γ(aj − ak )

Γ(bj )

j=1 ∑

q+1 j=1

j̸=k 1

= (2.78)

∏

q+1 ∏

q

z ak

Γ(aj ) k=1 Γ(bj − ak )

j=1 j=1

( )

ak , 1 − b1 + ak , . . . , 1 − bq + ak 1

× q+1 Fq k ;− .

1 − a1 + ak , . ∨. ., 1 − aq+1 + ak z

Example 2.6

( ) ( )

a 1 a 1

1 F0 ; −z = F

1 0 ; − , (2.79)

− za − z

( ) ( )

a, b Γ(c) Γ(b − a) 1 a, 1 − c + a 1

2 F1 ; −z = 2 F1 ; − (2.80)

c Γ(b) Γ(c − a) z a 1−b+a z

( )

Γ(c) Γ(a − b) 1 b, 1 − c + b 1

+ 2 F1 ;− .

Γ(a) Γ(c − b) z b 1−a+b z

Exercise 2.10 Check the following special case of (2.80):

( )

ν, s 1

2 F1 ;−

s+1 x

( ) (2.81)

s ν, ν − s Γ(s + 1)Γ(ν − s) s

= x ν 2 F1 ; −x + x

s−ν ν−s+1 Γ(ν)

([GR, p.1043]).

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 49

( )

s ν, ν − s

xν 2 F1 ; −x

s−ν ν−s+1

( ) (2.82)

Γ(s + 1)Γ(ν − s) s s, 0

+ x 2 F1 ; −x ,

Γ(ν) 1−ν+s

( ) ( )

b Γ(a − s) s 1−b+s

G2,0 z = z F ; −z

1,2 s, a Γ(b − s)

1 1

1−a+s

( )

Γ(s − a) a 1−b+a

+ z 1 F1 ; −z

Γ(b − a) 1−s+a

In what follows we shall use the power function z −s = e−s log z either by

considering log z (z ̸= 0) on its Riemann surface or by taking its principal

values, with z ∈

/ (−∞, 0].

We choose the path L which separates the poles of the integrand in such

a way that the poles of the gamma factor

∏

n

Γ(aj − Aj s)

j=1

∏

p ∏

q

Γ(aj − Aj s) Γ(bj − Bj s)

j=n+1 m+1

∏

m

Γ(bj + Bj s)

j=1

∏

p ∏

q

Γ(aj − Aj s) Γ(bj − Bj s)

j=n+1 j=m+1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 50

lie to the left of L. We call this the poles separation condition and

will tacitly assume this in almost all cases. For the coeﬃcients array

(0 ≤ n ≤ p, 0 ≤ m ≤ q)

( )

{(1 − aj , Aj )}nj=1 ; {(aj , Aj )}pj=n+1

∆= q (1.20)

{(bj , Bj )}m

j=1 ;{(1 − bj , Bj )}j=m+1

Γ(s | ∆) (1.21)

( )

{(1 − aj , Aj )}nj=1 ; {(aj , Aj )}pj=n+1

= Γ s q

{(bj , Bj )}m

j=1 ;{(1 − bj , Bj )}j=m+1

∏

m ∏n

Γ(bj + Bj s) Γ(aj − Aj s)

j=1 j=1

= (Aj , Bj > 0)

∏

p ∏

q

Γ(aj + Aj s) Γ(bj − Bj s)

j=n+1 j=m+1

H(z ∆)

( )

(1 − a1 , A1 ), . . . , (1 − an , An ), (an+1 , An+1 ), . . . , (ap , Ap )

= m,n

Hp,q

z

(b1 , B1 ), . . . , (bm , Bm ), (1 − bm+1 , Bm+1 ), . . . , (1 − bq , Bq )

∫

1

= Γ(s | ∆) z −s ds

2πi L

∏

m ∏

n

∫ Γ(bj + Bj s) Γ(aj − Aj s)

1 j=1 j=1

= z −s ds, (1.22)

2πi L ∏ p ∏q

Γ(aj + Aj s) Γ(bj − Bj s)

j=n+1 m+1

where the path L is subject to the poles separation conditions given above.

For L we principally use the following.

(1) L is a deformed Bromwich contour, i.e. an infinite vertical line

extending from γ − i∞ to γ + i∞ with indentation in finite part. We

denote an integral along such an L by

∫ ∫ γ+iX

= lim ,

L X→∞ γ−iX

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 51

∫

which we often denote by (γ)

. In the case

∑

n ∑

m ∑

p ∑

q

c

c= Aj + Bj − Aj − Bj > 0, and | arg(z)| < π

j=1 j=1 j=n+1 j=m+1

2

(2.84)

the integral converges absolutely (absolute convergence sometimes occurs

in the case c = 0 and z > 0, too).

In what follows we sometimes use the symbol to denote the coeﬃcients

array, which simply indicates that the coeﬃcients in capitals are positive:

( ∞

)4

⨿ ( )n

Ω= C × R+ . (2.85)

n=0

The following formulas play the most prominent role in transforming the

H-functions appearing in various contexts.

The following inversion formula is used to transform the H-function into

the form which is listed in the table and will be used without notice.

( )

{(1 − bj , Bj )}m , {(bj , Bj )}q

j=1 j=m+1

Hq,p z

n,m

{(aj , Aj )}nj=1 , {(1 − aj , Aj )}pj=n+1

( )

{(1 − aj , Aj )}n , {(aj , Aj )}p

m,n 1 j=1 j=n+1

= Hp,q . (2.86)

z {(bj , Bj )}m q

j=1 , {(1 − bj , Bj )}j=m+1

( ) ( )

H z Λ + c = z c H z Λ (c ∈ C), (2.87)

( )

1 ( )

H z Λ = C H zC Λ (C ∈ R+ ), (2.88)

C

( )

( ) 1

H z Λ∗ = H Λ , (2.89)

z

where for

( )

{(aj , Aj )}nj=1 ;{(aj , Aj )}pj=n+1

Λ= q ∈Ω

{(bj , Bj )}m

j=1 ;{(bj , Bj )}j=m+1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 52

and c ∈ C,

( )

{(aj + Aj c, Aj )}nj=1 ;{(aj + Aj c, Aj )}pj=n+1

Λ+c= q ∈ Ω, (2.90)

{(bj + Bj c, Bj )}m

j=1 ;{(bj + Bj c, Bj )}j=m+1

and for C ∈ R+ ,

( )

{(aj , CAj )}nj=1 ;{(aj , CAj )}pj=n+1

CΛ = q ∈ Ω, (2.91)

{(bj , CBj )}m

j=1 ;{(bj , CBj )}j=m+1

and

( q )

{(1 − bj , Bj )}m

j=1 ;{(1 − bj , Bj )}j=m+1

Λ∗ = ∈ Ω. (2.92)

{(1 − aj , Aj )}nj=1 ;{(1 − aj , Aj )}pj=n+1

( )

(a1 , 1 ), . . . , (an , 1 ), (an+1 , 1 ), . . . , (ap , 1 )

m,n

Hp,q z C C C C (2.93)

(b1 , C1 ), . . . , (bm , C1 ), (bm+1 , C1 ), . . . , (bq , C1 )

( )

C a1 , . . . , an , an+1 , . . . , ap

= C Gm,n

p,q z b1 , . . . , bm , bm+1 , . . . , bq (C > 0).

Proof follows from the following relations between the Gamma factors:

Γ(s | CΛ) = Γ(Cs | Λ) (C ∈ R+ ),

Γ(s | Λ∗ ) = Γ(−s | Λ).

( ′ ′ )

′ {(a′j , A′j )}nj=1 ;{(a′j , A′j )}pj=n′ +1

Λ = ′ ′

{(b′j , Bj′ )}m ′ ′ q

j=1 ;{(bj , Bj )}j=m′ +1

of Ω, we define

Λ ⊕ Λ′ (2.94)

( ′ ′ )

{(aj , Aj )}nj=1 ,{(a′j , A′j )}nj=1 ; {(aj , Aj )}pj=n+1 ,{(a′j , A′j )}pj=n′ +1

= ′ ′ m′ q ′ ′ q

′ .

{(bj , Bj )}mj=1 ,{(bj , Bj )}j=1 ; {(bj , Bj )}j=m+1 ,{(bj , Bj )}j=m′ +1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 53

Proposition 2.4 Suppose that there exists a path L parallel to the imag-

inary axis, such that

∫

1

H(z | Λ) = Γ(s | Λ) z −s ds,

2πi L

∫

′ 1

H(z | Λ ) = Γ(s | Λ′ ) z −s ds,

2πi L

∫

1

H(z | Λ ⊕ Λ′ ) = Γ(s | Λ ⊕ Λ′ ) z −s ds.

2πi L

Then we have

∫ ( ) ( )

∞ ( ) 1 ′ dt z

H tz Λ H Λ =H ′ Λ ⊕ Λ′ . (2.95)

tz ′ t z

0

Proof. As we have

∫ ∞(∫ ∞ ( ) )

( ) 1 ′ dt s−1

H tz Λ H Λ z dz

tz ′ t

0 0

∫ ∞(∫ ∞ ) ( )

( ) s−1 1 ′ dt

=

H tz Λ z dz H Λ

0 0 tz ′ t

∫ ∞ ∫ ∞

( ) s−1 ( ′ ) s−1

H x Λ x dx · H y Λ y dy · z ′

s

=

(

0

( 0

(

= Γ s Λ) Γ s Λ′ ) z ′ = Γ s Λ ⊕ Λ′ ) z ′ ,

s s

( ( ))

(c, C); −

H zΛ ⊕

− ;(c, C)

( ( ))

− ;(c, C) (2.96)

= H z Λ ⊕

(c, C); −

( )

= H z Λ .

We now state the duplication formula and the reciprocity formula which

are most often used in practice.

For

( )

{(aj , Aj )}nj=1 ;{(aj , Aj )}pj=n+1

Λ= q ∈Ω

{(bj , Bj )}m

j=1 ;{(bj , Bj )}j=m+1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 54

and

( )

a a +1 a a +1

′ {( 2j , Aj ), ( j2 , Aj )}nj=1 ;{( 2j , Aj ), ( j2 , Aj )}pj=n+1

Λ = b b +1 bj bj +1 q ∈Ω

{( 2j , Bj ), ( j2 , Bj )}m

j=1 ;{( 2 , Bj ), ( 2 , Bj )}j=m+1

∑

p ∑

q

( ) √ aj − bj +m−(p−n)

H z Λ′ ⊕ Ξ = ( π)m+n−(q−m)−(p−n) 2j=1 j=1

∑

p ∑

q

− Aj + Bj

× H 4 j=1 j=1

z 2Λ ⊕ Ξ. (2.97)

In particular, we have

( )

H z Λ′

∑

p ∑

q ∑

p ∑

q

aj − bj − p−q − Aj + Bj

z 2Λ

p+q 2

= (2π)m+n− 2 2 j=1 j=1

H 4 j=1 j=1

(2.98)

∑

p ∑

q ∑

p ∑

q

aj − bj − p−q

2 −1 − Aj + Bj √

z Λ.

p+q

= (2π)m+n− 2 2 j=1 j=1

H 2 j=1 j=1

(

{( aj , A ), ( aj +1 , A )}n , {(a′ , A′ )}n′ ,

2m+m′ ,2n+n′ j j j=1

H2p+p′ ,2q+q′ z b2j 2

′

j j j=1

′ m′

{( 2 , Bj ), ( bj2+1 , Bj )}m

j=1 , {(bj , Bj )}j=1 ,

′ )

{( 2j , Aj ), ( j2 , Aj )}pj=n+1 , {(a′j , A′j )}pj=n′ +1

a a +1

′

{( 2j , Bj ), ( j2 , Bj )}qj=m+1 , {(b′j , Bj′ )}qj=m′ +1

b b +1

∑

p ∑

q

√ aj − bj +m−(p−n)

= ( π)m+n−(q−m)−(p−n) 2j=1 j=1

(2.99)

∑

p ∑

q

Bj {(a , 2A )}n , {(a′ , A′ )}n ,

′

− Aj +

m+m′ ,n+n′ j j j=1

× Hp+p ′ ,q+q ′ 4 j=1 j=1

z j j j=1

{(bj , 2Bj )}m , {(b′ , B ′ )}m′ ,

j=1 j j j=1

p ′ ′ p′

{(aj , 2Aj )}j=n+1 , {(aj , Aj )}j=n′ +1

′ .

{(bj , 2Bj )}qj=m+1 , {(b′j , Bj′ )}qj=m′ +1

Theorem 2.3 (concentrating

) ( on one) group, say B(m) in Mnemonics. I.e., if

b b +1

the elements 2j , Bj , j2 , Bj contract to (bj , 2Bj ) with the new factor

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 55

∑m

2m− bj

π m/2 , then

j=1

( )

−,

′ ′

−,{(a′j , A′j )}nj=1 ,{(a′j , A′j )}pj=n′ +1

2m+m′ ,n′

Hp′ ,q′ z bj ′ m′

′

{( , Bj ), ( bj +1 , Bj )}m ′ ′ ′ q

j=1 ,−,{(bj , Bj )}j=1 ,{(bj , Bj )}j=m′ +1

2 2

∑

m

√ m− bj

= ( π)m 2 j=1 (2.100)

( ∑ )

Bj

m ′

′

′ ′ −, −,{(a′j , A′j )}nj=1 ,{(a′j , A′j )}pj=n′ +1

× Hpm+m

′ ,q ′

,n

4j=1

z ′ m′

′ .

{(bj , 2Bj )}m ′ ′ ′ q

j=1 ,−,{(bj , Bj )}j=1 ,{(bj , Bj )}j=m′ +1

Theorem 2.4 (Reciprocity formula) The reciprocity formula for the

gamma function and Euler’s identity lead to

( ( ))

−;(c, C)

H zΛ ⊕

−;(c, C)

(2.101)

1 { cπi ( −Cπi ) −cπi

( Cπi )}

= e H e z Λ −e H e z Λ .

2πi

In the traditional notation, (2.101) reads

( )

{(aj , Aj )}nj=1 , {(aj , Aj )}pj=n+1 , (c, C)

m,n

Hp+1,q+1 z q

{(bj , Bj )}m

j=1 , {(bj , Bj )}j=m+1 , (c, C)

{ ( )

1 {(aj , Aj )}nj=1 , {(aj , Aj )}pj=n+1

= ecπi Hp,q

m,n

e−Cπi z q (2.102)

2πi {(bj , Bj )}mj=1 , {(bj , Bj )}j=m+1

( )}

−cπi

p

Cπi {(aj , Aj )}j=1 , {(aj , Aj )}j=n+1

n

−e m,n

Hp,q e z q ,

j=1 , {(bj , Bj )}j=m+1

{(bj , Bj )}m

( )

a , . . . , an , an+1 , . . . , ap , c

Gp+1,q+1 z 1

m,n

b1 , . . . , bm , bm+1 , . . . , bq , c

{ ( )

1 cπi m,n

−πi a1 , . . . , an , an+1 , . . . , ap

= e Gp,q e z (2.103)

2πi b1 , . . . , bm , bm+1 , . . . , bq

( )}

πi a1 , . . . , an , an+1 , . . . , ap

− e−cπi Gm,n p,q e z b1 , . . . , bm , bm+1 , . . . , bq .

( ( )) ( ( ))

−;(c + 1, C) −;(c, C)

H z Λ ⊕ = −H z Λ ⊕ , (2.104)

−;(c + 1, C) −;(c, C)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 56

( )

{(aj , Aj )}nj=1 , {(aj , Aj )}pj=n+1 , (c + 1, C)

m,n

Hp+1,q+1 z q

{(bj , Bj )}m j=1 , {(bj , Bj )}j=m+1 , (c + 1, C)

( ) (2.105)

{(aj , Aj )}nj=1 , {(aj , Aj )}pj=n+1 , (c, C)

m,n

= −Hp+1,q+1 z q ,

{(bj , Bj )}mj=1 , {(bj , Bj )}j=m+1 , (c, C)

The gamma transform (or the Euler gamma integral) (1.42) is a special

case of

( )

−

Γ(a + s) : G0,1 z

1,0

= z a e−z . (2.106)

a

1

z b (1 − z)a−b−1 , |z| < 1

( )

Γ(a − b)

a

G1,0 = 1

1,1 z (2.107)

b

z = 1, a = b + 1

2

0, |z| > 1.

The beta transform (or the beta integral) (4.2) is a special case of

( )

a

G1,1 = Γ(1 − a + b) z b (z + 1)a−b−1 .

1,1 z

b

a companion to (2.31), reads

( )

− ( √ )

G2,0 z 1

2 (a+b) K

0,2 a, b = 2 z a−b 2 z .

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 57

( )

− √ a −2 √z

G0,2 z

2,0

1 = πz e .

a, a + 2

(2.108) and (2.109) are the only cases of two gamma factors that have

been essentially used in converting the integrals into the Riesz sum and the

modified K-Bessel function, respectively.

A companion to (2.24) is

( )

− ( √ )

G1,0 z 1

2 (a+b) J

0,2 a, b = z a−b 2 z (2.110)

( )

− za ( √ )

G1,0

0,2 z a, a + 1 = √π cos 2 z ,

2

( )

− za ( √ )

1,0

G0,2 z = √ sin 2 z .

1

a + ,a 2 π

( ) ( )

1−a Γ(b) a Γ(b)

G1,2 z

1,1

= 1 F1 ; −z = Φ(a, b; −z). (2.111)

0, 1 − b Γ(a) b Γ(a)

table of formulas for G-functions in [ErdH, pp.216-222] but they do appear

as integral representations (Mellin-Barnes integrals, or the G-functions),

[ErdH, (4), p.256] and [ErdH, (6), p.256], respectively.

(2.111) as an integral representation is essentially used in the proof of

Theorem 4.11.

Exercise 2.12 Show that (2.111) may be put into another form

( ) ( )

a b Γ(1 − a + b) 1−a+b

1,1

G1,2 z =z 1 F1 ; −z . (2.112)

b, c Γ(1 − c + b) 1−c+b

( α =1−a+

Solution Writing ) b, β = 1 − c + b, we may write the left-hand

1−α+c

side as G1,2 z

1,1

. Hence by (2.87), (2.112) follows.

1 − β + c, c

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 58

( )

1 ( )

1,1

G1,2 z = Γ(a) − Γ a, z . (2.113)

a, 0

( )

a

2,0

G1,2 z = z 2 (b+c−1) e− 2 z Wκ,µ (z),

1 1

(2.114)

b, c

µ = 21 (b − c) in Example 2.4 and [ErdH, (8), p.216] reads

( )

a

2,1

G1,2 z 1 1

= Γ(1 − a + b) Γ(1 − a + c) z 2 (b+c−1) e 2 z Wκ,µ (z), (2.115)

b, c

2 (b − c).

1

1

Wκ,µ (z) = e− 2 z z 2 c U (a, c; z),

1 1

a= − κ + µ, c = 2µ + 1, (2.116)

2

(2.114) is equi-vocal to

( )

a ( )

2,0

G1,2 z = e−z z b U a − c, b − c + 1, z , (2.117)

b, c

introduced in Example 2.3 while (2.115), to

( )

a

G2,1 z (2.118)

1,2 b, c

( )

= Γ(1 − a + b) Γ(1 − a + c) z b U b − a + 1, b − c + 1, z .

In view of

to

( )

1 ( )

G2,0 z

1,2 a, 0 = Γ a, z , (2.120)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 59

( )

a ( ) z b

G2,1

1,2 z a, b = Γ(1 − a + b) Γ a − b, z e z . (2.121)

and (2.122) in their specifications:

which is [ErdH, (22), p.266]. We also note

which is [PBM, 7.11.4.11, p.584].

(2.115) also reduces when b = a to

( )

1−a 1

G1,2 2z

− a− 1

2,1

= √ Γ(a)Γ(1 − a)ez (2z) ( 2 ) Ka− 12 (z) (2.124)

0, 1 − 2a π

in view of the relation [ErdH, (13), p.265]

( )12

2z

W0,µ (2z) = Kµ (z). (2.125)

π

For G3,1

1,3 , cf. (2.8) above. Other formulas will be given at their occur-

rences, cf. especially, Chapters 6, 8 and 9.

November 19, 2014 14:27 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 60

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 61

Chapter 3

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

M,0 Ñ ,0

3.1 The H0,M ↔ H0, Ñ

formula

functional equation with r a real number,

∏M

Γ(dj + Dj s) φ(s), Re(s) > σφ

j=1

χ(s) = (3.1)

∏Ñ

Γ(ej + Ej (r − s)) ψ(r − s), Re(s) < r − σψ

j=1

Dj , Ej > 0.

such that they squeeze a compact set S with boundary C for which sk ∈

S (1 ≤ k ≤ L)

Under these conditions the χ-function, key-function, X(s, z | ∆) is usu-

ally defined in our theory as

∫

1

X(s, z | ∆) = Γ(w − s | ∆)χ(w) z −(w−s) dw (3.2)

2πi L1 (s)

consider the case where there is no processing gamma factor, Γ(s | ∆) = 1

and we write X(s, z) = X(s, z | ∅).

Then we have the following modular relation.

61

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 62

Theorem 3.1

X(s, z) (3.3)

∞ ( )

∑ αk M,0 −

s H0, M zλk

λk {(dj + Dj s, Dj )}M

k=1 j=1

if L1 (s) can be taken to the right of σφ ,

( )

= ∑ ∞ −

βk Ñ ,0 µk

H

µk r−s 0,Ñ z {(ej + Ej (r − s), Ej )}Ñ

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

k=1 j=1

( )

∑ L

+ Res χ(w) z s−w , w = sk

k=1

if L2 (s) can be taken to the left of r − σψ

equivalence is proved of (3.4) to the modular relation ([BeI, Theorem 1]),

∞

∑ ∞

∑

αk Em (λk x) = x−r βk Em (µk /x) + P(x), (3.5)

k=1 k=1

∫ −

1

m,0

Em (x) = Γ (s)x ds = G0,m z 0, . . . , 0 (x > 0),

m −s

(3.6)

2πi (c) | {z }

m

1 ∑′

αk (x − λk )κ

Γ(κ + 1)

λk ≤x

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 63

∫ ( )−s

zκ Γ(s) Γm−1 (s) z2

Kκ (z; r, m) = ds (3.7)

2πi L Γ(κ + 1 − s) Γ m−1 (r − s) 22m

( z )2 −

m,0

= z κ H0,m (0, 1), . . . , (0, 1), (−κ, 1), (1 − r, 1), . . . , (1 − r, 1) ,

2m | {z } | {z }

m m

tions given below. Cf. (9.54) in Chapter 10.

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

in 1949 in [Bel2], [Bel3]. In [Bel4, (1b)], he refers to the Siegel integral,

which has been introduced by Wishart [Wis] (cf. [Sa] for more details).

In [Bel5] he further refers to the most general Steen’s function V =

V (x; a1 , . . . , an ) ([St]) as:

∫ ∞

1 dx

xs V (x; a1 , . . . , an ) = Γ(s + a1 ) · · · Γ(s + an ). (3.8)

2πi 0 x

( )

−

V (x; a1 , . . . , an ) = Gn,0

0,n x (3.9)

a1 , . . . , an

∞

∑ −

αk

Gm,0 zλk s, . . . , s (3.10)

k=1

λsk 0,m | {z }

m

∑ ∞ − ∑ ( )

βk µ

m,0 k

L

= G r − s, . . . , r − s + Res χ(w) z s−w

, w = s ,

µr−s 0,m z |

k

{z }

k=1 k k=1

m

For curiosity’s sake, we remark that in Chaudhry and Zubair [CZ] the

most far-reaching formula for G-functions seems to be [CZ, (4.14), p.197]

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 64

γν (α, x; b) + Γν (α, x; b)

= 2α−2 π −1 b 2

1

( 2 ( ) ( ) ( ) ( ))

b 1 1 1 1 1 1 1 1

× G4,0 ν + , − ν + , α + , α − ,

0,4

16 2 2 2 2 2 2 2 2

where

( )1/2∫ x ( )

2b α− 32 1

γν (α, x; b) = t exp(−t)Kν+ 12 dt.

π 0 t

We now refer to various generalized Bessel functions which have

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

hitherto been introduced by many authors and are integrals of the gamma

quotient G(s) (cf. the passage after Table 2.2). Most of them contains the

Γ(s)

Riesz sum factor Γ(s+κ+1) for the sake of enhancing the convergence (cf.

§6.1.6 in this regard) as well as the anticipated application to the Riesz

sums (§6.1.1).

Lavrik [Lav5] introduces (with κ = 0)

JN (ω, z, κ)

( )

(κ + 1, 1)

N +1,0

= H1,N

+1 z

(0, 1), (α1 ω + β1 , α1 ), . . . , (αN ω + βN , αN )

(3.11)

∫ ∏

N

1 Γ(s)

= Γ(αj s + αj ω + βj ) z −s ds,

2πi L Γ(κ + 1 + s) j=1

∫

z κ+r Γ(r − s)G(r − s) −s

fκ (z) = fκ,G (z) = z ds

2πi L Γ(κ + 1 + r − s)

∏

N

∫ Γ(r − s) Γ(αj s + βj )

z κ+r j=1

= z −s ds (3.12)

2πi L ∏

N

Γ(κ + 1 + r − s) Γ(αj r + βj − αj s)

j=1

( )

(1 − r, 1)

=z H1,N +1 z

κ+r N,1

,

(β1 , α1 ), . . . , (−κ − r, 1), (1 − (α1 r + β1 ), α1 ), . . .

where

Γ(β1 + α1 r − α1 s) · · · Γ(βN + αN r − αN s)

G(s) = (3.13)

Γ(β1 + α1 s) · · · Γ(βN + αN s)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 65

tation in finite part and may be thought of as a Bromwich path satisfying

the poles separation condition. He shows that the function fκ shares anal-

ogous properties with the J Bessel function.

A. Z. Walfisz, in connection with the m-dimensional Piltz divisor prob-

lem in algebraic number fields ([AZW2]) introduced the function

∫ ( )mκs−1

1 Γ(s) z

Jρ1 ,ρ2 ,m (z) = G(s)m ds (3.14)

2πi ( 34 ) Γ(s + 1) mκ2r1 /κ

( ( ) ( ))

z 1 − 1 , 1 , . . . , (1 − r, 1), (1, 1), . . . , 0, 1

1,ρ1 +ρ2

= Hρ1 +ρ2 +1,1 2 2 2

mκ2r1 /κ (0, 1)

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

special case of (3.13) with αj = 12 , βj = 2r , 1 ≤ j ≤ r1 , etc.). Noting that

Hafner’s generalized Bessel function may be written as

fκ (z) (3.15)

( )

1 (1 − (α1 r + β1 ), α1 ), (κ + 1, 1), . . . (β1 , α1 ), . . .

= z κ HN

1,N

,

+1,1

z (0, 1)

Walfisz’s function is a special case of that of Hafner’s. Walfisz’s function

Jr1 ,r2 (z) = Jρ1 ,ρ2 ,1 (z) with m = 1 appears in the study of the Idealfunktion

which was studied by him in [AZW1] mainly as a generalization of, as well as

a sequel to, Hardy’s Ω-results paper [HarDP]. We record Walfisz’s function

[AZW1, (3.12), p.31]

∫ ( z )κs−1

1 Γ(s)

Jr1 ,r2 (z) = G(s) ds (3.16)

2πi ( 34 ) Γ(s + 1) τ

with τ = κ2r1 /κ . We state the special case J2,0 which appears in the case

of a real quadratic field and is a relative of Koshlyakov’s L-function (cf.

(3.178)):

∫ ( z )2s−1

1 1

J2,0 (z) = G(s) ds

2πi ( 34 ) s 4

(( ) ( ) ( ) ( ) ( ) ) (3.17)

4 1 − 12 , 12 , 1 − 12 , 12 , 1, 21 , 0, 21

2

2 0,2

= H4,0 .

z z −

(π )

J2,0 (z) = − K1 (z) + Y1 (z) (3.18)

2

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 66

that in exactly the same way, we may deduce A. A. Walfisz’s result [AAW4,

(17)] (z > 0)

√ (π √ √ )

J˜2,0 (z) = − z K1 (4 z) + Y1 (4 z) , (3.19)

2

where

∫

1 1

J˜2,0 (z) = G(s) z s ds (3.20)

2πi (c) s

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

(( ) )

η−5/2 z 2

with c = H . Since J2,0 (z) = z4 J˜2,0 4 , (3.18) follows again.

1,0 1,0

3.1.1 The H0,1 ↔ H0,1 formula

The special case of Theorem 3.1 with M = Ñ = 1 includes the rational and

imaginary quadratic cases of Theorem 3.12:

X(s, z | ∆) (3.22)

( )

∑∞ −

αk 1,0

H zλ

λ s 0,1 k

(d1 + D1 s, D1 )

k=1 k

if L1 (s) can be taken to the right of max (σφh )

1≤h≤H

( )

= ∑ βk

∞ −

1,0 µ k

r−s H0,1

µ z (e1 + E1 (r − s), E1 )

k=1

k

∑

L ( )

+

Res χ(w) z s−w , w = sk

k=1

if L2 (s) can be taken to the left of min (r − σψi )

1≤i≤I

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 67

or

d1 +D1 s

∞

∑ 1

z D1 d1

αk λk D1 e−(zλk )

D1

(3.23)

D1

k=1

( ) e1 +EE1 (r−s) ∞ ( )

z −1 1 ∑ e1 µk

1 ∑

L

= βk µk E1 e−( z ) E1 + Res χ(w) z s−w , w = sk .

E1

k=1 k=1

1 ∑ βk ( µk ) 1 E11

∞ ∞

1 ∑ αk 1 e +E (r−s) 1

d1 +D1 s µk E

−(zλk ) D1

e− ( z )

1

s (zλk ) D1

e = r−s

D1 λk E1 µ z

k=1 k=1 k

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

∑

L ( )

+ Res χ(w) z s−w , w = sk . (3.24)

k=1

the use of the gamma transform (1.42).

Remark 3.1 It may be remarked that the K-Bessel function Ks (z) de-

fined by (2.31) and its reduction to exponential functions (2.32) and (2.33)

appeared in various contexts in relation to the lattice point problem as will

be enunciated in §4.7. It was [HarMB] who first used the integral (2.31),

without noticing that it is a K-Bessel function, in his research on the Ep-

stein zeta-function. Bellman [Bel3, (3)] already notices the relevance of

Hardy’s paper in the context of Maass forms.

(2.33) with plus sign while [Sieg1] is (2.33) with minus sign. We note

the following chain of diﬀerent symbols for the K-Bessel function, the

first equality being [Bel5, (4.1)]

2

Va (x, y) = x2a−1 Va (|xy|), Va (|xy|) = π a W2−a ((πxy) ) (3.25)

and

∫ ∞ √

e−zv− v v −a dv = 2z

1 a−1

Wa (z) = 2 Ka−1 (2 z), Re z > 0. (3.26)

0

(ii) Siegel’s proof of Hamburger’s theorem. We notice that the main in-

gredient of Siegel’s proof of Hamburger’s theorem is indeed the partial

fraction expansion for the cotangent function in Example 4.2:

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 68

( ) ( )

1 1

Γ s φ(s) = Γ (r − s) ψ(r − s) (3.27)

2 2

with a simple pole at s = r > 0 with residue ρ is equivalent to

( )∑ ∞

1 αk

2Γ s 1

2 2 2 2s

k=1 (λk + z )

∑∞ ( )

βk 1

s φ(0) z −s

1

(r−s)

= 4z 2 1 K 12 (r−s) (2µk z) + 2Γ

(r−s) 2

k=1 µk

2

( )

1

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

+ ρΓ (s − r) z r−s . (3.28)

2

This reduces for s = 2, r = 1 to the partial fraction expansion for a

generalized cotangent function

∞

∑ ∞

αk √ −1 ∑ √

2 2 2

= 2 πz βk e−2µk z + 2φ(0)z −2 + ρ πz −1 , (3.29)

λk + z

k=1 k=1

Remarkably enough, Siegel deduces (3.29) not as the Fourier-Bessel

expansion but as a consequence of the modular relation, a special case

of (3.23):

∞

∑ ∞

∑ µk 2

αk e−(zλk ) = 2z −(r−s) βk e−(

2

2z 2 z ) + P(z). (3.30)

k=1 k=1

[Sieg1, l.2, p.156]

∞

∑ ∞

2 ∑ πn2

αk e−πn βk e− x + P(x),

2

2z 2 x

=√ x > 0. (3.31)

x

k=1 k=1

x

1,0

thereby incorporating (2.33) with minus sign K− 21 , we uplift H0,1 ↔

1,0 1,1 2,0

H0,1 to H1,1 ↔ H0,2 .

(iii) Chowla and Selberg leave the function undefined, denoting it e.g. by

∫ ∞ ( √ )

√

− π 2a

∆n

(y+y −1 ) s− 32 π ∆n

Ia = e y dy = 2Ks− 12 . (3.32)

0 a

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 69

0,2 ↔ G0,2 formula

Anticipating an application to (3.114) (where φ is defined by (3.110)) and

(3.141) we state the special case of (9.6) as

∞

∑ ( )

αk −

G2,0 zλk

λsk 0,2 s + ν2 , s − ν2

k=1

∑ ∞ ( )

βk µk −

= G2,0

z r − s + ν2 , r − s −

r−s 0,2 ν (3.33)

k=1 k

µ 2

∑ ( )

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

L

+ Res χ(w) z s−w , w = sk .

k=1

∞

∑ ( √ )

2 αk Kν 2 zλk

k=1

∞

∑ ( √ ) ∑L ( )

(3.34)

µk

= 2z −r βk Kν 2 + Res χ(w) z −w , w = sk .

z

k=1 k=1

and the functional equation (9.1) are equivalent, which is the most relevant

to our work. This also follows from (3.5) in view of

( )

− ( √ )

E2 (z) = G2,0

z = 2 K0 2 z .

0,2

0, 0

There are two ways leading to the rephrased Ramanujan formula (3.87).

In the beginning all appealed to the use of the special diﬀerence of the shift

of the argument, i.e. the odd integer value shift, which brings in mind the

reduction of the product of two gamma factors into a single gamma factor,

thus amounting to the modular relation of Bochner. It seems Guinand who

suggested a more general reasoning based on the inverse Heaviside integral.

By this, the functional equation is transformed into the K-Bessel series and

vice versa. Then taking into account the special feature that the shift is a

half-integer, we are led to the degeneracy (2.32) of the Bessel function into

the exponential function. Cf. §3.3.3 below.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 70

Q] = κ = r1 + 2r2 (in standard notation; r1 and 2r2 indicates the number

of real and imaginary conjugates of Ω, respectively) and discriminant ∆.

For κ ≤ 2, cf. Let

2r2 π κ/2

A= √ (3.35)

|∆|

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

and let r = r1 + r2 − 1 denote the rank of the unit group. The Dedekind

zeta-function ζΩ (s) of the number field Ω is defined by

∑ ∞

∑ F (k)

1

ζΩ (s) = s = (3.36)

(Na) ks

0̸=a⊂Ω k=1

F (k) = FΩ (k) indicates the number of ideals of norm k, sometimes called

“Idealfunction”. Hence αk = βk = F (k) and λk = µk = Ak, r = 1. ζΩ (s)

satisfies the functional equation

(s)

A−s Γr1 Γr2 (s) ζΩ (s)

2

( ) (3.37)

−(1−s) r1 1 − s

=A Γ Γr2 (1 − s) ζΩ (1 − s),

2

residue (cf. (3.152) for κ ≤ 2)

2r+1 π r2 Rh

ρ = λh = √ . (3.38)

w |∆|

number and the regulator, respectively.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 71

symbol meaning

κ = r1 + 2r2 degree of Ω

r1 number of real conjugates of Ω

2r2 number of imaginary conjugates of Ω

r = r1 + r2 − 1 rank of the unit group of Ω

∆ discriminant of Ω

h class number

R regulator

w number of roots of unity in Ω

ζΩ (s) Dedekind zeta-function of Ω

F (k) = FΩ (k) number of ideals whose norm is k

2 π κ/2

2r√

A= coeﬃcient of functional equation for ζΩ (s)

|∆|

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

2r+1√π r2 Rh

λh = residue at s = 1 of ζΩ (s)

w |∆|

(s)

χ(s) = Γr1 Γr2 (s)φ(s), (3.39)

2

where

∞

∑ α(k)

φ(s) = A−s ζΩ (s) = s. (3.40)

(Ak)

k=1

(1 )

χ(s) = Γr1 s Γr2 (s)φ(s)

2

(1 1 )

= Γ r1 − s Γr2 (1 − s)φ(1 − s) (3.41)

2 2

= χ(1 − s)

which will not be relevant to us).

(3.41) is a special case of (3.1) with dj = 0, ej = 0, Dj = 1 (1 ≤ j ≤ r1 ),

Dj = 12 (r1 + 1 ≤ j ≤ r1 + r2 = M ) and Dj = 1 (1 ≤ j ≤ r1 ), Ej =

2 (r1 + 1 ≤ j ≤ r1 + r2 = Ñ = M ) and r = 1.

1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 72

Theorem 3.3

∞

∑ ( )

αk −

H M,0

zλ

k

s

(Ak) 0, M {( 2s , 12 )}rj=1

1

, {(s, 1)}rj=r

2

1 +1

k=1

∞

∑ ( )

αk µk −

= H Ñ ,0

(Ak)

s 1−s 0,Ñ z {( 1−s

2 , 2 )}j=1 , {(1 − s, 1)}j=r1 +1

1 r1 r2

k=1

∑

L ( )

+ Res χ(w) z s−w , w = sk . (3.42)

k=1

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

( )

−

be H0, κ x

κ,0

, i.e.

{(0, 12 )}rj=1

1

, {(0, 1)}rj=r

2

1 +1

∫ (s)

1

Xr1 ,r2 (x) = Γ r1 Γr2 (s)x−s ds, c > 0, (3.43)

2πi (c) 2

for x > 0 (eventually Re(x) > 0 being allowed), which proviso will remain

true in what follows.

∞

∑ ∞

∑

FΩ (k)Xr1 ,r2 (Akρ) = P(x) + FΩ (k)Xr1 ,r2 (Akρ−1 ), (3.44)

k=1 k=1

( )

Rj = Res χ(s)z −s , s = j , j = 0, 1. (3.45)

whence Ω is either the rational or a quadratic field. Corollary 3.1 will be

made more explicit as Theorem 3.12 below.

In what follows we shall state special cases of the following theorem, which

in turn is a special case of Theorem 3.2. Cf. (3.89), (3.129) and (3.129)

below.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 73

equation

1,0 1,0

we have the Bochner modular relation H0,1 ↔ H0,1

∞

∑ ( )r+ Cc ∑ ∞ 1

1

1 µk C

−(zλ

βk µk C e−( z )

c c c

) C

zC αk λk C e k

=

z

k=1 k=1

(3.47)

∑

L

( )

+C Res χ(w) z −w , w = sk .

k=1

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

First we follow [Win2] to introduce the Lambert series. Let {an } ⊂ C be

such that

1/n

lim sup |an | ≤ 1, (3.48)

∑∞

i.e. such that the power series n=1 an z n is absolutely convergent in |z| <

1. Then the Lambert series

∞

∑ zn

f (z) = an (3.49)

n=1

1 − zn

and moreover the power series of this function can be obtained by formal

rearrangement of (3.49), i.e.

∞

∑

f (z) = bn z n (|z| < 1), (3.50)

n=1

where

∑

bn = ad . (3.51)

d|n

∞

∑ zn

f+ (z) := an . (3.52)

n=1

1 + zn

Nova, §40 [Ja], consists of two parts, Fragment I and Fragment II, the

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 74

latter of which contains only formulas and almost no text. Dedekind, en-

trusted with editing and publishing the paper, succeeded in elucidating the

genesis of all the formulas in Fragment II by introducing the most cele-

brated Dedekind eta-function defined by (3.92) below. All the results in

Fragment II deal with the asymptotic behavior of those modular functions

from Jacobi’s Fundamenta Nova, for which the variable tends to rational

points on the unit circle. After Dedekind, several authors including Smith

[Sm], Hardy [HarLam], Rademacher [Ra1] made some more incorporations

of Fragment II. In 2004, Arias-de-Reyna [AdR] analyzed all the formulas

in Fragment II again by applying ad hoc methods to each of the formu-

las. N. Wang [NLW, Theorem 3], referring to the more informative paper

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

of Fragment I and some far-reaching comments on Fragment II but was

not mentioned by Arias-de-Reyna, chose a suitable R-function in taking

the radial limits and applied a form of Dirichlet-Abel Theorem, whereby

he almost immediately got the expression in terms of the (diﬀerences of)

polylogarithm function of order 1, without singularity. This is precisely

what Riemann intended to do, i.e. to eliminate the singular part, which

turns out to be the Clausen function, by taking the odd part, but could not

because of lack of time, and thought it could have been within reach of Rie-

mann (at least in his mind, Riemann must have felt the truth of this type

of Dirichlet-Abel theorem), those formulas were simply noted for record by

Riemann.

Riemann’s procedure in Fragment I. Riemann divides the Lambert series

by z and integrate the result from 0 to z = reiθ , r = |z| < 1, and then

puts r = 1 to obtain his results, i.e. “Fourier expansions”, in a formal

way. Wintner’s legitimation reads as follows. Slightly modifying Riemann’s

procedure by the trivial factor eiθ , which makes the function F = F (r, θ) a

function in z, one obtains

∫ r ∑∞

f (reiθ ) bn n

F (reiθ ) = dr = (reiθ ) ,

0 r n=0

n

or

∫ r ∑∞

f (reiθ )

F (z) = dr = cn z n , (3.53)

0 r n=0

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 75

with

1∑

cn = an . (3.54)

n

d|n

1∑ n

cn = (−1) an . (3.55)

n

d|n

as r → 1 (within the Stolz path) and if so, then whether or not the bound-

ary function is of class Lp , so that Riemann’s formal trigonometric series

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

Use is made not only of the celebrated Riesz-Fischer condition [Vi2,

Theorem 7.1, p. 139]

∞

∑ 2

|cn | < ∞ (3.56)

n=−∞

∞

∑ ε 2

|n| |cn | < ∞, (3.57)

n=−∞

∞

∑ p/(p−1)

|cn | <∞ (p ≥ 2) (3.58)

n=−∞

∑∞

the trigonometric series n=−∞ cn einθ is convergent a.e. (almost every-

where) and represents the values of the function a.e. We refer to this as

the Hausdorﬀ-Paley extension of Fischer-Riesz theorem to be used

again in §6.3.4. The contents of this subsection has a close relationship to

that of §6.3.4.

Theorem 3.5 (Wintner) Suppose

an = O(nλ−δ ) (3.59)

for some δ > 0 and a fixed 0 < λ ≤ 12 . Then the boundary function F (eiθ )

exists and is measurable such that

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 76

along the Stoltz path. If λ < 12 , then F (eiθ ) is of class L1/λ and if in (3.59),

the exponent can be taken arbitrarily small, then it is of class L∞ .

Proof. Recall that the divisor function, which is the special case of the

sum-of-divisors function (3.86) satisfies the estimate

∑

d(n) = 1 = O(nε ) (3.61)

d|n

cn = O(nλ−1−δ ) (3.62)

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

for some δ > 0. Hence if λ < 12 , then the Lp -condition (3.58) is satisfied

and if λ = 12 , then the series for F (reiθ ) is Cauchy in L2 and so there exists

a function F (eiθ ) of class L2 such that

∞

∑

F (eiθ ) ∼ cn einθ . (3.63)

n=−∞

∞

∑

iθ

F (e ) = cn einθ a.e. (3.64)

n=−∞

Example 3.1

n+1

(i) In the case a2n = 0, a2n+1 = 4(−1) , we obtain the Lambert series

∑∞ n+1 2n+1

2K (−1) z

f (z) = =4 + 1, (3.65)

π n=1

1−z 2n+1

∞

∑

2K ℓ 2

f (z) = =4 d4,1 (n)z 2 (4m−1) n + 1, (3.66)

π

ℓ,m,n=0

Hence

∑∞

d4,1 (n) n

F (z) = z , |z| < 1. (3.67)

n=0

n

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 77

∑∞

d4,1 (n) inθ

F (eiθ ) = e a.e. (3.68)

n=1

n

himself ([Lam])

∞

∑ ∑∞

zn

f (z) = = d(n)z n , |z| < 1. (3.69)

n=1

1 − z n

n=1

Hence

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

∑∞

d(n) n

F (z) = z , |z| < 1. (3.70)

n=1

n

∑∞

d(n) inθ

F (eiθ ) = e a.e. (3.71)

n=0

n

Let

1

B̄1 (x) = x − [x] − (3.72)

2

be the first periodic Bernoulli polynomial, where [x] indicates the great-

est integer function. As is well-known, we have the Fourier expansion for

x∈/Z

∞

1 ∑ sin 2πmx

B̄1 (x) = − , (3.73)

π m=1 m

Wintner [Win1], [Win2], we denote the Fourier series (3.73) by ψ(x). Then

∞

∑

− 1 sin 2πmx

, x∈ /Z

ψ(x) = π m=1 m (3.74)

0, x∈Z

is the sawtooth Fourier series. Indeed, this was also established as one

of the earliest instances of the modular relation by A. Z. Walfisz (6.113).

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 78

Exercise 3.2 Prove that the imaginary part of (3.71) gives the expan-

sion treated in [Win1], which in turn gives rise to the series considered by

Riemann [Rie1].

Indeed, this was proved by Chowla and Walfisz [CW]. Cf. (6.122) and

(6.123) and §6.3.4.

We note that the estimate (3.61) gives rise to one for the sums over all

divisors including the estimate for the sum-of-divisors function in the proof

of Corollary 6.1 and (6.164).

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

For a Dirichlet character χ modulo M , let Lχ (t) define the Lambert series

associated to χ:

∞

∑

Lχ (t) = χ(n)e−nt , Re t > 0, (3.75)

n=1

Let N be a multiple of M , say N = uM and let r be a positive integer

prime to N . The essential case is u < r, which we so assume. Then Szmidt,

Urbanowicz and Zagier [SUZ, p. 275] deduced the elegant formula

∑ χ̄(r) −N t ∑

χ(n)e−rnt = Lχ (rt) − e ψ(−N )Lχψ (t), (3.76)

φ(r)

0<n< N

r

ψ mod r

Exercise 3.3 Deduce (3.76).

∑

Solution We divide the sum (3.75) into two parts: S1 = and

∑ 0<n<N/r

S2 = n≥N/r and write rn = m in the latter:

∑

S1 = χ(n)e−rnt = Lχ (t) − S2 .

0<n<N/r

∑

Then the second sum becomes S2 = m≥N χ(mr−1 )e−mt . Writing m =

∑∞ r|m

−(n+N )t

n + N , we obtain S2 = m=1 χ̄(r)χ(m)e .

r|n+N

Finally, we use the orthogonality relation

1 ∑

ψ(n)ψ(−N ) (3.77)

φ(r)

ψ mod r

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 79

It is rather remarkable that (3.76), though it is just a division of the sum

into two parts, can be used to generate the identity as its Taylor coeﬃcients:

∑

mrm−1 χ(n)nm−1

0<n<N/r

∑ (3.78)

χ̄(r)

= −Bm,χ r m−1

+ ψ̄(−N )Bm,χψ (N ),

φ(r)

ψ mod r

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

(3.78) constitutes a counterpart of Y. Yamamoto’s formula, cf. [Ya].

We prove Theorem 3.6 below, forming Riemann-Hecke-Bochner corre-

spondence (cf. Knopp [MKn2]) with Szmidt, Urbanowicz and Zagier’s for-

mula (3.76), by applying the Mellin transform technique. This aspect is

already stated in [SUZ, p. 275, l. 8] to deduce their formula [SUZ, (5)], in

the spirit of Riemann, splitting up the range of integral.

N

primitive. Let N be a multiple of M and let r > M be a positive inte-

ger prime to N . Then we have the modular relation

∑ χ(n) χ(r) s ∑

= L(s, χ) − r ψ(−N )(L(s, χψ) − Ms (χψ)) (3.79)

ns φ(r)

0<n< N

r

ψ

∑N −1 χ(n)

valid for all values of s, where Ms (χ) = n=1 ns is the weighted complete

character sum.

Proof. Taking the Mellin transform of both sides of (3.76) and using

(2.106), we obtain for σ > 1

∑ χ(n) χ(r) ∑

−s

Γ(s) = Γ(s)r L(s, χ) − Γ(s) ψ(−N )Φ(χψ, N, s),

(rn)s φ(r)

0<n< N

r

ψ

where

∞

∑ χ(n)

Φ(χ, a, s) = s (3.80)

n=0

(n + a)

November 19, 2014 14:27 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 80

ically continued over the whole plane. Hence the formula

∑ χ(n) χ(r) ∑

−s

= r L(s, χ) − ψ(−N )Φ(χψ, N, s)

(rn)s φ(r)

0<n< N

r

ψ

Or more simply, using the expression

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

the functional equation (1.8), we deduce (3.79) for all s ∈ C.

276]), Ms (χ) may be considered as known for s = m ∈ N:

1

Mm (χ) = (Bm+1,χ (M ) − Bm+1,χ )

m

m ( )

1 ∑ m (3.81)

= Bm+1−k,χ M k .

m+1 k

k=1

formula

The following remarkable formula of Ramanujan is probably one of the

most famous ones named after him, which is stated as Entry 21 (i), Chap-

ter 14 of Ramanujan’s Notebook II [Berndt8] (which is also stated as I,

Entry 15 of Chapter 16 [Berndt7], cf. also IV, Entry 20 of [Berndt9]). The

most extensive account of information surrounding Ramanujan’s formula

is [Be11] and [IO] regarding special values of the zeta-functions in ques-

tion. We remark in passing that the intersection of references in these two

excellent survey papers is a null set.

αβ = π 2 (3.82)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 81

{1 ∞

∑ k −2n−1 }

α−n ζ(2n + 1) + (3.83)

2 e2αk − 1

k=1

{1 ∞

∑ k −2n−1 }

= (−β)−n ζ(2n + 1) +

2 e2βk − 1

k=1

∑

n+1

B2j B2n+2−2j

− 22n (−1)j αn+1−j β j ,

j=0

(2j)! (2n + 2 − 2j)!

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

∞

∑

x Bm m

= x , |x| < 2π. (3.84)

ex − 1 m=0 m!

sum as 0. It looks like at a first glance that the case n ≥ 0 counts more,

giving an analytic expression for zeta values at odd positive integers, but

the negative case is equally important in that it expresses the automorphic

property of the corresponding Eisenstein series; the case n = 0 is no less

important as we shall see below.

By the procedure similar to the one given above for treating Lambert

series, we may apply Liouville’s formula (cf. Koshlyakov [KoshI]) to

include the case n = 0.

∞

∑ ∞

∑

ka

= σa (k)e−2πkx (3.85)

e2πkx − 1

k=1 k=1

where

∑

σa (k) = da (3.86)

d|k

divisors of k.

Incorporating (3.85) with x = α/π, in (3.83), we may express Ramanu-

jan’s formula as its rephrased Ramanujan formula

∞

∑ ∞

∑

σ−2n−1 (k)e−2πkx + (−1)n+1 x2n σ−2n−1 (k)e−

2πk

x = P(x) (3.87)

k=1 k=1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 82

where

(2π)2n+1 ∑

n+1

B2j B2n+2−2j

P(x) = (−1)j x2n+2−2j (3.88)

2x j=0

(2j)! (2n + 2 − 2j)!

{ }

1

− ζ(2n + 1) 1 + (−1)n+1 x2n if n ≥ 1,

+ 12

log x if n = 0,

2

valid for Re x > 0.

Indeed, P(x) is the residual function given as the sum of the residues:

∑

Res (2π)−s Γ(s)ζ(s)ζ(s + 2n + 1)x−s ,

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

P(x) = (3.89)

s=ξ

ξ∈R

a double pole only when n = 0 (others are simple poles). Here (6.153) is to

be used to express the residues in terms of Bernoulli numbers. The same

remark applies to (3.83).

Ramanujan’s formula (3.83), in the rephrased form (3.87), amounts to a

modular relation and is equivalent to Bochner’s formula (Theorem 3.4) (cf.

also [KTY1, Example 1], where the reason why we should have the relation

(3.82) is elucidated). It is also a special case of Theorem 3.9 with a = 0.

Cf. Remark 3.2 above. And therefore we may include the case n = 0, in

which case Liouville’s formula (3.85) is nothing but the sum of the Taylor

expansion of the logarithm function:

∞

∑ ∞

∑ ∞

∑

1 lm

σ−1 (k)q k = q =− log(1 − q k ) := F (τ ), (3.90)

l

k=1 l,m=1 k=1

say, where

q = e2πiτ (3.91)

H defined by (3.100), i.e. Im τ > 0.

Dedekind introduced the eta function (1892) by

∞

∏ ∞

∏

1 πiτ

η(τ ) = q 24 (1 − q k ) = e 12 (1 − e2πikτ ), Im τ > 0 (3.92)

k=1 k=1

and proved its transformation formula (3.94) (under the Spiegelung) and a

1

fortiori (3.137). Note that this is the 24 -th root of the discriminant function

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 83

famous transformation formula for η(τ ) (under the Spiegelung (3.102))

reads

( )

1 1 τ

log η − = log η(τ ) + log , (3.93)

τ 2 i

or

( ) √

1 τ

η − = η(τ ). (3.94)

τ i

The most comprehensive account of the eta-transformation formula is

[KKlam]. See the references therein. For the general transformation for-

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

Theorem 3.7 Ramanujan would have proved the eta-transformation for-

mula (3.94) via his formula (3.87).

Proof. We put x = τ /i (Re x > 0) in (3.87) and (3.88) with n = 0 to

obtain (for F (τ ) defined by (3.90))

( ) ( )

1 π i τ 1 τ

F (τ ) = F − + − + log , (3.95)

τ 12 τ i 2 i

which is the formula deduced by A. Weil [We1, p.401].

Hence we see that

πi

F (τ ) = τ − log η(τ ). (3.96)

12

Rewriting (3.95) in terms of η, on using (3.96), gives (3.93).

∑∞

In (3.90) we summed the infinite sum l,m=1 1l q lm by grouping those l, m

which give the same value n, say, of lm. If instead, we sum over m first,

then we will get a geometric series, so that

∞

∑ ∞

1 i∑1

F (τ ) = −l

= (cot πlτ + i). (3.97)

l(q − 1) 2 l

l=1 l=1

( ) ∞ ( )

1 i∑1 πl

F (τ ) − F − = cot πlτ + cot . (3.98)

τ 2 l τ

l=1

(3.97) suggests a definition of the Dedekind sum based on cotangent values,

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 84

and (3.98) led C. L. Siegel [Sie5] to apply the residue theorem to the function

z −1 f (νz), where f (z) = cot z cot τz and ν = (n + 12 )π (n = 0, 1, . . .), giving

a simple proof of (3.94). As Siegel’s proof has its genesis in the modular

relation, the calculation done in [Sie5] is to find the residual function (and

the right-hand side of (3.98)).

This also assures the validity of Theorem 3.7, i.e. (3.93) was within easy

reach of Ramanujan.

Remark 3.4 Ogg [Ogg] continued the argument of Hecke and Weil [We1]

to give the product expansion of the basic theta-function ϑ = ϑ3 . For a

new treatment of Ramanujan’s formula based on the Barnes multiple zeta-

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

Chan [HH] considers the modular relation for the inverse of the eta-

function, which is the generating function of the partition function.

It would be appropriate to give a brief review here of the theory of modular

forms (or sometimes automorphic forms) in order to make clear the meaning

and importance of the eta-transformation formula (3.94) or in general the

transformation formula under the Spiegelung. There are many good books,

but the most accessible ones for beginners would be Serre [Ser], Doi and

Miyake [DM], Rankin [Ran] or Shimura [Shm1].

Let Γ = Γ(1) = SL(2, Z) denote the special linear group with integer

coeﬃcients:

{ ( ) }

a b

Γ = SL(2, Z) = γ = ad − bc = 1, a, b, c, d ∈ Z . (3.99)

cd

Γ acts on the upper half-plane

H = {z ∈ C| Im z > 0} (3.100)

az + b

γz = . (3.101)

cz + d

We note that Γ is generated by the translation T τ = τ + 1 and the

Spiegelung

1

Sτ = − , (3.102)

τ

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 85

Spiegelung.

Let κ be an even integer greater than or equal to 4. Let Mκ denote

the space of modular forms f (z) of weight κ with respect to the full mod-

ular group SL2 (Z), where f (z) is called a modular form with the above-

mentioned description if it satisfies the following conditions:

M1. f (z) is a holomorphic function on the upper half-plane H and is

holomorphic at infinity.

( az + b ) ( )

M2. f (z) = (cz + d)−κ f for all ac db ∈ SL2 (Z).

cz + d

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

The discriminant function is a cusp form of weight 12, which is de-

fined by

∞

∏ ∞

∑

24

∆(z) = q (1 − q k ) = τ (k)q k , (3.103)

k=1 k=1

Here τ (k) indicates the celebrated Ramanujan’s function (cf. [Ser,

pp.97-98]).

The most important modular forms include Eisenstein series Gκ (z)

defined by

∑ 1

′

Gκ (z) = , (κ ≥ 4) (3.104)

m,n

(mz + n)κ

where the summation runs over all pairs (m, n) of integers other than (0, 0).

Cf. Definition 4.1 below. It is well known that Gκ is a modular form of

weight κ with Fourier expansion

∞

(2πi)κ ∑

Gκ (z) = 2ζ(κ) + 2 σκ−1 (m)e2πimz , (3.105)

(κ − 1)! m=1

and that Mκ has for its basis {Ga4 Gb6 } with a, b non-negative integers satis-

fying 4a + 6b = κ.

The discriminant function appears in the context of the most famous

j-invariant

1728g23

j= , (3.106)

∆

where g2 = 60G2 .

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 86

∑∞

Theorem 3.8 Let f (z) = n=0 c(n)q n be a modular form of weight κ.

Then the zeta-function attached to f defined by

∑∞

c(n)

Φf (s) = , σ>κ (3.107)

n=1

ns

−s k −(κ−s)

(2π) Γ(s)Φf (s) = (−1) (2π) Γ(κ − s)Φf (κ − s). (3.108)

Note that (3.114) and (3.141) below are special cases of (3.108).

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

In this subsection we shall illustrate a remarkable principle that diﬀer-

entiation of Ramanujan’s formula, or the modular relation, gives rise to

automorphy of the corresponding automorphic form.

To this end we introduce the Mellin inversion with shifted argu-

ment Ia (x) for a real and x > 0 by

∫

1

Ia (x) = Ia,n (x) = (2π)−s Γ(s)φ(s − a)x−s ds (3.109)

2πi (κ)

where

with n a non-negative integer and κ > max{1, 1 + a}. It is easily seen that

∞

∑

Ia (x) = σ−2n−1 (k)k a e−2πkx . (3.111)

k=1

∫

1

I(x) = I0 (x) = (2π)−s Γ(s)φ(s)x−s ds. (3.112)

2πi (κ)

We note that this is nothing other than the Lambert series (3.85). In the

sequel suppose a is a non-negative integer referring to the number of times

of diﬀerentiation. Diﬀerentiating I(x) a-times with respect to x, whereby

we perform diﬀerentiation under integral sign, we have the additional factor

∏

a−1

(−s − j),

j=0

November 19, 2014 14:27 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 87

Γ(s) , whence we deduce the remarkable formula

da

I(x) = (−2π)a Ia (x), (3.113)

dxa

i.e. a-times diﬀerentiation of the Lambert series (3.112) is eﬀected by shift-

ing the argument of φ(s) by a in (3.112) and multiplying by (−2π)a . In

view of (3.113), the a-times diﬀerentiated form of Ramanujan’s formula

(3.87) amounts to a counterpart of the modular relation for Ia (x).

We shall obtain it by incorporating (cf. (3.129)) the functional equation

satisfied by φ(s)

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

noting that the horizontal integrals vanish in the limit as |t| → ∞, we have

∫

1

Ia (x) = Γ(s)φ(s − a)(2πx)−s ds + Pa (x), (3.115)

2πi (−κ1 )

where P(x) = Pa (x) denotes the sum of residues of the integrand at its

poles at s = a − 2n − 1, a − 2n, a − 2n + 1, a − 2n + 3, . . . , 0, a + 1.

Substitute the formula

Γ(a − 2n − s)

φ(s − a) = (−1)n (2π)2n+2s−2a φ(a − 2n − s)

Γ(s − a)

which follows from (3.114) in the integral, say, Ja (x) on the right-hand side

of (3.115). Then

Ja (x) (3.116)

∫

(−1) n

Γ(s)Γ(a − 2n − s)

= φ(a − 2n − s)(2π)2n+s−2a x−s ds

2πi (−κ1 ) Γ(s − a)

∫

(−1)n Γ(a − 2n − s)Γ(s)

= φ(s)(2π)−s−a xs−a+2n ds.

2πi (a−2n+κ1 ) Γ(−2n − s)

We note that by the reciprocal relation for the gamma function, we have

= (3.117)

Γ(−2n − s) Γ(1 + s + 2n − a) sin π(s − a)

Γ(a + 2n + 1)

= (−1)a .

Γ(s + 2n + 1 − a)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 88

Hence

Ja (x) (3.118)

n+a ∫

(−1) Γ(s + 2n + 1)Γ(s)

= φ(s)(2π)−s−a xs−a+2n ds.

2πi (a−2n+κ1 ) Γ(s + 2n + 1 − a)

Since the gamma factor can be computed as follows for 0 ≤ a ≤ 2n,

a ( )

Γ(s)Γ(s + 2n + 1) ∑ a (2n − k)!

= Γ(s + k), (3.119)

Γ(s + 2n + 1 − a) k (2n − a)!

k=0

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

a ( )

∑ a (2n − k)!

Ia (x) = (−1)n+a (2π)−a x−a+2n (3.120)

k (2n − a)!

k=0

∫ ( )−s

1 2π

× Γ(s + k)φ(s) ds + Pa (x).

2πi (a−2n+κ1 ) x

Finally, we note that the integral on the right-hand side of (3.120) be-

comes

∫ ( )−s+k

1 2π

Γ(s)φ(s − k) ds,

2πi (a−2n+κ1 +k) x

which is identical with

( )k ( )

2π 1

Ik .

x x

Thus we have proved

Theorem 3.9 (Ramanujan-Guinand formula) For the Mellin transform

Ia (x) with shifted argument as defined by (3.109), we have the modular

relation for n ≥ 0 and 0 ≤ a ≤ 2n,

∑a ( ) ( )k ( )

a (2n − k)! 2π 1

Ia (x) = (−1)n+a (2π)−a x−a+2n Ik (3.121)

k (2n − a)! x x

k=0

+ Pa (x),

Particular cases are

P0 (x) = P(x),

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 89

1

P2n (x) = (2n)! ζ(2n + 2)(2πx)−2n−1 + ζ(−2n − 1)(2πx) (3.122)

2

(−1)n −2n

+ (2π) (2n)!ζ(2n + 1), n ≥ 1.

2

On the other hand, for a = 2n + 1, we have

( )

1

I2n+1 (x) = (−1)n+1 (2π)−2n−1 x−2n−2 I2n+1 + P2n+1 (x), (3.123)

x

where

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

1

P2n+1 (x) = (2n + 1)!ζ(2n + 2)(2πx)−2n−2 − ζ(−2n − 1) (n ≥ 1). (3.124)

2

The meaning of this equality is discussed below.

Remark 3.5 The case a = 0 is Ramanujan’s formula (3.87), the case a =

2n is Guinand’s formula (cf. [KTY1, Theorem 3]), and the case a = 2n + 1

gives, on writing m = −n − 1, the negative case (m < 0) of Ramanujan’s

formula (3.87).

The special case of (3.122) with n = 1, i.e. once diﬀerentiated form of

Ramanujan’s formula, yields Terras’ formula [Te2], [Te3]

∞

∑ ( )

2 3 −2πk 1

ζ(3) = π −4 e 2 2

σ−3 (k) 2π k + πk + . (3.125)

45 2

k=1

perform numerical calculation of zeta values in a remarkably eﬃcient way.

Corollary 3.2 ([KTY1, Corollary 1]) For n ≥ 1 we have

1 (2π)2n+1 B2n+2 ( 1 )

(2n)! ζ(2n + 1) + − (−1)n (3.126)

2 2(2n + 2) 2n + 1

∑

2n+1

= (−1)n (2π)2n (r − 1)!S(2n + 1, r)A(1, r)

r=1

∑2n

(2n)! ∑

m+1

− (2π)m (r − 1)! S(m + 1, r)A(2n + 1 − m, r),

m=0

m! r=1

where

∞

∑ 1

A(p, q) = p (e2πm − 1)q

m=1

m

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 90

and S(k, r) denote the Stirling number of the second kind defined by

∑

k

xk = S(k, r)x(x − 1) · · · (x − r + 1).

r=1

{∞

2 3 ∑ 1( 1 1 ) 1

ζ(3) = π − 8π 2 1+ + (i)

45 k 2πk (2πk)2 e2πk−1

k=1 }

1( 1 )

∞

∑ ∞

∑

1 2 1

+ 3+ +

k 2πk (e2πk − 1)2 k (e2πk − 1)3

k=1 k=1

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

= 1.20205690315959428539973816151144999076498629234049

888179227155534183820578631309018 . . .

8π 3 ∑ 1 ( 6 )

∞

22 π 5 3 6 1

ζ(5) = − 1 + + + (ii)

33 · 5 · 7 3 k2 2πk (2πk)2 (2πk)3 e2πk − 1

k=1

1( 6 )

∞

3 ∑

8π 9 1

− 7 + +

3 k2 2πk (2πk)2 (e2πk − 1)2

k=1

1( 1 )

∞

∑ ∑ 1 ∞

1 1

− 16π 3

2 + − 16π 3

k 2 2πk (e 2πk − 1)3 2

k (e2πk − 1)4

k=1 k=1

= 1.03692775514336992633136548645703416805708091950191

28119741926779038035897862814845600 . . . .

Remark 3.6 In the book [SC] Srivastava and Choi record that by their

formula, they were able to compute the value of ζ(3) accurate up to 7 decimal

places by taking only first 50 terms.

It should be noted that their formula is a variant of Euler’s classical

formula and is easily derivable from [KKY1] or [KKY2] by diﬀerentiating

and forming linear combinations of resulting formulas. Note that by their

method one could not attain the acceleration of order e2π ,

e2π = 535.5491655524764736503049326 . . . ,

Remark 3.7 The method of proof of Theorem 3.9, which was not per-

ceived in [KTY1], the Mellin inversion with shifted argument is an additive

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 91

logues have been successfully applied in [KTY2]–[KTY5]. However, from

our present view point, this is simply a processed modular relation with the

processing gamma factor Γ(s + a) (cf. (3.114) and (3.115)). In this special

case, however, the main formula (3.113) is the manifestation of the state-

ment of Razar [GoRa] that the diﬀerentiation of Lambert series essentially

corresponds to the shift of the argument of the associated Dirichlet series

(cf. Remark 3.1).

Remark 3.8 We have seen that the special case n = 0, a = 0 of Theorem

3.9, the Ramanujan-Guinand formula, is nothing other than the automor-

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

diﬀerentiated form of Guinand’s formula. We note that in the case of n = 0,

the twice diﬀerentiated Guinand’s formula, with a suitable modification, co-

incides with the automorphic property of the Eisenstein series E2 (cf. e.g.

Rankin [Ran, p.196]). It is related to F defined by (3.90) by way of

∞

∑ kq k

3

E2 (τ ) + = 1 − 24

πy 1 − qk

k=1

12

= 1 − F ′ (τ )

πi

12 ′

= (log η(τ )) .

πi

In order to grasp the meaning of Formula (3.123) in the light of the trans-

formation formula for Eisenstein series, we are now in a position to interpret

(3.123), once diﬀerentiated form of Guinand’s formula (= 2n + 1 times dif-

ferentiated form of Ramanujan’s formula = the negative case thereof), in

terms of G2n+2 .

Stating (3.123) in explicit form

∞

∑ B2n+2

−2 σ2n+1 (k)e−2πkx +

2n + 2

k=1

{ ∞

}

∑ B2n+2

n+1 −2n−2 − 2πk

= (−1) x −2 σ2n+1 (k)e x + ,

2n + 2

k=1

( )

1

G2n+2 − = (ix)2n+2 G2n+2 (ix), (3.127)

ix

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 92

The following theorem is to be compared with Theorem 3.7. It is re-

markable that both are touched upon by Indian mathematicians.

Theorem 3.10 The functional equation (1.5) for the Riemann zeta-

function implies the functional equation (3.114), which in turn implies the

transformation formula for the discriminant function:

( )

1

∆ − = τ 24 ∆(τ ), (3.128)

τ

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

should read

where n can be any integer. Then we may simply apply Theorem 3.2

to deduce (3.123) for Ia,n (x) and the remaining thing is to compute the

residual function P(x) given by (3.89). We need the case I1 (X) = I0,1 (x) =

∑∞ −2πkx

k=1 σ1 (k)e in (3.111) and the modular relation reads

( )

1 1 1 −1 1

I1 (x) = −x−2 I1 + x−2 − x +

x 24 4π 24

or

( )

1 1 1 τ

− + I1 = −τ 2 (− + I1 (τ )) − , (3.130)

24 τ 24 4π

∞

∑ ∞

∑ 6iτ

σ1 (k)e−2πkτ ) = 1 − 24 σ1 (k)e−2πk τ +

1

τ 2 (1 − 24 .

π

k=1 k=1

Since

1 ∆′ ∆′

(τ ) = q (q) = 1 − 24I1 (τ ), (3.131)

2πi ∆ ∆

(3.130) amounts to

( )

∆′ 1 ∆′ 1 12

(τ ) = 2 − − ,

∆ τ ∆ τ τ

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 93

( )

1

log ∆(τ ) = log ∆ − − 12 log τ + log k,

τ

k being a constant. Exponentiating, we find that

∆(τ ) k

( ) = 12 . (3.132)

∆ − τ1 τ

the proof.

Exercise 3.4 Prove that the j-invariant defined by (3.106) is a modular

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

( )

1

j − = j(τ ). (3.133)

τ

Solution Note that (3.106) reads

1 + 240I3 (τ )

j(τ ) = , (3.134)

∆(τ )

where

∞

∑

I3 (τ ) = I3,1 (τ ) = σ3 (k)e−2πkτ .

k=1

( )

1

1 + 240I3 = τ 4 (1 + 240I3 (τ )). (3.135)

τ

Substituting (3.128) and (3.134) in (3.135) leads to (3.133).

For the history of Dedekind sums, cf. [Asa], [HiTa] and the beginning

of §3.3.1. Since the 24-th power ∆ of η is the more classically known

fundamental discriminant function, which is a modular form of weight 12,

it is natural that the η-function underwent much attention and has been a

fundamental example of half-integral weight modular forms.

Another transformation formula is

πib

log η(z + b) = log η(z) + , (3.136)

12

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 94

The general transformation formula for η(z) for c > 0 reads

( )

az + b

log η(γz) = log η

cz + d

(3.137)

1 cz + d πi a + d

= log η(z) + log + − πis(c, d),

2 i 12 c

where s(h, k) is the celebrated Dedekind sum defined by

∑

k−1 (( m )) (( hm )) ∑ (( m )) (( hm ))

s(h, k) = = , (3.138)

k k k k

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

m=1 m mod k

sums only and coincides with the sawtooth Fourier series ψ(x) in (3.74).

Hence we have another expression

∑

k−1 (m) ( ) ∑ (m) ( )

hm hm

s(h, k) = B̄1 B̄1 = B̄1 B̄1 . (3.139)

m=1

k k k k

m mod k

m̸≡0

N, (h, k) = 1

( )

1 1 h 1 k

s(h, k) + s(k, h) = − + + + . (3.140)

4 12 k hk h

There exist many generalizations of (3.140) and proofs thereof for which

we refer to [RG].

Theorem 3.11 The functional equation (1.5) for the Riemann zeta-

function implies the reciprocity relation (3.140) up to (3.136).

Lemma 3.1 (i) The functional equation (1.5) for the Riemann zeta-

function ζ(s) implies

(ii) The functional equation (1.5) for the Riemann zeta-function implies the

transformation formula (3.93).

November 19, 2014 14:27 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 95

formula and the duplication formula for the gamma function. Proof of (ii)

can be found in [RG, pp.47-48], which amounts to the following. For any

( )

ab

g=

cd

in Γ, we have

( )( ) ( )

ab 0 1 b −a

gS = = .

cd −1 0 d −c

on one

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

cSτ +d

hand and

bτ − a

τ ′′ = , (3.142)

dτ − c

we obtain

( )

1 d c ad − bc 1

s(d, c) − s(−c, d) = + + log(−i), (3.143)

12 c d dc 2πi

on the other, whence by oddness s(−h, k) = −s(h, k), we arrive at (3.140).

By Lemma 3.1, we complete the proof of Theorem 3.11.

This section will serve to get the reader familiar with the setting in number

fields as well as fixing some notation, including the choice of the fundamen-

tal sequence (by incorporating the associated constants in them). We shall

elucidate the trilogy of Koshlyalov and show that Koshlyakov’s theory of

Dedekind zeta-functions of quadratic fields [KoshI] reduces in the long run

to Theorem 4.8: the G1,1 2,0

1,1 ↔ G0,2 formula, or the Fourier-Bessel expan-

sion (the K-Bessel series for the perturbed Dirichlet series in the context

of [KTY7]). This trilogy hitherto has not been well-known compared with

his other papers [Kosum] and [Kosvor] but is rich in contents and can be

thought of as the compilation of his work. To expound it is beneficial in

two respects. First, one may come to know this relatively unknown but

important work, and secondly, to our benefit, we can exhibit three typical

cases of the functional equation corresponding to the rational, imaginary

and real quadratic field. The latter two corresponds to the zeta-functions

of definite and indefinite quadratic forms, respectively. The zeta-functions

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 96

theory has been developed rather fully. We refer to [Te5], [Vista, Chapter

6] and [Vi2, Chapter 6] for the definite case, and refer to [Sie3] and [Sie4]

for the latter. This view point may be found in [Coh] as well as in [Dav],

which refers to [LanZT] for proofs.

We follow the notation in §3.2, which is a slight deviation of Koshlyakov’s.

We assume that the extension degree of the number field Ω satisfies κ =

[Ω : Q] ≤ 2, whence Ω is either the rational or a quadratic field.

√ Hence if

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

the convention that ∆ = 1 in the case of the prime field Ω = Q.

{

∆=1 n=1

(3.144)

∆ = a square-free integer n = 2.

case may be expressed as

(r)

2r+1 π r2 Rh 2r+1 π r2 ζΩ (s)

ρ= √ =− √ , (3.145)

w |∆| |∆|

where (r) means the r-th derivative. But since the rank of the unit group

of Ω satisfies r = r1 + r2 − 1 ≤ 1, the derivative appears only in the case of

real quadratic fields.

(3.39), (3.40), (3.41) may be expressed concretely as

(s)

Γ π − 2 ζ(s)

s

n=1

2

( )−s

2π

χ(s) = Γ(s) √ ζΩ (s) n = 2, ∆ < 0 (3.146)

|∆|

( s )( π )−s

Γ2 √ ζΩ (s) n = 2, ∆ > 0

2 ∆

= χ(1 − s),

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 97

where

∑∞

FΩ (n)

φ(s) = A−s ζΩ (s) = s (3.147)

n=1

(An)

∞

∑ 1

− 2s

√

π ζ(s) = s n=1

( πn)

∞

n=1

∑ FΩ (n)

s n = 2, ∆ < 0

=

( √2π n)

n=1 |∆|

∑

∞

FΩ (n)

√ n = 2, ∆ > 0.

n=1 ( √ π n)

s

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

|∆|

nel function), K(x) = Kr1 ,r2 (x) defined by (4.57), which plays the most

prominent role in his theory; along with K, also introduced are X, Y , Z,

L, M , N , see the following table ([KoshI, p.122]). We shall treat some

(properties) of them.

ftn’s rat. imag. quadr. real quadr.

2

X 2e−x e−x 4K0 (2x)

2

Y e−x si(x)

( √ ) J0 (2x)

√ −x ( √ )

K πe −2kei0 4 x4 4ker0 4 x

√ √ ( √ √ )

L π cos 2x π

J (2 x) 2 K0 (4 x) − π2 Y0 (4 x)

√ √2 0 π √ √

M π sin 2x K0 (2 x) + 2 Y0 (2 x) πJ0 (4 x)

√

π −2x

√

x ( √ √ ) √ ( √ √ )

N 2

e i

εK1 (2ε̄ x) − ε̄K1 (2ε x x εK1 (4ε̄ x) + ε̄K1 (4ε x

Exercise 3.5 Prove the formulas in Table 3.2 for the X-function defined

by (3.43).

2

(3.148)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 98

and

Exercise 3.6 Recall( that

) Rj denote the residues at j = 0, 1 of the mero-

morphic function Γr1 2s Γr2 (s)A−s ζΩ (s)z −s ((3.45)). Prove in the case

κ ≤ 2 that

R1 = −2r1 ζΩ (0)z −1 ,

(r) (r)

R0 = 2r1 ζΩ (0), (3.151)

whence P(x) = 2r1 ζΩ (0) − 2r1 ζΩ (0)z −1 , which is the corrected form of

(r) (r)

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

( ( s )r1 ) (r −1)

ζ 1 (0)

R0 = Res Γ A−s ζΩ (s) z −s , s = 0 = 2r1 Ω

2 (r1 − 1)!

and

( ( s )r1 ) 1 (r −1)

R1 = Res Γ A−s ζΩ (s) z −s , s = 1 = −2r1 ζ 1 (0) z −1 .

2 (r1 − 1)! Ω

Solution

( ( )r )

1 − s 1 −(1−s) −s

R1 = Res Γ A ζΩ (1 − s) z , s = 1

2

(( )r ( )r

−2 1 3 s 1 −(1−s)

= Res Γ − A

s−1 2 2

( ) )

(r −1)

ζΩ 1 (0) r1 −1 −s

··· + (1 − s) + ··· z ,s = 1 .

(r1 − 1)!

( ) (r −1)

ζ 1 (0)

Note that in [BeI], Res Γ(s)r1 A−2s ζΩ (s) z −s , s = 0 = 2r1 −1 Ω(r1 −1)!

and

( ( s )r1 ) ( )r1

−s −s 1

Res Γ A ζΩ (s) z , s = 1 = Γ A−1 λhz −1 ,

2 2

where λh is the residue of ζΩ (s) at s = 1 given by (3.38). It may be also

expressed in the case κ ≤ 2 as

(r)

2r+1 π r2 ζΩ (s)

λh = − √ . (3.152)

|∆|

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 99

“On [KoshI, p. 114], the erroneous value ζΩ (0) = −1 of the Dedekind

zeta-function for the imaginary quadratic field is given. The true value is

ζΩ (0) = − h2 with the class number h in my previous paper · · · .”

However, the correction is valid except for the cases of the Gauss field

Ω = Q(i) (w = 4) and the Eisenstein field Ω = Q(ρ) (w = 6), where

2πi 2πi

i = e 4 and ρ = e 6 indicates the piervot’nyi primitive root of unity. In

our case of κ ≤ 2, we may unify the values into the form

h

ζΩ (0) = − (3.153)

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

w

with understanding that h = 1, w = 2 in the rational field and w = ∞

in the real quadratic field. Of course, in the rational case, a more famil-

iar expression for the values of the Riemann zeta-function at non-positive

integral arguments is that of Euler (cf. [NTA, Proposition 5.1, p. 138]),

ζ(0) = − B20 = − 12 . Euler’s formula is a consequence of the functional

equation. Historically, Euler first prove the expression and conjectured the

general form of the functional equation.

Using Exercises 3.5 and 3.6, we may state Corollary 3.1 in a more explicit

form, which is [KoshI, (4.1)]. It combines three identities (3.155), (3.156),

and (3.157) below.

Theorem 3.12 In the case of the rational or a quadratic field, the un-

processed modular relation reads

{ ∞

}

√ ∑

r1 (r)

ρ 2 ζΩ (0) − F (k)Xr1 ,r2 (Akρ)

k=1

{ ∞ ( )} (3.154)

1 ∑ 1

=√ 2r1 ζΩr (0) − F (k)Xr1 ,r2 Ak .

ρ ρ

k=1

(4.8)]:

{ ∞

} { ∞

}

√ ∑ 1 ∑ −πk 2 1

e−πk ρ

2 2

ρ 1+2 =√ 1+2 e ρ2 (3.155)

ρ

k=1 k=1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 100

or writing ρ2 for z,

∞

∑ ∞ (

∑ )

−πk2 z 1 − 12 −πk2 z −1 1

e + =z e + ,

2 2

k=1 k=1

In the imaginary quadratic case, (3.154) reads

∞

( ∞

)

∑ − √2π z ∑ − √2π z −1

−ζΩ (0) + F (k)e |∆| = z −1 −ζΩ (0) + F (k)e |∆| ,

k=1 k=1

or

( )

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

∞

∑

√ − √2π ρ

|∆|

ρ h+w F (k)e

k=1

( ∞

) (3.156)

1 ∑ − √2π ρ−1

=√ h+w F (k)e |∆|

,

ρ

k=1

(4.9)].

Finally, in the real quadratic case, (3.154) reads

{ ∞ ( )}

√ ∑ 2πk

ρ 4ζΩ′ (0) − 4 F (k)K0 √ ρ

k=1

∆

{ ∞ ( )} (3.157)

1 ∑ 2πi 1

′

=√ 4ζΩ (0) − 4 F (k)K0 √ .

ρ ∆ρ

k=1

0,2 ↔ G0,2 formula (§3.1.2). A

more general case has been treated as [BeI, Example 3] and a more general

form of (3.157) for general r1 is given on [BeI, p.358].

Berndt treats on [BeI, p.358] A−2s ζΩ (2s) as the Dirichlet series φ(s), so

that χ(s) = Γ(s)r1 φ(s)z −s , replaces z/A2 by y, whence z − 2 = Ay − 2 , and

1 1

so

( ) ( )r1

1 1 1

Res Γ(s)r1 A−2s ζΩ (2s) z −s , s = λhy − 2 .

1

=Γ

2 2 2

Because of this replacement, the modular relation leads to a slightly diﬀer-

ent form ([BeI, p.358]). For this cf. [KTY7, §7].

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 101

Complementary to the X-function is the Y -function which is the Mellin

factor of the K-function in the sense that ([KoshI, (5.8)], Example 6.1):

∫ ∞ ( )

a 1

X Y (bx) dx = K(ab), a > 0, b > 0. (3.158)

0 x b

∫

1 π

= ( ) ( ) x−s ds

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

(3.159)

∫

2

( )

1 Γ(s)Γ(1 − s) cos π2 s −s

= ( ) x ds, c > 0, x > 0,

2πi (c) Γr1 2−s 2 Γr2 (2 − s)

( )

1

(0, 1), ( 2 , 2 )1

1,1

= H2,r+3 x

(0, 1), ( 21 , 12 ), (0, 12 ), . . . , (0, 12 ), (−1, 1), . . . , (−1, 1)

( )

1 1,1 (0, 1)

e 2 i x

π

= H1,r+2

2 (0, 1), (0, 21 ), . . . , (0, 12 ), (−1, 1), . . . , (−1, 1)

( )

1 1,1 −π i

(0, 1)

+ H1,r+2 e 2 x .

2 (0, 1), (0, 21 ), . . . , (0, 12 ), (−1, 1), . . . , (−1, 1)

is trivial, we may view Y as the case where the processing gamma factor

Γ(1−s)

((7.24)) is Γ(s | ∆) = Γr1 2−s and the functional equation (3.41) is

( 2 )Γr2 (2−s)

the case of the imaginary quadratic field. This fact will show its eﬀect in

due course.

We distinguish three cases. In the case r1 = 1, r2 = 0,

∫ ( ) ( )

1 s −s ds −

= G0,1 x2 = e−x .

1,0 2

Y1,0 (x) = Γ x (3.161)

2πi (c) 2 2 0

( ( ) ( ))

1 0 0

Y0,1 (x) = 1,1

G1,2 iz 1,1

+ G1,2 −iz . (3.162)

2 0, −1 0, −1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 102

By (2.112),

( ) ( )

0 s Γ(1 + s) 1+s

G1,1

z =z 1 F1 ; −z .

1,2

s, −1 Γ(2 + s) 2+s

( )

0 Γ(1 + s)

1,1

G1,2 z = zs (s + 1)z −(s+1) γ(s + 1, z)

s, −1 Γ(2 + s) (3.163)

−1

=z γ(s + 1, z).

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

( )

0 1 − e−z

1,1

G1,2 z = z −1 γ(1, z) = . (3.164)

0, −1 z

Hence

sin(x)

Y0,1 (x) = = si(x), (3.165)

x

the sinus cardinalis function which plays an important role in signal trans-

mission (cf. [Spl]). Below we will give a modular relation version of the

sampling theorem in Theorem 3.14.

Finally, in the case of

∫

1 π −s

Y2,0 (x) = ( ) ( ) x ds,

2πi (c) 2 sin π s Γ 1 − s 2

2 2

Γ( 2s ) x−s

we rewrite the integrand as Γ(1− 2s ) 2 and then express Y2,0 (x) as the

G-function:

( ) ( )

1 1,0 −

2 −

Y2,0 (x) = H0,2 x = G1,0

x

2 (0, 21 ), (0, 12 ) 0,2 0, 0 = J0 (2x)

We recall that the X- and Y -functions play an important role in the

derivation of the Riemann-Siegel integral formula [KTY7, §4.2], in that a

recourse to the formula ([KoshI, (9.10)])

∫ ∞

Z(a, b)

X(ax)Y (bx) x dx = 2 , a > 0, b > 0

1 a + b2

was essential.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 103

main theorem of modular relations. One of the forms of the main theorem

in Chapter 7 reads

Theorem 3.13 If

( )

{(1 − aj , Aj )}nj=1 ; {(aj , Aj )}pj=n+1

∆= q ∈Ω (3.166)

{(bj , Bj )}m

j=1 ;{(1 − bj , Bj )}j=m+1

∏

m

( ) ∏

n

( )

Γ bj + Bj s Γ aj − Aj s

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

j=1 j=1

Γ(s | ∆) = , (3.167)

∏p

( ) ∏

q

( )

Γ aj + Aj s Γ bj − Bj s

j=n+1 j=m+1

then

( )

∞

∑ {(1 − aj , Aj )}n , {(aj , Aj )}p

αk j=1 j=n+1

H m+1,n zλk (3.168)

λsk p,q+1 (s, 1), {(bj , Bj )}m q

j=1 , {(1 − bj , Bj )}j=m+1

k=1

( )

∑∞ {(1 − bj , Bj )}m q

βk n+1,m µk j=1 , {(bj , Bj )}j=m+1

= H

µr−s q,p+1 z (r − s, 1), {(aj , Aj )}nj=1 , {(1 − aj , Aj )}pj=n+1

k=1 k

+ P(z)

∑L ( )

k=1 Res Γ(w − s | ∆) χ(w) z

s−w

where P(z) = , w = sk indicates the

residual function.

χ(s) = Γ(s)φ(s) has simple poles at s = 1 and s = 0.

Formally we apply the case m = 0, n = 1, p = 2, q = 2 and a1 = 1, A1 =

1, a2 = 12 , A2 = 21 b1 = −1, B1 = 1, b2 = 12 , B2 = 21 of Theorem 3.13 with

Γ(1 + s − w))

Γ(w − s | ∆) = ( ) ( 1−s 1 ) .

Γ(2 + s − w)Γ 1+s

2 − 2w Γ

1

2 + 2w

Γ(1−(w−s))

cos π2 (w − s) and expressing the cosine function as the sum of

Γ(2−(w−s))

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 104

puted to be

Γ(1 + s) πs Γ(s) s−1 πs

P(z) = cos z s φ(0) + z sin ρφ (3.170)

Γ(2 + s) 2 Γ(1 + s) 2

1 πs s 1 s−1 πs

= cos z φ(0) + z sin ρφ .

s+1 2 s 2

The modular relation (3.168) amounts to the case m = 0, n = 1, p = 1,

q = 1:

( ( ) ( ))

∑∞ 0 0

αk 1,1 1,1

G1,2 izλk + G1,2 −izλk (3.171)

s, −1 s, −1

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

λs

k=1 k

( ( ) ( ))

∑∞ 2 2

βk 2,0 µ k 2,0 µ k

= G1,2 i + G1,2 −i

µr−s z r − s, 1 z r − s, 1

k=1 k

+ 2P(z).

The left-hand side function is (3.162), which is the sinus cardinalis func-

tion in (3.165).

On the right-hand side, by (2.117)

( )

2 ( )

2,0

G1,2 z = e−z z r−s U 1, r − s, z .

r − s, 1

( )

2

G2,0 z −z r−s 1−(r−s) z

e Γ(r − s − 1, z)

1,2 r − s, 1 = e z z

( )

2

G2,0 z −z

1,2 1, 1 = zΓ(1, z) = ze , (3.173)

we conclude that

( ( ) ( ))

1 2 2

G1,2 iz

2,0

+ G1,2 −iz

2,0

(3.174)

2 1, 1 1, 1

= z sin z.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 105

Theorem 3.14

∞

∑ ∞

1∑ µk π

αk si(zλk ) = βk sin + φ(0) + z −1 ρφ . (3.175)

z z 2

k=1 k=1

This is interesting in the light of the sampling theorem, i.e. the left-hand

side of (3.175) may be thought of as the Fourier series of the function f (z)

in question and the right-hand side is the sampling series as given in the

sampling theorem.

§4.6) and are related through

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

1( )

Lr1 ,r2 (x) = Kr1 ,r2 (−ix) + Kr1 ,r2 (ix) ([KoshI, (7.11)]). (3.176)

2

It is defined by ([KoshI, (8.9)])

∫

1 π

Lr1 ,r2 (x) = G(1 − s) x−s ds

2πi (c) 2

∫ ( ) (3.177)

1 π Γr1 2s Γr2 (s)

= ( ) x−s ds, c > 0, x > 0.

2πi (c) 2 Γr1 1−s

2 Γr2 (1 − s)

Exercise 3.8 Prove the formulas for L-function in Table 3.2.

Solution The L1,0 (x) is similar to Y0,1 (x) and

( )

π

2 − 1

L1,0 (x) = · 2G1,0 0,2 x 0, 1 = πx J 12 (2x)

2

2 2

L0,1 (x) is just the same as Y2,0 (x).

Finally, L2,0 (x) may be transformed by (2.93) into

( )

π 2,0 −

L2,0 (x) = H0,4 x (3.178)

2 (0, 21 ), (0, 12 ), ( 12 , 12 ), ( 21 , 12 )

( )

π 2

−

= 2G2,0 x 0, 0, 1 , 1

2 0,4 2 2

( )

2 ( √ ) ( √ )

=π K0 4 x − Y0 4 x

π

by (9.21) in §9.1.2.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 106

“It should be noted that deduction of the functional equation for the

Dedekind zeta-function from formulas in our general theory is a total vicious

circle since the former are already used in deriving the latter.”

Such deductions occur

p.129, ll. 1-12 from below

which turns out to be the same as

p.234, ll. 1-10 from below and p.235, ll. 1-16.

by KAINAN UNIVERSITY on 02/26/15. For personal use only.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 107

Chapter 4

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

overstated. Along with this, there has been applied equally eﬀective method

of the beta transform, or the binomial expansion, leading to the K-Bessel

series. The purpose here is to elucidate the explicit and implicit use of

the beta transform in the transformation of the perturbed Dirichlet series

and reveal the hidden structure as the Fourier-Bessel expansion. As the

first instance, we shall clarify Stark’s method along with the Murty-Sinha

theorem as a manifestation of the beta-transform. As the second and main

part, we shall resurrect the long-forgotten important work of Koshlyakov

by revealing that Koshlyakov’s formula for the perturbed Dedekind zeta-

functions for a real quadratic field is in fact Lipschitz summation formula

and that for an imaginary quadratic field the K-Bessel function is intrinsic

to the Hecke functional equation. Similarly, we shall elucidate Koshlyalov’s

K-series in terms of Kelvin functions.

4.1 Introduction

the original Dirichlet series are assumed to satisfy the functional equation,

as their Fourier-Bessel expansion, Theorem 4.2.

Then we shall elucidate Stark’s method for deriving the Dirichlet class

number formula and the special values at negative integers of the Hurwitz

zeta-function as a manifestation of the beta-transform. A remarkable fea-

ture is that there is no need to appeal to the functional equation. However,

to derive Lerch’s formula (4.15), we need to make a recourse to a formula

of Ramanujan and Yoshimoto.

Then we will discuss why the work in [KoshI]-[KoshE] is dominated by

107

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 108

pp.243-247] in this vein and show that Koshlyakov’s formula for the per-

turbed Dedekind zeta-functions for a real quadratic field is in fact Lips-

chitz summation formula [Lip2] (cf. also [Ler3]). While in the imaginary

quadratic case, Proposition 4.3 explains the situation where the K-Bessel

function is intrinsic with the Hecke type functional equation (i.e. with the

gamma factor Γ(s)).

In the end we shall elucidate the nature of Koshlyakov’s σ-function in

terms of the modified Kelvin functions.

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

One can find in the literature many works devoted to the “analytic contin-

uation by Taylor expansion” of a class of zeta-functions. One direction is

to appeal to the binomial expansion due to Wilton [Wil0], e.g. [Mik, (4.6),

p.152], [MS], [Sta], etc. and another is to apply the Euler transformation

[Gs], [Has], [So] with diﬀerencing, which has recently been elucidated in

[XHW]. The paper of Murty and Sinha [MS] contains a general theorem

which gives an analytic continuation for the perturbed Dirichlet series once

an analytic continuation is known for the original Dirichlet series. We state

the Murty-Sinha theorem and materialize it in view of the binomial expan-

sion, or the beta transform.

∑∞

αn

φ(s) =

λs

n=1 n

tion to the whole plane, then its perturbation

∞

∑ αn

φ(s, a) = s

n=0

(λn + a)

also extends meromorphically to the whole plane and the poles of φ(s, a)

are contained in the positive integral translates of poles of φ(s).

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 109

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 109

ciple of analytic continuation” stated as

∑∞ ( )

−s

φ(s, a) = φ(s + r)ar (4.1)

r=0

r

We may interpret (4.1) as the beta-transform (cf. [Vi2, p.34–35]) (which

turns out to be an analytic expression for the binomial expansion),

∫

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

1 Γ(z + s) Γ(−s) s

(1 + x)−z = x ds

2πi (c) Γ(z)

( ) (4.2)

1 1

= G1,1 x−1

Γ(z) 1,1 z

for x > 0, −ξ < c < 0, where (c) denotes the vertical Bromwich path

s = c + it with −∞ < t < ∞. Here G1,1 1,1 is the Meijer G-function.

This beta-transform has been extensively used in the literature by many

authors including Mellin, Barnes, Hardy, Berndt, Koshlyakov, Katsurada,

Matsumoto et al., who referred to it as the Mellin-Barnes integral, which

1,1

turns out to be the beta-function integral H1,1 , where H indicates the Fox

H-function ((1.22)).

∑∞

e2πinλ

ϕ(λ, a, z) = (4.3)

n=0

(n + a)z

Cf. e.g. [SC, p.122]. The Hurwitz zeta-function is a special case of (4.3)

with λ ∈ Z:

∞

∑

ζ(z, a) = ϕ(λ, a, z) = (m + a)−z

m=0

for a > 0.

We illustrate Theorem 4.1 by

obtain the integral representation

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 110

∫

1 Γ(−w)Γ(s + w)

ϕ(λ, a + x, s) = ϕ(λ, a, s + w)xw dw. (4.4)

2πi (c) Γ(s)

Moving the line of integration to the right (thereby using Stirling’s for-

mula to make sure that the horizontal integrals vanish) up to Re w = aK .

Here

K < aK < K + 1, K ∈ N.

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

As a result we get

∫

1 Γ(−w)Γ(s + w)

ϕ(λ, a + x, s) = ϕ(λ, a, s + w) xw dw

2πi (aK ) Γ(s)

1 ∑

K

(−1)k

+ Γ(s + k) ϕ(λ, a, s + k) xk . (4.5)

Γ(s) k!

k=0

The relation (4.5) holds true for σ > −K and the integral is absolutely

convergent. Thus it gives an analytic continuation for ϕ(λ, a + x, s) in that

domain. The sum in (4.5) is

K (

∑ )

−s

ϕ(λ, a, s + k)xk .

k

k=0

The Taylor expansion [SC, p.125] (for |a| > 0) due to Klusch [Kl1] is:

∞

∑ ν (s)ν

L(x, s, a + ξ) = (−1) L(x, s + ν, a)ξ ν (4.6)

ν=0

ν!

symbol. Now (4.6) with x ∈ Z reduces to Wilton’s formula [Wil0]

∞

∑ ν (s)ν ν

ζ(s, a + ξ) = (−1) ζ(s + ν, a)(ξ) . (4.7)

ν=0

ν!

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 111

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 111

The relation (4.7) entails the generalized Ramaswami’s formula [SC, pp.145-

146]:

∑∞

(s)ν ζ(s + ν, a)

ζ(s, 2a − 1) − 21−s ζ(s, a) = . (4.8)

ν=0

ν! 2s+ν

∑∞

(s)ν ζ(s + ν)

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

(1 − 21−s )ζ(s) =

ν=0

ν! 2s+ν

∞

∑ ( )

ν −s ζ(s + ν)

= (−1)

ν=0

ν 2s+ν

= (−1)

ν! ν!

( )

ν −s

= (−1) .

ν

The Taylor series being convergent for |x| < |a|, we see that (4.5) leads to

a generalized Ramaswami’s formula.

equation, then the Murty-Sinha theorem amounts to the correspondence

1,1

H1,1 and its counterpart. Our Theorems 4.4, 4.5 and 4.6 illustrate this

principle.

We now move to consider the special case ζ(s, x) = ϕ(1, x, s) of the Hurwitz

zeta-function. While applying (4.2), we need to separate the term with

n = 0 and apply it with nx . Then we obtain, corresponding to (4.4),

∫

1 Γ(−w)Γ(s + w)

ζ(s, x) = ζ(s + w)xw dw (4.9)

2πi (c) Γ(s)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 112

valid for 1 − c < σ with c < 0. In the same way, we obtain corresponding

to (4.5),

∫

−s 1 Γ(−w)Γ(s + w)

ζ(s, x) = x + ζ(s + w)xw dw

2πi (aK ) Γ(s)

K (

∑ )

−s

+ ζ(s + k)xk (4.10)

k

k=0

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

which is valid for σ > −K and the integral is absolutely convergent and

thus gives an analytic continuation for ζ(s, x) in that domain. If we move

the line of integration far to the right as in the proof of Proposition 4.2, we

obtain the Taylor expansion (4.7) with a = 1.

H. Stark [Sta] uses only first a few terms to deduce the Dirichlet class

number formula and the special values of the relevant zeta and L-functions.

The Dirichlet L-function has the representation in terms of the Hurwitz

zeta-function

( )

1 ∑

q−1

a

L(s, χ) = s χ(a)ζ s, (4.11)

q a=1 q

formulas for the latter. Moving further and comparing [Sta, (13)] and

(4.5), we obtain

∞ ( K ( )

∑ −s

∑ −s −n−j j )

(n + x) − n x (4.12)

n=1 j=0

k

∫

1 Γ(−w)Γ(s + w)

= ζ(s + w)xw dw.

2πi (aK ) Γ(s)

If one follows Stark’s method more closely and write K + 1 for K in (4.5)

and set s = −K, 0 ≤ K ∈ Z in (4.5) one can deduce

∑(

K+1

K

)

K

ζ(−K, x) = x + ζ(−K + k)xk (4.13)

k

k=0

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 113

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 113

( )

−s

ζ(−s + K + 1) xK+1

K +1 s=−K

(−s) · · · (−s − K)ζ(−s + K + 1) xK+1

s=−K K+1

= x =

(K + 1)! K +1

on using the fact that Res ζ(s) = 1 and the term with k = K is ζ(0)xK .

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

Proposition 4.1

xK+1 ( ) ∑ (K )

K−1

ζ(−K, x) = + 1 + ζ(0) xK + ζ(−K + k) xk . (4.14)

K +1 k

k=0

1

ζ(0, x) = − x = −B1 (x)

2

where B1 (x) is the first Bernoulli polynomial (cf. (3.72)).

Γ(−K) = 0 for 0 ≤

K ∈ Z, the proof follows.

Bn+1

If we use ζ(−n) = n+1 for 0 ≤ n ∈ Z, we have (from (4.14)),

BK+1 (x)

ζ(−K, x) = − .

K +1

Now we give a√proof of Lerch’s formula by using the well known facts that

ζ ′ (0) = − log 2π and

1

ζ(s) = − γ + O((s − 1)).

s−1

Proposition 4.2 (Lerch’s formula)

Γ(x)

ζ ′ (0, x) = log √ . (4.15)

2π

Proof. Stark’s method of diﬀerentiating the left-hand side of (4.12) and

appealing to the Weierstrass product for the gamma function may in fact

be simpler.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 114

∫

1 Γ(−w)Γ(s + w)

ζ(s, x) = x−s + ζ(s + w)xw dw

2πi (aK ) Γ(s)

those terms except for the integral, say I, becomes

d

ζ ′ (0, x) = log x + ζ ′ (0) + γx + I . (4.17)

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

dt s=0

′ ′

Clearly, only the term with − ΓΓ2(s)

(s) = −

sΓ (s)−Γ(s+1)

Γ2 (s+1) counts, other terms

1

vanish in view of Γ(0) = 0. Therefore we only need to evaluate

∫

d 1

I = Γ(−w)Γ(w)ζ(w)xw dw

dt s=0 2πi (aK )

∑∞ ∫ ( x )w

1 π

= dw

n=1

2πi (aK ) w sin πw n

with 1 < aK < 2. Moving the line of integration far to the left, counting

the residues and noting that the resulting integral tends to 0, we conclude

that

1 ( x )k

∑∞ ∑ ∞

d

I s=0 = −

dt n=1

k n

k=2

∞

∑ ζ(k) k

= (−x) . (4.18)

k

k=2

Now we recall the formula [Vista, Theorem 3.2, p.62] (also [SC, (2), p.159])

of Ramanujan and Yoshimoto with α = 1, which is

∞

∑ ζ(k) k

(−x) = log Γ(x + 1) + γx, |x| < 1.

k

k=2

d

I = log Γ(x + 1) − γx.

dt s=0

Now substituting this in (4.17) and appealing to the diﬀerence equation

log Γ(x + 1) = log Γ(x) + log x for x > 0, we conclude (4.15).

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 115

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 115

In this section, we state Theorem 4.3, the concrete version of our main

theorem, Theorem 7.1 and give a number of applications. This will be

applied in subsequent chapters.

We have two sets of Dirichlet series

that satisfy the generalized functional equation (4.19) in the following sense.

Let {λk }∞

(h) (i) ∞ (h) ∞ (i) ∞

k=1 , {µk }k=1 denote sequences and {αk }k=1 , {βk }k=1 are

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

∞

∑ (h)

αk

φh (s) = (h)s

k=1 λk

and

∞

∑ (i)

βk

ψi (s) = (i)s

,

k=1 µk

with finite abscissa of absolute convergence σφh and σψi respectively.

Let there exist a meromorphic function χ(s), which satisfies (for a real

number r) the functional equation:

χ(s) (4.19)

M∏ (h) ( ) N∏

(h) ( )

(h) (h) (h)

Γ cj − Cj s

(h)

∑H Γ d j + Dj s

j=1 j=1

P∏(h) ( ) Q∏ (h) ( ) φh (s),

h=1 (h) (h) (h)

Γ dj − Dj s

(h)

Γ cj + Cj s

j=N (h) +1 j=M (h) +1

Re(s) > max (σφh )

1≤h≤H

=

Ñ∏

(i) ( ) M̃∏(i) ( )

(i) (i) (i) (i)

Γ e + E (r − s) Γ f − F (r − s)

∑

I j j j j

j=1 j=1

ψ (r − s),

( ) P̃∏ ( ) i

Q̃∏

(i) (i)

i=1 (i) (i)

Γ fj + Fj (r − s)

(i) (i)

Γ ej − Ej (r − s)

j=M̃ (i) +1 j=Ñ (i) +1

Re(s) < min (r − σψi )

1≤i≤I

( )

(h) (h) (i) (i)

Cj , Dj , Ej , Fj >0 .

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 116

of χ(s) are neither a pole of

N∏

(h) ( )

(h) (h)

Γ cj − Cj s

j=1

P∏

(h) ( ) Q∏

(h) ( )

(h) (h) (h) (h)

Γ cj + Cj s Γ d j − Dj s

j=N (h) +1 j=M (h) +1

nor a pole of

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

∏

M̃ (i) ( )

(i) (i) (i)

Γ fj − F j r + F j s

j=1

.

P̃∏

(i) ( ) ∏

Q̃(i) ( )

(i) (i) (i) (i) (i) (i)

Γ ej − E j r + E j s Γ fj + Fj r − Fj s

j=Ñ (i) +1 j=M̃ (i) +1

|v|→∞

uniformly in u1 ≤ u ≤ u2 .

We choose L1 (s) so that the poles of

∏

n N∏

(h) ( )

(h) (h)

Γ(aj + Aj s − Aj w) Γ cj − Cj w

j=1 j=1

∏

p P∏

(h) ( )

(h) (h)

Γ(aj − Aj s + Aj w) Γ cj + Cj w

j=n+1 j=N (h) +1

1

× ( )

∏

q Q∏

(h)

(h) (h)

Γ(bj + Bj s − Bj w) Γ dj − Dj w

j=m+1 j=M (h) +1

∏

m M∏

(h) ( )

(h) (h)

Γ(bj − Bj s + Bj w) Γ dj + Dj w

j=1 j=1

∏

q Q∏

(h) ( )

(h) (h)

Γ(bj + Bj s − Bj w) Γ dj − Dj w

j=m+1 j=M (h) +1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 117

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 117

1

× ( )

∏

p P∏

(h)

(h) (h)

Γ(aj − Aj s + Aj w) Γ cj + Cj w

j=n+1 j=N (h) +1

lie on the left of L1 (s), and choose L2 (s) so that the poles of

∏

m ∏

M̃ (i) ( )

(i) (i) (i)

Γ(bj − Bj s + Bj w) Γ fj − Fj r + Fj w

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

j=1 j=1

∏

q ∏

Q̃(i) ( )

(i) (i) (i)

Γ(bj + Bj s − Bj w) Γ fj + Fj r − Fj w

j=m+1 j=M̃ (i) +1

1

× ( )

∏

p P̃∏

(i)

(i) (i) (i)

Γ(aj − Aj s + Aj w) Γ ej − Ej r + Ej w

j=n+1 j=Ñ (i) +1

∏

n Ñ∏

(i) ( )

(i) (i) (i)

Γ(aj + Aj s − Aj w) Γ ej + E j r − E j w

j=1 j=1

∏

p P̃∏

(i) ( )

(i) (i) (i)

Γ(aj − Aj s + Aj w) Γ ej − E j r + E j w

j=n+1 j=Ñ (i) +1

1

×

∏

q ∏

Q̃(i) ( )

(i) (i) (i)

Γ(bj + Bj s − Bj w) Γ fj + Fj r − Fj w

j=m+1 j=M̃ (i) +1

lie on the right of L2 (s). Further, they squeeze a compact set S such that

sk ∈ S(1 ≤ k ≤ L).

Under these conditions the χ-function, key-function, X(s, z | ∆) is

defined by (3.2) above.

Then we have the following modular relation (cf. [Ts]) which is a

traditional version of our main Theorem 7.1.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 118

Theorem 4.3

X(s, z | ∆) (4.21)

(

∑ H ∑ ∞ (h)

αk m+M (h),n+N (h) (h)

H zλ k

λk

(h)s p+P (h) ,q+Q (h)

h=1 k=1

(h) (h) (h) (h)

{(1 − aj , Aj )}nj=1 , {(1 − cj + Cj s, Cj )}N j=1 ,

{(bj , Bj )}m

(h) (h) (h) M (h)

j=1 , {(dj + Dj s, Dj )}j=1 ,

)

(h) (h) (h) P (h)

{(aj , Aj )}pj=n+1 , {(cj + Cj s, Cj )}j=N

(h) +1

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

(h)

{(1 − bj , Bj )}qj=m+1 , {(1 − dj + Dj s, Dj )}j=M

(h) (h) Q(h)

(h) +1

if L1 (s) can be taken to the right of max (σφh )

1≤h≤H

( (i)

= ∑ I ∑ ∞ (i)

βk n+Ñ (i),m+M̃ (i) µk

H

(i)r−s q+Q̃(i) ,p+P̃ (i) z

i=1 k=1 µk

{(1 − bj , Bj )}m

(i) (i)

j=1 , {(1 − fj + Fj (r − s), Fj )}j=1 ,

(i) M̃ (i)

(i) (i)

{(aj , Aj )}nj=1 , {(ej + Ej (r − s), Ej )}Ñ

(i) (i)

j=1 ,

)

(i)

(i) (i)

{(bj , Bj )}qj=m+1 , {(fj + Fj (r − s), Fj )}Q̃

(i)

j= M̃ (i) +1

(i) (i) (i) (i)

{(1 − aj , Aj )}pj=n+1 , {(1 − ej + Ej (r − s), Ej )}P̃

j= Ñ (i) +1

∑L ( )

+ Res Γ(w − s | ∆)χ(w) z s−w , w = sk

k=1

if L2 (s) can be taken to the left of min (r − σψi ),

1≤i≤I

We now apply the modular relation (4.21) in the following special case.

values proc.gamma LHSFE RHSFE values

m + M (h) = 1 m=0 M (h) = 1 M (i) = 0 m + M̃ (i) = 0

n + N (h) = 1 n=1 N (h) = 0 N (i) = 1 n + Ñ (i) = 2

p + P (h) = 1 p=1 P (h) = 0 P (i) = 1 p + P̃ (i) = 2

q + Q(h) = 1 q=0 Q(h) = 1 Q(i) = 0 q + Q̃(i) = 0

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 119

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 119

X(s, z | ∆) (4.23)

( )

∑∞ (1 − a , A )

αk 1,1 1 1

H zλ

λ s 1,1 k

(d1 + D1 s, D1 )

k=1 k

if L1 (s) can be taken to the right of max (σφh )

1≤h≤H

( )

= ∑ βk

∞ −

2,0 µk

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

H0,2

µr−s z (a1 , A1 ), (e1 + E1 (r − s), E1 )

k=1 k

∑L ( )

+ Res Γ(a − A (w − s))χ(w) z s−w

, w = s

1 1 k

k=1

if L2 (s) can be taken to the left of min (r − σψi ).

1≤i≤I

We are ready to specify (4.23) with D1 = E1 = A1 . The right-hand side

H-function of the modular relation (4.23) becomes

( ) ( )

(1 − a1 , A1 ) 1 − a1

zλk

1

1,1

H1,1 = A−1

1 G1,1

1,1 (zλk )

A1

d1 + A1 s

(d1 + A1 s, A1 )

d1 1 −a1 −d1

= A−1

1 Γ(1 − a1 + d1 + A1 s)λk (zλk )

s A1

((zλk ) A1 + 1) (4.24)

1

×( 1 )s .

1

A1

λk + (z )−1 A1

(4.24) into

( )

(1, A1 )

1,1

H1,1 z −1 λk

(A1 s, A1 )

1

= A−1

1 Γ(1 + A1 s)λk ( 1

s

)s . (4.25)

1

A1

λk + z A1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 120

E1 = A1 :

( )

−

2,0

H0,2 µk z (4.26)

(a1 , A1 ), (e1 + A1 (r − s), A1 )

( )

−

A1

1

−1 2,0

= A1 G0,2 (µk z) a1 , e1 + A1 (r − s)

1

( 1

)

= 2A−1 1 (µk z)

2 (a1 +e1 +A1 (r−s))

Ka1 −e1 −A1 (r−s) 2(µk z) 2A1 .

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

∑ ∞

1 αk

X(s, z −1 |∆) = Γ(A1 s) ( 1 )A1 s

A1 1

k=1

λkA1 + z A1

∞

∑ βk ( 1

)

= 2A1 −1 z 2 A1 (r−s)

1

1 KA1 (r−s) 2(µk z) 2A1

µk 2 A1 (r−s)

k=1

∑

L ( )

+ Res Γ(A1 (s − w))Γ(A1 s)φ(s) z w−s , w = sk . (4.28)

k=1

Example 4.2 A special case of Corollary 4.1 is the Dirichlet series sat-

isfying the Riemann type functional equation with A1 = 21 , i.e.,

( ) ( )

1 1

Γ s φ(s) = Γ (r − s) ψ(r − s) (4.29)

2 2

with a simple pole at s = r > 0 with residue ρ. Thus from (4.28) we have

() ∞ ∞

1 ∑ αk 1

∑ βk

2Γ s = 4z 2 (r−s) K 12 (r−s) (2µk z)

1 1

2 2 2 2 s 2 (r−s)

k=1 (λk + z ) k=1 µk

∑

L ( (1 ) ( )

1 )

+ Res Γ (s − w) Γ w φ(w) z w−s , w = 0, r

2 2

k=1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 121

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 121

or

( ) ∞

1 ∑ αk

2Γ s 1

2 2 2 2s

k=1 (λk + z )

∑∞ ( )

βk 1

s φ(0) z −s

1

(r−s)

= 4z 2 1 K 12 (r−s) (2µk z) + 2Γ

(r−s) 2

k=1 µk

2

( )

1

+ ρΓ (s − r) z r−s . (4.30)

2

√

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

√

πk, r = 1 and ρ = 1, so that it satisfies (4.29) with φ(s) = ψ(s) =

− 2s

π ζ(s). Hence expressing the left-hand side of (4.30) as the sum over all

integers, we have

( ) ∑

1 1

Γ s 1 (4.31)

2 2 2 2s

0̸=k∈Z (πk + z )

∑∞ ( )

1 √ 1

s ζ(0) z −s .

1

= 4z 4 (1−s) 1 K 1

2 (1−s)

(2 πkz) + 2Γ

k 2 (1−s) 2

k=1

That is,

∑ 1

1

s

k∈Z(πk 2 + z 2 ) 2

∞

z 4 (1−s) ∑ √

1

1

= 4 (1 ) 1 K 1

(1−s) (2 πkz), (4.32)

Γ 2 s k=1 k 2 (1−s) 2

Example 4.3 Another more special case of Corollary 4.1 is the Dirichlet

series satisfying the Hecke type functional equation with A1 = 1, i.e.,

∞

∑ ∑ βk ∞

αk 1

2 (r−s)

√

Γ (s) s = 2z 1

(r−s)

Kr−s (2 µk z)

(λk + z)

k=1 µk

2

k=1

( )

+ Res Γ (s − w) χ(w) z w−s , w = r ,

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 122

which becomes

∞

∑ αk

Γ (s) s (4.34)

(λk + z)

k=1

∞

∑

1 βk √

= 2z 2 (r−s) 1 Kr−s (2 µk ) + ρΓ (s − r) z r−s .

2 (r−s)

k=1 µk

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

of a number field Ω with degree κ = r1 + 2r2 ≤ 2 and discriminant ∆. For

standard notation we refer to §3.4.1.

We recall the functional equation (3.1) as

(1 ) (1 1 )

Γ r1 s Γr2 (s)φ(s) = Γr1 − s Γr2 (1 − s)φ(1 − s) (4.35)

2 2 2

Now Corollary 4.1 with A1 = 1 gives

∞

∑ α(k)

Γ (s) s

(Ak + z)

k=1

1

∞

∑ α(k) ( √ )

= 2z 2 (1−s) 1 K1−s 2 Akz

2 (1−s)

k=1 (Ak)

+ Γ (s) φ(0) z −s + ρΓ (s − 1) z 1−s . (4.36)

z

Now if we replace A = ω with Re ω > 0, we get

∞

∑ α(k)

A−s s

(k + ω)

k=1

2

∞

∑ α(k) ( √ )

2A kω + A−s φ(0) ω −s

1

= ω 2 (1−s) 1 K 1−s

Γ(s) k 2 (1−s)

k=1

Γ (s − 1) 1−s

+ A1−s ρ ω ,

Γ(s)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 123

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 123

which leads to

∞

∑ α(k)

s (4.37)

(k + ω)

k=1

2

∞

∑ α(k) ( √ ) 1

2A kω + φ(0) ω −s + Aρ ω 1−s

1

= As ω 2 (1−s) K1−s .

Γ(s) 1

k 2 (1−s) s−1

k=1

Therefore in the case of an imaginary quadratic field one gets from (4.37)

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

to the Fourier-Bessel series

∞

∑ α(k)

s

(k + ω)

k=1

2 1

∞

∑ α(k) ( √ )

s (1−s)

=A ω 2

1 K1−s 2A kω

Γ(s) k 2 (1−s)

k=1

−s 1

+ ζΩ (0) ω + Aρ ω 1−s . (4.38)

s−1

We shall now show that Proposition 4.3 gives the Lipschitz summation

formula as in Koshlyakov [KoshII, p.245-246], where he takes a weighted

sum of ζΩ (s) (one reason may be to cancel the pole at s = 1). The essential

reason being discussed in more detail in §4.5. For a character analogue

of the Lipschitz summation formula, (cf. Berndt [Ber75-2]), as in [KoshII,

p.243], we consider for 0 < Re ω < 1 (in [KoshII, p.243], the limiting case

Re ω = 0, i.e. ω = iw, 0 < w < 1 is considered)

∞ ∞

1 πis ∑ α(k) 1 − πis ∑ α(k)

e 2

s+ e

2

s

2 (k+iw) 2 (k−iw)

k=1 k=1

∞

1 2 ∑ α(k)

= As 1 (4.39)

2 Γ(s) k 2 (1−s)

k=1

( )

πis 1

( √ ) −πis 1

( √ )

2 (1−s) (1−s)

× e 2 (iw) K1−s 2A ikw + e 2 (−iw) 2 K1−s 2A −ikw

+ ζΩ (0) w−s .

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 124

πi

If one writes ε = e 4 , then the right-hand side of (4.39) becomes

∞

1 s 2w 2 (1−s) ∑ α(k)

1

= A 1

2 Γ(s) k 2 (1−s)

k=1

( )

( √ ) ( √ )

εs+1 K1−s 2Aε kw + ε̄s+1 K1−s 2Aε̄ kw

ζΩ (0)

+ . (4.40)

ws

At this point, in conformity with Koshlyakov [KoshII, p.241] (cf. (3.40)),

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

∞

2ζΩ (0) ∑ α(k)

ζΩ (s, ω) = − + s. (4.41)

w (k + ω)

k=1

∞

∑ ′ α(n)

2 s ζ (s, iw)+e− 2 s ζ (s, −iw)

πi πi

s =e Ω Ω (4.42)

n=−∞

(w+in)

( )

2 (1−s) ∑

1 ∞ ( √ ) s+1 ( √ )

sw α(k) s+1

=A 1 ε K1−s 2Aε kw + ε̄ K1−s 2Aε̄ kw

Γ(s) k 2 (1−s)

k=1

ζΩ (0)

− .

ws

This is the corrected form of [KoshII, (23.15)] and here the prime on the

summation sign means that the term corresponding to n = 0 is excluded.

The Dedekind zeta-function of a real quadratic field satisfies the func-

tional equation (3.41) with r1 = 2, r2 = 0:

(1 ) (1 1 )

Γ2 s φ(s) = Γ2 − s φ(1 − s) (4.43)

2 2 2

where φ(s) is defined by (3.40).

1,1

In this form, there would be no proper modular relation with H1,1 unless

we combine the perturbed Dirichlet series, which we shall state in §4.5.

Here we view (4.43) to mean

( ) ( )

Γ 12 s Γ 1−s

2 s

( ) φ(s) = ( ) φ(1 − s). (4.44)

Γ 1−s

2 s Γ 1−(1−s)

2 s

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 125

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 125

values proc.gamma LHSFE RHSFE values

m + M (h) = 1 m=0 M (h) = 1 M (i) = 0 m + M̃ (i) = 0

n + N (h) = 1 n=1 N (h) = 0 N (i) = 1 n + Ñ (i) = 2

p + P (h) = 1 p=1 P (h) = 0 P (i) = 2 p + P̃ (i) = 3

q + Q(h) = 1 q=0 Q(h) = 1 Q(i) = 0 q + Q̃(i) = 0

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

φ(s) = ψ(r − s), (4.45)

Γ(d2 − D2 s) Γ(e2 − E2 (r − s))

X(s, z | ∆) (4.46)

( )

∞

∑ αk 1,1

(1 − a1 , A1 )

H1,1 zλk

λs (d1 + D1 s, D1 )

k=1 k

if L1 (s) can be taken to the right of σφ

(

∑∞ −

βk 2,0 µk

H

= k=1 µk r−s 0,3

z (a1 , A1 ), (e1 + E1 (r − s), E1 ),

)

−

(1 − e2 + E2 (r − s), E2 )

∑

L ( )

+

Res Γ(a − A (w − s))χ(w) z s−w

, w = s

1 1 k

k=1

if L2 (s) can be taken to the left of r − σψ .

(

−

2,0

H0,3 µk z

(a1 , A1 ), (e1 + (1/2)A1 (r − s), (1/2)A1 ),

)

( )

−

2,0

= H0,3 µk z

(0, 1), ( 1 (1 − s), 1 ), ( 1 + 1 (1 − s), 1 ) . (4.47)

2 2 2 2 2

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 126

We may work with a slightly more general form of the right-hand side

of the modular relation

( )

− (

2,0

H0,3 z ( 1 ) ) (4.48)

s, 2 A1 , (a1 , A1 ), s + 12 , 12 A1

( ( ) )

s + 1 1

, A

3,0

= H1,4 z ( 1 ) ( )

2 2 1 ( )

s, 2 A1 , s + 12 , 12 A1 , (a1 , A1 ), s + 12 , 12 A1

( ( ) )

√ −s 2,0 A1 s + 21 , 12 (A1

= 2 π 4 H1,3 2 z )

(2s, A1 ), (a1 , A1 ), s + 21 , 12 A1

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

{ ( )

4−s −

2A1 e− 2 A1 i z

1 π

=√ 2,0

e(s+ 2 )πi H0,2

πi (2s, A1 ), (a1 , A1 )

( )}

−

−e H0,2 2 e 2 z

−(s+ 21 )πi 2,0 A1 π A 1 i

(2s, A1 ), (a1 , A1 )

{ ( )

4−s 1

(s+ 12 )πi 2,0 −π i A1 −

=√ e G0,2 2 e 2 z

π iA1 2s, a1

( )}

1

−(s+ 12 )πi 2,0 π

i A1 −

−e G0,2 2 e z

2

2s, a1

( π 1 )s

2 { − π i A1 a1 e− 2 i z A1 ( √ 1

)

K2s−a1 2 2 e− 4 z 2A1

πi

=√ (2e 2 z 1)

πA1 2

( )s

π 1 a1

π 1

e 2 i z A1 ( √ πi 1 ) }

+ (2e 2i z A1

) K2s−a1 2 2 e 4 z 2A1 ,

2

2 and

A1 = 1 takes the form:

( )

− (

2,0

H0,3 z ( 1−s 1

) )

2 ,2 , (a1 , 1), 1−s 1 1

2 + 2, 2

( )1−s

2 { e− 2 i z ( πi √

)

π

2

a1

= √ (2e− 2 i z) K1−s−a1 2 e− 4 2z

π

π 2

( π )1−s ( πi √ ) }

π a1 e 2 iz 2

2

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 127

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 127

( )

−(

2,0

H0,3 z ( 1−s

1

) 1−s 1 1

) (4.50)

2 , 2 , (0, 1), 2 + 2, 2

( π )1−s

2 { e− 2 i z 2 ( πi √

)

=√ K1−s 2 e− 4 2z

π 2

( π i )1−s ( πi √ ) }

e2 z 2

+ K1−s 2 e 4 2z .

2

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

sequences get squared (since we must choose E1 = 2 in (4.45), we are forced

to choose A1 = 21 ).

case

We elucidate the reason why the K-Bessel function appears for the real

quadratic field Ω. Let us briefly recall his method. Koshlyakov considers

the integral with the integrand being of the form

( )

Γ2 1−z

( 2)

Γ(z)φ(z) = Γ(z) 2 z φ(1 − z) (4.51)

Γ 2

tion formula to the first factor and cancels Γ z2 in the denominator.

( ) sin π2 z

Then he writes Γ 1z as Γ 1 − z2 π and also expresses the product

( z 1 ) ( 1−z ) ( 2 ) ( )

Γ 2+2 Γ 2 as sin ππ1−z . Then finally, he combines Γ 1−z 2 and

( ) −(1−z) 12

2

Γ 1 − 2 to give 2

z

π Γ (1 − z). It follows that

π

Γ(z)ζΩ (z) = tan zΓ(1 − z)ζΩ (1 − z) (4.52)

2

it suggests the following

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 128

divisor problem

ζΩ (s, ω)

∞ ∫ ( )−z

ω −s ∑ α(k) 1 π 1 (4.53)

= tan zΓ(1 − z)Γ(s − z) dz

Γ(s) k 2πi (c) 2 kω

k=1

∞

1 ∑ α(n) 1 πi 1 πi

= e 2 s ζΩ (s, iw) + e− 2 s ζΩ (s, −iw) (4.54)

2 n=−∞ (w + in)s 2 2

(

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

∞

2 (1−s) ∑

1

εs+1 ( √ )

sw α(k)

=A 1 K1−s 2Aε kw

Γ(s) k 2 (1−s) i

k=1

)

ε̄s+1 ( √ )

− K1−s 2Aε̄ kw .

i

This in fact is the corrected form of [KoshII, (23.16)]. Proving the above re-

sult amounts to showing that, in contrast to the case of imaginary quadratic

fields in Proposition 4.3, the left-hand side cancels the denominator cos π2 z,

which is given in

ζΩ (s, iw) + e−

πi πi

e 2 s 2 s ζΩ (s, −iw). (4.55)

wz π

sin zΓ(1 − z)ζΩ (1 − z)

π 2

and so applying the inverse Heaviside integral (2.109) one can deduce (4.54).

( )

Γr1 2s Γr2 (s)

G(1 − s) = r ( 1−s ) r (4.56)

Γ 1 2 Γ 2 (1 − s)

and consider the kernel, which we call Koshlyakov’s K-function

∫

1 π G(1 − s)

K(x) = Kr1 ,r2 (x) = (π ) ds (4.57)

2πi (c) 2 cos 2 s xs

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 129

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 129

∞

A∑ ( )

σ(x) = σr1 ,r2 (x) = F (n)Kr1 ,r2 A2 xn (4.58)

π n=1

x > 0 (Re(x) > 0). He transforms the series (4.58) by Cauchy’s residue

theorem and apply in the last stage a form of the beta transform (4.2)

([KoshI, (3.3)])

∫

1 π dw 1

(π ) w =

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

2πi (c) 2 sin 2 w α 1 + αs

∞

1 ζΩ (0) 1 x ∑ α(n)

σ(x) = − ρ − + (4.60)

2 π x π n=1 n2 + x2

for Re x > 0, where ρ is defined in (3.38). Thus one gets the closed forms

for the K-function:

√ −2x

K1,0 (x) = πe . (4.61)

( √ )

x 1( ( √ ) ( √ ))

K0,1 (x) = −2 kei0 4 = K0 2ε x − K0 2ε̄ x (4.62)

4 i

√ ( ( √ ) ( √ ))

K2,0 (x) = 4 ker0 (4 x ) = 2 K0 ε x + K0 ε̄ x , (4.63)

where kei0 and ker0 are modified Kelvin functions ([ErdH, p.6], [PBM]) and

ε = e 4 i , ε̄ = e− 4 i .

π π

(4.64)

The main diﬀerence is that we express the resulting functions as Kelvin

functions.

We turn to the elucidation of Koshlyakov’s K-function (4.57). We dis-

tinguish three cases.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 130

In the case of Q one has r1 = 1 and (4.60) amounts to Example 4.2 and

(4.61) corresponds to the degenerate case of the K-Bessel function.

1 1

∫ (s) (s 1)

K1,0 (x) = Γ Γ + x−s ds

2 2πi(c) 2 2 2

( )

1 2,0 −

= H0,2 x (4.65)

2 (0, 21 ), ( 12 , 12 )

( )

2,0

2 − 1

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

= G0,2 x 1 = 2x 2 K 21 (2x),

0, 2

where in the last step we applied (2.109), Ks (x) being the K-Bessel func-

tion defined e.g. by (2.31) (cf. [Wat]). By (2.32), the right-hand side of

√

(4.65) further reduces to π e−2x , and we have (4.61) ([KoshI, (2.4)]).

Koshlyakov’s formula [KoshI, (2.5)] (except for the middle member in

(4.62)):

1( ( √ ) ( √ ))

K0,1 (x) = K0 2ε̄ x − K0 2ε x . (4.66)

i

Since

( )

1 2,1 (1, 1)

K0,1 (x) = H1,3 x 1 1 2 2 . (4.67)

2 ( 2 , 2 ), (0, 1), (0, 1)

( )

1 3,1 x ( 12 , 12 )

K0,1 (x) = H1,5

4 4 ( 12 , 12 ), (0, 12 ), ( 21 , 12 ), (0, 21 ), ( 12 , 21 )

( )

1 3,0 x −

= H0,4 (4.68)

4 4 (0, 12 ), ( 12 , 12 ), ( 21 , 12 ), (0, 21 )

( )

1 4,0 x (0, 12 )

= H1,5 .

4 4 (0, 12 ), ( 12 , 12 ), (0, 12 ), ( 21 , 12 ), (0, 12 )

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 131

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 131

Hence by (2.101)

{ ( )

e− 2 i x

π

1 4,0 −

K0,1 (x) = H0,4

8πi 4 (0, 21 ), ( 12 , 12 ), (0, 21 ), ( 12 , 21 )

( πi )}

4,0 e 2 x

−

− H0,4

4 (0, 12 ), ( 12 , 12 ), (0, 21 ), ( 12 , 21 )

{ ( ) ( )}

1 − −

−π2 ix 2 ix

π

= G2,0

0,2 e 0, 0 − G 2,0

0,2 e 0, 0 . (4.69)

2i

In the last step we applied (2.96) and then (2.93). The equation (4.69)

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

Exercise 4.1 Indicate the concrete procedure for the above computation.

Solution Applying (2.3) to the integrand, we deduce the second equality of

Γ( w )

(4.68). Multiplying the integrand by Γ w2 , we see that

(2)

∫ ( ( ) ( ))2

1 22w−2 Γ w2 Γ w+1

K0,1 (x) = ( ) ( ) 2

x−w dw (4.70)

2πi (c) Γ 1 − w2 Γ w2

∫ ( ) ( ( ) ( ))2

1 sin π2 w 2w−2 w w+1

= 2 Γ Γ x−w dw

2πi (c) π 2 2

− e− 2 s

πi πi

e 2 s

Γ(w)2 .

2i

Hence

( ∫

1 1

Γ(w)2 22w−1 (4e− 4 x)−w dw

πi

K0,1 (x) = (4.71)

i 2πi (c)

∫ )

1

Γ(w)2 22w−1 (4e x)−w dw

πi

− 4

2πi (c)

( ) ∫

2,0 u − 1 u −s (√ )

G0,2 = Γ2 (s) ds =2K0 u , (4.72)

4 0, 0 2πi (c) 4

Re(c) > 0,

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 132

We observe that we are to transit from (4.68) to

( )

1 3,0 ( x )2 −

K0,1 (x) = G0,4 . (4.73)

2 4 0, 12 , 12 , 0

We notice that the right-hand side is one of the modified Kelvin’s func-

tions:

( √ )

1 x

· (−4) kei0 4

2 4

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

Modified Kelvin functions satisfy the relation

( )

kerν (z) ± ikeiν (z) = e∓ 2 π Kν ze± 4 ,

iν πi

(4.74)

whence

( πi ) ( )

2ikeiν (z) = e− Kν ze 4 − e 2 i Kν ze− 4 .

νπ νπ πi

2 i (4.75)

( √ )

x 1( ( √ ) ( √ ))

K0,1 (x) = −2 kei0 4 = K0 2ε x − K0 2ε̄ x . (4.76)

4 i

Note that in deriving (4.76), we have given a proof of (4.75).

∫ ( ) ( )

1 1 Γ2 w2 Γ 1+w

K2,0 (x) = ( )2

x−w dw. (4.77)

2πi (c) 2 Γ 1−w

2

( )

1 3,0 −

K2,0 (x) = H0,4 x

2 (0, 21 ), (0, 12 ), ( 21 , 12 ), ( 12 , 21 )

( )

2 −

= G3,0

0,4 x 0, 0, 1 , 1

2 2

√

by (2.93), which is 4 ker0 (4 x ) ([PBM, 13, p.675]).

By (4.74), we have

( πi ) ( )

2 kerν (z) = e− 2 Kν ze 4 + e 2 Kν ze− 4 ,

νπi νπi πi

(4.78)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 133

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 133

( )

2 −

K2,0 (x) = G3,0

0,4 x 0, 0, 1 , 1

2 2

√ (4.79)

= 4 ker0 (4 x )

( ( √ ) ( √ ))

= 2 K0 ε x + K0 ε̄ x ,

Exercise 4.2 As in Exercise 4.1, work out the above formal computation.

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

( ( ) ( ) − 1 w−1 )2 2−2w

w w+1

1 Γ 2 Γ 2 π 22 π2

( 1−w ) ( 1+w ) x−w

2 Γ 2 Γ 2

(π )

= 2 cos w Γ2 (w)2−2w x−w

2

by (2.3) and (2.4).

Then we use Euler’s identity to obtain

( ∫

1 ( πi

)−w

K2,0 (x) = 2 22w−1 Γ(w)2 42 e 2 x dw

2πi (c)

∫ )

1 ( )−w

2w−1 2 2 − πi

+ 2 Γ(w) 4 e 2 x dw

2πi (c)

( ( √ ) ( √ ))

= 2 K0 4ε x + K0 4ε̄ x

1,1 ↔ G0,2

Apparently, the functional equation of Hecke’s type has been most exten-

sively studied and it has created such tremendous amount of results that

it is natural to dwell on this case rather fully and more thoroughly than in

other cases.

Γ(s) φ(s), Re(s) > σφ ,

χ(s) = (4.80)

Γ(r − s) ψ(r − s), Re(s) < r − σ .

ψ

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 134

∑∞ ( ) ∑ ( )

βk

1,1 µk 1 − a

L

G + Res Γ(a − s + w) χ(w) z s−w , w = sk

µ r−s 1,1

z r−s

k=1 k k=1

∞

∑ ( )

αk 2,0 −

= G zλk . (4.81)

λsk 0,2 s, a

k=1

∞

∑ αk ( √ )

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

s+a

2 s−a z 2 Ks−a 2 zλk

k=1 λk 2

∞

∑ za

= Γ(a + r − s) βk a+r−s

k=1

(µk + z)

∑

L ( )

+ Res Γ(a − s + w) χ(w) z s−w , w = sk .

k=1

the perturbed Dirichlet series (or K-Bessel expansion):

for the perturbed Dirichlet series

∞

∑ 1

Γ(r − s) βk r−s

(µk + z)

k=1

∑ αk s ( √

∞ )

=2 s z 2K

s 2 zλ k (4.82)

k=1 λk

2

∑

L ( )

− Res Γ(−s + w) χ(w) z s−w , w = sk

k=1

Theorem 4.8 was first proved by Berndt [Be1] and is stated as [BK,

Theorem 8.3, p.120]. In [BeIII] another proof is given and in [BeIV] a

general form (with multiple gamma factor) is given. It is stated as [KTY7,

(1.6)], one of the equivalent assertions to the functional equation ([KTY7,

Theorem 1]) (for prototype of these, cf. Exercises 1.1 and 1.2). For its

history, cf. §5.3 [KTY7]. As is stated there, the most essential ingredient

in the Fourier-Bessel expansion is the beta transform or the beta integral

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 135

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 135

Barnes integral in [Vista, p.126] and its special case appeared also on [Vista,

p.34, p.98 and p.196]. Since the Mellin-Barnes integrals refer to a much

wider class of special functions ([ErdH, p.49]), the Fox H-functions, we

shall speak about the beta transform (or beta integral) in the same spirit

as the gamma transform ((1.42), cf. [Vi2, p.25]). In the present book, it is

(2.108), which we restate in a specialized and concrete form

∫

1 Γ(s − z) Γ(z) −z

(1 + x)−s = x dz (4.83)

2πi (c) Γ(s)

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

for x > 0, 0 < c < σ. This has been used extensively in the context of the

perturbed Dirichlet series

∑

φ(s, a) = an (λn + a)−s (a > 0, σ ≫ 0). (4.84)

[Mat] and [PK], which were referred to in [KTY7], cf. also [HarMB] and

[KoshI]-[KoshIII]). Also Katsurada and Matsumoto made an extensive use

of the beta-transform in dealing with mean square of the Hurwitz type

zeta-functions [KatM1]-[KatM3]. [KoshI, p. 116, p. 130], [KoshII, p.243]

extensively uses a special case of (4.83) in the form ([Vi2, p.35])

∫

1 π 1

x−z dz = , x > 0, 0 < c < 2. (4.85)

2πi (c) 2 sin π2 z 1 + x2

form of the integral (4.83) G1,1

1,1 , applying the functional equation,

and then finally appealing to the inverse Heaviside integral G2,0 0,2

or to some more general G-function according to the form of the

processing gamma factor.

∞ ∞

√ ∑ α √ ( )∑ βk

π √ k e−2 zλk = Γ r − 21

λk r− 12

k=1 k=1 (µk + z)

∑

L ( ( ) )

Res Γ − 12 + w χ(w) z 2 −w , w = sk ;

1

+

k=1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 136

∞ ∞

√ −1 ∑ √ ( )∑ βk

πz 2 αk e−2 zλk = Γ r + 12 r+ 21

(4.86)

k=1 k=1 (µk + z)

∑

L ( ( ) )

Res Γ 12 + w χ(w) z − 2 −w , w = sk .

1

+

k=1

Exercise 4.5 Reviewing the proof of [Vi2, Proposition 2.1, p. 35], check

that the beta transform (4.2) is an analytic expression for the binomial

expansion.

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

we introduce (

new notation.

)

A B

Let Y = t be a block decomposition with A an n × n matrix

BC

and B an m × m matrix. Set

D = C − t BA−1 B.

( In) accordance

( ) with this decomposition, we decompose the vectors g =

g1 h1

,h = , g 1 , h1 ∈ Zn , g 2 , h2 ∈ Zm .

g2 h2

der the above notation, we have

Λ(Y, g, h, s) (4.87)

1 ( n )

= δ(g 2 ) e−2πig2 ·h2 Λ(A, g 1 , h1 , s) + δ(h1 ) √ Λ D, g 2 , h2 , s −

|A| 2

2e −2πig 1 ·h1 ∑ ∑ −1

+ √ e2πi(−g1 ·a+h2 ·b) e−2πiA B(b+g2 )·(a+h1 )

|A| a∈Zn b∈Zm

a+h1 ̸=z b+g 2 ̸=z

√ s− n

A−1 [a + b1 ]

2

( √ )

× Ks− n2 2 A−1 [a + h1 ] D[b + g 2 ] π .

D[b + g 2 ]

[SC] (cf. also Bateman and Grosswald [BG]), the case m = 1 is due to

Berndt [BeVI] and the general case with g = h = z is due to Terras [Te2,

Example 4, p.208].

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 137

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 137

∑′ 1

Z(s) = s (4.88)

m,n

(am2 + 2bmn + cn2 )

for σ > 1, where m, n run through integer values except for (m, n) = (0, 0)

and a, c > 0 and the discriminant ∆ = 4(b2 − ac) < 0. Then

√ ( )

22s as−1 π 1

Γ(s)Z(s) = 2Γ(s)ζ(2s)a−s + s− 1

Γ s − ζ(2s−1)+Q(s), (4.89)

|∆| 2 2

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

where

∞

( √ )

4π s 2s− 2 ∑ s− 1

1

b π |∆|

Q(s) = √ s

− 1 n 2 σ1−2s (n) cos nπ Ks− 12 n . (4.90)

a|∆| 2 4 n=1 a a

√ √

|∆| b d |∆|

d= = ac − b , x = , y =

2

= > 0. (4.91)

4 a a 2a

Then

( )1−s

2s s−1 √y

2 a d

s− 12

=2 1

|∆| ds− 2

and

∞

4π s 2s− 2 ∑ s− 1

1

2 − 4 n=1

s 1

d

y 4d

The special case of s = 1 of Corollary 4.2 gives rise to the Kronecker limit

formula in the notation of Theorem 1.2.

√2π

|∆|

lim Z(s) −

s→1 s−1

( )

2π a 4π

=√ 2γ + log −√ log η(ω1 )η(ω2 ). (4.93)

|∆| |∆| |∆|

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 138

Eisenstein series is defined by

∑′ 1

E(z; α) = as Z(s) = s, (4.94)

m,n

|m + nα|

for σ > 1, where m, n run through integer values except for (m, n) = (0, 0)

(a > 0) and α = x + iy with x, y in (4.91).

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

√

2 πΓ(z − 12 )

E(z; α) = 2ζ(2z) + ζ(2z − 1)y −2z+1 (4.95)

Γ(z)

∞

4 ∑ cos(2πmnx)

+ (ny)−2z Im,n (z, α),

Γ(z) m,n=1 m

( )

1 2,0

2 −

Im,n (z, α) = √ G0,2 (πmny) . (4.96)

π z, 12

0,2 and so

Corollary 4.2 above

z+ 12

Im,n (z, α) = 2π z (mny) Kz− 12 (2πmny), (4.98)

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 139

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 139

( )

2 − z+ 12

G2,0

0,2 x z, 1 = 2x Kz− 12 (2x) (4.100)

2

√

= πxz W0,z− 12 (4x)

( )

√ 1

1

= 2 πxz+ 2 e2x G2,0 4x 2

(4.101)

1,2

z − 1, − z + 1

2 2

√

πe−2x (2x)2z U (z, 2z; 4x)

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

= (4.102)

√ −2x z ( )

πe x 1

= G1,2 4x

2,1

. (4.103)

Γ(z)Γ(1 − z) z, 1 − z

Our task is now to ascend the hierarchy from (4.100) to (4.103) to prove

Theorem 4.11. Before turning to the proof, we give the Katsurada’s beta-

transform (cf. (4.2)). By the diﬀerence equation we have for any x and

K∈N

Γ(K + x)

Γ(x) = (4.104)

(K − 1 + x)(K − 2 + x) · · · (1 + x)x

(−1)K Γ(K + x)Γ(−K + 1 − x)

=

Γ(−x + 1)

(−1)K Γ(K + x)

= B(−K + 1 − x, K),

(K − 1)!

∫ 1

Γ(x)Γ(y)

B(x, y) = = tx−1 (1 − t)y−1 dt.

Γ(x + y) 0

Hence

∫ 1

(−1)K Γ(K + x)

Γ(x) = t−K−x (1 − t)K−1 dt, (4.105)

(K − 1)! 0

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 140

Theorem 4.11

√

2 π Γ(z − 12 )

E(z; α) = 2ζ(2z) + ζ(2z − 1)y −2z+1 (4.106)

Γ(z)

∞

4 ∑ cos(2πmnx) −2πmny

+ (ny)−2z e

Γ(z) m,n=1 m

∑

K−1

(−1)k

× (z)k (1 − z)k (πmny)−k+z

22k k!

k=0

∞

∑

4 cos(2πmnx) (−1)K −2πmny (2πmny)2z

(ny)−2z

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

+ e

Γ(z) m,n=1 m (K − 1)! Γ(z)Γ(1 − z)

∫ 1 ( )

4πmny 1 − z − K −K−z

× G2,1

1,2 0, 1 − 2z t (1 − t)K−1 dt.

0 t

Proof of Theorem 4.11 To make a clear comparison with Katsurada’s rea-

soning ([Kat5]), we use (4.97) in what follows (we may equally use other

expressions in Theorem 4.10). We assume that 0 < Re z = ξ < 1 and the

line of integration (c′1 ) satisfies

∫

1 Γ(s + z)Γ(−s)Γ(1 − 2z − s)

U (z, 2z; 4πmny) = (4πmny)s ds.

2πi (c′ ) Γ(z)Γ(1 − z)

Moving the line of integration to the left up to Re s = cK we encounter

the residues at s = −k − z, k = 0, 1, . . . , K − 1, where −ξ − K < cK <

−ξ − K + 1, K ∈ N. The sum of residues is

∑

K−1

(−1)k Γ(k + z)Γ(1 + k − z)

(4πmny)−k−z .

k!Γ(z)Γ(1 − z)

k=0

Hence

∑

K−1

(−1)k

= (z)k (1 − z)k (4πmny)−k−z

k!

k=0

∫

1 Γ(s + z)Γ(−s)Γ(1 − 2z − s)

+ (4πmny)s ds.

2πi (cK ) Γ(z)Γ(1 − z)

The restriction on ξ can be relaxed to −K < ξ < K + 1.

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 141

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 141

∑

K−1

(−1)k

= (z)k (1 − z)k (4πmny)−k−z

k!

k=0

∫

1 Γ(s + z + K)Γ(−s)Γ(1 − 2z − s)

+

2πi (cK ) Γ(z)Γ(1 − z)

∫ 1

(−1)K

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

(K − 1)! 0

∑

K−1

(−1)k

= (z)k (1 − z)k (4πmny)−k−z

k!

k=0

∫ 1

(−1)K 1

+ t−K−z (1 − t)K−1 I(t) dt,

(K − 1)! Γ(z)Γ(1 − z) 0

where

∫ ( )s

1 4πmny

I(t) = Γ(s + z + K)Γ(−s)Γ(1 − 2z − s) ds.

2πi (cK ) t

Since in the integral I(t) all the poles of Γ(s + z + K) are to the left side

and the all poles of Γ(−s) and Γ(1 − 2z − s) are to the right side of the line

cK , it is the G-function:

( )

4πmny 1 − z − K

I(t) = G2,1

1,2

t 0, 1 − 2z .

Hence

∑

K−1

(−1)k (−1)K 1

= (z)k (1 − z)k (4πmny)−k−z +

k! (K − 1)! Γ(z)Γ(1 − z)

k=0

∫ 1 ( )

4πmny 1 − z − K

× t−K−z (1 − t)K−1 G2,1

1,2 0, 1 − 2z dt,

0 t

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 142

which leads to

∑

K−1

(−1)k

−2πmny

=e (z)k (1 − z)k (πmny)−k+z

22k k!

k=0

K

(−1) (2πmny)2z

+ e−2πmny

(K − 1)! Γ(z)Γ(1 − z)

∫ 1 ( )

2,1 4πmny 1 − z − K −K−z

× (1 − t)K−1 dt,

by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

G1,2 0, 1 − 2z t

0 t

[ChN2] (along with [ChN3] for the multiple gamma case) seems to be the

most fundamental work on equivalent formulations of the Hecke type func-

tional equation, centering around (the summability of) the Riesz sums,

whose treatment was excluded in [KTY7]. We make a closer analysis of the

Riesz sums, incorporating the work of Anna Walfisz in §9.1.2, and in the

following example we elucidate the formula in [ChN2, Lemma 6] which was

proved in a more general form in Bochner and Chandrasekharan [BCh1].

Its prototype is due to Hardy [HarMB], [HarRam]. For the completion of

Hardy’s argument [HarMB], see [CZS2].

[ChN2]) In this example we use the residual function for the sum of

residues.

old λk µk s z

√ s2

new 2πλk 2πµk 2 2πz 8π

∞

∑ ( )ν2 ( √ )

αk s2

2 ν Kν s λk

(2πλ k) 2 8π

k=1

October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 143

1,1 2,0

Fourier-Bessel Expansion H1,1 ↔ H0,2 143

∞

∑ (8π)r−ν

= Γ(r − ν) βk r−ν

(s2 + 16π 2 µk )

k=1

∫ ( 2 )ν−z

1 s

+ Γ(z − ν)χ(z) dz, (4.110)

2πi C 8π

where C is a curve encircling all of sk , 1 ≤ k ≤ L.