Anda di halaman 1dari 316

8711_9789814449618_TP.

indd 1 20/10/14 12:07 pm


Series on Number Theory and Its Applications ISSN 1793-3161

Series Editor: Shigeru Kanemitsu (Kinki University, Japan)

Editorial Board Members:


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


LaiFun - Contributions to the Theory of Zeta.indd 1 6/11/2014 10:06:14 AM


World Scientific

8711_9789814449618_TP.indd 2 20/10/14 12:07 pm


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

Library of Congress Cataloging-in-Publication Data


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

British Library Cataloguing-in-Publication Data


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

Copyright © 2015 by World Scientific Publishing Co. Pte. Ltd.


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

LaiFun - Contributions to the Theory of Zeta.indd 2 6/11/2014 10:06:14 AM


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

To Professor Kok Khoo Phua

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

This page intentionally left blank


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

Preface

This is a book on zeta-functions as viewed from their symmetry—functional


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

viii Contributions to the Theory of Zeta-functions

Chapter 5 is devoted to the Ewald expansion, i.e. hypergeometric expan-


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. Grocery of Special Functions 29


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

3. Unprocessed Modular Relations 61


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

x Contributions to the Theory of Zeta-functions

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. The Ewald Expansion or the Incomplete Gamma Series 145


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 Atkinson-Berndt Abel mean . . . . . . . . . . . . . . . . 154


5.2.1 Landau’s exposition . . . . . . . . . . . . . . . . . 156
5.2.2 Screened Coulomb potential . . . . . . . . . . . . 158

6. The Riesz Sums 169


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. The General Modular Relation 211


7.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . 211
7.2 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . 213
7.3 Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
7.4 The Main Formula (basic version) . . . . . . . . . . . . . 220

8. The Hecke Type Zeta-functions 225


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

xii Contributions to the Theory of Zeta-functions

8.8 The second J-Bessel expansion: G1,1 2,0


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. The Product of Zeta-functions 241


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. Miscellany 267


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.

Good master, what shall I do to inherit eternal life? — Luke 18:18.

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

2 Contributions to the Theory of Zeta-functions

of prime elements associated with the zeta-functions (or the zeta-functions


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

Things done without example in their issue are to be feared.

— Shakespeare, Henry VIII, 1.2.

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

Volume II, modular relation ultimatum

and

Volume III, modular relation niezwyciȩnzȯny,

which are hoped to be inherited by the coming generations.

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

4 Contributions to the Theory of Zeta-functions

Selberg class S of zeta-functions have, for which we refer to the work of


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).

As will be expounded at the end of §1.3, we mean by the Riemann-


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)

In the right half-plane, however, there is the principle of analytic continua-


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

It is arguable whether we should treat the Riemann zeta-function or


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

as one that satisfies the Hecke type functional equation


(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

Let χ be a Dirichlet character to the modulus q and let τ (χ) be the


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

6 Contributions to the Theory of Zeta-functions

The Dirichlet L-function L(s, χ) is defined by


∑∞
χ(n)
L(s, χ) = , σ > 1. (1.7)
n=1
ns

It satisfies the functional equation, which takes the following concise


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

which we assume is absolutely convergent for σ > σφ we call the increasing


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

which we assume is absolutely convergent for σ > σψ , we shall discuss


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)

which is absolutely convergent for σ := Re s > 1. Then (1.5) is equi-vocal


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

Then we think of (1.9) as


( ) ( )
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

8 Contributions to the Theory of Zeta-functions

the Dedekind zeta-function of an imaginary quadratic field we also refer to


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

where C encircles all the singularities of χ(s) in S, where χ is the key-


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).

(v) As the summation variable we almost always use k instead of more


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

With 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

associated is the processing gamma factor (Aj , Bj > 0)

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

and the H-function

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.


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

10 Contributions to the Theory of Zeta-functions

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.

This is essentially Lavrik’s Example 1 [Lav5, p.522] that indicates the


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

where the term corresponding to n = 0 on the right-hand side is to be


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

(1.30) is [MKn2, (2.10)] and is the main ingredient in Siegel’s proof of


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

12 Contributions to the Theory of Zeta-functions

1.2 Eternal return or every 50 years

1.2.1 Poincaré recurrence theorem


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].

Theorem 1.1 (Poincaré recurrence theorem) Let V be a finite and


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).

Proof. Consider the sequence of actions of f on U :

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

Since f is volume-preserving, all of the terms of this sequence have


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. 

1.2.2 Lerch-Chowla-Selberg formula


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

Although we referred to [Kur] in [Vi2], there are no comments on that


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

ζ ′ (0, x) = log Γ(x). (1.31)

Using (1.31), he obtained [Ler3] an important relation between the values of


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

14 Contributions to the Theory of Zeta-functions

rational and the imaginary quadratic fields.


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

ζΩ (s) = ζ(s)L(s, χΩ ). (1.33)

The evaluation of L′ (1, χΩ ) amounts to that of the Fourier series


∑∞
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

which is in a long run a consequence of the functional equation for the


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

fα (x + 1) − fα (x) = logα x x > 0. (1.36)

Corollary 1.2 For α = 1, f1 (x) = log Γ(x) and for α = 2, f2 (x) =


′′ 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

The higher-order Rα -function has been applied to the expressing the


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

1.2.3 Knopp-Hasse-Sondow formula


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=−∞

in case κ is even and



∑ κ
ϑκ (t) = e−n t
(1.39)
n=0

in case κ is odd, for Re t > 0. Note that


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

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

16 Contributions to the Theory of Zeta-functions

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.

Theorem 1.3 Suppose f (z) is an analytic in the circle |z −1| ≤ r (r > 1)


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

is analytic over the whole plane.

Proof. In the above-mentioned region of analyticity of f (z), it can be


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

term-wise integration are legitimate. By the Taylor expansion of f (z), we


obtain
( ∞
)
f (z) − z 1 ∑ f (n) (1)

= 1 + f (1)(z − 1) + (z − 1) − z .
n
z−1 z−1 n=2
n!

Substituting z = e−t and using the binomial expansion, we obtain


∑∞ ( )
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

Separating the term with k = 0 from the inner sum, we obtain


∑∞
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

whose first term is f (0), which is 0. Hence


∑∞ ( )
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

Substituting (1.46) in (1.45) and using (1.42), we have


∑∞ ( )
f (n+1) (1) ∑
n
n −s
F (s) = (−1)n−k k (1.47)
n=1
(n + 1)! k
k=1

or by changing the order of summation



∑ ∞
(−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

whence the Majorant series for F (s) is


∞ ∞
(∞ )
∑ 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

18 Contributions to the Theory of Zeta-functions

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!

Example 1.1 To elucidate Goss’ procedure, we choose


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

We may easily obtain the partial fraction expansion of g(z):


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,



m−1
1 − ζj
f (ℓ) (z) = δℓ,1 m + (−1) ℓ! ℓ+1
, (1.56)
j=1 (z − ζ j )

where δℓ,1 = 1 or 0 according as ℓ = 1 or ℓ ≥ 2, whence



m−1
1
f (ℓ) (1) = δℓ,1 m + (−1) ℓ! ℓ
. (1.57)
j=1 (1 − ζ j )

To secure the analyticity condition in Theorem 1.3, which is given by



1 − e 2πi
m > r(≥ 1),

it would be better to have smaller values of m, say, m = 2. In this case,


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

is a corollary to Theorem 1.3, where the difference operator is defined by


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

(−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

20 Contributions to the Theory of Zeta-functions

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

which amounts to the difference equation



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

by (1.60), it follows from (1.63) that


∑∞ ( )
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

whence (1.58) follows, completing the proof. 

1.3 The theta-transformation formula

We begin with Problem 17 in [WW] which gives a proof of the theta-


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

function (or series) which is the κ = 2 case of (1.38) by ϑ:




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

say. Expanding the denominator of the integrand of I1 into the geometric


series


1 ( )−1
= −e−2πiz 1 − e−2πiz = −e−2πiz e−2πikz ,
e2πiz − 1
k=0

which is absolutely convergent, we see that I1 in (1.67) is


∞ ∫
∑ ∞−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 −∞

Hence it follows that


∑∞ ∫ ∞ ∫ ∞
−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

22 Contributions to the Theory of Zeta-functions

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)
−∞ −∞

This is also proved by the Cauchy residue theorem.


Substituting (1.69) in (1.68), we obtain
( )∫ ∞
1 1
e−x dx.
2
ϑ(t) = √ ϑ (1.70)
πt t −∞

Putting t = 1, we may evaluate the probability integral


∫ ∞

e−x dx = π.
2
(1.71)
−∞

Substituting (1.71) in (1.70), we conclude the theta-transformation


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

We remark that in deriving (1.72), we applied the Cauchy residue the-


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

is equivalent to the functional equation for the associated zeta-function,


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)
−∞

for τ ∈ H, the upper half-plane (3.100), i.e. Im τ > 0.

Θ(iz) = ϑ (z) , (1.75)

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).

1.4 Summation formulas

There are some summation formulas known of number-theoretic nature.


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.

1.4.1 Poisson summation formula


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

24 Contributions to the Theory of Zeta-functions

Definition 1.2 For a function f (t) ∈ L1 (Rn ), its Fourier transform fˆ =


fˆ(t) is defined by
∫ ∞
ˆ
f (t) = f (t)e(−t · x) dx, (1.77)
−∞

where t · x means the inner product. It is a continuous function on Rn ,


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)
−∞

holds. If f (t) is periodic in each variable tj and integrable over the n-


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

The formal series



F(f )(t) = fˆ(n)e(t · x) (1.80)
n∈Zn

is the Fourier series for f .


∑ 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

are satisfied, then we have the Poisson summation formula


∑ ∑
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

1.4.2 Generalization of the Plana summation formula


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

f (ix) − f (−ix) = o(x), σ → 0+, (1.86)

(1.85) holds true for −1 < σ < 0.


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)

uniformly in σ for σ1 ≤ σ ≤ σ2 , where a > 0, b > 0 are subject to


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

26 Contributions to the Theory of Zeta-functions

β ≤ σ ≤ α (−1 < β < 1, 0 < α) except for a simple pole at s = 0. Then


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)

by (1.87), where α > 1 and we made the change of variable s by 1 − s in the


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

R1 = ρΩ F(1)z −1 , R0 = −ζΩ (0) Res F(s).


s=0

We apply the functional equation in the form


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

Now we use Lemma 1.2 in the form (−1 < σ < 0)


∫ ∞
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

and change the order of integration in (1.92) (which is legitimate by the


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

provided that the domain of f (x) is extended to

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.

Theorem 1.5 For Re a > 0, we have


∫ ∞
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

28 Contributions to the Theory of Zeta-functions

Corollary 1.5 (Hermite’s formula [SC, p. 91, (12)]). For Re a > 0 we


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

Fig. 1.1 Joseph Fourier


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

Chapter 2

Grocery of Special Functions

2.1 Formulas for the gamma function and their use

We refer to [Vista] for a concise theory of the Euler gamma function. It


is introduced most commonly as the principal solution to the difference
equation

Γ(s + 1) = sΓ(s), log Γ(s + 1) − log Γ(s) = log s (2.1)

with the normalization

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

More generally, we have the multiplication formula for the gamma


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

30 Contributions to the Theory of Zeta-functions

Exercise 2.1 Prove by (2.1) and (2.2) that


( )ρ
1 d 2ρ Γ(z − ν + ρ + 1) 2ν−2z−2ρ
− s2ν−2z = s . (2.6)
s ds Γ(z − ν)

Exercise 2.2 By (2.4) prove that


( )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 .

The workable procedure is always the same.

Make the denominators of coefficients coincide. Cancel com-


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

Grocery of Special Functions 31

Applying (2.4) three times, we transform the integrand into

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) ).

Hence by the change of variable formula (2.93),


( ( ) ( ))
√ √ 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) ,

which is (2.8) and (2.8) in turn leads to (2.9) in view of (2.119).

Exercise 2.3 Prove the identities


( )
(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) .

Also work out another case:


( )
(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

(2.10) is used in [KTT1] and (2.11) will be used in §8.9.


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

32 Contributions to the Theory of Zeta-functions

Solution By (2.4), the left-hand side of (2.10) reduces to


( 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 )

to which we apply (2.93) and then (2.8).

Definition 2.1 The incomplete gamma function which will play a


central role in Ewald expansion (Chapter 5) is defined by
∫ ∞
dt
Γ(s, z) = ts e−t , σ > 0. (2.12)
z t

For generalizations of the incomplete gamma function, we refer to [CZ].

Exercise 2.4 (i) Prove that


( )
1 √ √
Γ(1, z) = e−z , Γ ,z = π erfc( z), (2.13)
2

where erfc is the error function.


(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

for ν = 12 . Also deduce


( )
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

Grocery of Special Functions 33

(iii) Check the formula [ErdH, II, p.149]


∫ ∞
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

There are two excellent surveys on the theory of zeta-functions in general.


[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

34 Contributions to the Theory of Zeta-functions

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].

Table 2.1 Zeta-table


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].

There is a large class of perturbed zeta-functions including the gen-


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

Grocery of Special Functions 35

Table 2.2 Functional equation-table


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.

2.3 Bessel functions

In this section we assemble definitions and properties of well-known Bessel


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

36 Contributions to the Theory of Zeta-functions

relevant to us is the integral representation [ErdH, III,(34) p.21]



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

which in turn amounts to [ErdH, III,(36) p.83]


∫ ( 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

Grocery of Special Functions 37

Associated to the J-Bessel function is its complex version (ν ∈ Z)

Iν (z) = e− 2 iν Jν (ze− 2 iν )
π π
(2.28)

called the modified Bessel function of the first kind. It is a solution to


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

for 0 ≤ Re(s) < c < Re(s) + 34 < 32 [ErdH].


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)

A special case of (2.32) which is often used without notice is



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

K−ν (z) = Kν (z). (2.36)


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

38 Contributions to the Theory of Zeta-functions

For an interesting interpretation of (2.35), cf. [CZ, p.9].

2.4 Ψ-functions

Definition of the Ψ-function. For p, q = 0, 1, 2, . . . , we call


( )
(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

and its normalization


( )
*
(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.

Relation between H and Ψ-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

Grocery of Special Functions 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 )

In Chapter 5, we shall apply the following complementary formula (cf.


[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)

where 1 Ψ1 is one of the Ψ-functions defined by (2.37):


( ) ∞

(a, A) Γ(a + Ak) z k
1 Ψ1 ;z = .
(b, B) Γ(b + Bk) k !
k=0

2.5 Generalized hypergeometric functions

Generalized hypergeometric function. We recall the Pochhammer


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)

for k ∈ N and (a)0 = 1 by (2.1). For p, q = 0, 1, 2, . . .,


( )
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

40 Contributions to the Theory of Zeta-functions

In the case, where a1 , a2 , . . . , ap , b1 , b2 , . . . , bq ̸= 0, −1, −2, . . . , the ra-


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

where in (2.45), Iν is defined by (2.28).

Example 2.3 (Confluent hypergeometric functions) We note the follow-


ing notation ([ErdH, p.248])
( )
a
1 F1 ;z ≡ Φ(a, c; z), (2.47)
c

with Φ Humbert’s symbol, where 1 F1 is defined by (2.42). Its counterpart


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

Grocery of Special Functions 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:

U (a, c; z) = Ψ(a, c; z) (2.50)


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)

[ErdH, (7), p.257], which is sometimes used as a definition of the Tricomi


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)

Example 2.4 (Whittaker functions) The Whittaker function Wκ,µ is one


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

Integral representations for p+1 Fq+1 in terms of p Fq .

Proposition 2.1 We have a general integral expression ([PBM, 9, p.438])


( )
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

42 Contributions to the Theory of Zeta-functions

Proof. We perform the termwise integration in


∫ 1 ( ∞ )

t (1 − t)
b−1 c−b−1 k
ak (tz) dt
0 k=0

which is permissible in our immediate application to the hypergeometric


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

(ii) We now dwell on novel applications of hypergeometric functions in


[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

where B(α, β) is the beta-function. Formula (2.57) is valid only if

| arg α − arg β| < π (2.58)


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

Grocery of Special Functions 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

| arg α| < π, | arg β| < π.

Exercise 2.6 Under condition (2.58), deduce (2.57). Similarly, deduce


∫ ∞
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

(x + α)−s (x + β)−s (2.60)


∫ ( )
α−s β −s s, ν α
= αν B(ν, 2s − ν)2 F1 ;1 − x−ν dν
2πi (c) 2s β
and

(x + α)−s (x + β)−s (2.61)


∫ ( )
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

44 Contributions to the Theory of Zeta-functions

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

(x + α)−s (x + β)−s (2.64)


−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

(x + α)−s (x + β)−s (2.65)


∫ ( )
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

valid for 0 < c < 2σ, in place of them, respectively.

Proof. It suffices to note that the true form of analytic continuation of


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

Grocery of Special Functions 45

Theorem 2.1 We have for σ > 0



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

(i) Deduce from (2.56) that


( ) ∫ 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

(iii) Check the following ([PBM, 8, p.431]).


( ) ( )
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

46 Contributions to the Theory of Zeta-functions

(iv) Check the following ([PBM, p.455, 28, p.797]).


( )
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

indicates the incomplete beta-function.

The relation between the G-function and generalized hypergeo-


metric functions. We suppose that all the poles of


m ∏
n
Γ(bj + s) Γ(aj − s)
j=1 j=1

are of degree 1. Then we have


( )
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

Grocery of Special Functions 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

48 Contributions to the Theory of Zeta-functions

In particular, in the case p = q + 1, the right-hand side of the following


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

Grocery of Special Functions 49

Solution Indeed, the left-hand side becomes


( )
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

whence (2.81) follows by (2.72).

Exercise 2.11 Check that (2.74) with m = q = 2, n = 0 and p = 1:


( ) ( )
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

leads to (2.50) in view of

U (a, c; z) = z 1−c U (a + 1 − c, 2 − c; z). (2.83)

2.6 Fox H-functions

Definition of the (Fox) H-function


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

lie to the right of L and the poles of the gamma factor



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

50 Contributions to the Theory of Zeta-functions

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

we associate the processing gamma factor (Aj , Bj > 0)

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

and define the Fox H-function by




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

Grocery of Special Functions 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

2.7 Formulas for Fox and Meijer functions

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

Proposition 2.3 We have the following relations between H-functions:


( ) ( )
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

52 Contributions to the Theory of Zeta-functions

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

In particular, Formula (2.88) entails the H → G formula:


( )
(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 | Λ + c) = Γ(s + c | Λ) (c ∈ C),


Γ(s | CΛ) = Γ(Cs | Λ) (C ∈ R+ ),
Γ(s | Λ∗ ) = Γ(−s | Λ).

For another element


( ′ ′ )
′ {(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

Grocery of Special Functions 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

the proposition follows from the Mellin inversion theorem. 

Theorem 2.2 The reduction-augmentation formula reads


( ( ))
(c, C); −

H z Λ ⊕
− ;(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.

Theorem 2.3 (Duplication formula)


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

54 Contributions to the Theory of Zeta-functions

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

and for any Ξ ∈ Ω, we have



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

In the traditional notation, (2.97) reads


(
{( 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

Remark 2.3 In practice, we use a concise version of the bulky formula in


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

Grocery of Special Functions 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

Formulas for other groups are similar.


Theorem 2.4 (Reciprocity formula) The reciprocity formula for the
gamma function and Euler’s identity lead to
( ( ))
−;(c, C)

H z Λ ⊕
−;(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

which entails the following formula for G-functions:


( )
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 .

Corollary 2.1 We have


( ( )) ( ( ))
−;(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

56 Contributions to the Theory of Zeta-functions

which in the traditional notation, reads


( )
{(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)

which entails the G-function version.

2.8 Special cases of G-functions

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

The Riesz kernel which produces the Riesz sum is defined by


 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.

Cf. Remark 6.1.

The beta transform (or the beta integral) (4.2) is a special case of

Γ(a − s)Γ(b + s) : (2.108)


( )
a
G1,1 = Γ(1 − a + b) z b (z + 1)a−b−1 .
1,1 z
b

The inverse Heaviside integral formula ([Vista, p.108], [Vi2, p.98]),


a companion to (2.31), reads

Γ(a + s)Γ(b + s) : (2.109)


( )
− ( √ )
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

Grocery of Special Functions 57

Its special case is (cf. (2.33))


( )
− √ 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)

whose special cases read (cf. [PBM, p.635, 3])


( )
− 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

For a slight generalization, cf. (9.11).

[ErdH, (4), p.256] (or [PBM, 8.4.45.1, p.715]) reads


( ) ( )
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)

We note that neither (2.111) or its complement (2.118) appears in the


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

58 Contributions to the Theory of Zeta-functions

(2.112) reduces, in view of (2.122), to


( )
1 ( )
1,1
G1,2 z = Γ(a) − Γ a, z . (2.113)
a, 0

[ErdH, (6), p.216] reads


( )
a
2,0
G1,2 z = z 2 (b+c−1) e− 2 z Wκ,µ (z),
1 1
(2.114)
b, c

where Wκ,µ is the Whittaker function with indexes κ = 12 (b + c + 1) − a,


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

where the indexes of the Whittaker function are κ = a − 21 (b + c + 1), µ =


2 (b − c).
1

In view of [ErdH, (2), p.264]


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

where U (a, c; z) is the confluent hypergeometric function of the second kind


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

Γ(a, z) = e−z U (1 − a, 1 − a; z), (2.119)

which is [ErdH, (21), p.266], (2.117) with a = 1, c = 0 and a for b reduces


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

Grocery of Special Functions 59

while (2.118) with b = a reduces to


( )
a ( ) z b
G2,1
1,2 z a, b = Γ(1 − a + b) Γ a − b, z e z . (2.121)

As noted above, (2.118) is complementary to (2.111) and so are (2.119)


and (2.122) in their specifications:

γ(a, z) = Γ(a) − Γ(a, z) = a−1 z a Φ(a, a + 1; z), (2.122)


which is [ErdH, (22), p.266]. We also note

Γ(a, z) = e−z z a Ψ(1, a + 1; z), (2.123)


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

60 Contributions to the Theory of Zeta-functions

Fig. 2.1 Charles C. Fox


October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 61

Chapter 3

Unprocessed Modular Relations


by KAINAN UNIVERSITY on 02/26/15. For personal use only.

M,0 Ñ ,0
3.1 The H0,M ↔ H0, Ñ
formula

We suppose the existence of the meromorphic function χ satisfying the


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

and having a finite number of poles sk (1 ≤ k ≤ L), where

Dj , Ej > 0.

In the w-plane we take two deformed Bromwich paths

L1 (s) : γ1 − i∞ → γ1 + i∞, L2 (s) : γ2 − i∞ → γ2 + i∞ (γ2 < γ1 )

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)

where Γ(s | ∆) is the processing gamma factor (1.21). In this chapter we


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

62 Contributions to the Theory of Zeta-functions

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

is equivalent to the functional equation (3.1).

Both [BeI] and [BeII] treat the functional equation

Γm (s)φ(s) = Γm (r − s)ψ(r − s) (3.4)

for m a positive integer, extending Hecke’s functional equation. In [BeI] the


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

where Em (x) indicates Voronoĭ’s function


 
∫ −
1
m,0 
Em (x) = Γ (s)x ds = G0,m z 0, . . . , 0  (x > 0),
m −s
(3.6)
2πi (c) | {z }
m

and to the Riesz sum of order κ (cf. (6.1)):

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

Unprocessed Modular Relations 63

involving the generalized Bessel function Kν (x; µ; m) ([BeI, Definition 4])

∫ ( )−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

which is slightly different from other definitions of generalized Bessel func-


tions given below. Cf. (9.54) in Chapter 10.
by KAINAN UNIVERSITY on 02/26/15. For personal use only.

We remark that Bellman had already introduced the Voronoĭ function


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

This is nothing but the G-function formula


( )

V (x; a1 , . . . , an ) = Gn,0
0,n x (3.9)
a1 , . . . , an

and will be applied in §8.16 as well as in the following exercise.

Exercise 3.1 Prove the following as an example of (9.45)


 

∑ −
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

which is a generalization (3.5) above.

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

64 Contributions to the Theory of Zeta-functions

to the effect that


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

while Hafner [Haf] introduces the most general



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

Unprocessed Modular Relations 65

and L = Ca,b is a special contour described on [Haf, pp.151-152] with inden-


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.

with G(s) the gamma quotient of the Dedekind zeta-function (which is a


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 −

The remaining argument being the same as in Exercise 3.8, we obtain


(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

66 Contributions to the Theory of Zeta-functions

which corresponds to [AZW2, (10.39)] in slightly different form. We note


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:

Theorem 3.2 The functional equation

Γ(d1 + D1 s)φ(s) = Γ(e1 + E1 (r − s))ψ(r − s), (3.21)

is equivalent to the modular relation

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

Unprocessed Modular Relations 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

Proof follows from


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

Theorem 3.2 with no perturbation d1 = e1 = 0 is a typical example of


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.

(i) K-Bessel-allied functions in Bellman’s papers. Indeed, [Bel3, (3)] is


(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

68 Contributions to the Theory of Zeta-functions

The Riemann type functional equation


( ) ( )
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

which amounts to [Sieg1, (10)].


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

By a trivial change of variables and specifications, (3.30) amounts to


[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

Then multiplying (3.31) by e−πt and integrating in x over (0, ∞),


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

Unprocessed Modular Relations 69

3.1.2 The G2,0 2,0


0,2 ↔ G0,2 formula

We keep Convention in §9.1, i.e. φ(s) = ψ(s) = π −s ζ(2s + ν)ζ(2s − ν).


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

By (2.109), (3.33) reduces to the K-Bessel formula:



∑ ( √ )
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

As a parenthetical remark on [BeIII, p.324], Berndt notes that (3.34)


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

Remark 3.2 The above reasoning can be perceived already by (3.89).


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

70 Contributions to the Theory of Zeta-functions

3.2 Dedekind zeta-function I

We consider the Dedekind zeta-function of a number field Ω with degree [Ω :


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

for σ = Re s > 1, where a runs through all non-zero ideals of Ω and


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

and is continued to a meromorphic function with a pole at s = 1 with


residue (cf. (3.152) for κ ≤ 2)

2r+1 π r2 Rh
ρ = λh = √ . (3.38)
w |∆|

Here w, h and R are the number of roots of unity in Ω, the class


number and the regulator, respectively.
October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 71

Unprocessed Modular Relations 71

Table 3.1 Invariants of a number field


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 |∆|

We define the meromorphic function in (3.1) as


(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

The functional equation (3.1) in this case reads

(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)

and χ(s) has poles at s = 1, 0 (and possibly at negative integer points,


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

Hence Theorem 3.1 reads


October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 72

72 Contributions to the Theory of Zeta-functions

Theorem 3.3

∑ ( )
αk −
H M,0

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.

Following Koshlyakov, we define the X-function, X(x) = Xr1 ,r2 (x), to


( )

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.

Corollary 3.1 Theorem 3.3 asserts that



∑ ∞

FΩ (k)Xr1 ,r2 (Akρ) = P(x) + FΩ (k)Xr1 ,r2 (Akρ−1 ), (3.44)
k=1 k=1

where P(x) = R0 + R1 is the residual function and


( )
Rj = Res χ(s)z −s , s = j , j = 0, 1. (3.45)

We note that R1 = ρz −1 , where ρ is defined by (3.38).

Remark 3.3 In Koshlyakov’s case which we will study in §3.4, κ ≤ 2,


whence Ω is either the rational or a quadratic field. Corollary 3.1 will be
made more explicit as Theorem 3.12 below.

3.3 Transformation formulas for Lambert series

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

Unprocessed Modular Relations 73

Theorem 3.4 For Dirichlet series φ, ψ that satisfy the functional


equation

Γ(c + Cs)φ(s) = Γ(c + C(r − s))ψ(r − s), (3.46)


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.

3.3.1 Lambert series


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

is absolutely convergent in |z| < 1 and represents an analytic function


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

Along with (3.49), we may consider



∑ zn
f+ (z) := an . (3.52)
n=1
1 + zn

Riemann’s posthumous Fragment [Rie1], based on Jacobi’s Fundamenta


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

74 Contributions to the Theory of Zeta-functions

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 Wintner [Win2] published in 1941, which already gave a close analysis


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.

For the moment we shall dwell on Wintner’s paper [Win2], elucidating


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

Unprocessed Modular Relations 75

with
1∑
cn = an . (3.54)
n
d|n

In the case of (3.49), we should have


1∑ n
cn = (−1) an . (3.55)
n
d|n

Then Wintner asks if F (z) tends to a measurable boundary function F (eiθ )


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.

actually is the Fourier series of the boundary function.


Use is made not only of the celebrated Riesz-Fischer condition [Vi2,
Theorem 7.1, p. 139]

∑ 2
|cn | < ∞ (3.56)
n=−∞

but also a weaker condition



∑ ε 2
|n| |cn | < ∞, (3.57)
n=−∞

which lies between (3.56) and the Lp -condition



∑ p/(p−1)
|cn | <∞ (p ≥ 2) (3.58)
n=−∞

for an f to be of class Lp for some p > 2. Condition (3.57) implies that


∑∞
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

F (reiθ ) → F (eiθ ) a.e., as r → 1 (3.60)


October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 76

76 Contributions to the Theory of Zeta-functions

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

for every ε > 0 (e.g. [HaWr]). Hence it follows that

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=−∞

This together with the condition (3.57) implies that





F (e ) = cn einθ a.e. (3.64)
n=−∞

Hence (3.60) follows by Abel’s continuity theorem. 


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

which in the notation of (3.50) reads




2K ℓ 2
f (z) = =4 d4,1 (n)z 2 (4m−1) n + 1, (3.66)
π
ℓ,m,n=0

with d4,1 (n) denoting the number of divisors of n of the form 4k + 1.


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

Unprocessed Modular Relations 77

Hence by Theorem 3.5, the boundary function F (eiθ ) exists and


∑∞
d4,1 (n) inθ
F (eiθ ) = e a.e. (3.68)
n=1
n

(ii) In the case an = 1, we obtain the Lambert series considered by Lambert


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

Hence by Theorem 3.5, the boundary function F (eiθ ) exists and


∑∞
d(n) inθ
F (eiθ ) = e a.e. (3.71)
n=0
n

In connection with (3.71), we introduce the notation and terminology.


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

while for x ∈ Z, the right-hand side series of (3.73) converges to 0. Following


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

78 Contributions to the Theory of Zeta-functions

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).

3.3.2 Lambert series and short character sums


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

the series being absolutely convergent in the right half-plane.


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

with φ being the Euler φ-function.


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

Unprocessed Modular Relations 79

to express the congruence n ≡ −N (mod r) to deduce (3.76).


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

which is then effectively used to obtain non-trivial congruences, cf. [SUZ],


by KAINAN UNIVERSITY on 02/26/15. For personal use only.

[KUW], [Ur1] etc.


(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.

Theorem 3.6 Let χ be a Dirichlet character mod M , not necessarily


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

80 Contributions to the Theory of Zeta-functions

is the Hurwitz-Lerch L-function (cf. §5.2.1 and [Mor]), which is analyt-


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

is valid for all s.


Or more simply, using the expression

Φ(χψ, N, s) = L(s, χψ) − Ms (χψ)


by KAINAN UNIVERSITY on 02/26/15. For personal use only.

and appealing to the analyticity of the Dirichlet L-function furnished by


the functional equation (1.8), we deduce (3.79) for all s ∈ C. 

In view of [Ber75-1, Example 2] with slight modifications (or [SUZ, p.


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

In view of (3.81), Theorem 3.6 entails the values at non-positive integers.

3.3.3 Ramanujan’s formula leading to the eta-transformation


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.

Let α > 0, β > 0 satisfy the relation (cf. (1.2))

αβ = π 2 (3.82)
October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 81

Unprocessed Modular Relations 81

and let n be any positive integer. Then


{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)!

where Bm denotes the m-th Bernoulli number defined by


by KAINAN UNIVERSITY on 02/26/15. For personal use only.



x Bm m
= x , |x| < 2π. (3.84)
ex − 1 m=0 m!

Formula (3.83) is true for all integers n ̸= 0 by interpreting a vacuous


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

indicates the sum-of-divisors function, the sum of a-th powers of all


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

82 Contributions to the Theory of Zeta-functions

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

where R = {−2n − 1, −2n, −2n + 1, −2n + 3, . . . , −3, −1, 0, 1}, and s = 0 is


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)

indicates the local uniformization parameter and τ is in the upper-half plane


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

Unprocessed Modular Relations 83

(3.103) below, whence it is an automorphic form of weight “1/2”. The most


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.

mula see (3.137) below.


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

Considering F (τ ) − F (− τ1 ), we are led to


( ) ∞ ( )
1 i∑1 πl
F (τ ) − F − = cot πlτ + cot . (3.98)
τ 2 l τ
l=1

Theorem 3.7 should be compared with Theorem 3.10 below. Formula


(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

84 Contributions to the Theory of Zeta-functions

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.

functions. cf. [KMaT].


Chan [HH] considers the modular relation for the inverse of the eta-
function, which is the generating function of the partition function.

3.3.4 A brief account of modular forms


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)

by the linear fractional transformations


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

Unprocessed Modular Relations 85

whence it usually suffices to know the transformation formula under the


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.

A modular form f (z) is called a cusp form if it vanishes at infinity.


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

which is 24-th power of the Dedekind eta-function (3.92) defined above.


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

86 Contributions to the Theory of Zeta-functions

∑∞
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

is continued to a meromorphic function and satisfies the functional equation


−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.

3.3.5 The Ramanujan-Guinand formula


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

φ(s) = ζ(s)ζ(s + 2n + 1), (3.110)

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

For simplicity we put



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

Unprocessed Modular Relations 87

which is (−1)a Γ(s+a)


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.

(2π)−s Γ(s)φ(s) = (−1)n (2π)2n+s Γ(−2n − s)φ(−2n − s). (3.114)

Moving the line of integration to σ = −κ1 (κ1 > 2n + 1 − a), whereby


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

Γ(a − 2n − s) Γ(1 + s + 2n) sin πs


= (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

88 Contributions to the Theory of Zeta-functions

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

we conclude from (3.115) and (3.118) that


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),

where Pa (x) is the residual function.


Particular cases are

P0 (x) = P(x),
October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 89

Unprocessed Modular Relations 89

given by (3.88), and


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

We now state a corollary to Theorem 3.9, which makes it possible to


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

90 Contributions to the Theory of Zeta-functions

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

Example 3.2 We have


{∞
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 . . .

correct up to 84 decimal places (by computing first 30 terms).

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 . . . ,

(cf. the Research Notices at the end of [KTY7]).


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

Unprocessed Modular Relations 91

version of the “pseudo modular relation principle” whose multiplicative ana-


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.

phic property of the Dedekind eta-function. This can be regarded as once


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

we see that it is nothing but


( )
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

92 Contributions to the Theory of Zeta-functions

i.e. the automorphy of G2n+2 (z).


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

whence follows (3.94) up to the root of unity.


by KAINAN UNIVERSITY on 02/26/15. For personal use only.

Proof. Proof is due to ([Cho]). According to Convention in §1.1, (3.114)


should read

Γ(s)φ(s) = Γ(2n + 1 − s)ψ(2n + 1 − s), (3.129)

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

or in the form of the Lambert series



∑ ∞
∑ 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

Unprocessed Modular Relations 93

whence integrating we deduce that


( )
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

Putting τ = i gives k = 1, so that (3.130) amounts to (3.128), completing


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.

form of weight 0, i.e.


( )
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

By (3.123) with a = 2, we have


( )
1
1 + 240I3 = τ 4 (1 + 240I3 (τ )). (3.135)
τ
Substituting (3.128) and (3.134) in (3.135) leads to (3.133).

3.3.6 The reciprocity law for Dedekind sums


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

94 Contributions to the Theory of Zeta-functions

which is immediate from (3.92).


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

where (h, k) = 1, k ∈ N and (((x)) is the special notation used in Dedekind


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

The Dedekind sums obey the famous reciprocity law: For h, k ∈


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).

We divide the proof of Theorem in two parts as stated in the following

Lemma 3.1 (i) The functional equation (1.5) for the Riemann zeta-
function ζ(s) implies

Φ(s) = (2π)−s Γ(s)ζ(s)ζ(s + 1) = Φ(−s). (3.141)

(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

Unprocessed Modular Relations 95

Proof of Lemma 3.1. Proof of (i) is elementary using the reciprocal


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

By comparing the two expressions for η(τ ′′ ), where τ ′′ = aSτ +b


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.

3.4 Koshlyakov’s method [KoshI]

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

96 Contributions to the Theory of Zeta-functions

of (positive) definite quadratic forms are the Epstein zeta-functions whose


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.

3.4.1 Dedekind zeta-function II


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.

∆ stands for the discriminant of Ω, then we may write Ω = Ω( ∆) under


the convention that ∆ = 1 in the case of the prime field Ω = Q.

{
∆=1 n=1
(3.144)
∆ = a square-free integer n = 2.

The residue ρ at s = 1 of the Dedekind zeta-function (cf. (3.38)) in this


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

Unprocessed Modular Relations 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.

|∆|

3.5 Koshlyakov’s functions

Koshlyakov introduced several functions starting from the K-function (ker-


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.

Table 3.2 Koshlyakov’s functions


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

3.5.1 Koshlyakov’s X-functions


Exercise 3.5 Prove the formulas in Table 3.2 for the X-function defined
by (3.43).

X1,0 (x) = 2e−x


2
(3.148)

X0,1 (x) = e−x (3.149)


October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 98

98 Contributions to the Theory of Zeta-functions

and

X2,0 (x) = 4K0 (2x). (3.150)

Solution These follow from (2.93), (2.106), (2.109).


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.

[KoshI, p.119, l.11 from below].

Exercise 3.7 In the case r2 = 0 (r1 ≥ 2), show that


( ( 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

Unprocessed Modular Relations 99

Remark 3.9 In [KoshE], Koshlyakov mentions:


“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

(3.154) leads, on incorporating (3.153), to Koshlyakov’s formula [KoshI,


(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

100 Contributions to the Theory of Zeta-functions

or writing ρ2 for z,

∑ ∞ (
∑ )
−πk2 z 1 − 12 −πk2 z −1 1
e + =z e + ,
2 2
k=1 k=1

the theta-transformation formula (1.72) (cf. §1.72).


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

by (3.153). Formula (3.156) is the corrected form of Koshlyakov’s [KoshI,


(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

Formula (3.157) is a special case of the G2,0 2,0


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

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

Unprocessed Modular Relations 101

3.5.2 Koshlyakov’s Y -function


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

The Y -function is defined by ([KoshI, (5.4)])

Y (x) = Yr1 ,r2 (x)



1 π
= ( ) ( ) x−s ds
by KAINAN UNIVERSITY on 02/26/15. For personal use only.

2πi (c) 2 sin π2 s Γr1 2−s Γ r2 (2 − s)


(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)

or after some transformation,

Y (x) = Yr1 ,r2 (x) (3.160)


( )
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)

Though contradicting to our claim that the processing gamma factor


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

In the case r1 = 0, r2 = 1, by (3.160),


( ( ) ( ))
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

102 Contributions to the Theory of Zeta-functions

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

By (2.122), this reduces further to


( )
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).

In the special case s = 0, r = 1, (3.163) reduces to


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)

by (2.110), Js (x) being the Bessel function defined by (2.24).


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

Unprocessed Modular Relations 103

In what follows we shall deduce Theorem 3.14 as a special case of the


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

with its associated Gamma factor



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)

is equivalent to the functional equation

Γ(s)φ(s) = Γ(r − s)φ(r − s), (3.169)


∑L ( )
k=1 Res Γ(w − s | ∆) χ(w) z
s−w
where P(z) = , w = sk indicates the
residual function.

We assume that φ(s) has a simple pole at s = 1 with residue ρφ , so that


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

However, we appeal to (3.162) by putting Γ(w − s | ∆) =


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

104 Contributions to the Theory of Zeta-functions

exponential functions. Then this amounts to the residual function is com-


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

By (2.123), this reduces further to


( )
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

= zΓ(r − s, z). (3.172)

In the special case s = 0, r = 1 of (3.171), since (3.172) reduces to


( )
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.

(3.170) reduces for s = 0 to φ(0) + π2 z −1 ρφ . Thus we have proved


October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 105

Unprocessed Modular Relations 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.

We now consider the L-function which are relatives of K-functions (in


§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

by (2.110), which further reduces e.g. by (2.27) to the required formula


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.

In [KoshE], Koshlyakov mentions:


October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 106

106 Contributions to the Theory of Zeta-functions

“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.

The importance of the Hecke gamma transform in number theory cannot be


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

We begin by materializing the Murty-Sinha theorem, Theorem 4.1, when


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

108 Contributions to the Theory of Zeta-functions

the modified K-Bessel function. In particular, we shall elucidate [KoshII,


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.

4.2 Stark’s method

4.2.1 Murty-Sinha theorem


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.

Theorem 4.1 ([MS, Theorem 3.1]) If the Dirichlet series

∑∞
αn
φ(s) =
λs
n=1 n

with abscissa σφ of absolute convergence admits a meromorphic continua-


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

The proof of the Murty-Sinha theorem depends on the “binomial prin-


ciple of analytic continuation” stated as

∑∞ ( )
−s
φ(s, a) = φ(s + r)ar (4.1)
r=0
r

for 0 < |a| < 1.


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)).

Let ξ = Re z > 1, then the Lipschitz-Lerch transcendent is defined by

∑∞
e2πinλ
ϕ(λ, a, z) = (4.3)
n=0
(n + a)z

where λ ∈ R and a ∈ C − (Z ∪ {0}), but we usually restrict to Re a > 0.


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

Example 4.1 We apply (4.2) to the Lipschitz-Lerch transcendent and


obtain the integral representation
October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 110

110 Contributions to the Theory of Zeta-functions


1 Γ(−w)Γ(s + w)
ϕ(λ, a + x, s) = ϕ(λ, a, s + w)xw dw. (4.4)
2πi (c) Γ(s)

This is valid for 1 − σ < c < 0.


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

This is the K-th partial sum of the Taylor series around a.

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!

in |ξ| < |a|, where (s)ν = s(s + 1) · · · (s + ν − 1) denotes the Pochhammber


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

This reduces to the Jensen-Ramaswami formula [Tit, (2.14.2)]

∑∞
(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

in view of the well-known relations

(s)ν ν (−s)(−s − 1) · · · (−s − ν + 1)


= (−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.

Theorem 4.2 If the original Dirichlet series satisfy the functional


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.

4.2.2 Stark’s method


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

112 Contributions to the Theory of Zeta-functions

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

where χ is a Dirichlet character mod q. Thus it suffices to give closed


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

with the understanding that the term with k = K + 1 is


( )
−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 .

We have for any 0 ≤ K ∈ Z,


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

In particular, if we use ζ(0) = − 12 , then

1
ζ(0, x) = − x = −B1 (x)
2
where B1 (x) is the first Bernoulli polynomial (cf. (3.72)).

Proof. Putting s = −K in (4.10) and noting that 1


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)

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

114 Contributions to the Theory of Zeta-functions

We use the integral representation (4.10). Putting K = 1 it becomes



1 Γ(−w)Γ(s + w)
ζ(s, x) = x−s + ζ(s + w)xw dw
2πi (aK ) Γ(s)

+ ζ(s) − sζ(s + 1)x. (4.16)

Now differentiating and putting s = 0 in (4.16), we find as in Stark that


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

Thus (4.18) leads to


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

4.3 The main formula for modular relations

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

{ϕh (s)} (1 ≤ h ≤ H) and {ψi (s)} (1 ≤ i ≤ I)

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.

complex sequences. Consider Dirichlet series



∑ (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

116 Contributions to the Theory of Zeta-functions

We further assume that only finitely many of the poles sk (1 ≤ k ≤ L)


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

We suppose that for any real numbers u1 , u2 (u1 < u2 )

lim Γ(u + iv − s | ∆) χ(u + iv) = 0 (4.20)


|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

lie on the right of L1 (s), and those of


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

lie on the left of L2 (s), and those of


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

118 Contributions to the Theory of Zeta-functions

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

is equivalent to the functional equation (4.19).

We now apply the modular relation (4.21) in the following special case.

Table 4.1 Choice of parameters:


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

Theorem 4.4 The functional equation

Γ(d1 + D1 s)φ(s) = Γ(e1 + E1 (r − s))ψ(r − s), (4.22)


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

with Γ(s|∆) = Γ(a1 − A1 s), is equivalent to

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

4.3.1 Specification of Theorem 4.4


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

We write z for z −1 and for simplicity we choose a1 = d1 = 0 to reduce


(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

120 Contributions to the Theory of Zeta-functions

Similarly the series on right-hand side of (4.24) becomes for z → z −1 and


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.

Further specifying a1 = d1 = e1 = 0, we deduce the following corollary.

Corollary 4.1 The functional equation

Γ(A1 s)φ(s) = Γ(A1 (r − s))ψ(r − s) (4.27)

is equivalent to (with Γ(w|∆)) = Γ(−A1 w))

∑ ∞
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.

In the case of the Riemann zeta-function, αk = βk = 1, λk = πk, µk =



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

i.e. Watson’s formula again. Cf. (1.28) and (1.25) above.

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.,

Γ (s) φ(s) = Γ (r − s) ψ(r − s) (4.33)

with a simple pole at s = r with residue ρ. In this case (4.28) gives



∑ ∑ 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

122 Contributions to the Theory of Zeta-functions

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

4.4 Dedekind zeta-function III


by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

In this section we specify the above results to the Dedekind zeta-function


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

with the coefficient A defined by (3.35).


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.

the following result.

Proposition 4.3 The functional equation (4.33) with r = 1 is equivalent


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

124 Contributions to the Theory of Zeta-functions

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.

we introduce the perturbed Dedekind zeta-function



2ζΩ (0) ∑ α(k)
ζΩ (s, ω) = − + s. (4.41)
w (k + ω)
k=1

Then (4.39) and (4.40) give the Lipschitz summation formula



∑ ′ 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

As before we choose the values of the parameters as in the following:


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

Table 4.2 Choice of parameters


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

In this set-up the modular relation reads

Theorem 4.5 The functional equation


by UNIVERSITY OF QUEENSLAND on 08/15/15. For personal use only.

Γ(d1 + D1 s) Γ(e1 + E1 (r − s))


φ(s) = ψ(r − s), (4.45)
Γ(d2 − D2 s) Γ(e2 − E2 (r − s))

with Γ(s|∆) = Γ(a1 − A1 s), is equivalent to

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 .

The right-hand side of the above contains the summand


(

2,0
H0,3 µk z
(a1 , A1 ), (e1 + (1/2)A1 (r − s), (1/2)A1 ),
)

(1 − e2 + (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

126 Contributions to the Theory of Zeta-functions

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

by (2.96), (2.99), (2.101) and (2.109). Equation (4.48) with s ↔ 1−s


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

+ (2e 2i z) K1−s−a1 2 e 4 2z . (4.49)


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

Using a1 = 0, the equation (4.49) becomes

( )
−(
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.

This leads to an analogue of Koshlyakov’s result, however with the basic


sequences get squared (since we must choose E1 = 2 in (4.45), we are forced
to choose A1 = 21 ).

4.5 Elucidation of Koshlyakov’s result in the real quadratic


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

which is in a form similar to that of (4.43). (He ) applies the duplica-


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

which is a reminiscent of the Wigert-Bellman divisor problem [Bel2]. Thus


it suggests the following

Theorem 4.6 The generating Dirichlet series for the Wigert-Bellman


October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 128

128 Contributions to the Theory of Zeta-functions

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

amounts to the Lipschitz summation formula



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)

The integral expression for (4.55) contains the integrand


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).

4.6 Koshlyakov’s K-series

Let G(s) = GΩ (s) denote the quotient of gamma factors in (3.41):


( )
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

for c > 0 and x > 0 (Re(x) > 0 being allowed).


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

Then Koshlyakov considers the basic series



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.

, α > 0, 0 < c < 2. (4.59)


2πi (c) 2 sin 2 w α 1 + αs

He proves the following theorem.

Theorem 4.7 (Koshlyakov) One has



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 proof of (4.60) follows by taking the limit as s → 0 in formulas (4.42).


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

130 Contributions to the Theory of Zeta-functions

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)]).

In the imaginary quadratic case, r1 = 0, r2 = 1, we need to locate


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)

Applying (2.96), we see that

( )
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.

amounts to (4.66) by (4.72) above which is a special case of (2.109).


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

by (2.3). Now by (2.4) and Euler’s identity, the integrand is

− 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)

which amounts to (4.66) by


( ) ∫

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

132 Contributions to the Theory of Zeta-functions

which is a special case of (2.109).


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.

[PBM, 13, p.675].


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)

Thus, in our case


( √ )
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).

Finally, we deal with the real quadratic case, where


∫ ( ) ( )
1 1 Γ2 w2 Γ 1+w
K2,0 (x) = ( )2
x−w dw. (4.77)
2πi (c) 2 Γ 1−w
2

Formula (4.77) is already


( )
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

whence it follows that


( )

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 ,

i.e. [KoshI, (2.6)].

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.

We transform the integrand into


( ( ) ( ) − 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

in view of (4.72), i.e. [KoshI, (2.6)] follows.

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


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

134 Contributions to the Theory of Zeta-functions

The special case of (3.6) described in the title reads

∑∞ ( ) ∑ ( )
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

Exercise 4.3 Check by (2.108) and (2.109) that (4.81) reduces to



∑ 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

(4.81) in the special case a = 0 gives the Fourier-Bessel expansion for


the perturbed Dirichlet series (or K-Bessel expansion):

Theorem 4.8 (Fourier-Bessel expansion) The Fourier-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

is equivalent to the functional equation.

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

for which, we refer to [Vi2, p.34]. There we referred to it as the Mellin-


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)

In addition to the papers [Be1], [BeIII], [BeIV], [BeVI], [CZS1], [CZS2],


[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

The overall procedures consist of expressing the sum in the


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.

Exercise 4.4 Deduce the following from (4.82).


∞ ∞
√ ∑ 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

136 Contributions to the Theory of Zeta-functions

∞ ∞
√ −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.

To state the generalized Chowla-Selberg type formula (Theorem 4.9),


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

Theorem 4.9 (generalized Chowla-Selberg type formula, cf. [Vista]) Un-


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 ]

The case g = h = z, n = m = 1 is due to Chowla and Selberg [CS],


[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

Corollary 4.2 Let


∑′ 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

Corollary 4.2 is often stated in slightly different form. Let


√ √
|∆| 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

Corollary 4.3 (4.90) may be written as



4π s 2s− 2 ∑ s− 1
1

Q(s) = √ √ n 2 σ1−2s (n) cos 2πnxKs− 21 (2πyn) . (4.92)


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

138 Contributions to the Theory of Zeta-functions

Definition 4.1 In conjunction with the Epstein zeta-function (4.88) the


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.

Theorem 4.10 (Kuzumaki) We have the Fourier-Bessel expansion


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

where α = x + iy (y > 0) and

( )
1 2,0
2 −
Im,n (z, α) = √ G0,2 (πmny) . (4.96)
π z, 12

We have several well-known expressions for G2,0


0,2 and so

Im,n (z, α) = e−2πmny (2πmny)2z U (z, 2z; 4πmny). (4.97)

Under (4.100), (4.95) reduces to the Chowla-Selberg integral formula in


Corollary 4.2 above

z+ 12
Im,n (z, α) = 2π z (mny) Kz− 12 (2πmny), (4.98)

while (4.102) leads to the Whittaker function series expression

Im,n (z, α) = (πmny)z W0,z− 12 (4πmny). (4.99)


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

The expressions depend on the G-function hierarchy

( )

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)!

where B(x, y) is the beta function which satisfies

∫ 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

which we refer to as the Katsurada transform ([Kat5]).


October 28, 2014 9:23 BC: 8711 – Contributions to the Theory of Zeta-Functions modularrelation page 140

140 Contributions to the Theory of Zeta-functions

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

−ξ < c′ < min{0, 1 − 2ξ}.

Under this condition, we have from (2.111)



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

U (z, 2z; 4πmny) (4.107)



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

Using (4.105) for Γ(s + z) in (4.107), we have

U (z, 2z; 4πmny)



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.

× (4πmny)s t−K−s−z (1 − t)K−1 dt ds


(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

U (z, 2z; 4πmny) (4.108)



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

142 Contributions to the Theory of Zeta-functions

which leads to

Im,n (z, α) (4.109)



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

which leads, in view of (4.97), to (4.106), completing the proof.

4.8 Bochner-Chandrasekharan and Narasimhan formula

[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].

Example 4.4 (Bochenr-Chandrasekharan and Narasimhan


[ChN2]) In this example we use the residual function for the sum of
residues.

Table 4.3 Relation among variables


old λk µk s z
√ s2
new 2πλk 2πµk 2 2πz 8π

With slight changes (table of variables) in (4.82), we have


∑ ( )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.

Exercise 4.6 Deduce the following from Exercise 4.3