Lecture Notes in Mathematics 1877

Editors:
J.-M. Morel, Cachan
F. Takens, Groningen
B. Teissier, Paris
Jrn Steuding

Value-Distribution
of L-Functions

ABC
Author
Jrn Steuding
Department of Mathematics
Chair of Complex Analysis
University of Wrzburg
Am Hubland
97074 Wrzburg
Germany
e-mail: [email protected]

Library of Congress Control Number: 2007921875

Mathematics Subject Classification (2000): 11M06, 11M26, 11M41, 30D35, 30E10,

60B05, 60F05

ISSN print edition: 0075-8434

ISSN electronic edition: 1617-9692
ISBN-10 3-540-26526-0 Springer Berlin Heidelberg New York
ISBN-13 978-3-540-26526-9 Springer Berlin Heidelberg New York

DOI 10.1007/978-3-540-44822-8

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication
or parts thereof is permitted only under the provisions of the German Copyright Law of September 9,
1965, in its current version, and permission for use must always be obtained from Springer. Violations are
liable for prosecution under the German Copyright Law.
Springer is a part of Springer Science+Business Media
springer.com
c Springer-Verlag Berlin Heidelberg 2007

The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,
even in the absence of a specific statement, that such names are exempt from the relevant protective laws
and regulations and therefore free for general use.
Typesetting by the author and SPi using a Springer LATEX macro package
Cover design: WMXDesign GmbH, Heidelberg
Printed on acid-free paper SPIN: 11504160 VA41/3100/SPi 543210
Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 The Riemann Zeta-Function and the Distribution
of Prime Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Bohrs Probabilistic Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Voronins Universality Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4 Dirichlet L-Functions and Joint Universality . . . . . . . . . . . . . . . . 19
1.5 L-Functions Associated with Newforms . . . . . . . . . . . . . . . . . . . . . 24
1.6 The LinnikIbragimov Conjecture . . . . . . . . . . . . . . . . . . . . . . . . . 28

2 Dirichlet Series and Polynomial Euler Products . . . . . . . . . . . . 35

2.1 General Theory of Dirichlet Series . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.2 A Class of Dirichlet Series: The Main Actors . . . . . . . . . . . . . . . . 37
2.3 Estimates for the Dirichlet Series Coecients . . . . . . . . . . . . . . . 40
2.4 The Mean-Square on Vertical Lines . . . . . . . . . . . . . . . . . . . . . . . . 43

3 Interlude: Results from Probability Theory . . . . . . . . . . . . . . . . 49

3.1 Weak Convergence of Probability Measures . . . . . . . . . . . . . . . . . 49
3.2 Random Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3 Denjoys Probabilistic Argument for Riemanns Hypothesis . . . 54
3.4 Characteristic Functions and Fourier Transforms . . . . . . . . . . . . 56
3.5 Haar Measure and Characters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.6 Random Processes and Ergodic Theory . . . . . . . . . . . . . . . . . . . . 58
3.7 The Space of Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 59

4 Limit Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1 Associated Random Elements and the Main Limit Theorem . . . 63
4.2 Limit Theorems for Dirichlet Polynomials . . . . . . . . . . . . . . . . . . 67
4.3 An Ergodic Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.4 Approximation in the Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
VI Contents

4.5 A Limit Theorem for Absolutely Convergent Series . . . . . . . . . . 76

4.6 Proof of the Main Limit Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.7 Generalizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.8 A Discrete Limit Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

5 Universality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.1 Dense Sets in Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.2 Application to the Space of Analytic Functions . . . . . . . . . . . . . . 94
5.3 Entire Functions of Exponential Type . . . . . . . . . . . . . . . . . . . . . . 96
5.4 The Positive-Density Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.5 The Support of the Limit Measure . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.6 The Universality Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.7 Discrete Universality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

6 The Selberg Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6.1 Denition and First Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.2 Primitive Functions and the Selberg Conjectures . . . . . . . . . . . . 116
6.3 Non-Vanishing and Prime Number Theorems . . . . . . . . . . . . . . . 119
6.4 Pair Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.5 The PhragmnLindelf Principle . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.6 Universality in the Selberg Class . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.7 Lindelfs Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.8 Symmetric Power L-Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

7 Value-Distribution in the Complex Plane . . . . . . . . . . . . . . . . . . 137

7.1 Sums Over c-Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.2 Riemannvon Mangoldt-Type Formulae . . . . . . . . . . . . . . . . . . . . 142
7.3 Nevanlinna Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.4 Uniqueness Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

8 The Riemann Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

8.1 Uniform Approximation and Zeros . . . . . . . . . . . . . . . . . . . . . . . . . 155
8.2 Bagchis Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
8.3 A Generalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
8.4 An Approach Towards Riemanns Hypothesis? . . . . . . . . . . . . . . 162
8.5 Further Equivalents of the Riemann Hypothesis . . . . . . . . . . . . . 163

9 Eective Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

9.1 The Problem of Eectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
9.2 Upper Bounds for the Density of Universality . . . . . . . . . . . . . . . 169
9.3 Value-Distribution on Arithmetic Progressions . . . . . . . . . . . . . . 174
9.4 Making Universality Visible . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
9.5 Almost Periodicity in the Half-Plane
of Absolute Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
9.6 Eective Inhomogeneous Diophantine Approximation . . . . . . . . 182
9.7 c-Values Revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
 Contents VII

10 Consequences of Universality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

10.1 Dense Sets in the Complex Plane . . . . . . . . . . . . . . . . . . . . . . . . . . 193
10.2 Functional Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
10.3 Joint Functional Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
10.4 Anderssons Disproof of a Mean-Square Conjecture . . . . . . . . . . 200
10.5 Voronins Theorems and Physics . . . . . . . . . . . . . . . . . . . . . . . . . . 201
10.6 Shifts of Universal Dirichlet Series . . . . . . . . . . . . . . . . . . . . . . . . . 202

11 Dirichlet Series with Periodic Coecients . . . . . . . . . . . . . . . . . 209

11.1 Zero-Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
11.2 A Link to the Selberg Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
11.3 Strong Universality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
11.4 Hurwitz Zeta-Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

12 Joint Universality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229

12.1 A Joint Limit Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
12.2 A Transfer Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
12.3 Twisted L-Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
12.4 First Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
12.5 A Conjecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246

13 L-Functions of Number Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249

13.1 Dedekind Zeta-Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
13.2 Grssencharacters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
13.3 Hecke L-Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
13.4 Universality for Hecke L-Functions . . . . . . . . . . . . . . . . . . . . . . . . 260
13.5 Artins Reciprocity Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
13.6 Artin L-Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
13.7 The Artin Conjecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
13.8 Joint Universality for Artin L-Functions . . . . . . . . . . . . . . . . . . . . 274
13.9 L-Functions to Automorphic Representations . . . . . . . . . . . . . . . 278

A Appendix: A Short History of Universality . . . . . . . . . . . . . . . . 285

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293

Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
Preface

L-functions are important objects in modern number theory. They are gener-
ating functions formed out of local data associated with either an arithmetic
object or with an automorphic form. They can be attached to smooth pro-
jective varieties dened over number elds, to irreducible (complex or p-adic)
representations of the Galois group of a number eld, to a cusp form or to
an irreducible cuspidal automorphic representation. All the L-functions have
in common that they can be described by an Euler product, i.e., a product
taken over prime numbers. In view of the unique prime factorization of integers
L-functions also have a Dirichlet series representation. The famous Riemann
zeta-function


1   1
1
(s) = = 1 s
n=1
ns p
p prime

may be regarded as the prototype. L-functions encode in their value-

distribution information on the underlying arithmetic or algebraic structure
that is often not obtainable by elementary or algebraic methods. For instance,
Dirichlets class number formula gives information on the deviation from
unique prime factorization in the ring of integers of quadratic number elds
by the values of certain Dirichlet L-functions L(s, ) at s = 1. In particu-
lar, the distribution of zeros of L-functions is of special interest with respect
to many problems in multiplicative number theory. A rst example is the
Riemann hypothesis on the non-vanishing of the Riemann zeta-function in
the right half of the critical strip and its impact on the distribution of prime
numbers. Another example are L-functions L(s, E) attached to elliptic curves
E dened over Q. The yet unproved conjecture of Birch and Swinnerton-Dyer
claims that L(s, E) has a zero at s = 1 whose order is equal to the rank of
the Mordell-Weil group of the elliptic curve E.
These notes present recent results in the value-distribution theory of such
L-functions with an emphasis on the phenomenon of universality. The starting
point of this theory is Bohrs achievement at the rst half of the twentieth
X Preface

century. He proved denseness results and rst limit theorems for the values of
the Riemann zeta-function. Maybe the most remarkable result concerning the
value-distribution of (s) is Voronins universality theorem from 1975, which
roughly states that any non-vanishing analytic function can be approximated
uniformly by certain shifts of the zeta-function in the critical strip. More
precisely: let 0 < r < 14 and suppose that g(s) is a non-vanishing continuous
function on the disc |s| r which is analytic in its interior. Then, for any
 > 0, there exists a real number such that
   
 3 
max  s + + i g(s) < ;
|s|r 4

moreover, the set of these has positive lower density:

     
1  3 
 
lim inf meas [0, T ] : max  s + + i g(s) <  > 0.
T T |s|r 4

This is a remarkable property! We say that (s) is universal since it allows

uniform approximation of a large class of functions. Voronins universality
theorem, in a spectacular way, indicates that Riemanns zeta-function is a
transcendental function; clearly, rational functions cannot be universal. In
some literature the validity of the Riemann hypothesis for abelian varieties
(proved by Hasse for elliptic curves and by Weil in the general case) is regarded
as evidence for the truth of Riemanns hypothesis for (s). However, the
zeta-function of an abelian variety is a rational function and so its value-
distribution is of a rather dierent type.
The LinnikIbragimov conjecture asserts that any Dirichlet series (which
has a suciently rich value-distribution) is universal. Meanwhile we know
quite many universal Dirichlet series; for instance, Dirichlet L-functions
(Voronin, 1975), Dedekind zeta-functions (Reich, 1980), Lerch zeta-functions
(Laurinikas, 1997), and L-functions associated with newforms (Laurinikas,
Matsumoto and Steuding, 2003). One aim of these notes is to prove an exten-
sion of Voronins universality theorem for a large class of L-functions which
covers (at least conjecturally) all known L-functions of number-theoretical
signicance.
These notes are organized as follows. In the introduction, we give an
overview on the value-distribution theory of the classical Riemann zeta-
function and Dirichlet L-functions; also we touch some allied zeta-functions
which we will not consider in detail in the following chapters. In Chap. 2,
we introduce a class S of Dirichlet series, satisfying certain analytic and
arithmetic axioms. The members of this class are the main actors in the
sequel. Roughly speaking, an L-function in S has a polynomial Euler product
and satises some hypothesis which may be regarded as some kind of prime
number theorem; besides, we require analytic continuation to the left of
the half-plane of absolute convergence for the associated Dirichlet series in
addition with some growth condition. The axioms dening S are kept quite
 Preface XI

general and therefore they may appear to be rather abstract and technical;
however, as we shall discuss later for many examples (in Chaps. 6, 12 and
13), they hold (or at least they are expected to hold) for all L-functions of
number theoretical interest. This abstract setting has the advantage that we
can derive a rather general universality theorem.
Our proof of universality, in the main part, relies on Bagchis probabilistic
approach from 1981. For the sake of completeness we briey present in Chap. 3
some basic facts from probability theory and measure theory. In Chap. 4, we
prove along the lines of Laurinikas extension of Bagchis method a limit
theorem (in the sense of weakly convergent probability measures) for functions
in the class S. In the following chapter we give the proof of the main result, a
universality theorem for L-functions in S. The proof depends on the limit theo-
rem of the previous chapter and the so-called positive density method, recently
introduced by Laurinikas and Matsumoto to tackle L-functions attached to
cusp forms. Furthermore, we discuss the phenomenon of discrete universality;
here the attribute discrete means that the shifts are taken from arithmetic
progressions. This concept of universality was introduced by Reich in 1980.
In Chap. 6, we introduce the Selberg class S consisting of Dirichlet
series with Euler product and a functional equation of Riemann-type (and
a bit more). It is a folklore conjecture that the Selberg class consists of all
automorphic L-functions. We study basic facts about S and discuss the main
conjectures, in particular, the far-reaching Selberg conjectures on primitive
elements. We shall see that the class S ts rather well into the setting of
the Selberg class S (especially with respect to Selbergs conjectures). Hence,
our general universality theorem extends to the Selberg class, unconditionally
for many of the classical L-function and conditionally to all elements of S
subject to some widely believed but rather deep conjectures. However, the
Selberg class is too small with respect to universality; for instance, a Dirichlet
L-function to an imprimitive character does not lie in the Selberg class (by
lack of an appropriate functional equation) but it is known to be universal.
Furthermore, some important L-functions are only conjectured to lie in the
Selberg class, and, in spite of this, for some of them we can derive universality
unconditionally.
In the following chapter, we consider the value-distribution of Dirichlet
series L(s) with functional equation in the complex plane. Following Levin-
sons approach from the 1970s, we shall prove asymptotic formulae for the
c-values of L, i.e., roots of the equation L(s) = c, and give applications in
Nevanlinna theory. In particular, we give an alternative proof of the Riemann
von Mangoldt formula for the elements in the Selberg class.
The main themes of Chap. 8 are almost periodicity and the Riemann
hypothesis. Universality has an interesting feedback to classical problems.
Bohr observed that the Riemann hypothesis for Dirichlet L-functions asso-
ciated with non-principal characters is equivalent to almost periodicity in
the right half of the critical strip. Applying Voronins universality theorem,
Bagchi was able to extend this result to the zeta-function in proving that if
XII Preface

the Riemann zeta-function can approximate itself uniformly in the sense of

Voronins theorem, then Riemanns hypothesis is true, and vice versa. We
sketch an extension of Bagchis theorem to other L-functions.
Chapter 9 deals with the problem of eectivity. The known proofs of uni-
versality are ineective, giving neither bounds for the rst approximating shift
nor for their density (with the exception of particular results due to Garunk-
tis, Good, and Laurinikas). We give explicit upper bounds for the density
of universality; more precisely, we prove upper bounds for the frequency with
which a certain class of target functions (analytic isomorphisms) can be uni-
formly approximated. Moreover, we apply eective results from the theory of
inhomogeneous diophantine approximation to prove several explicit estimates
for the value-distribution in the half-plane of absolute convergence.
In Chap. 10, we discuss further applications of universality, most of them
classical, e.g., an extension of Bohrs and Voronins results concerning the
value-distribution inside the critical strip, and the functional independence
which covers Ostrowskis solution of the Hilbert problem on the hyper-
transcendence of the zeta-function and some of its generalizations. Here a
function is called hyper-transcendental, if it does not satisfy any algebraic
dierential equation. Further, we study the value-distribution of linear com-
binations of (strongly) universal Dirichlet series. A subtle consequence of this
strong concept of universality, and a big contrast to L-functions, can be found
in the distribution of zeros o the critical line. Very likely a (universal) Dirich-
let series satisfying a functional equation of Riemann-type has either many
zeros to the right of the critical line (as a generic Dirichlet series with periodic
coecients) or none (as it is expected for L-functions). This seems to be the
heart of many secrets in the value-distribution theory of Dirichlet series.
Chapter 11 deals with Dirichlet series associated with periodic arithmeti-
cal functions. In general, these functions do not have an Euler product but
they are additively related to Dirichlet L-functions. Consequently, they share
certain properties with L-functions, e.g., a functional equation similar to the
one for Riemanns zeta-function. We prove universality for a large class of
these Dirichlet series; in contrast to L-functions they can approximate uni-
formly analytic functions having zeros (provided their Dirichlet coecients are
not multiplicative). Moreover, we study joint universality for Hurwitz zeta-
functions with rational parameters.
We conclude with joint universality; here joint stands for simultaneous
uniform approximation. In Chap. 12, we prove a theorem which reduces joint
universality for L-functions in S to a denseness property in a related function
space. Of course, we cannot have joint universality for any set of L-functions;
for example, (s) and (s)2 cannot approximate any given pair of admissi-
ble target functions simultaneously. However, we shall prove that in some
instances twists of L S with pairwise non-equivalent characters fulll this
condition (e.g., Dirichlet L functions). In the following chapter we present sev-
eral further applications. For instance, we prove joint universality for Artin
 Preface XIII

L-functions (which lie in the Selberg class if and only if the deep Artin con-
jecture is true). This universality theorem holds unconditionally despite the
fact that Artin L-functions might have innitely many poles in their strip of
universality; this was rst proved by Bauer in 2003 by a tricky argument.
At the end of these notes an appendix on the history of the general phe-
nomenon of universality in analysis is given. It is known that universality is a
quite regularly appearing phenomenon in limit processes, but among all these
universal objects only universal Dirichlet series are explicitly known. At the
end an index and a list of the notations and axioms which were used are given.
Value-distribution theory for L-functions with emphasis on aspects of uni-
versality was treated in the monographs of Karatsuba and Voronin [166] and
Laurinikas [186]. However, after the publication of these books, many new
results and applications were discovered; we refer the reader to the surveys of
Laurinikas [196] and of Matsumoto [242] for some of the progress made in the
meantime. The content of this book forms an extract of the authors habilita-
tion thesis written at Frankfurt University in 2003. We have added Chaps. 12
and 13 on joint universality and its applications as well as several remarks
and comments concerning the progress obtained in the meantime. Unfortu-
nately, we could not include the most current contributions as, for example,
the promising work [245] of Mauclaire which relates universality with almost
periodicity.
I am very grateful to Springer for publishing these notes; especially, I want
to thank Stefanie Zller and Catriona M. Byrne from Springer, the editors of
the series Lecture Notes in Mathematics, and, of course, the anonymous referees
for their excellent work, their valuable remarks and corrections. Furthermore,
I am grateful to my family, my friends and my colleagues for their interest
and support, in particular those from the Mathematics Departments at the
universities of Frankfurt, Madrid, and Wrzburg. Especially, I would like to
thank Ramunas Garunktis and Antanas Laurinikas for introducing me to
questions concerning universality, Ernesto Girondo, Aleksander Ivi, Roma
Kainskaite, Kohji Matsumoto, Georg Johann Rieger, Jrgen Sander, Wolf-
gang Schwarz, and Jrgen Wolfart for the fruitful discussions, helpful remarks
and their encouragement. Last but not least, I would like to thank my wife
Rasa.

Jrn Steuding
Wrzburg, December 2006
1
Introduction

The grandmother of all zeta-functions is the Riemann zeta-function.

David Ruelle

In this introduction we give some hints for the importance of the

Riemann zeta-function for analytic number theory and present rst classic
results on its amazing value-distribution due to Harald Bohr but also the re-
markable universality theorem of Voronin (including a sketch of his proof).
Moreover, we introduce Dirichlet L-functions and other generalizations of the
zeta-function, discuss their relevance in number theory and comment on their
value-distribution. For historical details we refer to Narkiewiczs monograph
[277] and Schwarzs surveys [317, 318].

1.1 The Riemann Zeta-Function and the Distribution

of Prime Numbers
The Riemann zeta-function is a function of a complex variable s = + it, for
> 1 given by


1  1
1
(s) = = 1 s ; (1.1)
n=1
ns p
p

here and in the sequel the letter p always denotes a prime number and the
product is taken over all primes. The Dirichlet series, and the Euler product,
converge absolutely in the half-plane > 1 and uniformly in each compact
subset of this half-plane. The identity between the Dirichlet series and the
Euler product was discovered by Euler [76] in 1737 and can be regarded as
an analytic version of the unique prime factorization of integers. The Euler
product gives a rst glance on the intimate connection between the zeta-
function and the distribution of prime numbers. A rst immediate consequence
is Eulers proof of the innitude of the primes. Assuming that there were only
2 1 Introduction

nitely many primes, the product in (1.1) is nite, and therefore convergent
for s = 1, contradicting the fact that the Dirichlet series dening (s) reduces
to the divergent harmonic series as s 1+. Hence, there exist innitely many
prime numbers. This fact is well known since Euclids elementary proof, but
the analytic access gives deeper knowledge on the distribution of the prime
numbers. It was the young Gauss [94] who conjectured in 1791 for the number
(x) of primes p x the asymptotic formula

(x) li(x), (1.2)

where the logarithmic integral is given by

 1
 x   x
du du
li(x) = lim + = 1.04 . . . ;

0+ 0 1+
log u 2 log u

this integral is a principal value in the sense of Cauchy. Gauss conjecture

states that, in rst approximation, the number of primes x is asymptotically
x
log x . By elementary means, Chebyshev [54, 55] proved around 1850 that for
suciently large x
log x
0.921 . . . (x) 1.055 . . . .
x
Furthermore, he showed that if the limit
log x
lim (x)
x x
exists, the limit is equal to one, which supports relation (1.2).
Riemann was the rst to investigate the Riemann zeta-function as a func-
tion of a complex variable. In his only one but outstanding paper [310] on
number theory from 1859 he outlined how Gauss conjecture could be proved
by using the function (s). However, at Riemanns time the theory of func-
tions was not developed suciently far, but the open questions concerning
the zeta-function pushed the research in this eld quickly forward. We shall
briey discuss Riemanns memoir. First of all, by partial summation

1 
N 1s [u] u
(s) = + +s du; (1.3)
ns s1 N us+1
nN

here and in the sequel [u] denotes the maximal integer less than or equal to u.
This gives an analytic continuation for (s) to the half-plane > 0 except for
a simple pole at s = 1 with residue 1. This process can be continued to the
left half-plane and shows that (s) is analytic throughout the whole complex
plane except for s = 1. Riemann gave the functional equation
s
 
1s
s/2 (s) = (1s)/2 (1 s), (1.4)
2 2
 1.1 The Riemann Zeta-Function and the Distribution of Prime Numbers 3

0.1

0.05

14 12 10 8 6 4 2

0.05

0.1

0.15

Fig. 1.1. (s) in the range s [14.5, 0.5]

where (s) denotes Eulers Gamma-function; it should be noted that

Euler [76] had partial results in this direction (namely, for integral s and
for half-integral s; see [7]). In view of the Euler product (1.1) it is easily seen
that (s) has no zeros in the half-plane > 1. It follows from the functional
equation and from basic properties of the Gamma-function that (s) vanishes
in < 0 exactly at the so-called trivial zeros s = 2n with n N (see Fig. 1.1
for the rst trivial zeros). All other zeros of (s) are said to be non-trivial,
and we denote them by  = + i. Obviously, they have to lie inside the
so-called critical strip 0 1, and it is easily seen that they are non-real.
The functional equation (1.4), in addition with the identity
(s) = (s),
shows some symmetries of (s). In particular, the non-trivial zeros of (s) are
distributed symmetrically with respect to the real axis and to the vertical line
= 12 . It was Riemanns ingenious contribution to number theory to point
out how the distribution of these non-trivial zeros is linked to the distribution
of prime numbers. Riemann conjectured that the number N (T ) of non-trivial
zeros  = + i with 0 < T (counted according multiplicities) satises
the asymptotic formula
T T
N (T ) log .
2 2e
This was proved in 1895 by von Mangoldt [235, 236] who found more precisely
T T
N (T ) = log + O(log T ). (1.5)
2 2e
Riemann worked with the function t  ( 12 + it) and wrote that very likely
all roots t are real, i.e., all non-trivial zeros lie on the so-called critical line
4 1 Introduction

0
10
0
10
20
30
40

Fig. 1.2. The reciprocal of the absolute value of (s) for [4, 4], t [10, 40].
The zeros of (s) appear as poles

= 12 . This is the famous, yet unproved Riemann hypothesis which we rewrite

equivalently as

Riemanns Hypothesis. (s) = 0 for > 12 .

In support of his conjecture, Riemann calculated some zeros; the rst one
with positive imaginary part is  = 12 + i14.134 . . . (see Fig. 1.2 and also
Fig. 8.1). Furthermore, Riemann conjectured that there exist constants A
and B such that
1 s
 s
  
s
s/2
s(s 1) (s) = exp(A + Bs) 1 exp .
2 2 
 

This was shown by Hadamard [113] in 1893 (recall the Hadamard product
theorem from the theory of functions). Finally, Riemann conjectured the so-
called explicit formula which states that

In 1932, Siegel [329] published an account of Riemanns work on the zeta-function

found in Riemanns private papers in the archive of the university library in
Gttingen. It became evident that behind Riemanns speculation there was
extensive analysis and computation.
 1.1 The Riemann Zeta-Function and the Distribution of Prime Numbers 5


(x1/n ) 
(x) + = li(x) li(x ) + li(x1 ) (1.6)
n=2
n
=+i
>0


du
+ log 2
x u(u2 1) log u

for any x 2 not being a prime power (otherwise a term 2k 1

has to be added on
k
the left-hand side, where x = p ); the appearing integral logarithm is dened
by
 (+i) log x
+i exp(z)
li(x )= dz,
(+i) log x z + i

where = +1 if > 0 and = 1 otherwise. The explicit formula was proved

by von Mangoldt [235] in 1895 as a consequence of both product representa-
tions for (s), the Euler product (1.1) on the one hand and the Hadamard
product on the other hand.
Riemanns ideas led to the rst proof of Gauss conjecture (1.2), the cel-
ebrated prime number theorem, by Hadamard [114] and de la Valle-Poussin
[357] (independently) in 1896. We give a very brief sketch (for the details we
refer to Ivi [141]). For technical reasons it is of advantage to work with the
logarithmic derivative of (s) which is for > 1 given by

 (n)
(s) = ,
n=1
ns

where the von Mangoldt -function is dened by

log p if n = pk with k N,
(n) = (1.7)
0 otherwise.

A lot of information concerning the prime counting function (x) can be

recovered from information about

(x) := (n) = log p + O x1/2 log x .

nx px

Partial summation gives

(x)
(x) .
log x
First of all, we shall express (x) in terms of the zeta-function. If c is a positive
constant, then
 c+i s
1 x 1 if x > 1,
ds = (1.8)
2i ci s 0 if 0 < x < 1.
6 1 Introduction

This yields the Perron formula: for x  Z and c > 1,

 c+i 
1 xs
(x) = (s) ds. (1.9)
2i ci s
Moving the path of integration to the left, we nd that the latter expression
is equal to the corresponding sum of residues, that are the residues of the
integrand at the pole of (s) at s = 1, at the zeros of (s), and at the pole of
the integrand at s = 0. The main term turns out to be
     s
xs 1 x
Res s=1 (s) = lim (s 1) + O(1) = x,
s s1 s1 s
whereas each non-trivial zero  gives the contribution
  
xs x
Res s= (s) = .
s 
By the same reasoning, the trivial zeros contribute

 
x2n 1 1
= log 1 2 .
n=1
2n 2 x

Incorporating the residue at s = 0, this leads to the the exact explicit formula

x  
1 1
(x) = x log 1 2 log(2),

 2 x

which is equivalent to Riemanns formula (1.6). Notice that the right-hand side
of this formula is not absolutely convergent. If (s) would have only nitely
many non-trivial zeros, the right-hand side would be a continuous function of
x, contradicting the jumps of (x) for prime powers x. However, going on it
is much more convenient to cut the integral in (1.9) at t = T which leads to
the truncated version

x x

(x) = x +O (log(xT ))2 , (1.10)

 T
||T

valid for all values of x. Next we need information on the distribution of the
non-trivial zeros. The largest known zero-free region for (s) was found by
Vinogradov [359] and Korobov [173] (independently) who proved
c
(s) = 0 in 1 ,
(log |t| + 3)1/3 (log log(|t| + 3))2/3
where c is some positive absolute constant; the rst complete proof due
to Richert appeared in Walsz [366]. In addition with the Riemannvon
Mangoldt formula (1.5) one can estimate the sum over the non-trivial zeros
in (1.10). Balancing out T and x, we obtain the prime number theorem with
the strongest existing remainder term:
 1.1 The Riemann Zeta-Function and the Distribution of Prime Numbers 7

Theorem 1.1. There exists an absolute positive constant C such that for suf-
ciently large x
  
(log x)3/5
(x) = li (x) + O x exp C .
(log log x)1/5
By the explicit formula (1.10) the impact of the Riemann hypothesis on the
prime number distribution becomes visible. Von Koch [172] showed that for
xed [ 12 , 1)
(x) li (x)  x+
(s) = 0 for > ; (1.11)
here and in the sequel  stands for an arbitrary small positive constant, not
necessarily the same at each appearance. With regard to known zeros of (s)
on the critical line it turns out that an error term with < 12 is impossible.
Thus, the Riemann hypothesis states that the prime numbers are as uniformly
distributed as possible!
Many computations were done to nd a counterexample to the Riemann
hypothesis. Van de Lune, te Riele and Winter [232] localized the rst
1 500 000 001 zeros, all lying without exception on the critical line; moreover
they all are simple! By observations like this it is conjectured, that all or
at least almost all zeros of the zeta-function are simple. This so-called essen-
tial simplicity hypothesis has arithmetical consequences. Cramr [63] showed,
assuming the Riemann hypothesis,
 X 2  
1 (x) x
 m() 2
dx  
log X 1 x    , (1.12)


where the sum is taken over distinct zeros and m() denotes their multiplicity.
The right-hand side is minimal if all the zeros are simple. Going further,
Goldston, Gonek and Montgomery [103] observed interesting relations
between the essential simplicity hypothesis, mean-values of the logarith-
mic derivative of (s), the error term in the prime number theorem, and
Montgomerys pair correlation conjecture.
A classical density theorem due to Bohr and Landau states that most
of the zeros lie arbitrarily close to the critical line. Denote by N (, T ) the
number of zeros  = + i of (s) for which > and 0 < T (counting
multiplicities). Bohr and Landau [34] (see also [224]) proved that
N (, T )  T = O(N (T )) (1.13)
1
for any xed > 2. Hence, almost all zeros of the zeta-function are clustered
around the critical line. A renement of their method led to bounds N (, T ) 
T with < 1 (see [35, 49]); for instance:
1
Theorem 1.2. For any xed with 2 < < 1,

N (, T )  T 4(1)+
.
For the various improvements of this density estimate we refer to Ivi [141].
8 1 Introduction

Hardy [116] showed that innitely many zeros lie on the critical line, and
Selberg [321] was the rst to prove that a positive proportion of all zeros lies
exactly on = 12 . Let N0 (T ) denote the number of zeros  of (s) on the
critical line with imaginary part 0 < T . The idea to use molliers to
dampen the oscillations of |( 12 + it)| led Selberg to
N0 (T + H) N0 (T )
lim inf > 0,
T N (T + H) N (T )
as long as H T 1/2+
. Karatsuba [164] improved this result to H T 27/82+

by some technical renements. The proportion is very small, about 106 as
2
Min calculated; a later renement by Zhuravlev gives after all 21 if H = T (cf.
[165, p. 36]). However, the localized zeros are not necessarily simple. By an
ingenious new method, working with molliers of nite length, Levinson [216]
localized more than one third of the non-trivial zeros of the zeta-function on
the critical line, and as Heath-Brown [122] and Selberg (unpublished) discov-
ered, they are all simple. By optimizing the technique, Levinson himself and
others improved the proportion 13 slightly, but more recognizable is Conreys
idea in introducing Kloosterman sums. So Conrey [57] was able to choose a
longer mollier to show that more than two-fths of the zeros are simple and
on the critical line. Bauer [17, 18] improved this proportion slightly. The use
of longer molliers leads to larger proportions. Farmer [77] observed that if it
is possible to take molliers of innite length, then almost all zeros lie on the
critical line and are simple. In [339], Steuding found a new approach (com-
bining ideas and methods of Atkinson, Jutila and Motohashi) to treat short
intervals [T, T + H], i.e., H = O(T ). It was proved that for H T 0.552 a
positive proportion of the zeros of the zeta-function with imaginary parts in
[T, T + H] lie on the critical line and are simple.
Recently, Garaev [81] showed that


|  ()|1  (log T )1/2 ;
0<<T

(
)=0


the divergence of the series  |  ()|1 was known before only subject to
the truth of the Riemann hypothesis (see [353, p. 374]). Garaevs result was
slightly improved by himself [82] and leeviiene and Steuding [333]