Lecture Notes in Mathematics Editors: A. Dold, Heidelberg B. Eckmann, ZUrich E Takens, Groningen Subseries: Mathematisches Institut der Universit~it und Max-Planck-Institut fiir Mathematik Bonn - vol. 21 Advisor: F. Hirzebruch
1593
Jay Jorgenson & Serge Lang Dorian Goldfeld
Explicit Formulas for Regularized Products and Series
Springer-Verlag Berlin Heidelberg NewYork London Paris Tokyo Hong Kong Barcelona Budapest
Authors Jay Jorgenson Serge Lang Mathematics Department Box 208283 Yale Station 10 Hillhouse Ave New Haven CT 06520-8283, USA Dorian Goldfeld Mathematics Department Columbia Unversity New York, NY 10027, USA
Mathematics Subject Classification (1991 ): 11 M35, 11M41,11 M99, 30B50, 30D 15, 35P99, 35S99, 42A99 Authors Note: there is no MSC number for regularized products, but there should be.
ISBN 3-540-58673-3 Springer-Verlag Berlin Heidelberg New York
CIP-Data applied for 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, re-use of illustrations, recitation, broadcasting, reproduction on microfilms 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-Verlag. Violations are liable for prosecution under the German Copyright Law. 9 Springer-Verlag Berlin Heidelberg 1994 Printed in Germany Typesetting: Camera-ready TEX output by the authors SPIN: 10130182 46/3140-543210 - Printed on acid-free paper
EXPLICIT
FORMULAS
REGULARIZED
FOR
PRODUCTS
AND SERIES
Jay Jorgenson and Serge Lang
A SPECTRAL
INTERPRETATION
OF WEIL'S EXPLICIT
FORMULA
Dorian Goldfeld
EXPLICIT
FORMULAS FOR REGULARIZED PRODUCTS AND SERIES
Jay J o r g e n s o n and Serge Lang
Introduction I A s y m p t o t i c e s t i m a t e s of regularized h a r m o n i c series 1. 2. 3. 4. 5. 6. 7.
Regularized products and harmonic series Asymptotics in verticM strips Asymptotics in sectors Asymptotics in a sequence to the left Asymptotics in a parMlel strip Regularized product and series type Some examples
II Cramdr's T h e o r e m as an Explicit F o r m u l a
3
11 14 20 22 24 34 36 39 43
Euler sums and functional equations The general Cram~r formula Proof of the Cramdr theorem An inductive theorem
45 47 51 57
III Explicit Formulas u n d e r Fourier A s s u m p t i o n s
61
1. 2. 3. 4.
1. 2. 3. 4. 5.
Growth conditions on Fourier transforms The explicit formulas The terms with the q's The term involving The Well functional and regularized product type
62 66 73 78 79
IV From F u n c t i o n a l E q u a t i o n s to T h e t a Inversions 85 1. An application of the explicit formulas 2. Some examples of theta inversions
87 92
Viii
V From Theta Inversions to Functional Equations 1. 2. 3. 4.
The Weft functional of a Gaussian test function Gauss transforms Theta inversions yield zeta functions A new zeta function for compact quotients of M3
V I A G e n e r a l i z a t i o n o f Fujii's T h e o r e m
97
99 101 109 113 119
122 125 128
1. Statement of the generalized Fujii theorem 2. Proof of Fujii's theorem 3. Examples
131
Bibliography
A SPECTRAL INTERPRETATION WEIL'S EXPLICIT FORMULA
OF
Dorian Goldfeld
1. 2. 3. 4. 5. 6. 7.
Introduction Notation Construction of the indefinite space s Spectral theory of s Eisenstein series Cusp forms The zeta function associated to an automorphic form on L2(T) 8. The Rankin-Selberg convolution 9. Higher rank generalizations 10. References
Index
137 139 140 141 142 145 147 148 148 152
153
EXPLICIT
FORMULAS
REGULARIZED
FOR
PRODUCTS
AND SERIES
Jay Jorgenson and Serge Lang
Introduction Explicit formulas in number theory were originally motivated by the counting of primes, and Ingham's exposition of the classical computations is still a wonderful reference [In 32]. Typical of these formulas is the Riemann-von Mangoldt formula log
pn<x
p
=
x -
}; P
x P --
-
1 log(1 - x - 2 ) .
P
Here the sum on the left is taken over all prime powers, and the sum on the right is taken over the non-trivial zeros of the Riemann zeta function. Later, Weil [We 52] pointed out that these formulas could be expressed much more generally as stating that the sum of a suitable test function taken over the prime powers is equal to the sum of the Mellin transform of the function taken over the zeros of the zeta function, plus an analytic term "at infinity", viewed as a functional evaluated on the test function. It is the purpose of these notes to carry through the derivation of the analogous so-called "explicit formulas" for a general zeta function having an Euler sum and functional equation whose fudge factors are of regularized product type. As a result, our general theorem applies to many known examples, some of which are listed in w of [JoL 93c]. The general Parseval formula from [JoL 93b] provides an evaluation of the "term at infinity", which we call the Weil functional. Also, as an example of our results, let us note that even in the well-studied case of the Selberg zeta function of a compact Riemann surface, our computations show that one may deal with a larger class of test functions than previously known. For some time, analogies between classical analytic number theory and spectral theory have been realized. MinakshisundaramPleijel defined a zeta function in connection with the Laplacian on an arbitrary Riemannian manifold [MiP 49], and subsequently Selberg defined his zeta function [Se 56]. In [JoL 93a,b] we developed a general theory of regularized products and series applicable equally to the classical analytic number theory and to some of these
analogous spectral situations. In particular, we proved the basic properties of the Weil functional at infinity in the context of regularized products and series, with a view to using the functional for the explicit formulas in this general context. A f u n d a m e n t a l class o f z e t a f u n c t i o n s . In [JoL 93c] we defined a fundamental class of functions to which we could apply these properties and carry out analogues of results in analytic number theory. Roughly speaking, the functions Z in our class are those which satisfy the three conditions: there is a functional equation; the logarithm of the function admits a generalized Dirichlet series converging in some half plane (we call this Dirichlet series an Euler sum for Z); - the fudge factors in the functional equation are of regularized product type. The precise definition of our class of functions is recalled in Chapter II, w The explicit formula can be formulated and proved for functions in this class. In Chapter II, w we discuss the extent to which this class is a much broader class than a certain class defined by Selberg [Se 91]. Furthermore, certain applications require an even broader class of functions to which all the present techniques can be applied. We shall describe the need for such a class in greater detail below. Just as we did for the analogue of Cram~r's theorem proved in [JoL 93c], we emphasize that the explicit formula involves an inductive step which describes a relation between some of the zeros and poles of the fudge factors and some of those of the principal zeta function Z. Such a step can be viewed as a step in the ladder of regularized products, because our generalized Cram~r theorem insures that a function Z in our class is also of regularized product type provided the fudge factors are of regularized product type. If Z is a function in our class, and, for Re(s) sufficiently large, the expression log Z(s) -- ~ q
is the Euler sum for log Z(s), with a sequence {q} of real numbers 1 tending to infinity, and complex coefficients c(q), then such q play the role of prime powers. However, readers should keep in mind cases when q does not look at all like a prime power. For example, the general theory applies to the case when Z(s) is a general
Dirichlet polynomial, up to an exponential fudge factor; a precise definition is given in Chapter II, w Such polynomials contain as special cases the local factors of more classical zeta functions and L-functions. In examples having to do with Riemannian geometry, log q is the Riemannian distance between two points in the universal covering space. The general version of Cram6r's theorem in [JoL 93c] was carried out for the original Cram6r's test function Cz(s) = r One can also view this version as a special case of an explicit formula with more general test functions. This is carried out in Chapter II. In [JoL 93c], w we gave a number of examples for our Cram6r-type theorem. To these we are adding not only the general Dirichlet polynomials as mentioned above, but also Fujii-type L-functions, obtained from a zeta function by inserting what amounts to a generalized character as coefficient of the Dirichlet series defining the zeta function (see the papers by Fujii listed in the bibliography). In Chapter V we show both how to recover Fujii's theorems for the functions he considered, namely the Riemann zeta function and the Selberg zeta function for PSL(2,Z), as well as an analogous theorem for the general zeta functions in our class, all as corollaries of our Cram6r's theorem. Similarly, a result of Venkov, which relates the eigenvalues of the Laplacian relative to PSL(2,Z) to the classical von Mangoldt function, will be generalized to any non-compact finite volume hyperbolic Riemann surface in [JoL 94]. The generalization involves another inductive type argument, using the fact that the fudge factor in the functional equation of the non-compact Selberg zeta function involves the determinant of the scattering matrix, which itself is in our class of functions since it has an Euler sum and a simple functional equation with constant fudge factors. In this case, the Euler sum exists whereas a classical Euler product does not. Thus, the general theory simultaneously contains previous results and gives new ones which were not proved previously by authors using such tools as the Selberg trace formula. A n a l y t i c e s t i m a t e s for t h e p r o o f . In addition to the Parseval formula of [JoL 93b], the proof of the general explicit formula relies on certain analytic estimates for regularized harmonic series, including the logarithmic derivatives of regularized products in strips. We gave such estimates already in [JoL 93a,b], but we need further such estimates which we present in Chapter I of the present work, using the technique of our generalized Gauss formula. Hard-core analytic estimates having thus been put out of the way, the rest of the work is then relatively formal. It is noteworthy that to each regularized product we associate naturally two non-
negative integers determined directly from the definition. Then the fundamental estimates of Chapter I show that the order of growth of the logarithmic derivatives of such products in strips is determined by these two integers. In the application to the evaluation of certain integrals involving test functions, one can then see that the order of decay of these test functions, needed to insure that the integrals converge, is also determined by these two integers. Our systematic approach both improves known estimates for the Selberg zeta function (cf. Chapter I, w and provides estimates for functions in our class which had not been considered previously. T h e t a i n v e r s i o n s . We shall postpone to still another work the application of explicit formulas to the counting of those objects which play the role of prime powers. Here we shall emphasize an entirely different type of application, obtained by taking Gaussian type functions as the test functions instead of other test functions which lead to the counting. Applying the general explicit formula to such Gaussians gives rise to relations which are vast generalizations of the classical Jacobi inversion formula for the classical Jacobi theta function, where t on one side gets inverted to lit on the other side of the formula. The classical Jacobi inversion formula is the relation 1 2rr E
n=--OO
1 ~e_(2,~,)2/4t, 4v/-~t
e-"~'-
n=--
which holds for all t > 0. Here, log q = 27rn where n is a positive integer. The zeta function Z(s) giving rise to the above theta series is essentially the special Dirichlet polynomial sin(
is)
= -
1
Ir,(1 _ e_2rts).
Thus, the most classical theta series appears in a new context, associated to a "zeta function" which looks quite different from those visualized classically. The general context of Chapter IV and Chapter V allows a formulation of a theta inversion when the theta series is of type
•
ake-,xk*
k
with various coefficients ak. Theta inversion applies in certain cases when the sequence {Ak} is the sequence of eigenvalues of an operator. For example, as we will show in Chapter V, w such an
inversion formula comes directly from considering the heat kernel on the compact quotient of an odd dimensional hyperbolic space which has metric with constant negative sectional curvature. For certain manifolds, the theta inversion already gives rise to an extended class of zeta functions, which instead of an Euler sum may have a Bessel sum. For manifolds of even dimension, the class of functions having an Euler sum or Bessel sum is still not adequate, and it is necessary to define an even further extended class, which we shall describe briefly below. At this moment, it is not yet completely clear just how far an extension we shM1 need, but so far, whatever the extension of the fundamental class we have met, the techniques of [JoL 93a,b,c] and of Chapter I apply. In [JoL 94], we show how the general explicit formula also applies to the scattering determinant of Eisenstein series. Here, the Euler sum exists, and scattering determinants are in the fundamental class.
An additive theory rather than multiplicative theory, and a n e x t e n d e d class o f f u n c t i o n s . The conditions defining our fundamental class of functions are phrased in a manner still relatively close to the classical manner, involving the functions multiplicatively. However, it turns out that many essential properties of these functions involve only their logarithmic derivative, and thus give rise to an additive theory. For a number of applications, it is irrelevant that the residues are integers, and in some applications we are forced to deal with the more general notions of a regularized harmonic series (suitably normalized Mittag-Leffier expansions, with poles of order one) whose definition is recalled in Chapter I, w In general, the residues of such a series are not integers, so one cannot integrate back to realize this series as a logarithmic derivative of a meromorphic function. Even for the Artin L-functions, although they can be defined by an Euler product, it was natural for Artin to define them via their logarithmic derivative, and at the time, Artin could only prove that the residues were rational numbers. It took many years before the residues were finally proved to be integers. The systematic approach of [JoL 93a,b,c] in fact has been carried out so that it applies to this additive situation. The example of Chapter V, w shows why such an additive theory is essential. Thus we are led to define not only the fundamental class of functions whose logarithmic derivative admits a Dirichlet series expression as mentioned above, but an extended class of functions where this condition is replaced by another one which will allow appli-
cations to more situations, starting with applications to various spectral theories as in [JoL 94]. Nevertheless, we still defined the fundamental class of functions having Euler sums, and we phrase some results multiplicatively, partly because at the present time, we feel that a complete change of notation with existing works would only make the present work less accessible, and partly because the class of functions admitting Euler sums is still a very important one including the classical functions of algebraic number theory and representation theory. However, we ask readers to keep in mind the additive rather than multiplicative formalism. Many sections, e.g. Chapter I and w and w of Chapter V, are written so that they apply directly to the additive situation. Functions in the multiplicative fundamental class are obtained as Mellin transforms of theta functions having an inversion formula. Functions in the extended additive class are obtained as regularized harmonic series which are Gaussian transforms of such theta functions. For example, the (not regularized) harmonic series obtained from the heat kernel theta function in the special case of compact quotients of the three dimensional, complete, simply connected, hyperbolic manifold is essentially
Ck(x)r k
Observe how the presence of s ( s - 2) in the series formally insures a trivial functional equation, that is invariance under s ~-* 2 - s. Conversely, given a function in our extended additive class, one may go in reverse and see that the original theta inversion is only a special case of the general explicit formula valid for much more general test functions. The existence of an explicit formula with a more general test function will then allow us to obtain various counting results in subsequent publications. Finally, let us note that many examples of explicit formulas using various test functions involving many examples of zeta functions have been treated in the literature, providing a vast number of papers on the subject. Most of the papers dealing with such explicit formulas are not directly relevant for what we do here, which is to lay out a general inductive "ladder principle" for explicit formulas in line with our treatment of Cram6r's theorem. For instance, Deninger in [Den 93] emphasizes the compatibility of an explicit formula for the Riemann zeta function with a conjectural formalism of a Lefschetz trace formula. Such a formalism might occur in the
presence of an operator whose eigenvalues are zeros of the zeta function. Our inductive hypotheses cover a wider class of functions than in [Den 93], and our treatment emphasizes another direction in the study of regularized products and series. Factors of regularized product type behave as if there were an operator, but no operator may be available. We also mention Gallagher's attempt to unify a treatment of Selberg's trace formula with treatments of ordinary analytic number theory [Ga 84]. However, the conditions under which Gallagher proves his results are very restrictive compared to ours, and, in particular, are too restrictive to take into account the inductive ladder principle which we are following. A c k n o w l e d g e m e n t : During the preparation of this work, the first author received support from NSF grant DMS-93-07023. Both authors benefited from visits to the Max-Planck-Institut in Bonn.
CHAPTER I A s y m p t o t i c e s t i m a t e s of regularized harmonic series. The proof of the general explicit formulas for functions whose fudge factors are of regularized product type will require a number of asymptotic estimates of general regularized harmonic series. The purpose of this chapter is to establish and tabulate these estimates in convenient form. The main definitions and results of this chapter are stated in w and w These asymptotic formulas are needed just as one needs the asymptotic behavior of the gamma function and the zeta function in classical analytic number theory (see, for example, Chapter XVII of [La 70]). However, classical arguments which estimate this behavior cannot be applied in general, and must be replaced by more powerful tools, such as our extension of Cram~r's theorem, proved in [JoL 93c], as well as our systematic analysis of the regularized harmonic series, given in [JoL 93a] and [JoL 935]. Following the notation of [JoL 93a] and [JoL 93b], we let R(z) be the regularized harmonic series associated to the theta function 8(t) = ~ ake-Xkt; in other words
R(z) -- CTs=ILM~9(s,z) where L M is the Laplace-Mellin transform, and CTs=I is the constant term of the power series in s at s = 1. As is shown in w of [JoL 93a], the function R(z) has a meromorphic continuation to all z E C whose singularities are simple poles located at z - --Ak with corresponding residue ak. In what we call t h e s p e c t r a l c a s e , meaning ak E Z for all k, we have a regularized product D(z) which is a meromorphic function defined for all z C C and which satisfies the relation
R(z) = D'/D(z).
12
However, in this chapter, we will work in the more general situation by considering a regularized harmonic series which is not necessarily the logarithmic derivative of a regularized product. Our basic tool for estimating R(z) is our general Gauss formula, which we shall recall at the end of w1. From the general Gauss formula, we shall determine the asymptotic behavior of R(z) as z ~ oo in each of the following cases: 1) in a vertical strip obtained by restricting Re(z) to a compact interval; 2) in a sector IIm(z)l << Re(z); 3) in a sequence of vertical line segments z = -Tm + iy with Tm ---, oe and y in a compact interval so that R(z) grows as slowly as possible; 4) in a strip :parallel and disjoint from a strip which contains all --Ak, assuming that Ak ~ oo in a horizontal strip. In all cases, the asymptotic behavior of R(z + w) will be determined from the general Gauss formula through judicious choices of z and W.
In w we apply these results to functions which are obtained from regularized products,by a suitable change of variables z ~ a z +/~ with (~, 3 E C. Such a change of variables is needed, for example, because zeta functions usually have their zeros in vertical strips and not in horizontal strips. Power products of regularized products subject to such changes of variables will be said to be of regularized product type, formally defined in w We conclude this chapter by comparing our results to various examples that exist elsewhere in the literature. Recall that a regularized harmonic series has a natural "reduced order M" which is closely related to the exponent of the leading term in the asymptotic expansion of the associated theta function near t = 0. We recall the precise definition at the end of w1. There is another characterization of the reduced order in the spectral case, since we can then write the regularized product as a Weierstrass product
D(z) =
P(Z)E(z)
where P has degree m + 1 and E is a canonical Weierstrass product of order _< rn. The smallest integer rn for which such an expression is possible is the reduced order of D. In the classical case of analytic
13 n u m b e r theory, the g a m m a and zeta functions of number fields have reduced order O.
From w to w we let R be a regularized harmonic series whose definition will be recalled in w as well as other basic definitions used throughout the chapter.
14
w
Regularized products and h a r m o n i c series.
Let us briefly recall necessary background material from the theory of regularized products and series, as established in [3oL 93a] and [3oL 93b], to which we refer for details and further results. We let L = { Ak} and A = {ak } be sequences of complex numbers which may be subject to the following conditions. D I R 1. For every positive real number c, there is only a finite number of k such that Re(Ak) < c. We use the convention that A0 = 0 a n d A k # 0 f o r k > 1. Under condition D I R 1 we delete from the complex plane C b e horizontal half lines going from - ~ to -Ak for each k, together, when necessary, the horizontal half line going from - o o to 0. We define the open set: U L =
the complement of the above half lines in C.
If all .kk are real and positive, then we note that UL is simply C minus the negative real axis R<0. D I R 2. (a) The Dirichlet series
E ak converges absolutely for some real a, say a0. (b) The Dirichlet series 1 k
J
converges absolutely for some real a, say al. D I R 3. There is a fixed c > 0 such that for all k sufficiently large, we have 7r
- - ~ -4- ~ < arg()~k) < --
--
7r m 2
m
~~
15
We will consider a t h e t a s e r i e s or t h e t a f u n c t i o n , which is defined by oo
OA,L(t) = O(t) = ao + Z ake-X*t' k----1
and, for each integer N >_ 1, we define the a s y m p t o t i c e x p o n e n t i a l p o l y n o m i a l s by N-1
QN(t) = ao + Z
ake-X*t"
k=l
We are also given a sequence of complex numbers {p} = {pj} with Re(p0) < Re(p1) < . . . < Re(pj) < . . . increasing to infinity, and, to every p in this sequence, we associate a polynomial Bp of degree np and set
bp(t) = Bp(logt). We then define the a s y m p t o t i c p o l y n o m i a l s a t 0 by
Pq(,) =
~
bA,)e.
Re(p)
We define
re(q) - max deg Bp for Re(p)= Re(q) n(q) = max deg Bp for Re(p) < Re(q), n(q') -- max deg Bp for
Re(p) < Re(q).
We shall use the term s p e c i a l case to describe the instance when n(q) = 0 for all q. The p r i n c i p a l p a r t of O(t) is defined to be
BoO(t)= ~
bp(t)e.
Re(p) <0 Let C (T) be the algebra of polynomials in T p with arbitrary complex powers p e C. Then, with this notation, Pq(t) E C [log t] (t) .
16 T h e function 0 on (0, oo) = R>0 is subject to a s y m p t o t i c ditions:
con-
A S 1. Given a positive n u m b e r C and to > 0, there exists N a n d K > 0 such t h a t
IO(t) - O v(t)l _<
-c'
for t _> to.
A S 2. For every q, we have
O(t) - Pq(t) = o(tRe(q) llogt[ re(q)) for t --~ O, which shall w r i t t e n as
o(t) ~
b,(t)t,. p
A S 3. Given 5 > 0, t h e r e exists an a > 0 a n d a constant C > 0 such t h a t for all N and 0 < t < 5 we have 1 0 ( t ) - QN(t)[ <_ C/t". We shall assume t h r o u g h o u t t h a t the t h e t a series converges absolutely for t > 0. F r o m D I R 1 it follows t h a t the convergence of the t h e t a series is uniform for t > 5 > 0 for every 5. The Laplace-Mellin transform (0, oo) is defined by
of a m e a s u r a b l e function f on
LMf(s,z) = /~176
d---~.
T h e o r e m 1.1. Let 8 satisfy A S 1, A S 2 a n d A S 3. Then L M 8 has a meromorphic continuation for s 6 C and z 6 UL. For each z, the function s ~-+ LMS(s, z) has poles only at the points --(p+ n) with bp • 0 in the asymptotic expansion of 8 at O. A pole at - ( p + n) has order at most n(p') + 1. /n the special
17
case when the asymptotic expansion at 0 has no poles are simple.
log terms,
the
We shall use a systematic notation for the coefficients of the Laurent expansion of Z M e ( s , z ) near s = so. Namely we let Rj(s0;z) be the coefficient of (s - s0)J, so that
LMO(s,z) = E
Rj(s~
- so) j.
The constant term R0(s0; z) is so important that we give it a special notation, namely,
CTs=8oLMO(s,z) = Ro(so;Z). In particular, we define the r e g u l a r i z e d h a r m o n i c s e r i e s R(z) to be the meromorphic function defined by
R(z) = CTs=lLMO(s,z) = R0(1;z). T h e o r e m 1.2. Let O satisfy A S 1, AS 2 and AS 3. a) For every z C UL and s near O, the function LMO(s, z) has a pole at s = 0 of order at most n(O I) + 1, and the function LMS(s, z) has the Laurent expansion L M S ( s , z ) = R-n(~ sn(0,)+l
+...+Ro(O;z)+Rl(O;z)s+...
where, for each j < 0, Rj(0; z) C C[z] is a polynomiaJ of degree < -Re(p0). b) One has the differential equation
OzLM~(s,z) = - L M ~ ( s + 1,z), and hence the relation
-0zR0(0; z) = R(z). The regularized harmonic series is a particular meromorphic function that has simple poles at z = --Ak with residue a k . Next we recall conditions when _R(z) is the logarithmic derivative of a meromorphic function. We define the s p e c t r a l case to be when all ak C Z.
18
T h e o r e m 1.3. In the spectral case, there exists a unique meromorphic function D(z), called the r e g u l a r i z e d p r o d u c t , such that - logD(z) = CT~=oLMS(s,z) = R0(0;z).
We have the relation D ' / D ( z ) = R(z). This regularized product is meromorphic of finite order having zeros at the points z = --Ak with multiplicity ak. To make the situation more explicit, and to compare with results that exist elsewhere in the literature, let us record the following formula. In the spectral case, define 1 (0(s,z) = F(s) L M 0 ( s , z ) If we assume that oo
~(t) = ~
ake -~kt
k----1
satisfies the asymptotic conditions AS 1, AS 2 and AS 3, then we have the equality oO
r
ak
=
(z + k=l
for Re(z) and Re(s) sufficiently large. By Theorem 1.1 and Theorem 1.2, r z) is holomorphic at s = 0 for z E UL and (~(0, z) = CTs=0LMO(s; z) + 7R-1(0; z) = R0(0; z) + 7R-1(0; z). Thus, in the spectral and special case, -R0(0; z) amounts to a normalization of the analytic torsion - ( r ( 0 , z). We have the Lerch formula D ' / D ( z ) = - ( ' ( 0 , z) + a constant. However, even in the most generM case, for many applications one can work just as well with the regularized harmonic series R(z) or
19
R0(0; z) even through ~(s,z) is not holomorphic at s = 0, and D does not exist. For further comments, see w of Chapter V. For any regularized harmonic series R, or regularized product D in the spectral case, we define the r e d u c e d o r d e r to be the pair of integers (M, m) where, in the notation of the asymptotic condition A S 2: M is the largest integer < -Re(p0);
m = m(po) + 1 if there is a complex index p with Re(p0)=Re(p)CZ<0 otherwise simply set m =
and
bp#0,
m(po).
Finally we recall the Gauss formula of [JoL 93a], w For any complex index q with Re(q) > 0, and variables z and w with Re(w) > 0 and Re(w) > - R e ( z ) - Re(Ak)
for all k
we then have the equality
R(z + w) = I~(z; q) + Sw(z; q) where
O0
Iw(z; q) =- J [Oz(t) PqOz(t)] e-Wtdt, -
o with and Sw(z;q)
=
Re(p)-4-k
(-z) k---~. CTs=oBp(c3s) [ F(s ws+p+k.4_ + p + k +1 l)]
20
w Asymptotics in vertical strips. In this section we will d e t e r m i n e t h e a s y m p t o t i c b e h a v i o r of a regularized h a r m o n i c series in a vertical strip. We shall see t h a t t h e a s y m p t o t i c s in a vertical strip are given by t h e t e r m Sw(z; q). We let
An = ~ 2 ( - R e ( ~ ) } . N o t e t h a t An is b o u n d e d above for all n a n d An --~ - o c as n --~ oo. T h e o r e m 2 . 1 . Let xl,x2 C R with Xl < x2. Select n su//iciently large so that An < xl + 1. Then for all q with Re(q) > 0 and uniformly for x E Ix1, x2], we have the asymptotic relation n--1
R ( x + l + i y ) = Sl+iy(X;q)+~-~
ak x + 1 + iy + s k----0
rot
lyl
+o (lYl-[R~(q)])
--* ~ .
Proof. W i t h n chosen as above, let Ln = { ~ n , . . . }. We d e c o m pose R(z) as n--1 ak
R(z) =
z +
+
k=O
w h e r e Rn(z) is t h e regularized h a r m o n i c series f o r m e d f r o m t h e sequence An w i t h coefficients {ak}, k > n. T h e desired e s t i m a t e is obvious for t h e finite s u m , so we m a y a s s u m e t h a t we are w o r k i n g w i t h t h e sequence Ln, a n d we will s u p p r e s s this s u b s c r i p t . We a p p l y t h e general Gauss f o r m u l a by s e t t i n g z = x a n d w = 1 + iy w i t h y C R . It suffices to prove t h e e s t i m a t e
zi+i,(x; q) =
o(lyl -R~(~))
for lYl -~ ~ .
T h e i n t e g e r n has b e e n chosen sufficiently large so t h a t , as a function of t,
O x ( t ) e - t = O ( e - ( ~ l + l - A n ) t)
for x E [Xl, x2] a n d t > 1.
In p a r t i c u l a r , this implies, by A S 1 a n d D I R Ox(t)e - t e
2, t h a t we h a v e
Ll[1, cx)) FI C~176 oo).
21 Since PqO,(t) has polynomial growth in t, PqOz(t)e -t e Lx[1, o~) f) C~[1, cx)). Directly from A S 2, we have [Ox(t) - PqOz(t)] e-* 9 sl[0, 1] f-? ctRe(q)][0, 1], and [0~(t) -
PqO~(t)]e -t = o(tRe(q)llogtlm(q) )
f o r t --* 0.
We now apply the Riemann-Lebesgue lemma to obtain the bound
Zl+iy(X; q) = f [Ox(t) -- PqO~(t)] e - t e - i y t d t 0
o(lyl-tRe(q)l)
=
which holds uniformly for all x E [Xl,X2] as lyl * the proof of the theorem is complete. []
o~ With this,
For many purposes, it suffices simply to know the lead asymptotic of R(z), which is obtained from the bound
n--1 E k=O
ak x + 1 + iy + Ak
= o(1)
for
lyl
-~ ~,
and the following result. C o r o l l a r y 2.2. Let m = m(po ) + 1 if there is an index p with Re(p0) = Re(p) E Z<0 and bp 7s O, otherwise set rn = m(po). Then for any q with Re(q) > 0, we have
Sl-t-iy(Z; q) ~-- 0 ([y[-Re(P~ Since-Re(p0)-1
as
lyl) m)
/'or lYl ~
~.
_< M < -Re(p0), this estimate can be writ.ten
S l + i y ( X ; q ) = O ( l y l M ( l o g l y l ) m)
f o r l y I --+co.
Proof. Immediate from the definition of S~o(z; q) as applied to Sl+iy(x; q). []
22
w
A s y m p t o t i c s in s e c t o r s .
In this section we will determine the asymptotic behavior of the regularized harmonic series R(w) as w ~ cx~ in a sector of the form 71"
Sec, = {w e C l - ~ + ~ < arg(~) <
71" __ ~}
for some e > 0. As in the previous section, the asymptotics are determined by the term Sw(0; q) in the general Gauss formula. Theorem
3.1.
Let
x = Re(w). For
edl q with Re(q)
> 0, we
have
R(w) =
Sw(0;
q) + O(x-Re(q)-l(log x) re(q))
as w --+ oo in Sec, where, as above, Sw(O;q)=
Z CTs=~ rte(p)
[F(swS+p+l + p+ I)]
Proof. We apply the general Gauss formula with z = 0. The proof follows from estimating the integrals q) = j [o(t)- e.o<~)]e-~'at + f EO
•
1
0
By AS
2, we have 1
[0(t)- PqO(t)]e-~tdt
<< j
tRY(q)[log tlm(q)e-~tdt
0 O~
< (os)m(q [~(S -~- 1) xs+l
=
o
+ j tR~(q)(l~ s----Re(q)
1
t)m(q)e-Xtdt
23 in See,. For t _> 1, O(t) - PqO(t) = O(e ct) for some c > 0 and so, for w 6 See, with x sufficiently large, we have
as w
--~ o o
/
OO
[0(t) - PqO(t)] e-Wtdt <
/
K
=
o
1
thus yielding the stated estimate.
[]
As in the case of asymptotics in vertical strips, one often needs to know only the lead asymptotics of R(z). Using Corollary 2.2, we can state this result. C o r o l l a r y 3.2. Let rn = m(po ) + 1 if there is an index p with Re(po) = Re(p) C Z
x) m)
for w ~ oo in See,.
In particular, since -Re(po) - 1 < M < -Re(po), this estimate can be written as R(w)=O(xM(logx)
m)
f o r w - - + ~ inSec,.
It is interesting to note that the asymptotic relation given in Corollary 3.2 is identical to the result given in Corollary 2.2 even though the directions in which w ~ oo are quite different. We can integrate the result in Theorem 3.1 in order to obtain an alternative proof of the generM Stirling's formula for a regularized product, first established in w of [JoL 93a]. For completeness, we shall simply state this result. T h e o r e m 3.3. Let D be a regularized product, so that we have R = D'/D. Let
Bq(w)
"~ Re(p)
[F(Sw~_~+ P)]j 6 C(w) [log w],
and set x = Re(w). Then for M1 q with Re(q) > O, we have
log D ( w ) = Bq(w) + O(x-R~(O(log x) re(O) as w --+ oo in Sec,.
24
w
A s y m p t o t i c s in a s e q u e n c e to t h e left.
In this section we will determine the asymptotic behavior of the regularized harmonic series R(z) for a particular sequence of vertical line segments going to the left so that R(z) grows as slowly as possible. Roughly, these line segments must pass between certMn consecutive pairs of poles of R(z) that are "sufficiently far apart". We shall state and prove results for general regularized harmonic series and for regularized harmonic series which are the logarithmic derivatives of regularized products. The main result of this section is the following theorem. T h e o r e m 4.1. There is a sequence of read numbers Tn --~ oo such that for all y in any compact interval of R the following asymptotic relations hold. a) In the notation of D I R 2, we have
R(-Tn q-iy)=o(Tan1+a~
asTn - - ~ .
b) Let m = m(po ) + 1 if there is a complex index p with Re(p0) = Re(p) C Z<0 and bp # O, otherwise simply set m = m(;o). Then for any q with Re(q) > O, we have R ( - T n q- iy) = O(TnRe(P~
m)
as Tn --+ cxD.
In particular, since -Re(p0) - 1 < M < -Re(p0), we have R ( - T n q- iy) = O(TnM(logTn) m)
as Tn "--+cx~.
The proof of Theorem 4.1 comes from considering the general Gauss formula for R ( u + w) with w=-T-l+iy
and
u E [0,1],
for various choices of T and u in R which are to be made later. For T sufficiently large, let us write
(1)
ak
R(u + I~kl<2T
u+w+Ak
+ S (u; o) +
o),
25 where ~(2T) (t) ---- E ake-~kt'e-Ut I~l>2T
-= Ou(t) --
E
ake-~*t
9 e -ut
I,Xkl<2T
and
,
e-ZOt dt.
= i0
In (1) we have used the relation
PoCJT)(t) = Poe (t), which, in particular, gives the identity
sw(2T) Theorem 4.1 will follow by estimating each term in (1) separately. In a manner similar to that of Corollary 2.2, we can estimate the term Sw(u; 0). Specifically, with the above choice of variables u and w = - T - 1 + iy, we immediately have (2)
S.,(u; 0 ) :
O (T-Re(P~
m)
for
ITI
+
oo.
This bound fits the result stated in Theorem 4.1(a) after noting the relations a0 > 0 and ~rl > -Re(p0). Throughout this section, which is devoted to the proof of Theorem 4.1, we will consider both general regularized harmonic series and those regularized harmonic series which are logarithmic derivatives of regularized products. In this latter case, we will, for this section only, assume that all coefficients ak are equal to -4-1 and count the zeros )~k with multiplicities. In order to estimate the remaining two terms in (1), we need the following bounds.
26
L e m m a 4.2. W i t h the notation as above, the following asymptotic relations hold: a) In the notation of D I R 2, we have
and la~l=o(lAkl~~176
k--o(IAkl ~')
b) In the spectral case, there is a positive constant C' such that k ,,~ C' IAk[-Re(p~ (log [Ak[)m(r~ , or, equivalently, there is a positive constant C such that
l.~kl ~
c (]g/(Xogk)m(P~ \
-I/Re(p~ /
Proof. Define the a b s o l u t e z e t a f u n c t i o n associated to the sequence L by OO
~bs(~) = ~
I~kl-s
k=l
By the inverse Mellin transform (see w of [JoL 93a]), we have, for all a > o-1
oo
a+ioo
k=l
a--ioo
Since R e ( - s + (71) < 0, we can apply the dominated convergence theorem to conclude (3)
lira t ~1 9
t--*0
e -I~klt
= 0.
k=l
Similarly, one shows that
(4)
lim t~0.
t---*0
lakl~ -I~'~l' = 0 . k=l
In the spectral case, we have, by AS 2, the existence of a positive constant c such that (5)
lira t-Re('~
t---*0
-m('~
9E
e-I'~klt
= c.
t:=1
At this point, we apply the following general result.
27
T h e K a r a m a t a T h e o r e m . Let c~ E R>0 and ~ C Z>0. Let # be a positive measure on R + such that the integraJ OO
f0 converges for t > O, and such that
t ~ ( - logt) ~ / e-tX#( A
lim
t---*O
=C
0
for some positive constant C. I f f ( x ) is a continuous function on the interval [0, 1], then
lim
t--*0
[
to'(- logt) ~
/
f(e-t)~)e-~
0
C 0
The proof of the K a r a m a t a theorem in the above form with the factor ( - l o g t) ~ follows the proof given on page 94 of [BGV 92]. We apply the K a r a m a t a theorem as on page 95 of [BGV 92], namely by considering a decreasing sequence of continuous test functions converging to the function x -1 on the interval [l/e, 1] and zero on the interval [0, l/e). In the setup of (a), we use the spectral measure that places unit mass at the points IAkl E l:t +, taking into account multiplicities. From (3), we obtain the bound
(6)
N(A) = #{k:
Since N ( A k ) >_ k,
and we conclude (7)
k = o(l&kl~l),
=
28
as was claimed in the first assertion of part (a). To continue, let us again apply the Karamata theorem, this time with the measure that places a mass of lakl at the points [.~kl E R +. The Karamata theorem then yields the estimate
,ao, = o < 0 ) , [xnl_
~
Combining this estimate with (7), we have lakl =
which completes the proof of (a). In the spectral case, we apply the Karamata theorem to (5) with the family of test functions described above and a=-Re(p0)
and
fl=-rn(p0),
from which we obtain the estimate (8)
N ( A ) - - # { k : IAkl _< A} ~ C'A~(log A) -~
for some positive constant C I. This estimate gives the relation k = X(Ak) ~ C'lAkl~(loglAkl)-~, as asserted in the statement of the lemma. Then k 1/'~ ~ C'l/~[Ak I (log
lakl)
and
logk ~ a . log I kl. Hence, for some positive constant C, we obtain the asymptotic relation IAkl ~ c ' - l / ~
1/~ (log IAkl) ~ / ~ ~ C]~ 1/~ (log k) ~/'~ ,
which completes the proof of the lemma.
[]
Lemma 4.2 allows us to estimate the term I(2T)(u;O) from (1) in the following proposition.
29
P r o p o s i t i o n 4.3. Let w = - T - 1 + iy and u E [0, 1]. For all y in any compact interval of R the following asymptotic relations hold. a) In the notation of D I R 2, we have
I(w2T)(u;O) : o ( T al+a~
logT)
for T --+ oo,
b) L e t m : m ( p 0 ) + l ifthereisa p withRe(po) = Re(p) 6 Z<0 and bp ~ O, otherwise set m : m(po). Then, in the spectral case, we have
I(w2T)(u;O): 0 (T-Re(p~
m)
for T --+ oo.
Proof. For n sufficiently large, one can write n (2T) OuI~; (u;0) = ( - 1 ) " r ( n + 1)
(u + IXkl>2T
ak w
+ Ak)-+l "
Since y lies in a compact interval, we have the estimate
lu+wl ~T, so we have the bound
lakl
O~r(2T)(,, 9O) = 0
)
E (IAkI _ T),~+, ]~kl>2T
i
Using Lemma 4.2(a) and an integral comparison, there exist positive constants Cl and c2 such that
o~nzw(2T)(~; ~ 0) =o
l ~
x ~~ (c2x11~, _ T ) , + , ) d x ) \
x
~ol~l-(~+1)l~l dx I
/ =0
(Tao+al-(n+l)).
30
To finish one integrates n-times. An extra power of l o g T occurs precisely in the case when a0 + al E Z<0. The proof of (b) is similar to that of (a). As above, let a=-Re(p0)
and
/3=-re(p0).
Lemma 4.2(b) and an integral comparison yield the estimate
,, ~(2T)( u ,.O ) OuI
= 0
CJ
1 T ~ (log T ) - #
=0
dx (c2(x(logx)#) 1/'~ x-('~+l)/~(log
_
x)-#(n+l)/adx I
.
1 T '~ (log T ) - #
If we integrate by parts, we obtain ~u.w~nT(2T){~"~, 0) ---- O
(
x-(n+l)/a+l(log
x)-#(r~+l)/adx
4
c l T ~ (log T ) - #
o ((~O(~o~)~) -(~247176247 : o (r~247 ~)-~) -- o (T--o(~o)-(o§ T)~(~0))
_-
)
~(o§
To finish one integrates n-times. An extra power of log T occurs precisely in the case when Re(p0) E Z<0. [] The results stated in (2) and Proposition 4.3 are for T --~ ~ . To estimate the remaining term in (1), namely the finite sum
>~ I~kl<2T
ak
u+w+Ak'
it is necessary to choose carefully a sequence Tn. Roughly speaking, the sequence must be as far from any --~k as possible. The following lemma makes this statement precise.
31
L e m m a 4.4. A s s u m e the n o t a t i o n as above. a) Let rn = ]Anl ~ . T h e r e is a positive constant c and a sequence {n/c} o f positive integers w i t h nk ---* oo such t h a t r n ~ + l -- rn~ > C
lr
all nk.
Equivalently, there is a positive constant c a n d a sequence {n/c} o f positive integers w i t h n/c --+ oc such that
I '-1
b)
foralln/c.
Let rn = I ~ l ~ ( l o g I .1) There is a positive constant c a n d a sequence {nk } o f positive integers w i t h n/c ~ o~ such that r n ~ + l - - rn~ ~ C for all nk. Equivalently, there is a positive constant c and a sequence {n/c} o f positive integers w i t h n/c ~ oo such t h a t -
:/,:,,-i (log n/c) -S~/~
I:,,,, I _> c 9r,/c
forallnk.
Proof. For t h e first s t a t e m e n t in (a), one has, by (6), the b o u n d k = o(r/c), which would be contradicted if no such constant c or subsequence {n/c} would exist. T h e second assertion in (a) and b o t h assertions in (b) are established similarly. []
By a s s u m p t i o n D I R 3, there is a constant C such t h a t for all k, we have the inequalities Re(Ak) < IX/Cl < CRe(A/C). One t h e n has results analogous to L e m m a 4.2 and L e m m a 4.4 for the sequence {Re(),k)}. Specifically, we shall need the following version of L e m m a 4.4. L e m m a 4.5. A s s u m e the n o t a t i o n as above. a) L e t rn = Re(An) ~1 9 T h e r e is a positive constant c arid a sequence {nk} Of positive integers w i t h nk ~ oc such t h a t for all nk, rn~+l
--rn~
~_ C.
32
Equivalently, there is a positive constant c and a sequence {nk } of positive integers with nk ~ oo such that for d l nk,
Re(/~nk+l) -- Re(/~nk)
>_ c . n ~ /'rl - 1
b) Let rn = Re(A,~)~(logRe(kn)) - z . There is a positive constant c and a sequence {nk} of positive integers with nk such that t'or a11 nk, r n k + l -- rnk ~ C.
Equivalently, there is a positive constant c and a sequence {nk} of positive integers with nk --* ee such that for a11 nk, Re(,~n~+l) - Re(,~n~) ~ c" rtl/a-l(log~tk)-~/a.
The proof of Lemma 4.5 is identical to that of Lemma 4.4, hence will be omitted. The following proposition estimates the finite sum in (1) for a particular sequence Tn of real numbers with T,, ---+ ~ so that the finite sum grows as slowly as possible. With the proposition, the proof of Theorem 4.1 is complete.
P r o p o s i t i o n 4.6. There is a sequence of real n u m b e r s T~ with Tn ~ c~ such that for all y in any compact interval os R, the following a s y m p t o t i c relations hold. a) In the notation of D I R 2, we have > ~, "
IAkl~2rn
ak -Tn+iy+Ak
=o(T~n,+'~
b) Let rn = rn(po) + 1. Then, in the spectral case, we have > ~, -"
ak = 0 (TnRe(p~ - T n + iy +,~k
33
Proof. For any z with Re(z) sufficiently large and
Izl
#
I kl,
we
can estimate the finite sum in (a) by ak E z+.~k IXkl<2T,~
(9)
0
(
----o
), _ [ kl<2T.
a
IRe(Ak) + Re(z)l
/
E I kl_<2T.
Let {&-k } be the sequence determined by Lemma 4.5(a), and set 1 Tk = ~ (Re(~n~) + ae(~n~+l)) 9 For any fixed positive number h and any number T, there is a uniform bound on the number of elements of the sequence {rk } in the interval [T, T + h]. Also, there is a positive constant c such that Irk - Tnl > c for all k and n. With these facts and Lemma 4.2, (9) can be bounded by 0
E k=l
Tn
-
k 1/al
-
'
which we can estimate by the integral
x,~O/,,1 o
T. - xl/~1 dx
0
u,,O d[u,,l] = o
T,~--~u
0
= o (Tr~~
log Tn) .
Similarly, for part (b), we can bound the finite sum in the spectral case by the integral 0
d [x a (log x)-#]
= 0 (Tn~-1 (log Tn)-fl+l )
0 = 0 (TnRe(p~176
With this, the proof of Proposition 4.6, and Theorem 4.1, is complete. []
34
w
Asymptotics in a parallel strip.
In this section we will consider a regularized harmonic series R ( z ) corresponding to a sequence {Ak} that converges to infinity in a strip, meaning there exists a constant C such that for all k, we have the inequality IIm(Ak)l < C.
Recall that vidual term Ak. Under bounding z
the sequence {Ak} is such that Re(Ak) ~ oo. An indi1 / ( z + Ak) may become arbitrarily large if z approaches the assumption [Im(Ak)[ < C, we will show that by to a parallel horizontal strip given by 0 < c < [ I m ( z ) - Im(Ak)l < c' < ~ ,
we shall be able to determine the asymptotic behavior of the regularized harmonic series R(z). T h e o r e m 5.1. Assume there is a constant C such that for all k we have [Im(Ak)[ ~_ C. Let z = - T + iy and assume there are constants c, c' > 0 such that for all k, we have 0 < c < lira(z)
-
Im(Ak)[ < c' < oo
a) In the notation of D I R 2, we have R ( - T + iy) = o(T al+a~
log T)
uniformly as T -~ oo.
b) Let rn = m(po) + 1 if there is a complex index p with Re(p0) = Re(p) 6 Z<0 and bp # O, otherwise simply set m = r e ( p 0 ) T h e n for a n y q with Re(q) > 0, we have
R(-T
+ iy) = O(T-I~e(P~
m)
uniformly as T ---, oo. Since - R e ( p 0 ) - 1 _< M < -Re(p0), we h a v e
R(-T
+ iy) = O ( T M ( l o g T ) m)
uniformly as T --+ oo.
35
Proof. The proof follows that of Theorem 5.1. The only aspect of the proof which needs to be reconsidered is the asymptotic behavior of the finite sum E
ak
u+w+Ak" [Ak[_<2T However, we immediately have the estimate
E ]Aki_<2T
ak z+Ak
=0
(~
,a~
\1 A~I_T
~0 \1 ~I_T
c + IRe(Ak) + Re(z)[
c + IIRe(ak)l-
)
IRe(~)ll
)
'
which can be bounded by the sum
c~1 k=l
k~'~
)
c + Ic2kl/*~ -
TI "
As in the proof of Proposition 4.6, this sum can be estimated by the integral cl] ~1 o
o
x,0/~l
IT - xX/~l + cdx
) = o
(Y
iT u ~~ u] + i d [ u ~ , ]
0
: o (T "~
)
logZ).
The proof of the remaining parts of the theorem follows the same pattern. []
36
w
R e g u l a r i z e d p r o d u c t a n d series t y p e .
In w we defined what is meant for a function to be a regularized product. In this section, we define functions which are of regularized product type. After this, we will summarize the asymptotic formulae of the previous sections for this new class of functions. D e f i n i t i o n 6.1. A meromorphic function of finite order is said to be of r e g u l a r i z e d p r o d u c t t y p e if
G(z)
G(z) : Q(z)e p(~) I I Dj(oljz + flj)kj j=l where: i) Q(z) is a rational function; ii) P(z) is a polynomial; iii) Dj(z) is a regularized product, with ~ j , ~ j 9 C and kj 9 Z; iv) the c~j, /3j are restricted so that the zeros and poles of Dj(&jz + flj) lie in the the union of regions of the form: 9 a sector opening to the right, meaning {zCC:--
7r
2 +e<arg(z)
7r
< ~-e}
for s o m e e > 0 ,
9 a sector opening to the left, meaning 7r {z 9 C: ~+e<arg(z)
37r <-~-e}
for some e > 0 ,
9 a vertical strip, meaning {z C C : a < Re(z) < b} for some a, b E R.
If G is of regularized product type, then any vertical strip which contains at most a finite number of zeros of G is called an a d m i s sible strip. Observe that if the zeros and poles of the function G are in sectors, as described in the first two conditions of Definition 6.1(iv) above, then every vertical strip is admissible. The main results of [JoL 93c] will be assumed (see in particular Theorem 1.5), and will not be repeated in these notes. Many examples of functions of regularized product type are given in w of [JoL 93c].
37
Corresponding to the multiplicative group of regularized product types, we have the additive group of their logarithmic derivatives, namely n
a'la( ) = aoD'o/Do(aoz +/30) + ~ kjahD}/Dj(ajz +/35)" j=l
This leads us to define the additive notion for its own sake. A function R will be said to be of r e g u l a r i z e d h a r m o n i c series t y p e if it is a finite linear combination R(z) = F_, chR,(
jz +/35) + P ' ( z ) +
c~,, z - k
where cj, c~, aj,/3j,/3~ E (3, each Rj is a regularized harmonic series as defined in w P ' is a polynomial, and aJ and/3j are subject to condition (iv) with respect to the poles of Rj(ajz +/3j). Sections w to w provide systematic estimates for functions of regularized harmonic series type in vertical and horizontal strips, not just for those which happen to be logarithmic derivatives of regularized product types. Indeed, a transformation z ~-+ az +/3 amounts to a dilation, rotation and translation, so the estimates of w and w apply to each term in the above linear combination for a function of regularized harmonic series type. The resulting estimates will be summarized in Theorem 6.2 below. Let Do(z) = Q(z)e P(*) and define the r e d u c e d o r d e r of Do to be (M0,0) where M0 = max{0, d e g P - 1}. Suppose G is of regularized product type of the form n
G(z) = Do(aoZ +/30) I I Dj(o~jz +/3j)kj. j=l
Let (Mj,mj) be the reduced order of D j, as defined at the end of Chapter I, w then we define the r e d u c e d o r d e r of G to be the pair of integers (M, m) where M = max{Mj } and m is the maximum over all raj for which M = Mj. We make a similar definition for the reduced order (M, m) of a regularized harmonic series type. In this case, the reduced order of P ' is (deg P', O) if P' # O.
38
We shall use for a vertical strip the notation Str(xl,x2) : {z e C : X 1 _ ( R e ( z )
_< x2}.
The asymptotic formulas of the previous sections can now be easily summarized for any function R of regularized harmonic series type of reduced order (M, m). T h e o r e m 6.2. Let R be of regularized harmonic series type of reduced order (M, m). a) Let S = S t r ( z l , x 2 ) be an admissible strip. Then uniformly for x E [Xl, X2], we have the asymptotic relation
n(x q- iy) = o(yM(logy) m+l)
as [y[--+ oo.
b) Let S = Str(xa, x2) be a vertical strip which contains an int~nite number of poles of the function R. Then there is a sequence of real numbers Tn ~ oo such that for all x C [Xl, x2], we have the following uniform asymptotic relation n ( x = k i T n ) = O(TM(logTn) re+l)
as Tn --+ cx~.
Proof. Assertion (a) follows directly from the definition of regularized harmonic series type, admissible strip, and Corollary 2.2, Corollary 3.2, and Theorem 5.1. Assertion (b) follows from the same results together with Theorem 4.1. Note that in order to obtain the symmetry as asserted, one needs to consider the union of the sequences in the regularized harmonic series decomposition of G, which a possible change of sign. []
39
w
S o m e examples.
To conclude this chapter, we shall briefly discuss how the results of the previous sections contain various classical asymptotic formula for the gamma function and the Riemann zeta function, and improve an existing bound for the Selberg zeta function associated to a compact hyperbolic Riemann surface.
Example 1: T h e g a m m a f u n c t i o n . The classical gamma function F(s) is, up to a factor of the form e a'+b, the regularized product associated to the sequence Z>0. In the notation used above, we have -Re(p0) = 1 and re(p0) = 0, hence the gamma function is of regularized product type of reduced order (0, 0). Theorem 2.3 yields the classical Stirling's formula for log F(s). As for asymptotics in vertical strips, Corollary 2.2 yields the equally classical asymptotic formula
r'/r(s)
= log Isl + o(1)
as s ~ c~ in any vertical strip of finite width. This result is an important ingredient in the proof of the classical explicit formula for zeta functions of number fields (see, for example, Chapter XVII of [La 70]).
Example 2: Dirichlet polynomials. We define a Dirichlet polynomial to be a holomorphic function of the form N P ( s ) = ~-'~anb~ n=l
where {an } is a finite sequence of complex numbers, {bn } is a finite sequence of positive real numbers. In Chapter II we will apply the general Cram@r theorem from [JoL 93c] to show that P is of regularized product type of reduced order (0, 1).
Example 3: The Riemann zeta f u n c t i o n . Let CQ(S) be the Riemann zeta function and consider the sequences A§ = {p/i E C l 6~(p) = O, 0 ~ Re(p) ~ I,
Im(p) > O}
and
A_ ={p/(-i) EC ] CQ(P)=O, O
40
By Corollary 1.3 of [JoL 93c], the theta function 0A+ associated to the sequence A+ satisfies the asymptotic conditions AS 1, AS 2, and AS 3 with -Re(p0) = 1 and m(po) = 1. Similarly, the theta function 0A_ associated to the sequence A_ satisfies the asymptotic conditions A S 1, AS 2, and A S 3, again with -Re(p0) = 1 and m(po) = 1. Theorem 1.6 of [JoL 93a] then implies that the functions D+(z) = exp ( - C T s : 0 L M O A + (s,z)) and
D_(z)
= exp (-CTs=0LMOA_ (s, z))
are holomorphic functions of finite order with
and
D+(z)=O
if and only i f - z E A + ,
D_(z)=O
if and only i f - z E A _ .
The argument of w of [JoL 93c] produces the relation
(ct(s)F(s/2) = D+(-s/i)D_(s/i)(s(s-
[))-leas+b
for some constants a and b. Therefore, the Riemann zeta function is of regularized product type of reduced order (0, 1). Example 3 and Theorem 5.1 combine to assert the existence of a sequence of numbers {Tn} with Tn --~ cx) such that
• To) = o ((logTo) :) for x E [Xl, x2] and T~ --* oc. This result is proved classically in a different manner, see, for instance [In 32], pages 71-73.
E x a m p l e 4: T h e Selberg zeta function Z(s) associated to a c o m p a c t R i e m a n n surface. The results of [JoL 93c] show that the Selberg zeta function Z(s) associated to any compact hyperbolic Riemann surface is an entire function of regularized product type and reduced order (1, 1), since, in this case, -Re(p0) = 2 and m(po) = 1. Theorem 5.1 asserts the existence of a sequence of numbers {Tn} with Tn --+ oe such that
IZ'/Z(x =l=iTn)l =
O (Tn(log T~) 2)
41 for x E [Xl, X2] and T,, ---+ oo. This improves a result stated on page 80 of [He 76], giving an upper bound of the form O ( T 2 ) . As we will see in subsequent chapters, this improvement is significant since it allows us to apply the explicit formula for the Selberg zeta function to any test function that satisfies the three basic conditions to order one, which essentially amounts to requiring that the test function and its first derivative satisfies certain smoothness conditions. With the weaker bound from page 80 of [He 76], one would be forced to require smoothness conditions on a test function and its first two derivatives. E x a m p l e 5: T h e S e l b e r g z e t a f u n c t i o n Z(s) a s s o c i a t e d to a n o n - c o m p a c t R i e m a n n s u r f a c e . One can apply the results of this section to the Selberg zeta function associated to any noncompact hyperbolic Riemann surface of finite volume X since, by Theorem 7.1 of [JoL 93c], this Selberg zeta function is of regularized product type of reduced order (1, 1). Associated to X there is a meromorphic function called the scattering determinant, which is the determinant of the constant terms in the Fourier expansions in the cusps of the Eisenstein series (see the discussion starting on page 498 of [He 83] or page 49 of [Sel 56]). Theorem 7.1 of [JoL 93c] shows that the scattering determinant is of regularized product type of reduced order (0, 1). In [JoL 94] we will give a more complete discussion of the application of our results to scattering determinants and to Eisenstein series on non-compact hyperbolic Riemann surfaces of finite volume. A larger but non-exhaustive list of further examples is given at the end of [JoL 93c], including zeta functions arising in representation theory and the theory of modular forms, and zeta functions associated with certain higher dimensional manifolds. We shall return to these specific applications in a subsequent publication devoted exclusively to them.
CHAPTER
II
Cram~r's theorem as an explicit formula. In [Cr 19] Cram6r showed that if {Pk} ranges over the non-trivial zeros of the Riemann zeta function with Im(pk) > 0, then the series
V(z) : E ~ converges for Im(z) > 0 and has a singularity at the origin of the type log z/(1 - e-z), by which we mean that the function
F(z) = 2 . i V ( z )
log z 1_ r
which is defined for Ira(z) > 0, has a meromorphic continuation to where n ranges over all C, with simple poles at the points • the integers, and at the points :t: logp m, where pm ranges over the prime powers. In [JoL 93c] we proved an analogous theorem for any meromorphic function with an Euler sum and functional equation whose fudge factors which are of regularized product type. In this chapter we will prove a Cram6r-type theorem by considering the same contour integral as analyzed in [JoL 93c] for a more general class of test functions. Specifically, we consider the contour integral
1f r
2rci
over a semi-infinite vertical rectangle Ts which is assumed to contain the top half of the "critical strip" of Z. The test function r is assumed to be holomorphic on the closure of Ts and to have reasonably weak growth conditions, essentially what is needed to make the proof go through. In the Cram6r theorem from [JoL 93c], the test function r depended on a complex parameter z, namely
r
= r
= ~sz
44 For Im(z) > 0, the function $z has exponential decay when s lies in a finite strip and Ira(s) ~ oo. As a result, the analysis needed in [JoL 93c] required a very weak growth result, which we proved for general meromorphic functions of prescribed order (see, in particular, Lemma 2.1 of [JoL 93c]). In w we recall the fundamental class of functions which we have defined, and discuss its relation to the Selberg class of functions defined in [Sel 91]. The definitions of w are used throughout, but the subsequent sections may be logically omitted for the rest of this work. Taking into account the asymptotic estimates of Chapter I, our method of proof from [JoL 93c] applies to the more general class of test functions considered here. In w we establish notation and state the main result of this chapter. The proof is given in w and various applications are discussed in w As remarked on page 397 of [JoL 93c], our proof does differ from the original proof due to Cram6r, which is one of the reasons why we can easily generalize the theorems to the class of functions which have Euler sums and functional equations with fudge factors which are of regularized product type.
45 w
Euler sums and functional equations.
We shall say that the functions Z and ,~ have an E u l e r s u m and f u n c t i o n a l e q u a t i o n if the following properties are satisfied: 1. M e r o m o r p h y . The functions Z and Z are meromorphic functions of finite order. . E u l e r S u m . There are sequences {q} and {(t} of real numbers > I that depend on Z and Z, respectively, and that converge to infinity, such that for every q and ~, there exist complex numbers c(q) and C(~l) and o~ > 0 such that for all Re(s) > cry,
log z ( s ) = ~
c(q)
and
log 2 ( s ) = ~
q
c(O)
0
The series are assumed to converge uniformly and absolutely in any half-plane of the form Re(s) > 0~ + ~ > a~. 3. F u n c t i o n a l E q u a t i o n . There are functions G and G, meromorphic and of finite order, and there exists 00 with 0 <_ Oo _< 0~ such that
z(s)G(s) = 2(00 - s)~(o0 - s). We let r
= c(s)/d(~0
- s),
so the functional equation can be written in the form
z(s)r
= 2(~0 - s).
We call G, G or q~ the f u d g e f a c t o r s of the functional equation. R e m a r k 1. We inadvertently took a0 = o~ in [JoL 93c], but it is important to allow a0 ~ o~ for some applications, e.g. the scattering determinants which we considered in w of [JoL 93c]. No change is needed in the proof of the Cram@r theorem except for ! choosing a > 0 such that 00 + a > a 0 as in w below. We dealt with
46 the scattering determinant in [JoL 93c], in the context of Cram6r's theorem. The Euler sum for Z implies that Z is uniformly bounded for Re(s) _> a~ + e for every e > 0. Notice that Z has no zeros or poles for Re(s) > a~, and all zeros and poles of Z in the region Re(s) < - a agree in location and order with poles and zeros of as. R e m a r k 2. A Dirichlet series expression is assumed only for Z and Z, so the fudge factors do not occur symetrically for the zeta function in the above conditions, although they might appear to do so in the functional equation. For example, in the most classical case of the Riemann zeta function, the fudge factor is essentially the gamma function, which does not have a Dirichlet series expansion but is of regularized product type. We define a triple (Z, 2, ~) to be in t h e f u n d a m e n t a l class if Z, 2 have an Euler sum and functional equation, and the fudge factors are of regularized product type. Selberg has defined a "Selberg class" of functions in analytic number theory (see [Se 91], [CoG 93]). Our class is much wider than Selberg's class in several major respects: 1) Selberg's fudge factors are of gamma type, i.e. ['(as +/~). 2) Selberg assumes a Ramanujan-Petersson estimate on the coefficients of the Euler sum, but we do not. 3) Selberg's Euler sum involves ordinary integers and ordinary prime powers. We allow arbitrary positive numbers {q}. Our conditions allow for a much wider domain of applicability in spectral theory as illustrated by our varied examples, and many more to be treated in subsequent papers. For example, our conditions allow the fudge factors to include F2, Fd for general d, or for and L functions themselves, or any function of regularized product type. R e m a r k 3. Even so, the Euler sum condition is still not sufficiently general for our purposes and will be ultimately be generalized to a Bessel sum condition. For further comments on this point of view, see w and w of Chapter V. In the same vein, the functional equation will also be replaced by an additive relation where the additive fudge factor will be assumed to be a regularized harmonic series, or regularized harmonic series type, in analogy with regularized product type.
47
w
The general Cram6r formula.
Let a > 0 be such that a0 + a > a~. We define the following regions in the complex plane: ~ + = semi-infinite open rectangle bounded by the lines Re(s) = - a ,
Re(s) = cr0 + a,
Im(s) = 0.
7E+(T) = the portion of g + below the line Im(s) = T. W e allow Z and 9 to have zeros or poles on the finite real s e g m e n t
[ - a , a0 + a], but we assume that 9 and Z have no zeros or poles on the vertical edges with Re(s) = - a and Re(s) = cro + a and Im(s) > O. Let r be any function which is holomorphic on the closure of the semi-infinite rectangle 7~+, and let H be a meromorphic function on this closure. We axe interested in studying the (formal) sums (1)
div+, a ( r
E
VH(Z)r
zETr +
where VH(Z) = ordH(z)
is the order of the zero or pole of H at z, so the sum (1) is actually over the divisor of H which lies in Tea+. Such sums do not converge a priori, so we need to define them as limits in a suitable sense, and for suitable functions r On a space of functions decaying sufficiently fast (depending on H), the divisor div+, ~ = E
VH(Z)(Z)
gives rise to the functional defined by the sum (1). The functional itself may be denoted by [div+,z] to distinguish the functional from the divisor. To determine such a space of functions, we proceed as follows. Let H i , . . . , Hr be a finite number of functions which are meromorphic on the closure of 7r + , and let r be a function which is holomorphic on this closure. We say that a sequence {Tin} of positive
48 real n u m b e r s t e n d i n g to infinity is J - a d m i s s i b l e for {H1,. 9 9 Hr} if for any k, Hk has no zero or pole on the segment
Sm = [-a + iTm,ao + a + iTm] and
qS(s)H~/Hk(s) --+ 0
for s C Srn with rn --+ oo.
W h e n { H 1 , . . . , Hr} is the set of functions {Z, z~, ~}, we say simply that {Tin } is a d m i s s i b l e . W i t h respect to a ~b-admissible sequence {Tin}, we define the divisor f u n c t i o n a l for H, which we denote by to be the limit
(2)
div+, (4)
= li~moo
VH(Z)C~(z),
E
zen+(T.) if such a limit exists. In particular, for H = Z, we let {p} = set of zeros and poles of Z in T~+ so the sum (2) can be written as
div+,a(q~)=~]i~rnoo E
vz(p)~(p).
We define other functionals as follows. Here we do not try to give subtle conditions on what a m o u n t s to a half-Fourier transform, so we simply assume t h a t the derivative $'(s) is in L 1 of each vertical half line [ - a , - a + ioo] and [~r0 + a, ~r0 + a + ic~]. We define the positive C r a m 6 r f u n c t i o n a l for q, which we denote by Cr+,a($), to be t h e integral ~o+a+ioo
Cr+,~(~) =
/
q~'(s)q-'ds.
ao+a
Similarly, the n e g a t i v e C r a m ~ r f u n c t i o n a l for q is defined by the integral ~oq-a =
/
~o+a-ioo
~'(~o - s)q-Sds.
49
Also, with respect to an H-admissible sequence {Tin}, where H is a meromorphic function which is holomorphic on Re(s) = - a , we define the functional -a+iTm
f r
U+,_a(r = Jim
--a
When all the above functionals are defined, we then consider The Cram~r formula. 2~ridiv+,=(r
: Z a0-ba
+ U;,_a(r
+
+ r
+
I
r
--a
(log Z(o'0 + a) - log Z ( - a ) ) .
Such a formula is derived formally by considering the contour integral
1 i r 2~ri o7r which can be evaluated in one way by using the residue theorem, and in another way by using the Euler sum and functional equation for Z. We are interested in conditions on r for which the Cram6r formula holds. For this purpose, we consider the following growth assumptions on r G R 1. r 1 6 2 1 6 2
is in L' on the vertical ray [ - a , - a + ic~].
G R 2. The derivative r [--a, --a + icx~] and
is in L 1 of each vertical half line
[ao + a, ao + a + icxD].
The first condition compares the decay of r with the exponential growth of r As in [3oL 93c], one often considers the situation
50 when 9 is of regularized product type. In this case, the above growth conditions can be verified as follows. P r o p o s i t i o n 2.1. a) If 9 is of regularized product type of reduced order (M, m) and the ray [ - a , - a + i o o ] lies in an admissiblestrip (defined in Chapter I, w then
O'/O(s) = O(blM(log Isl) m) for Isl ~ ~ and s on the ray I--a,--a + i ~ ] .
b) If for some 6 > O, r has the decay
r
= O(1/IslM+~(loglsl) m+l+a) for Isl--+ ~ on the ray, then G R 1 is satisfied.
The hypotheses in the above criterion have been shown to be satisfied in several cases which are of direct interest in our theory of regularized products, for instance Theorem 5.2 of Chapter I. The next theorem asserts that the divisor functional [div+,,] is defined on the vector space of functions satisfying G R 1 and G R 2, and satisfies the Cram6r formula. T h e o r e m 2.2. Assume r and 9 satisfy the two growth conditions G R 1 and G R 2. Then a11 the functionMs div+,a(r Ug,_a(r , Crq+,a(r and Crq, a(r &re defined, and the following formula holds: 27ridiv+,a(r = E
c(q)Cr~, a(r --E
c(q)Cr+, a(r
q
+ u+,_o(r
+ f
r
z(~)ds
--a
+ r
(log Z(cr0 + a) - log Z ( - a ) ) .
As stated above, the proof of Theorem 2.2 will be given in the following section, and various applications of the theorem will be discussed in w
51
w
P r o o f of t h e C r a m 4 r t h e o r e m .
The pattern of proof of Theorem 2.2 follows w of [JoL 93c], which, as we shall remark below, contains one significant technical improvement over the proof of the original theorem given by Cram@r [Cr 19] for the Riemann zeta function. Choose an e > 0 sufficiently small so that Z has no zeros or poles in the open rectangle with vertices --a,
--a + ie,
ao + a + ie,
ao + a
or on the line segment [ - a + i e , ao + a + i e ] . Note that the function Z may have zeros or poles on the horizontal line segment I - a , a0 + a]. For T sufficiently large, we shall study the contour integral --a+ie
27riVz( z, e; T ) =
o'o+a+ie
f
+ f
--a+iT
--a-l-ie
o'o+a+iT
(1)
-a+iT
+ f
f
o'o+a+i~
o'o+a+iT
We may assume that Z has no zeros or poles on the line segment connecting the points - a + i T and Oo + a + i T , because we will pick T = T m for m sufficiently large. Let: 7~T(e) = the finite rectangle with vertices - a + iT,
- a + ie,
ao + a + ie,
ao + a + iT.
By the residue theorem, we have
2.iyz(z,
;T)=
v(p)r peaT(e)
Theorem 2.2 will be established by studying each of the four integrals in (1). For simplicity, let us call these integrals the left, bottom, right, and top integrals, respectively. We begin with the top integral which will be shown to be arbitrarily small upon letting T = Tm approach infinity.
52
L e m m a 3.1. Let {Tm} be an admissible sequence relative to Z. T h e n we have -a+iT, n
i
lim m---~ oo
r
=0.
"o+a+iTm
The proof of Lemma 3.1 follows directly from the definition of an admissible sequence. To continue, we have, from the growth assumption G R 1, the limit --a+ie
lim
m-...+ ~
--a+ie
r
i
r
= f --a+ioo
-a+iTm
and ao+a+iTm
ao+a+i~
ao+a+ie
ao+a+ie
By combining these equations with Lemma 3.1, we have the following preliminary result. P r o p o s i t i o n 3.2. W i t h notation as above, we have 2~riVz(z,e)=
lim 2rciVz(z,e;Tm)
m --+r
--a-J-ie
= f
o'o+a+ie
+ f
r
--a+ioo
r
--a+ie
ao+a+ioo
+ f
r
ao+a-J-ie
As before, let us call the integrals in Proposition 3.2 the left, bottom and right integrals, respectively. By the above stated assumption on e, we have
Y~(z, ~) = Y~(z).
53
To continue our proof of T h e o r e m 2.2, we will c o m p u t e t h e three integrals in Proposition 3.2 using the axioms of Euler sum and functional equation. After these computations, we will let e approach 0, which will complete t h e proof. Let us use the functional equation to re-write the left integral as the s u m of three integrals involving Z and (I). Specifically, we have -a+ie
--a+ie
f r
f
--a+ioo
,.
--a+ic~ --a+ioo
(2)
= f
r
-a+ie aoTa--ie
+ f r
(3)
ao+a--ioo
After we let e --+ 0, the integral in (2) appears in the s t a t e m e n t of T h e o r e m 2.2 as the functional U+(r Note t h a t letting e --~ 0 is justified since (I) was assumed to be holomorphic and non-zero on the vertical lines of integration. As for (3), we can re-write this integral using the Euler s u m of Z, yielding o'oTa--ie
f r ao+a-ioo ao+a--ie
--~(a~ - s)l~ z~(s);i:;-;i -J- i
r
(ro+a--icx)
---- r
+
it)log 2(ao +
a-
ie) ao+a--ie
+~c(~) f (1 aoTa--ioo By
the Euler sum
condition
and
r
- s)q-'ds.
the fact that a > 0, we can let r
54 approach zero to get the equality O'o+a
f r
ao+a--ioo
(4)
= r
log Z(a0 + a) + E
c(q)Cr~, a(r
Both terms in (4) appear in the statement of Theorem 2.2. In the same manner as above, the right integral can be re-written using the Euler sum of Z, yielding ao+a+ioo
f
r
ao+a+ie o'o+a+ioo
=r
log z(~) ~0+o+-~~ -
/
r
Z(~)ds
ao+a+ie
= -r
+ a + ie) log Z(ao + a + ie) ao+a+ioo
--E
c(q) q
r
/ o.o+a+ie
Again, we can let e approach zero to obtain the equality o'o+a+ioo
r O" O -JF a
(5)
= -r
+ a) log Z(a0 + a ) - E
c(q)Crq+, a(r
q The second term in (5) appears in Theorem 2.2. The first term in (5) does not appear in Theorem 2.2 because this term cancels with a term that appears in the evaluation of the bottom integral, as we shall now see. In the evaluation of the bottom integral, we see the importance of choosing e > 0 before integrating by parts. By the choice of e,
55 Z has no zeros or poles on the line segment [ - a + ie, ~0 + a + ie], so we have tro+a+ie
r --a+ie
(6)
= r
log
Z ( 8 ) ao+a+ie --a+ie
-
ao+a+ie /
r
log Z(s)ds.
--a+ie
Now let e ~ 0 to get the equality O'o--ba
f
--a
= r
+
a)logZ(ao -4-a) -r
~to-~a
(7)
- / r --a
To complete the proof of Theorem 2.2, simply combine equations (2) through (7). Note the cancellation of one term in (5) with a term in (7). R e m a r k 1. The value of log Z(-a) is obtained by the analytic continuation of the Euler sum of Z along the horizontal line segment [~0 + a + i e , - a + ie], followed by the continuation along the vertical line segment I - a , - a + ie], which is equivalent to the analytic continuation along the top of the horizontal line segment I - a , ~0 + a] for small e. To be precise, one should write the integral in (7) as
f
--a ao-ba
(s)
=f
--a
ao-}-a
f
--a
56 R e m a r k 2. In the case that Z(s) is real on the real axis, t h e n arg(Z(s)) is a step function on [-a, ao + a] and takes on values in Z 9~ri, except at the zeros and poles of Z, where t h e a r g u m e n t is u n d e t e r m i n e d . In this case the integral with arg Z(s) in (8) can be evaluated directly and trivially, as an elementary integral.
57
w
An inductive theorem.
When comparing our work with that of Cram6r in [Cr 19], the reader should note that we have overcome a point of substantiM technical difficulty that Cram& encountered when proving Theorem 2.2 for the Riemann zeta function ~Q(S). By choosing a suitably, we have avoided having to consider the convergence of the Euler sum of Z on the line Re(s) = cr0' . Cram& used the fact that ~q(s) does not vanish on the vertical line Re(s) = 1 as well as specific knowledge about the distribution of prime numbers, namely
1 ~ < i - = p O(loglogx) as x--+oo
and the Landau theorem which states that the limit 1
li~Ino~E pl +it p<_x converges uniformly for t in compact subsets of R \ {0}. By following Cram~r's original proof exactly, we would have greatly increased the complexity of the axioms of meromorphy, Euler sum, and functional equation. The simplest example is that of the original Cram& theorem, for the Riemann zeta function ~Q. The g a m m a function obviously satisfies the growth conditions v i s a vis the test function r
sz
Im(z) > 0 .
In fact, for the zeta functions coming from modular forms, with gamma factors as fudge factors, the same remark applies. When applying Theorem 2.2 to the test function Cz(s) = e sz, we can then set z = it, and we determined the asymptotic behavior of e ipt as t approaches zero in complete detail in [JoL 93c] under the assumption that r is of regularized product type. The result of these calculations is the following theorem.
T h e o r e m 4.1. Let (Z, Z, g2) be in the fundamental class, and assume that ~2 has reduced order (M, m). Then Z and 2 are
58
of regularized product type and of reduced order (M, m + 1) if -Re(p0) ~ Z, otherwise the reduced order is (M,m). E x a m p l e s . In w of [JoL 93c] we applied Theorem 4.1 to give many examples of zeta functions which are thus sknown to be of regularized product type. We add to this list any D i r i c h l e t polyn o m i a l , which we define to be any holomorphic function of the form N
P(s)
=
Z anbSn n=l
where {a . } is a finite sequence of complex numbers, { bn } is a finite sequence of positive real numbers, which we may assume, without loss of generality, to satisfy the inequalities
0 < bl < b2 < ... < bN. Let
N
Q(s) = P(-s) -- Z anb'~*' n=l and write
P(s)=aNbSN [1 + ~ 1 aN a---2-n ~N
= aNbSN 9Z(s)
Q(s):alblS [1+~-:-'an (bn) ~1
= albl s.
and Z(s).
It is immediate that there exists some a~ > 0 such that for all s with Re(s) > a~ we have
n=l
an \ bn ]
----hi
~1
<
1.
Therefore, Z and Z have Euler sums, which means there exist sequences {q}, {~l} and {c(q)}, {C((l)} such that for Re(s) > a~ we have logZ(s) = Z q
c(__q)q8 and
log..7(s) = Z c(ocl~) el
59
If we set
V(s) = aNbSN a n d
G ( s ) = albl s,
t h e n the trivial relation P(s) = Q ( - s ) can be written as
G(s)Z(s) =
d(-s)2(-s),
so we also have a functional equation with a0 = 0. Notice that the functional equation implies that all the zeros of P lie in some vertical strip. Further, we can apply T h e o r e m 4.1 to conclude that the Dirichlet polynomial P is of regularized product type, with reduced order (0, 1). As a result, the estimates from Chapter I, specifically T h e o r e m 6.2, hold for any Dirichlet polynomial. Finally, observe that the local factors of the more classical zeta functions are Dirichlet polynomials. Indeed, such factors are of the form Polp(p - s ) + l where Polp is a polynomial with constant t e r m 1, and p is a prime number. For the R i e m a n n zeta function, this local factor is simply Polp(T) = 1 - T, so we have Polp(p - s ) = 1 - p-~. In the representation theory of GL(n), the polynomial Polp has degree n. For representations in GL(2) associated to an elliptic curve, say, we have Polp(T) = 1 - apT + pT 2, so in terms of p - S the local factor is
1 - app -s -4- pp-2S
Thus the local factors of classical zeta functions are themselves of regularized product type.
C H A P T E R III Explicit formulas under Fourier A s s u m p t i o n s The classical "explicit formulas" of analytic number theory show that the sum of a certain function taken over the prime powers is equal to the sum of the Mellin transform taken over the zeros of the zeta function. Historically, only very special functions were used until Weft pointed out that the formulas could be proved for a much wider class of test functions (see [We 52]). We shall give here a version of these explicit formulas applicable to a wide class of test functions in connection with general zeta functions which have an Euler sum and functional equation whose fudge factor is of regularized product type. As a result, our general theorem contains the known explicit formulas for zeta functions of number fields and Selberg type zeta functions as well as new examples of explicit formulas such as that corresponding to the scattering determinant and Eisenstein series associated to any non-compact finite volume hyperbolic Riemann surface. Various facts from analysis which we shall use in this chapter have been proved in our papers [JoL 93a] and [JoL 93b], as well as Chapter I. As a result, most of the steps taken here are relatively formal. We carry out the steps by integrating over a rectangle in the classical manner, but one aspect of this classical procedure emerges more clearly than in the case of classical zeta functions, namely the inductive procedure arising from a functional equation of the type =
2(
0 -
with zeta functions Z and z~ and fudge factor (I) which is of regularized product type. For instance, for the Selberg zeta function of compact Riemann surfaces, these factors involve the Barnes double g a m m a function, and for the non-compact case, these fudge factors may involve the Riemann zeta function itself at the very least. Ultimately, arbitrarily complicated regularized products will occur as fudge factors in such a functional equation.
62
w
G r o w t h conditions on Fourier t r a n s f o r m s .
We shall consider growth conditions on Fourier transforms and logarithmic derivatives of regularized products, and we begin by estimating Fourier transforms. Following Barner [Ba 81], we require the test functions g to satisfy the following two basic Fourier conditions. F O U 1. g e BV(R) A LI(R). F O U 2. g is n o r m a l i z e d , meaning 1
g(x) = ~ (g(x+) + g(x-))
for all x C R.
These will be the only relevant conditions in this section, but in the next section to apply a Parseval formula, we shall consider a third condition at the origin, namely: F O U 3. There exists e > 0 such that g ( x ) = g(O) + O(Ixl ~)
for x ~
0.
If we let N be any integer > 0, then we say that g satisfies the basic F o u r i e r c o n d i t i o n s t o o r d e r N if g is N times differentiable and its first N derivatives satisfy the above three basic conditions. L e m m a 1.1. Assume g satisfies F O U 1 and F O U 2 to order
M. Then
gn(t) -= O(1/Itl M+I)
for ltl ~ c~.
Proof. We integrate by parts M times to give g^(t)
-
-~
g(M)(x)e-itZdx. --
00
To finish, note that for any h C B V ( R ) A L I ( R ) , we have the Stieltjes integration by parts formula O0
hA(t) _ i
1
1 / --
00
e_it~dh(x) '
63
from which we o b t a i n the e s t i m a t e 1
R(h), Ih^(t)l <- -v- Vltl whence the proposition follows.
[]
Let f be a m e a s u r a b l e function on R + so t h a t u n d e r certain convergence conditions we have the M e l l i n t r a n s f o r m (DO
0
Let ~o E R . We define M~o/2 f to be the t r a n s l a t e of M f by ao/2, meaning
M ~ o / J ( s ) = M f(s - ao/2). We put
F(x) = f(e -~) so T h e n letting s = a + it, we find
f(u) = F ( - l o g u ) .
O0
M,o/2f(s)
= / F(x) --00 O0
= /
F(x) e-(~-a~
--(X)
which is v ~
times the Fourier t r a n s f o r m of
F<,(x) = F(x)e -(<'-~~ T h a t is,
M~o/2f(a + it) = v / ~ F ~ ( t ) . In particular, if we let a = ao/2, t h e n we obtain:
On the line Re(s) = ao/2, the Mellin transform is a constant multiple of the Fourier transform, namely M~o/2f(ao/2 + it) = v / ~ F ^ ( t ) .
64 L e m m a 1.2. Let F be a function on R and assume there is an e > 0 such that F( x )e( ~o/2-~l +~) lxl
Assume F satist~es F O U 1 and F O U 2 to order M a n d delqne the function f on R + by f ( u ) = F ( - logu). Let 0-1,02 E R be fixed T e a / n u m b e r s with 0-1 • 0-2 and consider the strip consisting of all s C C with Re(s) C [0-1,0-2]. Then for any s in the strip Re(s) C [0-1,0-2], we have M,0/2f(0- + it) = v / ~ F ~ ( t ) .
and M z o / 2 f ( s ) = O(1/Isl M+')
for Isl T h e proof of L e m m a 1.2 follows that of L e m m a 1.1, hence will be omitted. As in previous chapters, we consider a m e r o m o r p h i c function H, and we want to compare the rate of growth of the logarithmic derivative H ' / H and M~o/2 f on vertical lines. Specifically, we consider the following growth condition. G R . There exists a sequence {Tin} tending to eo such t h a t
M~o/2f(s)H'/H(s ) ~ 0
for rn ~ c~
for any s on the horizontal line segment S+m defined by 0- + iTm with 0-1 < 0- < 02. R e m a r k . Of course, we are interested in conside~ng the growth condition G R when H is one of the functions Z, Z, or q'. In the present variation, we need only one growth condition of type G R to insure the existence of certain limiting integrals. T h e point is t h a t u n d e r the additional Fourier theoretic conditions, the convergence of the integrals on the vertical lines will be reduced to Fourier inversion on the middle, or critical, line Re(s) = 0-o/2, after shifting the line of integration, and picking up residues corresponding to zeros and poles of the factor O. In this chapter, we shall apply the estimates s u m m a r i z e d in w of Chapter I to obtain a criterion u n d e r which the growth condition G R is satisfied, as in the next theorem.
65
T h e o r e m 1.3. Let H be of regularized product type of reduced order (M, rn ), and assume that F satisties the growth conditions of Lemma 1.2 to order M. Then H satist~es the growth condition G R whenever F satisfies the three basic conditions to order M The proof of Theorem 1.3 follows directly from Theorem 2.5 of Chapter I and Lemma 1.2 above. Recall that the basic Fourier conditions F O U 1 and F O U 2 insure the possibility of applying one of the most classicM inversion theorems of Fourier analysis, stemming from Dirichlet, and attributed more directly to Pringsheim, Prasad and Hobson by Titchmarsch [Ti 48], page 25 (see also Theorem 2.5 of Chapter X in [La 93b]). For completeness, let us recall this result.
Basic Fourier Inversion Formula. Let/3 C BV(R)NLI(R), and suppose/3 is normalized. Then T
lim
T--+oo ~
1
/ / 3 ^ ( t ) e i t , dt =/3(x) -T
for a l l x E R .
66 w
The explicit formulas.
Let (Z, Z, (I)) be in the f u n d a m e n t a l class. Let a > 0 be such t h a t or0 + a > a~ and such that Z, Z and (~ do not have a zero or pole on the lines Re(s) = - a and Re(s) = o0 + a. We let f and F be measurable functions related by
F(x) = f(e -~)
f(u) = F ( - logu).
so
For the m o m e n t , assume there is some a' > a for which we have the b o u n d
(I)
tF(x)l << e
We shall actually assume s o m e t h i n g stronger later, but for now we just want to deal with a region of absolute convergence for a Mellin integral. Assuming the b o u n d stated in (1), the function M,o/2f(s ) is holomorphic in the closed strip - a < ~ < or0 + a. We let: T~a be the infinite rectangle b o u n d e d by the vertical lines Re(s)---a
and
Re(s) = o 0 + a .
T~a (T) be the finite rectangle b o u n d e d by the above vertical lines and the lines Ira(s) = - T
and
I m ( s ) = T.
We assume at first t h a t (I) has no zeros or poles on the line Re(s) = 00/2. A variation without this restriction will also be treated. T h e line with o0/2 is especially useful for the more classical applications to counting q's, or primes in the n u m b e r theoretic case. We let: {p} = the set of zeros and poles of Z in the full strip --a_(a
67
Assume that T is chosen so that the functions Z, Z, 9 have no zeros or poles on the horizontal lines that border 7~a(T). Then, we may form the finite sum
Vz,,~(f,T) =
E
v(p)M~'o/2f(p)
pen~,(T)
= divz,~(M~,o/2f , T). Similarly, we define
V,~,a,~o/2(f,T) = E v(a)M"o/2f(a)" O:
Let f--.a(T) be the boundary of the rectangle T~(T) and consider the integral
/
M~o/2f(s)Z'/Z(s)ds.
s
By the residue theorem we have the equality
/
M~o/2f(s)Z'/Z(s)ds
s
= 2ri
E
v(p)M"o/2f(p) = 2~riVz,a(f,T).
p e n , (T)
At this point, we want to take a limit as T --~ ec. For this, one must have a choice of T ~ ec such that the integrand on the top and bottom segments of the rectangle tends to 0, as hypothesized in the growth condition G R . Roughly speaking, the more zeros and poles the function Z has in the strip T~a, the larger Zr/Z(s) could be on such horizontal segments, and so the smoother the function f must be so that its Mellin transform approaches 0 sufficiently fast, meaning faster than Zr/Z(s) approaches infinity. Assuming the growth condition G R , we are interested in the infinite sum
Vz,~(f) = E pE7~a
v(p)M"o/2f(p)'
68
which is understood in the limiting sense
E v(p)M"o/2f(p) = mli~m~ E
(2)
v(p)M"o/2f(p)"
peT~=(T,-n)
pE'R~G
Since the similar sum over the family {a} is taken on the left half interval, we use the notation
v(a)M~o/2f(a) Re(c0
=
Z
lim
m--+oo
v(a)M"o/2f(a)"
aeT~,, ( Tm),Re( oO~ao /2
The Fourier conditions F O U 1 and F O U 2 are imposed in order to deal with questions of Fourier inversion in connection with sums over {q}. As we shall see, the condition F O U 3 is concerned with our evaluation of the W e l l f u n c t i o n a l We, which is defined to be the limit Tm
w (r)
=
1
f F^(t)#2'/O(ao/2 + it)dt, -Tm
where #2 is assumed to be holomorphic on the line Re(s) = In addition to these assumptions, we will require:
ao/2.
F O U 4. There exists a constant a ~ > 0 such that the function
x ~ F(x)e (~~ is in BV(R). Under suitable conditions on the test function F, which will be expressed in terms of the above four Fourier conditions, we shall prove:
The Explicit Formula.
Vz,a(f) + Y,La,ao/2(f) = E
- c (qq~) qlog _ g ~ )(q)' + E
-c~g61f(1/~t)+q W,~(F).
More specifically, the main result of this chapter is the following theorem.
69
2.1. Let (Z, Z, ~) be in the fundamental class, and assume that if2 has reduced order (M, m) with no zeros or poles on the line Re(s) = ao/2. Then for any function F which satisfies the four Fourier conditions to order M , the functionals W e ( F ) , Vz,~(f) and Ve,~,~o/2(f) are defined and the explicit Theorem
formula holds. Observe that, as in our formulation of Crame~r's theorem, t h e above t h e o r e m is an inductive one, expressing t h e sum Vz,~ (f) in terms of a similar sum concerning Ve,~,~0/2, the Weil functional, and terms involving the families {q} and {4}. Note that t h r o u g h a "change of n o t a t i o n " , T h e o r e m 2.1 can be used to express an explicit formula involving the zeros and poles of Z. Remark such t h a t
1. Suppose there are m e r o m o r p h i c functions G a n d 9 (s) = C ( s ) / d ( a 0 - s).
T h e n we m a y write the Weil functional as
We(F) = W+(F) + Wj(F) where
1/
Tm
W+(F) =
ji%
F^(t)G'/G(ao/2 + it)dr
-Tm
and Tm
w~(F) = Ji%
1
/
FA(-t)Gt/G(cro/2 § it)dr.
--Trn
Further, in the case where G = (~, we can write the Weil functional as
1/
Trn
we(F) = limoo vS~
--Tm
[FA(t) § FA(--t)] G'/G(ao/2 -~ it)dr.
70
R e m a r k 2. Observe that the sum Vz, a + Vr is independent of a even though neither term is independent of a. For any u such that 9 has no zero or pole on the line Re(s) = u, we define u+iT~
W : , , ( f ) = lirno~ 2 z c i
M~~
u-iTm
The proof of the explicit formula will go through the following intermediate stage. T h e o r e m 2.2. Assume that 9 is of regularized product type of reduced order (M, m), and F satisi~es the Fourier conditions to order M. Then
Vz, a(f) = Eq- c ( q )qaO/2 l~ q f ( q ) + Z~I -c(s176 61~o/2 ~Iff' l/~l~ + W r# Theorem 2.2 will be proved in w After Theorem 2.2 has been proved, what will remain is to analyze the last term containing O'/r Different applications require different analyses of this term. For classical analytic number theory, one moves the line of integration from - a to ao/2, and then one applies the general Parseval formula. This will be carried out in w thus yielding Theorem 2.1. In w we give a further determination of the Weil functional, based on the general Parseval formula from [JoL 93b]. However, for other applications and notably those occuring later in this book in Chapter V, w and w we shall move the line of integration far to the right. In these applications, one can thus completely bypass the Parseval formula, and the final result is therefore much simpler to prove. We now carry this out. Any regularized product type can be expressed as a product of two such types, one of which has all of its zeros and poles in a left half plane and one of which has all of its zeros and poles in a right half plane, say r = r
71
We let A > a0 be a number such that all the zeros and poles of r are in the half plane Re(s) < A - 6, for some 6 > 0. Similarly, select a is such that all the zeros and poles of ~right are in the half plane Re(s) _> - a + 6, for some 6 > 0. In Theorem 2.2 we move the line of integration for (]}left to Re(s) = A, thus picking up the residues of all the poles between these two lines. At the same time, we define
V,~,.ft,a,A(f ) = ~ v(r where {(} = the set of all zeros and poles of Oleft such that Re(() > - a . Then, from Theorem 2.2, we arrive at the following formula. T h e o r e m 2.3. Let (Z, Z, ~) be in the fundamentM class, a n d assume that 9 is decomposed as above. Then for any function F which satisfies the four Fourier conditions to order M , a11 the functioneds in the next formula are defined and the following formula holds:
Vz, a(f) + Vo~at,a,A(f) =
-c(q) log qf(q) + ~ -C(~l)log ~lf/1/~ q
(t*,o/2
qa0/2
+ W'~,~,t,A(f) +
~
"
/
Ori,ht,--a(f)"
Steps which justify moving the line of integration are given in w in the case considered in Theorem 2.1. The same argument applies to Theorem 2.2, thus yielding Theorem 2.3. R e m a r k 3. In Chapter n, for the proof of Cram6r's theorem, we did not use a Mellin transform but worked directly on a half strip with a test function r One may do the same in a full strip to get the explict formula directly for such a function. In that case, the Well functionals are expressed in terms of r instead of the Mellin transform Mo, o/2f. For the applications to Chapter IV and Chapter V, this way of proceeding eliminates completely all Fourier conditions, and we could deal directly with the special test functions
r
= e (~-~~
72 or
r
= (2~ - ~o)~r
just as we dealt with the function [JoL 93c].
e ~z
~>'', for the Cram6r t h e o r e m in
73
w
T h e t e r m s w i t h t h e q's.
In this and the next section we will begin our proof of the explicit formula based on the axioms F O U 1 through F O U 4. In this section, we shall compute the sums over q that appear in Theorem 2.1. The remainder of the proof, namely the determination of the terms involving the fudge factor (I) and its set of zeros and poles {(~} will be given in the next section. Consider the rectangle TC~(Tm), as defined above, and integrate over the boundary of this rectangle. By the residue formula, we have
(1)
1
Z v(p)M~~ pOto(Tm)
27ri / M~~ L~(T.~)
Throughout we assume that (I) is of regularized product type of reduced order (M, rn). By Theorem 4.1 of Chapter II, Z and ~z are of regularized product type of reduced order (M, m + 1). We now can apply Lemma 1.2 to H = Z to find, for rn ~ cxz,
1 2 i
ao+a-iTm
f
M~o/2f(s)Z,/Z(s)ds
--a--iTm - a T iTra
(2)
M~o/2f(s)Z'/Z(s)ds
+ 2~ri
= o(1).
o'o+a+iTm
Upon combining (1) and (2), we obtain
E
v(p)M~o/2f(p) + o(1) -a-iTm
1 f 2~ri
M,,o/2f(s)Z,/Z(s)d s
--a+iTm ao+a-t-iTm
4--27ri -1
/
ao+a--iTm
M~o/2f(s)Z,/Z(s)ds"
74
Using the functional equation, we can express this equality as
1/
--a-iT~
2~ri
-aA-iTm -a-iTm
1 (3)
f M.ol2f(s)[_@,(s)/@(s)lds
+ 27r----i
-a+iTm ao+a+iT~
1
/
+ 2~r---i
M~,o/2f(s)Z,/Z(s)ds.
ao+a--iT,~
As stated above, in this section we will deal with the Z and Z integrals, and in the next section we will evaluate the @ integral. For the Z integral, we obtain ao+a+iT~
1 2zri
/
M,~o/2f(s)Z,/Z(s)ds
tro+a-iTm Tm
= -1 21r
(4)
f dt / F(x)e-(~'~176 --Tra
--oo
qao+a+itdx. q
This step follows by differentiating the Euler sum for log Z to obtain the series for Z'/Z. We shall give formal arguments to change this last expression (4), and then we give estimates to justify the calculations. We make a change of variables x=y-logq
and
dx=dy
in the integral of each term with q. Then the whole expression (4) becomes Trn
-- --1 27r / _
--Tm
dtZ
J
~ logq__: z'~yc(q)
q --c~
Trn
-I
= 27r f d t Z f Hq(')e-it'd' -T,~
q -oo
logq)e-(a~
75 where Hq(y) - c(q) log q F(y - log q)e -(~~ We let
H(Y)=EHq(y). q
With the definition of H, we may express our desired limit integral in (4) as being Trn
-127r,~--~lim / -Tm
dt ff H(y)e-itYdy:-H^^(O) : - H ( O ) . -oo
by the Basic Fourier Inversion Formula. We then see that -H(0) = E
- c ( q ) log q F ( qao/2 _~ log q ) = E
q
--c(q) log q ~ q--9-~o/2 J(q)'
q
which is a term in the statement of Theorem 2.1. We shall now justify the above steps. First, by F O U 4, there is a constant C such that IF(x)l <_
Ce -(~~
from which we get the estimates
IF(y
-
log q)l e -(~~
< -
-(~~
{Cq
,
Cq ~~
(~'-'~)y
-
e -(~~
and, in particular,
[F(y - log q)[ e-('~
< C q -(~~
for all y. From this it follows that (5)
IHq(y)l < 2C Ic(q)ll~ -
qao+a
y < logq
'
y > logq,
76
and after actually performing the integration, we obtain oo
-
q~0+~
a ~ -
a
cr 0 q - a q -
aI
--(X)
Estimate (5) shows that the series
H(y)=ZHq(y) q
is absolutely and uniformly convergent, and estimate (6) shows that this series defines a function y ~ H(y) in LI(R). Since each term Hq is of bounded variation and normalized, the uniform convergence of the series (6) shows that H is normalized. Furthermore, the total variation VR satisfies the triangle inequality for an infinite sum, as one verifies directly from the RiemannSteiltjes sums defining this variation. Similarly for a product, we h ave
WR(gh) <_ IlgllWR(h) + IlhllWR(g) Then VR(H) < ~ --
q
Ic(q)ll~ q g R q~o+a
(F(x)e -(~~
Hence, by F O U 4, H has bounded variation, and we have justified all the formal operations and the application of Fourier inversion in the evaluation of the Z integral as - H ( 0 ) . One may carry out similar arguments for the z~ integral -a-iT~
2rri
M~o/2Z(s)Z'/Z(ao - s)ds,
--a+iT~
which we write as Tm
l~ ~ dt / F(x)e -(~~/2+~+it)~ E_ c(q) ~t~ i?dx"
27r --rim
-- r
q
77
In this case, one uses the inequalities
{ C~i-(ao/2+a')e(a-a')v y >_ - log ~1 <-- C~laO/2+a,e(aO+a+a,)yy < - log ~t,
IF(y + log Cl)l e(~176 and
IF(y + log 4)1 e (*~
-< co -(~~
which holds for all y. Inequalities similiar to (5) and (6) t h e n follow when we define
Hel(y) - c( (:l) l~ ft F(y -4- log (t)e (<'~ and
H(y) = E Hel(y), el with the b o u n d e d variation
et In this way, we obtain the t e r m involving the z~ integral in T h e o r e m 2.1. This concludes the proof of the T h e o r e m 2.1 as far as it involves the Z and Z integrals. At this point, we have proved T h e o r e m 2.2.
78
w
The term involving r
In this section we will compute the terms in the explicit formula containing the fudge factor r and its set of roots {a}. To do so, we begin by considering the integral containing r namely - a + iTm
1 f 27ri
M~o/2f(s)~,/~2(s)ds.
--a-iTm
We want to move the line of integration to the critical line a = a0/2. Upon doing this, we pick up the residues of M~o/2f(s)~'/~(s) at the points a, and, hence, by using F O U 1, we find --aWiTm
1 / 2~ri
M~,o/2f(s)O,/O(s)ds
--a--iTm ao/2+iTm
= 2rci 1
f
Mvo/2f(s),~,/@(s)ds
r
-
E
v(a)Mao/2f(a ) + o(1)
a e R ~ (Tin) Tm 1
/ F^(t)O'/O(ao/2 + it)dt -Tm
(1)
E
-
v(a)M~o/2f(a) + o(1).
~eno(Tm) Using F O U 1 and F O U 2, we obtain the equality -a+iTm
1 lira 2~'i m~oo
ff
M~o/2f(s)~'/r
= We(F)-
Vr
--a--iTm
With this, as well as the calculations from the previous section, we have concluded the proof of Theorem 2.1, up to the evaluation of the Weil functional, which will now be dealt with.
79
w
The Well functional and regularized product type.
In this section we consider the evaluation of the Weil functional We when 9 is of regularized product type of reduced order (M, rn) and F satisfies the Fourier conditions to order M. Essentially, we will apply the general Parseval formula of [JoL 93b]. Quite generally, for suitable functions g and H, and u 6 R, we shall consider the Weil functional, which we define to be T
WH,u(g) = lim
1
T---*oo V / ~
g ^ ( t ) H ' / H ( u + it)dt.
/ -T
Our analysis involves cases when H is one of the following three types of functions. C a s e 1. H = Q for some rational function Q. C a s e 2. H
---- c P
for some polynomial P.
C a s e 3. H = D for some regularized product D. In the remainder of this section, we will evaluate the Weil functional in each of the above three cases. C a s e 1: H = Q for s o m e r a t i o n a l f u n c t i o n Q. It suffices to consider the function H(z) = z + a for some complex number a. T h e o r e m 5.1. Assume g satist~es F O U 1 and F O U 2. Then CO
/
-i
I F
9^(t) t +1. dt =
9 dx, g(x)e '"~
Im(a) > O,
o O0
i/g(-x)e-i~dx,
Im(~) < 0.
0
Proof. Lemma 1.1 shows that g^(t)/(t + a) is in L I ( R ) , so if
80
Im(a) > 0, we have O0
1
/
O0
1 dt_
r
+
1
/
-- oo
O0
/ _iei(t+~)Xdxdt
--oo
0
oo
= - i f g(x)ei~dx, o
by the Fubini Theorem and the Basic Fourier Inversion Formula. The case Im(a) < 0 is treated similarly. [] C a s e 2: H
=
e P
for s o m e p o l y n o m i a l P .
In this case, the evaluation of the Weil functional reduces to the Basic Fourier Inversion Formula as the following theorem demonstrates. T h e o r e m 5.2. Assume g satisfies F O U 1 and F O U 2 to order M. Let pi be a polynomial of degree < M, and let P'(-iO) be the corresponding constant coefficient partial differential operator. Then T
lira
/ gn(t)P'(t)dt = P'(-iO)g(O).
1
T----*ov
--T
Proof. The Basic Fourier Inversion Formula states T
lim
1
T----+cr X / ~
/ g^(t)P,(t)eit~d t = P ' ( - i o ) g ( x ) . -T
The assertion follows after we set x = 0.
[]
Case 3: H = D for s o m e r e g u l a r i z e d p r o d u c t D. This case is handled by the results from [JoL 92b], which generalizes Barner's formulation [Ba 81] of Weil's formula [We 52] in the special case H ' / H is the logarithmic derivative of the classical
81 gamma function. For completeness, let us recall the main theorem from [JoL 92b]. Let L and A be as in w and let P(x) be the corresponding theta function
P(x) = E
ake-)'~t
that satisfies the asymptotic axioms A S 1, A S 2 and A S 3. For any n let Ln = { A n + l , . . . , } , so then we have n
DIL/DL(z) = E
(1)
z +ak~--~~- D~Ln/DL"(Z)"
k=0
Since Theorem 5.1 applies to the sum in (1), we may assume, without loss of generality, that L is such that max{--Re()~k)} < u.
~kEL
The principal part of the theta function P(x) is BoP(X)=
b,(x)x',
E Re(p)<0
which, by AS 2, is such that one has the asymptotic behavior
P(x) - PoP(x) = 0(I log xl m(~
as x ---~ 0.
For any z ~ C, let
P~(x) =P(x)~ - ~ . By expanding e -z~ in a power series, we see that the principal part
of P~(x) is e0P~(x) = B0 [ , - ~ P ( x ) ]
4(x)xp+k
= rte(p)+k<0
k!
(-z)k"
82
Recall from w (Theorem 4.1 of [JoL 92a]) that, for any constant > 0, the logarithmic derivative of the regularized product can be written as
D ' / D ( u + it) = I~(u - a + it) + S~(u - a + it) where O0
_r,,,(z) = f [e.(x) - eoe.(x)] e-WZdx o
and
(-z) k
k---T--CT'=~176
Sw(z) =
[r(s + p + k + l)
[
;~-;~-vT~
Re(p)Wk
Note that, as a function of z, S~(z) is a polynomial of degree _< M, in which case Theorem 5.2 applies to show that S~(z) satisfies G R for any test function g that satisfies the basic conditions to order M. Therefore, in order to evaluate T
lim
T---*oo
1
/
g ^ ( t ) D ' / D ( u + it)dt,
-T
it suffices to evaluate T
lira
1
/
g^(t)I~(u-
a + it)dt.
--T
If we restrict the complex variable z to a vertical line by setting z = u - a + it, we get
~p(x)xp+k
Poe~(x) =
k!
(-u
+ ~ - it) k
Re(p)Wk
=
E k<-Re(po)
ck(u-~,x)(-it) ~,
83
where the coefficients Ck(U -- a, x) depend on the variables u - a and x through the coefficients of t p for Re(p) < 0. With this, we have
I~(u-a+it)
=/[ O(x)e-z(u-a+it)-O
E
Ck(U--a'x)(--it)k
e-~dx"
k<-Re(p0)
Let us define
d#~(x) = e - ~ d x
and
O~_~(x) = O(x)e -(~-~)~.
Therefore, the above calculations yield the equality
I~(u-a+it) O0
=
f 0~_~,{X', le-itx 0
)_~
Ck(U--a,x)(--it) k d # , ( x ) ,
k(--Re(p0)
Applying Theorem 4.3 from [JoL 92b] we obtain: T h e o r e m 5.3. Assume g satisfies the Fourier conditions to order M. Then, with notation as above, we have T
lim
T--*c~
1
f
g^(t)G(u-
a + it)dt
-T OO
=f o
G_~(x)g(-x)-
ck(u
k<--Re(po)
-
In summary, we have shown: T h e o r e m 5.4. Assume that ~ is of regularized product type or reduced order M. If g is a test function satisfying the four basic Fourier conditions to order M, then the WeB functional Wv,~(g) is defined. Further, Theorems 5.1, 5.2 and 5.3 combine to provide an explicit evaluation of this functionaJ. For the sake of space, we will explicitly evaluate the Weil functional only for special functions r namely those involving the classical g a m m a function.
84
E x a m p l e : T h e g a m m a f u n c t i o n . Many known zeta functions, such as those associated to number fields or modular forms, have Euler sums and functional equations with fudge factors which involve the gamma function. For example, the zeta function corresponding to an ideal class in a number field k has a functional equation with a0 = 1 and
AsI2r(sl2)r.r(~) r~,
a(~) = d(s) = SO
G ' / G ( s ) = ~ logA +
r'/r(42)
+ r~r'/r(~).
So, the evaluation of the Weil functional in this case reduces to considering the logarithmic derivative of the gamma function. As recalled on page 52 of [JoL 93a] and w of Chapter I, the (lesser known) Gauss formula for the gamma function states that for any z E C with Re(z) > - 1 , we have oc)
e-tdt o
where
0(3
o(t) =
1 -- e -t" n=0
From this, and Theorem 4.3 of [JoL 93b], for any f which satisfies the four basic Fourier conditions and a > - 1 , we have
ff"(t)r'/r(a+it+l)dt T
lim
T--~
1
-T
e - x dz.
This is Weil's formula as reworked by Barner (see [Ba 81], page 146).
CHAPTER
IV
From Functional Equations to Theta Inversions The classical aacobi theta inversion formula for the Riemann theta function of one variable states that for all t > 0, we have the equality 1 ~--~n=-oo E e- n 2 t If we set
1
v~
n=E-ooe-(2rtr02/4t
OO
then the Jacobi inversion formula can be stated as the equality
e(u) = ~ e ( 1 / u )
for u > 0.
SpectrMly, the inversion formula can be viewed as expressing a sum over all the eigenvalues of the Laplace operator on the circle (namely the squares of the integers) as equal to another similar sum, with the inversion t ~-+ 1/t. We give the following very simple spectral interpretation of the Jacobi inversion formula. Let X = 2 r r Z \ R be the circle. The heat kernel for the usual Laplacian on R is
KR(x,t,v) =
1 _(x_v)2/4t 4vqTi~
The heat kernel on X is the 27rZ periodization of the heat kernel KR on R. On the other hand, the eigenfunction expansion of the
86 heat kernel K x can be easily computed. When we equate this periodization with the eigenfunction expansion of the heat kernel, we obtain what amounts to a theta inversion formula, namely 1
oo
oo
1
r~=--oo
e--nZt einX e--iny.
271" n = - o o
In Theorem 1.1 below, we show how the above classical theta inversion formula admits a vast extension to much more general theta functions, essentially formed with the sequence of zeros and poles of functions in the fundamental class as defined in Chapter II, w Specifically, inversion formulas follow from our general explicit formula when using Gaussian type test functions. In this context, the Jacobi inversion formula comes from the explicit formulas associated to the sine function e--~rs
sin(zris) --
i
2i
e ~s
elrs
--
2i
(1 - e -2'~8)
which, when written in this form, can be seen to have an Euler sum and functional equation with cr0 = 0 and a simple exponential fudge factor. We will prove our general inversion formulas in w and give various examples in w We will show conversely in Chapter V how inversion formulas for theta functions satisfying AN 1, AS 2, and A S 3 yield Dirichlet series with an additive functional equation.
87
w
A n a p p l i c a t i o n of t h e explicit f o r m u l a s .
We shall apply the general explicit formula of Chapter III, to the test function ft defined for t > 0 by (1)
ft(u)=
1,~--e-O~ x/ 41rt
Ft(x)-
so
_ ~ e--x2/4t . 4~_
Note that Ft is the heat kernel on R. It is immediate that Ft satisfies the four basic Fourier conditions needed in the proof of the explicit formulas. By a direct calculation, we have (2)
M~o/2ft(s)=eCs-~~
and
FtA(r)= Vr~ 1 e _r2,
For example, to derive the first formula in (2), write OO
M~'~
= f f'(u)uS-~'~ 0 O0
_ =
1
f e_z214t+z(s_aol2)dx
e(S-ao/2) 2t.
With this, the explicit formula yields the following result, which we call a t h e t a inversion formula. T h e o r e m 1.1. Let (Z, 2, ~) be in the fundamental class, and assume that 9 has no zeros or poles on the line Re(s) = a0/2. Let {p) and {~} be as in Chapter m , w Then ~or ~I t > 0 we h ave
1
E -c(q) log q e_(logq)2/4t
4v~7 +-- 1
qao/2
q
E el
+
E~(t),
-c(~)log ~o12
qe_(log ~1)2/4t
88
where
1
Er
= ~
O0
f
+ it)dr.
~-r%'/r
--CX)
R e m a r k 1. As in Chapter III, sums over p, a and the integral for Er are to be understood as limits of sums and integral from -Tin to Tin, taken over a suitably defined sequence {Tin}, depending on Z and (I). R e m a r k 2. If (I)(s) =
G(s)/G(ao - s),
then
E,(~) : Ea(~) + Ea(~) where O0
Ea(,-) =
1 / e_r2tG,/G(ao/2+ir)dr. -- 00
In particular, if G = G, then E r
=
2EG(r).
Let us establish some notation in order to write the formula in Theorem 1.1 in a form fitting AS 1, AS 2 and AS 3. Let L = {#k} be the set of numbers
r = {~k } = { - ( p - ~0/2) 2, - ( ~ - ~0/2)2}, ordered as a sequence, with integral multiplicities
{a(~k)} = {v(p),v(~)}. Define the associated theta function
(3)
0L(~) = ~
v(p)~ (.-~0/~)~' + ~ p
v(.)~ ("-~o/~)2' c~
Similarly, let L v be the family of numbers L V = {#vk} = {(1ogq)2/4,(l~ with (not necessarily integral) "multiplicities" {a(#~)} = { - c ( q ) l o g q -c((])1og 61 qc,ol2 ' ~ao/2 j,
89
and define the associated theta function --c(q)log qe_[(,og q)2/4]t qao/2
OLV(t) = ~ q
+
(4)
et With this notation, we can now write Theorem 1.1 in the following form. T h e o r e m 1.2. With notation as above, we have the inversion formula 1 On(t) .~_~:8nv(1/t)+ E , ( t ) . x/ 41rt In general, we define a theta inversion formula to be a relation between two theta functions, such as (3) and (4), with an additional term such as Ev(t) which we require to satisfy the asymptotic condition AS 2. T h e o r e m 1.3. With notation as above, the theta function
p
satistles the asymptotic conditions AS 1, AS 2, and AS 3. Proof. By the Cram6r theorem, we have that the sequences {p/i:Z(p)=O
with Im(p) > 0
and
-a
with I m ( p ) < 0
and
-a
and
{p/i:Z(p)=O
are such that the associated theta functions satisfy the three basic asymptotic conditions. Therefore, by applying Theorem 7.6 and Theorem 7.7 of [JoL 93a], we conclude that the theta function
~v(p)e(p-~ol2) ~t p
90
satisfies the asymptotic conditions.
[]
The theta series taken over {a} is incomplete, and it will be more useful in this chapter to deal with the alternate version of the explicit formula given as Theorem 2.3 in w of Chapter III. We let R be of regularized harmonic series type. For each real number u such that R has no pole on the line Re(z) = u, we have the E ~ - t r a n s f o r m u+iTn
Eun(t)=
,~--~lim27ril
/
e(Z_~o/2)~tn(z)d z
u-iT,.,
where {Tn} is a sequence selected as in Theorem 6.2 of Chapter I. Directly from Theorem 2.3 of Chapter III, we have the following result. T h e o r e m 1.4. Let (Z, Z, if) be in the fundamental class. Decompose ~ as a product ~ ~left ~right
such that (I}left ha8 all its zeros and poles in a left hedf plane and r has all its zeros and poles in a right half plane. Select A > ao, 6 > O, and a such that: ffleft has all its zeros and poles in Re(z) _< A - 6; ffright has all its zeros and poles in Re(z) >_ - a + 6. Let {r } = set of zeros and poles of ~,o. with R e ( ( ) > - a .
Then for M1 t > 0 we have p
_1~
+~
[~q--C(q)logqe_OOgq)2/4t]
4~o/~
+ E~(elo./e~o.)(O + E-o(e',i~h,/e~h,)(t).
91 We note that in Theorem 1.4 both theta series satisfy the asymptotic conditions A S 1, A S 2, and A S 3. This assertion is true for the sum over ~ since we assumed ~ was of regularized product type, because we choose - a sufficiently far to the left, and so we can apply Theorem 7.6 and Theorem 7.7 of [JoL 93a]. As for the sum over p, one applies Cram~r's theorem and Theorem 7.6 and Theorem. 7.7 of [JoL 93a]. R e m a r k 3. In the present application, and the subsequent ones in this chapter and the next, we apply the explicit formula to the simplest types of Gaussian test functions. Even for such test functions, one can give examples where instead of (47rt) 1/2 or (4~rt) 3/2 (as in Chapter V, w one takes (4~rt) n/2 for an odd integer n, and in addition one also has an arbitrary polynomial as a coefficient. If one uses (4~rt) n/2 with an even integer n, then one gets Bessel series instead of Dirichlet series. All these cases deserve to be treated systematically, since they apply to several important situations of spectral analysis for classical manifolds, and we shall do so elsewhere. Here we selected the simplest cases to serve as examples.
92
w
Some examples of theta inversions
To emphasize the significance of the theta inversion formula given in Theorem 1.1, let us now discuss a few specific applications of the theorem. E x a m p l e 1: T h e s i n e f u n c t i o n . Let us write the sine function as 6 -Trs
sin(Tris) --
__ e 7rs
elrs
2i
= a(4Z(s)
2i
where G(s) = - e ' ~ ' / 2 i and Z(s) = 1 - e -2'~'. The fact that the sine function is odd trivially yields the functional equation = -a(-4z(-,),
so a0 = 0. Further, Z satisfies the Euler sum condition since log Z(s) = log (1 - e - 2 ~ ' ) ---- ' ~ n.~--I
le-2'~ns, n
whence {q} = {e 2rrn } for n > I and c(e 2rrn) = 1/n. Since sin(Tris)is zero only when s E Z, and then with multiplicity one, the inversion formula Theorem 1.1 specializes to the equality oo
1
(X)
E e - " h = 2 " - -4x/-~tE 2~re-(2'~n)~/4t+ Ee(t),
(1)
n----1
n.~----O0
where the factor of 2 appears since the sums over {q} and over {~} coincide in this example. Also, we have (2)
rr
v'~ "
Combining the terms in (1) and (2), we obtain the classical Jacobi inversion formula, which is the relation I
oo
27r n=--c~ E 6-n2t --
I
~
oo
E
e--(21rn)2/4t
for t > O.
93
Similarly, one can derive other classical formulas using the cosine function and hyperbolic trigonometric functions. E x a m p l e 2: D i r i c h l e t p o l y n o m i a l s . Recall the definition of a Dirichlet polynomial given in w of Chapter I. We saw in Chapter II that any Dirichlet polynomial P can be written as
P(s)=aNbSN [ 1+ ~1 aN a'--2-n\bN,] J =aNbsN'Z(s)' where Z has an Euler sum and functional equation with fudge factor ~(S)=
aN ( b N ~ s a, \ - C 1 /
Hence, one can directly evaluate the Weil functional, yielding the formula O0
E,(t) =
1 / e_r2t(log(bN/bl))dr_ log(bN/b1 ) --00
Therefore, the associated theta inversion formula is simply v(p)e" 2* _
log(bN/bl) x/47rt
~
1
~
c(q)(log q)e -(l~
a)~/4t
q 1
~ C(~l)(log (])e -O~ eO'/4t. el
The specific case of N = 2 with b2 = b~-1 = e '~s
and
a2 = - a l =
i/2
yields the Jacobi inversion formula. We find this example particularly interesting for the following reason. In Example 1, the Jacobi inversion formula is a relation involving two well-known sets of data, meaning that both theta functions in the inversion formula involves the squares of integers. However, in the case of a Dirichlet polynomial, we have one wellknown set of numbers, namely the sets {q} and {(t} which are
94 explicitly and simply expressed in terms of the initial set of numbers {aj, bj}, and one unknown set of numbers, namely the set of zeros.
Example 3: Zeta functions of number fields. In w of Chapter III we gave an evaluation of the Weil functional associated to the classical g a m m a function. Hence, there is a theta inversion formula associated to any zeta function of a number field as in Theorem 1.1. The term Ev or EG is given explicitly as follows. If =
,
then, by page 146 of [Ba Sl], see also w of [JoL 93b], we find
EG(t) --
log A 1/2
-2-~
i e-~h dr --OO
i r'/r(l/2+ it)e-r2tdr --OO
OO
+ 2---T-
0
_rl/2 _
__ __ 2e-X2/4te3x/2O(2x)] e-2Xdx 0
where
oo
n---~0
1 1 - e -x
It is important to note that to zeta functions, we are associating theta functions which admit inversion formulas but are different from the theta functions used in Hecke's proof of the functional equation and meromorphic continuation (see, for example, Chapter XIII of [La 70]).
Example 4: A connection with regularized products. In spectral theory, one meets the situation when a certain operator has the sequence of eigenvalues L = {~k } with integer multiplicities {ak}. In such a situation, when the theory of regularized products
95
applies (namely our axioms from [JoL 93a], recalled in Chapter I, w one may have the additional relation DL(~(8--ao)) = Z(s)G(8).
This occurs for instance for the Selberg zeta function and its analogues. Then (pk - * 0 / 2 ) 2 = - x k + o0 /4. Thus we have the simple relation v(pk)e (ph-a~ k
--
e(ao2/4)t ~"~ake-~
t --_ e(a~/4)teL(t),
k
expressing the theta series formed with the squares (Pk - a0/2) 2 in terms of the theta series with the eigenvalues. As stated above, in the case of the operator d 2/dx 2 on the circle S 1 = R/2~rZ, this relation reduces to the Jacobi theta series formed with )~k = k 2 and a0 = 0. In both cases we are dealing with the eigenvalues of the positive Laplacian on some Riemannian manifold.
CHAPTER
V
From Theta Inversions to F u n c t i o n a l E q u a t i o n s In this chapter we shall carry out the inverse construction of the preceeding sections; that is, from a theta series satisfying an inversion formula, we derive, by means of a Gauss transform, a Dirichlet series satisfying an additive functional equation. Of course, Riemann's proof of the functional equation of the Riemann zeta function also relied on a theta inversion formula, but in the present situation, our use of theta inversion is different from that of Riemann because we take a Laplace transform, with a quadratic change of variables, of the regularized theta series instead of the Mellin transform of the theta series. Hence we construct new types of zeta functions which are essentially regularized harmonic series. In case the residues are integers, such series are logarithmic derivatives of functions in the fundamental class. Thus, we see that the general theory requires that we consider the additive class rather than the more restrictive multiplicative class. Let R be of regularized harmonic series type. In w1, we analyze the transform E~ defined by u+iT,~
E,,R(t) = 2ira
1 J e(~_,,o/2)2tR(z)dz. u-iT.
In w we carry out the properties of the Gauss transform of theta These propseries, inverted theta series, and the transform erties are shown to imply the functional equation of the new zeta function in w In w we work out an example to obtain a zeta function for any compact quotient of the three dimensional hyperbolic space Ma. This zeta function lies in our additive class but, in general, not in our multiplicative class. For extensions of this example, see the remarks at the end of w
E,,R.
Throughout the remainder of this chapter, unless otherwise specified, we use the following basic conditions.
98
T h e basic conditions We let ~r0 be a real number > 0. We let R be a function of regularized harmonic series type. We let {#k} and {ak ) satisfy D I R 1, D I R 2, and D I R 3, and we assume that the corresponding theta series oo
O(t) = OL(t) = ~
ake - ~ `
k=O
satisfies AS 1, AS 2, and AS 3. We let (Mo, m) be the reduced order of the theta series 8. Note that we have used #k instead of Ak in the notation. This is because in a subsequent application to spectral theory, there will be an operator with eigenvalues /kk which are translates of #k, namely
#k -- Ak - a2o/4. We let R0 be the regularized harmonic series associated to the theta function 8, as defined in Chapter I, w that is,
Ro(z)=CT,=ILMO(s,z). We let {q} be a sequence of real numbers > 1 converging to infinity. We let {c(q)} be a sequence of complex numbers, and we let c(q)
=Z
q
qS
be the associated Dirichlet series, which we assume converges absolutely in some right half plane. Thus, JI'(s) = ~ q
- c ( q ) log q q~
We let OLv (t) : ~ q
--c(q)log qe_[(logq),/4], q#o/2
99 w
T h e WeU f u n c t i o n a l of a G a u s s i a n t e s t f u n c t i o n .
In this section, we are concerned with a function R of regularized harmonic series type. For each ( E C we let:
aa(~) = a(~) = residue of R at 4Observe that if R = Re = ~ / ~ , where ~ is of regularized product type, then a(4) = v(() is the order of 9 at ~. Any function R of regularized series type can be expressed as a sum of two such functions, each of which has poles only in a half plane to the left or a half plane to the right. Let us write such a decomposition as R ( z ) -~ Rleft(z) + Rright(Z). Assume that A > 0 is chosen so that all the poles of Rle~t lie in a left half plane of the form Re(z) < A - 5, for some ~ > 0. Similarly, assume that a > 0 is chosen so that all the poles of Rright lie in a right half plane of the form Re(z) k - a + 5, for some ~ > 0. Assume that Rleft has reduced order (Mleft,mleft) and Rright has reduced order (Mright, mright) As a direct application of Theorem 6.2(a) of Chapter I, we obtain the following result. L e m m a 1.1. W i t h notation as above, we have E a R l e f t ( t ) = 0(• -(Ml*rt+2)/2)
for $ ---+0
and E-aRright(t) = O(t -(M~'ght+2)/2)
for t -~ O.
Proof. By Theorem 6.2(a) of Chapter I, we know that on the line Re(z) = A the function Rlea has polynomial growth. In fact, the integral giving EARle~t can be coarsely estimated by
with some number b > 0. The first estimate asserted then follows from the standard change of variables y = x/tu. The second estimate is proved similarly. [] The proof of Lemma 1.1 also applies to prove the following asymptotic bounds.
100 Lemma
1.2. W i t h notation as above, we have
EARleft(t) = O(t -(Ml'ft+2)/2)
for t --~ co
and E-aRright(t) =- O(t -(M~ight+2)/2)
for t ---~ co.
Note t h a t the b o u n d given in T h e o r e m 6.2(a) of C h a p t e r I is stronger t h a n w h a t is used in the proof of L e m m a 1.1 a n d L e m m a 1.2 above. However, t h e above results are sufficient for our purposes. In fact, one can easily improve the exponent to - ( M + 1)/2 + e for any e > 0. T h e b o u n d s in L e m m a 1.1 and L e m m a 1.2 are e n o u g h to allow us to deal with the Gauss transforms of EARle~t a n d E-aRright in the next section.
101
w
Gauss transforms.
We shall need some analysis concerning Bessel integrals and what we call the Gauss transform, which we carry out in this section. Throughout we let r, t, u, x be real variables with t, u, x > 0. We start by recalling the collapse of the Bessel integral Ks(x, u) under certMn conditions. L e m m a 2.1. Let oo
f
dt
K~(x, u) = I g-x2, e-u2/tts a
7"
Then: a) In the case s = 1/2, we have the evaluation Kll2(X, u12) = Vl-~e-uz.
X
b) Let ~[~ and Z[~ be the differential operators
~I~ -
10 2x Ox
10 and
2lu=-~uuOuu.
Then we have the relations ( 2 I , ) " K s ( x , u ) = Ks+n(x,u)
and (2l~)"Ks(x,u) = K , _ n ( x , u ) . For further properties of the Bessel integral, including a proof of Lemma 2.1, the reader is referred to [La 87]. W a r n i n g . For our purposes, we use a normalization of the Bessel integral slightly different from the classical one. If, for instance, K ~ ( c ) denotes the g - S e s s e l function which one finds in tables (see Magnus, Oberhettinger, Erdelyi, Whittaker and Watson, etc.), then we have the relation
2K (2c) = Ks(c)
102
where OO
g,(c)
=f 0
The classical normalization gets rid of some factors in the differential equation, and our normalization gets rid of extraneous factors in the above integral. For a suitable function f of a single real variable, we recall the Laplace-Mellin transform is defined as
o
We shall deal with f = /9, where 0 is a t h e t a series of the sort considered previously. We shall put s = N + 1, where N is an integer > 0, sufficiently large to cancel the singularity at 0, and we shall also make a change of variables with z = (s - a0/2) 2. W i t h this, we obtain an integral operator which we call the G a u s s transform. More precisely, let N be a positive integer and let or0 be a real n u m b e r > 0. For a suitable test function f , we define OO
Gauss(~oN~2( f ) ( s ) = ( 2 s -
ao) / f(t)e-(S-a~
dtt
o
Then
Gauss(No]2(f)(s) = dz. ds
L M f ( N + 1 , ( s - a0/2)2).
Because of the change of variables from z to s, we are led to consider the differential operator :Ds_ao/2 -- - - -
d
1
o
ds 2s - ao"
Let x = s - a0/2. For every integer N > 0, we t h e n get the formula 1 __DN
2x" 21 N o 2x - -
s-~0/2
~
103 G a u s s t r a n s f o r m o f a n i n v e r t e d t h e t a series
We now shall prove inversion formulas for Gauss transforms of certain series. Such series occurred in the previous section in connection with functions in the fundamental class. Here we start with series for which we assume only the basic conditions. Directly from Lemma 2.1, we have the following theorem. T h e o r e m 2.2. Let G(t) -
1
~-, - c ( q ) log q e--(log q)2/4t.
4X/4' X/4' X/4' X/~4' X/q4~
qa0/2
Then for s r e M and sufl]ciently large, and N >_ O, we have.
GausS~o~]2(a)(s) = vf_,,o/2.n'(s ). Proof. The proof follows immediately from the definitions and andu=logqand Lemma2.1. For e a c h q , we put x = s - a o / 2 compute the transform of each individual term gq(,)
1
=
e
_u2/4t
as follows:
r
_
2x ~KN+l/2(z,u/2) 2x
- f---v,qlI.
N
9
g,12(x, u12)
- ~-~llx \2x = ,DN e-OOg ,:l)~
]
= DN_a0/2 (q-Sqa~ from which the formula in Theorem 2.2 is now clear.
, []
104
Gauss transform of a theta series
Next, let us apply the Gauss transform to a theta series itself; that is, let us evaluate CX~
~_ (28 -- (TO)/ O(t)E--(s-a~ 0
d-~
where OO
O(t) = E
ake-~kt"
k----O
Let Pk be complex numbers, all but a finite number of which have positive imaginary part, and such that (Pk - a0/2) 2 = --#k.
T h e o r e m 2.3. For any positive integer N s u ~ c i e n t l y large, and s t e a / a n d sut~ciently large, we have O0
/ O(t)e_(s_~o/2)~ttN+ 1 dt 0
k=0
Ek=0a k
r(N +
oo [( ~ -
1)
pk)(~ + ;~ - ~~
r ( g + 1) [(~ - ~ 0 / 2 ) ~
- (pk - ~ 0 / 2 ) ~ ] N + ~ '
hence the series gives the meromorphic continuation of the integrad to adl s C C with poles at the points s = Pk and s = cro - Pk. Proof. This identity follows routinely by integrating term by term, using the change of variables t ~ t / ( s - ao/2) 1/2. There is no problem with the square root since s was assume to be real and sufficiently large. []
Next we obtain the inversion analogous to Theorem 2.2, but for the theta series instead of the q series.
105
T h e o r e m 2.4. For any positive integer N sut~ciently large, a n d s real and sumciently large, we have
Gauss 0 /)2(O)(s) :
v f o/2
Proof. As a function of s, (2s - ao)Ro((s - a 0 / 2 ) 2) is a merom o r p h i c function whose singularities are simple poles at t h e points s : Pk a n d s ---- cr0 - Pk w i t h residue ak. For any individual k, we have - - -~Z
Z "gy ~ k
- - ( Z -~- ~ k ) N - b 1 "
So, N
2s --- a0
1
(~
-
--[(8 - r
~o/2)2
-
(pk
-
~o/2)2
r ( N + 1) 2 - ( P k -- (Yo/2)2] N-b1'
which can be w r i t t e n as 1 2 8 - - O"0
~N s--ao/2
2s - a0
(s - o o / 2 ~ -- (-pk - o 0 / 2 ) 2
]
J
"
T h e result now follows from T h e o r e m 2.3 after m u l t i p l y i n g by the t e r m 2s - a0 a n d t h e n s u m m i n g over k, which is valid since for N sufficiently large the series is absolutely convergent. [] R e m a r k 1. F r o m T h e o r e m 7.6 a n d T h e o r e m 7.7 of [JoL 93a], one can show t h a t it suffices to take N > (Mo + 2)/2, w h e r e has r e d u c e d order (Me, m). In fact, this r e q u i r e m e n t can be easily i m p r o v e d to N > (M0 + 1)/2 + e for any e > 0. R e m a r k 2. In the d e g e n e r a t e case w h e n Pk = a0/2, t h e regularized h a r m o n i c series w i t h the q u a d r a t i c change of variables has a double pole at s = a 0 / 2 w i t h zero residue, which is consistent with t h e above evaluation of residues.
106
I n v e r s e G a u s s t r a n s f o r m of a r e g u l a r i z e d h a r m o n i c series t y p e Finally, we deal with the Gauss transform applied to the terms
EARleft E-aRright which arose in Chapter IV. The next lemma evaluates this Gauss transform, and amounts to an inversion formula, showing that up to taking derivatives, the Gauss transform is the inverse of the EA transform. T h e o r e m 2.5. With notation as above, we have, for N sutl]ciently large and s real and sufficiently large, the equalities Causs(0N]2(EARleft)(s) = 7)Ea0/2Rleft(s ) and
GausS~o~/2 (E--a ~right )( S ) = "D;Lc~o/2 ~right (0.0 -- '~).
Proof. Let F(s) = / EARleft (t)e -(s-a~
ZttN+l dtt
0
which is defined for N sufficiently large by Lemma 1.1, and s real and sufficiently large by Lemma 1.2. Then
oo A+icx~ (1) F(s) -- 1 f / e(Z_ao/2)2te_(S_ao/2)2tRleft(z)tN+ldzdt 2~i ---t" 0 A--ioo By changing the order of integration, which is valid by Theorem 6.2(a) of Chapter I, we may write (1) in the form
,/
A+ioo r ( N + 1) [(8 - 0-0/2) 2 - (z - 0-0/2)2] N + I Rleft
(2) F(~)- ~ i
A-ioo Let us write
r(N + 1) [(s - ~
= ~,-~o/~
- (z - 0.0/2)2]N+,
(~ _ ~ o / 2 ) ~
-
(z - ~o/27
'
(z)dz.
107
which allows us to express (2) as
F(s) (3) A+ioo
27ri
~I~-~~
(s--
A-/oo
o0/2)5
(z
-
-
~0/2)2
Because of the s y m m e t r y in s and z, we can write (3), as
F(s) A+ic~
-
1 i 21ri
(~z-~0/2 )N
(-1) g
(~
- o0/27 : ~ - ~0/2)~ ]
R,e,<(z)dz
A-ioo A+ioo
1
- 2~i
(s - z)(s + z - a o )
(-JL-~~
Rleft(z)dz.
A--i~
Integration by parts shows t h a t -~[z-ao12 and :Dz_~0/2 are adjoint operators. Hence, we obtain the formula
G ausS(~oN]2(EA Rleft )(s ) = (2s -- a0 ) F ( s ) A+ic~ --
2~ri
- -
A--ioo
s-
z
s + z - a0
('Pz-ao/2)
left(z)dz.
From the proof of T h e o r e m 6.2(a) of C h a p t e r I, one obtains t h a t on the line Re(z) = A we have (4)
(T)z_ao/2)kRleit(A
-]- it) : O(It] M + e - 2 k )
f o r t ----> (x),
for any e > 0 a n d positive integer k. So, for N > ( M + 1)/2 + e, the integration by parts is valid. Now we can evaluate F(s) using a contour involving a halfcircle, opening to the right with diameter along the vertical line Re(s) = A. In this region, Rle~t(z) is holomorphic, so the only pole of the i n t e g r a n d occurs when z = s, and t h a t with residue 1,
108 after correcting the orientation of the contour. The integral along the half-circle will approach zero as the radius approaches infinity since, again by the proof of Theorem 6.2(a) of Chapter I, since the function (77z_,0/2)kRleft(z) has polynomial growth as in (4). With all this, the first part of Theorem 2.5 is proved. Concerning the second part of Theorem 2.5, one applies the same argument as above, except using a contour integral opening to the left, together with the identity 7:)z = D-z. [] R e m a r k 2. Lemma 1.1 and the estimate in (4) show that it suffices to take N > (M + 1)/2 + e where R has reduced order
(M,m).
109
w
T h e t a inversions yield zeta functions
We now put the results of w together to show how theta inversion gives rise to an additive functional equation. Given R of regularized harmonic series type, there exist a and A sufficiently large positive so that we can decompose R into a sum R : Rleft -[-
Rright
where Rleft has it poles in a left half plane Re(z) < A - 6, and where Rright has it poles in a right half plane Re(z) > - a + 6.
T h e o r e m 3.1. Assume that the theta series satisfies the inversion formula
eL(~) -
1
4v~eL,,(1/~) + EAR~o~t(t) + E-onri~ht(~).
For N sufficiently large, we have D~_~0/2 [(2s
-
a0)RoL ((s
-
~0/2)2)]
= vY-~0/2 [n'(~) + nle~(~) + n r i ~ ( ~ 0 - s)]. Proof. We simply apply Theorem 2.2, Theorem 2.3, and Theorem 2.5. [] R e m a r k 1. It suffices to take N for any e > O.
> l(max{MRl~ft,MR~ight,Mo}
-t- 1) +
e,
110
T h e o r e m 3.2. There is a p o l y n o m i M P ' os degree < 2N - 1 such that for all real s sumciently large, we have
(2s - a0)RoL
((s
-
a0/2) 2)
= JI'(s) + Rleft(s) + nright(G0 -- S) + P'(s). Proof. Let r ( ~ ) = (2~ - o0)RoL ((~ -- ~ 0 / 2 ) 2) -- a ' ( ~ ) -- R l e f t ( s ) - Rright(O'0 -- 3),
and let us write the formula in Theorem 3.1 as O 1 ) ~)N--1 Os 2s r ~_~o/2F(s) = O.
Hence, there exists a polynomial P1 of degree < 1 such that DN-1 s _ , o / 2 F ( s ) = Pl(s). The proof finishes by continuing to unwind the differential operator ~DN - 1
s_ero/2 9
[]
C o r o l l a r y 3.3. T h e Dirichlet series JU(s) has a m e r o m o r p h i c continuation such that under that m a p s ~-~ ao -- s the function J-I'(8) + Rleft(8) + Rright@r0
-- 8) + P ' ( s )
is odd.
This is immediate from Theorem 3.2 and the formula in Theorem 2.3. Upon integrating, the functional equation obtained in the present case is what we call an additive functional equation, to distinguish it from the multiplicative functional equation satisfied by functions in the fundamental class. We have now concluded the process from which a theta function with an inversion formula leads to a Dirichlet series having an additive functional equation with fudge terms which are of regularized harmonic series type.
111
E x a m p l e 1. If we apply the above construction to the Riemann zeta function, then we have
R(~) = ~1 r /,r ( , / 2 )
+ ~r'/r((1 - ~ ) / 2 ) - log ~,
hence we have R l e f t ( s ) ~-~
~r'/r(~/2)-log
7r,
and
Rright(S)
=
-~r'/r((1 - ~)/2).
It is easy to see that JI'(s) = 2(~/(Q(S) and M = 0, hence the above theorems apply with N = 1. Therefore, Corollary 3.3 reconstructs the fact that the function
2r162
= 2~/r
+ r'/r(~/2) - log~r
satisfies the functional equation ~ / ~ Q ( S ) = - ~ / ~ Q ( 1 - s). R e m a r k 2. It may seem odd that whereas we started with a triple (Z, Z, if) in our fundamental class with functional equation which is not symmetric, we derived a function in Corollary 3.3 which has a symmetric functionM equation up to an additive polynomial factor. The reason for this is the following. Since the test function to which we apply Theorem 2.3 of Chapter III is even, we can not distinguish between elements in the set {q} and elements in the set {~}. Therefore, we actually ended up considering the zeta function H i ( s ) = Z ( s ) Z ( s ) which satisfies the symmetric functionM equation Hl(s)~(s)
= Hx(ao - s ) ~ ( a o - s).
One could equally well use the odd test function Ft(x)-
-x/2t ~ e -
~/4" . "
so then M ~ o / 2 f ( s ) = - ( s - cro/2)e (~-~~
112 After carrying through the computations as above, we end up with a formula of the form ( 2~ - ~o ) n o ( ( ~ - ~o / 2 )2 )
= ,n'(~) -(~
-0-0/2
[n~oft(,~) - .n~ht(o-o - s)] + P'(,~).
This equation corresponds to the functional equation H2(s)H2(.o
- ~) = 1
where H~(~) - Z(.~)= .~(s). Z(~) Then, combining the two functional equations above, we recover the original functional equation (up to a sign) (z(.~),i,(~)) ~ = (2(0-o - ~))~, in other words z(~),I,(~) = +2(o-o - ~). E x a m p l e 2. The functional equation for the function //2 in Remark 2 is like that of the scattering determinant associated to any non-compact hyperbolic Riemann surface of finite volume (see [He 83]).
113
w176A new zeta function for compact quotients of M3. In this section we will give, what we believe, is a new example of a zeta function. The approach is that of the previous sections, namely to every theta inversion formula there is a corresponding zeta function. As we will see, the "zeta function" has a Dirichlet series in a hMf-plane, but is a regularized harmonic series whose singularities are simple poles whose residues are not necessarily integers. Hence the function is not the logarithmic derivative of a regularized product type. Let M3 be the simply connected three dimensional Riemannian manifold whose metric has constant sectional curvature equal to - 1 ; see page 38 of [Ch 84] for a precise model. On M3 there is a transitive group of isometries, so the heat kernel relative to the Laplacian acting on smooth functions, which we denote by KM3, is a function of distance r and t. In fact, let
(1)
h3(r,t)-
1 _r2/4te_ , r (47rt)3/2 e sinh r"
Then gM3(X,t,~)
=
h3(dM3(~,~/),t).
For details of this computation, see page 150 of [Ch 84] or page 397 of [DGM 76]. The result is basically due to Millson. Let X = F \ M 3 be a compact quotient of M3. Using the spectral decomposition of the heat kernel K x on X, one has the expression (2)
Kx(x,t,y) = E Ck(X)r k=O
where {r is a complete system of orthonormal eigenfunctions of the Laplacian and {Ak } is the corresponding system of eigenvalues. Choose 2 and ~ in M3 lying above x and y in X. Then
(3)
Kx(x,t,y) : E KM~(2,t,7~). yEF
View the pair (x, y) as fixed, and for each 7 C F define logq-y(x, y) = logq7 = dM3(~,TY).
114
Thus we obtain a sequence {q~}eEr, which we reindex by itself as simply {q}. Suppose x # y. Then by (1), we have (4)
Kx(x,t,y)-
2 --t e--(log q)Z/at (4~r~)3/2 ~ q _lo.__g_q_ q-1 q
which we call the F - e x p a n s i o n or the q - e x p a n s i o n . Combining (2) and (4), we immediately obtain the theta inversion formula OO
k----0
(5)
2
-
(4~rt)3/2
e- t
E q _lo_g.q q-1 q
e_(log q)2/4t.
Observe that the factor e -t could be brought to the other side. In the notation of w cr0 - 2 and #k = Ak -- 1. We note that (5) holds for any points x ~ y on X, and we have not taken a trace of the heat kernel, although we have taken the trace with respect to the infinite Galois group F, i.e. the fundamental group of X. Formula (5) simply reflects a combination of the existence and uniqueness of the heat kernel on X together with the spectral expression (2) and the group expression (3) for K x . As in the previous sections, any theta inversion formula can be used to obtain a zeta function with additive functional equation. Let us carry through the computations in this case. We are still assuming x ~ y. From (5), let us derive two expressions for the function OO
(6)
Fx,y(s) = (2s
-
2) / Kx(x,t,y)e-s(s-2)tt2dt. 0
Using the spectral expansion (2), the integral in (6) yields the equality (7)
F,,y(s) = (2.s - 2) E
k----0
(s(s - 2) + Ak)a'
which can be shown to converge uniformly and absolutely on X by combining standard asymptotic formulas for eigenvalues on compact manifolds (see [BGV 92], for example) and Sogge's theorem on sup-norm bounds for eigenfunctions (see page 226 of [St 90]).
115
On the other hand, we shall also take the Gauss transform of the right side of (5), for which we need the collapse of the Bessel integral OO
K3/2(s - 1,log q)= f e -(s-1)2te -(l~ q)2/4tl~3/2d_.~ 0
--
2s
-
2 0s
q-(~-l)
s
)
by Lemma 2.1. Therefore, taking the Gauss transform of (3), or the right side of (5), we obtain a Bessel sum for Fx,v which collapses to a Dirichlet series, for Re[(s - 1)2] > 0, namely 2 F,,~(s) - (4~r)3/2 E
q_l~
(2s - 2)K3/2(s - 1, log q)
q
which becomes
(s)
1 a ( -1 logq -s~ F',y(s)-2rOs 2s--2E1_q-2 q )" q
Using a simple argument involving volumes of fundamental domains, one can show directly that the Dirichlet series (8) converges for Re(s) > 2. Therefore, we have for s in this half-plane, the formula 1 0 ( Fx,v(s)-27rO s
-1
logq
)
2s_2 E 1Zq-2q-" q
oo
= (2~
-
2r162
2) Z (~(~ 21 + ~1~ _
k----O
From (7) it follows that the Dirichlet series in (8) has a meromorphic continuation to all s E C with the additive functional equation
F,,~(~)=-r,,~(2-~). Further, the meromorphic continuation has singularities precisely when s(s - 2) = --Ak, that is (9)
s ---- 1 :k iL/Ak -- 1
116
each singularity being a double pole with zero residue. U p o n integrating with respect to s, we obtain the Dirichlet series (10)
$Ix'y(S) = E
1 q-S 1_ q-2 = E E
q
q-S-2k,
q k=O
which satisfies the relation ~Ds_l JItz,y(8) -~ -27rFx,y(._q ).
The function JI = dI:~,y plays the role of the logarithm of a zeta function in the f u n d a m e n t a l class, and would be such a l o g a r i t h m if the residues of JI ~ were integers, but they are not in general. Theorem i) ii) iii) iv)
4.1. The function dI has the following properties:
The series for dI(s) converges for Re(s) > 2; JI ~ has a meromorphic continuation to edl s C C; dI ~ is odd u n d e r the m a p s ~ 2 - s; The singularities of the meromorphic continuation of the function ( 2 s - 2 ) d I ' ( s ) are all simple poles at the points (9), with corresponding residue 2~rr162
P r o p e r t y (iv) should be viewed as a type of R i e m a n n hypothesis for the zeta function (10). We conclude with some remarks in the case x = y. We pick = #. T h e n the sum in (1) must be d e c o m p o s e d into the t e r m with 7 = id and the other terms. We have qid = 1 and q.y > 1 for all 7 ~ id. In the present case, the t e r m with 7 = id is easily computed, and instead of (4) we t h e n obtain
(11)
Kx(x,t,x)
1
-- (4rot)3~2 e
-t +
2 (47rt)3/2
e-t E q
log q q _ q-1
e-(l~
q)2/4 t
where the sum is over q = q'r for 7 # id. Thus the identity t e r m must be h a n d l e d separately. The integral transformation considered in (6) is sufficiently simple so t h a t we have oo
2 s - 2 / e_(~_l)htl/2dt 1 (4-zr)3]~ t --4Ir" 0
117 As a result, when considering the series (10) for x = y, summing over q.~(x, x) = dM~ (x, "y:~) with -y r id, all properties (i) through (iv) hold with property (iii) being changed to allow an additive factor which is a polynomial of degree three in s. The coefficients of this polynomial can be determined by considering the limit as s ~ oc and using the asymptotic formulas from Chapter I. R e m a r k 1. Computations similar to the above hold for compact quotients of odd dimensional hyperbolic spaces, since, for these manifolds, one can obtain simple expressions similar to (1) for their heat kernels; see page 151 of [Ch 84]. For non-compact quotients, one must take into account the appearance of the Eisenstein series in the spectral decomposition of the heat kernels, and the subsequent appearance of other terms in the additive functional equation. Examples of such formulas, as well as the more complicated situation of even dimensional hyperbolic manifolds, will be treated in a future publication. R e m a r k 2. Now that we have constructed a zeta function with additive functional equation and a Dirichlet series, we can apply the methods of [JoL 93c] to study the function (11)
~
Cn(X)r
z'iv/-~-I ,
n~0
defined for Ira(z) > O. As in [JoL 93c1, there is a meromorphic continuation of (11) to complex z with singularities at the points {+ log q}. This should be viewed as a Duistermaat-Guillemin type theorem, as in [DG 75], since (11) is a wave kernel. We shall deal with this situation at greater length elsewhere, especially since it requires a systematic exposition of the additive fundamental class and its corresponding Cram6r theorem and explicit formulas. R e m a r k 3. The above example shows the necessity of an "additive class" of zeta functions, as discussed in the introduction, by which we mean functions which are meromorphic with simple poles, have Dirichlet series in a half-plane, and have an additive functional equation with additive fudge factors expressible as linear combinations of regularized harmonic series types. The above example is not expressible in the fundamental multiplicative class since the residues, as determined in (iv) above, will certainly not be integers.
118 R e m a r k 4. Readers will note the distinction between the zeta function defined by Minakshisundaram-Pleijel in IMP 49] by the Dirichlet series OO
r162
and the new zeta function JI~,y.I The Minakshisundaram-Pleijel zeta function is essentially the one obtained as the Mellin transform of the theta series satisfying the basic theta conditions, whereas the new zeta function is obtained as the Gauss transform of the theta series.
CHAPTER
VI
A g e n e r a l i z a t i o n of Fujii's t h e o r e m Let L~ denote the set of zeros of the Riemann zeta function with positive imaginary part, meaning L~={pEC:fQ(p)=0
and
Ira(p) > 0 } .
Write p = fl + i7 for any p C L~. The zeta function
v(p)
(1)
,(p)
defined for Re(s) > 1, was studied in [Del 66] and [Gu 45], and it was shown that (1) admits a meromorphic continuation to all s E C with explicitly computable singularities, including a double pole at s = 1. Building on these results, Fujii considered the zeta functions
,(p) sin(o~)
(2) pEL +
and
~ ,(p) cos(a')') 7s
(3)
pEL +
for non-zero a e R and Re(s) > 1. It was shown in [Fu 831 that (2) admits a holomorphic continuation to all s E C for any non-zero a, and (3) admits a mermorphic continuation to all s C C with a simple pole at s = 1 having residue equal to -(27r)-lA(e~)e -a/2, where
(4)
A(x)
S logp,
t
0,
if x = p k w h e r e p i s a p r i m e a n d k C Z > 0 otherwise.
120
Instead of functions formed separately with sine and cosine, one may as well consider what we call the Fujil f u n c t i o n eiC~7
=
v(p) ~/--T-
and its meromorphic continuation. We should note that Fujii's proof of the meromorphic continuation involves a very detailed study of many integrals arising from a generalization of Delsarte's work [Del 66] involving various integral transforms of the classical Riemann-von Mangoldt formula. As remarked on page 233 of [Fu 83], one can prove analogous results for the eigenvalues of the Laplace-Beltrami operator on the fundamental domain of the modular group PSL(2, Z). These theorems are given in [Fu 84b] and are as follows. Let {Ak} be the set of eigenvalues of the hyperbolic Laplacian on the space PSL(2, Z)\h and set Aj = 1/4 + rj with rj > 0. For any a E R +, Fujii considered the function sin(arj) ,
U
3
which is defined for Re(s) > 2. Through a rather lengthy and involved application of the Selberg trace formula, it was proved in [Fu 84b] that (5) has an analytic continuation to all s E C to a holomorphic function. Again, one could consider the Fujii function
rj
for all o~ E R and study its meromorphic continuation, thus capturing many of the results obtained by Fujii in the papers [Fu 84b] and [Fu 88a]. At this point, one could envision a series of articles in which one would define and study a Fujii function associated to every special zeta function, such as zeta functions and L-functions from the theory of modular forms, zeta functions and L-functions of number fields, spectral zeta functions constructed from the eigenvalues
121 associated to the Laplacian acting on any non-compact arithmetic Riemann surface, to name a few examples, with each example yielding a new paper. Such a case-by-case extension of the classical Cram6r theorem [Cr 19] has begun to appear in the literature. However, in [JoL 93c], we gave a vast generalization of Cram6r's theorem containing all previously known special cases and many more. Similarly, in this chapter, we obtain a generalization of Fujii's theorem which applies to any zeta function with Euler sum and functional equation whose fudge factors are of regularized product type. This generalization is simply a corollary of the generalized Cram6r theorem, and, in particular, applies both to the zeta functions arising from algebraic number theory and to those arising from spectral theory. In w we will state the generalized Fujii theorem, and the proof will be given in w To conclude this chapter, we will give various examples of the generalized Fujii theorem in w
122
w
S t a t e m e n t o f t h e g e n e r a l i z e d Fujii t h e o r e m .
Let us assume the notation of the previous chapters. With this, we can state the following result, which we call the generalized Fujii t h e o r e m . T h e o r e m 1.1. Let (Z, 2 , ~) be in the f u n d a m e n t a l class. Let a be such that ao + a > a'o, and let {p} be the set of zeros and poles of Z in the open infinite rectangle ~ a with vertices at the four points - a + ioo,
-a,
ao + a,
ao + a + icxD.
Let v(p) = ordpZ and set {A} = { p / i } . Then:
i) For any non-zero a E It, the Fujii function eia)~
Fz, o(s;
v(p)
= pE'R=
A8
has a m e r o m o r p h i c continuation to all s E C.
ii) For any non-zero a E R , the continuation of the Fujii function Fz,a(s; a) is holomorphicfor aJ1 s E C except for simple pole at s = 1 with residue
--(27/') lc(q) log q
if a = log q
-- (2~') -1C(~)(log ~)~--ao
if a = - log
0
otherwise.
In [Fu 83], Fujii considered the zeta function formed with the imaginary parts of the non-trivial zeros of the Riemann zeta function, assuming the Riemann hypothesis. The following theorem generalizes this result. T h e o r e m 1.2. W i t h notation as in Theorem 1.1, assume there is a real constant flo such that p = flo + i7 for all p C ~ a . Then: i) For any non-zero a C R , the Fujii function e i oL"y
RH
v(p) V
Fio (s; pE'R=
123 has a m e r o m o r p h i c c o n t i n u a t i o n to M1 s C C.
ii) For a n y non-zero a E R , the c o n t i n u a t i o n o f the Fajii function F~,~ RH (s; a) is h o l o m o r p h i c for all s C C except for simple pole at s = 1 w i t h residue
--(27r) lc(q)(log q)q-~O
if a = log q
--(27r) -1 c(q)(log cl)q -a~176
if a = - log
0
otherwise.
One obtains the Fujii theorem by considering the above series with a and - a since, for example, if Z = ~Q, we have
pEL+
7s
Also, as remarked on page 23 of [Fu 84a], we obtain a meromorphic continuation of the series
cos( 7)
2
-
Fz,a(s;
+
pEL~
both with and without a R i e m a n n hypothesis type assumption. Thus what appeared up to now to be a p h e n o m e n o m associated to more or less arithmetic situations, for instance the location of poles at the logs of prime powers or their analogues for the Selberg zeta function, is now seen to be quite a general property of our broad class of functions. Directly from Theorem 1.2, we have the following corollary. C o r o l l a r y 1.3. In addition to the above conditions, a s s u m e Z = Z , a n d a s s u m e all coet~cients c(q) are reM. I f all zeros o f Z in Tea lie on a v e r t i c a / l i n e Re(s) =/30, then we necessarily have =
o/2.
Finally, let us note t h a t the case of a = 0 is handled by our Cram@r theorem, specifically Corollary 1.3 of [JoL 93c], and our results from [JoL 93a], specifically Theorem 1.8 and Corollary 1.10. For completeness, let us list this theorem and refer to the above mentioned references in our work for a proof.
124
T h e o r e m 1.4. W i t h notation as in Theorem 1.1, assume that 42 has reduced order ( M , rn ). Then the zeta function
v(p)
1
pE'R.=
which is defined for Re(s) > M + 1, has a m e r o m o r p h i c continuation to all s E C. If there is a constant/3o such that p =/3o + i7 for MI p E 7~a, then the zeta function
pERa
which is defined for Re(s) > M + 1, has a m e r o m o r p h i c continuation to M1 s E C.
The asymptotic expansion in w of [JoL 93c] and Corollary 1.10 of [JoL 93a] combine to give an explicit description of the poles of the zeta functions in Theorem 1.4. We will not state these results here, but will simply remark that the location and order of poles of the zeta functions in Theorem 1.4 are determined by the asymptotic expansion near t = 0 of the theta function associated to the fudge factor 42.
125
w
P r o o f of F u j i i ' s t h e o r e m .
Let z = a + it for any non-zero a E R, and, with notation as above, let p = iA. In [JoL 93c] we proved various analytic properties of the Cram~r function
Vz(z) = ~ , v(p)~-, pE~
which is defined when Im(z) > 0. In particular, Theorem 1.1 of [JoL 93c] and subsequent discussion imply that the function aoWa
(1)
2~iVz,a(z) - ~o~
f
~-sZ~,l~(~)d ~
ao+a--ioo
has a meromorphic continuation to all z E C, whose only singularities are simple poles at the points log q and - l o g q. The residues of these poles are given on page 390 of [JoL 93c]. Now assume that 9 is of regularized product type of reduced order (M, m). By combining Lemma 3.1, Proposition 3.2, Lemma 3.3, and, quite importantly, Lemma 4.2 of [JoL 93c], we conclude that the integral in (1) has a holomorphic continuation to any nonzero z E R. Therefore, for a ~ 0, the function
v.,z,o(t)
=
vz, o(.
+
it)=
~
v(p)ei"*e-'~
pET~
has an asymptotic behavior of the form oO
V~,z,a(t) "~ E
(2)
cn(a)tn
as t ~ O,
n-~--I
for some constants ca(a) which depend on a. Further, from the formula in Theorem 1.1 in [JoL 93c], we have
(3)
C _ I (Ot~) =
-(27r) 1c(q) log q
if a = log q
-- (27r)- lc(~1) (log ~1)~1-~~
if a = - log Cl
0
otherwise.
126
By applying the Mellin transform, we conclude from (2) that the function O0
eie~ s = / VZ,a(~ + it)t" v(p)
r(,) ~ PE?~-a
= r(,)MV~,z,a(,)
0
has a meromorphic continuation to all s E C whose only singularities are simple poles at the points s E Z<0 and s = 1 (see Theorem 1.5 of [JoL 93a]). Since F(s) has simple poles at these points s E Z<0, Theorem 1.1 follows. Let us now assume that any zero or pole p of Z in Tea is such that Re(p) =/30, for fixed fl0. Then we can write p = /30 + iv, so A=7-ifl0. Withz=a+it, wehave eP z ~ eaflo eiflo t . e i ' Y a e - t ' Y .
Therefore, the function
E
(4)
v(p)ei'la e-'~t = e-a~~176
pE~,=
has asymptotics as in (2). With this, the proof of Theorem 1.2(i) is completed by applying the Mellin transform and the argument given above. Finally, from (3) and (4), one has the asymptotic formula
.(p)~-~*
~ ~ - ~ 0 c _ l ( ~ ) t -1
as t -~ 0,
pE~a
which completes the proof of Theorem 1.2(ii). The proof of Corollary 1.3 is as follows. Since all numbers 7 are real, we have
v(p) cos(~7) + i 2_., vkP),sin(aT)
FsR H (s; ~) = ~ pE?~=
7*
pE'R~=
7*
If we assume that all numbers c(q) and q are real, then the residue of the only possible pole of F RZ H I\s , 9a ) is real, hence the series
v(p) sin(o~) 7* pET~=
127
h a s a h o l o m o r p h i c c o n t i n u a t i o n to all s E C. H o w e v e r , we c a n also e x p r e s s this series as
2i ~ v(p)
eio~7
= F sRH (~; ,~) - F ; R, a.
(s;-,~),
pETr
w h i c h m e a n s t h a t t h e r e s i d u e s at s = 1 n e c e s s a r i l y cancel, h e n c e ~0 = cro/2.
128
w
Examples.
As before, specific applications of our general theorems yield several classical theorems, many recent results, and new applications. We list a few examples here. E x a m p l e 1: T h e sine f u n c t i o n . Let Z = sin(wis). In this case, the associated Fujii function is the Dirichlet-like L-function eian
(1)
Fz(s;o~) = ~
ns
n----1
since {q} = {e 2~n} and c(e 2"~n) = 1/n. Theorem 1.2 states that the series (1) has a holomorphic continuation whenever a r 2~rn for some n C Z, and when c~ = 2~rn, the series (1) has a simple pole at s = 1 with residue 1. E x a m p l e 2: T h e R i e m a n n z e t a f u n c t i o n . Let Z = ~Q. If we apply Theorem 1.2 to the Riemann zeta function, assuming the Riemann hypothesis, we obtain the Fujii theorem from [Fu 83]. However, without assuming the Riemann hypothesis, we do have the following result. Let S denote the non-triviM zeros of the Riemann zeta function in the upper half plane, and let c C C. Then the function ~-'~ e(P+{C) z : e icz ~ pES
e pz
pES
satisfies the asymptotic axiom AS 2. Therefore, the proof of Theorem 1.1 implies that for non-zero a E R, the function
(2)
eap/i
pES
(pli)+c)s
has a meromorphic continuation to all s C C. The only singularity of the continuation of (2) is a simple pole at ~ -- 1 and then only when a -- =t=log p'~ where p is a prime. Without any modification, the above argument applies to Dirichlet L-functions and L-functions of number fields. A list of zeta functions for which the Cram~r theorem of [JoL 93c] holds is given in section 7 of [JoL 93c].
129 E x a m p l e 3: Selberg zeta functions of c o m p a c t R i e m a n n surfaces. If Z is the Selberg zeta function associated to a compact Riemann surface, then the theta function coming from the Cram@r theorem is
~_, e(1/2+ix/~,j-1/4)z
where Aj is an eigenvalue of the Laplacian. Hence, this theta function can be viewed as a type of trace of the wave operator. E x a m p l e 4: Selberg zeta functions of n o n - c o m p a c t Riem a n n surfaces. As in [Fu 84b], consider the Selberg zeta function associated to PSL(2, Z ) \ h . It is known that the set of zeros of the Selberg zeta function is the union of two sets: One set being the eigenvalues of the Laplacian, as described above, and the other set associated to the zeros of the Riemann zeta function (see pages 498 and 508 of [He 83]). From the meromorphy of (2), we conclude that the Fujii function formed with the eigenvalues of the Laplacian acting on PSL(2, Z ) \ h has a meromorphic continuation to all C. This theorem is the main result of [Fu 84b]. Similarly, since the scattering determinant for any congruence group is expressible in terms of Dirichlet L-series (see [He 83]), the above argument applies to yield the analogue of the Fujii theorem in these cases. The case of a general non-compact hyperbolic Riemann surface, including those associated to non-congruence subgroups, will be considered in [JoL 94].
BIBLIOGRAPHY
[Ba 81] BARNER, K.: On Weil's explicit formula, d. reine angew. Math. 323, 139-152 (1981). [Ba 90] BARNER, K.: Einfiihrung in die Analytische Zahlentheotie. Preprint (1990). [BGV 92] BERLINE, N., GETZLER, E., and VERGNE, M.: Heat Kernels and Dirac Operators, Grundlehren der mathematicschen Wissenschaften 298 , New York: Springer-Verlag (1992). [CoG 93] CONREY, J.B., and GHOSH, A.: On the Selberg class of Dirichlet series: small weights. Duke Math. g. 72 (1993), 673-693. [Cr 19] CRAMI~R, H.: Studien ~iber die Nullstellen der Riemannschen Zetafunktion. Math. Z. 4 (1919), 104-130.
[DGM 76] DEBIARD, A., GAVEAU, B., and MAZET, E.: Th6or~ms de Comparison en G6ometrie Riemannienne. Publ. R I M S Kyoto Univ. 12 (1976) 391-425. [Del 66] DELSARTE, J.: Formules de Poisson avec reste. J. Analyse Math. 17 (1966) 419-431. [Den 92] DENINGER, C.: LocM L-factors of motives and regularized products. Invent. Math. 107, (1992) 135-150. [Den 93] DENINGER, C.: Lefschetz trace formulas and explicit formulas in analytic number theory. J. reine angew. Math.
441 (1993) 1-15. [DG 75] DUISTERMAAT, J., and GUILLEMIN, V.: The spectrum of positive elliptic operators and periodic bicharacteristics. Invent. Math. 29~ (1975) 39-79.
[Fu S3] FUJII, A.: The zeros of the Riemann zeta function and Gibbs's phenomenon. Comment. Math. Univ. St. Paul 32 (1983), 99-113. [Fu 84a] FUJII, A.: Zeros, eigenvalues, and arithmetic. Proc. Japan Acad. 60 Ser. A. (1984), 22-25.
132 [Fu 84b] FUJII, A.: A zeta function connected with the eigenvalues of the Laplace-Beltrami operator on the fundamental domain of the modular group. Nagoya Math. J. 96 (1984) 167-174. [Fu 88a] FUJII, A.: Arithmetic of some zeta function connected with the eigenvalues of the Laplace-Beltrami operator. Adv. Studies in Pure Math. 13 (1988) 237-260. [Fu 88b] FUJII, A.: Some generalizations of Chebyshev's conjecture. Proc. Japan Acad. 64 Set. A. (1988) 260-263.
[Fu 93]
FUJII, A.: Eigenvalues of the Laplace-Beltrami operator and the von-Mangoldt function. Proc. Japan Acad. 69 Ser. A. (1993), 125-130.
[Gal 84] GALLAGHER, P. X.: Applications of Guinand's formula. pp 135-157, volume 70 of Progress in Mathematics Boston: Birkhauser (1984). [Gu 45] GUINAND, A. P.: A summation formula in the theory of prime numbers. Proc. London Math. Soc (2) 50 (1945) 107-119.
[Hej 76]
HEJHAL, D.: The SeIberg Trace Formula for PSL(2, R), volume 1. Lecture Notes in Mathematics vol. 548 New York: Springer-Verlag (1976).
[Hej S3]
HEJHAL, D.:The Selberg trace formula for PSL(2, R), vol. 2. Springer Lecture Notes in Mathematics 1001 (1983).
[In 32] INGHAM, A. E.: The Distribution of Prime Numbers, Cambridge University Press, Cambridge, (1932).
[JoL 93a] JORGENSON, J., and LANG, S.: Complex analytic properties of regularized products and series. Springer Lecture Notes in Mathematics 1564 (1993), 1-88. [JoL 93b] JORGENSON, J., and LANG, S.: A Parseval formula for functions with a singular asymptotic expansion at the origin. Springer Lecture Notes in Mathematics 1564 (1993), 1-88.
[JoL 93c] JORGENSON, J., and LANG, S.: On Cram6r's theorem for general Euler products with functional equation. Math. Ann. 297 (1993), 383-416.
133
[JoL 94] JORGENSON, J., and LANG, S.: Applications of explicit formulas to scattering determinants of finite volume hyperbolic Riemann surfaces. In preparation. [Kur 91] KUROKAWA, N.: Multiple sine functions and Selberg zeta functions. Proc. Japan Acad., Set A 67, (1991) 61-64. [La 70] LANG, S.: Algebraic Number Theory, Menlo Park, CA.: Addison-Wesley (1970), reprinted as Graduate Texts in Mathematics 110, New York: Springer-Verlag (1986); third edition, Springer-Verlag (1994). [La 87] LANG, S.: Elliptic Functions, second edition. Graduate Texts in Mathematics 112 New York: Springer-Verlag (1987).
[La 93a] LANG, S.:
Complex Analysis, Graduate Texts in Mathematics 103, New York: Springer-Verlag (1985), Third Edition (1993).
[La 93b] LANG, S.: Real and Functional Analysis, Third Edition, New York: Springer-Verlag (1993). [Lay 93] LAVRIK, A. F.: Arithmetic equivalents of functional equations of Riemann type. Proc. StekIov Inst. Math. 2 (1993) 237-245. [MiP 49] MINAKSHISUNDARAM, S., and PLEIJEL, A.: Some properties of the eigenfunctions of the Laplace operator on Riemannian manifolds. Canadian Jour. Math. 1 (1949) 242-256. [Mo 76] MORENO, C. J.: Explicit formulas in automorphic forms. Lecture Notes in Mathematics 626 Springer-Verlag (1976). [Se 56] SELBERG, A.: Harmonic analysis and discontinous groups in weakly symmetric Riemannian spaces with applications to Dirichlet series, J. Indian Math. Soc. B. 20 (1956) 47-87 (Collected papers volume I, Springer-Verlag (1989) 423-463). [Sel 91] SELBERG, A.: Old and new conjectures and results about a class of Dirichlet series. Collected papers volume II, Springer-Verlag: New York (1991) 47-63.
134
[st 90] STRICHARTZ, R. S.: Book review of Heat Kernels and Spectral Theory, by E. B. Davies, Bull. Amer. Math. Soc. 23 (1990) 222-227.
[Ti 48] TITCHMARSH, E. C.: Introduction to the Theory of Fourier-Integrals, 2nd Edition Oxford University Press, Oxford (1948).
[Ve 78a]
VENKOV, A. B.: A formula for the Chebyshev psi function, Math. Notes of USSR 23 (1978) 271-274.
[Ve 78b] VENKOV, A. B.: Selberg's trace formula for the Hecke operator generated by an involution, and the eigenvalues of the Laplace-Beltrami operator on the fundamental domain of the modular group PSL(2, Z ) . Math. USSR Izv. 42
(1978) 448-462. [Ve 81] VENKOV, A. B.: RemMnder term in the Weyl-Selberg asymptotic formula. Y. Soviet Math. 17 (1981) 2083-2097. [We 52] WEIL, A.: Sur les "formules explicites" de la th6orie des nombres premiers, Comm. Lund (vol. d6di@ s Marcel Riesz), 252-265 (1952). [We 72] WEIL, A.: Sur les formules explicites de la th@orie des nombres, Izv. Mat. Nauk (Ser. Mat.) 36, 3-18 (1972).
A SPECTRAL INTERPRETATION OF W E I L ' S E X P L I C I T F O R M U L A
Dorian Goldfeld
w
Introduction:
The classical theory of automorphic functions over Q is based on Selberg's spectral decomposition of the space s
Q)\GL(2, A))
where A denotes the adele group of Q. It is well known that this space has an orthogonal decomposition into cusp forms, Eisenstein series, and residues of Eisenstein series. Further, each of these basic functions has a Fourier expansion which defines a canonical zeta function satisfying properties similar to the classical Riemann zeta function. The Rankin-Selberg [It], [Sl] convolution of an automorphic form for GL(2,Q) yields a new zeta function associated to an automorphic function on GL(3, Q). This is the Gelbart-Jacquet [G-J] lift. By studying the spectral decomposition of a kernel function on GL(2,Q), Selberg [$2] went considerably further and found the trace formula. This led to the discovery of the Selberg zeta function whose explicit formula is precisely the trace formula. The zeros of the Selberg zeta function determine the discrete spectrum of the Laplacian on the space L:2(GL(2, Q)\GL(2, A)). The fact that the Laplacian is a self adjoint operator gave the analogue of the Riemann hypothesis for the Selberg zeta function. The classical theory of automorphic forms is based on the spectral properties of the group GL(2,Q) acting on GL(2,A). We think of this group as generated by additions and one inversion. The object of this paper is to show that an analogue of the classical theory of automorphic forms exists for a group defined over Q generated by Research s u p p o r t e d in p a r t by NSF grant no. DMS 9003907
138
multiplications and one inversion which acts on A x , the idele group over Q. The space of 1;2 functions on the factor space has an orthogonal decomposition into cusp forms, Eisenstein series, and residues of Eisenstein series. Each of these basic functions has a Mellin expansion which defines a canonical zeta function, and there is a generalization of the Rankin-Selberg method and Gelbart-Jacquet lift. The trace formula for this space is precisely the explicit formula of A. Weil [W], and the discrete spectrum of the Laplacian is given by the zeros of the Riemann zeta function which is just the Selberg zeta function for this space. Our method gives the first interpretation of the integral involving the logarithmic derivative of the gamma function (in Weil's explicit formula) as an integral over the continuous spectrum of a Laplace operator. The Riemann hypothesis remains unproven, however, since it is not known that the analogue of the Petersson inner product for this space is positive definite on the cuspidal spectrum. In fact, the Riemann hypothesis is equivalent to the fact that the space generated by the cusp forms is positive definite. All that we are able to show at present is that this is an indefinite space. If the Riemann zeta function has a zero off the line Re(s) = 1 then the cusp form associated to this zero will have norm zero. Manin [M] asks if there exists a category where one can define "absolute Descartes powers," SpecZ x . - . x S p e c Z . Following Kurokawa [K], he shows that at least the zeta function of such an object can be defined which agrees with Deninger's [D1-2] representation of zeta functions as regularized infinite determinants. As in Marlin [M], we define a left directed family to be a set A of complex numbers A (which are discrete in C and where each A occurs with multiplicity rn~ E C) which satisfies (1) V r C R , (2) 3/~ > 0,
Card{AEAIRe(A)>r s.t.
E
Im~l =
} < oc,
o(g #)
as ( H ~ ~ ) .
AEA
Re(A)>-H
The tensor product A1 | A2 of two left directed families is defined to be A1 | A2 ={A='~lq-/~2
I~1 c A 1 ' A2 c A 2 ' m)~=
E AI+A2=A
maim)'2}"
139
Let A, A' be two left directed families and let s be a complex variable. The Kurokawa tensor product | of the products associated Kur
with the directed families is defined by the formula
II
~EA
II
Xt EAt
AEA|
I
By taking logarithmic derivatives this lifts to what we shall call a Kurokawa sum
)tEA
)t I E A I
AEA|
I
The author finds it remarkable that this sum which is expected to be associated to the tensor products of "motives," should be precisely the Rankin-Selberg convolution as exemplified in w of this paper.
w
Notation:
For a rational prime p let lip denote the p-adic valuation on Q normalized so that [pip = p-1. Let Qp, Zp, Up be the p-adic completion of Q with respect to [[p, the p-adic integers, and the p-adic units, respectively. If xp E Qp ' we set e(xp) = los(l~rlr) the log p ' exponent of xp and
s(xp)
= ~(~P) the sign of xp. I~(xp)l' Let A denote the group of adeles over Q, and let A x denote the group of ideles over Q where each x E A• is of the form x = (xoo,x2,x3,..., xp,... ) with xo~ E R, xp E Qp, and xp E Up for almost all primes p. For a subset B C Qp define
Ixlp [XplB =
0
for xp E B for xp ~ B.
We also set Ilxll = Hv Ixvlv to be the norm on QX where Ixlv = Ixl, the ordinary absolute value when v = oo. If we consider Q• embedded diagonally in A • then Q• acts on A• by multiplication. For a E QX, x E A x define this action by OlX :
(O[.Xoo, O~X2, 0 / . X 3 , . . . ).
140
Let w denote the involution wx
= x -I,
q5 = < QX,w
and define the group >
as the group generated by Qx and w. Then ~ acts on A x and may be realized as a matrix group generated by {(0
and
01)(~CQ}
(01)0 " Zdeoo o ho oo e o O, naao oe ho
gr~176176176
( ~
or
1
0
w
C o n s t r u c t i o n of the indefinite space s
01)
"
A function f : ]R ---+ C is said to be a symmetric Schwartz function if it is smooth and satisfies the conditions f(x) = f(1/x), xm dxm d f(z)
f(x) = f(-x), = 0(1)
(for all m , n > O, n e Z).
Let S denote the space of symmetric Schwartz functions. For f E S the Mellin transform of f is given by /(A) =
f ( x ) x ) ` dx
~0~176
X
so that f ( x ) is given by the inverse Mellin transform
1 [~+'~176 f(x)
for suitable a.
-
2~i
~-~oo
x '
141
We now construct a vector space T over C with an indefinite inner product. Every element of T will be a function F : F \ A x ~ C of the form
F(x) = ~ ( l o g p ) ~ p
[o~xv Q~ - 89-vp
f(ax~),
"
o~EP
where f : R --~ C is a symmetric Schwartz function. computation shows that
p
n=l
A simple
--
We shall say that F is associated to f. Given F, G E T associated to f, g, respectively, we define an inner product f
< F, G > =
/
Jr
\A•
F(x)g([[x[[)
dz
Ilxll"
After some computations we obtain
=
Jfr
F(x)g(llxll)
4zri J2-i~o
(
dx
Ilxll + A
j~(A)~(A) dA.
It immediately follows that < F, G > = < G, F >, and thus the inner product endows T with the structure of an indefinite inner product space.
w Spectral theory of s Let v be a place of Q and let ~ E Qv. Consider the multiplication operator Me,v = Me, where Me : T -~ T is given by 1
MeF(x) = F ( ( x ~ , x 2 , . . . ,Zxv,... )) + F ( ( x ~ , x 2 , . . . , ~ x v , . . . )).
142
Clearly < MtF, G > = < F, M t G > so the collection of Mt form a commuting system of symmetric operators. Each of these operators commute with the Laplacian
We shall show that each operator M t is a bounded operator on /~2(T) with discrete spectrum
I~lv~ + I~lv p, where i1 + p is a zero or pole of the Riemann zeta function. Further, M~ has a continuous spectrum
I~1~ + I~l~-~, with A E iR. Similarly, the Laplacian A is an unbounded operator on s with discrete spectrum p2 + p - 2 ,
where ~1 + p is a zero or pole of the Riemann zeta function. The Laplacian A also has a continuous spectrum A2 + A-2, with A C ill(. In analogy with the Selberg theory for SL(2, R) the space ~2(T) decomposes into cusp forms, Eisenstein series, and residues of Eisenstein series.
w Eisenstein series: Formally, we define the Eisenstein series as the element of T associated to the symmetric function ]lxll* + ][xl[-s, which is given in the form
E(x, ~) : ~-~(logp) ~ ~ p
(P"llxll) s + (p-"ll~lD s
n=l
+ (p~llxll) -~ + (p-"llxll) -~] _ ~'
+
-~]
143
T h e only problem is t h a t Ilxll , + Ilxll -~ is not Schwartz, and hence, the above series does not formally converge. T h e problem can be circumvented by defining the Eisenstein series as a distribution. Set
r =
s/ 1
-
F'
= - log~- + ~ -
1
F' ~
+
I
+ ~m
-
s)
We have the identity
~(~)(~+ _,)= I [~'+~ 4~'i ,, a - i o o
i
~( ~)
~ +-----~ i - - ~
)(x~+x_~,)dA,
which is valid for 0 < a < 89 a > IRe(s)l. Let S denote the test function space on F \ A x , and define K to be the space of all s m o o t h functions r \ A • - ~ C.
T h e n for g E S, G E l:2(T), and F C K, we consider the continuous linear functional
< F,G > = ~
F(x)g(llxll) dx IIxll'
\Ax
which defines F as a distribution. In this m a n n e r for g E S and G E T associated to g define dx
<
E,G
>
=
fr\Ax E(x,~)g(llxll)
Ilxll
I [~+~~176('(1 4~rij~,_~
-('2
+A)s--~+
which is valid for 0 < a < 89 a > a distribution.
)
r (,2-I -s+A
]Re(s)l. This
9 ~(A) dA,
defines
E(x,s)
as
144
Now, for f E S, define
EF(X) = 1 [+iooE ( x , s ) f ( s ) d s , 47ri ., - i ~
the projection of F onto the space of Eisenstein series. Then EF(x)=O(1),
so that EF is in s
We compute
< EF, G > -
1
9 :
-
1)
- s +--s + s 47ril ., -ioo[+i~f ( s ) . [ ~ + i ~ ~ ( ~ ) . ( -s + o'--io0
~(s163
~(~)](~)~(~) d~,
2hi ~ ~ - i ~
which is valid for 0 < a < 7" 1 It follows that
< EF, G > = < F, EG > = < E G , F > . 1, a > iRe(s)l,
In the same manner, forO < a <
< EF,E>
-
1 / "+~~176 2~ri.,~_ioo
(1 s+,k
1) -s+,k
=/(~>(s) = <E, EF>. It follows that
<(F-EF),E>=
O,
< E, ( F - EF) > = O, and finally
< ( G - Ea), ( F - EF) > = < G , ( F - E F )
>
= <(G-EG),F> = < ( F - EF),(G - EG) >.
145
The above computations establish the fact that F - E F lies in the orthogonal complement of the space of Eisenstein series. To recapitulate, the Eisenstein series
E(x,s)
=
[
-logTr+
lr' ( 89 ~-2
1F' (~___~)] + 5~ (11=11=+ Ilxll-*)'
may be defined as a distribution on s
which satisfies
Ae(x,s)=s2E(x,s),
(11~11"+ IlZll -~)
M~E(x, ~) =
E(x, ~),
and, hence, determines the continuous spectrum of these operators. 1 1 3 The Eisenstein series has poles at s - 2, 2, 2, 32, 52, 52, "'" with residues given by a constant multiple of
IIx[I ~ + Ilxll -~, for
w
n -
1 2,
1 2,
3 2,
3 2,
5 2,
5 2,'"
Cusp Forms:
Let ~(s) = ~--~ F (2) C(s) 9The functional equation of the zeta function may be written in the form
-#-(s) ---- ~(1--s). Let f be a symmetric Schwartz function. A consequence of the functional equation of the zeta function is the well known explicit formula
~ . (." + =-.) i(p) = 2~i ~ _ , ~
r
+ ~) (x~ + ~-~') ](~)d~
+ ~.1 [ ~ + ~ _ [~' 1 ~' ~-,~ ~(~ + :,) + ~(~1
A)] x ~ ](~)d:~,
146
1 T h e sum on the left in the above formula runs over where c > 7" complex n u m b e r s p which satisfy
r
+
p) = 0, ~.
Each zero or pole has multiplicity m o (taken negatively if ~ ( 71 + p ) = 0). T h e only pole is a simple pole at p = ~1 so that m 1 = +1. T h e only difference between this formula and Weil's [W] explicit formula is that it involves the Mellin transform instead of the Fourier transform. Of course one can easily pass from one to the other by a logarithmic transformation. Let F E Y be associated to f. Replacing x by [[x[[, the n o r m of an idele (in the above formula), we may rewrite the explicit formula in the form
F(x) =
~
mp/(p) (ll~ll p + Ilxll -~)
-
El(x).
~( 89 It now follows from the results of w that the cusp forms are given by the functions
Ilxll ~ + Ilxl1-0, where 71 + P is a zero of the zeta function. In fact these functions can be constructed explicitly by forming F ( x ) - E F ( X ) , where F is associated to I and f(A) vanishes at all the zeros or poles of the zeta function with one exception. Every F C Z:2(T) can be expressed as a linear combination of cusp forms plus an integral of Eisenstein series plus a multiple of
Ilxll + + Ilxll-+, which is a residue of the Eisenstein series. This establishes the spectral decomposition of the space/32(T). It is not hard to show that a cusp form [[x[[~ + [[xl[ - ~ has n o r m zero if and only if p is not pure imaginary. This establishes the fact that the R i e m a n n hypothesis is equivalent to the positivity of the inner product < > on the cuspidal spectrum.
147 w T h e z e t a f u n c t i o n a s s o c i a t e d to a n a u t o m o r p h i c f o r m on s Let F(x) e / : 2 ( T ) be associated to f E S. Then
C'
1 / a+i~
A)](A)(xX+x-X)dA,
1 for a > ~. We define the zeta function associated to F by the formula
CF(~) - 2~-i .,~-,oo
+ ~
C
](A) ->,-+~ +
~ ~- ~
d~,
which is the Mellin transform _~(s). To see this, we note that the Mellin transform of x A + x -A (denoted by (~+~-~-~) (s)) is defined as a distribution by the formula
l/~+i~
-(~+~-=~) (s)~(~) d~ = 27ri ~ a - i o ~
-- 2~Fi ,,~_ioo
/oo~ (x ~ + x_~)g(x ) dx x
s - A + ~s+ )~ ~(s) ds
= 2~(~), which defines a continuous linear functional on [0, ~ ] with invariant multiplicative measure d_~. We may thus interpret _L~( 89+ A)fi(A) as the Ath Mellin coefficient of F(x). This is analogous to the Fourier coefficient in the classical theory of automorphic forms where the expansion is taken with respect to the additive group. The zeta function associated to F satisfies the functional equation
CF(~) = - C F ( - ~ ) . The zeta function associated to a cusp form IIx]]p + such that ((1 + p ) = 0, will be - -
s
1 -p
-31-
1 s+p
[]xl]-P with
p
148
w
The Rankin-Selberg convolution: Let
~ = I1~11~ + I1~11-< ~ , = I1~11r + II~ll - r
((89
be two cusp forms where ((1 + p) = + p,) = 0 . Let F C s be associated to f E S. We form the inner product
< F, ~-W~r~,>
=
o ~ f(llxli)(llxllP
+ I1~11-~ (llxllr + I1~11-r
dx
= 2i(p + p') + 2/(p - p') The above formula can be further generalized by enlarging the allowable class of functions F to include distributions. If we consider the case
r(x)_
E(~,s) ~(~) ,
and recall the previously proved fact that the Mellin transform N
(x~+~-~)
(s) -
a-
1
--'
~ +
1
we obtain the Rankin-Selberg convolution
< E ( , s), ~,~,, >
= 2
[1 s+p+
pl +
1 s -p-
pt +
1 s+p-
pl +
1 ]pl
s-p+
"
This may be interpreted as the zeta function associated to an automorphic function for a certain group acting on (A x )2 in analogy with the Gelbart-Jacquet (see [G-J]) lift. These ideas will be briefly explained ill the next section.
w
H i g h e r rank g e n e r a l i z a t i o n s :
For n > 2 let Wn denote the Weyl group of G L ( n ) (which consists of all n • n matrices with zero entries except for exactly one
149 1 in each row and column). For an arbitrary multiplicative group K let 7?n(K) denote the diagonal group of all matrices of the form
kn "'. k2 kl )
with
ki 9
(l
The Weyl group Wn acts on :Dn(K)/K by conjugation. For a normal subgroup H <1 K, the group :D,,(H)/H acts on 7?n(K)/g by left matrix multiplication. We define the group ~ . ( H ) to be the group generated by the group 7?,,(H)/H and the Weyl group W.; by construction G,,(H) acts naturally on :Dr,(K)/K. We now consider the group F,, = F,,(Q x ) = Gn(Q• x which acts on :D,(AX)/A x. Every Z 9 :D,,(AX)/A x can be put in the canonical form
X~
I xn-1
I
xl 1
which we write succinctly as X = ( x l , . . . ,Xn--1). Let Xoo = 1 (Zoo,... , x ~ -1) and Xp = ( x l , . . . ,xpn--l) denote the infinite and p-component of X , respectively. A function
f:
•
C,
is said to be a Schwartz function if it and its partial derivatives (of all orders) with respect to x l , . . . ,x n-1 are bounded. Denote by Sn the space of all such Schwartz functions. We shall now construct a vector space (over I2) of automorphic functions for the group F,,. Let
x=(x',...,
x
-l) c
For a prime q, the function
"~q(Xql)= xlq uq"'" x q - l [ u q
• .
150
is invariant under the group W~. For an n-tuple of primes P = ( p l , P 2 , . . . ,Pn), we adopt the simplifying notation:
x~,-- {xp~,... ,x,,~ log P = ( l o g p l ) . . . (logpn-1). The function n-1
Ap(xp)=
II w/;, w6Wr, i=1
is automatically invariant under Wn. We define the vector space T~ over C to be all linear combinations (with complex coefficients) of automorphic functions of the form F(X)
=
~ 6 F . P = ( P l ..... P,~-I)
\qqLP
with f 6 S~. We say F is associated to f. As in the rank one case, an inner product on this space may be defined as follows. Let F, G C T~ be associated to f, g 6 S,~, respectively. Define the indefinite inner product n--1 dx j
F(x).g(llxll[,..
=
" , Ilxn-l[[)
Hi l
j=l
In the special case where F is associated to an f given in the form
[Iwx'll ~1..- ilwx--ll[s~ -, w 6 W,~
we obtain a minimal parabolic Eisenstein series which lies in the continuous spectrum of the space of differential operators which commute with our group qSn.
151
As an example, we consider t h e case n -- 3. T h e group Fa, which acts on T)n(A x ) / A x , is g e n e r a t e d by t h e t r a n s f o r m a t i o n s
(...11)
.lxl )
a n d the s y m m e t r i e s
x21)
(" / (" (.. /( X1
7i
~
for oL1,ol ~ Q,
1
yr
1
X1
1
~
1
1
x1 1
/ x2 X1 1 / V--+( x~l 1
associated to t h e Weyl g r o u p e l e m e n t s
(Z1~ (i ~176(i~ 0 0
0 1
,
0 1
1 0
,
( 1o) ( ol) 0 0
1 0
,
1 0
0 1
,
0 0
respectively. T h e cusp forms will be of t y p e
Ilxlll p 9IIx211 p' + ii~lii-P-r
li~lil p'.ilx~l Ip +
ill, it p' +
ilxlll r . lix~ll-P-p' + +
ilx lli p 9lix211-P-p',
152
where ~1 + p, and $1 + p! are nontrivial zeros of the Riemann zeta function. In this manner, the lift from G2 to Ga can be established. w
References:
[D1] Ch. Deninger, On the F-factors attached to motives, Inv. Math. 104 (1991), 245-261. [D2] Ch. Deninger, Local L-factors of motives and regularized determinants, Inv. Math. 107 (1992), 135-150. [G-J] Gelbart, S. and Jacquet, H. A relation between automorphic representations of GL(2) and GL(3), Ann. Sci. Ecole Normale Sup. 4 e s6rie. 11 (1978), 471-552. [K] Kurokawa, N. Multiple zeta functions: an example, in Zeta Functions in Geometry, ed. Kurokawa, N. and Sunada, T., Advanced Studies in Pure Mathematics 21 (1992), 219-226. [M] Manin, Y. Lectures on zeta functions and motives, to appear in Ast6risque. JR] Rankin, R. Contributions to the theory of Ramanujan's function r(n) and similar arithmetic functions, I and II, Proc. Cambridge Phil. Soc. 35 (1939), 351-356, 357-372. [S1] Selberg, A. Bemerkungen iiber eine Dirichletsche Reihe, die mit der Theorie der Modulformen nahe verbunden ist, Arch. Math. Naturvid. 43 (1940), 47-50. [$2] Selberg, A. Harmonic analysis and discontinuous groups in weakly symmetric Riemannian spaces with applications to Dirichlet series, J. Indian Math. Soc. 20 (1956), 47-87. [W] Weil, A. Sur les 'formule, explicite,' de la thdorie des hombres premiers, Comm. S6m. Math. Univ. Lund 1952, Tome suppl6mentaire (1952), 252-265.
Department of Mathematics Columbia University New York, NY 10027 [email protected]
INDEX Additive functional equation Additive zeta function Adeles Admissible strip AS axioms Asymptotic polynomial Asymptotics in parallel strips Asymptotics in sectors Asymptotics in vertical strips Asymptotics to the left Basic Fourier Inversion formula Bessel integral Cram5r formula Cram~r functional Cram~r theorem Cusp forms DIR axioms Dirichlet polynomials Duistermaat- Guillemin theorem
7, II0, 117 98, 110 139 36 36 15 2O 22 2O 24 65 110, 115
49 48 5, 57 145 14 39, 58, 93 117
E-transform Eisenstein series Euler sum Explicit formula
90, 99 142 45 68
Fourier conditions Fudge factors Fujii theorem Functional equation Fundamental class
62 45 122 45 46, 90
Gamma function Gauss formula Gauss transform GR (growth) axioms
39, 83 19 102 49
Heat kernel Higher rank Hyperbolic 3-space
85, 103 148 113
154 Indefinite space
140
Karamata theorem Laplace-Mellin transform
27 16, 102
Mellin transform Minakshisunaram-Pleijel Principal part
63 3, 118 15
Rankin-Selberg convolution Reduced order Regularized harmonic series Regularized harmonic series type Regularized product type Riemann zeta function Selberg zeta function Sine function Special case Spectral case Spectral functional equation Spectral theory Symmetric Schwartz function Theta inversion Theta series
148 19, 37 11, 17, 37 37, 38 36 39, 128 39, 129 92, 128 15 11, 17 94 141 144
6, 87, 109 87, 103
Venkov theorem Wave kernel Weil functional Zeta function of automorphic form Zeta function of hyperbolic 3-space Zeta function of number fields
117 68, 79
147 113-117 94
Vol. 1502: C. Simpson, Asymptotic Behavior of Monodromy. V, 139 pages. 1991.
Vol. 1526: J. Aztma, P. A. Meyer, M. Yor (Eds.), Stminaire de Probabilit~s XXVI. X, 633 pages. 1992.
Vol. 1503: S. Shokranian, The Selberg-Arthur Trace Formula (Lectures by J. Arthur). VII, 97 pages. 1991.
Vol. 1527: M. I. Freidlin, J.-F. Le Gall, Ecole d'Et6 de Probabilitts de Saint-Flour XX - 1990. Editor: P. L. Hennequin. VIII, 244 pages. 1992.
Vol. 1504: J. Cheeger, M. Gromov, C. Okonek, P. Pansu, Geometric Topology: Recent Developments. Editors: P. de Bartolomeis, F. Tricerri. VII, 197 pages. 1991. Vol. 1505: K. Kajitani, T. Nishitani, The Hyperbolic Cauchy Problem. VII, 168 pages. 1991. Vol. 1506: A. Buium, Differential Algebraic Groups of Finite Dimension. XV, 145 pages. 1992. Vol. 1507: K. Hulek, T. Peternell, M. Schneider, F.-O. Schreyer (Eds.), Complex Algebraic Varieties. Proceedings, 1990. VII, 179 pages. 1992. Vol. 1508: M. Vuorinen (Ed.), Quasiconformal Space Mappings. A Collection of Surveys 1960-1990. IX, 148 pages. 1992. Vol. 1509: J. Aguadt, M. Castellet, F. R. Cohen (Eds.), Algebraic Topology - Homotopy and Group Cohomology. Proceedings, 1990. X, 330 pages. 1992.
Vol. 1528: G. Isac, Complementarity Problems. VI, 297 pages. 1992. Vol. 1529: J. van Neerven, The Adjoint of a Semigroup of Linear Operators. X, 195 pages. 1992. Vol. 1530: L G. Heywood, K. Masuda, R. Rantmann, S. A. Solonnikov (Eds.), The Navier-Stokes Equations II Theory and Numerical Methods. IX, 322 pages. 1992. Vol. 1531: M. Stoer, Design of Survivable Networks. IV, 206 pages. 1992. Vol. 1532: J. F. Colombeau, Multiplication of Distributions. X, 184 pages. 1992. Vol. 1533: P. Jipsen, H. Rose, Varieties of Lattices. X, 162 pages. 1992. Vol. 1534: C. Greither, Cyclic Galois Extensions of Commutative Rings. X, 145 pages. 1992.
Vol. 1510: P. P. Kulish (Ed.), Quantum Groups. Proceedings, 1990. XII, 398 pages. 1992.
Vol. 1535: A. B. Evans, Orthomorphism Graphs of Groups. VIII, 114 pages. 1992.
Vol. 1511: B. S. Yadav, D. Singh (Eds.), Functional Analysis and Operator Theory. Proceedings, 1990. VIII, 223 pages. 1992.
Vol. 1536: M. K. Kwong, A. Zettl, Norm Inequalities for Derivatives and Differences. VII, 150 pages. 1992.
Vol. 1512: L. M. Adleman, M.-D. A. Huang, Primality Testing and Abelian Varieties Over Finite Fields. VII, 142 pages. 1992.
Vol. 1537: P. Fitzpatrick, M. Martelli, J. Mawhin, R. Nussbaum, Topological Methods for Ordinary Differential Equations. Montecatini Terme, 1991. Editors: M. Furi, P. Zecca. VII, 218 pages. 1993.
Vol. 1513: L. S. Block, W. A. Coppel, Dynamics in One Dimension. VIII, 249 pages. 1992.
Vol. 1538: P.-A. Meyer, Quantum Probability for Probabilists. X, 287 pages. 1993.
Vol. 1514: U. Krengel, K. Richter, V. Warstat (Eds.), Ergodic Theory and Related Topics III, Proceedings, 1990. VIII, 236 pages. 1992.
Vol. 1539: M. Coornaert, A. Papadopoulos, Symbolic Dynamics and Hyperbolic Groups. VIII, 138 pages. 1993.
Vol. 1515: E. Ballico, F. Catanese, C. Ciliberto (Eds.), Classification of Irregular Varieties. Proceedings, 1990. VII, 149 pages. 1992. Vol. 1516: R. A. Lorentz, Multivariate Birkhoff Interpolation. IX, 192 pages. 1992. Vol. 1517: K. Keimel, W. Roth, Ordered Cones and Approximation. VI, 134 pages. 1992. Vol. 1518: H. Stichtenoth, M. A. Tsfasman (Eds.), Coding Theory and Algebraic Geometry. Proceedings, 1991. VIII, 223 pages. 1992. Vol. 1519: M. W. Short, The Primitive Soluble Permutation Groups of Degree less than 256. IX, 145 pages. 1992. Vol. 1520: Yu. G. Borisovich, Yu. E. Gliklikh (Eds.), Global Analysis - Studies and Applications V. VII, 284 pages. 1992. Vol. 1521: S. Busenberg, B. Forte, H. K. Kuiken, Mathematical Modelling of Industrial Process. Bari, 1990. Editors: V. Capasso, A. Fasano. VII, 162 pages. 1992. Vol. 1522: J.-M. Delort, F. B. I. Transformation. VII, 101 pages. 1992. Vol. 1523: W. Xue, Rings with Morita Duality. X, 168 pages. 1992. Vol. 1524: M. Coste, L. Maht, M.-F. Roy (Eds.), Real Algebraic Geometry. Proceedings, 1991. VIII, 418 pages. 1992. Vol. 1525: C. Casacuberta, M. Castellet (Eds.), Mathematical Research Today and Tomorrow. VII, 112 pages. 1992.
Vol. 1540: H. Komatsu (Ed.), Functional Analysis and Related Topics, 1991. Proceedings. XXI, 413 pages. 1993. Vol. 1541: D. A. Dawson, B. Maisonneuve, J. Spencer, Ecole d" Et6 de Probabilitts de Saint-Flour XXI - 1991. Editor: P. L. Hennequin. VIII, 356 pages. 1993. Vol. 1542: J.Frthlich, Th.Kerler, Quantum Groups, Quantum Categories and Quantum Field Theory. VII, 431 pages. 1993. Vol. 1543: A. L. Dontchev, T. Zolezzi, Well-Posed Optimization Problems. XII, 421 pages. 1993. Vol. 1544: M.SchiJrmann, White Noise on Bialgebras. VII, 146 pages. 1993. Vol. 1545: J. Morgan, K. O'Grady, Differential Topology of Complex Surfaces. VIII, 224 pages. 1993. Vol. 1546: V. V. Kalashnikov, V. M. Zolotarev (Eds.), Stability Problems for Stochastic Models. Proceedings, 1991. VIII, 229 pages. 1993. Vol. 1547: P. Harmand, D. Werner, W. Werner, M-ideals in Banach Spaces and Banach Algebras. VIII, 387 pages. 1993. Vol. 1548: T. Urabe, Dynkin Graphs and Quadrilateral Singularities. VI, 233 pages. 1993. Vol. 1549: G. Vainikko, Multidimensional Weakly Singular Integral Equations. XI, 159 pages. 1993. Vol. 1550: A. A. Gonchar, E. B. Saff (Eds.), Methods of Approximation Theory in Complex Analysis and Mathematical Physics IV, 222 pages, 1993.
Vol. 1551: L. Arkeryd, P. L. Lions, P.A. Markowich, S.R. S. Varadhan. Nonequilibrium Problems in Many-Particle Systems. Montecatini, 1992. Editors: C. Cercignani, M. Pulvirenti. VII, 158 pages 1993. VoL 1552: J. Hilgert, K.-H. Neeb, Lie Semigroups and their Applications. XII, 315 pages. 1993. Vol. 1553: J.-L- Colliot-Th61bne, J. Kato, P. Vojta. Arithmetic Algebraic Geometry. Trento, 1991. Editor: E. Ballico. VII, 223 pages. 1993. Vol. 1554: A. K. Lenstra, H. W. Lenstra, Jr. (Eds.), The Development of the Number Field Sieve. VIII, 131 pages. 1993. Vol. 1555: O. Liess, Conical Refraction and Higher Microlocalization. X, 389 pages. 1993. Vol. 1556: S. B. Kuksin, Nearly Integrable InfiniteDimensional Hamiltonian Systems. XXVII, 101 pages. 1993.
Vol. 1574: V. P. Havin, N. K. Nikolski (Eds.), Linear and Complex Analysis - Problem Book 3 - Part II. XXII, 507 pages. 1994. Vol. 1575: M. Mitrea, Clifford Wavelets, Singular Integrals, and Hardy Spaces. XI, 116 pages. 1994. Vol. 1576: K. Kitahara, Spaces of Approximating Functions with Haar-Like Conditions. X, 110 pages. 1994. Vol. 1577: N. Obata, White Noise Calculus and Foek Space. X, 183 pages. 1994. Vol. 1578: J. Bernstein, V. Lunts, Equivariant Sheaves and Functors. V, 139 pages. 1994. Vol. 1579: N. Kazamaki, C o n t i n u o u s Exponential Martingales and BMO. VII, 91 pages. 1994. Vol. 1580: M. Milman, E x t r a p o l a t i o n and Optimal Decompositions with Applications to Analysis. XI, I61 pages. 1994.
Vol. 1557: J. Az6ma, P. A. Meyer, M. Yor (Eds.), S6minaire de Probabilit6s XXVII. VI, 327 pages. 1993.
Vol. 1581: D. Bakry, R, D. Gill, S. A. Molchanov, Lectures on Probability Theory. Editor: P. Bernard. VIII, 420 pages. 1994.
Vol. 1558: T. J. Bridges, J. E. Furter, Singularity Theory and Equivariant Symplectic Maps. VI, 226 pages. 1993.
Vol. 1582: W. Balser, From Divergent Power Series to Analytic Functions. X, 108 pages. 1994.
Vol. 1559: V. G. Sprind~uk, Classical Diophantine Equations. XII, 228 pages. 1993.
Vol. 1583: J. Az~ma, P. A. Meyer, M. Yor (Eds.), S6minaire de Probabilit6s XXVIII. VI, 334 pages. 1994.
Vol. 1560: T. Bartsch, Topological Methods for Variational Problems with Symmetries. X, 152 pages. 1993. Vol. 1561 : 1. S. Molchanov, Limit Theorems for Unions of Random Closed Sets. X, 157 pages. 1993.
Vol. 1584: M. Brokate, N. Kenmochi, I. Mtiller, J. F. Rodriguez, C. Verdi, Phase Transitions and Hysteresis. Montecatini Terme, 1993. Editor: A. Visintin. VII. 291 pages. 1994.
Vol. 1562: G. Harder, Eisensteinkohomologie und die Konstruktion gemischter Motive. XX, 184 pages. 1993.
Vol. 1585: G. Frey (Ed.), On Artin's Conjecture for Odd 2dimensional Representations. VIII, 148 pages. 1994.
VoL 1563: E. Fabes, M. Fukushima, L. Gross, C. Kenig, M. R6ckner, D. W. Stroock, Dirichlet Forms. Varenna, 1992. Editors: G. Dell'Antonio, U. Mosco. VII, 245 pages. 1993.
Vol. 1586: R. Nillsen, Difference Spaces and Invariant Linear Forms. Xll, 186 pages. 1994.
Vol. 1564: J. Jorgenson, S. Lang, Basic Analysis of Regularized Series and Products. IX, 122 pages. 1993.
Vol. 1588: C. Scheiderer, Real and l~tale Cohomology. XXIV, 273 pages. 1994.
Vol. 1565: L. Boutet de Monvel, C. De Concini, C. Procesi, P. Schapira, M. Vergne. D-modules, Representation Theory, and Quantum Groups. Venezia, 1992. Editors: G. Zampieri, A. D'Agnolo. VII, 217 pages. 1993.
Vol. 1589: J. Bellissard, M. Degli Esposti, G. Forni, S. Graffi, S. Isola, J. N. Mather, Transition to Chaos in Classical and Quantum Mechanics. Montecatini, 1991. Editor: S. Graffi. VII, 192 pages. 1994.
Vol. 1566: B. Edixhoven, J.-H. Evertse (Eds.), Diophantine Approximation and Abelian Varieties. XIII, 127 pages. 1993.
Vol. 1590: P. M. Soardi, Potential Theory on Infinite Networks. VIII, 187 pages. 1994.
Vol. 1567: R. L. Dobrushin, S. Kusuoka, Statistical Mechanics and Fractals. VII, 98 pages. 1993. Vol. 1568: F. Weisz, Martingale Hardy Spaces and their Application in Fourier Analysis. VIII, 217 pages. 1994. Vol. 1569: V. Totik, Weighted Approximation with Varying Weight. VI, 117 pages. 1994. Vol. 1570: R. deLaubenfels, Existence Families, Functional Calculi and Evolution Equations. XV, 234 pages. 1994. Vol. 1571: S. Yu. Pilyugin, The Space of Dynamical Systems with the C~ X, 188 pages. 1994, Vol. 1572: L. GOttsche, Hilbert Schemes of ZeroDimensional Subschemes of Smooth Varieties. IX, 196 pages. 1994. Vol. 1573: V. P. Havin, N. K. Nikolski (Eds.), Linear and Complex Analysis - Problem Book 3 - Part 1. XXII, 489 pages. 1994.
Vol. 1587: N. Xi, Representations of Affine Hecke Algebras. VIII, 137 pages. 1994.
Vol. 1591 : M. Abate, G. Patrizio, Finsler Metrics - A Global Approach. IX, 180 pages. 1994. Vol. 1592: K. W. Breitung, Asymptotic Approximations for Probability Integrals. IX, 146 pages. 1994. Vol. 1593: J. Jorgensou & S. Lang, D. Goldfeld, Explicit Formulas for Regularized Products and Series. VIII, 154 pages. 1994.