Professional Documents
Culture Documents
Basic Analysis
of Regularized Series
and Products
Springer-Verlag
Berlin Wdelbup NewYork
--=w
Hong Koq Barcelona
Budapest
Authors
Jay A. Jorgenson
Serge Lang
Department of Mathematics Foreword
Yale University
Box 2155 Yale Station
New Haven. CT 06520. USA The two papers contained in this volume provide results on
which a series of subsequent papers will be based, starting with
[JoL 92b], [JoL 92d) and [JoL 931. Each of the two papers contains
an introduction dealing at greater length with the mathematics
involved.
The two papers were &st submitted in 1992 for publication in
J. mine angew. Math. A referee emitted the opinion: "While such
generalized products are of interest, they are not of such central
interest as to justify a series of long papers in expensive journals."
The referee was cautious, stating that this "view is subjective" , and
adding that he "will leave to the jud ement of the editors whether
%
to pass on this recommendation to t e authors". The recornmen-
dation, in addition not to publish "in expensive journals", urged
US to publish a monograph instead. In any case, the editors took
full responsibility for the opinion about the publication of our se-
ries "in expensive journalsn. We disagree very strongly with this
opinion. In fact, one of the applications of the complex analytic
properties of r e e i z e d products contained in our first paper is
to a generalization of Crank's theorem, which we prove in great
generality, and which appears in Math. Annalen [JoL 92bj. The
MathematicsSubjectClassification(l991): 1lM35.l lM41,llM99,30BSO,30D15, referee for Math. Ann. characterized this result as "important and
35P99,35899,39B99,42A99 basic in the field".
(Authors' Note: then is no MSC number for regularized products, but there should be.)
Our papers were written in a self-contained way, to provide a
suitable background for an open-ended series. Thus we always con-
sidered the possible alternative to put them in a Sprin er Lecture
ISBN 3-540-57488-3 Springer-Verlag Berlin Heidelberg New York
ISBN 0-387-57488-3 Springer-Verlag New York Berlin Heidelberg B
Note, and we are very grateful to the SLN editors and pringer for
publishing them.
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, reproductionon microfilms or in any other Acknowledgement: During the preparation of these papers,
way. and storage in data banks. Duplication of this publication or puts thereof is the first author received support from the NSF Postdoctoral Fel-
permitted only under the provisions ofthe GermPn Copyri%ltLaw of Ssptember 9, 1owsMg WS-S-65681 and from FBF grant DMS-93-07023.The
1965. in its cumnt version, and permission for use must dways be obtained from second author benefited from his visits at the Max Planck Institut
Springer-Verlag. Violations are liable for prosecution under the German Copyright in Bonn.
Law.
Q Springer-Verlag Berlin Heidelberg 1993
Printed in Germany
214613140-543210 - Printed on acid-free paper
Part I
Content
Part I1
. . . ..... . .. . .. . .. . ... .. .
Introduction . . . .. . . . . . . . . 9 1
. . .....
1. A Theorern on Fourier Integrals .... ... .. .. 93
2. A P a n e d Formula . ............. . ..... .....
.. . . 105
. .. . . ... . ...
3. The General P a t 4 Formula . . . . . . 109
. .
4. The ParseVal Formula for I,(a+ it) . . .... .. ... 114
may be viewed a s giving a value for the meaningless infinite product where a ranges over the (non-zero) ideals of the ring of algebraic
on the right. Similarly, using the sequence { A k + z ) instead of { A k } , integers of F, and Na is the absolute norm, in other words, the
we would obtain a value for the meaningless infinite product index Na = (a : a). Then
where the derivative here is the partial derivative with respect to where a, is the number of ideals a such that Na = k. This theta
the variable s. To make sense of this procedure in the spectral function is different from the one which occurs in Hecke's classical
case, under certain conditions one shows that the sequence { A k ) proof of the functional equation of the Dedekind zeta functions (cf.
-
-
also determines: [La 701, Chapter XIII). Of ~ m cthis , example extends in a nat-
ural way to Dirichkt, Hedp, Artin, and other Gaeries classically
- The regularized product associated to number fields. In these extesIlSions, ak is usually not
an integer.
D(z) = cP(') ~ ( z ) ,
Zaa functions a r k q frola representation theory and the theory
where P is a normalizing polynomial, and E(z) is a standard of automorphic ibctiew comtatutc au &BILlljOn of the present
,but we omit hss fmtba meattaat of them to avoid having
Weierstrass product ha zeros at the numbers -4 with mul-
3
tipliuty a,. The degree P and the order of the Weierstrass ex-g
to el rate on their more complicated definitions.
Example 3. Regularized determinant of a n operator. In The multiplication formula may be viewed as a special case of
this case, we let {Ak} be the sequence of eigenvalues of an operator. the Artin formalism treated in [JoL 92dJ. The Parseval formula,
In the most classical case, the operator is the positive Laplacian on which determines the Fourier transform of r'/r as a distribution
a compact Riemannian manifold, but other much more complicated on a vertical line will be addressed in the context of a general re-
examples also arise naturally, involving possibly non-compact man- sult in Fourier analysis in [JoL 92~1.Here we show that the other
ifolds or pseudo differential operators. Suitably normalized, the formulas can be expressed and proved in a general context, under
function D(z) is viewed as a regularized determinant (generalizing certain axioms (covering all four examples and many more com-
the characteristic polynomial in fhite dimensions). plicated analogues). We shall find systematically how the simple
expression 1/(1 - e-') is replaced by theta functions throughout
Example 4. Zeros of t h e zeta function. Let {pk}range over the formalism developed in this part. More generally, whenever the
the zeros of the Riemann zeta function with positive imaginary above expression occurs in mathematics, one should be on the look-
part. Let a t be the multiplicity of pk, conjecturally equal to 1. out for a similar more general structure involving a theta function
Put A; = pk/i. The sequence {Xi} is thus obtained by rotating associated to a sequence having a regularized product.
the vertical strip to the right, so that it becomes a horizontal strip.
A theorem of Cram& [Cr 191 gives a meromorphic continuation T h e Asymptotic Expansion Axiom. The main axiom is a
(with a logarithmic singularity at the origin) for the function certain asymptotic expansion of the theta function at the origin,
given a s AS 2 in 1; namely, we assume that there exists a se-
quence of complex numbers {p) whose real parts tend to infinity,
and polynomials Bp such that
which amounts to a theta function in this case (after the change of
variables z = it).
This fourth example generalizes as follows. Suppose given a se-
quence {Ak)such that the corresponding zeta function has an Euler The presence of log terms is essential for some applications.
product and functional equation whose fudge factors are of regular-
ized product type. (These notions will be defined quite generally In Example 1, the asymptotic expansion of the theta function is
in [JoL 92bl.) We are then led to consider the sequence {A;} de- immediate, since B(t) = 1/(1 - e-'). In Example 2, this expansion
fined as above. From $7, one sees that a regularized product exists follows from the consideration of $7 of the present part. In both
for the sequence {Ak}. We will show in [JoL 92b] that a regular- Examples 1 and 2, Bp is constant for all p, so we say that there are
ized product also exists for the sequence {AJ}. Passing from {Ak} no log terms.
to {A;} will be called climbing t h e ladder in the hierarchy of In the spectral theory of Example 3, Minakshisundaram-Pleijel
regularized products. (MP 491 introduced the zeta function formed with the sequence of
Basic Formulas. The example of the gamma function provides eigenvalues Ak, and Ray-Singer introduced the so-called analytic
a basic table of properties which can be formulated and proved torsion [RS 731, namely ('(0). Voros [Vo 871 and Cartier-Voros
under some additional basic conditions which we shall list in a [CaV 901 gave further examples and raults, dealing with a se-
=
' moment . The table ineludes quence of n u m h whose real part ten& to inffnity. We have found
The multiplication formula their axio91atizatian caPlcernigg the oomsponding theta function
useful. However, both artides [Vo 871 w d [CaV 901 leave some
-
* The Lerch formula
The (other) Gauss formula basic questions open in laying doan t h foundations of regularized
The Stirling formula products. V o w himself states: "In the present work, we shall
The Hankel formula not be concerned with rigorous p r d , which certainly imply addi-
The Mellin inversion formula tional regularity pmpcrtiea for the q w n c e (Ak)." Furthermore,
The Parseval formula. Cartier-Vorw have only certain spkiflc and special applications in
mind (the Poisson summation formula and the Selberg trace for- types, the formula is new as far as we know. In Example 3, the for-
mula in the case of compact Riemann surfaces). Because of our mula occurs in many special cases of the theory of analytic torsion
more general asymptotic expansion for the theta function, the the- of Ray-Singer and in Vome [Vo 871, formula (4.1). In Example 4,
ory becomes applicable to arbitrary compact manifolds with arbi- the formula specializes to a formula discovered by Deninger for the
trary Riemannian metrics and elliptic pseudwdifferential operators Riemann zeta function Cg (see [De 921, Theorem 3.3).
where the log terms appear starting with [DUG 751, and continu- Applications. Aside from developing a formalism which we
ing with [BrS 851, [Gr 861 and [Ku 881 for the spectral theory. In find interesting for its own sake, we also give systematically funda-
this case, the theta function is the trace of the heat kernel, and its mental analytic results which are used in the subsequent series of
asymptotic expansion is proved as a consequence of an asymptotic parts, including not only the generalization of Crarnhr's theorem
expansion for the heat kernel itself. mentioned above, but for instance our formulation of general ex-
In Example 4 for the Riemann zeta function, the asymptotic plicit formulas analogous to those of analytic number theory (see
expansion follows as a corollary of Cramhr's theorem. A log term [JoL 931). These particular applications deal with cases when the
appears in the expansion. The generalization in [JoL 92b] involves zeta function has an Euler product and functional equation. Such
some extra work. Some of the arguments used to prove the asymp- cases may arise from a regularized product by a change of variables
totic expansion AS 2 are given in 55 of the present part (especially I = s(s - 1). The Selberg zeta function itself falls in this category.
, Theorem 5.11), because they are directly related to those used to However, so far the Euler product does not play a role. A prema-
prove Stirling's formula. Indeed, the Stirling formula gives an as- ture change of variables z = s(s - 1) obscures the basic properties
ymptotic expansion for the log of a regularized product at infinity; of the regularized product and Dirichlet series which do not depend
in [JoL 92b] we require in addition an asymptotic expansion for on the Euler product.
the Laplace transform of the log of the regularized product in a
neighborhood of zero. Although, as we have pointed out, some special cases of our
formulas are known, many others are new. Our results and formulas
Normalization of t h e Weierstrass Product by t h e Lerch concerning regularized products are proved in sufficient generality
Formula. We may now return to describe more accurately our to apply in several areas of mathematics where zeta functions occur,
normalization of the Weierstrass product. When the Hurwitz zeta e.g. analytic number theory, representation theory, spectral theory,
function ((s, z) is holomorphic at s = 0, and all numbers ak are ergodic theory and dynamical systems, etc. For example, our Lerch
positive integers, there is a unique entire functions D(z) whose formula is seen to apply to Selberg type zeta functions not only for
zeros are the numbers -Ak with multiplicities ah, with a normalized Riemann
-- surfaces but for certain higher
- dimensional manifolds as
Weierstrass product such that the Lerch formula is valid, namely well.
log D(z) = -('(0, z), Furthermore, our general principle of climbing the ladder of reg-
ularized products applies to the scattering determinant associated
where the derivative on the right is with respect to the variable s.
As to the Weierstrass order of D, let po be a leading exponent
<
in the asymptotic expansion for B(t), i.e. Re(po) Re(p) for all p
such that Bp # 0. Let M be the largest integer < -Re(po). Then
-
-
+
M 1 is the order of D.
In the general case with the log terms present in the asymptotic previously.
expansion, the Hurwitz zeta function ((s, z ) may be meromorphic
at s = 0 instead of being holomorphic, and we show in 53 how to Our theory also applies to R u d e type zeta functions arising in
make the appropriate definitions so that a similar formula is valid. ergodic theory and dynamicd systems (for example, see [Fr 861 and
references given in that part).
In Example 1, this formula is the classical Lerch formula. In
Example 2, and various generalizations to E h c t i o n s of various Therefore, we feel that it is timely to deal systematically with
51. Laplace-Mellin Transforms
the theory of regularized products, which we find central in math-
ematics.
We first recall some standard results concerning Laplace-Mellin
For the convenience of the reader, a table of notation is included transforms. The Mellin transform of a measurable function f on
at the end of this part. (0,m) is defined by
0 .
The Laplace transform is defined to be
0
The Laplace-Mellin transform combines both, with the defini-
tion
1 f 4 dt
0 9
since
is obviously holomorphic in z E C and Re(s) > -Re(q), by Lemma The preceding lemmas give us information for
1.1. This concludes the proof of Lemma 1.3. 0
g(t) = o(tRe(9)llog tim).
We are now going to show the effect of introducing logarithmic
terms. For p E C, we let Bp denote a polynomial with complex We end our sequence of lemmas by describing more precisely the
coefficients and we put behavior due to the term bp(t)tP.
bp(t) = Bp(logt). Lemma 1.4. Let f be piecewise continuous on ( 0 , ~ ) Assume:
.
We then let B(d, ) be the corresponding constant coefficient partial (a) There is some c E R such that f (t) = O(ect) for t -,oo.
differential operator.
- Lemma 1.3. General Case. Let f be piecewise continuous +
(b) f(t) = bp(t)tp 0(tRe(4jlog tim) ~ i t hbp(t) = B,(IO~t),
on (0, oo ) . Assume: k ( p ) < W q ) , m f Z>O
- and t 4 0.
A -
L
I
(a) f (t) is bounded for t oo. Then LM f (s, 2) is meromorphic in each variable for
p (b) f (t) = bp(t)tP + O(tRe(q)1 log t lm) for some function Re(z) > e and Re(s) > -rRe(q)
bp(t) = Bp(logt) E C[logt],
-
such that Re(p) < Re(q), m f Z>O,and t + 0.
when necessary, the horizontal half line going from -00 to 0. We
Proof. We split the integral: define the open set:
U L = the complement of the above half lines in C.
If all Xk are real and positive, then we note that U L is simply C
minus the negative real axis Rco.
-
where We also suppose given a sequence A = {ak) of distinct complex
numbers. With L and A, we form the asymptotic exponential
I1(s,p, Z) = I(
0
dt
f (t) - bp(t)tp)e-zita-t '
polynomials for integers N 2 1:
The specified poles in s are going to come only from 12.That is:
I1(s,p,z) is holornorphic for z E C and Re($)> -Re(q). with
12(s,p,z) has a meromorphic continuation given by expanding
e-" in its Taylor series and integrating term by term to get
increasing to infinity. To every p in this sequence, we associate a
polynomial Bp and, as before, we set
bp(t ) = Bp(log t ).
I3 (s, 2) is holomorphic for Re(z) > c and all s E C. The theorem We then define the asymptotic polynomials at 0 to be
follows. 0
We now consider an i&te sequence L = f Xk) of distinct corn-
plex numbers satisfying:
DIR 1. For every positive real number c, there is only a finite
number of k such that &(Ak) 5 c. " m a t Perhaps most, applications the exponents p are real. Be-
cause t ere are significant cases when the exponents are not nec-
We use the convention that Xo = 0 and Xk # 0 for k 2 1. essarily real, we lay the foundations in appropriate generality. We
Under condition DIR 1 we delete from the complex plane C the define
horizontal half lines going from -00 to -Ak for each k, together, m(q) = max deg Bp for Re(p) = R.e(q).
Let C ( T ) be the algebra of polynomials in TP with arbitrary Define n(q) to be
complex powers p E C. Then, with this notation,
For this we can apply Lemma 1.4 to f - QN, which shows that
for &(I) > -C the function
is meromor hic in C. In addition, Lemma 1.4 shows that the only so we include Re(p) = 0 in the sum. We decompose the integral
P
possible po es are as described in the statement of the theorem.
Note that the set of these poles is discrete, because the values p + n
into a sum:
we have
in the sense that the series on the right converges absolutely, to which completes the proof of the theorem. 0
M6(s). L particular, the convergence condition DIR 2(a) is
satisfied for the Dirichlet series +
Remark 1. If -Re(po) is an integer, then M 2 is the smallest
integer m for which the Dirichlet series
00
k=l
Indeed
From AS 1we have
+
The first integral on the right has its &st pole at s po = 0, so
: which is uniformly bounded if u > a. Therefore, we have that for at -pe, whence me > -F&(B). The first inequality in (6) follows
+
a 3 a e > a, the difference from the minimality of mo since Re(po) 0.
The following axidition on the v e n c o L uires that, beyond
"PI.
what is stated in the convergence condition DI 1, the sequence
X k approaches infinity in a sector.
DIR 3. There is a fixed r > 0 such that for all k sufficiently Let us apply this inequality to x = c2lAklt and then sum for k 2 N
large, we have to obtain, for any t > 0,
Now if we let
Q=uo+ul,
Theorem 1.12. Let (L, A) be sequences for which the associ-
and
a ted Dirichlet series
e$
k= 1
k=l
Now let
CN = min(c3IXk 1) for k 2 N,
and
r*
"Wote that for any x > 0, there is a constant c = c(uo + 01) such By DIR 1and DIR 3 we have
-that
Z ~ ~ + ue-'l < c. lim CN = oo,
N-roo
which gives the bound
and since
1
I
1
-CN 5 -c3(XkJ for k
2 2
>N,
we can write
Therefore, which shows that AS 1holds, and completes the proof of the the-
orem. 0
Em(z,A) = (1 - ): exp 00
n=m
= ao log z + co + x ( - l ) n - ' c n -
zn
n
or also n=1
We shall work under the assumptions DIR 2 and DIR 3, and, for
>
m mo, we define the Weierstrass product If D(0) # 0, meaning a0 = 0,then D(0) = eCO,and we have
00
- log [D(z)/D(O)] = c ( - 1 ) " ~ . -.zn
n=l
n
From this we see that the coefficients of PD(z) can be determined
By the elementary theory of Weierstrass products, Dm,~(z)is an from the expansion (2) and the numbers
- ?tire function of strict order 5 m, that is
Let us define IIn(r) for negative n through this recursive relation. Proof. If we count Xo = 0 with multiplicity ao, we find for Re(s)
-
- large and r >
mo,
-
Because of the absolute convergence of the theta series qt),we
have, by Theorem 1.8, for all z such that Re(z+Ak) > Ofor all k,
n
abdanyr2m0,
where, as before, conditions AS 1, AS 2 and AS 3. Then L" and OL1l(t)necessarily
satisfy these conditions and
We now look at the constant term in a Laurent expansion at s = 0. From this, we immediately have
If r > n o , then the series
and
DL(%)= DL' (%)DL"( z ) .
converges. At s = 0 the gamma function has a first order pole with A particular example of such a decomposition is the case when L'
residue 1. Note that the n,(O) = 0. So, Theorem 1.6 gives us a +
is a finite subset of L, in which case DL.(%)= n [ ( z X k ) e T ] ; see
bound mo - 1 for the degrees of the polynomials in z occuring as Remark 2 below.
coefficients in the negative powers in s of the Laurent expansion It now becomes of interest to determine in some fashion the coef-
of t ( s ,z ) . From this, we conclude that (&)'((s, z ) is holomorphic ficients of the polynomial PL( z ) ,and for this it suffices to determine
near s = 0. With all this, we can set s = 0 and obtain the equality
P ~ ) ( o )We
. define the reduced sequence Lo to be the sequence
which is the same as L except that we delete Xo = 0. Then
@LO(t)= @(t)- a0
and, as discussed in Fkmark 1, we have
from which we obtain, using (3),
Dmo , L O ( % ) = z-'ODmo ,L(%)-
P , ! ~ ( o=) -B~CT.=o[(s,0 )
for Re(%)sufficiently large, which completes the proof of the theo- for0 5 r 5 mo - 1. That is
rem upon setting r = mo. 0
-
-
so Rl ,((O; Z)is the coefficient of s in the Laurent expansion of ((s, z) satisfies AS 2. The Lerch formula for the sequence {Xi} (with mul-
at s = 0. In the special case we have that tiplicities ak) then specializes to a formula discovered by Deninger
[De 921, Theorem 3.3. In [JoL 92bJwe extend C r a d r ' s theorem
to a wide class of functions having an Euler product and func-
tional equation, including the L-series of a number field (Hecke and
Note that log D ~ ( zand
) log DL(z) differ by an obvious polynomial Artin), L-functions arising in representation theory and modular
in z coming from the Laurent expansion of r ( s ) at s = 0. For forms, Selberg-type zeta funtions and L-functions for Riemann sur-
+
example, if L is a finite sequence, then DL(z) = n ( z At). Also, faces and certain higher dimensional manifolds, etc. The present
let us record the formula section therefore applies to these functions as well, and thus the
Lerch formula is valid for them. Note that Deninger's method did
not give him the analytic continuation for the expressions in his
formula, but such continuation occurs naturally in our approach.
and 00
$3. Expressions at s = 1. Proof. Let us write
In this section we return to general Dirichlet series, as in $1,
meaning we are given sequences (L, A) whose Dirichlet series that
satisfy the convergence conditions DIR 1, DIR 2 and DIR 3, with
associated theta series eL,A= 8 that satisfies the three asymptotic
conditions AS 1, AS 2, and AS 3. With this, we will study the
xi-function
Recall that, by definition, m(0) = max deg Bp for Re(p) = 0 so Using AS 2, we have that the integral in (3) is holomorphic for all
z E C and Re(s) > 0, as is the integral in (5), and the integral in
(4) is holomorphic for all s E C and Re(%) > 4.Combining this,
the stated claim has been proved. 0
(2) e(t) - Poe(t) = { O(l log t lrn(0)) inin the special case
the general case, Corollary 3.2. The function
as t approaches zero.
Theorem 3.1. Let C and N be related as in AS 1. Then the
function is meromorphic at s = 1 for all z with singularities that. are
simple poles at z = -Ak. Also, the residue at z = -Ak is equal
to ak.
that appears in Corollary 3.2. To do so, we expand e-zt in a power Theorem 3.4. Near s = 1, the Hurwitz xi function ((s, z) has
series and apply the following lemma. the expansion
R-n(l)-l(l; 2)
Lemma 3.3. Given p, there is an entire function hp(z) such
that as s approaches 1, such that: ((~9t) = ( 8 - 1)nW+l +-..+R-l(l;Z)
(5 - 1)
+ &(l; z) + O(s - 1),
(a) Special Case: where:
(a) For j < 0, Rj(l; z) is a polynomial of degree < -Re(po ); in
fact, the polar part of ((s, z) near s = 1 is expressed by
with
If all numbers a k are integers, then one can assert the existence of
a meromorphic function D(z), unique up to constant factor, which
satisfies the relation
(9) j
0
je(t) - Pqe(t)l e-"t8
dt
-
t
Re@) > -Re(q), all z
up to constant factor, such that (11) + J [ Q ~ 8 ( t-) Pq8(t)]e-"V- dtt all s, z E UL.
1
(7) D1/D(z) = CT,,l((s, z).
For (9), the asserrtion of meromorphy on the right follows from
Further, we have, by the Lerch formula (Theorem 2.1), condition AS 2 and Lemma 1.3; for (10) the assertion follows by
Lemma 1.4, picking N and C as in AS .l. As for (ll), the integral
irhe Basic Identity: is a sum of integrals of elementary functions which we now recall.
First we have
R(z) = CT8,1(~(s, z) = -dzCT8=o(~(s,z) = D',/DL(~).
We shall normalize the constant factor in (7) so that
Next we shall give another type of expression for the singular so that
term, which also gives an expression for CT8,1(L(s, z) leading into
the Gauss formula of the next section.
Consider the function
M
where 00
F(1,z) = 1
0
[e(t) - Po8(t)] e-"dt.
From this, the conclusion stated above follows since the integral in The constant term on the right is obtained simply by multiply-
(11) is simply a sum of terms of the form given in (12), (13) and ing the Laurent expansions of r ( s +p) and z-"-p and is a universal
(14). 0 expression, meaning an expression that depends solely only on Bp
for Re(p) < 0. A direct calculation shows that there exists a poly-
Theorem 3.6. For any q and for all (s, z) in the region of +
nomial gp of degree at most deg Bp 1 such that
meromorphy, we have:
a) In the special case,
The possible pole of r(s+p) at s = 1accounts for the possibility of
deg gp exceeding deg Bps For convenience of the reader, we present
these calculations explicitly in the special case.
b) in the general case, Let p E C and write
and
Proof. For Re(s) large, one can interchange the sum and integral
in the definition of Fq(s,z) and use that I=O
..
Then
where
/
w
~ ( 1Z ), = [ e ( t )- ~ ~ e ( e-"dt.
t)]
where b-1 is a constant that depends solely on the genus g of X,
0
namely
Ll = 27r(2g - 2).
Example 1. Suppose we are in the spectral case and also that
B(t) is such that the principal part PoO(t) is simply L i l t . Then The iogarithmic derivative of the Selberg zeta function is defined
the equations in Theorem 3.6 become by the equation
we can write
This relation was uscd in [JoL 92d) to prove that the Selberg zeta
function satisifies Artin's formalism ([La 701).
+
P In this case, the integral expression for &(l; z ) reminds one of
7 tHe Gauss formula for the 1 rrithmic derivative of the gamma
"d
function, which will be studie in the next section.
Example 2. The regularid harmonic series is simply related
to the classical Selberg zeta function Z x ( s ) associated to a compact
48
and
The possible pole of I'(s + p) at s = 0 accounts for the possibility Hw(0) = D(w).
of deg B; exceeding deg B,. Using the relation
Furthermore, we have
Now set w = 1 to get and, therefore, can be viewed as a generalization of the classical
Stirling formula.
To begin, we will study the asymptotics of
00
~ ( sz), = J ~(t)e-"t8-
dt
t
for Re(z) oo.
which is the classical Gauss formula. 0
Fix Re(q) > 0 and, as before, let Pq8 be as in AS 2. Let us write
Example 2. Let (L, A) be the sequences of eigenvalues and
multiplicities associated to the Laplacian that acts on smooth sec-
tions of a power of the canonical sheaf over a compact hyperbolic
Riemann surface. Then one can combine Example 2 from $3, The-
orem 4.1 and the Lerch formula to establish the main theorem of
[DP 861 and [Sa 871, without using the Selberg trace formula. In
brief, this theorem states that the Selberg zeta function Zx(s) is
expressible, up to universal gamma-like functions, in terms of the
regularized product DL(s(s - 1)).
Recall that
m(q) = max deg Bp for Re(p) = Re(q).
The terms in (1) and (2) are handled in the two following lemmas.
54
Lemma 5.1. Let x = Re(%)and fix q with Re(q) > 0. Then Lemma 5.2. We have
Jz(0, z) = O(e-"1s) for x + w.
1
yielding the stated estimate. O
0
As for J ~ ( sz),
, recall that
For any complex number so, we have the power series expansion
r(s)
-x8 = (kE - so,k)
C.(SO,~S
e
l=O
,! x)' (S - So)'
(- log
) Lemma 5.3. We have
Thenforx 4 00, ,
+
CTa=o<(s,I) = Bq(r) ~ ( x - ~ ~ ( q )x)"(q)).
(lo~
It is im~ortantto note that Theorem 5.4 shows that the asymp- ~ - ~ (= (-1)" and ~ ( n=) -(-7).
n )-
totics of ~ T , = ~ ( ( S , Zas) & ( z ) -+ oo are governed by the as>m& n! n!
totics of 9(t) as t -+ 0. Also, in the spectral case, the Lerch formula
(Theorem 2.1) applies to give One can prove Lemma 5.5 by using the power series expansion
(4) and the formula
- log D L ( ~=) CTa=o(~(s,4,
and, hence, Theorem 5.4 determines the asymptotics of the reg-
r(s + 1) = s ~ ( s ) and r t ( i ) = -7,
ularized product log DL(%)as Re(z) -+ oo and, consequently, the which comes from the expansion
asymptotics of the the characteristic determinant
log D L(z) as &(z) -+ oo.
Remark 1. In Remark 1 of 34 we defined the polynomials B,+ Proposition 5.6. In the special case, we have, for &(q) > 0,
by the formula -
as x = Re(%)-+ oo. This restatement emphasizes the point that for Re(z) + m.
:' the asymptotics of 9(t) as t -+ 0 determine the asymptotics 0
CT,&(s, z) as Re(z) oo. -+ Example 1. Assume L is such that
To further develop the asymptotics given in Theorem 5.4, one
can use the following lemma.
Remark 2. Lemma 5.7 will be applied to an integral of the which is defined for Re(z) > 0, has a holomorphic extension to
form include a neighborhood of z = 0.
Proof. We have
and, in particular,
We now apply Lemma 5.7 to the function
noting that, by AS 2,
where
-- -
hp(z) = CT8=0hp(s, z).
-
as t -,0. With this, the proof of the lemma is complete. D
+
Proof. It suffices to assume that -Re(s p) is a large, non-
integral, real number, from which the result follows by analytic
Lemma 5.9. The function continuation in s. For this, note that
From this we have are given by universal formulas, depending linearly on El;, hence
on Bp.
As is shown in the proof of Lemma 5.3, we can write
where an
so we can write
R e m a r k 3. A direct calculation shows that there exists a poly-
+
nomial Bp# of degree at most deg Bp 1 such that
The possible zero of sin[?r(s+p)j at s = 0 accounts for the possibility Theorem 5.11. Let {gg), h, {B?), {f,} and {h,} be the
of deg Bf exceeding deg Bp. Further, by combining Remark 1 and above sequences of polynomials and entire functions for those p
Lemma 5.10, we arrive at the formula for which Bp # 0. Then for each q with Re(q) > 3
-
i
4
e-zuu-pB;(log u)du = zp-' ~ f ( l zo) ~ hp(z). +
As is clear from (6), the function hp doeti depend on the choice of
a. We note that the power series hp(z) and the polynomials B f
$6. Hankel Formula
as the real variable t approaches zero from the right.
The gamma function and the classical zeta function are well
- we have that
Example 2. In the case L = Z>O known to satisfy a Hankel formula, that is they are representable
as an integral over a Hankel contour. The existence of such a for-
mula depends on the integrand having at least an analytic contin-
uation over the Hankel contour, and expecially having an analytic
continuation around 0. Indeed, the integrands in these classical
Hankel transforms are essentially theta functions. Unfortunately,
as t approaches zem. In particular, the sequence {p} is simply theta functions cannot always be analytically continued around 0.
Z>- and all polynomials Bp have degree zero. Since For instance, C c-"" cannot, although c~~~ can. Thus the ana-
logues of the classical Hankel transform representations are missing
I =(4)"
log t
- t '
in general.
However, we shall give here one possible Hankel type formula ex-
pressing the Hurwitz xi function associated to the sequences (L, A)
as a complex integral of the regularized harmonic series. Observe
we have, by using the Lerch formula (Theorem 2.1), the equation
that the Hankel formula which we prove here is different from the
classical Hankel representation of the gamma function or the zeta
function. We note that such a formula was used by Deninger for
the Riemann zeta function (see 93 of [De 921).
As in previous sections, let (L, A) be sequences d o s e Dirichlet
series satisfy the three cower ence conditions DIR 1,DIR 2 and
7
Let q -+ co in Theorem 5.11. Usin the absolute convergence of
(8), we get (9). In order to deduce (9 from Theorem 5.11, we have
used that, upon letting q approach infinity, the power series (8)
8
DIR 3, and whose associate theta function satisfies the asymp-
totic conditions AS 1, AS 2 and AS 3, We suppose that the
converges in a neighborhood of the origin. In general, questions sequence L = {Ak}is such that
concerning the convergence of the above stated power series, which
are necessarily questions concerning the growth of the coefficients
b,, must be addressed. For now, let us complete this example by
combining the above results to conclude that Since we deal with arbitrary Dirichlet series ((9) = C akA;,' it is
i" - - 1 + qt).
e - t u log I'(u)du = logt
t 1-et
not the case in general that there is a regularized product whose
logarithmic deriviative gives the constant term CTaZI[(s, 2 ) ; in-
deed, this exists only in the case when ak Z for all k. However,
important a plications will be made to the spectral case, meaning
when ak E $>o for all k, in which case such a regularized product
1
This formula verifies calculations that appear in [Cr 191 (see also DL exists. T h ,for this case, we record here once more the
- f3oL 92b)). Bask Mentity:
Remark 4. The formal power series arising from the asymp-
$ totic expansion in Theorems 5.4 and 5.11 (letting Re(() -+ oo) are
interesting beyond their truncations mod ~ ( t l q l - ' ) . In important
applications, and notably to itive elliptic operators, these power All the formulas of this section involving
series are convergent, and del? e entire functions. This is part of the
theory of Voltma operators, c.f. [Di 781, Chapter XXIII, (23.6.5.3).
from z to any -Ak is at least cm-O1 with a suitable small constant
may then be used with R(z) replaced by D;/DL(z) in the appli- c. From D I R 2(a) we have that
cations to the spectral case.
Recall from D I R 2, D I R 3 and Theorem 1.8 that for any integer
n > uo we have the formula
so, if IAkI < 2m, we have for those k for which IXJ < 2m, the bound
lak( = CmeO,for a suitable large constant C. With all this, the
first sum can be bounded by
For the remainder of this section we will study the analytic con-
tinuation of the integral given in Theorem 6.2. The proof of Theo-
rem 6.2 shows that the problem in extending the region for which
Symbolically, let us set the integral conver es arises because of the asymptotic behavior of
R(r - t) for -Re($? large. In any case, the following lemma takes
care of the integral over the circle S6.
Lemma 6.3. For any fixed 6 sufticiently small, and any z such
that &(z) is sufficiently large, the integral
When taking the sum of the two integrals on the negative real axis,
it is of course understood that for the second integral, we deal with
the analytic continuation of G(w) over the circle S6.
is holomorphic for all s E C.
We call the result of the following theorem the Hankel formula. Proof. Since Re(z) is sufficiently large, the integrand is abso-
Theorem 6.2. Let s E C be such that &(s) > a 2 + 1. Then lutely bounded on S6.
for any z with Re(%)sufficiently large, we have
Lemma 6.3 implies that the analytic continuation of the Hankel
< +
formula to any s for which Re($) 6 2 1 must take account the
asymptotic behavior of R(z - w) for fixed z and as -&(w) -t oo.
We can use the eneralized Stirling's formula, Theorem 5.4, to
f
extend the Hanke formula in much the same way AS 2 is used to
extend the Hurwitz xi function (see,in particular, Lemma 1.3 and
- PT'OOf. k t T 6 R>o be SU& that 1 41# T for d l k, and let C6.T Theorem 1.5).
denote the contour consisting of: Lemma 6.4. For fixed 6 > 0, for any z with Fk(z) sufficiently
- the lower edge of the cut from -T to -6 in UL; large and for any q and p for which &(q) > Re(p), the integral
- the circle Sa, given by w = 6ei4 for 4 ranging from -a to
a;
- the upper edge of the cut from -6 to -2' in UL;
- and the circle ST,given by w = ~ e ' 4for ) ranging from a
to -7r.
is holomorphic for Re(s) = a > -RR(q) + 1. we have, by Lemma 6.3,
Proof. For any T > 1, the integral over the segments from -T
to -6, both the lower cut and the upper cut, is finite for all s since
the integrand is absolutely bounded. By Theorem 5.4, we have
lim
6+0
[([+ n -
I
[R(z - t) - Bq(z t)] log t dt
which establishes part (a). Note that if a > 00, then, by the
(b) For all s in any fixed compact subset of the half plane convergence condition DIR 2(a),
Re(s) = a > ao,
the convergence
which shows that. the upper bound in (a) approaches zero a. N
approaches m, for fixed s with Re($) sufficiently large. As for (b),
N -+00
I
lim ((5) -
k=l
akX;' = 0, since +N is bounded, if s lies in a compact subset K of the half
plane Re(s) > 00, then l t l is
~ bounded
~ independent of N for all
s E K. This completes the proof of the lemma. 0
is uniform.
Let cp be a suitable function, which will be appropriately char-
acterized below. For any a E R, let &(a) be the vertical line
Proof. If we write logXk = loglXkl + iarg(Xk), we have the Re($) = o in C. Under suitable conditions on cp which guarantee
bound the absolute convergence of the following integral, we define the
vertical transform V,cp of cp to be
which is valid if
a > -Re(Xk) for all k.
The inversion formula (3) is essentially a standard elementary in- +
is the L1-norm of eIu1*~I'(u iu) on the vertical line L(u) By the
version obtained for each individual term from the relation convergence condition DIR 3 and (2), we conclude there exists a
constant C, independent of N ,such that
As before, we put
satisfies the asymptotic conditions AS 1,AS 2 and AS 3. The
m(q) = maxdeg Bp for Re(p) = Re(q). asymptotic expansion of AS 2 is given by (9) and Theorem 7.4
with cp = C.
With this, we can combine Lemma 7.3 and (9) to obtain: Proof. he or em 1.12 shows that if (L, A) satisifies the three con-
vergence conditions, then the associated theta series satisfies the
Theorem 7.4. Let cp E Dom(V), and let {-p) be the sequence asymptotic conditions AS 1 and AS 3. Theorem 7.4 shows that
of poles of yI'. Let u be a positive, sufficientlylarge, real number the meromorphy assumption on C, namely that C is in Dom(V),
such that neither cp nor I' have poles for Re(s) 2 u. Then the implies AS 2.
vertical transform V,cp of cp satisfies the asymptotic condition
AS 2, or, brieffy stated,
To conclude this section, we will use Theorem 7.4 to show how,
given sequences whose theta series satisfies the asymptotic condi-
tions, one can construct a new sequence with the same property.
The following theorem follows immediately from the definitions and
the holomorphy of the exponential function.
+
Proof. If we let s = ut it with u' a large negative number, then Theorem 7.6. Let z E C be fixed and assume that (L, A) is
we have the estimate such that the associated theta series satisifies the three asymp-
totic conditions AS 1, AS 2 and AS 3. Then the sequence
+
(L z, A) satisfies the three asymptotic conditions AS 1,AS 2
and AS 3.
-
-
where IlyI'lll,,~ is the L1-norm of cpr on the line L(u1). By V 2,
Pmej Tke theta k b i o r t asm&kd to (L + z,A) is simply
this integral is finite. Now simply combine Lemma 7.3 and (9), and e-"&,,~(t), hence the proof follows.
let t 0 to conclude the proof. 0
Theorem 7.7. Let r > 0 and assume that (L, A) is such that
With all this, we can now summarize the connection between the associated theta -es s a t i d e s the three apymptotic condi-
the convergence conditions that apply to sequences (L, A) and the tions AS 1, AS 2 and AS 3. Then the theta series associated
asymptotic conditions that apply to the associated theta series. to
({A:''}, A) = (L~I', A)
Theorem 7.8. Let (Ll, A , ) and (L2, A2) be sequences such
satisfies satisfies the three asymptotic conditions AS 1, AS 2 that the associated theta series satisify the three asymptotic
and AS 3. If the exponents {p} in the asymptotic expansion conditions AS 1, AS 2 and AS 3. Assume there exists an
of BrtA near t = 0 are real, then so are the exponents in the E > 0 such that for all j and k, we have
asymptotic expansion of eLIIr,A near t = 0.
.= To study 6 and 83, we shdt apply Lcmma 7.1. For this, at d&e
Example 1. Let L be the sequence of eigenvalues associated
the truncated zeta function to be to a Laplacian that acts on Coofunctions on a compact hyperbolic
Riemann surface X. The parametrix construction of the heat ker-
N-1 nel shows that L satisfies the three asymptotic conditions AS 1,
00
-
AS 2 and AS 3. From Theorem 7.6 we have that L 1/4 satisfies
the three asymptotic conditions, and Theorem 7.7 implies that the
sequence
4
~
- = {,/-I ={~k}
also satisfies the three asymptotic conditions. The sequence TABLE OF NOTATION
Because of an accumulation of conditions and notation, and their
112 + ,Ki - ,/zip = (112 + irk} use in the current series of papers, we tabulate here the main ob-
jects and conditions that we shall consider.
is1 precisely the set of zeros { p } with Im(p) > 0 for the Selberg We let L = {Ak} and A = {ak} be sequences of complex num-
zeta function associated to X. Note that the general Cram& the- bers. To these sequences we associate various objects.
orem proved in [JoL 92b] also proves that the sequence d m
satisfies the three asymptotic conditions AS 1, AS 2 and AS 3. A Dirichlet series or zeta function:
Essentially the same argument, applied in reverse, holds for gen-
eral zeta functions, such as those associated to the theta function
as in CramCr's theorem ([JoL 92bl).
and, more generally, for each positive integer N, the truncated
Example 2. The Dedekind zeta function of a number field Dirichlet series
DIR 3. There is a fixed c > 0 such that for all L sufficiently increasing to infinity. To every p in this sequence, we associate a
large, we have polynomial Bp and we set
bp(t) = Bp(logt).
We then define:
Equivalently, there exists positive constants Cland C2
such that for all k, The asymptotic polynomials at 0:
f ( t )- P'(t) = 0(tR.(~)l
logtlm(q') for t -+ 0.
We will write AS 2 as
Part I1
Given a polynomial Bp a direct calculation shows that there
exist polynomials B;, $ and B f of degree 5 degBp 1 such +
that: A Parseval Formula for Functions with a
Singular Asymptotic Expansion at the Origin
= 9 - I B: (log z )
- S(X) = - 1
0
sin t
lim ~ A ( x=) f (x) for all x E R.
A-00
=- 7 S[At)a'f [t - x).
The following quantitative formulationof the Riemann-Lebesgue
lemma is proved by Barner in fBa 901, Sat8 82,in $21, to which the
reader is referred for a proof.
Lemma 1.4. Assume: Corollary 1.7. Assume g has M derivatives and all the func-
tions g , g('), . .. ,g ( M ) are in BV(R) n L1( R ) .Assume that
(a) s E BV(R).
g ( x ) = O(lxlM+') for 1x1 -,0.
( b ) g ( x ) = O ( ( x l t )for some e > 0 as x 4 0.
Then:
Then the improper integral that follows exists for A > 0 and
satisfies the bound (a) The function g~ has M derivatives ( g A ) ( l ),. ..,( g A ) ( M ) and
Proof. Since
and then we estimate the desired inte a1 in the coarsest way with
f
the sup norm to conclude the proof o the present case. 0
--
-
0
sin Ay sin A(x
3-Y I dy = og(1xI6)for x 4 0,
For the second integral 12, again with u = z - y, we find that (5)
For th.e integral J: let xarbe the characteristic function of [O,Y),
that is put
If g is continuous at y, then
where the interval over which we take the sup norm of g is
20
Using the growth estimate g(x) = O(lxlf) for x near zero and the
This is a convenient expression to plug into an application of Fu-
assumption A'l _< $'I4, we get bini's theorem which gives
1 1
>
For simplicity, we just look at y xo. The other piece for I3is
done in the same way. We then decompose the integral over y 2 xo
into a sum:
I[ sin Ay
dY)ydYl < ;vR(g),
the estimate on the right being independent of A. as asserted. 0
The proof of Proposition 1.10, which is much easier than that of
Proposition 1.8, will be given through the following lemmas. We now consider the integral in Proposition 1.10 in two separate
cases, when x 5 1/A4 aad when x > l/A4.
Lemma 1.11. Assume g E BV(R) n L1(R). Then
00
Case I. Assume x < l/A4. Then there is a universal constant
sin Ay 3 C such that
/g(Y)TdY < ;iv~(g).
1
1
Y . 2 - Y ] dy
5 c,xli4,
where the constant C, is independent of A, and depends only For f,g in the Schwartz space, we assume the elementary Parseval
on g. Form ula
Proof. Let us estimate the integrals separately. In fact, write (f,!l) = (fA,gA).
the integrals as The point of this section is to prove this formula under less restric-
tive conditions. We shall extend conditions of Barner [Bg 901, for
00
T g ( sin A(Y- sin Ay which the formula is true. The basic facts from real analysis (func-
tions of bounded variation and Stieltjes integral, Fourier transforms
y - 2 ")dar - Jg(y)Tdy under smooth conditions) used here are contained in [La 831.
1 1
We call the following conditions from [Ba 901 the basic condi-
tions on a function f:
Condition 1. f E BV(R) n L1(R).
sin Au
sin Au
du+
J g(u+x)-
U
du Condition 2. There exists e > 0 such that
f(x) = f(O)+O(lxle) for x -,0.
+ y)
Its "Fourier transform" as a distribution is the function W,1, such
= JJ
RxR+
f (9) sin A(x
7[ ~ ( x ) + 1-
sin Ay
- xY dy4w
that sin A(x - y ) sin Ay
2-Y
R+ R
1 eit~
= -Wp,JXt) where xt(x) =
6
The following theorem generalizes the Barner-Weil formula [Ba 901. At this point, we are finished with the proof in the case f = a is
a function satisfying the conditions of Lemma 1.3, namely f2.4 = a
Theorem 2.1. Let f satisfy the three basic conditions, and let for sufficiently large A. For the general case, we write the above
(p, 9) be a specid pair. Then integral as a sum
j fAct,wL,ctidt
-A
Its "Fourier transform" as a distribution is the function W,l,(t) As in the proof of Theorem 2.1, the assumptions on (p, 9 ) and f al-
such that low us to interchange the integrals which are absolutely convergent.
By integrating with respect to t, this expression becomes
sin A(s+ y)
M
RxR+ k=O
1 Continuing, we have:
= -W,,,(nt) where xt(x) = elt?
6
The proof of Theorem 3.1 now finishes as did the proof of The-
orem 2.1. By an appropriate extension of Lemma 1.3, choose
an o for which o~ = o for sufficiently large A and the num-
bers o(O), . . . , d M ) ( 0 ) have been chosen to agree with the first
M derivatives of f at zero, and set g = f - a. The above integrals
are linear in the function f , so proving the theorem for g will im-
ply the theorem for f , so we work with g. By Theorem 1.2, the
boundedness of g~ (as stated in Lemma 1.1), and assumption (b)
above, one can apply the dominated convergence theorem to the
first two integrals above. Using Corollary 1.7 and the definition of
the function uk(x), we can write the third integral as
M+Re(po)+61-1+6,
115
54. T h e Parseval Formula for Iw(a + i t ) Recall that the principal part of the theta function is
To conclude our investigation of the regularized harmonic series,
let us show how Theorem 3.1 applies to prove a Parseval formula Po+) = C bp(.)xP,
associated to the regularized harmonic series encountered in $3 of M P <o
)
[JoL 92a]. We will assume the notation defined in 53.
so
Recall as in 54 of [JoL 92a] that the classical Gauss formula e ( ~-)Pee($) = qllog as -+ 0.
states that for Fte(z) > 0,
As in [JoL 92a], let Bz(x) = e-zz6(x). By expanding e-Zz in a
power series, we see that the principal part of B,(x) is
where
eVaz
%(x) = I - e-z and dp(x) = e-'dx. the integral
Then
where the coefficients
Thus, we obtain:
Corollary 4.2. For any w E C with Re(w) > maxk{-Re(Xk)}
and Re(w) > 0, and any a E R+ define
Then
=- +
Iw(a it) =
k<-Re(po)
[La 701 LANG, S.: Algebraic Number Theory, Menlo Park: Add- [Ve 831 VENKOV, A. B., and ZOGRAF, P. G.: On analogues of
the Artin factorization formulas in the s ctral theory of
- ison-Wesley (1970), reprinted as Graduate Texts in Math-
ematics 110, New York: Springer-Verlag (1986). at? ~ i e a s Jd
witat represents-
tions of b c h s i a n groups, Math. USSR-Izu. 21, 435-443
,'a [La 831 LANG, S.: Real Analysis, Menlo Park: Addison-Wesley (1983).
(1983).
[Vo 871 VOROS, A.: Spectral functions, special hctions, and the
[La 851 LANG, S.: Complex Analysis, Graduate Texts in Mathe- Wberg zeta function. Comm. Math. Phya 110, 439-465
matics 103, New York: Springer-Verlag (1985). (1987).
Lecture Notes in Mathematics
[We 521 WEIL, A.: Sur les ''formules explicitd de la thkorie des For information about !Ms. 1-1384
please contact your bookseller or Springer-Vedag
nombres premiers, Comm. Lund (vol. dkdiC ir Marcel
Riesz), 252-265 (1952).
[We 721 WEIL, A.: Sur les formules explicites de la thCorie des nom-
bres, Im. Mat. Nauk (Ser. Mat.) 36, 3-18 (1972). Vol. 1385: A.M. Anile. Y. Cboquet-Brubrt (Edr.). V d 110(: M.-P. M8Uiavin(E!d.X W n d r e d ' A I g h Paul
Relativistic FluidDy.rmia.Seminu, 1987. V. 309pages. Dakdla khriePad Mdliavin. Roorcdiap. 1987-1988.
1989. N.410 p q a . 1989.
Vol. 1386: A. Bellem. C.W. O w . E. Russo (Edr.). VoL 1405: S. Ddrcti (Ed.). OpUmiutios. Fmcedags,
Nu~llericJMethod8 for Ckdinuy Mtkrunlrl E ~ W ~ O I U IW. V. m pga 1989. vd. 1406: L ~reobrw(FA.).
Procadiags, 1987. W.1% p g r ~1989. . Amlytk~dConUlvedPrauiowIII.~ap.
Vol. 1381: M. Pettovi'c. l t a t i v c Method8 for S i d - 1988. nI42 p8ga. 1989.
t l l l s a u I n c h u l o n d P d ~ Z P O L X 1989. ~ 6 ~ ~ Vd. 1401:W.PoWm.Roof1Lc#r. VI, 213pgw 1989.
Vol. 13883 J. Sbi.oQ. T.A. S*suq T. To@ (Ed8.X Vd. 14W W. Ukk,Tnuforrmdom Omups rod Algcbnic
Mrr&rmdul Logic and Applicuio~.M y r . 1987. K-Tbeor)..XU, 4 3 w.1989.
V. 223 pages. 1989. Vol. 1409: E WJror. Cb. Lub&b, M. M . . lEeNumaid
Vol. 1000: S d EdiUonH. Hopf, DiflenntW Oeoramy S o ~ d ~ - A l g e b n i c S y r ( e bym RungaKutta
t
in the h g e . W. 184pag48.1989. Metbodr. VII. 139 pagr~.1989.
Vol. 1410: PJ. Crmu. 0. Gi1-Medrw. A.M. Naveira
(Ms.), Difkrrntirl Ocomscry. Rocerdings. 1988. V. 308
1989.
VoL 141I: B, Jimg (Ed.). TopdoIicll Fixed Point Theory
and Appllutiolu. Rocesdings. 1981VI, 203 mu. 1989.
Vd. 1412: V.V. Kalnbdh. V.M. Zolotarev (Eds.).
Vol. 1391: S. Cmulual.. A. WQO. (Bb.). P d a M t y Stability Probbnu for Stocbutic MoQlr. Roceedings.
T h b o y o a V o c r o r S p r t r l V . ~ ~ , 1 9 8 7 . V m . 4 2 4 1987. x, 388ppl; 1989.
p a p . 1989. V d 1413: S. Wright, U.ic(oslwuof UICInjeuivelll,Factor.
V d 1 3 9 1 : R S f l b o l , R d A l p b n i c ~ . X 2 I 5 ~ I& M pap^. 1989.
1989. Vol. 1414: & Raminz & Amllrw (Ed.), Algebraic
Vd. 1393: N. Bouhrrr.D. Fey4 P. H*rL.0.MolroboQtLi Geome&y a d CospIcx M y & Rocredingr. 1987. VI.
(Ed8.X Stmidm de lWxk du Porcaticl Pmir. No. 9. ISOpyor. 1989.
Rocredinlr.W, 265 p o p . 1989. VoL 1415: M. Lanpi&M. Waldadmidt (Us.). Ciaqu.l)te
Vol. 1394 T.L. 0111, W.W. Zlcbvy (Bb.). No.lhKlr Ans de 301yn6mes. Fifty Years of Polynomials.
S e m I ~ p h ~ d ~ E q ~ a n d A m r c c Pmcecdinlr. o n . 1988.X 235 pages.1990.
Proaediyl. 1987. M,233 paga 1989. Vol. 1416: C Alkn (Bd.). G&mCllie Symplectiquc et
Vol. 1395. K. AlLdi (Ed.), NumbaTbor), Madras 1987. hwmdqw. Roerccdlq* 1988. v. 289 pga 1990.
Roaediqs. m234 wa. 1989. VoL 1411: AJ. Soamem, A. Bhcdiom, E.L. Uvorni
Vol. 13%: L Acardi, W. voa Walddela (l3b.X Que (Ed&), AIgckrk G a w r y . FVocadiny. 1988. V. 320
turn ProbrMlily and AppUeuioar IV. Rocscdtngs, 1987. page& 1990.
VI, 355 pages. 1989. Vol. 1419: M. Mi- (Ed ), Homotopy Tbtory aad
Vol. 1397: P.R. l h a (BaXNupaialh a l y d a Pa- RcLcrQTopia.R0ardl.p. 1968. V. 241 pages. 1990.
dklRoaulng. sd.y,1987. VL 264 ppr. * l. Vol. 1419: P.S. Bulh.. P.Y. Lee. J.L. Mawbin. P.
Muldowney. W.P. Plcffcr (Bds.). New Integrals.
m
Rocrcrdir,~1988. v. wa. 1990.
1
VatI4&YOllbl*LLT~~bIlalAnrlytb
Ocoagl.hoadbaI)#W.W6pages. 1990.
Vol. 1421: HA. Bin&& A Nodincar Theory of
~ ~ X H . 2 1 4 ~ 1990. p l .
[We 72) WEIL, A.: Sur les formules explicites de la thbrie des nom-
bres, Izv. Mat. Nauk (Ser. Mat.) 36, 3-18 (1972). Vol. 1385: A.M. Anile. Y. Cboquet-Brubrt (Edr.).
Reluivirtic Fluid Dynrmiu. Scminu. 1987. V. 308pycr
1989.
Vol. 1386: A. Bellen, C.W. Gear, E. Rsuo (Us.), Vd. 1105: S. Dokdd (Ed.). Optiducion. Roceedinp.
Numeriul Mctboda f a Oldbry M f k k r d d Quaiom. 1988. V. 223 pqa. 1989. Vol. IJ06: L Jmbaen (Ed.).
1987. W, 136 1989. M y t i c Tbea), dCoMid F d o u In. Proceedings.
Vd. 1387: M. 1Wravi%, (tcnoiva McUnnl8 far Shd. 1!#8. W,142 pap. 3999.
t r n e a u I l u & d ~ ' l l a a . X M J p r p ~ 1 9 9 9 . Vol. lUn: W.PoblqRooflbxy. VI,213 pqea. 1989.
Vol. 1388: 1. Sbi.ob. T.A. S h u n , T. TtlrpC (Eds.),
Mrtbermtial Logic ud AppUcrtionr. Roaedi.y. 1987.
V, 223 paps. 1989.
Vol. 10b0:SecondWtioa.H. Hopf, Mff& Gmmet~~
in the L ~ W, E 184pyer. 1989.
Vol. 1389: E Bllliea C Wbum(Ed&). Algebraic Cuwr Vd. 14la F.J. Cmnh 0. GiI-A4edr~0,A.M. Navein
and FYojecdveGamwry.Proctdiqh 1988. V. 288 pap.
1989. (Ed&), Diffacnt*I G a w t r y . Roceed*y. 1988. V. 308
pages. 1989.
Vol. 1 3 9 0 : 0 . ~ P n t o . L . ~ ~ . ~ . S I o c b w l c p n i .VoL a).
l 1411: B. J i a q Topological FixedPoint Tbcay
Differrotid BpDItioar ud Applicrdolu If. Pmcdiny. aedpd~ppllertion~ Roacd&y. 1988. VI. 203 pages. 1989.
1918. VI. 258 pyol. 1999. Vd. 1412: V.V. wrl.lbdlor. V.M. Zolotuev (Eds.).
Vol. 1391: S. Crmh.*. A. Wcm. (Ma.), Robability Smbility Robbnu for Stocbutic Modela. Procetdinga.
T b c a y o n V ~ S p r c u I V . ~ y . 1 9 8 7 . M I . 4 2 4 1987. x 380 pqa 1 m .
pages. 1989.
Vol. 1392: R SilhoL Rul Algcbnic Surf-. X 215 pages.
1989. Vol. 1414: E. Runires dc Amllrw (Ed.). Algebraic
Vol. 1393: N. Boulmu. D. k y d , F. Hkrds0. MoWodzki Oeolnay ud Complex Anrlyais. Prooeedinga. 1987. VI.
(Eds.). StnJtuinds Tb(orio du PoccPakl M r . No. 9. 180pgu. 1989.
Roacdiyr. VI. 265 ma. 1989. Vol. 141%M. Laagwia,M. Wddadddt (Eds.). Cioqumtc
Vol. 1394: T.L. 0111, W.W. Zrhry w.). Noalillur Ana de Polynbmer. Fifty Yeur of Polynoriirls.
Semiproup. P a W Diffaw~pirlEqurtiom ud A m r t m . Roceediqs, 1988. IX. 235 pagcs.1990.
Roceedinga 1987. IX, 233 prpr. 1989.
Vol. 1416: C. Alkrc (Ed.). OColdoic Symplcctique et
Vol. 1395: K. Alldi (Ed.), Number lbxy.h d w lM7. Mbdque. Roceedinp, 1988. V. 289 pager. 1990.
Roaediyr. W, 234 paps. 1989. Vol. 1417: AJ. Sonunew A. B i d o n . EL. L i v m i
Vol. 13%: L A d . W.vor W W & (Ed&). Quu- (Ed&), Alpbnic Geometry. hxedingr. 1988. V. 320
turn Probability ud AppUcleio~N. Roceediry, 19111. pages. 1990.
VI. 355 WU 1969. Vol. 1418: M. Mialun (Ed. ). Homotopy Tbeory and
Vol. 13W P.R. 'ILwr (Ed.). NmaiaiM y & md Pa- RslrrcdTopies. h c d i y l . 19s. V. 241 pages. 1990.
nlkl Pmccning. Seminu, 1987. Vl, 264 w.1989.
Vol. 1419: P.S. Bullen. P.Y. Lee. J.L. Mrwhin. P.
Vol. 139& LC. Kim, B.H. W. (Ed&). Group - Muldowaay. W.F. Pfeffsr (Wa.). New Integrrls.
K o m I ~ ~ V . 1 8 9 p l 1989. r r -19n.v.aQZprlu. 1990.
Vot. t399: W.-t. mala, K L.tp mf. lbiUNm& d Vot. W&U.WbW, A. Togu~U@t&.), R u t Asdytlc
Compkx Manifolds. Roacdiry, 1988. V. 171 pycr Olomrc*.- 19U.N.#6pyu. 1990.
1999. Vol. 1421: H.A. B4iomi. A Nonlinear Theory of
Vol. 1400: U. J u MiadMoUve~ud Alpk.ic b ~ ~ X & 2 11990. 4 ~ .
lhlry.Ma246psp. 1990. Val. 1411: V. VHL.I ($d). Cwplcx Geometry and
Vol. 141: J. A W r W Qldr,X Set Tbmq d k J y d ~Fmmdllp.
. IW1. V. 109 p a p . 1990.
I I a ~ R P a d l l l pS l l . V , T n w ~ ~lm9. . V d lltl: U k I(.eLr* Strblr hmompy Groups of
Vd. 1 ~ . C . ~ ? . ~ , B . H w r r o g J . - I , . h l y A Caqukc-AubtdACpoub. WII. 330 pages.
E B & f . ~ t t s r ~ ~ ~1990.1 9 a
V,249pe#s. 1989. Vd. 1424: P.R bw*8. J.H. Rawdey, T w i w Theory
Printing: Wcihert-DrucL GmbH, Dumrt.dt
Vol. 140%B. Shoom (Ed.). Ceasluairl Opdmiudocl. f a ~ S p l l n a r i c S p r c s r I I L1 1 2 ~1990.
~ .
Bindiw: B u & SablUtw,
~ OrOnrtrdt S d n a , 1986. V. 314 pe#a. 1909.