Вы находитесь на странице: 1из 3

Maass Cusp Forms

J.-M. Deshouillers, H. Iwaniec, R. S. Phillips, and P. Sarnak


PNAS 1985;82;3533-3534
doi:10.1073/pnas.82.11.3533
This information is current as of March 2007.

This article has been cited by other articles:


www.pnas.org#otherarticles
E-mail Alerts Receive free email alerts when new articles cite this article - sign up in the box at the top
right corner of the article or click here.
Rights & Permissions To reproduce this article in part (figures, tables) or in entirety, see:
www.pnas.org/misc/rightperm.shtml
Reprints To order reprints, see:
www.pnas.org/misc/reprints.shtml

Notes:
Proc. Natl. Acad. Sci. USA
Vol. 82, pp. 3533-3534, June 1985
Mathematics

Maass cusp forms


(L function/Teichmfiller space)
J.-M. DESHOUILLERS*, H. IWANIECt, R. S. PHILLIPSt, AND P. SARNAK*
*Universit6 de Bordeaux I, F-33405 Talence, France; Institute for Advanced Study, Princeton University, Princeton, NJ 08540; and tDepartment of
Mathematics, Stanford University, Stanford, CA 94305
Communicated by Peter D. Lax, January 24, 1985

ABSTRACT It is shown that, under certain standard as- We state this more precisely: Let FO(q) be the Hecke con-
sumptions, such as extended Riemann hypotheses, the scatter- gruence subgroup of level q; i.e., IO(q) = {(a) E SL(2,
ing matrix 4(s) for generic r S SL(2, R) is unexpectedly of Z); q c}. For simplicity we assume that q is prime. Let Q(z)
order 2. This leads to the conjecture that the generic cofinite r be a holomorphic cusp form of weight 4 for FO(q). Assume
has very few Maass cusp forms. further that Q is a Hecke eigenform. Let uj be, as above, an
orthonormal basis for the Maass cusp forms for 1O(q). These
In his 1954 Gottingen lectures on the theory of Eisenstein too are assumed to be Hecke eigenforms. Denote by L(Q 0
series, Selberg (1) raises the question of the asymptotic den- uj, s) the Rankin-Selberg L function of Q and u;. We assume
sity of the poles of the Eisenstein series. For r a cofinite that we have normalized this function so that it has a func-
(noncompact) discrete subgroup of PSL(2, R), let +(s) be the tional equation s -+ 1 - s, it is entire, and of course it has an
determinant of the scattering matrix (2, 3). Selberg shows Euler product [it has a factor of ;4(2s) where 4q(s) = (1 -
that as R -X 00 q-S)g(s) to make it entire]. The extended Riemann hypothe-
sis referred to earlier is that L(Q 0 uj, s) satisfies the ana-
M(R) = - -('/2 + ir)dr = O(R2). [1] logue of the Riemanp hypothesis. Actually it is not quite this
that we need but rather the analogue of the Lindelof hypoth-
esis (5).
The poles of 4 coincide with those of the Eisenstein series
and M gives a measure of the density of these poles. Selberg The Extended Lindelof Hypothesis
also shows that when F is a congruence subgroup then M(R)
= O(R log R). He asks the question as to what is the true size For E > 0,
of M(R) for a general F.
The importance of this question lies in its intimate connec- L(Q 0 uj, 1/2 + it) << [(1 + ItI)(1 + IrjI)]e(coshTrj)1/2. [5]
tion with the problem of the existence of Maass cusp forms
for F. Denote by b the upper half plane and by A the Lapla- The factor (cosh irr1)'/2 comes from the normalization 11uj112
cian for F\I with its Poincard metric. Let ul, u2, ... be a = 1. As in the case of the classical zeta function the extended
basis for the Maass cusp forms for F\b (2). Thus uj is a cusp Lindelof hypothesis follows from the extended Riemann hy-
form and satisfies pothesis and the Ramanujan conjectures (unpublished
work).
AUj + (1/4 + rj)uj = O- [2] For a given real number r, we define %r to be the subspace
of Maass cusp forms of Fo(q) with eigenparameter r and we
It is known (1-3) that if N(R) = #{ jI rj ' R} then set m(r) equal to the dimension of %r. At present all that is
known for sure is that m(r) <<rilog r. In the case of the
M(R) + N(R) Vol(\i)R2 [3 modular group it is conjectured that all eigenvalues are sim-
ple; i.e., m(r) c 1. For a congruence subgroup the analogous
conjecture would be that m(r) is bounded. We shall say that
In particular it follows from 3 and the previous remarks that JO(q) has a cusp form degeneracy of order 3 if
for a congruence subgroup there are a lot of cusp forms, in
fact m(r) <<r' [6]
N(R) -o4Fi)R . [4] Finally for any F let T(F) denote its Teichmuller space (6).
The assertion that a given property holds generically in a
Teichmuller space means that it holds except on a set of first
In general, however, it is not known which, if either, of the Baire category.
two terms in 3 is responsible for the asymptotics. On the THEOREM 1. Fix q and set T = T(Fo(q)). Assuming the
basis of the above result for congruence subgroups, Selberg extended Lindelof hypothesis for the L functions L(Q 0 uj,
conjectured that N(R) is always dominant. In its strongest s) and that the degeneracy of cusp forms for Fo(q) is of order
form his conjecture is that for every F there is an &O > 0 such I3, then for generic F in T we have
that M(R) =O(R2--c)
Our main result may be stated as follows: Under certain
hypotheses (such as an extended Riemann hypothesis), the M(R) >> R2-p-E V E> 0. [7]
generic F does not satisfy the strong form of the above con-
jecture. As mentioned above, we expect that /3 may in fact be cho-
sen arbitrarily small for Fo(q). In this case, we would have
The publication costs of this article were defrayed in part by page charge
for generic F in T the lower bound
payment. This article must therefore be hereby marked "advertisement"
in accordance with 18 U.S.C. §1734 solely to indicate this fact. M(R) >> R2'-, V E > 0,

3533
3534 Mathematics: Deshouillers et aLPProc. Nad Acad Sci. USA 82 (1985)
contrary to the strong form of Selberg's conjecture.
The proof of Theorem I is based on the methods devel-
#{Ilrjl . R IL(Q 0 uj, '/2 + irj)l 7 0) >> R2- e
oped in refs. 4 and 7. In ref. 7, it was shown that destroying V E> 0. [g9
cusp forms under a deformation is closely related to the van-
ishing or not of the numbers L(Q 0 uj, '/2 + irj) (these are Parenthetically we note that assuming only the Ramanujan
special values for these L functions!). On the other hand us- conjectures, one can prove
ing the methods in ref. 4 we can prove Theorem 2.
THEOREM 2. There is C > 0 depending on Q such that #{Ilrjl R L(Q 0 u>, /2 + irj) # O}»>>R9/8-e
VE >0 [101
E h ) IL(Q 0 uj, '/2 + irj)12 - C R2log R. [8]
JrjI.R cosh(irrj) (unpublished work).
The estimate 9, together with the previous remarks and
This shows that on average these numbers are not zero. Lemma 1 below, are the main ingredients in the proof of The-
However, in applying this result, the presence of multiple orem I for simple eigenvalues.
eigenvalues introduces an added complication. In this case Let {pj(E)} denote the eigenparameters of the pseudo-La-
another basis for the Maass cusp forms, which we call the placian A(e); these include the cusp eigenvalues rj(e). As
Kato basis and denote by {vj}, plays a more central role than shown in ref. 7, pj(E) depend analytically on e.
the {Ju} (see below). In any case Theorem 2 allows us to LEMMA 1. For any ,q there is E. > 0 such that
prove, without any assumptions at all, that an infinite num-
ber of zeros of the Selberg zeta function Zr~,q)(S) move off Ipj(e)I 21pj(iq) + l for 1c - qj < E,.
the line Re(s) = 1/2 under a deformation in T. In fact at least
R'-I of the R2 zeros whose height is less than R move off. In the case that ro(q) has a multiple eigenvalue r, we can
[See ref. 2 for the definition and properties of Z(s).] Theorem choose a basis vj(0), v2(0), . . ., Vk(O) of the r eigenspace 'tr
1 together with the methods of this paper further support the so that v,(e), . .., VM(e) are analytic in e. We call this basis of
possibility alluded to in ref. 7, which we now state as a pre- ,r the Kato basis (10). Unfortunately we have little control
cise conjecture. over this basis and it may have no relationship with the more
Conjecture. For generic r E T(Fo(q)) there are only a finite natural Hecke eigenbasis of Wr. In particular we can no long-
number of Maass cusp forms; that is, N(R) in 3 is a bounded er rely only on the estimate 5 and it is this that forces us to
function. make the assumption on the order of degeneracies stated in
We remark that this conjecture is not that far out of reach. Theorem 1. Complete details of the proofs will appear in ref.
If we ignore the technical difficulties of degeneracy of the 11 and elsewhere.
eigenvalues then the conjecture would follow from knowing
that all but a finite number of L(Q 0 u;, 1/2 + rj) are not zero. 1. Selberg, A. (1954) Gottingen Lectures.
All evidence points to this being the case. 2. Hejhal, D. (1983) S.L.N., 1001.
We turn to a brief outline of the main steps in the proof of 3. Lax, P. & Phillips, R. (1976) Ann. Math. Stud.
Theorem 1. For simplicity assume first that the cusp spec- 4. Deshouillers, J. M. & Iwaniec, H. (1982) Invent. Math. 70,
trum of r0(q) is simple (or that we are dealing with the simple 219-288.
part of it). For ye, a real analytic curve in T with tangent 5. Titchmarsh, E. (1951) The Theory of the Riemann Zeta Func-
vector Q at ro(q), it is shown in ref. 7 that if L(Q 0 u>, /2 + tion (Clarendon, Oxford).
irj) 7 0, then for all but a countable number of e, the corre- 6. Ahlfors, L. (1966) Lectures on Quasi Conformal Mappings
sponding uj(e) is not a cusp form. [Here uj(E) depend analyti- (Van Nostrand-Reinhold, Princeton, NJ).
cally on E and are eigenfunctions of the pseudo-Laplacian Aa 7. Phillips, R. & Sarnak, P., Inventiones, in press.
(see refs. 8 and 9 for the definition of Aa).I It follows that the 8. Colin de Verdiere, Y. (1983) Ann. Inst. Fourier 32, 275-286.
9. Colin de Verdiere, Y. (1982) Ann. Inst. Fourier 33, 87-113.
set of E for which all of the uj(e) are not cusp forms whenever 10. Kato, T. (1966) Perturbation Theory for Linear Operators
L(Q 0 uy, 1/2 + irs) # 0 is all but a countable set of y,. The (Springer, Berlin).
destruction of cusp forms is readily related to the growth of 11. Phillips, R. & Sarnak, P., Comm. Pure Applied Math., in
M(R) (see ref. 3). It follows from Theorem 2 and 5 that press.

Вам также может понравиться