26.12.2013 Views

A NULLSTELLENSATZ FOR AMOEBAS

A NULLSTELLENSATZ FOR AMOEBAS

A NULLSTELLENSATZ FOR AMOEBAS

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

ENDOSCOPIC LIFTING 491<br />

CONJECTURE 1<br />

Let π denote an irreducible generic cuspidal representation of the group H (A), and let<br />

ɛ denote an irreducible cuspidal representation of H 2 (A) according to Section 6.1(1) –<br />

(5). Assume that τ = τ(ɛ) is a cuspidal representation. Then the following are<br />

equivalent.<br />

(1) The partial tensor L-function L S (π × ɛ, s) = L S (π × τ(ɛ),s) has a simple<br />

pole at s = 1.HereS is a finite set of places, including the archimedean ones,<br />

such that outside of S, all data is unramified.<br />

(2) There is a choice of data such that the period integral P(π, τ(ɛ)) is not zero<br />

for some choice of data.<br />

(3) There is a generic cuspidal representation σ of H 1 (A) such that π is the weak<br />

endoscopic lift from σ and ɛ.<br />

Two parts of the conjecture are, in fact, a theorem. The implication that (1) implies<br />

(2) follows from the usual Rankin-Selberg theory. Indeed, it follows from the above<br />

references that when we unfold the global integrals, we represent the above tensor<br />

product L-functions. It also follows from the above references that for any finite<br />

place, data can be chosen so that the integral is not zero. From this, it follows that<br />

if the partial L-function L S (π × τ(ɛ),s) has a simple pole at s = 1, the period<br />

integral P(π, τ(ɛ)) is not zero for some choice of data. The implication that (3)<br />

implies (1) follows from the definition of the weak lift. Indeed, if we assume (3), then<br />

L S (π × ɛ, s) = L S (σ × ɛ, s)L S (ɛ × ɛ, s). Since all data are generic, we know from<br />

[CKPS] that all representations have a lift to an automorphic representation of GL.<br />

By the result of [Sh], we know that the tensor product L-function of two automorphic<br />

representations does not vanish at s = 1. From this, it follows that L S (π × ɛ, s) has a<br />

simple pole at s = 1.<br />

We note that the implication that (2) implies (1) in Conjecture 1 should, in<br />

principle, follow also from the Rankin-Selberg integral representations given above.<br />

In this part, we study the implication that (2) implies (3). We use the lifting studied<br />

in the previous sections to prove this implication in one case. The other four cases<br />

stated in Conjecture 1 are different. The main problem is that the representations ɛ<br />

involve a representation µ(ɛ) defined on a classical group. Therefore, another step<br />

is required. This step requires the study of the descent of a certain residue of an<br />

Eisenstein series and to prove that this descent is by itself a residue. At this point, it is<br />

not clear how to prove this.<br />

THEOREM 6<br />

Let H and H i be as in Definition 2(5). In other words, suppose that H = SO 2(n+m)+1 ,<br />

that H 1 = SO 2n+1 , and that H 2 = SO 2m+1 . Then Conjecture 1 holds.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!