I would like to understand better the L-function and epsilon factors attached to a Gelbart-Jacquet lift. Does the fact of being a Gelbart-Jacquet lift translates into strong properties concerning these objects? In particular I would like to understand, for $\pi$ an automorphic representation of $GL(3)$ that is a Gelbart-Jacquet lift :

  • could $\varepsilon(1/2, \pi)$ be zero, or can all the twists by quadratic characters $\varepsilon(1/2, \pi, \chi)$ be zero?
  • can $L(1/2, \pi)$ or $L(1/2, \pi, \chi)$ be zero?
  • are these situations possible for other automorphic representations, that are not necessarily Gelbart-Jacquet lifts?

Any clue on how to understand Gelbart-Jacquet lifts (computationally and in practice I mean) is welcome!


The function $s \mapsto \epsilon(s, \pi)$ has the form $s \mapsto A e^{Bs}$ for some constants $A, B$, so it vanishes nowhere on $\mathbb{C}$. That has nothing to do with being a GJ lift, it's a general property of epsilon factors. Similarly, being a GJ lift doesn't really tell you much about the value at $s = \tfrac{1}{2}$: analytically, the standard L-function of a Gelbart--Jacquet lift doesn't look much different from that of any other automorphic representation of $GL(3)$.

What really separates the GJ lifts from other kinds of cuspidal auto reps of $GL(3)$ is that they are self-dual; so the Rankin--Selberg L-function of $\pi$ with itself, $L(\pi \times \pi, s)$, has a pole at $s = 1$ if and only if $\pi$ is a GJ lift (I hope I've remembered that correctly). See also Peter Humphries' excellent answer to this question: How strong is the requirement of being a Gelbart-Jacquet lift?

  • 1
    $\begingroup$ For what it's worth, I've found Gelbart and Jacquet's paper (doi.org/10.24033/asens.1355) quite useful for computing the local epsilon factors and local $L$-functions, given one knows the local data. $\endgroup$ – Peter Humphries Sep 15 at 16:32
  • $\begingroup$ Thanks a lot for these clarifications. Even if the epsilon factor is nonzero, can we have L-value zero at 1/2? Is that possible for instance for all characters ? $\endgroup$ – Damon Sep 17 at 7:09
  • $\begingroup$ Certainly it is expected that for any cuspidal auto $\pi$ on $GL(3)$, and any finite set of places $S$, we have $L(\pi, \chi, 1/2) \ne 0$ for all but finitely many finite-order characters $\chi$ unram outside $S$; but I don't think this is known (whether or not $\pi$ is a GJ lift). $\endgroup$ – David Loeffler Sep 17 at 9:02
  • $\begingroup$ With that being said, the $L$-function of a self-dual Gelbart-Jacquet lift $\Pi$ whose epsilon factor satisfies $\epsilon(1/2,\Pi) = -1$ trivially vanishes at $s = 1/2$. $\endgroup$ – Peter Humphries Sep 17 at 12:29
  • $\begingroup$ Agreed, but twisting by a Dirichlet char will change the $\epsilon$-factor. $\endgroup$ – David Loeffler Sep 17 at 14:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.