Let $F$ be a number field. For an irreducible cuspidal automorphic representation $\pi$ of $\operatorname{GL}_n(\mathbb{A}_F)$, we say that $\pi$ is symplectic (or orthogonal) if $L(s,\pi,\bigwedge^{2})$ (or $L(s,\pi,\operatorname{Sym}^2)$) has a pole at $s=1$.

I am wondering whether if $\pi=\bigotimes \pi_v$ is symplectic (or orthogonal), then $\pi_v$’s are also symplectic (or orthogonal) for all places $v$?

(Here, $\pi_v$ is symplectic (or orthogonal) means that its corresponding Weil–Deligne group representation by local Langlands correspondence is of such type.)

Any comments are welcome!

  • $\begingroup$ It's easy to see that ($\pi$ is either symplectic or orthogonal) $\Leftrightarrow$ ($\pi$ is selfdual) $\Leftrightarrow$ ($\pi_v$ is selfdual for all $v$), but it's not so obvious if the local factors are symplectic / orthogonal if $\pi$ is. $\endgroup$ Jan 2, 2022 at 8:03
  • $\begingroup$ @DavidLoeffler, thank you very much for the helpful comment! The definition of orthogonal (or symplectic) of $\pi_v$ I mean is that if $\phi: WD_F \to \operatorname{GL}(n)$ is the Weil-Deligne representation corresponding to $\pi_v$, then we say $\pi_v$ is orthogonal (or symplectic) if the image of $\phi$ can be replaced to $\operatorname{SO}(n)$ (or $\operatorname{Sp}(n)$.) So if $\pi$ is symplectic (orthogonal), then it seems $\pi_v$’s are symplectic (or orthogonal) for all places $v$. $\endgroup$
    – Andrew
    Jan 2, 2022 at 13:15

1 Answer 1


Here is a proof of the claim using results from Arthur's monograph The Endoscopic Classification of Representations: Orthogonal and Symplectic Groups.

Let $N = 2n$ be an even integer, and $\pi$ a cuspidal automorphic representation satisfying $\pi^\vee \cong \pi$.

  • Since $\pi \cong \pi^\vee$, the Rankin-Selberg $L$-function $L(\pi \times \pi, s)$ has a simple pole at $s = 1$. We have the factorisation $L(\pi \times \pi, s) = L(\pi, S^2, s) L(\pi, \wedge^2, s)$ and both factors are non-vanishing at $s = 1$, so exactly one of them must have a pole; thus $\pi$ is orthogonal or symplectic, in the sense of the question, but never both.
  • The discussion preceding Theorem 1.5.3 of Arthur shows that $\pi$ defines an element of his set $\tilde{\Phi}_{\mathrm{sim}}(N)$ of global parameters, and this lies in the subset $\tilde{\Phi}_{\mathrm{sim}}(G)$ for a uniquely determined quasi-split group $G$ whose Langlands dual $\widehat{G}$ is either $SO_{2n}$ or $Sp_{2n}$.
  • Theorem 1.5.3 of Arthur shows that $\widehat{G}$ is symplectic if $\pi$ is of symplectic type, and $\widehat{G}$ is orthogonal if $\pi$ is of orthogonal type. (This is a very deep theorem, despite sounding like a tautology!)
  • Theorem 1.4.2 of op.cit. now shows that there is a cuspidal automorphic representation $\sigma$ of $G$ such that, for every $v$, the Weil–Deligne representation associated to $\pi_v$ is the image in $GL_N$ of the ${}^L G$-valued parameter associated to $\sigma_v$. So, in particular, it is symplectic (resp. orthogonal) if $\pi$ is.
  • $\begingroup$ thank you so much for sharing your knowledge and wisdom. It helped me a lot! In the meanwhile, may I ask one more question? If $\phi$ is the Weil-Deligne representation associated to $\pi_v$, we can decompose $\phi$ as $\phi_0 + \phi’ + \phi’^{\vee}$, where $\phi_0$ is the sum of irreducible subrepresentation of $\phi$ of the same type with $\phi$ and $\phi’$ is the sum of irreducible subrepresentation of $\phi$ which is not of the same type with $\phi$. Then I am wondering if it is possible that $\phi_0$ could be zero? In that case, the component group of $\pi_v$ is trivial. $\endgroup$
    – Andrew
    Jan 4, 2022 at 2:30
  • $\begingroup$ I am wondering that $\phi_0$ could be zero at places v where $\pi_v$ is unramified. I think it is possible if $\pi$ is of symplectic type cuspidal automorphic representation of $GL(2)$. $\endgroup$
    – Andrew
    Jan 4, 2022 at 6:36
  • $\begingroup$ It's not clear to me if such a decomposition is always well-defined, but for a symplectic-type $GL(2)$ rep, at almost all finite places the representation will look like $\phi_v \cong \chi_v \oplus \chi_v^\vee$ for a 1-dimensional unramified character $\chi_v$. $\endgroup$ Jan 4, 2022 at 7:07
  • $\begingroup$ Thank you very much for the comment! In the $GL(2)$ case, I also think as you because square-integrable representation cannot be unramified. So the component group $S_{\phi_v}$ of $\phi_v$ should be trivial group at unramified places $v$. Thank you very much again! $\endgroup$
    – Andrew
    Jan 4, 2022 at 7:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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