Take a cuspidal automorphic representation $\pi$ of $GL(3)$ over a number field. My question is quite straightforward and can be related to this one :

Can $L(1, \pi, \mathrm{sym}^2)$ be zero? If yes, is there any extra assumption ensuring it cannot?

I know that we have the splitting $L(s, \pi \times \tilde{\pi}) = L(s, \pi, \mathrm{sym}^2)L(s, \pi, \wedge^2 \pi)$. If we assume $\pi$ self-contragredient for instance, then the Rankin-Selberg convolution has a simple pole at $s=1$. Either $L(s, \pi, \mathrm{sym}^2)$ has a pole at $s=1$ (Gelbart-Jacquet lift case) and in that case it does not vanish ; or it has none and in that case the question rephrases as : can $L(s, \pi, \wedge^2 \pi)$ have a pole of order 2 at $s=1$?

Any new insight or reference on this precise question is welcome.


For $GL(3)$, the exterior square $L$-function $L(s,\wedge^2\pi)$ is entire as it agrees with $L(s,\tilde\pi\otimes\omega)$, where $\omega$ is the central character of $\pi$. Therefore, $L(1,\mathrm{sym}^2\pi)=0$ would imply that $L(1,\pi\otimes\pi)=0$, contradicting a result of Shahidi (1980). The best known zero-free region for general Rankin-Selberg $L$-functions is due to Brumley (2012): see the Appendix here.


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.