In this question, we will follow the notations and definitions from the book "Automorphic Representations and L-Functions for the General Linear Group" by Goldfeld and Hundley. For simplicity, let's focus of $GL_2$. The definition of an (cuspidal) automorphic representation with central character $\omega$ is given in Definition 5.1.12, (5.1.14), where $\omega$ is a unitary Hecke character $$\omega: \mathbb{Q}^{\times} \backslash \mathbb{A}_{\mathbb{Q}}^{\times} \rightarrow \mathbb{C}^{\times}$$

i.e. it is a smooth $(\mathfrak{g},K_{\infty}) \times GL(2,\mathbb{A}_{\text{finite}})$-module which is also isomorphic to a subquotient of the complex vector space of adelic (cusp) automorphic forms. As I literally know nothing about the theory of automorphic representations, I have two questions to bother the community.

  1. Suppose we have two cuspidal automorphic representations with the same central character $\omega$, and if they are isomorphic as $(\mathfrak{g},K_{\infty}) $ modules, are they isomorphic as automorphic representations? If not, do the two automorphic representations have any relations?

  2. Given a $(\mathfrak{g},K_{\infty}) $ module, how to check whether there exists a cuspidal automorphic representation whose local representation at $\infty$ is just this module? E.g. in order for this to be true, are there restrictions on this module? On the other hand, if there exists such a cuspidal automorphic representation, is it unique?

Any references are welcomed!

  • $\begingroup$ Did you look at the $GL_1$ case, if $\chi$ is a Dirichlet character modulo $q^k$ with $q$ prime then $\omega(x) = sign(x_\infty)^{(1-\chi(-1))/2} \chi(\frac{x_q}{q^{v_q(x_q)}} \bmod q^{k}) \prod_{p \ne q} \chi(p)^{-v_p(x_p)}$ is an automorphic form and representation. $\endgroup$
    – reuns
    Jan 6, 2019 at 17:13
  • $\begingroup$ @reuns Yes I did, and this also shows the $\infty$ part only contains very limited information. I thought the $GL_2$ case might be very different. $\endgroup$
    – Wenzhe
    Jan 6, 2019 at 19:07

1 Answer 1


1) No, two distinct cuspidal automorphic representations can have the same underlying $(\mathfrak{g},K)$-module. In particular, if the $(\mathfrak{g},K)$-module is a weight $k$ discrete series, then the cuspidal automorphic representations corresponds to a holomorphic newform of weight $k$, and there are plenty of these. On the other hand, if the $(\mathfrak{g},K)$-module is a principal series representation, then there conjecturally are very few cuspidal automorphic representations that share the same $(\mathfrak{g},K)$-module, though they do exist: the known ones are quadratic twists. See my recent paper for further discussion on this problem.

2) For weight $k$ discrete series, there are formulae for the dimensions of spaces of holomorphic cusp forms of weight $k$, so this is easily done. For principal series, this is a very hard question. These correspond to Maass cusp forms, so you are asking whether given a positive real number $\lambda$, there exists a Maass cusp form with Laplacian eigenvalue $\lambda$. We only explicitly know a very thin sequence of such Laplacian eigenvalues, namely those corresponding to Hecke Größencharaktere, whose Laplacian eigenvalues are related to logarithms of fundamental units of real quadratic fields.

  • $\begingroup$ Hi Peter, I have a question on your answer of 1). If I assume that the two cuspidal representations also share one of (resp almost all, all) their principal series component? Can we then say that they are the same? Also, the weight $k$ you're talking about is contained in the $\omega$ character right? $\endgroup$ Jan 9 at 21:17
  • $\begingroup$ No, you can have two automorphic representations that are quadratic twists of one another and have the same local components at 50% of their places. And no, the weight of a discrete series is not contained in the central character, except that there is a compatibility $\omega(-1) = (-1)^k$. $\endgroup$ Jan 9 at 22:11
  • $\begingroup$ Ok thank you very much! $\endgroup$ Jan 9 at 23:24

Your Answer

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

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