On the decomposition of symmetric powers of automorphic representations for \(\mathrm{GL}(3)\) and \(\mathrm{GL}(4)\) (Q2167500)

Let \(F\) be a number field, and \(\pi\) be a cuspidal automorphic representation of \(\mathrm{GL}_n(\mathbb{A}_{F})\). Let \(\mathrm{Sym}^{k}:\mathrm{GL}_{n}(\mathbb{C}) \rightarrow \mathrm{GL}_{m}(\mathbb{C})\) be the \(k\)-th symmetric power map. The Langlands functoriality conjecture predicts the existence of an automorphic representation \(\mathrm{Sym}^{k} \pi\) of \(\mathrm{GL}_{m}(\mathbb{A}_{F})\) such that its Satake parameters are the image of the Satake parameters of \(\pi\) under \(\mathrm{Sym}^{k}\). In general, \(\mathrm{Sym}^{k} \pi\) is not cuspidal, but rather an isobaric sum \[ \mathrm{Sym}^{k} \pi = \Pi_1 \boxplus \cdots \boxplus \Pi_r, \] where \(\Pi_i\) is a cuspidal automorphic representations of \(\mathrm{GL}_{m_i}(\mathbb{A}_F)\) and \(m_1 + \cdots + m_r = m\). In terms of L-functions, this means \[ L(s, \mathrm{Sym}^{k} \pi) = \prod_{i=1}^{r} L(s, \Pi_i). \] Then it is a basic problem to determine the cuspidality of \(\mathrm{Sym}^{k} \pi\). The case \(n =2\) has been studied by \textit{D. Ramakrishnan} [Contemp. Math. 488, 237--256 (2009; Zbl 1262.11064)]. The paper under review gives some bounds on the number \(r\) of isobaric summands when \(n = 3,4\) under some assumptions on the functoriality and cuspidality of functorial lifts. The paper also discusses when the bounds are sharp. The main tool in the proof is the Rankin-Selberg \(L\)-function \[ L(s, \mathrm{Sym}^{k} \pi \times \widetilde{\tau}), \] whose order of poles at \(s = 1\) detects whether a cuspidal automorphic representation \(\tau\) is an isobaric summand. The author also uses some interesting identities among partial automorphic \(L\)-functions such as \[ L^{X}(s, \pi, Sym^{k-1} \times Std) L^{X}(s, \pi, Sym^{k-3} \times \wedge^{3}) = L^{X}(s, \pi, Sym^{k})L^{X}(s, \pi, Sym^{k-2} \times \wedge^{2}) \] in case \(n=3\), where \(X\) is a finite set of places.