On the gamma factor of the triple product \(L\)-functions (Q626828)

Let \(f\in S_{k}\left(\text{SL}_{2}(\mathbb{Z})\right)\) be a Hecke-normalized cusp form of weight \(k\) on \(\text{SL}_{2}(\mathbb{Z})\) and suppose that \(f\) has the Fourier expansion \[ f\left(z\right) = \sum_{n\geq1}a_{f}\left(n\right)q^{n},\quad q=e^{2\pi iz}. \] Let the normalized Fourier coefficients (Hecke eigenvalues) of \(f\) be \(\lambda_{f}\left(n\right)=a_{f}\left(n\right)n^{\frac{1-k}{2}}\). We define the (analytically normalized) \textit{L-function} associated to \(f\) by \[ \begin{aligned} L\left(f,s\right) &= \sum_{n\geq1}\lambda_{f}\left(n\right)n^{-s} = \prod_{p\,\text{prime}}L_{p}\left(f,s\right), \\ L_{p}\left(f,s\right) &= \left(1-\lambda_{f}\left(p\right)p^{-s}+p^{-2s}\right)^{-1} \end{aligned} \] and the \textit{completed L-function} is given by \[ \Lambda\left(f,s\right)=\gamma\left(f,s\right)L\left(s,f\right), \] where \[ \gamma\left(f,s\right) = \Gamma_{\mathbb{R}}\left(s+\frac{k-1}{2}\right)\Gamma_{\mathbb{R}}\left(s+\frac{k+1}{2}\right),\quad\Gamma_{\mathbb{R}}\left(s\right)=\pi^{-\frac{s}{2}}\Gamma\left(\frac{s}{2}\right) \] is the \textit{gamma factor} corresponding to \(L\left(s,f\right)\). Writing \(\Lambda\left(f,s\right)\) in terms of a Mellin transform of \(f\) one can easily show that \(\Lambda\left(f,s\right)\) can be analytically continued to a meromorphic function on the upper half-plane and satisfies a functional equation of the form \[ \Lambda\left(1-s,f\right)=\epsilon\Lambda\left(s,f\right), \] where \(\epsilon\in\left\{ \pm1\right\} \) is the so-called \textit{root number} of \(L\left(f,s\right)\). It is sometimes better to rewrite the local factor \(L_{p}\left(f,s\right)\) in terms of the Satake parameters \(\alpha_{p}\) and \(\beta_{p}\) defined as the two roots of the equation \(1-\lambda_{f}\left(p\right)X+X^{2}=0\), i.e., \[ L_{p}\left(f,s\right)^{-1}=\left(1-\alpha_{p}p^{-s}\right)\left(1-\beta_{p}p^{-s}\right)=\det\left(1-A_{p}p^{-s}\right), \] where \(A_{p}\) is the \(2\times2\) diagonal matrix \(\text{diag}(\alpha_{p},\beta_{p})\). If \(f_{i}\in S_{k_{i}}\left(\text{SL}_{2}(\mathbb{Z})\right)\) (\(i=1,2\)) we can define the \(L\)-function of the \textit{Rankin-Selberg convolution} \(f_{1}\otimes f_{2}\): \[ \begin{aligned} L(f_{1}\otimes f_{2},s) &= \prod_{p}L_{p}(f_{1}\otimes f_{2},s),\\ L_{p}(f_{1}\otimes f_{2},s) &= \det\left(1_{4}-A_{1,p}\otimes A_{2,p}p^{-s}\right)^{-1} \end{aligned} \] where \(1_{4}\) is the \(4\times4\) identity matrix and \(\otimes\) denotes the usual Kronecker product. The gamma factor of \(L\left(f_{1}\otimes f_{2},s\right)\) can then be written as \[ \gamma\left(f_{1}\otimes f_{2},s\right)=\Gamma_{\mathbb{R}}\left(s+k^{-}\right)\Gamma_{\mathbb{R}}\left(s+k^{+}\right)\Gamma_{\mathbb{R}}\left(s+k^{-}+1\right)\Gamma_{\mathbb{R}}\left(s+k^{+}-1\right), \] where \(k^{+}=\frac{k_{1}+k_{2}}{2}\) and \(k^{-}=\frac{\left|k_{1}-k_{2}\right|}{2}\). The completed L-function \[ \Lambda\left(f\otimes g,s\right)=\gamma\left(f\otimes g,s\right)L\left(f\otimes g,s\right) \] once again has a meromorphic continuation and satisfies a functional equation. Similarly, for three cusp forms \(f_{i}\in S_{k_{i}}\left(\text{SL}_{2}(\mathbb{Z})\right)\) of weight \(k_{i}\) (\(i=1,2,3\)) one can define the \textit{triple-product L-function}: \[ \begin{aligned} L\left(f_{1}\times f_{2}\times f_{3},s\right) &= \prod_{p}L_{p}\left(f_{1}\times f_{2}\times f_{3},s\right),\\ L_{p}\left(f_{1}\times f_{2}\times f_{3},s\right) &= \det\left(1_{8}-A_{1,p}\otimes A_{2,p}\otimes A_{3,p}\cdot p^{-s}\right)^{-1}. \end{aligned} \] According to the general philosophy it should be possible to write the gamma factor in the form \[ \prod_{i=1}^{8}\Gamma_{\mathbb{R}}\left(s+\mu_{i}\right) \] for some real parameters \(\mu_{i}\). The first explicit computation of this gamma factor was by \textit{P.~B. Garrett} [Ann. Math. (2) 125, 209--235 (1987; Zbl 0625.10020)] who treated the case of \(k_{1}=k_{2}=k_{3}=k\). Instead of considering \(L\)-functions of holomorphic modular forms on \(\text{SL}_{2}(\mathbb{Z})\) it is also possible to consider \(L\)-functions of holomorphic or non-holomorphic automorphic forms on congruence subgroups of \(\text{SL}_{2}(\mathbb{Z})\). In general, if \(\pi\) is an irreducible, cuspidal, automorphic representation of \(\text{GL}_{2}\left(\mathbb{A}\right)\), where \(\mathbb{A}\) is the adèle ring of an algebraic number field one can define \(L\left(\pi,s\right)\) in the same manner as \(L\left(f,s\right)\), using the Satake parameters of \(\pi_{p}\) (\(p<\infty\)). The corresponding gamma factor is then given in terms of the parameters at the infinite places of \(\pi\). In the paper under review, the author uses techniques similar to \textit{T. Ikeda} [Duke Math. J. 97, No. 2, 301--318 (1999; Zbl 0971.11029) and J. Reine Angew. Math. 499, 199--223 (1998; Zbl 0923.11078)] to compute the gamma factor of the triple product \(L\)-function \(L\left(s,\pi_{1}\times\pi_{2}\times\pi_{3}\right)\) in the case \(\pi_{1}\) and \(\pi_{2}\) are discrete series and \(\pi_{3}\) is a principal series representation of \(\text{GL}_{2}(\mathbb{R})\). The main theorem expresses \(L\left(s,\pi_{1}\times\pi_{2}\times\pi_{3}\right)\) in terms of an integral of a function \(f^{\left(s\right)}\) (essentially corresponding to an Eisenstein series) against a Whittaker function \(W\) belonging to the tensor product of the Whittaker spaces of \(\pi_{1},\) \(\pi_{2}\) and \(\pi_{3}.\) The proof is quite technical and involves (amongst other things) several non-trivial integral representations of, and relationships between, sums of quotients of gamma functions. In the last section there is a summary of what has been calculated and it is useful to review this section while reading the previous sections. (Note that the author states that \(\pi_{1}\) and \(\pi_{2}\) are ``essentially (limits of) discrete series representation''. This terminology is not explained further, neither in the paper under review, nor in the paper by Ikeda referenced above; it is not clear to the reviewer what the precise meaning of such an object is.)