Pullback of Klingen Eisenstein series and certain critical L-values identities (Q2038991)

\textit{P. B. Garrett} [Prog. Math. 46, 114--137 (1984; Zbl 0544.10023)] discovered in the 1980's a remarkable integral representation for the degree \(2n+1\) \(L\)-function of an automorphic representation of \(\mathrm{GSp}_{2n}\) involving Siegel's Eisenstein series and a Siegel cusp form. This approach was generalized by \textit{G. Shimura} [Euler products and Eisenstein series. Providence, RI: American Mathematical Society (1997; Zbl 0906.11020); Arithmeticity in the theory of automorphic forms. Providence, RI: American Mathematical Society (AMS) (2000; Zbl 0967.11001)]. Following this line of ideas, the author proves the following pullback formula in Theorem 1.1. Let \(f\in S_k(\Gamma_0(N))\) denote an elliptic new form of level \(N\) of weight \(k>4\) and put for an element \(Z\in{\mathbb H}_2\) in the Siegel upper half space \[ E_{2,1}^k(Z,f,N):= \sum_{\gamma\in Q({\mathbb Q})\cap\Gamma_0^{\mathrm{GSp(4)}}(N)\backslash\Gamma_0^{\mathrm{GSp(4)}}(N)} f(\gamma\langle Z\rangle^*)\det(j(\gamma,Z))^{-k}. \] In the case \(N=1\) this is the classical Klingen Eisenstein series for the Klingen parabolic \[ Q= \left\{ \begin{pmatrix} \ast & & \ast & \ast\\ \ast & \ast & \ast & \ast\\ \ast & & \ast & \ast\\ \ast & & & \ast \end{pmatrix} \right\}. \] For \(s=k-2\) consider the Eisenstein series \[ E_1(s,k,N,\tau):= \sum_{A\in \Gamma_\infty\backslash\Gamma(N)} \frac{1}{|c\tau+d|^{s+2-k}(c\tau+d)^k}. \] Then the following formula is proved: \begin{align*} E_{2,1}^k(\mathrm{diag}(\tau_1,\tau_2),f,N) &=E_1(s,k,N,\tau_1)f(\tau_2)+E_1(s,k,N,\tau_2)f(\tau_1)\\ &\quad+2\sum_{\begin{subarray}cc,d\geq 0\\ (c,d)=1\end{subarray}} \sum_{\gamma_1,\gamma_2\in\Gamma_\infty\backslash\Gamma(N)} (j(\gamma_1,\tau_1)j(\gamma_2,\tau_2))^{-(s+2)} f(d^2\gamma_1\langle\tau_1\rangle+c^2\gamma_2\langle\tau_2\rangle). \end{align*} In Theorem 1.2 an almost identical formula is proved for a Klingen Eisenstein series \[ \tilde{E}_{2,1}^k(Z,f,N):= \sum_{\gamma\in L_N^{-1}Q({\mathbb Q})L_N\cap K(N^2)\backslash K(N^2)} f((L_N\gamma\langle Z\rangle)^*)\det(j(\gamma,Z))^{-k} \] for the paramodular subgroup \(K(N)\) and suitable \(L_N\in\mathrm{GSp}_4({\mathbb Z})\). Combining the above pullback formula of Theorem 1.1 with results of \textit{S.-i. Mizumoto} [Invent. Math. 65, 115--135 (1981; Zbl 0452.10032); Kodai Math. J. 7, 86--110 (1984; Zbl 0536.10019)] relating the Fourier coefficients of Klingen Eisenstein series to \(L\)-values the following formulae for an elliptic cusp form \(f\) of full level are obtained: \begin{align*} \frac{4}{\zeta(1-k)}+2\sum_{\begin{subarray}cc,d\geq 0\\ (c,d)=1\end{subarray}}a_{c,d}(1,1)&= 2+\frac{(-1)^{k/2}(k-1)!(2\pi)^{k-1}}{(2k-2)!L(2k-2,\mathrm{Sym}^2f)}\\ &\quad\times\left[2^{2k-3}L(k-1,\chi_{-4})L(k-1,f,\theta_1)\right.\\ &\quad+\left.2^{2k-3}L(k-1,\chi_{-3})L(k-1,f,\theta_2)\right], \end{align*} where \[ \sum_{\gamma_1,\gamma_2\in\Gamma_\infty\backslash\Gamma(N)} (j(\gamma_1,\tau_1)j(\gamma_2,\tau_2))^{-(s+2)} f(d^2\gamma_1\langle\tau_1\rangle+c^2\gamma_2\langle\tau_2\rangle)= \sum_{n_1,n_2}a_{c,d}(n_1,n_2)e^{2\pi i\tau_1n_1}e^{2\pi i\tau_2n_2}, \] \[ \theta_1(z)=\sum_{m,n\in{\mathbb Z}}q^{m^2+n^2}, \] \[ \theta_2(z)=\sum_{m,n\in{\mathbb Z}}q^{m^2+mn+n^2}. \] And for \((n_1,n_2)=1\): \begin{align*} &\sum_{\begin{subarray}cb\in{\mathbb Z}\\ 4n_1n_2-b^2\geq 0\end{subarray}} (-1)^{k/2}\Delta_{{\mathbb Q}\left(\sqrt{b^2-4n_1n_2}\right)}^{k-3/2} \frac{L(k-1,\chi_{b^2-4n_1n_2)})}{L(2k-2,\mathrm{Sym}^2 f)}\times\\ &\sum_{0<v\mid\sqrt{\frac{b^2-n_1n_2}{\Delta_{b,n_1,n_2}}}} L(k-1,f\otimes'\phi_{b,n_1,n_2}^{(v)})\\ &= \frac{(2k-2)!}{(k-1)!(2\pi)^{k-1}}\times\left(A_f(n_1,n_2)\right.\\ &\quad+\left.\frac{2}{\zeta(1-k)}\left[a(n_1)\sigma_{k-1}(n_2)+a(n_2)\sigma_{k-1}(n_1)\right]\right), \end{align*} where \( \Delta_{b,n_1,n_2}= \Delta_{{\mathbb Q}\left(\sqrt{b^2-4n_1n_2}\right)} \) is the discriminant of a quadratic extension, \[ \phi_{b,n_1,n_2}(z)=\sum_{m,n}q^{n_1m^2+bmn+n_2n^2}, \] and \(L(k-1,f\otimes'\phi_{b,n_1,n_2}^{(v)})\) is a \(v\)-twisted Rankin-Selberg \(L\)-function.