Elliptic Selberg integrals (Q2764643)

The Selberg integral [\textit{A. Selberg}, Bemerkninger om et multipelt integral, Norsk Mat. Tiddskr. 26, 71-78 (1948)] NEWLINE\[NEWLINE\int_0^1 \cdots \int_0^1 \prod_{i=1}^n x_i^{\alpha-1}(1-x_i)^{\beta-1} \prod_{1\leq j<k\leq n}|x_j-x_k|^{2\gamma} dx_1\cdots dx_n NEWLINE\]NEWLINE NEWLINE\[NEWLINE=\prod_{j=1}^n {\Gamma(\alpha+(j-1)\gamma)\Gamma(\beta+(j-1)\gamma)\Gamma(1+j\gamma) \over \Gamma(\alpha+\beta+(n+j-2)\gamma)\Gamma(1+\gamma)}NEWLINE\]NEWLINE for \(\operatorname {Re}(\alpha),\operatorname {Re}(\beta)>0\) and \(\operatorname {Re}(\gamma)> -\min(1/n,\operatorname {Re}(\alpha)/(n-1), \operatorname {Re}(\beta)/(n-1))\) is a famous multivariable generalization of Euler's beta integral, associated with the root system BC\(_n\). The Selberg integral has many interesting applications, and perhaps best known is its connection to constant term identities such as [\textit{I. G. Macdonald}, SIAM J. Math. Anal. 13, 988-1007 (1982; Zbl 0498.17006); \textit{W. G. Morris}, Constant term identities for finite and affine root systems: Conjectures and theorems, Ph.D. thesis, Univ. Wisconsin (1982)] NEWLINE\[NEWLINE\begin{multlined} \text{CT}\biggl(\prod_{i=1}^n (1-x_i)^{\alpha-1}(1-x_i^{-1})^{\beta-1} \prod_{j\neq k}(1-x_j/x_k)^{\gamma}\biggr) \cr =\prod_{j=1}^n {\Gamma(\alpha+\beta-1+(j-1)\gamma)\Gamma(1+j\gamma) \over \Gamma(\alpha+(j-1)\gamma)\Gamma(\beta+(j-1)\gamma)\Gamma(1+\gamma)}\end{multlined} NEWLINE\]NEWLINE and NEWLINE\[NEWLINE \begin{aligned} \text{CT}\biggl(\Big[\prod_{i=1}^n (1-x_i)&(1-x_i^{-1}) (1-x_i^2)(1-x_i^{-2})\times\\ &\times\prod_{1\leq j<k\leq n} (1-x_j/x_k)(1-x_k/x_j) (1-x_j x_k)(1-x_j^{-1}x_k^{-1}) \Big]^a\biggr) \\ &=\prod_{j=1}^n {\Gamma(1+(2j+2)a) \over \Gamma(1+(2j+1)a)\Gamma(1+a)}.\end{aligned} NEWLINE\]NEWLINE NEWLINENEWLINENEWLINEIn the article under review, two elliptic Selberg integrals are proposed involving Ruijsenaars elliptic gamma function [\textit{S. N. M. Ruijsenaars}, J. Math. Phys. 38, No. 2, 1069-1146 (1997; Zbl 0877.39002)]. With \((a;p,q)_{\infty}=\prod_{j,k=0}^{\infty}(1-ap^jq^k)\) (\(|p|,|q|<1\)), the elliptic gamma function is defined as \(\Gamma(z;p,q)=(pq/z;p,q)_{\infty}/ (z;p,q)_{\infty}\). The two elliptic Selberg integrals can then be stated as NEWLINE\[NEWLINE \int_{T^n} \Delta_n^{\text{I}}(z;p,q){\text{d}z_1\over z_1}\dots {\text{d}z_n\over z_n}={2^n n! \over (p;p)_{\infty}^n(q;q)_{\infty}^n} {\prod_{0\leq r<s\leq 2n+2}\Gamma(t_rt_s;p,q)\over \prod_{r=0}^{2n+2}\Gamma(A/t_r;p,q)} NEWLINE\]NEWLINE for \(|p|,|q|,|t_0|,\dots,|t_{2n+2}|<1\) such that \(|pq|<|t_0\dots t_{2n+2}|\), and and NEWLINE\[NEWLINE\int_{T^n} \Delta_n^{\text{II}}(z;p,q){dz_1\over z_1}\dots {dz_n\over z_n}={2^n n! \over (p;p)_{\infty}^n(q;q)_{\infty}^n} \prod_{j=1}^n{\Gamma(t^j;p,q)\over \Gamma(t;p,q)} {\prod_{0\leq r<s\leq 4}\Gamma(t^{j-1}t_rt_s;p,q)\over \prod_{r=0}^4\Gamma(t^{1-j}B/t_r;p,q)},NEWLINE\]NEWLINE for \(|p|,|q|,|t|,|t_0|,\dots,|t_4|<1\) such that \(|pq|<|t^{2n-2}t_0\dots t_4|\) Here \((q;q)_{\infty}=\prod_{j=1}^{\infty}(1-q^j)\), \(A=t_0\cdots t_{2n+2}\), \(B=t^{2n-2}t_0\cdots t_4\), NEWLINE\[NEWLINE \Delta_n^{\text{I}}(z;p,q)={1\over (2\pi\text{i})^n} \prod_{1\leq j<k\leq n}{1\over \Gamma(z_j z_k,z_j/z_k,z_k/z_j,1/z_jz_k;p,q)} \prod_{j=1}^n{\prod_{r=0}^{2n+2}\Gamma(t_rz_j,t_r/z_j;p,q)\over \Gamma(z_j^2,1/z_j^2,A z_j,A/z_j;p,q)} NEWLINE\]NEWLINE and NEWLINE\[NEWLINE \Delta_n^{\text{II}}(z;p,q)={1\over (2\pi\text{i})^n} \prod_{1\leq j<k\leq n}{\Gamma(tz_j z_k,tz_j/z_k,tz_k/z_j,t/z_jz_k;p,q) \over \Gamma(z_j z_k,z_j/z_k,z_k/z_j,1/z_jz_k;p,q)} \prod_{j=1}^n{\prod_{r=0}^4\Gamma(t_rz_j,t_r/z_j;p,q)\over \Gamma(z_j^2,1/z_j^2,B z_j,B/z_j;p,q)}. NEWLINE\]NEWLINE For \(n=1\) the ``type I'' and ``type II'' elliptic Selberg integrals coincide, and correspond to an elliptic Selberg integral of the second author [\textit{V. P. Spiridonov}, Russ. Math. Surv. 56, No. 1, 185-186 (2001; Zbl 0997.33009)]. For \(p=0\) the elliptic Selberg integrals reduce to the Nassrallah-Rahman integral [\textit{B. Nassrallah} and \textit{M. Rahman}, SIAM J. Math. Anal. 16, 186-197 (1985; Zbl 0564.33009)] and Gustafson's \(\text{Sp}(n)\) multivariable generalization of the Nassrallah-Rahman integral [\textit{R. A. Gustafson}, SIAM J. Math. Anal. 23, No. 2, 525-551 (1992; Zbl 0764.33008); ibid. 25, No. 2, 441-449 (1994; Zbl 0819.33009)]. NEWLINENEWLINENEWLINEFor \(n>1\) the proof of both elliptic Selberg integrals depends on a yet unproved ``vanishing hypothesis''. This hypothesis claims that the type I Selberg integral vanishes if \(t_{2n+2}\) tends to \(A\) when the integration contour is deformed from the unit circle to a Jordan curve on which some technical conditions are imposed.