Gaussian analytic functions in the polydisk

From MaRDI portal



Abstract: We study hyperbolic Gaussian analytic functions in the unit polydisk of mathbbCn. Following the scheme previously used in the unit ball we first study the asymptotics of fluctuations of linear statistics as the directional intensities Lj, j=1,dots,n tend to infty. Then we estimate the probability of large deviations of such linear statistics and use the estimate to prove a hole theorem. Our proofs are inspired by the methods of M. Sodin and B. Tsirelson for the one-dimensional case, and B. Shiffman and S. Zelditch for the study of the analogous problem for compact K"ahler manifolds.


The authors extend some well-known probabilistic properties of the hyperbolic Gaussian analytic functions in the unit disk \(\mathbb{D}\) to the unit polydisk \(\mathbb{D}^n\). For \(L=(L_1,\cdots,L_n)\), \(L_1,\cdots,L_n>1\), let \(B_L(\mathbb{D}^n)\) be the weighted Bergman space of all holomorphic functions in \(L^2_L(\mathbb{D}^n)=L^2(\mathbb{D}^n;[1-|z|^2]^L d\nu_n(z))\), where \([1-|z|^2]^L=\prod_{j=1}^n(1-|z_j|^2)^{L_j}\) and \(d\nu_n\) is the normalized invariant measure on \(\mathbb{D}^n\). The hyperbolic Gaussian analytic function of intensity \(L\) is defined by \[ f_L(z)=\sum_{\alpha} a_\alpha\frac{z^\alpha}{\|z^\alpha\|_{L^2_L}} \qquad (z\in\mathbb{D}^n), \] where \(a_\alpha\) are i.i.d. complex Gaussians of mean 0 and variance 1. The authors study the asymptotics of fluctuations of linear statistics as the intensities \(L_j\) tend to infinity. The main results establish probabilistic properties of the random variable \(I_L(\varphi)=\int_{Z(f_L)} \varphi\), where \(\varphi\) is a smooth \((n-1,n-1)\)-form and \(Z(f_L)\) is the zero variety of \(f_L\). As a consequence of these results the authors prove the following hole theorem: Let \(E(w,r)\) be an open pseudo-hyperbolic polydisk centered at \(w\) and of multiradius \(r=(r_1,\ldots,r_n),\,r_j\in (0,1)\). There exist positive constants \(C_1(n,r)\) and \(C_2(n,r)\) and a multi-index \(L^0\) such that, for all \(L\geq L^0\), \[ e^{-A_1(n,r,L)}\leq\,\text{Prob}[Z(f_L)\cap E(w,r)=\emptyset]=\text{Prob}[Z(f_L) \cap E(0,r)=\emptyset]\leq e^{-A_2(n,r,L)}, \] where \(A_j(n,r,L)=C_j(n,r)(\sum_k L_k)(\prod_k L_k)\), \(j=1,2\).











This page was built for publication: Gaussian analytic functions in the polydisk

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2363034)