Interlacing adjacent levels of \(\beta\)-Jacobi corners processes (Q1626623)

The authors study global fluctuations for the difference between two adjacent levels in the multilevel and general \(\beta\)-extension of the classical Jacobi ensemble of random matrices. The asymptotic covariance for linear statistics \({\mathcal \Sigma}_f = \sum\limits_{i=1}^{N} f(x_i)\) is understood through the corners processes. The jumps \({\mathcal \Sigma}_f^{\prime} = {\mathcal \Sigma}_f(k) - {\mathcal \Sigma}_f(k - 1)\) are studied in two separate asymptotic regimes, namely, for individual \(k = \lfloor y L\rfloor\) as \(L\rightarrow \infty\), and in the integrated form by averaging \({\mathcal \Sigma}_f^{\prime}(\lfloor y L\rfloor)\) with a smooth weight function on \(y\). For the first regime \({\mathcal \Sigma}_f^{\prime}(\lfloor y L\rfloor)\) converges as \(L\rightarrow \infty\) to a constant, while \({\mathcal \Sigma}_f^{\prime}(\lfloor y L\rfloor) - {\mathbb E} {\mathcal \Sigma}_f^{\prime}(\lfloor y L\rfloor)\) decays as \(L^{-1/2}\) and becomes asymptotically Gaussian upon rescaling. The main results of the paper strengthen the convergence of \({\mathcal \Sigma}_f(\lfloor y L\rfloor) - {\mathbb E} {\mathcal \Sigma}_f(\lfloor y L\rfloor)\) to the pullback of the Gaussian free field up to the convergence of the derivatives in the \(y\)-direction.