A note on the eigenvalue density of random matrices (Q1284182)

The paper focuses on the eigenvalue distribution of ensembles of \(N\times N\) random matrices \(M_N\) whose probability distribution has the density \[ Q_N^{-1} \exp\bigl\{- \kappa N \text{Tr} V(M) \bigr\} dM.\tag{1} \] It is known since the papers by F. Dyson that the eigenvalues \(\lambda_j\), \(j=1, \dots,N\) of such matrices are distributed as the charged particles in the potential determined by \(V\). The variational principle formulated by \textit{E. Wigner} for Hermitian random matrices (1) [Can. Math. Proc. 174-198 (1957)] leads to the following equation for the limiting eigenvalue density \(\rho(\lambda)= \lim_{N\to \infty}\sum^N_{j=1} \delta( \lambda- \lambda_j)N^{-1}\) \[ \kappa\int_{\text{supp}}{\rho(\mu)d\mu\over\lambda-\mu}= V'(\lambda). \tag{2} \] In a faper by \textit{A. Boutet de Monvel}, \textit{L. Pastur} and \textit{M. Shcherbina} [J. Stat. Phys. 79, 585-611 (1995)] equation (2) was derived rigorously. In present paper the authors prove an abstract statistical mechanics theorem on particle systems with unstable interactions. As a consequence of this theorem, existence of the limiting eigenvalue distributions is proved for a series of random matrix ensembles (1) generalising the ensemble considered in the paper cited above.