The Dirichlet Markov ensemble (Q847412)

Let \({\mathcal M}_n\) be the set of \(n \times n\) Markov matrices. Let this set be a convex compact polytope with \(n(n-1)\) degrees of freedom if \(n >1\) and with uniform distribution \({\mathcal U}({\mathcal M}_n)\) on \({\mathcal M}_n\) defined as the normalized trace of the Lebesgue measure of \({\mathbb R}^{n^2}\). It is shown that if the rows of \({\mathbf M}\) are independent and indentically distributed and follow the Dirichlet law of mean \((1/n, \dots, 1/n)\) then the probability distribution \({\mathcal U}({\mathcal M}_n)\) is invariant under permutations of rows and columns. Also, the probability distribution of the singular values of \(\sqrt{n}~{\mathbf M}\) tends to a Wigner quarter-circular distribution as \(n \rightarrow \infty\). It is conjectured that with probability one, the empirical distribution of the complex spectrum of \(\sqrt{n}~{\mathbf M}\) tends as \(n \rightarrow \infty\) to the uniform distribution on the unit disc of the complex plane and that the spectral gap of \(\mathbf M\) is of order \(1 - 1/ \sqrt{n}\) for large~\(n\).