Random permutation matrices under the generalized Ewens measure (Q1950260): Difference between revisions

Random unitary \(N\times N\)-matrices, the distributions of eigenvalues, traces and characteristic polynomials and their convergence behaviour (for \(N\to\infty\)) have been extensively investigated in the past, Also for matrices belonging to a closed infinite subgroup \(G\) of \(\mathcal{U}_N\). In the paper under review, the authors concentrate on the case of finite subgroups, in particular permutation groups \(\mathcal{S}_N\). This restriction forces them to use different methods and tools, cf., e.g., [\textit{K. Wieand}, Ann. Probab. 28, No. 4, 1563--1587 (2000; Zbl 1044.15017); J. Theor. Probab. 16, No. 3, 599--623 (2003; Zbl 1043.60007); \textit{B. M. Hambly} et al., Stochastic Processes Appl. 90, No. 2, 335--346 (2000; Zbl 1047.60013); \textit{G. Ben Arous} and \textit{K. Dang}, ``On fluctuations of eigenvalues of random permutation matrices'', \url{arXiv:1106.2108}]. The authors consider the above mentioned problems for a variety of distributions (`generalized Ewens measures') including the uniformly distributed case. All functions (and measures) considered here are conjugation invariant, hence class functions, and may therefore be considered as functions/measures on \(\mathcal{S}_N\) which are constant on classes, or as objects on the set of classes \(\{C_\lambda\}\), \(\lambda =(\lambda_i)\) denoting the `cycle type' of a permutation (defined by the representation as a product of cycles of length \(\lambda_i\)). Sections 1 and 2 are concerned with basic definitions and properties, in particular with the definition of the afore-mentioned generalized Ewens measures \(P_\Theta\) on \(\mathcal{S}_N\), \(\Theta =(\theta_k)\) denoting a sequence of positive numbers (cf. [\textit{A. Nikeghbali} and \textit{D. Zeindler}, ``The generalized weighted probability measure on the symmetric group and the asymptotic behaviour of the cycles'', \url{arXiv:1105.2315}].) A main tool in the investigations are generating functions \(t\mapsto G_\Theta(t)=\exp(\sum \theta_k t^k/k)\) represented as series \(\sum h_n t^n\), and the asymptotic behaviour of the coefficients if the functions satisfy certain smoothness conditions. Section 3 is concerned with moments of characteristic polynomials, more general, multiplicative class functions, and Section 4 with traces of powers of permutations, more general, of functions \(F(\sigma): F:S^1\to \mathbb{C}\) of permutations (additive class functions). In both sections, the emphasis is laid on the asymptotic behaviour (for \(N\to\infty\)), e.g., in case of traces, if \(F\) is a power function, more general, a Laurent polynomial, for suitable generating functions, \(\mathrm{Tr}(F)(\sigma)-Nb_0\) converges in distribution to a random variable representable as series of Poisson variables (\(b_0\) denoting the constant term of \(F\), and \(\mathrm{Tr}(F)(\sigma):=\sum F(\omega_k)\), \(\{\omega_k\}\) denoting the eigenvalues of \(\sigma\). Section 5 generalizes the above models to wreath products of the torus \(S^1\) and \(\mathcal{S}_N\) (consisting of matrices of the form \(\mathrm{diag}(z_1, \dots ,z_N)\sigma\) with \(z_i\in S^1\) and \(\sigma\in\mathcal{S}_N\)). While in Sections 4 and 5, the conditions imply boundedness of variances, in Section 6, for unbounded variances, the authors obtain a central limit theorem for \(\mathrm{Tr}(F)\), under the restriction that \(P_\Theta\) is a `classical' Ewens measure, i.e., with constant \(\theta_k\equiv \theta\), proved again within the framework of wreath products, generalizing a CLT obtained by G. Ben Arous and K. Dang [loc. cit.].