Tensorial free convolution, semicircular, free Poisson and R-transform in high order
Remi Bonnin
TL;DR
This work extends Voiculescu's freeness to tensors by defining a tensorial free convolution for compactly supported measures. It introduces higher-order semicircular and free Poisson laws with explicit moments and free cumulants, proves convergence results for Wigner and Wishart tensors, and establishes a tensorial additive convolution with an $R$-transform (and a $Q$-transform) that is additive under convolution. Basic examples illustrate how tensorial free convolution operates on the high-order laws and highlight connections to Fuss-Catalan and Fuss-Narayana combinatorics. The framework provides a foundation for high-dimensional tensor statistics and links combinatorial map invariants to free probabilistic convolution in the tensor setting.
Abstract
This work builds on our previous developments regarding a notion of freeness for tensors. We aim to establish a tensorial free convolution for compactly supported measures. First, we define higher-order analogues of the semicircular (or Wigner) law and the free Poisson (or Marcenko-Pastur) law, giving their moments and free cumulants. We prove the convergence of a Wishart-type tensor to the free Poisson law and recall the convergence of a Wigner tensor to the semicircular law. We also present a free Central Limit Theorem in this context. Next, we introduce a tensorial free convolution, define the $R$-transform, and provide the first examples of free convolution of measures.
