The support of the free additive convolution of multi-cut measures

Abstract

We consider the free additive convolution $\mu_\alpha\boxplus\mu_\beta$ of two probability measures $\mu_\alpha$ and $\mu_\beta$, supported on respectively $n_\alpha$ and $n_\beta$ disjoint bounded intervals on the real line, and derive a lower bound and an upper bound that is strictly smaller than $2n_\alpha n_\beta$, on the number of connected components in its support. We also obtain the corresponding results for the free additive convolution semi-group $\{\mu^{\boxplus t}\,:\, t\ge 1\}$. Throughout the paper, we consider classes of probability measures with power law behaviors at the endpoints of their supports with exponents ranging from $-1$ to $1$. Our main theorem generalizes a result of Bao, Erd\H{o}s and Schnelli~[4] to the multi-cut setup.