Bi-free extreme values

Abstract

In this paper, we continue Voiculescu's recent work on the analogous extreme value theory in the context of bi-free probability theory. We derive various equivalent conditions for a bivariate distribution function to be bi-freely max-infinitely divisible. A bi-freely max-infinitely divisible distribution function can be expressed in terms of its marginals and a special form of copulas. Such a distribution function is shown to be also max-infinitely divisible in the classical sense. In addition, we characterize the set of bi-free extreme value distribution functions. A distribution function of this type is also bi-freely max-stable and represented by its marginals and one copula composing of a Pickands dependence function, as in the classical extreme value theory. As a consequence, the determination of its bi-free domain of attraction is the same as the criteria in the classical theory. To illustrate these connections, some concrete examples are provided.