Phase transitions and equilibrium measures in random matrix models (Q2515012)

This paper deals with phase transitions in the Hermitian random matrix models with a polynomial potential or equivalently, with a family of equilibrium measures on the real line with a polynomial external field, parameterized by the total mass of the measures. The authors study the evolution of this family of measures by means of a system of ODE's and derive a number of known and new results, in particular on the local behavior of the system at phase transitions. Section 1 is an extended introduction on the equilibrium problem for the logarithmic potential and its solution, known as the equilibrium measure. This important topic arose in several fields of mathematics such as analysis, approximation theory, and statistical mechanics. Here, the focus is on the dependence of the equilibrium measure on the real line, and in particular its singularities, from the parameter \(t\) which may be viewed as the temperature, the time, or the total mass of the measure. One of the key problems is finding the support of such measures. In presence of a real-analytic external field, the support is comprised of a finite number of disjoint intervals, or ``cuts''. The authors study a system of differential equations ruling the evolution of the endpoints of such cuts and analyze several magnitudes of the system close to these singular points. This section also reviews connections with random matrix theory and the Dyson gas, weighted equilibria in analysis and constrained Green's equilibirum problems in soliton theory. Section 2 deals with families of the equilibrium and critical measures on the real line. Here, two kinds of critical measures are used to derive an equation for the Cauchy transform of the equilibrium measure (Theorem 2). Then, differentiation formulas of quantities such as the equilibrium density, the equilibrium constant and the infinite volume free energy with respect to the time parameter are discussed and expressed in terms of the Robin measure and constant of the support (Theorem 3). Theorem 4 belongs to the third author and gives conditions on the external field under which the support of the equilibrium measure is comprised of a single interval (one-cut case). In Section 3, the authors study the equilibrium measures in a polynomial field, reformulating Theorem 2 and its corollaries. They introduce the \(R\)-representation, where \(R\) is a certain polynomial used to express the density of the equilibrium measure with respect to the Lebesgue measure on the line. This function can itself be expressed in terms of two other polynomials \(A\) and \(B\), leading to the so-called \((A,B)\)-representation. The advantage is that the evolution of the support and the phase transitions can be recast in terms of the zeros of \(A\) and \(B\). In this context, singularities may be classified in three types following the nature of these zeros. Then, differentiation formulas may be rewritten (1) in a unified way, using time and the polynomial coefficients as parameters, and (2) using the zeros of \(A\) and \(B\). The local behavior at phase transitions, still in the case of the polynomial external field, is investigated in Section 4. There, a previously established result is extended to a general multi-cut case, showing that for the three types of singularities, the phase transition for the energy is at least of third order. Finally, Section 5 deals with the special case of the quartic external field, which is the first non-trivial case beyond that of a quadratic polynomial.