Transport maps for \(\beta\)-matrix models and universality (Q2351057)

In [\textit{C. Villani}, Topics in optimal transportation. Graduate Studies in Mathematics 58. Providence, RI: American Mathematical Society (AMS) (2003; Zbl 1106.90001)], it is shown that there are maps \(T^N: \mathbb{R}^N\to \mathbb{R}^N\) that transport the \(\beta \)-matrix model \(\mathbb{P}^N_V\) to \(\mathbb{P}^N_{V+W}\). The probability measure used in the construction of \(T^N\) is defined by a potential \(V:\mathbb{R}\to \mathbb{R}\). In the present paper, the authors obtain a robust and very flexible method to compare the asymptotics of local statistics. They consider one-matrix models (in fact their generalization given by \(\beta\)-models) and construct approximate transport maps with a very precise dependence on the dimension. The smooth approximate map constructed by the authors is, at the first order, a tensor product, then has first order corrections of order \(N^{-1}\). Thus, the first order term of the map considered by the authors is the monotone transport map between the asymptotic equilibrium measures. The transport map is constructed as the flow of a vector field obtained by approximately solving a linearized Jacobian equation and then making a suitable ansatz on the structure of the vector field. This approach has the advantage of being pretty robust and leads to some generalizations of many other mean field models. The paper is structured as follows. First the authors describe the general strategy to construct the transport maps as the flow of vector fields obtained by approximately solving a linearization of the Monge-Ampère equation. Such an idea comes from the optimal transport theory. Next the authors make an ansatz on the structure of an approximate solution and show that the obtained solution is a smooth solution that enjoys very nice regularity estimates that are uniform as \(N\to \infty\). Then they reconstruct the approximate map from \(\mathbb{P}^N_V\) to \(\mathbb{P}^N_{V+W}\) via a flow argument. The obtained estimates are crucial in showing the universality property.