A large deviation principle for weighted Riesz interactions (Q1702114)

Let \(d>2\) be a positive integer. For any \(y=(y_1, y_2,\ldots , y_d) \in \mathbb{R}^d\) put \(|y|^2=\sum_{j=1}^d y_j^2\) and \(W(y)=\frac{1}{|y|^{\alpha }},\) where \(0<\alpha < d.\) Assume that \(K\) is a compact set in \(\mathbb{R}^d.\) Denote by \(\mathcal{M}(K)\) the space of probability measures on \(K\) endowed with the topology of weak convergence. For \(K\) of positive Riesz \(\alpha\)-capacity and a continuous weight function \(Q:K \to \mathbb{R},\) the authors consider the ensemble of probability measures \(\text{Prob}_n\) on \(K^n\): \[ \text{Prob}_n:= \frac{1}{Z_n} \exp \Big[ -\sum_{1\leq i \neq j \leq n} W(x_i-x_j) \Big]\cdot \exp \Big[ -2n\sum_{ j =1}^n Q(x_j) \Big] d\nu (x_1) \cdots d\nu (x_n). \] They assume that \(\nu\) satisfies a strong Bernstein-Markov type property for Riesz potentials. Let \(j_n: K^n \to \mathcal{M}(K)\) be the empirical measure map \(j_n(x_1, \ldots, x_n)=\frac{1}{n}\sum _{j=1}^n \delta_{x_j}\) and \((j_n)_{*}\) denote the push-forward. The main result of the paper states that the sequence \(\{\sigma_n=(j_n)_{*}(\text{Prob}_n) \}\) of probability measures on \(\mathcal{M}(K)\) satisfies a large deviation principle with speed \(n^2\) and good rate function \(\mathcal{I}:=\mathcal{I}_{K,Q}\) such that \( \;\forall \mu \in \mathcal{M}(K): \;\mathcal{I}(\mu)=I^Q(\mu)-I^Q(\mu_{K,Q}),\) where \(I^Q\) denotes the weighted Riesz energy functional and \(\mu_{K,Q} \in \mathcal{M}(K)\) is the unique weighted equilibrium measure. The authors also provide sufficient conditions on \(K\) (which are satisfied if \(K\) is a smooth submanifold) so that a measure on \(K\) that satisfies a mass-density condition will also satisfy this strong Bernstein-Markov property.