De Finetti Structures in Networks of Quantum Systems

Abstract

Consider $N$ identical quantum systems at the vertices of a graph, interacting via pairwise edge couplings. We designate $K$ vertices as a system of interest $\mathrm S$ and the remaining $N-K$ as a reservoir $\mathrm R$. Under generic connectivity assumptions between $\mathrm S$ and $\mathrm R$, we show that for $N\to\infty$, the reduced state of $\mathrm S$ converges to a convex combination of symmetric product states of $K$ single-vertex density matrices -- called a de Finetti representation. Unlike in standard derivations for symmetric (exchangeable) systems, we establish de Finetti representations for quantum networks, which are inherently non-exchangeable. We construct the representations explicitly and quantify the convergence rate. We show that Erdős--Rényi random graphs exhibit the de Finetti structure: as the reservoir grows, the reduced state of any fixed finite subgraph converges almost surely to an explicit deterministic product state. As applications, we establish a general area law for the entanglement entropy of $\mathrm S$, analyze mean-field and fluctuation scalings of the $\mathrm S$-$\mathrm R$ interaction, and illustrate our results on qubit networks. We also derive quantitative convergence estimates for information-theoretic quantities, including von Neumann entropy, mutual information, and entanglement within $\mathrm S$ and between $\mathrm S$ and $\mathrm R$.