Embezzlement as a "Self-Test" for Infinite Copies of Entangled States
Li Liu
TL;DR
The paper addresses exact entanglement embezzlement in infinite-dimensional quantum systems through a C$^*$-algebraic framework, clarifying how a catalyst state $f$ must internally contain infinite, locality-respecting copies of a target state $g$ to enable exact embezzlement. It proves an equivalence between standard and no-input embezzlement, and then shows that exact embezzlement necessarily yields local containment of infinitely many copies of $g$ within $f$, formalized via the construction of $g^ extinfty$ and a hierarchy of lifted observables with certification, commutativity, and independence. This self-test-like certification of infinite copies implies that universal exact embezzlement requires catalysts with non-separable, infinite-copy structures, particularly to accommodate countable or uncountable families of target states, and raises questions about extending these results to approximate embezzlement. Overall, the work provides conceptual tools linking embezzlement, operator-algebraic locality, and state certification in infinite quantum systems, with potential implications for nonlocal games and quantum cryptography.
Abstract
We investigate the operator-algebraic structure underlying entanglement embezzlement, a phenomenon where a fixed entangled state (the catalyst) can be used to generate arbitrary target entangled states without being consumed. We show that the ability to embezzle a target state $g$ imposes strong internal constraints on the catalyst state $f$: specifically, $f$ must contain infinitely many mutually commuting, locally structured copies of $g$. This property is formalized using C*-algebraic tools and is analogous to a form of self-testing, certifying the presence of infinite copies of $g$ within $f$. Using this infinite-copy certification. Our results clarify the structural requirements for embezzlement and provide new conceptual tools for analyzing state certification in infinite-dimensional quantum systems.
