arXiv:2610.37785
Entanglement measures and LOCC conversion criteria from Schur-Weyl duality
Anton van Cleemput, Otfried Gühne, Albert Rico
Abstract
The characterization of possible transformations between quantum states and the quantification of entanglement are central for understanding quantum resources. Explicit results are, however, scarce since they require complicated optimizations over transformation protocols or state decompositions. We introduce a method to optimize functions of the entanglement spectrum of bipartite states over mixed-state decompositions with arbitrary accuracy. Our results are based on a convergent semidefinite programming hierarchy of optimizations over entanglement spectra using a natural polynomial basis and Schur-Weyl duality. Our approach can also be used to determine non-polynomial measures, such as the entanglement of formation, and to characterize the dimensionality of entanglement. Moreover, extracting the decomposition allows us to formulate criteria to determine state transformations under local operations and classical communication (LOCC). In particular, we provide a hierarchy of certificates of LOCC non-convertibility from pure to mixed states in any dimension and a heuristic approach for LOCC convertibility.
AI Summary
The paper addresses computing convex-roof entanglement quantities for bipartite mixed states and deciding deterministic LOCC conversion from a pure state to a mixed target. It develops a common optimization framework for continuous symmetric functions of the pure-state entanglement spectrum, including entanglement of formation and Vidal monotones. Schur–Weyl duality expresses symmetric spectral polynomials as expectation values of isotypic projectors on multiple copies. Convex-roof optimization then becomes linear optimization over mixtures of identical pure-state tensor powers. Relaxing these mixtures to Bose-symmetric, positive-partial-transpose extensions gives nondecreasing SDP lower bounds converging to the convex roof. The convergence proof uses a de Finetti approximation with trace-norm error at most $4Dm/k$ for an $m$-copy marginal of a $k$-copy extension, where $D=d_Ad_B$. Symmetry and support restrictions reduce the main variable’s dimension to $\binom{k+r-1}{k}$ for a rank-$r$ target. For non-polynomial functions, an embedded linear program selects state-dependent polynomial lower approximations on a spectral grid; inter-grid violations require postprocessing corrections. Tensor decomposition followed by an exact ensemble-repair step supplies upper bounds, with an asymptotic convergence argument. Combining all Vidal inequalities within one optimization yields a hierarchy detecting every impossible pure-to-mixed LOCC conversion at finite level in the continuous-grid formulation. Extracted ensembles can also certify possible conversions. Numerical examples include bound-entangled states and two-qutrit conversions. Computational size limits accessible hierarchy levels; practical tensor extraction is heuristic, and no complete extraction guarantee is given for LOCC convertibility. The framework is restricted to bipartite spectral convex roofs.
The numerical results, plots and proof calculations were assisted with AI (Claude Opus 4.8-5.5), and validated by the authors.
Central claim may not hold: For a pure two-qubit input whose reduced purity is absent from a finite spectral grid, the proposed non-polynomial optimization is unbounded at degree four and above, making its merged dual SDP infeasible. The main-text non-polynomial and LOCC certificate formulas incorrectly compute correction errors only at grid points, whereas valid corrections require global spectral maxima, as correctly specified in the appendices.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.