Optimal Local Simulations of a Quantum Singlet

Bell's seminal work showed that no local hidden variable (LHV) model can fully reproduce the quantum correlations of a two-qubit singlet state. His argument and later developments by Clauser et al. effectively rely on gaps between the anticorrelations achievable by classical models and quantum theory for projective measurements along randomly chosen axes separated by a fixed angle. However, the size of these gaps has to date remained unknown. Here we numerically determine the LHV models maximizing anticorrelations for random axes separated by any fixed angle, by mapping the problem onto ground state configurations of fixed-range spin models. We identify angles where this gap is largest and thus best suited for Bell tests. These findings enrich the understanding of Bell non-locality as a physical resource in quantum information theory and quantum cryptography.