Negativity and steering: a stronger Peres conjecture

Abstract

The violation of a Bell inequality certifies the presence of entanglement even if neither party trusts their measurement devices. Recently Moroder et. al. showed how to make this statement quantitative, using semidefinite programming to calculate how much entanglement is certified by a given violation. Here I adapt their techniques to the case where Bob's measurement devices are in fact trusted, the setting for "EPR-steering" inequalities. Interestingly, all of the steering inequalities studied turn out to require negativity for their violation. This supports a significant strengthening of Peres' conjecture that negativity is required to violate a bipartite Bell inequality.