Title:Optimal error intervals for properties of the quantum state

Abstract:Quantum state estimation aims at determining the quantum state from the observed data as precisely as possible, which usually requires a considerable effort. One is, however, often not interested in all the details of the fully estimated state. Rather, one needs to know one or a few properties of the state such as, for example, its fidelity with a target state or the degree of correlation for a specific bipartite structure. Since the best guess for a particular property is, typically, not equal to its value in the best-guess state, it is preferable to directly estimate the properties of interest from the recorded data-and, in fact, this is possible. As a natural and meaningful way of stating the accuracy of the estimated property values, we propose the use of optimal error intervals, which are immediate analogs of the optimal error regions for quantum state estimation [see New J. Phys. 15, 123026 (2013)]. The optimal error intervals are easily characterized by the likelihood for the data conditioned on the values of the properties of interest. We report an iterative algorithm for the computation of this likelihood. The step that is most costly (in CPU time) is the sampling, for the purpose of Monte Carlo integration, of the quantum state space in accordance with a sequence of prior densities. We illustrate the concepts and methods by identifying the optimal error intervals for the fidelity and purity of single-qubit states, as well as the CHSH quantity for two-qubit states, both for pre-chosen observables and for a state-dependent optimized choice of observables.