Optimal quantum violations of n-locality inequalities with conditional dependence on inputs

Abstract: Bell experiment in the network gives rise to a form of quantum nonlocality which is conceptually different from traditional multipartite Bell nonlocality. Conventional multipartite Bell experiment features a single source that distributes physical systems to multiple parties. In contrast, the network Bell experiment features multiple independent sources. This work considers a nontrivial quantum network, the star-network configuration in an arbitrary input scenario involving n independent sources and (n+1) parties, including n edge parties and one central party. Each of the n edge parties shares a physical system with the central party. We consider that the central party received an arbitrary m number of inputs, and each edge party receives 2^{m-1} number of inputs. The joint probabilities of the system are bounded by some linear constraints. We show that this behaviour of the joint probabilities in turn imposes conditional dependence on the inputs of the edge parties such that the observables of each edge party are bounded by few linear constraints. We derive a family of generalized n-locality inequalities and demonstrate its optimal quantum violation. We introduce an elegant sum-of-squares approach that enables the optimization in quantum theory without specifying the dimension of the quantum system. The optimal quantum value self-tests the observables of each edge party along with the conditional dependence. The observables of the central party along with the quantum state are also self-tested from the optimization procedure itself. Further, we characterize the network nonlocality and examine its correspondence with suitably derived standard Bell nonlocality.