We estimate the probability of random
N-qudit pure states violating full-correlation Bell inequalities with two dichotomic observables per site. These inequalities can show violations that grow exponentially with
N, but we prove this is not the typical case. For many-qubit states the probability to violate any of these inequalities by an amount that grows linearly with
N is vanishingly small. If each system's Hilbert space dimension is larger than two, on the other hand, the probability of seeing any violation is already small. For the qubits case we discuss, furthermore, the consequences of this result for the probability of seeing arbitrary violations (i.e., of any order of magnitude) when experimental imperfections are considered.
For any finite number of parts, measurements, and outcomes in a Bell scenario, we estimate the probability of random N-qudit pure states to substantially violate any Bell inequality with uniformly bounded coefficients. We prove that under some conditions on the local dimension, the probability to find any significant amount of violation goes to zero exponentially fast as the number of parts goes to infinity. In addition, we also prove that if the number of parts is at least 3, this probability also goes to zero as the local Hilbert space dimension goes to infinity.
We consider the expected violations of Bell inequalities from random pure states. More precisely, we focus on a slightly generalized version of the Collins-Gisin-Linden-Massar-Popescu inequality, which concerns Bell experiments of two parties, two measurement options, and
N outcomes, and analyze their expected quantum violations from random pure states for varying
N, assuming the conjectured optimal measurement operators. It is seen that for small
N the Bell inequality is not violated on average, while for larger
N it is. Both ensembles of unstructured as well as structured random pure states are considered. Using techniques from random matrix theory this is obtained analytically for small and large
N and numerically for intermediate
N. The results show a beautiful interplay of different aspects of random matrix theory, ranging from the Marchenko-Pastur distribution and fixed-trace ensembles to the
O(n) model.