Figure 1

Conceptual representation of (a) nonlocality using entanglement and (b) nonlocality without entanglement. The former requires joint preparation through an entangling interaction, but can be observed with local measurement and classical communication (LOCC). The latter involves separate (local) preparation, but is revealed by joint measurements, i.e., measurements not realizable with LOCC.Reuse & Permissions

Figure 2

Conceptual experimental setup. Pairs of degenerate unentangled photons in the state $|H⟩\otimes |V⟩$ are generated from a 2.5 mm type-II beta-barium borate parametric down-conversion source operating in the collapsed cone mode [31, 32], pumped by an ${\mathrm{Ar}}^{+}$ ion laser operating at 351.1 nm and 600 mW. The down-converted photons are collected into fibers to spatially filter the modes, and are prepared in identical/orthogonal product states by a half and quarter wave plate in each arm. To perform the joint measurement, we use a two photon linear optics controlled-not gate [25] operating with an average gate fidelity of $>90\%$, followed by polarization analyzers, 0.36 nm FWHM interference filters, and single photon counting modules. With the output analyzers set to measure diagonal and antidiagonal in the control arm, and horizontal and vertical in the target arm, a fully resolving Bell measurement is realized whenever two photons are detected in coincidence. For our demonstration, only two detectors are used, and the probabilities of the four Bell states are built up from measurement statistics on a large ensemble of identical inputs.Reuse & Permissions

Figure 3

Probability of correctly identifying the input state correlation (identical or orthogonal) for each of the input states used in the experiment. The bars represent a theoretical optimal joint measurement $\{{\Pi }_S,{\Pi }_A\}$; the values measured experimentally, where the cnot is used to realize $\{{\Pi }_S,{\Pi }_A\}$; and a theoretical optimal local measurement in the $H,V$ basis [33]. The overall average payoffs, represented by horizontal lines, are 0.75, $0.72\pm 0.02$, and $2/3$, respectively. Poissonian counting errors contribute to errors in the individual bars on the order of 0.005.Reuse & Permissions