Boosting linear-optical Bell measurement success probability with predetection squeezing and imperfect photon-number-resolving detectors

Abstract

Linear optical realizations of Bell state measurement (BSM) on two single-photon qubits succeed with probability $p_s$ no higher than $0.5$. However pre-detection quadrature squeezing, i.e., quantum noise limited phase sensitive amplification, in the usual linear-optical BSM circuit, can yield ${p_s \approx 0.643}$. The ability to achieve $p_s > 0.5$ has been found to be critical in resource-efficient realizations of linear optical quantum computing and all-photonic quantum repeaters. Yet, the aforesaid value of $p_s > 0.5$ is not known to be the maximum achievable using squeezing, thereby leaving it open whether close-to-$100\%$ efficient BSM might be achievable using squeezing as a resource. In this paper, we report new insights on why squeezing-enhanced BSM achieves $p_s > 0.5$. Using this, we show that the previously-reported ${p_s \approx 0.643}$ at single-mode squeezing strength $r=0.6585$---for unambiguous state discrimination (USD) of all four Bell states---is an experimentally unachievable point result, which drops to $p_s \approx 0.59$ with the slightest change in $r$. We however show that squeezing-induced boosting of $p_s$ with USD operation is still possible over a continuous range of $r$, with an experimentally achievable maximum occurring at $r=0.5774$, achieving ${p_s \approx 0.596}$. Finally, deviating from USD operation, we explore a trade-space between $p_s$, the probability with which the BSM circuit declares a "success", versus the probability of error $p_e$, the probability of an input Bell state being erroneously identified given the circuit declares a success. Since quantum error correction could correct for some $p_e > 0$, this tradeoff may enable better quantum repeater designs by potentially increasing the entanglement generation rates with $p_s$ exceeding what is possible with traditionally-studied USD operation of BSMs.