The fundamental probabilistic nature of light makes it impossible to perfectly distinguish light from dark at very low intensity. Low power and high fidelity in reading data are especially important for secure communications and quantum computation; to facilitate such capabilities, a detector that sees well in the dark is crucial. Now, using a scheme based on quantum mechanics called unambiguous state discrimination (USD), researchers at the Joint Quantum Institute have achieved the lowest error rate yet for a photodetector deciphering a fourfold phase encoding of information in a light pulse. A scheme devised by scientists at the Joint Quantum Institute for carrying out unambiguous state discrimination (USD). Inset (i) shows the four nonorthogonal symmetric coherent states with phases equal to φ = {0, π/2, π, 3π/2}. The state under measurement, |ai>, has vertical (V) polarization, and the phase reference, |LO>, has horizontal (H) polarization. The pulse is distributed among four elimination stages using mirrors (M) and beamsplitters (BS). Each elimination stage uses phase shifters (PS), a polarizer (Pol) and a single-photon detector (SPD) to eliminate one possibility for the phase of the input state |ai>. Photo courtesy of JQI. The photodetection system the investigators demonstrated can make highly accurate readings of incoming information at the single-photon level by allowing the detector not to give a conclusive answer in some instances. Instead of the traditional two states – zeros and ones – the researchers used a more sophisticated data encoding scheme of four states: 0, 1, 2 and 3. These states correspond to four different phases of the light pulse, but with some overlap, which can produce ambiguity when trying to distinguish from which state the information was received. This inherent overlap means that the measurement system can sometimes provide an inconclusive answer. By implementing a USD of fourfold phase-encoded states, Alan Migdall and colleagues eliminated all but one possible value for the input state, achieving an error rate four times lower than is possible with conventional measurement techniques. Previous measurements were done using a minimum error discrimination scheme. “The former is a technique that always gets an answer, albeit with some probability of being mistaken, while the latter is designed to get answers that in principle are never wrong, but at the expense of sometimes getting an answer that is the equivalent of ‘don’t know,’ ” Migdall said. The approach will be useful for quantum information processing and quantum communications with many states, and for fundamental quantum measurement studies at low-light levels, the investigators say. The study appeared in Nature Communications (doi: 10.1038/ncomms3028).