Room-Temperature Quantum Non-Demolition Measurement Enhanced by Machine Learning

Projective measurements of qubits are a key resource for quantum computation. For qubits based on Nitrogen-Vacancy centers in diamond at room-temperature, projective measurement has been achieved using quantum non-demolition measurement schemes enabled by an ancillary qubit. In this scheme, the nuclear spin (qubit) state is repetitively read out by a mapping to the NV electronic spin (ancillae) until the photon number distribution from one qubit state is distinguishable from the other in a single shot. High readout fidelity requires a few to tens of thousands of repetitive readouts. This, unfortunately, imposes a heavy time overhead to any quantum algorithms. The readout time is on the same order of the decoherence time of the best physical qubit in the system, preventing feed forward protocols. In this work, we describe a method to improve the single shot readout fidelity using machine learning with information already recorded, but traditionally discarded in the experiment. Hence, our method does not impose an additional experimental time penalty. Combined with photonic structures that enhance photon collection efficiency, we expect this technique will enable room-temperature feed forward quantum information processing.