Spin-qubit readout analysis based on a hidden Markov model

Abstract

Across most qubit platforms, the readout fidelities do not keep up with the gate fidelities, and new ways to increase the readout fidelities are searched for. For semiconductor spin qubits, a typical qubit-readout signal consists of a finite stretch of a digitized charge-sensor output. Such a signal trace is usually analyzed by compressing it into a single value, either maximum or sum. The binary measurement result follows by comparing the single value to a decision threshold fixed in advance. This threshold method, while simple and fast, omits information that could potentially improve the readout fidelity. Here, we analyze what can be achieved by more sophisticated signal-trace processing using the hidden Markov model (HMM). The HMM is a natural choice, being the optimal statistical processing if the noise is white. It also has a computationally efficient implementation, known as the forward-backward algorithm, making HMM processing practical. However, unlike in many computer-simulation studies, in real experiments the noise is correlated. How this change affects the HMM implementation and reliability is our subject. We find that the HMM using white noise as the system statistical model is surprisingly sensitive to correlations; it only tolerates very small correlation times. We suggest alleviating this deficiency by a signal prefiltering. The correlations have a similar strongly negative impact on the HMM model calibration (the Baum-Welch algorithm). Besides studying the effects of noise correlations, as a specific application of the HMM we calculate the readout fidelity at elevated temperatures, relevant to recent experimental pursuits of hot spin qubits.