Noise reconstruction in quantum hardware via convex optimization

Interactions between a quantum system and noisy control hardware, or its environment, critically limits the performance and capabilities of noisy intermediate-scale quantum (NISQ) devices, as well as future quantum computing technologies. Accurately characterizing the noise profile of these systems is of central importance in developing techniques to improve hardware performance. This includes detailed microscopic characterization of time-dependent noise processes. We introduce a novel machine-learning technique allowing the efficient, flexible, and quantitatively accurate reconstruction of a noise process’s frequency-resolved power spectral density. By reformatting PSD reconstruction based on a measurement record as a convex optimization problem, our algorithm does not need to assume any specific shape of the noise spectral and can easily incorporate physically-motivated constraints for reconstruction without the loss of numerical efficiency. Moreover, this technique permits reconstruction using an arbitrary set of measurements, relaxing constraints previously imposed to limit the introduction of numerical artefacts. We present details of the approach as well as experimental demonstrations of trapped-ion motional-mode noise characterization via spin-motional entanglement.