Multi-Axis Control of a Qubit in the Presence of Unknown Non-Markovian Quantum Noise

TL;DR

This paper develops a method for controlling a qubit affected by unknown non-Markovian quantum noise using empirical modeling and optimization techniques, enabling effective open-loop control with finite sampling considerations.

Contribution

It introduces a graybox modeling approach combined with gradient and genetic optimization for open-loop qubit control under complex quantum noise.

Findings

Effective control pulses derived from empirical models

Comparison of gradient descent and genetic optimization methods

Demonstrated control in single- and multi-axis scenarios

Abstract

In this paper, we consider the problem of open-loop control of a qubit that is coupled to an unknown fully quantum non-Markovian noise (either bosonic or fermionic). A graybox model that is empirically obtained from measurement data is employed to approximately represent the unknown quantum noise. The estimated model is then used to calculate the open-loop control pulses under constraints on the pulse amplitude and timing. For the control pulse optimization, we explore the use of gradient descent and genetic optimization methods. We consider the effect of finite sampling on estimating expectation values of observables and show results for single- and multi-axis control of a qubit.