Instrumental distortions in quantum optimal control

Abstract

Quantum optimal control methods, such as gradient ascent pulse engineering (GRAPE), are used for precise manipulation of quantum states. Many of those methods were pioneered in magnetic resonance spectroscopy where instrumental distortions are often negligible. However, that is not the case elsewhere: the usual jumble of cables, resonators, modulators, splitters, amplifiers, and filters can and would distort control signals. Those distortions may be non-linear, their inverse functions may be ill-defined and unstable; they may even vary from one day to the next, and across the sample. Here we introduce the response-aware gradient ascent pulse engineering (RAW-GRAPE) framework, which accounts for any cascade of differentiable distortions within the GRAPE optimisation loop, does not require filter function inversion, and produces control sequences that are resilient to user-specified distortion cascades with user-specified parameter ensembles. The framework is implemented into the optimal control module supplied with versions 2.10 and later of the open-source Spinach library; the user needs to provide function handles returning the actions by the distortions and, optionally, parameter ensembles for those actions.