Learning to Calibrate Quantum Control Pulses by Iterative Deconvolution

TL;DR

This paper introduces an iterative, learning-based deconvolution method to improve quantum control pulse calibration, effectively correcting model inaccuracies and reducing errors caused by finite sampling rates.

Contribution

It presents a novel iterative deconvolution scheme that learns to correct both linear and nonlinear model errors in quantum control pulse calibration.

Findings

The method achieves high-precision correction of model errors.

It effectively reduces calibration errors caused by finite sampling rates.

Active nonlinear components can suppress inter-sampling errors.

Abstract

In experimental control of quantum systems, the precision is often hindered by imperfect applied electronics that distort control pulses delivered to target quantum devices. To mitigate such error, the deconvolution method is commonly used for compensating the distortion via an identified convolutional model. However, its effectiveness is limited by model inaccuracies (e.g., imprecise parameters or unmodeled distortion dynamics). In this paper, we propose a learning-based scheme to eliminate the residual calibration error by repeatedly applying the deconvolution operations. The resulting iterative deconvolution method is shown to be able to correct both linear and nonlinear model errors to the highest precision allowed by available finite sampling rates. The calibration error induced by finite sampling rates is also analyzed, from which we propose that the inter-sampling error can be suppressed by actively introducing nonlinear components in the control electronics.