Experimental non-Markovian process characterisation and control on a quantum processor

As experimentally available quantum devices increase in precision and accessibility, attention is turning to understanding and eliminating correlated -- or non-Markovian -- noise. Here, we develop and experimentally test a framework for characterising non-Markovian dynamics in quantum systems. Applying this technique over four different \emph{IBM Q} superconducting quantum devices, we achieve a consistent reconstruction infidelity of $10^{-3}$. Our approach is based on the recently proposed process tensor. With this, we infer the non-Markovian process by measuring the system's response to a basis of control operations. As a consequence, the effects of \emph{any} control operation in the span are discernible, independent of the interaction with the environment. With our technique, we demonstrate several applications: We first estimate a statistically significant lower-bound on memory size for all of the devices. We demonstrate that the dynamical characterisation remains high in fidelity where conventional Markovian characterisation models suffer from an appreciable reduction in quality. Finally, we turn these high fidelity predictions into an adaptive control technique for non-Markovian systems and demonstrate decoupling of a qubit interacting with another qubit with an unknown Hamiltonian. We further show how the coupling can be manipulated to implement a non-unitary gate of our choosing. The results and methods have widespread application both to the study of fully general dynamics, and to optimal device control.