The control of quantum dynamics is key at both the fundamental and technological level. Over the years, various techniques have been devised to reach the important goal of controlling the fundamental processes that are responsible fro the evolution of a quantum system, from optimal control to shortcuts to adiabaticity. In a visionary manner, the development of such techniques will leads us to the design of controlled dynamics able to showcase quantum features all the way up to mesoscopic and macroscopic size-scales.
A very promising approach to controlled quantum dynamics is built on the merging of methods of machine learning with the fundamental building blocks of quantum dynamics. The application of such techniques is finding a growing interest in the community working on quantum technologies and open quantum systems in light of their potential and promises.
The work by TEQ member Mauro Paternostro (Queen’s University Belfast) with PhD student Pierpaolo Sgroi and in collaboration with Prof G Massimo Palma (University of Palermo) demonstrates the successful application of reinforcement learning techniques to the control and engineering of energy-exchange processes occurring at the quantum level, showing the possibility to control energy dissipation and irreversibility resulting from non-equilibrium quantum dynamics. This work paves the way to the design of energy-efficient protocols able to harness closed and open quantum evolution and will impact in fields such as quantum computation and quantum transport.
The paper made it to the cover of Physical Review Letters.
Link to the publication: Phys. Rev. Lett. 126, 020601 (2021) – Reinforcement Learning Approach to Nonequilibrium Quantum Thermodynamics (aps.org)