Project description
Advancing nonequilibrium quantum control with AI-driven quantum simulators
Controlling quantum matter in out-of-equilibrium states presents a significant challenge in modern physics. The ERC-funded QuSimCtrl project aims to address this by expanding control from equilibrium to out-of-equilibrium states. Using quantum simulation, it will explore phenomena not seen in conventional materials and tackle the challenge of manipulating intensely driven nonequilibrium systems. Integrating quantum control with AI, particularly reinforcement learning (RL), the project will develop frameworks for nonadiabatic many-body state control, improving manipulation in cold atoms, trapped ions, and quantum solids. Its goal is to uncover guiding principles for many-body control away from equilibrium, linking quantum dynamics, statistical mechanics, and machine learning. This research is crucial for innovating materials and technologies based on nonequilibrium processes in condensed matter, quantum optics and quantum computing.
Objective
The ability to control quantum matter in a state of equilibrium is a milestone of 20th-century physics. A major goal of modern physics is to extend this knowledge to out-of-equilibrium systems. Located at the boundary between equilibrium and nonequilibrum, quantum simulation appears particularly suitable for this purpose. Using periodic drives, quantum simulators can experimentally emulate phenomena hitherto inaccessible in conventional materials, such as artificial gauge fields or topological and dynamically localized matter. However, our understanding of how to manipulate systems exposed to intense nonequilibrium drives is in its infancy, especially regarding strongly interacting models.
We propose to overcome the current limitations by combining ideas from quantum control and artificial intelligence (AI) algorithms. We will develop a new theoretical framework for nonadiabatic many-body state control on top of strong periodic drives underlying the optimal manipulation of ordered prethermal states of matter without equilibrium counterparts. Understanding this many-body dynamics will improve cutting-edge manipulation techniques in cold atoms, trapped ions, superconducting circuits, and quantum solids.
We will add reinforcement learning (RL), one of the most promising techniques in AI, to the quantum entanglement control toolbox. Deep RL has the potential to push the state-of-the-art of (dis-)entangling quantum states since it is capable of identifying effective degrees of freedom even when no underlying physical structure is immediately obvious.
Discovering guiding principles of physics for many-body control away from equilibrium has the potential to reveal new connections across quantum dynamics, statistical mechanics, optimal control, and machine learning. The proposed research establishes a missing link on the roadmap for designing future materials and technologies based on nonequilibrium processes in condensed matter, quantum optics, and quantum computing.
Fields of science
- natural sciencescomputer and information sciencesartificial intelligencemachine learningreinforcement learning
- natural sciencesphysical sciencesoptics
- natural sciencesphysical sciencesclassical mechanicsstatistical mechanics
- natural sciencesphysical sciencesquantum physicsquantum optics
- natural sciencesphysical scienceselectromagnetism and electronicssuperconductivity
Keywords
Programme(s)
- HORIZON.1.1 - European Research Council (ERC) Main Programme
Topic(s)
Funding Scheme
HORIZON-ERC - HORIZON ERC GrantsHost institution
80539 Munchen
Germany