RL-Synthesised Quantum Circuits: A Novel Lens for Phase Transitions in Many-Body Systems

Keywords: Reinforcement Learning, Tensor Networks, Quantum Computation, Many-body Physics, Phase Transitions

TL;DR: When an RL agent is trained on quantum circuit discovery for a many-body problem, the resultant circuit complexity encodes information about the phase transition.

Abstract:

Quantum computing utilises the fundamental
properties of quantum mechanics to carry out
computations. The quantum circuit complexity of
a computation has embedded information about
important questions in many-body physics. In this
paper, we train a reinforcement learning agent to
synthesise quantum circuits that retrieve the time
evolution operator of the transverse field Ising
Hamiltonian from a prepared starting state. We
formalise the problem as three Markov Decision
Processes and show that the tensor network implementation outperforms other implementations and
accurately encodes information about the phase
transition boundary of the Hamiltonian by showing a stark decrease in circuit complexity at the
transition point.

Submission Number: 58