Machine learning the dynamics of quantum kicked rotor

DOI10.1016/J.AOP.2021.168500MaRDI QIDQ2063544zbMATH OpenOpenAlexFDO

Authors Tomohiro Mano, Tomi Ohtsuki

Publication date 11 January 2022

Published in Annals of Physics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2101.09432

zbMATH Keywords

Anderson transitionconvolutional neural networklong short-term memory networkmachine learningquantum kicked rotorquantum phase transition

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05) Dynamic and nonequilibrium phase transitions (general) in statistical mechanics (82C26) Neural nets applied to problems in time-dependent statistical mechanics (82C32) Dynamics of disordered systems (random Ising systems, etc.) in time-dependent statistical mechanics (82C44)

Abstract: Using the multilayer convolutional neural network (CNN), we can detect the quantum phases in random electron systems, and phase diagrams of two and higher dimensional Anderson transitions and quantum percolations as well as disordered topological systems have been obtained. Here, instead of using CNN to analyze the wave functions, we analyze the dynamics of wave packets via long short-term memory network (LSTM). We adopt the quasi-periodic quantum kicked rotors, which simulate the three and four dimensional Anderson transitions. By supervised training, we let LSTM extract the features of the time series of wave packet displacements in localized and delocalized phases. We then simulate the wave packets in unknown phases and let LSTM classify the time series to localized and delocalized phases. We compare the phase diagrams obtained by LSTM and those obtained by CNN.

Review text

The aim here is to investigate the Anderson transitions by using a procedure based on the quantum kicked rotor (QKR) model and convolutional neural networks (CNN). One considers the following model for QKR with incommensurate modulation of the kick: $$H(t) =\frac {p^2}{2}+K\cos x \sum_{n}\delta(t-n)F(t).$$ There are two ways in which the system is analyzed. The Hamiltonian is mapped to static higher-dimensional tight binding models first. Then, the eigenfunctions are studied by means of convolution neural networks. The second way involves the study of the time series of the original time-dependent one-dimensional system by using bidirectional long short-term memory network. Simulation results are shown and discussed in the second section of the article. The authors state that, for the investigation on Anderson transitions in higher dimensions, ``the second approach works especially well.

Recommendations

• Machine learning versus semidefinite programming approach to a particular problem of the theory of open quantum systems
• Approximation of quantum control correction scheme using deep neural networks
• Learning dynamical systems in noise using convolutional neural networks
• Learning quantum structures in compact localized eigenstates
• Machine learning as a universal tool for quantitative investigations of phase transitions

Cites work

• Quantum signatures of chaos
• Viscosity of the Electron Gas in Metals

Cited in

(2)

• Editorial. Special issue ``Localisation 2020: editorial summary
• Dynamical transition of quantum scrambling in a non-Hermitian Floquet synthetic system

This page was built for publication: Machine learning the dynamics of quantum kicked rotor

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2063544)