Type: report
Abstract: Understanding quantum many-body phenomena is a fundamental problem in physics.Since it is exponentially expensive to exactly analyze quantum many-body problems, it is essential to develop powerful numerical methods.For this purpose, machine learning methods have gathered much attention.Using artificial neural networks, one may be able to extract essential features of quantum correlations.We have developed a state-of-the-art machine learning method and elucidated the physics of quantum spin liquid in frustrated quantum spin systems.For these achievements, the researcher was awarded the "Condensed-Matter Science Prize" in November 2022.To make this powerful method even more powerful, it is essential to study machine learning itself, which is considered a black box.To this end, we decided to compare it to another powerful method, tensor networks.As a first step, the eigenstates of the 1D Heisenberg model on 18 spin chains were approximated by an artificial neural network and a tensor network, and their performance was compared.The results show that the artificial neural network has the same level of representational capability as the tensor network.This result reveals part of the reason why the artificial neural network method is so powerful in analyzing quantum many-body systems.In the future, we will try to develop a more powerful method based on these findings.
Author: 野村, 悠祐
Language: English
Topic: Quantum many-body systems
Subfield: Atomic and Molecular Physics, and Optics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7145354284