NQS Study of Deconfined Quantum Criticality from a U(1) Quantum spin Liquid to a Spinon Bose-Einstein Condensate

Abstract

Artificial intelligence has significantly impacted modern scientific research, with quantum physics standing at the forefront of these developments. In this thesis, we employ Neural Quantum States (NQS), a variational framework based on neural network ansatze, for ab initio simulations of quantum many-body systems. Building upon this foundation, we address key challenges in the application of NQS, particularly their efficiency and expressivity in frustrated quantum magnets. Frustrated systems, in which competing interactions hinder classical ordering, are essential for studying emergent quantum phases. As a case study, we focus on quantum spin ice, a highly frustrated system known to host exotic excitations and a potential U(1) quantum spin liquid phase. By applying NQS to this system, we assess the accuracy, scalability, and limitations of this method in comparison to established approaches, such as variational Monte Carlo and tensor-network-based techniques. Our results highlight the capability of NQS to capture complex correlations in frustrated magnets, thereby advancing their role in the simulation of strongly correlated quantum matter. Keywords: Quantum many-body system, Neural quantum state, Frustrated Magnets, Quantum spin ice, U(1) quantum spin liquid.