Learning effective spin Hamiltonian of quantum magnet
李伟 副教授
报告摘要:Finding the effective model of correlated quantum materials has long been a challenging problem for the community, as it constitutes a inverse quantum many-body problem. In this talk, we present a novel methodology of finding the microscopic Hamiltonian parameters by optimizing the goodness of fit between the theoretical many-body calculations and experimental thermodynamics data. We implement two different optimization strategies: a local gradient-based search with LBFGS restart and a global Bayesian optimization, and compare them with the conventional random grid search. We employ the automatic searching methodology to find the microscopic quantum models of several realistic magnetic materials, including the spin-chain compound Copper Nitrate, as well as frustrated 2D magnets TmMgGaO4 and RuCl3.