• Current opened records

  • Inferring underlying Heisenberg Hamiltonian from a spin spectral function for a quantum spin liquid by a neural network

Awards
Author(s):
Category:
Institution:
Region:
Winner Category:
Year:
Abstract:

  • A quantum spin liquid is a phase of matter where the spins are disordered and frustrated, and the system is highly entangled even at absolute zero. Due to the exotic excitations quantum spin liquids can host, it is believed that they could be used in quantum computers and other advance quantum technologies. Quantum spin liquids have become more topical and intriguing research topic over the past few years, as couple experimental studies claim to have discovered the existence of this non-magnetic quantum phase. Due to these statements, there is a need to develop access to the Hamiltonian of the system, which describes the exact quantum structure of material. Because
    the determination of Heisenberg Hamiltonian is impossible via analytical calculations due to the nature of the quantum spin liquid, neural networks have been used in this research.

    The aim of the conducted research was to offer a way to predict Heisenberg exchange interaction couplings J1, J2, and J3 for a rectangle nano lattice in a quantum spin liquid state. The prediction of the parameters was performed by using supervised machine learning. An artificial neural network was trained with computational data that contained spin spectral functions and the parameters of the lattice from around ten thousand samples. Computational spectral functions were generated by using exact diagonalization when Hamiltonians of the lattices were first specified. The network predicted the parameters from spin spectral functions and was found to be suitable for lattices 2x3,
    2x4, 2x5, 2x6, 2x8, and 4x4 with less than 2% error.

    In this research it was shown that the trained network is able to predict Heisenberg exchange interaction coupling constants accurately from noisy computational spctral data. With further
    research, the network could be used to predict the exchange coupling energies of a potential quantum spin liquid from experimental data. The investigation process on the theory and applications of quantum spin liquids could thereby be facilitated by giving access to a quantum mechanical
    structure of the system.

Attached Documents: