Operator subspace based method for the extraction of higher energy levels in Lattice field theoretic systems

Abstract

Given the importance of spectral analysis of lattice quantum field theory data, the advancement of techniques for the same remains critically important. Therefore, we propose an algorithm that uses the first three time slices of the correlation function matrix to construct the transfer matrix and extract its energy eigenvalues. Reduction of systematic error arising from truncating the operator basis is achieved by varying the operator subspace dimension and applying eigenvalue-variance extrapolation. Since the correlation functions in early time slices typically exhibit small statistical errors and retain a strong signal of higher-excited states, we expect our proposed method to perform well in extracting higher-excited-state energies. In this paper, we conduct two lattice Monte Carlo simulations for quantum mechanical systems, harmonic and anharmonic oscillators, to evaluate the efficiency of our proposed method for higher-excited-state energies. We compare the energy levels obtained using the standard GEVP method and our proposed method. We find that our method consistently outperforms the standard GEVP method in extracting intermediate excited states for both systems. Because our method has access only to the first three time slices, it is not preferable to the GEVP method for extracting the ground-state energy in its current form.