Multiscale Space-Time Ansatz Based on Quantics Tensor Trains for Quantum Field Theories

Abstract

We propose a new compression method for correlation functions, i.e. quantics tensor train (QTT), based on length-scale separations in space and time. In QTT, a correlation function is represented as a Matrix Product State (MPS). Discrete Fourier Transformation and convolution can be performed efficiently using a Matrix Product Operator (MPO). Numerous examples, ranging from equilibrium to non-equilibrium problems, demonstrate the high efficiency of QTT.