Optimization of path-integral tensor-multiplication schemes in open quantum systems

TL;DR

This paper introduces an optimized tensor-multiplication scheme for path-integral methods in open quantum systems, significantly reducing computational resources and enabling analysis of more complex, long-memory quantum systems.

Contribution

It presents a novel tensor optimization method using matrix representation and SVD, improving efficiency and extending the applicability of path-integral techniques in quantum dynamics.

Findings

Reduces tensor sizes and computational time dramatically.

Enables simulations requiring over 50 million GB of RAM on standard computers.

Successfully applied to quantum dot-microcavity systems with strong coupling.

Abstract

Path-integral techniques are a powerful tool used in open quantum systems to provide an exact solution for the non-Markovian dynamics. However, the exponential scaling of the tensor size with quantum memory length of these techniques limits the applicability when applied to systems with long memory times. Here we provide a general optimization of tensor multiplication schemes for systems with pair correlations and finite memory times. This optimization effectively reduces the tensor sizes by using a matrix representation of tensors combined with singular value decomposition to filter out negligible contributions. This approach dramatically reduces both computational time and memory usage of the traditional tensor-multiplication schemes. Calculations that would require over 50 million GB of RAM in the original approach are now available on standard computers, allowing access to new regimes and more complex systems. While more memory-efficient representations exist, this approach enables an extrapolation scheme that other techniques do not offer. As a demonstration, we apply it to the Trotter decomposition with linked cluster expansion technique, and use it to investigate a quantum dot-microcavity system at larger coupling strengths than previously achieved. We also apply the optimization in a case where the memory time is very long -- specifically in a system containing two spatially separated quantum dots in a common phonon bath.