A fast hybrid classical-quantum algorithm based on block successive over-relaxation for the heat differential equation

TL;DR

This paper introduces a hybrid classical-quantum algorithm leveraging block successive over-relaxation to efficiently solve high-dimensional heat equations, significantly reducing computation time with limited quantum resources.

Contribution

It develops a novel hybrid classical-quantum approach using block SOR and Advantage quantum computers to accelerate PDE solutions with limited qubits.

Findings

Achieves up to 2x speedup over existing methods.

Effectively decomposes large systems into smaller subsystems.

Demonstrates efficiency in solving high-dimensional heat equations.

Abstract

The numerical solution of partial differential equations (PDEs) is essential in computational physics. Over the past few decades, various quantum-based methods have been developed to formulate and solve PDEs. Solving PDEs incur high time complexity for real-world problems with high dimensions, and using traditional methods becomes practically inefficient. This paper presents a fast hybrid classical-quantum paradigm based on successive over-relaxation (SOR) to accelerate solving PDEs. Using the discretization method, this approach reduces the PDE solution to solving a system of linear equations, which is then addressed using the block SOR method. Due to limitations in the number of qubits, the block SOR method is employed, where the entire system of linear equations is decomposed into smaller subsystems. These subsystems are iteratively solved block-wise using Advantage quantum computers developed by D-Wave Systems, and the solutions are subsequently combined to obtain the overall solution. The performance of the proposed method is evaluated by solving the heat equation for a square plate with fixed boundary temperatures and comparing the results with the best existing method. Experimental results show that the proposed method can accelerate the solution of high-dimensional PDEs by using a limited number of qubits up to 2 times the existing method.