Entanglement-Efficient Distribution of Quantum Circuits over Large-Scale Quantum Networks

TL;DR

This paper proposes a modified quantum circuit partitioning method optimized for large-scale quantum networks, reducing entanglement costs and improving scalability through network and problem coarsening techniques.

Contribution

It introduces a simple modification to existing partitioning schemes that efficiently considers network topology and entanglement constraints, enabling scalable quantum circuit distribution.

Findings

Lower entanglement costs compared to state-of-the-art methods

Coarsened methods improve solution quality and reduce run-times

Effective for various network topologies and large-scale quantum networks

Abstract

Quantum computers face inherent scaling challenges, a fact that necessitates investigation of distributed quantum computing systems, whereby scaling is achieved through interconnection of smaller quantum processing units. However, connecting large numbers of QPUs will eventually result in connectivity constraints at the network level, where the difficulty of entanglement sharing increases with network path lengths. This increases the complexity of the quantum circuit partitioning problem, since the cost of generating entanglement between end nodes varies with network topologies and existing links. We address this challenge using a simple modification to existing partitioning schemes designed for all-to-all connected networks, that efficiently accounts for both of these factors. We investigate the performance in terms of entanglement requirements and optimisation time of various quantum circuits over different network topologies, achieving lower entanglement costs in the majority of cases than state-of-the-art methods. We provide techniques for scaling to large-scale quantum networks employing both network and problem coarsening. We show that coarsened methods can achieve improved solution quality in most cases with significantly lower run-times than direct partitioning methods.