Fast Stabiliser Simulation with Quadratic Form Expansions

TL;DR

This paper introduces an efficient method for simulating stabiliser quantum circuits using quadratic form expansions, achieving comparable or better performance than existing techniques, especially for measurements and low-rank states.

Contribution

It presents a novel quadratic form expansion approach that enables faster simulation of stabiliser operations, measurements, and syndromes, with complexity improvements for specific cases.

Findings

Simulates stabiliser operations in O(n^2) time.

Allows constant-time simulation of certain measurements.

Achieves O(n) time for local stabiliser syndrome measurements.

Abstract

This paper builds on the idea of simulating stabiliser circuits through transformations of quadratic form expansions. This is a representation of a quantum state which specifies a formula for the expansion in the standard basis, describing real and imaginary relative phases using a degree-2 polynomial over the integers. We show how, with deft management of the quadratic form expansion representation, we may simulate individual stabiliser operations in $O(n^2)$ time matching the overall complexity of other simulation techniques [arXiv:quant-ph/0406196, arXiv:quant-ph/0504117, arXiv:1808.00128]. Our techniques provide economies of scale in the time to simulate simultaneous measurements of all (or nearly all) qubits in the standard basis. Our techniques also allow single-qubit measurements with deterministic outcomes to be simulated in constant time. We also describe throughout how these bounds may be tightened when the expansion of the state in the standard basis has relatively few terms (has low 'rank'), or can be specified by sparse matrices. Specifically, this allows us to simulate a 'local' stabiliser syndrome measurement in time $O(n)$, for a stabiliser code subject to Pauli noise -- matching what is possible using techniques developed by Gidney [arXiv:2103.02202] without the need to store which operations have thus far been simulated.