Efficient Simulation of High-Level Quantum Gates

TL;DR

This paper introduces a gadget-based quantum circuit simulator that directly handles high-level gates, reducing overhead and improving simulation efficiency by leveraging stabilizer decompositions and establishing bounds on stabilizer ranks.

Contribution

The authors develop a novel simulator that directly simulates high-level quantum gates, avoiding costly compilation, and provide bounds on stabilizer ranks to enhance simulation performance.

Findings

Improved simulation efficiency over standard methods.

Established small stabilizer rank bounds for common high-level gates.

Derived exponential lower bounds for stabilizer ranks under complexity hypotheses.

Abstract

Quantum circuit simulation is paramount to the verification and optimization of quantum algorithms, and considerable research efforts have been made towards efficient simulators. While circuits often contain high-level gates such as oracles and multi-controlled X ($C^k$X) gates, existing simulation methods require compilation to a low-level gate-set before simulation. This, however, increases circuit size and incurs a considerable (typically exponential) overhead, even when the number of high-level gates is small. Here we present a gadget-based simulator which simulates high-level gates directly, thereby allowing to avoid or reduce the blowup of compilation. Our simulator uses a stabilizer decomposition of the magic state of non-stabilizer gates, with improvements in the rank of the magic state directly improving performance. We then proceed to establish a small stabilizer rank for a range of high-level gates that are common in various quantum algorithms. Using these bounds in our simulator, we improve both the theoretical complexity of simulating circuits containing such gates, and the practical running time compared to standard simulators found in IBM's Qiskit Aer library. We also derive exponential lower-bounds for the stabilizer rank of some gates under common complexity-theoretic hypotheses. In certain cases, our lower-bounds are asymptotically tight on the exponent.