Symbolic Synthesis of Clifford Circuits and Beyond

TL;DR

This paper introduces methods for analyzing and synthesizing Clifford circuits from path sums, a symbolic formalism for quantum operations, including algorithms for circuit extraction and generalization beyond Clifford cases.

Contribution

It provides the first polynomial-time algorithm for extracting Clifford circuits from Clifford path sums and proposes a heuristic for more general path sums.

Findings

Unitarity problem is co-NP-hard in general but in P for Clifford path sums.

Algorithm successfully synthesizes Clifford circuits in C1-H-C2 form.

Heuristic generalization often produces natural-looking circuits, demonstrated on quantum Fourier transform.

Abstract

Path sums are a convenient symbolic formalism for quantum operations with applications to the simulation, optimization, and verification of quantum protocols. Unlike quantum circuits, path sums are not limited to unitary operations, but can express arbitrary linear ones. Two problems, therefore, naturally arise in the study of path sums: the unitarity problem and the extraction problem. The former is the problem of deciding whether a given path sum represents a unitary operator. The latter is the problem of constructing a quantum circuit, given a path sum promised to represent a unitary operator. In this paper, we show that the unitarity problem is co-NP-hard in general, but that it is in P when restricted to Clifford path sums. We then provide an algorithm to synthesize a Clifford circuit from a unitary Clifford path sum. The circuits produced by our extraction algorithm are of the form C1-H-C2, where C1 and C2 are Hadamard-free circuits and H is a layer of Hadamard gates. We also provide a heuristic generalization of our extraction algorithm to arbitrary path sums. While this algorithm is not guaranteed to succeed, it often succeeds and typically produces natural looking circuits. Alongside applications to the optimization and decompilation of quantum circuits, we demonstrate the capability of our algorithm by synthesizing the standard quantum Fourier transform directly from a path sum.