Automatic Depth-Optimized Quantum Circuit Synthesis for Diagonal Unitary Matrices with Asymptotically Optimal Gate Count

TL;DR

This paper introduces a depth-optimized quantum circuit synthesis algorithm for diagonal unitaries that nearly halves circuit depth while maintaining asymptotically optimal gate count, crucial for NISQ device limitations.

Contribution

It presents a novel synthesis method that reduces circuit depth by nearly 50% compared to previous approaches, with theoretical analysis and experimental validation.

Findings

Achieves ~50% depth reduction for random diagonal unitaries up to 16 qubits.

Reduces average circuit depth by 22.05% for QAOA diagonal parts up to 14 qubits.

Ensures asymptotically optimal gate count while optimizing circuit depth.

Abstract

Current noisy intermediate-scale quantum (NISQ) devices can only execute small circuits with shallow depth, as they are still constrained by the presence of noise: quantum gates have error rates and quantum states are fragile due to decoherence. Hence, it is of great importance to optimize the depth/gate-count when designing quantum circuits for specific tasks. Diagonal unitary matrices are well-known to be key building blocks of many quantum algorithms or quantum computing procedures. Prior work has discussed the synthesis of diagonal unitary matrices over the primitive gate set $\{\text{CNOT}, R_Z\}$. However, the problem has not yet been fully understood, since the existing synthesis methods have not optimized the circuit depth. In this paper, we propose a depth-optimized synthesis algorithm that automatically produces a quantum circuit for any given diagonal unitary matrix. Specially, it not only ensures the asymptotically optimal gate-count, but also nearly halves the total circuit depth compared with the previous method. Technically, we discover a uniform circuit rewriting rule well-suited for reducing the circuit depth. The performance of our synthesis algorithm is both theoretically analyzed and experimentally validated by evaluations on two examples. First, we achieve a nearly 50\% depth reduction over Welch's method for synthesizing random diagonal unitary matrices with up to 16 qubits. Second, we achieve an average of 22.05\% depth reduction for resynthesizing the diagonal part of specific quantum approximate optimization algorithm (QAOA) circuits with up to 14 qubits.