量子电子学报 ›› 2026, Vol. 43 ›› Issue (4): 577-587.doi: 10.3969/j.issn.1007-5461.2026.04.007

• 量子线路设计自动化 • 上一篇    下一篇

BlockCut: 一种有效切割量子门块的方法(特邀)

张明飞 , 庞 浦 , 刘佳琦 , 张润哲 , 吴安邦 *   

  1. 上海交通大学计算机学院(网络空间安全学院、密码学院), 上海 200240
  • 收稿日期:2025-11-03 修回日期:2026-03-29 出版日期:2026-07-28 发布日期:2026-07-27
  • 通讯作者: E-mail: anbangwu@sjtu.edu.cn. E-mail:E-mail: anbangwu@sjtu.edu.cn.
  • 作者简介:张明飞 ( 2002 - ), 贵州铜仁人, 博士研究生, 主要从事量子电路优化方面的研究。E-mail: zloveq1314@sjtu.edu.cn
  • 基金资助:
    国家自然科学基金青年项目 (62502300)

BlockCut: An effective scheme to cut quantum gate blocks(Invited)

ZHANG Mingfei, PANG Pu, LIU Jiaqi, ZHANG Runzhe, WU Anbang   

  1. School of Computer Science (School of Cybersecurity, School of Cryptography),Shanghai Jiao Tong University, Shanghai 20024
  • Received:2025-11-03 Revised:2026-03-29 Published:2026-07-28 Online:2026-07-27

摘要: 线路切割技术为大型量子线路在当前量子硬件上运行提供了可能。该类技术将量子线路划分为若干子线路, 在硬件上分别执行后, 再通过对子线路的运行结果进行后处理从而重构整体线路的计算结果。然而, 现有的线路切割方法主要针对单一类型量子门进行划分, 导致昂贵的后处理和子线路执行成本。本文提出了一种新的线路切割方案—BlockCut, 该方案能够对更一般的、由多个量子门组成的门块进行切割, 突破了单一类型量子门的限制。BlockCut 对任意 n 量子比特门/门块都适用, 显著降低了子线路的数量和后处理代价。以 controlled Z(CZ)门为例, BlockCut只将其分解为5个子线路, 而现有的方法最少产生6个子线路。此外, BlockCut还支持与硬件架构的协同优化。以跨设备Toffoli门为例, BlockCut最多只产生10个设备局部可执行的子线路, 而现有方法将产生多达46656个子线路。

关键词: 量子计算, 量子编译, 量子线路切割

Abstract: Circuit cutting enables the execution of large-scale quantum circuits on current quantum hardware. This technique partitions a large quantum circuit into sub-circuits, which are executed separately on hardware, and then reconstructs the overall computation result through classical post-processing of the sub-circuit outcomes. However, existing circuit cutting methods consider only a few specific quantum gates, resulting in expensive post-processing and sub-circuit execution overhead. This paper introduces a new circuit cutting scheme, named BlockCut, which enables cutting more general gate blocks composed of multiple quantum gates, surpassing existing methods restricted to only a few gate types. BlockCut is applicable to arbitrary n-qubit gate blocks and significantly reduces both the number of sub-circuits and the post-processing overhead. For the controlled Z (CZ) gate, BlockCut decomposes it into only 5 sub-circuits, whereas existing methods require at least 6 sub-circuits. Furthermore, BlockCut also supports co-optimization with hardware architectures. For a Toffoli gate distributed across multiple devices, BlockCut generates at most 10 device-local sub-circuits, while existing methods will generate up to 46,656 sub-circuits.

Key words: quantum computation, quantum compilation, quantum circuit cutting

中图分类号: