Scalable Circuit Cutting: A Framework for Combined Gate and Wire Cuts Using Gate Groups

As quantum circuits grow in complexity, they often exceed the qubit capacity of current hardware. Circuit cutting is a technique to partition these large circuits into smaller, executable subcircuits. However, the process introduces a significant sampling overhead that grows exponentially with the number of cuts. This paper addresses the challenge of finding optimal cut placements—deciding where to cut to minimize this overhead—by proposing a unified framework that considers gate cuts, wire cuts, and joint gate group cuts simultaneously.

The authors propose a two-stage, multi-level partitioning strategy to solve the cut placement problem. First, they construct an interaction graph based on the spatial structure of the circuit to identify an initial, balanced gate cut. Second, they refine this partition using a circuit topology graph that incorporates temporal information, allowing the algorithm to swap gate cuts for wire cuts where it reduces the total sampling overhead. Throughout both stages, the framework identifies and exploits gate groups—such as parallel gates or cascades—to perform joint cuts, which are more efficient than cutting gates individually.

The framework provides a scalable way to handle large circuits by treating cut placement as a graph partitioning problem. By integrating joint cutting, the method achieves significant reductions in sampling overhead compared to approaches that treat gate and wire cuts in isolation. The authors demonstrate that their approach is not only more efficient but also provides diagnostic feedback, helping researchers determine if a specific circuit is suitable for cutting given the hardware constraints.

Practical quantum computing is currently limited by the number of available qubits. By enabling more efficient circuit decomposition, this framework allows researchers to execute larger algorithms on existing hardware. Because it scales to circuits with hundreds of qubits, it bridges the gap between theoretical circuit cutting and practical, large-scale quantum software stacks.