What performance gap emerges when separator size grows relative to total circuit vertices?

This is a solid question about quantum circuit optimization, but the preprints provided don't directly address the specific performance trade-off you're asking about—namely, how circuit quality degrades as separator size grows relative to total vertices.

The closest relevant work is [AG-2024.05-1606], which introduces the Deferred Constraint Quantum Divide and Conquer Algorithm (DC-QDCA). This method uses a *separator*—a set of vertices whose removal partitions a graph—to reduce communication costs in distributed quantum computing [AG-2024.05-1606]. The core idea is elegant: by deferring constraints associated with separator vertices, the algorithm simplifies circuit topology and cuts inter-device operations.

However, the abstract doesn't characterize the performance *trade-off* as separator size scales. A larger separator removes more vertices, which might partition the graph more cleanly (reducing circuit depth), but it also means more vertices have deferred constraints, which could increase overhead elsewhere. The paper presents experimental validation on simulated systems with "potentially thousands of variables" [AG-2024.05-1606], but doesn't quantify this specific gap.

Related work on quantum circuit mapping and routing—[AG-2024.02-1271], [AG-2024.07-2356]—addresses communication overhead from qubit movement, but focuses on SWAP gate insertion rather than separator-based decomposition.

To answer your question rigorously, you'd need the full paper [AG-2024.05-1606] or a dedicated ablation study on separator size versus circuit quality metrics (depth, gate count, fidelity).