发表机构
South East Technological University; Trinity College Dublin（东南理工大学; 都柏林三一学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究提出量子网络中控制信息传输导致的几何退相干限制，通过功能图抽象比较集中与本地资源分配，发现本地决策延迟几乎不随规模增长，并指出控制器放置等为物理设计参数。
AI 中文摘要
量子网络通常受两个硬件限制的约束：纠缠可以被宣告的速度，以及存储器能保持纠缠的时间。我们建立了第三个几何限制：当控制信息传输时，纠缠发生退相干，因此到资源分配者的距离直接计入保真度预算。我们开发了一种功能图抽象，每条链路根据其对端到端保真度的预测乘法贡献进行加权，将全局但延迟的状态与局部且即时的状态分开，并在复制的离散事件模型中比较集中式与本地分配。集中式交付延迟随网络直径增长，而本地延迟几乎与规模无关，对于短距离流量差异可达二十倍；全局视图优于新鲜视图的相干阈值在所测试的规模上不发生变化；随着宣告加速，主导成本转移到控制平面。时隙同步则遵循时隙速率与单向链路延迟的乘积，因此受中继器间距而非直径的限制。控制器放置、决策局部性、中继器间距和时隙速率是物理设计参数，而非实现细节。
英文摘要
Quantum networks are usually framed by two hardware limits: how fast entanglement can be heralded, and how long a memory can hold it. We establish a third, geometric limit: entanglement decoheres while control information travels, so the distance to whoever allocates resources enters the fidelity budget directly. We develop a functional-graph abstraction weighting each link by its predicted multiplicative contribution to end-to-end fidelity, separating global but delayed state from local and immediate state, and compare centralized against local allocation in a replicated discrete-event model. Centralized delivery latency grows with network diameter while local latency is nearly scale invariant, differing up to twenty-fold for short-range traffic; the coherence threshold at which a global view outweighs a fresh one does not shift across the tested sizes; and as heralding accelerates, the dominant cost moves into the control plane. Slot synchronization instead obeys the product of slot rate and one-way link delay, so it is bounded by repeater spacing rather than diameter. Controller placement, decision locality, repeater spacing and slot rate are physical design parameters, not implementation details.
