AI 中文总结
本文提出一种针对中性原子硬件的无测量酉制备表面码逻辑态方法，采用双向稳定子扩展CNOT级联，在保持故障距离的同时降低逻辑错误率，并低于阈值。
AI 中文摘要
制备逻辑态是任何基于表面码的量子计算开始时不断重复的基本操作。标准基于测量的协议在中性原子上代价高昂，因为测量比门操作慢几个数量级，并且在单区域架构中，还需要将原子穿梭到读出区域。在这项工作中，我们提出了一种针对中性原子量身定制的、无测量的表面码泡利本征态的酉制备方法。双向稳定子扩展CNOT级联从中间线向外以深度$(d{+}3)/2$增长补丁，同时在受保护的错误方向上保持故障距离$d$。每一层对应一次集体原子移动，随后进行一次全局里德伯脉冲，大约每两层拾取一行。我们使用开源的bloqade工具链，在硬件校准的电路级中性原子噪声模型下，编译并模拟了所得调度。双向级联在每个距离上都实现了最低的逻辑错误率，并且是所测试的构造中唯一在所考虑的设备级噪声下对两种逻辑态均低于阈值的构造。
英文摘要
Preparing logical states is a constantly recurring primitive at the start of any surface-code-based quantum computation. The standard measurement-based protocol is expensive on neutral atoms, since measurements are orders of magnitude slower than gates and, on single-zone architectures, additionally require shuttling the atoms to a readout zone. In this work, we present a measurement-free, unitary preparation of surface-code Pauli eigenstates tailored to neutral atoms. A bidirectional stabilizer-expanding CNOT cascade grows the patch from its middle line outward in depth $(d{+}3)/2$, while retaining fault distance $d$ in the protected error direction. Each layer maps to one collective atom move followed by a single global Rydberg pulse, with about one row pickup per two layers. We compile and simulate the resulting schedules with the open-source bloqade toolchain under a hardware-calibrated, circuit-level neutral-atom noise model. The bidirectional cascade achieves the lowest logical error rate at every distance, and is the only construction tested that operates below threshold for both logical states at the considered device-level noise.
Comments4 pages, 4 figures
