几何校准的平衡感知用于非局域拓扑光子晶格的逆设计
Geometry-calibrated equilibrium sensing for inverse design of nonlocal topological photonic lattices
AI总结：
该研究针对非局域拓扑光子晶格逆设计，提出几何校准的量子平衡传播框架，训练两量子位平衡传感器并结合物理门控评分，实现突破近邻规则的拓扑设计，验证了特定等离激元几何的优异响应。
AI中文摘要：
拓扑光子晶格通常采用短程哈密顿量设计，但实际纳米光子结构受几何和波长依赖的长程电磁相互作用支配。本文引入一种几何校准的量子平衡传播框架，用于有限非局域等离激元Su-Schrieffer-Heeger（SSH）晶格的推理与逆设计。在有效耦合空间中训练一个两量子位平衡传感器，以区分边界局域响应与平凡有限晶格响应。由于在非局域跃迁存在时，从近邻SSH模型继承的标签变得不可靠，因此利用非局域绕数和有限带隙准则对每个样本进行独立重新标记。在该经物理验证的评估集上，传感器达到99.8%的灵敏度和98.1%的特异性。随后，物理门控鲁棒性评分根据边界响应、响应对比度、带隙稳定性以及与所选非局域 regime 的兼容性对验证构型进行排序。孤立和成对金纳米颗粒的全波消光光谱建立了几何到耦合的校准，将颗粒尺寸和间距与波长分辨的对耦合关联起来。将该校准投影到已验证的耦合态势上，确定了一种有限等离激元几何，其在546 nm和642 nm处分别支持光谱上不同的角主导和边主导响应，扇区与体的强度对比度分别为2.4×10⁴和1.5×10³。该框架将实际电磁几何与非局域拓扑设计关联起来，突破了近邻设计规则，实现了经物理验证、几何分辨的推理。
英文摘要：
Topological photonic lattices are commonly designed using short-range Hamiltonians, yet realistic nanophotonic structures are governed by geometry- and wavelength-dependent long-range electromagnetic interactions. Here we introduce a geometry-calibrated quantum equilibrium-propagation framework for inference and inverse design in finite nonlocal plasmonic Su-Schrieffer-Heeger lattices. A two-qubit equilibrium sensor is trained in effective-coupling space to distinguish boundary-localized from trivial finite-lattice responses. Because labels inherited from the nearest-neighbor SSH model become unreliable in the presence of nonlocal hopping, each sample is independently relabeled using nonlocal winding numbers and a finite-gap criterion. On this physics-verified evaluation set, the sensor achieves 99.8% sensitivity and 98.1% specificity. A physics-gated robustness score then ranks verified configurations by boundary response, response contrast, gap stability and compatibility with the selected nonlocality regime. Full-wave extinction spectra of isolated and paired gold nanoparticles establish a geometry-to-coupling calibration linking particle size and separation to wavelength-resolved pair couplings. Projecting this calibration onto the verified coupling landscape identifies a finite plasmonic geometry supporting spectrally distinct corner- and edge-dominated responses at 546 and 642 nm, with sector-to-bulk intensity contrasts of $2.4 \times 10^{4}$ and $1.5 \times 10^{3}$, respectively. The framework links realistic electromagnetic geometry to nonlocal topological design, moving beyond nearest-neighbor design rules with physics-verified, geometry-resolved inference.
