2026-10-26 · arXiv Quantum Physics
arXiv:2610.32287v1 Announce Type: new
Abstract: Self-assembled two-dimensional molecular aggregates provide a versatile platform for engineering excitonic states, including topologically nontrivial edge states. However, these edge-states are difficult to probe optically: they have weak far-field transition strengths, and far-field spectra do not reveal whether an excitation is localized at the edge. For an aggregate consisting of two sublattices of molecules with genuinely complex transition dipoles, we show that spatially resolved near-field spectroscopy encodes the full spatial structure of these states. The resulting near-field maps do not mirror the site populations, but instead exhibit strongly nonuniform, interference-like features. We demonstrate that a convolutional neural network can reconstruct the complex expansion coefficients of the eigenstates from these maps, and that this reconstruction is robust against substantial random noise.
原文链接：https://arxiv.org/abs/2610.32287
查看更多资讯请见 AI 行业资讯。