表面（针尖）增强拉曼光谱的多尺度理论模拟与机理研究

Surface enhanced Raman spectroscopy (SERS) and tip-enhanced Raman spectroscopy (TERS) are widely used for the identification of the interfacial structures because of their ability of providing the `fingerprint`. To solve the difficulties in the SERS and TERS simulation under periodic boundary condition (PBC) and high local plasmonic field conditions, here we develop multiscale theoretical methods. Firstly, we use combined classic electromagnetic field theory and high accuracy quantum chemistry calculations to study the effect of the locality of the plasmonic field on SERS and TERS. Secondly, we develop an approach to accurately calculate the SERS and TERS spectra based on the PBC model. Based on this approach, we study the effect of the tip and substrate on the molecular geometrical and electronic structures as well as the Raman spectra. Thirdly, we employ PBC and cluster model together to develop a hybrid approach, where the Raman enhancement caused by the charge transfer state and the resonant state are investigated. Finally, we take the electromagnetic and chemical enhancements into account to obtain more accurate Raman spectra, building a reliable relationship between interfacial structures and the Raman spectra. The developed approach may provide theoretical basis to understand the Raman enhancements and help us to easily identify the interfacial structures.

表面增强拉曼光谱（SERS）与针尖增强拉曼光谱（TERS）能够提供界面的“指纹信息”，因此被广泛地应用在界面构型的识别中。针对目前拉曼光谱在周期性边界条件和局域场条件下模拟的难点，本项目拟发展多尺度理论方法，用以计算模拟SERS与TERS光谱。首先，采用经典电磁场模拟与高精度量子化学计算相结合的技术路线，系统地研究非均匀电磁场对SERS与TERS光谱的调制作用。其次，发展并完善周期性边界条件模型下SERS与TERS的理论模拟方法，并在此基础上，研究探针与衬底对分子的几何结构与电子结构的影响，以及由此导致的拉曼光谱的变化。再次，结合周期性边界条件模型与团簇模型，发展多尺度杂化的方法，用以探究电荷转移态与分子共振态对拉曼光谱的增强作用。最终，结合电磁场增强与化学增强，得到更加精确的拉曼光谱，建立全面、准确的“界面结构－光谱”关系，为识别界面构型，研究界面性质提供坚实的理论依据。

表面增强拉曼光谱（SERS）与针尖增强拉曼光谱（TERS）在实际中的应用，高度依赖于理论模拟与实验观测的紧密结合。针对目前的理论模拟难点，本项目开发了SERS与TERS光谱的新算法、新程序，用以提高计算精度和模拟速度。具体的，在计算精度方面，本项目发展并完善了周期性边界条件模型下拉曼光谱的理论模拟方法，并将范德华矫正包含在光谱的计算模拟中。在引入周期性边界条件的同时巧妙地节约了计算量，实现了大尺寸金属表面上的分子拉曼强度的准确计算。在计算速度方面，本项目开发了SERS光谱的机器学习(ML)算法。以反式-1,2-双（4-吡啶基）乙烯（BPE）为例，利用随机森林算法，在表界面结构与SERS信号之间建立ML模型，同时进行重要性分析。相比较传统量子化学算法，该ML模型的计算速度提升4个数量级。本项目发展的新算法，将与传统的量化软件相结合，为探测界面微观结构、理解界面光电作用、揭示界面构效关系提供知识基础和软件工具。