固体表面自旋交叉分子的高通量计算筛选与发现

自旋交叉分子可作为磁开关应用于自旋电子器件中。然而，从分子到器件往往涉及分子在固体表面的负载问题。分子的磁学特性会因此发生一定的变化，甚至完全消失。因此，迫切需要在单分子层次上明确吸附对自旋自由度的影响，以找到合适的基底。数据驱动的计算材料发现适用于这类问题，但是尚缺乏有效的材料描述符。申请人发现自旋态分裂能可作为衡量固体表面分子自旋交叉活性的关键参数。为了准确计算这个量，需要同时处理好非共价相互作用和强关联效应，且引入依赖于化学环境的局域关联效应至关重要。本项目拟基于这些理论发现，以稳定吸附分子的自旋态分裂能为描述符，对固定在常见金属、半导体表面，以及二维材料上的Fe(II)配合物开展高通量计算筛选，以寻找可用的自旋交叉体系。同时，本研究将建立表面分子材料数据库，以结合机器学习技术挖掘结构与磁性质之间的深层次关系，从而加快新材料的发现。本研究有望设计出可供实验进一步探索的表面分子磁体。

Spin-crossover molecules could act as magnetic switches with applications in spintronics. Fabricating devices from molecules remains, however, a challenging task since it requires the deposition of molecules on solid state surfaces. Due to molecule-substrate interactions, the molecular spin-crossover properties can be drastically tailored or even completely lost in some cases. Thus, it is essential to understand how deposition dictates the spin degrees of freedom at single-molecule level, in order to find substrates that preserve or enhance spin-crossover behaviors of molecular adsorbates. Data-driven computational materials discovery appears to be a promising tool to tackle this problem. However, it is still difficult to perform this task due to the lack of an appropriate descriptor of the magnetic properties of molecule-substrate systems. Recently we have shown that the spin-state splitting can well characterize spin-crossover behaviors at surface. To evaluate this physical parameter accurately, it is critical to incorporate strong correlation effects and dispersion interactions within density-functional theory. Further, the prediction is improved considerably by introducing the local Coulomb correlation that depends on the chemical environment. Based on these findings, we will perform high throughput computational screening of Fe(II) complexes deposited on metallic and semiconductor surfaces as well as two-dimensional materials to search for appropriate spin-crossover systems, using the spin-state splitting of stable molecular adsorbates as a simple descriptor. Meanwhile, we will create a computational molecule-surface materials database, which could be used by machine learning techniques to identify deep correlations between structures and magnetic properties, and thus should be useful for accelerating the chemical discovery of new materials. This study would provide a set of promising surface-supported molecular magnets for experimental investigation.