Correlated first-principles methods for applications in exploration-supported synthesis planning

在这个建议中，第一性原理量子化学计算将被利用来探测所提出的化学反应级联的可行性。特别是当合成新的化合物时，通常会对通向该化合物的多条路径进行评估。这些路径通常是从最终目标开始产生的，然后分解成更小的所谓合成子，这个过程被称为反合成。在开始合成该化合物的实验工作之前，将对这些反合成建议进行计算评估。允许这种验证的灵活算法的关键要求是一个自动基准测试系统，该系统允许创建的程序在内部验证其计算。对于一般未知的反应路径，必须有一种鲁棒的电子结构方法，以便在探索反应空间时精确计算遇到的任何系统。在这里，我们建议开发一种计算上可行的多配置方法来实现准确的内部基准测试。我们计划结合密度泛函理论（DFT）对短程电子-电子相互作用的描述，以及密度矩阵重整化群（DMRG）算法优化的完全有效空间波函数对远程电子-电子相互作用的描述。特别是，我们提出了一种无自旋污染的方法，用于开壳电子结构，如果用无限制的Kohn-Sham理论处理，已知会受到自旋污染的困扰。最终的实现将集成到软件中，该软件允许自动探索反应空间，以验证反合成建议。为此，现有的勘探软件将得到扩展，使其沿着预定的路径进行勘探，并评估可能的副反应。最后将在已知药物化合物的合成上进行实施测试，以便将来应用于未知化合物的合成。

In this proposal, first-principles quantum chemical calculations shall be exploited to probe the viability of proposed chemical reaction cascades. Specifically when new chemical compounds are synthesized, there is usually an evaluation of multiple paths leading to the compound.The paths are often generated starting from the final target, then breaking into smaller so-called synthons, this process is called retrosynthesis.These retrosynthetic suggestions will be evaluated computationally before experimental efforts to synthesize the compund are started.The key requirement for a flexible algorithm that allows for such a validation is an automated benchmarking system that allows the created program to validate its calculations internally.For a generally unknown reaction path it is imperative to have a robust electronic structure method that allows for the accurate calculation of any system encountered during the exploration of the reaction space.Here, we propose to develop a computationally feasible multi-configuration approach for accurate internal benchmarking.We plan to combine the description of short-range electron-electron interactions by density functional theory (DFT), and the description of long-range electron-electron interactions by a complete-active-space wave function optimized by the density matrix renormalization group (DMRG) algorithm.In particular, we propose a spin-contamination-free approach for open-shell electronic structures, which are known to be plagued by spin contamination if treated by unrestricted Kohn-Sham theory. The resulting implementation will be integrated into software that allows for the automated exploration of reaction space in order to validate retrosynthetic suggestions.To this end, the existing exploration software will be extended to drive its exploration along a predetermined path and evaluate possible side reactions.Finally the implementation will be tested on the syntheses of known pharmaceutical compounds, in order to be applied to the syntheses of unkown compunds in the future.