计算化学Computational chemistry

按精度和计算量排列：高精度从头算（CCSD(T)，适合几十个原子）→ 密度泛函理论（数百到上千个原子）→ 半经验方法 → 分子力学和分子动力学（百万原子）→ 粗粒化模型和连续介质模型。QM/MM 多尺度方法把不同层次结合起来。化学信息学处理化合物数据库和定量构效关系（QSAR），机器学习用于性质预测、逆合成设计和势能面拟合。

In order of accuracy and cost: high-level ab initio methods (CCSD(T), tens of atoms) → density functional theory (hundreds to thousands of atoms) → semi-empirical methods → molecular mechanics and dynamics (millions of atoms) → coarse-grained and continuum models. Multiscale QM/MM methods combine levels. Cheminformatics handles compound databases and quantitative structure–activity relationships (QSAR), and machine learning predicts properties, plans retrosyntheses and fits potential energy surfaces.

1950 年代电子计算机出现后，科学家开始用其求解量子化学方程。1970 年波普尔发布 Gaussian 70 程序，使从头算方法普及。1998 年科恩和波普尔、2013 年卡普拉斯、莱维特和沃谢尔、2024 年贝克（计算蛋白质设计）以及哈萨比斯和詹珀（AlphaFold）先后因计算化学相关工作获诺贝尔化学奖。

Once electronic computers appeared in the 1950s, scientists began using them to solve quantum-chemical equations. Pople's Gaussian 70 program of 1970 popularised ab initio methods. Nobel Prizes in Chemistry for computational work went to Kohn and Pople (1998), Karplus, Levitt and Warshel (2013), and Baker for computational protein design together with Hassabis and Jumper for AlphaFold (2024).

计算化学用于药物设计（虚拟筛选、分子对接、结合自由能计算）、催化剂设计（计算反应能垒和火山曲线）、材料筛选（电池正极、光伏、金属有机框架的高通量计算）、光谱预测与归属、大气和星际化学反应速率估算，以及解释实验难以观测的过渡态和短寿命中间体。

Computational chemistry is used in drug design (virtual screening, docking, binding free energies), catalyst design (reaction barriers and volcano plots), materials screening (high-throughput calculation of cathodes, photovoltaics and metal–organic frameworks), predicting and assigning spectra, estimating atmospheric and interstellar reaction rates, and explaining transition states and short-lived intermediates that experiments struggle to observe.