Theoretical Studies of Intermolecular Forces

在化学系化学理论、模型和计算方法(CTMC)计划的支持下，特拉华大学的Krzysztof Szalewicz教授正在对分子簇、分子凝聚相和生物分子系统进行量子力学研究。这类系统的性质受分子间力(范德华力)的支配：根据两个分子之间的距离，它们要么相互吸引，要么相互排斥，理查德·费曼认为这是人类最大的发现。Szalewicz和他的同事开发了计算分子间力的方法，这些方法不仅是最准确和计算效率最高的方法之一，而且还为研究人员提供了一种独特的能力，根据四种基本的物理机制来解释依赖于分子间力的性质：静电、交换排斥、极化和色散相互作用。由于直接的量子力学计算仅限于具有100个左右原子的分子组合，Szalewicz的团队将开发机器学习方法，将量子结果外推到任意大小的系统。这项工作的重要性源于它能够根据第一原理预测物质的性质，即从量子力学方程中推导出任意分子材料的性质。一个例子是对晶体结构的可靠预测。晶体的计算机设计对于制药、农用化学品、半导体和含能材料工业具有重要意义。这里正在进行的关于分子间力的基础研究有可能在从材料科学到生物分子，从大气科学到分子光谱学和天体物理学的广泛领域产生广泛的科学影响。这项活动的更广泛影响包括培训具有不同背景的研究生和博士后助理，与其他研究小组广泛合作，组织传播知识的会议和研讨会，参与颁奖委员会和国际学术界的活动，以及开发供其他研究人员使用的自由软件。在这一领域，Szalewicz小组正在开发和利用基于密度泛函理论(DFT)描述的单体的对称适应微扰理论(SAPT)，一种称为SAPT(DFT)的方法。用于从SAPT(DFT)计算中产生力场的机器学习方法将被扩展，以能够处理具有软内部自由度的分子。这些力场将用于当前实验、观测或技术感兴趣的几个系统，特别是用于根据第一性原理预测晶体结构，包括与单体不同构象有关的多态困难情况。理论的其他发展将包括能够可靠地预测色散能量的非局部相关DFT方法的工作，可以与准确的色散能量配对的改进的DFT方法，将机器学习的力场生成方法扩展到三体非加性相互作用，以及基于从头计算的单体性质的普适力场，这是朝着基于物理的生物分子力场的方向迈出的一步。这些研究有望促进分子间相互作用领域的知识，并有望在电子结构方法、力场开发技术和晶体结构预测方面发生革命性变化。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

With support from Chemical Theory, Models and Computational Methods (CTMC) program in the Division of Chemistry, Professor Krzysztof Szalewicz of University of Delaware is performing quantum-mechanical investigations of clusters of molecules, molecular condensed phases, and biomolecular systems. The properties of such systems are governed by intermolecular (van der Waals) forces: depending on the distance between two molecules, they will either attract or repel each other, a physical law that Richard Feynman considered to be the biggest finding of humanity. Szalewicz and coworkers have developed methods for computing intermolecular forces that are not only among the most accurate and computationally efficient ones, but also provide researchers with a unique ability to interpret properties dependent on intermolecular forces in terms of the four fundamental physical mechanisms: the electrostatic, exchange-repulsion, polarization, and dispersion interactions. Since direct quantum-mechanical calculations are limited to molecular assemblies with a hundred or so atoms, Szalewicz’s group will develop machine-learning methods of extrapolating quantum results to arbitrary-size systems. The importance of this work stems from its ability to predict properties of matter from first principles, i.e., deriving them from equations of quantum mechanics, for arbitrary molecular materials. One example are reliable predictions of crystal structures. Computational design of crystals is of significant importance for pharmaceutical, agrochemical, semiconductor, and energetic materials industries. The fundamental research on intermolecular forces being conducted here has the potential for broad scientific impact across a wide array of fields from materials science to biomolecular, and atmospheric science, to molecular spectroscopy and astrophysics. Broader impacts of this activity include training of graduate students and postdoctoral associates with diverse backgrounds, extensive collaborations with other research groups, organization of conferences and workshops that will disseminate knowledge, participation in award committees and in activities of an international academia, and developments of free software to be used by other researchers.Under this ward, the Szalewicz group is developing and utilizing symmetry-adapted perturbation theory (SAPT) based on monomers described by density-functional theory (DFT), an approach denoted as SAPT(DFT). Machine-learning methods for the generation of force fields derived from SAPT(DFT) calculations will be extended to enable treatment of molecules with soft internal degrees of freedom. These force fields will be used for several systems of current experimental, observational, or technological interest, in particular for predictions of crystal structures from first principles, including difficult cases with polymorphism related to varying conformations of monomers. Other developments of theory will include work on nonlocal-correlation DFT methods that can reliably predict dispersion energies, improved DFT methods that can be paired with accurate dispersion energies, extensions of machine-learning force-field generation methods to three-body nonadditive interactions, and universal force fields based on ab initio computed monomer properties, a step in the direction of physics-based biomolecular force fields. These studies are expected to advance knowledge of the field of intermolecular interactions and to have promise for transformative changes in electronic structure methods, in force-field development techniques, and in crystal structure predictions.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.