Structure Prediction and Design of Molecular Crystals with the GAtor Genetic Algorithm

非技术总结该奖项支持旨在开发和应用计算方法来预测分子晶体晶体结构的研究和教育活动。分子晶体是由分子结构单元组成的固体。它们被广泛用于各种需要将分子成分包装成具有均匀、可再现特性的结构的应用中，包括药物、高能材料和有机电子产品。分子固体的性质和性能与其晶体结构密不可分。由于分子晶体通过相对较弱的分子间相互作用结合在一起，而不是将原子结合在一起形成分子的强化学键，因此同一分子可以结晶成几种不同的晶体结构，称为多晶型物。相同分子的多晶型物可能具有显著不同的物理和化学性质。通过计算机模拟预测给定分子所有可能的多晶型及其特性的能力对于产品以分子晶体形式销售的行业至关重要。在这个研究项目中，PI和她的团队将开发用于预测分子晶体结构和设计具有改进性能的晶体结构的算法。该奖项还支持PI的教育和推广活动，包括对研究生和本科生进行高性能计算和机器学习的培训，新课程开发，以及组织“科学和工程领域的女性领导人”午餐会，让女学生与女性研讨会发言人会面，为他们提供导师和网络机会，以帮助他们留在科学和工程，并促进他们的职业生涯。该奖项支持旨在开发和应用计算方法来预测分子晶体晶体结构的研究和教育活动。分子晶体由弱色散相互作用束缚，这些相互作用产生具有许多局部极小值的势能景观，这些极小值在能量上可能非常接近。这就产生了多晶型现象，即同一分子以不同结构结晶。晶体结构可以深刻地影响物理和化学性质，因此影响分子固体在各种应用中的功能，包括药物，高能材料和有机电子。因此，预测分子晶体的结构和性质的能力是至关重要的。在这个研究项目中，PI和她的团队将开发用于预测分子晶体结构和改进性能的晶体结构逆向设计的算法。具体而言，多组分晶体和柔性分子晶体将成为目标。这将通过开发配置空间探索的新方法来实现，在无缝集成的工作流程中结合第一原理模拟，优化算法和机器学习。这些将以开放源代码实施，并行化，旨在在高性能计算机上高效执行。晶体结构预测数据集将作为其他研究人员的资源公开提供。该奖项还支持PI的教育和推广活动，包括对研究生和本科生进行高性能计算和机器学习的培训，新课程开发，以及组织“科学和工程领域的女性领导人”午餐会，让女学生与女性研讨会发言人会面，为他们提供指导和网络机会，以帮助他们留在科学和工程，并促进他们的职业生涯。这个奖项反映了NSF的法定使命，并已被认为是值得的通过使用基金会的知识价值和更广泛的影响审查标准进行评估来提供支持。

NONTECHNICAL SUMMARYThis award supports research and educational activities aimed at developing and applying computational methods to predict crystal structures of molecular crystals. Molecular crystals are solids comprised of molecular building blocks. They are used extensively in a variety of applications that require packing molecular ingredients into structures with uniform, reproducible properties, including pharmaceuticals, energetic materials, and organic electronics. The properties and performance of molecular solids are inextricably linked to their crystal structure. Because molecular crystals are held together by relatively weak inter-molecular interactions, as opposed to the strong chemical bonds that hold atoms together to form molecules, the same molecule may crystallize in several different crystal structures, known as polymorphs. Polymorphs of the same molecule may have markedly different physical and chemical properties. The ability to predict all the possible polymorphs of a given molecule and their properties by computer simulations is of paramount importance across industries whose products are marketed in the form of molecular crystals. In this research project, the PI and her team will develop algorithms for prediction of molecular crystal structures and for design of crystal structures with improved properties. This award also supports the PI's educational and outreach activities, which include training of graduate and undergraduate students in high performance computing and machine learning, new curriculum development, and organization of "Women Leaders in Science and Engineering" luncheon meetings where female students meet with women seminar speakers, providing them with mentorship and networking opportunities to help retain them in science and engineering and advance their careers. TECHNICAL SUMMARYThis award supports research and educational activities aimed at developing and applying computational methods to predict crystal structures of molecular crystals. Molecular crystals are bound by weak dispersion interactions that generate potential energy landscapes with many local minima that may be extremely close in energy. This gives rise to polymorphism, the crystallization of the same molecule in different structures. Crystal structure may profoundly influence the physical and chemical properties, and hence the functionality of molecular solids in diverse applications, including pharmaceuticals, energetic materials, and organic electronics. Therefore, the ability to predict the structure and properties of molecular crystals is of paramount importance. In this research project, the PI and her team will develop algorithms for prediction of molecular crystal structures and for inverse design of crystal structures with improved properties. Specifically, multi-component crystals and crystals of flexible molecules will be targeted. This will be achieved through the development of new methods for configuration space exploration, combining first-principles simulations, optimization algorithms, and machine learning in seamlessly integrated workflows. These will be implemented in open source codes, parallelized and designed for efficient execution on high-performance computers. Crystal structure prediction datasets will be made publicly available as a resource for other researchers. This award also supports the PI's educational and outreach activities, which include training of graduate and undergraduate students in high performance computing and machine learning, new curriculum development, and organization of "Women Leaders in Science and Engineering" luncheon meetings where female students meet with women seminar speakers, providing them with mentorship and networking opportunities to help retain them in science and engineering and advance their careers.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.