The Valence Multipole Model: Linking Structure and Reactivity

更广泛的含义。了解矿物和其他固体表面上的化学过程(分子和原子发生了什么)是支持矿物在整个自然界中的行为的科学的关键部分。这类表面的科学对涉及固体表面的重要工业过程也有影响，例如用于塑料制造的催化剂。科学家们发现，为了在分子尺度上理解表面上发生的事情，对这些过程进行计算建模是绝对必要的。这项工作的重点是推进矿物表面的建模，这对地球化学和化学都很重要。目前，用于模拟原子与分子相互作用的理论和模型在结果质量和可以解决的问题种类上都表现出很大的范围。非常需要数学上足够简单，可以应用于大型原子系统的模型，但要有更高的准确性和可转移性。研究小组将致力于创建这样一个模型，通过扩展结晶学家常用的简单成键模型-键价模型，以一种易于应用于原子模拟的方式。该奖项的另一个组成部分是建设一支未来的美国STEM劳动力队伍，他们不仅在地球科学方面有能力，而且在计算机编程、建模和数据分析方面也有能力。在地球科学(以及STEM的所有领域)，无论是在工业界还是学术界，舒适地使用计算的能力正在变得越来越重要。传统上，地球科学学位课程不要求学生掌握重要的计算机编程和数据分析技能。该项目将要求本科生研究人员使用MatLab科学编程环境进行数据挖掘、探索性数据分析和优化。这些技能的培训将由首席研究人员提供给学生。这个项目对地球化学以外的科学有影响，因此，由环境化学科学计划(NSF化学部)和跨学科的NSF计算和数据启用的科学和工程计划联合资助。技术描述。在这个项目中，首席研究人员Barry Bickmore和他的团队将继续工作，创建一个简化的化学键模型，基于键价模型(BVM)，可以很容易地转移到用于原子模拟的分子力学框架。该模型中的势能项主要基于每个原子的键价的多极(单极、偶极和四极)展开。这些术语往往出人意料地可预测，并描述了分子结构的所有主要方面，包括电子结构效应造成的扭曲(Lone-Pair和Jahn-Teller)。它们也是本质上的多体术语，描述了关于每个原子的总成键环境的各个方面，而不是关注单个原子对。这允许使用少量参数对非常复杂的交互进行建模。目前的工作集中在开发键合模型来预测基于BVM的结构描述符的理想值，以及偏离理想值的能量成本函数。此外，我们正在开发一些初步的分子力学力场，以进一步测试这一概念。许多本科生研究人员将参与该项目，并将学习计算机编程、数据分析和优化技术。

Broader implications.Understanding chemical processes (what happens to molecules and atoms) on mineral and other solid surfaces is a critical part of the science underlying how minerals behave throughout nature. The science of such surfaces also has implications for industrially important processes involving solid surfaces, such as the catalysis used for plastics manufacturing. Scientists have found that in order to understand what happens on surfaces at the molecular scale, computational modeling of these processes is absolutely essential. This work focuses on advancing modeling of mineral surfaces, which is important for both geochemistry and chemistry. Currently, the theories and models used to simulate interactions between atoms and molecules exhibit a large range in both the quality of results and the kinds of problems that can be addressed. There is a great need for models that are mathematically simple enough to be applied to large systems of atoms, but that are capable of greater accuracy and transferability. The research team will work on creating such a model by expanding a simple bonding model commonly used by crystallographers, the Bond-Valence Model, in a manner that is easily applied to atomistic simulations. An additional component of this award entails building a future US STEM workforce that is capable not only in geoscience, but also in computer programming, modeling, and data analysis. In geoscience (as well as all fields of STEM), both in industry and academia, the ability to comfortably use computing is growing in its importance. Traditionally, geoscience degree programs do not require students to acquire significant computer programming and data analysis skills. This project will require undergraduate researchers to do data mining, exploratory data analysis, and optimization, using the MATLAB scientific programming environment. Training in these skills will be provided to the students by the principal investigator.This project has implications for science beyond geochemistry, and, as such, is being jointly funded by the Environmental Chemical Sciences program (NSF Division of Chemistry) and the cross-disciplinary NSF Computational and Data-Enabled Science and Engineering program.Technical description.In this project, principal investigator Barry Bickmore and his group will continue work to create a simplified chemical bonding model, based on the bond-valence model (BVM), that can be easily transferred to a molecular mechanics framework for atomistic simulations. The potential energy terms in this model are primarily based on multipole (monopole, dipole, and quadrupole) expansions of bond valences about each atom. These terms tend to be surprisingly predictable, and describe all major aspects of molecular structure, including distortions due to electronic structure effects (lone-pair and Jahn-Teller). They are also intrinsically multi-body terms that describe aspects of the total bonding environment about each atom, rather than focusing on individual atom pairs. This allows for very complex interactions to be modeled using a small number of parameters. The present work focuses on developing the bonding model to predict ideal values of the BVM-based structural descriptors, as well as energy cost functions for deviations from the ideal values. In addition, we are developing some preliminary molecular mechanics force fields to further test the concept. A number of undergraduate researchers will be involved in the project, and will learn computer programming, data analysis, and optimization techniques.