Computational methods for the exploration of potential energy surfaces with applications to nanoalloy materials.

我们开发了高性能的计算方法来探索势能面(PES，描述分子的能量如何随着分子的几何结构变化而变化的数学函数)，从而为研究分子结构和化学反应活性提供了基础。这一提议的成功实施将创造一套独特的计算工具，用于研究分子结构、过渡态和反应路径，并用于模拟具有数百个原子的系统。这些工具将对加拿大和其他地区的研究人员非常重要。我们的一种方法模拟了DFT，计算速度提高了4到6个数量级。它将使纳米合金的研究在催化和数据存储设备中具有潜在的应用。以我们最近发现的双金属笼子为线索，我们将寻找新的、更大的全金属笼子。PES的极小值是分子的平衡几何构型；鞍点是过渡态；连接鞍点和极小值的最陡下降线是反应路径；最陡下降线附近的低能区域对于详细的反应机理模型很重要。我们将继续利用量子化学和密度泛函理论(DFT)的专家知识，结合我们自己的计算机程序来研究PES。学生将接受将方法开发和应用相结合的研究项目的培训，并将获得各种技能：Unix和量子化学软件、编程和外壳脚本编写、数学建模和一般科学素养方面的专业知识。这些项目是在不同层面上设计的，以满足本科生、研究生和博士后的需求和能力。我们将从三个方面研究纳米合金，这些合金具有有趣的性质，并在催化和信息技术中具有潜在的用途。(1)全局优化。进化计算方法，如遗传算法(GA)，将被开发来确定纳米合金团簇的几何结构。(2)反应机理。我们使用粒子群优化算法(PSO)在没有人类监督的情况下发现了化学反应过程中分子中发生的几何变换序列(反应机理)。我们将创造一种新的方法，基于化学直觉和优化技术，寻找先验(不需要任何DFT计算)反应机理。我们将在个案的基础上，找出这个机制是否很好地近似于最低能量的反应路径(真正的机制)，并将相应地对化学反应进行分类。(3)原子模拟。我们开发了一种新的拟合方法，用于创建高维PES，它是完全自动化的，不需要用户输入函数形式，并且几乎适用于任何化学成分。此PES函数模拟精确方法(DFT)并提供相对于DFT的4到6个数量级的加速。我们将使用它来进行精确的蒙特卡罗(AIMC)模拟，其中PES与“重要性抽样”一起使用，以提供极大的计算速度。我们还将采取双管齐下的方法进行全局优化，其中我们使用PES函数来筛选大量的几何结构，并生成一个更小的子集，为其计算DFT能量。在这三个方面的工作中，我们将试图揭示支配双金属团簇的相对稳定性和几何结构以及磁性、电子和化学性质的原理。基于这些原理和我们团队独有的新方法，我们将设计稳定的纳米合金，可能应用于催化和磁数据存储设备。

We develop high performance computational methods to explore potential energy surfaces (PES, mathematical functions that describe how the energy of molecules changes when the geometry of the molecule changes), thus providing the foundation for studies of molecular structure and chemical reactivity. Successful implementation of this proposal will create a unique set of computational tools for studying molecular structures, transition states, and reaction paths, and for the simulation of systems with hundreds of atoms. These tools will be of great importance to researchers in Canada and beyond. One of our methods mimics DFT and gives a computational speed-up of 4 to 6 orders of magnitude. It will allow the study of nanoalloys with potential applications in catalysis and data storage devices. Taking clues from our recent discovery of bimetallic cages, we will search for new and bigger all-metal cages. Minima of the PES are equilibrium geometries of molecules; saddle points are transition states; steepest descent lines connecting saddle points to minima are reaction paths; low energy regions in the vicinity of steepest-descent lines are important for detailed models of reaction mechanisms. We will continue to make expert use of quantum chemistry and density functional theory (DFT), combined with our own computer codes, to investigate PES. Students will get trained in research projects that combine method development and applications and will acquire a variety of skills: expertise in UNIX and quantum chemistry software, programming and shell scripting, mathematical modeling, and general scientific literacy. The projects are designed at various levels to address the needs and capabilities of undergraduate, postgraduate, and postdoctoral students. We will work on three fronts to study nanoalloys that have interesting properties and potential use in catalysis and information technologies. (1) Global optimization. Evolutionary computing methods, like Genetic Algorithms (GA), will be developed to determine the geometric structure of nanoalloy clusters. (2) Reaction mechanisms. We have used Particle Swarm Optimization (PSO) to discover, without human supervision, the sequence of geometric transformations that occur in molecules during chemical reactions (``reaction mechanisms''). We will create a new method, based on chemical intuition and optimization techniques, to find a priori (without any DFT calculation) reaction mechanisms. We will find, on a case-by-case basis, whether this mechanism is a good approximation to the lowest-energy reaction path (the true mechanism), and will classify chemical reactions accordingly. (3) Atomistic simulations. We developed a new fitting method for creating high-dimensional PES that is fully automated, does not require the user to input a function form, and which works for practically any chemical composition. This PES function mimics an accurate method (DFT) and offers a speed-up of 4 to 6 orders of magnitude relative to DFT. We will use it to carry out accurate Monte Carlo (AIMC) simulations where the PES is used with "importance sampling" to give great computational speed-up. We will also take a two-pronged approach to global optimization where we use the PES function for screening a large number of geometric structures and generate a much smaller subset for which the DFT energy gets calculated. Working on those three fronts we will try to uncover the principles that govern the relative stability and geometric structure, and the magnetic, electronic and chemical properties of bimetallic clusters. Based on these principles, and novel methods unique to our group, we will design stable nanoalloys for possible applications in catalysis and magnetic data storage devices.