ReaxFF MD模拟结果的化学反应网络自动构建及可视化
文献类型:学位论文
作者 | 贺巧鑫 |
答辩日期 | 2019-07-01 |
文献子类 | 硕士 |
授予单位 | 中国科学院大学 |
导师 | 李晓霞 |
关键词 | 反应分子动力学,Varxmd,化学反应机理,反应路径,化学反应网络 |
学位专业 | 化学工程 |
英文摘要 | 反应分子动力学(ReaxFF MD)是一种极具潜力可应用于模拟较大规模分子体系(>10,000个原子)化学反应的新方法。大规模ReaxFF MD模拟得到的原子轨迹和键级信息中蕴含着复杂的化学反应信息,人工分析机理并不现实。作者课题组开发的国际上首个反应分子动力学分析与可视化工具VARxMD,可从模拟结果中自动获取物种信息和详细的化学反应列表,并实现了基于子结构等特征对反应物种及其参与的反应进行自动分类。本工作在此基础上,进一步扩展了对指定起始反应物与目标产物之间的所有反应路径进行搜索、进而自动构建化学反应网络的功能。论文主要包括以下内容:(1) 针对ReaxFF MD提出基于反应列表和基于原子自动识别反应路径的两种策略。基于反应列表的构建策略需基于完整的反应列表且搜索目标的设定单一。相较之下,基于原子的反应路径搜索和反应网络构建策略可支持多对象搜索、且搜索得到的反应网络更加完整。(2) 设计并实现了三种搜索策略,分别为仅指定起始反应物向前追溯、仅指定目标产物向后回溯和同时指定起始反应物和目标产物搜索两者之间的所有反应路径,并设计了支持指定原子、化学键、分子和物种等多对象作为反应起始物和目标产物的基于原子进行反应路径搜索的算法,有利于研究者从多角度探索反应机理。反应路径搜索算法的实现基于宽度优先搜索策略,可根据指定的搜索目标,搜索得到目标在特定时间范围内的反应路径,并可进一步分析反应路径总结出有规律的反应机理。(3)选用合适的有向图布局算法Sugiyama算法并基于Qt的视图框架实现了化学反应网络的可视化,在此基础上分别实现了基于物种合并的化学反应网络和未经合并的含时反应路径图的可视化显示。基于Qt的信号与槽机制实现了网络图的交互功能,有利于获得化学反应网络中蕴含的丰富的物种与化学反应信息。本论文在VARxMD已有的功能基础上扩展其反应路径自动追溯、构建化学反应网络的功能。随着方法的逐步完善,本文扩展的化学反应网络构建功能有望被ReaxFF MD模拟的研究人员广泛使用。;Reactive molecular dynamics (ReaxFF MD) is a promising new method that can be applied to study complex chemical reactions in large scale reactive molecular systems (>10,000 atoms). There are huge amount of underlying chemical reactions in the trajectory and bond order files obtained in ReaxFF MD simulations, from which is not practical to discover reactions manually. VARxMD is a reaction analysis and visualization tool for ReaxFF MD developed by the author’s research group. VARxMD can automatically obtain unique species, detailed chemical reaction lists with detailed reaction sites from ReaxFF MD simulation results. In addition, reactions can be classified into different groups based on the characteristics of substructures and reaction sites of specified species. On the basis of the analysis data of VARxMD, this work extends the functions of reaction pathway search and construction of the chemical reaction network between specified reactant and product in the reaction list.Two strategies based on reaction lists and based on atoms for automatic identification of reaction paths are proposed. The advantages and disadvantages of these two strategies are further analyzed. The strategy based on the reaction lists needs to be performed based on complete reaction lists and with the limitation of single set of search targets designation. In contrast, the strategy based on atoms supports multi-target search and the detailed network with time stamp can be obtained. An atom-based reaction path search algorithm is designed and implemented that supports searching targets of atoms, chemical bonds, molecules and species, which is beneficial to explore the reaction mechanism from different perspectives. The reaction search algorithm is based on breadth-first search(BFS). The results and reaction pathways of reaction search during specified period between the range of two timesteps can be further analyzed for understanding of reaction mechanism.The visualization of obtained chemical reaction network is implemented with Sugiyama algorithm for directed graph layout and Qt-based view framework. The visualization of the chemical reaction network supports both the reduced reaction pathway network without time stamps or detailed reaction pathway with time stamps. The Qt-based signal and slot mechanism are employed to realize the interactive depiction of the reaction network diagram, which is beneficial to obtain the rich information of reactive species and chemical reaction sites for the contained in the chemical reaction network.Based on the existing functions of VARxMD, this work expands the function of automatic retrospect of reaction pathways and construction of chemical reaction networks. The extended functions for construction and visualization of chemical reaction network in this work is general and can be widely applied in ReaxFF MD simulations. |
语种 | 中文 |
源URL | [http://ir.ipe.ac.cn/handle/122111/40710] ![]() |
专题 | 中国科学院过程工程研究所 |
推荐引用方式 GB/T 7714 | 贺巧鑫. ReaxFF MD模拟结果的化学反应网络自动构建及可视化[D]. 中国科学院大学. 2019. |
入库方式: OAI收割
来源:过程工程研究所
其他版本
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。