Structure searching methods: general discussion
Structure searching methods: general discussion
复制标题
结构搜索方法：一般讨论
DOI：
10.1039/c8fd90030b
复制
发表时间：
2018
影响因子：
3.4
通讯作者：
Cooper, Andrew
中科院分区：
文献类型：
--
作者：
Addicoat, Matthew;Adjiman, Claire S.;Arhangelskis, Mihails;Beran, Gregory J.;Brandenburg, Jan Gerit;Braun, Doris E.;Burger, Virginia;Burow, Asbjoern;Collins, Christopher;Cooper, Andrew
格雷姆·M戴开始讨论萨拉L. Price：您已经很好地概述了晶体结构预测代码的愿景及其功能。在这篇文章中，您还没有详细说明这段代码应该使用什么分析方法来理解和询问生成的结构集。晶体结构预测界，至少在分子有机领域，主要是从实验晶体结构的分析中借用方法。这些可能是有用的，但我们需要更复杂的方法来从计算的结构集合中获得最多的信息吗？我很想听听你对什么样的分析方法需要成为一般晶体结构预测代码的一部分的想法。Price回答：这是一个困难的问题，因为当与实验结构一起工作时，两种结构如何不同才能算作多晶型物仍然是一个争论的问题，其中许多困难与多型或无序有关，CSP分析需要进一步发展。大多数用于比较实验结构的方法适用于CSP生成的结构，假设已经消除了仅在单元选择上不同的重复。我们当然需要能够将结构聚集成家族，以便同一家族的成员可以作为实验结构的无序组分被观察到，或者被认为是如此容易地能够沿着成核和生长途径在沿着的某个地方转变成家族中最稳定的结构，以至于它们不能被观察到，或者被紧密地观察到。
Graeme M. Day opened the discussion of the introductory lecture by Sarah L. Price: You have nicely outlined a vision for a crystal structure prediction code and what its capabilities should be. In this, you have not said much about what analysis methods this code should have for understanding and interrogating the structure sets that are generated. The crystal structure prediction community, at least in the molecular organic eld, has mainly borrowed methods from the analysis of experimental crystal structures. These can be useful, but do we need more sophisticated approaches to get the most information from computed sets of structures? I am interested in hearing your thoughts on what analysis methods need to be part of a general crystal structure prediction code.Sarah L. Price answered: This is a difficult question, as it is still a matter of debate how different two structures have to be to count as polymorphs when working with experimental structures, with many of the difficulties relating to polytypes or disorder, where CSP analysis needs further development. Most of the methods used to compare experimental structures are appropriate for CSP generated structures, assuming that duplicates which only differ in the cell choice have already been eliminated. We certainly need to be able to cluster structures into families so that members of the same family could be observed as disorder components of experimental structures or be deemed to be so readily able to transform into the most stable structures of the family somewhere along the nucleation and growth pathway that they could not be observed, or are closely