算法
晶体结构
计算机科学
数据结构
字节
标识符
晶体结构预测
NIST公司
Python(编程语言)
禁忌搜索
试验数据
结晶学
程序设计语言
化学
自然语言处理
作者
Silvia Bahmann,Jens Kortus
标识
DOI:10.1016/j.cpc.2013.02.007
摘要
We present EVO—an evolution strategy designed for crystal structure search and prediction. The concept and main features of biological evolution such as creation of diversity and survival of the fittest have been transferred to crystal structure prediction. EVO successfully demonstrates its applicability to find crystal structures of the elements of the 3rd main group with their different spacegroups. For this we used the number of atoms in the conventional cell and multiples of it. Running EVO with different numbers of carbon atoms per unit cell yields graphite as the lowest energy structure as well as a diamond-like structure, both in one run. Our implementation also supports the search for 2D structures and was able to find a boron sheet with structural features so far not considered in literature. Program title: EVO Catalogue identifier: AEOZ_v1_0 Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEOZ_v1_0.html Program obtainable from: CPC Program Library, Queen’s University, Belfast, N. Ireland Licensing provisions: GNU General Public License version 3 No. of lines in distributed program, including test data, etc.: 23488 No. of bytes in distributed program, including test data, etc.: 1830122 Distribution format: tar.gz Programming language: Python. Computer: No limitations known. Operating system: Linux. RAM: Negligible compared to the requirements of the electronic structure programs used Classification: 7.8. External routines: Quantum ESPRESSO (http://www.quantum-espresso.org/), GULP (https://projects.ivec.org/gulp/) Nature of problem: Crystal structure search is a global optimisation problem in 3N+3 dimensions where N is the number of atoms in the unit cell. The high dimensional search space is accompanied by an unknown energy landscape. Solution method: Evolutionary algorithms transfer the main features of biological evolution to use them in global searches. The combination of the “survival of the fittest” (deterministic) and the randomised choice of the parents and normally distributed mutation steps (non-deterministic) provides a thorough search. Restrictions: The algorithm is in principle only restricted by a huge search space and simultaneously increasing calculation time (memory, etc.), which is not a problem for our piece of code but for the used electronic structure programs. Running time: The simplest provided case runs serially and takes 30 minutes to one hour. All other calculations run for significantly longer time depending on the parameters like the number and sort of atoms and the electronic structure program in use as well as the level of parallelism included.
科研通智能强力驱动
Strongly Powered by AbleSci AI