xiaoqiugood的个人博客分享 http://blog.sciencenet.cn/u/xiaoqiugood

博文

USPEX算例简介

已有 9180 次阅读 2015-2-22 16:04 |个人分类:物质结构及其预测|系统分类:科研笔记|关键词:学者

关注:

1) USPEX、CALYPSO等的最新发表文章

2) USPEX、CALYPSO等的经典算例分析


附-各算例简介:

http://han.ess.sunysb.edu/uspex_examples/

 

****************************************************
*Here we give examples of calculations using USPEX.*
****************************************************
We recommend that before doing production calculations, you try to run these tests to make sure that everything works. If any problems, contact us.


We also recommend that when you do production calculations, you start with one of these "test" folders and edit input files - instead of creating
them from scratch.



Before testing, put executables of VASP, SIESTA, GULP,... (whichever codes you are using) in folders Specific as ./Specific/vasp,
./Specific/siesta, ./Specific/gulp.

【将可执行文件拷入Specific目录,不是必须的?】


To run USPEX, simply type:
nohup matlab < USPEX.m > log &
(or analogous command line with Octave). Well, this is described in the Manual - and before doing anything, you should read the Manual!



When analyzing your tests - you can compare your results with the "standard" testing results that can be found in folders "reference".



Folders AuxiliaryFiles and generationNN are removed from reference in all tests (except test #4) to decrease the size of the test folders.



Most tests are tailored to be fast, rather than accurate or physically meaningful. They test the basic machinery - the variation operators and
different functionalities of the code. Keep this in mind! Most tests here are done with GULP, whose speed is ideal for testing purposes.



But beware - most of the standard interatomic potentials result in wrong ground states! For MgAl2O4 the Buckingham potential we used in these tests gives
reasonable structures at high pressures, but fails at low pressure!!!
 
If you want good results, you should have "more expensive" calculations and should use ab initio calculations.

【如何在保持结果准确性的前提下,提高计算效率,节省时间】


Also, in real calculations you should not stop the code after 5-10 generations (as is done in some of the examples). Please give
USPEX some time to explore the energy landscape and find with confidence the globally optimum solution.



++++++++++++++++
+LIST OF TESTS:+
++++++++++++++++

(**)T0_Physics_VASP: Silicon (8 atoms/cell) at zero pressure.

Variable-cell DFT calculation using VASP, PBE96 functional.

Many thanks to G.Kresse for
permission to include his PAW files (POTCAR) in our distribution.  Test created by ARO, updated by AOL and ARO.

(**)T1_Physics_GULP: MgAl2O4 (28 atoms/cell) at 100 GPa pressure.


Variable-cell calculation using Buckingham potentials, GULP code. This test has
direct bearing on the physics of the Earth's interior! Test created by ARO, updated by AOL and ARO.  

(**)T2_conv_GULP:

this example shows how to do structure prediction when you know cell parameters. MgSiO3 (20 atoms/cell) with Buckingham potentials,
GULP code.

Cell parameters correspond to post-perovskite. Test created by ARO, updated by AOL and ARO.


(**)T3_antiseeds_GULP:

Calculations for Lennard_Jones nanoparticle with 38 atoms are done with different antiseeds (ageing) settings. 【什么是antiseeds ?!】

(**)T4_restart_GULP:

this example shows how to restart from a previous calculation.

The previous results are in folder results1, and when you run
folder results2 appears.

MgAl2O4 (14 atoms/cell) at 100 GPa pressure.

Variable-cell calculation using Buckingham potentials, GULP code.
Test created by ARO, updated by AOL.
!!! IMPORTANT !!! You can not restart matlab calculation using octave and octave calculation using matlab, since the formats of .mat files are
different.

Therefore please use the appropriate results1 folder (rename results1_octave into results1 if you use octave; default results1 is
the matlab calculation).

【不明白】

(**)T5_seeds_GULP:


this example shows how to use seed structures (i.e. starting not from fully random structures, but to include some structures
that you think might be reasonable).


Seed structures are given in the file Seeds/POSCARS. MgAl2O4 (14 atoms/cell) at 100 GPa pressure. Variable-cell
calculation using Buckingham potentials, GULP code. Test created by ARO, updated by AOL.  

【Seed structures 不是必须的?】


(**)T6_hardness_GULP:

this test shows the optimisation of hardness (search for hardest structure in a given chemical system).


MgO (8 atoms/cell). Variable-cell calculation using Buckingham potentials, GULP code.

Test created by ARO, updated by AOL and ARO.


Based on GULP and hardness formula of Li (2009) embedded in valence-topological approach of Lyakhov and Oganov (2010, 2011).
 
(**)T7_varcomp_GULP:

Lennard-Jones binary system, GULP, and variable-composition method of Wang, Lyakhov and Oganov (2008, 2010, 2012). Test created by ARO, updated by AOL.


(**)T8_clusters_GULP:

cluster (nanoparticle) structure optimisation.

Lennard-Jones cluster with 13 atoms. GULP code. Test created by AOL.


(**)T9_GULP_VASP:


Calculations in which optimisation starts with GULP and the last step is done with VASP for better accuracy.

Test created by ARO, updated by AOL and ARO. This test is very short - only 5 generations, just to test the gulp-vasp functionality.


(**)T10_Performance_GULP:

SrTiO3 (50 atoms/cell) at zero pressure.


Variable-cell calculation using Buckingham potentials, GULP code. Running this test you
can see that even for such a relatively large system USPEX code scores a >90% success rate and remarkable efficiency. This contrasts with a 7-12% success
rate reported for the same system and using the same potential by Zurek & Lonie. Clearly, USPEX outperforms the poor reimplementation of our method by
Zurek and Lonnie. We have witnessed excellent performance of our code also for much larger systems. Test created by ARO, updated by AOL.


(**)T11_molecule_DMACRYS: Methane molecular crystal with 8 molecules in the variable unit cell.


Molecule is described in the file MOL_1.
DMACRYS code is used for optimization. Please put executables dmacrys, neighcrys-pp, neighcrys-vv in the folder Specific. Test created by Q. Zhu.

(**)T12_molecule_VASP: Methane molecular crystal with 4 molecules in the variable unit cell.


The input is similar to that in T11, except VASP code is used for optimization. Test created by Q. Zhu.

(**)T13_Castep: Carbon (8 atoms/cell) at zero pressure.


Variable-cell DFT calculation using CASTEP. Test created by Z. Raza and X. Dong.



(**)T14_Physics_PSO_GULP: MgAl2O4 (28 atoms/cell) at 100 GPa pressure.


Variable-cell calculation using Buckingham potentials, GULP code.
Optimization is done using the Particle Swarm Optimization (PSO) algorithm of Lyakhov and Oganov (2012), which introduces important corrections to
earlier PSO algorithm of Boldyrev (and a very similar to it algorithm of Wang-Lv-Zhu-Ma).


To activate it, file PSO_input has to be present in the main folder. Test created by AOL.


(**)T15_random_sampling_GULP:

Test of random sampling approach (one large generation) with symmetric initialization (Zhu et al(2012), Lyakhov et al (2012)).
Shortened form of the input file is used (most parameters are default ones in this case).
MgAl2O4 (14 atoms/cell) at 100 GPa pressure. Variable-cell calculation using Buckingham potentials, GULP code. Test created by AOL.


(**)T16_metadynamics_GULP:

Evolutionary metadynamics (Zhu, Oganov, Lyakhov (2012)) illustrated for MgSiO3 (20 atoms/cell) using the GULP code.


The initial structure should be provided in the vasp 5.2 format (i.e. with added line containing element names) in the file POSCAR_1. Test created by Zhu.

(**)T17_order_optimization_GULP: Optimization of the structural order (Oganov and Valle (2009); Lyakhov Oganov Valle (2010)) for MgAl2O4 (14 atoms/cell) at 100 GPa pressure.

Variable-cell calculation using Buckingham potentials, GULP code. Note that the most ordered structure is not the ground state for this system! Test created by AOL.

(**)T18_order_without_relaxation: Optimization of the structural order (Oganov and Valle (2009); Lyakhov Oganov Valle (2010)) without relaxation (abinitioCode = 0).

Seed structure (supercell of Ti-Cu-O-structure) is permutated in a search of the permutant with the minimum/maximum order. Minimizing order in this situation, one gets a generalized version of the "special quasirandom structure". Test created by AOL and ARO.

(**)T19_QuantumEspresso: Carbon (8 atoms/cell) at zero pressure. Variable-cell DFT calculation using Quantum Espresso. Test created by AOL.

(**)T20_ATK: Carbon (8 atoms/cell) at zero pressure. Variable-cell DFT calculation using ATK. Test created by X Dong.


(**)T21_Lammps: Carbon (16 atoms/cell) at zero pressure. Variable-cell. Test created by X Dong.


(**)T22_VCNEB: Test of AlN B3->B1 phase transition at 14GPa.  Test created by Qian.
***********************
***********************
NOTE: USPEX works with a variety of other codes (CP2K, MD++). We will include test cases for at least some of these codes gradually.




https://m.sciencenet.cn/blog-567091-869488.html

上一篇:晶体结构预测-example 解读:变胞预测
下一篇:USPEX手册解读:学习从手册开始

0

该博文允许注册用户评论 请点击登录 评论 (0 个评论)

数据加载中...

Archiver|手机版|科学网 ( 京ICP备07017567号-12 )

GMT+8, 2024-6-2 20:10

Powered by ScienceNet.cn

Copyright © 2007- 中国科学报社

返回顶部