Artificial neural network potential for gold clusters

doi:10.1088/1674-1056/abc15d

中国物理B ›› 2020, Vol. 29 ›› Issue (11): 117304-.doi: 10.1088/1674-1056/abc15d

所属专题： SPECIAL TOPIC — Machine learning in condensed matter physics

Ling-Zhi Cao(曹凌志)¹^,², Peng-Ju Wang(王鹏举)³, Lin-Wei Sai(赛琳伟)⁴, Jie Fu(付洁)¹^,²^,†(), Xiang-Mei Duan(段香梅)¹^,²^,^‡()

收稿日期:2020-08-10 修回日期:2020-09-14 接受日期:2020-10-15 出版日期:2020-11-05 发布日期:2020-11-03

Artificial neural network potential for gold clusters

Ling-Zhi Cao(曹凌志)^1,2, Peng-Ju Wang(王鹏举)³, Lin-Wei Sai(赛琳伟)⁴, Jie Fu(付洁)^{1,2, †}, and Xiang-Mei Duan(段香梅)^{1,2,, ‡}

1 School of Physical Science and Technology, Ningbo University, Ningbo 315211, China
2 Laboratory of Clean Energy Storage and Conversion, Ningbo University, Ningbo 315211, China
3 Key Laboratory of Materials Modification by Laser, Ion and Electron Beams, Ministry of Education, Dalian University of Technology, Dalian 116024, China
4 College of Science, Hohai University, Changzhou 213022, China

Received:2020-08-10 Revised:2020-09-14 Accepted:2020-10-15 Online:2020-11-05 Published:2020-11-03
Contact: ^†Corresponding author. E-mail: fujie@nbu.edu.cn ^‡Corresponding author. E-mail: duanxiangmei@nbu.edu.cn
Supported by:
the National Natural Science Foundation of China (Grant Nos. 11804175, 11874033, 11804076, and 91961204) and the K.C. Wong Magna Foundation in Ningbo University.

摘要/Abstract

Abstract:

In cluster science, it is challenging to identify the ground state structures (GSS) of gold (Au) clusters. Among different search approaches, first-principles method based on density functional theory (DFT) is the most reliable one with high precision. However, as the cluster size increases, it requires more expensive computational cost and becomes impracticable. In this paper, we have developed an artificial neural network (ANN) potential for Au clusters, which is trained to the DFT binding energies and forces of 9000 Au_N clusters (11 ≤ N ≤ 100). The root mean square errors of energy and force are 13.4 meV/atom and 0.4 eV/Å, respectively. We demonstrate that the ANN potential has the capacity to differentiate the energy level of Au clusters and their isomers and highlight the need to further improve the accuracy. Given its excellent transferability, we emphasis that ANN potential is a promising tool to breakthrough computational bottleneck of DFT method and effectively accelerate the pre-screening of Au clusters’ GSS.

Key words: empirical potential, artificial neural network, gold cluster, first-principles

Ling-Zhi Cao(曹凌志), Peng-Ju Wang(王鹏举), Lin-Wei Sai(赛琳伟), Jie Fu(付洁), Xiang-Mei Duan(段香梅). [J]. 中国物理B, 2020, 29(11): 117304-.

Ling-Zhi Cao(曹凌志), Peng-Ju Wang(王鹏举), Lin-Wei Sai(赛琳伟), Jie Fu(付洁), and Xiang-Mei Duan(段香梅). Artificial neural network potential for gold clusters[J]. Chin. Phys. B, 2020, 29(11): 117304-.

图/表 6

参考文献 59

[1]	Daniel M C Astruc D 2004 Chem. Rev. 104 293 DOI: 10.1021/cr030698+
[2]	Ahmed S Annu Ikram S Yudha S S 2016 J. Photochem. Photobiol. B 161 141 DOI: 10.1016/j.jphotobiol.2016.04.034
[3]	Turkevich J Stevenson P C Hillier J 1951 Discuss. Faraday Soc. 11 55 DOI: 10.1039/df9511100055
[4]	Mingos D M 2015 Dalton Trans. 44 6680 DOI: 10.1039/C5DT00253B
[5]	Wang J Wang G Zhao J 2002 Phys Rev B 66 035418 DOI: 10.1103/PhysRevB.66.035418
[6]	Thorn A Rojas-Nunez J Hajinazar S Baltazar S E Kolmogorov A N 2019 J. Phys. Chem. C 123 30088 DOI: 10.1021/acs.jpcc.9b08517
[7]	Yang D Pei W Zhou S Zhao J Ding W Zhu Y 2020 Angew. Chem. Int. Ed. 59 1919 DOI: 10.1002/anie.v59.5
[8]	Zhou S Pei W Du Q Zhao J 2019 Phys. Chem. Chem. Phys. 21 10587 DOI: 10.1039/C9CP01517E
[9]	Hu Y L Zhu H R Wei S H 2019 Chin. Phys. B 28 113101 DOI: 10.1088/1674-1056/ab4cdd
[10]	Du Q Wu X Wang P Wu D Sai L King R B Park S J Zhao J 2020 J. Phys. Chem. C 124 7449 DOI: 10.1021/acs.jpcc.9b11588
[11]	Chen D D Kuang X Y Zhao Y R Shao P Li Y F 2011 Chin. Phys. B 20 063601 DOI: 10.1088/1674-1056/20/6/063601
[12]	Haruta M Yamada N Kobayashi T Iijima S 1989 J. Catal. 115 301 DOI: 10.1016/0021-9517(89)90034-1
[13]	Brust M Walker M Bethell D Schiffrin D J Whyman R 1994 J. Chem. Soc., Chem. Commun. 1994 801 DOI: 10.1039/C39940000801
[14]	Mingos D M P 1996 J. Chem. Soc., Dalton Trans. 5 561 DOI: 10.1039/DT9960000561
[15]	Häberlen O D Chung S C Stener M Rösch N 1997 J. Chem. Phys. 106 5189 DOI: 10.1063/1.473518
[16]	Oliveira L F L Tarrat N Cuny J Morillo J Lemoine D Spiegelman F Rapacioli M 2016 J. Phys. Chem. A. 120 8469 DOI: 10.1021/acs.jpca.6b09292
[17]	Dong Y Springborg M 2007 J. Phys. Chem. C 111 12528 DOI: 10.1021/jp071120x
[18]	Tarrat N Rapacioli M Cuny J Morillo J Heully J L Spiegelman F 2017 Comput. Theor. Chem. 1107 102 DOI: 10.1016/j.comptc.2017.01.022
[19]	Koskinen P Häkkinen H Seifert G Sanna S Frauenheim T Moseler M 2006 New. J. Phys. 8 9 DOI: 10.1088/1367-2630/8/1/009
[20]	Pyykko P 2004 Angew. Chem. Int. Ed. 43 4412 DOI: 10.1002/(ISSN)1521-3773
[21]	Pyykko P 2008 Chem. Soc. Rev. 37 1967 DOI: 10.1039/b708613j
[22]	Xiao L Tollberg B Hu X Wang L 2006 J. Chem. Phys. 124 114309 DOI: 10.1063/1.2179419
[23]	Sutton A P Chen J 1990 Phil. Mag. Lett. 61 139 DOI: 10.1080/09500839008206493
[24]	Cleri F Rosato V V 1993 Phys Rev B 48 22 DOI: 10.1103/PhysRevB.48.22
[25]	Artrith N Kolpak A M 2015 Comp. Mater. Sci. 110 20 DOI: 10.1016/j.commatsci.2015.07.046
[26]	Lee K Yoo D Jeong W Han S 2019 Comput. Phys. Commun. 242 95 DOI: 10.1016/j.cpc.2019.04.014
[27]	Sumpter B G Noid D W 1992 Chem. Phys. Lett. 192 455 DOI: 10.1016/0009-2614(92)85498-Y
[28]	Tai No K Ha Chang B Yeon Kim S Shik Jhon M Scheraga H A 1997 Chem. Phys. Lett. 271 152 DOI: 10.1016/S0009-2614(97)00448-X
[29]	Lorenz S Groß A Scheffler M 2004 Chem. Phys. Lett. 395 210 DOI: 10.1016/j.cplett.2004.07.076
[30]	Handley C M Popelier P L 2010 J. Phys. Chem. A. 114 3371 DOI: 10.1021/jp9105585
[31]	Guo P Zheng J M Zhao P Zheng L L Ren Z Y 2010 Chin. Phys. B 19 083601 DOI: 10.1088/1674-1056/19/8/083601
[32]	Artrith N Urban A 2016 Comp. Mater. Sci. 114 135 DOI: 10.1016/j.commatsci.2015.11.047
[33]	Artrith N Urban A Ceder G 2017 Phys Rev B 96 014112 DOI: 10.1103/PhysRevB.96.014112
[34]	Broyden C G 1970 IMA J. Appl. Math. 6 76 DOI: 10.1093/imamat/6.1.76
[35]	Fletcher R 1970 The Computer Journal 13 317 DOI: 10.1093/comjnl/13.3.317
[36]	Goldfarb D 1970 Math. Comput. 24 23 DOI: 10.1090/S0025-5718-1970-0258249-6
[37]	Shanno D F 1970 Math. Comput. 24 647 DOI: 10.1090/S0025-5718-1970-0274029-X
[38]	Byrd R H Lu P Nocedal J Zhu C 1995 SIAM. J. Sci. Comput. 16 1190 DOI: 10.1137/0916069
[39]	Behler J 2011 Phys. Chem. Chem. Phys 13 17930 DOI: 10.1039/c1cp21668f
[40]	Behler J Parrinello M 2007 Phys. Rev. Lett. 98 146401 DOI: 10.1103/PhysRevLett.98.146401
[41]	Artrith N Kolpak A M 2014 Nano. Lett. 14 2670 DOI: 10.1021/nl5005674
[42]	Khaliullin R Z Eshet H Kühne T D Behler J Parrinello M 2010 Phys Rev B 81 100103 DOI: 10.1103/PhysRevB.81.100103
[43]	Artrith N Morawietz T Behler J 2011 Phys Rev B 83 153101 DOI: 10.1103/PhysRevB.83.153101
[44]	Morawietz T Behler J 2013 J. Phys. Chem. A. 117 7356 DOI: 10.1021/jp401225b
[45]	Artrith N Behler J 2012 Phys Rev B 85 045439 DOI: 10.1103/PhysRevB.85.045439
[46]	Artrith N Hiller B Behler J 2013 Phys Status Solidi B 250 1191 DOI: 10.1002/pssb.v250.6
[47]	Delley B 1990 J. Chem. Phys. 92 508 DOI: 10.1063/1.458452
[48]	Delley B 2000 J. Chem. Phys. 113 7756 DOI: 10.1063/1.1316015
[49]	Hohenberg P Kohn W 1964 Phys. Rev. 136 B864 DOI: 10.1103/PhysRev.136.B864
[50]	Kohn W Sham L J 1965 Phys. Rev. 140 A1133 DOI: 10.1103/PhysRev.140.A1133
[51]	Perdew J P Burke K Ernzerhof M 1996 Phys. Rev. Lett. 77 3865 DOI: 10.1103/PhysRevLett.77.3865
[52]	Perdew J P Burke K Ernzerhof M 1997 Phys. Rev. Lett. 78 1396 DOI: 10.1103/PhysRevLett.78.1396
[53]	Behler J 2011 J. Chem. Phys. 134 074106 DOI: 10.1063/1.3553717
[54]	Doye J P K Wales D J 1998 New. J. Chem. 22 733 DOI: 10.1039/A709249K
[55]	http://www-wales.ch.cam.ac.uk/CCD.html
[56]	Nhat P V Si N T Nguyen M T 2020 J. Phys. Chem. A. 124 1289 DOI: 10.1021/acs.jpca.9b09287
[57]	Plimpton S 1995 J. Comput. Phys. 117 1 DOI: 10.1006/jcph.1995.1039
[58]	G W Rodrigue G 1989 Math. Comput. 53 775 DOI: 10.2307/2008751
[59]	https://lammps.sandia.gov/

Network	Number of weight	RMSE/(meV/atom)		MAE/(meV/atom)
architecture	parameters	Training set	Test set	Training set	Test set
26–2–2–1	63	19.2	20.4	14.1	14.5
26–5–5–1	171	18.6	17.8	13.7	13.2
26–10–10–1	391	16.2	16.1	12.0	11.9
26–15–15–1	661	16.0	16.7	11.9	12.4
26–20–20–1	981	15.3	17.2	11.4	12.5

Artificial neural network potential for gold clusters

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 6

参考文献 59

相关文章 0

Metrics

本文评价

推荐阅读 0