Pushing the frontiers of density functionals by solving the fractional electron problem

被引:266
作者
Kirkpatrick, James [1 ]
McMorrow, Brendan [1 ]
Turban, David H. P. [1 ]
Gaunt, Alexander L. [1 ]
Spencer, James S. [1 ]
Matthews, Alexander G. D. G. [1 ]
Obika, Annette [1 ]
Thiry, Louis [2 ]
Fortunato, Meire [1 ]
Pfau, David [1 ]
Castellanos, Lara Roman [1 ]
Petersen, Stig [1 ]
Nelson, Alexander W. R. [1 ]
Kohli, Pushmeet [1 ]
Mori-Sanchez, Paula [3 ,4 ]
Hassabis, Demis [1 ]
Cohen, Aron J. [1 ,5 ]
机构
[1] DeepMind, 6 Pancras Sq, London N1C 4AG, England
[2] PSL Univ, Dept Informat, CNRS, ENS, Paris, France
[3] UAM, Dept Quim, Madrid 28049, Spain
[4] UAM, IFIMAC, Madrid 28049, Spain
[5] Max Planck Inst Solid State Res, D-70569 Stuttgart, Germany
关键词
GENERALIZED GRADIENT APPROXIMATION; MAIN-GROUP THERMOCHEMISTRY; NONCOVALENT INTERACTIONS; ATOMIZATION ENERGIES; KINETICS; HYBRID; DFT; OPTIMIZATION; MECHANISMS; STEPWISE;
D O I
10.1126/science.abj6511
中图分类号
O [数理科学和化学]; P [天文学、地球科学]; Q [生物科学]; N [自然科学总论];
学科分类号
07 ; 0710 ; 09 ;
摘要
Density functional theory describes matter at the quantum level, but all popular approximations suffer from systematic errors that arise from the violation of mathematical properties of the exact functional. We overcame this fundamental limitation by training a neural network on molecular data and on fictitious systems with fractional charge and spin. The resulting functional, DM21 (DeepMind 21), correctly describes typical examples of artificial charge delocalization and strong correlation and performs better than traditional functionals on thorough benchmarks for main-group atoms and molecules. DM21 accurately models complex systems such as hydrogen chains, charged DNA base pairs, and diradical transition states. More crucially for the field, because our methodology relies on data and constraints, which are continually improving, it represents a viable pathway toward the exact universal functional.
引用
收藏
页码:1385 / +
页数:50
相关论文
共 73 条
[1]   Concerted vs Stepwise Mechanisms in Dehydro-Diels-Alder Reactions [J].
Ajaz, Aida ;
Bradley, Alexander Z. ;
Burrell, Richard C. ;
Li, William Hoi Hong ;
Daoust, Kimberly J. ;
Bovee, Laura Boddington ;
DiRico, Kenneth J. ;
Johnson, Richard P. .
JOURNAL OF ORGANIC CHEMISTRY, 2011, 76 (22) :9320-9328
[2]   Efficient Self-Consistent Implementation of Local Hybrid Functionals [J].
Bahmann, Hilke ;
Kaupp, Martin .
JOURNAL OF CHEMICAL THEORY AND COMPUTATION, 2015, 11 (04) :1540-1548
[3]   Combining active-space coupled-cluster approaches with moment energy corrections via the CC(P;Q) methodology: connected quadruple excitations [J].
Bauman, Nicholas P. ;
Shen, Jun ;
Piecuch, Piotr .
MOLECULAR PHYSICS, 2017, 115 (21-22) :2860-2891
[4]   DENSITY-FUNCTIONAL THERMOCHEMISTRY .3. THE ROLE OF EXACT EXCHANGE [J].
BECKE, AD .
JOURNAL OF CHEMICAL PHYSICS, 1993, 98 (07) :5648-5652
[5]   Fractional Kohn-Sham Occupancies from a Strong-Correlation Density Functional [J].
Becke, Axel D. .
DENSITY FUNCTIONALS: THERMOCHEMISTRY, 2015, 365 :175-186
[6]   Density functionals for static, dynamical, and strong correlation [J].
Becke, Axel D. .
JOURNAL OF CHEMICAL PHYSICS, 2013, 138 (07)
[7]   Isomerization of Bicyclo[1.1.0]butane by Means of the Diffusion Quantum Monte Carlo Method [J].
Berner, Raphael ;
Luechow, Arne .
JOURNAL OF PHYSICAL CHEMISTRY A, 2010, 114 (50) :13222-13227
[8]   The S66x8 benchmark for noncovalent interactions revisited: explicitly correlated ab initio methods and density functional theory [J].
Brauer, Brina ;
Kesharwani, Manoj K. ;
Kozuch, Sebastian ;
Martin, Jan M. L. .
PHYSICAL CHEMISTRY CHEMICAL PHYSICS, 2016, 18 (31) :20905-20925
[9]   Systematic optimization of long-range corrected hybrid density functionals [J].
Chai, Jeng-Da ;
Head-Gordon, Martin .
JOURNAL OF CHEMICAL PHYSICS, 2008, 128 (08)
[10]   DeePKS: A Comprehensive Data-Driven Approach toward Chemically Accurate Density Functional Theory [J].
Chen, Yixiao ;
Zhang, Linfeng ;
Wang, Han ;
Weinan, E. .
JOURNAL OF CHEMICAL THEORY AND COMPUTATION, 2021, 17 (01) :170-181