Neural network to predict ²³Na NMR spectra of Na_n clusters

Abstract

In order to understand the charging and discharging processes of sodium-ion batteries, we are interested in the relationship between the size of sodium clusters inserted into the hard carbon anode and the solid-state ²³Na NMR chemical shifts. In this study, we investigated the predictability of the size dependence of ²³Na NMR shielding constants by SchNet, a deep learning model that uses the distance between Na atoms without graph connection information. The data set required for training the neural network was constructed by density functional theory (DFT) calculations. This study shows that the neural network model, which only used structural data, achieved comparable accuracy in predicting the shielding constant to the Lasso model, which utilized gross orbital population predicted from DFT calculations. Moreover, by introducing a penalty term to the neural network's loss function, the neural network was able to reproduce the skewed distribution of the shielding constant without modifying its architecture.

Keywords

Neural network SchNet / ²³Na NMR chemical shifts / Na_n clusters / DFT calculations

References

[1] Yabuuchi N,Dahbi M.Research development on sodium-ion batteries.Chem Rev2014;114:11636-82

[2] Morita R,Kubota K.Correlation of carbonization condition with metallic property of sodium clusters formed in hard carbon studied using ²³Na nuclear magnetic resonance.Carbon2019;145:712-5

[3] Morita R,Fukunishi M.Combination of solid state NMR and DFT calculation to elucidate the state of sodium in hard carbon electrodes.J Mater Chem A2016;4:13183-93

[4] Kipf TN.Semi-supervised classification with graph convolutional networks.arXiv preprint arXiv2016;1609.02907

[5] Kwon Y,Choi YS,Kang S.Neural message passing for NMR chemical shift prediction.J Chem Inf Model2020;60:2024-30

[6] Schütt KT,Chmiela S,Tkatchenko A.Quantum-chemical insights from deep tensor neural networks.Nat Commun2017;8:13890 PMCID:PMC5228054

[7] Schütt KT,Kindermans PJ,Müller KR.SchNet - a deep learning architecture for molecules and materials.J Chem Phys2018;148:241722

[8] Zhai H.Ensemble-average representation of pt clusters in conditions of catalysis accessed through GPU accelerated deep neural network fitting global optimization.J Chem Theory Comput2016;12:6213-26

[9] Frisch MJ,Schlegel HB. Gaussian 16. Gaussian Inc.: Wallingford, USA; 2016.

[10] Lee C,Parr RG.Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density.Phys Rev B Condens Matter1988;37:785-9

[11] Zhao Y,Truhlar DG.Design of density functionals by combining the method of constraint satisfaction with parametrization for thermochemistry, thermochemical kinetics, and noncovalent interactions.J Chem Theory Comput2006;2:364-82

[12] Santosa F.Linear inversion of band-limited reflection seismograms.SIAM J Sci Stat Comput1986;7:1307-30

[13] Schütt KT,Gastegger M,Tkatchenko A.SchNetPack: a deep learning toolbox for atomistic systems.J Chem Theory Comput2019;15:448-55

[14] Kingma DP.Adam: a method for stochastic optimization.arXiv preprint arXiv2014:1412.6980

[15] Noya EG,Wales DJ.Geometric magic numbers of sodium clusters: Interpretation of the melting behaviour.Eur Phys J D2007;43:57-60

[16] Sun WG,Lu C,Kuang XY.Evolution of the structural and electronic properties of medium-sized sodium clusters: a honeycomb-like Na₂₀ cluster.Inorg Chem2017;56:1241-8

[17] Elstner M,Jungnickel G.Self-consistent-charge density-functional tight-binding method for simulations of complex materials properties.Phys Rev B1998;58:7260-8

[18] Frauenheim T,Elsterner M.A self-consistent charge density-functional based tight-binding method for predictive materials simulations in physics, chemistry and biology.Physica Status Solidi2000;217:41-62

[19] Oliveira LFL,Cuny J.Benchmarking density functional based tight-binding for silver and gold materials: from small clusters to bulk.J Phys Chem A2016;120:8469-83

[20] Doye JPK.Identifying structural patterns in disordered metal clusters.Phys Rev B2003;68

[21] Doye JP.Lead clusters: different potentials, different structures.Comput Mater Sci2006;35:227-31

[22] Le Chen B,Kuang XY.Structural stability and evolution of medium-sized tantalum-doped boron clusters: a half-sandwich-structured TaB₁₂^- cluster.Inorg Chem2018;57:343-50

[23] Jin S,Kuang X.Structural and electronic properties of medium-sized aluminum-doped boron clusters AlB n and their anions.J Phys Chem C2019;123:6276-83

[24] Sun W,Lu C,Hermann A.Probing the structural and electronic properties of zirconium doped boron clusters: Zr distorted B₁₂ ligand framework.Phys Chem Chem Phys2018;20:23740-6

[25] Tian Y,Jin Y,Lu C.Exhaustive exploration of MgB_n (n = 10-20) clusters and their anions.Phys Chem Chem Phys2019;21:6935-41

[26] Xia XX,Kuang XY,Lu C.Study of the structural and electronic properties of neutral and charged niobium-doped silicon clusters: niobium encapsulated in silicon cages.J Phys Chem C2016;120:677-84