PINK: physical-informed machine learning for lattice thermal conductivity

Abstract

Lattice thermal conductivity (κ_L) is crucial for efficient thermal management in electronics and energy conversion technologies. Traditional methods for predicting κ_L are often computationally expensive, limiting their scalability for large-scale material screening. Empirical models, such as the Slack model, offer faster alternatives but require time-consuming calculations for key parameters such as sound velocity and the Grüneisen parameter. This work presents a high-throughput framework, physical-informed kappa (PINK), which combines the predictive power of crystal graph convolutional neural networks (CGCNNs) with the physical interpretability of the Slack model to predict κ_L directly from crystallographic information files (CIFs). Unlike previous approaches, PINK enables rapid, batch predictions by extracting material properties such as bulk and shear modulus from CIFs using a well-trained CGCNN model. These properties are then used to compute the necessary parameters for κ_L calculation through a simplified physical formula. PINK was applied to a dataset of 377,221 stable materials, enabling the efficient identification of promising candidates with ultralow κ_L values, such as Ag₃Te₄W and Ag₃Te₄Ta. The platform, accessible via a user-friendly interface, offers an unprecedented combination of speed, accuracy, and scalability, significantly accelerating material discovery for thermal management and energy conversion applications.

Keywords

Physical-informed machine learning / thermoelectrics / lattice thermal conductivity / phonon engineering

References

[1] Li N,Wang L,Hänggi P.Colloquium: Phononics: manipulating heat flow with electronic analogs and beyond.Rev Mod Phys2012;84:1045

[2] van Erp R, Soleimanzadeh R, Nela L, Kampitsis G, Matioli E. Co-designing electronics with microfluidics for more sustainable cooling.Nature2020;585:211-6

[3] Kim HS,Chen G,Ren Z.Relationship between thermoelectric figure of merit and energy conversion efficiency.Proc Natl Acad Sci U S A2015;112:8205-10 PMCID:PMC4500231

[4] Ouyang H,Gao Z.Kirigami-inspired thermal regulator.Phys Rev Appl2023;19:L011001

[5] Kang JS,Wu H,Aoki T.Integration of boron arsenide cooling substrates into gallium nitride devices.Nat Electron2021;4:416-23

[6] Xu L,Wang Y,Ding X.Enhanced average power factor and ZT value in PbSe thermoelectric material with dual interstitial doping.Energy Environ Sci2024;17:2018-27

[7] Broido DA,Birner G,Stewart DA.Intrinsic lattice thermal conductivity of semiconductors from first principles.Appl Phys Lett2007;91:231922

[8] Stackhouse S.Theoretical methods for calculating the lattice thermal conductivity of minerals.Rev Mineral Geochem2010;71:253-69

[9] Tadano T.Self-consistent phonon calculations of lattice dynamical properties in cubic SrTiO₃ with first-principles anharmonic force constants.Phys Rev B2015;92:054301

[10] Carbogno C,Scheffler M.Ab initio green-kubo approach for the thermal conductivity of solids.Phys Rev Lett2017;118:175901

[11] Fan Z,Wang H,Donadio D.Force and heat current formulas for many-body potentials in molecular dynamics simulations with applications to thermal conductivity calculations.Phys Rev B2015;92:094301

[12] Yusuf A.Electrical, thermomechanical and cost analyses of a low-cost thermoelectric generator.Energy2022;241:122934

[13] Luo Y,Yuan H,Fang Y.Predicting lattice thermal conductivity via machine learning: a mini review.npj Comput Mater2023;9:964

[14] Mishin Y.Machine-learning interatomic potentials for materials science.Acta Mater2021;214:116980

[15] Liu C,Zhao Y.Actively and reversibly controlling thermal conductivity in solid materials.Phys Rep2024;1058:1-32

[16] Shi XL,Lyu W.Advancing flexible thermoelectrics for integrated electronics.Chem Soc Rev2024;53:9254-305

[17] Callaway J.Model for lattice thermal conductivity at low temperatures.Phys Rev1959;113:1046-51

[18] Zhang Y.First-principles Debye–Callaway approach to lattice thermal conductivity.J Materiomics2016;2:237-47

[19] Slack G.Nonmetallic crystals with high thermal conductivity.J Phys Chem Solids1973;34:321-35

[20] Morelli DT.High lattice thermal conductivity solids. In: Shindé SL, Goela JS, editors. High thermal conductivity materials. New York: Springer-Verlag; 2006. pp. 37-68.

[21] Nielsen MD,Heremans JP.Lone pair electrons minimize lattice thermal conductivity.Energy Environ Sci2013;6:570-8

[22] Chang C.Anharmoncity and low thermal conductivity in thermoelectrics.Mater Today Phys2018;4:50-7

[23] Virkar AV.Fabrication of high-thermal-conductivity polycrystalline aluminum nitride: thermodynamic and kinetic aspects of oxygen removal. In: Shindé SL, Goela JS, editors. High thermal conductivity materials. New York: Springer-Verlag; 2006. pp. 143-66.

[24] Chen S,Yan W.First-principles investigation of elastic anisotropy and thermal transport property of transition metal monosilicides CrSi, TiSi, and ZrSi under pressure.Mater Today Commun2024;39:108958

[25] Cao Y,Wang X.High-throughput screening of potentially ductile and low thermal conductivity ABX₃ (X = S, Se, Te) thermoelectric perovskites.Appl Phys Lett2024;124:092101

[26] Li R,Xi L,Singh DJ.High-throughput screening for advanced thermoelectric materials: diamond-like ABX₂ compounds.ACS Appl Mater Interfaces2019;11:24859-66

[27] Qin G,Liu Y.High-throughput computational evaluation of lattice thermal conductivity using an optimized Slack model.Mater Adv2022;3:6826-30

[28] Cao Y,Li X.Unraveling the relationships between chemical bonding and thermoelectric properties: n-type ABO₃ perovskites.J Mater Chem A2022;10:11039-45

[29] Wang X,Zhu G.An interpretable formula for lattice thermal conductivity of crystals.Mater Today Phys2024;48:101549

[30] Anderson OL.A simplified method for calculating the debye temperature from elastic constants.J Phys Chem Solids1963;24:909-17

[31] Belomestnykh VN.The acoustical Grüneisen constants of solids.Tech Phys Lett2004;30:91-3

[32] Ju S,Liu C.Exploring diamondlike lattice thermal conductivity crystals via feature-based transfer learning.Phys Rev Mater2021;5:053801

[33] Xie T.Crystal graph convolutional neural networks for an accurate and interpretable prediction of material properties.Phys Rev Lett2018;120:145301

[34] Zhu T,Gong S.Charting lattice thermal conductivity for inorganic crystals and discovering rare earth chalcogenides for thermoelectrics.Energy Environ Sci2021;14:3559-66

[35] Vaitesswar US,Huang T.Machine learning based feature engineering for thermoelectric materials by design.Digit Discov2024;3:210-20

[36] Omee SS,Dong R,Hu J.Structure-based out-of-distribution (OOD) materials property prediction: a benchmark study.npj Comput Mater2024;10:1316

[37] Meredig B,Church C.Can machine learning identify the next high-temperature superconductor? Examining extrapolation performance for materials discovery.Mol Syst Des Eng2018;3:819-25

[38] Merchant A,Schoenholz SS,Cheon G.Scaling deep learning for materials discovery.Nature2023;624:80-5 PMCID:PMC10700131

[39] Griesemer SD,Wolverton C.Accelerating the prediction of stable materials with machine learning.Nat Comput Sci2023;3:934-45

[40] Wang X,Zhu G.Role of high-order anharmonicity and off-diagonal terms in thermal conductivity: a case study of multiphase CsPbBr₃.Phys Rev B2023;107:214308

[41] Herring C.Role of low-energy phonons in thermal conduction.Phys Rev1954;95:954-65

[42] Jia T,Zhang Y.Lattice thermal conductivity evaluated using elastic properties.Phys Rev B2017;95:155206

[43] Knoop F,Scheffler M.Anharmonicity measure for materials.Phys Rev Mater2020;4:083809

[44] Belomestnykh VN.Interrelation between anharmonicity and lateral strain in quasi-isotropic polycrystalline solids.Tech Phys2004;49:1098-100

[45] Chung DH.The Voigt-Reuss-Hill approximation and elastic moduli of polycrystalline MgO, CaF₂, β-ZnS, ZnSe, and CdTe.J Appl Phys1967;38:2535-40

[46] Xiao Y,Pei Y.Origin of low thermal conductivity in SnSe.Phys Rev B2016;94:125203

[47] Kresse G.Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set.Phys Rev B Condens Matter1996;54:11169-86

[48] Blöchl PE.Projector augmented-wave method.Phys Rev B Condens Matter1994;50:17953-79

[49] Kresse G.From ultrasoft pseudopotentials to the projector augmented-wave method.Phys Rev B1999;59:1758-75

[50] Perdew JP,Ernzerhof M.Generalized gradient approximation made simple.Phys Rev Lett1996;77:3865-8

[51] Li W,A .Katcho N, Mingo N. ShengBTE: A solver of the Boltzmann transport equation for phonons.Comput Phys Commun2014;185:1747-58

[52] Dunn A,Ganose A,Jain A.Benchmarking materials property prediction methods: the Matbench test set and Automatminer reference algorithm.npj Comput Mater2020;6:406

[53] AFLOW. https://aflowlib.org/. (accessed 2025-02-17)

[54] Beasley JD.Thermal conductivities of some novel nonlinear optical materials.Appl Opt1994;33:1000-3

[55] Toberer ES,Snyder GJ.Phonon engineering through crystal chemistry.J Mater Chem2011;21:15843

[56] Slack GA.Thermal conductivity of MgO, Al₂O₃, MgAl₂O₄, and Fe₃O₄ crystals from 3° to 300° K.Phys Rev1962;126:427-41

[57] Bjerg L,Madsen GKH.Modeling the thermal conductivities of the zinc antimonides ZnSb and Zn₄Sb₃.Phys Rev B2014;89:024304

[58] Toher C,Plata JJ.Combining the AFLOW GIBBS and elastic libraries to efficiently and robustly screen thermomechanical properties of solids.Phys Rev Mater2017;1:015401

[59] Curtarolo S,Hart GL.AFLOW: an automatic framework for high-throughput materials discovery.Comput Mater Sci2012;58:218-26

[60] Wang Y,Wang X.Anomalous thermal conductivity in 2D silica nanocages of immobilizing noble gas atom.Appl Phys Lett2024;124:122205

[61] Zhou F,Xia Y.Lattice anharmonicity and thermal conductivity from compressive sensing of first-principles calculations.Phys Rev Lett2014;113:185501

[62] Hao Y,Zheng J.Machine learning for predicting ultralow thermal conductivity and high ZT in complex thermoelectric materials.ACS Appl Mater Interfaces2024;16:47866-78

[63] Zhang Y,Chen C,Kent PRC.Thermodynamic properties of PbTe, PbSe, and PbS: first-principles study.Phys Rev B2009;80:024304

[64] Zhao LD,Zhang Y.Ultralow thermal conductivity and high thermoelectric figure of merit in SnSe crystals.Nature2014;508:373-7

[65] Pashinkin AS,Malkova AS.Heat capacity and thermodynamic properties of lead selenide and lead telluride.Inorg Mater2009;45:1226-9

[66] Tian Z,Esfarjani K,Shiomi J.Phonon conduction in PbSe, PbTe, and PbTe_1-xSe_x from first-principles calculations.Phys Rev B2012;85:184303

[67] Zhou Z,Liu J.High-throughput prediction of the carrier relaxation time via data-driven descriptor.npj Comput Mater2020;6:417

[68] Madsen GK.BoltzTraP. A code for calculating band-structure dependent quantities.Comput Phys Commun2006;175:67-71

[69] Li X,Xi J.TransOpt. A code to solve electrical transport properties of semiconductors in constant electron–phonon coupling approximation.Comput Mater Sci2021;186:110074

[70] Karamad M,Shi Y,Gates ID.Orbital graph convolutional neural network for material property prediction.Phys Rev Mater2020;4:093801

[71] Chen C,Zuo Y,Ong SP.Graph networks as a universal machine learning framework for molecules and crystals.Chem Mater2019;31:3564-72

[72] Cheng J,Dong L.A geometric-information-enhanced crystal graph network for predicting properties of materials.Commun Mater2021;2:194

[73] Choudhary K.Atomistic line graph neural network for improved materials property predictions.npj Comput Mater2021;7:650

[74] Louis SY,Nasiri A.Graph convolutional neural networks with global attention for improved materials property prediction.Phys Chem Chem Phys2020;22:18141-8

[75] Omee SS,Fu N.Scalable deeper graph neural networks for high-performance materials property prediction.Patterns2022;3:100491 PMCID:PMC9122959

[76] Isayev O,Toher C,Curtarolo S.Universal fragment descriptors for predicting properties of inorganic crystals.Nat Commun2017;8:15679 PMCID:PMC5465371

[77] Ward L,Krishna A.Including crystal structure attributes in machine learning models of formation energies via Voronoi tessellations.Phys Rev B2017;96:024104

[78] Ruff R,Stühmer J.Connectivity optimized nested line graph networks for crystal structures.Digit Discov2024;3:594-601