Capturing Anharmonicity in a Lattice Thermal Conductivity Model for High-Throughput Predictions

Samuel A. Miller, Prashun Gorai, Brenden R. Ortiz, Anuj Goyal, Duanfeng Gao, Scott A. Barnett, Thomas O. Mason, G. Jeffrey Snyder, Qin Lv, Vladan Stevanović, Eric S. TobererChemistry 2017 · DOI

Abstract: High-throughput, low-cost, and accurate predictions of thermal properties of new materials would be beneficial in fields ranging from thermal barrier coatings and thermoelectrics to integrated circuits. To date, computational efforts for predicting lattice thermal conductivity ($κ_L$) have been hampered by the complexity associated with computing multiple phonon interactions. In this work, we develop and validate a semiempirical model for $κ_L$ by fitting density functional theory calculations to experimental data. Experimental values for $κ_L$ come from new measurements on $SrIn_2O_4$, $Ba_2SnO_4$, $Cu_2ZnSiTe_4$, $MoTe_2$, $Ba_3In_2O_6$, $Cu_3TaTe_4$, $SnO$, and $InI$ as well as 55 compounds from across the published literature. To capture the anharmonicity in phonon interactions, we incorporate a structural parameter that allows the model to predict $κ_L$ within a factor of 1.5 of the experimental value across 4 orders of magnitude in $κ_L$ values and over a diverse chemical and structural phase space, with accuracy similar to or better than that of computationally more expensive models.

The Challenge of Predicting Thermal Conductivity

Predicting how heat moves through a solid via lattice vibrations (phonons) is notoriously difficult. Accurate values for lattice thermal conductivity ($κ_L$) typically require calculating "anharmonic" interatomic force constants, which involves simulating complex multi-phonon interactions. While accurate, these ab initio calculations are computationally expensive, often requiring weeks of supercomputer time for a single complex material. This creates a significant bottleneck for high-throughput discovery, where researchers aim to screen thousands of candidate materials for applications like thermoelectric waste-heat recovery or heat dissipation in microelectronics.

A Semiempirical Model for Anharmonicity

To address this, researchers have developed a semiempirical approach that blends the speed of simple analytical models with the accuracy of experimental data. The base model uses a simplified Debye–Callaway framework, treating heat transport as a sum of acoustic and optical phonon contributions.

The critical innovation lies in finding a proxy for the Grüneisen parameter ($γ$), which describes how vibrational frequencies change with volume—a direct measure of anharmonicity. Instead of calculating $γ$ directly, the authors discovered that the average coordination number (CN) serves as an excellent structural proxy. By integrating Density Functional Theory (DFT) to calculate "easy" properties like bulk modulus and density, the model can rapidly determine the speed of sound and subsequent thermal conductivity.

Comparison of Average Factor Difference (AFD) across various structural and chemical parameters, identifying average coordination number (Avg. CN) as the most effective proxy for anharmonicity.

Experimental Validation and Performance

The study utilized a dataset of 63 compounds to train and validate the model, including eight specific compounds synthesized and measured by the authors to fill gaps in the literature for low-conductivity materials. The model was refined by fitting parameters against this diverse experimental dataset using the Average Factor Difference (AFD) metric.

As shown in the parity plot below, the refined model predicts $κ_L$ within a factor of 1.5 of experimental values across four orders of magnitude. This performance is comparable to or better than much more expensive ab initio methods. By adding the coordination number parameter, the authors reduced the AFD from a baseline of 1.88 to 1.48.

Crystal structures of newly measured compounds and a parity plot comparing modeled vs. experimental lattice thermal conductivity (κL) across four orders of magnitude.

Key Insights into Structural Physics

  • Coordination as a Physics Shortcut: Local geometry dictates anharmonicity. Materials with low coordination (e.g., tetrahedral structures) tend to have lower Grüneisen parameters ($∼$0.75), while highly coordinated octahedral structures have higher values ($∼$1.5). This simple geometric rule replaces weeks of complex scattering calculations.
  • Complexity vs. Conductivity: While intuition suggests heavier atoms and complex unit cells lead to lower $κ_L$, the data showed an order of magnitude variation for compounds with similar masses, highlighting the necessity of a nuanced theory involving anharmonicity.
  • Universal Scaling: The model maintains accuracy across a massive range, from highly conductive materials like diamond to near-perfect thermal insulators.

Limitations and Future Outlook

The current model is designed for stoichiometric, bulk crystalline materials and does not yet account for solid solution alloys or nanostructured materials where grain boundary scattering dominates. Additionally, while the chemical space is diverse, the model was primarily trained on covalent and ionic materials; its performance on purely metallic or intermetallic systems may require further refinement. As more targeted experimental data becomes available, the model can be further tuned to provide even higher precision for specific classes of functional materials.

Summary updated

Abstract

High-throughput, low-cost, and accurate predictions of thermal properties of new materials would be beneficial in fields ranging from thermal barrier coatings and thermoelectrics to integrated circuits. To date, computational efforts for predicting lattice thermal conductivity (κL) have been hampered by the complexity associated with computing multiple phonon interactions. In this work, we develop and validate a semiempirical model for κL by fitting density functional theory calculations to experimental data. Experimental values for κL come from new measurements on SrIn2O4, Ba2SnO4, Cu2ZnSiTe4, MoTe2, Ba3In2O6, Cu3TaTe4, SnO, and InI as well as 55 compounds from across the published literature. To capture the anharmonicity in phonon interactions, we incorporate a structural parameter that allows the model to predict κL within a factor of 1.5 of the experimental value across 4 orders of magnitude in κL values and over a diverse chemical and structural phase space, with accuracy similar to or better than that of computationally more expensive models.
BibTeX
@inproceedings{miller2017capturing,
  author = {Samuel A. Miller and Prashun Gorai and Brenden R. Ortiz and Anuj Goyal and Duanfeng Gao and Scott A. Barnett and Thomas O. Mason and G. Jeffrey Snyder and Qin Lv and Vladan Stevanović and Eric S. Toberer},
  title = {{Capturing Anharmonicity in a Lattice Thermal Conductivity Model for High-Throughput Predictions}},
  booktitle = {Chemistry},
  year = {2017},
  doi = {10.1021/acs.chemmater.6b04179},
}
Figure