Development of Machine Learning Models for Thermoelectric Figure of Merit Predictions of Doped Chalcogenides

Abstract:

Machine Learning (ML) approach seeks to open new frontiers in the search for novel thermoelectric materials that convert heat waste into useful electrical energy. Five regression-type ML algorithms Linear Regression, Random Forest, XGBoost, Bagging Regressor, and Gradient Boosting Regressor were employed in this study to forecast the thermoelectric figure of merit (ZT) of doped chalcogenide compounds. Gradient Boosting Regressor achieved the best baseline performance (R² = 94.5%, MAE = 0.073, RMSE = 0.128), further improved with hyperparameter tuning to R² = 95.8%, MAE = 0.065, and RMSE = 0.112. Compared to the baseline, tuning reduced RMSE by 12.6% and MAE by 10.8%. The optimized model reliably reproduced experimental ZT trends in doped Bi₂Te₂Se and Ag₂Te, validating its predictive capacity. Our findings show that hyperparameter tuning is greatly recommended for high-fidelity predictions in thermoelectrics.

• Periodical: Applied Mechanics and Materials (Volume 933)
• Pages: 155-175
• DOI: https://doi.org/10.4028/p-2TYjM0
• Online since: February 2026
• Authors: Eli Usheunepa Yunana*, Jesse Usheujawa Yunana, Mustapha Abdullahi, Ibrahim Mohammed Inuwa
• Keywords: Chalcogenides, Figure of Merit, Hyperparameter, Machine Learning, Thermoelectric

References

[1] Zada R., Ali Z. and Mehmood S., Optoelectronic, elastic and thermoelectric properties of the perovskites (Sr3N)Sb and (Sr3N)Bi. Materials Science in Semiconductor processing 147, 106734 (2022).

DOI: 10.1016/j.mssp.2022.106734

[2] Zoui M.A, Bentouba S., Velauthapillai D., Zioui N. and Bourouis M., Design and characterization of a novel finned tubular thermoelectric generator for waste heat recovery. Energy 253, 124083 (2022).

DOI: 10.1016/j.energy.2022.124083

[3] Hicks L.D., Harman T.C., Sun X., and Dresselhaus M.S., Experimental Studies of the effect of quantum well structures on the thermoelectric figure of merit, Physical Review B 53, R10493–R10496 (1996).

DOI: 10.1103/PhysRevB.53.R10493

[4] Balandin A.A., Review: Nanophononics: phonon engineering in nanostructures and nanodevices, Journal of Nanoscience and Nanotechnology 5, (2005) 1015–1022.

DOI: 10.1166/jnn.2005.124

[5] Li J.F., Liu W.S., Zhao L.D. and Zhou M., High-performance nanostructured thermoelectric materials. Asia Materials, 2, (2010) 152–158.

DOI: 10.1038/asiamat.2010.24

[6] Hicks L.D and Dresselhaus M.S., Effect of Quantum well structures on the thermoelectric figure of merit, Physical Review B 47, 12727 (1993).

DOI: 10.1103/PhysRevB.47.12727

[7] Katyal P., Rathi M., Mehra P. and Panwar A.K., Evaluation of Figure of merit of Thermoelectric materials using Machine Learning. International Journal of Advanced Science and Technology 9, (2020) 2858-2863.

[8] Na, G.S., Jang, S. & Chang, H., Predicting thermoelectric properties from chemical formula with explicitly identifying dopant effects. Computational Materials, 7,106 (2021).

DOI: 10.1038/s41524-021-00564-y

[9] Sheng, Y., Wu, Y., Yang, J., Lu W., Villas P. and Zhang W., Active learning for the power factor prediction in diamond-like thermoelectric materials. Computational Materials 6, 171(2020).

DOI: 10.1038/s41524-020-00439-8

[10] Shuaibu A., Abdullahi Y.Z, Tanko Y.A, Olusola O.P and Kafayat O.A., Thermoelectric Transport Properties of Ternaries Mixed Chalcogenide Bi2Te2 X (X= S and Se): A First Principle study. Physics Memoir 1, (2019) 67-87.

[11] Gan, Y., Wang, G., Zhou, J. and Sun Z., Prediction of thermoelectric performance for layered IV-V-VI semiconductors by high-throughput ab initio calculations and machine learning. Computational Materials 7, (2021) 176.

DOI: 10.1038/s41524-021-00645-y

[12] Kim, D.H., Kim, H.S., Rahman, J.U., Shin H.W., Kim T. and Kim S., Thermoelectric transport properties of S-doped In0.9Si0.1Se. Journal of the Korean Ceramic Society, 59, (2022) 64–69.

DOI: 10.1007/s43207-021-00153-1

[13] Olayinka A.S., Nwankwo W. and Olayinka T.C., Model based Machine Learning Approach to Predict Thermoelectric Figure of Merit, Archive of Science & Technology 1, (2020) 55 - 67.

[14] Wang T., Zhang C., Snoussi H. and Zhang G., Review: Machine Learning Approaches for Thermoelectric Materials Research. Advanced Functional Materials 30, 1906041 (2020).

DOI: 10.1002/adfm.201906041

[15] Yang Z., Sheng Y., Zhu C., Ni J., Zhu Z., Xi J., Zhang W. and Yang J., Accurate and explainable machine learning for the power factors of diamond-like thermoelectric materials. Journal of Materiomics 8, (2022) 633 - 639.

DOI: 10.1016/j.jmat.2021.09.008

[16] Parse, N., & Pinitsoontorn, S.. Machine learning for predicting ZT values of high-performance thermoelectric materials in mid-temperature range. APL Materials, 11(8), 085101 (2023).

DOI: 10.1063/5.0160055

[17] Gaultois, M. W., Sparks, T. D., Borg, C. K. H., Seshadri, R., Bonificio, W. D., and Clarke, D.R., Web-based machine learning models for real-time screening of thermoelectric materials properties. APL Materials, 4(5), (2016) 053213.

DOI: 10.1063/1.4945682

[18] Wang, H., Zhou, Y., Zhang, H., & Kanatzidis, M. G., Interpretable machine learning models for thermoelectric material screening using experimental datasets. ACS Applied Materials and Interfaces, 16(18), (2024) 22401–22413.

DOI: 10.1021/acsami.4c19149

[19] Chen, M., Li, D., & Zhang, G., Machine learning prediction of ZT values in Bi₂Te₃–ₓSeₓ solid solutions using experimental and compositional features. Journal of Materials Chemistry C, Advance Article (2024).

DOI: 10.1039/D4TC01058B

[20] Xu, W., Zhao, Y., Zhang, X., & Liu, Q., Prediction of thermoelectric figure of merit based on autoencoder and LightGBM. Journal of Applied Physics, 135(7), 074901 (2024).

DOI: 10.1063/5.0186324

[21] Jiang, T., Zhou, C., Ren, J., & Liang, Q., Hybrid DFT–ML framework for screening chalcogenide thermoelectrics (2024). arXiv preprint arXiv:2405.02618. https://arxiv.org/abs/2405.02618.

[22] Zhang, Y., Liu, Z., Wu, T., Yang, J., and Wang, X., Enhanced thermoelectric performance in Cu₂₊ₓCd₁₋ₓSnSe₄ by controlled Cu substitution. Journal of Materials Chemistry A, 10(25), (2022) 13211–13220.

[23] Wang, Y., Li, J., Chen, J., Sun, F., Zhang, H., and Liu, W., High-performance Na- and I-doped PbTe thermoelectrics. ACS Applied Materials & Interfaces, 16(5), (2024) 11425–11434.

DOI: 10.1021/acsami.4c01042

[24] Li, W., Zhang, C., Zhao, H., Yuan, X., and Tang, X., Thermoelectric properties of Sn-doped Ag₂Se polycrystals. ACS Applied Materials & Interfaces, 16(8), (2024) 14021–14030.

DOI: 10.1021/acsami.4c04085

[25] Liu, R., Hu, L., Liu, Y., Yang, W., and He, Y., Enhanced performance of Ba-doped BiCuSeO bulk materials. Journal of Materials Chemistry C, 12(4), (2024) 987–995.

[26] Chen, J., Zhao, Y., Wang, D., Li, Q., and Xu, X., Band engineering and thermoelectric enhancement in Sb-doped GeTe. Journal of Physics D: Applied Physics, 56(10), 105502 (2023).

[27] Shao, H., Zeng, X., Luo, Y., Huang, Y., and Lin, S., Enhanced thermoelectric performance in Se-doped AgSbTe₂. Materials Today Physics, 24, (2022) 100683.

DOI: 10.1016/j.mtphys.2022.100683

[28] Tan, G., Zhao, L. D., He, J., and Kanatzidis, M. G., Iodine doping in SnSe for thermoelectric performance optimization. Acta Materialia, 255, 118941 (2023).

DOI: 10.1016/j.actamat.2023.118941

[29] Zhou, M., Chen, Y., Zhang, H., Wang, S., and Zhu, Y., Bismuth-doped Cu₂Se with low thermal conductivity and good TE efficiency. Journal of Alloys and Compounds, 960, 170097 (2023).

[30] Kim, H., Park, Y., Jeong, S., Lee, H., and Choi, M., Na-doped PbSe thermoelectric materials with reduced lattice thermal conductivity. npj Computational Materials, 9 (2023) 89.

[31] Park, J., Ahn, S., Kim, B., and Lee, D., Structural control of Zn-doped Bi₂Te₃ and its thermoelectric performance. Proceedings of Thermoelectrics International Conference, Paper No. 317 (2022).

[32] Gao, W., Li, Z., Zhang, M., Chen, J., and Wang, Y., Antimony-doped Mg₂Si with improved electrical transport. Journal of Materials Science, 58(14), (2023) 5573–5585.

[33] Xu, L., Jiang, Z., Wang, F., Liu, Q., and Yang, X., Thermoelectric properties of SnTe with controlled Sb doping. Solid State Communications, 378, 115149 (2023).

DOI: 10.1016/j.ssc.2023.115149

[34] Cheng, X., Luo, X., Zhao, C., Lin, D., and Yang, P., Thermoelectric transport in Zn-doped Cu₃SbSe₄. Energy Materials Advances, 1(3), (2022) 133–139.

[35] Cheng, W., Xie, T. and Ong, S.P., A comparative evaluation of machine learning models for predicting thermoelectric properties. npj Computational Materials, 5(1), 59 (2019).

[36] Ward, L., Agrawal A. Choudhary, A., andWolverton, C., A general purpose Machine Learning framework for predicting properties of inorganic materials. npj Computational Materials 4, 25 (2018).

DOI: 10.1038/npjcompumats.2016.28

[37] Xie, T. and Grossman, J.C., Heirachical graph neural networks for materials property predictions. Nature Communications, 11, 890 (2021).