Machine learning-enabled prediction of oxide glasses’ dielectric constants via augmented data and physicochemical descriptors

Abstract

Precise tuning of dielectric constants (ε_r) in oxide glasses is critical for high-frequency devices in 5G/6G systems, where ε_r directly governs signal propagation efficiency. A machine learning framework combining data augmentation and physicochemical descriptor integration is developed to address data scarcity. Validated pseudo-labels are generated via ensemble learning, expanding the dataset from 1503 to 11,029 compositions without distributional shift. The XGBoost model trained on the augmented dataset achieved superior accuracy, with an R² of 0.96 and an MSE of 0.14. For prediction tasks on unseen data, it reduced the error rate by 48% compared to the non-augmented model and improved generalization performance by 43% over GlassNet. B₂O₃ and SiO₂ are identified as ε_r suppressors and BaO and TiO₂ as enhancers through SHAP analysis, aligning with network former/modifier roles. Cation-specific polarizabilities are derived via Clausius–Mossotti regression (R² = 0.909). Integration of physicochemical descriptors (coordination number and bond strength) enables transferable predictions for Y₂O₃ and La₂O₃ containing glasses, with mean deviation 2.46%–4.76%. Crucially, structural descriptors dominate polarizability with 69.9% feature importance, establishing network engineering as the optimal design paradigm. A data-driven pathway for rational dielectric glass development is thus established.

Keywords

Clausius–Mossotti model / dielectric properties / machine learning / oxide glass

References

[1] Ahmad I, Kumar T, Liyanage M, Okwuibe J, Ylianttila M, Gurtov A. Overview of 5G security challenges and solutions. IEEE Comm Stand Mag. 2018;2(1):36-43.

[2] Cai L, Wu J, Lamberson L, et al. Glass for 5G applications. Appl Phys Lett. 2021;119(8):082901.

[3] Chen S, Kang S. A tutorial on 5G and the progress in China. Frontiers Inf Technol Electronic Eng. 2018;19(3):309-321.

[4] Jang D, Kong NK, Choo H. Design of an on-glass 5G monopole antenna for a vehicle window glass. IEEE Access. 2021;9:152749-152755.

[5] Xiang H, Yao L, Chen J, Yang A, Yang H, Fang L. Microwave dielectric high-entropy ceramic li(Gd0.2Ho0.2Er0.2Yb0.2Lu0.2)GeO4 with stable temperature coefficient for low-temperature cofired ceramic technologies. J Mater Sci Technol. 2021;93:28-32.

[6] Qin J, Liu Z, Ma M, Li Y. Machine learning approaches for permittivity prediction and rational design of microwave dielectric ceramics. J Materiomics. 2021;7(6):1284-1293.

[7] Qin J, Liu Z, Ma M, Li Y. Optimizing and extending ion dielectric polarizability database for microwave frequencies using machine learning methods. Npj Comput Mater. 2023;9(1):132.

[8] Viegas JI, Thomas S, Gontijo RN, Righi A, Moreira RL, Dias A. Vibrational spectroscopy and intrinsic dielectric properties of Sr2RE8(SiO4)6O2 (RE = rare Earth) ceramics. Mater Res Bull. 2022;146:111616.

[9] Kadathala L, Park Y-O, Oh M-K, Han W-T, Kim BH. Analysis of the dielectric properties of alkali-free aluminoborosilicate glasses by considering the contributions of electronic and ionic polarizabilities in the GHz frequency range. Materials. 2024;17(6):1404.

[10] Lanagan MT, Cai L, Lamberson LA, Wu J, Streltsova E, Smith NJ. Dielectric polarizability of alkali and alkaline-earth modified silicate glasses at microwave frequency. Appl Phys Lett. 2020;116(22):222902.

[11] Gerace KS, Lanagan MT, Mauro JC. Dielectric polarizability of SiO₂ in niobiosilicate glasses. J Am Ceram Soc. 2023;106(8):4546-4553.

[12] Shannon RD. Dielectric polarizabilities of ions in oxides and fluorides. J Appl Phys. 1993;73(1):348-366.

[13] Chen L, Kim C, Batra R, et al. Frequency-dependent dielectric constant prediction of polymers using machine learning. Npj Comput Mater. 2020;6(1):61.

[14] Lin B-N, Shih Y-T. Temperature and frequency-dependent dielectric properties prediction of oxide glasses by machine learning. Ceram Int. 2024;51(16):S0272884224056359.

[15] Bhattoo R, Bishnoi S, Zaki M, Krishnan NMA. Understanding the compositional control on electrical, mechanical, optical, and physical properties of inorganic glasses with interpretable machine learning. Acta Mater. 2023;242:118439.

[16] Cassar DR. GlassNet: a multitask deep neural network for predicting many glass properties. Ceram Int. 2023;49(22):36013-36024.

[17] Kang Z, Wang L, Li X, et al. Interpretable machine learning accelerates development of high-specific modulus glass. Comput Mater Sci. 2025;246:113482.

[18] Deng B. Machine learning on density and elastic property of oxide glasses driven by large dataset. J Non-Cryst Solids. 2020;529:119768.

[19] Pedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: machine learning in python. J Mach Learn Res. 2011;12:2825-2830.

[20] Chen T, Guestrin C. XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM; 2016:785-794.

[21] Bergstra J, Yamins D, Cox DD. Making a science of model search: hyperparameter optimization in hundreds of dimensions for vision architectures. In: Proceedings of the 30th International Conference on International Conference on Machine Learning - Volume 28 (ICML'13). JMLR.org; 2013:I–115–I–123.

[22] Chicco D, Warrens MJ, Jurman G. The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation. PeerJ Comput Sci. 2021;7:e623.

[23] Jēkabsons G, Lavendels J, Sitikovs V. Model evaluation and selection in multiple nonlinear regression analysis. Math Model Anal. 2007;12(1):81-90.

[24] Shorten C, Khoshgoftaar TM. A survey on image data augmentation for deep learning. J Big Data. 2019;6(1):60.

[25] Goodfellow IJ, Pouget-Abadie J, Mirza M, et al. Generative adversarial nets. In: Ghahramani Z, Welling M, Cortes C, Lawrence N, Weinberger KQ, eds. Advances in Neural Information Processing Systems. Curran Associates, Inc.; 2014. https://proceedings.neurips.cc/paper_files/paper/2014/file/f033ed80deb0234979a61f95710dbe25-Paper.pdf

[26] Kingma DP, Welling M Auto-encoding variational bayes. 2013.

[27] Jimenez Rezende D, Mohamed S, Wierstra D. Stochastic backpropagation and approximate inference in deep generative models. In: Xing EP, Jebara T, eds. Proceedings of the 31st International Conference on Machine Learning. PMLR; 2014:1278-1286. https://proceedings.mlr.press/v32/rezende14.html

[28] Zhou Z-H. Ensemble Methods: Foundations and Algorithms. 0 ed. Chapman and Hall/CRC; 2012.

[29] Chapelle O, Scholkopf B, Zien A. Semi-supervised learning (Chapelle, O. et al., eds.; 2006) [book reviews]. IEEE Trans Neural Netw. 2009;20(3):542.

[30] Devlin J, Chang M-W, Lee K, Toutanova K. BERT: pre-training of deep bidirectional transformers for language understanding. In: Proceedings of the 2019 Conference of the North. Association for Computational Linguistics; 2019:4171-4186.

[31] Lundberg SM, Lee S-I. A unified approach to interpreting model predictions. In: Proceedings of the 31st International Conference on Neural Information Processing Systems. Curran Associates Inc.; 2017:4768-4777.

[32] Lundberg SM, Erion G, Chen H, et al. From local explanations to global understanding with explainable AI for trees. Nat Mach Intell. 2020;2(1):56-67.

[33] Rodriguez-Cano R, Clark NL, Mauro JC, Lanagan MT. Borosilicate glass with low dielectric loss and low permittivity for 5G/6G electronic packaging applications. AIP Adv. 2024;14(11):115105.

[34] Hsieh CH, Jain H, Kamitsos EI. Correlation between dielectric constant and chemical structure of sodium silicate glasses. J Appl Phys. 1996;80(3):1704-1712.

[35] Li B, Yue Z, Zhou J, Gui Z, Li L. Low dielectric constant borophosphosilicate glass–ceramics derived from sol–gel process. Mater Lett. 2002;54(1):25-29.

[36] Sheikholeslam SA, López-Zorrilla J, Manzano H, Pourtavakoli S, Ivanov A. Relationship between atomic structure, composition, and dielectric constant in zr–SiO₂ glasses. ACS Omega. 2021;6(43):28561-28568.

[37] Welch RS, Salrin TC, Greiner T, Bragatto CB, Mauro JC. Molecular dynamics simulations of magnesium aluminosilicate glass structure: high-coordinated alumina and oxygen tricluster formation. J Am Ceram Soc. 2024;107(4):2155-2171.

[38] Zhang X, Yue Y, Wu H. Effects of MgO/CaO on the structural, thermal and dielectric properties of aluminoborosilicate glasses. J Mater Sci Mater Electron. 2013;24(8):2755-2760.

[39] Mentel L. mendeleev—a python resource for properties of chemical elements, ions and isotopes. version 1.1.0. 2014. https://github.com/lmmentel/mendeleev

[40] Haynes WM, Lide DR, Bruno TJ, eds. CRC handbook of chemistry and physics. 97th ed. CRC Press; 2016.

[41] Sun K. Fundamental condition of glass formation*. J Am Ceram Soc. 1947;30(9):277-281.

RIGHTS & PERMISSIONS

2025 The Author(s). Materials Genome Engineering Advances published by Wiley-VCH GmbH on behalf of University of Science and Technology Beijing.