Oxide ceramics of A 2 M 3 O 12 family with negative and close-to-zero thermal expansion coefficients: Machine learning-based modeling of functional characteristics

Ceramic materials with negative and close -to -zero coefficients of thermal expansion may open the avenue to the technologies that have so far been constrained by physical limitations concerned with the thermal stress or with the insufficient structural stability. Two important characteristics of NTE materials which could be used for the evaluation of the possible area and limitations of the sphere of application for negative thermal expansion (NTE) materials are the linear thermal expansion coefficient and the transition temperature from monoclinic to orthorhombic phase. In this study, the machine learning methods were involved in the analysis of experimental data for NTE oxide ceramics of A 2 M 3 O 12 family (where M is Mo 6 + , W 6 + , V 5 + or P 5 + while A position may be accommodated by the wide range of metal cations). The models are characterized by the following statistical coefficients: the determination coefficient R 2 = 0.81 and prediction error RMSE = 1.170 for linear thermal expansion coefficient; the corresponding parameters for the phase transition temperature were assessed as 0.81 and 82.239, respectively. Ionic conductivity in this class of compounds has been discussed as a tandem functional characteristic, emphasizing the role of anharmonicity in both characteristics. The role of synthesis route and defect chemistry in NTE was analyzed. A conclusion on the expected enhancement of NTE resulted from the intentional introduction of cation A vacancies has been made. The principal possibility of combining two functional characteristics, - an ion conductor and a phase with negative thermal expansion, with some compromise in the characteristics of each of them is substantiated.