Dielectric constant prediction of perovskite microwave dielectric ceramics via machine learning

With the development of communication technology, microwave dielectric ceramics are in increasingly urgent need. Perovskite ceramics, as a kind of microwave dielectric ceramics with large dielectric constant span, have broad application prospects. Predicting material properties before experiments can greatly accelerate the development of materials. Although the existing methods, including classical theory and density functional theory, are of practical use for dielectric constant prediction, unsatisfactory universality and predictability limit rational design of microwave dielectric ceramics. This work aims to develop an uncomplicated method to quickly predict the dielectric constant of perovskite ceramics. According to the element and content of the compound, the dielectric constant can be accurately predicted by our machine learning model. Moreover, the model provides prediction results that are consistent with the experiment, but are completely different from those calculated by C-M equation.

Automated summary

With the advancement of communication technology, the need for microwave dielectric ceramics has become increasingly urgent. Perovskite ceramics, with a wide range of dielectric constant, hold great potential for various applications. However, existing methods for predicting dielectric constant have limitations in universality and predictability, hindering the rational design of microwave dielectric ceramics. This research aims to develop a simple method using machine learning to accurately predict the dielectric constant of perovskite ceramics based on their elemental composition. The predicted results from our model align with experimental data, but differ significantly from those calculated using the C-M equation.