Designing high-entropy ceramics via incorporation of the bond-mechanical behavior correlation with the machine-learning methodology

Abstract

Although high-entropy ceramics (HECs) are greatly attractive because of their superior properties over conventional ceramics, there is a lack of reliable and effective design guidelines for producing HECs with the wished-for mechanical properties. The often-used trial-and-error testing approach or case-by-case calculations without clear design guidelines are ineffective and expensive. Here, we propose a machine-learning accelerated strategy to design HECs with the desired mechanical properties. Using rock-salt ceramics as representative examples, we demonstrate that their mechanical properties are determined synergistically by different types of bonds, and bond properties of multi-element ceramics can be weighted from those of the involved constituents. Machine-learning models are developed to describe the correlations between bond characteristics and macro-mechanical properties, which show good prediction accuracy, as verified by computational and experimental data. The strategy for the HEC design, developed based on bond-mechanical property correlations and machine-learning methodology, provides a low-cost, highly efficient, and reliable method for developing advanced ceramics with superior mechanical properties.