Inverse Design of MXenes for High-Capacity Energy Storage Materials Using Multi-Target Machine Learning
Sichao Li, Amanda S. Barnard
Abstract
There is significant interest in discovering high-capacity battery materials, prompting the investigation of the electrochemical energy storage potential of the two-dimensional early transition metal carbides known as MXenes. Predicting the relationship between the composition of a MXene and electrochemical properties is a focus of considerable research. In this paper we classify the specific MXene chemical formula using a new categorical descriptor and simultaneously predict multiple target electrochemical properties. We then invert the design challenge and predict the formula for MXenes based on a set of battery performance criteria. This approach involves a workflow that includes multi-target regression and multi-target classification, focusing on the physicochemical features most pertinent to battery design. The final inverse model recommends Li2M2C and Mg2M2C (M = Sc, Ti, Cr) as candidates for more focused research, based on desirable ranges of gravimetric capacity, voltage, and induced charge.