Establishing Structure-Function Relationships of MXenes Using Machine Learning

Two-dimensional materials like MXenes have exquisitely tunable electronic properties leading to promising applications in energy storage/conversion and electromagnetic shielding. One such electronic property that underpins structure-function relationships is the work function. Experiments indicate that the work function can be altered across a wide range through changes to the MXene composition or altering the surface termination of MXenes. Recent advances have expanded the range of possible surface terminations to several p-block elements like O*, N*, F*, Cl* S* etc. MXenes also function as supports in low-dimensional heterostructures consisting of metal layers adsorbed on the MXene. Such heterostructures have distinctive electronic properties due to charge transfer between the metal and the MXene. The stability of these heterostructures is governed by the adhesion energy between the metal and the MXene. This metric of stability determines whether a metal sheet exists in a low-dimensional heterostructure or as 3D nanoparticles. The ever-expanding space of MXenes presents a new challenge for efficiently predicting properties like the work function and the adhesion energy.<br/><br/>We construct machine learning models that predict functional properties of MXenes like the work function and the adhesion energy of metal overlayers on MXenes. A feature space consisting of physico-chemical properties of the constituting elements is used. The feature space includes properties like the electronegativity, ionization potential, orbital radii, etc that are available in databases. Several regression models like ordinary linear regression, neural networks, and random forests are used. To lend physical interpretability to our non-linear models, we also employ symbolic transformers. We employ two independent techniques to identify the most important features that govern these structure-property relationships. First, we perform a sensitivity analysis using permutation feature importance. Next, we compute the occurrence probability of different features in the most accurate models.<br/><br/>A 15-feature neural network model emerges as the best performing model and can predict the work function of MXenes from the properties of constituting atoms with a training and testing error of 0.13 eV and 0.25 eV respectively. The sensitivity analyses illustrates that the work function is governed by properties of the surface termination of the MXene. Leveraging insights from our sensitivity analyses, we construct a simpler 5-feature neural network that predicts the work function of MXenes with errors of 0.28 eV, which is comparable to the best performing model. We implement a transfer learning approach that efficiently re-trains the neural network to compositions beyond the training set. Experimental measurements of the work function of Ti₃C₂ terminated with various p-block elements confirm the trends in the work function established through our model.<br/><br/>While non-linear models are necessary to predict the work function of MXenes, the adhesion energies of metal films on MXenes is robustly estimated using simple linear models. Key features that govern these adhesion energies include the electronegativity of the termination, the strain on the metal film, and the work function of the support. Model predicted adhesion energies are estimated with errors of 0.01-0.03 eV/A^2. These adhesion energies are used to assess the thermal and electrochemical stability of the MXene based heterostructures. We leverage this framework to identify thermally and electrochemically stable heterostructures from a space of 5000+ structures. Key trends are validated against experimental reports taken from the literature. Taken together, this work showcases the power of machine learning methods in predicting functional properties of MXenes using readily available physico-chemical properties as inputs.