Understanding the Trend in Core–Shell Preferences\nfor Bimetallic Nanoclusters: A Machine Learning Approach

Abstract

Finding\nout the driving factors in core–shell preference\nof nanoscale binary metal alloys is important due to their ubiquitous\npresence in applications ranging from catalysis to biomedical. We\nconsider binary-alloyed metallic nanoparticles encompassing a vast\nrange of alkali, alkaline, basic, 3d, 4d, and 5d transition metals,\nand p-block metals and determine the core–shell preference\nby calculating the segregation energies of single-atom alloy clusters\nby density functional theory. Application of machine learning to this\nlarge database, built on features characterizing the constituents,\nleads to the identification of four key factors: (i) cohesive energy\ndifference, (ii) atomic radius difference, (iii) coordination number\ndifference, and (iv) magnetism, providing the core-to-shell preference\nof a given constituent. Interestingly, the relative importance of\none key feature over another is found to be decided by the metal type.\nOur analysis also predicts that, for very small and very large differences\nof cohesive energy of the constituents, instead of core–shell\nstructure, mixed and Janus structures are stabilized, respectively.\nOur exhaustive study will be useful in designing bimetallic nanoalloys\nwith specific chemical ordering of the constituent species.