Barak Ratzker

Abstract The discovery of high-performance multicomponent alloys is constrained by the vastness of composition space and the scarcity of experimentally validated data. We develop VIBANN, a variational information bottleneck-augmented attention-based neural network framework, for uncertainty-aware inverse design of exceptionally hard bulk multicomponent metallic glasses.

1 Introduction Steel production and primary ironmaking remain among the most significant industrial contributors to anthropogenic CO2 emissions [1, 2]. Hydrogen-based direct reduction has therefore emerged as a leading decarbonization route, replacing carbonaceous reductants with H2, yielding H2O rather than CO2 as the byproduct [3-5].

MXenes are 2D transition-metal carbides, nitrides, or carbonitrides, typically synthesized by HF- or molten salt etching of a parent MAX phase metallic A-layer, followed by intercalation (e.g., alkali ions like Li+ or Na+, or organic molecules), and exfoliation. (1) During synthesis, MXenes surfaces acquire an abundance of termination groups (e.g., −O, −OH, −F, −Cl) with their chemistry depending on the synthesis process.