Tungsten exhibits exceptional thermal, mechanical, and radiation-resistant properties, making it well-suited for extreme environments applications such as nuclear fusion reactors. Additive manufacturing offers geometrical design freedom and rapid prototyping capabilities for these applications, surpassing the limitations of traditional powder metallurgy manufacturing. However, tungsten has low ductility at room temperature, which presents significant challenges for manufacturing and practical use. Alloying tungsten with refractory metals has the potential to enhance its ductility while preserving its high-temperature properties. Ab initio methods, such as density functional theory (DFT), are commonly used to compute elastic properties as indicators of ductility, guiding experimental alloy design. Machine learning interatomic potentials (MLIPs) have emerged as generalizable and efficient surrogates for DFT, enabling rapid screening across a vast compositional and elemental space. While MLIPs are typically assessed based on their ability to predict the energies and forces in atomistic simulations, we evaluate their accuracy in predicting relevant experimentally elastic properties, such as shear and bulk moduli.
In this work, we assess the accuracy of MLIPs using a dataset from the Materials Project, comprising 12,000 structures, including unary, binary, ternary, and higher-order alloys. Our analysis shows that the accuracy depends on the smoothness of the potential energy surface (PES) constructed by the MLIP. We find that the eSEN MLIP, which was designed to have a smoothly-varying PES, has the lowest errors with 5.6 and 7.6 GPa on the bulk and shear moduli, respectively. Finally, we assess the Pugh ratio (the ratio of the shear modulus and bulk modulus) as a proxy for printability and formation of cracks in the additively-manufactured tungsten-tantalum-niobium alloys.
