First-principles data for solid-solution Nb–Zr, Ta–Zr, and V–Zr binary alloys with body-centered-cubic structures

Abstract

We present three open-access datasets containing density functional theory (DFT) calculations for refractory solid-solution binary alloys niobium–zirconium (NbZr), tantalum–zirconium (TaZr), and vanadium–zirconium (VZr) in 128-atom body-centered-cubic (BCC) supercells. These datasets extend our previously published first-principles collections for NbTa, NbV, TaV, and NbTaV alloys and were generated with the same automated high-throughput workflow based on the Vienna Ab-Initio Simulation Package (VASP). For each of the three NbZr, TaZr, and VZr systems, we uniformly sample 31 chemical compositions across the full composition range in 4-atom increments. For every composition we construct 100 distinct disordered atomic configurations by randomizing the element assignments on the BCC lattice sites. All structures are fully relaxed, and the released data include complete input files, detailed VASP outputs for each relaxation stage, and derived physical quantities such as ground-state formation energy and root mean-squared displacement (RMSD) relative to an ideal BCC reference. These Zr-containing binaries are motivated by our earlier demonstration that graph neural networks (GNNs) trained on binary NbTa, NbV, and TaV data can interpolate formation enthalpies and structural descriptors in ternary NbTaV alloys while dramatically reducing the required number of ternary DFT calculations. The new NbZr, TaZr, and VZr datasets increase the diversity of available binary training data, in particular in systems with larger size mismatch and stronger lattice distortion due to Zr, and are intended to support the development of transferable machine learning (ML) interatomic models for accelerated exploration of high-dimensional multicomponent alloy spaces.