Tracking Out-Of-Equilibrium Growth of Bimetallic Nanoalloys Using Deep Learning and X-ray Scattering

Reconstructing the growth of metallic nanoparticles in real time remains a long-standing challenge, particularly for nanoalloy systems that exhibit complex atomic spatial arrangements and display structures and properties distinct from those of pure metal particles. Here, we present a methodology that integrates Molecular Dynamics and in silico wide-angle X-ray scattering signals, which are calculated using the Debye Scattering Equation, with a deep learning regressor, enabling a direct connection between experiments and data-driven structural interpretation. Atomic-scale descriptors of size and composition of bimetallic nanoparticles, as well as elemental ordering spanning from Janus-like to core-shell arrangements, previously accessible primarily through computational methods, are now extracted using X-ray scattering measured patterns as the sole input information. Applying this model, we reconstruct the growth mechanism of Ag-Co nanoparticles formed in ultra high vacuum by Co vapor deposition onto Ag seeds, monitored in situ and in real time by grazing incidence wide-angle X-ray scattering and occurring under strongly out-of-equilibrium conditions. This analysis provides evidence that the nanoparticle structural evolution is not governed solely by the amount of Co supplied but instead results from a kinetically controlled evolution pathway, in line with theoretical predictions. Overall, the developed framework provides a strategy that significantly improves the interpretation of nanoalloys growth processes in real time via X-ray scattering data that would otherwise be difficult or even impossible to analyze through conventional fitting approaches. As such, the presented method is inherently compatible with on-the-fly and high-throughput data collection strategies.