Machine-Learning-Assisted X-ray Spectroscopy Decoding an Elemental Segregation Mechanism in Pt-Ru Binary Alloy...

Area of Science:

• Materials Science
• Nanotechnology
• Catalysis

Background:

• Elemental segregation in alloys dictates critical properties like catalytic performance and stability.
• Understanding nanoscale compositional heterogeneity is key for designing advanced alloy catalysts.
• Conventional X-ray absorption fine structure (XAFS) methods struggle with ensemble-averaging limitations.

Purpose of the Study:

• To develop a machine learning-assisted framework for interpreting X-ray absorption near-edge structure (XANES) spectra.
• To quantitatively deconvolve surface and bulk elemental signals in alloy nanoparticles.
• To investigate surface segregation in platinum-ruthenium (Pt-Ru) alloy nanoparticles.

Main Methods:

• Developed a cascade multiscale convolutional neural network to predict Pt coordination numbers from XANES spectra.
• Employed computational-experimental spectral deconvolution to separate surface and bulk XANES profiles.
• Validated the methodology using Pt-Ru alloy nanoparticles as a model system.

Main Results:

• Successfully extracted surface-specific XANES profiles.
• Confirmed significantly higher surface Pt-Pt coordination numbers compared to the bulk phase.
• Observed a reverse trend in Pt-Ru coordination, verifying Pt surface segregation.

Conclusions:

• The integrated machine learning, computational modeling, and XAFS approach enables atomistic-level decoding of physicochemical properties in nanostructured alloys.
• This methodology offers a novel way to analyze localized physicochemical variations within nanomaterials.
• Provides new paradigms for the rational design of advanced alloy catalysts.