Sign in Subscribe
Thin films

Machine Learning Breakthrough in Material Science: Quantitative XPS Analysis for Heteroatoms

Strategic Insights

— 1 min read

Researchers at National Taiwan University have made a significant breakthrough in material science by applying machine learning to quantitative X-ray Photoelectron Spectroscopy (XPS) analysis for heteroatoms. The study found that XPS spectra can be influenced by lattice distortions caused by alloyed heteroatoms, and developed an Artificial Neural Network (ANN) machine learning model to quantify the concentration of heteroatoms. The model was able to precisely predict the concentration of heteroatoms in HfO2 samples, with a high degree of accuracy.

Key Takeaways:

  • The study aimed to identify the correlation between XPS spectral features and lattice distortion, specifically in terms of alloying element concentration.
  • The ANN machine learning model was able to quantify the concentration of heteroatoms (F, La, and N) in HfO2 samples with a high degree of accuracy.
  • The model was able to precisely predict the concentration of heteroatoms as long as the element of interest was within its training process.
  • The study used the SHapley Additive exPlanation (SHAP) method to characterize the model, which suggests that the characterization mechanism is varied from element to element.
  • The study provides a novel approach in complex material characterization, and the success of the prediction provides a proof-of-concept for the application of machine learning in material science.
  • The research was supported by the National Science and Technology Council and the Center for Emergent Materials and Advanced Devices at National Taiwan University.
  • The study has the potential to be applied in various fields, including electronics, energy, and biomedical applications.

Statistics:

  • The study used an Artificial Neural Network (ANN) machine learning model to quantify the concentration of heteroatoms.
  • The model was trained on a dataset of XPS spectra and was able to achieve a high degree of accuracy (>90%) in predicting the concentration of heteroatoms.
  • The model was able to precisely predict the concentration of heteroatoms as long as the element of interest was within its training process.
  • The study characterized the model using the SHapley Additive exPlanation (SHAP) method, which suggests that the characterization mechanism is varied from element to element.

Sources:

  • Surfaces and Interfaces, 2025;72.
  • Elsevier, Radarweg 29, 1043 Nx Amsterdam, Netherlands.
  • Hsiu-Wei Cheng, National Taiwan University, Dept. of Chemistry, Taipei, Taiwan.

Read more