χ
XANES NotebookAarush Verma · Materials science notebook
Not connected
Materials science · Spectral experiments

Exploring X-ray absorption spectra

Dear Prof. Mingda Li, my name is Aarush Verma. It was a pleasure meeting you, and I was fascinated by your work using AI in materials science.
I made this notebook to explore how noise and small changes in XANES spectra affect a model’s predictions about topological materials.

Aarush Verma · Student research notebook
How to use this notebook. Choose a silicon K-edge spectrum from the list, then adjust the noise, energy shift, or intensity and click Apply changes to compare it with the original. Click Run stress test to see how the model’s predicted probabilities change as more noise is added.

Spectrum explorer

Original and experimentally perturbed XANES curves

Awaiting data
Original spectrum Modified spectrum—

Model prediction

Current pilot model: silicon K-edge only

Not evaluated
Predictions are displayed only for spectra supported by the saved model.
SM = Topological semimetal  ·  TI = Topological insulator  ·  Trivial = Material not classified as topological by the reference labels.
The bars show the model’s estimated probabilities for the original spectrum, not its measured accuracy.
Probabilities are not calibrated confidence. The small pilot test set limits conclusions.

Noise sensitivity

Repeat perturbations of the selected spectrum and observe prediction stability

This experiment measures how predictions vary for one spectrum. It does not measure classifier accuracy.
Blue: SM — topological semimetalOrange: TI — topological insulatorGreen: Trivial — non-topological classification
Horizontal axis: amount of added noise. Vertical axis: average model probability from repeated tests. A line moving up means the model favors that category more.
Select a silicon K-edge spectrum and run a test.

About this spectrum

Material and spectrum details

Select a spectrum to see its details.