Modalities
X-Ray Absorption Spectroscopy
X-ray absorption spectroscopy (XAS) probes the electronic state and local atomic environment of a chosen element. It measures how absorption changes as X-ray energy passes through an absorption edge.
XAS is used to study catalysts, battery electrodes and disordered materials. The near-edge region (XANES) is sensitive to electronic structure and local geometry. Oscillations above the edge (EXAFS) arise from interference between outgoing and scattered photoelectron waves, revealing information about neighbouring atoms. Measurements during operation can capture changes that disappear when a specimen is removed.
AI Agent Potential
AI agents can make spectral analysis more systematic by calibrating measurements, comparing physical models and checking predictions. Reliable assistance also requires recognising when another measurement is needed.
In one internal XAS investigation, an agent completed a substantial analysis but drew conclusions that failed under further measurements. The lesson is to test an interpretation beyond the data used to develop it.
Interpreting the Evidence
A good spectral match need not identify a unique chemical state. An edge can shift with oxidation state, bonding or coordination, while a mixture can resemble an intermediate state. Chemically appropriate standards help distinguish these possibilities.
Measurement conditions also shape the spectrum. In fluorescence measurements, absorption within the specimen can distort features. Detector nonlinearity and beam damage can alter results, even when repeated scans agree.
Structural parameters can be ambiguous too. In EXAFS, coordination and disorder can compensate for one another in a fit, and scattering phase shifts must be accounted for when inferring bond lengths from Fourier-transform peaks. Fitted uncertainties remain conditional on the chosen model.
Example Task Demo
A changed spectrum can reflect both material changes and energy drift. In this fictional task, an agent compares a catalyst before and after a gas change, using an unchanged reference recorded on the same energy axis.
Compare the alignment choices for these synthetic curves. The agent must check the reference before interpreting the remaining specimen difference and choosing its next measurement.
BeforeAfterDifference
1 Specimen, before alignment
2 Check the unchanged reference
Same correction as the specimen comparison below.
Estimated reference displacement: +1.20 eVApplied to both after curves: 1.20 eV left.
3 Reference-aligned specimen comparison
- Specimen mismatch
- 0.068
- Reference mismatch
- 0.000
Root-mean-square differences in edge-step units; not significance tests.
The reference aligns, but the specimen still differs. Chemical standards or structural models are needed to interpret the change.
Check for beam damage with a lower-dose repeat, then compare appropriate chemical standards.
Model and limits of the example
These synthetic curves use smooth steps and peaks, with no experimental or benchmark data. They illustrate analysis choices; no AI model runs in the demo.
Both alignment choices fit an energy shift and apply it to both after spectra. Axes stay fixed across comparisons.
The model assumes rigid energy drift, a stable reference and no noise. Exact overlap is possible here; real measurements require checks of normalisation, detector response, repeatability and beam damage.
This is purely an example task and is not part of GDPx.