Modalities
Powder X-Ray Diffraction
Powder X-ray diffraction (PXRD) probes the arrangement of atoms in crystalline materials. Its pattern of peaks helps identify phases and track structural change.
The technique connects processing to structure. Measurements during heating, for example, can reveal the formation of a new crystalline phase.
AI Agent Potential
AI agents could make diffraction analysis more efficient. Across a heating series, they could compare candidate structures and flag reflections that suggest a phase change.
The value lies in reliable interpretation. Useful assistance would check that lattice parameters are chemically plausible and that reported statistics agree with the fitted profiles.
Interpreting the Evidence
A close fit does not uniquely identify a structure. Overlapping reflections can support several models, so chemical plausibility and unexplained features in the difference curve need attention.
Refinement results depend on their assumptions. Formal uncertainties may exclude calibration error, while fractions normalised to crystalline phases cannot establish the amorphous content.
Peak widths combine several effects. Inferring coherent domain size, which can be smaller than particle size, requires accounting for microstrain and instrumental broadening.
Illustrative Example
An agent can test a proposed structure against the whole diffraction pattern. Here, it must decide which of two single-phase cubic models to retain and identify the reflections that justify its choice.
Both models match the first peak. Switch to the full pattern to compare missing and extra reflections; peak positions follow the two lattices, while heights are schematic.
Synthetic observationSelected modelDifference
Both predict the reflection near 36.8°. An agent needs the wider pattern before choosing between them.
A retained model remains a provisional assignment. An agent would also check calibration and sample chemistry, and consider other structures or mixtures.
How this illustration is constructed
The curves are generated for this illustration. The hypothetical monatomic lattices are face-centred cubic (A, a = 4.230 Å) and body-centred cubic (B, a ≈ 3.454 Å), chosen to share their first reflection at a wavelength of 1.5406 Å.
The observation is generated from A with a smooth background and a small deterministic perturbation. Both models use Gaussian peaks of 0.24° full width at half maximum, schematic heights and the same background. No parameters are fitted: this illustrates peak positions, not Rietveld refinement or phase quantification.
This is purely an example task and is not part of GDPx.