Modalities
Diffraction Qualification
Diffraction qualification uses crystal structure and composition to judge whether a material meets its specification.
For an incoming powder, that means establishing how much of the intended product is present and how evenly it is distributed through the lot. Identifying the right diffraction peaks is only the beginning.
AI Agent Potential
AI agents can coordinate calibration, compare crystal structures and repeat quantitative analysis across many specimens. Their value depends on testing whether a convincing fit supports the material decision.
In one simulated qualification study, an agent met the forecast criteria and made correct release decisions, yet overstated variation within a lot. A later check found that a different crystal-orientation model better explained the same observations and reduced the error. This suggests that the agent had partly mistaken orientation effects for differences in composition.
Measurement and Interpretation
Reliable interpretation separates instrument response from specimen behaviour. An error in detector geometry, for example, can appear to be a change in lattice spacing. Measurements of a suitable reference help establish the instrument’s angular calibration and contribution to peak width.
Composition must account for the whole powder. A product can dominate the crystalline phases while substantial noncrystalline material remains. Adding a known crystalline standard provides a mass reference, after correcting for the addition. Unaccounted mass still needs interpretation: omitted crystalline phases and systematic errors can also contribute.
Purity and uniformity require separate checks. A satisfactory average can conceal a poorly mixed portion. An agent must compare portions and test alternative physical explanations before treating an apparent difference as a material defect.
Example Task Demo
An agent must decide whether to release a powder lot or gather more evidence. Here, release requires at least 95% product by original mass and a range between portions no greater than 4 percentage points.
Choose a fictional lot, then follow the evidence from a crystalline-phase fit through a known standard addition to separate portion measurements. Each stage shows what the agent can conclude and what it should check next. The results are illustrative; the demo does not fit diffraction data.
Illustrative values for three fictional lots.
Establish product content
ProductOther crystalsNoncrystalline
Check variation within the lot
Weighted lot meanIndividual portion
Add 20 g of standard to 100 g of original powder. The illustrative refined product-to-standard mass ratio is 4.757. That gives 20 × 4.757 = 95.14 g of product per 100 g of original powder.
- Original-mass product
- 95.14% Pass · minimum 95%
- Portion range
- 11.00 pp Fail · maximum 4 pp
The mean passes, but the range is too large. The agent should hold the lot and report the lower product content in portion P3.
Assumptions and numerical values
Three homogeneous portions make up the entire lot, with mass weights of 50%, 30% and 20%. The composite sample represents those weights. Values are exact for this illustration; no measurement uncertainty is assigned.
| Portion | Lot weight | Product | Other crystals | Noncrystalline |
|---|---|---|---|---|
| P1 | 50% | 97.40% | 1.60% | 1.00% |
| P2 | 30% | 97.20% | 1.80% | 1.00% |
| P3 | 20% | 86.40% | 3.60% | 10.00% |
| Weighted lot | 100% | 95.14% | 2.06% | 2.80% |
The added standard is absent initially, fully crystalline, non-reactive and uniformly mixed. All crystalline phases are assumed correctly quantified. The displayed mass ratio is calculated from the invented composition, not from peak heights or an actual refinement.
The lot mean weights each portion by its mass. Product among crystals divides total product mass by total crystalline mass. The range is the largest minus the smallest original-mass product percentage, without weighting. It is expressed in percentage points (pp).
Equality passes both example limits. The purity requirement applies to the lot mean, not to each portion separately. Mass bars span 0–100%; the portion plot enlarges 80–100%. These are illustrative purchase conditions, not universal materials standards.
In practice, these fractions need calibrated measurements, defensible physical models and uncertainty estimates. A defensible sampling plan is also needed to estimate lot-wide variation from the measured portions.
This is purely an example task and is not part of GDPx.