Modalities
High-Performance Liquid Chromatography
High-performance liquid chromatography (HPLC) separates compounds in a liquid sample to help identify and measure them.
In materials research, it helps characterise additives, residual reactants and degradation products. A liquid carries the sample through a packed column, and a detector records a signal as compounds emerge.
AI Agent Potential
AI agents can make chromatographic analysis more consistent and reproducible. They can integrate peaks, compare calibration models and carry dilution factors through a batch, leaving the calculations open to inspection.
The crucial judgement is whether a signal supports a chemical claim. An overlapping compound can inflate a peak area while leaving the estimated concentration inside the calibration range. A plausible number does not establish that the intended compound was measured.
Dependable assistance connects the calculation to the laboratory procedure. An agent should check units and preparation steps, test whether calibration remains valid for new samples, and retain uncertain assignments. When evidence is incomplete, it should propose a measurement that could resolve the ambiguity.
Example Task Demo
Quantification includes deciding when a result is not ready to report. In this fictional coating assay, an agent uses a sample trace, a fresh reference injection of A and calibration standards to estimate compound A in the extract. All signals and calibration points are invented.
Fictional coating assay · synthetic signals
For retention drift, centre the interval on the fresh reference. Revealing constituents shows how the example was made, not extra evidence available to the agent.
1 Inspect and integrate
Total responseIntegration intervalFresh A referenceDashed: A (white), interferent (amber)
Fresh A reference: 5.00 minIntegration: 4.65–5.35 min
2 Compare with calibration
Invented standardsApparent estimate
Within the illustrated 2–16 mg/L calibration range. Range alone does not establish validity.
A candidate estimate
A separated peak permits a candidate estimate. Here, it lies within range; the agent would check identity, sample preparation and quality-control evidence before reporting.
Different ambiguities call for different checks. A fresh reference injection can help investigate retention drift; an overlap may require a more selective separation or complementary detection. Neither adjustment replaces evidence that sample preparation and calibration are suitable.
How the illustration works
The synthetic UV signals use Gaussian peaks on a constant baseline of 5 mAU, without noise. Compound A has a total area of 40 mAU·min. The ideal calibration has a slope of 5 mAU·min per mg/L, using standards from 2 to 16 mg/L. Integration subtracts the known baseline and assumes identical injection volumes and detector response.
The drift case moves both the sample peak and the fresh A reference from 5.00 to 5.48 min, keeping area and response unchanged. The overlap adds a second response at 5.12 min. The shaded integration interval stays 0.70 min wide. Revealed constituents are the signals used to construct the illustration, not information an agent could infer from one real UV trace.
The estimate describes the injected extract only. It does not account for extraction recovery, dilution or measurement uncertainty. The suggested checks illustrate analytical judgement, not a validated method or an automatic release decision.
This is purely an example task and is not part of GDPx.