Section 03 · The Bench E-15
Integration
Where the operator decides a peak starts and stops.
The boundary the operator draws
A chromatogram does not arrive pre-labelled. The detector produces a continuous signal — voltage rising and falling as compounds elute from the column — and it is the software, guided by the operator, that decides where each peak begins, where it ends, and therefore how much area falls inside.
That decision is integration. The area under a peak, bounded left and right by a baseline the operator either accepts or redraws, is what gets compared against the calibration curve and converted into a mass. Everything upstream — sampling, desorption, separation — reduces to this number.
The difficulty is that peaks are rarely textbook-clean. Baselines drift. Two compounds elute close together and their peaks merge into a shoulder. Noise mimics small signals at the edges. Software handles these cases with rules — minimum peak width, peak-to-valley ratios for co-eluting pairs, baseline anchoring algorithms — but every ruleset involves a choice, and choices can be made poorly or inconsistently.
For a well-resolved peak on a flat baseline, integration is almost mechanical and two operators will agree closely. For a tailing peak, a merged doublet or a peak sitting on a rising solvent front, the outcome depends heavily on how the baseline is drawn. Moving the start or end point by a second or two of retention time can shift the reported area by ten percent or more, and that propagates directly into the final concentration.
This is why integration parameters — the settings used — should be fixed in the method and applied uniformly, not adjusted run-by-run to produce a tidier result. When a laboratory audit examines raw chromatograms alongside reported values, one of the first things a reviewer checks is whether the baseline was placed consistently or whether it migrated toward a preferred answer. A reintegrated chromatogram that was not flagged in the run record is a chain-of-custody problem, not just a technical one.
The peak is real. The boundary is a human decision.
Filed inSection 03 — The Bench