HPLC and GC in practice

Why it matters

Chromatography separates mixtures so each component can be measured on its own. High-performance liquid chromatography (HPLC) is the most used quantitative technique in pharmaceutical QC: assay, impurities, dissolution and stability testing all run on it. Gas chromatography (GC) does the same for volatile compounds: residual solvents, fuels, flavors, environmental pollutants. The basics of how a column separates were covered in lab-techniques-instrumentation. This lesson is about what an analyst does every day: read a chromatogram, check that the system is suitable, and integrate peaks correctly.

Scope. Instrument operation, solvents and gases are hazardous and are learned on the job or in a teaching lab with supervision. The activity uses simulated data only.

The instruments in one paragraph each

• HPLC: a pump pushes a liquid mobile phase at high pressure through a packed column. The sample is injected by an autosampler. Components interact differently with the stationary phase (most often C18, reversed phase) and leave the column at different retention times. A detector (UV-Vis diode array is the most common; also fluorescence, refractive index, or a mass spectrometer) records signal vs time.
• GC: the sample is vaporized in a hot inlet and carried by an inert gas (helium, hydrogen or nitrogen) through a long, thin capillary column inside an oven. Temperature programming speeds elution of high-boiling compounds. Common detectors: flame ionization (FID) for organics, electron capture for halogenated compounds, and mass spectrometry.

Reading a chromatogram

• Retention timeThe time a compound takes to leave the column; compared with a reference standard to identify a peak. (t_R): identifies a peak by comparison with a reference standard run under the same conditions. Retention time alone does not prove identity; a diode-array spectrum or MS adds confidence.
• Peak areaThe area between the signal and the baseline, proportional to the amount of compound and used for quantitation.: proportional to the amount of compound, which is why it is the basis of quantitation. Peak height is used sometimes, but area is more robust.
• ResolutionHow well two neighboring peaks are separated, Rs = 2 (t_R2 - t_R1) / (w1 + w2); 1.5 or more means baseline separation. (Rs): how well two neighboring peaks are separated: Rs = 2 (t_R2 - t_R1) / (w1 + w2), where w is the baseline peak width (about 4 standard deviations for a Gaussian peak). Rs of at least 1.5 means baseline separation and is a common requirement.
• Tailing factor: asymmetry. A peak with a long tail integrates less reliably; many methods require a tailing factor at most 2.
• Plate count (N): column efficiency. Higher N gives sharper peaks.

These go into system suitabilityChecks on a standard injection that must pass before any sample results are valid.: before samples are run, a standard is injected and Rs, tailing, plate count and the RSD of replicate injections must meet the method's criteria. If system suitability fails, no sample results are valid.

Peak integration

The chromatography data system (CDS) finds where each peak starts and ends, draws a baseline, and calculates the area between the signal and the baseline. Numerically this is the trapezoidal ruleSumming (y_i + y_i+1) / 2 x (t_i+1 - t_i) over adjacent points to approximate the area under a curve.: for adjacent points (t_i, y_i) and (t_i+1, y_i+1), the area slice is (y_i + y_i+1) / 2 x (t_i+1 - t_i), summed over the peak.

Baselines are where judgement comes in:

• Drop-line (perpendicular) split: two overlapping peaks are separated by a vertical line at the valley.
• Valley-to-valley: the baseline follows the valleys, which removes a small peak riding on a larger one.
• Tangent skim: a small peak on the tail of a large one is skimmed off with a tangent line.

The integration events (threshold, peak width, baseline type) are part of the validated method. Changing them per injection to get a desired result is manual integrationChanging integration settings for an injection; a data-integrity risk that must be justified, documented and reviewed., which regulators scrutinize closely as a data-integrity risk. Any manual integration must be justified, documented and reviewed.

Worked example: area percent

A simulated chromatogram has a constant baseline of 2 mAU and two Gaussian peaks:

For a Gaussian, area = height x sigma x sqrt(2 pi):

• A: 50 x 0.10 x 2.507 = 12.53 mAU x min
• B: 80 x 0.15 x 2.507 = 30.08 mAU x min
• Area % of A = 12.53 / (12.53 + 30.08) x 100 = 29.4 %
• Resolution: widths are 0.40 and 0.60 min, so Rs = 2 x 2.00 / 1.00 = 4.0 (well separated)

Area percent assumes every component gives the same detector response, which is rarely true. Real assays use external standard calibration (compare sample area with a reference standard's area) or an internal standard (ratio of analyte area to a compound added at a fixed amount), which corrects for injection-volume variation.

Common mistakes

• Identifying a peak by retention time alone when a co-eluting impurity is possible.
• Running samples after system suitability fails.
• Integrating with a baseline drawn through noise or a drifting baseline.
• Changing integration parameters for one injection without documentation.
• Reporting area percent as if it were a weight percent.

Quick recap

• HPLC for non-volatile compounds; GC for volatile ones. Both give peaks vs time.
• Retention time identifies (with a standard); area quantifies.
• System suitability (Rs, tailing, plates, replicate RSD) must pass before samples count.
• Integration is the trapezoidal sum above a baseline; integration rules are part of the validated method.

Try it

Open the Do tab. A NumPy integrator is written except for the trapezoidal sum.