How to read an HPLC chromatogram

1. The axes: signal vs time

The horizontal axis is time from the moment of injection, almost always in minutes. The vertical axis is the detector response, the units depend on the detector: mAU for a UV/DAD, RIU for a refractive-index detector (RID), relative fluorescence for an FLD, or ion counts for an MS. A flat stretch is the baseline (mobile phase only); every bump above it is a compound leaving the column.

2. Dead time (t₀): the starting line

The first thing to find is t₀, the dead time, the time an unretained species needs to travel from injector to detector. It marks the column's void volume (t₀ = V₀ / flow). Everything meaningful is measured from t₀, not from injection, because the seconds a molecule spends simply being carried through the void carry no separation information.

3. Retention time and retention factor: what a peak is

Each peak's retention time (t_R) is its identity under fixed conditions: same column, same mobile phase, same temperature and flow, and a given compound elutes at the same t_R. But t_R changes if you change flow or column length, so chromatographers prefer the dimensionless retention factor:

k = (t_R − t₀) / t₀

A good working range is roughly k = 2 to 10. Below ~1 the peak crowds the void and co-elutes with junk; above ~20 the run drags on and peaks broaden. Identification by retention time alone is presumptive: confirm with a standard, a spiked sample, a diode-array spectrum, or a mass.

4. Area vs height: how much is there

For quantitation, peak area is the gold standard: the integrated area is proportional to the amount of analyte that reached the detector. You convert area to concentration with a calibration curve (area vs known standards) or an internal standard. Peak height can work for very sharp, reproducible peaks or trace-level work, but it is more sensitive to tailing and column ageing, so area is safer in routine work.

Two cautions when reading the table your software prints: check that the integration baseline was drawn correctly (a dropped or tilted baseline silently changes area), and beware co-eluting shoulders hidden inside one reported peak.

5. Resolution: can you trust adjacent peaks?

Resolution (R_s) measures how cleanly two neighbouring peaks are separated, combining their spacing and their widths:

R_s = 1.18 × (t_R2 − t_R1) / (W_1,½ + W_2,½)

R_s ≥ 1.5 is baseline resolution: the two peaks are essentially fully separated and each integrates cleanly, which is the usual target. At R_s ≈ 1.0 you still see two peaks but they overlap enough to bias the areas. The critical pair, the closest-eluting pair in the run, is the one that decides whether a method passes.

6. Efficiency and symmetry: plate count and tailing

Plate count (N) is the column's efficiency, how narrow a peak is for its retention: N = 16 (t_R / W)² (or 5.54 (t_R / W_½)² at half height). Higher N means sharper peaks and more separating power. A column losing N over time is wearing out.

Tailing factor (T) is symmetry, measured at 5 % of peak height. A perfect Gaussian peak gives T = 1.0; most methods accept T ≤ 2.0. Above that, integration and resolution degrade. If your peaks tail, see why HPLC peaks tail and how to fix it.

7. A quick reading routine

1. Find t₀ (first disturbance / solvent front).
2. List the peaks by t_R and compute k for each.
3. Check the critical pair's R_s (≥ 1.5?).
4. Check shape: tailing factor ≤ 2, no hidden shoulders.
5. Quantify from area against your calibration.
6. Confirm identity with a standard, spectrum or mass.

Common things you will see (and what they mean)

• A tailing peak (slow return to baseline) → see the tailing guide.
• A negative peak → usually a refractive-index or detector-response mismatch, common with RID, not always a problem.
• Two peaks merging into one hump → low resolution; change selectivity (column, %B, pH, temperature). A resolution map finds the robust conditions.
• Drifting or wavy baseline → gradient artefact, contamination, temperature, or a detector equilibrating.

Reading a chromatogram is just answering five questions in order: where, how much, how separated, how sharp, how sure. Once those are second nature, the plot stops being a picture and becomes data.