Chromatographic principles and peak interpretation

A chromatogram records response against time

In column chromatography, constituents move with a mobile phase while interacting with a stationary phase. Repeated exchange or other retention interactions delay them differently. Diffusion, finite exchange and flow differences can broaden the bands. A detector records a response as those bands reach it; the response need not be equally sensitive to all constituents.

A chromatogram is therefore an observation of separation and detection together. Its horizontal coordinate may be elapsed time, while the vertical coordinate is a detector-specific response. Retention time, width, height and integrated area describe different features. None by itself supplies a unique molecular identity or complete composition.

A baseline and the chosen integration boundaries also matter. A mathematical zero-baseline Gaussian is useful for learning, but real drifting backgrounds, unresolved shoulders and asymmetric peaks require additional interpretation and validation.

A hold-up reference must be justified

Define t_M as the hold-up time for an appropriate verified unretained reference under the same time convention. Adjusted retention is t_R′=t_R−t_M, and retention factor k=t_R′/t_M. The first detected feature in a sample is not automatically such a reference: it may be retained, absent from the relevant phase volume or affected by a different delay.

An original constant-condition model gives t_M=1.10 min and t_R=5.50 min, with negligible or corrected extra-column delay. Then t_R′=4.40 min and k=4.00. The raw ratio t_R/t_M=5.00 equals 1+k; it is not k. A truly unretained constituent in this stipulated reference convention has t_R=t_M and k=0.

Retention factor here is a dimensionless chromatographic delay ratio, not a chemical reaction rate constant. Retention can support comparison with a suitable reference, but it changes with method conditions and is not a universal chemical fingerprint.

Resolution compares time separation with both widths

For two peaks A and B with t_R,B>t_R,A, define baseline-width resolution R_s=2(t_R,B−t_R,A)/(w_b,A+w_b,B). Both widths belong in the denominator. Replacing them with the width of only the later peak requires an approximation, such as sufficiently similar widths, that should be stated.

For exact Gaussian peaks, w_b=4σ_t and w_h=2√(2 ln 2)σ_t. The same resolution can then be written R_s=√(2 ln 2)(t_R,B−t_R,A)/(w_h,A+w_h,B). The often rounded coefficient 1.18 approximates √(2 ln 2); it is not an exact new identity. Tailing peaks need more than a Gaussian-width summary.

Invent retention times 4.80 and 5.40 min and tangent baseline widths 0.400 and 0.500 min. The time separation is 0.600 min and R_s=1.33333. The Gaussian half-height widths are 0.23548 and 0.29435 min, which reproduce that result using the exact coefficient. This chosen R_s value does not establish identity, purity or a universally acceptable separation.

Equal areas can arise from unequal heights and amounts

For an exact Gaussian response R(t)=h exp[−(t−t_R)²/(2σ_t²)] on a zero baseline, the complete area is hσ_t√(2π). Height h has response units U; integrated area has U·min when time is in minutes. A broader peak can have a lower height while retaining the same area.

Give symbolic A mean 3.00 min, σ_t=0.0800 min and h=10.0 U; give B mean 6.00 min, σ_t=0.160 min and h=5.00 U. Both complete mathematical areas are 2.00530262 U·min. Their heights differ twofold and their widths differ twofold in the opposite direction. The graph shows invented mathematical curves, not a measured chromatogram.

Now stipulate known exact linear zero-intercept area-response factors: 2.00 U·min mmol⁻¹ for A and 4.00 U·min mmol⁻¹ for B. Amount equals area divided by that constituent’s factor, giving 1.00265131 mmol A and 0.501325655 mmol B. Equal areas therefore do not imply equal amounts.

The displayed interval samples the curves; the quoted areas come from integrating the complete Gaussian over all time. These supplied factors omit calibration uncertainty, drift, matrix effects and detector changes. A real amount result needs the relevant calibrated response and appropriate integration, rather than assuming one common response for every peak.

One peak shape can conceal two inventories

Consider another original model with two known toy constituents that coelute exactly. Their common shape is g(t)=exp[−(t−3.00 min)²/(2(0.100 min)²)], with peak value one. Assign peak-height response coefficients 2.00 U mmol⁻¹ for A and 3.00 U mmol⁻¹ for B. The entire signal is R(t)=(2n_A+3n_B)g(t), with amounts expressed in mmol.

Inventory n_A=2.00 mmol and n_B=2.00 mmol gives coefficient 10.0 U. Inventory n_A=5.00 mmol and n_B=0 gives the same coefficient and every time point has the same signal. Sampling that same common shape more densely does not supply a second independent amount relation.

The indistinguishable signal is a consequence of the stated model, not a claim about every detector. An additional selective response, changed separation or other independent evidence might distinguish the inventories. A symmetrical single peak alone does not establish a pure constituent.

Interpret peaks within the analytical question

A retention match, narrow band and clean-looking baseline can each contribute evidence, but the target entity and matrix determine what is sufficient. Quantification needs suitable calibration and recovery; identity needs evidence that relevant alternatives cannot produce the same observations.

Treat integration settings and source-processing decisions as part of the result. Document the time reference, width convention, response units and unresolved alternatives. The next lesson examines how separation mechanisms and transport conditions change those observations.

Worked examples

Try it yourself

1. For the same reference convention, t_M=1.20 min and t_R=4.80 min. Find adjusted time and k.

   Show the answer and reasoning

   3.60 min and k=3.00.

   Subtract 1.20 from 4.80, then divide by the hold-up time.

2. Keep the time separation 0.600 min but assign both tangent widths 0.400 min. Find R_s.

   Show the answer and reasoning

   R_s=1.50.

   Use 2×0.600/(0.400+0.400); the changed summed width changes resolution.

3. Give one reason why the two equal areas in the plotted model represent different amounts, without changing either peak shape.

   Show the answer and reasoning

   The calibrated constituent-specific area-response factors are different.

   A requires 2.00 U·min per mmol and B 4.00 U·min per mmol, so the same area corresponds to twice as much A.

4. Why do extra time samples of the exactly shared g(t) fail to distinguish the two coeluting inventories?

   Show the answer and reasoning

   Every sample is another value of the same single coefficient 2n_A+3n_B.

   No new independent relation between the two amounts is introduced; one needs evidence with different dependence on them.

What to carry forward

• A chromatogram combines separation with a detector response.
• Justify the hold-up reference before subtracting it.
• Resolution includes both peak widths and a defined convention.
• Height, area and amount are distinct quantities.
• An apparent single peak can remain compositionally ambiguous.