Read an HPLC system suitability run

What these three injections can show

A system suitability test checks the measurement system and associated analytical operations against criteria chosen for the procedure. ICH Q14 places it within the control strategy for pharmaceutical analytical procedures. This exercise examines just one resolution check and a separate blank-review rule.

ICH Q14, §6

Each case is a retrospective, fictional excerpt: injection 01 is a solvent blank, 02 a two-component reference mixture and 03 a sample. Earlier injection history is not supplied. The recorded sample in case B is part of the exercise; it is not permission to continue a real sequence after a failed check.

All signals are noise-free sums of Gaussian components in arbitrary signal units (AU). A flat modeled blank means no components were included; it is not a measured “not detected” result. Sample component labels S1/S2 carry no chemical identity, purity or concentration claim.

Use the stated widths and rules

Here w_b is the tangent-base width, not the width at half height. For the supplied Gaussian components, w_b = 4σ. The curve’s tails never actually reach zero; the width comes from the baseline intersections of tangents at the two inflection points.

Rₛ = 2 × (tᵣ₂ − tᵣ₁) / (w_b1 + w_b2)

Use minutes for both retention times and widths. Both unequal widths belong in the denominator. Do not insert a half-height width into this baseline-width formula.

IUPAC: Rₛ · IUPAC: w_b

Teaching rule 1: the reference pair meets the resolution check when Rₛ ≥ 1.50, including equality. This number is a fictional exercise choice, not a general HPLC requirement.

Teaching rule 2: flag a modeled blank component whose apex lies from 3.80 to 4.20 min inclusive and whose height is ≥ 0.020 AU. This simplified component rule is not a detection limit or a real blank acceptance policy.

Compare the three sequence excerpts

Sequence A

01 · Solvent blank

02 · Reference mixture

03 · Sample

Reference centers: 4.00 / 5.60 min. Baseline widths: 0.80 / 1.20 min. No components were included in this modeled blank.

Sequence B

01 · Solvent blank

02 · Reference mixture

03 · Sample

Reference centers: 4.00 / 5.20 min. Baseline widths: 0.80 / 1.20 min. No components were included in this modeled blank.

Sequence C

01 · Solvent blank

02 · Reference mixture

03 · Sample

Reference centers: 4.00 / 5.60 min. Baseline widths: 0.80 / 1.20 min. Unknown blank component U1: center 4.02 min, height 0.060 AU, σ 0.20 min.

Synthetic total signal only; all nine traces use the same 0–8 min and 0–1.2 AU scales. Model parameters and numerical traces are downloadable below.

Reference inputs and blank observations — all synthetic

This table scrolls horizontally. Focus the table region and use the left and right arrow keys.

Sequence	tᵣ₁ / min	tᵣ₂ / min	w_b1 / min	w_b2 / min	Blank component
A	4.00	5.60	0.80	1.20	None modeled
B	4.00	5.20	0.80	1.20	None modeled
C	4.00	5.60	0.80	1.20	U1: 4.02 min; 0.060 AU

Write a decision for each case

Calculate Rₛ from the reference pair, keeping both baseline widths in the denominator. Compare it with the fictional ≥ 1.50 rule.
Assess the blank separately using the stated window and height rule. Record the observation without assigning a chemical identity or cause.
State which teaching check is met or unresolved. Then name evidence still missing before a real analytical result could be accepted.

In C, why can the blank signal not be attributed to the reference injection shown after it? What earlier records or targeted checks would help investigate it?

Worked decisions

Show the three calculations and decisions

A · The two specified checks are met

Rₛ = 2 × (5.60 − 4.00) / (0.80 + 1.20) = 1.60

The reference reaches the fictional 1.50 criterion. The model contains no blank component, so no component triggers the separate blank flag. These two teaching checks are met. The excerpt does not establish that all requirements of a real system suitability test, or all conditions for reporting this sample, were met.

B · The reference resolution is insufficient

Rₛ = 2 × (5.20 − 4.00) / (0.80 + 1.20) = 1.20

The reference gives Rₛ = 1.20, below 1.50. An unflagged blank and a visually plausible sample do not repair that failed reference check. The recorded sample result is not supported by the stated sequence checks. Document the failure and follow the applicable procedure to investigate and re-establish suitability; the model does not prescribe reinjection or release.

C · Resolution is met; the blank remains unresolved

Rₛ = 2 × (5.60 − 4.00) / (0.80 + 1.20) = 1.60

The reference gives Rₛ = 1.60, but the blank’s 4.02 min apex lies within 3.80–4.20 min and its 0.060 AU height exceeds 0.020 AU. Flag the unexplained blank signal and leave its impact on the sample open. Retention-time proximity does not identify the signal. A passing reference resolution does not resolve it.

Keep the cause open

The blank was recorded before the reference and sample shown here. Neither later injection can explain it as carryover from that later injection. Earlier history is missing. Prior carryover, reagents or preparation, and system contributions are investigation possibilities, not established causes. Review the actual sequence history, preparation records and method-defined blank controls; choose follow-up work from that evidence. Do not silently subtract this blank from the sample or turn the signal ratio into an impurity percentage.

Record the boundary of the conclusion

Only one reference injection per case is supplied. There is no repeatability series, sensitivity or noise test, tailing assessment, full bracketing sequence, validated sample matrix or reporting range. Thus no case demonstrates complete system suitability, sample identity, purity or fitness for use. A real decision needs the current procedure and its complete criteria, observations and investigation record.

A useful training note records the sequence and injection IDs, width convention, calculation, criterion, separate blank observation, provisional decision and missing evidence. Keep the method revision and original traces with an actual review.

Download and reproduce the evidence

Download simulate.py and run the command below with Python 3.9 or newer. It needs no additional packages or network and writes all five CSVs into a new reproduced directory. If output already exists, it stops and preserves those files.

python3 simulate.py --output-dir reproduced

CSV files use English field names, commas and decimal points in every language. sequence.csv contains nine injection records; peak-parameters.csv lists the 13 modeled components. traces.csv contains 401 time points per injection, 3,609 rows in total. decisions.csv reports the two teaching checks separately and marks complete_sst_demonstrated as false in every case. Empty component lists are represented by modeled_components = 0 in sequence.csv.

For a group exercise, give learners the worksheet and input files before the decision key. The numerical traces let them compare visual appearance with the calculation. The three sample signals are modeled observations, not quantities from which a validated assay can be inferred.

Sources and reuse

Original synthetic model and teaching text: ChemVerify / Peak & Mass, created with AI assistance, 12 November 2026. CC BY 4.0; credit this publication and link to this guide. External references retain their own terms. These fictional examples and criteria are ours; no endorsement by ICH or IUPAC is implied.

CC BY 4.0