What a Calibration Curve Does in GC
In gas chromatography, a calibration curve is the mathematical link between what the detector sees and how much analyte is actually in the sample. By running standards with known concentrations, you build a reference that lets you convert peak area or height into a concentration for every unknown that follows. Without that curve, a chromatogram is just a pattern of peaks; with it, the data becomes quantitative.
More from this site
Keep reading the latest coverage
The curve is typically constructed as a plot of response (y-axis) against concentration (x-axis), and the resulting equation — often linear, sometimes quadratic or polynomial — travels with the analytical method into routine use. When the curve is sound, your reported concentrations carry defensible accuracy; when it is flawed, every result downstream inherits that error.
Preparing Standards and the Calibration Range
A reliable curve starts with standards prepared from a certified stock solution, diluted with the same matrix as the samples wherever possible. The concentration range should bracket the expected sample levels, with the lowest standard sitting near the limit of quantitation and the highest near the upper end of the detector's linear range. Spacing matters: equal logarithmic or arithmetic intervals make the regression easier to interpret and defend.
Each standard goes through the same injection sequence as samples, and replicates at key levels help reveal pipetting or injection variability. Matrix-matched blanks are critical because a contaminated solvent or vial can distort the intercept and inflate apparent concentrations at low levels.
Linearity, Regression, and the Meaning of R-Squared
Most GC quantitations assume a linear relationship, which is why a simple least-squares regression of y = mx + b is the starting point. The slope (m) reflects detector sensitivity, while the intercept (b) captures the background or blank response. Analysts often check R-squared as a quick quality flag, but R-squared alone is not enough; a high value can mask a poor intercept or a single outlier that bends the line.
To dig deeper, examine the residual plot — the differences between observed and predicted responses across the concentration range. Random scatter around zero suggests a good fit; a curve in the residuals hints at non-linearity, heteroscedasticity, or an interferent that grows with concentration. When the response flattens at the top of the range, a quadratic or weighted regression may be more appropriate than forcing a straight line.
Weighting and Non-Linear Models
In GC, detector response often scales with concentration at low levels but shows constant absolute error at high levels, a pattern called heteroscedasticity. Applying a weighting factor — commonly 1/x or 1/x² — gives less influence to high-concentration points and stabilises the curve across the range. The choice of weighting should be documented in the method and justified by the residual behaviour.
When the detector response is inherently curved, a polynomial or multi-point calibration can capture that shape, but complexity brings risk: overfitting to noise can make the curve look perfect in the range tested yet fail badly for samples that fall outside it. Whenever a non-linear model is used, carry the calibration through to verify that unknown results land where the model is reliable.
Quality Checks That Protect the Curve
Quality control is not an afterthought; it is built into every calibration run. A continuing calibration blank checks for carryover and contamination, while a continuing calibration verification (CCV) at a mid-range concentration confirms that the response has not drifted since the initial calibration. If the CCV falls outside a predefined acceptance window — often ±10% or ±20%, depending on the regulatory framework — the run is suspect and may require re-calibration.
Internal standards add another layer of defence by correcting for injection volume variation and matrix effects. By monitoring the response ratio of analyte to internal standard rather than absolute peak area, you can spot problems that a single-point calibration would miss entirely.
Common Pitfalls and How to Avoid Them
The most frequent errors come from mismatched matrices, expired standards, or a calibration range that does not cover the sample concentration. Injecting a sample that falls above the upper limit of the curve forces dilution, which introduces uncertainty; conversely, samples below the lowest standard cannot be reliably quantified. Always note where an unknown sits relative to the calibrated range, and flag out-of-range results clearly.
Another subtle trap is ignoring detector saturation. Modern GC detectors like the mass spectrometer can handle high concentrations, but the ion source or electron multiplier can saturate, compressing the response and breaking linearity. Running a quick check of the highest standard's response against the expected slope catches this before it poisons a batch of samples.
When to Re-Calibrate
Calibration curves do not last forever. Changes in the column, liner, detector tuning, or mobile gas purity can shift the response factor. Regulatory methods typically require a fresh calibration at the start of a sequence, a mid-sequence check, and a new calibration after a defined number of injections or a set time interval. Following that schedule, and documenting every deviation, is what separates a defensible dataset from one that cannot stand up to review.