Peak Purity: What a Diode Array Detector Can and Cannot Prove
Start with the case that matters most and that the method handles worst. Two peptides separated only by a deleted non-aromatic residue, or by deamidation, or by a D-to-L inversion, carry essentially the same ultraviolet spectrum. Their chromophores are identical. Put both under one chromatographic peak and every spectral purity test in the software will report a clean pass. The impurities most likely to hide beneath a peptide peak are, awkwardly, the impurities most likely to share its spectrum.
That is the boundary of the technique. Everything else about it makes sense once the boundary is fixed in view.
The extra dimension the hardware supplies
Absorbance at one wavelength as a peak passes is all a single-wavelength detector gives you: a curve. Disperse the light across an array instead and the instrument captures many wavelengths simultaneously, writing a complete ultraviolet spectrum many times per second for the entire width of the peak. The peak stops being a curve and becomes a stack of spectra. That stack is the raw material for the analysis, and the detector was already collecting it.
The reasoning applied to the stack
One compound has one spectral shape. Concentration changes the height of that spectrum, never its shape. It follows that a peak made of a single compound should present the same normalized shape at its leading edge, at its apex and at its tail.
Shape that drifts across the elution window means something additional is present and its share of the signal is changing as the peak passes. Shape that holds steady is consistent with one compound. Consistent with, note, rather than proof of.
How vendors express the result
Numbers from one implementation cannot be set against numbers from another. Three reporting styles predominate.
| Purity angle versus threshold angle | Spectra are handled as vectors and the angle between them computed. Alongside it the software derives a threshold, the angle noise alone would be expected to produce. When the purity angle falls under the threshold, the observed variation cannot be separated from noise. |
| Match or similarity factor | Typically scaled so that 1000, or 100, denotes a perfect match, accompanied by a nominated pass value. |
| Purity plot | Deviation drawn across the peak itself. This shows where within the peak trouble arises, which carries more information than any single figure. |
Conditions the test quietly depends on
Four requirements have to hold before either outcome means anything. Signal has to be sufficient, because noisy spectra at low absorbance drive the threshold angle upward until every peak passes; a pass earned on a small peak may carry no information whatsoever. The peak must not be overloaded either, since detector response turns non-linear at high absorbance, spectra distort around the apex, and a genuinely pure peak can be failed on that basis alone. Baseline stability matters for a related reason: solvent absorbance shifting through a gradient injects spectral variation of its own, which is what background correction exists to remove. Finally, a co-eluting species has to absorb at all. Anything lacking a chromophore is simply invisible here.
Asymmetric verdicts
Failure and success do not carry equal weight. A fail is strong evidence, provided signal and baseline conditions were met: more than one species occupies that peak.
A pass says far less. Its honest reading is that no spectrally distinguishable second component was detected, under those conditions, at that signal level. Nothing there demonstrates that the peak holds one compound, and a certificate that presents the pass as such a demonstration is claiming more than the method delivers.
Techniques that do settle the question
Separating on a different property is the direct route. Switch column chemistry, shift pH, or move from reversed phase to hydrophilic interaction chromatography, and pairs that the original method merged will often resolve. The wider issue is treated in co-elution and peptide purity.
Stronger again is mass spectrometric detection taken across the peak. Because it discriminates by mass rather than by spectrum, it catches precisely the deletion and truncation impurities that render the spectral test blind.
What a useful report looks like
When a peak purity result appears on a document, the informative version names the metric the software used, states both the value and the threshold it was judged against, and where possible reproduces the purity plot. A bare assertion that the peak was pure, with neither metric nor threshold attached, documents that the software was run. It does not document what the software found. Related reading: reading an HPLC chromatogram.
