Purity angle vs purity threshold: what PDA peak purity tells you, and what it can't

A peak purity result in Empower is two numbers: a purity angle and a purity threshold. If the angle is below the threshold, the peak passes, and most reports stop there.

That pass is a narrower statement than it looks. It says the UV spectra across the peak are the same shape, within what noise and solvent can explain. It does not say only one compound is under the peak.

Waters is clear about this in its own Empower tips: “Peak Purity is spectral purity or spectral homogeneity. Peak Purity is not chemical purity” (Empower tip #251). This piece is about what that difference means in practice, and how to set the method up so the result means something.

How PDA peak purity works

A photodiode array detector records a full UV spectrum many times per second. Across a single peak, you get a series of spectra from the front, the apex and the tail.

If only one compound is eluting, those spectra have the same shape. Their size changes with concentration, but their shape does not. If a second compound with a different spectrum is co-eluting, the shape changes across the peak, because the mixture changes from front to tail.

Empower measures that shape difference with the spectral contrast algorithm (tip #252):

1. Each spectrum is baseline-corrected.
2. Each spectrum is turned into a vector, one dimension per wavelength.
3. The vectors are normalised to the same length, which removes concentration.
4. The angle between two vectors is the spectral contrast angle.

An angle of 0° means identical shape. An angle of 90° means no overlap at all.

Purity angle and purity threshold

The purity angle summarises the contrast angles between the spectra across the peak and the apex spectrum.

The purity threshold is the angle you would expect from non-ideal effects alone, with one pure compound. Detector noise, photometric error at high absorbance, and changes in solvent composition or pH during a gradient all change a spectrum a little, even for a pure peak (tip #252). In Empower the threshold is usually the sum of a noise angle, from a noise interval you choose, and a solvent angle (tip #255).

So the comparison is:

• purity angle < purity threshold: the spectral differences can be explained by noise and solvent. No evidence of a co-eluting compound.
• purity angle > purity threshold: the differences are larger than noise and solvent explain. Evidence of a spectrally different co-elutant.

Notice the wording in the first line. “No evidence” is what Waters writes, and it is the accurate phrase.

What a passing result cannot see

The spectral contrast angle depends on how different the two spectra are. Waters gives three examples in tip #252:

• two very different compounds: 53°
• two structurally related compounds: 10°
• two compounds that differ only by a CH₂ group: 0.5°

Half a degree is inside the noise of most measurements. The PDA getting started guide shows the same compound at two concentrations giving 3.4° from noise alone (ACQUITY UPLC PDA Detector Getting Started Guide, 71500108703, p. 6-8).

So a homologue, an isomer, or a degradant that kept the chromophore can sit under your main peak and pass peak purity. These are the impurities a stability-indicating method most needs to separate.

There is a second blind spot. A co-eluting compound that does not absorb in your wavelength range adds nothing to the spectra, so it cannot change their shape.

Keep the peak below 1 AU

This is the most common reason a pure peak fails.

Above about 1 AU, photometric error produces departures from Beer’s law of roughly 1%. Waters notes this has a negligible effect on quantitation, but it distorts the spectrum, and spectral distortion is exactly what peak purity measures. Their recommendation is that “the maximum spectral absorbance of a compound should be less than 1.0 AU” (tip #251; PDA guide, p. 6-9).

Two practical points follow:

• Check the absorbance of the main peak in the MaxPlot, not only at your quantitation wavelength. The maximum may be at another wavelength.
• Mobile-phase absorbance eats into that range. If your buffer absorbs at the low end and is autozeroed out, the usable range for the analyte is smaller (tip #251).

If your assay concentration puts the main peak above 1 AU, a purity failure on that peak tells you very little. Inject a diluted sample for the purity assessment.

Set the method up so the threshold means something

Peak purity is decided at acquisition as much as at processing. Waters’ recommended settings:

In the instrument method (tip #251)

• Resolution: 1.2 nm, for the best spectral resolution.
• Sampling rate: at least 12 spectra across the narrowest peak of interest; 15 to 20 is optimum.
• Wavelength range: start above the UV cutoff of the mobile phase and cover all the analyte’s absorbance.

In the processing method (tip #254)

• Wavelength limit: the same logic, set from the MaxPlot and the spectra of the peaks of interest.
• Noise interval: a stretch of baseline with no significant absorbance, 2 to 4 peak widths at half height long. It must contain at least 12 spectra, or Empower will not calculate peak purity and posts a message in the Message Center.
• Check the noise spectrum. It should look random. If it looks like your analyte’s spectrum, the interval is not baseline.

The threshold (tips #255 and #256)

• Try AutoThreshold first. Validate it with six injections of the standard: if the purity angle is below the threshold for every peak in every injection, it works.
• AutoThreshold then applies to unknowns only while their maximum absorbance is below 1.0 AU and below five times that of the standard used to set it up.
• If any standard injection fails, set the solvent angle yourself: process the six injections with the noise threshold only, take the highest purity angle reported, and enter it as the solvent angle.

The Active Peak Region setting can exclude the noisy baseline spectra at the very start and end of a peak, which is useful when the baseline is noisy (tip #255).

Read the purity plot as well as the numbers

The purity angle is an average over the whole peak. A small co-elutant in the tail can be diluted by all the clean spectra near the apex.

Waters’ own review example shows this (tip #257). Two caffeine samples had a purity angle below the threshold, so they passed on the numbers. But in the purity plot, the purity angle line rose above the threshold line in the tail. And the caffeine purity angle was about 3 times larger than that of another peak at similar absorbance, which is a second warning sign.

So before you accept a pass:

1. Look at the purity plot across the whole peak, especially the tail and the front.
2. Compare the purity angle with other peaks of similar absorbance in the same run.
3. If the peak fails, use Purity Passes to estimate how many spectrally different components are under it (tip #258).

Retention time is not identity either

The same detector is often used to confirm identity, and the same caution applies in the other direction.

In a Waters application note on vitamins and caffeine in drinks, peaks at the retention times of vitamins B12 and B9 were not B12 or B9 when their spectra were checked. A peak at the caffeine retention time was caffeine, and it was not on the label (Waters application note 720003188, pp. 4–5). Retention time alone would have reported two vitamins that were not there.

A pass is one piece of evidence

ICH Q2(R2) describes specificity in terms of absence of interference and, where impurities cannot be obtained, comparison with an orthogonal procedure that uses a different separation or measurement principle (ICH Q2(R2), section 3.1.1.2). Waters gives the same advice at the end of its peak purity series: during method development, use an orthogonal technique, a different separation mode or at least a different column, to add to the evidence (tip #257).

That is how I would treat PDA peak purity. It is fast, it is already in your data, and it catches co-elutants with different spectra. For the impurities that matter most in a stability-indicating method, the ones that look like the drug, it needs help from a second column, a second mode, or a mass detector.

Sources

• Waters Empower Tips #251–#259 (peak purity series), Neil Lander, Waters Corporation.
• Waters. ACQUITY UPLC PDA Detector Getting Started Guide. 71500108703, Rev A, chapter 6 (spectral contrast theory).
• Waters application note 720003188, water-soluble vitamins, caffeine and dyes by ACQUITY UPLC with PDA detection.
• ICH Q2(R2), Validation of analytical procedures, 2023 (error-corrected 2025), section 3.1.

If you have a peak that passes purity and still looks wrong to you, send me the purity plot on LinkedIn and I will tell you what I would check next. The validation-readiness checklist covers specificity and the other questions to settle before validation.