DAD/PDA: Extracted and Reference wavelengths

If you gather full DAD/PDA spectra, then instead of extracting a wavelength during acquisition, you can do this digitally offline. Unless you’re standing next to the instrument and want to know peak shapes right away, there’s little value (and may be some harm) in extracting wavelengths right in the acquisition CDS. You can always do this offline after all spectra are gathered.

Problems with online wavelength extraction

First, the way your instrument sets up wavelength extraction isn’t obvious. From Waters and Agilent documentation it’s not clear if they acquire requested wavelength separately. Meaning, does DAD switch between gathering full-spectra and selected-wavelength? Or maybe it keeps acquiring the spectra and simply extracts the chromatogram afterwards in the software? If it’s the former, then extracting digitally afterwards will give slightly different values. Which is probably not a big deal, but it’s not perfect either.

A bigger problem with this is that when reading files by offline processing software, you don’t know if it displays whatever was acquired. And if that software does display the original trace, the results may differ if you do this again in the offline software. Because integration algorithms can be different between 2 software. For this reason if spectra are available, Peaksel will discard the original traces and will re-extract it itself. This way we ensure all chromatograms are extracted the same way and guarantee some consistency.

Reference wavelength - how does it work?

If you use a gradient elution, you may end up with a baseline drift. To compensate for this some scientists use a Reference Wavelength. The thinking is:

  1. If changing solvent composition has a global impact on all wavelengths (e.g. due to changes in refractive index), then it adds/subtracts xAU both at 254nm as well as at 300nm.
  2. Therefore if we subtract whatever the signal is at 300nm from 254nm, we correct for the drift and end up with a straight baseline.

There are problems with this thinking though: what if at some point a substance is eluted that absorbs at 300nm? Then by subtracting 300nm from 254nm we get an unjustifiable signal reduction. This is one of the reasons you may have dips on UV chromatograms. But if there’s no clear dip - you may just get a smaller peak area. So the data will lie and you wouldn’t know it.

Now let’s say you extract a 254nm - 300±50nm and discard spectra. Without the original spectra, how do you check if there weren’t true absorptions at 300nm? This is a reason you can’t discard spectra in case you subtract a Reference Wavelength. Or at least you should extract Reference Wavelength as a separate trace - in addition to the wavelength of interest.

Additionally, suppose there’s a range of wavelengths that aren’t absorbed by any substance. So you can subtract it. But how do you know that range? You’ll first have to extract Reference Wavelength to ensure there are no peaks. Would you want to do 2 injections then? Probably not.

These are the reasons it doesn't make sense to subtract online and always prefer post-processing. See also an article by HPLC EXPERT .

How Peaksel extracts UV chromatograms

Suppose you want to extract a 254±4nm, what Peaksel will do is:

  1. Sum intensities of all wavelengths between 250nm and 258nm. Note, that other software may integrate area (instead of summing) in the range instead.
  2. And divide by the number of points summed to get an average. So if your instrument gathered 250, 251, …, 258 nm, then Peaksel will divide by 9.

Why averaging? In general the reason you specify 254±4 instead of just 254nm is because you want to reduce noise. Since any measured signal always has random noise, sometimes we get a little bigger signal than the true value, sometimes we get a little smaller value. So by summing multiple values we cancel it out. Then we calculate the average to give an approximation to whatever was at 254nm.

Averaging also means that we most likely will have a lower signal. So the wider the bandwidth, the smaller the peaks get. This is because your λmax is usually the apex of the peak on a spectrum and the rest of the points are located lower. Well, unless there's an overlap with another spectral peak and a shoulder is formed. But in that case you're better off selecting a different wavelength anyway.

Note that the Total (aka TWC) signal is different. This one is just a summation without averaging. So they can’t be compared.

How Peaksel subtracts Reference Wavelength

A chromatogram 254±4 - 300±50 is calculated this way:

  1. Get an average extracted wavelength at 254±4
  2. Get an average of the Reference Wavelength at 300±50nm
  3. Subtract 2nd from the 1st