PDA Peak Purity

[MestizoJoe]

Hi Everyone,

I am using Empower to determine peak purity.

What I am observing is when I scan the UV spectrum from 190-400nm the peak of interest is not pure. Scanning at 210-400nm shows the peak is pure.

My question is, can I just ignore the spectrum from 190-210? Or do I have something co-eluting with my peak of interest?

My guess is that the absorbance at that lower wavelength range varies disproportionately with respect to analyte concentration which causes the empower software to show the peak as impure. But, I don't KNOW that and I don't want to just guess. Anyone know where I can find information on this?

Thanks,

[aceto_81]

You can overlay your peak spectra of the top of the peak and the start, end and apex.
Then you can see where the spectrum is different.
If you are right, then there should be a lot of noise in your 190-210 nm range visible (which is possible of course).

Ace

[HW Mueller]

This could just be another example of the shortcomings of this sort of peak purity function. Maybe you didn´t zero the apparatus at the lower wavelength correctly such that the absorbance of the mobile phase interferes.
You can not just ignore unpleasant things, you have to find out the underlying cause. To determine purity with UV is either impossible or difficult in most cases.

[Uwe Neue]

Between 190 and 210 nm, you see all kinds of things that you do not see at higher wavelength, so the peak purity algorithm may actually be doing exactly what it is supposed to do. If you do not need to quantitate in this wavelength range but can do the work at a higher wavelength, your peak is pure at this higher wavelength, and you can do quantitation witout difficulties.

[Koen Hollebekkers]

I agree that it’s better to avoid working below 210 nm. Peak purity will start looking very weird because enormous high absorption al lower wavelength. In PDA the diodes can get saturated, giving weird spectra.

[tom jupille]

First of all, we have to realize the limitations of *any* PDA-based peak purity measurements. They will fail to detect an underlying peak (false negative) under four conditions:
1. The underlying peak has a very similar UV spectrum.
2. The underlying peak is very small.
3. The underlying peak has no UV absorbance over the wavelength range involved.
4. The underlying peak exactly coelutes with the peak of interest.

They will detect a non-existant underlying peak (false positive) only when the signal/noise ratio is low (very noisy baseline or very small peak of interest).

In this case (false positive?)Low-wavelength noise and/or saturation of the detector (as has been suggesting) is certainly a possibility. The other possibility is that you do, indeed have an underlying peak, which does not have a significant chromophore above 210 nm but *does* have an "end absorbance" below 210 nm (an aliphatic alcohol, for example).

What to do depends on your purpose. If its an assay (where you want to quantitate the main peak), then sitting at the absorbance maximum (presumably above 210 nm) would be fine. If you are looking for impurities, then there *is* a chance that you're missing something.

[HW Mueller]

Tom, I am not sure I understand all of what you mention, but your 4th "condition" is suspect. One runs spectra to check whether a substance with a different spec. coelutes.

[Alex Buske]

If the peaks coelute perfectly, the spectra at apex and on the sides are just additions of the spectra of both components and can not be distinguished.
So if the peaks coelute and their spectra are summing up in the same ratio at all points of the peak, a peak purity test won't work.
The only solution would be a comparison with a library.

Alex

[HW Mueller]

Alex, that´s exactly what I don´t see in Tom´point 4. If the specs of two overlappiing peaks are identical than specs don´t help. If the specs are completely different it will show even in a conventional spectrophotometer where the two compounds are both in the same solution in the same vial (if both compounds are at sufficient conc.).

[Alex Buske]

All the Peak Purity algorithms I know just compare spectra _within_ the peak. So perfectly coeluting peaks cannot be detected as two different compounds. In that case the LC is just as usefull as a spectrophotometer.
Comparing spectra with library spectra is usually not considered as peak purity testing. It can be used for identification Purposes.

Alex

[HW Mueller]

So, lets say you have two perfectly overlapping (coeluting peaks), one has a fairly sharp maximum at 250 nm, the other a fairly sharp max at 350 nm. One can clearly see this, again this is no different from just mixing the two substances and measuring them in a spectrophotometer. If the specs are not overlapping one can read off their concentrations without any algorithms. If your algorithms, or any other jazz, can´t handle that it is time to throw it out.
This appears so simple to me that I think we must be talking about different things?? For instance, what do you mean to say with "In that case the LC is just as usefull as a spectrophotometer."?

[tom jupille]

If the specs are not overlapping one can read off their concentrations without any algorithms. If your algorithms, or any other jazz, can´t handle that it is time to throw it out.

Very true -- assuming that you know those two compounds are there and know what their UV spectra look like.


In the absence of that knowledge, you will see a single homogeneous peak with the same spectrum across the board (the sum of the spectra of the coeluting peaks).

Don' get me wrong; I am not a great fan of peak purity measurements. They have their place, but they do tend to give us a false sense of security.

[HW Mueller]

We have actually discussed that many times, I am also one of those who are highly skeptical that this is used properly. Nevertheless, I have often used it, semiquantitatively, with my scanning detectors (Linear).
To understand some phenomena I always go to extreme examples. Here I used an example which is highly ideal. I assumed that one knows the spectrum of the analyte. Lets say it has a max at 250 nm, starts at 200 nm and ends at 300 nm. The "dirt" which coelutes completely has a spectrum with max at 350 nm, starts at 300 nm and ends at 400 nm.
Now if you have detectable amounts of both in the coeluting peak, your spec scan will show a max at 250 and another at 350, whether you know the spec. of the dirt or not, nothing homogeneous here. If you can not determine the absorption coefficient of the dirt separatly you can´t determine the purity of even this ideal case. However if you do know the coefficient and it is the same as that of the analyte, and the two spectral peaks are of the same hight you know immediatly that the analyte is actually only 50% of the chrom. peak, algorythms not needed.

[Alex Buske]

Peak purity testing is not about quantitation. Of course you can quantitate perfectly overlapping compounds, as long as they have different spectral properties and you KNOW these properties.

Peak purity testing as it is implemented in many programms should asses peak purity WITHOUT any further knowledge. It is so easy and convinient that it is often used without any knowledge at all.

When an inspector asks "why do you think this peak is pure" and you say "PDA peak purity testing" everthing should be ok. If you talk about heartcut 2D-LC she/he gets probably confused, wants to see the chromatograms and then asks why the solvent peaks are labelled as solvent peaks.

[MestizoJoe]

So, I took the UV spectrum from the PDA at several time points within the peak. What I found was variation in the spectrum. I guess this can be considered noise. Correct me if I am wrong.



Also, I injected gradually smaller volumes of sample (10, 5, 2, and 1uL) and saw the peak purity angle shrink below the purity threshold. This seems to be consistent with the idea that the sample has high absorbance in the range 190-210nm and that it may be saturating the detector.

Could one argue that this is actually a false conclusion for purity due to the overlapping peak is too dilute at these smaller injection volumes to be detected?

Am I looking too deeply into this?

This is USP standard which is being tested on a method similar to a compendial method. I am confident it is pure but I'd like to be able to experimentally prove that it is pure.