how to determine UV detector up limit

[ym3142]

We do not want to saturate UV detector but do not have resource to test the linearity for every analyte and but we do want to make use of the full scale of the uv detector. Do we have a simple way to judge when/udner what concentrate x inj vol the UV has been saturated?

Tom mentioned the peak could be flat top if saturated. Does this mean as long as no flat top peak, the detector is not saturated and under linear range giving accurate result such as for purity test by %area?

In my UPLC PDA, my peak is sharp at >3AU but the UV spectrum has spikes. I assume the spikes are due to the saturation, am I right?

Thanks

[Consumer Products Guy]

Running linearity is not that difficult, especially if you have an autosampler. I can't understand why you'd want to work so close to the upper limit of your detector either, there just doesn't seem to be good reason. I think 3AU is quite high, most use 1AU as approximate "ceiling" and either dilute samples or use a less-sensitve wavength if over that. You have UPLC and don't have resources to do it right?

[mbicking]

You could analyze calibration curves for linearity, using standard statistical tests (not r^2), and/or look at residuals plots.

You can also calculate response factors for all concentrations (RF = area/amount). This number should be a constant over the linear range. It will begin to decrease as you begin to saturate the detector. An aribitrary limit could be used (say, 5% change), or use a statistical test.

[zokitano]

Also, working at the upper end of the detector UV scale (>2 AU) has its own risks associated with the linearity between the analyte conc and the detector response.

Also the Beer's law, which defines the linear relationship between the analyte concentration and the absorbance, deviates or is non-applicable when working at the upper limit of UV detector (that assumes also higher concentration of the target analyte(s) in solution and the Beer law works only for diluted samples at chosen wavelength)

[unmgvar]

the upper linearity limit of a detector is something that your vendor should know. ask them.
checking for that in general is yes very simple.
inject different concentration/volumes of a compound and look at the ratio height response/concentration.

in general, for single UV of last generation models linearity can go up to 2500 mAU
before that 1500 mAU is for most; some could do reasonably at 1800 mAU.
PDA are 1500-1800 mAU these days for a decent linearity range.

remember that PDA and UV have diferrent optics and therefore different sensitivities. UV are more sensitive by factors of 6-7 most of the time so do not extrapolate a result you get with one detector to the second one

[HW Mueller]

zokitano, it seems that nonlinearity due to deviations from the Beer-Lambert law (like molecular association) do not necessarily go parallel to nonlinearity caused by blacking out (saturating) the detector.
Also, one can have a very strong background (baseline) absorption, a very low concentration of analyte may easily add enough absorption to cause a blackout.

ym3142, don´t you have a horrendously noisy baseline at 3AU?
Why would you want to do an analysis if you can´t do a linearity test of your analytes?? Unless they are tested how will anybody know whether there is not a deviation from the B-L law, even if one is far from saturating the detector?

[zokitano]

zokitano, it seems that nonlinearity due to deviations from the Beer-Lambert law (like molecular association) do not necessarily go parallel to nonlinearity caused by blacking out (saturating) the detector.
Also, one can have a very strong background (baseline) absorption, a very low concentration of analyte may easily add enough absorption to cause a blackout.

I agree with you Hans. But don't you think that if one observes such high background absorption that means he/she already have "enough" component(s) (besides the analyte of interest) which can in some way alter the physico-chemical properties of the analyte molecules and cause deviation from the Lambert-Beer's Law (molecular association, dissociation, influence of the increased ionic strength of the solution...)?

And, what will be the purpose (practically speaking) of such spectrometric analysis of that sample in which the matrix blacks out the analyte of interest? In this cases you surely cannot determine low(or lower) concentration of your analyte in the sample (due to the black-out effect)

The black-out effect here, in this posted problem (in this thread) is not a case. I personally think that one should avoid working at the upper limit of detector's response, just to be sure that he/she in that way will prevent systematic errors (due to the deviations from the Beer's Law) to occur.

Regards

[JGK]

As a' "Rule of Thumb", I prefer maximum absorbances for UV to be about 1000 mAU

[lmh]

For me, the bottom line is that you should never be working outside the range of your calibration curve. If you are, then it doesn't matter how well it passes any number of statistical tests, and whether its R-squared is 0.9999, you are quite possibly getting a bad answer.

If you have a calibration curve that is OK, and you are working well within its range, then it doesn't really matter what's going on with detector linearity. Your results should be as good as the standards.

It's also good to keep in mind that absorbance is log-10, which means that at 3 you have only 1/1000th of the light getting through. Since the actual light detector frequently isn't logarithmic, at the very least you may start to see big steps in what changes it can measure at this level. You may start to see the analogue-to-digital conversion process in action. As someone wrote, noise is likely to be horrible. It's not a good place to be working. If you can't dilute the sample (because you need to analyse other, small peaks, and need the results in a single run), consider finding a different wavelength.

[HW Mueller]

zokitano, definitely there is a better chance for association, etc., at higher concentrations, but it doesn´t have to be that way, it depends on the substances. If one gets a linear response, has acceptable precision and accuracy, one can go ahead with the analysis. Most likely it is not very robust, and sensitivity is probably lousy. I have done it, but I hate it.

[tom jupille]

Tom mentioned the peak could be flat top if saturated. Does this mean as long as no flat top peak, the detector is not saturated and under linear range giving accurate result such as for purity test by %area?

The problem is that you will see non-linearity well before you flat-top the peak.

While unmgvar's numbers are about right, I personally use the same "rule of thumb" cited by JGK: stay under 1 AU to to be safe. That buys some insurance in case of a high background absorbance.

Non-linearity from chemical causes (e.g., association) are a whole different story and can only be evaluated by plotting response factor versus sample size (or maximum absorbance), but that means you have to run individual calibration curves for each analyte