Hi folks, we have a problem that has us
stumped but I'm sure someone out there can
help us out.

We are analysing the asymmetric epoxidation
of chalcone on a Gilson system with a
chiralpak AD (9:1 hexane/EtOH) at 254nm
(Gilson UV/VIS-151 detector). We have tried to
calculate a correction factor for the difference
in UV absorption between chalcone and
chalcone epoxide but there is a problem.

Going by the UV spectra of the two
compounds (at the same concentration), at
254nm there is a ratio of 1:2.7 in the
absorbencies. However, when a 1:1 mixture of
the two compounds is injected the peak areas
are in a ratio of 1:1.5.

Are we missing something obvious? Can
anyone shed some light on why we are
seeing a difference?

Did you measure the UV absorbance in MP? I guess that the diluent changes the absorbance significantly (both lambda and max wavelength).

Best regards.

I would also use a syringe and fill the flowcell first with one compound then the other. The absorbances measured by the detector may not be the same as the absorbances measured by a stand alone UV Spectrophotmeter. You might want to look at the bandwidth, slitwidth, etc.

Other obvious things:

Make sure your basis for your absorbances is the same in both cases: i.e., moles or grams.

I take it your mobile phase is isocratic. If you happen to be using flow programming (Can a Gilson HPLC do flow programming?) you can expect the areas ratio to differ from the absorbance ratio by a factor involving the flow rates.

Real loss of an analyte due to irreversable adsorption, catalytic decomposition, dimerization, etc., will throw off the measurement.

Was there some particular reason for using a separate UV specctrophotometer for the absorbance data rather than just using external calibration by HPLC-UV in the first place?

It appears that you are comparing apples and apples. However, chromatography separates stuff, while an absorbance measurement includes all the stuff in your sample. If the chalcone epoxide contains a large amount of other stuff, it may show up in a different part of the chromatogram. For example, maybe the epoxide is not all epoxide any more.
I would trust the chromatography more than the simple UV measurement.

You may be seeing the effect of averaging of the absorbances of the wavelengths around your lambda max. In a LC UV detector, you have a bandwidth of probably 8 nm. The output you get at from the detector is the average of all of these wavelengths. In a spectrophotometer, you have a very narrow bandwidth and therefore it is probably a true absorbance.

Will! You are comparing apples and pears.
In UV spectrophotometer result is absorbency.
Peak area in LS is function of absorbency, and time of flowing the compound through detector cell. Area difference is coming from differences in peaks shape for both compounds.
Jan

I disagree with J Poskrobko. It is impossible to set up a data system that badly that you would get such a large difference in peak area.

Thanks for all the feed-back!

We shall have to think some more about how
we calculate the conversion from the HPLC
trace. As 254nm is not the lambda max for
chalcone or the epoxide, It is probably the
case that small variations in the wavelength
translate into a relatively large range of
absorbancies. We will be checking the
conversion factor with standard solutions on a
daily basis so as to see exactly how much it
varies.

I would like to ask to J Poskrobko exactly what
you meant in your responce. I probably
misunderstood you, but I would have thought
that the absorbance measured in a UV
specrotophotometer is dependant upon the
amount of material in the solution. The area
under a peak can ONLY be a function of the
amount of material and how strongly that
material absorbs UV light. This is regardless
of the peak shape. The more material there is,
the longer it will take to pass through the cell
and therefore the bigger the area under the
peak.

In responce to BF's question, we have been
using a correction factor of 2.1 since I joined
the group (as determined by a previous
member using UV spectroscopy), but a
collaborator in another university with an
identical HPLC system checked this value by
HPLC of standard solutions and this resulted
in a much lower result (1.5). We then decided
to carry out the calculation with the UV
spectrometer, and this gave yet another,
higher result (2.6). When we injected standerd
solutions on our HPLC machine the results
were closer to those of our collaborator, but
still much less than the values from the
UV spectrometer.

Apologies for such a long message, but the
initial inquiry seems to have generated much
interest so I would like to give as much
information as possible.

Thanks, Will

Will and Uve
Realy in UV output signal is absorbency as function of molar absorbency (at defined wavelength) and concentration. Result is stable, independent how long we measure.
In nowadays LC signal is converted from analog to digital. Measures of absorbency (dependant upon molar absorbency and concentration) are performed in small bits of time (defined by time constants of detector and analog/digital conversion parameters) and next INTEGRATED! Thus area is proportional to time of measuring. You can test it by yourself: if you decrease twice the flow of eluent, you will obtain area of peaks twice (approximately) higher.
Because of that difference, I told that we speak about two different measurands.
The best way to determine correction factor is to perform standard analyse in this same route (method, procedure and machine) as sample analysis.
Sorry my English is poor, although if any other questions or comments, I am ready to discuss.
Jan.

Will, try a simple mental stretch of the chromatography: If you stop flow until eternity, even during a tiny peak, your area will be infinite. See the time factor now?

Ups, didnīt see Janīs second message, anyway, if one compares two peaks in one run one should have the necessary constancy of factors that Jan seems to "request". Therefore, my guess would be the mentioned slit width as the main culprit if the specs of the two compunds have a different shape.

Jan, I understand you now! Thanks.

If a mixture of the two compounds in solution
is injected, run at 1mL/min and both
compounds come off giving two peaks,
shouldn't the time factor be the same for both
compounds? And therefore not an issue in
determining the difference in absorbance? Or
is it the case that small irregularities in the
flow rate are enough to put the measurements
out?

Will

Will!
Of course, typical small flow irregularities in the flow rate changes answer (peak area) negligibly. My example (including different peak shapes) was only to explain difference between absorbency in UV machine and peak area in LC machine. That I understand is fixed. I think we now do not discuss about difference in Your factors 2,6 or 2,1 obtained in UV and factor 1,5 obtained in LC.
Separate (and general) problem is (if existing)repeatability to factor (?) obtained in collaborate university HPLC laboratory.
Jan.

Jan, thanks for the reply. I think the correction
factors we are obtaining from the HPLC
experiments and those obtained in the same
way by our collaborators agree closely
enough. I think I understand now why the
values obtained from the UV
spectrophotometer are so different.

In hindsight we probably should have just
carried out the calibration measurements with
the HPLC machine and not involved the UV
spectrophotometer at all.

So, now we know! Thanks for all the advice.

Will

Jan, I don't get it. Of course, you get different peak areas when you change the flow rate. But we are not doing this here, the flow remains constant. For the response ratio, it makes no difference whether there is flow or no flow or what the flow is. One can get into difficulty only, if either the integration does not work, or if the number of data points per peak is getting into single digit numbers. Even in the last case it is difficult to get into trouble, since reduced sampling rates are usually accomplished by accumulating data, instead of true low sampling.

But since peaks have different retention time, they will have different width (and peak shape generaly, different sampling point number, etc.). All chromatographic measurements are RELATIVE, you have DETECTOR, but not SPECTROPHOTOMETER in your system, so shouldnīt expect full agreement in results.

Although the width will increase for the later
peaks the height will reduce because the size
of the peak is down to the amount of stuff and
how much UV it absorbs. What else can it be?

Rolandas, to do anything quantitative with HPLC you have to calibrate your system, after that you know the needed RELATIONS. The probable causes for discrepancies have already been mentioned above.

Hans

We are measuring peak area, and not peak height. Therefore there should not be a difference due to chromatographic retention.

Dear H W Mueller, you have to calibrate against standard or using some relation calculated from measurements on chromatographic system. This relation is valid just under particular set of conditions. It hardly can be related to measurements with other type of instrumentation.

Most HPLC UV detectors are standardized with respect to wavelength and absorbence.

Will,
Thank you for provoking the great discussion.

J.Poskrobko said it right. The rest of the discussion refer to other reasons for non-reproducability in chromatography or confusion between spectroscopy and chromatography.

UV detectors do not measure absorbance, they measure transmitted light. From transmittance in hardware or software, absorbance is computed. Imagine a compound at a certain concentration transmitting 50% of the light at 270nm and 60% of the light at 280nm. If the fall off is linear, the extinction coefficient is .86 of the value for a narrow bandwidth at 270nm. As the detector response changes less with concentration, noise will affect the precision more greatly also.

If you are using a Hg lamp with an interference filter, (very good wavelength accuracy, precision, and low bandwidth,) You may do as well as a dedicated spectrometer, so bandwidth may not be a consideration if the quality of the spectrometer is good (bandwidth =1)

In any case, do not switch between instruments, for establishing a parameter, and calibrate the instrument you are using it and a set of standards appropriate to the assay.

Jim

Assuming that we all are familiar with the Beer-Lambert-(Bouguet) law one does not need to go into that any further. The measuring principle behind a spectrometer and detector is the same, many people use spectrometers as detectors. The difference is that in chromatography you are at a flux (dynamic), but one can calibrate this to get at the concentration, necessary to calc. the absorption (A). What I am saying is that if you use a spectrometer to get at the absorption coefficient (a)the traditional way, you should be able to get (a) by running a chrom., through the spectrometer, provided that you keep the slit constant. Remember though, (a) in the literature are at lamda max. Or stated differently, especially if you measure absorbance off the lamda max, if you determine an (A)ratio (or (a) ratio) in a spectrometer, then change the slit so that your bandwidth is more than one wavelength you may get a different (A) or (a)ratio as well.
This is theoretical, for instance, if your substances are not pure you will have trouble with a traditional spectrometer determination, as Uwe mentioned.
If I am off here I guess I should go back to the books and do some penitence in sack and ashes.

Hans