I remember sometime ago, somewhere, that plates are determined when an LC or GC method is isocratic or isothermal. Is it meaningful to calculate plates for a peak eluting during an HPLC gradient? In such a case would the plate number be still useful as a gauge of column performance? Would it be better to term the plates calculated during a gradient as Apparent Plates?

No plate number is useless. Just add an isocratic hold at the begining of your gradient and your retention times and therfore plate numbers will increase.

You should be using peak capacity. see http://www.laboratorytalk.com/books/chem/chrom/rs_6/rs_6_62.html

The peak capacity can be calculated simply by taking the gradient time tg (which you know, because you programmed it) dividing it by the average peak width w. Add 1 to it, and you are done:

P = 1 + tg/w

Uwe:
The average peak width of which peak - the last peak, which would be the widest? Assume I am quantitating numerous impurities in a drug. The drug peak will be very large compared to the impurity peaks. Each impurity may or may not be present, so the last peak that is known to elute may not be there or may be very small. Would a separate injection of an authentic sample of the last impurity to elute be needed? IS there a web site which covers the equation you posted?

Tom:
The labtalk article seemed to be too theoretical and I did not see a practical equation like the one Uwe posted.

One other clarification. If I have a hold time of 5 minutes then the gradient begins then stops at 40 min, tg = 35 mins?
If the last eluting peak elutes around 35 minutes but the gradient contiues to 40 mins (to allow for column differences and more retention) why not calculate tg = 35-5 = 30 minutes?

The average peak width is the average from the widths of all peaks. There is no reason that the last peak should be the broadest (unless you use very flat gradients).
to the second posting: as long as beginning and end compositions are not given you can use any gradient time you want, just make sure you use for a given method always the same time. In your example 35 min would be rigth.

actually the time that you should proabably be using for peak capacity would be the difference between your last eluting analyte and an unretained compound. What it tells you then is the number of peaks you could fit in between your start and end point. as alex said, just be consistent and document.

The problem with peak capacity seems to be that you need to be able to see all the impurity peaks all the time. The only way around it would be to run a check sample containing authentic impurities. Plate count for the standard seems easier and less complicated.

For simplicity and consistency, I recommend to use the time over which the gradient is executed as your gradient time. Your calculation in your example is correct: you subtract the initial hold time and use 30 minutes for the gradient time.

By average peak width, I meant that you average the width of several peaks. I assume that you have several peask in the chromatogram that are clean enough to do a peak-width calculation. If you can't do it from the sample chromatogram du to overlap, you can still do it from the standards.

In gradients, most of the peaks have approximately the same width. If you look carefully, you may sometimes see a somewhat narrower width at the beginning of the gradient than later in the gradient, especially with flat gradients. This is why the averaging does a good job in compensating for such small difficulties.

Also, another clarification: by width, I mean the value of 4 standard deviations of the peak. This width can be obtained by drawing tangents on the peaks. Or you can take the width at half height, and multiply it by 1.7.

The general concept of peak capacity is a good way of measuring separation poer when other measures fail. I believe that it was invented by Giddings, and he uses versions of it in his "Unified Separation Science". I have outlined the simple concept above in an article in the Advances in Chromatography 41(Marcel Dekker) an in an article in the Journal of Separation Science 24 (2001), 921-929. There is a brief discussion in Colin Poole's "Chromatography Today", which may be accessible to many readers.

In my opinion, it is a terrific concept that can be used to sort out the generally optimal conditions for a gradient. It just has not been used much in the past, unfortunately.

Uwe:

I typically do not have several peaks in the chromatogram. The sample (parent drug) will normally only contain one impurity peak and even then the impurity will be close to the LOQ of the method. The standard is just a dilution of the parent drug and so is a single peak. In that case what should I do?

Thanks.

Since you have only one peak, you just take the peak width of this peak for the calculation. No need for averaging...

:-)