I would like to compare separation on different columns (reversed phase, cyano, diol, etc.), even in different chromatographic modes. I want to concentrate on the effect of different chemistries, different adsorption mechanisms for my separation rather than on the effects caused by packing quality. How should I practically compensate for the differencies in quality of packing (for the same particle size), differencies in plate numbers? Any reference available? Thanks.

If I read your question right, what you want to do is to focus on retention and selectivity rather than efficiency (right?). If that's the case, then just measure the k values (and, by extension, the alphas) for your analytes, and ignore the plate numbers.

I am doing separation of distereomers. Well, because my study involves SFC, I still would like to be able to evaluate differencies in efficiencies using same columns in SFC and HPLC modes as well as selectivities for different columns in one mode. What I want to exclude is the factor of good/bad packing. I am thinking of comparing results after compensating for differencies in N. I am considering of using N that is provided with the column rather than getting N experimentally with my own (not "ideal") compounds. Another option that I see is to use the formula Rs=1.18(t2-t1)/(W0.5,1 + W0.5,2) which does not include any N and (may be) still can be used for comparison during the separation of not base-line resolved pairs. Your opinion?

If you want to characterize selectivity, it would seem that the best bet would be the alpha values for selected pairs of peaks. This leaves the efficiency part out entirely.

If you want to characterize efficiency while ignoring variations in how well packed the column is, the situation is more complex. Any value that you calculate directly that involved width will necessarily be dependent on the quality of the column packing. The standard was of getting around this is to evaluation the C-term of a Knox (van Deemter) plot for your analytes. The catch is that you can't do a single measurement per system.

I think you have a good grounding in this, but for any other readers, this is a plot of the reduced plate height as a function of the reduced velocity. At sufficiently high velocity, this is close to linear; the slope of the plot in this region is the "C-term" of the overall relationship (the A- and B-terms are neglible at high flow velocity). The C-term essentially quantifies the effects of mass transfer kinetics on peak broadening (the A-term is eddy diffusion, and is influenced by how well packed the column is, and the B-term is molecular diffusion and is influenced mostly by diffusivity in the mobile phase; this leaves the C-term as the part most profoundly influenced by column chemistry).

-- Tom Jupille / LC Resources