Chromatography Forum

Anal. Chem. 1997, 69, 2575-2581
Determination of Lipophilicity by Gradient Elution High-Performance Liquid Chromatography

This paper reports linear relationship with high correlation (R^2 > 0.97) between gradient retention times and isocratic logk' for a solutes analyzed on an ODS column. However, the test solute set was small, and composed of compounds that I would consider "relatively similar." Would you expect this correlation to deteriorate with increasing structural diversity of the solute set?

Solute set:
n-propylbenzene
p-xylene
ethylbenzene
toluene
anisole
benzene
2,4,6-trinitrotoluene
nitrobenzene
p-dinitrobenzene
benzaldehyde
benzyl alcohol
phenol
aniline

To clarify, by "increasing structural diversity" I meant including additional compounds of similar (ballpark) size (MW< 600) but from different chemical classes eg., steroids, organic bases, carboxylic acids (all analyzed as neutral compounds).

That's a large, economy-sized can of worms. It basically takes a whole book to answer:
High-Performance Gradient Elution: The Practical Application of the Linear-Solvent-Strength Model; Lloyd Snyder and John Dolan

The upshot is that it depends on the S-values of the compounds in question (S is the slope of the log(k') vs. fraction-strong-solvent relationship). The more similar the S values, the more likely that correlation is to hold. And, yes, structurally similar compounds are more likely to have similar S-values.

I was hoping to hear from you, Tom.

Let me ask you this- If a plot of S vs logKw is, for sake of discussion, perfectly linear R=1 for a group of solutes, then would you say that, not only would selectivity never change regardless of gradient steepness, but also that retention times would be linearly related between two gradients of different tG?

I wish I could find this reference. I'll look for it, but in the mean time you can take my word for it

For a traditional ODS column, with what authors would call "a chemically-diverse set of neutral small molecules" the linearity of logKw vs S is usually not so hot. However, this correlation improves (often dramatically) when 0.25% octanol is included in the mobile phase.

More modern ODS with polar embedded groups tend to have much better linearity between logkw and S to start with. However inclusion of 0.25% octanol improves this relationship slightly.

How would you interpret this?