log k vs volume fraction organic plot

[DJ]

In polycratic RP-HPLC with acetonitrile as organic, the correlation coeff for plot of organic% fraction vs log k is best when fit with a quadratic function. Theoretical log k at 0% organic, (y intercept) is obtained as a regression coefficient in either (linear, quadratic) fit. What is the "meaning" of the other regression coefficients in quadratic fit?

[tom jupille]

The linear slope ("S-value") can be interpreted as the number of strong solvent molecules required to displace one analyte molecule from the stationary phase. In that sense, it's analogous to the Z-value in ion exchange. S values tend to cluster around 5 for small molecules and increase roughly in proportion to the square root of the analyte molecular weight.

I'm not aware of a physical interpretation of the quadratic coefficient; I've always lumped it under "secondary effects"