Gradient Curve

[pipettemonkey]

When would one use a non-linear (convex, concave) gradient slope?

[tom jupille]

If one is smart -- never. Curved gradients are extremely difficult to transfer from one brand of instrument to another. Multi-linear gradients can accomplish the same thing and are much easier to transfer.

The simplest (and most common) case involves the situation where you have a few low-level late eluters. A steeper segment can be added to sharpen the late ones after the "difficult" peaks have eluted.

More complicated cases use two- or three-segment gradients to improve resolution of "critical" peak pairs. In general, the first step is to optimize the linear gradient range and steepness. If selectivity changes as a function of steepness, you will then have two critical pairs, each pair having roughly equal resolution. The overall steepness will be a compromise between these pairs. If the two pairs are at opposite ends of the gradient, then a two-segment gradient, with each steepness optimized for one critical pair, can improve things. If the two pairs are close together, then an isocratic hold in the middle of the chromatogram can (sometimes) improve things. The optimization can be done fairly easily using one of the commercially available modeling programs (e.g.g, DryLab, ChromSword, ACD Chromatography Simulator), but is extremely time-consuming otherwise. My personal feeling is that it's more effective to optimize other selectivity parameters (e.g., temperature or solvent type).

[alwelch]

Concave and convex gradients came from the fact that the very first gradient programmers used resistors and capacitors to form an RC time constant ramp. The shape of the curve for a charging and discharging capacitor is concave and convex and that is where those curves came from.

Al Welch