Developing a gradient method

[smartin]

Hi all,

I would like to know if somebody has a specific workflow on how to build a gradient elution.
Generally you have to start with 95:5 (A:B) for 60 min, and from the results obtained, you can start to build the gradient. But.... any well-structured workflow about it?

Hope you can help me!!
Thanks!!

[tom jupille]

Take a look at this article by John Dolan (accessible from the LC Troubleshooting Bible): http://www.lcresources.com/tsbible/18052000.pdf

Basically, you start with a full-range gradient with a gradient time scaled to give a k* of somewhere around 5. We have a spreadsheet that will estimate the appropriate time based on your column dimensions and flow rate: http://www.lcresources.com/resources/ex ... ents.2.xls (it's the second tab, "Select").


Enter the retention times of the first peak of interest and the last peak to elute (whether or not it's of interest) into the first tab of that spreadsheet to estimate isocratic vs. gradient conditions.


I'll put in a "commercial" here. We'll cover that and a lot more in our Advanced HPLC Method Development course running from September 14-30. Details are here: http://www.lcresources.com/training/tramd.html

[smartin]

Thank you very much!
But, what about if in this first gradient there are to peaks with nearly the same retention time?

[mattmullaney]

Hi Smartin,

Probably what I'd try first in the case where you've two co-eluting peaks is to change the slope of the gradient. In the example, the slope is about 5.3 %B per minute...one could halve this to about 2.5 %B per minute and see what happens. One important thing to remember is to allow sufficient time for all of the peaks within the sample time to elute, this may take some trial-and-error.

Please, see what you think and thank you.

[tom jupille]

I'll expand a bit on what mattmullaney said.

Once you have established a gradient range that elutes all your peaks and a steepness that gives you a reasonable k*, the rest of systematic method development consists of optimizing the selectivity (and then tweaking the efficiency if necessary). In that respect, gradient and isocratic behave the same way. There are six parameters that can change selectivity (move peaks around relative to one another):
- Solvent strength (isocratic) or gradient steepness (gradient). As suggested by mattmullaney, this is probably the first thing you should look at.
- Temperature
- Organic solvent type (acetonitrile / methanol / THF)
- pH (if your analytes have weak acid / base functionality
- buffer concentration
- column chemistry

In a pharmaceutical company, you would turn to "Quality by Design" or "Design of Experiments" ("QbD" or "DOE") to set up experimental matrices to evaluate the combined effects of the different paramters. In many cases it's easier and equally effective to use the "OFAT" ("One Factor At a Time") approach. Take your initial method and make a big change in each parameter in turn (changing only one parameter at a time!). If/when you see the peak spacing change, you focus on optimizing that parameter until you get the resolution you need. If you're lucky, the first parameter you look at (usually gradient steepness) will do the trick. If you're not lucky, you'll be in for a lot of work.

[lmh]

It's also often a good idea to pick the easiest things first. If you have coelution in acetonitrile on your first, wide, gradient, you will have near-coelution even if you make the gradient flatter, so fiddling with gradient-shape has only limited possibilities. Of the very helpful list that Tom gave, solvent is extremely easy to change, particularly if you're using a quaternary system or a binary system with solvent selection valves. It's probably no work at all to try methanol instead of acetonitrile.
pH is fiddly, especially with columns with limited pH stability ranges. After changing solvent, my usual approach would be a wildly different column(for example switch to something phenolic if a C18 column has failed you). Sometimes a different C18 will do the job, but I am suspicious of sales-reps who suggest that their C18 is magically better than everyone else's; often a change of C18 will merely convert a near co-elution at 5 minutes to a near co-elution at 6 minutes.
If you want to be logical, look at the two compounds that are nearly coeluting and see if there is anything feature that they do not share, and exploit it.