Drylab retention simulation-ternary mobile phase

[Austin]

Hi, dear all,

In Drylab simulation software an isocratic ternay mobile phase mode is available, and I know the simulation analog is based on the linear logarithm of retention factor(k') versus organic modifier strength(B%) in binary mobile phase, but what's the analog is appropriate for ternary mobile phase mode?

And this simulation software supplies an approach for ternary mobile phase prediction: First, carry out two binary mobile phase run with different modifier strength(as Run 1 and Run 2) and produces the corresponding retention, then mixes the above two binary solvent mixtures with equvalent volume(1:1) to make a ternary mobile phase, which generates different retentions(Run 3), finally those peak parameters are inputted into the software and makes the Resolution-Retention Map, so what's idea leads to this approach? Information, if any, be appreciated greatly! Thanks in advance!:)

Best Regards

Austin

[Uwe Neue]

The underlying idea is that the retention and selectivity in mobile phases with different organic modifiers is different, and that at an intermediate composition, you get intermediate retention and selectivity.

It is one of the most powerful and rational approaches in method development.

[tom jupille]

To expand on Uwe's answer: if the intermediate results were a linear function of composition, you would only need two calibration data points.

Let's say you had found that the best results with acetonitrile occurred at 50/50 ACN/water, and the best results with methanol occurred at 60/40 methanol water. If the behavior were linear, you could safely predict that the retention times at 30/25/45 ACN/MeOH/H20 (i.e., the middle) would be the exact average.

But in real life, the behavior is non-linear. That's why you need an intermediate point (DryLab uses a quadratic fit for interpolation; hence a total of three calibration points).

Years ago, I suggested calling this approach "BOB" (Blending Optimum Binaries), but that never caught on

[Alex Buske]

[quote="Uwe Neue"]
It is one of the most powerful and rational approaches in method development.[/quote]

I am glad to read that, as it is also my opinion, at least for neutral compounds.

I would do however the scouting runs for Drylab in the gradient mode (doesn't have to be 5-95% organic, could also be e.g. 40-70%, depending on polarity). With this data bases gradient as well as isocratic (with different water fractions) methods can be calculated. With a isocratic data set just a isocratic method with the same water fraction can be simulated.

Alex

PS: Are there any good models for PFP-phases?

[tom jupille]

PS: Are there any good models for PFP-phases?

The way DryLab is set up, you can add data points to any of the single-variable isocratic models (up to a maximum of 16). The interpolation function is generally a "cubic spline" (which, as I understand, reverts to linear for 2 data points, quadratic for 3 data points, cubic for 4 data points, and a combination of cubics such that the tangents at the merge points are equal for 5 or more data points). As long as you don't try to extrapolate, 3 or 4 data points should work for pretty much anything.

With regard to actual scouting runs, the general strategy behind DryLab has been to start with two gradients (wide enough range to be sure that you see your peaks!) covering the same range, but with different times. Those are used to calibrate either an isocratic or gradient model (your choice). Always check the predictions against an actual experimental run. If you can get acceptable results with a binary system, that's the ideal. If you can't, try again, with a different solvent, different temperature, different column, or different pH (your choice; the program doesn't care).

If the second attempt was made with a different solvent, and selectivity was different (i.e., different critical pairs), then the ternary blend makes sense. At that point run a 50/50 blend of the optimum ACN/water and MeOH/water phases (assuming those were the solvents used!), and use the data from those three runs to calibrate the ternary model as described.

At one point years ago, I tried to call this technique "BOB" ("Blending Optimum Binaries"), but that never caught on

[Alex Buske]

Always check the predictions against an actual experimental run.

Playing around with Drylab can be great fun, however my boss doesn't care and wants to see perfect methods, so I have to check the results.

For sufficiently complicated separations I used to run two gradients (two runs), at two tempratures (four runs), with three organic solvents (ACN, MeOH, 50/50 thereof, => 12 runs) and a number of columns, hoping that at least in some combinations all peaks ccan be identified. It doesn't make sense to enter values of 9 peaks into Drylab when you know there should be 10.

[tom jupille]

Alex, that's very similar to what we recommend in our method development courses. The only (slight) difference is that I don't like to run the 50/50 runs unless I see a selectivity difference between the organic solvents (I'd rather use those runs to screen a different column or pH). A quick visual examination of the chromatograms lets me quickly focus in on the solvents and/or columns that look promising, and then I model only those combinations. I've sometimes referred to that as the ". . . there must be a pony in here somewhere"* approach.






*from the punch line of the very bad joke about the father who tried to cure his over-optimistic child by filling the child's room with horse manure. The kid opened the door and joyfully started to dig through the pile with the explanation, "With all this horse manure, there must be a pony in here somewhere!"

[tadah]

Is there really no other chromatography simulation package out there? The only freely available (but limited) program is from http://www.virtualcolumn.com/download.html.
I would also like to run a few calculations and plot Van Deemter curves and kinetic plots. Is there a good simulation software or Excel spreadsheet that can be useful for doing that? Thank you.

[tom jupille]

I don't have the links handy, but two others are ChromSword and ACD Chrom Simulator. Neither is free.