Fast HPLC

[Mat]

What pressures are we talking about here?
In other words: A learner (me) is interested in what pressure ranges are "normal", and which are considered as high.

[unmgvar]

As "fast HPLC" goes, you of course first of all mean faster eluting peaks for shorter run times.
A part for the given tips about UPLC, Higher temparatures, you might want to try the monolith colunm from merck. They already have a lot of existing aplications. Also for most aplications it will be easily transferable. Necessary adjustements with existing HPLC's are minimal (generally uv specifications, column lenght and gradient mixing).

[adam]

It seems that no one has yet mentioned one of the more important issues for high-speed HPLC when using a gradient method. That is the maximum gradient rate at which the system can maintain equilibrium.

Anyone have any knowledge on this subject; or can suggest any references?

[unmgvar]

when you talk about the rate. do you mean the rate at which you change the composition or the flow rate of your pump?

I have seen that to many laboratories do not follow 2 simple rules about work with gradients
the first one is not to work when A and B are both 100% buffer and 100% organic and then expect for a good reproducibility.
the second one is that they also forget to check the point of first crystal appearance in the mobile phase compostion. too many times people will clogg their HPLC with buffer deposit by moving to a 100% organic composition when their system at 90% organic and 10 % buffer already get salt deposits. Pressure rise in part of the systems due to clogging and also the check valves get stuck.
basically this is bad for the instrument and bad for the column.
I have not yet seen a problem with the rate at which you change the composition itself rather at which composition you find yourself.

As for monolithic columns a new factor comes into play. it is the mixing capacity of your instrument. All HPLC's can safely mix all type of compositions up to 2 ml/min. but if you go higher then 4 ml/min you might have to add or change your mixer when working from 2 channels.
you know that you have a mixing problem because you get a very noisy baseline as well as very bad reproducebility.

[tom jupille]

Mat: for at least a couple of decades, the upper pressure limit on most LC systems was around 5,000 psi (about 350 bar or 35 MPa). The "ultra pressure" systems from Waters or Jasco raise that by a factor of 3.

Adam: the "rate" issue with gradient systems is usually associated with the rate of change in composition. Mixing volume in the system "rounds" the corners of the gradients (the points at which changes in gradient rate occur). If the gradient takes place over a volume which is much larger than the effective mixing volume, then the effect is negligible. At the other extreme, if the gradient takes place over a volume which is smaller than the mixing volume, then the mixing characteristics determine the actual gradient profile. There is, obviously, a continuum of intermediate cases. There is no sharp boundary at which we can say "Aha!, the system can track only so fast". In fact, if you always use the same system, then it doesn't matter much: if the method works, it works. The major problem arises when you transfer a method to a different system with different mixing characteristics.

[Uwe Neue]

Adam: there was a publication by Pet Carr not too long ago in J. Chrom on ultra-fast gradients. They looked at column euqilibration issues, and did not find a proble,. In a gradient, you are always out of equilibrium, so the only thing that counts is how reproducibly you are out of equilibrium, which in turn is a question of the hardware, as Tom has pointed out.

[Uwe Neue]

Adam, we also have recently run rapid, 10 second gradients without difficulties. While we scanned this subject only briefly, it looked on first glance just as theory predicts.

[adam]

Interesting. I has always thought there was a limit on how fast one could ramp the gradient. For example, on a lot of our methods we have a fast ramp at the end (basically part of the column washout). But in cases where we have a late eluting peak, which comes out on the fast-ramp part of the gradient, we notice that the retention time of that peak is usually variable. We have always attributed that to the system not being in equilibration.

Out of curiousity what is the gradient range for your 10-second runs, and do you see any retention time variation?

[Uwe Neue]

Adam,

We have to differentiate two things: one is a "fast" ramp in the sense that a large change in solvent composition is happening over a single or a few column volumes. The second one is a rapid gradient execution, where true gradients (incorporating a large number of column volumes) are executed over a short period of time. The latter is reproducible, the former is not.

[adam]

Ok. So - at least in a rough sense - there needs to be some correlation between the column flow rate and the gradient time and gradient range.

I don't suppose there is any rule of thumb on this.

[Uwe Neue]

You can look at many gradients, and you can build yourself an impression on what works and what does not work. Or you can look at the evolution of peak capacity with gradient duration. You will find that most gradients are excuted over roughly 20 to 30 column volumes.

My rough rule of thumb is that you do not want to run a gradient over less than 10 column volumes, unless the gradient is really really flat (meaning the solvent composition from the beginning to the end does not vary by much).

So, if your column dead volume is 0.1 mL (e.g. 5 cm x 2 mm column) you can run a 1-minute gradient at 1 mL/min to get to this minimum performance. However, if you play with this you will find that you will get better results at a still higher flow rate for this 1 minute separation.

[tom jupille]

Adam, another way of looking at this is by considering the "average" k' of your peaks during a gradient. By definition, it's the k' that peaks have when theyhit the midpoint. The steeper you make the gradient, the lower the average k'. Just like in isocratic separations, you don't want to make it too low or you run into problems. As Uwe points out, this relates back to the number of column volumes run during the gradient. If you're interested, I can give you more details taken from our method development course.