Step Gradients: Where's the Love?

[Triple Quad]

I''ve been doing some quantitative LCMSMS (MRM) method development.

My A mobile phase is 5mM Ammonium Formate in Water.
The B mobile phase is Methanol with 5mM Ammonium Formate.

My analyte has two major metabolites with MW 732 and 746.

I found a step gradient that beautifully separates the target analytes from the rest of the junk in the wheat sample.

I'm using a 3um 2.1mm X 50mm dC-18 column.

The empirically derived step gradient is as follows.
Start off with
10% B at time 0
10% B to 10 min
80% B at 11 min
80% B to 20 min
95% B at 21 min
95% B to 25 min
10% B at 26 min
10% B to 30 min
end of run.

I full scan mode the some stuff washes off by 2.5 minutes,
other substantial unknowns elute 14-15.5 minutes, then big blob of eluents are observed 25-29min.

In MRM mode, my two analytes appear at 19.3 and 20.5 minutes.

I reinject a known sample 11 times in a row and had rock solid retention times. Peak areas varied from the initial injection within 5%. I used a known standard that was in 60% methanol/40% water.

Granted my LC method is a bit unconventional, it was just a test run to see what would come off in the aqueous and the organic phases and when.

I would use this method for our work, my attempts so far to ramp the gradient just hasn't produced the same results. I would therefore go ahead and use it as is. However, my colleague doesn't like step gradients and I should strictly use and gradient ramp. I understand the reasoning, but in this case the results are good.

Would you ever use a step gradient? Steps produce an effect that is different than ramps that in my case is desireable, IMHO.

Your comments are welcome.
Thanks.

[rhaefe]

just a very quick response:
you do actually have a gradient here: you are going from 10% B to 80%B in 60 seconds. That is quite steep but a gradient nevertheless. You also have to take your system dwell volume into account. Depending on this volume and your flow rate you end up again with a gradient. And since most dwell volumes are different from LC system to LC system it will be hard to transfer this method.
A good way to monitor your steps/gradients is to run your pump program with eluent A=water, eluent B=water +0.1% acetone and record the detector response (UV at 254nm). Make sure to replace your column with a restrictor capillary.

[shaun78]

Step gradients tend taking this large of a step tend to shock the column. This will cause swelliing and shrinking of the solid support. This will degrade column life.

As perviously mentioned, considering individual instrument dwell volumes, this will be hard to transfer to anything else.

The use of a nonlinear gradient curve may be a better option.

[Triple Quad]

Thanks for the quality replies.

I obviously was a bit myopic and seeing the results from my instrument and didn't think about the transferability of the method. Nor did I realize the effect of degrading column life via repeatedly shocking the column with such large steps.

I'm going try starting at 10% methanol at time 0 and ramp to 95% over 22 minutes and see how that works out.

Thank you for those insights.

[tom jupille]

Depending on your flow rate, that "step" gradient is probably steep, but not outrageous. The time of a gradient is less important than the number of column volumes involved, and you are using a small column. As long as you're using a silica-based RP packing, I wouldn't worry too much about column damage.

The dwell volume issue is another matter, but you can compensate for that by varying that initial hold time so that your composition change hits the column at the same time on all systems. That does require that your users measure their dwell volume.

[juddc]

I agree with Tom.

I've used step gradients successfully and they work well. They're also no different to transfer between instruments than any other gradient. Why would they be? All you have to do is know both instrument's dwell volumes and adjust your gradient times accordingly. I've done just this and it works well.

Also, as long as your're using a silica support, don't worry too much about swelling. Most folks use steps all the time in chromatography when they're resetting to original conditions in order to inject the next sample. No harm is typically done. Polymeric supports are another matter entirely, however.

I'd also condsider nonlinear gradients very carefully and make sure the gradient profiles between any different machines the method might be run on are equivalent before even thinking about transferring a method.

If your results are good, don't worry about your colleague's bias. Unless he can give a reasoned explanation as to why step gradients are NG, then go with what works. The only thing I can say against them is that there are areas in the chromatogram (in the "step") where you wouldn't want important analytes due to potential baseiline disruptions, but otherwise, they're no differrent in my experience than anything else.

[Triple Quad]

OK, so going back to my original step gradient, I was asked why such a long hold time on 10%B?

My original answer was I simply was doing an elution profile to see where the bulk of sample constituents would appear. Getting to know the column and the sample.

Now, I see that my answer could be to allow for compensation for transfer to other systems due to varied dwell volumes.

I won't worry so much about it now and I'll see if I can "sell" the step gradient method.

Thanks for the terrific responses.

[Kostas Petritis]

10 min is too much indeed and probably unecessary. Assuming that your injection volume is reasonably small, 2-3 minutes should be enough especially as your compounds of interest are eluted 10 after your step gradient. 2-3 minutes should also be enough to acommondate for varied dwell volumes from instrument to instrument (that would be even more true if you were using a 4.6 ID column -> higher flow rates)

[Triple Quad]

10 min is too much indeed and probably unecessary. Assuming that your injection volume is reasonably small, 2-3 minutes should be enough especially as your compounds of interest are eluted 10 after your step gradient. 2-3 minutes should also be enough to acommondate for varied dwell volumes from instrument to instrument (that would be even more true if you were using a 4.6 ID column -> higher flow rates)

I agree. My 5ul injection volume is small.

[Bruce Hamilton]

As it works for you, then that may be all you need, but if others have problems, you need to ensure their hardware can also deliver identical steep gradient changes, as well as compensate for instrument dwell volumes, as suggested above.

Being totally ignorant about 50mm columns, why do you need a 30 min run on such a short column?.

I would have assumed that the unwanted junk is going to sit and sulk near the beginning of the column until hit by the wall of steep gradient solvent - which then washes it off in a tight band before the second analyte. One question would be, are the samples always going to be consistent, and how does sample variation, especially junk content, affect retention behaviour?.

Wouldn't shortening the whole process offer some time benefits, perhaps even precision benefits. I'd consider using a slightly less aggresive gradient and different mobile phase compositions and shorter times at the equilibrium points.

Please keep having fun,

Bruce Hamilton

[Triple Quad]

If I am understanding the common theme of the responses, then any gradient (not just step gradients) can have difficulty when transferring the method between instruments. Isocratic conditions make the transfers easier.

How do we then write an SOP with a gradient not knowing what the instrumental differences will be between the method development instrument and the next analysts instrument? My very general assumption is that "shallower" gradients better than steep gradients.

I obviously didn't know as much about HPLC as I previously thought. I am learning a ton with my first real world project.

Thanks for helping me along.

You're right Bruce Hamilton, I probably don't need such a long run with a short column. I'll attempt to shorten it. I really wasn't very excited with a 30 minute run, but I was excited by the separation that I achieved. I'll see if I can reproduce it in a shorter time frame.

I am having fun. I'm a tweaker. I'm getting plenty of opportunities to tweak.

[Triple Quad]

When mobile phase hits the C-18 column lets say it's 25/75 (water/methanol), does the mobile phase separate inside the HPLC column? Do tiny bands of separated water and methanol move through the column? Does the mobile phase stay thoroughly mixed?

[Bruce Hamilton]

I should leave this to the experts, but my intuitive response ( which could be seriously wrong ) would be:-

Once mobile phases are properly mixed and the column equilibriated, they will move through the C18 column as the original mixture.

Life may get more messy when you have a mobile phase component that seriously interacts with the column, eg water onto silica, ion-pair agents onto phases, etc.. They may be depleted/enriched until new equilibria are established, then the composition should be unchanged.
Gradients would be non-equilibrium conditions.

Please keep having fun,

Bruce Hamilton

[tom jupille]

How do we then write an SOP with a gradient not knowing what the instrumental differences will be between the method development instrument and the next analysts instrument?

For what you have described (an initial weak-solvent hold to wash crud through, followed by a jump to a stronger-solvent isocratic run), the key parameter is the "dwell time" (gradient delay time). Instrument-to-instrument differences in dwell time are easily compensated by making the (isocratic hold + dwell time = constant). That means that your users will have to measure the dwell time (it's not hard to do; you can find a simple procedure on our web site:
http://www.lcresources.com/resources/TSWiz/hs410.htm).

The next most important concern would be the mixing volume (which will distort the gradient profile at the beginning and end of your ramp). As long as your peaks of interest are well separated from the ramp (which yours seem to be), that should not be a major issue.

My very general assumption is that "shallower" gradients better than steep gradients.

If peaks of interest are coming off during the gradient (which is the usual case), then there is actually an optimum range of gradient steepness based on column dimensions and flow (which can be calculated, although that's outside the scope of this discussion). Assuming a flow rate of 0.2 mL/min on your small column, it would be from about 4%/min to about 20%/min. Steeper than that would cost you resolution, shallower than that would cost you sensitivity.

[Triple Quad]

How do we then write an SOP with a gradient not knowing what the instrumental differences will be between the method development instrument and the next analysts instrument?

For what you have described (an initial weak-solvent hold to wash crud through, followed by a jump to a stronger-solvent isocratic run), the key parameter is the "dwell time" (gradient delay time). Instrument-to-instrument differences in dwell time are easily compensated by making the (isocratic hold + dwell time = constant). That means that your users will have to measure the dwell time (it's not hard to do; you can find a simple procedure on our web site:
http://www.lcresources.com/resources/TSWiz/hs410.htm).

The next most important concern would be the mixing volume (which will distort the gradient profile at the beginning and end of your ramp). As long as your peaks of interest are well separated from the ramp (which yours seem to be), that should not be a major issue.

My very general assumption is that "shallower" gradients better than steep gradients.

If peaks of interest are coming off during the gradient (which is the usual case), then there is actually an optimum range of gradient steepness based on column dimensions and flow (which can be calculated, although that's outside the scope of this discussion). Assuming a flow rate of 0.2 mL/min on your small column, it would be from about 4%/min to about 20%/min. Steeper than that would cost you resolution, shallower than that would cost you sensitivity.

Thank you, Tom, that (4%/min to 20%/min) is exactly the sort of info I was seeking. I'm gonna see if I can make time to do the dwell volume (gradient delay time) exercise.

This has been a greatly informative thread for me.
Thanks for all the terrific contribuitions.