Problems using analytical method for microbore column

[ballz2wallz]

I have this method that works very well (and is well documented) for an analytical column for the separation of some peptides on rpHPLC.

Now I want to use a microbore column to do the same thing. The problem is, using the same method on that column shows extreme peak broadening and late retention times, and terrible separation.

I was wondering if it might have something to do with the hardware of the system? I called Waters and they said that we should be able to use a microbore column on the system, but my colleagues and I feel like there is some diffusion going on somewhere creating this terrible separation. The flow rate is only .05 ml/min so diffiusion of the sample following injection certainly seems plausible.

Does anybody have any suggestions?

[goxy43]

Hello,
microbore separations are extremely "sensitive" to void volume and generally to tubing dimension, dimension of the UV cell. Can you specify your hardware setup a bit more detailed? I am doing nano HPLC for years now and must say that I never had a problem neither with a method transfer nor with such terrible separation as you describe it.
So, waiting for details.
Cheers

[tom jupille]

microbore separations are extremely "sensitive" to void volume and generally to tubing dimension, dimension of the UV cell.

Just to give you an idea how sensitive, here's an example we use in our training courses (modeled using DryLab). The extra-column volume was assumed to be 16 microliters:



and here's a plot showing the effect of extra-column volume on resolution for various column sizes:



I would say that extra-column volume is the most likely culprit. If you're running a gradient (with peptides, you probably are) you also have to take into account the dwell volume (gradient delay volume) of your system. A standard analytical scale system might have 0.5 mL or so between the mixer and the head of the column. At 0.05 mL/min, it will take 10 minutes for the gradient to even get to your column! The next most likely is the dwell volume.

And all of that doesn't eliminate the possibility that your flow may be lower than optimum (depends on your column diameter and the particle size).

[Victor]

Tom,

Nice example. Can you define exactly what you mean by the extra column volume here? I guess this is just a hypothetical example but perhaps you might indicate how this value would be measured in this case (if it was being measured).

[Hollow]

@Tom
Dear Tom
I wonder about the "translated conditions" in the above chrom. Are they corrected for the new dimensions?
When I use my transfer calculator I would get the following parameters for the 50x2.1/2um column: 1.25 ml/min; 20-90% / 1.67 min.
When adjusting the flow to 0.25 ml/min, the gradient time would be 8.34 min (dwell volume assumed to be 0)
Additionaly the Injection volume should be decreased to only 7% of the original one.

How would the chromatograms looks like with these conditions?
Above you decreased the plates by about 15%, changed the cross section velocity to its half and lowered the gradient slope from about to 70%/18.1 column volumes, right?
I wonder if the changes in the chromatogram are only due to the extra column volume?

[tom jupille]

By definition, it's the total volume of liquid traversed by the sample, minus the volume inside the column. In practice, it's more complicated than that because unswept volume (like a poorly assembled fitting) is worse than a straight run of tubing, very narrow tubing is more prone to laminar-flow problems, and so on.

Measuring extra-column volume is conceptually straightforward, but a PITA in practice. You run a set of "well-behaved" compounds and measure the peak width (baseline width = 4σ) and then plot σ² versus tR². The y-intercept gives an estimate of the effective extra-column volume, and the slope gives an estimate of the column plate number (independent of extra column effects). Here's a slide from that same course (the "?"s on the image are actually "≈"s on the original slide) :



In practice, knowing the extra-column volume exactly is not terribly useful (what's important it to be careful to minimize it!), so I usually figure on the injection volume + 1/2 the cell volume + the tubing volume (0.010" id tubing contributes around 0.5 μL/cm of length).

[Victor]

Tom- in your original example I guess therefore that the extra column volume is the value sigma? I gues it is not sigma^2 because the units of this would be ul^2
It is also possible to measure a delay volume where you inject a compound with no column in place but substitute a zero dead volume fitting. This is something different. I was not sure what value you were referring to.

A value of sigma of 16ul is quite large and I could easily believe could give rise to the effects you have shown

[Victor]

Tom-basically my point is

a) There is a delay volume that one can measure by connecting the injector and detector directly via a ZDV connection. This is not too difficult to do experimentally-you just inject something and measure its retention time. The retention volume is then the delay volume. Your approximation of taking the (injection volume + connection tubing volume + half the cell volume) would seem a reasonable way of estimating this value.

b) There is a the bandspreading for a peak. This can again be measured by connecting the injector and detector as above, but now we measure the peak width, for instance at base, and calculate sigma or sigma^2. This value might be influenced by a number of things...even to a small extent by the diffusion coefficient of the probe in the mobile phase.


You seem to be using the values of the measurements in a) and b) interchangeably, as if they are the same thing. Is this correct?

[tom jupille]

Victor:

You seem to be using the values of the measurements in a) and b) interchangeably, as if they are the same thing. Is this correct?

I am assuming that extra-column volume and extra-column contribution to band spreading are the same thing. This is a gross oversimplification, which is why I hedged my statements:

In practice, it's more complicated than that because unswept volume (like a poorly assembled fitting) is worse than a straight run of tubing, very narrow tubing is more prone to laminar-flow problems, and so on.

As a matter of practice, in all the years that I worked with the DryLab program (in which you can specify an "extra-column volume" which is actually treated as a band-broadening contribution), I found that "fudging" the value to make the predicted and observed chromatograms match in terms of peak width required a value which was in reasonable agreement with the estimated extra-column volume of the system.

Measuring the extra-column volume directly is appealing in principle, but runs into practical problems on a standard analytical scale system. At 1 mL/min, 16 microliters implies a transit time of less than 1 second, so that injection valve latency and detector response times become an issue. The y-intercept approach is more tedious, ultimately more reliable, and gives a measure of what we're really interested in, which is band-spreading.

That said, I'll reiterate my earlier statement that as a matter of practice, actually measuring the extra-column contribution to band spreading is not as useful as recognizing that we have to keep the extra-column volume to a minimum. This is especially true when we try to run very small columns on a standard system.

[Uwe Neue]

to ballz2wallz

I assume that you are running a gradient for your peptide separation. Several things can go wrong when scaling down a separation to a smaller column. The first one is correct gradient scaling, the second is the gradient delay volume, and the third is the post-column volume, especially the detector volume.

The late retention is caused by the gradient delay volume of your system. For the large i.d. column, it played only a small role. For the much smaller column, the gradient delay volume would be (for a fixed gradient) MUCH longer. Assuming that you ran the fat column at 1 mL/min and now the thin column at 0.05 mL/min, it will take 20 times longer to purge the gradient delay volume. This is what you are seeing.

This can be fixed by using a delayed injection. Look at your software C/U manual for that feature, or call the service and let them explain to you how to use that feature.

In gradient separations, the wide peaks are caused by the post-column tubing and the detector. For such a small column, you will need a much smaller detector cell than for your standard column. Look for what microbore flow cells are available for your detector.

[tom jupille]

Hollow:

I wonder about the "translated conditions" in the above chrom. Are they corrected for the new dimensions?

A bit of "back story" is probably in order here. Those figures were originally put together for a paper presented at the 2006 Pittsburgh Conference (I think a copy of it is still buried on our web site:
http://www.lcresources.com/resources/ex ... 060223.pdf)
No attempt was made to "optimize" any of the separations. That particular figure was based on what a naive user might do, which is to run the same gradient time, but cut the flow down to give about the same maximum pressure (ca 2000 psi). The bottom line is that the peak width issues are due largely to extra-column band broadening, while the retention time issues are due primarily to dwell volume.

[Hollow]

Thank you Tom for the reference link.

But I'm still a bit surprised, if (in the real world) the width issue is only/mainly due the extra colum volume or will be the sum of all the changes?

Nevertheless the conclusions stays the same: reduce extra column volume when using smaller columns.

[tom jupille]

But I'm still a bit surprised, if (in the real world) the width issue is only/mainly due the extra colum volume or will be the sum of all the changes?

The observed width is, of course, the cumulative result of all the individual contributions. Assuming they are all independent, the contributions are root-mean-square additive. (the observed width is the square root of the sum of the squares of the independent contributions). In practice that means that the largest contribution has a disproportionately large effect. In the case of very small columns (e.g., 50 x 2.1 mm) in "conventional" systems with effective extra-column contributions of 10 microliters or so, the extra-column contribution does, indeed, dominate.

[goxy43]

Instead of talking long on theoretical values and experiments:

Mitulović G, Smoluch M, Chervet JP, Steinmacher I, Kungl A, Mechtler K.
Abstract
An improved method for tracking and reducing the void volume in nano HPLC-MS with micro trapping columns.
Anal Bioanal Chem. 2003 Aug;376(7):946-51. Epub 2003 Jul 11.

A practical example of how void volume can ruin your "micro" or "nano" separation. Let us look at this problem from the practical point: we need an exact description of the separation system, connecting tubing, UV cell volume.... in order to round up and fix the problem.