scalability b/w analytical and capillary LC

[smiley]

anyone here know the scalability between regular LC column(4.6mm ID) and capillary LC(~300um ID or other size?)?
what i mean is if there is any transfer issue if i have an analytical method on 4.6mm column and want to apply on capillary LC, or vice versa?
i know the injector dead volume will be an issue.
and can i find the corespondent capillary columns for regular C18 phase? and what about normal phase?
the initial look on LC packings website shows the flow rate is not proportional to that of 4.6mm column, the 0.8-1.0 ml/min flow on 4.6mm column should be around 3-4ul/min on capillary column, but the data from LC Packings shows either 10 times higher on 10 times lower.
is there any such comparison study published somewhere?
thanks,

[Uwe Neue]

To get the same results, the flow rate should scaled by the square of the column internal diameter.

[tom jupille]

. . . and the extra-column volume should be scaled to the column volume (square of the diameter times length).

Additionally, if you're running a gradient:
- the gradient delay volume should also be scaled to the column volume.
- the gradient time should be scaled to keep the ratio of flow rate to column volume constant. {t1*F1/V1 = t2*F2/V2 where t is gradient time, F is flow, and V is column volume}

[smiley]

what about resolution? is there any adjustment necessary?

[tom jupille]

Getting the same resolution is what scaling is all about. If you can get the same retention, selectivity, and efficiency, then the resolution will be the same . . .
. . . but that's a big "if".

[smiley]

Tom,
do you know any such comparison study done in the past, by either academic or industry?
thanks,

[Uwe Neue]

We do this routinely to make sure that there are no strange issues that cause difficulties. Either from analytical columns to prep columns, or from analytical columns to smaller bore columns. It works.

If you do not get the same results, than there is a problem somewhere, which can be diagnosed depending on the observations.

[smiley]

Uwe,
so you mean it is mostly based on observation but no systematic ways to configure it from the very beginning?

[tom jupille]

I'll jump back in here, Smiley.

It *is* systematic, as described in the various earlier posts. Uwe's point is that the systematic predictions are routinely checked in the real world. This is neither new nor undocumented; whole books have been written on the subject. In fact, Uwe is the author of an excellent one:

http://www.amazon.com/exec/obidos/tg/de ... s&n=507846

[Uwe Neue]

I still do no yet understand your question. If I want to change a separation from one column to another one with different dimensions, the scaling parameter is the ratio of the column volumes. In the simplest case, I keep the column length constant, and then the saling is done by the column crosssections, or squares of the column diameters.

If I have a 4.6 mm column running at 1 mL/min, I should get the same results on a 1 mm column of equal length and particle size at a low rate of 0.047 mL/min. I also need to decrease the injection volume by the same factor. Now, if you try this on an instrument that has not been designed for such small i.d. columns, your peaks will be wider than on the 4.6 mm column. If you have a suitable system, and the peaks look good, then you still may have some changes in retention time, if you are running a gradient. These changes in retention time are due to the gradient delay volume of the two systems that you are running. This can be addressed also by proper timing of the injection, if your instrument lets you do this.

Just a few thoughts on how things should work, and what possible complications might be there if you do not get the expected results.

[adam]

OK, this is important - and commonly missunderstood - stuff. So at the risk of further complicating this thread, let me take a stab at it.

With respect to the following items:

- flow rate
- injection volume
- extra column volume
- dwell (or delay) volume

It seems to me the scaling should be based on the square of the column diameter. Tom: you had stated that delay volume should be scaled to column volume. This doesn't seem right to me. Here's my logic: if the flow rate is reduced by the square of the column diameter then, for a constant dwell volume, the delay time would increase by this same factor. So to make the delay time equivalent (time is really what's important on the chromatogram), we would have to reduce the dwell volume by the same factor (square of column diameter).

The one point I'm not entirely clear on is the scaling of the actual gradient. We all seem to agree that flow should be scaled to the square of column diameter, so let's focus exclusively on the length issue (i.e. let's assume that we're comparing columns of equivalent ID but different length). Here it seems to me that the chromatography would look different in most cases when a gradient method is run. It seems to me the only situation where reasonably similar chromatography would be obtained is where - on both columns - none of the peaks elute until the end of the gradient (then the only difference would be slightly longer retention times on the longer column, and slightly wider peaks).

By all means let me know if you see any flaws in my logic.

Adam

[adam]

Quick followup: I should have directed my challenge to both Uwe and Tom. You've both emphasized column volume as the relevant parameter instead of column diameter.

[MG]

What I understand from this thread is that a capillary or nano-bore column, of the same length and containing the same packing material as a 4.6mm column, should have the same resolving power as the 4.6mm column. No better, no worse. (Assuming flow rates have been scaled properly and extra-column volumes have been accounted for). Do I understand correctly?

[adam]

MG

Throughout this thread I think there's been an assumption that - if all the scaling issues are properly dealt with, there should be no change in efficiency or resolution as a function of column ID (for packed columns of course). This is also the conclusion one would reach based on the van Deemter equation: in that H has no dependence on column ID. In reality, there have been papers published that show that one can get better efficiency with smaller ID columns. That's a whole other topic (if you want to start a new thread on it, I can throw in what I know - though I'm not the world's leading authority on it).

Right now we are working off the first approximation that column ID has no effect on efficiency assuming scaling issues have been addressed (basically extra column effects).

I am in the uncomfortable position of having challenged both Tom and Uwe, because I think they place to much emphasis on column volume, when it seems that column ID is the relevant parameter for most scaling issues.

[tom jupille]

MG: yes.

adam: both volume and id are important in different ways.

For efficiency:

- flow rate should scale to the square of the column id.
- extra-column volume should scale to the column volume.
- assuming you scale flow rate and extra-column volume as above, the resolution will scale to the square root of the column length.d

For gradient systems: if you want to run under "equivalent" conditions, you need to keep the average k' of your analytes (k*) constant (assuming linear gradients). That whole topic was discussed earlier:
http://www.sepsci.com/chromforum/viewto ... highlight=
but the bottom line is that k* is proportional to (tG/D%)(F/Vm)(1/S), where:
- tG is gradient time
- D% is the gradient range (final%-initial%)
- F is flow
- Vm is column internal volume
- S is a characteristic of the sample (it's the slope of the k' vs isocratic mobile phase composition).
Given that you've scaled flow and column dimensions as described above, you can tweak the gradient time as necessary to keep the same k*.

Strictly speaking, if your starting % organic is sufficiently low that your analytes stick to the head of the column, dwell volume is irrelevant. In practice, when you get down to small columns (and resulting low flow rates and fast gradients, the time to flush the dwell volume can be the limiting factor on throughput.

I think a lot of the confusion arises because id affects columm volume.