excsessive void volume estimation

[Hollow]

Hi

I have one more question on this often discused topic.

Is there any rule of thumb, for the estimation of the void volume (flow x retention time of an unretained comp.) based on the column dimension?

Which values are common/acceptable for normal HPLC and high/rapid resolution (i.e. UPLC/RRLC)?

When should I have a closer look if the column or the system have excessive "void volume"/"extra colum volume"?

In the archive I found an answer from Uwe, where he stated that the column internal volume can be estimated as ~2/3 of the column's cylinder Volume

Are there any empirical / recommended factors on this?

Thanks

[tom jupille]

When should I have a closer look if the column or the system have excessive "void volume"/"extra colum volume"?

Those are two different things. The column void volume (internal volume) is whatever it is. Extra-column volume should be kept to a minimum. The smaller the column and/or the higher the plate number, the more rigorous the requirement for small extra-column volume. You will know you have a problem if the plate number is below what it should be (especially for early peaks).

In the archive I found an answer from Uwe, where he stated that the column internal volume can be estimated as ~2/3 of the column's cylinder Volume

That's a "rule of thumb"; I've used 60%, which is about the same thing. It's based on the assumption that about 25% of the column volume is "interstitial" (between the particles) and about 35-40% of the column volume is internal (inside the pores). As you might imagine, the internal volume can vary considerably based on the porosity of the packing.

Are there any empirical / recommended factors on this?

No recommended value (it is what it is). If you have a 4.6-mm id column, you can take the column length in cm and divide by ten and be reasonably close (give or take 15% or so). Thus, a 150 X 4.6 mm column has a void volume of around 1.5 mL.

[Hollow]

Thank you Tom for your answer.

Yes I know the differences but what I can evaluate quickly is the "overall volume" of column and system, so my idea was to have a short estimation on hand, if the observed value is acceptable or if there is some "extra performance" or problem hidden in the system.
Especially on our older systems which were designed for 250mm columns and now are more often being used for 150mm or even smaller columns too. In routine analysis there usually are no thoughts on this nor the time to do some tests or optimization and as long as the chromatogram looks good, everything is fine but if not, maybe the fault is searched anywhere else but there...

Cause a lot of our methods are done in gradient mode, the plate count is not calculated on a regular basis, so maybe we overlook that there is actually a need to improve the system.

For example, when I observe a void time corresp. to about 75% of the column volume, is this still ok, or is this probly too much and I should take the time to analyze and optimize this issue, or at least keep it in mind if troubles occure.

Are there any "recommendations / rule of thumb" for the ratio of "extra column volume" : "column volume", how much could be ok or when do I probably lose too much performance?

How do I best determine the extra column volume?
I found a procedure on a Waters column booklet (no column installed, 5sigma @4.4%height), but which substance and flow rate should I use?
Would be acetophenon @ 0.5 ml/min MeOH ok?
I once read that this should be done on a UV detector and not on a PDA. Why?
(Nevertheless I have to do it on PDAs..)

[tom jupille]

The problem with the "rules of thumb" is that they are +/- about 15%, which is a lot more than what you're trying to measure (150 microliters of extra column volume is a lot!)

You can get a more accurate measure of the extra-column contribution to band broadening (which is what really matters) by doing an isocratic run with three well-behaved compounds (usually neutral, no tailing, etc.) that give you reasonable (and different) k' values, then plot σ^2 versus tR^2. The three values should lie on a straight line. The slope is the reciprocal of the "true" plate number for the column (i.e., unaffected by extra-column broadening) and the intercept is the extra-column contribution to dispersion σ^2ec. Here's a slide from our Troubleshooting course that shows what I mean:



This assumes an ideal peak shape, and it assumes all peaks give the same plate number, neither of which is strictly true, but close enough.

The fact is that most people don't go to that much trouble. Unless you're going down to very small columns (2.1 mm ID and/or 50 mm long), common sense prevails:
- stay with 0.010" or smaller ID tubing
- keep the runs of tubing as short as you can

[Hollow]

Thank you Tom

I will try to reevaluate some existing chroms in this manner and maybe I find some "free" time to do it the other way and compare the two results.

If I can use the data from the existing chroms that would make it easier to implement this "test" on the other department, cause it won't generate additional work.

[Uwe Neue]

Hollow,

It appears that you are interested in more than one thing, since you mentioned above that your colleagues are running gradients, and you are interested in things that can influence the chromatograms. The extra-column volume (between injector and detector) are not very important and are typically small. The gradient delay volume though can be large and affect the chromatogram more significantly. The real problem with this is that you do not "see" this volume by looking at the column void volume...

[Hollow]

Dear Uwe

thank you for your response.
Do you remember our mail exchange about a year ago?
I created my own gradient transfer calculator and had asked you about some details.
My calculator is working well (afaik) and I'm using it quiet often for my work. There, I implemented the correction for the delay volume, so this parameter is fine, thanks.
Slowly but more and more, I'm getting my colleagues to be aware of such details.

But, now I "discovered" this system parameter (bandspreading volume), which I think won't hurt to know its value.

Background:
Since we run into some issues when we applied existing methods on our Agilent RRLCs, I would like to know our systems better, so I can fasten the troubleshooting/-solving process or even avoid problems.

What we often get now on the RRLC, is the "volume overload" issue due to too high solvent strength of the sample. Since there were no problems in the past on the other HPLCs, this parameter wasn't realy considered.
But now, the "optimized" RRLC with its short and narrow tubing seems to be very sensitiv to this.
What we conlcuded from my last post (http://www.sepsci.com/chromforum/viewtopic.php?t=8252) is, that our Waters systems are less senitive to this cause of the injector design.
But we don't have a clear occasion why the third systems (Varian) are less sensitive to this too.
One of my guess is that these systems have more bandspreading volume due to thicker tubing.
(I won't be surprised if we didn't used all the performace of the system and falsely assigned problems to column failure in the past...)


Back to the topic:
I re-evaluated some existing data from the PQ-test (Waters) in the manner of Tom.
I took the width@4.4% and divided it by 5, raised it to the power of 2 and plotted it versus tR^2
Linear fit (n=3) gave a r^2 of 0.9998, N=(1/slope)= 10211, and the extra column volume = 33µL, for our Alliance system.
(Column was a Symmetry C18, 3.5µm, 4.6x75mm)
For the second system, a Waters 600 modular: r^2=0.9997; N=10811; V(ec)=31µL

In addition I carried out some other test suggested on a Waters troubleshooting presentation:
- replaced column with a union
- MeOH 1 ml/min, isocratic
- injected various µL of a solution of Aceton 6 g/L and Acetophenon 60 mg/L in MeOH (gave about 700 mAU with 5µL)
- set the PDA (2996) on single mode to 254 nm, 10 Hz, no filters, no delay
- evaluated the width@4.4% and tR

For the Alliance I get the a system void volume (peak appex) of 100µL, and a bandspreading volume of 54µL.
For the 600er sytem: V(sys)=100µL; V(bs)=63µL

On the 600 the value of 63 staied +/- constant from 1 to 5 µL and increased then for increasing injection volumes.
(If I subtracted the injection volume from the V(bs). the values were constant up to about 10µL, @58µL)

The value of 33µL looks a bit too low for me, while I think the 60µL are plausible and quiet good.

QuestionS:
Are the values of the two methods comparable or are they measuring some different volumes (e.g. volume column-to-PDA vs. inj-to-PDA)?
Or should they be more or less similar and probably my data from the PQ-tests aren't good enough?
Is the correction for the injection volume not made cause it's not relevant or would it be even false to do it?

Thanks a lot!

[Uwe Neue]

I prefer to use the direct method, rather than the method proposed by Tom. However, I usually set it up at a lower flow rate than what you have used. If I want to be really picky, one needs to use the retention time at the center of gravity of the peak for the calculation of the volume, but for all practical purposes, this is not very exciting. I also would use the fastest sampling rate that my system allows, as well as the smallest injection volume possible.

[Hollow]

I used to do the test @0.5 ml/min on the Alliance.
For V(sys) it gave a some lower result about 89µL while the V(bs) gave the same value of 54µL.
At 0.2 ml/min the system shortly "lost prime" (I think due to few pressure)

What do you mean by "center of gravity"?
Is it the real middle of the peak? (=appex only for symmetrical peak)
How to calculate it? "Peak start" + "width@4.4%"/2 ?

[mardexis]

Hi Hollow,

Yes. Center of gravity is more like the real center of the peak. The first moment of the chromatogram is the center of gravity while the 2nd moment is the 'true' measure of the peak dispersion. In Chemstation you can go to 'reports' and choose 'extended performance' which should include the first and second moment values. I don't use Millenium but I'm sure that it must do it as well.

I typically measure system volume at 100 uL/min. You should add a restriter after your detector to keep bubbles from forming being careful not to exceed the max for your flowcell. If you need more pressure to keep your Varian pumps stable then add some restriction (coiled tubing or column) between your pump and injector.

[Hollow]

thank you for your suggestions, mardexis. I will keep them in mind for the future.

Update:
First measure on the oldest Varian revealed no good results (as almost expected...)
V0,sys = 170µL
Vbs =190µL (!) (manually measured, cause CDS won't give width@4.4%)

The PQ test on a symmetry c18 column (3.5µm, 4.6x75) only turned about 3200 plates for butyrophenon, where as the same column on waters 600 (s.a.) gave about 7300 plates!

I'm realy impressed, the PQ-chromatogramm looks terrible...
(almost like a pic of a shark meeting )

[Uwe Neue]

Well, what you see from different sytems is not a surprise. I always recommend to measure all these values so that you know which systems are more similar than others. Especially the system bandspreading can bother you a lot in isocratic chromatography. On the other hand, in gradient chromatography, it is usually irrelevant.