Solvent Analysis

[MikeD]

I can confirm what Peter Apps says about splitting your sample between two columns. If you don't have MS then get a second FID and a non-polar phase (same dimensions) to complement your 624 and split at the inlet. Choose your non-polar column to have a phase thickness about half that of the 624 and the run times will be similar on the two columns. We did this for many years when dealing with solvent mixtures. Out of 150 common VOCs in industrial use we always got separation on at least one column. I would never attempt it on a single column without MS.

Taking the scope of your sample to be waste VOCs gives quite a few coelutions or close elutions on a phase closely related to your 624:

n-pentane/Freon 11
n-hexane/iso-propanol
n-heptane/carbon tetrachloride
MEK/1,1,1-trichloroethane
benzene/1,2-dichloroethane
ethoxyethanol/2,4-dimethylheptane
n-decane/ethoxyethyl acetate
1,3,5-trimethylbenzene/cyclohexanone
p-ethyltoluene/furfuryl alcohol

and many more.

[MikeD]

Sorry, I got the phase thickness ratio the wrong way round in the message above. Assuming a 50:50 inlet split between the non-polar column and the 624 then the phase thickness of the non-polar column needs to be twice that of the 624 column in order for run times to be similar on the parallel columns. Otherwise you end up wasting say 15 mins per sample.

[sadsal123]

Hi Peter,

On this, I altered the flow to 0.5 throughout the run. What do you think? The Peak after Butoxy Ethanol is Cyclohexanone.

Regards,

Salma

[Peter Apps]

Hi Salma

OK, for a two minute shorter run there are two pairs of peaks not quite as well resolved. Next step; run a 2C/min temperature programme from time 0 (in other words no isothermal at the beginning).

It is probably an artifcat of the printer or screen display, but you have quite coarse steps on the peaks. What is your data sampling rate ?

Peter

[sadsal123]

Hi Peter,

Should I use the original method (0.3 Flow then 0.4) to do the ramp that you suggested or should I do it on the 0.5 Flow?

Also, my Sampling rate is 6.25pts/sec. Yes I also thought that the peaks were coarse, but wasnt sure how to improve this yet. Should I increase or decrease the sampling rate?

Also, How from 0 to what temp should I do the ramp?

Thanks.

Salma

[Peter Apps]

Hi Salma

Use the 0.5 ml/min flow rate, start at 38 C and programme at 2C/min. I expect the MEK and EtAc peaks to get narrower because of the programme, and as long as the different temperature does not change the column selectivity they will stay the same distance apart, giving better resolution.

NB that chromatography is designed not to work as predicted !!

You need 20 points on a peak profile to give robust integration, so 6 points per second is bordeline OK for peaks 3 seconds wide at the base (yours are wider than this). The steps on the chromatogram are just printer/screen artifacts.

Peter

[sadsal123]

The Label of EtAc is incorrect as it comes off very close to MEK. The MEK and Ethyl Acetate Peaks are not very well resolved. Also, The Butoxy Ethanol and the Cyclohexanone are not well resolved either.

Have you any more ideas? I think the selectivity is changing with temperature because at about 40oC, the MEK and EtAc are resolving but anything lower or higher, its not working. Is this possible?

[Peter Apps]

Hi Salma

Thanks for trying

If it happens, it must be possible. The only other thing that you can try is a column with a different statioanry phase, but Murphy's law will ensure that you always have co-elutions (Calvin Giddings actually quantitifed the chances of getting a chromatogram full of resolved peaks - and they are depressingly low ).

I would go with MikeD's advice - simultaneous analyses on two columns going to two detectors. The only alternative is GC - mass spec, and that will only be viable if the co-eluting compounds have very different mass spectra.

Good luck Peter

[chromatographer1]

This is an example where GCxGC can really be useful and give a full analysis of the solvents.

If this is critical to your business and you cannot do multiple analyses on multiple columns then this path may be one you may wish to consider.

A less difficult direction would be to use a multidimensional analysis with two detectors using a simple heartcut valve to perform multiple heartcuts to a column of different selectivity where coelutions are known to exist.

This procedure is similar to how I once developed a fuel test method to measure 0-100 ppm of total aromatics (C6 to C16) in 12 minutes (later changed for more robustness to 20 minutes). It can work and work well.

best wishes,

Rod

[sadsal123]

MikeD,
Any ideas as to what 2nd column would be best? Because you have worked in similar area, you may have the 'right answer'

Also, lack of experience means I havent a clue as to how to work with a second column. How would I connect? Do I connect the two columns or did you mean put the second column parallel to the first and connect to the second detector? Help!

Rod, can you plz elaborate?

Peter, I really thought you might suggest something easier than 2 FID's and 2 Columns

I am still learning the ropes and it has been very interesting.

Thank you all for your suggestions.

Salma

[Peter Apps]

Hi Salma

On the basis that you have to do the analysis on at least two columns to get each component as a clean peak on at least one chromatogram, and on the assumption that your boss would rather not spend the money for an extra FID (GCxGC only needs one detector but for the price of the fancy GCxGC hardware and software you could buy a dozen retro-fit FIDs !) there is one cheaper option to having two columns going to two detectors.

Connect two different columns (the phases etc that MikeD suggests would be the ones to go for) to the inlet, using a two-hole ferrule, which you can get from any of the GC consumables and spares suppliers. Connect one of the columns to the detector and do an injection. Half the sample goes down the column to the detector and gives you a chromatogram, the other half goes to waste down the other column. Then swap the columns at the detector end (good to have a nut and ferrule on both columns) and inject again. Then you get the complementary chromatogram on which (if the gods are kind) what co-eluted on the first column is now separated.

The advantage of leaving both columns connected to the inlet all the time is that they stay conditioned - if you swap them completely you have to run the program once to clean them up. The only constraint is that the maximum oven temperature has to suit the lowest maximum operating temperature for the two columns, but with the nice volatile analytes you have this would not be a problem. If you have to automate, then run the batch on one column, swap and then run them all again.

At the least the results might convice your boss to get another FID to save you burning your fingers on the hot detector nuts.

Enjoy Peter

[chromatographer1]

Salma

I assume you are testing waste solvents which may vary in composition from only one or two solvents to the maximum possible to the imagination.

If you base your primary separation on the 1301 column (a good choice)l, on your list of possible solvents you can determine where the coelutions exist. For example, MEK and EtoAc, benzene and dichloroethane. among others.

By research you can find another column which you know (by experiment) will separate as many as possible of the coelutions from the 1301, perhaps a 100% methyl silicone (not the only choice but a fairly good one)

Now, say you run your test sample and you find 8 solvents, and one of these is MEK/EtoAC. Which is it? You can't tell from the one column.

If you only have 1 FID in your (valve or Deans switch equipped) GC, you remove the 1301 COLUMN from the detector and install the second column, whatever it is, leaving the end of the other column in the oven. If you have two FIDs you install one column into each detector and leave them there.

You inject your test sample again and heartcut the peak of MEK/EtoAc onto the second column (using a (Valco?) valve or Dean switching device) and determine the composition of the mystery MEK/EtoAc peak. It may be one, or the other, or some of both, right?

You make as many heartcuts from the 1301 as you have possible coelutions and ID as many of them as you can. I hope this is clear to you.
GCxGC is much more expensive and would require research before you have a viable solution. It 'heartcuts' all the peaks onto a second column for identification. Not trivial to set up but when successful it would require little technical adjustment and would allow automated determination of results.

Now the other idea Peter and Mike proposed is to install 2 columns (both in one injector, or 1 column each into two injectors). Inject your sample onto both columns (in the same injector or separately, at the same time or separately) and use the different selectiivity of the two columns to separate coelutions from the 1301 (which you cannot report with certainty) on the second column. The possible problem with this is that you might have a new coelution on the second column and still be unable to report the true amount or the true identity of the solvents.

The possible problem with the first heartcut solution is not having a second column that will separate all possible coelutions from the 1301.

BUT, you should find a solution that will separate almost all of your coelutions, either the first heartcut multidimensional solution or the second duplicate confirmation column solution.

and I wish you good luck. Be sure to publish your results as many would love to have the solution at hand.

best wishes,

ROD

[chromatographer1]

I fotgot to add Salma, that if you don't have to use the maximum separation ability of your column (long run times) you can screen samples much faster and get more work (more samples) done. Then you can go back and test the possible coelutions for ID confirmation and measurement.

This holds true for both the dual parallel column configuration and for the multidimensional (optional GCxGC) configuration.

The first is the cheapest. The second may be the most successful (possibly).

Rod

[sadsal123]

Thank yo all for your suggestions. They sound very interesting and I shall definitely think about trying them in the near future. I am off next week but will try them after.

Peter, any idea what kind of run times I am looking at? I am sorry its a little wild to ask as i havent even looked at columns but what I mean is, as Rod says, my solvents could be a mixture of anything and it would be fantastic to try an separate everything, but because I have a high throughput, I am not sure that I can afford very long run times. If I was to use two columns and two FIDs, am I looking at very long run times?

I have mentioned changing our carrier gas and will definitely do it in the next 2 months. I have a huge backlog to catch up with so dont want to start messing with the methods that I have managed to get to work with current settings.


Have a great weekend,

Salma

[Peter Apps]

Hi Salma

The run times (for each separation) will be shorter than you have now - probably 30 - 40 minutes, or even shorter because you will be able to tolerate some co-elutions on each column.

If you have two detectors the separations run on both columns at the same time - you generate two chromatograms in the time that it now takes to do one. If you only have one detector your total time for each sample is approx doubled because you have to run one separation after the other.

Having a huge backlog is exactly why you should change to helium now, it will double your throughput. Just remember to change the volume flow rate (best of all set the linear flow velocity at 35 - 40 cm/s).

Rod is right that the only way to be completely sure that you will resolve every component in every sample is to use multidimensional systems, but comphrehensive is very expensive and fiddly, and you have very limited sample capacity on the narrow bore second dimension column so it does will not work well if you have a very wide range oc component concentrations. Deans switch or valve heartcutting is cheaper to set up because you can do the plumbing yourself, but you have to do one second dimension run for every coelution on the first column - which can become very time consuming.

The logic behind doing two parallel one-dimension separations on different columns is that you only have to have each component as a clean peak on one of the chromatograms. You do not need either separation to have no co-elutions, as long as the peaks that co-elute on one column do not co-elute on the other. Suppose that you have compounds ABCDEFGHI, and if co-eluting compounds are in brackets the chromatogram on one column looks like this A B C (DE) F G H I. On the second column the chromatogram could be A (BC) D E (FG) H I, or even, in the extreme (ABCFGHI) D E and you would still be able to identify and quantify every component because you have a clean peak on at least one of the two chromatograms for every component of the mixture.

Two columns in parallel is also far simpler and cheaper than comphrehensive or heartcutting 2-dimensional.

Peter