New GC columns

[julia]

Hi everyone,
I never worked with new GC columns before (we just bought 2 new columns, the old ones were about 6 years old (my boss didn't want us to change them, and he never performed any check on the old columns and I was not using this GC before)), so I was wondering if you could help me to build a performance check schedule.

I bought a DB-WAX and a HP-5MS and the standard test mixtures for each two. Included with the columns, there was a GC performance summary, that includes the retention times of the test mixture components and the theorical plates per meter. The chromatogram was made using Hydrogen as carrier gas, but I'm using Helium. How will I be able to determine if the column is good? With the theorical plates count or
is there a way to compare the retention times, even if I don't use the same carrier gas?

If I pass about 100 samples per month by headspace in the DB-WAX column, when should I perform a performance check on my column?

If I pass about 150 samples per month by direct injection in the HP-5MS column, when should I perform a performance check on my column?

Thank you for your help!!! I'm a beginner in GC, but I really want to learn and to do my analysis correctly!

[Peter Apps]

Hi Julia

I am not a great fan of running specific performance checks on columns as such. My approach is, and some people will argue that it is heresy, is that as long as the results are OK, the column is OK.

In other words, the best test mixture for any column is a real sample, and the best test schedule is to keep an eye on peak shapes, baseline drift etc on the sample chromatograms.

Peter

[Victor]

While I partially agree with Peter, in that your own sample may be the best test mix, such an approach will not give you the whole story. For example, such an approach would not reveal a fault in your GC system which may have degraded your analysis for a long time. If the analysis has just about worked for years with the fault in place, then you may lose out because as long as you are able to reproduce the conditions you got last week you may be satisfied.

If you already have the test mixes from the manufacturer, I would go ahead and inject them under similar conditions. The diffusivity of sample components in hydrogen is greater than that in helium, so it is possible to run an analysis with hydrogen at faster carrier gas flow than with helium to gain approximately the same column efficiency. However, the results you get with helium should not be massively different and this should serve as a test to see if there are such gross problems in your system. If your test does not compare too badly, this tells you your instrument is not working too badly. Then I would establish your own test with your own compounds. This will give you a reference point for how your analysis looks with a new column with a working instrument under a given set of conditions. Then you can show the difference to your boss when you think the columns need changing.

Sounds simple-but there are difficulties. I doubt if the manufacturer supplies a test mix specifically for headspace, and thus using the manufacturers' mix and conditions will not reveal any problems in your headspace technique. However, I still think you will get something out of performing the manufacturers' test. However, when you have done it once, I might as Peter advocates continue to use your own sample as a test which you perform daily/weekly or whatever.

[Peter Apps]

Victor, you are right, running a test mix could point to inefficiencies in the whole system that might be compromising run times or repeatability for instance, but once a method is optimised the samples and standards will tracjk its performance. I am assuming of course that there is some sort of QC on calibrations, repeatability of duplicates etc.

Agilent supplies a headspace test mix with their 1888 headspace sampler. You can probably buy it separately, and other manufacturers most likely have something similar; EPA VOC mixes for example.

Peter

[HW Mueller]

Victor, do you have any real life examples of "test mixes" showing a malfunction when samples plus standards did not? I sure would like to have some examples of that type of thing. Of course, it wouldn´t be fair to forward an example in which the sample QC was established within one week, and the "test mix" QC was checked over years.
Theoretically, one can imagine that there might be an advantage of a "test mix" if it is designed to show particular aspects of the chromatographic equipment..

[Victor]

Peter- I am almost entirely in agreement with you.

My point about headspace analysis was that the column purchased for this analysis (DB wax) will probably have been supplied with a test chromatogram performed with liquid injection. Similarly, the Agilent headspace mix (or other test mixes) are unlikely to be supplied with a test chromatogram for the the particular DB Wax column in question. So strictly we need to perform two tests, although now we are becoming extremely pedantic and nit picking and I would not particular advocate this unless there seemed to be a particular problem which could not be solved by just seeing if it worked.

If the same analyst repeats the same analysis with a particular GC and gets the same result, this is an indication that perhaps the instrument is working correctly-or at least reproducibly. If the analyst gets the same result as an independent analyst in a different lab using a different GC the analyst has a lot more confidence in this result. The manufacturers test chromatogram provides this external check.

The analysis of free sterols by GC comes to mind-not a recommended procedure, but just an example. Let us say that the analysis worked but that the peaks tailed quite seriously. I assume that this is normal (free sterols are difficult to analyse!!) Some test mixtures contain octanol which can mimic the adsorption of free sterols-let's say the manufactuers test contains octanol, and this is shown as a virtually symmetric peak on the supplied test chromatgram. However, on my instrument octanol tails badly. I clean out the injector, and I get a nice peak for octanol and subsequently a nice peak for my sterols.

Without the manufacturer's test, I have no comparison point for how well my instrument is working.

I don't advocate analysis of free sterols, but this is just an example.

[Peter Apps]

Victor,

let's look at your sterol example. You could have performed the same diagnosis by injecting a pure sterol - in other words a sterol standard. Nonetheless, tailing of sterols could be due to (at least) two causes; hydrogen bonding adsorption, and non-specific adsorption or absorption. Octanol (being small but with an -OH group) will suffer hydrogen bonding adsorption but not as much non-specific ads- and absorption. If you want to know whether to just clean out the inlet, or to deactivate it as well, you need a test mix with a probe for -OH binding (e.g. octanol) and a probe for ads- and abs- (e.g. a heavy alkane). If both tail you need to clean and deactivate, if only the octanol tails you need to deactivete, if only the alkane tails you need to clean.

Column test mixes are designed to demonstrate that the columns can handle a wide range of analytes - wider in fact than is likely to be present in any given sample. For this reason they have a lot of redundancy for nearly all trouble shooting - as an extreme example; if I am fingerprinting the hydrocarbons in paint thinners the amines and acids in a column test mix are pretty much irrelevent, but the non-polar components (e.g. alkanes, esters, ketones) that probe for efficiency per se will tell me something useful (but nothing that I could not also get from running, say, the set of alkanes that I use as retention markers).

In the area of headspace analysis, you can get application specific test mixtures (e.g. from Restek for blood alcohol) that probe the selectivity, specificity and LOQ of the analysis - but these are mixtures of the analyte (ethanol) and real life interferences at realistic concentrations; in other words they are system standards rather than column test mixtures.

Peter

[Victor]

Peter-this is an interesting discussion and your points are good ones. In many cases the operator's own test is the best one. However, especially for a newbie in GC (as Julia says she is) an external reference point is most useful prior to establishing her own test.

If I analyse sterol X and get a moderately tailing peak- how do I know if this is the expected result for a compound which can hydrogen bond, even to a clean FS surface? However, if Peter Apps supplies me with a test chromatogram on his (know clean system) with sterol X and his result is rather similar to mine, I think-well my system is working as well as his, and this is probably the best that can be done with this analysis. Of course, instead I could analyse sterol X, then strip the system done completely and re-analyse it. Perhaps I may discover a fault, and my analysis improves somewhat. But it will still help me to know how my analysis compares with yours. Of course I may have to do all this still, if my result on the test compares unfavorably with yours.

Of course test chromatograms hardly ever contain the analyte that you are particularly interested in. Some manufacturers' test chromatograms contain only very simple compounds which bare no resemblance to the structure of the desired compounds. But this independent comparison system with at least vaguely similar compounds (e.g. octanol and sterols)can be invaluable .

[HW Mueller]

Victor, a test mix might be better than nothing, but a better reference for comparison would be a publication where your compound was analyzed with the same or a similar system.
Actually, since my teachers hammered in the shortcomings of chromatography, I am inclined to repeat that if one really wanted to be sure about the quality of ones analysis, one would have to develop two fairly independent methods. If you can get two such methods to agree your peak shape, etc., is good enough.