Hi Jens, maybe try connecting the "good" GC (the one with the FID)  to the IRMS using a capillary after the column and before the backflush valve on the "bad" GC and manually operate everything. This way, you hopefully will be able to tell if the problem lies before or after the backflush valve, and move on from there.
If you don't want or can't do that, maybe inject air and monitor the argon peaks coming out. Argon should not react with active sites (I hope; if I am wrong someone please correct me) and you should be able to see argon peaks even for tiny injections. Maybe this way you could at least discover a possible leak.
You can of course mix the two approaches and compare argon peak heights with the two GC, but this will be tricky because of differences in capillary length, valves, etc
By the way, did you spray argon around the system, while monitoring mass 40? It's the best way to find small leaks.  

Matheus C. Carvalho
Centre for Coastal Biogeochemistry, Southern Cross University, Australia
Editor for Plos ONE and HardwareX (Elsevier)



Online course about lab automation: bit.ly/2Hg5gfj

Book: Practical Laboratory Automation Made Easy with AutoIt
Open source autosampler for liquids: bit.ly/2QVnDJU
Open source autosampler for solids: bit.ly/3erjacu
Open source autosampler for field work: bit.ly/34cHKut


On Wed, Jun 9, 2021 at 10:58 PM Dyckmans, Jens <[log in to unmask]> wrote:
Thanks for the responses so far, but unfortunately, problem is still not solved.
Of course we changed liner (self-deactivated and brand new ones) and cut of column and replaced oxidation oven.

What puzzles me is that we have a wonderful linear behaviour of peak size vs sample amount, but lose a fixed amount (apparently). If it was some kind of surface effect or reaction I would expect a percentage loss, i.e. a low slope through the origin.
We observe increased (proportional) loss if we insert deactivated glass wool into the liner, but still have the cut off at 70g/L (but a nicely linear regression). We observe a non-linear loss (somehow quadratic regression) if we use non-deactivated glass wool, but still zero signal at 70g/L.

We tried split injection, resulting in lower peaks, but still linear regression going to zero signal at 70 g/L.

What can cause the loss of a fixed amount of sample irrespective of the amount added??

So desperate for help…

Jens

Von: Stable Isotope Geochemistry <[log in to unmask]> Im Auftrag von mwolhowe
Gesendet: Dienstag, 1. Juni 2021 22:07
An: [log in to unmask]
Betreff: Re: [ISOGEOCHEM] substance loss in GC-C

I would echo other comments to clean your system of potential active locations that are 'titrating' your analyte; clean the inlet and replace your liner, take a length off the front end of the column, maybe even clean out your split lines/replace the split vent trap if your system has one (analyte can expand into that space and condense during splitless time, and all sorts of gross things can build up since it's not heated)... maybe swap in a new needle. Stuff can build up in the t-piece between the column and the reactor/backflush too (which your FID presumably doesn't have). We would need to know the exact ways in which the inlets on your two systems are different, though, to give better advice. One question I might ask is if you have glass wool in one liner and not in another; some compounds will break down on glass wool depending on the presence/type/integrity of any deactivation. 

However, given that it sounds like you have a hard cut-off at 70 g/L, and not a decreasing response under that, it sounds too... perfect to be an active site thing (in my limited experience with reactive compounds). It almost sounds more like something is happening to your sample in the preparation for the injection, wet-chemically. Are you diluting the samples in the same way with the same solvents, into the same kinds of vials, using the same needle rinse solvents on the GCs, etc?  

Another question: how does the response factor for your >70 g/L crotonic acid ethyl ester injections compare to things with similar elemental makeup?

Dr. Matthew Wolhowe
Research Scientist/Engineer III
Lab Manager, Oceanographic and Geochemical Stable Isotope Facility
School of Oceanography
College of the Environment
University of Washington  

On Mon, May 31, 2021 at 7:10 AM Dyckmans, Jens <mailto:[log in to unmask]> wrote:
Dear all,

we observe a strange behaviour for some substances in our GC-C system. While measurements of amino acids, PFLA and amino sugars run smoothly (most of the time...), we have problems with the peak heights of several derivatives of polyhydroxybutyrate. We tried several compounds and found either no peak in IRMS (while there is no problem in GC-FID) for crotonic acid ethyl ester, or a linear relationship between sample amount and peak height that intersects the x-axis at (for example) 70g/L, i.e. no peaks for lower concentrations but a nice linear, reproducible relationship above that threshold for crotonic acis and ethyl-3-hydroxybutyrate.
Apparently some part of the GC-C system eats parts (or all) of the substances. But why does this not happen for GC-FID, where the same setup (although on a different GC with different inlet) is used?

Did anyone experience this kind of behaviour or has any idea what the reason might be?

Best,
Jens