How consistent are different FIDs?

[nalandmark]

Hello I am new to the forum and had a question about FID consistency. I am in the essential oils industry and we inject a 0.2ul neat sample onto a 15m DB-Wax column using % area for results.
We use 4 GCs; 2 Agilent 6890s and 2 Varian 3400s. I routinely replace the septa and insert liners but the results may still vary by as much as 1-2% between instruments. Shouldn't FIDs be able to achieve greater consistency than this? Thank you for your reply.

[chromatographer1]

Are you splitting the sample?

If so, then it is not the GC detectors, but a distribution of injection variance that may be the primary cause.

The same happens when you change the position of the installation of a GC column, even in the same injector. 1% variance between split injections on the same injector is not unusual. In the extreme it is called sample discrimination.

Plus you have the possible decomposition (injection port, activity of the column, etc) of the sample changing the final purity.

Responses of different FID models to different analytes is generally very consistent if the gas flows are set properly.

If you give more detail there are many on the forum who can go into more detail with your particular situation.

best wishes,

Rod

[AICMM]

I agree with Rod, far more likely front end than back end. Something else you might consider. Pick a component or add a component and then do the calculations based on that component as an internal standard. Lessens the impact of the inlet discrimination. I'm not necessarily suggesting you do this all the time, but rather to look at consistency from a new perspective.

Best regards.

[MikeD]

1 % to 2% difference in FID response between instruments is good. I would not expect much better. I have had some experience of splitting at the injector into two columns and two "identical" FIDs. The relative response of hydrocarbons was no more consistent this. Halogenated compounds were about the same. Other hetero- atoms, particularly oxygen, induced greater differences between two FIDs sometimes up to 10% discrepancy in rrf., even when the flows were carefully adjusted to be the same. I am talking about relative response to a hydrocarbon here, so inlet conditions were not a factor.

[Peter Apps]

Hi Mike

Did you publish any of this ?

Peter

[MikeD]

Peter

No I didn't publish that aspect because it wasn't very repeatable in the long-term.

It does rather conflict with some very old classical GC work on assigning FID contributions to component atoms/groups, but that is what I found. It's just speculation but what may happen after a time is that the two FID jets get differentially clogged with silica because the parallel columns are bleeding differently. So perhaps the effect is less when you have new jets or you clean the old ones. It's possible the effect would be less if the two columns were identical and bleeding the same amount, although that would have been a rather pointless system in one GC. You are trying to get two different chromatograms for the price of one. In my case the columns were a methylsilicone and a phenylmethylsilicone similar to 624.

I did publish on that dual-column setup once but only with respect to the influence of column load on retention time and misidentification. All history now with wall to wall MS.

Regards.

[nalandmark]

Thanks to all who replied. I'm sorry if I was a little vague in my earlier post.
Yes we do split the sample @ 100:1 ratio. Of course it is alot easier to set my spilt ratio on the Agilent instruments.
For varying products we report between 3-15 compounds, many of which are in the <1%-3% area range. The smaller responses add to the necessity for consistency.
I have thought of using an IS in the past, but there were three problems with the idea. There are minor unidentified constituents throughout the chromatogram, had a problem finding a compound relevant in response to all my trouble compounds, and the industry standard is area %.
Also, I mispoke(mistyped) in my prior post. The 1-2% variance is in absolute terms, i.e. 34.7% on instrument A and 36.1% on instrument B. In relative terms, variations of 10-20% between instruments can happen on some compounds.
By the way MikeD-After your comment on oxygenated compounds, I reviewed some of my data and the majority of inconsistency is in my alcohols.

[chromatographer1]

nalandmark

OK, now that I see what you are doing I have additional remarks.

To summarize: You are analyzing essential oils, probably composed of alcohols, aldehydes, and hydrocarbons, both unsaturated and saturated.

You are injecting small amounts (0.2µL) and splitting the sample 100:1 or so.

#1 point: Your injection liner is not perfectly inert. You are open to all kinds of reactions catalyzed by the silanols and trace metals in the injection liner as well as possible deposits left from earlier injections. These reactions can vary from injection to injection, producing different amounts of reactants which often will co-elute with other peaks causing a variation in the measured area of these peaks.

#2 point: Carbowax is not a perfectly inert surface. Especially when hot it can also catalyze reactions, especially dehydrogenation of alcohols to alkenes with the same result as above, sometimes maddeningly variable.

#3 point: Your metal needle of your injection syringe can also cause reactions noted in points 1 + 2 above, ESPECIALLY IN NEAT INJECTIONS.

#4 point: Your analytes in your essential oil have different volatilities and can produce variable peaks 'natural' to any splitter injection system. No splitter design is perfect. There is always some discrimination of the sample. Fun Experiment: adjust the position of the column tip in the injector and see how your area percent values will change.

Add to these points that certain components can be reacting all along the column length and with metal components of the detector and suddeningly it becomes reasonable to expect some variation in results, even 1 or 2 % variation between different injections and GCs and detectors of the same design.

With capillary columns this variation is usually minimized, especially when compared to analyses performed on phase coated diatomaceous earth packed columns.

(age give-away) I can remember analyses with catalytic decomposition of cholesterol to dehydrocholesterol, and 1-decanol to 1-decene, usually because I was performing neat or too high concentration injections with too high an injection port temperature.

Chromatography is fun. I have been having fun in QA and research labs since the 1970s.

best wishes,

Rod

[nalandmark]

Rod,

Thank you for your remarks and let me give you a little more background on the comparisons I have made.
I started by injecting the same sample 10 times on each instrument. For the Agilent instruments the only real variation was a peak with inconsistent integration. For the Varian instruments the results were also very reproducable, except for one or two intermitent runs. I'm sure that the four points you mention play a role in the intermitent analyses. With a more regular PM program I believe I have eliminated these issues.
It is the reproducable differences between instruments that concerns me. We are in the process of reviewing all of our specifications, and without uniform chromatography the control limits will be too broad to mean anything.

[chromatographer1]

Each brand of GC was reproducible.

That tells the story. Design differences create differences in output numbers.

If you want accurate and precise numbers with GC analysis you must eliminate chemical reactions of the analytes being measured, and

eliminate obvious variables like splitter discrimination by using direct injection.

Process analyzers are GCs used to control processes almost always use direct injection. Even there they are concerned with trend analysis and not absolute numbers.

Good luck in reaching your goals.

Rod