Poor linearity FID Shimadzu GC2010

[EUROFINS]

Eurofins Food in The Netherlands checked the linearity of the Shimadzu GC-2010 FID. Linearity came out very poor; measured peak areas for lower concentrations where over 30% too low compared to the calibration curve (0.3mm jet, N2 make-up). The GC-2014 uses the same FID for capillary columns.
Shimadzu Corporation does a thouroug investigation on this issue (I hope).

Have a look at http://www.jhtolsma.nl

Soon the 150 EUROFINS labs (Worldwide) will get informed via the internal newsletter.



Best Regards,
Jan-Hendrik Tolsma

[Yuva]

This is may be because of Wide Range Amplifier in the system. The Wide range amplifier use the Log amplifier electronic. In the Log amplifier we will loose the precision at the higher range.

[EUROFINS]

The amplifier is set on Linear mode, this is the correct setting.

[Peter Apps]

This does look like a log response, are you sure that there is no log-linear setting anywhere else in the data system.

More likely is that you are operarating above the linear range of the detector - how much (in ng) is in each peak ? Even if the peaks are not excessively large the system zero may have drifted upwards (especially if you have an autozero at the start of each run) so that the full range of the detector and electronics is no longer available.

Peter

[EUROFINS]

Let me explain the non-linearity better:

undiluted standard: peak area = 1000 mV*s
10-fold dilution: peak area = 90, this is 10% too low
100-fold dilution: peak area = 8, this is 20% too low

So the deviation compared to the expected value seems to follow a logaritmic function. There are no other settings in the data system.

When I turn of the detector gasses the signal drops 14 mV, so this is the background level.
A peak that has an area of 1000mV*s corresponds to 50 ng.
So I don't believe the level is too high.

Thanks

[AICMM]

Eurofins,

Are you using make-up gas?

Best regards.

[EUROFINS]

Yes, 30 ml/min N2.

[Jumpshooter]

EuroFins:
Clearly you have not over-loaded the column as 50 ng is well within the optimum amount of analyte you want on the column. Indeed there is no reason to suspect that the detector response should anything other than nearly perfect linearity. However, you may be under-loading it.

Logistics of your current study are: to prep a series of 1/10 dilutions of a stock STD that has conc = 50 ng/uL. Then you task the instrument to conduct a 1:20 split of this, thus loading onto the column 50 ng x 0.05 = 2.5 ng 'on-column". No?

Thus your assay does injection 1 = 2.50 ng STD injected, inj 2 = 1.25 ng; inj 3 = 0.625 ng. Since your peak areas are decreasing in a non-linear fashion, then you may be under the LOQ for the assay.
Please try this:

Prep from your stock STD a series of four levels of 1:2 dilutions (2-fold dilutions instead of 10 fold dil'ns) and notate the peak area response. In this scenario: inj 1 = 50 ng, inj 2 = 25 ng; inj 3 = 12.5 ng; inj 4 = 6.25 ng.
I'd wager that this will yield a "more" linear dilution than the 10 x dilutions.
Is it feasible to do this experiment? this would help to get a handle on the LOQ for the assay.

Thanks for your question.

[EUROFINS]

Jumpshooter,
Thank you for your reply.

The 50 ng is calculated with the incorporation of the split ratio, before splitting the amount is 1000 ng.
I did very extensive linearity tests with a standard mix of 20 FAME and made a series of 23 dilutions out of it. The lowest amount corresponded to the LOQ, which is 0.5 ng before split.
All the 20 FAME show the same non-linearity as with undecane.
I only mentioned the 10 and 100-fold dilution to simplify things.

Best regards

[BZK]

I am analysing fatty acid methyl esters (FAME) on a Shimadzu GC2010 with FID detector and having trouble with linearity.


Same problem? Suggestions? Please let me know.

Best regards,
Jan-Hendrik Tolsma

column: Supelco SP2560 100m 0.25mm 0.20µm.
liner: Restek Siltek deactivated with wool.
split injection 1:20.

Maybe not a detector problem but discrimination between injection.
With a 100 m column, 0.25 mm id, the head pressure is very high and any cold spot can cause solvent condensation. I made the same analysis with GC2010 but I use on-column injection and without any unlinearity phenomena.
Try to do the same test with very high split ratio (1:200-1:500) and observe if the unlinearity remain the same. If not the problem is into injection side.

Roberto Barcarolo, Italy

[EUROFINS]

Thank you Roberto,

In my understanding discrimination is when high boiling compounds are too low compared to lower boiling compounds. This is not the problem.
My problem is: any component in lower concentrations is measured too low.
I varied the split ratio and the injector temperature but this had no effect on linearity.

Best regards

[Peter Apps]

Undiluted standard is 50 ng on the the column, 10 fold dilution is 5 ng, 100 fold dilution is 0.5.ng and this is certainly inthe region where adsorption or absortion could be removing significant quantities of the analyte. This is not commonly a problem with tractable compounds like FAMES or alkanes, but if you have a bit of septum or some dirt in the inlet (or at the head of the column) it could well cause the symptoms that you describe.

Have a careful look at your peak shapes - if they are tailing you have ad- or absorption.

That the problem goes away when you change the air flow may be due to one non-linearity cancelling out the other when you force the FID to run under very sub-optimal conditions.

Peter

[Jumpshooter]

This is quite the enigma!
I am still pondering along the lines of how when you varied the gas flow rate to the FID, that the linearity problem became attenuated. Quite enigmatic.

[Jake]

Glad (or sorry?) to hear we're not the only ones contending with this problem.

We've been dealing with the same issue when measuring FAMEs on a Shimadzu 2010. What we noticed was that FAME ratios of NuChek standards varied significantly over relatively small on-column ranges (less than 10X), and that the greater the diffence in area % of the FAMEs, the more exagerated the range would be. I.e., high area % FAMEs over-reported, low area % FAMEs under-reported.

We've really spent a lot of time trying to sort this out, with minor success. Out of frustration, we set our system up identically to the Shimadzu FAME application note. Surprisingly, the data improved. The difference? The Shimadzu application note uses helium as a make-up gas. We originally set the system up with N2 as the make-up gas, which is what I've always used. The problem persists, but isn't as dramatic as with N2 as the make-up. Unfortunately we weren't able to improve it any further by varying detector gas flows. Give He make-up a try as a partial fix.

As several have suggested, it does seem to me to be amplifier or perhaps electrometer related. My understanding from the documentation is that the instrument's keypad setting of 'linear' is only for the analog-out. We're using a workstation. The results are the same using GCSolutions or ClassVP. We have a dual channel system. One channel's a little better than the other in regards to the problem, and when we swapped electrometers, better data followed the electrometer. We did have an electrometer replaced due to noise issues. We were hoping that we just had a bad set, but unfortunately the problem remained after the replacement.

Any progress from your end EUROFINS?

[Peter Apps]

Hi Jake

Welcome to the forum.

I can't offer a solution, but it is very surprising that changing the makeup gas makes a difference to the FID's linearity. Any idea why ?

Peter