Stability/repeatability of Agilent TCD

[sigsto]

Dear all,

I am quite new in the field of gas chromatography and have some questions regarding area repeatability:

I am analysing samples of CO2 and N2 of both known and unknown concentration (in the range from 50/50 to 10/90). The samples are pre-mixed in 10L gas cylinders. My GC system is an Agilent 7890A with a Supelco Carboxen-1010 PLOT Capillary Column, using the thermal conductivity detector. The sample is fed to the GC column through a 25uL sample loop at ambient pressure and temperature.

My problem is when I analyse several consecutive samples from the same gas cylinder, using the same GC method under no obvious changing parameters, the absolute area of the two component peaks vary significantly. The retention time is very stable. Also, the internal relation between the two peak areas (fraction of one compound to the other -> Area_CO2/(Area_CO2+Area_N2) ) is very stable, giving a relative standard deviation in the ball park of 0,1%. The relative standard deviaton of the absolute peak areas, on the other hand, usually turns out in the ball park of 10%.

According to the GC data sheet (http://www.chem.agilent.com/Library/dat ... 6317EN.pdf), the area repeatability should be less than 1% RSD. At the moment I am quite far from this.

What I wonder is how I can get such varying results? I have experienced having two consecutive samples from the same sample mixture giving an absolute area increase or decrease of more than 30%. To me it more or less seems like different amounts of sample reaches the detector in each run, something which I find truly strange.

Also, which area repeatibility will one typically be able to achieve using a PLOT column and a thermal conductivity detector? When Agilent says 1% RSD, I would guess better repeatability than this us quite common?

Thanks for any help!

-Sigmund

[Dr Lou]

Just a thought - Could the gases in the cylinder not be mixed properly?

I used to do gas bag sampling and found I had to rock my cylinder to mix my standard gas mixtures together otherwise I got strange results.

[Johnny Rod]

Not very likely Dr. Lou but it is a concern where you have a gas which is realtively easy to condense. The standards here would have to be at low pressure to get around liquefying CO2 anyway.

What gas flow are you using when you make the injection? Ideally you should have a rotameter (flowmeter tube) or similar on the sample loop outlet. either inject at the same flow rate each time e.g. 100ml/min, or purge it all out, stop teh flow, and inject as the flowmeter hits zero. You should get RSDs under 2% for the flow technique and even down below 0.1% by the latter, without accounting for wandering atmospheric pressure. Also hopefully you're piping your gases into the injection loop not using a syringe, as nitrogen is all around. Once your regulator is purged (how do you do this?) then it'll need a few minutes of gas flow before you make your first injection.

Do you get big peaks or are you splitting the sample? 25ul of pure gas (or % level) is rather a lot. We use 10ul on 1/8" packed columns and the peaks are pretty big to say the least.

[sigsto]

Thanks for the answers.

I do not believe the mixing is the problem, as fraction of each compound is very stable.


For the injection into the sample loop, I use permanent tubings from the cylinder regulator to a six port valve connected to the sample loop. The other end is connected from the six port valve and to a soap film meter to roughly monitor the volume filled/flushed through the sample loop. Usually I open the regulater to fill an internal volume between the regulator and a second valve inserted in the tubings. From this volume (pressure appr. 10bara) i fill/flush the sample loop. After doing this, I usually wait 2-3 minutes before I switch the six port valve and start the analysis to be sure to depressurize the system. At least after this time interval, the soap film does not move anymore in the soap film meter.

Our peaks are quite big, yes, but I'm quite confident that the TCD is not saturated. We also have a smaller sample loop (5 uL), I'll try that one next.

[chromatographer1]

It certainly sounds like you are not purging the sample lines fully, or there is some vortexing of the gas flow and it is not purged uniformly.

The other option is that you have a leak.

After purging and filling the sample loop with a characteristic sample of gas there should not be a large variation in area for your peaks. It is not necessary to wait but only a few seconds after the flow is stopped to inject your sample.

best wishes,

Rod

[rb6banjo]

I do a lot of this. Generally, I'm looking for fairly trace levels of impurities in CO2. I have a small rotometer connected to the outlet of the line used to fill the loop. I just wait until the flow drops to zero (a second or 2 at most) before injection. I get precision like 1-2% RSD on nearly all of my analytes.

Gases diffuse very fast and 2-3 minutes might be long enough for the concentrations in the sample line to change. Just a thought.

[Johnny Rod]

As said above, you're not purging enough, especially for analysing for atmospheric gases. Don't be afraid to use some gas up, leave it with the cylinder open for a few minutes, then shut off the last valve, see the flow drop to zero (rotameter is better than soap filmmeter here but all roads lead to Rome), and inject immediately. In between injections, put the gas flow back on (low if you want). That wasy it keeps the air out, and all you have to do for subsequent injections is crank the flow up for a minute as it'll already be purged.

When you first connect the regulator to the cylinder, do as you are doing now by static pressure purging the regulator, do this three times, then proceed to purge the gas lines as I've described above.

[thohry]

I think purging is important. I got this problem with gas analysis some time ago. And finally, I increased the purging time, the area repeatability was rather good.

[sigsto]

Thank you all, these were very usefull replies.

I'll try to increase my flushing time and let the gas flow through the pipings between injections. This makes sense, I hope it'll take me forward.

Best,
Sigmund

[AICMM]

Sigsto,

I doubt a leak since it would probably change your CO2/N2 ratio significantly. Likewise with flushing since you would see an oxygen peak if you did not flush enough and it doesn't take that much to flush if the sample cylinder is relatively near the sample loop.

Couple of questions then couple of comments. What carrier/reference are you using for the TCD? What temperature do you have the TCD set to? What are the peak areas for CO2/N2?

I would suggest what I call bubble/bubble sampling. Put a piece of tubing on the exhaust of the sampling valve and run it into a flask of water. Turn on your cylinder to purge the sample loop. Do this several times. Then, on the last purge, wait until you see almost no bubbles out the tubing in water and start your run. This gets the sample to the same (atmospheric) pressure for every run.

Best regards,

AICMM