best way to calibrate for permanent gases?

[rekuci]

I am in the process of analyzing the composition of unknown samples that contain H2, O2, N2, CO, CH4, and CO2 on a capillary column (Carboxen 1010, oven at 25C for 10 min then 5C/min to 60C, Ar carrier). The samples are collected in miniature cylinders during experiments, with varying pressures, which I can hook up to fill a 10 µL evacuated loop on a six-port valve, while reading the pressure in the loop on a gauge, to know how much is really being injected.

What is the best way of performing a calibration to determine a % composition? I have a Scotty calibration mix with 4-5% of all gases above in helium, so all that I can do is vary the pressure injected and plot partial pressure of each gas vs. peak area. I did this and got some very nice, slightly nonlinear curves, however when I then use these curves to get partial pressures in my unknown, the total pressure is far less than what I read on the gauge! It is possible, but unlikely that this is really true. Is there a better way of doing this?

Several months ago, I did some very detailed calibration curves of %H2 and %air vs. peak area for varying compositions of 44.7 psia of air/hydrogen mixtures, which I can make myself, but this curve (a line with R^2 = 0.997) predicts a zero peak area for 22% hydrogen! And this was actually the case when I ran a sample with 22% H2. This, by the way, is in contradiction to the partial pressure vs. area calibration curves generated from the calibration gas.

[pi3832]

Where is the gauge in relation to the loop? How are you controlling the pressure?

Not that it matters, but is it really a 10 micro-L sample? That seems rather small to me. I do a lot of permanent gas analyses, and I can't off the top of my head recall using anything smaller than 1 mL.

I'd suggest using a larger sample loop and doing a "drop-ball" injection. Which means: purge your sample through the loop, shut off flow and let the loop drop to atmospheric pressure, then do the injection.

A drop-ball's usually the easiest way to get consistent pressures in your sample loop.

As for calibration: what concentrations of the analytes are you expecting? If they're consistent, the easiest thing is to get a standard or two with the components at the roughly the same concentration as what you expect in the sample.

If you're in the U.S. or Europe, there should be any number of specialty gas suppliers that can make those standards for you, at ppm-levels and up. (PPB level standards are going to be a little harder to source.)

[Bruce Hamilton]

I am in the process of analyzing the composition of unknown samples that contain H2, O2, N2, CO, CH4, and CO2 on a capillary column (Carboxen 1010, oven at 25C for 10 min then 5C/min to 60C, Ar carrier). The samples are collected in miniature cylinders during experiments, with varying pressures, which I can hook up to fill a 10 µL evacuated loop on a six-port valve, while reading the pressure in the loop on a gauge, to know how much is really being injected.

What is the best way of performing a calibration to determine a % composition? I have a Scotty calibration mix with 4-5% of all gases above in helium, so all that I can do is vary the pressure injected and plot partial pressure of each gas vs. peak area. I did this and got some very nice, slightly nonlinear curves, however when I then use these curves to get partial pressures in my unknown, the total pressure is far less than what I read on the gauge! It is possible, but unlikely that this is really true. Is there a better way of doing this?

Several months ago, I did some very detailed calibration curves of %H2 and %air vs. peak area for varying compositions of 44.7 psia of air/hydrogen mixtures, which I can make myself, but this curve (a line with R^2 = 0.997) predicts a zero peak area for 22% hydrogen! And this was actually the case when I ran a sample with 22% H2. This, by the way, is in contradiction to the partial pressure vs. area calibration curves generated from the calibration gas.

Basically, your system seems fine ( assuming that you are evacuating all the way back to the cylinder valve to clear the line, rather than just purging, and you are allowing an equilibrium to form in the loop.

Such a small loop implies the samples are at high pressure, in which case you should match the pressure with your standard mixture. I'd actually use a larger sample loop ( 100 - 500 ul ) for near-ambient pressures, but it depends on your system, and loop pressure ).

The problem appears to be your calibrations, and possibly the fact that your standard mixture contains helium. I assume your system is leak-free, and that you are purging the sample lines.

Provided the sample doesn't change, you may find using a constant sampling pressure will give more consistent results. I would definately want at least one standard mixture concentration and composition that more closely reflects your sample composition and pressure.

If you know the range of your samples, it's good to bracket them. You can always use dried ambient air for the O2 and N2, taking care of where the argon could appear on your column, which I'm not familiar with.

The alternative is to use different-sized sample loops, which if you have, I would strongly recommend using to investigate your problem and the linearity versus pressure for each compound calibration.

The issue with the hydrogen may simply be the well-known behaviour of thermal conductivity detectors and typical GC carrier gases with varying hydrogen concentrations in samples. The hydrogen thermal conductivity is so different that as the concentration increases the conductivity passes through the baseline. There are carrier gas blends that overcome the problem.

Please keep having fun,

Bruce Hamilton

[rekuci]

It's a little tough to explain the whole setup, but so far it's worked really well, the problem is just coming up with a calibration method so I can back out % composition. I realize using capillary GC is strange for this!

The samples are collected in little fixtures consisting of a 10 mL cylinder, a bellows-sealed valve, and a female quick connect fitting.

I have a six-port valve installed before the GC inlet. In the 'load' position, there is a continuous line (with various fittings and tubing) as follows:

vacuum pump -- bellows-sealed valve -- sample loop -- pressure gauge -- male quick connect fitting

So I can hook up the sample by the quick connect fitting, evacuate the entire loop line including dead space before the valve holding the sample in the cylinder, then close the valve between the loop and the vacuum pump, and introduce the sample to the loop by opening the valve before the cylinder. The pressure read on the gauge at this point should be the pressure of the entire line including the loop, so when I move the six-port valve position to 'inject', 10 µL of sample at the read pressure is injected. I'm not really controlling pressure - the cylinder is at a certain pressure, and when I open it up to the loop line, a slightly lower pressure results due to the larger volume. Theoretically, I know the absolute quantity of gas that is injected.

There isn't enough of each sample to purge through the loop, which is why I had to come up with this evacuation method...I can get it down to 28 in. Hg vac and purge with air at atmospheric pressure in between runs, and I've also done runs on the composition of the 'evacuated' loop, there's obviously a little air left over but it's very little.

For concentrations, they're on the order of 5-50% each. I don't need nearly down to the ppm level...I'd be happy to get within +/- 2%, and the collected samples are typically 5-50 psia (yes some of them are at reduced pressures relative to atmospheric!). Do I need to calibrate with mixtures with different compositions or can I use a single composition with different loop pressures (like what I'm doing with the 4-5% calibration gas)?

I might have to come up with more conspiracy theories about thermal conductivity of multicomponent mixtures with hydrogen...

[rekuci]

Thanks for your reply. Actually the loop pressure is not necessarily high and is often below atmospheric pressure, and it can vary quite a bit from sample to sample. The last set of unknowns that I analyzed ranged from 18.5 in. Hg vacuum to 18 psig. It isn't possible to get a consistent sampling pressure, or at least I don't know how - the samples are just collected during a reaction in a closed system, and the pressure changes so quickly during this reaction that I don't really know the pressure in the collection vessels until I hook them up to the GC loop line.

The last calibration that I did using the gas with 4-5% of each component in helium, the pressures analyzed were representative of the range seen in the unknowns. I plotted partial pressure of each gas vs. area, used this to back out the partial pressures in the unknowns, and then added the partial pressures back up in the unknown - the result is far less than the measured pressure in the loop. The gases in the sample result mostly from high temperature pyrolysis, so I think it's very unlikely that the sample is mostly larger species that aren't eluting (and nothing shows up randomly on repetitive runs, maybe it would just get stuck in the column).

I do have 3 different sized sample loops, although it is quite an effort to change this out each time. It seems like the varying pressure calibration with the 4-5% gas is more representative of what's really happening (the ratio of the gases in the unknowns doesn't seem to change that much). What do you think? Is this partial pressure method I made up even a valid way of calibrating? I can't seem to find any guidance on the 'correct' way to do this.

[Bruce Hamilton]

You're sure that your samples are homogeneous and not multi-phase or containg water vapour?

I was suggesting changing the loop size to investigate the linearity and the fact that your samples could be in a different gas than the standards, not for all future work .

Given that your analyte gases have differing thermal conductivity, the only way to identify you propblem is to perform calibrations using standards over the same range as your samples.

What you could try is evacuating your sample containers and filling them with standard ( or compressed air ) at different pressures.

Although rather dodgy, you could put a needle vave on a larger sample container and collect more. Then you should be able to vent subsamples into your loop at a range of different pressures.

It's all about playing until you understand the cause of the apparent mass loss, which is probably a calibration issue. I'd still want a standards at different concentration, but even air would help confirm N2 and O2 behaviour.

Please keep having fun,

Bruce Hamilton

[rekuci]

Thanks so much for your help, I think I've bored everyone else away long ago. The calibration standard is in a helium matrix (none is available without special order in argon), which differs from the samples. Is this critical? Since I don't see any sign of a helium peak, I figured it's probably just flying through the column... (in my current setup, hydrogen elutes at 3.5 min)

What you suggest about running the standard at different pressures with an evacuated sample container is exactly what I did (and there is enough that I don't need to keep refilling it to get a range of pressures representative of the unknown samples).

So it sounds like the ideal calibration will be to get a few different compositions of the gases made up in argon, and then inject a range of pressures representative of the unknowns?

And this is probably a dumb question, but is this thing I'm doing with relating peak area to a *partial pressure* correct? Since the loop volume and temperature is constant, the partial pressure should be an absolute quantity like moles. Every calibration method I can find simply gets a "response factor", often from a single run of a standard, which is area/mol%, but the mol% in my standard stays constant and only pressure changes! Or should I find some kind of response factor as a function of pressure? Doing this gives drastically different results than the partial pressure vs. area method (that seem more wrong).

[Bruce Hamilton]

It's over two decades since I used partial pressures to calculate gas compositions ( it used to be the default method for natural gas here ), so I'm probably the wrong person to respond.

I assumed you wanted partial pressures for some further calculations, if you don't, then I'd lose them very quickly.

In general, the normal method is to calculate an area response based on the mass or moles of the analyte in the injection loop, so halving the pressure should halve the analyte area counts, which should result in the same response factor. Mole % or volume % are often the preferred unit for final calculations, but mass can be convenient, depending on how standards are certified.

Because each gas component has different thermal conductuvity, the total area count will change with sample composition. What we used to do was inject pure standards of each gas, then mixtures at 10 and 5 mole %. If you samples are in an inert carrier, eg argon, you should calibrate for that as well.

Each analyte would have it's own response factor curve from 5 - 100 % based on response factor/moles injected, and other calculations would derive from that.

Your procedure for using different pressures should provide similar data, provided peak counts for sample peaks are within the standard range ( increase standard pressure if you can't get higher % standards ).

Once you'ce calculated all the mole % (or volume % ), including inert carriers, it should add up to 100%, but water and heavy gas condensation can really mess up the calculation in process samples from reaction rigs.

I hope this is helpful,

Bruce Hamilton

[rekuci]

I think I just figured out the entire source of confusion.

The problem was that it looked like hydrogen was not nearly as much as it should have been when I used the calibration curve for the standard gas vs. when I used my prior calibration with an air/hydrogen mixture.

The standard is in a helium matrix, which I assumed was eluting too fast to detect. But if the helium is co-eluting with hydrogen, it'll vastly inflate the peak area I've been attributing to hydrogen!!! This is so obvious now...and I absolutely need a standard in argon matrix. I need to go right now and inject some helium and see where it elutes. This is what happens when you blindly assume something.

[pi3832]

I think I just figured out the entire source of confusion.

The problem was that it looked like hydrogen was not nearly as much as it should have been when I used the calibration curve for the standard gas vs. when I used my prior calibration with an air/hydrogen mixture.

The standard is in a helium matrix, which I assumed was eluting too fast to detect. But if the helium is co-eluting with hydrogen, it'll vastly inflate the peak area I've been attributing to hydrogen!!! This is so obvious now...and I absolutely need a standard in argon matrix. I need to go right now and inject some helium and see where it elutes. This is what happens when you blindly assume something.

Bingo. I was thinking the same thing when you said earlier that you never saw the helium but that the hydrogen took 3.5 minutes to elute.

Definitely go with an argon-balance standard.

[pi3832]

And this is probably a dumb question, but is this thing I'm doing with relating peak area to a *partial pressure* correct? Since the loop volume and temperature is constant, the partial pressure should be an absolute quantity like moles. Every calibration method I can find simply gets a "response factor", often from a single run of a standard, which is area/mol%, but the mol% in my standard stays constant and only pressure changes! Or should I find some kind of response factor as a function of pressure? Doing this gives drastically different results than the partial pressure vs. area method (that seem more wrong).

Intuitively, I don't think what you're doing works. Need to play with the math more, though.

Have you run through the calculations on something other than hydrogen, which is now suspect?

[pi3832]

And this is probably a dumb question, but is this thing I'm doing with relating peak area to a *partial pressure* correct? Since the loop volume and temperature is constant, the partial pressure should be an absolute quantity like moles. Every calibration method I can find simply gets a "response factor", often from a single run of a standard, which is area/mol%, but the mol% in my standard stays constant and only pressure changes! Or should I find some kind of response factor as a function of pressure? Doing this gives drastically different results than the partial pressure vs. area method (that seem more wrong).

Intuitively, I don't think what you're doing works. Need to play with the math more, though.

Have you run through the calculations on something other than hydrogen, which is now suspect?

Belay that. Did the math. You're right. It works out. You can develop an RF that when multiplied by the area counts gives you the partial pressure of the analyte in the sample.

[rekuci]

Just an update - right after my last post I ran some 100% helium and it was eluting at 3.5 min just like hydrogen. DOH!

Everything seems to work out if I use a separate calibration for hydrogen (from air/hydrogen mixtures), but there is still the strange artifact of having zero peak area for hydrogen at 22%, the lowest concentration I've run yet. I really need to run lower than 22% and see if the peak is just zero or if it turns negative.

Also, now when I add up the partial pressures, the total of the 6 gases is now MORE than the measured pressure (and the lower the measured pressure, the bigger the difference between measured and back-calculated numbers). My first thought was that if you have a sudden expansion of an ideal gas, as is happening when I open the valve between my sample and the line with the sample loop, you'll have a decrease in temperature which would make your pressure look lower than it really is at room temp until it thermally equilibrates...but to account for some of these pressure differences, the temperature in the loop would have to have been -70C or less! Obviously that's not happening, so I'm not sure what's going on, exactly.