Low Conc. Water Detection

[polster]

I've read that for certain concentrations of hydrogen, separation in helium can be difficult to quantify by TCD because of a shift in thermal conductivity of the hydrogen/helium mixture.

I think I'm seeing a similar phenomenon in separation of low ppm water mixtures, where I get a negative response for very low concentrations that then shifts to a positive response as the concentration pushes higher.

I wanted to know if anyone else has seen something like this, and if it is possible to accurately quantify water in this case. Thanks.

[chromatographer1]

Chris,

I have never seen what you describe and I have done analyses for ppm water at very low levels and performed linearity tests for validation.

But your data would interest me if you might share more details of your analysis and perhaps some chromatograms of the negative peaks changing to positive peaks. Please describe what was injected on your column before your demonstated chromatogram series. (What sort of sample matrix)

thanks.

best wishes,

Rod

[HbJ]

I second what chromatographer1 said.

Please tell us about your sample matrix, column etc.

I've done water analysis for higher (around 200 ppm) concentrations with a satisfying degree of precision (boss was happy).

So if you tell us more about your analysis we may be able to help


All the best,

HbJ

[polster]

My samples were largely composed of hydrogen and helium, with low concentrations of oxygen (~50 ppm), carbon monoxide (~100 ppm), carbon dioxide (~10-20 ppm), and of course water. I am also able to get satisfactory data for water concentrations of around 200 ppm, but I am more concerned with concentrations of a few ppm and lower. My sample baselines have no water in them (with the exception of what little amount may sneak through various filters I have), and the chromatogram exhibits a slight negative peak. As water concentration climbs, I notice a peak forming, eventually dwarfing the original negative response. The negative peak is very small, but nonetheless present.

This is an example of what I see with "zero" water.


Here is what it looks like with very low amounts of water. This is ~6 ppm by my analysis.


The samples are analyzed using an Agilent MicroGC, using a plot Q column, helium carrier at 120 C and 20 psi. Elution time is roughly 2.2 minutes. If you need any more information, let me know. Otherwise, any insight into the issue would be helpful.

Thanks.

[chromatographer1]

Chris,

Your negative peak may be an artifact of the injection valve, a pressure pulse. It could also be something caused by a mechanical leak in your valve eluting air into your carrier gas at trace levels which when you inject your sample gas stops leaking in the inject possition but then returns to leaking in the carrier when in the load position.

It would appear that your water peak elutes on top of this negative artifact and that you can measure water accurately if you can integrate accurately.

That is my guess.

Rod

[polster]

What I have been doing for quantification purposes is subtracting this negative response from my samples prior to integration. Does this seem like a reasonable quantification method?

[chromatographer1]

Chris,

Normally if your baseline returned to its previous level I would subtract the negative area (add a positive area) from the positive area measured by the integrator for the water peak. If your negative area was 500 and the water peak area was 200 I would count that peak as 700.

However, your 6 ppm water chromatogram indicates the baseline did not return to its previous level. Is this reproducible? or does your baseline shift after every elution of a water peak? This may be indicative of a hardware problem.

Your column length in a MicroGC may only be 3 to 10 meters in length. How much separation do you have from the other peaks in your sample?

I would expect at 120°C and with a fast linear velocity, the amount of separation might be minimal.

You are welcome to send me additional chromatograms off forum to my work address. I would be glad to try to resolve your baseline problem if I can.

Rodney George
Senior Research and Development Scientist
Gas Separations Research
Supelco
595 North Harrison Road
Bellefonte, PA 16823

814-359-5737 voice
814-359-5459 fax
rgeorge@sial.com

[polster]

Sorry for the slow reply.

Given my sample composition, I get excellent separation from other species. H2, O2, and CO2 all elute around 1 min, with CO2 being fairly distinct, only a little bit on a shoulder. Water elutes around 2.2 min, with nothing else in the vicinity. I know you're right about the length of my column being relatively short, but separation isn't a problem.

The baseline does drift slightly after every water peak. It will recover, but it takes some time.

When you say to add the negative area to the positive area, I'm sure of what the negative area is, but how should I integrate the positive area? Should it be from the bottom point just before the peak to where it returns to the baseline?

[chromatographer1]

Chris,

I would determine if the negative peak area when injecting a pure dry sample of gas is reproducible first.

If it is, then I would use the level baseline to determine the area (A) of this negative peak (or baseline upset).

When injecting a sample with water in it, if you still have a small negative peak (B) say half the negative area of the negative peak previously established, I take the smaller negative peak and subtract it from the original negative peak (A-B) and call that the area of my water peak (C).

If your water peak increases to a size above the level baseline then I would add that positive area peak (D) to the peak area (C).

At some point there will be no negative peak at all. Then I would add the area of the negative peak to the positive peak to give me the true area of the water peak.

Example: negative area of baseline upset = 500 counts

1ppm water peak (smaller negative peak) = 250 counts (250 counts for actual water peak)

10 ppm water positive (no negative) area = 2000 counts (2000+500) counts) or 2500 counts actual area

I hope this makes sense.

best wishes,

Rod

[polster]

I believe I understand what you're saying. I just want to clarify it for the example I show above for 6 ppm. If I have some negative area (E) and some positive area (F), I should just integrate everything from the baseline, subtract the negative area from A and then add that to F? So (A-E) + F?

I can give that a try and see if I get a linear response. I can compare it to my current method and let you know what works best.

Thanks.

[chromatographer1]

Yes, Chris, just be careful how the software draws the baseline.

I think we have a common understanding.

You may have to perform more than one integration in order to measure the negative peak accurately and measure the water peak above the level baseline accurately.

Then if you measure the negative peak E and then measure the positive peak F above the level baseline, then A-E + F = the actual water peak area.

Remember one thing:

if it were easy, anybody could do it.

best wishes,

Rod

[aeidt]

Polster,

Your negative peak when you analyze a completely dry sample likely arises because your injected sample is actually drier than your carrier gas. The TCD measures the thermal conductivity of the column effluent relative to a reference carrier stream. If there is less water in your injected sample than in your reference stream, a negative peak results.

The problem can be solved by placing a high efficiency moisture trap on your carrier gas stream, with as few fittings/connections in between it and the GC as possible (water tends to ingress at any connection or elastomeric seal in your system). Water is everywhere and hard to keep out of your system. Alltech sells a good moisture trap (part # 7036).

Hope this helps.