CO2 Sampling - Unexpected Correlation Between Measurements

[MWeaver]

All right this is my first post. I'm grad student doing research on greenhouse gas concentrations, currently I am trying to measure CO2 mineralization rates from soil samples obtained in high desert mountain systems. I am measuring CO2 rates obtained from the static headspace of 125 ml flasks containing about 10 mg of sample soil. Each day I sample I first flush the flasks with room air next I take a series of headspace samples, typically three over the course of 2-5 hours. I then plot concentration of CO2 vs. time to determine a rate of mineralization or creation of CO2 by the soil. Although I have some experience and have researched the GC to a large extent I would consider myself a neophyte in its use.

Here is my setup:
I am using an SRI 8610C GC with an FID sensor, and a CTR-I column from Alltech to measure CO2. I process samples at a constant 60 deg-C. I use Helium for my carrier gas and air and hydrogen as fuel for the FID. I process samples for about two minutes. I inject samples into my sample loop with gas tight syringe. I calibrate the GC with a multi-segment line using three points 0 ppm, 998 ppm, and 5000 ppm, my calibration range covers my anticipated sample concentrations. I run samples one day a week, and calibrate each day samples are run, it ususally takes an entire day, and the GC is not on or used at any other time of the week. At the end of my sample day I run my standard gases again and usually get closure within 4-5%. I run blank samples with every round of samples and the concentration of blans vaires by ~10% and area counts vary by 5% week in and week out.

My problem:
I apologize for my lack of vocabulary but the problem is I see a correlation between changes in concentration from one week to the next between my individual soil samples that are not anticipated. I have ruled out ambient influences that could be causing the correlation, i.e. changes in soil moisture, changes in temperature, sample handling, etc. I believe whatever is introducing the correlation is coming from the GC and/or my use of it. The following graph depicts CO2 mineralization rates (ug/g/day) from two different soil samples and the inverse of the area counts of the 998 ppm standard gas. The correlation (rises and falls from week to week) of the values is undeniable and I feel not representative of the real system. What could I be doing wrong with my setup and/or use of the GC that could lead to this type of correlation, i.e. I would not expect to see any relationship between the rise and fall of CO2 mineralization rates from different soil samples and the correlation to the inverse of the area counts of my standard gas suggests it is coming from the GC.

Many thanks for any help.

Well I can't figure out how to attach a picture so I gues that's another questions. My picture was going to illustrate that over the course of 148 days the trends in changes in CO2 concentrations were identical between different soil samples for every round of sampling, i.e. when one concnetration went up or down on one sample it did the same up or down trend on the other sample.

[Bruce Hamilton]

I don't have any GC reference books handy to confirm, but I would have expected an FID detector to be rather insensitive to CO2 concentrations, consequently even minor changes in sample composition may have profound effects on response.

Bruce Hamilton

[Carvone]

Yeah, I was taught that an FID was best for organic compounds and would give a stronger signal based on more -C-C- bonds and to a lesser extent -C-H bonds, with a substantially minimal signal with -C-OH or -C=O bonds.

I think your trend of different soil samples following the same increases and decreases of COi2 reveals the detector's sensitivity variation that week (H2 and Compressed Air gas pressure changes perhaps), as mentioned by Bruce above. You may want to check into different detectors better suited for CO2 detection. For examples see below:

http://www.chromatography-online.org/GC ... /rs31.html

http://www.chromatography-online.org/GC ... /rs90.html

[MWeaver]

That is great feedback I really appreciate it. One thing I forgot to mention is the fact that I have a methanizer on the front end of the FID converting CO and CO2 to CH4 which allows for measurement by the FID. Does this change your comments? Also, I have a TCD detector in series with the FID and I could use it as well. I noticed a TCD was not one of the two detectors you recommended, would it be better suited for this experiment than an FID?

Thanks again for the informative comments.

[Bruce Hamilton]

OK, The Thermal Conductivity Detector is similar to the Katharometer.
Yes, the methaniser makes the FID much more sensitive ( and suitable ).

Depending on the CO2 concentrations, it would be useful to hook up the TCD ( presumably before the Methaniser/FID - if it's in series, but probably should be in parallel - if permitted by detector and instrument design ), as it would give an indication whether some other component in your samples is interacting with your FID response.

I suspect the problem may be your sampling technique, rather than the instrument, although you should ensure that you analyse the closest standard several times during a day and week to confirm it's not the instrument.

It may be that the sample water content is changing and affecting samples/instrument. Also, how do you ensure the composition of the vapour in the flask is homogeneous before sampling?.

Syringe sampling and injection is often a lottery, and have you confirmed repeatability on real samples - rather than standards?.

Please keep having fun,

Bruce Hamilton

[krickos]

Hi

This thread explains how to include pictures etc:

viewtopic.php?t=2617

[Peter Apps]

How, and when do you generate your standard mixtures, and how do you introduce them into the GC ?

What are you running as blanks ?

Peter

[AICMM]

Mweaver,

I'm sorry, I am a bit thick here, but why do you flush your samples with room air? How do you then determine mineralization rates? By being in the room you are affecting the concentration of the CO2 in the room so....

If you care to explain, that would really help me understand. Sorry for my ignorance.

Best regards.

[Ron]

AICMM, that is exactly what I was thinking, there may be some steps missing in the description of the sample preparation and the sampling, but I see a couple of potential problems. Room air CO2 is not necessarily constant from day to day, so I would use a gas mixture that has a known concentration of CO2. Secondly, are the reaction flasks being filled to a consisent pressure every time. If the flasks are not pressurized to above ambient pressure, the repetitive sampling from the flasks will cause the pressure to be lowered, and if the syringe does not have a valve then room air could be mixing with the sample in the syringe, biasing the results.

Methanizers can be a bit unstable, if flow of hydrogen is shut off while the catalyst is still hot conversion efficiency can be changed.

If you do a series of replicate injections of room air do you get good reproducibility?

[MWeaver]

Oops see below.

[MWeaver]

A lot of great questions. I'm amazed at how well you all seem to understand my set up, even if you don't quite know all the details. To answer some of the questions.

My standard gases came premixed in "k" cylinders from a local gas supplier (Norco). My introduction of the standards into the GC is different from my sample introduction, in that the tanks are directly plumbed into my sample loop, where as the samples are inject via syringe. I have taken great pains to make sure the flow rate and pause (to allow gas to decompress) before sampling are the same for both methods. That being said I realize it is not desirable to have different methods, but I do.

Because my samples flasks are a closed environment carbon mineralization can only occur for so long until the flask head space reaches saturation, pushes back on the soil environment and alters the kinetics of the mineralization reaction enough that the "bugs" stop breaking down organic carbon - or at least really slow down. Therefore, to measure the mineralization rate I need to first purge the headspace, or wipe the slate clean, so that optimum mineralization can occur. Because I have many flasks (48 total) that I am required to purge I just landed on room air because it was easy to come by and via a compressor made the purging step much faster. I have considered using an inert gas source but really my sample concentrations are much larger than background concentration of CO2 in the room air so I did not think changes in 25-50 ppm of CO2 week in and week out would make that big of a difference. And, the error I am seeing is much larger than the variation you would expect from changes in CO2 concentrations in the room.

After purging I wait about 10 minutes before I take my first sample from the flask, which is usually right around room concentrations, which is what you would expect. I then take several more samples from the headspace over the course of 2-3 hours. What I develop is a plot of CO2 concentrations vs. time. The slope of a regressed line through the scatter plot represents the rate of mineralization. I run my experiment with three replicates for each soil sample. I get amazingly good correlation amongst the replicates (RSV ~.999) within the same week, its just when you start comparing results from one week to the next that things go wonky.

My blanks consist of three flasks (to mimic three replicates of a soil sample) with no soil in them. I purge the empty flask the same as if they were soil samples, sample the head space, and inject them into the GC - all just as if they had actual soil samples in them. In essence my blanks are room air values and I get outstanding agreement and repeatability between my blanks each week and from week to week, as far as area counts are concerned (better than 95% agreement), however, my concentration readings from the GC only get about 88% agreement week in and week out. That could be due to changes in concentrations in the room, or maybe some other error.

From my reading I understand that syringe injection can be a less desirable way to introduce samples, but it is a simple and cheap way to go. I can't for example afford an autosampler. Does anyone have other recommendations for how to introduce the headspace gases from my flasks into the GC?

I'm in much appreciation of all of your comments and I always try and have fun.

Viva la science.

[Peter Apps]

How big is the sample loop and how big is the gas tight syringe ?. Do you use the same syringe all the time ? How many loop volumes do you flush through when you load the loop with standards ?

Please have another try at loading the picture.

Peter

[Ron]

Since you are seeing good precision short term and a change in response over time on both the blanks and samples I would look to the instrument to see if there is a problem there. I'm not trying to be negative about SRI, but their reputation is one of producing low end instruments at low cost. I suspect their long term reproducibility is not going to be as good as Agilent, Shimadzu, Thermo, Varian, etc. Is there another instrument from a different manufacturer around where you could run a few samples a week for a few weeks to see if the second instrument is more stable?

The experimantal design looks pretty good, I think it is now time to focus on the hardware instead of the method.

Good luck.