Problems making a calibration curve for CO2

[jsfoodbk]

Hello from Uruguay, I´m trying to make a calibration curve for CO2. I am using an Alltech® CTR I column on an Shimadzu GC 14 A/B, using He as carrier and TCD as detector. The oven temperature is 60ºC. The inject method was manually with a syring (1 ml).

To obtain the different points of the curve I took diferent volumens from a jar filled with a gas mixture and completed the volumen with N2. The curve was linear but it didn´t pass by zero. So I repeated the procedure with other gas mixture, and the same happend.


In the image I attached you can see, what I tryied to explain. The chart shows the CO2 peak area vs the sample concentration. The above curve was did with a gas mixture with 15.1% CO2 (20%O2 and the rest N2). The second curve with a gas mixture with 5.6% CO2 (6%O2 and the rest N2). When I inject air (0.04% CO2) the CO2 peak area average is 200, so both curves are not correct.

So I have two problems:

1) The curves don´t pass by zero.
2) The curves have similar slope but different y-intercept

Is there a problem with the GC? the column?, or is the procedure I used to generate the differents point is bad?

I hope you can help me and sorry for my english.

[chromatographer1]

There doesn't seem to be anything wrong here except your assumptions.

You are assuming that the gases when diluted with nitrogen behave ideally. CO2 doesn't act ideally in my experience. But this is probably a small factor in the discrepancy of the data.

You are assuming that the dual (column inside a column) CTR I column will sample (that is split) consistently with two different blends of gas. From the points generated, this looks like it is true.

You are assuming the blends of gases are accurately made AND you are sampling them precisely in the same mole percent in which they were prepared. Not knowing the pressures involved in the bulk tank I can only speculate this may not be true. It certainly accounts for the difference in the x-axis intercepts.

I must say you do very nice work in consistently sampling and injecting your gas samples. I am impressed.

My conclusion then is you are not sampling gas from the main cylinders at its documented concentrations (sample discrimination) or the bulk cylinders have been sampled enough so that their contents are not accurate to the original composition at which they were originally prepared due to sampling discrimination conducted previously.

And don't forget to account for losses of CO2 on the tubing of the column when injecting very dilute samples, as in the 400 ppm air sample you injected.

What to do? Inject your samples and calculate the amount of CO2 present from the slope of your calibration.

Nice work.

Rod

[Johnny Rod]

Do you accurately know the volume of the glass flask you are using to make the standards?

[jsfoodbk]

Thanks rod for the answer. Like always in science one answer brings more questions.

You are assuming that the dual (column inside a column) CTR I column will sample (that is split) consistently with two different blends of gas. From the points generated, this looks like it is true.

Yes, I´m begging that. How do you know it form the points?

You are assuming the blends of gases are accurately made AND you are sampling them precisely in the same mole percent in which they were prepared. Not knowing the pressures involved in the bulk tank I can only speculate this may not be true. It certainly accounts for the difference in the x-axis intercepts.

Yes I´m assuming that I´m sampling them precisely in the same mole percent in which they were prepared. My thought was that no matter the pressure, the mole percent will be constant, so I can fill a flask and use it like a tank. It is a bad assumption?

Is better to take the sample from the bulk tank? what happend with the pressure? I need to inject at 1 ATM pressure, is the pressure an important factor?

I must say you do very nice work in consistently sampling and injecting your gas samples. I am impressed.

Thanks

And don't forget to account for losses of CO2 on the tubing of the column when injecting very dilute samples, as in the 400 ppm air sample you injected.

How can I account that?

What to do? Inject your samples and calculate the amount of CO2 present from the slope of your calibration.

But how can I do it with only the slope? (when I say slope I mean [y=ax+b] only the a factor, are you reffering to the same?)
Maybe I need one calibration cuve for each gas blend? It has no sense for me, because the GC must determine CO2 independently from another gases present in the sample.

Do you accurately know the volume of the glass flask you are using to make the standards?

No, but it is about 4-5 l

[chromatographer1]

chromatographer1 wrote:
You are assuming that the dual (column inside a column) CTR I column will sample (that is split) consistently with two different blends of gas. From the points generated, this looks like it is true.

Yes, I´m begging that. How do you know it form the points?

You produced straight lines over a range of concentrations using two different concentrations of calibration gas. You would have curved lines, or two lines with different slopes if otherwise.

chromatographer1 wrote:
You are assuming the blends of gases are accurately made AND you are sampling them precisely in the same mole percent in which they were prepared. Not knowing the pressures involved in the bulk tank I can only speculate this may not be true. It certainly accounts for the difference in the x-axis intercepts.

Yes I´m assuming that I´m sampling them precisely in the same mole percent in which they were prepared. My thought was that no matter the pressure, the mole percent will be constant, so I can fill a flask and use it like a tank. It is a bad assumption?

Yes it is. If you mixed oil and water in a flask would you assume you will always sample the two components consistently from the top? Do you think because it is a gas this discrimination will never happen if CO2 is involved? What if you were discussing water as a gas. If the cylinder was at -40°C and 120°C would you think you would measure water at the same concentration at both temperatures?

Is better to take the sample from the bulk tank? NO. Just the opposite.

chromatographer1 wrote:
And don't forget to account for losses of CO2 on the tubing of the column when injecting very dilute samples, as in the 400 ppm air sample you injected.

How can I account that?

CO2 will coat (bind to) the metal surface and you will have losses from any sample passing through the metal tubes. Small losses certainly, but unused metal surfaces may lose measurable amounts of CO2 during initial injections of air. Depending upon the equilibrium this loss should drop after each following injection of CO2. Again this loss should be SMALL, barely measurable if at all. Don't worry about it.

chromatographer1 wrote:
What to do? Inject your samples and calculate the amount of CO2 present from the slope of your calibration.

But how can I do it with only the slope? (when I say slope I mean [y=ax+b] only the a factor, are you reffering to the same?)

It is simple math. Divide the area by the slope and you get the concentration of that point. x = y/a You can ignore the b factor since this is the 'calculated' concentration of the contamination of the dilution gas with the analyte or the error in the assignment of concentration of the initial gas sample)

Calculate the concentration of the gas in each blend from your injections slope and see how they differ from the documented concentration.

Maybe I need one calibration cuve for each gas blend? It has no sense for me, because the GC must determine CO2 independently from another gases present in the sample.

You should get the same slope for any CO2 gas blend you use as long as you accurately dilute the calibration gas and sample and inject accurately. You showed that already with two calibration blends when diluted precisely the gas response was precisely lowered at the same factor as the dilution factor (You got a straight line when plotted).

You have calculated a response factor for CO2 twice now, they match almost exactly (370989 and 375948). Use the response factor to calculate the concentration of the CO2 in the gas. Try to think using the units of your equation, don't ignore them. Perhaps that will help you understand more clearly.

best wishes,

Rod

[jsfoodbk]

then, what is the best way to take a sample??

About the curve, Sorry because I didn´t explain it very well. In the y-axis there is the CO2 peak area (mv), and in the x-axis there is the % of CO2 in the sample.
To build the curve I assumed that the blend A has 15.1% of CO2, so the peak area of that injection is 15.1% in the curve. To build the others points, I filled a percent of the syringe and the rest with N2. This percent multiplied by 15.1% is the x-axis valor of the peak area of that injection. For example, I filled 50% of the syringe with the gas blend and completed with N2, so I assumed that the CO2 percent in the sample was 15.1%*50%=15.1/2=7.55% of CO2.

So if I change the 15.1%, it changes all the other points, and don´t solve the problem. By the way I built the calibration curve, I suppose that necesary the curve must pass by zero, so I think there is another problem causing this differences, could be?

[chromatographer1]

You don't need to change anything. I just thought you might like to see how the slope calculation would compare to the documented value of the gas blend.

You explained the curve quite well, and I understood clearly what you were stating.

You do not need to explain the error of the blends. You have demonstrated clearly an accurate response for CO2.

If your problem is to measure the % CO2 in gas samples your solution has been reached. If your problem is to generate a plot that intercepts zero based on the documented value and the present condition of the gas blends in your cylinder that is another issue.

"By the way I built the calibration curve, I suppose that necesary the curve must pass by zero, so I think there is another problem causing this differences, could be?"

If 1) the CO2 content was accurate and 2) your sampling of the cylinder accurate then the curve SHOULD pass through zero. But with both gas standards this did not happen.

One of these two assumptions is incorrect and is the cause of your non-zero x-axis intercept and your concern about the calibration plot.

You decide which one or both are contributing causes of the bias. If you cannot decide which or both is the cause it does not matter.

You can still measure the CO2 content of your samples from your present calibration curve with a small error. Please note that if the documented value is wrong on the low side then your plot will not intercept zero but will look exactly like the plot you generated. If your cylinder has been sampled many times before then that supposition is quite possible.

Use the slope of the smaller concentration plot (the line intercepts closer to zero and thus should have a smaller error in its slope value.

Let me repeat: I do not think you have done anything wrong in your procedure. Are your blends certified? Are they fresh or weeks or months old? What is the error range from the vendor, ±2% or ±5% ?
Have they been sampled many times previously?

Have you considered getting a 100% CO2 cylinder and performing a calibration with it?

You have demonstrated that your detector is linear and you can sample your gas sample in a precise manner.

Congratulations. Now use the slope for calculations of CO2 percent content and get your work done.

best wishes,

Rod

[Johnny Rod]

As Rod says your slopes are consistent so it's looking good. You don't seem to have any points conrresponding to 0% CO2, have you run blanks? This could be why you have an offset.

[JI2002]

If you want to measure CO2 in ambient air which is about 300 ppm to 400 ppm, then you should make the calibration curve from 100 ppm to 20000 ppm (2%). It will fix both of your problems. Based on my experience, with 1 ml injection volume, CO2 above 5% may saturate the detector.

[chromatographer1]

jsfoodbk

Do you think your questions have been answered satisfactorily?

Are you still concerned?

Have you been helped?

best wishes,

Rod

[jsfoodbk]

Please excuse me for the delayed answer. I was dealing with another problem. I think you have given me good answers but to totally solve the problem I must run some more test. But I can´t in this moment. We are going to buy some pure CO2 but probable not in a short time. When I can I will try to compare results with another GC and view if there is a difference.

One thing in my head is about the curve, I don´t think that is correct only use the slope, but how I can´t be sure that the concentrations I used to built the curves are correct? I can´t assure that until I get some certified gas blend or pure CO2.

Somebody asked if I ran blanks, the blanks gave me 0, and the air (400 ppm CO2) around 200 mv.

Why I don´t think is correct only use the slope?
1)I´m not sure about blends concentration
2)If I use the slope in the data that I used to built the calibration curve it doesn´t give the original values, so I don´t return to the origin point.

chromatographer1, thanks for all your help

[chromatographer1]

" Why I don´t think is correct only use the slope?
1)I´m not sure about blends concentration
2)If I use the slope in the data that I used to built the calibration curve it doesn´t give the original values, so I don´t return to the origin point. "

Response:

1) That is always the case unless you buy from a reputable supplier and the cylinder is unused. Even then you may have to 'stir' the bottle before using. (and I hate doing this sometimes necessary procedure)

2) - TRUE If the blend is off you don't want to use the curve (regression line) itself.

HOWEVER, using the slope WILL give you an answer that is inaccurate to the same degree as the concentration given differs from the true value, so it COULD be quite close to the true answer. Remember, it could only be off by 10% or even less. Only you can decide if that is a problem. The good news is that your results appear linear, so your answers will be COMPARATIVELY accurate.

Good luck in your work.

Rod