Chromatography Forum

Hi,
I'm new to gas chromatography and I've been having systemic difficulties using an external standard as a means of quantifying samples. Its a GC-FID setup, and i get excellent linearity when looking at the calibration curve, but when i go to analyse a sample of the reference material it invariably comes out about 5% lower than it was specified. Area normalisation shows the specified value to be correct so I'm not sure whats going wrong. Reproducibility is fine, I suspect I'm doing something wrong in the way I've been calculating the calibration curve.
I weigh out a sample, then dissolve it in a weighed amount of acetone - near enough to ten ml in a volumetric flask. Then based on the stated purity of the reference i calculate the concentration by multiplying by 0.97 (97% reference purity). Then divide this figure by the total volume.
Anyone have any thoughts on a better system? I'd be grateful for any advice

What are you using to generate the calibration curve ? Presumably it is the same chemical substance, made up to a range of different concentrations. What solvent do you use for this, and what is the purity of the substance that you use to generate the claibration curve ? Do you make up standards by weighing in the same way as you dissolve the reference material ?

What do you mean by area normalization, and how do you calculate from that to show that the specified value is correct ?

What is the reference material - is it a (nearly) pure chemical (97%) or is it a sample in some kind of matrix ? I suspect the former, but you talk about corecting the concentration of a solution of sample by correcting for a 97% stated purity of a reference .

Peter

sextonrick,

I apologize, but I don't quite follow what you are doing. You weigh the sample, say 100mg and then dissolve it in acetone (say 12grams.) So far, no problem. Then multiply by 0.97 (which, like Peter, I don't understand why, if a sample.) Then divide by total volume? Now you have mg/grams/mL?

Sorry but I am a bit slow on the uptake here.

Best regards.

Sorry about the ambiguity of my first post.
To generate a calibration curve I am using a 97% pure reference material of the same compound i am trying to quantify. The calibration curve is based on area response. I'm using 4 points 50,100,150,and 200ul microlitres, which i also weigh to check the accuracy of my pippeting. These are made up to ten ml in volumetric flasks using acetone.
from this i get concentrations in mg/g or using the density of acetone (probably a flawed step) mg/ml.
To account for the reference material being only 97% pure, i had been multiplying the reference weight value by 0.97. for example if i had 50mg in a volumetric flask i would assume that only 48.5 of this was the desired compound. Then i would use these corrected values to calculate the concentration which i would then use to make a calibration curve.
From the curve I used the software to calculate the concentration of my unknown.
Hope this helps

Lets make some standards for a calibration curve:

80% of nominal or Low STD = 8 mg x 0.994 (99.4% Potency)/100 mL = 0.08 mg/mL

100% of nominal or Mid STD = 10 mg x 0.994/ 100 mL = 0.1 mg/mL

120% of nominal or High STD = 12 mg x 0.994/ 100 mL = 0.12 mg/mL

Note: When making the three standards as above, they were corrected for purity or potency.

They will be used as External standards in generating the calibration curve.

The generated sample concentration, say 0.11 mg/mL x 100 (dilution factor) will give your answer.

Hi Rick

Your calibration procedure looks OK (weighing voaltile solvents has some procedural issues but we can leave them for the moment).

You say that your result for a sample is 5% low - I presume that the sample in the first post is the same as the unknown in your second post. How do you know the answer that you "should" get ? - in other words how do you know the concentration of the unknown is 5% above your experimental result ?

Peter

Hi Peter,
- The product that we were testing was specified to be 90% pure by those who made it. We also had another company perform a GCMS of it and it came out at approximately 90%.
In a further crosscheck i ran the sample neat with a high split ratio and then used the software to look at the areas - without the use of any external standard. In this test the compound of interest again came out at 90%.
The calculations from the external standard were giving values of around 85%. I think there's probably something going wrong with the process somewhere

Hi Rick

The value that you get for purity when you take the fraction of the total area on a chromatogram that is due to the compound of interest is the maximum purity for that sample, because there may be impurities present that are not detected by GC-MS or GC-FID. By far the most likely is water - can you do a moisture determination ?

Can you please post in step by step detail how you are making up your solutions for both standards and samples - with a volatile solvent like acetone you can easily get 5% negative bias if you are weighing by difference.

Peter

Hi Peter,
This process is the same for my reference points and unknown samples:
I take a clean and dry ten ml volumetric flask and weigh it, I then tare the scales. Then i pippette the amount of reference / sample and weigh the amount pipetted.
I then fill the volumetric to the 10ml point with acetone and weigh the acetone.
I thoroughly mix the solution with a pasteur pippette, then decant some into a 2ml GC vial.
Thats it!
Thanks for all your responses

Hi Rick

The devil is in the details.

When you say that you mix the solution with a pasteur pipette do you suck up into the pipete and squirt out again a few times, or do you stir it with the tip of the pipette.

What is you split ratio, and the rest of the GC conditions while we are about it ?. Your solutions are quite concentrated (e.g. 20 ug/ul), so if you have a low split ratio you might be getting into a non-linear response from the detector (although this would probably show as poor linearity on the claibration line).

How do the calibrations and the results look if you work volumetrically rather than by weighing ?

How much sample (ul or mg) do you dilute ?

What is the resolution on your balance - i.e. how many decimals ?

How repeatable are the areas for replicate injections of a single solution, and for replicate solutions at a given level ?

Peter

Hi Peter,
I mix the samples by repeatedly sucking up the solution with a pasteur pipette. I run a split ratio of 50:1, injection of 1ul and an isothermal oven temperature of 190C. injector temp 250C, FID 250C. constant linear velocity program (not that this would make a difference with the isothermal ramp).
I'll have a look at how the numbers and calibrations work out when based on volume, but I don't think it will make a big difference.
The balance i use measures 4 decimal places.
In unknown samples i use 100ul diluted up to ten ml in acetone.
Repeatability is fine, although I'll have a look at the data for actual figures.

Thanks

Hi Rick

As far as I can see there is nothing wrong with what you are doing. From the total area calculations the sample is not more than 90% pure, and that agress with the MS results (exactly how was that done ?) so it is entirely possible that the 85% result is accurate.

What about moisture ?

Peter

If we consider this backwards, if we assume that where your calculated concentration is low, would be indicating that your standards are too high. Are your purity specifications possibly > specs, meaning that maybe when you adjust the amount down, you're over-correcting?

What we're really trying to figure out is why are your standards coming out too high, and reproducibly too high (since your calibration is very good). If the error was due to some type of random error happening, we would tend to see increased error (scatter) in the calibration. If it is in fact tight, then we're trying to sort out something that is systemic; hence my interest in the purity specifications.

I'd recommend checking the certificate of analysis for your STD material. As well as a % Purity (by Technique) there should be moisture content (and possibly some additional % impurity data).


Calculate a "Potency" value = (100 - total % impurities) * % purity

Do your results improve if the Potency value is applied to the calibration data?

Have Tonto take care of it???