Recently, I quickly calibrated a method on a GC-MS, using headspace SPME, for the analysis of ethyl esters in a VERY complex matrix. I was surprised by the resulting QUADRATIC calibration curves. My calibration chromatograms look good (no tailing, etc), and all the components are correctly identified.
Is this QUADRATIC behavior surprising?
Thank you in advance for any suggestions.
Here are my extraction conditions:
-Internal standard calibration.
-Fiber: Supelco 50/30um DVB/Carboxen/PDMS
-Calibration range: 0-1000ppm ethyl esters (2 calibration levels of 500 and 1000ppm), from ethyl butyrate to ethyl decanoate.
-Extraction time: 15 sec (autosampler)
-Desorption: 3 min
-GC-MS calibration/quantitation: Selected ions
Note: Due to my very complex matrix, and to the level of analytes present, I optimized the extraction time to 15 seconds.
By Chris on Tuesday, April 9, 2002 - 11:21 am:
15 secs seems like a pretty low time for extraction-are you sure that is the optimal time? Also wasn't sure what your matrix was. If it is a liquid (water), did you try adding a salt? You might want to try directly injecting your standards (if that's possible)to see if you can get a linear calibration that way. That will prove that you need to further develop your spme method!
By Anonymous on Tuesday, April 9, 2002 - 11:42 am:
Remember, your SPME will only pick up what's actually in the headspace, and the longer esters are less volatile. We always get different profiles with fragrance headspace SPME than just injecting the same fragrance directly.
By Anonymous on Tuesday, April 9, 2002 - 02:47 pm:
I agree. 15 secs seems like a very short time for extraction. However, I tried 15, 30, and 60 secs, and 15 was the best (at 30 secs I started seeing poor chromatography on some analytes, already).
The matrix is water-based with a high pH (10-11). I tried adding a salt, but without success.
My chromatograms look good (I ran 3 injections of each calibration level). And yes, as I expected, the heavier esters, being less volatile, show a lower response than lighter esters.
However, I am still puzzled by these QUADRATIC fits for my calibration curves...
(So far, thank you both for your help)
By Anonymous on Wednesday, April 10, 2002 - 09:22 am:
Are you sure that you are not overloading your system (either exceeding the SPME fiber adsoprtion capacity or operating outside of the linear range of your chromatography system?) I don't believe that an increased extraction time should effect your chromatography unless you are loading too much material into your system. Just a thought. Good Luck.
By Anonymous on Friday, April 12, 2002 - 06:17 am:
I am extremely surprised by the quadratic response if I read your conditions correctly. If you are using only 2 calibration points you should have a straight line. Are you forcing the curves through 0?
By Anonymous on Monday, April 15, 2002 - 06:37 am:
No, I am not forcing the curves through the origin.
HOWEVER, I am including the origin as a "fictious" third data point.
By Anonymous on Monday, April 15, 2002 - 08:00 am:
Apparently the calibration curves would not pass through the origin, and adding the origin as a zero (which is actually the same as forcing through the origin) makes the curves quadratic. If you want to include a zero concentration standard run a blank and set the concentration to 0, then run a fiber treated in the same way as the other standards. Better yet, run a three point curve using three concentrations, say 250, 500, and 1000, then use that as your calibration. There is an in-depth discussion on the LC board on calibration and linearity.
By Anonymous on Monday, May 6, 2002 - 11:30 am:
To bring some closure to this discussion: I decided to add a third point (ca. 250ppm) to my calibration data, and ignore the origin. Further, instead of using specific ions (from mass spectra), I calibrated this method using RIC (this gave me better reproducibility, and lower RSD). As a result, my curves "became" linear, with R-square values of 0.992 and above, and a small positive intercept.
By Gellért Karvaly on Tuesday, May 25, 2004 - 07:29 am:
I am using a PFPD detector in the quadratic mode to analyze a thioether and have found no data as to calculating sensitivity, LOD, LOC etc from the equation of the calibration curve. I also have found the curve to be difficult to reproduce. Has anyone encountered the same problem? I am truly grateful for any help.
Posting is currently disabled in this topic. Contact your discussion moderator for more information.