I wonder if anybody can indicate the reasons of non-linear but exponential calibration in splitless analysis. References to the corresponding literature will be of great help.

Are you using an ECD?

No, I'm using ion trap MS detector.

Probably not much useful literature around, but I've been running a Saturn for 10 years doing pesticide work. It is NOT inherent to the detector design.

Reason 1...
It's not the detector, it's the GC that's attached. Columns and inlets that absorb compounds usually cause this problem. If you continue to run the system over a long period of time, non-stop, it often improves. We are currently dealing with a Fluridone analysis problem on an MSD that shows this behavior. The internal standards and surrogates have linear curves, the Fluridone is quadratic. Once upon a time, it, too, was linear. Then we took time off for Thanksgiving.

The problem is often concentration related. Unfortunately, it isn't always possible to run high levels on a trap because it will overload, or alter the spectrum. It's not as forgiving as an MSD in this respect. But it's still my favorite instrument in the lab.

Reason 2...
Compound dependent--that is just the way the ions form. I usually have perfectly linear curves and quadratic curves from the same splitless injection, for different compounds. It's not the injector that's at fault, either (assuming it is set up properly).

Sometimes lowering temps can help, and there are all sorts of other things to try. A big part of the problem is the capillary column. Analytes that work fine on a megabore system won't behave on a 0.25 capillary. I think part of the reason is that megabores have much thicker film thicknesses--we use 2.65 um a lot--versus the 0.25 um of a typical MS grade capillary. There are probably a lot more active sites on the cap column.

The solution....
Use LC wherever possible. In general, it does not suffer from this problem, and I am drawing on not only my experience, but that of my colleagues that represent hundreds of thousands of analytical runs using LC/UV, FLD, post-column FLD, Antek N detector, and MS detectors running pesticides. We are running more and more compounds on LC/MS. We can't run everything yet, but we can do a lot. Another generation or two of interfaces, and we will be able to run almost everything we need to on those systems.

In the meantime, you must learn to live with the quadratic curves. I feel it is more important that they be reproducible, no matter what the shape.

Scott, I also think that it is GC part that is responsible for curvature in calibration graphs and not Detector. Mainly it is claimed that injector and injection mode are at fault. And of cause it is sample dependent and that's why there is a variety of injectors and liners. And of cause it is concentration related, but not at high as at low concentrations.
It will be interesting and very helpful if GC folks could describe cases with non-linearity and perhaps the solutions found in these cases. This is my appeal to all forum members