There are two possibilities for getting carbon dioxide into your instrument. The first is a leak in the GC/MS system, which would give a continuous signal for CO2. The other would be related to your sampling system, which would introduce CO2 with the sample and would give a chromatographic peak for CO2. (And, with a cold trap a leak in the GC system ahead of the cold trap can give a signal for CO2 when the trap is warm and reduced signal for CO2 when the trap is cold, if the trap is cold enough.)
I assume this is a continuous CO2 signal?
Some things to look at in that case:
The first thing to do is to be sure the instrument is leak tight. How does your air/water profile look on the instrument? Depending on the vendor you may be able to ratio the m/z 28 peak from air to a reference, such as m/z 69 in PFTBA to evaluate this.
If the instrument shows CO2 as a continuous background, it is a reasonable assumption that the CO2 is coming from the air. You should see a corresponding signal of m/z 18 from water if you will scan that low. With the instrument showing a profile on the display and the peak for m/z 18 showing, turn a can of "dust-off" or a similar product upside down and spray a short burst across the column a few cm from the detector end of the column so that a bit of frost forms on the column. Do this away from the ferrules - we do not want to cause thermal expansion/contraction. If the intensity of the signal drops, the water was coming through the column. If there was no effect on the m/z 18 signal, the water is not passing by that cold spot. Look for a leak downstream - the mass spectrometer. (Note that the frost will disappear and trapped water will be released and you will see a brief spike in signal.)
I have heard of a thermal desorbtion system that gave the user continuous trouble with leaks. I know they removed the thermal desorbtion unit, but that decision may have been made for unrelated reasons.
