Why infrared absorption is a nonlinear technique? I am looking for a good explanation explaining why infrared absorption technique is essentially nonlinear (eg. for carbon monoxide quantification).
When using UV/visible/near-IR absorption technique, Beer-Lambert Law often is valid and then you have quite straight way to convert transmittance into absorbance which linearly respond to concentration (provided BLL limitations are not met).
Above technique relies on electronic transitions, when IR absorption technique relies on vibrational and rotational transitions for molecules which has permanent or transient dipole moments. I would like to figure out why such technique is essentially nonlinear.
 A: Edit: putting the summary of discussion in comments into my answer.
Beer-Lambert law assumes that every photon has equal probability to be absorbed by every molecule. It is only valid for sufficiently monochromatic light 
– that is, the bandwidth of light source should be smaller than the width of the absorption line. When absorption is measured with a laser tuned to a particular rovibrational line of a molecule, the Beer-Lambert law works well.
The method discussed here uses a broadband light source that covers a spectral range containing many rovibrational lines of $CO$, but also regions between those lines, where light is not absorbed at all. There is no simple model that could describe such behavior. In theory, one can calculate the absorption spectrum of a molecule, including temperature- and pressure-dependent line broadening, and then calculate it's convolution with the spectrum of the light source. In practice, using gas samples of known composition one can build a calibration curve Absorption vs gas concentration and use it to convert absorption to gas concentration.
.
My old answer described another case where Beer-Lambert law may be not valid:
The only nonlinear effect that comes to my mind is saturation. Beer-Lambert Law is valid if the population of the ground state is not depleted. However, vibrational bands in mid-IR have very strong absorption, and with intense light source (like laser) one could deplete the ground state - see answer to this question.
