Simple: The CO2 content of the air is different. If the CO2 content decreases after the background measurement, the corresponding peaks are negative in the absorbance spectrum otherwise if the content increases you get the usual positive peaks.

One second thought: Why do your water bands show rotational fine structure? Most probably because it is water vapor and not because it is present inside the sample (but there certainly might be water bands below that belong to this kind of water). It might well be that you have problems with the nitrogen purging...

Did you try to run a background spectrum again before doing the measurement of your sample?

Simple: The CO2 content of the air is different. If the CO2 content decreases after the background measurement, the corresponding peaks are negative in the absorbance spectrum otherwise if the content increases you get the usual positive peaks.

One second thought: Why do your water bands show rotational fine structure? Most probably because it is water vapor and not because it is present inside the sample (but there certainly might be water bands below that belong to this kind of water). It might well be that you have problems with the nitrogen purging...

the answer which Thomas Mayerhöfer noted above is right. positive or negative peak is a function of CO2 content in the air and also manner of CO2 adsorption which take place during FTIR analysis.

Thanks Thomas and Seyed for your inputs. Yes Thomas, the rotational bands were could be from water vapor not the PEI solution. More analysis would be done to verify this.

Danila, not right before the analysis. Though, I think a background was taken before analyzing a polystyrene standard

I agree with Thomas Mayerhöfer completely. I have two different comments. The aspect of rotational structure of water. The main point is what is your resolution. The rotational lines can be observer only when the separation between the lines is much lower than your resolution. So if you have the the lines separated by 2 cm-1 and you are using resolution 4 cm-1 you can see only the contour of the rotational structure. In your spectra both CO2 and H2O are gas phase spectra. Your sample is a solution placed inside the closed cell as I assume. So if the cell is not leaking there is no possibility to induce the vapor to the well maintained sample chamber. Second aspect is the connected with the details of your experimental procedure. I have no idea what type of FTIR are you using. Nitrogen purging , vacuum system or atmospheric compensation (in these case the measurement is done under atmospheric conditions assuming the air in the room has the same CO2 and water concentration all the time). Remember you have to open the chamber to place the sample in the holder and you cannot hold your breath. So after opening the chamber the air more rich in CO2 is getting inside. Good laboratory practice says that you should record you background and spectrum under same conditions. So set your purging before taking the background. Wait for some time. Than open the chamber like you usually do. Wait some time e.g. 5 min. Record the background. Than open the chamber again and place your sample. Wait the same period of time and than collect the spectra. In 99% it helps. If no... yes it the problem with you purging system.

How possible is it that the water absorption bands 'overshadowed' the peaks (e.g. N-H) that I'm looking for?

Thomas Mayerhöfer's answer is basically correct.

More specifically, the upper spectrum was taken with the reference spectrum with good purging of the spectrometer, whereas the sample spectrum was taken with almost no purge. Thus, all the water vapor and CO2 spectra appear in positive direction in the absorbance mode. The P and Q branches of the water and CO2 appear because those are in the poorly purged atmosphere as a molecularly dispersed gas. The CO2 band around 2350 cm-1 shows smooth doublet whereas the water bands around 1600 cm-1 show many sharp bands due to the band separation of those fine structures. As Carolina Haupa correctly pointed out, the separation of the CO2 fine structure is much closer than 2 or 4 cm-1 resolution typically used for solid material characterization, forcing only the sharp band peak contour (smooth band) can be seen. If you used the resolution of 0.1 cm-1 or better, you will also be able to see many sharp bands in the 2350 cm-1 region.

For the bottom spectrum, you purged the spectrometer for obtaining reference and sample spectra relatively well. However, the degree of purge of the sample spectrum is better than the reference spectrum, causing the CO2 band around 2350 cm-1 to be inverted.

Are you looking for N-H (secondary amine) or -NH2 (primary amine)? If the former, it should show rather broad single band around 3300 cm-1 and thus should be visible if you have the secondary amine in your sample, which does not heavily overlap with the water bands above 3500 cm-1. If you meant it to be the primary amine, it would show more complex bands. According to the theoretical expectation, you would predict two bands (antisymmetric and symmetric -NH2 bands) in the region of 3300-3200 cm-1. However, the antisymmetric bands at higher frequency accidentally coincide with the overtone of the NH2 bending modes in the lower frequency, forcing the antisymmetric mode to split into two nearly equal intensity bands (called Fermi resonance). Thus, you would see three bands in the area of 3400 - 3200 cm-1 for the primary amine instead of expected two. Still, this range is below the water band region and should be observable should your sample contain -NH2 groups.