The rule to remember is this:
“When you measure absorption spectra (X-ray, UV, Vis, IR…) in transmission mode then you must not have any feature in the spectrum with an absorbance higher than approximately 1.2.”
The principles from which this follows are nicely explained here:
For an absorbance of 1 90% of the incident light is absorbed. So 10% of the light transmits the sample and is detected. This ensures that the detector for the transmitted light has a signal with reliable intensity.
If you have an absorbance of 2 then 99% of the light is absorbed by your sample – so the transmission detector only sees 1% of the original intensity, which is usually not quite enough to ensure low noise and a linear response of the detector.
For an absorbance of 3 the detector only gets 0.1% of the original light intensity. This intensity is already approaching the noise level in many laboratory spectrometers. So what you are measuring is probably not significant – you are probably just seeing a flat noise line due to … – electronic noise.
So if the concentration of your substance is so that the absorbance is >1.2 then you must dilute your sample or shorten the thickness of the sample.
Most of the time dilution is the better way to do it:*
Just take a defined volume (in practice usually 1 mL or 10 mL) from your solution with a volumetric pipette and fill it up to 100 mL or perhaps even 1000 mL, in a clean volumetric flask. Use clean glassware for all of this – plastics may introduce contaminants.
Accuracy is important here – especially in pipetting. A 10% error in the pipette volume will mean a 10% error in the absorbance. You should aim for 1% accuracy.
*Those of you interested in molecular interactions at high concentrations can’t dilute of course – you will need to get special cuvettes providing a low sample thickness.