Spectra from KBR pellets are usually superior to ATR (lower noise, fewer anomalies, sharper peaks) especially for solid samples that are too hard to deform against the surface of the ATR crystal. This is only true, however, if you have the ability to make high quality KBR pellets. This requires spectral grade dry KBR, dry samples, the ability to make a fine powder (mortar and pestle usually sufficient) and a high pressure press. In my experience, it is difficult to get good pellets from the screw type presses.
ATR is great for liquids, soft deformable solids, and for anyone who lacks the equipment to make high quality pellets. ATR is also good for quant work because the path length is fixed.
Dear Alaa, while Henrik and Theodore above have provided basically fair appraisal of KBr-disk vs ATR spectra, i'd like to note the following.
In fact, there always will be a BIG difference between the spectra of the same sample recorded by the two techniques. Because of Water. Even trace amounts of it in KBr give broad and intense peak in 3500-3200 cm-1 range masking (or distorting) absorptions of OH or NH groups (if present) in the sample. KBr pellet without such interfering contribution is almost unattainable ideal.
Be aware also that not every sample tolerates grinding and pressing with KBr. The latter literally flows under the high pressures in the die, and the analyte molecules or macromolecules can change their molecular or submolecular structure under the high shearing strains developing within this flow.Thus, running the spectra of fragile molecules, you can observe VERY BIG differences among the spectra recorded in KBr and on an ATR.
Yet, possible (under the high pressure) in KBr reactions of ionic exchange can change your sample to make it unrecognizable.
KBr spectra of some polymers (hard to disintegrate to few-micron-sized particles) also may look awful as compared to those obtained by ATR.
In more depth these (and many other very useful) IR sampling aspects are discussed on www.ijvs.com pages, and I cordially advise you to visit this site. Good luck!
In case of my samples (which are viscous), I get best solutions whenever I use KBr pellets.... Likewise, the sample requirement is less than required in ATR