When evaluating the band structure or doing any calculation in ATK, why is it necessary to do a geometry optimization? The unit cell dimensions and distance between atoms is already determined from X-ray diffraction experiments and we have a list of all those data. Then why is geometry optimization required?
If you want to see what the band structure of your crystal looks like for measured lattice parameters, you are free to do it. There is nothing wrong about it.
Atomistic simulations that are based the approximate DFT approach (which deals with the density functional approximations such as LDA, GGA and others) allow you to predict equilibrium lattice parameters that can then be compared to experiment. But if you adopt experimental values, your structure will be somewhat strained due to approximate nature of the DFT functionals.The agreement with experiment is usually within 1 %, and can better or worse depending on the material and DFT scheme used.
In many cases, it is important to have your system in equilibrium state, e.g., when calculating phonon spectrum or studying effect of strain on the material properties and so on. In this case, you have to optimize relax the structure for the sake of consistency in the model, even so the lattice parameters are not exactly as in experiment. I would like to notice that measurements of physical quantities also come with an error bar, and using different measurement techniques may give somewhat different results.
Regarding your second question, you may consult our documentation, as well as a vast literature on the topic of atomistic modeling.
