Can lammps perform QM/MM md runs using LATTE, like it does with fix qmmm and quantum espresso? I tried giving fix latte for a single molecule but fix nve for others, and it looks fine on small scale i tried.
So is this approach of selectively applying fix latte correct?
also in all the examples of LATTE pair_style zero is used, but i cannot find a justification for it.
your best chance to get a competent answer on these questions is to contact the LATTE developers.
my own $0.02 on this are: to do an ONIOM type QM/MM calculation, you would need to run 3 concurrent calculations (e.g. as a multi-partition run): 1) the whole system with classical MD, 2) the QM system with DFTB/LATTE, 3) the QM system with classical MD. and then you would need to subtract the forces of calculation 3) from the matching atoms in calculation 1) and add those from calculation 2). for full QM/MM with electrostatic embedding and boundary interactions (so you can cut you QM system along bonds), you would need additional modeling of the interactions of the QM and the MM interface region.