Sergey
qcplab@gmail.com
I try to optimize adenine-adenine stack minimum using MP2/cc-pVTZ (pls. see att.1, att. 2), and I witness unexpected behaviour. As this system is a weakly, dispersion bound di -molecular system, the overall energy is 2*(adenine)+dispersion energy, and in the beginning of optimization, the overall energy quickly drops - as mainly molecular but not intermolecular geometry is optimized.
Thus, pretty soon we have a minima which corresponds to 2 separate optimized molecules of adenine that 'do not feel' each other so far. As optimization proceeds to lower gradients (frequencies of Hessian), stacking becomes evident and 2 adenine molecules attract. MP2 overestimates stacking energy so instead of 3.4 A it goes to 3.2 A.
However, we have an issue: as 2 adenines approach each other, they perturb each other MO geometry, so the energies of single adenines rise; while stacking energy lowers; at some point they should find balance? But if you take a look at the optimization energy vs. cycle graph (attached), you find it climbs up and only change this tendency over large number of cycles when (see attached animation) a mutual orientation of adenine molecules also changes.
The question is, why the algorithm climbs up for such the extended number of steps? I supposed, it is BFGS -driven code, i.e. it finds minima of energy on PES eventhough the gradient may rise. Indeed, how the algorithm resolve such the situation, where there are 2 coupled, but of different order, contributors to the overall energy term?
Thank you!
This message contains the 46 kb attachment [ AAstack.bmp ] Adenine-Adenine Energy vs. Cycles |