The adsorption energy between 45 atoms molecule and 10x10 unit cell of graphene needed to be calculated. The potential energy surface is very shallow and i can't reach a minimum structure. Any suggestions how to solve this problem?
OK, so your actual problem differs from the major question you asked because getting useful adsorption energies and being non-cpu expensive are two very different goals. Especially if you have a shallow potential energy surface it's going to be even harder. Some ideas:
0) Preoptimize the graphene layer without the adsorbate with whatever combination you use. I presume you are already doing that but just to be sure I'll put it here.
1) If the PES is shallow, it is a good idea not to play too much with functionals, basis sets and dispersion corrections in the beginning but rather to think about SCF and geometry convergence limits because that might be the reason why you didn't reach a minimum in the first place.
2) Regarding the actual functional choice: that's a heavily debated point in the literature and there is a lot of movement. In a 2015 comparative study (DOI: 10.1088/0953-8984/27/41/415502) the combination PBE+TS+SDS+MBE came closest to the non-DFT "gold standard" values. I myself also measured adsorption energies of flat PAHs to HOPG (DOI: 10.1063/1.5052728 + image correction in 10.1063/1.5109235 ) for which the vdW-DF method happened to match the measurements best, but note that in other benchmark studies vdW-DF was quite far off so that might have been more luck than systematics. In any case, check the literature for more recent benchmarks, I haven't done a thorough search since the publication of my aforementioned paper anymore.
What did you try about the SCF? Does your SCF calculation itself converge and, if so, have you checked whether the gradients in the geometry optimization have a significant magnitude?
HF calculation.....Only forces are converged but displacement never get converged. Can you send me ur email address so that I can send you the log file. Hopefully, you can find a clue for this system
Probably you are not doing your calculations in Turbomole, right? I only have deeper experience in that software, with logs from other QC packages I probably can't do much.
Most good QC suites have their own form where such a question will probably be placed better, if you're using VASP it's: