Andrea Pedrielli

In search of the ground-state crystal structure of Ta2O5 from ab initio and Monte Carlo simulations Andrea Pedrielli, Nicola M. Pugno, Maurizio Dapor, and Simone Taioli Computational Materials Science, Volume 216, 2023, 111828 https://doi.org/10.1016/j.commatsci.2022.11182

Hi Kay, thanks you for the answer. Do you mean the two atoms in the large box or the fcc TaO (my last post)? I would like to know if you managed to converge the calculation with standard species and you relaxed the atomic position only at pbe level or also at pbe0 one. Could you post your input as reference? I'll try to relax also the larger cells, but they were not much strechedin theory. At most by the difference between gga and pbe0 positions. Regards, Andrea

Hi Kay, I obtained converged results for a TaO fcc cell (at least a real material). The crucial parameters are the MT radii, set to very small values, and rgkmax, that has to be at most 3.0 to not produce a divergence of the energy. Also I moved some electrons of the Ta on the valence. Currently in the code is used msmooth to stabilize the calculation, what is the difference with the old trimgv keyword? Why did you delete it? Regards, Andrea Input file tasks 0 1 5 120 320 vecql 0.00 0.00 0.00 xctype...

Hi Kay, coud I ask if you have any ideas on possible solutions? I forgot to say that the lowest distance for which I got the convergence (Ta z coordinate 0.40), was otained with this custom specie without local orbitals and two more core orbitals (O specie standard but also without local orbitals). This seems strange to me as I expected (and you confirmed) that the ideal species for small MT radii should have more orbitals treated as valence that the default ones. Regards, Andrea 'Ta' : spsymb 'tantalum'...

Hi Kay, the pbe calculation is smooth under most of the parameters setting and also for high nempty. The problem emerges with the pbe0 for example with this settings (however is somewhat independent on the settings (lowq, ngridq, ngridk etc ), aside the MT radii): tasks 0 1 5 xctype 20 hybrid .true. hybridc 0.25 !lowq !.true. spinorb .true. spinpol .true. !nempty ! 20.0 swidth 0.01 avec 6.906362882 0.000000000 0.000000000 0.000000000 8.693394351 0.000000000 0.000000000 0.000000000 10.93371025 atoms...

Hi Kay, it still concerns the calculation for the Ta205 with pbe/pbe0 where the MT radius decreases from 2.8000 to 2.1726 for the tantalum and from 1.8000 to 1.3967 for the oxygen. Sorry to bother you again, however I find very difficult to solve this problem. Since the difficult in converging the large cells with pbe0, I tried to converge at least 2 atoms in the vacuum, but for a distance of less than 2.3 angstrom (z of Ta =0.4 in the coordinates below) it doesn't converge and breaks with a linearization...

I'll give up. In the end I think that this simulation is not feasible in elk, due to the contrast between the low number of empty bands needed for the pbe0 calculation, and the high number of empty bands needed for the energy loss function calculation for which the pbe0 calculation oscillates badly. Thanks anyway for the support, best, Andrea

Dear all, could you guide me in building up a specie file for a case in which the MT radius is very small, like for example the half of the default one? I understand that one should promote to the valence some of orbitals, but how many? Best, Andrea