without additional knowledge of the electronic structure of the system I think a reasonable criterion is that the density of points in all directions is more or less the same. so I would say that in an hcp with c/a ~ 1.6 a more balanced grid could have been 18 18 12 (~18/1.6) 0 0 0. but how many points are needed depends on the dispersion of the band structure in the different directions. If on a given direction the band structure varies strongly and in another it is almost flat It is likely you'll need many more k-points in the former direction than in the latter.
The golden rule is as usual to check convergence in your particular case and if the behavior of your convergence test does not agree with your expectation think which of the assumptions you based your expectation on does not apply. hope this helps, regards, stefano On 05/03/2012 01:28 PM, Divya Srivastava wrote: > Dear QE Developers/Users > > I have used K_POINTS (automatic) 18 18 18 0 0 0 for HCP (hexagonal close > packed) lattice in scf calculation. My lattice parameters are a= 4.0 a. u. > and c/a = 1.59. I would like to know whether I can use nk1=nk2=nk3 for the > HCP lattice or not. If not what is the criteria for choosing them in case of > non cubic (HCP) lattices. > > Sincerely, > Sincerely, > > Dr. Divya Srivastava > Research Associate > TSU, JNCASR > Jakkur, Bangalore-64 > > Tel: +9122082834 > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum
