[Pw_forum] can't minimize electrons by cp.x in some case

Nicola Marzari marzari at MIT.EDU
Wed Jul 18 16:16:25 CEST 2007


Liping YU wrote:
> Dear CP users,
> By using cp.x (Version 3.2.2), I successfully run the electronic
> minimization for alpha-Al2O3 in the orthorhombic supercell with 0001
> along z. However, for the same system, but with all atoms' original
> x-coordinates changed into z-coordinates [ correspondingly, y->x, z->y,
> and celldm(1)->celldm(3), celldm(2)->cellcm(1) and celldm(3)->celldm(2)
> ], I obtained exploded results in the same calculation of electronic
> minimization. The forces in the later calculation are unreasonable huge.
> I played with restart_mode, dt, emass, emass_cutoff, electro_dynamics
> and orthogonalization, but obtained the similar results for the later.
> Does anybody have an idea on it? Thanks a lot!


Dear Liping,


this is interesting. It looks like your cell was rotated correctly
(esr and eself are the same).

Note that you are using conjugate gradients, so emass and dt and 
electron_dynamics should not matter (those are for Car-Parrinello).

Starting from scratch, and with Gram-Schmidt orthogonalization, the only
thing that could go wrong is that your initial step in the first 
steepest descent direction is too large (or too small), so that
the quadratic minimization is not working (cg takes an electronic
step in the direction of the steepest descent/conj grad for the
electrons, and with the initial value of the energy, initial
value of the derivate of the energy with respect to the search
direction, and value of the energy for a finite step, it guesses
by parabolic interpolation where it should go (cg_sub.f90).

Can you send me the first ~20 lines of convergence.dat, for the run
you did, starting from scratch, with the Gram-Schnmidt ortho ?

This is just a reminder for all of us that the CG/ensemble-DFT output 
should be made more clear.

				nicola


-- 
---------------------------------------------------------------------
Prof Nicola Marzari   Department of Materials Science and Engineering
13-5066   MIT   77 Massachusetts Avenue   Cambridge MA 02139-4307 USA
tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu



More information about the users mailing list