[Wien] Relaxation (mini) problem

[Peter Blaha]

> I have a problem with relaxation of forces. Let,s use InGaN as an example
> - this is a simplest unit cell for InGaN with 50% of Ga replaced by In
> (x=0.5), 'optimized' atomic positions from struct file are at the end of
> this e-mail. This is inM file for mini:
>
> NEWT 2.0           #(NOSE, NEWT, BFGS, MOLD, tolf (a4,f5.2))
> 0.0 0.0 0.0 0.2    # Atom1   (NOSE, MOLD:Masse, delta t, T, nose-frequency)
> 0.2 0.2 0.2 0.2    # Atom2   (NEWT: 1,2,3:delta, 4:eta(1=MOLD))
> 0.2 0.2 0.2 0.2    # Atom3   (BFGS, NEWT: [ 1 || 2 || 3 ] = 0 constraint)
> 0.2 0.2 0.2 0.2
>
> These are forces at the end of scf_mini file:
>
>        TOTAL FORCE IN mRy/a.u. = |F|     Fx             Fy             Fz
> :FOR001:   1.ATOM         14.750          0.000          0.000         14.750
> :FOR002:   2.ATOM          0.847          0.000          0.000         -0.847
> :FOR003:   3.ATOM          1.809          0.000          0.000         -1.809
> :FOR004:   4.ATOM          1.301          0.000          0.000         -1.301
>
> I didn't move atom #1 because I guess its only necessary to move 3 atoms
> when there are 4 in the supercell but then the force on the atom #1 is not
> optimized (its 14.750). Why it is the case?

I don't think there is any reason not to optimize the position of the 1st
atom!
Maybe the sysmmetry is such that you could fix one atom, however, then it
may take very very long until ALL other atoms have slowly adjusted. This
would probably mean that you have to go on until the forces on the other
atoms are very small too (much smaller than you have).
It is simply a "mass problem". Suppose you have an adatom at an surface:
Of course you could fix the position of this add-atom and then "move" the
whole surface slab until it is at equillibrium, but probably you will wait
foerever until this happens. (While it is simple to fix the slab and
relax the ad-atom).


> 0.97398187). However, I noticed that similar things happen when relaxaing
> the surface with typical 3 layers relaxation - I don't like this because
> it means its a lot of strain when 3 layers are relaxed and one should
> maybe relax many more layers - this puts in question surface BS calculated
> later with such relaxation.
?????
Did you use the optimized bulk lattice parameter or the experimental one ?
With the latter distances you get of course forces for all layers, because
all layers want to come closer/further apart since they are not constraint
as in a 3D calculation.

> PS. To ML-administrator: this time I use again my original e-mail
> plucinks at mail.desy.de to send this e-mail to the mailing list. When I use
> my new e-mail pluto at physics.bu.edu I always have to wait for the
> 'moderator' decision - maybe its related to the fact that 'physics.bu.edu'
> is alias of 'buphy.bu.edu'?

You are subscribed as:   pluto at physics.bu.edu
and only if the "from" line of your email has this adress, you can post.
plucinsks... certainly cannot work since there is no subscription.

You must configure your mailing system such that the "from" line is the
same as the address you want to have.

                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------