[Wien] No convergence during Volume Optimization

[pieper]

The 'difference between energy and force approaches' is just that: a 
different approach to find the equilibrium positions in space for the 
atoms and their electrons within the structural model given by the 
.struct file.

The equilibrium is the energy minimum in the parameter space under 
consideration. An energy minimum means that the derivatives with respect 
to (atom) positions - that is the forces - vanish. Knowing the 
derivatives (forces) helps, of course, a lot in finding the minimum - 
they point the way to the next (local) minimum.

In the user guide you see:

------User guide ------
5.3.1
  Lattice parameters (Volume, c/a, lattice parameters)

Package optimize

The auxilliary program optimize (x optimize) generates from an existing 
case.struct (or
case initial.struct, which is generated at the first call of optimize) a 
series of struct files
with various volumes (or c/a ratios, or other modified parameters) 
(depending on your input):

.
.
.

After execution of this script you should have a series of scf-files 
with energies corresponding to the
modified parameters, which should allow you to find the corresponding 
equillibrium parameters.
For the volume optimization an analysis tool is available, other tools 
are under development).

----------------------

The minimum total energy defines the equilibrium. The derivatives with 
respect to lattice parameters are not easy to obtain during the scf 
cycle so for lattice parameters an 'energy approach' is used.

For the internal parameters this is different. Persons as ingenious as 
Prof. Marks can calculate the derivatives with respect to internal 
parameters from the charge distribution at affordable computational cost 
during the scf.


------User guide ------

5.3.2
  Minimization of internal parameters (min lapw)

Most of the more complicated structures have free internal structural 
parameters, which can either
be taken from experiment or optimized using the calculated forces on the 
nuclei.
Starting with WIEN2k 11.1 there are two possibilities to determine the 
equilibrium position of all
individual atoms automatically (obeying the symmetry constraints of a 
certain space group). One
can use either

> the shell script min lapw, together with the program mini, which will 
> run a scf-cycle, update
the positions using the calculated forces and restarts a new scf cycle. 
This continues until
forces drop below a certain value;

> or use the normal scf-scripts run lapw -min where in case.inm the 
> switch MSR1 will be
modified to MSR1a such that the charge density and the positions are 
simultaneously opti-
mized during the scf-cycle.

------------------------------

The first option uses what you call the 'force approach': in 
equilibrium, no forces should push the atoms around.

The second option indicates a mixed approach: The positions of the atoms 
(according to forces acting on them) and the ones of the electrons (to 
minimize the total energy) are BOTH adjusted in each step of the scf.

As I said, there can be many reasons why your calculation did not reach 
convergence for some structural parameters. Did the scf stop without 
errors because the maximum number of iterations was reached? If yes, 
what was this number of iterations? Maybe your convergence criteria are 
too strong for the numerical precision you set by parameters like RKMAX 
or the k-mesh? Maybe the scf oscillates between several good solutions 
(is it a magnetic case?). Maybe your starting configuration and/or the 
model Hamiltonian is completely off or missing some ingredient and the 
poor scf wanders helpless around, lost in a multidimensional world ...

Best regards,

Martin Pieper


---
Dr. Martin Pieper
Karl-Franzens University
Institute of Physics
Universitätsplatz 5
A-8010 Graz
Austria
Tel.: +43-(0)316-380-8564


Am 13.11.2016 19:14, schrieb Abderrahmane Reggad:
> Thank you Dr Pieper for your interesting to my questions.
> 
> I have optimized the atomic positions before doing calculation.
> 
> Tha thing that I didn't understand is that the convergence is reached
> for some points but not for others.
> 
> For the "optimization notes " , there is no mention on the difference
> betwwen the energy and force approaches.
> 
> Best regards
> 
>  
> 
> --
> 
> Mr:
> A.Reggad                                          
> 
> Laboratoire de Génie Physique
> Université Ibn Khaldoun - Tiaret
> 
> Algerie
> 
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:
> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html