The ABINIT Users Mailing List ( CLOSED )

[Dan Ricinschi <ricinski@semi.ee.es.osaka-u.ac.jp>]

Hi!

You may try what you described, but I think you should first re-examine
the calculation parameters. I have done R3c optimizations for similar
crystals for some time and normally there are no problems with this, i.e.
the lengths, angles as well as atomic coordinates keep their rhombohedral
symmetry until the last step. I do recall to have observed this anomaly
for P4mm structures and perhaps even for R3c couple of times, but I have
managed to finally avoid it by making a better initial guess and/or
changing the calculation parameters. I agree that perhaps the
optimization code should keep the space group symmetry unchanged "in
a perfect world", but I think that in difficult cases it just cannot
proceed further to decrease the total energy without altering the
symmetry (by the way I normally use ionmov 3).

To quickly exemplify that it can work, I have adapted my old R3c
calculations for your BiAlO3 crystal, using pseudopotentials available on
the website. I started from:

   natom 10
   typat 1 2 3 3 1 2 3 3 3 3
   znucl 13 83 8

  acell 1.0647333039E+01  1.0647333039E+01 
1.0647333039E+01
  angles 59.348366 59.348366 59.348366

       xred  
2.2077000000E-01  2.2077000000E-01  2.2077000000E-01
             
0.0000000000E+00  0.0000000000E+00  0.0000000000E+00
             
3.9480000000E-01  5.2790000000E-01  9.3330000000E-01
             
5.2790000000E-01  9.3330000000E-01  3.9480000000E-01
             
7.2077000000E-01  7.2077000000E-01  7.2077000000E-01
             
5.0000000000E-01  5.0000000000E-01  5.0000000000E-01
             
4.3330000000E-01  2.7900000000E-02  8.9480000000E-01
             
9.3330000000E-01  3.9480000000E-01  5.2790000000E-01
             
2.7900000000E-02  8.9480000000E-01  4.3330000000E-01
             
8.9480000000E-01  4.3330000000E-01  2.7900000000E-02

and after 15 optimization steps, I got:

  acell 
1.0066476476E+01  1.0066476476E+01  1.0066476476E+01
  angles 59.73236210  59.73236210  59.73236210

  xred
  2.1763655340E-01   2.1763655340E-01  
2.1763655340E-01
 -1.3269553677E-02  -1.3269553677E-02 
-1.3269553677E-02
  4.1881033138E-01   5.0551522013E-01  
9.4807744877E-01
  5.0551522013E-01   9.4807744877E-01  
4.1881033138E-01
  7.1763655340E-01   7.1763655340E-01  
7.1763655340E-01
  4.8673044632E-01   4.8673044632E-01  
4.8673044632E-01
  4.4807744877E-01   5.5152201272E-03  
9.1881033138E-01
  9.4807744877E-01   4.1881033138E-01  
5.0551522013E-01
  5.5152201272E-03   9.1881033138E-01  
4.4807744877E-01
  9.1881033138E-01   4.4807744877E-01  
5.5152201272E-03

Although the numbers themselves are probably meaningless due to
overly-simplified conditions and improper psp-s, you can see that the
symmetry is preserved in this example. You may try the above initial
configuration or other similar ones to see if it works. However I think
that other calculation parameters should be adjusteded as well, starting
with Ecut and ngkpt as you already started to do. Also, your PAWs may not
be good enough (transferability problems?) for BiAlO3. Do not forget to
implement the calculation parameters according to what is generally known
for this crystal.

I am also suggesting to try to optimize as a metal first. Perhaps you do
not expect your crystal to be a metal, but you know that sometimes there
are such problems for these types of compounds.

Good luck!

DR