How to do geometry optimization with defect?

Total energy, geometry optimization, DFT+U, spin....

Moderator: bguster

Locked
raul_l
Posts: 74
Joined: Sun Jan 08, 2012 7:45 pm

How to do geometry optimization with defect?

Post by raul_l » Mon Nov 16, 2015 1:37 pm

I'm studying a material with 52 atoms in the unit cell. For the pure material I can do structural optimization and calculate other things just fine. However, when I substitute one of the atoms with an impurity atom, structural optimization is unable to converge. That is, the maximal force value (maxfor) goes like this:

Code: Select all

2.539E-02
2.003E-02
1.416E-02
1.231E-02
1.480E-02
1.319E-02
1.930E-02
2.289E-02
2.301E-02
...

Is it possible that it has something to do with changes in symmetry? I can see the following in my log file:

Code: Select all

--- !COMMENT
message: |
    The Bravais lattice determined only from the primitive
    vectors, bravais(1)=  4, is more symmetric
    than the real one, iholohedry=  1, obtained by taking into
    account the atomic positions. Start deforming the primitive vector set.
src_file: symbrav.F90
src_line: 201
...

 symlatt : the Bravais lattice is mC (one-face-centered monoclinic)

--- !COMMENT
message: |
    The Bravais lattice determined from modified primitive
    vectors, bravais(1)=  2, has a lower symmetry than before,
    but is still more symmetric than the real one, iholohedry=  1
    obtained by taking into account the atomic positions.
src_file: symbrav.F90
src_line: 210

The pure material material has the tetragonal lattice. With a defect, I think it becomes monoclinic or triclinic. Should I modify the lattice constants and/or the primitive vectors slightly to help break the symmetry?
Raul Laasner
Netherlands Institute for Space Research

raul_l
Posts: 74
Joined: Sun Jan 08, 2012 7:45 pm

Re: How to do geometry optimization with defect?

Post by raul_l » Mon Nov 23, 2015 11:43 am

It seems to have been just a really difficult structure to relax. With ionmov 7 it converged after about 130 cycles. Using that as input, it also converges with ionmov 2.
Raul Laasner
Netherlands Institute for Space Research

Locked