[Wien] ,IRelast: incorrect structure for e=0.3 in C44 for TiO2

[Peter Blaha]

Your email must be smaller than 40 kB !!!!

A couple of comments:

Very small distortions are always problematic. When do you consider a 
certain symmetry fulfilled ? Due to the limited digits of certain 
numbers rounding problems (or factors like sqrt(2), or cos(45), ... may 
cause such rounding problems, which later on cause problems.

I'm not surprised to see energy-jumps when comparing energies from a 
tetragonal case with an orthorhombic one, where the lattice vectors have 
been rotated. Such calculations would require "fully converged" meshes 
(k-mesh, FFT mesh,..). You should do the "high-symmetry" case also in 
the low-symmetry structure. For this, you have to initialize a 
low-symmetry distorted structure and AFTER  init_lapw, change manually 
the lattice parameters such that you obtain the high-symmetry case. Do 
NOT rerun  init_lapw afterwards.

In essence: when symmetry (or better: the bravais lattice) changes, you 
should either not include the e=0 case in the analysis or run the e=0 
case as I indicated above.

PS: With a struct file, maybe the developer of IRestast could play 
around ....

Best regards
Peter Blaha
---------------------
Von:
"Park, Ken" <Kenneth_Park at baylor.edu>
Datum:
13.11.2022, 16:18
An:
"wien at zeus.theochem.tuwien.ac.at" <wien at zeus.theochem.tuwien.ac.at>

Dear Wien2k developers and experts,

I have been using wien2k 21.1 and IRelast package to calculate the 
elastic constants for rutile TiO2 at pressure =0.

When I used a small strain such as e = 0.3% was used for C44 (monoclinic 
distortion), the script generated an incorrect structure causing a 
crash. Below is a top portion of the structure with beta not equal to 
90, for which wien2k expects alpha=beta=90.



[parkk at login001 tio2r]$ head -10 tio2r.struct

TiO2 


CXZ LATTICE,NONEQUIV.ATOMS:  4 12 C2/m

MODE OF CALC=RELA unit=bohr 


  12.390517 13.568280 12.390517 90.000000 90.000562155.460807

ATOM   1: X=0.00000000 Y=0.00000000 Z=0.00000000

           MULT= 1          ISPLIT= 0

Ti1        NPT=  781  R0=0.00005000 RMT= 1.78000     Z: 22.0

LOCAL ROT MATRIX:    1.0000000 0.0000000 0.0000000

                      0.0000000 1.0000000 0.0000000

                      0.0000000 0.0000000 1.0000000



The structures with e >= 0.5 seem working all right. Could you look into 
this issue?



I also have been getting negative values of C44 and C66 in higher order 
fit (4). So, initially I thought I needed higher rkmax in calculating 
those elastic moduli, but then, I found out that the energy at e=0 
remains higher compared to those up to about e < 1, despite of using 
rkmax up to 10. (See the attached figure for C66.) With more data 
points, it becomes clear that it is due to the e=0 data point lying 
above the neighboring points by about 1.5 mRy. I believe that different 
symmetry constraints are responsible for the energy discrepancy between 
e=0 (tetragonal) and e != 0 (orthorhombic ) as the distortion for C66 
lowers the symmetry from the tetragonal to the orthorhombic. This would 
explain why I see a similar energy jump at e=0 also for C44, which 
involves the deformation from tetragonal to monoclinic whereas the other 
three distortions c11+c12, c33, and czz (all tetragonal) don’t show 
such. So, I think that in the calculation of the elastic moduli with the 
distortions involving symmetry change, the e=0 structure should not be 
included. Could you comment on this?



If it may be convenient for you, I will be more than happy to provide 
you the structure and outputeos files.

Respectfully,

Ken Park

-- 
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300
Email: peter.blaha at tuwien.ac.at    WIEN2k: http://www.wien2k.at
WWW:   http://www.imc.tuwien.ac.at
-------------------------------------------------------------------------