[Wien] Ionic versus covalent systems

[Peter Blaha]

Make sure the Na sphere is not too big ! Eventually, use the same sphere
as for Nb. Otherwise use "defaults".
Such simple system should converge very easily.

jadhikari at clarku.edu schrieb:
> Dear Wien2k users,
> 
> Thank you very much for the suggestion for my question on ionic versus
> covalent systems.
> 
> I have tried with mixing parameter of 0.1 most of the time and changed it
> to 0.05 or 0.2 or 0.25 or 0.4 when things do not work. As for example,
> NaNbO3-
> 
> 1. Most of the time I get ghost bands which is corrected by changing the
> energies in the case.in1 file and changing the global energy parameter
> based on the :FER energy. I have tried adding local orbitals and changed
> its energy until the ghost bands are gone. Energies are used based on the
> DOS plot rather than putting something inaccurate or wild.
> 
>  The term ghost band I mean is when QTL-B value is higher than 5.00. There
> are other ionic systems in which I could manage to go below 5.00 and the
> system showed no error. The minimum that I got was 2.10 and never it went
> below this value. I tried more than 100 times to fix this. So, moving up
> in k-points with this QTL-B value is not appropriate as the calculation
> surely shows the ghost band with higher QTL-B value.
> 
> 2. RMT for the three atoms are optimized. Values are chosen so that there
> is very very small leackage of charges. It has two extreme values- the
> minimum RMT is the one in which the calculation is stuck at lapw2 and the
> maximum is the one in which there is a nearest neighbor error. Value in
> between these two with negligible leakage and close to that of other atoms
> is the optimized RMT. (UG's specification is also followed)
> 
> 3. RKmax is altered from 4.00 to 7.00. I got the system to converg with
> 5.50 and not with other ones. There is no basis for its optimization as I
> am aware of (considering the no of plane waves required to accurately
> describe the system). Maybe the output changes with 6.00 or 6.50 or 7.00.
> So convergence with respect to RKmax is still under investigation. EFG
> seems to be extremely sensitive to even a slight change.
> 
> 4. No of points in the irreducible brillouin zone is started with the
> minimum. Then, it is increased to the next higher one possible based on
> the symmetry of the system generated by "x kgen". Now the problem comes.
> Either the RMT has to be changed or ghost band has to be fixed.......
> 
> 5. The TEMP smearing of 0.002 Ry was found to produce no errors in most of
> the cases. I have tested this with other methods and different values.
> 
> 6. PBE-XC functional is the one chosen to be more accurate than LDA-XC for
> the present case. But I have tried both.
> 
> 7. s and p orbitals receive 0 and d receives 1 for LAPW and APW+lo
> respectively. Other combinations did not seem to be right. The energy
> window is -10.0 to 2.0 and the values upto -12.0 for minimum and upto 3.0
> for maximum have also been tried.
> 
> Hence, this ionic system needs something that I have not understood yet. I
> have thought of the following now,
> 
> 1. Lowering the space group symmetry or
> 2. Using supercells or
> 3. Using all the system's default parameters or
> 4. ........
> 
> Any suggestion would be highly appreciated.
> 
> Once again thanking you.
> 
> Subin Adhikari
> 
> 
> 
> 
> 
> 
> 
>  Dear Subin,
>> You can change the parameter in case.inm from the default 0.40 to 0.01,
> as
> %
> 
> 
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

-- 

                                       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/
--------------------------------------------------------------------------