[gpaw-users] Anatase TiO2 lattice parameter determination with PBE

[Rasmus Karlsson]

Hello again Marcin,
sorry for taking so long to get back to you.

On 02/23/2013 06:36 PM, Marcin Dulak wrote:
> when finding lattice constants one should fix gpts instead of h, as 
> the 
> https://wiki.fysik.dtu.dk/gpaw/tutorials/lattice_constants/lattice_constants.html 
> tutorial shows,
> you can verify your outputs if the number of gpts is constant - if not 
> - does fixing gpts change the result?
I didn't have much luck setting gpoints. However, the number of grid 
points seemed more or less constant from what I could see in my output.
>> I am using GPAW version 0.9.0.8965 (with ASE 3.6.0.2515) and numpy 
>> version 1.3.0. My gpaw-setups version is 0.8.7929.
> Note that changing c parameter results in very small energy changes.
> In order to find a minimum one needs to use quite tight scf/optimizer 
> thresholds.
> I'm getting a=3.81 and c=9.69 (see attached script). I tried also an 
> fhi-aims with light basis set, getting a=3.80, c=9.74,
> so it looks like the correct PBE result is closer to the reference 45 
> from the article you mentioned.
> You need GPAW trunk to run that script.
> Could you verify the consistency of the stress optimization and ps 
> scan methods in PW mode on this example?
>
I was able to use the stress tensor to get values similar to yours.

I had a lot of trouble getting the potential surface scanning method to 
work. In the end, what fixed it was making sure I relaxed the contents 
of each unit cell in each single point calculation. This relaxation 
could change the energy of the system by about 0.2eV, which turned out 
to be important. After doing this, I was able to see that a single point 
PW(600) calculation at the minimum energy geometry found by the stress 
tensor gave an energy that was very similar (within 20 meV).

I was also able to use a PAW basis and a similar grid of single point 
energies, and was able to get optimized anatase lattice constants that 
differed from the result of the PW(600) stress tensor by about 0.02-0.04 
Å. The stress optimization and potential scan methods seem consistent.

Thank you,
Rasmus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20130307/7797418b/attachment.html