anatase and rutile

Message

giacomo giorgi · #1 Post by **giacomo giorgi** » Wed Feb 03, 2010 7:23 am

Dear all,
I am trying to minimize the structures of both anatase and rutile. I obtain both at PAW/PBE and PAW/LDA that anatase is more stable of rutile (0.08 eV at PBE, 0.002 eV at LDA). Usually it is the opposite

Can anybody help me in understanding the reason?

I used The standard pseudo for Ti (4elec in the val) and the soft one for O (O_s), i.e,
(I got the same result using the standard pseudo for O)

PAW Ti 03Oct2001
4.00000000000000000
parameters from PSCTR are:
VRHFIN =Ti: d3 s1

and

PAW O_s 06May1998
6.00000000000000000
parameters from PSCTR are:
VRHFIN =O: s2p4

KPOINTS is 8 x 8 x 4 MP for anatase and 8x8x16 MP for rutile
(no difference found centering the sampling on Gamma)

I attach my INCAR
SYSTEM=Cella TiO2
ISTART=1
ISMEAR=0
PREC=High
ISIF=2
NSW=300
IBRION=2
#ADDGRID= .True.
EDIFF=1E-6
EDIFFG=-0.001
GGA=PE

Thanks in advance.

Best,
G

alex · #2 Post by **alex** » Wed Feb 03, 2010 7:30 am

Hi Giacomo,

did you optimise the cell parameters (ISIF = 3) with at least 500eV cutoff energy?

I can't see that in your INCAR!

Cheers,

alex

giacomo giorgi · #3 Post by **giacomo giorgi** » Wed Feb 03, 2010 8:28 am

Dear Alex
Thanks.
with ENCUT=520 ENAUG=600 ...the result is even worse.... (ISIF=3)... at least at PBE

anatase= -26.58987425

rutile= -26.5095320

Any further idea?
Thanks in advance!

G

Thanks

giacomo giorgi · #4 Post by **giacomo giorgi** » Wed Feb 03, 2010 9:06 am

Moreover I am reading in literature (Theor Chem Acc (2007) 117, 565

For the LDA calc the optimsed parameters were obtained with a 5x5x5 kpoint mesh and a 400eV cutoff....

For GGA same sampling with (!) 350 eV of cutoff.

Here they obtain rutile more stable than anatase by -75 meV at GGA and -11 meV at LDA.....

!dodge

thanks!
g

alex · #5 Post by **alex** » Wed Feb 03, 2010 9:18 am

Well, in case of the citation I wonder then if they compared the right thing anyway.

I've crosschecked with 800eV and standard O. The same ...

One could go futher and include zero-point vibrational energy and see what happens. And dig literature.

The other way out would be to ask the titania grand master herself: Anabella Selloni, Princeton.

Cheers,

Alex

zoowe · #6 Post by **zoowe** » Wed Feb 03, 2010 1:06 pm

Is it reasonable to compare total energy of Rutile & anatase?

alex · #7 Post by **alex** » Wed Feb 03, 2010 1:23 pm

Why not?

aron · #8 Post by **aron** » Wed Feb 03, 2010 8:46 pm

If you're concerned with reproducing such small energy differences, I suggest using the larger Ti POTCAR (12 valence electrons), setting a high plane wave cutoff and doing a full optimization of each structure (ideally with an equation of state fit to get the equilibrium volume).

giacomo giorgi · #9 Post by **giacomo giorgi** » Thu Feb 04, 2010 7:28 am

same INCAR , K-POINTS and restarting from CONTCAR but using 12 electrons in the valence for Ti (Ti_sv) and the standard O pseudo resulted in

-26.8905 eV for rutile

&

-26.9828 eV for anatase

ENCUT=520 eV

:?

best,
G

zoowe · #10 Post by **zoowe** » Thu Feb 04, 2010 2:54 pm

Why don't you convert the total energies to other quantities which can be compared between the systems in different volume?

If you want to compare calculated total energies, make sure that reference for energy, Cell, k-mesh (k density), number of planwaves.... are the same.

Tell me if I am wrong.