Problems with Localizing CeO2 f electrons with HSE06 functional

[askhetan]

Hi, I am a beginner. I want to study role of ceria as a substrate using accurate calculations of energy and DOS, (band gaps etc not important). To deal with the localization problem, I want to use HSE06 rather than the DFT+U method, but on running the calculations, I find that the even after 16 hours, the stdlog.out file never goes beyond the fourth DAV step, though the calculations are up and running. I am using gamma k-point mesh with 7X7X1 for a (111) slab model

This is my INCAR file:
SYSTEM = Ceria(111)
ENCUT=500
ISMEAR=-5
ISIF=2
IBRION = 2
MAGMOM = 18*1 # had 18 atoms in the sytem so just started with this to be safe
NSW=2000
ISPIN=2
ISPIND=2
LWAVE=.FALSE.
LREAL=AUTO
GGA=PE
HFSCREEN=0.2 # standard value for CeO2 from (kresse, ganduglia, sauer) hse06 papers
LHFCALC=.TRUE.
AEXX=0.25 # standard value for CeO2 from (kresse, ganduglia, sauer) hse06 papers

and this is my POSCAR files (bottom two layers are fixed):
CeO2(111)
1.0
3.8262000084 0.0000000000 0.0000000000
-1.9131000042 3.3135864072 0.0000000000
0.0000000000 0.0000000000 32.1822013855
O Ce
12 6
Selective Dynamics
Direct
0.000000000 0.000000000 0.097070001 F F F
0.000000000 0.000000000 0.388289988 T T T
0.333330005 0.666670024 0.194150001 T T T
0.333330005 0.666670024 0.485370010 T T T
0.666670024 0.333330005 0.000000000 F F F
0.666670024 0.333330005 0.291220009 T T T
0.333330005 0.666670024 0.048540000 F F F
0.333330005 0.666670024 0.339760005 T T T
0.000000000 0.000000000 0.242679998 T T T
0.000000000 0.000000000 0.533900023 T T T
0.666670024 0.333330005 0.145610005 F F F
0.666670024 0.333330005 0.436830014 T T T
0.000000000 0.000000000 0.024270000 F F F
0.000000000 0.000000000 0.315490007 T T T
0.666670024 0.333330005 0.218419999 T T T
0.666670024 0.333330005 0.509639978 T T T
0.333330005 0.666670024 0.121339999 F F F
0.333330005 0.666670024 0.412559986 T T T

Can you please suggest what parameters in the kpoints/incar/poscar should i play with to move the calculations further ? please please please help. i don't like calculations at all but thats a part of my job

[alex]

Yep, it takes ages. For that setup (esp. the number of k-point's I'd guess days(!) for one(!) electronic step even with many cores in action!

In your case you want to start with PBE+U.


Cheers,

alex

[askhetan]

alright, so i modified my INCAR file for just a CeO2(111) surface slab calculation. The incar file now has this desciption:

LASPH = .TRUE.
LDAU = .TRUE.
LDAUTYPE = 2 3 # for O, Ce
LDAUL = -1 3
LDAUU = 0 5
LDAUJ = 0 1
LMAXMIX=6

The question I have here is - can I specify different types of LDAUTYPE for different species in the system (like i've done hereby setting 2 for O and 3 for Ce, in the same order that they appear in my POSCAR) ? or should i just put one value of LDAUTYPE?

Please help!! the hybrid dft HSE06 calculations don't seem to move forward, need to use DFT+U

[alex]

Looks good except for LDAUTYPE. Only one type should be allowed.
And yes, you may have different U_eff for the same species. You would also need extra POTCAR entry for that.

Cheers,

alex

[askhetan]

Cheers alex!

I removed multiple values of LDAUTYPE and just used 3. Of course you're right, the POTCAR needs to have both the entries for O and Ce. Funnily though, my calculations with even multiple values of LDAUTYPE converged.

Another question I had was if there is any particular type of ISMEAR & SIGMA values that one should use for CeO2 system. with the default settings the entropy value (T*S) i get is about 20 meV but VASP manual recommends adjusting SIGMA till the (T*S) reaches a value less than 1 meV??

[alex]

Hi askhetan,

I'd guess your 'misplaced' LDAUTYPE calculation just took one value.

About ceria: I'd also stick to 1 meV for SIGMA, otherwise you might easily end up in some metallic states during the iterations and then ... you are really in trouble.

Cheers,

alex