HF convergence problem

[wcwang]

Hi, All,

I got convengence problem when I calculated HfO2 system(96 atoms) with impurity doping. Before doing doping in HfO2, pure bulk(96 atoms include monoclic and cubic phases) calcualtion showed the convergence of HSE correction went well. It means the HSE correction could be finished within 6 interactions. In this successful pure bulk calculation, I used one kpoint and INCAr is posted here:
SYSTEM = HfO2

ISTART =1

ICHARG =11

ISMEAR = 0

LREAL= Auto

#NELM=120



# HF calculation

LHFCALC=.TRUE.

HFSCREEN=0.2

ENCUTFOCK=0

#NKRED=2

AEXX = 0.25

ALGO=ALL

NBANDS=350

TIME=0.4

#ISYM=0

#SHIFTRED= T

#LMAXMIX=6

However, I got trouble to do the correction with impurity doping in HfO2 system including monoclinic and cubic phase. Whatever substitution or interstitials, I never succeed. INCAR was as same as that in pure bulk calcuation. And ISYM was also swithed on and off to change the symetry of the system. Whatever I tried,it did not work. It ended up with the following errors:

....
ROT: 59 -0.723136219943E+05 -0.11747E+05

ROT: 60 -0.605663335891E+05 0.11747E+05

gam= 8.094 g(H,U,f)= 0.442E+04 0.511E+04-0.605E+03 ort(H,U,f) =-0.621E-01 0.143E+04 0.112E+04

LAPACK: Routine ZPOTRF failed! 157

.......

Some recent literature reported Al2O3 doping with HF(HSE) correction. So that I believe VASP has no problem to do HF correction, most likely, I missed something in the INCAR which is not what I know. I've been working on this issue for a couple of day, still no clue about this.

Any advice and suggestion is appreciated.

[Danny]

have yu tried switching ALGO to Damped?
(I have one case this solved my problem...partially)

Danny

[wcwang]

Hi, Danny, thanks for the advice. However It did not work since I tried it. I also tried to switch ISMEAR to metal cases, it did not work either. Furthermore, I varied TIME flag from 0.1 to 1.0, still it failed. I feel fustrated about the correction now.

[admin]

please check (eg from spectra) whether the f-electrons in Hf are in valence or localized (in the core) at the Hf sites next to the doping sites. It may be necessary to use 2 different Hf PPs (f in valence and in core) and define 2 in-equivalent types of Hf atoms therefore (those next to the impurity and those in the unperturbed bulk)

[Danny]

[quote="wcwang"]Hi, Danny, thanks for the advice. However It did not work since I tried it. I also tried to switch ISMEAR to metal cases, it did not work either. Furthermore, I varied TIME flag from 0.1 to 1.0, still it failed. I feel fustrated about the correction now.[/quote]

Bummer, I seem to have some similar mess, but with a simple 1atom alpha-Ce system...convergence is not really reasonable using HSE.

@admin: If you know the answer (i.e. the electronic structure of the atoms, localized of delocalized f's) than doing the calculations seems an irrelevant exercise to me, total loss of predictive power for new systems and I do not think this should be the idea behind the use of a complex code like VASP.

Danny

[wcwang]

In PBE potetials, Hf has two potentials. f electrons are not included in valence part in both cases. However, I used two different potentials for Hf as admin suggested, it failed again.