ABINIT Discussion Forums

Dear all,
I'm trying to determine the U parameter for ferromagnetic FePO4 using the linear response approach described in the relative tutorial: I produced a WFK file using the geometry with full symmetry using:

#Unit cell
acell 10.332 6.010 4.692 angstrom
chkprim 0 # 0: do not check if uc primitive
natom 24
ntypat 3
typat 1 2 3 3 3 1 2 3 3 3 1 2 3 3 3 1 2 3 3 3 3 3 3 3
znucl 26 15 8
xred
2.8200000000E-01 2.5000000000E-01 9.7500000000E-01 # Fe
9.5000000000E-02 2.5000000000E-01 4.1800000000E-01 # P
9.7000000000E-02 2.5000000000E-01 7.4300000000E-01 # O
4.5700000000E-01 2.5000000000E-01 2.0600000000E-01 # O
1.6500000000E-01 4.6000000000E-02 2.8400000000E-01 # O
2.1800000000E-01 7.5000000000E-01 4.7500000000E-01 # Fe
4.0500000000E-01 7.5000000000E-01 9.1800000000E-01 # P
4.0300000000E-01 7.5000000000E-01 2.4300000000E-01 # O
4.3000000000E-02 7.5000000000E-01 7.0600000000E-01 # O
3.3500000000E-01 9.5400000000E-01 7.8400000000E-01 # O
7.1800000000E-01 7.5000000000E-01 2.5000000000E-02 # Fe
9.0500000000E-01 7.5000000000E-01 5.8200000000E-01 # P
9.0300000000E-01 7.5000000000E-01 2.5700000000E-01 # O
5.4300000000E-01 7.5000000000E-01 7.9400000000E-01 # O
8.3500000000E-01 5.4600000000E-01 7.1600000000E-01 # O
7.8200000000E-01 2.5000000000E-01 5.2500000000E-01 # Fe
5.9500000000E-01 2.5000000000E-01 8.2000000000E-02 # P
5.9700000000E-01 2.5000000000E-01 7.5700000000E-01 # O
9.5700000000E-01 2.5000000000E-01 2.9400000000E-01 # O
6.6500000000E-01 4.5400000000E-01 2.1600000000E-01 # O
8.3500000000E-01 9.5400000000E-01 7.1600000000E-01 # O
6.6500000000E-01 4.6000000000E-02 2.1600000000E-01 # O
1.6500000000E-01 4.5400000000E-01 2.8400000000E-01 # O
3.3500000000E-01 5.4600000000E-01 7.8400000000E-01 # O

#Spin polarization
nsppol 2 #1 unpolarized / 2 polarized
spinat 0 0 2.843062
0 0 0
0 0 0
0 0 0
0 0 0
0 0 2.843062
0 0 0
0 0 0
0 0 0
0 0 0
0 0 2.843062
0 0 0
0 0 0
0 0 0
0 0 0
0 0 2.843062
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0

# Parameters
ecut 8 # Very low Energy cutoff, just to verify the procedure
pawecutdg 20

#Definition of the SCF procedure
nstep 2 # max number SCF cycles
tolvrs 10d-12

#Definition of the k-point grid
kptopt 1 # 1: automatic generation of k points
ngkpt 1 2 3 # n x n x n
nshiftk 1
shiftk 0.5 0.5 0.5

#Smearing
occopt 4
tsmear 0.05 eV

#DFT+U
usepawu 1 # 1 at lim dble cnt / 2 rnd m fld dle cnt
lpawu 2 -1 -1 -1 -1 2 -1 -1 -1 -1 2 -1 -1 -1 -1 2 -1 -1 -1 -1 -1 -1 -1 -1 # ang moments corrrected

#Save disk space & Miscelaneous
prteig 0
prtden 1 # This is the default value
optforces 2

...
then I've broken the symmetry (from nsym 8 to 2) substituting the fisrt Fe atoms with another species. The symrel obtained were inserted in the new input file and the WFK from the first run was used to as a starting wave function:

---- same as input 1 -----

#DFT+U
usepawu 1 # 1 at lim dble cnt / 2 rnd m fld dle cnt
lpawu 2 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 # ang moments corrrected
nsym 2 # nsym&symrel: break symmetry of crystal: allow individual ionicoccupations
symrel 1 0 0 0 1 0 0 0 1 1 0 0 0 -1 0 0 0 1
tnons 0.0000000 0.0000000 0.0000000 0.0000000 0.5000000 0.0000000
pawujat 1 # default, the atom on which U is determined
pawujv 0.5 eV # default, size of the potential shift
macro_uj 3 # activate determination of U
pawujrad 2.5 # optional, radius ASA-sphere to which U should be extrapolated
dmatpuopt 2
#Only to accelerate test
irdwfk 1 # default for macro_uj = 1
# nline 2
# nnsclo 2
tolvrs 10d-9 # default for macro_uj = 1

#Save disk space & Miscelaneous
prteig 0
prtden 1 # This is the default value
optforces 2

...

Abinit starts smoothly and then crashes (segmentation fault) during the first SCF iteration with the following message:

ITER STEP NUMBER 1
vtorho : nnsclo_now= 2, note that nnsclo,dbl_nnsclo,istep= 0 0 1
**** In vtorho for isppol= 1
-P-0000 fxphas : BUG -
-P-0000 The eigenvector number 1 has zero norm.
-P-0000
-P-0000 leave_new : decision taken to exit ...

I tried to change various parameter with no success. Is this a real bug or I'm missing something?

Thanks a lot in advance and best regards,
Mauro.