Dear all,
In the G0W0 calculation for Europium oxide, from the version upper to 6.4 (including the 6.9) of Abinit I obtain a systematic segmentation fault at the level of the sigma calculation in the PAW case.
Indeed I can read in the log file :
Calculating <nk|Sigma_x|nk> at k= 0.000 0.000 0.000
bands n = from 5 to 25
m_wfs.F90:4706:COMMENT
Changing FFT mesh
-P-0000 Will sum 267 (b,k,s) occupied states in Sigma_x.
calc_sigx_me: calculation status ( 64 to be completed):
[node031:07851] *** Process received signal ***
[node031:07851] Signal: Segmentation fault (11)
[node031:07851] Signal code: (128)
[node031:07851] Failing at address: (nil)
Notice that the G0W0 for the Europium oxide for release 6.2 of Abinit is normally ended (despite the fact that we observe an over correction for the f-states).
The main characteristic of EuO is its Ferromagnetic insulator character with 4f states (use of tsmear, spinpol, etc). However it is a half-metal in the DFT calculation and the GW correction is supposed to open the gap.
This different avenue I tried but it did not solve the problem
- Use different values of gwcalctyp, COHSEX calculation
- Use of contour deformation or plasmon-pole model
- Use different PAW atomic data,
- Including or not the compensation charge density with the PAW atomic data
- Use or not of semi-core states in the Eu PAW atomic data
- Make the calculation in serie or in parallel over the number of bands
Note that an equivalent Hartree-Fock calculation give rise to the same segmentation fault, so the problem can not be uniquely ascribed to the correlation part.
This is the input file...
Bruno Bertrand
# #########################
# # Europium monoxide EuO #
# #########################
# Calculation of the GW corrections of EuO at Gamma
# -------------------------------------------------
# Dataset 1: Calculation of the density file
# Dataset 2: calculation of the kss file for k-points in IBZ
# Dataset 3: calculation of the screening (epsilon^-1 matrix for W)
# Dataset 4: calculation of the Self-Energy matrix elements (GW corrections)
chkexit 1
ndtset 2 # 4 dataset
jdtset 3 4 # Only one run with the generic variables and those with index 'j'
# ---------------------------------------------
#| Dataset 1: Calculation of the density file |
# ---------------------------------------------
# Definition of the k-point grid
# ------------------------------
# Shifted to optimize the calculation of density
kptopt1 1 # Option for the automatic generation of k points
ngkpt1 4 4 4 # Great Density of k points = great precision for density
nshiftk1 4
shiftk1 0.5 0.5 0.5
0.5 0.0 0.0
0.0 0.5 0.0
0.0 0.0 0.5
nbdbuf1 5 # Good compromise between tolvrs (toldff) and tolwfr
#nband1 11 # valence 11 => HOMO half-filled
nband1 30 # Specify slightly more than the occupied bands
# to help convergence, so including the buffering bands
# Definition of the SCF procedure
# -------------------------------
nstep1 80 # Maximal number of SCF cycles
tolvrs1 1.0d-14 # Calcul of density => Tolerance on Potential energy residual
diemac1 4.0
prtden1 1 # Print out density
# -------------------------------------------------------------
#| Dataset 2 Generation of the Khom-Sham structure : KSS file |
# -------------------------------------------------------------
kssform2 3
# Definition of the k-point grid
# ------------------------------
kptopt2 1
ngkpt2 4 4 4
nshiftk2 1
shiftk2 0.0 0.0 0.0
# Number of bands
# ---------------
nbdbuf2 100
nband2 1100 # Number of (occ and empty) bands to be computed
nbandkss2 1000 # Number of bands in KSS file
# Optimization
# ------------
nline2 6 # More CG minimisation steps
getden2 1 # Take the output density of a previous dataset (1)
#and use them as input density (2), with iscf<0
# Definition of the SCF procedure
# -------------------------------
iscf2 -2 # Do a non-self consistent calc. reading
# in the density (_DEN) file
tolwfr2 1.0d-14 # when iscf<0
nstep2 60
# -------------------------------------------
#| Dataset 3 : Calculation of the screening |
# -------------------------------------------
optdriver3 3 # Screening calculation (epsilon^-1 matrix)
# Number of bands
# ---------------
nband3 991
# Planewave set
# -------------
ecutwfn3 45
ecuteps3 6
# Optimization
# ------------
gwcalctyp3 2
nfreqim3 6
nfreqre3 25
freqremax3 100
getkss3 2 # Get KSS file from dataset 2
fftgw3 21 # 11 : Reduce the memory requirement by using a coarse fft grid
gwpara3 2 # Use parallellism over bands
awtr3 1 # Use time-reversal symmetry
symchi3 1 # Use crystal symmetries in calc. of Chi
inclvkb 2 # The commutator of the non-local part of the Hamiltonian with the position operator is correctly included in the q => 0 contribution
# Definition of the k-points
# --------------------------
kptopt3 1
ngkpt3 4 4 4
nshiftk3 1
shiftk3 0.0 0.0 0.0
# -----------------------------------
#| Dataset 4 : Calculation of Sigma |
# -----------------------------------
optdriver4 4 # Calculation of the Self-Energy matrix elements (GW corrections)
# Number of bands
# ---------------
nband4 1000
bdgw4 5 25 # Calculate GW corrections for bands from 91 to 95
5 25
5 25
5 25 # Two before the gap and 3 after due to intertwining and degeneracy
5 25
5 25
5 25
5 25
# bdgw4 1 1 20 20 1 1 20 20 1 1 20 20 1 1 20 20
# Planewave set
# -------------
ecutwfn4 45.0
ecutsigx4 45.0
# Optimization
# ------------
getkss4 2 # Obtain KSS file from dataset 7
getscr4 3
gwcalctyp4 2
nfreqim4 6
nfreqre4 25
freqremax4 100
# gwmem4 00 # Screening matrix read just a q-vector after another
# => reduce the memory needs but make the calculus slower
fftgw4 21
gwpara4 2 # Use parallellism over bands
# Definition of the k-points
# --------------------------
kptopt4 1
ngkpt4 4 4 4
nshiftk4 1
shiftk4 0.0 0.0 0.0
nkptgw4 4 # number of k-point where to calculate the GW correction
kptgw4 0.000 0.000 0.000 # Gamma
0.500 0.000 0.000 # L
0.500 0.250 0.000 # W
0.500 0.500 0.000 # X'
# ==================================================================
# || ||
# || THE INPUT VARIABLES BELOW ARE IDENTICAL FOR ALL DATASETS ||
# || ||
# ==================================================================
# Atoms in the base
# -----------------
natom 2
ntypat 2
typat 1*1 1*2
xred 0.00000 0.00000 0.00000 # Eu
1/2 1/2 1/2 # 0
znucl 63.0 8.0
nsym 0 # Find autom. the symmetry operations leaving atomic sublattices invariant. Checks whether the cell is primitive
# Definition of the unit cell
# ---------------------------
rprim 0.0 0.5 0.5 # FCC primitive vectors (to be scaled by acell)
0.5 0.0 0.5
0.5 0.5 0.0
acell 3*9.72 # experimental value
# Optimisation in ecut
# --------------------
ecut 45 # Maximal plane-wave kinetic energy cut-off, in Hartree
# To be slightly raised
pawecutdg 120
# To be raised to 100 ! see log files
ecutsm 0.5
# Spin polarisation and half-metallic character
# ---------------------------------------------
nsppol 2
nspinor 1
nspden 2 # scalar magnetization (axis arbitrarily fixed in the z direction)
# Spin up chosen
occopt 7
tsmear 0.001
spinat 0 0 2
0 0 0
# Other
# -----
enunit 2
istwfk *1 # Bug of abinit related to symmetry
ixc 11 #GGA calculus
GW+PAW - EuO - Segmentation fault
Moderators: maryam.azizi, bruneval