I am trying to calculate Raman Tensor and Raman Intensities of a 56 atoms size non-magnetic cubic spinel structure. I am using PAW pseudopotentials. The approach is using sternheimer like under Miwa scheme following the tutorial presented in abinit tutorial page (tnlo_6.abi). Below is the input file I am using now.
ndtset 7
#DATASET1 : scf calculation: GS WF in the IBZ
#********************************************
prtden1 1 # save density on disk, will be used in other datasets
prtwf1 1 # save WF on disk, will be used in other datasets
kptopt1 1 # use Irreducible Brillouin Zone (all symmetry taken into account)
toldfe1 2.0d-7 # SCF convergence criteria (could be tolwfr or tolvrs)
#DATASET2 : non scf calculation: GS WF in half BZ
#*****************************************************
getden2 1 # use density from dataset 1
kptopt2 2 # use Half Brillouin Zone (only time-reversal symmetry taken into account)
getwfk2 1 # use GS WF from dataset 1 (as input)
iscf2 -2 # non-self-consistent calculation
tolwfr2 1.5d-3 # convergence criteria on WF, need high precision for response
prtwf2 1 # save WF on disk, will be used in other datasets
#DATASET3 : derivative of WF with respect to k points (d/dk)
#**********************************************************
getwfk3 2 # use GS WF from dataset 2
kptopt3 2 # use Half Brillouin Zone (only time-reversal symmetry taken into account)
rfelfd3 2 # compute 1st-order WF derivatives (d/dk)
tolwfr3 4.0d-3 # convergence criteria on WF, need high precision for response
prtwf3 1 # save 1st-order WF on disk, will be used in other datasets
#DATASET4 : response functions (2nd derivatives of E) and corresponding 1st order WF derivatives phonons, electric fields, and strains are all done
#**************************************************************
getwfk4 2 # use GS WF from dataset 2
kptopt4 2 # use Half Brillouin Zone (only time-reversal symmetry taken into account)
getddk4 3 # use ddk WF from dataset 3 (needed for electric field)
rfphon4 1 # compute 1st-order WF derivatives with respect to atomic displacements…
rfelfd4 3 # compute 1st-order WF derivatives with respect to electric field
rfstrs4 3 # compute 1st-order WF derivatives with respect to strains
toldfe4 2.0d-4 # SCF convergence criteria (could be tolwfr)
prepanl4 1 # make sure that response functions are correctly prepared for a non-linear computation
prtwf4 1 # save 1st-order WF on disk, will be used in other datasets
prtden4 1 # save 1st-order density on disk, will be used in other datasets
DATASET5 : 2nd order WF derivatives with respect to k-points (d/dkdk)
rf2_dkdk5 1 # compute 2nd-order WF derivatives (d/dkdk)
getwfk5 2 # use GS WF from dataset 2
kptopt5 2 # use Half Brillouin Zone (only time-reversal symmetry taken into account)
getddk5 3 # use 1st-order WF from dataset 3
prepanl5 1 # compute only needed directions for the non-linear computation
tolwfr5 4.0d-3 # convergence criteria on WF, need high precision for response
prtwf5 1 # save 2nd-order WF on disk, will be used in other datasets
#DATASET6 : 2nd order WF derivatives with respect to electric field and k-points (d/dkde)
#*****************************************************************************************
rf2_dkde6 1 # compute 2nd-order WF derivatives (d/dkde)
getwfk6 2 # use GS WF from dataset 2
getddk6 3 # use 1st-order WF (d/dk) from dataset 3
get1den6 4 # use 1st-order densities from dataset 4
getdelfd6 4 # use 1st-order WF (d/de) from dataset 4
getdkdk6 5 # use 2nd-order WF (d/dkdk) from dataset 5
prepanl6 1 # compute only directions needed for nonlinear computation
kptopt6 2 # use Half Brillouin Zone (only time-reversal symmetry taken into account)
prtwf6 1 # save 2nd-order WF on disk, will be used in other datasets
tolwfr6 4.0d-3 # convergence criteria on WF, need high precision for response
#DATASET7 : 3rd derivatives of E
#*********************************
getwfk7 2 # use GS WF from dataset 2
getddk7 3 # use 1st-order WF (d/dk) from dataset 3
get1den7 4 # use 1st-order densities from dataset 4
get1wf7 4 # use 1st-order WFs from dataset 4
getdkde7 6 # use 2nd-order WF (d/dkde) from dataset 6
kptopt7 2 # use Half Brillouin Zone (only time-reversal symmetry taken into account)
optdriver7 5 # compute 3rd order derivatives of the energy
usepead7 0 # no use of pead method, so use full DFPT instead
d3e_pert1_elfd7 1 # activate electric field for 1st perturbation…
d3e_pert1_phon7 1 # …and also atomic displacements…
d3e_pert1_atpol7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 # …for all atoms (so here only 1 and 2)…
d3e_pert2_elfd7 1 # activate electric field for 2nd perturbation…
d3e_pert3_elfd7 1 # activate electric field for 3rd perturbation…
ecut 60 # this value is very low but is used here to achieve very low calculation times.
# in a production environment this should be checked carefully for convergence and
# a more reasonable value is probably around 30
pawecutdg 120
pawxcdev 0
ixc 7
ecutsm 0.5
dilatmx 1.05
ngkpt 2 2 2 # must be checked carefully for convergence, Raman calculations converge slowly with kpt
nshiftk 1 # this Monkhorst-Pack shift pattern is used so that the symmetry of the shifted grid
# is correct. A gamma-centered grid would also have the correct symmetry but would be
# less efficient.
shiftk 0 0 0
nstep 100 # limit number of steps for the tutorial, in general this should be set to its default (30) or higher
#suppress printing of density, wavefunctions, etc except what is
#explicitly requested above in the ndtset section
prtwf 0
prtden 0
prteig 0
When I submit the job it runs upto DATASET 4 completely and converges in each dataset. Later when it starts to calculate DATASET 5 it crashes and doesn’t print any error message at the end. Here’s the few lines from the end -
================================================================================
== DATASET 5 ==================================================================
- mpi_nproc: 320, omp_nthreads: 1 (-1 if OMP is not activated)
— !DatasetInfo
iteration_state: {dtset: 5, }
dimensions: {natom: 56, nkpt: 8, mband: 232, nsppol: 1, nspinor: 1, nspden: 1, mpw: 97137, }
cutoff_energies: {ecut: 60.0, pawecutdg: 120.0, }
electrons: {nelect: 4.64000000E+02, charge: 0.00000000E+00, occopt: 1.00000000E+00, tsmear: 1.00000000E-02, }
meta: {optdriver: 1, }
…
mkfilename : getwfk/=0, take file _WFK from output of DATASET 2.
mkfilename : getddk/=0, take file _1WF from output of DATASET 3.
Exchange-correlation functional for the present dataset will be:
LDA: Perdew-Wang 92 LSD fit to Ceperley-Alder data - ixc=7
Citation for XC functional:
J.P.Perdew and Y.Wang, PRB 45, 13244 (1992)
Real(R)+Recip(G) space primitive vectors, cartesian coordinates (Bohr,Bohr^-1):
R(1)= 15.5581242 0.0000000 0.0000000 G(1)= 0.0642751 0.0000000 0.0000000
R(2)= 0.0000000 15.5581242 0.0000000 G(2)= 0.0000000 0.0642751 0.0000000
R(3)= 0.0000000 0.0000000 15.5581242 G(3)= 0.0000000 0.0000000 0.0642751
Unit cell volume ucvol= 3.7659253E+03 bohr^3
Angles (23,13,12)= 9.00000000E+01 9.00000000E+01 9.00000000E+01 degrees
Coarse grid specifications (used for wave-functions):
getcut: wavevector= 0.0000 0.0000 0.0000 ngfft= 120 120 120
ecut(hartree)= 66.150 => boxcut(ratio)= 2.10666
Fine grid specifications (used for densities):
getcut: wavevector= 0.0000 0.0000 0.0000 ngfft= 162 162 162
ecut(hartree)= 132.300 => boxcut(ratio)= 2.01100
==> initialize data related to q vector <==
The list of irreducible perturbations for this q vector is:
================================================================================
Perturbation wavevector (in red.coord.) 0.000000 0.000000 0.000000
Perturbation : 2nd derivative wrt k, idir1 = 1 idir2 = 1
dfpt_looppert : COMMENT -
In this case, iscf is set to -3 automatically.
symkpt : not enough symmetry to change the number of k points.
Initialisation of the first-order wave-functions :
ireadwf= 0
- dfpt_looppert: read the DDK wavefunctions from file: tnlo_6o_DS3_1WF169
— !BeginCycle
iteration_state: {dtset: 5, }
solver: {iscf: 7, nstep: 100, nline: 4, wfoptalg: 10, }
tolerances: {tolwfr: 4.00E-03, }
…
iter 2DEtotal(Ha) deltaE(Ha) residm vres2
In this situation how should I proceed? What could be the possible reason of this to get failed? This is the performance of the code -
State: FAILED (exit code 127)
Nodes: 8
Cores per node: 40
CPU Utilized: 191-11:56:47
CPU Efficiency: 99.25% of 192-22:29:20 core-walltime
Job Wall-clock time: 14:28:13
Memory Utilized: 770.98 GB (estimated maximum)
Memory Efficiency: 38.55% of 1.95 TB (250.00 GB/node)