abinit/tests/paral/Input/t77.abi

107 lines
4.0 KiB
Plaintext

# Molecular SiH4 in GWLS (G0W0 with Lanczos basis and Sternheimer equations)
# There are 3 datasets
# 1: Calculation of the Ground State density
# 2: Calculation of the Ground State wavefunctions
# 3: Calculation of the Self-Energy matrix element <gwls_band_index|\Sigma(gw_freqsp)|gwls_band_index>
ndtset 3
### Specific variables
# Ground state calculation (density)
toldfe1 1d-10
paral_kgb1 0 # Parallel ground state calculations don't work for molecules (the minimisation algorithm becomes unstable).
# Ground state calculation (wavefunctions)
tolwfr2 1d-25
getden2 1
iscf2 -2
paral_kgb2 0
# GWLS calculation for the HOMO
tolwfr3 1d-25 # This is the precision to which the Sternheimer eqs. are solved in GWLS.
getden3 1
getwfk3 2
optdriver3 66 # This is the optdriver triggering a GWLS calculation.
gwls_band_index3 4 # This is the band for which the matrix element of the Self-energy is requested : <gwls_band_index|\Sigma(gw_freqsp)|gwls_band_index>
gw_customnfreqsp3 1 # This is the number if frequencies at which the matrix element of the Self-energy is requested.
gw_freqsp3 -0.5558150504 # These are the frequencies at which the matrix element of the Self-energy is requested.
gwls_stern_kmax3 8 # This controls the convergence with respect to the size of the dielectric matrix.
gwls_kmax_complement3 8 # This controls the convergence with respect to the size of the MODEL dielectric matrix.
gw_icutcoul3 0 # For molecules, the coulomb potential needs to be spherically truncated.
rcut3 9.0 # The spherical truncation has this radius.
prtwf3 0 # By default, WFs would be printed with mpiio. However, the NAG compiler (on petrus) takes forever to do so. Since these WFs are totally unuseful and this is not a mpiio test, printing WFs is disabled.
prtden3 0 # Same reason as WFs.
gwls_recycle3 1 # The writing to file (gwls_recycle3 2) is more desirable as it saves a lot of RAM on production runs, but petrus + NAG 5.3.1 + OpenMPI 1.6.2 cause a BUG with writing / reading unformatted fortran files. For the time being, we won't test this, then.
### Common variables
# Other parameters
istwfk *1 # GWLS doesn't support time reversal symetry.
nline 100 # GWLS linear equation solver may require many steps at problematic frequencies.
# Number of bands
# NOTE : The last band must be higher in energy that the higher gw_customnfreqsp requested by the user.
nband 10
# Parallelisation : GWLS supports parallelism over bands and FFTs (from the ground state implementation),
# but FFT parallelism tends to have poor efficiency in GWLS.
# The optimum is usually to set the highest possible npband.
paral_kgb 1
npfft 2
npband 2
bandpp 1 # old default value
# Cutoff energy
ecut 1.0
# Definition of the unit cell
acell 3*18
rprim 1.0 0.0 0.0
0.0 1.0 0.0
0.0 0.0 1.0
# Definition of the atom types
ntypat 2
znucl 14 1
# Definition of the atomic structure
natom 5
typat 1 2 2 2 2
xcart
0.00000000000000 0.00000000000000 0.00000000000000
1.63281519635864 1.63281519635864 1.63281519635864
-1.63281519635864 -1.63281519635864 1.63281519635864
1.63281519635864 -1.63281519635864 -1.63281519635864
-1.63281519635864 1.63281519635864 -1.63281519635864
# BZ sampling for GW.
kptopt 1 # Option for the automatic generation of k points,
nkpt 1
ngkpt 1 1 1
nshiftk 1
shiftk 0.0 0.0 0.0
pp_dirpath "$ABI_PSPDIR"
pseudos "PseudosGTH_pwteter/14si.pspgth, PseudosGTH_pwteter/01h.pspgth"
#%%<BEGIN TEST_INFO>
#%% [setup]
#%% executable = abinit
#%% [files]
#%% files_to_test =
#%% [paral_info]
#%% nprocs_to_test = 4
#%% max_nprocs = 4
#%% [NCPU_4]
#%% files_to_test = t77_MPI4.abo, tolnlines= 20, tolabs = 1.0e-4, tolrel = 1.0e-4, fld_options = -ridiculous;
#%% [extra_info]
#%% authors = J. Laflamme Janssen
#%% keywords = GW, GWLS
#%% description =
#%% Molecular SiH4 in GWLS (G0W0 with Lanczos basis and Sternheimer equations)
#%% topics = GWls
#%%<END TEST_INFO>