2013-07-23 19:49:42 +02:00
|
|
|
|
|
|
|
################################################################################
|
|
|
|
#
|
|
|
|
# TRIQS: a Toolbox for Research in Interacting Quantum Systems
|
|
|
|
#
|
|
|
|
# Copyright (C) 2011 by M. Aichhorn, L. Pourovskii, V. Vildosola
|
|
|
|
#
|
|
|
|
# TRIQS is free software: you can redistribute it and/or modify it under the
|
|
|
|
# terms of the GNU General Public License as published by the Free Software
|
|
|
|
# Foundation, either version 3 of the License, or (at your option) any later
|
|
|
|
# version.
|
|
|
|
#
|
|
|
|
# TRIQS is distributed in the hope that it will be useful, but WITHOUT ANY
|
|
|
|
# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
|
|
|
|
# FOR A PARTICULAR PURPOSE. See the GNU General Public License for more
|
|
|
|
# details.
|
|
|
|
#
|
|
|
|
# You should have received a copy of the GNU General Public License along with
|
|
|
|
# TRIQS. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
#
|
|
|
|
################################################################################
|
|
|
|
|
|
|
|
from types import *
|
|
|
|
import numpy
|
|
|
|
from pytriqs.archive import *
|
|
|
|
import pytriqs.utility.mpi as mpi
|
|
|
|
import string
|
|
|
|
|
|
|
|
|
|
|
|
def read_fortran_file (filename):
|
|
|
|
""" Returns a generator that yields all numbers in the Fortran file as float, one by one"""
|
|
|
|
import os.path
|
2014-09-22 19:24:33 +02:00
|
|
|
if not(os.path.exists(filename)) : raise IOError, "File %s does not exist."%filename
|
2013-07-23 19:49:42 +02:00
|
|
|
for line in open(filename,'r') :
|
|
|
|
for x in line.replace('D','E').split() :
|
|
|
|
yield string.atof(x)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class Wien2kConverter:
|
|
|
|
"""
|
2014-09-22 19:24:33 +02:00
|
|
|
Conversion from Wien2k output to an hdf5 file that can be used as input for the SumkLDA class.
|
2013-07-23 19:49:42 +02:00
|
|
|
"""
|
|
|
|
|
|
|
|
def __init__(self, filename, lda_subgrp = 'SumK_LDA', symm_subgrp = 'SymmCorr', repacking = False):
|
|
|
|
"""
|
2014-09-22 19:24:33 +02:00
|
|
|
Init of the class. Variable filename gives the root of all filenames, e.g. case.ctqmcout, case.h5, and so on.
|
2013-07-23 19:49:42 +02:00
|
|
|
"""
|
|
|
|
|
|
|
|
assert type(filename)==StringType,"LDA_file must be a filename"
|
|
|
|
self.hdf_file = filename+'.h5'
|
|
|
|
self.lda_file = filename+'.ctqmcout'
|
|
|
|
self.symm_file = filename+'.symqmc'
|
|
|
|
self.parproj_file = filename+'.parproj'
|
|
|
|
self.symmpar_file = filename+'.sympar'
|
|
|
|
self.band_file = filename+'.outband'
|
|
|
|
self.lda_subgrp = lda_subgrp
|
|
|
|
self.symm_subgrp = symm_subgrp
|
|
|
|
|
|
|
|
# Checks if h5 file is there and repacks it if wanted:
|
|
|
|
import os.path
|
|
|
|
if (os.path.exists(self.hdf_file) and repacking):
|
|
|
|
self.__repack()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def convert_dmft_input(self):
|
|
|
|
"""
|
|
|
|
Reads the input files, and stores the data in the HDFfile
|
|
|
|
"""
|
|
|
|
|
|
|
|
|
2014-09-22 19:24:33 +02:00
|
|
|
# Read and write only on the master node
|
|
|
|
if not (mpi.is_master_node()): return
|
2013-07-23 19:49:42 +02:00
|
|
|
mpi.report("Reading input from %s..."%self.lda_file)
|
|
|
|
|
|
|
|
# R is a generator : each R.Next() will return the next number in the file
|
|
|
|
R = read_fortran_file(self.lda_file)
|
|
|
|
try:
|
|
|
|
energy_unit = R.next() # read the energy convertion factor
|
|
|
|
n_k = int(R.next()) # read the number of k points
|
|
|
|
k_dep_projection = 1
|
|
|
|
SP = int(R.next()) # flag for spin-polarised calculation
|
|
|
|
SO = int(R.next()) # flag for spin-orbit calculation
|
|
|
|
charge_below = R.next() # total charge below energy window
|
|
|
|
density_required = R.next() # total density required, for setting the chemical potential
|
|
|
|
symm_op = 1 # Use symmetry groups for the k-sum
|
|
|
|
|
|
|
|
# the information on the non-correlated shells is not important here, maybe skip:
|
|
|
|
n_shells = int(R.next()) # number of shells (e.g. Fe d, As p, O p) in the unit cell,
|
|
|
|
# corresponds to index R in formulas
|
|
|
|
shells = [ [ int(R.next()) for i in range(4) ] for icrsh in range(n_shells) ] # reads iatom, sort, l, dim
|
2014-09-22 19:24:33 +02:00
|
|
|
|
2013-07-23 19:49:42 +02:00
|
|
|
n_corr_shells = int(R.next()) # number of corr. shells (e.g. Fe d, Ce f) in the unit cell,
|
|
|
|
# corresponds to index R in formulas
|
|
|
|
# now read the information about the shells:
|
|
|
|
corr_shells = [ [ int(R.next()) for i in range(6) ] for icrsh in range(n_corr_shells) ] # reads iatom, sort, l, dim, SO flag, irep
|
|
|
|
|
|
|
|
self.inequiv_shells(corr_shells) # determine the number of inequivalent correlated shells, has to be known for further reading...
|
|
|
|
|
|
|
|
use_rotations = 1
|
|
|
|
rot_mat = [numpy.identity(corr_shells[icrsh][3],numpy.complex_) for icrsh in xrange(n_corr_shells)]
|
|
|
|
|
|
|
|
# read the matrices
|
|
|
|
rot_mat_time_inv = [0 for i in range(n_corr_shells)]
|
|
|
|
|
|
|
|
for icrsh in xrange(n_corr_shells):
|
|
|
|
for i in xrange(corr_shells[icrsh][3]): # read real part:
|
|
|
|
for j in xrange(corr_shells[icrsh][3]):
|
|
|
|
rot_mat[icrsh][i,j] = R.next()
|
|
|
|
for i in xrange(corr_shells[icrsh][3]): # read imaginary part:
|
|
|
|
for j in xrange(corr_shells[icrsh][3]):
|
|
|
|
rot_mat[icrsh][i,j] += 1j * R.next()
|
|
|
|
|
|
|
|
if (SP==1): # read time inversion flag:
|
|
|
|
rot_mat_time_inv[icrsh] = int(R.next())
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-09-22 19:24:33 +02:00
|
|
|
# Read here the info for the transformation of the basis:
|
2013-07-23 19:49:42 +02:00
|
|
|
n_reps = [1 for i in range(self.n_inequiv_corr_shells)]
|
|
|
|
dim_reps = [0 for i in range(self.n_inequiv_corr_shells)]
|
|
|
|
T = []
|
|
|
|
for icrsh in range(self.n_inequiv_corr_shells):
|
|
|
|
n_reps[icrsh] = int(R.next()) # number of representatives ("subsets"), e.g. t2g and eg
|
|
|
|
dim_reps[icrsh] = [int(R.next()) for i in range(n_reps[icrsh])] # dimensions of the subsets
|
|
|
|
|
2014-09-22 19:24:33 +02:00
|
|
|
# The transformation matrix:
|
|
|
|
# is of dimension 2l+1 without SO, and 2*(2l+1) with SO!
|
2013-07-23 19:49:42 +02:00
|
|
|
ll = 2*corr_shells[self.invshellmap[icrsh]][2]+1
|
|
|
|
lmax = ll * (corr_shells[self.invshellmap[icrsh]][4] + 1)
|
|
|
|
T.append(numpy.zeros([lmax,lmax],numpy.complex_))
|
|
|
|
|
|
|
|
# now read it from file:
|
|
|
|
for i in xrange(lmax):
|
|
|
|
for j in xrange(lmax):
|
|
|
|
T[icrsh][i,j] = R.next()
|
|
|
|
for i in xrange(lmax):
|
|
|
|
for j in xrange(lmax):
|
|
|
|
T[icrsh][i,j] += 1j * R.next()
|
|
|
|
|
|
|
|
|
|
|
|
# Spin blocks to be read:
|
|
|
|
n_spin_blocs = SP + 1 - SO
|
|
|
|
|
|
|
|
|
|
|
|
# read the list of n_orbitals for all k points
|
|
|
|
n_orbitals = numpy.zeros([n_k,n_spin_blocs],numpy.int)
|
|
|
|
for isp in range(n_spin_blocs):
|
|
|
|
for ik in xrange(n_k):
|
|
|
|
n_orbitals[ik,isp] = int(R.next())
|
|
|
|
|
|
|
|
|
|
|
|
# Initialise the projectors:
|
|
|
|
proj_mat = numpy.zeros([n_k,n_spin_blocs,n_corr_shells,max(numpy.array(corr_shells)[:,3]),max(n_orbitals)],numpy.complex_)
|
2014-09-22 19:24:33 +02:00
|
|
|
|
2013-07-23 19:49:42 +02:00
|
|
|
|
|
|
|
# Read the projectors from the file:
|
|
|
|
for ik in xrange(n_k):
|
|
|
|
for icrsh in range(n_corr_shells):
|
|
|
|
no = corr_shells[icrsh][3]
|
|
|
|
# first Real part for BOTH spins, due to conventions in dmftproj:
|
|
|
|
for isp in range(n_spin_blocs):
|
|
|
|
for i in xrange(no):
|
|
|
|
for j in xrange(n_orbitals[ik][isp]):
|
|
|
|
proj_mat[ik,isp,icrsh,i,j] = R.next()
|
|
|
|
# now Imag part:
|
|
|
|
for isp in range(n_spin_blocs):
|
|
|
|
for i in xrange(no):
|
|
|
|
for j in xrange(n_orbitals[ik][isp]):
|
|
|
|
proj_mat[ik,isp,icrsh,i,j] += 1j * R.next()
|
2014-09-22 19:24:33 +02:00
|
|
|
|
2013-07-23 19:49:42 +02:00
|
|
|
|
|
|
|
# now define the arrays for weights and hopping ...
|
|
|
|
bz_weights = numpy.ones([n_k],numpy.float_)/ float(n_k) # w(k_index), default normalisation
|
|
|
|
hopping = numpy.zeros([n_k,n_spin_blocs,max(n_orbitals),max(n_orbitals)],numpy.complex_)
|
2014-09-22 19:24:33 +02:00
|
|
|
|
2013-07-23 19:49:42 +02:00
|
|
|
# weights in the file
|
|
|
|
for ik in xrange(n_k) : bz_weights[ik] = R.next()
|
|
|
|
|
|
|
|
# if the sum over spins is in the weights, take it out again!!
|
|
|
|
sm = sum(bz_weights)
|
|
|
|
bz_weights[:] /= sm
|
2014-09-22 19:24:33 +02:00
|
|
|
|
2013-07-23 19:49:42 +02:00
|
|
|
# Grab the H
|
|
|
|
# we use now the convention of a DIAGONAL Hamiltonian!!!!
|
|
|
|
for isp in range(n_spin_blocs):
|
|
|
|
for ik in xrange(n_k) :
|
2014-09-22 19:24:33 +02:00
|
|
|
no = n_orbitals[ik,isp]
|
2013-07-23 19:49:42 +02:00
|
|
|
for i in xrange(no):
|
|
|
|
hopping[ik,isp,i,i] = R.next() * energy_unit
|
|
|
|
|
2014-09-22 19:24:33 +02:00
|
|
|
# keep some things that we need for reading parproj:
|
2013-07-23 19:49:42 +02:00
|
|
|
self.n_shells = n_shells
|
|
|
|
self.shells = shells
|
|
|
|
self.n_corr_shells = n_corr_shells
|
|
|
|
self.corr_shells = corr_shells
|
|
|
|
self.n_spin_blocs = n_spin_blocs
|
|
|
|
self.n_orbitals = n_orbitals
|
|
|
|
self.n_k = n_k
|
|
|
|
self.SO = SO
|
|
|
|
self.SP = SP
|
|
|
|
self.energy_unit = energy_unit
|
|
|
|
except StopIteration : # a more explicit error if the file is corrupted.
|
2014-09-22 19:24:33 +02:00
|
|
|
raise "Wien2k_converter : reading file lda_file failed!"
|
2013-07-23 19:49:42 +02:00
|
|
|
|
|
|
|
R.close()
|
|
|
|
|
|
|
|
#-----------------------------------------
|
|
|
|
# Store the input into HDF5:
|
|
|
|
ar = HDFArchive(self.hdf_file,'a')
|
|
|
|
if not (self.lda_subgrp in ar): ar.create_group(self.lda_subgrp)
|
|
|
|
# The subgroup containing the data. If it does not exist, it is created.
|
|
|
|
# If it exists, the data is overwritten!!!
|
|
|
|
|
|
|
|
ar[self.lda_subgrp]['energy_unit'] = energy_unit
|
|
|
|
ar[self.lda_subgrp]['n_k'] = n_k
|
|
|
|
ar[self.lda_subgrp]['k_dep_projection'] = k_dep_projection
|
|
|
|
ar[self.lda_subgrp]['SP'] = SP
|
|
|
|
ar[self.lda_subgrp]['SO'] = SO
|
|
|
|
ar[self.lda_subgrp]['charge_below'] = charge_below
|
|
|
|
ar[self.lda_subgrp]['density_required'] = density_required
|
|
|
|
ar[self.lda_subgrp]['symm_op'] = symm_op
|
|
|
|
ar[self.lda_subgrp]['n_shells'] = n_shells
|
|
|
|
ar[self.lda_subgrp]['shells'] = shells
|
|
|
|
ar[self.lda_subgrp]['n_corr_shells'] = n_corr_shells
|
|
|
|
ar[self.lda_subgrp]['corr_shells'] = corr_shells
|
|
|
|
ar[self.lda_subgrp]['use_rotations'] = use_rotations
|
|
|
|
ar[self.lda_subgrp]['rot_mat'] = rot_mat
|
|
|
|
ar[self.lda_subgrp]['rot_mat_time_inv'] = rot_mat_time_inv
|
|
|
|
ar[self.lda_subgrp]['n_reps'] = n_reps
|
|
|
|
ar[self.lda_subgrp]['dim_reps'] = dim_reps
|
|
|
|
ar[self.lda_subgrp]['T'] = T
|
|
|
|
ar[self.lda_subgrp]['n_orbitals'] = n_orbitals
|
|
|
|
ar[self.lda_subgrp]['proj_mat'] = proj_mat
|
|
|
|
ar[self.lda_subgrp]['bz_weights'] = bz_weights
|
|
|
|
ar[self.lda_subgrp]['hopping'] = hopping
|
|
|
|
|
|
|
|
del ar
|
|
|
|
|
|
|
|
# Symmetries are used,
|
|
|
|
# Now do the symmetries for correlated orbitals:
|
|
|
|
self.read_symmetry_input(orbits=corr_shells,symm_file=self.symm_file,symm_subgrp=self.symm_subgrp,SO=SO,SP=SP)
|
|
|
|
|
|
|
|
|
|
|
|
def convert_parproj_input(self, par_proj_subgrp='SumK_LDA_ParProj', symm_par_subgrp='SymmPar'):
|
|
|
|
"""
|
|
|
|
Reads the input for the partial charges projectors from case.parproj, and stores it in the symm_par_subgrp
|
|
|
|
group in the HDF5.
|
|
|
|
"""
|
|
|
|
|
|
|
|
if not (mpi.is_master_node()): return
|
|
|
|
|
|
|
|
self.par_proj_subgrp = par_proj_subgrp
|
|
|
|
self.symm_par_subgrp = symm_par_subgrp
|
|
|
|
|
|
|
|
mpi.report("Reading parproj input from %s..."%self.parproj_file)
|
|
|
|
|
|
|
|
dens_mat_below = [ [numpy.zeros([self.shells[ish][3],self.shells[ish][3]],numpy.complex_) for ish in range(self.n_shells)]
|
|
|
|
for isp in range(self.n_spin_blocs) ]
|
|
|
|
|
|
|
|
R = read_fortran_file(self.parproj_file)
|
|
|
|
|
|
|
|
n_parproj = [int(R.next()) for i in range(self.n_shells)]
|
|
|
|
n_parproj = numpy.array(n_parproj)
|
|
|
|
|
|
|
|
# Initialise P, here a double list of matrices:
|
|
|
|
proj_mat_pc = numpy.zeros([self.n_k,self.n_spin_blocs,self.n_shells,max(n_parproj),max(numpy.array(self.shells)[:,3]),max(self.n_orbitals)],numpy.complex_)
|
|
|
|
|
|
|
|
rot_mat_all = [numpy.identity(self.shells[ish][3],numpy.complex_) for ish in xrange(self.n_shells)]
|
|
|
|
rot_mat_all_time_inv = [0 for i in range(self.n_shells)]
|
|
|
|
|
|
|
|
for ish in range(self.n_shells):
|
|
|
|
# read first the projectors for this orbital:
|
|
|
|
for ik in xrange(self.n_k):
|
|
|
|
for ir in range(n_parproj[ish]):
|
|
|
|
for isp in range(self.n_spin_blocs):
|
|
|
|
|
|
|
|
for i in xrange(self.shells[ish][3]): # read real part:
|
|
|
|
for j in xrange(self.n_orbitals[ik][isp]):
|
|
|
|
proj_mat_pc[ik,isp,ish,ir,i,j] = R.next()
|
|
|
|
|
|
|
|
for isp in range(self.n_spin_blocs):
|
|
|
|
for i in xrange(self.shells[ish][3]): # read imaginary part:
|
|
|
|
for j in xrange(self.n_orbitals[ik][isp]):
|
|
|
|
proj_mat_pc[ik,isp,ish,ir,i,j] += 1j * R.next()
|
|
|
|
|
|
|
|
|
|
|
|
# now read the Density Matrix for this orbital below the energy window:
|
|
|
|
for isp in range(self.n_spin_blocs):
|
|
|
|
for i in xrange(self.shells[ish][3]): # read real part:
|
|
|
|
for j in xrange(self.shells[ish][3]):
|
|
|
|
dens_mat_below[isp][ish][i,j] = R.next()
|
|
|
|
for isp in range(self.n_spin_blocs):
|
|
|
|
for i in xrange(self.shells[ish][3]): # read imaginary part:
|
|
|
|
for j in xrange(self.shells[ish][3]):
|
|
|
|
dens_mat_below[isp][ish][i,j] += 1j * R.next()
|
|
|
|
if (self.SP==0): dens_mat_below[isp][ish] /= 2.0
|
|
|
|
|
|
|
|
# Global -> local rotation matrix for this shell:
|
|
|
|
for i in xrange(self.shells[ish][3]): # read real part:
|
|
|
|
for j in xrange(self.shells[ish][3]):
|
|
|
|
rot_mat_all[ish][i,j] = R.next()
|
|
|
|
for i in xrange(self.shells[ish][3]): # read imaginary part:
|
|
|
|
for j in xrange(self.shells[ish][3]):
|
|
|
|
rot_mat_all[ish][i,j] += 1j * R.next()
|
|
|
|
|
|
|
|
#print Dens_Mat_below[0][ish],Dens_Mat_below[1][ish]
|
|
|
|
|
|
|
|
if (self.SP):
|
|
|
|
rot_mat_all_time_inv[ish] = int(R.next())
|
|
|
|
|
|
|
|
R.close()
|
|
|
|
|
|
|
|
#-----------------------------------------
|
|
|
|
# Store the input into HDF5:
|
|
|
|
ar = HDFArchive(self.hdf_file,'a')
|
|
|
|
if not (self.par_proj_subgrp in ar): ar.create_group(self.par_proj_subgrp)
|
|
|
|
# The subgroup containing the data. If it does not exist, it is created.
|
|
|
|
# If it exists, the data is overwritten!!!
|
|
|
|
thingstowrite = ['dens_mat_below','n_parproj','proj_mat_pc','rot_mat_all','rot_mat_all_time_inv']
|
|
|
|
for it in thingstowrite: exec "ar['%s']['%s'] = %s"%(self.par_proj_subgrp,it,it)
|
|
|
|
del ar
|
|
|
|
|
|
|
|
# Symmetries are used,
|
|
|
|
# Now do the symmetries for all orbitals:
|
|
|
|
self.read_symmetry_input(orbits=self.shells,symm_file=self.symmpar_file,symm_subgrp=self.symm_par_subgrp,SO=self.SO,SP=self.SP)
|
|
|
|
|
|
|
|
|
|
|
|
def convert_bands_input(self, bands_subgrp = 'SumK_LDA_Bands'):
|
|
|
|
"""
|
|
|
|
Converts the input for momentum resolved spectral functions, and stores it in bands_subgrp in the
|
|
|
|
HDF5.
|
|
|
|
"""
|
|
|
|
|
|
|
|
if not (mpi.is_master_node()): return
|
|
|
|
|
|
|
|
self.bands_subgrp = bands_subgrp
|
|
|
|
mpi.report("Reading bands input from %s..."%self.band_file)
|
|
|
|
|
|
|
|
R = read_fortran_file(self.band_file)
|
|
|
|
try:
|
|
|
|
n_k = int(R.next())
|
|
|
|
|
|
|
|
# read the list of n_orbitals for all k points
|
|
|
|
n_orbitals = numpy.zeros([n_k,self.n_spin_blocs],numpy.int)
|
|
|
|
for isp in range(self.n_spin_blocs):
|
|
|
|
for ik in xrange(n_k):
|
|
|
|
n_orbitals[ik,isp] = int(R.next())
|
|
|
|
|
|
|
|
# Initialise the projectors:
|
|
|
|
proj_mat = numpy.zeros([n_k,self.n_spin_blocs,self.n_corr_shells,max(numpy.array(self.corr_shells)[:,3]),max(n_orbitals)],numpy.complex_)
|
|
|
|
|
|
|
|
# Read the projectors from the file:
|
|
|
|
for ik in xrange(n_k):
|
|
|
|
for icrsh in range(self.n_corr_shells):
|
|
|
|
no = self.corr_shells[icrsh][3]
|
|
|
|
# first Real part for BOTH spins, due to conventions in dmftproj:
|
|
|
|
for isp in range(self.n_spin_blocs):
|
|
|
|
for i in xrange(no):
|
|
|
|
for j in xrange(n_orbitals[ik,isp]):
|
|
|
|
proj_mat[ik,isp,icrsh,i,j] = R.next()
|
|
|
|
# now Imag part:
|
|
|
|
for isp in range(self.n_spin_blocs):
|
|
|
|
for i in xrange(no):
|
|
|
|
for j in xrange(n_orbitals[ik,isp]):
|
|
|
|
proj_mat[ik,isp,icrsh,i,j] += 1j * R.next()
|
|
|
|
|
|
|
|
hopping = numpy.zeros([n_k,self.n_spin_blocs,max(n_orbitals),max(n_orbitals)],numpy.complex_)
|
|
|
|
|
|
|
|
# Grab the H
|
|
|
|
# we use now the convention of a DIAGONAL Hamiltonian!!!!
|
|
|
|
for isp in range(self.n_spin_blocs):
|
|
|
|
for ik in xrange(n_k) :
|
|
|
|
no = n_orbitals[ik,isp]
|
|
|
|
for i in xrange(no):
|
|
|
|
hopping[ik,isp,i,i] = R.next() * self.energy_unit
|
|
|
|
|
|
|
|
# now read the partial projectors:
|
|
|
|
n_parproj = [int(R.next()) for i in range(self.n_shells)]
|
|
|
|
n_parproj = numpy.array(n_parproj)
|
|
|
|
|
|
|
|
# Initialise P, here a double list of matrices:
|
|
|
|
proj_mat_pc = numpy.zeros([n_k,self.n_spin_blocs,self.n_shells,max(n_parproj),max(numpy.array(self.shells)[:,3]),max(n_orbitals)],numpy.complex_)
|
|
|
|
|
|
|
|
|
|
|
|
for ish in range(self.n_shells):
|
|
|
|
|
|
|
|
for ik in xrange(n_k):
|
|
|
|
for ir in range(n_parproj[ish]):
|
|
|
|
for isp in range(self.n_spin_blocs):
|
|
|
|
|
|
|
|
for i in xrange(self.shells[ish][3]): # read real part:
|
|
|
|
for j in xrange(n_orbitals[ik,isp]):
|
|
|
|
proj_mat_pc[ik,isp,ish,ir,i,j] = R.next()
|
|
|
|
|
|
|
|
for i in xrange(self.shells[ish][3]): # read imaginary part:
|
|
|
|
for j in xrange(n_orbitals[ik,isp]):
|
|
|
|
proj_mat_pc[ik,isp,ish,ir,i,j] += 1j * R.next()
|
|
|
|
|
|
|
|
except StopIteration : # a more explicit error if the file is corrupted.
|
2014-09-22 19:24:33 +02:00
|
|
|
raise "Wien2k_converter : reading file band_file failed!"
|
2013-07-23 19:49:42 +02:00
|
|
|
|
|
|
|
R.close()
|
|
|
|
# reading done!
|
|
|
|
|
|
|
|
#-----------------------------------------
|
|
|
|
# Store the input into HDF5:
|
|
|
|
ar = HDFArchive(self.hdf_file,'a')
|
|
|
|
if not (self.bands_subgrp in ar): ar.create_group(self.bands_subgrp)
|
2014-09-22 19:24:33 +02:00
|
|
|
|
2013-07-23 19:49:42 +02:00
|
|
|
# The subgroup containing the data. If it does not exist, it is created.
|
|
|
|
# If it exists, the data is overwritten!!!
|
|
|
|
thingstowrite = ['n_k','n_orbitals','proj_mat','hopping','n_parproj','proj_mat_pc']
|
|
|
|
for it in thingstowrite: exec "ar['%s']['%s'] = %s"%(self.bands_subgrp,it,it)
|
|
|
|
del ar
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def read_symmetry_input(self, orbits, symm_file, symm_subgrp, SO, SP):
|
|
|
|
"""
|
|
|
|
Reads input for the symmetrisations from symm_file, which is case.sympar or case.symqmc.
|
|
|
|
"""
|
|
|
|
|
|
|
|
if not (mpi.is_master_node()): return
|
|
|
|
|
|
|
|
mpi.report("Reading symmetry input from %s..."%symm_file)
|
|
|
|
|
|
|
|
n_orbits = len(orbits)
|
|
|
|
R=read_fortran_file(symm_file)
|
|
|
|
|
|
|
|
try:
|
|
|
|
n_s = int(R.next()) # Number of symmetry operations
|
|
|
|
n_atoms = int(R.next()) # number of atoms involved
|
|
|
|
perm = [ [int(R.next()) for i in xrange(n_atoms)] for j in xrange(n_s) ] # list of permutations of the atoms
|
|
|
|
if SP:
|
|
|
|
time_inv = [ int(R.next()) for j in xrange(n_s) ] # timeinversion for SO xoupling
|
|
|
|
else:
|
|
|
|
time_inv = [ 0 for j in xrange(n_s) ]
|
|
|
|
|
|
|
|
# Now read matrices:
|
|
|
|
mat = []
|
|
|
|
for in_s in xrange(n_s):
|
|
|
|
|
|
|
|
mat.append( [ numpy.zeros([orbits[orb][3], orbits[orb][3]],numpy.complex_) for orb in xrange(n_orbits) ] )
|
|
|
|
for orb in range(n_orbits):
|
|
|
|
for i in xrange(orbits[orb][3]):
|
|
|
|
for j in xrange(orbits[orb][3]):
|
|
|
|
mat[in_s][orb][i,j] = R.next() # real part
|
|
|
|
for i in xrange(orbits[orb][3]):
|
|
|
|
for j in xrange(orbits[orb][3]):
|
|
|
|
mat[in_s][orb][i,j] += 1j * R.next() # imaginary part
|
|
|
|
|
|
|
|
# determine the inequivalent shells:
|
2014-09-22 19:24:33 +02:00
|
|
|
#SHOULD BE FINALLY REMOVED, PUT IT FOR ALL ORBITALS!!!!! (PS: FIXME?)
|
2013-07-23 19:49:42 +02:00
|
|
|
#self.inequiv_shells(orbits)
|
|
|
|
mat_tinv = [numpy.identity(orbits[orb][3],numpy.complex_)
|
|
|
|
for orb in range(n_orbits)]
|
|
|
|
|
|
|
|
if ((SO==0) and (SP==0)):
|
|
|
|
# here we need an additional time inversion operation, so read it:
|
|
|
|
for orb in range(n_orbits):
|
|
|
|
for i in xrange(orbits[orb][3]):
|
|
|
|
for j in xrange(orbits[orb][3]):
|
|
|
|
mat_tinv[orb][i,j] = R.next() # real part
|
|
|
|
for i in xrange(orbits[orb][3]):
|
|
|
|
for j in xrange(orbits[orb][3]):
|
|
|
|
mat_tinv[orb][i,j] += 1j * R.next() # imaginary part
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
except StopIteration : # a more explicit error if the file is corrupted.
|
2014-09-22 19:24:33 +02:00
|
|
|
raise "Wien2k_converter : reading file symm_file failed!"
|
2013-07-23 19:49:42 +02:00
|
|
|
|
|
|
|
R.close()
|
|
|
|
|
|
|
|
# Save it to the HDF:
|
|
|
|
ar=HDFArchive(self.hdf_file,'a')
|
|
|
|
if not (symm_subgrp in ar): ar.create_group(symm_subgrp)
|
|
|
|
thingstowrite = ['n_s','n_atoms','perm','orbits','SO','SP','time_inv','mat','mat_tinv']
|
|
|
|
for it in thingstowrite: exec "ar['%s']['%s'] = %s"%(symm_subgrp,it,it)
|
|
|
|
del ar
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def __repack(self):
|
|
|
|
"""Calls the h5repack routine, in order to reduce the file size of the hdf5 archive.
|
|
|
|
Should only be used BEFORE the first invokation of HDFArchive in the program, otherwise
|
|
|
|
the hdf5 linking is broken!!!"""
|
|
|
|
|
|
|
|
import subprocess
|
|
|
|
|
|
|
|
if not (mpi.is_master_node()): return
|
|
|
|
|
|
|
|
mpi.report("Repacking the file %s"%self.hdf_file)
|
|
|
|
|
|
|
|
retcode = subprocess.call(["h5repack","-i%s"%self.hdf_file, "-otemphgfrt.h5"])
|
|
|
|
if (retcode!=0):
|
|
|
|
mpi.report("h5repack failed!")
|
|
|
|
else:
|
|
|
|
subprocess.call(["mv","-f","temphgfrt.h5","%s"%self.hdf_file])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def inequiv_shells(self,lst):
|
|
|
|
"""
|
|
|
|
The number of inequivalent shells is calculated from lst, and a mapping is given as
|
|
|
|
map(i_corr_shells) = i_inequiv_corr_shells
|
|
|
|
invmap(i_inequiv_corr_shells) = i_corr_shells
|
|
|
|
in order to put the Self energies to all equivalent shells, and for extracting Gloc
|
|
|
|
"""
|
|
|
|
|
|
|
|
tmp = []
|
|
|
|
self.shellmap = [0 for i in range(len(lst))]
|
|
|
|
self.invshellmap = [0]
|
|
|
|
self.n_inequiv_corr_shells = 1
|
|
|
|
tmp.append( lst[0][1:3] )
|
|
|
|
|
|
|
|
if (len(lst)>1):
|
|
|
|
for i in range(len(lst)-1):
|
|
|
|
|
|
|
|
fnd = False
|
|
|
|
for j in range(self.n_inequiv_corr_shells):
|
|
|
|
if (tmp[j]==lst[i+1][1:3]):
|
|
|
|
fnd = True
|
|
|
|
self.shellmap[i+1] = j
|
|
|
|
if (fnd==False):
|
|
|
|
self.shellmap[i+1] = self.n_inequiv_corr_shells
|
|
|
|
self.n_inequiv_corr_shells += 1
|
|
|
|
tmp.append( lst[i+1][1:3] )
|
|
|
|
self.invshellmap.append(i+1)
|
|
|
|
|