mirror of
https://github.com/triqs/dft_tools
synced 2024-11-18 12:03:50 +01:00
688 lines
26 KiB
Python
688 lines
26 KiB
Python
|
|
##########################################################################
|
|
#
|
|
# TRIQS: a Toolbox for Research in Interacting Quantum Systems
|
|
#
|
|
# Copyright (C) 2019 by A. D. N. James, M. Zingl and M. Aichhorn
|
|
#
|
|
# 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 triqs_dft_tools.converters.converter_tools import *
|
|
import os.path
|
|
from locale import atof
|
|
|
|
|
|
class readElkfiles:
|
|
"""
|
|
Read in the Elk output files for the elk_converter.py routines.
|
|
"""
|
|
|
|
def __init__(self):
|
|
self.dft_file = 'PROJ.OUT'
|
|
self.band_file = 'BAND.OUT'
|
|
self.eval_file = 'EIGVAL.OUT'
|
|
self.efermi_file = 'EFERMI.OUT'
|
|
self.kp_file = 'KPOINTS.OUT'
|
|
self.geom_file='GEOMETRY.OUT'
|
|
|
|
def read_elk_file(self, filename, to_replace):
|
|
#this function is almost identical to read_fortran_file, but it removes words after ':'
|
|
"""
|
|
Returns a generator that yields all numbers in the Fortran file as float, with possible replacements.
|
|
|
|
Parameters
|
|
----------
|
|
filename : string
|
|
Name of Fortran-produced file.
|
|
to_replace : dict of str:str
|
|
Dictionary defining old_char:new_char.
|
|
|
|
Yields
|
|
------
|
|
string
|
|
The next number in file.
|
|
|
|
"""
|
|
#import os.path
|
|
#import string
|
|
import locale
|
|
from locale import atof
|
|
if not(os.path.exists(filename)):
|
|
raise IOError("File %s does not exist." % filename)
|
|
for line in open(filename, 'r'):
|
|
for old, new in to_replace.items():
|
|
line = line.replace(old, new)
|
|
# removes the substring after ':' character - if it exists
|
|
if ":" in line:
|
|
line = self.split_string(line)
|
|
if "(" in line:
|
|
line = self.split_string3(line)
|
|
for x in line.split():
|
|
yield atof(x)
|
|
|
|
def read_elk_file2(self, filename, to_replace):
|
|
#this function is almost identical to read_fortran_file, but it removes words after '('
|
|
"""
|
|
Returns a generator that yields all numbers in the Fortran file as float, with possible replacements.
|
|
|
|
Parameters
|
|
----------
|
|
filename : string
|
|
Name of Fortran-produced file.
|
|
to_replace : dict of str:str
|
|
Dictionary defining old_char:new_char.
|
|
|
|
Yields
|
|
------
|
|
string
|
|
The next number in file.
|
|
|
|
"""
|
|
import os.path
|
|
import string
|
|
if not(os.path.exists(filename)):
|
|
raise IOError("File %s does not exist." % filename)
|
|
for line in open(filename, 'r'):
|
|
for old, new in to_replace.items():
|
|
line = line.replace(old, new)
|
|
# removes the substring after '(' character - if it exists
|
|
if "(" in line:
|
|
line = self.split_string2(line,'(')
|
|
# removes the substring after '+' character - if it exists
|
|
if "+" in line:
|
|
line = self.split_string2(line,'+')
|
|
# removes the substring after '|' character - if it exists
|
|
if "|" in line:
|
|
line = self.split_string2(line,'|')
|
|
#only include lines which have multiple characters
|
|
if(len(line)>1):
|
|
yield line.split()
|
|
|
|
def split_string(self,line):
|
|
"""
|
|
This removes the excess information after ':' of the string read in from the file
|
|
"""
|
|
temp = line.split(':')
|
|
line = temp[0]
|
|
return line
|
|
|
|
def split_string2(self,line,str):
|
|
"""
|
|
This removes the excess information after 'str' of the string read in from the file
|
|
"""
|
|
temp = line.split(str)
|
|
line = temp[0]
|
|
return line
|
|
|
|
def split_string3(self,line):
|
|
"""
|
|
This removes the excess information after '(' of the string read in from the file
|
|
"""
|
|
temp = line.split('(')
|
|
line = temp[0]
|
|
return line
|
|
|
|
|
|
def read_proj(self,dft_file):#,things_to_set):
|
|
"""
|
|
This function reads the contents of PROJ.OUT and returns them for general use.
|
|
"""
|
|
#read in PROJ.OUT file
|
|
R = self.read_elk_file( dft_file, self.fortran_to_replace)
|
|
try:
|
|
#arrays of entry names
|
|
gen_info_entries = ['nproj', 'n_k', 'spinpol', 'SO','natm']
|
|
proj_entries = ['sort', 'natom', 'l', 'dim']
|
|
at_entries = ['spatom','atom']
|
|
#Read in the No. of projectors, no. of k-points, spin and spinorb info
|
|
gen_info = {name: int(val) for name, val in zip(gen_info_entries, R)}
|
|
#mpi.report(gen_info)
|
|
#read in the information of all the projectors
|
|
#initialized variables
|
|
icorr=0
|
|
lmax=3 #maximum l value that will be used
|
|
n_shells=0
|
|
n_inequiv_shells=0
|
|
#local arrays
|
|
neqatom=[]#numpy.zeros([n_shells], int)
|
|
proj=[]
|
|
shells=[]#numpy.zeros([n_shells], int)
|
|
corr_shells=[]#numpy.zeros([n_shells], int)
|
|
prjtype=[]
|
|
wan=[]
|
|
proj_info=[]#numpy.zeros([n_shells], int)
|
|
T=[]
|
|
basis=[]
|
|
inequiv_to_corr=[]
|
|
corr_to_inequiv=[]
|
|
ind=[]
|
|
proj_idx=[]
|
|
at=[]
|
|
idxlm=[]
|
|
#assign global variables
|
|
n_k=gen_info['n_k']
|
|
SO=gen_info['SO']
|
|
#loop over projector flags
|
|
for ip in range(gen_info['nproj']):
|
|
#projector index
|
|
proj_idx.append(int(next(R)))
|
|
proj.append({name: int(val) for name, val in zip(proj_entries, R)})
|
|
na=proj[ip]['natom'] # integer which reduces when reading in atom info
|
|
while(na>0):
|
|
neqatom.append(int(next(R)))
|
|
na-=neqatom[n_inequiv_shells]
|
|
#appends the mapping array to index of inequivalent atom in shells
|
|
inequiv_to_corr.append(icorr)
|
|
for ia in range(neqatom[n_inequiv_shells]):
|
|
corr_to_inequiv.append(n_inequiv_shells)
|
|
at.append({name: int(val) for name, val in zip(at_entries, R)})
|
|
shells.append(proj[ip].copy())
|
|
shells[n_shells].update({'atom':at[n_shells]['atom'], 'spatom':at[n_shells]['spatom']})
|
|
corr_shells.append(shells[n_shells].copy())
|
|
n_orb=2*shells[n_shells]['l']+1
|
|
#lm submatrix indices
|
|
idxlm.append(numpy.zeros(2*lmax+1, dtype=int))
|
|
nrep=proj[ip]['dim']
|
|
for i in range(nrep):
|
|
idxlm[n_shells][i]=next(R)-1
|
|
ind.append(idxlm[n_shells][0:nrep])
|
|
#determine basis type and transformation matrix
|
|
basis.append(int(next(R)))
|
|
#determine whether which basis the projectors where generated in
|
|
#spherical harmonics
|
|
T.append(numpy.zeros([n_orb, n_orb], dtype=complex))
|
|
#Elk generated unitary basis
|
|
if (basis[n_shells]==2):
|
|
#reads the transformation matrix
|
|
for i in range(n_orb):
|
|
for j in range(n_orb):
|
|
T[n_shells][i, j] = next(R)
|
|
for i in range(n_orb):
|
|
for j in range(n_orb):
|
|
T[n_shells][i, j] += 1j * next(R)
|
|
#transformation matrix from spherical harmonics to cubic basis
|
|
elif (basis[n_shells]==1):
|
|
T[n_shells]=self.determine_T(shells[n_shells]['l'])
|
|
else:
|
|
for i in range(n_orb):
|
|
T[n_shells][i,i] = 1.0
|
|
#index for the next inequivalent atom (+1 to index is not needed as this is incorporated in
|
|
#neqatom[ish]
|
|
n_shells+=1
|
|
#increase the inequiv_to_corr value
|
|
icorr+=neqatom[n_inequiv_shells]
|
|
#increase the numer of inequivalent atoms
|
|
n_inequiv_shells+=1
|
|
#end reading the file if read the last projector
|
|
if (ip+1==gen_info['nproj']):
|
|
break
|
|
#determine the irreducible representation
|
|
dim_reps=[]
|
|
n_reps=[]
|
|
irep=[]
|
|
for ish in range(n_inequiv_shells):
|
|
isheq=inequiv_to_corr[ish]
|
|
[n_reps,dim_reps,irep] = self.determine_rep(isheq,ish,corr_shells,basis,ind,n_reps,dim_reps,irep)
|
|
for ish in range(n_shells):
|
|
ishin=corr_to_inequiv[ish]
|
|
corr_shells[ish].update({'SO':SO, 'irep':irep[ishin]})
|
|
|
|
except StopIteration: # a more explicit error if the file is corrupted.
|
|
#raise "Elk_converter : reading PROJ.OUT file failed!"
|
|
raise IOError("Elk_converter : reading PROJ.OUT file failed!")
|
|
|
|
R.close()
|
|
#output desired information
|
|
return (gen_info,n_shells,n_inequiv_shells,corr_to_inequiv,inequiv_to_corr,corr_shells,n_reps,dim_reps,ind,basis,T)
|
|
|
|
def determine_T(self,l):
|
|
"""
|
|
Current version calculates the transformation matrix T to convert the inputs from spherical
|
|
harmonics to the cubic basis (as used in TRIQS and Wien2k).
|
|
This routine is currently very similar to spherical_to_cubic in the TRIQS library.
|
|
This routine can be extended to include other unitary rotation matrices
|
|
"""
|
|
from math import sqrt
|
|
size=2*l+1
|
|
T = numpy.zeros((size,size),dtype=complex)
|
|
if(l == 0):
|
|
T[0,0]= 1.0
|
|
elif(l == 1):
|
|
T[0,0] = 1.0/sqrt(2); T[0,2] = -1.0/sqrt(2)
|
|
T[1,0] = 1j/sqrt(2); T[1,2] = 1j/sqrt(2)
|
|
T[2,1] = 1.0
|
|
elif l == 2:
|
|
T[0,2] = 1.0
|
|
T[1,0] = 1.0/sqrt(2); T[1,4] = 1.0/sqrt(2)
|
|
T[2,0] =-1.0/sqrt(2); T[2,4] = 1.0/sqrt(2)
|
|
T[3,1] = 1.0/sqrt(2); T[3,3] =-1.0/sqrt(2)
|
|
T[4,1] = 1.0/sqrt(2); T[4,3] = 1.0/sqrt(2)
|
|
elif l == 3:
|
|
#This needs to be checked
|
|
T[0,0] = 1.0/sqrt(2); T[0,6] = -1.0/sqrt(2)
|
|
T[1,1] = 1.0/sqrt(2); T[1,5] = 1.0/sqrt(2)
|
|
T[2,2] = 1.0/sqrt(2); T[2,4] = -1.0/sqrt(2)
|
|
T[3,3] = 1.0
|
|
T[4,2] = 1j/sqrt(2); T[4,4] = 1j/sqrt(2)
|
|
T[5,1] = 1j/sqrt(2); T[5,5] = -1j/sqrt(2)
|
|
T[6,0] = 1j/sqrt(2); T[6,6] = 1j/sqrt(2)
|
|
else: raise ValueError("determine: implemented only for l=0,1,2,3")
|
|
return numpy.matrix(T)
|
|
|
|
|
|
def determine_rep(self,ish,ishin,corr_shells,basis,ind,n_reps,dim_reps,irep):
|
|
"""
|
|
Determines the irreducible representation used for projection calculation.
|
|
Only for Cubic harmonics at the moment
|
|
"""
|
|
#if all of the orbitals were used to construct the projectors
|
|
rep=corr_shells[ish]['dim']
|
|
if(rep==2*corr_shells[ish]['l']+1):
|
|
n_reps.append(1)
|
|
dim_reps.append([corr_shells[ish]['dim']])
|
|
irep.append(1)
|
|
#if a subset of obitals were used for generating the projectors (l>1)
|
|
#if Elk generated T matrix is used
|
|
elif((basis[ish]==2)&(corr_shells[ish]['l']>1)):
|
|
if (corr_shells[ish]['l']==2):
|
|
n_reps.append(2)
|
|
dim_reps.append([3, 2])
|
|
irep.append(dim_reps[ishin].index(rep)+1)
|
|
elif (corr_shells[ish]['l']==3):
|
|
n_reps.append(3)
|
|
dim_reps.append([1, 3, 3])
|
|
#cubic T matrix
|
|
elif((basis[ish]==1)&(corr_shells[ish]['l']>1)):
|
|
if (corr_shells[ish]['l']==2):
|
|
n_reps.append(2)
|
|
dim_reps.append([2, 3])
|
|
irep.append(dim_reps[ishin].index(rep)+1)
|
|
elif (corr_shells[ish]['l']==3):
|
|
n_reps.append(3)
|
|
dim_reps.append([2, 2, 3])
|
|
#determine the dim_reps from the lm indices in ind
|
|
if(rep==3):
|
|
irep.append(3)
|
|
else:
|
|
for i in range(2):
|
|
#look for ind[i]==0
|
|
if(ind[i]==0):
|
|
irep.append(1)
|
|
break
|
|
#default to irep=2
|
|
elif(i==1):
|
|
irep.append(2)
|
|
|
|
else:
|
|
raise ValueError("Elk_converter (determine_rep) : irreducible representations were not found!")
|
|
|
|
return (n_reps,dim_reps,irep)
|
|
|
|
def read_projector(self,shell,n_spin_blocks,ish,proj_mat,ind,T,basis,filext):
|
|
"""
|
|
This function reads the contents of WANPROJ_L**_S**_A****.OUT
|
|
"""
|
|
import string
|
|
l = str(shell[ish]['l']).zfill(2)
|
|
s = str(shell[ish]['sort']).zfill(2)
|
|
a = str(shell[ish]['spatom']).zfill(4)
|
|
#construct file name from atoms info
|
|
wan_shells_file = 'WANPROJ_L'+l+'_S'+s+'_A'+a+filext
|
|
R = self.read_elk_file(wan_shells_file, self.fortran_to_replace)
|
|
try:
|
|
#no. of kpts and number of orbital
|
|
gen_entries = ['n_k', 'lmmax','dim_rep']
|
|
gen = {name: int(val) for name, val in zip(gen_entries, R)}
|
|
#projector lm size
|
|
dim=gen['lmmax']
|
|
#no. of lm used to construct projector
|
|
dim_rep=gen['dim_rep']
|
|
lat=[]
|
|
n_k=gen['n_k']
|
|
n_orbitals = numpy.zeros([n_k, n_spin_blocks], int)
|
|
band_window = [None for isp in range(n_spin_blocks)]
|
|
for isp in range(n_spin_blocks):
|
|
band_window[isp] = numpy.zeros([n_k, 2], dtype=int)
|
|
for ik in range(0,n_k):
|
|
#k-point index and correlated band window indices
|
|
lat_entries = ['ik','vklx','vkly','vklz']
|
|
lat.append({name: int(val) for name, val in zip(lat_entries, R)})
|
|
#loop over each spin
|
|
for isp in range(0,n_spin_blocks):
|
|
#band window indices
|
|
proj_entries = ['isp', 'ist_min','ist_max']
|
|
proj_dim={name: int(val) for name, val in zip(proj_entries, R)}
|
|
if(isp+1!=proj_dim['isp']):
|
|
raise IOError("Elk_converter (",wan_shells_file,") : reading spin projecotrs failed!")
|
|
return
|
|
#setting band window arrays
|
|
n_orbitals[ik,isp]=proj_dim['ist_max']-proj_dim['ist_min']+1
|
|
band_window[isp][ik, 0] = proj_dim['ist_min']
|
|
band_window[isp][ik, 1] = proj_dim['ist_max']
|
|
#define temporary matrix for reading in the projectors
|
|
mat = numpy.zeros([dim, n_orbitals[ik,isp]], complex)
|
|
# Real part
|
|
for j in range(dim):
|
|
for i in range(n_orbitals[ik,isp]):
|
|
mat[j, i] = next(R)
|
|
# Imag part:
|
|
for j in range(dim):
|
|
for i in range(n_orbitals[ik,isp]):
|
|
mat[j, i] += 1j * next(R)
|
|
#rotate projectors into basis T
|
|
if(basis[ish]!=0):
|
|
mat[:,:]=numpy.matmul(T[ish],mat[:,:])
|
|
|
|
#put desired projector subset into master array
|
|
proj_mat[ik,isp,ish,0:dim_rep,0:n_orbitals[ik,isp]]=mat[ind[ish],0:n_orbitals[ik,isp]]
|
|
#delete temporary index
|
|
del mat
|
|
#resize array length of band indices to maximum number used
|
|
proj_mat=proj_mat[:,:,:,:,:numpy.max(n_orbitals)]
|
|
|
|
except StopIteration: # a more explicit error if the file is corrupted.
|
|
raise IOError("Elk_converter (read projectors): reading file failed!")
|
|
R.close()
|
|
return(n_orbitals,band_window,dim_rep,proj_mat)
|
|
|
|
def read_eig(self,filext=".OUT"):
|
|
"""
|
|
This function reads the contents of EIGVAL.OUT and EFERMI.OUT
|
|
"""
|
|
import string
|
|
#construct file name from atoms info
|
|
R = self.read_elk_file( self.efermi_file, self.fortran_to_replace)
|
|
try:
|
|
efermi=next(R)
|
|
except StopIteration: # a more explicit error if the file is corrupted.
|
|
raise IOError("Elk_converter (EFERMI.OUT): reading file failed!")
|
|
eval_file = 'EIGVAL'+filext
|
|
R = self.read_elk_file( eval_file, self.fortran_to_replace)
|
|
try:
|
|
#no. of kpts
|
|
n_k=int(next(R))
|
|
#no. second variation states (no. bands)
|
|
nstsv=int(next(R))
|
|
en=[]
|
|
occ=[]
|
|
kp2=[]
|
|
for ik in range(0,n_k):
|
|
#k index and corresponing lattice coordinates
|
|
k_entries = ['ik', 'vklx','vkly','vklz']
|
|
kp2.append([{name: val for name, val in zip(k_entries, R)}])
|
|
en.append([])
|
|
occ.append([])
|
|
for ist in range(0,nstsv):
|
|
#stores the band index, energy eigenvalues and occupancies
|
|
en_entries = ['ist', 'hopping','occ']
|
|
read_en={name: val for name, val in zip(en_entries, R)}
|
|
#Energy eigenvalues and occupations
|
|
en[ik].append(read_en['hopping']-efermi)
|
|
occ[ik].append(read_en['occ'])
|
|
|
|
except StopIteration: # a more explicit error if the file is corrupted.
|
|
raise IOError("Elk_converter (EIGVAL.OUT): reading file failed!")
|
|
R.close()
|
|
return(en,occ,nstsv)
|
|
|
|
def read_kpoints(self,filext=".OUT"):
|
|
"""
|
|
This function reads the contents of KPOINTS.OUT
|
|
"""
|
|
import string
|
|
kp_file = 'KPOINTS'+filext
|
|
R = self.read_elk_file( kp_file, self.fortran_to_replace)
|
|
try:
|
|
#no. of kpts
|
|
n_k=int(next(R))
|
|
kp=[]
|
|
#reads in the k index, lattice vectors, weights and nmat for each kpt
|
|
#array for bz weights
|
|
bz_weights = numpy.ones([n_k], float) / float(n_k)
|
|
#array for lattice vectors
|
|
vkl = numpy.ones([n_k,3], float)
|
|
for ik in range(n_k):
|
|
#k-grid info
|
|
k_entries = ['ik', 'vklx','vkly','vklz', 'bz_weights', 'nmat']
|
|
kp.append({name: val for name, val in zip(k_entries, R)})
|
|
#convert to integers
|
|
kp[ik]['ik']=int(kp[ik]['ik'])
|
|
#read in BZ weights
|
|
bz_weights[ik]=kp[ik]['bz_weights']
|
|
j=0
|
|
for i in range(1,4):
|
|
vkl[ik,j]=kp[ik][k_entries[i]]
|
|
j=j+1
|
|
except StopIteration: # a more explicit error if the file is corrupted.
|
|
raise IOError("Elk_converter (KPOINTS.OUT): reading file failed!")
|
|
R.close()
|
|
return(bz_weights,vkl)
|
|
|
|
def readsym(self):
|
|
"""
|
|
Read in the (crystal) symmetries in lattice coordinates
|
|
"""
|
|
dft_file='SYMCRYS.OUT'
|
|
R = self.read_elk_file2( dft_file, self.fortran_to_replace)
|
|
try:
|
|
symmat=[]
|
|
spinmat=[]
|
|
tr=[]
|
|
#read the number of crystal symmetries
|
|
x = next(R)
|
|
nsym = int(atof(x[0]))
|
|
#set up symmetry matrices
|
|
for isym in range(nsym):
|
|
symmat.append(numpy.zeros([3, 3], float))
|
|
spinmat.append(numpy.zeros([3, 3], float))
|
|
tr.append(numpy.zeros([3], float))
|
|
#read each symmetry
|
|
for isym in range(nsym):
|
|
#read the symmetry index and check it
|
|
x = next(R)
|
|
isymm = int(x[3])
|
|
if(isym+1!=isymm):
|
|
raise IOError("Elk_converter : reading symmetries failed!")
|
|
#next read has no useful infomation
|
|
x = next(R)
|
|
#read in the translation vector used for symmetries
|
|
x = next(R)
|
|
for i in range(3):
|
|
tr[isym][i]=atof(x[i])
|
|
#read in the spatial symmetries
|
|
#string with no useful information
|
|
x = next(R)
|
|
#read in the spatial symmetry
|
|
for i in range(3):
|
|
x = next(R)
|
|
for j in range(3):
|
|
symmat[isym][i,j]=int(atof(x[j]))
|
|
#string with no useful information
|
|
x = next(R)
|
|
#read in the spin symmetries
|
|
for i in range(3):
|
|
x = next(R)
|
|
for j in range(3):
|
|
spinmat[isym][i,j]=int(atof(x[j]))
|
|
|
|
except StopIteration: # a more explicit error if the file is corrupted.
|
|
raise IOError("Elk_converter : reading SYMCRYS.OUT file failed!")
|
|
R.close()
|
|
return nsym,spinmat,symmat,tr
|
|
|
|
def readlat(self):
|
|
"""
|
|
Read in information about the lattice.
|
|
"""
|
|
dft_file='LATTICE.OUT'
|
|
R = self.read_elk_file2( dft_file, self.fortran_to_replace)
|
|
try:
|
|
amat = numpy.zeros([3, 3], float)
|
|
amatinv = numpy.zeros([3, 3], float)
|
|
bmat = numpy.zeros([3, 3], float)
|
|
bmatinv = numpy.zeros([3, 3], float)
|
|
#real space lattice matrices
|
|
#cycling through information which is not needed
|
|
for i in range(4):
|
|
x = next(R)
|
|
#reading in the lattice vectors as matrix
|
|
for i in range(3):
|
|
x = next(R)
|
|
for j in range(3):
|
|
amat[i,j] = atof(x[j])
|
|
#cycling through information which is not needed
|
|
x = next(R)
|
|
#reading in the inverse lattice matrix
|
|
for i in range(3):
|
|
x = next(R)
|
|
for j in range(3):
|
|
amatinv[i,j] = atof(x[j])
|
|
#read in cell volume (for transport)
|
|
x = next(R)
|
|
cell_vol = atof(x[-1])
|
|
#cycling through information which is not needed
|
|
for i in range(4):
|
|
x = next(R)
|
|
#reading in the reciprocal lattice vectors as matrix
|
|
for i in range(3):
|
|
x = next(R)
|
|
for j in range(3):
|
|
bmat[i,j] = atof(x[j])
|
|
#cycling through information which is not needed
|
|
x = next(R)
|
|
#reading in the inverse reciprocal lattice matrix
|
|
for i in range(3):
|
|
x = next(R)
|
|
for j in range(3):
|
|
bmatinv[i,j] = atof(x[j])
|
|
|
|
except StopIteration: # a more explicit error if the file is corrupted.
|
|
raise IOError("Elk_converter : reading PROJ.OUT file failed!")
|
|
R.close()
|
|
return amat,amatinv,bmat,bmatinv,cell_vol
|
|
|
|
def read_geometry(self):
|
|
"""
|
|
This function reads the contents of GEOMETRY.OUT
|
|
"""
|
|
import string
|
|
#read in the Geometery file
|
|
dft_file='GEOMETRY.OUT'
|
|
#atom positions in lattice coordinates
|
|
R = self.read_elk_file2( dft_file, self.fortran_to_replace)
|
|
nspecies=0
|
|
na=[]
|
|
atpos_entries = ['kx', 'ky','kz','Bkx', 'Bky', 'Bkz']
|
|
atpos=[]
|
|
try:
|
|
i=0
|
|
#cycle through file until "atoms" is found
|
|
while(i<1000):
|
|
x = next(R)
|
|
if "atoms" in x:
|
|
break
|
|
elif(i==1000):
|
|
raise IOError("Elk_converter : Could not find 'atoms' in GEOMETRY.OUT!")
|
|
i+=1
|
|
#read in the number of species
|
|
x = next(R)
|
|
ns = int(atof(x[0]))
|
|
#loop over species
|
|
for js in range(ns):
|
|
#Skip species name
|
|
x = next(R)
|
|
#read in the number of atoms per species
|
|
x = next(R)
|
|
na.append(int(atof(x[0])))
|
|
#loop over atomss pre species
|
|
atpos.append([])
|
|
for ia in range(na[js]):
|
|
atpos[js].append(numpy.zeros(6, float))
|
|
x = next(R)
|
|
for j in range(6):
|
|
atpos[js][ia][j]=atof(x[j])
|
|
except StopIteration: # a more explicit error if the file is corrupted.
|
|
raise IOError("Elk_converter : Could not open GEOMETRY.OUT!")
|
|
R.close()
|
|
return ns, na, atpos
|
|
|
|
#band character dependent calculations
|
|
def read_bc(self):
|
|
"""
|
|
Read in the ELK generated band characters from BC.OUT
|
|
"""
|
|
|
|
#import string
|
|
file = 'BC.OUT'
|
|
R = self.read_elk_file(file, self.fortran_to_replace)
|
|
try:
|
|
#no. of kpts and number of orbital
|
|
gen_entries = ['maxlm', 'nspinor','natmtot','nstsv','nkpt','irep']
|
|
gen = {name: int(val) for name, val in zip(gen_entries, R)}
|
|
#projector lm size
|
|
#check the read in information complies with previous read in data
|
|
nspinor=self.SP+1
|
|
if(gen['nspinor'] != nspinor):
|
|
mpi.report('HDF file nspinor = %s'%nspinor)
|
|
mpi.report('BC.OUT nspinor = %s'%gen['nspinor'])
|
|
raise IOError("Elk_converter (",file,") : reading nspinor failed!")
|
|
return
|
|
if(gen['natmtot'] != self.n_atoms):
|
|
raise IOError("Elk_converter (",file,") : reading no. of atoms failed!")
|
|
return
|
|
if(gen['nstsv'] != self.nstsv):
|
|
raise IOError("Elk_converter (",file,") : reading all states failed!")
|
|
return
|
|
if(gen['nkpt'] != self.n_k):
|
|
raise IOError("Elk_converter (",file,") : reading kpoints failed failed!")
|
|
return
|
|
if(gen['irep'] == 0):
|
|
raise IOError("Elk_converter (",file,") : Band characters are in spherical hamonics, may have issues with the PDOS!")
|
|
return
|
|
|
|
dim=gen['maxlm']
|
|
lmax=numpy.sqrt(dim)-1
|
|
bc = numpy.zeros([dim,nspinor,self.n_atoms,self.nstsv,self.n_k], float)
|
|
|
|
for ik in range(0,self.n_k):
|
|
for iatom in range(0,self.n_atoms):
|
|
for ispn in range(0,nspinor):
|
|
entry = ['ispn','ias','is','ia','ik']
|
|
ent = {name: int(val) for name, val in zip(entry, R)}
|
|
#k-point index and correlated band window indices
|
|
#check read in values
|
|
if(ent['ispn'] != ispn+1):
|
|
raise IOError("Elk_converter (",file,") : reading ispn failed!")
|
|
return
|
|
if(ent['ias'] != iatom+1):
|
|
raise IOError("Elk_converter (",file,") : reading iatom failed!")
|
|
return
|
|
if(ent['ik'] != ik+1):
|
|
raise IOError("Elk_converter (",file,") : reading ik failed!")
|
|
return
|
|
|
|
for ist in range(self.nstsv):
|
|
for lm in range(dim):
|
|
bc[lm,ispn,iatom,ist,ik] = next(R)
|
|
|
|
except StopIteration: # a more explicit error if the file is corrupted.
|
|
raise IOError("Elk_converter (read BC.OUT): reading file failed!")
|
|
R.close()
|
|
return(bc,dim)
|
|
|