def parse_scf_conv(scf_out):
from qharv.reel import ascii_out
mm = ascii_out.read(scf_out)
idxl = ascii_out.all_lines_with_tag(mm, 'iteration #')
data = []
for idx in idxl:
mm.seek(idx)
# read iteration number
iternow = ascii_out.name_sep_val(mm, 'iteration', sep='#', dtype=int)
# find total energy and other info (!!!! must be in order)
try:
time = ascii_out.name_sep_val(mm,
'cpu time spent up to now',
sep='is')
enow = ascii_out.name_sep_val(mm, 'total energy')
except:
continue
entry = {'istep': iternow, 'energy': enow, 'time': time}
data.append(entry)
return data
def get_dsk_amat(floc):
""" extract A matrix from qmcfinitesize output
k->0 behavior of 3D structure factor S(k) is fitted to a Gaussian
S(k) = k^T A k
Args:
floc (str): location of qmcfinitesize output
Returns:
np.array: A matrix (3x3)
"""
mm = ascii_out.read(floc)
amat = np.zeros([3,3])
# step 1: fill upper triangular part of amat
xyzm = {'x':0,'y':1,'z':2} # map x,y,z to index
keyl = ['a_xx','a_yy','a_zz','a_xy','a_xz','a_yz']
for key in keyl: # order of key matters!
val = ascii_out.name_sep_val(mm,key)
xyz_xyz = key.split('_')[-1]
idx = tuple([xyzm[xyz] for xyz in xyz_xyz])
amat[idx] = val
# end for
# step 2: symmetrize amat
amat[(1,0)] = amat[(0,1)]
amat[(2,1)] = amat[(1,2)]
amat[(2,0)] = amat[(0,2)]
return amat
def read_kpoints(scf_out):
from qharv.reel import ascii_out
mm = ascii_out.read(scf_out)
# get lattice units
alat = ascii_out.name_sep_val(mm, 'lattice parameter (alat)')
blat = 2 * np.pi / alat
# start parsing k points
idx = mm.find(b'number of k points')
mm.seek(idx)
# read first line
# e.g. number of k points= 32 Fermi-Dirac smearing ...
line = mm.readline().decode()
nk = int(line.split('=')[1].split()[0])
# confirm units in second line
line = mm.readline().decode()
assert '2pi/alat' in line
# start parsing kvectors
data = np.zeros([nk, 4]) # ik, kx, ky, kz, wk
for ik in range(nk):
line = mm.readline().decode()
kvec, wk = parse_kline(line, ik=ik)
data[ik, :3] = kvec * blat
data[ik, 3] = wk
return data
def get_elem_pos(nscf_in):
from qharv.reel import ascii_out
mm = ascii_out.read(nscf_in)
natom = ascii_out.name_sep_val(mm, 'nat', '=', dtype=int)
idx = mm.find(b'ATOMIC_POSITIONS')
mm.seek(idx)
header = mm.readline().decode()
eleml = []
posl = []
for iatom in range(natom):
line = mm.readline()
tokens = line.split()
eleml.append(tokens[0])
posl.append(tokens[1:])
mm.close()
elem = np.array(eleml, dtype=str)
pos = np.array(posl, dtype=float)
return elem, pos, header
def parse_output(floc):
""" get energy, volume and pressure from QE output """
etot = read_first_energy(floc)
entry = {'energy': etot / 2.} # Ry to ha
mm = read(floc)
label_map = {'volume': 'unit-cell volume', 'natom': 'number of atoms/cell'}
for key in label_map.keys():
val = name_sep_val(mm, label_map[key])
entry[key] = val
# end for
au_stressl, kbar_stressl = read_stress(floc)
assert len(au_stressl) == 1
au_stress = au_stressl[0]
entry['pressure'] = np.diag(au_stress).mean() / 2. # Ry to ha
entry['stress'] = au_stress / 2. # Ry to ha
return entry
def read_fsc_out(fout, chi2_tol=1e-12, rtol=0.01):
from qharv.reel import ascii_out
mm = ascii_out.read(fout)
# check chi^2
idxl = ascii_out.all_lines_with_tag(mm, 'fitpn: Chi^2 =')
for idx in idxl:
mm.seek(idx)
chi2 = ascii_out.name_sep_val(mm, 'fitpn: Chi^2', '=')
assert chi2 < chi2_tol
# rewind
mm.seek(0)
# get nelec
nelec = ascii_out.name_sep_val(mm, 'nparts', '=', dtype=int)
# read potential
idx = mm.find(b'Potential energy correction')
mm.seek(idx)
idx = mm.find(b'using optimized potential')
mm.seek(idx)
dv = ascii_out.name_sep_val(mm, 'dV/N', '=')
# check potential
idx = mm.find(b'interpolating optimized potential')
mm.seek(idx)
dv0 = ascii_out.name_sep_val(mm, 'dV/N', '=')
assert dv > 0
assert dv0 > 0
rdv = abs(dv - dv0) / dv0
assert rdv < rtol
# read kinetic
mm = ascii_out.read(fout)
idx = mm.find(b'Kinetic energy correction')
mm.seek(idx)
idx = mm.find(b'using optimized potential')
mm.seek(idx)
dt = ascii_out.name_sep_val(mm, 'dT/N', '=')
# check kinetic
idx = mm.find(b'interpolating optimized potential')
mm.seek(idx)
dt0 = ascii_out.name_sep_val(mm, 'dT/N', '=')
assert dt > 0
assert dt0 > 0
rdt = abs(dt - dt0) / dt0
assert rdt < rtol
mm.close()
return dv * nelec, dt * nelec
def get_volume(fout): mm = ascii_out.read(fout) omega = ascii_out.name_sep_val(mm, 'Vol', pos=1) return omega
def get_efermi(fout):
from qharv.reel import ascii_out
mm = ascii_out.read(fout)
efermi = ascii_out.name_sep_val(mm, 'the Fermi energy', sep='is')
return efermi
def parse_nscf_bands(nscf_out, span=7, trailer='occupation numbers'):
data = {} # build a dictionary as return value
def scanf_7f(line, n):
""" implement scanf("%7.*f") """
numl = []
for i in range(n):
token = line[span * i:span * (i + 1)]
num = float(token)
numl.append(num)
return numl
def parse_float_body(body):
""" parse a blob of floats """
lines = body.split('\n')
numl = []
for line in lines:
if len(line) == 0: continue
numl += map(float, line.split())
return numl
from qharv.reel import ascii_out
ndim = 3
mm = ascii_out.read(nscf_out)
alat = ascii_out.name_sep_val(mm, 'lattice parameter (alat)')
blat = 2 * np.pi / alat
# find the beginnings of each band
bhead = ' k ='
idxl = ascii_out.all_lines_with_tag(mm, bhead)
nkpt = len(idxl)
data['nkpt'] = nkpt
# estimate the end of the last band
idx1 = ascii_out.all_lines_with_tag(mm, trailer)[-1]
# trick to use no if statement in the loop
idxl = idxl + [idx1]
kvecs = [] # (nkpt, ndim)
mat = [] # (nkpt, nbnd)
for ikpt in range(nkpt):
# specify beginning and end of the band output
idx0 = idxl[ikpt]
idx1 = idxl[ikpt + 1]
# parse band output
# first read header
mm.seek(idx0)
header = mm.readline().decode()
if not 'bands (ev)' in header: continue
kxkykz = ascii_out.lr_mark(header, '=', '(')
kvec = scanf_7f(kxkykz, ndim)
kvecs.append(kvec)
# then read body
body = mm[mm.tell():idx1].decode().strip('\n')
if trailer in body:
idx2 = mm.find(trailer.encode())
body = mm[mm.tell():idx2].strip('\n')
row = parse_float_body(body)
mat.append(row)
# end for ikpt
data['kvecs'] = blat * np.array(kvecs)
data['bands'] = np.array(mat)
return data
def input_structure(scf_in, put_in_box=True):
ndim = 3 # assume 3 dimensions
with open(scf_in, 'r+') as f:
mm = mmap(f.fileno(), 0)
# end with
from qharv.reel.ascii_out import name_sep_val
ntyp = name_sep_val(mm, 'ntyp', dtype=int)
if ntyp != 1:
raise NotImplementedError('only support 1 type of atom for now')
# end if
# read lattice
mm.seek(0)
idx = mm.find(b'ibrav')
mm.seek(idx)
ibrav_line = mm.readline()
ibrav = int(ibrav_line.split('=')[-1])
if ibrav != 0:
raise NotImplementedError('only ibrav = 0 is supported')
# end if
idx = mm.find(b'CELL_PARAMETERS')
mm.seek(idx)
header = mm.readline()
unit = header.split()[-1]
axes = np.zeros([ndim, ndim])
for idim in range(ndim):
line = mm.readline()
axes[idim, :] = map(float, line.split())
# end for
cell = {'unit': unit, 'axes': axes}
# read atomic positions
mm.seek(0) # rewind
idx = mm.find(b'nat')
mm.seek(idx)
nat_line = mm.readline()
nat = int(nat_line.split('=')[-1])
idx = mm.find(b'ATOMIC_POSITIONS')
mm.seek(idx)
header = mm.readline()
unit = header.split()[-1]
pos = np.zeros([nat, ndim])
for iat in range(nat):
line = mm.readline()
pos[iat, :] = map(float, line.split()[-3:])
# end for iat
try:
line = mm.readline()
float(line.split()[-3:])
raise RuntimeError('next lines looks like positions too!\n%s' % line)
except:
pass # expect to see an empty line
# end try
if put_in_box:
atpos = {
'pos_unit': unit,
'pos': pos_in_box(np.array(pos), np.array(axes)).tolist()
}
else:
atpos = {'pos_unit': unit, 'pos': pos}
# end if
entry = {'infile': scf_in}
entry.update(cell)
entry.update(atpos)
return entry
def read_forces(scf_out, ndim=3, which='total'):
""" read the forces in a pwscf output, assume only one force block
'which' decides which block of forces to read, choices are:
['total', 'non-local', 'local', 'ionic', 'core', 'Hubbard', 'scf']
!!!! assuming QE uses Ry, will convert to Ha """
Ry = 0.5 # Ha
begin_tag_dict = {
'total': 'Forces acting on atoms',
'non-local': 'The non-local contrib. to forces',
'ionic': 'The ionic contribution to forces',
'local': 'The local contribution to forces',
'core': 'The core correction contribution to forces',
'Hubbard': 'The Hubbard contrib. to forces',
'scf': 'The SCF correction term to forces'
}
end_tag_dict = {
'total': 'The non-local contrib. to forces',
'non-local': 'The ionic contribution to forces',
'ionic': 'The local contribution to forces',
'local': 'The core correction contribution to forces',
'core': 'The Hubbard contrib. to forces',
'Hubbard': 'The SCF correction term to forces',
'scf': 'Total force ='
}
fhandle = open(scf_out, 'r+')
mm = mmap(fhandle.fileno(), 0)
natom = name_sep_val(mm, 'number of atoms', dtype=int)
# locate force block
begin_tag = begin_tag_dict[which]
end_tag = end_tag_dict[which]
begin_idx = mm.find(begin_tag.encode())
end_idx = mm.find(end_tag.encode())
if begin_idx == -1:
raise RuntimeError('cannot locate %s' % begin_tag)
elif end_idx == -1:
# maybe verbosity='low'
end_idx = mm.find(b'Total force =')
if end_idx == -1:
raise RuntimeError('cannot locate %s' % end_tag)
# end if
# end if
force_block = mm[begin_idx:end_idx]
# parse force block for forces
forces = np.zeros([natom, ndim])
iatom = 0
for line in force_block.split(b'\n'):
if line.strip().startswith(b'atom'):
tokens = line.split()
if len(tokens) == 9: # found an atom
myforce = np.array(tokens[-3:], dtype=float)
forces[iatom, :] = tokens[-3:]
iatom += 1
# end if
# end if
# end for
if iatom != natom:
raise RuntimeError('found %d forces for %d atoms' % (iatom, natom))
# end if
fhandle.close()
return forces * Ry