def _compute(self):
nsteps = len(self._pos)
# Setup k vectors and tabulate rho
k_sorted, k_selected = self._decimate_k()
kmax = max(self.kvec.keys()) + self.dk
cnt = [0 for k in k_sorted]
# Note: actually rho_av is not calculated because it is negligible
rho_av = [complex(0.,0.) for k in k_sorted]
rho2_av = [complex(0.,0.) for k in k_sorted]
cell_variable = is_cell_variable(self.trajectory)
for i in range(0, nsteps, self.skip):
# If cell changes we have to update
if cell_variable:
self._setup(i)
k_sorted, k_selected = self._decimate_k()
kmax = max(self.kvec.keys()) + self.dk
expo = expo_sphere(self.k0, kmax, self._pos[i])
for kk, knorm in enumerate(k_sorted):
for k in k_selected[kk]:
ik = self.kvec[knorm][k]
Ri = self._radius[i]
mk = 4 * numpy.pi / knorm**3 * (numpy.sin(knorm*Ri) - (knorm*Ri) * numpy.cos(knorm*Ri))
rho = numpy.sum(mk*expo[...,0,ik[0]]*expo[...,1,ik[1]]*expo[...,2,ik[2]])
rho2_av[kk] += (rho * rho.conjugate())
cnt[kk] += 1
# Normalization.
volume = numpy.average([s.cell.volume for s in self.trajectory])
self.grid = k_sorted
self.value = [(rho2_av[kk] / cnt[kk] - rho_av[kk]*rho_av[kk].conjugate() / cnt[kk]**2).real / volume
for kk in range(len(self.grid))]
self.value_nonorm = [rho2_av[kk].real / cnt[kk]
for kk in range(len(self.grid))]
def test_cell_variable(self):
finp = '/tmp/test_utils.xyz'
# with open(finp, 'w') as fh:
# fh.write("""\
# 1
# step:0 cell:1.0,1.0,1.0
# A 1.0 -1.0 0.0
# 1
# step:1 cell:1.0,1.0,1.0
# A 1.0 -1.0 0.0
# 1
# step:2 cell:1.0,1.0,1.0
# A 1.0 -1.0 0.0
# 1
# step:3 cell:1.0,1.0,1.0
# A 1.0 -1.0 0.0
# """)
# t = Trajectory(finp)
# self.assertFalse(utils.is_cell_variable(t, tests=-1))
with open(finp, 'w') as fh:
fh.write("""\
1
step:0 cell:1.0,1.0,1.0
A 1.0 -1.0 0.0
1
step:1 cell:2.0,1.0,1.0
A 1.0 -1.0 0.0
1
step:2 cell:1.0,1.0,1.0
A 1.0 -1.0 0.0
1
step:3 cell:1.0,1.0,1.0
A 1.0 -1.0 0.0
""")
with TrajectoryXYZ(finp) as th:
self.assertTrue(utils.is_cell_variable(th, tests=-1))
self.assertFalse(utils.is_cell_variable(th, tests=1))
self.assertTrue(utils.is_cell_variable(th, tests=2))
def _compute(self):
from atooms.trajectory.utils import is_cell_variable
nsteps = len(self._pos_0)
# Setup k vectors and tabulate rho
k_sorted, k_selected = self._decimate_k()
kmax = max(self.kvec.keys()) + self.dk
cnt = [0 for k in k_sorted]
rho_av = [complex(0., 0.) for k in k_sorted]
rho2_av = [complex(0., 0.) for k in k_sorted]
variable_cell = is_cell_variable(self.trajectory)
for i in range(0, nsteps, self.skip):
# If cell changes we have to update the wave vectors
if variable_cell:
self._setup(i)
k_sorted, k_selected = self._decimate_k()
kmax = max(self.kvec.keys()) + self.dk
# Tabulate exponentials
# Note: tabulating and computing takes about the same time
if self._pos_0[i] is self._pos_1[i]:
# Identical species
expo_0 = expo_sphere(self.k0, kmax, self._pos_0[i])
expo_1 = expo_0
else:
# Cross correlation
expo_0 = expo_sphere(self.k0, kmax, self._pos_0[i])
expo_1 = expo_sphere(self.k0, kmax, self._pos_1[i])
for kk, knorm in enumerate(k_sorted):
for k in k_selected[kk]:
ik = self.kvec[knorm][k]
# In the absence of a microscopic field, rho_av = (0, 0)
if not self._field:
if expo_0 is expo_1:
# Identical species
rho_0 = numpy.sum(expo_0[..., 0, ik[0]] *
expo_0[..., 1, ik[1]] *
expo_0[..., 2, ik[2]])
rho_1 = rho_0
else:
# Cross correlation
rho_0 = numpy.sum(expo_0[..., 0, ik[0]] *
expo_0[..., 1, ik[1]] *
expo_0[..., 2, ik[2]])
rho_1 = numpy.sum(expo_1[..., 0, ik[0]] *
expo_1[..., 1, ik[1]] *
expo_1[..., 2, ik[2]])
else:
# We have a field as a weight
rho_0 = numpy.sum(
self._field[i] * expo_0[..., 0, ik[0]] *
expo_0[..., 1, ik[1]] * expo_0[..., 2, ik[2]])
rho_1 = rho_0
rho_av[kk] += rho_0
rho2_av[kk] += (rho_0 * rho_1.conjugate())
cnt[kk] += 1
# Normalization.
npart_0 = sum([p.shape[0]
for p in self._pos_0]) / float(len(self._pos_0))
npart_1 = sum([p.shape[0]
for p in self._pos_1]) / float(len(self._pos_1))
self.grid = k_sorted
self.value, self.value_nonorm = [], []
for kk in range(len(self.grid)):
norm = float(npart_0 * npart_1)**0.5
value = (rho2_av[kk] / cnt[kk] -
rho_av[kk] * rho_av[kk].conjugate() / cnt[kk]**2).real
self.value.append(value / norm)
self.value_nonorm.append(value)
def _compute(self):
from atooms.trajectory.utils import is_cell_variable
nsteps = len(self._pos_0)
# Setup k vectors and tabulate rho
k_sorted, k_selected = self._decimate_k()
kmax = max(self.kvec.keys()) + self.dk
cnt = [0 for k in k_sorted]
rho_av = [complex(0., 0.) for k in k_sorted]
rho2_av = [complex(0., 0.) for k in k_sorted]
variable_cell = is_cell_variable(self.trajectory)
for i in range(0, nsteps, self.skip):
# If cell changes we have to update the wave vectors
if variable_cell:
self._setup(i)
k_sorted, k_selected = self._decimate_k()
kmax = max(self.kvec.keys()) + self.dk
# Tabulate exponentials
# Note: tabulating and computing takes about the same time
if self._pos_0[i] is self._pos_1[i]:
# Identical species
expo_0 = expo_sphere(self.k0, kmax, self._pos_0[i])
expo_1 = expo_0
else:
# Cross correlation
expo_0 = expo_sphere(self.k0, kmax, self._pos_0[i])
expo_1 = expo_sphere(self.k0, kmax, self._pos_1[i])
import atooms.postprocessing.fourierspace_wrap
from atooms.postprocessing.fourierspace_wrap import fourierspace_module
for kk, knorm in enumerate(k_sorted):
ikvec = numpy.ndarray((3, len(k_selected[kk])),
order='F',
dtype=numpy.int32)
i = 0
for k in k_selected[kk]:
ikvec[:, i] = self.kvec[knorm][k]
i += 1
rho = numpy.zeros(ikvec.shape[1], dtype=numpy.complex128)
fourierspace_module.sk_bare(expo_0, ikvec, rho)
rho_0 = rho
rho_1 = rho
rho2_av[kk] += numpy.sum(rho_0 * rho_1.conjugate())
cnt[kk] += rho.shape[0]
# for kk, knorm in enumerate(k_sorted):
# for k in k_selected[kk]:
# ik = self.kvec[knorm][k]
# # In the absence of a microscopic field, rho_av = (0, 0)
# if not self._field:
# if expo_0 is expo_1:
# # Identical species
# rho_0 = numpy.sum(expo_0[...,0,ik[0]] *
# expo_0[...,1,ik[1]] *
# expo_0[...,2,ik[2]])
# rho_1 = rho_0
# else:
# # Cross correlation
# rho_0 = numpy.sum(expo_0[...,0,ik[0]] *
# expo_0[...,1,ik[1]] *
# expo_0[...,2,ik[2]])
# rho_1 = numpy.sum(expo_1[...,0,ik[0]] *
# expo_1[...,1,ik[1]] *
# expo_1[...,2,ik[2]])
# else:
# # We have a field as a weight
# rho_0 = numpy.sum(self._field[i] *
# expo_0[...,0,ik[0]] *
# expo_0[...,1,ik[1]] *
# expo_0[...,2,ik[2]])
# rho_1 = rho_0
# rho_av[kk] += rho_0
# rho2_av[kk] += (rho_0 * rho_1.conjugate())
# cnt[kk] += 1
# Normalization.
npart_0 = sum([p.shape[0]
for p in self._pos_0]) / float(len(self._pos_0))
npart_1 = sum([p.shape[0]
for p in self._pos_1]) / float(len(self._pos_1))
self.grid = k_sorted
self.value, self.value_nonorm = [], []
for kk in range(len(self.grid)):
norm = float(npart_0 * npart_1)**0.5
value = (rho2_av[kk] / cnt[kk] -
rho_av[kk] * rho_av[kk].conjugate() / cnt[kk]**2).real
self.value.append(value / norm)
self.value_nonorm.append(value)
def _compute(self):
nsteps = len(self._pos_0)
ndims = len(self.k0)
# Setup k vectors and tabulate rho
kgrid, selection = self.kgrid, self.selection
kmax = max(self.kvector.keys()) + self.dk
cnt = [0 for k in kgrid]
rho_0_av = [complex(0., 0.) for k in kgrid]
rho_1_av = [complex(0., 0.) for k in kgrid]
rho2_av = [complex(0., 0.) for k in kgrid]
variable_cell = is_cell_variable(self.trajectory)
for i in progress(range(0, nsteps, self.skip),
total=nsteps // self.skip):
# If cell changes we have to update the wave vectors
if variable_cell:
self._setup(i)
kgrid, selection = self._decimate_k()
kmax = max(self.kvector.keys()) + self.dk
# Tabulate exponentials
# Note: tabulating and computing takes about the same time
if self._pos_0[i] is self._pos_1[i]:
# Identical species
expo_0 = expo_sphere(self.k0, kmax, self._pos_0[i])
expo_1 = expo_0
else:
# Cross correlation
expo_0 = expo_sphere(self.k0, kmax, self._pos_0[i])
expo_1 = expo_sphere(self.k0, kmax, self._pos_1[i])
# Define weights
if self._weight is None:
weight_0, weight_1 = 1.0, 1.0
else:
weight_0, weight_1 = self._weight_0[i], self._weight_1[i]
# Nice spaghetti here
for kk, knorm in enumerate(kgrid):
for k in selection[kk]:
ik = self.kvector[knorm][k]
if expo_0 is expo_1:
# Identical species
if ndims == 3:
rho_0 = numpy.sum(
weight_0 * expo_0[..., 0, ik[0]] *
expo_0[..., 1, ik[1]] * expo_0[..., 2, ik[2]])
elif ndims == 2:
rho_0 = numpy.sum(weight_0 *
expo_0[..., 0, ik[0]] *
expo_0[..., 1, ik[1]])
else:
tmp = weight_0 * expo_0[..., 0, ik[0]]
for idim in range(1, ndims):
tmp *= expo_0[..., idim, ik[idim]]
rho_0 = numpy.sum(tmp)
rho_1 = rho_0
else:
# Cross correlation
if ndims == 3:
rho_0 = numpy.sum(
weight_0 * expo_0[..., 0, ik[0]] *
expo_0[..., 1, ik[1]] * expo_0[..., 2, ik[2]])
rho_1 = numpy.sum(
weight_1 * expo_1[..., 0, ik[0]] *
expo_1[..., 1, ik[1]] * expo_1[..., 2, ik[2]])
elif ndims == 2:
rho_0 = numpy.sum(weight_0 *
expo_0[..., 0, ik[0]] *
expo_0[..., 1, ik[1]])
rho_1 = numpy.sum(weight_1 *
expo_1[..., 0, ik[0]] *
expo_1[..., 1, ik[1]])
else:
tmp = weight_0 * expo_0[..., 0, ik[0]]
for idim in range(ndims):
tmp *= expo_0[..., idim, ik[idim]]
rho_0 = numpy.sum(tmp)
tmp = weight_0 * expo_1[..., 0, ik[0]]
for idim in range(ndims):
tmp *= expo_1[..., idim, ik[idim]]
rho_1 = numpy.sum(tmp)
# Cumulate averages
rho_0_av[kk] += rho_0
rho_1_av[kk] += rho_1
rho2_av[kk] += (rho_0 * rho_1.conjugate())
cnt[kk] += 1
# In the absence of a microscopic field, the average density is zero.
# We get rid of the average and compute <rho(k)rho*(k)>.
if self._weight_0 is None:
rho_0_av = [complex(0., 0.) for k in kgrid]
rho_1_av = [complex(0., 0.) for k in kgrid]
# Normalization
npart_0 = sum([p.shape[0]
for p in self._pos_0]) / float(len(self._pos_0))
npart_1 = sum([p.shape[0]
for p in self._pos_1]) / float(len(self._pos_1))
self.grid = kgrid
self.value, self.value_nonorm = [], []
for kk in range(len(self.grid)):
norm = float(npart_0 * npart_1)**0.5
value = (rho2_av[kk] / cnt[kk] -
rho_0_av[kk] * rho_1_av[kk].conjugate() / cnt[kk]**2).real
self.value.append(value / norm)
self.value_nonorm.append(value)
def _compute(self):
from atooms.trajectory.utils import is_cell_variable
try:
from atooms.postprocessing.fourierspace_wrap import fourierspace_module
except ImportError:
_log.error('f90 wrapper missing or not functioning')
raise
nsteps = len(self._pos_0)
# Setup k vectors and tabulate rho
kgrid, selection = self.kgrid, self.selection
kmax = max(self.kvector.keys()) + self.dk
cnt = [0 for k in kgrid]
rho_av = [complex(0., 0.) for k in kgrid]
rho2_av = [complex(0., 0.) for k in kgrid]
variable_cell = is_cell_variable(self.trajectory)
for i in range(0, nsteps, self.skip):
# If cell changes we have to update the wave vectors
if variable_cell:
self._setup(i)
kgrid, selection = self._decimate_k()
kmax = max(self.kvector.keys()) + self.dk
# Tabulate exponentials
# Note: tabulating and computing takes about the same time
if self._pos_0[i] is self._pos_1[i]:
# Identical species
expo_0 = expo_sphere(self.k0, kmax, self._pos_0[i])
expo_1 = expo_0
else:
# Cross correlation
# TODO: cross correlation wont work
expo_0 = expo_sphere(self.k0, kmax, self._pos_0[i])
expo_1 = expo_sphere(self.k0, kmax, self._pos_1[i])
for kk, knorm in enumerate(kgrid):
ikvec = numpy.ndarray((3, len(selection[kk])),
order='F',
dtype=numpy.int32)
i = 0
for k in selection[kk]:
ikvec[:, i] = self.kvector[knorm][k]
i += 1
rho = numpy.zeros(ikvec.shape[1], dtype=numpy.complex128)
fourierspace_module.sk_bare(expo_0, ikvec, rho)
rho_0 = rho
rho_1 = rho
rho2_av[kk] += numpy.sum(rho_0 * rho_1.conjugate())
cnt[kk] += rho.shape[0]
# Normalization.
npart_0 = sum([p.shape[0]
for p in self._pos_0]) / float(len(self._pos_0))
npart_1 = sum([p.shape[0]
for p in self._pos_1]) / float(len(self._pos_1))
self.grid = kgrid
self.value, self.value_nonorm = [], []
for kk in range(len(self.grid)):
norm = float(npart_0 * npart_1)**0.5
value = (rho2_av[kk] / cnt[kk] -
rho_av[kk] * rho_av[kk].conjugate() / cnt[kk]**2).real
self.value.append(value / norm)
self.value_nonorm.append(value)