def matrix(self, shape=None):
"""
This function returns the object as a matrix of DFT or iDFT resp.
"""
N=self.N
prodN=np.prod(N)
if shape is not None:
dim=np.prod(np.array(shape))
elif hasattr(self, 'shape'):
dim=np.prod(np.array(shape))
else:
raise ValueError('Missing shape of the DFT.')
proddN=dim*prodN
ZN_input=Grid.get_ZNl(N, fft_form=0)
ZN_output=Grid.get_ZNl(N, fft_form='c')
if self.inverse:
DFTcoef=lambda k, l, N: np.exp(2*np.pi*1j*np.sum(k*l/N))
else:
DFTcoef=lambda k, l, N: np.exp(-2*np.pi*1j*np.sum(k*l/N))/np.prod(N)
DTM=np.zeros([self.pN(), self.pN()], dtype=np.complex128)
for ii, kk in enumerate(itertools.product(*tuple(ZN_output))):
for jj, ll in enumerate(itertools.product(*tuple(ZN_input))):
DTM[ii, jj]=DFTcoef(np.array(kk, dtype=np.float),
np.array(ll), N)
DTMd=npmatlib.zeros([proddN, proddN], dtype=np.complex128)
for ii in range(dim):
DTMd[prodN*ii:prodN*(ii+1), prodN*ii:prodN*(ii+1)]=DTM
return DTMd
def matrix(self, shape=None):
"""
This function returns the object as a matrix of DFT or iDFT resp.
"""
N = self.N
prodN = np.prod(N)
if shape is not None:
dim = np.prod(np.array(shape))
elif hasattr(self, 'shape'):
dim = np.prod(np.array(shape))
else:
raise ValueError('Missing shape of the DFT.')
proddN = dim * prodN
ZN_input = Grid.get_ZNl(N, fft_form=0)
ZN_output = Grid.get_ZNl(N, fft_form='c')
if self.inverse:
DFTcoef = lambda k, l, N: np.exp(2 * np.pi * 1j * np.sum(k * l / N)
)
else:
DFTcoef = lambda k, l, N: np.exp(-2 * np.pi * 1j * np.sum(
k * l / N)) / np.prod(N)
DTM = np.zeros([self.pN(), self.pN()], dtype=np.complex128)
for ii, kk in enumerate(itertools.product(*tuple(ZN_output))):
for jj, ll in enumerate(itertools.product(*tuple(ZN_input))):
DTM[ii, jj] = DFTcoef(np.array(kk, dtype=np.float),
np.array(ll), N)
DTMd = npmatlib.zeros([proddN, proddN], dtype=np.complex128)
for ii in range(dim):
DTMd[prodN * ii:prodN * (ii + 1),
prodN * ii:prodN * (ii + 1)] = DTM
return DTMd
def get_weights_circ(r, Nbar, Y):
"""
it evaluates integral weights,
which are used for upper-lower bounds calculation,
with constant circular inclusion
Parameters
----------
r - the parameter determining the size of inclusion
Nbar - no. of points of regular grid where the weights are evaluated
Y - the size of periodic unit cell
Returns
-------
Wphi - integral weights at regular grid sizing Nbar
"""
d = np.size(Y)
ZN2l = Grid.get_ZNl(Nbar, fft_form='c')
meas_puc = np.prod(Y)
circ = 0
for m in range(d):
Nshape = np.ones(d, dtype=np.int)
Nshape[m] = Nbar[m]
Nrep = np.copy(Nbar)
Nrep[m] = 1
xi_p2 = np.tile(np.reshape((ZN2l[m]/Y[m])**2, Nshape), Nrep)
circ += xi_p2
circ = circ**0.5
ind = mean_index(Nbar, fft_form='c')
circ[ind] = 1.
Wphi = r**2 * sp.jn(1, 2*np.pi*circ*r) / (circ*r)
Wphi[ind] = np.pi*r**2
Wphi = Wphi / meas_puc
return Wphi
def get_weights_circ(r, Nbar, Y):
"""
it evaluates integral weights,
which are used for upper-lower bounds calculation,
with constant circular inclusion
Parameters
----------
r - the parameter determining the size of inclusion
Nbar - no. of points of regular grid where the weights are evaluated
Y - the size of periodic unit cell
Returns
-------
Wphi - integral weights at regular grid sizing Nbar
"""
d = np.size(Y)
ZN2l = Grid.get_ZNl(Nbar, fft_form='c')
meas_puc = np.prod(Y)
circ = 0
for m in range(d):
Nshape = np.ones(d, dtype=np.int)
Nshape[m] = Nbar[m]
Nrep = np.copy(Nbar)
Nrep[m] = 1
xi_p2 = np.tile(np.reshape((ZN2l[m] / Y[m])**2, Nshape), Nrep)
circ += xi_p2
circ = circ**0.5
ind = mean_index(Nbar, fft_form='c')
circ[ind] = 1.
Wphi = r**2 * sp.jn(1, 2 * np.pi * circ * r) / (circ * r)
Wphi[ind] = np.pi * r**2
Wphi = Wphi / meas_puc
return Wphi
def get_weights_lin(h, Nbar, Y):
"""
it evaluates integral weights,
which are used for upper-lower bounds calculation,
with bilinear inclusion at rectangular area
Parameters
----------
h - the parameter determining the size of inclusion (half-size of support)
Nbar - no. of points of regular grid where the weights are evaluated
Y - the size of periodic unit cell
Returns
-------
Wphi - integral weights at regular grid sizing Nbar
"""
d = np.size(Y)
meas_puc = np.prod(Y)
ZN2l = Grid.get_ZNl(Nbar, fft_form='c')
Wphi = np.ones(Nbar) / meas_puc
for ii in np.arange(d):
Nshape = np.ones(d, dtype=np.int)
Nshape[ii] = Nbar[ii]
Nrep = np.copy(Nbar)
Nrep[ii] = 1
Wphi *= h[ii]*np.tile(np.reshape((np.sinc(h[ii]*ZN2l[ii]/Y[ii]))**2,
Nshape), Nrep)
return Wphi
def get_weights_lin(h, Nbar, Y):
"""
it evaluates integral weights,
which are used for upper-lower bounds calculation,
with bilinear inclusion at rectangular area
Parameters
----------
h - the parameter determining the size of inclusion (half-size of support)
Nbar - no. of points of regular grid where the weights are evaluated
Y - the size of periodic unit cell
Returns
-------
Wphi - integral weights at regular grid sizing Nbar
"""
d = np.size(Y)
meas_puc = np.prod(Y)
ZN2l = Grid.get_ZNl(Nbar, fft_form='c')
Wphi = np.ones(Nbar) / meas_puc
for ii in np.arange(d):
Nshape = np.ones(d, dtype=np.int)
Nshape[ii] = Nbar[ii]
Nrep = np.copy(Nbar)
Nrep[ii] = 1
Wphi *= h[ii] * np.tile(
np.reshape((np.sinc(h[ii] * ZN2l[ii] / Y[ii]))**2, Nshape), Nrep)
return Wphi
def get_shift_inclusion(N, h, Y):
N = np.array(N, dtype=np.int)
Y = np.array(Y, dtype=np.float)
dim = N.size
ZN = Grid.get_ZNl(N)
SS = np.ones(N, dtype=np.complex128)
for ii in np.arange(dim):
Nshape = np.ones(dim, dtype=np.int)
Nshape[ii] = N[ii]
Nrep = N
Nrep[ii] = 1
SS *= np.tile(np.reshape(np.exp(-2*np.pi*1j*(h[ii]*ZN[ii]/Y[ii])),
Nshape), Nrep)
return SS
def get_shift_inclusion(N, h, Y):
N = np.array(N, dtype=np.int)
Y = np.array(Y, dtype=np.float)
dim = N.size
ZN = Grid.get_ZNl(N)
SS = np.ones(N, dtype=np.complex128)
for ii in np.arange(dim):
Nshape = np.ones(dim, dtype=np.int)
Nshape[ii] = N[ii]
Nrep = N
Nrep[ii] = 1
SS *= np.tile(
np.reshape(np.exp(-2 * np.pi * 1j * (h[ii] * ZN[ii] / Y[ii])),
Nshape), Nrep)
return SS
def matrix(self):
"""
This function returns the object as a matrix of DFT or iDFT resp.
"""
N=self.N
prodN=np.prod(N)
proddN=self.d*prodN
ZNl=Grid.get_ZNl(N, fft_form='c')
if self.inverse:
DFTcoef=lambda k, l, N: np.exp(2*np.pi*1j*np.sum(k*l/N))
else:
DFTcoef=lambda k, l, N: np.exp(-2*np.pi*1j*np.sum(k*l/N))/np.prod(N)
DTM=np.zeros([self.pN(), self.pN()], dtype=np.complex128)
for ii, kk in enumerate(itertools.product(*tuple(ZNl))):
for jj, ll in enumerate(itertools.product(*tuple(ZNl))):
DTM[ii, jj]=DFTcoef(np.array(kk, dtype=np.float),
np.array(ll), N)
DTMd=npmatlib.zeros([proddN, proddN], dtype=np.complex128)
for ii in range(self.d):
DTMd[prodN*ii:prodN*(ii+1), prodN*ii:prodN*(ii+1)]=DTM
return DTMd
def matrix(self):
"""
This function returns the object as a matrix of DFT or iDFT resp.
"""
N=self.N
prodN=np.prod(N)
proddN=self.d*prodN
ZNl=Grid.get_ZNl(N, fft_form='c')
if self.inverse:
DFTcoef=lambda k, l, N: np.exp(2*np.pi*1j*np.sum(k*l/N))
else:
DFTcoef=lambda k, l, N: np.exp(-2*np.pi*1j*np.sum(k*l/N))/np.prod(N)
DTM=np.zeros([self.pN(), self.pN()], dtype=np.complex128)
for ii, kk in enumerate(itertools.product(*tuple(ZNl))):
for jj, ll in enumerate(itertools.product(*tuple(ZNl))):
DTM[ii, jj]=DFTcoef(np.array(kk, dtype=np.float),
np.array(ll), N)
DTMd=npmatlib.zeros([proddN, proddN], dtype=np.complex128)
for ii in range(self.d):
DTMd[prodN*ii:prodN*(ii+1), prodN*ii:prodN*(ii+1)]=DTM
return DTMd
