def get_leadsiw(self, leadsw, w_hyb, nleads, beta, niw):
norb = self.norbitals
nsp = self.nspins
if nleads:
leadsiw = np.zeros((niw, nleads, norb, nsp, norb, nsp),
dtype=complex)
#wre2mat wants the total # of Matsubara, positive + negative
# we pass the hybridization spectral density to the transformation
leadtrf = tr.transform(tr.wre2mat(beta, w_hyb, 'fermi', niw),
-leadsw.imag / np.pi)
leadtrf = leadtrf.transpose(
5, 0, 1, 2, 3, 4) # iw_n, lead, band, spin, band, spin
if not self.spinorbit:
#GS: self.leadsiw = np.zeros_like(leadsiw)
# in case of spin-independent Hamiltonians, we copy twice to the diagonal
leadsiw[:, :, :, :1, :, :1] = leadtrf
leadsiw[:, :, :, 1:2, :, 1:2] = leadtrf
else:
leadsiw = leadtrf
# leadsiw.real += self.hk[0,...].real #adding a (band, spin, band, spin) array to a (w, lead, band, spin, band, spin) one
#GS:!!!!!! leadsiw.real += self.hk.mean(0).real #here I am less sure about the right indices, but it seems to work. Ask Markus!
leadsmom = tr.transform(tr.wre_int(beta, w_hyb),
-leadsw.imag / np.pi)
print("leadsmom = ", leadsmom.shape)
leadsmom = leadsmom.transpose(5, 0, 1, 2, 3, 4)
# print "leadsmom = ", leadsmom
else:
leadsiw = np.zeros((niw, 1, norb, nsp, norb, nsp), dtype=complex)
leadsmom = np.zeros((1, norb, nsp, norb, nsp))
return leadsiw, leadsmom
def get_giw(cls, config, problem, result):
#typ = "legendre" # FIXME: QMC config does not contain FTtype
typ = config["General"]["FTType"]
if config["QMC"]["offdiag"] == 1 and typ == "legendre":
typ = "legendre_full"
if typ == "none":
giw = -result["giw-meas"]
elif typ == "none_worm":
giw = result["giw-worm"]
giw = orbspin.extract_diagonal(giw.transpose(4,0,1,2,3)) \
.transpose(1,2,0)
elif typ == "plain":
ntau = config["QMC"]["Ntau"]
tf_matrix = tf.tau2mat(problem.beta, ntau, 'fermi', problem.niw)
giw = tf.transform(tf_matrix, result["gtau"], None)
elif typ == "legendre":
nleg = config["QMC"]["NLegOrder"]
tf_matrix = -tf.leg2mat(nleg, problem.niw)
giw = tf.transform(tf_matrix,
result["gleg"],
None,
onmismatch=tf.Truncate.tofirst,
warn=False)
elif typ == "legendre_full":
nleg = config["QMC"]["NLegOrder"]
tf_matrix = -tf.leg2mat(nleg, problem.niw)
nbands = result["gleg-full"].shape[0]
giw = np.zeros(shape=(nbands, 2, nbands, 2, problem.niw),
dtype=complex)
for b in range(0, nbands):
for s in range(0, 2):
giw[b, s, :, :, :] = tf.transform(
tf_matrix,
result["gleg-full"][b, s, :, :, :],
None,
onmismatch=tf.Truncate.tofirst,
warn=False)
#### the old one
#tmp = tf.transform(tf_matrix, result["gleg-full"][b,s,:,:,:], None,
#onmismatch=tf.Truncate.tofirst, warn=False)
#for b2 in range(0,nbands):
#for s2 in range(0,2):
#for w in range(0,problem.niw):
#print "w", w
#giw[b,s,b2,s2,w] = tmp[b2,s2,w]
giw = giw.reshape(nbands * 2, nbands * 2, problem.niw)
giw_new = np.zeros_like(giw, dtype=complex)
for i in range(0, problem.niw):
tmp = giw[:, :, i]
giw_new[:, :, i] = 0.5 * (tmp.transpose(1, 0) + tmp)
giw = giw_new
else:
raise RuntimeError("Invalid transform type: `%s'" % typ)
return giw # copy not necessary here.
# Use hard legendre order cut-off
if options.maxorder:
print >> err, "Hard cut of %d coefficients" % (gleg.shape[-1]-options.maxorder-1)
gleg = gleg[...,:options.maxorder+1]
gleg_error = gleg_error[...,:options.maxorder+1]
# Use soft sigma cut-off
if options.threshold:
noisy = np.abs(gleg)/gleg_error < options.threshold
print >> err, "Number of cut noisy coefficients:",
print >> err, ", ".join(map(str, np.sum(noisy, -1).flat))
gleg[noisy] = 0.
iw = 1j * hf["axes/iw"].value
gliw, gliw_error = tf.transform(-tf.leg2mat(gleg.shape[-1], iw.size),
gleg, gleg_error)
grp = iter.create_group("giw-smooth")
grp.attrs["desc"] = "Smoothened Green's function from Legendre"
grp.attrs["axes"] = ["ineq", "band", "spin", "iw"]
grp.create_dataset("value", data=gliw)
grp.create_dataset("error", data=gliw_error)
# Now, compute G0(iw) in order to find the self energy
fiw = iter["fiw/value"].value
if "mu-imp" in iter:
g0iw_inv = iw - iter["mu-imp/value"][...,np.newaxis] - fiw[::-1]
else:
g0iw_inv = iw + iter["mu/value"] - fiw[::-1]
# Checks for old run
if not ineq_name.startswith("ineq-"):
continue
ineq_node = iter_node[ineq_name]
if "gloctau" in ineq_node: # replace with your quantity name
print("WARNING: overriding")
del ineq_node["gloctau"]
gtau_mean = ineq_node["gtau/value"].value
gtau_error = ineq_node["gtau/error"].value
glociw = ineq_node["gloc/value"].value
mat2tau_matrix = tf.mat2tau(beta, 'fermi', glociw.shape[-1], taus)
gloctau = -tf.transform(mat2tau_matrix, glociw - 1 / iw) + .5
num_dbands = gtau_mean.shape[0]
d_part = slice(None, num_dbands)
p_part = slice(num_dbands, None)
if not np.allclose(gloctau.imag, 0):
raise RuntimeError("Gloctau is not real????")
if not np.allclose(gloctau[d_part], gtau_mean):
print("WARNING: d-part of G(tau) does not match with impurity")
gloctau = gloctau.real
gloctau_err = np.empty_like(gloctau)
gloctau_err[d_part] = gtau_error
gloctau_err[p_part] = gloctau_err[d_part].mean(0)[None]
gloctau_group = ineq_node.create_group("gloctau")
