mirror of
https://github.com/triqs/dft_tools
synced 2024-11-01 11:43:47 +01:00
f7fad85fca
This is an iteration over the doc mainly thank to Priyanka. I fixed another couple of details on the way.
191 lines
6.6 KiB
Python
191 lines
6.6 KiB
Python
|
|
################################################################################
|
|
#
|
|
# TRIQS: a Toolbox for Research in Interacting Quantum Systems
|
|
#
|
|
# Copyright (C) 2011 by M. Ferrero, O. Parcollet
|
|
#
|
|
# 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/>.
|
|
#
|
|
################################################################################
|
|
|
|
import types,string,itertools
|
|
from operator import isSequenceType
|
|
import numpy
|
|
|
|
class DOS :
|
|
r"""
|
|
* Stores a density of state of fermions
|
|
|
|
.. math::
|
|
|
|
\rho (\epsilon) \equiv \sum'_k \delta( \epsilon - \epsilon_k)
|
|
|
|
* The sum is normalized
|
|
|
|
.. math::
|
|
|
|
\int_{-\infty}^{\infty} d\epsilon \rho (\epsilon) = 1
|
|
|
|
* Implement :ref:`Plot Protocol <plotting>`.
|
|
|
|
"""
|
|
def __init__(self, eps, rho, name = ''):
|
|
"""
|
|
Parameters
|
|
------------
|
|
eps : 1d array-type
|
|
eps[i] is value of epsilon.
|
|
rho : 1d array-type
|
|
The corresponding value of the dos.
|
|
name : string
|
|
Name of the dos/orbital
|
|
|
|
"""
|
|
self.name = name
|
|
try :
|
|
self.eps = numpy.array( eps )
|
|
assert len(self.eps.shape) ==1
|
|
except :
|
|
raise RuntimeError, "Argument eps mismatch"
|
|
try :
|
|
self.rho = numpy.array( rho )
|
|
assert len(self.rho.shape) ==1
|
|
except :
|
|
raise RuntimeError, "Argument rho mismatch"
|
|
assert self.eps.shape[0] == self.rho.shape[0], "Dimensions of eps and rho do not match"
|
|
|
|
|
|
#-------------------------------------------------------------
|
|
|
|
def __reduce__(self) :
|
|
return self.__class__, (self.eps,self.rho, self.name)
|
|
|
|
def __reduce_to_dict__(self) :
|
|
return {'epsilon' : self.eps, 'rho': self.rho, 'name' : self.name}
|
|
|
|
@classmethod
|
|
def __factory_from_dict__(cls,D) :
|
|
return cls(D['epsilon'],D['rho'], D['name'])
|
|
|
|
def __repr__(self) :
|
|
return """
|
|
DOS object :
|
|
"""%self.__dict__
|
|
|
|
def _plot_(self, Options) :
|
|
return [ {'type' : "XY", 'label' : self.name, 'xlabel' :r'$\epsilon$', 'ylabel' : r'%s$(\epsilon)$'%self.name, 'xdata' : self.eps,'ydata' : self.rho } ]
|
|
|
|
def density(self,mu=0):
|
|
"""Calculates the density of free fermions for the given DOS for chemical potential mu."""
|
|
|
|
dens = 0.0
|
|
a = [ (e>mu) for e in self.eps ]
|
|
try:
|
|
ind = a.index(True)
|
|
except:
|
|
ind = self.eps.shape[0]
|
|
|
|
de = self.eps[1]-self.eps[0]
|
|
#for e,r in itertools.izip(self.eps[0:ind],self.rho[0:ind]):
|
|
# dens += r
|
|
dens = (sum(self.rho[0:ind]) - self.rho[0]/2.0 - self.rho[ind-1]/2.0) * de
|
|
#dens2 = dens + (self.rho[ind-1]/2.0 + self.rho[ind]/2.0) * de
|
|
if (ind<self.eps.shape[0]): dens += (mu-self.eps[ind-1]) * (self.rho[ind-1] + self.rho[ind])/2.0
|
|
return dens
|
|
|
|
##########################################################################
|
|
|
|
def dos_from_file(Filename, name = '', single_orbital = None):
|
|
"""
|
|
Read the DOS from a file
|
|
|
|
:param Filename: a string : name of the file
|
|
:param name: name of the DOS
|
|
:param single_orbital: can be None or an integer.
|
|
|
|
:rtype:
|
|
* if single_orbital== None, returns a tuple of DOS (even if there is one dos !).
|
|
* If single_orbital==i, return only ONE DOS corresponding to ith orbital (starting at 1).
|
|
|
|
Format of the file :
|
|
* N_orbitals +1 columns,
|
|
* the first column is the value of epsilon
|
|
* the N_orbitals other columns are the values of the dos for various orbitals
|
|
"""
|
|
f = open(Filename); s=''
|
|
while not(s.strip()) :
|
|
s= f.readline()
|
|
assert s, "File is empty !"
|
|
N_Orbitals = len (s.split()) - 1
|
|
assert N_Orbitals >0, "File : wrong format"
|
|
# not very safe : fromfile routine can crashes if given non numerics
|
|
r = numpy.fromfile(Filename,sep=' ')
|
|
l,div = r.shape[0], N_Orbitals +1
|
|
assert l%(div)==0,"File does not contains N*%d numbers !"%(div)
|
|
r.shape = l//(div) , div # reshape the array
|
|
eps = r[:,0]
|
|
if single_orbital :
|
|
assert single_orbital>0 and single_orbital <= N_Orbitals, " single_orbital "
|
|
return DOS (r[:,0] ,r[:,single_orbital], name)
|
|
else :
|
|
return [ DOS (r[:,0] ,r[:,i +1 ], name) for i in range (N_Orbitals)]
|
|
|
|
|
|
##########################################################################
|
|
|
|
class DOSFromFunction(DOS):
|
|
"""
|
|
* A DOS class, but constructed from a function.
|
|
|
|
* The number of points can be variable and self-adjusted in the Hilbert transform to adapt precision.
|
|
"""
|
|
def __init__(self, function, x_min, x_max, n_pts=100, name=''):
|
|
"""
|
|
:param function: * a function :math:`\\epsilon \\rightarrow \\rho(\\epsilon)`
|
|
* The result type can be anything from which a 1d-array can be constructed by numpy
|
|
:param x_min,x_max: Bound of the mesh (domain of the function).
|
|
:param n_pts: Number of points in the mesh.
|
|
:param name: Name of the DOS.
|
|
"""
|
|
assert callable(function), "function is not callable"
|
|
self.function,self.x_min,self.x_max = function,x_min,x_max
|
|
try :
|
|
e = function(0.001)
|
|
len(numpy.array(e).shape) ==1
|
|
except :
|
|
raise RuntimeError , "Value of the function must be a 1d-array"
|
|
self.__f(n_pts) # compute arrays
|
|
DOS.__init__(self,self.eps,self.rho,name)
|
|
|
|
#-------------------------------------------------------------
|
|
|
|
def __reduce__(self) :
|
|
return self.__class__, (self.function,self.x_min, self.x_max, len(self.eps), self.name)
|
|
|
|
#-------------------------------------------------------------
|
|
|
|
def __f(self,N) :
|
|
r = (self.x_max - self.x_min)/float(N-1)
|
|
self.eps = numpy.array( [self.x_min + r* i for i in range(N) ] )
|
|
self.rho = numpy.array( [self.function(e) for e in self.eps])
|
|
|
|
#-----------------------------------------------------
|
|
# Register the class for HDFArchive
|
|
#-----------------------------------------------------
|
|
|
|
from pytriqs.archive.hdf_archive_schemes import register_class
|
|
register_class (DOS)
|
|
|