dft_tools/doc/tour/dmft.rst

Dynamical mean-field theory on a Bethe lattice
----------------------------------------------

.. note::
  
   Requires TRIQS and the :doc:`application cthyb_matrix <../../applications>`
  

In the case of Bethe lattice the dynamical mean-field theory (DMFT) self-consistency condition takes a particularly simple form

.. math::

  G^{-1}_{0,\sigma} (i \omega_n) = i \omega_n + \mu - t^2 G_{\sigma} (i \omega_n).


Hence, from a strictly technical point of view, in this case the DMFT cycle can be implemented by modifying 
the previous single-impurity example to the case of a bath with semi-circular density of states and adding a python loop to update :math:`G_0` as function of :math:`G`.

Here is a complete program doing this plain-vanilla DMFT  on a half-filled one-band Bethe lattice:


.. runblock:: python

   from pytriqs.gf.local import *
   from pytriqs.operators import *
   from pytriqs.archive import *
   import pytriqs.utility.mpi as mpi

   # Set up a few parameters
   U = 2.5
   half_bandwidth = 1.0
   chemical_potential = U/2.0
   beta = 100
   n_loops = 5

   # Construct the CTQMC solver
   from pytriqs.applications.impurity_solvers.cthyb_matrix import Solver
   S = Solver(beta = beta, gf_struct = [ ('up',[1]), ('down',[1]) ])

   # Initalize the Green's function to a semi circular
   S.G <<= SemiCircular(half_bandwidth)

   # Now do the DMFT loop
   for IterationNumber in range(n_loops):

       # Compute S.G0 with the self-consistency condition while imposing paramagnetism
       g = 0.5 * ( S.G['up'] + S.G['down'] )
       for name, g0block in S.G0:
           g0block <<= inverse( iOmega_n + chemical_potential - (half_bandwidth/2.0)**2  * g )

       # Run the solver
       S.solve(H_local = U * N('up',1) * N('down',1),                          # Local Hamiltonian
               quantum_numbers = { 'Nup': N('up',1), 'Ndown': N('down',1) }, # Quantum Numbers (operators commuting with H_Local)
               n_cycles = 5000,                                                # Number of QMC cycles
               length_cycle = 200,                                             # Length of a cycle
               n_warmup_cycles = 1000,                                         # How many warmup cycles
               n_legendre = 30,                                                # Use 30 Legendre coefficients to represent G(tau)
               random_name = "mt19937",                                        # Use the Mersenne Twister 19937 random generator
               use_segment_picture = True)                                     # Here we can use the segment picture

       # Some intermediate saves
       if mpi.is_master_node():
         R = HDFArchive("single_site_bethe.h5")
         R["G-%s"%IterationNumber] = S.G
         del R

       # Here we would usually write some convergence test
       # if Converged: break