mirror of
https://github.com/triqs/dft_tools
synced 2024-11-07 06:33:48 +01:00
b90e1e80e2
* recompute maps every time SK called rather than saving them * user saves and feeds in chemical potential and dc manually * set_dc sets dc to known values (eg from previous calculations) while calc_dc computes them * find_mu -> calc_mu to match API for other functions
130 lines
7.7 KiB
ReStructuredText
130 lines
7.7 KiB
ReStructuredText
.. index:: DFT+DMFT calculation
|
|
|
|
.. _DFTDMFTmain:
|
|
|
|
The DFT+DMFT calculation
|
|
========================
|
|
|
|
After having set up the hdf5 arxive, we can now do our DFT+DMFT calculation. It consists of
|
|
initialisation steps, and the actual DMFT self consistency loop.
|
|
|
|
.. index:: initialisation of DFT+DMFT
|
|
|
|
Initialisation of the calculation
|
|
---------------------------------
|
|
|
|
Before doing the calculation, we have to intialize all the objects that we will need. The first thing is the
|
|
:class:`SumkDFT` class. It contains all basic routines that are necessary to perform a summation in k-space
|
|
to get the local quantities used in DMFT. It is initialized by::
|
|
|
|
from pytriqs.applications.dft.sumk_dft import *
|
|
SK = SumkDFT(hdf_file = filename)
|
|
|
|
The only necessary parameter is the filename of the hdf5 archive. In addition, there are some optional parameters:
|
|
|
|
* `mu`: The chemical potential at initialization. This value is only used if no other value is found in the hdf5 arxive. The default value is 0.0.
|
|
* `h_field`: External magnetic field. The default value is 0.0.
|
|
* `use_dft_blocks`: If true, the structure of the density matrix is analysed at initialisation, and non-zero matrix elements
|
|
are identified. The DMFT calculation is then restricted to these matrix elements, yielding a more efficient solution of the
|
|
local interaction problem. Degeneracies in orbital and spin space are also identified and stored for later use. The default value is `False`.
|
|
* `dft_data`, `symmcorr_data`, `parproj_data`, `symmpar_data`, `bands_data`: These string variables define the subgroups in the hdf5 arxive,
|
|
where the corresponding information is stored. The default values are consistent with those in :ref:`interfacetowien`.
|
|
|
|
At initialisation, the necessary data is read from the hdf5 file. If a calculation is restarted based on a previous hdf5 file, information on
|
|
degenerate shells, the block structure of the density matrix, the chemical potential, and double counting correction is also read in.
|
|
|
|
.. index:: Multiband solver
|
|
|
|
Setting up the Multi-Band Solver
|
|
--------------------------------
|
|
|
|
There is a module that helps setting up the multiband CTQMC solver. It is loaded and initialized by::
|
|
|
|
from pytriqs.applications.dft.solver_multiband import *
|
|
S = SolverMultiBand(beta, n_orb, gf_struct = SK.gf_struct_solver[0], map=SK.map[0])
|
|
|
|
The necessary parameters are the inverse temperature `beta`, the Coulomb interaction `U_interact`, the Hund's rule coupling `J_hund`,
|
|
and the number of orbitals `n_orb`. There are again several optional parameters that allow the tailoring of the local Hamiltonian to
|
|
specific needs. They are:
|
|
|
|
* `gf_struct`: The block structure of the local density matrix given in the format calculated by :class:`SumkDFT`.
|
|
* `map`: If `gf_struct` is given as parameter, `map` also must be given. This is the mapping from the block structure to a general
|
|
up/down structure.
|
|
|
|
The solver method is called later by this statement::
|
|
|
|
S.solve(U_interact,J_hund,use_spinflip=False,use_matrix=True,
|
|
l=2,T=None, dim_reps=None, irep=None, n_cycles =10000,
|
|
length_cycle=200,n_warmup_cycles=1000)
|
|
|
|
The parameters for the Coulomb interaction `U_interact` and the Hund's coupling `J_hund` are necessary input parameters. The rest are optional
|
|
parameters for which default values are set. Generally, they should be reset for the problem at hand. Here is a description of the parameters:
|
|
|
|
* `use_matrix`: If `True`, the interaction matrix is calculated from Slater integrals, which are computed from `U_interact` and
|
|
`J_hund`. Otherwise, a Kanamori representation is used. Attention: We define the intraorbital interaction as
|
|
`U_interact`, the interorbital interaction for opposite spins as `U_interact-2*J_hund`, and interorbital for equal spins as
|
|
`U_interact-3*J_hund`.
|
|
* `T`: The matrix that transforms the interaction matrix from spherical harmonics to a symmetry-adapted basis. Only effective for Slater
|
|
parametrisation, i.e. `use_matrix=True`.
|
|
* `l`: The orbital quantum number. Again, only effective for Slater parametrisation, i.e. `use_matrix=True`.
|
|
* `use_spinflip`: If `True`, the full rotationally-invariant interaction is used. Otherwise, only density-density terms are
|
|
kept in the local Hamiltonian.
|
|
* `dim_reps`: If only a subset of the full d-shell is used as correlated orbtials, one can specify here the dimensions of all the subspaces
|
|
of the d-shell, i.e. t2g and eg. Only effective for Slater parametrisation.
|
|
* `irep`: The index in the list `dim_reps` of the subset that is used. Only effective for Slater parametrisation.
|
|
* `n_cycles`: Number of CTQMC cycles (a sequence of moves followed by a measurement) per core. The default value of 10000 is the minimum, and generally should be increased.
|
|
* `length_cycle`: Number of CTQMC moves per one cycle.
|
|
* `n_warmup_cycles`: Number of initial CTQMC cycles before measurements start. Usually of order of 10000, sometimes needs to be increased significantly.
|
|
|
|
Most of above parameters can be taken directly from the :class:`SumkDFT` class, without defining them by hand. We will see a specific example
|
|
at the end of this tutorial.
|
|
|
|
|
|
.. index:: DFT+DMFT loop, one-shot calculation
|
|
|
|
Doing the DMFT loop
|
|
-------------------
|
|
|
|
Having initialised the SumK class and the Solver, we can proceed with the DMFT loop itself. As explained in the tutorial, we have to
|
|
set up the loop over DMFT iterations and the self-consistency condition::
|
|
|
|
n_loops = 5
|
|
for iteration_number in range(n_loops) : # start the DMFT loop
|
|
|
|
SK.put_Sigma(Sigma_imp = [ S.Sigma ]) # Put self energy to the SumK class
|
|
chemical_potential = SK.calc_mu() # calculate the chemical potential for the given density
|
|
S.G << SK.extract_G_loc()[0] # extract the local Green function
|
|
S.G0 << inverse(S.Sigma + inverse(S.G)) # finally get G0, the input for the Solver
|
|
|
|
S.solve(U_interact,J_hund,use_spinflip=False,use_matrix=True, # now solve the impurity problem
|
|
l=2,T=None, dim_reps=None, irep=None, n_cycles =10000,
|
|
length_cycle=200,n_warmup_cycles=1000)
|
|
|
|
dm = S.G.density() # Density matrix of the impurity problem
|
|
SK.calc_dc( dm, U_interact = U, J_hund = J, use_dc_formula = 0) # Set the double counting term
|
|
SK.save(['chemical_potential','dc_imp','dc_energ']) # Save data in the hdf5 archive
|
|
|
|
These basic steps are enough to set up the basic DMFT Loop. For a detailed description of the :class:`SumkDFT` routines,
|
|
see the reference manual. After the self-consistency steps, the solution of the Anderson impurity problem is calculation by CTQMC.
|
|
Different to model calculations, we have to do a few more steps after this, because of the double-counting correction. We first
|
|
calculate the density of the impurity problem. Then, the routine `calc_dc` takes as parameters this density matrix, the
|
|
Coulomb interaction, Hund's rule coupling, and the type of double-counting that should be used. Possible values for `use_dc_formula` are:
|
|
|
|
* `0`: Full-localised limit
|
|
* `1`: DC formula as given in K. Held, Adv. Phys. 56, 829 (2007).
|
|
* `2`: Around-mean-field
|
|
|
|
At the end of the calculation, we can save the Greens function and self energy into a file::
|
|
|
|
from pytriqs.archive import HDFArchive
|
|
import pytriqs.utility.mpi as mpi
|
|
if mpi.is_master_node():
|
|
ar = HDFArchive("YourDFTDMFTcalculation.h5",'w')
|
|
ar["G"] = S.G
|
|
ar["Sigma"] = S.Sigma
|
|
|
|
This is it!
|
|
|
|
These are the essential steps to do a one-shot DFT+DMFT calculation. For full charge-self consistent calculations, there are some more things
|
|
to consider, which we will see later on.
|