Next: How to run the
Up: NCA_descrip
Previous: A note on calculating
How to generate the NCA input file
In general case, the matrix
is a full matrix between the Hilbert subspaces of
and electrons. The best choice of base for solving NCA
equations is the local eigenbase, where the impurity energies
are diagonal. In this case, one needs to
write the matrix in local eigenbase as well. First, one can write
in direct base, where it is very simple and mostly zero. However,
one needs to pay some attention to the fermionic sign that appears due
to the fermionic nature of the electron operator . Then, one
can transform to the local eigenbase.
The matrix is greatly simplified when the hybridization
is diagonal and Coulomb repulsion is of the form
since in this case, the direct base is also the
local eigenbase. For this case, the program called generate
generates the necessary base and calculates the matrix, which is
necessary to run the NCA code.
Let me give a simple example of what the program generate
does. In the two band case, with the crystal field splitting,
the local states are
|
(25) |
where is the total number of electrons on the impurity, is
degeneracy of the state and a state represents an atomic state
with 2 electrons in the first band and 1 electron in the second. To
generate this output, one needs to give the following data to the
program
- number of baths : 2
- degeneracy of the first bath : 2
- degeneracy of the second bath: 2
In the program generate, user can then take only some part of
the atomic states (to speed up the calculation) or all of the
states. For example, if one is interested in the
and is large, there is no point of keeping the state with 4
electrons on the local level since the probability for the state
number 8 is very small. The program generate than calculates
the matrix and writes it to a file which needs to be given to the
NCA program.
If one is interested in the SU(N) problem, it is best to choose
- number of baths : 1
- degeneracy of the bath : N .
Next: How to run the
Up: NCA_descrip
Previous: A note on calculating
Kristjan Haule
2004-08-23