Question regarding DMFT

structure determination,...

Moderators: jzwanzig, jolafc

Post Reply
Posts: 19
Joined: Wed Apr 15, 2015 1:05 am

Question regarding DMFT

Post by pipidog » Thu Nov 17, 2016 12:10 am

Dear all,

I'm trying to run DMFT calculation. I have a few question about the tutorial

1). What are the meaning of nnscol and nline? I checked abinit user manual but I still don't
get it. Why are they important to DMFT? Actually, I'm own calculation. I also noticed that
if nnscol and nnline are not set to higher values, the calculation will crash because wave
functions are not convergent. Can anyone tell me more about the two parameters?

2). Can anyone explain what are typical values of dmftqmc_therm and dmftctqmc_gmove?

3). The variables dmftbandi and dmftbandf are used to wannierize the bands. However, if I
just want to study the dmft effects on d-band but the chosen energy range has other band
involved. If so, I still need to chose the whole energy range to include all the other bands
even if they don't have strong d-band characters, right?

4). Why running DMFT, we will still set U and J for the solver, right? If so, why I always got
warning message in the log file saying that:

src_file: invars1.F90
src_line: 893
message: |
usedmft and usepawu are both activated
This is not an usual calculation:
usepawu will be put to 10:
LDA+U potential and energy will be put to zero

Any suggestion about the above questions are welcome !


Posts: 44
Joined: Mon Aug 24, 2009 10:58 am

Re: Question regarding DMFT

Post by amadon » Thu Nov 17, 2016 10:35 am


1) nnsclo and nline are used to insure that the Hamiltonian is correctly diagonalized and that the empty states
are meaningful. They are required.

2)dmftqmc_therm is the thermalization time. Typical values are 10000. But you should check that the results of your calculation are insensitive to this parameter (by increasing it).
dmftctqmc_gmove are related to global moves (see e.g section IX G of Rev. Mod. Phys. 83, 349 (2011)) . They should be used if the calculation is not ergodic. In most of the cases, this variable is not useful (but it can be checked by using small values of dmftctqmc_gmove and decreasing it).

3) The problem of the choice of Wannier orbitals for a specific systems is related to the choice of U and J. So, a value of these parameters is related
to the choice of correlated orbitals. In other words, there is no unique choice of Wannier functions. A discussion about this can be found in e.g. Phys. Rev. B 77, 205112 (2008) section III B and references cited.

4) This is a warning to remind the reader that DFT+U and DFT+DMFT cannot be done at the same time. As you are aware of this, you can thus ignore it without any trouble ;).

Best regards
Bernard Amadon

Post Reply