OmegaMaxEnt: The imaginary time step can be either too small or too large

Running Abinit Many-body theory
1

Dear everyone,
Follow the DMFT tutorial, I’m try to extract the Spectral function of SrVO3, but the OmegaMaxEnt show me the errors below. Anyone Knows what’s going on?

Any suggestions will be welcome!
Best,
Chao

OmegaMaxEnt Copyright (C) 2015 Dominic Bergeron (dominic.bergeron@usherbrooke.ca)
This program comes with ABSOLUTELY NO WARRANTY; This is free software, and you
are welcome to redistribute it under certain conditions; For details refer to
the GNU General Public License, of which you should have received a copy
along with the program, or see http://www.gnu.org/licenses/.

OTHER PARAMETERS (different from default ones):

INPUT PARAMETERS:
data file: Gtau.dat
data file loaded
bosonic data (yes/[no]): no
imaginary time data (yes/[no]): yes

PREPROCESSING

fermionic Green function given in imaginary time
For a Green function given in imaginary time, the first two moments are required to obtain the Matsubara frequency Green function.
The program will try to extract those moments from the behavior of the Green function around tau=0 and tau=beta.
Number of imaginary time slices: 50
-G(0)-G(beta): 1
not enough information provided to define the real frequency grid. The program will try to extract moments from G(tau) around tau=0 and tau=beta.
temperature: 0.00380018
no errors provided
using a constant error
COMPUTING MOMENTS
compute_moments_tau(): unable to compute the moments from G(tau). The imaginary time step can be either too small or too large. You can either change the step, provide the first moment, or increase parameter R_sv_min in file “OmegaMaxEnt_other_params.dat”.
Computation of moments failed.
continue execution? ([y]/n):
y
continue execution? ([y]/n):
^C

2

I increased the dmftqmc_l from 50 to 100, and dmft_nwlo from 100 to 200 , the errors are disappeared. I don’t know what’s happenned but it worked.

Best,
Chao

3

Dear Chao,

When the input data for OmegaMaxEnt is an imaginary time Green’s function G(tau), OmegaMaxEnt makes first a polynomial fit of your Green’s function before doing the Fourier transform.
If the number of Matsubara frequencies is to small, i.e. if the highest frequency is to low, the program will not be able to get the asymptotic behavior of your Green’s function G(iw).

To solve this issue, you can increase the number of Matsubara frequencies with a new DMFT calculation (That’s what you did by increasing dmft_nwlo from 100 to 200).
Another solution is simply to play with the input parameters of OmegaMaxent. In the OmegaMaxEnt_other_params.dat file, you can increase the number of values of tau used for the polynomial fit (“FNfitTauW”) to 5 instead of 4, in order to help the program for the fit.
This is well explained in the user manual of OmegaMaxEnt.

Best,
Fred.