ABINIT Discussion Forums

I need to have the SCF converge for my band structure calculation. However, I tried both toldfe and toldff for this convergence and does not work even with nstep 100. But, I was able get a converged ecut, ngkpt, and did the relaxation for the 2D material. Can somebody share some tips on how to ease the convergence?(I'm using PAW pseudo). Below is part of the output.

Best,
Csun








ETOT 992 -338.19358479038 1.77E-02 4.52E-04 1.46E-01 5.0E-03 1.25E-01 2.000
ETOT 993 -338.24828206241 -5.47E-02 2.21E-04 2.20E-01 2.2E-02 1.03E-01 2.000
ETOT 994 -338.26468308427 -1.64E-02 8.00E-05 2.31E-01 5.4E-03 9.73E-02 2.000
ETOT 995 -338.17977483809 8.49E-02 1.00E-03 1.28E-01 3.3E-02 1.30E-01 2.000
ETOT 996 -338.22406092845 -4.43E-02 5.43E-04 1.72E-01 1.8E-02 1.12E-01 2.000
ETOT 997 -338.09253641020 1.32E-01 2.40E-03 1.35E-01 4.5E-02 1.57E-01 2.000
ETOT 998 -338.30619257465 -2.14E-01 1.07E-03 2.78E-01 7.8E-02 7.87E-02 2.000
ETOT 999 -338.07376582877 2.32E-01 1.34E-02 4.38E-02 7.4E-02 1.52E-01 2.000
ETOT 1000 -337.68432279831 3.89E-01 1.68E-02 1.13E-01 4.9E-02 2.02E-01 2.000

Cartesian components of stress tensor (hartree/bohr^3)
sigma(1 1)= 1.74578136E-04 sigma(3 2)= 0.00000000E+00
sigma(2 2)= 1.74578136E-04 sigma(3 1)= 0.00000000E+00
sigma(3 3)= -7.78109029E-04 sigma(2 1)= 0.00000000E+00


scprqt: WARNING -
nstep= 1000 was not enough SCF cycles to converge;
maximum force difference= 4.943E-02 exceeds toldff= 1.000E-06

Dear Csun,
I do not fully understand what do you mean by not converging the bandstructure, do you mean the SCF part or the non-SCF?
Best wishes,
Eric

Hello Eric,

I am referring to the the SCF cycle. The values just never converge. It goes up and down.

Best,
Csun

Dear Csun,
OK, you can follow what is advised here:
viewtopic.php?f=9&t=4530
There are other posts on this forum discussing how to tune the input flags to get SCF converged in troublesome cases.
Best wishes,
Eric