extremely large forces outputted by abinit for the equilibrium structure

Total energy, geometry optimization, DFT+U, spin....

Moderator: bguster

Locked
whw19850730
Posts: 2
Joined: Thu Dec 17, 2020 3:16 am

extremely large forces outputted by abinit for the equilibrium structure

Post by whw19850730 » Thu Dec 17, 2020 4:19 am

Dear Abinit developer,

I am Hongwei Wang, a postdoctoral fellow at the University of Minnesota.
I am trying to study the two-dimensional material SnSe with Abinit code.
When I use Abinit to calculate the equilibrium structure optimized by VASP code,
the output forces are extremely large,indicating the structure is unstable for
Abinit. I have tried with both official PAW and norm conservation pseudo potentials,
but this problem is unsolved. However, the output forces are small and acceptable when I use Quantum Espresso to calculate with the same pseudopotentials. I have investigated many three-dimensional oxide materials before,Abinit can always give consistent results with VASP and Quantum Espresso. I think there must be certain specific parameters needed to add to the Abinit input file, I hope you could help me solve this problem.
I look forward to your reply, Thanks.


The attachments are my Abinit input files for SnSe. I cannot upload pseudopotentials to
the mail-list system, which are psp8 format downloaded from Abinit homepage.

#####################################
forces calculated by Abinit

cartesian forces (hartree/bohr) at end:
1 0.00000000000000 -0.20198131688915 0.03608491415769
2 0.00000000000000 -0.20198131688915 -0.03608491415769
3 0.00000000000000 0.20198131688915 0.03888216878455
4 0.00000000000000 0.20198131688915 -0.03888216878455

cartesian forces (eV/Angstrom) at end:
1 0.00000000000000 -10.38629604072255 1.85556073610250
2 0.00000000000000 -10.38629604072255 -1.85556073610250
3 0.00000000000000 10.38629604072255 1.99940134029009
4 0.00000000000000 10.38629604072255 -1.99940134029009
################################
forces calculated by Quantum Espresso
Forces acting on atoms (cartesian axes, Ry/au):

atom 1 type 1 force = 0.00000000 -0.00226581 0.00006872
atom 2 type 1 force = 0.00000000 -0.00226581 -0.00006872
atom 3 type 2 force = 0.00000000 0.00226581 -0.00149451
atom 4 type 2 force = 0.00000000 0.00226581 0.00149451
####################################

Best Regards

Hongwei
Attachments
SnSe.out
(24.91 KiB) Downloaded 291 times
SnSe.in
(1.72 KiB) Downloaded 290 times

gonze
Posts: 412
Joined: Fri Aug 14, 2009 8:29 pm

Re: extremely large forces outputted by abinit for the equilibrium structure

Post by gonze » Thu Dec 24, 2020 5:00 pm

Dear Hongwei Wang,

I tried to reproduce your problem with ABINIT 9.2.2, but I got good forces.
I note that you are using v8.8.4, with the following warning
appearing in the output file that you have posted:

The starting date is more than 2 years after the initial release
of this version of ABINIT, namely Jul 2018.
Note that the use beyond 3 years after the release will not be supported.
Action: please, switch to a more recent version of ABINIT.

I also suppressed the input variables
chksymbreak 0
chkprim 0
because there is no need to have them, and finally, I suppressed
tolsym 1.e-5
since this is the default value for ABINIT9.2.2.

Fro the attached input file, I got the attached output file, with small residual forces.

fcart -0.0000000000E+00 -2.6846053275E-06 1.5633831339E-04
-0.0000000000E+00 -2.6846053275E-06 -1.5633831339E-04
-0.0000000000E+00 2.6846053275E-06 -1.4704458212E-04
-0.0000000000E+00 2.6846053275E-06 1.4704458212E-04

So, the problem does not show up, on my machine (also 16 processors like you, but I used gnu9.3 while you used intel15.0) with the use of v9.2.2 and suppression of the above-mentioned input variables.
I have not tried to reproduce your problem with v8.8.4 or on another machine.
Please, may I suggest that you switch to v9.2.2. If you still have the problem, then I will try to see what is happening when I compile with intel15.0.

Best wishes,
Xavier
Attachments
SnSe.in
(1.72 KiB) Downloaded 314 times
SnSe.out
(26.08 KiB) Downloaded 286 times

Locked