Two runs gets significantly different energies with ISPIN = 1 and 2 with mag = 0

Message

yhan_vasp · #1 Post by **yhan_vasp** » Sat Dec 04, 2021 9:36 am

Dear VASP developer,

I first ran the first job "PBE-600-21x22x23-m" (see the attached file "TwoRuns.rar") with ISPIN = 2 and got the converged energy:
-7.55964159
and the mag = -0.0000, indicating no magnetic contribution.

Then, I copied CONTCAR of the above job to a new POSCAR and set ISPIN = 1 as the second job "PBE-600-21x22x23". After I ran the second job, I got a significantly different energy:
-6.27543820

I also checked the final coordinates of atoms by comparing two CONTCAR files. There are almost the same. The difference less than 0.0001 in fractional coordinates is not possible to result in such large energy difference > 1 eV.

How do you understand this behavior that obviously does not make any sense?

Thank you for your time!

#2 Post by **henrique_miranda** » Sat Dec 11, 2021 3:34 pm

First of all two comments regarding your calculation:
1. You are using 192 cores to run a molecule with 2 atoms, this is a waste. You can run this calculation on a single-core machine without a significant slowdown.
2. You might also be able to reduce the amount of vacuum without a significant loss in total energy accuracy.

Regarding the energy differences, you can easily verify that they are not due to the relaxation of the ionic positions.
To do so just take one POSCAR and run a static calculation (no ionic relaxation) with (ISPIN=2) and without spin (ISPIN=1) and you will see that you get different total energies.

The reason for these differences is explained here:
https://vasp.at/wiki/index.php/Calculat ... y_of_atoms

Let me know if this explanation is not clear.

My Community

Two runs gets significantly different energies with ISPIN = 1 and 2 with mag = 0

Two runs gets significantly different energies with ISPIN = 1 and 2 with mag = 0

Re: Two runs gets significantly different energies with ISPIN = 1 and 2 with mag = 0