never seen before output, support inunderstanding it

[patrizio_graziosi]

Hello everybody,

I have started to run calculations on a structure given to me by experimental colleagues. During the convergence tests I find in the OUTCAR file what follows:

eigenvalues EBANDS = ******************
atomic energy EATOM = 32269.71645653
Solvation Ediel_sol = 0.00000000
---------------------------------------------------
free energy TOTEN = ****************** eV

energy without entropy =****************** energy(sigma->0) =******************

Would you mind to help in understanding that meaning? ...I have never seen it...

Thanks
Patrizio

[michael_wolloch]

Dear Patrizio Graziosi,

This output is created if the number has too many digits to fit in the output field for the write statement.

Please attach your input and output files, according to the posting guidelines, so I can analyze what goes wrong and if we need to fix this issue.

Thanks, Michael

[patrizio_graziosi]

Dear Michael,

thanks for the fast reply. I attach the INCAR and OUTCAR files for two k-grids, 2 2 2 and 3 3 3, together with the POSCAR; the message appears only in the first case.

Thanks,
Patrizio

[michael_wolloch]

Dear Patrizio,

Thanks for the files. It looks like the calculation with the lower k-point mesh diverged after around 12 electronic steps.
Here are the free energies (TOTEN) for both runs from step 10 to step 15:

Code: Select all

 2x2x2                        3x3x3
-1983.81941312               -1975.89172162
-1872.02127735               -1801.08461317
-1780.33849634               -1573.03982291
-2625.77392198               -1166.65860979
-4667.88520473               -1042.59180719
-34279912.71741207           -1003.08212304

Soon after, the energies are too large to fit into the output field. And the 2x2x2 calculation never converges. I will do some tests to check what goes wrong. Here are a few initial thoughts:

• For this system, which I would call a molecule in a box, you should do the simulation with the gamma point only.

• The vacuum space in the x and y directions seems small, especially with vdW forces enabled. You will get lots of interactions between the replicas.

• ALGO=All should be more robust if you keep having convergence troubles.

• Electrostatic corrections might play a role.

I hope this helps,
Michael

[michael_wolloch]

Hi Patrizio,

I have converged your system (initially without vdW interactions) without much trouble using

• ALGO = All

• ISEARCH = 1

• LREAL = Auto

• SIGMA = 0.1

• NELMDL = -10

• NELM = 200

Otherwise, I copied your settings (other than parallelization) and made a \(\Gamma\)-point only calculation, using the appropriate version of VASP. It takes a while ( around 150 steps), but the system converges well.

I then started from a WAVECAR using DFT-D3 corrections, and convergence was achieved in a few steps.

Let me know if you have success with this as well, and if I can lock the topic.

Cheers, Michael

[patrizio_graziosi]

Hello Michael,

so with these input tags you could converge the system with a k-points mesh not restricted to Gamma-only? If so, I will try these information.

In the meantime, I am also trying to understand a realistic periodicity together with my experimental colleagues because the system must be a 3D crystal, not a single-molecule...
Patrizio

[michael_wolloch]

Hi Patrizio,

I am just checking in to see if you still have any questions on this topic? Did you succeed with a tighter mesh and a periodic structure?
If there are no more issues, please let me know so I can lock the topic.

Cheers, Michael