Errors in AIMD simulation

[manoj_wijesingha]

Hi Colleagues,
When I was running the AIMD simulation, it was popped up the error message "Error EDDDAV: Call to ZHEGV failed. Returncode = 23 1 24". I have attached the files for your review.
Thank you.

Best
Manoj

[manuel_engel1]

Hi Manoj,

Thanks for reaching out! This error that you encounter means that the Davidson algorithm during the electronic minimization step has failed. However, this algorithm failing is usually an indicator that something else is going on.

During the final few MD steps that lead up to the crash, VASP issues the following information

Code: Select all

Information: wavefunction orthogonal band   72  0.8598

The final SCF loop that fails also shows really huge energies:

Code: Select all

       N       E
DAV:   1    -0.879874029851E+02 ...
DAV:   2    -0.187511517486E+06 ... <- here
DAV:   3    -0.782599980741E+02 ...
DAV:   4    -0.830074247797E+02 ...

This means that the AIMD calculation you are performing is not stable enough. First, I encourage you to look at the final few atomic configuration in the XDATCAR file and see if the dynamics make sense (are the atoms too close to each other? are they accelerating? etc.). This could give you an impression of what is going on. However, in any case, you will have to change a few computational parameters to improve the stability of the calculation. Here are a few suggestions:

• Try increasing the number of k-points. This system is relatively small and might not properly converge with just the Gamma point, leading to instabilities.

• Try setting IWAVPR=11. This is a simpler prediction algorithm for the wavefunctions compared to the default for MD (IWAVPR=12). However, in your case, it might improve the stability. See this forum post for more context.

• Depending on what you see in the trajectory file (XDATCAR), you might want to decrease your time step via POTIM.

[manoj_wijesingha]

Thank you so much for your response and the explanation, Manuel.

Best,
Manoj