My Community

Hi

I am trying to calculate the elastic constant for a perfect Zr supercell using 'ISIF = 3 and IBRION = 6'. I have found this calculation extremely hard to converge. Though, with the use of a different IBRION for the same supercell only took 5 to 6 hours to finish the calculation. I wonder if there is anything I could do to speed up the calculation or if there is another way to calculate the elastic constant for the supercell. Please see the attachment for the input files.

Best

Hi,

Thanks for bringing up the topic. I will try to have a look at it. Could you please share both the input and the main output files? Please also share the other calculation where you used a different IBRION that was faster.

Cheers
Marie-Therese

Hi,

Please see the attachment for the input and output files for my calculations using different 'IBRION'. For the IBRION = 6 calculations, I restart the calculation by changing the CONTCAR to POSCAR each time. In the attached folder, the output files are for my first and latest calculations.

Best

Hi,

The settings IBRION = 2 (ionic relaxation, conjugate gradient algorithm) and IBRION = 6 (second derivatives, finite differences) are two very different calculations, so you cannot compare the computational cost. But more importantly, the calculation does not follow the same workflow.

For ionic relaxation, it is correct to copy the CONTCAR to POSCAR to continue the ionic relaxation; For instance, if you set a stricter convergence criterion or if the maximum number of ionic steps was reached before convergence was achieved.

For IBRION = 6, all the ionic displacements are performed to compute the Hessian matrix, which is the second derivative of the Born–Oppenheimer energy surface w.r.t. ionic displacement and the main ingredient to compute phonon frequencies. When ISIF = 3, the cell volume, cell shape, and ionic positions are allowed to change. This is a prerequisite to computing the second derivative of the Born–Oppenheimer energy surface w.r.t. strain, which is the ion-clamped elastic constant. Thus, you only need to run VASP once.

Your calculation suddenly stopped without any reason being indicated in the OUTCAR file. Maybe there was a message printed to the standard output of the terminal when the calculation terminated. You can try to perform the calculation for some small unit cell (maybe even with small ENCUT and few KPOINTS) to make the calculation cheap and see what output should be printed. The elastic constant is written toward the end of the OUTCAR file after all ionic displacements are done.

Best,
Marie-Therese