Page 1 of 1

Question about ALGO=48 setting

Posted: Fri Jun 09, 2023 3:44 am
by Leon1999526
Hello,
I was dealing with a problem in electronic step convergence and took the suggestion of this post and changed ALGO from ALGO=Fast to ALGO=48: forum/viewtopic.php?p=17600

It worked for my system, but I didn't see on the wiki page saying that ALGO tag may have the value of ALGO=48, so I am not truly sure about what it actually is. (I only changed ALGO, not IALGO)

Re: Question about ALGO=48 setting

Posted: Mon Jun 12, 2023 7:01 am
by marie-therese.huebsch
Hi,

ALGO = 48 is a typo. The previous post was meant to suggest setting IALGO = 48.

The reason "it worked" for you is that if you set ALGO=48 instead of ALGO=Fast, the setting cannot be interpreted, and VASP uses the default setting for ALGO. The default is ALGO=Normal, which corresponds to blocked Davidson (IALGO=38). You can check that by searching for IALGO in your OUTCAR file.

ALGO=Fast is computationally cheaper but less robust, i.e., you may converge faster or never, and less accurate. ALGO=Normal is quite robust and accurate. For some purposes, when you, for instance, need unoccupied states, e.g., ALGO=Exact is the better choice because it is more accurate. For other calculations, direct optimization (IALGO=53-58) may be better for finding a local minimum, e.g., employing Hartree-Fock type and meta-GGA functionals. It is difficult to cover this topic here exhaustively, so I will stop thinking loudly. If you have convergence issues, I suggest you post the input and main output, and we look at possible strategies for your calculation.

Marie-Therese

Re: Question about ALGO=48 setting

Posted: Fri Jun 16, 2023 4:24 am
by Leon1999526
Ok. Thank you very much!