IRC not going to reactants

[fabio_colasuonno]

Hello everyone,
I am running an IRC calculation starting from a TS optimised with the improved dimer. I already checked that it has only one imaginary frequency, so it is the correct TS. When setting up the IRC calculation using VASP 5.4.4 I use the following keywords:

Code: Select all

irc settings
IBRION=40           !  use the irc method as optimization engine
IRC_DIRECTION = -1
IRC_STOP = 40
IRC_VNORM0 = 0.0005

ionic relaxation options
EDIFFG    = -0.01   ! if negative, relax until forces are below this value (in eV/A) use high values when running NEB
ISIF      = 2
NSW       = 5000   ! Number of relaxation steps
POTIM     = 0.02

IVDW = 12

Despite the IRC_DIRECTION being set to -1, when opening the XDATCAR with a visualisation software, the geometry follows the reaction coordinate towards the products rather than going back to the reactants.
Moreover, when doing 'grep IRC OUTCAR' all the reaction coordinate points have positive values, almost like the code did not read the IRC_DIRECTION value.

I tried changing IRC_VNORM0 to a different value (0.002) but the geometry is still following the reaction coordinate in the other direction.

Any suggestion?

[henrique_miranda]

I think the solution in this case is to set IRC_DIRECTION = 1.
Note that -1 or 1 don't necessarily mean products or reactants.
Currently, there is no implementation in VASP to try to find in which direction is the product or reactant, only which mode to follow.
The phonon mode displacement is computed up to a phase, so it might be pointing in either the product or the reactant in a seemingly arbitrary way (in fact it depends on the numerics).
This is why there is the IRC_DIRECTION tag so that the user can perform two calculation in opposite directions and find both the product and the reactant.