The "h2o-repimd" example: possible contradiction?

53 views

DriversDynamics

Skip to first unread message

dfipi

unread,

Jun 7, 2019, 5:29:04 AM6/7/19

to ipi-users

Dear Forum,

The h2o-repimd example contains input script for path-integral run using replica exchange on a water model.

The different replica (10 in total) use different coordinates as initial conditions (files *_init.xyz).

I wonder whether this is necessary, or is it possible to use the same equilibrated structure as initial coordinates for each (exchange) replica?

I'm asking because with a DFTB+ driver, the accompanying input script (dftb_in.psd) can have only *one* geometry.gen file defined for each input.xml file for i-pi.

Is there a contradiction or a mere misunderstanding on my side?

Best,

David

venkat kapil

unread,

Jun 7, 2019, 10:31:43 AM6/7/19

to ipi-users

Hello David,

If you want to launch M instances of DFTB+ then you can just execute it M times using the same geometry.gen file. This initial structure is not used by i-pi and it required to just start DFTB+. It doesn't even have to be the same structures (*_init.xyz) that i-PI uses. Once DFTB+ begins, i-PI will send updated positions and cell at every step and receive energetics.