Re: [AMBER] Sequential equilibration versus parallel runs in TI

From: David A Case <>
Date: Thu, 5 Dec 2019 08:54:39 -0500

On Wed, Dec 04, 2019, Debarati DasGupta wrote:
>For TI calculations, there are two parts equilibration at various clamda
>values, followed by production at these different clambda values.
>Now if my lambda values are not equidistant, what I mean is *lambda
>is not fixed, so do I have to do my equilibriations runs in a
>hierarchical manner; start from 0.0159..then go to 0.1981..then to
> smaller lambda to higher lambda sequentially or can I run
>my equlibrations in parallel, and then do production runs.
>I m using 9 lambda values reported in Table 21.1 Page 425 of AMBER18 manual.

The point of "equilibration" is to remove dependence on the details of
the starting structure, and to adapt that structure to what the force
field thinks are low-energy configurations.

Hence, if you equilibration is complete enough, it won't matter whether
you do them serially, or in parallel. Nor will it matter whether you
lambda values are equidistant or use Gaussian quadrature nodes, or
follow some other scheme.


AMBER mailing list
Received on Thu Dec 05 2019 - 06:00:06 PST
Custom Search