Re: [AMBER] Assigning new velocities from Maxwell-Boltzmann distribution (ntx = 1)

From: David A Case <>
Date: Mon, 10 Aug 2020 07:57:40 -0400

On Sun, Aug 09, 2020, German P. Barletta wrote:

> When I start new replicas, I give them 5ns for "re-equilbration". The
>input for this re-equilibration is the same as the one for production but
>instead of reading velocities from the input file (ntx = 5), I assign new
>velocities at random (ntx = 1).
> The problem is, for some reason, Amber says the starting temperature
>(~450K) is way higher than it is supposed to be (temp0 = tempi = 300.0) .
>100ps after, it's back to normal.

This is not really a problem: it has to do with the fact that the random
velocity assignment in pmemd/sander doesn't know about SHAKE. This extra
temperature goes away very quickly -- try a short run with ntpr=1.

That said, cpptraj has a solution to this (I think), so that you can assign
random velocities there (making a new restart file). That algorithm will
respect SHAKE. Somebody should put this into the main code.


AMBER mailing list
Received on Mon Aug 10 2020 - 05:00:02 PDT
Custom Search