On Sun, Apr 28, 2013 at 8:17 PM, Sebastian Petrik <
sebastian.petrik.griffithuni.edu.au> wrote:
> Hello,
>
> I have been attempting a simple minimize-warmup-dynamics run and seem to be
> running into a problem with both SANDER and PMEMD. My latest trial
> minimizes fine, warms up fine but as the dynamics run is attempted it fails
> on:
>
> | ERROR: I could not understand line 2597
> 412.9567261 427.1826530-780.4936416
>
> Inspection of the warmup RST file shows at about 2500 lines in some values
> start to get replaced with ************* which i assume is out of bounds
> and is probably not a good thing?
>
> A snippet of the lines around 2597 is provided below, which I hope someone
>
> can make sense of. It looks as if it might be truncated?
>
The issue is that the formatted restart file uses a fixed number of
characters in 12.7 precision (12 total characters with 7 decimal places).
Therefore, there is not enough space to represent numbers -1000 and lower
or 10000 and higher.
This can happen when you have systems comprised of multiple molecules that
are not connected by bonds, and these molecules naturally diffuse and drift
away from each other. This is typical for explicit solvent calculations,
only, since implicit solvent calculations are infinitely dilute (and the
lack of explicit solvent molecules means diffusion is impossible to model).
For explicit solvent calculations, there are two solutions: use NetCDF
restart files (Amber 12 or later, I believe), which can accommodate any
double precision value (not just those between -1000 and 10000). The
second is to use the iwrap=1 variable so that all particles that diffuse
out of the 'main' unit cell are 'wrapped' back into the 'main' unit cell on
the other side.
It typically takes a very long time for these particles to diffuse far
enough away to cause these types of errors (but with GPUs becoming more
common, this issue pops up more). The fact that this is happening in your
heating phase leads me to believe that you are doing this in implicit
solvent, and that you have numerous molecules that are not interconnected.
In this case, you will need to find a way to prevent this kind of particle
diffusion. You can either jump to explicit solvent, or you can use some
kind of restraint system (but think about the questions you are asking and
how these simulations will be able to answer those questions).
HTH,
Jason
P.S., Did you not get a helpful message from pmemd when you got this error?
You hopefully saw something similar to this email...
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Apr 28 2013 - 18:30:03 PDT
