From: Todd Trimble (TODD.TRIMBLE_at_asu.edu)
Date: Tue Nov 28 2006 - 17:18:24 CST
Hi Brian,
Thanks for responding. I'm running v. 2.5 on a cluster with Xeon
processors running Red Hat Linux.
Originally, I suspected that the PME grid or patch grid were the problem
in my NPT simulations. Using the Langevin piston for pressure control,
the system volume increased approx. linearly until the volume increased
very suddenly and the system exploded with the error "bad global
exclusion count." However, if I ran a shorter simulation and used the
restart files to begin another simulation, although the volume was
initially the same as the final volume of the previous simulation, it
quickly decayed to a value that was close to the initial value of the
previous simulation. Then the volume began to increase again as the run
continued. I thought this behavior might be due to the reinitialization
of the PME grid or patch grid. It's the only thing I can think would
change on restart. However, I never got any messages of "margin
violations". Also, even for fairly short simulations, in which atoms on
the DNA molecule have made only moderate changes in position (and so the
grids should still be OK), I still see this behavior.
This is why I started looking at NVT simulations of the same system.
The pressure just keeps increasing, or jumps back down on restart. I
then fixed the DNA in an NVT run. It looks like the pressure is not
increasing, but I need to run a longer simulation to make sure. I'm not
sure how exactly the pressure is calculated in the presence of the
field. There is some vague remark in the pressure control section of
the manual that the calculation is slightly different.
Anyway, please let me know if you've had similar experiences or any more
insight. Thanks,
Todd
-----Original Message-----
From: Brian Bennion [mailto:bennion1_at_llnl.gov]
Sent: Tuesday, November 28, 2006 3:06 PM
To: Todd Trimble
Cc: namd-l_at_ks.uiuc.edu
Subject: Re: namd-l: pressure calculation in applied field
Hello Todd,
This sounds like a difficult problem (which is why no one has
responded) and appears at first blush to be similar to problems I see
in my membrane simulations
where the box continually expands in one dimension until it crashes
due to pme grid/small patch errors.
I am using namd2.6 which version are you using? You probably stated
this in email #1 but for my sake will you give the machine and
version specifics again?
Thanks
Brian
At 10:36 AM 11/28/2006, you wrote:
>Hi All,
>
>Well I haven't got any response so far, but maybe the 10th time is the
>charm.
>
>I'm running an NPT simulation of an ssDNA molecule tethered at one end
>and surrounded by water and counterions. I apply an electric field as
>well. If my run has a large number of steps the box will explode but
if
>I break up into multiple runs it looks OK. The volume of the box in
any
>run increases pretty linearly up until failure. I figured the pressure
>control might be the problem, so I ran NVT simulations. I expected the
>pressure to be pretty constant, although the change in DNA conformation
>should affect the pressure through the interparticle force part of the
>virial. Well, the pressure showed a systematic increase and doesn't
>level off. Even worse, when I use the binary restart files in a new
>simulation, the pressure drops way down again. How can this be when
the
>virial depends only on the coordinates and the velocities and these are
>identical to those at the end of the previous simulation? These
>problems with pressure calculation must be responsible for my exploding
>NPT system. The calculated pressure keeps increasing, which forces the
>volume to increase in order to maintain constant pressure until the
>system is unstable.
>
>Is the pressure calculation not reliable in the presence of an applied
>field or forces?
>
>Thanks, Todd
>
>**********************************************************************
>Dr. Todd M. Trimble
>Center for Applied Nanobioscience at the Biodesign Institute
>Arizona State University
>
>todd.trimble_at_asu.edu
>work: 480-727-0435
>fax: 480-727-8283
>**********************************************************************
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:42:52 CST