Constant pressure MD problems (sander)

From: Pratul Agarwal <>
Date: Mon 12 Aug 2002 07:45:15 -0700 (PDT)

A short summary of the problem first:
Problems with constant pressure MD (with slow heating by coupling to
bath), the sander simulation either crashes or the system expands with
unreasonable density.

Full description of the problem:
System - a protein complex solvated in a TIP3 waterbox.
Size - ~19000 atoms (~6000 solute atoms + 6200 water molecules)
AMBER v. - 7.0
Hardware - IBM pSeries (Power 4)
OS - AIX 5.1
Parallel - YES (16 nodes)

After I have solvated the protein complex (tLeap), I have performed
minimization of the protein only and then water only. The next step I
want to perform is to remove the "bubbles" and get the system to a
reasonable density. For this, I plan to do a 25 ps (25000 steps in 1 fs
steps) simulation at constant pressure (NTP=1) with a coupling to bath
(NTT=1). I found that my job was bombing out with an a short error

EWALD BOMB in subroutine ewald_list
volume of ucell too big!!
Any ideas why the job is failing? How can I get the system to a
reasonable density?
I tabulated this behavior with various values of 'tautp' and 'taup'. In
some cases the job does complete but did not give a reasonable density.
Note: Starting density of system 0.8444 and the target was 25000 steps,
the number in parenthesis after Bombs indicates the number of steps
after which the job bombs.
Also note that where the job completes the RMS fluctuations in pressure
are too high.
                   Bombs or   Ending   Pressure
SIM  tautp  taup  Completes   Density    RMS
 1   0.2    0.2   Bombs(1304)  0.6724     -
 2   0.2    0.5   Bombs(2150)  0.7044     -   
 3   0.2    1.0   Completes    0.8424   7333.7
 4   0.5    0.2   Bombs(1841)  0.6744     -
 5   0.5    0.5   Bombs(1930)  0.6678     -
 6   0.5    1.0   Bombs(2475)  0.7015     -
 7   1.0    0.2   Bombs(2780)  0.6422     -
 8   1.0    0.5   Bombs(2725)  0.6941     -
 9   1.0    1.0   Completes    0.8031    779.0
Another thing which I noticed in the output files is that there are
"vlimit resets". The velocities  component are extremely high -- upto a
million times more than the default limit of 20. I thought maybe this
is a "cold solute/hot solvent" problem, but I see that the option NTT=5
has been discontinued after AMBER v 5.0.
I will appreciate any help.
Do You Yahoo!?
HotJobs - Search Thousands of New Jobs
Received on Mon Aug 12 2002 - 07:45:15 PDT
Custom Search