Re: Mysterious Formation of Solvent Spheres

From: Charles Zhao (
Date: Wed Aug 11 2010 - 15:23:53 CDT

That's an interesting point. I was following the instructions provided here:
 , which, among other things, tells you to turn it off if using a water box
(which I am). because of the possibility of the system becoming very flat in
one dimension. would make sense if that were the issue. I'll think
about it, perhaps perform a trial run.

From: "Bennion, Brian" <>
To: Charles Zhao <>
Sent: Wed, August 11, 2010 10:31:46 AM
Subject: RE: namd-l: Mysterious Formation of Solvent Spheres


If it is an NPT ensemble that you want to use then I am curious as to why the
"useflexiblecell" is disabled. The langevin piston barostat is useless with
this option disabled, from my experience.


-----Original Message-----
From: [] On Behalf Of
Charles Zhao
Sent: Wednesday, August 11, 2010 9:42 AM
To: Axel Kohlmeyer
Subject: Re: namd-l: Mysterious Formation of Solvent Spheres

Thanks for the response.

Well, I did spend some time to go back and recheck my configuration files, so
I'm pretty certain there's no copy/paste errors at least.
That being said, it was NPT ensemble. Since production simulation did not
include any pulling, etc. the overall procedure was just 5000 step minimization,
followed by slow heating to 310 K and extended equilibration. Perhaps it was
the fact that I fixed some atoms (which I have not done in the past but I did do
for the simulations that didn't become droplets)? Or maybe the slow heating

A sample equilibration file for one of the systems that ended up being water
# input
coordinates Fullrancon.pdb
structure Fullrancon.psf
parameters par_all27_prot_lipid_na.inp
paratypecharmm on
fixedAtoms on
fixedAtomsFile Fullranconfixed.pdb
fixedAtomsCol B
# output
set output "Fullrancon"
outputname $output
dcdfile ${output}.dcd
xstFile ${output}.xst
dcdfreq 5000
xstFreq 500

binaryoutput yes
binaryrestart yes
outputEnergies 100
restartfreq 1000

# Basic dynamics
exclude scaled1-4
1-4scaling 1
COMmotion no
dielectric 1.0

# Simulation space partitioning
switching on
switchdist 10
cutoff 12
pairlistdist 14

# Multiple timestepping
firsttimestep 0
timestep 2
stepspercycle 10
nonbondedFreq 1
fullElectFrequency 2
rigidbonds all

#Periodic Boundary Conditions
cellBasisVector1 240 0 0
cellBasisVector2 0 225 0
cellBasisVector3 0 0 250
cellOrigin -124.77 -155.36 -67.71
wrapwater on
wrapAll on

set temperature 310
temperature $temperature; # run temperature

#Particle Mesh Ewald Electrostatics
PME yes
PMEGridSizeX 240
PMEGridSizeY 225
PMEGridSizeZ 250

# Constant Pressure Control (variable volume)
useGroupPressure yes ;# needed for rigid bonds
useFlexibleCell no ;# no for water box, yes for membrane
useConstantArea no ;# no for water box, maybe for membrane

langevinPiston on
langevinPistonTarget 1.01325 ;# pressure in bar -> 1 atm
langevinPistonPeriod 100. ;# oscillation period around 100 fs
langevinPistonDecay 50. ;# oscillation decay time of 50 fs
langevinPistonTemp $temperature ;# coupled to heat bath

# Constant Temperature Control
langevin on ;# langevin dynamics
langevinDamping 2. ;# damping coefficient of 5/ps
langevinTemp $temperature ;# random noise at this level
langevinHydrogen no ;# don't couple bath to hydrogens

# Minimization
minimize 5000
set freq 20

for {set i 0} {$i <= $temperature} {incr i} {
langevinTemp $i
reinitvels $i
run $freq
# Scripting
run 2500000


From: Axel Kohlmeyer <>
To: Charles Zhao <>
Sent: Wed, August 11, 2010 3:23:36 AM
Subject: Re: namd-l: Mysterious Formation of Solvent Spheres


i don't think that there is something mysterious going on.

computer programs are not very smart and only do
what they got told to do (unless they have bugs).
for something that is used as much as NAMD, the chances
are _much_ higher that you didn't tell it what you
wanted it to do. this is also known as the GI-GO principle.

On Wed, Aug 11, 2010 at 2:32 AM, Charles Zhao <> wrote:
> I constructed a group of similar solvated protein systems in periodic box
> boundary conditions using the exact same procedure. Then I ran them all in
> namd 2.7b2 for 10ns using virtually identical configuration files (I
> literally copy/pasted, changed the filenames, and changed the periodic box
> sizes to properly match the system).

please note, that copy-n-paste is a frequent source of errors.
it is very easy to miss a character here or there.

> The problem is, now I have some systems where the solvent congeals into this
> sphere shape (with a few waters flying off into the distance), and others
> where it stays in the proper shape of a box. The only thing I can think of
> is that the water density is somehow too low in the sphere systems. But
> then why is it perfectly fine in the remaining systems? They're all of

who says that it is perfect in the other systems? perhaps they are
just in a metastable configuration. have you calculated the total
density in those systems?

> similar size, differing perhaps by 30 or so angstroms in various dimensions,
> and constructed by the same method. And since when does water in a periodic
> box system, no matter how low the pressure, decide to form a sphere in the
> middle?

water does this since there are three phases: solid, liquid and gaseous.
if your water potential can reproduce those, it can form droplets,
periodic boundaries or not. TIP3P definitely can form droplets.

your description is lacking detail, so it is very difficult to assess
what is going wrong, but it almost looks as if your equilibration
protocol seems to be way off, and that most likely _all_ your
simulation results may be useless. please let us know in more
detail what steps you performed in order to equilibrate your
initial configurations for the production simulations. in what
ensemble were you running when you observed the formation
of the water droplet? nvt, nve, npt?



Dr. Axel Kohlmeyer
Institute for Computational Molecular Science
Temple University, Philadelphia PA, USA.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:54:24 CST