numerical inaccuracy upon restart

From: Harald Tepper (h.l.tepper_at_amolf.nl)
Date: Wed May 19 2004 - 08:45:07 CDT

Hello,

Trying to get acquainted with NAMD, I did a test with restarting a
simulation. I first ran a simulation of 200 steps, and then two
consecutive simulations of 100 steps starting from the same initial
conditions. The first 100 steps in both cases give exactly the same
results, but from step 100 onwards they differ slightly numerically.
I would expect that on using binary restart files the numerical output
should exactly match the non-restarted simulation. Unfortunately I could
not find a definite answer to this in either the user manual or the
mailing list archive.
I also use tCouple for thermostatting and not the Langevin method, so I
would guess random number seeds would not be an issue.
Can anyone explain this?

Thanks very much in advance,

Harald Tepper
(attached are the three configuration files)

-- 
--------------------------------------------------------------------
      'Wat baten kaars en bril als de uil niet zien wil?'
Dr. Harald Tepper
FOM Institute for Atomic and Molecular Physics [AMOLF]
P.O.Box 41883
1009 DB Amsterdam
The Netherlands
Tel:   +31 (20) 6081389
Fax:   +31 (20) 6684106
mailto:H.L.Tepper_at_amolf.nl
WWW:   http://www.hec.utah.edu/~harald/
--------------------------------------------------------------------

# molecular system
structure ../Equilibrate/solvate.psf
cellBasisVector1 93.965 0.000 0.000
cellBasisVector2 0.000 80.356 0.000
cellBasisVector3 0.000 0.000 74.544

# force field
paratypecharmm on
parameters /usr/cof1/tepper/ForceFields/CHARMM/par_all_22_27.inp
exclude scaled1-4
1-4scaling 1.0

# approximations
switching on
switchdist 8
cutoff 12
pairlistdist 13.5
margin 0
stepspercycle 20

#integrator
timestep 1.0

#output
outputenergies 10
outputtiming 100
binaryoutput no

# molecular system
binaryrestart yes
Coordinates ../Equilibrate/solvate.coor
Velocities ../Equilibrate/solvate.vel
numsteps 200

#output
outputname run200
dcdfreq 500
restartfreq 500
restartname run200_restart

#protocol
tCouple on
tCoupleTemp 300


# molecular system
structure ../Equilibrate/solvate.psf
cellBasisVector1 93.965 0.000 0.000
cellBasisVector2 0.000 80.356 0.000
cellBasisVector3 0.000 0.000 74.544

# force field
paratypecharmm on
parameters /usr/cof1/tepper/ForceFields/CHARMM/par_all_22_27.inp
exclude scaled1-4
1-4scaling 1.0

# approximations
switching on
switchdist 8
cutoff 12
pairlistdist 13.5
margin 0
stepspercycle 20

#integrator
timestep 1.0

#output
outputenergies 10
outputtiming 100
binaryoutput yes

# molecular system
binaryrestart yes
Coordinates ../Equilibrate/solvate.coor
Velocities ../Equilibrate/solvate.vel
numsteps 100

#output
outputname run100_1
dcdfreq 500
restartfreq 100
restartname run100_1_restart

#protocol
tCouple on
tCoupleTemp 300


# molecular system
structure ../Equilibrate/solvate.psf
cellBasisVector1 93.965 0.000 0.000
cellBasisVector2 0.000 80.356 0.000
cellBasisVector3 0.000 0.000 74.544

# force field
paratypecharmm on
parameters /usr/cof1/tepper/ForceFields/CHARMM/par_all_22_27.inp
exclude scaled1-4
1-4scaling 1.0

# approximations
switching on
switchdist 8
cutoff 12
pairlistdist 13.5
margin 0
stepspercycle 20

#integrator
timestep 1.0

#output
outputenergies 10
outputtiming 100
binaryoutput yes

# molecular system
# continuing a run
binaryrestart yes
binCoordinates ../Equilibrate_TCouple_1bin/run100_1.coor
Coordinates ../Equilibrate/solvate.coor
binVelocities ../Equilibrate_TCouple_1bin/run100_1.vel
extendedSystem ../Equilibrate_TCouple_1bin/run100_1.xsc
firstTimestep 100
numsteps 200

#output
outputname run100_2
dcdfreq 500
restartfreq 100
restartname run100_2_restart

#protocol
tCouple on
tCoupleTemp 300

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:37:37 CST