flying ice cube

Dears
I have a problem during equilibration of system.
I have a simulation box include 2 different crystal. I define 3 groups ( lower,upper,mobile) and I am using the rerax/c potential to simulate the simple tensile test.
as the following commands, I have “flying ice cube” problem in my second step of equilibration. also the lower group separate from mobile group when the “flying ice cube” problem occur.
at first equilibration step I used Langvin thermostat and I didn’t have any problem, but in second step I used berendsen thermostat. I know from Axel advises that this syndrome occurs more when using berendsen thermostat. I used NVT instead but the problem exist yet.
is there any suggestion to avoid this problem corresponding to my commands?
Thanks
Mohammad

boundary p p s

neighbor 3 bin
neigh_modify once no every 2 delay 0 check yes

fix boxrelax all box/relax x 0 y 0

min_style cg
min_modify dmax 0.1 line quadratic
unfix boxrelax

reset_timestep 0

compute new3d mobile temp
compute new2d mobile temp/partial 1 1 0

#-----------#Run-Step 1

velocity mobile create 300 12345 temp new3d mom yes rot yes dist gaussian

fix fixnve all nve
fix fixstep1 mobile langevin 300 300 100 699483
fix_modify fixstep1 temp new3d

fix upperforce upper setforce 0.0 0.0 NULL
fix lowerforce lower setforce 0.0 0.0 NULL

fix loweraveforce lower aveforce NULL NULL 0.0
fix upperaveforce upper aveforce NULL NULL 0.0

thermo 100
thermo_style custom step pe vol temp press lx ly lz
thermo_modify temp new3d
timestep 0.5
run 100000

unfix fixnve
unfix fixstep1

unfix upperforce
unfix lowerforce
unfix loweraveforce
unfix upperaveforce

reset_timestep 0

#-----------#Run-Step 2 #

fix fixnve all nve
fix fixstep2 mobile temp/berendsen 300 300 100
fix_modify fixstep2 temp new3d

fix upperforce upper setforce 0.0 0.0 NULL
fix lowerforce lower setforce 0.0 0.0 NULL

fix loweraveforce lower aveforce NULL NULL 0.0
fix upperaveforce upper aveforce NULL NULL 0.0

thermo 100
thermo_style custom step pe vol temp press lx ly lz
thermo_modify temp new3d

timestep 0.5
run 200000

Dears
I have a problem during equilibration of system.
I have a simulation box include 2 different crystal. I define 3 groups (
lower,upper,mobile) and I am using the rerax/c potential to simulate the
simple tensile test.
as the following commands, I have "flying ice cube" problem in my second
step of equilibration. also the lower group separate from mobile group when
the "flying ice cube" problem occur.
at first equilibration step I used Langvin thermostat and I didn't have any
problem, but in second step I used berendsen thermostat. I know from Axel
advises that this syndrome occurs more when using berendsen thermostat. I
used NVT instead but the problem exist yet.
is there any suggestion to avoid this problem corresponding to my commands?

have you tried to remove the residual translational momentum after the
first step? unless you have a dissipative thermostat like fix langevin
or some kind of position restraint, the drift of your system only
grows.

axel.

Dear Axel
Thanks for your reply.
As I understand from your email, I should use a method to zero the residual translational momentum,

because of I don’t use langevin thermostat in second step and that mean I should use “fix 1 all momentum 1 linear 1 1 1” when I use NVT or berendsen.

if I understand it Ok, I want to know is it necessary to use the “fix momentum” in tensile loading step ( by set same velocity to each lower & upper groups) and use berendsen or NVT?
Thanks
Mohammad

Dear Axel
Thanks for your reply.
As I understand from your email, I should use a method to zero the residual
translational momentum,
because of I don't use langevin thermostat in second step and that mean I
should use "fix 1 all momentum 1 linear 1 1 1" when I use NVT or berendsen.
if I understand it Ok, I want to know is it necessary to use the "fix
momentum" in tensile loading step ( by set same velocity to each lower &
upper groups) and use berendsen or NVT?

if your simulation is properly set up, fix momentum should not be needed.

you should only need to remove the residual translational momentum
once and that can be done with "velocity zero". if you still get a
drifting system, you need to find out which part of your setup is
introducing the drift. if you start a fix nvt simulation for an
initial configuration that has no translational momentum, then it
should not accumulate it.

axel.