ERROR: Computed temperature for fix temp/rescale cannot be 0.0

Dear All,

I was using “compute temp/profile”

as a temperature compute for the “fix temp/rescale” in order

to ensure the rescaling of the unbiased velocities and additionally in order to calculate the temperature profile.

My simulation is split in two runs (first of 2000001 and second of 6000000 steps). During the 1st run everything was fine and temperature of the system was around the desired 235K.

As the second run starts the temperature of the system get completely wrong values and after several hundreds of steps LAMMPS outputs an error.

I would appreciate a lot any hint regarding what might went wrong.

Here is the input script

units metal
dimension 3
atom_style atomic
boundary p p p
neighbor 2.0 bin

read_data data.dd

pair_style dd 4

pair_coeff 1 1 1 1

fix 1 all nve
velocity all create 235.0 4928459 rot yes dist gaussian
compute tz all temp/profile 0 0 1 z 200 out bin

fix 2 all temp/rescale 100 235.0 235.0 10.0 1.0
fix_modify 2 temp tz

timestep 0.001
thermo_style custom step pe ke etotal c_tz

Thermolize

thermo 10
run 2000001

fix 5 all ave/time 10 600000 8000000 c_tz mode vector file temp_prof

Run a simulation

run 6000000

Thanks a lot!

Anton

Dear All,

I was using "compute temp/profile"
as a temperature compute for the "fix temp/rescale" in order
to ensure the rescaling of the unbiased velocities and additionally in order
to calculate the temperature profile.
My simulation is split in two runs (first of 2000001 and second of 6000000
steps). During the 1st run everything was fine and temperature of the system
was around the desired 235K.

As the second run starts the temperature of the system get completely wrong
values and after several hundreds of steps LAMMPS outputs an error.

I would appreciate a lot any hint regarding what might went wrong.

there are two things wrong independently from the problem you are
seeing. a) you don't -provide a complete input. b) you use
temp/rescale in a production calculation.

axel.

Why using temperature rescaling for the production run is wrong? I understand that temperature rescaling ruins the dynamics but the average distribution should not be affected. At this point I wanted just to have some thermostating without bias. Do mean that temp/rescale cannot be used together with compute temp/profile?

Please find the full input in the attachment.

data.Hg.dp.N1000.lxlz.d13.69.relaxed (49.3 KB)

in.nodd (1.35 KB)

Why using temperature rescaling for the production run is wrong? I understand that temperature rescaling ruins the dynamics but the

Not just that.

average distribution should not be affected. At this point I wanted just to have some thermostating without

But why use a bad thermostat when there are better options?

bias. Do mean that temp/rescale cannot be used together with compute temp/profile?