system temperature too high?

Hi all,

I simulated a collision system with two groups. One is the active region inner, the other “side” is the thermostat region outside the system.
And group s1 is one of the atoms in the active region. The thermostat region is to absorb the inner energy of the system during the collision.
The partial input script is as follows.

group side …
group s1 …
fix 3 side nve
compute ru side temp
fix 4 side temp/rescale 10 300 300 5.0 1.0
fix_modify 4 temp ru
velocity s1 set 0.0 1000 0.0
dump all cfg 100 cu16/cu.*.cfg id type xs ys zs
#thermo_modify temp
timestep 0.001
run 10000

The running result is the following when close the themo_modify command.
Step Temp E_pair E_mol TotEng Press Volume
110 52084.603 -542881.98 0 342524.48 555533.73 1829687.8
120 52091.315 -542996.13 0 342524.42 555068.5 1829687.8
130 52102.062 -543178.92 0 342524.32 554622.1 1829691.4
140 52112.125 -543350.05 0 342524.25 554333.76 1829691.4
150 52119.66 -543478.23 0 342524.17 554239.94 1829691.4

The running result is the following when open the themo_modify command.
Step Temp E_pair E_mol TotEng Press Volume
110 300 -542881.98 0 -542266.92 39018.649 1829687.8
120 300 -543004.85 0 -542389.79 38490.658 1829687.8
130 300 -543213.99 0 -542598.93 37884.08 1829687.8
140 300 -543423 0 -542807.94 37387.837 1829691.9
150 300 -543603.09 0 -542988.03 37041.153 1829691.9

My question:

  1. The system temperature was stable when open the themo_modify command.
    The system temperature was so high, up to 52000 K, when close themo_modify command. Why?
    Didn’t the thermostat play a role?
  2. I can’t seemingly see the collision phenomenon. Where is the problem of the above settings?

Any suggestions are welcome.
Thanks.

Best regards,
Pan

global temp computation include all velocity contribution in kinetic energy.

Just use and read about “compute temp/com” command.

06.03.2012, 13:47, “Hua Pan” <phq323@…24…>:

the result using compute temp/com command is the same as the compute temp command.
1220 52074.233 -543624.17 0 341606 554125.59 1829760.3
1230 52073.818 -543631.71 0 341591.4 554051.16 1829760.3
1240 52074.15 -543637.36 0 341591.4 553922.99 1829760.3
1250 52074.122 -543636.88 0 341591.4 553758.21 1829746.6
1260 52073.742 -543630.42 0 341591.4 553557.54 1829746.6

2012/3/6 Maxim <esmax@…2375…>

if you off thremo_modify you are see a global temp.

06.03.2012, 15:36, “Hua Pan” <phq323@…24…>:

the result using compute temp/com command is the same as the compute temp command.
1220 52074.233 -543624.17 0 341606 554125.59 1829760.3
1230 52073.818 -543631.71 0 341591.4 554051.16 1829760.3
1240 52074.15 -543637.36 0 341591.4 553922.99 1829760.3
1250 52074.122 -543636.88 0 341591.4 553758.21 1829746.6
1260 52073.742 -543630.42 0 341591.4 553557.54 1829746.6
#thermo_modify temp
the results have not changed the same as above.

2012/3/6 Maxim <esmax@…2375…>

default calculation of global temperature(that you see with off thermo_modify) does not depend on your script. if the system contains velocities not related to temperature it’s may be not physical temperature value of system.

and may be you need to see visualization for checking.

06.03.2012, 16:34, “Hua Pan” <phq323@…24…>:

hi hua,

Hi all,

I simulated a collision system with two groups. One is the active region
inner, the other "side" is the thermostat region outside the system.
And group s1 is one of the atoms in the active region. The thermostat region
is to absorb the inner energy of the system during the collision.

first off, doing that with temp/rescale is very unphysical,
since temp/rescale is a horrible model for connecting a
finite system to a bulk for exchange of kinetic energy.

second, you need to take a step back and look at what
"temperature" means in the case of a finite size classical
MD simulation. it is based on the equipartitioning theorem
and thus derived from the instantaneous kinetic energy,
the boltzman constant and the degrees of freedom.

if you model a projectile, you have to give it kinetic energy,
quite a lot actually, and that cranks up your overall
"temperature" for the reasons i just outlined. if you then
use temp/rescale it will also rescale the kinetic energy
of your projectile.

a proper strategy would be the following:

set up your projectile without a center of mass momentum.
similar for your impact surface. split that surface in three parts.
the lowest layer is fixed in space (can be just one layer of atoms),
the middle area (multiple layers) is coupled to a thermostat
(e.g. fix langevin) to emulate the thermal exchange with bulk.
on top of that you have a few layers of atoms without a thermostat.
those are coupled to the underlying atoms, which is the physically
correct representation.

now you equilibrate this system.

then you add the center of mass velocity to your projectile.
(whether you initially thermostat that or not is probably of
little impact, unless you want it to be very hot, e.g. molten)
and keep integrating its equations of motion with plain nve.

the global temperature of this system is going to be
very high. that is just the way it is. the only thermostat
to care about is the one that is supposed to dissipate
the kinetic energy that would go into the bulk of your
target area.

i suggest you ponder this a little bit, have a look at
your book on statistical mechanics about the relation
between kinetic energy and temperature and *after*
you feel that you are no longer fishing in the dark,
you set up your simulation again. perhaps it also
helps to discuss with some colleagues or an
adviser. that is what they are for!

good luck,
    axel.