Temperature is not constant in NVT simulation

LAMMPS LAMMPS Mailing List Mirror
1

Hello,

To give some detail of my message.

During the simulation, I print : Step Temp E_pair E_mol Tot_ Eng Press
53 600 0 27.672161 28.525278 3356.784
20000 638.4112 0 27.871209 28.778941 6776.9679
40000 666.05221 0 27.78253 28.729564 -833.93178
60000 486.86297 0 27.774174 28.466426 1578.8261
80000 520.00695 0 28.187929 28.927307 12075.273

I see that temperature change rapidly, each steps. Or I execute my simulation with NVT condition.
I tested my script in the simple system, Ge, and I don’t observe this behaviour. For Ge, I used the Tersoff potential.

You can also see a high change of the press in the supercell.
From these observations, I suggested a problem with the initial structure, but my structure has been relaxed by DFT calculation.
And I think, it is correct.

Do you know the cause of this problem “temperature behavior” ?

I saw 2 Warning in my log file:

WARNING: Inconsistent image flags (…/domain.cpp:756)
WARNING: Bond/angle/dihedral extent > half of periodic box length (…/domain.cpp:877)

From the previous talks, I can understand that these warnings don’t change the MD simulation.
I don’t know the cause of these warnings !

Regards,
Pierre-François.

Hello,

The structure for my system is cubic (6.50 Angstrum). time step = 1fs with equilibration 200ps and product phase of 20 ns (I stopped this latter).
I run with my force field: bond, angle, dihedral, van der wall. I checked the parameters for all functions.

I see a clear oscillation of the temperature during the simulation.

And?

You see the log file for my simulation.

Let me the error in this log file.

What error?

2

Hello,

To give some detail of my message.

During the simulation, I print : Step Temp E_pair
E_mol Tot_ Eng Press
                                                 53 600
0 27.672161 28.525278 3356.784
                                                20000 638.4112
0 27.871209 28.778941 6776.9679
                                                40000 666.05221
0 27.78253 28.729564 -833.93178
                                                60000 486.86297
0 27.774174 28.466426 1578.8261
                                                80000 520.00695
0 28.187929 28.927307 12075.273

I see that temperature change rapidly, each steps. Or I execute my
simulation with NVT condition.
I tested my script in the simple system, Ge, and I don't observe this
behaviour. For Ge, I used the Tersoff potential.

You can also see a high change of the press in the supercell.
From these observations, I suggested a problem with the initial structure,
but my structure has been relaxed by DFT calculation.
And I think, it is correct.

you may have a DFT optimized structure, but that is not likely
consistent with your classical force field, and only the latter
matters for dynamics.
more importantly, your system is *tiny* (12 atoms, IIRC), so it *has*
to fluctuate... a *lot*.

please note that the conventional concept of temperature about
macroscopic objects, i.e with effectively an infinite number of
particles.
in MD simulations, temperature is computed on the basis of
equipartitioning of kinetic energy and that will only work in case a
system is in equilibrium and you average over a long time. just take
the most extreme example: a harmonic oscillator. its kinetic energy
oscillates between 0 and 2x the average kinetic energy.

this is all basic stuff and has been discussed many many times on this
very mailing list

Do you know the cause of this problem "temperature behavior" ?

the main problem here is an incomplete knowledge of statistical mechanics.

I saw 2 Warning in my log file:

WARNING: Inconsistent image flags (../domain.cpp:756)
WARNING: Bond/angle/dihedral extent > half of periodic box length
(../domain.cpp:877)

From the previous talks, I can understand that these warnings don't change
the MD simulation.
I don't know the cause of these warnings !

http://lammps.sandia.gov/doc/Section_errors.html