Equilibration in NVT MD run

[Raghav]

Dear cp2k users,

                        I am trying to equilibrate my system. I am doing NVT simulation but the mean temperature of my system is not reaching the target Temperature (310K) as can be seen in the figure attached. I am using Nose Hoover Thermostat, Is there some problem with the parameters of the thermostat or something else is the problem.

Input file is attached.

Regards

Raghav

[M. Brehm]

Dear Raghav,

I could imagine that this effect is due to the "TEMP_TOL   100" line. Please try to comment out this line.

Apart from that, a thermostat time constant of 100 fs is relatively weak for an early equilibration. If there is much change in the structure of the system, the coupling of the thermostat might simply be too weak. I recommend to use a lower value (e.g. "TIMECON 20") in the early equilibration stages. The value of 100 fs is fine for production runs or late equilibration.

One more comment: Your time step of 1.0 fs is fairly large for a system which contains hydrogen atoms. This is always a matter of taste, but I would go for 0.5 fs if hydrogens are involved. Otherwise, the integration error might become significant, leading to a drifting conserved quantity.

Best regards,
Martin

[raghav saxena]

Hey Martin,

                I removed the TEMP_TOL=100 line and changed the time constant to 20 and I got this but the mean temperature is still not 310 even after 3.4 ps. Any suggestions?

Regards

Raghav

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

[Lucas Lodeiro]

I am not an expert in that, but i can suggest you some things:

The timestep is not the ideal for X-H bonds, less for gas phase molecules, since the oscillation frecuency is bigger. I would put 0.5 fs.

Other important thing is, the temperature is not well defined for one step (is a statistical property), less if the sistem just have 50 atoms, maybe that do not allow a "fast or regular" convergence of temperature.

Also, you are not using a PBC system, i do not know if the kinetic energy, and therefore the temperature, is well defined.

In my case, I usually use a liquid system (300 atoms) and PBC conditions, with the same nose-hoover parameters, except for TIMECON=50, but I use a clasical Molecular Dinamics first the ab-initio MD to obtain a better guess, with a better correspondence between the kinetic energy and the potential energy (positions); those systems reached the reference temperature and equilibrium in less than 1 ps (with the tipical oscilation around the reference temperature, but well matched the average value). I feel for the temperature plot, and the initial molecular geometry, you use an optimiced/relax potential energy geometry, and that can make it longer time to reached the reference temperature, because the thermasthat must inject a great amount of kinetic energy to reach the equilibrium.

Sorry if it is not helpful. Regards