[gmx-users] temperature shooting up

Mark Abraham Mark.Abraham at anu.edu.au
Wed Nov 8 03:40:50 CET 2006

Cherry Y. Yates wrote:
> Dear Mark,
> Thanks for your help, I followed you instruction, and I can get stable 
> temperature after I remove the temperature coupling. In this case I used 
> the flexible spc itp file. However if I use a rigid spc.itp,  after 
> remveing the temperature coupling, the system temperature is shooting up 
> from 300K to thousands degree, the total energy is going up also instead 
> of fluctuating. I attached both the itp files and grompp.mdp files. Let 
> me know what could be wrong. It is really wired to me.

The change in the model physics from flexible to rigid is severe one - 
almost all of the water molecules won't fit the rigid geometry, and will 
thus be in a high-potential-energy state. If the constraint-enforcing 
algorithm conserves energy, then that potential energy has to be 
converted to kinetic energy. Then if the simulation lacks temperature 
coupling, it is at constant energy and the temperature will be 
necessarily high. Actually you say you observe rising energy - this is 
probably because your timestep is too large to conserve energy at this 
very high temperature (and thus atomic velocity).

So the take-home lesson is not to do two perturbations of your model 
physics at one time. If you really want to start with a flexible water 
model with temperature regulation and end up with rigid water model at 
without temperature regulation, do the water transition with temperature 
coupling still on to give the energy somewhere to flow - to the external 
bath. I expect you'd need a short temperature coupling constant to make 
this work in practice. Once that's stable, *now* try taking off 
temperature coupling.

This suggests the question, why you'd want to do this flexible-rigid 
transition in an equilibration? It would be simpler to start in the 
rigid domain, and to do energy minimization to correct any geometries 
before starting MD simulations.


More information about the gromacs.org_gmx-users mailing list