DPD/Tstat:Ineffective temperature control

Dear LAMMPS users and developers,

I am encountering a temperature control issue with dpd/tstat for a non-equilibrium system.

It’s a very big system with almost 0.6 million particles(sorry I have to go for this big). The shear rate I applied is erate=0.05, and the compute temp/deform is invoked to remove the streaming velocity.The dpd/tstat friction can be tuned to realize the purpose, but it might be unphysical though… For a similar system with smaller system size, the temperature is well-controlled.
If needed, I can upload the results as well.

I uploaded my input file, and output file. The coord. data file seems to be too big to to be accepted by server.
Would someone please take a look and provide some ideas for this situation?

Thanks very much in advance.

input_s005nonp (843 Bytes)

output_nonp (87.8 KB)

I am encountering a temperature control issue with dpd/tstat for a
non-equilibrium system.

No one is going to run your system. You have to explain your problem.


Hi Steve,
Sorry for the confusion. The problem is that the temperature is not well controlled, even after the deduction of streaming velocity. However, this problem is not seen for a much smaller system with the same condition, thus I wonder if there is a ceiling for the system size so that dpd/tstat will work well, and how to resolve it if I really need such a big system.

Thanks a lot!

Did you check what happens if you do not apply the box deformation?
Is the temp well controlled then? This will tell you if the problem
comes only from your system size.

Hi Carlos,

Thanks for your advice. I ran a system without any shear applied, and the temperature is well controlled (at kbT = 0.5). It takes longer to reach the target temperature as expected due to the system size. I copied a snapshot of screen output here.

So the problem is actually caused by the shear flow.
Thanks again for your help!



Yes, no surprise there with DPD taking longer to equilibrate the system. As the damping term in DPD depends on how large is the projection of the relative velocity of the pair of particles along the relative particles’ distance, this thermostat will act in a “softer way” on the modes that imply concerted motions of atoms, i.e, longer wavelength modes.
See for example the extreme case of two atoms moving at the same speed in the same direction. There \delta v_ij = 0 and thus the DPD damping (without the random term) will have no effect. Or when \delta v_ij is perpendicular to \delta r_ij, which could be seen as a coupled translation of the center of mass (if v_i != v_j) plus a rotation about it. There again DPD will damp nothing. Pure translations and rotations are not affected by the damping term.

Your problem with the shear flow could have several origins. I recommend you read the following paper:
I am convinced it will provide you plenty of guidance on how to tackle the issues you are facing.