When lamda < 0.6* hgrid, we do the shifting of srd cells using the option shift.
In this case the output temperature of the system is increasing with the number of processors used.
Please see the attached plot, where the srd temperature is plotted with the number of steps for number of processors 1, 2, 4 and 8. The lammps script is also attached.
To do the grid shifting we communicate the coordinates of srd particles among the processors.
The temperature dependence on the number of processors is suggesting there could be a possible bug in the shift algorithm.
To satisfy the criteria lamda > 0.6* hgrid I have used a very small mass for srd particles and in this case shifting is not needed. Now the temperature is in agreement with the target temperature and independent of the number of processors used.
So I am wondering if there is a possible bug in the shift algorithm.
no - have you run the examples/srd input scripts and seen
if the problem exists in those? If not, what is different
about your system vs those examples?
The script is very basic and simple. I just modified the system parameters to have values used in your JCP papers.
So the problem exist in the example scripts too. Without shifting the grids the temperature is fine. The temperature is
increasing with the number of processors used when grid shifting is done. So, I think the problem must be in the “shift” algorithm, where we communicate the coordinates of the atoms between the processors.
Below is the script. Both the script and temperature are attached the first email.
Dan looked into this - we actually had a comment in the code that
we were communicating incomplete info for shared bins (on processor
boundaries) to insure perfect energy conservation. So try out
the latest patch (28Jan) and see if it fixes your problem.
Note that you should really be monitoring the bin-wise SRD temperature,
which is the 9th fix srd output, rather than the kinetic energy of
the entire system (global temp), since the former is what is actually
being thermostatted and conserved. Also if you are using SRD with
big particles, you should definitely use a thermostat, whether bin
shifting or not. There are other issues associated with mixtures
that can influence temperature drift.