Dear Sirs,
I would like to ask how I could implement continuous deformation by box
change in a parallel replica dynamics simulation. I have tried fix
deform (not allowed according to manual), but it gives unexpected
discontinuities in strain rate after quench stages.
Thank you,
Dear Sirs,
I would like to ask how I could implement continuous deformation by box
change in a parallel replica dynamics simulation. I have tried fix
please explain how it should be possible to use a method that is
designed to enhance sampling in a system that is in equilibrium
together with a non-equilibrium method. what is the rationale, that
would make this a viable approach?
axel.
Hello,
in the manual the prd method is described as applicable for 'performing accelerated dynamics that is suitable for infrequent-event systems that obey first-order kinetics'. Along these lines I would like to use it for simulating cracking of crystalline structures at slow deformation rates.
Thanks,
Osvalds
Hello,
in the manual the prd method is described as applicable for 'performing
accelerated dynamics that is suitable for infrequent-event systems that obey
first-order kinetics'. Along these lines I would like to use it for
simulating cracking of crystalline structures at slow deformation rates.
that is not answering my question.
axel.
Hello,
the viability, as I have tested it by normal simulations of identical
setting, would be based on being able to identify infrequent events in
crack propagation, in particular threshold changes in number of atom
neighbors.
Thanks,
Osvalds
One issue is probably that PRD goes thru a sequence of dynamics,
quench, dephasing, etc where it is advancing the timestep in
different ways and backing up. Fix deform is applying strain based on the
current timestep. For some of those stages you probably
want to turn off deformation, and also insure the timestep
and box deformation are consistent. So I can imagine that
fix deform is confused by what PRD is doing. There would
have to be some code added to make something like
that work. Ditto if you were using TAD with fix deform.
Aidan can probably comment.
Steve
For illustration I have attached 2 plots from a test simulation, which
ran without event occurrence until failure.
Thanks,
Osvalds
P_E.eps (22.8 KB)
V_e.eps (30.5 KB)
Hi Osvalds,
Even though the physical picture is one of continuous deformation, I suspect that your desired rate of deformation is much larger than the femtosecond timescale of MD simulation. Hence, there is no need for the deformation to be continuous within the MD simulation. I suggest you try constructing a simulation that alternates between moderate-length PRD runs and instantaneous box deformations., for which you can use the change_box command. LAMMPS scripting capabilities allow you to program the PRD and deformation stages within do/for loop that can be repeated an arbitrary number of times.
Aidan
Or you could likely do the same thing with continuous deformation. You would
need to do 2 things:
a) insure that PRD does not continue to apply deformations during de-phasing
and quenching.
This would likely require some additional logic be added to prd.cpp to allow
for turning off certain fixes at certain stages
b) write a new event detection compute which accounts for events occurring
in a system undergoing active strain. I imagine the vanilla compute we
provide could be broken by strain.
Steve