# MD for nanocavities filling with melt polymer: Feasible?

Dear all,

I am a PhD student and my research is focused on how a melted plastic fills a nanocavity using injection moulding manufacturing process.

Until now, I have carried out simulations using CFD solvers for holes with sizes above the 100nm. However, I’m currently interested in going further and simulating for smaller dimensions. I am aware that for such smaller dimensions CFD wouldn’t be appropiate and I would need a MD approach.

My question is the following:

• Is it possible to study via MD how a polymer goes into a hole due to a defined pressure profile and cools due to heat transfer with the walls until reaches the no flow temperature?

My concern is that, since I have a poor knowledge of MD, I could start studying something that is somehow not feasible.

I would appreciate your experienced point of view of what I expect to do: It is feasible, it has no sense, I should go through other path, etc.

Thanks a lot.

The chief Q is what length scale (# or atoms or coarse-grained
particles) and what time scale you need. If it is
too big, MD cannot do your problem.

Steve

Dear Steve,

my domain would be around 10nmx10nmx30nm. Taking into account a C diameter of 0.3nm I estimate about 12000 atoms in the domain.

Concerning the time scale, I would need to simulate until the polymer freezes, what happens with approx. 1e-4s.

Am I inside the MD capabilities? Is it possible to simulate the solidification of a polymer via MD? My purpose is to know how deep the polymer goes into the cavity before it solidifies and stops its movement.

Thanks for this initial orientation, it is important for me in order to know whether MD is a good path for my problem or not.

If it’s an all-atom model then you have 10^4 atoms and
a fm sec timestep for 10^11 timesteps. A simple LJ
potential is about 1 us per atom per step on a single
CPU core (see Benchmarks
page on LAMMPS web site for a table of these timings
for many potentials). So you can do the math from there.

Steve

If it's an all-atom model then you have 10^4 atoms and
a fm sec timestep for 10^11 timesteps. A simple LJ
potential is about 1 us per atom per step on a single
CPU core (see Benchmarks
page on LAMMPS web site for a table of these timings
for many potentials). So you can do the math from there.

there are two more issues to take under consideration here:
1) you'll also have to include explicit atoms to model the cavity and
its environment, so you can increase the number of atoms in your
estimate by a significant chunk (3x? 5x?).
2) you have to use thermostatting of the outer environment (i.e. the
outer part of the cavity atoms), to move away kinetic energy, i.e. to
model that the cavity is embedded in a macroscopic system. the
settings of that thermostat will have a significant impact on the time
until polymer freezes. on one hand, you use this to accelerate
freezing and reduce the total simulation time. on the other hand,
you'll have to do some calibrating to match this with macroscopic
data. melting/freezing processes on the atomic level are not always as
straightforward as for macroscopic systems.

axel.