I am trying to model the flow development within a nanochannel. I need to control the inlet velocity which I suppose I can do by adding a force on a small portion of the atoms at the start of the channel.
Of course, the first thing that comes to mind, is to have periodic boundary conditions in the direction of the flow. However, I will be running sub-sonic cases information from the inlet will travel downstream and affect the development of my flow (and vice versa). I need my inlet and outlet conditions to be independent. Furthermore, the channel width might be changing along the direction of the flow so again, periodic boundary conditions will not work!
Is there any way around such issue? For example, could I have a reservoir feeding atoms to my inlet (at a specific mass flow) and then simply evaporating any atoms that leave through the outlet (although this sounds like an overly complicated process)?
I am trying to model the flow development within a nanochannel. I need to
control the inlet velocity which I suppose I can do by adding a force on a
small portion of the atoms at the start of the channel.
Of course, the first thing that comes to mind, is to have periodic boundary
conditions in the direction of the flow. However, I will be running
sub-sonic cases information from the inlet will travel downstream and affect
the development of my flow (and vice versa). I need my inlet and outlet
conditions to be independent. Furthermore, the channel width might be
changing along the direction of the flow so again, periodic boundary
conditions will not work!
Is there any way around such issue? For example, could I have a reservoir
feeding atoms to my inlet (at a specific mass flow) and then simply
evaporating any atoms that leave through the outlet (although this sounds
like an overly complicated process)?
what you want to do *is* complicated. i don't think there is a way
around that. having a large (enough) reservoir and moving atoms
through the channel at a fixed rate is the simplest approach. you can
define a group of atoms at the outermost region of the reservoir and
move it like a piston at constant speed with fix move and then simply
have a fixed boundary after the outlet and use the thermo_modify lost
ignore command to continue the simulation, when the atoms continue
beyond that boundary. the downside of this approach is, that you have
a limited time that you can run the simulation.
if you want to have a system at a steady state, you *have* to use
periodic boundary conditions *and* a large (enough and wide enough)
reservoir, where you have multiple zones: 1) atoms are moved away from
the outlet, 2) atoms are slowed down and re-equilibrated to a desired
temperature using a dissipative thermostat, 3) inlet flow conditions
are set up using a suitable (smaller) force.
You can also look at the doc pages for fix append/atom and fix pour.
Neither probably do what you are asking, but they are similar
in the sense they insert atoms (continuously) at one end of the
simulation, then you can just delete atoms as they leave the other end.
And do this in for a non-perioidic box.
What you are asking for subsonic BC I have seen for gases,
but not for liquids (or solids). The tricky part for LAMMPS for
a high density material is how to insert randomized atoms
at high enough density w/out overlaps. Fix append/atoms
gets around this by using a lattice (for a solid). Fix pour
is more random, but just simple random, meant for granular
particles falling under gravity.
But possibly one of those could be modified to do what
you want by you or someone else.
Thanks for your response. Indeed this sounds like a good option. So, I can append the atoms on a crystal lattice with the lattice constant corresponding to the desired liquid density. These atoms can be inserted in my inlet where a force acts on them in the direction of the flow and then have another smaller, opposite force at the outlet to emulate a pressure driven flow. Does that sound right?
The only problem is that I need to insert the same number of atoms as the ones being deleted.
Regardless what method you use, you will have to construct a type of
filter that will prevent passage of all pressure disturbances, while
transmitting atoms, fully equilibrated, at the desired flowrate.
This filter could be be a strongly-coupled Langevin thermostat with a
specified streaming velocity, as Axel suggested. The filter length
could be quite small, a few multiples of the force cut-off, I think.
This will work best for a periodic system, where you will only have a
single filter region that will both absorb outlet particles and
provide a steady supply of inlet particles. I think a non-periodic
approach combining fix append and atom deletion could be made to work,
but it would require two filters, and seems to create a lot of extra
complications, like having to melt the crystal before entering the
inlet filter and deleting atoms after they exit the outlet filter.