How can I update the position of a wall throughout the simulation?

[aldorvasq]

Dear HOOMD users,

I am trying to implement a movable wall in my current simulation (HOOMD 4.4.1). As of now I am trying to implement this by using hoomd.custom.Action and I can not find the right way to use it. I am basing most of the code I use (shown below) from the example provided in "How to continuously vary potential parameters" and from a discussion post from HOOMDv3 ("Compressing Walls in HOOMD 3") where it is suggested to update the walls' position using:

wall_potential.walls = [# new walls as a function of the current simulation state, timestep, etc...]

I have tried to implement that, but I get a RuntimeError: Buffer is full message which I am not sure if its due to my system being ran in a newer version of HOOMD or to something else.

Here is a snippet of my attempted implementation, where I employ two walls (planes) parallel to each other separated by 70 units of distance in the x-direction. I attempt to move the left wall towards the right (+x-direction) 10 units of distance during a time span of 20,000 timesteps.

left_wall = hoomd.wall.Plane(origin=(-35, 0, 0), normal=(1, 0, 0))
right_wall = hoomd.wall.Plane(origin=(35, 0, 0), normal=(-1, 0, 0))
lj_wall = hoomd.md.external.wall.ForceShiftedLJ(walls=[left_wall, right_wall])
lj_wall.params[['W']] = {"epsilon": 1.0, "sigma": 1.0, "r_cut": 2.0 ** (1 / 6)}
integrator.forces.append(lj_wall)

sim.operations.integrator = integrator
# Step 1: Subclass hoomd.custom.Action.
class WallParameterModifier(hoomd.custom.Action):
    def __init__(self, lj_wall):
        super().__init__()
        self.lj_wall = lj_wall

    def act(self, timestep):
        wall_position = -35 + timestep * 0.0005 
        left = hoomd.wall.Plane(origin=(wall_position, 0, 0), normal=(1, 0, 0))
        self.lj_wall.walls = [left, right_wall]

# Step 2: Create a hoomd.update.CustomUpdater
wall_parameter_modifier = WallParameterModifier(lj_wall)
wall_parameter_updater = hoomd.update.CustomUpdater(
    trigger=hoomd.trigger.Periodic(1), action=wall_parameter_modifier)

# Step 3: Add the updater to the operations
sim.operations.updaters.append(wall_parameter_updater)

I also tried different things in the act(self, timestep) function and the simulation does run, but in everything I tried the walls remain fixed in the initial positions they were assigned (hoomd.wall.Plane(origin=(-35, 0, 0), normal=(1, 0, 0))). Here is one of the things I tried:

    def act(self, timestep):
        wall_position = -35 + timestep * 0.0005 
        left = hoomd.wall.Plane(origin=(wall_position, 0, 0), normal=(1, 0, 0))
        self.lj_wall = hoomd.md.external.wall.ForceShiftedLJ(walls=[left, right_wall])

Could someone point in the right direction as to how implement a movable wall in HOOMD?

I will leave a front-facing image of the simulation box (80x60x60) I am using, where I left extra space in the x-direction to be able to test if the walls were working.

Thank you,
Aldo V

[joaander]

Use:

    def act(self, timestep):
        wall_position = -35 + timestep * 0.0005 
        left = hoomd.wall.Plane(origin=(wall_position, 0, 0), normal=(1, 0, 0))
        self.lj_wall.walls[0] = left

Your original script should have worked. I am able to reproduce the Buffer is full error message but did not see an obvious reason for the error. I have opened #1879 to track the defect.