Implementing a Mixed Active and Non-Active Brownian Particle Simulation in Hoomd-blue: Configuration and Resizing

[idop11]

Hello,

I am new to using Hoomd-blue and I am attempting to set up a simulation with the following specifications:
I aim to construct a rectangle with a height of 2L and a width of L. In the upper half of this rectangle, I plan to distribute N=1000 active Brownian particles. These particles are to be positioned within a LxL area, based on a predefined configuration stored in a .gsd file, where each particle is assigned a specific location. In the lower half of the rectangle, I intend to place 500 non-active Brownian particles, also arranged based on a specific LxL configuration from a different .gsd file.

My question concerns the implementation of this setup. Should I first introduce the 1000 particles within an LxL box, then expand the box to a 2LxL rectangle before adding the additional 500 particles? If this is the case, which functions would allow me to do so? I have read in the documentation about using hoomd.update.BoxResize for resizing the simulation box, but I need this resizing to happen instantly. Moreover, how can I manage the addition of particles so that, after placing the first 1000 particles, I can incorporate the subsequent 500 particles from the second configuration?
To clarify, I am not seeking a complete code for this setup, but rather guidance on the methodology and the specific functions that would enable me to achieve the intended configuration.

Thank you!

[joaander]

All your particle positions are pre-defined, so there is no need to involve HOOMD-blue when placing them. Use your preferred method to place items in a list when constructing the gsd file. For example:

frame.configuration.box = [L, 2*L, 0, 0, 0, 0]
frame.particles.position = active_brownian_particle_positions + non_active_particle_positions

Assuming you would like the active and non-active particles to have different types:

frame.particles.typeid = [0] * len(active_brownian_particle_positions) + [1] * len(non_active_particle_positions)

[joaander]

I don't know how you compute the color of the particles, so I can't say for certain. I would guess that you forgot to tell the GSD writer that typeid is dynamic.

[idop11]

Indeed, this solved the problem. Thank you.
I have one more question regarding this setup (I hope it's fine that I'm asking multiple questions in this post – I think it's more straightforward than creating a new one).
I want to adapt this system so that if a non-active Brownian particle (from the lower half) crosses to the upper half, then its new self-propelling velocity (the parameter vzero) will align with the direction of the previous velocity vector it had the moment before it crossed into the upper box when it was still non-active. For example, if at the step before changing its type to an active particle, the unit vector of velocity was $(\frac{\sqrt{2}}{2},\frac{\sqrt{2}}{2},0)$, then after crossing, its self-propelling velocity should be: $vzero\cdot(\frac{\sqrt{2}}{2},\frac{\sqrt{2}}{2},0)$.
I understand that it's not possible to change the active force for each particle individually, so I thought about determining the orientation of the Brownian particle before it changed its type, ensuring that after crossing and becoming an active particle, it will retain the same orientation.
However, I haven't managed to figure out how to calculate the orientation of a non-active Brownian particle (that performs a random walk). To clarify, by the orientation of the non-active particle, I mean the direction in which it fluctuates during a specific time step. Is there a straightforward way to find its orientation in this state?

I'm attaching my code once again (I made some minor changes following your advice):
sample_code_new.txt

[joaander]

Yes, you should set your particle's orientation to achieve the desired direction of the active vector in the world frame. The active force vector you supply to force.Active is in the particle's reference frame. You can access the properties snap.particles.velocity and snap.particles.orientation to implement this. You may find a library like rowan useful.

Note that embedding unrelated questions in discussion threads makes it harder for others to discover the question and the answer.

[idop11]

Thank you. If I have any new questions, I'll make sure to post them separately.

As for my current question, I've observed that the orientation of non-active Brownian particles remains unchanged from their initial state, despite their orientation ostensibly changing at each timestep (perhaps the fact that their motion is solely due to random thermal fluctuations is influencing this?). Do you have any insights on how to address this issue?

[joaander]

Please see the documentation on "Degrees of freedom". Your simulation does not meet these conditions:

• DPDConservative applies 0 torque to particles.
• You do not set a moment of inertia.
• You do not set integrate_rotational_degrees_of_freedom=True.

Therefore, the integration method will not integrate the rotational degrees of freedom (orientation and angular_momentum) and they will remain fixed. I don't think this is an issue and it does not need to be addressed. Your particles are spherical!

You previously asked about setting a particle's orientation when it switched from non-active to active and I answered that question. Once active, your ActiveRotationalDiffusion updater will update the orientations of the active particles to implement rotational diffusion.