Preserving Orientation when Switching Particle Types in Overdamped BD

[idop11]

Hello everyone,

I’m running a 2D simulation in HOOMD‐blue where the simulation box is conceptually split into two regions along the y-axis:

• The top region $(y>0)$ contains type A particles that have a nonzero active force.
• The bottom region $(y\leq0)$ contains type B particles that have zero active force.

As soon as a particle crosses from $(y>0)$ into $(y\leq0)$ (or vice versa), I change its typeid accordingly. In principle, I want all particles to maintain a continuous orientation and to undergo rotational diffusion with time scale $\tau_r = 1/D_r$. However, in practice, once a particle switches type, it doesn’t appear to keep the orientation it had before the switch - it looks like the orientation resets or changes abruptly.

Here’s a minimal reproducible example that illustrates my approach. The essential points are:

1. I use a custom TypeSwitchAction to change typeid based on a particle’s $y$-position.
2. I have two forces:

• active (filter=type A) with a nonzero active force,
• activeB (filter=type B) with zero active force.

3. I apply ActiveRotationalDiffusion updaters to both active forces so that all particles should keep rotating.

Despite this setup, particles sometimes fail to preserve their pre‐switch orientation. I’d greatly appreciate any pointers or insights from folks who have encountered similar behavior or have ideas on how to fix this.

My script:
sample_code.txt

Problem:
I need each particle’s orientation quaternion to remain continuous (and keep rotating stochastically) when crossing from $(y>0)$ to $(y\leq0)$, but it appears that the orientation is reset or changed abruptly at the type switch.
Any help identifying the cause of this reset (and how to preserve orientation seamlessly) would be greatly appreciated.

Thanks in advance!

[joaander]

The only obvious problem I see in your code is that the filters are not updated. type_C and type_B will continue to refer to the original particles that matched the filter even though you change the type. Add a FilterUpdater to reevaluate the filters: https://hoomd-blue.readthedocs.io/en/v5.0.0/hoomd/update/filterupdater.html

Could you provide a smaller minimum example? Ideally you would place only one particle in the simulation box and set the initial position/orientation such that running a few timesteps is enough to trigger the behavior you observe. No GSD file is necessary, take_snapshot and print() should be sufficient to demonstrate that orientation is reset.