Particle Localization on the Smallest Entity Instead of Elements

[Fuad-HH]

In trace_particle_through_mesh or search_mesh, they take a parameter called elem_ids which stores the current element the particle is located inside. But in some scenarios, it can be useful to know whether the particle is on a side or a node of an element. For example, in neutron transport, particles can move in any direction and miss elements to check for intersection with.

As a solution, Prof. @jacobmerson and I thought it could be solved by having the entity ID and entity dimension, replacing elem_id.

[jacobmerson]

@cwsmith just want to confirm that you are on the same page with us here.

By storing the classification of the particle on the mesh entity (dimension + ID), we resolve two problems that are not easy to resolve otherwise:

1. Ensures consistency when asking about the point containment from the red triangle and the blue triangle.
2. Clarifies to downstream algorithms that they must handle situations where a particle lands on a element boundary (i.e., face, edge, vertex). This should help eliminate some number of lost particles that are currently being observed.

Note: This change to use mesh entities (dimension+ID) has also been made to the uniform grid localization in PCMS so we could make these interfaces consistent.

[cwsmith]

I'm fine with adding a search procedure/option that supports that, but hesitant about the breaking change to the current search that downstream apps are using.

Can this effectively be added as a post-processing option or does the search procedure itself need to know which entity on the element closure is being traversed?

[zhangchonglin]

@jacobmerson: just saw this discussion:

• Long term, I think this could a good strategy to be able to handle particles precisely on the material surface for particle-surface interaction, or generating new particles from interface for the particle simulation tools I am developing.
• I also agree with Cameron @cwsmith. This could be implemented as a new search function to avoid breaking existing code and to allow for thorough testing and comparison in existing code.

[jacobmerson]

This should not be done as post processing. But we can implement it on the generalized search method Fuad’s been working on since no one else uses that yet.

We can maintain the current behavior and even implement the old behavior in terms of the new behavior.