Cell Importances Discussion

[pshriwise]

Description

There's been some recent (and IIRC, previous) discussion around adding cell importance. I'm creating this issue as a place for that discussion and for the placement of motivating use cases.

One use-case brought up in slack recently was the ability to easily terminate particles that enter a specific cell. I believe this can be done with the correct set of surfaces and boundary conditions, but there are possibly other cases I'm not thinking of that we may only be able to support with such a capability.

Alternatives

I've considered that cell-based weight windows could be used to accomplish the same behavior as importance with more flexibility, though they would be more difficult to define and are somewhat less intuitive.

Compatibility

This should be a pure addition in terms of code, but consideration would need to be given as to how it will interact with other forms of variance reduction like survival biasing and mesh-based weight windows.

[yrrepy]

I find cell importances very useful. They are often very easy to setup, or add-on to an existing simulation.

You can have very simply geometries where you just throw on some importances and you can get your relevant particles through the shield, or to the relevant tallying location.
In some situations I have heavy lead shields, and I can very easily onion layer the lead shield and drive the photons through, without doing the same to the neutrons.
No need to go through the setup process of generating, iterating an applicable weight window mesh.

Another aspect, from a beginner/learner POV; before understanding weight windows VR, etc. I would encourage anyone to start with importance and witness their effect. It's just simpler and more intuitive.
The concept of particle weight is somewhat more abstract (I like the explanation that it is like importance, but onboarded into the particle rather than inherent to the geometry) (of you course you still have the geometry/mesh weight bounds).

I'm pro Cell Importances.

[tjlaboss]

This would be nice for compatibility between OpenMC and MCNP models.

[GuySten]

I've opened PR #3843 to fix this issue.