Reproducing a 1986 MORSE-CG fusion-shielding streaming study in OpenMC + weight windows — looking for a technical read

User Support
1

I’ve been using OpenMC to reproduce a 1986 fusion-reactor shielding paper I co-authored as a grad student (Halley & Miller, “Neutron Streaming Through Gaps in Fusion Reactor Shielding,” Fusion Technology, 1986, https://doi.org/10.13182/FST86-A24782; original calcs done in MORSE-CG). The motivation: revisit the original analog Monte Carlo results with modern variance reduction, and generate a clean parametric dataset for a small ML surrogate.

What’s done:

  • Straight-slot streaming reproduces the original transverse falloff (Fig IV-1) to within ~0.1 decade.
  • Analog MC hits the same “wall” the original 1986 referee flagged for deep-penetration cases — solid-shield and multi-step baselines sit at the noise floor (FSD > 0.3) with plain analog tracking.
  • MAGIC-generated weight windows (Davis & Turner, 2011) converge those same cases cleanly, recovering a three-step-staircase result within 0.08 dec of the original (Fig IV-11), and a qualitative two-peak reproduction of the single-step dog-leg case (Fig IV-8) — quantitatively weaker than the original, likely due to under-specified corner geometry in both the original and reproduction.
  • A two-Gaussian-Process surrogate (straight/stepped) fit to the parametric sweep, LOO mean error 0.19 dec.

Background: the underlying 1984 MS thesis (more detail on the original MORSE-CG model than the paper) is at DSpace

Code, model, and dataset are on GitHub (alanhalle/openmc-starfire), archived at Zenodo: OpenMC STARFIRE gap-streaming reproduction | Zenodo

What I’m looking for:

I’d like an independent technical read before submitting — ideally from someone who works with OpenMC weight windows / CADIS-FW-CADIS or has shielding/deep-penetration MC experience, to sanity-check the WW setup and the physics framing (I’m an independent researcher, not currently affiliated with an institution). Also planning an arXiv preprint (physics.ins-det) and would appreciate a pointer to anyone willing to endorse an unaffiliated submitter in that category.

Happy to share the manuscript draft directly. Thanks for any pointers.