Reproducibility artifacts for "Learning low-rank neutron transport dynamics on linear subspaces and nonlinear manifolds"

This dataset provides the large numerical artifacts needed to reproduce theone- and two-dimensional results in “Learning Low-Rank Neutron Transport Dynamics on Linear Subspaces andNonlinear Manifolds: A Semi-Intrusive Operator-Inferred Approach”by Francesco A.B. Silva and Jean C. Ragusa. For the one-dimensional problem, the deposit preserves the completedreproduction and audit artifacts associated with Figures 1–5. These includethe full-order snapshot data, reduced-order-model data, figure-data bundles,final figures, configuration and provenance records, independent referencedata, and the execution/audit history used to establish the regeneratedresults. The one-dimensional reproduction retains an important provenance distinction.The manuscript states a zero initial angular flux, whereas the numericalfigures were generated using the localized sigmoid initialization preservedin the deposited data and source. In addition, the original historicalrank-dependent nonlinear regularization selections for Figure 4 were notavailable and were regenerated using the documented search procedure.Accordingly, the deposited one-dimensional results constitute a transparentregenerated reproduction rather than a claim of bit-for-bit historicalidentity. For the two-dimensional problem, the deposit contains the two large OpenSnangular-flux datasets that are impractical to store in Git: - `3newh_aflx.tar`, containing the 1001 transient angular-flux VTU solutions;- `aflux_3newss_1000_0.vtu`, containing the steady-state angular-flux solution used to center the transient snapshots. The associated GitHub repository contains the OpenSn inputs and a three-stage2-D reproduction workflow: 1. run the OpenSn full-order transport problem;2. construct the DG mass matrices, full-order operators, centered snapshot matrix, and steady-state centering vector;3. compute the POD basis, nonlinear-manifold reduced models, inferred operators, and paper figures. The 2-D preprocessing workflow has been validated against preserved historicaldata. The regenerated centered snapshot matrix, DG mass matrices, andfull-order operators agree exactly in numerical values with the historicalreferences. Derived 2-D quantities such as the centered snapshot matrix, PODbasis, reduced operators, ROM solutions, and figures are therefore notduplicated in this Zenodo record because they can be regenerated from thedeposited OpenSn outputs and the tracked source code. Reserved dataset DOI:10.5281/zenodo.21762243 Source repository:https://github.com/Open-Sn/Nonlinear-Manifolds-for-Sn-Transport Source commit:[FINAL REPRODUCIBILITY COMMIT TO BE INSERTED BEFORE PUBLICATION] The deposited numerical artifacts and documentation are licensed underCC BY 4.0. Source code in the associated repository retains the repository’sMIT license.