Conservative volatility transport, Hodge circulation, and state-gated boundary forcing on time-varying equity networks: code and data

Code, processed data, results and manuscript sources for G. Lim, "Conservative volatility transport, Hodge circulation, and state-gated boundary forcing on time-varying equity networks", submitted to Physical Review E (2026). The study treats the 20-day realized variance of individual stocks as a conserved mass transported on time-varying residual planar maximally filtered graphs: an upwind conservative advection that splits exactly into pure transport and a compressive velocity-divergence term, a symmetric Laplacian diffusion, the discrete Hodge decomposition of the edge current, the Schnakenberg entropy production with its pure-diffusion counterfactual and the 3-clique cycle affinities, a purged out-of-sample open-system inverse problem (M0-M5n, constant and stress-gated macro boundary, fixed-gate grid), three surrogate nulls, and hub persistence under volatility-quintile and mass controls, on 1,304 rolling windows of a point-in-time S&P 500 universe (2000-2026) and 1,283 windows of the Korean exchange. The deposit contains the Python pipeline (scripts 01-12, runnable from the unpacked archive without path changes), the public point-in-time constituent history, the US rolling universe and survivorship diagnostics, the log-return and dollar-volume panels, the window-, node-, edge- and face-level network outputs, the end-of-window state panels, every inverse-problem, surrogate and hub-persistence output, and the REVTeX sources, figures and compiled PDF. Every number in the manuscript is generated from the CSV files in this deposit (see manuscript/numbers_check.csv). Raw daily prices are not redistributed; script 01 re-downloads them from the public sources named in README.md. The Korean panel, universe and PMFG edges are those of the companion deposit https://doi.org/10.5281/zenodo.23912552; the fixed ten-year universe and physics-informed graph network of the earlier study are in https://doi.org/10.5281/zenodo.23891620. Code under scripts/ is MIT-licensed (see LICENSE); data, results and figures are CC BY 4.0. Funding: National Research Foundation of Korea, grant RS-2024-00507484.