Density-Gated Transport in Galaxy Rotation Curves A Regime-Aware Effective Model for the Full SPARC Sample

We present a regime-aware phenomenological model for galaxy rotation curves using the publicSPARC mass-model sample. The model treats rotation-curve residuals as structured responses tobaryonic organization rather than as deviations from a single universal scaling relation. Using143 usable SPARC galaxies, we construct structural diagnostics based on stellar anchoring, gasdominance, radial coherence, outer-demand behavior, and effective disk surface density. Thesediagnostics define transport regimes in which stellar-anchor support, gas-driven outer response,shape-gated recovery, or bounded outer-reservoir support may be activated.The Version 4 pipeline reproduces the full benchmark suite, including core, final, shape-gated,and active outer-reservoir system stages. The best full-sample model achieves a mean chi-squared per galaxy of 37.61 and a median of 9.55, while the clean sample excluding active outer-reservoir systems yields a mean of 27.30 and a median of 9.10. Removing the 20 highest-residualsystems reduces the mean to 14.33 while leaving the median near 8.24, indicating that theelevated full-sample mean is dominated by a small number of structurally identifiable tailsystems.Tests of global amplitude and global surface-density gating do not provide comparableorganization, indicating that the fitted amplitude does not act as a universal physical coupling butinstead serves as an effective summary of local, regime-dependent transport structure. Theseresults support a reproducible empirical framework in which galaxy diversity is organized bystructural regime rather than treated as random scatter. Reproducibility and Data Access A fully reproducible single-cell analysis pipeline, including all data parsing, model stages, and figure generation, is publicly archived and available at: https://doi.org/10.5281/zenodo.19957968 The pipeline reproduces all results presented in this work using publicly available SPARC mass-model files (Lelli, McGaugh, & Schombert 2016). No proprietary data or hidden processing steps are required.