An E8 Fine-Structure Signal in Sparse Geographic Catalogs

We test whether projections of the E8 root system's 6,720-edge network onto the Earth's surface align with sparse geographic catalogs beyond what chance permits. From 270 random 8→3 orthonormal projections tested against 160 sacred sites using a best-of-search Monte Carlo null model (1,000 trials each), four projections survive Bonferroni correction with p < 0.05 and effect sizes d = 2.13–3.52. The alignment is a fine-structure phenomenon: specific E8 edges run within 1–11 km of specific sites at orientations requiring 2–3° precision. Two rounds of synthetic projection engineering (200 matrices matching confirmed seeds' dimension fingerprint, a second round matching inter-row-angle patterns) demonstrate that the mechanism cannot be reduced to any structural property of the projection matrix (cross-validated classifier AUC = 0.52, chance level). Shape controls confirm E8 specificity within the tested set: Fibonacci spheres (n = 240, 600) and 120-cell polytope projections (20 seeds) fail comprehensively under search-aware null models, with RMS values 85–150× worse than E8. Independent geophysical replication using Indo/Atlantic mantle hotspots (36 sites, zero overlap with sacred sites) yields significant alignment at p = 0.005 (seed 68). Sacred-sites seed 46 cross-replicates on hotspots (p = 0.025), providing the first cross-catalog replication in this study. Additional sparse datasets (plate boundaries, earthquake epicentres, subduction zones) show significant RMS-level alignment at p = 0.010–0.020. Dense catalogs (n ≥ 300) produce flat landscapes with no signal — a sparsity principle constraining the phenomenon to a density regime of roughly 30–160 sites relative to E8's 6,720-edge network. Version 2.1.0 adds a closest-neighbour 8D shape control: a 50-seed pilot of D₈ root-system projections (112 roots, 1,344 canonical edges; D₈ ⊂ E8) was run on both the 160-site sacred catalog and the 36-site Indo/Atlantic hotspot catalog using the identical pipeline. No D₈ seed reached p < 0.05 on either catalog (best p+ = 0.104 and 0.095 respectively). This strengthens the "E8-specific within the tested set" claim against the strongest available 8-dimensional control. The closest neighbour to E8 in 8 dimensions does not produce the alignment signal. The alignment is statistically robust across independent catalogs and survives every control applied. Its origin remains unknown.