What horizon-scale imaging features would reveal fractal geometry versus nonextensive correlations in competition?

This is a sophisticated question about distinguishing two competing signatures in large-scale structure: *fractal geometry* (self-similar patterns across scales) versus *nonextensive correlations* (long-range statistical dependencies that don't obey standard additivity). Let me work through what the archive suggests.

The core distinction

Fractal geometry produces self-similar clustering—the same branching or filamentary pattern repeats at smaller and smaller scales [AG-2024.03-477]. Nonextensive correlations, by contrast, are about how statistical moments scale; they preserve spatial structure but violate the usual independence assumptions [AG-2025.03-103]. These are genuinely different physical mechanisms, even though both can create deviations from Gaussian randomness.

What horizon-scale imaging would reveal

Fractal signatures: Look for *recursive bifurcations* and *eddy cascades*. In cosmological large-scale structure, a fractal turbulent cascade should show the same Jeans-length-driven fragmentation pattern repeating across hierarchies—from galaxy cluster seeds down to galaxy-scale seeds [AG-2024.03-477]. Observationally, this means power-law galaxy counts with a constant fractal dimension $D$ across redshift bins; recent data favor $D ≈ 1.5$ at $0.1 < z < 1.0$ [AG-2025.04-196]. The key signature is *scale-invariant topology*: Minkowski functionals (genus, perimeter, area) should follow simple power laws without breaks.

Nonextensive correlation signatures: These show up as *anomalous clustering of rare events*. If spatial correlations are long-ranged but sub-fractal (i.e., not purely self-similar), then extreme overdensities—say, beyond 7.5σ—will cluster non-randomly and develop non-spherical morphologies that don't match Poisson expectations [AG-2025.01-300]. The Euler characteristic (a topological measure of connectivity) will deviate from mean-field predictions *more severely* for rare events than for typical ones. Rayleigh-Lévy flights, which model hierarchical correlations without strict self-similarity, show this signature: their density PDFs and perimeter statistics remain non-Gaussian even in the beyond-mean-field regime [AG-2025.03-103].

In black-hole shadows

A strikingly concrete test: non-Kerr compact objects produce *fractal eyebrow formations* in their photon shadows when light orbits resonate between multiple escape routes [AG-2024.05-307]. This is true fractal geometry—infinitely fine structure from chaotic resonance. By contrast, if a shadow's brightness correlations (lagged images from light echoes) show long-range temporal memory without spatial self-similarity, that signals nonextensive correlation rather than fractal structure [AG-2025.12-230].

The observational separator

Measure the Euler characteristic as a function of density threshold and cluster size:

• Pure fractals show Euler characteristic scaling consistent with a single fractal dimension across all thresholds.
• Nonextensive correlations produce threshold-dependent topology: the Euler characteristic changes *shape* (not just scale) as you probe rarer and rarer events.

Complementarily, compute power-spectrum slopes at different angular scales in the CMB or matter distribution. A true fractal has a single power-law index; nonextensive correlations can produce multi-slope behavior or curvature in log–log plots [AG-2024.03-506].

---

The archive does not directly address a head-to-head observational protocol for *competing* fractal versus nonextensive models in a single dataset, though the papers strongly suggest that topology (Minkowski functionals) and rare-event clustering are the right knobs to turn.