Stage 3 Results: What We Learned

We tested our hypothesis rigorously. Here is what happened — the part that worked, the part that did not, and where the work goes from here.

Stage 3 summary
  • 5D Einstein–Gauss-Bonnet solutionresidual < 6×10⁻⁶passed
  • Holographic dictionary → T(k)uniquely determinedpassed
  • CAMB pipeline validationmethodology soundpassed
  • S₈ constraint satisfiedonly in isolationfailed
  • Lyman-α constraint satisfiedonly in isolationfailed
  • Both satisfied simultaneouslyexcluded — structuralfailed
Every mathematical and methodological step passed. The failure is observational, and it appears only when both constraints are imposed at once.
Why S₈ and Lyman-α cannot both be satisfiedA schematic of suppression required versus wavenumber. Weak lensing near k of 0.5 h per Mpc allows only a few percent of suppression, while the Lyman-alpha forest near k of 5 h per Mpc requires far deeper suppression. A monotonic power-law transfer function rising from left to right passes through only one of the two allowed boxes at a time.wavenumber k [h/Mpc] — logarithmicsuppression requiredS₈ allowed (k ≈ 0.5)Lyman-α required (k ≈ 5)derived monotonic T(k) ∝ k^(−1/2)≈ 1 decade in k
Schematic of the structural conflict: the two constraints sit about a decade apart in wavenumber. A curve deep enough for the Lyman-α box overshoots the S₈ box, and vice versa. Tuning the parameters slides the curve; it does not bend it into both boxes.

A. The Success — the mathematics works

  • The full 5D Einstein–Gauss-Bonnet equations were solved to a boundary-value residual below 6×10⁻⁶, giving a self-consistent warp factor, radion profile, and radion potential.
  • Holographic entanglement entropy uniquely determines the transfer function from that geometry. There is no fitted ansatz left in this step — which is exactly what Stage 3 set out to achieve.
  • The resulting power-law scaling, T(k) ∝ k^(−1/2), follows from the geometry rather than from a choice.

This part of the programme stands, independent of the observational verdict. See the Framework page for the derivation and the technical report on GitHub .

B. The Failure — the observations do not match

Verdict: observational viability of the power-law radion-leakage mechanism is below 20%. We treat the mechanism as excluded and report it as such.

C. Why This Matters

  • Eliminates one pathway — power-law radion leakage — so nobody has to re-walk it.
  • Narrows the possibility space for cosmology: the wavenumber gap between S₈ and Lyman-α is a hard constraint on any monotonic small-scale suppressor, not just on this model.
  • Shows that rigorous testing is mandatory. A derivation that is mathematically flawless can still be physically wrong; mathematics is not physics.
  • Offers a concrete, testable data point in the wider extra-dimensions programme, including the conceptual "Dark Dimension" proposal of Vafa et al. (2022).

D. What’s Next

Three resurrection paths are under consideration. None is derived yet, and we are not claiming any of them works.

Path A
Non-monotonic transfer function

Allow a localized feature rather than a power law. Requires a mechanism that switches off between the two scales — currently unmotivated.

Path B
Time-dependent coupling

Let the leakage epoch itself be scale-dependent. Adds freedom, but risks reintroducing the fitting that Stage 3 removed.

Path C
Partially-coupled dark matter

Only a fraction of the dark sector couples to the leakage. Breaks the monotonicity that produced the exclusion, and connects to broader dark-matter phenomenology. Most promising — and speculative.

Path C is where the current effort is going. If you work on dark-sector phenomenology or Boltzmann codes, we would rather have your criticism early than late — see Get involved.

Reproduce this

The 5D solution, the derived transfer function, the CAMB pipeline, and the falsification test are all published. If the result is wrong, it should be straightforward to show that.