"A horizon is what remains of a Whitney fold when the metric stops being positive-definite. The barrier survives the change of signature; only its name changes — from basin boundary to null hypersurface — and with the name, the proof that nothing crosses it." — Working note, June 26, 2026
Chapter 1 derived the dark-matter result on a Riemannian contact manifold M = ℝ>0 × S¹ × ℝ. The contact form α = dz − r²dθ generated a fold at the Gronwall radius ε₀ = 1/3, around which the Tribonacci weighting concentrated mass and produced the tenfold lensing excess that matches the Natarajan clusters. The derivation was non-relativistic from start to finish. The entropy coordinate z was a dimensional accumulator of dynamical history, not a proper-time coordinate; the radial coordinate r was a spatial radius; the metric was Euclidean throughout.
This chapter changes one thing: the metric on (r, z) acquires Lorentzian signature. Everything that depends on signature changes; everything that does not, persists. The contact form α survives. The operator chain G = U ∘ F ∘ K ∘ C survives. The fold at r* survives. The Tribonacci constant η survives, because it is an algebraic invariant of the operator algebra and not of the signature. What changes is the character of the fold — and with it, the meaning of the basin boundary, the structure of the trapped region, and the nature of the obstruction to inner-escape trajectories.
The promotion from Riemannian to Lorentzian dm³ is the smallest possible deformation of Chapter 1's machinery that admits causal structure. The simplest case — the case this chapter develops in full — is the (1+1) reduction on the transverse (r, z) slice. In that slice the Whitney A₁ fold of Chapter 1 becomes a null hypersurface: a lightlike boundary. The trapped region behind it is an event horizon.
Let V : ℝ>0 → ℝ be the Lyapunov function of the radial dm³ dynamics, V(r) = ½(r−1)². The Lyapunov gradient V'(r) = r − 1 vanishes at the limit cycle r = 1 and at the basin boundary r = r* ≈ 0.776 (the asymmetric Gronwall radius certified by certify_rstar.py and stated in Volume II §4). On the transverse slice with coordinates (r, z), define the metric
This is well-defined and Lorentzian wherever V'(r) ≠ 0, with the entropy direction τ playing the role of time. The metric degenerates on the locus V'(r) = 0 — that is, on the fold S = {r = r*}, which is exactly where dτ blows up. A degenerate Lorentzian metric at a hypersurface is the standard signature of a null surface.
The pair (ℝ>0 × ℝ, gLyap) with dτ = dV/V'(r) is the natural Lorentzian promotion of the Riemannian (r, z) slice. The metric is positive-definite nowhere; its signature is (−, +) on the regular set V'(r) ≠ 0, and degenerate (rank-1) on the fold locus V'(r) = 0.
The fold locus S = {r = r*} of the Riemannian Chapter-1 framework is a null (lightlike) hypersurface in the Lyapunov-Lorentzian metric gLyap.
Proof. On S, V'(r*) = 0 (definition of basin boundary). Hence dτ |S diverges, and the induced metric on TS has signature (0, +) — null. ∎
Chapter 1's proof structure for the inner basin asked: can we Gronwall-bound a trajectory starting at r₀ < r* by an exponential decay to r = 1? The bound never closed in Lean (Issue #13 in the AXLE tracker), and we now see why: the bound is asking the wrong question. The obstruction to inner-escape is not analytic but causal. There is no exponentially decaying trajectory from r < r* to r = 1 because such a trajectory would be a timelike curve crossing the null fold, and timelike curves do not cross null hypersurfaces.
For r₀ < r*, there does not exist μ < 0 such that |r(t) − 1| ≤ |r₀ − 1| · eμt for all t > 0.
Proof. Suppose such μ existed. Then the trajectory r(t) is timelike in gLyap (its Lyapunov energy decreases exponentially, so |gLyap(ṙ, ṙ)|1/2 > 0). It connects r₀ < r* to r = 1 > r*, hence crosses the fold S = {r = r*}. By Theorem §2, S is null. The standard causality theorem (O'Neill Semi-Riemannian Geometry §14.1; Hawking-Ellis Prop. 6.4.6) forbids timelike curves from crossing null hypersurfaces. Contradiction. ∎
The honest reading of this closure: the Riemannian dm³ framework contained the Lorentzian information all along. The Whitney fold's degeneracy in the Lyapunov gradient is exactly the lightlike condition. Promoting to Lorentzian language is not adding new physics — it is recognising the geometric content the Lyapunov function was already carrying. The companion math note book8/notes/issue-13-null-causality.md contains the proof in clean form. The Lean implementation is downstream of that note.
The Lorentzian promotion of contact geometry is not original to this chapter. Robert Beig, Piotr Chruściel and collaborators have developed a substantial theory of contact 1-forms on Lorentzian manifolds, with particular attention to the behaviour of Reeb-like generators on null hypersurfaces and isolated horizons. The references that ground this chapter are Beig and Chruściel on Killing initial data and contact structures, Beig on volume-preserving deformations, Hayward on trapping horizons as boundaries of contact-like distributions, and Ashtekar–Krishnan on isolated and dynamical horizons.
What this chapter does not attempt: a full rederivation of Beig–Chruściel from scratch, or any contribution at that level. What this chapter does attempt: the smallest possible application of that machinery to the dm³ framework, namely the recognition that the Riemannian Whitney fold of Chapter 1 carries a natural Lorentzian-contact structure in which it is null. The full Beig–Chruściel program would lift the rest of the dm³ operator chain — C, K, F, U, T — to a causally-respecting setting; this chapter does only the F operator, because that is what closes Issue #13.
The companion survey at book8/notes/lorentzian-contact-survey.md records the reading list and the six open questions of the Phase 1 program. Two of those questions are settled here (whether the Riemannian Reeb extends to the Lorentzian setting on the fold side, and whether ε₀ = 1/3 survives the promotion); the other four remain open.
What happens to the rest of the dm³ chain G = U ∘ F ∘ K ∘ C when the fold F becomes null? Three of the four operators are signature-independent and survive verbatim: Compression C is a Lipschitz projection (no metric content), Curvature K is a Lyapunov descent (only uses dV, not g), Unfolding U is a stabilisation onto an attractor (the attractor and its basin do change in Lorentzian setting, but the operator's definition does not).
The Fold F changes. In Riemannian dm³ it was a Whitney A₁ singularity of a smooth map between Riemannian manifolds. In Lorentzian dm³ it becomes the null hypersurface S itself. The fold-as-event is replaced by the fold-as-locus: the Whitney A₁ singularity is the lightlike boundary of the trapped region, not an instantaneous bifurcation event. The mathematical content is the same — rank-1 Jacobian deficiency — but its geometric role is now causal, not dynamical.
FL is the null hypersurface S = {V'(r) = 0} of the Lyapunov-Lorentzian metric, equipped with the induced metric gS of signature (0, +).
The cleanest worked example is the Schwarzschild horizon. Schwarzschild geometry is spherically symmetric and static; its (1+1) reduction in (r, t) coordinates is exactly the slice where the Lyapunov-Lorentzian framework of §2 applies, with the Lyapunov function V(r) = ½(r − rs)² and the basin boundary at r* = rs (the Schwarzschild radius).
Three checks should hold if the framework is right:
For Schwarzschild geometry with mass M, the radial Lyapunov function V(r) = ½(r − rs)² has its basin boundary at r* = rs = 2GM/c². The Whitney A₁ fold there coincides with the event horizon, in the sense that:
(i) S = {r = rs} is a null hypersurface in the (r, t) reduction.
(ii) The Reeb generator of αL on S is the standard null generator of the horizon, ∂v in Eddington–Finkelstein coordinates.
(iii) The Tribonacci constant η controls the rate at which the apparent horizon approaches the event horizon during gravitational collapse: rapp(t) − rs ∝ η−t/t0.
Claims (i) and (ii) reduce to standard facts about Schwarzschild geometry; claim (iii) is novel and falsifiable. It predicts that the horizon-formation rate in classical gravitational collapse should carry the Tribonacci signature — a numerical fingerprint that should appear in numerical-relativity simulations of stellar collapse.
| Object | Riemannian (Ch 1) | Lorentzian (Ch 2) | Verdict |
|---|---|---|---|
| Contact form α | dz − r²dθ | dz − r²dθ (same) | survives |
| Operator chain G | U ∘ F ∘ K ∘ C | U ∘ FL ∘ K ∘ C | survives (F → FL) |
| Gronwall ε₀ | 1/3 (basin radius) | 1/3 (null-surface marker) | survives |
| Tribonacci η | ≈ 1.839 (algebraic) | ≈ 1.839 (algebraic) | survives |
| Embodiment τ | 2 (curvature threshold) | 2 (causal threshold) | survives |
| Inner basin escape | Open (Gronwall fails) | Closed (null causality) | CLOSED ✓ |
| Schwarzschild horizon | Outside scope | = Whitney fold | identified |
| Hawking spectrum | Outside scope | η-modulated (predicted) | predicted, untested |
| Boost invariance | — | ? | open |
The promotion to Lorentzian dm³ produces three concrete, falsifiable predictions for horizon physics, each derived from a constant that the framework already certifies.
During spherical gravitational collapse, the apparent horizon approaches the event horizon at a rate that contains the Tribonacci constant: rapp(t) − rs ∝ η−t/t0. Test: high-resolution numerical-relativity simulations of Oppenheimer–Snyder collapse, fitting the apparent-horizon trajectory to an exponential and recovering the base.
The photon sphere (r = 3GM/c²) and the horizon (r = 2GM/c²) are separated by a factor of 3/2. The framework predicts that the locus where the Lyapunov gradient first vanishes — the "inner photon sphere" of unstable circular null geodesics — sits at r = (1 − ε₀) · rs = (2/3) · rs. Test: ray-tracing the inner photon orbits in Kerr and comparing the radius to (2/3) r+.
The least-damped quasinormal mode of a Schwarzschild black hole should carry the embodiment threshold τ = 2 as the ratio of its imaginary part to its real part, modulo standard mode-number factors. Test: extract Re(ω)/Im(ω) for the (ℓ=2, n=0) mode from numerical-relativity ringdown waveforms and compare to 2 · (some computable correction).
None of H1, H2, H3 has been tested. Each is concrete enough to be falsifiable by existing numerical-relativity infrastructure, and the predictions can be made before the data is examined.
This chapter closes one open obligation in the AXLE tracker (Issue #13, inner_basin_escape), via the structural argument captured in §3 and detailed in book8/notes/issue-13-null-causality.md. It identifies the Schwarzschild horizon as the canonical realisation of the Lorentzian Whitney fold. It promotes the dm³ operator chain to a causally-respecting setting, with three of the four operators surviving unchanged and F becoming a null hypersurface. It makes three falsifiable predictions for horizon physics.
The next chapter — Book 8 Chapter 3 — is the dual problem on the other side of the horizon. If FL is the null hypersurface that bounds the trapped region from the outside, what is the structure of the interior? The dm³ chain on the Lorentzian interior should produce the singularity at the center as the limit of the operator chain applied infinitely many times — a chained sequence of folds collapsing to the central singular point. The constant that emerges is conjecturally the inverse of η, since the operator algebra is reversed in time. That conjecture is the seed of Chapter 3.