Book 3 · The Mini-Beast · Chapter 4 of 44
The Critical Curvature Threshold κ*
One number governs every regime change in every domain.
κ ≥ κ* ⇒ generative transition
OrientationOne Number, Every Room
The claim of this chapter is narrow and testable: a single geometric quantity governs regime change in systems that share no physics, no substrate and no timescale. It is not claimed to be a fitted constant, but no computation of it from the geometry of any system is shown on this page, and the later section “Two Kinds of Number” classes the per-domain values as fitted parameters awaiting calibration.
The critical focal curvature is defined through the focal radius foc(x), the distance from a point to the nearest focal point of the submanifold. When curvature is sufficient to trigger a fold, the threshold is bounded from both sides — by the second fundamental form and by the sectional curvature.
Reading the DefinitionWhy a Minimum, and Why Two Terms
The second fundamental form ‖II_x‖
II measures how the submanifold bends inside its ambient space — extrinsic curvature. Where the embedding bends sharply, the focal point is close, and a small displacement along the trajectory is enough to reach it.
The sectional curvature √K_sec(x)
K_sec is intrinsic: it does not care how the manifold sits in anything. With positive sectional curvature, geodesics converge on their own, and that convergence sets a second, independent ceiling on the focal radius.
Why the minimum is the right combination
The fold triggers on whichever mechanism reaches its limit first. Taking the minimum is not a modelling convenience — it is the statement that either route to focal convergence is sufficient. This is also what makes κ* portable: a plasma physicist and a market microstructure analyst compute different terms and land on the same object.
Rank loss at the transition
In each domain the accompanying Morse assumption does the same work: the relevant energy or stability functional has non-degenerate critical points away from the transition, and its Hessian loses rank exactly one at the fold. Rank-one degeneracy is the analytic signature that κ has reached κ*.
- The displayed identity is false in general. For a hypersurface in flat space the nearest focal point lies at distance 1/kmax, so 1/foc = kmax, the largest principal curvature. With Gauss's formula K = k₁k₂ one has √K ≤ kmax ≤ ‖II‖. So 1/foc is sandwiched between √Ksec (below) and ‖II‖ (above), and min{‖II‖, √Ksec} is just √Ksec, the lower end. The two agree only at umbilic points (k₁ = k₂, e.g. a sphere). Example: principal curvatures 1 and 0.25 give 1/foc = 1, ‖II‖ = 1.031, √K = 0.5, so min{‖II‖, √K} = 0.5 ≠ 1. A cylinder of radius R in ℝ³ has 1/foc = 1/R but √K = 0, so the min is 0.
- What stands. The sandwich √Ksec ≤ 1/foc ≤ ‖II‖ (for a convex hypersurface in flat ambient space, K ≥ 0) is correct and is the sense in which the threshold is “bounded from both sides”. The “minimum” reading (“either route is sufficient”) is not what the bounds say. The scope of this check is hypersurfaces in Euclidean space; the chapter does not state its setting.
- The scaling-failure test uses df = 1 + log μ / log λ, which is undefined for μ < 0 and unit-dependent (see the Chapter 3 audit note); Chapter 3 writes log|μmax|. The “sub-threshold transition falsifies the model” test also needs the Chapter 1 point that K as written reaches κ* only asymptotically.
- “Proved invariants” (T*, μmax, τ) = (2π, −2, 2): this audit has not located a Lean theorem for the triple. The autophagy file proves −V″(1)/2 = −3 for its potential and carries −2 as a separate constant (Chapter A audit), so “proved” is unverified for μmax = −2.
ch04ks-audit-verify.py.Stated ValuesThe Threshold in Four Rooms
The values below are stated per domain; no computation from that domain's geometry is shown, and (see “Two Kinds of Number” below) they are to be treated as fitted proxies awaiting the calibration pipeline, not as computed results. The units differ because the manifolds differ; the role of the number does not.
| Domain | μ_max (s⁻¹) | ω (rad/s) | β | κ* |
|---|---|---|---|---|
| HPA stress | −0.38 | 0.21 | 1.9 | 0.15–0.22 |
| Neural oscillations | −0.55 | 0.45 | 2.1 | 0.25–0.35 |
| Circadian clock | −0.29 | 2π/86400 | 1.6 | 0.08–0.12 |
| Immune adaptation | −0.44 | 0.18 | 2.0 | 0.11–0.19 |
| Plasma reconnection | −0.42 | 0.015 | 1.8 | 0.8–1.2 × 10⁻³ km⁻¹ |
| Market volatility | −0.67 | 0.28 | 2.4 | 0.12–0.18 |
Two entries deserve comment. The plasma threshold carries physical units (km⁻¹) because the current-sheet manifold is embedded in physical space; MMS instrument resolution is approximately 10⁻⁴ km⁻¹, an order of magnitude finer than the predicted band, so the measurement is feasible. The market threshold is dimensionless because the Fisher information metric on volatility space is dimensionless by construction.
Three conditions, applied in every domain
- Sub-threshold transition. A generative transition observed at κ < κ* falsifies the model outright.
- Selection failure. Post-transition topology not selected by gradient descent on Φ falsifies the U operator.
- Scaling failure. Fractal dimension deviating from d_f = 1 + log μ / log λ falsifies the compression–fold link.
These three conditions are stated once here and inherited by every domain chapter. Where a domain adds a fourth condition, it is because the instrumentation permits a sharper test, not because the general protocol is weaker.
Two Kinds of NumberProved Invariants and Fitted Parameters
A reader moving between Book 3 and the later volumes will meet μ_max twice, with different values, and the difference is not an error. It is the most important methodological point in the series.
The canonical invariants — (T*, μ_max, τ) = (2π, −2, 2), ε* = 1/3 — are dimensionless facts about an abstract operator chain, proved, carrying no physical unit anywhere in the Lean sources.
The per-domain values in this book — −0.38, −0.42, −0.67 — are fitted parameters: operationalised proxies estimated from measured data. They are not the proved constant expressed in other units, and they should never be cited as though they were.
Getting from the first kind to the second requires a calibration pipeline — name a measurable observable, fit a parameter vector against a real time series under a stated loss, validate out of sample, and report the error bars. That step has been attempted once in this corpus, for μ_max, and it did not clear. Treat every domain number in Book 3 as awaiting that pipeline.
BridgesWhere This Connects
- WP-31 · The Calibration PipelineHow a dimensionless fixed point becomes a fitted, falsifiable function of measured data — and the five ways that step goes wrong.
- WP-29 · The Numerology SweepA corpus-wide check of every 'not a coincidence' claim attached to a specific number. Required reading before accepting a universality claim.
- Pointers · Brazil–China–Portugal, UFRNWhether κ* is a flow-invariant threshold in any Ricci-flow sense, or the correspondence is nominal — flagged there as the bridge most likely to be a metaphor.
- Book 6 · G6 CrystalWhere the canonical dimensionless invariants (T*, μ_max, τ) = (2π, −2, 2) and ε* = 1/3 are inherited from Vol V.
- The 1/3 Invariant · Positional Dominance in Network GamesThe same ε₀ = 1/3 stability radius, argued from a two-player stochastic game rather than a biological orbit — cite v2 (WP-38), the v1 analytical threshold is superseded.