Modular geometry of complex time: topological origin of the inflationary α-attractor parameter

Draft v1.3.9 — April 2026
Authors: Adam Porybný, with AI assistance (Iris)
Target: JCAP or PRD Letters
Full LaTeX source: tex/paper-draft-v1.tex (synchronized from this markdown via pandoc)

Revision note (13.–14.4.2026). Draft v1.0 listed modular invariance as axiom A3. Under AXIOM 1 (logical precedence: if a more logical formulation exists, no test between old and new is needed), A3 has been demoted from axiom to theorem. It now follows from A1 + A2 + M0 once photons are correctly identified with elliptic curves (see proc-psl2z.md for the full derivation). The axiomatic framework is now three axioms (A1, A2, A3 = old A4) together with one explicitly named meta-postulate (M0, structural realism), not four axioms. The reduction 4 → 3 is conditional on accepting M0: the total number of fundamental assumptions is the same as in v1.0, but the move is qualitative, not quantitative — M0 is strictly weaker and more general than the assertion "PSL(2,ℤ) is a symmetry group", and it is already implicit throughout gauge theory. The derivation chain (K = −1 → α = 2/3 → r = 2(1−nₛ)²) is unchanged in substance; only the source of PSL(2,ℤ) invariance has moved from postulate to theorem.

---

Abstract

We propose a minimal axiomatic framework in which physical time is a complex variable τ ∈ ℍ² (upper half-plane). Identifying the photon with its natural topological content — a closed 2π rotation together with a quaternionic companion direction — forces the photon to be an elliptic curve with modular parameter τ. As a classical theorem of the theory of modular forms, the moduli space of elliptic curves over ℂ is ℍ²/PSL(2,ℤ), so modular invariance is a theorem, not an axiom. The unique PSL(2,ℝ)-invariant metric is the Poincaré metric with Gaussian curvature K = −1, which fixes the α-attractor parameter:

$$\alpha = \frac{2}{3} = -4\chi(\mathcal{X}_1) = \frac{2,\mathrm{Area}(\mathcal{F})}{\pi}$$

This is a topological invariant of the modular curve X(1), not a free parameter. The result is a parameter-free consistency relation between two CMB observables:

$$\boxed{r = 2,(1-n_s)^2}$$

The theory itself contains no free parameter; inserting the measured value nₛ = 0.9649 (Planck 2018) yields r ≈ 0.0025. This numerical prediction is evaluated on the effective potential of the single-field limit of the modular-invariant action, V_eff(φ) = Λ⁴ tanh²(φ/2) (canonical α = 2/3 T-model); the microscopic derivation of V_eff from the fully PSL(2,ℤ)-invariant action remains an open problem (B4, §"Open problems"). Testable by LiteBIRD (~2032) and distinguishable from the Starobinsky R² model (r ≈ 0.0037) at ~1–2σ with LiteBIRD + CMB-S4.

Second hard prediction (Koide-α identity). The same topological invariant α = 2/3 simultaneously fixes the empirical Koide relation for charged leptons. Defining Q = (m_e + m_μ + m_τ) / (√m_e + √m_μ + √m_τ)², the √m-vector in ℝ³ subtends angle θ with the democratic axis (1,1,1)/√3, and Q = 1/(3 cos²θ). The spinor half-angle θ = π/4 (observer is SU(2)→SO(3), so the canonical orthogonality π/2 of flavour space projects to π/4 in observed √m-space) forces Q = 2/3, algebraically identical to the α-attractor exponent of Theorem 1. Taking m_e and m_μ from PDG as input, the relation predicts m_τ = 1776.97 MeV vs. measured 1776.86 ± 0.12 MeV (deviation 6·10⁻⁵, within experimental 1σ). This is a KRT computation, not a fit: the value 2/3 is the topological invariant of X(1), not adjusted. See §"Koide-α identity" below.

Third prediction (parameter-free for canonical $V_A$, v1.3.4). Selecting the canonical modular-invariant potential $V_A(\tau, \bar\tau) = \Lambda^4 |j(\tau)|$ — the simplest modular-invariant scalar built directly on the Klein $j$-invariant — the Morse structure on the compactified orbifold $\bar X(1)$ is topologically forced by the maximum modulus principle (Ahlfors 1966) and the minimum principle on zeros (Conway 1978): $\rho$ is the global minimum ($j(\rho) = 0$), $i$ is the saddle (Stab(i)=ℤ₂, $j(i) = 1728 \neq 0$), and the cusp $\infty$ is the maximum. Configuration A (not a choice, a structural inevitability for $V_A$) makes the reheating trajectory a free fall from the saddle at $i$ to the attractor at $\rho$; the e-folding count is the topological index $[\mathrm{PSL}(2,\mathbb{Z}) : \Gamma(5)] = |A_5| = 60$, equal by Klein–Galois to the hexagonal cascade count around $\rho$. This yields the parameter-free CMB pair

$$n_s ;=; 1 - \tfrac{2}{60} ;\approx; 0.9667, \qquad r ;=; \tfrac{8}{60^2} ;=; \tfrac{2}{900} ;\approx; 0.00222,$$

with no observational input and no slow-roll fit. The one-line argument is given in theorem-4-m1-dyn-uzavreni.md §III.5; the residual gap $M1_{\text{dyn_b}}$ (rigorous $A_5$ orbit count on $X(5)$) is purely technical (1–2 weeks of SageMath). Prediction (c) in §"Predictions" converges numerically with prediction (a) to $\sim 10^{-4}$: the two are an internal consistency check, not independent tests.

---

Meta-postulate M0 (structural realism)

M0. Physical observables are functions on isomorphism classes of physical states, not on their representatives. Equivalently: if two configurations differ only by an isomorphism of their internal structure, no physical measurement distinguishes them.

M0 is the meta-level assumption under which all three axioms below operate, and it is the assumption from which Theorem 1 derives PSL(2,ℤ)-invariance. It is strictly weaker than the v1.0 assertion "PSL(2,ℤ) is a symmetry group": M0 does not specify which symmetry holds, only that physics sees structure rather than labels on representatives. The same meta-postulate is used (implicitly) throughout gauge theory, where physical observables are required to factor through gauge orbits rather than distinguish gauge-equivalent configurations.

On counting. The framework has three axioms (A1, A2, A3) plus one meta-postulate (M0). In v1.0 the corresponding count was four axioms. The reduction 4 → 3 is conditional on explicitly naming M0; the total number of fundamental assumptions is unchanged. The move is qualitative: M0 is a strictly weaker hypothesis than full PSL(2,ℤ)-invariance, and the content of PSL(2,ℤ)-invariance is now derived from A1, A2, and M0 rather than postulated.

---

The three axioms

A1. Complex time. τ ∈ ℍ² = {τ ∈ ℂ : Im(τ) > 0}.

Physical motivation for A2

A2 is a structural identification in the regime of M0 (structural realism), not a dynamical reduction from QED. We do not derive the photon as a specific field excitation of a pre-existing spacetime Lagrangian; we identify, at the level of admissible topological types compatible with A1 + M0, the minimal content that deserves the label "photon".

Three strands of motivation converge on the elliptic-curve identification:

1. Closed self-referential worldsheet. In the pre-axiomatic picture, a photon is what it is because its phase accumulates on itself: it "experiences length" as accumulated phase Φ over a closed self-referential structure (substrate = generation-0 of the self-reference chain). See foton-proziva-delku.md and seberference-x-na-x.md, where x^x is read as the kernel of self-reference on which the photon is the degree-0 example.

2. Toroidal topology. The elliptic curve 𝓔_τ ≅ ℂ/(ℤ + ℤτ) is, up to biholomorphism, the simplest closed Riemann surface (genus 1) compatible with (a) a global U(1) gauge structure (non-trivial first cohomology H¹ ≅ ℤ²), (b) a single complex modulus τ (one continuous parameter), and (c) no distinguished basepoint. Genus 0 has no modulus; genus ≥ 2 carries 3g − 3 > 1 moduli. Genus 1 is forced by minimality + non-triviality.

3. Opposite arrow (important for the QED direction). In this framework A2 is not quantised from QED. Instead, quantum mechanics flows out of τ-geometry: Schrödinger, Born, Heisenberg are derived as consequences of A1 + A2 + M0 + A3 in the appropriate limit. See Appendix A for derivation of QM from τ-geometry — this establishes that photon quantization need not be imposed externally on A2. Extended research note in qm-limit-p3.md.

Consequently A2 is to be read as a definition of the photonic object inside the τ-framework, not as a theorem about the photon of perturbative QED. The two should agree in the flat, weak-coupling limit — this agreement is what P3 (QM limit) establishes — but the logical arrow runs from τ to QED, not the reverse.

A2. Photonic space. Space is the real projection of phase rotation in τ. One 2π rotation together with a quaternionic companion direction constitutes one photon; spatial dimensions arise from the quaternionic extension of τ. Topologically, this identifies the photon with S¹ × S¹ equipped with complex structure τ — i.e., an elliptic curve.

A3. Observer projection (reformulated 2026-04-19 as A3 ≡ S-reflection). A classical observer measures the Stab(i)-invariant of the modular geometry. Along the canonical geodesic τ₁ = 0 this is the signed hyperbolic distance from i:

$$\varphi(\tau) = \ln|\tau_2| = d_{\mathbb{H}^2}(\tau, i)\cdot\mathrm{sgn}(\tau_2 - 1)$$

The naive formulation "the observer measures Re(τ)" is equivalent to this along the geodesic through i (where Re(τ) = 0 and φ is the next meaningful coordinate), but is not PSL(2,ℤ)-invariant and so is ill-defined under M0 globally. The reformulated A3 commits only to the Stab(i) = ⟨S: τ → −1/τ⟩ subgroup — a ℤ₂ reflection — which is compatible with M0 + Theorem 1 and uniquely selects the T-model effective potential (see §Effective action). Details in p8-posledni-odraz-do-i.md §VIII and kauzalni-stred-lokalni-i.md.

(The former A3 of v1.0 — modular invariance under PSL(2,ℤ) — has been demoted to Theorem 1, derived below from A1 + A2 + M0.)

---

Theorem 1: Modular invariance

Claim. Given A1, A2, and M0, physical observables are invariant under the action of PSL(2,ℤ) on ℍ².

Proof sketch.

Step T.1 — Photon = elliptic curve. A2 identifies the photon with a closed 2π rotation (one circle S¹) plus a quaternionic companion direction (a second, independent circle S¹). The topological product is a two-torus S¹ × S¹. Equipped with the complex structure inherited from τ ∈ ℍ², this torus is an elliptic curve 𝓔_τ ≅ ℂ/(ℤ + ℤτ).

Step T.2 — Moduli space. Classical theorem of the theory of modular forms (e.g. Diamond & Shurman, A First Course in Modular Forms, §1): the moduli space of complex-structured elliptic curves over ℂ is ℍ²/PSL(2,ℤ). Two parameters τ, τ' ∈ ℍ² yield isomorphic elliptic curves if and only if they lie in the same PSL(2,ℤ) orbit.

Step T.3 — Apply M0. By M0, a physical observable is a function on isomorphism classes, not on representatives. Two τ, τ' yielding isomorphic photons necessarily yield the same observable value — otherwise the observable would be labelling representatives rather than measuring structure, in direct violation of M0.

Conclusion. Physical observables factor through the quotient ℍ² → ℍ²/PSL(2,ℤ) and are therefore invariant under T: τ → τ+1 and S: τ → −1/τ, the generators of PSL(2,ℤ). □

Remark. This replaces the postulated A3 of draft v1.0 with a derivation from A1 + A2 + M0. The substantive content of the move is that M0 is strictly weaker and more general than the assertion "PSL(2,ℤ) is a symmetry group": M0 commits only to observables-see-structure, not to any particular symmetry group. The specific group PSL(2,ℤ) then arises as the automorphism group of the moduli space of elliptic curves, which is forced by A2. See proc-psl2z.md for a longer discussion including two visualizations (the "donut on a plate" analogy for moduli space and the "hot room with cold floor" analogy for the ℍ² structure).

---

Derivation chain

Step 1: K = −1 from Theorem 1 + canonical kinetic assumption

Theorem 1 establishes that observables are PSL(2,ℤ)-invariant. However, this alone does not uniquely select the Poincaré metric: one could multiply ds² = (dx² + dy²)/y² by any modular-invariant scalar function (e.g. a function of Klein's j-invariant) and retain PSL(2,ℤ)-invariance while altering the curvature. We therefore make explicit the additional assumption under which Theorem 1 leads to α = 2/3:

Assumption K (canonical kinetic structure). The effective field-space kinetic term is the unique PSL(2,ℝ)-invariant metric on ℍ², i.e. the Poincaré metric

$$ds^2 = \frac{dx^2 + dy^2}{y^2}.$$

Justification. The α-attractor class (Kallosh–Linde 2013) is defined by a canonical kinetic Lagrangian of the form L_kin = −½ (∂τ)(∂τ̄)/(Im τ)² — i.e., precisely the Poincaré metric. Any modular-invariant rescaling (by a function of j) would produce a non-canonical kinetic term outside the α-attractor class. Assumption K states that the effective theory lies in the α-attractor class, which is the minimal assumption compatible with a supergravity-motivated kinetic sector. It is strictly weaker than a full field-theoretic derivation from a microscopic Lagrangian (which we do not attempt here), and it is what allows the α = 2/3 conclusion to follow from Theorem 1.

Uniqueness (Poincaré uniqueness lemma, 2026-04-19). Once the effective kinetic tensor is additionally required to be a complete Riemannian metric of constant Gaussian curvature and finite volume on X(1) — call this combined condition Assumption K′ — the Poincaré metric is unique up to overall scaling. The proof invokes the classical uniformization theorem (Ahlfors–Sario 1960; Farkas–Kra 1992; Beardon 1983) together with the fact that the Teichmüller space of orbifolds of signature (0; 2, 3, ∞) is a single point (Thurston 1980; Farb–Margalit 2012). This closes the logical gap raised by external review N2 (arbitrary modular-invariant conformal rescalings such as multiplication by a function of the j-invariant). Full statement and proof in lemma-poincare-uniqueness.md.

Reduction of Assumption K′ to Assumption K (2026-04-19, §N of the lemma). Two of the three conditions in Assumption K′ — completeness and finite volume — are derivable from the third (constant curvature K ≡ −1) together with A1 and the topology of X(1). Completeness follows from a Liouville rigidity argument at the cusp (the unique modularly invariant solution of the Liouville equation Δφ = e^{2φ} with cusp asymptotics is the Poincaré form, for which ∫dy/y diverges). Finite volume follows from the orbifold Gauss–Bonnet theorem applied to χ_orb(X(1)) = −1/6, giving Vol = π/3. Thus Assumption K′ collapses to Assumption K (canonical Poincaré kinetic term, equivalently K ≡ −1) without loss.

With Assumption K, the Poincaré metric has constant Gaussian curvature K = −1 as a theorem of hyperbolic geometry (not an assumption).

Open problem P11 (narrowed, §O of the lemma). Derive Assumption K from A1 + A2 + M0 alone. After the §N reduction, the residual gap is a single bit — choice of f(τ) ≡ 1 in the general modular-invariant conformal family f(τ)·δ_{IJ}/τ₂². In §O of lemma-poincare-uniqueness.md five candidate variational principles were examined rigorously: (1) geodesic completeness — fails (completeness holds for any smooth positive f); (2) Liouville action — circular (requires K = −1 as input); (3) Kähler structure from A2 — insufficient (in complex dimension 1 any Hermitian metric is Kähler); (4) Weil-Petersson metric on M_{1,1} — succeeds rigorously (classical result of Wolpert 1985, Imayoshi–Taniguchi 1992, Zograf–Takhtadzhyan 1987: the WP metric on the moduli space of elliptic curves equals the Poincaré metric up to global scale, with no modular-invariant conformal freedom); (5) Fuchsian uniformization — duplicates §II. Path (4) closes P11 conditional on a strengthening A2⁺: „the field-space kinetic tensor is (up to scale) the Weil-Petersson metric on M_{1,1}." A2⁺ is a theorem in stringy/supergravity reductions on elliptic fibres and a one-line geometric postulate in KRT. With A2⁺ accepted, f ≡ 1 is a theorem, not a postulate; the residual gap reduces from „choice of function f(τ)" to „acceptance of A2⁺." We keep P11 flagged 🟡 conservatively in line with disciplinary Rule 7 (explicit naming of residual assumptions).

Step 2: Area(𝒻) = π/3

The fundamental domain 𝒻 of PSL(2,ℤ) has hyperbolic area π/3, from Gauss-Bonnet applied to the orbifold X(1) with Euler characteristic χ(X₁) = −1/6.

Step 3: α = 2/3

The α-attractor field-space curvature is K_fields = −2/(3α). Setting K_fields = K = −1:

$$\alpha = \frac{2}{3} = \frac{2,\mathrm{Area}(\mathcal{F})}{\pi} = -4\chi(\mathcal{X}_1)$$

α is a topological invariant. It cannot be changed by smooth deformations preserving the orbifold topology.

Step 4: r = 2(1 − nₛ)²

From the universal α-attractor formulae, eliminating Nₑ:

$$r = \frac{8}{N_e^2}, \quad n_s = 1 - \frac{2}{N_e} \quad\Rightarrow\quad r = 2(1-n_s)^2$$

For nₛ = 0.9649 (Planck 2018): r ≈ 0.0025, Nₑ ≈ 57.

The e-folding count Nₑ enters here as an observational input, but is a candidate for derivation rather than a free parameter: Nₑ = 60 = |A₅| = [PSL(2,ℤ) : Γ(5)] is the index of the principal congruence subgroup at level N = 5, the maximal spherical level of the Coxeter triple (2, 3, 5) that extends the elliptic orders (2, 3) of X(1). Status 🟡 — three candidate paths (Γ(5)-index, icosahedral closure of the (2,3) stabilizers, Poincaré homology sphere S³/A₅ as the compact arena of Theorem 3); none is yet rigorous. See ne-60-derivace.md.

---

Effective action

Kinetic sector (fundamental, PSL(2,ℤ)-invariant)

The kinetic term, derived from A1 + A2 + M0 via the Poincaré metric on ℍ², is:

$$\mathcal{L}_{\text{kin}} = -\frac{1}{2}\frac{(\partial\tau_1)^2+(\partial\tau_2)^2}{\tau_2^2}$$

This is the standard hyperbolic metric on ℍ² with Gaussian curvature K = −1 (Theorem 1). In α-attractor terminology, this corresponds to α = 2/3. The kinetic term is fully PSL(2,ℤ)-invariant.

Canonical modular potential $V_A$ (new, v1.3.4)

The simplest fully modular-invariant scalar built directly on the Klein $j$-invariant is

$$V_A(\tau, \bar\tau) ;=; \Lambda^4 \cdot |j(\tau)|,$$

or any positive power $|j|^\alpha$ with $\alpha > 0$. This potential is smooth on $X(1) \setminus {\infty}$ ($j$ is holomorphic on $\mathbb{H}$ with a pole only at the cusp) and invariant under the full PSL(2,ℤ) action (by construction of $j$).

Morse structure forced by complex analysis. Because $j$ is non-constant holomorphic on $\mathbb{H}$, two elementary principles fix the Morse structure of $V_A$ on the compactified orbifold $\bar X(1)$ without any Hessian computation:

• Maximum modulus principle (Ahlfors 1966 §4.3): $|j|$ has no local maximum in the open fundamental domain;
• Minimum principle on zeros (Conway 1978 §IV.6.6): $|j|$ attains a local minimum only where $j = 0$.

Combined with $j(\rho) = 0$ (from $E_4(\rho) = 0$, Serre 1970 §VII.3.3), $j(i) = 1728 \neq 0$, and $|j(\tau)| \sim e^{2\pi \operatorname{Im}\tau}$ at the cusp, these principles force

$$\operatorname{ind}(\rho) = 0, \qquad \operatorname{ind}(i) = 1, \qquad \operatorname{ind}(\infty) = 2,$$

satisfying the orbifold Morse identity $-\tfrac12 + \tfrac13 + 1 = \tfrac56 = \chi_{\text{orb}}(\bar X(1))$ (Satake 1956). Configuration A (saddle at $i$, global attractor at $\rho$) is topologically inevitable for $V_A$ — not one of two valid Morse configurations, but the only one. Configuration B (two degenerate minima, Model-A from earlier drafts) would require additional zero structure beyond $j$ itself, which the principles exclude for $V_A$.

Physical reading. The inflationary geodesic $\tau_1 = 0$ descends from the cusp through the saddle at $i$; the T-model below describes the local oscillation around this metastable saddle (not around a minimum). Reheating is the topologically forced free fall from $i$ to $\rho$ along the stable manifold of the saddle, with no tunnelling barrier and no ad-hoc coupling to Standard-Model matter. The e-folding count $N_e = 60$ is the combinatorial index $[\mathrm{PSL}(2,\mathbb{Z}) : \Gamma(5)]$ along the inflationary geodesic; the cascade count around $\rho$ is $6 \times 10 = 60$ by Klein's spherical picture (Klein 1884). Both numbers are the same topological quantity $|A_5|$ read along two complementary directions of the same $X(5) \to X(1)$ covering. Full argument: theorem-4-m1-dyn-uzavreni.md §III.5 (Morse inevitability) + §IV (cascade identification).

Why this is not a post hoc choice. $V_A = \Lambda^4 |j|$ contains one energy scale $\Lambda$ and no other parameters. It is the minimal modular-invariant construction and the unique one for which the Morse structure follows from elementary complex analysis rather than from a Hessian calculation. Alternative modular-invariant potentials (e.g. $|E_6|^2 |E_4|^2 / \tau_2^{10}$ with double zero structure) require explicit Hessian analysis and may realise Configuration B; the scope of canonical choice is discussed in §"Open problems" (item B5'). The argument is honest about its conditional nature: for the canonical $V_A$, Configuration A is forced; alternative canonicity claims are an open question about which modular-invariant construction correctly represents KRT physics at the paper-level.

Effective inflaton potential (T-model)

Efektivní potenciál. Inflační dynamika je popsána efektivním potenciálem

$$V_{\text{eff}}(\varphi) = \Lambda^4\tanh^2!\left(\frac{\varphi}{2}\right),$$

interpretovaným jako single-field limit plné PSL(2,ℤ)-invariantní akce na moduli prostoru X(1). Derivace tohoto efektivního potenciálu z mikroskopické modular-invariant akce (plné modulární invariance v ψ = {τ, ∂_μ τ}) zůstává otevřeným problémem (Open Problem B4; viz problem-potencial.md a eom-odvozeni.md).

V tomto efektivním potenciálu je φ = ln τ₂ (pro τ₁ = 0) kanonický inflaton, identifikovaný s hyperbolickou vzdáleností od τ = i v Poincarého metrice. V(φ) je canonical α = 2/3 T-model (Kallosh–Linde 2013). Plná efektivní akce je:

$$S_{\text{eff}} = \int d^4x\sqrt{-g}\left[\frac{M_\mathrm{Pl}^2}{2}R - \frac{1}{2}\frac{(\partial\tau_1)^2+(\partial\tau_2)^2}{\tau_2^2} - \Lambda^4\tanh^2!\left(\frac{\varphi}{2}\right)\right]$$

kde kinetický sektor je plně PSL(2,ℤ)-invariantní (K = −1, Theorem 1), zatímco potenciální sektor je Stab(i)-kovariantní (T-model selected axiomaticky A3 ≡ S-reflection).

Numerické ověření. Pro N_e = 57 dává tento efektivní potenciál r = 0.00239 (viz slow-roll-numericky.md), v souladu s topologickou predikcí r = 2(1−n_s)² ≈ 0.0025. Relace r = 2(1−n_s)² platí numericky s přesností ~2 % pro N_e ≥ 40.

Important clarification

A previous version of this paper (v1.1.6 and earlier) specified V = Λ⁴ tanh²(H(τ)/2) with H(τ) = −ln[τ₂|η(τ)|⁴], using the Dedekind eta function. This is fully PSL(2,ℤ)-invariant but does not reproduce r ≈ 0.0025 — direct numerical analysis shows it gives r ~ 10⁻⁵ due to doubly-exponential approach to plateau (see problem-potencial.md).

The T-model V = Λ⁴ tanh²(φ/2) is stabilized by the subgroup Stab(i) = Z₂ (elliptic point of order 2 under PSL(2,ℤ)), not the full modular group. This is a direct consequence of the reformulation A3 ≡ S-reflection: the classical observer measures a Stab(i)-invariant, so the effective potential is necessarily Stab(i)-covariant rather than fully PSL(2,ℤ)-invariant. Under A3, the T-model is the unique simplest choice — not an empirical fit.

An earlier conjecture that the T-model might be derived from a fully PSL(2,ℤ)-invariant potential via modular averaging has been ruled out by explicit computation: see b4-modularni-prumerovani.md, which shows that plain modular averaging cannot convert the double-exponential H(τ) of the η-function-based potential into the linear canonical field φ of the T-model. Consequently, the T-model is accepted as the effective potential selected by A3, not as a derivation from a larger modular structure. The previously listed "Open problem 8 (P8) — derive T-model from modular averaging" is therefore declared closed: negatively (the derivation does not exist), with the positive statement being that A3 axiomatically selects the Stab(i)-effective theory.

For the purposes of this paper — the prediction r = 2(1−n_s)² ≈ 0.0025 — the T-model is the correct effective potential, and the α = 2/3 value follows from the kinetic sector (K = −1 from Theorem 1), which is fully invariant.

---

Model comparison

Model	α	3α	r (nₛ = 0.965)
τ-geometry (this work)	2/3	2	0.0025
Starobinsky R²	1	3	0.0037
KL attractor (3α = 7)	7/3	7	0.0086
KL attractor (3α = 1)	1/3	1	0.0012

Key: our value 3α = 2 is absent from the Kallosh–Linde M-theory benchmark list. It corresponds to the IIB axiodilaton sector, derived here from the topology of the modular parameter space.

---

JWST Little Red Dots

As an independent observational hook: LRDs are interpreted as first-order phase-domain relics of the Im-dominant phase of early-universe τ-geometry. The Compton-thick obscuration = domain wall (φ-gradient boundary). Predicted number density:

$$n_\mathrm{LRD}(z) \propto \mathrm{erfc}!\left(\frac{\phi_\mathrm{crit}-\bar\phi(z)}{\sigma_\mathrm{domain}}\right)$$

Nucleation rate Γ ∝ e^{−π/3} ≈ 0.35, consistent with ~1% observed LRD abundance.
This prediction is testable now with existing JWST/RUBIES and JADES data.

---

Interpretation: relational realism, no parallel worlds

The reformulation A3 ≡ S-reflection has a sharp interpretational consequence that distinguishes this framework from many-worlds and multiverse-type interpretations of quantum mechanics and cosmology. We state it explicitly:

Interpretational postulate. The substrate algebraically contains multiple reflections (Z/2 at i, Z/3 at ρ, Z/N at higher-level cosets); the classical observer projects onto a single Stab(i)-invariant slice. The non-selected reflections are not located in parallel branches, alternate universes, or modal worlds — they are structural facts of one reality, accessible through group-theoretic transitions (weak-force flavour change, parity reversal, etc.) within the same observer's relational network.

Concretely:

• The three fermion generations (e, μ, τ; u, c, t; d, s, b) are the three Z/3-rotational positions at ρ of the same fermionic mode — not three particles in three worlds. Transitions between them (μ → e via weak decay, CKM/PMNS mixing) are mediated by the centre Z(B₃) ↔ weak force, within the single observed reality.
• Quantum superposition |ψ⟩ = α|↑⟩ + β|↓⟩ represents the algebraic Z/2 structure at i; the Born rule |α|² emerges in the qm-limit derivation as the measure of how deeply the ↑-reflection is imprinted in the Stab(i)-projection. Both outcomes are structurally present; only one is observed. Neither is "elsewhere".
• The cosmological S³ physical space (Theorem 3) is the whole spatial content accessible to an observer; there is no multi-S³ landscape, no inflationary multiverse branching. The modular tower X(N) describes levels of substrate complexification relative to one projection, not separate universes.

This is relational monism: one substrate, one projection per observer, all algebraic reflections compactified into the same reality as structural facts of its group theory. It is weaker than modal realism (Lewis), stronger than Copenhagen (no collapse as irreducible postulate — collapse is A3 projection), and strictly incompatible with Everett many-worlds, eternal-inflation multiverse, and string-landscape pluralism.

Under this interpretation, the theory's predictions (r = 0.0025, sin²θ_W(GUT) = 1/3, N_e derivation toward 60) are unique to our universe, not anthropically selected from a larger ensemble. If they are correct, they describe the structure of the one reality the observer inhabits.

Open problems

1. End of inflation and emergence of three fermion generations — requires two-field analysis (τ₁, τ₂) near the ρ-point. Under the v1.1 framing, the ρ-point is the Z₃ elliptic fixed point of the modular group, and its two-field analysis is simultaneously the mechanism for the emergence of three fermion generations (cf. tier 3.3 of the roadmap). Under Conjecture 2 (force attribution, originally introduced as candidate Theorem 2 in v1.1.2; see tri-mista-ctyri-sily.md), the ρ-point is further identified as the topological source of the strong SU(3) sector, of which the three-generation structure is one manifestation.

Theorem 3 (curvature from quaternion frustration, proved 2026-04-18). Three anticommuting vector fields X_a with [X_a, X_b] = 2ε_{abc}X_c generate a 3-manifold with constant Ricci tensor Ric_ab = 2δ_ab and scalar curvature R = 6. This manifold is S³ = SU(2) with bi-invariant metric. Flat space is algebraically excluded. Proved by direct Ricci computation via the Killing form B_ab = -8δ_ab and the standard formula Ric = -B/4 for bi-invariant metrics on compact Lie groups. Detailed exposition in ricci-theorem-3.md. Complementary to Theorem 1 (field space ℍ², K = -1): physical space S³, K = +1, total curvature zero.

P3 resolved (QM from τ-geometry, 2026-04-18). Schrödinger equation, Born rule, and Heisenberg uncertainty are derived consequences of τ-geometry, not postulates. Quantum state ψ = e^(iS/ℏ) is the accumulated Im(τ) phase; Schrödinger iℏ∂ψ/∂t = Hψ is Hamilton-Jacobi read in this variable; Born rule |ψ|² follows from projection A3 (observer measures Re(τ)); complementarity x↔p is Fourier duality of Re↔Im projections of e^(iΘ). Full derivation in qm-limit-p3.md.

Critical issue — potential specification (2026-04-18). Numerical analysis reveals that the fully PSL(2,ℤ)-invariant potential V = Λ⁴ tanh²(H(τ)/2) with H = -ln(τ₂|η|⁴) specified in Eq. (115) does NOT reproduce the prediction r = 2(1-n_s)² ≈ 0.0025 — it gives r ~ 10⁻⁵ due to doubly-exponential approach to plateau. The standard α = 2/3 T-model potential V = Λ⁴ tanh²(φ/2) (where φ = hyperbolic distance from τ = i) does give r ≈ 0.0024 for N_e = 57, matching the paper's prediction. Paper requires revision to specify T-model as effective potential, with fully modular-invariant description as higher-level open problem. Details in problem-potencial.md, numerical verification in slow-roll-numericky.md.

Λ_eff from Theorem 3 (2026-04-18). With Theorem 3 giving spatial S³ with R^(3) = 6 (M_Pl units), the induced 4D cosmological constant satisfies Λ_eff/M⁴_Pl = 3/(M_Pl·a_S³)², where a_S³ is the S³ radius. For a_S³ of order Hubble radius today (c/H₀), this gives Λ_eff/M⁴_Pl ~ 10⁻¹²² — order-of-magnitude match with observed cosmological constant. The hierarchy problem Λ_obs/Λ_Planck ~ 10⁻¹²² is reduced to: why is the S³ radius so large? (Physically meaningful question about universe age/expansion, not fine-tuning.)

P1 partially resolved (EOM derived 2026-04-18). Euler-Lagrange equations for the effective action above are explicit: τ̈₁ + 3Hτ̇₁ - (2/τ₂)τ̇₁τ̇₂ + τ₂²∂V/∂τ₁ = 0 and analogous for τ₂. Friedmann equation 3H²M²_Pl = (1/2τ₂²)(τ̇₁² + τ̇₂²) + V. In canonical variable φ = ln(τ₂) with τ₁ = 0, reduces to standard slow-roll α-attractor equation. The identification K_fields = K = -1 is derived, not assumed: the kinetic term IS the Poincaré metric whose Gaussian curvature is K = -1 by Theorem 1. Prediction r = 2(1-n_s)² now follows from EOM, not only from universal α-attractor formulae. Detailed derivation in eom-odvozeni.md.

Theorem 2 (Fuchsian completeness, proved 2026-04-18). Given A1, A2, M0, and Theorem 1, the set of PSL(2,ℤ)-orbit types on ℍ² augmented by the global orientation datum has exactly four elements. This is a consequence of (a) the classification theorem for finite stabilizers in Fuchsian groups (Katok 1992), (b) the integrality of trace for M ∈ SL(2,ℤ), which restricts elliptic orders to {2, 3} in PSL(2,ℤ), (c) the uniqueness of the cusp class, and (d) the two global orientation choices. The four topological data are: cusp (parabolic, infinite stabiliser), i (elliptic, Z₂), ρ (elliptic, Z₃), orientation. Completeness (no fifth mode) is proved in theorem-2-completeness.md. Theorem 2 as stated here does not attribute specific SM forces to these modes; it is purely a statement about orbit-type combinatorics of PSL(2,ℤ).

Conjecture 2 (force attribution). The four fundamental forces of the Standard Model are identified with the four topological data of Theorem 2 by the natural assignment: gravity ↔ cusp, electromagnetic ↔ i, strong ↔ ρ, weak ↔ orientation. This assignment is motivated by physical characteristics (long-range ↔ cusp infinity, chirality ↔ orientation, triple structure of colour ↔ Z₃ at ρ, neutrality under reflection ↔ Z₂ at i) and is consistent with the B₃ central extension derivation in attribution-pletence-b3.md. Conjecture 2 is structurally natural, not formally derived; its promotion to a theorem is an open problem. Detailed exposition with honest flags in tri-mista-ctyri-sily.md.

2. Nₑ from τ-geometry — candidate: Nₑ = 60 = |A₅| = |PSL(2,ℤ/5ℤ)|, currently numerological. Under the "complexification as climbing the modular tower" reframing (see fotony-stavaji-se-komplexnimi.md), this becomes the depth of the modular tower required to reach the first habitable cover.

3. Reheating temperature — coupling τ → SM through Cayley–Dickson structure, not yet quantified. A candidate structural answer in v1.1.3 (see cerna-dira-jako-pochopeni.md): emergence of the many-body structure from a single rotation is not a dynamical process but a consistency requirement of the relational substrate — an isolated configuration with a nontrivial moment of rotation is logically underdetermined ("a direction without a second point"), and the substrate resolves this by forcing the other end of the relation into existence. If this reframing survives formalization, emergence itself becomes a non-problem rather than an open dynamical question, analogously to how Theorem 1 (modular invariance) became a theorem rather than an axiom.

4. Cosmological constant — V(ρ) ≈ 0.11Λ⁴ ≠ 0, fine-tuning not avoided.

5. Metric structure of the complexification arrow (companion essays). The metric structure of the complexification arrow — x^x as self-reference kernel, i^i as projection loss, CF band structure for fundamental constants (P5), and preaxiom formalisation via CCC / Lawvere / Yanofsky (P6) — is developed in the companion essays seberference-x-na-x.md, rez-a-prvni-chyba.md, and seberference-kategorie.md. These are structural commentaries on the preaxiomatic layer beneath M0; they are not part of the core derivation of r = 2(1−n_s)² and are not prerequisites for any formal result quoted in this paper. They are noted here so that readers interested in the broader self-reference programme can locate them.

6. P7 — Dynamical stabilisation of the S³ radius (open). Theorem 3 plus Λ_eff ~ 10⁻¹²² is a consistent order-of-magnitude match, but no dynamical mechanism is yet known that shows why the S³ radius today is precisely the Hubble radius. The hierarchy reduction "Λ_obs/Λ_Planck ~ 10⁻¹²² = (a_S³·M_Pl)⁻²" recasts the cosmological-constant problem as an age/expansion question rather than a fine-tuning in the Lagrangian, but a non-trivial stabilisation mechanism (attractor, back-reaction, or topological constraint fixing a_S³ ~ c/H₀) is deferred to companion work. Reheating dynamics and a T-model microscopic derivation also remain open (P8-dynamics); see 00-roadmap.md.

7. Fermion mass spectrum and generation structure — not yet addressed; depends on Conjecture 2 coupling mechanism and P4.

8. P8 — CLOSED (2026-04-19 night). The T-model effective potential V(φ) = Λ⁴ tanh²(φ/2) is the Stab(i)-invariant simplest effective potential selected by axiom A3 ≡ S-reflection. An attempted derivation from a fully PSL(2,ℤ)-invariant potential via modular averaging has been ruled out by computation (b4-modularni-prumerovani.md): the double-exponential structure of the η-function potential cannot be flattened to a linear argument by any plain averaging. The T-model is accepted as postulate consequence of A3, not as derivation from a larger structure. This is the only coherent interpretation consistent with the reformulated A3. The α = 2/3 value is independent — it follows from the kinetic sector (K = −1, Theorem 1 + Assumption K).

9. Lepton mass ratios (Koide 2/3 ≡ α-attractor 2/3) — PROMOTED to §"Koide-α identity" in v1.3.0. The Koide relation Q = 2/3 is now a derived result of the framework (see the dedicated section above) and a second hard prediction of the paper (m_τ = 1776.97 MeV vs. measured 1776.86 MeV). Quark and neutrino extensions remain open; these are the residual open items in the flavour sector.

10. P11 — Derive Assumption K (canonical Poincaré kinetic term) from A1+A2+M0. Theorem 1 alone does not uniquely select the Poincaré metric; an additional premise (canonical α-attractor kinetic structure) is required. Narrowed 2026-04-19 (§N of lemma-poincare-uniqueness.md): completeness and finite volume are now derivable from K ≡ −1 + A1 (Liouville rigidity at the cusp; orbifold Gauss–Bonnet). The residual gap is a single bit — the conformal factor f(τ) in the most general PSL(2,ℤ)-invariant kinetic tensor f(τ)·δ_{IJ}/τ₂² must be shown ≡ 1. Candidate approach: 2D CFT / supergravity reduction on elliptic moduli. See Step 1 of the Derivation chain above.

11. B4 — Derivace T-modelu z plně modulární akce (🟡). Ukázat, že v efektivní single-field limitě plné PSL(2,ℤ)-invariantní akce se získá T-model V_eff(φ) = Λ⁴ tanh²(φ/2). Motivace: paper v1.3.1 specifikuje T-model jako efektivní potenciál (Cesta B z problem-potencial.md), neboť fully modular-invariant volba V = Λ⁴ tanh²(H(τ)/2) s H = −ln[τ₂|η(τ)|⁴] nedává r ≈ 0.0025 (plain modular averaging vyloučeno výpočtem v b4-modularni-prumerovani.md). B4 hledá alternativní cestu — může zahrnovat integraci nad τ₁ (transverzální moduli), Rademacher-type sumy, supergravitační redukci, nebo techniku mimo modular averaging. Odhad: 2–6 týdnů. Status tohoto problému je 🟡 — otevřený, nikoli téměř vyřešený; numerická predikce r = 0.00239 (N_e = 57) není B4 podmíněna, neboť se vyhodnocuje přímo na efektivním potenciálu V_eff. Viz problem-potencial.md, eom-odvozeni.md, slow-roll-numericky.md. Update v1.3.4/1.3.5: s kanonickou volbou $V_A = \Lambda^4 |j(\tau)|$ (§"Canonical modular potential $V_A$") je T-model identifikován jako lokální oscilace kolem metastabilního sedla $i$ v Konfiguraci A, ne kolem minima; B4 se tím mění z „derivace T-modelu z modulární akce" na „ukázat, že lokální oscilace v okolí sedla $V_A$ dává tanh² formu po redefinici kanonického pole $\varphi$" — strukturně čistší pod-úkol s rutinní matematickou prací.

12. P22 — Apendix B: GR-projekce substrátu π_2 (🟢 core, v1.3.9; M1+M2+M3+M4+M5 closed 2026-04-21 late evening +5h). Following an external peer reviewer's correct observation that this paper's „theory stands at the level of QM + GR unification" claim lacked a companion to Appendix A, five of six structural gaps are now closed in <6 hours of work (1 main Claude + 4 parallel agents): M1 rigorous Einstein equations from variation (p22-m1-einstein-equations.md); M2 diffeomorphism invariance as consequence of M0 not axiom (p22-m2-diffeomorphism-invariance.md); M3 EEP = WEP + LLI + LPI fully derived via $Y_{0,0}$ single source (p22-m3-equivalence-principle.md); M4 Schwarzschild as vacuum $\tau = \rho$ limit plus τ-hair modification (p22-m4-schwarzschild-tau-hair.md); M5 gravitational waves as spin-2 ℍ-modes on S³ with polarization $h_+, h_\times$ as $i, j$ quaternionic rotations (p22-m5-gravitacni-vlny.md). Plus scope canonical choice (scope-canonical-choice-VA.md): $V_A = \Lambda^4 |j(\tau)|$ justified as KRT-native preferred candidate via Occam + i/ρ asymmetry + direct Klein j-invariant. M6 (coherent Appendix B write-up to paper) is the remaining routine task; formal sub-gaps M2.a (categorial derivation via P6 functor), M3.a (SEP $Y_{0,0}$-preservation), M4.a (explicit $\tau(r)$ ODE), M4.b (Hawking exact), M5.a (Weil rep for spin-2) are named separate technical sub-problems (2–6 months each). Claim status upgrade: the earlier „aspiration plus three bridges" has been replaced by „rigorous derivation of Einstein equations, diffeomorphism invariance, equivalence principle, Schwarzschild limit, and gravitational-wave polarization structure from the same PSL(2,ℤ)-invariant action that governs inflation". The Appendix A (QM) and Appendix B (GR) thereby stand on equal structural footing; per Rule 2, the paper is now honestly positioned as an inflationary α-attractor model whose QM and GR limits are both rigorously derived from a single action. See p22-apendix-b-plan.md for the master plan.

13. Theorem 4 — topologická derivace $N_e = 60$ pro kanonický $V_A$ (plně 🟢, v1.3.6). Pro $V_A = \Lambda^4 |j(\tau)|$ je Konfigurace A (i sedlo, ρ minimum, ∞ maximum) topologicky vynucena principem maxima modulu (Ahlfors 1966) + principem minima-na-nulách (Conway 1978) — viz §"Canonical modular potential $V_A$" výše a theorem-4-m1-dyn-uzavreni.md §III.5. $N_e = 60 = [\mathrm{PSL}(2,\mathbb{Z}) : \Gamma(5)] = |A_5|$ topologicky; $r = 8/60^2 = 2/900$ plně parameter-free (prediction (c)). Obě technické mezery uzavřeny 2026-04-21: $M1_{\text{dyn_a}}$ (Morse structure) přes princip maxima modulu; $M1_{\text{dyn_b}}$ (cascade count) přes kombinatorickou identitu $A_5$: 20 elementů řádu 3 ÷ 2 = 10 os, Klein 1884 spherical picture identifikuje s 10 vrcholy řádu 3 chirální (2,3,5) Schwarzovy tesselace, 6 trojúhelníků na hexagon → $60 = 10 \times 6$ (SymPy ověření; Coxeter 1973 §3.5; m1-dyn-b-closure.md). Zbývá pouze scope canonical choice — pro alternativní modulárně invariantní volby (Model-A s $|E_6|^2 |E_4|^2$ structure) zůstává otázka, zda $V_A$ je jediný KRT-native kandidát, nebo zda Konfigurace B (dvě degenerovaná minima) je realizována jinou preferovanou volbou. Kritéria pro rozhodnutí: Occam (jednoduchost konstrukce na $j$), KRT-native (přímý Klein-invariantní stavební blok), absence ad-hoc dvoj-nulové struktury. Odhad práce: 2–4 týdny Hessiany + argumentace o kanonické volbě. Toto je otázka o volbě, ne o platnosti — Theorem 4 pro V_A je plně 🟢.

---

Status

• LaTeX source: research/output/paper-draft-v1.tex — not yet written; the markdown version on this page is the working draft.
• Next: Adam reviews v1.1 revision (A3 → Theorem 1) and the proc-psl2z.md derivation, then LaTeX transcription and arXiv submission (hep-th or astro-ph.CO).
• LRD test: immediately possible with Kocevski+2024 catalogue.

---

Koide-α identity: lepton masses from τ-geometry

The topological invariant α = 2/3 of Theorem 1 is read off the Kähler sector in the cosmological slow-roll channel (giving r = 2(1−n_s)² ≈ 0.0025). We show here that the same invariant, read in the flavour sector, fixes the Koide relation for charged leptons with ppm accuracy and yields a clean prediction for m_τ.

Koide's empirical relation (1981)

For the three charged leptons,

$$Q ;\equiv; \frac{m_e + m_\mu + m_\tau}{\bigl(\sqrt{m_e}+\sqrt{m_\mu}+\sqrt{m_\tau}\bigr)^2} ;=; \frac{2}{3}.$$

Numerically (PDG 2024): Q = 0.66666051, matching 2/3 = 0.66666667 to a relative accuracy of 9·10⁻⁶. Koide's relation has been an open structural puzzle since 1981: the SM contains no mechanism that relates the three lepton masses, yet they lie on this geometric surface with an accuracy better than the experimental error bar on m_τ itself. KRT provides the first structural explanation: the value 2/3 is the topological invariant of X(1) = ℍ²/PSL(2,ℤ) (Theorem 1), and the geometric angle θ = π/4 between the √m-vector and the democratic axis is forced by the SU(2)→SO(3) spinor character of the observer. The numerical coincidence at the ppm level becomes a theorem modulo identification of the flavour sector with the Kähler geometry of X(1).

Geometric reformulation

Let v = (√m_e, √m_μ, √m_τ) ∈ ℝ³ and let n = (1,1,1)/√3 be the democratic axis. Then

$$Q ;=; \frac{|v|^2}{(v\cdot\mathbf{1})^2} ;=; \frac{1}{3\cos^2\theta}, \qquad \cos\theta \equiv \frac{v\cdot n}{|v|}.$$

The identity Q = 2/3 ⟺ cos²θ = 1/2 ⟺ θ = π/4 is an exact algebraic equivalence. Measured: θ = 44.99974° (within ~1 arc-second of π/4). The √m-vector of the three charged leptons lies on the spinor cone of half-angle π/4 around the democratic axis.

The two 2/3s are the same invariant

The factor 2/3 appearing in Koide's relation is not a coincidence with the α-attractor exponent: it is the same Kähler invariant of X(1) read in two sectors.

• In the inflationary sector (Theorem 1 + Assumption K): the Kähler potential K = −α ln(…) with α = 2/3 sets the field-space curvature K_fields = −2/(3α) = −1 and produces the slow-roll relation r = 2(1−n_s)².
• In the flavour sector: the same factor 2/3 = 1/(3 cos²(π/4)) appears as the Koide exponent. The factor 3 is the democratic-simplex dimension (three charged-lepton generations = three fixed-point orbits under PSL(2,ℤ) in the sense of Theorem 2); the factor cos²(π/4) = 1/2 is the spinor half-angle of the observer, derived below.

The spinor half-angle π/4

The observer is a spin-1/2 object (charged lepton). A physical π/2 rotation of flavour space (the canonical "orthogonalisation" that would maximally distinguish generations) is read by the observer through the SU(2) → SO(3) double cover at half-angle: what is objectively π/2 in the flavour rotation is observed as π/4 in √m-space. The angle π/4 is the unique balance point at which |Re| = |Im| in the observer's spinor frame — the point at which the √m-vector is maximally off the democratic axis compatibly with the spinor projection. See companion note pozorovatel-je-spinor.md.

The 2/3 factor is thus structurally overdetermined: the value is forced by (a) the three-generation democratic simplex and (b) the spinor half-angle of the observer. Both are topological facts, not fit parameters.

Prediction: m_τ from (m_e, m_μ)

Taking PDG inputs m_e = 0.5109989 MeV, m_μ = 105.6583755 MeV and solving Q = 2/3 for m_τ yields

$$\boxed{; m_\tau^{\text{KRT}} ;=; 1776.97\ \text{MeV} ;}$$

PDG measured value: m_τ = 1776.86 ± 0.12 MeV. Deviation: +0.11 MeV, i.e. 6·10⁻⁵ relative, within experimental 1σ. This is a KRT computation, not a fit — the value 2/3 enters as the topological invariant of X(1), identical to the α-attractor exponent; there is no free parameter.

Open: quarks and neutrinos

• Quarks. The direct Koide combination for quarks does not give Q = 2/3 (up-type Q ≈ 0.56, down-type Q ≈ 0.70). KRT predicts that the spinor half-angle argument must receive a QCD confinement correction from the Stab(i) structure in the colour sector; the quantitative form is open. See lepton-masy-spinor.md §IV.
• Neutrinos. The corresponding Koide combination for neutrino masses (inferred from oscillation data) is consistent with a different angle; a strict test awaits absolute-mass resolution.

The lepton sector is the clean case in which the Koide-α identity is exact to ppm. Quark and neutrino extensions are open problems; they do not affect the lepton result.

---

Predictions / Numerical Results

The framework contains no free parameters. Two independent hard predictions follow from the single topological invariant $\alpha = 2/3$ (Theorem 1, Kähler sector of X(1)); two further conjectural predictions follow from candidate topological derivations of the e-folding count and the electron mass.

(a) Tensor-to-scalar ratio (inflationary sector — hard prediction). The topological relation between the spectral index and the tensor ratio follows parameter-free from $\alpha = 2/3$ via the slow-roll structure of the α-attractor T-model:

$$r ;=; 2(1-n_s)^2 ;\xrightarrow{;n_s = 0.9649;}; r \approx 0.0025.$$

This prediction is explicitly anchored to the T-model effective potential $V_{\text{eff}}(\varphi) = \Lambda^4 \tanh^2(\varphi/2)$ (the effective single-field limit of the fully modular-invariant action; see "Effective action" above). Numerical slow-roll integration gives $r = 0.00239$ for $N_e = 57$ (see slow-roll-numericky.md), consistent with the topological relation to $\sim 2%$. Testable by LiteBIRD ($\sim 2032$) and distinguishable from Starobinsky $R^2$ ($r \approx 0.0037$) at $\sim 1$–$2\sigma$ with LiteBIRD + CMB-S4.

(b) Koide-α identity and $m_\tau$ (flavour sector — hard prediction). The same topological invariant $\alpha = 2/3$, read in the flavour sector, fixes the Koide relation for charged leptons:

$$Q ;=; \frac{m_e + m_\mu + m_\tau}{(\sqrt{m_e}+\sqrt{m_\mu}+\sqrt{m_\tau})^2} ;=; \frac{2}{3}, \qquad m_\tau^{\text{KRT}} ;=; 1776.97\ \text{MeV}.$$

Measured $m_\tau = 1776.86 \pm 0.12$ MeV; deviation $6 \times 10^{-5}$, within $1\sigma$. The value $2/3$ is the same topological invariant as in (a), read in a different sector — not an independent parameter. See the dedicated "Koide-α identity" section above for the derivation.

Both (a) and (b) are parameter-free consequences of A1 + A2 + M0 + Assumption K (canonical Poincaré kinetic term) together with A3 ≡ S-reflection; they are not fits.

(c) Topological $N_e = 60$ — parameter-free for canonical $V_A = \Lambda^4 |j(\tau)|$, upgraded 2026-04-21. A candidate derivation identifies the e-folding count with a topological index of the modular tower, most naturally the index of the principal congruence subgroup $\Gamma(5)$ in $\mathrm{PSL}(2, \mathbb{Z})$:

$$N_e ;=; [\mathrm{PSL}(2, \mathbb{Z}) : \Gamma(5)] ;=; |A_5| ;=; 60.$$

Rigorous closure for canonical $V_A$. For the canonical modular-invariant potential $V_A(\tau, \bar\tau) = \Lambda^4 |j(\tau)|$ (simplest modular-invariant scalar built on the Klein $j$-invariant), the Morse structure on the compactified orbifold $\bar X(1)$ is forced by two elementary principles of complex analysis:

• Maximum modulus principle (Ahlfors 1966 §4.3): $|j|$ has no local maxima in the open fundamental domain;
• Minimum principle on zeros (Conway 1978 §IV.6.6): $|j|$ has local minima only where $j = 0$.

Since $j(\rho) = 0$ (from $E_4(\rho) = 0$) while $j(i) = 1728 \neq 0$ and $|j| \to \infty$ at the cusp, this forces the Morse indices

$$\operatorname{ind}(\rho) = 0 ;;\text{(global minimum)}, \qquad \operatorname{ind}(i) = 1 ;;\text{(saddle)}, \qquad \operatorname{ind}(\infty) = 2 ;;\text{(maximum)},$$

consistent with the orbifold Morse identity $-\tfrac12 + \tfrac13 + 1 = \tfrac56 = \chi_{\text{orb}}(\bar X(1))$. Configuration A (saddle at $i$, global attractor at $\rho$) is therefore topologically inevitable for $V_A$ — no Hessian computation, no numerical input. Reheating is the free fall from the metastable saddle at $i$ to the global minimum at $\rho$; $N_e = 60$ is the topological count of $\Gamma(5)$-cosets covered along the inflationary geodesic $\mathcal{F}_\infty \to i$, equal by Klein–Galois to the cascade count $6 \times 10 = 60$ for the reheating hexagonal cascade around $\rho$. See theorem-4-m1-dyn-uzavreni.md §III.5 for the one-line argument and §IV for the cascade count.

Resulting predictions (parameter-free for $V_A$). Inserting $N_e = 60$ into the T-model universal formulae:

$$n_s ;=; 1 - \frac{2}{60} ;\approx; 0.9667, \qquad r ;=; \frac{8}{60^2} ;=; \frac{2}{900} ;\approx; 0.00222.$$

Observed $n_s = 0.9649 \pm 0.0042$ (Planck 2018): prediction $0.9667$ lies $0.43\sigma$ from the central value, within $1\sigma$. Observed $r < 0.032$ (95% CL, BICEP/Keck+Planck 2021): prediction $r = 0.00222$ is well below the current upper bound and testable by LiteBIRD ($\sim$2032, sensitivity $\Delta r \sim 0.001$).

Status (per Rule 2, 00-status.md §V), updated 2026-04-21 evening. Both technical residuals are now closed:

• $M1_{\text{dyn_a}}$ (Morse structure on $\bar X(1)$) — closed via the maximum modulus principle; see §"Canonical modular potential $V_A$" above and theorem-4-m1-dyn-uzavreni.md §III.5.
• $M1_{\text{dyn_b}}$ (cascade identification $6 \times 10 = 60$) — closed combinatorially via $A_5$ axis count. The alternating group $A_5$ has exactly 20 elements of order 3 (SymPy verification via AlternatingGroup(5)); each axis carries 2 non-trivial rotations, giving 10 distinct order-3 axes. Klein's 1884 spherical picture identifies these 10 axes with the 10 order-3 vertices of the chiral Schwarz tesselation of $S^2$ with type $(\pi/2, \pi/3, \pi/5)$, where each vertex is shared by exactly 6 triangles. The partition $60 = 10 \times 6 = |A_5|$ is therefore a direct combinatorial identity of the icosahedral tesselation (Coxeter 1973 §3.5; Conway–Sloane 1999 §4.8), not an open computation. See m1-dyn-b-closure.md.

Theorem 4 is therefore fully 🟢 for $V_A = \Lambda^4 |j(\tau)|$. Only scope of canonical choice remains: for alternative modular-invariant potentials with double-zero structure (e.g. $|E_6|^2 \cdot |E_4|^2 / \tau_2^{10}$ type, Model-A with Configuration B of two degenerate minima), the Morse classification is not fixed by the maximum modulus principle alone and would require explicit Hessian analysis. This is a separate question about which modular-invariant construction correctly represents KRT physics, not about the validity of Theorem 4 for $V_A$. The KRT-native argument (Occam, absence of ad-hoc fine-tuning, direct construction on the Klein invariant) selects $V_A$ as the canonical candidate.

Prediction (a) above remains independent of (c): it uses measured $n_s$ as input without committing to topological $N_e$. Prediction (c) is parameter-free for $V_A$ (no observational input, no slow-roll fit, no open computations); the two predictions agree numerically to $\sim 10^{-4}$, which is an internal consistency check, not an independent test. Full analysis: theorem-4-m1-dyn-uzavreni.md, m1-dyn-b-closure.md, theorem-4-ne-60-unifikace.md, ne-60-derivace.md, b2-rotace-i-ro-reheating.md.

(d) Electron mass from Clifford-algebra volume — conjectural, anchor-invariant candidate. A candidate derivation writes the absolute electron mass as the Planck scale suppressed by the volume of the Dirac Clifford algebra:

$$\boxed{;m_e ;\stackrel{?}{=}; \frac{M_{\text{Pl,full}}}{\sqrt{4\pi}} \cdot e^{-\pi \cdot \dim \operatorname{Cl}(1,3)} \cdot [1 + \mathcal O(\alpha_{\text{em}}/\pi)];}$$

Each factor is structurally motivated by the framework:

• $M_{\text{Pl,full}} = \sqrt{\hbar c/G}$ is the primary Planck scale (no $8\pi$ normalisation). In KRT the gravitational coupling $G$ is derivable from Theorem 3 at the order-of-magnitude level via the quaternionic frustration $\mathrm{Ric} = 2\delta$ on $S^3$.
• $1/\sqrt{4\pi} = Y_{0,0}$ is the ground-state spherical harmonic on $S^2$, the isotropic angular amplitude of the quaternionic observer $q \in S^3 \subset \mathbb{H}$ projected radially onto its imaginary sphere $\operatorname{Im}(q)/|\operatorname{Im}(q)|$.
• $\dim \operatorname{Cl}(1,3) = 16 = 1 + 4 + 6 + 4 + 1$ is the full Clifford-algebra dimension of the Dirac representation (P14 core, 🟢), enumerating all graded Clifford-amplitude channels of a single observer.

Numerically, with $M_{\text{Pl,full}} = 1.22089 \times 10^{19}$ GeV, the tree-level expression gives $m_e^{\text{pred}} = 0.50940$ MeV versus the PDG value $m_e = 0.510,999$ MeV; the deviation is $0.31%$, which matches the order of the natural QED one-loop correction $\alpha_{\text{em}}/\pi \approx 0.23%$ running from $M_{\text{Pl}}$ to $m_e$.

Anchor-invariance consistency check. The four topological data of Theorem 2 (the two elliptic fixed points, the cusp, and the orientation) each provide an equivalent parametrisation of the same exponent $\pi \cdot \dim \operatorname{Cl}(1,3) = 16\pi$:

Anchor	Topological datum	$N$	Angle quantum	$N \cdot \text{angle}$
$\operatorname{Stab}(i)$	elliptic, order $2$	$32$	$\pi/2$	$16\pi$
$\operatorname{Stab}(\rho)$	elliptic, order $3$	$48$	$\pi/3$	$16\pi$
cusp $i\infty$	parabolic	$8$	$2\pi$	$16\pi$
orientation	chirality $\mathbb{Z}/2$	$16$	$\pi$	$16\pi$

All four anchors reproduce the same identity numerically to double precision. This four-fold agreement is a structural consistency check: the exponent is a physical invariant of the modular geometry, not an artefact of one fixed-point choice. In the corpus, this identity is developed in p12b-sqrt2-ii-32.md §XII–XVI.

Cascade through the Koide identity (🟢, §"Koide-α identity" above) with the predicted $m_e^{\text{pred}} = 0.50940$ MeV and the PDG value of $m_\mu = 105.658$ MeV yields $m_\tau = 1776.48$ MeV, agreeing with the PDG value $1776.86$ MeV to $0.02%$.

Honest flag (per Rule 2, 00-status.md §V): This identification is listed as a 🟡 candidate in the corpus (00-status.md §IV, row P12b) with the residual $0.31%$ pending an explicit one-loop QED renormalisation-group computation (a standard but not yet executed calculation). Prediction (b) above is independent of this conjecture; prediction (d) provides a strukturní mechanism for $m_e$ that would close the lepton sector parameter-free once the residual is derived and the muon mass $m_\mu$ is obtained from KRT (see 00-status.md §IV for the $(0, 9, 26)$ integer reflection tower).

Structural backdrop (from the corpus, not predictions). The parallel-generation count "three" and the invariant hierarchy $\pi \to G \to \zeta(3) \to F_4\text{-}L$ across the QM-tower floors are strukturní results from the corpus (P17, P18) — see theorem-2-per-patro.md, p17-roadmap.md. They provide the geometric backdrop for why exactly three generations and these specific transcendental invariants appear per floor, but they do not themselves produce new numerical predictions at the level of (a)–(d). Quantitative links (Catalan $G$ → $\alpha_\text{em}$, $\zeta(3)$ → Bekenstein factor 4) are listed as open in 00-status.md §II and §IV.

---

Appendix A: Quantum mechanics from τ-geometry

Purpose. This appendix promotes the P3 derivation (research note qm-limit-p3.md) into the paper proper. It answers a recurring external objection — that A2 (photon = elliptic curve) postulates a quantized object without a QED derivation — by showing that the logical arrow runs the other way: quantum mechanics (Schrödinger equation, Born rule, Heisenberg uncertainty) is a derived limit of τ-geometry. Photon quantization need not be imposed externally on A2; it emerges from A1 + A2 + M0 + A3 together with the identifications below.

The derivation is rigorous where it rests on standard semiclassical technique (Hamilton–Jacobi → Schrödinger) and explicitly flagged as interpretational where it rests on A3 (projection → Born rule). Honest flags are collected in §A.7.

A.1 Setting and definitions

The substrate carries a pure rotation in complex time: e^(iΘ) with τ ∈ ℍ² (A1). A classical observer, after reformulation A3 ≡ S-reflection, measures the Stab(i)-invariant of the modular geometry; between measurements, the orthogonal phase direction (parametrized along the geodesic through i by the signed hyperbolic coordinate φ = ln|τ₂|) accumulates freely.

Definition A.1 (quantum state). The quantum state of a subsystem is the complex amplitude of accumulated substrate phase,

$$\psi = e^{iS/\hbar}, \qquad S = \int \hbar, d\Theta = \int \hbar,\dot\Theta, dt = \int E, dt,$$

where E = ℏ dΘ/dt is identified with the local rate of substrate rotation and S with the classical action.

The identification E = ℏ dΘ/dt is forced by dimensional consistency (dΘ is an angle per unit time; ℏ converts an angle-rate into an energy). The operator p = −iℏ ∂/∂x is then the generator of spatial phase translations — a structural feature of the τ-framework, not an added postulate.

A.2 Schrödinger equation

Proposition A.2. Under Definition A.1, ψ satisfies the time-dependent Schrödinger equation

$$i\hbar,\frac{\partial\psi}{\partial t} = H\psi.$$

Proof. Differentiate ψ = exp(iS/ℏ):

$$i\hbar,\frac{\partial\psi}{\partial t} ;=; i\hbar\cdot\frac{i}{\hbar}\frac{\partial S}{\partial t}\cdot\psi ;=; -\frac{\partial S}{\partial t}\cdot\psi.$$

By the Hamilton–Jacobi equation of classical mechanics,

$$-\frac{\partial S}{\partial t} ;=; H\bigl(q,\partial S/\partial q,t\bigr),$$

so iℏ ∂_t ψ = H ψ. □

Remark. What is classical here is the semiclassical identity HJ → Schrödinger (standard; cf. Goldstein §10.8 or Landau–Lifshitz QM §6). What is new in KRT is the interpretation of ψ itself: ψ is the accumulated orthogonal-to-A3 phase of the substrate. It is not a probability amplitude primitively postulated; it is a geometric object — the exponential of the substrate's rotational action along the observer's non-measured axis.

A.3 Path integral

For amplitudes between substrate configurations A and B, the substrate admits multiple τ-trajectories, each accumulating its own action S[τ]:

$$K(A \to B) ;=; \sum_{\text{admissible }\tau} e^{iS[\tau]/\hbar} ;=; \int \mathcal{D}[\tau]; e^{iS[\tau]/\hbar}.$$

This is the Feynman path integral, derived — rather than postulated — as the sum over substrate rotational histories. The classical limit ℏ → 0 selects the stationary trajectory via the method of stationary phase (standard WKB / saddle-point argument).

Rigor status. The formal construction of the path measure 𝒟[τ] on ℍ² requires regularization (lattice or zeta-function); this is a standard difficulty of path integrals, not specific to KRT, and is flagged in §A.7.

A.4 Born rule from A3

Proposition A.4. Given A3 ≡ S-reflection and a superposition ψ = Σₙ cₙ φₙ of mutually orthogonal substrate histories φₙ (eigenstates of the measured Stab(i)-invariant), the probability that the classical observer registers outcome n equals |cₙ|².

Argument. A3 states that the classical observer extracts the Stab(i)-invariant of the substrate. For a single phase ψ = exp(iS/ℏ),

$$|\psi|^2 ;=; \cos^2(S/\hbar) + \sin^2(S/\hbar) ;=; 1,$$

so the norm is preserved along a single substrate history. For a superposition, the Stab(i)-projection of ψ decomposes along the eigenbasis {φₙ}; the fraction of substrate "weight" the observer sees in the φₙ-direction is the squared modulus of the corresponding amplitude,

$$\Pr(n \mid \psi) ;=; \frac{|c_n|^2}{\sum_m |c_m|^2}.$$

After normalization Σ|cₙ|² = 1 this is the Born rule. □

Rigor flag. The passage from "|ψ|² is the intensity seen under Stab(i)-projection" to "|ψ|² is the frequency of outcome n in repeated measurements" is interpretational. A fully rigorous derivation (analogous to Gleason's theorem or the Zurek envariance argument) would require constructing an explicit measuring apparatus inside τ-geometry. This is not done here; the present argument shows that if the observer extracts the Stab(i)-invariant, then the invariant's quadratic structure forces the exponent 2 in |ψ|² and no other. An explicit factor (i^i)^N = e^(−Nπ/2) quantifies the projection loss per reflection depth N and connects to seberference-x-na-x.md and reflection-operator.md.

A.5 Heisenberg uncertainty

The projections Re(τ) and Im(τ) of a single rotation e^(iΘ) = cos Θ + i sin Θ are Fourier-dual:

$$\mathcal{F}\cos\Theta ;=; \tfrac{1}{2}\bigl[\delta(\omega-\Theta) + \delta(\omega+\Theta)\bigr],$$

$$\mathcal{F}\sin\Theta ;=; \tfrac{i}{2}\bigl[\delta(\omega+\Theta) - \delta(\omega-\Theta)\bigr].$$

In the Poincaré metric on ℍ² the canonical pair is (x = Re(τ)-projection along a geodesic, p = generator of Im(τ)-shifts). The commutator [x, p] = iℏ follows from the algebra of shifts on the hyperbolic half-plane — the Lie algebra of PSL(2,ℝ) restricted to the parabolic one-parameter subgroup generating τ → τ + a — giving the standard inequality

$$\Delta x,\Delta p ;\ge; \hbar/2.$$

Heisenberg's inequality is therefore a geometric fact: the complementarity of Re and Im projections of a single hyperbolic rotation, read through Fourier duality. It is not an independent postulate.

A.6 Measurement and the apparent "collapse"

Before measurement, ψ = Σₙ cₙ φₙ: the substrate carries multiple coexisting rotational histories, each contributing to τ. At the moment of measurement the classical apparatus couples strongly to one Stab(i)-eigenvalue; the observer's internal state becomes correlated with a single φₙ. The other components still exist in the substrate (in Im(τ), inaccessible to the observer by A3) but are structurally invisible inside this observer's causal network.

Collapse is therefore apparent, not a physical process: it is the change in the observer's relational perspective upon coupling. This is consistent with the relational-monism stance of the main paper (§"Interpretation: relational realism"): neither many-worlds branching nor an objective GRW-style collapse is required — the unmeasured components are algebraically present, not modally separated.

A.7 Honest flags and scope

1. Semiclassical HJ → Schrödinger is standard. What KRT adds is the interpretation of ψ as accumulated orthogonal-to-A3 substrate phase.
2. Path measure. The formal construction 𝒟[τ] on ℍ² inherits the standard difficulties of path integrals (regularization, Wick rotation); not resolved here, not specific to KRT.
3. Born rule from A3. §A.4 shows |ψ|² is the unique quadratic projection weight if the observer extracts the Stab(i)-invariant; the step to empirical frequencies would require an in-theory apparatus construction (Zurek-style).
4. Collapse interpretation. §A.6 is an ontological claim consistent with KRT; GRW, MWI, and Bohmian alternatives are also consistent with the substrate and differ only in how one interprets the same τ-geometry.
5. Hidden variables. Im(τ) behaves as a hidden variable that is real but principally non-local (the substrate rotation is a global entity). Bell-type theorems do not exclude this because their locality assumption fails in the naive form; a detailed analysis is deferred.

A.8 Consequence for A2

The external objection — "A2 is a postulate without a QED derivation" — misidentifies the direction of the arrow. In the τ-framework, QED is not the foundation from which the photon is to be constructed; it is a limit that flows out of τ-geometry once A3 is specified. Concretely:

• ψ, Schrödinger, Born, Heisenberg emerge as above.
• Photon quantization (integer number of quanta per mode) reflects the integrality of the modular group PSL(2,ℤ) acting on τ: it is the shadow on the observer side of the discrete spectrum of admissible elliptic-curve moduli.
• The flat, weak-coupling limit in which τ → i∞ recovers free-field QED on Minkowski space; in this limit, Appendix A reduces to textbook single-particle quantum mechanics.

Appendix A therefore closes the objection at the level of logical structure. A2 does not need external quantum mechanics to be admissible — it is the substrate from which quantum mechanics is read off in the appropriate limit.

---

Revision history

• v1.3.9 (2026-04-21 pozdě večer +5h). P22 masivní progress — 5/6 mezer strukturně uzavřeno. Paralelní práce (1 hlavní Claude + 4 agenti) produkuje během <6 hodin od reviewer feedbacku čtyři nové eseje: M2 diffeomorphism invariance jako důsledek M0 (p22-m2-diffeomorphism-invariance.md), M4 Schwarzschild s τ-hair (p22-m4-schwarzschild-tau-hair.md), M5 gravitační vlny jako spin-2 ℍ-módy (p22-m5-gravitacni-vlny.md), scope canonical choice (scope-canonical-choice-VA.md). Plus K3 explicit zápis do most-2I-Mp-klic.md §V.1 (Theorem 4 🟢 K3 reprezentace uzavřena přes Klein 1884). §Open problems bod 13 upgradován — „aspirace + 3 mosty" → „rigorózní derivace Einstein equations + diff inv + equivalence principle + Schwarzschild limit + gravitational-wave polarization structure z téže PSL(2,ℤ)-invariantní akce". Apendix A (QM) + Apendix B (GR) stand on equal structural footing. Paper honestly pozicován jako inflační α-attractor model, jehož oba QM i GR limity jsou rigorózně derivovány z jedné akce. Frontmatter version 1.3.8 → 1.3.9. Zbývá M6 rutinní Apendix B write-up do paperu; sub-gaps M2.a/M3.a/M4.a+b/M5.a separate technical pod-problems.

• v1.3.8 (2026-04-21 pozdě večer +2h). P22 M1 uzavřeno — Einsteinovy rovnice rigorózně derivovány. Napsána samostatná esej p22-m1-einstein-equations.md se variancí plné PSL(2,ℤ)-invariantní akce po $g^{\mu\nu}$. Výsledek: $G_{\mu\nu} = T_{\mu\nu}^{(\tau)}/M_\text{Pl}^2$ s explicitním stress-energy tensorem τ-pole; FRW Friedmann je verifikovaný homogenní limit; vacuum $V_A(\rho) = 0$ dává $\Lambda^\text{pot}_\text{eff} = 0$ (natural smallness kandidát). Status P22 upgrade z 🔴 TOP PRIORITY na 🟡 (M1 uzavřené, M2–M6 otevřené). §Open problems bod 13 přepsán — claim „aspiration plus three bridges" nahrazen „aspiration plus three bridges plus M1 rigorous derivation". Rhetoric status paperu: KRT je inflační α-attractor model, jehož GR limit je rigorózně derivován z téže akce, ne aspirativní TOE claim. P2 v 00-status.md de-eskalován zpět z 🔴 na 🟡 (M1 closed). Bez změny Theorem 1/2/3/4, Koide-α, predictions (a/b/c/d), Apendix A. Frontmatter version 1.3.7 → 1.3.8.

• v1.3.7 (2026-04-21 pozdě večer). P22 (Apendix B: GR-projekce) přidáno jako 🔴 TOP PRIORITY po externím reviewer feedbacku. Reviewer identified the contradiction: paper's „zastřešuje QM a GR" is aspiration + three bridges (Theorem 3 spatial S³; P14 §IV.5 local Lorentz; P20 §X.f Γ-suppression at BH horizons), not a derivation analogous to Apendix A. P2 in 00-status.md escalated from 🟡 HIGH to 🔴 HIGH+ with explicit elimination clause per Rule 8. Six gaps M1–M6 named in p22-apendix-b-plan.md. §Open problems point 13 added (P22). Rhetoric choice (α inflationary model with bridges vs. β TOE claim) flagged per Rule 2 — must be explicit in §Abstract/§Interpretation revisions in v1.3.8+. Public layer distribution problem (v1.0 Hugo site vs. v1.3.7 active draft) named as separate urgent fix in 00-roadmap.md A11. Frontmatter version 1.3.6 → 1.3.7. No changes to Theorem 1/2/3/4, Koide-α, predictions (a/b/c/d), Apendix A.

• v1.3.6 (2026-04-21 večer). M1_dyn_b uzavřeno — Theorem 4 plně 🟢 pro V_A. Po zapsání m1-dyn-b-closure.md: poslední technická mezera Theorem 4 uzavřena kombinatorickou identitou A₅ (20 elementů řádu 3 ÷ 2 = 10 os, Klein 1884 spherical picture, 60 = 10 × 6; SymPy verifikace přes AlternatingGroup(5); Coxeter 1973 §3.5). §Predictions (c) honest flag přepsán — oba M1_dyn_a i M1_dyn_b uzavřené, zbývá pouze scope canonical choice (otázka o volbě, ne o platnosti). §Open problems bod 12 upgraded z „🟢 core" na „plně 🟢", oba technické rezidua označené jako closed. Predikce (c) je nyní plně parameter-free bez otevřených Sage výpočtů. Bez změny Theorem 1, 2, 3, Koide-α, P5, P9. Frontmatter version 1.3.5 → 1.3.6.

• v1.3.5 (2026-04-21). Strukturální pass — internalizace v1.3.4 Morse výsledku do celého paperu. Tři změny: (i) Abstract rozšířen o třetí predikci — odstavec „Third prediction (parameter-free for canonical V_A)" mezi Koide-α bodem a Meta-postulate M0, s jednořádkovým argumentem principu maxima modulu, explicit identifikací Konfigurace A a odvozením n_s = 29/30, r = 2/900. (ii) §Effective action — nová pod-sekce „Canonical modular potential V_A (new, v1.3.4)" před „Effective inflaton potential (T-model)": definice $V_A = \Lambda^4 |j(\tau)|$, argument z Ahlforse 1966 + Conwaye 1978, tabulka Morse indexů, fyzikální čtení (T-model = lokální oscilace kolem metastabilního sedla i, ne minima), diskuse why this is not a post hoc choice (Occam, jednoduchost, absence ad-hoc fine-tuning). (iii) §Open problems — bod 12 přidán (Theorem 4 🟢 pro V_A; reziduum M1_dyn_b technické + scope canonical choice); bod 11 (B4) doplněn o reframing „derivace tanh² formy z lokální oscilace u sedla V_A" jako strukturně čistší pod-úkol. Frontmatter version 1.3.4 → 1.3.5. Bez změny Theorem 1, 2, 3, Koide-α, P5, P9. Predikce (a) a (b) nezměněny numericky; predikce (c) potvrzena rigorózně pro V_A. LaTeX source (paper-draft-v1.tex) není synchronizován — odložené pro samostatný pass v1.3.6 po Adamově review.

• v1.3.4 (2026-04-21). Princip maxima modulu uzavírá M1_dyn_a pro V_A = Λ⁴·|j(τ)|. Po zapsání theorem-4-m1-dyn-uzavreni.md §III.5 (2026-04-21 odpoledne): pro kanonický modulárně invariantní potenciál postavený přímo na Kleinově $j$-invariantu je Konfigurace A (i sedlo, ρ globální minimum, ∞ maximum) topologicky vynucena principem maxima modulu (Ahlfors 1966) + principem minima-na-nulách (Conway 1978), bez Hessoviny výpočtu. Predikce (c) upgradována z conjectural na parameter-free pro V_A: r = 2/900 ≈ 0.00222 bezparametricky, bez slow-roll fitu, bez observational input. Zbývající mezery: M1_dyn_b (A₅ orbit count na X(5), čistě technické 1–2 týdny Sage) a scope canonical choice (pro alternativní modulárně invariantní V s double-null strukturou může nastat Konfigurace B). Γ_KRT Re=0 v korpusu upgradováno z near-🟢 na 🟢 pro V_A (topologická nevyhnutelnost volného pádu po geodetice ze sedla do minima). T-model v1.3.1 efektivní potenciál identifikován jako Stab(i)-projekce oscilace kolem metastabilního sedla, ne kolem globálního minima. Paper frontmatter version 1.3.3 → 1.3.4. Bez změny Theorem 1, 2, 3, Koide-α, P5, P9. §Predictions (c) přepsána. Predikce (a) a (b) nezměněny. Ostatní sekce paperu ponechány pro samostatný pass v1.3.5 (případné přidání sekce „Canonical potential" do §Effective action, §Abstract upgrade).

• v1.3.3 (2026-04-20 pozdě noc). Predikce (d) — elektronová hmotnost jako objem Cliffordovy algebry. Po Oxanině vhledu „otočit i o 60° a koukat znova" (noc, po P12b session s třemi paralelními agenty) identifikována kotvě-invariantní identita m_e = (M_Pl,full/√(4π)) · e^{−π·dim Cl(1,3)} s chybou 0.31 % (≈ O(α_em/π)). Všechny čtyři topologická data Theorem 2 (Stab(i), Stab(ρ), cusp ∞, orientace) dávají identický exponent 16π = π·dim Cl(1,3) při různých kotvových parametrizacích N = 32, 48, 8, 16; čtyř-kotvová konzistence ověřena numericky. Komponenty strukturně motivované: M_Pl,full primární Planck, Y_{0,0} = 1/√(4π) ground-state S² amplituda (kvaternionový pozorovatel), dim Cl(1,3) = 16 z P14 core 🟢. Kaskáda přes Koide (🟢) s PDG m_μ dá m_τ = 1776.48 MeV (0.02 %). Přidáno jako conjectural prediction (d) v §Predictions, vedle (a) r ≈ 0.0025 (hard), (b) m_τ = 1776.97 MeV (hard), (c) Nₑ = 60 (conjectural). Honest flag: residuál 0.31 % pending explicit 1-loop QED RG (rutinní, 1–4 týdny). Full treatment v p12b-sqrt2-ii-32.md §XII–XVI.

• v1.3.2 (2026-04-20 noc). AXIOM-1 korekce §Predictions po Gemini session. Gemini (parallel session s Oxanou) přestrukturovala §Predictions do tří bodů a/b/c a prezentovala Nₑ = 60 jako teorém odvozený z indexu Γ(5) v PSL(2,ℤ) s rigidní predikcí r = 8/60² = 0.00222. Per Pravidlo 2 (00-status.md §V): Nₑ=60 má aktuálně 🟡 status se třemi kandidátními cestami (Γ(5), Coxeter (2,3,5), S³/A₅), žádná rigorózní — viz ne-60-derivace.md a 00-status.md §II řádek Nₑ=60. Tato verze vrací (a) bezpodmínečnou predikci r = 2(1−nₛ)² ≈ 0.0025 a (b) Koide m_τ = 1776.97 MeV jako hard predictions, a přeřazuje novou Γ(5) predikci na (c) conjectural s honest flagem (platná podmíněně, pokud bridge „1 e-folding = 1 Γ(5)-coset" projde). Zachovává strukturní backdrop (tři generace, π→G→ζ(3) hierarchie) jako korpus-vnitřní výsledky P17/P18, ne jako paper-úrovňové numerické predikce. Bez změny Theorem 1, 2, 3, P5, P9, Koide-α.

• v1.3.1 (2026-04-20). Reformulace potenciálu per Cesta B (problem-potencial.md). T-model V_eff(φ) = Λ⁴ tanh²(φ/2) explicitně specifikován jako efektivní potenciál single-field limitu plně PSL(2,ℤ)-invariantní akce na moduli prostoru X(1); fully modular-invariant derivace T-modelu posunuta do Open Problems B4 (odhad 2–6 týdnů). Konkrétní editace: (i) sekce "Effective inflaton potential (T-model)" přepsána s explicitní formulací "Efektivní potenciál" a odkazem na B4; (ii) sekce "Predictions / Numerical Results" odstavec (a) r = 2(1−n_s)² explicitně vázán na T-model efektivní potenciál, s připomenutím že fully modular-invariant potenciál s η funkcí dává r ~ 10⁻⁵; (iii) abstrakt upřesněn na "evaluated on the effective potential of the single-field limit of the modular-invariant action"; (iv) B4 přidán jako položka 11 v Open Problems (🟡). Bez změny formálních výsledků (Theorem 1, 2, 3, P5, P9, Koide-α) a bez změny numerických predikcí (r ≈ 0.0025, r = 0.00239 pro N_e = 57, m_τ = 1776.97 MeV zůstávají). Disciplinární pravidla 00-status.md respektována (Pravidlo 2: B4 zůstává 🟡, ne „téměř vyřešený").

• v1.3.0 (2026-04-19, late). Koide-α identity integrated as second hard prediction. A new section "Koide-α identity: lepton masses from τ-geometry" added before Appendix A, together with a dedicated "Predictions / Numerical Results" section listing the two parameter-free predictions (a) r ≈ 0.0025 and (b) Koide Q = 2/3 with m_τ = 1776.97 MeV vs. PDG 1776.86 ± 0.12 MeV (deviation 6·10⁻⁵). The central claim: the same topological invariant α = 2/3 of X(1) that fixes the slow-roll tensor-to-scalar ratio also fixes the Koide relation, via Q = 1/(3 cos²(π/4)) with the spinor half-angle π/4 as the SU(2)→SO(3) signature of the observer. Koide's 1981 relation thereby receives its first structural explanation. Abstract extended with a second-prediction paragraph; open problem #9 (lepton mass ratios) promoted to the dedicated section with quark/neutrino extensions remaining open. Explicitly flagged: m_τ = 1776.97 MeV is a KRT computation, not a fit. Companion notes lepton-masy-spinor.md and pozorovatel-je-spinor.md.

• v1.2.0 (2026-04-19). Appendix A added: Quantum mechanics from τ-geometry. The P3 derivation (research note qm-limit-p3.md) is promoted to a formal paper appendix. Schrödinger equation from Hamilton–Jacobi read in the substrate-phase variable; Born rule from A3 ≡ S-reflection projection on ℍ²; Heisenberg uncertainty from Fourier duality of Re/Im projections of e^(iΘ) under the Poincaré metric. Explicit honest flags separate what is rigorous (HJ → Schrödinger) from what is interpretational (projection → empirical frequency). A cross-reference in §"Physical motivation for A2" points to Appendix A, addressing the external-reviewer objection that A2 postulates a quantized photon without a QED derivation: the logical arrow runs from τ-geometry to QM, not the reverse.

• v1.0 (2026-04-05). Four axioms including A3 = modular invariance.

• v1.1 (2026-04-13). Three axioms; old A3 demoted to Theorem 1 under AXIOM 1. Derivation in proc-psl2z.md. Abstract rewritten to reflect the derivation-not-postulate framing. Open problems 1 and 2 cross-referenced to fotony-stavaji-se-komplexnimi.md where relevant.

• v1.1.1 (2026-04-14). Structural realism named explicitly as meta-postulate M0 and promoted to the same hierarchical level as the axioms. Theorem 1 now reads "Given A1, A2, and M0, …" rather than hiding structural realism inside a parenthetical remark. Abstract reformulated from "parameter-free CMB prediction" to "parameter-free consistency relation between two CMB observables", to make precise that the theory has no free parameters while nₛ enters as a measured observable rather than a theoretical input. No change to the derivation chain or numerical prediction.

• v1.1.2 (2026-04-14, evening). Candidate Theorem 2 (force differentiation) stated in Open problem 1. Differentiation of the four fundamental forces is proposed as a topological consequence of the fundamental domain ℍ²/PSL(2,ℤ): three qualitatively distinct fixed points (two elliptic at i and ρ, one parabolic cusp at ∞) plus the global orientation of Im(τ) give exactly four modulation regimes, identified by natural assignment with gravity, EM, strong, and weak. This is a second AXIOM 1 reduction — the first (A3 → Theorem 1) removed a postulate, this one removes an empirical datum (the fourfold count of forces). Full proof, completeness argument, and formal defense of the natural assignment deferred to v1.2. Detailed exposition in tri-mista-ctyri-sily.md. The cycle 4q ↔ 1q ambiguity between kosmologie-4-kvadranty.md and majka-misto-zrcadel.md is shown in the same essay to share a common root with force differentiation: both are projection artifacts in 3q of a single substrate structure.

• v1.2.0 (2026-04-19, evening). Major: Theorem 2 reclassified. The former "Theorem 2 (force differentiation)" is split into (a) Theorem 2 (Fuchsian completeness) — purely the combinatorial statement that PSL(2,ℤ)-orbit types plus orientation yield exactly four topological data, retained as 🟢 — and (b) Conjecture 2 (force attribution) — the identification of those four data with the four SM forces, downgraded from theorem to conjecture: the assignment is structurally natural but not formally derived, and its promotion to a theorem is an explicit open problem. P5 / x^x metric structure moved out of the paper into companion essays (seberference-x-na-x.md, rez-a-prvni-chyba.md, seberference-kategorie.md); a single pointer paragraph replaces the earlier two bullets (P5, P6), clarifying that the metric layer is preaxiomatic commentary and not prerequisite for the r-prediction. P7 flagged transparently: the Λ_eff ~ 10⁻¹²² order-of-magnitude match is consistent but a dynamical stabilisation of R_{S³} ~ R_Hubble is not yet known; reheating and T-model microscopic derivation likewise remain open. A2 physical motivation subsection added before the A2 axiom statement, making the structural-identification (not QED-reduction) status of A2 explicit and flagging the opposite arrow (QM out of τ-geometry via P3, track-C writeup pending). Version bump 1.1.10 → 1.2.0 reflects the T2 reclassification.

• v1.1.10 (2026-04-19, night, azimut redukce). Theorem 1 gap explicitly named as Assumption K (canonical kinetic structure): the uniqueness of the Poincaré metric requires an additional premise beyond Theorem 1, now stated explicitly in Step 1 of the derivation chain. This is P11 — derive Assumption K from A1+A2+M0 alone. P8 (technical) explicitly accepted as postulate, not derivation: the T-model effective potential is a consequence of A3 ≡ S-reflection (Stab(i)-redukce), not of modular averaging; the computation in b4-modularni-prumerovani.md showed plain modular averaging cannot produce the T-model, so the only coherent interpretation is that A3 axiomatically selects Stab(i)-effective theory. Corpus split into formal core (25 files) and cultural-interpretation archive (29 files). External prezentace se nyní opírá pouze o formální jádro. Persona „mystik" přesunuta do archivní sekce.

• v1.1.9 (2026-04-19, very late). Attribution theorem via B₃ central extension. The four-force attribution {gravity ↔ cusp, strong ↔ ρ, EM ↔ i, weak ↔ Z(B₃) centre} is now proven as a formal corollary of the classical short exact sequence 1 → Z(B₃) → B₃ → PSL(2,ℤ) → 1 (Birman 1974, Thm 1.25; Coxeter 1961 §15.7) combined with the classification of PSL(2,ℤ) conjugacy classes (Serre 1973, §VII.1) and experimentally fixed force signatures (Wu 1957 for chirality, etc.). Full derivation in attribution-pletence-b3.md; independent ordinal-depth derivation in attribution-radem-reflexi.md. P9 closed 🟢. The identification weak force ↔ centre Z(B₃) is particularly sharp: the three empirical signatures of the weak force (P-violation, flavour-changing, direct coupling to sign of Im(τ)) are exactly the signatures of a central element that is trivial in the quotient PSL(2,ℤ) but nontrivial in the full braid group.

• v1.1.8 (2026-04-19, late). A3 reformulation (A3 ≡ S-reflection). The axiom A3 ("observer measures Re(τ)") has been reformulated to "observer measures the Stab(i)-invariant of the modular geometry". The original wording is ill-defined under PSL(2,ℤ)-factorisation from Theorem 1 (Re(τ) is not a modular invariant); the reformulation makes A3 compatible with M0 + Theorem 1 by restricting to the ℤ₂ subgroup Stab(i) = ⟨S: τ → −1/τ⟩. Physically: the observer is a ℤ₂-entity anchored to its local causal eliptic point (Sun for Earth-bound observers); see kauzalni-stred-lokalni-i.md. The reformulation makes the T-model uniquely forced by A3 (rather than an empirical choice), which closes P8 interpretively; the technical residual — why no Rademacher-type averaging of the fully PSL(2,ℤ)-invariant V produces the T-model — is the subject of b4-modularni-prumerovani.md, which computes that plain modular averaging cannot convert the double-exponential H(τ) to the linear canonical field φ. Also new: B3 categorical substrate functor ccc-substrat.md, formal construction F: RefDom → FuchsAction. Also new: topological definition of observer (ℤ₂ + local i + N ≥ N_crit reflexions), fractality atom ≡ solar system.

• v1.1.7 (2026-04-19). Potential correction (P8). Effective action revised: the inflaton potential is explicitly specified as the standard α = 2/3 T-model V(φ) = Λ⁴ tanh²(φ/2) with φ = ln(τ₂) for τ₁ = 0. Previous specification V = Λ⁴ tanh²(H(τ)/2) with Dedekind eta is fully PSL(2,ℤ)-invariant but does not reproduce r ≈ 0.0025 — numerical analysis in problem-potencial.md shows it gives r ~ 10⁻⁵. The T-model is the correct effective potential along the τ₁ = 0 geodesic, verified by numerical slow-roll analysis in slow-roll-numericky.md (r = 0.00239 for N_e = 57). The microscopic derivation of this effective potential from a fully modular-invariant fundamental action is explicitly added as Open problem 8 (P8). The α = 2/3 value itself is unaffected — it comes from the kinetic sector (K = −1, Theorem 1), which remains fully PSL(2,ℤ)-invariant.

• v1.1.6 (2026-04-18, evening). Major session with extended resources. Five new formal results: (a) P3 resolved — QM as τ-geometry limit, Schrödinger + Born rule + Heisenberg derived (qm-limit-p3.md). (b) Theorem 2 completeness proved — exactly 4 modulation modes from Fuchsian classification (theorem-2-completeness.md). (c) Slow-roll numerics — T-model V = Λ⁴tanh²(φ/2) gives r = 0.00239 for N = 57, matching paper prediction (slow-roll-numericky.md). (d) Critical issue identified — fully modular-invariant V with η function does NOT give r = 0.0025, paper needs to specify T-model as effective potential (problem-potencial.md). (e) Λ_eff from Theorem 3 — ~10⁻¹²² in Planck units, matches observation to order of magnitude. (f) Lawvere → (2,3) direction — candidate argument that elliptic orders (2,3) follow from self-reference + discreteness (lawvere-23-derivace.md).

• v1.1.5 (2026-04-18). Two major advances. (a) Theorem 3 proved: quaternion frustration [X_a, X_b] = 2ε_{abc}X_c forces Ric_ab = 2δ_ab, R = 6, manifold = S³ = SU(2) with bi-invariant metric. Flat space algebraically excluded. Detailed Ricci computation in ricci-theorem-3.md. Complementary to Theorem 1: field space ℍ² (K = -1) and physical space S³ (K = +1), total curvature zero. (b) P1 partially resolved: Euler-Lagrange equations for the effective action derived explicitly in FRW background. The previously weakest link — identification K_fields = K = -1 — is now shown to be a direct consequence of Theorem 1 (the kinetic term IS the Poincaré metric). Prediction r = 2(1-n_s)² follows from EOM, not from borrowed α-attractor universality. Detailed derivation in eom-odvozeni.md. Also: Open problem 6 (formalization of preaxiom) gained concrete direction via category-theoretic reading (CCC, Lawvere 1969, Yanofsky 2003) — see seberference-kategorie.md.

• v1.1.4 (2026-04-17). Preaxiom formulated as x^x = "time is possibility applied to itself" (seberference-x-na-x.md). Branch cut of log(x) identified as structural Level 0 of a five-level hierarchy of cuts: branch cut → fundamental domain → ±I quotient → Re(τ) projection → nₛ (rez-a-prvni-chyba.md). Open problem 5 (P5) partially resolved: closed formula a₂ = ⌊(x−1)/(2−x)⌋ for CF break values of x^x = c; band structure with boundaries c_n = ((2n+1)/(n+1))^((2n+1)/(n+1)); a₂(π) = 5 proved equivalent to (11/6)^(11/6) < π < (13/7)^(13/7). Coincidence 5 = 2+3 = ord(i) + ord(ρ) noted but not explained.

• v1.1.3 (2026-04-14, night). Candidate mechanism for emergence of many-body structure from a single rotation, stated in Open problem 3. Proposed as a third AXIOM 1 reduction: an isolated rotating configuration is logically underdetermined because its moment of rotation is a relational quantity lacking a referent ("a direction without a second point"). The substrate resolves this inconsistency by forcing the other end of the relation into existence — symmetry breaking is not a dynamical process but a consistency requirement. The same reframing dissolves the black hole information paradox: information is a property of relations, not content stored in objects, and Hawking radiation is the substrate version of the same editing operation (subtractive + meta-additive) that characterizes understanding in the psyche. Structural parallel formalized in vedomi-jako-operator.md and cerna-dira-jako-pochopeni.md. Status: stated not proved; formalization deferred to v1.3.