HFG
Article III · HFG Dispatch · June 2026

The CP Phase Is a Manifold Invariant

The leptonic CP-violating phase is not a property of specific words in the fundamental group — it is forced by the arithmetic geometry of m003(−2,3). Zero free parameters.

For the past several months the Hyperbolic Flavor Geometry programme has stated: the leptonic CP-violating phase is given by δ = π + φ(aaB) + φ(baa), using the dominant eigenvalue branch of the holonomy representation, giving 195.91° against the PDG value of 197.0°.

A natural question arose. Is this number tied to the particular words aaB and baa? Could a different word choice give a different phase? Or does the value emerge from the manifold itself, independent of any word choice?

We now have the answer.

The Test

We enumerated all reduced words of length ≤ 6 in the fundamental group of the Meyerhoff manifold m003(−2,3), computed their holonomy eigenphases, and grouped them by their homology class in H₁(M) = ℤ/5. For each nontrivial class, we identified the first primitive geodesic whose eigenphase enters a window around the manifold's natural resonance — which turned out to be θ* = −180°, the condition that the eigenvalue is purely real and negative.

H₁ classFirst-in-window geodesicPhase φDistance D = |φ + 180°|
1aab−167.362°12.638°
2aaBABB−159.192°20.808°
3aaBAB−151.857°28.143°
4aaB−176.731°3.269°

The ℤ/5 group has two inverse pairs: (1,4) where 1+4 = 0 mod 5, and (2,3) where 2+3 = 0 mod 5. The PMNS pair is (1,4). Their sum of distances to the resonance is:

Key Result
D(1) + D(4) = 12.638° + 3.269° = 15.907°
D(2) + D(3) = 20.808° + 28.143° = 48.951°

The pair (1,4) is 3.1× closer to the −180° resonance than the competing pair (2,3) — without reference to the PDG value.

The Conjugacy Collapse

The full enumeration found 216 candidate word triples satisfying the H₁ constraints. All of them — every single one — gave the same CP phase: 195.91°, with the same error of 1.09° from the PDG value. This is not a coincidence. After collapsing by conjugacy, the apparent multiplicity of 216 reduces to a single canonical geometric object:

216 candidate word triples
length ≤ 3, H₁-correct
2 primitive geodesic classes
Class A: φ = −176.73°  ·  Class B: φ = −167.36°
1 inverse pair in ℤ/5
H₁ classes 1 and 4 = ±1 mod 5
δ_CP = 195.91°
PDG 197.0°  ·  1.09° error  ·  0 free parameters

The 216 apparent solutions are not 216 independent predictions. They are 216 representatives of the same canonical geometric pair — related by the symmetry group of the fundamental group of m003(−2,3).

What Is the −180° Resonance?

The condition φ = −180° means the dominant holonomy eigenvalue is a negative real number: λ = −|λ|. This forces the trace of the holonomy matrix to be real and negative — a distinguished locus in the space of PSL(2,ℂ) representations. It is an intrinsic arithmetic condition on the trace field K = ℚ(w), where w⁴ = w + 1 and disc(K) = −283.

The four geodesic classes in the table above are ordered by proximity to this locus. The pair (1,4) sits dramatically closer — the class-4 geodesic aaB has D = 3.27°, nearly at the resonance. This ordering is a property of the manifold's arithmetic structure, not of any fitting procedure.

The Unified Picture: PMNS and CKM

The same mechanism applies to the CKM manifold m006(−5,2), but with a different resonance angle. The CKM geodesics cluster near θ* = +90° — purely imaginary trace — rather than −180°. This corresponds to the Iwasawa factorization rather than the Borel factorization, which produces small quark mixing rather than large lepton mixing.

Unified Selection Rule
For each manifold, the mixing matrix is determined by the inverse H₁ pair whose first-in-window geodesics are closest to the manifold's phase resonance θ*:

m003(−2,3): θ* = −180° → Borel factorization → large mixing → PMNS
m006(−5,2): θ* = +90° → Iwasawa factorization → small mixing → CKM

Both manifolds select the same inverse pair (1,4) — the generators of ℤ/5. Both show a ≈ 3× advantage over the competing pair. The resonance angle is different, the factorization method is different, the matrix is different — but the selection mechanism is the same.

Falsification

This is a quantitative, testable prediction. The Meyerhoff manifold's geometry forces a specific value: δ = 195.91°. If future experiments (Hyper-Kamiokande, DUNE) move the central value of the CP phase away from 195.91°, this selection principle is falsified. If they move toward it — if the error bar narrows and includes 195.91° — the principle is supported.

The current PDG central value is 197.0° ± 27°. The HFG prediction sits well within the current uncertainty. The experimental error will eventually decide.

Publication Status
The arithmetic foundations underlying this result are established in the paper The Sextic-Octic Decomposition of the Meyerhoff Manifold Shape Polynomial, submitted to Research in Number Theory (manuscript RNTB-D-26-00299, June 8 2026). The selection principle mechanism is the subject of ongoing work. All computations are reproducible via github.com/drmlgentry/hyperbolic-flavor-scan.
← Article II: Why Hyperbolic Space Looks Impossible Substack version →