This paper presents a QMU-first model of charged lepton flavor violation (CLFV) focused on the radiative decay ^+ e^+. In the Aether Physics Model (APM), the charged leptons correspond to distinct packing (holonomy) sectors on the Aether-unit lattice. In that setting, CLFV is modeled as holonomy leakage: a controlled mismatch between flavor-sector connections that survives the 5D 4D projection. A ledger-first workflow is used. Before introducing a channel model, the paper verifies a minimal QMU structural identity (the Ledger Closure Check) that anchors the unit system employed throughout. In the paper this is expressed as Aᵤ curl=Fq^2C^2. Using the correspondence FqC=c, this can be recognized as Aᵤ curl=c^2 and therefore matches the familiar electromagnetic propagation constant written in SI as c^2=1/ (₀₀). The role of this check here is to certify internal ledger closure before interpreting any dimensionless experimental observable as a bound on a QMU-defined invariant. The experimental observable is the branching ratio B ₄: = (^+ e^+) / (^+ e^+), which is intrinsically dimensionless and therefore directly compatible with QMU holonomy predictions without importing field-strength narratives. The central construction is a dimensionless leakage invariant defined as a normalized closed-path integral of an inter-sector defect one-form. Let _ and ₑ denote the effective connection one-forms for the muon and electron packing sectors (including chronovibrational closure and the observational projection). Define the defect one-form ₄: =_-ₑ and the leakage invariant ₄: = (2) ^-1 ₄. Exact sector orthogonality implies ₄=0, enforcing B ₄=0 in the present model; any nonzero branching ratio signals a specific geometric permission (an obstruction that survives domain pairing and projection). To leading order the paper writes the channel rate asB ₄ = K ₄\, ₄^2, where K ₄ is a dimensionless geometric coefficient determined by radiative emission geometry and selection rules. The paper emphasizes that the inferred scale of ₄ depends sensitively on whether K ₄ is order-unity or carries an explicit fine-structure suppression; this distinction materially changes the leakage bound extracted from a given branching-ratio limit. A multi-channel extension is developed in the same holonomy language. The same leakage invariant ₄ is taken to control the three ``golden'' muon CLFV channels ^+ e^+, ^+ e^+e^-e^+, and coherent conversion ^-N e^-N. Each observable is written in the factorized form R₎=K₎ ₄^2, implying channel-ratio tests independent of the leakage amplitude: B ₃₄/B ₄=K ₃₄/K ₄ andR ₄^ (N) /B ₄=K ₄^ (N) /K ₄. This yields an immediately falsifiable program: once the K-ratios are computed from seam/packing geometry, any one measured (or bounded) channel predicts the others without re-tuning. The paper then commits to a discrete obstruction picture for the leakage invariant on an octant domain decomposition of the positive electrostatic sheet (shared by e^+] and [^+). Assuming local exactness on domains, the loop integral reduces to quantized boundary jumps, producing a discrete spectrum ₄=^p ₄/N with ₄ and N=8. A seam-orientation sign rule derived from loxodromic geometry determines when the leading ^1 contribution cancels (forcing p 2) or survives (allowing p=1). On the pinned symmetric branch the paper commits to p=2 and shows that, when the radiative readout coefficient carries an explicit, the scaling B (e) ^1+2p=^5 places the natural ₄=1 level near the current 10^-13 experimental sensitivity. Finally, for coherent conversion the paper introduces a QMU-native nuclear packing overlap factor N and a mandatory timing (dwell) factor TN for pulsed experiments with delayed live windows. With muonic-atom mean life N, dead-time td, and live width t, the dwell factor is defined by TN (td, t) =e^-td/N (1-e^- t/N). This yields a clean separation between (i) the universal lepton-sector leakage invariant and (ii) how a chosen target and timing program exposes it. The resulting target+timing scaling relations are proposed as direct experimental discriminators and motivate concrete near-term tests (timing scans on a fixed target, target swaps at fixed timing, and a heavy-target stress test with earlier gating). The MEG II full-dataset null limit is interpreted as a direct bound on ₄ once K ₄ is fixed geometrically, and the paper outlines derivation milestones to compute the sector connections, classify obstruction classes, and publish falsifiable channel-ratio tables for standard conversion targets.
David Thomson (Tue,) studied this question.