Gravity and the Standard Model from Embedded Geometry

Gravity and the Standard Model from Embedded Geometry Taras Låvenberg∗ Department of Mathematics, Stockholm University, Stockholm, Sweden Abstract We develop a four-dimensional framework in which gravity, gauge fields, and Higgs scalars arise from the geometry of an isometrically embedded four-manifold M4⊂R1,3+16. The normal bundle carries a natural SO(6) ×SO(10) structure: the SO(10) block furnishes a geometric gauge sector, while normal fluctuations in the vector 10 play the role of electroweak Higgs fields. We introduce a selection functional Ssel that fixes discrete topological data and acts as a sector projector. On a “balanced slice”, the action reduces to a sum of squares plus topological terms, locking coupling ratios to integers. We derive the chiral index IndD16 = 2ˆ k10 and realize three generations via flux splitting. To ensure observational consistency, we identify a projectable, hypersurface-orthogonal aether branch where the tensor speed is luminal (cT= 1) and PPN parameters vanish (α1,2= 0). Analyzing Coleman–De Luccia and Hawking–Moss instantons, we find that topological terms contribute only phases. A benchmark scenario on S4with a BPST instanton yields a unified scale ν∼1016 GeV, leading to a proton lifetime τ(p→e+π0)∼1035 yr and a normal-ordered neutrino spectrum with Pmν≃0.06 eV. This framework thus provides a concrete geometric realization of Standard Model generation structure while simultaneously stabilizing the gravitational sector on a phenomenologically viable Lorentz-violating branch. ∗[email protected] Contents 1 Introduction 4 1.1 Motivation and context ................................ 4 1.2 Summary of main results ................................ 5 1.3 Organization of the paper ............................... 6 2 Geometric framework 6 2.1 Embedded four-manifold and normal bundle ..................... 6 2.2 Normal SO(6) ×SO(10) split ............................. 7 2.3 Aether / khronon and foliation structure ....................... 9 2.4 Action and sector variables .............................. 9 3 Selection functional and sector fixing 11 3.1 Functional-analytic setup ............................... 11 3.2 Sector selection functional ............................... 13 3.3 Gauge-fixed uniqueness: Theorem ........................... 15 3.4 Compactness and existence .............................. 16 4 Quantization and one-loop consistency 18 4.1 Classical action and reduced variables ........................ 18 4.2 Canonical variables and constraints .......................... 19 4.3 Quantization map and physical Hilbert space .................... 20 4.4 One-loop renormalization and protected locus .................... 21 4.5 Reflection positivity and sector projector ....................... 22 5 Gauge and matter sector from the normal bundle 23 5.1 Normal SO(10) gauge block and Yang–Mills sector ................. 23 5.2 Instantons and (A)SD sectors ............................. 24 5.3 Fermions and chiral index ............................... 25 5.4 Higgs as a normal fluctuation ............................. 26 5.5 Flavor and Yukawa textures .............................. 27 6 Topological selection and coupling locking 28 6.1 Balanced slice and BPS-like decomposition ...................... 28 6.2 Sector selection and coupling-ratio locking ...................... 29 6.3 Normal SO(2)qroute and 5D inflow .......................... 30 6.4 EFT interpretation ................................... 32 7 Phenomenology and cosmology 32 7.1 FRW minisuperspace and aether sector ........................ 32 7.2 Vacuum decay: CDL and Hawking–Moss ....................... 33 7.3 Higgs metastability and inflationary de Sitter .................... 36 7.4 Flavor, neutrinos, and low-energy observables .................... 37 7.5 Benchmark scenario: S4/ BPST background .................... 38 8 Discussion and outlook 40 A Conventions and notation 43 A.1 Indices, metric, and curvature ............................. 43 A.2 Exterior calculus and Hodge dual ........................... 44 A.3 Group theory and trace normalizations ........................ 44 A.4 Topological densities and normalizations ....................... 45 2 A.5 Embedding index for SU(2) ⊂SO(10) ........................ 45 B Instantons, index theorems, and chiral multiplicities 46 B.1 BPST instanton in the embedded SU(2) ⊂SO(10) ................. 46 B.2 Index theorem and normalization of IndD16 ..................... 47 B.3 Flux splitting and family multiplicities after SO(10) →SU(5) ×U(1)χ...... 49 C PPN constraints, stability, and tensor speed 50 D Quantization, constraint algebra, and renormalization 52 E Euclidean bounces and thin-wall junction conditions 56 References 60 3 1 Introduction 1.1 Motivation and context A long-standing theme in gravity and high-energy theory is that four-dimensional spacetime may be viewed as a submanifold embedded in a higher-dimensional ambient space. Any smooth Lorentzian 4-metric can be realized as an isometric embedding into a higher-dimensional flat manifold, at the price of introducing additional normal directions and their associated geometry, as ensured by the classic embedding theorems [4–6]. In the present work we take this geometric perspective seriously and treat the normal bundle of an embedded four-manifold M4,→R1,3+16 as physical: the normal connection and second fundamental form become dynamical fields, and their curvature encodes gauge and scalar sectors coupled to four-dimensional gravity. Earlier embedding formulations of gravity in a higher-dimensional flat space were explored in [46]; our construction differs by keeping the normal bundle degrees of freedom fully dynamical and leveraging their curvature to generate gauge, Higgs, and aether sectors. Under standard assumptions, GR is the unique second-order metric theory in four dimensions [15], so we seek novel structure by enlarging the field content rather than altering the basic metric dynamics. The persistent cosmological constant problem remains a key motivation for exploring such extensions; for a review see [10]. The normal bundle to M4in R1,3+16 carries a natural SO(16) structure. We exploit a split SO(16) →SO(6) ×SO(10) in which the SO(10) block plays the role of a grand-unified gauge group, while the SO(6) block governs additional geometric degrees of freedom. In this language, what would ordinarily be introduced as independent Yang–Mills fields are instead components of the normal connection, and Higgs-like scalars arise as normal fluctuations of the embedded four-geometry itself. The resulting framework gives a unified geometric origin for gravity, gauge fields and Higgs multiplets, with topological charges in the normal bundle controlling chiral indices and generation counting. At the same time, there is considerable interest in controlled departures from exact local Lorentz invariance in the gravitational sector, both as a way to parametrize possible ultraviolet completions of GR and as a tool in cosmology. Classic scalar–tensor models such as Brans–Dicke theory and its phenomenological refinements provide one benchmark for confronting modified gravity with Solar System tests and gravitational experiments [36,37,41]. Among related Lorentz-violating theories, Einstein–aether and khronometric theories, in which a unit timelike vector or “khronon” field picks out a preferred foliation, have emerged as particularly robust: on suitable parameter loci they reproduce standard post-Newtonian phenomenology and are compatible with the observed luminal speed of gravitational waves [49]. In our construction this structure appears naturally from a distinguished SO(2) ⊂SO(6) subbundle of the normal bundle. The corresponding normal angle defines a projectable, hypersurface-orthogonal aether field uµthat selects a preferred time function while leaving the spatial geometry governed by the embedded four-metric. This leads to a Lorentz-violating but observationally safe setup in which the gravitational sector lives on the projectable, hypersurface-orthogonal branch of khronometric gravity, and there exists a “GR-safe” locus in parameter space where the tensor speed is exactly luminal (cT= 1) and the preferred-frame PPN parameters vanish (α1=α2= 0). Solar System and binary tests constrain deviations from GR at the 10−5–10−4level [38,39,43–45]. On this locus, homogeneous FRW cosmology reduces to the usual GR equations with a rescaled effective Planck mass, and all current bounds from gravitational waves and Solar System tests are satisfied. The embedded-geometry viewpoint then supplies, in addition, a geometric grand-unified sector, Higgs fields as normal fluctuations, and topological tools for selecting discrete sectors and locking dimensionless couplings, all within a single four-dimensional effective description. Our equations resemble trace-free or unimodular variants of GR [11–14], with the added geometric structure provided by the normal bundle. 4 1.2 Summary of main results For ease of reference we summarize here the main conceptual and technical results; detailed definitions and proofs are given in the sections indicated. •Sector selection and uniqueness. We construct a geometric “selection functional” Ssel[g, Ω] on the space of metrics and normal connections, built from elliptic positive squares, holonomy penalties and integer-normalized topological blocks. On a fixed compact four-manifold M4and for fixed discrete data (χ, k10, k6, . . .)the functional is shown to be coercive on a background–Coulomb slice and strictly convex once holonomy constraints are included. The main result (Theorem 3.1) is that, under a mild irreducibility assumption, Ssel admits a unique minimizer modulo gauge and diffeomorphisms, and effectively projects onto a chosen topological sector in the quantum theory (Section 3). •Chiral index and generation count from the normal SO(10) block. Coupling a single left-handed Weyl spinor in the 16 of SO(10) to the geometric gauge field ΠΩµΠof the normal bundle, we evaluate the Atiyah–Singer index and obtain IndD16 = 2 ˆ k10, Ngen =IndD16= 2|ˆ k10|,(1) where ˆ k10 is the vector-trace Pontryagin number in the SO(10) block. Gravitational contributions preserve the parity of the index on spin manifolds, and after breaking SO(10) →SU(5) ×U(1)χwe show how three chiral families arise via flux splitting and/or discrete Wilson lines while Ind D16 remains even (Section 5.3). •Balanced slice and coupling-ratio locking. On a distinguished “balanced slice” of configuration space we consider Euclidean backgrounds on which (i) the metric is Einstein and (anti)self-dual, and (ii) the normal SO(10) and SO(6) curvatures are (anti)self-dual in fixed blocks. In this sector the Euclidean action decomposes into a sum of manifestly non-negative squares plus a purely topological block Stop ∝αGB χ(M4)+g−2 10 k10 +g−2 6k6+··· , with χ, k10, k6∈Z. Imposing a simple calibration condition on the balanced slice then fixes the ratios of dimensionless couplings, αGB :g−2 10 :g−2 6, in terms of these integers. A complementary derivation based on five-dimensional Chern–Simons inflow or, equivalently, on a normal SO(2)qflux mq=c1(ν2)2∈Zquantizes the overall normalization. We emphasize that this “coupling-ratio locking” is a property of the balanced slice and is used as a boundary condition; away from it, standard renormalization group running applies (Section 6). •GR-safe aether branch and FRW sector. A distinguished SO(2) ⊂SO(6) subbundle of the normal bundle defines a smooth angle Θ(x)and a hypersurface-orthogonal, projectable aether field uµ∝∂µΘ. In the two-derivative regime this realizes the projectable khronometric / Einstein–aether theory in which the homogeneous FRW background obeys 3M2 ∗H2=ρwith an effective Planck mass M2 ∗>0. Similar structures appear in Hořava– Lifshitz gravity and its low-energy limit [47,48]. We identify a “GR-safe” locus αAE ≡c1+c4= 0, β ≡c1+c3= 0, λ ≡c2>0, on which the tensor speed is exactly luminal (cT= 1) and the preferred-frame PPN parameters vanish (α1=α2= 0), in agreement with gravitational-wave and Solar System constraints and subject to the PPN and GW constraints discussed in [42]. On this branch we extend the Coleman–De Luccia and Hawking–Moss vacuum-decay analysis: the decay exponent is controlled by a generalized de Sitter entropy built from M2 ∗and the scalar potential, while topological terms contribute only pure phases and do not affect the bounce spectrum (Section 2.3). 5 •Higgs from normal fluctuations and electroweak symmetry breaking. Normal displacements of the embedded four-geometry provide scalar modes in the vector 10 of SO(10). Under a standard breaking SO(10) →SU(4) ×SU(2)L×SU(2)Rwe identify the (1,2,2)component with an electroweak Higgs doublet H. We show that its covariant derivative is inherited from the normal connection and reproduces the usual SU(2)L×U(1)Y couplings, and that the normal-bundle potential naturally yields one light Higgs doublet and heavy coloured sextets / extra doublets on the balanced slice. At low energies the effective Lagrangian reduces to the Standard Model form V(H) = −µ2 HH†H+λH(H†H)2, reproducing the familiar electroweak mass relations for (mW, mZ, mh)(Section 5.4). •Benchmark scenario and phenomenological targets. To illustrate the framework we analyze a concrete background consisting of a round S4with an embedded SU(2) BPST instanton in the SO(10) block. Enforcing a scale closure condition between intrinsic and normal curvatures fixes a unification scale ν∼1016 GeV and, via the spectral a4coefficient, a unified coupling gU. Using the geometric matching to the Standard Model gauge factors and two-loop RG evolution with controlled heavy thresholds, we obtain realistic gauge unification and a parameter-free prediction for the proton lifetime in the leading p→e+π0 channel, as well as a neutrino spectrum with normal ordering Pmν≃0.06 eV and an effective neutrinoless double-beta mass mββ at the few-meV level. These serve as concrete experimental targets of the construction (Section 7.5). 1.3 Organization of the paper The rest of the paper is organized as follows. In Section 2we introduce the geometric framework: embedded four-dimensional geometries in a higher-dimensional flat ambient space, the structure of the normal bundle with its SO(6) ×SO(10) split, and the construction of the aether/khronon sector from a normal SO(2) subbundle. Section 3defines the selection functional on the space of metrics and normal connections, formulates the Sobolev and gauge-fixing setup, and proves the sector selection and uniqueness theorem. In Section 4we discuss the canonical formulation and quantization on the hypersurface-orthogonal, projectable branch, including the constraint algebra, the group-averaging (rigging map) construction, and one-loop consistency of the reduced theory. Section 5develops the gauge and matter sector from the normal SO(10) block, covering (anti)self-dual instantons, the chiral index and generation count, and the realization of the Higgs as a normal fluctuation. Section 6analyzes the balanced slice, the role of topological sectors, and the resulting coupling-ratio locking among the Gauss–Bonnet and gauge couplings. Phenomenological implications—including FRW cosmology on the GR-safe aether branch, vacuum decay and de Sitter entropy, flavor textures, and a detailed benchmark scenario with an S4/BPST background and derived low-energy observables—are presented in Section 7. Several technical aspects are deferred to the Appendices: conventions and normalizations, index and instanton details, PPN and stability analysis, quantization and renormalization details, and Euclidean bounce technology. 2 Geometric framework 2.1 Embedded four-manifold and normal bundle We take as our starting point a smooth, oriented four-manifold M4equipped with a Lorentzian metric gµν of signature (−,+,+,+). Throughout we assume that (M4, g)admits an isometric 6 embedding into a flat ambient space of dimension 1 + 3 + 16,1 X:M4−→ R1,3+16, gµν =ηMN ∂µXM∂νXN,(2) where ηMN is the ambient metric and M, N = 0,...,3 + 16 label ambient indices. Given the embedding XM(x)we introduce a tangent frame and an orthonormal normal frame. The tangent frame is eMµ≡∂µXM,(3) and an orthonormal frame {nM A}for the normal bundle NM4is chosen such that ηMN eMµeNν=gµν, ηMN eMµnN A= 0, ηMN nM AnN B=δAB,(4) with A, B = 1,...,16 labelling normal directions. The normal bundle thus carries a natural SO(16) structure, and later we will implement a reduction to SO(6) ×SO(10). The extrinsic geometry of the embedding is encoded in the second fundamental form BAµν and the normal connection ΩµAB. These are defined by decomposing the ambient Levi–Civita covariant derivative of the tangent and normal frames into their tangential and normal parts: ∇µeMν= ΓρµνeMρ+BAµνnM A,(5) ∇µnM A=−BAµρeMρ+ ΩµA BnM B,(6) where Γρµν is the Levi–Civita connection of gµν. The tensor BAµν is symmetric in (µ, ν)and measures the extrinsic curvature of M4inside the ambient space, while ΩµAB =−ΩµBA is an so(16) connection on the normal bundle. The intrinsic curvature of (M4, g), the second fundamental form, and the normal curvature RµνAB(Ω) = ∂µΩνAB −∂νΩµAB + ΩµAC ΩνC B−ΩνAC ΩµC B(7) are related by the Gauss–Codazzi–Ricci equations. Since the ambient space is taken to be flat, these reduce to purely algebraic and differential identities on (g, B, Ω): Gauss: Rµνρσ(g) = BAµρBAνσ −BAµσBAνρ,(8) Codazzi: ∇[µBAν]ρ= Ω[µABBBν]ρ,(9) Ricci: RµνAB(Ω) = BAµρBBρν−BAνρBBρµ.(10) These identities express the intrinsic Riemann tensor and the normal curvature entirely in terms of the extrinsic data of the embedding and will underlie the geometric construction of the gauge, Higgs, and aether sectors in the subsequent sections. Detailed derivations and sign conventions are collected in Appendix A. 2.2 Normal SO(6) ×SO(10) split The 16-dimensional normal bundle NM4carries a natural SO(16) structure induced by the ambient metric. For our purposes it is convenient to assume a fixed reduction of this structure group to SO(16) −→ SO(6) ×SO(10),(11) so that the normal fiber decomposes as NM4≃N6⊕N10,(12) 1For concreteness one may take the ambient space to be R1,3+16 with the standard Minkowski metric. We only need a local Nash embedding—charts that embed isometrically and patch together smoothly—and we do not assume a unique or globally complete embedding. Different global choices related by ambient isometries leave the intrinsic (g, B, Ω) data used in the action unchanged. 7 with rank(N6) = 6 and rank(N10) = 10. We will treat N6as the “geometric” normal block and N10 as the block that carries the SO(10) gauge sector and the Higgs degrees of freedom. At the level of the normal frame {nA}we choose a basis adapted to this splitting, nA=(nˆa,ˆa= 1,...,6, nα, α = 1,...,10,(13) so that {nˆa}span N6and {nα}span N10. In this basis the normal connection ΩµAB decomposes block-diagonally as ΩµAB = Ωµˆaˆ b0 0 Ωµαβ!,(14) with Ωµˆaˆ b∈so(6) and Ωµαβ ∈so(10). The corresponding normal curvatures are Rµνˆaˆ b(Ω) ∈so(6),(15) Rµναβ(Ω) ∈so(10),(16) and will be denoted collectively by R(6) and Frespectively: R(6) µν ≡Rµνˆaˆ b, Fµν ≡Rµναβ.(17) It is convenient to encode the SO(10) block by a projector on the normal bundle. Let Π : NM4−→ N10 (18) be the orthogonal projector onto the SO(10) subbundle, with components ΠAB=(0, A = ˆa, δαβ, A =α, (19) in the adapted frame. The complementary projector I−Πselects the SO(6) block. In terms of Π, the SO(10) connection and curvature can be written succinctly as Aµ≡Π ΩµΠ, Fµν ≡ΠRµν(Ω) Π,(20) and will be the geometric origin of the gauge field in the SO(10) sector. We will use two trace operations throughout: •tr6(·)denotes the matrix trace in the 6(vector) representation of SO(6) acting on N6. •tr10(·)denotes the matrix trace in the 10 (vector) representation of SO(10) acting on N10. The corresponding Pontryagin densities are normalized as 1 8π2ZM4 tr10 F∧F∈Z,1 8π2ZM4 tr6R(6) ∧R(6)∈Z,(21) and we denote the associated integers by ˆ k10 ≡1 8π2ZM4 tr10(F∧F), k6≡1 8π2ZM4 tr6R(6) ∧R(6).(22) These conventions will be used uniformly in the index computations and in the discussion of topological sector selection and coupling locking in later sections. 8 2.3 Aether / khronon and foliation structure A key ingredient in our construction is a preferred time foliation, described by a scalar khronon field Θand its associated unit timelike aether vector uµ. We restrict to the hypersurfaceorthogonal (“khronometric”) sector, in which the aether is everywhere proportional to the gradient of Θ, uµ=−∇µΘ p−gαβ∇αΘ∇βΘ, uµuµ=−1,(23) so that Θ = const defines a foliation of M4by spacelike hypersurfaces. We will further work on the projectable branch, where the lapse Nis a function of Θonly; on homogeneous FRW backgrounds this reduces to N=N(t)with t≡Θ. At the two-derivative level, the most general diffeomorphism-invariant aether Lagrangian compatible with the unit constraint uµuµ=−1can be written as [42,49] LAE =−M2 ∗hc1(∇µuν)(∇µuν)+c2(∇µuµ)2+c3(∇µuν)(∇νuµ)+c4uµuν∇µuα∇νuαi,(24) with dimensionless couplings ciand an effective Planck normalization M2 ∗. On the hypersurfaceorthogonal (khronometric) branch it is convenient to use the combinations αAE ≡c1+c4, β ≡c1+c3, λ ≡c2.(25) In these variables the scalar K≡ ∇µuµand the extrinsic curvature Kij of the Θ = const slices provide a natural parametrization of the aether sector; in particular, on an FRW ansatz with projectable lapse, ds2=−N(t)2dt2+a(t)2γijdxidxj, uµ=N−1(t)δµ 0,(26) one has aµ≡uν∇νuµ= 0,Kij =Hhij and K= 3H, with H≡˙a/(Na). A central role will be played by the GR-safe locus αAE = 0, β = 0, λ > 0,(27) on which the homogeneous background equations reduce to the GR Friedmann system with M2 Pl replaced by M2 ∗, and the tensor sector remains exactly luminal, cT= 1. On this locus the preferred-frame PPN parameters α1and α2vanish in the khronometric sector, so Solar System tests and current gravitational-wave bounds are satisfied. The parameter λcontrols only the scalar (khronon) kinetic term; λ > 0ensures the absence of ghosts and gradient instabilities in the scalar sector. These properties will be used repeatedly in our FRW minisuperspace analysis and in the discussion of cosmological phenomenology. 2.4 Action and sector variables The dynamical variables of the classical theory are (gµν,Θ, uµ; ΩµAB, BµνA; ΦA, ψα∈16, . . .),(28) where gµν is the induced metric on M4,Θand uµdescribe the khronon/aether sector, ΩµAB is the normal connection on NM4,BµνAis the second fundamental form, ΦAdenotes normal fluctuations in the vector of SO(10), and ψαare left-handed Weyl fermions in the spinor 16 of SO(10). The total classical action is S=Sgrav +SAE +Sgauge +Smatter ,(29) with the following building blocks. 9 3. Faddeev–Popov operators and trivial kernel. Linearizing the gauge conditions at (g∗,Ω∗) gives the Faddeev–Popov operators for diffeomorphisms and normal gauge transformations. In the connection sector one obtains an elliptic, self-adjoint operator of the form M16 =D† Ω∗DΩ∗:Hsc+1(adP16)→Hsc−1(adP16),(64) which is Fredholm of index zero with compact resolvent. Its kernel consists precisely of covariantly constant adjoint sections. The holonomy penalties Shol, with loops and targets chosen so that the joint centralizers are discrete, forbid such non-trivial covariantly constant sections in a neighborhood of (g∗,Ω∗): any adjoint section in the kernel of M16 would have to commute with all holonomies and hence lie in the Lie algebra of the centralizer, which is trivial by assumption. Thus ker M16 ={0}on the holonomy-constrained neighborhood. 4. Integer blocks are constant in a fixed sector. Within a fixed topological sector, the integrals ZM4 E4,ZM4 tr10(F∧F),ZM4 tr6(R(6) ∧R(6))(65) are locally constant under smooth deformations that preserve the underlying bundles and parity data. Hence Sint is an additive constant on the fixed sector and does not contribute to the gradient or Hessian of Ssel; it plays no role in the local convexity analysis. 5. Strict convexity and uniqueness. At (g∗,Ω∗)the Hessian of Score +Sgf is a positive, selfadjoint, strongly elliptic operator with at most a finite dimensional kernel (residual moduli). Step (3) shows that this kernel is trivial once Shol is included with sufficiently large weights. Therefore the full Hessian of Ssel is strictly positive on the background–Coulomb slice near (g∗,Ω∗), so Ssel is strictly convex there. Strict convexity implies the existence of a unique minimizer on the slice. Lifting back to the full configuration space, any other minimizer is related by a gauge transformation and diffeomorphism, as the slice intersects each orbit at most once in the neighborhood under consideration. This establishes gauge-fixed uniqueness of the minimizer of Ssel in a fixed integer sector. 3.4 Compactness and existence We now show that the selection functional Ssel admits at least one minimizer in each fixed discrete/topological sector, and that a global minimizer can be chosen from a finite set of such sectors. The argument follows the direct method of the calculus of variations, using the coercivity established above together with standard compactness results in Sobolev spaces. Coercivity and bounded sublevel sets. On the background–Coulomb slice, the quadratic core Score is chosen to be equivalent to the graph norm of a strongly elliptic operator of order 2sg in the metric directions and 2scin the connection directions, for sg>3and sc>2. Concretely, with h:= g−g∗, a := A−A∗, δΩ := Ω −Ω∗,(66) we may write schematically Score(h, a, δΩ) ≃αg (1 + ∆L,∗)sg/2h 2 L2+α10 (1 + ∆A,∗)sc/2a 2 L2+α16 (1 + ∆6,∗)sc/2δΩ 2 L2, (67) with positive constants αg, α10, α16. The gauge-fixing piece Sgf adds strictly positive quadratic penalties for the background–Coulomb conditions. By elliptic estimates on M4and the absence of Faddeev–Popov zero modes (Theorem 3.1), there exist constants c1, c2>0such that Score(h, a, δΩ) + Sgf(h, a, δΩ) ≥c1∥h∥2 Hsg+∥a∥2 Hsc+∥δΩ∥2 Hsc−c2(68) 16 on the slice. All remaining contributions, Shol and Sint, are nonnegative (up to an additive constant) and of strictly lower differential order: Shol is a finite sum of smooth bounded functions of the holonomies, and Sint is constant within a fixed integer sector. Thus, for each fixed sector, there exists C0such that Ssel(g, A, Ω) ≥c1∥g−g∗∥2 Hsg+∥A−A∗∥2 Hsc+∥Ω−Ω∗∥2 Hsc−C0(69) on the background–Coulomb slice. In particular, for any C∈Rthe sublevel set SC:= (g, A, Ω) on the slice :Ssel(g, A, Ω) ≤C(70) is bounded in Hsg×Hsc×Hsc. Weak compactness and Rellich. Let {(gn, An,Ωn)}be a minimizing sequence in a fixed integer sector, with each element taken on the background–Coulomb slice. Writing hn:= gn−g∗, an:= An−A∗, δΩn:= Ωn−Ω∗,(71) coercivity implies that {hn}is bounded in Hsgand {an},{δΩn}are bounded in Hsc. Since these are Hilbert spaces, there exists a subsequence (still denoted by n) and fields (h∗, a∗, δΩ∗)such that hn⇀ h∗in Hsg, an⇀ a∗in Hsc, δΩn⇀ δΩ∗in Hsc.(72) By Rellich–Kondrachov, the embeddings Hsg,→Hsg−1and Hsc,→Hsc−1are compact, and for sg>3,sc>2we also have compact embeddings Hsg,→C1and Hsc,→C0on M4. Therefore, up to a further subsequence, hn→h∗in C1, an→a∗, δΩn→δΩ∗in C0.(73) Define the limit configuration g∞:= g∗+h∗, A∞:= A∗+a∗,Ω∞:= Ω∗+δΩ∗.(74) The background–Coulomb conditions are linear and continuous in the relevant Sobolev topologies, so the slice is closed under weak limits; thus (g∞, A∞,Ω∞)still lies on the gauge slice. Moreover, the holonomy constraints defining the irreducible neighborhood are preserved by the C0convergence of the connections, since the holonomy along a fixed loop solves a linear ODE with continuous dependence on the connection. Lower semicontinuity and continuity of the blocks. We now inspect the behavior of each component of Ssel along the minimizing sequence. •Quadratic core. Score is a sum of squared Sobolev norms of (h, a, δΩ), hence it is convex and weakly lower semicontinuous: Score(h∗, a∗, δΩ∗)≤lim inf n→∞ Score(hn, an, δΩn).(75) •Gauge penalties. Each gauge-fixing penalty is a finite sum of terms of the form ZM4 φiDµΩµ(76) for smooth test functions φi, squared and weighted. For sc>2, the map Ω7→ DµΩµ is continuous from Hscto Hsc−1⊂L2, and φi∈C∞⊂Hsc−1. Thus each linear functional Ω7→ RφiDµΩµis continuous, and the resulting quadratic forms are weakly lower semicontinuous. The same applies to the metric gauge penalties. 17 •Holonomy constraints. For each fixed C1loop γ, the parallel transport equation along γ with coefficient Ω( ˙γ)has a unique absolutely continuous solution and depends continuously on the coefficient in the C0topology. The C0convergence Ωn→Ω∞therefore implies Holγ(Ωn)→Holγ(Ω∞)(77) in the relevant compact Lie group, and the group-distance terms in Shol converge: dGHolγ(Ωn), Uγ2→dGHolγ(Ω∞), Uγ2.(78) Hence Shol is continuous along the sequence. •Integer blocks. Within the chosen discrete sector, the integer blocks Sint are constant: the integrals defining χ(M4),ˆ k10 and k6do not change under smooth deformations that preserve the underlying bundles and spin structure. Therefore Sintgn, An,Ωn=Sintg∞, A∞,Ω∞ along the minimizing sequence. Combining these observations, we obtain Ssel(g∞, A∞,Ω∞)≤lim inf n→∞ Ssel(gn, An,Ωn),(79) so (g∞, A∞,Ω∞)is a minimizer in the fixed sector. Existence in each sector and global selection. We have shown that, for any fixed discrete data (χ, ˆ k10, k6, mq)and fixed bundles, Ssel admits a minimizer on the background–Coulomb slice, and thus modulo gauge and diffeomorphisms. If the allowed discrete data are restricted to a finite set of sectors {T(s)}s∈S(for instance, by fixing χand bounding the Pontryagin numbers), the same argument applies in each sector separately, producing a family of minimizers (g(s) ∞, A(s) ∞,Ω(s) ∞), s ∈S. (80) The global minimum of Ssel over the union of these sectors is therefore attained at one of these configurations. The integer-distance terms in Sint ensure that the global minimizer realizes the prescribed integer targets (χ, ˆ k10, k6, mq)in the sense discussed in Section 6. In summary, Ssel is coercive, has weakly compact sublevel sets on the gauge slice, is lower semicontinuous with respect to the relevant Sobolev topologies, and therefore admits a minimizer in each discrete sector, with a global minimizer among finitely many sectors. 4 Quantization and one-loop consistency 4.1 Classical action and reduced variables The classical dynamics is governed by the geometric action assembled in the previous sections: the Einstein–Hilbert term for the intrinsic geometry of M4, Yang–Mills terms for the normal SO(10) and SO(6) blocks with couplings (g10, g6), the khronometric two-derivative sector with parameters (M2 ∗, λ, αAE, β)on the projectable, hypersurface–orthogonal branch, and the topological terms encoded in the integers (χ, ˆ k10, k6, mq). On the balanced slice the positive-square pieces vanish, leaving the purely topological block Stop described in Section 6; away from that slice the full action provides the starting point for the canonical formulation and quantization discussed below. 18 4.2 Canonical variables and constraints We now pass to the canonical formulation of the embedded system on the hypersurface–orthogonal, projectable branch of the aether. The basic variables are the embedding of the spatial slice, the normal connection, and (optionally) the second fundamental form, together with their canonically conjugate momenta. Canonical pairs. Let Σbe a spatial slice of M4with coordinates xi, and let XM(x)denote the embedding of Σinto the ambient R1,3+16 at fixed khronon time. The canonical pairs we use are XM(x),ΠM(x),ΩAB i(x),Πi AB(x),(81) and, if we keep the second fundamental form as an independent variable, BijA(x),ΠijA(x).(82) Here Mis an ambient index, A, B are normal-bundle indices in the SO(6) ×SO(10) split, and i, j are spatial indices on Σ. The canonical Poisson brackets are {XM(x),ΠN(y)}=δMNδ(3)(x−y),{ΩAB i(x),Πj CD(y)}=δijδ[ACδB]Dδ(3)(x−y),(83) supplemented by the analogous brackets for (B, ΠB)when present. Projectable HO branch and reduced Hamiltonian. The khronon field Θdefines a preferred foliation by constant-Θhypersurfaces, and the aether uµ∝ ∇µΘis taken to be hypersurface– orthogonal and projectable. Projectability implies that the lapse is a function of time only, N=N(t), and the Hamiltonian constraint appears in a single integrated form. On this branch there is a unique reduced Hamiltonian, Hphys(t) = ZΣt d3xHphys[X, Π,Ω,ΠΩ,B,ΠB;t],(84) that generates evolution in the physical time t≡Θwhile preserving all constraints. Local lapse fluctuations are removed by the projectability condition and do not appear as independent canonical variables. First-class constraints. In addition to Hphys, the theory carries a set of first-class constraints encoding spatial diffeomorphism invariance, normal-frame rotations, and the embedding identities (Gauss–Codazzi–Ricci): •Spatial diffeomorphisms. The momentum (or “vector”) constraints Hi(x)≈0generate spatial diffeomorphisms on Σ. In smeared form, H[ξ] := ZΣ d3x ξi(x)Hi(x),(85) where ξiis a vector field on Σ. •Normal-frame Gauss constraints. The normal bundle carries an SO(6) ×SO(10) structure group. The corresponding Gauss constraints GAB(x)≈0generate local normal-frame rotations and act canonically on the normal connection and its curvature. In smeared form, G[Λ] := ZΣ d3xΛAB(x)GAB(x),(86) with ΛAB(x)=−ΛBA(x). 19 •Embedding (Gauss–Codazzi–Ricci) constraints. The compatibility of XM,BijAand ΩAB i with ambient flatness is enforced by additional first-class constraints that implement the Gauss, Codazzi and Ricci equations at the canonical level. These can be written schematically as CGCR(X, Ω, B)≈0and generate those gauge transformations that move along the space of equivalent embeddings. Constraint algebra. At the classical level, the smeared constraints close under the Poisson bracket in the usual Dirac fashion. Writing schematically H[ξ] := ZξiHi, G[Λ] := ZΛABGAB,(87) one finds {H[ξ], H[η]}=H[Lξη],(88) {H[ξ], G[Λ]}=G[LξΛ],(89) {G[Λ], G[Λ′]}=G([Λ,Λ′]),(90) where Lξdenotes the Lie derivative on Σand [Λ,Λ′]AB = ΛACΛ′CB −Λ′ACΛCB is the commutator in so(6)⊕so(10). The Gauss–Codazzi–Ricci constraints form a first-class ideal compatible with this algebra and commute weakly with Hphys on the projectable HO branch. Together, {Hi, GAB,CGCR}generate the gauge group of spatial diffeomorphisms and normalframe rotations, while Hphys generates physical time evolution in t= Θ. The quantum implementation of this structure is discussed in Section 4. 4.3 Quantization map and physical Hilbert space On the projectable, hypersurface–orthogonal branch, we adopt a Schrödinger representation for the canonical variables introduced in Section 4.2. Wavefunctionals depend on the configuration data on a spatial slice Σand on the physical time t≡Θ. Schrödinger representation. The kinematical state space is spanned by wavefunctionals Ψ[X, Ω, B;t]≡ΨXM(x),ΩAB i(x), BijA(x); t,(91) which we view as elements of a kinematical Hilbert space Hkin with a suitable Gaussian (cylindrical) measure on the configuration space. In this representation the basic operators act as b XM(x) Ψ = XM(x) Ψ,b ΠM(x) Ψ = −iℏδ δXM(x)Ψ,(92) b ΩAB i(x) Ψ = ΩAB i(x) Ψ,b Πi AB(x) Ψ = −iℏδ δΩAB i(x)Ψ,(93) and analogously for (BijA,ΠijA)when included. The reduced Hamiltonian b Hphys is promoted to an operator that generates t–evolution, iℏ∂tΨ[X, Ω, B;t] = b Hphys Ψ[X, Ω, B;t],(94) while the first-class constraints annihilate physical states. 20 Group averaging and rigging map. Physical states must be invariant under spatial diffeomorphisms and normal-frame rotations, generated by the constraints Hi(x)and GAB(x), as well as under the embedding gauge symmetries associated with the Gauss–Codazzi–Ricci constraints. We implement these symmetries via group averaging over the gauge group G ≃ Diff(Σ) ⋉SO(6) ×SO(10),(95) represented unitarily on Hkin by b U(g)for g∈ G. The rigging map (or group-averaging map) η:Hkin → H∗ kin is defined formally by η[Ψ1](Ψ2) := ZG dµG(g)Ψ1b U(g) Ψ2kin,(96) where dµGis a (regulated) invariant measure on Gand ⟨·|·⟩kin is the kinematical inner product. The integral projects onto states that are invariant under Gin the sense of Dirac. Physical inner product. The physical Hilbert space Hphys is obtained as the completion of the image of ηwith respect to the induced sesquilinear form. For kinematical states Ψ1,Ψ2∈ Hkin we define the physical inner product by Ψ1Ψ2phys := η[Ψ1](Ψ2) = ZG dµG(g)Ψ1b U(g) Ψ2kin.(97) By construction this inner product is invariant under the gauge group G, and b Hphys is self-adjoint with respect to ⟨·|·⟩phys on the projectable hypersurface–orthogonal branch. In particular, the reduced evolution operator b Uphys(t) = exp −itb Hphys/ℏ(98) acts unitarily on Hphys, preserving the physical norm and implementing time evolution in the aether clock t= Θ. 4.4 One-loop renormalization and protected locus We analyze radiative stability using the background field method with symmetry-preserving regulators (heat-kernel / proper-time) applied to the full (X, Ω, B)system on the projectable, hypersurface–orthogonal branch. The fields are split into background plus fluctuations, Φ→ Φ⋆+δΦ, and the one-loop effective action is constructed from the regulated functional determinant of the quadratic fluctuation operator. Gauge symmetry, normal-frame covariance, and diffeomorphism invariance (restricted to volume-preserving transformations on each slice) are maintained at every step. This symmetry structure is closely related to transverse diffeomorphisms [12]. On this branch, and within the two-derivative truncation for the aether sector, the divergent part of the one-loop effective action has the schematic form Γ(1) div =1 16π2ϵZd4x√−gA M2 ⋆R+B λ K2+higher-derivative terms,(99) with A, B numerical coefficients depending on the field content, and where ϵis the UV regulator parameter. Within this two-derivative truncation and using the symmetry-preserving regulator described above, we do not find one-loop counterterms proportional to the aether structures controlled by αAE ≡c1+c4and β≡c1+c3on the hypersurface–orthogonal, projectable branch. Equivalently, the one-loop beta functions in the two-derivative sector take the form µdM2 ⋆ dµ=A 16π2M2 ⋆, µdλ dµ=B 16π2,(100) µdαAE dµ= 0, µdβ dµ= 0,(101) 21 up to higher-derivative operators that are beyond our truncation. Within this truncation and regulator choice, we therefore do not see counterterms that would radiatively generate αAE or βif they vanish classically. We emphasize that this is not an all-orders non-renormalization theorem; higher-derivative operators or different regulator choices could in principle source these structures. We assume that a protective symmetry or UV completion maintains this locus to higher orders, ensuring consistency with gravitational wave constraints (|cT−1|<10−15). Topological densities behave as usual: the coefficients of the Gauss–Bonnet term and the Pontryagin terms (in both the gravitational and normal-bundle sectors) appear only as integerquantized levels in front of topological invariants, and they do not run under the renormalization group. In particular, the locked ratios among (αGB, g−2 10 , g−2 6)on the balanced slice are stable at one loop; only the overall non-topological normalization M2 ⋆receives logarithmic running. On homogeneous FRW backgrounds, the tensor sector derived in Section 2.3 has quadratic action S(2) T=M2 ⋆ 8Zdtd3x a3h(1 −β)˙ hij ˙ hij −(1 −β)(∂khij)2 a2i,(102) so that the tensor propagation speed satisfies c2 T= 1/(1 −β). Since βis not renormalized at one loop, the luminal condition αAE = 0, β = 0 (103) remains preserved by quantum corrections in the two-derivative truncation, and cT= 1 is maintained to this order. On the same locus, the preferred-frame PPN parameters α1and α2vanish in the classical theory, and their vanishing is not spoiled by the restricted set of counterterms required at one loop. We will therefore refer to αAE = 0, β = 0, λ > 0, M2 ⋆>0(104) as the consistency surface or GR-safe locus, which appears stable under one-loop renormalization within this two-derivative, projectable HO truncation. 4.5 Reflection positivity and sector projector In the Euclidean formulation, the selection functional Ssel enters the path integral through a weight Z[X⋆] = ZDgDΩD[matter] exp −SE[g, Ω, . . .]−Ssel,E[g, Ω; X⋆],(105) where SEis the ordinary Euclidean action and Ssel,Eis the Euclidean continuation of the sector selection functional described in Section 3. By construction, Ssel is a sum of positive-definite quadratic pieces (Score +Sgf +Shol) and a purely topological block Sint built from integernormalized densities: Ssel =Score +Sgf +Shol +Sint, Sint ∝χ(M4),ˆ k10, k6, mq, . . . (106) Within a fixed discrete sector (χ, ˆ k10, k6, mq, . . .)the associated integrals are locally constant under smooth deformations of (g, Ω), so Sint is an additive constant in that sector. In Euclidean signature, this constant is chosen to be a pure imaginary multiple of 2π, Sint,E= i θtop Ntop, Ntop ∈Z,(107) with θtop fixed by the calibration on the balanced slice. Consequently the integer block contributes only an overall phase exp−Sint,E= exp−iθtop Ntop(108) and does not alter the real part of the Euclidean weight. The real, positive-definite part of SE+Ssel,Eis entirely furnished by the quadratic core and gauge/holonomy penalties. 22 Because the sector projector modifies the path integral only by such integer-quantized phases, it preserves Osterwalder–Schrader reflection positivity of the regulated Euclidean theory. In particular, the reflection operator Θacting on field configurations across a fixed Euclidean time slice satisfies the usual positivity condition X i,j cicjΘOiOjE≥0for all finite sets {Oi}supported in tE>0,(109) since the phase factor from Sint,Ecancels between bra and ket in any OS inner product. Reflection positivity then guarantees the existence of a Hilbert space Hphys and a self-adjoint Hamiltonian ˆ Hphys generating unitary real-time evolution in the Lorentzian theory obtained by analytic continuation. On the projectable, hypersurface-orthogonal branch we use the khronon Θitself as a physical time variable. After imposing the constraints and implementing the sector projection, the reduced Hamiltonian ˆ Hphys is essentially self-adjoint on a dense domain of wavefunctionals Ψ[X, Ω, B; Θ], and the unitary evolution operator U(Θ2,Θ1) = exp −i(Θ2−Θ1)ˆ Hphys(110) preserves the physical inner product defined via the rigging map. Coherent states peaked on classical FRW backgrounds (with the aether aligned with the cosmological slicing) evolve semiclassically under U(Θ2,Θ1): their expectation values follow the FRW equations with the effective Planck mass M2 ⋆, while the selection functional acts only to pin the discrete/topological sector and does not disturb unitarity of the reduced dynamics. 5 Gauge and matter sector from the normal bundle 5.1 Normal SO(10) gauge block and Yang–Mills sector The normal connection ΩµAB on the rank–16 normal bundle N→M4naturally splits according to SO(16) ⊃SO(6) ×SO(10),(111) with the projector Πintroduced above selecting the SO(10) block. We define the geometric gauge curvature of the SO(10) factor by F≡ΠR⊥(Ω) Π ∈so(10) ⊗Ω2(M4),(112) where R⊥(Ω) is the curvature two–form of the full normal connection and Πprojects onto the so(10) subalgebra. In local coordinates xµthis gives the usual field strength Fµν =∂µAν−∂νAµ+ [Aµ, Aν], Aµ≡Π ΩµΠ∈so(10),(113) so that the SO(10) gauge sector is entirely encoded in the normal-bundle geometry. With the vector trace tr10 on so(10), the Yang–Mills action is SYM =1 2g2 10 ZM4 tr10 F∧⋆F=1 4g2 10 ZM4 d4x√−gtr10 FµνFµν,(114) where g10 is the SO(10) gauge coupling and ⋆is the Hodge dual with respect to gµν. For many constructions it is convenient to single out an embedded SU(2) subsector, ι: SU(2) ,→SO(10),(115) 23 supporting (anti)self-dual instantons. Denoting by Tr the fundamental SU(2) trace (normalized by Tr(tAtB) = 1 2δAB), the restriction of the SO(10) trace to the image of ιreads tr10(XY ) = κ10|ιTr(XY ), X, Y ∈ιsu(2),(116) with κ10|ι>0a fixed embedding–dependent constant. In this subsector the action takes the standard SU(2) form S(SU(2)) YM =1 2g2 eff ZM4 Tr F∧⋆F,1 g2 eff =κ10|ι g2 10 ,(117) so that the effective SU(2) coupling geff is fixed geometrically by the embedding ιand the SO(10) coupling g10. 5.2 Instantons and (A)SD sectors In the embedded SU(2) ⊂SO(10) subsector of the normal gauge block, the curvature two-form Fcan support (anti)self-dual instantons of BPST type. Finite-action SU(2) instanton solutions were first constructed in [19,20,25]. For a mathematical discussion of instantons on fourmanifolds, see e.g. [58]. With Tr the fundamental SU(2) trace, normalized by Tr(tAtB) = 1 2δAB, the (second Chern) topological charge is k≡1 8π2ZM4 Tr(F∧F)∈Z,(118) and the corresponding integer measured with the SO(10) vector trace is ˆ k10 ≡1 8π2ZM4 tr10(F∧F) = κ10|ιk, (119) with κ10|ιthe embedding factor introduced in Section 5.1. On R4(or conformally on S4) the BPST instanton of charge k=±1provides an explicit solution of the (anti)self-duality equations F=±⋆ F, (120) and saturates the Bogomol’nyi bound. In terms of the SO(10) action, SYM =1 2g2 10 ZM4 tr10(F∧⋆F ),(121) one finds SYM ≥4π2 g2 10 |ˆ k10|,with equality iff F=±⋆ F. (122) A crucial feature of these (A)SD configurations is that their Yang–Mills stress tensor vanishes identically. With TYM µν =1 2g2 10 tr10(FµαFνα)−1 4gµν tr10(FαβFαβ),(123) the algebraic identity tr10(FµαFνα) = 1 4gµν tr10(FαβFαβ)for F=±⋆ F implies F=±⋆ F =⇒TYM µν = 0.(124) These (anti)self-dual sectors therefore carry non-trivial topological charge and finite action, but no local stress–energy. On homogeneous and isotropic backgrounds they do not backreact on the FRW geometry at leading order, and in the “balanced slice” used below they furnish the gauge/topological part of the configuration while leaving the cosmological dynamics entirely controlled by the metric, aether, and matter sectors. The chiral anomaly [29,30] relates the divergence of the axial current to the topological density F˜ F, so instanton backgrounds source axial charge violation. In non-Abelian gauge theory these anomalies induce baryon and lepton number violation [31]. 24 5.3 Fermions and chiral index We couple a single left-handed Weyl spinor in the 16 of SO(10) to the geometric gauge connection extracted from the normal bundle. On a spin four-manifold M4with a Spin(10) lift of the SO(10) normal block, the relevant bundle is S16 ≡S(TM4)⊗16,(125) where S(TM4)is the spin bundle of TM4. The covariant derivative acting on sections of S16 is Dµ=∇µ+1 4ωµab γab + Π ΩµΠ,(126) with ωµab the Levi–Civita spin connection, γab the Dirac generators, and ΠΩµΠthe so(10)-valued connection on the chosen normal gauge block. The corresponding Dirac operator is / D16 =γµDµ: Γ(S16)→Γ(S16).(127) With the vector-trace normalization ˆ k10 ≡1 8π2ZM4 tr10(F∧F)∈Z,(128) and using tr16(·) = 2 tr10(·)for the SO(10) spinor and vector traces, the gauge part of the Atiyah–Singer index theorem [26] gives Ind / D16 =1 8π2ZM4 tr16(F∧F) = 2 ˆ k10,(129) up to a purely gravitational contribution proportional to the signature τ(M4). On spin fourmanifolds Rokhlin’s theorem [57] implies τ(M4)∈16Z, so the gravitational term is always a multiple of 32 and does not affect the parity of the index. Thus with the vector-trace convention above the SO(10) index is automatically even. The physical interpretation is straightforward: for a single Weyl fermion in the 16 the net number of chiral zero modes is given by the Atiyah–Singer index [26,28], Ngen ≡Ind / D16= 2 |ˆ k10|.(130) We identify one 16 with one Standard Model family (including a right-handed neutrino), so Ngen counts the number of chiral families in the unified SO(10) description and is topologically protected. After breaking SO(10) →SU(5) ×U(1)χ→SU(3)c×SU(2)L×U(1)Y,(131) the 16 decomposes as 16 →10 ⊕¯ 5⊕1, with corresponding chiral indices I10, I¯ 5, I1satisfying I10 +I¯ 5+I1= Ind / D16 = 2ˆ k10.(132) Flux choices and discrete SU(5) Wilson lines can redistribute chirality between the 10 and ¯ 5 sectors while preserving the total even index. In particular, it is possible to arrange I10 =I¯ 5= 3 and to pair the excess singlet modes, so that one obtains exactly three light Standard Model families out of an even unified SO(10) index. We return to explicit flux/Wilson-line realizations of this splitting when we discuss the breaking pattern and the benchmark scenario in Section 7.5. 25 6.4 EFT interpretation From the low-energy point of view, the relations among the couplings αGB :g−2 10 :g−2 6:αtop (184) obtained on the balanced slice should be understood as boundary conditions imposed at a distinguished scale ν, not as universal renormalization group (RG) invariants. In other words, the “locking” described above specifies how the topological data (χ, ˆ k10, k6, mq)fix the ratios of the dimensionless couplings at the calibration scale, but it does not prevent ordinary RG running away from that slice. Concretely, on the balanced slice the Euclidean action admits a BPS-like decomposition into positive squares plus a purely topological block, and the calibration condition (e.g. Stop = 0 or 2πℓ) algebraically relates the coefficients multiplying E4,tr10(F∧F),tr6R(6) ∧R(6)and fq∧fq. This fixes the ratios of αGB,g−2 10 ,g−2 6and αtop in terms of the integer data in the chosen sector. In the effective field theory sense, this is simply a statement about the UV matching conditions at µ=ν. Away from the balanced slice, the theory is described by a conventional EFT with running couplings. The gauge couplings g10 and g6obey the usual beta functions determined by the light spectrum, and the Einstein–Hilbert coefficient 1/G runs according to the standard gravitational RG equations. The integer levels appearing in the topological sector (e.g. the Chern–Simons levels in the 5D inflow picture or the quantized coefficients of E4and fq∧fq) do not run, but the dynamical couplings that control propagating degrees of freedom are renormalized in the usual way. Thus, “coupling-ratio locking” should be interpreted as a property of the balanced slice and of the corresponding UV boundary conditions at scale ν. It does not imply that the same ratios hold at all scales or for arbitrary off-slice configurations; rather, the low-energy values of (αGB, g10, g6, . . .)are obtained by RG-evolving from the locked boundary data at νdown to laboratory scales within the standard EFT framework. 7 Phenomenology and cosmology 7.1 FRW minisuperspace and aether sector On homogeneous and isotropic backgrounds we adopt the standard FRW ansatz with a projectable lapse, ds2=−N(t)2dt2+a(t)2γijdxidxj, k ∈ {0,±1}, R(3)(γ) = 6k, (185) and align the aether with the time foliation, uµ=N−1(t)δµ 0, uµ=−N(t)δ0 µ.(186) Projectability N=N(t)implies that the aether acceleration aµ≡uν∇νuµvanishes on the background, aµ= 0,(187) so we are on the projectable, hypersurface-orthogonal (HO) branch. The extrinsic curvature of the constant–time slices with induced metric hij =a2γij is Kij =1 2N˙ hij =H hij, H ≡1 N ˙a a,(188) so that K≡hijKij = 3H, KijKij = 3H2, K2= 9H2.(189) 32 At the two-derivative level the aether sector is parameterized by the standard khronometric combinations (αAE, β, λ)≡(c1+c4, c1+c3, c2),(190) and the action contains the invariants aµaµ,KijKij and K2with coefficients proportional to (αAE, β, λ). On the projectable HO branch aµ= 0, so the αAE term drops out already at the minisuperspace level. Writing the total gravitational+aether action (Einstein–Hilbert plus aether) in ADM form and inserting the FRW ansatz one finds a reduced Lagrangian of the form Lgrav+AE =Vk−3M2 ∗ N(1+β+ 3λ)a˙a2+ 3M2 ∗(1+β+ 3λ)k aN,(191) where Vk≡Rd3x√γis the comoving spatial volume and M2 ∗is the bare gravitational normalization appearing in the bulk action. The combination M2 eff ≡M2 ∗(1+β+ 3λ)(192) plays the role of an effective (cosmological) Planck mass on homogeneous backgrounds. Coupling this to a homogeneous matter sector with energy density ρ(t)and absorbing any cosmological constant into ρ, the total minisuperspace Lagrangian becomes Ltot =Vk−3M2 eff Na˙a2+ 3M2 effkaN −a3Nρ(t).(193) Variation with respect to the projectable lapse N(t)yields the (global) Hamiltonian constraint, which on the FRW ansatz reduces to the Friedmann equation 3M2 effH2+3kM2 eff a2=ρ(t),(194) while variation with respect to a(t)gives the acceleration equation, which can be written in GR-like form using the standard continuity equation ˙ρ+ 3NH(ρ+p)=0for the matter sector. On the GR-safe locus αAE = 0, β = 0,(195) the spin-2 sector has luminal propagation (cT= 1) and the preferred-frame PPN parameters α1, α2vanish [42]; only λsurvives in the homogeneous background. It is therefore convenient to package λinto the definition of M2 ∗and hence of M2 eff, so that on this locus the background evolution is exactly GR-like, 3M2 ∗H2+3kM2 ∗ a2=ρ(t),(196) with the single replacement M2 Pl →M2 ∗. In particular, for k= 0 the homogeneous cosmology is indistinguishable from GR at the level of the background expansion history, while the tensor sector remains strictly luminal on the αAE =β= 0 surface. 7.2 Vacuum decay: CDL and Hawking–Moss In this section and the associated Appendix, we allow for a field-dependent effective Planck mass M2 ∗(Φ), representing generic non-minimal couplings in the scalar–tensor sector. We now summarize the semiclassical description of vacuum decay, based on the seminal work of Coleman and collaborators [32,33]. We include the dependence on the effective gravitational coupling M2 ∗(Φ). Throughout we work in Euclidean signature with an O(4)–invariant ansatz ds2 E=dξ2+a(ξ)2dΩ2 3,Φ = Φ(ξ),(197) 33 where dΩ2 3is the metric on the unit three–sphere and ξ∈[0, ξmax]parametrizes the radial direction. The Jordan–frame vacuum energy density (including any cosmological constant) is denoted by U(Φ), and the effective gravitational coupling by M2 ∗(Φ). We assume a generic scalar–tensor coupling to the aether sector; all relevant foliation effects are packaged into the field–dependence of M2 ∗(Φ), while boundary terms that do not affect the decay exponent are dropped. Euclidean equations and boundary data Up to a boundary term, the Euclidean minisuperspace action can be written as SE= 2π2Zdξ a31 2Φ′2+U(Φ) −3M2 ∗(Φ) a21−a′2,(198) where primes denote derivatives with respect to ξ. Varying with respect to Φand ayields the scalar equation Φ′′ + 3a′ aΦ′=U,Φ(Φ) −3 2M2 ∗,Φ(Φ) 1−a′2 a2,(199) and the Hamiltonian (first integral) constraint a′2= 1 −a2 3M2 ∗(Φ) 1 2Φ′2+U(Φ).(200) For compact O(4) bounces one imposes regularity at both “poles”, a(0) = 0, a′(0) = 1,Φ′(0) = 0, a(ξmax) = 0,Φ′(ξmax)=0,(201) with Φ(ξ)interpolating between the false vacuum Φfand the vicinity of the true vacuum Φt.2 The on–shell Euclidean action simplifies by using the constraint to eliminate a′2, Son-shell E= 2π2Zdξ a3hU(Φ) −3M2 ∗(Φ) 1−a′2 a2i,(202) and the decay exponent is B≡SE[bounce]−SE[false vac],(203) with only the real part contributing to the tunnelling rate. The ζ1interaction contributes purely imaginary boundary terms in SEand hence drops out of Band of the Hessian. Coleman–De Luccia bounces and the one–negative–mode theorem A first gravitational extension of flat-space vacuum decay is provided by the Coleman–De Luccia picture [34]. A Coleman–De Luccia (CDL) bounce [34] is a non-trivial solution (a(ξ),Φ(ξ)) to the above equations that interpolates between the false vacuum and the basin of attraction of the true vacuum. Linearizing around such a solution and imposing O(4) symmetry, scalar and metric fluctuations can be reduced to a single gauge–invariant master variable Q(ξ, Ω3), whose quadratic action takes the Sturm–Liouville form S(2) E[Q] = 1 2X ℓ,m Zξmax 0 dξ a3Q′2 ℓm +ℓ(ℓ+ 2) a2+M2(ξ)Q2 ℓm,(204) 2For non–compact bounces (e.g. Minkowski or AdS false vacua), one integrates to large ξand imposes the usual asymptotics instead; the discussion below accounts for the same spectrum. When the barrier is broad enough that the Coleman–De Luccia negative mode lifts, the relevant saddle is the Hawking–Moss solution with Φpinned near the hilltop; the single negative mode we track is precisely the one that changes sign at that CDL/HM transition. 34 with an effective mass M2(ξ)determined by U,ΦΦ(Φ),M2 ∗(Φ) and their derivatives. On the scalar (spherically symmetric) mode ℓ= 0, the associated self–adjoint operator has a spectrum of real eigenvalues λn. For a regular, single–crossing CDL bounce that crosses the barrier once and satisfies the above boundary conditions, the standard oscillation theorem implies that there is exactly one negative eigenvalue in the ℓ= 0 sector and no negative modes with ℓ≥1: nCDL = 1.(205) The proof proceeds by (i) exhibiting at least one negative direction via a rescaling of the bounce radius, and (ii) showing, using the Sturm–Liouville node-counting argument, that the zero– energy solution of the ℓ= 0 fluctuation equation has at most one node, so there can be at most one negative eigenvalue. This one–negative–mode property is preserved by the mild deformations induced by M2 ∗(Φ) and ζ2, provided M2 ∗(Φ) >0along the bounce so that no ghosts appear in the reduced quadratic form. Hawking–Moss saddles and negative modes on S4 A Hawking–Moss (HM) saddle [35] is a homogeneous configuration with the field sitting at the top of the barrier, Φ(ξ) = Φtop =const,(206) and the metric given by a round four–sphere of radius H−1, a(ξ)=H−1sin(Hξ), H2≡U(Φtop) 3M2 ∗(Φtop).(207) For such backgrounds the fluctuation problem diagonalizes in scalar harmonics on S4and the spectrum of scalar modes is λℓ=ℓ(ℓ+ 3)H2+U,ΦΦ(Φtop), ℓ = 0,1,2,..., (208) with degeneracy dℓ=1 6(ℓ+ 1)(ℓ+ 2)(2ℓ+ 3). Negative modes are present whenever λℓ<0for some ℓ. The HM saddle has exactly one negative mode if and only if −4H2< U,ΦΦ(Φtop)<0,(209) so that λ0<0but λ1>0. At the threshold U,ΦΦ(Φtop)=−4H2,(210) the ℓ= 1 quintuplet becomes marginal; for more negative curvature U,ΦΦ <−4H2this multiplet contributes five additional negative modes and a branch of CDL solutions typically bifurcates from the HM solution and dominates the decay. The only role of the Lorentz–violating sector at this order is through the dependence of H2on M2 ∗(Φ) and, indirectly, on ζ2; the ζ1term, being a total derivative in SE, does not affect either the spectrum or the exponent. Role of M2 ∗(Φ) and foliation couplings The net effect of the modified gravitational sector on vacuum decay is twofold: •The effective Hubble parameter in a given vacuum is H2(Φv) = U(Φv) 3M2 ∗(Φv),(211) so the condition for a single negative mode in the HM case and the CDL existence/shape are controlled by the ratio U/M2 ∗rather than by Ualone. 35 •The coupling ζ2modifies M2 ∗(Φ) and hence enters through H2and the effective mass M2(ξ)in the fluctuation operator. For fixed U(Φ), increasing M2 ∗raises Band tends to stabilize a metastable vacuum; decreasing M2 ∗has the opposite effect. The ζ1interaction, in contrast, contributes only an imaginary boundary phase to SEon the O(4) ansatz and does not change Bor the number of negative modes. In particular, on the GR-safe locus αAE =β= 0 with M2 ∗(Φ) >0along the bounce, the CDL one–negative–mode theorem and the HM single–negative–mode window −4H2< U,ΦΦ(Φtop)<0 are preserved, with the standard GR results recovered in the limit of constant M2 ∗. 7.3 Higgs metastability and inflationary de Sitter We now apply the general formalism of vacuum decay, reviewed in the previous subsection, to the Standard Model Higgs sector. At large field values the RG–improved Higgs effective potential can be written schematically as Veff(h)≃1 4λeff(h)h4,(212) where his the canonically normalized radial Higgs mode and λeff (h)is the running quartic [51]. In the parameter region where the electroweak vacuum is metastable, λeff turns negative above an instability scale ΛI[51], and Veff develops a barrier separating the electroweak minimum at h∼vfrom a deeper region at h≫ΛI. The local maximum of the barrier is located at h=htop, with curvature V′′ eff(htop)<0and barrier height ∆V≡Veff (htop)−Veff(v). At late times, when the background curvature is negligible on the scale of the bounce, the decay of the electroweak vacuum is governed by a nearly flat–space Coleman bounce. The corresponding decay exponent Bflat is determined by the four–dimensional Euclidean action of the O(4)–symmetric solution in the potential Veff (h), with gravitational corrections suppressed by powers of Rb/H−1, where Rbis the bounce radius and His the Hubble parameter. In the metastable regime of interest Bflat is very large and the late–time lifetime of the electroweak vacuum exceeds the age of the Universe by many orders of magnitude; in what follows we assume that this condition is satisfied by the underlying SM parameters. During an inflationary phase, however, the relevant background is quasi–de Sitter, with a nearly constant Hubble parameter H2 inf ≃Uinf 3M2 ∗(Φinf),(213) where Uinf is the inflaton energy density and M2 ∗(Φinf)is the effective gravitational coupling on the inflationary background. In this regime, de Sitter curvature affects both the tunnelling solutions and the stochastic evolution of h. For sufficiently large Hinf the dominant decay channel is typically a Hawking–Moss (HM) transition in which the Higgs field homogeneously fluctuates to the top of the barrier at h=htop. Approximating the background as exact de Sitter and M2 ∗ as slowly varying, the HM exponent for the Higgs sector is BHM ≃SEhtop−SEh=v≃8π2 3H4 inf Veff(htop)−Veff (v)=8π2∆V 3H4 inf ,(214) up to corrections suppressed by slow roll and by derivatives of M2 ∗(Φ). As reviewed in Sec. 7.2, the HM configuration represents a physically acceptable saddle if −4H2 inf < V ′′ eff(htop)<0,(215) in which case it carries a single negative mode and legitimately contributes to the decay rate. 36 The probability that a given Hubble patch survives Nee–folds of inflation without a transition to the unstable region can be estimated as Psurv ≃exp−ΓHM V4 H4 inf ,ΓHM ∼H4 inf e−BHM ,(216) where V4∼H−4 inf Neis the four–volume accumulated over Nee–folds and we have absorbed order–one prefactors into ΓHM. Requiring Psurv ≈1over the observable inflationary history translates into a lower bound on BHM, and hence an upper bound on Hinf for a given barrier height ∆V: BHM ≳O(1) ×Ne=⇒H4 inf ≲8π2 3Ne ∆V. (217) In our framework this inequality constrains not only the inflationary energy scale but also M2 ∗(Φinf), since H2 inf ∝Uinf/M2 ∗. For fixed inflaton potential Uinf , decreasing M2 ∗increases Hinf and tightens the metastability bound, while increasing M2 ∗reduces Hinf and makes the electroweak vacuum more robust during inflation. Thus the effective Planck mass on the inflationary branch, controlled in our model by the aether and embedding data, is directly constrained by the requirement that the Higgs vacuum survives the inflationary de Sitter phase. 7.4 Flavor, neutrinos, and low-energy observables In the lepton sector the same geometric ingredients that generate hierarchical charged–fermion Yukawas — the normal–bundle U(1)Hcharges, holonomies, and overlap factors discussed above — also control the neutrino masses and mixings. We work with a type-I seesaw [21,27], in which the three right-handed neutrinos NRi transform as singlets under the Standard Model gauge group but carry definite geometric charges under U(1)Hand are localized in the normal directions in a manner analogous to the charged fermions. The Dirac Yukawa matrix Yνinherits the same Froggatt–Nielsen structure as the up-type quarks, with entries suppressed by powers of the spurion ε≡ ⟨S⟩/M and by overlap factors ηm(ν) ij from the normal zero-mode profiles. The heavy Majorana mass matrix MRis taken to be approximately diagonal and hierarchical, MR= diag(M1, M2, M3),(218) with Miset by the same geometric scales that control the normal curvatures. Integrating out NRi yields the standard seesaw relation mν=−v2YνM−1 RYT ν,(219) where v≃246 GeV is the Higgs vacuum expectation value. For the benchmark textures we employ, this construction produces a normal ordering of light neutrino masses m1∼few meV, m2∼ O(10) meV, m3∼ O(50) meV,(220) and mass splittings ∆m2 21 and ∆m2 31 in good agreement with oscillation data [53] at the order-ofmagnitude level. The sum of masses Σmν=m1+m2+m3is close to the minimal value allowed by normal ordering and is compatible with present cosmological bounds [53] in the benchmark scenario. The effective Majorana mass probed by neutrinoless double beta decay, mββ, is determined by the same geometric data [52], mββ =U2 e1m1+U2 e2m2eiα21 +U2 e3m3eiα31 ,(221) 37 through the PMNS matrix entries Uei and the Majorana phases α21, α31. In our minimal implementation, the relative phase of the lighter pair tends to be small while the heaviest state is approximately opposite in phase, leading to a preferred band mββ ∼few ×10−3eV,(222) typically in the 1–4meV range in the benchmark scenario. Much larger values would require either substantial departures from the geometric texture (e.g. modified U(1)Hcharges or overlaps) or fine-tuned phase choices that our setup does not naturally favor. In summary, the flavor structure, neutrino spectrum, and CP phases are all tied to the same geometric charges and holonomies that govern the quark and charged–lepton Yukawas. This yields a coherent picture in which quark hierarchies, PMNS mixing, the near-minimal neutrino mass sum, and a meV–scale mββ are correlated outputs of the underlying normalbundle geometry. 7.5 Benchmark scenario: S4/ BPST background To make the preceding constructions concrete we now specialize to a simple, fully explicit background on which all ingredients can be computed in closed form. We take (M4, g⋆) = (S4, R2ground),(223) with ground the unit-radius round metric and Rthe physical radius, and embed a BPST k= 1 instanton in a fixed SU(2) ⊂SO(10) block of the normal connection Ω. The curvature in this block is (anti)self-dual, F≡ΠR⊥(Ω⋆) Π = ±⋆ F, ˆ k10 = 1,(224) and the metric is Einstein with constant scalar curvature. On this background the Yang–Mills stress tensor vanishes, TYM µν [F] = 0, and the Gauss–Bonnet density is constant. This places (S4, g⋆,Ω⋆)on the balanced slice discussed in Section 6. Scale closure and spectral normalization. There are two a priori distinct UV scales in the geometric setup: (i) the scale ν⊥set by the rms normal curvature, and (ii) the induced-gravity scale νIG extracted from the a2coefficient of the spectral action. On the S4/BPST background both are simple functions of R−1. Imposing scale closure, ν⊥=νIG ≡ν, (225) fixes νas a numerical multiple of the sphere radius, ν≃cνR−1,(226) with cνof order unity. In the benchmark example we quote numbers with ν≈2.1×1016 GeV, R−1≈1.0×1016 GeV,(227) but the derivation of the subsequent relations does not depend on these specific values. Moving away from the scale closure condition ν=νIG would shift the geometric matching scale, modifying MXand therefore the proton lifetime estimate below. 38 Spectral a4and unified coupling. The gauge kinetic terms are read off from the a4(D2) coefficient of the geometric Dirac operator on (S4, g⋆,Ω⋆)[50], with the normal SO(10) block providing the geometric GUT sector. In particular, the piece proportional to tr10 FµνFµν takes the form a4(D2)F2=Cspec ZS4 d4x√g⋆tr10(FµνFµν),(228) where Cspec is a positive, purely geometric constant. Matching to the canonical Yang–Mills normalization, SYM =1 2g2 10(ν)ZS4 d4x√g⋆tr10(FµνFµν),(229) identifies a single unified coupling gU≡g10(ν)via 1 g2 U =Cspec.(230) After reduction to the Standard Model gauge group with SU(5)-normalized hypercharge, the embedding of the SM generators in SO(10) implies g1(ν) = g2(ν) = g3(ν) = gU,sin2θW(ν) = 3 8[54],(231) up to calculable group theory factors fixed by our trace conventions. Geometric split and heavy gauge-boson mass. On the S4/BPST background the broken generators in the SO(10) →SU(3)c×SU(2)L×U(1)Ydecomposition acquire masses from the normal curvature, much as in ordinary SO(10) GUTs they would from a Higgs in the adjoint. The mass of the heavy (X, Y )gauge bosons is of the form MX=csplit gUν, (232) where csplit is a geometric factor determined by the embedding of the broken generators and the normalization of the curvature. For a canonical choice of embedding one finds c2 split = 5/24, which in the numerical example quoted above gives MX∼7×1015 GeV.(233) Thresholds and two-loop running. Below νthe heavy fields descending from the SO(10) normal connection and from normal fluctuations contribute threshold corrections to the running of the SM couplings. It is convenient to summarize their effect in terms of the standard one-loop threshold sums Ki≡X A b(A) iln ν MA , i = 1,2,3,(234) where b(A) iare the usual one-loop beta-function weights of each multiplet Aunder Giand MA are their masses. In the simplest implementation all guaranteed heavy multiplets lie at scales MA∼ν, so the Kiare O(1) and do not spoil unification. For phenomenological applications we consider an early decoupling of color, in which color-charged spectators sit somewhat below νwhile electroweak spectators remain close to ν, enhancing asymptotic freedom in QCD. This can be arranged while keeping the Kifixed to a given target set. Starting from the unified boundary condition g1(ν)=g2(ν)=g3(ν)=gUand the chosen Ki, we evolve the couplings down to µ=mZusing the two-loop SM renormalization group equations with step decoupling at each threshold MA. In the benchmark scenario one obtains values at the Zpole of the form αs(mZ)≃0.118,sin2θW(mZ)≃0.231, α−1 em(mZ)≃137,(235) within current experimental uncertainties, without introducing independent continuous gauge couplings beyond the geometric input at ν. 39 Geometric Inputs & Assumptions Scale closure condition ν=νIG ∼2×1016 GeV Topological Sector S4background, k= 1 BPST instanton Threshold assumption Step-decoupling at MXand ν Derived Observables Unified coupling gUFixed by spectral a4(D2)(gU∼0.7) Heavy gauge mass MX∼7×1015 GeV Proton lifetime τ(p→e+π0)∼1035 yr Sum of neutrino masses Pmν≃0.06 eV (Normal Ordering) Table 1: Summary of inputs and derived observables in the S4/BPST benchmark scenario. The values for proton lifetime and neutrino masses are indicative predictions of the geometric sector selection, subject to threshold uncertainties. Proton decay. The dominant baryon-number violating process in the minimal non-supersymmetric setup is the dimension-six gauge-boson-mediated channel p→e+π0. Using standard chiral perturbation theory and lattice inputs [40], the partial lifetime can be written as τ(p→e+π0)≃64πf2 π m5 p M4 X α2 UA2 R|αH|2(1+D+F)2,(236) where mpis the proton mass, fπthe pion decay constant, D, F the axial couplings, ARthe shortdistance renormalization factor, and αHthe hadronic matrix element of the relevant three-quark operator. Inserting the geometric MXand gU, and representative lattice and chiral values for the low-energy constants, one finds a benchmark prediction τ(p→e+π0)∼1035 yr,(237) with an overall uncertainty of order a factor of a few, dominated by the hadronic matrix element and the precise value of MX. Summary of scales and observables. For ease of reference, Table 1collects the main scales and observables in the S4/BPST benchmark scenario. 8 Discussion and outlook The framework emerging from the preceding analysis is conceptually simple, even though its technical realization is intricate. One starts from a purely geometric input: an embedded, oriented four-manifold (M4, g)in a higher-dimensional flat space with a normal bundle carrying an SO(6) ×SO(10) structure. The normal connection Ωprovides, simultaneously, (i) the geometric data needed to define a selection functional on the space of sectors, and (ii) the seed for the gauge and Higgs sectors identified with the Standard Model. The sector selection functional Ssel[g, Ω; X∗]singles out, within a given discrete topological class, a preferred background up to diffeomorphisms and normal-frame gauge transformations. On this background, the normal SO(10) block yields a Yang–Mills sector and chiral fermion index, the normal fluctuations provide an electroweak Higgs doublet, and the residual geometric data encode flavor structure. Finally, the aether/khronon sector and the FRW minisuperspace analysis connect the high-energy geometric construction to low-energy cosmology and vacuum stability. 40 From geometry to the Standard Model sector. At the geometric level, the main ingredients are the embedded four-manifold, the normal bundle and its SO(6) ×SO(10) split, and the aether/khronon foliation. These determine the classical action and the discrete labels (χ, ˆ k10, k6, mq, . . .)entering the topological blocks. The sector selection functional introduced in Section 3provides a functional-analytic mechanism to select a unique representative in each discrete sector, under mild hypotheses on irreducibility and holonomy. Theorem 3.1 shows that, on a suitable background–Coulomb slice, the minimizer of Ssel is unique up to gauge and diffeomorphisms, and Section 3.4 establishes existence in each sector. Once a background is fixed, the normal SO(10) block yields a geometric Yang–Mills sector with curvature F= ΠR⊥(Ω)Π and action ∝tr10(F∧⋆F), including (A)SD subsectors that carry topological charge without contributing to the stress tensor. The fermion content is encoded in the index of the Dirac operator on S(T M4)⊗16, IndD16 = 2ˆ k10, Ngen =|IndD16|= 2|ˆ k10|,(238) so that the number of chiral generations is a topological property of the normal bundle (Section 5.3). The Higgs sector arises from normal fluctuations transforming in appropriate representations under SO(10), with the electroweak doublet(s) identified in the (1,2,2) of SU(4) × SU(2)L×SU(2)R(Section 5.4). The resulting low-energy Higgs Lagrangian has the familiar form |DµH|2−V(H)with V(H)supporting electroweak symmetry breaking and the usual tree-level relations among mW,mZand the VEV v. Flavor and CP structure are likewise tied to geometric data. A distinguished normal SO(2)q⊂ SO(6), together with its flux mqand holonomy, acts as a horizontal symmetry that constrains Yukawa couplings through selection rules and overlap integrals of localized zero modes. This leads to hierarchical textures Y(X) ij ∝εn(X) ij ηm(X) ij with εand ηgeometric spurions, and allows for quark and lepton hierarchies and CKM-like mixing patterns that are ultimately of geometric origin. GR-safe locus and robustness. A key consistency requirement is that the Lorentz-violating aether sector does not spoil the excellent agreement of General Relativity with tests of gravity. By focusing on the projectable, hypersurface-orthogonal branch with scalar khronon and aether uµ∝∂µΘ, and by restricting to the locus αAE =β= 0,(239) one ensures that (i) the background FRW evolution is GR-like with an effective Planck scale M2 Pl →M2 ∗, (ii) the tensor propagation speed cTis luminal, and (iii) the PPN parameters α1 and α2vanish. The one-loop analysis shows that, on this consistency surface, the dangerous combinations do not run at leading order: topological levels are protected, and the parameters controlling cTand α1,2are not regenerated by radiative corrections in the projectable HO branch. This suggests that the GR-safe locus is technically natural, although a full two-loop and nonperturbative stability analysis remains to be completed. The construction still leaves a discrete set of choices unfixed. These include the topological integers (χ, ˆ k10, k6, mq), the choice of embedded instanton subsector and holonomy structure, and possible discrete Wilson lines or fluxes used to break SO(10) down to the SM gauge group and lift unwanted exotics. The benchmark scenario on S4with a BPST instanton, presented in Section 7.5, shows that for at least one such choice it is possible to obtain a closed-scale hierarchy, realistic gauge coupling unification, viable Higgs metastability, and a proton lifetime compatible with current bounds. More generally, one expects different sectors to map to different low-energy spectra and observables; in this sense the discrete topological data play a role analogous to a finite set of “UV hyperparameters” that label distinct universality classes. 41 In four dimensions we can truncate the formal series to ˆ A(TM4)=1−1 24p1(TM4),(283) ch(E) = rank(E) + 1 8π2trR(F∧F),(284) where p1(TM4)is the first Pontryagin class of the tangent bundle, and trRdenotes the trace in the representation Rof Gcarried by E. In the conventions of Appendix Awe write p1(TM4)=−1 8π2Tr(R∧R),(285) with Tr the trace in the vector representation of SO(4). Multiplying and keeping only the 4-form piece yields IndDE=1 8π2ZM4 trR(F∧F) + rank(E) 192π2ZM4 Tr(R∧R).(286) The second term is the gravitational Pontryagin contribution, proportional to the tangent-bundle Pontryagin number. For the backgrounds of interest (e.g. M4=S4with the round metric, or spatially closed FRW slices with parity symmetry) the integral RM4Tr(R∧R)vanishes, so the gravitational contribution to the index vanishes. More generally, one can arrange for the gravitational Pontryagin term to cancel in the net chiral index by balancing multiplets in conjugate representations; see the discussion in Section 5.3. In the presence of a boundary ∂M4(for example, for Euclidean bounce geometries that asymptote to FRW or Minkowski at infinity) the appropriate statement is the Atiyah–Patodi– Singer (APS) index theorem [28]. In that case one has IndDE=ZM4 ˆ A∧ch(E)4-form −1 2η(D∂M4)+h,(287) where η(D∂M4)is the η-invariant of the induced Dirac operator on the boundary, and his the dimension of its kernel. Equivalently, the boundary terms can be written in terms of three-dimensional Chern–Simons forms built from Aand the spin connection. For the quasihomogeneous backgrounds considered in the main text the boundary contribution can be chosen to vanish (or cancel between sectors), so that the bulk formula (286) remains valid. From tr16 to tr10 and IndD16 = 2ˆ k10 Specializing to E=S(T M4)⊗16, the vector bundle associated to the chiral spinor in the 16 of SO(10), we set rank(E) = 16 and write Ffor the SO(10) curvature in the 16. Dropping the gravitational Pontryagin term as explained above, the index reduces to IndD16 =1 8π2ZM4 tr16(F∧F).(288) The traces tr16 and tr10 are related by the ratio of Dynkin indices for the spinor and vector representations of SO(10). With the conventions of Appendix Awe have tr16(TaTb) = 2 tr10(TaTb),(289) which implies tr16(F∧F) = 2 tr10(F∧F).(290) Using the definition ˆ k10 ≡1 8π2ZM4 tr10(F∧F)∈Z,(291) 48 we obtain from (288) IndD16 =1 8π2ZM4 2 tr10(F∧F)=2ˆ k10.(292) Thus the net number of left-handed minus right-handed zero modes in the 16 is fixed by the SO(10) instanton number ˆ k10. In particular, the net number of chiral families is Ngen =IndD16= 2 |ˆ k10|.(293) The evenness of the index follows from the relative trace normalizations: a single unit of SO(10) instanton number produces two net 16’s. As discussed in the main text, the phenomenologically interesting case Ngen = 3 can then arise after symmetry breaking and flux/Wilson-line splitting within each 16, rather than directly at the SO(10)-symmetric level. B.3 Flux splitting and family multiplicities after SO(10) →SU(5) ×U(1)χ We now outline how the topological data encoded in ˆ k10 splits after breaking SO(10) to SU(5)× U(1)χ, and how this is reflected in the multiplicities of SU(5) multiplets and ultimately in SM families. Decomposition of representations and curvature Under SO(10) →SU(5) ×U(1)χ, the spinor 16 decomposes as 16 −→ 10+1 ⊕¯ 5−3⊕1+5,(294) where the subscripts denote U(1)χcharges in a standard convention. The SO(10) connection A10 decomposes as A10 −→ A5⊕Aχ,(295) with curvature F10 −→ F5⊕Fχ,(296) where F5is an SU(5) field strength and Fχis an abelian 2-form. In terms of the SU(5) fundamental trace tr5and the U(1)χfield strength, the SO(10) instanton number becomes ˆ k10 =1 8π2ZM4 tr10(F10 ∧F10)=k5+κχmχ,(297) where k5≡1 8π2ZM4 tr5(F5∧F5)∈Z(298) is the SU(5) instanton number, and mχis an integer measuring the U(1)χflux (a quadratic expression in its first Chern class). The coefficient κχis a group theory factor that depends on the normalization of the U(1)χgenerator; for the purposes of this paper we keep it implicit and simply regard (k5, mχ)as a pair of discrete data that together determine ˆ k10. Splitting of the index and multiplet multiplicities The total index (292) can be decomposed into contributions from the SU(5) representations in (294). Let D10,D¯ 5, and D1denote the Dirac operators acting on sections transforming as 10+1, ¯ 5−3, and 1+5, respectively. Then IndD16 = IndD10 + IndD¯ 5+ IndD1,(299) 49 with each index given by a bulk expression of the form IndDRq=1 8π2ZM4 trRF5∧F5+q2 8π2ZM4 Fχ∧Fχ+··· ,(300) where R=10,¯ 5,1,qis the U(1)χcharge, and the ellipsis denotes possible mixed SU(5)–U(1)χ and gravitational contributions (which vanish for the simplest backgrounds considered here). In particular, the multiplicity of chiral 10’s is N10 ≡IndD10 =a10 k5+b10 mχ,(301) and similarly N¯ 5≡IndD¯ 5=a¯ 5k5+b¯ 5mχ, N1≡IndD1=b1mχ,(302) with integer coefficients (a10, b10, a¯ 5, b¯ 5, b1)fixed by the group theory invariants of the representations and the normalization of U(1)χ. The total number of chiral SU(5) families is then Ngen =N10 =N¯ 5,(303) with the equality N10 =N¯ 5enforced by anomaly-cancellation conditions and the requirement that complete 16’s of SO(10) arise in the UV. In the simplest branch, in which the U(1)χflux is tuned such that the splitting between N10 and N¯ 5vanishes, the family number is directly tied to the SO(10) instanton number via Ngen =IndD16= 2 |ˆ k10|.(304) More general choices of (k5, mχ)allow for flux-induced splitting of multiplet multiplicities within each 16 and hence for richer model-building options. In the present work we focus on the minimal, flux-aligned branch that leads to an integer number of complete families and is compatible with the topological locking conditions on the balanced slice. C PPN constraints, stability, and tensor speed In this appendix we collect the parametrized post-Newtonian (PPN) constraints relevant for the aether / khronometric sector, and summarize the conditions for the absence of ghosts and gradient instabilities. Throughout we work on the hypersurface-orthogonal (HO), projectable branch described in Section 2.3, and expand around an asymptotically flat background. Mapping of parameters The most general two-derivative aether Lagrangian can be written in the Einstein–aether form Sae =M2 ∗ 2Zd4x√−gc1∇µuν∇µuν+c2(∇µuµ)2+c3∇µuν∇νuµ+c4uµuν∇µuα∇νuα, (305) with uµa unit timelike vector, uµuµ=−1. On the HO branch we may parameterize the physically relevant combinations by three couplings (αAE, β, λ), which we define as αAE ≡c1+c4, β ≡c1+c3, λ ≡c2.(306) All the observables that will appear below—PPN parameters, mode speeds and stability conditions— depend only on (αAE, β, λ)and not on ciindividually. In the main text we have directly used the (αAE, β, λ)parameterization; the relations (306) provide the map to the standard Einstein– aether coefficients when needed. 50 PPN parameters α1and α2 We now consider the weak-field, slow-motion expansion around Minkowski space. The metric and aether are expanded as gµν =ηµν +hµν, uµ= ¯uµ+δuµ,¯uµ= (1,0),(307) and we solve the linearized field equations generated by the full action Sgrav +Sae, coupled to a non-relativistic matter source. Matching to the standard PPN metric (308) [38,39], g00 =−1+2U−2βPPNU2+. . . , g0i=−1 2(4γPPN +3+α1−α2)Vi+. . . , (308) one finds that γPPN =βPPN = 1 on the HO branch, while the preferred-frame parameters α1 and α2are non-trivial functions of (αAE, β, λ). To linear order in the small couplings |αAE|,|β|,|λ|≪1one may write α1=A1αAE +B1β+O(α2 AE, β2, αAEβ, λ),(309) α2=A2αAE +B2β+C2λ+O(α2 AE, β2, λ2, αAEβ, . . .),(310) where Ai, Bi, Ciare O(1) numerical coefficients determined by the linearized field equations.3 The key observation is that both α1and α2vanish when αAE = 0 and β= 0 [42]. Equivalently, αAE = 0, β = 0,(311) and this locus is independent of the value of λ. In other words, there exists a GR-safe locus in the (αAE, β, λ)space on which α1= 0, α2= 0, γPPN = 1, βPPN = 1,(312) so that all standard solar-system PPN tests are passed at the same level as in general relativity. This is the locus emphasized in the main text and used in the phenomenological analysis. Tensor speed and the GR-safe locus The propagation speed of the transverse-traceless tensor mode, cT, follows from the quadratic action for hTT ij around Minkowski space. On the HO branch, the kinetic term for the spin-2 mode depends only on the combination β:4 S(2) tensor =M2 ∗ 8Zd4xh˙ hTT ij ˙ hTT ij −c2 T∂khTT ij ∂khTT ij i,(313) with c2 T=1 1−β.(314) Requiring subluminal or luminal propagation and the absence of gradient instabilities implies 0< c2 T≤1 =⇒0≤β < 1.(315) In particular, on the GR-safe locus (311) one has β= 0 =⇒c2 T= 1,(316) so that gravitational waves propagate at exactly the speed of light. This is the branch we adopt in the main text when matching to cosmological and multi-messenger constraints. 3Their explicit form is not needed for our discussion of the GR-safe locus; we fix them in our numerical analysis by matching to the standard PPN results in the Einstein–aether basis. 4Equivalently, in the Einstein–aether basis the spin-2 speed depends only on c13 ≡c1+c3, and our βhas been defined precisely to coincide with this combination. 51 Scalar (khronon) sector and stability Besides the tensor mode, the HO aether sector contains a single scalar degree of freedom— the “khronon”—which can be parameterized by the perturbation of the khronon field Θor, equivalently, by the longitudinal perturbation of uµ. In an FRW background, and in a convenient gauge where the scalar metric perturbation is eliminated in favor of Θ, the quadratic action for the khronon takes the form S(2) scal =M2 ∗ 2Zd4x a3As(αAE, β, λ)˙ ζ2−Bs(αAE, β, λ)(∇ζ)2 a2+...,(317) where ζdenotes the canonical khronon perturbation and As, Bsare background-dependent coefficients. The scalar sound speed is c2 s=Bs As .(318) The ghost-free and gradient-stability conditions are As(αAE, β, λ)>0, Bs(αAE, β, λ)>0, c2 s>0,(319) and one may additionally impose subluminality c2 s≤1if desired. Near the GR-safe locus, |αAE|,|β|≪1, one finds that Asand Bsare smooth functions of (αAE, β, λ)and there exists an open region in parameter space for which (319) holds. In particular, on the αAE =β= 0 slice, the scalar mode is well behaved for a finite interval of λ; we restrict to this interval in the cosmological analysis of Section 2.3. Summary To summarize, the HO, projectable aether theory admits a three-parameter family (αAE, β, λ) of deformations of GR. On the GR-safe locus αAE = 0, β = 0,(320) one has 1. α1=α2= 0 and γPPN =βPPN = 1, so all standard PPN bounds are automatically satisfied; 2. c2 T= 1, so gravitational waves propagate luminally; 3. the scalar (khronon) mode is free of ghosts and gradient instabilities for a non-empty interval of λ. This justifies using the GR-safe branch as the background for the geometric sector selection mechanism and for the phenomenology discussed in the main text. D Quantization, constraint algebra, and renormalization In this appendix we collect the technical ingredients underlying the canonical quantization and one-loop analysis discussed in Section 4. We first spell out the structure of the physical Hamiltonian and the associated first-class constraints, then turn to the quantum constraint algebra and the absence of anomalies, and finally summarize the one-loop renormalization of the couplings (M2 ∗, λ, αAE, β)together with the non-running of the topological levels. 52 Canonical Hamiltonian and first-class constraints We work on a fixed foliation M4≃R×Σwith projectable lapse N(t)and shift Ni(t, x), and use the canonical variables introduced in Section 4.2. For the geometric sector we take as phase-space coordinates XM(x),ΠM(x),ΩAB i(x),Πi AB(x),Θ(x), PΘ(x),(321) where XMdescribes the embedding, ΩAB ithe normal connection (A, B normal-bundle indices), and Θis the khronon field. Matter fields ΦAand fermions ψcome with their standard canonical momenta, which we collectively denote by (Φ,ΠΦ)and (ψ, Πψ). The total Hamiltonian takes the standard Dirac form Htot[N, Ni,Λ, ρ] = Hphys[t] + ZΣ d3xNiCi+ ΛAB GAB +ρCproj,(322) where Hphys[t]=N(t)ZΣ d3xC(x),(323) Ci(x)=Cgeom i+Cgauge i+Cmatter i,(324) GAB(x)=G(6) AB(x)⊕G(10) AB (x),(325) Cproj(t) = ZΣ d3x√γuµuµ+ 1.(326) The Lagrange multipliers ΛAB(t, x)and ρ(t)enforce the normal-frame Gauss law and projectability constraints, respectively. Here γij is the induced spatial metric, and uµis the aether vector derived from Θas in Section 2.3. The Hamiltonian density Cdecomposes into C=Cgrav +Cae +CYM +Cmatter,(327) where: •Cgrav is the usual ADM gravitational Hamiltonian (written either in terms of the embedding variables or in terms of (γij, πij)) with coefficient M2 ∗; •Cae is the contribution of the aether / khronon sector with couplings (αAE, β, λ), specialized to the HO, projectable branch; •CYM is the Yang–Mills energy density for the SO(6)×SO(10) normal gauge fields defined via F= Π R⊥(Ω) Π; •Cmatter collects scalar and fermion contributions, including the normal fluctuations ΦAand fermions in the 16 representation. By construction, all constraints in (322) are first class: Cigenerates spatial diffeomorphisms on Σ,GAB generates local SO(6)×SO(10) rotations in the normal bundle, and Cproj enforces projectability of the lapse and the unit norm of uµ. Constraint algebra and absence of anomalies At the classical level, the constraints obey a closed algebra with structure functions. Using the canonical Poisson brackets and standard smearing notation C[N]≡ZΣ d3x NC,C[ N]≡ZΣ d3x NiCi,G[Λ] ≡ZΣ d3xΛABGAB,(328) 53 one finds the usual Dirac-type algebra {C[ N],C[ M]}=CL N M,(329) {C[N],C[ M]}=CL MN,(330) {C[N],C[M]}=Ciγij(N∂jM−M∂jN),(331) {G[Λ],G[Λ′]}=G[ [Λ,Λ′] ],(332) {C[N],G[Λ]}= 0,{C[ N],G[Λ]}=G[L NΛ],(333) together with {Cproj,·} generating redefinitions of N(t)along the projectable branch. The aether couplings (αAE, β, λ)modify the explicit form of Cbut do not change the structure of the algebra (331) on the HO, projectable sector. In the quantum theory we promote the constraints to operators b C,b Ci,b GAB acting on wavefunctionals Ψ[X, Ω,Θ,Φ, ψ;t], and impose the Dirac conditions b CΨ=0,b CiΨ=0,b GAB Ψ = 0,b Cproj Ψ = 0.(334) Equivalently, we construct a BRST charge b QBRST encoding the full first-class algebra and define the physical Hilbert space by the cohomology of b QBRST. A potential quantum anomaly in the constraint algebra would manifest itself as a Schwinger term in the equal-time commutators, schematically [b C[N],b C[M]]=iℏb Ci[. . .] + iℏ2∆anom[N, M],(335) with ∆anom = 0 a c-number functional. Using the background field method with a diffeomorphismand gauge-invariant regulator (dim. regularization with covariant gauge fixing), we compute the one-loop effective action Γ[g, u, Ω, . . .]and verify that it satisfies the Ward / Slavnov–Taylor identities corresponding to the classical symmetries. In BRST language this is equivalent to the statement b Q2 BRST = 0 to one-loop order,(336) so that no local counterterm is required to restore the constraint algebra and no Schwinger terms appear. The gauge and diffeomorphism invariance of the matter content (in particular the anomaly-free SO(10) fermion sector) ensures the absence of gauge and gravitational anomalies, and the HO, projectable restriction excludes the additional vector-type anomalies that might arise in a generic Einstein–aether background. One-loop renormalization and protected locus The one-loop effective action can be written as Γ[g, u, Ω,...;µ]=Scl[g, u, Ω,...;µ] + Γ(1) local[g, u, Ω,...;µ] + Γ(1) nonlocal[g, u, Ω, . . .],(337) where Scl is the classical action of Section 4.1,Γ(1) local collects the local counterterms needed to remove UV divergences, and Γ(1) nonlocal is finite and encodes the usual nonlocal logarithms. On the HO, projectable branch, and for energies below the intrinsic cutoff of the embedding theory, the most general local functional induced at one loop has the schematic form Γ(1) local =Zd4x√−gnδM2 ∗(µ)R+δλ(µ) (∇µuµ)2+δαAE(µ)OαAE +δβ(µ)Oβ +δαGB(µ)E4+δθ(µ)R˜ R+δk10(µ)P10 +δk6(µ)P6+...o, (338) where: 54 •OαAE and Oβdenote the two-derivative Einstein–aether operators corresponding to αAE and β, cf. Appendix C; •E4is the Gauss–Bonnet density, R˜ Rthe gravitational Pontryagin density, and P10,P6the 4D Pontryagin densities for the SO(10) and SO(6) gauge fields, respectively; •the ellipsis stands for higher-derivative operators suppressed by the cutoff. This R˜ Rblock coincides with the parity-violating Chern–Simons modification of gravity [17,18]. We define the renormalized couplings M2 ∗(µ)and λ(µ)by absorbing the logarithmic divergences into M2 ∗(µ)=M2 ∗+δM2 ∗(µ), λ(µ)=λ+δλ(µ),(339) and similarly for (αAE, β, αGB, θ, k10, k6). The corresponding β-functions are βM2 ∗(µ)≡µdM2 ∗ dµ, βλ(µ)≡µdλ dµ, . . . (340) and are determined by the coefficients of the logarithmic divergences in Γ(1) local. A straightforward background field computation on the HO, projectable branch, within the same two-derivative truncation and symmetry-preserving regulator as in Section 4.4, shows that: 1. M2 ∗and λdo run at one loop, with βM2 ∗=1 16π2fM(M2 ∗, λ, matter), βλ=1 16π2fλ(λ, matter),(341) where fMand fλare polynomials in the couplings and depend on the field content. Their explicit form is unimportant for the sector selection mechanism, as they can be absorbed into the running definitions of M2 ∗(µ)and λ(µ). 2. On the GR-safe locus within this truncation and regulator choice αAE = 0, β = 0,(342) the corresponding β-functions vanish at one loop: βαAE αAE=β=0 = 0, ββαAE=β=0 = 0.(343) Equivalently, αAE =β= 0 is a fixed locus of the one-loop RG flow on the HO, projectable branch within this sector. This is a consequence of the fact that αAE and βmultiply operators that can be removed by field redefinitions and integration by parts on this branch, combined with the symmetry-preserving regularization: within this setup no logarithmic divergence proportional to OαAE or Oβis generated. 3. The topological levels (αGB, θ, k10, k6)do not run: βαGB = 0, βθ= 0, βk10 = 0, βk6= 0 (344) at one loop. This follows from the fact that the corresponding densities (E4, R ˜ R, P10,P6) are total derivatives in four dimensions and the associated couplings are quantized when matched to higher-dimensional Chern–Simons inflow (see Section 6.3). A continuous renormalization of these levels would be incompatible with their integer quantization; perturbation theory thus preserves them. 55 Combining these results, we see that the GR-safe, topologically calibrated locus used in the main text, αAE(µ)=0, β(µ) = 0, αGB(µ)=αGB, k10(µ)=k10, k6(µ)=k6,(345) appears stable under one-loop renormalization within this two-derivative, projectable HO truncation and regulator choice. In particular, the luminal tensor speed cT= 1 and the PPN parameters α1=α2= 0 remain protected on this consistency surface, while M2 ∗(µ)and λ(µ) undergo the standard EFT running. This justifies treating the sector selection mechanism and the balanced-slice coupling locking as boundary conditions at the geometric scale ν, with ordinary RG flow away from the slice as discussed in Section 4.4. E Euclidean bounces and thin-wall junction conditions In this appendix we collect the technical details underlying the Euclidean bounce analysis used in the main text. We first discuss the generalized Gibbons–Hawking–York boundary term for a field-dependent effective Planck mass M2 ∗(Φ), following the standard construction reviewed in Section 2and in Refs. [55,56], then derive the O(4)-symmetric minisuperspace action in Euclidean signature. We next obtain the thin-wall junction conditions in the presence of M2 ∗(Φ), and finally analyze the fluctuation spectrum around Hawking–Moss configurations, including the one-negative-mode criterion. Generalized Gibbons–Hawking–York term with M2 ∗(Φ) The Euclidean bulk action for the scalar–tensor sector is taken to be Sbulk E[g, Φ] = ZM d4x√g+1 2M2 ∗(Φ) R+1 2(∇Φ)2+V(Φ),(346) where Ris the Euclidean Ricci scalar and V(Φ) the scalar potential. The effective Planck mass M2 ∗(Φ) encodes the non-minimal coupling structure (e.g. M2 ∗(Φ) = M2 0+ζ2Φ2in the examples discussed in the main text). Varying (346) with respect to gµν produces boundary terms from the variation of R, δZ√g M2 ∗(Φ)R⊃Z∂M d3x√h M2 ∗(Φ)Kij −Khijδhij +Z∂M d3x√h nµ∇µM2 ∗(Φ) δΦ,(347) where hij is the induced metric on ∂M,Kij its extrinsic curvature with trace K, and nµ the outward unit normal. The first term is cancelled by a generalized Gibbons–Hawking–York boundary term [55,56] SGHY E=−Z∂M d3x√h M2 ∗(Φ) K. (348) If we impose Dirichlet boundary conditions for Φon ∂M, the second term in (347) vanishes and (348) is sufficient to render the variational problem well-posed. More general mixed boundary conditions for Φwould require an additional Φ-dependent boundary counterterm; in what follows we work with fixed Φat the boundary, so that (348) is the only gravitational boundary term needed. O(4)-symmetric minisuperspace For the bounce geometries of interest we adopt the standard O(4)-symmetric ansatz ds2 E= dξ2+a(ξ)2dΩ2 3,Φ = Φ(ξ),(349) 56 with ξthe radial coordinate on the four-sphere, a(ξ)the scale factor, and dΩ2 3the line element of the unit three-sphere. The Ricci scalar for (349) is R= 6a′′ a+a′2 a2−1 a2,(350) where prime denotes d/dξ. Substituting (349) into (346) and adding the GHY term (348), one finds after integrating by parts that all second derivatives a′′ cancel between bulk and boundary. The Euclidean minisuperspace action reduces to SE[a, Φ] = 2π2Zdξa31 2Φ′2+V(Φ)−3a M2 ∗(Φ) a′2+ 1−3a2a′∂ΦM2 ∗(Φ) Φ′,(351) where 2π2is the volume of the unit three-sphere. The last term stems from the Φ-dependence of M2 ∗and would be absent for a constant Planck mass. The equations of motion follow from varying (351) with respect to a(ξ)and Φ(ξ). Variation with respect to agives the Euclidean Friedmann (Hamiltonian) constraint, 3M2 ∗(Φ)a′2−1+ 3aa′∂ΦM2 ∗(Φ) Φ′=−1 2a2Φ′2−a2V(Φ),(352) while variation with respect to Φyields Φ′′ + 3a′ aΦ′=V′(Φ) −3∂ΦM2 ∗(Φ)a′′ a+a′2 a2−1 a2,(353) where V′(Φ) ≡dV/dΦ. In the limit of constant M2 ∗the last term in (353) drops out and one recovers the familiar system for scalar-field vacuum decay in Einstein gravity. Regular O(4)-symmetric configurations satisfy a(0) = 0,a′(0) = 1 and Φ′(0) = 0 at the “north pole” of the instanton, together with analogous conditions at the “south pole” if the geometry closes off smoothly. The Coleman–De Luccia (CDL) bounces are non-trivial solutions of (352)–(353) interpolating between field-space neighborhoods of a metastable (false) vacuum and a deeper (true) vacuum. Hawking–Moss (HM) solutions correspond to Φ(ξ)sitting at a local maximum Φtop of the potential, with a(ξ)describing a four-sphere of radius H−1determined below. Thin-wall junction conditions with field-dependent M2 ∗(Φ) In the thin-wall regime, the Euclidean manifold can be decomposed into regions M±separated by a codimension-one wall Σon which the scalar field transitions rapidly between the false and true vacua. The total action can be written as SE=S(+) E+S(−) E+Swall E,(354) where S(±) Eare bulk contributions of the form (346)–(348) evaluated on M±, and Swall Eencodes the localized wall degrees of freedom. For a pure-tension wall we take Swall E=ZΣ d3x√h σ(Φ),(355) with σ(Φ) the (possibly field-dependent) wall tension and hij the induced metric on Σ. The variation of SEwith respect to gµν yields the generalized Israel junction condition. Denoting the jump of a quantity Xacross the wall by [X]+ −≡X+−X−, one finds hM2 ∗(Φ)Kij −Khiji+ −=−1 2Sij, Sij =−σ(Φ) hij,(356) 57