Skip to content
Shadow Theory

Geometric Realization of Completed Source Relations Descent, Orbit Spaces, Invariant Relations, and Variational Response in Shadow Theory

The geometric source domain

Jeremy RodgersSection 3 of 12Original PDF ↗Publication DOI ↗

The geometric source domain

Configurations

Convention 2.1 (Smoothness, signature, and index conventions). Throughout the general theory, MM is a smooth, connected, paracompact, oriented four-manifold, possibly with boundary ∂M\partial M. The choice dim⁡M=4\dim M=4 is made for definiteness in the physical application and is used nowhere essentially: Sections 2–5 hold verbatim in any dimension. Section 6 is a deliberate one-dimensional specialization, declared as such in Convention 6.1, and the dimensional stipulation of the present convention is suspended there. All fields are smooth unless a weaker regularity class is explicitly declared as part of the admissibility data. A Lorentzian metric has signature (−,+,+,+)(-,+,+,+); a time-oriented Lorentzian metric is a pair (g,t)(g,\mathfrak{t}) consisting of a Lorentzian metric gg together with a choice t\mathfrak{t} of one of the two continuous global orientations of the timelike cone bundle of gg (equivalently, the choice of a connected component of the set of gg-timelike vector fields, when one exists). Not every Lorentzian metric admits such a choice; admitting one is a nontrivial global condition on gg, and it is imposed below as part of admissibility rather than derived (see Remark 2.3). Greek indices run over 0,1,2,30,1,2,3 and are raised and lowered with gg. GG denotes a finite-dimensional Lie group with Lie algebra g\mathfrak{g}, equipped with a fixed Ad⁡\operatorname{Ad}-invariant nondegenerate bilinear form ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle on g\mathfrak{g} (for compact semisimple GG one may take the negative Killing form; for G=U(1)G=U(1) the standard normalization on u(1)≅iR\mathfrak{u}(1)\cong i\mathbb{R}). Finally ρ ⁣:G→GL(W)\rho\colon G\to \mathrm{GL}(W) is a fixed finite-dimensional representation; ρ\rho is not assumed unitary, and where unitarity is needed it is hypothesized explicitly.

Definition 2.2 (Global geometric configuration). A global geometric configuration is a tuple Ψ=(M,  g,  P,  A,  Φ,  B),(2) \Psi=\bigl(M,\;g,\;P,\;A,\;\Phi,\;\mathcal{B}\bigr),\tag{2} in which

  1. gg is a Lorentzian metric on MM;

  2. π ⁣:P→M\pi\colon P\to M is a principal GG-bundle;

  3. A∈Ω1(P,g)A\in\Omega^1(P,\mathfrak{g}) is a principal connection on PP [6, 3];

  4. Φ\Phi is a section of the associated bundle P×ρW→MP\times_\rho W\to M;

  5. B\mathcal{B} denotes the boundary, asymptotic, causal, and global data required to make the model well posed. Among these data we always include a choice of time orientation t\mathfrak{t} for gg; the remaining components of B\mathcal{B} are model-specific and may comprise, for instance, fixed induced data on ∂M\partial M, a decay class at infinity, a fixed conformal boundary structure, a global hyperbolicity requirement, a spin structure, or a fixed topological sector.

Remark 3 (Where time orientation lives). Time orientation is a property of the pair (M,g)(M,g), not of the bare smooth manifold MM: it is meaningless to speak of MM as “time-orientable” independently of a Lorentzian metric, since the timelike cone bundle whose components are being selected is defined by gg. Accordingly the choice t\mathfrak{t} is carried in B\mathcal{B}, and the arrows of the redundancy groupoid are required below to preserve it. This has a consequence for the descent theorem, recorded explicitly in Theorem 3.4 and Remark 3.6: descent reconstructs the metric, but membership of the reconstructed (g,t)(g,\mathfrak{t}) in a time-oriented admissibility class is an additional global condition on the glued object, which either must be imposed after the fact or must be built into the descent data as compatible local time orientations.

Definition 2.4 (Admissibility class). An admissibility class Cgeo\mathcal{C}_{\mathrm{geo}} is a specified class of tuples (2) satisfying declared conditions on regularity, signature, causal structure, boundary and asymptotic behaviour, and topological sector, and in which every gg is time-orientable with a distinguished time orientation t∈B\mathfrak{t}\in\mathcal{B}. The class Cgeo\mathcal{C}_{\mathrm{geo}} is part of the data of the model. All subsequent statements are relative to a fixed Cgeo\mathcal{C}_{\mathrm{geo}}.

The insistence that Cgeo\mathcal{C}_{\mathrm{geo}} be declared is not pedantry. Every separation, realization, and conservation statement below is sensitive to it: a relation family may separate orbits in one admissibility class and fail to separate them in a larger one, and a Ward identity may hold with compactly supported variations and fail without a boundary term.

The redundancy groupoid

Gauge and coordinate freedom in a bundle-theoretic setting is not, in general, a group acting on a fixed configuration space: different admissible configurations may carry non-isomorphic bundles, and the diffeomorphisms that lift to a given bundle need not lift to another. The correct structure is a groupoid.

Definition 2.5 (Ambient allowed automorphisms). Fix a principal GG-bundle P→MP\to M. Let Diff⁡B(M)\operatorname{Diff}_{\mathcal{B}}(M) denote the group of diffeomorphisms of MM preserving the orientation of MM and the non-metric structural data in B\mathcal{B} (boundary, asymptotic regime, topological sector), and let Aut⁡B(P):={F^∈Aut⁡(P)  :  F^ covers some f∈Diff⁡B(M)}(3) \operatorname{Aut}_{\mathcal{B}}(P) :=\bigl\{\hat F\in\operatorname{Aut}(P)\;:\;\hat F \text{ covers some } f\in\operatorname{Diff}_{\mathcal{B}}(M)\bigr\}\tag{3} be the group of GG-equivariant diffeomorphisms of PP covering such diffeomorphisms. Define Diff⁡P,B(M)  :=  image of  Aut⁡B(P)  in  Diff⁡B(M),(4) \operatorname{Diff}_{P,\mathcal{B}}(M)\;:=\;\text{image of }\ \operatorname{Aut}_{\mathcal{B}}(P)\ \text{ in }\ \operatorname{Diff}_{\mathcal{B}}(M),\tag{4} that is, the subgroup of allowed base diffeomorphisms that admit an equivariant lift to PP.

Proposition 2.6 (Structure of the ambient group). With the definitions (3)–(4) there is a short exact sequence of groups 1  ⟶  Gau⁡(P)  ⟶  Aut⁡B(P)  ⟶  Diff⁡P,B(M)  ⟶  1,(5) 1\;\longrightarrow\;\operatorname{Gau}(P)\;\longrightarrow\;\operatorname{Aut}_{\mathcal{B}}(P)\;\longrightarrow\; \operatorname{Diff}_{P,\mathcal{B}}(M)\;\longrightarrow\;1 ,\tag{5} where Gau⁡(P)  =  {F^∈Aut⁡(P):F^ covers idM}  ≅  Γ(P×conjG)\operatorname{Gau}(P)\;=\;\bigl\{\hat F\in\operatorname{Aut}(P):\hat F\ \text{covers}\ \mathrm{id}_M\bigr\} \;\cong\;\Gamma\bigl(P\times_{\mathrm{conj}}G\bigr) is the gauge group. The sequence is exact by construction: Diff⁡P,B(M)\operatorname{Diff}_{P,\mathcal{B}}(M) is defined as the image of the projection, and the kernel of that projection is exactly the set of automorphisms covering the identity. The sequence need not split.

Proof. Exactness at Gau⁡(P)\operatorname{Gau}(P) and at Diff⁡P,B(M)\operatorname{Diff}_{P,\mathcal{B}}(M) is immediate from the definitions of kernel and image. That the extension need not split is a statement about bundle topology: a splitting would furnish, for each allowed f∈Diff⁡P,B(M)f\in\operatorname{Diff}_{P,\mathcal{B}}(M), an equivariant lift σ(f)∈Aut⁡B(P)\sigma(f)\in\operatorname{Aut}_{\mathcal{B}}(P) depending on ff homomorphically, and there is no reason for such a coherent choice to exist. Consequently Aut⁡B(P)\operatorname{Aut}_{\mathcal{B}}(P) is not in general a semidirect product Gau⁡(P)⋊Diff⁡P,B(M)\operatorname{Gau}(P)\rtimes\operatorname{Diff}_{P,\mathcal{B}}(M), and nothing below assumes it is. ◻

Remark 7 (The ambient group is not the isotropy group).

Proposition 2.6 describes the ambient group of allowed bundle automorphisms over a fixed PP. It does not describe the isotropy group of the groupoid at a configuration. The isotropy (stabilizer) group at Ψ=(M,g,P,A,Φ,B)\Psi=(M,g,P,A,\Phi,\mathcal{B}) is the subgroup Aut⁡(Ψ):={F^∈Aut⁡B(P) covering some f :  f∗g=g,f∗t=t,(F^−1)∗A=A,F^ρ∘Φ=Φ∘f,f∗B=B}  ≤  Aut⁡B(P),(6) \begin{aligned} \operatorname{Aut}(\Psi) :=\bigl\{\hat F\in\operatorname{Aut}_{\mathcal{B}}(P)\ \text{covering some}\ f\ :\ \ &f_*g=g,\quad f_*\mathfrak{t}=\mathfrak{t},\quad (\hat F^{-1})^*A=A,\\[-0.15em] &\hat F_\rho\circ\Phi=\Phi\circ f,\quad f_*\mathcal{B}=\mathcal{B} \bigr\}\;\le\;\operatorname{Aut}_{\mathcal{B}}(P), \end{aligned}\tag{6} that is, the stabilizer of the whole tuple: it preserves the metric, the chosen time orientation and orientation, the connection, the matter section, and the boundary data. Aut⁡(Ψ)\operatorname{Aut}(\Psi) is typically far smaller than Aut⁡B(P)\operatorname{Aut}_{\mathcal{B}}(P): for a generic (g,A,Φ)(g,A,\Phi) it is a small or even trivial group, its base part being contained in the isometry group of gg. In particular the exact sequence (5) is a statement about Aut⁡B(P)\operatorname{Aut}_{\mathcal{B}}(P) and not about Aut⁡(Ψ)\operatorname{Aut}(\Psi).

Definition 2.8 (Redundancy groupoid).

Let Ggeo\mathfrak{G}_{\mathrm{geo}} be the groupoid whose objects are the elements of Cgeo\mathcal{C}_{\mathrm{geo}}. Given two objects Ψ=(M,g,P,A,Φ,B),Ψ′=(M,g′,P′,A′,Φ′,B′),\Psi=(M,g,P,A,\Phi,\mathcal{B}), \qquad \Psi'=(M,g',P',A',\Phi',\mathcal{B}'), the arrows F ⁣:Ψ→Ψ′F\colon\Psi\to\Psi' are the pairs F=(F^,f)F=(\hat F,f) consisting of

  • an allowed diffeomorphism f∈Diff⁡B(M)f\in\operatorname{Diff}_{\mathcal{B}}(M), and

  • a GG-equivariant bundle isomorphism F^ ⁣:P→P′\hat F\colon P\to P' covering ff,

subject to f∗g=g′,f∗t=t′,(F^−1)∗A=A′,F^ρ∘Φ=Φ′∘f,f∗B=B′,(7) f_*g=g', \qquad f_*\mathfrak{t}=\mathfrak{t}', \qquad (\hat F^{-1})^{*}A = A', \qquad \hat F_{\rho}\circ\Phi = \Phi'\circ f, \qquad f_*\mathcal{B}=\mathcal{B}',\tag{7} where F^ρ\hat F_\rho is the induced isomorphism P×ρW→P′×ρWP\times_\rho W\to P'\times_\rho W. The second condition is the requirement that arrows preserve the chosen time orientation; it is not implied by the first, since f∗g=g′f_*g=g' alone permits ff to exchange the two components of the timelike cone bundle. Composition is composition of bundle maps; identities are (idP,idM)(\mathrm{id}_P,\mathrm{id}_M). The isotropy group of Ggeo\mathfrak{G}_{\mathrm{geo}} at Ψ\Psi is Aut⁡(Ψ)\operatorname{Aut}(\Psi) of (6).

The groupoid formulation, rather than a single group acting on a single configuration space, is forced by two features. First, the arrows relate objects carrying different bundle representatives, which is exactly what is needed when descent produces a bundle only up to isomorphism (Theorem 3.4). Second, isotropy jumps from configuration to configuration, since Aut⁡(Ψ)\operatorname{Aut}(\Psi) depends on Ψ\Psi; there is no ambient group acting with constant stabilizer.

Definition 2.9 (Physical equivalence and the reduced source).

Write Ψ≃geoΨ′⟺∃ F ⁣:Ψ→Ψ′ in Ggeo.(8) \Psi\simeq_{\mathrm{geo}}\Psi' \quad\Longleftrightarrow\quad \exists\,F\colon\Psi\to\Psi' \text{ in } \mathfrak{G}_{\mathrm{geo}}.\tag{8} This is an equivalence relation on Cgeo\mathcal{C}_{\mathrm{geo}} because Ggeo\mathfrak{G}_{\mathrm{geo}} is a groupoid. The reduced geometric source is the orbit space Sgeo  :=  Cgeo/ ⁣≃geo,(9) S_{\mathrm{geo}}\;:=\; \mathcal{C}_{\mathrm{geo}}/\!\simeq_{\mathrm{geo}} ,\tag{9} with quotient map q ⁣:Cgeo→Sgeoq\colon\mathcal{C}_{\mathrm{geo}}\to S_{\mathrm{geo}}, q(Ψ)=[Ψ]q(\Psi)=[\Psi].

Remark 10 (Set-theoretic hygiene).

SgeoS_{\mathrm{geo}} is used below only as a set (equivalently, as the set of isomorphism classes of objects of the essentially small groupoid Ggeo\mathfrak{G}_{\mathrm{geo}} obtained after fixing a set of bundle representatives). No smooth or stratified structure on SgeoS_{\mathrm{geo}} is used or assumed; the orbit space of a gauge groupoid is in general not a manifold, precisely because isotropy jumps, as Remark 2.7 notes, and the arguments of Sections 5–6 do not require it to be. Where a topology is needed, as for the moduli space Mflat(S1,U(1))\mathcal{M}_{\mathrm{flat}}(S^1,U(1)) in Section 6, it is supplied and identified explicitly.

Convention 2.11 (Status of relation coordinates).

A relation coordinate rαr_\alpha is a function on SgeoS_{\mathrm{geo}}, not an independent field on MM. It is not varied independently in Section 7 unless a model explicitly promotes it to a field by adjoining it to the configuration tuple (2) and enlarging Cgeo\mathcal{C}_{\mathrm{geo}} accordingly. Where rαr_\alpha is a functional of (g,A,Φ)(g,A,\Phi) its variation is computed by the functional chain rule of Section 7.2; where it labels a fixed topological or boundary sector it is held fixed under variations within that sector. Both cases are handled in Section 7; they are not conflated.

Every realization relation used in this paper is, by construction, a function on SgeoS_{\mathrm{geo}} and never a function of an arbitrary coordinate chart or gauge potential.

Back to top ↑