Introduction
Classical Yang-Mills (YM) theory is formulated on a principal G G G -bundle π : P → M \pi:P\to M π : P → M over an orientable n n n -dimensional semi-Riemannian manifold ( M , g ) (M,g) ( M , g ) , where G G G is a compact semi-simple Lie group with Lie algebra g \mathfrak{g} g [20 , 12 ] . Given a connection 1-form D ∈ Conn ( P ; g ) ⊂ Ω 1 ( P ; g ) D\in\operatorname{Conn}(P;\mathfrak{g})\subset\Omega^1(P;\mathfrak{g}) D ∈ Conn ( P ; g ) ⊂ Ω 1 ( P ; g ) , its curvature (or field strength) is the horizontal, G G G -equivariant g \mathfrak{g} g -valued 2-form
F D = d D + 1 2 D [ ∧ ] g D ∈ Ω h e q 2 ( P ; g ) , F_D=dD+\frac{1}{2}D[\wedge]_{\mathfrak{g}}D\in\Omega^2_{\mathrm{heq}}(P;\mathfrak{g}), F D = d D + 2 1 D [ ∧ ] g D ∈ Ω heq 2 ( P ; g ) ,
where [ ∧ ] g [\wedge]_{\mathfrak{g}} [ ∧ ] g denotes the exterior product induced by the Lie bracket of g \mathfrak{g} g . Using the canonical Weil isomorphism j : Ω h e q 2 ( P ; g ) → ∼ Ω 2 ( M ; E g ) j:\Omega^2_{\mathrm{heq}}(P;\mathfrak{g})\xrightarrow{\sim}\Omega^2(M;E_{\mathfrak{g}}) j : Ω heq 2 ( P ; g ) ∼ Ω 2 ( M ; E g ) onto 2-forms with values in the adjoint bundle E g = P × A d g E_{\mathfrak{g}}=P\times_{\mathrm{Ad}}\mathfrak{g} E g = P × Ad g , the classical Yang-Mills action functional is given by
S Y M [ D ] = ∫ M ⟨ F D , F D ⟩ g d v o l g = ∫ M tr ( j ( F D ) ∧ ∗ g j ( F D ) ) , (1) S_{\mathrm{YM}}[D]=\int_M\langle F_D,F_D\rangle_{\mathfrak{g}}\,\mathrm{dvol}_{g}=\int_M\operatorname{tr}\bigl(j(F_D)\wedge*_g j(F_D)\bigr), \tag*{(1)} S YM [ D ] = ∫ M ⟨ F D , F D ⟩ g dvol g = ∫ M tr ( j ( F D ) ∧ ∗ g j ( F D ) ) , (1)
where ∗ g *_g ∗ g is the Hodge star operator of g \mathfrak{g} g and tr \operatorname{tr} tr represents the Killing form B g B_{\mathfrak{g}} B g of g \mathfrak{g} g .
While physically foundational, the classical action (1) relies on three structural assumptions:
The base manifold M M M must admit a semi-Riemannian metric g g g of prescribed signature (which is subject to topological obstructions, such as χ ( M ) = 0 \chi(M)=0 χ ( M ) = 0 for compact Lorentzian manifolds [16 , 5 ] );
The internal symmetry algebra must be a Lie algebra g \mathfrak{g} g admitting a non-degenerate Ad-invariant quadratic form (typically requiring G G G to be semi-simple or compact, or belong to specific classes of quadratic solvable Lie algebras [26 ] );
(3) The gauge potential D D D takes values in a Lie algebra rather than a more general (possibly non-associative or m-graded) algebra ( A , ∗ ) (A,\ast) ( A , ∗ ) modeling internal, worldsheet, supergeometric, Malcev, or reductive Cartan symmetries [3 , 37 , 28 , 9 ] .
In the context of first-order tetradic gravity, it was shown in [21 ] that the Einstein-Hilbert-Palatini (EHP) action functional can be detached from Lorentzian Cartan geometry and formulated algebraically over an arbitrary m-graded algebra ( A , ∗ ) (A,\ast) ( A , ∗ ) decomposed as A ≃ A 0 ⊕ A 1 A \simeq A_0 \oplus A_1 A ≃ A 0 ⊕ A 1 . Doing so revealed that the non-triviality of the A A A -valued EHP functional is governed by the graded polynomial identities of ( A , ∗ ) (A,\ast) ( A , ∗ ) : specifically, if A 0 A_0 A 0 is weak ( k , s ) (k,s) ( k , s ) -solvable, the cosmological constant decouples for n ≥ k + s + 1 n \geq k+s+1 n ≥ k + s + 1 and the entire A A A -EHP functional vanishes identically for n ≥ k + s + 3 n \geq k+s+3 n ≥ k + s + 3 [21 ] .
The purpose of the present paper is to formulate algebra-valued Yang-Mills-type (A-YMT) theories directly in a concrete differential-geometric and algebraic setting (without requiring category-theoretic functorial bundle systems), establish their variational and gauge-theoretic foundations, and determine the algebraic, parenthesization, and holonomy obstructions that govern them. Given a principal G G G -bundle π : P → M \pi:P \to M π : P → M and a representation ρ : G → Aut A l g ( A ) \rho:G \to\operatorname{Aut}_{\mathrm{Alg}}(A) ρ : G → Aut Alg ( A ) of G G G by algebra automorphisms of an m-graded algebra ( A , ∗ ) (A,\ast) ( A , ∗ ) , horizontal ρ \rho ρ -equivariant A A A -valued differential forms Ω h e q , ρ ∙ ( P ; A ) \Omega^{\bullet}_{\mathrm{heq},\rho}(P;A) Ω heq , ρ ∙ ( P ; A ) descend via the canonical Weil algebra isomorphism j : Ω h e q , ρ ∙ ( P ; A ) ⟶ ∼ Ω ∙ ( M ; E A ) j:\Omega^{\bullet}_{\mathrm{heq},\rho}(P;A) \overset{\sim}{\longrightarrow} \Omega^\bullet(M;E_A) j : Ω heq , ρ ∙ ( P ; A ) ⟶ ∼ Ω ∙ ( M ; E A ) to forms valued in the associated algebra bundle E A = P × ρ A → M E_A=P\times_\rho A \to M E A = P × ρ A → M . We simultaneously decouple the quadratic curvature pairing from both the metric Hodge star ∗ g \ast_g ∗ g and the Killing form B g B_g B g by allowing a general R \mathbb{R} R -bilinear or C ∞ ( M ) C^\infty(M) C ∞ ( M ) -bilinear pairing ⟨ ⋅ , ⋅ ⟩ A \langle\cdot,\cdot\rangle_A ⟨ ⋅ , ⋅ ⟩ A on Ω 2 ( M ; E A ) \Omega^2(M;E_A) Ω 2 ( M ; E A ) .
A fundamental structural contrast immediately emerges between A-EHP gravity and A-YMT gauge theory:
In A-EHP theory, the Hilbert-Palatini n n n -form α n , Λ = ∧ ∗ n − 2 e ∧ ∗ Ω + Λ ( n − 1 ) ! ∧ ∗ n e \alpha_{n,\Lambda}=\wedge^{n-2}_\ast e\wedge_\ast\Omega+\frac{\Lambda}{(n-1)!}\wedge^n_\ast e α n , Λ = ∧ ∗ n − 2 e ∧ ∗ Ω + ( n − 1 )! Λ ∧ ∗ n e contains a number of factors of e e e that grows linearly with dim M = n \dim M=n dim M = n , requiring either quotienting by associativity-like polynomial ideals or fixing a parenthesization scheme whenever ( A , ∗ ) (A,\ast) ( A , ∗ ) is non-associative, and converting ( k , s ) (k,s) ( k , s ) -solvability directly into upper bounds on the spacetime dimension n n n [21 ] .
In pure A-YMT theory, the curvature 2-form F D = d D + 1 2 D ∧ ∗ D = d D + 1 4 D ∧ [ ⋅ , ⋅ ] ∗ D F_D=dD+\frac{1}{2}D\wedge_\ast D=dD+\frac{1}{4}D\wedge_{[\cdot,\cdot]_\ast}D F D = d D + 2 1 D ∧ ∗ D = d D + 4 1 D ∧ [ ⋅ , ⋅ ] ∗ D involves only a single binary product D ∧ ∗ D D\wedge_\ast D D ∧ ∗ D , which automatically projects onto the skew-symmetric commutator bracket [ x , y ] ∗ = x ∗ y − y ∗ x [x,y]_\ast=x\ast y-y\ast x [ x , y ] ∗ = x ∗ y − y ∗ x while annihilating the symmetric Jordan product x ∘ ∗ y = 1 2 ( x ∗ y + y ∗ x ) x\circ_\ast y=\frac{1}{2}(x\ast y+y\ast x) x ∘ ∗ y = 2 1 ( x ∗ y + y ∗ x ) . Consequently, F D F_D F D is unambiguously defined for any non-associative algebra ( A , ∗ ) (A,\ast) ( A , ∗ ) without choosing parenthesizations or quotienting by associativity ideals, and the action S A [ D ] = ∫ M ⟨ j ( F D ) , j ( F D ) ⟩ A S^A[D]=\int_M\langle j(F_D),j(F_D)\rangle_A S A [ D ] = ∫ M ⟨ j ( F D ) , j ( F D ) ⟩ A is universally quartic in D D D across all spacetime dimensions n > 2 n>2 n > 2 . However, whenever higher curvature powers ∧ ∗ r F D \wedge_\ast^rF_D ∧ ∗ r F D , constitutive soldering powers ∧ ∗ n − 4 e \wedge_\ast^{n-4}e ∧ ∗ n − 4 e , or differential identities such as the Bianchi identity d F D + 1 2 [ D , F D ] ∗ dF_D+\frac{1}{2}[D,F_D]_\ast d F D + 2 1 [ D , F D ] ∗ and Chern-Simons transgression are formed, the symmetric product ∘ ∗ \circ_\ast ∘ ∗ , the parenthesization scheme, the associator Assoc ∗ \operatorname{Assoc}_\ast Assoc ∗ , and the commutator Jacobiator Jac [ ⋅ , ⋅ ] ∗ \operatorname{Jac}_{[\cdot,\cdot]_\ast} Jac [ ⋅ , ⋅ ] ∗ play decisive roles.
Specifically, the main results of this paper are organized as follows:
Ehresmann Admissibility, the Lie Nucleus Constraint, Variational Field Equations, and the Jacobiator Bianchi/Gauge Obstructions (Proposition 2.4 and Theorem 2.7): On a principal G G G -bundle π : P → M \pi:P \to M π : P → M with automorphism representation ρ : G → Aut A l g ( A ) \rho:G \to\operatorname{Aut}_{\mathrm{Alg}}(A) ρ : G → Aut Alg ( A ) , we prove that an affine Ehresmann connection D ∈ Ω h e q , ρ 1 ( P ; A ) D\in\Omega^1_{\mathrm{heq},\rho}(P;A) D ∈ Ω heq , ρ 1 ( P ; A ) with vertical map D ( X ) = η ( X ) D(X)=\eta(X) D ( X ) = η ( X ) has horizontal curvature F D ∈ Ω h e q , ρ 2 ( P ; A ) F_D\in\Omega^2_{\mathrm{heq},\rho}(P;A) F D ∈ Ω heq , ρ 2 ( P ; A ) if and only if d ρ ( X ) ( y ) = 1 2 [ η ( X ) , y ] ∗ d\rho(X)(y)=\frac{1}{2}[\eta(X),y]_\ast d ρ ( X ) ( y ) = 2 1 [ η ( X ) , y ] ∗ , which in turn forces the vertical image η ( g ) \eta(\mathfrak{g}) η ( g ) to lie in the Lie nucleus N L i e ( A ) \mathcal{N}_{\mathrm{Lie}}(A) N Lie ( A ) of ( A , [ ⋅ , ⋅ ] ∗ ) (A,[\cdot,\cdot]_\ast) ( A , [ ⋅ , ⋅ ] ∗ ) . We then derive the Euler-Lagrange equations of a general A-YMT theory and prove that under an infinitesimal gauge transformation δ ϵ D = d ϵ + 1 2 D ∧ [ ⋅ , ⋅ ] ∗ ϵ \delta_\epsilon D=d\epsilon+\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}\epsilon δ ϵ D = d ϵ + 2 1 D ∧ [ ⋅ , ⋅ ] ∗ ϵ , the failure of homogeneous curvature covariance δ ϵ F D − 1 2 F D ∧ [ ⋅ , ⋅ ] ∗ ϵ \delta_\epsilon F_D-\frac{1}{2}F_D\wedge_{[\cdot,\cdot]_\ast}\epsilon δ ϵ F D − 2 1 F D ∧ [ ⋅ , ⋅ ] ∗ ϵ and the failure of the differential Bianchi identity d F D + 1 2 D ∧ [ ⋅ , ⋅ ] ∗ F D dF_D+\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}F_D d F D + 2 1 D ∧ [ ⋅ , ⋅ ] ∗ F D are governed purely by the horizontal Jacobiator 2-form − 1 8 Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) -\frac{1}{8}\operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,\epsilon) − 8 1 Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) and 3-form − 1 24 Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) -\frac{1}{24}\operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,D) − 24 1 Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) . In particular, principal G G G -gauge transformations ϵ ∈ Ω h e q , ρ 0 ( P ; η ( g ) ) ⊂ Ω h e q , ρ 0 ( P ; N L i e ( A ) ) \epsilon\in\Omega^0_{\mathrm{heq},\rho}(P;\eta(\mathfrak{g}))\subset\Omega^0_{\mathrm{heq},\rho}(P;\mathcal{N}_{\mathrm{Lie}}(A)) ϵ ∈ Ω heq , ρ 0 ( P ; η ( g )) ⊂ Ω heq , ρ 0 ( P ; N Lie ( A ))
are universally anomaly-free for any algebra ( A , ∗ ) (A,\ast) ( A , ∗ ) , whereas extended A A A -gauge transformations ϵ ∈ Ω h e q , ρ 0 ( P ; A ) \epsilon\in\Omega^{0}_{\mathrm{heq},\rho}(P;A) ϵ ∈ Ω heq , ρ 0 ( P ; A ) outside N L i e ( A ) \mathcal{N}_{\mathrm{Lie}}(A) N Lie ( A ) are obstructed by Jac [ ⋅ , ⋅ ] ∗ \operatorname{Jac}_{[\cdot,\cdot]_{\ast}} Jac [ ⋅ , ⋅ ] ∗ . Using a rigorous pointwise localization argument relative to the basepoint connection D 0 = η ∘ ω 0 D_{0}=\eta\circ\omega_{0} D 0 = η ∘ ω 0 , we also establish the exact converse when ⟨ ⋅ , ⋅ ⟩ A , s y m \langle\cdot,\cdot\rangle_{A,\mathrm{sym}} ⟨ ⋅ , ⋅ ⟩ A , sym is C ∞ ( M ) C^{\infty}(M) C ∞ ( M ) -bilinear (tensorial), pointwise ad [ ⋅ , ⋅ ] ∗ \operatorname{ad}_{[\cdot,\cdot]_{\ast}} ad [ ⋅ , ⋅ ] ∗ -invariant, and non-degenerate.
(2) Internal-Product Topological and Self-Dual A A A -YMT Obstructions (Theorem 3.2 and Proposition 3.3): After recording the commutator projection, abelianization of weak ( 1 , 1 ) (1,1) ( 1 , 1 ) -solvable algebras, and derived-filtration localization (Proposition 3.1 ), we study internal-trace topological and self-dual pairings constructed purely from the internal multiplication ∧ ∗ \wedge_{\ast} ∧ ∗ and a bundle linear functional tr ξ : E A → M × R \operatorname{tr}_{\xi}:E_{A}\to M\times\mathbb{R} tr ξ : E A → M × R (as opposed to an external symmetric bilinear invariant form κ ∈ inv 2 ( g ) \kappa\in\operatorname{inv}^{2}(\mathfrak{g}) κ ∈ inv 2 ( g ) ). We prove that weak ( 2 , s ) (2,s) ( 2 , s ) -solvability forces the r r r -th degree internal Chern-Weil action ∫ M tr ξ ( j ( ∧ ∗ r F D ) ) \int_{M}\operatorname{tr}_{\xi}(j(\wedge_{\ast}^{r}F_{D})) ∫ M tr ξ ( j ( ∧ ∗ r F D )) in dimension n = 2 r n=2r n = 2 r to vanish for n ≥ 2 ( s + 1 ) n\geq2(s+1) n ≥ 2 ( s + 1 ) , and we show that these obstructions are preserved under ( Q , S ) (Q,S) ( Q , S ) -splitting extensions. For s = 1 s=1 s = 1 , this internal-product vanishing encompasses all skew-commutative algebras (including Lie algebras under [ ∧ ] g [\wedge]_{\mathfrak{g}} [ ∧ ] g , Berger holonomy algebras, Hitchin’s generalized complex geometry, and Cayley-Dickson unitary algebras u ( k ; C D l ( A ) ) u(k;\mathrm{CD}^{l}(A)) u ( k ; CD l ( A )) under the skew-symmetric matrix product ∧ \wedge ∧ ).
(3) Parenthesization Schemes, 4D First-Bianchi Contractions, and Berger Holonomy Reductions in Cartan-Yang-Mills / MacDowell-Mansouri Theories (Lemma 2.1 , Lemma 4.3, Theorem 4.5 , and Proposition 4.6 ): For reductive connections ∇ = e + ω \nabla=e+\omega ∇ = e + ω valued in A = R k ⊕ h A=\mathbb{R}^{k}\oplus\mathfrak{h} A = R k ⊕ h with h ⊂ s o ( k 1 ) ⊕ ⋯ ⊕ s o ( k r ) \mathfrak{h}\subset\mathfrak{so}(k_{1})\oplus\cdots\oplus\mathfrak{so}(k_{r}) h ⊂ so ( k 1 ) ⊕ ⋯ ⊕ so ( k r ) , we formulate the internal product ∧ × \wedge_{\times} ∧ × via an H H H -equivariant soldering map σ : R k ∧ R k → h \sigma:\mathbb{R}^{k}\wedge\mathbb{R}^{k}\to\mathfrak{h} σ : R k ∧ R k → h (including the G 2 G_{2} G 2 -equivariant vector cross product on R 7 \mathbb{R}^{7} R 7 when n = k = 7 n=k=7 n = k = 7 ). We distinguish between two non-associative parenthesization schemes for exterior powers of e e e : (a) quartic-block parenthesization ∧ ∗ , b l k m e \wedge_{\ast,\mathrm{blk}}^{m}e ∧ ∗ , blk m e (which factors out ∧ ∗ , b a l 4 e = ( e ∧ ∗ e ) ∧ ∗ ( e ∧ ∗ e ) = 0 \wedge_{\ast,\mathrm{bal}}^{4}e=(e\wedge_{\ast}e)\wedge_{\ast}(e\wedge_{\ast}e)=0 ∧ ∗ , bal 4 e = ( e ∧ ∗ e ) ∧ ∗ ( e ∧ ∗ e ) = 0 and vanishes for all m ≥ 4 m\geq4 m ≥ 4 ), and (b) recursive binary-tree halving ∧ ∗ , b a l m e \wedge_{\ast,\mathrm{bal}}^{m}e ∧ ∗ , bal m e (which we prove vanishes algebraically for all m ≥ 4 m\geq4 m ≥ 4 outside the consecutive Jacobsthal pairs m ∈ ⋃ r ≥ 3 { 2 J r , J r + 1 } = { 5 , 6 } ∪ { 10 , 11 } ∪ { 21 , 22 } ∪ { 42 , 43 } ∪ ⋯ m\in\bigcup_{r\geq3}\{2J_{r},J_{r+1}\}=\{5,6\}\cup\{10,11\}\cup\{21,22\}\cup\{42,43\}\cup\cdots m ∈ ⋃ r ≥ 3 { 2 J r , J r + 1 } = { 5 , 6 } ∪ { 10 , 11 } ∪ { 21 , 22 } ∪ { 42 , 43 } ∪ ⋯ where J r = 2 r − ( − 1 ) r 3 J_{r}=\frac{2^{r}-(-1)^{r}}{3} J r = 3 2 r − ( − 1 ) r , and vanishes by coframe rank for all m > k m>k m > k , so for k ≤ 9 k\leq9 k ≤ 9 it vanishes for all m ≥ 7 m\geq7 m ≥ 7 ). Under quartic-block parenthesization, the cosmological term vanishes for n ≥ 4 n\geq4 n ≥ 4 , the EHP term vanishes for n ≥ 6 n\geq6 n ≥ 6 , and the quadratic Yang-Mills term survives up to $n=7 before vanishing for n ≥ 8 n\geq8 n ≥ 8 (while under recursive binary halving the YM term vanishes algebraically for all n ≥ 8 n\geq8 n ≥ 8 with n − 4 ∉ ⋃ r ≥ 3 { 2 J r , J r + 1 } n-4\notin\bigcup_{r\geq3}\{2J_{r},J_{r+1}\} n − 4 ∈ / ⋃ r ≥ 3 { 2 J r , J r + 1 } and by coframe rank for all n ≥ k + 5 n\geq k+5 n ≥ k + 5 , so for k ≤ 9 k\leq9 k ≤ 9 it vanishes at n = 8 , 9 n=8,9 n = 8 , 9 and for all n ≥ 11 n\geq11 n ≥ 11 ). Furthermore, at n = k = 4 n=k=4 n = k = 4 , we prove an exact proportionality and on-shell/off-shell dichotomy for U ( 2 ) U(2) U ( 2 ) and S U ( 2 ) SU(2) S U ( 2 ) connections (Lemma 4.3): for any h ⊂ s o ( k ) \mathfrak{h}\subset\mathfrak{so}(k) h ⊂ so ( k ) with χ H = 0 \chi_{H}=0 χ H = 0 , the right-nested EHP 4-form satisfies e ∧ × ( e ∧ Ω ω ) = ( 0 , − 1 2 σ ( e ∧ ( Ω ω e ) ) ) = 1 2 ( e ∧ × e ) ∧ Ω ω = ( 0 , 1 4 σ ( e ∧ e ) , Ω ω ) e\wedge_{\times}(e\wedge_{\Omega}\omega)=(0,-\frac{1}{2}\sigma(e\wedge(\Omega_{\omega}e)))=\frac{1}{2}(e\wedge_{\times}e)\wedge_{\Omega}\omega=(0,\frac{1}{4}\sigma(e\wedge e),\Omega_{\omega}) e ∧ × ( e ∧ Ω ω ) = ( 0 , − 2 1 σ ( e ∧ ( Ω ω e ))) = 2 1 ( e ∧ × e ) ∧ Ω ω = ( 0 , 4 1 σ ( e ∧ e ) , Ω ω ) ; at n = k = 4 n=k=4 n = k = 4 , because the anti-self-dual 2-forms ω a ∈ Λ − 2 R 4 \omega^{a}\in\Lambda^{2}_{-}\mathbb{R}^{4} ω a ∈ Λ − 2 R 4 satisfy ω a ∧ ω b = − 2 δ a b vol 4 \omega^{a}\wedge\omega^{b}=-2\delta^{ab}\operatorname{vol}_{4} ω a ∧ ω b = − 2 δ ab vol 4 and the algebraic first Bianchi identity forces the curvature matrix R a b = R b a R_{ab}=R_{ba} R ab = R ba on s u ( 2 ) \mathfrak{su}(2) su ( 2 ) to be symmetric, both ( e ∧ × e ) ∧ Ω Ω ω (e\wedge_{\times}e)\wedge_{\Omega}\Omega_{\omega} ( e ∧ × e ) ∧ Ω Ω ω and e ∧ × ( e ∧ × Ω ω ) e\wedge_{\times}(e\wedge_{\times}\Omega_{\omega}) e ∧ × ( e ∧ × Ω ω ) vanish identically on-shell for every torsion-free U ( 2 ) U(2) U ( 2 ) and S U ( 2 ) SU(2) S U ( 2 ) connection on Kähler and K 3 K3 K 3 surfaces while remaining non-zero off-shell when torsion Θ e , ω ≠ 0 \Theta_{e,\omega}\neq0 Θ e , ω = 0 is present (whereas in dimension n = k = 6 n=k=6 n = k = 6 , the pure quadratic YM term Ω ω ∧ × ( ( e ∧ × e ) ∧ × Ω ω ) \Omega_{\omega}\wedge_{\times}((e\wedge_{\times}e)\wedge_{\times}\Omega_{\omega}) Ω ω ∧ × (( e ∧ × e ) ∧ × Ω ω ) is non-vanishing on-shell for torsion-free U ( 3 ) U(3) U ( 3 ) and S U ( 3 ) SU(3) S U ( 3 ) connections!). At the critical dimension n = k = 7 n=k=7 n = k = 7 , we prove a complete representation-theoretic structure proposition (Proposition 4.6 ) showing that for H = G 2 ⊂ S O ( 7 ) H=G_{2}\subset SO(7) H = G 2 ⊂ S O ( 7 ) , Schur’s lemma forces the g 2 \mathfrak{g}_{2} g 2 -component B ( 3 ) = σ ( ( e ∧ × G 2 e ) ∧ e ) ∈ Hom G 2 ( Λ 3 R 7 , Λ 14 2 ) B^{(3)}=\sigma((e\wedge_{\times_{G_{2}}}e)\wedge e)\in\operatorname{Hom}_{G_{2}}(\Lambda^{3}\mathbb{R}^{7},\Lambda^{2}_{14}) B ( 3 ) = σ (( e ∧ × G 2 e ) ∧ e ) ∈ Hom G 2 ( Λ 3 R 7 , Λ 14 2 ) of Λ × , b a l 3 e \Lambda^{3}_{\times,\mathrm{bal}}e Λ × , bal 3 e to vanish identically, while the R 7 \mathbb{R}^{7} R 7 -component ( v ( 3 ) ) a = − 7 i a ( ∗ φ 0 ) ≠ 0 (v^{(3)})_{a}=-7i_{a}(\ast\varphi_{0})\neq0 ( v ( 3 ) ) a = − 7 i a ( ∗ φ 0 ) = 0 satisfies ( Ω ω ⋅ v ( 3 ) ) b = 7 ∑ c , d = 1 7 R b c , c d e d ∧ ∗ φ 0 (\Omega_{\omega}\cdot v^{(3)})_{b}=7\sum_{c,d=1}^{7}R_{bc,cd}e^{d}\wedge\ast\varphi_{0} ( Ω ω ⋅ v ( 3 ) ) b = 7 ∑ c , d = 1 7 R b c , c d e d ∧ ∗ φ 0 whenever Ω ω ∈ Λ 14 2 ⊗ g 2 \Omega_{\omega}\in\Lambda^{2}_{14}\otimes\mathfrak{g}_{2} Ω ω ∈ Λ 14 2 ⊗ g 2 by the identity Λ 14 2 ∧ ∗ φ 0 = 0 \Lambda^{2}_{14}\wedge\ast\varphi_{0}=0 Λ 14 2 ∧ ∗ φ 0 = 0 ; consequently, just as at n = k = 4 n=k=4 n = k = 4 , L 7 C Y M \mathcal{L}^{\mathrm{CYM}}_{7} L 7 CYM vanishes on-shell on the torsion-free G 2 G_{2} G 2 -curvature space W 77 ⊂ S 2 ( Λ 14 2 ) \mathcal{W}_{77}\subset S^{2}(\Lambda^{2}_{14}) W 77 ⊂ S 2 ( Λ 14 2 ) (where the first Bianchi identity and G 2 G_{2} G 2 -holonomy force Ric b d = ∑ c = 1 7 R b c , c d = 0 \operatorname{Ric}_{bd}=\sum_{c=1}^{7}R_{bc,cd}=0 Ric b d = ∑ c = 1 7 R b c , c d = 0 ) and is non-vanishing off-shell on Ω 2 ( P ; g 2 ) \Omega^{2}(P;\mathfrak{g}_{2}) Ω 2 ( P ; g 2 ) whenever torsion Θ e , ω ≠ 0 \Theta_{e,\omega}\neq0 Θ e , ω = 0 excites the Λ 7 2 ⊗ g 2 \Lambda^{2}_{7}\otimes\mathfrak{g}_{2} Λ 7 2 ⊗ g 2 or non-Bianchi curvature sectors.
(4) Associator-Quotient Collapse and Jacobiator Dichotomy in Internal Chern-Simons Theories, with Applications to B F BF B F , Higgs, and Dual Theories (Theorem 5.3 and Propositions 5.4–5.6): We extend our framework to odd-dimensional internal A A A -valued Chern-Simons (A A A -CS) theories on linear connection spaces. We prove a three-way structural dichotomy for skew-commutative algebras ( A , ∧ ) (A,\wedge) ( A , ∧ ) : (a) in the associative quotient, all ≥ 3 \geq3 ≥ 3 -fold products vanish identically; (b) in the unquotiented grouped formulation, if ( A , ∧ ) (A,\wedge) ( A , ∧ ) is a Lie algebra (Jac ∧ = 0 \operatorname{Jac}_{\wedge}=0 Jac ∧ = 0 ), the Jacobi identity forces ( D ∧ D ) ∧ D = 1 3 Jac ∧ ( D , D , D ) = 0 (D\wedge D)\wedge D=\frac{1}{3}\operatorname{Jac}_{\wedge}(D,D,D)=0 ( D ∧ D ) ∧ D = 3 1 Jac ∧ ( D , D , D ) = 0 , annihilating the 3D cubic vertex and killing all ≥ 5 \geq5 ≥ 5 D internal A A A -CS theories under both quartic-block and recursive binary-tree parenthesizations; and (c) if ( A , ∧ ) (A,\wedge) ( A , ∧ ) is a non-Lie skew-commutative algebra (Jac ∧ ≠ 0 \operatorname{Jac}_{\wedge}\neq0 Jac ∧ = 0 ), the 3D cubic and 5D quartic Jacobiator vertices survive under quartic-block grouping (with an additional 5D quintic Jacobsthal vertex under recursive binary halving), while for n ≥ 7 n\geq7 n ≥ 7 all block-grouped A A A -CS theories vanish identically and binary-tree A A A -CS theories vanish outside the sparse odd Jacobsthal pairs n ∈ ⋃ m ≥ 4 { J m , J m + 2 } ∩ [ 7 , ∞ ) = { 7 } ∪ { 11 , 13 } ∪ { 21 , 23 } ∪ ⋯ n\in\bigcup_{m\geq4}\{J_m,J_m+2\}\cap[7,\infty)=\{7\}\cup\{11,13\}\cup\{21,23\}\cup\cdots n ∈ ⋃ m ≥ 4 { J m , J m + 2 } ∩ [ 7 , ∞ ) = { 7 } ∪ { 11 , 13 } ∪ { 21 , 23 } ∪ ⋯ (and for n ≥ dim A + 3 n\geq\dim A+3 n ≥ dim A + 3 by rank). We conclude with applications to Cattaneo-Plebanski A A A -B F BF B F equivalences, explicit octonionic and reductive Proca mass spectra in the A A A -valued Yang-Mills-Higgs mechanism, and reductive dualities.
Geometric-Algebraic Foundations of A-YMT Theories
Non-Associative Algebras, Parenthesization, and (k,s)-Solvability
Let R R R be a commutative unital ring of characteristic zero (throughout, R = R R=\mathbb{R} R = R or C \mathbb{C} C ) and let ( A , ∗ ) (A,\ast) ( A , ∗ ) be an R R R -algebra, i.e., an R R R -module A A A equipped with an R R R -bilinear multiplication ∗ : A × A → A \ast:A\times A\to A ∗ : A × A → A , without assuming associativity, commutativity, or a unit element [32 ] . Every multiplication ∗ \ast ∗ decomposes canonically into its skew-symmetric commutator bracket and its symmetric Jordan product:
x ∗ y = 1 2 [ x , y ] ∗ + x ∘ ∗ y , where [ x , y ] ∗ : = x ∗ y − y ∗ x , x ∘ ∗ y : = 1 2 ( x ∗ y + y ∗ x ) . (2) x\ast y=\frac{1}{2}[x,y]_{\ast}+x\circ_{\ast}y,\qquad\text{where}\qquad[x,y]_{\ast}:=x\ast y-y\ast x,\qquad x\circ_{\ast}y:=\frac{1}{2}(x\ast y+y\ast x). \tag*{(2)} x ∗ y = 2 1 [ x , y ] ∗ + x ∘ ∗ y , where [ x , y ] ∗ := x ∗ y − y ∗ x , x ∘ ∗ y := 2 1 ( x ∗ y + y ∗ x ) . (2)
We denote the associator of ( A , ∗ ) (A,\ast) ( A , ∗ ) and the Jacobiator of ( A , [ ⋅ , ⋅ ] ∗ ) (A,[\cdot,\cdot]_{\ast}) ( A , [ ⋅ , ⋅ ] ∗ ) for x , y , z ∈ A x,y,z\in A x , y , z ∈ A by
Assoc ∗ ( x , y , z ) : = ( x ∗ y ) ∗ z − x ∗ ( y ∗ z ) , Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) : = [ [ x , y ] ∗ , z ] ∗ + [ [ y , z ] ∗ , x ] ∗ + [ [ z , x ] ∗ , y ] ∗ . (3) \operatorname{Assoc}_{\ast}(x,y,z):=(x\ast y)\ast z-x\ast(y\ast z),\qquad\operatorname{Jac}_{[\cdot,\cdot]_{\ast}}(x,y,z):=[[x,y]_{\ast},z]_{\ast}+[[y,z]_{\ast},x]_{\ast}+[[z,x]_{\ast},y]_{\ast}. \tag*{(3)} Assoc ∗ ( x , y , z ) := ( x ∗ y ) ∗ z − x ∗ ( y ∗ z ) , Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) := [[ x , y ] ∗ , z ] ∗ + [[ y , z ] ∗ , x ] ∗ + [[ z , x ] ∗ , y ] ∗ . (3)
Thus ( A , [ ⋅ , ⋅ ] ∗ ) (A,[\cdot,\cdot]_{\ast}) ( A , [ ⋅ , ⋅ ] ∗ ) is a Lie algebra if and only if Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) = 0 \operatorname{Jac}_{[\cdot,\cdot]_{\ast}}(x,y,z)=0 Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) = 0 for all x , y , z ∈ A x,y,z\in A x , y , z ∈ A (in which case ( A , ∗ ) (A,\ast) ( A , ∗ ) is called Lie-admissible [32 ] ; every associative algebra is Lie-admissible since
Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) = ∑ σ ∈ S 3 sign ( σ ) Assoc ∗ ( x σ ( 1 ) , x σ ( 2 ) , x σ ( 3 ) ) . \operatorname{Jac}_{[\cdot,\cdot]_{\ast}}(x,y,z)=\sum_{\sigma\in S_3}\operatorname{sign}(\sigma)\operatorname{Assoc}_{\ast}(x_{\sigma(1)},x_{\sigma(2)},x_{\sigma(3)}). Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) = σ ∈ S 3 ∑ sign ( σ ) Assoc ∗ ( x σ ( 1 ) , x σ ( 2 ) , x σ ( 3 ) ) .
More generally, the Lie nucleus (or Jacobi nucleus ) of ( A , [ ⋅ , ⋅ ] ∗ ) (A,[\cdot,\cdot]_{\ast}) ( A , [ ⋅ , ⋅ ] ∗ ) is the Lie subalgebra
N L i e ( A ) : = { x ∈ A | Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) = 0 for all y , z ∈ A } . (4) \mathcal{N}_{\mathrm{Lie}}(A):=\left\{x\in A\ \middle|\ \operatorname{Jac}_{[\cdot,\cdot]_{\ast}}(x,y,z)=0\text{ for all }y,z\in A\right\}. \tag*{(4)} N Lie ( A ) := { x ∈ A Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) = 0 for all y , z ∈ A } . (4)
By definition of Jac [ ⋅ , ⋅ ] ∗ \operatorname{Jac}_{[\cdot,\cdot]_{\ast}} Jac [ ⋅ , ⋅ ] ∗ , the inner adjoint map ad x [ ⋅ , ⋅ ] ∗ ( y ) : = [ x , y ] ∗ \operatorname{ad}_{x}^{[\cdot,\cdot]_{\ast}}(y):=[x,y]_{\ast} ad x [ ⋅ , ⋅ ] ∗ ( y ) := [ x , y ] ∗ is a derivation of the commutator bracket [ y , z ] ∗ [y,z]_{\ast} [ y , z ] ∗ if and only if x ∈ N L i e ( A ) x\in\mathcal{N}_{\mathrm{Lie}}(A) x ∈ N Lie ( A ) .
Given a smooth manifold P P P , let Ω ( P ; A ) = ⨁ k ≥ 0 Ω k ( P ; A ) \Omega(P;A)=\bigoplus_{k\geq0}\Omega^k(P;A) Ω ( P ; A ) = ⨁ k ≥ 0 Ω k ( P ; A ) denote the N \mathbb{N} N -graded R R R -module of smooth A A A -valued differential forms on P P P , where Ω k ( P ; A ) = Γ ( Λ k T ∗ P ⊗ A ) \Omega^k(P;A)=\Gamma(\Lambda^kT^*P\otimes A) Ω k ( P ; A ) = Γ ( Λ k T ∗ P ⊗ A ) . The bilinear multiplication ∗ : A ⊗ A → A \ast:A\otimes A\to A ∗ : A ⊗ A → A induces a canonical exterior product
∧ ∗ : Ω k ( P ; A ) × Ω l ( P ; A ) ⟶ Ω k + l ( P ; A ) , (5) \wedge_{\ast}:\Omega^k(P;A)\times\Omega^l(P;A)\longrightarrow\Omega^{k+l}(P;A), \tag*{(5)} ∧ ∗ : Ω k ( P ; A ) × Ω l ( P ; A ) ⟶ Ω k + l ( P ; A ) , (5)
defined pointwise at p ∈ P p\in P p ∈ P on tangent vectors v 1 , … , v k + l ∈ T p P v_1,\ldots,v_{k+l}\in T_pP v 1 , … , v k + l ∈ T p P by
( α ∧ ∗ β ) p ( v 1 , … , v k + l ) = 1 k ! l ! ∑ σ ∈ S k + l sign ( σ ) α p ( v σ ( 1 ) , … , v σ ( k ) ) ∗ β p ( v σ ( k + 1 ) , … , v σ ( k + l ) ) . (6) (\alpha\wedge_{\ast}\beta)_p(v_1,\ldots,v_{k+l})=\frac{1}{k!l!}\sum_{\sigma\in S_{k+l}}\operatorname{sign}(\sigma)\,\alpha_p(v_{\sigma(1)},\ldots,v_{\sigma(k)})\ast\beta_p(v_{\sigma(k+1)},\ldots,v_{\sigma(k+l)}). \tag*{(6)} ( α ∧ ∗ β ) p ( v 1 , … , v k + l ) = k ! l ! 1 σ ∈ S k + l ∑ sign ( σ ) α p ( v σ ( 1 ) , … , v σ ( k ) ) ∗ β p ( v σ ( k + 1 ) , … , v σ ( k + l ) ) . (6)
In terms of the decomposition (2) , for any α ∈ Ω k ( P ; A ) \alpha\in\Omega^k(P;A) α ∈ Ω k ( P ; A ) and β ∈ Ω l ( P ; A ) \beta\in\Omega^l(P;A) β ∈ Ω l ( P ; A ) we have
α ∧ [ ⋅ , ⋅ ] ∗ β = α ∧ ∗ β − ( − 1 ) k l β ∧ ∗ α , α ∧ ∘ ∗ β = 1 2 ( α ∧ ∗ β + ( − 1 ) k l β ∧ ∗ α ) . (7) \alpha\wedge_{[\cdot,\cdot]_{\ast}}\beta=\alpha\wedge_{\ast}\beta-(-1)^{kl}\beta\wedge_{\ast}\alpha,\qquad
\alpha\wedge_{\circ_{\ast}}\beta=\frac{1}{2}\left(\alpha\wedge_{\ast}\beta+(-1)^{kl}\beta\wedge_{\ast}\alpha\right). \tag*{(7)} α ∧ [ ⋅ , ⋅ ] ∗ β = α ∧ ∗ β − ( − 1 ) k l β ∧ ∗ α , α ∧ ∘ ∗ β = 2 1 ( α ∧ ∗ β + ( − 1 ) k l β ∧ ∗ α ) . (7)
Setting α = β ∈ Ω k ( P ; A ) \alpha=\beta\in\Omega^k(P;A) α = β ∈ Ω k ( P ; A ) in (7) yields the basic parity dichotomy of exterior squares:
Odd-Degree Forms (k k k odd, e.g., connection 1-forms D ∈ Ω 1 ( P ; A ) D \in\Omega^{1}(P; A) D ∈ Ω 1 ( P ; A ) ): ( − 1 ) k 2 = − 1 (-1)^{k^{2}}=-1 ( − 1 ) k 2 = − 1 , so α ∧ ∘ ∗ α = 0 \alpha\wedge_{\circ*}\alpha=0 α ∧ ∘∗ α = 0 identically and
α ∧ ∗ α = 1 2 α ∧ [ ⋅ , ⋅ ] ∗ α ( α ∈ Ω o d d ( P ; A ) ) . (8) \alpha\wedge_{*}\alpha=\frac{1}{2}\alpha\wedge_{[\cdot,\cdot]_{*}}\alpha\quad(\alpha\in\Omega^{\mathrm{odd}}(P; A)). \tag*{(8)} α ∧ ∗ α = 2 1 α ∧ [ ⋅ , ⋅ ] ∗ α ( α ∈ Ω odd ( P ; A )) . (8)
Hence the exterior square of any odd-degree form depends solely on the commutator bracket [ x , y ] ∗ [x,y]_{*} [ x , y ] ∗ and vanishes identically for all odd forms (when dim P ≥ k \dim P\ge k dim P ≥ k ) if and only if ( A , ∗ ) (A,*) ( A , ∗ ) is commutative ([ x , y ] ∗ = 0 [x,y]_{*}=0 [ x , y ] ∗ = 0 ).
Even-Degree Forms (k k k even, e.g., curvature 2-forms F D ∈ Ω 2 ( P ; A ) F_{D}\in\Omega^{2}(P; A) F D ∈ Ω 2 ( P ; A ) ): ( − 1 ) k 2 = + 1 (-1)^{k^{2}}=+1 ( − 1 ) k 2 = + 1 , so α ∧ [ ⋅ , ⋅ ] ∗ α = 0 \alpha\wedge_{[\cdot,\cdot]_{*}}\alpha=0 α ∧ [ ⋅ , ⋅ ] ∗ α = 0 identically and
α ∧ ∗ α = α ∧ ∘ ∗ α ( α ∈ Ω e v e n ( P ; A ) ) . (9) \alpha\wedge_{*}\alpha=\alpha\wedge_{\circ*}\alpha\quad(\alpha\in\Omega^{\mathrm{even}}(P; A)). \tag*{(9)} α ∧ ∗ α = α ∧ ∘∗ α ( α ∈ Ω even ( P ; A )) . (9)
Hence the exterior square of any even-degree form depends solely on the symmetric Jordan product x ∘ ∗ y x\circ_{*}y x ∘ ∗ y and vanishes identically whenever ( A , ∗ ) (A,*) ( A , ∗ ) is skew-commutative (x ∘ ∗ y = 0 x\circ_{*}y=0 x ∘ ∗ y = 0 ).
Notice that when passing from algebra elements x i ∈ A x_i\in A x i ∈ A to differential forms α i ∈ Ω k i ( P ; A ) \alpha_i\in\Omega^{k_i}(P; A) α i ∈ Ω k i ( P ; A ) , permuting odd-degree forms introduces Koszul signs ( − 1 ) k i k j (-1)^{k_i k_j} ( − 1 ) k i k j that twist symmetric algebra identities into alternating ones (and vice versa). Thus, whereas a non-permuting multilinear identity p ( x 1 , … , x m ) = 0 p(x_1,\ldots,x_m)=0 p ( x 1 , … , x m ) = 0 (such as a parenthesized nilpotency identity x 1 ∗ ⋯ ∗ x m = 0 x_1*\cdots*x_m=0 x 1 ∗ ⋯ ∗ x m = 0 ) or an identity evaluated on even-degree forms lifts directly by multilinearity [19 , 13 ] , polynomial nilpotency conditions on k k k -forms must be defined directly on Ω k ( P ; A ) \Omega^{k}(P; A) Ω k ( P ; A ) [21 ] .
When ( A , ∗ ) (A,*) ( A , ∗ ) is non-associative, products of m ≥ 3 m\ge3 m ≥ 3 forms depend essentially on the choice of parenthesization. To avoid any ambiguity in higher-degree powers ∧ ∗ m α \wedge_{*}^{m}\alpha ∧ ∗ m α , we distinguish three explicit parenthesization conventions for an A A A -valued form α ∈ Ω k ( P ; A ) \alpha\in\Omega^{k}(P; A) α ∈ Ω k ( P ; A ) :
Full parenthesization: All non-associative monomial parenthesizations p m ( α , … , α ) p_m(\alpha,\ldots,\alpha) p m ( α , … , α ) of degree m m m .
Recursive binary-tree parenthesization ( ∧ ∗ , b a l m α ) (\wedge_{*,\mathrm{bal}}^{m}\alpha) ( ∧ ∗ , bal m α ) : Defined inductively for m ≥ 1 m\ge1 m ≥ 1 by ∧ ∗ , b a l 1 α : = α \wedge_{*,\mathrm{bal}}^{1}\alpha:=\alpha ∧ ∗ , bal 1 α := α and by exact binary halving
∧ ∗ , b a l m α : = ( ∧ ∗ , b a l ⌈ m / 2 ⌉ α ) ∧ ∗ ( ∧ ∗ , b a l ⌊ m / 2 ⌋ α ) ( m ≥ 2 ) . (10) \wedge_{*,\mathrm{bal}}^{m}\alpha:=\left(\wedge_{*,\mathrm{bal}}^{\lceil m/2\rceil}\alpha\right)\wedge_{*}\left(\wedge_{*,\mathrm{bal}}^{\lfloor m/2\rfloor}\alpha\right)\quad(m\ge2). \tag*{(10)} ∧ ∗ , bal m α := ( ∧ ∗ , bal ⌈ m /2 ⌉ α ) ∧ ∗ ( ∧ ∗ , bal ⌊ m /2 ⌋ α ) ( m ≥ 2 ) . (10)
In particular, for 1 ≤ m ≤ 8 1\le m\le8 1 ≤ m ≤ 8 , recursive binary halving produces:
∧ ∗ , b a l 2 α = α ∧ ∗ α , ∧ ∗ , b a l 3 α = ( α ∧ ∗ α ) ∧ ∗ α , ∧ ∗ , b a l 4 α = ( α ∧ ∗ α ) ∧ ∗ ( α ∧ ∗ α ) , \wedge_{*,\mathrm{bal}}^{2}\alpha=\alpha\wedge_{*}\alpha,\qquad
\wedge_{*,\mathrm{bal}}^{3}\alpha=(\alpha\wedge_{*}\alpha)\wedge_{*}\alpha,\qquad
\wedge_{*,\mathrm{bal}}^{4}\alpha=(\alpha\wedge_{*}\alpha)\wedge_{*}(\alpha\wedge_{*}\alpha), ∧ ∗ , bal 2 α = α ∧ ∗ α , ∧ ∗ , bal 3 α = ( α ∧ ∗ α ) ∧ ∗ α , ∧ ∗ , bal 4 α = ( α ∧ ∗ α ) ∧ ∗ ( α ∧ ∗ α ) ,
∧ ∗ , b a l 5 α = ( ( α ∧ ∗ α ) ∧ ∗ α ) ∧ ∗ ( α ∧ ∗ α ) , ∧ ∗ , b a l 6 α = ( ( α ∧ ∗ α ) ∧ ∗ α ) ∧ ∗ ( ( α ∧ ∗ α ) ∧ ∗ α ) , \wedge_{*,\mathrm{bal}}^{5}\alpha=((\alpha\wedge_{*}\alpha)\wedge_{*}\alpha)\wedge_{*}(\alpha\wedge_{*}\alpha),\qquad
\wedge_{*,\mathrm{bal}}^{6}\alpha=((\alpha\wedge_{*}\alpha)\wedge_{*}\alpha)\wedge_{*}((\alpha\wedge_{*}\alpha)\wedge_{*}\alpha), ∧ ∗ , bal 5 α = (( α ∧ ∗ α ) ∧ ∗ α ) ∧ ∗ ( α ∧ ∗ α ) , ∧ ∗ , bal 6 α = (( α ∧ ∗ α ) ∧ ∗ α ) ∧ ∗ (( α ∧ ∗ α ) ∧ ∗ α ) ,
∧ ∗ , b a l 7 α = ( ∧ ∗ , b a l 4 α ) ∧ ∗ ( ∧ ∗ , b a l 3 α ) , ∧ ∗ , b a l 8 α = ( ∧ ∗ , b a l 4 α ) ∧ ∗ ( ∧ ∗ , b a l 4 α ) . (11) \wedge_{*,\mathrm{bal}}^{7}\alpha=(\wedge_{*,\mathrm{bal}}^{4}\alpha)\wedge_{*}(\wedge_{*,\mathrm{bal}}^{3}\alpha),\qquad
\wedge_{*,\mathrm{bal}}^{8}\alpha=(\wedge_{*,\mathrm{bal}}^{4}\alpha)\wedge_{*}(\wedge_{*,\mathrm{bal}}^{4}\alpha). \tag*{(11)} ∧ ∗ , bal 7 α = ( ∧ ∗ , bal 4 α ) ∧ ∗ ( ∧ ∗ , bal 3 α ) , ∧ ∗ , bal 8 α = ( ∧ ∗ , bal 4 α ) ∧ ∗ ( ∧ ∗ , bal 4 α ) . (11)
Quartic-block parenthesization ( ∧ ∗ , b l k m α ) (\wedge_{*,\mathrm{blk}}^{m}\alpha) ( ∧ ∗ , blk m α ) : Defined so that it agrees with ∧ ∗ , b a l m α \wedge_{*,\mathrm{bal}}^{m}\alpha ∧ ∗ , bal m α for 1 ≤ m ≤ 4 1\le m\le4 1 ≤ m ≤ 4 , and for m ≥ 5 m\ge5 m ≥ 5 factors out leading quartic blocks ∧ ∗ , b a l 4 α \wedge_{*,\mathrm{bal}}^{4}\alpha ∧ ∗ , bal 4 α recursively:
∧ ∗ , b l k m α : = { ∧ ∗ , b a l m α , if 1 ≤ m ≤ 4 , ( ∧ ∗ , b a l 4 α ) ∧ ∗ ( ∧ ∗ , b l k m − 4 α ) , if m ≥ 5. (12) \wedge_{*,\mathrm{blk}}^{m}\alpha:=
\begin{cases}
\wedge_{*,\mathrm{bal}}^{m}\alpha, & \text{if }1\le m\le4,\\
(\wedge_{*,\mathrm{bal}}^{4}\alpha)\wedge_{*}(\wedge_{*,\mathrm{blk}}^{m-4}\alpha), & \text{if }m\ge5.
\end{cases} \tag*{(12)} ∧ ∗ , blk m α := { ∧ ∗ , bal m α , ( ∧ ∗ , bal 4 α ) ∧ ∗ ( ∧ ∗ , blk m − 4 α ) , if 1 ≤ m ≤ 4 , if m ≥ 5. (12)
By convention, when m = 0 m=0 m = 0 , ( ∧ ∗ , b a l 0 α ) ∧ ∗ β = ( ∧ ∗ , b l k 0 α ) ∧ ∗ β : = β (\wedge_{*,\mathrm{bal}}^{0}\alpha)\wedge_{*}\beta=(\wedge_{*,\mathrm{blk}}^{0}\alpha)\wedge_{*}\beta:=\beta ( ∧ ∗ , bal 0 α ) ∧ ∗ β = ( ∧ ∗ , blk 0 α ) ∧ ∗ β := β omits the α \alpha α -factor.
Lemma 2.1 (Comparison of Parenthesization Schemes and All-Orders Jacobsthal Characterization on 1-Forms in Skew-Commutative Algebras). Let ( A , ∗ ) (A,*) ( A , ∗ ) be a skew-commutative algebra (x ∗ y = − y ∗ x x*y=-y*x x ∗ y = − y ∗ x , so x ∘ ∗ y = 0 x\circ_{*}y=0 x ∘ ∗ y = 0 ), and let e ∈ Ω 1 ( P ; A ) e\in\Omega^{1}(P; A) e ∈ Ω 1 ( P ; A ) be an A A A -valued 1-form.
(1) Quartic-Block Vanishing: Because e ∧ ∗ e ∈ Ω 2 ( P ; A ) e\wedge_{*}e\in\Omega^{2}(P; A) e ∧ ∗ e ∈ Ω 2 ( P ; A ) is an even-degree form, (9) gives
∧ ∗ , b a l 4 e = ( e ∧ ∗ e ) ∧ ∗ ( e ∧ ∗ e ) = 0. (13) \wedge_{*,\mathrm{bal}}^{4}e=(e\wedge_{*}e)\wedge_{*}(e\wedge_{*}e)=0. \tag*{(13)} ∧ ∗ , bal 4 e = ( e ∧ ∗ e ) ∧ ∗ ( e ∧ ∗ e ) = 0. (13) Consequently, under quartic-block parenthesization (12) , ∧ ∗ , b l k m e = 0 \wedge_{*,\mathrm{blk}}^{m}e=0 ∧ ∗ , blk m e = 0 identically for all m ≥ 4 m\ge4 m ≥ 4 .
(2) Recursive Binary-Tree Halving and Jacobsthal Survivor Pairs: Under recursive binary-tree parenthesization (10) , let J r : = 2 r − ( − 1 ) r 3 J_r := \frac{2^r-(-1)^r}{3} J r := 3 2 r − ( − 1 ) r (r ≥ 1 r \ge1 r ≥ 1 ) denote the r r r -th Jacobsthal number (J 1 = 1 , J 2 = 1 , J 3 = 3 , J 4 = 5 , J 5 = 11 , J 6 = 21 , J 7 = 43 , … J_1 = 1, J_2 = 1, J_3 = 3, J_4 = 5, J_5 = 11, J_6 = 21, J_7 = 43, \ldots J 1 = 1 , J 2 = 1 , J 3 = 3 , J 4 = 5 , J 5 = 11 , J 6 = 21 , J 7 = 43 , … , satisfying J r + 1 = 2 J r + ( − 1 ) r J_{r+1} = 2J_r + (-1)^r J r + 1 = 2 J r + ( − 1 ) r ), and define the set of binary-tree survivor degrees
S b a l : = { 1 , 2 , 3 } ∪ ⋃ r = 3 ∞ { 2 J r , J r + 1 } = { 1 , 2 , 3 } ∪ { 5 , 6 } ∪ { 10 , 11 } ∪ { 21 , 22 } ∪ { 42 , 43 } ∪ ⋯ (14) \mathcal{S}_{\mathrm{bal}} := \{1,2,3\} \cup\bigcup_{r=3}^{\infty}\{2J_r,J_{r+1}\} = \{1,2,3\} \cup\{5,6\} \cup\{10,11\} \cup\{21,22\} \cup\{42,43\} \cup\cdots \tag*{(14)} S bal := { 1 , 2 , 3 } ∪ r = 3 ⋃ ∞ { 2 J r , J r + 1 } = { 1 , 2 , 3 } ∪ { 5 , 6 } ∪ { 10 , 11 } ∪ { 21 , 22 } ∪ { 42 , 43 } ∪ ⋯ (14)
(a) For every integer m ∉ S b a l m \notin\mathcal{S}_{\mathrm{bal}} m ∈ / S bal (in particular, for m = 4 m = 4 m = 4 , 7 ≤ m ≤ 9 7 \le m \le9 7 ≤ m ≤ 9 , 12 ≤ m ≤ 20 12 \le m \le20 12 ≤ m ≤ 20 , 23 ≤ m ≤ 41 23 \le m \le41 23 ≤ m ≤ 41 , etc.), ⋀ ∗ , b a l m e = 0 \bigwedge_{\ast,\mathrm{bal}}^m e = 0 ⋀ ∗ , bal m e = 0 holds identically in Ω m ( P ; A ) \Omega^m(P;A) Ω m ( P ; A ) . Moreover, if ⋀ ∗ , b a l 3 e = 0 \bigwedge_{\ast,\mathrm{bal}}^3 e = 0 ⋀ ∗ , bal 3 e = 0 as well (which holds whenever ( A , ∗ ) (A,\ast) ( A , ∗ ) is a Lie algebra), then ⋀ ∗ , b a l m e = 0 \bigwedge_{\ast,\mathrm{bal}}^m e = 0 ⋀ ∗ , bal m e = 0 for all m ≥ 3 m \ge3 m ≥ 3 ; more generally, if ⋀ ∗ , b a l J r e = 0 \bigwedge_{\ast,\mathrm{bal}}^{J_r} e = 0 ⋀ ∗ , bal J r e = 0 for some r ≥ 3 r \ge3 r ≥ 3 , then ⋀ ∗ , b a l m e = 0 \bigwedge_{\ast,\mathrm{bal}}^m e = 0 ⋀ ∗ , bal m e = 0 for all m ≥ 2 J r − 1 m \ge2J_r - 1 m ≥ 2 J r − 1 .
(b) Conversely, at each dyadic level r ≥ 3 r \ge3 r ≥ 3 , the consecutive pair { 2 J r , J r + 1 } ⊂ S b a l \{2J_r,J_{r+1}\} \subset\mathcal{S}_{\mathrm{bal}} { 2 J r , J r + 1 } ⊂ S bal consists of the exterior square of the odd-degree J r J_r J r -form ⋀ ∗ , b a l J r e ∈ Ω J r ( P ; A ) \bigwedge_{\ast,\mathrm{bal}}^{J_r}e \in\Omega^{J_r}(P;A) ⋀ ∗ , bal J r e ∈ Ω J r ( P ; A ) ,
⋀ ∗ , b a l 2 J r e = ( ⋀ ∗ , b a l J r e ) ∧ ∗ ( ⋀ ∗ , b a l J r e ) = 1 2 ( ⋀ ∗ , b a l J r e ) ∧ [ ⋅ , ⋅ ] ∗ ( ⋀ ∗ , b a l J r e ) , \bigwedge_{\ast,\mathrm{bal}}^{2J_r}e
=
\left(\bigwedge_{\ast,\mathrm{bal}}^{J_r}e\right)\wedge_\ast
\left(\bigwedge_{\ast,\mathrm{bal}}^{J_r}e\right)
=
\frac{1}{2}\left(\bigwedge_{\ast,\mathrm{bal}}^{J_r}e\right)
\wedge_{[\cdot,\cdot]_\ast}
\left(\bigwedge_{\ast,\mathrm{bal}}^{J_r}e\right), ∗ , bal ⋀ 2 J r e = ∗ , bal ⋀ J r e ∧ ∗ ∗ , bal ⋀ J r e = 2 1 ∗ , bal ⋀ J r e ∧ [ ⋅ , ⋅ ] ∗ ∗ , bal ⋀ J r e ,
which by (8) projects onto the commutator bracket [ x , y ] ∗ = 2 x ∗ y [x,y]_\ast= 2x\ast y [ x , y ] ∗ = 2 x ∗ y rather than the Jordan product, together with the odd-degree product of ⋀ ∗ , b a l J r e \bigwedge_{\ast,\mathrm{bal}}^{J_r}e ⋀ ∗ , bal J r e and its even neighbor ⋀ ∗ , b a l J r + ( − 1 ) r e = ⋀ ∗ , b a l 2 J r − 1 e \bigwedge_{\ast,\mathrm{bal}}^{J_r+(-1)^r}e = \bigwedge_{\ast,\mathrm{bal}}^{2J_r-1}e ⋀ ∗ , bal J r + ( − 1 ) r e = ⋀ ∗ , bal 2 J r − 1 e :
⋀ ∗ , b a l J r + 1 e = { ( ⋀ ∗ , b a l J r e ) ∧ ∗ ( ⋀ ∗ , b a l 2 J r − 1 e ) , if r is odd ( J r + 1 = 2 J r − 1 ) , ( ⋀ ∗ , b a l 2 J r − 1 e ) ∧ ∗ ( ⋀ ∗ , b a l J r e ) , if r is even ( J r + 1 = 2 J r + 1 ) . \bigwedge_{\ast,\mathrm{bal}}^{J_{r+1}}e =
\begin{cases}
\left(\bigwedge_{\ast,\mathrm{bal}}^{J_r}e\right)\wedge_\ast
\left(\bigwedge_{\ast,\mathrm{bal}}^{2J_r-1}e\right), & \text{if }r\text{ is odd }(J_{r+1}=2J_r-1),\\
\left(\bigwedge_{\ast,\mathrm{bal}}^{2J_r-1}e\right)\wedge_\ast
\left(\bigwedge_{\ast,\mathrm{bal}}^{J_r}e\right), & \text{if }r\text{ is even }(J_{r+1}=2J_r+1).
\end{cases} ∗ , bal ⋀ J r + 1 e = ⎩ ⎨ ⎧ ( ⋀ ∗ , bal J r e ) ∧ ∗ ( ⋀ ∗ , bal 2 J r − 1 e ) , ( ⋀ ∗ , bal 2 J r − 1 e ) ∧ ∗ ( ⋀ ∗ , bal J r e ) , if r is odd ( J r + 1 = 2 J r − 1 ) , if r is even ( J r + 1 = 2 J r + 1 ) .
(c) Furthermore, if e e e takes values in a k k k -dimensional subspace V ⊂ A V \subset A V ⊂ A (or has pointwise rank ≤ k \le k ≤ k , so that e = ∑ a = 1 k e a ⊗ v a e = \sum_{a=1}^k e^a \otimes v_a e = ∑ a = 1 k e a ⊗ v a is spanned by k k k scalar 1 1 1 -forms e 1 , … , e k ∈ Ω 1 ( P ) e^1,\ldots,e^k \in\Omega^1(P) e 1 , … , e k ∈ Ω 1 ( P ) ), then ⋀ ∗ , P m e = 0 \bigwedge_{\ast,\mathcal{P}}^m e = 0 ⋀ ∗ , P m e = 0 holds identically on P P P for all m > k m > k m > k and every parenthesization scheme P \mathcal{P} P (since Λ m ( R k ) ∗ = 0 \Lambda^m(\mathbb{R}^k)_\ast= 0 Λ m ( R k ) ∗ = 0 for m > k m > k m > k ). In particular, if rk ( e ) ≤ 9 \operatorname{rk}(e) \le9 rk ( e ) ≤ 9 , then ⋀ ∗ , b a l m e = 0 \bigwedge_{\ast,\mathrm{bal}}^m e = 0 ⋀ ∗ , bal m e = 0 for all m ≥ 7 m \ge7 m ≥ 7 .
(3) Left-Normed Products: In contrast to both (10) and (12) , left-normed (iterated adjoint) products ( ( ( … ( e ∧ ∗ e ) ∧ ∗ e ) … ) ∧ ∗ e ) (((\ldots(e\wedge_\ast e)\wedge_\ast e)\ldots)\wedge_\ast e) ((( … ( e ∧ ∗ e ) ∧ ∗ e ) … ) ∧ ∗ e ) do not vanish at degree 4 in general Lie algebras s o ( k ) \mathfrak{so}(k) so ( k ) (k ≥ 4 k \ge4 k ≥ 4 ).
Proof. Since e ∧ ∗ e e\wedge_\ast e e ∧ ∗ e has even degree 2 and ( A , ∗ ) (A,\ast) ( A , ∗ ) is skew-commutative, (9) yields ⋀ ∗ , b a l 4 e = ( e ∧ ∗ e ) ∧ ∗ ( e ∧ ∗ e ) = 0 \bigwedge_{\ast,\mathrm{bal}}^4 e = (e\wedge_\ast e)\wedge_\ast(e\wedge_\ast e)=0 ⋀ ∗ , bal 4 e = ( e ∧ ∗ e ) ∧ ∗ ( e ∧ ∗ e ) = 0 . By (12) , ⋀ ∗ , b l k m e = ( ⋀ ∗ , b a l 4 e ) ∧ ∗ ( ⋀ ∗ , b l k m − 4 e ) = 0 \bigwedge_{\ast,\mathrm{blk}}^m e = (\bigwedge_{\ast,\mathrm{bal}}^4 e)\wedge_\ast(\bigwedge_{\ast,\mathrm{blk}}^{m-4}e)=0 ⋀ ∗ , blk m e = ( ⋀ ∗ , bal 4 e ) ∧ ∗ ( ⋀ ∗ , blk m − 4 e ) = 0 for all m ≥ 4 m\ge4 m ≥ 4 , proving (1).
For (2a)–(2b), note that under recursive binary halving (10) , an even degree m = 2 q m=2q m = 2 q satisfies ⋀ ∗ , b a l 2 q e = ( ⋀ ∗ , b a l q e ) ∧ ∗ ( ⋀ ∗ , b a l q e ) \bigwedge_{\ast,\mathrm{bal}}^{2q}e=(\bigwedge_{\ast,\mathrm{bal}}^q e)\wedge_\ast(\bigwedge_{\ast,\mathrm{bal}}^q e) ⋀ ∗ , bal 2 q e = ( ⋀ ∗ , bal q e ) ∧ ∗ ( ⋀ ∗ , bal q e ) , which vanishes identically by (9) whenever q q q is even, and can survive only when q q q is odd and ⋀ ∗ , b a l q e ≠ 0 \bigwedge_{\ast,\mathrm{bal}}^q e\ne0 ⋀ ∗ , bal q e = 0 . Meanwhile, an odd degree m = 2 q + 1 m=2q+1 m = 2 q + 1 splits into ⌈ m / 2 ⌉ = q + 1 \lceil m/2\rceil=q+1 ⌈ m /2 ⌉ = q + 1 and ⌊ m / 2 ⌋ = q \lfloor m/2\rfloor=q ⌊ m /2 ⌋ = q , so ⋀ ∗ , b a l 2 q + 1 e = ( ⋀ ∗ , b a l q + 1 e ) ∧ ∗ ( ⋀ ∗ , b a l q e ) \bigwedge_{\ast,\mathrm{bal}}^{2q+1}e=(\bigwedge_{\ast,\mathrm{bal}}^{q+1}e)\wedge_\ast(\bigwedge_{\ast,\mathrm{bal}}^q e) ⋀ ∗ , bal 2 q + 1 e = ( ⋀ ∗ , bal q + 1 e ) ∧ ∗ ( ⋀ ∗ , bal q e ) can survive only if the consecutive pair { q , q + 1 } \{q,q+1\} { q , q + 1 } both …
At base level r = 2 r=2 r = 2 , the surviving pair in { 1 , 2 , 3 , 4 } \{1,2,3,4\} { 1 , 2 , 3 , 4 } of degree ≥ 2 \ge2 ≥ 2 is { 2 , 3 } = { 2 J 2 , J 3 } \{2,3\}=\{2J_2,J_3\} { 2 , 3 } = { 2 J 2 , J 3 } , where 2 is even and J 3 = 3 J_3=3 J 3 = 3 is odd (since m = 4 = 2 ⋅ 2 m=4=2\cdot2 m = 4 = 2 ⋅ 2 vanishes).
Inductively, suppose for r ≥ 2 r\ge2 r ≥ 2 that the only surviving degrees at level r r r are the consecutive pair { 2 J r − 1 , J r } \{2J_{r-1},J_r\} { 2 J r − 1 , J r } (which equals { J r − 1 , J r } \{J_r-1,J_r\} { J r − 1 , J r } when r r r is odd and { J r , J r + 1 } \{J_r,J_r+1\} { J r , J r + 1 } when r r r is even), where 2 J r − 1 2J_{r-1} 2 J r − 1 is even and J r J_r J r is odd. At level r + 1 r+1 r + 1 , any surviving even degree m = 2 q m=2q m = 2 q requires q ∈ { 2 J r − 1 , J r } q\in\{2J_{r-1},J_r\} q ∈ { 2 J r − 1 , J r } , which uniquely forces q = J r q=J_r q = J r and m = 2 J r m=2J_r m = 2 J r (whereas m = 2 ( 2 J r − 1 ) = 4 J r − 1 m=2(2J_{r-1})=4J_{r-1} m = 2 ( 2 J r − 1 ) = 4 J r − 1 vanishes by (9) because 2 J r − 1 2J_{r-1} 2 J r − 1 is even!). Any surviving odd degree m = 2 q + 1 m=2q+1 m = 2 q + 1 requires both q q q and q + 1 q+1 q + 1 to survive at level r r r , which uniquely forces { q , q + 1 } = { 2 J r − 1 , J r } \{q,q+1\}=\{2J_{r-1},J_r\} { q , q + 1 } = { 2 J r − 1 , J r } and hence m = 2 J r − 1 + J r = J r + ( − 1 ) r − 1 + J r = 2 J r + ( − 1 ) r = J r + 1 m=2J_{r-1}+J_r=J_r+(-1)^{r-1}+J_r=2J_r+(-1)^r=J_{r+1} m = 2 J r − 1 + J r = J r + ( − 1 ) r − 1 + J r = 2 J r + ( − 1 ) r = J r + 1 .
Since { 2 J r , J r + 1 } \{2J_r,J_{r+1}\} { 2 J r , J r + 1 } is again a consecutive pair consisting of one even integer 2 J r 2J_r 2 J r and one odd integer J r + 1 J_{r+1} J r + 1 , the induction closes for all r ≥ 3 r \ge3 r ≥ 3 , proving that ∧ ∗ , b a l m e = 0 \wedge_{\ast,\mathrm{bal}}^{m}e=0 ∧ ∗ , bal m e = 0 identically for all m ∉ S b a l m\notin S_{\mathrm{bal}} m ∈ / S bal . Moreover, if ∧ ∗ , b a l J r e = 0 \wedge_{\ast,\mathrm{bal}}^{J_r}e=0 ∧ ∗ , bal J r e = 0 for some r ≥ 2 r\ge2 r ≥ 2 , then both ∧ ∗ , b a l 2 J r e \wedge_{\ast,\mathrm{bal}}^{2J_r}e ∧ ∗ , bal 2 J r e and ∧ ∗ , b a l J r + 1 e \wedge_{\ast,\mathrm{bal}}^{J_{r+1}}e ∧ ∗ , bal J r + 1 e vanish, which inductively annihilates all higher levels.
For (2c), expanding e = ∑ a = 1 k e a ⊗ v a e=\sum_{a=1}^{k}e^a\otimes v_a e = ∑ a = 1 k e a ⊗ v a shows by multilinearity that every m m m -fold exterior product ∧ ∗ , P m e \wedge_{\ast,\mathcal{P}}^{m}e ∧ ∗ , P m e is a linear combination of m m m -forms e a 1 ∧ ⋯ ∧ e a m e^{a_1}\wedge\cdots\wedge e^{a_m} e a 1 ∧ ⋯ ∧ e a m , all of which vanish identically for m > k m>k m > k .
Associated Algebra Bundles, A-Connections, and the Lie Nucleus Constraint
Definition 2.2 ((k , s k,s k , s )-Nil and (k , s k,s k , s )-Solvable Algebras [21 ] ). Let ( A , ∗ ) (A,\ast) ( A , ∗ ) be an R R R -algebra and let k , s ≥ 1 k,s\ge1 k , s ≥ 1 be integers. Fix a parenthesization scheme P ∈ { f u l l , b a l , b l k } \mathcal{P}\in\{\mathrm{full},\mathrm{bal},\mathrm{blk}\} P ∈ { full , bal , blk } .
We say that ( A , ∗ ) (A,\ast) ( A , ∗ ) is P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nilpotent of index ≤ s \le s ≤ s (or simply ( k , s ) (k,s) ( k , s ) -nil when P = f u l l \mathcal{P}=\mathrm{full} P = full , binary-balanced ( k , s ) (k,s) ( k , s ) -nil when P = b a l \mathcal{P}=\mathrm{bal} P = bal , and block-balanced ( k , s ) (k,s) ( k , s ) -nil when P = b l k \mathcal{P}=\mathrm{blk} P = blk ) if for every smooth manifold P P P and every A A A -valued k k k -form α ∈ Ω k ( P ; A ) \alpha\in\Omega^k(P;A) α ∈ Ω k ( P ; A ) , we have ∧ ∗ , P m α = 0 \wedge_{\ast,\mathcal{P}}^m\alpha=0 ∧ ∗ , P m α = 0 for all m ≥ s + 1 m\ge s+1 m ≥ s + 1 . (For s = 1 s=1 s = 1 , ∧ ∗ 2 α = α ∧ ∗ α \wedge_{\ast}^{2}\alpha=\alpha\wedge_{\ast}\alpha ∧ ∗ 2 α = α ∧ ∗ α is a single binary product, so all three schemes coincide: ( 1 , 1 ) (1,1) ( 1 , 1 ) -nilpotency is equivalent to commutativity of ( A , ∗ ) (A,\ast) ( A , ∗ ) and ( 2 , 1 ) (2,1) ( 2 , 1 ) -nilpotency is equivalent to skew-commutativity of ( A , ∗ ) (A,\ast) ( A , ∗ ) . By Lemma 2.1(1) [21 ] , every skew-commutative algebra is block-balanced ( 1 , 3 ) (1,3) ( 1 , 3 ) -nil since ∧ ∗ , b l k m α = 0 \wedge_{\ast,\mathrm{blk}}^m\alpha=0 ∧ ∗ , blk m α = 0 for all α ∈ Ω 1 ( P ; A ) \alpha\in\Omega^1(P;A) α ∈ Ω 1 ( P ; A ) and m ≥ 4 m\ge4 m ≥ 4 .)
A submodule V ⊂ A V\subset A V ⊂ A is a P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nilpotent submodule if every V V V -valued k k k -form α ∈ Ω k ( P ; V ) \alpha\in\Omega^k(P;V) α ∈ Ω k ( P ; V ) satisfies ∧ ∗ , P m α = 0 \wedge_{\ast,\mathcal{P}}^m\alpha=0 ∧ ∗ , P m α = 0 for all m ≥ s + 1 m\ge s+1 m ≥ s + 1 in Ω ( P ; A ) \Omega(P;A) Ω ( P ; A ) .
We say that A A A (resp. V ⊂ A V\subset A V ⊂ A ) is P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -solvable—denoted ( k , s ) (k,s) ( k , s ) -solv—if it decomposes as a finite direct sum A = ⨁ i = 1 r A i A=\bigoplus_{i=1}^{r}A_i A = ⨁ i = 1 r A i of mutually orthogonal P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nilpotent submodules satisfying A i ∗ A j = 0 A_i\ast A_j=0 A i ∗ A j = 0 for i ≠ j i\ne j i = j . More generally, if A = ⨁ i = 1 r A i A=\bigoplus_{i=1}^{r}A_i A = ⨁ i = 1 r A i is a vector space direct sum of P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nilpotent subalgebras without A i ∗ A j = 0 A_i\ast A_j=0 A i ∗ A j = 0 , a k k k -form α = ∑ i = 1 r α ( i ) ∈ Ω k ( P ; A ) \alpha=\sum_{i=1}^{r}\alpha^{(i)}\in\Omega^k(P;A) α = ∑ i = 1 r α ( i ) ∈ Ω k ( P ; A ) is called algebraically split-adapted if α ( i ) ∧ ∗ α ( j ) = 0 \alpha^{(i)}\wedge_{\ast}\alpha^{(j)}=0 α ( i ) ∧ ∗ α ( j ) = 0 for all i ≠ j i\ne j i = j ; we denote the set of algebraically split-adapted k k k -forms on P P P by C s p l i t k ( P ; A ) C^k_{\mathrm{split}}(P;A) C split k ( P ; A ) .
Suppose A = ⨁ m ∈ m A m A=\bigoplus_{m\in\mathfrak{m}}A^m A = ⨁ m ∈ m A m is an m \mathfrak{m} m -graded algebra for an abelian group (or monoid) ( m , + ) (\mathfrak{m},+) ( m , + ) . A graded submodule V = ⨁ m ∈ m V m V=\bigoplus_{m\in\mathfrak{m}}V^m V = ⨁ m ∈ m V m is homogeneous weak P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nilpotent if ∧ ∗ , P r α m = 0 \wedge_{\ast,\mathcal{P}}^r\alpha^m=0 ∧ ∗ , P r α m = 0 for all r ≥ s + 1 r\ge s+1 r ≥ s + 1 and each individual homogeneous degree m ∈ m m\in\mathfrak{m} m ∈ m [21 ] , and (multidegree) weak P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nilpotent if pr m ( α m 1 , … , α m r ) = 0 \operatorname{pr}_{\mathfrak{m}}(\alpha^{m_1},\ldots,\alpha^{m_r})=0 pr m ( α m 1 , … , α m r ) = 0 for all P \mathcal{P} P -parenthesizations of degree r ≥ s + 1 r\ge s+1 r ≥ s + 1 , every V V V -valued k k k -form α = ∑ m ∈ m α m \alpha=\sum_{m\in\mathfrak{m}}\alpha^m α = ∑ m ∈ m α m , and all m 1 , … , m r ∈ m m_1,\ldots,m_r\in\mathfrak{m} m 1 , … , m r ∈ m . (Notice that for finite-dimensional algebras or finitely supported graded sums, multidegree weak P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nilpotency of V V V is equivalent by multilinearity to ungraded P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nilpotency of V V V and differs only for infinite-dimensional m \mathfrak{m} m -gradings, whereas homogeneous weak P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nilpotency requires vanishing only on single-degree forms α m ∈ Ω k ( P ; V m ) \alpha^m\in\Omega^k(P;V^m) α m ∈ Ω k ( P ; V m ) and is strictly weaker even in finite dimensions.) We say that V V V (and in particular A A A ) is weak P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -solvable (resp. homogeneous weak P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -solvable) if it decomposes as a finite direct sum V = ⨁ i = 1 r V i V=\bigoplus_{i=1}^{r}V_i V = ⨁ i = 1 r V i of mutually orthogonal weak (resp. homogeneous weak) P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nilpotent graded submodules (V i ∗ V j = 0 V_i\ast V_j=0 V i ∗ V j = 0 for i ≠ j i\ne j i = j ).
Recall that the derived series of an algebra A A A is defined inductively by A ( 0 ) = A A^{(0)}=A A ( 0 ) = A and A ( r ) = A ( r − 1 ) ∗ A ( r − 1 ) A^{(r)}=A^{(r-1)}\ast A^{(r-1)} A ( r ) = A ( r − 1 ) ∗ A ( r − 1 ) [32 , 34 ] ; A A A is solvable of derived length s s s if A ( s + 1 ) = 0 A^{(s+1)}=0 A ( s + 1 ) = 0 . For any form α ∈ Ω k ( P ; A ) \alpha\in\Omega^k(P;A) α ∈ Ω k ( P ; A ) , defining the iterated squaring sequence α [ 0 ] : = α \alpha^{[0]}:=\alpha α [ 0 ] := α and α [ m + 1 ] : = α [ m ] ∧ ∗ α [ m ] \alpha^{[m+1]}:=\alpha^{[m]}\wedge_{\ast}\alpha^{[m]} α [ m + 1 ] := α [ m ] ∧ ∗ α [ m ] immediately implies by induction on m m m that α [ m ] ∈ Ω 2 m k ( P ; A ( m ) ) , \alpha^{[m]}\in\Omega^{2^m k}(P;A^{(m)}), α [ m ] ∈ Ω 2 m k ( P ; A ( m ) ) , and hence α [ s + 1 ] = 0 \alpha^{[s+1]}=0 α [ s + 1 ] = 0 whenever A ( s + 1 ) = 0 A^{(s+1)}=0 A ( s + 1 ) = 0 .
2.2. Associated Algebra Bundles, A A A -Connections, and the Lie Nucleus Constraint.
Rather than introducing category-theoretic functorial bundle systems, we formulate our gauge-theoretic framework directly in terms of principal G G G -bundles and associated algebra bundles [20 , 22 ] .
Let M M M be a smooth n n n -manifold, let π : P → M \pi:P\to M π : P → M be a principal G G G -bundle with structure Lie group G G G and Lie algebra g \mathfrak{g} g , and let ( A , ∗ ) (A,\ast) ( A , ∗ ) be a finite-dimensional real (or complex) algebra equipped with a smooth representation by algebra automorphisms:
ρ : G → Aut Alg ( A ) ⊂ GL ( A ) , ρ ( g ) ( x ∗ y ) = ρ ( g ) ( x ) ∗ ρ ( g ) ( y ) ( ∀ g ∈ G , x , y ∈ A ) . (15) \rho:G\to\operatorname{Aut}_{\operatorname{Alg}}(A)\subset\operatorname{GL}(A),\qquad\rho(g)(x\ast y)=\rho(g)(x)\ast\rho(g)(y)\quad(\forall g\in G,x,y\in A). \tag*{(15)} ρ : G → Aut Alg ( A ) ⊂ GL ( A ) , ρ ( g ) ( x ∗ y ) = ρ ( g ) ( x ) ∗ ρ ( g ) ( y ) ( ∀ g ∈ G , x , y ∈ A ) . (15)
Differentiating ρ \rho ρ at the identity element e ∈ G e \in G e ∈ G yields a Lie algebra representation by derivations of ( A , ∗ ) (A,\ast) ( A , ∗ ) :
d ρ : g ⟶ Der ( A ) , d ρ ( X ) ( x ∗ y ) = d ρ ( X ) ( x ) ∗ y + x ∗ d ρ ( X ) ( y ) ( ∀ X ∈ g , x , y ∈ A ) . (16) d\rho:\mathfrak{g}\longrightarrow\operatorname{Der}(A),\qquad d\rho(X)(x\ast y)=d\rho(X)(x)\ast y+x\ast d\rho(X)(y)\quad(\forall X\in\mathfrak{g},\ x,y\in A). \tag*{(16)} d ρ : g ⟶ Der ( A ) , d ρ ( X ) ( x ∗ y ) = d ρ ( X ) ( x ) ∗ y + x ∗ d ρ ( X ) ( y ) ( ∀ X ∈ g , x , y ∈ A ) . (16)
Because G G G acts on A A A by algebra automorphisms, the associated vector bundle
E A : = P × ρ A ⟶ M (17) E_A:=P\times_\rho A\longrightarrow M \tag*{(17)} E A := P × ρ A ⟶ M (17)
is a smooth algebra bundle over M M M with typical fiber ( A , ∗ ) (A,\ast) ( A , ∗ ) , whose fiberwise multiplication is well-defined by [ p , x ] ∗ [ p , y ] : = [ p , x ∗ y ] [p,x]\ast[p,y]:=[p,x\ast y] [ p , x ] ∗ [ p , y ] := [ p , x ∗ y ] . Moreover, for every G G G -invariant subspace V ⊂ A V\subset A V ⊂ A (in particular, for every Aut Alg ( A ) \operatorname{Aut}_{\operatorname{Alg}}(A) Aut Alg ( A ) -invariant subspace such as the commutator subspace [ A , A ] ∗ [A,A]_\ast [ A , A ] ∗ , the derived terms A ( r ) A^{(r)} A ( r ) , or the Lie nucleus N L i e ( A ) \mathcal{N}_{\mathrm{Lie}}(A) N Lie ( A ) ), E V : = P × ρ V ⊂ E A E_V:=P\times_\rho V\subset E_A E V := P × ρ V ⊂ E A is a canonical smooth vector subbundle of E A E_A E A .
For each k ≥ 0 k\geq0 k ≥ 0 , let Ω e q , ρ k ( P ; A ) \Omega_{\mathrm{eq},\rho}^{k}(P;A) Ω eq , ρ k ( P ; A ) denote the space of ρ \rho ρ -equivariant A A A -valued k k k -forms α ∈ Ω k ( P ; A ) \alpha\in\Omega^k(P;A) α ∈ Ω k ( P ; A ) satisfying R g ∗ α = ρ ( g − 1 ) ∘ α R_g^\ast\alpha=\rho(g^{-1})\circ\alpha R g ∗ α = ρ ( g − 1 ) ∘ α for all g ∈ G g\in G g ∈ G , and let
Ω h e q , ρ k ( P ; A ) : = { α ∈ Ω e q , ρ k ( P ; A ) ∣ ι X α = 0 for all X ∈ g } (18) \Omega_{\mathrm{heq},\rho}^{k}(P;A):=\left\{\alpha\in\Omega_{\mathrm{eq},\rho}^{k}(P;A)\mid\iota_X\alpha=0\ \text{for all }X\in\mathfrak{g}\right\} \tag*{(18)} Ω heq , ρ k ( P ; A ) := { α ∈ Ω eq , ρ k ( P ; A ) ∣ ι X α = 0 for all X ∈ g } (18)
denote the subspace of horizontal , ρ \rho ρ -equivariant A A A -valued k k k -forms on P P P , where X p : = d d t ∣ t = 0 p ⋅ exp ( t X ) X_p:=\left.\frac{d}{dt}\right|_{t=0}p\cdot\exp(tX) X p := d t d t = 0 p ⋅ exp ( tX ) is the fundamental vertical vector field generated by X ∈ g X\in\mathfrak{g} X ∈ g . Because ρ ( g ) ∈ Aut Alg ( A ) \rho(g)\in\operatorname{Aut}_{\operatorname{Alg}}(A) ρ ( g ) ∈ Aut Alg ( A ) and ι X ( α ∧ ∗ β ) = ( ι X α ) ∧ ∗ β + ( − 1 ) deg α α ∧ ∗ ( ι X β ) \iota_X(\alpha\wedge_\ast\beta)=(\iota_X\alpha)\wedge_\ast\beta+(-1)^{\deg\alpha}\alpha\wedge_\ast(\iota_X\beta) ι X ( α ∧ ∗ β ) = ( ι X α ) ∧ ∗ β + ( − 1 ) d e g α α ∧ ∗ ( ι X β ) , the direct sum Ω h e q , ρ ∙ ( P ; A ) = ⨁ k = 0 n Ω h e q , ρ k ( P ; A ) \Omega_{\mathrm{heq},\rho}^{\bullet}(P;A)=\bigoplus_{k=0}^{n}\Omega_{\mathrm{heq},\rho}^{k}(P;A) Ω heq , ρ ∙ ( P ; A ) = ⨁ k = 0 n Ω heq , ρ k ( P ; A ) is a Z \mathbb{Z} Z -graded subalgebra of ( Ω ∙ ( P ; A ) , ∧ ∗ ) (\Omega^\bullet(P;A),\wedge_\ast) ( Ω ∙ ( P ; A ) , ∧ ∗ ) . The canonical Weil isomorphism [20 , 22 ]
j : Ω h e q , ρ ∙ ( P ; A ) → ∼ Ω ∙ ( M ; E A ) , j ( α ) π ( p ) ( d π p ( v 1 ) , … , d π p ( v k ) ) : = [ p , α p ( v 1 , … , v k ) ] , (19) j:\Omega_{\mathrm{heq},\rho}^{\bullet}(P;A)\xrightarrow{\sim}\Omega^\bullet(M;E_A),\qquad
j(\alpha)_{\pi(p)}(d\pi_p(v_1),\ldots,d\pi_p(v_k)):=[p,\alpha_p(v_1,\ldots,v_k)], \tag*{(19)} j : Ω heq , ρ ∙ ( P ; A ) ∼ Ω ∙ ( M ; E A ) , j ( α ) π ( p ) ( d π p ( v 1 ) , … , d π p ( v k )) := [ p , α p ( v 1 , … , v k )] , (19)
is an isomorphism of Z \mathbb{Z} Z -graded algebras: j ( α ∧ ∗ β ) = j ( α ) ∧ ∗ j ( β ) j(\alpha\wedge_\ast\beta)=j(\alpha)\wedge_\ast j(\beta) j ( α ∧ ∗ β ) = j ( α ) ∧ ∗ j ( β ) for all α , β ∈ Ω h e q , ρ ∙ ( P ; A ) \alpha,\beta\in\Omega_{\mathrm{heq},\rho}^{\bullet}(P;A) α , β ∈ Ω heq , ρ ∙ ( P ; A ) , and for every G G G -invariant subspace V ⊂ A V\subset A V ⊂ A , j j j maps Ω h e q , ρ k ( P ; V ) \Omega_{\mathrm{heq},\rho}^{k}(P;V) Ω heq , ρ k ( P ; V ) isomorphically onto Ω k ( M ; E V ) \Omega^k(M;E_V) Ω k ( M ; E V ) .
A smooth bundle trace functional on E A E_A E A is a fiberwise linear bundle morphism tr ξ : E A → M × R \operatorname{tr}_{\xi}:E_A\to M\times\mathbb{R} tr ξ : E A → M × R determined by a smooth global section ξ ∈ Γ ( E A ∗ ) \xi\in\Gamma(E_A^\ast) ξ ∈ Γ ( E A ∗ ) of the dual bundle E A ∗ = P × ρ ∗ A ∗ E_A^\ast=P\times_{\rho^\ast}A^\ast E A ∗ = P × ρ ∗ A ∗ via tr ξ ( Ψ ) : = ⟨ ξ , Ψ ⟩ ∈ Ω ∙ ( M ; R ) \operatorname{tr}_{\xi}(\Psi):=\langle\xi,\Psi\rangle\in\Omega^\bullet(M;\mathbb{R}) tr ξ ( Ψ ) := ⟨ ξ , Ψ ⟩ ∈ Ω ∙ ( M ; R ) for Ψ ∈ Ω ∙ ( M ; E A ) \Psi\in\Omega^\bullet(M;E_A) Ψ ∈ Ω ∙ ( M ; E A ) . Equivalently, ξ = j ( ξ ^ ) \xi=j(\widehat{\xi}) ξ = j ( ξ ) corresponds to a G G G -equivariant map ξ ^ : P → A ∗ \widehat{\xi}:P\to A^\ast ξ : P → A ∗ satisfying ξ ^ ( p ⋅ g ) = ρ ( g ) ∗ ξ ^ ( p ) \widehat{\xi}(p\cdot g)=\rho(g)^\ast\widehat{\xi}(p) ξ ( p ⋅ g ) = ρ ( g ) ∗ ξ ( p ) . Such trace sections arise in two geometric ways:
(T1) G-Invariant Linear Functional: If τ 0 ∈ ( A ∗ ) G \tau_0\in(A^\ast)^G τ 0 ∈ ( A ∗ ) G is a constant G G G -invariant linear functional on A A A (e.g., the matrix trace on g l ( k ; R ) \mathfrak{gl}(k;\mathbb{R}) gl ( k ; R ) or the central projection u ( k ) → u ( 1 ) ≃ R \mathfrak{u}(k)\to\mathfrak{u}(1)\simeq\mathbb{R} u ( k ) → u ( 1 ) ≃ R ), then ξ ^ ( p ) ≡ τ 0 \widehat{\xi}(p)\equiv\tau_0 ξ ( p ) ≡ τ 0 defines a strictly G G G -invariant bundle trace morphism tr τ 0 : E A → M × R \operatorname{tr}_{\tau_0}:E_A\to M\times\mathbb{R} tr τ 0 : E A → M × R . Note that if ρ = Ad \rho=\operatorname{Ad} ρ = Ad on a Lie algebra A = g A=\mathfrak{g} A = g , any Ad G \operatorname{Ad}_G Ad G -invariant linear functional τ 0 ∈ ( g ∗ ) G \tau_0\in(\mathfrak{g}^\ast)^G τ 0 ∈ ( g ∗ ) G vanishes identically on the derived subalgebra [ g , g ] [\mathfrak{g},\mathfrak{g}] [ g , g ] (though for a reductive Lie algebra g = z ( g ) ⊕ [ g , g ] \mathfrak{g}=\mathfrak{z}(\mathfrak{g})\oplus[\mathfrak{g},\mathfrak{g}] g = z ( g ) ⊕ [ g , g ] it may be non-zero on the center z ( g ) \mathfrak{z}(\mathfrak{g}) z ( g ) ; however, any interaction term taking values strictly inside [ A , A ] ∗ [A,A]_\ast [ A , A ] ∗ or [ g , g ] [\mathfrak{g},\mathfrak{g}] [ g , g ] is still annihilated by τ 0 \tau_0 τ 0 ).
(T2) Coadjoint Background Section: When ( A ∗ ) G = { 0 } (A^\ast)^G=\{0\} ( A ∗ ) G = { 0 } on [ A , A ] ∗ [A,A]_\ast [ A , A ] ∗ , any smooth non-zero section ξ ∈ Γ ( E A ∗ ) \xi\in\Gamma(E_A^\ast) ξ ∈ Γ ( E A ∗ ) defines a bundle linear functional tr ξ : E A → M × R \operatorname{tr}_{\xi}:E_A\to M\times\mathbb{R} tr ξ : E A → M × R . Wherever ξ ( x ) ≠ 0 \xi(x)\ne0 ξ ( x ) = 0 , choosing ξ \xi ξ acts as a symmetry-breaking background field that reduces the internal structure group G G G at x x x to the stabilizer subgroup G ξ ( x ) ⊂ G G_{\xi(x)}\subset G G ξ ( x ) ⊂ G .
Definition 2.3 (A-Connections, Field Strength, and Covariant Exterior Derivative ). Let π : P → M \pi:P\to M π : P → M be a principal G G G -bundle and let ( A , ∗ ) (A,\ast) ( A , ∗ ) be an R R R -algebra with automorphism representation ρ : G → Aut Alg ( A ) \rho:G\to\operatorname{Aut}_{\operatorname{Alg}}(A) ρ : G → Aut Alg ( A ) .
A pure A A A -connection on P P P is an A A A -valued 1 1 1 -form D ∈ Ω 1 ( P ; A ) D\in\Omega^1(P;A) D ∈ Ω 1 ( P ; A ) . Its field strength (or curvature ) is the A A A -valued 2 2 2 -form F D ∈ Ω 2 ( P ; A ) F_D\in\Omega^2(P;A) F D ∈ Ω 2 ( P ; A ) given by
F D : = d D + 1 2 D ∧ ∗ D = d D + 1 4 D ∧ [ , ] ∗ D . (20) F_D:=dD+\frac{1}{2}D\wedge_\ast D=dD+\frac{1}{4}D\wedge_{[,]\ast}D. \tag*{(20)} F D := d D + 2 1 D ∧ ∗ D = d D + 4 1 D ∧ [ , ] ∗ D . (20)
and its commutator covariant exterior derivative d D [ ⋅ , ⋅ ] ∗ : Ω k ( P ; A ) → Ω k + 1 ( P ; A ) d_D^{[\cdot,\cdot]_*}:\Omega^k(P;A)\to\Omega^{k+1}(P;A) d D [ ⋅ , ⋅ ] ∗ : Ω k ( P ; A ) → Ω k + 1 ( P ; A ) is defined by
d D [ ⋅ , ⋅ ] ∗ α : = d α + 1 2 D ∧ [ ⋅ , ⋅ ] ∗ α = d α + 1 2 ( D ∧ ∗ α − ( − 1 ) k α ∧ ∗ D ) . (21) d_D^{[\cdot,\cdot]_*}\alpha:=d\alpha+\frac{1}{2}D\wedge_{[\cdot,\cdot]_*}\alpha=d\alpha+\frac{1}{2}\left(D\wedge_*\alpha-(-1)^k\alpha\wedge_*D\right). \tag*{(21)} d D [ ⋅ , ⋅ ] ∗ α := d α + 2 1 D ∧ [ ⋅ , ⋅ ] ∗ α = d α + 2 1 ( D ∧ ∗ α − ( − 1 ) k α ∧ ∗ D ) . (21)
For any k k k -form Φ ∈ Ω k ( P ; A ) \Phi\in\Omega^k(P;A) Φ ∈ Ω k ( P ; A ) and 0 0 0 -form ϵ ∈ Ω 0 ( P ; A ) \epsilon\in\Omega^0(P;A) ϵ ∈ Ω 0 ( P ; A ) , we also write [ Φ , ϵ ] ∗ : = Φ ∧ [ ⋅ , ⋅ ] ∗ ϵ = Φ ∧ ∗ ϵ − ϵ ∧ ∗ Φ = − [ ϵ , Φ ] ∗ ∈ Ω k ( P ; A ) [\Phi,\epsilon]_*:=\Phi\wedge_{[\cdot,\cdot]_*}\epsilon=\Phi\wedge_*\epsilon-\epsilon\wedge_*\Phi=-[\epsilon,\Phi]_*\in\Omega^k(P;A) [ Φ , ϵ ] ∗ := Φ ∧ [ ⋅ , ⋅ ] ∗ ϵ = Φ ∧ ∗ ϵ − ϵ ∧ ∗ Φ = − [ ϵ , Φ ] ∗ ∈ Ω k ( P ; A ) .
(2) Suppose A A A is equipped with a G G G -invariant R R R -module decomposition A ≃ A 0 ⊕ A 1 A\simeq A_0\oplus A_1 A ≃ A 0 ⊕ A 1 . A reductive A A A -connection on P P P is a pair ∇ = e + ω ∈ Ω 1 ( P ; A 0 ) ⊕ Ω 1 ( P ; A 1 ) \nabla=e+\omega\in\Omega^1(P;A_0)\oplus\Omega^1(P;A_1) ∇ = e + ω ∈ Ω 1 ( P ; A 0 ) ⊕ Ω 1 ( P ; A 1 ) . Its full Cartan field strength is F ∇ = d ∇ + 1 2 ∇ ∧ ∗ ∇ ∈ Ω 2 ( P ; A ) F_\nabla=d\nabla+\frac{1}{2}\nabla\wedge_*\nabla\in\Omega^2(P;A) F ∇ = d ∇ + 2 1 ∇ ∧ ∗ ∇ ∈ Ω 2 ( P ; A ) , which expands as
F ∇ = Ω ω + Θ e , ω + E e , (22) F_\nabla=\Omega_\omega+\Theta_{e,\omega}+E_e, \tag*{(22)} F ∇ = Ω ω + Θ e , ω + E e , (22)
where Ω ω : = d ω + 1 2 ω ∧ ∗ ω \Omega_\omega:=d\omega+\frac{1}{2}\omega\wedge_*\omega Ω ω := d ω + 2 1 ω ∧ ∗ ω is the pure curvature of ω \omega ω , Θ e , ω : = d e + 1 2 ( ω ∧ ∗ e + e ∧ ∗ ω ) = d e + 1 2 ω ∧ [ ⋅ , ⋅ ] ∗ e \Theta_{e,\omega}:=de+\frac{1}{2}(\omega\wedge_*e+e\wedge_*\omega)=de+\frac{1}{2}\omega\wedge_{[\cdot,\cdot]_*}e Θ e , ω := d e + 2 1 ( ω ∧ ∗ e + e ∧ ∗ ω ) = d e + 2 1 ω ∧ [ ⋅ , ⋅ ] ∗ e is the generalized torsion , and E e : = 1 2 e ∧ ∗ e E_e:=\frac{1}{2}e\wedge_*e E e := 2 1 e ∧ ∗ e is the quadratic soldering form (all evaluated via the ambient multiplication ∗ * ∗ of A A A ).
(3) If A = ⨁ m ∈ m A m A=\bigoplus_{m\in\mathfrak{m}}A^m A = ⨁ m ∈ m A m is an m \mathfrak{m} m -graded algebra ( A m 1 ∗ A m 2 ⊂ A m 1 + m 2 ) (A^{m_1}*A^{m_2}\subset A^{m_1+m_2}) ( A m 1 ∗ A m 2 ⊂ A m 1 + m 2 ) , an A A A -connection D ∈ Ω 1 ( P ; A ) D\in\Omega^1(P;A) D ∈ Ω 1 ( P ; A ) decomposes algebraically along the constant vector space direct sum A = ⨁ m ∈ m A m A=\bigoplus_{m\in\mathfrak{m}}A^m A = ⨁ m ∈ m A m as D = ∑ m ∈ m D m D=\sum_{m\in\mathfrak{m}}D^m D = ∑ m ∈ m D m with D m ∈ Ω 1 ( P ; A m ) D^m\in\Omega^1(P;A^m) D m ∈ Ω 1 ( P ; A m ) , and its curvature F D = ∑ m ∈ m ( F D ) m ∈ ⨁ m ∈ m Ω 2 ( P ; A m ) F_D=\sum_{m\in\mathfrak{m}}(F_D)^m\in\bigoplus_{m\in\mathfrak{m}}\Omega^2(P;A^m) F D = ∑ m ∈ m ( F D ) m ∈ ⨁ m ∈ m Ω 2 ( P ; A m ) has homogeneous algebra components
( F D ) m = d ( D m ) + 1 2 ∑ m 1 + m 2 = m D m 1 ∧ ∗ D m 2 ∈ Ω 2 ( P ; A m ) . (23) (F_D)^m=d(D^m)+\frac{1}{2}\sum_{m_1+m_2=m}D^{m_1}\wedge_*D^{m_2}\in\Omega^2(P;A^m). \tag*{(23)} ( F D ) m = d ( D m ) + 2 1 m 1 + m 2 = m ∑ D m 1 ∧ ∗ D m 2 ∈ Ω 2 ( P ; A m ) . (23)
We now establish the exact algebraic and geometric conditions under which the curvature F D F_D F D , the covariant exterior derivative d D [ ⋅ , ⋅ ] ∗ α d_D^{[\cdot,\cdot]_*}\alpha d D [ ⋅ , ⋅ ] ∗ α , and the Jacobiator forms of D D D are horizontal and ρ \rho ρ -equivariant on P P P , so that they descend to E A → M E_A\to M E A → M via the Weil isomorphism j j j .
Given any trilinear bracket J ( x , y , z ) J(x,y,z) J ( x , y , z ) on A A A (such as Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) \operatorname{Jac}_{[\cdot,\cdot]_*}(x,y,z) Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) or Jac ⋏ ( x , y , z ) : = ( x ⋏ y ) ⋏ z + ( y ⋏ z ) ⋏ x + ( z ⋏ x ) ⋏ y ) , f o r a n y f o r m s \operatorname{Jac}_{\curlywedge}(x,y,z):=(x\curlywedge y)\curlywedge z+(y\curlywedge z)\curlywedge x+(z\curlywedge x)\curlywedge y), for any forms Jac ⋏ ( x , y , z ) := ( x ⋏ y ) ⋏ z + ( y ⋏ z ) ⋏ x + ( z ⋏ x ) ⋏ y ) , f or an y f or m s \alpha\in\Omega^{k_1}(P;A)$, β ∈ Ω k 2 ( P ; A ) \beta\in\Omega^{k_2}(P;A) β ∈ Ω k 2 ( P ; A ) , and γ ∈ Ω k 3 ( P ; A ) \gamma\in\Omega^{k_3}(P;A) γ ∈ Ω k 3 ( P ; A ) with K = k 1 + k 2 + k 3 K=k_1+k_2+k_3 K = k 1 + k 2 + k 3 , we define J ( α , β , γ ) ∈ Ω K ( P ; A ) J(\alpha,\beta,\gamma)\in\Omega^K(P;A) J ( α , β , γ ) ∈ Ω K ( P ; A ) by
J ( α , β , γ ) p ( v 1 , … , v K ) : = 1 k 1 ! k 2 ! k 3 ! ∑ σ ∈ S K sign ( σ ) J ( α p ( v σ ( 1 ) , … ) , β p ( … ) , γ p ( … , v σ ( K ) ) ) . (24) J(\alpha,\beta,\gamma)_p(v_1,\ldots,v_K):=\frac{1}{k_1!k_2!k_3!}\sum_{\sigma\in S_K}\operatorname{sign}(\sigma)J\left(\alpha_p(v_{\sigma(1)},\ldots),\beta_p(\ldots),\gamma_p(\ldots,v_{\sigma(K)})\right). \tag*{(24)} J ( α , β , γ ) p ( v 1 , … , v K ) := k 1 ! k 2 ! k 3 ! 1 σ ∈ S K ∑ sign ( σ ) J ( α p ( v σ ( 1 ) , … ) , β p ( … ) , γ p ( … , v σ ( K ) ) ) . (24)
In particular, for D ∈ Ω 1 ( P ; A ) D\in\Omega^1(P;A) D ∈ Ω 1 ( P ; A ) and ϵ ∈ Ω 0 ( P ; A ) \epsilon\in\Omega^0(P;A) ϵ ∈ Ω 0 ( P ; A ) , evaluating (24) on v 1 , v 2 ∈ T p P v_1,v_2\in T_pP v 1 , v 2 ∈ T p P gives J ( D , D , ϵ ) p ( v 1 , v 2 ) = 2 J ( D p ( v 1 ) , D p ( v 2 ) , ϵ ( p ) ) J(D,D,\epsilon)_p(v_1,v_2)=2J(D_p(v_1),D_p(v_2),\epsilon(p)) J ( D , D , ϵ ) p ( v 1 , v 2 ) = 2 J ( D p ( v 1 ) , D p ( v 2 ) , ϵ ( p )) , while on v 1 , v 2 , v 3 ∈ T p P v_1,v_2,v_3\in T_pP v 1 , v 2 , v 3 ∈ T p P it gives J ( D , D , D ) p ( v 1 , v 2 , v 3 ) = 6 J ( D p ( v 1 ) , D p ( v 2 ) , D p ( v 3 ) ) J(D,D,D)_p(v_1,v_2,v_3)=6J(D_p(v_1),D_p(v_2),D_p(v_3)) J ( D , D , D ) p ( v 1 , v 2 , v 3 ) = 6 J ( D p ( v 1 ) , D p ( v 2 ) , D p ( v 3 )) .
Proposition 2.4 (Ehresmann Admissibility and the Lie Nucleus Constraint). Let π : P → M \pi:P\to M π : P → M be a principal G G G -bundle, let ρ : G → Aut A l g ( A ) \rho:G\to\operatorname{Aut}_{\mathrm{Alg}}(A) ρ : G → Aut Alg ( A ) be a representation of G G G by algebra automorphisms of ( A , ∗ ) (A,*) ( A , ∗ ) , and let η : g → A \eta:\mathfrak{g}\to A η : g → A be a G G G -equivariant linear map ( η ( Ad g X ) = ρ ( g ) ( η ( X ) ) (\eta(\operatorname{Ad}_gX)=\rho(g)(\eta(X)) ( η ( Ad g X ) = ρ ( g ) ( η ( X )) for all g ∈ G g\in G g ∈ G , X ∈ g ) X\in\mathfrak{g}) X ∈ g ) . Define the space of η \eta η -compatible Ehresmann A A A -connections on P P P by
Conn ρ , η ( P ; A ) : = { D ∈ Ω e q , ρ 1 ( P ; A ) ∣ D ( X ) = η ( X ) for all X ∈ g } . (25) \operatorname{Conn}_{\rho,\eta}(P;A):=\left\{D\in\Omega^1_{\mathrm{eq},\rho}(P;A)\mid D(X)=\eta(X)\ \text{for all }X\in\mathfrak{g}\right\}. \tag*{(25)} Conn ρ , η ( P ; A ) := { D ∈ Ω eq , ρ 1 ( P ; A ) ∣ D ( X ) = η ( X ) for all X ∈ g } . (25)
Assume that the intertwining condition
d ρ ( X ) ( y ) = 1 2 [ η ( X ) , y ] ∗ = 1 2 ( η ( X ) ∗ y − y ∗ η ( X ) ) ( ∀ X ∈ g , y ∈ A ) (26) d\rho(X)(y)=\frac{1}{2}[\eta(X),y]_*=\frac{1}{2}\left(\eta(X)*y-y*\eta(X)\right)\qquad(\forall X\in\mathfrak{g},\ y\in A) \tag*{(26)} d ρ ( X ) ( y ) = 2 1 [ η ( X ) , y ] ∗ = 2 1 ( η ( X ) ∗ y − y ∗ η ( X ) ) ( ∀ X ∈ g , y ∈ A ) (26)
holds. (Notice that when ( A , ∗ ) (A,*) ( A , ∗ ) is a Lie algebra ( g , [ ⋅ , ⋅ ] g ) (\mathfrak{g},[\cdot,\cdot]_{\mathfrak{g}}) ( g , [ ⋅ , ⋅ ] g ) with ∗ = [ ⋅ , ⋅ ] g *=[\cdot,\cdot]_{\mathfrak{g}} ∗ = [ ⋅ , ⋅ ] g and η = id g \eta=\operatorname{id}_{\mathfrak{g}} η = id g , the commutator of ∗ * ∗ is [ x , y ] ∗ = [ x , y ] g − [ y , x ] g = 2 [ x , y ] g [x,y]_*=[x,y]_{\mathfrak{g}}-[y,x]_{\mathfrak{g}}=2[x,y]_{\mathfrak{g}} [ x , y ] ∗ = [ x , y ] g − [ y , x ] g = 2 [ x , y ] g , so 1 2 [ η ( X ) , y ] ∗ = [ X , y ] g = ad X ( y ) \frac{1}{2}[\eta(X),y]_*=[X,y]_{\mathfrak{g}}=\operatorname{ad}_X(y) 2 1 [ η ( X ) , y ] ∗ = [ X , y ] g = ad X ( y ) recovers the standard adjoint representation.) Then:
(1) Lie Nucleus Constraint: Because d ρ ( X ) ∈ Der ( A ) d\rho(X)\in\operatorname{Der}(A) d ρ ( X ) ∈ Der ( A ) is a derivation of ( A , ∗ ) (A,*) ( A , ∗ ) and hence of ( A , [ ⋅ , ⋅ ] ∗ ) (A,[\cdot,\cdot]_*) ( A , [ ⋅ , ⋅ ] ∗ ) , condition (26) forces the image of η \eta η to lie in the Lie nucleus (4) of ( A , [ ⋅ , ⋅ ] ∗ ) (A,[\cdot,\cdot]_*) ( A , [ ⋅ , ⋅ ] ∗ ) :
η ( g ) ⊂ N L i e ( A ) . (27) \eta(\mathfrak{g})\subset\mathcal{N}_{\mathrm{Lie}}(A). \tag*{(27)} η ( g ) ⊂ N Lie ( A ) . (27)
More precisely, under (26) , 1 2 ad η ( X ) [ ⋅ , ⋅ ] ∗ \frac{1}{2}\operatorname{ad}^{[\cdot,\cdot]_*}_{\eta(X)} 2 1 ad η ( X ) [ ⋅ , ⋅ ] ∗ is a derivation of the commutator algebra ( A , [ ⋅ , ⋅ ] ∗ ) (A,[\cdot,\cdot]_*) ( A , [ ⋅ , ⋅ ] ∗ ) if and only if η ( X ) ∈ N L i e ( A ) \eta(X)\in\mathcal{N}_{\mathrm{Lie}}(A) η ( X ) ∈ N Lie ( A ) (which alone suffices for all horizontality assertions in (2) below), and is a derivation of the full algebra ( A , ∗ ) (A,\ast) ( A , ∗ ) if and only if η ( X ) \eta(X) η ( X ) also satisfies the Jordan-derivation identity [ η ( X ) , y ∘ ∗ z ] ∗ = [ η ( X ) , y ] ∗ ∘ ∗ z + y ∘ ∗ [ η ( X ) , z ] ∗ [\eta(X),y\mathbin{\circ_\ast} z]_\ast=[\eta(X),y]_\ast\mathbin{\circ_\ast} z+y\mathbin{\circ_\ast}[\eta(X),z]_\ast [ η ( X ) , y ∘ ∗ z ] ∗ = [ η ( X ) , y ] ∗ ∘ ∗ z + y ∘ ∗ [ η ( X ) , z ] ∗ for all y , z ∈ A y,z\in A y , z ∈ A (which holds automatically on N L i e ( A ) N_{\mathrm{Lie}}(A) N Lie ( A ) whenever ( A , ∗ ) (A,\ast) ( A , ∗ ) is skew-commutative or associative). In particular, if ( A , [ ⋅ , ⋅ ] ∗ ) (A,[\cdot,\cdot]_\ast) ( A , [ ⋅ , ⋅ ] ∗ ) has trivial Lie nucleus N L i e ( A ) = { 0 } N_{\mathrm{Lie}}(A)=\{0\} N Lie ( A ) = { 0 } (such as the 7-dimensional octonionic Malcev algebra Im ( O ) \operatorname{Im}(\mathbb{O}) Im ( O ) , where Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) = 6 Assoc O ( x , y , z ) \operatorname{Jac}_{[\cdot,\cdot]_\ast}(x,y,z)=6\operatorname{Assoc}_{\mathbb{O}}(x,y,z) Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) = 6 Assoc O ( x , y , z ) and N a s s o c ( O ) = R ⋅ 1 N_{\mathrm{assoc}}(\mathbb{O})=\mathbb{R}\cdot1 N assoc ( O ) = R ⋅ 1 , so N L i e ( Im ( O ) ) = Im ( O ) ∩ R ⋅ 1 = { 0 } N_{\mathrm{Lie}}(\operatorname{Im}(\mathbb{O}))=\operatorname{Im}(\mathbb{O})\cap\mathbb{R}\cdot1=\{0\} N Lie ( Im ( O )) = Im ( O ) ∩ R ⋅ 1 = { 0 } ), then (26) requires η = 0 \eta=0 η = 0 and d ρ = 0 d\rho=0 d ρ = 0 .
Definition of A-YMT Theories, Variational Equations, and the Jacobiator Anomaly
(2) Horizontality and Descent of Curvature, Covariant Derivative, and Jacobiator Forms: For every D ∈ Conn ρ , η ( P ; A ) D\in\operatorname{Conn}_{\rho,\eta}(P;A) D ∈ Conn ρ , η ( P ; A ) , every α ∈ Ω h e q , ρ k ( P ; A ) \alpha\in\Omega_{\mathrm{heq},\rho}^{k}(P;A) α ∈ Ω heq , ρ k ( P ; A ) , and every ϵ ∈ Ω h e q , ρ 0 ( P ; A ) \epsilon\in\Omega_{\mathrm{heq},\rho}^{0}(P;A) ϵ ∈ Ω heq , ρ 0 ( P ; A ) :
(a) The curvature F D = d D + 1 2 D ∧ ∗ D F_D=dD+\frac{1}{2}D\wedge_\ast D F D = d D + 2 1 D ∧ ∗ D lies in Ω h e q , ρ 2 ( P ; A ) \Omega_{\mathrm{heq},\rho}^{2}(P;A) Ω heq , ρ 2 ( P ; A ) and descends via j j j to j ( F D ) ∈ Ω 2 ( M ; E A ) j(F_D)\in\Omega^{2}(M;E_A) j ( F D ) ∈ Ω 2 ( M ; E A ) ;
(b) The commutator covariant exterior derivative d D [ ⋅ , ⋅ ] ∗ α d_D^{[\cdot,\cdot]_\ast}\alpha d D [ ⋅ , ⋅ ] ∗ α lies in Ω h e q , ρ k + 1 ( P ; A ) \Omega_{\mathrm{heq},\rho}^{k+1}(P;A) Ω heq , ρ k + 1 ( P ; A ) and descends via j j j to d j [ ⋅ , ⋅ ] ∗ j ( α ) : = j ( d D [ ⋅ , ⋅ ] ∗ α ) ∈ Ω k + 1 ( M ; E A ) d_j^{[\cdot,\cdot]_\ast}j(\alpha):=j(d_D^{[\cdot,\cdot]_\ast}\alpha)\in\Omega^{k+1}(M;E_A) d j [ ⋅ , ⋅ ] ∗ j ( α ) := j ( d D [ ⋅ , ⋅ ] ∗ α ) ∈ Ω k + 1 ( M ; E A ) ;
(c) By the Lie nucleus inclusion (27) , the Jacobiator forms Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) ∈ Ω h e q , ρ 3 ( P ; A ) \operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,D)\in\Omega_{\mathrm{heq},\rho}^{3}(P;A) Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) ∈ Ω heq , ρ 3 ( P ; A ) and Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ∈ Ω h e q , ρ 2 ( P ; A ) \operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,\epsilon)\in\Omega_{\mathrm{heq},\rho}^{2}(P;A) Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ∈ Ω heq , ρ 2 ( P ; A ) are horizontal and ρ \rho ρ -equivariant, i.e. lie in Ω h e q , ρ 3 ( P ; A ) \Omega_{\mathrm{heq},\rho}^{3}(P;A) Ω heq , ρ 3 ( P ; A ) and Ω h e q , ρ 2 ( P ; A ) \Omega_{\mathrm{heq},\rho}^{2}(P;A) Ω heq , ρ 2 ( P ; A ) , and descend via j j j to Ω 3 ( M ; E A ) \Omega^{3}(M;E_A) Ω 3 ( M ; E A ) and Ω 2 ( M ; E A ) \Omega^{2}(M;E_A) Ω 2 ( M ; E A ) .
(3) Affine vs. Linear Connection Regimes:
Affine Ehresmann Regime (η ≠ 0 \eta\ne0 η = 0 ): Choosing any principal connection ω 0 ∈ Conn ( P ; g ) \omega_0\in\operatorname{Conn}(P;\mathfrak{g}) ω 0 ∈ Conn ( P ; g ) , D 0 : = η ∘ ω 0 ∈ Conn ρ , η ( P ; A ) D_0:=\eta\circ\omega_0\in\operatorname{Conn}_{\rho,\eta}(P;A) D 0 := η ∘ ω 0 ∈ Conn ρ , η ( P ; A ) is a basepoint connection and Conn ρ , η ( P ; A ) = D 0 + Ω h e q , ρ 1 ( P ; A ) \operatorname{Conn}_{\rho,\eta}(P;A)=D_0+\Omega_{\mathrm{heq},\rho}^{1}(P;A) Conn ρ , η ( P ; A ) = D 0 + Ω heq , ρ 1 ( P ; A ) is a non-empty affine space modeled on Ω h e q , ρ 1 ( P ; A ) ≃ Ω 1 ( M ; E A ) \Omega_{\mathrm{heq},\rho}^{1}(P;A)\simeq\Omega^{1}(M;E_A) Ω heq , ρ 1 ( P ; A ) ≃ Ω 1 ( M ; E A ) .
Linear/Horizontal Regime (η = 0 \eta=0 η = 0 , d ρ = 0 d\rho=0 d ρ = 0 ): When G G G acts trivially on A A A (or on a trivial bundle P = M × G P=M\times G P = M × G relative to the flat background exterior derivative d d d ), η = 0 \eta=0 η = 0 satisfies (26) and (27) for every algebra ( A , ∗ ) (A,\ast) ( A , ∗ ) without any Lie-nucleus restriction, and Conn l i n ( P ; A ) : = Conn ρ , 0 ( P ; A ) = Ω h e q , ρ 1 ( P ; A ) ≃ Ω 1 ( M ; E A ) \operatorname{Conn}^{\mathrm{lin}}(P;A):=\operatorname{Conn}_{\rho,0}(P;A)=\Omega_{\mathrm{heq},\rho}^{1}(P;A)\simeq\Omega^{1}(M;E_A) Conn lin ( P ; A ) := Conn ρ , 0 ( P ; A ) = Ω heq , ρ 1 ( P ; A ) ≃ Ω 1 ( M ; E A ) is a real vector space.
Proof. For (1), since ρ ( G ) ⊂ Aut A l g ( A ) \rho(G)\subset\operatorname{Aut}_{\mathrm{Alg}}(A) ρ ( G ) ⊂ Aut Alg ( A ) , each d ρ ( X ) d\rho(X) d ρ ( X ) is a derivation of ( A , ∗ ) (A,\ast) ( A , ∗ ) and therefore a derivation of both [ y , z ] ∗ = y ∗ z − z ∗ y [y,z]_\ast=y\ast z-z\ast y [ y , z ] ∗ = y ∗ z − z ∗ y and y ∘ ∗ z = 1 2 ( y ∗ z + z ∗ y ) y\mathbin{\circ_\ast}z=\frac{1}{2}(y\ast z+z\ast y) y ∘ ∗ z = 2 1 ( y ∗ z + z ∗ y ) . On the other hand, expanding 1 2 [ η ( X ) , [ y , z ] ∗ ] ∗ − [ 1 2 [ η ( X ) , y ] ∗ , z ] ∗ − [ y , 1 2 [ η ( X ) , z ] ∗ ] ∗ \frac{1}{2}[\eta(X),[y,z]_\ast]_\ast-\left[\frac{1}{2}[\eta(X),y]_\ast,z\right]_\ast-\left[y,\frac{1}{2}[\eta(X),z]_\ast\right]_\ast 2 1 [ η ( X ) , [ y , z ] ∗ ] ∗ − [ 2 1 [ η ( X ) , y ] ∗ , z ] ∗ − [ y , 2 1 [ η ( X ) , z ] ∗ ] ∗ gives − 1 2 Jac [ ⋅ , ⋅ ] ∗ ( η ( X ) , y , z ) -\frac{1}{2}\operatorname{Jac}_{[\cdot,\cdot]_\ast}(\eta(X),y,z) − 2 1 Jac [ ⋅ , ⋅ ] ∗ ( η ( X ) , y , z ) , so 1 2 ad η ( X ) [ ⋅ , ⋅ ] ∗ \frac{1}{2}\operatorname{ad}_{\eta(X)}^{[\cdot,\cdot]_\ast} 2 1 ad η ( X ) [ ⋅ , ⋅ ] ∗ is a derivation of [ ⋅ , ⋅ ] ∗ [\cdot,\cdot]_\ast [ ⋅ , ⋅ ] ∗ if and only if Jac [ ⋅ , ⋅ ] ∗ ( η ( X ) , y , z ) = 0 \operatorname{Jac}_{[\cdot,\cdot]_\ast}(\eta(X),y,z)=0 Jac [ ⋅ , ⋅ ] ∗ ( η ( X ) , y , z ) = 0 for all y , z ∈ A y,z\in A y , z ∈ A , i.e., η ( X ) ∈ N L i e ( A ) \eta(X)\in N_{\mathrm{Lie}}(A) η ( X ) ∈ N Lie ( A ) (and is a derivation of the full product ∗ \ast ∗ if and only if 1 2 ad η ( X ) [ ⋅ , ⋅ ] ∗ \frac{1}{2}\operatorname{ad}_{\eta(X)}^{[\cdot,\cdot]_\ast} 2 1 ad η ( X ) [ ⋅ , ⋅ ] ∗ also acts as a derivation on y ∘ ∗ z y\mathbin{\circ_\ast}z y ∘ ∗ z , which is automatic on N L i e ( A ) N_{\mathrm{Lie}}(A) N Lie ( A ) when y ∘ ∗ z = 0 y\mathbin{\circ_\ast}z=0 y ∘ ∗ z = 0 or when ∗ \ast ∗ is associative).
For (2a), ρ \rho ρ -equivariance of D D D and ρ ( g ) ∈ Aut A l g ( A ) \rho(g)\in\operatorname{Aut}_{\mathrm{Alg}}(A) ρ ( g ) ∈ Aut Alg ( A ) immediately imply R g ∗ F D = ρ ( g − 1 ) ∘ F D R_g^\ast F_D=\rho(g^{-1})\circ F_D R g ∗ F D = ρ ( g − 1 ) ∘ F D . For horizontality, differentiating R exp ( t X ) ∗ D = ρ ( exp ( − t X ) ) ∘ D R_{\exp(tX)}^\ast D=\rho(\exp(-tX))\circ D R e x p ( tX ) ∗ D = ρ ( exp ( − tX )) ∘ D at t = 0 t=0 t = 0 via Cartan’s magic formula L X D = ι X d D + d ( ι X D ) \mathcal{L}_XD=\iota_XdD+d(\iota_XD) L X D = ι X d D + d ( ι X D ) and using ι X D = η ( X ) = const \iota_XD=\eta(X)=\operatorname{const} ι X D = η ( X ) = const gives ι X d D = − d ρ ( X ) ( D ) = − 1 2 [ η ( X ) , D ] ∗ \iota_XdD=-d\rho(X)(D)=-\frac{1}{2}[\eta(X),D]_\ast ι X d D = − d ρ ( X ) ( D ) = − 2 1 [ η ( X ) , D ] ∗ . Meanwhile, by (8) , ι X ( 1 2 D ∧ ∗ D ) = 1 4 ι X ( D ∧ [ ⋅ , ⋅ ] ∗ D ) = 1 2 [ ι X D , D ] ∗ = 1 2 [ η ( X ) , D ] ∗ \iota_X(\frac{1}{2}D\wedge_\ast D)=\frac{1}{4}\iota_X(D\wedge_{[\cdot,\cdot]_\ast}D)=\frac{1}{2}[\iota_XD,D]_\ast=\frac{1}{2}[\eta(X),D]_\ast ι X ( 2 1 D ∧ ∗ D ) = 4 1 ι X ( D ∧ [ ⋅ , ⋅ ] ∗ D ) = 2 1 [ ι X D , D ] ∗ = 2 1 [ η ( X ) , D ] ∗ . Summing the two terms gives ι X F D = 0 \iota_XF_D=0 ι X F D = 0 , so F D ∈ Ω h e q , ρ 2 ( P ; A ) F_D\in\Omega_{\mathrm{heq},\rho}^{2}(P;A) F D ∈ Ω heq , ρ 2 ( P ; A ) .
For (2b), if α ∈ Ω h e q , ρ k ( P ; A ) \alpha\in\Omega_{\mathrm{heq},\rho}^{k}(P;A) α ∈ Ω heq , ρ k ( P ; A ) , then ι X α = 0 \iota_X\alpha=0 ι X α = 0 and L X α = ι X d α = − d ρ ( X ) ( α ) = − 1 2 [ η ( X ) , α ] ∗ \mathcal{L}_X\alpha=\iota_Xd\alpha=-d\rho(X)(\alpha)=-\frac{1}{2}[\eta(X),\alpha]_\ast L X α = ι X d α = − d ρ ( X ) ( α ) = − 2 1 [ η ( X ) , α ] ∗ , while ι X ( 1 2 D ∧ [ ⋅ , ⋅ ] ∗ α ) = 1 2 [ η ( X ) , α ] ∗ − 1 2 D ∧ [ ⋅ , ⋅ ] ∗ ( ι X α ) = 1 2 [ η ( X ) , α ] ∗ \iota_X(\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}\alpha)=\frac{1}{2}[\eta(X),\alpha]_\ast-\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}(\iota_X\alpha)=\frac{1}{2}[\eta(X),\alpha]_\ast ι X ( 2 1 D ∧ [ ⋅ , ⋅ ] ∗ α ) = 2 1 [ η ( X ) , α ] ∗ − 2 1 D ∧ [ ⋅ , ⋅ ] ∗ ( ι X α ) = 2 1 [ η ( X ) , α ] ∗ , so ι X ( d D [ ⋅ , ⋅ ] ∗ α ) = 0 \iota_X(d_D^{[\cdot,\cdot]_\ast}\alpha)=0 ι X ( d D [ ⋅ , ⋅ ] ∗ α ) = 0 .
For (2c), for any X ∈ g X\in\mathfrak{g} X ∈ g and tangent vectors v 1 , v 2 ∈ T p P v_1,v_2\in T_pP v 1 , v 2 ∈ T p P , (24) and (27) give
( ι X Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) ) p ( v 1 , v 2 ) = 6 Jac [ ⋅ , ⋅ ] ∗ ( η ( X ) , D p ( v 1 ) , D p ( v 2 ) ) = 0 , (\iota_X\operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,D))_p(v_1,v_2)=6\operatorname{Jac}_{[\cdot,\cdot]_\ast}(\eta(X),D_p(v_1),D_p(v_2))=0, ( ι X Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) ) p ( v 1 , v 2 ) = 6 Jac [ ⋅ , ⋅ ] ∗ ( η ( X ) , D p ( v 1 ) , D p ( v 2 )) = 0 ,
and ( ι X Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ) p ( v 1 ) = 2 Jac [ ⋅ , ⋅ ] ∗ ( η ( X ) , D p ( v 1 ) , ϵ ( p ) ) = 0 (\iota_X\operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,\epsilon))_p(v_1)=2\operatorname{Jac}_{[\cdot,\cdot]_\ast}(\eta(X),D_p(v_1),\epsilon(p))=0 ( ι X Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ) p ( v 1 ) = 2 Jac [ ⋅ , ⋅ ] ∗ ( η ( X ) , D p ( v 1 ) , ϵ ( p )) = 0 because η ( X ) ∈ N L i e ( A ) \eta(X)\in N_{\mathrm{Lie}}(A) η ( X ) ∈ N Lie ( A ) . Since ρ ( G ) ⊂ Aut A l g ( A ) \rho(G)\subset\operatorname{Aut}_{\mathrm{Alg}}(A) ρ ( G ) ⊂ Aut Alg ( A ) preserves Jac [ ⋅ , ⋅ ] ∗ \operatorname{Jac}_{[\cdot,\cdot]_\ast} Jac [ ⋅ , ⋅ ] ∗ , both forms are ρ \rho ρ -equivariant and horizontal. Part (3) is immediate from ι X ( D 1 − D 2 ) = η ( X ) − η ( X ) = 0 \iota_X(D_1-D_2)=\eta(X)-\eta(X)=0 ι X ( D 1 − D 2 ) = η ( X ) − η ( X ) = 0 .
2.3. Definition of A-YMT Theories, Variational Equations, and the Jacobiator Anomaly. Throughout the remainder of the paper, we work in the geometric setting of Proposition 2.4: π : P → M \pi:P\to M π : P → M is a principal G G G -bundle over a smooth n n n -manifold M M M , ρ : G → Aut A l g ( A ) \rho:G\to\operatorname{Aut}_{\mathrm{Alg}}(A) ρ : G → Aut Alg ( A ) is an algebra automorphism representation with associated algebra bundle E A = P × ρ A → M E_A=P\times_\rho A\to M E A = P × ρ A → M and Weil isomorphism j : Ω h e q , ρ ∙ ( P ; A ) ⟶ ∼ Ω ∙ ( M ; E A ) j:\Omega^\bullet_{\mathrm{heq},\rho}(P;A)\overset{\sim}{\longrightarrow}\Omega^\bullet(M;E_A) j : Ω heq , ρ ∙ ( P ; A ) ⟶ ∼ Ω ∙ ( M ; E A ) , and Conn ρ , η ( P ; A ) \operatorname{Conn}_{\rho,\eta}(P;A) Conn ρ , η ( P ; A ) is the space of admissible A A A -connections satisfying (26) (including the linear/horizontal case η = 0 \eta=0 η = 0 ).
Let Dens ( M ) \operatorname{Dens}(M) Dens ( M ) denote the C ∞ ( M ) C^\infty(M) C ∞ ( M ) -module of smooth 1-densities on M M M (identified with Ω n ( M ; R ) \Omega^n(M;\mathbb{R}) Ω n ( M ; R ) when M M M is oriented). For any C ∞ ( M ) C^\infty(M) C ∞ ( M ) -module V V V , let V R V_{\mathbb{R}} V R denote the real vector space obtained by restriction of scalars along R ↪ C ∞ ( M ) \mathbb{R}\hookrightarrow C^\infty(M) R ↪ C ∞ ( M ) .
Definition 2.5 (A A A -Valued Yang-Mills-Type (A A A -YMT) Theory). Let E A = P × ρ A → M E_A=P\times_\rho A\to M E A = P × ρ A → M be an associated algebra bundle over a smooth n n n -manifold M M M .
An A A A -valued Yang-Mills-type (A A A -YMT) pairing on E A E_A E A is an R \mathbb{R} R -bilinear map
⟨ ⋅ , ⋅ ⟩ A : Ω 2 ( M ; E A ) R ⊗ R Ω 2 ( M ; E A ) R ⟶ Dens ( M ) R , (28) \langle\cdot,\cdot\rangle_A:\Omega^2(M;E_A)_{\mathbb{R}}\otimes_{\mathbb{R}}\Omega^2(M;E_A)_{\mathbb{R}}\longrightarrow\operatorname{Dens}(M)_{\mathbb{R}}, \tag*{(28)} ⟨ ⋅ , ⋅ ⟩ A : Ω 2 ( M ; E A ) R ⊗ R Ω 2 ( M ; E A ) R ⟶ Dens ( M ) R , (28) with symmetrization ⟨ Φ , Ψ ⟩ A , s y m : = 1 2 ( ⟨ Φ , Ψ ⟩ A + ⟨ Ψ , Φ ⟩ A ) \langle\Phi,\Psi\rangle_{A,\mathrm{sym}}:=\frac{1}{2}(\langle\Phi,\Psi\rangle_A+\langle\Psi,\Phi\rangle_A) ⟨ Φ , Ψ ⟩ A , sym := 2 1 (⟨ Φ , Ψ ⟩ A + ⟨ Ψ , Φ ⟩ A ) . We say that ⟨ ⋅ , ⋅ ⟩ A , s y m \langle\cdot,\cdot\rangle_{A,\mathrm{sym}} ⟨ ⋅ , ⋅ ⟩ A , sym is tensorial (or C ∞ ( M ) C^\infty(M) C ∞ ( M ) -bilinear) if ⟨ f Φ , g Ψ ⟩ A , s y m = f g ⟨ Φ , Ψ ⟩ A , s y m \langle f\Phi,g\Psi\rangle_{A,\mathrm{sym}}=fg\langle\Phi,\Psi\rangle_{A,\mathrm{sym}} ⟨ f Φ , g Ψ ⟩ A , sym = f g ⟨ Φ , Ψ ⟩ A , sym for all f , g ∈ C ∞ ( M ) f,g\in C^\infty(M) f , g ∈ C ∞ ( M ) (which implies support locality supp ( ⟨ Φ , Ψ ⟩ A , s y m ) ⊂ supp ( Φ ) ∩ supp ( Ψ ) \operatorname{supp}(\langle\Phi,\Psi\rangle_{A,\mathrm{sym}})\subset\operatorname{supp}(\Phi)\cap\operatorname{supp}(\Psi) supp (⟨ Φ , Ψ ⟩ A , sym ) ⊂ supp ( Φ ) ∩ supp ( Ψ ) and induces a smooth bundle bilinear form on Λ 2 T ∗ M ⊗ E A \Lambda^2T^*M\otimes E_A Λ 2 T ∗ M ⊗ E A ), pointwise ad [ ⋅ , ⋅ ] ∗ \operatorname{ad}_{[\cdot,\cdot]^*} ad [ ⋅ , ⋅ ] ∗ -invariant if ⟨ [ Φ , j ( ϵ ) ] ∗ , Ψ ⟩ A , s y m + ⟨ Φ , [ Ψ , j ( ϵ ) ] ∗ ⟩ A , s y m = 0 \langle[\Phi,j(\epsilon)]_*,\Psi\rangle_{A,\mathrm{sym}}+\langle\Phi,[\Psi,j(\epsilon)]_*\rangle_{A,\mathrm{sym}}=0 ⟨[ Φ , j ( ϵ ) ] ∗ , Ψ ⟩ A , sym + ⟨ Φ , [ Ψ , j ( ϵ ) ] ∗ ⟩ A , sym = 0 in Dens ( M ) \operatorname{Dens}(M) Dens ( M ) for all Φ , Ψ ∈ Ω 2 ( M ; E A ) \Phi,\Psi\in\Omega^2(M;E_A) Φ , Ψ ∈ Ω 2 ( M ; E A ) and ϵ ∈ Ω h e q , ρ 0 ( P ; A ) \epsilon\in\Omega^0_{\mathrm{heq},\rho}(P;A) ϵ ∈ Ω heq , ρ 0 ( P ; A ) , and non-degenerate if ∫ M ⟨ Φ , Ψ ⟩ A , s y m = 0 \int_M\langle\Phi,\Psi\rangle_{A,\mathrm{sym}}=0 ∫ M ⟨ Φ , Ψ ⟩ A , sym = 0 for all compactly supported Φ ∈ Ω c 2 ( M ; E A ) \Phi\in\Omega^2_c(M;E_A) Φ ∈ Ω c 2 ( M ; E A ) implies Ψ = 0 \Psi=0 Ψ = 0 .
Assuming M M M is compact (or restricting to connections with compactly supported curvature density), the A A A -YMT action functional associated with ⟨ ⋅ , ⋅ ⟩ A \langle\cdot,\cdot\rangle_A ⟨ ⋅ , ⋅ ⟩ A is S A : Conn ρ , η ( P ; A ) → R S^A:\operatorname{Conn}_{\rho,\eta}(P;A)\to\mathbb{R} S A : Conn ρ , η ( P ; A ) → R ,
S A [ D ] : = ∫ M ⟨ j ( F D ) , j ( F D ) ⟩ A . (29) S^A[D]:=\int_M\langle j(F_D),j(F_D)\rangle_A. \tag*{(29)} S A [ D ] := ∫ M ⟨ j ( F D ) , j ( F D ) ⟩ A . (29) We distinguish three classes of A A A -YMT theories according to the pairing ⟨ ⋅ , ⋅ ⟩ A \langle\cdot,\cdot\rangle_A ⟨ ⋅ , ⋅ ⟩ A :
General A A A -YMT Theory: ⟨ ⋅ , ⋅ ⟩ A \langle\cdot,\cdot\rangle_A ⟨ ⋅ , ⋅ ⟩ A is an arbitrary R \mathbb{R} R -bilinear map (28) .
Tensorial A A A -YMT Theory: ⟨ ⋅ , ⋅ ⟩ A \langle\cdot,\cdot\rangle_A ⟨ ⋅ , ⋅ ⟩ A is C ∞ ( M ) C^\infty(M) C ∞ ( M ) -bilinear, e.g., given a fiberwise symmetric bilinear form κ : E A × M E A → M × R \kappa:E_A\times_M E_A\to M\times\mathbb{R} κ : E A × M E A → M × R and a bundle Hodge-type morphism ⋆ M : Ω 2 ( M ; E A ) → Ω n − 2 ( M ; E A ) \star_M:\Omega^2(M;E_A)\to\Omega^{n-2}(M;E_A) ⋆ M : Ω 2 ( M ; E A ) → Ω n − 2 ( M ; E A ) on an oriented n n n -manifold M M M ,
⟨ Φ , Ψ ⟩ κ , ⋆ M : = κ ( Φ ∧ ⋆ M Ψ ) ∈ Ω n ( M ; R ) . (30) \langle\Phi,\Psi\rangle_{\kappa,\star_M}:=\kappa(\Phi\wedge\star_M\Psi)\in\Omega^n(M;\mathbb{R}). \tag*{(30)} ⟨ Φ , Ψ ⟩ κ , ⋆ M := κ ( Φ ∧ ⋆ M Ψ ) ∈ Ω n ( M ; R ) . (30) Multiplicative (or Internal-Trace) A A A -YMT Theory: M M M is oriented, E A E_A E A is equipped with a bundle trace functional tr ξ : E A → M × R \operatorname{tr}_{\xi}:E_A\to M\times\mathbb{R} tr ξ : E A → M × R induced by ξ ∈ Γ ( E A ∗ ) \xi\in\Gamma(E_A^*) ξ ∈ Γ ( E A ∗ ) , and ⋆ ^ : Ω h e q , ρ 2 ( P ; A ) → Ω h e q , ρ n − 2 ( P ; A ) \hat{\star}:\Omega^2_{\mathrm{heq},\rho}(P;A)\to\Omega^{n-2}_{\mathrm{heq},\rho}(P;A) ⋆ ^ : Ω heq , ρ 2 ( P ; A ) → Ω heq , ρ n − 2 ( P ; A ) is an internal constitutive operator such that the pairing is formed via the internal product ∧ ∗ \wedge_* ∧ ∗ of A A A followed by tr ξ \operatorname{tr}_{\xi} tr ξ :
S ⋆ ^ A [ D ] : = ∫ M tr ξ ( j ( F D ) ∧ ∗ ⋆ ^ F D ) . (31) S^A_{\hat{\star}}[D]:=\int_M\operatorname{tr}_{\xi}\bigl(j(F_D)\wedge_*\hat{\star}F_D\bigr). \tag*{(31)} S ⋆ ^ A [ D ] := ∫ M tr ξ ( j ( F D ) ∧ ∗ ⋆ ^ F D ) . (31)
Theorem 2.7 (Variational Field Equations, Infinitesimal Gauge Covariance, and the Jacobiator Bianchi Anomaly). Let π : P → M \pi:P\to M π : P → M be a principal G G G -bundle over a closed smooth n n n -manifold
M M M , let ρ : G → Aut A l g ( A ) \rho:G\to\operatorname{Aut}_{\mathrm{Alg}}(A) ρ : G → Aut Alg ( A ) and η : g → N L i e ( A ) \eta:\mathfrak{g}\to\mathcal{N}_{\mathrm{Lie}}(A) η : g → N Lie ( A ) satisfy (26) , and let D ∈ Conn ρ , η ( P ; A ) D\in\operatorname{Conn}_{\rho,\eta}(P;A) D ∈ Conn ρ , η ( P ; A ) be an admissible A A A -connection with curvature F D = d D + 1 2 D ∧ ∗ D = d D + 1 4 D ∧ [ ⋅ , ⋅ ] ∗ D ∈ Ω h e q , ρ 2 ( P ; A ) F_D=dD+\frac{1}{2}D\wedge_*D=dD+\frac{1}{4}D\wedge_{[\cdot,\cdot]_*}D\in\Omega^2_{\mathrm{heq},\rho}(P;A) F D = d D + 2 1 D ∧ ∗ D = d D + 4 1 D ∧ [ ⋅ , ⋅ ] ∗ D ∈ Ω heq , ρ 2 ( P ; A ) .
(1) First Variation of Curvature and Action: For any tangent variation D ˙ ∈ Ω h e q , ρ 1 ( P ; A ) ≃ Ω 1 ( M ; E A ) \dot{D}\in\Omega^1_{\mathrm{heq},\rho}(P;A)\simeq\Omega^1(M;E_A) D ˙ ∈ Ω heq , ρ 1 ( P ; A ) ≃ Ω 1 ( M ; E A ) , the directional derivative of the curvature is given for every algebra ( A , ∗ ) (A,*) ( A , ∗ ) (without assuming associativity or the Jacobi identity) by
δ D ˙ F D : = d d t ∣ t = 0 F D + t D ˙ = d D ˙ + 1 2 ( D ∧ ∗ D ˙ + D ˙ ∧ ∗ D ) = d D [ ⋅ , ⋅ ] ∗ D ˙ ∈ Ω h e q , ρ 2 ( P ; A ) . (33) \delta_{\dot{D}}F_D:=\left.\frac{d}{dt}\right|_{t=0}F_{D+t\dot{D}}=d\dot{D}+\frac{1}{2}(D\wedge_*\dot{D}+\dot{D}\wedge_*D)=d_D^{[\cdot,\cdot]_*}\dot{D}\in\Omega^2_{\mathrm{heq},\rho}(P;A). \tag*{(33)} δ D ˙ F D := d t d t = 0 F D + t D ˙ = d D ˙ + 2 1 ( D ∧ ∗ D ˙ + D ˙ ∧ ∗ D ) = d D [ ⋅ , ⋅ ] ∗ D ˙ ∈ Ω heq , ρ 2 ( P ; A ) . (33)
and the first variation of the general A-YMT action S A [ D ] S^A[D] S A [ D ] is
δ D ˙ S A [ D ] = 2 ∫ M ⟨ j ( F D ) , d D [ ⋅ , ⋅ ] ∗ j ( D ˙ ) ⟩ A , s y m . (34) \delta_{\dot{D}}S^A[D]=2\int_M\langle j(F_D),d_D^{[\cdot,\cdot]_*}j(\dot{D})\rangle_{A,\mathrm{sym}}. \tag*{(34)} δ D ˙ S A [ D ] = 2 ∫ M ⟨ j ( F D ) , d D [ ⋅ , ⋅ ] ∗ j ( D ˙ ) ⟩ A , sym . (34)
In particular, consider a tensorial A-YMT theory (30) on a closed oriented n n n -manifold M M M , where κ : E A × M E A → M × R \kappa:E_A\times_M E_A\to M\times\mathbb{R} κ : E A × M E A → M × R is fiberwise symmetric, non-degenerate, and ad [ ⋅ , ⋅ ] ∗ \operatorname{ad}_{[\cdot,\cdot]_*} ad [ ⋅ , ⋅ ] ∗ -invariant (κ ( [ x , y ] ∗ , z ) + κ ( y , [ x , z ] ∗ ) = 0 \kappa([x,y]_*,z)+\kappa(y,[x,z]_*)=0 κ ([ x , y ] ∗ , z ) + κ ( y , [ x , z ] ∗ ) = 0 ), and ⋆ M : Ω 2 ( M ; E A ) → Ω n − 2 ( M ; E A ) \star_M:\Omega^2(M;E_A)\to\Omega^{n-2}(M;E_A) ⋆ M : Ω 2 ( M ; E A ) → Ω n − 2 ( M ; E A ) satisfies κ ( Φ ∧ ⋆ M Ψ ) = κ ( Ψ ∧ ⋆ M Φ ) \kappa(\Phi\wedge\star_M\Psi)=\kappa(\Psi\wedge\star_M\Phi) κ ( Φ ∧ ⋆ M Ψ ) = κ ( Ψ ∧ ⋆ M Φ ) . Then D D D is a critical point of S A [ D ] S^A[D] S A [ D ] if and only if it satisfies the A-Yang-Mills Euler-Lagrange equation:
d ∇ D [ ⋅ , ⋅ ] ∗ ( ⋆ M j ( F D ) ) = j ( d D [ ⋅ , ⋅ ] ∗ ( j − 1 ( ⋆ M j ( F D ) ) ) ) = 0 in Ω n − 1 ( M ; E A ) . (35) d_{\nabla_D}^{[\cdot,\cdot]_*}(\star_Mj(F_D))=j\left(d_D^{[\cdot,\cdot]_*}\left(j^{-1}(\star_Mj(F_D))\right)\right)=0\quad\text{in }\Omega^{n-1}(M;E_A). \tag*{(35)} d ∇ D [ ⋅ , ⋅ ] ∗ ( ⋆ M j ( F D )) = j ( d D [ ⋅ , ⋅ ] ∗ ( j − 1 ( ⋆ M j ( F D )) ) ) = 0 in Ω n − 1 ( M ; E A ) . (35)
(2) Generalized Bianchi Identity and Jacobiator Obstruction: For every admissible connection D ∈ Conn ρ , η ( P ; A ) D\in\operatorname{Conn}_{\rho,\eta}(P;A) D ∈ Conn ρ , η ( P ; A ) , the covariant exterior derivative of the curvature F D F_D F D satisfies in Ω h e q , ρ 3 ( P ; A ) \Omega^3_{\mathrm{heq},\rho}(P;A) Ω heq , ρ 3 ( P ; A ) :
d D [ ⋅ , ⋅ ] ∗ F D = d F D + 1 2 D ∧ [ ⋅ , ⋅ ] ∗ F D = − 1 8 ( D ∧ [ ⋅ , ⋅ ] ∗ D ) ∧ [ ⋅ , ⋅ ] ∗ D = − 1 24 Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) . (36) d_D^{[\cdot,\cdot]_*}F_D=dF_D+\frac{1}{2}D\wedge_{[\cdot,\cdot]_*}F_D=-\frac{1}{8}(D\wedge_{[\cdot,\cdot]_*}D)\wedge_{[\cdot,\cdot]_*}D=-\frac{1}{24}\operatorname{Jac}_{[\cdot,\cdot]_*}(D,D,D). \tag*{(36)} d D [ ⋅ , ⋅ ] ∗ F D = d F D + 2 1 D ∧ [ ⋅ , ⋅ ] ∗ F D = − 8 1 ( D ∧ [ ⋅ , ⋅ ] ∗ D ) ∧ [ ⋅ , ⋅ ] ∗ D = − 24 1 Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) . (36)
Consequently, the classical Bianchi identity d ∇ D [ ⋅ , ⋅ ] ∗ j ( F D ) = 0 d_{\nabla_D}^{[\cdot,\cdot]_*}j(F_D)=0 d ∇ D [ ⋅ , ⋅ ] ∗ j ( F D ) = 0 holds identically for all connections D ∈ Conn ρ , η ( P ; A ) D\in\operatorname{Conn}_{\rho,\eta}(P;A) D ∈ Conn ρ , η ( P ; A ) on manifolds of dimension n ≥ 3 n\ge3 n ≥ 3 if and only if ( A , ∗ ) (A,*) ( A , ∗ ) is Lie-admissible (Jac [ ⋅ , ⋅ ] ∗ ≡ 0 \operatorname{Jac}_{[\cdot,\cdot]_*}\equiv0 Jac [ ⋅ , ⋅ ] ∗ ≡ 0 ).
(3) Infinitesimal Gauge Transformations, Principal vs. Extended A-Symmetries, and Covariance Anomaly: Given a 0 0 0 -form gauge parameter ϵ ∈ Ω h e q , ρ 0 ( P ; A ) ≃ Γ ( E A ) \epsilon\in\Omega^0_{\mathrm{heq},\rho}(P;A)\simeq\Gamma(E_A) ϵ ∈ Ω heq , ρ 0 ( P ; A ) ≃ Γ ( E A ) , define the infinitesimal gauge variation of D ∈ Conn ρ , η ( P ; A ) D\in\operatorname{Conn}_{\rho,\eta}(P;A) D ∈ Conn ρ , η ( P ; A ) by
δ ϵ D : = d D [ ⋅ , ⋅ ] ∗ ϵ = d ϵ + 1 2 [ D , ϵ ] ∗ ∈ Ω h e q , ρ 1 ( P ; A ) . (37) \delta_\epsilon D:=d_D^{[\cdot,\cdot]_*}\epsilon=d\epsilon+\frac{1}{2}[D,\epsilon]_*\in\Omega^1_{\mathrm{heq},\rho}(P;A). \tag*{(37)} δ ϵ D := d D [ ⋅ , ⋅ ] ∗ ϵ = d ϵ + 2 1 [ D , ϵ ] ∗ ∈ Ω heq , ρ 1 ( P ; A ) . (37)
Then D + t δ ϵ D ∈ Conn ρ , η ( P ; A ) D+t\delta_\epsilon D\in\operatorname{Conn}_{\rho,\eta}(P;A) D + t δ ϵ D ∈ Conn ρ , η ( P ; A ) , and the induced infinitesimal variation of the curvature F D F_D F D is
δ ϵ F D = d D [ ⋅ , ⋅ ] ∗ ( d D [ ⋅ , ⋅ ] ∗ ϵ ) = 1 2 [ F D , ϵ ] ∗ − 1 8 Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ∈ Ω h e q , ρ 2 ( P ; A ) . (38) \delta_\epsilon F_D=d_D^{[\cdot,\cdot]_*}(d_D^{[\cdot,\cdot]_*}\epsilon)=\frac{1}{2}[F_D,\epsilon]_*-\frac{1}{8}\operatorname{Jac}_{[\cdot,\cdot]_*}(D,D,\epsilon)\in\Omega^2_{\mathrm{heq},\rho}(P;A). \tag*{(38)} δ ϵ F D = d D [ ⋅ , ⋅ ] ∗ ( d D [ ⋅ , ⋅ ] ∗ ϵ ) = 2 1 [ F D , ϵ ] ∗ − 8 1 Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ∈ Ω heq , ρ 2 ( P ; A ) . (38)
Consequently, we have a sharp dichotomy between principal G G G -gauge transformations and extended A A A -gauge transformations:
(a) Principal G G G -Gauge Transformations (ϵ ∈ Ω h e q , ρ 0 ( P ; η ( g ) ) ⊂ Ω h e q , ρ 0 ( P ; N L i e ( A ) ) \epsilon\in\Omega^0_{\mathrm{heq},\rho}(P;\eta(\mathfrak{g}))\subset\Omega^0_{\mathrm{heq},\rho}(P;\mathcal{N}_{\mathrm{Lie}}(A)) ϵ ∈ Ω heq , ρ 0 ( P ; η ( g )) ⊂ Ω heq , ρ 0 ( P ; N Lie ( A )) ): For every gauge parameter ϵ \epsilon ϵ valued in the Lie nucleus N L i e ( A ) \mathcal{N}_{\mathrm{Lie}}(A) N Lie ( A ) (and in particular for every vertical principal-bundle automorphism ϵ = η ( ϵ ~ ) \epsilon=\eta(\widetilde{\epsilon}) ϵ = η ( ϵ ) with ϵ ~ ∈ Ω h e q , Ad 0 ( P ; g ) \widetilde{\epsilon}\in\Omega^0_{\mathrm{heq},\operatorname{Ad}}(P;\mathfrak{g}) ϵ ∈ Ω heq , Ad 0 ( P ; g ) ), Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ≡ 0 \operatorname{Jac}_{[\cdot,\cdot]_*}(D,D,\epsilon)\equiv0 Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ≡ 0 vanishes identically and F D F_D F D transforms homogeneously (δ ϵ F D = 1 2 [ F D , ϵ ] ∗ \delta_\epsilon F_D=\frac{1}{2}[F_D,\epsilon]_* δ ϵ F D = 2 1 [ F D , ϵ ] ∗ ) for every algebra ( A , ∗ ) (A,*) ( A , ∗ ) .
(b) Extended A A A -Gauge Transformations (ϵ ∈ Ω h e q , ρ 0 ( P ; A ) \epsilon\in\Omega^0_{\mathrm{heq},\rho}(P;A) ϵ ∈ Ω heq , ρ 0 ( P ; A ) ): F D F_D F D transforms homogeneously for all ( D , ϵ ) ∈ Conn ρ , η ( P ; A ) × Ω h e q , ρ 0 ( P ; A ) (D,\epsilon)\in\operatorname{Conn}_{\rho,\eta}(P;A)\times\Omega^0_{\mathrm{heq},\rho}(P;A) ( D , ϵ ) ∈ Conn ρ , η ( P ; A ) × Ω heq , ρ 0 ( P ; A ) on manifolds of dimension n ≥ 2 n\ge2 n ≥ 2 if and only if ( A , ∗ ) (A,*) ( A , ∗ ) is Lie-admissible (N L i e ( A ) = A \mathcal{N}_{\mathrm{Lie}}(A)=A N Lie ( A ) = A ). Furthermore, if ⟨ ⋅ , ⋅ ⟩ A , s y m \langle\cdot,\cdot\rangle_{A,\mathrm{sym}} ⟨ ⋅ , ⋅ ⟩ A , sym is pointwise ad [ ⋅ , ⋅ ] ∗ \operatorname{ad}_{[\cdot,\cdot]_*} ad [ ⋅ , ⋅ ] ∗ -invariant, then
δ ϵ S A [ D ] = − 1 4 ∫ M ⟨ j ( F D ) , j ( Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ) ⟩ A , s y m , (39) \delta_\epsilon S^A[D]=-\frac{1}{4}\int_M\langle j(F_D),j(\operatorname{Jac}_{[\cdot,\cdot]_*}(D,D,\epsilon))\rangle_{A,\mathrm{sym}}, \tag*{(39)} δ ϵ S A [ D ] = − 4 1 ∫ M ⟨ j ( F D ) , j ( Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ )) ⟩ A , sym , (39)
which vanishes for all ( D , ϵ ) (D,\epsilon) ( D , ϵ ) whenever ( A , ∗ ) (A,\ast) ( A , ∗ ) is Lie-admissible, and conversely forces Jac [ ⋅ , ⋅ ] ∗ ≡ 0 \operatorname{Jac}_{[\cdot,\cdot]_\ast}\equiv0 Jac [ ⋅ , ⋅ ] ∗ ≡ 0 on A A A whenever ⟨ ⋅ , ⋅ ⟩ A , s y m \langle\cdot,\cdot\rangle_{A,\mathrm{sym}} ⟨ ⋅ , ⋅ ⟩ A , sym is tensorial (C ∞ ( M ) C^\infty(M) C ∞ ( M ) -bilinear), pointwise ad [ ⋅ , ⋅ ] ∗ \operatorname{ad}_{[\cdot,\cdot]_\ast} ad [ ⋅ , ⋅ ] ∗ -invariant, and non-degenerate and n ≥ 2 n\ge2 n ≥ 2 .
Proof. By Proposition 2.4(2), F D F_D F D , d D [ ⋅ , ⋅ ] ∗ D d_D^{[\cdot,\cdot]_\ast}D d D [ ⋅ , ⋅ ] ∗ D , Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) \operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,D) Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) , δ ϵ D \delta_\epsilon D δ ϵ D , and Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) \operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,\epsilon) Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) all lie in Ω h e q , ρ n ( P ; A ) \Omega^n_{\mathrm{heq},\rho}(P;A) Ω heq , ρ n ( P ; A ) and descend via j j j to Ω n ( M ; E A ) \Omega^n(M;E_A) Ω n ( M ; E A ) .
For part (1), expanding F D + t D ˙ F_{D+t\dot D} F D + t D ˙ to first order in t t t and using D ∧ ∗ D ˙ + D ˙ ∧ ∗ D = D ∧ [ ⋅ , ⋅ ] ∗ D ˙ D\wedge_\ast\dot D+\dot D\wedge_\ast D=D\wedge_{[\cdot,\cdot]_\ast}\dot D D ∧ ∗ D ˙ + D ˙ ∧ ∗ D = D ∧ [ ⋅ , ⋅ ] ∗ D ˙ gives (33) and (34) . For the tensorial pairing ⟨ Φ , Ψ ⟩ A = κ ( Φ ∧ ⋆ M Ψ ) \langle\Phi,\Psi\rangle_A=\kappa(\Phi\wedge\star_M\Psi) ⟨ Φ , Ψ ⟩ A = κ ( Φ ∧ ⋆ M Ψ ) , let Ξ : = j − 1 ( ⋆ M j ( F D ) ) ∈ Ω h e q , ρ n − 2 ( P ; A ) \Xi:=j^{-1}(\star_Mj(F_D))\in\Omega^{n-2}_{\mathrm{heq},\rho}(P;A) Ξ := j − 1 ( ⋆ M j ( F D )) ∈ Ω heq , ρ n − 2 ( P ; A ) . On P P P , the graded Leibniz rule for the 1 1 1 -form D ˙ \dot D D ˙ gives d ( κ ( D ˙ ∧ Ξ ) ) = κ ( d D ˙ ∧ Ξ ) − κ ( D ˙ ∧ d Ξ ) d(\kappa(\dot D\wedge\Xi))=\kappa(d\dot D\wedge\Xi)-\kappa(\dot D\wedge d\Xi) d ( κ ( D ˙ ∧ Ξ )) = κ ( d D ˙ ∧ Ξ ) − κ ( D ˙ ∧ d Ξ ) . On decomposable elements D = x ⊗ α D=x\otimes\alpha D = x ⊗ α , D ˙ = y ⊗ β \dot D=y\otimes\beta D ˙ = y ⊗ β (α , β ∈ Ω 1 ( P ) \alpha,\beta\in\Omega^1(P) α , β ∈ Ω 1 ( P ) ) and Ξ = z ⊗ γ \Xi=z\otimes\gamma Ξ = z ⊗ γ (γ ∈ Ω n − 2 ( P ) \gamma\in\Omega^{n-2}(P) γ ∈ Ω n − 2 ( P ) ), the Koszul sign β ∧ α ∧ γ = − α ∧ β ∧ γ \beta\wedge\alpha\wedge\gamma=-\alpha\wedge\beta\wedge\gamma β ∧ α ∧ γ = − α ∧ β ∧ γ combines with ad [ ⋅ , ⋅ ] ∗ \operatorname{ad}_{[\cdot,\cdot]_\ast} ad [ ⋅ , ⋅ ] ∗ -invariance κ ( [ x , y ] ∗ , z ) = − κ ( y , [ x , z ] ∗ ) \kappa([x,y]_\ast,z)=-\kappa(y,[x,z]_\ast) κ ([ x , y ] ∗ , z ) = − κ ( y , [ x , z ] ∗ ) to give
κ ( ( D ∧ [ ⋅ , ⋅ ] ∗ D ˙ ) ∧ Ξ ) = κ ( [ x , y ] ∗ , z ) α ∧ β ∧ γ = + κ ( y , [ x , z ] ∗ ) β ∧ α ∧ γ = + κ ( D ˙ ∧ ( D ∧ [ ⋅ , ⋅ ] ∗ Ξ ) ) . \kappa((D\wedge_{[\cdot,\cdot]_\ast}\dot D)\wedge\Xi)=\kappa([x,y]_\ast,z)\,\alpha\wedge\beta\wedge\gamma=+\kappa(y,[x,z]_\ast)\,\beta\wedge\alpha\wedge\gamma=+\kappa(\dot D\wedge(D\wedge_{[\cdot,\cdot]_\ast}\Xi)). κ (( D ∧ [ ⋅ , ⋅ ] ∗ D ˙ ) ∧ Ξ ) = κ ([ x , y ] ∗ , z ) α ∧ β ∧ γ = + κ ( y , [ x , z ] ∗ ) β ∧ α ∧ γ = + κ ( D ˙ ∧ ( D ∧ [ ⋅ , ⋅ ] ∗ Ξ )) .
Applying Stokes’ theorem on the closed manifold M M M yields δ D ˙ S A [ D ] = 2 ∫ M κ ( j ( D ˙ ) ∧ j ( ⋆ M j ( F D ) ) ) \delta_{\dot D}S^A[D]=2\int_M\kappa(j(\dot D)\wedge j(\star_Mj(F_D))) δ D ˙ S A [ D ] = 2 ∫ M κ ( j ( D ˙ ) ∧ j ( ⋆ M j ( F D ))) , which vanishes for all D ˙ ∈ Ω h e q , ρ 1 ( P ; A ) \dot D\in\Omega^1_{\mathrm{heq},\rho}(P;A) D ˙ ∈ Ω heq , ρ 1 ( P ; A ) if and only if (35) holds.
For part (2), since d 2 D = 0 d^2D=0 d 2 D = 0 and d ( D ∧ [ ⋅ , ⋅ ] ∗ D ) = 2 d D ∧ [ ⋅ , ⋅ ] ∗ D = − 2 D ∧ [ ⋅ , ⋅ ] ∗ d D d(D\wedge_{[\cdot,\cdot]_\ast}D)=2dD\wedge_{[\cdot,\cdot]_\ast}D=-2D\wedge_{[\cdot,\cdot]_\ast}dD d ( D ∧ [ ⋅ , ⋅ ] ∗ D ) = 2 d D ∧ [ ⋅ , ⋅ ] ∗ D = − 2 D ∧ [ ⋅ , ⋅ ] ∗ d D , we compute:
d D [ ⋅ , ⋅ ] ∗ F D = d ( d D + 1 4 D ∧ [ ⋅ , ⋅ ] ∗ D ) + 1 2 D ∧ [ ⋅ , ⋅ ] ∗ ( d D + 1 4 D ∧ [ ⋅ , ⋅ ] ∗ D ) = − 1 2 D ∧ [ ⋅ , ⋅ ] ∗ d D + 1 2 D ∧ [ ⋅ , ⋅ ] ∗ d D + 1 8 D ∧ [ ⋅ , ⋅ ] ∗ ( D ∧ [ ⋅ , ⋅ ] ∗ D ) = − 1 8 ( D ∧ [ ⋅ , ⋅ ] ∗ D ) ∧ [ ⋅ , ⋅ ] ∗ D . \begin{aligned}
d_D^{[\cdot,\cdot]_\ast}F_D=d\left(dD+\frac{1}{4}D\wedge_{[\cdot,\cdot]_\ast}D\right)+\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}\left(dD+\frac{1}{4}D\wedge_{[\cdot,\cdot]_\ast}D\right) \\
&=-\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}dD+\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}dD+\frac{1}{8}D\wedge_{[\cdot,\cdot]_\ast}(D\wedge_{[\cdot,\cdot]_\ast}D)=-\frac{1}{8}(D\wedge_{[\cdot,\cdot]_\ast}D)\wedge_{[\cdot,\cdot]_\ast}D.
\end{aligned} d D [ ⋅ , ⋅ ] ∗ F D = d ( d D + 4 1 D ∧ [ ⋅ , ⋅ ] ∗ D ) + 2 1 D ∧ [ ⋅ , ⋅ ] ∗ ( d D + 4 1 D ∧ [ ⋅ , ⋅ ] ∗ D ) = − 2 1 D ∧ [ ⋅ , ⋅ ] ∗ d D + 2 1 D ∧ [ ⋅ , ⋅ ] ∗ d D + 8 1 D ∧ [ ⋅ , ⋅ ] ∗ ( D ∧ [ ⋅ , ⋅ ] ∗ D ) = − 8 1 ( D ∧ [ ⋅ , ⋅ ] ∗ D ) ∧ [ ⋅ , ⋅ ] ∗ D .
Since ( D ∧ [ ⋅ , ⋅ ] ∗ D ) p ( u , w ) = 2 [ D p ( u ) , D p ( w ) ] ∗ (D\wedge_{[\cdot,\cdot]_\ast}D)_p(u,w)=2[D_p(u),D_p(w)]_\ast ( D ∧ [ ⋅ , ⋅ ] ∗ D ) p ( u , w ) = 2 [ D p ( u ) , D p ( w ) ] ∗ , evaluating the 3 3 3 -form ( ( D ∧ [ ⋅ , ⋅ ] ∗ D ) ∧ [ ⋅ , ⋅ ] ∗ D ) ( v 1 , v 2 , v 3 ) ((D\wedge_{[\cdot,\cdot]_\ast}D)\wedge_{[\cdot,\cdot]_\ast}D)(v_1,v_2,v_3) (( D ∧ [ ⋅ , ⋅ ] ∗ D ) ∧ [ ⋅ , ⋅ ] ∗ D ) ( v 1 , v 2 , v 3 ) on tangent vectors v 1 , v 2 , v 3 ∈ T p P v_1,v_2,v_3\in T_pP v 1 , v 2 , v 3 ∈ T p P via (6) sums the terms 2 [ [ D ( v σ ( 1 ) ) , D ( v σ ( 2 ) ) ] ∗ , D ( v σ ( 3 ) ) ] ∗ 2[[D(v_{\sigma(1)}),D(v_{\sigma(2)})]_\ast,D(v_{\sigma(3)})]_\ast 2 [[ D ( v σ ( 1 ) ) , D ( v σ ( 2 ) ) ] ∗ , D ( v σ ( 3 ) ) ] ∗ with prefactor 1 2 ! 1 ! \frac{1}{2!1!} 2 ! 1 ! 1 over σ ∈ S 3 \sigma\in S_3 σ ∈ S 3 . Grouping the 6 6 6 permutations into the 3 3 3 cyclic pairs ( 123 ) , ( 231 ) , ( 312 ) (123),(231),(312) ( 123 ) , ( 231 ) , ( 312 ) gives 2 Jac [ ⋅ , ⋅ ] ∗ ( D ( v 1 ) , D ( v 2 ) , D ( v 3 ) ) = 1 3 Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) ( v 1 , v 2 , v 3 ) 2\operatorname{Jac}_{[\cdot,\cdot]_\ast}(D(v_1),D(v_2),D(v_3))=\frac{1}{3}\operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,D)(v_1,v_2,v_3) 2 Jac [ ⋅ , ⋅ ] ∗ ( D ( v 1 ) , D ( v 2 ) , D ( v 3 )) = 3 1 Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) ( v 1 , v 2 , v 3 ) by (24) , proving (36) .
When n ≥ 3 n\ge3 n ≥ 3 , choosing a local trivialization of P P P over an open chart U ⊂ M U\subset M U ⊂ M around x ∈ M x\in M x ∈ M and evaluating (36) on horizontal tangent vectors v 1 , v 2 , v 3 ∈ T p P v_1,v_2,v_3\in T_pP v 1 , v 2 , v 3 ∈ T p P projecting to linearly independent vectors in T x M T_xM T x M with D p ( v i ) = y i ∈ A D_p(v_i)=y_i\in A D p ( v i ) = y i ∈ A arbitrary shows that Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) = 0 \operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,D)=0 Jac [ ⋅ , ⋅ ] ∗ ( D , D , D ) = 0 for all D D D if and only if Jac [ ⋅ , ⋅ ] ∗ ( y 1 , y 2 , y 3 ) = 0 \operatorname{Jac}_{[\cdot,\cdot]_\ast}(y_1,y_2,y_3)=0 Jac [ ⋅ , ⋅ ] ∗ ( y 1 , y 2 , y 3 ) = 0 for all y 1 , y 2 , y 3 ∈ A y_1,y_2,y_3\in A y 1 , y 2 , y 3 ∈ A .
For part (3), substituting D ˙ = δ ϵ D = d ϵ + 1 2 [ D , ϵ ] ∗ \dot D=\delta_\epsilon D=d\epsilon+\frac{1}{2}[D,\epsilon]_\ast D ˙ = δ ϵ D = d ϵ + 2 1 [ D , ϵ ] ∗ into (33) and using the graded Leibniz identity d ( 1 2 [ D , ϵ ] ∗ ) = 1 2 [ d D , ϵ ] ∗ − 1 2 D ∧ [ ⋅ , ⋅ ] ∗ d ϵ d(\frac{1}{2}[D,\epsilon]_\ast)=\frac{1}{2}[dD,\epsilon]_\ast-\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}d\epsilon d ( 2 1 [ D , ϵ ] ∗ ) = 2 1 [ d D , ϵ ] ∗ − 2 1 D ∧ [ ⋅ , ⋅ ] ∗ d ϵ (since D D D has degree 1 1 1 and ϵ \epsilon ϵ has degree 0 0 0 ) gives
δ ϵ F D = d ( d ϵ + 1 2 [ D , ϵ ] ∗ ) + 1 2 D ∧ [ ⋅ , ⋅ ] ∗ ( d ϵ + 1 2 [ D , ϵ ] ∗ ) = 1 2 [ d D , ϵ ] ∗ − 1 2 D ∧ [ ⋅ , ⋅ ] ∗ d ϵ + 1 2 D ∧ [ ⋅ , ⋅ ] ∗ d ϵ + 1 4 D ∧ [ ⋅ , ⋅ ] ∗ [ D , ϵ ] ∗ = 1 2 [ F D − 1 4 D ∧ [ ⋅ , ⋅ ] ∗ D , ϵ ] ∗ + 1 4 D ∧ [ ⋅ , ⋅ ] ∗ [ D , ϵ ] ∗ . \begin{aligned}
\delta_\epsilon F_D=d\left(d\epsilon+\frac{1}{2}[D,\epsilon]_\ast\right)+\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}\left(d\epsilon+\frac{1}{2}[D,\epsilon]_\ast\right) \\
&=\frac{1}{2}[dD,\epsilon]_\ast-\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}d\epsilon+\frac{1}{2}D\wedge_{[\cdot,\cdot]_\ast}d\epsilon+\frac{1}{4}D\wedge_{[\cdot,\cdot]_\ast}[D,\epsilon]_\ast\\
&=\frac{1}{2}\left[F_D-\frac{1}{4}D\wedge_{[\cdot,\cdot]_\ast}D,\epsilon\right]_\ast+\frac{1}{4}D\wedge_{[\cdot,\cdot]_\ast}[D,\epsilon]_\ast.
\end{aligned} δ ϵ F D = d ( d ϵ + 2 1 [ D , ϵ ] ∗ ) + 2 1 D ∧ [ ⋅ , ⋅ ] ∗ ( d ϵ + 2 1 [ D , ϵ ] ∗ ) = 2 1 [ d D , ϵ ] ∗ − 2 1 D ∧ [ ⋅ , ⋅ ] ∗ d ϵ + 2 1 D ∧ [ ⋅ , ⋅ ] ∗ d ϵ + 4 1 D ∧ [ ⋅ , ⋅ ] ∗ [ D , ϵ ] ∗ = 2 1 [ F D − 4 1 D ∧ [ ⋅ , ⋅ ] ∗ D , ϵ ] ∗ + 4 1 D ∧ [ ⋅ , ⋅ ] ∗ [ D , ϵ ] ∗ .
Evaluating 1 4 D ∧ [ ⋅ , ⋅ ] ∗ [ D , ϵ ] ∗ − 1 8 [ D ∧ [ ⋅ , ⋅ ] ∗ D , ϵ ] ∗ \frac{1}{4}D\wedge_{[\cdot,\cdot]_\ast}[D,\epsilon]_\ast-\frac{1}{8}[D\wedge_{[\cdot,\cdot]_\ast}D,\epsilon]_\ast 4 1 D ∧ [ ⋅ , ⋅ ] ∗ [ D , ϵ ] ∗ − 8 1 [ D ∧ [ ⋅ , ⋅ ] ∗ D , ϵ ] ∗ on ( v 1 , v 2 ) (v_1,v_2) ( v 1 , v 2 ) gives − 1 4 Jac [ ⋅ , ⋅ ] ∗ ( D ( v 1 ) , D ( v 2 ) , ϵ ) = − 1 8 Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ( v 1 , v 2 ) -\frac{1}{4}\operatorname{Jac}_{[\cdot,\cdot]_\ast}(D(v_1),D(v_2),\epsilon)=-\frac{1}{8}\operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,\epsilon)(v_1,v_2) − 4 1 Jac [ ⋅ , ⋅ ] ∗ ( D ( v 1 ) , D ( v 2 ) , ϵ ) = − 8 1 Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ( v 1 , v 2 ) by (24) , proving (38) . Whenever ϵ ∈ Ω h e q , ρ 0 ( P ; N L i e ( A ) ) \epsilon\in\Omega^0_{\mathrm{heq},\rho}(P;\mathcal{N}_{\mathrm{Lie}}(A)) ϵ ∈ Ω heq , ρ 0 ( P ; N Lie ( A )) , equation (4) yields Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) = 0 \operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,\epsilon)=0 Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) = 0 , proving (3a). In general, substituting (38) into δ ϵ S A [ D ] = 2 ∫ M ⟨ j ( F D ) , j ( δ ϵ F D ) ⟩ A , s y m \delta_\epsilon S^A[D]=2\int_M\langle j(F_D),j(\delta_\epsilon F_D)\rangle_{A,\mathrm{sym}} δ ϵ S A [ D ] = 2 ∫ M ⟨ j ( F D ) , j ( δ ϵ F D ) ⟩ A , sym and using pointwise ad [ ⋅ , ⋅ ] ∗ \operatorname{ad}_{[\cdot,\cdot]_\ast} ad [ ⋅ , ⋅ ] ∗ -invariance ⟨ [ j ( F D ) , j ( ϵ ) ] ∗ , j ( F D ) ⟩ A , s y m = 0 \langle[j(F_D),j(\epsilon)]_\ast,j(F_D)\rangle_{A,\mathrm{sym}}=0 ⟨[ j ( F D ) , j ( ϵ ) ] ∗ , j ( F D ) ⟩ A , sym = 0 yields (39) .
Finally, to prove the converse in (3b) rigorously when ⟨ ⋅ , ⋅ ⟩ A , s y m \langle\cdot,\cdot\rangle_{A,\mathrm{sym}} ⟨ ⋅ , ⋅ ⟩ A , sym is tensorial (C ∞ ( M ) C^\infty(M) C ∞ ( M ) -bilinear), pointwise ad [ ⋅ , ⋅ ] ∗ \operatorname{ad}_{[\cdot,\cdot]_\ast} ad [ ⋅ , ⋅ ] ∗ -invariant, and non-degenerate and n ≥ 2 n\ge2 n ≥ 2 , suppose (39) vanishes identically for all D ∈ Conn ρ , η ( P ; A ) D\in\operatorname{Conn}_{\rho,\eta}(P;A) D ∈ Conn ρ , η ( P ; A ) and ϵ ∈ Ω h e q , ρ 0 ( P ; A ) \epsilon\in\Omega^0_{\mathrm{heq},\rho}(P;A) ϵ ∈ Ω heq , ρ 0 ( P ; A ) . Because ⟨ ⋅ , ⋅ ⟩ A , s y m \langle\cdot,\cdot\rangle_{A,\mathrm{sym}} ⟨ ⋅ , ⋅ ⟩ A , sym is C ∞ ( M ) C^\infty(M) C ∞ ( M ) -bilinear and non-degenerate, it is given by a fiberwise non-degenerate symmetric bilinear bundle form ( ⋅ , ⋅ ) A , x : ( Λ 2 T x ∗ M ⊗ ( E A ) x ) × ( Λ 2 T x ∗ M ⊗ ( E A ) x ) → Dens x ( M ) (\cdot,\cdot)_{A,x}:(\Lambda^2T_x^\ast M\otimes(E_A)_x)\times(\Lambda^2T_x^\ast M\otimes(E_A)_x)\to\operatorname{Dens}_x(M) ( ⋅ , ⋅ ) A , x : ( Λ 2 T x ∗ M ⊗ ( E A ) x ) × ( Λ 2 T x ∗ M ⊗ ( E A ) x ) → Dens x ( M ) at every point x ∈ M x\in M x ∈ M . Replacing ϵ \epsilon ϵ by f ϵ f\epsilon f ϵ for arbitrary f ∈ C ∞ ( M ) f\in C^\infty(M) f ∈ C ∞ ( M ) and using C ∞ ( M ) C^\infty(M) C ∞ ( M ) -linearity Jac [ ⋅ , ⋅ ] ∗ ( D , D , f ϵ ) = f Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) \operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,f\epsilon)=f\operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,\epsilon) Jac [ ⋅ , ⋅ ] ∗ ( D , D , f ϵ ) = f Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) implies that the pointwise pairing ( j ( F D ) x 0 , j ( Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ) x 0 ) A , x 0 = 0 (j(F_D)_{x_0},j(\operatorname{Jac}_{[\cdot,\cdot]_\ast}(D,D,\epsilon))_{x_0})_{A,x_0}=0 ( j ( F D ) x 0 , j ( Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ) x 0 ) A , x 0 = 0 vanishes identically at every point x 0 ∈ M x_0\in M x 0 ∈ M for all D ∈ Conn ρ , η ( P ; A ) D\in\operatorname{Conn}_{\rho,\eta}(P;A) D ∈ Conn ρ , η ( P ; A ) and all ϵ ∈ Ω h e q , ρ 0 ( P ; A ) \epsilon\in\Omega^0_{\mathrm{heq},\rho}(P;A) ϵ ∈ Ω heq , ρ 0 ( P ; A ) . Now fix x 0 ∈ M x_0\in M x 0 ∈ M , a coordinate chart ( U , x 1 , … , x n ) (U,x^1,\ldots,x^n) ( U , x 1 , … , x n ) centered at x 0 x_0 x 0 (x μ ( x 0 ) = 0 x^\mu(x_0)=0 x μ ( x 0 ) = 0 ), a basepoint connection D 0 = D_0= D 0 = η ∘ ω 0 ∈ Conn ρ , η ( P ; A ) \eta\circ\omega_0 \in\operatorname{Conn}_{\rho,\eta}(P; A) η ∘ ω 0 ∈ Conn ρ , η ( P ; A ) , and a local section s 0 : U → P s_0 : U \to P s 0 : U → P chosen in synchronous gauge at x 0 x_0 x 0 so that ( s 0 ∗ D 0 ) x 0 = 0 (s_0^*D_0)_{x_0}=0 ( s 0 ∗ D 0 ) x 0 = 0 . Writing D = D 0 + A D=D_0+A D = D 0 + A with A ∈ Ω h e q , ρ 1 ( P ; A ) A\in\Omega^1_{\mathrm{heq},\rho}(P;A) A ∈ Ω heq , ρ 1 ( P ; A ) , given any y 1 , y 2 , z ∈ A y_1,y_2,z\in A y 1 , y 2 , z ∈ A and any 2-form Ψ 0 = 1 2 ∑ μ , ν = 1 n Ψ μ ν d x μ ∧ d x ν ∈ Λ 2 T x 0 ∗ M ⊗ A \Psi_0=\frac{1}{2}\sum_{\mu,\nu=1}^{n}\Psi_{\mu\nu}\,dx^\mu\wedge dx^\nu\in\Lambda^2T_{x_0}^*M\otimes A Ψ 0 = 2 1 ∑ μ , ν = 1 n Ψ μν d x μ ∧ d x ν ∈ Λ 2 T x 0 ∗ M ⊗ A (Ψ μ ν = − Ψ ν μ ∈ A \Psi_{\mu\nu}=-\Psi_{\nu\mu}\in A Ψ μν = − Ψ ν μ ∈ A ), choose A ∈ Ω h e q , ρ 1 ( P ; A ) A\in\Omega^1_{\mathrm{heq},\rho}(P;A) A ∈ Ω heq , ρ 1 ( P ; A ) such that on U U U ,
s 0 ∗ A = y 1 d x 1 + y 2 d x 2 + 1 2 ∑ μ , ν = 1 n x μ Ψ μ ν d x ν s_0^*A=y_1\,dx^1+y_2\,dx^2+\frac{1}{2}\sum_{\mu,\nu=1}^{n}x^\mu\Psi_{\mu\nu}\,dx^\nu s 0 ∗ A = y 1 d x 1 + y 2 d x 2 + 2 1 μ , ν = 1 ∑ n x μ Ψ μν d x ν
and choose ϵ ∈ Ω h e q , ρ 0 ( P ; A ) \epsilon\in\Omega^0_{\mathrm{heq},\rho}(P;A) ϵ ∈ Ω heq , ρ 0 ( P ; A ) with ( s 0 ∗ ϵ ) ( x 0 ) = z (s_0^*\epsilon)(x_0)=z ( s 0 ∗ ϵ ) ( x 0 ) = z . Evaluating at x 0 x_0 x 0 (where x μ ( x 0 ) = 0 x^\mu(x_0)=0 x μ ( x 0 ) = 0 and ( s 0 ∗ D 0 ) x 0 = 0 (s_0^*D_0)_{x_0}=0 ( s 0 ∗ D 0 ) x 0 = 0 ) gives ( s 0 ∗ D ) x 0 = ( s 0 ∗ A ) x 0 = y 1 d x 1 + y 2 d x 2 (s_0^*D)_{x_0}=(s_0^*A)_{x_0}=y_1\,dx^1+y_2\,dx^2 ( s 0 ∗ D ) x 0 = ( s 0 ∗ A ) x 0 = y 1 d x 1 + y 2 d x 2 , while ( s 0 ∗ d A ) x 0 = Ψ 0 (s_0^*dA)_{x_0}=\Psi_0 ( s 0 ∗ d A ) x 0 = Ψ 0 can be chosen arbitrarily so that ( s 0 ∗ F D ) x 0 = ( s 0 ∗ d D 0 ) x 0 + Ψ 0 + 1 2 [ y 1 , y 2 ] ∗ d x 1 ∧ d x 2 (s_0^*F_D)_{x_0}=(s_0^*dD_0)_{x_0}+\Psi_0+\frac{1}{2}[y_1,y_2]_*\,dx^1\wedge dx^2 ( s 0 ∗ F D ) x 0 = ( s 0 ∗ d D 0 ) x 0 + Ψ 0 + 2 1 [ y 1 , y 2 ] ∗ d x 1 ∧ d x 2 equals any prescribed element of Λ 2 T x 0 ∗ M ⊗ A \Lambda^2T_{x_0}^*M\otimes A Λ 2 T x 0 ∗ M ⊗ A . By pointwise non-degeneracy of ( ⋅ , ⋅ ) A , x 0 (\cdot,\cdot)_{A,x_0} ( ⋅ , ⋅ ) A , x 0 , it follows that j ( Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ) x 0 = 2 Jac [ ⋅ , ⋅ ] ∗ ( y 1 , y 2 , z ) d x 1 ∧ d x 2 = 0 j(\operatorname{Jac}_{[\cdot,\cdot]_*}(D,D,\epsilon))_{x_0}=2\operatorname{Jac}_{[\cdot,\cdot]_*}(y_1,y_2,z)\,dx^1\wedge dx^2=0 j ( Jac [ ⋅ , ⋅ ] ∗ ( D , D , ϵ ) ) x 0 = 2 Jac [ ⋅ , ⋅ ] ∗ ( y 1 , y 2 , z ) d x 1 ∧ d x 2 = 0 , so Jac [ ⋅ , ⋅ ] ∗ ( y 1 , y 2 , z ) = 0 \operatorname{Jac}_{[\cdot,\cdot]_*}(y_1,y_2,z)=0 Jac [ ⋅ , ⋅ ] ∗ ( y 1 , y 2 , z ) = 0 for all y 1 , y 2 , z ∈ A y_1,y_2,z\in A y 1 , y 2 , z ∈ A . □
Algebraic and Polynomial Obstructions on A-YMT Theories
Substituting F D = d D + 1 2 D ∧ ∗ D F_D=dD+\frac{1}{2}D\wedge_*D F D = d D + 2 1 D ∧ ∗ D into the A A A -YMT action functional (29) yields the expansion
S A [ D ] = S f r e e A [ D ] + S c u b i c A [ D ] + S q u a r t i c A [ D ] , (40) S^A[D]=S^A_{\mathrm{free}}[D]+S^A_{\mathrm{cubic}}[D]+S^A_{\mathrm{quartic}}[D], \tag*{(40)} S A [ D ] = S free A [ D ] + S cubic A [ D ] + S quartic A [ D ] , (40)
where S f r e e A [ D ] : = ∫ M ⟨ j ( d D ) , j ( d D ) ⟩ A S^A_{\mathrm{free}}[D]:=\int_M\langle j(dD),j(dD)\rangle_A S free A [ D ] := ∫ M ⟨ j ( d D ) , j ( d D ) ⟩ A , S c u b i c A [ D ] : = ∫ M ⟨ j ( d D ) , j ( D ∧ ∗ D ) ⟩ A , s y m S^A_{\mathrm{cubic}}[D]:=\int_M\langle j(dD),j(D\wedge_*D)\rangle_{A,\mathrm{sym}} S cubic A [ D ] := ∫ M ⟨ j ( d D ) , j ( D ∧ ∗ D ) ⟩ A , sym , and S q u a r t i c A [ D ] : = 1 4 ∫ M ⟨ j ( D ∧ ∗ D ) , j ( D ∧ ∗ D ) ⟩ A S^A_{\mathrm{quartic}}[D]:=\frac{1}{4}\int_M\langle j(D\wedge_*D),j(D\wedge_*D)\rangle_A S quartic A [ D ] := 4 1 ∫ M ⟨ j ( D ∧ ∗ D ) , j ( D ∧ ∗ D ) ⟩ A (in the linear/horizontal setting η = 0 \eta=0 η = 0 , or relative to a basepoint connection D 0 D_0 D 0 in the affine setting).
Commutator Projection, Abelianization, and Subbundle Localization
By (8) , for any 1-form D ∈ Ω 1 ( P ; A ) D\in\Omega^1(P;A) D ∈ Ω 1 ( P ; A ) , the binary product D ∧ ∗ D = 1 2 D ∧ [ ⋅ , ⋅ ] ∗ D D\wedge_*D=\frac{1}{2}D\wedge_{[\cdot,\cdot]_*}D D ∧ ∗ D = 2 1 D ∧ [ ⋅ , ⋅ ] ∗ D depends solely on the commutator bracket [ x , y ] ∗ = x ∗ y − y ∗ x [x,y]_* = x*y-y*x [ x , y ] ∗ = x ∗ y − y ∗ x . We record the immediate consequences of this projection in a single proposition.
Proposition 3.1 (Abelianization, Derived Filtration Shift, and Subbundle Localization). Let π : P → M \pi:P\to M π : P → M be a principal G G G -bundle over an n n n -manifold M M M (n ≥ 2 n\ge2 n ≥ 2 ) and let ρ : G → Aut A l g ( A ) \rho:G\to\operatorname{Aut}_{\mathrm{Alg}}(A) ρ : G → Aut Alg ( A ) be an algebra automorphism representation on an m m m -graded algebra A = ⨁ m ∈ Z A m A=\bigoplus_{m\in\mathbb{Z}}A^m A = ⨁ m ∈ Z A m .
(1) Abelianization: If A A A is a weak ( 1 , 1 ) (1,1) ( 1 , 1 ) -solvable m m m -graded algebra (i.e., A = ⨁ i = 1 r V i A=\bigoplus_{i=1}^{r}V_i A = ⨁ i = 1 r V i decomposes into mutually orthogonal commutative graded submodules), or if A = ⨁ i = 1 r A i A=\bigoplus_{i=1}^{r}A_i A = ⨁ i = 1 r A i is a vector space direct sum of commutative subalgebras and D ∈ C s p l i t 1 ( P ; A ) D\in C^1_{\mathrm{split}}(P;A) D ∈ C split 1 ( P ; A ) is algebraically split-adapted, then D ∧ ∗ D = 0 D\wedge_*D=0 D ∧ ∗ D = 0 identically, reducing F D = d D F_D=dD F D = d D and S A [ D ] = S f r e e A [ D ] S^A[D]=S^A_{\mathrm{free}}[D] S A [ D ] = S free A [ D ] to a free abelian theory in all dimensions n ≥ 2 n\ge2 n ≥ 2 .
(2) Derived Filtration Shift: If A A A is a solvable algebra of derived length s s s (A ( s + 1 ) = 0 A^{(s+1)}=0 A ( s + 1 ) = 0 ), then for every D ∈ Ω 1 ( P ; A ) D\in\Omega^1(P;A) D ∈ Ω 1 ( P ; A ) , D ∧ ∗ D ∈ Ω 2 ( P ; A ( 1 ) ) D\wedge_*D\in\Omega^2(P;A^{(1)}) D ∧ ∗ D ∈ Ω 2 ( P ; A ( 1 ) ) , and the iterated squares ( D ∧ ∗ D ) [ m ] ∈ Ω 2 m + 1 ( P ; A ( m + 1 ) ) (D\wedge_*D)^{[m]}\in\Omega^{2^{m+1}}(P;A^{(m+1)}) ( D ∧ ∗ D ) [ m ] ∈ Ω 2 m + 1 ( P ; A ( m + 1 ) ) vanish identically at step m = s m=s m = s .
(3) Commutator Subbundle Localization: Because [ A , A ] ∗ ⊂ A ( 1 ) [A,A]_*\subset A^{(1)} [ A , A ] ∗ ⊂ A ( 1 ) is Aut A l g ( A ) \operatorname{Aut}_{\mathrm{Alg}}(A) Aut Alg ( A ) -invariant, E [ A , A ] ∗ = P × ρ [ A , A ] ∗ ⊂ E A ( 1 ) ⊂ E A E_{[A,A]_*}=P\times_\rho[A,A]_*\subset E_{A^{(1)}}\subset E_A E [ A , A ] ∗ = P × ρ [ A , A ] ∗ ⊂ E A ( 1 ) ⊂ E A is a canonical vector subbundle. For every D ∈ Conn l i n ( P ; A ) D\in\operatorname{Conn}^{\mathrm{lin}}(P;A) D ∈ Conn lin ( P ; A ) , j ( D ∧ ∗ D ) ∈ Ω 2 ( M ; E [ A , A ] ∗ ) j(D\wedge_*D)\in\Omega^2(M;E_{[A,A]_*}) j ( D ∧ ∗ D ) ∈ Ω 2 ( M ; E [ A , A ] ∗ ) , so S q u a r t i c A [ D ] S^A_{\mathrm{quartic}}[D] S quartic A [ D ] vanishes identically whenever Ω 2 ( M ; E [ A , A ] ∗ ) \Omega^2(M;E_{[A,A]_*}) Ω 2 ( M ; E [ A , A ] ∗ ) lies in the radical of ⟨ ⋅ , ⋅ ⟩ A \langle\cdot,\cdot\rangle_A ⟨ ⋅ , ⋅ ⟩ A .
Proof. In part (1), writing D = ∑ i = 1 r ∑ m ∈ Z D i m D=\sum_{i=1}^{r}\sum_{m\in\mathbb{Z}}D_i^m D = ∑ i = 1 r ∑ m ∈ Z D i m with D i m ∈ Ω 1 ( P ; V i m ) D_i^m\in\Omega^1(P;V_i^m) D i m ∈ Ω 1 ( P ; V i m ) , orthogonality V i ∗ V j = 0 V_i*V_j=0 V i ∗ V j = 0 (i ≠ j i\ne j i = j ) and weak ( 1 , 1 ) (1,1) ( 1 , 1 ) -nilpotency D i m ∧ ∗ D i m = 0 D_i^m\wedge_*D_i^{m}=0 D i m ∧ ∗ D i m = 0 give D ∧ ∗ D = 0 D\wedge_*D=0 D ∧ ∗ D = 0 (and similarly for D ∈ C s p l i t 1 ( P ; A ) D\in C^1_{\mathrm{split}}(P;A) D ∈ C split 1 ( P ; A ) ). Part (2) follows by induction from A ( m ) ∗ A ( m ) = A ( m + 1 ) A^{(m)}*A^{(m)}=A^{(m+1)} A ( m ) ∗ A ( m ) = A ( m + 1 ) . For part (3), ( D ∧ ∗ D ) p ( v 1 , v 2 ) = [ D p ( v 1 ) , D p ( v 2 ) ] ∗ ∈ [ A , A ] ∗ ⊂ A ( 1 ) (D\wedge_*D)_p(v_1,v_2)=[D_p(v_1),D_p(v_2)]_*\in[A,A]_*\subset A^{(1)} ( D ∧ ∗ D ) p ( v 1 , v 2 ) = [ D p ( v 1 ) , D p ( v 2 ) ] ∗ ∈ [ A , A ] ∗ ⊂ A ( 1 ) , which descends via j j j to Ω 2 ( M ; E [ A , A ] ∗ ) \Omega^2(M;E_{[A,A]_*}) Ω 2 ( M ; E [ A , A ] ∗ ) . □
Obstructions on Internal Multiplicative, Self-Dual, and Topological A-YMT Theories
We now examine internal multiplicative A A A -YMT theories (31) and higher-degree internal topological (Chern–Weil) A A A -YMT theories (32) , in which the pairing on F D F_D F D is constructed from the internal product ∧ ∗ \wedge_* ∧ ∗ of Ω ( P ; A ) \Omega(P;A) Ω ( P ; A ) and a bundle trace functional tr : E A → M × R \operatorname{tr}:E_A\to M\times\mathbb{R} tr : E A → M × R (Remark 2.6 ). By (9) , the exterior square of the 2-form F D ∧ ∗ F D = F D ∧ ∘ ∗ F D F_D\wedge_*F_D=F_D\wedge_{\circ_*}F_D F D ∧ ∗ F D = F D ∧ ∘ ∗ F D projects onto the symmetric Jordan product x ∘ ∗ y = 1 2 ( x ∗ y + y ∗ x ) x\circ_*y=\frac{1}{2}(x*y+y*x) x ∘ ∗ y = 2 1 ( x ∗ y + y ∗ x ) , so ( 2 , s ) (2,s) ( 2 , s ) -solvability of ( A , ∗ ) (A,*) ( A , ∗ ) yields direct dimensional and self-duality obstructions.
Theorem 3.2 (Obstruction to Internal-Trace Topological and Self-Dual A-YMT Theories). Let π : P → M \pi:P\to M π : P → M be a principal G G G -bundle over an oriented n n n -manifold M M M , let ρ : G → Aut A l g ( A ) \rho:G\to\operatorname{Aut}_{\mathrm{Alg}}(A) ρ : G → Aut Alg ( A ) be an automorphism representation on a weak P \mathcal{P} P -( 2 , s ) (2,s) ( 2 , s ) -solvable m m m -graded algebra A = ⨁ m ∈ m A m A=\bigoplus_{m\in\mathfrak{m}} A^m A = ⨁ m ∈ m A m (for P ∈ { f u l l , b a l , b l k } \mathcal{P}\in\{\mathrm{full},\mathrm{bal},\mathrm{blk}\} P ∈ { full , bal , blk } ), and let tr ξ : E A → M × R \operatorname{tr}_{\xi}:E_A\to M\times\mathbb{R} tr ξ : E A → M × R be any bundle trace functional induced by ξ ∈ Γ ( E A ∗ ) \xi\in\Gamma(E_A^*) ξ ∈ Γ ( E A ∗ ) .
Higher-Dimensional Internal Chern-Weil Obstruction: Suppose n = 2 r n=2r n = 2 r is even. If n ≥ 2 ( s + 1 ) n\ge2(s+1) n ≥ 2 ( s + 1 ) (i.e., r ≥ s + 1 r\ge s+1 r ≥ s + 1 ), then for every admissible connection D ∈ Conn ρ , η ( P ; A ) D\in\operatorname{Conn}_{\rho,\eta}(P;A) D ∈ Conn ρ , η ( P ; A ) (including m m m -graded connections of Definition Definition 2.3(3); or for every homogeneous connection D ∈ Ω 1 ( P ; A m ) D\in\Omega^1(P;A^m) D ∈ Ω 1 ( P ; A m ) with m + m = m m+m=m m + m = m when A A A is only homogeneous weak P \mathcal{P} P -( 2 , s ) (2,s) ( 2 , s ) -solvable), ∧ ∗ , P r F D = 0 \wedge_{\ast,\mathcal{P}}^rF_D=0 ∧ ∗ , P r F D = 0 identically in Ω 2 r ( P ; A ) \Omega^{2r}(P;A) Ω 2 r ( P ; A ) , and hence the r r r -th internal topological A-YMT functional
S t o p , r , P A [ D ] = ∫ M tr ξ ( j ( ∧ ∗ , P r F D ) ) S_{\mathrm{top},r,\mathcal{P}}^A[D]=\int_M\operatorname{tr}_{\xi}\left(j\left(\wedge_{\ast,\mathcal{P}}^rF_D\right)\right) S top , r , P A [ D ] = ∫ M tr ξ ( j ( ∧ ∗ , P r F D ) ) vanishes identically on Conn ρ , η ( P ; A ) \operatorname{Conn}_{\rho,\eta}(P;A) Conn ρ , η ( P ; A ) for every trace section ξ ∈ Γ ( E A ∗ ) \xi\in\Gamma(E_A^*) ξ ∈ Γ ( E A ∗ ) .
Four-Dimensional Internal Topological and Self-Dual Obstruction (s = 1 s=1 s = 1 ): Suppose s = 1 s=1 s = 1 (so A A A is weak ( 2 , 1 ) (2,1) ( 2 , 1 ) -solvable, which by (9) includes every skew-commutative algebra since F D ∧ ∗ F D = F D ∧ ∘ ∗ F D = 0 ) . T h e n f o r e v e r y F_D\wedge_{\ast}F_D=F_D\wedge_{\circ\ast}F_D=0). Then for every F D ∧ ∗ F D = F D ∧ ∘∗ F D = 0 ) . T h e n f or e v er y D\in\Omega^1(P;A)$, F D ∧ ∗ F D = 0 F_D\wedge_{\ast}F_D=0 F D ∧ ∗ F D = 0 in Ω 4 ( P ; A ) \Omega^4(P;A) Ω 4 ( P ; A ) . Consequently, in dimension n = 4 n=4 n = 4 : (a) The $4D$ internal topological Yang-Mills functional S t o p , 2 A [ D ] = ∫ M tr ξ ( j ( F D ∧ ∗ F D ) ) S_{\mathrm{top},2}^A[D]=\int_M\operatorname{tr}_{\xi}(j(F_D\wedge_{\ast}F_D)) S top , 2 A [ D ] = ∫ M tr ξ ( j ( F D ∧ ∗ F D )) is identically zero; (b) Any internal multiplicative 4 D 4D 4 D A-YMT functional S ∗ A [ D ] = ∫ M tr ξ ( j ( F D ∧ ∗ ⋆ ^ F D ) ) S_{\ast}^A[D]=\int_M\operatorname{tr}_{\xi}(j(F_D\wedge_{\ast}\widehat{\star}F_D)) S ∗ A [ D ] = ∫ M tr ξ ( j ( F D ∧ ∗ ⋆ F D )) vanishes identically on all ⋆ ^ \widehat{\star} ⋆ -self-dual and ⋆ ^ \widehat{\star} ⋆ -anti-self-dual connections (⋆ ^ F D = ± F D \widehat{\star}F_D=\pm F_D ⋆ F D = ± F D ).
Proof. Decomposing A = ⨁ i = 1 k W i A=\bigoplus_{i=1}^k W_i A = ⨁ i = 1 k W i into mutually orthogonal weak P \mathcal{P} P -( 2 , s ) (2,s) ( 2 , s ) -nilpotent graded submodules W i = ⨁ m ∈ m W i m W_i=\bigoplus_{m\in\mathfrak{m}}W_i^m W i = ⨁ m ∈ m W i m (W i ∗ W j = 0 W_i\ast W_j=0 W i ∗ W j = 0 for i ≠ j i\ne j i = j ), we write F D = ∑ i = 1 k F D , i F_D=\sum_{i=1}^kF_{D,i} F D = ∑ i = 1 k F D , i with F D , i = ∑ m ∈ m F D , i m ∈ Ω 2 ( P ; W i ) F_{D,i}=\sum_{m\in\mathfrak{m}}F_{D,i}^m\in\Omega^2(P;W_i) F D , i = ∑ m ∈ m F D , i m ∈ Ω 2 ( P ; W i ) (where F D , i m F_{D,i}^m F D , i m is given by (23) ). Expanding ∧ ∗ , P r F D \wedge_{\ast,\mathcal{P}}^rF_D ∧ ∗ , P r F D and using W i ∗ W j = 0 W_i\ast W_j=0 W i ∗ W j = 0 (i ≠ j i\ne j i = j ) yields ∑ i = 1 k ∑ m 1 , … , m r ∈ m p r ( F D , i m 1 , … , F D , i m r ) \sum_{i=1}^k\sum_{m_1,\ldots,m_r\in\mathfrak{m}}p_r(F_{D,i}^{m_1},\ldots,F_{D,i}^{m_r}) ∑ i = 1 k ∑ m 1 , … , m r ∈ m p r ( F D , i m 1 , … , F D , i m r ) , where each P \mathcal{P} P -parenthesized monomial of degree r ≥ s + 1 r\ge s+1 r ≥ s + 1 vanishes by Definition 2.2(4). Thus ∧ ∗ , P r F D = 0 \wedge_{\ast,\mathcal{P}}^rF_D=0 ∧ ∗ , P r F D = 0 for all r ≥ s + 1 r\ge s+1 r ≥ s + 1 , proving (1). Setting s = 1 s=1 s = 1 and r = 2 r=2 r = 2 gives F D ∧ ∗ F D = 0 F_D\wedge_{\ast}F_D=0 F D ∧ ∗ F D = 0 , which immediately implies (2a) and (2b). □
Finally, given a surjective R R R -linear projection Q : E ↠ A Q:E\twoheadrightarrow A Q : E ↠ A and an R R R -linear section S : A ↪ E S:A\hookrightarrow E S : A ↪ E with Q ∘ S = id A Q\circ S=\operatorname{id}_A Q ∘ S = id A (for instance, a split short exact sequence 0 → H → E → Q A → 0 0\to H\to E\xrightarrow{Q}A\to0 0 → H → E Q A → 0 ), endowing E E E with the ( Q , S ) (Q,S) ( Q , S ) -pulled-back product x ∗ Q , S y : = S ( Q ( x ) ∗ Q ( y ) ) x\ast_{Q,S}y:=S(Q(x)\ast Q(y)) x ∗ Q , S y := S ( Q ( x ) ∗ Q ( y )) yields the following invariance property.
Proposition 3.3 (Invariance under ( Q , S ) (Q,S) ( Q , S ) -Splitting Extensions). For x ∗ ′ y : = S ( Q ( x ) ∗ Q ( y ) ) x\ast' y:=S(Q(x)\ast Q(y)) x ∗ ′ y := S ( Q ( x ) ∗ Q ( y )) on E E E with Q ∘ S = id A Q\circ S=\operatorname{id}_A Q ∘ S = id A , any non-associative monomial p r p_r p r satisfies the relation p r ∗ ′ ( α 1 , … , α r ) = ( Ω S ) ( p r ∗ ( ( Ω Q ) ( α 1 ) , … , ( Ω Q ) ( α r ) ) ) p_r^{\ast'}(\alpha_1,\ldots,\alpha_r)=(\Omega S)(p_r^{\ast}((\Omega Q)(\alpha_1),\ldots,(\Omega Q)(\alpha_r))) p r ∗ ′ ( α 1 , … , α r ) = ( Ω S ) ( p r ∗ (( Ω Q ) ( α 1 ) , … , ( Ω Q ) ( α r ))) on Ω ( P ; E ) \Omega(P;E) Ω ( P ; E ) . Since Ω S \Omega S Ω S is injective and Ω Q \Omega Q Ω Q is surjective, ( E , ∗ ′ ) (E,\ast') ( E , ∗ ′ ) is P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nil if and only if ( A , ∗ ) (A,\ast) ( A , ∗ ) is P \mathcal{P} P -( k , s ) (k,s) ( k , s ) -nil.
Proof. Follows by induction on r ≥ 2 r\ge2 r ≥ 2 from α ∧ ∗ ′ β = ( Ω S ) ( ( Ω Q ) ( α ) ∧ ∗ ( Ω Q ) ( β ) ) \alpha\wedge_{\ast'}\beta=(\Omega S)((\Omega Q)(\alpha)\wedge_{\ast}(\Omega Q)(\beta)) α ∧ ∗ ′ β = ( Ω S ) (( Ω Q ) ( α ) ∧ ∗ ( Ω Q ) ( β )) and ( Ω Q ) ∘ ( Ω S ) = id Ω ( P ; A ) (\Omega Q)\circ(\Omega S)=\operatorname{id}_{\Omega(P;A)} ( Ω Q ) ∘ ( Ω S ) = id Ω ( P ; A ) . □
Parenthesization Cutoffs and Holonomy Reductions in Reductive Cartan-Yang-Mills Theories
In approaches that unify gravity and Yang-Mills gauge theory—such as MacDowell-Mansouri gravity [36 , 37 ] and Cartan-Yang-Mills theories—one works with a reductive A A A -connection ∇ = e + ω ∈ Ω 1 ( P ; A 0 ) ⊕ Ω 1 ( P ; A 1 ) \nabla=e+\omega\in\Omega^1(P;A_0)\oplus\Omega^1(P;A_1) ∇ = e + ω ∈ Ω 1 ( P ; A 0 ) ⊕ Ω 1 ( P ; A 1 ) and uses the soldering 1 1 1 -form e e e to construct an internal constitutive operator on the Cartan curvature F ∇ F_{\nabla} F ∇ . We now analyze the dimensional cutoffs of reductive Cartan-Yang-Mills theories under both quadratic-block parenthesization (12) and recursive binary-tree parenthesization (10) , establish the 4D first-Bianchi contraction identity and proportionality between left-grouped and right-nested EHP terms, pair the dimensional cutoffs with Berger’s holonomy classification in dimensions 4 ≤ n ≤ 7 4\le n\le7 4 ≤ n ≤ 7 , and establish the representation-theoretic dichotomy at the critical G 2 G_2 G 2 dimension n = k = 7 n=k=7 n = k = 7 .
H-Equivariant Reductive Soldering Products and k k k -Proper Bundles
Let us establish the algebraic and geometric setup:
Let h ⊂ s o ( k 1 ) ⊕ ⋯ ⊕ s o ( k r ) ⊂ s o ( k ) \mathfrak{h} \subset\mathfrak{so}(k_1) \oplus\cdots\oplus\mathfrak{so}(k_r) \subset\mathfrak{so}(k) h ⊂ so ( k 1 ) ⊕ ⋯ ⊕ so ( k r ) ⊂ so ( k ) (k 1 + ⋯ + k r = k k_1+\cdots+k_r=k k 1 + ⋯ + k r = k ) be a matrix Lie subalgebra equipped with the skew-commutative product X ⋏ Y : = 1 2 ( X Y − Y X ) = 1 2 [ X , Y ] X\curlywedge Y := \frac{1}{2}(XY-YX)=\frac{1}{2}[X,Y] X ⋏ Y := 2 1 ( X Y − Y X ) = 2 1 [ X , Y ] in h \mathfrak{h} h (a k k k -algebra, with connected k k k -group H ⊂ S O ( k 1 ) × ⋯ × S O ( k r ) H\subset SO(k_1)\times\cdots\times SO(k_r) H ⊂ S O ( k 1 ) × ⋯ × S O ( k r ) ) [21 ] .
On the reductive direct sum A = A 0 ⊕ A 1 = R k ⊕ h A=A_0\oplus A_1=\mathbb{R}^k\oplus\mathfrak{h} A = A 0 ⊕ A 1 = R k ⊕ h (where H H H acts on A 0 = R k A_0=\mathbb{R}^k A 0 = R k via the defining vector representation and on A 1 = h A_1=\mathfrak{h} A 1 = h via Ad H \operatorname{Ad}_H Ad H ), we identify R k ∧ R k ≃ s o ( k ) \mathbb{R}^k\wedge\mathbb{R}^k\simeq\mathfrak{so}(k) R k ∧ R k ≃ so ( k ) via ( u ∧ v ) a b : = u a v b − u b v a (u\wedge v)_{ab}:=u_av_b-u_bv_a ( u ∧ v ) ab := u a v b − u b v a for u , v ∈ R k u,v\in\mathbb{R}^k u , v ∈ R k , and let σ : R k ∧ R k → h \sigma:\mathbb{R}^k\wedge\mathbb{R}^k\to\mathfrak{h} σ : R k ∧ R k → h be the canonical H H H -equivariant skew-symmetric soldering projection defined by the orthogonal projection s o ( k ) ↠ h \mathfrak{so}(k)\twoheadrightarrow\mathfrak{h} so ( k ) ↠ h with respect to the positive-definite inner product ⟨ X , Y ⟩ s o ( k ) : = − 1 2 Tr ( X Y ) = 1 2 ∑ a , b = 1 k X a b Y a b \langle X,Y\rangle_{\mathfrak{so}(k)}:=-\frac{1}{2}\operatorname{Tr}(XY)=\frac{1}{2}\sum_{a,b=1}^{k}X_{ab}Y_{ab} ⟨ X , Y ⟩ so ( k ) := − 2 1 Tr ( X Y ) = 2 1 ∑ a , b = 1 k X ab Y ab . Explicitly, for any ⟨ ⋅ , ⋅ ⟩ s o ( k ) \langle\cdot,\cdot\rangle_{\mathfrak{so}(k)} ⟨ ⋅ , ⋅ ⟩ so ( k ) -orthonormal basis { T A } \{T_A\} { T A } of h \mathfrak{h} h ,
σ ( u ∧ v ) = ∑ A ⟨ T A , u ∧ v ⟩ s o ( k ) T A = ∑ A ( ∑ a , b = 1 k ( T A ) a b u a v b ) T A (41) \sigma(u\wedge v)=\sum_A\langle T_A,u\wedge v\rangle_{\mathfrak{so}(k)}T_A=\sum_A\left(\sum_{a,b=1}^{k}(T_A)_{ab}u_av_b\right)T_A \tag*{(41)} σ ( u ∧ v ) = A ∑ ⟨ T A , u ∧ v ⟩ so ( k ) T A = A ∑ a , b = 1 ∑ k ( T A ) ab u a v b T A (41)
so that for R k \mathbb{R}^k R k -valued forms α ∈ Ω p ( P ; R k ) \alpha\in\Omega^p(P;\mathbb{R}^k) α ∈ Ω p ( P ; R k ) and β ∈ Ω q ( P ; R k ) \beta\in\Omega^q(P;\mathbb{R}^k) β ∈ Ω q ( P ; R k ) , the h \mathfrak{h} h -valued ( p + q ) (p+q) ( p + q ) -form σ ( α ∧ β ) = ∑ A ( ∑ a , b = 1 k ( T A ) a b α a ∧ β b ) T A \sigma(\alpha\wedge\beta)=\sum_A(\sum_{a,b=1}^{k}(T_A)_{ab}\alpha^a\wedge\beta^b)T_A σ ( α ∧ β ) = ∑ A ( ∑ a , b = 1 k ( T A ) ab α a ∧ β b ) T A is the orthogonal projection of the skew-symmetric matrix ( α ∧ β ) a b = α a ∧ β b − α b ∧ β a = α a ∧ β b − ( − 1 ) p q β a ∧ α b ∈ Ω p + q ( P ; s o ( k ) ) (\alpha\wedge\beta)_{ab}=\alpha^a\wedge\beta^b-\alpha^b\wedge\beta^a=\alpha^a\wedge\beta^b-(-1)^{pq}\beta^a\wedge\alpha^b\in\Omega^{p+q}(P;\mathfrak{so}(k)) ( α ∧ β ) ab = α a ∧ β b − α b ∧ β a = α a ∧ β b − ( − 1 ) pq β a ∧ α b ∈ Ω p + q ( P ; so ( k )) . Let × H : R k ∧ R k → R k \times_H:\mathbb{R}^k\wedge\mathbb{R}^k\to\mathbb{R}^k × H : R k ∧ R k → R k be an H H H -equivariant skew-symmetric bilinear vector product on R k \mathbb{R}^k R k (for irreducible H ⊂ S O ( k ) H\subset SO(k) H ⊂ S O ( k ) , × H = 0 \times_H=0 × H = 0 except when k = 7 k=7 k = 7 and H = G 2 ⊂ S O ( 7 ) H=G_2\subset SO(7) H = G 2 ⊂ S O ( 7 ) , where ( u × G 2 v ) a = φ a b c u b v c (u\times_{G_2}v)^a=\varphi^a{}_{bc}u^bv^c ( u × G 2 v ) a = φ a b c u b v c is the octonionic cross product induced by the associative G 2 G_2 G 2 -invariant 3 3 3 -form φ 0 ∈ ( Λ 3 ( R 7 ) ∗ ) G 2 \varphi_0\in(\Lambda^3(\mathbb{R}^7)^*)_{G_2} φ 0 ∈ ( Λ 3 ( R 7 ) ∗ ) G 2 ) [2 , 29 ] . We define the H H H -equivariant reductive Cartan-MacDowell-Mansouri product ⋏ ⋊ \curlywedge_{\rtimes} ⋏ ⋊ on A = R k ⊕ h A=\mathbb{R}^k\oplus\mathfrak{h} A = R k ⊕ h by
( v 1 , X 1 ) ⋏ ⋊ ( v 2 , X 2 ) : = ( 1 2 ( X 1 v 2 − X 2 v 1 ) + v 1 × H v 2 , X 1 ⋏ X 2 + σ ( v 1 ∧ v 2 ) ) (42) (v_1,X_1)\curlywedge_{\rtimes}(v_2,X_2):=\left(\frac{1}{2}(X_1v_2-X_2v_1)+v_1\times_Hv_2,\ X_1\curlywedge X_2+\sigma(v_1\wedge v_2)\right) \tag*{(42)} ( v 1 , X 1 ) ⋏ ⋊ ( v 2 , X 2 ) := ( 2 1 ( X 1 v 2 − X 2 v 1 ) + v 1 × H v 2 , X 1 ⋏ X 2 + σ ( v 1 ∧ v 2 ) ) (42)
where ( X v ) a : = ∑ b = 1 k X a b v b (Xv)^a:=\sum_{b=1}^{k}X_{ab}v^b ( X v ) a := ∑ b = 1 k X ab v b denotes the standard matrix action of X ∈ h ⊂ s o ( k ) X\in\mathfrak{h}\subset\mathfrak{so}(k) X ∈ h ⊂ so ( k ) on v ∈ R k v\in\mathbb{R}^k v ∈ R k . The product (42) satisfies three key properties:
(P1) H H H -Equivariance: H H H acts on ( A , ⋏ ⋊ ) (A,\curlywedge_{\rtimes}) ( A , ⋏ ⋊ ) by algebra automorphisms, inducing the associated algebra bundle E A = P × H ( R k ⊕ h ) → M E_A=P\times_H(\mathbb{R}^k\oplus\mathfrak{h})\to M E A = P × H ( R k ⊕ h ) → M .
(P2) Skew-Commutativity, Parenthesization, and Rank Identities: Because (42) is skew-commutative on A = R k ⊕ h A=\mathbb{R}^k\oplus\mathfrak{h} A = R k ⊕ h , ( A , ⋏ ⋊ ) (A,\curlywedge_{\rtimes}) ( A , ⋏ ⋊ ) is ( 2 , 1 ) (2,1) ( 2 , 1 ) -nil (Φ ⋏ ⋊ Φ = 0 \Phi\curlywedge_{\rtimes}\Phi=0 Φ ⋏ ⋊ Φ = 0 for all Φ ∈ Ω even ( P ; A ) \Phi\in\Omega^{\operatorname{even}}(P;A) Φ ∈ Ω even ( P ; A ) ) and block-balanced ( 1 , 3 ) (1,3) ( 1 , 3 ) -nil (⋏ ⋊ , b l k m α = 0 \curlywedge_{\rtimes,\mathrm{blk}}^m\alpha=0 ⋏ ⋊ , blk m α = 0 for all α ∈ Ω 1 ( P ; A ) \alpha\in\Omega^1(P;A) α ∈ Ω 1 ( P ; A ) and m ≥ 4 m\geq4 m ≥ 4 by Lemma 2.1(1)). Furthermore, because any A 0 A_0 A 0 -valued soldering 1 1 1 -form e = ( e , 0 ) ∈ Ω 1 ( P ; R k ) e=(e,0)\in\Omega^1(P;\mathbb{R}^k) e = ( e , 0 ) ∈ Ω 1 ( P ; R k ) takes values in the k k k -dimensional subspace A 0 = R k A_0=\mathbb{R}^k A 0 = R k , Lemma 2.1(2c) implies that ⋏ ⋊ , P m e = 0 \curlywedge_{\rtimes,\mathrm{P}}^m e=0 ⋏ ⋊ , P m e = 0 identically on P P P for all m > k m>k m > k . Moreover, when × H = 0 \times_H=0 × H = 0 , A = A 0 ⊕ A 1 A=A_0\oplus A_1 A = A 0 ⊕ A 1 is Z 2 \mathbb{Z}_2 Z 2 -graded (A 0 ⋏ ⋊ A 0 ⊂ A 1 A_0\curlywedge_{\rtimes}A_0\subset A_1 A 0 ⋏ ⋊ A 0 ⊂ A 1 , A 1 ⋏ ⋊ A 0 ⊂ A 0 A_1\curlywedge_{\rtimes}A_0\subset A_0 A 1 ⋏ ⋊ A 0 ⊂ A 0 , A 1 ⋏ ⋊ A 1 ⊂ A 1 A_1\curlywedge_{\rtimes}A_1\subset A_1 A 1 ⋏ ⋊ A 1 ⊂ A 1 ), so e ⋏ ⋊ e = ( 0 , σ ( e ∧ e ) ) ∈ Ω 2 ( P ; h ) e\curlywedge_{\rtimes}e=(0,\sigma(e\wedge e))\in\Omega^2(P;\mathfrak{h}) e ⋏ ⋊ e = ( 0 , σ ( e ∧ e )) ∈ Ω 2 ( P ; h ) lies in A 1 = h A_1=\mathfrak{h} A 1 = h and ⋏ ⋊ , b a l 3 e = ( e ⋏ ⋊ e ) ⋏ ⋊ e ∈ Ω 3 ( P ; R k ) \curlywedge_{\rtimes,\mathrm{bal}}^3e=(e\curlywedge_{\rtimes}e)\curlywedge_{\rtimes}e\in\Omega^3(P;\mathbb{R}^k) ⋏ ⋊ , bal 3 e = ( e ⋏ ⋊ e ) ⋏ ⋊ e ∈ Ω 3 ( P ; R k ) lies in A 0 = R k A_0=\mathbb{R}^k A 0 = R k .
(P3) Classical Cartan Torsion on A 0 = R k A_0=\mathbb{R}^k A 0 = R k . For ω ∈ Ω 1 ( P ; h ) \omega\in\Omega^1(P;\mathfrak{h}) ω ∈ Ω 1 ( P ; h ) and e ∈ Ω 1 ( P ; R k ) e\in\Omega^1(P;\mathbb{R}^k) e ∈ Ω 1 ( P ; R k ) , the generalized torsion in (22) is Θ e , ω = ( d e + 1 2 ω ∧ e , 0 ) \Theta_{e,\omega}=(de+\frac{1}{2}\omega\wedge e,0) Θ e , ω = ( d e + 2 1 ω ∧ e , 0 ) (since by (6) and (42) , ( ω ⋏ ⋊ e + e ⋏ ⋊ ω ) p ( v 1 , v 2 ) = 2 ω p ( v 1 ) ⋅ 1 2 e p ( v 2 ) − 2 ω p ( v 2 ) ⋅ 1 2 e p ( v 1 ) = ( ω ∧ e ) p ( v 1 , v 2 ) (\omega\curlywedge_{\rtimes}e+e\curlywedge_{\rtimes}\omega)_p(v_1,v_2)=2\omega_p(v_1)\cdot\frac{1}{2}e_p(v_2)-2\omega_p(v_2)\cdot\frac{1}{2}e_p(v_1)=(\omega\wedge e)_p(v_1,v_2) ( ω ⋏ ⋊ e + e ⋏ ⋊ ω ) p ( v 1 , v 2 ) = 2 ω p ( v 1 ) ⋅ 2 1 e p ( v 2 ) − 2 ω p ( v 2 ) ⋅ 2 1 e p ( v 1 ) = ( ω ∧ e ) p ( v 1 , v 2 ) ), which vanishes if and only if ω ~ : = 1 2 ω \widetilde{\omega}:=\frac{1}{2}\omega ω := 2 1 ω satisfies the classical Cartan torsion-free equation d e + ω ~ ∧ e = 0 de+\widetilde{\omega}\wedge e=0 d e + ω ∧ e = 0 .
Let N N N be a simply connected smooth k k k -manifold (k ≤ n k\leq n k ≤ n ). Following [21 ] , a principal H H H -bundle π : P → M \pi:P\to M π : P → M over an n n n -manifold M M M is called k k k -proper if there exists an immersed simply connected k k k -manifold ι : N → M \iota:N\to M ι : N → M whose orthonormal frame bundle F N FN F N reduces to an H H H -subbundle P N ⊂ F N P_N\subset FN P N ⊂ F N isomorphic as a principal H H H -bundle over N N N to i ∗ P i^*P i ∗ P via κ : P N → ∼ i ∗ P \kappa:P_N\xrightarrow{\sim}i^*P κ : P N ∼ i ∗ P , and we define the partially torsionless irreducible configuration space Conn Θ N = 0 , i r r ( P ) \operatorname{Conn}^{\Theta_{N=0,\mathrm{irr}}}(P) Conn Θ N = 0 , irr ( P ) to be the space of pairs of horizontal H H H -equivariant soldering 1 1 1 -forms e ∈ Ω h o r 1 ( P ; R k ) e\in\Omega^1_{\mathrm{hor}}(P;\mathbb{R}^k) e ∈ Ω hor 1 ( P ; R k ) and Ehresmann connections ω ~ = 1 2 ω ∈ Conn ( P ; h ) \widetilde{\omega}=\frac{1}{2}\omega\in\operatorname{Conn}(P;\mathfrak{h}) ω = 2 1 ω ∈ Conn ( P ; h ) whose pullback along i ∘ κ : P N → P i\circ\kappa:P_N\to P i ∘ κ : P N → P satisfies the Cartan torsion-free equation ( i ∘ κ ) ∗ ( d e + ω ~ ∧ e ) = 0 (i\circ\kappa)^*(de+\widetilde{\omega}\wedge e)=0 ( i ∘ κ ) ∗ ( d e + ω ∧ e ) = 0 on P N P_N P N (and d e + ω ~ ∧ e = 0 de+\widetilde{\omega}\wedge e=0 d e + ω ∧ e = 0 globally on P P P whenever k = n k=n k = n and M = N M=N M = N , in which case we write
Conn A Θ = 0 , i r r ( P ) \operatorname{Conn}^{\Theta=0,\mathrm{irr}}_{A}(P) Conn A Θ = 0 , irr ( P ) such that ( i ∘ κ ) ∗ e ∈ Ω 1 ( P N ; R k ) (i\circ\kappa)^*e\in\Omega^1(P_N;\mathbb{R}^k) ( i ∘ κ ) ∗ e ∈ Ω 1 ( P N ; R k ) is the canonical orthonormal soldering 1-form on P N ⊂ F N P_N\subset F N P N ⊂ F N inducing a locally irreducible, locally non-symmetric Riemannian metric g N g_N g N on N N N (so that ( N , g N ) (N,g_N) ( N , g N ) is a Berger k k k -manifold [6 , 7 , 25 ] with Levi-Civita connection ω ~ N : = 1 2 ( i ∘ κ ) ∗ ω \widetilde{\omega}_N:=\frac{1}{2}(i\circ\kappa)^*\omega ω N := 2 1 ( i ∘ κ ) ∗ ω and curvature Ω ~ N = 1 2 ( i ∘ κ ) ∗ Ω ω \widetilde{\Omega}_N=\frac{1}{2}(i\circ\kappa)^*\Omega_\omega Ω N = 2 1 ( i ∘ κ ) ∗ Ω ω ).
Remark 4.1 (Geometric Realizations of k k k -Proper Bundles and the Role of Mixed Torsion when k < n k<n k < n ). When k = n k=n k = n and M = N M=N M = N is simply connected, P ⊂ F M P\subset FM P ⊂ F M is an H H H -structure on M M M with globally torsion-free Levi-Civita connection ω ~ = 1 2 ω \widetilde{\omega}=\frac{1}{2}\omega ω = 2 1 ω (Θ e , ω = d e + ω ~ ∧ e = 0 \Theta_{e,\omega}=de+\widetilde{\omega}\wedge e=0 Θ e , ω = d e + ω ∧ e = 0 globally on P P P ) and orthonormal coframe e e e . When k < n k<n k < n , e ∈ Ω h e q 1 ( P ; R k ) e\in\Omega^1_{\mathrm{heq}}(P;\mathbb{R}^k) e ∈ Ω heq 1 ( P ; R k ) has a horizontal kernel distribution ν ⊥ = ker ( e ) ⊂ T M \nu^\perp=\ker(e)\subset TM ν ⊥ = ker ( e ) ⊂ T M of rank n − k n-k n − k . Suppose π M : M → N \pi_M:M\to N π M : M → N is a smooth submersion (for instance, a Riemannian product M = N k × K n − k M=N^k\times K^{n-k} M = N k × K n − k ) with i : N ↪ M i:N\hookrightarrow M i : N ↪ M a section, P = π M ∗ P N P=\pi_M^*P_N P = π M ∗ P N , and e = π M ∗ e N e=\pi_M^*e_N e = π M ∗ e N . Any connection ω ~ = π M ∗ ω ~ N + β \widetilde{\omega}=\pi_M^*\widetilde{\omega}_N+\beta ω = π M ∗ ω N + β with β ∈ Ω h e q 1 ( P ; h ) \beta\in\Omega^1_{\mathrm{heq}}(P;\mathfrak{h}) β ∈ Ω heq 1 ( P ; h ) vanishing along T N TN T N (ι v β = 0 \iota_v\beta=0 ι v β = 0 for v ∈ T N v\in TN v ∈ T N ) satisfies the pulled-back torsion-free condition ( i ∘ κ ) ∗ ( d e + ω ~ ∧ e ) = d e N + ω ~ N ∧ e N = 0 (i\circ\kappa)^*(de+\widetilde{\omega}\wedge e)=de_N+\widetilde{\omega}_N\wedge e_N=0 ( i ∘ κ ) ∗ ( d e + ω ∧ e ) = d e N + ω N ∧ e N = 0 on P N P_N P N . Notice that on M n = N k × K n − k M^n=N^k\times K^{n-k} M n = N k × K n − k with e = π M ∗ e N e=\pi_M^*e_N e = π M ∗ e N , d e = π M ∗ d e N = − π M ∗ ( ω ~ N ∧ e N ) de=\pi_M^*de_N=-\pi_M^*(\widetilde{\omega}_N\wedge e_N) d e = π M ∗ d e N = − π M ∗ ( ω N ∧ e N ) , so the full torsion on M n M^n M n is Θ e , ω = d e + ω ~ ∧ e = β ∧ π M ∗ e N \Theta_{e,\omega}=de+\widetilde{\omega}\wedge e=\beta\wedge\pi_M^*e_N Θ e , ω = d e + ω ∧ e = β ∧ π M ∗ e N . Since β \beta β is along T ∗ K T^*K T ∗ K , e N e_N e N is an orthonormal coframe along T ∗ N T^*N T ∗ N , and H ⊂ S O ( k ) H\subset SO(k) H ⊂ S O ( k ) acts faithfully on R k \mathbb{R}^k R k , imposing global torsion-freeness Θ e , ω = 0 \Theta_{e,\omega}=0 Θ e , ω = 0 on all of M n M^n M n would force β ( w ) ⋅ e N ( v ) = 0 \beta(w)\cdot e_N(v)=0 β ( w ) ⋅ e N ( v ) = 0 for all w ∈ T K w\in TK w ∈ T K , v ∈ T N v\in TN v ∈ T N and hence β = 0 \beta=0 β = 0 , which would make both e e e and Ω ω \Omega_\omega Ω ω pullbacks from N k N^k N k and force every n n n -form built from ( e , Ω ω ) (e,\Omega_\omega) ( e , Ω ω ) to vanish identically for n > k n>k n > k (since Λ n ( T ∗ N k ) = 0 ) ! \Lambda^n(T^*N^k)=0)! Λ n ( T ∗ N k ) = 0 )! Conversely, because Conn A Θ N = 0 , i r r ( P ) \operatorname{Conn}^{\Theta_N=0,\mathrm{irr}}_{A}(P) Conn A Θ N = 0 , irr ( P ) imposes torsion-freeness ( i ∘ κ ) ∗ Θ e , ω = 0 (i\circ\kappa)^*\Theta_{e,\omega}=0 ( i ∘ κ ) ∗ Θ e , ω = 0 only along N k N^k N k while allowing non-zero mixed torsion Θ e , ω ( w , v ) = β ( w ) ⋅ e N ( v ) ≠ 0 \Theta_{e,\omega}(w,v)=\beta(w)\cdot e_N(v)\ne0 Θ e , ω ( w , v ) = β ( w ) ⋅ e N ( v ) = 0 on T K × T N TK\times TN T K × T N , a non-zero horizontal 1-form β \beta β along T K TK T K contributes non-zero transverse curvature Ω ω \Omega_\omega Ω ω along T ∗ K ∧ T ∗ K T^*K\wedge T^*K T ∗ K ∧ T ∗ K (or T ∗ K ∧ T ∗ N T^*K\wedge T^*N T ∗ K ∧ T ∗ N ). Since the two factors of Ω ω \Omega_\omega Ω ω in the pure Yang-Mills n n n -form Ω ω ⋏ λ ( ( ⋏ λ , P n − 4 e ) ⋏ λ Ω ω ) \Omega_\omega\curlywedge_\lambda((\curlywedge^{n-4}_{\lambda,P}e)\curlywedge_\lambda\Omega_\omega) Ω ω ⋏ λ (( ⋏ λ , P n − 4 e ) ⋏ λ Ω ω ) have total degree 4 and can saturate at most 4 cotangent directions along T ∗ K T^*K T ∗ K while ⋏ λ , P n − 4 e \curlywedge^{n-4}_{\lambda,P}e ⋏ λ , P n − 4 e saturates n − 4 ≤ k n-4\le k n − 4 ≤ k directions along T ∗ N T^*N T ∗ N , the pure Yang-Mills n n n -form is non-vanishing on Conn A Θ N = 0 , i r r ( P ) \operatorname{Conn}^{\Theta_N=0,\mathrm{irr}}_{A}(P) Conn A Θ N = 0 , irr ( P ) whenever n − 4 ≤ k < n n-4\le k<n n − 4 ≤ k < n .
Let us clarify the relationship between the quadratic constitutive expansion of the full Cartan curvature F ∇ = Ω ω + Θ e , ω + E e F_\nabla=\Omega_\omega+\Theta_{e,\omega}+E_e F ∇ = Ω ω + Θ e , ω + E e (E e = 1 2 e ∧ e E_e=\frac{1}{2}e\wedge e E e = 2 1 e ∧ e ) in (31) and the canonical left-grouped Cartan-Yang-Mills sectors. In a non-associative algebra ( A , ⋏ λ ) (A,\curlywedge_\lambda) ( A , ⋏ λ ) , if one applies the internal constitutive operator ⋆ ^ P ( Φ ) : = ( ⋏ λ , P n − 4 e ) ⋏ λ Φ \hat{\star}_P(\Phi):=(\curlywedge^{n-4}_{\lambda,P}e)\curlywedge_\lambda\Phi ⋆ ^ P ( Φ ) := ( ⋏ λ , P n − 4 e ) ⋏ λ Φ directly to the torsionless Cartan curvature F ∇ Θ = 0 : = Ω ω + 1 2 e ∧ e F_{\nabla}^{\Theta=0}:=\Omega_\omega+\frac{1}{2}e\wedge e F ∇ Θ = 0 := Ω ω + 2 1 e ∧ e , expanding F ∇ Θ = 0 ⋏ λ ⋆ ^ P ( F ∇ Θ = 0 ) F_{\nabla}^{\Theta=0}\curlywedge_\lambda\hat{\star}_P(F_{\nabla}^{\Theta=0}) F ∇ Θ = 0 ⋏ λ ⋆ ^ P ( F ∇ Θ = 0 ) produces the quadratic Yang-Mills term Ω ω ⋏ λ ( ( ⋏ λ , P n − 4 e ) ⋏ λ Ω ω ) \Omega_\omega\curlywedge_\lambda((\curlywedge^{n-4}_{\lambda,P}e)\curlywedge_\lambda\Omega_\omega) Ω ω ⋏ λ (( ⋏ λ , P n − 4 e ) ⋏ λ Ω ω ) , together with the cross-terms 1 2 ( e ∧ e ) ⋏ λ ( ( ⋏ λ , P n − 4 e ) ⋏ λ Ω ω ) + 1 2 Ω ω ⋏ λ ( ( ⋏ λ , P n − 4 e ) ⋏ λ ( e ∧ e ) ) \frac{1}{2}(e\wedge e)\curlywedge_\lambda((\curlywedge^{n-4}_{\lambda,P}e)\curlywedge_\lambda\Omega_\omega)+\frac{1}{2}\Omega_\omega\curlywedge_\lambda((\curlywedge^{n-4}_{\lambda,P}e)\curlywedge_\lambda(e\wedge e)) 2 1 ( e ∧ e ) ⋏ λ (( ⋏ λ , P n − 4 e ) ⋏ λ Ω ω ) + 2 1 Ω ω ⋏ λ (( ⋏ λ , P n − 4 e ) ⋏ λ ( e ∧ e )) and the cosmological term 1 4 ( e ∧ e ) ⋏ λ ( ( ⋏ λ , P n − 4 e ) ⋏ λ ( e ∧ e ) ) \frac{1}{4}(e\wedge e)\curlywedge_\lambda((\curlywedge^{n-4}_{\lambda,P}e)\curlywedge_\lambda(e\wedge e)) 4 1 ( e ∧ e ) ⋏ λ (( ⋏ λ , P n − 4 e ) ⋏ λ ( e ∧ e )) . Because ( A , ⋏ λ ) (A,\curlywedge_\lambda) ( A , ⋏ λ ) is non-associative, in order to place all three gravitational and gauge sectors on a uniform footing where the constitutive soldering factors of e e e are grouped into a single monomial ⋏ λ , P m e \curlywedge^m_{\lambda,P}e ⋏ λ , P m e prior to multiplying with Ω ω \Omega_\omega Ω ω (matching the A A A -EHP convention of [21 ] ), we define for P ∈ { b l k , b a l } P\in\{\mathrm{blk},\mathrm{bal}\} P ∈ { blk , bal } in (10) –(12) the left-grouped Cartan-Yang-Mills Lagrangian n n n -form L n , P C Y M ∈ Ω h e q n ( P ; A ) \mathcal{L}^{\mathrm{CYM}}_{n,P}\in\Omega^n_{\mathrm{heq}}(P;A) L n , P CYM ∈ Ω heq n ( P ; A ) by
L n , P C Y M : = Ω ω ⋏ λ ( ( ⋏ λ , P n − 4 e ) ⋏ λ Ω ω ) + ( ⋏ λ , P n − 2 e ) ⋏ λ Ω ω + 1 4 ⋏ λ , P n e , (43) \mathcal{L}^{\mathrm{CYM}}_{n,P}:=\Omega_\omega\curlywedge_\lambda\left((\curlywedge^{n-4}_{\lambda,P}e)\curlywedge_\lambda\Omega_\omega\right)+(\curlywedge^{n-2}_{\lambda,P}e)\curlywedge_\lambda\Omega_\omega+\frac{1}{4}\curlywedge^n_{\lambda,P}e, \tag*{(43)} L n , P CYM := Ω ω ⋏ λ ( ( ⋏ λ , P n − 4 e ) ⋏ λ Ω ω ) + ( ⋏ λ , P n − 2 e ) ⋏ λ Ω ω + 4 1 ⋏ λ , P n e , (43)
where the three terms are respectively the quadratic Yang-Mills term , the left-grouped Einstein-Hilbert-Palatini (EHP) cross-term , and the cosmological term . (Notice that at n = 6 n=6 n = 6 , the ungrouped cross-term 1 2 ( e ∧ e ) ⋏ λ ( ( e ∧ e ) ⋏ λ Ω ω ) \frac{1}{2}(e\wedge e)\curlywedge_\lambda((e\wedge e)\curlywedge_\lambda\Omega_\omega) 2 1 ( e ∧ e ) ⋏ λ (( e ∧ e ) ⋏ λ Ω ω ) does not factor through ⋏ λ , b a l 4 e = 0 \curlywedge^4_{\lambda,\mathrm{bal}}e=0 ⋏ λ , bal 4 e = 0 , whereas in (43) each sector depends directly on ⋏ λ , P m e \curlywedge^m_{\lambda,P}e ⋏ λ , P m e for m ∈ { n − 4 , n − 2 , n } m\in\{n-4,n-2,n\} m ∈ { n − 4 , n − 2 , n } .) Recall from § 2 that when n = 4 n=4 n = 4 (n − 4 = 0 n-4=0 n − 4 = 0 ), our m = 0 m=0 m = 0 convention ( ⋏ λ , P 0 e ) ⋏ λ Ω ω : = Ω ω (\curlywedge^0_{\lambda,P}e)\curlywedge_\lambda\Omega_\omega:=\Omega_\omega ( ⋏ λ , P 0 e ) ⋏ λ Ω ω := Ω ω omits the e e e -factor, so the quadratic Yang-Mills term at n = 4 n=4 n = 4 is simply Ω ω ⋏ λ Ω ω \Omega_\omega\curlywedge_\lambda\Omega_\omega Ω ω ⋏ λ Ω ω , which vanishes identically by ( 2 , 1 ) (2,1) ( 2 , 1 ) -nilpotency of ( A , ⋏ λ ) (A,\curlywedge_\lambda) ( A , ⋏ λ ) !
To compare the two natural parenthesizations of the EHP cross-term at n = 4 n=4 n = 4 , we also consider the right-nested EHP cross-term
L 4 E H P , r i g h t : = e ⋏ λ ( e ⋏ λ Ω ω ) ∈ Ω h e q 4 ( P ; A ) , (44) \mathcal{L}^{\mathrm{EHP,right}}_4:=e\curlywedge_\lambda(e\curlywedge_\lambda\Omega_\omega)\in\Omega^4_{\mathrm{heq}}(P;A), \tag*{(44)} L 4 EHP , right := e ⋏ λ ( e ⋏ λ Ω ω ) ∈ Ω heq 4 ( P ; A ) , (44)
as opposed to the left-grouped EHP cross-term L 4 E H P , l e f t : = ( e ⋏ λ e ) ⋏ λ Ω ω \mathcal{L}^{\mathrm{EHP,left}}_4:=(e\curlywedge_\lambda e)\curlywedge_\lambda\Omega_\omega L 4 EHP , left := ( e ⋏ λ e ) ⋏ λ Ω ω in (43) .
Definition 4.2 (Cartan–Yang–Mills Functional). For n ≥ 4 n \ge4 n ≥ 4 , P ∈ { b l k , b a l } \mathcal{P} \in\{\mathrm{blk},\mathrm{bal}\} P ∈ { blk , bal } , and any bundle trace functional tr ξ : E A → M × R \operatorname{tr}_{\xi}: E_A \to M \times\mathbb{R} tr ξ : E A → M × R induced by ξ ∈ Γ ( E A ∗ ) \xi\in\Gamma(E_A^*) ξ ∈ Γ ( E A ∗ ) , the P \mathcal{P} P -balanced Cartan–Yang–Mills (A A A -CYM) functional on Conn A Θ = 0 , i r r ( P ) \operatorname{Conn}_{A}^{\Theta=0,\mathrm{irr}}(\mathcal{P}) Conn A Θ = 0 , irr ( P ) is
S n , ξ , P C Y M [ e , ω ] : = ∫ M tr ξ ( j ( L n , P C Y M ) ) . (45) S_{n,\xi,\mathcal{P}}^{\mathrm{CYM}}[e,\omega] := \int_M \operatorname{tr}_{\xi}\bigl(j(\mathcal{L}_{n,\mathcal{P}}^{\mathrm{CYM}})\bigr). \tag*{(45)} S n , ξ , P CYM [ e , ω ] := ∫ M tr ξ ( j ( L n , P CYM ) ) . (45)
Dimensional Cutoffs, the 4D First-Bianchi Contraction Identity, and Berger Holonomy Reductions
Before stating our main theorem on S n , ξ , P C Y M [ e , ω ] S_{n,\xi,\mathcal{P}}^{\mathrm{CYM}}[e,\omega] S n , ξ , P CYM [ e , ω ] , we prove a fundamental geometric lemma establishing the exact proportionality between the left-grouped EHP 4-form ( e ∧ × e ) ∧ × Ω ω (e \wedge_{\times} e) \wedge_{\times} \Omega_{\omega} ( e ∧ × e ) ∧ × Ω ω and the right-nested EHP 4-form e ∧ × ( e ∧ × Ω ω ) e \wedge_{\times} (e \wedge_{\times} \Omega_{\omega}) e ∧ × ( e ∧ × Ω ω ) , and showing that on 4-manifolds with U ( 2 ) U(2) U ( 2 ) or S U ( 2 ) SU(2) S U ( 2 ) holonomy both parenthesizations vanish identically on-shell (when torsion Θ e , ω = 0 \Theta_{e,\omega}=0 Θ e , ω = 0 ) while remaining non-zero off-shell (when Θ e , ω ≠ 0 \Theta_{e,\omega} \ne0 Θ e , ω = 0 ).
Lemma 4.3 (4D First-Bianchi Contraction Identity and Proportionality of Left-Grouped and Right-Nested EHP Terms). Let n = k = 4 n=k=4 n = k = 4 and let H ⊂ S O ( 4 ) H \subset SO(4) H ⊂ S O ( 4 ) be either S U ( 2 ) − SU(2)_{-} S U ( 2 ) − (hyperkähler holonomy) or U ( 2 ) = ( U ( 1 ) + × S U ( 2 ) − ) / Z 2 U(2)=(U(1)_{+}\times SU(2)_{-})/\mathbb{Z}_{2} U ( 2 ) = ( U ( 1 ) + × S U ( 2 ) − ) / Z 2 (Kähler 4-manifold holonomy), under the self-dual/anti-self-dual Lie algebra decomposition s o ( 4 ) ≃ s u ( 2 ) + ⊕ s u ( 2 ) − \mathfrak{so}(4)\simeq\mathfrak{su}(2)_{+}\oplus\mathfrak{su}(2)_{-} so ( 4 ) ≃ su ( 2 ) + ⊕ su ( 2 ) − . Let ( e , ω ) ∈ Conn A Θ = 0 ( P ) (e,\omega)\in\operatorname{Conn}_{A}^{\Theta=0}(\mathcal{P}) ( e , ω ) ∈ Conn A Θ = 0 ( P ) be any torsion-free pair (d e + ω ~ ∧ e = 0 de+\widetilde{\omega}\wedge e=0 d e + ω ∧ e = 0 , ω ~ = ω \widetilde{\omega}=\omega ω = ω ). Then:
On-Shell Vanishing of the Left-Grouped EHP Term ( e ∧ × e ) ∧ × Ω ω (e\wedge_{\times}e)\wedge_{\times}\Omega_{\omega} ( e ∧ × e ) ∧ × Ω ω : Under (42) ,
( e ∧ × e ) ∧ × Ω ω = ( 0 , 1 2 [ σ ( e ∧ e ) , Ω ω ] ) ∈ Ω 4 ( P ; [ h , h ] ) . (e\wedge_{\times}e)\wedge_{\times}\Omega_{\omega}
=
\left(0,\frac{1}{2}[\sigma(e\wedge e),\Omega_{\omega}]\right)
\in\Omega^{4}(P;[\mathfrak{h},\mathfrak{h}]). ( e ∧ × e ) ∧ × Ω ω = ( 0 , 2 1 [ σ ( e ∧ e ) , Ω ω ] ) ∈ Ω 4 ( P ; [ h , h ]) .
Because d e + ω ~ ∧ e = 0 de+\widetilde{\omega}\wedge e=0 d e + ω ∧ e = 0 implies the algebraic first Bianchi identity Ω ω ∧ e = 0 \Omega_{\omega}\wedge e=0 Ω ω ∧ e = 0 , the curvature 2-form Ω ω \Omega_{\omega} Ω ω has symmetric Riemann components R a b = R b a R_{ab}=R_{ba} R ab = R ba in the standard basis { ω 1 − , ω 2 − , ω 3 − } \{\omega_{1}^{-},\omega_{2}^{-},\omega_{3}^{-}\} { ω 1 − , ω 2 − , ω 3 − } of Λ 2 R 4 ≃ s u ( 2 ) − \Lambda^{2}\mathbb{R}^{4}\simeq\mathfrak{su}(2)_{-} Λ 2 R 4 ≃ su ( 2 ) − . Since ω a − ∧ ω b − = − 2 δ a b vol 4 \omega_{a}^{-}\wedge\omega_{b}^{-}=-2\delta_{ab}\operatorname{vol}_{4} ω a − ∧ ω b − = − 2 δ ab vol 4 is also symmetric in ( a , b ) (a,b) ( a , b ) while [ T a , T b ] = ϵ a b c T c [T_{a},T_{b}]=\epsilon_{abc}T_{c} [ T a , T b ] = ϵ ab c T c is skew-symmetric, we have
( e ∧ × e ) ∧ × Ω ω = ( 0 , 1 2 [ σ ( e ∧ e ) , Ω ω ] ) = 0 identically on Conn A Θ = 0 ( P ) (46) (e\wedge_{\times}e)\wedge_{\times}\Omega_{\omega}
=
\left(0,\frac{1}{2}[\sigma(e\wedge e),\Omega_{\omega}]\right)
=0
\quad\text{identically on }\operatorname{Conn}_{A}^{\Theta=0}(\mathcal{P}) \tag*{(46)} ( e ∧ × e ) ∧ × Ω ω = ( 0 , 2 1 [ σ ( e ∧ e ) , Ω ω ] ) = 0 identically on Conn A Θ = 0 ( P ) (46)
for both H = S U ( 2 ) − H=SU(2)_{-} H = S U ( 2 ) − and H = U ( 2 ) H=U(2) H = U ( 2 ) . (Off-shell, i.e., for connections ω \omega ω with non-zero torsion Θ e , ω ≠ 0 \Theta_{e,\omega}\ne0 Θ e , ω = 0 whose curvature matrix R a b ≠ R b a R_{ab}\ne R_{ba} R ab = R ba has a non-zero skew-symmetric part, [ σ ( e ∧ e ) , Ω ω ] = − 2 ∑ a , b R a b ϵ a b c T c vol 4 ≠ 0 [\sigma(e\wedge e),\Omega_{\omega}]=-2\sum_{a,b}R_{ab}\epsilon_{abc}T_{c}\operatorname{vol}_{4}\ne0 [ σ ( e ∧ e ) , Ω ω ] = − 2 ∑ a , b R ab ϵ ab c T c vol 4 = 0 does not vanish.)
Exact Proportional Relation with the Right-Nested EHP Term e ∧ × ( e ∧ × Ω ω ) e\wedge_{\times}(e\wedge_{\times}\Omega_{\omega}) e ∧ × ( e ∧ × Ω ω ) : For any subalgebra h ⊂ s o ( k ) \mathfrak{h}\subset\mathfrak{so}(k) h ⊂ so ( k ) with χ H = 0 \chi_{H}=0 χ H = 0 , the right-nested parenthesization (44) satisfies
e ∧ × ( e ∧ × Ω ω ) = ( 0 , − 1 2 σ ( e ∧ ( Ω ω ⋅ e ) ) ) = 1 2 ( e ∧ × e ) ∧ × Ω ω = ( 0 , 1 4 [ σ ( e ∧ e ) , Ω ω ] ) . (47) e\wedge_{\times}(e\wedge_{\times}\Omega_{\omega})
=
\left(0,-\frac{1}{2}\sigma(e\wedge(\Omega_{\omega}\cdot e))\right)
=
\frac{1}{2}(e\wedge_{\times}e)\wedge_{\times}\Omega_{\omega}
=
\left(0,\frac{1}{4}[\sigma(e\wedge e),\Omega_{\omega}]\right). \tag*{(47)} e ∧ × ( e ∧ × Ω ω ) = ( 0 , − 2 1 σ ( e ∧ ( Ω ω ⋅ e )) ) = 2 1 ( e ∧ × e ) ∧ × Ω ω = ( 0 , 4 1 [ σ ( e ∧ e ) , Ω ω ] ) . (47)
Consequently, at n = k = 4 n=k=4 n = k = 4 with H = S U ( 2 ) − H=SU(2)_{-} H = S U ( 2 ) − or U ( 2 ) U(2) U ( 2 ) , both ( e ∧ × e ) ∧ × Ω ω (e\wedge_{\times}e)\wedge_{\times}\Omega_{\omega} ( e ∧ × e ) ∧ × Ω ω and e ∧ × ( e ∧ × Ω ω ) e\wedge_{\times}(e\wedge_{\times}\Omega_{\omega}) e ∧ × ( e ∧ × Ω ω ) vanish identically on the torsion-free space Conn A Θ = 0 ( P ) \operatorname{Conn}_{A}^{\Theta=0}(\mathcal{P}) Conn A Θ = 0 ( P ) and are non-vanishing off-shell when Θ e , ω ≠ 0 \Theta_{e,\omega}\ne0 Θ e , ω = 0 .
Proof. Let us verify both parts directly. In any subalgebra h ⊂ s o ( k ) \mathfrak{h}\subset\mathfrak{so}(k) h ⊂ so ( k ) with χ H = 0 \chi_{H}=0 χ H = 0 , for e ∈ Ω 1 ( P ; R k ) e\in\Omega^{1}(P;\mathbb{R}^{k}) e ∈ Ω 1 ( P ; R k ) and Ω ω ∈ Ω 2 ( P ; h ) \Omega_{\omega}\in\Omega^{2}(P;\mathfrak{h}) Ω ω ∈ Ω 2 ( P ; h ) , (42) gives e ∧ × Ω ω = ( − 1 2 Ω ω ⋅ e , 0 ) ∈ Ω 3 ( P ; R k ⊕ h ) e\wedge_{\times}\Omega_{\omega}=\left(-\frac{1}{2}\Omega_{\omega}\cdot e,0\right)\in\Omega^{3}(P;\mathbb{R}^{k}\oplus\mathfrak{h}) e ∧ × Ω ω = ( − 2 1 Ω ω ⋅ e , 0 ) ∈ Ω 3 ( P ; R k ⊕ h ) , and multiplying on the left by e = ( e , 0 ) e=(e,0) e = ( e , 0 ) gives e ∧ × ( e ∧ × Ω ω ) = ( 0 , − 1 2 σ ( e ∧ ( Ω ω ⋅ e ) ) ) e\wedge_{\times}(e\wedge_{\times}\Omega_{\omega})=\left(0,-\frac{1}{2}\sigma(e\wedge(\Omega_{\omega}\cdot e))\right) e ∧ × ( e ∧ × Ω ω ) = ( 0 , − 2 1 σ ( e ∧ ( Ω ω ⋅ e )) ) . For any basis { T A } \{T_{A}\} { T A } of h ⊂ s o ( k ) \mathfrak{h}\subset\mathfrak{so}(k) h ⊂ so ( k ) orthonormal with respect to ⟨ T A , T B ⟩ s o ( k ) = − 1 2 Tr ( T A T B ) = δ A B \langle T_{A},T_{B}\rangle_{\mathfrak{so}(k)}=-\frac{1}{2}\operatorname{Tr}(T_{A}T_{B})=\delta_{AB} ⟨ T A , T B ⟩ so ( k ) = − 2 1 Tr ( T A T B ) = δ A B , note that since e i e^{i} e i is a 1-form and ( Ω ω ⋅ e ) j = ( Ω ω ) j k ∧ e k (\Omega_{\omega}\cdot e)^{j}=(\Omega_{\omega})_{jk}\wedge e^{k} ( Ω ω ⋅ e ) j = ( Ω ω ) j k ∧ e k is a 3-form, the ( i , j ) (i,j) ( i , j ) -entry of the skew-symmetric matrix e ∧ ( Ω ω ⋅ e ) ∈ Ω 4 ( P ; s o ( k ) ) e\wedge(\Omega_{\omega}\cdot e)\in\Omega^{4}(P;\mathfrak{so}(k)) e ∧ ( Ω ω ⋅ e ) ∈ Ω 4 ( P ; so ( k )) is e i ∧ ( Ω ω ⋅ e ) j − e j ∧ ( Ω ω ⋅ e ) i = e i ∧ ( Ω ω ⋅ e ) j + ( Ω ω ⋅ e ) i ∧ e j e^{i}\wedge(\Omega_{\omega}\cdot e)^{j}-e^{j}\wedge(\Omega_{\omega}\cdot e)^{i}=e^{i}\wedge(\Omega_{\omega}\cdot e)^{j}+(\Omega_{\omega}\cdot e)^{i}\wedge e^{j} e i ∧ ( Ω ω ⋅ e ) j − e j ∧ ( Ω ω ⋅ e ) i = e i ∧ ( Ω ω ⋅ e ) j + ( Ω ω ⋅ e ) i ∧ e j . By (41) , the T A T_{A} T A -component of − 1 2 σ ( e ∧ ( Ω ω ⋅ e ) ) -\frac{1}{2}\sigma(e\wedge(\Omega_{\omega}\cdot e)) − 2 1 σ ( e ∧ ( Ω ω ⋅ e )) is therefore
1 4 Tr ( T A ⋅ ( e ∧ ( Ω ω ⋅ e ) + ( Ω ω ⋅ e ) ∧ e T ) ) = − 1 2 ( T A ) i j ( Ω ω ) j k ∧ e i ∧ e k = − 1 4 [ T A , Ω ω ] i k ∧ e i ∧ e k , \frac{1}{4}\operatorname{Tr}\left(T_{A}\cdot\left(e\wedge(\Omega_{\omega}\cdot e)+(\Omega_{\omega}\cdot e)\wedge e^{T}\right)\right)
=
-\frac{1}{2}(T_{A})_{ij}(\Omega_{\omega})_{jk}\wedge e^{i}\wedge e^{k}
=
-\frac{1}{4}[T_{A},\Omega_{\omega}]_{ik}\wedge e^{i}\wedge e^{k}, 4 1 Tr ( T A ⋅ ( e ∧ ( Ω ω ⋅ e ) + ( Ω ω ⋅ e ) ∧ e T ) ) = − 2 1 ( T A ) ij ( Ω ω ) j k ∧ e i ∧ e k = − 4 1 [ T A , Ω ω ] ik ∧ e i ∧ e k ,
because the symmetric part 1 2 ( T A Ω ω + Ω ω T A ) i k \frac{1}{2}(T_{A}\Omega_{\omega}+\Omega_{\omega}T_{A})_{ik} 2 1 ( T A Ω ω + Ω ω T A ) ik vanishes when contracted with the skew-symmetric 2-form e i ∧ e k e^{i}\wedge e^{k} e i ∧ e k . On the other hand, since ⟨ X , [ Y , Z ] ⟩ s o ( k ) = − ⟨ [ X , Z ] , Y ⟩ s o ( k ) \langle X,[Y,Z]\rangle_{\mathfrak{so}(k)}=-\langle[X,Z],Y\rangle_{\mathfrak{so}(k)} ⟨ X , [ Y , Z ] ⟩ so ( k ) = − ⟨[ X , Z ] , Y ⟩ so ( k ) by ad-invariance of ⟨ ⋅ , ⋅ ⟩ s o ( k ) \langle\cdot,\cdot\rangle_{\mathfrak{so}(k)} ⟨ ⋅ , ⋅ ⟩ so ( k ) , (41) shows that the T A T_{A} T A -component of + 1 4 [ σ ( e ∧ e ) , Ω ω ] +\frac{1}{4}[\sigma(e\wedge e),\Omega_{\omega}] + 4 1 [ σ ( e ∧ e ) , Ω ω ] is
1 4 ⟨ T A , [ σ ( e ∧ e ) , Ω ω ] ⟩ s o ( k ) = − 1 4 ⟨ [ T A , Ω ω ] , σ ( e ∧ e ) ⟩ s o ( k ) = − 1 4 [ T A , Ω ω ] i k ∧ e i ∧ e k . \frac{1}{4}\langle T_{A},[\sigma(e\wedge e),\Omega_{\omega}]\rangle_{\mathfrak{so}(k)}
=
-\frac{1}{4}\langle[T_{A},\Omega_{\omega}],\sigma(e\wedge e)\rangle_{\mathfrak{so}(k)}
=
-\frac{1}{4}[T_{A},\Omega_{\omega}]_{ik}\wedge e^{i}\wedge e^{k}. 4 1 ⟨ T A , [ σ ( e ∧ e ) , Ω ω ] ⟩ so ( k ) = − 4 1 ⟨[ T A , Ω ω ] , σ ( e ∧ e ) ⟩ so ( k ) = − 4 1 [ T A , Ω ω ] ik ∧ e i ∧ e k .
Consequently, for every subalgebra h ⊂ s o ( k ) \mathfrak{h}\subset\mathfrak{so}(k) h ⊂ so ( k ) with χ H = 0 \chi_{H}=0 χ H = 0 , the associative matrix identity ( T A ) i j ( T B ) j k e i ∧ e k = 1 2 [ T A , T B ] i k e i ∧ e k (T_{A})_{ij}(T_{B})_{jk}e^{i}\wedge e^{k}=\frac{1}{2}[T_{A},T_{B}]_{ik}e^{i}\wedge e^{k} ( T A ) ij ( T B ) j k e i ∧ e k = 2 1 [ T A , T B ] ik e i ∧ e k implies (47) .
Now suppose n = k = 4 n=k=4 n = k = 4 and h ⊂ u ( 2 ) = u ( 1 ) + ⊕ s u ( 2 ) − \mathfrak{h}\subset\mathfrak{u}(2)=\mathfrak{u}(1)_{+}\oplus\mathfrak{su}(2)_{-} h ⊂ u ( 2 ) = u ( 1 ) + ⊕ su ( 2 ) − . Since u ( 1 ) + \mathfrak{u}(1)_{+} u ( 1 ) + is central in u ( 2 ) \mathfrak{u}(2) u ( 2 ) , the commutator [ σ ( e ∧ e ) , Ω ω ] [\sigma(e\wedge e),\Omega_{\omega}] [ σ ( e ∧ e ) , Ω ω ] depends only on the s u ( 2 ) − \mathfrak{su}(2)_{-} su ( 2 ) − projections σ − ( e ∧ e ) = ∑ a = 1 3 ω a − ⊗ T a \sigma_{-}(e\wedge e)=\sum_{a=1}^{3}\omega^{-}_{a}\otimes T_{a} σ − ( e ∧ e ) = ∑ a = 1 3 ω a − ⊗ T a and ( Ω ω ) − = ∑ b = 1 3 Ω b − ⊗ T b (\Omega_{\omega})_{-}=\sum_{b=1}^{3}\Omega^{-}_{b}\otimes T_{b} ( Ω ω ) − = ∑ b = 1 3 Ω b − ⊗ T b . On a 4-manifold, the three anti-self-dual 2-forms ω 1 − , ω 2 − , ω 3 − \omega^{-}_{1},\omega^{-}_{2},\omega^{-}_{3} ω 1 − , ω 2 − , ω 3 − satisfy ω a − ∧ ω b − = − 2 δ a b vol 4 \omega^{-}_{a}\wedge\omega^{-}_{b}=-2\delta_{ab}\operatorname{vol}_{4} ω a − ∧ ω b − = − 2 δ ab vol 4 . Decomposing Ω b − = ∑ a = 1 3 R a b ω a − + Ω b , + − \Omega^{-}_{b}=\sum_{a=1}^{3}R_{ab}\omega^{-}_{a}+\Omega^{-}_{b,+} Ω b − = ∑ a = 1 3 R ab ω a − + Ω b , + − where Ω b , + − ∈ Λ + 2 M \Omega^{-}_{b,+}\in\Lambda^{2}_{+}M Ω b , + − ∈ Λ + 2 M is the self-dual part of Ω b − \Omega^{-}_{b} Ω b − (which represents the traceless Ricci tensor Ric 0 ∈ Λ + 2 ⊗ Λ − 2 \operatorname{Ric}_{0}\in\Lambda^{2}_{+}\otimes\Lambda^{2}_{-} Ric 0 ∈ Λ + 2 ⊗ Λ − 2 ), since Λ − 2 ∧ Λ + 2 = 0 \Lambda^{2}_{-}\wedge\Lambda^{2}_{+}=0 Λ − 2 ∧ Λ + 2 = 0 we have ω a − ∧ Ω b − = − 2 R a b vol 4 \omega^{-}_{a}\wedge\Omega^{-}_{b}=-2R_{ab}\operatorname{vol}_{4} ω a − ∧ Ω b − = − 2 R ab vol 4 , and hence
[ σ ( e ∧ e ) , Ω ω ] = ∑ a , b = 1 3 ( ω a − ∧ Ω b − ) ⊗ [ T a , T b ] = − 2 ∑ a , b , c = 1 3 R a b ϵ a b c T c vol 4 . (48) [\sigma(e\wedge e),\Omega_{\omega}]=\sum_{a,b=1}^{3}(\omega^{-}_{a}\wedge\Omega^{-}_{b})\otimes[T_{a},T_{b}]=-2\sum_{a,b,c=1}^{3}R_{ab}\epsilon_{abc}T_{c}\operatorname{vol}_{4}. \tag*{(48)} [ σ ( e ∧ e ) , Ω ω ] = a , b = 1 ∑ 3 ( ω a − ∧ Ω b − ) ⊗ [ T a , T b ] = − 2 a , b , c = 1 ∑ 3 R ab ϵ ab c T c vol 4 . (48)
If ( e , ω ) ∈ Conn A Θ = 0 ( P ) (e,\omega)\in\operatorname{Conn}^{\Theta=0}_{A}(P) ( e , ω ) ∈ Conn A Θ = 0 ( P ) is torsion-free (d e + ω ~ ∧ e = 0 de+\widetilde{\omega}\wedge e=0 d e + ω ∧ e = 0 ), the algebraic first Bianchi identity Ω ω ~ ∧ e = 0 \Omega_{\widetilde{\omega}}\wedge e=0 Ω ω ∧ e = 0 forces the anti-self-dual block R a b ∈ End ( Λ − 2 ) R_{ab}\in\operatorname{End}(\Lambda^{2}_{-}) R ab ∈ End ( Λ − 2 ) of the Riemann curvature operator to be symmetric (R a b = R b a R_{ab}=R_{ba} R ab = R ba : specifically, R a b = W a b − + R 12 δ a b R_{ab}=W^{-}_{ab}+\frac{R}{12}\delta_{ab} R ab = W ab − + 12 R δ ab ). Contracting the symmetric matrix R a b = R b a R_{ab}=R_{ba} R ab = R ba with the antisymmetric structure constants ϵ a b c \epsilon_{abc} ϵ ab c in (48) gives zero identically. Conversely, if ω \omega ω has non-zero torsion Θ e , ω ≠ 0 \Theta_{e,\omega}\ne0 Θ e , ω = 0 , then Ω ω ~ ∧ e = d ω ~ Θ e , ω ≠ 0 \Omega_{\widetilde{\omega}}\wedge e=d_{\widetilde{\omega}}\Theta_{e,\omega}\ne0 Ω ω ∧ e = d ω Θ e , ω = 0 allows R a b − R b a ≠ 0 R_{ab}-R_{ba}\ne0 R ab − R ba = 0 , so (48) is non-zero off-shell. □
Remark 4.4 (Contrast Between n = k = 4 n=k=4 n = k = 4 and n = k = 6 n=k=6 n = k = 6 for U ( k / 2 ) U(k/2) U ( k /2 ) and S U ( k / 2 ) SU(k/2) S U ( k /2 ) ). Lemma 4.3 reveals a striking geometric distinction between k = 4 k=4 k = 4 and k = 6 k=6 k = 6 :
In dimension n = k = 4 n=k=4 n = k = 4 , both the quadratic YM term Ω ω ∧ λ Ω ω = 0 \Omega_{\omega}\wedge_{\lambda}\Omega_{\omega}=0 Ω ω ∧ λ Ω ω = 0 (by ( 2 , 1 ) (2,1) ( 2 , 1 ) -nilpotency) and the EHP term ( e ∧ λ e ) ∧ λ Ω ω = 0 (e\wedge_{\lambda}e)\wedge_{\lambda}\Omega_{\omega}=0 ( e ∧ λ e ) ∧ λ Ω ω = 0 (by (48) and the algebraic first Bianchi identity R a b = R b a R_{ab}=R_{ba} R ab = R ba on Λ − 2 \Lambda^{2}_{-} Λ − 2 ) vanish on-shell on the torsion-free space Conn A Θ = 0 ( P ) \operatorname{Conn}^{\Theta=0}_{A}(P) Conn A Θ = 0 ( P ) for H = U ( 2 ) H=U(2) H = U ( 2 ) and S U ( 2 ) SU(2) S U ( 2 ) , whereas the EHP term ( e ∧ λ e ) ∧ λ Ω ω (e\wedge_{\lambda}e)\wedge_{\lambda}\Omega_{\omega} ( e ∧ λ e ) ∧ λ Ω ω is non-trivial off-shell when torsion Θ e , ω ≠ 0 \Theta_{e,\omega}\ne0 Θ e , ω = 0 is present (and in dimension n = 6 n=6 n = 6 with k = 4 k=4 k = 4 , the pure YM term Ω ω ∧ λ ( ( e ∧ λ e ) ∧ λ Ω ω ) \Omega_{\omega}\wedge_{\lambda}((e\wedge_{\lambda}e)\wedge_{\lambda}\Omega_{\omega}) Ω ω ∧ λ (( e ∧ λ e ) ∧ λ Ω ω ) is non-trivial on Conn A Θ N = 0 , i r r ( P ) \operatorname{Conn}^{\Theta_{N}=0,\mathrm{irr}}_{A}(P) Conn A Θ N = 0 , irr ( P ) when the pulled-back connection on N 4 N^{4} N 4 is torsion-free while transverse curvature and mixed torsion along M 6 → N 4 M^{6}\to N^{4} M 6 → N 4 are non-zero; see Remark 4.1!).
In dimension n = k = 6 n=k=6 n = k = 6 with H = U ( 3 ) H=U(3) H = U ( 3 ) or S U ( 3 ) SU(3) S U ( 3 ) , the surviving Lagrangian is the pure quadratic Yang-Mills 6-form Ω ω ∧ λ ( ( e ∧ λ e ) ∧ λ Ω ω ) = ( 0 , 1 4 [ Ω ω , [ σ ( e ∧ e ) , Ω ω ] ] ) ∈ Ω 6 ( P ; s u ( 3 ) ) \Omega_{\omega}\wedge_{\lambda}((e\wedge_{\lambda}e)\wedge_{\lambda}\Omega_{\omega})=\left(0,\frac{1}{4}[\Omega_{\omega},[\sigma(e\wedge e),\Omega_{\omega}]]\right)\in\Omega^{6}(P;\mathfrak{su}(3)) Ω ω ∧ λ (( e ∧ λ e ) ∧ λ Ω ω ) = ( 0 , 4 1 [ Ω ω , [ σ ( e ∧ e ) , Ω ω ]] ) ∈ Ω 6 ( P ; su ( 3 )) . Unlike dimension 4 (where Λ 0 1 , 1 R 4 \Lambda^{1,1}_{0}\mathbb{R}^{4} Λ 0 1 , 1 R 4 is 3-dimensional and ω a − ∧ ω b − ∝ δ a b vol 4 \omega^{-}_{a}\wedge\omega^{-}_{b}\propto\delta_{ab}\operatorname{vol}_{4} ω a − ∧ ω b − ∝ δ ab vol 4 is diagonal), in dimension 6 the space Λ 0 1 , 1 R 6 ≃ s u ( 3 ) \Lambda^{1,1}_{0}\mathbb{R}^{6}\simeq\mathfrak{su}(3) Λ 0 1 , 1 R 6 ≃ su ( 3 ) is 8-dimensional, and the wedge product of two ( 1 , 1 ) (1,1) ( 1 , 1 ) -forms is a ( 2 , 2 ) (2,2) ( 2 , 2 ) -form (spanning a 15-dimensional space Λ 2 , 2 R 6 \Lambda^{2,2}\mathbb{R}^{6} Λ 2 , 2 R 6 , not a 1-dimensional volume space!). Consequently, for a torsion-free Calabi–Yau 3-fold (H = S U ( 3 ) H=SU(3) H = S U ( 3 ) ) or Kähler 3-fold (H = U ( 3 ) H=U(3) H = U ( 3 ) ) with curvature Ω ω = ∑ A , B = 1 8 W A B η A ⊗ T B \Omega_{\omega}=\sum_{A,B=1}^{8}W_{AB}\eta_{A}\otimes T_{B} Ω ω = ∑ A , B = 1 8 W A B η A ⊗ T B , the intermediate 4-form [ σ ( e ∧ e ) , Ω ω ] = ∑ A , B , C f A B C W B D ( η A ∧ η D ) ⊗ T C ∈ Ω 4 ( P ; s u ( 3 ) ) [\sigma(e\wedge e),\Omega_{\omega}]=\sum_{A,B,C}f_{ABC}W_{BD}(\eta_{A}\wedge\eta_{D})\otimes T_{C}\in\Omega^{4}(P;\mathfrak{su}(3)) [ σ ( e ∧ e ) , Ω ω ] = ∑ A , B , C f A B C W B D ( η A ∧ η D ) ⊗ T C ∈ Ω 4 ( P ; su ( 3 )) is non-zero because η A ∧ η D ∈ Λ 2 , 2 R 6 \eta_{A}\wedge\eta_{D}\in\Lambda^{2,2}\mathbb{R}^{6} η A ∧ η D ∈ Λ 2 , 2 R 6 is not proportional to δ A D \delta_{AD} δ A D , and wedging again with Ω ω \Omega_{\omega} Ω ω produces a non-zero s u ( 3 ) \mathfrak{su}(3) su ( 3 ) -valued 6-form 1 4 [ Ω ω , [ σ ( e ∧ e ) , Ω ω ] ] ≠ 0 \frac{1}{4}[\Omega_{\omega},[\sigma(e\wedge e),\Omega_{\omega}]]\ne0 4 1 [ Ω ω , [ σ ( e ∧ e ) , Ω ω ]] = 0 on-shell!
Theorem 4.5 (Parenthesization Cutoffs and Berger Holonomy Reductions in Cartan-Yang-Mills Theories). Let P → M P\to M P → M be a k k k -proper H H H -bundle over an oriented n n n -manifold M M M (4 ≤ k ≤ n 4\le k\le n 4 ≤ k ≤ n ), where h ⊂ s o ( k 1 ) ⊕ ⋯ ⊕ s o ( k r ) \mathfrak{h}\subset\mathfrak{so}(k_{1})\oplus\cdots\oplus\mathfrak{so}(k_{r}) h ⊂ so ( k 1 ) ⊕ ⋯ ⊕ so ( k r ) properly (k 1 + ⋯ + k r = k k_{1}+\cdots+k_{r}=k k 1 + ⋯ + k r = k ) is a k k k -algebra. Consider the Cartan-Yang-Mills functional S n , ξ C Y M ( e , ω ) S^{\mathrm{CYM}}_{n,\xi}(e,\omega) S n , ξ CYM ( e , ω ) with values in A = R k ⊕ h A=\mathbb{R}^{k}\oplus\mathfrak{h} A = R k ⊕ h under (42) and (43) .
(1) Hierarchical Dimensional Cutoff under Quartic-Block Grouping (P = b l k \mathcal{P}=\mathrm{blk} P = blk ):
For all n ≥ 4 n\ge4 n ≥ 4 , the cosmological term 1 4 ∧ × , b l k n e \frac{1}{4}\wedge^{n}_{\times,\mathrm{blk}}e 4 1 ∧ × , blk n e vanishes identically in Ω n ( P ; A ) \Omega^{n}(P;A) Ω n ( P ; A ) ;
For all n ≥ 6 n\ge6 n ≥ 6 , the EHP cross-term ( ∧ × , b l k n − 2 e ) ∧ λ Ω ω (\wedge^{n-2}_{\times,\mathrm{blk}}e)\wedge_{\lambda}\Omega_{\omega} ( ∧ × , blk n − 2 e ) ∧ λ Ω ω vanishes identically in Ω n ( P ; A ) \Omega^{n}(P;A) Ω n ( P ; A ) , reducing S n , ξ , b l k C Y M [ e , ω ] S^{\mathrm{CYM}}_{n,\xi,\mathrm{blk}}[e,\omega] S n , ξ , blk CYM [ e , ω ] to the pure quadratic Yang-Mills term ∫ M tr ξ ( j ( Ω ω ∧ λ ( ( ∧ × , b l k n − 4 e ) ∧ λ Ω ω ) ) ) \int_{M}\operatorname{tr}^{\xi}(j(\Omega_{\omega}\wedge_{\lambda}((\wedge^{n-4}_{\times,\mathrm{blk}}e)\wedge_{\lambda}\Omega_{\omega}))) ∫ M tr ξ ( j ( Ω ω ∧ λ (( ∧ × , blk n − 4 e ) ∧ λ Ω ω ))) ;
For all n ≥ 8 n\ge8 n ≥ 8 , the entire block-balanced Cartan-Yang-Mills functional S n , ξ , b l k C Y M [ e , ω ] S^{\mathrm{CYM}}_{n,\xi,\mathrm{blk}}[e,\omega] S n , ξ , blk CYM [ e , ω ] vanishes identically on M M M for every trace section ξ ∈ Γ ( E A ∗ ) \xi\in\Gamma(E^{*}_{A}) ξ ∈ Γ ( E A ∗ ) .
(2) Comparison with Recursive Binary-Tree Halving (P = b a l \mathcal{P}=\mathrm{bal} P = bal ): Under exact recursive binary halving (10) , ∧ × , b a l m e \wedge^{m}_{\times,\mathrm{bal}}e ∧ × , bal m e coincides with ∧ × , b l k m e \wedge^{m}_{\times,\mathrm{blk}}e ∧ × , blk m e for 1 ≤ m ≤ 4 1\le m\le4 1 ≤ m ≤ 4 , so S n , ξ , b a l C Y M = S n , ξ , b l k C Y M S^{\mathrm{CYM}}_{n,\xi,\mathrm{bal}}=S^{\mathrm{CYM}}_{n,\xi,\mathrm{blk}} S n , ξ , bal CYM = S n , ξ , blk CYM for all 4 ≤ n ≤ 8 4\le n\le8 4 ≤ n ≤ 8 , and in particular S 8 , ξ , b a l C Y M = 0 S^{\mathrm{CYM}}_{8,\xi,\mathrm{bal}}=0 S 8 , ξ , bal CYM = 0 at n = 8 n=8 n = 8 . For n ≥ 8 n\ge8 n ≥ 8 :
By the all-orders Jacobsthal characterization in Lemma 2.1 (2a), the binary-tree quadratic Yang–Mills term Ω ω ∧ × ( ( ∧ × , b a l n − 4 e ) ∧ × Ω ω ) \Omega_\omega\wedge_{\times} ((\wedge_{\times,\mathrm{bal}}^{n-4}e) \wedge_{\times} \Omega_\omega) Ω ω ∧ × (( ∧ × , bal n − 4 e ) ∧ × Ω ω ) vanishes algebraically for every n ≥ 8 n \ge8 n ≥ 8 such that n − 4 ∉ ⋃ r = 3 ∞ { 2 J r , J r + 1 } n-4 \notin\bigcup_{r=3}^{\infty}\{2J_r,J_{r+1}\} n − 4 ∈ / ⋃ r = 3 ∞ { 2 J r , J r + 1 } where J r = 2 r − ( − 1 ) r 3 J_r = \frac{2^r-(-1)^r}{3} J r = 3 2 r − ( − 1 ) r (so for n ≥ 8 n \ge8 n ≥ 8 , the only dimensions not killed algebraically by even-degree squaring are the shifted Jacobsthal pairs n ∈ { 9 , 10 } ∪ { 14 , 15 } ∪ { 25 , 26 } ∪ { 46 , 47 } ∪ ⋯ n \in\{9,10\} \cup\{14,15\} \cup\{25,26\} \cup\{46,47\} \cup\cdots n ∈ { 9 , 10 } ∪ { 14 , 15 } ∪ { 25 , 26 } ∪ { 46 , 47 } ∪ ⋯ ).
Moreover, because e ∈ Ω 1 ( P ; R k ) e \in\Omega^1(P;\mathbb{R}^k) e ∈ Ω 1 ( P ; R k ) has rank ≤ k \le k ≤ k , Lemma 2.1 (2c) guarantees that ∧ × , b a l m e = 0 \wedge_{\times,\mathrm{bal}}^m e = 0 ∧ × , bal m e = 0 on P P P for all m > k m > k m > k , so S n , ξ , b a l C Y M S_{n,\xi,\mathrm{bal}}^{\mathrm{CYM}} S n , ξ , bal CYM vanishes identically for all n ≥ k + 5 n \ge k+5 n ≥ k + 5 . In particular, for any k ≤ 9 k \le9 k ≤ 9 (which includes all Berger holonomy reductions with k ≤ 7 k \le7 k ≤ 7 below), ∧ × , b a l 10 e = ∧ × , b a l 11 e = 0 \wedge_{\times,\mathrm{bal}}^{10}e = \wedge_{\times,\mathrm{bal}}^{11}e = 0 ∧ × , bal 10 e = ∧ × , bal 11 e = 0 by coframe rank, so S n , ξ , b a l C Y M S_{n,\xi,\mathrm{bal}}^{\mathrm{CYM}} S n , ξ , bal CYM vanishes identically for n = 8 n=8 n = 8 and for all n ≥ 11 n \ge11 n ≥ 11 (and for k = 4 k=4 k = 4 , also at n ∈ { 9 , 10 } n \in\{9,10\} n ∈ { 9 , 10 } ).
(3) Berger Holonomy Reductions (4 ≤ n ≤ \le n \le ≤ n ≤ 7): For any k k k -proper H H H -bundle P → M P \to M P → M with non-empty configuration space Conn A Θ N = 0 , i r r ( P ) ≠ ∅ \operatorname{Conn}_{A}^{\Theta_N=0,\mathrm{irr}}(P) \ne\varnothing Conn A Θ N = 0 , irr ( P ) = ∅ , de Rham irreducibility of ( N , g N ) (N,g_N) ( N , g N ) forces k 1 = k k_1=k k 1 = k and k i > 1 = 0 k_{i>1}=0 k i > 1 = 0 (h ⊊ s o ( k ) \mathfrak h \subsetneq\mathfrak{so}(k) h ⊊ so ( k ) ), and Berger’s holonomy theorem restricts Hol ( ω ~ N ) ⊂ H ⊊ S O ( k ) \operatorname{Hol}(\widetilde{\omega}_N) \subset H \subsetneq SO(k) Hol ( ω N ) ⊂ H ⊊ S O ( k ) to the holonomy groups of Table 1 . (Note that for k = 2 k=2 k = 2 , s o ( 2 ) ≃ u ( 1 ) \mathfrak{so}(2) \simeq\mathfrak{u}(1) so ( 2 ) ≃ u ( 1 ) is 1-dimensional and abelian, so it admits no non-flat proper holonomy subalgebras.) Combining Conn A Θ N = 0 , i r r ( P ) ≠ ∅ \operatorname{Conn}_{A}^{\Theta_N=0,\mathrm{irr}}(P) \ne\varnothing Conn A Θ N = 0 , irr ( P ) = ∅ with the dimensional cutoff n ≤ 7 n \le7 n ≤ 7 (which forces 4 ≤ k ≤ n ≤ 7 4 \le k \le n \le7 4 ≤ k ≤ n ≤ 7 ), S n , ξ , b l k C Y M [ e , ω ] S_{n,\xi,\mathrm{blk}}^{\mathrm{CYM}}[e,\omega] S n , ξ , blk CYM [ e , ω ] can be non-trivial only if 4 ≤ n ≤ 7 4 \le n \le7 4 ≤ n ≤ 7 and Hol ( ω ~ N ) ⊂ H \operatorname{Hol}(\widetilde{\omega}_N) \subset H Hol ( ω N ) ⊂ H on ( N , g N ) (N,g_N) ( N , g N ) belongs to one of the following geometric regimes:
Table 1. Comparison of block-balanced dimensional cutoffs and Berger holonomy regimes between A A A -EHP gravity [21 ] and Cartan-Yang-Mills (A A A -CYM) theory for proper k k k -algebras h ⊊ s o ( k ) \mathfrak h\subsetneq\mathfrak{so}(k) h ⊊ so ( k ) when k = n k=n k = n .
(B1) k = 6 k=6 k = 6 and ( N , g N ) (N,g_N) ( N , g N ) is a 6D Kähler or Calabi–Yau Berger 6-manifold (Hol ( ω ~ N ) = U ( 3 ) \operatorname{Hol}(\widetilde{\omega}_N)=U(3) Hol ( ω N ) = U ( 3 ) or S U ( 3 ) SU(3) S U ( 3 ) ) with n ∈ { 6 , 7 } n \in\{6,7\} n ∈ { 6 , 7 } , where at n = k = 6 n=k=6 n = k = 6 the pure quadratic Yang–Mills term 1 4 [ Ω ω , [ σ ( e ∧ e ) , Ω ω ] ] \frac{1}{4}[\Omega_\omega,[\sigma(e \wedge e),\Omega_\omega]] 4 1 [ Ω ω , [ σ ( e ∧ e ) , Ω ω ]] is non-vanishing on-shell on Conn A Θ = 0 , i r r ( P ) \operatorname{Conn}_{A}^{\Theta=0,\mathrm{irr}}(P) Conn A Θ = 0 , irr ( P ) (Remark 4.4);
(B2) k = 4 k=4 k = 4 and ( N , g N ) (N,g_N) ( N , g N ) is a Kähler or hyperkähler Berger 4-manifold (Hol ( ω ~ N ) = U ( 2 ) \operatorname{Hol}(\widetilde{\omega}_N)=U(2) Hol ( ω N ) = U ( 2 ) or S U ( 2 ) ≃ S p ( 1 ) SU(2)\simeq Sp(1) S U ( 2 ) ≃ S p ( 1 ) , e.g. a K3 surface when N N N is compact): for 5 ≤ n ≤ 7 5 \le n \le7 5 ≤ n ≤ 7 (k = 4 < n k=4<n k = 4 < n ), the pure Yang–Mills term survives on Conn A Θ N = 0 , i r r ( P ) \operatorname{Conn}_{A}^{\Theta_N=0,\mathrm{irr}}(P) Conn A Θ N = 0 , irr ( P ) via transverse curvature and mixed torsion along M n → N 4 M^n \to N^4 M n → N 4 as shown in Remark 4.1; whereas at n = k = 4 n=k=4 n = k = 4 , Lemma 4.3 shows that both Ω ω ∧ × Ω ω = 0 \Omega_\omega\wedge_{\times} \Omega_\omega=0 Ω ω ∧ × Ω ω = 0 and ( e ∧ × e ) ∧ × Ω ω = 0 (e\wedge_{\times}e)\wedge_{\times}\Omega_\omega=0 ( e ∧ × e ) ∧ × Ω ω = 0 vanish on-shell for torsion-free connections by the first Bianchi identity, and ( e ∧ × e ) ∧ × Ω ω ≠ 0 (e\wedge_{\times}e)\wedge_{\times}\Omega_\omega\ne0 ( e ∧ × e ) ∧ × Ω ω = 0 survives only off-shell in the presence of torsion Θ e , ω ≠ 0 \Theta_{e,\omega}\ne0 Θ e , ω = 0 ;
(B3) k = 7 k=7 k = 7 (which since k ≤ n ≤ 7 k \le n \le7 k ≤ n ≤ 7 occurs strictly at the critical dimension n = k = 7 n=k=7 n = k = 7 , where only the pure quadratic Yang–Mills term survives) and ( N , g N ) = ( M , g M ) (N,g_N)=(M,g_M) ( N , g N ) = ( M , g M ) is a Berger 7-manifold with exceptional G 2 G_2 G 2 -holonomy (Hol ( ω ~ N ) = G 2 \operatorname{Hol}(\widetilde{\omega}_N)=G_2 Hol ( ω N ) = G 2 ). As shown in Proposition 4.6 below, at n = k = 7 n=k=7 n = k = 7 the quadratic Yang–Mills 7-form L 7 C Y M \mathcal{L}^{\mathrm{CYM}}_7 L 7 CYM takes values solely in R ⊕ { 0 } \mathbb{R}^{\oplus}\{0\} R ⊕ { 0 } , vanishes on-shell on the torsion-free G 2 G_2 G 2 -curvature space W 77 ⊂ S 2 ( Λ 14 2 ) \mathcal{W}_{77}\subset S^2(\Lambda^2_{14}) W 77 ⊂ S 2 ( Λ 14 2 ) , and is non-vanishing off-shell on Ω 2 ( P ; g 2 ) \Omega^2(P;\mathfrak{g}_2) Ω 2 ( P ; g 2 ) whenever torsion Θ e , ω ≠ 0 \Theta_{e,\omega}\ne0 Θ e , ω = 0 excites the Λ 7 2 ⊗ g 2 \Lambda^2_7\otimes\mathfrak{g}_2 Λ 7 2 ⊗ g 2 or non-Bianchi curvature sectors.
(4) Coadjoint Trace Obstruction and Symmetry Breaking: Under (42) , for k ∈ { 4 , 6 } k\in\{4,6\} k ∈ { 4 , 6 } (χ H = 0 \chi_H=0 χ H = 0 ), e ∧ × e = ( 0 , σ ( e ∧ e ) ) ∈ Ω 2 ( P ; h ) e\wedge_{\times}e=(0,\sigma(e\wedge e))\in\Omega^2(P;\mathfrak h) e ∧ × e = ( 0 , σ ( e ∧ e )) ∈ Ω 2 ( P ; h ) , so the surviving Lagrangian L n C Y M \mathcal{L}^{\mathrm{CYM}}_n L n CYM for 4 ≤ n ≤ 6 4\le n\le6 4 ≤ n ≤ 6 takes values strictly inside the derived subalgebra ( 0 , [ h , h ] ) (0,[\mathfrak h,\mathfrak h]) ( 0 , [ h , h ]) . At n = k = 7 n=k=7 n = k = 7 (H = G 2 H=G_2 H = G 2 ), L 7 C Y M = Ω ω ∧ × ( ( ∧ × , b a l 3 e ) ∧ × Ω ω ) \mathcal{L}^{\mathrm{CYM}}_7=\Omega_\omega\wedge_{\times}((\wedge_{\times,\mathrm{bal}}^3e)\wedge_{\times}\Omega_\omega) L 7 CYM = Ω ω ∧ × (( ∧ × , bal 3 e ) ∧ × Ω ω ) takes values in R ⊕ { 0 } \mathbb{R}^{\oplus}\{0\} R ⊕ { 0 } , where ( R 7 ) G 2 = { 0 } (\mathbb{R}^7)^{G_2}=\{0\} ( R 7 ) G 2 = { 0 } . Consequently, any constant H H H -invariant linear functional τ 0 ∈ ( A ∗ ) H \tau_0\in(A^*)^H τ 0 ∈ ( A ∗ ) H annihilates L n C Y M \mathcal{L}^{\mathrm{CYM}}_n L n CYM identically, so non-triviality of S n , ξ C Y M [ e , ω ] S_{n,\xi}^{\mathrm{CYM}}[e,\omega] S n , ξ CYM [ e , ω ] requires a non-zero coadjoint background section ξ ∈ Γ ( E A ∗ ) \xi\in\Gamma(E_A^*) ξ ∈ Γ ( E A ∗ ) with ξ ∣ E [ h , h ] ⊕ E R k ≠ 0 \xi|_{E_{[\mathfrak h,\mathfrak h]}\oplus E_{\mathbb{R}^k}}\ne0 ξ ∣ E [ h , h ] ⊕ E R k = 0 , reducing the internal H H H -symmetry along ξ \xi ξ to H ξ ⊊ H H_\xi\subsetneq H H ξ ⊊ H .
Proof. Part (1) follows immediately from Lemma 2.1(1) and (2c): since ( A , ∧ × ) (A,\wedge_{\times}) ( A , ∧ × ) is skew-commutative, ∧ × , b l k m e = 0 \wedge_{\times,\mathrm{blk}}^m e=0 ∧ × , blk m e = 0 for all m ≥ 4 m\ge4 m ≥ 4 (and for all m > k m>k m > k ). Thus the cosmological term (m = n m=n m = n ) vanishes for n ≥ 4 n\ge4 n ≥ 4 , the EHP term (m = n − 2 m=n-2 m = n − 2 ) vanishes for n − 2 ≥ 4 ⟺ n ≥ 6 n-2\ge4\Longleftrightarrow n\ge6 n − 2 ≥ 4 ⟺ n ≥ 6 (and n − 2 > k n-2>k n − 2 > k ), and the quadratic Yang–Mills term (m = n − 4 m=n-4 m = n − 4 ) vanishes for n − 4 ≥ 4 ⟺ n ≥ 8 n-4\ge4\Longleftrightarrow n\ge8 n − 4 ≥ 4 ⟺ n ≥ 8 (and n − 4 > k n-4>k n − 4 > k ). Part (2) follows from Lemma 2.1 (2a)–(2c): ∧ × , b a l m e = ∧ × , b l k m e \wedge_{\times,\mathrm{bal}}^m e=\wedge_{\times,\mathrm{blk}}^m e ∧ × , bal m e = ∧ × , blk m e for m ≤ 4 m\le4 m ≤ 4 (covering n ≤ 8 n\le8 n ≤ 8 ), ∧ × , b a l m e = 0 \wedge_{\times,\mathrm{bal}}^m e=0 ∧ × , bal m e = 0 algebraically for all m ≥ 4 m\ge4 m ≥ 4 with m ∉ ⋃ r = 3 ∞ { 2 J r , J r + 1 } m\notin\bigcup_{r=3}^{\infty}\{2J_r,J_{r+1}\} m ∈ / ⋃ r = 3 ∞ { 2 J r , J r + 1 } , and ∧ × , b a l m e = 0 \wedge_{\times,\mathrm{bal}}^m e=0 ∧ × , bal m e = 0 by coframe rank for all m > k m>k m > k (so for k ≤ 9 k\le9 k ≤ 9 , m = 10 , 11 m=10,11 m = 10 , 11 also vanish, eliminating all m ≥ 7 m\ge7 m ≥ 7 , i.e., n ≥ 11 n\ge11 n ≥ 11 ).
For part (3), on Conn A Θ N = 0 , i r r ( P ) ≠ ∅ \operatorname{Conn}_{A}^{\Theta_N=0,\mathrm{irr}}(P)\ne\varnothing Conn A Θ N = 0 , irr ( P ) = ∅ , the pullback connection ω ~ N : = 1 2 ( ι ∘ κ ) ∗ ω \widetilde{\omega}_N:=\frac{1}{2}(\iota\circ\kappa)^*\omega ω N := 2 1 ( ι ∘ κ ) ∗ ω on P N ⊂ F N P_N\subset FN P N ⊂ F N is the Levi-Civita connection of the locally irreducible, locally non-symmetric Riemannian metric g N g_N g N on the simply connected k k k -manifold N N N (4 ≤ k ≤ n ≤ 7 4 \le k \le n \le7 4 ≤ k ≤ n ≤ 7 ), with holonomy group Hol ( ω ~ N ) ⊂ H ⊂ S O ( k 1 ) × ⋯ × S O ( k r ) ⊂ S O ( k ) \operatorname{Hol}(\widetilde{\omega}_N) \subset H \subset SO(k_1) \times\cdots\times SO(k_r) \subset SO(k) Hol ( ω N ) ⊂ H ⊂ S O ( k 1 ) × ⋯ × S O ( k r ) ⊂ S O ( k ) [20 , 21 ] . By de Rham’s decomposition theorem [10 , 38 ] , irreducibility of ( N , g N ) (N,g_N) ( N , g N ) forces k 1 = k k_1=k k 1 = k and k j > 1 = 0 k_{j>1}=0 k j > 1 = 0 , so Hol ( ω ~ N ) ⊂ H ⊊ S O ( k ) \operatorname{Hol}(\widetilde{\omega}_N) \subset H \subsetneq SO(k) Hol ( ω N ) ⊂ H ⊊ S O ( k ) is an irreducible proper holonomy subgroup. Berger’s holonomy classification [6 , 7 , 25 ] combined with 4 ≤ k ≤ n ≤ 7 4 \le k \le n \le7 4 ≤ k ≤ n ≤ 7 , Lemma 4.3, and Proposition 4.6 restricts Hol ( ω ~ N ) \operatorname{Hol}(\widetilde{\omega}_N) Hol ( ω N ) to Table 1 , yielding (B1)–(B3).
Finally, for part (4), when × H = 0 \times_H=0 × H = 0 , e ∧ × e = ( 0 , σ ( e ∧ e ) ) ∈ Ω 2 ( P ; h ) e\mathbin{\wedge_{\times}}e=(0,\sigma(e\wedge e))\in\Omega^2(P;\mathfrak h) e ∧ × e = ( 0 , σ ( e ∧ e )) ∈ Ω 2 ( P ; h ) and Ω ω ∈ Ω 2 ( P ; h ) \Omega_\omega\in\Omega^2(P;\mathfrak h) Ω ω ∈ Ω 2 ( P ; h ) , so both ( e ∧ × e ) ∧ × Ω ω (e\mathbin{\wedge_{\times}}e)\mathbin{\wedge_{\times}}\Omega_\omega ( e ∧ × e ) ∧ × Ω ω (n = 4 n=4 n = 4 ) and Ω ω ∧ × ( ( e ∧ × e ) ∧ × Ω ω ) \Omega_\omega\mathbin{\wedge_{\times}}((e\mathbin{\wedge_{\times}}e)\mathbin{\wedge_{\times}}\Omega_\omega) Ω ω ∧ × (( e ∧ × e ) ∧ × Ω ω ) (n = 6 n=6 n = 6 ) take values in ( 0 , [ h , h ] ) (0,[\mathfrak h,\mathfrak h]) ( 0 , [ h , h ]) (while at n = 5 n=5 n = 5 the surviving terms Ω ω ∧ × ( e ∧ × Ω ω ) \Omega_\omega\mathbin{\wedge_{\times}}(e\mathbin{\wedge_{\times}}\Omega_\omega) Ω ω ∧ × ( e ∧ × Ω ω ) and ( Λ × , b a l 3 e ) ∧ × Ω ω (\Lambda^3_{\times,\mathrm{bal}}e)\mathbin{\wedge_{\times}}\Omega_\omega ( Λ × , bal 3 e ) ∧ × Ω ω take values in R k ⊕ { 0 } \mathbb{R}^k\oplus\{0\} R k ⊕ { 0 } , where ( R k ) H = { 0 } (\mathbb{R}^k)^H=\{0\} ( R k ) H = { 0 } by irreducibility of H ⊂ S O ( k ) H\subset SO(k) H ⊂ S O ( k ) ). At n = k = 7 n=k=7 n = k = 7 (H = G 2 H=G_2 H = G 2 ), Proposition 4.6 shows that Λ × , b a l 3 e = ( v ( 3 ) , 0 ) ∈ Ω 3 ( P ; R 7 ⊕ { 0 } ) \Lambda^3_{\times,\mathrm{bal}}e=(v^{(3)},0)\in\Omega^3(P;\mathbb{R}^7\oplus\{0\}) Λ × , bal 3 e = ( v ( 3 ) , 0 ) ∈ Ω 3 ( P ; R 7 ⊕ { 0 }) , and multiplying twice with Ω ω = ( 0 , Ω ω ) \Omega_\omega=(0,\Omega_\omega) Ω ω = ( 0 , Ω ω ) via (42) yields L 7 C Y M = ( − 1 4 Ω ω ⋅ ( Ω ω ⋅ v ( 3 ) ) , 0 ) ∈ Ω 7 ( P ; R 7 ⊕ { 0 } ) \mathcal{L}^{\mathrm{CYM}}_7=(-\frac14\Omega_\omega\cdot(\Omega_\omega\cdot v^{(3)}),0)\in\Omega^7(P;\mathbb{R}^7\oplus\{0\}) L 7 CYM = ( − 4 1 Ω ω ⋅ ( Ω ω ⋅ v ( 3 ) ) , 0 ) ∈ Ω 7 ( P ; R 7 ⊕ { 0 }) . Since any τ 0 ∈ ( A ∗ ) H \tau_0\in(A^*)^H τ 0 ∈ ( A ∗ ) H vanishes on [ h , h ] [\mathfrak h,\mathfrak h] [ h , h ] and ( R k ) H = { 0 } (\mathbb{R}^k)^H=\{0\} ( R k ) H = { 0 } , tr τ 0 \operatorname{tr}_{\tau_0} tr τ 0 annihilates these terms identically. □
We now establish the exact representation-theoretic structure and on-shell/off-shell dichotomy of L 7 C Y M \mathcal{L}^{\mathrm{CYM}}_7 L 7 CYM at the critical dimension n = k = 7 n=k=7 n = k = 7 .
Proposition 4.6 (Explicit G 2 G_2 G 2 -Contraction Formulas and On-Shell/Off-Shell Dichotomy of L 7 C Y M \mathcal{L}^{\mathrm{CYM}}_7 L 7 CYM at n = k = 7 n=k=7 n = k = 7 ). Let n = k = 7 n=k=7 n = k = 7 and H = G 2 ⊂ S O ( 7 ) H=G_2\subset SO(7) H = G 2 ⊂ S O ( 7 ) acting on A = R 7 ⊕ g 2 A=\mathbb{R}^7\oplus\mathfrak g_2 A = R 7 ⊕ g 2 under (42) , where ( u × G 2 v ) a = φ a b c u b v c (u\times_{G_2}v)_a=\varphi_{abc}u^bv^c ( u × G 2 v ) a = φ ab c u b v c is induced by the associative 3-form φ 0 = 1 6 φ a b c e e a ∧ e b ∧ e c \varphi_0=\frac16\varphi_{abce}e^a\wedge e^b\wedge e^c φ 0 = 6 1 φ ab ce e a ∧ e b ∧ e c , with dual coassociative 4-form ∗ φ 0 = 1 24 ( ∗ φ ) a b c d e e a ∧ e b ∧ e c ∧ e d *\varphi_0=\frac1{24}(*\varphi)_{abcde}e^a\wedge e^b\wedge e^c\wedge e^d ∗ φ 0 = 24 1 ( ∗ φ ) ab c d e e a ∧ e b ∧ e c ∧ e d (satisfying φ a b c ( ∗ φ ) b c d e = 4 φ a d e \varphi^{abc}(*\varphi)_{bcde}=4\varphi^a{}_{de} φ ab c ( ∗ φ ) b c d e = 4 φ a d e under the Fano-plane orientation) and orthogonal projection σ : Λ 2 R 7 ↠ Λ 14 2 = g 2 \sigma:\Lambda^2\mathbb{R}^7\twoheadrightarrow\Lambda^2_{14}=\mathfrak g_2 σ : Λ 2 R 7 ↠ Λ 14 2 = g 2 given on vectors u , v ∈ R 7 u,v\in\mathbb{R}^7 u , v ∈ R 7 by
σ ( u ∧ v ) a b = 2 3 ( u a v b − u b v a ) − 1 3 ( ∗ φ ) a b c d u c v d = 2 3 ( u a v b − u b v a ) − 1 6 ( ∗ φ ) a b c d ( u c v d − u d v c ) (49) \sigma(u\wedge v)_{ab}=\frac23(u_av_b-u_bv_a)-\frac13(*\varphi)_{abcd}u^cv^d=\frac23(u_av_b-u_bv_a)-\frac16(*\varphi)_{abcd}(u^cv^d-u^dv^c) \tag*{(49)} σ ( u ∧ v ) ab = 3 2 ( u a v b − u b v a ) − 3 1 ( ∗ φ ) ab c d u c v d = 3 2 ( u a v b − u b v a ) − 6 1 ( ∗ φ ) ab c d ( u c v d − u d v c ) (49) (so that φ a b c σ ( u ∧ v ) b c = 0 \varphi^{abc}\sigma(u\wedge v)_{bc}=0 φ ab c σ ( u ∧ v ) b c = 0 ). For an orthonormal coframe e = ( e , 0 ) ∈ Ω 1 ( P ; R 7 ⊕ g 2 ) e=(e,0)\in\Omega^1(P;\mathbb{R}^7\oplus\mathfrak g_2) e = ( e , 0 ) ∈ Ω 1 ( P ; R 7 ⊕ g 2 ) with volume form vol 7 = e 1 ∧ ⋯ ∧ e 7 \operatorname{vol}_7=e^1\wedge\cdots\wedge e^7 vol 7 = e 1 ∧ ⋯ ∧ e 7 :
(1) Quadratic Soldering Form: Evaluating the exterior product e ∧ × e = ∑ c , d = 1 7 ( e c ∧ e d ) ⊗ ( ( ∂ c , 0 ) × ( ∂ d , 0 ) ) e\mathbin{\wedge_{\times}}e=\sum_{c,d=1}^7(e^c\wedge e^d)\otimes((\partial_c,0)\mathbin{\times}(\partial_d,0)) e ∧ × e = ∑ c , d = 1 7 ( e c ∧ e d ) ⊗ (( ∂ c , 0 ) × ( ∂ d , 0 )) via (6) , (42) , and (49) gives e ∧ × e = ( e ∧ G 2 e , σ ( e ∧ e ) ) ∈ Ω 2 ( P ; R 7 ⊕ g 2 ) e\mathbin{\wedge_{\times}}e=(e\wedge_{G_2}e,\sigma(e\wedge e))\in\Omega^2(P;\mathbb{R}^7\oplus\mathfrak g_2) e ∧ × e = ( e ∧ G 2 e , σ ( e ∧ e )) ∈ Ω 2 ( P ; R 7 ⊕ g 2 ) , with ( e ∧ G 2 e ) b = φ b c d e c ∧ e d (e\wedge_{G_2}e)_b=\varphi_{bcd}e^c\wedge e^d ( e ∧ G 2 e ) b = φ b c d e c ∧ e d and
σ ( e ∧ e ) a b = ∑ c , d = 1 7 σ ( ∂ c ∧ ∂ d ) a b e c ∧ e d = 4 3 e a ∧ e b − 1 3 ( ∗ φ ) a b c d e c ∧ e d ( φ a b c σ ( e ∧ e ) b c = 0 ) . (50) \sigma(e\wedge e)_{ab}=\sum_{c,d=1}^{7}\sigma(\partial_c\wedge\partial_d)_{ab}e^c\wedge e^d=\frac43e_a\wedge e_b-\frac13(*\varphi)_{abcd}e^c\wedge e^d \qquad(\varphi^{abc}\sigma(e\wedge e)_{bc}=0). \tag*{(50)} σ ( e ∧ e ) ab = c , d = 1 ∑ 7 σ ( ∂ c ∧ ∂ d ) ab e c ∧ e d = 3 4 e a ∧ e b − 3 1 ( ∗ φ ) ab c d e c ∧ e d ( φ ab c σ ( e ∧ e ) b c = 0 ) . (50) (2) Cubic Soldering Form and Schur Vanishing of the g 2 \mathfrak g_2 g 2 -Sector: Multiplying e ∧ × e e\mathbin{\wedge_{\times}}e e ∧ × e (degree 2) on the right by e = ( e , 0 ) e=(e,0) e = ( e , 0 ) (degree 1) via (6) and (42) gives Λ × , b a l 3 e = ( e ∧ × e ) ∧ × e = ( v ( 3 ) , B ( 3 ) ) ∈ Ω 3 ( P ; R 7 ⊕ g 2 ) \Lambda^3_{\times,\mathrm{bal}}e=(e\mathbin{\wedge_{\times}}e)\mathbin{\wedge_{\times}}e=(v^{(3)},B^{(3)})\in\Omega^3(P;\mathbb{R}^7\oplus\mathfrak g_2) Λ × , bal 3 e = ( e ∧ × e ) ∧ × e = ( v ( 3 ) , B ( 3 ) ) ∈ Ω 3 ( P ; R 7 ⊕ g 2 ) , where:
In the R 7 \mathbb{R}^7 R 7 -component , since ( v 1 , X 1 ) ∧ × ( e , 0 ) (v_1,X_1)\mathbin{\wedge_{\times}}(e,0) ( v 1 , X 1 ) ∧ × ( e , 0 ) has R 7 \mathbb{R}^7 R 7 -part 1 2 σ ( e ∧ e ) a b ∧ e b + ( ( e ∧ G 2 e ) ∧ G 2 e ) a \frac12\sigma(e\wedge e)_{ab}\wedge e^b+((e\wedge_{G_2}e)\wedge_{G_2}e)_a 2 1 σ ( e ∧ e ) ab ∧ e b + (( e ∧ G 2 e ) ∧ G 2 e ) a , the G 2 G_2 G 2 contraction identities 1 2 σ ( e ∧ e ) a b ∧ e b = − 1 6 ( ∗ φ ) a b c d e e b ∧ e c ∧ e d = − i ∂ a ( ∗ φ 0 ) \frac12\sigma(e\wedge e)_{ab}\wedge e^b=-\frac16(*\varphi)_{abcde}e^b\wedge e^c\wedge e^d=-\mathrm{i}_{\partial_a}(*\varphi_0) 2 1 σ ( e ∧ e ) ab ∧ e b = − 6 1 ( ∗ φ ) ab c d e e b ∧ e c ∧ e d = − i ∂ a ( ∗ φ 0 ) (from (50) since e b ∧ e b = 0 e_b\wedge e^b=0 e b ∧ e b = 0 ) and $((e\wedge_{G_2}e)\wedge_{G_2}e)_a=\varphi_{abe}\varphi_{ecd}e^c\wedge
e d ∧ e b = − ( ∗ φ ) a b c d e b ∧ e c ∧ e d = − 6 ı ∂ a ( ∗ φ 0 ) e^{d} \wedge e^{b}=-(\ast\varphi)_{abcd}e^{b}\wedge e^{c}\wedge e^{d}=-6\,\imath_{\partial_a}(\ast\varphi_0) e d ∧ e b = − ( ∗ φ ) ab c d e b ∧ e c ∧ e d = − 6 ∂ a ( ∗ φ 0 ) sum with the same negative sign to give
( v ( 3 ) ) a = 1 2 σ ( e ∧ e ) a b ∧ e b + ( ( e ∧ × G 2 e ) ∧ × G 2 e ) a = − 7 6 ( ∗ φ ) a b c d e b ∧ e c ∧ e d = − 7 ı ∂ a ( ∗ φ 0 ) . (51) (v^{(3)})_a=\frac{1}{2}\sigma(e\wedge e)_{ab}\wedge e^b+\bigl((e\wedge_{\times G_2}e)\wedge_{\times G_2}e\bigr)_a=-\frac{7}{6}(\ast\varphi)_{abcd}e^b\wedge e^c\wedge e^d=-7\imath_{\partial_a}(\ast\varphi_0). \tag*{(51)} ( v ( 3 ) ) a = 2 1 σ ( e ∧ e ) ab ∧ e b + ( ( e ∧ × G 2 e ) ∧ × G 2 e ) a = − 6 7 ( ∗ φ ) ab c d e b ∧ e c ∧ e d = − 7 ∂ a ( ∗ φ 0 ) . (51)
In particular, v ( 3 ) ≠ 0 v^{(3)}\ne0 v ( 3 ) = 0 (and even if one sets × G 2 = 0 \times_{G_2}=0 × G 2 = 0 , the first term alone gives 1 2 σ ( e ∧ e ) a b ∧ e b = − ı ∂ a ( ∗ φ 0 ) ≠ 0 \frac{1}{2}\sigma(e\wedge e)_{ab}\wedge e^b=-\imath_{\partial_a}(\ast\varphi_0)\ne0 2 1 σ ( e ∧ e ) ab ∧ e b = − ∂ a ( ∗ φ 0 ) = 0 ).
In the g 2 \mathfrak{g}_2 g 2 -component, B e ( 3 ) = σ ( ( e ∧ × G 2 e ) ∧ e ) ∈ Ω 3 ( P ; g 2 ) B_e^{(3)}=\sigma((e\wedge_{\times G_2}e)\wedge e)\in\Omega^3(P;\mathfrak{g}_2) B e ( 3 ) = σ (( e ∧ × G 2 e ) ∧ e ) ∈ Ω 3 ( P ; g 2 ) defines a G 2 G_2 G 2 -equivariant linear map Λ 3 R 7 → Λ 14 2 = g 2 \Lambda^3\mathbb{R}^7\to\Lambda^2_{14}=\mathfrak{g}_2 Λ 3 R 7 → Λ 14 2 = g 2 . Because Λ 3 R 7 = Λ 1 3 ⊕ Λ 7 3 ⊕ Λ 27 3 \Lambda^3\mathbb{R}^7=\Lambda^3_1\oplus\Lambda^3_7\oplus\Lambda^3_{27} Λ 3 R 7 = Λ 1 3 ⊕ Λ 7 3 ⊕ Λ 27 3 contains no Λ 14 \Lambda_{14} Λ 14 irreducible summand, Schur’s lemma (and direct expansion of (49) ) yields
B e ( 3 ) = σ ( ( e ∧ × G 2 e ) ∧ e ) = 0 identically in Ω 3 ( P ; g 2 ) . (52) B_e^{(3)}=\sigma((e\wedge_{\times G_2}e)\wedge e)=0\quad\text{identically in }\Omega^3(P;\mathfrak{g}_2). \tag*{(52)} B e ( 3 ) = σ (( e ∧ × G 2 e ) ∧ e ) = 0 identically in Ω 3 ( P ; g 2 ) . (52)
On-Shell/Off-Shell Dichotomy of L 7 C Y M \mathcal{L}^{\mathrm{CYM}}_7 L 7 CYM
For any curvature 2 2 2 -form Ω ω = ( 0 , Ω ω ) ∈ Ω 2 ( P ; g 2 ) \Omega_\omega=(0,\Omega_\omega)\in\Omega^2(P;\mathfrak{g}_2) Ω ω = ( 0 , Ω ω ) ∈ Ω 2 ( P ; g 2 ) , the Cartan-Yang-Mills 7 7 7 -form is
L 7 C Y M = Ω ω ∧ × ( ( Λ × , b a l 3 ) ∧ × Ω ω ) = ( − 1 4 Ω ω ⋅ ( Ω ω ⋅ v ( 3 ) ) , 0 ) ∈ Ω 7 ( P ; R 7 ⊕ { 0 } ) . (53) \mathcal{L}^{\mathrm{CYM}}_7=\Omega_\omega\wedge_{\times}\bigl((\Lambda^3_{\times,\mathrm{bal}})\wedge_{\times}\Omega_\omega\bigr)=\left(-\frac{1}{4}\Omega_\omega\cdot(\Omega_\omega\cdot v^{(3)}),0\right)\in\Omega^7(P;\mathbb{R}^7\oplus\{0\}). \tag*{(53)} L 7 CYM = Ω ω ∧ × ( ( Λ × , bal 3 ) ∧ × Ω ω ) = ( − 4 1 Ω ω ⋅ ( Ω ω ⋅ v ( 3 ) ) , 0 ) ∈ Ω 7 ( P ; R 7 ⊕ { 0 }) . (53)
Decomposing the 2 2 2 -form factors of Ω ω ∈ Ω 2 ( P ; g 2 ) \Omega_\omega\in\Omega^2(P;\mathfrak{g}_2) Ω ω ∈ Ω 2 ( P ; g 2 ) along Λ 2 R 7 = Λ 7 2 ⊕ Λ 14 2 \Lambda^2\mathbb{R}^7=\Lambda^2_7\oplus\Lambda^2_{14} Λ 2 R 7 = Λ 7 2 ⊕ Λ 14 2 :
(a) Ricci-Type Contraction on Λ 14 2 ⊗ g 2 \Lambda^2_{14}\otimes\mathfrak{g}_2 Λ 14 2 ⊗ g 2 and On-Shell Vanishing on W 77 \mathcal{W}_{77} W 77 : For any Ω ω = 1 2 R b c , d f e d ∧ e f ⊗ E b c ∈ Γ ( P , Λ 14 2 T ∗ M ⊗ g 2 ) \Omega_\omega=\frac{1}{2}R_{bc,df}e^d\wedge e^f\otimes E_{bc}\in\Gamma(P,\Lambda^2_{14}T^*M\otimes\mathfrak{g}_2) Ω ω = 2 1 R b c , df e d ∧ e f ⊗ E b c ∈ Γ ( P , Λ 14 2 T ∗ M ⊗ g 2 ) , the identity ( Ω ω ) b c ∧ ∗ φ 0 = 0 (\Omega_\omega)_{bc}\wedge\ast\varphi_0=0 ( Ω ω ) b c ∧ ∗ φ 0 = 0 implies
( Ω ω ⋅ v ( 3 ) ) b = − 7 ∑ c = 1 7 ( Ω ω ) b c ∧ ı ∂ c ( ∗ φ 0 ) = 7 ∑ c , d = 1 7 R b c , c d e d ∧ ∗ φ 0 ∈ Ω 5 ( P ; R 7 ) . (54) (\Omega_\omega\cdot v^{(3)})_b=-7\sum_{c=1}^{7}(\Omega_\omega)_{bc}\wedge\imath_{\partial_c}(\ast\varphi_0)=7\sum_{c,d=1}^{7}R_{bc,cd}e^d\wedge\ast\varphi_0\in\Omega^5(P;\mathbb{R}^7). \tag*{(54)} ( Ω ω ⋅ v ( 3 ) ) b = − 7 c = 1 ∑ 7 ( Ω ω ) b c ∧ ∂ c ( ∗ φ 0 ) = 7 c , d = 1 ∑ 7 R b c , c d e d ∧ ∗ φ 0 ∈ Ω 5 ( P ; R 7 ) . (54)
In particular, for every torsion-free G 2 G_2 G 2 -connection ( e , ω ) ∈ Conn A Θ = 0 ( P ) (e,\omega)\in\operatorname{Conn}^{\Theta=0}_A(P) ( e , ω ) ∈ Conn A Θ = 0 ( P ) (whose curvature lies in the torsion-free G 2 G_2 G 2 Riemann curvature space W 77 = S 2 ( Λ 14 2 ) ∩ ker ( ∧ ) ⊂ Λ 14 2 ⊗ g 2 \mathcal{W}_{77}=S^2(\Lambda^2_{14})\cap\ker(\wedge)\subset\Lambda^2_{14}\otimes\mathfrak{g}_2 W 77 = S 2 ( Λ 14 2 ) ∩ ker ( ∧ ) ⊂ Λ 14 2 ⊗ g 2 , where the algebraic first Bianchi identity and G 2 G_2 G 2 -holonomy force Ric b d = ∑ c = 1 7 R b c , c d = 0 \operatorname{Ric}_{bd}=\sum_{c=1}^{7}R_{bc,cd}=0 Ric b d = ∑ c = 1 7 R b c , c d = 0 ), we have Ω ω ⋅ v ( 3 ) = 0 \Omega_\omega\cdot v^{(3)}=0 Ω ω ⋅ v ( 3 ) = 0 in Ω 5 ( P ; R 7 ) \Omega^5(P;\mathbb{R}^7) Ω 5 ( P ; R 7 ) and hence L 7 C Y M = 0 \mathcal{L}^{\mathrm{CYM}}_7=0 L 7 CYM = 0 on-shell;
(b) Off-Shell Non-Vanishing on Ω 2 ( P ; g 2 ) \Omega^2(P;\mathfrak{g}_2) Ω 2 ( P ; g 2 ) : Whenever torsion Θ e , ω ≠ 0 \Theta_{e,\omega}\ne0 Θ e , ω = 0 (or a general g 2 \mathfrak{g}_2 g 2 -connection ω \omega ω ) excites the Λ 7 2 ⊗ g 2 \Lambda^2_7\otimes\mathfrak{g}_2 Λ 7 2 ⊗ g 2 curvature component or violates the first Bianchi identity on Λ 14 2 ⊗ g 2 \Lambda^2_{14}\otimes\mathfrak{g}_2 Λ 14 2 ⊗ g 2 , L 7 C Y M \mathcal{L}^{\mathrm{CYM}}_7 L 7 CYM is non-vanishing in Ω 7 ( P ; R 7 ⊕ { 0 } ) \Omega^7(P;\mathbb{R}^7\oplus\{0\}) Ω 7 ( P ; R 7 ⊕ { 0 }) .
Proof. Part (1) and the R 7 \mathbb{R}^7 R 7 -contraction (51) in part (2) follow directly from (42) , (49) , (50) , and the identity φ e a b φ e c d = δ a c δ b d − δ a d δ b c − ( ∗ φ ) a b c d \varphi_{eab}\varphi_{ecd}=\delta_{ac}\delta_{bd}-\delta_{ad}\delta_{bc}-(\ast\varphi)_{abcd} φ e ab φ ec d = δ a c δ b d − δ a d δ b c − ( ∗ φ ) ab c d : contracting with e c ∧ e d ∧ e b e^c\wedge e^d\wedge e^b e c ∧ e d ∧ e b annihilates δ a c δ b d − δ a d δ b c \delta_{ac}\delta_{bd}-\delta_{ad}\delta_{bc} δ a c δ b d − δ a d δ b c (since e b ∧ e b = 0 e^b\wedge e^b=0 e b ∧ e b = 0 ) and leaves φ a b e φ e c d e c ∧ e d ∧ e b = − φ e a b φ e c d e d ∧ e d ∧ e b = + ( ∗ φ ) a b c d e c ∧ e d ∧ e b = − ( ∗ φ ) a b c d e b ∧ e c ∧ e d = − 6 ı ∂ a ( ∗ φ 0 ) \varphi_{abe}\varphi_{ecde^c}\wedge e^d\wedge e^b=-\varphi_{eab}\varphi_{ecde^d}\wedge e^d\wedge e^b=+(\ast\varphi)_{abcde^c}\wedge e^d\wedge e^b=-(\ast\varphi)_{abcde^b}\wedge e^c\wedge e^d=-6\imath_{\partial_a}(\ast\varphi_0) φ ab e φ ec d e c ∧ e d ∧ e b = − φ e ab φ ec d e d ∧ e d ∧ e b = + ( ∗ φ ) ab c d e c ∧ e d ∧ e b = − ( ∗ φ ) ab c d e b ∧ e c ∧ e d = − 6 ∂ a ( ∗ φ 0 ) , while 1 2 σ ( e ∧ e ) a b ∧ e b = − 1 6 ( ∗ φ ) a b c d e b ∧ e c ∧ e d = − ı ∂ a ( ∗ φ 0 ) \frac{1}{2}\sigma(e\wedge e)_{ab}\wedge e^b=-\frac{1}{6}(\ast\varphi)_{abcde^b}\wedge e^c\wedge e^d=-\imath_{\partial_a}(\ast\varphi_0) 2 1 σ ( e ∧ e ) ab ∧ e b = − 6 1 ( ∗ φ ) ab c d e b ∧ e c ∧ e d = − ∂ a ( ∗ φ 0 ) , summing to − 7 ı ∂ a ( ∗ φ 0 ) -7\imath_{\partial_a}(\ast\varphi_0) − 7 ∂ a ( ∗ φ 0 ) in (51) . For (52) , the 3 3 3 -form B e ( 3 ) = σ ( ( e ∧ × G 2 e ) ∧ e ) B_e^{(3)}=\sigma((e\wedge_{\times G_2}e)\wedge e) B e ( 3 ) = σ (( e ∧ × G 2 e ) ∧ e ) is G 2 G_2 G 2 -invariant in ( Λ 3 R 7 ) ∗ ⊗ Λ 14 2 ≃ Hom G 2 ( Λ 3 R 7 , Λ 14 2 ) (\Lambda^3\mathbb{R}^7)^*\otimes\Lambda^2_{14}\simeq\operatorname{Hom}_{G_2}(\Lambda^3\mathbb{R}^7,\Lambda^2_{14}) ( Λ 3 R 7 ) ∗ ⊗ Λ 14 2 ≃ Hom G 2 ( Λ 3 R 7 , Λ 14 2 ) , which is zero by Schur’s lemma since Λ 3 R 7 = Λ 1 3 ⊕ Λ 7 3 ⊕ Λ 27 3 \Lambda^3\mathbb{R}^7=\Lambda^3_1\oplus\Lambda^3_7\oplus\Lambda^3_{27} Λ 3 R 7 = Λ 1 3 ⊕ Λ 7 3 ⊕ Λ 27 3 has no Λ 14 \Lambda_{14} Λ 14 summand (explicitly, in (49) , contracting 1 3 ( ∗ φ ) a b e f φ e c d = 1 3 ( δ a c φ b d f − δ a d φ b c f − δ b c φ a d f + δ b d φ a c f ) \frac{1}{3}(\ast\varphi)_{abef}\varphi_{ecd}=\frac{1}{3}(\delta_{ac}\varphi_{bdf}-\delta_{ad}\varphi_{bcf}-\delta_{bc}\varphi_{adf}+\delta_{bd}\varphi_{acf}) 3 1 ( ∗ φ ) ab e f φ ec d = 3 1 ( δ a c φ b df − δ a d φ b c f − δ b c φ a df + δ b d φ a c f ) with e c ∧ e d ∧ e f e^c\wedge e^d\wedge e^f e c ∧ e d ∧ e f gives 2 3 ( φ a c d e e c ∧ e d ∧ e b − φ b c d e e c ∧ e d ∧ e a ) \frac{2}{3}(\varphi_{acde}e^c\wedge e^d\wedge e_b-\varphi_{bcde}e^c\wedge e^d\wedge e_a) 3 2 ( φ a c d e e c ∧ e d ∧ e b − φ b c d e e c ∧ e d ∧ e a ) , which cancels the first term 2 3 ( φ a c d e e c ∧ e d ∧ e b − φ b c d e e c ∧ e d ∧ e a ) \frac{2}{3}(\varphi_{acde}e^c\wedge e^d\wedge e_b-\varphi_{bcde}e^c\wedge e^d\wedge e_a) 3 2 ( φ a c d e e c ∧ e d ∧ e b − φ b c d e e c ∧ e d ∧ e a ) of σ ( ( e ∧ × G 2 e ) ∧ e ) a b \sigma((e\wedge_{\times G_2}e)\wedge e)_{ab} σ (( e ∧ × G 2 e ) ∧ e ) ab identically).
For part (3), since B e ( 3 ) = 0 B_e^{(3)}=0 B e ( 3 ) = 0 , multiplying Λ × , b a l 3 e = ( v ( 3 ) , 0 ) \Lambda^3_{\times,\mathrm{bal}}e=(v^{(3)},0) Λ × , bal 3 e = ( v ( 3 ) , 0 ) twice on the left by ( 0 , Ω ω ) (0,\Omega_\omega) ( 0 , Ω ω ) via (42) gives (53) . To prove (3a), suppose ( Ω ω ) b c = 1 2 ∑ d , f = 1 7 R b c , d f e d ∧ e f ∈ Λ 14 2 (\Omega_\omega)_{bc}=\frac{1}{2}\sum_{d,f=1}^{7}R_{bc,df}e^d\wedge e^f\in\Lambda^2_{14} ( Ω ω ) b c = 2 1 ∑ d , f = 1 7 R b c , df e d ∧ e f ∈ Λ 14 2 for each pair ( b , c ) (b,c) ( b , c ) . Recall that a 2 2 2 -form α ∈ Λ 2 R 7 \alpha\in\Lambda^2\mathbb{R}^7 α ∈ Λ 2 R 7 lies in Λ 14 2 = ker ( ι ( ⋅ ) φ 0 ) \Lambda^2_{14}=\ker(\iota_{(\cdot)}\varphi_0) Λ 14 2 = ker ( ι ( ⋅ ) φ 0 ) if and only if α ∧ ∗ φ 0 = 0 \alpha\wedge\ast\varphi_0=0 α ∧ ∗ φ 0 = 0 in Λ 6 R 7 \Lambda^6\mathbb{R}^7 Λ 6 R 7 . Applying the interior derivation ı ∂ c \imath_{\partial_c} ∂ c to ( Ω ω ) b c ∧ ∗ φ 0 = 0 (\Omega_\omega)_{bc}\wedge\ast\varphi_0=0 ( Ω ω ) b c ∧ ∗ φ 0 = 0 gives
0 = ı ∂ c ( ( Ω ω ) b c ∧ ∗ φ 0 ) = ( ı ∂ c ( Ω ω ) b c ) ∧ ∗ φ 0 + ( Ω ω ) b c ∧ ı ∂ c ( ∗ φ 0 ) , 0=\imath_{\partial_c}\bigl((\Omega_\omega)_{bc}\wedge\ast\varphi_0\bigr)=\bigl(\imath_{\partial_c}(\Omega_\omega)_{bc}\bigr)\wedge\ast\varphi_0+(\Omega_\omega)_{bc}\wedge\imath_{\partial_c}(\ast\varphi_0), 0 = ∂ c ( ( Ω ω ) b c ∧ ∗ φ 0 ) = ( ∂ c ( Ω ω ) b c ) ∧ ∗ φ 0 + ( Ω ω ) b c ∧ ∂ c ( ∗ φ 0 ) ,
so ( Ω ω ) b c ∧ ı ∂ c ( ∗ φ 0 ) = − ( ı ∂ c ( Ω ω ) b c ) ∧ ∗ φ 0 (\Omega_\omega)_{bc} \wedge\imath_{\partial_c}(*\varphi_0) = -(\imath_{\partial_c}(\Omega_\omega)_{bc}) \wedge*\varphi_0 ( Ω ω ) b c ∧ ∂ c ( ∗ φ 0 ) = − ( ∂ c ( Ω ω ) b c ) ∧ ∗ φ 0 . Since ı ∂ c ( Ω ω ) b c = ∑ d = 1 7 R b c , c d e d \imath_{\partial_c}(\Omega_\omega)_{bc} = \sum_{d=1}^{7} R_{bc,cd}e^d ∂ c ( Ω ω ) b c = ∑ d = 1 7 R b c , c d e d , summing over c ∈ { 1 , … , 7 } c \in\{1,\ldots,7\} c ∈ { 1 , … , 7 } yields (54) for every b ∈ { 1 , … , 7 } b \in\{1,\ldots,7\} b ∈ { 1 , … , 7 } :
( Ω ω ⋅ v ( 3 ) ) b = − 7 ∑ c = 1 7 ( Ω ω ) b c ∧ ı ∂ c ( ∗ φ 0 ) = 7 ∑ c , d = 1 7 R b c , c d e d ∧ ∗ φ 0 . (\Omega_\omega\cdot v^{(3)})_b=-7\sum_{c=1}^{7}(\Omega_\omega)_{bc}\wedge\imath_{\partial_c}(*\varphi_0)=7\sum_{c,d=1}^{7}R_{bc,cd}e^d\wedge*\varphi_0. ( Ω ω ⋅ v ( 3 ) ) b = − 7 c = 1 ∑ 7 ( Ω ω ) b c ∧ ∂ c ( ∗ φ 0 ) = 7 c , d = 1 ∑ 7 R b c , c d e d ∧ ∗ φ 0 .
For every torsion-free G 2 G_2 G 2 -connection ( e , ω ) ∈ Conn A Θ = 0 ( P ) (e,\omega)\in\operatorname{Conn}^{\Theta=0}_{A}(P) ( e , ω ) ∈ Conn A Θ = 0 ( P ) , the algebraic first Bianchi identity Ω ω ∧ e = 0 \Omega_\omega\wedge e=0 Ω ω ∧ e = 0 and the G 2 G_2 G 2 -invariance L E ( ∗ φ 0 ) = ∑ b , c = 1 7 E b c e b ∧ ı ∂ c ( ∗ φ 0 ) = 0 L_E(*\varphi_0)=\sum_{b,c=1}^{7}E_{bc}e^b\wedge\imath_{\partial_c}(*\varphi_0)=0 L E ( ∗ φ 0 ) = ∑ b , c = 1 7 E b c e b ∧ ∂ c ( ∗ φ 0 ) = 0 (E ∈ g 2 E\in\mathfrak{g}_2 E ∈ g 2 ) imply that Ω ω ∈ W 77 = S 2 ( Λ 14 2 ) ∩ ker ( ∧ ) \Omega_\omega\in W_{77}=\mathcal{S}^{2}(\Lambda^{2}_{14})\cap\ker(\wedge) Ω ω ∈ W 77 = S 2 ( Λ 14 2 ) ∩ ker ( ∧ ) is Ricci-flat (Ric b d = ∑ c = 1 7 R b c , c d = 0 \operatorname{Ric}_{bd}=\sum_{c=1}^{7}R_{bc,cd}=0 Ric b d = ∑ c = 1 7 R b c , c d = 0 ). Thus Ω ω ⋅ v ( 3 ) = 0 \Omega_\omega\cdot v^{(3)}=0 Ω ω ⋅ v ( 3 ) = 0 and L 7 C Y M = 0 \mathcal{L}^{\mathrm{CYM}}_7=0 L 7 CYM = 0 on-shell.
Finally, to prove (3b), under the standard Fano-plane convention φ 0 = e 123 + e 145 + e 167 + e 246 − e 257 − e 347 − e 356 \varphi_0=e^{123}+e^{145}+e^{167}+e^{246}-e^{257}-e^{347}-e^{356} φ 0 = e 123 + e 145 + e 167 + e 246 − e 257 − e 347 − e 356 , we have ∗ φ 0 = e 4567 + e 2367 + e 2345 + e 1357 − e 1346 − e 1256 − e 1247 *\varphi_0=e^{4567}+e^{2367}+e^{2345}+e^{1357}-e^{1346}-e^{1256}-e^{1247} ∗ φ 0 = e 4567 + e 2367 + e 2345 + e 1357 − e 1346 − e 1256 − e 1247 and ( v ( 3 ) ) 5 = − 7 ı ∂ 5 ( ∗ φ 0 ) = 7 ( e 467 − e 234 − e 137 + e 126 ) (v^{(3)})_5=-7\imath_{\partial_5}(*\varphi_0)=7(e^{467}-e^{234}-e^{137}+e^{126}) ( v ( 3 ) ) 5 = − 7 ∂ 5 ( ∗ φ 0 ) = 7 ( e 467 − e 234 − e 137 + e 126 ) . Take the g 2 \mathfrak{g}_2 g 2 basis matrices E 1 = e 67 − e 45 ∈ g 2 E_1=e_{67}-e_{45}\in\mathfrak{g}_2 E 1 = e 67 − e 45 ∈ g 2 and E 2 = e 46 + e 57 ∈ g 2 E_2=e_{46}+e_{57}\in\mathfrak{g}_2 E 2 = e 46 + e 57 ∈ g 2 , and set Ω ω = α 1 ⊗ E 1 + α 2 ⊗ E 2 ∈ Ω 2 ( P ; g 2 ) \Omega_\omega=\alpha_1\otimes E_1+\alpha_2\otimes E_2\in\Omega^2(P;\mathfrak{g}_2) Ω ω = α 1 ⊗ E 1 + α 2 ⊗ E 2 ∈ Ω 2 ( P ; g 2 ) with α 1 = e 12 − e 45 ∈ Λ 14 2 \alpha_1=e^{12}-e^{45}\in\Lambda^2_{14} α 1 = e 12 − e 45 ∈ Λ 14 2 and α 2 = e 35 + e 46 ∈ Λ 2 R 7 \alpha_2=e^{35}+e^{46}\in\Lambda^2\mathbb{R}^7 α 2 = e 35 + e 46 ∈ Λ 2 R 7 (which has a non-zero Λ 7 2 \Lambda^2_7 Λ 7 2 projection since ı α 2 φ 0 = e 2 − e 6 ≠ 0 \imath_{\alpha_2}\varphi_0=e^2-e^6\ne0 α 2 φ 0 = e 2 − e 6 = 0 ). For a = 6 a=6 a = 6 , the only non-zero entries in row 6 of E 1 , E 2 E_1,E_2 E 1 , E 2 are ( E 1 ) 67 = + 1 (E_1)_{67}=+1 ( E 1 ) 67 = + 1 (followed by ( E 2 ) 75 = − 1 (E_2)_{75}=-1 ( E 2 ) 75 = − 1 to c = 5 c=5 c = 5 ) and ( E 2 ) 64 = − 1 (E_2)_{64}=-1 ( E 2 ) 64 = − 1 (followed by ( E 1 ) 45 = − 1 (E_1)_{45}=-1 ( E 1 ) 45 = − 1 to c = 5 c=5 c = 5 ). Summing over b , c ∈ { 1 , … , 7 } b,c\in\{1,\ldots,7\} b , c ∈ { 1 , … , 7 } gives
( Ω ω ⋅ ( Ω ω ⋅ v ( 3 ) ) ) 6 = ( ( Ω ω ) 67 ∧ ( Ω ω ) 75 + ( Ω ω ) 64 ∧ ( Ω ω ) 45 ) ∧ ( v ( 3 ) ) 5 = − 2 α 1 ∧ α 2 ∧ ( v ( 3 ) ) 5 . \bigl(\Omega_\omega\cdot(\Omega_\omega\cdot v^{(3)})\bigr)_6=\bigl((\Omega_\omega)_{67}\wedge(\Omega_\omega)_{75}+(\Omega_\omega)_{64}\wedge(\Omega_\omega)_{45}\bigr)\wedge(v^{(3)})_5=-2\alpha_1\wedge\alpha_2\wedge(v^{(3)})_5. ( Ω ω ⋅ ( Ω ω ⋅ v ( 3 ) ) ) 6 = ( ( Ω ω ) 67 ∧ ( Ω ω ) 75 + ( Ω ω ) 64 ∧ ( Ω ω ) 45 ) ∧ ( v ( 3 ) ) 5 = − 2 α 1 ∧ α 2 ∧ ( v ( 3 ) ) 5 .
Since α 1 ∧ α 2 = e 1235 + e 1246 \alpha_1\wedge\alpha_2=e^{1235}+e^{1246} α 1 ∧ α 2 = e 1235 + e 1246 and e 1235 ∧ ( 7 e 467 ) = − 7 vol 7 e^{1235}\wedge(7e^{467})=-7\operatorname{vol}_7 e 1235 ∧ ( 7 e 467 ) = − 7 vol 7 , we obtain
( Ω ω ⋅ ( Ω ω ⋅ v ( 3 ) ) ) 6 = − 2 ( e 1235 + e 1246 ) ∧ 7 ( e 467 − e 234 − e 137 + e 126 ) = + 14 vol 7 ≠ 0. \bigl(\Omega_\omega\cdot(\Omega_\omega\cdot v^{(3)})\bigr)_6=-2(e^{1235}+e^{1246})\wedge7(e^{467}-e^{234}-e^{137}+e^{126})=+14\operatorname{vol}_7\ne0. ( Ω ω ⋅ ( Ω ω ⋅ v ( 3 ) ) ) 6 = − 2 ( e 1235 + e 1246 ) ∧ 7 ( e 467 − e 234 − e 137 + e 126 ) = + 14 vol 7 = 0.
Internal Chern-Simons Dichotomy and Applications to BF, Higgs, and Dual Theories
Algebra-Valued Chern-Simons (A-CS) Theories: Associator Quotient Collapse and Jacobiator Dichotomy
Let M M M be a compact oriented smooth manifold of odd dimension n = 2 r − 1 n=2r-1 n = 2 r − 1 (r ≥ 2 r\ge2 r ≥ 2 ), and let Conn l i n ( P ; A ) = Ω h e q , ρ 1 ( P ; A ) \operatorname{Conn}^{\mathrm{lin}}(P;A)=\Omega^1_{\mathrm{heq},\rho}(P;A) Conn lin ( P ; A ) = Ω heq , ρ 1 ( P ; A ) be the linear space of horizontal equivariant connections (Proposition 2.4 (3)) so that for any D ∈ Conn l i n ( P ; A ) D\in\operatorname{Conn}^{\mathrm{lin}}(P;A) D ∈ Conn lin ( P ; A ) and t ∈ [ 0 , 1 ] t\in[0,1] t ∈ [ 0 , 1 ] , the scaled connection D t : = t D ∈ Conn l i n ( P ; A ) D_t:=tD\in\operatorname{Conn}^{\mathrm{lin}}(P;A) D t := t D ∈ Conn lin ( P ; A ) has curvature F t : = F D t = t d D + t 2 2 D ∧ ∗ D ∈ Ω h e q , ρ 2 ( P ; A ) F_t:=F_{D_t}=tdD+\frac{t^2}{2}D\wedge_*D\in\Omega^2_{\mathrm{heq},\rho}(P;A) F t := F D t = t d D + 2 t 2 D ∧ ∗ D ∈ Ω heq , ρ 2 ( P ; A ) . (In the affine Ehresmann regime D = D 0 + A D=D_0+A D = D 0 + A of Proposition 2.4 (3) with D 0 = η ∘ ω 0 D_0=\eta\circ\omega_0 D 0 = η ∘ ω 0 , the same formulas apply to the horizontal connection difference A = D − D 0 ∈ Ω h e q , ρ 1 ( P ; A ) A=D-D_0\in\Omega^1_{\mathrm{heq},\rho}(P;A) A = D − D 0 ∈ Ω heq , ρ 1 ( P ; A ) via the linear homotopy D t = D 0 + t A D_t=D_0+tA D t = D 0 + t A relative to a flat background F D 0 = 0 F_{D_0}=0 F D 0 = 0 .) We compare two ways of forming internal A A A -valued Chern-Simons functionals:
The Associative Quotient Ω h e q , ρ ( P ; A ) / A s s o c ( P , A ) \Omega_{\mathrm{heq},\rho}(P;A)/\mathfrak{Assoc}(P,A) Ω heq , ρ ( P ; A ) / Assoc ( P , A ) : Let A s s o c ( A ) ⊂ A \mathfrak{Assoc}(A)\subset A Assoc ( A ) ⊂ A be the two-sided ideal generated by all algebra associators Assoc ∗ ( x , y , z ) = ( x ∗ y ) ∗ z − x ∗ ( y ∗ z ) \operatorname{Assoc}_*(x,y,z)=(x*y)*z-x*(y*z) Assoc ∗ ( x , y , z ) = ( x ∗ y ) ∗ z − x ∗ ( y ∗ z ) , which induces the two-sided graded ideal A s s o c ( P , A ) = Ω h e q , ρ ( P ; A s s o c ( A ) ) ⊂ Ω h e q , ρ ( P ; A ) \mathfrak{Assoc}(P,A)=\Omega_{\mathrm{heq},\rho}(P;\mathfrak{Assoc}(A))\subset\Omega_{\mathrm{heq},\rho}(P;A) Assoc ( P , A ) = Ω heq , ρ ( P ; Assoc ( A )) ⊂ Ω heq , ρ ( P ; A ) generated by all form associators Assoc ∗ ( α , β , γ ) : = ( α ∧ ∗ β ) ∧ ∗ γ − α ∧ ∗ ( β ∧ ∗ γ ) \operatorname{Assoc}_*(\alpha,\beta,\gamma):=(\alpha\wedge_*\beta)\wedge_*\gamma-\alpha\wedge_*(\beta\wedge_*\gamma) Assoc ∗ ( α , β , γ ) := ( α ∧ ∗ β ) ∧ ∗ γ − α ∧ ∗ ( β ∧ ∗ γ ) , with quotient projection π a s : Ω h e q , ρ ( P ; A ) → Ω h e q , ρ ( P ; A ) / A s s o c ( P , A ) ≃ Ω h e q , ρ ( P ; A / A s s o c ( A ) ) \pi_{\mathrm{as}}:\Omega_{\mathrm{heq},\rho}(P;A)\to\Omega_{\mathrm{heq},\rho}(P;A)/\mathfrak{Assoc}(P,A)\simeq\Omega_{\mathrm{heq},\rho}(P;A/\mathfrak{Assoc}(A)) π as : Ω heq , ρ ( P ; A ) → Ω heq , ρ ( P ; A ) / Assoc ( P , A ) ≃ Ω heq , ρ ( P ; A / Assoc ( A )) . Assuming the trace section ξ ∈ Γ ( E A ∗ ) \xi\in\Gamma(E_A^*) ξ ∈ Γ ( E A ∗ ) annihilates the subbundle E A s s o c ( A ) E_{\mathfrak{Assoc}(A)} E Assoc ( A ) (so ξ ∈ Γ ( E A / A s s o c ( A ) ∗ ) ⊂ Γ ( E A ∗ ) \xi\in\Gamma(E_{A/\mathfrak{Assoc}(A)}^*)\subset\Gamma(E_A^*) ξ ∈ Γ ( E A / Assoc ( A ) ∗ ) ⊂ Γ ( E A ∗ ) ), tr ξ \operatorname{tr}_\xi tr ξ descends to a well-defined quotient bundle trace tr ‾ ξ : E A / A s s o c ( A ) → M × R \overline{\operatorname{tr}}_\xi:E_{A/\mathfrak{Assoc}(A)}\to M\times\mathbb{R} tr ξ : E A / Assoc ( A ) → M × R satisfying tr ‾ ξ ( ȷ ˉ ( π a s ( Ψ ) ) ) = tr ξ ( ȷ ( Ψ ) ) \overline{\operatorname{tr}}_\xi(\bar{\jmath}(\pi_{\mathrm{as}}(\Psi)))=\operatorname{tr}_\xi(\jmath(\Psi)) tr ξ ( ˉ ( π as ( Ψ ))) = tr ξ ( ( Ψ )) .
Unquotiented Grouped Parenthesization in Ω h e q , ρ ( P ; A ) \Omega_{\mathrm{heq},\rho}(P;A) Ω heq , ρ ( P ; A ) : Working directly in Ω h e q , ρ ( P ; A ) \Omega_{\mathrm{heq},\rho}(P;A) Ω heq , ρ ( P ; A ) without quotienting by A s s o c ( P , A ) \mathfrak{Assoc}(P,A) Assoc ( P , A ) , we define the grouped Chern-Simons ( 2 r − 1 ) (2r-1) ( 2 r − 1 ) -form C S 2 r − 1 g r p , P ( D ) ∈ Ω h e q , ρ 2 r − 1 ( P ; A ) \mathrm{CS}^{\mathrm{grp},\mathcal{P}}_{2r-1}(D)\in\Omega^{2r-1}_{\mathrm{heq},\rho}(P;A) CS 2 r − 1 grp , P ( D ) ∈ Ω heq , ρ 2 r − 1 ( P ; A ) for P ∈ { b l k , b a l } \mathcal{P}\in\{\mathrm{blk},\mathrm{bal}\} P ∈ { blk , bal } by grouping the ( r − 1 − j ) (r-1-j) ( r − 1 − j ) factors of d D dD d D and the ( 2 j + 1 ) (2j+1) ( 2 j + 1 ) factors of D D D :
C S 2 r − 1 g r p , P ( D ) : = ∑ j = 0 r − 1 r − 1 ( r + j ) 2 j ( r j ) ( Λ ∗ , P r − 1 − j d D ) ∧ ∗ ( Λ ∗ , P 2 j + 1 D ) ∈ Ω h e q , ρ 2 r − 1 ( P ; A ) . (55) \mathrm{CS}^{\mathrm{grp},\mathcal{P}}_{2r-1}(D):=\sum_{j=0}^{r-1}\frac{r-1}{(r+j)2^j}\binom{r}{j}\left(\Lambda^{r-1-j}_{*,\mathcal{P}}dD\right)\wedge_*\left(\Lambda^{2j+1}_{*,\mathcal{P}}D\right)\in\Omega^{2r-1}_{\mathrm{heq},\rho}(P;A). \tag*{(55)} CS 2 r − 1 grp , P ( D ) := j = 0 ∑ r − 1 ( r + j ) 2 j r − 1 ( j r ) ( Λ ∗ , P r − 1 − j d D ) ∧ ∗ ( Λ ∗ , P 2 j + 1 D ) ∈ Ω heq , ρ 2 r − 1 ( P ; A ) . (55)
In particular, for n = 3 n=3 n = 3 (r = 2 r=2 r = 2 , where Λ ∗ , b l k m \Lambda^m_{*,\mathrm{blk}} Λ ∗ , blk m and Λ ∗ , b a l m \Lambda^m_{*,\mathrm{bal}} Λ ∗ , bal m coincide on all factors) and n = 5 n=5 n = 5 (r = 3 r=3 r = 3 , where the j = 2 j=2 j = 2 term 3 20 Λ ∗ , P 5 D \frac{3}{20}\Lambda^5_{*,\mathcal{P}}D 20 3 Λ ∗ , P 5 D equals 3 20 ( Λ ∗ , b a l 4 D ) ∧ ∗ D \frac{3}{20}(\Lambda^4_{*,\mathrm{bal}}D)\wedge_*D 20 3 ( Λ ∗ , bal 4 D ) ∧ ∗ D under P = b l k \mathcal{P}=\mathrm{blk} P = blk and 3 20 ( ( D ∧ ⋆ D ) ∧ ⋆ D ) ∧ ⋆ ( D ∧ ⋆ D ) \frac{3}{20}((D\wedge_{\star}D)\wedge_{\star}D)\wedge_{\star}(D\wedge_{\star}D) 20 3 (( D ∧ ⋆ D ) ∧ ⋆ D ) ∧ ⋆ ( D ∧ ⋆ D ) under P = b a l ) \mathcal{P}=\mathrm{bal}) P = bal ) :
CS 3 g r p ( D ) = d D ∧ ⋆ D + 1 3 ( D ∧ ⋆ D ) ∧ ⋆ D , \operatorname{CS}^{\mathrm{grp}}_{3}(D)=dD\wedge_{\star}D+\frac{1}{3}(D\wedge_{\star}D)\wedge_{\star}D, CS 3 grp ( D ) = d D ∧ ⋆ D + 3 1 ( D ∧ ⋆ D ) ∧ ⋆ D ,
CS 5 g r p , P ( D ) = ( d D ∧ ⋆ d D ) ∧ ⋆ D + 3 4 d D ∧ ⋆ ( ( D ∧ ⋆ D ) ∧ ⋆ D ) + 3 20 ζ ⋆ , P 5 D . \operatorname{CS}^{\mathrm{grp},\mathcal{P}}_{5}(D)=(dD\wedge_{\star}dD)\wedge_{\star}D+\frac{3}{4}dD\wedge_{\star}((D\wedge_{\star}D)\wedge_{\star}D)+\frac{3}{20}\zeta^{5}_{\star,\mathcal{P}}D. CS 5 grp , P ( D ) = ( d D ∧ ⋆ d D ) ∧ ⋆ D + 4 3 d D ∧ ⋆ (( D ∧ ⋆ D ) ∧ ⋆ D ) + 20 3 ζ ⋆ , P 5 D .
Definition 5.1 (Internal Algebra-Valued Chern-Simons (A A A -CS) Functionals). Let dim M = 2 r − 1 \dim M=2r-1 dim M = 2 r − 1 (r ≥ 2 r\ge2 r ≥ 2 ). The quotiented A A A -CS functional (for ξ ∈ Γ ( E ( A / A s s o c ( A ) ) ∗ ) \xi\in\Gamma(E_{(A/\mathsf{Assoc}(A))^{*}}) ξ ∈ Γ ( E ( A / Assoc ( A ) ) ∗ ) ) and the unquotiented grouped A A A -CS functional on Conn l i n ( P ; A ) \operatorname{Conn}^{\mathrm{lin}}(P;A) Conn lin ( P ; A ) are defined respectively by
S C S , 2 r − 1 A , q u o t [ D ] : = r ∫ M ∫ 0 1 tr ‾ ξ ( j ( π a s ( D ∧ ⋆ ( ∧ ⋆ r − 1 F t ) ) ) ) d t , S C S , 2 r − 1 A , g r p , P [ D ] : = ∫ M tr ξ ( j ( CS 2 r − 1 g r p , P ( D ) ) ) . (56) S^{A,\mathrm{quot}}_{\mathrm{CS},2r-1}[D]:=r\int_{M}\int_{0}^{1}\overline{\operatorname{tr}}_{\xi}\left(j\left(\pi_{\mathrm{as}}\left(D\wedge_{\star}\left(\wedge_{\star}^{r-1}F_{t}\right)\right)\right)\right)\,dt,\qquad S^{A,\mathrm{grp},\mathcal{P}}_{\mathrm{CS},2r-1}[D]:=\int_{M}\operatorname{tr}_{\xi}\left(j\left(\operatorname{CS}^{\mathrm{grp},\mathcal{P}}_{2r-1}(D)\right)\right). \tag*{(56)} S CS , 2 r − 1 A , quot [ D ] := r ∫ M ∫ 0 1 tr ξ ( j ( π as ( D ∧ ⋆ ( ∧ ⋆ r − 1 F t ) ) ) ) d t , S CS , 2 r − 1 A , grp , P [ D ] := ∫ M tr ξ ( j ( CS 2 r − 1 grp , P ( D ) ) ) . (56)
Remark 5.2 (Internal A-CS vs. Classical Bilinear Chern-Simons Theory). Just as noted for Chern-Weil theory in Remark 2.6 , classical 3D Chern-Simons theory on a Lie algebra g \mathfrak{g} g uses an external symmetric Ad \operatorname{Ad} Ad -invariant bilinear form κ ∈ S 2 ( g ∗ ) Ad \kappa\in S^{2}(\mathfrak{g}^{*})^{\operatorname{Ad}} κ ∈ S 2 ( g ∗ ) Ad to form ∫ M ( κ ( D ∧ d D ) + 1 3 κ ( D ∧ [ D ∧ D ] g ) ) \int_{M}(\kappa(D\wedge dD)+\frac{1}{3}\kappa(D\wedge[D\wedge D]_{\mathfrak{g}})) ∫ M ( κ ( D ∧ d D ) + 3 1 κ ( D ∧ [ D ∧ D ] g )) (or uses the associative matrix product in Mat k × k ( R ) \operatorname{Mat}_{k\times k}(\mathbb{R}) Mat k × k ( R ) ). Because κ \kappa κ is symmetric while [ ⋅ , ⋅ ] g [\cdot,\cdot]_{\mathfrak{g}} [ ⋅ , ⋅ ] g is skew-symmetric, κ ( x , [ y , z ] g ) \kappa(x,[y,z]_{\mathfrak{g}}) κ ( x , [ y , z ] g ) is totally skew-symmetric in ( x , y , z ) (x,y,z) ( x , y , z ) , so the classical cubic term is non-zero for every non-abelian semi-simple Lie algebra. In contrast, Definition 5.1 multiplies all factors internally within ( A , ∗ ) (A,\ast) ( A , ∗ ) before applying a linear functional tr ξ : E A → M × R \operatorname{tr}_{\xi}:E_{A}\to M\times\mathbb{R} tr ξ : E A → M × R . When ( A , ∧ ) (A,\wedge) ( A , ∧ ) is skew-commutative, the outer multiplication in ( D ∧ D ) ∧ D (D\wedge D)\wedge D ( D ∧ D ) ∧ D is also skew-symmetric, which forces ( D ∧ D ) ∧ D (D\wedge D)\wedge D ( D ∧ D ) ∧ D to project onto the Jacobiator Jac ∧ ( D , D , D ) \operatorname{Jac}_{\wedge}(D,D,D) Jac ∧ ( D , D , D ) ! Moreover, for a skew-commutative product ∧ \wedge ∧ , because d D dD d D has degree 2 2 2 and D D D has degree 1 1 1 , the quadratic term d D ∧ D dD\wedge D d D ∧ D takes values in the commutator subspace [ A , A ] ∧ [A,A]_{\wedge} [ A , A ] ∧ : thus d D ∧ D dD\wedge D d D ∧ D vanishes identically if ( A , ∧ ) (A,\wedge) ( A , ∧ ) is abelian, is annihilated by any G G G -invariant trace τ 0 ∈ ( A ∗ ) G \tau_{0}\in(A^{*})^{G} τ 0 ∈ ( A ∗ ) G , and is globally exact on P P P since d ( D ∧ D ) = d D ∧ D − D ∧ d D = d D ∧ D − ( − 1 ) 1 ⋅ 2 d D ∧ D = 2 d D ∧ D d(D\wedge D)=dD\wedge D-D\wedge dD=dD\wedge D-(-1)^{1\cdot2}dD\wedge D=2dD\wedge D d ( D ∧ D ) = d D ∧ D − D ∧ d D = d D ∧ D − ( − 1 ) 1 ⋅ 2 d D ∧ D = 2 d D ∧ D (so d D ∧ D = 1 2 d ( D ∧ D ) dD\wedge D=\frac{1}{2}d(D\wedge D) d D ∧ D = 2 1 d ( D ∧ D ) and d ( d D ∧ D ) = − d D ∧ d D = 0 d(dD\wedge D)=-dD\wedge dD=0 d ( d D ∧ D ) = − d D ∧ d D = 0 ).
Theorem 5.3 (Associator Quotient Collapse, Jacobiator Dichotomy, and Transgression Anomaly in Internal-Trace A-CS Theories). Let M M M be a compact oriented manifold of odd dimension n = 2 r − 1 n=2r-1 n = 2 r − 1 (r ≥ 2 r\ge2 r ≥ 2 ), and let E A = P × ρ A → M E_{A}=P\times_{\rho}A\to M E A = P × ρ A → M be equipped with a trace functional tr ξ \operatorname{tr}_{\xi} tr ξ .
(1) Cubic Annihilation of Skew-Commutative Algebras in the Associative Quotient: Suppose ( A , ∗ ) (A,\ast) ( A , ∗ ) is any skew-commutative algebra ( x ∗ y = − y ∗ x ) (x\ast y=-y\ast x) ( x ∗ y = − y ∗ x ) . Then in the associative quotient algebra A / A s s o c ( A ) A/\mathsf{Assoc}(A) A / Assoc ( A ) , every 3-fold product vanishes identically ( x ∗ y ∗ z ∈ A s s o c ( A ) (x\ast y\ast z\in\mathsf{Assoc}(A) ( x ∗ y ∗ z ∈ Assoc ( A ) for all x , y , z ∈ A ) x,y,z\in A) x , y , z ∈ A ) . Consequently, in Ω h e q , ρ ( P ; A ) / A s s o c ( P , A ) \Omega_{\mathrm{heq},\rho}(P;A)/\mathsf{Assoc}(P,A) Ω heq , ρ ( P ; A ) / Assoc ( P , A ) :
For n = 3 n=3 n = 3 (r = 2 r=2 r = 2 ), the cubic term 1 3 π a s ( D ∧ ⋆ D ∧ ⋆ D ) = 0 \frac{1}{3}\pi_{\mathrm{as}}(D\wedge_{\star}D\wedge_{\star}D)=0 3 1 π as ( D ∧ ⋆ D ∧ ⋆ D ) = 0 vanishes identically, leaving only ∫ M tr ξ ( j ( π a s ( D ∧ ⋆ d D ) ) ) = − ∫ M tr ξ ( j ( π a s ( d D ∧ ⋆ D ) ) ) \int_{M}\operatorname{tr}_{\xi}(j(\pi_{\mathrm{as}}(D\wedge_{\star}dD)))=-\int_{M}\operatorname{tr}_{\xi}(j(\pi_{\mathrm{as}}(dD\wedge_{\star}D))) ∫ M tr ξ ( j ( π as ( D ∧ ⋆ d D ))) = − ∫ M tr ξ ( j ( π as ( d D ∧ ⋆ D ))) (where D ∧ ⋆ d D = − ( − 1 ) 1 ⋅ 2 d D ∧ ⋆ D = − d D ∧ ⋆ D D\wedge_{\star}dD=-(-1)^{1\cdot2}dD\wedge_{\star}D=-dD\wedge_{\star}D D ∧ ⋆ d D = − ( − 1 ) 1 ⋅ 2 d D ∧ ⋆ D = − d D ∧ ⋆ D by (7) );
For all odd dimensions n = 2 r − 1 ≥ 5 n=2r-1\ge5 n = 2 r − 1 ≥ 5 (r ≥ 3 r\ge3 r ≥ 3 ), S C S , 2 r − 1 A , q u o t [ D ] ≡ 0 S^{A,\mathrm{quot}}_{\mathrm{CS},2r-1}[D]\equiv0 S CS , 2 r − 1 A , quot [ D ] ≡ 0 vanishes identically.
(2) Unquotiented Vertex Truncation under P \mathcal{P} P -(1, s s s )-Solvability: If A = ⨁ m ∈ Z A m A=\bigoplus_{m\in\mathbb{Z}}A^{m} A = ⨁ m ∈ Z A m is a weak P \mathcal{P} P -(1, s s s )-solvable m m m -graded algebra (P ∈ { b l k , b a l } \mathcal{P}\in\{\mathrm{blk},\mathrm{bal}\} P ∈ { blk , bal } ), then for every j j j satisfying 2 j + 1 ≥ s + 1 2j+1\ge s+1 2 j + 1 ≥ s + 1 (i.e., j ≥ ⌈ s / 2 ⌉ j\ge\lceil s/2\rceil j ≥ ⌈ s /2 ⌉ ), the ( r + j ) (r+j) ( r + j ) -point vertex ( ∧ ⋆ , P r − 1 d D ) ∧ ⋆ ( ∧ ⋆ , P 2 j + 1 D ) (\wedge_{\star,\mathcal{P}}^{r-1}dD)\wedge_{\star}(\wedge_{\star,\mathcal{P}}^{2j+1}D) ( ∧ ⋆ , P r − 1 d D ) ∧ ⋆ ( ∧ ⋆ , P 2 j + 1 D ) in (55) vanishes identically in Ω 2 r − 1 ( P ; A ) \Omega^{2r-1}(P;A) Ω 2 r − 1 ( P ; A ) .
(3) Unquotiented Skew-Commutative Dichotomy (Lie vs. Non-Lie Algebras): Suppose ( A , ∧ ) (A,\wedge) ( A , ∧ ) is any skew-commutative algebra ( x ∧ y = − y ∧ x ) (x\wedge y=-y\wedge x) ( x ∧ y = − y ∧ x ) , with Jacobiator Jac ∧ ( x , y , z ) = ( x ∧ y ) ∧ z + ( y ∧ z ) ∧ x + ( z ∧ x ) ∧ y \operatorname{Jac}_{\wedge}(x,y,z)=(x\wedge y)\wedge z+(y\wedge z)\wedge x+(z\wedge x)\wedge y Jac ∧ ( x , y , z ) = ( x ∧ y ) ∧ z + ( y ∧ z ) ∧ x + ( z ∧ x ) ∧ y . For every 1-form D ∈ Ω 1 ( P ; A ) D\in\Omega^{1}(P;A) D ∈ Ω 1 ( P ; A ) , direct evaluation via (6) and (24) gives the exact identity
( D ∧ D ) ∧ D = 1 3 Jac ∧ ( D , D , D ) ∈ Ω 3 ( P ; A ) , d D ∧ D = 1 2 d ( D ∧ D ) ∈ Ω 3 ( P ; A ) . (57) (D\wedge D)\wedge D=\frac{1}{3}\operatorname{Jac}_{\wedge}(D,D,D)\in\Omega^{3}(P;A),\qquad dD\wedge D=\frac{1}{2}d(D\wedge D)\in\Omega^{3}(P;A). \tag*{(57)} ( D ∧ D ) ∧ D = 3 1 Jac ∧ ( D , D , D ) ∈ Ω 3 ( P ; A ) , d D ∧ D = 2 1 d ( D ∧ D ) ∈ Ω 3 ( P ; A ) . (57) Consequently, for the unquotiented grouped Chern-Simons functional S C S , 2 r − 1 A , g r p , P [ D ] S^{A,\mathrm{grp},\mathcal{P}}_{\mathrm{CS},2r-1}[D] S CS , 2 r − 1 A , grp , P [ D ] :
(a) Lie Case (Jac ∧ ≡ 0 \operatorname{Jac}_{\wedge}\equiv0 Jac ∧ ≡ 0 , e.g., any Lie subalgebra h ⊂ s o ( k ) \mathfrak{h}\subset\mathfrak{so}(k) h ⊂ so ( k ) under X ∧ Y = 1 2 [ X , Y ] X\wedge Y=\frac{1}{2}[X,Y] X ∧ Y = 2 1 [ X , Y ] ):
By (57) , ∧ b a l 3 D = ( D ∧ D ) ∧ D = 0 \wedge_{\mathrm{bal}}^{3}D=(D\wedge D)\wedge D=0 ∧ bal 3 D = ( D ∧ D ) ∧ D = 0 vanishes identically in Ω 3 ( P ; h ) \Omega^{3}(P;\mathfrak{h}) Ω 3 ( P ; h ) , and hence by Lemma 2.1 (2a), ∧ b a l m D = ∧ b l k m D = 0 \wedge_{\mathrm{bal}}^{m}D=\wedge_{\mathrm{blk}}^{m}D=0 ∧ bal m D = ∧ blk m D = 0 vanishes for all m ≥ 3 m\ge3 m ≥ 3 under both blk and bal parenthesizations! Therefore:
In dimension n = 3 n = 3 n = 3 (r = 2 r = 2 r = 2 ), the cubic vertex 1 3 ( D ∧ D ) ∧ D = 0 \frac{1}{3}(D \wedge D) \wedge D = 0 3 1 ( D ∧ D ) ∧ D = 0 vanishes identically, reducing S C S , 3 h , g r p [ D ] S^{\mathfrak h,\mathrm{grp}}_{\mathrm{CS},3}[D] S CS , 3 h , grp [ D ] to the quadratic commutator functional 1 2 ∫ M tr ξ ( j ( d ( D ∧ D ) ) ) \frac{1}{2}\int_M \operatorname{tr}_{\xi}(j(d(D \wedge D))) 2 1 ∫ M tr ξ ( j ( d ( D ∧ D ))) (which vanishes identically by Stokes’ theorem on the closed manifold M M M whenever the equivariant lift ξ ^ : P → h ∗ \hat{\xi}:P \to\mathfrak h^* ξ ^ : P → h ∗ is constant, d ξ ^ = 0 d\hat{\xi}=0 d ξ ^ = 0 , in the linear/horizontal regime d ρ = 0 d\rho=0 d ρ = 0 *);
In all odd dimensions n = 2 r − 1 ≥ 5 n=2r-1\geq5 n = 2 r − 1 ≥ 5 (r ≥ 3 r\geq3 r ≥ 3 ), S C S , 2 r − 1 h , g r p , P [ D ] ≡ 0 S^{\mathfrak h,\mathrm{grp},P}_{\mathrm{CS},2r-1}[D]\equiv0 S CS , 2 r − 1 h , grp , P [ D ] ≡ 0 vanishes identically in Ω 2 r − 1 ( P ; h ) ∗ u n d e r b o t h ∗ \Omega^{2r-1}(P;\mathfrak h) *under both* Ω 2 r − 1 ( P ; h ) ∗ u n d er b o t h ∗ \mathcal P=\mathrm{blk}$ and P = b a l \mathcal P=\mathrm{bal} P = bal .
(b) Non-Lie Skew-Commutative Case (Jac ∧ ≢ 0 \operatorname{Jac}_{\wedge}\not\equiv0 Jac ∧ ≡ 0 , e.g., the octonionic Malcev algebra Im ( O ) \operatorname{Im}(\mathbb O) Im ( O ) , reductive extensions A = R k ⊕ h A=\mathbb R^k\oplus\mathfrak h A = R k ⊕ h under ∧ λ \wedge_{\lambda} ∧ λ , or Cayley-Dickson algebras u ( k ; C D l ( R ) ) \mathfrak u(k;\mathrm{CD}^{l}(\mathbb R)) u ( k ; CD l ( R )) for l ≥ 3 l\geq3 l ≥ 3 ):
In dimension n = 3 n=3 n = 3 (r = 2 r=2 r = 2 ), the cubic vertex 1 3 ( D ∧ D ) ∧ D = 1 3 Jac ∧ ( D , D , D ) \frac{1}{3}(D\wedge D)\wedge D=\frac{1}{3}\operatorname{Jac}_{\wedge}(D,D,D) 3 1 ( D ∧ D ) ∧ D = 3 1 Jac ∧ ( D , D , D ) is non-vanishing in Ω 3 ( P ; A ) \Omega^3(P;A) Ω 3 ( P ; A ) ;
In dimension n = 5 n=5 n = 5 (r = 3 r=3 r = 3 ), under quartic-block grouping (P = b l k \mathcal P=\mathrm{blk} P = blk ), both ( d D ∧ d D ) ∧ D = 0 (dD\wedge dD)\wedge D=0 ( d D ∧ d D ) ∧ D = 0 and 3 20 ∧ b l k 5 D = 3 20 ( ∧ b a l 4 D ) ∧ D = 0 \frac{3}{20}\wedge_{\mathrm{blk}}^5D=\frac{3}{20}(\wedge_{\mathrm{bal}}^4D)\wedge D=0 20 3 ∧ blk 5 D = 20 3 ( ∧ bal 4 D ) ∧ D = 0 vanish by ( 2 , 1 ) (2,1) ( 2 , 1 ) -nilpotency, collapsing the unquotiented 5D block-grouped Chern-Simons functional to the pure quartic Jacobiator vertex:
S C S , 5 A , g r p , b l k [ D ] = 3 4 ∫ M tr ξ ( j ( d D ∧ ( ( D ∧ D ) ∧ D ) ) ) = 1 4 ∫ M tr ξ ( j ( d D ∧ Jac ∧ ( D , D , D ) ) ) (58) S^{A,\mathrm{grp},\mathrm{blk}}_{\mathrm{CS},5}[D]
=\frac{3}{4}\int_M\operatorname{tr}_{\xi}(j(dD\wedge((D\wedge D)\wedge D)))
=\frac{1}{4}\int_M\operatorname{tr}_{\xi}(j(dD\wedge\operatorname{Jac}_{\wedge}(D,D,D))) \tag*{(58)} S CS , 5 A , grp , blk [ D ] = 4 3 ∫ M tr ξ ( j ( d D ∧ (( D ∧ D ) ∧ D ))) = 4 1 ∫ M tr ξ ( j ( d D ∧ Jac ∧ ( D , D , D ))) (58)
(whereas under recursive binary halving P = b a l \mathcal P=\mathrm{bal} P = bal , S C S , 5 A , g r p , b a l [ D ] S^{A,\mathrm{grp},\mathrm{bal}}_{\mathrm{CS},5}[D] S CS , 5 A , grp , bal [ D ] additionally retains the J 4 = 5 J_4=5 J 4 = 5 quintic Jacobsthal vertex 3 20 ∫ M tr ξ ( j ( ∧ b a l 5 D ) ) = 1 20 ∫ M tr ξ ( j ( Jac ∧ ( D , D , D ) ∧ ( D ∧ D ) ) ) \frac{3}{20}\int_M\operatorname{tr}_{\xi}(j(\wedge_{\mathrm{bal}}^5D))=\frac{1}{20}\int_M\operatorname{tr}_{\xi}(j(\operatorname{Jac}_{\wedge}(D,D,D)\wedge(D\wedge D))) 20 3 ∫ M tr ξ ( j ( ∧ bal 5 D )) = 20 1 ∫ M tr ξ ( j ( Jac ∧ ( D , D , D ) ∧ ( D ∧ D ))) *);
In all odd dimensions n = 2 r − 1 ≥ 7 n=2r-1\geq7 n = 2 r − 1 ≥ 7 (r ≥ 4 r\geq4 r ≥ 4 ), the block-grouped A A A -valued Chern-Simons functional S C S , 2 r − 1 A , g r p , b l k [ D ] S^{A,\mathrm{grp},\mathrm{blk}}_{\mathrm{CS},2r-1}[D] S CS , 2 r − 1 A , grp , blk [ D ] vanishes identically in Ω 2 r − 1 ( P ; A ) \Omega^{2r-1}(P;A) Ω 2 r − 1 ( P ; A ) , while under recursive binary halving (P = b a l \mathcal P=\mathrm{bal} P = bal ), S C S , n A , g r p , b a l [ D ] S^{A,\mathrm{grp},\mathrm{bal}}_{\mathrm{CS},n}[D] S CS , n A , grp , bal [ D ] vanishes algebraically for every odd dimension n ≥ 7 n\geq7 n ≥ 7 with n ∉ ⋃ m = 4 ∞ { J m , J m + 2 } ∩ [ 7 , ∞ ) = { 7 } ∪ { 11 , 13 } ∪ { 21 , 23 } ∪ { 43 , 45 } ∪ ⋯ n\notin\bigcup_{m=4}^{\infty}\{J_m,J_m+2\}\cap[7,\infty)=\{7\}\cup\{11,13\}\cup\{21,23\}\cup\{43,45\}\cup\cdots n ∈ / ⋃ m = 4 ∞ { J m , J m + 2 } ∩ [ 7 , ∞ ) = { 7 } ∪ { 11 , 13 } ∪ { 21 , 23 } ∪ { 43 , 45 } ∪ ⋯ (where J m = 2 m − ( − 1 ) m 3 J_m=\frac{2^m-(-1)^m}{3} J m = 3 2 m − ( − 1 ) m is the m m m -*th odd Jacobsthal number) and vanishes by rank for all n ≥ dim R A + 3 n\geq\dim_{\mathbb R}A+3 n ≥ dim R A + 3 .
(4) Exterior Derivative in 3D and 5D: For any skew-commutative algebra ( A , ∧ ) (A,\wedge) ( A , ∧ ) , since d ( d D ∧ D ) = − d D ∧ d D = 0 d(dD\wedge D)=-dD\wedge dD=0 d ( d D ∧ D ) = − d D ∧ d D = 0 , the exterior derivatives of C S 3 g r p ( D ) \mathrm{CS}^{\mathrm{grp}}_3(D) CS 3 grp ( D ) and C S 5 g r p , b l k ( D ) \mathrm{CS}^{\mathrm{grp},\mathrm{blk}}_5(D) CS 5 grp , blk ( D ) are given purely by their Jacobiator terms:
d C S 3 g r p ( D ) = 1 3 d ( Jac ∧ ( D , D , D ) ) , d C S 5 g r p , b l k ( D ) = − 1 4 d D ∧ d ( Jac ∧ ( D , D , D ) ) , (59) d\mathrm{CS}^{\mathrm{grp}}_3(D)=\frac{1}{3}d(\operatorname{Jac}_{\wedge}(D,D,D)),\qquad
d\mathrm{CS}^{\mathrm{grp},\mathrm{blk}}_5(D)=-\frac{1}{4}dD\wedge d(\operatorname{Jac}_{\wedge}(D,D,D)), \tag*{(59)} d CS 3 grp ( D ) = 3 1 d ( Jac ∧ ( D , D , D )) , d CS 5 grp , blk ( D ) = − 4 1 d D ∧ d ( Jac ∧ ( D , D , D )) , (59)
where 1 3 d ( Jac ∧ ( D , D , D ) ) = d ( ( D ∧ D ) ∧ D ) = 2 ( d D ∧ D ) ∧ D − d D ∧ ( D ∧ D ) \frac{1}{3}d(\operatorname{Jac}_{\wedge}(D,D,D))=d((D\wedge D)\wedge D)=2(dD\wedge D)\wedge D-dD\wedge(D\wedge D) 3 1 d ( Jac ∧ ( D , D , D )) = d (( D ∧ D ) ∧ D ) = 2 ( d D ∧ D ) ∧ D − d D ∧ ( D ∧ D ) .
Proof. To prove (1), in the associative quotient A / Assoc ( A ) A/\operatorname{Assoc}(A) A / Assoc ( A ) , alternating use of skew-commutativity and associativity gives ( x ∗ y ) ∗ z = − ( y ∗ x ) ∗ z = − y ∗ ( x ∗ z ) = + ( x ∗ z ) ∗ y = x ∗ ( z ∗ y ) = − x ∗ ( y ∗ z ) = − ( x ∗ y ) ∗ z (x\ast y)\ast z=-(y\ast x)\ast z=-y\ast(x\ast z)=+(x\ast z)\ast y=x\ast(z\ast y)=-x\ast(y\ast z)=-(x\ast y)\ast z ( x ∗ y ) ∗ z = − ( y ∗ x ) ∗ z = − y ∗ ( x ∗ z ) = + ( x ∗ z ) ∗ y = x ∗ ( z ∗ y ) = − x ∗ ( y ∗ z ) = − ( x ∗ y ) ∗ z , so 2 ( x ∗ y ∗ z ) = 0 ⟹ x ∗ y ∗ z = 0 2(x\ast y\ast z)=0\Longrightarrow x\ast y\ast z=0 2 ( x ∗ y ∗ z ) = 0 ⟹ x ∗ y ∗ z = 0 in A / Assoc ( A ) A/\operatorname{Assoc}(A) A / Assoc ( A ) . Directly on differential forms in Ω h e q , ρ ( P ; A ) / Assoc ( P , A ) \Omega_{\mathrm{heq},\rho}(P;A)/\operatorname{Assoc}(P,A) Ω heq , ρ ( P ; A ) / Assoc ( P , A ) , since D ∧ ∗ D D\wedge_{\ast}D D ∧ ∗ D is a 2-form and D D D is a 1-form while α ∧ ∗ β = − ( − 1 ) k l β ∧ ∗ α \alpha\wedge_{\ast}\beta=-(-1)^{kl}\beta\wedge_{\ast}\alpha α ∧ ∗ β = − ( − 1 ) k l β ∧ ∗ α by (7) , we have ( D ∧ ∗ D ) ∧ ∗ D = − ( − 1 ) 2 ⋅ 1 D ∧ ∗ ( D ∧ ∗ D ) = − D ∧ ∗ ( D ∧ ∗ D ) ≡ − ( D ∧ ∗ D ) ∧ ∗ D ( m o d Assoc ( P , A ) ) (D\wedge_{\ast}D)\wedge_{\ast}D=-(-1)^{2\cdot1}D\wedge_{\ast}(D\wedge_{\ast}D)=-D\wedge_{\ast}(D\wedge_{\ast}D)\equiv-(D\wedge_{\ast}D)\wedge_{\ast}D\pmod{\operatorname{Assoc}(P,A)} ( D ∧ ∗ D ) ∧ ∗ D = − ( − 1 ) 2 ⋅ 1 D ∧ ∗ ( D ∧ ∗ D ) = − D ∧ ∗ ( D ∧ ∗ D ) ≡ − ( D ∧ ∗ D ) ∧ ∗ D ( mod Assoc ( P , A )) , so π a s ( ( D ∧ ∗ D ) ∧ ∗ D ) = 0 \pi_{\mathrm{as}}((D\wedge_{\ast}D)\wedge_{\ast}D)=0 π as (( D ∧ ∗ D ) ∧ ∗ D ) = 0 (and similarly any ≥ 3 \geq3 ≥ 3 -fold product of D D D and F t F_t F t vanishes in the quotient).
Part (2) is immediate since ∧ ∗ 2 j + 1 D = 0 \wedge_{\ast}^{2j+1}D=0 ∧ ∗ 2 j + 1 D = 0 whenever 2 j + 1 ≥ s + 1 2j+1\geq s+1 2 j + 1 ≥ s + 1 .
For (3), since ( D ∧ D ) ρ ( u , w ) = 2 D ρ ( u ) ∧ D ρ ( w ) (D\wedge D)_{\rho}(u,w)=2D_{\rho}(u)\wedge D_{\rho}(w) ( D ∧ D ) ρ ( u , w ) = 2 D ρ ( u ) ∧ D ρ ( w ) , evaluating ( ( D ∧ D ) ∧ D ) ρ ( v 1 , v 2 , v 3 ) ((D\wedge D)\wedge D)_{\rho}(v_1,v_2,v_3) (( D ∧ D ) ∧ D ) ρ ( v 1 , v 2 , v 3 ) directly from (6) sums ( 2 D ρ ( v σ ( 1 ) ) ∧ D ρ ( v σ ( 2 ) ) ) ∧ D ρ ( v σ ( 3 ) ) (2D_{\rho}(v_{\sigma(1)})\wedge D_{\rho}(v_{\sigma(2)}))\wedge D_{\rho}(v_{\sigma(3)}) ( 2 D ρ ( v σ ( 1 ) ) ∧ D ρ ( v σ ( 2 ) )) ∧ D ρ ( v σ ( 3 ) ) with prefactor 1 2 ! 1 ! \frac{1}{2!1!} 2 ! 1 ! 1 over σ ∈ S 3 \sigma\in S_3 σ ∈ S 3 ; grouping into the 3 cycles ( 123 ) , ( 231 ) , ( 312 ) (123),(231),(312) ( 123 ) , ( 231 ) , ( 312 ) yields 2 Jac ∧ ( D ρ ( v 1 ) , D ρ ( v 2 ) , D ρ ( v 3 ) ) = 1 3 Jac ∧ ( D , D , D ) ρ ( v 1 , v 2 , v 3 ) 2\operatorname{Jac}_{\wedge}(D_{\rho}(v_1),D_{\rho}(v_2),D_{\rho}(v_3))=\frac{1}{3}\operatorname{Jac}_{\wedge}(D,D,D)_{\rho}(v_1,v_2,v_3) 2 Jac ∧ ( D ρ ( v 1 ) , D ρ ( v 2 ) , D ρ ( v 3 )) = 3 1 Jac ∧ ( D , D , D ) ρ ( v 1 , v 2 , v 3 ) by (24) . Furthermore, since α ∧ β = − ( − 1 ) k l β ∧ α \alpha\wedge\beta=-(-1)^{kl}\beta\wedge\alpha α ∧ β = − ( − 1 ) k l β ∧ α , we have d ( D ∧ D ) = d D ∧ D − D ∧ d D = d D ∧ D − ( − 1 ) 1 ⋅ 2 d D ∧ D = 2 d D ∧ D d(D\wedge D)=dD\wedge D-D\wedge dD=dD\wedge D-(-1)^{1\cdot2}dD\wedge D=2dD\wedge D d ( D ∧ D ) = d D ∧ D − D ∧ d D = d D ∧ D − ( − 1 ) 1 ⋅ 2 d D ∧ D = 2 d D ∧ D , proving (57) . In the j j j -th vertex ( ∧ ρ r − 1 − j d D ) ∧ ( ∧ ρ 2 j + 1 D ) (\wedge_{\rho}^{r-1-j}dD)\wedge(\wedge_{\rho}^{2j+1}D) ( ∧ ρ r − 1 − j d D ) ∧ ( ∧ ρ 2 j + 1 D ) of (55) , d D ∧ d D = 0 dD\wedge dD=0 d D ∧ d D = 0 kills all terms with r − 1 − j ≥ 2 ⟺ j ≤ r − 3 r-1-j\geq2\Longleftrightarrow j\leq r-3 r − 1 − j ≥ 2 ⟺ j ≤ r − 3 . If Jac ∧ ≡ 0 \operatorname{Jac}_{\wedge}\equiv0 Jac ∧ ≡ 0 , then ∧ b a l 3 D = 0 \wedge_{\mathrm{bal}}^3D=0 ∧ bal 3 D = 0 and ∧ b a l 4 D = 0 \wedge_{\mathrm{bal}}^4D=0 ∧ bal 4 D = 0 , and by Lemma 2.1(2a), ∧ b a l m D = ∧ b l k m D = 0 \wedge_{\mathrm{bal}}^mD=\wedge_{\mathrm{blk}}^mD=0 ∧ bal m D = ∧ blk m D = 0 for all m ≥ 3 m\geq3 m ≥ 3 (2 j + 1 ≥ 3 ⟺ j ≥ 1 2j+1\geq3\Longleftrightarrow j\geq1 2 j + 1 ≥ 3 ⟺ j ≥ 1 ), leaving only j = 0 j=0 j = 0 at r = 2 r=2 r = 2 (n = 3 n=3 n = 3 ) and killing all terms for r ≥ 3 r\geq3 r ≥ 3 (n ≥ 5 n\geq5 n ≥ 5 ). If Jac ∧ ≢ 0 \operatorname{Jac}_{\wedge}\not\equiv0 Jac ∧ ≡ 0 , under P = b l k \mathcal P=\mathrm{blk} P = blk we have ∧ b l k 2 j + 1 D = 0 \wedge_{\mathrm{blk}}^{2j+1}D=0 ∧ blk 2 j + 1 D = 0 for 2 j + 1 ≥ 4 ⟺ j ≥ 2 2j+1\geq4\Longleftrightarrow j\geq2 2 j + 1 ≥ 4 ⟺ j ≥ 2 , so only r − 2 ≤ j ≤ 1 r-2 \le j \le1 r − 2 ≤ j ≤ 1 can survive: j ∈ { 0 , 1 } j \in\{0,1\} j ∈ { 0 , 1 } survive at n = 3 n=3 n = 3 , j = 1 j=1 j = 1 survives at n = 5 n=5 n = 5 (r = 3 r=3 r = 3 ), and r − 2 ≤ j ≤ 1 r-2 \le j \le1 r − 2 ≤ j ≤ 1 is empty for r ≥ 4 r \ge4 r ≥ 4 (n ≥ 7 n \ge7 n ≥ 7 ). Under P = b a l P=\mathrm{bal} P = bal , since d D ∧ d D = 0 dD \wedge dD=0 d D ∧ d D = 0 restricts j j j to j ∈ { r − 2 , r − 1 } j \in\{r-2,r-1\} j ∈ { r − 2 , r − 1 } where 2 j + 1 ∈ { n − 2 , n } 2j+1 \in\{n-2,n\} 2 j + 1 ∈ { n − 2 , n } is odd, Lemma 2.1(2a),(2c) implies that ∧ b a l 2 j + 1 D = 0 \wedge_{\mathrm{bal}}^{2j+1}D=0 ∧ bal 2 j + 1 D = 0 unless 2 j + 1 = J m 2j+1=J_m 2 j + 1 = J m for an odd Jacobsthal number J m J_m J m (m ≥ 3 m \ge3 m ≥ 3 ) with J m ≤ dim R A J_m \le\dim_{\mathbb{R}} A J m ≤ dim R A .
Finally, for (4), since d ( d D ∧ D ) = 1 2 d 2 ( D ∧ D ) = 0 d(dD \wedge D)=\frac{1}{2}d^2(D \wedge D)=0 d ( d D ∧ D ) = 2 1 d 2 ( D ∧ D ) = 0 , differentiating C S 3 g r p ( D ) = d D ∧ D + 1 9 Jac ∧ ( D , D , D ) \mathrm{CS}^{\mathrm{grp}}_3(D)=dD \wedge D+\frac{1}{9}\operatorname{Jac}_{\wedge}(D,D,D) CS 3 grp ( D ) = d D ∧ D + 9 1 Jac ∧ ( D , D , D ) and C S 5 g r p , b l k ( D ) = 1 4 d D ∧ Jac ∧ ( D , D , D ) \mathrm{CS}^{\mathrm{grp,blk}}_5(D)=\frac{1}{4}dD \wedge\operatorname{Jac}_{\wedge}(D,D,D) CS 5 grp , blk ( D ) = 4 1 d D ∧ Jac ∧ ( D , D , D ) immediately gives (59) . □
Applications to A-BF Theories, the Yang-Mills-Higgs Mechanism, and Reductive Dualities
We conclude this section by applying our framework to A A A -valued B F BF B F theories [1 , 8 ] , the A A A -valued Yang-Mills-Higgs mechanism [17 , 14 , 35 , 30 ] (including an explicit octonionic Malcev Proca mass spectrum computation), and reductive dualities.
Proposition 5.4 (Cattaneo-Plebanski A A A -B F BF B F Relations). Let M M M be a compact oriented n n n -manifold (n ≥ 2 n \ge2 n ≥ 2 ) and E A = P × ρ A → M E_A=P\times_\rho A \to M E A = P × ρ A → M .
(1) Given pairings ⟨ ⋅ , ⋅ ⟩ B F \langle\cdot,\cdot\rangle_{\mathrm{BF}} ⟨ ⋅ , ⋅ ⟩ BF and ⟨ ⋅ , ⋅ ⟩ B \langle\cdot,\cdot\rangle_B ⟨ ⋅ , ⋅ ⟩ B on Ω 2 ( M ; E A ) \Omega^2(M;E_A) Ω 2 ( M ; E A ) and Ω n − 2 ( M ; E A ) \Omega^{n-2}(M;E_A) Ω n − 2 ( M ; E A ) related via a linear isomorphism ⋆ M : Ω 2 ( M ; E A ) → ∼ Ω n − 2 ( M ; E A ) \star_M:\Omega^2(M;E_A)\xrightarrow{\sim}\Omega^{n-2}(M;E_A) ⋆ M : Ω 2 ( M ; E A ) ∼ Ω n − 2 ( M ; E A ) by ⟨ Φ , B ⟩ B F = ⟨ ⋆ M Φ , B ⟩ B \langle\Phi,B\rangle_{\mathrm{BF}}=\langle\star_M\Phi,B\rangle_B ⟨ Φ , B ⟩ BF = ⟨ ⋆ M Φ , B ⟩ B with ⟨ ⋅ , ⋅ ⟩ B \langle\cdot,\cdot\rangle_B ⟨ ⋅ , ⋅ ⟩ B symmetric and non-degenerate, the deformed A A A -B F BF B F functional
S B F , μ A [ D , B ] : = ∫ M ( ⟨ j ( F D ) , j ( B ) ⟩ B F − μ 2 ⟨ j ( B ) , j ( B ) ⟩ B ) ( μ ≠ 0 ) (60) S^{A}_{\mathrm{BF},\mu}[D,B]:=\int_M\left(\langle j(F_D),j(B)\rangle_{\mathrm{BF}}-\frac{\mu}{2}\langle j(B),j(B)\rangle_B\right)\qquad(\mu\ne0) \tag*{(60)} S BF , μ A [ D , B ] := ∫ M ( ⟨ j ( F D ) , j ( B ) ⟩ BF − 2 μ ⟨ j ( B ) , j ( B ) ⟩ B ) ( μ = 0 ) (60) has the unique auxiliary Euler-Lagrange solution j ( B o n - s h e l l ( D ) ) = 1 μ ⋆ M j ( F D ) j(B_{\mathrm{on\text{-}shell}}(D))=\frac{1}{\mu}\star_M j(F_D) j ( B on - shell ( D )) = μ 1 ⋆ M j ( F D ) , whose substitution into (60) recovers 1 2 μ S A [ D ] \frac{1}{2\mu}S^A[D] 2 μ 1 S A [ D ] with ⟨ Φ , Ψ ⟩ A : = ⟨ ⋆ M Φ , ⋆ M Ψ ⟩ B \langle\Phi,\Psi\rangle_A:=\langle\star_M\Phi,\star_M\Psi\rangle_B ⟨ Φ , Ψ ⟩ A := ⟨ ⋆ M Φ , ⋆ M Ψ ⟩ B .
(2) In the multiplicative A A A -B F BF B F functional S B F , ∗ A [ D , B ] : = ∫ M tr ξ ( j ( F D ∧ ∗ B ) ) S^A_{\mathrm{BF},*}[D,B]:=\int_M\operatorname{tr}_{\xi}(j(F_D\wedge_*B)) S BF , ∗ A [ D , B ] := ∫ M tr ξ ( j ( F D ∧ ∗ B )) , if A ≃ A 0 ⊕ A 1 A\simeq A_0\oplus A_1 A ≃ A 0 ⊕ A 1 has a weak P P P -( k , s ) (k,s) ( k , s ) -solvable submodule A 0 A_0 A 0 and B e = ∧ ∗ , P n e B_e=\wedge^n_{*,\mathcal{P}}e B e = ∧ ∗ , P n e for e ∈ Ω h e q , ρ 0 ( P ; A 0 ) e\in\Omega^0_{\mathrm{heq},\rho}(P;A_0) e ∈ Ω heq , ρ 0 ( P ; A 0 ) , then B e = 0 B_e=0 B e = 0 for n ≥ k + s + 3 n\ge k+s+3 n ≥ k + s + 3 [21 ] ; if n = 2 r n=2r n = 2 r is even, A A A is weak P P P -( 2 , s ) (2,s) ( 2 , s ) -solvable, and B = ∧ ∗ , P r − 1 β B=\wedge^{r-1}_{*,\mathcal{P}}\beta B = ∧ ∗ , P r − 1 β for β ∈ Ω h e q , ρ 0 ( P ; A ) \beta\in\Omega^0_{\mathrm{heq},\rho}(P;A) β ∈ Ω heq , ρ 0 ( P ; A ) , then B = 0 B=0 B = 0 for n ≥ 2 ( s + 2 ) n\ge2(s+2) n ≥ 2 ( s + 2 ) .
Proof. In (1), varying (5.6) with respect to B ˙ \dot B B ˙ gives ∫ M ⟨ ⋆ M j ( F D ) − μ j ( B ) , j ( B ˙ ) ⟩ B = 0 ⟺ j ( B ) = 1 μ ⋆ M j ( F D ) \int_M\langle\star_Mj(F_D)-\mu j(B),j(\dot B)\rangle_B=0\Longleftrightarrow j(B)=\frac{1}{\mu}\star_Mj(F_D) ∫ M ⟨ ⋆ M j ( F D ) − μ j ( B ) , j ( B ˙ ) ⟩ B = 0 ⟺ j ( B ) = μ 1 ⋆ M j ( F D ) . Part (2) follows directly from ∧ ∗ , P k + s + 1 e = 0 \wedge^{k+s+1}_{*,\mathcal{P}}e=0 ∧ ∗ , P k + s + 1 e = 0 and ∧ ∗ , P s + 1 β = 0 \wedge^{s+1}_{*,\mathcal{P}}\beta=0 ∧ ∗ , P s + 1 β = 0 . □
Next, let ι : G ↪ G ^ \iota:G\hookrightarrow\widehat G ι : G ↪ G be a closed Lie subgroup inclusion inducing the extended principal G ^ \widehat G G -bundle P ^ = P × G G ^ → M \widehat P=P\times_G\widehat G\to M P = P × G G → M and canonical reduction ι P : P ↪ P ^ \iota_P:P\hookrightarrow\widehat P ι P : P ↪ P [20 , 22 ] , which defines the canonical global Higgs section ϕ P ∈ Higgs ( P ^ ; G ) : = Γ ( P ^ × G ^ ( G ^ / G ) ) \phi_P\in\operatorname{Higgs}(\widehat P;G):=\Gamma(\widehat P\times_{\widehat G}(\widehat G/G)) ϕ P ∈ Higgs ( P ; G ) := Γ ( P × G ( G / G )) . Suppose ( A ^ , ∗ ) (\widehat A,*) ( A , ∗ ) is a finite-dimensional real algebra with representation ρ : G ^ → Aut A l g ( A ^ ) \rho:\widehat G\to\operatorname{Aut}_{\mathrm{Alg}}(\widehat A) ρ : G → Aut Alg ( A ) admitting a G ^ \widehat G G -equivariant map θ : G ^ / G → A ^ \theta:\widehat G/G\to\widehat A θ : G / G → A , θ ( a g G ) = ρ ( a ) ( θ ( g G ) ) \theta(agG)=\rho(a)(\theta(gG)) θ ( a g G ) = ρ ( a ) ( θ ( g G )) , and let Φ ϕ : = j − 1 ( θ ∗ ( ϕ ) ) ∈ Ω h e q , ρ 0 ( P ^ ; A ^ ) \Phi_\phi:=j^{-1}(\theta_*(\phi))\in\Omega^0_{\mathrm{heq},\rho}(\widehat P;\widehat A) Φ ϕ := j − 1 ( θ ∗ ( ϕ )) ∈ Ω heq , ρ 0 ( P ; A ) be the lifted Higgs 0-form (where θ ∗ \theta_* θ ∗ denotes the induced bundle map P ^ × G ^ ( G ^ / G ) → E A ^ \widehat P\times_{\widehat G}(\widehat G/G)\to E_{\widehat A} P × G ( G / G ) → E A ). Along the unitary-gauge reduction ι P : P ↪ P ^ \iota_P:P\hookrightarrow\widehat P ι P : P ↪ P , the canonical section ϕ P \phi_P ϕ P takes the constant value Φ 0 : = θ ( e G ) ∈ A ^ G \Phi_0:=\theta(eG)\in\widehat A^G Φ 0 := θ ( e G ) ∈ A G , so d ( ι P ∗ Φ ϕ P ) = 0 d(\iota_P^*\Phi_{\phi_P})=0 d ( ι P ∗ Φ ϕ P ) = 0 . Consequently:
In the Ehresmann-compatible setting of Proposition 2.4 (d ρ ( X ) ( y ) = 1 2 [ η ( X ) , y ] ∗ d\rho(X)(y)=\frac{1}{2}[\eta(X),y]_* d ρ ( X ) ( y ) = 2 1 [ η ( X ) , y ] ∗ ), the commutator covariant derivative ∇ D ^ A ^ , [ ⋅ , ⋅ ] ∗ Φ ϕ : = d Φ ϕ + 1 2 D ^ ∧ [ ⋅ , ⋅ ] ∗ Φ ϕ ∈ Ω h e q , ρ 1 ( P ^ ; A ^ ) \nabla^{\widehat A,[\cdot,\cdot]_*}_{\widehat D}\Phi_\phi:=d\Phi_\phi+\frac{1}{2}\widehat D\wedge_{[\cdot,\cdot]_*}\Phi_\phi\in\Omega^1_{\mathrm{heq},\rho}(\widehat P;\widehat A) ∇ D A , [ ⋅ , ⋅ ] ∗ Φ ϕ := d Φ ϕ + 2 1 D ∧ [ ⋅ , ⋅ ] ∗ Φ ϕ ∈ Ω heq , ρ 1 ( P ; A ) pulls back along ι P : P → P ^ \iota_P:P\to\widehat P ι P : P → P at ϕ = ϕ P \phi=\phi_P ϕ = ϕ P to R Φ 0 ( ι P ∗ D ^ ) \mathcal{R}_{\Phi_0}(\iota_P^*\widehat D) R Φ 0 ( ι P ∗ D ) , where R Φ 0 : A ^ → A ^ \mathcal{R}_{\Phi_0}:\widehat A\to\widehat A R Φ 0 : A → A is the linear mass operator R Φ 0 ( X ) = 1 2 [ X , Φ 0 ] ∗ \mathcal{R}_{\Phi_0}(X)=\frac{1}{2}[X,\Phi_0]_* R Φ 0 ( X ) = 2 1 [ X , Φ 0 ] ∗ .
In the horizontal/flat setting (η = 0 \eta=0 η = 0 , d = d 0 d=d_0 d = d 0 ) with one-sided coupling ∇ D ^ A ^ Φ 0 = D ^ ∧ ∗ Φ 0 \nabla^{\widehat A}_{\widehat D}\Phi_0=\widehat D\wedge_*\Phi_0 ∇ D A Φ 0 = D ∧ ∗ Φ 0 , the linear mass operator is R Φ 0 ( X ) = X ∗ Φ 0 \mathcal{R}_{\Phi_0}(X)=X* \Phi_0 R Φ 0 ( X ) = X ∗ Φ 0 .
Given a positive-definite fiberwise inner product ⟨ ⋅ , ⋅ ⟩ A ^ \langle\cdot,\cdot\rangle_{\widehat A} ⟨ ⋅ , ⋅ ⟩ A on A ^ \widehat A A and a Riemannian metric g g g on M M M inducing the kinetic pairing ⟨ α , β ⟩ k i n = ⟨ α ∧ ∗ g β ⟩ A ^ \langle\alpha,\beta\rangle_{\mathrm{kin}}=\langle\alpha\wedge*_g\beta\rangle_{\widehat A} ⟨ α , β ⟩ kin = ⟨ α ∧ ∗ g β ⟩ A on Ω 1 ( M ; E A ^ ) \Omega^1(M;E_{\widehat A}) Ω 1 ( M ; E A ) , let N m a s s ( Φ 0 , x ) = rk ( R Φ 0 ) N_{\mathrm{mass}}(\Phi_0,x)=\operatorname{rk}(\mathcal{R}_{\Phi_0}) N mass ( Φ 0 , x ) = rk ( R Φ 0 ) denote the rank of the Proca mass quadratic form ⟨ j ( R Φ 0 ( D ^ ) ) , j ( R Φ 0 ( D ^ ) ) ⟩ k i n , x \langle j(\mathcal{R}_{\Phi_0}(\widehat D)),j(\mathcal{R}_{\Phi_0}(\widehat D))\rangle_{\mathrm{kin},x} ⟨ j ( R Φ 0 ( D )) , j ( R Φ 0 ( D )) ⟩ kin , x at x ∈ M x\in M x ∈ M , and let g Φ 0 : = ker ( R Φ 0 ) ⊂ A ^ \mathfrak{g}_{\Phi_0}:=\ker(\mathcal{R}_{\Phi_0})\subset\widehat A g Φ 0 := ker ( R Φ 0 ) ⊂ A be the algebraic stabilizer of Φ 0 \Phi_0 Φ 0 .
Proposition 5.5 (Unitary-Gauge Proca Mass Bounds and Octonionic/Reductive Spectra in A A A -YMH Theories). Let Φ 0 = θ ( e G ) ∈ A ^ G \Phi_0=\theta(eG)\in\widehat A^G Φ 0 = θ ( e G ) ∈ A G be the unitary-gauge Higgs vacuum on P ↪ P ^ P\hookrightarrow\widehat P P ↪ P .
(1) Splitting Kernel Decoupling: If A ^ \hat{A} A ^ arises from a split extension 0 → H → A ^ → q A → 0 0 \to H \to\hat{A} \xrightarrow{q} A \to0 0 → H → A ^ q A → 0 with pulled-back product x ∗ ′ y = s ( q ( x ) ∗ q ( y ) ) x *' y = s(q(x) * q(y)) x ∗ ′ y = s ( q ( x ) ∗ q ( y )) and Φ 0 ∈ H = ker ( q ) \Phi_0 \in H = \ker(q) Φ 0 ∈ H = ker ( q ) , then R Φ 0 ≡ 0 R_{\Phi_0} \equiv0 R Φ 0 ≡ 0 and N m a s s ( Φ 0 , x ) = 0 N_{\mathrm{mass}}(\Phi_0,x)=0 N mass ( Φ 0 , x ) = 0 .
(2) Derived and Commutator Rank Bounds: Because im ( R Φ 0 ) ⊂ [ A ^ , A ^ ] ∗ \operatorname{im}(R_{\Phi_0}) \subset[\hat{A},\hat{A}]_* im ( R Φ 0 ) ⊂ [ A ^ , A ^ ] ∗ for the commutator coupling (resp. im ( R Φ 0 ) ⊂ A ^ ( 1 ) = A ^ ∗ A ^ \operatorname{im}(R_{\Phi_0}) \subset\hat{A}^{(1)} = \hat{A} * \hat{A} im ( R Φ 0 ) ⊂ A ^ ( 1 ) = A ^ ∗ A ^ for the one-sided coupling), the Proca mass rank and the algebraic stabilizer dimension dim R g Φ 0 = dim R A ^ − rk ( R Φ 0 ) \dim_{\mathbb{R}}\mathfrak{g}_{\Phi_0} = \dim_{\mathbb{R}}\hat{A} - \operatorname{rk}(R_{\Phi_0}) dim R g Φ 0 = dim R A ^ − rk ( R Φ 0 ) satisfy
N m a s s ( Φ 0 , x ) ≤ rk ( R Φ 0 ) ≤ dim R [ A ^ , A ^ ] ∗ , dim R g Φ 0 ≥ dim R A ^ − dim R [ A ^ , A ^ ] ∗ ( c o m m . ) (61) N_{\mathrm{mass}}(\Phi_0,x) \le\operatorname{rk}(R_{\Phi_0}) \le\dim_{\mathbb{R}}[\hat{A},\hat{A}]_*, \qquad\dim_{\mathbb{R}}\mathfrak{g}_{\Phi_0} \ge\dim_{\mathbb{R}}\hat{A} - \dim_{\mathbb{R}}[\hat{A},\hat{A}]_* \quad(\mathrm{comm.}) \tag*{(61)} N mass ( Φ 0 , x ) ≤ rk ( R Φ 0 ) ≤ dim R [ A ^ , A ^ ] ∗ , dim R g Φ 0 ≥ dim R A ^ − dim R [ A ^ , A ^ ] ∗ ( comm. ) (61)
N m a s s ( Φ 0 , x ) ≤ rk ( R Φ 0 ) ≤ dim R A ^ ( 1 ) , dim R g Φ 0 ≥ dim R A ^ − dim R A ^ ( 1 ) ( o n e − s i d e d ) . (62) N_{\mathrm{mass}}(\Phi_0,x) \le\operatorname{rk}(R_{\Phi_0}) \le\dim_{\mathbb{R}}\hat{A}^{(1)}, \qquad\dim_{\mathbb{R}}\mathfrak{g}_{\Phi_0} \ge\dim_{\mathbb{R}}\hat{A} - \dim_{\mathbb{R}}\hat{A}^{(1)} \quad(\mathrm{one-sided}). \tag*{(62)} N mass ( Φ 0 , x ) ≤ rk ( R Φ 0 ) ≤ dim R A ^ ( 1 ) , dim R g Φ 0 ≥ dim R A ^ − dim R A ^ ( 1 ) ( one − sided ) . (62)
If Φ 0 ∈ A ^ ( r ) \Phi_0 \in\hat{A}^{(r)} Φ 0 ∈ A ^ ( r ) (resp. Φ 0 ∈ A ^ r \Phi_0 \in\hat{A}_r Φ 0 ∈ A ^ r ), then N m a s s ( Φ 0 , x ) ≤ dim R ( A ^ ∗ A ^ ( r ) + A ^ ( r ) ∗ A ^ ) N_{\mathrm{mass}}(\Phi_0,x) \le\dim_{\mathbb{R}}(\hat{A} * \hat{A}^{(r)} + \hat{A}^{(r)} * \hat{A}) N mass ( Φ 0 , x ) ≤ dim R ( A ^ ∗ A ^ ( r ) + A ^ ( r ) ∗ A ^ ) (resp. ≤ dim R A ^ r + 1 \le\dim_{\mathbb{R}}\hat{A}_{r+1} ≤ dim R A ^ r + 1 ).
(3) Explicit Octonionic Malcev and Reductive Cartan Mass Spectra:
(a) Octonionic Gauge Theory (A ^ = O = R ⋅ 1 ⊕ Im ( O ) \hat{A} = \mathbb{O} = \mathbb{R}\cdot1 \oplus\operatorname{Im}(\mathbb{O}) A ^ = O = R ⋅ 1 ⊕ Im ( O ) ): For any vacuum Φ 0 = c 0 ⋅ 1 + ϕ ⃗ 0 ∈ O \Phi_0 = c_0 \cdot1 + \vec{\phi}_0 \in\mathbb{O} Φ 0 = c 0 ⋅ 1 + ϕ 0 ∈ O with ∣ ϕ ⃗ 0 ∣ = v > 0 |\vec{\phi}_0| = v > 0 ∣ ϕ 0 ∣ = v > 0 under the commutator coupling R Φ 0 ( X ) = 1 2 [ X , Φ 0 ] ∗ = X ⃗ × G 2 ϕ ⃗ 0 R_{\Phi_0}(X) = \frac{1}{2}[X,\Phi_0]_* = \vec{X} \times_{G_2} \vec{\phi}_0 R Φ 0 ( X ) = 2 1 [ X , Φ 0 ] ∗ = X × G 2 ϕ 0 , the central unit R ⋅ 1 \mathbb{R}\cdot1 R ⋅ 1 and the parallel imaginary line R ⋅ ϕ ⃗ 0 \mathbb{R}\cdot\vec{\phi}_0 R ⋅ ϕ 0 span a 2-dimensional massless stabilizer g Φ 0 = R ⋅ 1 ⊕ R ⋅ ϕ ⃗ 0 ≃ C \mathfrak{g}_{\Phi_0} = \mathbb{R}\cdot1 \oplus\mathbb{R}\cdot\vec{\phi}_0 \simeq\mathbb{C} g Φ 0 = R ⋅ 1 ⊕ R ⋅ ϕ 0 ≃ C , while on the orthogonal 6-plane ϕ ⃗ 0 ⊥ ⊂ Im ( O ) \vec{\phi}_0^\perp\subset\operatorname{Im}(\mathbb{O}) ϕ 0 ⊥ ⊂ Im ( O ) the operator R Φ 0 2 = − v 2 id ϕ ⃗ 0 ⊥ R_{\Phi_0}^2 = -v^2\operatorname{id}_{\vec{\phi}_0^\perp} R Φ 0 2 = − v 2 id ϕ 0 ⊥ gives N m a s s ( Φ 0 ) = 6 N_{\mathrm{mass}}(\Phi_0)=6 N mass ( Φ 0 ) = 6 degenerate massive Proca gauge bosons of mass m = v m=v m = v (breaking G 2 → S U ( 3 ) G_2 \to SU(3) G 2 → S U ( 3 ) , the stabilizer of ϕ ⃗ 0 \vec{\phi}_0 ϕ 0 in G 2 G_2 G 2 ).
(b) Reductive Cartan Algebra (A ^ = R k ⊕ s o ( k ) \hat{A} = \mathbb{R}^k \oplus\mathfrak{so}(k) A ^ = R k ⊕ so ( k ) under ⋉ \ltimes ⋉ with × H = 0 \times_H=0 × H = 0 ): For a pure translational vacuum Φ 0 = ( v 0 , 0 ) ∈ R k ⊕ { 0 } \Phi_0=(v_0,0)\in\mathbb{R}^k\oplus\{0\} Φ 0 = ( v 0 , 0 ) ∈ R k ⊕ { 0 } with ∣ v 0 ∣ = v > 0 |v_0|=v>0 ∣ v 0 ∣ = v > 0 , R Φ 0 ( u , X ) = ( u , X ) ⋉ ( v 0 , 0 ) = ( 1 2 X v 0 , σ ( u ∧ v 0 ) ) R_{\Phi_0}(u,X)=(u,X)\ltimes(v_0,0)=(\frac{1}{2}Xv_0,\sigma(u\wedge v_0)) R Φ 0 ( u , X ) = ( u , X ) ⋉ ( v 0 , 0 ) = ( 2 1 X v 0 , σ ( u ∧ v 0 )) . Here ker ( R Φ 0 ) = R v 0 ⊕ s o ( k − 1 ) v 0 \ker(R_{\Phi_0})=\mathbb{R}v_0\oplus\mathfrak{so}(k-1)_{v_0} ker ( R Φ 0 ) = R v 0 ⊕ so ( k − 1 ) v 0 has dimension 1 + ( k − 1 2 ) 1+\binom{k-1}{2} 1 + ( 2 k − 1 ) , while N m a s s ( Φ 0 ) = 2 ( k − 1 ) N_{\mathrm{mass}}(\Phi_0)=2(k-1) N mass ( Φ 0 ) = 2 ( k − 1 ) gauge modes (k − 1 k-1 k − 1 boost/coset generators in s o ( k ) / s o ( k − 1 ) \mathfrak{so}(k)/\mathfrak{so}(k-1) so ( k ) / so ( k − 1 ) with mass m r o t = 1 2 v m_{\mathrm{rot}}=\frac{1}{2}v m rot = 2 1 v and k − 1 k-1 k − 1 orthogonal soldering modes in v 0 ⊥ ⊂ R k v_0^\perp\subset\mathbb{R}^k v 0 ⊥ ⊂ R k with mass m s o l d = v m_{\mathrm{sold}}=v m sold = v ) acquire Proca masses.
Proof. Part (1) follows from q ( Φ 0 ) = 0 ⟹ R Φ 0 ( X ) = 0 q(\Phi_0)=0\Longrightarrow R_{\Phi_0}(X)=0 q ( Φ 0 ) = 0 ⟹ R Φ 0 ( X ) = 0 , and part (2) follows from rk ( R Φ 0 ) = dim R im ( R Φ 0 ) \operatorname{rk}(R_{\Phi_0})=\dim_{\mathbb{R}}\operatorname{im}(R_{\Phi_0}) rk ( R Φ 0 ) = dim R im ( R Φ 0 ) together with 1 2 [ X , Φ 0 ] ∗ ∈ [ A ^ , A ^ ] ∗ \frac{1}{2}[X,\Phi_0]_*\in[\hat{A},\hat{A}]_* 2 1 [ X , Φ 0 ] ∗ ∈ [ A ^ , A ^ ] ∗ and X ∗ Φ 0 ∈ A ^ ( 1 ) X*\Phi_0\in\hat{A}^{(1)} X ∗ Φ 0 ∈ A ^ ( 1 ) . For (3a), in O \mathbb{O} O , 1 2 [ x 0 + x ⃗ , c 0 + ϕ ⃗ 0 ] ∗ = x ⃗ × G 2 ϕ ⃗ 0 \frac{1}{2}[x_0+\vec{x},c_0+\vec{\phi}_0]_*=\vec{x}\times_{G_2}\vec{\phi}_0 2 1 [ x 0 + x , c 0 + ϕ 0 ] ∗ = x × G 2 ϕ 0 , and by the octonionic cross-product identity ( x ⃗ × G 2 ϕ ⃗ 0 ) × G 2 ϕ ⃗ 0 = ⟨ x ⃗ , ϕ ⃗ 0 ⟩ ϕ ⃗ 0 − ∣ ϕ ⃗ 0 ∣ 2 x ⃗ (\vec{x}\times_{G_2}\vec{\phi}_0)\times_{G_2}\vec{\phi}_0=\langle\vec{x},\vec{\phi}_0\rangle\vec{\phi}_0-|\vec{\phi}_0|^2\vec{x} ( x × G 2 ϕ 0 ) × G 2 ϕ 0 = ⟨ x , ϕ 0 ⟩ ϕ 0 − ∣ ϕ 0 ∣ 2 x , the operator R Φ 0 ∗ R Φ 0 R_{\Phi_0}^*R_{\Phi_0} R Φ 0 ∗ R Φ 0 has eigenvalue 0 0 0 on span { 1 , ϕ ⃗ 0 } \operatorname{span}\{1,\vec{\phi}_0\} span { 1 , ϕ 0 } (multiplicity 2) and eigenvalue v 2 v^2 v 2 on ϕ ⃗ 0 ⊥ ⊂ Im ( O ) \vec{\phi}_0^\perp\subset\operatorname{Im}(\mathbb{O}) ϕ 0 ⊥ ⊂ Im ( O ) (multiplicity 6). For (3b), X v 0 = 0 ⟺ X ∈ s o ( k − 1 ) v 0 Xv_0=0\Longleftrightarrow X\in\mathfrak{so}(k-1)_{v_0} X v 0 = 0 ⟺ X ∈ so ( k − 1 ) v 0 and u ∧ v 0 = 0 ⟺ u ∈ R v 0 u\wedge v_0=0\Longleftrightarrow u\in\mathbb{R}v_0 u ∧ v 0 = 0 ⟺ u ∈ R v 0 , giving dim ker ( R Φ 0 ) = 1 + ( k − 1 2 ) \dim\ker(R_{\Phi_0})=1+\binom{k-1}{2} dim ker ( R Φ 0 ) = 1 + ( 2 k − 1 ) and rk ( R Φ 0 ) = 2 ( k − 1 ) \operatorname{rk}(R_{\Phi_0})=2(k-1) rk ( R Φ 0 ) = 2 ( k − 1 ) . □ \square □
Finally, given a reductive decomposition A ≃ A 0 ⊕ A 1 A\simeq A_0\oplus A_1 A ≃ A 0 ⊕ A 1 where all products are evaluated via the ambient multiplication ∗ * ∗ of A A A , we define the geometric dual ∗ S A [ e , ω ] : = ∫ M ⟨ j ( ∗ F ( e , ω ) ) , j ( ∗ F ( e , ω ) ) ⟩ A {}^*S^A[e,\omega]:=\int_M\langle j({}^*F_{(e,\omega)}),j({}^*F_{(e,\omega)})\rangle_A ∗ S A [ e , ω ] := ∫ M ⟨ j ( ∗ F ( e , ω ) ) , j ( ∗ F ( e , ω ) ) ⟩ A on Ω h e q , ρ 1 ( P ; A 0 ) ⊕ Ω h e q , ρ 1 ( P ; A 1 ) \Omega^1_{\mathrm{heq},\rho}(P;A_0)\oplus\Omega^1_{\mathrm{heq},\rho}(P;A_1) Ω heq , ρ 1 ( P ; A 0 ) ⊕ Ω heq , ρ 1 ( P ; A 1 ) by interchanging the roles of e e e and ω \omega ω in the differential and quadratic sectors (∗ F ( e , ω ) : = ( d e + 1 2 e ∧ Λ e ) + ( d ω + 1 2 ω ∧ [ , ] ∗ e ) + 1 2 ω ∧ λ ω {}^*F_{(e,\omega)}:=(de+\frac{1}{2}e\wedge_{\Lambda}e)+(d\omega+\frac{1}{2}\omega\wedge_{[,]_*}e)+\frac{1}{2}\omega\wedge_{\lambda}\omega ∗ F ( e , ω ) := ( d e + 2 1 e ∧ Λ e ) + ( d ω + 2 1 ω ∧ [ , ] ∗ e ) + 2 1 ω ∧ λ ω ), and the algebraic dual ( S A ) ∗ [ e ′ , ω ′ ] : = ∫ M ⟨ j ( F ( e ′ , ω ′ ) ) , j ( F ( e ′ , ω ′ ) ) ⟩ A (S^A)^*[e',\omega']:=\int_M\langle j(F_{(e',\omega')}),j(F_{(e',\omega')})\rangle_A ( S A ) ∗ [ e ′ , ω ′ ] := ∫ M ⟨ j ( F ( e ′ , ω ′ ) ) , j ( F ( e ′ , ω ′ ) ) ⟩ A on the swapped configuration space ( e ′ , ω ′ ) ∈ Ω h e q , ρ 1 ( P ; A 1 ) ⊕ Ω h e q , ρ 1 ( P ; A 0 ) (e',\omega')\in\Omega^1_{\mathrm{heq},\rho}(P;A_1)\oplus\Omega^1_{\mathrm{heq},\rho}(P;A_0) ( e ′ , ω ′ ) ∈ Ω heq , ρ 1 ( P ; A 1 ) ⊕ Ω heq , ρ 1 ( P ; A 0 ) with F ( e ′ , ω ′ ) : = ( d ω ′ + 1 2 ω ′ ∧ Λ ω ′ ) + ( d e ′ + 1 2 ω ′ ∧ [ , ] ∗ e ′ ) + 1 2 e ′ ∧ Λ e ′ F_{(e',\omega')}:=(d\omega'+\frac{1}{2}\omega'\wedge_{\Lambda}\omega')+(de'+\frac{1}{2}\omega'\wedge_{[,]_*}e')+\frac{1}{2}e'\wedge_{\Lambda}e' F ( e ′ , ω ′ ) := ( d ω ′ + 2 1 ω ′ ∧ Λ ω ′ ) + ( d e ′ + 2 1 ω ′ ∧ [ , ] ∗ e ′ ) + 2 1 e ′ ∧ Λ e ′ . Applying Proposition 3.1 and Theorem 4.5 (1) directly to the swapped 1-forms yields:
Proposition 5.6 (Duality Transfer of Obstructions). Let A ≃ A 0 ⊕ A 1 A\simeq A_0\oplus A_1 A ≃ A 0 ⊕ A 1 be an m m m -graded algebra. If A 0 A_0 A 0 is weak ( 1 , 1 ) (1,1) ( 1 , 1 ) -solvable under ∗ * ∗ , then e ∧ Λ e = 0 e\wedge_{\Lambda}e=0 e ∧ Λ e = 0 and ω ′ ∧ Λ ω ′ = 0 \omega'\wedge_{\Lambda}\omega'=0 ω ′ ∧ Λ ω ′ = 0 , abelianizing the differential curvature sector for both ∗ S A {}^*S^A ∗ S A and ( S A ) ∗ (S^A)^* ( S A ) ∗ in all dimensions n ≥ 2 n\ge2 n ≥ 2 . If A 1 = h ⊂ s o ( k 1 ) ⊕ ⋯ ⊕ s o ( k r ) A_1=\mathfrak{h}\subset\mathfrak{so}(k_1)\oplus\cdots\oplus\mathfrak{so}(k_r) A 1 = h ⊂ so ( k 1 ) ⊕ ⋯ ⊕ so ( k r ) is a k k k -algebra under λ \lambda λ , then the block-and-balanced dual Chern-Simons functionals ∗ S n , s , b l k C Y M [ e , ω ] {}^*S^{\mathrm{CYM}}_{n,\mathfrak{s},\mathrm{blk}}[e,\omega] ∗ S n , s , blk CYM [ e , ω ] and ( S n , s , b l k C Y M ) ∗ [ e ′ , ω ′ ] (S^{\mathrm{CYM}}_{n,\mathfrak{s},\mathrm{blk}})^*[e',\omega'] ( S n , s , blk CYM ) ∗ [ e ′ , ω ′ ] (formed with constitutive powers λ b l k m ω \lambda_{\mathrm{blk}}^m\omega λ blk m ω and λ b l k m e ′ \lambda_{\mathrm{blk}}^m e' λ blk m e ′ ) have vanishing cosmological terms for n ≥ 4 n\ge4 n ≥ 4 , vanishing EHP terms for n ≥ 6 n\ge6 n ≥ 6 , and vanish identically for n ≥ 8 n\ge8 n ≥ 8 .
Realizations and Concrete/Abstract Examples
We conclude by applying our theorems to concrete matrix geometries, Cayley-Dickson geometries, and abstract non-associative and graded algebras.
Classical, Generalized Complex, and Exceptional Geometries
By Theorem 3.2 (2) and Theorem 4.5 (1), any algebra ( A , ∗ ) (A, \ast) ( A , ∗ ) that is ( 2 , 1 ) (2,1) ( 2 , 1 ) -nil and block-balanced ( 1 , 3 ) (1,3) ( 1 , 3 ) -nil obstructs 4D internal-product topological/self-dual multiplicative A-YMT theories and cuts off block-balanced Cartan-Yang-Mills theories for n ≥ 8 n \ge8 n ≥ 8 . Since every subalgebra h ⊂ s o ( k 1 ) ⊕ ⋯ ⊕ s o ( k r ) \mathfrak{h} \subset\mathfrak{so}(k_{1}) \oplus\cdots\oplus\mathfrak{so}(k_{r}) h ⊂ so ( k 1 ) ⊕ ⋯ ⊕ so ( k r ) (as well as h ⊗ R C ⊂ s o ( k ; C ) \mathfrak{h} \otimes_{\mathbb{R}} \mathbb{C} \subset\mathfrak{so}(k;\mathbb{C}) h ⊗ R C ⊂ so ( k ; C ) and R k ⊕ h \mathbb{R}^{k} \oplus\mathfrak{h} R k ⊕ h under ∧ × \wedge_{\times} ∧ × ) is skew-commutative, Lemma 2.1 (1) shows that this encompasses:
Classical G G G -Structures in Berger’s List and Spinor Embeddings: s o ( k ) \mathfrak{so}(k) so ( k ) , u ( k ) \mathfrak{u}(k) u ( k ) , s u ( k ) \mathfrak{su}(k) su ( k ) , s p ( k ) \mathfrak{sp}(k) sp ( k ) , s p ( k ) ⊕ s p ( 1 ) \mathfrak{sp}(k) \oplus\mathfrak{sp}(1) sp ( k ) ⊕ sp ( 1 ) , g 2 ⊂ s o ( 7 ) \mathfrak{g}_{2} \subset\mathfrak{so}(7) g 2 ⊂ so ( 7 ) , s p i n ( 7 ) ⊂ s o ( 8 ) \mathfrak{spin}(7) \subset\mathfrak{so}(8) spin ( 7 ) ⊂ so ( 8 ) , and s p i n ( 9 ) ⊂ s o ( 16 ) \mathfrak{spin}(9) \subset\mathfrak{so}(16) spin ( 9 ) ⊂ so ( 16 ) , as well as their covering groups Spin ( k ) \operatorname{Spin}(k) Spin ( k ) and product distributions [6 , 7 , 21 ] .
Type II String and Hitchin’s Generalized Geometries: The Narain orthogonal reduction O ( k ) × O ( k ) ↪ O ( k , k ) O(k) \times O(k) \hookrightarrow O(k,k) O ( k ) × O ( k ) ↪ O ( k , k ) [23 , 9 ] , Hitchin and Gualtieri’s generalized complex geometry U ( k , k ) ↪ O ( 2 k , 2 k ) U(k,k) \hookrightarrow O(2k,2k) U ( k , k ) ↪ O ( 2 k , 2 k ) [18 , 15 ] , generalized Calabi-Yau geometry S U ( k , k ) ↪ O ( 2 k , 2 k ) SU(k,k) \hookrightarrow O(2k,2k) S U ( k , k ) ↪ O ( 2 k , 2 k ) , and generalized Kähler geometry U ( k ) × U ( k ) ↪ O ( 2 k , 2 k ) U(k) \times U(k) \hookrightarrow O(2k,2k) U ( k ) × U ( k ) ↪ O ( 2 k , 2 k ) [15 ] .
Exotic Orthogonal Subgroups and Exceptional Lie Algebras: Obata’s 15 infinite families of subgroups H ⊂ O ( k ) H \subset O(k) H ⊂ O ( k ) [24 ] , maximal tori T ( q ; K ) ↪ O ( q ; K ) T(q;\mathbb{K}) \hookrightarrow O(q;\mathbb{K}) T ( q ; K ) ↪ O ( q ; K ) (ch ( K ) ≠ 2 \operatorname{ch}(\mathbb{K}) \ne2 ch ( K ) = 2 [4 ] ), and the exceptional orthogonal embeddings O ( 3 ) ↪ G 2 O(3) \hookrightarrow G_{2} O ( 3 ) ↪ G 2 , O ( 9 ) ↪ F 4 O(9) \hookrightarrow F_{4} O ( 9 ) ↪ F 4 , O ( 10 ) ↪ E 6 O(10) \hookrightarrow E_{6} O ( 10 ) ↪ E 6 , O ( 12 ) ↪ E 7 O(12) \hookrightarrow E_{7} O ( 12 ) ↪ E 7 , and O ( 16 ) ↪ E 8 O(16) \hookrightarrow E_{8} O ( 16 ) ↪ E 8 .
Infinite Families of Real and Generalized Cayley-Dickson Geometries
Let ( A , ∗ ) (A,\ast) ( A , ∗ ) be a finite-dimensional real algebra with involution ( − ) ‾ : A → A \overline{(-)}: A \to A ( − ) : A → A , and let CD ( A ) = A ⊕ A \operatorname{CD}(A)=A\oplus A CD ( A ) = A ⊕ A be its Cayley-Dickson double with ( x , y ) ∗ ( z , w ) = ( x ∗ z − w ‾ ∗ y , w ∗ x + y ∗ z ‾ ) (x,y)\ast(z,w)=(x\ast z-\overline{w}\ast y,w\ast x+y\ast\overline{z}) ( x , y ) ∗ ( z , w ) = ( x ∗ z − w ∗ y , w ∗ x + y ∗ z ) and ( x , y ) ‾ = ( x ‾ , − y ) \overline{(x,y)}=(\overline{x},-y) ( x , y ) = ( x , − y ) [31 , 2 ] . For k ≥ 1 k \ge1 k ≥ 1 , let U ( k ; A ) ⊂ Mat k × k ( A ) U(k;A)\subset\operatorname{Mat}_{k\times k}(A) U ( k ; A ) ⊂ Mat k × k ( A ) denote the A A A -unitary matrices preserving s ( x , y ) = ∑ i = 1 k x ‾ i ∗ y i s(x,y)=\sum_{i=1}^{k}\overline{x}_{i}\ast y_{i} s ( x , y ) = ∑ i = 1 k x i ∗ y i [21 ] . Iterating U ( k ; CD ( A ) ) ↪ U ( 2 k ; A ) U(k;\operatorname{CD}(A))\hookrightarrow U(2k;A) U ( k ; CD ( A )) ↪ U ( 2 k ; A ) yields for every l ≥ 1 l \ge1 l ≥ 1 [21 ] :
U ( k ; CD l ( R ) ) ↪ ⋯ ↪ U ( 2 l − 2 k ; H ) ↪ U ( 2 l − 1 k ; C ) ↪ O ( 2 l k ; R ) (63) U(k;\operatorname{CD}^{l}(\mathbb{R}))\hookrightarrow\cdots\hookrightarrow U(2^{l-2}k;\mathbb{H})\hookrightarrow U(2^{l-1}k;\mathbb{C})\hookrightarrow O(2^{l}k;\mathbb{R}) \tag*{(63)} U ( k ; CD l ( R )) ↪ ⋯ ↪ U ( 2 l − 2 k ; H ) ↪ U ( 2 l − 1 k ; C ) ↪ O ( 2 l k ; R ) (63)
U ( k , k ; CD l ( A ) ) ↪ O ( 2 l k , 2 l k ; CD l ( A ) ) ≅ O ( 2 l + 1 k ; CD l ) (64) U(k,k;\operatorname{CD}^{l}(A))\hookrightarrow O(2^{l}k,2^{l}k;\operatorname{CD}^{l}(A))\cong O(2^{l+1}k;\operatorname{CD}^{l}) \tag*{(64)} U ( k , k ; CD l ( A )) ↪ O ( 2 l k , 2 l k ; CD l ( A )) ≅ O ( 2 l + 1 k ; CD l ) (64)
recovering Hitchin’s generalized complex geometry for l = 1 l=1 l = 1 [15 ] and Pantilie’s generalized quaternionic geometry for l = 2 l=2 l = 2 [27 , 11 ] . Because the algebra o ( m ; B ) \mathfrak{o}(m;B) o ( m ; B ) of B B B -skew-Hermitian matrices under X ∧ Y = 1 2 ( X Y − Y X ) X \wedge Y=\frac{1}{2}(XY-YX) X ∧ Y = 2 1 ( X Y − Y X ) is skew-commutative for any involutive algebra B B B , u ( k ; CD l ( R ) ) \mathfrak{u}(k;\operatorname{CD}^{l}(\mathbb{R})) u ( k ; CD l ( R )) is ( 2 , 1 ) (2,1) ( 2 , 1 ) -nil and block-balanced ( 1 , 3 ) (1,3) ( 1 , 3 ) -nil for all l ≥ 1 l \ge1 l ≥ 1 . Moreover, u ( k ; CD l ( R ) ) \mathfrak{u}(k;\operatorname{CD}^{l}(\mathbb{R})) u ( k ; CD l ( R )) is a Lie algebra (Jac ∧ = 0 \operatorname{Jac}_{\wedge}=0 Jac ∧ = 0 ) for l ∈ { 1 , 2 } l\in\{1,2\} l ∈ { 1 , 2 } (C , H \mathbb{C},\mathbb{H} C , H ), whereas for l ≥ 3 l\ge3 l ≥ 3 (O \mathbb{O} O , sedenions, etc.) it is a non-Lie skew-commutative algebra with Jac ∧ ≠ 0 \operatorname{Jac}_{\wedge}\ne0 Jac ∧ = 0 .
Corollary 6.1 (Obstructions in Cayley-Dickson Geometries of Arbitrary Order). For every k ≥ 1 k \ge1 k ≥ 1 and l ≥ 1 l \ge1 l ≥ 1 , the algebras u ( k ; CD l ( R ) ) \mathfrak{u}(k;\operatorname{CD}^{l}(\mathbb{R})) u ( k ; CD l ( R )) and u ( k , k ; CD l ( A ) ) \mathfrak{u}(k,k;\operatorname{CD}^{l}(A)) u ( k , k ; CD l ( A )) (and their reductive extensions by R 2 l k \mathbb{R}^{2^{l}k} R 2 l k ) are ( 2 , 1 ) (2,1) ( 2 , 1 ) -nil and block-balanced ( 1 , 3 ) (1,3) ( 1 , 3 ) -nil under ∧ \wedge ∧ and ∧ × \wedge_{\times} ∧ × . Therefore:
(1) Every 4D internal-product topological A-YMT theory ∫ M tr ξ ( j ( F D ∧ F D ) ) \int_{M}\operatorname{tr}_{\xi}(j(F_{D}\wedge F_{D})) ∫ M tr ξ ( j ( F D ∧ F D )) and every 4D self-dual internal multiplicative A-YMT theory vanishes identically;
(2) Every block-balanced Cartan-Yang-Mills theory (45) (and its geometric and algebraic duals) has a vanishing cosmological term for n ≥ 4 n\ge4 n ≥ 4 , a vanishing EHP term for n ≥ 6 n\ge6 n ≥ 6 , and vanishes identically for n ≥ 8 n\ge8 n ≥ 8 ;
(3) For the unquantized grouped A-valued Chern-Simons functional S C S , 2 r − 1 A , g r p [ D ] S_{\mathrm{CS},2r-1}^{A,\mathrm{grp}}[D] S CS , 2 r − 1 A , grp [ D ] valued in the algebra u ( k ; CD l ( R ) ) \mathfrak{u}(k;\operatorname{CD}^{l}(\mathbb{R})) u ( k ; CD l ( R )) : when l ∈ { 1 , 2 } l\in\{1,2\} l ∈ { 1 , 2 } (Jac ∧ = 0 \operatorname{Jac}_{\wedge}=0 Jac ∧ = 0 ), the 3D cubic vertex vanishes and all ≥ 5 \ge5 ≥ 5 D internal A-CS theories vanish identically under both blk and bal; when l ≥ 3 l\ge3 l ≥ 3 (Jac ∧ ≠ 0 \operatorname{Jac}_{\wedge}\ne0 Jac ∧ = 0 ), the 3D cubic and 5D quartic Jacobiator vertex (58) survive under blk (together with the 5D quintic Jacobsthal vertex under bal), while for odd dimensions n = 2 r − 1 ≥ 7 n=2r-1\ge7 n = 2 r − 1 ≥ 7 all block-grouped A-CS theories vanish identically (and binary-balanced A-CS theories vanish outside n ∈ ⋃ m ≥ 4 { J m , J m + 2 } ∩ [ 7 , ∞ ) = { 7 } ∪ { 11 , 13 } ∪ ⋯ n\in\bigcup_{m\ge4}\{J_{m},J_{m}+2\}\cap[7,\infty)=\{7\}\cup\{11,13\}\cup\cdots n ∈ ⋃ m ≥ 4 { J m , J m + 2 } ∩ [ 7 , ∞ ) = { 7 } ∪ { 11 , 13 } ∪ ⋯ and for n ≥ dim R u ( k ; CD l ( R ) ) + 3 n\ge\dim_{\mathbb{R}}\mathfrak{u}(k;\operatorname{CD}^{l}(\mathbb{R}))+3 n ≥ dim R u ( k ; CD l ( R )) + 3 ).
Abstract Examples: Lie Algebras, Octonionic Malcev Algebras, Graded Lie Superalgebras, and Nil Algebras
(1) Lie Algebras and Graded Lie Superalgebras under [ ∧ ] g [\wedge]_{\mathfrak{g}} [ ∧ ] g : Let ( g , [ ⋅ , ⋅ ] ) (\mathfrak{g},[\cdot,\cdot]) ( g , [ ⋅ , ⋅ ]) be a Lie algebra. Because the Lie bracket is skew-commutative and satisfies the Jacobi identity, Jac[ ⋅ , ⋅ ] ≡ 0 _{[\cdot,\cdot]} \equiv0 [ ⋅ , ⋅ ] ≡ 0 (so d D [ ⋅ , ⋅ ] F D = 0 d_{D}^{[\cdot,\cdot]}F_D=0 d D [ ⋅ , ⋅ ] F D = 0 and infinitesimal gauge covariance hold universally by Theorem 2.7 ), while g \mathfrak{g} g is ( 2 , 1 ) (2,1) ( 2 , 1 ) -nil under [ ∧ ] g [\wedge]_{\mathfrak{g}} [ ∧ ] g , meaning F D [ ∧ ] g F D = 0 F_D[\wedge]_{\mathfrak{g}}F_D=0 F D [ ∧ ] g F D = 0 identically. More generally, if g = g 0 ⊕ g 1 \mathfrak{g}=\mathfrak{g}^{0}\oplus\mathfrak{g}^{1} g = g 0 ⊕ g 1 is an m m m -graded Lie superalgebra, its even part g 0 \mathfrak{g}^{0} g 0 is ( 2 , 1 ) (2,1) ( 2 , 1 ) -nil. By Theorem 3.2 and Remark 2.6 , any topological or self-dual multiplicative A-YMT functional built solely from the internal Lie bracket [ ∧ ] g [\wedge]_{\mathfrak{g}} [ ∧ ] g and a linear trace functional tr ξ : E g → M × R \operatorname{tr}_{\xi}:E_{\mathfrak{g}}\to M\times\mathbb{R} tr ξ : E g → M × R vanishes identically, underscoring why classical 4D Chern-Weil theory requires an external symmetric bilinear invariant polynomial κ ∈ inv 2 ( g ) \kappa\in\operatorname{inv}_{2}(\mathfrak{g}) κ ∈ inv 2 ( g ) .
(2) Non-Lie Octonionic and Malcev Gauge Theories: Let A = O A=\mathbb{O} A = O be the octonions (or its 7-dimensional imaginary subspace Im ( O ) ≃ R 7 \operatorname{Im}(\mathbb{O})\simeq\mathbb{R}^{7} Im ( O ) ≃ R 7 ) equipped with [ x , y ] ∗ = x y − y x = 2 x × G 2 y [x,y]_{*}=xy-yx=2\,x\times_{G_{2}}y [ x , y ] ∗ = x y − y x = 2 x × G 2 y , which is a non-Lie Malcev algebra [2 , 32 ] . For any O \mathbb{O} O -valued connection 1-form D = D R e ⋅ 1 + D ⃗ ∈ Ω 1 ( P ; R ⊕ Im ( O ) ) D=D_{\mathrm{Re}}\cdot1+\vec{D}\in\Omega^{1}(P;\mathbb{R}\oplus\operatorname{Im}(\mathbb{O})) D = D Re ⋅ 1 + D ∈ Ω 1 ( P ; R ⊕ Im ( O )) , the real unit part D R e ⋅ 1 D_{\mathrm{Re}}\cdot1 D Re ⋅ 1 lies in the center ([ O , 1 ] ∗ = 0 [\mathbb{O},1]_{*}=0 [ O , 1 ] ∗ = 0 ) and the real inner-product part − ⟨ D ⃗ ∧ D ⃗ ⟩ = 0 -\langle\vec{D}\wedge\vec{D}\rangle=0 − ⟨ D ∧ D ⟩ = 0 vanishes by symmetry of ⟨ ⋅ , ⋅ ⟩ \langle\cdot,\cdot\rangle ⟨ ⋅ , ⋅ ⟩ , so D R e ⋅ 1 D_{\mathrm{Re}}\cdot1 D Re ⋅ 1 decouples from D ∧ ∗ D = 1 2 D ∧ [ ⋅ , ⋅ ] ∗ D = D ⃗ ∧ × G 2 D ⃗ ∈ Ω 2 ( P ; Im ( O ) ) D\wedge_{*}D=\frac{1}{2}D\wedge_{[\cdot,\cdot]_{*}}D=\vec{D}\wedge_{\times_{G_{2}}}\vec{D}\in\Omega^{2}(P;\operatorname{Im}(\mathbb{O})) D ∧ ∗ D = 2 1 D ∧ [ ⋅ , ⋅ ] ∗ D = D ∧ × G 2 D ∈ Ω 2 ( P ; Im ( O )) by Proposition 3.1 (3). Because Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) = 6 Assoc ∗ ( x , y , z ) ≠ 0 \operatorname{Jac}_{[\cdot,\cdot]_{*}}(x,y,z)=6\operatorname{Assoc}_{*}(x,y,z)\ne0 Jac [ ⋅ , ⋅ ] ∗ ( x , y , z ) = 6 Assoc ∗ ( x , y , z ) = 0 on non-associative octonionic triples, the Lie nucleus of Im ( O ) \operatorname{Im}(\mathbb{O}) Im ( O ) is trivial (N L i e ( Im ( O ) ) = { 0 } \mathcal{N}_{\mathrm{Lie}}(\operatorname{Im}(\mathbb{O}))=\{0\} N Lie ( Im ( O )) = { 0 } , so admissible connections on Im ( O ) \operatorname{Im}(\mathbb{O}) Im ( O ) are horizontal by Proposition 2.4 (1), while on full O \mathbb{O} O the Lie nucleus is N L i e ( O ) = R ⋅ 1 \mathcal{N}_{\mathrm{Lie}}(\mathbb{O})=\mathbb{R}\cdot1 N Lie ( O ) = R ⋅ 1 ), and Theorem 2.7 (2)–(3) shows that d D [ ⋅ , ⋅ ] ∗ F D = − 1 4 Assoc ∗ ( D ⃗ , D ⃗ , D ⃗ ) ≠ 0 d_{D}^{[\cdot,\cdot]_{*}}F_D=-\frac{1}{4}\operatorname{Assoc}_{*}(\vec{D},\vec{D},\vec{D})\ne0 d D [ ⋅ , ⋅ ] ∗ F D = − 4 1 Assoc ∗ ( D , D , D ) = 0 whenever D ⃗ \vec{D} D spans a non-associative 3-plane in Im ( O ) \operatorname{Im}(\mathbb{O}) Im ( O ) (for n ≥ 3 n\ge3 n ≥ 3 ), whereas d D [ ⋅ , ⋅ ] ∗ F D = 0 d_{D}^{[\cdot,\cdot]_{*}}F_D=0 d D [ ⋅ , ⋅ ] ∗ F D = 0 holds identically whenever dim M = 2 \dim M=2 dim M = 2 or whenever D ⃗ \vec{D} D takes values in an associative quaternionic subalgebra H ⊂ O \mathbb{H}\subset\mathbb{O} H ⊂ O .
(3) Solvable Lie Algebras and Affine Nil Algebras: For a solvable Lie algebra g \mathfrak{g} g of derived length s s s (g ( s + 1 ) = 0 \mathfrak{g}^{(s+1)}=0 g ( s + 1 ) = 0 ), Proposition 3.1 shows that D [ ∧ ] g D ∈ Ω 2 ( P ; [ g , g ] ) D[\wedge]_{\mathfrak{g}}D\in\Omega^{2}(P;[\mathfrak{g},\mathfrak{g}]) D [ ∧ ] g D ∈ Ω 2 ( P ; [ g , g ]) shifts up the derived filtration and its iterated squares ( D [ ∧ ] g D ) [ m ] ∈ Ω 2 m + 1 ( P ; g ( m + 1 ) ) (D[\wedge]_{\mathfrak{g}}D)^{[m]}\in\Omega^{2m+1}(P;\mathfrak{g}^{(m+1)}) ( D [ ∧ ] g D ) [ m ] ∈ Ω 2 m + 1 ( P ; g ( m + 1 ) ) vanish at step m = s m=s m = s , while confining the quartic interaction sector to E [ g , g ] ⊊ E g E_{[\mathfrak{g},\mathfrak{g}]}\subsetneq E_{\mathfrak{g}} E [ g , g ] ⊊ E g . Finally, for any affine nil algebra A A A of nilpotency degree s s s [33 ] , A A A is ( 1 , s ) (1,s) ( 1 , s ) -nil and ( 2 , s ) (2,s) ( 2 , s ) -nil, so Theorem 3.2 (1) implies that all internal topological Chern-Weil A A A -YMT functionals S t o p , r , P A [ D ] S_{\mathrm{top},r,P}^{A}[D] S top , r , P A [ D ] vanish in dimensions n ≥ 2 ( s + 1 ) n\ge2(s+1) n ≥ 2 ( s + 1 ) .
References
Lemma Group, Brazil
Email address: ximenesyuri@gmx.ie
UTFPR, Curitiba, Brazil
Email address: willianvelasco@utfpr.edu.br