Introduction
The geometrical foundations of gravitation continue to motivate extensions of general relativity in which the metric and the affine connection are treated as independent fields. While the Levi–Civita connection of a pseudo-Riemannian spacetime is torsion-free and metric-compatible, a general affine connection admits both torsion and nonmetricity. This broader geometrical setting provides a natural framework for examining modified gravitational dynamics and the structure of compact objects [1 , 2 , 3 ] . Nonlinear curvature Lagrangians have a long history, dating at least to the work of Buchdahl [4 ] ; more recent metric–affine constructions allow curvature, torsion, and nonmetricity invariants to enter the gravitational action simultaneously [5 , 6 , 7 , 8 ] .
A terminological distinction is essential. Two physically different classes of theories are commonly denoted by f ( R , T ) f(R,T) f ( R , T ) or F ( R , T ) F(R,T) F ( R , T ) . In the curvature–matter theory of Harko et al., T T T denotes the trace T m = g μ ν T μ ν T_{\mathrm{m}}=g^{\mu\nu}T_{\mu\nu} T m = g μν T μν of the matter energy–momentum tensor [9 ] . By contrast, in Myrzakulov F ( R , T ) F(R,T) F ( R , T ) gravity, also referred to as Myrzakulov gravity-I (MG-I), T T T denotes a geometrical torsion scalar and R R R is a curvature scalar [10 , 5 , 11 ] . The two constructions have distinct field content and field equations, despite the similar notation. Throughout this paper, T T T denotes exclusively the torsion scalar constructed from the independent connection; the notation T m T_{\mathrm{m}} T m is reserved for the matter trace. The distinction remains important in extended models where geometric and matter invariants occur together [6 , 7 ] .
Myrzakulov gravity was developed as a framework combining geometric invariants in generalized gravitational actions [10 , 11 ] . The wider family includes models built from curvature, torsion, nonmetricity, and additional boundary or matter-dependent terms. The classification presented in Table 1 of Ref. [11 ] situates MG-I within this broader landscape. In the present work we restrict attention to the curvature–torsion subclass, rather than treating the various Myrzakulov models as dynamically interchangeable. Metric–affine formulations and generalizations have been developed in Refs. [5 , 6 , 7 , 8 ] , and recent overviews provide additional context for the corresponding geometric and phenomenological constructions [12 ] . The metric-affine Myrzakulov gravity theories introduced in Ref. [11 ] are summarized in Table 1 .
Table 1. Classification of metric-affine Myrzakulov gravity theories MG-I–MG-XIX, adapted from the classification in Ref. [11 ] . Here R R R , T T T , Q Q Q , and G G G denote the curvature, torsion, nonmetricity, and Gauss–Bonnet scalars, respectively, while T m T_m T m denotes the trace of the matter energy–momentum tensor. The identical arguments assigned to MG-IX and MG-XVIII reproduce the original classification.
The geometric invariants entering the different Myrzakulov models are not interchangeable: in particular, metric-affine curvature–torsion models need not satisfy the flat-connection condition characteristic of teleparallel geometry. The classification of geometric actions in Ref. [11 ] , together with subsequent metric-affine formulations [5 , 6 , 8 ] , provides the context for our choice of MG-I. The present solutions belong to a Riemann–Cartan sector with Q λ μ ν = 0 Q^{\lambda}{}_{\mu\nu}=0 Q λ μν = 0 and totally antisymmetric torsion, but the connection equations are obtained before either restriction is imposed [2 , 3 ] .
Cosmological aspects of Myrzakulov gravity have been investigated extensively. Early applications and observational constraints were studied in Refs. [13 , 14 ] , while a dynamical-systems treatment was developed in Ref. [15 ] . Subsequent work considered baryon asymmetry, quasi-dilaton massive-gravity settings, transit and exact cosmologies, and extensions involving nonmetricity [16 , 17 , 18 , 19 , 20 , 21 , 22 ] . Other recent studies have examined cosmological evolution in alternative formulations and generalized geometric sectors [23 , 24 ] . These results motivate a complementary investigation of strongly gravitating configurations, for which the independent connection can carry geometric information not captured by the metric alone.
Black-hole solutions in Myrzakulov-type theories have also begun to receive attention. In particular, Momeni and Myrzakulov have reported an unpublished study entitled Black Hole Solutions in Metric-Affine Myrzakulov Gravity [25 ] ; related developments address extensions involving Gauss–Bonnet terms and horizon entropy [26 , 27 ] . These works make it especially important to identify the precise gravitational Lagrangian, affine restrictions, and solution-generating assumptions in any new black-hole construction. The present paper focuses on a polynomial MG-I model and on solutions verified against the field equations obtained from independent metric and affine variations. We do not assume that a spherically reduced point-like Lagrangian automatically captures every connection equation.
Specifically, we consider the vacuum model
F ( R , T ) = α R + β T + γ T 3 , α ≠ 0 , (1) F(R,T)=\alpha R+\beta T+\gamma T^3,\qquad\alpha\ne0, \tag*{(1)} F ( R , T ) = α R + β T + γ T 3 , α = 0 , (1)
where R = R ( g , Γ ) R=R(g,\Gamma) R = R ( g , Γ ) is the Ricci scalar of the independent connection and T = T ( g , Γ ) T=T(g,\Gamma) T = T ( g , Γ ) is the quadratic torsion scalar defined below. The metric and all 64 ordered components of Γ μ ν λ \Gamma^\lambda_{\mu\nu} Γ μν λ are varied independently before specializing to a Riemann–Cartan geometry with vanishing nonmetricity and totally antisymmetric torsion. This order of operations is essential: a symmetry or geometric restriction imposed at the level of the action need not reproduce the unrestricted Euler–Lagrange equations [28 , 29 , 2 ] .
In the axial sector we identify an exceptional nonvanishing-torsion branch satisfying
α + β + 3 γ T ∗ 2 = 0. (2) \alpha+\beta+3\gamma T_*^2=0. \tag*{(2)} α + β + 3 γ T ∗ 2 = 0. (2)
On this branch the metric equation reduces to an Einstein equation,
G ˚ μ ν + Λ ∗ g μ ν = 0 , Λ ∗ : = γ T ∗ 3 α , (3) \mathring{G}_{\mu\nu}+\Lambda_*g_{\mu\nu}=0,\qquad\Lambda_*:=\frac{\gamma T_*^3}{\alpha}, \tag*{(3)} G ˚ μν + Λ ∗ g μν = 0 , Λ ∗ := α γ T ∗ 3 , (3)
where G ˚ μ ν \mathring{G}_{\mu\nu} G ˚ μν is computed from the Levi–Civita connection. With the curvature conventions used here, the static spherical metric belongs locally to the Schwarzschild–(anti-)de Sitter class. The full affine connection nevertheless need not coincide with the Levi–Civita connection: a nonzero axial torsion three-form can remain compatible with the complete vacuum field equations.
Our analysis has several complementary objectives. First, we derive and check the independent metric and affine equations, including the fixed-connection metric derivative of the torsion invariant and the unrestricted connection variation. Second, we classify the local solutions within the metric-compatible axial sector, making explicit that this is not a classification of all metric–affine connections. Third, we construct spherical representatives in horizon-penetrating coordinates and investigate their torsion profiles, local smoothness, and affine curvature invariants. Finally, we analyze the degeneracy of the restricted axial perturbative Hessian, while distinguishing this calculation from a full dynamical stability analysis under unrestricted metric–affine perturbations. These distinctions are particularly important because finiteness of scalar invariants does not by itself establish global regularity or stability [30 , 31 , 32 ] .
The paper is organized as follows. § 2 fixes the geometric conventions, while § 3 presents the metric–affine field equations. The role and limitations of spherical action reductions are discussed next. The Riemann–Cartan and axial equations are then developed, followed by the spherical reduction. We subsequently present the local classification of axial metric-compatible connections, construct explicit black-hole solutions and examine their affine invariants and horizon behavior, and discuss the restricted perturbative degeneracy and the open problem of general metric–affine stability. The final section summarizes the results and their limitations.
Geometric framework and conventions
We work on a four-dimensional Lorentzian manifold with signature ( + − − − ) (+---) ( + − − − ) , a metric g μ ν g_{\mu\nu} g μν and an independent affine connection Γ μ ν λ \Gamma^\lambda_{\mu\nu} Γ μν λ . Our index convention is
∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ . (4) \nabla_\nu V^\lambda=\partial_\nu V^\lambda+\Gamma^\lambda_{\rho\nu}V^\rho. \tag*{(4)} ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ . (4)
The torsion, its half-normalized version, and the nonmetricity are defined by
X μ ν λ : = 2 Γ [ μ ν ] λ , S μ ν λ : = 1 2 X μ ν λ , Q λ μ ν : = − ∇ λ g μ ν . (5) X^\lambda_{\mu\nu}:=2\Gamma^\lambda_{[\mu\nu]},\qquad S^\lambda_{\mu\nu}:=\frac{1}{2}X^\lambda_{\mu\nu},\qquad Q_{\lambda\mu\nu}:=-\nabla_\lambda g_{\mu\nu}. \tag*{(5)} X μν λ := 2 Γ [ μν ] λ , S μν λ := 2 1 X μν λ , Q λ μν := − ∇ λ g μν . (5)
We use X X X for the torsion tensor in formulas in which the scalar torsion invariant T T T also appears. The curvature convention and its Ricci contraction are
R ρ μ ν ( Γ ) = 2 ∂ [ μ Γ ∣ ρ ∣ ν ] ρ + 2 Γ σ [ μ ρ Γ ∣ ρ ∣ ν ] σ , R μ ν ( Γ ) : = R λ μ λ ν ( Γ ) , R ( Γ ) : = g μ ν R μ ν ( Γ ) . (6) R^{\rho}{}_{\mu\nu}(\Gamma)=2\partial_{[\mu}\Gamma^{\rho}_{\ \lvert\rho\lvert\nu]}+2\Gamma^{\rho}_{\ \sigma[\mu}\Gamma^{\sigma}_{\ \lvert\rho\lvert\nu]},\qquad R_{\mu\nu}(\Gamma):=R^{\lambda}{}_{\mu\lambda\nu}(\Gamma),\qquad R(\Gamma):=g^{\mu\nu}R_{\mu\nu}(\Gamma). \tag*{(6)} R ρ μν ( Γ ) = 2 ∂ [ μ Γ ∣ ρ ∣ ν ] ρ + 2 Γ σ [ μ ρ Γ ∣ ρ ∣ ν ] σ , R μν ( Γ ) := R λ μ λ ν ( Γ ) , R ( Γ ) := g μν R μν ( Γ ) . (6)
The antisymmetric Ricci tensor need not vanish for a general affine connection. Our curvature and torsion conventions should be read together with the explicit coordinate checks in Appendix B and the variation conventions in Appendix C. General treatments of metric–affine geometry and torsion may be found in Refs. [2 , 3 , 5 ] .
Explicit check of curvature and Ricci conventions. To fix the index ordering unambiguously, our convention ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ \nabla_{\nu}V^{\lambda}=\partial_{\nu}V^{\lambda}+\Gamma^{\lambda}_{\ \rho\nu}V^{\rho} ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ implies
[ ∇ μ , ∇ ν ] V λ = R λ ρ μ ν ( Γ ) V ρ , (7) [\nabla_{\mu},\nabla_{\nu}]V^{\lambda}=R^{\lambda}{}_{\rho\mu\nu}(\Gamma)V^{\rho}, \tag*{(7)} [ ∇ μ , ∇ ν ] V λ = R λ ρ μν ( Γ ) V ρ , (7)
R λ ρ μ ν ( Γ ) = ∂ μ Γ ρ ν λ − ∂ ν Γ ρ μ λ + Γ σ μ λ Γ ρ ν σ − Γ σ ν λ Γ ρ μ σ , (8) R^{\lambda}{}_{\rho\mu\nu}(\Gamma)=\partial_{\mu}\Gamma^{\lambda}_{\ \rho\nu}-\partial_{\nu}\Gamma^{\lambda}_{\ \rho\mu}+\Gamma^{\lambda}_{\ \sigma\mu}\Gamma^{\sigma}_{\ \rho\nu}-\Gamma^{\lambda}_{\ \sigma\nu}\Gamma^{\sigma}_{\ \rho\mu}, \tag*{(8)} R λ ρ μν ( Γ ) = ∂ μ Γ ρ ν λ − ∂ ν Γ ρ μ λ + Γ σ μ λ Γ ρ ν σ − Γ σ ν λ Γ ρ μ σ , (8)
R ρ ν ( Γ ) : = R λ ρ λ ν ( Γ ) , R ( Γ ) = g ρ ν R ρ ν ( Γ ) . (9) R_{\rho\nu}(\Gamma):=R^{\lambda}{}_{\rho\lambda\nu}(\Gamma),\qquad R(\Gamma)=g^{\rho\nu}R_{\rho\nu}(\Gamma). \tag*{(9)} R ρ ν ( Γ ) := R λ ρ λ ν ( Γ ) , R ( Γ ) = g ρ ν R ρ ν ( Γ ) . (9)
The curvature convention in Eq. (6) fixes both the order of the connection indices and the Ricci contraction. In a coordinate basis, the operator commutator [ ∇ μ , ∇ ν ] [\nabla_{\mu},\nabla_{\nu}] [ ∇ μ , ∇ ν ] acting on a vector field yields the curvature endomorphism, whereas the antisymmetrized tensorial second covariant derivative includes the torsion term displayed in Eq. (10) . This commutator identity uses coordinate-direction covariant derivatives acting on vector components. The Ricci contraction in Eq. (6) is on the first and third indices of the four-index curvature, not on its first and fourth indices. In general R μ ν ( Γ ) R_{\mu\nu}(\Gamma) R μν ( Γ ) need not be symmetric in metric–affine geometry [1 , 2 , 3 ] .
The commutator in Eq. (7) is understood as the operator commutator of covariant derivatives along coordinate vector fields. It should be distinguished from the antisymmetrized second covariant derivative of a vector field, for which a torsion-dependent term appears:
∇ μ ∇ ν V λ − ∇ ν ∇ μ V λ = R λ ρ μ ν V ρ − X ρ μ ν ∇ ρ V λ . (10) \nabla_{\mu}\nabla_{\nu}V^{\lambda}-\nabla_{\nu}\nabla_{\mu}V^{\lambda}=R^{\lambda}{}_{\rho\mu\nu}V^{\rho}-X^{\rho}{}_{\mu\nu}\nabla_{\rho}V^{\lambda}. \tag*{(10)} ∇ μ ∇ ν V λ − ∇ ν ∇ μ V λ = R λ ρ μν V ρ − X ρ μν ∇ ρ V λ . (10)
Here the left-hand side denotes the tensorial second covariant derivatives, including the connection acting on the lower derivative index.
As a direct coordinate test, consider the torsion-free advanced Eddington–Finkelstein metric
d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 ( d θ 2 + sin 2 θ d ϕ 2 ) . (11) ds^{2}=f(r)\,dv^{2}-2\,dv\,dr-r^{2}(d\theta^{2}+\sin^{2}\theta\,d\phi^{2}). \tag*{(11)} d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 ( d θ 2 + sin 2 θ d ϕ 2 ) . (11)
Computing its Levi–Civita curvature with (6) and contracting according to (9) gives
R ˚ v v = f 2 ( f ′ ′ + 2 f ′ r ) , R ˚ v r = R ˚ r v = − 1 2 ( f ′ ′ + 2 f ′ r ) , (12) \mathring{R}_{vv}=\frac{f}{2}\left(f^{\prime\prime}+\frac{2f^{\prime}}{r}\right),\qquad\mathring{R}_{vr}=\mathring{R}_{rv}=-\frac{1}{2}\left(f^{\prime\prime}+\frac{2f^{\prime}}{r}\right), \tag*{(12)} R ˚ v v = 2 f ( f ′′ + r 2 f ′ ) , R ˚ v r = R ˚ r v = − 2 1 ( f ′′ + r 2 f ′ ) , (12)
R ˚ r r = 0 , R ˚ θ θ = 1 − f − r f ′ , R ˚ ϕ ϕ = sin 2 θ R ˚ θ θ . (13) \mathring{R}_{rr}=0,\qquad\mathring{R}_{\theta\theta}=1-f-rf^{\prime},\qquad\mathring{R}_{\phi\phi}=\sin^{2}\theta\,\mathring{R}_{\theta\theta}. \tag*{(13)} R ˚ r r = 0 , R ˚ θ θ = 1 − f − r f ′ , R ˚ ϕϕ = sin 2 θ R ˚ θ θ . (13)
Since g v r = − 1 g^{vr}=-1 g v r = − 1 , g r r = − f g^{rr}=-f g r r = − f and g θ θ = − r − 2 g^{\theta\theta}=-r^{-2} g θ θ = − r − 2 , the scalar contraction is
R ˚ = − 2 R ˚ v r − 2 R ˚ θ θ r 2 = f ′ ′ + 4 f ′ r + 2 ( f − 1 ) r 2 . (14) \mathring{R}=-2\mathring{R}_{vr}-\frac{2\mathring{R}_{\theta\theta}}{r^{2}}=f^{\prime\prime}+\frac{4f^{\prime}}{r}+\frac{2(f-1)}{r^{2}}. \tag*{(14)} R ˚ = − 2 R ˚ v r − r 2 2 R ˚ θ θ = f ′′ + r 4 f ′ + r 2 2 ( f − 1 ) . (14)
In particular, for f ( r ) = 1 − 2 M / r + Λ ∗ r 2 / 3 f(r)=1-2M/r+\Lambda_{*}r^{2}/3 f ( r ) = 1 − 2 M / r + Λ ∗ r 2 /3 one obtains
R ˚ μ ν = Λ ∗ g μ ν , R ˚ = 4 Λ ∗ , (15) \mathring{R}_{\mu\nu}=\Lambda_{*}g_{\mu\nu},\qquad\mathring{R}=4\Lambda_{*}, \tag*{(15)} R ˚ μν = Λ ∗ g μν , R ˚ = 4 Λ ∗ , (15)
which fixes the sign used in the spherical field equations below. If instead the conventional black-hole parameter is defined by f = 1 − 2 M / r − Λ p h y s r 2 / 3 f=1-2M/r-\Lambda_{\mathrm{phys}}r^{2}/3 f = 1 − 2 M / r − Λ phys r 2 /3 , then Λ p h y s = − Λ ∗ \Lambda_{\mathrm{phys}}=-\Lambda_{*} Λ phys = − Λ ∗ in the present conventions [30 , 31 , 32 ] .
Distortion and the principal geometric sectors
The difference between the independent connection and the Levi–Civita connection Γ ˚ μ ν λ \mathring{\Gamma}^{\lambda}_{\mu\nu} Γ ˚ μν λ is the distortion tensor,
Γ μ ν λ = Γ ˚ μ ν λ + N μ ν λ . (16) \Gamma^{\lambda}_{\mu\nu}=\mathring{\Gamma}^{\lambda}_{\mu\nu}+N^{\lambda}_{\mu\nu}. \tag*{(16)} Γ μν λ = Γ ˚ μν λ + N μν λ . (16)
Writing ∇ ˚ \mathring{\nabla} ∇ ˚ for the Levi–Civita derivative gives the exact curvature decomposition
R ρ μ ν λ ( Γ ) = R ˚ ρ μ ν λ + 2 ∇ ˚ [ μ N ∣ ρ ∣ ν ] λ + 2 N [ μ σ N ∣ ρ ∣ ν ] λ . (17) R^{\lambda}_{\ \rho\mu\nu}(\Gamma)=\mathring{R}^{\lambda}_{\ \rho\mu\nu}+2\mathring{\nabla}_{[\mu}N^{\lambda}_{\ |\rho|\nu]}+2N^{\sigma}_{\ [\mu}N^{\lambda}_{\ |\rho|\nu]}. \tag*{(17)} R ρ μν λ ( Γ ) = R ˚ ρ μν λ + 2 ∇ ˚ [ μ N ∣ ρ ∣ ν ] λ + 2 N [ μ σ N ∣ ρ ∣ ν ] λ . (17)
Consequently,
R ( Γ ) = R ˚ + u ( g , N ) , u ( g , N ) : = g ρ ν ( 2 ∇ ˚ [ λ N ∣ ρ ∣ ν ] λ + 2 N [ λ σ N ∣ ρ ∣ ν ] λ ) . (18) R(\Gamma)=\mathring{R}+u(g,N),\qquad u(g,N):=g^{\rho\nu}\left(2\mathring{\nabla}_{[\lambda}N^{\lambda}_{\ |\rho|\nu]}+2N^{\sigma}_{\ [\lambda}N^{\lambda}_{\ |\rho|\nu]}\right). \tag*{(18)} R ( Γ ) = R ˚ + u ( g , N ) , u ( g , N ) := g ρ ν ( 2 ∇ ˚ [ λ N ∣ ρ ∣ ν ] λ + 2 N [ λ σ N ∣ ρ ∣ ν ] λ ) . (18)
Thus u u u is a derived geometric expression, not an independently prescribed radial function. In the Riemann–Cartan sector Q λ μ ν = 0 Q_{\lambda\mu\nu}=0 Q λ μν = 0 , the distortion becomes the contorsion K μ ν λ K^{\lambda}_{\ \mu\nu} K μν λ . No condition of vanishing affine curvature is imposed. This distinguishes the present setting from teleparallel geometry, for which R ρ μ ν λ ( Γ ) = 0 R^{\lambda}_{\ \rho\mu\nu}(\Gamma)=0 R ρ μν λ ( Γ ) = 0 ; symmetric teleparallel geometry instead has R ρ μ ν λ = 0 R^{\lambda}_{\ \rho\mu\nu}=0 R ρ μν λ = 0 and X μ ν λ = 0 X^{\lambda}_{\ \mu\nu}=0 X μν λ = 0 [1 , 2 , 3 ] .
The quadratic torsion invariant entering the MG-I action is fixed as
T : = 1 4 X λ μ ν X λ μ ν + 1 2 X λ μ ν X ν μ λ − t μ t μ , t μ : = X λ μ λ . (19) T:=\frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu}+\frac{1}{2}X_{\lambda\mu\nu}X^{\nu\mu\lambda}-t_{\mu}t^{\mu},\qquad t_{\mu}:=X^{\lambda}_{\ \lambda\mu}. \tag*{(19)} T := 4 1 X λ μν X λ μν + 2 1 X λ μν X ν μ λ − t μ t μ , t μ := X λ μ λ . (19)
This definition refers to the independent affine connection, not to an arbitrary tetrad-dependent torsion scalar. If a reference connection Γ 0 \Gamma_{0} Γ 0 is chosen, T = T 0 + v T=T_{0}+v T = T 0 + v with v : = T ( g , Γ ) − T ( g , Γ 0 ) v:=T(g,\Gamma)-T(g,\Gamma_{0}) v := T ( g , Γ ) − T ( g , Γ 0 ) is merely an identity. In particular, the functions u u u and v v v cannot be freely varied as substitutes for the components of Γ \Gamma Γ [10 , 5 ] .
Axial torsion and the spherical metric
Our black-hole construction subsequently restricts the solutions of the unrestricted field equations to the metric-compatible, totally antisymmetric sector,
Q λ μ ν = 0 , X λ μ ν = X [ λ μ ν ] . (20) Q_{\lambda\mu\nu}=0,\qquad X_{\lambda\mu\nu}=X_{[\lambda\mu\nu]}. \tag*{(20)} Q λ μν = 0 , X λ μν = X [ λ μν ] . (20)
In this sector t μ = 0 t_{\mu}=0 t μ = 0 and the convention (19) gives
T = − 1 4 X λ μ ν X λ μ ν , K λ μ ν = 1 2 X λ μ ν . (21) T=-\frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu},\qquad K_{\lambda\mu\nu}=\frac{1}{2}X_{\lambda\mu\nu}. \tag*{(21)} T = − 4 1 X λ μν X λ μν , K λ μν = 2 1 X λ μν . (21)
The axial curvature identities used throughout the axial analysis below, subject to the independent checks discussed below, are
R ( μ ν ) ( Γ ) = R ˚ μ ν + H μ ν , R ( Γ ) = R ˚ + T , (22) R_{(\mu\nu)}(\Gamma)=\mathring{R}_{\mu\nu}+H_{\mu\nu},\qquad R(\Gamma)=\mathring{R}+T, \tag*{(22)} R ( μν ) ( Γ ) = R ˚ μν + H μν , R ( Γ ) = R ˚ + T , (22)
where H μ ν H_{\mu\nu} H μν is the fixed-connection metric derivative of T T T , as specified in § 3 . The component-level geometric calculation is collected in Appendix § 14 ; the independent variation and its scope are discussed in Appendix § 15 . We do not assume these axial identities for arbitrary torsion or nonmetricity.
Before choosing a radial gauge, the static spherical line element is
d s 2 = A ( r ) d t 2 − B ( r ) d r 2 − C ( r ) ( d θ 2 + sin 2 θ d ϕ 2 ) , (23) ds^{2}=A(r)\,dt^{2}-B(r)\,dr^{2}-C(r)\left(d\theta^{2}+\sin^{2}\theta\,d\phi^{2}\right), \tag*{(23)} d s 2 = A ( r ) d t 2 − B ( r ) d r 2 − C ( r ) ( d θ 2 + sin 2 θ d ϕ 2 ) , (23)
with A , B , C > 0 A,B,C>0 A , B , C > 0 in a static region. The areal-radius choice C = r 2 C=r^{2} C = r 2 and the conformal radial gauge B = A B=A B = A are different coordinate conditions; they should not be imposed simultaneously on a generic metric. In the Einstein-type branch derived in § 6 –§ 7 we shall use the justified Schwarzschild-type representative A = f A=f A = f , B = f − 1 B=f^{-1} B = f − 1 , C = r 2 C=r^{2} C = r 2 . With the curvature convention used for the spherical expressions of the manuscript, this representative satisfies
R ˚ = f ′ ′ + 4 f ′ r + 2 ( f − 1 ) r 2 . (24) \mathring{R}=f^{\prime\prime}+\frac{4f^{\prime}}{r}+\frac{2(f-1)}{r^{2}}. \tag*{(24)} R ˚ = f ′′ + r 4 f ′ + r 2 2 ( f − 1 ) . (24)
Metric–affine Myrzakulov F ( R , T ) F(R,T) F ( R , T ) gravity
Myrzakulov gravity-I combines the affine Ricci scalar and a torsion scalar in a single Lagrangian [10 , 13 , 5 ] . We adopt the action
S [ g , Γ , φ ] = 1 2 κ ∫ d 4 x − g F ( R ( g , Γ ) , T ( g , Γ ) ) + S m [ g , Γ , φ ] , κ = 8 π G . (25) S[g,\Gamma,\varphi] = \frac{1}{2\kappa}\int d^4x\sqrt{-g}\,F\bigl(R(g,\Gamma),T(g,\Gamma)\bigr)+S_{\mathrm{m}}[g,\Gamma,\varphi],\qquad\kappa=8\pi G. \tag*{(25)} S [ g , Γ , φ ] = 2 κ 1 ∫ d 4 x − g F ( R ( g , Γ ) , T ( g , Γ ) ) + S m [ g , Γ , φ ] , κ = 8 π G . (25)
The independent gravitational variables are the ten symmetric metric components g μ ν g^{\mu\nu} g μν and the 64 connection components Γ μ ν λ \Gamma^\lambda_{\mu\nu} Γ μν λ . The latter are not assumed symmetric or metric-compatible during variation. We write F R : = ∂ F / ∂ R F_R:=\partial F/\partial R F R := ∂ F / ∂ R and F T : = ∂ F / ∂ T F_T:=\partial F/\partial T F T := ∂ F / ∂ T . Here T T T is the torsion scalar, not the matter stress-tensor trace.
Independent metric variation
The metric is varied with Γ \Gamma Γ held fixed. Defining
H μ ν : = ∂ T ∂ g μ ν ∣ Γ , T μ ν : = − 2 − g δ S m δ g μ ν , (26) H_{\mu\nu}:=\left.\frac{\partial T}{\partial g^{\mu\nu}}\right|_{\Gamma},\qquad\mathcal{T}_{\mu\nu}:=-\frac{2}{\sqrt{-g}}\frac{\delta S_{\mathrm{m}}}{\delta g^{\mu\nu}}, \tag*{(26)} H μν := ∂ g μν ∂ T Γ , T μν := − − g 2 δ g μν δ S m , (26)
we obtain the metric equation
F R R ( μ ν ) ( Γ ) − 1 2 g μ ν F + F T H μ ν = κ T μ ν . (27) F_R R_{(\mu\nu)}(\Gamma)-\frac{1}{2}g_{\mu\nu}F+F_T H_{\mu\nu}=\kappa\mathcal{T}_{\mu\nu}. \tag*{(27)} F R R ( μν ) ( Γ ) − 2 1 g μν F + F T H μν = κ T μν . (27)
Its trace is
F R R − 2 F + F T H = κ T , H : = g μ ν H μ ν . (28) F_RR-2F+F_TH=\kappa\mathcal{T},\qquad H:=g^{\mu\nu}H_{\mu\nu}. \tag*{(28)} F R R − 2 F + F T H = κ T , H := g μν H μν . (28)
The definition (26) is fundamental: the torsion tensor with a lowered first index is not held fixed during this variation. For the axial sector, the identity to be checked with these independent variables is
H μ ν ∣ a x i a l = − 1 4 X μ ρ σ X ν ρ σ , H ∣ a x i a l = T . (29) H_{\mu\nu}\big|_{\mathrm{axial}}=-\frac{1}{4}X_{\mu\rho\sigma}X_{\nu}{}^{\rho\sigma},\qquad H\big|_{\mathrm{axial}}=T. \tag*{(29)} H μν axial = − 4 1 X μ ρ σ X ν ρ σ , H axial = T . (29)
The variational calculation and the scope of the symbolic component tests are discussed in Appendix § 15 ; the fixed-connection derivative must be independently checked against the unrestricted scalar definition. In particular, a test of a restricted axial Lagrangian is not a replacement for the unrestricted metric variation [2 , 3 ] .
Independent affine variation
In vacuum, variation of all connection components produces an equation which we denote schematically by
P μ ν ( F R ) + F T J μ ν = 0 , (30) \mathcal{P}^{\mu\nu}(F_R)+F_T\mathcal{J}^{\mu\nu}=0, \tag*{(30)} P μν ( F R ) + F T J μν = 0 , (30)
where P \mathcal{P} P is the Palatini operator arising from the integrated-by-parts variation of − g R \sqrt{-g}R − g R , and J \mathcal{J} J denotes the derivative of the torsion invariant with respect to the independent connection. In the notation of the original field equations, J = 2 K \mathcal{J}=2\mathcal{K} J = 2 K , so that P + 2 F T K = 0 \mathcal{P}+2F_T\mathcal{K}=0 P + 2 F T K = 0 . The explicit index conventions and the comparison of the two normalizations are recorded in Appendix § 15 . This separation makes clear that varying only the algebraic connection-squared part of R R R does not reproduce the full Palatini operator. The affine equation is essential even when the final solution has Q λ μ ν = 0 Q_{\lambda\mu\nu}=0 Q λ μν = 0 [2 , 5 ] .
For definiteness, the torsion coefficient is defined by the fixed-metric variation of the invariant itself,
δ Γ T = J λ μ ν δ Γ λ μ ν , J λ μ ν : = ∂ T ∂ Γ λ μ ν ∣ g . (31) \delta_{\Gamma}T=\mathcal{J}_{\lambda}{}^{\mu\nu}\delta\Gamma^{\lambda}{}_{\mu\nu},\qquad\mathcal{J}_{\lambda}{}^{\mu\nu}:=\left.\frac{\partial T}{\partial\Gamma^{\lambda}{}_{\mu\nu}}\right|_{g}. \tag*{(31)} δ Γ T = J λ μν δ Γ λ μν , J λ μν := ∂ Γ λ μν ∂ T g . (31)
Since T T T depends on Γ \Gamma Γ only through its antisymmetric torsion, J λ ( μ ν ) = 0 \mathcal{J}_{\lambda}{}^{(\mu\nu)}=0 J λ ( μν ) = 0 identically. The explicit unrestricted component expression and its axial specialization are given in Appendix § 15 .§ 4 ; in particular, no metric-compatibility or axial restriction enters the definition (31) .
Polynomial model and axial consistency
The remainder of the paper focuses on
F ( R , T ) = α R + β T + γ T 3 , F R = α , F T = Ψ ( T ) : = β + 3 γ T 2 . (32) F(R,T)=\alpha R+\beta T+\gamma T^{3},\qquad F_{R}=\alpha,\qquad F_{T}=\Psi(T):=\beta+3\gamma T^{2}. \tag*{(32)} F ( R , T ) = α R + β T + γ T 3 , F R = α , F T = Ψ ( T ) := β + 3 γ T 2 . (32)
For this model the vacuum equations become
0 = α R ( μ ν ) − 1 2 g μ ν ( α R + β T + γ T 3 ) + Ψ ( T ) H μ ν , (33) 0=\alpha R_{(\mu\nu)}-\frac{1}{2}g_{\mu\nu}(\alpha R+\beta T+\gamma T^{3})+\Psi(T)H_{\mu\nu}, \tag*{(33)} 0 = α R ( μν ) − 2 1 g μν ( α R + β T + γ T 3 ) + Ψ ( T ) H μν , (33)
0 = P λ μ ν ( α ) + Ψ ( T ) J λ μ ν . (34) 0=\mathcal{P}_{\lambda}^{\mu\nu}(\alpha)+\Psi(T)\mathcal{J}_{\lambda}^{\mu\nu}. \tag*{(34)} 0 = P λ μν ( α ) + Ψ ( T ) J λ μν . (34)
The axial specialization of the unrestricted affine equation, provided the unrestricted variational identities in Appendix § 15 are validated, is
[ α + Ψ ( T ) ] X λ μ ν = 0. (35) [\alpha+\Psi(T)]X_{\lambda}^{\mu\nu}=0. \tag*{(35)} [ α + Ψ ( T )] X λ μν = 0. (35)
Thus nonvanishing axial torsion requires α + β + 3 γ T 2 = 0 \alpha+\beta+3\gamma T^{2}=0 α + β + 3 γ T 2 = 0 . For γ ≠ 0 \gamma\ne0 γ = 0 a continuous real branch has T = T ∗ T=T_{*} T = T ∗ with T ∗ 2 = − ( α + β ) / ( 3 γ ) ≥ 0 T_{*}^{2}=-(\alpha+\beta)/(3\gamma)\ge0 T ∗ 2 = − ( α + β ) / ( 3 γ ) ≥ 0 ; for γ = 0 \gamma=0 γ = 0 the exceptional coupling is β = − α \beta=-\alpha β = − α . The detailed algebra, including the null-norm case, belongs to § 6 –§ 7 rather than being repeated here.
Static spherical reduction with an independent affine connection
A static spherical ansatz is useful for organizing the field equations of Myrzakulov gravity-I, but the independent affine degrees of freedom must not be eliminated before variation. In particular, the point-like formulations employed in cosmological studies of F ( R , T ) F(R,T) F ( R , T ) gravity [10 , 13 , 14 , 5 ] do not, without further hypotheses, replace the unrestricted metric–affine equations. The latter distinction is especially important when torsion and nonmetricity are independent geometric fields [2 , 5 , 6 , 8 ] .
Metric, distortion, and radial action
In a static region consider the general spherically symmetric metric
d s 2 = A ( r ) d t 2 − B ( r ) d r 2 − C ( r ) d Ω 2 , d Ω 2 = d θ 2 + sin 2 θ d ϕ 2 , (36) ds^{2}=A(r)\,dt^{2}-B(r)\,dr^{2}-C(r)\,d\Omega^{2},\qquad d\Omega^{2}=d\theta^{2}+\sin^{2}\theta\,d\phi^{2}, \tag*{(36)} d s 2 = A ( r ) d t 2 − B ( r ) d r 2 − C ( r ) d Ω 2 , d Ω 2 = d θ 2 + sin 2 θ d ϕ 2 , (36)
where A , B , C > 0 A,B,C>0 A , B , C > 0 . Before fixing the radial coordinate, the independent connection is represented schematically as
Γ μ ν λ = Γ ˚ μ ν λ [ g ] + N μ ν λ ( r , θ ; ξ I ( r ) ) , (37) \Gamma^{\lambda}_{\mu\nu}=\mathring{\Gamma}^{\lambda}_{\mu\nu}[g]+N^{\lambda}_{\mu\nu}(r,\theta;\xi^{I}(r)), \tag*{(37)} Γ μν λ = Γ ˚ μν λ [ g ] + N μν λ ( r , θ ; ξ I ( r )) , (37)
where the angular dependence is constrained by rotational invariance and ξ I \xi^{I} ξ I denotes the remaining independent radial functions. Equation (37) is a bookkeeping notation, not a claim that a complete classification of invariant affine connections has been performed. In particular, the number and transformation properties of the ξ I \xi^{I} ξ I must be established before treating the reduced system as closed [2 , 29 ] .
With N = Γ − Γ ˚ N=\Gamma-\mathring{\Gamma} N = Γ − Γ ˚ , the exact curvature identity is
R λ ρ μ ν ( Γ ) = R ˚ λ ρ μ ν + 2 ∇ ˚ [ μ N λ ∣ ρ ∣ ν ] + 2 N λ σ [ μ N σ ∣ ρ ∣ ν ] , (38) R^{\lambda}{}_{\rho\mu\nu}(\Gamma)=\mathring{R}^{\lambda}{}_{\rho\mu\nu}+2\mathring{\nabla}_{[\mu}N^{\lambda}{}_{|\rho|\nu]}+2N^{\lambda}{}_{\sigma[\mu}N^{\sigma}{}_{|\rho|\nu]}, \tag*{(38)} R λ ρ μν ( Γ ) = R ˚ λ ρ μν + 2 ∇ ˚ [ μ N λ ∣ ρ ∣ ν ] + 2 N λ σ [ μ N σ ∣ ρ ∣ ν ] , (38)
R ( Γ ) = R ˚ + u ( g , N ) , R(\Gamma)=\mathring{R}+u(g,N), R ( Γ ) = R ˚ + u ( g , N ) ,
u ( g , N ) = g μ ν ( 2 ∇ ˚ [ λ N λ ∣ ρ ∣ ν ] + 2 N λ σ [ λ N σ ∣ ρ ∣ ν ] ) . (39) u(g,N)=g^{\mu\nu}\left(2\mathring{\nabla}_{[\lambda}N^{\lambda}{}_{|\rho|\nu]}+2N^{\lambda}{}_{\sigma[\lambda}N^{\sigma}{}_{|\rho|\nu]}\right). \tag*{(39)} u ( g , N ) = g μν ( 2 ∇ ˚ [ λ N λ ∣ ρ ∣ ν ] + 2 N λ σ [ λ N σ ∣ ρ ∣ ν ] ) . (39)
Thus u u u is determined by the distortion and is not an independent potential. The torsion scalar T ( g , Γ ) T(g,\Gamma) T ( g , Γ ) is likewise the invariant defined in § 2 , not an arbitrary radial function.
For the vacuum action S = ( 2 κ ) − 1 ∫ d 4 x − g F ( R , T ) S=(2\kappa)^{-1}\int d^{4}x\sqrt{-g}F(R,T) S = ( 2 κ ) − 1 ∫ d 4 x − g F ( R , T ) , integration over the two-sphere gives, per unit coordinate time,
S v a c = ∫ d t d r L r a d , L r a d = 2 π κ C A B F ( R ˙ + u ( g , N ) , T ( g , Γ ) ) . (40) S_{\mathrm{vac}}=\int dt\,dr\,L_{\mathrm{rad}},\qquad L_{\mathrm{rad}}=\frac{2\pi}{\kappa}C\sqrt{AB}\,F\left(\dot{R}+u(g,N),T(g,\Gamma)\right). \tag*{(40)} S vac = ∫ d t d r L rad , L rad = κ 2 π C A B F ( R ˙ + u ( g , N ) , T ( g , Γ ) ) . (40)
For F = α R + β T + γ T 3 F=\alpha R+\beta T+\gamma T^{3} F = α R + β T + γ T 3 , this becomes
L r a d = 2 π κ C A B [ α ( R ˙ + u ( g , N ) ) + β T ( g , Γ ) + γ T ( g , Γ ) 3 ] . (41) L_{\mathrm{rad}}=\frac{2\pi}{\kappa}C\sqrt{AB}\left[\alpha\left(\dot{R}+u(g,N)\right)+\beta T(g,\Gamma)+\gamma T(g,\Gamma)^{3}\right]. \tag*{(41)} L rad = κ 2 π C A B [ α ( R ˙ + u ( g , N ) ) + β T ( g , Γ ) + γ T ( g , Γ ) 3 ] . (41)
Boundary terms generated by integrations by parts must be retained or specified consistently. Neither u u u nor T T T may be varied as a freely prescribed function in place of the underlying affine variables.
Auxiliary scalars and the scope of symmetry reduction
To compare with point-like formulations of MG-I [10 , 13 , 14 , 15 ] , introduce auxiliary scalars R , T \mathcal{R},\mathcal{T} R , T and multipliers λ R , λ T \lambda_{\mathcal{R}},\lambda_{\mathcal{T}} λ R , λ T :
S a u x = 1 2 κ ∫ d 4 x − g [ F ( R , T ) − λ R ( R − R ( g , Γ ) ) − λ T ( T − T ( g , Γ ) ) ] . (42) S_{\mathrm{aux}}=\frac{1}{2\kappa}\int d^{4}x\sqrt{-g}\left[F(\mathcal{R},\mathcal{T})-\lambda_{\mathcal{R}}\left(\mathcal{R}-R(g,\Gamma)\right)-\lambda_{\mathcal{T}}\left(\mathcal{T}-T(g,\Gamma)\right)\right]. \tag*{(42)} S aux = 2 κ 1 ∫ d 4 x − g [ F ( R , T ) − λ R ( R − R ( g , Γ ) ) − λ T ( T − T ( g , Γ ) ) ] . (42)
The multiplier equations enforce R = R ( g , Γ ) \mathcal{R}=R(g,\Gamma) R = R ( g , Γ ) and T = T ( g , Γ ) \mathcal{T}=T(g,\Gamma) T = T ( g , Γ ) , while variation of the auxiliary scalars gives λ R = F R \lambda_{\mathcal{R}}=F_{R} λ R = F R and λ T = F T \lambda_{\mathcal{T}}=F_{T} λ T = F T . This rewriting does not remove the independent affine field equations. Furthermore, when the Legendre map is degenerate, these relations do not necessarily allow one to solve uniquely for both auxiliary scalars; the multiplier formulation remains the appropriate statement.
Symmetry reduction of a variational principle requires care. The principle of symmetric criticality gives conditions under which the symmetric Euler–Lagrange equations follow from a restricted action [28 , 29 ] ; it does not license discarding affine variations or fixing the radial gauge prematurely. In the present work, Eqs. (40) –(42) serve as an organizational description. The black-hole solutions are checked against the full equations derived in § 3 .
Geometric specialization used below
Only after the unrestricted variations do we evaluate the equations in the metric-compatible, totally antisymmetric torsion sector,
Q λ μ ν = 0 , X λ μ ν = X [ λ μ ν ] . (43) Q_{\lambda\mu\nu}=0,\qquad X_{\lambda\mu\nu}=X_{[\lambda\mu\nu]}. \tag*{(43)} Q λ μν = 0 , X λ μν = X [ λ μν ] . (43)
A rotation-invariant axial three-form has two independent radial amplitudes. In the advanced Eddington–Finkelstein chart employed in § 10 , it can be written as
τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ . (44) \tau=r^{2}\sin\theta\left[U(r)\,dv-V(r)\,dr\right]\wedge d\theta\wedge d\phi. \tag*{(44)} τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ . (44)
This ansatz is a specialization of the tensorial torsion field, not a substitute for the unrestricted connection variation. Its algebraic norm constraint and its horizon regularity are treated in § 10 .
Variational consistency and the route to black-hole solutions
The construction rests on two independent equations: the metric equation and the complete affine equation, each obtained before imposing Q = 0 Q=0 Q = 0 or axiality. Their simultaneous satisfaction is necessary for a solution of metric–affine MG-I [5 , 6 , 2 ] . The radial formulation is useful for checking, rather than replacing, these equations.
Radial Euler–Lagrange equations and constraints
For independent radial variables q I ( r ) q^{I}(r) q I ( r ) and a reduced Lagrangian L ( q , q ′ , q ′ ′ ) L(q,q',q'') L ( q , q ′ , q ′′ ) , the variational equations are
∂ L ∂ q I − d d r ∂ L ∂ q ′ I + d 2 d r 2 ∂ L ∂ q ′ ′ I = 0. (45) \frac{\partial L}{\partial q^{I}}-\frac{d}{dr}\frac{\partial L}{\partial q^{\prime I}}+\frac{d^{2}}{dr^{2}}\frac{\partial L}{\partial q^{\prime\prime I}}=0. \tag*{(45)} ∂ q I ∂ L − d r d ∂ q ′ I ∂ L + d r 2 d 2 ∂ q ′′ I ∂ L = 0. (45)
If L L L has no explicit radial dependence, the associated formal Ostrogradsky energy is
E = ∑ I [ q ′ I ( ∂ L ∂ q ′ I − d d r ∂ L ∂ q ′ ′ I ) + q ′ ′ I ∂ L ∂ q ′ ′ I ] − L . (46) \mathcal{E}=\sum_{I}\left[q^{\prime I}\left(\frac{\partial L}{\partial q^{\prime I}}-\frac{d}{dr}\frac{\partial L}{\partial q^{\prime\prime I}}\right)+q^{\prime\prime I}\frac{\partial L}{\partial q^{\prime\prime I}}\right]-L. \tag*{(46)} E = I ∑ [ q ′ I ( ∂ q ′ I ∂ L − d r d ∂ q ′′ I ∂ L ) + q ′′ I ∂ q ′′ I ∂ L ] − L . (46)
Its formal conservation does not imply E = 0 \mathcal{E}=0 E = 0 . Any gravitational constraint must instead be obtained from the appropriate unfixed metric or affine equation. In particular, a radial coordinate choice does not eliminate an independent affine amplitude. The older point-like relations are discussed in Appendix § 13 , with their restricted scope made explicit [10 , 13 , 28 , 29 ] .
Independent consistency checks
The route to the axial black-hole branch requires four logically distinct checks: (i) the metric derivative H μ ν H_{\mu\nu} H μν at fixed Γ \Gamma Γ , calculated before imposing axiality; (ii) the full affine variation with all 64 ordered connection components; (iii) the curvature and Ricci contractions for the reconstructed connection; and (iv) smoothness of the torsion three-form in a horizon-penetrating chart. Appendices B and C supply the tensorial and symbolic checks, while § 10 addresses the horizon analysis. These tests have different logical roles: an invariant curvature calculation cannot replace an affine field equation, and finite scalar invariants alone do not establish smoothness of the connection [2 , 3 , 30 , 32 ] .
For the polynomial model the full vacuum equations are
0 = α R ( μ ν ) ( Γ ) − 1 2 g μ ν ( α R + β T + γ T 3 ) + Ψ ( T ) H μ ν , (47) 0=\alpha R_{(\mu\nu)}(\Gamma)-\frac{1}{2}g_{\mu\nu}\left(\alpha R+\beta T+\gamma T^{3}\right)+\Psi(T)H_{\mu\nu}, \tag*{(47)} 0 = α R ( μν ) ( Γ ) − 2 1 g μν ( α R + β T + γ T 3 ) + Ψ ( T ) H μν , (47)
0 = α P λ μ ν + Ψ ( T ) J λ μ ν , Ψ ( T ) = β + 3 γ T 2 . (48) 0=\alpha P_{\lambda}{}^{\mu\nu}+\Psi(T)J_{\lambda}{}^{\mu\nu},\qquad\Psi(T)=\beta+3\gamma T^{2}. \tag*{(48)} 0 = α P λ μν + Ψ ( T ) J λ μν , Ψ ( T ) = β + 3 γ T 2 . (48)
Here P P P is the integrated-by-parts Palatini coefficient, and J J J is the unrestricted connection derivative of the torsion scalar, with the index and sign conventions of § 2 and Appendix § 15 . § 6 now specializes these equations to Riemann–Cartan geometry; the local existence statement is reserved for § 7 .
Linear and cubic MG-I models in Riemann–Cartan geometry
We evaluate the full metric–affine equations of
F ( R , T ) = α R + β T + γ T 3 , α ≠ 0 , (49) F(R,T)=\alpha R+\beta T+\gamma T^{3},\qquad\alpha\ne0, \tag*{(49)} F ( R , T ) = α R + β T + γ T 3 , α = 0 , (49)
on the metric-compatible sector. The linear model is obtained when γ = 0 \gamma=0 γ = 0 . In contrast with teleparallel geometry, Riemann–Cartan geometry permits both nonzero torsion and nonzero affine curvature [1 , 33 , 34 , 2 , 3 ] .
Metric compatibility and axial identities
For Q λ μ ν = 0 Q_{\lambda\mu\nu}=0 Q λ μν = 0 , the independent connection takes the form
Γ λ μ ν = Γ ˚ λ μ ν + K λ μ ν . (50) \Gamma^{\lambda}{}_{\mu\nu}=\mathring{\Gamma}^{\lambda}{}_{\mu\nu}+K^{\lambda}{}_{\mu\nu}. \tag*{(50)} Γ λ μν = Γ ˚ λ μν + K λ μν . (50)
The curvature decomposition is Eq. (38) with N = K N=K N = K . In the totally antisymmetric sector,
X λ μ ν = X [ λ μ ν ] , t μ = X λ λ μ = 0 , K λ μ ν = 1 2 X λ μ ν . (51) X_{\lambda\mu\nu}=X_{[\lambda\mu\nu]},\qquad t_{\mu}=X^{\lambda}{}_{\lambda\mu}=0,\qquad K^{\lambda}{}_{\mu\nu}=\frac{1}{2}X^{\lambda}{}_{\mu\nu}. \tag*{(51)} X λ μν = X [ λ μν ] , t μ = X λ λ μ = 0 , K λ μν = 2 1 X λ μν . (51)
With the torsion invariant and curvature conventions fixed in § 2 , the following identities hold:
T = − 1 4 X λ μ ν X λ μ ν , (52) T=-\frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu}, \tag*{(52)} T = − 4 1 X λ μν X λ μν , (52)
H μ ν : = ∂ T ∂ g μ ν ∣ Γ = − 1 4 X μ ρ σ X ν ρ σ , g μ ν H μ ν = T , (53) H_{\mu\nu}:=\left.\frac{\partial T}{\partial g^{\mu\nu}}\right|_{\Gamma}=-\frac{1}{4}X_{\mu\rho\sigma}X_{\nu}{}^{\rho\sigma},\qquad g^{\mu\nu}H_{\mu\nu}=T, \tag*{(53)} H μν := ∂ g μν ∂ T Γ = − 4 1 X μ ρ σ X ν ρ σ , g μν H μν = T , (53)
R ( μ ν ) ( Γ ) = R ˚ μ ν + H μ ν , R ( Γ ) = R ˚ + T , (54) R_{(\mu\nu)}(\Gamma)=\mathring{R}_{\mu\nu}+H_{\mu\nu},\qquad R(\Gamma)=\mathring{R}+T, \tag*{(54)} R ( μν ) ( Γ ) = R ˚ μν + H μν , R ( Γ ) = R ˚ + T , (54)
R [ μ ν ] ( Γ ) = 1 2 ∇ ˚ λ X λ μ ν . (55) R_{[\mu\nu]}(\Gamma)=\frac{1}{2}\mathring{\nabla}_{\lambda}X^{\lambda}{}_{\mu\nu}. \tag*{(55)} R [ μν ] ( Γ ) = 2 1 ∇ ˚ λ X λ μν . (55)
The derivative defining H μ ν H_{\mu\nu} H μν is taken with the mixed-index torsion X λ μ ν X^{\lambda}{}_{\mu\nu} X λ μν fixed; one must not vary the already-restricted expression (52) as if X λ μ ν X_{\lambda\mu\nu} X λ μν were fixed. The full derivations and component checks are given in Appendices § 14 –§ 15 . The identities above do not assume affine flatness [2 , 3 ] .
Metric trace and axial field equation
For arbitrary torsion the trace of Eq. (47) reads
− α R − 2 β T − 2 γ T 3 + Ψ ( T ) H = 0 , H = g μ ν H μ ν , (56) -\alpha R-2\beta T-2\gamma T^{3}+\Psi(T)H=0,\qquad H=g^{\mu\nu}H_{\mu\nu}, \tag*{(56)} − α R − 2 β T − 2 γ T 3 + Ψ ( T ) H = 0 , H = g μν H μν , (56)
and its trace-free part is
α ( R ( μ ν ) − 1 4 R g μ ν ) + Ψ ( T ) ( H μ ν − 1 4 H g μ ν ) = 0. (57) \alpha\left(R_{(\mu\nu)}-\frac{1}{4}Rg_{\mu\nu}\right)+\Psi(T)\left(H_{\mu\nu}-\frac{1}{4}Hg_{\mu\nu}\right)=0. \tag*{(57)} α ( R ( μν ) − 4 1 R g μν ) + Ψ ( T ) ( H μν − 4 1 H g μν ) = 0. (57)
No identity H = T H=T H = T is asserted for unrestricted torsion.
After the axial specialization (53)–(54), the complete metric equation becomes
α G ˚ μ ν + [ α + Ψ ( T ) ] H μ ν − 1 2 g μ ν [ ( α + β ) T + γ T 3 ] = 0. (58) \alpha\mathring{G}_{\mu\nu}+\left[\alpha+\Psi(T)\right]H_{\mu\nu}-\frac{1}{2}g_{\mu\nu}\left[(\alpha+\beta)T+\gamma T^{3}\right]=0. \tag*{(58)} α G ˚ μν + [ α + Ψ ( T ) ] H μν − 2 1 g μν [ ( α + β ) T + γ T 3 ] = 0. (58)
Its trace and trace-free part, respectively, are
α R ˚ + ( α + β ) T − γ T 3 = 0 , (59) \alpha\mathring{R}+(\alpha+\beta)T-\gamma T^{3}=0, \tag*{(59)} α R ˚ + ( α + β ) T − γ T 3 = 0 , (59)
α ( R ˚ μ ν − 1 4 R ˚ g μ ν ) + [ α + Ψ ( T ) ] ( H μ ν − 1 4 T g μ ν ) = 0. (60) \alpha\left(\mathring{R}_{\mu\nu}-\frac{1}{4}\mathring{R}g_{\mu\nu}\right)+\left[\alpha+\Psi(T)\right]\left(H_{\mu\nu}-\frac{1}{4}Tg_{\mu\nu}\right)=0. \tag*{(60)} α ( R ˚ μν − 4 1 R ˚ g μν ) + [ α + Ψ ( T ) ] ( H μν − 4 1 T g μν ) = 0. (60)
These expressions retain the distinction between the Levi–Civita curvature and the torsion contribution.
Exceptional axial branch and Einstein reduction
The affine coefficients obtained by unrestricted variation reduce on the axial sector to
P μ ν λ ∣ a x i a l = − X λ μ ν , J μ ν λ ∣ a x i a l = − X λ μ ν . (61) \left.P^{\mu\nu}{}_{\lambda}\right|_{\mathrm{axial}}=-X_{\lambda}{}^{\mu\nu},\qquad\left.J^{\mu\nu}{}_{\lambda}\right|_{\mathrm{axial}}=-X_{\lambda}{}^{\mu\nu}. \tag*{(61)} P μν λ ∣ axial = − X λ μν , J μν λ ∣ axial = − X λ μν . (61)
Consequently, the complete connection equation is
[ α + β + 3 γ T 2 ] X λ μ ν = 0. (62) \left[\alpha+\beta+3\gamma T^{2}\right]X_{\lambda}{}^{\mu\nu}=0. \tag*{(62)} [ α + β + 3 γ T 2 ] X λ μν = 0. (62)
For nonzero axial torsion and γ ≠ 0 \gamma\ne0 γ = 0 , this implies, on a connected smooth branch,
T = T ∗ , T ∗ 2 = − α + β 3 γ ≥ 0 , Ψ ( T ∗ ) = − α . (63) T=T_{\ast},\qquad T_{\ast}^{2}=-\frac{\alpha+\beta}{3\gamma}\ge0,\qquad\Psi(T_{\ast})=-\alpha. \tag*{(63)} T = T ∗ , T ∗ 2 = − 3 γ α + β ≥ 0 , Ψ ( T ∗ ) = − α . (63)
The sign of a nonzero T ∗ T_{\ast} T ∗ is constant on such a connected branch. A nonzero Lorentzian three-form can also have T ∗ = 0 T_{\ast}=0 T ∗ = 0 , provided α + β = 0 \alpha+\beta=0 α + β = 0 . The geometric and algebraic details of these possibilities are developed in § 7 .
On the exceptional branch the H μ ν H_{\mu\nu} H μν term in Eq. (58) cancels. Since ( α + β ) T ∗ = − 3 γ T ∗ 3 (\alpha+\beta)T_{\ast}=-3\gamma T_{\ast}^{3} ( α + β ) T ∗ = − 3 γ T ∗ 3 , one obtains
α G ˚ μ ν + γ T ∗ 3 g μ ν = 0 , G ˚ μ ν + Λ ∗ g μ ν = 0 , Λ ∗ : = γ T ∗ 3 α . (64) \alpha\mathring{G}_{\mu\nu}+\gamma T_{\ast}^{3}g_{\mu\nu}=0,\qquad\mathring{G}_{\mu\nu}+\Lambda_{\ast}g_{\mu\nu}=0,\qquad\Lambda_{\ast}:=\frac{\gamma T_{\ast}^{3}}{\alpha}. \tag*{(64)} α G ˚ μν + γ T ∗ 3 g μν = 0 , G ˚ μν + Λ ∗ g μν = 0 , Λ ∗ := α γ T ∗ 3 . (64)
With the curvature sign convention of § 2 ,
R ˚ = 4 Λ ∗ , R ( Γ ) = 4 Λ ∗ + T ∗ . (65) \mathring{R}=4\Lambda_{\ast},\qquad R(\Gamma)=4\Lambda_{\ast}+T_{\ast}. \tag*{(65)} R ˚ = 4 Λ ∗ , R ( Γ ) = 4 Λ ∗ + T ∗ . (65)
For the exceptional linear model γ = 0 \gamma=0 γ = 0 and β = − α \beta=-\alpha β = − α , Eq. (62) instead imposes no algebraic condition on the axial torsion norm, while Eq. (58) reduces to G ˚ μ ν = 0 \mathring{G}_{\mu\nu}=0 G ˚ μν = 0 . These results concern the specified axial sector; they are not a classification of every metric–affine solution.
Implications for spherical black holes
Under the usual local spherical vacuum assumptions, the Einstein metric sector is Schwarzschild–(anti-)de Sitter, with
d s 2 = f ( r ) d t 2 − d r 2 f ( r ) − r 2 d Ω 2 , f ( r ) = 1 − 2 M r + Λ ∗ r 2 3 . (66) ds^{2}=f(r)dt^{2}-\frac{dr^{2}}{f(r)}-r^{2}d\Omega^{2},\qquad f(r)=1-\frac{2M}{r}+\frac{\Lambda_{\ast}r^{2}}{3}. \tag*{(66)} d s 2 = f ( r ) d t 2 − f ( r ) d r 2 − r 2 d Ω 2 , f ( r ) = 1 − r 2 M + 3 Λ ∗ r 2 . (66)
The sign of the cosmological term follows the Ricci convention adopted in § 2 ; relative to the common convention f = 1 − 2 M / r − Λ p h y s r 2 / 3 f=1-2M/r-\Lambda_{\mathrm{phys}}r^{2}/3 f = 1 − 2 M / r − Λ phys r 2 /3 , one has Λ p h y s = − Λ ∗ \Lambda_{\mathrm{phys}}=-\Lambda_{\ast} Λ phys = − Λ ∗ [30 , 31 , 32 ] . The metric alone does not fix the axial three-form. Section 7 proves the local existence statement for smooth constant-norm torsion, while § 8 and § 10 construct explicit spherical representatives and examine their horizon behavior and affine curvature invariants. No teleparallel condition R λ ρ μ ν ( Γ ) = 0 R^{\lambda}{}_{\rho\mu\nu}(\Gamma)=0 R λ ρ μν ( Γ ) = 0 is imposed.
Metric-compatible connections with totally antisymmetric torsion
We now determine a class of solutions of the unrestricted metric–affine equations of the polynomial Myrzakulov gravity-I model. In particular, metric compatibility and axial torsion are imposed on the resulting equations, not on the variational principle. This distinction is essential in metric–affine gravity and in applications of symmetric criticality [5 , 6 , 1 , 2 , 3 , 28 , 29 ] . Our curvature and connection-index conventions are those fixed in § 2 and checked in Appendix § 15 .
Axial Riemann–Cartan geometry
Let
X λ μ ν = 2 Γ λ [ μ ν ] , t μ = X λ λ μ , Q λ μ ν = − ∇ λ g μ ν . (67) X^{\lambda}{}_{\mu\nu}=2\Gamma^{\lambda}{}_{[\mu\nu]},\qquad t_{\mu}=X^{\lambda}{}_{\lambda\mu},\qquad Q_{\lambda\mu\nu}=-\nabla_{\lambda}g_{\mu\nu}. \tag*{(67)} X λ μν = 2 Γ λ [ μν ] , t μ = X λ λ μ , Q λ μν = − ∇ λ g μν . (67)
We evaluate the full field equations on the sector
Q λ μ ν = 0 , X λ μ ν = X [ λ μ ν ] . (68) Q_{\lambda\mu\nu}=0,\qquad X_{\lambda\mu\nu}=X_{[\lambda\mu\nu]}. \tag*{(68)} Q λ μν = 0 , X λ μν = X [ λ μν ] . (68)
The torsion trace vanishes, t μ = 0 t_{\mu}=0 t μ = 0 , and the metric-compatible connection is
Γ λ μ ν = Γ ˚ λ μ ν + 1 2 X λ μ ν . (69) \Gamma^{\lambda}{}_{\mu\nu}=\mathring{\Gamma}^{\lambda}{}_{\mu\nu}+\frac{1}{2}X^{\lambda}{}_{\mu\nu}. \tag*{(69)} Γ λ μν = Γ ˚ λ μν + 2 1 X λ μν . (69)
The tensor X λ μ ν X_{\lambda\mu\nu} X λ μν defines the three-form
τ = 1 3 ! X λ μ ν d x λ ∧ d x μ ∧ d x ν . (70) \tau=\frac{1}{3!}X_{\lambda\mu\nu}dx^{\lambda}\wedge dx^{\mu}\wedge dx^{\nu}. \tag*{(70)} τ = 3 ! 1 X λ μν d x λ ∧ d x μ ∧ d x ν . (70)
The affine curvature of (69) is generally nonzero; these configurations are Riemann–Cartan, not necessarily teleparallel [1 , 2 , 3 ] .
Quadratic torsion scalar and fixed-connection metric derivative
The torsion scalar is defined for an arbitrary independent connection by
T = 1 4 X λ μ ν X λ μ ν + 1 2 X λ μ ν X ν μ λ − t μ t μ . (71) T = \frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu} + \frac{1}{2}X_{\lambda\mu\nu}X^{\nu\mu\lambda} - t_\mu t^\mu. \tag*{(71)} T = 4 1 X λ μν X λ μν + 2 1 X λ μν X ν μ λ − t μ t μ . (71)
Since X ν μ λ = − X λ μ ν X^{\nu\mu\lambda}=-X^{\lambda\mu\nu} X ν μ λ = − X λ μν for a three-form, the axial value is
T = − 1 4 X λ μ ν X λ μ ν . (72) T = -\frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu}. \tag*{(72)} T = − 4 1 X λ μν X λ μν . (72)
It is important that (72) is not varied as a restricted action. In the unrestricted metric variation, X b c a X^{a}_{\ bc} X b c a is fixed and X a b c = g a d X b c d X_{abc}=g_{ad}X^{d}_{\ bc} X ab c = g a d X b c d is not. An exact expression prior to imposing axial torsion is
T = 1 4 g a d g b p g c q X b c a X p q d + 1 2 g b p X b c a X p a c − g b c t b t c . (73) T = \frac{1}{4}g_{ad}g^{bp}g^{cq}X^{a}_{\ bc}X^{d}_{\ pq}+\frac{1}{2}g^{bp}X^{a}_{\ bc}X^{c}_{\ pa}-g^{bc}t_b t_c. \tag*{(73)} T = 4 1 g a d g b p g c q X b c a X pq d + 2 1 g b p X b c a X p a c − g b c t b t c . (73)
Define H μ ν : = ( ∂ T / ∂ g μ ν ) Γ H_{\mu\nu}:=\left(\partial T/\partial g^{\mu\nu}\right)_{\Gamma} H μν := ( ∂ T / ∂ g μν ) Γ . Differentiating (73) first, and then evaluating on (68) , gives
H μ ν ∣ a x i a l = − 1 4 X μ ρ σ X ν ρ σ , g μ ν H μ ν = T . (74) \left.H_{\mu\nu}\right|_{\mathrm{axial}}=-\frac{1}{4}X_{\mu\rho\sigma}X_{\nu}^{\ \rho\sigma},\qquad g^{\mu\nu}H_{\mu\nu}=T. \tag*{(74)} H μν ∣ axial = − 4 1 X μ ρ σ X ν ρ σ , g μν H μν = T . (74)
The trace also follows from the degree-one homogeneity of (73) in the inverse metric, with the mixed torsion held fixed. An independent exact component check of (74) has been performed for all ten symmetric inverse-metric variations at a Lorentz-orthonormal point with four arbitrary axial components. This is a pointwise algebraic verification; the tensorial derivation is supplied by (73) and Appendix § 15 .
Curvature decomposition
With the curvature convention of § 2 , the exact distortion formula is
R ρ μ ν λ ( Γ ) = R ˚ ρ μ ν λ + 2 ∇ ˚ [ μ K ∣ ρ ∣ ν ] λ + 2 K σ [ μ λ K ∣ ρ ∣ ν ] σ , K μ ν λ = 1 2 X μ ν λ . (75) R^{\lambda}_{\ \rho\mu\nu}(\Gamma)=\mathring{R}^{\lambda}_{\ \rho\mu\nu}+2\mathring{\nabla}_{[\mu}K^{\lambda}_{\ |\rho|\nu]}+2K^{\lambda}_{\ \sigma[\mu}K^{\sigma}_{\ |\rho|\nu]},\qquad K^{\lambda}_{\ \mu\nu}=\frac{1}{2}X^{\lambda}_{\ \mu\nu}. \tag*{(75)} R ρ μν λ ( Γ ) = R ˚ ρ μν λ + 2 ∇ ˚ [ μ K ∣ ρ ∣ ν ] λ + 2 K σ [ μ λ K ∣ ρ ∣ ν ] σ , K μν λ = 2 1 X μν λ . (75)
The divergence of a three-form is antisymmetric in its remaining two indices. Consequently, the symmetric and antisymmetric Ricci parts are
R ( μ ν ) ( Γ ) = R ˚ μ ν − 1 4 X μ ρ σ X ν ρ σ = R ˚ μ ν + H μ ν , (76) R_{(\mu\nu)}(\Gamma)=\mathring{R}_{\mu\nu}-\frac{1}{4}X_{\mu\rho\sigma}X_{\nu}^{\ \rho\sigma}=\mathring{R}_{\mu\nu}+H_{\mu\nu}, \tag*{(76)} R ( μν ) ( Γ ) = R ˚ μν − 4 1 X μ ρ σ X ν ρ σ = R ˚ μν + H μν , (76)
R [ μ ν ] ( Γ ) = 1 2 ∇ ˚ λ X μ ν λ , (77) R_{[\mu\nu]}(\Gamma)=\frac{1}{2}\mathring{\nabla}_{\lambda}X^{\lambda}_{\ \mu\nu}, \tag*{(77)} R [ μν ] ( Γ ) = 2 1 ∇ ˚ λ X μν λ , (77)
R ( Γ ) = R ˚ + T . (78) R(\Gamma)=\mathring{R}+T. \tag*{(78)} R ( Γ ) = R ˚ + T . (78)
These are geometric identities on the axial Riemann–Cartan sector, independent of the field equations. In particular, (77) need not vanish. Their index ordering should be kept consistent with Appendix § 14 [2 , 3 ] .
Unrestricted affine field equation
Consider the vacuum action
S [ g , Γ ] = 1 2 κ ∫ d 4 x − g F ( R , T ) , F ( R , T ) = α R + β T + γ T 3 , (79) S[g,\Gamma]=\frac{1}{2\kappa}\int d^{4}x\sqrt{-g}\,F(R,T),\qquad F(R,T)=\alpha R+\beta T+\gamma T^{3}, \tag*{(79)} S [ g , Γ ] = 2 κ 1 ∫ d 4 x − g F ( R , T ) , F ( R , T ) = α R + β T + γ T 3 , (79)
with α , β , γ \alpha,\beta,\gamma α , β , γ constant. Introduce
F R = α , F T = Ψ ( T ) : = β + 3 γ T 2 . (80) F_R=\alpha,\qquad F_T=\Psi(T):=\beta+3\gamma T^{2}. \tag*{(80)} F R = α , F T = Ψ ( T ) := β + 3 γ T 2 . (80)
All 64 ordered components of Γ μ ν λ \Gamma^\lambda_{\mu\nu} Γ μν λ are varied independently. Writing the integrated-by-parts connection derivative of − g R \sqrt{-g}R − g R as P λ μ ν P_\lambda{}^{\mu\nu} P λ μν , and the algebraic derivative of T T T as J λ μ ν J_\lambda{}^{\mu\nu} J λ μν , the affine equation is
α P λ μ ν + Ψ ( T ) J λ μ ν = 0. (81) \alpha P_\lambda{}^{\mu\nu}+\Psi(T)J_\lambda{}^{\mu\nu}=0. \tag*{(81)} α P λ μν + Ψ ( T ) J λ μν = 0. (81)
For clarity, in the convention ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ \nabla_\nu V^\lambda=\partial_\nu V^\lambda+\Gamma^\lambda_{\rho\nu}V^\rho ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ , the coefficient of δ Γ μ ν λ \delta\Gamma^\lambda_{\mu\nu} δ Γ μν λ can be obtained without restricting the connection:
P λ μ ν = − 1 − g ∂ λ ( − g g μ ν ) + δ λ ν − g ∂ σ ( − g g μ σ ) + δ λ ν g ρ σ Γ ρ σ μ + g μ ν Γ λ ρ ρ − g ρ ν Γ λ ρ μ − g μ ρ Γ λ ρ ν . (82) P_\lambda{}^{\mu\nu}=-\frac{1}{\sqrt{-g}}\partial_\lambda\left(\sqrt{-g}g^{\mu\nu}\right)+\frac{\delta^\nu_\lambda}{\sqrt{-g}}\partial_\sigma\left(\sqrt{-g}g^{\mu\sigma}\right)+\delta^\nu_\lambda g^{\rho\sigma}\Gamma^\mu_{\rho\sigma}+g^{\mu\nu}\Gamma^\rho_{\lambda\rho}-g^{\rho\nu}\Gamma^\mu_{\lambda\rho}-g^{\mu\rho}\Gamma^\nu_{\lambda\rho}. \tag*{(82)} P λ μν = − − g 1 ∂ λ ( − g g μν ) + − g δ λ ν ∂ σ ( − g g μ σ ) + δ λ ν g ρ σ Γ ρ σ μ + g μν Γ λ ρ ρ − g ρ ν Γ λ ρ μ − g μ ρ Γ λ ρ ν . (82)
The coefficient J λ μ ν J_\lambda{}^{\mu\nu} J λ μν is obtained from (71) by δ X λ μ ν = 2 δ Γ λ [ μ ν ] \delta X^\lambda{}_{\mu\nu}=2\delta\Gamma^\lambda{}_{[\mu\nu]} δ X λ μν = 2 δ Γ λ [ μν ] at fixed metric. Evaluating these unrestricted derivatives on (68) yields
P λ μ ν ∣ a x i a l = − X λ μ ν , J λ μ ν ∣ a x i a l = − X λ μ ν . (83) \left.P_\lambda{}^{\mu\nu}\right|_{\mathrm{axial}}=-X_\lambda{}^{\mu\nu},\qquad\left.J_\lambda{}^{\mu\nu}\right|_{\mathrm{axial}}=-X_\lambda{}^{\mu\nu}. \tag*{(83)} P λ μν ∣ axial = − X λ μν , J λ μν ∣ axial = − X λ μν . (83)
The symmetric components in ( μ , ν ) (\mu,\nu) ( μ , ν ) vanish individually, while the antisymmetric components of (81) give
[ α + β + 3 γ T 2 ] X λ μ ν = 0. (84) \left[\alpha+\beta+3\gamma T^2\right]X_\lambda{}^{\mu\nu}=0. \tag*{(84)} [ α + β + 3 γ T 2 ] X λ μν = 0. (84)
The local coefficients in (83) have also been checked algebraically in all 64 ordered connection directions at a Levi–Civita normal point. This component check is complementary to, not a replacement for, the integration-by-parts derivation of (82) [1 , 2 , 3 , 5 , 6 ] .
Classification of exceptional branches
If X λ μ ν ≠ 0 X_{\lambda\mu\nu}\ne0 X λ μν = 0 at a point, (84) requires
α + β + 3 γ T 2 = 0. (85) \alpha+\beta+3\gamma T^2=0. \tag*{(85)} α + β + 3 γ T 2 = 0. (85)
For γ ≠ 0 \gamma\ne0 γ = 0 and − ( α + β ) / ( 3 γ ) > 0 -(\alpha+\beta)/(3\gamma)>0 − ( α + β ) / ( 3 γ ) > 0 , smoothness on a connected region with nonzero axial torsion selects one of the two constant-norm branches
T = T ∗ , T ∗ = ± − α + β 3 γ . (86) T=T_\ast,\qquad T_\ast=\pm\sqrt{-\frac{\alpha+\beta}{3\gamma}}. \tag*{(86)} T = T ∗ , T ∗ = ± − 3 γ α + β . (86)
When α + β = 0 \alpha+\beta=0 α + β = 0 , the exceptional value is T ∗ = 0 T_\ast=0 T ∗ = 0 . A nonzero Lorentzian three-form can have zero quadratic norm, so this null axial branch is not torsion-free. If γ = 0 \gamma=0 γ = 0 , nonzero axial torsion instead requires β = − α \beta=-\alpha β = − α , and the affine equation places no algebraic restriction on T ( x ) T(x) T ( x ) . For α = 0 \alpha=0 α = 0 the effective Einstein reduction below is unavailable and the metric equations require a separate analysis; we do not include this degenerate case in the existence proposition.
Metric equation and Einstein reduction
The vacuum metric equation obtained by varying g μ ν g^{\mu\nu} g μν at fixed Γ \Gamma Γ is
α R ( μ ν ) ( Γ ) − 1 2 g μ ν ( α R + β T + γ T 3 ) + Ψ ( T ) H μ ν = 0. (87) \alpha R_{(\mu\nu)}(\Gamma)-\frac{1}{2}g_{\mu\nu}\left(\alpha R+\beta T+\gamma T^3\right)+\Psi(T)H_{\mu\nu}=0. \tag*{(87)} α R ( μν ) ( Γ ) − 2 1 g μν ( α R + β T + γ T 3 ) + Ψ ( T ) H μν = 0. (87)
Substituting (76) and (78) , without making any additional field-equation assumption, gives
α G ˚ μ ν + [ α + Ψ ( T ) ] H μ ν − 1 2 g μ ν [ ( α + β ) T + γ T 3 ] = 0. (88) \alpha\mathring{G}_{\mu\nu}+\left[\alpha+\Psi(T)\right]H_{\mu\nu}-\frac{1}{2}g_{\mu\nu}\left[(\alpha+\beta)T+\gamma T^3\right]=0. \tag*{(88)} α G ˚ μν + [ α + Ψ ( T ) ] H μν − 2 1 g μν [ ( α + β ) T + γ T 3 ] = 0. (88)
On the nonlinear exceptional branch, Ψ ( T ∗ ) = − α \Psi(T_\ast)=-\alpha Ψ ( T ∗ ) = − α and ( α + β ) T ∗ = − 3 γ T ∗ 3 (\alpha+\beta)T_\ast=-3\gamma T_\ast^3 ( α + β ) T ∗ = − 3 γ T ∗ 3 . Therefore,
α G ˚ μ ν + γ T ∗ 3 g μ ν = 0. (89) \alpha\mathring{G}_{\mu\nu}+\gamma T_\ast^3g_{\mu\nu}=0. \tag*{(89)} α G ˚ μν + γ T ∗ 3 g μν = 0. (89)
For α ≠ 0 \alpha\ne0 α = 0 , set
G ˚ μ ν + Λ ∗ g μ ν = 0 , Λ ∗ : = γ T ∗ 3 α . (90) \mathring{G}_{\mu\nu}+\Lambda_*g_{\mu\nu}=0,\qquad\Lambda_*:=\frac{\gamma T_*^3}{\alpha}. \tag*{(90)} G ˚ μν + Λ ∗ g μν = 0 , Λ ∗ := α γ T ∗ 3 . (90)
With the Ricci-sign convention of § 2 , the trace gives R ˚ = 4 Λ ∗ \mathring{R}=4\Lambda_* R ˚ = 4 Λ ∗ and hence R ( Γ ) = 4 Λ ∗ + T ∗ R(\Gamma)=4\Lambda_*+T_* R ( Γ ) = 4 Λ ∗ + T ∗ . For the linear exceptional model γ = 0 \gamma=0 γ = 0 , β = − α \beta=-\alpha β = − α , the same calculation instead yields G ˚ μ ν = 0 \mathring{G}_{\mu\nu}=0 G ˚ μν = 0 without fixing T ( x ) T(x) T ( x ) . The interpretation of the cosmological term is convention dependent: in the conventional static metric f = 1 − 2 M / r − Λ p h y s r 2 / 3 f=1-2M/r-\Lambda_{\mathrm{phys}}r^2/3 f = 1 − 2 M / r − Λ phys r 2 /3 used in much of the black-hole literature, Λ p h y s = − Λ ∗ \Lambda_{\mathrm{phys}}=-\Lambda_* Λ phys = − Λ ∗ for the sign convention of this manuscript [30 , 31 , 32 ] .
Local existence proposition and its scope
The following statement concerns classical solutions of the unrestricted metric–affine Euler–Lagrange equations, subsequently evaluated in the metric-compatible, totally antisymmetric torsion sector. In particular, metric compatibility and axiality are not imposed in the action before variation; this distinction is important in metric–affine gravity and in symmetry reductions [5 , 6 , 1 , 2 , 28 , 29 ] .
Proposition (local axial solutions of polynomial MG-I). Let U U U be an open subset of a four-dimensional Lorentzian manifold, and let g g g be a smooth, nondegenerate Lorentzian metric on U U U . Fix real constants α , β , γ \alpha,\beta,\gamma α , β , γ with α ≠ 0 \alpha\ne0 α = 0 . Suppose that a real constant T ∗ T_* T ∗ obeys
α + β + 3 γ T ∗ 2 = 0 , Λ ∗ : = γ T ∗ 3 α , (91) \alpha+\beta+3\gamma T_*^2=0,\qquad\Lambda_*:=\frac{\gamma T_*^3}{\alpha}, \tag*{(91)} α + β + 3 γ T ∗ 2 = 0 , Λ ∗ := α γ T ∗ 3 , (91) and that the metric satisfies
G ˚ μ ν + Λ ∗ g μ ν = 0 on U . (92) \mathring{G}_{\mu\nu}+\Lambda_*g_{\mu\nu}=0\qquad\text{on }U. \tag*{(92)} G ˚ μν + Λ ∗ g μν = 0 on U . (92) Let X ∈ Ω 3 ( U ) X\in\Omega^3(U) X ∈ Ω 3 ( U ) be a smooth, real three-form, not identically zero, whose torsion scalar, defined with the conventions of § 2 , satisfies
T ( g , X ) = − 1 4 X λ μ ν X λ μ ν = T ∗ throughout U . (93) T(g,X)=-\frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu}=T_*\qquad\text{throughout }U. \tag*{(93)} T ( g , X ) = − 4 1 X λ μν X λ μν = T ∗ throughout U . (93) Then the affine connection
Γ μ ν λ = Γ ˚ μ ν λ + 1 2 g λ ρ X ρ μ ν (94) \Gamma^\lambda_{\mu\nu}=\mathring{\Gamma}^\lambda_{\mu\nu}+\frac{1}{2}g^{\lambda\rho}X_{\rho\mu\nu} \tag*{(94)} Γ μν λ = Γ ˚ μν λ + 2 1 g λ ρ X ρ μν (94) is smooth and metric-compatible, has torsion 2 Γ [ μ ν ] λ = X λ μ ν 2\Gamma^\lambda_{[\mu\nu]}=X^\lambda{}_{\mu\nu} 2 Γ [ μν ] λ = X λ μν , and, together with g g g , satisfies all vacuum metric and affine field equations of F ( R , T ) = α R + β T + γ T 3 F(R,T)=\alpha R+\beta T+\gamma T^3 F ( R , T ) = α R + β T + γ T 3 on U U U .
Proof. Because g g g is nondegenerate and smooth, its Levi–Civita connection is smooth. The tensor g λ ρ X ρ μ ν / 2 g^{\lambda\rho}X_{\rho\mu\nu}/2 g λ ρ X ρ μν /2 is smooth and antisymmetric in its lowered first two indices; hence (94) is metric-compatible and has the prescribed totally antisymmetric torsion. The connection is evaluated in the field equations obtained by independently varying all ten symmetric metric components and all 64 ordered affine components. According to the unrestricted affine variation summarized in Eqs. (81) –(84) and Appendix § 15 , its complete affine equation reduces to
[ α + β + 3 γ T ( g , X ) 2 ] X λ μ ν = 0. (95) [\alpha+\beta+3\gamma T(g,X)^2]X_\lambda{}^{\mu\nu}=0. \tag*{(95)} [ α + β + 3 γ T ( g , X ) 2 ] X λ μν = 0. (95)
The coefficient vanishes pointwise by (91) and (93) ; in particular, this verifies the antisymmetric affine equations, while the symmetric affine components vanish separately on this sector.
For the metric equation, the axial curvature identities (76)–(78), established with the conventions of § 2 and checked in Appendix § 14 , reduce the unrestricted metric equation to Eq. (87) . The coefficient of H μ ν H_{\mu\nu} H μν vanishes by (91) , and ( α + β ) T ∗ = − 3 γ T ∗ 3 (\alpha+\beta)T_*=-3\gamma T_*^3 ( α + β ) T ∗ = − 3 γ T ∗ 3 gives
α G ˚ μ ν + γ T ∗ 3 g μ ν = 0. (96) \alpha\mathring{G}_{\mu\nu}+\gamma T_*^3g_{\mu\nu}=0. \tag*{(96)} α G ˚ μν + γ T ∗ 3 g μν = 0. (96)
This is exactly (92) multiplied by α \alpha α . Both independent sets of vacuum field equations therefore hold at every point of $U.
Corollary (linear exceptional model). Let α ≠ 0 \alpha\ne0 α = 0 , γ = 0 \gamma= 0 γ = 0 , and β = − α \beta= -\alpha β = − α . Let ( U , g ) (U,g) ( U , g ) be a smooth Lorentzian vacuum Einstein spacetime satisfying
G ˚ μ ν = 0. \mathring{G}_{\mu\nu}=0. G ˚ μν = 0. Then any smooth axial torsion three-form X ∈ Ω 3 ( U ) X \in\Omega^3(U) X ∈ Ω 3 ( U ) defines, through
Γ μ ν λ = Γ ˚ μ ν λ + 1 2 g λ ρ X ρ μ ν , \Gamma^\lambda_{\mu\nu}=\mathring{\Gamma}^\lambda_{\mu\nu}+\frac{1}{2}g^{\lambda\rho}X_{\rho\mu\nu}, Γ μν λ = Γ ˚ μν λ + 2 1 g λ ρ X ρ μν , a solution of the unrestricted vacuum metric-affine field equations. No constant-norm condition on T ( g , X ) T(g,X) T ( g , X ) is required.
Proof. In the axial sector, the unrestricted affine equation reduces to ( α + β ) X λ μ ν = 0 (\alpha+\beta)X_\lambda{}^{\mu\nu}=0 ( α + β ) X λ μν = 0 , which vanishes identically. The metric equation reduces to α G ˚ μ ν = 0 \alpha\mathring{G}_{\mu\nu}=0 α G ˚ μν = 0 . Both equations are therefore satisfied. □
Regularity assumptions and local meaning. Smoothness of g g g and X X X is a sufficient hypothesis for the classical pointwise field equations and for a smooth affine connection. The proposition is conditional: it does not assert that a smooth three-form of prescribed norm exists globally on a given spacetime, or even on an arbitrarily prescribed region. In a chosen local coordinate chart, regularity must be assessed from the tensorial three-form and the reconstructed connection, rather than from components in a frame that becomes singular. In particular, smoothness through a Killing horizon requires a horizon-penetrating chart and a smooth extension of X X X in that chart; the construction in Sec. 10 supplies such representatives on suitable future-horizon neighborhoods [30 , 31 , 32 ] .
For the explicit spherical representative of Sec. 10, the strict inequality a 2 + c f ( r ) > 0 a^2+cf(r)>0 a 2 + c f ( r ) > 0 is a local domain condition ensuring that the chosen square-root branch is real and smooth. At a horizon f ( r h ) = 0 f(r_h)=0 f ( r h ) = 0 with a > 0 a>0 a > 0 , it holds automatically on a sufficiently small neighborhood. Neither the proposition nor this inequality proves a smooth extension across every horizon, a bifurcation two-sphere, or a maximal spacetime extension. The vanishing of the square-root radicand may obstruct this particular representative without excluding other local torsion profiles.
The condition T = T ∗ T=T_* T = T ∗ is algebraic and does not imply ∇ ˚ ρ X λ μ ν = 0 \mathring{\nabla}_{\rho}X_{\lambda\mu\nu}=0 ∇ ˚ ρ X λ μν = 0 , nor does it impose R λ ρ μ ν ( Γ ) = 0 R^\lambda{}_{\rho\mu\nu}(\Gamma)=0 R λ ρ μν ( Γ ) = 0 ; the solutions are generally Riemann–Cartan rather than teleparallel [1 , 2 , 3 ] . The exceptional value T ∗ = 0 T_*=0 T ∗ = 0 allows nonzero null axial torsion in Lorentzian signature. For γ = 0 \gamma=0 γ = 0 and β = − α \beta=-\alpha β = − α , the same field-equation argument gives G ˚ μ ν = 0 \mathring{G}_{\mu\nu}=0 G ˚ μν = 0 for any smooth axial three-form, without requiring its norm to be constant; this is a broader linear exceptional case than the constant-norm statement above.
Local realizability of the norm condition. The prescribed-norm hypothesis is locally nonempty. On any sufficiently small Lorentzian coordinate neighborhood, choose a smooth orthonormal coframe and a nonzero axial three-form with constant frame components. Because the quadratic form on three-forms has indefinite signature, its coefficients can be selected to realize either sign of T ∗ T_* T ∗ , as well as T ∗ = 0 T_*=0 T ∗ = 0 with nonzero torsion. This is a local construction; it makes no assertion about the existence of a globally defined constant-norm three-form on an arbitrary spacetime. See also the geometric discussion of torsion in Refs. [2 , 3 ] .
Finally, the proposition establishes neither uniqueness nor global regularity, perturbative stability, a propagating torsion degree of freedom, nor physical inequivalence of different torsion profiles. Such conclusions require additional dynamical or invariant analyses; local solvability alone is insufficient [2 , 3 ] . The spherical Ricci-symmetric subclass of this proposition is characterized further in Sec. 10.6 by ( f V − U ) ′ = 0 (fV-U)'=0 ( f V − U ) ′ = 0 ; its local parameterization and horizon-regular representatives are not a classification of unrestricted metric-affine connections.
Spherical axial data and the horizon problem
For the orthonormal spherical coframe used in Sec. 8, a rotation-invariant axial three-form can be written
τ = x ( r ) e 0 ∧ e 2 ∧ e 3 − y ( r ) e 1 ∧ e 2 ∧ e 3 , (97) \tau=x(r)e^0\wedge e^2\wedge e^3-y(r)e^1\wedge e^2\wedge e^3, \tag*{(97)} τ = x ( r ) e 0 ∧ e 2 ∧ e 3 − y ( r ) e 1 ∧ e 2 ∧ e 3 , (97)
so that
T = 3 2 [ y ( r ) 2 − x ( r ) 2 ] , y ( r ) 2 − x ( r ) 2 = 2 T ∗ 3 . (98) T=\frac{3}{2}\left[y(r)^2-x(r)^2\right],\qquad y(r)^2-x(r)^2=\frac{2T_*}{3}. \tag*{(98)} T = 2 3 [ y ( r ) 2 − x ( r ) 2 ] , y ( r ) 2 − x ( r ) 2 = 3 2 T ∗ . (98)
The algebraic constraint leaves one local functional freedom within this ansatz. It does not by itself identify a gauge symmetry or a propagating mode. In a static frame, x x x and y y y may become singular at a Killing horizon even when the three-form is regular. Section 10 therefore reconstructs the torsion and affine connection in horizon-penetrating coordinates and checks their curvature invariants. These local tests must be distinguished from maximal analytic extension and global regularity [2 , 3 , 30 , 32 ] .
Static spherical reduction and axial torsion branches
We now apply the local axial-connection result of Sec. 7 to static, spherically symmetric spacetimes. The independent metric and affine variations are performed before restricting the solution to the Riemann–Cartan sector. This order of operations is important in metric–affine theories and is not guaranteed by a symmetry-reduced action alone [5 , 6 , 1 , 2 , 28 , 29 ] . Throughout this section the signature is ( + − − − ) (+---) ( + − − − ) , the connection is ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ \nabla_\nu V^\lambda=\partial_\nu V^\lambda+\Gamma^\lambda_{\rho\nu}V^\rho ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ , and X μ ν λ = 2 Γ [ μ ν ] λ X^\lambda_{\ \mu\nu}=2\Gamma^\lambda_{[\mu\nu]} X μν λ = 2 Γ [ μν ] λ .
Invariant coframe and axial three-form
In a static region let
d s 2 = A ( r ) d t 2 − B ( r ) d r 2 − C ( r ) d Ω 2 , d Ω 2 = d θ 2 + sin 2 θ d ϕ 2 , A , B , C > 0. (99) ds^2=A(r)dt^2-B(r)dr^2-C(r)d\Omega^2,\qquad d\Omega^2=d\theta^2+\sin^2\theta\,d\phi^2,\qquad A,B,C>0. \tag*{(99)} d s 2 = A ( r ) d t 2 − B ( r ) d r 2 − C ( r ) d Ω 2 , d Ω 2 = d θ 2 + sin 2 θ d ϕ 2 , A , B , C > 0. (99)
An orthonormal coframe is
e 0 = A d t , e 1 = B d r , e 2 = C d θ , e 3 = C sin θ d ϕ . (100) e^0=\sqrt{A}\,dt,\qquad e^1=\sqrt{B}\,dr,\qquad e^2=\sqrt{C}\,d\theta,\qquad e^3=\sqrt{C}\sin\theta\,d\phi. \tag*{(100)} e 0 = A d t , e 1 = B d r , e 2 = C d θ , e 3 = C sin θ d ϕ . (100)
The most general static, rotation-invariant axial three-form on the chosen connected static patch can be written as
τ : = 1 3 ! X a b c e a ∧ e b ∧ e c = x ( r ) e 0 ∧ e 2 ∧ e 3 − y ( r ) e 1 ∧ e 2 ∧ e 3 . (101) \tau:=\frac{1}{3!}X_{abc}e^a\wedge e^b\wedge e^c=x(r)e^0\wedge e^2\wedge e^3-y(r)e^1\wedge e^2\wedge e^3. \tag*{(101)} τ := 3 ! 1 X ab c e a ∧ e b ∧ e c = x ( r ) e 0 ∧ e 2 ∧ e 3 − y ( r ) e 1 ∧ e 2 ∧ e 3 . (101)
The distinction between rotational invariance of a torsion field and more restrictive cosmological symmetry assumptions is discussed in the classical literature on spacetime torsion [35 ] . Here X 023 = x X_{023}=x X 023 = x and X 123 = − y X_{123}=-y X 123 = − y . Defining T a = 1 2 X b c a e b ∧ e c T^a=\frac{1}{2}X^a_{\ bc}e^b\wedge e^c T a = 2 1 X b c a e b ∧ e c , raising the first index with η a b = diag ( 1 , − 1 , − 1 , − 1 ) \eta^{ab}=\operatorname{diag}(1,-1,-1,-1) η ab = diag ( 1 , − 1 , − 1 , − 1 ) gives
T 0 = x e 2 ∧ e 3 , T 1 = y e 2 ∧ e 3 , (102) T^0=x e^2\wedge e^3,\qquad T^1=y e^2\wedge e^3, \tag*{(102)} T 0 = x e 2 ∧ e 3 , T 1 = y e 2 ∧ e 3 , (102)
T 2 = x e 0 ∧ e 3 − y e 1 ∧ e 3 , T 3 = − x e 0 ∧ e 2 + y e 1 ∧ e 2 . (103) T^2=x e^0\wedge e^3-y e^1\wedge e^3,\qquad T^3=-x e^0\wedge e^2+y e^1\wedge e^2. \tag*{(103)} T 2 = x e 0 ∧ e 3 − y e 1 ∧ e 3 , T 3 = − x e 0 ∧ e 2 + y e 1 ∧ e 2 . (103)
In particular, the x e 0 ∧ e 3 xe^0\wedge e^3 x e 0 ∧ e 3 term in T 2 T^2 T 2 has a plus sign. This follows directly from X 03 2 = η 22 X 203 = x X^2_{\ 03}=\eta^{22}X_{203}=x X 03 2 = η 22 X 203 = x ; the opposite sign would contradict total antisymmetry. The metric-compatible connection is
Γ μ ν λ = Γ ˚ μ ν λ + 1 2 g λ ρ X ρ μ ν . (104) \Gamma^\lambda_{\ \mu\nu}=\mathring{\Gamma}^\lambda_{\ \mu\nu}+\frac{1}{2}g^{\lambda\rho}X_{\rho\mu\nu}. \tag*{(104)} Γ μν λ = Γ ˚ μν λ + 2 1 g λ ρ X ρ μν . (104)
This is a Riemann–Cartan connection; its curvature is not assumed to vanish [33 , 34 , 1 , 2 , 3 ] .
Quadratic contractions and component field equations
For the torsion scalar defined in Sec. 2, the contractions of (101) are
X a b c X a b c = 6 ( x 2 − y 2 ) , T = − 1 4 X a b c X a b c = 3 2 ( y 2 − x 2 ) . (105) X_{abc}X^{abc}=6(x^2-y^2),\qquad T=-\frac{1}{4}X_{abc}X^{abc}=\frac{3}{2}(y^2-x^2). \tag*{(105)} X ab c X ab c = 6 ( x 2 − y 2 ) , T = − 4 1 X ab c X ab c = 2 3 ( y 2 − x 2 ) . (105)
The fixed-connection metric derivative, evaluated after unrestricted variation, is H a b = − 1 4 X a c d X b c d H_{ab}=-\frac{1}{4}X_{acd}X_b{}^{cd} H ab = − 4 1 X a c d X b c d . Its nonzero components are
( H a b ) = 1 2 ( − x 2 x y 0 0 x y − y 2 0 0 0 0 x 2 − y 2 0 0 0 0 x 2 − y 2 ) , η a b H a b = T , H a b H a b = 1 3 T 2 . (106) (H_{ab})=\frac{1}{2}\begin{pmatrix}
-x^2 & xy & 0 & 0 \\
xy & -y^2 & 0 & 0 \\
0 & 0 & x^2-y^2 & 0 \\
0 & 0 & 0 & x^2-y^2
\end{pmatrix},\qquad\eta^{ab}H_{ab}=T,\qquad H_{ab}H^{ab}=\frac{1}{3}T^2. \tag*{(106)} ( H ab ) = 2 1 − x 2 x y 0 0 x y − y 2 0 0 0 0 x 2 − y 2 0 0 0 0 x 2 − y 2 , η ab H ab = T , H ab H ab = 3 1 T 2 . (106)
The last identity follows by direct contraction and does not require constant T T T .
For F ( R , T ) = α R + β T + γ T 3 F(R,T)=\alpha R+\beta T+\gamma T^3 F ( R , T ) = α R + β T + γ T 3 , set
Ψ ( T ) = β + 3 γ T 2 , V ( T ) = ( α + β ) T + γ T 3 . (107) \Psi(T)=\beta+3\gamma T^2,\qquad\mathcal{V}(T)=(\alpha+\beta)T+\gamma T^3. \tag*{(107)} Ψ ( T ) = β + 3 γ T 2 , V ( T ) = ( α + β ) T + γ T 3 . (107)
The axial curvature identities and the fixed-connection derivative established in § 7 reduce the full metric equation to
α G ˙ a b + [ α + Ψ ( T ) ] H a b − 1 2 η a b V ( T ) = 0. (108) \alpha\dot{G}_{ab}+[\alpha+\Psi(T)]H_{ab}-\frac{1}{2}\eta_{ab}\mathcal{V}(T)=0. \tag*{(108)} α G ˙ ab + [ α + Ψ ( T )] H ab − 2 1 η ab V ( T ) = 0. (108)
In the orthonormal coframe its independent components read
0 = α G ˙ 00 − 1 2 ( α + Ψ ) x 2 − 1 2 V , (109) 0=\alpha\dot{G}_{00}-\frac{1}{2}(\alpha+\Psi)x^2-\frac{1}{2}\mathcal{V}, \tag*{(109)} 0 = α G ˙ 00 − 2 1 ( α + Ψ ) x 2 − 2 1 V , (109)
0 = α G ˙ 11 − 1 2 ( α + Ψ ) y 2 + 1 2 V , (110) 0=\alpha\dot{G}_{11}-\frac{1}{2}(\alpha+\Psi)y^2+\frac{1}{2}\mathcal{V}, \tag*{(110)} 0 = α G ˙ 11 − 2 1 ( α + Ψ ) y 2 + 2 1 V , (110)
0 = α G ˙ 22 + 1 2 ( α + Ψ ) ( x 2 − y 2 ) + 1 2 V , (111) 0=\alpha\dot{G}_{22}+\frac{1}{2}(\alpha+\Psi)(x^2-y^2)+\frac{1}{2}\mathcal{V}, \tag*{(111)} 0 = α G ˙ 22 + 2 1 ( α + Ψ ) ( x 2 − y 2 ) + 2 1 V , (111)
0 = α G ˙ 33 + 1 2 ( α + Ψ ) ( x 2 − y 2 ) + 1 2 V , (112) 0=\alpha\dot{G}_{33}+\frac{1}{2}(\alpha+\Psi)(x^2-y^2)+\frac{1}{2}\mathcal{V}, \tag*{(112)} 0 = α G ˙ 33 + 2 1 ( α + Ψ ) ( x 2 − y 2 ) + 2 1 V , (112)
0 = α G ˙ 01 + 1 2 ( α + Ψ ) x y . (113) 0=\alpha\dot{G}_{01}+\frac{1}{2}(\alpha+\Psi)xy. \tag*{(113)} 0 = α G ˙ 01 + 2 1 ( α + Ψ ) x y . (113)
Here G ˙ 01 = 0 \dot{G}_{01}=0 G ˙ 01 = 0 and G ˙ 22 = G ˙ 33 \dot{G}_{22}=\dot{G}_{33} G ˙ 22 = G ˙ 33 for a static spherical metric. The off-diagonal equation is a useful check, but the exceptional condition follows more generally from the unrestricted affine equation of § 7 , including when x y = 0 xy=0 x y = 0 .
Exceptional, null, and linear branches
For nonzero axial torsion, the full connection equation requires
α + Ψ ( T ) = α + β + 3 γ T 2 = 0. (114) \alpha+\Psi(T)=\alpha+\beta+3\gamma T^2=0. \tag*{(114)} α + Ψ ( T ) = α + β + 3 γ T 2 = 0. (114)
When γ ≠ 0 \gamma\ne0 γ = 0 , the real branches satisfy
T = T ∗ , T ∗ 2 = − α + β 3 γ ≥ 0 , y 2 − x 2 = c : = 2 T ∗ 3 . (115) T=T_\ast,\qquad T_\ast^2=-\frac{\alpha+\beta}{3\gamma}\ge0,\qquad y^2-x^2=c:=\frac{2T_\ast}{3}. \tag*{(115)} T = T ∗ , T ∗ 2 = − 3 γ α + β ≥ 0 , y 2 − x 2 = c := 3 2 T ∗ . (115)
On a connected region with T ∗ > 0 T_\ast>0 T ∗ > 0 , one may write
x = ϵ c sinh ζ ( r ) , y = ϵ c cosh ζ ( r ) , ϵ = ± 1. (116) x=\epsilon\sqrt{c}\sinh\zeta(r),\qquad y=\epsilon\sqrt{c}\cosh\zeta(r),\qquad\epsilon=\pm1. \tag*{(116)} x = ϵ c sinh ζ ( r ) , y = ϵ c cosh ζ ( r ) , ϵ = ± 1. (116)
For T ∗ < 0 T_\ast<0 T ∗ < 0 , writing d = − c > 0 d=-c>0 d = − c > 0 , a corresponding parametrization is
x = ϵ d cosh ζ ( r ) , y = ϵ d sinh ζ ( r ) . (117) x=\epsilon\sqrt{d}\cosh\zeta(r),\qquad y=\epsilon\sqrt{d}\sinh\zeta(r). \tag*{(117)} x = ϵ d cosh ζ ( r ) , y = ϵ d sinh ζ ( r ) . (117)
The function ζ ( r ) \zeta(r) ζ ( r ) remains algebraically undetermined. This is freedom in the axial connection, not by itself proof of a gauge symmetry or of a propagating mode.
For T ∗ = 0 T_\ast=0 T ∗ = 0 and γ ≠ 0 \gamma\ne0 γ = 0 , Eq. (114) requires α + β = 0 \alpha+\beta=0 α + β = 0 , and a nonzero null three-form may have y = x ≠ 0 y=x\ne0 y = x = 0 or y = − x ≠ 0 y=-x\ne0 y = − x = 0 .
If γ = 0 \gamma=0 γ = 0 and β = − α \beta=-\alpha β = − α , the axial affine equation is satisfied without imposing a constant norm: T ( r ) T(r) T ( r ) may vary. These branches should not be identified with the torsion-free solution.
Integration of the spherical metric equations
Assume α ≠ 0 \alpha\ne0 α = 0 and a nonlinear exceptional branch. Then Ψ ( T ∗ ) = − α \Psi(T_*)=-\alpha Ψ ( T ∗ ) = − α and V ( T ∗ ) = − 2 γ T ∗ 3 V(T_*)=-2\gamma T_*^3 V ( T ∗ ) = − 2 γ T ∗ 3 , so (108) becomes
G ˚ μ ν + Λ ∗ g μ ν = 0 , Λ ∗ : = γ T ∗ 3 α . (118) \mathring{G}_{\mu\nu}+\Lambda_*g_{\mu\nu}=0,\qquad\Lambda_*:=\frac{\gamma T_*^3}{\alpha}. \tag*{(118)} G ˚ μν + Λ ∗ g μν = 0 , Λ ∗ := α γ T ∗ 3 . (118)
Where the areal radius ρ = C ( r ) \rho=\sqrt{C(r)} ρ = C ( r ) is a valid local coordinate, introduce
d s 2 = e 2 Φ ( ρ ) F ( ρ ) d t 2 − d ρ 2 F ( ρ ) − ρ 2 d Ω 2 . (119) ds^2=e^{2\Phi(\rho)}F(\rho)dt^2-\frac{d\rho^2}{F(\rho)}-\rho^2d\Omega^2. \tag*{(119)} d s 2 = e 2Φ ( ρ ) F ( ρ ) d t 2 − F ( ρ ) d ρ 2 − ρ 2 d Ω 2 . (119)
On a nondegenerate static interval, the difference of the temporal and radial Einstein equations gives Φ ′ = 0 \Phi'=0 Φ ′ = 0 ; rescale t t t to set Φ = 0 \Phi=0 Φ = 0 . With the curvature convention of § 2 the remaining equations are
ρ F ′ + F − 1 ρ 2 = Λ ∗ , 1 2 F ′ ′ + F ′ ρ = Λ ∗ . (120) \frac{\rho F'+F-1}{\rho^2}=\Lambda_*,\qquad\frac{1}{2}F''+\frac{F'}{\rho}=\Lambda_*. \tag*{(120)} ρ 2 ρ F ′ + F − 1 = Λ ∗ , 2 1 F ′′ + ρ F ′ = Λ ∗ . (120)
The second equation follows by differentiating the first after multiplication by ρ 2 \rho^2 ρ 2 . Integration gives
F ( ρ ) = 1 − 2 M ρ + Λ ∗ 3 ρ 2 , R ˚ = 4 Λ ∗ . (121) F(\rho)=1-\frac{2M}{\rho}+\frac{\Lambda_*}{3}\rho^2,\qquad\mathring{R}=4\Lambda_*. \tag*{(121)} F ( ρ ) = 1 − ρ 2 M + 3 Λ ∗ ρ 2 , R ˚ = 4 Λ ∗ . (121)
Thus the metric is locally Schwarzschild–(anti-)de Sitter, while the connection may carry nonzero axial torsion. In the more common convention F = 1 − 2 M / ρ − Λ p h y s ρ 2 / 3 F=1-2M/\rho-\Lambda_{\mathrm{phys}}\rho^2/3 F = 1 − 2 M / ρ − Λ phys ρ 2 /3 , Λ p h y s = − Λ ∗ \Lambda_{\mathrm{phys}}=-\Lambda_* Λ phys = − Λ ∗ [30 , 31 , 32 ] . For the exceptional linear model γ = 0 \gamma=0 γ = 0 , β = − α \beta=-\alpha β = − α , Eq. (118) is replaced by G ˚ μ ν = 0 \mathring{G}_{\mu\nu}=0 G ˚ μν = 0 , without a fixed torsion norm.
Horizon-penetrating axial variables
The static coframe ceases to be regular at F ( ρ h ) = 0 F(\rho_h)=0 F ( ρ h ) = 0 . On a future-horizon patch define the advanced coordinate
v = t + ∫ d ρ F ( ρ ) , d s 2 = F ( ρ ) d v 2 − 2 d v d ρ − ρ 2 d Ω 2 . (122) v=t+\int\frac{d\rho}{F(\rho)},\qquad ds^2=F(\rho)dv^2-2\,dv\,d\rho-\rho^2d\Omega^2. \tag*{(122)} v = t + ∫ F ( ρ ) d ρ , d s 2 = F ( ρ ) d v 2 − 2 d v d ρ − ρ 2 d Ω 2 . (122)
Writing the same tensorial three-form as
τ = ρ 2 sin θ [ U ( ρ ) d v − V ( ρ ) d ρ ] ∧ d θ ∧ d ϕ , (123) \tau=\rho^2\sin\theta\,[U(\rho)dv-V(\rho)d\rho]\wedge d\theta\wedge d\phi, \tag*{(123)} τ = ρ 2 sin θ [ U ( ρ ) d v − V ( ρ ) d ρ ] ∧ d θ ∧ d ϕ , (123)
comparison with (101) in the static region gives
U = F x , V = x + y F , x = U F , y = F V − U F . (124) U=\sqrt{F}\,x,\qquad V=\frac{x+y}{\sqrt{F}},\qquad x=\frac{U}{\sqrt{F}},\qquad y=\sqrt{F}V-\frac{U}{\sqrt{F}}. \tag*{(124)} U = F x , V = F x + y , x = F U , y = F V − F U . (124)
Consequently, the constant-norm condition y 2 − x 2 = c y^2-x^2=c y 2 − x 2 = c becomes
F V 2 − 2 U V = c , T = 3 2 ( F V 2 − 2 U V ) . (125) FV^2-2UV=c,\qquad T=\frac{3}{2}(FV^2-2UV). \tag*{(125)} F V 2 − 2 U V = c , T = 2 3 ( F V 2 − 2 U V ) . (125)
This equation is regular at F = 0 F=0 F = 0 . If U ( ρ h ) > 0 U(\rho_h)>0 U ( ρ h ) > 0 , its horizon-regular root is
V = U − U 2 + c F F = − c U + U 2 + c F . (126) V=\frac{U-\sqrt{U^2+cF}}{F}=-\frac{c}{U+\sqrt{U^2+cF}}. \tag*{(126)} V = F U − U 2 + c F = − U + U 2 + c F c . (126)
In particular, the local choice U = a > 0 U=a>0 U = a > 0 gives
V = − c a + a 2 + c F , a 2 + c F > 0 , V ( ρ h ) = − c a . (127) V=-\frac{c}{a+\sqrt{a^2+cF}},\qquad a^2+cF>0,\qquad V(\rho_h)=-\frac{c}{a}. \tag*{(127)} V = − a + a 2 + c F c , a 2 + c F > 0 , V ( ρ h ) = − a c . (127)
The radicand is strictly positive at the horizon and therefore remains positive on a sufficiently small neighborhood. Since q μ ν q^{\mu\nu} q μν and τ \tau τ are smooth in the ingoing chart, the contorsion and the reconstructed affine connection (104) are smooth there. This establishes local future-horizon regularity ; it does not establish regularity at a bifurcation sphere, maximal extension, or global boundedness of affine curvature invariants [30 , 32 , 2 , 3 ] . The explicit affine curvature tests are deferred to § 9 and Appendix § 14 .
Classification of axial metric-compatible connections
We distinguish the classification of solutions within the metric-compatible, totally antisymmetric torsion sector from the classification of arbitrary metric–affine connections. The latter includes nonmetricity and non-axial torsion and is not established here. This distinction is essential because imposing geometric restrictions before varying the action can eliminate independent affine equations [2 , 3 , 28 , 29 ] .
Consider the vacuum model
S = 1 2 κ ∫ d 4 x − g F ( R , T ) , F ( R , T ) = α R + β T + γ T 3 , α ≠ 0. (128) S = \frac{1}{2\kappa}\int d^{4}x\sqrt{-g}\,F(R,T), \qquad F(R,T)=\alpha R+\beta T+\gamma T^{3}, \qquad\alpha\ne0. \tag*{(128)} S = 2 κ 1 ∫ d 4 x − g F ( R , T ) , F ( R , T ) = α R + β T + γ T 3 , α = 0. (128)
Here the curvature, torsion scalar, and connection follow the conventions fixed in § 2 . The independent metric and affine equations are
α R ( μ ν ) − 1 2 g μ ν F + ( β + 3 γ T 2 ) H μ ν = 0 , (129) \alpha R_{(\mu\nu)}-\frac{1}{2}g_{\mu\nu}F+(\beta+3\gamma T^{2})H_{\mu\nu}=0, \tag*{(129)} α R ( μν ) − 2 1 g μν F + ( β + 3 γ T 2 ) H μν = 0 , (129)
α P λ μ ν + ( β + 3 γ T 2 ) J λ μ ν = 0. (130) \alpha P_{\lambda}{}^{\mu\nu}+(\beta+3\gamma T^{2})J_{\lambda}{}^{\mu\nu}=0. \tag*{(130)} α P λ μν + ( β + 3 γ T 2 ) J λ μν = 0. (130)
Only after taking these unrestricted variations do we specialize to
Q λ μ ν = 0 , X λ μ ν = X [ λ μ ν ] , Γ λ μ ν = Γ ˚ λ μ ν + 1 2 X λ μ ν . (131) Q_{\lambda\mu\nu}=0,\qquad X_{\lambda\mu\nu}=X_{[\lambda\mu\nu]},\qquad\Gamma^{\lambda}{}_{\mu\nu}=\mathring{\Gamma}^{\lambda}{}_{\mu\nu}+\frac{1}{2}X^{\lambda}{}_{\mu\nu}. \tag*{(131)} Q λ μν = 0 , X λ μν = X [ λ μν ] , Γ λ μν = Γ ˚ λ μν + 2 1 X λ μν . (131)
The independently established identities in § 7 and Appendices B–C give
P λ μ ν = J λ μ ν = − X λ μ ν , R ( Γ ) = R ˚ + T , R ( μ ν ) ( Γ ) = R ˚ μ ν + H μ ν . (132) P_{\lambda}{}^{\mu\nu}=J_{\lambda}{}^{\mu\nu}=-X_{\lambda}{}^{\mu\nu},\qquad R(\Gamma)=\mathring{R}+T,\qquad R_{(\mu\nu)}(\Gamma)=\mathring{R}_{\mu\nu}+H_{\mu\nu}. \tag*{(132)} P λ μν = J λ μν = − X λ μν , R ( Γ ) = R ˚ + T , R ( μν ) ( Γ ) = R ˚ μν + H μν . (132)
Consequently the affine equation becomes
[ α + β + 3 γ T 2 ] X λ μ ν = 0. (133) [\alpha+\beta+3\gamma T^{2}]X_{\lambda}{}^{\mu\nu}=0. \tag*{(133)} [ α + β + 3 γ T 2 ] X λ μν = 0. (133)
Proposition (local classification in the axial sector). On a connected open region with a smooth Lorentzian metric and smooth metric-compatible axial torsion, the solutions of Eqs. (129) –(130) are described locally by the following alternatives.
(i) Torsion-free branch. If X = 0 X=0 X = 0 , the connection is Levi–Civita, T = 0 T=0 T = 0 , and G ˚ μ ν = 0 \mathring{G}_{\mu\nu}=0 G ˚ μν = 0 .
(ii) Nonzero-torsion nonlinear branch. If γ ≠ 0 \gamma\ne0 γ = 0 and X X X is nonzero on an open region, Eq. (133) implies
T 2 = − α + β 3 γ ≥ 0. (134) T^{2}=-\frac{\alpha+\beta}{3\gamma}\ge0. \tag*{(134)} T 2 = − 3 γ α + β ≥ 0. (134) For T ∗ ≠ 0 T_{\ast}\ne0 T ∗ = 0 , continuity fixes the sign of T = T ∗ T=T_{\ast} T = T ∗ on each connected nonzero-torsion region. The metric equation reduces to
G ˚ μ ν + Λ ∗ g μ ν = 0 , Λ ∗ = γ T ∗ 3 α . (135) \mathring{G}_{\mu\nu}+\Lambda_{\ast}g_{\mu\nu}=0,\qquad\Lambda_{\ast}=\frac{\gamma T_{\ast}^{3}}{\alpha}. \tag*{(135)} G ˚ μν + Λ ∗ g μν = 0 , Λ ∗ = α γ T ∗ 3 . (135) Conversely, any smooth Einstein metric satisfying Eq. (135) and any smooth axial three-form of constant scalar norm T ∗ T_{\ast} T ∗ define, through Eq. (131) , a solution of the unrestricted field equations. No spherical symmetry is needed for this local construction.
(iii) Null-norm nonlinear branch. When γ ≠ 0 \gamma\ne0 γ = 0 and α + β = 0 \alpha+\beta=0 α + β = 0 , a nonzero axial three-form may have T ∗ = 0 T_{\ast}=0 T ∗ = 0 . The metric is Ricci-flat, but X X X need not vanish. Thus a zero torsion scalar is not equivalent to zero torsion [1 , 3 ] .
(iv) Linear degenerate branch. When γ = 0 \gamma=0 γ = 0 and β = − α \beta=-\alpha β = − α , Eq. (133) is identically satisfied for every axial three-form, without a fixed-norm condition; the metric is Ricci-flat. For γ = 0 \gamma=0 γ = 0 and β ≠ − α \beta\ne-\alpha β = − α , Eq. (133) instead forces X = 0 X=0 X = 0 .
Proof. For X ≠ 0 X \ne0 X = 0 , Eq. (133) yields the exceptional condition. Using Eq. (132) , the metric equation is
α G ˚ μ ν + [ α + β + 3 γ T 2 ] H μ ν − 1 2 g μ ν [ ( α + β ) T + γ T 3 ] = 0. (136) \alpha\mathring{G}_{\mu\nu}+\left[\alpha+\beta+3\gamma T^2\right]H_{\mu\nu}-\frac{1}{2}g_{\mu\nu}\left[(\alpha+\beta)T+\gamma T^3\right]=0. \tag*{(136)} α G ˚ μν + [ α + β + 3 γ T 2 ] H μν − 2 1 g μν [ ( α + β ) T + γ T 3 ] = 0. (136)
For the nonlinear exceptional branch, α + β = − 3 γ T ∗ 2 \alpha+\beta=-3\gamma T_*^2 α + β = − 3 γ T ∗ 2 , hence the bracketed potential equals − 2 γ T ∗ 3 -2\gamma T_*^3 − 2 γ T ∗ 3 and Eq. (135) follows. For γ = 0 \gamma=0 γ = 0 , β = − α \beta=-\alpha β = − α , the potential vanishes identically. The torsion-free case follows directly from Eqs. (129) –(130) . Conversely, substitution into the unrestricted equations proves sufficiency. □
In four dimensions, the axial three-form is Hodge-dual to a vector: X μ ν ρ = ε μ ν ρ σ A σ X_{\mu\nu\rho}=\varepsilon_{\mu\nu\rho\sigma}A^\sigma X μν ρ = ε μν ρ σ A σ . For the signature ( + − − − ) (+---) ( + − − − ) and the orientation convention ε μ ν ρ σ ε μ ν ρ λ = − 6 δ σ λ \varepsilon_{\mu\nu\rho\sigma}\varepsilon^{\mu\nu\rho\lambda}=-6\delta_\sigma^\lambda ε μν ρ σ ε μν ρ λ = − 6 δ σ λ , one has
X λ μ ν X λ μ ν = − 6 A μ A μ , T = 3 2 A μ A μ . (137) X_{\lambda\mu\nu}X^{\lambda\mu\nu}=-6A_\mu A^\mu,\qquad T=\frac{3}{2}A_\mu A^\mu. \tag*{(137)} X λ μν X λ μν = − 6 A μ A μ , T = 2 3 A μ A μ . (137)
Thus the nonlinear axial classification is equivalently a local fixed-norm-vector problem on an Einstein geometry. This is a classification of solutions in the axial subsector , not of all inequivalent physical degrees of freedom. Global smoothness, boundary conditions, and possible gauge identifications remain separate questions [2 , 30 , 32 ] .
Local branches and their zero sets. The classification is pointwise and local. On an open subset where X ≠ 0 X\ne0 X = 0 , the nonlinear affine equation enforces a constant value of T T T selected by the coupling constants. Smooth solutions may possess zero sets of X X X , and the existence of a smooth continuation across such sets requires a separate analysis. In particular, a nonzero null three-form can have T = 0 T=0 T = 0 , so the condition T = 0 T=0 T = 0 must not be used as an abbreviation for vanishing torsion. The classification is not a classification modulo diffeomorphisms or any putative affine gauge symmetry [2 , 3 ] .
Scope beyond the axial sector. For a general affine connection, write Γ = Γ ˚ + N \Gamma=\mathring{\Gamma}+N Γ = Γ ˚ + N . The 64 components of N N N encode 24 torsion and 40 nonmetricity components. Since F R = α F_R=\alpha F R = α is constant, the Einstein–Hilbert contribution differs from its Levi–Civita form by a divergence and terms algebraic and quadratic in N N N ; the torsion potential is also algebraic. The general affine equations therefore constitute a nonlinear pointwise algebraic system. Its exhaustive classification requires the ranks and kernels of the unrestricted affine operator in the vector, axial, tensor-torsion, and nonmetricity sectors, followed by the full metric equations. Neither the proposition above nor the spherically symmetric examples establish that broader classification [2 , 5 , 6 ] .
Black-hole solutions, horizon regularity, and affine curvature
We investigate the metric and the independent Riemann–Cartan connection on the exceptional axial branch constructed in Secs. 7–8. The metric and affine field equations are understood as the unrestricted metric–affine equations, subsequently evaluated on the metric-compatible axial sector. We distinguish Levi–Civita invariants from invariants of the independent connection throughout [5 , 6 , 1 , 2 , 3 ] . We assume vacuum, α ≠ 0 \alpha\ne0 α = 0 , and the curvature convention fixed in § 2 .
Einstein metric, horizons, and surface gravity
For F ( R , T ) = α R + β T + γ T 3 F(R,T)=\alpha R+\beta T+\gamma T^3 F ( R , T ) = α R + β T + γ T 3 , the nonzero axial branch obeys
α + β + 3 γ T ∗ 2 = 0 , Λ ∗ : = γ T ∗ 3 α , G ˚ μ ν + Λ ∗ g μ ν = 0. (138) \alpha+\beta+3\gamma T_*^2=0,\qquad\Lambda_*:=\frac{\gamma T_*^3}{\alpha},\qquad\mathring{G}_{\mu\nu}+\Lambda_*g_{\mu\nu}=0. \tag*{(138)} α + β + 3 γ T ∗ 2 = 0 , Λ ∗ := α γ T ∗ 3 , G ˚ μν + Λ ∗ g μν = 0. (138)
For γ ≠ 0 \gamma\ne0 γ = 0 , real constant-norm branches require T ∗ 2 = − ( α + β ) / ( 3 γ ) ≥ 0 T_*^2=-(\alpha+\beta)/(3\gamma)\ge0 T ∗ 2 = − ( α + β ) / ( 3 γ ) ≥ 0 . The metric equation in the areal-radius gauge is
d s 2 = f ( r ) d t 2 − d r 2 f ( r ) − r 2 d Ω 2 , d Ω 2 = d θ 2 + sin 2 θ d ϕ 2 . (139) ds^2=f(r)\,dt^2-\frac{dr^2}{f(r)}-r^2d\Omega^2,\qquad d\Omega^2=d\theta^2+\sin^2\theta\,d\phi^2. \tag*{(139)} d s 2 = f ( r ) d t 2 − f ( r ) d r 2 − r 2 d Ω 2 , d Ω 2 = d θ 2 + sin 2 θ d ϕ 2 . (139)
With the curvature sign convention of the present work,
R ˚ = f ′ ′ + 4 f ′ r + 2 ( f − 1 ) r 2 , (140) \mathring{R}=f^{\prime\prime}+\frac{4f^{\prime}}{r}+\frac{2(f-1)}{r^2}, \tag*{(140)} R ˚ = f ′′ + r 4 f ′ + r 2 2 ( f − 1 ) , (140)
r f ′ + f − 1 r 2 = Λ ∗ , f ′ ′ 2 + f ′ r = Λ ∗ . (141) \frac{rf^{\prime}+f-1}{r^2}=\Lambda_\ast,\qquad\frac{f^{\prime\prime}}{2}+\frac{f^{\prime}}{r}=\Lambda_\ast. \tag*{(141)} r 2 r f ′ + f − 1 = Λ ∗ , 2 f ′′ + r f ′ = Λ ∗ . (141)
The second radial equation follows by differentiating ( r f ′ ) ′ = 1 + Λ ∗ r 2 (rf^{\prime})^{\prime}=1+\Lambda_\ast r^2 ( r f ′ ) ′ = 1 + Λ ∗ r 2 ; integration yields
f ( r ) = 1 − 2 M r + Λ ∗ r 2 3 , R ˚ = 4 Λ ∗ . (142) f(r)=1-\frac{2M}{r}+\frac{\Lambda_\ast r^2}{3},\qquad\mathring{R}=4\Lambda_\ast. \tag*{(142)} f ( r ) = 1 − r 2 M + 3 Λ ∗ r 2 , R ˚ = 4 Λ ∗ . (142)
This is the Schwarzschild–(anti-)de Sitter metric with the manuscript’s sign convention: the more usual parameter Λ p h y s \Lambda_{\mathrm{phys}} Λ phys in f = 1 − 2 M / r − Λ p h y s r 2 / 3 f=1-2M/r-\Lambda_{\mathrm{phys}}r^2/3 f = 1 − 2 M / r − Λ phys r 2 /3 equals − Λ ∗ -\Lambda_\ast − Λ ∗ . Birkhoff-type uniqueness applies to the Einstein metric, not to the independent affine connection [30 , 31 , 32 ] .
A nonzero Killing horizon r h r_h r h is a positive root of
Λ ∗ 3 r h 3 + r h − 2 M = 0 , f ′ ( r h ) = 1 r h + Λ ∗ r h . (143) \frac{\Lambda_\ast}{3}r_h^3+r_h-2M=0,\qquad f^{\prime}(r_h)=\frac{1}{r_h}+\Lambda_\ast r_h. \tag*{(143)} 3 Λ ∗ r h 3 + r h − 2 M = 0 , f ′ ( r h ) = r h 1 + Λ ∗ r h . (143)
For M > 0 M>0 M > 0 , Λ ∗ > 0 \Lambda_\ast>0 Λ ∗ > 0 gives one positive horizon, whereas Λ ∗ < 0 \Lambda_\ast<0 Λ ∗ < 0 gives two distinct positive horizons precisely when 0 < 9 M 2 ( − Λ ∗ ) < 1 0<9M^2(-\Lambda_\ast)<1 0 < 9 M 2 ( − Λ ∗ ) < 1 ; equality gives a degenerate horizon. For a nondegenerate horizon the magnitude of the metric Killing surface gravity is
κ h = 1 2 ∣ f ′ ( r h ) ∣ = 1 2 ∣ 1 r h + Λ ∗ r h ∣ . (144) \kappa_h=\frac{1}{2}\lvert f^{\prime}(r_h)\rvert=\frac{1}{2}\left\lvert\frac{1}{r_h}+\Lambda_\ast r_h\right\rvert. \tag*{(144)} κ h = 2 1 ∣ f ′ ( r h )∣ = 2 1 r h 1 + Λ ∗ r h . (144)
The absolute value is necessary at a cosmological horizon, where the radial derivative may be negative. The associated metric temperature, when the standard semiclassical interpretation is applicable, is T H = κ h / ( 2 π ) T_H=\kappa_h/(2\pi) T H = κ h / ( 2 π ) in units ℏ = k B = c = 1 \hbar=k_B=c=1 ℏ = k B = c = 1 . The behavior of spinful matter coupled to Γ \Gamma Γ is a separate question [30 , 32 , 2 ] .
Smooth axial torsion at a future Killing horizon
Introduce the advanced Eddington–Finkelstein coordinate v = t + ∫ d r / f ( r ) v=t+\int dr/f(r) v = t + ∫ d r / f ( r ) on a future-horizon patch. Then
d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 d Ω 2 . (145) ds^2=f(r)\,dv^2-2\,dv\,dr-r^2d\Omega^2. \tag*{(145)} d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 d Ω 2 . (145)
The rotation-invariant torsion three-form can be written
τ = 1 3 ! X μ ν ρ d x μ ∧ d x ν ∧ d x ρ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ . (146) \tau=\frac{1}{3!}X_{\mu\nu\rho}dx^\mu\wedge dx^\nu\wedge dx^\rho=r^2\sin\theta\,[U(r)\,dv-V(r)\,dr]\wedge d\theta\wedge d\phi. \tag*{(146)} τ = 3 ! 1 X μν ρ d x μ ∧ d x ν ∧ d x ρ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ . (146)
The area form sin θ d θ ∧ d ϕ \sin\theta\,d\theta\wedge d\phi sin θ d θ ∧ d ϕ is globally smooth on the symmetry sphere. Direct contraction gives
T = 3 2 ( f V 2 − 2 U V ) , f V 2 − 2 U V = c , c : = 2 T ∗ 3 . (147) T=\frac{3}{2}(fV^2-2UV),\qquad fV^2-2UV=c,\qquad c:=\frac{2T_\ast}{3}. \tag*{(147)} T = 2 3 ( f V 2 − 2 U V ) , f V 2 − 2 U V = c , c := 3 2 T ∗ . (147)
For any smooth U U U with U ( r h ) > 0 U(r_h)>0 U ( r h ) > 0 , the root regular at f ( r h ) = 0 f(r_h)=0 f ( r h ) = 0 is
V = U − U 2 + c f f = − c U + U 2 + c f , V ( r h ) = − c 2 U ( r h ) . (148) V=\frac{U-\sqrt{U^2+cf}}{f}=-\frac{c}{U+\sqrt{U^2+cf}},\qquad V(r_h)=-\frac{c}{2U(r_h)}. \tag*{(148)} V = f U − U 2 + c f = − U + U 2 + c f c , V ( r h ) = − 2 U ( r h ) c . (148)
A useful representative is
U = a > 0 , q ( r ) : = a 2 + c f ( r ) , V = − c a + q , q 2 > 0 , W : = f V − U = − q . (149) U=a>0,\qquad q(r):=\sqrt{a^2+cf(r)},\qquad V=-\frac{c}{a+q},\qquad q^2>0,\qquad W:=fV-U=-q. \tag*{(149)} U = a > 0 , q ( r ) := a 2 + c f ( r ) , V = − a + q c , q 2 > 0 , W := f V − U = − q . (149)
Since q ( r h ) = a > 0 q(r_h)=a>0 q ( r h ) = a > 0 , this is smooth on some neighborhood of every finite-radius horizon of the metric, including a degenerate one. The independent connection is
Γ μ ν λ = Γ ˚ μ ν λ + 1 2 g λ ρ X ρ μ ν . (150) \Gamma^\lambda_{\mu\nu}=\mathring{\Gamma}^\lambda_{\mu\nu}+\frac{1}{2}g^{\lambda\rho}X_{\rho\mu\nu}. \tag*{(150)} Γ μν λ = Γ ˚ μν λ + 2 1 g λ ρ X ρ μν . (150)
Smoothness of g g g and τ \tau τ in the regular chart implies smoothness of the contorsion and of Γ \Gamma Γ there. This proves local regularity at the selected future horizon, not smoothness at the bifurcation sphere or across all past horizons [30 , 32 , 31 ] . For c = 0 c=0 c = 0 , U = a U=a U = a , V = 0 V=0 V = 0 is a nonzero torsion three-form with vanishing quadratic torsion scalar; it is not the torsion-free limit.
Local regularity versus global extension. The Eddington–Finkelstein construction establishes smoothness of the metric and torsion three-form on a neighborhood of a selected future Killing horizon. It does not, by itself, establish smoothness at a bifurcation two-sphere, across a past horizon, or on a maximal extension. Moreover, bounded polynomial curvature invariants do not replace a componentwise test in a regular frame. These distinctions are standard in the analysis of black-hole extensions [30 , 32 ] .
Ricci tensor and curvature invariants
The Levi–Civita invariants are
R ˚ = 4 Λ ∗ , R ˚ μ ν R ˚ μ ν = 4 Λ ∗ 2 , K L C : = R ˚ α β μ ν R ˚ α β μ ν = 48 M 2 r 6 + 8 3 Λ ∗ 2 . (151) \mathring{R}=4\Lambda_{*},\qquad\mathring{R}_{\mu\nu}\mathring{R}^{\mu\nu}=4\Lambda_{*}^{2},\qquad K_{\mathrm{LC}}:=\mathring{R}_{\alpha\beta\mu\nu}\mathring{R}^{\alpha\beta\mu\nu}=\frac{48M^{2}}{r^{6}}+\frac{8}{3}\Lambda_{*}^{2}. \tag*{(151)} R ˚ = 4 Λ ∗ , R ˚ μν R ˚ μν = 4 Λ ∗ 2 , K LC := R ˚ α β μν R ˚ α β μν = r 6 48 M 2 + 3 8 Λ ∗ 2 . (151)
On the axial sector, the identities of § 7 give
R ( μ ν ) ( Γ ) = Λ ∗ g μ ν + H μ ν , R ( Γ ) = 4 Λ ∗ + T ∗ , H μ μ = T ∗ , H μ ν H μ ν = T ∗ 2 3 . (152) R_{(\mu\nu)}(\Gamma)=\Lambda_{*}g_{\mu\nu}+H_{\mu\nu},\qquad R(\Gamma)=4\Lambda_{*}+T_{*},\qquad H^{\mu}_{\ \mu}=T_{*},\qquad H_{\mu\nu}H^{\mu\nu}=\frac{T_{*}^{2}}{3}. \tag*{(152)} R ( μν ) ( Γ ) = Λ ∗ g μν + H μν , R ( Γ ) = 4 Λ ∗ + T ∗ , H μ μ = T ∗ , H μν H μν = 3 T ∗ 2 . (152)
Consequently,
R ( μ ν ) R ( μ ν ) = 4 Λ ∗ 2 + 2 Λ ∗ T ∗ + T ∗ 2 3 , X α β γ X α β γ = − 4 T ∗ . (153) R_{(\mu\nu)}R^{(\mu\nu)}=4\Lambda_{*}^{2}+2\Lambda_{*}T_{*}+\frac{T_{*}^{2}}{3},\qquad X_{\alpha\beta\gamma}X^{\alpha\beta\gamma}=-4T_{*}. \tag*{(153)} R ( μν ) R ( μν ) = 4 Λ ∗ 2 + 2 Λ ∗ T ∗ + 3 T ∗ 2 , X α β γ X α β γ = − 4 T ∗ . (153)
Define W = f V − U W=fV-U W = f V − U . Raising the indices of the three-form gives
X r θ ϕ = W r 2 sin θ , ∇ ˚ λ X λ θ ϕ = W ′ r 2 sin θ . (154) X^{r\theta\phi}=\frac{W}{r^{2}\sin\theta},\qquad\mathring{\nabla}_{\lambda}X^{\lambda\theta\phi}=\frac{W'}{r^{2}\sin\theta}. \tag*{(154)} X r θ ϕ = r 2 sin θ W , ∇ ˚ λ X λ θ ϕ = r 2 sin θ W ′ . (154)
The antisymmetric Ricci identity R [ μ ν ] = 1 2 ∇ ˚ λ X μ ν λ R_{[\mu\nu]}=\frac{1}{2}\mathring{\nabla}_{\lambda}X^{\lambda}_{\ \mu\nu} R [ μν ] = 2 1 ∇ ˚ λ X μν λ then implies
R [ θ ϕ ] = 1 2 r 2 sin θ W ′ , R [ v r ] = 0 , R [ μ ν ] R [ μ ν ] = 1 2 ( W ′ ) 2 . (155) R_{[\theta\phi]}=\frac{1}{2}r^{2}\sin\theta\,W',\qquad R_{[vr]}=0,\qquad R_{[\mu\nu]}R^{[\mu\nu]}=\frac{1}{2}(W')^{2}. \tag*{(155)} R [ θ ϕ ] = 2 1 r 2 sin θ W ′ , R [ v r ] = 0 , R [ μν ] R [ μν ] = 2 1 ( W ′ ) 2 . (155)
The symmetric–antisymmetric cross contraction vanishes, so
R μ ν ( Γ ) R μ ν ( Γ ) = 4 Λ ∗ 2 + 2 Λ ∗ T ∗ + T ∗ 2 3 + 1 2 [ f V − U ] ′ 2 . (156) \boxed{R_{\mu\nu}(\Gamma)R^{\mu\nu}(\Gamma)=4\Lambda_{*}^{2}+2\Lambda_{*}T_{*}+\frac{T_{*}^{2}}{3}+\frac{1}{2}\left[fV-U\right]^{\prime2}}. \tag*{(156)} R μν ( Γ ) R μν ( Γ ) = 4 Λ ∗ 2 + 2 Λ ∗ T ∗ + 3 T ∗ 2 + 2 1 [ f V − U ] ′2 . (156)
For (149) , W = − q W=-q W = − q and W ′ = − c f ′ / ( 2 q ) W'=-cf'/(2q) W ′ = − c f ′ / ( 2 q ) , yielding
R [ μ ν ] R [ μ ν ] = c 2 ( f ′ ) 2 8 ( a 2 + c f ) , f ′ = 2 M r 2 + 2 Λ ∗ r 3 . (157) R_{[\mu\nu]}R^{[\mu\nu]}=\frac{c^{2}(f')^{2}}{8(a^{2}+cf)},\qquad f'=\frac{2M}{r^{2}}+\frac{2\Lambda_{*}r}{3}. \tag*{(157)} R [ μν ] R [ μν ] = 8 ( a 2 + c f ) c 2 ( f ′ ) 2 , f ′ = r 2 2 M + 3 2 Λ ∗ r . (157)
At a horizon,
R [ μ ν ] R [ μ ν ] ∣ r = r h = c 2 8 a 2 ( 1 r h + Λ ∗ r h ) 2 . (158) \left.R_{[\mu\nu]}R^{[\mu\nu]}\right|_{r=r_{h}}=\frac{c^{2}}{8a^{2}}\left(\frac{1}{r_{h}}+\Lambda_{*}r_{h}\right)^{2}. \tag*{(158)} R [ μν ] R [ μν ] r = r h = 8 a 2 c 2 ( r h 1 + Λ ∗ r h ) 2 . (158)
This vanishes for a degenerate horizon, and also for the null-norm representative c = 0 c=0 c = 0 , without implying that its full affine curvature vanishes.
Full affine Kretschmann invariant, radial domains, and regular continuation
The affine curvature of a metric-compatible connection with totally antisymmetric torsion is not determined by its symmetric Ricci tensor. In particular, the full quadratic invariant must be constructed from the four-index curvature tensor, with the curvature and contraction conventions of § 2 [1 , 2 , 3 ] . Writing
D λ ρ μ ν : = R λ ρ μ ν ( Γ ) − R ˚ λ ρ μ ν , K Γ = K L C + 2 R ˚ λ ρ μ ν D λ ρ μ ν + D λ ρ μ ν D λ ρ μ ν , (159) D_{\lambda\rho\mu\nu} := R_{\lambda\rho\mu\nu}(\Gamma) - \mathring{R}_{\lambda\rho\mu\nu}, \qquad K_{\Gamma} = K_{\mathrm{LC}} + 2\mathring{R}_{\lambda\rho\mu\nu}D^{\lambda\rho\mu\nu} + D_{\lambda\rho\mu\nu}D^{\lambda\rho\mu\nu}, \tag*{(159)} D λ ρ μν := R λ ρ μν ( Γ ) − R ˚ λ ρ μν , K Γ = K LC + 2 R ˚ λ ρ μν D λ ρ μν + D λ ρ μν D λ ρ μν , (159)
we employ the ingoing Eddington–Finkelstein metric and axial three-form
d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 d Ω 2 , τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ . (160) ds^{2} = f(r)\,dv^{2} - 2\,dv\,dr - r^{2}d\Omega^{2}, \qquad\tau= r^{2}\sin\theta\left[U(r)\,dv - V(r)\,dr\right]\wedge d\theta\wedge d\phi. \tag*{(160)} d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 d Ω 2 , τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ . (160)
Here U , V U,V U , V are smooth radial functions wherever the representative is defined. The torsion scalar and the auxiliary function W W W are
T = 3 2 ( f V 2 − 2 U V ) , W : = f V − U . (161) T = \frac{3}{2}\left(fV^{2} - 2UV\right), \qquad W := fV - U. \tag*{(161)} T = 2 3 ( f V 2 − 2 U V ) , W := f V − U . (161)
Independent curvature contractions. Direct reconstruction of the full affine curvature using Γ μ ν λ = Γ ˚ μ ν λ + X μ ν λ / 2 \Gamma^{\lambda}_{\mu\nu} = \mathring{\Gamma}^{\lambda}_{\mu\nu} + X^{\lambda}_{\mu\nu}/2 Γ μν λ = Γ ˚ μν λ + X μν λ /2 gives the cross contraction
2 R ˚ λ ρ μ ν D λ ρ μ ν = 2 V ( f V − 2 U ) r 2 ( f + r f ′ − 1 ) . (162) 2\mathring{R}_{\lambda\rho\mu\nu}D^{\lambda\rho\mu\nu} = \frac{2V(fV - 2U)}{r^{2}}\left(f + rf' - 1\right). \tag*{(162)} 2 R ˚ λ ρ μν D λ ρ μν = r 2 2 V ( f V − 2 U ) ( f + r f ′ − 1 ) . (162)
For completeness, the distortion self-contraction is
D λ ρ μ ν D λ ρ μ ν = 1 4 [ 3 f 2 V 4 + 8 f 2 ( V ′ ) 2 + 8 f f ′ V V ′ − 12 f U V 3 − 16 f U ′ V ′ + 4 ( f ′ ) 2 V 2 + 8 f ′ U V ′ − 8 f ′ U ′ V + 12 U 2 V 2 − 4 ( U ′ ) 2 ] − 4 U U ′ r − 8 U 2 r 2 . (163) D_{\lambda\rho\mu\nu}D^{\lambda\rho\mu\nu} = \frac{1}{4}\left[3f^{2}V^{4} + 8f^{2}(V')^{2} + 8ff'VV' - 12fUV^{3} - 16fU'V' + 4(f')^{2}V^{2} + 8f'UV' - 8f'U'V + 12U^{2}V^{2} - 4(U')^{2}\right] - \frac{4UU'}{r} - \frac{8U^{2}}{r^{2}}. \tag*{(163)} D λ ρ μν D λ ρ μν = 4 1 [ 3 f 2 V 4 + 8 f 2 ( V ′ ) 2 + 8 f f ′ V V ′ − 12 f U V 3 − 16 f U ′ V ′ + 4 ( f ′ ) 2 V 2 + 8 f ′ U V ′ − 8 f ′ U ′ V + 12 U 2 V 2 − 4 ( U ′ ) 2 ] − r 4 U U ′ − r 2 8 U 2 . (163)
These expressions are to be read together with
K L C = ( f ′ ′ ) 2 + 4 ( f ′ ) 2 r 2 + 4 ( f − 1 ) 2 r 4 . (164) K_{\mathrm{LC}} = (f'')^{2} + \frac{4(f')^{2}}{r^{2}} + \frac{4(f-1)^{2}}{r^{4}}. \tag*{(164)} K LC = ( f ′′ ) 2 + r 2 4 ( f ′ ) 2 + r 4 4 ( f − 1 ) 2 . (164)
The contraction was independently compared at θ = π / 2 \theta= \pi/2 θ = π /2 and θ = π / 3 \theta= \pi/3 θ = π /3 before imposing the constant-norm constraint; both calculations yield the same polynomial. Angular independence also follows from spherical symmetry. This is a two-angle symbolic cross-check, not a claim of a separate arbitrary-angle symbolic reconstruction. The decomposition provides a direct reproducibility test for Appendix § 14 .
On the exceptional nonlinear branch, let
c : = 2 T ∗ 3 , f V 2 − 2 U V = c , f ( r ) = 1 − 2 M r + Λ ∗ r 2 3 , Λ ∗ : = α T ∗ 3 α . (165) c := \frac{2T_{*}}{3}, \qquad fV^{2} - 2UV = c, \qquad f(r) = 1 - \frac{2M}{r} + \frac{\Lambda_{*}r^{2}}{3}, \qquad\Lambda_{*} := \frac{\alpha T_{*}^{3}}{\alpha}. \tag*{(165)} c := 3 2 T ∗ , f V 2 − 2 U V = c , f ( r ) = 1 − r 2 M + 3 Λ ∗ r 2 , Λ ∗ := α α T ∗ 3 . (165)
Since ( f + r f ′ − 1 ) / r 2 = Λ ∗ (f + rf' - 1)/r^{2} = \Lambda_{*} ( f + r f ′ − 1 ) / r 2 = Λ ∗ , Eq. (162) simplifies exactly to
2 R ˚ ⋅ D = 2 c Λ ∗ . (166) 2\mathring{R}\cdot D = 2c\Lambda_{*}. \tag*{(166)} 2 R ˚ ⋅ D = 2 c Λ ∗ . (166)
For the horizon-regular constant-U U U representative
U = a > 0 , h ( r ) : = a 2 + c f ( r ) , q ( r ) : = h ( r ) , V = − c a + q , h > 0 , (167) U = a > 0, \qquad h(r) := a^{2} + cf(r), \qquad q(r) := \sqrt{h(r)}, \qquad V = -\frac{c}{a+q}, \qquad h > 0, \tag*{(167)} U = a > 0 , h ( r ) := a 2 + c f ( r ) , q ( r ) := h ( r ) , V = − a + q c , h > 0 , (167)
we have W = − q W=-q W = − q and V ′ = c 2 f ′ / [ 2 q ( a + q ) 2 ] V' = c^{2}f'/[2q(a+q)^{2}] V ′ = c 2 f ′ / [ 2 q ( a + q ) 2 ] . Substituting in Eq. (163) yields
D ⋅ D = c 2 ( f ′ ) 2 2 h + 2 c f ′ r + 3 c 2 4 + 2 c f − 8 a 2 − 2 c r 2 − 2 c Λ ∗ . (168) D\cdot D = \frac{c^{2}(f')^{2}}{2h} + \frac{2cf'}{r} + \frac{3c^{2}}{4} + \frac{2cf - 8a^{2} - 2c}{r^{2}} - 2c\Lambda_{*}. \tag*{(168)} D ⋅ D = 2 h c 2 ( f ′ ) 2 + r 2 c f ′ + 4 3 c 2 + r 2 2 c f − 8 a 2 − 2 c − 2 c Λ ∗ . (168)
Consequently, the full affine Kretschmann invariant is
K Γ = 48 M 2 r 6 + 8 Λ ∗ 2 3 + c 2 ( f ′ ) 2 2 ( a 2 + c f ) + 2 c f ′ r + 3 c 2 4 + 2 c f − 8 a 2 − 2 c r 2 . (169) K_{\Gamma} = \frac{48M^{2}}{r^{6}} + \frac{8\Lambda_{*}^{2}}{3} + \frac{c^{2}(f')^{2}}{2(a^{2}+cf)} + \frac{2cf'}{r} + \frac{3c^{2}}{4} + \frac{2cf - 8a^{2} - 2c}{r^{2}}. \tag*{(169)} K Γ = r 6 48 M 2 + 3 8 Λ ∗ 2 + 2 ( a 2 + c f ) c 2 ( f ′ ) 2 + r 2 c f ′ + 4 3 c 2 + r 2 2 c f − 8 a 2 − 2 c . (169)
This formula follows from the four-index contraction, not from a Ricci-only identity. Because the spacetime metric is Lorentzian, K Γ K_{\Gamma} K Γ is not positive definite; a negative value is not intrinsically inconsistent [30 , 2 ] . At any finite positive Killing horizon r h r_h r h with f ( r h ) = 0 f(r_h)=0 f ( r h ) = 0 , one has h ( r h ) = a 2 > 0 h(r_h)=a^2>0 h ( r h ) = a 2 > 0 , so g g g , τ \tau τ , Γ \Gamma Γ , and their curvature are smooth on a sufficiently small ingoing patch. In particular,
K Γ ( r h ) = K L C ( r h ) + c 2 ∣ f ′ ( r h ) ∣ 2 2 a 2 + 2 c f ′ ( r h ) r h + 3 c 2 4 − 8 a 2 + 2 c r h 2 . (170) K_{\Gamma}(r_h)=K_{\mathrm{LC}}(r_h)+\frac{c^2\lvert f'(r_h)\rvert^2}{2a^2}+\frac{2cf'(r_h)}{r_h}+\frac{3c^2}{4}-\frac{8a^2+2c}{r_h^2}. \tag*{(170)} K Γ ( r h ) = K LC ( r h ) + 2 a 2 c 2 ∣ f ′ ( r h ) ∣ 2 + r h 2 c f ′ ( r h ) + 4 3 c 2 − r h 2 8 a 2 + 2 c . (170)
This is future-horizon regularity in a horizon-penetrating chart, not regularity at a bifurcation sphere or a statement about a maximal extension [30 , 31 , 32 ] . The regularity statement is stronger than the finiteness of the displayed scalar invariant: the metric and axial three-form are smooth in the chosen advanced chart, and the reconstructed connection and curvature tensor are consequently smooth there. Conversely, finiteness of K Γ K_{\Gamma} K Γ alone would not establish regularity of all affine curvature components.
Positive zeros of the radicand. The real principal-root domain is D + : = { r > 0 : h ( r ) > 0 } \mathcal{D}_{+}:=\{r>0:h(r)>0\} D + := { r > 0 : h ( r ) > 0 } . If c ≠ 0 c\ne0 c = 0 , a positive zero is a positive root of
P ( r ) : = c Λ ∗ 3 r 3 + ( a 2 + c ) r − 2 M c = 0 , h ( r ) = P ( r ) r . (171) P(r):=\frac{c\Lambda_{*}}{3}r^3+(a^2+c)r-2Mc=0,\qquad h(r)=\frac{P(r)}{r}. \tag*{(171)} P ( r ) := 3 c Λ ∗ r 3 + ( a 2 + c ) r − 2 M c = 0 , h ( r ) = r P ( r ) . (171)
No such root is a Killing horizon, because h ( r h ) = a 2 h(r_h)=a^2 h ( r h ) = a 2 . If c = 0 c=0 c = 0 , h = a 2 h=a^2 h = a 2 has no zeros. For M > 0 M>0 M > 0 ,
h ′ = c ( 2 M r 2 + 2 Λ ∗ r 3 ) , h ′ ′ = c ( − 4 M r 3 + 2 Λ ∗ 3 ) . (172) h'=c\left(\frac{2M}{r^2}+\frac{2\Lambda_{*}r}{3}\right),\qquad h''=c\left(-\frac{4M}{r^3}+\frac{2\Lambda_{*}}{3}\right). \tag*{(172)} h ′ = c ( r 2 2 M + 3 2 Λ ∗ r ) , h ′′ = c ( − r 3 4 M + 3 2 Λ ∗ ) . (172)
If Λ ∗ ≥ 0 \Lambda_{*}\ge0 Λ ∗ ≥ 0 , h h h is monotone: for c > 0 c>0 c > 0 there is exactly one positive simple zero, with h > 0 h>0 h > 0 beyond it; for c < 0 c<0 c < 0 and Λ ∗ > 0 \Lambda_{*}>0 Λ ∗ > 0 there is exactly one positive simple zero, with h > 0 h>0 h > 0 below it. For Λ ∗ = 0 \Lambda_{*}=0 Λ ∗ = 0 and c < 0 c<0 c < 0 , a positive zero exists precisely if a 2 + c < 0 a^2+c<0 a 2 + c < 0 ; when a 2 + c ≥ 0 a^2+c\ge0 a 2 + c ≥ 0 , h > 0 h>0 h > 0 for every finite r > 0 r>0 r > 0 .
If Λ ∗ < 0 \Lambda_{*}<0 Λ ∗ < 0 , the unique positive stationary radius and the corresponding extremal value are
r m = ( − 3 M Λ ∗ ) 1 / 3 , h m = h ( r m ) = a 2 + c ( 1 + Λ ∗ r m 2 ) . (173) r_m=\left(-\frac{3M}{\Lambda_{*}}\right)^{1/3},\qquad h_m=h(r_m)=a^2+c(1+\Lambda_{*}r_m^2). \tag*{(173)} r m = ( − Λ ∗ 3 M ) 1/3 , h m = h ( r m ) = a 2 + c ( 1 + Λ ∗ r m 2 ) . (173)
For c > 0 c>0 c > 0 , h m h_m h m is a maximum: two positive simple zeros occur if h m > 0 h_m>0 h m > 0 , one double zero if h m = 0 h_m=0 h m = 0 , and no real principal-root domain if h m < 0 h_m<0 h m < 0 . For c < 0 c<0 c < 0 , h m h_m h m is a minimum: two positive simple zeros occur if h m < 0 h_m<0 h m < 0 , one double zero if h m = 0 h_m=0 h m = 0 , and no zeros if h m > 0 h_m>0 h m > 0 . The condition h m = 0 h_m=0 h m = 0 with c > 0 c>0 c > 0 is algebraically possible, but then h ≤ 0 h\le0 h ≤ 0 near r m r_m r m and does not give a real open neighborhood of the chosen representative.
Simple zeros and obstruction to regular extension. Let r 0 > 0 r_0>0 r 0 > 0 satisfy h ( r 0 ) = 0 h(r_0)=0 h ( r 0 ) = 0 and h ′ ( r 0 ) = c f ′ ( r 0 ) ≠ 0 h'(r_0)=cf'(r_0)\ne0 h ′ ( r 0 ) = c f ′ ( r 0 ) = 0 . On the admissible side h > 0 h>0 h > 0 ,
c 2 ( f ′ ) 2 2 h = c f ′ ( r 0 ) 2 ( r − r 0 ) + O ( 1 ) , r ⟶ r 0 . (174) \frac{c^2(f')^2}{2h}=\frac{cf'(r_0)}{2(r-r_0)}+O(1),\qquad r\longrightarrow r_0. \tag*{(174)} 2 h c 2 ( f ′ ) 2 = 2 ( r − r 0 ) c f ′ ( r 0 ) + O ( 1 ) , r ⟶ r 0 . (174)
The leading expression is positive on that side, while all remaining terms of Eq. (169) are finite at r 0 > 0 r_0>0 r 0 > 0 . Thus K Γ K_{\Gamma} K Γ diverges, ruling out a smooth finite-curvature extension of this constant-U U U representative through a simple radicand zero. It does not rule out other smooth axial profiles satisfying the same torsion-norm constraint.
Double zeros and a signed-root continuation. Assume M > 0 M>0 M > 0 , Λ ∗ < 0 \Lambda_{*}<0 Λ ∗ < 0 , c < 0 c<0 c < 0 , a > 0 a>0 a > 0 , and h ( r m ) = 0 h(r_m)=0 h ( r m ) = 0 . Then
f ′ ( r m ) = 0 , f ′ ′ ( r m ) = 2 Λ ∗ , c = − a 2 1 + Λ ∗ r m 2 < 0 , (175) f'(r_m)=0,\qquad f''(r_m)=2\Lambda_{*},\qquad c=-\frac{a^2}{1+\Lambda_{*}r_m^2}<0, \tag*{(175)} f ′ ( r m ) = 0 , f ′′ ( r m ) = 2 Λ ∗ , c = − 1 + Λ ∗ r m 2 a 2 < 0 , (175)
so 1 + Λ ∗ r m 2 > 0 1+\Lambda_{*}r_m^2>0 1 + Λ ∗ r m 2 > 0 . Taylor’s theorem factors the smooth function h h h locally as
h ( r ) = ( r − r m ) 2 A ( r ) , A ( r m ) = c Λ ∗ > 0 , s ( r ) : = ( r − r m ) A ( r ) , s 2 = h . (176) h(r)=(r-r_m)^2A(r),\qquad A(r_m)=c\Lambda_{*}>0,\qquad s(r):=(r-r_m)\sqrt{A(r)},\qquad s^2=h. \tag*{(176)} h ( r ) = ( r − r m ) 2 A ( r ) , A ( r m ) = c Λ ∗ > 0 , s ( r ) := ( r − r m ) A ( r ) , s 2 = h . (176)
After shrinking the neighborhood to ensure A > 0 A > 0 A > 0 and a + s ≠ 0 a + s \ne0 a + s = 0 , the function
V e x t ( r ) = − c a + s ( r ) (177) V_{\mathrm{ext}}(r) = -\frac{c}{a+s(r)} \tag*{(177)} V ext ( r ) = − a + s ( r ) c (177)
is smooth and satisfies f V e x t 2 − 2 a V e x t = c fV_{\mathrm{ext}}^2-2aV_{\mathrm{ext}}=c f V ext 2 − 2 a V ext = c . It agrees with the principal root on r > r m r>r_m r > r m and switches to the other algebraic root on r < r m r<r_m r < r m . The metric, torsion three-form, reconstructed connection, and curvature are consequently smooth across r m r_m r m in this local chart. The principal-root function alone has a cusp at r m r_m r m . In agreement with the smooth continuation,
lim r → r m c 2 ( f ′ ) 2 2 h = 2 c Λ ∗ . (178) \lim_{r\to r_m}\frac{c^2(f')^2}{2h}=2c\Lambda_\ast. \tag*{(178)} r → r m lim 2 h c 2 ( f ′ ) 2 = 2 c Λ ∗ . (178)
The finite scalar limit alone would not establish smoothness; the explicit signed-root construction supplies that stronger conclusion. No global completeness statement follows [30 , 32 ] .
For the nontrivial null-norm representative c = 0 c=0 c = 0 , U = a ≠ 0 U=a\ne0 U = a = 0 , V = 0 V=0 V = 0 , the torsion three-form is nonzero even though T = 0 T=0 T = 0 , and
K Γ = K L C − 8 a 2 r 2 , R [ μ ν ] R [ μ ν ] = 0. (179) K_{\Gamma}=K_{\mathrm{LC}}-\frac{8a^2}{r^2},\qquad R_{[\mu\nu]}R^{[\mu\nu]}=0. \tag*{(179)} K Γ = K LC − r 2 8 a 2 , R [ μν ] R [ μν ] = 0. (179)
Neither vanishing torsion norm nor vanishing antisymmetric Ricci invariant implies a torsion-free connection. For M ≠ 0 M\ne0 M = 0 , K L C K_{\mathrm{LC}} K LC diverges at r = 0 r=0 r = 0 , independently of the local horizon analysis.
Exceptional branches, profile dependence, and physical interpretation
The axial sector contains three distinct exceptional cases. First, for γ ≠ 0 \gamma\ne0 γ = 0 and T ∗ ≠ 0 T_\ast\ne0 T ∗ = 0 , real branches require
T ∗ 2 = − α + β 3 γ > 0 , Λ ∗ = γ T ∗ 3 α , α ≠ 0. (180) T_\ast^2=-\frac{\alpha+\beta}{3\gamma}>0,\qquad\Lambda_\ast=\frac{\gamma T_\ast^3}{\alpha},\qquad\alpha\ne0. \tag*{(180)} T ∗ 2 = − 3 γ α + β > 0 , Λ ∗ = α γ T ∗ 3 , α = 0. (180)
The metric is locally Schwarzschild–(anti-)de Sitter with the manuscript’s curvature convention, while the axial connection can retain nonzero torsion. Second, for γ ≠ 0 \gamma\ne0 γ = 0 , α + β = 0 \alpha+\beta=0 α + β = 0 , and T ∗ = 0 T_\ast=0 T ∗ = 0 , nonzero null axial torsion is allowed and the metric is Ricci-flat. Third, for γ = 0 \gamma=0 γ = 0 , β = − α \beta=-\alpha β = − α , the metric is Ricci-flat and the affine equation imposes no condition on the smooth axial torsion norm T ( x ) T(x) T ( x ) ; constant-norm profiles form only a subfamily. The torsion-free Levi–Civita Schwarzschild metric is also a vacuum solution of the polynomial model without exceptional tuning. These conclusions follow from the unrestricted metric and affine equations specialized after variation, as discussed in § 3 –§ 7 and the metric-affine literature [1 , 2 , 3 ] .
An invariant test of the residual axial freedom. On a constant-norm branch, W = f V − U W=fV-U W = f V − U obeys
W 2 = U 2 + c f , 2 W W ′ = 2 U U ′ + c f ′ . (181) W^2=U^2+cf,\qquad2WW'=2UU'+cf'. \tag*{(181)} W 2 = U 2 + c f , 2 W W ′ = 2 U U ′ + c f ′ . (181)
On any open interval where W ≠ 0 W\ne0 W = 0 , the antisymmetric affine Ricci invariant is
I A : = R [ μ ν ] ( Γ ) R [ μ ν ] ( Γ ) = 1 2 ( W ′ ) 2 = ( U U ′ + 1 2 c f ′ ) 2 2 ( U 2 + c f ) . (182) I_A:=R_{[\mu\nu]}(\Gamma)R^{[\mu\nu]}(\Gamma)=\frac{1}{2}(W')^2=\frac{\left(UU'+\frac{1}{2}cf'\right)^2}{2(U^2+cf)}. \tag*{(182)} I A := R [ μν ] ( Γ ) R [ μν ] ( Γ ) = 2 1 ( W ′ ) 2 = 2 ( U 2 + c f ) ( U U ′ + 2 1 c f ′ ) 2 . (182)
This formula extends the constant-U U U result of § 10.3 and is independent of the angular coordinates. At a finite horizon r h r_h r h , take two smooth representatives with the same metric, T ∗ T_\ast T ∗ , and U 1 ( r h ) = U 2 ( r h ) = a > 0 U_1(r_h)=U_2(r_h)=a>0 U 1 ( r h ) = U 2 ( r h ) = a > 0 , but with U 1 ′ ( r h ) = 0 U_1'(r_h)=0 U 1 ′ ( r h ) = 0 and U 2 ′ ( r h ) = b U_2'(r_h)=b U 2 ′ ( r h ) = b . Choosing the same smooth root W ( r h ) = − a W(r_h)=-a W ( r h ) = − a , their invariants satisfy
I A ( 1 ) ( r h ) = c 2 [ f ′ ( r h ) ] 2 8 a 2 , I A ( 2 ) ( r h ) = 1 2 ( b + c f ′ ( r h ) 2 a ) 2 . (183) I_A^{(1)}(r_h)=\frac{c^2[f'(r_h)]^2}{8a^2},\qquad I_A^{(2)}(r_h)=\frac{1}{2}\left(b+\frac{cf'(r_h)}{2a}\right)^2. \tag*{(183)} I A ( 1 ) ( r h ) = 8 a 2 c 2 [ f ′ ( r h ) ] 2 , I A ( 2 ) ( r h ) = 2 1 ( b + 2 a c f ′ ( r h ) ) 2 . (183)
The invariant I A I_A I A provides a sufficient test of geometric distinction when its values differ, but it is not a complete invariant of the spherical axial connection. Equality of I A I_A I A does not imply equality of the torsion profiles or of the full affine curvature tensor. These are unequal for generic b b b (although special choices may coincide). Thus the residual functional freedom is generally detectable by an affine scalar invariant, even when the metric and torsion norm agree. This proves geometric distinguishability for such profiles, not a propagating degree of freedom, a new conserved charge, or observable black-hole hair. The latter claims require an analysis of boundary conditions, allowed field redefinitions, matter couplings, and linearized dynamics [1 , 2 , 3 , 30 ] .
The results of § 10.4 further distinguish three notions: smoothness at a selected future Killing horizon, breakdown of a particular constant-U U U square-root representative at a radicand zero, and global extension of the metric-affine spacetime. The first is established locally; a simple radicand zero obstructs smooth extension of that representative, whereas a suitable double zero admits a signed-root continuation. Neither outcome settles the global classification of smooth torsion three-forms on maximally extended black-hole geometries [31 , 32 ] .
The constant-W W W axial family, Ricci symmetry, and affine curvature
We use the advanced Eddington–Finkelstein chart and the curvature convention of § 2 ,
d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 d Ω 2 , f ( r ) = 1 − 2 M r + Λ ∗ r 2 3 , τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ . (184) ds^2=f(r)\,dv^2-2\,dv\,dr-r^2d\Omega^2,\qquad f(r)=1-\frac{2M}{r}+\frac{\Lambda_*r^2}{3},\qquad\tau=r^2\sin\theta\,[U(r)\,dv-V(r)\,dr]\wedge d\theta\wedge d\phi. \tag*{(184)} d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 d Ω 2 , f ( r ) = 1 − r 2 M + 3 Λ ∗ r 2 , τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ . (184)
The metric-compatible connection is Γ μ ν λ = Γ ˚ μ ν λ + 1 2 g λ ρ X ρ μ ν \Gamma^\lambda_{\mu\nu}=\mathring{\Gamma}^\lambda_{\mu\nu}+\frac{1}{2}g^{\lambda\rho}X_{\rho\mu\nu} Γ μν λ = Γ ˚ μν λ + 2 1 g λ ρ X ρ μν , where τ = X λ μ ν d x λ ∧ d x μ ∧ d x ν / 3 ! \tau=X_{\lambda\mu\nu}dx^\lambda\wedge dx^\mu\wedge dx^\nu/3! τ = X λ μν d x λ ∧ d x μ ∧ d x ν /3 ! . In the exceptional nonlinear axial sector,
α + β + 3 γ T ∗ 2 = 0 , Λ ∗ = γ T ∗ 3 α , c = 2 T ∗ 3 , f V 2 − 2 U V = c . (185) \alpha+\beta+3\gamma T_*^2=0,\qquad\Lambda_*=\frac{\gamma T_*^3}{\alpha},\qquad c=\frac{2T_*}{3},\qquad fV^2-2UV=c. \tag*{(185)} α + β + 3 γ T ∗ 2 = 0 , Λ ∗ = α γ T ∗ 3 , c = 3 2 T ∗ , f V 2 − 2 U V = c . (185)
No affine-flatness condition is imposed [2 , 3 , 1 ] . Set W = f V − U W=fV-U W = f V − U . Then W 2 = U 2 + c f W^2=U^2+cf W 2 = U 2 + c f .
Local classification of Ricci-symmetric spherical axial connections. For arbitrary differentiable U , V U,V U , V in this ansatz, the antisymmetric Ricci tensor has the sole independent component
R [ θ ϕ ] = r 2 sin θ 2 ( f V − U ) ′ = r 2 sin θ 2 W ′ , R [ μ ν ] R [ μ ν ] = 1 2 ( W ′ ) 2 . (186) R_{[\theta\phi]}=\frac{r^2\sin\theta}{2}(fV-U)'=\frac{r^2\sin\theta}{2}W',\qquad R_{[\mu\nu]}R^{[\mu\nu]}=\frac{1}{2}(W')^2. \tag*{(186)} R [ θ ϕ ] = 2 r 2 sin θ ( f V − U ) ′ = 2 r 2 sin θ W ′ , R [ μν ] R [ μν ] = 2 1 ( W ′ ) 2 . (186)
Consequently, on a connected radial interval I ⊂ ( 0 , ∞ ) I\subset(0,\infty) I ⊂ ( 0 , ∞ ) ,
R [ μ ν ] = 0 ⟺ W ′ = 0 ⟺ W = W 0 (constant on I ) . (187) R_{[\mu\nu]}=0\Longleftrightarrow W'=0\Longleftrightarrow W=W_0\ \text{(constant on }I\text{)}. \tag*{(187)} R [ μν ] = 0 ⟺ W ′ = 0 ⟺ W = W 0 (constant on I ) . (187)
This is a necessary and sufficient statement within the spherical metric-compatible axial ansatz , not a classification of unrestricted affine connections. For constant W 0 W_0 W 0 , the norm condition gives
U 2 = W 0 2 − c f , U = σ W 0 2 − c f , V = − c U − W 0 , σ = ± 1 , (188) U^2=W_0^2-cf,\qquad U=\sigma\sqrt{W_0^2-cf},\qquad V=-\frac{c}{U-W_0},\qquad\sigma=\pm1, \tag*{(188)} U 2 = W 0 2 − c f , U = σ W 0 2 − c f , V = − U − W 0 c , σ = ± 1 , (188)
where the radicand is positive and U − W 0 ≠ 0 U-W_0\ne0 U − W 0 = 0 . The rationalized formula follows from ( U − W 0 ) ( U + W 0 ) = − c f (U-W_0)(U+W_0)=-cf ( U − W 0 ) ( U + W 0 ) = − c f and extends smoothly through a zero of f f f whenever its denominator is nonzero. If U − W 0 = 0 U-W_0=0 U − W 0 = 0 at a point, one must instead use the original constraint and analyze the branch separately; the division in (188) is not valid there. The parametrization (188) describes the regular nondegenerate branches of the Ricci-symmetric spherical axial sector. It is not a global parametrization across zeros of its denominator or across points where the square-root branches cease to be differentiable. The degenerate algebraic cases must be analyzed directly from the original norm constraint and the defining relation for W W W .
We select W 0 = − a W_0=-a W 0 = − a with a > 0 a>0 a > 0 and the positive root
u = a 2 − c f , U = u , V = c a + u , a 2 − c f > 0. (189) u=\sqrt{a^2-cf},\qquad U=u,\qquad V=\frac{c}{a+u},\qquad a^2-cf>0. \tag*{(189)} u = a 2 − c f , U = u , V = a + u c , a 2 − c f > 0. (189)
Direct substitution gives f V = u − a fV=u-a f V = u − a , W = − a W=-a W = − a , and f V 2 − 2 U V = c fV^2-2UV=c f V 2 − 2 U V = c . Thus T = 3 c / 2 = T ∗ T=3c/2=T_* T = 3 c /2 = T ∗ exactly and the local existence proposition of § 7.7 verifies the full field equations, not merely the spherical reduced action [28 , 29 ] .
Degenerate denominators and the null-norm limit. The rationalized expression for V V V in Eq. (188) is valid only when U − W 0 ≠ 0 U-W_0 \ne0 U − W 0 = 0 . To analyze its complement without dividing by a potentially vanishing factor, retain the original relations
f V = U + W 0 , f V 2 − 2 U V = c , ( U − W 0 ) ( U + W 0 ) = − c f . (190) fV=U+W_0,\qquad fV^2-2UV=c,\qquad(U-W_0)(U+W_0)=-cf. \tag*{(190)} f V = U + W 0 , f V 2 − 2 U V = c , ( U − W 0 ) ( U + W 0 ) = − c f . (190)
Suppose U U U , V V V , f f f are finite at a point r 0 > 0 r_0>0 r 0 > 0 and W = W 0 W=W_0 W = W 0 is constant. If U ( r 0 ) = W 0 U(r_0)=W_0 U ( r 0 ) = W 0 and c ≠ 0 c\ne0 c = 0 , the last identity forces f ( r 0 ) = 0 f(r_0)=0 f ( r 0 ) = 0 . The first identity then gives 2 W 0 = 0 2W_0=0 2 W 0 = 0 , so U ( r 0 ) = W 0 = 0 U(r_0)=W_0=0 U ( r 0 ) = W 0 = 0 ; the second identity would imply c = 0 c=0 c = 0 , a contradiction. Thus no smooth, finite-profile constant-W W W solution with c ≠ 0 c\ne0 c = 0 can attain U = W 0 U=W_0 U = W 0 . This is a local regularity statement for the chosen advanced-coordinate ansatz, not an obstruction to every possible coordinate representation of an affine connection [30 , 32 ] .
For c = 0 c=0 c = 0 , the equations instead factorize as
V ( f V − 2 U ) = 0 , f V − U = W 0 . (191) V(fV-2U)=0,\qquad fV-U=W_0. \tag*{(191)} V ( f V − 2 U ) = 0 , f V − U = W 0 . (191)
On an interval with f ≠ 0 f\ne0 f = 0 , the two algebraic branches are
V = 0 , U = − W 0 , (192) V=0,\qquad U=-W_0, \tag*{(192)} V = 0 , U = − W 0 , (192)
U = W 0 , V = 2 W 0 f . (193) U=W_0,\qquad V=\frac{2W_0}{f}. \tag*{(193)} U = W 0 , V = f 2 W 0 . (193)
The second branch is generally singular at a zero of f f f when W 0 ≠ 0 W_0\ne0 W 0 = 0 ; its apparent singularity must be assessed through the three-form rather than through the scalar norm alone.
If W 0 = 0 W_0=0 W 0 = 0 , the branches coincide with U = V = 0 U=V=0 U = V = 0 wherever f ≠ 0 f\ne0 f = 0 ; at an isolated horizon the continuity of smooth profiles excludes assigning independent nonzero values solely at that point. In particular, the nontrivial smooth null-norm representative U = a U=a U = a , V = 0 V=0 V = 0 belongs to W 0 = − a W_0=-a W 0 = − a , not to the singular branch U = W 0 U=W_0 U = W 0 .
For W 0 ≠ 0 W_0\ne0 W 0 = 0 , the second representative does not extend as a smooth three-form through a zero of f f f in the regular advanced Eddington–Finkelstein chart: its coefficient V V V diverges while d r ∧ d θ ∧ d ϕ dr\wedge d\theta\wedge d\phi d r ∧ d θ ∧ d ϕ is a smooth, nonvanishing coordinate three-form away from the polar coordinate singularities. This is a failure of regularity of that representative, not merely a pole introduced by rationalizing the norm constraint. The first representative, by contrast, has finite constant coefficients in the same chart. These statements are local and do not preclude the use of other horizon-adapted charts or other torsion profiles [30 , 32 ] .
Derivative and invariant checks. Differentiating u 2 = a 2 − c f u^2=a^2-cf u 2 = a 2 − c f gives
U ′ = − c f ′ 2 u , U ′ ′ = − c f ′ ′ 2 u − c 2 ( f ′ ) 2 4 u 3 . (194) U'=-\frac{cf'}{2u},\qquad U''=-\frac{cf''}{2u}-\frac{c^2(f')^2}{4u^3}. \tag*{(194)} U ′ = − 2 u c f ′ , U ′′ = − 2 u c f ′′ − 4 u 3 c 2 ( f ′ ) 2 . (194)
V ′ = − c 2 f ′ 2 u ( a + u ) 2 , V ′ ′ = − c 2 f ′ ′ 2 u ( a + u ) 2 − c 3 ( f ′ ) 2 ( a + 3 u ) 4 u 3 ( a + u ) 3 . (195) V'=-\frac{c^2f'}{2u(a+u)^2},\qquad V''=-\frac{c^2f''}{2u(a+u)^2}-\frac{c^3(f')^2(a+3u)}{4u^3(a+u)^3}. \tag*{(195)} V ′ = − 2 u ( a + u ) 2 c 2 f ′ , V ′′ = − 2 u ( a + u ) 2 c 2 f ′′ − 4 u 3 ( a + u ) 3 c 3 ( f ′ ) 2 ( a + 3 u ) . (195)
The identities 2 U U ′ = − c f ′ 2UU'=-cf' 2 U U ′ = − c f ′ and 2 ( U ′ ) 2 + 2 U U ′ ′ = − c f ′ ′ 2(U')^2+2UU''=-cf'' 2 ( U ′ ) 2 + 2 U U ′′ = − c f ′′ provide independent differentiation checks. Since W ′ = 0 W'=0 W ′ = 0 , the affine Ricci tensor is symmetric, with
R μ ν ( Γ ) = Λ ∗ g μ ν + H μ ν , R ( Γ ) = 4 Λ ∗ + T ∗ , (196) R_{\mu\nu}(\Gamma)=\Lambda_*g_{\mu\nu}+H_{\mu\nu},\qquad R(\Gamma)=4\Lambda_*+T_*, \tag*{(196)} R μν ( Γ ) = Λ ∗ g μν + H μν , R ( Γ ) = 4 Λ ∗ + T ∗ , (196)
R μ ν R μ ν = 4 Λ ∗ 2 + 2 Λ ∗ T ∗ + T ∗ 2 3 , X λ μ X λ μ = − 4 T ∗ . (197) R_{\mu\nu}R^{\mu\nu}=4\Lambda_*^2+2\Lambda_*T_*+\frac{T_*^2}{3},\qquad X_{\lambda\mu}X^{\lambda\mu}=-4T_*. \tag*{(197)} R μν R μν = 4 Λ ∗ 2 + 2 Λ ∗ T ∗ + 3 T ∗ 2 , X λ μ X λ μ = − 4 T ∗ . (197)
These contractions follow from H μ μ = T ∗ H^\mu{}_\mu=T_* H μ μ = T ∗ and H μ ν H μ ν = T ∗ 2 / 3 H_{\mu\nu}H^{\mu\nu}=T_*^2/3 H μν H μν = T ∗ 2 /3 ; their independence of a a a does not imply that the full affine Riemann tensor is independent of a a a [2 , 3 ] .
Four-index curvature and reproducible algebraic reduction. Write D = R ( Γ ) − R ˚ D=R(\Gamma)-\mathring{R} D = R ( Γ ) − R ˚ at the four-index level, and K Γ = R λ ρ μ ν ( Γ ) R λ ρ μ ν ( Γ ) K_{\Gamma}=R_{\lambda\rho\mu\nu}(\Gamma)R^{\lambda\rho\mu\nu}(\Gamma) K Γ = R λ ρ μν ( Γ ) R λ ρ μν ( Γ ) . Appendix § 14 .§ 3 gives
K Γ = K L C + 2 R ˚ ⋅ D + D ⋅ D , K L C = 48 M 2 r 6 + 8 Λ ∗ 2 3 . (198) K_{\Gamma}=K_{\mathrm{LC}}+2\mathring{R}\cdot D+D\cdot D,\qquad K_{\mathrm{LC}}=\frac{48M^2}{r^6}+\frac{8\Lambda_*^2}{3}. \tag*{(198)} K Γ = K LC + 2 R ˚ ⋅ D + D ⋅ D , K LC = r 6 48 M 2 + 3 8 Λ ∗ 2 . (198)
For f V 2 − 2 U V = c fV^2-2UV=c f V 2 − 2 U V = c and f + r f ′ − 1 = Λ ∗ r 2 f+rf'-1=\Lambda_*r^2 f + r f ′ − 1 = Λ ∗ r 2 ,
2 R ˚ ⋅ D = 2 V ( f V − 2 U ) r 2 ( f + r f ′ − 1 ) = 2 c Λ ∗ . (199) 2\mathring{R}\cdot D=\frac{2V(fV-2U)}{r^2}(f+rf'-1)=2c\Lambda_*. \tag*{(199)} 2 R ˚ ⋅ D = r 2 2 V ( f V − 2 U ) ( f + r f ′ − 1 ) = 2 c Λ ∗ . (199)
In the notation of Appendix § 14 .§ 3 , define
P = 3 f 2 V 4 + 8 f 2 ( V ′ ) 2 + 8 f f ′ V V ′ − 12 f U V 3 − 16 f U ′ V ′ P=3f^2V^4+8f^2(V')^2+8ff'VV'-12fUV^3-16fU'V' P = 3 f 2 V 4 + 8 f 2 ( V ′ ) 2 + 8 f f ′ V V ′ − 12 f U V 3 − 16 f U ′ V ′
+ 4 ( f ′ ) 2 V 2 + 8 f ′ U V ′ − 8 f ′ U ′ V − 4 ( U ′ ) 2 + 12 U 2 V 2 . (200) {}+4(f')^2V^2+8f'UV'-8f'U'V-4(U')^2+12U^2V^2. \tag*{(200)} + 4 ( f ′ ) 2 V 2 + 8 f ′ U V ′ − 8 f ′ U ′ V − 4 ( U ′ ) 2 + 12 U 2 V 2 . (200)
P 4 = 3 c 2 4 − 3 c 2 ( f ′ ) 2 4 u 2 , − 4 U U ′ r = 2 c f ′ r . (201) \frac{P}{4}=\frac{3c^2}{4}-\frac{3c^2(f')^2}{4u^2},\qquad-\frac{4UU'}{r}=\frac{2cf'}{r}. \tag*{(201)} 4 P = 4 3 c 2 − 4 u 2 3 c 2 ( f ′ ) 2 , − r 4 U U ′ = r 2 c f ′ . (201)
Substituting (189) , its derivatives, and f = ( a 2 − u 2 ) / c f=(a^2-u^2)/c f = ( a 2 − u 2 ) / c (for c ≠ 0 c\ne0 c = 0 ), yields the exact identities
Both identities extend continuously to c = 0 c=0 c = 0 in the selected family. In particular, the first was checked by exact symbolic polynomial simplification without imposing the Einstein equation for f f f ; this checks the substitution into the generic curvature formula, not a separate reconstruction of all four-index components. The resulting affine Kretschmann contraction is
K Γ ( W ) = K L C + 2 c Λ ∗ + 3 c 2 4 − 3 c 2 ( f ′ ) 2 4 ( a 2 − c f ) + 2 c f ′ r − 8 ( a 2 − c f ) r 2 . (202) K_{\Gamma}^{(W)}=K_{\mathrm{LC}}+2c\Lambda_*+\frac{3c^2}{4}-\frac{3c^2(f')^2}{4(a^2-cf)}+\frac{2cf'}{r}-\frac{8(a^2-cf)}{r^2}. \tag*{(202)} K Γ ( W ) = K LC + 2 c Λ ∗ + 4 3 c 2 − 4 ( a 2 − c f ) 3 c 2 ( f ′ ) 2 + r 2 c f ′ − r 2 8 ( a 2 − c f ) . (202)
Its sign is not constrained to be positive in Lorentzian signature [30 , 3 ] .
Comparison with the constant-U U U representative. For U = a U=a U = a and V = − c / [ a + a 2 + c f ] V=-c/[a+\sqrt{a^2+cf}] V = − c / [ a + a 2 + c f ] , Appendix § 14 .§ 4 yields
K Γ ( U ) = K L C + c 2 ( f ′ ) 2 2 ( a 2 + c f ) + 2 c f ′ r + 3 c 2 4 + 2 c f − 8 a 2 − 2 c r 2 . (203) K_{\Gamma}^{(U)}=K_{\mathrm{LC}}+\frac{c^2(f')^2}{2(a^2+cf)}+\frac{2cf'}{r}+\frac{3c^2}{4}+\frac{2cf-8a^2-2c}{r^2}. \tag*{(203)} K Γ ( U ) = K LC + 2 ( a 2 + c f ) c 2 ( f ′ ) 2 + r 2 c f ′ + 4 3 c 2 + r 2 2 c f − 8 a 2 − 2 c . (203)
The common principal-root domain is
D c o m m o n = { r > 0 : ∣ c f ( r ) ∣ < a 2 } . (204) \mathcal{D}_{\mathrm{common}}=\{r>0:|cf(r)|<a^2\}. \tag*{(204)} D common = { r > 0 : ∣ c f ( r ) ∣ < a 2 } . (204)
Subtracting the two expressions gives
Δ K : = K Γ ( W ) − K Γ ( U ) = 2 c Λ ∗ + 6 c f + 2 c r 2 − c 2 ( f ′ ) 2 4 ( 3 a 2 − c f + 2 a 2 + c f ) . (205) \Delta K:=K_{\Gamma}^{(W)}-K_{\Gamma}^{(U)}=2c\Lambda_*+\frac{6cf+2c}{r^2}-\frac{c^2(f')^2}{4}\left(\frac{3}{a^2-cf}+\frac{2}{a^2+cf}\right). \tag*{(205)} Δ K := K Γ ( W ) − K Γ ( U ) = 2 c Λ ∗ + r 2 6 c f + 2 c − 4 c 2 ( f ′ ) 2 ( a 2 − c f 3 + a 2 + c f 2 ) . (205)
Moreover,
I A ( W ) = 0 , I A ( U ) = c 2 ( f ′ ) 2 8 ( a 2 + c f ) . (206) I_A^{(W)}=0,\qquad I_A^{(U)}=\frac{c^2(f')^2}{8(a^2+cf)}. \tag*{(206)} I A ( W ) = 0 , I A ( U ) = 8 ( a 2 + c f ) c 2 ( f ′ ) 2 . (206)
Thus for c ≠ 0 c\ne0 c = 0 and f ′ ≠ 0 f'\ne0 f ′ = 0 the two connections are invariantly distinct. An isolated zero of Δ K \Delta K Δ K does not imply equality of the connections. If c = 0 c=0 c = 0 , the selected representatives coincide, U = a U=a U = a , V = 0 V=0 V = 0 , and K Γ = K L C − 8 a 2 / r 2 K_{\Gamma}=K_{\mathrm{LC}}-8a^2/r^2 K Γ = K LC − 8 a 2 / r 2 , although the axial torsion three-form is nonzero for a > 0 a>0 a > 0 .
Horizon regularity and limitations. At a finite future Killing horizon f ( r h ) = 0 f(r_h)=0 f ( r h ) = 0 , assuming finite f ′ ( r h ) f'(r_h) f ′ ( r h ) and a > 0 a>0 a > 0 ,
U h = a , V h = − c 2 a , U h ′ = − c f h ′ 2 a , V h ′ = − c 2 f h ′ 8 a 3 . (207) U_h=a,\qquad V_h=-\frac{c}{2a},\qquad U'_h=-\frac{cf'_h}{2a},\qquad V'_h=-\frac{c^2f'_h}{8a^3}. \tag*{(207)} U h = a , V h = − 2 a c , U h ′ = − 2 a c f h ′ , V h ′ = − 8 a 3 c 2 f h ′ . (207)
Therefore
K Π ( W ) ( r h ) = K L C ( r h ) + 2 c Λ ∗ + 3 c 2 4 − 3 c 2 ( f h ′ ) 2 4 a 2 + 2 c f h ′ r h − 8 a 2 r h 2 , (208) K_{\Pi}^{(W)}(r_h)=K_{\mathrm{LC}}(r_h)+2c\Lambda_*+\frac{3c^2}{4}-\frac{3c^2(f'_h)^2}{4a^2}+\frac{2cf'_h}{r_h}-\frac{8a^2}{r_h^2}, \tag*{(208)} K Π ( W ) ( r h ) = K LC ( r h ) + 2 c Λ ∗ + 4 3 c 2 − 4 a 2 3 c 2 ( f h ′ ) 2 + r h 2 c f h ′ − r h 2 8 a 2 , (208)
Δ K ( r h ) = 2 c Λ ∗ + 2 c r h 2 − 5 c 2 ( f h ′ ) 2 4 a 2 . (209) \Delta K(r_h)=2c\Lambda_*+\frac{2c}{r_h^2}-\frac{5c^2(f'_h)^2}{4a^2}. \tag*{(209)} Δ K ( r h ) = 2 c Λ ∗ + r h 2 2 c − 4 a 2 5 c 2 ( f h ′ ) 2 . (209)
The strict inequality a 2 − c f ( r h ) = a 2 > 0 a^2-cf(r_h)=a^2>0 a 2 − c f ( r h ) = a 2 > 0 ensures a smooth local extension of U , V U,V U , V , the torsion three-form, and the reconstructed connection in this chart. This is a local future-horizon statement; it does not establish regularity at r = 0 r=0 r = 0 , at a radicand zero, on a bifurcation sphere, or throughout a maximal extension [30 , 31 , 32 ] . Different affine invariants distinguish geometries, but their physical observability requires specifying how matter couples to the independent connection [2 , 1 ] .
Comparison of horizon jets. The constant-U U U and constant-W W W representatives have the same zeroth-order torsion data at a finite Killing horizon r h r_h r h , f ( r h ) = 0 f(r_h)=0 f ( r h ) = 0 , but their first radial derivatives generally differ. Writing Δ F : = F ( W ) − F ( U ) \Delta F:=F^{(W)}-F^{(U)} Δ F := F ( W ) − F ( U ) , direct differentiation of U ( U ) = a U^{(U)}=a U ( U ) = a , U ( W ) = a 2 − c f U^{(W)}=\sqrt{a^2-cf} U ( W ) = a 2 − c f and their corresponding rationalized V V V profiles yields
Δ U h = Δ V h = 0 , Δ U h ′ = − c f h ′ 2 a , Δ V h ′ = − c 2 f h ′ 4 a 3 . (210) \Delta U_h=\Delta V_h=0,\qquad\Delta U'_h=-\frac{cf'_h}{2a},\qquad\Delta V'_h=-\frac{c^2f'_h}{4a^3}. \tag*{(210)} Δ U h = Δ V h = 0 , Δ U h ′ = − 2 a c f h ′ , Δ V h ′ = − 4 a 3 c 2 f h ′ . (210)
Hence the two torsion three-forms agree at the horizon but, for c f h ′ ≠ 0 cf'_h\ne0 c f h ′ = 0 , have distinct first radial jets in the same smooth advanced Eddington–Finkelstein chart. This provides a local component-level explanation for why their affine curvatures can differ even where the torsion profiles themselves coincide. Such a comparison does not, by itself, establish a coordinate-independent classification of the connections; the scalar invariant R [ μ ν ] R [ μ ν ] R_{[\mu\nu]}R^{[\mu\nu]} R [ μν ] R [ μν ] supplies an invariant distinction where it is nonzero [2 , 3 ] .
If the horizon is degenerate, f h = f h ′ = 0 f_h=f'_h=0 f h = f h ′ = 0 , the first-jet differences vanish, whereas the second derivatives satisfy
Δ U h ′ ′ = − c f h ′ ′ 2 a , Δ V h ′ ′ = − c 2 f h ′ ′ 4 a 3 . (211) \Delta U''_h=-\frac{cf''_h}{2a},\qquad\Delta V''_h=-\frac{c^2f''_h}{4a^3}. \tag*{(211)} Δ U h ′′ = − 2 a c f h ′′ , Δ V h ′′ = − 4 a 3 c 2 f h ′′ . (211)
These identities require a > 0 a>0 a > 0 and twice-differentiable f f f and profiles. The existence of a degenerate horizon is an additional restriction on the metric parameters, not a generic feature of the Schwarzschild–(anti-)de Sitter family [30 , 32 ] .
Actual extremality conditions for the Einstein metric. The degenerate-horizon limit is further constrained by the explicit Schwarzschild–(anti-)de Sitter lapse,
f ( r ) = 1 − 2 M r + Λ ∗ r 2 3 . (212) f(r)=1-\frac{2M}{r}+\frac{\Lambda_*r^2}{3}. \tag*{(212)} f ( r ) = 1 − r 2 M + 3 Λ ∗ r 2 . (212)
At a finite positive double root r h r_h r h , the conditions f ( r h ) = f ′ ( r h ) = 0 f(r_h)=f'(r_h)=0 f ( r h ) = f ′ ( r h ) = 0 imply
Λ ∗ = − 1 r h 2 , M = r h 3 , f ′ ′ ( r h ) = − 2 r h 2 . (213) \Lambda_*=-\frac{1}{r_h^2},\qquad M=\frac{r_h}{3},\qquad f''(r_h)=-\frac{2}{r_h^2}. \tag*{(213)} Λ ∗ = − r h 2 1 , M = 3 r h , f ′′ ( r h ) = − r h 2 2 . (213)
Substitution into Eq. (209) yields the stronger on-shell identity
Δ K ( r h ) = 0 at an actual double root of f . (214) \Delta K(r_h)=0\qquad\text{at an actual double root of }f. \tag*{(214)} Δ K ( r h ) = 0 at an actual double root of f . (214)
Coincidence of curvature data at the double horizon. At an actual double root r h r_h r h of the Einstein lapse, the two representatives satisfy
Δ U h = Δ V h = Δ U h ′ = Δ V h ′ = 0 , Δ F : = F ( W ) − F ( U ) . (215) \Delta U_h=\Delta V_h=\Delta U'_h=\Delta V'_h=0,\qquad\Delta F:=F^{(W)}-F^{(U)}. \tag*{(215)} Δ U h = Δ V h = Δ U h ′ = Δ V h ′ = 0 , Δ F := F ( W ) − F ( U ) . (215)
Because the metric is common to both solutions and the distortion K μ ν λ = 1 2 g λ ρ X ρ μ ν K^\lambda_{\mu\nu} = \frac{1}{2}g^{\lambda\rho}X_{\rho\mu\nu} K μν λ = 2 1 g λ ρ X ρ μν depends algebraically on the torsion three-form, these equalities imply
Γ μ ν ( W ) λ ∣ r h = Γ μ ν ( U ) λ ∣ r h , R ρ μ ν ( W ) λ ∣ r h = R ρ μ ν ( U ) λ ∣ r h . (216) \left.\Gamma^{(W)\lambda}_{\mu\nu}\right|_{r_h} = \left.\Gamma^{(U)\lambda}_{\mu\nu}\right|_{r_h}, \qquad\left.R^{(W)\lambda}_{\ \ \rho\mu\nu}\right|_{r_h} = \left.R^{(U)\lambda}_{\ \ \rho\mu\nu}\right|_{r_h}. \tag*{(216)} Γ μν ( W ) λ r h = Γ μν ( U ) λ r h , R ρ μν ( W ) λ r h = R ρ μν ( U ) λ r h . (216)
Here the comparison is made in the same smooth advanced Eddington–Finkelstein chart. In particular, Eq. (206) gives
I A ( W ) ( r h ) = I A ( U ) ( r h ) = 0 , (217) I_A^{(W)}(r_h) = I_A^{(U)}(r_h) = 0, \tag*{(217)} I A ( W ) ( r h ) = I A ( U ) ( r h ) = 0 , (217)
in agreement with Δ K ( r h ) = 0 \Delta K(r_h)=0 Δ K ( r h ) = 0 . These pointwise coincidences do not extend to an open neighborhood in general: the nonzero second-jet differences in Eq. (211) can distinguish the radial derivatives of the affine curvature, and the invariant I A ( U ) I_A^{(U)} I A ( U ) is generically nonzero away from the double horizon.
Nevertheless, Eq. (211) gives, for c ≠ 0 c \ne0 c = 0 ,
Δ U h ′ ′ = c a r h 2 , Δ V h ′ ′ = c 2 2 a 3 r h 2 , (218) \Delta U_h^{\prime\prime} = \frac{c}{a r_h^2}, \qquad\Delta V_h^{\prime\prime} = \frac{c^2}{2a^3 r_h^2}, \tag*{(218)} Δ U h ′′ = a r h 2 c , Δ V h ′′ = 2 a 3 r h 2 c 2 , (218)
which need not vanish. Although equality of a single scalar curvature invariant does not generally imply equality of the full curvature tensors, the stronger pointwise equality established in Eq. (216) follows here from the coincidence of the torsion profiles and their first radial derivatives. The nonvanishing second-jet differences in Eq. (218) show that this pointwise coincidence need not extend to a neighborhood of the degenerate horizon. The double-root conditions also require M > 0 M > 0 M > 0 and Λ ∗ < 0 \Lambda_* < 0 Λ ∗ < 0 for r h > 0 r_h > 0 r h > 0 in the present sign convention; Λ p h y s = − Λ ∗ > 0 \Lambda_{\mathrm{phys}} = -\Lambda_* > 0 Λ phys = − Λ ∗ > 0 in the conventional notation [30 , 32 ] .
The distinction between the two families is not merely a consequence of their different radial parametrizations. On any open region where c f ′ ≠ 0 cf' \ne0 c f ′ = 0 , the constant-U U U representative has a nonvanishing antisymmetric affine Ricci tensor, whereas the constant-W W W representative has R [ μ ν ] = 0 R_{[\mu\nu]} = 0 R [ μν ] = 0 identically. Hence the two affine geometries cannot be related by a diffeomorphism that identifies their full curvature tensors on such a region. This conclusion concerns geometric inequivalence and does not, by itself, establish observational distinguishability.
Comparison of the two horizon-regular representatives. The two representatives have the same metric and constant torsion scalar, but their affine Ricci symmetry properties differ. The constant-W W W representative satisfies R [ μ ν ] = 0 R_{[\mu\nu]} = 0 R [ μν ] = 0 , whereas the constant-U U U representative generally has R [ μ ν ] ≠ 0 R_{[\mu\nu]} \ne0 R [ μν ] = 0 when c f ′ ≠ 0 cf' \ne0 c f ′ = 0 . Their torsion three-forms agree at a selected horizon, although their radial jets need not agree. At an actual double root of the Einstein lapse the first jets and the displayed Kretschmann values agree, but the second jets may differ. These observations concern local geometric data and do not alone establish physical inequivalence of the solutions [2 , 3 ] .
Perturbative degeneracy and the unrestricted metric–affine stability problem
We separate a pointwise algebraic degeneracy in the restricted axial sector from physical propagating modes and black-hole stability. Let
g μ ν = g ˉ μ ν + ϵ h μ ν + O ( ϵ 2 ) , Γ μ ν λ = Γ ˉ μ ν λ + ϵ C μ ν λ + O ( ϵ 2 ) . (219) g_{\mu\nu} = \bar{g}_{\mu\nu} + \epsilon h_{\mu\nu} + O(\epsilon^2), \qquad\Gamma^\lambda_{\mu\nu} = \bar{\Gamma}^\lambda_{\mu\nu} + \epsilon C^\lambda_{\mu\nu} + O(\epsilon^2). \tag*{(219)} g μν = g ˉ μν + ϵ h μν + O ( ϵ 2 ) , Γ μν λ = Γ ˉ μν λ + ϵ C μν λ + O ( ϵ 2 ) . (219)
The unrestricted variables comprise ten metric and 64 ordered connection components; fixed-metric axial variations occupy only four algebraic directions [2 , 3 ] .
Restricted axial quadratic form and its rank
On the metric-compatible axial submanifold of field space, R ( Γ ) = R ˚ + T R(\Gamma)=\mathring{R}+T R ( Γ ) = R ˚ + T and the restricted functional is
S a x = 1 2 κ ∫ d 4 x − g [ R ˚ + V ( T ) ] , V ( T ) = ( α + β ) T + γ T 3 . (220) S_{\mathrm{ax}}=\frac{1}{2\kappa}\int d^{4}x\sqrt{-g}\left[\mathring{R}+\mathcal{V}(T)\right],\qquad\mathcal{V}(T)=(\alpha+\beta)T+\gamma T^{3}. \tag*{(220)} S ax = 2 κ 1 ∫ d 4 x − g [ R ˚ + V ( T ) ] , V ( T ) = ( α + β ) T + γ T 3 . (220)
This identity is suitable for computing a restricted Hessian , not for replacing the unrestricted affine variation [28 , 29 ] . On the exceptional branch, V ′ ( T ∗ ) = 0 \mathcal{V}'(T_{*})=0 V ′ ( T ∗ ) = 0 and V ′ ′ ( T ∗ ) = 6 γ T ∗ \mathcal{V}''(T_{*})=6\gamma T_{*} V ′′ ( T ∗ ) = 6 γ T ∗ . At fixed metric, write X = X ˉ + ϵ δ X X=\bar{X}+\epsilon\delta X X = X ˉ + ϵδ X and
T = T ∗ + ϵ t 1 + ϵ 2 t 2 , t 1 = − 1 2 X ˉ λ μ ν δ X λ μ ν , t 2 = − 1 4 δ X λ μ ν δ X λ μ ν . (221) T=T_{*}+\epsilon t_{1}+\epsilon^{2}t_{2},\qquad t_{1}=-\frac{1}{2}\bar{X}_{\lambda\mu\nu}\delta X^{\lambda\mu\nu},\qquad t_{2}=-\frac{1}{4}\delta X_{\lambda\mu\nu}\delta X^{\lambda\mu\nu}. \tag*{(221)} T = T ∗ + ϵ t 1 + ϵ 2 t 2 , t 1 = − 2 1 X ˉ λ μν δ X λ μν , t 2 = − 4 1 δ X λ μν δ X λ μν . (221)
The exact cubic Taylor expansion is
V ( T ) = V ( T ∗ ) + 3 γ T ∗ ϵ 2 t 1 2 + ϵ 3 ( 6 γ T ∗ t 1 t 2 + γ t 1 3 ) + ϵ 4 ( 3 γ T ∗ t 2 2 + 3 γ t 1 2 t 2 ) + 3 γ ϵ 5 t 1 t 2 2 + γ ϵ 6 t 2 3 . (222) \mathcal{V}(T)=\mathcal{V}(T_{*})+3\gamma T_{*}\epsilon^{2}t_{1}^{2}+\epsilon^{3}\left(6\gamma T_{*}t_{1}t_{2}+\gamma t_{1}^{3}\right)+\epsilon^{4}\left(3\gamma T_{*}t_{2}^{2}+3\gamma t_{1}^{2}t_{2}\right)+3\gamma\epsilon^{5}t_{1}t_{2}^{2}+\gamma\epsilon^{6}t_{2}^{3}. \tag*{(222)} V ( T ) = V ( T ∗ ) + 3 γ T ∗ ϵ 2 t 1 2 + ϵ 3 ( 6 γ T ∗ t 1 t 2 + γ t 1 3 ) + ϵ 4 ( 3 γ T ∗ t 2 2 + 3 γ t 1 2 t 2 ) + 3 γ ϵ 5 t 1 t 2 2 + γ ϵ 6 t 2 3 . (222)
Here ϵ 2 \epsilon^{2} ϵ 2 denotes the quadratic coefficient, not the second derivative of the action. The fixed-metric axial quadratic coefficient is
S a x [ 2 ] = 3 γ T ∗ 2 κ ∫ d 4 x − g ˉ t 1 2 . (223) S_{\mathrm{ax}}^{[2]}=\frac{3\gamma T_{*}}{2\kappa}\int d^{4}x\sqrt{-\bar{g}}\,t_{1}^{2}. \tag*{(223)} S ax [ 2 ] = 2 κ 3 γ T ∗ ∫ d 4 x − g ˉ t 1 2 . (223)
The bilinear pairing on Λ 3 T p ∗ M \Lambda^{3}T_{p}^{*}M Λ 3 T p ∗ M is nondegenerate, so for X ˉ ≠ 0 \bar{X}\ne0 X ˉ = 0 the linear functional δ X ↦ t 1 \delta X\mapsto t_{1} δ X ↦ t 1 is nonzero even when X ˉ \bar{X} X ˉ is null. Consequently, for γ T ∗ ≠ 0 \gamma T_{*}\ne0 γ T ∗ = 0 the pointwise algebraic Hessian has rank one and a three-dimensional kernel,
rank H a x i a l = 1 , ker H a x i a l = { δ X : t 1 = 0 } , dim ker H a x i a l = 3. (224) \operatorname{rank}\mathcal{H}_{\mathrm{axial}}=1,\qquad\ker\mathcal{H}_{\mathrm{axial}}=\{\delta X:t_{1}=0\},\qquad\dim\ker\mathcal{H}_{\mathrm{axial}}=3. \tag*{(224)} rank H axial = 1 , ker H axial = { δ X : t 1 = 0 } , dim ker H axial = 3. (224)
For the spherical ( δ U , δ V ) (\delta U,\delta V) ( δ U , δ V ) subspace, the generic rank is one and the kernel has dimension one. These counts are not counts of physical degrees of freedom.
The rank statement in Eq. (224) refers exclusively to the four-dimensional vector space of axial three-form variations at fixed background metric. It is not the rank of the unrestricted connection Hessian, nor of the coupled metric–connection quadratic action. In particular, metric variations contribute to the first-order variation of the torsion scalar and must be retained in the complete linearized constraint system.
Linearized constraints and the null-norm branch
The exact axial affine equation is
V ′ ( T ) X μ ν λ = ( α + β + 3 γ T 2 ) X μ ν λ = 0. (225) \mathcal{V}'(T)X^{\lambda}_{\ \mu\nu}=(\alpha+\beta+3\gamma T^{2})X^{\lambda}_{\ \mu\nu}=0. \tag*{(225)} V ′ ( T ) X μν λ = ( α + β + 3 γ T 2 ) X μν λ = 0. (225)
At X ˉ ≠ 0 \bar{X}\ne0 X ˉ = 0 , its first variation on a non-null exceptional background is
6 γ T ∗ δ T X ˉ μ ν λ = 0 , ⟹ δ T = 0 ( γ T ∗ ≠ 0 ) . (226) 6\gamma T_{*}\,\delta T\,\bar{X}^{\lambda}_{\ \mu\nu}=0,\qquad\Longrightarrow\qquad\delta T=0\quad(\gamma T_{*}\ne0). \tag*{(226)} 6 γ T ∗ δ T X ˉ μν λ = 0 , ⟹ δ T = 0 ( γ T ∗ = 0 ) . (226)
When the metric also varies, the scalar variation is
δ T = H ˉ μ ν δ g μ ν − 1 2 X ˉ λ μ ν δ X λ μ ν ∣ g ˉ , δ g μ ν = − g ˉ μ ρ g ˉ ν σ h ρ σ . (227) \delta T=\bar{H}_{\mu\nu}\delta g^{\mu\nu}-\frac{1}{2}\bar{X}_{\lambda\mu\nu}\delta X^{\lambda\mu\nu}\big|_{\bar{g}},\qquad\delta g^{\mu\nu}=-\bar{g}^{\mu\rho}\bar{g}^{\nu\sigma}h_{\rho\sigma}. \tag*{(227)} δ T = H ˉ μν δ g μν − 2 1 X ˉ λ μν δ X λ μν g ˉ , δ g μν = − g ˉ μ ρ g ˉ ν σ h ρ σ . (227)
The notation ∣ g ˉ \big|_{\bar{g}} g ˉ indicates that only the independent mixed-index torsion is varied in that term. Any exact smooth family of nonzero-torsion exceptional solutions with fixed couplings and γ ≠ 0 \gamma\ne0 γ = 0 satisfies T = T ∗ T=T_{*} T = T ∗ pointwise; thus δ T = 0 \delta T=0 δ T = 0 is a tangent constraint, not a propagating field equation.
Tangent variations of the exceptional norm constraint. For a one-parameter family of exact exceptional solutions with fixed couplings, γ T ∗ ≠ 0 \gamma T_* \ne0 γ T ∗ = 0 and nonzero background axial torsion, T ( ϵ ) = T ∗ T(\epsilon)=T_* T ( ϵ ) = T ∗ implies
0 = δ T = H ˉ μ ν δ g μ ν + J ˉ λ μ ν δ Γ λ μ ν , (228) 0=\delta T=\bar{H}_{\mu\nu}\,\delta g^{\mu\nu}+\bar{J}_{\lambda}{}^{\mu\nu}\,\delta\Gamma^{\lambda}{}_{\mu\nu}, \tag*{(228)} 0 = δ T = H ˉ μν δ g μν + J ˉ λ μν δ Γ λ μν , (228)
where H ˉ \bar{H} H ˉ and J ˉ \bar{J} J ˉ are the derivatives of the unrestricted torsion scalar evaluated on the background. For γ T ∗ ≠ 0 \gamma T_* \ne0 γ T ∗ = 0 and nonzero background axial torsion, the linearized exceptional affine equation imposes δ T = 0 \delta T=0 δ T = 0 on perturbations satisfying Eq. (228) . The same condition is necessary for tangent vectors to smooth families of exact exceptional solutions. It is not an off-shell restriction on arbitrary metric–affine trial variations, and it is not sufficient to solve the remaining linearized field equations or to guarantee integrability to an exact solution family. At T ∗ = 0 T_*=0 T ∗ = 0 , the first-order affine equation does not enforce δ T = 0 \delta T=0 δ T = 0 ; the nonlinear constraint must then be analyzed at higher orders. These distinctions prevent identifying a restricted Hessian kernel with physical gauge or propagating degrees of freedom [2 , 3 ] .
Null-norm exceptional branch. For the null-norm exceptional branch T ∗ = 0 T_*=0 T ∗ = 0 , α + β = 0 \alpha+\beta=0 α + β = 0 and γ ≠ 0 \gamma\ne0 γ = 0 . The restricted quadratic potential vanishes, whereas for a generic fixed-metric variation
V ( T ) = γ ϵ 3 t 1 3 + 3 γ ϵ 4 t 1 2 t 2 + 3 γ ϵ 5 t 1 t 2 2 + γ ϵ 6 t 2 3 . (229) \mathcal{V}(T)=\gamma\epsilon^3 t_1^3+3\gamma\epsilon^4 t_1^2t_2+3\gamma\epsilon^5t_1t_2^2+\gamma\epsilon^6t_2^3. \tag*{(229)} V ( T ) = γ ϵ 3 t 1 3 + 3 γ ϵ 4 t 1 2 t 2 + 3 γ ϵ 5 t 1 t 2 2 + γ ϵ 6 t 2 3 . (229)
The affine equation begins at order ϵ 2 \epsilon^2 ϵ 2 :
V ′ ( T ) X λ μ ν = 3 γ ϵ 2 t 1 2 X ˉ λ μ ν + O ( ϵ 3 ) . (230) \mathcal{V}'(T)X^\lambda{}_{\mu\nu}=3\gamma\epsilon^2t_1^2\bar{X}^\lambda{}_{\mu\nu}+O(\epsilon^3). \tag*{(230)} V ′ ( T ) X λ μν = 3 γ ϵ 2 t 1 2 X ˉ λ μν + O ( ϵ 3 ) . (230)
Therefore its linearization alone imposes no norm constraint, while an exact family of solutions with nonzero torsion must still satisfy T = 0 T=0 T = 0 . In particular, if t 1 = 0 t_1=0 t 1 = 0 along a linear trial path, the first nonzero restricted potential term may be of order ϵ 6 \epsilon^6 ϵ 6 , namely γ ϵ 6 t 2 3 \gamma\epsilon^6t_2^3 γ ϵ 6 t 2 3 ; for an exact solution path, higher-order corrections to X X X can enforce T = 0 T=0 T = 0 order by order. This is a nonlinear constraint/linearization issue, not by itself a demonstration of dynamical strong coupling or instability. The absence of a first-order constraint at T ∗ = 0 T_*=0 T ∗ = 0 reflects a failure of the linearized equation to capture the full nonlinear constraint. It does not imply that arbitrary first-order torsion perturbations can be integrated into nearby exact solutions. Such integrability requires the higher-order field equations, including the metric and non-axial affine variations, to be satisfied. The functional freedom of the exact exceptional solutions and the kernel of the restricted axial Hessian are related manifestations of the degenerate axial field equations, but they are not equivalent characterizations of physical degrees of freedom. Establishing such an interpretation requires the complete constrained metric-affine perturbation theory.
What is required for unrestricted stability.
The full second variation must retain h μ ν h_{\mu\nu} h μν and all C λ μ ν C^\lambda{}_{\mu\nu} C λ μν , including nonmetricity and non-axial torsion. Its block form may be denoted schematically by
S [ 2 ] = 1 2 ∫ d 4 x − g ( h C ) ( K h h K h C K C h K C C ) ( h C ) . (231) S^{[2]}=\frac{1}{2}\int d^4x\sqrt{-g}\,
\begin{pmatrix}h & C\end{pmatrix}
\begin{pmatrix}
\mathcal{K}_{hh} & \mathcal{K}_{hC}\\
\mathcal{K}_{Ch} & \mathcal{K}_{CC}
\end{pmatrix}
\begin{pmatrix}h\\C\end{pmatrix}. \tag*{(231)} S [ 2 ] = 2 1 ∫ d 4 x − g ( h C ) ( K hh K C h K h C K C C ) ( h C ) . (231)
The blocks are differential/algebraic operators with appropriate index contractions, not a computed finite-dimensional kinetic matrix. Because F R = α F_R=\alpha F R = α is constant, the curvature term is linear in affine curvature and the independent distortion is algebraic in the bulk after separating a divergence; nonetheless, its constraint matrix can be singular and can mix with metric perturbations. A kernel of K C C \mathcal{K}_{CC} K C C does not by itself establish a gauge invariance. In particular, possible projective transformations must be tested against the complete F ( R , T ) F(R,T) F ( R , T ) action, since the torsion scalar generally changes under such transformations [2 , 3 ] .
Algebraic affine sector and unresolved mode count. Because F R = α F_R=\alpha F R = α is constant, the distortion decomposition Γ = Γ ˚ + N \Gamma=\mathring{\Gamma}+N Γ = Γ ˚ + N reduces the bulk dependence on N N N , up to a boundary term, to algebraic curvature–distortion contractions and the algebraic torsion potential. Thus the independent distortion equations contain no derivatives of N N N , although their linearized algebraic operator may be singular and may mix irreducible torsion and nonmetricity sectors. Its linearization may be organized schematically as
A A B δ N B + B A μ ν δ g μ ν = 0 , (232) \mathcal{A}_{AB}\,\delta N^B+\mathcal{B}_A{}^{\mu\nu}\delta g_{\mu\nu}=0, \tag*{(232)} A A B δ N B + B A μν δ g μν = 0 , (232)
where A , B A,B A , B label the 64 ordered distortion components. Determining physical degrees of freedom requires computing the rank and kernel of the complete 64 × 64 64\times64 64 × 64 operator in Eq. (232) , identifying genuine gauge identities, consistently eliminating auxiliary components, and analyzing the remaining metric kinetic and gradient operators. Neither the restricted axial Hessian rank nor the algebraicity of N N N establishes dynamical stability [2 , 3 , 36 , 37 ] .
Background dependence of the affine constraint rank. The rank and null space of the algebraic operator A A B \mathcal{A}_{AB} A A B are properties of a specified background, rather than universal numbers for the entire polynomial model. In particular, the non-null exceptional branch (T ∗ ≠ 0 T_\ast\ne0 T ∗ = 0 ) and the null-norm branch (T ∗ = 0 T_\ast=0 T ∗ = 0 ) need not have the same linearized constraint structure. Any rank calculation should therefore be repeated on the relevant backgrounds and combined with the mixed metric–distortion blocks before auxiliary variables are eliminated. A null vector of A A B \mathcal{A}_{AB} A A B alone neither identifies a gauge transformation nor establishes a propagating affine mode.
A stability analysis would require the complete second variation, constraint reduction, identification of genuine gauge symmetries, parity/harmonic decomposition, and the resulting kinetic and gradient operators with horizon and asymptotic boundary conditions. Standard metric black-hole perturbations provide a benchmark, but not a substitute for this coupled calculation [36 , 37 , 30 , 32 ] . The present results establish a rank-one restricted non-null axial Hessian, a nonlinear fixed-norm constraint, and a more degenerate null-norm branch. They establish neither unrestricted dynamical stability nor propagating torsion hair; physical charges and horizon entropy require a separate covariant phase-space treatment [38 , 27 ] .
Discussion and Conclusions
We have studied vacuum black-hole configurations in metric–affine Myrzakulov F ( R , T ) F(R,T) F ( R , T ) gravity (MG-I), distinguishing the affine Ricci scalar R R R and the quadratic torsion invariant T T T from the matter trace used in curvature–matter theories [5 , 6 , 9 ] . For the polynomial model F ( R , T ) = α R + β T + γ T 2 F(R,T)=\alpha R+\beta T+\gamma T^2 F ( R , T ) = α R + β T + γ T 2 with α ≠ 0 \alpha\ne0 α = 0 , we first varied the metric and all components of the independent affine connection, and only then restricted the resulting equations to metric-compatible, totally antisymmetric torsion. This ordering avoids assuming that a symmetry-reduced action automatically reproduces the complete affine Euler–Lagrange equations [2 , 3 , 28 , 29 ] .
For nonzero axial torsion, the affine equations admit the exceptional branch
α + β + 3 γ T ∗ 2 = 0. (233) \alpha+\beta+3\gamma T_\ast^2=0. \tag*{(233)} α + β + 3 γ T ∗ 2 = 0. (233)
For γ ≠ 0 \gamma\ne0 γ = 0 , the torsion scalar is constant on a connected smooth branch, and the metric equations reduce to
G ˙ μ ν + Λ ∗ g μ ν = 0 , Λ ∗ : = γ T ∗ 3 α . (234) \dot{G}_{\mu\nu}+\Lambda_\ast g_{\mu\nu}=0,\qquad\Lambda_\ast:=\frac{\gamma T_\ast^3}{\alpha}. \tag*{(234)} G ˙ μν + Λ ∗ g μν = 0 , Λ ∗ := α γ T ∗ 3 . (234)
The local static spherical metric is therefore of Schwarzschild–(anti–)de Sitter form, with f ( r ) = 1 − 2 M / r + Λ ∗ r 2 / 3 f(r)=1-2M/r+\Lambda_\ast r^2/3 f ( r ) = 1 − 2 M / r + Λ ∗ r 2 /3 in our curvature convention [30 , 31 ] . Its independent connection, however, can carry nonzero axial torsion and affine curvature; these are Riemann–Cartan solutions, not necessarily teleparallel geometries [1 , 2 , 3 ] .
A principal result is the residual functional freedom in the spherical axial connection at fixed T ∗ T_\ast T ∗ . Writing the torsion three-form in advanced Eddington–Finkelstein coordinates as τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ \tau=r^2\sin\theta[U(r)\,dv-V(r)\,dr]\wedge d\theta\wedge d\phi τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ and setting c = 2 T ∗ / 3 c=2T_\ast/3 c = 2 T ∗ /3 , we have constructed two families smooth across a selected finite-radius future Killing horizon. The first has U = a > 0 U=a>0 U = a > 0 and V = − c / ( a + a 2 + c f ) V=-c/(a+\sqrt{a^2+cf}) V = − c / ( a + a 2 + c f ) ; the second, developed in § 10.6 , is specified by W : = f V − U = − a W:=fV-U=-a W := f V − U = − a and
U = a 2 − c f , V = − c a + a 2 − c f , a 2 − c f > 0. (235) U=\sqrt{a^2-cf},\qquad V=-\frac{c}{a+\sqrt{a^2-cf}},\qquad a^2-cf>0. \tag*{(235)} U = a 2 − c f , V = − a + a 2 − c f c , a 2 − c f > 0. (235)
For the general spherical axial ansatz, R [ μ ν ] = 0 R_{[\mu\nu]}=0 R [ μν ] = 0 if and only if W ′ = 0 W'=0 W ′ = 0 on a regular radial interval. Thus the constant-W W W family has symmetric affine Ricci tensor, whereas the constant-U U U family generally does not. On their common real domain a 2 ± c f > 0 a^2\pm cf>0 a 2 ± c f > 0 , the two connections can support the same Einstein metric and torsion scalar while exhibiting different affine curvature invariants. The full affine Kretschmann contraction, including curvature–contorsion interference terms, is therefore essential to their local geometric comparison [2 , 3 ] . Although the constant-U U U and constant-W W W representatives agree in their torsion profiles at a selected future Killing horizon, their first radial jets generically differ, and so do affine curvature invariants. At a degenerate horizon their first-jet differences vanish, but differences in higher derivatives may remain. These statements concern the explicit local representatives and do not imply a global classification or dynamical stability.
The construction establishes smoothness of the torsion three-form and reconstructed connection on suitable future-horizon neighborhoods; it does not establish regularity at r = 0 r=0 r = 0 , at zeros of the square-root radicands, across a bifurcation sphere, or throughout a maximal extension [30 , 32 ] . The null-norm branch T ∗ = 0 T_* = 0 T ∗ = 0 may retain nonzero torsion: in the constant-W W W family U = a U=a U = a , V = 0 V=0 V = 0 , and K Γ = K L C − 8 a 2 / r 2 K_\Gamma=K_{\mathrm{LC}}-8a^2/r^2 K Γ = K LC − 8 a 2 / r 2 . In the exceptional linear case γ = 0 \gamma=0 γ = 0 , β = − α \beta=-\alpha β = − α , the axial torsion norm is not fixed by the affine equation, while the metric is Ricci flat.
For the actual double root of the Einstein lapse, extremality fixes Λ ∗ = − r h − 2 \Lambda_*=-r_h^{-2} Λ ∗ = − r h − 2 and M = r h / 3 M=r_h/3 M = r h /3 . The constant-U U U and constant-W W W representatives then have equal affine Kretschmann scalars at the horizon, Δ K ( r h ) = 0 \Delta K(r_h)=0 Δ K ( r h ) = 0 , even though their second radial jets can differ. This provides a concrete example of the limited discriminating power of a single scalar curvature invariant at an isolated hypersurface; it does not imply equivalence of the two affine geometries. At an actual double Killing horizon, the two representatives have identical affine curvature tensors pointwise, not merely equal Kretschmann scalars. Their second radial torsion jets may nevertheless differ, so this pointwise equality does not extend automatically to an open neighborhood.
Finally, the null directions of the restricted axial perturbative Hessian do not establish a propagating torsion mode, dynamical stability, or observable black-hole hair. A complete constrained analysis must include non-axial metric–affine perturbations, appropriate boundary conditions, and possible matter or spin couplings [33 , 34 , 36 , 37 ] . Conserved charges, horizon entropy, and global extensions also require separate treatment in the presence of an independent connection [38 , 27 ] . The results provide a local existence theorem and an invariant geometric comparison of torsionful black-hole configurations in a specified axial Riemann–Cartan sector, not a classification of all metric–affine black holes. Comparisons with other Myrzakulov black-hole constructions must keep their actions and connection restrictions explicit [12 , 25 ] .
Spherical minisuperspace reduction and its variational status
This appendix clarifies the relation between the formal point-like Lagrangian used in earlier treatments of Myrzakulov gravity and the unrestricted metric–affine variational problem. The distinction is particularly relevant when torsion and nonmetricity are independent geometric variables [10 , 5 , 2 ] . The arguments below do not assume that symmetry reduction commutes with variation; the relevant qualifications are described by the principle of symmetric criticality [28 , 29 ] .
Spherical variables and geometric identities
On a static patch consider
d s 2 = A ( r ) d t 2 − B ( r ) d r 2 − C ( r ) d Ω 2 , A , B , C > 0. (236) ds^2=A(r)\,dt^2-B(r)\,dr^2-C(r)\,d\Omega^2,\qquad A,B,C>0. \tag*{(236)} d s 2 = A ( r ) d t 2 − B ( r ) d r 2 − C ( r ) d Ω 2 , A , B , C > 0. (236)
Let ξ I ( r ) \xi^{I}(r) ξ I ( r ) denote a sufficiently general collection of spherically symmetric connection functions, so that Γ μ ν λ = Γ μ ν λ [ A , B , C ; ξ I ] \Gamma^{\lambda}_{\mu\nu}=\Gamma^{\lambda}_{\mu\nu}[A,B,C;\xi^{I}] Γ μν λ = Γ μν λ [ A , B , C ; ξ I ] . The independent connection must be included in both invariants:
R ( g , Γ ) = R ˚ ( g ) + u ( g , Γ ) , T ( g , Γ ) = T 0 ( g , Γ 0 ) + v ( g , Γ ; Γ 0 ) , (237) R(g,\Gamma)=\mathring{R}(g)+u(g,\Gamma),\qquad T(g,\Gamma)=T_{0}(g,\Gamma_{0})+v(g,\Gamma;\Gamma_{0}), \tag*{(237)} R ( g , Γ ) = R ˚ ( g ) + u ( g , Γ ) , T ( g , Γ ) = T 0 ( g , Γ 0 ) + v ( g , Γ ; Γ 0 ) , (237)
where
u : = R ( g , Γ ) − R ˚ ( g ) , v : = T ( g , Γ ) − T 0 ( g , Γ 0 ) . (238) u:=R(g,\Gamma)-\mathring{R}(g),\qquad v:=T(g,\Gamma)-T_{0}(g,\Gamma_{0}). \tag*{(238)} u := R ( g , Γ ) − R ˚ ( g ) , v := T ( g , Γ ) − T 0 ( g , Γ 0 ) . (238)
These are identities, not additional field equations or independent functions of r r r . In terms of the distortion N μ ν λ : = Γ μ ν λ − Γ ˚ μ ν λ N^{\lambda}_{\mu\nu}:=\Gamma^{\lambda}_{\mu\nu}-\mathring{\Gamma}^{\lambda}_{\mu\nu} N μν λ := Γ μν λ − Γ ˚ μν λ ,
u = g ρ ν ( 2 ∇ ˚ [ λ N ∣ ρ ∣ ν ] λ + 2 N σ [ λ λ N ∣ ρ ∣ ν ] σ ) . (239) u=g^{\rho\nu}\left(2\mathring{\nabla}_{[\lambda}N^{\lambda}_{|\rho|\nu]}+2N^{\lambda}_{\sigma[\lambda}N^{\sigma}_{|\rho|\nu]}\right). \tag*{(239)} u = g ρ ν ( 2 ∇ ˚ [ λ N ∣ ρ ∣ ν ] λ + 2 N σ [ λ λ N ∣ ρ ∣ ν ] σ ) . (239)
In particular, a reference connection Γ 0 \Gamma_{0} Γ 0 must be stated before v v v is used. The distinction between curvature-based, teleparallel, and general Riemann–Cartan connections is reviewed in [1 , 2 , 3 ] .
Introduce two auxiliary scalars R , T \mathcal{R},\mathcal{T} R , T and Lagrange multipliers λ R , λ T \lambda_{R},\lambda_{T} λ R , λ T in the vacuum action:
S a u x = 1 2 κ ∫ d 4 x − g [ F ( R , T ) − λ R ( R − R ( g , Γ ) ) − λ T ( T − T ( g , Γ ) ) ] . (240) S_{\mathrm{aux}}=\frac{1}{2\kappa}\int d^{4}x\sqrt{-g}\left[F(\mathcal{R},\mathcal{T})-\lambda_{R}\left(\mathcal{R}-R(g,\Gamma)\right)-\lambda_{T}\left(\mathcal{T}-T(g,\Gamma)\right)\right]. \tag*{(240)} S aux = 2 κ 1 ∫ d 4 x − g [ F ( R , T ) − λ R ( R − R ( g , Γ ) ) − λ T ( T − T ( g , Γ ) ) ] . (240)
The multiplier equations impose R = R ( g , Γ ) \mathcal{R}=R(g,\Gamma) R = R ( g , Γ ) and T = T ( g , Γ ) \mathcal{T}=T(g,\Gamma) T = T ( g , Γ ) , whereas the auxiliary-scalar equations yield
λ R = F R , λ T = F T . (241) \lambda_{R}=F_{R},\qquad\lambda_{T}=F_{T}. \tag*{(241)} λ R = F R , λ T = F T . (241)
Thus elimination of the multipliers gives
S a u x = 1 2 κ ∫ d 4 x − g [ F − F R ( R − R ( g , Γ ) ) − F T ( T − T ( g , Γ ) ) ] . (242) S_{\mathrm{aux}}=\frac{1}{2\kappa}\int d^{4}x\sqrt{-g}\left[F-F_{R}\left(\mathcal{R}-R(g,\Gamma)\right)-F_{T}\left(\mathcal{T}-T(g,\Gamma)\right)\right]. \tag*{(242)} S aux = 2 κ 1 ∫ d 4 x − g [ F − F R ( R − R ( g , Γ ) ) − F T ( T − T ( g , Γ ) ) ] . (242)
This expression is an auxiliary representation of the original variational system, provided the auxiliary constraints are retained. It does not eliminate the affine Euler–Lagrange equations. In particular, for F = α R + β T + γ T 3 F=\alpha\mathcal{R}+\beta\mathcal{T}+\gamma\mathcal{T}^{3} F = α R + β T + γ T 3 , F R = α F_{R}=\alpha F R = α is constant and the Hessian of F F F in ( R , T ) (\mathcal{R},\mathcal{T}) ( R , T ) is singular. Therefore one must not infer the independent auxiliary scalars from an invertible two-variable Legendre transformation. The multiplier formulation in (240) remains valid.
Integrating over the symmetry spheres gives, per unit coordinate time, the radial action
S r a d = 2 π κ ∫ d r C A B F ( R ( g , Γ ) , T ( g , Γ ) ) , (243) S_{\mathrm{rad}}=\frac{2\pi}{\kappa}\int dr\,C\sqrt{AB}\,F\left(R(g,\Gamma),T(g,\Gamma)\right), \tag*{(243)} S rad = κ 2 π ∫ d r C A B F ( R ( g , Γ ) , T ( g , Γ ) ) , (243)
up to the time-interval factor and boundary terms. A formal point-like density sometimes written as
L f o r m a l = C A B [ F − F R ( R − R ˚ − u ) − F T ( T − T 0 − v ) ] (244) \mathcal{L}_{\mathrm{formal}}=C\sqrt{AB}\left[F-F_{R}\left(\mathcal{R}-\mathring{R}-u\right)-F_{T}\left(\mathcal{T}-T_{0}-v\right)\right] \tag*{(244)} L formal = C A B [ F − F R ( R − R ˚ − u ) − F T ( T − T 0 − v ) ] (244)
is a closed reduced Lagrangian only after the connection functions ξ I \xi^{I} ξ I have been specified and u , v u,v u , v have been evaluated as functions of those fields and their derivatives. Omitting an overall constant factor does not affect the vacuum Euler–Lagrange equations.
Euler–Lagrange equations and boundary terms
Let q I = ( A , B , C , ξ 1 , … , ξ n , R , T , λ R , λ T ) q^{I}=(A,B,C,\xi^{1},\ldots,\xi^{n},\mathcal{R},\mathcal{T},\lambda_{R},\lambda_{T}) q I = ( A , B , C , ξ 1 , … , ξ n , R , T , λ R , λ T ) denote the independent reduced variables when the auxiliary formulation is used. For a radial Lagrangian L ( q , q ′ , q ′ ′ ) L(q,q',q'') L ( q , q ′ , q ′′ ) , the appropriate operator is
E I ( L ) : = ∂ L ∂ q I − d d r ∂ L ∂ q I ′ + d 2 d r 2 ∂ L ∂ q I ′ ′ = 0. (245) E_{I}(L):=\frac{\partial L}{\partial q^{I}}-\frac{d}{dr}\frac{\partial L}{\partial q^{I\prime}}+\frac{d^{2}}{dr^{2}}\frac{\partial L}{\partial q^{I\prime\prime}}=0. \tag*{(245)} E I ( L ) := ∂ q I ∂ L − d r d ∂ q I ′ ∂ L + d r 2 d 2 ∂ q I ′′ ∂ L = 0. (245)
It follows from two integrations by parts:
δ ∫ L d r = ∫ d r E I ( L ) δ q I + [ ( ∂ L ∂ q I ′ − d d r ∂ L ∂ q I ′ ′ ) δ q I + ∂ L ∂ q I ′ ′ δ q I ′ ] ∂ I . (246) \delta\int L\,\mathrm{d}r = \int\mathrm{d}r\,\mathcal{E}_{I}(L)\,\delta q^{I} + \left[\left(\frac{\partial L}{\partial q^{I\prime}}-\frac{\mathrm{d}}{\mathrm{d}r}\frac{\partial L}{\partial q^{I\prime\prime}}\right)\delta q^{I}+\frac{\partial L}{\partial q^{I\prime\prime}}\delta q^{I\prime}\right]_{\partial I}. \tag*{(246)} δ ∫ L d r = ∫ d r E I ( L ) δ q I + [ ( ∂ q I ′ ∂ L − d r d ∂ q I ′′ ∂ L ) δ q I + ∂ q I ′′ ∂ L δ q I ′ ] ∂ I . (246)
Repeated indices I I I are summed. Consequently, the expanded metric-only equations displayed in some earlier point-like formulations cannot be treated as the complete MG-I system unless the connection variables and their variations are also accounted for. In the polynomial model, α R \alpha R α R contains a divergence of the distortion; the admissibility of discarding this term depends on boundary conditions and on the fields held fixed.
Radial energy versus a gravitational constraint
If L L L has no explicit r r r dependence, its higher-derivative Noether–Ostrogradsky energy is
E = ∑ I [ q I ′ ( ∂ L ∂ q I ′ − d d r ∂ L ∂ q I ′ ′ ) + q I ′ ′ ∂ L ∂ q I ′ ′ ] − L . (247) E=\sum_I\left[q^{I\prime}\left(\frac{\partial L}{\partial q^{I\prime}}-\frac{\mathrm{d}}{\mathrm{d}r}\frac{\partial L}{\partial q^{I\prime\prime}}\right)+q^{I\prime\prime}\frac{\partial L}{\partial q^{I\prime\prime}}\right]-L. \tag*{(247)} E = I ∑ [ q I ′ ( ∂ q I ′ ∂ L − d r d ∂ q I ′′ ∂ L ) + q I ′′ ∂ q I ′′ ∂ L ] − L . (247)
Direct differentiation gives the identity
d E d r = − ∑ I E I ( L ) q I ′ . (248) \frac{\mathrm{d}E}{\mathrm{d}r}=-\sum_I\mathcal{E}_I(L)q^{I\prime}. \tag*{(248)} d r d E = − I ∑ E I ( L ) q I ′ . (248)
Thus E E E is constant on solutions, but E = 0 E=0 E = 0 is not a consequence of this identity. A vanishing radial Hamiltonian constraint, if present, must be derived from an unfixed metric equation or a properly established gauge symmetry. The frequently used first-order expression E f i r s t = ∑ I q I ′ ∂ L / ∂ q I ′ − L E_{\mathrm{first}}=\sum_I q^{I\prime}\partial L/\partial q^{I\prime}-L E first = ∑ I q I ′ ∂ L / ∂ q I ′ − L is applicable only to an equivalent first-order Lagrangian, after justified boundary-term manipulations. Moreover, the numerical value of a canonical energy can shift under addition of total derivatives; the equations of motion do not.
Consistency with unrestricted metric–affine equations
The fundamental vacuum equations are the independent variations
δ S δ g μ ν ∣ Γ = 0 , δ S δ Γ μ ν λ ∣ g = 0. (249) \left.\frac{\delta S}{\delta g^{\mu\nu}}\right|_{\Gamma}=0,\qquad\left.\frac{\delta S}{\delta\Gamma^{\lambda}_{\mu\nu}}\right|_{g}=0. \tag*{(249)} δ g μν δ S Γ = 0 , δ Γ μν λ δ S g = 0. (249)
A symmetry-reduced action reproduces their symmetry-restricted content only when the hypotheses of symmetric criticality and the boundary treatment have been checked [28 , 29 ] . In particular, a metric-only variation of (244) cannot replace the 64 independent affine variations. Neither can the radial coordinate freedom be used to remove an arbitrary function ξ I ( r ) \xi^{I}(r) ξ I ( r ) of the connection.
For the black-hole configurations considered in the main text, the reliable procedure is therefore to (i) derive the unrestricted metric and affine equations, (ii) specialize those equations to the metric-compatible, totally antisymmetric torsion sector, and (iii) substitute the spherical metric and torsion three-form into all remaining equations. The exceptional branch and its horizon-regular representatives are established by that route, not by counting the formal equations in the historical minisuperspace system [5 , 2 , 3 ] .
Independent coordinate verification of the axial black-hole geometry
The calculations in this appendix are independent of the field equations and of the axial Ricci identities invoked in the main text. We use the metric signature ( + − − − ) (+---) ( + − − − ) , the derivative convention ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ \nabla_{\nu}V^{\lambda}=\partial_{\nu}V^{\lambda}+\Gamma^{\lambda}_{\rho\nu}V^{\rho} ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ , and the curvature convention
R λ ρ μ ν = ∂ μ Γ ρ ν λ − ∂ ν Γ ρ μ λ + Γ σ μ λ Γ ρ ν σ − Γ σ ν λ Γ ρ μ σ , R ρ ν = R λ ρ λ ν . (250) R^{\lambda}{}_{\rho\mu\nu}=\partial_{\mu}\Gamma^{\lambda}_{\rho\nu}-\partial_{\nu}\Gamma^{\lambda}_{\rho\mu}+\Gamma^{\lambda}_{\sigma\mu}\Gamma^{\sigma}_{\rho\nu}-\Gamma^{\lambda}_{\sigma\nu}\Gamma^{\sigma}_{\rho\mu},\qquad R_{\rho\nu}=R^{\lambda}{}_{\rho\lambda\nu}. \tag*{(250)} R λ ρ μν = ∂ μ Γ ρ ν λ − ∂ ν Γ ρ μ λ + Γ σ μ λ Γ ρ ν σ − Γ σ ν λ Γ ρ μ σ , R ρ ν = R λ ρ λ ν . (250)
No vanishing-curvature (teleparallel) constraint is imposed [1 , 2 , 3 , 5 ] .
In the regular advanced chart x μ = ( v , r , θ , ϕ ) x^\mu=(v,r,\theta,\phi) x μ = ( v , r , θ , ϕ ) , take
d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 ( d θ 2 + sin 2 θ d ϕ 2 ) , (251) ds^2=f(r)\,dv^2-2\,dv\,dr-r^2(d\theta^2+\sin^2\theta\,d\phi^2), \tag*{(251)} d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 ( d θ 2 + sin 2 θ d ϕ 2 ) , (251)
τ = 1 3 ! X λ μ ν d x λ ∧ d x μ ∧ d x ν = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ , (252) \tau=\frac{1}{3!}X_{\lambda\mu\nu}\,dx^\lambda\wedge dx^\mu\wedge dx^\nu=r^2\sin\theta\,[U(r)\,dv-V(r)\,dr]\wedge d\theta\wedge d\phi, \tag*{(252)} τ = 3 ! 1 X λ μν d x λ ∧ d x μ ∧ d x ν = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ , (252)
Γ μ ν λ = Γ ˚ μ ν λ + 1 2 g λ ρ X ρ μ ν . (253) \Gamma^\lambda_{\ \mu\nu}=\mathring{\Gamma}^\lambda_{\ \mu\nu}+\frac{1}{2}g^{\lambda\rho}X_{\rho\mu\nu}. \tag*{(253)} Γ μν λ = Γ ˚ μν λ + 2 1 g λ ρ X ρ μν . (253)
The inverse metric has g v v = 0 g^{vv}=0 g v v = 0 , g v r = g r v = − 1 g^{vr}=g^{rv}=-1 g v r = g r v = − 1 , g r r = − f g^{rr}=-f g r r = − f , g θ θ = − r − 2 g^{\theta\theta}=-r^{-2} g θ θ = − r − 2 and g ϕ ϕ = − ( r 2 sin 2 θ ) − 1 g^{\phi\phi}=-(r^2\sin^2\theta)^{-1} g ϕϕ = − ( r 2 sin 2 θ ) − 1 . The only independent nonzero lowered torsion components are
X v θ ϕ = r 2 sin θ U , X r θ ϕ = − r 2 sin θ V , (254) X_{v\theta\phi}=r^2\sin\theta\,U,\qquad X_{r\theta\phi}=-r^2\sin\theta\,V, \tag*{(254)} X v θ ϕ = r 2 sin θ U , X r θ ϕ = − r 2 sin θ V , (254)
with all others fixed by total antisymmetry. This construction is manifestly metric-compatible, since the distortion in (253) is antisymmetric after lowering its first index. In particular,
X λ μ ν X λ μ ν = − 6 ( f V 2 − 2 U V ) , T = − 1 4 X λ μ ν X λ μ ν = 3 2 ( f V 2 − 2 U V ) . (255) X_{\lambda\mu\nu}X^{\lambda\mu\nu}=-6(fV^2-2UV),\qquad T=-\frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu}=\frac{3}{2}(fV^2-2UV). \tag*{(255)} X λ μν X λ μν = − 6 ( f V 2 − 2 U V ) , T = − 4 1 X λ μν X λ μν = 2 3 ( f V 2 − 2 U V ) . (255)
Direct Ricci contraction at generic polar angle
Construct Γ ˚ \mathring{\Gamma} Γ ˚ from (251) , add the contorsion in (253) , form every component of (250) , and contract R ρ λ ν λ R^\lambda_{\ \rho\lambda\nu} R ρ λ ν λ . With θ \theta θ retained as a variable, the nonzero Ricci components are
R v v = f 2 ( f ′ ′ + 2 f ′ r ) − U 2 2 , R v r = R r v = − 1 2 ( f ′ ′ + 2 f ′ r ) + U V 2 , (256) R_{vv}=\frac{f}{2}\left(f^{\prime\prime}+\frac{2f'}{r}\right)-\frac{U^2}{2},\qquad R_{vr}=R_{rv}=-\frac{1}{2}\left(f^{\prime\prime}+\frac{2f'}{r}\right)+\frac{UV}{2}, \tag*{(256)} R v v = 2 f ( f ′′ + r 2 f ′ ) − 2 U 2 , R v r = R r v = − 2 1 ( f ′′ + r 2 f ′ ) + 2 U V , (256)
R r r = − V 2 2 , R θ θ = 1 − f − r f ′ + r 2 ( U V − f V 2 2 ) , (257) R_{rr}=-\frac{V^2}{2},\qquad R_{\theta\theta}=1-f-rf'+r^2\left(UV-\frac{fV^2}{2}\right), \tag*{(257)} R r r = − 2 V 2 , R θ θ = 1 − f − r f ′ + r 2 ( U V − 2 f V 2 ) , (257)
R ϕ ϕ = sin 2 θ R θ θ , R θ ϕ = − R ϕ θ = r 2 sin θ 2 ( f V − U ) ′ . (258) R_{\phi\phi}=\sin^2\theta\,R_{\theta\theta},\qquad R_{\theta\phi}=-R_{\phi\theta}=\frac{r^2\sin\theta}{2}(fV-U)'. \tag*{(258)} R ϕϕ = sin 2 θ R θ θ , R θ ϕ = − R ϕθ = 2 r 2 sin θ ( f V − U ) ′ . (258)
All remaining ordered components vanish. The sixteen ordered Ricci entries have been compared symbolically with the direct connection construction at generic θ \theta θ , rather than only at the equator. In particular, the angular dependence in (258) follows from the full coordinate calculation, not merely from an appeal to covariance.
For comparison, the Levi–Civita components are
R ˚ v v = f 2 ( f ′ ′ + 2 f ′ r ) , R ˚ v r = − 1 2 ( f ′ ′ + 2 f ′ r ) , R ˚ r r = 0 , (259) \mathring{R}_{vv}=\frac{f}{2}\left(f^{\prime\prime}+\frac{2f'}{r}\right),\qquad\mathring{R}_{vr}=-\frac{1}{2}\left(f^{\prime\prime}+\frac{2f'}{r}\right),\qquad\mathring{R}_{rr}=0, \tag*{(259)} R ˚ v v = 2 f ( f ′′ + r 2 f ′ ) , R ˚ v r = − 2 1 ( f ′′ + r 2 f ′ ) , R ˚ r r = 0 , (259)
R ˚ θ θ = 1 − f − r f ′ , R ˚ ϕ ϕ = sin 2 θ R ˚ θ θ . (260) \mathring{R}_{\theta\theta}=1-f-rf',\qquad\mathring{R}_{\phi\phi}=\sin^2\theta\,\mathring{R}_{\theta\theta}. \tag*{(260)} R ˚ θ θ = 1 − f − r f ′ , R ˚ ϕϕ = sin 2 θ R ˚ θ θ . (260)
Thus the symmetric correction H μ ν = R ( μ ν ) − R ˚ μ ν H_{\mu\nu}=R_{(\mu\nu)}-\mathring{R}_{\mu\nu} H μν = R ( μν ) − R ˚ μν is explicitly
H v v = − U 2 2 , H v r = U V 2 , H r r = − V 2 2 , (261) H_{vv}=-\frac{U^2}{2},\qquad H_{vr}=\frac{UV}{2},\qquad H_{rr}=-\frac{V^2}{2}, \tag*{(261)} H v v = − 2 U 2 , H v r = 2 U V , H r r = − 2 V 2 , (261)
H θ θ = r 2 ( U V − f V 2 2 ) , H ϕ ϕ = sin 2 θ H θ θ . (262) H_{\theta\theta}=r^2\left(UV-\frac{fV^2}{2}\right),\qquad H_{\phi\phi}=\sin^2\theta\,H_{\theta\theta}. \tag*{(262)} H θ θ = r 2 ( U V − 2 f V 2 ) , H ϕϕ = sin 2 θ H θ θ . (262)
Direct contraction, without imposing the field equations, yields
R ( Γ ) = f ′ ′ + 4 f ′ r + 2 ( f − 1 ) r 2 + 3 2 ( f V 2 − 2 U V ) = R ˚ + T , (263) R(\Gamma)=f^{\prime\prime}+\frac{4f'}{r}+\frac{2(f-1)}{r^2}+\frac{3}{2}(fV^2-2UV)=\mathring{R}+T, \tag*{(263)} R ( Γ ) = f ′′ + r 4 f ′ + r 2 2 ( f − 1 ) + 2 3 ( f V 2 − 2 U V ) = R ˚ + T , (263)
g μ ν H μ ν = T , R [ μ ν ] R [ μ ν ] = 1 2 [ ( f V − U ) ′ ] 2 . (264) g^{\mu\nu}H_{\mu\nu}=T,\qquad R_{[\mu\nu]}R^{[\mu\nu]}=\frac{1}{2}[(fV-U)']^2. \tag*{(264)} g μν H μν = T , R [ μν ] R [ μν ] = 2 1 [( f V − U ) ′ ] 2 . (264)
These calculations independently corroborate R ( μ ν ) = R ˚ μ ν + H μ ν R_{(\mu\nu)}=\mathring{R}_{\mu\nu}+H_{\mu\nu} R ( μν ) = R ˚ μν + H μν and R ( Γ ) = R ˚ + T R(\Gamma)=\mathring{R}+T R ( Γ ) = R ˚ + T in the specified axial sector. They do not constitute a metric variation of the unrestricted action; that separate issue is addressed in Appendix C.
Independent four-index curvature contraction
We compute the full affine curvature invariant without imposing either the constant-norm constraint or the gravitational field equations. The curvature and Ricci conventions are those of § 2 . In particular, the connection is metric-compatible but is not assumed flat. This is the Riemann–Cartan setting discussed in Refs. [2 , 3 ] ; it should not be confused with a teleparallel connection.
For the advanced Eddington–Finkelstein metric and axial three-form
d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 ( d θ 2 + sin 2 θ d ϕ 2 ) , (265) ds^2=f(r)\,dv^2-2\,dv\,dr-r^2(d\theta^2+\sin^2\theta\,d\phi^2), \tag*{(265)} d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 ( d θ 2 + sin 2 θ d ϕ 2 ) , (265)
τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ , (266) \tau=r^2\sin\theta\,[U(r)\,dv-V(r)\,dr]\wedge d\theta\wedge d\phi, \tag*{(266)} τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ , (266)
we construct Γ λ μ ν = Γ ˚ λ μ ν + 1 2 g λ ρ X ρ μ ν \Gamma^\lambda{}_{\mu\nu}=\mathring{\Gamma}^\lambda{}_{\mu\nu}+\frac{1}{2}g^{\lambda\rho}X_{\rho\mu\nu} Γ λ μν = Γ ˚ λ μν + 2 1 g λ ρ X ρ μν and then calculate
R λ ρ μ ν ( Γ ) = ∂ μ Γ λ ρ ν − ∂ ν Γ λ ρ μ + Γ λ σ μ Γ σ ρ ν − Γ λ σ ν Γ σ ρ μ . (267) R^\lambda{}_{\rho\mu\nu}(\Gamma)=\partial_\mu\Gamma^\lambda{}_{\rho\nu}-\partial_\nu\Gamma^\lambda{}_{\rho\mu}+\Gamma^\lambda{}_{\sigma\mu}\Gamma^\sigma{}_{\rho\nu}-\Gamma^\lambda{}_{\sigma\nu}\Gamma^\sigma{}_{\rho\mu}. \tag*{(267)} R λ ρ μν ( Γ ) = ∂ μ Γ λ ρ ν − ∂ ν Γ λ ρ μ + Γ λ σ μ Γ σ ρ ν − Γ λ σ ν Γ σ ρ μ . (267)
The four-index contraction, with the first curvature index lowered before contraction, is
K Γ = R λ ρ μ ν ( Γ ) R λ ρ μ ν ( Γ ) . (268) K_\Gamma=R_{\lambda\rho\mu\nu}(\Gamma)R^{\lambda\rho\mu\nu}(\Gamma). \tag*{(268)} K Γ = R λ ρ μν ( Γ ) R λ ρ μν ( Γ ) . (268)
The Ricci tensor alone does not determine (268) for a general Riemann–Cartan connection [2 , 3 ] .
Unrestricted radial amplitudes. Write F = f F=f F = f , F 1 = f ′ F_1=f' F 1 = f ′ , F 2 = f ′ ′ F_2=f'' F 2 = f ′′ , U 1 = U ′ U_1=U' U 1 = U ′ and V 1 = V ′ V_1=V' V 1 = V ′ . Direct contraction of the complete four-index curvature gives
4 r 4 K Γ = 4 r 4 F 2 2 + 16 r 2 F 1 2 + 16 ( F − 1 ) 2 + r 4 [ 3 F 2 V 4 + 8 F 2 V 1 2 + 8 F F 1 V V 1 − 12 F U V 3 − 16 F U 1 V 1 + 4 F 1 2 V 2 + 8 F 1 U V 1 − 8 F 1 U 1 V − 4 U 1 2 + 12 U 2 V 2 ] + r 3 [ 8 F F 1 V 2 − 16 F 1 U V − 16 U U 1 ] + r 2 [ 8 F 2 V 2 − 16 F U V − 8 F V 2 − 32 U 2 + 16 U V ] . (269) \begin{aligned}
4r^4K_\Gamma={}&4r^4F_2^2+16r^2F_1^2+16(F-1)^2\\
&+r^4\left[3F^2V^4+8F^2V_1^2+8FF_1VV_1-12FUV^3-16FU_1V_1\right.\\
&\left.\quad+4F_1^2V^2+8F_1UV_1-8F_1U_1V-4U_1^2+12U^2V^2\right]\\
&+r^3\left[8FF_1V^2-16F_1UV-16UU_1\right]\\
&+r^2\left[8F^2V^2-16FUV-8FV^2-32U^2+16UV\right]. \tag*{(269)}
\end{aligned} 4 r 4 K Γ = 4 r 4 F 2 2 + 16 r 2 F 1 2 + 16 ( F − 1 ) 2 + r 4 [ 3 F 2 V 4 + 8 F 2 V 1 2 + 8 F F 1 V V 1 − 12 F U V 3 − 16 F U 1 V 1 + 4 F 1 2 V 2 + 8 F 1 U V 1 − 8 F 1 U 1 V − 4 U 1 2 + 12 U 2 V 2 ] + r 3 [ 8 F F 1 V 2 − 16 F 1 U V − 16 U U 1 ] + r 2 [ 8 F 2 V 2 − 16 F U V − 8 F V 2 − 32 U 2 + 16 U V ] . (269)
The torsion-free part is precisely
K L C = ( f ′ ′ ) 2 + 4 ( f ′ ) 2 r 2 + 4 ( f − 1 ) 2 r 4 . (270) K_{\mathrm{LC}}=(f'')^2+\frac{4(f')^2}{r^2}+\frac{4(f-1)^2}{r^4}. \tag*{(270)} K LC = ( f ′′ ) 2 + r 2 4 ( f ′ ) 2 + r 4 4 ( f − 1 ) 2 . (270)
Equation (269) holds for arbitrary differentiable f f f , U U U , V V V and is not contingent on f V 2 − 2 U V fV^2-2UV f V 2 − 2 U V being constant.
Independent angular check. The curvature was reconstructed independently from the full connection at θ = π / 2 \theta=\pi/2 θ = π /2 and θ = π / 3 \theta=\pi/3 θ = π /3 , retaining F F F , F 1 F_1 F 1 , F 2 F_2 F 2 , U U U , U 1 U_1 U 1 , V V V , V 1 V_1 V 1 as algebraically independent quantities. Both calculations give exactly (269) ; their symbolic difference simplifies to zero. This two-angle test is an independent check of the component implementation. In addition, spherical symmetry of the metric and three-form implies that the scalar (268) is angle-independent away from the polar coordinate singularities. The two-angle test should not be described as a third, independent symbolic calculation with a free polar-angle variable.
Curvature decomposition. Set D λ ρ μ ν : = R λ ρ μ ν ( Γ ) − R ˚ λ ρ μ ν \mathcal{D}^{\lambda}{}_{\rho\mu\nu}:=R^{\lambda}{}_{\rho\mu\nu}(\Gamma)-\mathring{R}^{\lambda}{}_{\rho\mu\nu} D λ ρ μν := R λ ρ μν ( Γ ) − R ˚ λ ρ μν . The exact distortion identity is
D λ ρ μ ν = 2 ∇ ˚ [ μ K λ ∣ ρ ∣ ν ] + 2 K λ σ [ μ K σ ∣ ρ ∣ ν ] , K λ μ ν = 1 2 X λ μ ν . (271) \mathcal{D}^{\lambda}{}_{\rho\mu\nu}=2\mathring{\nabla}_{[\mu}K^\lambda{}_{|\rho|\nu]}+2K^\lambda{}_{\sigma[\mu}K^\sigma{}_{|\rho|\nu]},\qquad K^\lambda{}_{\mu\nu}=\frac{1}{2}X^\lambda{}_{\mu\nu}. \tag*{(271)} D λ ρ μν = 2 ∇ ˚ [ μ K λ ∣ ρ ∣ ν ] + 2 K λ σ [ μ K σ ∣ ρ ∣ ν ] , K λ μν = 2 1 X λ μν . (271)
Consequently,
K Γ = K L C + 2 R ˚ ⋅ D + D ⋅ D . (272) K_\Gamma=K_{\mathrm{LC}}+2\mathring{R}\cdot\mathcal{D}+\mathcal{D}\cdot\mathcal{D}. \tag*{(272)} K Γ = K LC + 2 R ˚ ⋅ D + D ⋅ D . (272)
The cross term obtained by a separate contraction is
2 R ˚ ⋅ D = 2 V ( f V − 2 U ) r 2 ( f + r f ′ − 1 ) . (273) 2\mathring{R}\cdot D=\frac{2V(fV-2U)}{r^2}(f+rf'-1). \tag*{(273)} 2 R ˚ ⋅ D = r 2 2 V ( f V − 2 U ) ( f + r f ′ − 1 ) . (273)
The purely distortion-dependent contribution, without invoking the field equations, is
D λ ρ μ ν D λ ρ μ ν = 1 4 [ 3 f 2 V 4 + 8 f 2 ( V ′ ) 2 + 8 f f ′ V V ′ − 12 f U V 3 − 16 f U ′ V ′ + 4 ( f ′ ) 2 V 2 + 8 f ′ U V ′ − 8 f ′ U ′ V − 4 ( U ′ ) 2 + 12 U 2 V 2 ] − 4 U U ′ r − 8 U 2 r 2 . (274) D_{\lambda\rho\mu\nu}D^{\lambda\rho\mu\nu}=\frac{1}{4}\left[3f^2V^4+8f^2(V')^2+8ff'VV'-12fUV^3-16fU'V'+4(f')^2V^2+8f'UV'-8f'U'V-4(U')^2+12U^2V^2\right]-\frac{4UU'}{r}-\frac{8U^2}{r^2}. \tag*{(274)} D λ ρ μν D λ ρ μν = 4 1 [ 3 f 2 V 4 + 8 f 2 ( V ′ ) 2 + 8 f f ′ V V ′ − 12 f U V 3 − 16 f U ′ V ′ + 4 ( f ′ ) 2 V 2 + 8 f ′ U V ′ − 8 f ′ U ′ V − 4 ( U ′ ) 2 + 12 U 2 V 2 ] − r 4 U U ′ − r 2 8 U 2 . (274)
Indeed, subtracting the Levi–Civita term and the mixed contraction from the unrestricted result gives
D λ ρ μ ν D λ ρ μ ν = K Γ − K L C − 2 R ˚ λ ρ μ ν D λ ρ μ ν , (275) D_{\lambda\rho\mu\nu}D^{\lambda\rho\mu\nu}=K_{\Gamma}-K_{\mathrm{LC}}-2\mathring{R}_{\lambda\rho\mu\nu}D^{\lambda\rho\mu\nu}, \tag*{(275)} D λ ρ μν D λ ρ μν = K Γ − K LC − 2 R ˚ λ ρ μν D λ ρ μν , (275)
which reproduces Eq. (274) identically, without imposing the constant-torsion constraint. Substitution of (273) and (274) into (272) reproduces every term in (269) , furnishing an additional algebraic consistency check.
Exceptional Einstein branch. For the nonlinear exceptional solutions considered in § 10 ,
f = 1 − 2 M r + Λ ∗ r 2 3 , f V 2 − 2 U V = c : = 2 T ∗ 3 . (276) f=1-\frac{2M}{r}+\frac{\Lambda_*r^2}{3},\qquad fV^2-2UV=c:=\frac{2T_*}{3}. \tag*{(276)} f = 1 − r 2 M + 3 Λ ∗ r 2 , f V 2 − 2 U V = c := 3 2 T ∗ . (276)
Equation (273) then reduces to
2 R ˚ ⋅ D = 2 c Λ ∗ . (277) 2\mathring{R}\cdot D=2c\Lambda_*. \tag*{(277)} 2 R ˚ ⋅ D = 2 c Λ ∗ . (277)
For the constant-U U U representative U = a > 0 U=a>0 U = a > 0 , let q 2 = a 2 + c f q^2=a^2+cf q 2 = a 2 + c f and V = − c / ( a + q ) V=-c/(a+q) V = − c / ( a + q ) on a smooth real branch (q ≠ 0 q\ne0 q = 0 and a + q ≠ 0 a+q\ne0 a + q = 0 ). Differentiating gives
V ′ = c 2 f ′ 2 q ( a + q ) 2 . (278) V'=\frac{c^2f'}{2q(a+q)^2}. \tag*{(278)} V ′ = 2 q ( a + q ) 2 c 2 f ′ . (278)
Inserting these relations into the unrestricted polynomial yields
K Γ = K L C + c 2 ( f ′ ) 2 2 ( a 2 + c f ) + 2 c f ′ r + 3 c 2 4 + 2 c f − 8 a 2 − 2 c r 2 , K L C = 48 M 2 r 6 + 8 Λ ∗ 2 3 . (279) \begin{aligned}
K_{\Gamma}&=K_{\mathrm{LC}}+\frac{c^2(f')^2}{2(a^2+cf)}+\frac{2cf'}{r}+\frac{3c^2}{4}+\frac{2cf-8a^2-2c}{r^2},\\
K_{\mathrm{LC}}&=\frac{48M^2}{r^6}+\frac{8\Lambda_*^2}{3}. \tag*{(279)}
\end{aligned} K Γ K LC = K LC + 2 ( a 2 + c f ) c 2 ( f ′ ) 2 + r 2 c f ′ + 4 3 c 2 + r 2 2 c f − 8 a 2 − 2 c , = r 6 48 M 2 + 3 8 Λ ∗ 2 . (279)
An exact symbolic substitution into (269) , with q 2 = a 2 + c f q^2=a^2+cf q 2 = a 2 + c f , gives a vanishing difference from (279) . The expression is valid where the chosen representative is differentiable; zeros of the radicand require the separate analysis in § 10.4 . In the null case c = 0 c=0 c = 0 , a ≠ 0 a\ne0 a = 0 , this reduces to K Γ = K L C − 8 a 2 / r 2 K_{\Gamma}=K_{\mathrm{LC}}-8a^2/r^2 K Γ = K LC − 8 a 2 / r 2 . Since the metric is Lorentzian, K Γ K_{\Gamma} K Γ need not be nonnegative; its sign alone is not a regularity criterion. The interpretation of horizon regularity and curvature invariants follows the standard black-hole geometric framework [30 , 32 ] .
Exceptional axial branch: two horizon-regular representatives and invariant checks
We specialize the generic coordinate curvature identities of Appendix B.2–B.3 after deriving the unrestricted metric–affine field equations. In the advanced Eddington–Finkelstein chart, with signature ( + − − − ) (+---) ( + − − − ) ,
d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 ( d θ 2 + sin 2 θ d ϕ 2 ) , f ( r ) = 1 − 2 M r + Λ ∗ r 2 3 , (280) ds^2=f(r)\,dv^2-2\,dv\,dr-r^2(d\theta^2+\sin^2\theta\,d\phi^2),\qquad f(r)=1-\frac{2M}{r}+\frac{\Lambda_*r^2}{3}, \tag*{(280)} d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 ( d θ 2 + sin 2 θ d ϕ 2 ) , f ( r ) = 1 − r 2 M + 3 Λ ∗ r 2 , (280)
τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ , Γ μ ν λ = Γ ˚ μ ν λ + 1 2 g λ ρ X ρ μ ν . (281) \tau=r^2\sin\theta[U(r)\,dv-V(r)\,dr]\wedge d\theta\wedge d\phi,\qquad\Gamma^\lambda_{\mu\nu}=\mathring{\Gamma}^\lambda_{\mu\nu}+\frac{1}{2}g^{\lambda\rho}X_{\rho\mu\nu}. \tag*{(281)} τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ , Γ μν λ = Γ ˚ μν λ + 2 1 g λ ρ X ρ μν . (281)
Here τ = X λ μ ν d x λ ∧ d x μ ∧ d x ν / 3 ! \tau=X_{\lambda\mu\nu}dx^\lambda\wedge dx^\mu\wedge dx^\nu/3! τ = X λ μν d x λ ∧ d x μ ∧ d x ν /3 ! ; our curvature convention is R λ ρ μ ν = ∂ μ Γ ρ ν λ − ∂ ν Γ ρ μ λ + Γ σ μ λ Γ ρ ν σ − Γ σ ν λ Γ ρ μ σ R^\lambda{}_{\rho\mu\nu}=\partial_\mu\Gamma^\lambda_{\rho\nu}-\partial_\nu\Gamma^\lambda_{\rho\mu}+\Gamma^\lambda_{\sigma\mu}\Gamma^\sigma_{\rho\nu}-\Gamma^\lambda_{\sigma\nu}\Gamma^\sigma_{\rho\mu} R λ ρ μν = ∂ μ Γ ρ ν λ − ∂ ν Γ ρ μ λ + Γ σ μ λ Γ ρ ν σ − Γ σ ν λ Γ ρ μ σ . The exceptional nonlinear axial branch obeys
α + β + 3 γ T ∗ 2 = 0 , Λ ∗ = γ T ∗ 3 α , c : = 2 T ∗ 3 , f V 2 − 2 U V = c . (282) \alpha+\beta+3\gamma T_*^2=0,\qquad\Lambda_*=\frac{\gamma T_*^3}{\alpha},\qquad c:=\frac{2T_*}{3},\qquad fV^2-2UV=c. \tag*{(282)} α + β + 3 γ T ∗ 2 = 0 , Λ ∗ = α γ T ∗ 3 , c := 3 2 T ∗ , f V 2 − 2 U V = c . (282)
The connection is Riemann–Cartan and is not assumed teleparallel [1 , 2 , 3 ] . All results below concern connected radial intervals with r > 0 r>0 r > 0 and smooth real profiles.
Generic Ricci identities and contractions. Before choosing either representative, direct contraction of the full four-index affine curvature yields
R v v = f 2 ( f ′ ′ + 2 f ′ r ) − U 2 2 , R v r = R r v = − 1 2 ( f ′ ′ + 2 f ′ r ) + U V 2 (283) R_{vv}=\frac{f}{2}\left(f^{\prime\prime}+\frac{2f^\prime}{r}\right)-\frac{U^2}{2},\qquad R_{vr}=R_{rv}=-\frac{1}{2}\left(f^{\prime\prime}+\frac{2f^\prime}{r}\right)+\frac{UV}{2} \tag*{(283)} R v v = 2 f ( f ′′ + r 2 f ′ ) − 2 U 2 , R v r = R r v = − 2 1 ( f ′′ + r 2 f ′ ) + 2 U V (283)
R r r = − V 2 2 , R θ θ = 1 − f − r f ′ + r 2 ( U V − f V 2 2 ) (284) R_{rr}=-\frac{V^2}{2},\qquad R_{\theta\theta}=1-f-rf^\prime+r^2\left(UV-\frac{fV^2}{2}\right) \tag*{(284)} R r r = − 2 V 2 , R θ θ = 1 − f − r f ′ + r 2 ( U V − 2 f V 2 ) (284)
R ϕ ϕ = sin 2 θ R θ θ , R θ ϕ = − R ϕ θ = r 2 sin θ 2 ( f V − U ) ′ . (285) R_{\phi\phi}=\sin^2\theta\,R_{\theta\theta},\qquad R_{\theta\phi}=-R_{\phi\theta}=\frac{r^2\sin\theta}{2}(fV-U)^\prime. \tag*{(285)} R ϕϕ = sin 2 θ R θ θ , R θ ϕ = − R ϕθ = 2 r 2 sin θ ( f V − U ) ′ . (285)
These equations use the Ricci contraction R ρ ν = R λ ρ λ ν R_{\rho\nu}=R^\lambda{}_{\rho\lambda\nu} R ρ ν = R λ ρ λ ν , not an alternative contraction. With W : = f V − U W:=fV-U W := f V − U , the constant-norm condition gives W 2 = U 2 + c f W^2=U^2+cf W 2 = U 2 + c f and
R [ μ ν ] R [ μ ν ] = 1 2 ( W ′ ) 2 , R [ μ ν ] = 0 ⟺ W ′ = 0 ⟺ W = W 0 on a connected interval . (286) R_{[\mu\nu]}R^{[\mu\nu]}=\frac{1}{2}(W^\prime)^2,\qquad R_{[\mu\nu]}=0\Longleftrightarrow W^\prime=0\Longleftrightarrow W=W_0\quad\text{on a connected interval}. \tag*{(286)} R [ μν ] R [ μν ] = 2 1 ( W ′ ) 2 , R [ μν ] = 0 ⟺ W ′ = 0 ⟺ W = W 0 on a connected interval . (286)
This equivalence is restricted to the spherical axial ansatz. On the Einstein metric and constant-norm branch,
R ( Γ ) = 4 Λ ∗ + T ∗ , R ( μ ν ) R ( μ ν ) = 4 Λ ∗ 2 + 2 Λ ∗ T ∗ + T ∗ 2 3 . (287) R(\Gamma)=4\Lambda_*+T_*,\qquad R_{(\mu\nu)}R^{(\mu\nu)}=4\Lambda_*^2+2\Lambda_*T_*+\frac{T_*^2}{3}. \tag*{(287)} R ( Γ ) = 4 Λ ∗ + T ∗ , R ( μν ) R ( μν ) = 4 Λ ∗ 2 + 2 Λ ∗ T ∗ + 3 T ∗ 2 . (287)
Indeed R ( μ ν ) = Λ ∗ g μ ν + H μ ν R_{(\mu\nu)}=\Lambda_*g_{\mu\nu}+H_{\mu\nu} R ( μν ) = Λ ∗ g μν + H μν , g μ ν H μ ν = T ∗ g^{\mu\nu}H_{\mu\nu}=T_* g μν H μν = T ∗ and H μ ν H μ ν = T ∗ 2 / 3 H_{\mu\nu}H^{\mu\nu}=T_*^2/3 H μν H μν = T ∗ 2 /3 .
Constant-U U U representative. Choose U = a > 0 U=a>0 U = a > 0 , q : = a 2 + c f q:=\sqrt{a^2+cf} q := a 2 + c f , and the horizon-regular root
V = a − q f = − c a + q , W = − q , q ′ = c f ′ 2 q , V ′ = c 2 f ′ 2 q ( a + q ) 2 , a 2 + c f > 0. (288) V=\frac{a-q}{f}=-\frac{c}{a+q},\qquad W=-q,\qquad q^\prime=\frac{cf^\prime}{2q},\qquad V^\prime=\frac{c^2f^\prime}{2q(a+q)^2},\qquad a^2+cf>0. \tag*{(288)} V = f a − q = − a + q c , W = − q , q ′ = 2 q c f ′ , V ′ = 2 q ( a + q ) 2 c 2 f ′ , a 2 + c f > 0. (288)
The rationalized expression, not the quotient ( a − q ) / f (a-q)/f ( a − q ) / f , defines V V V at a zero of f f f . Since W ′ = − c f ′ / ( 2 q ) W^\prime=-cf^\prime/(2q) W ′ = − c f ′ / ( 2 q ) ,
R θ ϕ = r 2 sin θ 2 ( f V ) ′ , I A ( U ) : = R [ μ ν ] R [ μ ν ] = c 2 ( f ′ ) 2 8 ( a 2 + c f ) . (289) R_{\theta\phi}=\frac{r^2\sin\theta}{2}(fV)^\prime,\qquad I_A^{(U)}:=R_{[\mu\nu]}R^{[\mu\nu]}=\frac{c^2(f^\prime)^2}{8(a^2+cf)}. \tag*{(289)} R θ ϕ = 2 r 2 sin θ ( f V ) ′ , I A ( U ) := R [ μν ] R [ μν ] = 8 ( a 2 + c f ) c 2 ( f ′ ) 2 . (289)
On the Einstein background the remaining Ricci components simplify to
R v v = Λ ∗ f − a 2 2 , R v r = R r v = − Λ ∗ + a V 2 , R r r = − V 2 2 (290) R_{vv}=\Lambda_*f-\frac{a^2}{2},\qquad R_{vr}=R_{rv}=-\Lambda_*+\frac{aV}{2},\qquad R_{rr}=-\frac{V^2}{2} \tag*{(290)} R v v = Λ ∗ f − 2 a 2 , R v r = R r v = − Λ ∗ + 2 aV , R r r = − 2 V 2 (290)
R θ θ = − r 2 ( Λ ∗ + c 2 ) , R ϕ ϕ = sin 2 θ R θ θ . (291) R_{\theta\theta}=-r^2\left(\Lambda_*+\frac{c}{2}\right),\qquad R_{\phi\phi}=\sin^2\theta\,R_{\theta\theta}. \tag*{(291)} R θ θ = − r 2 ( Λ ∗ + 2 c ) , R ϕϕ = sin 2 θ R θ θ . (291)
The full affine Kretschmann scalar, not a Ricci-only contraction, is
K Γ ( U ) = K L C + c 2 ( f ′ ) 2 2 ( a 2 + c f ) + 2 c f ′ r + 3 c 2 4 + 2 c f − 8 a 2 − 2 c r 2 , K L C = 48 M 2 r 6 + 8 Λ ∗ 2 3 . (292) K_\Gamma^{(U)}=K_{\mathrm{LC}}+\frac{c^2(f^\prime)^2}{2(a^2+cf)}+\frac{2cf^\prime}{r}+\frac{3c^2}{4}+\frac{2cf-8a^2-2c}{r^2},\qquad K_{\mathrm{LC}}=\frac{48M^2}{r^6}+\frac{8\Lambda_*^2}{3}. \tag*{(292)} K Γ ( U ) = K LC + 2 ( a 2 + c f ) c 2 ( f ′ ) 2 + r 2 c f ′ + 4 3 c 2 + r 2 2 c f − 8 a 2 − 2 c , K LC = r 6 48 M 2 + 3 8 Λ ∗ 2 . (292)
Constant-W W W representative and local classification. Equation (286) establishes that every Ricci-symmetric spherical axial profile has W = W 0 W=W_0 W = W 0 constant. The algebraic constraint then requires
U 2 = W 0 2 − c f , U = σ W 0 2 − c f , V = − c U − W 0 , σ ∈ { + 1 , − 1 } , (293) U^2=W_0^2-cf,\qquad U=\sigma\sqrt{W_0^2-cf},\qquad V=-\frac{c}{U-W_0},\qquad\sigma\in\{+1,-1\}, \tag*{(293)} U 2 = W 0 2 − c f , U = σ W 0 2 − c f , V = − U − W 0 c , σ ∈ { + 1 , − 1 } , (293)
on intervals where the radicand is positive and U − W 0 ≠ 0 U-W_0\ne0 U − W 0 = 0 . At a zero of the denominator the undivided norm constraint must be used; the rationalized expression cannot be extended by division there. For the future-horizon representative selected in § 10.6 ,
W = − a , u : = a 2 − c f , U = u , V = − c a + u , a 2 − c f > 0. (294) W=-a,\qquad u:=\sqrt{a^2-cf},\qquad U=u,\qquad V=-\frac{c}{a+u},\qquad a^2-cf>0. \tag*{(294)} W = − a , u := a 2 − c f , U = u , V = − a + u c , a 2 − c f > 0. (294)
The exact differential identities are
2 u u ′ = − c f ′ , 2 ( u ′ ) 2 + 2 u u ′ ′ = − c f ′ ′ (295) \begin{aligned}
2uu'&=-cf', & 2(u')^2+2uu''&=-cf'' \tag*{(295)}
\end{aligned} 2 u u ′ = − c f ′ , 2 ( u ′ ) 2 + 2 u u ′′ = − c f ′′ (295)
U ′ = − c f ′ 2 u , U ′ ′ = − c f ′ ′ 2 u − c 2 ( f ′ ) 2 4 u 3 (296) \begin{aligned}
U'&=-\frac{cf'}{2u}, & U''&=-\frac{cf''}{2u}-\frac{c^2(f')^2}{4u^3} \tag*{(296)}
\end{aligned} U ′ = − 2 u c f ′ , U ′′ = − 2 u c f ′′ − 4 u 3 c 2 ( f ′ ) 2 (296)
V ′ = − c 2 f ′ 2 u ( a + u ) 2 , V ′ ′ = − c 2 f ′ ′ 2 u ( a + u ) 2 − c 3 ( f ′ ) 2 ( a + 3 u ) 4 u 3 ( a + u ) 3 (297) \begin{aligned}
V'&=-\frac{c^2f'}{2u(a+u)^2}, & V''&=-\frac{c^2f''}{2u(a+u)^2}-\frac{c^3(f')^2(a+3u)}{4u^3(a+u)^3} \tag*{(297)}
\end{aligned} V ′ = − 2 u ( a + u ) 2 c 2 f ′ , V ′′ = − 2 u ( a + u ) 2 c 2 f ′′ − 4 u 3 ( a + u ) 3 c 3 ( f ′ ) 2 ( a + 3 u ) (297)
Thus f V = u − a fV=u-a f V = u − a , W ′ ≡ 0 W'\equiv0 W ′ ≡ 0 , f V 2 − 2 U V = c fV^2-2UV=c f V 2 − 2 U V = c and I A ( W ) = 0 I_A^{(W)}=0 I A ( W ) = 0 . The symmetric Ricci contractions are precisely (287) , although the full affine curvature differs from that of the constant-U U U representative.
Four-index contraction and algebraic reduction. Define D λ ρ μ ν : = R λ ρ μ ν ( Γ ) − R ˚ λ ρ μ ν D^\lambda{}_{\rho\mu\nu}:=R^\lambda{}_{\rho\mu\nu}(\Gamma)-\mathring{R}^\lambda{}_{\rho\mu\nu} D λ ρ μν := R λ ρ μν ( Γ ) − R ˚ λ ρ μν . The generic Appendix B.3 decomposition reads
K Γ = K L C + 2 R ˚ ⋅ D + D ⋅ D , 2 R ˚ ⋅ D = 2 V ( f V − 2 U ) r 2 ( f + r f ′ − 1 ) . (298) K_\Gamma=K_{\mathrm{LC}}+2\mathring{R}\mathbin{\cdot}D+D\mathbin{\cdot}D,\qquad
2\mathring{R}\mathbin{\cdot}D=\frac{2V(fV-2U)}{r^2}(f+rf'-1). \tag*{(298)} K Γ = K LC + 2 R ˚ ⋅ D + D ⋅ D , 2 R ˚ ⋅ D = r 2 2 V ( f V − 2 U ) ( f + r f ′ − 1 ) . (298)
Since V ( f V − 2 U ) = c V(fV-2U)=c V ( f V − 2 U ) = c and f + r f ′ − 1 = Λ ∗ r 2 f+rf'-1=\Lambda_*r^2 f + r f ′ − 1 = Λ ∗ r 2 , the cross term is 2 c Λ ∗ 2c\Lambda_* 2 c Λ ∗ . For reproducibility, introduce
P = 3 f 2 V 4 + 8 f 2 ( V ′ ) 2 + 8 f f ′ V V ′ − 12 f U V 3 − 16 f U ′ V ′ + 4 ( f ′ ) 2 V 2 (299) \mathcal{P}=3f^2V^4+8f^2(V')^2+8ff'VV'-12fUV^3-16fU'V'+4(f')^2V^2 \tag*{(299)} P = 3 f 2 V 4 + 8 f 2 ( V ′ ) 2 + 8 f f ′ V V ′ − 12 f U V 3 − 16 f U ′ V ′ + 4 ( f ′ ) 2 V 2 (299)
P + 8 f ′ U V ′ − 8 f ′ U ′ V − 4 ( U ′ ) 2 + 12 U 2 V 2 , (300) \mathcal{P}+8f'UV'-8f'U'V-4(U')^2+12U^2V^2, \tag*{(300)} P + 8 f ′ U V ′ − 8 f ′ U ′ V − 4 ( U ′ ) 2 + 12 U 2 V 2 , (300)
so that the unrestricted axial identity is
D ⋅ D = P 4 − 4 U U ′ r − 8 U 2 r 2 . (301) D\mathbin{\cdot}D=\frac{\mathcal{P}}{4}-\frac{4UU'}{r}-\frac{8U^2}{r^2}. \tag*{(301)} D ⋅ D = 4 P − r 4 U U ′ − r 2 8 U 2 . (301)
For (294) the substitutions u 2 = a 2 − c f u^2=a^2-cf u 2 = a 2 − c f , 2 u u ′ = − c f ′ 2uu'=-cf' 2 u u ′ = − c f ′ yield
P 4 = 3 c 2 4 − 3 c 2 ( f ′ ) 2 4 ( a 2 − c f ) , − 4 U U ′ r = 2 c f ′ r . (302) \frac{\mathcal{P}}{4}=\frac{3c^2}{4}-\frac{3c^2(f')^2}{4(a^2-cf)},\qquad
-\frac{4UU'}{r}=\frac{2cf'}{r}. \tag*{(302)} 4 P = 4 3 c 2 − 4 ( a 2 − c f ) 3 c 2 ( f ′ ) 2 , − r 4 U U ′ = r 2 c f ′ . (302)
Consequently,
K Γ ( W ) = K L C + 2 c Λ ∗ + 3 c 2 4 − 3 c 2 ( f ′ ) 2 4 ( a 2 − c f ) + 2 c f ′ r − 8 ( a 2 − c f ) r 2 . (303) K_\Gamma^{(W)}=K_{\mathrm{LC}}+2c\Lambda_*+\frac{3c^2}{4}-\frac{3c^2(f')^2}{4(a^2-cf)}+\frac{2cf'}{r}-\frac{8(a^2-cf)}{r^2}. \tag*{(303)} K Γ ( W ) = K LC + 2 c Λ ∗ + 4 3 c 2 − 4 ( a 2 − c f ) 3 c 2 ( f ′ ) 2 + r 2 c f ′ − r 2 8 ( a 2 − c f ) . (303)
The Lorentzian four-index contraction is not required to be nonnegative. Its distinction from the Ricci contractions is essential for Riemann–Cartan curvature, which lacks some Levi–Civita index-pair symmetries [2 , 3 ] .
Invariant comparison and horizon limits. The two principal-root representatives are simultaneously real on
D c o m m o n = { r > 0 : ∣ c f ( r ) ∣ < a 2 } . (304) \mathcal{D}_{\mathrm{common}}=\{r>0:\ |cf(r)|<a^2\}. \tag*{(304)} D common = { r > 0 : ∣ c f ( r ) ∣ < a 2 } . (304)
Subtracting (292) from (303) gives
Δ K : = K Γ ( W ) − K Γ ( U ) = 2 c Λ ∗ + 6 c f + 2 c r 2 − c 2 ( f ′ ) 2 4 ( 3 a 2 − c f + 2 a 2 + c f ) . (305) \Delta K:=K_\Gamma^{(W)}-K_\Gamma^{(U)}
=2c\Lambda_*+\frac{6cf+2c}{r^2}-\frac{c^2(f')^2}{4}\left(\frac{3}{a^2-cf}+\frac{2}{a^2+cf}\right). \tag*{(305)} Δ K := K Γ ( W ) − K Γ ( U ) = 2 c Λ ∗ + r 2 6 c f + 2 c − 4 c 2 ( f ′ ) 2 ( a 2 − c f 3 + a 2 + c f 2 ) . (305)
In particular, I A ( W ) = 0 I_A^{(W)}=0 I A ( W ) = 0 while I A ( U ) = c 2 ( f ′ ) 2 / [ 8 ( a 2 + c f ) ] I_A^{(U)}=c^2(f')^2/[8(a^2+cf)] I A ( U ) = c 2 ( f ′ ) 2 / [ 8 ( a 2 + c f )] ; thus for c f ′ ≠ 0 cf'\ne0 c f ′ = 0 the connections are distinguished by a scalar curvature invariant, regardless of possible isolated zeros of Δ K \Delta K Δ K . At a finite Killing horizon r h > 0 r_h>0 r h > 0 with f ( r h ) = 0 f(r_h)=0 f ( r h ) = 0 ,
U h ( U ) = U h ( W ) = a , V h ( U ) = V h ( W ) = − c 2 a . (306) U_h^{(U)}=U_h^{(W)}=a,\qquad V_h^{(U)}=V_h^{(W)}=-\frac{c}{2a}. \tag*{(306)} U h ( U ) = U h ( W ) = a , V h ( U ) = V h ( W ) = − 2 a c . (306)
( U ( U ) ) h ′ = 0 , ( U ( W ) ) h ′ = − c f h ′ 2 a . (307) (U^{(U)})'_h=0,\qquad(U^{(W)})'_h=-\frac{cf'_h}{2a}. \tag*{(307)} ( U ( U ) ) h ′ = 0 , ( U ( W ) ) h ′ = − 2 a c f h ′ . (307)
( V ( U ) ) h ′ = c 2 f h ′ 8 a 3 , ( V ( W ) ) h ′ = − c 2 f h ′ 8 a 3 . (308) (V^{(U)})'_h=\frac{c^2f'_h}{8a^3},\qquad(V^{(W)})'_h=-\frac{c^2f'_h}{8a^3}. \tag*{(308)} ( V ( U ) ) h ′ = 8 a 3 c 2 f h ′ , ( V ( W ) ) h ′ = − 8 a 3 c 2 f h ′ . (308)
Δ K ( r h ) = 2 c Λ ∗ + 2 c r h 2 − 5 c 2 ( f h ′ ) 2 4 a 2 . (309) \Delta K(r_h)=2c\Lambda_*+\frac{2c}{r_h^2}-\frac{5c^2(f'_h)^2}{4a^2}. \tag*{(309)} Δ K ( r h ) = 2 c Λ ∗ + r h 2 2 c − 4 a 2 5 c 2 ( f h ′ ) 2 . (309)
The radicands are both a 2 > 0 a^2 > 0 a 2 > 0 at the horizon, hence all these functions, the torsion three-form, the contorsion, and the reconstructed connection are smooth on sufficiently small advanced-chart neighborhoods of any finite-radius horizon. This does not establish regularity at r = 0 r = 0 r = 0 , at radicand zeros, across a bifurcation sphere, or on a maximal extension [30 , 32 ] .
Degenerate-horizon specialization. For a degenerate horizon f ( r h ) = f ′ ( r h ) = 0 f(r_h) = f'(r_h) = 0 f ( r h ) = f ′ ( r h ) = 0 , the invariant difference in Eq. (305) reduces algebraically to
Δ K ∣ r h , f h ′ = 0 = 2 c Λ ∗ + 2 c r h 2 . (310) \Delta K\big|_{r_h,f'_h=0} = 2c\Lambda_* + \frac{2c}{r_h^2}. \tag*{(310)} Δ K r h , f h ′ = 0 = 2 c Λ ∗ + r h 2 2 c . (310)
The dependence on the profile parameter a a a therefore drops out of this particular scalar comparison at the degenerate horizon. For an actual double root of the Einstein lapse, the stronger first-jet coincidence in Eq. (210) implies equality of the full affine Riemann tensors at r = r h r = r_h r = r h , not merely equality of their Kretschmann contractions. This is a pointwise statement in the common regular chart and does not establish local equivalence of the connections: Eq. (211) permits distinct second jets of the torsion profiles, and affine curvature invariants generally differ away from the horizon [30 , 32 ] .
On-shell simplification at a double horizon. For the particular Einstein lapse f = 1 − 2 M / r + Λ ∗ r 2 / 3 f = 1 - 2M/r + \Lambda_*r^2/3 f = 1 − 2 M / r + Λ ∗ r 2 /3 , f ( r h ) = f ′ ( r h ) = 0 f(r_h) = f'(r_h) = 0 f ( r h ) = f ′ ( r h ) = 0 fixes Λ ∗ = − r h − 2 \Lambda_* = -r_h^{-2} Λ ∗ = − r h − 2 and M = r h / 3 M = r_h/3 M = r h /3 . Consequently, Eq. (310) reduces to
Δ K ∣ r h , f h = f h ′ = 0 = 2 c ( Λ ∗ + 1 r h 2 ) = 0. (311) \Delta K\big|_{r_h,f_h=f'_h=0} = 2c\left(\Lambda_* + \frac{1}{r_h^2}\right) = 0. \tag*{(311)} Δ K r h , f h = f h ′ = 0 = 2 c ( Λ ∗ + r h 2 1 ) = 0. (311)
This cancellation is special to an actual double root of the specified lapse; Eq. (310) remains the more general algebraic specialization. The second-jet differences do not contradict the pointwise coincidence of the affine connections and their curvature tensors at the degenerate horizon. Instead, they show that the corresponding radial derivatives of the affine curvature need not coincide. Consequently, the two connections can remain locally distinguishable away from the horizon, despite the pointwise equalities established above.
Null-norm and linear exceptional limits. For c = 0 c = 0 c = 0 , the two selected representatives coincide exactly:
U = a , V = 0 , W = − a , T ∗ = 0 , I A = 0 , K Γ = K L C − 8 a 2 r 2 . (312) U = a,\quad V = 0,\quad W = -a,\quad T_* = 0,\quad I_A = 0,\quad K_\Gamma= K_{\mathrm{LC}} - \frac{8a^2}{r^2}. \tag*{(312)} U = a , V = 0 , W = − a , T ∗ = 0 , I A = 0 , K Γ = K LC − r 2 8 a 2 . (312)
This is nonzero null torsion when a > 0 a > 0 a > 0 , not the torsion-free connection. For γ ≠ 0 \gamma\ne0 γ = 0 this branch requires α + β = 0 \alpha+ \beta= 0 α + β = 0 . In the exceptional linear case γ = 0 \gamma= 0 γ = 0 , β = − α \beta= -\alpha β = − α , the metric is Ricci flat but the axial torsion norm need not be constant; the generic identities above still apply, while the fixed-c c c representatives form only a subset of that broader solution space.
Verification scope. Equations (286)–(288) follow from the direct coordinate contraction of Appendix § 14 .§ 2 . Equations (292) and (303) are specializations of the unrestricted four-index polynomial in Appendix § 14 .§ 3 ; (305) is their exact algebraic difference. These checks establish consistency with the stated curvature identities; they should not be described as a new, independent exhaustive 64-component curvature reconstruction or as a perturbative stability analysis. The distinction between varying the full theory and restricting its action before variation remains essential [28 , 29 , 2 ] .
Unrestricted metric and affine variations and exact algebraic checks
Independent variables and torsion invariant
The independent gravitational variables are the symmetric inverse metric g μ ν g^{\mu\nu} g μν (ten components) and the ordered connection Γ μ ν λ \Gamma^\lambda_{\mu\nu} Γ μν λ (64 components). Neither metric compatibility nor total antisymmetry of torsion is imposed during variation. Define
X λ μ ν = 2 Γ λ [ μ ν ] , t ν = X λ λ ν , Q λ μ ν = − ∇ λ g μ ν . (313) X^\lambda{}_{\mu\nu} = 2\Gamma^\lambda{}_{[\mu\nu]},\qquad t_\nu= X^\lambda{}_{\lambda\nu},\qquad Q_{\lambda\mu\nu} = -\nabla_\lambda g_{\mu\nu}. \tag*{(313)} X λ μν = 2 Γ λ [ μν ] , t ν = X λ λ ν , Q λ μν = − ∇ λ g μν . (313)
with the torsion scalar of the manuscript
T = 1 4 X λ μ ν X λ μ ν + 1 2 X λ μ ν X ν μ λ − t μ t μ . (314) T = \frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu} + \frac{1}{2}X_{\lambda\mu\nu}X^{\nu\mu\lambda} - t_\mu t^\mu. \tag*{(314)} T = 4 1 X λ μν X λ μν + 2 1 X λ μν X ν μ λ − t μ t μ . (314)
In particular, T T T is not the matter stress-tensor trace. The metric–affine framework and the distinction between constrained and unconstrained variations are discussed in [5 , 6 , 1 , 2 , 3 , 33 , 34 ] .
Full fixed-connection metric variation
At fixed Γ \Gamma Γ , X b c a X^{a}_{\ bc} X b c a and t b = X a b a t_b = X^{a}_{\ ab} t b = X ab a are fixed, whereas X a b c = g a d X b c d X_{abc} = g_{ad}X^{d}_{\ bc} X ab c = g a d X b c d varies. Prior to imposing any torsion restriction, (314) is exactly
T = 1 4 g a d g b p g c q X b c a X p q d + 1 2 g b p X b c a X p a c − g b c t b t c . (315) T = \frac{1}{4}g_{ad}g^{bp}g^{cq}X^{a}_{\ bc}X^{d}_{\ pq} + \frac{1}{2}g^{bp}X^{a}_{\ bc}X^{c}_{\ pa} - g^{bc}t_b t_c. \tag*{(315)} T = 4 1 g a d g b p g c q X b c a X pq d + 2 1 g b p X b c a X p a c − g b c t b t c . (315)
The elementary identities δ g a b = − g a μ g b ν δ g μ ν \delta g_{ab} = -g_{a\mu}g_{b\nu}\delta g^{\mu\nu} δ g ab = − g a μ g b ν δ g μν and δ g a b = δ a ν δ b μ δ g μ ν \delta g^{ab} = \delta^\nu_a\delta^\mu_b\delta g^{\mu\nu} δ g ab = δ a ν δ b μ δ g μν give the unrestricted derivative H μ ν = ( ∂ T / ∂ g μ ν ) Γ H_{\mu\nu} = \left(\partial T/\partial g^{\mu\nu}\right)_\Gamma H μν = ( ∂ T / ∂ g μν ) Γ :
H μ ν = − 1 4 g a ( μ g ν ) d g b p g c q X b c a X p q d + 1 4 g a d g c q X ( μ ∣ c ∣ a X ν ) q d + 1 4 g a d g b p X b ( μ a X p ν ) d + 1 2 X ( μ ∣ c ∣ a X ν ) a c − t μ t ν . (316) \begin{aligned}
H_{\mu\nu} ={}& -\frac{1}{4}g_{a(\mu}g_{\nu)d}g^{bp}g^{cq}X^{a}_{\ bc}X^{d}_{\ pq}
+\frac{1}{4}g_{ad}g^{cq}X^{a}_{\ (\mu|c|}X^{d}_{\nu)q} \\
&+\frac{1}{4}g_{ad}g^{bp}X^{a}_{\ b(\mu}X^{d}_{p\nu)}
+\frac{1}{2}X^{a}_{\ (\mu|c|}X^{c}_{\nu)a}
-t_\mu t_\nu. \tag*{(316)}
\end{aligned} H μν = − 4 1 g a ( μ g ν ) d g b p g c q X b c a X pq d + 4 1 g a d g c q X ( μ ∣ c ∣ a X ν ) q d + 4 1 g a d g b p X b ( μ a X p ν ) d + 2 1 X ( μ ∣ c ∣ a X ν ) a c − t μ t ν . (316)
The second and third terms arise from the two distinct inverse-metric factors in the first term of Eq. (315) ; both are retained to make the unrestricted differentiation transparent. On a metric-compatible background with X a b c = X [ a b c ] X_{abc} = X_{[abc]} X ab c = X [ ab c ] , the unrestricted expression reduces to
H μ ν ∣ a x i a l = − 1 4 X μ ρ σ X ν ρ σ , g μ ν H μ ν = T = − 1 4 X λ μ ν X λ μ ν . (317) H_{\mu\nu}\big|_{\mathrm{axial}} = -\frac{1}{4}X_{\mu\rho\sigma}X_{\nu}^{\ \rho\sigma},\qquad
g^{\mu\nu}H_{\mu\nu} = T = -\frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu}. \tag*{(317)} H μν axial = − 4 1 X μ ρ σ X ν ρ σ , g μν H μν = T = − 4 1 X λ μν X λ μν . (317)
The trace also follows from the degree-one homogeneity of (315) in g μ ν g^{\mu\nu} g μν with the mixed torsion held fixed. At a Lorentz-orthonormal point with four arbitrary axial components, all ten independent metric-variation directions were compared with (317) : the four diagonal residuals and six off-diagonal residuals vanish exactly. For off-diagonal variables the directional derivative is 2 H μ ν 2H_{\mu\nu} 2 H μν , not H μ ν H_{\mu\nu} H μν .
For the vacuum action S = ( 2 κ ) − 1 ∫ d 4 x − g F ( R , T ) S = (2\kappa)^{-1}\int d^4x\sqrt{-g}F(R,T) S = ( 2 κ ) − 1 ∫ d 4 x − g F ( R , T ) , the metric equation is
F R R ( μ ν ) − 1 2 g μ ν F + F T H μ ν = 0. (318) F_R R_{(\mu\nu)} - \frac{1}{2}g_{\mu\nu}F + F_T H_{\mu\nu} = 0. \tag*{(318)} F R R ( μν ) − 2 1 g μν F + F T H μν = 0. (318)
Full connection variation and Palatini coefficient
We vary all 64 ordered connection components independently, keeping g μ ν g_{\mu\nu} g μν fixed. Neither metric compatibility nor total antisymmetry is imposed in taking the functional derivatives. We use ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ \nabla_\nu V^\lambda= \partial_\nu V^\lambda+ \Gamma^\lambda_{\ \rho\nu}V^\rho ∇ ν V λ = ∂ ν V λ + Γ ρ ν λ V ρ , X μ ν λ = 2 Γ [ μ ν ] λ X^\lambda_{\ \mu\nu} = 2\Gamma^\lambda_{\ [\mu\nu]} X μν λ = 2 Γ [ μν ] λ and t μ = X λ μ λ t_\mu= X^\lambda_{\ \lambda\mu} t μ = X λ μ λ , as in § 2 . The distinction between the unrestricted metric–affine variation and its evaluation on a Riemann–Cartan subsector is standard in the metric–affine literature [2 , 6 , 8 , 5 ] .
Palatini contribution. For W = − g W = \sqrt{-g} W = − g , integration by parts of the curvature variation, with the convention R ρ ν = R ρ λ ν λ R_{\rho\nu} = R^\lambda_{\ \rho\lambda\nu} R ρ ν = R ρ λ ν λ , gives
δ Γ ( W R ) = W P λ μ ν δ Γ μ ν λ + ∂ σ ( ⋯ ) , (319) \delta_\Gamma(WR) = W P_\lambda^{\ \mu\nu}\delta\Gamma^\lambda_{\ \mu\nu} + \partial_\sigma(\cdots), \tag*{(319)} δ Γ ( W R ) = W P λ μν δ Γ μν λ + ∂ σ ( ⋯ ) , (319)
P λ μ ν = − 1 W ∂ λ ( W g μ ν ) + δ λ ν W ∂ σ ( W g μ σ ) + δ λ ν g ρ σ Γ ρ σ μ + g μ ν Γ ρ λ ρ − g μ ρ Γ λ ρ ν − g μ ρ Γ λ ρ ν . (320) \begin{aligned}
P_\lambda^{\ \mu\nu} ={}&
-\frac{1}{W}\partial_\lambda(Wg^{\mu\nu})
+\frac{\delta^\nu_\lambda}{W}\partial_\sigma(Wg^{\mu\sigma})
+\delta^\nu_\lambda g^{\rho\sigma}\Gamma^\mu_{\ \rho\sigma}
+g^{\mu\nu}\Gamma^\rho_{\ \rho\lambda} \\
&-g^{\mu\rho}\Gamma^\nu_{\ \lambda\rho}
-g^{\mu\rho}\Gamma^\nu_{\ \lambda\rho}. \tag*{(320)}
\end{aligned} P λ μν = − W 1 ∂ λ ( W g μν ) + W δ λ ν ∂ σ ( W g μ σ ) + δ λ ν g ρ σ Γ ρ σ μ + g μν Γ ρ λ ρ − g μ ρ Γ λ ρ ν − g μ ρ Γ λ ρ ν . (320)
The derivative terms in (320) are indispensable: varying only the connection-quadratic part of R R R is not, in general, the Palatini variation. Formula (320) is the coordinate coefficient for the index ordering used throughout this work [2 , 6 , 8 ] .
Unrestricted derivative of the torsion scalar. The invariant is
T = 1 4 X λ μ ν X λ μ ν + 1 2 X λ μ ν X ν μ λ − t μ t μ . (321) T = \frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu} + \frac{1}{2}X_{\lambda\mu\nu}X^{\nu\mu\lambda} - t_\mu t^\mu. \tag*{(321)} T = 4 1 X λ μν X λ μν + 2 1 X λ μν X ν μ λ − t μ t μ . (321)
Treating the mixed-index X λ μ ν X^\lambda{}_{\mu\nu} X λ μν as the algebraic variable, first define the coefficient D D D of its unconstrained component variation:
D λ μ ν = 1 2 X λ μ ν + 1 2 g μ ρ X ν ρ λ + 1 2 g ν ρ X μ λ ρ − 2 δ λ μ t ν . (322) D_\lambda{}^{\mu\nu} = \frac{1}{2}X_\lambda{}^{\mu\nu} + \frac{1}{2}g^{\mu\rho}X^\nu{}_{\rho\lambda} + \frac{1}{2}g^{\nu\rho}X^\mu{}_{\lambda\rho} - 2\delta_\lambda^{\mu}t^\nu. \tag*{(322)} D λ μν = 2 1 X λ μν + 2 1 g μ ρ X ν ρ λ + 2 1 g ν ρ X μ λ ρ − 2 δ λ μ t ν . (322)
Because δ X λ μ ν = δ Γ λ μ ν − δ Γ λ ν μ \delta X^\lambda{}_{\mu\nu} = \delta\Gamma^\lambda{}_{\mu\nu} - \delta\Gamma^\lambda{}_{\nu\mu} δ X λ μν = δ Γ λ μν − δ Γ λ ν μ , its exact connection derivative is
δ Γ T = J λ μ ν δ Γ λ μ ν , (323) \delta_\Gamma T = J_\lambda{}^{\mu\nu}\delta\Gamma^\lambda{}_{\mu\nu}, \tag*{(323)} δ Γ T = J λ μν δ Γ λ μν , (323)
J λ μ ν = D λ μ ν − D λ ν μ = X λ μ ν + g μ ρ X ν ρ λ − g ν ρ X μ ρ λ + 2 δ λ ν t μ − 2 δ λ μ t ν . (324) J_\lambda{}^{\mu\nu} = D_\lambda{}^{\mu\nu} - D_\lambda{}^{\nu\mu}
= X_\lambda{}^{\mu\nu} + g^{\mu\rho}X^\nu{}_{\rho\lambda} - g^{\nu\rho}X^\mu{}_{\rho\lambda} + 2\delta_\lambda^\nu t^\mu- 2\delta_\lambda^\mu t^\nu. \tag*{(324)} J λ μν = D λ μν − D λ ν μ = X λ μν + g μ ρ X ν ρ λ − g ν ρ X μ ρ λ + 2 δ λ ν t μ − 2 δ λ μ t ν . (324)
In particular, J λ ( μ ν ) = 0 J_\lambda{}^{(\mu\nu)} = 0 J λ ( μν ) = 0 identically, for arbitrary torsion and nonmetricity. The trace terms in (324) must not be dropped before specializing to a trace-free torsion configuration.
Evaluation on the axial metric-compatible sector. Only now set Q λ μ ν = 0 Q_{\lambda\mu\nu} = 0 Q λ μν = 0 and X λ μ ν = X [ λ μ ν ] X_{\lambda\mu\nu} = X_{[\lambda\mu\nu]} X λ μν = X [ λ μν ] . Then t μ = 0 t_\mu= 0 t μ = 0 and Γ λ μ ν = Γ ˚ λ μ ν + 1 2 X λ μ ν \Gamma^\lambda{}_{\mu\nu} = \mathring{\Gamma}^\lambda{}_{\mu\nu} + \frac{1}{2}X^\lambda{}_{\mu\nu} Γ λ μν = Γ ˚ λ μν + 2 1 X λ μν . The three-form antisymmetry implies
g μ ρ X ν ρ λ = − X λ μ ν , − g ν ρ X μ ρ λ = − X λ μ ν , (325) g^{\mu\rho}X^\nu{}_{\rho\lambda} = -X_\lambda{}^{\mu\nu}, \qquad-g^{\nu\rho}X^\mu{}_{\rho\lambda} = -X_\lambda{}^{\mu\nu}, \tag*{(325)} g μ ρ X ν ρ λ = − X λ μν , − g ν ρ X μ ρ λ = − X λ μν , (325)
so that
J λ μ ν ∣ a x i a l = − X λ μ ν . (326) \left.J_\lambda{}^{\mu\nu}\right|_{\mathrm{axial}} = -X_\lambda{}^{\mu\nu}. \tag*{(326)} J λ μν ∣ axial = − X λ μν . (326)
Substituting the same connection into the complete Palatini coefficient (320) , using ∇ ˚ λ g μ ν = 0 \mathring{\nabla}_\lambda g_{\mu\nu} = 0 ∇ ˚ λ g μν = 0 , gives
P λ μ ν ∣ a x i a l = − X λ μ ν . (327) \left.P_\lambda{}^{\mu\nu}\right|_{\mathrm{axial}} = -X_\lambda{}^{\mu\nu}. \tag*{(327)} P λ μν ∣ axial = − X λ μν . (327)
Both tensors have vanishing symmetric ( μ , ν ) (\mu,\nu) ( μ , ν ) parts on this sector. These identities use the stated Ricci contraction and torsion conventions; alternative conventions may move signs between the definitions of P P P , J J J , and X X X [2 , 3 ] .
Independent component checks and field equation. We performed two exact symbolic tests. First, for a generic torsion tensor antisymmetric only in its last two indices, the directional derivative of the full scalar (321) agrees with (324) for all 64 ordered connection directions (zero nonzero residuals). Second, at a Levi–Civita normal point with diagonal Lorentzian metric and four arbitrary axial three-form components, the directly differentiated curvature and torsion expressions reproduce (327) –(326) for all 64 directions (zero nonzero residuals). The latter is a pointwise algebraic check, not a substitute for the integration by parts leading to (320) . Together with tensorial covariance, these calculations check the local specialization of the unrestricted affine equations.
For the polynomial model F ( R , T ) = α R + β T + γ T 3 F(R,T) = \alpha R + \beta T + \gamma T^3 F ( R , T ) = α R + β T + γ T 3 , the full vacuum connection equation is
α P λ μ ν + ( β + 3 γ T 2 ) J λ μ ν = 0. (328) \alpha P_\lambda{}^{\mu\nu} + (\beta+ 3\gamma T^2)J_\lambda{}^{\mu\nu} = 0. \tag*{(328)} α P λ μν + ( β + 3 γ T 2 ) J λ μν = 0. (328)
Equations (326) and (327) reduce this to
( α + β + 3 γ T 2 ) X λ μ ν = 0. (329) (\alpha+ \beta+ 3\gamma T^2)X_\lambda{}^{\mu\nu} = 0. \tag*{(329)} ( α + β + 3 γ T 2 ) X λ μν = 0. (329)
Thus a nonzero axial torsion field requires the exceptional coupling condition pointwise. For γ ≠ 0 \gamma\ne0 γ = 0 it fixes T 2 = − ( α + β ) / ( 3 γ ) T^2 = -(\alpha+ \beta)/(3\gamma) T 2 = − ( α + β ) / ( 3 γ ) ; the T = 0 T = 0 T = 0 case still permits nonzero null three-form torsion. When γ = 0 \gamma= 0 γ = 0 and β = − α \beta= -\alpha β = − α , the axial connection equation places no algebraic restriction on T T T . No affine-flatness condition has been used [5 , 2 ] .
Exceptional branch and reduction of the metric equation
For
F ′ ( R , T ) = α R + β T + γ T 3 , F R = α , F T = Ψ ( T ) = β + 3 γ T 2 , (330) F'(R,T)=\alpha R+\beta T+\gamma T^3,\qquad F_R=\alpha,\qquad F_T=\Psi(T)=\beta+3\gamma T^2, \tag*{(330)} F ′ ( R , T ) = α R + β T + γ T 3 , F R = α , F T = Ψ ( T ) = β + 3 γ T 2 , (330)
the full affine equation is α P λ μ ν + Ψ ( T ) J λ μ ν = 0 \alpha P_\lambda{}^{\mu\nu}+\Psi(T)J_\lambda{}^{\mu\nu}=0 α P λ μν + Ψ ( T ) J λ μν = 0 . The nonzero axial torsion requires
[ α + β + 3 γ T 2 ] X λ μ ν = 0 , T ∗ 2 = − α + β 3 γ ≥ 0 ( γ ≠ 0 ) . (331) [\alpha+\beta+3\gamma T^2]X^{\lambda\mu\nu}=0,\qquad T_*^2=-\frac{\alpha+\beta}{3\gamma}\geq0\quad(\gamma\ne0). \tag*{(331)} [ α + β + 3 γ T 2 ] X λ μν = 0 , T ∗ 2 = − 3 γ α + β ≥ 0 ( γ = 0 ) . (331)
The value T ∗ = 0 T_*=0 T ∗ = 0 is compatible with nonzero null axial torsion. When γ = 0 \gamma=0 γ = 0 , the exceptional coupling is β = − α \beta=-\alpha β = − α , without an algebraic restriction on the torsion norm.
The exact axial curvature identities checked independently in Appendix § 14 are
R ( μ ν ) = R ˚ μ ν + H μ ν , R ( Γ ) = R ˚ + T . (332) R_{(\mu\nu)}=\mathring{R}_{\mu\nu}+H_{\mu\nu},\qquad R(\Gamma)=\mathring{R}+T. \tag*{(332)} R ( μν ) = R ˚ μν + H μν , R ( Γ ) = R ˚ + T . (332)
Inserting them into (318) gives
α G ˚ μ ν + [ α + Ψ ( T ) ] H μ ν − 1 2 g μ ν [ ( α + β ) T + γ T 3 ] = 0. (333) \alpha\mathring{G}_{\mu\nu}+[\alpha+\Psi(T)]H_{\mu\nu}-\frac{1}{2}g_{\mu\nu}\left[(\alpha+\beta)T+\gamma T^3\right]=0. \tag*{(333)} α G ˚ μν + [ α + Ψ ( T )] H μν − 2 1 g μν [ ( α + β ) T + γ T 3 ] = 0. (333)
On the exceptional branch, Ψ ( T ∗ ) = − α \Psi(T_*)=-\alpha Ψ ( T ∗ ) = − α and ( α + β ) T ∗ = − 3 γ T ∗ 3 (\alpha+\beta)T_*=-3\gamma T_*^3 ( α + β ) T ∗ = − 3 γ T ∗ 3 . Thus, for α ≠ 0 \alpha\ne0 α = 0 ,
G ˚ μ ν + Λ ∗ g μ ν = 0 , Λ ∗ : = γ T ∗ 3 α , R ( Γ ) = 4 Λ ∗ + T ∗ . (334) \mathring{G}_{\mu\nu}+\Lambda_*g_{\mu\nu}=0,\qquad\Lambda_*:=\frac{\gamma T_*^3}{\alpha},\qquad R(\Gamma)=4\Lambda_*+T_*. \tag*{(334)} G ˚ μν + Λ ∗ g μν = 0 , Λ ∗ := α γ T ∗ 3 , R ( Γ ) = 4 Λ ∗ + T ∗ . (334)
These statements are local and require neither ∇ ˚ X = 0 \mathring{\nabla}X=0 ∇ ˚ X = 0 nor R λ ρ μ ν ( Γ ) = 0 R^\lambda{}_{\rho\mu\nu}(\Gamma)=0 R λ ρ μν ( Γ ) = 0 . They do not imply global extendibility, perturbative stability, or that distinct torsion profiles represent distinct observable degrees of freedom. For the role of symmetry reduction and its limitations, see [28 , 29 ] ; for black-hole geometry see [30 , 32 ] .
Reproducibility, independent checks, and limitations
The results of this work combine covariant identities, unrestricted variational calculations, and independent symbolic component checks. These levels of verification should be distinguished: a coordinate calculation can test a tensor identity, but it does not replace the variational derivation or establish the global properties of a solution. This distinction is particularly important in metric–affine gravity, where restricting the connection before variation can discard field equations [2 , 3 , 5 , 6 ] . Likewise, a symmetry-reduced action need not reproduce all unrestricted Euler–Lagrange equations without the relevant hypotheses of symmetric criticality [28 , 29 ] .
Unrestricted metric variation. The starting point is the torsion invariant defined for an arbitrary connection,
T = 1 4 X λ μ ν X λ μ ν + 1 2 X λ μ ν X ν μ λ − t μ t μ , t μ = X λ λ μ . (335) T=\frac{1}{4}X_{\lambda\mu\nu}X^{\lambda\mu\nu}+\frac{1}{2}X_{\lambda\mu\nu}X^{\nu\mu\lambda}-t_\mu t^\mu,\qquad t_\mu=X^\lambda{}_{\lambda\mu}. \tag*{(335)} T = 4 1 X λ μν X λ μν + 2 1 X λ μν X ν μ λ − t μ t μ , t μ = X λ λ μ . (335)
The mixed-index components X λ μ ν X^\lambda{}_{\mu\nu} X λ μν are held fixed in the metric variation, since the independent connection is fixed. The ten independent directions of the symmetric inverse metric were tested at a Lorentz-orthonormal point, using four arbitrary components of an axial three-form. The four diagonal and six off-diagonal residuals vanish algebraically when compared with the axial specialization of the unrestricted derivative,
H μ ν : = ∂ T ∂ g μ ν ∣ Γ , H μ ν ∣ a x i a l = − 1 4 X μ ρ σ X ν ρ σ , g μ ν H μ ν = T . (336) H_{\mu\nu}:=\left.\frac{\partial T}{\partial g^{\mu\nu}}\right|_{\Gamma},\qquad H_{\mu\nu}\big|_{\mathrm{axial}}=-\frac{1}{4}X_{\mu\rho\sigma}X_\nu{}^{\rho\sigma},\qquad g^{\mu\nu}H_{\mu\nu}=T. \tag*{(336)} H μν := ∂ g μν ∂ T Γ , H μν axial = − 4 1 X μ ρ σ X ν ρ σ , g μν H μν = T . (336)
For an off-diagonal symmetric metric variable, the directional derivative is 2 H μ ν 2H_{\mu\nu} 2 H μν rather than H μ ν H_{\mu\nu} H μν . The component test supports the algebraic identity; its tensorial derivation is given in Appendix § 15 .§ 2 .
Unrestricted connection variation. All 4 3 = 64 4^3 = 64 4 3 = 64 ordered components Γ μ ν λ \Gamma^\lambda_{\mu\nu} Γ μν λ are independent before the Riemann–Cartan and axial conditions are imposed. Two separate checks were used. First, the direct directional derivatives of (335) with respect to the ordered connection components agree with the unrestricted torsion coefficient J λ μ ν J_\lambda{}^{\mu\nu} J λ μν of Appendix § 15 .§ 4 , with identically vanishing algebraic residuals. In particular,
J λ ( μ ν ) = 0 (337) J_\lambda{}^{(\mu\nu)} = 0 \tag*{(337)} J λ ( μν ) = 0 (337)
holds before axial specialization. Second, at a Levi–Civita normal point, direct component differentiation of the curvature and torsion terms agrees with the Palatini and torsion coefficients for all 64 ordered directions. The analytic integration by parts in Appendix § 15 .§ 4 is essential to this comparison: differentiating only the connection-quadratic terms of the Ricci scalar would omit contributions to the full Palatini coefficient. On a metric-compatible, totally antisymmetric torsion background the independently derived coefficients satisfy
P λ μ ν ∣ a x i a l = J λ μ ν ∣ a x i a l = − X λ μ ν , P λ ( μ ν ) ∣ a x i a l = 0. (338) P_\lambda{}^{\mu\nu}\big|_{\rm axial}=J_\lambda{}^{\mu\nu}\big|_{\rm axial}=-X_\lambda{}^{\mu\nu},\qquad P_\lambda{}^{(\mu\nu)}\big|_{\rm axial}=0. \tag*{(338)} P λ μν axial = J λ μν axial = − X λ μν , P λ ( μν ) axial = 0. (338)
The normal-point check tests this local specialization; it does not replace the general integration-by-parts identity [2 , 5 , 6 ] .
Independent spherical curvature reconstruction. Appendix § 14 reconstructs the full connection from the advanced Eddington–Finkelstein metric and the axial three-form,
d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 d Ω 2 , τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ . (339) ds^2=f(r)\,dv^2-2\,dv\,dr-r^2d\Omega^2,\qquad\tau=r^2\sin\theta\,[U(r)\,dv-V(r)\,dr]\wedge d\theta\wedge d\phi. \tag*{(339)} d s 2 = f ( r ) d v 2 − 2 d v d r − r 2 d Ω 2 , τ = r 2 sin θ [ U ( r ) d v − V ( r ) d r ] ∧ d θ ∧ d ϕ . (339)
The Ricci components and their contractions were checked with the polar angle θ \theta θ left symbolic. Separately, the full four-index affine Kretschmann scalar
K Γ = R λ ρ μ ν ( Γ ) R λ ρ μ ν ( Γ ) (340) K_\Gamma=R_{\lambda\rho\mu\nu}(\Gamma)R^{\lambda\rho\mu\nu}(\Gamma) \tag*{(340)} K Γ = R λ ρ μν ( Γ ) R λ ρ μν ( Γ ) (340)
was calculated from the reconstructed connection at θ = π / 2 \theta=\pi/2 θ = π /2 and independently at θ = π / 3 \theta=\pi/3 θ = π /3 . In both computations, f f f , f ′ f' f ′ , f ′ ′ f'' f ′′ , U U U , U ′ U' U ′ , V V V , V ′ V' V ′ were retained as algebraically independent quantities, before imposing either the constant-norm condition or the field equations. Both expressions reduce to the unrestricted polynomial displayed in Appendix § 14 .§ 3 , and their difference vanishes identically. This is a two-angle component check, not a third independent computation with a symbolic generic angle. Angular independence also follows geometrically from the spherical symmetry of the scalar away from polar coordinate singularities [30 , 32 ] .
An additional cross-check uses the curvature decomposition R ( Γ ) = R ˚ + D R(\Gamma)=\mathring{R}+D R ( Γ ) = R ˚ + D at the four-index level:
K Γ = K L C + 2 R ˚ ⋅ D + D ⋅ D . (341) K_\Gamma=K_{\rm LC}+2\mathring{R}\cdot D+D\cdot D. \tag*{(341)} K Γ = K LC + 2 R ˚ ⋅ D + D ⋅ D . (341)
The separately contracted cross term and distortion-square term reproduce the unrestricted expression of Appendix § 14 .§ 3 . Only subsequently is the exceptional branch imposed,
f = 1 − 2 M r + Λ ∗ r 2 3 , f V 2 − 2 U V = c , (342) f=1-\frac{2M}{r}+\frac{\Lambda_*r^2}{3},\qquad fV^2-2UV=c, \tag*{(342)} f = 1 − r 2 M + 3 Λ ∗ r 2 , f V 2 − 2 U V = c , (342)
for which 2 R ˚ ⋅ D = 2 c Λ ∗ 2\mathring{R}\cdot D=2c\Lambda_* 2 R ˚ ⋅ D = 2 c Λ ∗ . With U = a > 0 U=a>0 U = a > 0 and the smooth real root V = − c / ( a + a 2 + c f ) V=-c/(a+\sqrt{a^2+cf}) V = − c / ( a + a 2 + c f ) , the result agrees with the closed expression used in § 10.4 . The torsion-free limit U = V = 0 U=V=0 U = V = 0 yields K Γ = K L C K_\Gamma=K_{\rm LC} K Γ = K LC , while the nonzero null-norm representative c = 0 c=0 c = 0 , U = a U=a U = a , V = 0 V=0 V = 0 gives K Γ = K L C − 8 a 2 / r 2 K_\Gamma=K_{\rm LC}-8a^2/r^2 K Γ = K LC − 8 a 2 / r 2 . These are checks of the contractions, not additional field equations.
Scope and reproducibility of the independent checks. The independent symbolic tests employ the index convention of § 2 , the explicitly stated independent metric and affine variables, and exact arithmetic for residual comparisons. The fixed-connection metric derivative of T T T is tested along the ten independent symmetric inverse-metric directions, while the algebraic connection derivative of T T T and the Palatini coefficient are checked along all 64 ordered affine directions at the stated normal-point backgrounds. Ricci contractions are compared componentwise; the four-index affine Kretschmann invariant is reconstructed and compared before the fixed-norm and field-equation substitutions in the explicitly specified test configurations. These results corroborate the corresponding tensor identities but do not constitute an independent symbolic proof on every possible background or a global regularity or stability theorem. For reproducibility, the supplementary archive should contain the metric and inverse metric, component orderings, independent symbolic variables, exact input configurations, executable scripts, and machine-checkable zero-residual assertions. The checks do not establish smoothness at a bifurcation sphere, uniqueness, or physical inequivalence of all torsion profiles [2 , 3 , 30 , 32 ] .
References