Preprint
Article

This version is not peer-reviewed.

Compactly Supported Nonlinear Constraint Deformations and Linearized Curvature Growth in Ori’s Chronology Model

Submitted:

31 October 2026

Posted:

04 November 2026

You are already at the latest version

Abstract
Ori's vacuum--dust spacetime has a regular initial hypersurface whose vacuum-core portion is umbilical. The core develops a chronology horizon.Let \(\theta_0\) be the angular cutoff of the core and choose \(0<a<b\) with \(b+2\log(2+\sqrt{3})<\theta_0\). Fix a nonzero real \(p\in C_c^\infty((a,b))\) that is odd about \((a+b)/2\). We construct a smooth, nontrivial one-parameter family of finite-amplitude vacuum constraint deformations, compactly supported in a fixed annulus of the core and defined for every real \(\varepsilon\) with \(\varepsilon^2\|p^2+(p')^2\|_\infty<1\).The family satisfies both nonlinear vacuum Einstein constraints exactly within the core and extends, with Ori's unchanged dust envelope and exterior, to global Einstein--dust initial data satisfying the full constraints. We propagate its axial first variation using the established Kodama--Ishibashi formalism. Under the specified core-size and support conditions, the resulting Weyl Pontryagin first variation satisfies a logarithmic \(L^2\) lower bound as the background horizon is approached. A local hyperbolic equation and the curvature reconstruction establish finite propagation for the invariant curvature observable, even though the auxiliary master potential need not be compactly supported. A radial Wronskian gives a nonzero horizon coefficient on a set of positive spectral measure. We also compute the exact initial twist and the leading term in the initial-slice Weyl Pontryagin expansion of the finite-amplitude data. The growth estimate concerns the linearized Einstein equations, not the curvature of any fixed nonzero-amplitude development. No nonlinear chronology-protection statement is implied.
Keywords: 
;  ;  ;  ;  

MSC:  Primary 83C05; Secondary 83C25, 83C75

1. Introduction

The vacuum–dust time-machine construction of Ori [1, Secs. III–VI] starts from a regular spacelike initial hypersurface and develops a chronology horizon. This construction builds on Ori’s earlier compact-vacuum-core time-machine model [2]. The existence and geometry of this background are established results, not part of the present claims. Dietz, Dirmeier and Scherfner later established the geodesic incompleteness and analytic inextendibility of the pseudo-Schwarzschild spacetime [3, Proposition 2]. This inextendibility concerns the pseudo-Schwarzschild spacetime as a whole and does not preclude Ori’s two inequivalent analytic extensions of its pre-horizon region across the Cauchy horizon [1, Sec. I]. Dietz, Dirmeier and Scherfner also constructed an exact rotating pseudo-Kerr vacuum metric [3, “Pseudo Kerr Spacetime” section]. Its first rotational variation has an angular one-form proportional to sinh 2 ρ d ϑ = * h H 2 d ( cosh ρ ) , which is neither compactly supported nor square integrable on H 2 . That work neither constructs the annular constraint data nor proves the linearized curvature bound established here.
Rieger subsequently placed the pseudo-Schwarzschild geometry in a Misner-type causal classification and established isocausality results under explicit covering and quotient conditions [4, Proposition 7.1]. That work does not study axial Einstein perturbations of Ori’s matched vacuum–dust initial data. Growth of a linearized field near a chronology horizon likewise does not establish an obstruction to nonlinear chronology violation.
We ask whether a compactly supported axial perturbation can be tangent to an exact finite-amplitude Einstein constraint family on Ori’s initial slice while its linearized vacuum development exhibits growth of a gauge-invariant curvature observable in a region causally protected from the dust interface. The same initial family underlies both conclusions, one about the nonlinear constraints and the other about the linearized evolution. Rauch and Rendall exhibited logarithmic behavior for particular circle-invariant test-field solutions near a Cauchy horizon [5, Sec. 2, Eq. (7)]. Our curvature estimate concerns instead a gauge-invariant first variation of the Einstein field, with compactly supported data confined to a protected region of Ori’s vacuum core. The explicit nonlinear constraint family has this compactly supported axial initial datum as its derivative at zero amplitude. The resulting curvature growth remains a linearized statement and does not establish nonlinear horizon instability.
Theorem 1.1. 
Let μ > 0 , r h = 2 μ , r 0 = 3 μ and θ c = log ( 2 + 3 ) . On the angular factor H 2 , let ρ be the distance from a fixed origin o, and write B R ( o ) for the metric ball of radius R. Fix an Ori core with angular cutoff ρ = θ 0 and numbers 0 < a < b satisfying b + 2 θ c < θ 0 . Let p ∈ C c ∞ ( ( a , b ) ) be a nonzero real function odd about ( a + b ) / 2 . For every real amplitude ε with ε 2 ∥ p 2 + ( p ′ ) 2 ∥ ∞ < 1 , the core family given in (18) and (19) satisfies both nonlinear vacuum Einstein initial constraints exactly and agrees with Ori’s original core data outside the annulus. With the dust envelope and exterior unchanged, these data extend to a global Einstein–dust initial-data family satisfying the full constraints.
Its nonzero axial first variation determines a linearized vacuum Einstein solution on the protected interior. For the Weyl Pontryagin first variation P ˙ of this solution, there exist c , δ > 0 such that
P ˙ ( r ) L 2 ( B θ 0 − θ c ( o ) , d V H 2 ) ≥ c log r − r h r h ( 0 < r − r h < δ ) .
The last assertion is about the first variation at ε = 0 and thebackgroundhorizon. It is not a statement of nonlinear curvature blow-up.
The proof combines the explicit constraint factorization in Section 4, the spectral transfer in Section 6, and the local curvature calculation and finite-propagation argument in Section 7. Exact initial twist and finite-amplitude curvature information are collected afterward. Section 9 proves the combined theorem.
The particular axial master equation and metric reconstruction used here follow Kodama and Ishibashi [6, Sec. 5.2]. An especially close predecessor is the analysis of odd gravitational perturbations of spatially closed Bianchi III vacuum spacetimes by Tanimoto, Moncrief and Yasuno [7, Secs. 2, 5 and 9, Eqs. (13), (52), (88), (105) and (189)]. Writing r = t TMY + k , k = μ and τ = z in their background metric, where t TMY is their time coordinate, shows that its universal-cover geometry is locally isometric to the untruncated Ori vacuum core. In their zero circle-frequency odd sector m = 0 , the gauge-invariant variable Q = γ 6 is the axial cross-metric amplitude, up to normalization of the angular vector harmonic. Their Proposition 7 gives the full odd metric reconstruction in a gauge used there. With m = 0 its other two terms vanish and q 2 / u = 1 , so the metric variation reduces to Q ( E 6 ) a b . Under the eigenvalue correspondence λ TMY 2 = λ , their Eq. (105) reduces exactly to (60). Thus their formalism already contains the local axial evolution equation, although the spectral resolutions differ. Their compact quotient has a discrete angular spectrum, whereas H 2 has a continuous spectrum.
Tanimoto later analyzed the large-time asymptotics of Bianchi III perturbations and the stability of the background [8, Secs. 2.4–2.5]. Those large-time results concern t TMY → ∞ , not the finite t TMY ↓ k horizon limit studied here.
In the harmonic one-form sector analyzed by Tanimoto, Moncrief and Yasuno at zero circle frequency, the general solution has a finite horizon limit and may contain a subleading term proportional to ( t TMY − k ) log ( t TMY − k ) as t TMY ↓ k [7, Eq. (112)]. The harmonic one-form sector is distinct from the scalar-derived sector with λ ≥ 1 / 4 used below. Its possible logarithmic term therefore does not by itself establish the Weyl Pontryagin lower bound for the first variation arising from our compactly supported data. Their compact hyperbolic quotient differs from the truncated hyperbolic disk in Ori’s vacuum core, which is surrounded by dust. Here we construct an exact nonlinear constraint family supported in an annulus. We use its tangent as physical Cauchy data for the linearized vacuum evolution and prove a logarithmic L 2 lower bound for the resulting Weyl Pontryagin first variation on a causally protected portion of Ori’s vacuum core.
The construction of Kodama and Ishibashi builds on Mukohyama’s gauge-invariant master equations for maximally symmetric backgrounds [9] and the extension of the vector formalism to more general warped-product backgrounds by Kodama, Ishibashi and Seto [10]. Cai and Cao later formulated gauge-invariant perturbation equations directly on warped products, without requiring a spectral expansion of the transverse Laplacian [11, Secs. III–V]. Earlier gauge-invariant odd-parity perturbation theory on general spherically symmetric backgrounds goes back at least to Gerlach and Sengupta [12], who also formulated invariant odd-parity junction conditions [13]. More generally, Karlovini derived a Maxwell-like equation for the linearized twist of a nonnull, hypersurface-orthogonal Killing field without requiring spherical symmetry [14, Sec. 2, Eq. (23)]. That general equation applies locally to the original Ori circle Killing field, but does not give the present annular constraint family or the curvature lower bound for its linearized tangent. Sarbach and Tiglio developed geometric Regge–Wheeler reconstruction in horizon-penetrating Schwarzschild coordinates [15, Secs. II–III]. Martel and Poisson gave a covariant odd-parity master-field formulation, including source terms and horizon-penetrating reconstructions [16, Secs. V and VII]. Chaverra, Ortiz and Sarbach later developed covariant, gauge-invariant perturbation equations for general spherically symmetric Einstein–matter backgrounds without requiring a spherical-harmonic expansion [17, Sec. IV]. The spherical-background works of Gerlach and Sengupta, Sarbach and Tiglio, Martel and Poisson, and Chaverra, Ortiz and Sarbach do not treat Ori’s hyperbolic core. The broader Kodama–Ishibashi and Cai–Cao formalisms allow more general transverse geometries, but do not construct the annular Ori initial data considered here. We do not claim to have introduced gauge-invariant odd-parity reconstruction, junction conditions, horizon-penetrating coordinates, or nonspectral perturbation equations. Negative-curvature topological black-hole perturbations have also been studied for asymptotically anti-de Sitter backgrounds [18]. That black-hole stability problem is distinct from the Cauchy-horizon initial-value problem in the periodic Ori vacuum core considered here. The use of Pontryagin density as an axial curvature diagnostic is familiar from the Chern–Simons and nonmodal perturbation literature [19,20,21,22]. The curvature interpretation of odd-parity gauge invariants on Schwarzschild backgrounds also predates the present calculation [23, Secs. IV and VI]. Fernández Tío and Dotti also studied the divergence of proper-time derivatives of invariant curvature perturbations near the Reissner–Nordström Cauchy horizon [22, Sec. IV]. Their background and initial-value problem differ from the Ori construction studied here. Classical gluing techniques and their relation to Killing initial data are also prior art [24,25]. Delay and Mazzieri proved localized vacuum-constraint gluing under a no-KID condition, with exact agreement outside a connected-sum neck neighborhood [26, 2010 preprint, Sec. 1.2, Theorem 2]. Their theorem does not directly apply to the present annulus, which carries the inherited circle KID. Carlotto and Schoen established localized Einstein-constraint data in asymptotically flat conical regions [27]. Their construction differs from the present compactly supported annular deformation on Ori’s initial slice. Czimek and Rodnianski established obstruction-free spacelike gluing of asymptotically flat data to Kerr initial data far out in the asymptotic region under conditions on the differences of their ADM parameters [28, Corollary 2.13]. Mao, Oh and Tao subsequently constructed annular gluing for initial data sufficiently close to Minkowski data using solution operators with prescribed support. Their annular theorem assumes the averaged-charge inequality Δ E > | Δ P | , together with further charge and smallness bounds [29, Theorem 1.7 and Lemma 2.2]. Neither theorem directly asserts the existence of a nontrivial deformation supported inside the fixed Ori annulus while keeping the same Ori data on both collars, nor does it supply the linearized curvature estimate established here. Chruściel and Delay constructed exotic gluings of asymptotically hyperbolic initial data that contain regions where the data are exactly hyperbolic [30, Secs. 1–3]. This is additional precedent for localization of Einstein-constraint data, but it does not directly give either the annular family constructed here or the linearized Weyl–Pontryagin estimate on the protected portion of Ori’s vacuum core. Chruściel, Cogo and Nützi constructed a support-preserving solution operator for the linearized constant-scalar-curvature equation at hyperbolic space, modulo the finite-dimensional static-KID obstruction [31, Proposition 1.1]. Their annular interpolation near hyperbolic data initially satisfies the constant-scalar-curvature equation modulo that obstruction [31, Sec. 4, Proposition 4.1 (Informal)]. The Ori initial metric instead has the product geometry S 1 × H 2 and a nonzero second fundamental form. The cited results therefore do not directly establish the present exact family for the full vacuum constraints. More closely related nonlinear work on spacetimes with a circle symmetry includes the following. Choquet-Bruhat and Moncrief obtained solutions of the constraints and analyzed expanding vacuum spacetimes with a circle symmetry over compact hyperbolic surfaces [32, Secs. 1–3]. Choquet-Bruhat subsequently treated the unpolarized, nonzero-twist case [33, Secs. 1–2]. Gicquaud and Huneau obtained vacuum-constraint solutions with a translational circle symmetry over compact hyperbolic surfaces under a mean-curvature condition via the conformal method [34, Corollary 2.3], while Huneau studied the noncompact asymptotically flat case [35, Secs. 1–2]. Berger and Moncrief had already constructed exact, inhomogeneous circle-symmetric vacuum solutions using a target-space S L ( 2 , R ) action [36, Secs. II–III]. Geroch showed that a vacuum solution with two commuting Killing fields generates a family of new exact solutions depending on two arbitrary functions of one variable, provided a certain pair of constants vanishes [37]. Consequently, we do not claim originality for the existence of nonlinear circle-symmetric constraint solutions, nonzero twist, or exact inhomogeneous circle-symmetric vacuum solutions. We construct an annular deformation of Ori’s initial data that agrees with the inherited data on both collars and throughout the dust envelope. For its tangent, the Weyl Pontryagin first variation satisfies a logarithmic L 2 lower bound on a causally protected portion of Ori’s vacuum core. These results do not amount to a general nonlinear gluing or horizon-instability theorem.

2. The Ori Initial Slice

On Ori’s vacuum core the inherited metric can be written as
g O = − f ( r ) − 1 d r 2 + f ( r ) d τ 2 + r 2 h H 2 , f ( r ) = 1 − 2 μ r , r > 2 μ ,
where τ is a periodic coordinate, μ > 0 , and h H 2 = d ρ 2 + sinh 2 ρ d ϑ 2 . We orient the hyperbolic factor so that * h H 2 d ρ = sinh ρ d ϑ . Here ρ is the radial coordinate denoted θ in [1]. Our τ and ϑ are Ori’s periodic t and the core hyperbolic azimuthal φ , respectively, and ρ = θ 0 is the core cutoff. We write t Ori = τ when referring to Ori’s periodic coordinate, reserving t below for Gaussian normal time. In Ori’s global gluing the axial angle is instead ϕ Ori = ( 2 π / ℓ Ori ) t Ori , where ℓ Ori denotes the period of t Ori . This is a rescaling of our τ , not the core angle φ [1]. The coordinate r decreases toward the future from the initial slice to the background chronology horizon. Ori’s core initial slice is r = r 0 = 3 μ with r h = 2 μ and f 0 = 1 / 3 . The background induced metric and second fundamental form are γ 0 = f 0 d τ 2 + r 0 2 h H 2 and K 0 = ( 3 r 0 ) − 1 γ 0 . We use the sign convention K = − 1 2 L n γ , where n is the future-pointing unit normal. The global vacuum–dust matching and chronology geometry are due to Ori [1, Secs. III–VI]. Ori requires θ 0 > θ c for the protected region and then chooses the circle period subject to his construction conditions [1, Sec. V.A, Eq. (23)]. The additional inequality b + 2 θ c < θ 0 restricts attention to Ori backgrounds with enough angular room for the present data. The period does not enter the local transfer calculation.
For later reference, on a general r-slice the principal extrinsic curvatures in the convention used here are
K τ τ = f ′ 2 f , K A B = f r δ A B .
They coincide at r 0 = 3 μ , as used in the nonlinear constraint construction below.

3. An Anisotropic Axial Constraint Ansatz

Fix the inherited umbilical Ori slice described above. Thus
γ 0 = f 0 d τ 2 + σ , f 0 = 1 3 , σ = r 0 2 h H 2 , r 0 = 3 μ , k 0 = 1 3 r 0 ,
where τ is periodic. Throughout, physical axial seed scalars ϕ ∈ C c ∞ ( H 2 ) are real-valued. The generalized harmonics used in spectral decompositions may be complex. Fix a nonzero real ϕ ∈ C c ∞ ( H 2 ) and set
V = * σ d ϕ , T = d τ ⊗ V + V ⊗ d τ , S = | T | γ 0 2 = 2 f 0 − 1 | V | σ 2 .
The Hodge star on one-forms in dimension two is independent of a conformal rescaling. Because V = * σ d ϕ is coexact, it also satisfies D σ A V A = 0 . We write D and Δ σ for the Levi-Civita derivative and scalar Laplacian of σ .
For a smooth real function w on H 2 , independent of τ , and a real amplitude ε , consider
γ [ w ] = f 0 e 2 w d τ 2 + e − 2 w σ , K [ w , ε ] = k 0 γ [ w ] + ε e − w ( d τ ⊗ V + V ⊗ d τ ) .
The circle and hyperbolic factors scale in opposite directions. The factor e − w in the trace-free term makes the momentum constraint hold exactly.
Proposition 3.1. 
For every smooth τ-independent w, the initial data (6) satisfy
tr γ [ w ] K [ w , ε ] = 3 k 0 , div γ [ w ] K [ w , ε ] − 3 k 0 γ [ w ] = 0 .
Their vacuum Hamiltonian residual has the exact pointwise form
H [ w , ε ] = c 0 ( 1 − e 2 w ) − 2 e 2 w | D w | σ 2 − ε 2 e − 2 w S , c 0 = 2 r 0 2 = 6 k 0 2 .
Neither an elliptic inverse nor a truncation in ε is needed for these identities.
Proof. 
The determinant and inverse blocks of γ [ w ] are
det γ [ w ] = f 0 e − w det σ , γ τ τ = f 0 − 1 e − 2 w , γ A B = e 2 w σ A B .
Put T ˜ = ε e − w T . It is off-diagonal, so its trace vanishes identically. The momentum constraint reduces to div γ [ w ] T ˜ = 0 . For a symmetric tensor,
∇ j T ˜ j i = 1 det γ ∂ j det γ T ˜ j i − 1 2 T ˜ j k ∂ i γ j k .
The last term is zero because T ˜ is off-diagonal and γ [ w ] is block diagonal and circle-independent. For i = τ , the remaining expression is the divergence of det γ T ˜ A τ = ε f 0 det σ V A , which vanishes since D σ A V A = 0 . For i = A , only the j = τ term survives and its circle derivative vanishes. This proves the momentum identity without assuming a special radial form of ϕ or w.
Regard γ [ w ] as a one-dimensional warped product over ( H 2 , e − 2 w σ ) with fiber radius f 0 e w . The two-dimensional conformal curvature formula and the warped-product formula give
R ( γ [ w ] ) = e 2 w R ( σ ) − 2 | D w | σ 2 , R ( σ ) = − 2 r 0 2 .
The Δ σ w terms cancel exactly. This cancellation is not a linearization artifact. The squared norm of the second fundamental form is
| K [ w , ε ] | γ [ w ] 2 = 3 k 0 2 + ε 2 e − 2 w S .
Substitution into R ( γ ) + ( tr K ) 2 − | K | 2 proves (8). □

4. An Exact Compactly Supported Nonlinear Constraint Family with an Axial First Variation

The exact Hamiltonian identity above admits a quadratic reduction. We use it to construct nonzero axial first-order data as tangents to smooth, compactly supported families satisfying the full nonlinear constraints exactly. Because the axial seed varies at order ε 2 , this construction does not establish existence for the reduced equation with a freely prescribed, amplitude-independent ϕ ∈ C c ∞ ( H 2 ) .

4.1. A Quadratic Reduction of the Hamiltonian Constraint

Put Y = e 2 w > 0 in the exact Hamiltonian identity (8). Multiplication by Y gives
Y H = c 0 Y ( 1 − Y ) − 1 2 | D Y | σ 2 − ε 2 S , c 0 = 2 r 0 2 .
This is an identity, rather than an approximation. Any exact solution with S ≥ 0 and Y > 0 obeys
0 < Y ≤ 1 , ε 2 S ≤ c 0 Y ( 1 − Y ) ≤ c 0 4 .
For a fixed nonnegative source, every exact solution must therefore satisfy the necessary amplitude bound ε 2 ∥ S ∥ ∞ ≤ c 0 / 4 . This is neither an obstruction at arbitrarily small amplitude nor an existence theorem. The reduced Hamiltonian equation is first order in Y. Although its linearization at Y = 1 is multiplication by − c 0 , the squared-gradient term causes derivative loss between the usual function spaces. A direct inverse-function argument therefore does not construct the desired compactly supported family.
We choose an amplitude-dependent axial potential so that the resulting exact constraint family has a nonzero axial first variation at ε = 0 . This differs from holding the finite-amplitude axial source fixed.

4.2. A Smooth Annular Seed with an Exact Nonlinear Completion

Write the hyperbolic-plane metric in polar coordinates as h H 2 = d ρ 2 + sinh 2 ρ d ϑ 2 . The inherited Ori initial data are
γ 0 = f 0 d τ 2 + r 0 2 ( d ρ 2 + sinh 2 ρ d ϑ 2 ) , K 0 = k 0 γ 0 ,
where r 0 = 3 μ , f 0 = 1 / 3 and k 0 = 1 / ( 3 r 0 ) . These data and their matching to Ori’s exterior are inherited from [1, Sec. III.B and Secs. IV–VI].
Fix 0 < a < b < θ 0 and put ρ * = ( a + b ) / 2 . Choose a real, nonzero function p ∈ C c ∞ ( ( a , b ) ) and extend it smoothly by zero to R . Impose the reflection parity
p ( 2 ρ * − ρ ) = − p ( ρ ) , ρ ∈ R .
For example, any nonzero smooth odd bump about ρ * supported strictly inside ( a , b ) has the required parity. Define
Q ( ρ ) = p ( ρ ) 2 + p ′ ( ρ ) 2 , D p : = { ε ∈ R : ε 2 ∥ Q ∥ ∞ < 1 } .
The explicit formulas and exact constraint identities below hold for every ε ∈ D p , including ε = 0 . For the first-variation and local-development statements, choose ε * > 0 with ε * 2 ∥ Q ∥ ∞ < 1 and restrict to | ε | < ε * . Set
Y ε ( ρ ) = 1 − ε 2 p ( ρ ) 2 , w ε ( ρ ) = 1 2 log Y ε ( ρ ) , ϕ ε ( ρ ) = f 0 ∫ 0 ρ p ( s ) 1 − ε 2 Q ( s ) d s , V ε = * σ d ϕ ε .
Here σ = r 0 2 h H 2 . Let T ε = d τ ⊗ V ε + V ε ⊗ d τ . We use the anisotropic ansatz with this smoothly varying potential,
γ ε = f 0 Y ε d τ 2 + Y ε − 1 σ , K ε = k 0 γ ε + ε Y ε − 1 / 2 T ε .
Proposition 4.1. 
The family (19) is smooth in space and in ε whenever ε 2 ∥ Q ∥ ∞ < 1 . For every such ε, both nonlinear vacuum Einstein initial constraints hold exactly,
R ( γ ε ) + ( tr γ ε K ε ) 2 − | K ε | γ ε 2 = 0 , div γ ε ( K ε − tr γ ε K ε γ ε ) = 0 .
The data agree identically with ( γ 0 , K 0 ) outside S 1 × { a < ρ < b } . Their first variation is
∂ ε γ ε 0 = 0 , ∂ ε K ε 0 = d τ ⊗ ( * σ d ϕ 0 ) + ( * σ d ϕ 0 ) ⊗ d τ ≠ 0 ,
where
ϕ 0 ( ρ ) = f 0 ∫ 0 ρ p ( s ) d s ∈ C c ∞ ( H 2 ) ∖ { 0 } .
Thus this particular nonzero, compactly supported axial linearized Einstein-constraint datum is tangent to an exact, nonlinear and compactly supported vacuum-constraint family.
Proof. 
Since p is odd about ρ * while p 2 and p ′ 2 are even, the integrand p ( s ) 1 − ε 2 Q ( s ) is odd about ρ * . Its integral over ( a , b ) therefore vanishes. Consequently, ϕ ε = 0 on both ρ ≤ a and ρ ≥ b . The integrand is smooth and vanishes to all orders at the ends of its compact support. The potential extends smoothly over the hyperbolic origin, where it is identically zero. It follows that ϕ ε , V ε , w ε and all departures of (19) from the background are smooth and supported in the stated annulus. In particular, Y ε > 0 throughout the stated open amplitude range, since ε 2 ∥ Q ∥ ∞ < 1 .
The momentum constraint and constant mean curvature are exact for every such w ε and every divergence-free V ε = * σ d ϕ ε by Proposition 3.1. It remains to check the Hamiltonian constraint without any asymptotic expansion. For a radial function, the σ -gradient has norm | D Y | σ 2 = r 0 − 2 ( Y ′ ) 2 . From (18),
Y ε ′ = − 2 ε 2 p p ′ , S ε = | T ε | γ 0 2 = 2 f 0 r 0 2 ( ϕ ε ′ ) 2 = c 0 p 2 1 − ε 2 ( p 2 + p ′ 2 ) .
Substituting these expressions into (13) gives the exact pointwise identity
Y ε H = c 0 ε 2 p 2 ( 1 − ε 2 p 2 ) − 2 ε 4 r 0 2 p 2 p ′ 2 − c 0 ε 2 p 2 ( 1 − ε 2 p 2 − ε 2 p ′ 2 ) = ε 4 p 2 p ′ 2 c 0 − 2 r 0 2 = 0 .
This holds even at the zeros of p, including the annular midpoint and the two boundary collars. No division by p is involved. Since Y ε > 0 , the original Hamiltonian constraint vanishes exactly.
Finally ϕ ε = ϕ 0 + O C m ( ε 2 ) , Y ε = 1 + O C m ( ε 2 ) and Y ε − 1 / 2 = 1 + O C m ( ε 2 ) for every fixed m. Differentiating (19) at zero therefore gives (21). The tangent is nonzero since p ≠ 0 and ϕ 0 ′ = f 0 p . □

4.3. Exact Assembly with the Inherited Ori Dust Initial Data

The annular data agree with Ori’s initial data on open vacuum collars. This exact initial matching does not by itself determine the perturbed spacetime junction conditions or the evolution of the dust interface. In particular, we make no claim that the odd-parity spacetime junction conditions of Gerlach and Sengupta [13] are satisfied for a moving Ori interface.
Proposition 4.2. 
Let ( γ O , K O , m O ) denote the smooth global Einstein–dust initial data of Ori [1]. Assume that its chosen core contains the annulus a < ρ < b < θ 0 . For each ε ∈ D p , replace the original vacuum data on that annulus by (19) and leave m O , the remainder of the vacuum core, the entire dust envelope and the asymptotically flat exterior unchanged. Then the resulting global initial data satisfy the full Einstein–dust constraint equations exactly. They have the original global topology, asymptotics and matter data.
Proof. 
The new data agree with the originals on open collars at either end of the annulus. They therefore define smooth, globally matched tensors without interpolation, a surface layer, or a junction correction. In the annulus the matter density and momentum vanish and the exact vacuum constraints hold by Proposition 4.1. Elsewhere, the unchanged Ori data already satisfy the Einstein–dust constraints. The constraint equations are local in the initial data and their derivatives, so the globally matched data satisfy them everywhere. The dust fields and the asymptotically flat end remain unchanged. This is a statement only about the initial hypersurface. □
The tangential Killing initial datum ( 0 , ∂ τ ) is present throughout the deformed vacuum core, not only in the untouched collar. Both γ ε and K ε are invariant under the circle action. In particular, the no-local-KID hypothesis of the general local gluing theorem of Chruściel, Isenberg and Pollack [24, Theorem 1.1] does not hold in this collar. Our construction is an explicit solution within a special family, not an application of that theorem. Both the axial potential ϕ ε and its quadratic source S ε vary with ε , as shown in (18) and (25). The fixed-source equation for an arbitrarily prescribed ϕ ∈ C c ∞ ( H 2 ) is not covered by this construction.

4.4. Scope of the Explicit Constraint Family

The reflection condition (16) ensures that the primitive defining ϕ ε vanishes on both sides of the annulus. Without a cancellation condition, that primitive need not return to zero outside the annulus. The present construction does not establish a completion for an arbitrary compactly supported scalar seed. Although ϕ ε = ϕ 0 + O ( ε 2 ) , the potential generally depends on the amplitude. The associated quadratic source satisfies
S ε = c 0 p 2 − c 0 ε 2 p 2 ( p 2 + p ′ 2 ) ,
so this finite-amplitude family must not be confused with a formal expansion for a fixed prescribed source. The family’s first variation is nevertheless the nonzero axial datum stated above.
Exact initial matching and persistence of the original vacuum–dust interface under evolution are distinct issues. The interface is unperturbed at the initial slice, but later causal propagation may transmit changes into it. Nor does the existence of KIDs rule out our explicit solution. It only precludes the unqualified use of a gluing theorem that assumes their absence.
The construction uses a restricted ansatz for which both vacuum constraint equations hold exactly. This establishes existence within that ansatz, not a gluing theorem for arbitrarily prescribed compactly supported axial data. The role of Killing initial data in integrability of linearized constraints is classical. The background Ori core has four independent local Killing fields. They are ∂ τ and the three local generators of the H 2 isometry group. The spacetime curvature invariant R a b c d R a b c d = 48 μ 2 / r 6 forces any local Killing field to preserve r, while the distinct eigenvalues of the intrinsic three-dimensional Ricci tensor of each r-slice separate the circle and hyperbolic tangent distributions. Thus all four induced local Killing initial data ( N , X ) have N = 0 . For a compactly supported tangent ( γ ˙ , K ˙ ) = ( 0 , T 0 ) , the second variation of the momentum constraint vanishes. With the metric fixed, that constraint is linear in K. Consequently, its quadratic Taub pairing with every local KID vanishes, even though the second variation of the Hamiltonian constraint need not vanish. This is a necessary second-order check, not a general integrability criterion on Ori’s noncompact and globally matched data. On compact CMC vacuum slices, the corresponding necessary-and-sufficient criterion was proved in the one-Killing-field case by Fischer, Marsden and Moncrief [38, Theorem 3.1] and extended to several Killing fields by Arms, Marsden and Moncrief [39]. See Carlotto [40, Theorem 4.7]. Here the explicit annular family, rather than an abstract linearization-stability theorem, establishes a compactly supported nonlinear completion for the chosen tangent.
Moncrief also studied all-order symmetry-preserving formal perturbations of generalized Taub–NUT backgrounds [41]. These earlier results do not, on their own, establish a vacuum-constraint family agreeing with Ori’s data outside the prescribed annulus, its matching to the unchanged dust envelope, or the Weyl–Pontryagin lower bound for the constructed tangent. We do not use these results in our explicit construction, nor do we claim a solution to general KID obstructions. The construction itself neither predicts an evolved chronology horizon nor rules one out.
The closest comparison is not limited to formal Taub–NUT deformations. The unpolarized U ( 1 ) reductions of Choquet-Bruhat [33, Sec. 1] and the exact target-space S L ( 2 , R ) construction of Berger and Moncrief [36, Sec. III] already permit nontrivial twist and nonlinear circle-symmetric vacuum geometry. The compact-hyperbolic constraint analysis of Gicquaud and Huneau [34, Secs. 2.2–2.3] likewise precedes our reduction. Those works address different geometric problems from the fixed annular matching on Ori’s vacuum–dust initial slice. Bernard and Julia developed a dressing group for vacuum solutions with two commuting, surface-orthogonal Killing fields [42]. Penna later used the Geroch group to derive the cylindrical Einstein–Rosen wave metric algebraically [43]. In particular, a single spatially constant target-space S L ( 2 , R ) isometry cannot transform Ori’s unperturbed norm-and-twist map into that of this family. At r = r 0 the original map is spatially constant, whereas the squared norm of ∂ τ in the initial data, λ ε : = γ ε ( ∂ τ , ∂ τ ) = f 0 ( 1 − ε 2 p 2 ) , and the compactly supported twist potential in Proposition 8.1 vary with ρ . This excludes only that fixed target-space transformation. It does not exclude a reduction of the annular family to Geroch’s two-Killing-field construction [37], the dressing-group construction just cited, or another combination of earlier results.

5. Linearized Axial Einstein Equations in Ori’s Vacuum Core

For the circle-independent linearization of the exact family, we use the established vector-type reduction of the vacuum Einstein equations due to Kodama and Ishibashi [6]. Version 3 of their arXiv preprint presents the invariant reconstruction in Eq. (5.12), the master variable in Eq. (5.13), and the effective potential for vector-type perturbations in Eq. (5.15) [6, Sec. 5.2, Eqs. (5.10)–(5.15)]. The specialization to the negative-curvature transverse factor uses n = 2 , K = − 1 , M = − μ and Λ = 0 in that reference. Here n and K are the dimension and transverse sectional-curvature parameters of Kodama–Ishibashi, not our future unit normal and second fundamental form. In the Kodama–Ishibashi convention, the static metric coefficient is f KI = K − 2 M / r = − f , where f = 1 − 2 μ / r in (2). Thus the orbit metric in Ori’s chronological region is the same Lorentzian metric, with r as the time coordinate. This sign correspondence must be made before specializing the master equation. Set A = − Δ H 2 . A generalized scalar harmonic Y λ has A Y λ = λ Y λ , where λ ≥ 1 / 4 , and the coexact one-form V = * H 2 d Y λ satisfies k V 2 = λ + 1 . Here D ^ is the Levi-Civita derivative of h H 2 , with indices raised by that metric. The hyperbolic Weitzenböck commutator and Ric ( h H 2 ) = − h H 2 give
D ^ B D ^ B ( * h H 2 d Y λ ) A = − ( λ + 1 ) ( * h H 2 d Y λ ) A .
In particular, the zero circle-frequency specialization of Kodama and Ishibashi’s Eq. (5.11) is precisely the radial equation (34) used below. This correspondence includes the negative-curvature shift k V 2 = λ + 1 . All spectral equalities below are in the L 2 functional calculus. Individual generalized harmonics are not assumed square integrable.
The gauge-invariant axial master variable Φ = Ω / r satisfies the specialized Kodama–Ishibashi equation
∂ r ( f ∂ r Φ ) − f − 1 ∂ τ 2 Φ − r − 2 Δ H 2 Φ + 6 μ r 3 Φ = 0 .
In the horizon-regular background chart v = − ( τ + r * ) , where r * ′ = f − 1 , the equation becomes
∂ r ( f ∂ r Φ ) − 2 ∂ v ∂ r Φ − r − 2 Δ H 2 Φ + 6 μ r 3 Φ = 0 .
For the vector metric perturbation h a i = r F a V i , the invariant orbit one-form is obtained from
r F a = ϵ a b ∂ b Ω , Ω = r Φ .
With the sign convention ϵ v r = 1 in the EF orbit chart g ( 2 ) = f d v 2 + 2 d v d r , this gives
r F v = f ∂ r Ω − ∂ v Ω , r F r = ∂ r Ω .
The reconstruction gives a linearized vacuum Einstein field, not merely an unconstrained metric perturbation. The orbital amplitudes and their coexact angular reconstruction are invariant under the admissible smooth vector-type infinitesimal gauges. The existence of nonzero L 2 harmonic one-forms on H 2 is well known [44, p. 396, Theorem (iii)]. For example, the one-form Re ( d ζ ) in a Poincaré disk coordinate ζ is L 2 harmonic. Its squared L 2 norm is π by conformal invariance in two dimensions. It lies outside the scalar-derived positive-spectral sector used here and is not considered.

6. Cauchy Transfer for Compactly Supported Axial Data

For the axial tangent of the exact family, the physical initial data are compactly supported while the auxiliary master potential need not be. We compute the physical-to-master Cauchy map and the regularity obstruction at the background horizon.

6.1. The Spectral Cauchy Map for Compactly Supported Axial Data

We use the vector reconstruction stated in Section 5. In particular, r h = 2 μ , r 0 = 3 μ , f = 1 − r h / r , and γ 0 = f 0 d τ 2 + σ with f 0 = 1 / 3 . Fix 0 ≠ ϕ ∈ C c ∞ ( H 2 ) . In the combined theorem, we take ϕ = ϕ 0 from (22). Consider the transverse-traceless axial seed
γ ˙ = 0 , K ˙ τ A = ( * σ d ϕ ) A , K ˙ τ τ = K ˙ A B = 0 .
The tensor is trace free. Its divergence vanishes since D σ A ( * σ d ϕ ) A = 0 and it is independent of τ . The initial slice is umbilical, so ( γ ˙ , K ˙ ) satisfies the linearized vacuum constraints. The map from this physical datum to the Kodama–Ishibashi master scalar need not be local. We explicitly compute it rather than presuming a compact master field.
In the Regge–Wheeler vector gauge and the original diagonal ( r , τ ) chart, a τ -independent mode has
h τ A = − f ∂ r Ω λ V A , h r A = 0 , V = * σ d Y λ , Ω λ = r u λ .
These follow from the vector reconstruction (29) and the coordinate change from τ to v. All the formulas in this subsection are restricted to the scalar-derived, nonexceptional vector sector with λ ≥ 1 / 4 . The static master equation for each − Δ H 2 spectral fiber is
( f u λ ′ ) ′ + λ r 2 + 6 μ r 3 u λ = 0 , λ ∈ [ 1 / 4 , ∞ ) .
Upon substituting u λ = Ω λ / r , this becomes the exact identity
f Ω λ ′ ′ + f ′ − 2 f r Ω λ ′ + λ + 2 r 2 Ω λ = 0 .
On the initial slice the Regge–Wheeler gauge has vanishing shift perturbation and lapse perturbation. The chosen future normal is n = − f ∂ r , so K = − 1 2 L n γ = ( f / 2 ) ∂ r γ when the shift vanishes. Consequently, in the extrinsic-curvature convention of (3),
K ˙ τ A ( r 0 ) = f 0 2 ∂ r h τ A r 0 = f 0 ( λ + 2 ) 2 r 0 2 Ω λ ( r 0 ) V A ,
where γ ˙ τ A = 0 has imposed Ω λ ′ ( r 0 ) = 0 . Here and in the next display, angular spectral coefficients are understood in the L 2 functional calculus, not as normalizable individual plane waves. The exact Cauchy map is therefore
Ω ( r 0 , · ) = 2 r 0 2 f 0 ( − Δ H 2 + 2 ) − 1 ϕ , ∂ r Ω ( r 0 , · ) = 0 .
The inverse in (36) is a bounded elliptic inverse on L 2 ( H 2 ) , but it is nonlocal and does not generally preserve compact support. For example, if ϕ ≥ 0 and ϕ ¬ ≡ 0 , positivity of the resolvent kernel implies that Ω ( r 0 , y ) > 0 at every y ∈ H 2 . Thus the KI master datum associated with a compactly supported axial constraint perturbation need not be compactly supported. This nonlocality does not, by itself, imply that the physical perturbation reconstructed in a regular gauge has noncompact angular support.

6.2. A Logarithmic Horizon Coefficient for Circle-Independent Axial Data

The static fiber equation (33) has a repeated horizon indicial root. Denote by F λ ( r ) the solution analytic at r = r h , normalized by
F λ ( r h ) = 1 , F λ ′ ( r h ) = − λ + 3 r h .
The radial weighted Wronskian of F λ and any other solution u λ is constant. The regular Frobenius coefficients depend analytically on the spectral parameter, including at the auxiliary value λ = − 6 used below. The regular series is obtained by a recursion whose jth denominator for j ≥ 1 is a nonzero multiple of j 2 , whereas its numerator is polynomial in λ . The regular series converges uniformly for λ in compact sets and | r − r h | ≤ z I r h for any fixed 0 < z I < 1 . For the particular data (36),
W λ = f ( F λ u λ ′ − F λ ′ u λ ) = − f 0 F λ ′ ( r 0 ) + F λ ( r 0 ) r 0 u λ ( r 0 ) , u λ ( r 0 ) = 2 r 0 f 0 ( λ + 2 ) ϕ ^ ( λ , ω ) .
Here ϕ ^ ( λ , ω ) denotes the Helgason–Fourier transform of ϕ on H 2 . The multiplicity parameter ω ∈ ∂ ∞ H 2 ≃ S 1 labels the ideal boundary, not a circle frequency. We write d ν ( λ , ω ) for the joint Plancherel measure. The transform and measure are normalized together so that ∥ ϕ ∥ L 2 ( H 2 ) 2 = ∫ [ 1 / 4 , ∞ ) × S 1 | ϕ ^ ( λ , ω ) | 2 d ν ( λ , ω ) . Thus ∫ I … d ν means integration over λ ∈ I and all boundary parameters ω . Writing s = r − r h , the coefficient of log ( s / r h ) is
C λ ( ω ) = r h W λ = − r h f 0 F λ ′ ( r 0 ) + F λ ( r 0 ) r 0 2 r 0 ϕ ^ ( λ , ω ) f 0 ( λ + 2 ) .
The passage from the fiberwise logarithmic coefficient to an L 2 spectral lower bound requires estimates uniform in the spectral parameter. We establish these estimates explicitly rather than taking a pointwise Frobenius limit through a Plancherel integral.
Lemma 1. 
Fix a compact interval I ⋐ ( 1 / 4 , ∞ ) and let u λ be specified by the Cauchy pair in (36). Set z = ( r − r h ) / r h . There are bandwise L 2 amplitudes A λ ( ω ) and C λ ( ω ) , linear in ϕ ^ ( λ , ω ) , with C λ given by (39), such that for some z I > 0 and κ I < ∞ ,
u λ ( r , ω ) = C λ ( ω ) log z + A λ ( ω ) + R λ ( z , ω ) ,
| R λ | + | z ∂ z R λ | ≤ κ I z ( 1 + | log z | ) | ϕ ^ ( λ , ω ) | ( 0 < z < z I ) ,
for almost every ( λ , ω ) in the spectral band over I. In particular, as z ↓ 0 , u λ / log z and z ∂ z u λ both converge strongly to C λ in the band-limited spectral L 2 space.
Proof. 
With x = r / r h = 1 + z , the static equation (33) is exactly
x 2 ( x − 1 ) u x x + x u x + ( λ x + 3 ) u = 0 .
For the horizon-regular solution write F λ = ∑ j ≥ 0 a j ( λ ) z j with a 0 = 1 and a − 1 = 0 . Comparing coefficients gives the explicit recurrence
( j + 1 ) 2 a j + 1 + ( 2 j 2 − j + λ + 3 ) a j + ( j − 1 ) ( j − 2 ) + λ a j − 1 = 0 , j ≥ 0 .
In particular, a 1 = − ( λ + 3 ) , agreeing with (37) after converting back from z to r. The nonzero factors ( j + 1 ) 2 show that every a j is a polynomial in λ . Regular-singular series theory gives convergence for | z | < 1 . Here is an explicit uniform majorant, sufficient for the spectral limit. Put L = sup I and M = 2 L + 10 . For j ≥ 1 , the absolute values of the two recurrence coefficients divided by ( j + 1 ) 2 are at most L + 5 and L + 1 , respectively. Since | a 0 | = 1 , | a 1 | ≤ L + 3 < M , and M 2 > ( L + 5 ) M + ( L + 1 ) , induction gives
| a j ( λ ) | ≤ M j ( j ≥ 0 , λ ∈ I ) .
In particular, for 0 ≤ z ≤ z I : = 1 / ( 4 M ) ,
| F λ ( z ) − 1 | ≤ 1 3 , | ∂ z F λ ( z ) | ≤ 16 M 9 , | F λ ( z ) | ≥ 2 3 .
These estimates are uniform on the entire spectral band, including its endpoints, and give a common zero-free horizon neighborhood without a pointwise-in- λ choice of radius.
Let W λ = f ( F λ u λ ′ − F λ ′ u λ ) . Reduction of order and f = z / ( 1 + z ) imply
∂ z u λ F λ = r h W λ ( 1 + z ) z F λ 2 = C λ z + C λ q λ ( z ) ,
where q λ is analytic near z = 0 and bounded there uniformly on I. More explicitly, q λ ( z ) = z − 1 ( ( 1 + z ) / F λ ( z ) 2 − 1 ) . Equation (45) bounds it by 7 M + 9 / 4 for 0 ≤ z ≤ z I (with its continuous value at z = 0 ). The initial Cauchy pair at z = 1 / 2 is
( u λ , ∂ z u λ ) | z = 1 / 2 = u λ ( r 0 ) , − 2 3 u λ ( r 0 ) , | u λ ( r 0 ) | ≤ c I | ϕ ^ ( λ , ω ) | .
The ordinary differential equation is uniformly regular for z ∈ [ z I , 1 / 2 ] and λ ∈ I . Its fundamental-matrix propagator is uniformly bounded on this compact set. Hence u λ ( z I ) and ∂ z u λ ( z I ) are bounded by c I | ϕ ^ | . The Wronskian identity then bounds | C λ | by the same quantity. Integrating (46) from zero gives
u λ ( z ) = F λ ( z ) C λ log z + A λ + C λ ∫ 0 z q λ ( z ˜ ) d z ˜ ,
where the value at z I uniquely determines A λ . The displayed bounds control | A λ | and the integral uniformly by c I | ϕ ^ | . The bounds on F λ and ∂ z F λ now prove both inequalities in (41). Plancherel on the compact band then proves the asserted strong limits. □
The horizon coefficient need not be nonzero at every spectral parameter. It is nonzero on a set of positive hyperbolic spectral measure for every 0 ≠ ϕ ∈ C c ∞ ( H 2 ) .
Proposition 6.1. 
For every 0 ≠ ϕ ∈ C c ∞ ( H 2 ) , take the physical axial constraint datum (31) on the r = r 0 slice of the auxiliary untruncated Ori vacuum background with periodic τ and angular factor H 2 . The corresponding circle-invariant master Cauchy pair is given by (36). For the resulting master potential Ω, define the reconstructed coexact angular one-form
F r ( r , · ) : = * σ d H 2 ∂ r Ω ( r , · ) .
There exist a compact interval I ⊂ ( 1 / 4 , ∞ ) and c I , δ I > 0 such that the gauge-invariant vector reconstruction in the smooth EF chart satisfies
Π I F r ( r , · ) L 2 ( T * H 2 ) ≥ c I r − r h ( 0 < r − r h < δ I ) .
Here Π I is the spectral-band projector for − Δ H 2 on scalars and the induced projector on scalar-derived coexact one-forms through * σ d H 2 . The assertion is a fixed-backgroundlinearized vacuumresult in the scalar-derived axial sector. It is not a claim about the globally matched spacetime or a theorem on the linearized Weyl curvature.
Proof. 
The regular Frobenius solution F λ , evaluated at a fixed r > r h , is real-analytic in λ . Therefore
b ( λ ) = F λ ′ ( r 0 ) + F λ ( r 0 ) r 0
is real-analytic. It is not identically zero. At the nonphysical value λ = − 6 one has F − 6 ( r ) = ( r / r h ) 3 , whence b ( − 6 ) = 4 r 0 2 / r h 3 ≠ 0 . This value is used only to rule out the zero analytic function and is not a normalizable hyperbolic mode. The zero set of b in the physical spectral interval is discrete. Helgason’s Plancherel theorem for the hyperbolic Fourier transform [45, Theorem 2.2 and Eq. (3.5)] gives an absolutely continuous spectral resolution with positive density. The disk metric in that source has curvature − 4 . After rescaling to curvature − 1 , the spectral variable in our notation is λ = ( ζ 2 + 1 ) / 4 , where ζ > 0 is the Fourier parameter. Since ϕ ≠ 0 , its transform cannot be supported on the discrete zero set of b . Consequently, the coefficient in (39) is nonzero on a set of positive spectral measure.
Choose a compact interval I of positive spectral mass avoiding the discrete zero set. Lemma 1 establishes the uniform remainder bounds on this band, so
( r − r h ) u λ ′ ( r , ω ) ⟶ C λ ( ω ) , ( r − r h ) ∂ r Ω λ ( r , ω ) ⟶ r h C λ ( ω )
strongly in the corresponding finite-band spectral L 2 space. The EF reconstruction (30) gives r F r = ∂ r Ω for the scalar orbit coefficient when the field is independent of v. The coexact angular reconstruction and spectral Plancherel identity give
∥ Π I F r ( r , · ) ∥ L 2 ( T * H 2 ) 2 = ∫ I λ | ∂ r Ω λ ( r , ω ) | 2 d ν ( λ , ω ) .
The bandwise strong limit then gives
lim r ↓ r h ( r − r h ) Π I F r ( r , · ) L 2 ( T * H 2 ) = r h ∫ I λ | C λ ( ω ) | 2 d ν ( λ , ω ) 1 / 2 > 0 .
This proves (47). □
Remark 1. 
The excluded spectral zeros are not merely a formal possibility. With z = ( r − r h ) / r h , the recurrence (43) and r 0 = 3 r h / 2 give
r h b ( λ ) = ∑ j ≥ 0 2 j + 2 3 2 − j a j ( λ ) .
For 6.74 ≤ λ ≤ 6.75 , set α j = ( − 1 ) j a j and β j = α j − α j − 1 . The recurrence implies
β j + 1 = ( j − 1 ) ( j − 2 ) + λ ( j + 1 ) 2 β j ( j ≥ 1 ) .
Here β 1 = λ + 2 > 0 , while 0 < β j + 1 ≤ β j for j ≥ 2 . The first two coefficients then give 0 < α j ≤ 16 j for j ≥ 2 . Truncating the displayed series after j = 150 has absolute error at most
48 2 − 151 ( 2 · 151 2 + 4 · 151 + 6 ) < 8 · 10 − 40 .
Set
λ − = 6.74673670987912949775 , λ + = 6.74673670987912949776 .
Exact rational evaluation of the partial sum at these two endpoints gives values in ( − 1.780 , − 1.779 ) × 10 − 21 and ( 2.355 , 2.356 ) × 10 − 21 , respectively. Both signs are certified by the error bound. Thus b has an actual zero λ * ∈ ( λ − , λ + ) . This observation is not needed for the positive-measure spectral-band argument.
The conclusion in the zero circle-frequency sector follows from the physical Cauchy data, not from independently prescribed master-field data. The F a in this argument are invariant under smooth vector-type infinitesimal coordinate changes. The ( v , r ) chart is regular on the background. The growth of F r alone is not a curvature statement. The next section computes a gauge-invariant Weyl response and shows how the nonlocal auxiliary master potential is compatible with finite propagation inside Ori’s vacuum core.

7. Growth of the Weyl Pontryagin First Variation in a Protected Region of Ori’s Vacuum Core

The auxiliary master potential may have noncompact angular support even when the physical perturbation is compactly supported. For the sourced linearized Einstein equations, the existence of spatially compactly supported metric perturbations is subject to Killing-charge conditions [46, Theorem 1.1]. That theorem is not used here. We work with source-free axial data and establish physical localization through a local hyperbolic Cauchy problem. For the circle-independent datum in (31), we then prove logarithmic growth of a local curvature observable within a protected portion of Ori’s finite vacuum core. The resulting statement remains a linearized vacuum result and does not control an exact perturbed Einstein–dust development.
We use the established relation between axial gravitational master fields and the first variation of the Weyl Pontryagin density. The analogous Schwarzschild identity appears in the Chern–Simons perturbation literature [19] and explicitly in [20, Appendix F.1, Eq. (F.2)]. The odd-parity curvature interpretation has also been developed in [23, Secs. IV and VI]. Dotti and Fernández Tío developed related invariant methods for Schwarzschild and Reissner–Nordström perturbations [21,22]. In the charged case they also analyzed the divergence of derivatives of the curvature invariants near the Cauchy horizon [22, Sec. IV]. We do not claim originality for the curvature diagnostic or the general phenomenon of linear horizon growth. We verify the negative-curvature Ori-core specialization and connect it to the explicit constraint datum within the causally protected vacuum region.

7.1. The Direct Curvature Identity and Its Initial-Data Check

Write A = − Δ H 2 and L = A + 2 . Let Y be a smooth angular scalar. In the diagonal chart of (32), set
V = * H 2 d Y , h τ A = H ( r ) V A , h r A = 0 , H = − f ∂ r Ω .
All other first-order metric components vanish in this vector gauge. We use the Weyl Pontryagin pseudoscalar P = ★ C a b c d C a b c d , with spacetime orientation d t ∧ d τ ∧ d ρ ∧ d ϑ > 0 and future Gaussian normal n = ∂ t . It vanishes identically on the unperturbed core. For an explicit local contraction, take upper-half-plane coordinates ( x , y ) with y > 0 and h H 2 = y − 2 ( d x 2 + d y 2 ) . We choose the upper-half-plane coordinates so that the oriented area form is − y − 2 d x ∧ d y , consistently with * h H 2 d ρ = sinh ρ d ϑ . Then the vector form is V x = ∂ y Y and V y = − ∂ x Y . Direct first variation of the four-dimensional Riemann tensor gives
P ˙ = 24 μ r 6 ( r H ′ − 2 H ) Δ H 2 Y .
The direct tensor calculation can be performed without fixing an angular eigenmode. Let g 0 : = g O , and denote its Levi-Civita connection by ∇ 0 . With h the off-diagonal perturbation, the first variations are obtained from
Γ ˙ a b c = 1 2 ( g 0 ) a d ( ∇ b 0 h d c + ∇ c 0 h d b − ∇ d 0 h b c ) , R ˙ a b c d = ∇ c 0 Γ ˙ a d b − ∇ d 0 Γ ˙ a c b .
We raise the Riemann indices with the background inverse and its first variation. Appendix A lists the coordinate formulas for a finite-jet check. The surviving angular factor is y 2 ( ∂ x 2 + ∂ y 2 ) Y = Δ H 2 Y . The displayed sign is fixed by the stated orientation and the duality convention. A reversal of orientation reverses P . As tr g 0 h = 0 , there is no first-order determinant correction to the volume tensor. The equality of the Riemann and Weyl Pontryagin contractions is algebraic in four dimensions and does not require the vacuum field equation [47, Appendix B, Eq. (B1)]. Thus (52) follows without imposing an equation of motion on H.
The static axial master equation (34) can also be written as
f Ω ′ ′ + f ′ − 2 f r Ω ′ + L r 2 Ω = 0 .
Consequently,
r H ′ − 2 H = − r f Ω ′ ′ − ( r f ′ − 2 f ) Ω ′ = 1 r L Ω .
Superposition in the angular spectral representation therefore gives the local curvature-transfer identity
P ˙ ( r , · ) = − 24 μ r 7 A L Ω ( r , · ) = − 24 μ r 7 A Ψ ( r , · ) , Ψ : = L Ω .
Equivalently, in a λ -mode its magnitude is 24 μ λ ( λ + 2 ) | Φ λ | / r 6 . The factor λ ( λ + 2 ) is the hyperbolic-curvature counterpart of the spherical ℓ ( ℓ + 1 ) ( ℓ − 1 ) ( ℓ + 2 ) factor in Dotti’s odd Weyl-curvature pseudoscalar [21, Eq. (22)]. The coefficient in (56) is independently verified here for the negative-curvature Ori core. Equation (56) computes the curvature observable directly from Ψ using an angular differential operator, without the elliptic inversion needed to recover Ω . Equivalently, for the physical metric amplitude Ξ introduced in (59),
P ˙ = 24 μ r 6 ( r ∂ r − 2 ) Δ H 2 Ξ .
An independent initial-slice calculation checks the normalization. Since (36) gives Ψ ( r 0 ) = 2 r 0 2 ϕ / f 0 , one obtains
P ˙ ( r 0 , · ) = − 48 μ r 0 5 f 0 A ϕ .
Using r 0 = 3 μ and Δ σ = r 0 − 2 Δ H 2 , this agrees including the sign with the initial-slice Gauss–Codazzi calculation of SubSection 8.3. In particular, it detects every 0 ≠ ϕ ∈ C c ∞ ( H 2 ) , since the only compactly supported harmonic scalar is zero. This agreement checks the coefficient and its orientation without equating a master amplitude with curvature by definition.

7.2. Local Cauchy Evolution of the Physical Axial Amplitude

The auxiliary master datum Ω ( r 0 ) = const · L − 1 ϕ may be noncompact. This does not obstruct a local Cauchy description of the physical metric amplitude. In the circle-independent Regge–Wheeler vector gauge, write
h τ A = ( * σ d Ξ ) A , h r A = 0 , Ξ = − f ∂ r Ω .
The direct linearized Ricci identity (A11) gives the local equation
f Ξ ′ ′ − 1 r 2 Δ H 2 Ξ + 4 μ r 3 Ξ = 0 .
The physical Cauchy data (31) give
Ξ ( r 0 , y ) = 0 , Ξ ′ ( r 0 , y ) = 2 f 0 ϕ ( y ) .
This equation also follows from the zero circle-frequency specialization of the gauge-invariant odd-mode equation of Tanimoto, Moncrief and Yasuno [7, Sec. 5, Eq. (105)]. Their Proposition 7 and Eq. (189) also already reconstruct the same physical axial cross-metric mode in this sector, up to normalization of the vector harmonic. Our direct Ricci calculation independently checks the normalization and sign of the Ori-core equation. Both data are compactly supported, without any angular elliptic inversion. The Kodama–Ishibashi reconstruction already supplies a linearized Einstein solution with this Cauchy pair. Uniqueness for (60) shows that its physical amplitude agrees with the solution of this local problem. Thus we are not replacing the full linearized Einstein system by a single unchecked Ricci component. On shell, the local differential relation
Ψ = L Ω = r ( r Ξ ′ − 2 Ξ )
follows from (55). As a consistency check, applying L to the master equation shows that Ψ also satisfies a local hyperbolic equation,
f Ψ ′ ′ + f ′ − 2 f r Ψ ′ − 1 r 2 Δ H 2 Ψ + 2 r 2 Ψ = 0 ,
with compactly supported initial data
Ψ ( r 0 , y ) = 2 r 0 2 f 0 ϕ ( y ) , Ψ ′ ( r 0 , y ) = 0 .
Both (60) and (63) are local Cauchy equations with compactly supported data. Mode-independent vector perturbation equations on general warped backgrounds are earlier work of Cai and Cao [11, Sec. III.B.2]. The Ori-core equation (60) is derived here directly from the Ricci variation as an independent check of its earlier Bianchi III version, rather than quoted from the formalism of Cai and Cao. Rauch and Rendall used compactly supported test-field data and finite propagation in certain Cauchy-horizon geometries with noncompact transverse sections [5, Sec. 3, concluding paragraphs]. For the present Ori core, both local equations have the principal propagation speed 1 / ( r f ( r ) ) in the unit hyperbolic metric. To check the full finite-speed assertion, introduce the radial conformal time
η = d ( r ) = ∫ r r 0 d s s f ( s ) , 0 ≤ η < θ c , ∂ r = − 1 r f ∂ η .
After multiplying the two local equations by r 2 they become
∂ η 2 Ξ − Δ H 2 Ξ + 1 + f 2 f ∂ η Ξ + 4 μ r Ξ = 0 ,
∂ η 2 Ψ − Δ H 2 Ψ + 7 f − 1 2 f ∂ η Ψ + 2 Ψ = 0 .
Here r = r ( η ) and f = f ( r ( η ) ) . For every η * < θ c , all lower-order coefficients in these equations are bounded and smooth on [ 0 , η * ] , and both principal parts are precisely ∂ η 2 − Δ H 2 . Uniqueness in hyperbolic backward light cones therefore gives unit angular propagation speed on each such closed time interval. Letting η * vary proves the same support inclusion for every r h < r ≤ r 0 . No uniform energy bound at the singular limiting coefficient f = 0 is asserted. If supp ϕ ⊂ B R ( o ) , finite propagation gives
supp Ψ ( r , · ) ∪ supp Ξ ( r , · ) ⊆ B R + d ( r ) ( o ) ¯ , d ( r ) = ∫ r r 0 d s s f ( s ) , 0 ≤ d ( r ) < θ c ,
where θ c = log ( 2 + 3 ) and r h < r ≤ r 0 = 3 μ . The lower-order potential term in (63) does not change the characteristic cones.
The field Ξ directly determines the physical metric perturbation through (59) and its EF counterpart through r F r = − Ξ / f , without angular elliptic inversion. An equivalent reconstruction from Ψ follows from the divergence form of (54). This equation reads
f r 2 Ω ′ ′ = − Ψ r 4 .
The condition Ω ′ ( r 0 ) = 0 therefore gives
H ( r , y ) = − f ( r ) Ω ′ ( r , y ) = − r 2 ∫ r r 0 Ψ ( s , y ) s 4 d s .
The right-hand side is supported in B R + d ( r ) ( o ) ¯ , since d ( s ) ≤ d ( r ) for r ≤ s ≤ r 0 . It follows that both components of the vector-type metric perturbation, including the EF component r F r = Ω ′ = − H / f for each fixed r h < r ≤ r 0 , vanish outside the same angular propagation cone. Even when L − 1 ϕ has a noncompact angular tail, that tail belongs to an auxiliary variable and does not force noncompact support of the reconstructed physical linear perturbation for r h < r ≤ r 0 . This concerns support alone and gives no uniform bound on reconstructed metric components as r ↓ r h . Equation (56) also shows directly that
supp P ˙ ( r , · ) ⊆ B R + d ( r ) ( o ) ¯ .
This conclusion is specific to the circle-independent axial Cauchy pair and its local field equation. It is not a localization theorem for arbitrary polar modes or for the nonlinear CMC correction.

7.3. A Linearized Curvature Growth Estimate in a Causally Protected Region

Proposition 7.1. 
Let ϕ ∈ C c ∞ ( H 2 ) be nonzero, supported in B R ( o ) . Consider the smooth circle-independent axial constraint datum (31) on r 0 = 3 μ and its linearized vacuum Einstein development for r h < r ≤ r 0 , approaching the background horizon r h = 2 μ . Assume Ori’s truncated core has angular radius θ 0 with
R + 2 θ c < θ 0 , θ c = log ( 2 + 3 ) .
Then the reconstructed metric perturbation in the stated vector gauge and the invariant P ˙ are supported within the protected vacuum domain for every r h < r ≤ r 0 . There exist c > 0 and δ > 0 , depending on the seed and background parameters, such that
∥ P ˙ ( r , · ) ∥ L 2 ( B θ 0 − θ c ( o ) , d V H 2 ) ≥ c log r − r h r h ( 0 < r − r h < δ ) .
Thus the first-order, spacetime-gauge-invariant Weyl Pontryagin response does not extend as a bounded continuous scalar to the protected horizon patch. The proposition does not address curvature singularities in exact finite-amplitude Einstein–dust developments.
Proof. 
The physical initial constraints and zero-frequency transfer have been established in (31) and (36). By (64)–(71), finite propagation confines the physical perturbation and P ˙ to B R + θ c ( o ) . More precisely, Ori’s causal-distance bound leaves the known vacuum region B θ 0 − d ( r ) ( o ) on each pre-horizon slice, while our support lies in B R + d ( r ) ( o ) . Since d ( r ) ≤ θ c , the strict inequality R + 2 θ c < θ 0 gives R + d ( r ) < θ 0 − d ( r ) at every such slice. At the horizon, Ori’s protected sector contains B θ 0 − θ c ( o ) by [1]. The bound R + 2 θ c < θ 0 is a sufficient support condition for this specified datum, not a necessary condition for all perturbations of Ori’s model. The size inequality keeps the perturbation strictly within this protected vacuum region. By local Cauchy uniqueness, the linearized field in the stated vector gauge agrees with the solution from the same initial data on the untruncated Ori vacuum background with periodic τ and angular factor H 2 . There is no need to prescribe a disturbance of the dust shell to obtain this interior linear conclusion.
In the hyperbolic spectral resolution the initial-data Wronskian calculation (38)–(39) gives
u λ ( r , ω ) = C λ ( ω ) log r − r h r h + O I | ϕ ^ ( λ , ω ) |
on any fixed compact angular spectral band I away from the zeros of b ( λ ) . This uniform error bound follows from the bound on A λ in the proof of Lemma 1 and the remainder estimate (41). Since the seed is nonzero and b is analytic, there is a compact interval I ⋐ ( 1 / 4 , ∞ ) on which
0 < ∫ I λ 2 ( λ + 2 ) 2 | C λ ( ω ) | 2 d ν ( λ , ω ) < ∞ .
Substituting Ω = r u into (56), Plancherel and the uniform Frobenius expansion give
lim r ↓ r h Π I P ˙ ( r ) L 2 ( H 2 ) log ( ( r − r h ) / r h ) = 24 μ r h 6 ∫ I λ 2 ( λ + 2 ) 2 | C λ ( ω ) | 2 d ν 1 / 2 > 0 .
Although the spectral projection Π I is nonlocal, it is used only to bound a global norm from below. The unprojected curvature scalar is compactly supported in the fixed protected angular ball by (71). Hence
∥ P ˙ ( r ) ∥ L 2 ( B θ 0 − θ c ( o ) ) = ∥ P ˙ ( r ) ∥ L 2 ( H 2 ) ≥ ∥ Π I P ˙ ( r ) ∥ L 2 ( H 2 ) ,
which proves (73). Since the protected angular ball has finite hyperbolic area, the L ∞ norm of P ˙ on that ball is also unbounded as r ↓ r h . Finally the background P [ g O ] vanishes identically, so its first variation is invariant under smooth spacetime infinitesimal diffeomorphisms. It cannot be removed as a coordinate effect. □

7.4. Scope and Limits of the Linearized Curvature Estimate

This proposition concerns a fixed background and the first-order Einstein equations for compactly supported, circle-invariant axial data with transverse-traceless extrinsic-curvature variation. The estimate concerns a curvature scalar that vanishes on the background, not merely a divergent Regge–Wheeler reconstruction coefficient. It does not prove a generic theorem for all polarizations or arbitrary circle dependence. The spectral argument explicitly allows exceptional individual fibers where b ( λ ) = 0 . Such isolated fibers have zero Plancherel measure on the full hyperbolic plane. The support condition (72) restricts the seed’s angular support relative to the chosen Ori core radius.
Even within the protected patch, the amplitude measured here is P ˙ , not P [ g ε ] for a family of exact evolved metrics. A first-order term of size | ε | | log ( ( r − r h ) / r h ) | leaves the perturbative regime arbitrarily close to the horizon at each fixed nonzero ε . The calculation neither determines the nonlinear endpoint nor proves that the original chronology horizon cannot be replaced by a different geometry. Our exact annular constraint family is matched to the unmodified dust initial data, but no corresponding claim about its full coupled nonlinear evolution follows. Nothing here proves finite-amplitude curvature blow-up, inextendibility of the coupled solution, or a universal chronology-protection principle.
The background geometry is due to Ori [1], and the axial master equation to Kodama and Ishibashi [6]. Pontryagin curvature diagnostics also appear in earlier axial perturbation studies [19,20]. The statement here is restricted to the specified compactly supported transverse-traceless data, the resulting gauge-invariant first variation, and the causally protected region of the Ori vacuum core. We do not claim a general Cauchy-horizon instability theorem.

8. Finite-Amplitude Twist Data and Local Vacuum Developments

Proposition 4.1 constructs exact finite-amplitude vacuum constraint data, not merely a formal perturbation series. Throughout this section, we use that same annular family. Wherever the circle Killing field is spacelike, the local vacuum equations admit the 2 + 1 Einstein–wave-map reduction developed by Geroch and Moncrief [48,49]. Moncrief gives the local formulation explicitly in his later exposition [50, p. 168]. We write Σ 0 for the common initial hypersurface on which these data families are compared. For fields defined through a local vacuum development, statements on Σ 0 refer only to its vacuum-core portion.

8.1. Persistence of the Inherited Symmetry in the Vacuum Development

Let ξ = ∂ τ on the original core initial surface. Since Y ε and ϕ ε depend only on ρ , the tensors γ ε and K ε in (19) are invariant under the τ -circle action and the rotations generated by ∂ ϑ . They are smooth across ρ = 0 , where the latter vector field vanishes. By local geometric uniqueness for vacuum Einstein evolution, these initial symmetries induce Killing fields on the corresponding vacuum domain of dependence [51, Sec. 2.2]. The τ -circle action is also the globally axial symmetry of Ori’s original initial data. His global azimuthal coordinate is a rescaling of t Ori = τ , not the core hyperbolic angle ϑ [1, Sec. III.B, after Eq. (6)]. Since our change is τ -independent and supported strictly inside the core, the modified global initial data retain this axial symmetry. The extra core rotation generated by ∂ ϑ need not extend across the dust envelope. The inherited periodic coordinate τ gives a free circle action on the core initial slice. In the local vacuum development, write λ ε : = g ε ( ξ , ξ ) . Its restriction to the initial slice agrees with the norm γ ε ( ∂ τ , ∂ τ ) used above. The Killing field ξ is spacelike there because
λ ε | Σ 0 = f 0 Y ε = f 0 ( 1 − ε 2 p 2 ) > 0 .
It remains spacelike in a sufficiently small vacuum neighborhood of that slice. We make no claim that it stays spacelike up to a perturbed chronology horizon. The vacuum Einstein–wave-map reduction used below is asserted only within the inherited vacuum region, not across the dust envelope. The latter is globally axially symmetric but need not inherit the extra ∂ ϑ symmetry of the core.
Where λ ε > 0 in the vacuum development, use the Einstein-frame metric h on the three-dimensional orbit space,
h a b = λ ε g a b − λ ε − 1 ξ a ξ b .
The induced two-metric q on the quotient of the vacuum-core portion of Σ 0 by the τ -circle action satisfies the exact identity
q ε = λ ε ( Y ε − 1 σ ) = f 0 σ = q 0 .
Thus the finite-amplitude deformation has no effect on the Einstein-frame quotient spatial metric at the initial instant. This is an identity for this particular ansatz. It says nothing about the subsequent quotient metric or its second fundamental form.

8.2. A Compactly Supported Initial Twist Potential

In the vacuum development define the Killing twist one-form by
ϖ a = ε a b c d ξ b ∇ c ξ d .
It is closed wherever Ric ( g ε ) = 0 . Consequently, it is locally d χ ε for a twist potential χ ε . We fix the orientation by d t ∧ d τ ∧ d ρ ∧ d ϑ > 0 and take the future unit normal to be n = ∂ t in Gaussian normal coordinates along Σ 0 . Our convention for extrinsic curvature is K i j = − 1 2 ( L n γ ) i j . A reversal of spacetime orientation reverses all the signs in the next twist identity but not its consequences.
Proposition 8.1. 
For the exact family (19) the restriction of the Killing twist to Σ 0 is
ϖ | Σ 0 = − 2 ε f 0 d ϕ ε , χ ε | Σ 0 = − 2 ε f 0 ϕ ε .
The potential is normalized to vanish in the two vacuum collars. The initial twist potential is smooth, compactly supported, and nonconstant for every ε ∈ D p ∖ { 0 } . The exact nonlinear vacuum development is therefore not polarized in any open neighborhood intersecting the nonzero-twist part of the annulus.
Proof. 
At t = 0 the metric is diagonal in ( τ , ρ , ϑ ) with spatial coefficients
A = f 0 Y ε , B = r 0 2 Y ε , C = r 0 2 sinh 2 ρ Y ε .
Writing a ε = ϕ ε ′ , the only mixed initial component of the second fundamental form is
K τ ϑ = ε Y ε − 1 / 2 a ε sinh ρ .
Gaussian evolution gives ∂ t g τ ϑ | t = 0 = − 2 K τ ϑ . The terms in ξ ♭ ∧ d ξ ♭ involving d A cancel because ξ ♭ = A d τ at t = 0 . The surviving term is
ξ ♭ ∧ d ξ ♭ t = 0 = 2 ε f 0 Y ε a ε sinh ρ d t ∧ d τ ∧ d ϑ .
Taking the four-dimensional Hodge dual using − det g = f 0 r 0 2 sinh ρ / Y ε gives precisely ϖ | Σ 0 = − 2 ε f 0 a ε d ρ in our orientation. Equation (18) then supplies the compactly supported potential. Where a ε ≠ 0 the three-form (85) is nonzero. Frobenius’ criterion therefore excludes hypersurface orthogonality of ξ in that region. □
The exact norm and twist data also allow a direct computation of the spatial contribution to the reduced wave-map energy. The target of the vacuum U ( 1 ) reduction is hyperbolic [50, pp. 168–171]. In the norm-and-twist target coordinates ( λ K , χ ) we choose the metric
G H = d λ K 2 + d χ 2 2 λ K 2 .
The coordinate λ K pulls back to λ ε along the reduced solution. We call the spatial gradient contribution
E sp = | d λ ε | q 0 2 + | d χ ε | q 0 2 2 λ ε 2
the gradient energy density associated with the quotient initial two-metric. It is not the total wave-map energy and need not be a conserved dynamical charge.
Proposition 8.2. 
For the annular data defined in (19),
E sp ( ρ ) = 2 ε 2 p ( ρ ) 2 f 0 r 0 2 [ 1 − ε 2 p ( ρ ) 2 ] ,
and its spatial integral is
∫ H 2 E sp d V q 0 = 4 π ε 2 ∫ a b p ( ρ ) 2 sinh ρ 1 − ε 2 p ( ρ ) 2 d ρ > 0
for every ε ∈ D p ∖ { 0 } . The integral over H 2 uses the smooth zero extension of this annular energy density, not an extension of the vacuum development. It therefore equals the integral over the physical vacuum-core quotient B θ 0 ( o ) . As ε → 0 , the integral is O ( ε 2 ) , and its leading quadratic coefficient is strictly positive.
Proof. 
From (78) and (82),
λ ε ′ = − 2 f 0 ε 2 p p ′ , χ ε ′ = − 2 ε f 0 ϕ ε ′ , q 0 = f 0 r 0 2 ( d ρ 2 + sinh 2 ρ d ϑ 2 ) .
Use ( ϕ ε ′ ) 2 = f 0 p 2 [ 1 − ε 2 ( p 2 + ( p ′ ) 2 ) ] . The numerator in (87) is
4 f 0 ε 2 p 2 r 0 2 ε 2 ( p ′ ) 2 + 1 − ε 2 ( p 2 + ( p ′ ) 2 ) = 4 f 0 ε 2 p 2 r 0 2 ( 1 − ε 2 p 2 ) .
Division by 2 λ ε 2 gives (88). Integrating with d V q 0 = f 0 r 0 2 sinh ρ d ρ d ϑ proves (89). □

8.3. A Finite-Amplitude Curvature Diagnostic at the Initial Slice

The initial twist rules out polarization wherever it is nonzero. We also derive an initial-slice expansion of the four-dimensional Weyl Pontryagin scalar for the exact finite-amplitude data. For the electric and magnetic Weyl formulas and their Pontryagin contraction, see [47, Appendix B, Eqs. (B6)– (B8)]. On a vacuum initial slice, let ∇ γ denote the Levi-Civita connection of the induced three-metric γ , and let ϵ i k l be its oriented volume tensor. Our curvature conventions give
E i j = Ric ( γ ) i j + ( tr K ) K i j − K i k K k j , B i j = ϵ i k l ∇ k γ K l j .
Our B has the opposite sign to the magnetic Weyl convention used in [47, Eq. (B8)]. Combining this choice with their Eq. (B6) fixes P = + 16 E i j B j i in our orientation. The background Pontryagin scalar vanishes, so its first variation is gauge invariant.
Proposition 8.3. 
Let ( g ε ) be any local smooth vacuum development of the exact family in (19), with a consistent orientation and identification of initial surfaces. For every fixed nonnegative integer m, the exact initial Weyl Pontryagin scalar satisfies
P [ g ε ] | Σ 0 = 16 ε f 0 r 0 2 Δ σ ϕ 0 + O C m ( ε 3 ) .
All O C m bounds in this subsection are understood on each fixed compact subset of the initial vacuum core, with constants depending on m and the compact set. For each sufficiently small nonzero ε, the scalar is not identically zero. It is an invariant of a finite-amplitude vacuum metric, not a formal linear curvature perturbation. For the admissible odd profiles with a simple midpoint zero, p ′ ( ρ * ) ≠ 0 , the initial scalar is nonzero at ρ = ρ * for every ε ∈ D p ∖ { 0 } .
Proof. 
The diagonal background electric Weyl eigenvalues are
( E 0 ) τ τ = 2 3 r 0 2 , ( E 0 ) ρ ρ = ( E 0 ) ϑ ϑ = − 1 3 r 0 2 ,
and B 0 = 0 . With y = Y ε and a ε = ϕ ε ′ , computing B i j from the diagonal metric (83) and the mixed tensor (84) gives
B τ τ = ε f 0 r 0 2 a ε ′ + coth ρ a ε − y ′ y a ε , B ρ ρ = ε f 0 r 0 2 y ′ y − coth ρ a ε , B ϑ ϑ = − ε f 0 r 0 2 a ε ′ .
There are no off-diagonal magnetic components for this radial annular ansatz. The displayed trace is exactly zero. Both y and a ε are even smooth functions of ε , with y = 1 + O C m ( ε 2 ) and a ε = ϕ 0 ′ + O C m ( ε 2 ) . The diagonal electric components are even in ε , whereas the only off-diagonal electric terms are odd and have no magnetic counterpart. Therefore E i j B j i is odd in ε and its next term after the linear one is cubic. By the trace-free identity for B, the eigenvalue gap in (94) gives
E i j B j i = ε f 0 r 0 4 ( ϕ 0 ′ ′ + coth ρ ϕ 0 ′ ) + O C m ( ε 3 ) .
This is (93) because Δ σ ϕ 0 = r 0 − 2 ( ϕ 0 ′ ′ + coth ρ ϕ 0 ′ ) . If Δ σ ϕ 0 vanished everywhere, integration by parts on H 2 and compact support would force ϕ 0 = 0 , contrary to (22). Consequently, for every sufficiently small nonzero ε , the finite-amplitude Pontryagin scalar is nonzero at some point of the annulus.
For the final assertion, fix any admissible odd profile with b * : = p ′ ( ρ * ) ≠ 0 . Such profiles exist. For example, take p ( ρ ) = ( ρ − ρ * ) β ( ρ − ρ * ) with β smooth, even, supported strictly inside ( − ( b − a ) / 2 , ( b − a ) / 2 ) , and β ( 0 ) ≠ 0 . At ρ * one has p = 0 , Y ε = 1 , Y ε ′ = 0 , Y ε ′ ′ = − 2 ε 2 b * 2 , and K ε = k 0 γ ε . Direct evaluation of the spatial Ricci tensor and (95) then gives
E τ τ − E ϑ ϑ = 1 + 2 ε 2 b * 2 r 0 2 , B τ τ = − B ϑ ϑ = ε b * r 0 2 1 − ε 2 b * 2 , B ρ ρ = 0 .
Consequently, the exact initial scalar is
P [ g ε ] Σ 0 , ρ * = 16 ε b * r 0 4 1 − ε 2 b * 2 ( 1 + 2 ε 2 b * 2 ) ≠ 0 .
This strengthens the initial-slice nonvanishing statement for this profile. It gives no finite-amplitude estimate at the future chronology horizon. □
There is a second, geometric explanation for the absence of even powers in (93). The reflection J ( τ , ρ , ϑ ) = ( − τ , ρ , ϑ ) leaves γ ε unchanged and pulls K ε back to K − ε . It reverses spatial orientation. Under geometric uniqueness the corresponding local vacuum developments are related by an orientation-reversing isometry, so their Pontryagin scalars have opposite signs under the initial-data correspondence induced by J. The oddness conclusion concerns the Pontryagin pseudoscalar P . It does not assert oddness of orientation-even curvature invariants such as the Kretschmann scalar.

8.4. Local Perturbative Control on Fixed Pre-Horizon Slabs

The general local existence and Cauchy stability results for the vacuum Einstein equations are classical [51, Sec. 2.2]. Damour and Schmidt studied the relation between smooth one-parameter families of exact Einstein solutions and their perturbative expansions [52]. Our use of local Cauchy stability is confined to the domain of dependence of the vacuum core. No well-posedness assertion for the moving vacuum–dust interface follows from these vacuum-only results.
Proposition 8.4. 
Fix a relatively compact smooth pre-horizon vacuum subslab K of the background Ori development. Assume that its causal antecedent lies in the interior of the original vacuum core and is separated from the core–dust interface, the initial-slice boundary and the background chronology horizon by positive margins. After decreasing ε * if necessary, the annular data have vacuum developments on a common neighborhood of K for all | ε | < ε * . In a consistent smooth equivariant identification of these developments,
P [ g ε ] − ε P ˙ C m ( K ) ≤ C K , m | ε | 3 ,
for any specified finite m and sufficiently small | ε | . The derivative P ˙ is the linearized vacuum curvature response of the same annular tangent data. In particular, near any initial point where P ˙ ≠ 0 , the exact scalar P [ g ε ] is nonzero for every sufficiently small nonzero amplitude.
Proof. 
The initial data depend smoothly on ε in every local Sobolev norm. Fix m and choose an integer Sobolev order s > m + 5 . Use a generalized wave gauge relative to the background metric. Set the initial lapse to one and the initial shift to zero, and determine their normal derivatives from the wave-gauge conditions. These choices depend smoothly on the data and commute with the reflection J. On compact vacuum Cauchy neighborhoods covering the causal past of K , the first-order variables U ε = ( g ε , ∂ g ε ) obey a smooth quasilinear symmetric-hyperbolic reduced system. The classical symmetric-hyperbolic formulation of Einstein evolution is established in [53]. Local Cauchy stability, constraint propagation and finite speed give a common regular neighborhood after restricting the amplitude range [51, Secs. 2.2–2.3].
To obtain the stated cubic remainder, expand the smooth gauged initial data through order two in ε . Let U 1 and U 2 satisfy the first and second variational reduced equations along the background solution U 0 , with those Taylor data. These are linear symmetric-hyperbolic equations with smooth coefficients and, at second order, a smooth quadratic source. The second-order approximate field U ε [ 2 ] = U 0 + ε U 1 + ε 2 U 2 therefore has an initial discrepancy and a reduced-equation residual bounded by C s | ε | 3 in H s on the fixed regular neighborhood. Applying the symmetric-hyperbolic energy estimate to the difference between the exact and approximate reduced systems gives
sup t ∈ [ 0 , T ] ∥ U ε ( t ) − U ε [ 2 ] ( t ) ∥ H s ≤ C s | ε | 3 .
Here the smooth initial family supplies as many higher Sobolev norms as required to bound the coefficients and the residual. Sobolev embedding and the local curvature formula consequently give a C m expansion of P [ g ε ] through order two with an O C m ( | ε | 3 ) remainder.
The reflection J from the preceding subsection maps the initial data at amplitude ε isometrically to those at − ε . Equivariance of the chosen gauge and uniqueness give g − ε = J * g ε in the common identification. Since J reverses spacetime orientation while each development is τ -invariant, the corresponding Pontryagin pseudoscalars satisfy
P [ g − ε ] ( t , τ , ρ , ϑ ) = − P [ g ε ] ( t , − τ , ρ , ϑ ) = − P [ g ε ] ( t , τ , ρ , ϑ ) .
Comparing the second-order expansions for the two amplitudes using (100) shows that both the zeroth and quadratic coefficients vanish. The cubic residual estimate (99) then gives (98). The first variation of g ε is the linearized Einstein development of (21) in the chosen gauge. Its induced Weyl Pontryagin variation is the scalar coefficient P ˙ , which is gauge invariant because the background Pontryagin scalar vanishes. □
The estimate above is established only on fixed compact slabs. The argument gives no bound for C K , m that is uniform as the slab approaches r = 2 μ , nor does it ensure that, for a fixed ε ≠ 0 , the exact development reaches a counterpart of the background Cauchy horizon. The odd function tanh ( ε P ˙ ) illustrates why a cubic Taylor remainder on every fixed compact slab need not control an endpoint where the linear coefficient grows without bound. Thus neither the oddness nor (98) proves finite-amplitude singularity of a chronology horizon. The 2 + 1 Einstein–wave-map formulation could be useful for studying the nonlinear endpoint, but our argument does not give estimates uniform as the Killing norm λ ε approaches zero.

8.5. Related Cauchy-Horizon Results and Limitations

We distinguish the present result from earlier analyses of smooth and singular Cauchy horizons, beginning with vacuum models that admit Killing symmetry. Moncrief constructed infinite-dimensional families containing smoothly extendible vacuum Cauchy horizons [54,55]. His analysis also established boundary curvature singularities for generic diagonal Gowdy solutions [54]. For certain extendible backgrounds, he showed that generic linear perturbations blow up near the Cauchy horizons of those backgrounds [55]. The same author studied symmetry-preserving perturbations of generalized Taub–NUT models to all formal orders [41]. In a separate paper, he analyzed a cubic nonlinear wave equation on fixed vacuum Cauchy-horizon backgrounds [56].
Rauch and Rendall proved test-field nonextendibility results on Taub and Moncrief backgrounds [5, Secs. 2–3]. They also explained how their flux argument extends to certain noncompact transverse sections for compactly supported initial data, assuming propagation limited by the light cones and a nonzero flux [5, Sec. 3, concluding paragraphs]. In their Taub example, the spatially homogeneous linear solution in Eq. (7) contains an integral whose integrand has a simple pole at the Cauchy horizon. Consequently, that example admits logarithmically divergent solutions [5, Sec. 2, Eq. (7)]. Their Theorem 2 requires a nonzero flux involving a derivative of the test scalar in the circle direction. For circle-invariant scalar modes this flux vanishes, so that theorem does not by itself imply our tensorial Pontryagin estimate.
Grubišić and Moncrief used all-order perturbative expansions to analyze curvature singularities in T 3 × R Gowdy vacuum spacetimes [57]. Their geometry and boundary differ from those of the matched Ori core.
Moncrief and Isenberg proved that analytic vacuum and electrovacuum spacetimes with compact null hypersurfaces ruled by closed generators satisfying a local product-bundle condition possess nontrivial Killing symmetries [58]. Isenberg and Moncrief subsequently removed the bundle condition while retaining analyticity and the closed-generator assumption [59]. Moncrief also classified analytic vacuum solutions with compact Hopf-fibered S 3 Cauchy horizons, obtaining an infinite-dimensional family [60]. In a related construction, he used Geroch transformations to obtain an infinite-dimensional family of curvature-singular circle-symmetric vacuum spacetimes from generalized Taub–NUT solutions [61]. His later exposition describes the accompanying change in circle-bundle topology [50, p. 173]. For smooth Ricci-flat spacetimes with compact, smooth, totally geodesic Cauchy horizons whose surface gravity can be normalized to a nonzero constant, Petersen and Rácz established the existence of a nontrivial Killing field [62, Theorem 1.2].
Under different global assumptions, Maeda and Ishibashi proved global hyperbolicity for vacuum spacetimes with compact S 1 -fibered spatial sections, provided their null generic and chronological-horizon conditions hold and all occurring curvature singularities are strong [63, Sec. IV]. Maeda, Ishibashi and Narita excluded a restricted class of chronology-violating regions in maximal finitely vicious spacetimes, assuming their specified boundary, strong-causality, energy and genericity conditions and that every incomplete causal geodesic terminates at a non-naked singularity [64, Theorem 15 in the cited preprint].
Yurtsever proved instability results for Killing–Cauchy horizons in plane-symmetric spacetimes [65]. For colliding parallel-polarized gravitational plane waves, Yurtsever proved nonlinear instability of Killing–Cauchy horizons under small generic initial-data perturbations [66]. He later extended the result to arbitrary polarizations under generic plane-symmetric initial-data perturbations [67].
For black-hole Cauchy horizons, Luk and Oh proved linear scalar-wave instability at the Reissner–Nordström Cauchy horizon [68], while Sbierski proved a linearized gravitational instability theorem at the Kerr Cauchy horizon [69].
At the nonlinear level, Gurriaran’s 2026 preprint [70, Theorem 4.5.1] establishes conditional Lipschitz inextendibility near timelike infinity for vacuum developments from smooth data on a hypersurface slightly inside a dynamical black hole settling down to subextremal Kerr. The hypotheses include a precise nonlinear Price-law-type estimate and a nonvanishing leading-mode condition. Under upper and lower bounds on smooth characteristic data representing an event horizon settling down to that of strictly rotating subextremal Kerr, Luk and Sbierski’s 2026 preprint [71] establishes weak null singularity formation in the corresponding dynamical rotating black-hole interiors. These results concern different backgrounds, perturbation classes, initial surfaces, or matter models. None is invoked as a theorem about the matched Ori 2007 vacuum–dust data.
The exact initial observables and fixed-slab control do not establish nonlinear horizon instability. Any such theorem would require control of the U ( 1 ) quotient evolution as λ ε = g ε ( ξ , ξ ) ↓ 0 , a proof of either a perturbed-horizon regularity obstruction or a geometric obstruction to extendibility, and control of the coupled evolution at the dust interface. None of these conclusions follows from the present argument.

9. Proof of the Combined Theorem and Conclusions

Proof of Theorem 1.1. 
Proposition 3.1 gives the exact constraint reduction. Proposition 4.1 constructs the smooth finite-amplitude family and computes its nonzero axial tangent. Proposition 4.2 assembles the global constraint data with the unchanged Ori envelope. For the tangent ϕ = ϕ 0 , the map (36) is the physical axial Cauchy map. The Wronskian proof of Proposition 6.1 selects a finite spectral band on which the logarithmic horizon coefficient has strictly positive spectral L 2 norm. Proposition 7.1 converts that coefficient to a lower bound for the gauge-invariant Weyl Pontryagin variation. Its strict support condition is satisfied because p ∈ C c ∞ ( ( a , b ) ) implies supp ϕ 0 ⋐ B b ( o ) and b + 2 θ c < θ 0 . This is precisely (1). □
The nonlinear and linearized conclusions must be kept separate. For every amplitude in the open positivity set D p , the matched initial data satisfy the full nonlinear Einstein–dust constraints exactly. The family’s tangent at ε = 0 has a linearized vacuum development on the fixed background whose Weyl Pontryagin response grows without bound toward the background horizon. The local Taylor estimate on each fixed pre-horizon compact slab has a slab-dependent constant and cannot justify interchanging ε → 0 with r ↓ r h . In particular, this article does not establish that an exact perturbed Einstein–dust solution forms a Cauchy horizon, that a new Cauchy horizon is necessarily singular, or that chronology violation is prevented.
The linearized curvature estimate applies to every nonzero ϕ ∈ C c ∞ ( H 2 ) with supp ϕ ⊂ B R ( o ) and R + 2 θ c < θ 0 , as in Proposition 7.1. The combined theorem specializes it to the tangent of the explicit finite-amplitude constraint family. The constraint, perturbation, and curvature arguments draw substantially on established results. A general nonlinear propagation theorem for the vacuum–dust interface, uniform endpoint estimates, and a global nonlinear horizon theorem remain outside this work.

Funding

No funding was received for this research.

Data Availability Statement

This theoretical study uses no datasets. The mathematical arguments are contained in the manuscript.

Conflicts of Interest

The author declares no competing financial or non-financial interests.

Appendix A. Coordinate Verification of the Off-Shell Pontryagin Variation

We give a direct calculation underlying (52). It does not assume the master equation. Use coordinates ( r , τ , x , y ) , with y > 0 , and write
g 0 = diag − 1 f , f , r 2 y 2 , r 2 y 2 , f = 1 − 2 μ r ,
with the only nonzero independent components of the first variation
h τ x = H ( r ) ∂ y Y ( x , y ) , h τ y = − H ( r ) ∂ x Y ( x , y ) .
Choose the upper-half-plane coordinate orientation compatible with Section 2. Since the future normal is n = − f ∂ r , the convention d t ∧ d τ ∧ d ρ ∧ d ϑ > 0 corresponds to [ r τ x y ] = 1 for the covariant alternating symbol. In Lorentzian signature ( − , + , + , + ) , the contravariant volume tensor consequently satisfies ε r τ x y = − 1 / | det g | , as in the Lorentzian volume conventions of [47]. The Riemann Pontryagin contraction is therefore
P = − 1 2 | det g | [ a b e f ] R a b c d R e f c d .
To differentiate this identity we use
g ˙ a b = − ( g 0 ) a c h c d ( g 0 ) d b , Γ ˙ a b c = ( g 0 ) a d 2 ( ∇ b 0 h d c + ∇ c 0 h d b − ∇ d 0 h b c ) .
Here R ˙ a b c d includes the variation of the inverse metric in both raised indices. Since tr g 0 h = 0 , we have ∂ ε | det ( g 0 + ε h ) | 0 = 0 . The background Pontryagin contraction also vanishes. Inserting the connection variation and its covariant curl into (A3) gives
P ˙ = 24 μ r 6 ( r H ′ − 2 H ) y 2 ( ∂ x 2 + ∂ y 2 ) Y .
To reproduce the calculation algebraically, one may treat the radial jets H , H ′ , H ′ ′ and the angular derivatives of Y through order three as independent symbols when varying the Riemann tensor. Every coefficient involving H ′ ′ , third angular derivatives and first or mixed second angular derivatives cancels. The remaining expression is (A5), identically for r > 2 μ and y > 0 . The scalar relation y 2 ( ∂ x 2 + ∂ y 2 ) Y = Δ H 2 Y recovers (52). The identity is algebraic and uses neither a spectral decomposition nor an Einstein field equation for H.
The coefficient can also be verified directly from the electric and magnetic Weyl tensors on the r = const slices. Let ϵ ^ A B be the contravariant area tensor of h H 2 in the angular orientation fixed above. For r-slices, set γ ( r ) = f d τ 2 + r 2 h H 2 and K ( r ) = ( f / 2 ) ∂ r γ ( r ) . At r = r 0 these restrict to ( γ 0 , K 0 ) . Thus
( K ( r ) ) τ τ = f ′ 2 f , ( K ( r ) ) A B = f r δ A B , ( E 0 ( r ) ) τ τ = 2 μ r 3 , ( E 0 ( r ) ) A B = − μ r 3 δ A B .
Here B 0 ( r ) = 0 . The perturbation h τ A = H V A changes the second fundamental form by K ˙ τ A = ( f / 2 ) H ′ V A . Raising an index also varies the inverse induced metric. Both contributions are needed to obtain
K ˙ B τ = H ′ − 2 H / r 2 f V B .
The diagonal magnetic Weyl component follows from the symmetrized spatial curl of K in the sign convention of (92). This convention is the negative of that in [47] Appendix B, Eq. (B8). The connection variation involving Γ ˙ A B τ is symmetric in A , B and disappears under contraction with the spatial volume tensor. Our future-normal and orientation conventions give
ϵ τ A B = f r 2 ϵ ^ A B , ϵ ^ A B D ^ A V B = Δ H 2 Y .
The second identity follows from d ( * h H 2 d Y ) = ( Δ H 2 Y ) d V h H 2 . Consequently,
B ˙ τ τ = H ′ − 2 H / r 2 r 2 Δ H 2 Y .
The magnetic Weyl tensor is trace free. Its two angular components therefore enter the contraction with E 0 ( r ) through their sum, which is − B ˙ τ τ . Since B 0 ( r ) = 0 , the Pontryagin identity with our magnetic sign convention gives
P ˙ = 16 ( E 0 ( r ) ) τ τ + μ r 3 B ˙ τ τ = 24 μ r 6 ( r H ′ − 2 H ) Δ H 2 Y .
No field equation for H is used in this calculation.
An independent field-equation check fixes the radial sign in the axial reduction. For a circle-independent coexact axial field, all linearized Ricci components vanish identically except possibly the mixed τ A components. The τ r component vanishes by angular co-closedness. The same off-shell connection variation gives the mixed τ A components for V = * h H 2 d Y .
2 Ric ˙ τ A = f H ′ ′ + 4 μ r 3 H V A − H r 2 * h H 2 d ( Δ H 2 Y ) A .
This identity is valid for any smooth angular scalar Y, without assuming a master equation. If − Δ H 2 Y = λ Y and V ≠ 0 , its vacuum condition is
f H ′ ′ + λ r 2 + 4 μ r 3 H = 0 .
Conversely, the master equation (34) with H = − f Ω ′ gives r H ′ − 2 H = ( λ + 2 ) Ω / r . Differentiating r ( r H ′ − 2 H ) = ( λ + 2 ) Ω and substituting Ω ′ = − H / f recovers (A12). This verifies the sign of the negative-curvature axial specialization against a direct component of the linearized vacuum Einstein equations.

References

  1. Ori, A. Formation of closed timelike curves in a composite vacuum/dust asymptotically-flat spacetime. Phys. Rev. D.;Author 2007, arXiv:gr76, 044002. [Google Scholar] [CrossRef]
  2. Ori, A. A class of time-machine solutions with a compact vacuum core. Phys. Rev. Lett. 2005, 95, 021101. [Google Scholar] [CrossRef] [PubMed]
  3. Dietz, J.; Dirmeier, A.; Scherfner, M. Geometric analysis of particular compactly constructed time machine spacetimes. J. Geom. Phys. 2012, arXiv:1111.282262, 1273–1283. [Google Scholar] [CrossRef]
  4. Rieger, N. E. Causal classification of pathological Misner-type spacetimes. Ann. Phys. 2026, arXiv:2501.00841v5489, 170455. [Google Scholar] [CrossRef]
  5. Rauch, J.; Rendall, A. D. Blow-up of test fields near Cauchy horizons. Lett. Math. Phys. 1993, arXiv:gr29, 241–248. [Google Scholar] [CrossRef]
  6. Kodama, H.; Ishibashi, A. A master equation for gravitational perturbations of maximally symmetric black holes in higher dimensions. Prog. Theor. Phys. 2003, arXiv:hep110, 701–722. [Google Scholar] [CrossRef]
  7. Tanimoto, M.; Moncrief, V.; Yasuno, K. Perturbations of spatially closed Bianchi III spacetimes. Class. Quantum Grav. 2003, arXiv:gr20, 1879–1927. [Google Scholar] [CrossRef]
  8. Tanimoto, M. Linear perturbations of spatially locally homogeneous spacetimes. Recent Adv. Riemannian Lorentzian Geom. Contemp. Math. 2003, 337, 171–185. [Google Scholar] [CrossRef]
  9. Mukohyama, S. Gauge-invariant gravitational perturbations of maximally symmetric spacetimes. Phys. Rev. D. 2000, 62, 084015. [Google Scholar] [CrossRef]
  10. Kodama, H.; Ishibashi, A.; Seto, O. Brane world cosmology: Gauge-invariant formalism for perturbation. Phys. Rev. D. 2000, 62, 064022. [Google Scholar] [CrossRef]
  11. Cai, R.-G.; Cao, L.-M. Generalized formalism in gauge-invariant gravitational perturbations. Phys. Rev. D. 2013, arXiv:1306.492788, 084047. [Google Scholar] [CrossRef]
  12. Gerlach, U. H.; Sengupta, U. K. Gauge-invariant perturbations on most general spherically symmetric space-times. Phys. Rev. D. 1979, 19, 2268–2272. [Google Scholar] [CrossRef]
  13. Gerlach, U. H.; Sengupta, U. K. Junction conditions for odd-parity perturbations on most general spherically symmetric space-times. Phys. Rev. D. 1979, 20, 3009–3014. [Google Scholar] [CrossRef]
  14. Karlovini, M. “Axial perturbations of general spherically symmetric space-times,”. Cl. Quantum Grav. 2002, 19, 2125–2140. [Google Scholar] [CrossRef]
  15. Sarbach, O.; Tiglio, M. Gauge-invariant perturbations of Schwarzschild black holes in horizon-penetrating coordinates. Phys. Rev. D. 2001, 64, 084016. [Google Scholar] [CrossRef]
  16. Martel, K.; Poisson, E. Gravitational perturbations of the Schwarzschild spacetime: A practical covariant and gauge-invariant formalism. Phys. Rev. D. 2005, 71, 104003. [Google Scholar] [CrossRef]
  17. Chaverra, E.; Ortiz, N.; Sarbach, O. Linear perturbations of self-gravitating spherically symmetric configurations. Phys. Rev. D. 2013, arXiv:1209.3731v287, 044015. [Google Scholar] [CrossRef]
  18. Birmingham, D.; Mokhtari, S. Stability of topological black holes. Phys. Rev. D. 2007, arXiv:0709.238876, 124039. [Google Scholar] [CrossRef]
  19. Yunes, N.; Sopuerta, C. F. Perturbations of Schwarzschild black holes in Chern–Simons modified gravity. Phys. Rev. D. 2008, arXiv:0712.102877, 064007. [Google Scholar] [CrossRef]
  20. Bhattacharyya, S. Distinguishing General Relativity and Modified Theories of Gravity Using Quasi-Normal Modes. doctoral thesis, thesis PDF. IISER Thiruvananthapuram, 2019. [Google Scholar]
  21. Dotti, G. Nonmodal linear stability of the Schwarzschild black hole. Phys. Rev. Lett. 2014, arXiv:1307.3340112, 191101. [Google Scholar] [CrossRef] [PubMed]
  22. Fernández Tío, J. M.; Dotti, G. Black hole nonmodal linear stability under odd perturbations: The Reissner–Nordström case. Phys. Rev. D. 2017, arXiv:1607.00975v395, 124041. [Google Scholar] [CrossRef]
  23. Shah, A. G.; Whiting, B. F.; Aksteiner, S.; Andersson, L.; Bäckdahl, T. “Gauge-invariant perturbations of Schwarzschild spacetime,”. arXiv. 2016. Available online: https://arxiv.org/abs/1611.08291.
  24. Chruściel, P. T.; Isenberg, J.; Pollack, D. Initial data engineering. Commun. Math. Phys. 2005, arXiv:gr-qc/0403066v2257, 29–42. [Google Scholar] [CrossRef]
  25. Beig, R.; Chruściel, P. T. “Killing initial data,”. Cl. Quantum Grav. 1997, arXiv:gr-qc/960404014, A83–A92. [Google Scholar] [CrossRef]
  26. Delay, E.; Mazzieri, L. Refined gluing for vacuum Einstein constraint equations. Geom. Dedicata 2014, arXiv:1003.4178173, 393–415. [Google Scholar] [CrossRef]
  27. Carlotto, A.; Schoen, R. Localizing solutions of the Einstein constraint equations. Invent. Math. 2016, arXiv:1407.4766205, 559–615. [Google Scholar] [CrossRef]
  28. Czimek, S.; Rodnianski, I. Obstruction-free gluing for the Einstein equations. arXiv. 2022. Available online: https://arxiv.org/abs/2210.09663.
  29. Mao, Y.; Oh, S.-J.; Tao, Z. Initial data gluing in the asymptotically flat regime via solution operators with prescribed support properties. arXiv. 2023. Available online: https://arxiv.org/abs/2308.13031.
  30. Chruściel, P. T.; Delay, E. Exotic hyperbolic gluings. J. Differ. Geom. 2018, arXiv:1511.07858v3108, 243–293. [Google Scholar] [CrossRef]
  31. Chruściel, P. T.; Cogo, A.; Nützi, A. “A Bogovskiǐ-type operator for Corvino–Schoen hyperbolic gluing,”. Cl. Quantum Grav. 2025, arXiv:2409.0750242, 147001. [Google Scholar] [CrossRef]
  32. Choquet-Bruhat, Y.; Moncrief, V. Future global in time Einsteinian spacetimes with U(1) isometry group. Ann. Henri Poincaré 2001, arXiv:gr-qc/0112049v12, 1007–1064. [Google Scholar] [CrossRef]
  33. Choquet-Bruhat, Y. Future complete U(1) symmetric Einsteinian spacetimes, the unpolarized case. In The Einstein Equations and the Large Scale Behavior of Gravitational Fields; Chruściel, P. T., Friedrich, H., Eds.; Birkhäuser, 2004; pp. 251–298. [Google Scholar] [CrossRef]
  34. Gicquaud, R.; Huneau, C. Limit equation for vacuum Einstein constraints with a translational Killing vector field in the compact hyperbolic case. J. Geom. Phys. 2016, arXiv:1409.3477v1107, 175–186. [Google Scholar] [CrossRef]
  35. Huneau, C. Constraint equations for 3+1 vacuum Einstein equations with a translational space-like Killing field in the asymptotically flat case. Ann. Henri Poincaré 2016, arXiv:1302.1473v217, 271–299. [Google Scholar] [CrossRef]
  36. Berger, B. K.; Moncrief, V. Exact U(1) symmetric cosmologies with local Mixmaster dynamics. Phys. Rev. D. 2000, arXiv:gr-qc/0001083v162, 023509. [Google Scholar] [CrossRef]
  37. Geroch, R. A method for generating new solutions of Einstein’s equation. II. J. Math. Phys. 1972, 13, 394–404. [Google Scholar] [CrossRef]
  38. Fischer, A. E.; Marsden, J. E.; Moncrief, V. The structure of the space of solutions of Einstein’s equations. I. One Killing field. Ann. Inst. H. Poincaré Sect. A 1980, 33, 147–194. Available online: https://www.numdam.org/item/AIHPA_1980__33_2_147_0/.
  39. Arms, J. M.; Marsden, J. E.; Moncrief, V. The structure of the space of solutions of Einstein’s equations. II. Several Killing fields and the Einstein–Yang–Mills equations. Ann. Phys. 1982, 144, 81–106. [Google Scholar] [CrossRef]
  40. Carlotto, A. The general relativistic constraint equations. Living Rev. Relativ. 2021, 24, 2. [Google Scholar] [CrossRef]
  41. Moncrief, V. Symmetry preserving deformations of generalized Taub–NUT space-times to all orders in perturbation theory. J. Math. Phys. 1989, 30, 2297–2314. [Google Scholar] [CrossRef]
  42. Bernard, D.; Julia, B. Twisted self-duality of dimensionally reduced gravity and vertex operators. Nucl. Phys. B 1999, arXiv:hep547, 427–470. [Google Scholar] [CrossRef]
  43. Penna, R. F. Einstein–Rosen waves and the Geroch group. J. Math. Phys. 2021, arXiv:2106.13252v262, 082503. [Google Scholar] [CrossRef]
  44. Dodziuk, J. L2 harmonic forms on rotationally symmetric Riemannian manifolds. Proc. Amer. Math. Soc. 1979, 77, 395–400. [Google Scholar] [CrossRef]
  45. Helgason, S. Non-Euclidean Analysis. In Non-Euclidean Geometries; Theorem 2.2 and Eq. (3.5) are cited from the 2004 preprint; Prékopa, A., Molnár, E., Eds.; Springer, 2006; vol. 581, pp. 367–384. [Google Scholar] [CrossRef] [PubMed]
  46. Hintz, P. The linearized Einstein equations with sources. Lett. Math. Phys. 2024, 114, 94. [Google Scholar] [CrossRef]
  47. Okounkova, M.; Stein, L. C.; Scheel, M. A.; Hemberger, D. A. Numerical binary black hole mergers in dynamical Chern–Simons gravity: Scalar field. Phys. Rev. D. 2017, arXiv:1705.07924v296, 044020. [Google Scholar] [CrossRef]
  48. Geroch, R. A method for generating solutions of Einstein’s equations. J. Math. Phys. 1971, 12, 918–924. [Google Scholar] [CrossRef]
  49. Moncrief, V. Reduction of Einstein’s equations for vacuum space-times with spacelike U(1) isometry groups. Ann. Phys. 1986, 167, 118–142. [Google Scholar] [CrossRef]
  50. Moncrief, V. The reduction problem for Einstein’s equations. In Conference on Mathematical Relativity, Centre for Mathematics and Its Applications, Australian National University, Proceedings; Bartnik, R., Ed.; 1989; Vol. 19, pp. 168–174. Available online: https://maths.anu.edu.au/files/CMAProcVol19-Moncrief.pdf.
  51. Rendall, A. D. Local and global existence theorems for the Einstein equations. Living Rev. Relativ. 2000, 3, article 1. [Google Scholar] [CrossRef] [PubMed]
  52. Damour, T.; Schmidt, B. Reliability of perturbation theory in general relativity. J. Math. Phys. 1990, 31, 2441–2453. [Google Scholar] [CrossRef]
  53. Fischer, A. E.; Marsden, J. E. The Einstein evolution equations as a first-order quasi-linear symmetric hyperbolic system. I. Commun. Math. Phys. 1972, 28, 1–38. [Google Scholar] [CrossRef]
  54. Moncrief, V. Infinite-dimensional family of vacuum cosmological models with Taub–NUT (Newman–Unti–Tamburino)-type extensions. Phys. Rev. D. 1981, 23, 312–315. [Google Scholar] [CrossRef]
  55. Moncrief, V. Neighborhoods of Cauchy horizons in cosmological spacetimes with one Killing field. Ann. Phys. 1982, 141, 83–103. [Google Scholar] [CrossRef]
  56. Moncrief, V. The asymptotic behavior of nonlinear waves near a cosmological Cauchy horizon. J. Math. Phys. 1989, 30, 1760–1768. [Google Scholar] [CrossRef]
  57. Grubišić, B.; Moncrief, V. Asymptotic behavior of the T3×R Gowdy space-times. Phys. Rev. D. 1993, arXiv:gr-qc/920900647, 2371–2382. [Google Scholar] [CrossRef] [PubMed]
  58. Moncrief, V.; Isenberg, J. Symmetries of cosmological Cauchy horizons. Commun. Math. Phys. 1983, 89, 387–413. [Google Scholar] [CrossRef]
  59. Isenberg, J.; Moncrief, V. Symmetries of cosmological Cauchy horizons with exceptional orbits. J. Math. Phys. 1985, 26, 1024–1027. [Google Scholar] [CrossRef]
  60. Moncrief, V. The space of (generalized) Taub–NUT spacetimes. J. Geom. Phys. 1984, 1, 107–130. [Google Scholar] [CrossRef]
  61. Moncrief, V. An infinite-dimensional family of curvature singular vacuum spacetimes with U(1) isometry groups. Class. Quantum Grav. 1987, 4, 1555–1563. [Google Scholar] [CrossRef]
  62. Petersen, O.; Rácz, I. Symmetries of vacuum spacetimes with a compact Cauchy horizon of constant nonzero surface gravity. Ann. Henri Poincaré 2023, 24, 3921–3943. [Google Scholar] [CrossRef]
  63. Maeda, K.; Ishibashi, A. Strong cosmic censorship and causality violation. Phys. Rev. D. 1997, arXiv:gr-qc/960705256, 2094–2099. [Google Scholar] [CrossRef]
  64. Maeda, K.; Ishibashi, A.; Narita, M. “Chronology protection and non-naked singularity,”. Cl. Quantum Grav. 1998, arXiv:gr-qc/9803020v115, 1637–1651. [Google Scholar] [CrossRef]
  65. Yurtsever, U. Instability of Killing–Cauchy horizons in plane-symmetric spacetimes. Phys. Rev. D. 1987, 36, 1662–1672. [Google Scholar] [CrossRef] [PubMed]
  66. Yurtsever, U. Structure of the singularities produced by colliding plane waves. Phys. Rev. D. 1988, 38, 1706–1730. [Google Scholar] [CrossRef] [PubMed]
  67. Yurtsever, U. Singularities and horizons in the collisions of gravitational waves. Phys. Rev. D. 1989, 40, 329–359. [Google Scholar] [CrossRef] [PubMed]
  68. Luk, J.; Oh, S.-J. Proof of linear instability of the Reissner–Nordström Cauchy horizon under scalar perturbations. Duke Math. J. 2017, arXiv:1501.04598166, 437–493. [Google Scholar] [CrossRef]
  69. Sbierski, J. Instability of the Kerr Cauchy horizon under linearised gravitational perturbations. Ann. PDE 2023, arXiv:2201.122959, 7. [Google Scholar] [CrossRef]
  70. Gurriaran, S. “Non-linear instability of the Kerr Cauchy horizon near i+,” preprint (2026). Available online: https://arxiv.org/abs/2603.17911v2.
  71. Luk, J.; Sbierski, J. “The formation of a weak null singularity in the interior of generic rotating black holes,”. arXiv 2026, arXiv:2604.04877v1. Available online: https://arxiv.org/abs/2604.04877v1.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.