Horizons of some asymptotically stationary spacetimes

Peter Hintz

REVIEW 3 major objections 4 minor 10 references

Horizons of some asymptotically stationary spacetimes

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read For spacetimes settling to a stationary black hole, a unique smooth horizon emerges at late times.

desk verdict A clean, useful extension of Chen–Klainerman's horizon-regularity result via a general unstable manifold theorem; the local result is solid, and the event-horizon identification is honest but conditional on a global escape assumption that the paper does not derive. read the letter →

arxiv 2411.12568 v1 pith:FQ5FSIIK submitted 2024-11-19 gr-qc math.DGmath.DS

Peter Hintz This is my paper · ORCID

classification gr-qcmath.DGmath.DS MSC 37C7083C5753B50

keywords eventhorizonnullhypersurfaceunstablemanifoldnormalsinkKerr–Newmanspacetimeasymptoticallystationaryspacetimesnull-geodesicflowdynamicalblackholes

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proves that a dynamical spacetime which decays to a stationary spacetime containing a horizon $\mathcal{H}_0$ must contain a unique smooth null hypersurface $\mathcal{H}$ asymptotic to $\mathcal{H}_0$. The result is cast as a general unstable-manifold theorem for perturbations of flows that translate in time and have a normal sink at an invariant manifold. For asymptotically subextremal Kerr–Newman spacetimes, the paper further shows that $\mathcal{H}$ is exactly the boundary of the black hole region, i.e. the future event horizon. A sympathetic reader cares because it turns a global, hard-to-access object—the event horizon—into a late-time local object that is uniquely determined by the asymptotic metric alone.

What carries the argument

The machinery is a general unstable-manifold theorem for maps that shift time by one and contract the normal bundle at an invariant submanifold. In the spacetime application, the invariant manifold is the conormal bundle of the unperturbed horizon, viewed in the spherical cotangent bundle, and the contracting map is the time-one null-geodesic flow. The key mechanism is that the linearization of the null-geodesic vector field along the horizon has a strictly negative normal Lyapunov exponent, making the horizon a normal sink; the unstable-manifold theorem then produces a unique invariant manifold carrying the perturbed horizon, and a geometric argument shows its base projection is null and ruled by null geodesics.

What would settle it

Construct an asymptotically stationary metric $g=g_0+h$ with $h\in\rho C^\infty_b$ and a compact far-field shell that traps null geodesics, so that some points at arbitrarily large radius cannot reach future null infinity while the metric outside the shell is exactly Kerr–Newman. If Theorem 3.4's conclusion is correct, the unique $\mathcal{H}$ from Theorem 3.2 would still have to bound the black hole region, whereas the engineered escape failure would make the true event horizon deviate from $\mathcal{H}$.

Watch

Extended reading notes

Core claim

The central discovery is Theorem 3.2: given a Lorentzian metric $g$ differing from a stationary Kerr–Newman-type metric $g_0$ by a term in $\rho C^\infty_b$, there is a $t_0$ and a unique smooth null hypersurface $\mathcal{H}\subset\{t_*\ge t_0\}$ that approaches the unperturbed horizon $\mathcal{H}_0$ in a $\rho C^\infty_b$ sense and is ruled by null geodesics. Theorem 3.4 adds that, under a mild global far-field condition, $\mathcal{H}$ coincides with the event horizon of $(M,g)$, defined as the boundary of the causal past of future null infinity. The argument constructs $\mathcal{H}$ as the base projection of an unstable manifold in the spherical cotangent bundle, obtained from the phase space of the null-geodesic flow; the horizon is therefore produced by a contraction mechanism, not by a global causal analysis.

Load-bearing premise

The load-bearing premise is the global condition in Theorem 3.4 that every sufficiently distant point can be joined to future null infinity by a future causal curve; if that fails, the locally constructed $\mathcal{H}$ need not be the boundary of $J^-(\mathcal{I}^+)$, only a unique asymptotic null hypersurface.

Editorial extensions

If this is right

If $g$ decays to any stationary Kerr–Newman-type metric with a non-degenerate horizon, the late-time horizon $\mathcal{H}$ exists and is unique, independent of gauge or foliation choices.
In subextremal Kerr and Kerr–Newman, $\mathcal{H}$ is the actual event horizon, so the event horizon of a perturbed black hole is smooth rather than merely Lipschitz.
The same construction covers event and cosmological horizons of subextremal Kerr–Newman–de Sitter spacetimes, with the analogous conclusion relative to the conformal boundary.
If the perturbation is small in $\rho C^2_b$, the horizon can be constructed from $t_0=0$, i.e. on the entire future.
Under the global far-field assumption, every future causal curve starting on the far side of $\mathcal{H}$ escapes to null infinity, while curves starting on the other side stay in the black hole region.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Beyond the paper: the same unstable-manifold mechanism may identify other distinguished null hypersurfaces—for instance inner horizons or Cauchy horizons—whenever the normal contraction is replaced by the appropriate expansion.
Beyond the paper: the uniqueness statement suggests that horizon extraction at late times is structurally stable; any two numerical or analytical procedures that track the apparent horizon should converge to the same hypersurface once the spacetime settles.
A testable extension would replace the $C^\infty_b$ decay weight by weaker or non-smooth weights and ask whether the perturbed horizon inherits exactly the regularity of the perturbation rather than remaining smooth.
The far-field escape condition of Theorem 3.4 could be probed numerically by constructing a metric with a localized shell that traps null geodesics, then checking whether the locally unique $\mathcal{H}$ still equals the boundary of $J^-(\mathcal{I}^+)$.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

First thing you should know: this is a worthwhile paper, and the local horizon-construction part holds up. Hintz recasts Chen–Klainerman's Kerr horizon regularity as an application of a general unstable manifold theorem for maps that translate in time and have a normal sink at an invariant manifold. The machinery is genuinely general—Theorem 2.2 is stated abstractly and proved by reduction to [Hin21, Theorem 2.3] plus an invariant-subbundle argument from [HPS77]. Then he checks the normal-sink condition for the projected null geodesic flow near subextremal Kerr–Newman horizons (Lemma 3.1) and gets existence and uniqueness of a smooth null hypersurface asymptotic to the stationary horizon (Theorem 3.2). That is a real step beyond Chen–Klainerman, which treats subextremal Kerr only, and the proof is honest: it cites the heavy PDE machinery rather than re-deriving it.

Where the paper is softest is Theorem 3.4, the event-horizon identification. The local theorem proves existence of H, but showing H equals the boundary of J^-(I+) needs a global far-field escape condition: every point far outside (r >= R0) must be able to reach future null infinity. That condition is assumed, not derived from the local decay hypotheses. Remark 3.5 gives sufficient conditions, but they are exactly that—extra assumptions. If they fail, H may still be a smooth null hypersurface in the local sense while not being the event horizon. So the advertised physical conclusion is conditional. This is a presentation gap rather than a discovered error, but it should be fixed in revision: either state Theorem 3.4 with the escape condition as an explicit hypothesis (it nearly does) or prove it from stronger decay/global geometric hypotheses.

Minor: the de Sitter/cosmological-horizon case, where mu'(r+) < 0, is only sketched as 'track the sign changes.' That is fine for a note, but it should be written out or explicitly deferred. Also, there is a small unjustified step in the nullness reduction where an arbitrary X in T H at t* = t1 is asserted to have g(L,X) = 0; the later B/E decomposition seems to give an independent route to nullness, so this does not threaten Theorem 3.2.

Bottom line: the central local theorem is solid and novel; the event-horizon statement is honest but conditional. This deserves peer review. A serious referee should focus on Theorem 3.4's hypotheses and the sign-case write-up. I would bring it to a reading group and would cite the abstract unstable manifold theorem.

Referee Report

3 major / 4 minor

Summary. This paper develops a general unstable-manifold theorem (Theorem 2.2) for smooth maps that translate in time and have a normal sink at an invariant manifold, extending the machinery of [Hin21] and [HPS77]. It then applies this theorem to the time-1 null geodesic flow lifted to the spherical cotangent bundle near the conormal bundle of a non-degenerate horizon in Kerr-type spacetimes. Theorem 3.2 establishes that for any Lorentzian metric g on a neighborhood of the horizon with g − g0 ∈ ρC∞b, there exists a unique smooth null hypersurface H asymptotic to the unperturbed horizon H0 with rate ρ. Theorem 3.4 identifies H with the event horizon of the perturbed spacetime in the subextremal Kerr–Newman case, conditional on an additional global far-field escape assumption (existence of R0); Remark 3.5 gives sufficient far-field decay conditions for that assumption. The abstract claims the event-horizon identification for Kerr(-Newman) without this qualification.

Significance. If the flagged issues are repaired, this is a valuable conceptual contribution: it recasts horizon regularity as a normal-sink/unstable-manifold statement and gives a short, transparent alternative proof of the recent Chen–Klainerman horizon-regularity theorem [CK24]. The reliance on [Hin21] and [HPS77] is a legitimate dependency, not a circularity, and the derivation is parameter-free. Theorem 3.2, the core result, is sound in substance: although the intended nullness reduction contains a circular step, the later decomposition argument (3.8) supplies a valid independent proof of nullness. The principal weaknesses are the unproved global escape condition behind Theorem 3.4, the abstract's unqualified statement of the event-horizon identification, and the deferred µ′(r+) < 0 (cosmological horizon) case.

major comments (3)

[Proof of Theorem 3.2, 'H is null: reduction'] The identity 'g(L, X̃) = g(L, X) = 0 at t∗ = t1' assumes that every X ∈ T_{z0}H on the slice t∗ = t1 is g-orthogonal to the null generator L; this is exactly the statement that T_{z0}H = L⊥, i.e. that H is null at that slice, the property the argument is meant to establish. For a hypersurface that is only known to be ruled by null geodesics, g(L, X) = 0 holds for X in a proper hyperplane of T_{z0}H, not for all X. The later argument 'H is null: conclusion' based on the decomposition (3.8) provides an independent and valid proof of nullness, so Theorem 3.2 survives, but the reduction step as written is circular and should be removed or replaced.
[Theorem 3.4 and Remark 3.5] The conclusion that H is the event horizon is conditional on the assumed existence of R0 such that every point in {t∗ ≥ 0, r ≥ R0} can be connected to future null infinity by a future causal curve. This is a genuine additional global hypothesis: it is not derived from g − g0 ∈ ρC∞b (which controls decay as t∗ → ∞ at fixed r) nor from the local normal-sink dynamics, and the proof uses it exactly once to show E ⊂ J−(I+). Remark 3.5 gives sufficient far-field decay conditions for the R0 assumption, but it does not show that these conditions follow from the hypotheses of Theorem 3.4, and the theorem does not state them as hypotheses. The abstract's claim that in the Kerr(-Newman) case 'we show that H is equal to the boundary of the black hole region' is therefore stronger than the statement proved in Theorem 3.4; please either add an explicit far-field hypothesis to Theorem 3.4 and qualify the abstract accordingly, or prove the escape condition from natural asymptotic-flatness assumptions.
[§3 footnote 1 and the abstract] The abstract and the introduction list event and cosmological horizons of subextremal Kerr–Newman–de Sitter as examples to which the results apply, but the construction of §3 assumes µ′(r+) > 0 throughout, and the positivity of κ = µ′(r+)/(2b(r²+ + a²)) is used in the normal-sink estimate of Lemma 3.1 and in the sign structure of the sets B and E in the proof of Theorem 3.2. The footnote deferring the µ′(r+) < 0 (cosmological horizon) case to 'sign changes in the arguments below' does not establish the claimed scope, since the sign of µ′ flips the timelike/spacelike character of dr on the two sides of the horizon and exchanges the roles of the two sides. Please either work out the cosmological-horizon case or restrict the claims in the abstract and introduction.

minor comments (4)

[Proof of (3.8) in Theorem 3.2] In the case of a maximal future null geodesic γ with t∗ ◦ γ bounded and ¯r ≥ r+, the assertion that γ 'forces r(γ(s)) ≥ r+ + δ for some s' is not justified: the geodesic could in principle exit through r = r+ − 2δ while having previously reached values ≥ r+. A monotonicity or no-return argument (for example using that dr is timelike below r+ for large t∗) is needed to exclude this alternative.
[Proof of Theorem 3.4] The proof works with B and E defined by (3.7) on the extended manifold X = [r+ − 2δ, ∞) × S² and invokes the decomposition (3.8), but (3.8) is proved in Theorem 3.2 only on the local spacetime with bounded r; the extension of the decomposition to the extended spacetime should be stated explicitly. Also, the assertion '(B ∪ H) ∩ J−(I+) = ∅' is made without proof; it follows from (3.8) together with E ⊂ J−(I+) and the fact that a causal curve reaching I+ witnesses membership in E, but the derivation should be spelled out, as should the derivation of items (1)–(3) of the theorem.
[§3, especially (3.5) and the proof of Theorem 3.2] The symbol X is used both for the physical space S² × (r+ − 2δ, r+ + 2δ) in the statement of Theorem 3.2 and for the phase space R_ˆσ × (T∗S²) in (3.5) and in the definition of the map F in the proof; this makes the proof of Theorem 3.2 very hard to follow, and distinct notation should be used for the two spaces.
[Statement of Theorem 3.4] In the statement of Theorem 3.4, B and E are defined as the components of (M ∩ {t∗ ≥ t0}) \ H on which r is bounded, resp. unbounded, while in the proof they are redefined by (3.7) with t0 in place of t1; the identification of the two definitions is part of the conclusion. Please clarify this by using separate symbols or by stating explicitly that the (3.7)-defined sets are provisional and their equality with the components is established in the proof.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the construction and event-horizon identification follow from an independent unstable-manifold theorem plus explicitly stated assumptions, not from the target conclusion.

full rationale

The derivation chain is self-contained in the relevant sense. Theorem 2.2 is a general unstable-manifold result proved by reducing to the author's earlier [Hin21, Theorem 2.3] and to the classical [HPS77, Theorem 3.5]; neither cited theorem contains the present paper's target conclusion, and the hypotheses of those theorems (normal sink, r-normal hyperbolicity) are checked here in Lemma 3.1 by a direct computation of the linearization L_{\bar V} and the induced normal contraction rate. Theorem 3.2 therefore genuinely proves existence and uniqueness of a smooth null hypersurface H asymptotic to H0 under the decay assumption g-g0 ∈ ρC^∞_b; uniqueness comes from the invariant-manifold uniqueness statement, not from an assumption that H is the horizon. Theorem 3.4's global identification of H with the event horizon is conditional on an explicitly stated far-field escape hypothesis: the existence of R0 such that every point in {t* ≥ 0, r ≥ R0} can be connected to future null infinity I+ by a causal curve. This is an additional assumption, not a disguised form of the conclusion; it does not mention H or the event horizon of (M,g), and the proof uses it to derive the inclusions E ⊂ J^-(I+) and (B ∪ H) ∩ J^-(I+) = ∅. No fitted parameter is renamed as a prediction, no quantity is defined in terms of what it is supposed to predict, and no ansatz is imported solely by self-citation. The reliance on [Hin21] is heavy but legitimate, since that earlier theorem's assumptions do not include the present result. Accordingly, no circular step is present and the paper should receive the lowest circularity score.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

No free parameters are fitted; the physical constants of Kerr-Newman are inputs from the stationary family. No new physical entities are introduced. The objects Gamma^u and H are mathematical constructs, and the contraction rate kappa is computed from mu'(r+) of the unperturbed metric.

assumptions (3)

standard math Hirsch-Pugh-Shub invariant manifold theory, specifically [HPS77, Theorem 3.5], and Hin21's Theorem 2.3 for normally hyperbolic trapping.
Used in Section 2 to construct the invariant subbundle and the unstable manifold Gamma^u; these are cited as established theorems, not proved in this paper.
domain assumption The spacetime is asymptotically stationary: the perturbed metric satisfies g - g0 in rho C^infinity_b on {t* >= 0}, with t* level sets spacelike and transversal to H0, and the horizon is non-degenerate with mu'(r+) > 0.
These are the hypotheses of Theorem 3.2 and Lemma 3.1 defining the class of spacetimes; without them the normal-sink argument does not run.
domain assumption Global far-field condition: there exists R0 such that every point in {t* >= 0, r >= R0} can be connected to future null infinity I+ by a future causal curve.
Invoked in Theorem 3.4 to identify the local hypersurface H with the event horizon; this is an external geometric condition on the far end of the spacetime.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Horizons of some asymptotically stationary spacetimes." pith.science (2026). https://pith.science/paper/FQ5FSIIK

@misc{pith2026241112568,
  author       = {Pith},
  title        = {Pith review of: Horizons of some asymptotically stationary spacetimes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FQ5FSIIK}},
  note         = {Machine review of arXiv:2411.12568}
}

abstract

On a class of dynamical spacetimes which are asymptotic as $t\to\infty$ to a stationary spacetime containing a horizon $\mathcal{H}_0$, we show the existence of a unique null hypersurface $\mathcal{H}$ which is asymptotic to $\mathcal{H}_0$. This is a special case of a general unstable manifold theorem for perturbations of flows which translate in time and have a normal sink at an invariant manifold in space. Examples of horizons $\mathcal{H}_0$ to which our result applies include event horizons of subextremal Kerr and Kerr-Newman black holes as well as event and cosmological horizons of subextremal Kerr-Newman-de Sitter black holes. In the Kerr(-Newman) case, we show that $\mathcal{H}$ is equal to the boundary of the black hole region of the dynamical spacetime.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 6 canonical work pages

[1]

Regularity of the F uture E vent H orizon in P erturbations of K err

Xuantao Chen and Sergiu Klainerman. Regularity of the F uture E vent H orizon in P erturbations of K err. Preprint, arXiv:2409.05700 , 2024

arXiv 2024
[2]

The nonlinear stability of the S chwarzschild solution to gravitational perturbations

Mihalis Dafermos, Gustav Holzegel, Igor Rodnianski, and Martin Taylor. The nonlinear stability of the S chwarzschild solution to gravitational perturbations. Preprint, arXiv:2104.08222 , 2021

arXiv 2021
[3]

The null-geodesic flow near horizons

Oran Gannot. The null-geodesic flow near horizons. Transactions of the American Mathematical Society , 371(7):4769--4791, 2019

work page 2019
[4]

Wave equations estimates and the nonlinear stability of slowly rotating K err black holes

Elena Giorgi, Sergiu Klainerman, and Jeremie Szeftel. Wave equations estimates and the nonlinear stability of slowly rotating K err black holes. Preprint, arXiv:2205.14808 , 2022

arXiv 2022
[5]

Normally hyperbolic trapping on asymptotically stationary spacetimes

Peter Hintz. Normally hyperbolic trapping on asymptotically stationary spacetimes. Probability and Mathematical Physics , 2(1):71--126, 2021. https://doi.org/10.2140/pmp.2021.2.71 doi:10.2140/pmp.2021.2.71

work page doi:10.2140/pmp.2021.2.71 2021
[6]

Hirsch, Charles C

Morris W. Hirsch, Charles C. Pugh, and Michael Shub. Invariant manifolds , volume 583. Springer-Verlag, Berlin-New York, 1977

work page 1977
[7]

Kerr stability for small angular momentum

Sergiu Klainerman and J \'e r \'e mie Szeftel. Kerr stability for small angular momentum. Pure and Applied Mathematics Quarterly , 19(3):791--1678, 2023

work page 2023
[8]

Griffiths

Ji r \' Podolsk \`y and Jerry B. Griffiths. A ccelerating K err-- N ewman black holes in (anti-)de S itter space-time. Physical Review D , 73(4):044018, 2006

work page 2006

Show all 10 references

[9]

Construction of GCM hypersurfaces in perturbations of K err

Dawei Shen. Construction of GCM hypersurfaces in perturbations of K err. Annals of PDE , 9(1):1--112, 2023

2023
[10]

Microlocal analysis of asymptotically hyperbolic and K err--de S itter spaces (with an appendix by S emyon D yatlov)

Andr \'a s Vasy. Microlocal analysis of asymptotically hyperbolic and K err--de S itter spaces (with an appendix by S emyon D yatlov). Invent. Math. , 194(2):381--513, 2013. https://doi.org/10.1007/s00222-012-0446-8 doi:10.1007/s00222-012-0446-8

2013 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Regularity of the F uture E vent H orizon in P erturbations of K err

Xuantao Chen and Sergiu Klainerman. Regularity of the F uture E vent H orizon in P erturbations of K err. Preprint, arXiv:2409.05700 , 2024

arXiv 2024

[2] [2]

The nonlinear stability of the S chwarzschild solution to gravitational perturbations

Mihalis Dafermos, Gustav Holzegel, Igor Rodnianski, and Martin Taylor. The nonlinear stability of the S chwarzschild solution to gravitational perturbations. Preprint, arXiv:2104.08222 , 2021

arXiv 2021

[3] [3]

The null-geodesic flow near horizons

Oran Gannot. The null-geodesic flow near horizons. Transactions of the American Mathematical Society , 371(7):4769--4791, 2019

work page 2019

[4] [4]

Wave equations estimates and the nonlinear stability of slowly rotating K err black holes

Elena Giorgi, Sergiu Klainerman, and Jeremie Szeftel. Wave equations estimates and the nonlinear stability of slowly rotating K err black holes. Preprint, arXiv:2205.14808 , 2022

arXiv 2022

[5] [5]

Normally hyperbolic trapping on asymptotically stationary spacetimes

Peter Hintz. Normally hyperbolic trapping on asymptotically stationary spacetimes. Probability and Mathematical Physics , 2(1):71--126, 2021. https://doi.org/10.2140/pmp.2021.2.71 doi:10.2140/pmp.2021.2.71

work page doi:10.2140/pmp.2021.2.71 2021

[6] [6]

Hirsch, Charles C

Morris W. Hirsch, Charles C. Pugh, and Michael Shub. Invariant manifolds , volume 583. Springer-Verlag, Berlin-New York, 1977

work page 1977

[7] [7]

Kerr stability for small angular momentum

Sergiu Klainerman and J \'e r \'e mie Szeftel. Kerr stability for small angular momentum. Pure and Applied Mathematics Quarterly , 19(3):791--1678, 2023

work page 2023

[8] [8]

Griffiths

Ji r \' Podolsk \`y and Jerry B. Griffiths. A ccelerating K err-- N ewman black holes in (anti-)de S itter space-time. Physical Review D , 73(4):044018, 2006

work page 2006