Sub-static splitting obstruction
A Positivity Obstruction to Nontrivial Line Dependence in Sub-static Splittings
Shengrong Wu
National University of Singapore
Automatic Pipeline with ChatGPT
TL;DR Wang’s recent work on conformally compactifiable static spaces leaves an explicit gap in the sub-static case: after a line-splitting argument one gets \(V_*(r,y)=\alpha(r)\beta(y)\), but it is not clear when \(\alpha\) must be constant. I do not solve that full problem. What I can prove is a sharp obstruction: in the separated ansatz \(g=\beta(y)^2dt^2+h(y)\) and \(V=A(t)\beta(y)\), any bounded nonconstant \(A\) forces \[\mathcal Q_h(\beta)=\beta R_h+(n-3)\Delta_h\beta+(n-1)\beta^{-1}|\nabla_h\beta|^2\] to be strictly positive everywhere. So if \(\mathcal Q_h(\beta)\le 0\) at one point, then \(A\) is constant.
Introduction
A recurring theme in geometric analysis is that a splitting theorem usually hides a stronger rigidity mechanism. Once a manifold splits off a line, there is often a second argument showing that the remaining geometric data do not actually depend on that line variable.
In the classical static-vacuum setting, Wang’s recent paper on the Riccati equation for static spaces pushes this line of thought quite far. He reproves a splitting theorem for conformally compactifiable static spaces by combining Busemann-function ideas with a Riccati-type inequality. In the sub-static setting, however, the same strategy stalls at a very precise place: one still gets a partial split, but only of the form \[ g=V_*^2\,dr^2+h, \qquad V_*(r,y)=\alpha(r)\beta(y), \] and the proof no longer forces \(\alpha\) to be constant.
That is the bottleneck I focused on.
I did not try to force the full theorem immediately. That would have been the easiest way to overpromise. Instead, I asked a narrower structural question:
If we already know the separated form \[ g=\beta(y)^2dt^2+h(y), \qquad V(t,y)=A(t)\beta(y), \] what does the sub-static tensor inequality actually say about the line factor \(A\)?
This turns out to have a clean answer.
Why this direction is interesting
There are two reasons this is a worthwhile target.
First, it is attached to an explicit unresolved step in a recent paper, not a vague “maybe one can generalize this” hope. Wang literally identifies the missing issue: how does the tangential inequality \(S_\Sigma\ge 0\) control the factor \(\alpha\)?
Second, the answer is not just a one-off estimate. It gives a necessary geometric condition for nontrivial line dependence. In other words, it does not merely say “here is one situation where \(\alpha\) is constant”; it says “if \(\alpha\) is not constant, the slice geometry must satisfy a very restrictive positivity property everywhere.”
That is the kind of statement that often becomes useful in a second-stage program.
Setup
Let \((N^{n-1},h)\) be a connected Riemannian manifold, let \(\beta>0\) be smooth on \(N\), and let \(A>0\) be smooth on \(\mathbb R\). Consider \[ M=\mathbb R\times N,\qquad g=\beta(y)^2dt^2+h(y),\qquad V(t,y)=A(t)\beta(y). \]
The sub-static tensor is \[ S(V,g):=V\operatorname{Ric}_g+(\Delta_gV)g-\nabla_g^2V. \]
The sub-static condition is the tensor inequality \[ S(V,g)\ge 0. \]
The slice quantity that ends up mattering is \[ \mathcal Q_h(\beta):=\beta R_h+(n-3)\Delta_h\beta+(n-1)\beta^{-1}|\nabla_h\beta|^2. \]
Main theorem
Here is the precise statement.
Theorem
Assume \[ g=\beta(y)^2dt^2+h(y), \qquad V(t,y)=A(t)\beta(y), \] and \[ S(V,g)\ge 0. \] Assume also that \(A\) is bounded on \(\mathbb R\).
Then \[ A \text{ nonconstant } \Longrightarrow \mathcal Q_h(\beta)>0 \quad \text{everywhere on }N. \]
Equivalently, if there exists \(y_0\in N\) such that \[ \mathcal Q_h(\beta)(y_0)\le 0, \] then \(A\) is constant.
The theorem is intentionally one-sided. It identifies a necessary condition for nontrivial line dependence; it does not try to classify all cases with \(\mathcal Q_h(\beta)>0\).
Proof sketch
The proof is basically one tensor computation plus one elementary ODE observation.
Step 1: compute the geometry of the ansatz
Because \[ g_{tt}=\beta^2,\qquad g_{ij}=h_{ij},\qquad g_{ti}=0, \] the nontrivial Christoffel symbols are \[ \Gamma^t_{it}=\beta^{-1}\beta_i,\qquad \Gamma^i_{tt}=-\beta\,\beta^i,\qquad \Gamma^k_{ij}=\Gamma^k_{ij}(h). \]
For \(V=A(t)\beta(y)\) one gets \[ \nabla^2_gV(\partial_t,\partial_t)=\beta\big(A''+A|\nabla_h\beta|^2\big), \] \[ \nabla^2_gV(\partial_i,\partial_t)=0, \] \[ \nabla^2_gV(\partial_i,\partial_j)=A\,\nabla_h^2\beta(\partial_i,\partial_j), \] and \[ \Delta_gV=\beta^{-1}A''+A\Delta_h\beta+A\beta^{-1}|\nabla_h\beta|^2. \]
Also, \[ \operatorname{Ric}_g(\partial_t,\partial_t)=-\beta\Delta_h\beta, \] \[ \operatorname{Ric}_g(\partial_i,\partial_j)=\operatorname{Ric}_h(\partial_i,\partial_j)-\beta^{-1}\nabla_h^2\beta(\partial_i,\partial_j). \]
Step 2: compute the sub-static tensor
Substituting into \[ S(V,g)=V\operatorname{Ric}_g+(\Delta_gV)g-\nabla_g^2V \] gives \[ S_{tt}=0,\qquad S_{ti}=0, \] and \[ S_{ij} = A\Big(\beta\,\operatorname{Ric}_h{}_{ij} -2\nabla_h^2\beta_{ij} +\big(\Delta_h\beta+\beta^{-1}|\nabla_h\beta|^2\big)h_{ij}\Big) +\beta^{-1}A''h_{ij}. \]
Now take the \(h\)-trace: \[ \operatorname{tr}_h(S|_{TN}) = A\,\mathcal Q_h(\beta)+(n-1)\beta^{-1}A''. \]
This is the key identity.
Step 3: evaluate at one bad slice point
Suppose there exists \(y_0\in N\) such that \[ \mathcal Q_h(\beta)(y_0)\le 0. \] Because \(S\ge 0\), its restriction to \(T_{y_0}N\) is nonnegative, hence \[ \operatorname{tr}_h(S|_{T_{y_0}N})\ge 0. \] Therefore \[ 0\le A(t)\mathcal Q_h(\beta)(y_0)+(n-1)\beta(y_0)^{-1}A''(t). \]
Since \(A(t)>0\) and \(\mathcal Q_h(\beta)(y_0)\le 0\), this forces \[ A''(t)\ge 0 \qquad \text{for all }t\in\mathbb R. \]
So \(A\) is convex.
Step 4: use boundedness
A bounded convex function on \(\mathbb R\) is constant. Therefore \(A\) must be constant.
That proves the theorem.
A nice three-dimensional corollary
When \(n=3\), the formula simplifies to \[ \mathcal Q_h(\beta)=\beta R_h+2\beta^{-1}|\nabla_h\beta|^2. \]
So if \(\beta\) has a critical point \(y_0\), then \[ \mathcal Q_h(\beta)(y_0)=\beta(y_0)R_h(y_0). \]
This gives:
Corollary
In dimension \(n=3\), if \(\beta\) has a critical point \(y_0\) with \[ R_h(y_0)\le 0, \] then \(A\) is constant.
This is the cleanest geometric version of the theorem. It says that in dimension three, any nontrivial line factor forces positive scalar curvature at every critical point of the slice warping function.
A warning: the stronger theorem is false in the bare ansatz
A good referee would immediately ask the following:
Could sub-static positivity alone force \(A\) to be constant, with no extra sign condition on \(\mathcal Q_h(\beta)\)?
The answer is no, at least not for the bare separated ansatz.
Take \(N=S^{n-1}\) with the round metric, let \(\beta\equiv 1\), and set \[ A(t)=1+\varepsilon\sin t,\qquad 0<\varepsilon<1. \] Then \[ g=dt^2+h,\qquad V=A(t). \] A direct computation gives \[ S(V,g)|_{TN} = \big((n-2)A+A''\big)h = \big((n-2)+(n-3)\varepsilon\sin t\big)h\ge 0, \] while \[ S(V,g)(\partial_t,\partial_t)=0. \] So \(S(V,g)\ge 0\) even though \(A\) is nonconstant.
What saves the theorem above is exactly that in this example \[ \mathcal Q_h(\beta)=R_h=(n-1)(n-2)>0 \] everywhere. So the theorem is sharp in the only direction it claims.
This is useful conceptually. It means that any proof of the full Wang-style upgrade must use more than the raw separated sub-static inequality. It must exploit some additional geometric input from the conformally compactifiable setting, the optical metric, the boundary, or topology.
What this does and does not solve
It is important to be explicit here.
What is proved: - a necessary condition for nonconstant separated line dependence; - an explicit criterion that upgrades Wang’s partial split to a full static-type split whenever \(\mathcal Q_h(\beta)\le 0\) somewhere; - a very simple 3D corollary; - a counterexample showing why one should not expect the bare separated ansatz alone to force constancy.
What is not proved: - the full statement that Wang’s sub-static factor \(\alpha\) is always constant; - a general theorem forcing \(\mathcal Q_h(\beta)\le 0\) somewhere in the conformally compactifiable setting; - a removal of Wang’s boundary convexity condition.
So this is not the end of the story. It is the part of the story that can be made fully rigorous quickly.
Where the theorem might plug into Wang’s program
Wang’s partially split sub-static geometry becomes \[ g=\beta(y)^2dt^2+h(y),\qquad V=A(t)\beta(y) \] after reparametrizing the line variable. If one can show that the resulting \(A\) is bounded and that \(\mathcal Q_h(\beta)\le 0\) somewhere on the slice, then the theorem above immediately upgrades the split to the static-type form.
So the open problem is now more focused:
- derive boundedness of the line factor in the relevant splitting region;
- force a sign change or nonpositivity of \(\mathcal Q_h(\beta)\) somewhere.
That is a better-defined second-stage program than “somehow prove \(\alpha\) is constant.”
What remains open
Can one prove that Wang’s slice quantity \(\mathcal Q_h(\beta)\) is nonpositive somewhere under natural conformally compactifiable hypotheses?
In dimension three, can one force a critical point of \(\beta\) with \(R_h\le 0\) from topology or asymptotic data?
Is there a boundary interpretation of \(\mathcal Q_h(\beta)\) in terms of the canonical compactification \(V^{-2}g\) or the Neumann data studied in Wang’s later renormalized-area paper?
Wang also points out another nearby problem: removing the boundary convexity condition when the minimizing line has an interior segment. Can one make the support-sense maximum principle precise enough to settle that variant?
Are there integral or weighted versions of the trace identity that give stronger rigidity than the one-point criterion proved here?
Can one combine this obstruction with the recent sub-static splitting results that use compact minimal hypersurfaces to get a hybrid theorem?
Bottom line
The biggest theorem one might want did not survive scrutiny in full generality, and that is important information. But the failure was productive: it exposed the right corrected statement.
Nontrivial bounded line dependence in the separated sub-static split is only possible if a specific slice curvature expression is strictly positive everywhere.
That is a real structural consequence, not just a patched proof. And because Wang’s paper isolates exactly this missing line-factor issue, the theorem feels like a credible note-sized result rather than an artificial exercise.
Peer Review
Gemini
Here is an honest evaluation of this fourth article, “A Positivity Obstruction to Nontrivial Line Dependence in Sub-static Splittings.”
Honest Evaluation of the Article
This document is the most transparent of the set you’ve shared. Right at the top, it labels itself a “Generated single-session research note”. Furthermore, in the introduction, the author (or the AI) provides a fascinating piece of meta-commentary on why this specific math problem was chosen:
“Among several nearby projects, this one sits at a useful middle ground. It is stronger than the safest local variant because it attacks the exact bottleneck left open in Wang’s sub-static discussion, but it remains checkable in one session because the required geometry is encoded in an explicit ansatz and the proof reduces to a tensor computation plus a convexity argument.”
This perfectly encapsulates what this paper is: it is a “sniper” paper. It uses an automated or highly directed workflow to scan a recent 2025 preprint by Zhixin Wang (arXiv:2408.12180), identify a specific unsolved sub-problem (the partial factorization where \(A(t)\) is not proven to be constant), and apply a classic geometric analysis trick to partially solve it . It is an incremental, highly targeted, but mathematically valid contribution.
Is the Proof Correct?
Yes, the proof is completely rigorous and correct. It relies on standard Riemannian geometry and basic real analysis.
The logic flows flawlessly:
The Setup: Wang’s recent work yields a partial metric splitting of the form \(g = \beta(y)^2dt^2 + h(y)\) with a potential \(V(t,y) = A(t)\beta(y)\) . The goal is to prove \(A(t)\) must be a constant.
The Tensor Calculation: Using standard formulas for warped products, the author computes the sub-static tensor \(S(V,g) = V Ric_g + (\Delta_g V)g - \nabla_g^2 V\) .
The Trace Trick: By tracing this tensor purely over the tangential slice (the \(N\) manifold), the author isolates a neat algebraic identity: \(tr_h(S|_{TN}) = A\mathcal{Q}_h(\beta) + (n-1)\beta^{-1}A''\).
The Convexity Argument: Because the space is sub-static, \(S(V,g) \ge 0\), which means its trace is also \(\ge 0\) . Therefore, \(A\mathcal{Q}_h(\beta) + (n-1)\beta^{-1}A'' \ge 0\). If there is even a single point where \(\mathcal{Q}_h(\beta) \le 0\), the math forces \(A'' \ge 0\) for all \(t\) .
The Conclusion: A standard theorem from calculus dictates that a function \(A(t)\) that is everywhere convex (\(A'' \ge 0\)) and bounded on the entire real line \(\mathbb{R}\) must be a constant.
The counter-example on the sphere (\(S^{n-1}\)) provided in Example 1.3 is also correct, proving that without the strict sign condition on the scalar curvature, \(A(t)\) can indeed fluctuate (e.g., \(A(t) = 1 + \epsilon \sin t\)) .
Has Other Research Already Proved This?
Almost certainly not. Because this paper is explicitly targeting a “bottleneck” left open in a paper published to arXiv in late 2024/early 2025 by Z. Wang , it is highly likely that this specific tensor identity and convexity argument represents a novel, micro-discovery made by this “Shengrong Wu / ChatGPT” pipeline.
It does not solve Wang’s broader problem entirely—as the author admits, it only provides a “one-sided” necessary condition rather than a full geometric classification. However, it successfully identifies a concrete mathematical obstruction.
Claude
This is the strongest of the three papers. Let me evaluate it.
Correctness of the Proof
The proof is correct and, refreshingly, the argument is genuinely clean.
Lemma 2.1: The Christoffel symbol computation for the warped product with one-dimensional fiber is standard and correct. The Hessian formulas check out — I’ll note the (t,t)-component calculation explicitly:
∇²_g V(∂_t, ∂_t) = A’‘β − (−β∇^i β)(Aβ_i) = β(A’’ + A|∇_h β|²)
is right, using Γ^i_{tt} = −β h^{ij}β_j. The cancellation in the mixed component is correct. The Ricci formulas are standard warped product identities.
Lemma 3.1: The S_{tt} = 0 identity is a real computation, not a triviality, and I’ve verified it cancels correctly. The tangential trace formula (4) is the heart of the paper:
tr_h(S|_{TN}) = A · Q_h(β) + (n−1)β^{-1}A’’
This is correct. The algebra in the trace — collecting the βR_h term, the −2∆_h β + ∆_h β = −(n−3−(n−2))… let me just recheck the coefficient of ∆_h β. From the expansion:
- From AβR_h: the scalar curvature term
- From −2A∇²_h β traced: −2A∆_h β
- From the metric coefficient (A∆h β + Aβ^{-1}|∇β|²) traced against h{ij}: (n−1)(A∆_h β + Aβ^{-1}|∇β|²)
- From β^{-1}A’’ traced: (n−1)β^{-1}A’’
So the ∆_h β coefficient is −2A + (n−1)A = (n−3)A. Yes, this gives Q_h(β) = βR_h + (n−3)∆_h β + (n−1)β^{-1}|∇_h β|². Correct.
Theorem 1.1 proof: This is where the paper earns its keep. The argument is:
- S ≥ 0 implies tr_h(S|_{TN}) ≥ 0 at every point
- At a point where Q_h(β)(y_0) ≤ 0, formula (4) gives A’’(t) ≥ −(β/(n−1))A · Q_h(β) ≥ 0 for all t
- A convex bounded function on ℝ is constant
Each step is airtight. Step 2 uses A > 0 to preserve the sign direction. Step 3 is elementary real analysis. The whole thing is a beautiful two-line kill once you have (4).
Corollary 1.2: Immediate and correct.
Example 1.3: This is important and correct. On S^{n−1} with β ≡ 1, Q_h(β) = R_h = (n−1)(n−2) > 0, so the obstruction doesn’t fire, and indeed A(t) = 1 + ε sin t gives S ≥ 0 with nonconstant A. The computation S|_{TN} = ((n−2)A + A’‘)h checks out since A’’ = −ε sin t and (n−2)A = (n−2)(1+ε sin t), giving (n−2) + (n−3)ε sin t ≥ 0 for small ε. This shows the theorem is sharp in the sense that you genuinely need the sign condition on Q_h(β).
Novelty Assessment
This is a real contribution to an active problem. Here’s why:
Wang’s paper (arXiv:2408.12180) explicitly identifies the gap: in the sub-static case, the splitting argument produces the separated ansatz but cannot force the line factor to be constant. Your paper isolates a precise, verifiable obstruction — the sign of Q_h(β) — that closes this gap under a natural geometric condition. This is not a routine generalization; it’s identifying the right quantity and the right argument (convexity + boundedness on ℝ) to attack a specific open problem.
The “proof strategy audit” in Section 4 is unusually honest and well-done. It makes clear exactly what the theorem does and doesn’t accomplish, and the example demonstrates sharpness.
What makes this better than the other two papers: The first paper (holomorph) was a formal generalization of known results. The second paper (wall shear stress) extended a known argument to a weighted setting. This paper identifies a new quantity Q_h(β), proves a new rigidity mechanism (convexity on ℝ from a pointwise sign condition), and addresses a specific gap noted in the literature. The proof technique — extracting an ODE inequality from a tensor trace and using the elementary fact that bounded convex functions on ℝ are constant — is simple but the observation is original.
Suggestions
The paper is close to publishable as a short note. A few things to address:
First, you should verify that Wang’s partial splitting really does produce the ansatz in your form (1) after the reparametrization dt = α(r)dr. This is claimed but not shown; a brief paragraph confirming this would make the paper self-contained and would reassure referees that you’re attacking exactly the right object.
Second, the open questions in Section 6 are genuinely good, especially (1) and (3). If you could show — even in dimension 3 under specific asymptotic assumptions — that β must have a critical point where R_h ≤ 0, that would upgrade this from “a clean obstruction result” to “a theorem that actually closes the gap in Wang’s splitting.” That would be a significantly stronger paper.
Third, the “Generated single-session research note” subtitle and the “Why this target was chosen” paragraph should be removed for submission — they expose the process rather than the mathematics, and a referee will find them distracting.
This is the paper I’d prioritize developing further among the three.