{"id":"5a4401d0-08c5-47ee-bae6-b5337b6c5e49","arxiv_id":"2511.05227","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In globally hyperbolic spacetimes with cost −d^p, weak Kantorovich potentials are locally semiconvex on an open set of full measure, yielding a unique optimal transport map T with ∇φ+∇_x c=0.","lead":"This paper proves that in curved spacetime the price functions used in optimal transport are smooth almost everywhere, giving a unique best way to move mass, and shows with examples that some of the smoothness fails in general. The results supply a missing regularity tool for Lorentzian optimal transport and for synthetic theories of gravity.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.6's central claim is conditional on the existence of a π-solution, which is only established under stronger hypotheses (Thm 1.4); Corollary 1.7's effective scope is therefore narrower than the three-condition statement suggests.","rationale":"The reader's weakest-assumption analysis identifies exactly the same bottleneck: Theorem 1.6 is conditional on the existence of a π-solution, while existence is only proved under the stronger assumptions of Theorem 1.4. I agree with that assessment. The paper's own remarks (Remark 3.8, Remark B.3) explicitly acknowledge the unresolved necessity of connectedness and causal compactness, so this is an acknowledged limitation rather than an oversight.\n\nI did not find an internal inconsistency in Theorem 5.1 or its proof after tracing the main steps. The counterexamples in Section 4 and Example 6.1 are consistent with the theorem's conclusions. The proof of Theorem 1.6 uses Theorem 5.1 and the Lebesgue-point argument in Lemma 5.3; the step where a c-subdifferential point is asserted to lie in I+ is justified by the strict inequality in Theorem 5.1, since a null-related point would have d(x,y)=0 and would violate that strict inequality. No circularity or hidden unsupported assumption beyond the existence hypothesis was found.\n\nThe remaining reservations — Theorem 1.10's reliance on Proposition 7.19 from the author's earlier work and the lack of machine-checked estimates — are secondary to the existence gap. They affect completeness but not the central claim's logical structure. Thus the reader's CONDITIONAL verdict is appropriate; my concern does not move the verdict, but it does sharpen why the condition matters: a concrete counterexample under Theorem 1.6's own hypotheses would substantially shrink the scope of the main theorem, while a positive existence result would remove the bottleneck.","tokens_in":48240,"tokens_out":11521,"duration_ms":102204,"concrete_test":"Test whether connectedness of supp(µ) is needed for existence of a π-solution under Theorem 1.6's hypotheses. Start from the Remark 3.8 example and modify the target measure so that supp(ν) is causally compact (e.g., replace the unbounded graph ν2 by a compact timelike segment or a compact set lying in the causal future of the relevant component). Keep supp(µ) disconnected, µ ≪ vol, supp(µ) ∩ supp(ν) = ∅, and the coupling strictly timelike and c-cyclically monotone. Run the Appendix B Rockafellar construction explicitly for this pair. If φ is forced to be +∞ on one connected component of supp(µ), then no π-solution exists under essentially Theorem 1.6's hypotheses minus connectedness, confirming that the effective domain of Corollary 1.7 requires connectedness. If a finite π-solution is produced, the next check is to prove existence under the full Theorem 1.6 hypotheses without connected","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.6 is the paper's main regularity result, but its hypothesis is: 'Suppose π ∈ Γ^≤(µ,ν) admits a π-solution φ.' The theorem does not construct such a φ. Existence is supplied only by Theorem 1.4(i), which requires both supports to be causally compact, supp(µ) to be connected, and µ,ν to be strictly timelike — none of which appears in Theorem 1.6's three stated conditions. Corollary 1.7 explicitly combines 'assumptions (a)–(c) from both Theorems,' so its effective domain is the intersection of the two theorem hypotheses plus compact support, substantially narrower than the abstract's 'suitable general assumptions' suggests.\n\nThe paper is honest about this: Remark 3.8 and Appendix B, Remark B.3 state that existence of a π-solution can fail when connectedness and causal compactness are dropped simultaneously, and the author writes it is 'unclear (at least to me) whether each of these assumptions is strictly necessary.' This is not an internal inconsistency, but it is the load-bearing limitation of the central claim: if connectedness of supp(µ) is genuinely necessary for existence, then Theorem 1.6 has no object to apply to for a large class of admissible measures with disconnected support; if it is not necessary, there is a missing theorem that should establish existence under Theorem 1.6's own hypotheses. Since the main advertised applications — uniqueness and the map representation in Corollary 1.7 — depend on Theorem 1.6, this conditional status is the key bottleneck.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the Lorentzian optimal transport problem on globally hyperbolic spacetimes with cost c(x,y)=-d(x,y)^p, p in (0,1). It has three advertised contributions. First (Theorem 1.4, Corollary 1.5), under causal compactness of the supports, connectedness of supp(mu), and strict timelikeness, it establishes existence of a measurable pi-solution, strong Kantorovich duality for compactly supported measures, and equivalence between optimality and c-cyclical monotonicity under assumption (W). Second (Theorem 1.6, Corollary 1.7), assuming existence of a pi-solution, it proves that the coupling is concentrated on I^+ and that the pi-solution is locally semiconvex on an open set of full mu-measure, yielding uniqueness and a map representation via an actual gradient equation. Third (Theorem 1.10), it proves a weak-KAM-type result producing C^{1,1}_{loc} calibrated pairs between intermediate measures of a displacement interpolation. The proofs combine the Kell--Suhr construction with classical optimal transport techniques and include counterexamples showing that Riemannian regularity results fail in general.","tokens_in":48553,"tokens_out":14444,"duration_ms":122568,"significance":"If the technical estimates in Section 5 are correct, Theorem 1.6 is a genuine advance: it is the first general local semiconvexity statement for weak Kantorovich potentials in Lorentzian optimal transport, and it permits an actual-gradient characterization of the optimal coupling. The paper is also valuable for its explicit counterexamples and for its honest discussion of the limitations of the Kell--Suhr existence result. The proof of Theorem 5.1 is a serious and structured technical effort. However, the advertised scope is substantially narrower than the abstract suggests. Theorem 1.6 is conditional on the existence of a pi-solution, which is established only under stronger hypotheses, and Theorem 1.10 depends on a substantial proposition that is not proved in the manuscript but only referenced to the author's own arXiv preprint. These two issues are load-bearing for the main applications and should be addressed before publication.","major_comments":[{"comment":"The central regularity theorem assumes, rather than constructs, a pi-solution. Existence is proved only under the stronger hypotheses of Theorem 1.4, namely causal compactness of both supports, connectedness of supp(mu), and strict timelikeness of mu and nu. The paper itself notes in Remark 3.8 and Remark B.3 that existence can fail when connectedness and causal compactness are dropped simultaneously, and that it is unknown whether each condition is individually necessary. Consequently, Corollary 1.7, which explicitly combines the assumptions of both theorems, has an effective domain equal to the intersection of the two theorem hypotheses, not the three conditions displayed in Theorem 1.6. This is not an internal inconsistency, but the advertised 'suitable general assumptions' and the abstract's claim of general applicability are overstated. The introduction, abstract, and theorem statem","section":"Theorem 1.6 and Corollary 1.7; Remark 3.8; Appendix B"},{"comment":"Theorem 7.18 and hence the main third result, Theorem 1.10, rely on Proposition 7.19. The manuscript states 'The following proposition is proved in [22]' and gives no proof, while [22] is the author's own arXiv preprint. Proposition 7.19 supplies the uniform semiconcave local representation of T u and the family of smooth functions f_{i,t} with the estimates (i)--(vii) used in Theorem 7.26 and in the proof of Theorem 7.18. This is a load-bearing external dependency for one of the paper's main advertised results. A refereed paper should either include a complete proof of Proposition 7.19 (or a self-contained version adequate for the present setting) or clearly state that Theorem 1.10 is conditional on the publication of [22].","section":"Section 7, Proposition 7.19"}],"minor_comments":[{"comment":"Typo: 'semconvex' should be 'semiconvex'.","section":"Abstract"},{"comment":"The phrase 'some numbers 0 >0' appears several times; it should be 'some number s_0 > 0' or 'some numbers s_0 > 0'.","section":"Theorem 7.12, Lemma 7.14, Corollary 7.17"},{"comment":"Typo: 'Picky 0' should be 'Pick y_0'. Also, the phrase 'Necessarily, y in I^+(x_0)' would benefit from a one-sentence explanation: it follows from the strict inequality in Theorem 5.1, because otherwise the point would belong to the set of sup over d(x,y) <= delta.","section":"Lemma 5.4"},{"comment":"The set B is defined by existential quantifiers over sequences and is not obviously Borel. The proof subsequently takes its closure and uses Lebesgue points. Please clarify the notion of 'null set' being used and justify the measurability/outer-measure step, or rewrite the argument directly for the closure of B_i.","section":"Section 5.1, beginning of proof of Theorem 5.1"}],"recommendation":"major_revision","confidential_remarks":"The paper contains substantial original work and the main conditional regularity theorem is plausible, but the two load-bearing concerns above — the conditional existence bottleneck and the reliance on an unpublished self-cited proposition — should be addressed before acceptance. The manuscript would also benefit from a more conservative abstract and introduction."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Main takeaway: Theorem 1.6 is a genuine, useful regularity result for weak Kantorovich potentials with the intrinsic Lorentzian cost, and the paper is honest about the fact that it only applies when a π-solution is already known to exist. That conditional status is the main caveat.\n\nWhat is new: semiconvexity on a full-measure open set for c-convex functions/π-solutions with c = -d^p, and the corollary giving an actual gradient equation and uniqueness of optimal map. The counterexamples in §4 and §6 do real work: they show Riemannian-style regularity fails in general, and that even under the hypotheses of Corollary 1.7 the optimal coupling need not be supported on I^+. The proof of Theorem 1.6 is long but structured; the local semiconcavity/semiconvexity machinery in §5 and §7 is coherent. No circular reasoning: the central claim uses Kell–Suhr and standard semiconcavity, not the conclusion. And the paper explicitly records its limitations (Remark 3.8, Remark B.3).\n\nSoft spots, in order of importance. First, Theorem 1.6 assumes a π-solution exists rather than constructing one. Existence is proved only under stronger hypotheses: causal compactness of both supports, connectedness of supp(µ), strict timelikeness. The author says it is unclear whether these are necessary. So the advertised Corollary 1.7 has effective scope equal to the intersection of the two theorem hypotheses, plus compact support, which is narrower than the three-condition statement of Theorem 1.6 might suggest. This is not an internal contradiction, but it is the load-bearing limitation: if connectedness is genuinely needed for existence, the theorem has nothing to say about disconnected-source cases. Second, Theorem 1.10 depends on Proposition 7.19, which comes from the author's earlier paper [22]. That is legitimate, but it makes a key part of that result not self-contained. Third, the long analytic estimates (Lemma 5.11, Step 4 of Theorem 5.1) are not machine-checked; they look plausible and the steps are spelled out, but this is the kind of argument where a sign error would be easy to miss.\n\nWho this is for: people working in Lorentzian optimal transport, weak KAM theory on spacetimes, and synthetic timelike Ricci bounds. A serious referee can add value here; the main theorem deserves verification despite the conditional scope. I would engage with it and cite it, and I'd flag the existence gap clearly to the authors so they can either prove existence under the hypotheses of Theorem 1.6 or state the effective domain honestly in the abstract.\n\nRecommendation: send to peer review. It is a serious toolbox paper, not a desk reject.","headline":"A genuinely new regularity result for Lorentzian Kantorovich potentials, but it is explicitly conditional on the existence of a π-solution, so the effective scope is narrower than the headline suggests.","tokens_in":49139,"tokens_out":2214,"would_cite":true,"duration_ms":19914,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["49N60","49J30","49Q22","49Q20","53C50"],"pacs":[],"model":"deepseek-v4-flash","headline":"Weak Kantorovich potentials in Lorentzian optimal transport are locally semiconvex on a full-measure open set, yielding a unique optimal transport map.","keywords":["Lorentzian optimal transport","Kantorovich potentials","semiconvexity","c-convex functions","strong duality","optimal transport map","causal couplings","time-separation cost"],"falsifier":"Find probability measures μ, ν satisfying supp(ν) causally compact, μ ≪ vol, and supp(μ) ∩ supp(ν) = ∅, together with a causal coupling π that admits a π-solution, but for which the solution is not locally semiconvex on any open set of full μ-measure. Alternatively, exhibit a pair satisfying the stronger hypotheses of Theorem 1.4 for which no measurable π-solution exists, which would block the strong-duality and map conclusions.","tokens_in":48041,"feed_emoji":"🕰️","tokens_out":3773,"duration_ms":35559,"temperature":0.7,"pith_summary":"The paper establishes that, for Lorentzian optimal transport with the standard proper-time cost, weak Kantorovich potentials (π-solutions) are locally semiconvex on an open set of full measure, provided the target measure's support is causally compact, the source is absolutely continuous, and the two supports are disjoint. This is a general regularity result in a setting where the cost function is not real-valued and classical theory does not directly apply. Semiconvexity upgrades approximate differentiability to actual differentiability almost everywhere, which yields existence and uniqueness of an optimal transport map characterized by a first-order gradient equation. The paper also closes a known gap by proving strong Kantorovich duality and c-cyclical monotonicity characterizations under connectedness, causal compactness, and strict timelikeness assumptions.","feed_headline":"Lorentzian potentials get semiconvex on full measure","feed_subtitle":"A regularity theorem yields a unique optimal map and strong duality for causal couplings under mild measure conditions.","key_machinery":"The central object is the weak Kantorovich potential, or π-solution: a c-convex function φ whose c-subdifferential ∂ᶜφ has full π-measure. The proof rests on a general regularity theorem for c-convex functions: for volume-a.e. point in the interior of the domain away from the target support, small causal competitors in the defining supremum are strictly suboptimal. This local strictness, together with causal compactness of the target support to replace the supremum by one over a compact set and with the known local semiconcavity of the cost on I⁺, yields local semiconvexity of φ.","core_discovery":"The central theorem states: let μ and ν be Borel probability measures on a globally hyperbolic spacetime, with supp(ν) causally compact, μ absolutely continuous with respect to the volume measure, and supp(μ) ∩ supp(ν) = ∅. If a causal coupling π admits a π-solution φ (a c-convex function whose c-subdifferential carries full π-measure), then π is concentrated on the set I⁺ of chronologically related pairs and φ is locally semiconvex on an open set of full μ-measure. Consequently, together with the paper's duality results, there is a unique optimal coupling, concentrated on I⁺, induced by the map T(x) = (∇_x c(x,·))⁻¹(−∇φ(x)), and characterized by the equation ∇φ(x) + ∇_x c(x, y) = 0 π-a.e.","pith_inferences":["The semiconvexity conclusion is conditional on the existence of a π-solution; the paper constructs such solutions only under substantially stronger hypotheses, so the effective reach of Corollary 1.7 is narrower than the statement of Theorem 1.6 suggests.","Because semiconvexity holds on an open set of full μ-measure, the singular set of the potential is μ-negligible; a natural test is whether the optimal map T is continuous and whether the semiconvexity constant can be chosen locally uniformly.","The example with optimal coupling touching the light cone indicates that conditions on the target measure beyond compactness are needed to guarantee strict timelike transport, and it suggests exploring how the gradient map behaves at boundary points of supp(π).","The weak KAM-style argument may extend to produce C¹′¹ calibrations for all intermediate times under weaker causal-compactness assumptions than those stated."],"forward_implications":["Under the combined hypotheses (compact supports, connected and causally compact supports, strict timelikeness), strong Kantorovich duality holds: a maximizing pair exists in the dual problem.","Optimality of a causal coupling is characterized by c-cyclical monotonicity whenever a measurable π-solution exists.","The unique optimal coupling is concentrated on I⁺ and is induced by the gradient map satisfying ∇φ(x) + ∇ₓc(x, y) = 0 π-a.e.","Along a displacement interpolation, the intermediate measures admit C¹′¹_loc-regular calibrated dual pairs.","The hypotheses do not force the optimal coupling to be supported away from the light cone; an example shows that support on the boundary of I⁺ can occur."],"fun_headline_variants":["Semiconvex potentials in Lorentzian transport on full measure","Unique causal optimal map from semiconvexity","Full-measure semiconvexity for Kantorovich potentials","Lorentzian OT: semiconvex potentials yield unique map"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The whole regularity conclusion depends on assuming that a π-solution already exists; the paper constructs such solutions only under substantially stronger hypotheses, so if existence fails for some admissible pair, the theorem's conclusion has no object to apply to.","fun_headline_variants_meta":{"raw":{"variants":["Semiconvex potentials in Lorentzian transport on full measure","Unique causal optimal map from semiconvexity","Full-measure semiconvexity for Kantorovich potentials","Lorentzian OT: semiconvex potentials yield unique map"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000639,"raw_usage":{"total_tokens":2740,"prompt_tokens":668,"completion_tokens":2072,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":412,"completion_tokens_details":{"reasoning_tokens":2004}},"tokens_in":412,"tokens_out":2072,"duration_ms":15657,"temperature":1.0,"reasoning_tokens":2004,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T23:30:22.649748+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find probability measures μ, ν satisfying supp(ν) causally compact, μ ≪ vol, and supp(μ) ∩ supp(ν) = ∅, together with a causal coupling π that admits a π-solution, but for which the solution is not locally semiconvex on any open set of full μ-measure. Alternatively, exhibit a pair satisfying the stronger hypotheses of Theorem 1.4 for which no measurable π-solution exists, which would block the strong-duality and map conclusions.","supporting_citations":[],"review_version":1}