{"id":"d1dc6290-430c-453a-ae59-5866ad12fba5","arxiv_id":"2412.00255","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Each unit costable cohomology class on a closed oriented Riemannian d-manifold (2 to 7) determines a unique largest lamination of minimal hypersurfaces calibrated by every calibration in the class, constraining the stable norm ball.","lead":"The paper constructs, on any closed manifold of dimension 2 to 7, a canonical lamination (a non-crossing stack of minimal hypersurfaces) attached to each unit cohomology class, whose leaves are exactly the hypersurfaces calibrated by every calibration in that class. The object constrains the shape of the stable norm ball by the topology of the manifold and sharpens an analogy with Thurston's earthquake norm.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.7 imports [Bac24, Theorem B] to the noncompact universal cover without verifying that the cited theorem applies there; this is the load-bearing bridge for Theorem 1.2.","rationale":"The reader identified the dependence on [Bac24, Theorem B] as the weakest assumption, and I agree that this imported theorem is the load-bearing bridge for Theorem 1.2. My concern is more specific: the theorem is applied on the noncompact universal cover, whereas the cited source may only establish the least-gradient-to-lamination correspondence for closed manifolds. If the hypotheses do not cover this case, the nontriviality lemma (Lemma 3.9) fails and the whole construction in Section 4 has no starting point. This is not a charge of circularity—[Bac24] is published and peer-reviewed, which is legitimate support—but the transfer to ~M is unverified in the text. I did not choose the other weaknesses (Lemma 5.6's coarea estimate, the max(0,b_1−1) vs max(1,b_1−1) discrepancy, Theorem 1.7's missing proof) because they are local to the applications in Section 5 and do not threaten Theorem 1.2. The reader's CONDITIONAL verdict remains appropriate: the central architecture is plausible, but the hypothesis check for Theorem 3.7 on the universal cover should be settled before accepting the main claim.","tokens_in":28321,"tokens_out":19755,"duration_ms":192722,"concrete_test":"Check the precise hypotheses of [Bac24, Theorem B]. If it is only stated for closed manifolds, supply and record the natural reduction: take a torsion-free finite-index subgroup Γ' of π_1(M) so that ~M/Γ' is closed, note that u is Γ'-equivariant, apply the theorem on that compact quotient, and descend the resulting lamination to M. If this reduction is valid and written down, the gap closes; if no such argument exists, Lemma 3.9 is unsupported and the main theorem lacks its nontriviality input.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 1.2 depends on Lemma 3.9, which selects an α-equivariant least-gradient function u on the noncompact universal cover ~M and invokes Theorem 3.7 (a restatement of [Bac24, Theorem B]) to convert level sets of u into a homologically minimizing lamination with Ruelle–Sullivan current du. If [Bac24, Theorem B] is proved only for closed manifolds, its hypotheses are not satisfied by ~M, and the paper gives no reduction by deck-group invariance or by passing to a closed quotient. Without this step there is no nontrivial F-calibrated lamination, so hypothesis (1) in Section 4 fails and Theorem 1.2 collapses. The paper also does not verify that the exact form of Theorem 3.7(c)—that leaves are level sets of u—matches the statement in [Bac24].","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs, for every unit-costable cohomology class ρ in H^{d−1}(M,R) on a closed oriented Riemannian d-manifold with 2≤d≤7, a canonical lamination λ_ρ whose leaves are exactly the complete immersed hypersurfaces calibrated by every calibration representing ρ. The construction first builds, for each calibration F in ρ, a lamination λ_F of F-calibrated minimal hypersurfaces, using α-equivariant functions of least gradient on the universal cover and a theorem (Theorem 3.7) that turns their level sets into homologically minimizing laminations. The canonical lamination is then the largest lamination contained in every λ_F. The paper derives structural consequences for the stable unit ball: the dual flat ρ* is exactly the set of homology classes of transverse probability measures on sublaminations of λ_ρ (Corollary 1.3); maximal flats are polytopes whose rational vertices correspond to closed leaves (Theorem 1.4); strict convexity of the stable ball follows from a condition on the derived series of π_1(M), and a line segment in the boundary of the stable ball forces vanishing intersection product (Theorem 1.6). The final section develops an analogy with the earthquake norm on Teichmüller space.","tokens_in":28337,"tokens_out":20826,"duration_ms":197878,"significance":"If the main theorem is correct, the paper provides a canonical lamination associated to every costable unit class, generalizing the Thurston–Guéritaud–Kassel maximally stretched lamination to calibrated hypersurfaces in higher dimensions. The structural corollaries—the polytope description of flats, the closed-leaf characterization of rational vertices, and the strict-convexity criteria—are concrete, falsifiable statements about the stable norm ball and go substantially beyond what was previously proved. The construction is parameter-free and I do not see a circularity: λ_ρ is built from calibrations and least-gradient functions, and the later theorems are derived from the structure of λ_ρ. However, the proof depends on a load-bearing import ([Bac24, Theorem B] via Theorem 3.7) and on a questionable level-set identity in the proof of Theorem 1.6(2), so the paper is not yet in final form.","major_comments":[{"comment":"Theorem 3.7 is the bridge from least-gradient functions to calibrated laminations: Lemma 3.9 uses it to produce the nontriviality input for the construction of λ_ρ in §4. The proof applies [Bac24, Theorem B] to the noncompact universal cover \\tilde M, but the manuscript does not state the hypotheses of that theorem or verify that \\tilde M satisfies them. If [Bac24, Theorem B] is proved only for closed manifolds, its hypotheses are not satisfied by \\tilde M, and no reduction by deck-group invariance or by passing to a closed quotient is given. Since the nontriviality condition (1) in §4 and therefore Theorem 1.2 collapse without this step, please quote the exact theorem used, show that it applies to \\tilde M, or supply the missing argument.","section":"§3.2 (Theorem 3.7)"},{"comment":"The proof asserts that for u := (u_α+u_β)/2 one has ∂{u>0} = ∂{u_α>0} ∪ ∂{u_β>0}. This equality is not a consequence of the definitions and is generally false: the support of the average Ruelle–Sullivan current is the union of the two supports, but the topological boundary of a superlevel set of the averaged function need not have both leaves as boundary components. Since the disconnectedness of ∂{u>0} is then fed into Lemma 5.10, the proof of Theorem 1.6(2) is incomplete as written. A corrected argument is needed, for example by choosing the constants or orientations so that the two leaves are boundary components of a single level set, or by working directly with the jump set rather than the superlevel boundary.","section":"§5.5 (proof of Theorem 1.6(2))"},{"comment":"The step 'The infinite sequence (α_n) is linearly dependent, so M \\setminus ⋃N_n must be disconnected; therefore there can be no leaf of λ_ρ which is dense in M' is too compressed. Linear dependence of the homology classes of infinitely many closed leaves does not by itself imply that their union disconnects M, and this implication is used before applying the Morgan–Shelan decomposition. Since Theorem 1.4 is a central structural claim, this topological step should be justified or replaced by a direct argument.","section":"§5.3 (proof of Theorem 1.4)"}],"minor_comments":[{"comment":"There are duplicated words ('the the geometry' in the abstract, 'a an affine map' in §5.1) and the title contains an errant space ('LAMINA TION'); these should be corrected.","section":"Abstract and §1"},{"comment":"For self-containedness, the exact statement of [Bac24, Theorem B] that is being invoked should be quoted, especially because Theorem 3.7 is presented as an equivalence but the proof refers to the cited theorem for both directions.","section":"§3.2 (Theorem 3.7)"},{"comment":"In the proof of Lemma 5.6, the sentence 'The image of T in S^1 is a point' is not explained; the intended reason is presumably that the sheets in the ball are lifts of the same leaf and the equivariant function changes by integers under deck transformations, but this should be stated explicitly.","section":"§5.2 (Lemma 5.6)"},{"comment":"The proof writes m ≤ max(1,b_1−1) while the theorem states max(0,b_1−1); the discrepancy should be resolved and the bookkeeping for b_1=0 clarified.","section":"§5.3 (Theorem 1.4 proof)"},{"comment":"Lemmas 5.9 and 5.10 are attributed to the unpublished manuscript [AB12]; the paper reproduces them with credit, but it would help the reader if the manuscript stated which parts of those lemmas are being reused verbatim and which are modified.","section":"§5.5"}],"recommendation":"major_revision","confidential_remarks":"The paper is ambitious and, if the identified gaps are repaired, likely a strong contribution to calibrated geometry and the theory of the stable norm. The main concern is not novelty or circularity but verifiability: Theorem 3.7 rests on a self-citation whose hypotheses are not stated, and the proof of Theorem 1.6(2) contains a level-set equality that looks wrong as written. I would encourage the editor to send the paper back for a careful revision addressing those two points before further review."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know upfront. First, if Theorem 1.2 is correct, this paper closes the Auer–Bangert program and delivers a genuinely useful object: for every unit costable class ρ there is a largest lamination λ_ρ calibrated by every calibration in the class, and the dual flat ρ* is exactly the set of homology classes of transverse measures on sublaminations of λ_ρ. Second, the proof's bridge from least-gradient functions to laminations—Theorem 3.7—is imported from the author's own [Bac24] and applied to the noncompact universal cover without stating or verifying that the cited theorem's hypotheses hold there. That is the load-bearing step, and it is what a referee should check first.\n\nThe construction in Section 4 is real work. The curvature bound via Schoen–Simon, the disjointness argument via the maximum principle, and the compactness argument via Vietoris limits are standard tools but assembled in a clean, readable way. Corollary 1.3 and Theorem 1.4 are plausible and mostly well derived. The earthquake-norm analogy is illuminating and not just decoration; it clearly shaped the statements.\n\nThe soft spots are local but real. Lemma 5.6's coarea estimate is not justified as written: the volume lower bound is established on a discrete set T and then integrated over full intervals. A referee should ask for a genuine coarea argument or a different proof. Theorem 1.4 states max(0,b_1−1) but the proof gives max(1,b_1−1); the discrepancy only matters when b_1=0, but it is still a mismatch. Lemma 5.4's second clause is mis-stated: as written it says the homology classes are linearly independent and if they span H_{d−1}(M,R) then H_{d−1}(M,R)=0, which cannot be right. Theorem 1.7 is a promissory note to [DU25], and the AB12 lemmas in §5.5 include a 'straightforward generalization' of [BGG69] that is not spelled out. None of these by itself breaks the central claim, but together they make the paper feel less finished than the main theorem deserves.\n\nOn the stress-test note: I think the concern lands. The paper does not show that [Bac24, Theorem B] applies on the universal cover, and it does not reconcile the exact form of Theorem 3.7(c) with the cited result. This is not circular—[Bac24] is published and peer-reviewed—so it is a legitimate gap, not a fatal flaw. It may be that a simple equivariant localization argument fills it, but the text needs to say so.\n\nThis paper is for people working on stable norms, calibrated geometry, and minimal laminations. It deserves a serious referee. I would send it to review, with a request to verify the [Bac24] bridge first and then clean up the local issues. If the bridge holds, it is a significant paper; if not, it needs substantial revision. Either way, worth the referee time.","headline":"Ambitious, mostly coherent construction of the canonical calibrated lamination; the main thing to referee is the unverified use of [Bac24] on the noncompact universal cover.","tokens_in":29075,"tokens_out":6593,"would_cite":true,"duration_ms":56303,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["49Q05","53C38","37F34"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every unit cohomology class yields a canonical lamination","keywords":["laminations","minimal hypersurfaces","calibrations","functions of least gradient","stable norm","costable norm","Ruelle-Sullivan current","earthquake norm"],"falsifier":"Take a flat $3$-torus and a unit costable class $\\rho$; the theorem predicts the hypersurfaces calibrated by every calibration in $\\rho$ are exactly the leaves of one foliation by parallel planes. If one can exhibit two non-parallel minimal hypersurfaces each calibrated by every calibration representing $\\rho$, the maximality or uniqueness of the canonical lamination fails.","tokens_in":27868,"feed_emoji":"📐","tokens_out":10787,"duration_ms":89909,"temperature":0.7,"pith_summary":"This paper proves that on any closed oriented Riemannian manifold of dimension $2$ through $7$, each cohomology class of degree $d-1$ with unit stable dual norm determines a unique largest lamination, a closed union of disjoint minimal hypersurfaces, whose leaves are exactly the hypersurfaces calibrated by every calibration representing the class. This gives a canonical geometric object attached to a purely cohomological datum, analogous to the maximally stretched lamination attached to a best Lipschitz map between hyperbolic surface metrics. The paper then reads geometric information about the stable norm ball, the unit ball of the area-minimizing homology norm, off this lamination: transverse measures on its sublaminations parameterize the dual face of the ball, vertices of maximal flats correspond to closed leaves, and a line segment on the stable unit sphere forces the vanishing of an intersection product. A strict-convexity criterion follows from a statement about the derived series of the fundamental group. The point of caring is that the shape of the stable norm ball is constrained by the topology of $M$ in a way that mirrors the earthquake norm on the tangent space to the space of hyperbolic surface metrics.","feed_headline":"Every unit cohomology class yields a canonical lamination","feed_subtitle":"Its leaves are exactly the minimal hypersurfaces calibrated by every calibration in the class.","key_machinery":"The load-bearing mechanism is the correspondence between functions of least gradient and homologically minimizing laminations. A function $u$ on the universal cover that is equivariant under deck transformations with a homomorphism $\\alpha:\\pi_1(M)\\to\\mathbb{R}$ and minimizes total variation has level sets that descend to a measured lamination $\\lambda_u$ of minimal hypersurfaces whose Ruelle-Sullivan current equals $du$; this is imported as Theorem 3.7 and holds for $d \\leq 7$. Lemma 3.9 selects any $\\alpha$ in the dual face $\\rho^*$, builds an $\\alpha$-equivariant least-gradient function, and converts it into a lamination calibrated by every calibration representing $\\rho$, giving the nontriviality input. The technical glue is a package of $L^\\infty$ calibration results: the normal trace theorem extends integration of closed $L^\\infty$ $(d-1)$-forms to Lipschitz hypersurfaces, the Anzellotti wedge product bounds the pairing between $du$ and a calibration, and the $L^\\infty$ Poincar\\'e lemma gives continuous potentials so that Stokes' theorem applies; together these let the paper treat calibrations that are only $L^\\infty$, not continuous. Curvature bounds for stable minimal hypersurfaces give the compactness that upgrades the set of calibrated hypersurfaces to a Lipschitz lamination.","core_discovery":"The central claim is Theorem 1.2: for every $\rho \\in H^{d-1}(M,\\mathbb{R})$ with $\\|\\rho\\|_\\infty = 1$, there exists a unique largest lamination $\\lambda_\\rho$ in $M$ such that every calibration $F$ representing $\\rho$ calibrates every leaf of $\\lambda_\\rho$. Here a calibration is a closed $(d-1)$-form of $L^\\infty$ comass $1$, and a hypersurface is calibrated if the form restricts to its area form. The lamination is largest in the sense that its leaf set is the intersection of the leaf sets of the individual calibrations' laminations $\\lambda_F$, and any hypersurface calibrated by every calibration in the class must be a leaf. The proof constructs for each calibration $F$ the lamination $\\lambda_F$ whose leaves are all complete connected $F$-calibrated hypersurfaces, proves these leaves have uniform curvature bounds and are pairwise disjoint, and then takes the intersection over all $F$ in the class. The non-emptiness and calibration-by-every-$F$ property come from a least-gradient function: for any homology class $\\alpha$ in the dual face $\\rho^*$, an $\\alpha$-equivariant function of least gradient produces a measured lamination calibrated by every calibration in $\\rho$. As corollaries, the dual face $\\rho^*$ is exactly the set of homology classes carried by transverse probability measures on sublaminations of $\\lambda_\\rho$, every extreme point of $\\rho^*$ corresponds to an ergodic measure, rational-direction vertices of maximal flats correspond to closed leaves, and if a maximal flat has too many vertices then the lamination has a spiraling part with no transverse measure.","pith_inferences":["A testable extension is to compute $\\lambda_\\rho$ numerically in low-dimensional manifolds such as flat tori or hyperbolic $3$-manifolds and check that the extreme points of $\\rho^*$ are realized by distinct ergodic sublaminations; this would provide explicit examples of the stable unit ball's non-strict convexity.","The $d \\leq 7$ restriction likely tracks the regularity theory for least-gradient level sets rather than any essential feature of calibrations, so a natural project is to extend the construction to $d \\geq 8$ for classes whose calibrations are continuous or have special structure, using the paper's $L^\\infty$ machinery.","The paper's analogy with the earthquake norm suggests that the canonical lamination $\\lambda_\\rho$ should be characterized as the maximal set on which every optimal Lipschitz representative of $\\rho$ is infinitesimally stretched; in dimension $2$ this is exactly what happens, and proving the analogue in higher dimensions would unify the two theories.","Because the proof identifies $\\rho^*$ with the set of transverse measures on sublaminations, the non-strict convexity of the stable norm can be probed by searching for two distinct ergodic calibrated sublaminations of $\\lambda_\\rho$; this gives a geometric, rather than algebraic, way to detect flats in the stable unit ball."],"forward_implications":["Every unit costable cohomology class carves out a canonical lamination, so the dual face $\\rho^*$ of the stable unit ball is geometrically realized: its points are exactly the homology classes of transverse probability measures on sublaminations of $\\lambda_\\rho$.","Vertices of maximal flats of the stable unit sphere coincide with closed leaves of $\\lambda_\\rho$ exactly when they have rational direction, and irrational vertices are limited by $b_1(M)-1$, so the combinatorial shape of stable flats is constrained by the topology of $M$.","If the stable unit ball is strictly convex, every ergodic calibrated lamination is uniquely ergodic; on manifolds with $b_1(M) \\geq 2$, all but countably many boundary homology classes are represented by uniquely ergodic calibrated laminations without closed leaves.","A line segment contained in the stable unit sphere forces the intersection product of its endpoints to vanish, and if the quotient of the first two terms of the derived series of $\\pi_1(M)$ is a torsion group, the stable unit ball is strictly convex, so every metric on a torus has a strictly convex stable unit ball and many uniquely ergodic laminations of minimal hypersurfaces.","In dimension $2$, the construction recovers a canonical maximally stretched lamination for homotopy classes of maps to the circle, placing the minimal-hypersurface result in the same family as the stretch lamination for best Lipschitz maps between hyperbolic surface metrics."],"supporting_citations":[{"why":"Supplies both directions of the least-gradient-to-lamination correspondence used as Theorem 3.7, as well as the compactness lemma for limits of minimal hypersurfaces.","marker":"[Bac24]"},{"why":"Establishes the duality that the costable norm equals the minimal $L^\\infty$ comass in a cohomology class, so unit classes admit calibrations.","marker":"[Fed74]"},{"why":"Gives the normal trace theorem and wedge-product estimate that let closed $L^\\infty$ forms be integrated along Lipschitz hypersurfaces.","marker":"[Anz83]"},{"why":"Provides the curvature bounds for stable minimal hypersurfaces used to obtain uniform second fundamental form estimates.","marker":"[SS81]"},{"why":"Supplies the laminar flow-box formalism and the decomposition of measured laminations used in the structure theorem for transverse measures.","marker":"[MS88]"},{"why":"Contributes the ergodic-theoretic lemma bounding the number of ergodic measures on sublaminations without closed leaves.","marker":"[AL86]"},{"why":"Supplies the two lemmas on connectedness of superlevel sets and boundaries of least-gradient functions used to prove strict convexity of the stable unit ball.","marker":"[AB12]"}],"fun_headline_variants":["Unit cohomology class yields canonical lamination","Unique largest lamination calibrated by every calibration","Cohomology class pins down a canonical lamination","Cohomology class determines a canonical lamination","Intersection of calibrations yields a unique lamination"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything rests on an imported theorem that every nonzero equivariant function of least gradient has level sets forming a homologically minimizing lamination of minimal hypersurfaces, which is only proved in dimensions at most $7$; if that correspondence fails for a class admitting unit calibrations, the canonical lamination need not exist.","fun_headline_variants_meta":{"raw":{"variants":["Unit cohomology class yields canonical lamination","Unique largest lamination calibrated by every calibration","Cohomology class pins down a canonical lamination","Cohomology class determines a canonical lamination","Intersection of calibrations yields a unique lamination"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002271,"raw_usage":{"total_tokens":8820,"prompt_tokens":1043,"completion_tokens":7777,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":659,"completion_tokens_details":{"reasoning_tokens":7703}},"tokens_in":659,"tokens_out":7777,"duration_ms":45546,"temperature":1.0,"reasoning_tokens":7703,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:35:56.702071+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a flat $3$-torus and a unit costable class $\\rho$; the theorem predicts the hypersurfaces calibrated by every calibration in $\\rho$ are exactly the leaves of one foliation by parallel planes. If one can exhibit two non-parallel minimal hypersurfaces each calibrated by every calibration representing $\\rho$, the maximality or uniqueness of the canonical lamination fails.","supporting_citations":[],"review_version":1}