{"id":"92389019-feed-44ad-a329-a002b067a514","arxiv_id":"2412.12078","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every coherent logarithmic scheme decomposes into a jointly surjective collection of fine logarithmic faces, computable by monoidal Gröbner basis algorithms.","lead":"A new mathematical construction breaks any 'coherent logarithmic scheme' into a collection of 'fine' pieces, so that tools from tropical geometry can be applied to schemes where they previously did not work. The authors also give algorithms to compute these pieces, with applications to enumerative geometry and moduli spaces.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem A depends on the unproved claim that the fine faces in Construction 3.11 are independent of the choice of global chart and descend from an étale cover; without this, the central decomposition is not well-defined.","rationale":"I identified the same load-bearing concern as the reader: Theorem A's canonicality rests on the chart-independence and descent assertions in Construction 3.11, neither of which is proved. This is more central than the other issues noted in the reader's verdict. The notation clash in §5.2 (where ∨ is defined as coordinatewise minimum but Algorithm 5.14 uses what must be the lcm) is real but affects only the algorithmic content of Theorem B, not the existence theorem. Theorem C is stated without proof but is an application rather than the core claim. The gap in Construction 3.11 is closer to the heart of the paper. The example of the log point with characteristic N charted by N and N^2 shows that the chart-independence claim is non-vacuous: the same face can arise from different prime ideals in different charts, so a proof of invariance is required. This does not mean the paper is wrong—the argument may be completable—but as written, the proof of Theorem A is incomplete. Hence the reader's CONDITIONAL verdict is appropriate and I would not change it.","tokens_in":22933,"tokens_out":37814,"duration_ms":335340,"concrete_test":"Run the chart-independence check for the log point X = Spec k with characteristic monoid N, using the two global charts: (a) Q=N, generator s mapping to 0∈k; (b) Q=N^2, generators x↦0, y↦1. For each chart, enumerate all prime ideals p of Q, form Z_p := X ×_{Spec Z[Q]} Spec(Z[Q]/(p)), take the nonempty connected components, and integralize each to obtain fine faces. Compare the two collections of fine log schemes (with their pointed maps to X). If they are not canonically isomorphic, Construction 3.11 is chart-dependent and Theorem A is false.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Construction 3.11 defines the fine faces of a coherent log scheme X. Theorem A asserts that these are canonical and jointly surjective, and Proposition 3.16 reduces the surjectivity to the case of a global chart. The definition, however, contains two unproved assertions on which everything rests: (i) 'the connected components of the Z_p ... do not depend on the choice of chart'; and (ii) the étale presheaf F^pre_T, defined on a strict étale cover T→X admitting a global chart, 'extends uniquely to a sheaf' on X_ét. If (i) is false, the phrase 'the set of fine faces' is not well-defined and Theorem A collapses. If (ii) is false, the reduction from local charts to X is invalid and the closed immersions F^int→X are not constructed. The paper gives no proof or reference for either assertion. This is an internal gap, not a disagreement: different charts for the same log structure can have different prime ideals—e.g., the log point Spec k with characteristic N can be charted by Q=N (generator s↦0) or by Q=N^2 (x↦0, y↦1), where the trivial face comes from the prime ideal N\\{0} in the first chart but from the face N·e2 in the second. It is plausible that the resulting collections of faces agree up to canonical isomorphism, but the paper does not show it. Until the chart-independence and descent statements are proved, Theorem A should be read as conditional.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a decomposition of a coherent logarithmic scheme X into a canonical collection of fine logarithmic subschemes, called fine faces, which are defined locally via prime ideals of charting monoids and then integralized. Theorem A asserts that the underlying schemes of these fine faces admit jointly surjective closed immersions into X. The construction is motivated by scheme-theoretic fiber products of toric varieties, where the fine faces recover the normalized components of the fiber product, and by logarithmic Gromov–Witten theory, where non-fine fiber products arise naturally. Section 5 provides algorithms, based on monoidal Gröbner bases and saturation, to compute the faces and their integralizations (Theorem B), and a formal statement about extended cone complexes of fs faces (Theorem C).","tokens_in":23249,"tokens_out":11551,"duration_ms":104850,"significance":"If the construction is made fully rigorous, the paper gives a useful new tool: a way to decompose arbitrary coherent log schemes into fine pieces, thereby allowing tropical and toric techniques to be imported into settings that are only coherent. The algorithms in Section 5 are concrete and potentially valuable for computations in logarithmic enumerative geometry. The main structural proof of Proposition 3.16 is clear assuming the face construction is well-defined, and the examples (toric fiber products, Chow quotients, double ramification cycles) illustrate genuine applications. The paper's central claims are not derived from the authors' prior work and the overall strategy is original.","major_comments":[{"comment":"The definition of the set of faces rests on two unproved assertions: (i) that the connected components of the subschemes Z_p 'do not depend on the choice of chart'; and (ii) that the étale presheaf F^pre_T on a strict étale cover T→X 'extends uniquely to a sheaf' F^pre_X on X_ét. These assertions are load-bearing for Theorem A, for Definition 3.15, and for Proposition 3.16, which assumes both chart-independence and descent when reducing to a global chart. The example of a log point with characteristic N charted by Q=N or by Q=N^2 shows that different charts can yield different prime ideals and different individual closed subschemes, so the claimed invariance is not formal. The paper should provide a proof of chart independence (for example, by comparing two charts via a common refined chart and showing that the resulting collections of connected components, with their induced log structures, are canonically identified) and a proof of étale descent for F^pre. Without these, the phrase 'the set of fine faces' is not well-defined and Theorem A is conditional.","section":"§3.2, Construction 3.11"},{"comment":"There is an internal notation inconsistency that affects the correctness of the main algorithm. The conventions state: 'we denote by a∨b the coordinatewise-minimum (min(a_i,b_i))_{i∈I} and by a∧b the coordinatewise-maximum.' However, Algorithm 5.14 defines the S-relation as (b_i + a_i∨a_j - a_i, b_j + a_i∨a_j - a_j) and the proof of that algorithm uses a_i∨a_j as the coordinatewise maximum (the least common multiple of the initial terms). With ∨ defined as coordinatewise minimum, the expression a_i∨a_j - a_i can have negative coordinates, so the relation is not a monoidal relation in F×F and the algorithm as literally written is not well-defined. The proof of Theorem B depends on this algorithm. The notation should be fixed (for instance, by defining ∨ as coordinatewise maximum and ∧ as coordinatewise minimum, or by writing max(a_i,a_j) explicitly throughout Section 5).","section":"§1 (Notation) and §5.3, Algorithm 5.14"}],"minor_comments":[{"comment":"The 'Observe' following Definition 3.5 asserts that for F = P\\I ∪ {∞} the natural map F → P/I is an isomorphism, but this is false for general ideals I that do not contain ∞. For example, P = N∪{∞} and I = {1,2,...} gives F = {0,∞}, while P/I has an additional absorbing element. The statement is true when I is replaced by I∪{∞} for a prime ideal I, which is the case used in Lemma 3.7; the general claim should be removed or corrected.","section":"§3.1, Definition 3.5"},{"comment":"Theorem C is stated as a 'formal consequence' of the construction but no proof is supplied. Since it is announced as a theorem, the authors should either provide a proof or explicitly demote it to a conjecture or a remark with a justification sketch.","section":"§4, Theorem C"},{"comment":"The 'semi-open hypercube ∏_{1≤j≤n}[0,1[v_j' is not a hypercube unless the v_j form an orthogonal basis; it is a fundamental parallelepiped. The terminology should be adjusted for clarity.","section":"§5.6, Step (2)"},{"comment":"There are numerous typographical errors, e.g., 'particulary' and 'particicularly' (p. 2), 'poarlized' (p. 3), 'varities' (p. 3), 'T echniques' (p. 1), and several missing/extra spaces in the bibliography. A careful proofreading pass is needed.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is promising and the main idea is attractive, but the two major comments above are load-bearing: without a proof of chart-independence and descent in Construction 3.11, Theorem A is not actually established; without fixing the coordinatewise min/max conflict, the Buchberger algorithm in Section 5 is not well-defined. Both issues appear fixable within the manuscript's scope, so I recommend major revision rather than rejection. The authors should also consider adding a short proof or reference for Theorem C."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The fine-face decomposition is the real thing here. The idea that every coherent log scheme has a canonical cover by fine log subschemes is new and worth having, and the examples from toric fiber products and moduli of stable maps show why it matters. The monoid algorithms in Section 5 are a genuine contribution: they give a concrete way to compute the faces and their integralizations, and the reduction to pure difference ideals is sensible. I agree with the reader that Theorems A and B are not repackaging, and the self-citations are motivational, not load-bearing. The structural proof of Proposition 3.16 is clear once the faces are defined.\n\nThe soft spot is exactly where the stress-test note lands. Construction 3.11 asserts that the connected components Z_p do not depend on the choice of chart, and that the étale presheaf F^pre_T extends uniquely to a sheaf, but no proof is given for either. That is an internal gap: Theorem A depends on it, and Proposition 3.16 only proves joint surjectivity after the faces are known to be well-defined. The log point example with N versus N^2 in the stress-test note shows this is not a phantom concern; different charts can present the same log structure with different prime ideals. It is plausible that the collections agree, but the paper does not show it. This needs to be fixed before I would rely on Theorem A as stated.\n\nThe second issue is smaller but real: the conventions define a∨b as coordinatewise minimum and a∧b as maximum, while Algorithm 5.14 and its proof use a∨b as coordinatewise maximum. The critical pair formula and the claim that m = a_i∨a_j + d both require ∨ to be max. As written, the algorithm is broken. This looks like a simple typo, but it needs correcting.\n\nTheorem C is also stated without proof, which is thin for a numbered theorem, though the statement is plausible.\n\nThis paper is for log geometers, especially people working on log Gromov-Witten theory and moduli of abelian varieties. It deserves a serious referee, not a desk rejection. I would send it out with a request that the authors prove or explicitly reference the chart-independence and descent in 3.11, fix the Buchberger notation, and either prove Theorem C or mark it as a conjecture.","headline":"A new and useful decomposition of coherent log schemes into fine pieces, with real algorithmic content, but the main construction has an unproved canonicality claim and the Buchberger algorithm's notation is internally inconsistent.","tokens_in":23784,"tokens_out":4717,"would_cite":true,"duration_ms":47006,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14A21","14M25","14N35","13P10"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that every coherent logarithmic scheme decomposes canonically into fine logarithmic subschemes, its fine faces, which jointly cover the scheme.","keywords":["coherent logarithmic schemes","fine faces","logarithmic structures","integralization","monoidal Gröbner bases","binomial schemes","tropical geometry","toric fiber products"],"falsifier":"Take a coherent logarithmic scheme with two different global charts and compare the collections of connected components obtained by pulling back Spec(ℤ[Q]/(p)) for all prime ideals p; if the two collections of integralized closed subschemes differ for some scheme, the canonicality asserted in Theorem A is false.","tokens_in":22738,"feed_emoji":"🧩","tokens_out":7905,"duration_ms":72842,"temperature":0.7,"pith_summary":"Logarithmic structures track monomial functions on a scheme, and the most useful such structures are the fine ones. But fineness is not preserved by fiber products and other limits, so many natural schemes only carry a coherent logarithmic structure. This paper shows that every coherent logarithmic scheme is covered by a finite collection of fine logarithmic subschemes, called its fine faces, each mapping into the original scheme by a closed immersion and with the integralization appearing as one of the faces. If true, this lets tropical geometry, normally confined to fine and saturated log schemes, be applied to the broader coherent setting by working face by face. The paper also conjectures that Alexeev's compactified moduli space of polarized abelian varieties carries such a coherent structure, with maximal fine faces matching the components of its normalization.","feed_headline":"Coherent log schemes decompose into fine faces","feed_subtitle":"Fine faces let tropical geometry reach beyond the fine setting, with algorithms to compute them.","key_machinery":"The central object is the fine face of a coherent logarithmic scheme, defined through faces of the charting monoid: for a chart X → Spec ℤ[Q], a face R of Q (a submonoid closed under taking summands) gives a closed subscheme Spec ℤ[R] ↪ Spec ℤ[Q], which pulls back to X. The key identity is Lemma 3.7: a prime ideal in a pointed monoid has complement a face, and Spec(ℤ[P]/(p)) = Spec ℤ[R], so prime ideals of the charting monoid cut out exactly the log schemes charted by finer monoids. The algorithms of Section 5 use monoidal Gröbner bases and rewriting rules to compute prime ideals, integralizations via localization, and saturations via lattice-point enumeration in cones, making the face decomposition effective from a monoid presentation.","core_discovery":"The central claim is Theorem A: for every coherent logarithmic scheme X, there is a canonically defined collection of fine logarithmic schemes F, the fine faces, whose underlying schemes admit jointly surjective closed immersions F ↪ X, and one of these faces is the integralization of X. The construction is chart-based: a chart X → Spec ℤ[Q] by a finitely generated monoid Q gives, for each prime ideal p of Q, a closed subscheme Spec(ℤ[Q]/(p)) which by Lemma 3.7 equals Spec ℤ[R] where R is the complementary face of Q; pulling this back to X, taking connected components, and integralizing produces the fine faces. The collection is asserted to be independent of the chosen chart and to descend étale-locally. The paper also proves Theorem B, that these faces can be computed by monoidal Buchberger-style algorithms from presentations of the charting monoids, and Theorem C, that the extended cone complex of an fs face of a fiber product of fs log schemes is a face of the fiber product of the extended cone complexes, so the faces are visible tropically.","pith_inferences":["If the chart-independence of the fine faces is made fully rigorous, every coherent log scheme would carry a canonical stratification by fine log schemes, potentially allowing moduli stacks of coherent log schemes to be studied by gluing fine charts.","The paper's analogy between coherent log schemes and binomial schemes suggests that enumerative invariants of non-fine targets could be computed by summing contributions from fine faces, provided virtual classes behave well under the closed immersions.","The saturation algorithm's exponential worst-case behavior, noted as analogous to the shortest vector problem, hints that computing fine faces is intrinsically hard in general, so structural or tropical shortcuts may be necessary for practical large-scale examples."],"forward_implications":["Tropical geometry, normally restricted to fine and saturated log schemes, becomes applicable to arbitrary coherent log schemes by working on each fine face.","For scheme-theoretic fiber products of toric varieties, the fine faces are exactly the normalizations of the irreducible components, which are toric, so the combinatorial description via extended tropicalization computes the full fiber product.","In logarithmic Gromov–Witten theory, non-fine points of the non-fs fiber product correspond to genuinely new maps, such as the example with constraint λ = μ, explaining why naive generalizations of degeneration formulas fail and how the fine faces capture the missing complexity.","The monoidal Gröbner basis algorithms of Section 5 compute fine faces and their integralizations from explicit charting monoids, making the decomposition effective in moduli contexts where tropical geometry supplies the charts."],"supporting_citations":[{"why":"Supplies the binomial ideal machinery, including the result that normal prime binomial schemes are toric (used in Theorem 4.3), and the term-elimination algorithms adapted in Section 5.","marker":"[ES96]"},{"why":"Provides the notion of pointed monoids, which underlies the definition of faces via prime ideals and absorbing elements.","marker":"[HMU19]"},{"why":"Introduces extended cone complexes, which are used in Theorem C to compare fine faces with fiber products of extended tropicalizations.","marker":"[ACP15]"},{"why":"Establishes the extended tropicalization of toroidal embeddings and the observation that extended tropicalization does not commute with fiber products, motivating the face construction.","marker":"[Pay09]"},{"why":"Supplies representability of moduli of stable log maps and the identification of the moduli space M(X,D) with an fs fiber product (Proposition 4.15).","marker":"[AC14]"},{"why":"Provides charts for moduli of log stable maps via tropical maps, giving the monoid descriptions used in Section 4.2.2.","marker":"[GS13]"},{"why":"Establishes that the Chow quotient is the normalized closure of the torus, used to prove the Chow quotient is the fs inverse limit of GIT quotients (Proposition 4.13).","marker":"[KSZ91]"},{"why":"Supplies tropical geometry as a source of charts in moduli contexts, providing the input data for the algorithms of Section 5.","marker":"[CCUW20]"}],"fun_headline_variants":["Coherent log schemes split into fine faces","Fine faces: canonical decomposition for log schemes","Log schemes get fine faces from monoid charts","Many faces of log schemes, now fine and computable","Canonical fine faces for any coherent log scheme"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the collection of fine faces is independent of the chosen chart and descends uniquely to a sheaf; if two different charts produced different connected components, Theorem A would fail.","fun_headline_variants_meta":{"raw":{"variants":["Coherent log schemes split into fine faces","Fine faces: canonical decomposition for log schemes","Log schemes get fine faces from monoid charts","Many faces of log schemes, now fine and computable","Canonical fine faces for any coherent log scheme"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000685,"raw_usage":{"total_tokens":3062,"prompt_tokens":853,"completion_tokens":2209,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":469,"completion_tokens_details":{"reasoning_tokens":2137}},"tokens_in":469,"tokens_out":2209,"duration_ms":15295,"temperature":1.0,"reasoning_tokens":2137,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:16:56.530415+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a coherent logarithmic scheme with two different global charts and compare the collections of connected components obtained by pulling back Spec(ℤ[Q]/(p)) for all prime ideals p; if the two collections of integralized closed subschemes differ for some scheme, the canonicality asserted in Theorem A is false.","supporting_citations":[],"review_version":1}