{"id":"3fd11e5d-16d3-4648-acb3-687d204998a0","arxiv_id":"2411.17909","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Stable reduction for log canonically polarized varieties holds in large characteristic assuming two standard MMP conjectures; for surfaces this recovers Hacon-Kovács properness of the moduli of stable surfaces.","lead":"A mathematician proposes a new general recipe for stable reduction in higher dimensions, a key step for building moduli spaces. The recipe works in large characteristic and, for surfaces, recovers a known theorem with an explicit but non-computable characteristic bound.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Surface corollary's LCM input may not cover mixed characteristic: HNT20 Theorem 4.11 as cited appears to be field-based, while Remark 1.5 claims mixed-characteristic stable reduction.","rationale":"The reader's weakest assumption is Conjecture LCM_{n+1}, and I agree that this is the main conditional input to Theorem 1.1. My stress-test identifies a more specific place where that assumption is discharged: in the surface case, Corollary 1.2 uses HNT20 Theorem 4.11 to construct the LCM over Spec R. If HNT20 only applies over a field of characteristic p, then the proof does not cover mixed-characteristic DVRs, contradicting Remark 1.5. The conditional Theorem 1.1 itself appears internally coherent: the vertical-component concern is resolved because any vertical component of Δ in the LCM would add to the already reduced coefficient 1 of a central fibre component, violating lc, so Δ is horizontal and coeff(Δ)=coeff(Δ_K); the volume bound from Conjecture V_n then gives m_i < p, making the base change tame. The surface moduli application over k only needs equicharacteristic DVRs, so Theorem 1.3 is not threatened by this issue. Therefore the reader's CONDITIONAL verdict remains appropriate, and the concrete check on HNT20's hypotheses would settle whether Corollary 1.2 and Remark 1.5 need a caveat.","tokens_in":19850,"tokens_out":39858,"duration_ms":389772,"concrete_test":"Read the statement of [HNT20, Theorem 4.11] and record its base hypotheses: does it allow a projective morphism X → Spec R where R is a DVR with perfect residue field k of char p > 5 and fraction field of characteristic 0, or only a morphism X → C with C a curve over a perfect field k of characteristic p? If the latter, check whether Corollary 1.2 can be repaired by adding an equicharacteristic hypothesis or by citing a mixed-characteristic MMP theorem such as [BMP+23] or [ABP24]; if no repair is available, Corollary 1.2 and Remark 1.5 should be weakened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Corollary 1.2 is the only unconditional result, and its proof obtains the required log canonical model over Spec R by invoking [HNT20, Theorem 4.11] after passing to a log resolution. The quoted phrase 'over any perfect field k with char k > 5' indicates that HNT20's theorem is a relative MMP statement for threefolds over a perfect field of characteristic p, i.e., for a base curve over that field. In the mixed-characteristic case allowed by Remark 1.5, Spec R is not a curve over the residue field k, and there is no k-algebra structure on R. Thus the cited theorem may not supply the LCM over Spec R in that case, and the proof of Corollary 1.2 would only establish equicharacteristic stable reduction. Since the moduli application (Theorem 1.3) only needs DVRs that are k-algebras, the main properness claim over k survives, but the unconditional statement of Corollary 1.2 and Remark 1.5 overclaim unless another source (e.g., ABP24 or BMP+23) supplies the mixed-characteristic LCM. This is not an internal inconsistency of Theorem 1.1, which explicitly assumes LCM_{n+1}; it is a load-bearing gap at the point where that assumption is discharged for surfaces.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes a higher-dimensional stable reduction conjecture (SR_n) for log canonically polarized pairs over a DVR with algebraically closed residue field, and proves a conditional theorem. Assuming (i) existence of a log canonical model over Spec R for the pair with the reduced central fibre (Conjecture LCM_{n+1}) and (ii) a uniform lower volume bound for n-dimensional stable log varieties with coefficients in a DCC set (Conjecture V_n), Theorem 1.1 shows that if char k > v/v(n,D(I)) and vol(K_{X_K}+Δ_K) ≤ v, then SR_n holds: the author bounds multiplicities of the central fibre components by the volume, then makes a tame base change of degree the lcm of these multiplicities and invokes Proposition 3.2. For surfaces, V_2 is known via [HK19] and the proof invokes [HNT20] to supply LCM_3, yielding unconditional stable reduction (Corollary 1.2); with Posva's gluing and [ABP24] this recovers properness and projectivity of the moduli of stable surfaces over large characteristic fields (Theorem 1.3).","tokens_in":126,"tokens_out":34626,"duration_ms":449112,"significance":"If correct, the conditional theorem is a clean and useful reduction: it replaces ad hoc stable reduction arguments with two standard MMP conjectures and gives an explicit characteristic bound and explicit base change. The proof of Theorem 1.1 is transparent, has no fitted parameters, and does not appear circular; the conjectures are honestly stated as hypotheses. The surface application is the main concrete evidence, while the higher-dimensional statement is conditional on substantial open problems. The paper is a well-written note whose value is conceptual and expository, and it recovers known properness and projectivity results for stable surfaces.","major_comments":[{"comment":"The proof of Corollary 1.2 discharges Conjecture LCM_3 by invoking [HNT20, Theorem 4.11] after passing to a log resolution, quoting it as a theorem 'over any perfect field k with char k > 5'. If that theorem is a statement about projective threefolds over a perfect field, it does not by itself produce a relative log canonical model over Spec R when R is a mixed-characteristic DVR, since Spec R is not a k-scheme and no k-algebra structure is available on R. The proof as written therefore establishes Corollary 1.2 only for equicharacteristic DVRs, and Remark 1.5 overclaims in asserting that the results hold in both mixed and equi-characteristic. This is not an internal inconsistency in Theorem 1.1, which explicitly assumes LCM_{n+1}, but it is load-bearing for the unconditional surface statement. Theorem 1.3 survives, because the valuative criterion over k only tests DVRs that are k-algebras; still, the author should either restrict Corollary 1.2 and Remark 1.5 to the equicharacteristic case or supply a valid mixed-characteristic reference for the lc model over Spec R (for instance [ABP24] or [BMP+23]).","section":"§3, proof of Corollary 1.2; Remark 1.5"}],"minor_comments":[{"comment":"There is a typo: 'De ligne–Mumford' should be 'Deligne–Mumford'.","section":"Abstract"},{"comment":"The sign in the displayed equivalence and the subsequent application of [Kol13, 1.17] are terse; please state explicitly which variant of the negativity lemma is used and why E1−E2 satisfies its hypotheses.","section":"Theorem 3.1, equation (4)"},{"comment":"The displayed equality coeff(D_i^K+Δ_i^K)=coeff(Δ_i^K) is not literally correct, since the conductor D_i^K has coefficient 1; the needed inclusion coeff(D_i^K+Δ_i^K) ⊆ {1} ∪ coeff(Δ_i^K) ⊆ I nevertheless holds because I contains 1.","section":"Proof of Corollary 3.7"},{"comment":"The sentence 'Since R contains k = \\bar{k}, g is Galois' assumes R is a k-algebra; if R is only a DVR with residue field k, this needs justification or a Henselization step.","section":"Corollary 3.10"},{"comment":"The example depends on an unpublished lecture [McQ]; please add a citable reference or explicitly mark the example as folklore.","section":"Example 3.5"}],"recommendation":"major_revision","confidential_remarks":"The central conditional argument appears sound as far as I checked, and the paper is honest about its hypotheses. The blocking issue is the mixed-characteristic citation gap in the surface corollary; if the author fixes that by restriction or by adding a valid reference, I would support publication. The automorphism-extension part of Theorem 3.1 is terse but I did not find a definitive error; asking for an expanded proof there would improve the note."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe central move here—take the log canonical model first, then do a tame base change—is genuinely new and makes the proof short. Theorem 3.1, characterizing when stable reduction is achieved after a tame extension, is the real contribution; its proof uses a clean negativity argument to extend automorphisms. The higher-dimensional theorem is honestly conditional: it assumes LCM_{n+1} and V_n explicitly, and the volume argument bounding central fiber multiplicities is coherent. The surface corollary recovers Hacon–Kovács properness and adds projectivity via Pos24 and ABP24. That is real progress, even if the surface result is not new.\n\nThe softest spot is the mixed-characteristic claim. Corollary 1.2 is stated for any DVR R with algebraically closed residue field k, and Remark 1.5 explicitly says mixed and equi-characteristic both work. But in the proof, existence of the LCM over Spec R is discharged by [HNT20, Theorem 4.11], which is a statement about threefolds over a perfect field—i.e., equicharacteristic. As cited, it does not provide a log canonical model over a mixed-characteristic DVR. The stress-test note is correct. This is not fatal: the moduli application Theorem 1.3 only needs DVRs that are k-algebras, so the properness and projectivity statements over k survive. But Corollary 1.2 and Remark 1.5 overclaim as written. The author should either cite a genuine mixed-characteristic source or restrict those statements to equicharacteristic DVRs.\n\nOne smaller annoyance: the paper sells the characteristic bound as explicit, but it depends on the positive constant v(2,D(I)) from Alexeev and Hacon–Kovács, which is not numerically computed. That is acceptable if phrased as 'up to the boundedness constant', but the current phrasing oversells it.\n\nThe citation pattern looks fine: heavy reliance on very recent papers, but honest, and the conditional statements are cleanly stated. I checked the surface proof and found no circularity. The verification burden is high—independent checking of the Kol23b-style automorphism extension is warranted—but I did not find an internal contradiction.\n\nShould this go to peer review? Yes. The core idea and Theorem 3.1 deserve a serious referee. The mixed-characteristic gap is a revision issue, not a desk-reject issue. I'd send it, with a referee asked to check the HNT20 citation carefully.\n\nBest.","headline":"A genuinely attractive LCM-first proof of stable reduction with a clean tame base-change characterization; the mixed-characteristic claim in the surface corollary overreaches its citation, but the core is sound and deserves refereeing.","tokens_in":20646,"tokens_out":3152,"would_cite":true,"duration_ms":28606,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14D06","14E30","14G17","14J17"],"pacs":[],"model":"deepseek-v4-flash","headline":"Assuming two standard MMP conjectures, this paper proves stable reduction for log canonically polarized pairs in large characteristic, and recovers properness and projectivity of the moduli of stable surfaces.","keywords":["stable reduction","log canonical model","positive characteristic","moduli of stable surfaces","minimal model program","boundedness of volume","tame base change","DCC sets"],"falsifier":"Exhibit a DVR with algebraically closed residue field of characteristic $p > v/v(n,D(I))$ and a projective geometrically normal pair $(X_K,\\Delta_K)$ over $K$ with $K_{X_K}+\\Delta_K$ log canonical and ample, volume at most $v$, and coefficients in a DCC set $I$, for which after every finite extension $L/K$ there is no stable log model; this would refute Theorem 1.1, and in dimension two it would refute Corollary 1.2. A less global test is to find an LCM whose central fibre has a component of multiplicity at least $p$ while $(K_X+\\Delta+(X_k)_{\\mathrm{red}})^n\\cdot F_i$ is still uniformly bounded below, which would violate the volume identity and show the characteristic threshold is sharp.","tokens_in":19623,"feed_emoji":"📐","tokens_out":12964,"duration_ms":98740,"temperature":0.7,"pith_summary":"The paper proposes a higher-dimensional analogue of the classical stable reduction theorem for curves and proves it in large characteristic: when the residue characteristic $p$ exceeds the ratio $v/v(n,D(I))$ of the volume of the generic fibre to a universal minimal volume, every component of the central fibre of the log canonical model has multiplicity smaller than $p$, so the natural base change is tame and yields the stable model. The argument runs on two standard Minimal Model Program conjectures: existence of the log canonical model over the base, and a uniform lower bound on volumes of stable log varieties with coefficients in a DCC set. In dimension two both are theorems, so the paper recovers, unconditionally, the properness of the moduli stack and the projectivity of the coarse moduli space of stable surfaces of fixed volume over algebraically closed fields of large characteristic. A characteristic-free theorem characterizes stable reduction after a tame base change in terms of the multiplicities of the central fibre of the log canonical model, generalizing the classical curve criterion.","feed_headline":"Stable reduction in large characteristic follows from a volume bound","feed_subtitle":"The proof bounds central-fibre multiplicities by volume, so the base change is tame when p is large.","key_machinery":"The load-bearing object is the log canonical model $(X, \\Delta+(X_k)_{\\mathrm{red}})$ over the base DVR — the pair obtained by adding the reduced central fibre to the given boundary — together with the identity $v \\ge \\operatorname{vol}(K_{X_K}+\\Delta_K) = \\sum_i m_i \\operatorname{vol}(K_{F_i}+B_i)$, where $F_i$ are the components of the central fibre with multiplicities $m_i$ and $B_i$ is the different on the normalization of $F_i$. This identity converts the total volume into a weighted sum of component volumes, each bounded below by $v(n,D(I))$ by Conjecture $V_n$, so each multiplicity is bounded by $v/v(n,D(I))$. The enabling criterion is Theorem 3.1: a log canonically polarized pair has stable reduction after a tamely ramified base change of degree $N$ if and only if there is an LCM whose central-fibre multiplicities have least common multiple dividing $N$. The forward direction uses the tame cover $\\pi = t^N$ and a standard discrepancy lemma for ramified covers; the reverse direction builds the LCM as the quotient of the stable model by a cyclic Galois group.","core_discovery":"The central claim is Theorem 1.1: for fixed dimension $n$, volume bound $v$, and DCC coefficient set $I$, if Conjecture $\\operatorname{LCM}_{n+1}$ (existence of a log canonical model of the pair together with the reduced central fibre over the base DVR) and Conjecture $V_n$ (a uniform positive lower bound on volumes of $n$-dimensional normal stable log varieties with coefficients in a DCC set) hold, then Conjecture $\\operatorname{SR}_n$ holds for all pairs with volume at most $v$ and coefficients in $I$, provided $\\operatorname{char} k > v/v(n,D(I))$. The proof starts with the log canonical model, writes the central fibre $X_k = \\sum_i m_i F_i$, and uses the identity $\\operatorname{vol}(K_{X_K}+\\Delta_K) = \\sum_i m_i \\operatorname{vol}(K_{F_i}+B_i)$ to show each multiplicity $m_i$ is at most $v/v(n,D(I))$. When the characteristic exceeds this ratio, the least common multiple $N$ of the multiplicities is prime to $p$, so the base change $\\pi = t^N$ is tame; the normalization of the pullback is then the stable log model, with reduced central fibre and log canonical total space. In the surface case both conjectures are known theorems, giving an unconditional stable reduction statement for log canonically polarized surfaces and recovering the properness of the moduli stack of stable surfaces and the projectivity of its coarse space over large characteristic fields.","pith_inferences":["If Conjectures $\\operatorname{LCM}_{n+1}$ and $V_n$ are eventually proved in all dimensions, the same volume argument would yield stable reduction for log canonically polarized pairs in arbitrary dimension over large characteristic fields, with an explicit characteristic threshold.","The volume identity gives a practical way to detect when wild base change is unavoidable: by Corollary 3.3, an LCM component whose multiplicity is divisible by $p$ forces any stabilizing base change to have degree divisible by $p$, so the open problem of finding the right base change in that case is the main obstruction to removing the large-characteristic assumption.","Making the boundedness constant $v(n,D(I))$ effective would turn the theorem into an explicit arithmetic bound on the characteristic, which could be tested computationally on examples of log surfaces.","The description of stable limits in Corollary 3.10 suggests a concrete recipe for constructing higher-dimensional stable limits in practice: pass to the LCM, make the tame base change reducing all multiplicities to one, and read off the limit components as covers of the LCM's central components."],"forward_implications":["Conjecture $\\operatorname{SR}_2$ holds unconditionally for log canonically polarized surfaces over algebraically closed fields of characteristic greater than $\\max\\{5, v/v(2,D(I))\\}$, for any fixed volume bound $v$ and DCC set $I$.","The moduli stack of stable surfaces of volume $v$ is proper, and its coarse moduli space is projective, over such fields — recovering the known properness theorem for stable surfaces in large characteristic.","The required base change is explicit: it is the tame cover $\\pi = t^N$ with $N$ the least common multiple of the central-fibre multiplicities of the log canonical model, and one can always take $N = (p-1)!$.","The stable limit is explicitly described: if $F$ is a component of the central fibre of the LCM with multiplicity $m_F$, its preimage splits into $\\gcd\\{m_F, n_j\\}_j$ reduced components, each a degree $m_F/\\#F$ cover of $F$ with volume $(m_F/\\#F)\\cdot\\operatorname{vol}(K_F+B)$.","In characteristic zero, the same circle of ideas proves Conjecture $\\operatorname{SR}_n$ for klt pairs, because the log canonical model exists under mild assumptions."],"supporting_citations":[{"why":"Supplies the curve stable reduction theorem that Conjecture SR_n extends to higher dimensions.","marker":"[DM69, Corollary 2.7]"},{"why":"Provides the ramified-cover discrepancy lemma used in Proposition 3.2 and in the tame base change proof.","marker":"[Kol13, 2.42-2.43]"},{"why":"One of the two boundedness results establishing Conjecture V_2 for surfaces.","marker":"[Ale94, Theorem 8.2]"},{"why":"Completes Conjecture V_2 and is the properness theorem whose surface moduli result is recovered.","marker":"[HK19, Theorem 2]"},{"why":"Supplies the log canonical model for threefolds in positive characteristic, making the surface case of Theorem 1.1 unconditional.","marker":"[HNT20, Theorem 4.11]"},{"why":"Gives the gluing theory that extends normal stable reduction to demi-normal slc surfaces.","marker":"[Pos24, Theorem 6.0.5]"},{"why":"Yields the projectivity of the coarse moduli space of stable surfaces.","marker":"[Pat17, Theorem 1.2]"},{"why":"Shows D(I) is a DCC set when I is, so Conjecture V_n applies with coefficient set D(I).","marker":"[MP04, Lemma 4.4]"}],"fun_headline_variants":["Stable reduction in large characteristic via log canonical model","Volume bound tames base change: stable reduction for large p","Conditional on MMP conjectures, stable reduction holds for large p","Log canonical model yields tame base change for stable reduction","Stable reduction for large p from volume bounds on central fibres"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument collapses without Conjecture $\\operatorname{LCM}_{n+1}$: there must exist a log canonical model of $(X, \\Delta+(X_k)_{\\mathrm{red}})$ over the DVR whose central fibre has components with bounded multiplicities, and in higher dimensions this existence is an open Minimal Model Program statement, while in the surface case it is supplied by a known existence theorem for threefolds in positive characteristic.","fun_headline_variants_meta":{"raw":{"variants":["Stable reduction in large characteristic via log canonical model","Volume bound tames base change: stable reduction for large p","Conditional on MMP conjectures, stable reduction holds for large p","Log canonical model yields tame base change for stable reduction","Stable reduction for large p from volume bounds on central fibres"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000924,"raw_usage":{"total_tokens":4214,"prompt_tokens":948,"completion_tokens":3266,"prompt_tokens_details":{"cached_tokens":896},"prompt_cache_hit_tokens":896,"prompt_cache_miss_tokens":52,"completion_tokens_details":{"reasoning_tokens":3183}},"tokens_in":52,"tokens_out":3266,"duration_ms":68247,"temperature":1.0,"reasoning_tokens":3183,"cache_read_input_tokens":896,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:47:23.052002+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a DVR with algebraically closed residue field of characteristic $p > v/v(n,D(I))$ and a projective geometrically normal pair $(X_K,\\Delta_K)$ over $K$ with $K_{X_K}+\\Delta_K$ log canonical and ample, volume at most $v$, and coefficients in a DCC set $I$, for which after every finite extension $L/K$ there is no stable log model; this would refute Theorem 1.1, and in dimension two it would refute Corollary 1.2. A less global test is to find an LCM whose central fibre has a component of multiplicity at least $p$ while $(K_X+\\Delta+(X_k)_{\\mathrm{red}})^n\\cdot F_i$ is still uniformly bounded below, which would violate the volume identity and show the characteristic threshold is sharp.","supporting_citations":[],"review_version":1}