{"id":"1cd76bfa-5aa0-48c3-a6a7-480e8413aad8","arxiv_id":"2412.03006","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"ACC for minimal log discrepancies is proved for every fixed normal threefold.","lead":"This paper proves that minimal log discrepancies on any fixed threefold satisfy the ascending chain condition (ACC), a central conjecture in birational geometry. It extends the author's earlier proof for smooth threefolds to arbitrary normal threefolds, including singular ones, via a detailed study of divisorial contractions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 7.4 cites Lemma 6.2 for factoriality of a Gorenstein terminal threefold point, but Lemma 6.2 does not imply factoriality; the ordinary double point is a counterexample, leaving the descent of l_i in Theorem 5.3 with an unproved step.","rationale":"The reader's weakest assumption identifies Section 7's termination argument as the load-bearing point, and I agree that Theorem 5.3 is where the proof must be checked most carefully. My concern is more specific than the reader's: rather than the completeness of the external classification, the immediate textual gap is in Lemma 7.4's use of Lemma 6.2. The assertion that X_j is factorial at x_j is not a consequence of the quoted lemma, and the ordinary double point shows that Gorenstein terminal threefold germs need not be factorial. If Lemma 7.4 cannot be repaired by replacing 'factorial' with the weaker and true statement that E_j is Cartier at x_j, and by a separate argument that E_j is the only exceptional divisor through x_j, then the strict descent of l_i is unsupported. This matters because Propositions 7.9 and 7.11 rely directly on Lemma 7.4, and Theorem 5.3 relies on them. I do not see this as a fatal objection: the missing step is local and checkable, and the surrounding classification may well supply the needed uniqueness. Hence the reader's CONDITIONAL verdict remains appropriate; no adjustment is needed, but the condition should explicitly include a complete justification of Lemma 7.4.","tokens_in":24808,"tokens_out":38733,"duration_ms":355634,"concrete_test":"Verify Lemma 6.2 in [14]: if it does not imply factoriality, test the critical local case where X_j is the ordinary double point x1x2+x3x4=0 and E_j is a non-Cartier ruling. Check whether a general z4 in m_i with z4 not in m_j^2 generates the ideal of E_j locally, and compute the cokernel lengths b and g in the diagram of Lemma 7.4. If g<b fails, identify whether the alternative m_j<m_i occurs; then confirm that Propositions 7.9 and 7.11 still cover that alternative. If the uniqueness of the exceptional divisor through x_j is instead the issue, inspect Proposition 5.2's choice of x_j=C_j∩E_j and determine from Table 1 whether x_j can lie on the strict transform of an earlier exceptional divisor.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 7.4 is the engine that produces strict descent of the invariant l_i in Propositions 7.9 and 7.11, and hence in the termination argument for Theorem 5.3. In its proof, after choosing z4 generally from m_i and assuming that m_i O_{X_j} is not contained in m_j^2, the text asserts that 'X_j is factorial at x_j by Lemma 6.2' and concludes that z4 cuts out the unique π_ij-exceptional prime divisor E_j through x_j. Lemma 6.2, as quoted from [14, Lemma 5.1], only says that every Q-Cartier divisor on a terminal threefold germ is linearly equivalent to an integer multiple of K_X. That does not make the local ring factorial: the ordinary double point x1x2+x3x4=0, which the paper itself uses as a terminal Gorenstein non-Q-factorial singularity, is a counterexample. What the length computation actually needs is that the ideal m_i O_{X_j} is principal at x_j with generator z4; this requires E_j to be Cartier at x_j and also that no other π_ij-exceptional divisor passes through x_j. The Cartier property follows from Lemma 6.2 because E_j is Q-Cartier and x_j is Gorenstein, but the uniqueness of the exceptional divisor through x_j is not proved. The cited factoriality is therefore both false and insufficient as written. If the point x_j constructed in Proposition 5.2 can lie on an earlier exceptional divisor in some case covered by Table 1, the inequality g<b in Lemma 7.4 may fail, and the termination argument has a real gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves Theorem 1.1: for a fixed normal threefold X over an algebraically closed field of characteristic zero and a DCC set I of positive real numbers, the set of minimal log discrepancies mld_η(X,Δ) over all scheme-theoretic points η and boundaries Δ with coefficients in I satisfies the ACC. The proof reduces to a boundedness statement on a fixed Q-factorial terminal threefold germ (Theorem 3.1), constructs a generic limit pair (Rhat X,Rhat Δ), builds birational models by means of divisorial contractions (Propositions 5.1 and 5.2), and proves termination of the resulting sequence of contractions through two invariants m_i and l_i (Theorem 5.3). Theorems 1.2–1.4 are derived from Theorem 3.1 via equivalences established in the author's earlier work.","tokens_in":25133,"tokens_out":7190,"duration_ms":70270,"significance":"If the proof is correct, this is a substantial result: it establishes the ACC for minimal log discrepancies on every fixed threefold, including singular threefolds, and with it the equivalent uniform m-adic semi-continuity, ACC for a-lc thresholds, and Nakamura boundedness on a fixed threefold. The strategy is coherent and builds naturally on the author's prior classification of threefold divisorial contractions and on the smooth-threefold case. The paper also benefits from a clear reduction to a terminal germ and from the use of generic limits. However, the proof is long and relies heavily on delegated classifications and computations; the points raised below need to be addressed before the central claim can be considered verified.","major_comments":[{"comment":"The proof contains the sentence “Recall that X_j is factorial at x_j by Lemma 6.2.” This is not correct: Lemma 6.2 only states that every Q-Cartier divisor on a terminal threefold germ is linearly equivalent to an integer multiple of K_X. It does not imply factoriality of the local ring; a terminal Gorenstein ordinary double point is a counterexample. The subsequent conclusion that E_j is the only π_ij-exceptional prime divisor through x_j and that it is cut out by z_4 requires a proof. Since Lemma 7.4 is the engine that produces the strict descent of l_i in Propositions 7.9 and 7.11, and hence the termination argument for Theorem 5.3, this is a load-bearing gap. Please either prove the required factoriality (or the weaker Cartier/uniqueness statement) from the actual hypotheses, or replace the argument.","section":"§7, Lemma 7.4"},{"comment":"The proof of Theorem 4.4 is delegated: the text says “The proof is the same as that of [13, theorem 4.10] ... one can write down the complete proof following the proof in [13] verbatim.” This theorem supplies the key equality mld_x(X,Δ_i) = mld_{y_i}(Y_i,Γ_i) or mld_{Rhat x}(Rhat X,Rhat Δ), so it is central to the reduction. The author explicitly notes that Rhat X may fail to be Q-factorial, which is a real difference from the smooth case. Please include a complete proof or a detailed lemma-by-lemma translation that makes the adaptation visible and verifiable, rather than leaving the proof to the reader.","section":"§4, Theorem 4.4"},{"comment":"In the final case of Lemma 7.10, the assertion “Now by a direct computation of the weighted blow-up π_{i+1}, one can check that X_{i+2} has only quotient singularities” is not shown. This computation is used in Proposition 7.11(iii) to dispose of the case (e13, 2, 3) with a_{j+1}/n_{j+1} = 3/2. Please provide the computation in detail, or at least a precise reference to the classification result that contains it.","section":"§7, Lemma 7.10"}],"minor_comments":[{"comment":"The symbol l is used both as the index of the family approximation and as a rational bound in the conclusion; this overloading makes the statement harder to read and should be fixed.","section":"§4, Lemma 4.1"},{"comment":"The phrase “For the counterpart of the divisor Rhat Q” introduces an object Rhat Q that has not been defined in the proof; the notation should be introduced or the passage reworded.","section":"§4, proof of Theorem 4.4"},{"comment":"The terms B_ι(4) and B_ι(6) in the formula for d(2) are not defined in the paper; they should be defined or the reader should be pointed to the exact formula in [12].","section":"§6, Lemma 6.5(iii)"},{"comment":"The proof treats quotient singularities via Kawamata's result [15] without stating the precise statement used; a short statement of the needed assertion would improve readability.","section":"§7, proof of Theorem 5.3"}],"recommendation":"major_revision","confidential_remarks":"The false citation in Lemma 7.4 is the most serious issue. If it cannot be repaired by proving the needed Cartier/uniqueness of the exceptional divisor at x_j, the termination argument for the central construction collapses. The paper also relies very heavily on the author's own earlier classification papers [6]–[9] and on [13]; for a journal submission, the delegated proof of Theorem 4.4 and the unshown computation in Lemma 7.10 should be regarded as part of the refereeing burden, not as optional references."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Kawakita proves the ACC for minimal log discrepancies on an arbitrary fixed normal threefold, extending his earlier smooth-threefold theorem. That is a substantial result: combined with lower semicontinuity it gives termination of flips for pairs on a fixed threefold, and via his previous equivalence it yields the boundedness statements in Theorems 1.2–1.4. The architecture is sensible: reduce to a fixed Q-factorial terminal germ, pass to a generic limit, construct models by divisorial contractions, and then use the numerical classification of threefold divisorial contractions to force a smooth model. The reliance on the author's own classification work is legitimate; those are published results and the paper is transparent about quoting them.\n\nThe proof as written has real presentation gaps. Theorem 4.4 is explicitly a reformulation of [13, Theorem 4.10] with the proof left to the reader; Lemma 7.10 ends with 'by a direct computation' for the final case. Those are not fatal, but they make independent verification slow.\n\nThe specific stress-test point about Lemma 7.4 is on target. The text says 'Xj is factorial at xj by Lemma 6.2.' That is false: Lemma 6.2 gives that every Q-Cartier divisor is linearly equivalent to a multiple of K_X, which, when xj is Gorenstein, makes every Q-Cartier divisor Cartier—but it does not make the local ring factorial. The ordinary double point is a terminal Gorenstein singularity that is not factorial. So the sentence as written is wrong. However, the conclusion the proof needs is not full factoriality but uniqueness of the π_ij-exceptional prime divisor through xj. That does follow from the non-containment assumption m_i O_Xj ⊄ m_j^2, using the facts that Xj is Q-factorial, xj is Gorenstein, and hence every exceptional divisor through xj is Cartier; two such divisors would put their product in m_j^2. The author doesn't supply that argument, so the proof has a genuine missing step, but it is repairable in a few lines.\n\nI did not find circularity or invented entities. The citation pattern is heavy on self-citations, but they are to prior published classifications and the author's own verified results, which is appropriate here.\n\nThis is a paper for specialists in birational geometry and singularities. It deserves serious refereeing; a referee should ask the author to fix Lemma 7.4 and to expand the delegated proofs at least to the level of clear statements of the reductions.","headline":"Kawakita proves the fixed-variety ACC for minimal log discrepancies on threefolds, a major result, but a key step in the termination argument contains a false factoriality claim that needs repair.","tokens_in":25692,"tokens_out":14314,"would_cite":true,"duration_ms":132345,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14E30","14B05","14J30"],"pacs":[],"model":"deepseek-v4-flash","headline":"Minimal log discrepancies on any fixed normal threefold satisfy the ascending chain condition.","keywords":["minimal log discrepancy","ACC","DCC","threefold","log canonical pairs","divisorial contractions","generic limits","birational geometry"],"falsifier":"A concrete disproof would be a fixed normal threefold $X$ and a DCC set $I$ for which the values $\\mathrm{mld}_x(X,\\Delta_i)$ at a closed point form an infinite strictly increasing sequence, or, failing that, an infinite sequence of divisorial contractions from Proposition 5.2 with constant $m_i$ and non-decreasing $l_i$ at singular Gorenstein points, which would undermine Theorem 5.3 and hence the bound $l$.","tokens_in":24566,"feed_emoji":"📐","tokens_out":15562,"duration_ms":148684,"temperature":0.7,"pith_summary":"This paper proves that on any fixed normal threefold, the minimal log discrepancies obtained from boundaries with coefficients in a DCC set satisfy the ascending chain condition: no infinite strictly increasing sequence of such values exists. This is the fixed-variety case of a central conjecture in birational geometry, and it was previously known for smooth threefolds but open for singular ones. The author reduces the statement to a uniform boundedness assertion on a fixed terminal threefold germ, then builds an infinite sequence of divisorial contractions from a log canonical limit pair with a one-dimensional centre and shows that an integer invariant attached to each germ strictly decreases until the germ becomes smooth. If the proof is correct, the same mechanism also yields the ACC for a-lc thresholds, a uniform m-adic semi-continuity statement, and a boundedness statement for divisors computing minimal log discrepancies on the fixed threefold.","feed_headline":"Minimal log discrepancies on a fixed threefold obey the ACC","feed_subtitle":"Settles the fixed-variety case in dimension three and yields uniform bounds for divisors computing the discrepancies.","key_machinery":"Three objects carry the argument. First, the generic limit: an infinite sequence of $\\mathbb{Q}$-divisors $\\Delta_i$ on the germ is replaced by one boundary $\\hat{\\Delta}$ on the completion $\\hat{X}$ of the local ring over a large algebraically closed field, so that the ACC problem becomes a statement about a single pair $(\\hat{X},\\hat{\\Delta})$. Second, a birational model: from a log canonical pair whose smallest lc centre is a regular curve $\\hat{C}$, the author constructs a contraction $Y_K\\to X_K$ and, by Proposition 5.2, an infinite sequence of threefold divisorial contractions $E_{i+1}\\subset X_{i+1}\\to x_i\\in X_i$ contracting a divisor to the point $x_i$ on the strict transform of $\\hat{C}$. Third, two integer invariants at the Gorenstein indices of the sequence: $m_i$, the least order of the maximal ideal of $x_i$ along the future exceptional divisors, and $l_i$, defined as the length of the cokernel of the natural map $\\bigwedge^2 I_i/I_i^{(2)}\\to \\omega_{X_i}\\otimes \\omega_{C_i}^{-1}$ for the ideal sheaf $I_i$ of the curve $C_i$. The classification of threefold divisorial contractions (Table 1) is used to show that, after $m_i$ is constant, $l_i$ strictly decreases until $x_i$ is smooth; this termination is Theorem 5.3, and it is the step from which the uniform bound $l$ is derived.","core_discovery":"The paper's central claim is Theorem 1.1: fix a normal threefold $X$ and a subset $I$ of the positive real numbers satisfying the DCC. Then the set $$\\{\\mathrm{mld}_\\eta(X,\\$\\Delta$) \\mid \\eta \\text{ scheme-theoretic point},\\ (X,\\$\\Delta$)\\text{ pair},\\ \\$\\Delta$\\in I\\}$$ satisfies the ACC. Since every minimal log discrepancy at a scheme-theoretic point with closure $Z$ equals the minimal log discrepancy at a general closed point of $Z$ minus $\\dim Z$, the closed-point case is the essential one. The proof of the theorem passes through Theorem 3.1, a boundedness statement on the germ of a $\\mathbb{Q}$-factorial terminal threefold: for every sequence of effective $\\mathbb{Q}$-divisors with denominators bounded by a fixed integer, there is a uniform bound $l$ such that infinitely many of the mld values are computed by divisors whose own log discrepancy with respect to the threefold is at most $l$.","pith_inferences":["The proof's termination step suggests a prototype for higher dimensions: a fixed-variety ACC would follow from a decreasing, integer-valued statistic attached to the minimal centre along a sequence of divisorial contractions, and the pair $(m_i,l_i)$ shows what such a statistic can look like in dimension three.","Because the only classification input is the table of threefold divisorial contractions, a classification-free version of the descent lemma would likely make the uniform bound $l$ effective; one could compute it on explicit terminal quotient germs and compare the rate of descent of $l_i$.","A step the paper does not take is to let the threefold vary in a bounded family and ask whether the bound $l$ in Theorem 3.1 depends only on the family; a positive answer would reduce the unfixed three-dimensional ACC to a stratification of threefold germs."],"forward_implications":["For every fixed normal threefold and every DCC coefficient set, the non-negative minimal log discrepancies at its scheme-theoretic points form an ACC set, so no infinite strictly increasing sequence of values occurs.","The ACC for a-lc thresholds follows (Theorem 1.2): for fixed $a$, the set of thresholds $t$ with $\\mathrm{mld}_\\eta(X,\\Delta+tA)=a$ satisfies the ACC.","Uniform $m$-adic semi-continuity follows (Theorem 1.3): there is an integer $l$ such that ideals sharing their $l$-th powers modulo the maximal ideal give equal minimal log discrepancies.","A boundedness statement for computing divisors follows (Theorem 1.4): on a fixed log terminal threefold, every boundary's minimal log discrepancy is computed by a divisor whose own log discrepancy is bounded by a fixed $l$.","Together with the previously settled smooth case, the theorem completes the ACC for minimal log discrepancies on every fixed threefold."],"supporting_citations":[{"why":"Settles the smooth-threefold case and supplies the final uniform bound on the model that is smooth at the relevant point.","marker":"[13]"},{"why":"Provides the connectedness and generic-limit results that the paper extends from smooth to terminal threefold germs.","marker":"[10]"},{"why":"Gives the ACC for canonical thresholds on terminal threefolds that terminates the model construction in Proposition 5.1.","marker":"[5]"},{"why":"Supplies the numerical classification of threefold divisorial contractions in Table 1 used to prove the strict descent of $m_i$ and $l_i$.","marker":"[9]"},{"why":"Establishes the general-elephant theorem for threefold divisorial contractions used in Lemmas 7.5 and 7.6.","marker":"[8]"},{"why":"Introduces the invariant $l_i$ as the length of a cokernel, the decreasing quantity that forces termination.","marker":"[23]"},{"why":"Defines log canonical thresholds on varieties with bounded singularities and the sheaf of special differentials underlying $l_i$.","marker":"[2]"},{"why":"Proves the ACC for log canonical thresholds on smooth varieties, a building block of the generic-limit argument.","marker":"[1]"},{"why":"Yields the uniform boundedness of the Cartier index on log terminal varieties used in the reduction to Theorem 3.1.","marker":"[3]"}],"fun_headline_variants":["ACC proven for minimal log discrepancies on threefolds","Threefold mlds satisfy ACC: fixed variety case settled","ACC for mlds on any fixed threefold: proof via boundedness","Minimal log discrepancies on fixed threefolds satisfy ACC","Dimension three: ACC for minimal log discrepancies holds"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument relies on a complete numerical classification of threefold divisorial contractions; if a contraction of a terminal threefold not listed in Table 1 exists, the descent of the invariant that forces a smooth point could fail, and the bound proving the ACC would not follow.","fun_headline_variants_meta":{"raw":{"variants":["ACC proven for minimal log discrepancies on threefolds","Threefold mlds satisfy ACC: fixed variety case settled","ACC for mlds on any fixed threefold: proof via boundedness","Minimal log discrepancies on fixed threefolds satisfy ACC","Dimension three: ACC for minimal log discrepancies holds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000363,"raw_usage":{"total_tokens":1853,"prompt_tokens":738,"completion_tokens":1115,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":354,"completion_tokens_details":{"reasoning_tokens":1032}},"tokens_in":354,"tokens_out":1115,"duration_ms":9990,"temperature":1.0,"reasoning_tokens":1032,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T22:51:29.260803+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete disproof would be a fixed normal threefold $X$ and a DCC set $I$ for which the values $\\mathrm{mld}_x(X,\\Delta_i)$ at a closed point form an infinite strictly increasing sequence, or, failing that, an infinite sequence of divisorial contractions from Proposition 5.2 with constant $m_i$ and non-decreasing $l_i$ at singular Gorenstein points, which would undermine Theorem 5.3 and hence the bound $l$.","supporting_citations":[{"cited_title":"Minimal log discrepancies on smooth threefolds","cited_arxiv_id":"2312.13599","evidence_quote":"Settles the smooth-threefold case and supplies the final uniform bound on the model that is smooth at the relevant point."},{"cited_title":"Kawakita","cited_arxiv_id":null,"evidence_quote":"Provides the connectedness and generic-limit results that the paper extends from smooth to terminal threefold germs."},{"cited_title":"Kawakita","cited_arxiv_id":null,"evidence_quote":"Supplies the numerical classification of threefold divisorial contractions in Table 1 used to prove the strict descent of $m_i$ and $l_i$."},{"cited_title":"Kawakita","cited_arxiv_id":null,"evidence_quote":"Establishes the general-elephant theorem for threefold divisorial contractions used in Lemmas 7.5 and 7.6."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the invariant $l_i$ as the length of a cokernel, the decreasing quantity that forces termination."},{"cited_title":"de Fernex, L","cited_arxiv_id":null,"evidence_quote":"Defines log canonical thresholds on varieties with bounded singularities and the sheaf of special differentials underlying $l_i$."},{"cited_title":"de Fernex, L","cited_arxiv_id":null,"evidence_quote":"Proves the ACC for log canonical thresholds on smooth varieties, a building block of the generic-limit argument."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Yields the uniform boundedness of the Cartier index on log terminal varieties used in the reduction to Theorem 3.1."}],"review_version":1}