{"id":"569b7555-3eeb-492b-b72b-bc102bdc3460","arxiv_id":"2505.14939","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The Gromov-Hausdorff limit of collapsing Calabi-Yau metrics is homeomorphic to the base variety, and the singular set has Hausdorff codimension at least two.","lead":"Calabi-Yau metrics that collapse along a fibration have a Gromov-Hausdorff limit homeomorphic to the base variety, with the singular discriminant locus of Hausdorff codimension at least two. The paper proves the general form of Tosatti's conjecture, replacing the special cases known before.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The RCD(0,2n) upgrade in §3 depends on an unproved transfer from Cheeger-Colding [5] eigenfunctions on the GH limit to W^{1,2} eigenfunctions on the incomplete smooth locus (X^circ, ω_can); Proposition 13 is stated without proof.","rationale":"The reader's weakest_assumption identifies exactly the same mechanism: the RCD(0,2n) property needed for Theorem 3 is obtained via Proposition 13 from Lipschitz eigenfunctions, and the transfer from the GH limit back to (X^circ,ω_can) is not proved. My reading of the full text confirms this is the single most load-bearing unproved step. The paper's other dependencies are heavy (preprints [8], [22], [36], [38]), but those are at least explicitly cited; Proposition 13 is a key structural input stated without proof, and the one-line appeal to Cheeger-Colding [5] is not automatic in a collapsing, dimension-changing limit. I do not claim the theorem is wrong: the intended argument is plausible, since X\\X^circ has real codimension at least 2 and should have zero W^{1,2}-capacity, making the Friedrichs and Neumann Laplacians coincide. But this capacity argument is nontrivial and absent, and the quantitative Lipschitz bounds needed for Honda's RCD criterion are not shown to follow from [5] in this collapsed setting. Because the gap is specific, localizable to the proof of Theorem 2, and potentially fixable, the appropriate verdict is CONDITIONAL rather than ACCEPT or REJECT: accept only if the missing transfer is supplied. The paper deserves credit for the substantial new technical machinery in Propositions 8, 9, 10, and 12, and Proposition 14's measure identification is carefully argued; the concern is confined to the final RCD upgrade, but that upgrade is essential for the headline result.","tokens_in":18226,"tokens_out":19125,"duration_ms":174636,"concrete_test":"Write out the proof of Proposition 13 following [38, Proposition 9] in the present setting, and identify the precise Dirichlet form whose eigenfunctions satisfy the Lipschitz hypothesis. Then verify against Cheeger-Colding [5, Theorem 7.3 and surrounding spectral results] whether that theorem applies to the measured GH limit (Z,d_Z,ν) from Proposition 14, and prove the identification W^{1,2}(Z,ν) = W^{1,2}(X^circ,ω_can) using the zero W^{1,2}-capacity of X\\X^circ (real codimension at least 2) together with ν = v ω_can^n. If the capacity argument cannot be completed, or if [5] only controls eigenfunctions of a different Laplacian, then Proposition 13's hypothesis is not established and the RCD(0,2n) conclusion in Theorem 2 fails.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Theorem 2 is reduced to Theorem 3 by proving that the metric completion (X-hat, d_Z, ω_can^n) is RCD(0,2n). The proof in Section 3 does this in two sentences: Proposition 14 identifies the renormalized limit measure with ω_can^n, then \"It follows by Cheeger-Colding [5] that the W^{1,2} eigenfunctions of the Laplacian on (X^circ, ω_can) are Lipschitz, and therefore by Proposition 13 the space (Z,d_Z,ω_can^n) is an RCD(0,2n)-space.\" The load-bearing step is Proposition 13, which is asserted exactly as in [38, Proposition 9] with no proof in the present text. More importantly, the premise of Proposition 13 concerns eigenfunctions on the incomplete smooth locus (X^circ, ω_can), whereas Cheeger-Colding [5] supplies Lipschitz regularity for eigenfunctions of the Dirichlet form on the measured Gromov-Hausdorff limit (Z,d_Z,ν). No argument is given that these two spectral problems coincide: there is no identification of W^{1,2}(Z,ν) with W^{1,2}(X^circ,ω_can), no proof that the singular sets X\\X^circ have zero W^{1,2}-capacity, and no check that the eigenfunctions controlled by [5] exhaust the spectrum of the Laplacian on (X^circ,ω_can) used in Proposition 13. Since the whole RCD bridge to Theorem 3 rests on this transfer, a gap here invalidates the proof of Theorem 2 even if the rest of the analysis is correct. The concern is not that the claim is false, but that a central, non-obvious analytic step is unproved in the manuscript.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies Ricci-flat Kähler metrics on a projective Calabi-Yau manifold M as their Kähler classes degenerate to the semiample class c_1(L) of a fibration f: M -> X. The main result, Theorem 2, states that the Gromov-Hausdorff limit (Z,d_Z) of the collapsing metrics is homeomorphic to the base X, with Hausdorff dimension bounds dim_H(X\\X_reg) <= 2n-4 and dim_H(X\\X^circ) <= 2n-2, where n = dim_C X. This is presented as resolving conjectures of Tosatti. The proof reduces Theorem 2 to a general statement, Theorem 3, about singular Kähler spaces (X,ω) whose metric completion is an RCD(K,2n)-space. The main analytic work in Section 2 proves the homeomorphism and dimension bounds under the RCD hypothesis, using a new approximation argument (Proposition 9) and tools from the author's prior work and Donaldson-Sun/Liu-Szekelyhidi. Section 3 then attempts to establish the RCD(0,2n) condition for the collapsing Calabi-Yau setting via the limit measure and spectral arguments, culminating in Proposition 13 and the proof of Theorem 2.","tokens_in":18587,"tokens_out":4269,"duration_ms":39538,"significance":"If correct, this result would resolve a long-standing conjecture of Tosatti and complete the program begun by Song-Tian-Zhang on collapsing Calabi-Yau fibrations. The dimension bounds and the homeomorphism statement are sharp and expected. The paper introduces a potentially useful new technique, Proposition 9, for handling points where the metric is regular in tangent cones but the underlying complex structure is singular along a divisor. The reliance on prior work, especially the author's own framework, is substantial but the chain of citations is credible. However, the central RCD bridge in Section 3 is currently supported by an unproved transfer statement, which is a load-bearing gap that must be fixed before the result can be considered fully established.","major_comments":[{"comment":"","section":"Section 3, Proof of Theorem 2"},{"comment":"","section":"Section 3, Proposition 13"},{"comment":"","section":"Section 3, Proposition 14 and the identification of ν with ω_can^n"}],"minor_comments":[{"comment":"","section":"Section 2, Proposition 7"},{"comment":"","section":"Section 2, around equation (9)"},{"comment":"","section":"Throughout"},{"comment":"","section":"Section 2, Proposition 12"}],"recommendation":"major_revision","confidential_remarks":"The paper is likely correct and the overall strategy is compelling. The main issue is the unproved spectral transfer in Section 3, which is exactly the kind of step that a referee must see in detail. The manuscript should not be rejected, but it needs a major revision to either prove the transfer or cite a theorem that directly covers it. The heavy self-citation is acceptable because the cited works are directly relevant, but the authors should ensure that every load-bearing assertion is backed by a proof or a precise reference."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Main thing: this paper is a serious attempt at Tosatti's conjecture, and it is the right kind of serious. The new content is real: Proposition 9 handles metric-regular points on the divisor, which was the obstacle, and the dimension bounds for X\\X_reg and X\\X^circ are genuinely new. The reduction of Theorem 3 to the RCD framework is coherent, and the proofs in Section 2 are detailed. The paper earns its keep.\n\nBut there is a load-bearing soft spot. Section 3's only job is to show that the metric completion (X-hat,d_Z,omega_can^n) is RCD(0,2n). That is done via Proposition 13, which is asserted verbatim from the author's own [38, Prop. 9] with no proof here. More to the point, the hypothesis of Prop. 13 concerns W^{1,2} eigenfunctions on the incomplete smooth locus (X^circ,omega_can). The paper says these are Lipschitz \"by Cheeger-Colding [5]\" because the limit space is RCD. But [5] gives Lipschitz eigenfunctions on the GH limit, not on the incomplete smooth locus. No identification of the two Sobolev spaces is given, no capacity argument for X\\X^circ, no check that the spectra match. This is not a trivial technicality: the whole RCD bridge to Theorem 3 passes through it. If that transfer fails, Theorem 2 is not proved, even if the rest of the analysis is correct.\n\nI should say the stress-test note is right about this. The reader's report flagged it as a weakness, but I think it is bigger than that: it is the one place where the paper states a non-obvious analytic fact without proof, and everything depends on it. It is not a question of honesty or fitting; the author is straight about the dependence. It is a question of whether the chain holds. If Prop. 13 is false, the main theorem does not follow; if it is true, it needs a proof in this paper or a precise reference that includes the transfer.\n\nEverything else is proportionately solid. The regularization in Proposition 9 is written out in detail and looks plausible. The dimension estimates for X\\X_reg and X\\X^circ are clever, especially the use of holomorphic functions to get codimension 2. The citation pattern is heavy on the author's own prior work, but those are independent results, not circular.\n\nRecommendation: send to peer review. It is a major claim, and the referee should demand a complete proof of Proposition 13, or a citation that actually contains the transfer. If that comes back, the paper is likely right. As it stands, Theorem 2 is conditional on an unproved lemma.","headline":"Proves the Tosatti conjecture modulo an unproved analytic transfer in Section 3; Proposition 13 needs a real proof before Theorem 2 is fully established.","tokens_in":19138,"tokens_out":2438,"would_cite":true,"duration_ms":21254,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C23","32Q25"],"pacs":[],"model":"deepseek-v4-flash","headline":"Calabi-Yau collapse limit is the base, singular locus codimension 2","keywords":["Gromov-Hausdorff convergence","Calabi-Yau metrics","collapsing limits","RCD spaces","semiample fibrations","discriminant locus","Hausdorff dimension","singular Kähler metrics"],"falsifier":"A reader could settle the weakest step by taking an explicit collapsing fibration, such as an elliptic fibration over a singular surface, and checking whether the $W^{1,2}$ eigenfunctions on $(X^\\circ,\\omega_{\\mathrm{can}})$ are Lipschitz; if a non-Lipschitz eigenfunction appears, the RCD bridge to Theorem 2 breaks. A more direct falsifier would be any semiample fibration whose Gromov-Hausdorff limit has Hausdorff dimension of $X\\setminus X^\\circ$ larger than $2n-2$ or is not homeomorphic to $X$.","tokens_in":17994,"feed_emoji":"📐","tokens_out":12284,"duration_ms":100425,"temperature":0.7,"pith_summary":"This paper resolves a long-standing conjecture about collapsing Calabi-Yau metrics: when the Kähler class degenerates to a semiample fibration class, the Ricci-flat metrics collapse in the Gromov-Hausdorff sense to the base variety of the fibration. The paper proves that the limit is homeomorphic to that base, and that the discriminant locus, where the fibration is singular, has Hausdorff codimension at least 2. This proves the statement in full generality, without smoothness, simple-normal-crossing, or low-dimension assumptions on the base. A curious reader should care because it settles what the collapsed geometry of Ricci-flat spaces actually is, and it gives a general mechanism for identifying such limits.","feed_headline":"Calabi-Yau collapse limit is the base, singular locus codimension 2","feed_subtitle":"Proof that collapsing Ricci-flat metrics converge to the fibration base, with singularity dimension bounds.","key_machinery":"The central object is the metric completion $\\hat X$ of $(X^\\circ,\\omega_{\\mathrm{can}})$ equipped with the measure $\\omega_{\\mathrm{can}}^n$. The load-bearing mechanism is the RCD$(K,2n)$ condition, a synthetic Riemannian Ricci-curvature lower bound for metric measure spaces; the paper transfers this structure from the Ricci-flat manifolds $(M,\\omega_t)$ to $\\hat X$ via measured Gromov-Hausdorff convergence. Once RCD is available, the proof uses a strict positivity result to approximate the singular metric $\\omega$ locally by smooth Kähler metrics with Ricci curvature bounded below, and then uses a theorem on Gromov-Hausdorff limits of such metrics (Proposition 9) to produce holomorphic charts around almost regular points. These charts separate points of $\\hat X$ and control which tangent cones can occur, yielding both the homeomorphism and the dimension bounds.","core_discovery":"The central claim is Theorem 2: the Gromov-Hausdorff limit $(Z,d_Z)$ of a collapsing family of Calabi-Yau metrics is homeomorphic to the base $X$ of the fibration, and the Hausdorff dimensions of the singular loci satisfy $\\dim_H(X\\setminus X_{\\mathrm{reg}})\\le 2n-4$ and $\\dim_H(X\\setminus X^\\circ)\\le 2n-2$, where $n=\\dim_{\\mathbb C}X$. The proof passes through a more general statement, Theorem 3: for any singular Kähler space $X$ satisfying three analytic conditions on its smooth open set, if the metric completion $\\hat X$ is an RCD$(K,2n)$ space, then $\\hat X$ is homeomorphic to $X$ and the same dimension estimates hold. The paper then shows the collapsing Calabi-Yau setting satisfies these hypotheses by proving the completion of $(X^\\circ,\\omega_{\\mathrm{can}})$ is an RCD$(0,2n)$-space, and it obtains the same conclusions for the canonical-model and continuity-method setting in Theorem 15.","pith_inferences":["Beyond the paper: the RCD hypothesis in Theorem 3 may be removable; the paper states this as its Conjecture 4 but does not prove it.","Beyond the paper: if the RCD bridge were replaced by a direct argument for Lipschitz eigenfunctions, the homeomorphism statement would follow without passing through measured Gromov-Hausdorff convergence of the collapsing family.","Beyond the paper: explicit examples with singular base strata should realize the dimension bounds, though the paper does not compute such examples, providing a test of sharpness."],"forward_implications":["The long-standing collapse-limit conjecture is true in full generality: the Gromov-Hausdorff limit is the base variety itself, not an exotic quotient.","The same homeomorphism and dimension conclusions hold for any singular Kähler space satisfying the paper's three conditions together with the RCD assumption.","The dimension bounds describe the singular set precisely: $X\\setminus X_{\\mathrm{reg}}$ has Hausdorff dimension at most $2n-4$ and $X\\setminus X^\\circ$ at most $2n-2$ in the limit metric.","The canonical-model and continuity-method setting (Theorem 15) inherits the same conclusions, so the result applies beyond Ricci-flat collapsing to Kähler-Ricci flow degenerations."],"supporting_citations":[{"why":"Supplies the renormalized limit measure on the Gromov-Hausdorff limit used to identify the RCD measure.","marker":"[4]"},{"why":"Used to infer that $W^{1,2}$ eigenfunctions on the smooth locus are Lipschitz from limit-space regularity.","marker":"[5]"},{"why":"Shows the GH limit is the metric completion of $(X^\\circ,\\omega_{\\mathrm{can}})$, the starting point for the homeomorphism proof.","marker":"[36]"},{"why":"Establishes strict positivity of the singular Kähler metric, needed for the smooth approximations with Ricci lower bounds.","marker":"[8]"},{"why":"Provides the construction of holomorphic charts on Gromov-Hausdorff limits of Kähler manifolds with Ricci curvature bounded below, used in Proposition 9.","marker":"[32]"},{"why":"Supplies the earlier strategy for identifying limits of singular Kähler-Einstein metrics via RCD spaces that this paper adapts.","marker":"[38]"},{"why":"Gives Hausdorff dimension estimates for singular sets of noncollapsed RCD spaces, used for the codimension bounds.","marker":"[10]"},{"why":"Provides diameter bounds and measure comparison used in the canonical-model and continuity-method setting (Theorem 15).","marker":"[16]"}],"fun_headline_variants":["Collapsing Calabi-Yau metrics converge to the base","CY collapse: GH limit is base, singular set codim≥2","Gromov-Hausdorff limit of CY fibration collapse equals base","Tosatti conjecture resolved: collapsing CY limit is base","Calabi-Yau collapse limit is base with codim≥2 discriminant"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on the claim that the eigenfunctions used to define the synthetic Ricci-bound structure on the smooth part extend to Lipschitz functions on the whole limit, a transfer from the limit space back to the incomplete smooth part that the paper states follows from standard convergence results but does not write out.","fun_headline_variants_meta":{"raw":{"variants":["Collapsing Calabi-Yau metrics converge to the base","CY collapse: GH limit is base, singular set codim≥2","Gromov-Hausdorff limit of CY fibration collapse equals base","Tosatti conjecture resolved: collapsing CY limit is base","Calabi-Yau collapse limit is base with codim≥2 discriminant"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001061,"raw_usage":{"total_tokens":4394,"prompt_tokens":836,"completion_tokens":3558,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":452,"completion_tokens_details":{"reasoning_tokens":3467}},"tokens_in":452,"tokens_out":3558,"duration_ms":23174,"temperature":1.0,"reasoning_tokens":3467,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T15:27:30.955197+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A reader could settle the weakest step by taking an explicit collapsing fibration, such as an elliptic fibration over a singular surface, and checking whether the $W^{1,2}$ eigenfunctions on $(X^\\circ,\\omega_{\\mathrm{can}})$ are Lipschitz; if a non-Lipschitz eigenfunction appears, the RCD bridge to Theorem 2 breaks. A more direct falsifier would be any semiample fibration whose Gromov-Hausdorff limit has Hausdorff dimension of $X\\setminus X^\\circ$ larger than $2n-2$ or is not homeomorphic to $X$.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the renormalized limit measure on the Gromov-Hausdorff limit used to identify the RCD measure."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Used to infer that $W^{1,2}$ eigenfunctions on the smooth locus are Lipschitz from limit-space regularity."},{"cited_title":"Gromov-Hausdorff limits of K¨ ahler manifolds with Ricci curvature bounded below","cited_arxiv_id":null,"evidence_quote":"Provides the construction of holomorphic charts on Gromov-Hausdorff limits of Kähler manifolds with Ricci curvature bounded below, used in Proposition 9."},{"cited_title":"Non-collapsed spa ces with Ricci curvature bounded from below","cited_arxiv_id":null,"evidence_quote":"Gives Hausdorff dimension estimates for singular sets of noncollapsed RCD spaces, used for the codimension bounds."},{"cited_title":"Geometric estimates for c omplex Monge-Amp` ere equations","cited_arxiv_id":null,"evidence_quote":"Provides diameter bounds and measure comparison used in the canonical-model and continuity-method setting (Theorem 15)."}],"review_version":1}