{"id":"5d9f8e4f-9e08-4e2a-ab22-6c9e5aba0adf","arxiv_id":"2607.06717","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"DAH is a birationally invariant persistence module for analytic pairs whose minimum gives half the real log canonical threshold and whose critical groups on normal surface germs equal the homology of critical weighted dual graphs.","lead":"A new invariant called Divisorial Asymptotic Homology (DAH) attaches a persistence module—a family of homology groups indexed by a scale—to any analytic singularity, recovering the classical real log canonical threshold as its first critical level. The paper proves tameness, functoriality, and for normal surface singularities identifies critical DAH groups with the homology of a weighted dual graph, claiming a topological refinement of numerical invariants.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Localization estimate in Lemma 4.3/Prop. 4.4 assumes a_E=0; for a_E>0 the homotopy cylinder bound fails, so Theorem D is not proven.","rationale":"The reader's weakest assumption identifies precisely the estimate that I find most load-bearing: the localization step in the surface rigidity theorem. The paper's central claim—Theorem D and its consequences in the examples—depends on Prop. 4.4, which in turn uses the radial contraction argument of Lemma 4.3. The proof's integral bound ∫_{t_0}^1 t^{-(a_E+1)} dt ≲ |log ε| is correct only for a_E = 0; for a_E > 0 the integral is a negative power of ε, changing the cylinder mass exponent. This is not a disagreement with convention or consensus but an internal estimate that does not follow from the displayed integrand. The same issue appears in Lemma 4.3 and Prop. 4.4, and it affects the claimed null-homology of chains near noncritical divisors and the homotopy equivalence C^{γ_i}_•(U^{(i)}) ≃ C_•(Exc^{(i)}). I also note the paper's other flagged gaps—functoriality is asserted rather than proved in detail, and Theorem E's promised explicit pair is only described—but these are secondary because the rigidity theorem is the source of the explicit DAH computations. In good faith, the concern is a proof gap rather than a demonstrated falsehood: a corrected estimate or a modified contraction could in principle restore Theorem D. Since the reader already assigned CONDITIONAL, my independent read does not move the verdict; hence UNCHANGED.","tokens_in":46984,"tokens_out":6249,"duration_ms":58990,"concrete_test":"Recompute the homotopy-cylinder estimate in Lemma 4.3 for the explicit pair X = C^2, I = (x^2,y^2), using the two-blow-up resolution. Let E be the first exceptional divisor (ν=2, a=1, γ=1/2) and F the second (ν=2, a=2, γ=3/4), and take γ_i = 1/2. For a compact subanalytic 1-cycle c whose strict transform meets F transversely, compute H_2(|r(c)| ∩ U_ε) in resolution coordinates where K_I∘π ≍ |u|^4 and |det Dπ| ≍ |u|^2 near F. If the cylinder mass is ≍ ε^{1/4} (as the corrected integral predicts) rather than O(ε^{1/2}), the contraction in Lemma 4.3 is not γ_i-admissible and the proof of DAH^{(1)}_k ≅ H_k(Exc^{(1)}) fails for this example. Then check whether an alternative admissible null-homology exists; if not, Theorem D and the examples built on it are not justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing concern is the radial-contraction estimate used to prove the surface rigidity theorem. Lemma 4.3 and Proposition 4.4 claim that, for a divisor E with γ_E > γ_i (Lemma 4.3) or γ_E ≥ γ_i (Prop. 4.4), the homotopy cylinder of the contraction r_t(u,w)=(tu,w) is γ_i-admissible. The proof bounds the cylinder measure by roughly ε^{γ_E} ∫_{t_0}^1 t^{-(a_E+1)} dt with t_0 = ε^{1/(2ν_E)}, and asserts this integral is ≲ |log ε|. That assertion is valid only when a_E = 0. For a_E > 0, ∫_{t_0}^1 t^{-(a_E+1)} dt ≍ t_0^{-a_E} = ε^{-a_E/(2ν_E)}, so the cylinder mass has exponent 1/(2ν_E), not γ_E. The desired bound O(ε^{γ_i}) then holds only if 1/(2ν_E) ≥ γ_i, or a_E = 0. The complementary case 1/(2ν_E) < γ_i < γ_E is compatible with the hypotheses of Lemma 4.3 and is real: for (X,I) = (C^2,(x^2,y^2)), after two blow-ups the second exceptional divisor has ν=2, a=2, γ=3/4, while a critical divisor below has γ_i=1/2; the corrected cylinder exponent is 1/4, not ≥1/2. Thus the null-homology argument for chains near noncritical divisors is not established, and the localization isomorphism DAH^{(i)}_k ≅ H_k(Exc^{(i)}) in Prop. 4.4 fails to follow. Since Theorem 4.7 / Theorem D and all Example computations depend on Prop. 4.4, this is a genuine gap in the paper's central rigidity claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces Divisorial Asymptotic Homology (DAH), a persistence module associated to an analytic pair (X,I). Chains are assigned a divisorial admissibility threshold δ via their intersection with exceptional divisors of a log-resolution, and the sublevel filtration of the intrinsic energy K_I yields a persistence module whose minimum is (1/2) rlct(I) and whose jump locus is a finite homological spectrum Γ_H. The main results are: tameness (Theorem 3.32), functoriality (Prop. 3.38), relative and Mayer–Vietoris sequences (Thm. 3.39, 3.40), birational invariance (Thm. 3.47), a surface rigidity theorem identifying critical DAH groups with homology of critical dual graphs (Thm. 4.7 / Theorem D), and an example (Theorem E) claiming two pairs with equal Γ_H but non-isomorphic DAH. The paper situates DAH relative to metric and vanishing homology.","tokens_in":47519,"tokens_out":2809,"duration_ms":21327,"significance":"If the main theorems hold, DAH is an original and potentially useful invariant: it upgrades the RLCT from a single number to a persistence module, proves a surface-level rigidity theorem with a clean combinatorial output, and the paper provides multiple checkable examples. Strengths include a purely valuative definition of δ, a proof of resolution-independence of δ via valuation-theoretic arguments (Lemmas 3.41–3.45, Theorem 3.47), and a clearly stated finiteness/tameness result (Prop. 3.22, Thm. 3.32). The claimed comparison with metric homology (Prop. 5.8) is honest about its conditional hypotheses.","major_comments":[{"comment":"The localization step contains a load-bearing estimate that appears incorrect. Lemma 4.3 and Prop. 4.4 bound the homotopy cylinder of r_t(u,w)=(tu,w) by roughly ε^{γ_E} ∫_{t_0}^1 t^{-(a_E+1)} dt with t_0 = ε^{1/(2ν_E)}, and the text asserts this integral is O(|log ε|). This is only true for a_E = 0. For a_E > 0 the integral is ≍ t_0^{-a_E} = ε^{-a_E/(2ν_E)}, so the cylinder exponent is 1/(2ν_E), not γ_E. The condition 1/(2ν_E) ≥ γ_i is not implied by the hypotheses. Thus the null-homology argument for chains near noncritical divisors in Lemma 4.3 is not established, and the isomorphism DAH^{(i)}_k ≅ H_k(Exc^{(i)}) in Prop. 4.4 does not follow. Since Theorem 4.7 / Theorem D and all examples in §4.5 depend on this result, the surface rigidity theorem is not proven as written.","section":"Lemma 4.3 / Proposition 4.4"},{"comment":"Functoriality is asserted with the proof: 'Since f is a morphism of analytic pairs, every divisor meeting g f# c also meets e c. Hence δ(f#c) ≥ δ(c).' For a general ideal morphism (only f*J ⊆ I and the energy condition), this incidence claim is not justified. The strict transform behavior under non-submersive maps is delicate; a divisor meeting the push-forward of c need not lift to a divisor meeting e c. This unproved incidence assertion is load-bearing for Theorem B and for the relative/Mayer–Vietoris sequences.","section":"Proposition 3.38"},{"comment":"The examples in §4.5 are all computed under the a priori identification DAH ≅ H_*(Γ(α)) of Theorem 4.7, so they do not provide independent evidence for the rigidity theorem. Theorem E is stated as a result about pairs with identical Γ_H but non-isomorphic DAH; in the main text I find a description of the pairs only through the critical weighted dual graphs being non-isomorphic. Given that Prop. 4.4 is the only bridge from DAH to these graphs, the theorem inherits the gap in the localization estimate.","section":"Section 4.5.3 and Theorem E"}],"minor_comments":[{"comment":"The remark correctly notes that interval decomposition is unavailable over GrAb, but the paper still speaks of 'the persistence barcode' in §4.5. Either restrict to field coefficients or explain what 'barcode' means for the group-valued module.","section":"Remark 1.6"},{"comment":"The proof of resolution-independence via K_I^{(1)} ≍ K_I^{(2)} is terse; the line 'because the exponents in E_ρ are non-negative and bounded above by the discrepancies of ρ' is unclear and should be expanded.","section":"Proposition 3.25"},{"comment":"After the displayed K_I ∘ π = ∏_{j=1}^k |u_j|^{2ν_{i_j}}, k ∈ {1,2}, the text says this holds 'up to a nowhere-vanishing analytic unit'; the unit is not tracked in the subsequent estimates. This is cosmetic because the unit is bounded, but should be stated.","section":"Section 4.1, displayed normal form"},{"comment":"The tameness statement is stated for the module over R_{>0}; the proof shows local constancy of each chain group as a function of α. It would help to state explicitly that the module has finite critical values and is isomorphic to a finite-indexed module.","section":"Theorem 3.32"}],"recommendation":"major_revision","confidential_remarks":"The central conceptual framing is attractive and the valuative construction is mostly clean. However, the surface rigidity theorem, which is the paper's main concrete payoff, rests on the homotopy-cylinder estimate in Lemma 4.3/Prop. 4.4; the reader's stress-test concern reproduces a genuine numerical gap for a_E>0. The authors should either fix the estimate (possibly by restricting to a_E=0, which is not the generic case) or replace the localization argument. A second, less central but still load-bearing issue is the unproved incidence claim in Prop. 3.38. I recommend major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take on Grulha's DAH paper. The basic idea is good: attach to an analytic pair (X,I) the persistence module of compact subanalytic chains filtered by the divisorial admissibility threshold δ(c) = min γ_E over divisors meeting the total transform. This is a natural construction, it is genuinely valuative, the tameness theorem (finite spectrum ⇒ locally constant filtration) is correct and simple, and the examples showing that Γ_H alone misses gluing data are convincing. The RLCT appears as the minimum of Γ_H, which is just the classical resolution formula read backwards; fine.\n\nThe problems are in the main rigidity claims. Lemma 4.3 and Proposition 4.4, which are needed for Theorem D, bound the radial-contraction homotopy cylinder by ∫_{t0}^1 t^{-(a_E+1)} dt · ε^{γ_E} and then assert this is O(|log ε|). That is only true when a_E = 0. For a_E > 0 the integral is ≍ ε^{-a_E/(2ν_E)}, so the cylinder measure has exponent 1/(2ν_E), not γ_i. The stress-test example (C^2, (x^2,y^2)) is real: the second exceptional divisor has (a,ν)=(2,2) and γ=3/4, while a critical divisor below has γ_i=1/2; the corrected cylinder exponent is 1/4, so the cylinder is not γ_i-admissible. Since Theorem D and all surface examples depend on this localization, the central rigidity statement is not proven as written.\n\nThere is also an unsupported incidence claim in the functoriality proof (Prop. 3.38): 'every divisor meeting f_#c also meets ec' is exactly what needs proof and does not follow from f^{-1}(V(I_Y)) ⊆ V(I_X). And Theorem E, the claim that Γ_H-equal pairs can have non-isomorphic DAH, is asserted but no explicit construction appears in the paper; the cusp example shows a pair with H_1 in DAH, but not two pairs with identical spectra.\n\nNone of this looks like a fatal blow to the framework — the definitions are solid, tameness is fine, and the rigidity gap may be repairable with a different homotopy or a weaker statement. But the paper in its current form overclaims: Theorem D and Theorem E should be labelled conjectural or proved. Worth a serious referee, but only with major revision.","headline":"Ambitious, cleanly defined persistence invariant for analytic pairs, but the surface rigidity theorem rests on a wrong estimate and Theorem E is not actually proved.","tokens_in":47936,"tokens_out":2647,"would_cite":false,"duration_ms":24070,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["32S05","32S45","14E15","14B05","55N31"],"pacs":[],"model":"deepseek-v4-flash","headline":"An analytic pair carries a canonical persistence module — divisorial asymptotic homology — whose minimum is the real log canonical threshold and whose full structure records how exceptional divisors glue; for normal surface germs its critic","keywords":["divisorial asymptotic homology","real log canonical threshold","homological spectrum","persistence module","analytic pair","critical weighted dual graph","birational invariance","normal surface singularity"],"falsifier":"For a normal surface germ with a critical divisor E having a_E>0, compute the homotopy-cylinder measure used in Lemma 4.3: H_{k+1}(|r(c)|∩U_ε) ≈ ε^{γ_E}∫_{ε^{1/(2ν_E)}}^1 t^{-(a_E+1)}dt. If for a_E>0 this quantity is not O(ε^{γ_i}), the claimed isomorphism DAH^{(i)}_k ≅ H_k(Γ(i)) fails at that level; a concrete check is to compare the resulting DAH groups for an A_n or cusp pair whose critical divisor has positive discrepancy against the homology of its critical dual graph.","tokens_in":46908,"feed_emoji":"🕸️","tokens_out":4682,"duration_ms":43046,"temperature":0.7,"pith_summary":"This paper tries to establish that numerical invariants of singularities, especially the real log canonical threshold, are only the first visible layer of a deeper topological structure. It constructs divisorial asymptotic homology (DAH), a persistence module attached canonically to an analytic pair through the sublevel sets of the intrinsic energy. Each chain receives a valuative admissibility threshold, and the resulting module has the RLCT as its minimum and a finite homological spectrum as its jump locus. For normal surface germs, the critical DAH groups are identified with the homology of critical weighted dual graphs, and examples show pairs with identical homological spectra but non-isomorphic DAH. A sympathetic reader should take away that the gluing of exceptional divisors, information discarded by exponents and spectra, is recoverable by a persistence theory whose filtration is determined by birational geometry rather than by a user-chosen scale.","feed_headline":"A homology theory recovers the topology lost by the RLCT","feed_subtitle":"Divisorial asymptotic homology turns the real log canonical threshold into the first level of a canonical persistence module.","key_machinery":"The central object is the admissibility threshold δ(c), defined purely valuatively as the infimum of divisorial exponents γ(v)=(a(v)+1)/(2ν(v)) over divisorial valuations reaching the chain, equivalently the minimum γ_E over exceptional divisors meeting the total transform on any log-resolution. This threshold filters the chain complex into α-admissible subcomplexes, whose homology defines the persistence module P(X,I). The divisorial exponents themselves, extracted from the order ν_E and Jacobian discrepancy a_E of each exceptional divisor, together with the weighted monomial integral estimates producing ε^{γ}|log ε|^{m−1}, carry the analytic content; in dimension two the critical weighted","core_discovery":"The paper claims that for an analytic pair the filtration by sublevel sets of the energy induces a divisorial persistence module whose critical values are exactly the divisorial levels γ_E=(a_E+1)/(2ν_E); the RLCT is its minimum, the homological spectrum is its jump locus, and the whole module is birationally invariant. In dimension two, the critical DAH groups are canonically and resolution-independently isomorphic to the homology of the critical weighted dual graph, making DAH an explicit combinatorial invariant of normal surface germs. The paper further claims that this persistent structure is strictly finer than the RLCT and the homological spectrum, exhibiting pairs with equal homologic","pith_inferences":["If DAH is as robust as claimed, singular learning theory could be stratified beyond the learning coefficient: the full homological spectrum and the birth times of higher-degree classes (as in the cusp example, where H_1 is born only at the last critical level) suggest testable refinements of asymptotic model selection.","The surface rigidity result points toward a higher-dimensional extension in which critical DAH groups are homology groups of a critical dual complex recording intersections of all orders; the paper notes this extension is not developed, making it a concrete open problem.","The comparison theorems with metric and vanishing homology are explicitly conditional: the metric-homology identification assumes normal embedding plus an asymptotic expansion, and the vanishing-homology correspondence is stated as a conjecture requiring projective-thinness control and flat-limit stability. A careful reader should treat those comparisons as programmatic, not established.","The most fragile step is the localization argument for surface rigidity: the radial contraction bound in Lemma 4.3 and Proposition 4.4 may require a_E=0; if the estimate fails for positive discrepancies, Theorem D would need an additional hypothesis on critical divisors."],"forward_implications":["The RLCT becomes the first level of a canonical hierarchy, with min Γ_H(X,I)=1/2 rlct(I).","DAH is tame: the persistence module changes only at the finitely many critical values of the homological spectrum.","For normal surface germs, the persistence module reduces to the homology of critical weighted dual graphs, making it computable and resolution-independent.","DAH is strictly finer than the homological spectrum and the RLCT: pairs with identical spectral data can have non-isomorphic DAH groups.","DAH is functorial and admits relative and Mayer–Vietoris long exact sequences, so it behaves like a homology theory rather than a single numerical invariant."],"fun_headline_variants":["Divisorial homology captures lost topology of analytic pairs","Persistence module refines the real log canonical threshold","Homological spectrum outdoes RLCT via divisorial persistence","Topology of exceptional divisors now a homology invariant","New invariant: DAH recovers topology from RLCT"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that admissibility survives localization to the critical exceptional set: the radial contraction homotopy in Lemma 4.3 and Proposition 4.4 is claimed to stay γ_i-admissible via the estimate ∫_{t_0}^1 t^{-(a_E+1)}dt ≲ |log ε|, but this bound appears to hold only for a_E=0; for a_E>0 the integral contributes ε^{-a_E/(2ν_E)}, so without an added hypothesis the isomorphism DAH^{(i)}_k ≅ H_k(Exc^{(i)}) is not established. The rigidity theorem also assum","fun_headline_variants_meta":{"raw":{"variants":["Divisorial homology captures lost topology of analytic pairs","Persistence module refines the real log canonical threshold","Homological spectrum outdoes RLCT via divisorial persistence","Topology of exceptional divisors now a homology invariant","New invariant: DAH recovers topology from RLCT"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000184,"raw_usage":{"total_tokens":1140,"prompt_tokens":712,"completion_tokens":428,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":456,"completion_tokens_details":{"reasoning_tokens":351}},"tokens_in":456,"tokens_out":428,"duration_ms":4199,"temperature":1.0,"reasoning_tokens":351,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T04:25:06.720668+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a normal surface germ with a critical divisor E having a_E>0, compute the homotopy-cylinder measure used in Lemma 4.3: H_{k+1}(|r(c)|∩U_ε) ≈ ε^{γ_E}∫_{ε^{1/(2ν_E)}}^1 t^{-(a_E+1)}dt. If for a_E>0 this quantity is not O(ε^{γ_i}), the claimed isomorphism DAH^{(i)}_k ≅ H_k(Γ(i)) fails at that level; a concrete check is to compare the resulting DAH groups for an A_n or cusp pair whose critical divisor has positive discrepancy against the homology of its critical dual graph.","supporting_citations":[],"review_version":3}