{"id":"52c6ed1b-afa1-4759-9c7d-df80453b5662","arxiv_id":"2608.13000","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Minimal surjective local homeomorphisms of finite-dimensional compact spaces give Deaconu-Renault groupoids with dynamical comparison, yielding UCT Kirchberg algebras and, in the Cantor case, Matui's AH conjecture.","lead":"A new proof shows that certain many-to-one maps on compact spaces generate groupoids with dynamical comparison, a strong rearrangement property. This makes the associated C*-algebras classifiable Kirchberg algebras and confirms Matui's AH conjecture in the zero-dimensional case.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No concrete flaw detected; the load-bearing issue is the unverified Section 5 thin-boundary proof (Lemma 5.2, Theorem 5.8) underpinning the higher-dimensional main theorem.","rationale":"The reader's weakest assumption correctly identified the load-bearing point: the higher-dimensional Theorem A (Theorem 4.7) depends on the freeness-free thin boundary result (Theorem 2.4 / Corollary 5.9), whose proof in Section 5 is intricate and, per the AI statement, partly AI-assisted. My own reading of Section 5 found no explicit contradiction, but the proof has two under-specified steps that are essential: the refinement in Lemma 5.2 must keep closures inside the chosen bisection sources, and the induction in Theorem 5.8 must apply Lemma 5.5 to a closed set, requiring an invariant that is not stated. Both appear repairable, but until explicitly verified, the conditional verdict is appropriate. I agree with the reader's assessment and recommend no change to the verdict.","tokens_in":50,"tokens_out":64147,"duration_ms":1148284,"concrete_test":"Re-derive Lemma 5.2 and Theorem 5.8 from scratch, explicitly checking: (a) the existence of a refinement P of the cover V such that closure(P)⊂V_P⊂s(B_P) for every P (via [BK04, Lemma 3.2]); (b) the inductive invariant closure(O_k)⊂O_k∪N_k in Theorem 5.8 and the application of Lemma 5.5 to the closed set closure(O_k). A Lean/Coq formalization of these two proofs would also settle the question; if both checks pass, the concern does not land.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The main theorem for non-zero-dimensional unit spaces depends entirely on Theorem 2.4 (Corollary 5.9), the thin boundary property for arbitrary minimal second countable Hausdorff etale groupoids with finite-dimensional compact unit space. That theorem is proved in Section 5 via two long inductions, Lemma 5.2 and Theorem 5.8, which the AI statement discloses were originally ChatGPT drafts. I could not find a definite mathematical error: the collision-set induction in Lemma 5.2 is structurally sound, and the diagonal construction in Theorem 5.8 follows Szabó's argument. The two places where the text is too terse to certify are: (i) Lemma 5.2's refinement step: the proof needs the refined cover P to satisfy closure(P) subset of the bisection source s(B_P), so that K_P = K∩closure(P) is a valid compact set for the C_m collision hypothesis; this is likely what [BK04, Lemma 3.2] gives, but it is not stated and must be checked. (ii) Theorem 5.8's induction: Lemma 5.5 requires a closed set A, but the text says to apply it to O_k (open); the natural fix is A=closure(O_k), which requires closure(O_k)⊂O_k∪N_k at each step; this follows only if the previously chosen N_{k+1}⊂N_k and boundary containment are maintained. These are fixable-looking but not written down. Because Theorem A's higher-dimensional case collapses if either fails, the argument is not yet certifiable as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves dynamical comparison for Deaconu–Renault groupoids of minimal surjective non-injective local homeomorphisms of compact metrizable spaces with finite Lebesgue covering dimension (Theorem A, Theorem 4.7). The strategy is to reduce the general case to local homeomorphisms with fibers of size at least two (Section 3), embed suitable partial actions of free groups with large domains (Section 4), and control the boundary of the embedded partial action using a new groupoid version of the thin boundary property (Theorem D, Corollary 5.9). Consequences include a dynamical proof that the associated C*-algebras are UCT Kirchberg algebras, verification of Matui's AH-conjecture for zero-dimensional Deaconu–Renault groupoids via Li's theorem, and a freeness-free small boundary property for minimal group actions (Corollaries B, C, E).","tokens_in":28238,"tokens_out":16549,"duration_ms":161267,"significance":"If correct, this is a substantial result: it places a large class of purely infinite étale groupoids in the dynamical comparison regime, recovers and extends known C*-algebraic results by Carlsen–Thomsen through purely dynamical methods, and identifies a general freeness-free small-boundary phenomenon. The paper's strengths include a clean reduction in Section 3, a concrete use of paradoxical towers from [GGKN23], and an explicit and honest disclosure of AI assistance. The main theorem, however, is conditional on the intricate Section 5 proof of Theorem D; as written, two technical points in that proof are not fully justified. These points are load-bearing for the higher-dimensional case, so the paper cannot be certified in its present form, although the issues appear repairable.","major_comments":[{"comment":"The proof defines K_P := K ∩ closure(P) and later applies the hypothesis K ∈ C_m to the compact sets K_P, which requires K_P ⊂ s(B_P). However, the refinement obtained from [BK04, Lemma 3.2] is only stated to satisfy P ⊂ s(B_P), not closure(P) ⊂ s(B_P). Since s(B_P) is open, closure(P) can stick out of s(B_P), and then K_P is not contained in the source of the bisection B_P. Please add the missing argument that the cover can be chosen with closure(P) ⊂ s(B_P) for every P, or adjust the construction accordingly. Without this, the C_m collision hypothesis cannot be applied to K_P.","section":"Section 5, Lemma 5.2"},{"comment":"At the induction step, Lemma 5.5 is invoked 'with respect to the inclusion O_k ⊂ O_k ∪ N_k', but Lemma 5.5 requires a closed set A. The natural intended set is A = closure(O_k), and this is admissible only if closure(O_k) ⊂ O_k ∪ N_k; that condition follows from the previously recorded ∂O_k ⊂ N_k, but it is not stated. The same issue recurs in the later closure estimates. Please rewrite this induction step explicitly with A = closure(O_k) and verify the boundary containment at each stage, so that the use of Lemma 5.5 is formally valid.","section":"Section 5, Theorem 5.8"}],"minor_comments":[{"comment":"In the proof of Theorem 4.4, the statement 'we have D_{h_i} = D_{a_1} = D_{a_2} = X' is not justified and appears inconsistent with the definition of the partial action by α_{a_i} = T|_{A_i}; for a positive word h_i of length at least two, the domain D_{h_i} is typically a proper subset of X. Since the zero-dimensional case can be deduced from the general theorem, please correct this step or remove the redundant proof.","section":"Section 4, Theorem 4.4"},{"comment":"In the displayed chain proving G(0) ≺ U, the summation indices are misprinted: the middle unions over i should run over j = 0, ..., n_i, not over i. The intended formula is clear but should be corrected.","section":"Section 3, Lemma 3.4"},{"comment":"The proof should explicitly record at the start that N_1 is chosen with N_1 ⊂ O, and that the inclusion N_{k+1} ⊂ N_k is maintained after refining N_{k+1}. These facts are used later in deriving \\(\\overline{W} \\subset O\\) and ∂W ⊂ N_k, but they are only implicit in the current text.","section":"Section 5, Theorem 5.8"},{"comment":"The main argument depends essentially on the unpublished preprints [Ste26] and [ES19]. Please confirm that the quoted theorems are publicly available in their final form, and consider stating the precise results used from [Ste26, Theorem 2.2] and [ES19, Theorem 9.6, Corollary 9.7] in the text or an appendix.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The two gaps in Section 5 are likely fixable, and I did not find a concrete mathematical error in the rest of the architecture. However, because the higher-dimensional main theorem collapses if Lemma 5.2 or Theorem 5.8 cannot be repaired as indicated, I could not certify the proof as written. The authors' AI disclosure is transparent and not itself a reason for concern. The dependence on [Ste26] and [ES19] should be checked editorially. I would be willing to review a revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, Theorem A is genuinely new: it gives dynamical comparison for Deaconu–Renault groupoids of non-injective local homeomorphisms, the purely infinite regime that previous comparison results did not touch. Second, the proof comes in two very different parts: a clean reduction to the “large fibers” case plus a paradoxical-towers argument (Sections 3–4), and an intricate freeness-free thin-boundary theorem (Section 5) that carries the whole higher-dimensional case. I found no concrete error in either part, but Section 5 is where I would spend referee time.\n\nWhat the paper does well: the reduction in Corollary 3.5 is elegant. Lemma 3.2 (some iterate has fibers of size at least two) and Lemma 3.3 (clopen invariant subset for T^n) are short and convincing. Theorem 4.4 gives a self-contained zero-dimensional proof using de Castro–Steinberg's realization and the paradoxical towers from GGKN23; that part is readable and likely correct. The byproduct, Corollary E (small boundary property for minimal actions without freeness), is a strong result in its own right. The authors are also honest in the AI statement about which proofs started as drafts.\n\nThe soft spot is the thin boundary property. Theorem 2.4/Corollary 5.9 asserts that every minimal second countable etale groupoid with compact finite-dimensional unit space has the thin boundary property, with no freeness assumption. That is a strong claim. The proof in Section 5 requires checking two delicate inductions, Lemma 5.2 and Theorem 5.8. The stress-test note identifies specific terse places: in Lemma 5.2, the refinement via [BK04, Lemma 3.2] needs to give closure(P) contained in s(B_P) so that K_P is compact; in Theorem 5.8, Lemma 5.5 is applied to open sets O_k rather than closed sets, and the natural fix is to use closures, which requires a containment the induction maintains but does not state. These look fixable, but they are not written down, and the higher-dimensional Theorem A collapses if they fail. The zero-dimensional result does not depend on Section 5 and stands on its own.\n\nOther wrinkles: the paper leans on several recent preprints (Ste26, ES19, Li25). These are one-way dependencies, not circular, but a referee should verify the hypotheses match. The citation pattern is fine. No manufactured flaws here.\n\nBottom line: this deserves a serious referee. I would send it out with explicit instruction to check Section 5 line by line and to confirm the assumptions in the cited preprints. If Section 5 holds, this is a significant paper for the comparison program and for classification. The zero-dimensional part is solid regardless.","headline":"New for non-injective local homeomorphisms; zero-dimensional proof is solid, but the higher-dimensional theorem rides on an unverified thin-boundary Section 5 that needs a line-by-line referee.","tokens_in":28814,"tokens_out":2671,"would_cite":true,"duration_ms":26616,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["22A22","54H20","37B05","20J06","46L35","54F45"],"pacs":[],"model":"deepseek-v4-flash","headline":"Minimal local homeomorphisms of finite-dimensional compact spaces satisfy dynamical comparison.","keywords":["dynamical comparison","Deaconu-Renault groupoids","local homeomorphisms","thin boundary property","partial actions","purely infinite C*-algebras","groupoid homology","AH-conjecture"],"falsifier":"Find a minimal surjective non-injective local homeomorphism $T$ of a compact metrizable finite-dimensional space and a closed set $A$ and non-empty open set $U$ for which finitely many open bisections with disjoint ranges inside $U$ cannot cover $A$; the paper claims no such pair exists. A direct place to look is a 2-to-1 expanding map of the circle, checking whether a small closed arc is dynamically below a tiny open arc.","tokens_in":27718,"feed_emoji":"🔄","tokens_out":12742,"duration_ms":118677,"temperature":0.7,"pith_summary":"The paper establishes dynamical comparison for Deaconu–Renault groupoids of minimal surjective local homeomorphisms of compact metrizable spaces with finite Lebesgue covering dimension. Dynamical comparison is a regularity property saying that whenever a closed set is smaller than an open set in the eyes of every invariant measure, the closed set can be cut into finitely many pieces and moved disjointly into the open set by the dynamics. In the non-injective case there are no invariant probability measures at all, so comparison says the whole space can be dynamically compressed into any nonempty open set, a purely infinite, paradoxical regime. The authors prove this by embedding partial actions of non-abelian free groups as large subgroupoids and controlling their boundaries with a new thin boundary property derived from finite dimension. If correct, this gives a dynamical proof that the associated C*-algebras are UCT Kirchberg algebras and, in the Cantor case, verifies the AH-conjecture for these groupoids.","feed_headline":"Non-injective local maps satisfy dynamical comparison","feed_subtitle":"Closed sets can always be dynamically moved into any open set; the resulting C*-algebras are UCT Kirchberg algebras.","key_machinery":"The central technical object is the thin boundary property for étale groupoids: a basis of open sets whose boundaries are $G$-thin, meaning each boundary is dynamically below every nonempty open set. The paper proves this property for minimal second-countable Hausdorff étale groupoids with compact metrizable unit space of finite covering dimension, without freeness assumptions, using a hierarchy of compact sets $\\mathcal{C}_m$ defined by collision sets: $K \\in \\mathcal{C}_m$ if moving two disjoint pieces of $K$ to the same place produces a collision set of rank lower than $m$; Lemma 5.2 converts membership in $\\mathcal{C}_m$ into thinness, Lemma 5.5 produces boundaries in general position by dimension-counting, and Theorem 5.8 combines these to produce boundaries in $\\mathcal{C}_{d-1}$. This machinery replaces the empty boundaries available in the zero-dimensional case and lets the free-group partial action argument survive in higher dimension.","core_discovery":"The paper's central claim is that every Deaconu–Renault groupoid $G_T$ associated with a minimal surjective local homeomorphism $T \\colon X \\to X$ of a compact metrizable space $X$ with finite Lebesgue covering dimension satisfies dynamical comparison. When $T$ is non-injective, the proof shows the stronger statement that $G_T$ has no invariant probability measures, so dynamical comparison takes the purely infinite form: every closed subset is dynamically below every non-empty open subset. When $T$ is injective, comparison is already known for minimal homeomorphisms, so the new content is the non-injective case. The argument reduces to maps whose fibers have at least two points, embeds a partial action of a non-abelian free group as an open subgroupoid, proves comparison for that partial action using paradoxical towers, and uses the thin boundary property to control the boundary of the embedding. The same comparison result is then used, via known theorems, to recover that $C^*_r(G_T)$ is a UCT Kirchberg algebra and to verify the AH-conjecture for the Cantor-space case.","pith_inferences":["A natural extension, not pursued in the paper, is that the same free-group embedding plus thin-boundary argument should yield dynamical comparison for any minimal second-countable étale groupoid containing a large amenable partial action of a non-elementary hyperbolic group, not just for Deaconu–Renault groupoids.","The freeness-free thin-boundary proof suggests that the small-boundary property for minimal actions on finite-dimensional spaces is a general phenomenon, so regularity results built on small boundaries may survive for non-free actions; the paper only explicitly draws the small-boundary conclusion for group actions.","In the zero-dimensional case, the proof via partial actions of free groups combined with recent work on groupoid homology opens a combinatorial route to the AH-conjecture for the separated-graph models of local homeomorphisms; the paper leaves a direct graph-based derivation as an open question."],"forward_implications":["Every minimal surjective non-injective local homeomorphism of a compact metrizable finite-dimensional space gives a Deaconu–Renault groupoid with no invariant probability measures, so dynamical comparison holds in the strongest purely infinite form: each closed set is dynamically below each nonempty open set.","For any such local homeomorphism, the reduced C*-algebra $C^*_r(G_T)$ is a UCT Kirchberg algebra, recovering a known structural theorem by dynamical methods.","When the unit space is the Cantor set, the associated groupoid satisfies the AH-conjecture for essentially principal minimal étale groupoids, recovering the AH-conjecture for graph groupoids as a special case.","Every minimal second-countable Hausdorff étale groupoid with compact metrizable unit space of finite covering dimension has the thin boundary property.","Every minimal action of a countable discrete group on a compact metrizable finite-dimensional space has the small boundary property, with no freeness assumption."],"supporting_citations":[{"why":"It realizes Deaconu–Renault groupoids with totally disconnected unit space as transformation groupoids of partial actions of free groups, giving the partial-action model used in Section 4.","marker":"[Ste26]"},{"why":"It supplies the paradoxical-towers technique for comparison of minimal amenable non-amenable group actions, adapted here to partial actions of free groups.","marker":"[GGKN23]"},{"why":"It provides the amenability and presentation results that identify the partial-action transformation groupoid with the Deaconu–Renault groupoid and show it is amenable.","marker":"[ES19]"},{"why":"It contributes the finite-dimensional general-position and small-boundary arguments reformulated in Section 5 as the groupoid thin boundary property.","marker":"[Sza15]"},{"why":"It contributes small-boundary techniques, including the version of Lemma 7.5 used to promote the $\\mathcal{C}_m$ hierarchy to thinness.","marker":"[KS20]"},{"why":"It supplies dimension-reduction and general-position tools for the small boundary property that Section 5 adapts to groupoids.","marker":"[Lin95]"},{"why":"It introduces the thin-set terminology and small-boundary ideas for $\\mathbb{Z}$-actions that motivate the thin boundary property.","marker":"[Buc13]"},{"why":"It provides the open-cover refinement lemma used in Lemma 5.2 to color covers in dimension $d$.","marker":"[BK04]"},{"why":"It converts dynamical comparison into the AH-conjecture statement for the Cantor case by relating groupoid homology to topological full group homology.","marker":"[Li25]"},{"why":"It defines groupoid comparison and supplies the pure-infiniteness criterion used to conclude that $C^*_r(G_T)$ is purely infinite.","marker":"[Ma22]"}],"fun_headline_variants":["Dynamical comparison for non-injective local maps","Local maps yield UCT Kirchberg algebras","Comparison proved for minimal local homeomorphisms","Non-injective local dynamics: comparison holds","Deaconu–Renault groupoids have dynamical comparison"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that in every minimal second-countable étale groupoid with compact metrizable unit space of finite covering dimension, the boundaries of basic open sets are small enough to be dynamically moved into any nonempty open set, without any freeness assumption; the higher-dimensional part of the proof collapses if this Section 5 assertion has a gap.","fun_headline_variants_meta":{"raw":{"variants":["Dynamical comparison for non-injective local maps","Local maps yield UCT Kirchberg algebras","Comparison proved for minimal local homeomorphisms","Non-injective local dynamics: comparison holds","Deaconu–Renault groupoids have dynamical comparison"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000246,"raw_usage":{"total_tokens":1553,"prompt_tokens":969,"completion_tokens":584,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":585,"completion_tokens_details":{"reasoning_tokens":513}},"tokens_in":585,"tokens_out":584,"duration_ms":5796,"temperature":1.0,"reasoning_tokens":513,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:49:13.579249+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a minimal surjective non-injective local homeomorphism $T$ of a compact metrizable finite-dimensional space and a closed set $A$ and non-empty open set $U$ for which finitely many open bisections with disjoint ranges inside $U$ cannot cover $A$; the paper claims no such pair exists. A direct place to look is a 2-to-1 expanding map of the circle, checking whether a small closed arc is dynamically below a tiny open arc.","supporting_citations":[],"review_version":1}