{"id":"fb160182-2d06-4d84-a26e-0993da1d732c","arxiv_id":"1908.01149","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Zero topological entropy is equivalent to unique ergodicity for every dynamical system with the approximate product property, with a minimality version under a stronger condition.","lead":"For a broad family of dynamical systems with the approximate product property, this paper proves that zero topological entropy and unique ergodicity are exactly equivalent. The result ties two classical dynamical notions together for this class, and it gives a complete description of interval maps with these properties.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.2 rests on Proposition 2.3, a self-cited 'cf. [27]' result not proved here; if Proposition 2.3 fails without asymptotic entropy expansiveness, both directions collapse.","rationale":"The reader's weakest assumption is exactly Proposition 2.3, and I agree. The proposition is the engine of both directions: without it, the sufficiency proof has no way to separate four measures into disjoint invariant sets, and the necessity proof has no way to produce low-entropy invariant sets whose intersection yields a zero-entropy measure. The paper labels it 'cf. [27]' and says it is 'implicitly proved' there, but [27] is a self-cited arXiv preprint not reproduced in the current text. Because the main theorem depends on an unverified external result, the reader's conditional verdict is appropriate. I considered the nonemptiness gap in Section 3.2, but it is a minor repair (take the first k with Gamma_k cap Lambda_{k+1} = empty); it does not threaten the central claim. I also checked Lemma 3.1; after correcting an apparent typo (a 't' that should read 's' in the last line of Case 2), the separated-set argument is coherent. The remaining open question is whether Proposition 2.3 itself is true in the stated generality, and that is the load-bearing concern.","tokens_in":12586,"tokens_out":22700,"duration_ms":229947,"concrete_test":"Read arXiv:1906.09862 (reference [27]) and verify that Proposition 2.3, or the implicit result it cites, is proved there without invoking asymptotic entropy expansiveness. Specifically, trace the step in [27] that constructs a compact invariant set from the approximate product property; if that step uses asymptotic entropy expansiveness, then Proposition 2.3 is unproved in the stated generality and Theorem 1.2 must be restricted or supplied with a full proof in an appendix. A complementary check: test Proposition 2.3 on the full 2-shift for the non-ergodic measure mu = (delta_0 + delta_1)/2, where APP holds, and confirm that a compact invariant set with all invariant measures close to mu and small scale entropy exists.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Both directions of Theorem 1.2 (Section 3) depend on Proposition 2.3, which asserts that for every invariant measure mu of an APP system and every eta, epsilon, beta > 0 there is a compact invariant set Lambda whose invariant measures all lie eta-close to mu and whose epsilon-scale entropy is < beta. The proposition is not proved in the text; it is only 'cf. [27]', the author's own unpublished preprint arXiv:1906.09862. In Section 3.1, the sufficiency proof needs Proposition 2.3 with non-ergodic mixtures mu3, mu4 to manufacture four pairwise-disjoint low-entropy invariant sets, and the separated-set construction then yields h(f) > 0. In Section 3.2, the necessity proof needs the same proposition to produce sets Lambda_k with h(Lambda_k, f, 1/k) < 1/k; the intersection argument then produces a zero-entropy ergodic measure alongside a positive-entropy one. If Proposition 2.3 is false, or true only under asymptotic entropy expansiveness, Theorem 1.2 as stated does not follow. The paper provides no proof or independent verification of this key step. A secondary gap in Section 3.2 case (1) — Gamma_k may be empty unless one takes the first index where Gamma_k cap Lambda_{k+1} = empty — is easily repaired and is not the main concern.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves Theorem 1.2: for a topological dynamical system with the approximate product property (APP), unique ergodicity is equivalent to zero topological entropy. The proof of sufficiency (§3.1) starts from two distinct ergodic measures, builds four pairwise disjoint compact invariant sets, and then uses the APP tracing property to construct exponentially many separated orbits, forcing positive entropy. The proof of necessity (§3.2) uses Proposition 2.3 to produce invariant sets with arbitrarily small scale entropy whose intersection carries a zero-entropy ergodic measure, while positive entropy supplies a positive-entropy measure. The paper also proves Theorem 1.3 (strict APP: minimality, unique ergodicity, and zero entropy are equivalent), Theorem 1.4 (if the system has a periodic point, APP plus zero entropy is equivalent to unique ergodicity), and Theorem 1.5 (continuous interval maps with APP and zero entropy are exactly those with a unique attracting fixed point). Several examples illustrate the boundary of the results, including a zero-entropy mixing non-minimal APP subshift and Herman's minimal positive-entropy diffeomorphisms.","tokens_in":12912,"tokens_out":10154,"duration_ms":105055,"significance":"If Theorem 1.2 is correct, it gives a clean structural dichotomy for APP systems: zero entropy corresponds to a single invariant measure, and positive entropy corresponds to a Poulsen simplex of invariant measures. This would be a substantial and attractive result, answering the Parry/Herman uniqueness-versus-entropy question within the APP class without asymptotic entropy expansiveness. Theorem 1.3 and the interval-map characterization Theorem 1.5 are also valuable and natural additions. The paper's strengths include an interesting four-measure separation construction in §3.1, a self-contained treatment of the strict APP case, and informative examples. However, the central theorem currently rests on Proposition 2.3, which is not proved in this manuscript, and the separation lemma in §3.1 has an indexing flaw as written. These issues affect the main proof and must be resolved before the result is fully established.","major_comments":[{"comment":"Proposition 2.3 is the engine for both directions of Theorem 1.2. It asserts that for every invariant measure μ and every η, ε, β > 0 there is a compact invariant set Λ whose invariant measures are all η-close to μ and for which h(Λ, f, ε) < β. The paper gives no proof of this proposition, only the reference 'cf. [27]' to arXiv:1906.09862. Because [27] is an unpublished preprint and because the proposition is applied precisely in the APP class without asymptotic entropy expansiveness, this is a load-bearing external dependency. In §3.1 the disjoint invariant sets Λ_i are obtained from Proposition 2.3, and in §3.2 the sets Λ_k with h(Λ_k, f, 1/k) < 1/k are obtained from it. The revision should either prove Proposition 2.3 in the present paper or state the exact theorem/lemma from a published or otherwise available source and verify that its hypotheses match the applications here. A 'cf.' citation is not sufficient for a step of this weight.","section":"§2, Proposition 2.3"},{"comment":"There is an indexing mismatch in the separation lemma. If ξ and ξ' first differ at the index n, the first differing entries of the traced sequences C_ξ and C_{ξ'} are at positions 2n−1 and 2n, not at position n. For example, taking ξ(1)=ξ'(1)=1 and ξ(2)≠ξ'(2) gives x_2(ξ)=x_2(ξ')=y_2, so the proof's comparison of x_n(ξ) and x_n(ξ') may compare identical points and the estimate (6) need not apply. The argument can likely be repaired by comparing blocks indexed 2n−1 and 2n and using a separated length of order (1+δ)(2N)m, which would still yield positive entropy, but as written Lemma 3.1 does not establish the stated lower bound for h(f).","section":"§3.1, Lemma 3.1 and Eq. (7)"},{"comment":"In the case Γ_k ∩ Λ_{k+1} = ∅, the text asserts that Γ_k and Λ_{k+1} support two distinct ergodic measures. This is only justified if Γ_k is nonempty, which need not hold for an arbitrary k. The proof should choose the least k with Γ_k ∩ Λ_{k+1} = ∅; then Γ_k is nonempty by induction and the conclusion follows. This is a small gap, but it occurs in the proof of the necessity direction and should be repaired explicitly.","section":"§3.2, case (1)"}],"minor_comments":[{"comment":"There are typographical errors, e.g. 'approxima te product' in the abstract and 'diffeomorphisms' in the introduction; these should be corrected in the final version.","section":"Abstract and title page"},{"comment":"Theorem 1.1 is cited from [27, Theorem 1.1], but [27] is listed as an arXiv preprint; if the companion paper has been published or accepted by now, the reference should be updated.","section":"§1, Theorem 1.1 citation"},{"comment":"The choice of m > M(δ, δ, γ) from Definition 2.2 is made once for all sequences C_ξ. Since the approximate product property requires the tracing length to be larger than M, this is fine, but the text should say explicitly that the same m is used for all ξ, as the subsequent entropy computation depends on it.","section":"§3.1 after Eq. (7)"},{"comment":"The expression for the uniform convergence of the maximal average density of 1's in X_1 is clear, but the notation '1/n max{|{m ≤ k < m+n : w_k = 1}| : m ∈ N}' would be easier to read with the set-builder written as 'm ∈ N, 0 ≤ m' since N is defined as nonnegative integers earlier; this is a minor notational point.","section":"§4, Example 4.3"}],"recommendation":"major_revision","confidential_remarks":"The main issue is the load-bearing reliance on Proposition 2.3, which is cited only to the author's own unpublished arXiv preprint. This is not a novelty or attribution problem in itself, but a correctness-dependency problem: the referee cannot verify the main theorem without seeing the proof of that proposition. If the companion paper is under review, the authors should be asked to include the needed statements or a detailed proof. The indexing error in Lemma 3.1 appears easily fixable without changing the main result, and the nonemptiness gap in §3.2 is also readily repaired. I do not see evidence of an intentional circular argument; the main proof adds substantial new separation arguments beyond [27]. The paper is a good fit for a dynamical systems journal if the dependency is resolved."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper proves a nice result: for systems with the approximate product property, zero entropy is equivalent to unique ergodicity. That is a clean dichotomy for a well-studied class, and the interval-map classification in Theorem 1.5 is a genuinely nice byproduct. The separation lemma in Section 3.1 is the real work here, and it looks convincing: using mixtures of two ergodic measures to build four pairwise disjoint invariant sets, then coding two full shifts on them, is a clever way around the variable gaps in the definition.\n\nThe soft spot is exactly where the reader put it. Proposition 2.3 is load-bearing for both directions of Theorem 1.2, and it is not proved in this paper. It is stated as \"cf. [27]\"—the author's own arXiv preprint—and that preprint is not yet published as far as I can tell. If Proposition 2.3 needs asymptotic entropy expansiveness, or fails for some other reason, both directions collapse. This is not a matter of a missing routine detail; it is the core of the proof. The referee will need to see the proof of Proposition 2.3, or the paper should include it as an appendix.\n\nThe secondary gap in Section 3.2, case (1), about Gamma_k possibly being empty, is minor and easily repaired: just take the first index where the intersection is empty. That is not a real problem.\n\nThe citation pattern is self-referential but not abusive: the author has a series of papers on this topic, and the prior results are relevant. The concern is purely that the central dependency is not self-contained.\n\nWho is this for? Ergodic theorists working on specification-like properties, and people studying interval dynamics. It deserves a serious referee, not a desk reject, because the claimed theorem is important and the proof strategy is original. But my recommendation would be: send it out, and make sure the referee checks [27] carefully. If Proposition 2.3 holds as stated, the paper is publishable as is modulo small fixes; if not, the theorem may need qualification.","headline":"A clean and useful dichotomy for approximate-product systems, but the main theorem leans on Proposition 2.3, which is not proved here; worth refereeing if the dependency gets resolved.","tokens_in":13360,"tokens_out":1558,"would_cite":false,"duration_ms":16063,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37B05","37B40","37C40","37C50","37E05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that, for topological dynamical systems with the approximate product property, zero topological entropy is equivalent to unique ergodicity.","keywords":["approximate product property","unique ergodicity","topological entropy","ergodic measure","minimality","specification","gluing orbit","interval map"],"falsifier":"A direct check would be to search for a compact metric system with the approximate product property that is uniquely ergodic while having positive topological entropy, or one with zero entropy and two ergodic measures; either would disprove Theorem 1.2. Short of that, one could try to exhibit an invariant measure for a non-asymptotically-entropy-expansive APP system for which Proposition 2.3 fails, since the theorem's proof depends entirely on that proposition.","tokens_in":12416,"feed_emoji":"🔄","tokens_out":9116,"duration_ms":83544,"temperature":0.7,"pith_summary":"This paper proves that for any compact metric topological dynamical system with the approximate product property — a weak specification-like condition that allows orbits to approximate arbitrary sequences of orbit segments with small gaps and occasional mistakes — zero topological entropy is equivalent to unique ergodicity. The result resolves Parry's long-standing question within this class, and it completes a dichotomy: with the approximate product property, the sign of the topological entropy determines whether the space of invariant measures is a single point or a Poulsen simplex. A slightly stronger condition, the strict approximate product property, is shown to make minimality equivalent as well. The proofs manage the variable gaps in the definition, which the author identifies as the main technical obstacle.","feed_headline":"Zero entropy equals unique ergodicity for weak specification systems","feed_subtitle":"The theorem completes a dichotomy: one measure at zero entropy, a rich measure space at positive entropy.","key_machinery":"The object that carries the argument is the approximate product property (Definition 2.2): for every δ1,δ2,ε>0 there is M such that for every n>M, any sequence of orbit segments can be approximated by a single point's orbit, with gaps of size at most 1+δ1n and with the approximating orbit matching each segment except on a δ2-fraction of its ticks. The second pillar is Proposition 2.3, imported from the author's preceding preprint [27]: every invariant measure μ is weak-* approximated by some compact invariant set Λ whose invariant measures all lie in an η-neighbourhood of μ and whose topological entropy at every scale ε is smaller than any prescribed β. The proof of Theorem 1.2 uses these together: the sufficiency direction turns the mutual separation of four such Λ's into an explicit lower entropy bound via the tracing of binary sequences, while the necessity direction employs the same proposition to force either disjoint invariant sets or a zero-entropy intersection, in either case breaking unique ergodicity. The variable gaps of the approximate product property are handled by a careful alignment lemma (Lemma 3.1) that compares two tracings even when their segment boundaries do not coincide.","core_discovery":"The central discovery is Theorem 1.2: for a topological dynamical system (X,f) with the approximate product property, (X,f) is uniquely ergodic if and only if h(f)=0. The sufficiency direction starts from two distinct ergodic measures, uses Proposition 2.3 to build four pairwise disjoint compact invariant sets whose invariant measures stay close to each original measure, and then codes binary sequences by tracing through pairs of these sets. Because the four sets are separated by a fixed distance γ, the tracing property produces 2^n points that are pairwise (1+δ)nm, γ-separated at every length, giving h(f) ≥ ln 2/((1+δ)m) > 0. The necessity direction shows that if h(f)>0, then either the scale-entropy compacta from Proposition 2.3 eventually separate and support distinct ergodic measures, or their intersection is a zero-entropy invariant set while the system also carries a positive-entropy ergodic measure; either way, multiple invariant measures exist. The paper further proves an analogous trichotomy under the strict approximate product property and, for systems with periodic points, that unique ergodicity plus a periodic point implies the approximate product property and zero entropy.","pith_inferences":["The explicit entropy estimate h(f) ≥ ln 2/((1+δ)m) produced by the binary-coding argument gives a constructive way to certify positive entropy in non-uniquely-ergodic APP systems directly from the tracing constants, without searching for separated sets.","The proof's reliance on Proposition 2.3 suggests that the sharpest test of the theorem lies in verifying that proposition for APP systems that are not asymptotically entropy expansive; such a verification would also clarify how much of the mechanism survives if the proposition is weakened.","The distinction between the approximate product property and its strict version appears to be exactly what separates non-minimal zero-entropy behavior from minimality; one could explore whether a similar strict/weak distinction governs minimality for other specification-like properties.","For interval maps, the classification implies a practical criterion: if a continuous interval map has a unique attracting fixed point, it automatically satisfies the approximate product property; this could be used to generate new examples of APP systems in one dimension."],"forward_implications":["For every system with the approximate product property, unique ergodicity is now characterized by zero topological entropy; asymptotic entropy expansiveness is not needed.","The structural dichotomy becomes: h(f)=0 if and only if the space of invariant measures is a singleton, and h(f)>0 if and only if it is a Poulsen simplex.","Under the strict approximate product property, minimality, unique ergodicity, and zero topological entropy are equivalent; the zero-entropy but non-minimal examples must fail the strict property.","If the system has a periodic point, the approximate product property plus zero entropy holds exactly when the periodic measure is the unique invariant measure.","For continuous interval maps, the approximate product property with zero entropy is equivalent to having a unique attracting fixed point."],"supporting_citations":[{"why":"Supplies Proposition 2.3, the entropy-denseness fact that every invariant measure can be approximated by an invariant set of arbitrarily small scale-entropy; also supplies Theorem 1.1 and the minimality-to-zero-entropy implication used here.","marker":"[27]"},{"why":"Introduces the approximate product property and proves the entropy denseness of ergodic measures that motivates the present study.","marker":"[17]"},{"why":"Provides the proof pattern for Proposition 2.6, showing that non-minimality under a strong tracing property implies non-unique ergodicity.","marker":"[23]"},{"why":"Constructs the hereditary shift used in Example 4.3, a zero-entropy mixing subshift with the approximate product property but without minimality.","marker":"[13]"},{"why":"Standard reference for topological entropy, invariant measures, and the variational principle used in the necessity direction.","marker":"[30]"}],"fun_headline_variants":["Zero entropy equals unique ergodicity for approximate product systems","Unique ergodicity iff zero entropy in approximate product systems","Approximate product property forces entropy-ergodicity equivalence","Zero entropy iff unique ergodicity under approximate product property","Zero entropy and unique ergodicity: equivalence under approximate product property"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central theorem rests on Proposition 2.3, imported from the author's earlier unpublished work: every invariant measure can be weak-* approximated by a compact invariant set of arbitrarily small entropy at every scale, a statement the paper does not prove here. If that proposition fails without asymptotic entropy expansiveness, the proof of Theorem 1.2 collapses.","fun_headline_variants_meta":{"raw":{"variants":["Zero entropy equals unique ergodicity for approximate product systems","Unique ergodicity iff zero entropy in approximate product systems","Approximate product property forces entropy-ergodicity equivalence","Zero entropy iff unique ergodicity under approximate product property","Zero entropy and unique ergodicity: equivalence under approximate product property"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000714,"raw_usage":{"total_tokens":3148,"prompt_tokens":823,"completion_tokens":2325,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":439,"completion_tokens_details":{"reasoning_tokens":2242}},"tokens_in":439,"tokens_out":2325,"duration_ms":18379,"temperature":1.0,"reasoning_tokens":2242,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:22:51.528857+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct check would be to search for a compact metric system with the approximate product property that is uniquely ergodic while having positive topological entropy, or one with zero entropy and two ergodic measures; either would disprove Theorem 1.2. Short of that, one could try to exhibit an invariant measure for a non-asymptotically-entropy-expansive APP system for which Proposition 2.3 fails, since the theorem's proof depends entirely on that proposition.","supporting_citations":[{"cited_title":"Ergodic measures of intermediate entropies for dynamical systems with the approximate product property","cited_arxiv_id":"1906.09862","evidence_quote":"Supplies Proposition 2.3, the entropy-denseness fact that every invariant measure can be approximated by an invariant set of arbitrarily small scale-entropy; also supplies Theorem 1.1 and the minimality-to-zero-entropy implication used here."},{"cited_title":"-E., Sullivan, W","cited_arxiv_id":null,"evidence_quote":"Introduces the approximate product property and proves the entropy denseness of ergodic measures that motivates the present study."},{"cited_title":"Discrete and Continuous Dynamical Systems - A , 39(7), 4041-4056 (2019)","cited_arxiv_id":null,"evidence_quote":"Provides the proof pattern for Proposition 2.6, showing that non-minimality under a strong tracing property implies non-unique ergodicity."},{"cited_title":"Discrete and Continuous Dynamical Systems - A , 33(6), 2451-2467 (2013)","cited_arxiv_id":null,"evidence_quote":"Constructs the hereditary shift used in Example 4.3, a zero-entropy mixing subshift with the approximate product property but without minimality."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Standard reference for topological entropy, invariant measures, and the variational principle used in the necessity direction."}],"review_version":1}