{"id":"9b760559-2624-4183-bf66-99697913c07e","arxiv_id":"2607.23526","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Canonical and microcanonical ensembles are equivalent at local energy scale for Euclidean lattices, with sharp counting asymptotics recovering Bost entropy, but exact-shell equivalence fails for some nonlattice energies.","lead":"The paper proves that for N noninteracting copies of a Euclidean lattice with quadratic energy, fixed subsystems of the microcanonical ensemble converge in total variation to the canonical Gibbs product measure, on exact shells when energies are lattice-valued and on fixed-width windows otherwise. It also gives sharp shell-counting asymptotics whose exponential rate recovers Bost’s Arakelov entropy, and exhibits a nonlattice counterexample where exact shells fail.","discovery_kind":"extension","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"No significant objection identified. The uniform local CLTs (Prop. 8, Lemma 11, Prop. 13) are proved, not assumed, and the ratio-of-probabilities arguments in (14) and (29) are exact algebra that I could not fault.","rationale":"The reader identified the correct load-bearing location — the parameter-uniform local CLTs — and correctly judged it to hold. My independent pass confirms this is the right place to look and that the concern does not land: what could have been an assumption is in fact proved, and the proof's key enabling fact is that the tilted family p_β has full support on Λ for every β > 0, which makes the lattice/nonlattice character and the span β-independent and turns the pointwise strict contraction |ϕ_β(t)| < 1 into a uniform one by compactness. The remaining ingredients (uniform exponential moments via Z(β−η)/Z(β), variance bounds via Prop. 1, exact counting identities (12) and (26), O(1) numerator displacements, Scheffé) are each checkable line by line and check out. The counting corollaries follow by solving the same identities for g_N / G_N, and the Bost normalization match S_B(πu) = s(u) is a correct change of variables. The counterexample is internally consistent and genuinely demonstrates the necessity of windowing in the nonlattice regime, which strengthens rather than weakens the main claim. Weaknesses are presentational only (e.g., Prop. 13's statement fixes h while Cor. 14 applies it with h = Δ/q per fixed q — legitimate, but the limit-swapping could be spelled out). ACCEPT with HIGH confidence stands; the proposed numerical test is a worthwhile end-to-end confirmation rather than a response to any identified flaw.","tokens_in":13138,"tokens_out":7642,"duration_ms":488915,"concrete_test":"Exact numerical replication of the lattice-case chain (Prop. 8 → Thm. 9 → Cor. 10) for Λ = Z², H(a,b) = a²+b² (maximal span h=1). Truncate the one-site energy distribution at H_max with tail mass < 1e-15 under p_β for, say, β = 1; compute the N-fold energy distribution exactly by FFT convolution for N = 25, 50, 100, 200, 400; take E_N = round(N·m(1)) adjusted to the nearest reachable value. Then (a) compute the exact 1-marginal of μ^sh_{N,E_N} via g_{N−1}(E_N − A)/g_N(E_N) and its TV distance to p_1, checking monotone decay roughly like N^{−1/2}; (b) compare (1/N) log g_N(E_N) against s(u) = ψ(1) + m(1). If (a) fails to decrease or (b) misses s(u) by more than O(1/N), the uniformity chain is suspect; if both pass, the central mechanism is confirmed end-to-end.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I looked hardest at the spot the reader flagged: the uniformity in β of the local limit theorems. The concern would land if the uniform bounds were hypotheses rather than theorems, but here they are derived, and the derivations go through for every Euclidean lattice for a structural reason the paper exploits correctly: p_β(x) > 0 for every x ∈ Λ, so the support of Y_β = H(X_β) is β-independent. Consequently (i) the maximal span / nonlattice dichotomy is a property of H(Λ) alone, not of β; (ii) |ϕ_β(t)| < 1 (resp. |φ_β(t)| < 1) holds pointwise on 0 < |t| ≤ π (resp. t ≠ 0) for all β, and continuity of (β,t) ↦ ϕ_β(t) on the compact set K × [δ,π] upgrades this to the uniform bound (10) / (17); (iii) uniform exponential moments come from E_β[e^{ηY_β}] = Z(β−η)/Z(β) with η < β_−, giving uniform third moments and hence the uniform Gaussian bound |ξ_β(t)| ≤ e^{−ct²} near 0, with v(β) bounded below by compactness and strict positivity of σ². The two ratio identities are then exact: (12) and (26) are pointwise counting identities, and in (14)/(29) the numerator displacement is ku_N − A = O(1), so the Gaussian factors and the 1/√(N−k) vs 1/√N prefactors both have ratio → 1; Scheffé on the countable space Λ^k converts pointwise convergence to TV. The sandwich argument in Prop. 13 (Fejér kernel for point masses, Beurling–Selberg for endpoints) is standard and correctly ordered (fixed T, N→∞, then T→∞; fixed q, N→∞, then q→∞ in Cor. 14). I also checked the counterexample (Prop. 17): H = 3a²+6b²−2√2ab is positive definite (discriminant 8−72 < 0), (M,Q) support contains (0,0),(1,0),(3,1) generating Z², and Q-linear independence of 3 and 2√2 forces the pair constraint; the exclusion of p_β via ratio comparison with (0,0) is sound. No load-bearing gap found.","agreement_with_reader":"agree"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The paper studies N labelled noninteracting copies of a Euclidean lattice (Λ,‖·‖) with one-site energy H=‖·‖². It proves equivalence of ensembles at local energy scale, split by the arithmetic nature of H(Λ). In the lattice case (H(Λ)⊆hZ with maximal span h), Theorem 9 shows that on reachable exact shells with E_N/N→u, every fixed k-marginal of the uniform shell measure converges in total variation to p_β^⊗k with β=m⁻¹(u); Corollary 10 gives the sharp shell count with prefactor and (1/N)log g_N(E_N)→s(u)=ψ(β)+βu. In the nonlattice case, Theorem 15 and Corollary 16 prove the analogous statements for fixed-width Δ-windows via a uniform version of Stone's integro-local theorem (Proposition 13) and a weighted variant (Corollary 14). Section 5 identifies s(u) with Bost's stable Arakelov entropy S_B up to the normalization H_B=πH, and derives the high-temperature expansion with the arithmetic correction ddeg(L)−½log|D_K|. Section 6 (Proposition 17) constructs a nonlattice quadratic H=3M−2√2 Q on Z² for which exact-shell marginals converge to a product law outside the one-parameter Gibbs family, showing that windowing is necessary in the nonlattice regime.","tokens_in":13556,"tokens_out":5147,"duration_ms":182392,"significance":"If correct, this is a clean, self-contained local refinement of Bost's cumulative thermodynamic formalism: Bost's [2, Thm 5.2.1] works at exponential scale for sublevel counts, while this paper resolves individual reachable levels (lattice case) and O(1) total-energy windows (nonlattice case), identifies the limiting law of every fixed subsystem in total variation, and supplies sharp prefactors. The strengths are concrete: there are no free parameters anywhere in the limit statements; the entropy s(u) is defined variationally from the theta series and then proved to equal shell/window growth, so the identification with S_B is a normalization change, not a circular import; the two parameter-uniform local limit theorems (Propositions 8 and 13) are proved rather than assumed, with the uniformity in β reduced to structural facts (β-independent support, strict positivity of theta weights, exact exponential-moment identity E_β[e^{ηY}]=Z(β−η)/Z(β)); and Proposition 17 is a genuinely informative counterexample whose arithmetic realization over K=Q(√2) ties the lattice/nonlattice dichotomy to metric data of Hermitian line bundles. The exact-shell counterexample is, to my knowledge, the news","major_comments":[],"minor_comments":[{"comment":"Proof of Theorem 9 (and similarly Theorem 15): the phrase 'since the sequence {β_N} is compact' should read 'since {β_N} is eventually contained in a compact subset of (0,∞)' — the sequence itself is not compact; only its range with its limit is.","section":"§3, proof of Theorem 9"},{"comment":"Proposition 13 uses h>0 for the interval length in [x,x+h), while h already denotes the maximal span in §3. The contexts do not overlap, but a different letter (e.g., ℓ) in Proposition 13 would avoid confusion, especially since Corollary 14 then partitions [0,Δ).","section":"§4, Proposition 13"},{"comment":"Proposition 17 invokes 'the two-dimensional aperiodic lattice local central limit theorem' and later 'the multivariate lattice local central limit theorem' without a reference. A citation (e.g., [5, Chapter 2] again, or Bhattacharya–Rao) would help, since the nondegeneracy/aperiodicity hypotheses verified in the text should match the cited statement.","section":"§6, proof of Proposition 17"},{"comment":"Theorem 9 assumes E_N∈im(H_N) reachable. It would be worth one remark that for every u>0 such reachable sequences with E_N/N→u actually exist: by the maximal-span condition (7), the additive semigroup generated by H(Λ) contains all sufficiently large multiples of h, so any u>0 is approachable. This makes the theorem visibly non-vacuous for arbitrary u.","section":"§3, Theorem 9"},{"comment":"The high-temperature expansion s(u)=(d/2)(1+log(2πu/d))+(ddeg−½log|D_K|)+o(1) uses β=d/(2u)+O(e^{−c/β}) implicitly; the exponentially small error in m(β) is stated, but the resulting error in ψ(β)+βu after substituting the approximate inverse is not estimated. Since the error is O(ue^{−cu}), the o(1) conclusion is fine, but one line of justification would close it.","section":"§5, entropy expansion"},{"comment":"In the Introduction, the sentence defining the criterion appears as a sentence fragment ('and μ^can_{N,β}=p_β^⊗N is the entropy-maximizing law with mean energy Nu'); please check the typesetting of this passage.","section":"§1, first page"},{"comment":"Lemma 11(i) is dismissed with 'can be proved similarly'; since this exponential-moment bound is the input for the uniform Gaussian estimate (21), repeating the one-line identity E_β[e^{ηY_β}]=Z(β−η)/Z(β) with η<inf K would make §4 self-contained.","section":"§4, Lemma 11"},{"comment":"Corollary 14's mesh argument ('fixed q, N→∞, then q→∞') is correct but compressed; a half-sentence noting that the two-dimensional array is controlled because the lower/upper Riemann sums are monotone in q would help readers.","section":"§4, Corollary 14"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript sits at the interface of probability (local limit theory) and Arakelov geometry; for a probability journal the main theorems stand alone, and §5–6 supply the arithmetic motivation without importing unproved input. The overlap with Bost's work is disclosed accurately — the paper is explicit that the exponential-rate identification is a renormalization of [2, Thm 5.2.1] and that its contribution is the local refinement plus marginal convergence. I verified the load-bearing points independently: the β-uniformity of the LLTs rests on β-independence of the support (theta weights are everywhere positive), the ratio identities (14) and (29) are exact algebra with O(1) displacement, and the Proposition 17 counterexample checks out (the quadratic form has determinant 16>0, and 3, 2√2 are Q-linearly independent, so the exact shell genuinely encodes two integer constraints). I found no load-bearing gaps."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The one thing worth knowing: for quadratic energy on a Euclidean lattice, once E_N/N \to u = m(eta), every fixed k-marginal of the exact-shell microcanonical (lattice, reachable) or fixed-width windowed microcanonical (nonlattice) converges in TV to the canonical product, and (1/N) log of the shell/window count recovers s(u). That is a genuine local refinement of Bost’s cumulative Arakelov counting, not just a rephrasing.\n\nWhat is new is the dichotomy itself, the O(1)-width window regime, the sharp prefactors, and especially Prop. 17: a concrete nonlattice quadratic where an exact energy equation secretly imposes two independent integer constraints, so the exact-shell marginals leave the one-parameter Gibbs family. Thickening restores canonicity. The proofs are the standard exponential tilt plus parameter-uniform local CLT (lattice Fourier inversion; Stone + Beurling–Selberg sandwich for nonlattice), written carefully enough that the ratio identities (14) and (29) are exact algebra and Scheffé finishes the TV step. Uniformity in eta is proved, not assumed: support of H is eta-independent, exponential moments come from the theta ratio, and the characteristic-function bounds upgrade by continuity on compacts. The Bost comparison in §5 is honest about scale (fixed total width vs his \\sqrt N windows) and normalization.\n\nSoft spots are minor. The uniformity arguments sketch Fourier tails rather than chase every constant; reachability of exact shells is correctly flagged as essential rather than automatic. No free parameters, no circularity, citation pattern is appropriate (Bost, Diaconis–Freedman, classical local CLTs). Significance is solid inside math.PR / arithmetic geometry, not field-changing.\n\nThis is for people who care about conditional limit theorems or Arakelov lattice-point counts. It deserves a serious referee. I would engage with it and expect to cite the local counting and the counterexample.","headline":"Clean local upgrade of Bost’s entropy to exact shells/windows plus subsystem TV equivalence; the lattice/nonlattice split and counterexample are the real additions.","tokens_in":14588,"tokens_out":502,"would_cite":true,"duration_ms":9485,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60F05","82B30","11H06"],"pacs":[],"model":"grok-4.5","headline":"For Euclidean lattices with quadratic energy, every fixed subsystem of the right microcanonical measure converges to the canonical Gibbs product once energy density is matched.","keywords":["ensemble equivalence","Euclidean lattices","canonical ensemble","microcanonical ensemble","local central limit theorem","thermodynamic entropy","lattice-point counting","Arakelov entropy"],"falsifier":"Exhibit a Euclidean lattice whose Gibbs characteristic functions fail the uniform bounds on some compact temperature range, or verify on the paper’s explicit nonlattice quadratic example that exact-shell k-marginals converge outside the Gibbs family while fixed-width window marginals converge to it.","tokens_in":14219,"feed_emoji":"⚖️","tokens_out":989,"duration_ms":30349,"temperature":0.7,"pith_summary":"This paper asks when the microcanonical picture of many noninteracting copies of a Euclidean lattice—equal weight on a fixed total energy—looks the same as the canonical Gibbs product at matching mean energy. The answer splits on arithmetic: if energies live on a lattice, reachable exact shells work; if not, one must thicken to a fixed-width energy window. In both regimes every fixed collection of sites has a marginal that converges in total variation to the Gibbs product, and the exponential growth rate of the number of configurations recovers the thermodynamic entropy density. A counterexample shows that sticking to exact shells in the nonlattice case can force hidden integer constraints and land outside the one-parameter Gibbs family. The local counting formulas refine cumulative lattice-point asymptotics to individual shells or O(1) windows while keeping the same entropy function.","feed_headline":"Lattice microcanonical shells match Gibbs on every fixed subsystem","feed_subtitle":"Exact shells work when energy is lattice-valued; fixed-width windows restore equivalence otherwise.","key_machinery":"An exponential tilt that centers the prescribed energy under the product Gibbs law, followed by parameter-uniform local central limit theorems (lattice Fourier form; Stone integro-local form for nonlattice windows). Ratios of the resulting shell or window probabilities yield the limiting subsystem law; the probabilities themselves give sharp counting asymptotics.","core_discovery":"Once the energy per site tends to a positive density u equal to the Gibbs mean m(β), every fixed k-marginal of the exact-shell microcanonical measure (lattice energies, reachable shells) or of the fixed-width windowed microcanonical measure (nonlattice energies) converges in total variation to the canonical product p_β^⊗k. Simultaneously, (1/N) log of the shell or window cardinality tends to the entropy density s(u) = ψ(β) + βu.","pith_inferences":["The lattice/nonlattice split suggests that ‘local’ ensemble equivalence is really about whether the energy map is a single aperiodic additive constraint or hides a higher-rank lattice of constraints.","Uniform local-CLT technology used here should extend, with work, to other one-site Hamiltonians that keep uniform exponential moments and nonlattice characteristic-function gaps.","Matching only the mean energy is enough for fixed marginals; large-deviation or moderate-deviation windows would likely be needed for extensive subsystems.","Arithmetic applications may extract effective error terms in shell counts for concrete number-field lattices once the o(1) in the local CLT is made explicit."],"forward_implications":["Reachable exact energy shells on lattice-valued H admit the sharp count g_N(E_N) ∼ h Z(β_N)^N e^{β_N E_N} / √(2π N σ²(β_N)).","Fixed-width nonlattice windows admit the analogous count with an extra factor ∫_0^Δ e^{β_N t} dt, independent of the limiting marginal.","Both exponential rates equal the same entropy density that appears in stable Arakelov cumulative lattice-point counting.","Exact-shell equivalence can fail for nonlattice quadratic energies that encode several independent integer constraints; fixed-width windows remove those constraints.","The same local equivalence applies to Euclidean lattices coming from Hermitian line bundles over rings of integers."],"fun_headline_variants":["Lattice shells match Gibbs on every fixed marginal","Exact energy shells equate canonical and microcanonical ensembles","Microcanonical lattice shells converge to Gibbs product measures","Fixed-k marginals of lattice shells equal canonical Gibbs","Shell counting recovers entropy; marginals match Gibbs at density u"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The one-site Gibbs family must obey local limit theorems that stay uniform when inverse temperature varies over compact intervals, which rests on uniform exponential-moment and characteristic-function bounds for that family.","fun_headline_variants_meta":{"raw":{"variants":["Lattice shells match Gibbs on every fixed marginal","Exact energy shells equate canonical and microcanonical ensembles","Microcanonical lattice shells converge to Gibbs product measures","Fixed-k marginals of lattice shells equal canonical Gibbs","Shell counting recovers entropy; marginals match Gibbs at density u"]},"model":"grok-4.5","effort":"low","cost_usd":0.002824,"raw_usage":{"total_tokens":955,"prompt_tokens":672,"num_sources_used":0,"completion_tokens":77,"cost_in_usd_ticks":28244000,"prompt_tokens_details":{"text_tokens":672,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":206,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":672,"tokens_out":77,"duration_ms":5119,"temperature":1.0,"reasoning_tokens":206,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-30T20:01:53.203748+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Exhibit a Euclidean lattice whose Gibbs characteristic functions fail the uniform bounds on some compact temperature range, or verify on the paper’s explicit nonlattice quadratic example that exact-shell k-marginals converge outside the Gibbs family while fixed-width window marginals converge to it.","supporting_citations":[],"review_version":1}