{"id":"04788b0f-8d2b-49cb-86f9-1671560acd4f","arxiv_id":"2501.01574","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"The double-dimer nesting field converges to the CLE(4) nesting field in Sobolev spaces as the mesh size tends to zero, and the local loop count satisfies a central limit theorem.","lead":"Crossing the double-dimer model on a fine square grid, this paper proves that the number of loops encircling a fixed point fluctuates like a Gaussian with explicit mean and variance, and that the whole nesting field converges to the CLE(4) nesting field. This is a rigorous step toward the long-open conjecture that double-dimer loops scale to conformal loop ensembles.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.7, Case 2 applies the pointwise asymptotic of Lemma 3.6 at arguments that diverge with s; if s log|w_s| → ∞ the claimed limit of F_s gains a vanishing factor, so the contour argument may yield no contradiction.","rationale":"The reader correctly identifies Lemma 3.7 as the most fragile load-bearing step, and I agree that the uniformity of (3.6) in s→0 is the crux of Theorem 1. However, the reader's statement frames the issue as the constant C_s in Lemma 3.4 possibly blowing up, which Lemma 3.7 is designed to rule out. My concern is more specific: inside Lemma 3.7 itself, Case 2 invokes Lemma 3.6 at a sequence of arguments that goes to infinity with s, and the pointwise nature of Lemma 3.6 makes this invocation invalid when s log|w_s| diverges. In that subcase the normalized kernel F_s can vanish in the limit, so the contour identity (3.22) no longer forces the constants Φ(0) and Φ*(0) to zero, and the compactness contradiction fails. Because Proposition 3.1's error estimate relies directly on Lemma 3.7, the proof of Theorem 1 as written is incomplete at this point. I do not claim the theorem is false; the gap appears repairable by an appropriate rescaling of (3.22), but the manuscript does not provide it. The main external input, Theorem 4.2, is published and not the source of this concern. Therefore I recommend CONDITIONAL rather than ACCEPT: the result is plausible and the structure is sound, but Lemma 3.7's Case 2 must be corrected or supplied with the missing argument before the central claims can be considered fully proven.","tokens_in":66845,"tokens_out":27138,"duration_ms":256921,"concrete_test":"Re-derive the limit used in Lemma 3.7, Case 2: starting from the explicit representation of K_s^{-1}(b,w0) in the proof of Lemma 3.6 (around eq. (3.15)), compute F_s(w)=δ_s^{-1}K_{-s}^{-1}(b1,δ_s^{-1}w) in the regime s→0+, |w_s|→∞ with s log|w_s|→∞. Check whether F_s(w) tends to 0 or to a nonzero constant multiple of -1/(2πw)-η_{b0}^2η_w^2/(2π\\bar w). If it tends to 0, repeat the contour argument with (3.22) multiplied by δ_s^{-s} and verify that the limiting equations still imply Φ(0)=Φ*(0)=0. A numerical corroboration: evaluate K_s^{-1}(b0,W) for W=2^20,2^40,2^80 with s=1/√(log W) and compare the ratio to W^{-s}=e^{-√(log W)}; the observed decay would confirm the missing factor.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of Theorem 1 depends on Proposition 3.1, whose error term requires the uniform bound (3.6) with C_s bounded as s→0, established in Lemma 3.7. Within that lemma, Case 2 rescales by δ_s=|w_s|^{-1} and defines Φ_s(b)=δ_s^{-1}M_s^{-1}K_s^{-1}(δ_s^{-1}b,w_s). To extract the constants Φ(0), Φ*(0) the proof introduces F_s(w)=δ_s^{-1}K_{-s}^{-1}(b1,δ_s^{-1}w) and states, citing Lemma 3.6, that F_s(w)→-1/(2πw)-η_{b0}^2η_w^2/(2π\\bar w) uniformly near a fixed contour γ. However, Lemma 3.6 is a pointwise-in-w asymptotic for fixed w; it does not control w depending on s. Here the second argument is W=δ_s^{-1}w, so |W|→∞. The leading term in (3.11) contains (b0/W)^s = b0^s δ_s^s w^{-s}. If s log|w_s|→∞, the factor δ_s^s=exp(-s log|w_s|) can tend to 0 (e.g. s=1/√(log|w_s|)), so the asserted limit of F_s is false. If F_s→0 on γ, the limiting contour identity (3.22) gives 0=0 rather than equations forcing Φ(0)=Φ*(0)=0, and the contradiction proving limsup M_s<∞ collapses. This is a genuine gap in the manuscript as written; the proof of Lemma 3.7 needs an additional argument (e.g. multiplying (3.22) by δ_s^{-s} before taking the limit) to handle this subcase.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the double-dimer model on the square lattice in the upper half-plane and analyses the random nesting field N_delta(x), the number of double-dimer loops surrounding a point x. Theorem 1 gives the one-point asymptotics E N_delta(x) = -(1/pi^2) log delta + O(1), Var N_delta(x) = -(2/(3pi^2)) log delta + O(1), and the central limit theorem (N_delta(x) - E N_delta(x))/sigma_delta => N(0,1). Theorem 2 states that the unnormalized random field N_delta - E N_delta converges in distribution to the CLE(4) nesting field of Miller-Watson-Wilson with respect to the topology of H^{-1-nu}_{loc}(C_+) for every nu>0. The proofs combine combinatorial identities relating loop counts to the height function, refined asymptotics for the inverse Kasteleyn operator with SL(2,C) monodromy, an approximate Wick rule for the height function, and published cylindrical-event convergence results for double-dimer loops to CLE(4).","tokens_in":67268,"tokens_out":21923,"duration_ms":209278,"significance":"If the proofs are correct, this is a strong and timely contribution: it establishes a CLT for loop nesting counts with the expected logarithmic mean and variance, and it proves field-level convergence to the CLE(4) nesting field without first proving full convergence of the loop ensemble. The constants are derived rather than fitted, the error terms are explicit in the main estimates, and the authors clearly separate what is proved in the upper half-plane from conditional extensions to Temperleyan domains. The reliance on Theorem 4.2 for cylindrical events is appropriate because that input is published and independent of the present conclusions. The main technical risk is concentrated in Lemma 3.7, which supplies the uniformity in s -> 0 of the inverse Kasteleyn estimate (3.6); this uniformity is load-bearing for Proposition 3.1 and hence for Theorem 1(3).","major_comments":[{"comment":"The proof that limsup_{s->0+} M_s < infinity contains a gap in Case 2. The text asserts that F_s(w) = delta_s^{-1} K_{-s}^{-1}(b_1, delta_s^{-1} w) converges to -1/(2 pi w) - eta_{b_0}^2 eta_w^2/(2 pi \\bar w) by Lemma 3.6. However, Lemma 3.6 is an asymptotic in the second argument with the first argument fixed; here the second argument is W = delta_s^{-1} w, so |W| -> infinity. Applying (3.11) with s replaced by -s, the leading terms of F_s have magnitudes delta_s^{-s} |w|^{s-1} and delta_s^s |w|^{-s-1} (up to constants), up to the error term. If s log|w_s| -> infinity, the first magnitude diverges, so the claimed o(1) limit is false unless s log|w_s| -> 0. If one instead normalizes the contour identity (3.22) by delta_s^s before passing to the limit, only the holomorphic term survives; this yields Phi(0)=0 but gives no information about Phi^*(0), so the desired contradiction collapses. Since Lemma 3.7 is the key step that makes the constant C_s in (3.6) uniform as s -> 0, and since (3.6) is the input for Proposition 3.1 and Theorem 1(3), this is a load-bearing gap. The proof needs an additional argument, for example a second contour identity using K_s^{-1} in place of K_{-s}^{-1} to recover Phi^*(0), or a direct uniform estimate showing that the case |w_s| -> infinity cannot occur.","section":"Section 3.1, Lemma 3.7, Case 2 (Eq. (3.22))"}],"minor_comments":[{"comment":"The statement of the asymptotics (3.11) should specify its quantifiers more carefully: it appears to be an asymptotic as |w| -> infinity uniformly in s -> 0+, rather than a pointwise-in-w statement. The notation O(w^{s-2}) is ambiguous and should be accompanied by a sentence explaining the allowed dependence of w on s, especially in the regime s log|w| -> infinity.","section":"Lemma 3.6"},{"comment":"Several technical steps are described as 'straightforward computation' or have details left to the reader, notably Lemma 3.2, the extension of the functions f_s and g_s to white vertices in Lemma 3.5, and part of the proof of Lemma 3.6. Since these computations feed into the uniformity estimates that are central to the paper, spelling out the key cancellations would improve verifiability.","section":"Section 2, Lemmas 3.2 and 3.5"},{"comment":"The manuscript contains a number of typographical inconsistencies, such as 'Tempereley/Tempreley' for 'Temperleyan' and some OCR-like artifacts in formulas. A careful proofreading pass is recommended.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern about Lemma 3.7, Case 2, is genuine in my reading: the subcase s log|w_s| -> infinity is not handled by the argument as written, and the proof of the uniform bound (3.6) needs an additional estimate. However, the overall strategy of the paper is coherent, the constants and main statements are plausible, and the gap appears repairable within the scope of the manuscript, so I would not recommend rejection. If the authors can fix Lemma 3.7 in a revision, the paper would be a substantial contribution to the field."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real result, not a rehash. Theorem 1's CLT with the explicit 1/pi^2 and 2/(3pi^2) constants, and Theorem 2's Sobolev convergence of the nesting field, are genuinely new. The paper earns its length by showing the work: explicit error terms, careful separation of what is proven in the half-plane from what is conditional for Temperleyan domains, and honest reliance on published external inputs like Theorem 4.2. The approximate Wick rule in Lemma 2.6 looks like a useful standalone tool.\n\nThe soft spot is real, and it is load-bearing. The proof of Theorem 1 rests on Lemma 3.7, the uniformity of the bound (3.6) as s -> 0. In Case 2 of that lemma, the authors cite Lemma 3.6 to claim uniform convergence of F_s(w) = delta_s^{-1} K_{-s}^{-1}(b1, delta_s^{-1} w) to a fixed nonzero function on a contour gamma. But Lemma 3.6 is a pointwise asymptotic in the second argument; here the argument delta_s^{-1} w diverges. The leading term contains (b0 / (delta_s^{-1} w))^s = b0^s delta_s^s w^{-s}. If s log|w_s| -> infinity, delta_s^s = exp(-s log|w_s|) -> 0, so the asserted limit can vanish identically. Then the contour identity (3.22) gives 0 = 0, and the contradiction that proves limsup M_s < infinity collapses. I do not see this subcase handled in the manuscript as written. It may be fixable (multiply (3.22) by delta_s^{-s} before taking the limit, or pass to a subsequence with s log|w_s| bounded), but as written it is a genuine gap in the proof of Theorem 1.\n\nThis does not sink the paper's overall value. The CLE-side arguments in Section 4 are independent of Lemma 3.7 and look solid; Theorem 2's proof uses the published Theorem 4.2 and the two-point approximation machinery carefully. The omitted \"straightforward computations\" are annoying but not disqualifying. The small constant a in Definition 3.1 is a technical choice, not a fitted parameter.\n\nWho should read it: people working on dimers, CLE, and discrete complex analysis. It deserves a serious referee, but I would send it with a specific request to verify Lemma 3.7 before accepting. If the gap is repaired, this is a strong paper; as it stands, Theorem 1 is conditional on a lemma I do not fully trust.","headline":"Serious new results in double-dimer to CLE(4) convergence, but Lemma 3.7 has a load-bearing gap that needs fixing before I'd fully trust Theorem 1.","tokens_in":67784,"tokens_out":2738,"would_cite":false,"duration_ms":25918,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60F05","82B20","60J67","60G60"],"pacs":[],"model":"deepseek-v4-flash","headline":"Proven: double-dimer loop counts are Gaussian; field limit CLE(4)","keywords":["double-dimer model","nesting field","conformal loop ensembles","Kasteleyn operator","monodromy","Gaussian free field","central limit theorem","Sobolev spaces"],"falsifier":"Compute the fourth cumulant of $N_\\delta(v)$ from the determinant identity in the paper: Theorem 1 predicts that the normalized fourth cumulant tends to zero, so any nonzero limiting excess kurtosis, or any variance slope differing from $-2/(3\\pi^2)$, would refute the single-point claim. For the field claim, evaluate the smoothed covariance $\\mathbb{E}[(\\phi_\\delta, f)(\\phi_\\delta, g)]$ on compactly supported test functions $f,g$ with supports approaching the diagonal; it should converge to the CLE(4) nesting-field covariance, whose logarithmic diagonal singularity is computable, and a mismatch in that singularity would refute Theorem 2.","tokens_in":66676,"feed_emoji":"🌀","tokens_out":12356,"duration_ms":113763,"temperature":0.7,"pith_summary":"The paper proves that the nesting statistics of the double-dimer model—the loop ensemble formed by superimposing two independent dimer coverings of the square lattice—have a rigid scaling limit. For a fixed point $v$ in the upper half-plane, the number $N_\\delta(v)$ of double-dimer loops encircling $v$ has mean $-(1/\\pi^2)\\log\\delta+O(1)$ and variance $-(2/(3\\pi^2))\\log\\delta+O(1)$, and the normalized count $(N_\\delta(v)-\\mu_\\delta)/\\sigma_\\delta$ converges in distribution to a standard Gaussian as the mesh size $\\delta\\to 0$. At the field level, the centered random field $\\phi_\\delta=N_\\delta-\\mathbb{E}N_\\delta$ converges in distribution to the CLE(4) nesting field in the local Sobolev space $H^{-1-\\nu}_{\\mathrm{loc}}(\\mathbb{C}_+)$ for every $\\nu>0$. This matters because it gives a rigorous, conformally invariant scaling limit for a loop observable without first proving convergence of the loops themselves, and it identifies the large-scale nesting of double-dimer loops with the universal CLE(4) nesting field.","feed_headline":"Proven: double-dimer loop counts are Gaussian; field limit CLE(4)","feed_subtitle":"Mean and variance take universal log(1/δ) form; centered field is conformally invariant.","key_machinery":"The load-bearing identity is $\\mathbb{E}e^{ith_\\delta(x)}=\\mathbb{E}(\\cos t)^{N_\\delta(x)}$, linking the Laplace transform of the loop count to the Fourier transform of the double-dimer height function $h_\\delta$. On the algebraic side, this expectation is a determinant of the Kasteleyn matrix—a weighted adjacency matrix whose inverse encodes dimer correlations—with a scalar monodromy $e^{it}$ around the marked face, and the proof needs a uniform-in-$s$ near-diagonal asymptotic for the inverse Kasteleyn operator with monodromy $e^{2\\pi is}$ as $s\\to 0$. On the probabilistic side, an approximate Wick rule for the height function controls correlations of the squared height, which encodes the two-point behavior of the nesting field after normal-ordering. The Sobolev-space topology $H^{-1-\\nu}_{\\mathrm{loc}}$ is used because the unnormalized field has no pointwise limit; convergence is obtained by comparing both the discrete field and the CLE(4) field to their two-point regularizations at small separation $\\varepsilon$.","core_discovery":"The central discovery is that the double-dimer nesting field has a scaling limit described by CLE(4), the conformal loop ensemble at parameter 4 whose loops are the conjectured scaling limit of double-dimer loops. The authors establish two precise statements. First, for a fixed bulk point $v$, the loop count $N_\\delta(v)$ satisfies a central limit theorem with the explicitly universal logarithmic mean and variance given above; the constants match what one obtains from CLE(4) loops with a cutoff at scale $\\delta$. Second, the unnormalized field $\\phi_\\delta=N_\\delta(\\cdot)-\\mathbb{E}N_\\delta(\\cdot)$, interpreted as a random distribution, converges to the CLE(4) nesting field, the conformally invariant random distribution obtained by subtracting the expected loop count in a regularized construction. The proof does not rely on convergence of individual double-dimer loops; it extracts the limit from algebraic identities relating loop counts to the height function and to determinants of the Kasteleyn operator with monodromy, supplemented by existing results on cylindrical loop events.","pith_inferences":["The same Laplace-transform/Kasteleyn route could be used to study joint nesting statistics at several points, such as the distribution of the number of loops surrounding both $x$ and $y$, with the two-point monodromy construction in Section 3.2 as the natural tool.","One can test the predicted universality numerically on finite Temperleyan domains: the fitted slope of $\\operatorname{Var}N_\\delta(v)$ against $\\log(1/\\delta)$ should converge to $2/(3\\pi^2)$, and the empirical covariance of smoothed $\\phi_\\delta$ should approach the CLE(4) nesting-field covariance with errors governed by the two-scale bound in Proposition 4.3.","A full proof of curve convergence to CLE(4) would make Theorem 2 a corollary; absent that, the nesting-field convergence is the strongest rigorous sense in which double-dimer loop statistics have a CLE(4) scaling limit.","The approximate Wick rule for the height function is likely of independent use for other normal-ordered observables, for example conformal Ward identities or stress-energy-type quantities in the dimer model."],"forward_implications":["If the central claim is right, a single-point loop count in any large half-plane double-dimer configuration has fluctuations of order $\\sqrt{\\log(1/\\delta)}$ around a mean of order $\\log(1/\\delta)$, with a universal Gaussian law once normalized.","The limiting field is conformally invariant, so the distribution of any test-function integral of the centered nesting field does not change under conformal maps of the upper half-plane.","The constants $-(1/\\pi^2)$ and $-(2/(3\\pi^2))$ would also appear if one counted CLE(4) loops whose conformal radius seen from the point is at least $\\delta$, so the double-dimer model and CLE(4) share the same nesting statistics at all scales down to the lattice cutoff.","Convergence of the nesting field provides a well-defined notion of scaling limit for double-dimer loop statistics that does not require convergence of the loops themselves, sidestepping the missing precompactness for curves.","For any test function $f$, $\\int \\phi_\\delta f$ converges in distribution to $\\int \\phi f$, making the CLE(4) nesting field a computable target for finite-size numerical comparisons."],"supporting_citations":[{"why":"Defines the CLE(4) nesting field and its ball-cutoff regularization, the target object of Theorem 2.","marker":"[31]"},{"why":"Introduces the Kasteleyn operator with monodromy and the near-diagonal inverse asymptotics that this paper refines.","marker":"[16]"},{"why":"Supplies the determinant identity expressing double-dimer loop statistics through Kasteleyn operators with monodromy.","marker":"[22]"},{"why":"Provides the infinite-volume double-dimer model in the half-plane and convergence of cylindrical loop events toward CLE(4).","marker":"[17]"},{"why":"Gives the cylindrical-event convergence for double-dimers to CLE(4) used as the external input for Theorem 2.","marker":"[4]"},{"why":"Supplies the uniform loop-count tail estimates for CLE(4) that control the truncated nesting-field approximations.","marker":"[2]"},{"why":"Establishes height-function convergence to the Gaussian free field, the basis of the approximate Wick rule.","marker":"[23]"},{"why":"Shows the electric-correlator identity connecting $\\mathbb{E}e^{ith_\\delta(x)}$ to Kasteleyn determinants.","marker":"[33]"},{"why":"Provides the conformal-invariance and inverse-Kasteleyn asymptotics used for the height function and loop observables.","marker":"[21]"},{"why":"Sets up the discrete complex analysis framework for the Kasteleyn operator used throughout the proofs.","marker":"[12]"}],"fun_headline_variants":["Double-dimer nesting field: Gaussian fluctuations, limit CLE(4)","Loop counts become Gaussian; field limit is CLE(4) nesting","Fixed-point CLT for loop count; nesting field limit is CLE(4)","Double-dimer loop fluctuations Gaussian; field converges to CLE(4)"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of the central limit theorem rests on a uniform bound for the inverse Kasteleyn operator with monodromy as the monodromy parameter tends to zero; if that bound degenerates, the Laplace-transform error term in Proposition 3.1 becomes uncontrolled and Gaussian convergence can fail, and the field-level convergence additionally relies on the already-published convergence of cylindrical double-dimer events to CLE(4).","fun_headline_variants_meta":{"raw":{"variants":["Double-dimer nesting field: Gaussian fluctuations, limit CLE(4)","Loop counts become Gaussian; field limit is CLE(4) nesting","Fixed-point CLT for loop count; nesting field limit is CLE(4)","Double-dimer loop fluctuations Gaussian; field converges to CLE(4)"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000713,"raw_usage":{"total_tokens":3184,"prompt_tokens":898,"completion_tokens":2286,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":514,"completion_tokens_details":{"reasoning_tokens":2209}},"tokens_in":514,"tokens_out":2286,"duration_ms":15132,"temperature":1.0,"reasoning_tokens":2209,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:25:14.737539+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the fourth cumulant of $N_\\delta(v)$ from the determinant identity in the paper: Theorem 1 predicts that the normalized fourth cumulant tends to zero, so any nonzero limiting excess kurtosis, or any variance slope differing from $-2/(3\\pi^2)$, would refute the single-point claim. For the field claim, evaluate the smoothed covariance $\\mathbb{E}[(\\phi_\\delta, f)(\\phi_\\delta, g)]$ on compactly supported test functions $f,g$ with supports approaching the diagonal; it should converge to the CLE(4) nesting-field covariance, whose logarithmic diagonal singularity is computable, and a mismatch in that singularity would refute Theorem 2.","supporting_citations":[{"cited_title":"The conforma l loop ensemble nesting field","cited_arxiv_id":null,"evidence_quote":"Defines the CLE(4) nesting field and its ball-cutoff regularization, the target object of Theorem 2."},{"cited_title":"Dimers and families of Cauchy-Riemann operat ors I","cited_arxiv_id":null,"evidence_quote":"Introduces the Kasteleyn operator with monodromy and the near-diagonal inverse asymptotics that this paper refines."},{"cited_title":"Conformal invariance of loops in the double-d imer model","cited_arxiv_id":null,"evidence_quote":"Supplies the determinant identity expressing double-dimer loop statistics through Kasteleyn operators with monodromy."},{"cited_title":"Double dimers, conformal loop ensembles and isomonodromic deformations","cited_arxiv_id":null,"evidence_quote":"Provides the infinite-volume double-dimer model in the half-plane and convergence of cylindrical loop events toward CLE(4)."},{"cited_title":"Tau-functions ` a la Dub´ edat and probabilities of cylindrical events for double-dimers and CLE(4)","cited_arxiv_id":null,"evidence_quote":"Gives the cylindrical-event convergence for double-dimers to CLE(4) used as the external input for Theorem 2."},{"cited_title":"On the crossing estimates for simple co nformal loop ensembles","cited_arxiv_id":null,"evidence_quote":"Supplies the uniform loop-count tail estimates for CLE(4) that control the truncated nesting-field approximations."},{"cited_title":"Dominos and the Gaussian free field","cited_arxiv_id":null,"evidence_quote":"Establishes height-function convergence to the Gaussian free field, the basis of the approximate Wick rule."},{"cited_title":"Rotational invariance and discrete analyticity in t he 2d dimer model","cited_arxiv_id":null,"evidence_quote":"Shows the electric-correlator identity connecting $\\mathbb{E}e^{ith_\\delta(x)}$ to Kasteleyn determinants."},{"cited_title":"Conformal invariance of domino tiling","cited_arxiv_id":null,"evidence_quote":"Provides the conformal-invariance and inverse-Kasteleyn asymptotics used for the height function and loop observables."},{"cited_title":"Dimer model and holomorphic functions on t-embeddings of planar graphs","cited_arxiv_id":"2001.11871","evidence_quote":"Sets up the discrete complex analysis framework for the Kasteleyn operator used throughout the proofs."}],"review_version":1}