{"id":"945f55a5-a51f-43f8-9c3e-5300efb62dd9","arxiv_id":"1908.04850","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Uniform connected planar graphs have a quenched local limit, a new infinite random graph called the uniform infinite planar graph (UIPG).","lead":"This paper proves that a random connected planar graph with n labelled vertices, viewed near a random vertex, converges to a single infinite random graph, the uniform infinite planar graph. It also proves matching local limits for 2-connected planar graphs and maps, and reproves the known asymptotic count of planar graphs.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Subcriticality inequality ν_C<1 in §8.1 rests on approximate constants; without error bounds the condensation core of the proof is not yet rigorous.","rationale":"The reader's weakest assumption is exactly the most load-bearing: the strict inequality ν_C<1 underpins the entire proof architecture (equation (9.103), Lemmas 3.2–3.3, and the subcritical Galton–Watson encoding). The paper's verification is an approximate numerical substitution with no error control. I see no reason to suspect falsehood—the margin is large—but the gap is real and fills the criterion for a conditional verdict. The manuscript contains many passages deferring tedious estimates, but none is individually as decisive as this inequality. A single rigorous interval-arithmetic computation would settle the issue. Hence no change to the reader's CONDITIONAL verdict.","tokens_in":55477,"tokens_out":14990,"duration_ms":152145,"concrete_test":"Perform a certified interval-arithmetic computation of the right side of Inequality (8.5) using the explicit analytic expressions for D0, D2, and ρ_B from Bender et al. (2002) (with the factor-t correction in D2 noted by Giménez–Noy 2009), together with a rigorous bound on the O(X^4) remainder in the singular expansion of N(x,1). If the interval upper bound for ν_C is strictly below 1, the subcriticality assumption is established; if the interval contains 1 or exceeds it, the condensation argument in §§8.1 and 9.9 is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central limit theorem needs E[ξ_P]=ν_C=ρ_B ∂²B/∂x²(ρ_B,1)<1 (Eqs. (8.3), (9.103)), since Lemmas 3.2–3.3 and Eq. (3.7) are applied to the simply generated tree T^P_n only in this subcritical regime. §8.1 establishes the inequality by substituting D0≈1.09417 and D2≈−0.13749 into Eq. (8.5), constants taken from Bender et al. (2002) without explicit error bounds for the singular expansion N(x,1)=D0+D2X²+D3X³+O(X⁴), nor rigorous enclosures for ρ_B and the approximated constants. The resulting numerical value 0.041302 is far below 1, so the inequality is almost certainly true, but the manuscript does not supply a proof of it; the check is a numerical evaluation, not a verified bound. If ν_C were actually ≥1, the tree encoding would leave the condensation regime and the quenched local convergence argument would collapse at its first probabilistic input. This is the same load-bearing gap the reader identified; the other 'details left to the reader' (e.g., Lemma 9.10) are secondary by comparison.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a probabilistic framework for the local convergence of random planar structures. Its main theorem, Theorem 1.1, states that the uniform connected simple planar graph P_n on n labelled vertices, rooted at a uniformly chosen vertex, converges in the quenched sense in the local topology to a novel infinite random graph called the uniform infinite planar graph (UIPG). Along the way the paper establishes analogous quenched local limits for vertex-weighted 2-connected planar graphs (Theorem 9.11 and Theorem 1.2) and non-separable planar maps (Theorems 9.9 and 1.3), and it recovers the Giménez–Noy asymptotic formula for the number of planar graphs (Theorem 1.4). The proof combines Tutte's decomposition with Gibbs partitions, enriched tree encodings, condensation phenomena in simply generated trees, and transfer arguments between random mixtures.","tokens_in":55771,"tokens_out":4562,"duration_ms":45120,"significance":"If the technical gaps described below are closed, this would be a major contribution: it provides the first quenched local limit for uniform connected planar graphs, strengthening the existing annealed picture and yielding natural applications such as subgraph-count asymptotics via Corollary 1.5. The high-level architecture is coherent, and the paper contains a genuinely new probabilistic view of the Tutte decomposition, reducing planar graph limits to condensation in subcritical Galton–Watson trees and Gibbs partition convergence. The dependence on the author's previously published theorems is legitimate because those results are stated with explicit assumptions and are not fitted to the present problem. The main risk is not circularity but rather the completeness of several load-bearing technical estimates.","major_comments":[{"comment":"The inequality ν_C < 1 is load-bearing for Theorem 1.1, since E[ξ_P] = ν_C is the only source of subcriticality in Eq. (9.103), and Lemmas 3.2–3.3 are applied to the simply generated tree T^P_n only in that regime. The verification is currently an unchecked numerical evaluation: the singular expansion N(x,1) = D0 + D2X^2 + D3X^3 + O(X^4) is quoted from Bender et al. (2002) with approximate constants D0 ≈ 1.09417 and D2 ≈ −0.13749, and no explicit error bounds or validated enclosures are given for these constants or for ρ_B. A failure of ν_C < 1 would collapse the condensation argument, so the manuscript must either prove this inequality rigorously (for example, by interval arithmetic applied to the analytic expressions) or cite a verified computation with explicit error bounds; the displayed value 0.041302 < 1 is not itself a proof.","section":"Section 8.1, Eq. (8.5)"},{"comment":"The offspring distribution ξ_M of the simply generated tree encoding non-separable maps is supported on even integers, so it does not satisfy Condition (3.4) as stated. The manuscript asserts in one sentence that Lemmas 3.2 and 3.3 'may be extended' to this setting by rescaling by 1/2. These lemmas are used repeatedly, for example in Eqs. (9.3), (9.26), (9.52), and (9.75), to obtain the local limit theorems for core sizes, so the periodic extension is not a purely cosmetic detail. The paper should state and prove the even-version lemma, or explicitly spell out the rescaling argument and verify that the uniformity of the o(1) error terms is preserved.","section":"Section 9.1, Eq. (9.2) and following paragraph"},{"comment":"The local limit theorem for O(K^n_t), which is later used in Sections 9.7–9.8 to derive quenched convergence of 2-connected planar graphs, depends on a long double-sum simplification. The proof delegates the main part to 'tedious but not difficult steps' and leaves the details to the reader. Because this estimate is the bridge from the local limit theorem for R(K^n_t) to that for O(K^n_t), the details should be included, or the reduction should be replaced by a direct derivation. As written, this is an assertion rather than a completed proof at a point that is load-bearing for Theorem 9.11.","section":"Lemma 9.10, proof of Eq. (9.75)"}],"minor_comments":[{"comment":"The phrase 'by performing analytic integration and man m la using analytic methods' appears to contain a garbled or corrupted word ('man m la'); it should be corrected to a readable sentence.","section":"Section 1.1, Introduction"},{"comment":"The notation ¯D(Mt_n) appears without definition; the context suggests it should be D(Mt_n), the D-network corresponding to V(Mt_n).","section":"Section 9.2, proof of Corollary 9.4"},{"comment":"Equation (9.105) is asserted to follow from Eq. (9.104) 'by identical arguments' to Corollary 9.4, but the proof is not given, and the remark only describes verbally the subtle distinction between B(P_n) and the largest 2-connected block. Since this statement is not used in the proof of the main theorem, it could be moved to a clearly marked sketch or proved in full.","section":"Remark 9.12"},{"comment":"The simplification leading to νM(t) would benefit from at least one intermediate algebraic step; as written, the reader must reproduce a lengthy reduction involving Eqs. (8.14), (8.15), (8.22), and (8.33).","section":"Section 8.4, Eq. (8.35)"},{"comment":"The rendering 'ρ−n C' should be ρ_C^{-n} to match the notation in Eq. (1.4).","section":"Abstract and Theorem 1.4"}],"recommendation":"major_revision","confidential_remarks":"The main theorem is likely correct, and the cited prior work provides independent support for the components of the proof. However, the numerical check in Section 8.1 is exactly the kind of threshold condition that interval arithmetic can settle rigorously, and I would ask for that before acceptance. The 'details left to the reader' passages in Section 9.1 and Lemma 9.10 should also be filled in, since they are not peripheral; they support the central local-limit theorems. The paper is long but well organized, and the proposed fixes appear feasible within the manuscript's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a serious paper. The quenched local limit for uniform connected planar graphs (the UIPG) is a real advance, and the enriched-tree/condensation approach is the right tool. The main caveat is the subcriticality check in Section 8.1, which relies on approximate constants from Bender–Gao–Wormald without explicit error bounds. That needs to be nailed down before I call the proof complete, but I would still send it to review.\n\nWhat is genuinely new: prior local limits were for maps or 3-connected maps, not the full connected graph. The UIPG is a canonical object, and the coupling between the UIPG and the 2-connected limit is natural and useful. The proof of the Giménez–Noy formula via large deviations and subexponential random walks is a nice alternative to analytic integration.\n\nWhat is done well: the overall architecture—going from maps to non-separable cores, then to O-cores, then transferring via Gibbs partitions—is coherent, and the main induction in Lemmas 9.2 and 9.6 is written out in enough detail to follow. The heavy reliance on the author's own earlier theorems is legitimate: those are published, parameter-free results with stated assumptions, not fitted to the data here. No circularity.\n\nThe soft spots: the load-bearing step is Inequality (8.5), ν_C < 1. The paper substitutes D0 ≈ 1.09417 and D2 ≈ −0.13749 from Bender et al. (2002), notes a typo in their expansion, and obtains 0.041302 < 1. There are no rigorous enclosures for the constants or for the remainder term in the singular expansion. The margin is large, so I am almost certain the inequality is true, but as written the verification is numerical rather than analytic. If it failed, the condensation regime for the tree encoding would break, so this is exactly where a referee should press. There are also smaller gaps: the periodic extension of the condensation lemmas in Section 9.1 is asserted (easy to fill), the double-sum simplification in Lemma 9.10 is left partly to the reader, and Remark 9.12 waves off a distinction between the largest block and the W-core. None are fatal, but they turn the preprint into a draft rather than a final version.\n\nWho gets value: anyone working on random planar graphs, local convergence, or combinatorial species. I would cite this once the ν_C point is resolved—though if I needed the UIPG as a known object, I might cite it before that. For peer review: yes. The result is important, the method is novel, and the gaps look closable. I would recommend major revision with a specific request for a verified bound on the right-hand side of (8.5), plus a fuller write-up of the skipped technical steps.","headline":"A major advance with a largely sound proof; the one load-bearing numerical inequality needs a rigorous bound.","tokens_in":56264,"tokens_out":3860,"would_cite":true,"duration_ms":37900,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C80","60C05","05C10","60J80","05A15"],"pacs":[],"model":"deepseek-v4-flash","headline":"Uniform random labelled planar graphs converge locally to a new infinite random graph, the UIPG.","keywords":["random planar graphs","local weak convergence","quenched limit","uniform infinite planar graph","Tutte decomposition","condensation","Gibbs partitions","enriched trees"],"falsifier":"Compute certified interval bounds for $\\nu_C = \\rho_B \\, \\partial^2 B/\\partial x^2(\\rho_B,1)$ using rigorous interval arithmetic on the singular expansion of the 2-connected planar graph generating series; a certified lower bound at least 1 would disprove the condensation premise and invalidate the proof of Theorem 1.1, while a certified upper bound below 1 would close the remaining numerical gap.","tokens_in":2143,"feed_emoji":"🕸️","tokens_out":7563,"duration_ms":134397,"temperature":0.7,"pith_summary":"This paper proves that a graph drawn uniformly from all connected planar graphs on $n$ labelled vertices has a well-defined asymptotic local shape: the neighbourhood of a uniformly random vertex converges, in the stronger quenched sense, to a single infinite random planar graph called the uniform infinite planar graph (UIPG). It establishes the analogous quenched limits for uniform 2-connected planar graphs and for 2-connected planar maps, and it identifies how the UIPG is assembled from the 2-connected limit by attaching independent Boltzmann-distributed connected pieces. The proof runs through the Tutte decomposition probabilistically, encoding each connectivity layer as a tree that exhibits a condensation phenomenon, and the classical asymptotic count of planar graphs falls out as a by-product of the same machinery. A sympathetic reader should care because the UIPG provides a canonical infinite object in which local statistics of finite random planar graphs, such as degree distributions and subgraph counts, can be read off directly.","feed_headline":"Random planar graphs converge locally to a new infinite graph","feed_subtitle":"The limit is built from 2-connected pieces and makes local statistics like degrees and subgraph counts readable at infinity.","key_machinery":"The engine of the argument is a fully recursive tree-like encoding of the Tutte decomposition. For graphs the paper introduces the species $K$ and $R$ of networks, related by $K \\equiv yR(x,K)$ with $R = J\\,\\mathrm{SEQ}(I^*)$, and for maps the analogous barred species $\\bar{K}$ and $\\bar{R}$; each identity turns the connectivity layer into a simply generated tree decorated by smaller networks. Sampling such an enriched tree produces a Galton-Watson tree with subexponential offspring in the subcritical regime, and the condensation phenomenon then forces a unique giant component at each layer, with fluctuations of order $n^{2/3}$ governed by a $3/2$-stable density. Local limit theorems for the giant component size and for fringe subtrees, together with Gibbs-partition transfer results, let quenched convergence pass successively from maps down to $\\bar{O}$-, $\\bar{R}$-, $\\bar{K}$-, and $V$-cores, then back up the graph-side chain to 2-connected and connected planar graphs.","core_discovery":"The central claim is Theorem 1.1: if $P_n$ is the uniform connected simple planar graph on $n$ labelled vertices and $v_n$ is a uniformly selected vertex, then the regular conditional law $\\mathcal{L}((P_n,v_n)\\mid P_n)$ converges in probability to the law of a limiting infinite planar graph $\\hat{P}$, the UIPG, in the local topology. This is quenched convergence, meaning the empirical distribution of rooted neighbourhoods inside a single large random graph approximates the limit, not merely the averaged law. The paper also proves quenched local limits for uniform 2-connected planar graphs (with limit $\\hat{B}$, the UI2PG) and for non-separable planar maps (with limit $\\hat{V}$, the UI2PM), and it shows that $\\hat{P}$ is obtained from $\\hat{B}$ by inserting i.i.d. Boltzmann-distributed vertex-marked connected planar graphs at non-root vertices and a doubly marked Boltzmann component at the root. Along the way the paper recovers the asymptotic formula $p_n \\sim c_G \\rho_C^{-n} n^{-7/2}$ for the number of planar graphs, without the analytic integration used in the original proof.","pith_inferences":["Inference: because the route only uses a subexponential $n^{-5/2}$ census tail and Tutte stability, the same enriched-tree-plus-condensation scheme should produce quenched local limits for other Tutte-stable graph classes with the same census profile, and comparing the resulting limits would test how universal the UIPG-type shape is.","Inference: the paper explicitly notes that the graph-side decomposition $K \\equiv yR(x,K)$ is not isomorphism-preserving, so the unlabelled random planar graph is not covered; controlling automorphism bias in that substitution would show whether the same UIPG appears for unlabelled graphs.","Inference: the proof identifies $n^{2/3}$-scale, $3/2$-stable fluctuations for the sizes of the successive giant cores; a concrete test is to generate moderately large planar graphs and check whether the largest 2-connected block size, rescaled by $n^{2/3}$, matches the stable density $h$ used throughout the paper.","Inference: quenched convergence suggests one can estimate UIPG subgraph probabilities from a single large sample graph, but the paper gives no rates; obtaining explicit rates would require quantitative versions of the condensation and transfer lemmas."],"forward_implications":["The UIPG exists as a quenched local limit, and the stationary-rooted version of the same convergence implies that $\\hat{P}$ is almost surely recurrent.","For any fixed finite connected graph $H$, the number of subgraph occurrences satisfies $\\mathrm{emb}(H,P_n)/n \\to \\mathbb{E}[\\mathrm{emb}^{\\bullet}(H^{\\bullet},\\hat{P})]$ in probability.","The asymptotic enumeration constant and exponent $p_n \\sim c_G\\rho_C^{-n}n^{-7/2}$ follow from the probabilistic condensation argument without a single analytic integration step.","The vertex-weighted versions of random 2-connected planar graphs and non-separable planar maps admit quenched local limits with explicitly described infinite limiting objects.","The root degree of the UIPG matches the known asymptotic degree distribution of uniform random planar graphs."],"supporting_citations":[{"why":"Supplies the asymptotic count and singular expansion of 2-connected planar graphs, including the constants $D_0$ and $D_2$ used to verify the condensation inequality $\\nu_C<1$ and to seed the enumeration chain.","marker":"Bender et al. (2002)"},{"why":"Provides the target asymptotic formula for planar graphs and the analytic expressions for $\\rho_C$ and $c_G$ that the paper recovers, plus the correction to the constant $D_2$.","marker":"Giménez and Noy (2009)"},{"why":"Shows that the 2-connected core of $P_n$ is asymptotically a mixture of weighted 2-connected planar graphs with vertex weight $\\rho_B$, and supplies the local limit for the largest block size used to transfer convergence to $B(P_n)$.","marker":"Giménez et al. (2013)"},{"why":"Establishes the quenched local convergence of vertex-weighted random planar maps, the starting convergence that is then passed down through the successive cores.","marker":"Stufler (2019b)"},{"why":"Supplies the Gibbs-partition convergence results used to pass from a random compound structure to its giant component at each connectivity layer.","marker":"Stufler (2018)"},{"why":"Provides the local limit theorems for the maximal offspring and fringe subtrees of subcritical Galton-Watson trees, which drive the condensation analysis of the tree encodings.","marker":"Stufler (2019a)"},{"why":"Contains the quenched inductive argument for neighbourhood probabilities and the block-weighted graph result used to assemble connected planar graphs from their 2-connected core.","marker":"Stufler (2016)"},{"why":"Supplies the basic theory of simply generated trees as conditioned Galton-Watson trees and the condensation facts formalized in Lemma 3.1.","marker":"Janson (2012)"},{"why":"Provides the big-jump large-deviation result for random walks under subexponentiality that converts condensation asymptotics into asymptotic coefficient estimates.","marker":"Denisov et al. (2008)"},{"why":"Provides the subexponential density results used in the Gibbs-partition and random-product-structure asymptotics throughout Sections 5 and 9.","marker":"Foss et al. (2013)"}],"fun_headline_variants":["Random planar graphs converge to a new infinite local limit","Uniform planar graphs: a novel infinite graph is the local limit","Local limit of planar graphs is a new infinite graph","New UIPG: local limit of uniform random planar graphs","Quenched local limit for planar graphs gives a new infinite graph"],"cache_read_input_tokens":58368,"weakest_assumption_plain":"The whole condensation route for planar graphs rests on the strict inequality $\\nu_C < 1$, where $\\nu_C$ is a constant built from the generating series of 2-connected planar graphs; the paper checks this with approximate constants from an earlier enumeration paper, without rigorous error bounds, so if the true value reached 1 the condensation step and the main convergence argument would fail.","fun_headline_variants_meta":{"raw":{"variants":["Random planar graphs converge to a new infinite local limit","Uniform planar graphs: a novel infinite graph is the local limit","Local limit of planar graphs is a new infinite graph","New UIPG: local limit of uniform random planar graphs","Quenched local limit for planar graphs gives a new infinite graph"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00059,"raw_usage":{"total_tokens":2754,"prompt_tokens":914,"completion_tokens":1840,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":530,"completion_tokens_details":{"reasoning_tokens":1758}},"tokens_in":530,"tokens_out":1840,"duration_ms":14118,"temperature":1.0,"reasoning_tokens":1758,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:31:21.636514+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute certified interval bounds for $\\nu_C = \\rho_B \\, \\partial^2 B/\\partial x^2(\\rho_B,1)$ using rigorous interval arithmetic on the singular expansion of the 2-connected planar graph generating series; a certified lower bound at least 1 would disprove the condensation premise and invalidate the proof of Theorem 1.1, while a certified upper bound below 1 would close the remaining numerical gap.","supporting_citations":[{"cited_title":"Asymptotic enumeration and limit laws of planar graphs","cited_arxiv_id":null,"evidence_quote":"Provides the target asymptotic formula for planar graphs and the analytic expressions for $\\rho_C$ and $c_G$ that the paper recovers, plus the correction to the constant $D_2$."},{"cited_title":"Dieker and Vsevolod Shneer","cited_arxiv_id":null,"evidence_quote":"Provides the big-jump large-deviation result for random walks under subexponentiality that converts condensation asymptotics into asymptotic coefficient estimates."}],"review_version":1}