{"id":"3c367a3d-6970-4140-8f51-f10c7486a960","arxiv_id":"1908.08857","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Maximal geometric thrackles without isolated vertices can have edge-vertex ratio arbitrarily close to 1/2, and there is an infinite family of maximal topological thrackles with ratio 5/6.","lead":"This paper studies 'thrackles', graph drawings in which every pair of edges meets exactly once, and asks how sparse a maximal such drawing can be. It shows that maximal geometric thrackles can have an arbitrarily small edge-to-vertex ratio, and constructs an infinite family of maximal topological thrackles with ratio exactly 5/6.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3 rests on the unproven simultaneous geometric interlacement claim of the Kynčl belt construction; if the copies cannot be placed without extra or missing intersections, the construction is not a thrackle and the theorem collapses.","rationale":"I read the manuscript in good faith. The combinatorial structure is coherent: the duplication step gives a thrackled cycle C_{4n+2}, the count of added vertices and edges gives ratio 5/6, and the maximality bookkeeping in Lemmas 2–7 and Property 3 is internally consistent assuming T2 is a thrackle. I found no independent algebraic or count-based error. The soft spot is exactly the geometric feasibility of simultaneously placing all Kynčl copies so that every pair of edges intersects precisely once. This is the reader's weakest assumption, and I agree with it. A concrete small-case instantiation or a formalization of the interlacement would settle whether the construction is realizable; until then the conditional verdict is appropriate. Section 5's caveat that iterating toward 4/5 is ongoing does not weaken Theorem 3 itself, since the theorem is only about the single belt construction.","tokens_in":18718,"tokens_out":23566,"duration_ms":244179,"concrete_test":"For the smallest case, n=2, take T1 to be C10 and implement the belt construction literally: place T1 as the doubled star-shaped thrackle, draw each Kynčl copy as offset curves in thin tunnels, and check with an arrangement intersection routine (for example, spline curves with CGAL or a custom arrangement checker) that every pair of the 50 edges of T2 intersects exactly once. If extra or missing intersections appear, the construction as written is not realizable. As an analytic cross-check, write down the prescribed cyclic sequence of bundle crossings for one edge of K_e and test whether that Gauss word is realizable by a simple curve.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim, Theorem 3, depends on the Kynčl belt construction in Section 4. In the paragraph following Figure 17, the paper asserts that for each edge e of T1 a copy K_e of Kynčl's example can be drawn in a thin tunnel around e so that: (i) each of its four edges crosses the adjacent edges g and f in the required vertex vicinities and passes through every other edge of T1 exactly once, and (ii) edges from different Kynčl copies cross one another exactly once. The text says \"This ensures three facts\" and then proceeds, but this is a simultaneous geometric existence claim, not a proof. No coordinates, explicit construction order, or argument that the local drawings in the vertex vicinities can be glued without unintended additional crossings is given. Lemmas 2–7 and Property 3 all presuppose this drawing of T2; if some pair of edges of T2 intersects twice, or if some edge of K_e misses an edge of T1 or of another copy, then T2 is not a thrackle, and the maximality argument has no object to apply to. The figures are suggestive evidence, but the load-bearing assertion is exactly the faithful realization of the belt interlacement.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies maximal thrackles, drawings of graphs in which every pair of edges intersects exactly once (at a common vertex or at a proper crossing), and investigates the possible values of the edge-vertex ratio ε(T)=|E|/|V|. It proves three existence results: (1) geometric maximal thrackles can have arbitrarily small ε, and, if isolated vertices are forbidden, can have ε arbitrarily close to the handshaking bound 1/2; (2) topological maximal thrackles with isolated vertices can have arbitrarily small ε; and (3) there is an infinite family of maximal topological thrackles without isolated vertices with ε exactly 5/6. The main construction in Theorem 3 starts from a star-shaped drawing of an odd cycle, duplicates vertices and edges to obtain a maximal thrackled cycle T1, and then applies a 'Kynčl belt construction' that attaches a copy of Kynčl's four-edge maximal thrackle to each edge of T1. The resulting graph T2 is shown to be maximal by a sequence of rerouting lemmas and structural properties (Lemmas 2–7, Properties 1–3).","tokens_in":18913,"tokens_out":19189,"duration_ms":184825,"significance":"If Theorem 3 is accepted, it is a valuable extremal result: apart from the trivial K1,1, it gives the first infinite family of maximal thrackles without isolated vertices whose edge-vertex ratio is strictly below 1, and it does so through a flexible gadget construction that may be adaptable to other saturation questions. The paper also contains a self-contained proof that Kynčl's example is maximal and a detailed proof that the duplicated cycle T1 is maximal, along with an independent verification of the specific case where Conway's conjecture for n≤12 is invoked. The constructions are explicit and the intermediate lemmas are clearly stated. The main weakness is that the Kynčl belt construction is described informally, with reference to figures, rather than by a formal existence proof, and the later lemmas inherit this informality. The paper does not include machine-checked proofs, but the case analysis is extensive and appears internally coherent.","major_comments":[{"comment":"The Kynčl belt construction is not proved to exist. The text states that the copies K_e are drawn in thin tunnels around each edge e and that 'This ensures three facts', but no argument is given that a simultaneous drawing with the required intersection pattern is realizable. In particular, the claims that each edge of K_e intersects each edge of K_f and K_g precisely once, and that each edge of K_e intersects each edge of every remaining Kynčl copy exactly once, are global assertions about the interaction of 4|E(T1)| new curves. The existence of pairwise disjoint vertex vicinities and of tunnels with the required crossing behavior needs a proof or at least a constructive ordering, such as an ε-tunnel argument. As written, Theorem 3 and all subsequent lemmas rest on an unverified geometric hypothesis.","section":"Section 4, paragraph after Figure 17"},{"comment":"The proofs of these results depend on the precise local layout of the Kynčl copies inside the vertex vicinities, but this layout is described only by figures and informal phrases such as 'as illustrated in Figure 17' and 'the red-shaded region in Figure 19'. For example, Lemma 3 uses regions R, L, and G without textual definitions, and Property 3 refers to the triangular region T_u that is only shown in Figure 11. Since these lemmas establish maximality of T2, the authors should provide a combinatorial description of the local drawing in each vicinity, including the cyclic order of edges around each vertex and the sectors through which edges leave the disk, so that the case analyses can be checked independently of the figures.","section":"Section 4, Lemmas 2–7 and Properties 1–3"},{"comment":"The maximality proof of T1, though extensive, is not fully formal. In Case 1, the region R and the face C are not precisely defined, and the claim that the new edge crosses the boundary of R an even number of times 'since it contains C' is stated without proof. Similar issues appear in Cases 3 and 4 with the definitions of lower, middle, and upper parts of edges. Because Proposition 2 is used in the concluding step of Theorem 3 via Property 3, the argument should be made fully rigorous, for example by defining the relevant regions explicitly and justifying the parity or crossing claims.","section":"Section 4, Proposition 2"}],"minor_comments":[{"comment":"In the discussion of k-simple graphs, 'th k-simple property' should read 'the k-simple property'.","section":"Introduction"},{"comment":"The phrase 'maximal trackles' appears twice; it should be 'maximal thrackles'.","section":"Section 5"},{"comment":"The phrase 'we apply the usual modification for removing multiple edge crossings' is vague; please specify how the modification works and why it preserves the thrackle condition.","section":"Lemma 3"},{"comment":"In the sentence about replacing sections of s, 'close to the boundary of DU' should be 'close to the boundary of D_u'.","section":"Property 3"},{"comment":"The argument uses the fact that a thrackle cannot contain a 4-cycle without proof or citation; this is a standard consequence of Woodall's characterization of thrackled cycles and should be stated explicitly.","section":"Theorem 2, direct proof in Case 3"},{"comment":"The sentence 'It is clear that by adding any number of segments in this way, we obtain a thrackle' is not fully justified; a short explanation of why the new segments intersect each other exactly once would be helpful.","section":"Section 2, proof of Theorem 1(b)"}],"recommendation":"major_revision","confidential_remarks":"The main gap is the missing proof of the geometric realizability of the Kynčl belt construction. I believe this is fixable by adding a standard thin-tunnel lemma, so I am not recommending rejection. The paper otherwise makes a solid contribution to the study of saturated thrackles. One concern for the editor: the manuscript contains an 'Ongoing work' section that is unusual for a journal paper; the authors should decide whether to present the 5/6 result alone or to frame the iteration conjecture purely as an open problem. The figures are essential and should be made unambiguous in the revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this paper has three genuinely new extremal constructions and one under-proved load-bearing geometric assertion. The constructions are a butterfly giving arbitrarily small edge-vertex ratios for geometric thrackles, a 6-cycle with isolated vertices for topological thrackles, and the Kynčl belt construction giving an infinite family of maximal topological thrackles without isolated vertices at ratio 5/6. The doubled odd cycle T1 is also proved maximal, which is a necessary scaffolding. The edge-vertex ratio computations are correct, and the maximality proofs are mostly careful case analyses. I spot-checked the easy parts and they hold.\n\nThe soft spot is Section 4's belt construction. The paper describes placing a copy of Kynčl's example in a thin tunnel around each edge of T1 and then says \"This ensures three facts.\" The first of those facts, which is load-bearing, is that every edge of every copy crosses every edge of every other copy exactly once, and also crosses T1 exactly once. That is asserted rather than proved. There are no coordinates, no explicit gluing order, no argument ruling out double crossings inside the vertex vicinities or along the tunnels. The figures are suggestive and I suspect the claim is true, but a referee should ask for a formalized version: either a lemma with explicit local drawings and a crossing-number check, or a more careful inductive construction. Lemmas 2-7 and Property 3 are coherent, but they presuppose that T2 exists as a thrackle, so they do not repair the gap.\n\nOther notes: Theorem 2 relies on Conway's conjecture for n≤12, but the authors include an independent direct proof, which is good. Section 5 honestly states that the iteration toward 4/5 is still incomplete. The citation pattern is fine; relevant prior work is cited.\n\nBottom line: this is a real contribution if the belt construction is formalized. The paper deserves a serious referee, and I would send it to review with a request to tighten Section 4. I would bring it to a reading group if people are interested in extremal graph drawings.","headline":"New extremal constructions for maximal thrackles, with the 5/6 family resting on a plausibly true but under-formalized belt construction.","tokens_in":19518,"tokens_out":2797,"would_cite":true,"duration_ms":29825,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C10","05C62","05C35"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper constructs maximal thrackles with arbitrarily small edge-vertex ratio in the geometric case, and an infinite family without isolated vertices with ratio exactly 5/6 in the topological case.","keywords":["thrackle","maximal thrackle","edge-vertex ratio","thrackle conjecture","geometric thrackle","topological thrackle","belt construction","saturated drawing"],"falsifier":"Take the smallest case of the construction, $n=2$, for which $T_1$ is a 10-cycle, build an explicit drawing of $T_2$, and check every potential new edge between nonadjacent vertices: if any curve joins two such vertices while crossing every edge of $T_2$ exactly once, then $T_2$ is not maximal and Theorem 3 collapses. A systematic search over rotation systems of the 10-vertex drawing would settle the check.","tokens_in":18490,"feed_emoji":"📐","tokens_out":10443,"duration_ms":95932,"temperature":0.7,"pith_summary":"The paper asks how sparse a maximal thrackle can be: a thrackle is a drawing in which every pair of edges meets exactly once, and maximal means no edge between existing vertices can be added without breaking that condition. It shows that in the straight-line (geometric) setting the edge-vertex ratio can be made arbitrarily small by adding isolated vertices, and it can be pushed arbitrarily close to the natural lower bound of $1/2$ when isolated vertices are forbidden. For topological thrackles, isolated vertices again let the ratio tend to zero. The main theorem constructs an infinite family of maximal topological thrackles without isolated vertices whose edge-vertex ratio is exactly $5/6$. This matters because it shows that maximality does not force a thrackle to be dense, and it gives the lowest nondegenerate ratio currently exhibited toward the question of how low the ratio can go.","feed_headline":"Maximal thrackles can be built with edge-vertex ratio 5/6","feed_subtitle":"Sparse drawings that cannot accept one more edge sharpen the context around the thrackle conjecture.","key_machinery":"The load-bearing mechanism is the belt construction: for each directed edge $e=uv$ of a cycle thrackle, place a copy of a four-edge, six-vertex local example with its vertices split between small disks around $u$ and $v$, and route every edge of the copy in a thin tunnel along $e$ so that it crosses all edges of the original drawing exactly once, while the two copies attached to consecutive edges also cross each other exactly once. The construction is engineered so that exactly four new edges and five new vertices are added per original edge, which fixes the edge-vertex ratio at $5/6$. The second mechanism is the maximality transfer encoded in Property 3: if a new edge can be added to the inflated drawing, it can be rerouted step by step until its endpoints lie in the underlying cycle, so maximality of the inflated drawing follows from maximality of the cycle.","core_discovery":"The central discovery is the infinite family of Theorem 3: there exist maximal topological thrackles without isolated vertices with edge-vertex ratio exactly $5/6$. The proof begins with a star-shaped drawing of the odd cycle $C_{2n+1}$, duplicates every vertex and edge to obtain a maximal thrackle $T_1$ on the cycle $C_{4n+2}$, and then applies the belt construction to every edge of $T_1$. For each edge it places a copy of a fixed four-edge, six-vertex local example in a thin tunnel around that edge, interlacing the copy with the edge and its two neighbours so that every new edge crosses every other edge of the drawing exactly once. Each original edge survives and gains four new companion edges, while five new vertices are introduced, so the ratio $5/6$ follows by counting. The paper proves maximality of $T_2$ by a rerouting argument: any hypothetical new edge in $T_2$ can be locally rerouted to one whose endpoints belong to $T_1$, contradicting the known maximality of $T_1$.","pith_inferences":["The paper states the iterated belt construction as ongoing work; this reader's extrapolation is that the exact counting would give $4/5$ per additional round, but the difficulty is the maximality transfer, not the ratio.","The rerouting strategy suggests a reusable design principle: if a small, non-extendable local drawing is placed in a tunnel around each edge of a maximal thrackle, and every hypothetical new edge can be pulled back into the underlying cycle, then the inflated drawing inherits maximality. Testing this principle on other local modules could produce ratios below $5/6$.","A natural next experiment is to replace the four-edge, six-vertex local module by other non-extendable drawings with fewer edges per vertex and check whether the belt construction still closes; the paper does not explore this."],"forward_implications":["Maximality does not force a thrackle to have as many edges as vertices; maximal topological thrackles without isolated vertices can have ratio $5/6$, and geometric ones can approach $1/2$.","The lower bound $1/2$ from the handshaking lemma is asymptotically tight for maximal geometric thrackles without isolated vertices.","Adding isolated vertices is enough to drive the ratio to zero in both geometric and topological settings, so any positive lower bound for maximal thrackles must exclude isolated vertices.","The belt construction preserves maximality while inflating the edge count, giving a local operation that builds larger maximal thrackles from smaller ones.","Iterating the same construction on the original edges is proposed in the paper as a route toward ratios approaching $4/5$, which would leave the gap between $1/2$ and $5/6$ open."],"supporting_citations":[{"why":"States the conjecture that every thrackle has at most as many edges as vertices, which frames the paper's investigation of the other extreme.","marker":"[6]"},{"why":"Introduces saturated k-simple graphs and motivates the study of maximal thrackles with few edges.","marker":"[10]"},{"why":"Provides the four-edge, six-vertex local example used as the module in the belt construction, including its non-extendability.","marker":"[11]"},{"why":"Uses the same local example in a related saturation problem, supporting the construction's context.","marker":"[12]"},{"why":"Proves the thrackle conjecture for straight-line (geometric) drawings, supplying the setting for Theorem 1.","marker":"[17]"},{"why":"Verifies the conjecture up to twelve vertices, used to rule out chords in the six-vertex cycle in the proof of Theorem 2.","marker":"[18]"},{"why":"Gives the classical characterization of thrackles and the odd-cycle drawings that underlie the doubling construction.","marker":"[21]"}],"fun_headline_variants":["Infinite maximal thrackle family hits 5/6 edge-vertex ratio","Edge-vertex ratio 5/6 for maximal topological thrackles","Maximal thrackles reach 5/6 edge-vertex density","Ratio 5/6 achieved in infinite maximal thrackle class"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction assumes that all these small copies can be drawn simultaneously in thin tunnels around the edges so that every pair of edges crosses exactly once and no unintended intersections appear; the paper demonstrates the required interlacing in figures but does not give a formal proof that the simultaneous placement is always achievable.","fun_headline_variants_meta":{"raw":{"variants":["Infinite maximal thrackle family hits 5/6 edge-vertex ratio","Edge-vertex ratio 5/6 for maximal topological thrackles","Maximal thrackles reach 5/6 edge-vertex density","Ratio 5/6 achieved in infinite maximal thrackle class"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000236,"raw_usage":{"total_tokens":1493,"prompt_tokens":921,"completion_tokens":572,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":537,"completion_tokens_details":{"reasoning_tokens":492}},"tokens_in":537,"tokens_out":572,"duration_ms":5252,"temperature":1.0,"reasoning_tokens":492,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:31:52.858561+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the smallest case of the construction, $n=2$, for which $T_1$ is a 10-cycle, build an explicit drawing of $T_2$, and check every potential new edge between nonadjacent vertices: if any curve joins two such vertices while crossing every edge of $T_2$ exactly once, then $T_2$ is not maximal and Theorem 3 collapses. A systematic search over rotation systems of the 10-vertex drawing would settle the check.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"States the conjecture that every thrackle has at most as many edges as vertices, which frames the paper's investigation of the other extreme."},{"cited_title":"Journal of Graph Algorithms and Applications 22(1), 117–138 (2018)","cited_arxiv_id":null,"evidence_quote":"Introduces saturated k-simple graphs and motivates the study of maximal thrackles with few edges."},{"cited_title":"The American Mathe- matical Monthly 118(6), 544–548 (2011)","cited_arxiv_id":null,"evidence_quote":"Proves the thrackle conjecture for straight-line (geometric) drawings, supplying the setting for Theorem 1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Verifies the conjecture up to twelve vertices, used to rule out chords in the six-vertex cycle in the proof of Theorem 2."},{"cited_title":"Combinatorial Mathematics and its Applications pp","cited_arxiv_id":null,"evidence_quote":"Gives the classical characterization of thrackles and the odd-cycle drawings that underlie the doubling construction."}],"review_version":1}