{"id":"7d3bb506-41d4-43f5-a880-c07f2df7f7ec","arxiv_id":"2502.03324","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Split Lagrangian tori in non-monotone S^2 × S^2 are Hamiltonian isotopic exactly when their base points lie on the same good billiard trajectory in a rectangle.","lead":"This paper classifies Lagrangian tori that are products of two circles in the four-manifold S^2 × S^2 with any non-monotone symplectic form, showing each equivalence class is encoded by the bouncing points of a billiard trajectory in a rectangle. The billiard reformulation yields answers to open questions on Lagrangian packing, ball embeddings, and the topology of the space of Lagrangian tori.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'only if' direction of Theorem 1.3 is delegated to [7, Theorem 4.7], an unstated to-appear result; until that theorem is verified for non-monotone X_alpha, the classification is conditionally supported.","rationale":"Applies good-faith reading. Theorem 1.3 is a biconditional; Lemmas 2.2 and 2.4 give the 'if' direction constructively and are internally consistent, including the folding-map algebra and the corner-reflection convention. The 'only if' direction, however, is not self-contained: Lemma 2.7 cites [7, Theorem 4.7] for the key step that a Hamiltonian map preserves the distinguished disk class. Without that step, the linear algebra in equations (28)-(31) has no starting point, and the Chekanov invariants stated in Theorem 2.6 are visibly incomplete (on Q, Gamma is dense for irrational y). Thus the classification's arithmetic is only as secure as an external to-appear theorem. This is a verifiability risk, not an internal inconsistency. I also checked the apparent 'only points with #d=3' claim in Lemma 2.7: the points (+/-1,0) also have #d=3, but the proof treats the whole y=0 segment separately, so this typo is not load-bearing. The applications and the explicit disclosure of unresolved monodromy cases (55) and (58) are honest and do not affect Theorem 1.3. Recommendation: CONDITIONAL rather than REJECT or UNCHANGED, because the construction side is solid and the obstruction is likely correct, but the unique classification should not be accepted as fully verified until [7, Theorem 4.7] is confirmed to apply in this setting.","tokens_in":31640,"tokens_out":12125,"duration_ms":119237,"concrete_test":"When [7] is publicly available, verify the exact statement and proof of Theorem 4.7: (i) check that its hypotheses cover X_alpha for every alpha > 0 with moment polytope square_alpha = [-1-alpha, 1+alpha] x [-alpha, alpha]; (ii) confirm that it indeed forces phi_*D1 = D'1 for p,q in Q, and the analogous assertion for the two-distinguished-class case on [-1,1] x {0}; (iii) check that no monotonicity or rationality assumption is hidden. If the theorem is not yet available, the only-if direction of Theorem 1.3 remains unverified and should be treated as conditional.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Lemma 2.7 is the only proof of the uniqueness half of the main classification. In the generic case (x,y),(x',y') in Q, after the first Chekanov invariant gives y'=y, the rest of the argument is carried by [7, Theorem 4.7]: this theorem is used to assert that a Hamiltonian diffeomorphism mapping T(x,y) to T(x',y) sends the distinguished relative homology class D1 to D'1, rather than to a combination of the D'_i. That assertion is what converts the area computation into the arithmetic relation x = ±x' + 2k1 + 2k2y in equations (30)-(31). The first and third Chekanov invariants from Theorem 2.6 are not sufficient here: in Q they yield y'=y and, when y is irrational, a dense subgroup Gamma, so they cannot separate the x-values. [7, Theorem 4.7] is not stated or proved in this paper, its hypotheses are not given, and [7] is listed as 'to appear'. If the theorem fails or does not apply to the non-monotone polytope square_alpha for alpha > 0, or to the boundary-like set Sigma, the billiard criterion over-classifies. The construction half (symmetric probes and folding) is self-contained; the soft spot is specifically the external obstruction that forces phi_*D1 = D'1.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper classifies, up to Hamiltonian isotopy, the split Lagrangian tori T(x,y)=S^1_x × S^1_y in S^2 × S^2 equipped with the non-monotone split symplectic form ωα. The main result (Theorems 1.3 and 2.5) states that two split tori are Hamiltonian isotopic if and only if a certain billiard trajectory in a rectangle, defined by the base point, hits one of the obvious reflected points as an admissible bouncing point; this is refined to an arithmetic condition y′=y and x=±x′+2k1+2k2y. The construction direction is proved by symmetric probes and a folding/unfolding argument. The obstruction direction is proved using Chekanov invariants and a relative-homology result, [7, Theorem 4.7], from the first author's earlier work. The paper then derives several applications: infinite packing numbers, counterexamples to Chassé–Leclercq conjectures on the space of Lagrangians, a partial determination of the Hamiltonian monodromy group, and a characterization of which split tori are images of Clifford, Chekanov, or nonmonotone product tori under symplectic ball embeddings.","tokens_in":31875,"tokens_out":5046,"duration_ms":47254,"significance":"If the main theorem is fully justified, this is a substantial contribution: it gives the first complete classification of split tori in a non-monotone four-dimensional symplectic manifold, it connects the problem to an elegant and explicit billiard model, and it yields a number of immediate applications, including answering a question of Polterovich–Shelukhin on packing numbers and determining ball-Chekanov, ball-Clifford, and ball-nonmonotone split tori. The paper is also commendably honest: it explicitly states that the Hamiltonian monodromy group is not determined in cases (55) and (58) and gives a cautionary example in §3.5. The billiard constructions are self-contained and the arithmetic reductions are presented with machine-checkable precision. The main weakness is that the 'only if' direction of the central classification is not self-contained: it relies on an unpublished theorem from [7] whose hypotheses are not stated.","major_comments":[{"comment":"The proof of the 'only if' direction of Theorem 2.5 uses [7, Theorem 4.7] to assert that a Hamiltonian diffeomorphism mapping T(x,y) to T(x′,y) sends the distinguished class D1 to D′1 (and, on Σ, controls the permutation of distinguished classes). This assertion is the load-bearing step that converts the area computation into the arithmetic relations (30)–(31). However, the theorem is not stated in the paper, its hypotheses are not given, and [7] is listed as 'to appear'. The first and second Chekanov invariants from Theorem 2.6 are not sufficient for this step: in Q they give y′=y and, for irrational y, a dense subgroup Γ(p)=Γ(q), which does not separate different x-values. The authors should either state [7, Theorem 4.7] with its precise hypotheses and verify those hypotheses for the non-monotone polytope □α, or provide a self-contained proof of the needed assertion. Without this, the 'only if' direction of the classification, and every application that relies on it, is conditional on an unverified external theorem.","section":"§2.3, Lemma 2.7"},{"comment":"The Hamiltonian monodromy group is not completely determined in cases (55) and (58), where only inclusions are proved; this is explicitly acknowledged in §1.2 and §3.5 and is a legitimate partial result. However, the cases that are determined also depend on Lemma 3.1, which in turn relies on Lemma 2.7 and hence on [7, Theorem 4.7]. The paper should state clearly that the monodromy classification for (x,y) ∈ Q inherits the same external dependence, so that the reader can separate the self-contained probe constructions from the conditional obstructions.","section":"§3.4, Theorem 3.5"}],"minor_comments":[{"comment":"In the definition of the folding map Fr, the reader is told only that (m,n) ∈ Z2 is chosen so that Fr(x,y) ∈ □r. It would be clearer to specify the admissible values of m and n for each (x,y), since the well-definedness of Fr at the preimages of the boundary is a subtle point that the current wording leaves implicit.","section":"§2.2, Definition 2.3"},{"comment":"The counting argument for b(x,y) in the rational case y=p/q is compressed in the sentence about the segment [0,1/q] containing exactly one admissible bouncing point. A short derivation, or at least a reference to equation (48), would help the reader verify the boundary cases p′≡p+q mod 2 and p′≢p+q mod 2.","section":"§5.2, Proposition 5.3"},{"comment":"In the discussion of symmetric probes, the paper states that for |k|>2 the probes of slope (k,−1) are redundant for the classification. Since this is used to justify restricting to the four slope types in (34), a one-sentence explanation of why these probes are redundant would be useful.","section":"§2.1"}],"recommendation":"major_revision","confidential_remarks":"The main theorem and many of the applications rest on [7, Theorem 4.7], which is cited as 'to appear' and is not stated in this manuscript. This is a serious dependency for a central load-bearing step. I recommend that the editors require the authors to either include a precise statement of that theorem and verify its hypotheses for the non-monotone setting, or provide a proof of the needed assertion in an appendix. The paper is otherwise well-written, the constructions are clear, and the applications are interesting, so the result seems worth publishing once this external dependence is resolved."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know: this paper is strong. It classifies split Lagrangian tori in non-monotone S^2 × S^2 up to Hamiltonian isotopy, via a billiard criterion: two split tori are equivalent iff the corresponding good billiard trajectory has the other's base point as an admissible bouncing point (Theorem 1.3). The refinement to an arithmetic condition (Theorem 2.5) is practical, and the applications are genuinely new: infinite packing numbers for generic split tori, counterexamples in dimension four to the Chassé–Leclercq Conjectures A and B, and a complete ball-Clifford/Chekanov/nonmonotone trichotomy. The construction half, via symmetric probes and the folding map, is self-contained and convincing. The paper is also unusually candid: it explicitly says where its methods fail, namely the Hamiltonian monodromy cases (55) and (58) and the cautionary example T(0,1).\n\nThe main soft spot is the 'only if' direction. Lemma 2.7, which gives uniqueness, relies on [7, Theorem 4.7] to force the distinguished relative homology class D1 to map to D'1. That theorem is not stated here, its hypotheses are not repeated, and [7] is listed as 'to appear'. If that theorem does not cover non-monotone X_alpha or the boundary set Sigma, the classification would over-classify. This is a verifiability problem rather than an internal contradiction. The self-contained constructions and the first author's record make me think the result is likely correct, but a referee will need access to [7] or an explicit statement of Theorem 4.7. Minor point: Lemma 2.7 says the points with #d=3 are (0,1) and (0,-1), which looks like a typo for (1,0) and (-1,0).\n\nWho is this for? Symplectic topologists working on Lagrangian tori, toric fibres, and packing questions. It deserves a serious referee. For publication I would ask the authors to state the needed result from [7] or make the dependency explicit in the introduction, and to fix the typo. I would engage with the paper.","headline":"Strong classification result with a real dependency on a to-appear theorem for the obstruction half; the billiard framing is new and the applications are clean.","tokens_in":32492,"tokens_out":3042,"would_cite":true,"duration_ms":26503,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53D12","53D35","37D50"],"pacs":[],"model":"deepseek-v4-flash","headline":"Billiard paths classify product tori in $S^2 \\times S^2$ up to Hamiltonian isotopy.","keywords":["Lagrangian tori","Hamiltonian isotopy","S^2 × S^2","toric fibres","mathematical billiards","symmetric probes","Chekanov invariants","Lagrangian packing numbers"],"falsifier":"Find an explicit Hamiltonian diffeomorphism of $X_\\alpha$ sending $T(x,y)$ to $T(x',y')$ where $(\\pm x', \\pm y')$ is not an admissible bouncing point of the good billiard trajectory of $(x,y)$; the theorem forbids this, so a single such map would falsify it. Short of that, compute a Hamiltonian-isotopy invariant not derived from [7], such as displacement energy or a filtered Floer group, on two points satisfying $y' = y$ and $x = \\pm x' + 2k_1 + 2k_2 y$ and check whether the invariant distinguishes them.","tokens_in":1963,"feed_emoji":"🎱","tokens_out":2293,"duration_ms":69133,"temperature":0.7,"pith_summary":"The paper classifies, up to Hamiltonian isotopy, all Lagrangian tori in $S^2 \\times S^2$ (equipped with a non-monotone split symplectic form) that split as a product of one circle in each factor. It claims that two such split tori are Hamiltonian isotopic exactly when one of their base points appears as an admissible bouncing point of a certain billiard trajectory in a rectangle, and in the main region this reduces to the arithmetic condition $y' = y$ and $x = \\pm x' + 2k_1 + 2k_2 y$ for integers $k_1, k_2$. This yields a complete answer for split tori, shows the answer is independent of the parameter $\\alpha > 0$, and supplies applications to Lagrangian packing numbers, the space of Lagrangian tori, Hamiltonian monodromy, and the question of which tori arise from symplectic ball embeddings.","feed_headline":"Billiards decide which split tori are isotopic","feed_subtitle":"One billiard trajectory determines when product tori in S^2 x S^2 are Hamiltonian isotopic.","key_machinery":"The billiard picture is the central organizing object: a split torus $T(x,y)$ is represented by a point in the moment rectangle $\\square_\\alpha$, and its equivalence class is read off from the trajectory of slope $\\pi/4$ that bounces in the rectangle $\\square_r$ for $r = \\max\\{|x|-1, |y|\\}$, with corner collisions reflected back along the same line. On the construction side, symmetric probes—segments in the Delzant polytope that intersect the boundary integrally transversely and lift to Hamiltonian isotopies of toric fibres—turn each billiard bounce into an actual Hamiltonian equivalence. On the obstruction side, the paper uses the Chekanov invariants from [7] and the relative second-homology classes $D_i$ of the four obvious disks bounding the circles in the product, whose symplectic areas are the integral affine distances to the facets; preserving these areas under a Hamiltonian diffeomorphism produces the arithmetic condition.","core_discovery":"The central result is Theorem 1.3: for split tori $T(x,y)$ and $T(x',y')$ in $X_\\alpha$, Hamiltonian isotopy holds if and only if one of the points $(\\pm x', \\pm y')$ is an admissible bouncing point of the good billiard trajectory of $(x,y)$ in the rectangle $\\square_{r(x,y)}$. In the region $Q \\cup \\Sigma$, Theorem 2.5 sharpens this to the arithmetic criterion $y' = y$ and $x = \\pm x' + 2k_1 + 2k_2 y$ for some integers $k_1, k_2$. The equivalence is proved in both directions: consecutive bouncing points lie on symmetric probes at equal distance from the boundary, so each billiard bounce is realized by a Hamiltonian isotopy, while Chekanov-type invariants together with a relative-homology computation force any Hamiltonian diffeomorphism between the tori to satisfy the arithmetic condition.","pith_inferences":["The paper leaves implicit that the same billiard combinatorics should describe toric fibres of the even Hirzebruch surfaces, since those toric structures are related to $S^2 \\times S^2$ by mutations; the authors note the mutation argument as beyond the present scope.","The open monodromy cases (55) and (58), together with the cautionary example $T(0,1)$ in Section 3.5, suggest that corner-hitting billiard trajectories carry the remaining complexity; a modified billiard rule that tracks corner bounces explicitly may close the gap.","The explicit counting function $b(x,y)$ from Proposition 5.3 gives a toric packing number that lower-bounds the true Lagrangian packing number, so comparing it with Floer-theoretic upper bounds in the rational cases could reveal where rigidity resumes."],"forward_implications":["Every split torus with $r(x,y)$ irrational has infinite Lagrangian packing number, because its good billiard trajectory has infinitely many admissible bouncing points.","There exist split tori whose Hamiltonian orbit is neither $C^\\infty$-closed nor locally path connected, giving dimension-four counterexamples to two conjectures about the space of Lagrangians.","Ball-Chekanov tori in $X_\\alpha$ are never exotic: every such torus is Hamiltonian isotopic to a split torus of the form $T(\\alpha - a - 1, \\alpha - a)$.","The classification is independent of $\\alpha > 0$, so equivalences between split tori persist across all non-monotone split symplectic forms on $S^2 \\times S^2$.","Theorem 1.3 confirms the symmetric-probe conjecture from [7] in the case of non-monotone $S^2 \\times S^2$."],"supporting_citations":[{"why":"Supplies the symmetric-probe isotopy theorem and the Chekanov-type invariants, including Theorem 4.7 on relative homology, used in both directions of the classification.","marker":"[7]"},{"why":"Provides the Chekanov classification of product tori in $\\mathbb{R}^{2n}$ from which the toric Chekanov invariants are derived.","marker":"[12]"},{"why":"Introduces the probe technique for displacing and equivalencing toric fibres that the paper adapts to symmetric probes.","marker":"[28]"},{"why":"Supplies the notion of symmetric probes and the basic Hamiltonian equivalences they induce among toric fibres.","marker":"[2]"},{"why":"Gives the Lagrangian packing upper bound that the paper adapts as Theorem 1.15 and whose sharpness it tests in Corollary 1.16.","marker":"[31]"},{"why":"Establishes connectedness of the space of symplectic ball embeddings into $X_\\alpha$, the key fact behind the ball-embedding results.","marker":"[27]"},{"why":"Shows the symplectomorphism group of $S^2 \\times S^2$ is connected, identifying the symplectomorphism and Hamiltonian classifications.","marker":"[1]"},{"why":"Provides displacement-energy computations used to distinguish product tori over the central segment $[-1,1] \\times \\{0\\}$.","marker":"[19]"}],"fun_headline_variants":["Split tori match iff billiard hits align","Billiard trajectories classify split tori in S2×S2","Hamiltonian tori: billiard test for equality","Isotopy of split tori from billiard bounces"],"cache_read_input_tokens":34560,"weakest_assumption_plain":"The 'only if' direction of the classification depends on the completeness of the Chekanov-type obstructions from [7], a source listed as to appear, and if those invariants fail to distinguish split tori in non-monotone $S^2 \\times S^2$ (including the boundary segments $\\Sigma$), the billiard criterion could identify tori that are not truly Hamiltonian isotopic.","fun_headline_variants_meta":{"raw":{"variants":["Split tori match iff billiard hits align","Billiard trajectories classify split tori in S2×S2","Hamiltonian tori: billiard test for equality","Isotopy of split tori from billiard bounces"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000354,"raw_usage":{"total_tokens":1885,"prompt_tokens":866,"completion_tokens":1019,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":482,"completion_tokens_details":{"reasoning_tokens":951}},"tokens_in":482,"tokens_out":1019,"duration_ms":9196,"temperature":1.0,"reasoning_tokens":951,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T05:08:37.166178+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find an explicit Hamiltonian diffeomorphism of $X_\\alpha$ sending $T(x,y)$ to $T(x',y')$ where $(\\pm x', \\pm y')$ is not an admissible bouncing point of the good billiard trajectory of $(x,y)$; the theorem forbids this, so a single such map would falsify it. Short of that, compute a Hamiltonian-isotopy invariant not derived from [7], such as displacement energy or a filtered Floer group, on two points satisfying $y' = y$ and $x = \\pm x' + 2k_1 + 2k_2 y$ and check whether the invariant distinguishes them.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the symmetric-probe isotopy theorem and the Chekanov-type invariants, including Theorem 4.7 on relative homology, used in both directions of the classification."},{"cited_title":"Chekanov","cited_arxiv_id":null,"evidence_quote":"Provides the Chekanov classification of product tori in $\\mathbb{R}^{2n}$ from which the toric Chekanov invariants are derived."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the probe technique for displacing and equivalencing toric fibres that the paper adapts to symmetric probes."},{"cited_title":"Abreu, M","cited_arxiv_id":null,"evidence_quote":"Supplies the notion of symmetric probes and the basic Hamiltonian equivalences they induce among toric fibres."},{"cited_title":"Polterovich and E","cited_arxiv_id":null,"evidence_quote":"Gives the Lagrangian packing upper bound that the paper adapts as Theorem 1.15 and whose sharpness it tests in Corollary 1.16."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes connectedness of the space of symplectic ball embeddings into $X_\\alpha$, the key fact behind the ball-embedding results."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows the symplectomorphism group of $S^2 \\times S^2$ is connected, identifying the symplectomorphism and Hamiltonian classifications."},{"cited_title":"Fukaya, Y.-G","cited_arxiv_id":null,"evidence_quote":"Provides displacement-energy computations used to distinguish product tori over the central segment $[-1,1] \\times \\{0\\}$."}],"review_version":1}