{"id":"a8b208da-baad-40e4-a813-7f25078faba9","arxiv_id":"2506.21130","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper defines an invariant whose image is completely described, showing infinitely many regular homotopy classes of triple-point-free immersed spheres.","lead":"The paper constructs a numeric invariant that tells apart certain self-intersecting spheres (those with only double, no triple, crossing points) and shows that such spheres fall into infinitely many distinct deformation classes. For anyone studying how surfaces move and transform in 3D space, this provides a sharp new tool and explains when a deformation must create triple crossings.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.4's surjectivity is asserted from figures, not proved: Examples 6.1/6.2 and the connected-sum formula (2) carry the entire image theorem without explicit parametrizations or degree checks.","rationale":"The reader's weakest assumption correctly identifies the surjectivity half of Theorem 1.4 as the least secure part. My reading of the manuscript confirms that Examples 6.1/6.2 and equation (2) are presented pictorially rather than derived. The stress-test adds a precise point: the connected-sum lemma requires a degree-1 polar vertex in every participating surface, and for negative k this is not automatic from F(f_k)=e_k. This is not an accusation of error; it is a genuine gap in the written proof of the central image theorem. If the explicit check succeeds, the paper's claim should be accepted as is; if it fails, the image description must be weakened, even though the infinitude of classes may still survive through the e_k examples. Therefore the appropriate verdict is CONDITIONAL rather than unconditional ACCEPT.","tokens_in":11049,"tokens_out":23717,"duration_ms":307871,"concrete_test":"Independently construct f_3 and g_5 from the generating curves indicated in Figures 18 and 19, compute their double point trees and F values, then form the pole connected sum of Figure 20 and verify both that no triple point is introduced and that F(f_3#g_5)=e_3+(2e_1-e_5)-e_1=e_3+e_1-e_5. If the computed F differs by even one basis vector, or if a triple point appears in any intermediate surface, the surjectivity proof of Theorem 1.4 fails and the image statement must be revised.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of Theorem 1.3 is detailed and self-contained, but the paper's headline image theorem (Theorem 1.4) rests on the realizability portion of Lemma 6.3. That lemma depends on three unexpanded assertions: (i) Examples 6.1 and 6.2 construct, for every odd k, triple-point-free surfaces of revolution f_k and g_k with F(f_k)=e_k and F(g_k)=2e1-e_k; (ii) each such surface has an extremal polar vertex of topological degree δ=1, with the oriented normal pointing into the unbounded component, as required by the connected-sum description; and (iii) the pole connected sum gives exactly F(f#g)=F(f)+F(g)-e1. These are justified only by Figures 18-20 and a short sentence, not by explicit curve parametrizations, double-point-tree computations, or a check that no hidden triple point is introduced. Since every h in U is then realized as an alternating connected sum of these f_k and g_k, a failure in any of (i)-(iii) would make U not the full image. For instance, if f_k with negative k does not admit a degree-1 polar vertex, the sequential gluing in Lemma 6.3 cannot be carried out as written. The infinitude of regular homotopy classes would still follow from the e_k alone, but the exact description of the image would be unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper defines an invariant F for triple-point-free immersed spheres S^2 -> R^3. For a generic immersion f, the double points are encoded in a directed tree G_f with a pairing on edges and a local topological degree δ_f on vertices; F(f) is the integer sequence whose k-th entry is the sum of 1 - deg^-(v) over vertices with δ_f(v)=k. Theorem 1.3 states that F is invariant under regular homotopies through Imm_{<3}(S^2,R^3); the proof reduces to E and H singularities and uses the tree-modification classification of Section 4. Theorem 1.4 characterizes the image of F as the set U of finitely supported integer sequences with zero even entries and total sum 1. The paper also derives consequences: Imm_{<3}(S^2,R^3) has infinitely many regular homotopy classes, and a specific surface of revolution j cannot be regularly homotoped to an embedded sphere without triple points (Theorem 1.7).","tokens_in":11409,"tokens_out":6871,"duration_ms":80602,"significance":"If the main results are correct, this is a significant advance: it upgrades the previously known lower bound of two components of Imm_{<3}(S^2,R^3) to infinitely many, and it gives an explicit, computable invariant that is not of finite order. The construction of the invariant from a double-point tree plus local degrees is natural and elegant, and the invariance proof via a classification of E/H tree modifications is a substantial and mostly self-contained contribution. The inclusion im(F) ⊆ U is proved cleanly from tree identities. The main weakness is that the surjectivity half of Theorem 1.4 rests on unexpanded constructions and a connected-sum formula that are presented through figures rather than proved.","major_comments":[{"comment":"The existence of surfaces of revolution f_k and g_k for every odd k, with F(f_k)=e_k and F(g_k)=2e_1-e_k, is asserted and illustrated only by Figures 18 and 19. No explicit generating curves, parametrizations, or proofs that these surfaces have no triple points are supplied, and no computation of their double-point trees or local degrees is given. Since Lemma 6.3 uses these surfaces to realize every h in U, this is a load-bearing step: the exact image description in Theorem 1.4 is not established without a rigorous construction of f_k and g_k.","section":"Section 6, Examples 6.1 and 6.2"},{"comment":"The connected-sum formula F(f#g)=F(f)+F(g)-e_1 is justified only by the sentence that the double-point tree of f#g is the two trees glued along extremal vertices of degree δ=1, together with Figure 20. This requires proof that the poles used for the connected sum correspond to vertices of topological degree 1, that the gluing does not introduce triple points or additional double-point curves, and that the fused vertex has indegree equal to the sum of the two indegrees. Without these checks, the alternating connected-sum construction in Lemma 6.3 and hence the surjectivity part of Theorem 1.4 are unsupported.","section":"Section 6, equation (2)"},{"comment":"The proof of Lemma 6.3 assumes that every f_k and g_k has an extremal polar vertex of topological degree δ=1 with the oriented normal pointing into the unbounded component of R^3 \\ f(S^2). This property is not verified for the surfaces in Examples 6.1 and 6.2, nor is it proved that the pole connected sum preserves the absence of triple points under repeated gluing. If this polar-vertex condition fails for some k, the sequential gluing ih := fa(1)#gb(1)#... cannot be carried out as written, and the image theorem would require a different argument.","section":"Section 6, Lemma 6.3"}],"minor_comments":[{"comment":"The statement that the invariant 'is therefore necessarily not of finite order' is asserted without proof or reference; please either provide a short argument or soften the claim, since it is not needed for the main theorems.","section":"Introduction, after Theorem 1.4"},{"comment":"The codomain of F is written as Z^Z, while the abstract and Theorem 1.4 describe the image in l^1(Z); this is harmless but the notation could be aligned, for example by defining the subset of finitely supported sequences.","section":"Definition 1.2"},{"comment":"The phrase 'double point tree with incomplete pairing' is used before a formal definition of an incomplete pairing is given; please define this notion explicitly.","section":"Definition 4.3"},{"comment":"The classification of H-singularity tree modifications is presented largely through figures and the phrase 'may (or may not)' for reattachments; a fully formal enumeration of the cases with explicit notation would make the proof easier to verify.","section":"Section 4, Proposition 4.4"},{"comment":"In the H-singularity case, the equation deg^-(v_f)=deg^-(v_g)+deg^-(w_g)-1 is stated without indicating which vertices in Figure 16 correspond to v_f, v_g, and w_g in each of the Hp- and Hs-split cases; labelling the figure would improve readability.","section":"Section 5, proof of Theorem 1.3"},{"comment":"Equation (2) is introduced with a bare reference to Figure 20; adding a numbered display and a short derivation of the change in indegree at the glued vertex would make the argument explicit.","section":"Section 6, equation (2)"}],"recommendation":"major_revision","confidential_remarks":"The referee report from the reader is more optimistic than mine. The invariance theorem and the inclusion im(F) ⊆ U appear sound and are supported by a substantial tree-modification analysis. However, Theorem 1.4 is the paper's headline result, and its surjectivity currently depends on figure-based existence claims and an unproved connected-sum formula. These gaps are fixable within the manuscript's scope by adding explicit descriptions of the generating curves, verifying the polar-vertex condition, and proving equation (2) directly, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: Theorem 1.3 is real new mathematics and the invariance proof holds up; the surjectivity half of Theorem 1.4 is the honest soft spot, and the stress-test note is right that Figures 18-20 and the one-sentence connected-sum formula carry the entire image theorem. The gap is fixable and local, not fatal.\n\nWhat is actually new: F is defined from the directed double-point tree with pairing plus the vertex degree function, and it is explicitly outside the finite-order invariants from Nowik's classification. The proof of invariance is the core of the paper. The E-singularity case is immediate; the H case rests on the Section 4 tree-modification classification, which is intricate and mostly convincing, with a nice use of the RP2 triple-point obstruction in Lemma 4.2. The inclusion im F subset U is completely clean: odd delta kills the even coordinates, the finite tree gives finite support, and |V|-|E|=1 gives total sum 1. And the big consequence, infinitely many regular homotopy classes in Imm<3, only needs the f_k of Example 6.1 plus Theorem 1.3 - it does not need the full image theorem.\n\nWhere it is soft: the surjectivity of Theorem 1.4 is asserted, not proved. Examples 6.1 and 6.2 state, for every odd k, the existence of surfaces of revolution with F(f_k)=e_k and F(g_k)=2e1-e_k, with no generating-curve equations and no computation of the double-point tree. The connected-sum formula (2) is justified in one sentence, and the conditions needed - degree-1 polar vertices with the normal pointing into the unbounded component, no triple point introduced by the neck - are stated but not verified for the specific surfaces. The stress-test's worry is legitimate. I would rate the soundness of the image theorem lower than the soundness of Theorem 1.3. That said, the worst case is not as bad as it sounds: the infinitude result survives on the firmer half, and the missing checks are routine, just not written down.\n\nMinor: Section 4's case analysis is figure-assisted in places (Lemma 4.1's 'odd number of double point curves' step, Lemma 4.2's deformation), so a referee should spend time there. The citation pattern is fine - Max-Banchoff, Nowik, Goryunov, Banchoff, and Smale are all cited in the right places.\n\nVerdict: worth a serious referee. Send it, but ask for explicit parametrizations or a real inductive construction for Examples 6.1-6.2 and a proof of (2). The paper will be stronger for it.","headline":"The invariant and its invariance proof are solid and genuinely new; the surjectivity half of the image theorem is carried by figures and a one-sentence connected-sum assertion, which is a fixable gap in rigor, not a fatal flaw.","tokens_in":11866,"tokens_out":7491,"would_cite":true,"duration_ms":71921,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["57R42","57R45","57M15"],"pacs":[],"model":"deepseek-v4-flash","headline":"Integer-sequence invariant separates triple-point-free immersed spheres into infinitely many classes","keywords":["triple points","double points","double point tree","Gauss code","regular homotopy","sphere eversion","Willmore energy"],"falsifier":"Carry out the construction of Lemma 6.3 for the sequence $h = e_7 + e_{-5} - e_1$: glue the surfaces $f_7$, $g_1$, and $f_{-5}$ along their poles and enumerate the double point tree of the result. If any gluing neck produces a triple point, if any vertex of the resulting tree has even local degree, or if the computed value of $F$ differs from $h$, then the surjectivity half of Theorem 1.4 fails.","tokens_in":2343,"feed_emoji":"🫧","tokens_out":2373,"duration_ms":220114,"temperature":0.7,"pith_summary":"This paper defines an invariant $F$ for immersed spheres in $\\mathbb{R}^3$ that are allowed to cross themselves only in double points, never in triple points. The invariant is a finitely supported integer sequence that survives every deformation avoiding triple points, and the paper proves its image is exactly the set of sequences with zero even entries and total sum $1$. Consequently, the space of such immersions has infinitely many regular homotopy classes, a sharp upgrade from the previously known lower bound of two. A concrete payoff is a topological explanation for why a certain initial surface for the Willmore flow must develop triple points when deforming to a round sphere.","feed_headline":"Infinitely many classes of triple-point-free immersed spheres","feed_subtitle":"A sequence-valued invariant separates these classes; eversions must cross triple points.","key_machinery":"The key object is the double point tree $(G_f, P, \\delta_f)$: the double point curves of a generic triple-point-free immersion are disjoint circles; cutting $S^2$ along them yields a finite tree with an edge for each double point curve, a pairing $P$ that identifies conjugate curves on the two sheets, and a vertex function $\\delta_f$ equal to the sum of topological degrees of the two ambient components adjacent to the vertex's image. The invariant $F$ is the weighted sum $\\sum_{v: \\delta_f(v)=k} (1 - \\deg^-(v))$ at each integer $k$. The proof of invariance hinges on Proposition 4.4, which classifies how a regular homotopy crossing an elliptic or hyperbolic tangency modifies the tree; in particular the H-split operations create a vertex with indegree one less than the sum of two new indegrees, so the contributions at each degree level cancel exactly. Surjectivity relies on Examples 6.1 and 6.2 (surfaces of revolution realizing basis vectors $e_k$ and $2e_1-e_k$) and on the connected-sum formula $F(f\\#g)=F(f)+F(g)-e_1$, which lets one realize any sequence in $U$ by a connected sum of such pieces.","core_discovery":"The central result, stated as Theorems 1.3 and 1.4, is that the map $F$ defined on generic triple-point-free immersions by $F(f)_k = \\sum_{v \\in \\delta_f^{-1}(\\{k\\})} (1 - \\deg^-(v))$ is invariant under regular homotopies staying within $\\mathrm{Imm}_{<3}(S^2,\\mathbb{R}^3)$, and its image is exactly $U = \\{ (h_k) \\in \\mathbb{Z}^{\\mathbb{Z}} : h_{2k}=0\\ \\forall k,\\ \\sum |h_k| < \\infty,\\ \\sum h_k = 1 \\}$. The invariant is read off from a double point tree: cutting the sphere along the double point curves gives a tree whose vertices are regions, with an edge for each double point curve, a pairing matching the two sheets along each curve, and a local topological degree $\\delta_f$ on each vertex. Invariance is shown by classifying the two types of tangency events (E and H) that a generic deformation crosses; for H-events, the tree modification replaces one vertex by two whose indegrees satisfy $\\deg^-(v_f) = \\deg^-(v_g) + \\deg^-(w_g) - 1$, making the weighted count per degree unchanged. Surjectivity is achieved by constructing surfaces of revolution with $F=e_k$ and $F=2e_1-e_k$ for every odd $k$, and a pole-gluing connected sum that obeys $F(f\\#g)=F(f)+F(g)-e_1$.","pith_inferences":["The connected-sum relation $F(f\\#g)+e_1 = (F(f)+e_1)+(F(g)+e_1)$ suggests that, if the invariant is ever shown to be complete, the regular homotopy classes of triple-point-free immersions would form a free abelian monoid under connected sum with the standard embedding as unit.","Because $\\mathrm{Imm}_{<4}(S^2,\\mathbb{R}^3)$ is a subspace of $\\mathrm{Imm}_{<3}(S^2,\\mathbb{R}^3)$, any value of $F$ realized by an immersion without quadruple points obstructs quadruple-point-free deformations; the surfaces of revolution realizing all of $U$ might be adapted to answer Question 1.6, asking whether $\\mathrm{Imm}_{<4}$ has more than two components.","The explicit image of $F$ on the Willmore initial surface $j$ and the triple-point obstruction could be used to search for minimisers of the Willmore energy within each regular homotopy class, provided the gradient flow can stay away from triple points for a long time."],"forward_implications":["The space $\\mathrm{Imm}_{<3}(S^2,\\mathbb{R}^3)$ of immersed spheres without triple points has infinitely many regular homotopy classes, whereas only two were known before.","Any two immersed spheres with different values of $F$ can only be connected by a regular homotopy that passes through triple points, so the invariant detects a topological obstruction to avoiding triple points.","For the Willmore-flow initial surface $j$ of Figure 1, $F(j)=e_3$; hence every regular homotopy from $j$ to the round sphere must cross a triple point, giving a topological reason for singular behaviour in the Willmore flow and restricting the components of energy sublevel sets.","Since known finite-order local invariants cannot separate triple-point-free immersions from both standard embedded spheres, $F$ is necessarily a new kind of non-finite-order invariant.","The invariant restricts to $\\mathrm{Imm}_{<3}(S^2,\\mathbb{R}^3)/\\mathrm{Diff}(S^2)$ by identifying a sequence with its reflection, giving a version of the classification that is independent of sphere reparametrisation."],"supporting_citations":[{"why":"Provides the earlier quadruple-point invariant and the lower bound of two components for Imm<3, which Theorem 1.4 upgrades to infinitely many.","marker":"[MB81]"},{"why":"The classical result that all immersed spheres in R3 form a single regular homotopy class, the baseline that Table 1 contrasts with the new infinitude.","marker":"[Sma58]"},{"why":"Supplies the theorem that immersed real projective planes in R3 have an odd number of triple points, used in Lemma 4.2 to rule out one H-singularity connection pattern.","marker":"[Ban74b]"},{"why":"Classifies the codimension-one strata E, H, T, Q of Imm(S2,R3), which justifies perturbing any regular homotopy to a generic one that only crosses those events.","marker":"[HK01]"},{"why":"Provides the order-one invariants and the H1/H2 singularity notation adopted in the tree-modification classification, and the comparison showing F is not of finite order.","marker":"[Now04]"},{"why":"Classifies which trees with a pairing are realizable by immersed spheres, used in Remark 4.5 to note that tree moves can produce non-realizable data.","marker":"[Lip04]"}],"fun_headline_variants":["Sequence-valued invariant forces triple points in sphere eversions","Counting double point degrees reveals infinite homotopy classes","Invariant in l1(Z) separates immersed spheres without triple points","Triple-point-free sphere immersions split into infinite families"],"cache_read_input_tokens":13952,"weakest_assumption_plain":"The surjectivity part of the image description rests on constructions presented mainly through figures: for every odd integer $k$, surfaces of revolution $f_k$ and $g_k$ exist without triple points and with $F(f_k)=e_k$ and $F(g_k)=2e_1-e_k$, and gluing such surfaces along their poles changes the invariant by $F(f\\#g)=F(f)+F(g)-e_1$; if any of these constructions accidentally creates a triple point or changes the invariant differently, the image description and the infinitude of classes could fail.","fun_headline_variants_meta":{"raw":{"variants":["Sequence-valued invariant forces triple points in sphere eversions","Counting double point degrees reveals infinite homotopy classes","Invariant in l1(Z) separates immersed spheres without triple points","Triple-point-free sphere immersions split into infinite families"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000946,"raw_usage":{"total_tokens":4060,"prompt_tokens":990,"completion_tokens":3070,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":606,"completion_tokens_details":{"reasoning_tokens":3002}},"tokens_in":606,"tokens_out":3070,"duration_ms":24147,"temperature":1.0,"reasoning_tokens":3002,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:33:29.642601+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Carry out the construction of Lemma 6.3 for the sequence $h = e_7 + e_{-5} - e_1$: glue the surfaces $f_7$, $g_1$, and $f_{-5}$ along their poles and enumerate the double point tree of the result. If any gluing neck produces a triple point, if any vertex of the resulting tree has even local degree, or if the computed value of $F$ differs from $h$, then the surjectivity half of Theorem 1.4 fails.","supporting_citations":[],"review_version":1}