{"id":"c9acd758-60d3-490f-bcc1-fde5b503a4a2","arxiv_id":"2608.02785","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Every smooth two-component split sphere link in S^4 admits infinitely many topologically non-isotopic splitting 3-spheres.","lead":"This paper proves that every two-component sphere link in four-dimensional space has infinitely many different splitting 3-spheres, even when the spheres are knotted. It completes a recent program on the non-uniqueness of splitting spheres in S^4.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation (3.1), the key computation for the knotted case, is asserted without derivation; if the signs, coset identification, or linear-in-k accumulation of the Whitehead terms are wrong, Proposition 3.3 and hence Theorem 1.1 fail.","rationale":"The reader's weakest assumption picks out Eq. (3.1); I agree. The proof of Proposition 3.3 reduces non-homotopy to non-π1-conjugacy, and that conclusion rests entirely on (3.1). The cited [BG25, Prop 2.8] is not included in the paper, so neither the exact Whitehead expression nor its behavior under iteration is independently justified. The rest of the argument—the universal-cover decomposition, Lemma 3.6, and the application to knot complements—appears coherent and is not where I would first attack. The trivial case contains a separate underived statement about smooth isotopy of homotopic circles, but that is less central because the unlink case is already known by Tatsuoka and the novel claim concerns knotted components. A direct model computation would settle whether (3.1) is correct; if it is correct, the central claim very likely holds. Thus the reader's conditional verdict remains appropriate, and no change to the verdict is needed.","tokens_in":11454,"tokens_out":32486,"duration_ms":277928,"concrete_test":"Recompute Eq. (3.1) in the model case X1 = S^2×D^2 (so π2(X1)=Z generated by [S]) and X2 = S^1×D^3, with α the generator of π1(X2). Use the explicit barbell on S^2×D^2 ♮ S^2×D^2 from [BG25, §3] to evaluate [F(Σ)] and [F^2(Σ)] in π3(X), via the universal-cover decomposition of §3. Verify that the coefficients of Wh([S],[S]^α) and Wh([S],[S]^{α^{-1}}) are ±1 for k=1 and ±2 for k=2, and that F_* fixes the corresponding summands. If the k=2 coefficients are not twice the k=1 coefficients, or if the two Whitehead terms land in the same summand under the π1-action, then Eq. (3.1) fails and Proposition 3.3 needs a different proof.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 3.3, the engine of the knotted case, depends on the single formula (3.1): [f^k(Σ)] = [Σ] ± k Wh([S],[S]^[α]) ± k Wh([S],[S]^([α]^{-1})). The formula is imported from [BG25, Prop 2.8] without derivation. Two load-bearing facts are asserted rather than proved. First, the two Whitehead products are placed in the distinct summands indexed by {1,[α]} and {1,[α]^{-1}}; the distinction uses [α]^2≠1, but the signs and coefficients in front of each term depend on orientations of S and S′ and on the choice of the α-twist, and these are not computed. Second, and more seriously, the coefficient k requires that the barbell diffeomorphism F acts trivially on the two Whitehead summands, so that each additional application of F adds the same two terms. [BG25, Prop 2.8] as cited concerns a single barbell; the iteration to k applications is not derived. If F_* permutes, conjugates, or rescales the Whitehead terms, the k-dependence need not be linear, or the terms could cancel at some nonzero k, and the conclusion that [f^k(Σ)] is not π1-conjugate to [Σ] breaks. Since Proposition 3.3 is the only route to Theorem 1.1 when a component is knotted, this is the most load-bearing unproved step in the paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that every smooth two-component split sphere link in S^4 admits infinitely many smoothly embedded splitting 3-spheres that are pairwise topologically non-isotopic. The proof has two parts. In the topologically trivial case, the authors extend Tatsuoka's theorem using Budney–Gabai's infinite-order diffeomorphisms of S^1×D^3 and a covering argument to show that certain pushed splitting spheres are inequivalent. In the knotted case, they establish a general criterion (Proposition 3.3) for a connected sum of 4-manifolds to have infinitely many non-homotopic splitting spheres, using barbell diffeomorphisms, Whitehead products, and an algebraic decomposition of the third homotopy group of the universal cover. They then apply this criterion to the exteriors of the two sphere-link components, using Freedman's unknotting theorem and Swarup's criterion to verify the hypotheses. The paper also includes proofs of auxiliary algebraic lemmas.","tokens_in":11799,"tokens_out":29528,"duration_ms":238423,"significance":"If the proof is completed, the result is a clean and broad generalization of Tatsuoka's theorem, covering all two-component split sphere links rather than only the unlink, and it gives a unified sufficient condition that recovers previously known nonuniqueness examples for surface links. The paper is written in a structured way, with the main geometric ideas clearly separated from the algebraic lemmas, and it makes explicit use of a number of deep external results (Budney–Gabai, Freedman–Quinn, Swarup, Dunwoody) that are appropriately cited. The algebraic decomposition in Lemma 3.2 and the torsion-freeness result in Proposition 3.1 are proved in the text and are of independent use. The main missing pieces are a small number of technical justifications in the two key propositions.","major_comments":[{"comment":"The key formula for the knotted case, [f^k(Σ)] = [Σ] ± k Wh([S],[S]^[α]) ± k Wh([S],[S]^([α]^{-1})), is asserted without derivation. The coefficient k is not justified: the cited [BG25, Prop 2.8] concerns a single barbell, and the paper does not prove that iterating the barbell diffeomorphism adds the same two Whitehead terms linearly. If the induced map on the relevant Whitehead summands permutes, conjugates, or rescales the terms, the conclusion that [f^k(Σ)] is not π1-conjugate to [Σ] could fail. Since Proposition 3.3 is the only route to Theorem 1.1 when a component is knotted, this step must be fully proved, or a reference covering the iterated barbell case must be provided.","section":"§3, equation (3.1)"},{"comment":"The last ∼ relation in (2.1) claims that γ and γ1 are homotopic and hence smoothly isotopic in X1#X̃2, and that the isotopy preserves the canonical framing. Homotopy of embedded circles does not imply smooth isotopy in general 4-manifolds, and this particular manifold (homotopy equivalent to S^1∨S^3) is not covered by a theorem the authors cite. Since the equality of the implementations (fγ)^k and (fγ1)^k in (2.1) is used to derive the contradiction in Proposition 2.5, this assertion needs a proof or a precise reference.","section":"§2, proof of Proposition 2.5, near (2.1)"},{"comment":"The homotopy equivalence X = X1#X2 ≃ X1∨X2∨S^3 is stated without proof. This equivalence underlies the universal-cover decomposition of π3(X) used to separate the components in (3.1) and to conclude that the Whitehead terms cannot cancel. A short justification or reference would remove the gap.","section":"§3, proof of Proposition 3.3"}],"minor_comments":[{"comment":"In the sentence 'S^4\\ν(L) and S^4\\ν(L) are both homeomorphic to S^1×D^3', the second occurrence should be S^4\\ν(R).","section":"§2, proof of Theorem 1.1 (trivial case)"},{"comment":"The superscripts for the π1-action are missing in places: the terms should read [S]^[α] and [S]^([α]^{-1}).","section":"§3, equation (3.1) and surrounding text"},{"comment":"The two Whitehead terms are both written with the same exponent [α]; the second should be [S]^([α]^{-1}), not [S]^[α].","section":"§3, paragraph after the universal-cover decomposition"},{"comment":"The equality H^2_c(X;Z) = lim H^2(X,X\\K;Z) should be written as an isomorphism with the correct direction of the direct limit; the intended statement is standard.","section":"§3, proof of Proposition 3.1"},{"comment":"The phrase 'the descriptions of the isomorphisms follow from straightforward diagram chasing' could be expanded by one sentence, since the identification of the tensor-product summand with the Whitehead product is used later in a load-bearing way.","section":"§3, proof of Lemma 3.2"},{"comment":"The role of the nonempty boundary in the homotopy equivalence X ≃ X1∨X2∨S^3 should be explained or a reference given, since the same formula would fail for closed manifolds.","section":"§3, proof of Proposition 3.3"}],"recommendation":"major_revision","confidential_remarks":"The main risk is the single underexplained formula (3.1) in the knotted case; the authors should be encouraged to provide full details even if lengthy. The homotopy-implies-isotopy claim in §2 is also likely to attract expert scrutiny; if it is true in this special setting, a precise reference is needed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nThis paper answers a natural question completely: every smooth two-component split sphere link in S^4, even with arbitrarily knotted components, has infinitely many splitting 3-spheres that are pairwise topologically non-isotopic. That's a genuine advance over the earlier examples of Hughes–Kim–Miller and Tatsuoka, and the main theorem is new.\n\nThe proof has two parts. The topologically trivial case is a careful extension of Tatsuoka's construction using Budney–Gabai diffeomorphisms of S^1×D^3; that portion is solid. The paper also proves a general criterion (Prop 1.3) for connected sums of 4-manifolds, which is a nice contribution in its own right and covers the previous positive-genus surface link examples.\n\nThe topologically nontrivial case is the soft spot. Everything rests on equation (3.1): applying the barbell diffeomorphism k times changes the homotopy class of the splitting sphere by ±k times two Whitehead products, each living in a distinct summand. The authors cite [BG25, Prop 2.8] for this, but that result covers a single barbell. The iteration to k applications, the linear coefficients, and the claim that the barbell action is trivial on the relevant summands are not proved. If any of these details fail, the conclusion of Prop 3.3 fails, and with it the knotted-component case. That's a real gap.\n\nTwo smaller points: the proof of Prop 2.5 asserts that homotopic circles are smoothly isotopic in X1#X~2 without justification, and the homotopy equivalence X1#X2 ≃ X1∨X2∨S^3 is stated as a fact. Both are probably true in this setting, but they need a reference or a sentence.\n\nNone of this makes the paper a write-off. The overall strategy is coherent, the unknotted case is complete, and the algebraic machinery is interesting. The gap in (3.1) is a missing computation, not a contradiction. I'd send this to a serious referee: the result is important enough, and the paper is serious enough, that it deserves referee time. With (3.1) properly proved, it would be a strong paper. As is, it's a strong paper with a genuine missing argument.\n\nRecommendation: accept for peer review; require a full proof of (3.1) and small clarifications elsewhere.","headline":"Answers a natural question completely; the knotted-component case rests on an unproved linearity formula (3.1) that needs a full derivation before the proof is airtight.","tokens_in":12304,"tokens_out":9861,"would_cite":true,"duration_ms":80003,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["57K45","57R52","57N13"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every smooth two-component split sphere link in $S^4$ admits infinitely many smooth splitting $3$-spheres that are pairwise topologically non-isotopic.","keywords":["splitting spheres","sphere links","4-manifolds","barbell diffeomorphism","topological isotopy","Whitehead products","connected sum","unknotting"],"falsifier":"Verify equation (3.1) in the model case $X_1 = S^2 \\times D^2$, $X_2 = S^1 \\times D^3$, where $\\pi_3$ can be computed explicitly from the wedge decomposition; if direct computation shows the coefficient of $\\operatorname{Wh}([S],[S]^{[\\alpha]^{-1}})$ is not $\\pm k$, or the two terms lie in the same summand when $[\\alpha]^2=1$, the knotted-case argument fails.","tokens_in":11246,"feed_emoji":"🔗","tokens_out":12027,"duration_ms":94271,"temperature":0.7,"pith_summary":"This paper proves that every smooth two-component split sphere link $L \\sqcup R \\subset S^4$ admits infinitely many smoothly embedded $3$-spheres that separate $L$ from $R$ and are pairwise non-isotopic even by topological ambient isotopies fixing the link. The theorem was previously known only for the unlink; the new proof handles components of arbitrary knot type. The key intermediate result, Proposition 1.3, is a general criterion: a connected sum of two $4$-manifolds has infinitely many topologically non-isotopic splitting $3$-spheres whenever one summand contains a non-null-homotopic $2$-sphere with trivial normal bundle and the other has a loop whose square is nontrivial. The criterion also accounts for the earlier examples coming from positive-genus surface links.","feed_headline":"Every split sphere link in S^4 has infinitely many splitting spheres","feed_subtitle":"Even knotted components admit endlessly many topologically distinct ways to separate them.","key_machinery":"The barbell diffeomorphism is the central object: a diffeomorphism supported in a neighborhood of an embedded arc joining two parallel copies of a $2$-sphere, which changes the homotopy class of the splitting sphere by Whitehead products $\\operatorname{Wh}([S],[S]^{[\\alpha]})$ and $\\operatorname{Wh}([S],[S]^{[\\alpha]^{-1}})$ with coefficient $\\pm k$. In the unknotted case the machinery is instead a diffeomorphism $f$ of $S^1 \\times D^3$ that is the identity near the boundary, has infinite order in the mapping class group modulo discs, and lifts to a finite cover as a diffeomorphism smoothly isotopic to the identity; implementing $f$ along a circle representing $1 * 1 \\in \\pi_1(X_1 \\# X_2)$ gives the infinite family of spheres.","core_discovery":"The central claim, Theorem 1.1, states that for any smooth two-component split sphere link $L \\sqcup R \\subset S^4$ there exists an infinite family $\\{\\Sigma_i\\}_{i \\in \\mathbb{N}}$ of smoothly embedded $3$-spheres in $S^4 \\setminus (L \\sqcup R)$, each separating $L$ from $R$, such that for $i \\neq j$ no topological ambient isotopy of $S^4$ fixing a neighborhood of $L \\sqcup R$ carries $\\Sigma_i$ to $\\Sigma_j$. Two cases are treated. When both components are topologically unknotted, the result follows from Proposition 1.2, which generalizes the previously known unlink result to connected sums of two copies of $S^1 \\times D^3$ using a diffeomorphism of infinite order in the mapping class group. When at least one component is knotted, Proposition 1.3 supplies infinitely many splitting spheres that are non-homotopic in the complement; the proof uses a barbell diffeomorphism built from two parallel copies of a non-null-homotopic $2$-sphere and an arc representing a loop with nontrivial square, and detects the non-homotopy via Whitehead products in $\\pi_3$.","pith_inferences":["The homotopy class of a splitting sphere in the complement may be a complete invariant for the topological isotopy problem among separating $3$-spheres produced by barbell constructions; if so, the main theorem gives a classification of these spheres up to isotopy.","The condition $[\\alpha]^2 \\neq 1$ suggests that when the fundamental group of one complement has only elements of order dividing two, the two Whitehead terms in equation (3.1) could coincide or cancel, so proving non-uniqueness for such links would require a different mechanism.","The same criterion might extend to split surface links with knotted positive-genus components, as long as the complement of one component contains a suitable embedded $2$-sphere; this would unify the existing examples beyond the specific cases already studied.","Because the proof only uses the homotopy action of $\\pi_1$ on $\\pi_2$, the construction may transfer to codimension-one separating submanifolds in other $4$-manifolds with nonempty boundary, such as those arising in exotic embedding problems."],"forward_implications":["Every smooth two-component split sphere link in $S^4$ has infinitely many smoothly embedded separating $3$-spheres, and since topological non-isotopy implies smooth non-isotopy, the family is also smoothly non-isotopic.","The general criterion (Proposition 1.3) yields infinite non-uniqueness for any connected sum of two compact $4$-manifolds with nonempty boundary where one side contains a non-null-homotopic $2$-sphere with trivial normal bundle and the other has a loop whose square is not trivial.","In the knotted case the constructed spheres are actually non-homotopic in the complement, so the obstruction is visible in $\\pi_3$ of the link complement rather than being a purely smooth phenomenon.","The proof covers all smooth sphere links, with no restriction on the knot type of the components, because the only input about a knotted component is that a parallel copy is non-contractible in its complement."],"supporting_citations":[{"why":"Supplies the diffeomorphism $f$ of $S^1\\times D^3$ with the three properties in Theorem 2.2, and Proposition 2.8 gives the Whitehead-product formula used in equation (3.1).","marker":"[BG25]"},{"why":"Provides the previously known theorem for the unlink and the strategy, generalized in Proposition 1.2, of implementing an infinite-order diffeomorphism along an embedded circle.","marker":"[Tat26]"},{"why":"Gives the unknotting criterion used to show that a parallel copy of a knotted sphere is not contractible in the complement.","marker":"[Swa75]"},{"why":"Proves accessibility of finitely presented groups, which is needed to apply Swarup's criterion in the knotted case.","marker":"[Dun85]"},{"why":"Provides the unknotting theorem converting a $\\pi_1=\\mathbb{Z}$ complement into topological unknottedness, contradicting the assumption of a knotted component.","marker":"[FQ14]"},{"why":"Establishes the topological annulus theorem in dimension 4, used to identify complements of balls with standard pieces in the unknotted case.","marker":"[Qui82]"},{"why":"Supplies uniqueness of topological tubular neighborhoods in dimension 4, used to identify the complement of an unknotted sphere with $S^1\\times D^3$.","marker":"[FNOP25]"},{"why":"Together with the next reference, shows homeomorphisms of smooth 3-manifolds are topologically isotopic to diffeomorphisms, justifying boundary choices near the collar.","marker":"[Mun60]"},{"why":"Provides local contractibility of homeomorphism groups, used in combination with the previous reference to choose diffeomorphisms near the boundary.","marker":"[Čer69]"}],"fun_headline_variants":["Every split S^2-link admits infinitely many non-isotopic splitting spheres","Even knotted components: infinitely many topologically distinct splittings","Infinite non-isotopic splitting spheres for all split sphere links","Generalizing Tatsuoka: infinite topologically distinct splitting spheres"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The knotted case relies on the quoted formula (3.1) that the barbell diffeomorphism changes the splitting sphere's homotopy class by exactly $\\pm k \\operatorname{Wh}([S],[S]^{[\\alpha]}) \\pm k \\operatorname{Wh}([S],[S]^{[\\alpha]^{-1}})$; if that formula's signs, coefficient, or identification of the two Whitehead summands is wrong, the proof that $[f^k(\\Sigma)]$ is not $\\pi_1$-conjugate to $[\\Sigma]$ collapses.","fun_headline_variants_meta":{"raw":{"variants":["Every split S^2-link admits infinitely many non-isotopic splitting spheres","Even knotted components: infinitely many topologically distinct splittings","Infinite non-isotopic splitting spheres for all split sphere links","Generalizing Tatsuoka: infinite topologically distinct splitting spheres"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00096,"raw_usage":{"total_tokens":4078,"prompt_tokens":923,"completion_tokens":3155,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":539,"completion_tokens_details":{"reasoning_tokens":3080}},"tokens_in":539,"tokens_out":3155,"duration_ms":20639,"temperature":1.0,"reasoning_tokens":3080,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:01:38.712467+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Verify equation (3.1) in the model case $X_1 = S^2 \\times D^2$, $X_2 = S^1 \\times D^3$, where $\\pi_3$ can be computed explicitly from the wedge decomposition; if direct computation shows the coefficient of $\\operatorname{Wh}([S],[S]^{[\\alpha]^{-1}})$ is not $\\pm k$, or the two terms lie in the same summand when $[\\alpha]^2=1$, the knotted-case argument fails.","supporting_citations":[],"review_version":1}