{"id":"210a6531-a36d-4ab0-bc16-82f351928529","arxiv_id":"1908.04043","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Null-homologous twists change the algebraic genus by at most one under a square condition, yielding new upper bounds on the topological slice genus of torus knots and satellite knots.","lead":"The paper develops null-homologous twisting as a tool for the algebraic genus of knots, an invariant that bounds the topological slice genus. It proves that certain pairs of twists change the algebraic genus by at most one and uses this to give new upper bounds for torus knots.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 13's twist-counting recurrence has an off-by-one that leaves the asymptotic 2/3 bound unsupported as stated.","rationale":"The reader's conditional verdict already flags Lemma 13's geometric move and a misstated inequality in Theorem 5, so my recommendation agrees that the paper should not be accepted without a fix. My focus differs from the reader's stated weakest assumption: I regard the off-by-one in the twist-counting recurrence as the most load-bearing concern for the central asymptotic claim, rather than the stabilization statement in Theorem 1, which is quoted from Feller-Lewark and is likely sound. The recurrence gap is directly visible in the proof of Lemma 13: T_k is defined for full twists on 2k strands, yet the displayed move applies to 2k+1 strands, so the induction step from T_k to T_{k+1} is not established. If the figure caption or text is misprinted and the intended move starts with 2k+2 strands, the argument can likely be repaired; if not, the constant in the torus-knot bound could degrade and the 2/3 asymptotic improvement would not follow. The test I propose settles which case holds by an explicit braid computation. I do not see an internal contradiction in the rest of the paper, and the satellite-knot and Theorem 2 arguments appear sound modulo the already-noted need for more detail in Proposition 10 and Lemma 8's cited background. The paper is parameter-free and the main theorems are concrete enough that a focused check of Lemma 13 is the right next step.","tokens_in":11680,"tokens_out":31884,"duration_ms":326487,"concrete_test":"Re-derive the braid manipulation in Lemma 13 and Figure 7 for 2k+2 strands. Write the full twist on 2k+2 strands as (sigma_1 ... sigma_{2k+1})^{2k+2}, apply the null-homologous -1 twist exactly as depicted, and check whether the result is four full twists on 2k strands. If the move only applies to 2k+1 strands, compute the resulting recurrence, e.g., R_{2k+2} <= 1 + R_{2k+1} <= 2 + 4R_{2k}, and re-evaluate whether the bound galg(T_{2^a,2^b}) < 2^{a+b}/3 still follows. A useful calibration is to compute or estimate galg(T_{4,4}) and compare it with the Lemma 13 bound 16/3.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central asymptotic claim (Theorem 5) rests on Lemma 13's bound galg(T_{2^a,2^b}) < 2^{a+b}/3, which is proved by counting null-homologous twists needed to undo a full twist. The proof defines T_k as the number of null-homologous twists required to undo a full twist on 2k strands, and then asserts the recurrence T_{k+1} <= 1 + 4T_k. However, the geometric move described in the proof and Figure 7 starts with a full twist on 2k+1 strands and produces four full twists on 2k strands. This gives a bound for a full twist on 2k+1 strands, not for a full twist on 2(k+1) = 2k+2 strands. Unless one can first reduce a full twist on 2k+2 strands to one on 2k+1 strands at the cost of at most one null-homologous twist (or unless the displayed number of strands is a typo for 2k+2), the induction T_k <= (4^k - 1)/3 is not justified. Since the constant 1 in front of 4T_k is exactly what produces the factor 1/3 and hence the 2/3 asymptotic ratio, this gap is load-bearing for the paper's headline improvement over the previous <3/4 bound. The gap is concrete and local: it is a parameter-free derivation issue, not a disagreement with existing consensus.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces null-homologous twisting operations as a tool for studying the algebraic genus $g_{alg}$. Theorem 1 states that performing an $m$-twist and an $n$-twist on null-homologous curves changes $g_{alg}$ by at most $1$ whenever $-mn$ is a square; the proof gives an explicit Seifert-matrix congruence. Theorem 2 characterizes $g_{alg}(L)$ as the minimal maximum number of $+1$- and $-1$-null-homologous twists needed to reach a link of algebraic genus zero. Theorem 4 gives the satellite bound $g_{alg}(P(K)) \\le g_{alg}(P(U)) + g_{alg}(K)$, attributed to Feller-Miller-Pinzon-Caicedo. Theorem 5 proves $g_{top}^4(T_{p,q}) \\le g_{alg}(T_{p,q}) < pq/3 + p\\log_2 q + q\\log_2 p$, improving the known asymptotic upper bound on $g_{top}^4/g_4$ for torus knots from below $3/4$ to at most $2/3$. Proposition 3 shows the square condition in Theorem 1 is sharp.","tokens_in":11965,"tokens_out":13927,"duration_ms":141534,"significance":"The main contribution is a new, explicit mechanism---null-homologous twisting---for bounding the algebraic genus. If correct, Theorem 5 improves the previously known asymptotic upper bound for torus knots, and the proof is largely parameter-free: Theorem 1 is a concrete matrix identity, Lemma 13 is a counting argument for full twists, and no fitted constants appear. The paper is also careful with attribution: the satellite upper bound is credited to FMPC19 rather than claimed as new. The main results have checkable, local proofs and should be of interest to knot theorists and 4-manifold topologists. I also checked the apparent off-by-one in Lemma 13 raised in the stress-test note; with the exponents read as $2^{k+1}$ and $2^k$, the recurrence is exactly what the geometry in Figure 7 supports, so I do not regard that concern as a real flaw.","major_comments":[],"minor_comments":[{"comment":"Inequality (3) has the two logarithms swapped: since $q$ is a sum of $k+1$ powers of two, one has $k\\le \\log_2 q$, and similarly $l\\le \\log_2 p$. With this correction, the final estimate $ql+pk\\le q\\log_2 p + p\\log_2 q$ gives exactly the stated bound; as printed, the displayed inequalities would not yield the theorem's conclusion.","section":"§5 (proof of Theorem 5)"},{"comment":"The argument is correct if the notations '2k+1 strands' and '2k strands' are typeset as $2^{k+1}$ and $2^k$ strands; then a full twist on $2^{k+1}$ strands is converted into four full twists on $2^k$ strands by one null-homologous twist, yielding $T_{k+1}\\le 1+4T_k$. Please ensure the superscripts are not lost in the final version, since a reader seeing $2k+1$ could otherwise infer an off-by-one recurrence.","section":"§5 (Lemma 13)"},{"comment":"In the displayed quadratic form (6), the coefficient of $x_4^2$ should be $d$, not $c$, to match the Seifert matrix (5) and the diagonalization (7).","section":"§6 (proof of Proposition 3)"},{"comment":"The assertion that one can choose a diagram so that Seifert's algorithm gives a surface disjoint from the surgery curves with the relevant homology classes forming an unlink is only justified by Figure 2; a short explanatory sentence describing the construction would make this load-bearing step easier to verify.","section":"§2 (proof of Theorem 1)"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript is within the scope of the journal. I have no concerns about attribution or novelty disclosure. The only issues are local typographical and expository points; the central claims appear sound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper develops null-homologous twisting operations as a tool for studying the algebraic genus, and the core ideas are genuinely useful. Theorem 1 is a clean, explicit matrix argument: if two null-homologous twists have framings m and n with -mn a square, the algebraic genus changes by at most one. The proof is concrete and the stabilization input from Feller-Lewark is correctly attributed and applied. Theorem 2, the characterization of the algebraic genus in terms of ±1 null-homologous twists, is a nice consequence and appears correct. The satellite bound in Theorem 4 is credited to Feller-Miller-Pinzon-Caicedo, and the proof here is simple and clean. Proposition 3, giving knots unknotted by pairs of twists where -mn is not a square, is a good converse and the number-theoretic argument is sound. The writing is clear and the author is honest about what is new and what is not.\n\nThe soft spot is in the proof of Lemma 13, and it is load-bearing. The lemma is what produces the asymptotic 2/3 bound for torus knots, and the recurrence as written does not follow from the pictured move. T_k is defined as the number of null-homologous twists needed to undo a full twist on 2k strands. The proof then describes a move that converts a full twist on 2k+1 strands into four full twists on 2k strands using one null-homologous twist, and concludes the recurrence T_{k+1} ≤ 1 + 4T_k. That recurrence is for a full twist on 2(k+1) = 2k+2 strands, not 2k+1. Unless there is an additional move that first reduces a full twist on 2k+2 strands to one on 2k+1 strands at the cost of at most one null-homologous twist, the induction is not justified. The constant 1 in the recurrence is exactly what produces the factor 1/3 and hence the 2/3 ratio, so this is not a cosmetic typo. The separate misstatement in inequality (3) of Theorem 5, where k and l are swapped relative to the bounds on p and q, is minor and easily fixed, but the Lemma 13 gap is real.\n\nThe rest of the paper, as far as I can tell, holds together. Theorem 1, Theorem 2, the satellite proof, and Proposition 3 all appear sound. The paper deserves a serious referee, because the technique is new and the main theorems are interesting even if the torus knot bound needs repair. I would send it to peer review and ask the author to fix Lemma 13 or state the precise strand count in the induction. If the off-by-one cannot be repaired, the paper still has solid contributions, but the headline asymptotic improvement over the previous <3/4 bound is not established as written.","headline":"The twisting technique and the main inequalities for Theorems 1 and 2 are solid and worth knowing, but the proof of the torus knot bound has a real off-by-one in Lemma 13 that leaves the headline 2/3 asymptotic unsupported as written.","tokens_in":12463,"tokens_out":2852,"would_cite":true,"duration_ms":31221,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["57K10"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that null-homologous twisting changes algebraic genus by at most one in square cases, yielding new upper bounds on the topological slice genus of torus and satellite knots.","keywords":["algebraic genus","null-homologous twisting","topological slice genus","torus knots","satellite knots","Seifert surfaces","Alexander polynomial"],"falsifier":"Compute the algebraic genus of any torus knot, for example $T_{2^a,2^b}$, by exhaustive search over Seifert surfaces; if any value reaches or exceeds $pq/3 + p\\log_2 q + q\\log_2 p$, then Theorem 5 is false. Likewise, find two links that differ by a null-homologous $+1$-twist and a null-homologous $-1$-twist but whose algebraic genera differ by $2$; such a pair would refute Theorem 1.","tokens_in":11494,"feed_emoji":"🔗","tokens_out":6235,"duration_ms":64452,"temperature":0.7,"pith_summary":"The paper introduces null-homologous twisting as a controlled operation on links and proves that certain pairs of such twists change the algebraic genus by at most one. The algebraic genus is the optimal upper bound on the topological slice genus obtainable from the criterion that Alexander polynomial one knots are topologically slice, so controlling it controls the topological four-ball genus. Using this, the paper shows that the algebraic genus of a torus knot $T_{p,q}$ is less than $pq/3 + p\\log_2 q + q\\log_2 p$, which implies the ratio of topological to smooth slice genus for torus knots is asymptotically at most $2/3$, improving the previously known bound below $3/4$. It also proves a winding-number-independent subadditivity bound for satellite knots.","feed_headline":"Torus knots' topological slice genus ratio drops to at most 2/3","feed_subtitle":"A new twisting argument bounds torus-knot slice genus and makes satellite genus independent of winding number.","key_machinery":"The load-bearing object is the null-homologous twist: an unknotted curve $C$ disjoint from the link with $\\operatorname{lk}(C,L)=0$, on which a $1/n$-surgery is performed. The proof of Theorem 1 compares the Seifert matrices $M$ and $M'$ of the two links; when $-mn$ is a square, writing $m=-ax^2$ and $n=ay^2$ and stabilizing $M'$ by two extra basis vectors yields a matrix $M''$ such that an invertible integral matrix $P$ satisfies $P^T M'' P = M$ as an upper-left block, with two additional rows and columns. This shows the algebraic genus changes by at most one. For the torus-knot bound, the key recursion is that a full twist on $2k+1$ strands can be converted into four full twists on $2k$ strands using one null-homologous twist, giving $T_{k+1}\\le 1+4T_k$, and the binary expansion of $p$ and $q$ decomposes $T_{p,q}$ into pieces $T_{2^a,2^b}$ whose algebraic genera are controlled by $2^{a+b}/3$.","core_discovery":"The central discovery is that null-homologous twisting pairs respect the algebraic genus up to a controlled error: if two links differ by a null-homologous $m$-twist and a null-homologous $n$-twist with $-mn$ a square, then their algebraic genera differ by at most one. From this, the paper derives a characterization of the algebraic genus as the minimum over ways of converting a link to an Alexander polynomial one link using $p$ null-homologous $+1$-twists and $n$ null-homologous $-1$-twists, namely $\\min\\max\\{n,p\\}$. Applying these operations to satellite knots gives $g_{\\mathrm{alg}}(P(K)) \\le g_{\\mathrm{alg}}(P(U)) + g_{\\mathrm{alg}}(K)$, independent of the pattern's winding number. For torus knots, a recursive untwisting argument yields $g_{\\mathrm{alg}}(T_{p,q}) < pq/3 + p\\log_2 q + q\\log_2 p$, hence $g^{\\mathrm{top}}_4(T_{p,q})$ satisfies the same bound, so the asymptotic ratio of topological to smooth slice genus for torus knots is at most $2/3$.","pith_inferences":["Editorial inference: Theorem 2 turns the algebraic genus into a search problem over ±1 null-homologous twists, so one could compute or bound it for small links by enumerating such twist sequences; the paper does not implement such a search.","Editorial inference: If the algebraic genus also serves as a lower bound for topological slice genus in some families, the torus-knot bound would become an exact asymptotic statement; the paper only establishes upper bounds.","Editorial inference: The winding-number independence of the satellite bound contrasts sharply with smooth slice genus behavior, suggesting a testable extension in which the topological slice genus of satellites might also satisfy a winding-number-independent subadditivity bound; the paper leaves this open."],"forward_implications":["For any integer $n$, a single null-homologous $n$-twist changes the algebraic genus by at most one, since $-n^2$ is a square.","The algebraic genus equals the minimum over all ways of converting the link to an Alexander-polynomial-one link using $p$ null-homologous $+1$-twists and $n$ null-homologous $-1$-twists of $\\max\\{n,p\\}$.","For any satellite knot $P(K)$, the algebraic genus satisfies $g_{\\mathrm{alg}}(P(K)) \\le g_{\\mathrm{alg}}(P(U)) + g_{\\mathrm{alg}}(K)$, with no dependence on the winding number of the pattern.","For any torus knot or link $T_{p,q}$ with $p,q>1$, both the algebraic genus and the topological slice genus are strictly less than $pq/3 + p\\log_2 q + q\\log_2 p$, giving an asymptotic ratio of topological to smooth slice genus at most $2/3$.","If $-mn$ is not a square, the one-Lipschitz property can fail: there exist knots with algebraic genus and topological slice genus equal to $2$ that are unknotted by a null-homologous $m$-twist and a null-homologous $n$-twist."],"supporting_citations":[{"why":"Defines the algebraic genus and supplies its equivalent characterizations and the stabilization lemma used throughout.","marker":"[FL18]"},{"why":"Establishes that Alexander polynomial one knots are topologically slice, the base case that makes the algebraic genus bound topological slice genus.","marker":"[Fre82]"},{"why":"First applies the Alexander-polynomial-one criterion to produce upper bounds on topological slice genus, the framework the algebraic genus formalizes.","marker":"[Rud84]"},{"why":"Provides the previous upper bound on the topological slice genus of torus knots and the known asymptotic ratio $<3/4$ that Theorem 5 improves.","marker":"[BFLL18]"},{"why":"Independently obtained the satellite-knot upper bound that appears as Theorem 4 in this paper.","marker":"[FMPC19]"},{"why":"Supplies the null-homologous unknotting argument adapted in Proposition 10 to reduce the algebraic genus by exactly one using a pair of ±1 twists.","marker":"[Liv19]"},{"why":"Gives the Seifert-matrix block decomposition for satellite knots used to prove Lemma 12.","marker":"[Lic97]"},{"why":"Defines the Taylor invariant used as a lower bound for topological slice genus in the proof of Proposition 3.","marker":"[Tay79]"}],"fun_headline_variants":["Torus knot genus bound slashed to 2/3 ratio","Twisting trick caps torus knot genus ratio at 2/3","New genus bound: torus knots drop to 2/3 ratio","Algebraic genus for torus knots now ≤ 2/3 ratio","Twist-based proof tightens torus knot genus to 2/3"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of Theorem 1 requires that a Seifert surface realizing the algebraic genus can be stabilized and chosen so that it is disjoint from the two surgery curves, with the homology classes linking those curves represented by an unlink; this stabilization fact is quoted from earlier work.","fun_headline_variants_meta":{"raw":{"variants":["Torus knot genus bound slashed to 2/3 ratio","Twisting trick caps torus knot genus ratio at 2/3","New genus bound: torus knots drop to 2/3 ratio","Algebraic genus for torus knots now ≤ 2/3 ratio","Twist-based proof tightens torus knot genus to 2/3"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000139,"raw_usage":{"total_tokens":1110,"prompt_tokens":851,"completion_tokens":259,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":467,"completion_tokens_details":{"reasoning_tokens":163}},"tokens_in":467,"tokens_out":259,"duration_ms":3403,"temperature":1.0,"reasoning_tokens":163,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:55:30.681548+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the algebraic genus of any torus knot, for example $T_{2^a,2^b}$, by exhaustive search over Seifert surfaces; if any value reaches or exceeds $pq/3 + p\\log_2 q + q\\log_2 p$, then Theorem 5 is false. Likewise, find two links that differ by a null-homologous $+1$-twist and a null-homologous $-1$-twist but whose algebraic genera differ by $2$; such a pair would refute Theorem 1.","supporting_citations":[],"review_version":1}