Smoothly slice knots with smoothly non-approximable topological slice discs

Min Hoon Kim, Mark Powell · 2025

Read Smoothly slice knots with smoothly non-approximable topological slice discs by Min Hoon Kim, Mark Powell (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-03705.txt.

Introduction

Min Hoon Kim and Mark Powell construct smoothly slice knots whose topological slice discs cannot be approximated by smooth ones, published in 2025. A knot is slice when it bounds a disc in four-dimensional space, and the smooth and topological versions of that question can diverge. The paper gives infinitely many examples: each knot is smoothly slice, yet it admits topological slice discs that no sequence of smooth discs can approximate. The result adds a concrete infinite family to the known separation between the smooth and topological categories.

Catalog record

CREATORMin Hoon Kim, Mark Powell
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE10 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-03705

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Smoothly slice knots with smoothly non-approximable topological slice discs Min Hoon Kim, Mark Powell Abstract . We construct infinitely many smoothly slice knots having topological slice discs that are non-approximable by smooth slice discs. 1991 Mathematics Subject Classification 57K10, 57K18, 57N70, 57R40. Let K be a smoothly slice knot in S^{3} and let D be a topological slice disc for K in D^{4} , that is, D is a locally flat, topologically embedded disc in D^{4} such that \partial(D^{4},D)=(S^{3},K) . We say that D is smoothly non-approximable if there exists \varepsilon>0 such that the \varepsilon -neighbourhood \nu_{\varepsilon}(D) of D does not contain any smooth slice disc for K . Theorem . There exist infinitely many smoothly slice knots K_{m} in S^{3} , each of which admits a smoothly non-approximable topological slice disc. This answers mathoverflow question https://mathoverflow.net/questions/419693 , asked by M. Winter. It is interesting to contrast our theorem with Venema’s results in [ Ven78 , Ven81 ] , which imply that every topological slice disc can be approximated by smoothly embedded disc if the boundary is allowed to move. In particular Venema’s smooth approximation could have a different knot type on the boundary, or could even lie in the interior of D^{4} . Figure 1. The knots R_{m} and K_{m} . There are 2m+1 crossings between the bands. On the left, the knot R_{m} has unknots \alpha_{1} and \alpha_{2} in its complement, each linking one of the bands of the obvious Seifert surface for R_{m} . Perform a satellite operation on \alpha_{2} using \Wh to obtain K_{m}:=R_{m}(U,{\Wh}) , as shown on the right. Proof. For an odd integer m\geq 1 , let {\Wh}:=\operatorname{Wh}(T_{2,3}) denote the positive-clasped zero-twisted Whitehead double of the right-handed trefoil, let U denote the unknot, and let K_{m}=R_{m}(U,{\Wh}) , as defined in the caption of Figure 1 . Since \alpha_{1} and \alpha_{2} link R_{m} trivially, the Alexander polynomial is unchanged by the satellites [ Sei50 ] , i.e. \Delta_{K_{m}}(t)=\Delta_{R_{m}}(t) . Also \Delta_{R_{m}}(t)=((m+1)t-m)(mt-(m+1)) . Thus if m\neq n , then \Delta_{K_{m}}(t)\neq\Delta_{K_{n}}(t) , and hence the knots K_{m} are mutually distinct. The configuration (R_{m},\alpha_{1},\alpha_{2}) is often considered as an operator R_{m}(-,-) that takes two knots as input, with R_{m}(J_{1},J_{2}) obtained by performing the satellite operation on \alpha_{i} with J_{i} , for i=1,2 . Since taking J_{1}=U has no effect, this explains our definition in the caption to Figure 1 . For each m , the knot K_{m} is smoothly slice, as can be seen by ‘cutting’ the right hand band in the Seifert surface. From the embedded Morse theory perspective, this corresponds to a saddle move, which results in a 2-component unlink, which can then be capped off by two smooth discs, corresponding to minima, to obtain a smooth slice disc for K_{m} . Let D_{m} be a topological slice disc for K_{m} obtained by cutting the left band and capping with two parallel copies of a topological slice disc for {\Wh} , which exists since \Delta_{\Wh}(t)=1 [ Fre84 ] . We claim that D_{m} is smoothly non-approximable for each m . For a contradiction, suppose that for every \varepsilon>0 , the \varepsilon -neighbourhood \nu_{\varepsilon}(D_{m}) contains a smooth slice disc D_{m,\varepsilon}^{\prime} for K . Let N(D_{m}) be an open tubular neighbourhood of the topologically locally flat slice disc D_{m} , which exists by [ FQ90 , Theorem 9.3] . Then D^{4}\smallsetminus N(D_{m}) is the exterior of D_{m} , and \partial(D^{4}\smallsetminus N(D_{m}))\cong M_{K_{m}} , the 3-manifold obtained by zero-framed surgery on K_{m} . Choose \varepsilon>0 small enough such that \nu_{\varepsilon}(D_{m})\subseteq N(D_{m}) . We have inclusion maps M_{K_{m}}\xrightarrow{i_{1}}D^{4}\smallsetminus N(D_{m})\xrightarrow{i_{2}}D^{4}\smallsetminus\nu_{\varepsilon}(D_{m})\xrightarrow{i_{3}}D^{4}\smallsetminus D_{m,\varepsilon}^{\prime}. It follows from Alexander duality that these inclusion maps induce isomorphisms H_{1}(M_{K_{m}};\mathbb{Z})\xrightarrow{\cong}H_{1}(D^{4}\smallsetminus N(D_{m});\mathbb{Z})\xrightarrow{\cong}H_{1}(D^{4}\smallsetminus D_{m,\varepsilon}^{\prime};\mathbb{Z}). Since H_{1}(M_{K_{m}};\mathbb{Z})\cong\mathbb{Z} we obtain consistent coefficient systems \pi_{1}(X)\to\mathbb{Z}\cong\langle t\rangle for all X\in\{M_{K_{m}},D^{4}\smallsetminus N(D_{m}),D^{4}\smallsetminus D_{m,\varepsilon}^{\prime}\} , and can therefore consider homology with \mathbb{Q}[t^{\pm 1}] coefficients. Consider the corresponding composition \iota\colon H_{1}(M_{K_{m}};\mathbb{Q}[t^{\pm 1}])\xrightarrow{(i_{1})_{*}}H_{1}(D^{4}\smallsetminus N(D_{m});\mathbb{Q}[t^{\pm 1}])\xrightarrow{(i_{3}\circ i_{2})_{*}}H_{1}(D^{4}\smallsetminus D_{m,\varepsilon}^{\prime};\mathbb{Q}[t^{\pm 1}]). We need to understand the kernel of this map \iota:=(i_{3}\circ i_{2}\circ i_{1})_{*} . By our description of D_{m} , (i_{1})_{*}(\alpha_{1})=0 and hence \langle\alpha_{1}\rangle\subseteq\ker\iota , where we abuse notation and use \alpha_{1} to also denote the corresponding curve in the complement of K_{m} . Since D_{m,\varepsilon}^{\prime} is a slice disc, the kernel of \iota is a metaboliser for the Blanchfield form of M_{K_{m}} (see e.g. [ COT03 , Theorem 4.4] or [ Hil12 , Theorem 2.4] ). The Blanchfield form is a Hermitian, sesquilinear, nonsingular pairing \Bl\colon H_{1}(M_{K_{m}};\mathbb{Q}[t^{\pm 1}])\times H_{1}(M_{K_{m}};\mathbb{Q}[t^{\pm 1}])\to\mathbb{Q}(t)/\mathbb{Q}[t^{\pm 1}], and we say that P\subseteq H_{1}(M_{K_{m}};\mathbb{Q}[t^{\pm 1}]) is a metaboliser if P=P^{\bot} . We can compute, using the presentation matrix tV_{m}-V_{m}^{T} , where V_{m}=\left[\begin{smallmatrix}0&m+1\\ m&0\end{smallmatrix}\right] is a Seifert matrix for K_{m} , that H_{1}(M_{K_{m}};\mathbb{Q}[t^{\pm 1}])\cong\mbox{\footnotesize$\displaystyle\frac{\mathbb{Q}[t^{\pm 1}]}{(mt-(m+1))}$}\oplus\mbox{\footnotesize$\displaystyle\frac{\mathbb{Q}[t^{\pm 1}]}{((m+1)t-m)}$}\cong\mathbb{Q}\oplus\mathbb{Q}. The first summand is generated by \alpha_{1} , and the other is generated by \alpha_{2} . We claim that \langle\alpha_{1}\rangle=\ker(\iota) . To see this, we consider H_{1}(M_{K_{m}};\mathbb{Q}[t^{\pm 1}]) as a \mathbb{Q} -vector space. Then \langle\alpha_{1}\rangle\subseteq\ker(\iota)\subseteq H_{1}(M_{K_{m}};\mathbb{Q}[t^{\pm 1}]) implies that 1\leq\dim_{\mathbb{Q}}\ker(\iota)\leq 2 . However \ker(\iota) is a metaboliser for the Blanchfield form, and the Blanchfield form is nonsingular, so we cannot have \ker\iota=H_{1}(M_{K_{m}};\mathbb{Q}[t^{\pm 1}]) . Hence \dim_{\mathbb{Q}}\ker(\iota)=1 . Since \langle\alpha_{1}\rangle\subseteq\ker(\iota) and their dimensions as \mathbb{Q} -vector spaces are equal, we deduce that \langle\alpha_{1}\rangle=\ker(\iota) as claimed. Now we apply the d -invariant from Heegaard-Floer homology, together with calculations of Cha-Kim [ CK21 ] and Cha [ Cha21 ] , to deduce that the smooth disc D_{m,\varepsilon}^{\prime} cannot exist. For the definition of the d -invariant we refer to Ozsváth–Szabó [ OS03 , Definition 4.1] . It suffices for our purposes to know that given a rational homology 3-sphere Y and a spin c structure \mathfrak{s} on Y , the d -invariant is a rational number d(Y,\mathfrak{s})\in\mathbb{Q} . Let \Sigma_{r} be the r -fold cyclic cover of S^{3} branched along K_{m} , which is a rational homology 3-sphere for all prime r\geq 2 . Then \alpha_{1} and \alpha_{2} lift to homology classes in H_{1}(\Sigma_{r};\mathbb{Z}) , say x_{1} and x_{2} respectively. Since H_{1}(\Sigma_{r};\mathbb{Z})\cong H_{1}(M_{K_{m}};\mathbb{Z}[t^{\pm 1}])/(t^{r}-1) , we have that H_{1}(\Sigma_{r};\mathbb{Z})\cong\mathbb{Z}/_{(m+1)^{r}-m^{r}}\oplus\mathbb{Z}/_{(m+1)^{r}-m^{r}}, where the summands are generated by x_{1} and x_{2} respectively. In particular, \Sigma_{r} has a unique spin structure \mathfrak{s}_{\Sigma_{r}} since it is a \mathbb{Z}/2 -homology 3-sphere. By [ CK21 , Lemma 5.2] and [ Cha21 , p. 17, Assertion] , the condition \langle\alpha_{1}\rangle=\ker(\iota) implies that there exists a prime r such that the kernel of the inclusion induced map H_{1}(\Sigma_{r};\mathbb{Z})\to H_{1}(V_{r};\mathbb{Z}) is generated by x_{1} , where V_{r} is the r -fold cyclic cover of D^{4} branched along D^{\prime}_{m,\varepsilon} . Since D^{\prime}_{m,\varepsilon} is a smooth slice disc, [ GRS08 , Theorem 1.1] implies that d(\Sigma_{r},\mathfrak{s}_{\Sigma_{r}}+k\widehat{x}_{1})=0 for all k\in\mathbb{Z} . Here the spin structure \mathfrak{s}_{\Sigma_{r}} uniquely determines a spin c structure, and \widehat{x}_{1}:=PD^{-1}(x_{1})\in H^{2}(\Sigma_{r};\mathbb{Z}) is the Poincaré dual of x_{1} . Then the spin c structure \mathfrak{s}_{\Sigma_{r}}+k\widehat{x}_{1} is defined by noting that spin c structures on \Sigma_{r} are a torsor over H^{2}(\Sigma_{r};\mathbb{Z}) . We obtain the desired contradiction since there exists an integer k\in\mathbb{Z} such that d(\Sigma_{r},\mathfrak{s}_{\Sigma_{r}}+k\widehat{x}_{1})\neq 0, by [ CK21 , Theorem 5.4] and [ Cha21 , Lemma 4.1] . This completes the proof. ∎ Remark . Our examples are topologically doubly slice and smoothly slice, but not smoothly doubly slice. Infinitely many such knots were first constructed by Meier [ Mei15 ] , though the earlier d -invariant arguments of Cochran, Harvey, and Horn [

Read the complete text