This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

On tunnel numbers of a cable knot and its companion

Junhua Wang School of Mathematics and Physics, Jiangsu University of Technology, Changzhou 213001, P. R. China [email protected]; [email protected]  and  Yanqing Zou School of Mathematical Sciences & Shanghai Key Laboratory of PMMP, East China Normal University [email protected]; [email protected]
Abstract.

Let KK be a nontrivial knot in S3S^{3} and t(K)t(K) its tunnel number. For any (p2,q)(p\geq 2,q)-slope in the torus boundary of a closed regular neighborhood of KK in S3S^{3}, denoted by KK^{\star}, it is a nontrivial cable knot in S3S^{3}. Though t(K)t(K)+1t(K^{\star})\leq t(K)+1, Example 1.1 in Section 1 shows that in some case, t(K)t(K)t(K^{\star})\leq t(K). So it is interesting to know when t(K)=t(K)+1t(K^{\star})=t(K)+1.

After using some combinatorial techniques, we prove that (1) for any nontrivial cable knot KK^{\star} and its companion KK, t(K)t(K)t(K^{\star})\geq t(K); (2) if either KK admits a high distance Heegaard splitting or p/qp/q is far away from a fixed subset in the Farey graph, then t(K)=t(K)+1t(K^{\star})=t(K)+1. Using the second conclusion, we construct a satellite knot and its companion so that the difference between their tunnel numbers is arbitrary large.

Key words and phrases:
cable knot, tunnel number, Heegaard distance
2010 Mathematics Subject Classification:
57M27
This work was partially supported by NSFC No.S 11601065, 11571110, 11726609, 11726610 and Science and Technology Commission of Shanghai Municipality (STCSM), grant No. 18dz2271000. We thank Qilong Guo for his suggestion that we should give out a way to construct a satellite knot and its companion with arbitrary large difference between their tunnel numbers.

1. Introduction

Let KK be a nontrivial knot in S3S^{3} and E(K)E(K) its closed complement in S3S^{3}. Then E(K)E(K) admits a Heegaard splitting VSWV\cup_{S}W with E(K)=W\partial E(K)=\partial_{-}W. Let g(K)g(K) be the minimal Heegaard genus of E(K)E(K) and the tunnel number t(K)=g(K)1t(K)=g(K)-1. For any pair of pairwise coprime numbers pp and qq, there is a slope crossing the meridian pp times and the longitude qq times, denoted by p/qp/q, in E(K)\partial E(K). Then it is a (p,q)(p,q)-cable knot over KK, denoted by KK^{\star}, and KK is a companion of KK^{\star}. Though KK is also a cable knot of itself, we only consider the nontrivial case and require p2p\geq 2. Since KK^{\star} is contained in the closed neighborhood of KK, it is interesting to know the difference between t(K)t(K) and t(K)t(K^{\star}).

There is a combinatorial description of E(K)E(K^{\star}) through E(K)E(K), in which way it gives an inequality between t(K)t(K^{\star}) and t(K)t(K). Let η(K)\eta(K) be the closed regular neighborhood of KK in S3S^{3}. Since KK^{\star} is a p/qp/q slope in T2=η(K)=E(K)T^{2}=\partial\eta(K)=\partial E(K), we can slightly push KK^{\star} into the interior of η(K)\eta(K). It is not hard to see that E(K)E(K^{\star}) is homeomorphic to the amalgamation of E(K)E(K) and η(K)\eta(K) along an annulus AA in their common torus boundary, where the core curve of AA is isotopic to the p/qp/q slope. Then E(K)E(K^{\star}) admits a Heegaard splitting as follows: Let aa be a fiber arc in A×IA\times I from A×{0}A\times\{0\} to A×{1}A\times\{1\} and η(a)\eta(a) the closed regular neighborhood in A×IA\times I. Then we define V=Vη(a)η(K)V^{\star}=V\cup\eta(a)\cup\eta(K) and W=E(K)V¯W^{\star}=\overline{E(K^{\star})-V^{\star}}. Since WW^{\star} is homeomorphic to the amalgamation of a handlebody and (T2disk)×I(T^{2}-disk)\times I along an once punctured annulus, it is a compression body, see Figure 1. Moreover, +W=+V=S\partial_{+}W^{\star}=\partial_{+}V^{\star}=S^{\star}. So VSWV^{\star}\cup_{S^{\star}}W^{\star} is a Heegaard splitting of E(K)E(K^{\star}) and g(K)g(S)=g(S)+1g(K^{\star})\leq g(S^{\star})=g(S)+1. Therefore g(K)g(K)+1g(K^{\star})\leq g(K)+1. Hence t(K)t(K)+1t(K^{\star})\leq t(K)+1, see also in [16]. However, t(K)+1t(K)+1 is not always the best upper bound of t(K)t(K^{\star}), see Example 1.1 as follows.

Refer to caption
Figure 1. Construction of a Heegaard splitting VSWV^{\star}\cup_{S^{\star}}W^{\star} of E(K)E(K^{\star})

Before stating Example 1.1, we introduce the definition of a rr-primitive knot KK in S3S^{3}. We assume that the torus boundary E(K)W\partial E(K)\subset W. Let rr be a slope in E(K)\partial E(K). If there is a spanning annulus A1WA_{1}\subset W with rAr\subset A and an essential disk D1VD_{1}\subset V such that A1D1A_{1}\cap D_{1} is a single point, then VSWV\cup_{S}W is called rr-primitive. Moreover, if E(K)E(K) admits a rr-primitive minimal Heegaard splitting, KK is called rr-primitive.

Example 1.1.

Let KK be a p/qp/q-primitive knot in S3S^{3}. Suppose VSWV\cup_{S}W is a p/qp/q-primitive mininal Heegaard splitting of E(K)E(K) with E(K)=W\partial E(K)=\partial_{-}W. Then there is a spanning annulus A1A_{1} and an essential disk D1D_{1} so that (1) |A1D1|=1|\partial A_{1}\cap\partial D_{1}|=1; (2) A1W\partial A_{1}\cap\partial_{-}W is a p/qp/q slope on T2T^{2}. In the above construction of the Heegaard splitting VSWV^{\star}\cup_{S^{\star}}W^{\star} of E(K)=E(K)Aη(K)E(K^{\star})=E(K)\cup_{A}\eta(K), let a0=aWa_{0}=a\cap W be the spanning arc of A1A_{1} which is disjoint from the point A1D1\partial A_{1}\cap\partial D_{1}. Since W=Wη(a0)¯(Adisk)[(T2disk)×I]W^{\star}=\overline{W-\eta(a_{0})}\cup_{(A-disk)}[(T^{2}-disk)\times I], [(A1a0)W]×I¯[(A1a0)+W]×I¯\overline{[(\partial A_{1}-a_{0})\cap\partial_{-}W]\times I}\cup\overline{[(\partial A_{1}-a_{0})\cap\partial_{+}W]\times I} is an essential disk of WW^{\star}, denoted by D2D_{2}. Note that D1D_{1} is an essential disk in VV^{\star} and |D1D2|=1|\partial D_{1}\cap\partial D_{2}|=1. Then VSWV^{\star}\cup_{S^{\star}}W^{\star} is stabilized, see Figure 2. Hence t(K)t(K)t(K^{\star})\leq t(K), see [12].

Refer to caption
(a) |D1A1|=1|\partial D_{1}\cap\partial A_{1}|=1
Refer to caption
(b) |D1D2|=1|\partial D_{1}\cap\partial D_{2}|=1
Figure 2. VSWV^{\star}\cup_{S^{\star}}W^{\star} is stabilized

Then there is a natural question: Can t(K)t(K)\mid t(K^{\star})-t(K)\mid be arbitrarily large? Unfortunately, the answer is negative. More precisely, we have the following theorem.

Theorem 1.1.

For any nontrivial cable knot KS3K^{\star}\subset S^{3} and its companion KS3K\subset S^{3}, t(K)t(K)t(K)+1t(K)\leq t(K^{\star})\leq t(K)+1.

Remark 1.1.

In a conference held in Xi’an Jiaotong University 2015, Tao Li announced that for any nontrivial satellite knot KS3K^{\star}\subset S^{3} and its companion KS3K\subset S^{3}, t(K)t(K)t(K^{\star})\geq t(K). But until now, we have found no preprint or publication containing this result.

By Example 1.1, if KK^{\star} is a (p,q)-cable knot while KK is p/qp/q-primitive, then t(K)t(K)t(K^{\star})\leq t(K). Hence in this case, t(K)=t(K)t(K^{\star})=t(K). So there is a problem as follows.

Problem 1.1.

For any nontrivial cable knot KS3K^{\star}\subset S^{3} and its companion KS3K\subset S^{3}, t(K)=t(K)t(K^{\star})=t(K) if and only if KK^{\star} is a (p,q)-cable knot and KK is p/qp/q-primitive.

There are some results related to this problem. For example, Moriah [16] proved that if KK is tunnel number one but no torus knot, then t(K)=t(K)+1t(K^{\star})=t(K)+1. If we consider E(K)E(K^{\star}) as the amalgamation of E(K)E(K) and η(K)\eta(K) along an annulus, it is more or less similar to the complement of the connected sum of two knots in S3S^{3}. For any two high distance knots, Gao, Guo and Qiu [3] proved that t(K1K2)=t(K1)+t(K2)+1t(K_{1}\sharp K_{2})=t(K_{1})+t(K_{2})+1. Since a p/qp/q-primitive Heegaard splitting has distance at most 2, Example 1.1 doesn’t hold on a high distance knot. Thus we guess that if KK is a high distance knot, t(K)=t(K)+1t(K^{\star})=t(K)+1.

Although there are many well properties of a high distance knot, given a knot KK in S3S^{3}, it is a little bit hard to determine whether E(K)E(K) admits a high distance Heegaard splitting or not. So for an arbitrary nontrivial knot KK, we consider the problem that how to properly choose KK^{\star} so that t(K)=t(K)+1t(K^{\star})=t(K)+1. In [11], Li introduced an idea to consider a sufficiently complicated gluing map in the calculation of minimal Heegaard genus of an amalgamated 3-manifold. As pointed earlier, E(K)E(K^{\star}) is also an amalgamation of two 3-manifolds along an annulus. Then we guess that there should be a similar result on tunnel numbers between KK^{\star} and KK . We present these two ideas in the following theorem.

Theorem 1.2.

Suppose KK^{\star} is a (p2,q)(p\geq 2,q)-cable knot over a nontrivial knot KK in S3S^{3}.

  1. (1)

    If E(K)E(K) admits a distance at least 2t(K)+52t(K)+5 Heegaard splitting, then t(K)=t(K)+1t(K^{\star})=t(K)+1.

  2. (2)

    Let InSInS be the collection of boundary slopes of essential surfaces properly embedded in E(K)E(K). Then there is a constant number 𝒩\mathcal{N} depending on KK so that if

    diam𝒞(T2)(p/q,InS)>𝒩,diam_{\mathcal{C}(T^{2})}(p/q,InS)>\mathcal{N},

    then t(K)=t(K)+1t(K^{\star})=t(K)+1.

Remark 1.2.
  • In fact, there are infinitely many high distance knots in S3S^{3}, see [15];

  • By the second conclusion in Theorem 1.2, for any given number nN+n\in N^{+}, we can construct a satellite knot KS3K^{\star}\subset S^{3} and its companion KK so that t(K)t(K)nt(K^{\star})-t(K)\geq n as follows. For any KS3K\subset S^{3}, let K1K^{1} be the cable knot so that t(K1)=t(K)+1t(K^{1})=t(K)+1. And then let K2K^{2} be the cable knot of K1K^{1} so that t(K2)=t(K1)+1t(K^{2})=t(K^{1})+1. It is not hard to see that K2K^{2} lies in the solid torus neighborhood of KK. By the same argument again and again, we have a knot KnK^{n} contained in the solid torus neighborhood of KK and t(Kn)=t(K)+nt(K^{n})=t(K)+n.

A Heegaard splitting V1S1W1V_{1}\cup_{S_{1}}W_{1} is a Dehn surgery of VSWV\cup_{S}W if there is an embedded simple closed curve cc in VV or WW so that V1V_{1} or W1W_{1} is a Dehn surgery of VV or WW along cc. Under the second condition in Theorem 1.2, there is no essential surface in E(K)E(K) with p/qp/q slopes as its boundary slopes. So this phenomenon derives out the following corollary.

Corollary 1.1.

Ler KK, KK^{\star} be the same as in Theorem 1.2. Suppose KK^{\star} satisfies the condition in Theorem 1.2 (2). Then each unstabilized Heegaard splitting of E(K)E(K^{\star}) is a Dehn surgery of one of E(K)E(K)’s.

The conclusion of Corollary 1.1 implies that each unstabilized Heegaard splitting of E(K)E(K^{\star}) is a Dehn surgery of one of E(K)E(K)’s. In reverse, doing a Dehn surgery on each unstabilized Heegaard splitting of E(K)E(K) also produces a Heegaard splitting of E(K)E(K^{\star}). Then there is a natural question:

Question 1.1.

For any two non isotopic but same genus Heegaard splittings of E(K)E(K), is it possible that doing two Dehn surgeries on them simultaneously produce two isotopic Heegaard splittings of E(K)E(K^{\star})?

Remark 1.3.

To our best knowledge, there is no evidence showing that whether it is true or false.

We will introduce some lemmas in Section 2, prove Theorem 1.1 in Section 3, Theorem 1.2 in Section 4 and Corollary 1.1 in Section 5.

2. Preliminaries

Let VSWV\cup_{S}W be a Heegaard splitting. It is stabilized if there is a pair of essential disks in VV and WW individually so that their boundary curves intersect in one point. Otherwise, it is unstabilized. A Heegaard splitting VSWV\cup_{S}W is reducible if there is an essential simple closed curve in SS bounding a disk in VV and also a disk in WW. Otherwise, it is irreducible. For any irreducible Heegaard splitting, Casson and Gordon [2] introduced a weakly reducible and irreducible Heegaard splitting as follows: VSWV\cup_{S}W is weakly reducible and irreducible if it is irreducible and there are a pair of disjoint essential disks in VV and WW individually. Otherwise, it is strongly irreducible. So if VSWV\cup_{S}W is unstabilized, it is either strongly irreducible or weakly reducible and irreducible.

If SS is a closed orientable and genus at least two surface, then there is a curve complex 𝒞(S)\mathcal{C}(S) defined on it as follows, see also in [5]. The vertices are the isotopic classes of essential simple closed curves in SS. A nn-simplex is a set of n+1n+1 isotopic classes of nonisotopic and pairwise disjoint essential simple closed curves in SS. Without further notation, we abuse a curve and its isotopic class. Then there is a distance defined on the one-skeleton 𝒞1(S)\mathcal{C}^{1}(S). Let α\alpha and β\beta be arbitrary two essential simple closed curves on SS. The distance of α\alpha and β\beta, denoted by d𝒞(S)(α,β)d_{\mathcal{C}(S)}(\alpha,\beta), is the minimal number of edges connecting α\alpha and β\beta in 𝒞1(S)\mathcal{C}^{1}(S), i.e., the minimal integer nn satisfying α0α1αn\alpha_{0}\rightarrow\alpha_{1}\rightarrow\cdots\rightarrow\alpha_{n} is an edge path in 𝒞1(S)\mathcal{C}^{1}(S) where α0=α\alpha_{0}=\alpha and αn=β\alpha_{n}=\beta and αiαi+1\alpha_{i}\rightarrow\alpha_{i+1} is an edge (i.e., αi\alpha_{i} is distinct and disjoint from αi+1\alpha_{i+1}) (i=0,1,,n1)(i=0,1,\cdots,n-1). It is not hard to see that linearly extending dd to the whole 𝒞1(S)\mathcal{C}^{1}(S) is a metric of it. It is well known that (𝒞1(S),d)(\mathcal{C}^{1}(S),d) is gromov hyperbolic, see [14]. If SS is a torus, there is also a curve complex defined on it, which is the Farey graph 𝒞(T2)\mathcal{C}(T^{2}). Similarly, the vertices are the isotopic classes of essential simple closed curves in T2T^{2}. We put an edge between two isotopic classes of essential, non isotopic, intersecting one point simple closed curves in T2T^{2}. Then for any two vertices α\alpha and β\beta in 𝒞(T2)\mathcal{C}(T^{2}), the distance d𝒞(T2)(α,β)d_{\mathcal{C}(T^{2})}(\alpha,\beta) is defined to be the minimal number of edges from α\alpha to β\beta in 𝒞(T2)\mathcal{C}(T^{2}). It is well known that the diameter of 𝒞(T2)\mathcal{C}(T^{2}) is infinite, i.e., diam(𝒞(T2))=diam(\mathcal{C}(T^{2}))=\infty.

There is a disk complex defined on either of VV and WW as follows. The vertices are the isotopic classes of essential simple closed curves in SS which bound essential disks in VV (resp. WW). A nn-simplex is a set of n+1n+1 isotopic classes of nonisotopic and pairwise disjoint essential simple closed curves in SS which bound essential disks in VV (resp. WW). Denote the disk complex of VV (resp. WW) by 𝒟(V)\mathcal{D}(V) (resp. 𝒟(W)\mathcal{D}(W)). It is not hard to see that 𝒟(V)𝒞(S)\mathcal{D}(V)\subset\mathcal{C}(S) and 𝒟(W)𝒞(S)\mathcal{D}(W)\subset\mathcal{C}(S). Hempel [9] defined the Heegaard distance d𝒞(S)(V,W)d_{\mathcal{C}(S)}(V,W) to be the distance between 𝒟(V)\mathcal{D}(V) and 𝒟(W)\mathcal{D}(W) in 𝒞(S)\mathcal{C}(S) and proved that VSWV\cup_{S}W is weakly reducible and irreducible if and only if d𝒞(S)(V,W)=1d_{\mathcal{C}(S)}(V,W)=1; VSWV\cup_{S}W is strongly irreducible if and only if d𝒞(S)(V,W)2d_{\mathcal{C}(S)}(V,W)\geq 2.

Let QQ be a properly embedded essential surface, i.e., incompressible and \partial-incompressible, in a compact orientable 3-manifold MM. Hartshorn [4] and Schalemann [19] proved that d𝒞(S)(V,W)2χ(Q)d_{\mathcal{C}(S)}(V,W)\leq 2-\chi(Q). Later, Scharlemann and Tomova [21], Li [10] extended this result into a general case.

Definition 2.1.

Let PP be a separating surface properly embedded in MM and MP¯=XY\overline{M-P}=X\cup Y. Suppose PP has compressing disks on both sides. We say PP is strongly irreducible if each compressing disk in XX meets each compressing disk in YY. Otherwise, PP is called weakly reducible.

Definition 2.2.

Let PP be a strongly irreducible surface. Then PP is called \partial-strongly irreducible if

  1. (1)

    every compressing and \partial-compressing disk in XX meets every compressing and \partial-compressing disk in YY, and

  2. (2)

    there is at least one compressing or \partial-compressing disk on each side of PP.

In Li’s proof of Theorem 1.1 in [10], it contains the following lemma:

Lemma 2.1.

Suppose MM is a compact orientable irreducible 3-manifold and PP is a separating strongly irreducible surface in MM. Let QQ be a properly embedded compact orientable surface in MM and suppose QQ is either essential or separating strongly irreducible. Then either

  1. (1)

    d(P)2χ(Q)d(P)\leq 2-\chi(Q), or

  2. (2)

    after isotopy, PtQ=P_{t}\cap Q=\emptyset for all tt, where Pt(t[0,1])P_{t}~{}(t\in[0,1]) is a level surface in a sweep-out for PP, or

  3. (3)

    PP and QQ are isotopic.

If M\partial M\neq\emptyset, there is at least one essential surface with boundary curves. When M\partial M are some tori, Hatcher [8] studied all boundary curves of essential surface in MM. In particular, if M\partial M is a torus, there is a finiteness result about boundary slopes of essential surfaces in MM as follows. For a general case, there are some results on it, see [6, 7, 13].

Lemma 2.2 (Corollary [8]).

Let MM be a compact, orientable, irreducible 3-manifold. If M=T2\partial M=T^{2}, there are finitely many slopes realized by boundary curves of essential surfaces in MM.

In particular, Bachman, Schleimer and Sedgewick [1] proved the following lemma:

Lemma 2.3 (Lemma 4.8[1]).

Suppose MM is compact orientable irreducible 3-manifold with a torus boundary. Let PP be a separating, properly embedded, connected surface in MM which is strongly irreducible, has non-empty boundary, and is not peripheral. Then either PP is \partial-strongly irreducible or P\partial P is at most distance one from the boundary of some properly embedded surface which is both incompressible and boundary-incompressible.

In general, for any essential surface and any strongly irreducible and \partial-strongly irreducible surface in MM, Li proved that there is an upper bound of distances between any two of their boundary curves in 𝒞(M)\mathcal{C}(\partial M).

Lemma 2.4 (Lemma 3.7 [11]).

Suppose MM is not an II-bundle and has a connected boundary. Let CgC_{g} be the collection of orientable surfaces properly embedded in MM with genus no more than gg and boundary is essential in M\partial M. Let PP and QQ be surfaces in CgC_{g}. Suppose QQ is essential and suppose PP is either essential or strongly irreducible and \partial-strongly irreducible. Then there exists a number KK^{{}^{\prime}} that depends only on gg, such that the distance d(P,Q)Kd(\partial P,\partial Q)\leq K^{{}^{\prime}} in 𝒞(M)\mathcal{C}(\partial M).

At the end of this section, we introduce a lemma about essential annuli and disks in a compression body.

Lemma 2.5 (Lemma 3.1 [17]).

Let VV be a nontrivial compression body, and let 𝒜\mathcal{A} be a collection of pairwise disjoint essential annuli properly embedded in VV. Then there is an essential disk properly embedded in VV disjoint from 𝒜\mathcal{A}.

3. t(K)t(K)t(K^{\star})\geq t(K)

Let VSWV^{\star}\cup_{S^{\star}}W^{\star} be a minimal genus Heegaard splitting of E(K)E(K^{\star}). Since E(K)E(K^{\star}) is irreducible, VSWV^{\star}\cup_{S^{\star}}W^{\star} is either strongly irreducible or weakly reducible and irreducible.

3.1. VSWV^{\star}\cup_{S^{\star}}W^{\star} is strongly irreducible.

Since AA is an essential annulus in E(K)E(K^{\star}), by Schultens lemma [22], SAS^{\star}\cap A consists of nonzero and finitely many essential simple closed curves in both of them. Under this condition, we assume that SA\mid S^{\star}\cap A\mid is minimal.

Claim 3.1.

One component of SA¯\overline{S^{\star}-A} is strongly irreducible while the others are incompressible in E(K)AE(K^{\star})-A.

Proof.

Not only AVA\cap V^{\star} but also AWA\cap W^{\star} consist of finitely many essential annuli. For if not, one of them contains at least one boundary parallel annulus. Then we do an isotopy on AA to reduce SA\mid S^{\star}\cap A\mid, which is against the minimal assumption of SA\mid S^{\star}\cap A\mid. By Lemma 2.5, there is an essential disk DVA¯D\subset\overline{V^{\star}-A} (resp. EWA¯E\subset\overline{W^{\star}-A} ). Then D\partial D is contained in one subsurface of SAS^{\star}-A, says S1S_{1}. Since VSWV^{\star}\cup_{S^{\star}}W^{\star} is strongly irreducible, ES1\partial E\subset S_{1}. Then any essential disk of VV^{\star} disjoint from AVA\cap V^{\star} has its boundary curve in S1S_{1}. So any essential disk of WW^{\star} disjoint from AWA\cap W^{\star} has its boundary curve in S1S_{1}. Hence S1S_{1} is the only compressible subsurface in SA¯\overline{S^{\star}-A}. Since S1SS_{1}\subset S^{*} is essential, i.e., the inclusion map on its fundamental group is injective, S1S_{1} is strongly irreducible.

Let S1S_{1} be the strongly irreducible surface of SA¯\overline{S^{\star}-A}. Recall that E(K)=E(K)Aη(K)E(K^{\star})=E(K)\cup_{A}\eta(K). Since its interior is disjoint from AA, S1S_{1} lies in either E(K)E(K) or the solid torus η(K)\eta(K), abbreviated by JJ. If S1E(K)S_{1}\subset E(K), then SJS^{\star}\cap J consists of finitely many nested annuli, denoted by {A1,..,An}\{A_{1},..,A_{n}\}, for some nN+n\in N^{+}.

For each 1in1\leq i\leq n, AiA\partial A_{i}\cap A is a pair of essential curves. Then there is a 1jn1\leq j\leq n so that AjA\partial A_{j}\cap A is the innermost in AA which bounds an annulus A0AA_{0}\subset A. So A0A_{0} is an essential annulus in one compression body, says in VV^{\star} for example. For if not, we do an isotopy on AA to reduce SA\mid S^{\star}\cap A\mid. On one side, since A0=Aj\partial A_{0}=\partial A_{j}, A0\partial A_{0} bounds an annulus in SS^{\star}. It means that there is a pair of two isotopic essential simple closed curves in SS^{\star} bounding an essential annulus in VV^{\star}. By the standard outermost disk argument, there is a boundary compression in A0A_{0} producing an essential disk D0D_{0}. It is not hard to see that D0D_{0} is separating and cuts out a solid torus STST in VV^{\star}. On the other side, A0AjA_{0}\cup A_{j} also bounds a solid torus in JJ, denoted by ST0ST_{0}. Moreover, ST0STST_{0}\subset ST.

Let ll be the longitude of ST0ST_{0}. Then we push it a little into the interior of ST0ST_{0}. For simplicity, it is still denoted by ll. Removing a regular neighbor of ll, denoted by η(l)\eta(l), in ST0ST_{0} and STST makes STST into a torus I-bundle, where the disk D0D_{0} is in one of its boundary surface. Since VV^{\star} is cutten into the solid torus STST and a genus less one compression body or handlebody, Vη(l)V^{\star}-\eta(l) is still a compression body but with one more negative boundary surface. In STη(l)ST-\eta(l), we attach a 2-handle along an essential simple closed curve in η(l)\partial\eta(l) and a 3-ball to cancel the resulted 2-sphere so that the resulted solid torus is actually the A0×IA_{0}\times I and also Aj×IA_{j}\times I. Similarly we attach a 2-handle addition along the same essential simple closed curve on Vη(l)V^{\star}-\eta(l). So it produces a new compression body VV, where +V=S\partial_{+}V=S^{\star}. Then VSWV\cup_{S^{\star}}W^{\star} is still a Heegaard splitting. For simplicity, we replace SS^{\star} by SS, WW^{\star} by WW. Then VSWV\cup_{S}W is a Heegaard splitting. As we do the Dehn surgery in ST0JST_{0}\subset J, the 3-manifold M=VSWM=V\cup_{S}W is an amalgamation of E(K)E(K) and a solid torus along AA. Moreover,

Claim 3.2.

VSWV\cup_{S}W is a Heegaard splitting of E(K)E(K).

Proof.

Since SJS^{\star}\cap J consists of finitely many nested annuli in JJ which are not parallel to AA, by doing the Dehn surgery in VV^{\star}, the solid torus ST0JST_{0}\subset J bounded by A0AjA_{0}\cup A_{j} is changed into a new solid torus so that AjA_{j} is parallel to A0A_{0}. It means that among of all these annuli {A1,..,An}\{A_{1},..,A_{n}\}, each one is parallel to A0A_{0}. So we do an isotopy on SS so that it is disjoint from AA. Therefore, AA is in either VV or WW. Since AA has the same core curve with A0A_{0}, AA is incompressible. Therefore, AA is boundary parallel in VV or WW. So AA cuts out the I-bundle A×IA\times I in MM. It means that MM is homeomorphic to E(K)E(K). ∎

By Claim 3.2, g(S)=g(S)g(S^{\star})=g(S). Since g(S)g(S^{\star}) is a minimal Heegaard genus of E(K)E(K^{\star}), g(K)g(K)g(K^{\star})\geq g(K). So t(K)t(K)t(K^{\star})\geq t(K).

Otherwise, S1S_{1} lies in the solid torus JJ. Recall that a strongly irreducible surface is bicompressible, i.e., being compressed in its two sides, and weakly incompressible, i.e., no disjoint compression disks from its two sides. Scharlemann[18] studied the bicompressible but weakly incompressible surfaces in a solid torus, proved the following lemma.

Lemma 3.1 (Proposition 3.2[18]).

Let cc be essential simple closed curve in J\partial J so that it neither bounds a disk in JJ nor intersects the essential disk in JJ in one point. Then for any bicompressible, weakly incompressible surface with cc as its boundary curve, it is either a boundary parallel incompressible annulus with a tube parallel to an arc in J\partial J attached or the tube sum of two boundary parallel incompressible annuli and the tube is parallel to an arc in J\partial J.

So S1S_{1} is either a boundary parallel annulus with a tube attached or the tube sum of two boundary parallel annuli in JJ. In case of a long argument, we divide its proof into these two lemmas 3.2 and 3.3.

Lemma 3.2.

If S1S_{1} is a boundary parallel annulus AA^{{}^{\prime}} with a tube attached in JJ, then t(K)t(K)t(K^{\star})\geq t(K).

Proof.

Since A\partial A^{{}^{\prime}} bounds a subannulus A′′AA^{{}^{\prime\prime}}\subset A, AA^{{}^{\prime}} is parallel to either A′′A^{{}^{\prime\prime}} or its complement annulus in J\partial J.

(1) If AA^{{}^{\prime}} is parallel to a subannulus A′′AA^{{}^{\prime\prime}}\subset A, then int(A′′)S=int(A^{{}^{\prime\prime}})\cap S^{\star}=\emptyset up to isotopy, see Figure 3. For if not, then there is an annulus in SJS^{\star}\cap J parallel to AA. So we do an isotopy along this annulus to reduce SA\mid S^{\star}\cap A\mid. Then A′′A^{{}^{\prime\prime}} is an essential annulus in one of VV^{\star} and WW^{\star}. Without loss of generality, we assume that A′′VA^{{}^{\prime\prime}}\subset V^{\star}. The other case is similar. So we omit it. Then cutting VV^{\star} along A′′A^{{}^{\prime\prime}} produces two handlebodies or one handlebody and a compression body. Let VV^{{}^{\prime}} be the compression body or the handlebody containing no S1S_{1}.

Refer to caption
Figure 3.

By Claim 3.1, except S1S_{1}, all other components of SJS^{\star}\cap J are nested annuli in JJ. If there is no annulus in SJS^{\star}\cap J, there are a pair of two spanning annuli, says A2′′A^{2^{\prime\prime}} and A3′′A^{3^{\prime\prime}} so that one component of WA2′′A3′′W^{\star}-A^{2^{\prime\prime}}\cup A^{3^{\prime\prime}}, says NN, lies in E(K)E(K). It is not hard to see that NN is a handlebody and A′′N\partial A^{{}^{\prime\prime}}\subset N. Let A′′×IA^{{}^{\prime\prime}}\times I be I-bundle in VV^{{}^{\prime}}. Then we remove this I-bundle from VV^{{}^{\prime}} and attach it to NN along these two annuli A′′×I\partial A^{{}^{\prime\prime}}\times I. So NN is changed into a new 3-manifold, says N0N_{0}. In this case, N0\partial N_{0} contains E(K)\partial E(K). Let aa be a vertical arc connecting A′′×{0,1}A^{{}^{\prime\prime}}\times\{0,1\} in A′′×IA^{{}^{\prime\prime}}\times I. So aa is a fiber arc in N0N_{0} connecting E(K)\partial E(K) and the other boundary surface of N0N_{0}. We remove the union of a regular neighborhood of E(K)\partial E(K) and a regular neighborhood of aa in N0N_{0} from N0N_{0}. Then N0N_{0} is changed into a 3-manifold WW, which is also the amalgamation of NN and a tubed annulus I-bundle along two essential annuli. It is not hard to see that WW is also a handlebody and furthermore E(K)W¯\overline{E(K)-W} is the disk sum of VV^{{}^{\prime}} and E(K)×I\partial E(K)\times I, denoted by VV. So VV is a compression body and V+V=WWV\cup_{\partial_{+}V=\partial W}W is a Heegaard splitting of E(K)E(K). During the process of this surgery, g(W)=g(S)g(\partial W)=g(S^{\star}). It means that g(K)g(K)g(K^{\star})\geq g(K) and t(K)t(K)t(K^{\star})\geq t(K).

So we assume that there is at least one annulus in SJS^{\star}\cap J. Let A2A^{2} be the innermost annulus in SJS^{\star}\cap J, i.e., no other component of SJS^{\star}\cap J lies between it and S1S_{1}. Then A2A^{2} is not boundary parallel into AA. For if not, then we can do an isotopy on AA to reduce SJS^{\star}\cap J. Let VV^{{}^{\prime}} be the handlebody or compression body as above. Then E(K)V¯\overline{E(K^{\star})-V^{{}^{\prime}}} contains a smaller copy of JJ. Then we do a Dehn surgery on this smaller copy of JJ along its longitude so that A2A^{2} is parallel to AA^{{}^{\prime}}. Then E(K)V¯\overline{E(K^{\star})-V^{{}^{\prime}}} is changed into a new 3-manifold, denoted by NN again. By the same argument of the case that SJS^{\star}\cap J is incompressible, NVN\cup V^{{}^{\prime}} is homeomorphic to E(K)E(K). Since A2AA^{2}\cup A^{{}^{\prime}} bounds an annulus I-bundle, let aa be a fiber arc connecting A2A^{2} and AA^{{}^{\prime}} in this I-bundle. Then we remove a regular neighborhood of aa and then NN is changed into a 3-manifold WW. By the same argument as above, WW is a compression body or handlebody. Without loss of generality, we assume that WW is a handlebody. Meanwhile, the union of VV^{{}^{\prime}} and the closed regular neighborhood of aa is a compression body. So V+V=WWV\cup_{\partial_{+}V=\partial W}W is a Heegaard splitting of E(K)E(K). By the same argument, t(K)t(K)t(K^{\star})\geq t(K).

(2) If AA^{{}^{\prime}} is parallel to the complement annulus of A′′A^{{}^{\prime\prime}} in J\partial J, then int(A′′)S=int(A^{{}^{\prime\prime}})\cap S^{\star}=\emptyset. For if not, by Claim 3.1, then there is an innermost annulus A′′′′A{{}^{\prime\prime\prime\prime}} of SJS^{\star}\cap J so that its boundary curves lie in the interior of A′′A^{{}^{\prime\prime}}. By minimality of SA\mid S^{\star}\cap A\mid, A′′′′A^{{}^{\prime\prime\prime\prime}} is not boundary parallel to AA. Then A′′′′A′′A^{{}^{\prime\prime\prime\prime}}\cup A^{{}^{\prime\prime}} bounds a smaller copy of JJ. By the similar argument as above, we do a Dehn surgery on this smaller copy of JJ and get a Heegaard splitting of E(K)E(K). Moreover, t(K)t(K)t(K^{\star})\geq t(K).

It is not hard to see that the dual core curve of the tube in S1S_{1} bounds an essential disk EE, says in WW^{\star}. Then we do a compression on WW^{\star} along the disk EE and get a new 3-manifold WW^{{}^{\prime}}. Moreover, S1S_{1} is changed into an annulus AA^{{}^{\prime}} in JJ. Then AA′′A^{{}^{\prime}}\cup A^{{}^{\prime\prime}} bounds a smaller copy of JJ in E(K)W¯\overline{E(K^{\star})-W^{{}^{\prime}}}. Denote N=E(K)W¯N=\overline{E(K^{\star})-W^{{}^{\prime}}}. If A′′A^{{}^{\prime\prime}} is an essential annulus in VV^{\star}, then NN is an amalgamation of a handlebody or a compression body VV^{{}^{\prime}} and a smaller copy of JJ along A′′A^{{}^{\prime\prime}}. So we do a Dehn surgery as above so that the smaller copy of JJ is changed into the I-bundle A′′×IA^{{}^{\prime\prime}}\times I and NN is changed into a 3-manifold VV, where VV is the union of VV^{{}^{\prime}} and A′′×IA^{{}^{\prime\prime}}\times I. Then VV is homeomorphic to VV^{{}^{\prime}}. By the same argument as above, VWV^{{}^{\prime}}\cup W is a Heegaard splitting of E(K)E(K) and t(K)t(K)+1t(K^{\star})\geq t(K)+1.

Otherwise, A′′A^{{}^{\prime\prime}} is an essential annulus in WW^{\star}. Then both of AA^{{}^{\prime}} and A′′A^{{}^{\prime\prime}} are in WW^{\star}, where they bound a smaller copy of JJ. Then we do a Dehn surgery on this smaller copy of JJ so that they are parallel. Hence WW^{\star} is changed into a compression body or handlebody WW. Replace VV^{\star} by VV. So VSWV\cup_{S^{\star}}W is a Heegaard splitting of E(K)E(K) and t(K)t(K)t(K^{\star})\geq t(K). ∎

Lemma 3.3.

If S1S_{1} is the tube sum of two annuli in JJ, then t(K)t(K)t(K^{\star})\geq t(K).

Proof.

Let S1S_{1} be the tube sum of two annuli A1A_{1} and A2A_{2} in JJ. We say there is no other component of SJS^{\star}\cap J in the region between A1A_{1} and A2A_{2} in JJ. For if not, then the tube would not connect them. It is not hard to see that A1\partial A_{1} (resp. A2\partial A_{2}) bounds an annulus in AA. Without loss of generality, we assume that the annulus bounded by A1\partial A_{1} in AA doesn’t contain A2\partial A_{2}. Then we say that there is no other component of SJS^{\star}\cap J, of which boundary curves lies in the annulus bounded by A1\partial A_{1} in AA. Otherwise, we either do an isotopy on AA to reduce SA\mid S^{\star}\cap A\mid or do a Dehn surgery as above and get t(K)t(K)t(K^{\star})\geq t(K).

There are three types of A1A_{1} and A2A_{2} as follows:

  • (1) one of them is parallel to AA, says A1A_{1} while A2A_{2} not;

  • (2) both of them are parallel to AA;

  • (3) neither of them is parallel to AA.

For the first case, either A1\partial A_{1} separate A2\partial A_{2} in AA or not. If A1\partial A_{1} separate A2\partial A_{2}, then A1A2A_{1}\cup A_{2} bounds a smaller copy of JJ. Then we do a compression on this tube along the disk DD bounded by the core curve. Without loss of generality, we assume that DVD\subset V^{\star}. So VV^{\star} is changed into a new 3-manifold, denoted by VV^{{}^{\prime}}, which is also a compression body or handlebody. Denoted E(K)V¯\overline{E(K^{\star})-V^{{}^{\prime}}} by NN. Again we do a Dehn surgery on NN as in the proof of Lemma 3.2 and so the smaller copy of JJ is changed into the annulus I-bundle A1×IA_{1}\times I. So NN is changed into NN^{{}^{\prime}} and NV=E(K)N^{{}^{\prime}}\cup V^{{}^{\prime}}=E(K). Let aa be the vertical arc of A1×IA_{1}\times I, which connects A1×{0,1}A_{1}\times\{0,1\}. Then W=NN(a)W=N^{{}^{\prime}}-N(a) is a compression body or handlebody. Meanwhile, the union of VV^{{}^{\prime}} and the closed regular neighborhood of aa is a handlebody or compression body, denoted by VV. Moreover, g(+V)=g(S)g(\partial_{+}V)=g(S^{\star}). So t(K)t(K)t(K^{\star})\geq t(K). If A1\partial A_{1} doesn’t separate A2\partial A_{2}, then there is a smaller copy of JJ bounded by A2A_{2}. So we do a dehn surgery as above. Then t(K)t(K)t(K^{\star})\geq t(K).

For the second case, A1A_{1} and A2A_{2} are not nested. Without loss of generality, we assume that the disk DD bounded by core curve of this tube lies in VV^{*}. Then we do a compression on S1S_{1} along an essential disk EE in WW^{*}. So WW^{*} is changed into a new compression body or handlebody, says WW^{{}^{\prime}}. Since A1\partial A_{1} and A2\partial A_{2} bound two essential annuli in VV^{*}, cutting VV^{*} along them produce a handlebody or compression body, says VV^{{}^{\prime}}. Then E(K)WE(K^{*})-W^{{}^{\prime}} is the amalgamation of VV^{{}^{\prime}} and an annulus I-bundle or a smaller copy of JJ. For the first case, removing a closed neighborhood of a fiber arc in this I-bundle changes E(K)WE(K^{*})-W^{{}^{\prime}} into a handlebody or compression body VV. For the later case, we do a dehn surgery on this smaller copy of JJ so that it is changed into an annulus I-bundle. In both of these two cases, t(K)t(K)t(K^{\star})\geq t(K).

For the third case, A1A_{1} and A2A_{2} are parallel in JJ and there is a smaller copy of JJ bounded by a subannulus in AA and A1A_{1}. For if not, then both of A1A_{1} and A2A_{2} lie in an I-bundle of AA. Then we do an isotopy to reduce SAS^{\star}\cap A. We do a compression on this tube along the disk DD bounded by the core curve. Without loss of generality, we assume that DVD\subset V^{\star}. So VV^{\star} is changed into a new 3-manifold, denoted by VV^{{}^{\prime}}, which is also a compression body or handlebody. Denoted E(K)V¯\overline{E(K^{\star})-V^{{}^{\prime}}} by NN. Since there is no other component of SJS^{\star}\cap J lie between A1A_{1} and AA, this smaller copy of JJ lies in VV^{{}^{\prime}}. By the same argument in the proof of Lemma 3.2, we do a Dehn surgery on this smaller copy of JJ so that it is changed into A1×IA_{1}\times I. So VV^{{}^{\prime}} is changed into a new compression body or handlebody V′′V^{{}^{\prime\prime}}. Moreover, NV′′=E(K)N\cup V^{{}^{\prime\prime}}=E(K). On one side, we attach a closed regular neighborhood of a vertical fiber arc aa in the I-bundle bounded by A1A_{1} and A2A_{2} to V′′V^{{}^{\prime\prime}} so that V′′V^{{}^{\prime\prime}} is changed into a compression body or handlebody, says VV. On the other side, W=NN(a)¯W=\overline{N-N(a)} is still a handlebody or compression body and +W=+V\partial_{+}W=\partial_{+}V. During this process, g(+V)=g(S)g(\partial_{+}V)=g(S^{\star}) and t(K)t(K)t(K^{\star})\geq t(K). ∎

Refer to caption
Refer to caption
Figure 4.

3.2. VSWV^{\star}\cup_{S^{\star}}W^{\star} is weakly reducible and irreducible.

By the main result in [20], VSWV^{\star}\cup_{S^{\star}}W^{\star} has an untelescoping, says (V1S1W1)F1Fn1(VnSnWn)(n2)(V_{1}\cup_{S_{1}}W_{1})\cup_{F_{1}}...\cup_{F_{n-1}}(V_{n}\cup_{S_{n}}W_{n})~{}(n\geq 2). We assume that each component of [(i=1nSi)(i=1n1Fi)]A[(\mathop{\cup}\limits_{i=1}^{n}S_{i})\cup(\mathop{\cup}\limits_{i=1}^{n-1}F_{i})]\cap A is essential in both of them up to isotopy. Under this condition, we assume that [(i=1nSi)(i=1n1Fi)]A\mid[(\mathop{\cup}\limits_{i=1}^{n}S_{i})\cup(\mathop{\cup}\limits_{i=1}^{n-1}F_{i})]\cap A\mid is minimal. Then each component of (i=1n1Fi)J(\mathop{\cup}\limits_{i=1}^{n-1}F_{i})\cap J is an incompressible annulus in JJ, which is not boundary parallel to AA in JJ. So (i=1n1Fi)J(\mathop{\cup}\limits_{i=1}^{n-1}F_{i})\cap J are nested annuli in JJ. Let A1A_{1} be the innermost one among (i=1n1Fi)J(\mathop{\cup}\limits_{i=1}^{n-1}F_{i})\cap J. Since A1\partial A_{1} bounds an annulus AAA^{{}^{\prime}}\subset A, A1AA_{1}\cup A^{{}^{\prime}} bounds a smaller copy of JJ, denoted by JJ^{{}^{\prime}}. Then JJ^{{}^{\prime}} lies in some ViSiWiV_{i}\cup_{S_{i}}W_{i}, for some 1in1\leq i\leq n.

By Claim 3.1, SiJS_{i}\cap J contains at most one strongly irreducible surface, denoted by Si,1S_{i,1}. If there is no strongly irreducible surface in SiJS_{i}\cap J, by the similar argument as above, we do a Dehn surgery on ViSiWiV_{i}\cup_{S_{i}}W_{i} so that JJ^{{}^{\prime}} is changed into the annulus I-bundle A×IA^{{}^{\prime}}\times I. Then the Heegaard splitting ViSiWiV_{i}\cup_{S_{i}}W_{i} is changed into Vi,1SiWi,1V_{i,1}\cup_{S_{i}}W_{i,1}. So the amalgamation (V1S1W1)F1Fi1(Vi,1SiWi,1)FiFn1(VnSnWn)(n2)(V_{1}\cup_{S_{1}}W_{1})\cup_{F_{1}}...\cup_{F_{i-1}}(V_{i,1}\cup_{S_{i}}W_{i,1})\cup_{F_{i}}...\cup_{F_{n-1}}(V_{n}\cup_{S_{n}}W_{n})~{}(n\geq 2) is a Heegaard splitting of E(K)E(K). Hence g(K)g(K)g(K^{\star})\geq g(K) and hence t(K)t(K)t(K^{\star})\geq t(K).

So we assume that Si,1S_{i,1} is strongly irreducible in JJ^{{}^{\prime}}. By Lemma 3.1, Si,1S_{i,1} is either a boundary parallel annulus with a tube attached or the tube sum of two boundary parallel annuli in JJ^{{}^{\prime}}. Though the argument is almost same to the proofs in Lemma 3.2 and 3.3, it is slightly different. Without loss of generality, we assume that the dual disk to the tube lies in ViV_{i}. For the first case, we do a compression on Si,1S_{i,1} along a non-separating essential disk DD in WiW_{i}. On one hand, WiW_{i} is changed into a genus less one compression body or handlebody Wi,1W_{i,1}. On the other hand, Si,1\partial S_{i,1} bounds an incompressible annulus A′′AA^{{}^{\prime\prime}}\subset A^{{}^{\prime}}. So A′′A^{{}^{\prime\prime}} is an essential annulus in ViV_{i}. Then one component of ViA′′¯\overline{V_{i}-A^{{}^{\prime\prime}}} contains no Si,1S_{i,1}, denoted by V′′V^{{}^{\prime\prime}}. It is not hard to see that V′′V^{{}^{\prime\prime}} is a handlebody or compression body. Since Si,1S_{i,1} is a genus one, two boundary curves compact surface, Si,1S_{i,1} is changed into an annulus A2A^{2} and A2=Si,1=A′′\partial A^{2}=\partial S_{i,1}=\partial A^{{}^{\prime\prime}}. It means that the complement of Wi,1W_{i,1} in Mi=ViSiWiM_{i}=V_{i}\cup_{S_{i}}W_{i}, denoted by NiN_{i}, is the amalgamation of V′′V^{{}^{\prime\prime}} and the solid torus bounded by A2A′′A^{2}\cup A^{{}^{\prime\prime}} along A′′A^{{}^{\prime\prime}}.

It is known that A2A^{2} is incompressible in JJ. Then either A2A^{2} is parallel to A′′A^{{}^{\prime\prime}} or there is a smaller copy of JJ^{{}^{\prime}} bounded by A2A′′A^{2}\cup A^{{}^{\prime\prime}}. If A2A^{2} is parallel to A′′A^{{}^{\prime\prime}}, then NiN_{i} is homeomorphic to V′′V^{{}^{\prime\prime}}. So it is a handlebody or compression body, denoted by Vi,1V_{i,1}. It means that Vi,1+Vi,1Wi,1V_{i,1}\cup_{\partial_{+}V_{i,1}}W_{i,1} is a genus [g(Si)1][g(S_{i})-1] Heegaard splitting of Vi+ViWiV_{i}\cup_{\partial_{+}V_{i}}W_{i}, which is impossible. So A2A′′A^{2}\cup A^{{}^{\prime\prime}} bounds a smaller copy of JJ^{{}^{\prime}}. Then by the same argument in Lemma 3.3, we do a Dehn surgery on JJ^{{}^{\prime}} in NiN_{i} so that NiN_{i} is changed into a compression body or handlebody Vi,1V_{i,1}. And Vi,1+Vi,1Wi,1V_{i,1}\cup_{\partial_{+}V_{i,1}}W_{i,1} is a genus [g(Si)1][g(S_{i})-1] Heegaard splitting. By the similar argument, (V1S1W1)F1Fi1(Vi,1SiWi,1)FiFn1(VnSnWn)(n2)(V_{1}\cup_{S_{1}}W_{1})\cup_{F_{1}}...\cup_{F_{i-1}}(V_{i,1}\cup_{S_{i}}W_{i,1})\cup_{F_{i}}...\cup_{F_{n-1}}(V_{n}\cup_{S_{n}}W_{n})~{}(n\geq 2) is a Heegaard splitting of E(K)E(K). So t(K)t(K)+1t(K^{\star})\geq t(K)+1.

The left case is that Si,1S_{i,1} is the tube sum of two boundary parallel annuli in JJ^{{}^{\prime}}. By the same argument in Lemma 3.3, ViSiWiV_{i}\cup_{S_{i}}W_{i} is changed into Vi,1+Vi,1Wi,1V_{i,1}\cup_{\partial_{+}V_{i,1}}W_{i,1}, where g(Si)=g(+Vi,1)g(S_{i})=g(\partial_{+}V_{i,1}). Moreover, (V1S1W1)F1Fi1(Vi,1SiWi,1)FiFn1(VnSnWn)(n2)(V_{1}\cup_{S_{1}}W_{1})\cup_{F_{1}}...\cup_{F_{i-1}}(V_{i,1}\cup_{S_{i}}W_{i,1})\cup_{F_{i}}...\cup_{F_{n-1}}(V_{n}\cup_{S_{n}}W_{n})~{}(n\geq 2) is a Heegaard splitting of E(K)E(K). So t(K)t(K)t(K^{\star})\geq t(K).

4. The proof of Theorem 1.2

Let KK, KK^{\star}, t(K)t(K), t(K)t(K^{\star}), T2T^{2} and 𝒞(T2)\mathcal{C}(T^{2}) be the same as in Section 1. We rewrite Theorem 1.2 as the following proposition:

Proposition 4.1.

Suppose KK^{\star} is a (p2,q)(p\geq 2,q)-cable knot over a nontrivial knot KK in S3S^{3}.

  1. (1)

    If E(K)E(K) admits a distance at least 2t(K)+52t(K)+5 Heegaard splitting, then t(K)=t(K)+1t(K^{\star})=t(K)+1.

  2. (2)

    Let InSInS be the collection of boundary slopes of essential surfaces properly embedded in E(K)E(K). Then there is a constant 𝒩\mathcal{N} depending on KK so that if

    diam𝒞(T2)(p/q,InS)>𝒩,diam_{\mathcal{C}(T^{2})}(p/q,InS)>\mathcal{N},

    then t(K)=t(K)+1t(K^{\star})=t(K)+1.

Proof.

Since KK^{\star} is in T2=η(K)=E(K)T^{2}=\partial\eta(K)=\partial E(K), we can slightly push KK^{\star} into the interior of η(K)\eta(K). Then E(K)=E(K)T2CE(K^{\star})=E(K)\cup_{T^{2}}C where C=(T2×I)Aη(K)C=(T^{2}\times I)\cup_{A}\eta(K), see Figure 5.

Refer to caption
Figure 5. C=(T2×I)Aη(K)C=(T^{2}\times I)\cup_{A}\eta(K)

On one side, since KK is nontrivial, T2T^{2} is incompressible in E(K)E(K). On the other side, T2T^{2} is incompressible in CC and not parallel to CT2=E(K)\partial C-T^{2}=\partial E(K^{\star}). Then T2T^{2} is essential in E(K)E(K^{\star}). We will prove Proposition 4.1 (1) in subsection 4.1 and 4.1 (2) in subsection 4.2.

4.1. Proof of Proposition 4.1 (1)

Let VSWV^{\star}\cup_{S^{\star}}W^{\star} be a minimal genus Heegaard splitting of E(K)E(K^{\star}) with V=E(K)\partial_{-}V^{\star}=\partial E(K^{\star}). Then it is either strongly irreducible or weakly reducible and irreducible.

Lemma 4.1.

VSWV^{\star}\cup_{S^{\star}}W^{\star} is weakly reducible and irreducible.

Proof.

Suppose the conclusion is false. Then VSWV^{\star}\cup_{S^{\star}}W^{\star} is strongly irreducible. Since T2E(K)T^{2}\subset E(K^{\star}) is essential, then T2T^{2} intersects SS^{\star} nontrivially up to isotopy. By Schultens’ lemma [22], we assume that (1) |ST2||S^{\star}\cap T^{2}| is minimal; (2) each simple closed curve of ST2S^{\star}\cap T^{2} is essential in both SS^{\star} and T2T^{2}.

Claim 4.1.

There is at most one strongly irreducible component in SE(K)S^{\star}\cap E(K) while others are essential in E(K)E(K).

Proof.

Since T2VT^{2}\cap V^{\star} (resp. T2WT^{2}\cap W^{\star}) is a collection of disjoint essential annuli in VV^{\star} (resp. WW^{\star}), by Lemma 2.5, there is a compressible disk BB (resp. DD) in VV^{\star} (resp. WW^{\star}) so that BB (resp. DD) disjoint from T2VT^{2}\cap V^{\star} (resp. T2WT^{2}\cap W^{\star}). Since SS^{\star} is strongly irreducible, both BB and DD lie in E(K)E(K) or CC . Furthermore, B\partial B and D\partial D lie in a same component S1S_{1} of ST2¯\overline{S^{\star}-T^{2}}. Moreover, S1S_{1} is strongly irreducible while other components of ST2¯\overline{S^{\star}-T^{2}} are essential. For if not, then there is another compressible component of ST2¯\overline{S^{\star}-T^{2}}. It means that SS^{\star} is weakly reducible, see Figure 6. ∎

Refer to caption
Figure 6. SE(K)S^{\star}\cap E(K)

Let VSWV\cup_{S}W be a distance at least 2t(K)+52t(K)+5 Heegaard splitting of E(K)E(K). Then SS is strongly irreducible in both E(K)E(K) and E(K)E(K^{\star}). By Claim 4.1, there is at most one strongly irreducible component in SE(K)S^{\star}\cap E(K) while others are essential in E(K)E(K). If SE(K)S^{\star}\cap E(K) contains only one strongly irreducible component S1S_{1}, then S1S_{1} is separating. Since S1S_{1} intersects T2T^{2} nontrivially up to isotopy, S1S_{1} and SS are not well-separated. Moreover, S1S_{1} is not isotopic to SS. Then by Lemma 2.1, d𝒞(S)(V,W)2χ(S1)d_{\mathcal{C}(S)}(V,W)\leq 2-\chi(S_{1}). Since S1\partial S_{1} is essential in SS^{\star}, 2χ(S1)2χ(SE(K))2χ(S)=2g(S)=2g(K)=2t(K)+22t(K)+42-\chi(S_{1})\leq 2-\chi(S^{\star}\cap E(K))\leq 2-\chi(S^{\star})=2g(S^{\star})=2g(K^{\star})=2t(K^{\star})+2\leq 2t(K)+4. Then d𝒞(S)(V,W)2t(K)+4d_{\mathcal{C}(S)}(V,W)\leq 2t(K)+4. A contradiction. Otherwise, SE(K)S^{\star}\cap E(K) contains an essential subsurface S1S_{1} in E(K)E(K). Then by the same argument, d𝒞(S)(V,W)2t(K)+4d_{\mathcal{C}(S)}(V,W)\leq 2t(K)+4. A contradicition.

So VSWV^{\star}\cup_{S^{\star}}W^{\star} is weakly reducible and irreducible. By [20], VSWV^{\star}\cup_{S^{\star}}W^{\star} has an untelescoping (V1S1W1)F1Fn1(VnSnWn)(V_{1}\cup_{S_{1}}W_{1})\cup_{F_{1}}...\cup_{F_{n-1}}(V_{n}\cup_{S_{n}}W_{n}), for n2n\geq 2, so that (1) FiF_{i} is essential in E(K)E(K^{\star}), for any 1in11\leq i\leq n-1; (2) VjSjWjV_{j}\cup_{S_{j}}W_{j} is a strongly irreducible Heegaard splitting, for any 1jn1\leq j\leq n, see Figure 7.

Refer to caption
Figure 7. Untelescoping of VSWV^{\star}\cup_{S^{\star}}W^{\star}

Since

g(S)=i=1ng(Si)j=1n1g(Fj)=i=1n1(g(Si)g(Fi))+g(Sn)g(S^{\star})=\sum\limits_{i=1}^{n}g(S_{i})-\sum\limits_{j=1}^{n-1}g(F_{j})=\sum\limits_{i=1}^{n-1}(g(S_{i})-g(F_{i}))+g(S_{n})
=i=1k1(g(Si)g(Fi))+g(Sk)+j=k+1n(g(Sj)g(Fj1))(2kn1)=\sum\limits_{i=1}^{k-1}(g(S_{i})-g(F_{i}))+g(S_{k})+\sum\limits_{j=k+1}^{n}(g(S_{j})-g(F_{j-1}))~{}(2\leq k\leq n-1)
=g(S1)+j=2n(g(Sj)g(Fj1))=g(S_{1})+\sum\limits_{j=2}^{n}(g(S_{j})-g(F_{j-1}))

and

g(Si)max{g(Fi),g(Fi1)}(2in1),g(S_{i})\geq max~{}\{g(F_{i}),~{}g(F_{i-1})\}~{}(2\leq i\leq n-1),
g(S1)g(F1),g(S_{1})\geq g(F_{1}),
g(Sn)g(Fn1),g(S_{n})\geq g(F_{n-1}),

we have

g(S)g(Sk)(1kn).g(S^{\star})\geq g(S_{k})~{}(1\leq k\leq n).

If for some 1in11\leq i\leq n-1, FiF_{i} intersects T2T^{2} nontrivially up to isotopy, then each component of FiE(K)F_{i}\cap E(K) is essential in E(K)E(K). Let F1,iFiE(K)F_{1,i}\subset F_{i}\cap E(K) be an essential surface in E(K)E(K). Then by Lemma 2.1, d𝒞(S)(V,W)2χ(F1,i)2χ(Fi)=2g(Fi)2g(Si)2g(S)2t(K)+4d_{\mathcal{C}(S)}(V,W)\leq 2-\chi(F_{1,i})\leq 2-\chi(F_{i})=2g(F_{i})\leq 2g(S_{i})\leq 2g(S^{\star})\leq 2t(K)+4. It contradicts the assumption that d𝒞(S)(V,W)2t(K)+5d_{\mathcal{C}(S)}(V,W)\geq 2t(K)+5. So for any 1in11\leq i\leq n-1, FiF_{i} is disjoint from T2T^{2}. By the similar argument in the proof of Lemma 4.1, for any 1jn1\leq j\leq n, SjS_{j} is disjoint from T2T^{2}. Hence T2T^{2} is disjoint from (i=1n1Fi)(j=1nSj)(\mathop{\cup}\limits_{i=1}^{n-1}F_{i})\cup(\mathop{\cup}\limits_{j=1}^{n}S_{j}).

Then T2T^{2} lies in some ViV_{i} or WiW_{i}, for some 1in1\leq i\leq n. Without loss of generality, we assume that T2T^{2} lies in ViV_{i}. Since T2T^{2} is essential in E(K)E(K^{\star}), i2i\geq 2. It is known that there is no essential closed surface in a compression body or handlebody. So T2T^{2} is isotopic to Fi1F_{i-1}. Therefore VSWV^{\star}\cup_{S^{\star}}W^{\star} is an amalgamation of a Heegaard splitting of E(K)E(K) and a Heegaard splitting of CC along T2T^{2}.

Fact 4.1.

g(C)=2g(C)=2.

Proof.

Since p2p\geq 2, KK^{\star} runs around the longitude of η(K)\eta(K) at least twice. So CC is not a torus II-bundle. On one hand, C\partial C consists of two tori. Then g(C)2g(C)\geq 2. On the other hand, there is a genus two Heegaard splitting of C=(T2×I)Aη(K)C=(T^{2}\times I)\cup_{A}\eta(K). Let J=η(K)(η(K)×I)¯J=\overline{\eta(K)-(\partial\eta(K)\times I)} and bb be a fiber arc in C(T2×[0,12])J¯\overline{C-(T^{2}\times[0,\frac{1}{2}])-J} with one endpoint in T2×{12}T^{2}\times\{\frac{1}{2}\} and the other in J\partial J and bint(A)b\cap int(A)\neq\emptyset. Then both V1=(T2×[0,12])η(b)JV_{1}=(T^{2}\times[0,\frac{1}{2}])\cup\eta(b)\cup J and W1=CV1¯W_{1}=\overline{C-V_{1}} are genus two compression bodies with +V1=+W1\partial_{+}V_{1}=\partial_{+}W_{1}, see Figure 8. So V1W1V_{1}\cup W_{1} is a genus two Heegaard splitting of CC and g(C)2g(C)\leq 2. Hence g(C)=2g(C)=2. ∎

Refer to caption
Figure 8. A genus two Heegaard splitting of CC

Therefore, g(K)=g(S)g(K)+g(C)g(T2)=g(K)+1g(K^{\star})=g(S^{\star})\geq g(K)+g(C)-g(T^{2})=g(K)+1. And t(K)t(K)+1t(K^{\star})\geq t(K)+1. However, t(K)t(K)+1t(K^{\star})\leq t(K)+1. Hence t(K)=t(K)+1t(K^{\star})=t(K)+1.

4.2. Proof of Proposition 4.1 (2)

Recall that InSInS is the collection of isotopy classes of boundary slopes of properly embedded essential surfaces in E(K)E(K). Then by Lemma 2.2, InSInS contains finitely many vertices in 𝒞(T2)\mathcal{C}(T^{2}) depending on KK. So there is a constant 𝒩1\mathcal{N}_{1} depending on KK so that diam𝒞(T2)(InS)𝒩1diam_{\mathcal{C}(T^{2})}(InS)\leq\mathcal{N}_{1}.

Claim 4.2.

E(K)E(K) is not a twisted II-bundle over a compact non-orientable surface.

Proof.

Suppose the conclusion is false. Then E(K)E(K) is a twisted II-bundle over a compact non-orientable surface FF. Then χ(E(K))=2χ(F)\chi(\partial E(K))=2\chi(F). Since E(K)\partial E(K) is a torus, χ(E(K))=2χ(F)=0\chi(\partial E(K))=2\chi(F)=0. So FF is either a Mobius band or Klein bottle. We say that FF is not a Mobius band. For if not, then E(K)E(K) is the twisted I-bundle of a Mobius band, i.e., a solid torus. So KK is a trivial knot in S3S^{3}. A contradiction. So FF is a Klein bottle. Let F~\widetilde{F} be a double cover of FF. Then F~\widetilde{F} is a torus and F~×I\widetilde{F}\times I, i.e., a torus I-bundle, is a double covering of E(K)E(K). For any slope rE(K)r\subset\partial E(K), F~×I(r,r)\widetilde{F}\times I(r,r) is a double covering of E(K)(r)E(K)(r). However, when rr is the meridian of KK, F~×I(m,m)\widetilde{F}\times I(m,m) is a Lens space or S3S^{3} while E(K)(m)E(K)(m) is S3S^{3}. It means that S3S^{3} is not simple connected. A contradiction. ∎

Let αInS\alpha\in InS be a boundary slope of an essential surface QQ properly embedded in E(K)E(K) such that the genus of QQ is minimal. Let g=max{g(Q),t(K)+2}g=max~{}\{g(Q),t(K)+2\}. By Claim 4.2, E(K)E(K) is not a twisted II-bundle over a non-orientable surface. Since KK is nontrivial, E(K)\partial E(K) is incompressible. So E(K)E(K) is not a product II-bundle over an orientable compact surface. By Lemma 2.4, for any properly embedded, strongly irreducible and \partial-strongly irreducible, genus at most gg surface PP in E(K)E(K), there is a number KK^{{}^{\prime}} depending only on gg so that d𝒞(T2)(P,α)Kd_{\mathcal{C}(T^{2})}(\partial P,\alpha)\leq K^{{}^{\prime}}. Let 𝒩=K+𝒩1\mathcal{N}=K^{{}^{\prime}}+\mathcal{N}_{1}.

Let VSWV^{\star}\cup_{S^{\star}}W^{\star} be a minimal genus Heegaard splitting of E(K)E(K^{\star}) with V=E(K)\partial_{-}V^{\star}=\partial E(K^{\star}). Since E(K)E(K^{\star}) is irreducible, VSWV^{\star}\cup_{S^{\star}}W^{\star} is either strongly irreducible or weakly reducible and irreducible.

Lemma 4.2.

VSWV^{\star}\cup_{S^{\star}}W^{\star} is weakly reducible and irreducible.

Proof.

Suppose that the conclusion is false. Then VSWV^{\star}\cup_{S^{\star}}W^{\star} is strongly irreducible. Since T2T^{2} is essential in E(K)E(K^{\star}), T2T^{2} intersects SS^{\star} nontrivially up to isotopy. By Claim 4.1, there is at most one strongly irreducible component in SE(K)S^{\star}\cap E(K) while others are essential in E(K)E(K). Moreover, SE(K)\partial S^{\star}\cap E(K) are isotopic to the slope p/qp/q in T2T^{2}.

If SE(K)S^{\star}\cap E(K) contains an essential surface in E(K)E(K), then p/qInSp/q\in InS and diam𝒞(T2)(p/q,InS)𝒩1𝒩diam_{\mathcal{C}(T^{2})}(p/q,InS)\leq\mathcal{N}_{1}\leq\mathcal{N}. A contradiction. Otherwise, SE(K)S^{\star}\cap E(K) contains only one strongly irreducible component S1S_{1}. By Lemma 2.3, either S1S_{1} is strongly irreducible, \partial-irreducible or there is an incompressible and boundary incompressible surface FF so that d𝒞(T2)(S1,F)1.d_{\mathcal{C}(T^{2})}(\partial S_{1},\partial F)\leq 1. For the first case, g(S1)g(SE(K))g(S)=g(K)=t(K)+1t(K)+2gg(S_{1})\leq g(S^{\star}\cap E(K))\leq g(S^{\star})=g(K^{\star})=t(K^{\star})+1\leq t(K)+2\leq g. Then d𝒞(T2)(S1,α)Kd_{\mathcal{C}(T^{2})}(\partial S_{1},\alpha)\leq K^{{}^{\prime}}, i.e., d𝒞(T2)(p/q,α)Kd_{\mathcal{C}(T^{2})}(p/q,\alpha)\leq K^{{}^{\prime}}. Hence diam𝒞(T2)(p/q,InS)K+𝒩1=𝒩diam_{\mathcal{C}(T^{2})}(p/q,InS)\leq K^{{}^{\prime}}+\mathcal{N}_{1}=\mathcal{N}. A contradiction. For the second case, diam𝒞(T2)(p/q,InS)𝒩1+1𝒩diam_{\mathcal{C}(T^{2})}(p/q,InS)\leq\mathcal{N}_{1}+1\leq\mathcal{N}. A contradiction. ∎

So VSWV^{\star}\cup_{S^{\star}}W^{\star} is weakly reducible and irreducible. By the main result in [20], VSWV^{\star}\cup_{S^{\star}}W^{\star} has an untelescoping (V1S1W1)F1Fn1(VnSnWn)(V_{1}\cup_{S_{1}}W_{1})\cup_{F_{1}}...\cup_{F_{n-1}}(V_{n}\cup_{S_{n}}W_{n}), n2n\geq 2 so that (1) for any 1in11\leq i\leq n-1, FiF_{i} is incompressible in E(K)E(K^{\star}) and not parallel to E(K)\partial E(K^{\star}); (2) for any 1in1\leq i\leq n, ViSiWiV_{i}\cup_{S_{i}}W_{i} is a strongly irreducible Heegaard splitting. We say that T2T^{2} is disjoint from (i=1n1Fi)(j=1nSj)(\mathop{\cup}\limits_{i=1}^{n-1}F_{i})\cup(\mathop{\cup}\limits_{j=1}^{n}S_{j}). For if not, (1) either FiT2F_{i}\cap T^{2}\neq\emptyset up to isotopy for some 1in11\leq i\leq n-1. Then each component of FiE(K)F_{i}\cap E(K) is essential in E(K)E(K). Note that FiE(K)\partial F_{i}\cap E(K) are isotopic to p/qp/q in T2T^{2}. Then diam𝒞(T2)(p/q,InS)𝒩1𝒩diam_{\mathcal{C}(T^{2})}(p/q,InS)\leq\mathcal{N}_{1}\leq\mathcal{N}. A contradiction. Or, (2) SiT2S_{i}\cap T^{2}\neq\emptyset. By the similar argument in the proof of Lemma 4.2, diam𝒞(T2)(p/q,InS)𝒩1𝒩diam_{\mathcal{C}(T^{2})}(p/q,InS)\leq\mathcal{N}_{1}\leq\mathcal{N}. A contradiction.

Therefore VSWV^{\star}\cup_{S^{\star}}W^{\star} is an amalgamation of a Heegaard splitting of E(K)E(K) and a Heegaard splitting of CC along T2T^{2}. So g(K)=g(S)g(K)+g(C)g(T2)=g(K)+1g(K^{\star})=g(S^{\star})\geq g(K)+g(C)-g(T^{2})=g(K)+1 and t(K)t(K)+1t(K^{\star})\geq t(K)+1. However, it is known that t(K)t(K)+1t(K^{\star})\leq t(K)+1. So t(K)=t(K)+1t(K^{\star})=t(K)+1.

5. The proof of Corollary 1.1

Let KK and KK^{\star} be the same as in Theorem 1.2 (2). For any unstabilized Heegaard splitting VSWV^{\circ}\cup_{S^{\circ}}W^{\circ} of E(K)E(K^{\star}), since E(K)E(K^{\star}) is irreducible, it is either strongly irreducible or weakly reducible and irreducible. So we divide it into two cases: (1) VSWV^{\circ}\cup_{S^{\circ}}W^{\circ} is strongly irreducible; (2) VSWV^{\circ}\cup_{S^{\circ}}W^{\circ} is weakly reducible and irreducible. We firstly prove Corollary 1.1 for the strongly irreducible case.

Since AA is an essential annulus in E(K)E(K^{\star}), by Schultens lemma [22], each curve of SAS^{\circ}\cap A is essential in both SS^{\circ} and AA. Then by the same argument in the proof of Claim 3.1, one subsurface of SA¯\overline{S^{\circ}-A} is strongly irreducible while the others are incompressible in their corresponding components of E(K)AE(K^{\star})-A. Moreover, each subsurface of SA¯\overline{S^{\circ}-A} has p/qp/q slopes as its boundary curves. By the condition that

diam𝒞(T2)(p/q,InS)>𝒩,diam_{\mathcal{C}(T^{2})}(p/q,InS)>\mathcal{N},

SE(K)S^{\circ}\cap E(K) is connected and strongly irreducible while each component of SJS^{\circ}\cap J is essential and an annulus. So SJS^{\circ}\cap J is a collection of nested annuli in JJ. By the similar argument of Case 1 of Theorem 1.1 in Section 3, we do a Dehn surgery on VSWV^{\circ}\cup_{S^{\circ}}W^{\circ} and obtain a Heegaard splitting VSWV\cup_{S}W of E(K)E(K). In reverse, VSWV^{\circ}\cup_{S^{\circ}}W^{\circ} is also a Dehn surgery of VSWV\cup_{S}W.

Otherwise, VSWV^{\circ}\cup_{S^{\circ}}W^{\circ} is weakly reducible and irreducible. Then it has an untelescoping, says VSW=(V1S1W1)F1Fn1(VnSnWn)V^{\circ}\cup_{S^{\circ}}W^{\circ}=(V_{1}\cup_{S_{1}}W_{1})\cup_{F_{1}}...\cup_{F_{n-1}}(V_{n}\cup_{S_{n}}W_{n}), for n2n\geq 2, where Fi(1in1)F_{i}~{}(1\leq i\leq n-1) is incompressible in E(K)E(K^{\star}). By the same argument as above, AA is disjoint from i=1n1Fi\mathop{\cup}\limits_{i=1}^{n-1}F_{i}. So AA is contained in ViSiWiV_{i}\cup_{S_{i}}W_{i}, for some 1in1\leq i\leq n. Then by the same argument as the strongly irreducible case, we do a Dehn surgery on ViSiWiV_{i}\cup_{S_{i}}W_{i} so that it is changed into V1,iS1,iW1,iV_{1,i}\cup_{S_{1,i}}W_{1,i}. Moreover, (V1S1W1)F1Fi1(V1,iS1,iW1,i)FiFn1(VnSnWn)(V_{1}\cup_{S_{1}}W_{1})\cup_{F_{1}}...\cup_{F_{i-1}}(V_{1,i}\cup_{S_{1,i}}W_{1,i})\cup_{F_{i}}...\cup_{F_{n-1}}(V_{n}\cup_{S_{n}}W_{n}) is a Heegaard splitting of E(K)E(K). In reverse, VSW=(V1S1W1)F1Fn1(VnSnWn)V^{\circ}\cup_{S^{\circ}}W^{\circ}=(V_{1}\cup_{S_{1}}W_{1})\cup_{F_{1}}...\cup_{F_{n-1}}(V_{n}\cup_{S_{n}}W_{n}) is also a Dehn surgery of (V1S1W1)F1Fi1(V1,iS1,iW1,i)FiFn1(VnSnWn)(V_{1}\cup_{S_{1}}W_{1})\cup_{F_{1}}...\cup_{F_{i-1}}(V_{1,i}\cup_{S_{1,i}}W_{1,i})\cup_{F_{i}}...\cup_{F_{n-1}}(V_{n}\cup_{S_{n}}W_{n}).

References

  • [1] Bachman, D., Schleimer, S. and Sedgwick, E., Sweepouts of amalgamated 3-manifolds, Algebr. Geom. Topol., 6, 2006, 171-194.
  • [2] Casson, A. J. and Gordon C. McA., Reducing Heegaard splittings, Topology Appl. 27 (1987), no. 3, 275-283.
  • [3] Gao X. T., Guo Q. L. and Qiu R. F., A note on tunnel number of composite knots, Topology Appl. 158 (2011), no. 16, 2240-2243.
  • [4] Hartshorn K., Heegaard splittings of Haken manifolds have bounded distance, Pacific J. Math. 204 (2002), no. 1, 61-75.
  • [5] Harvey W. J., Boundary structure of the modular group. Riemann surfaces and related topics: Proceedings of the 1978 Stony Brook Conference (State Univ. New York, Stony Brook, N.Y., 1978), pp. 245-251, Ann. of Math. Stud., 97, Princeton Univ. Press, Princeton, N.J., 1981.
  • [6] Hass J., Rubinstein J. H. and Wang S. C., Boundary slopes of immersed surfaces in 3-manifolds, J. Differential Geom. 52 (1999), no. 2, 303-325.
  • [7] Hass J., Wang S. C. and Zhou Q., On finiteness of the number of boundary slopes of immersed surfaces in 3-manifolds, Proc. Amer. Math. Soc. 130 (2002), no. 6, 1851-1857.
  • [8] Hatcher A. E., On the boundary curves of incompressible surfaces, Pacific J. Math. 99 (1982), no. 2, 373-377.
  • [9] Hempel J., 3-manifolds as viewed from the curve complex, Topology 40 (2001), no. 3, 631-657.
  • [10] Li T., Saddle tangencies and the distance of Heegaard splittings, Algebr. Geom. Topol. 7 (2007), 1119-1134.
  • [11] ——, Heegaard surfaces and the distance of amalgamation, Geom. Topol. 14 (2010), no. 4, 1871-1919.
  • [12] Li T., Qiu R. F., On the degeneration of tunnel numbers under a connected sum, Trans. Amer. Math. Soc. 368 (2016), no. 4, 2793-2807.
  • [13] Li T., Qiu R. F. and Wang S. C., A quadratic bound on the number of boundary slopes of essential surfaces with bounded genus, Geom. Dedicata 147 (2010), 131-138.
  • [14] Masur H. and Minsky Y.N., Geometry of the complex of curves. I. Hyperbolicity, Invent. Math. 138 (1999), no. 1, 103-149.
  • [15] Minsky Y. N., Moriah Y., Schleimer S., High distance knots, Algebr. Geom. Topol. 7 (2007), 1471-1483.
  • [16] Moriah Y., A note on satellites and tunnel number, Kobe J. Math. 8 (1991), no. 1, 73-79.
  • [17] Morimoto K., On the super additivity of tunnel number of knots, Math. Ann. 317 (2000), no. 3, 489-508.
  • [18] Scharlemann M., Local detection of strongly irreducible Heegaard splittings, Topology Appl. 90 (1998), 135-147.
  • [19] ——, Proximity in the curve complex: boundary reduction and bicompressible surfaces, Pacific J. Math. 228 (2006), no. 2, 325-348.
  • [20] Scharlemann M. and Thompson A., Thin position for 3-manifolds, Geometric topology (Haifa, 1992), 231-238, Contemp. Math., 164, Amer. Math. Soc., Providence, RI, 1994.
  • [21] Scharlemann M. and Tomova M., Alternate Heegaard genus bounds distance, Geom. Topol. 10 (2006), 593-617.
  • [22] Schultens J., Additivity of tunnel number for small knots, Comment. Math. Helv. 75 (2000), no. 3, 353-367.