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

Primitive, proper power, and Seifert curves in the boundary of a genus two handlebody

Sungmo Kang [email protected]
Abstract.

A simple closed curve α\alpha in the boundary of a genus two handlebody HH is primitive if adding a 2-handle to HH along α\alpha yields a solid torus. If adding a 2-handle to HH along α\alpha yields a Seifert-fibered space and not a solid torus, the curve is called Seifert. If α\alpha is disjoint from an essential separating disk in HH, does not bound a disk in HH, and is not primitive in HH, then it is said to be proper power.

As one of the background papers of the classification project of hyperbolic primitive/Seifert knots in S3S^{3} whose complete list is given in [BK20], this paper classifies in terms of R-R diagrams primitive, proper power, and Seifert curves. In other words, we provide up to equivalence all possible R-R diagrams of such curves. Furthermore, we further classify all possible R-R diagrams of proper power curves with respect to an arbitrary complete set of cutting disks of a genus two handlebody.

1. Introduction and main results

In this paper, we provide the classifications of three types of simple closed curves lying in the boundary of a genus two handlebody: primitive, proper power, and Seifert curves. These classifications will be used in the classification project of hyperbolic primitive/Seifert knots in S3S^{3} whose complete list is given in [BK20].

Primitive/Seifert(or simply P/SF) knots, which were introduced in [D03], are a natural generalization of primtive/primitive(or simply P/P) knots defined by Berge in [B90] or an available version [B18]. Both P/P knots and P/SF knots are represented by simple closed curves lying in a genus two Heegaard surface of S3S^{3} bounding two handlebodies such that 2-handle additions to the handlebodies along the curves are solid tori for P/P knots and one 2-handle addition is a solid torus and the other is a Seifert-fibered space and not a solid torus for P/SF knots. One component of the intersection of a regular neighborhood of a knot and the Heegaard surface defines a so-called surface-slope.

Berge constructed twelve families of P/P knots which are referred to as the Berge knots. The Berge knots admit lens space Dehn surgeries at surface-slopes. The Berge conjecture, which is still unsolved, says that if a knot in S3S^{3} admits a lens space Dehn surgery, then it is a Berge knot and the surgery is the corresponding surface-slope surgery. Therefore the conjecture implies the complete classification of knots admitting lens space Dehn surgeries. Toward the Berge conjecture, it is proved in [B08] or independently in [G13] that all P/P knots are the Berge knots. This implies that the Berge knots are the complete list of P/P knots.

Meanwhile, P/SF knot are also of interest, because P/SF knots admit Seifert-fibered Dehn surgeries at surface-surface slopes and knots with Dehn surgeries yielding Seifert-fibered spaces are not well understood. The classification project of hyperbolic primitive/Seifert knots in S3S^{3} has been carried out for years and has recently been completed. The complete list of hyperbolic primitive/Seifert knots in S3S^{3} is given in [BK20] where the surface-slope of the exceptional surgery on each P/SF knot that yields a Seifert-fibered space and the indexes of each exceptional fiber in the resulting Seifert-fibered space are also provided.

Now we describe the results of this paper, which is the classifications of primitive, proper power, and Seifert curves in the boundary of a genus two handlebody. The definitions of such curves are as follows.

Definition 1.1.

Let HH be a genus two handlebody, α\alpha an essential simple closed curve in H\partial H, and H[α]H[\alpha] the 3-manifold obtained by adding a 2-handle to HH along α\alpha.

  1. (1)

    α\alpha is said to be primitive if H[α]H[\alpha] is a solid torus.

  2. (2)

    α\alpha is said to be proper power if α\alpha is disjoint from an essential separating disk in HH, does not bound a disk in HH, and is not primitive in HH.

  3. (3)

    α\alpha is said to be Seifert if H[α]H[\alpha] is a Seifert-fibered space and not a solid torus.

There are subtypes of Seifert curves in HH. Since HH is a genus two handlebody, that α\alpha is Seifert in HH implies that H[α]H[\alpha] is an orientable Seifert-fibered space over D2D^{2} with two exceptional fibers, or an orientable Seifert-fibered space over the Möbius band with at most one exceptional fiber. Therefore, we further divide Seifert curves into two subtypes. If H[α]H[\alpha] is Seifert-fibered over D2D^{2}, we say that α\alpha is Seifert-d. If H[α]H[\alpha] is Seifert-fibered over the Möbius band, we say that α\alpha is Seifert-m.

The following theorems present the classifications of primitive, proper power, and Seifert curves. They are described in terms of R-R diagrams. For the definition and properties of R-R diagrams, see [K20].

Theorem 1.2.

Suppose α\alpha is a simple closed curve in the boundary of a genus two handlebody HH.

  1. (1)

    If α\alpha is a primitive curve, then α\alpha has an R-R diagram of the form shown in Figure 1.

  2. (2)

    If α\alpha is a Seifert-d curve, then α\alpha has an R-R diagram of the form shown in Figure 2 with n,s>1n,s>1, a,b>0a,b>0, and gcd(a,b)=1\gcd(a,b)=1.

  3. (3)

    If α\alpha is a Seifert-m curve, then α\alpha has an R-R diagram of the form shown in Figure 3 with s>1s>1.

  4. (4)

    If α\alpha is a proper power curve, then α\alpha has an R-R diagram of the form shown in Figure 4 with s>1s>1.

Regarding Seifert curves, if α\alpha has an R-R diagram of the form shown in Figure 2a (b, respectively), then H[α]H[\alpha] is a Seifert-fibered space over D2D^{2} with two exceptional fibers of indexes nn and ss (n(a+b)+bn(a+b)+b and ss, respectively). If α\alpha has an R-R diagram of the form shown in Figure 3, then H[α]H[\alpha] is a Seifert-fibered space over the Möbius band with one exceptional fiber of index ss.

Refer to caption
Figure 1. If α\alpha is a primitive curve in the boundary of a genus two handlebody HH, then α\alpha has an R-R diagram with the form of this figure.
Refer to caption
Figure 2. If α\alpha is a Seifert-d curve in the boundary of a genus two handlebody HH, then α\alpha has an R-R diagram with the form of one of these figures with n,s>1n,s>1, a,b>0a,b>0, and gcd(a,b)=1\gcd(a,b)=1. If α\alpha has an R-R diagram of the form shown in Figure 2a (2b, respectively), then H[α]H[\alpha] is a Seifert-fibered space over D2D^{2} with two exceptional fibers of indexes nn and ss (n(a+b)+bn(a+b)+b and ss, respectively).
Refer to caption
Figure 3. If α\alpha is a Seifert-m curve in the boundary of a genus two handlebody HH, then α\alpha has an R-R diagram with the form of this figure with s>1s>1 in which α=ABsA1Bs\alpha=AB^{s}A^{-1}B^{s} in π1(H)\pi_{1}(H) and H[α]H[\alpha] is a Seifert-fibered space over the Möbius band with one exceptional fiber of index ss.

We can further classify proper power curves in the following theorem. From now on, to distinguish proper power curves from primitive or Seifert curves, we use the letter β\beta instead of α\alpha to represent proper power curves. We will see such a situation in Sections 4 and 5.

Theorem 1.3 (Further classification of proper power curves).

Suppose HH is a genus two handlebody with a complete set of cutting disks {DA,DB}\{D_{A},D_{B}\} with π1(H)=F(A,B)\pi_{1}(H)=F(A,B), where the generators AA and BB are dual to DAD_{A} and DBD_{B} respectively. If β\beta is a proper power curve in HH, then β\beta has one of the following R-R diagrams with respect to the complete set of cutting disks {DA,DB}\{D_{A},D_{B}\} up to the homeomorphisms of HH inducing the automorphisms exchanging AA and BB, and replacing A1A^{-1} by AA or B1B^{-1} by BB:

  1. (1)

    Type I: β\beta has an R-R diagram with a 0-connection in at least one of the handles.

  2. (2)

    Type II: β\beta has an R-R diagram of the form shown in Figure 4 with s>1s>1.

  3. (3)

    Type III: β\beta has an R-R diagram of the form shown in Figure 5 with a,b>0a,b>0 and s>0s>0.

  4. (4)

    Type IV: β\beta has an R-R diagram of the form shown in Figure 6 with a,b,c>0a,b,c>0.

  5. (5)

    Type V: β\beta has an R-R diagram of the form shown in Figure 7 with a,b,c,d>0a,b,c,d>0.

Refer to caption
Figure 4. Proper power curves referred as α\alpha in Theorem 1.2 and Type II of proper power curves β\beta: [β]=Bs[\beta]=B^{s} in Theorem 1.3.
Refer to caption
Figure 5. Type III of proper power curves β\beta: [β]=(ABs)a+b[\beta]=(AB^{s})^{a+b}, where s>0s>0 and a,b>0a,b>0.
Refer to caption
Figure 6. Type IV of proper power curves β\beta: [β]=(AB)a+b+c[\beta]=(AB)^{a+b+c}, where a,b,c>0a,b,c>0.
Refer to caption
Figure 7. Type V of proper power curves β\beta, where a,b,c,d>0a,b,c,d>0.

The classifications of such curves play very important role in the classification of all hyperbolic primitive/Seifert(or simply P/SF) knots in S3S^{3} whose complete list is given in [BK20]. The classifications of primitive and Seifert curves are the first step in the classification of hyperbolic P/SF knots in S3S^{3}. Additionally if α\alpha is Seifert in HH such that H[α]H[\alpha] embeds in S3S^{3}, then H[α]H[\alpha] is homeomorphic to the exterior of a torus knot. Therefore the classification of Seifert curves in HH naturally carries that of all simple closed curves α\alpha such that H[α]H[\alpha] is homeomorphic to the exterior of a torus knot.

The classification of proper power curves has various applications. First it is used to determine if P/P(primitive/primitive) or P/SF knots are hyperbolic or not. It turns out that if P/P or P/SF knots are not hyperbolic, then there exists a proper power curve in some circumstances. Another application is that when a curve α\alpha is Seifert in HH, i.e., H[α]H[\alpha] is an orientable Seifert-fibered space, a proper power curve disjoint from α\alpha becomes a regular fiber of the Seifert-fibered space H[α]H[\alpha] and can be used to compute the indexes of exceptional fibers. Also in order to classify some type of primitive/Seifert knots, called knots in Once-Punctured Tori(or simply OPT), properties of proper power curves are used.

Acknowledgement. In 2008, in a week-long series of talks to a seminar in the department of mathematics of the University of Texas as Austin, John Berge outlined a project to completely classify and describe the primitive/Seifert knots in S3S^{3}. The present paper, which provides some of the background materials necessary to carry out the project, is originated from the joint work with John Berge for the project. I should like to express my gratitude to John Berge for his support and collaboration. I would also like to thank Cameron Gordon and John Luecke for their support while I stayed in the University of Texas at Austin.

2. R-R diagrams of primitive curves

In this section, we classify R-R diagrams of primitive curves in the boundary of a genus two handlebody. Let α\alpha be a primitive curve in a genus two handlebody HH. In other words, H[α]H[\alpha] is a solid torus. In order to obtain R-R diagrams of α\alpha we use the following lemma which shows equivalent conditions of primitivity. The proof may be found in [W36], [Z70], or [G87].

Lemma 2.1.

The following are equivalent:

  • (1)

    α\alpha is primitive in HH, i.e., H[α]H[\alpha] is a solid torus;

  • (2)

    α\alpha belongs to a basis for the free group π1(H)\pi_{1}(H);

  • (3)

    α\alpha is transverse to a properly embedded disk in HH.

Theorem 2.2.

If α\alpha is a nonseparating simple closed curve in the boundary of a genus two handlebody HH such that α\alpha is primitive on HH, then α\alpha has an R-R diagram with the form of Figure 1.

Proof.

By Lemma 2.1, there exists a cutting disk DAD_{A} in HH such that α\alpha meets DA\partial D_{A} transversely in a single point. Consider the regular neighborhood NN of DAα\partial D_{A}\cup\alpha in H\partial H. Then NN is a once-punctured torus which contains DA\partial D_{A}, and whose boundary N\partial N bounds a separating disk of HH. Let N=HN¯N^{\prime}=\overline{\partial H-N}. Then it follows by cutting HH open along DAD_{A} that there exists a unique cutting disk DBD_{B} of HH up to isotopy whose boundary DB\partial D_{B} lies on NN^{\prime}.

This partition {N,N}\{N,N^{\prime}\} of H\partial H with DAN\partial D_{A}\subset N and DBN\partial D_{B}\subset N^{\prime} gives rise to an R-R diagram of α\alpha in which NN and NN^{\prime} correspond to the AA-handle and BB-handle respectively. Since α\alpha lies in NN and intersects DA\partial D_{A} in a single point, α\alpha has an R-R diagram of the form shown in Figure 1. This completes the proof. ∎

3. R-R diagrams of a proper power curve and more classifications

In this section, we classify proper power curves. Let β\beta be a proper power curve in the boundary of a genus two handlebody HH. The following lemma is an easy consequence of Lemma 2.1.

Lemma 3.1.

The following are equivalent:

  • (1)

    β\beta is a proper power curve in HH, i.e., β\beta is disjoint from an essential separating disk, does not bound a disk, and is not primitive in HH;

  • (2)

    β\beta is conjugate to wn,n>1,w^{n},\hskip 4.26773ptn>1, of π1(H)\pi_{1}(H), where ww is a free generator of π1(H)\pi_{1}(H);

  • (3)

    There exists a complete set of cutting disks {DA,DB}\{D_{A},D_{B}\} of HH such that β\beta is disjoint to, say, DAD_{A} and is transverse to DBD_{B} n>1n>1 times.

The main results of this section are the following.

Theorem 3.2.

Suppose β\beta is a simple closed curve in the boundary of a genus two handlebody HH. If β\beta is a proper power curve in HH, then β\beta has an R-R diagram of the form shown in Figure 4.

Theorem 3.3.

Suppose HH is a genus two handlebody with a complete set of cutting disks {DA,DB}\{D_{A},D_{B}\} with π1(H)=F(A,B)\pi_{1}(H)=F(A,B), where the generators AA and BB are dual to DAD_{A} and DBD_{B} respectively. If β\beta is a proper power curve in HH, then β\beta has one of the following R-R diagrams with respect to the complete set of cutting disks {DA,DB}\{D_{A},D_{B}\} up to the homeomorphisms of HH inducing the automorphisms exchanging AA and BB, and replacing A1A^{-1} by AA or B1B^{-1} by BB:

  1. (1)

    Type I: β\beta has an R-R diagram with a 0-connection in at least one of the handles.

  2. (2)

    Type II: β\beta has an R-R diagram of the form shown in Figure 4 with s>1s>1..

  3. (3)

    Type III: β\beta has an R-R diagram of the form shown in Figure 5 with a,b>0a,b>0 and s>0s>0.

  4. (4)

    Type IV: β\beta has an R-R diagram of the form shown in Figure 6 with a,b,c>0a,b,c>0.

  5. (5)

    Type V: β\beta has an R-R diagram of the form shown in Figure 7 with a,b,c,d>0a,b,c,d>0.

The proof of Theorem 3.2 follows immediately from the definition of a proper power curve. In order to prove Theorem 3.3, we need the following minor generalization of a result of Cohen, Metzler, and Zimmerman [CMZ81] which allows one to determine easily if a given cyclically reduced word in a free group of rank two is a primitive or a proper power of a primitive.

Theorem 3.4.

[CMZ81] Suppose a cyclic conjugate of

W=An1Bm1AnlBmlW=A^{n_{1}}B^{m_{1}}\dots A^{n_{l}}B^{m_{l}}

is a member of a basis of F(A,B)F(A,B) or a proper power of a member of a basis of F(A,B)F(A,B), where l1l\geq 1 and each indicated exponent is nonzero. Then, after perhaps replacing AA by A1A^{-1} or BB by B1B^{-1}, there exists e>0e>0 such that:

n1==nl=1,and{m1,,ml}{e,e+1},n_{1}=\dots=n_{l}=1,\quad\text{and}\quad\{m_{1},\dots,m_{l}\}\subseteq\{e,e+1\},

or

{n1,,nl}{e,e+1},andm1==ml=1.\{n_{1},\dots,n_{l}\}\subseteq\{e,e+1\},\quad\text{and}\quad m_{1}=\dots=m_{l}=1.
The proof of Theorem 3.3.

In order to find all possible R-R diagrams of a proper power curve β\beta, we consider the following cases separately.

  1. (1)

    β\beta has a 0-connection in at least one of the handles in its R-R diagram.

  2. (2)

    β\beta has no 0-connections in either handle in its R-R diagram.

The case where β\beta has a 0-connection in one handle gives restriction to other curves in H\partial H disjoint from β\beta. In other words, if γ\gamma is a simple closed curve in H\partial H disjoint from β\beta, then γ\gamma must have only bands of connections labeled by 0 or 11 in that handle. Therefore we put this case into one type of possible proper power curves, which gives Type I in Theorem 1.2.

Now we assume that β\beta has no 0-connections in either handle.

Suppose β\beta has only one generator in π1(H)=F(A,B)\pi_{1}(H)=F(A,B). Then up to replacement of AA with A1A^{-1}, BB with B1B^{-1}, or exchange of AA and BB, we may assume that [β]=Bs[\beta]=B^{s} for some s>1s>1. Since β\beta has no 0-connections, this implies that β\beta has no connections in the AA-handle and only one connection in the BB-handle. Thus this case gives rise to Type II of a proper power curve in Theorem 3.2 with an R-R diagram of the form shown in Figure 4.

Now suppose that β\beta has the two generators AA and BB in π1(H)\pi_{1}(H). By Theorem 3.4, we may assume that β=ABm1ABml\beta=AB^{m_{1}}\cdots AB^{m_{l}}, where {m1,,ml}{s,s+ϵ}\{m_{1},\dots,m_{l}\}\subseteq\{s,s+\epsilon\} with ϵ=±1\epsilon=\pm 1 and min{s,s+ϵ}>0\{s,s+\epsilon\}>0. This implies that every connection in the AA-handle is labeled by 1. There are two cases to consider:

  • (1)

    β\beta has only one band of 1-connections in the AA-handle,

  • (2)

    β\beta has two bands of 1-connections in the AA-handle.

Refer to caption
Figure 8. An R-R diagram of β\beta when there are three bands of connections in the BB-handle.

Case (1): Assume that β\beta has only one band of 1-connections in the AA-handle.

Suppose {m1,,ml}{s,s+ϵ}\{m_{1},\dots,m_{l}\}\subsetneq\{s,s+\epsilon\}. Without loss of generality, we may assume that {m1,,ml}={s}\{m_{1},\dots,m_{l}\}=\{s\}. If s>1s>1, then there must be only one band of ss-connections in the BB-handle, in which case β\beta is a primitive curve with [β]=ABs[\beta]=AB^{s}. Therefore s=1s=1 and there must be two bands of 1-connections in the BB-handle. Then with AA and BB exchanged, this case belongs to Type III of a proper power curve in Theorem 3.2 with an R-R diagram of the form shown in Figure 5.

Suppose {m1,,ml}={s,s+ϵ}\{m_{1},\dots,m_{l}\}=\{s,s+\epsilon\}. Then β\beta has at least two bands of connections in the BB-handle. If β\beta has two bands of connections in the BB-handle, β\beta must have an R-R diagram of the form shown in Figure 9 and by Lemma 3.5, β\beta is not a proper power curve. If β\beta has three bands of connections in the BB-handle, then these three bands of connections should be labeled by 1, 2 and 1 respectively, and β\beta has the R-R diagram of the form shown in Figure 11. By Lemma 3.6, β\beta is not a proper power curve.

Case (2): β\beta has two bands of 1-connections in the AA-handle.

Suppose {m1,,ml}{s,s+ϵ}\{m_{1},\dots,m_{l}\}\subsetneq\{s,s+\epsilon\}. Without loss of generality, we may assume that {m1,,mp}={s}\{m_{1},\dots,m_{p}\}=\{s\}. Then if β\beta has only one band of connections in the BB-handle, then the band must be labeled by ss, in which case Type III of a proper power curve in Theorem 3.2 arises with an R-R diagram of the form shown in Figure 5. If β\beta has two bands of connections in the BB-handle, then ss must be 1 and this case yields Type IV of a proper power curve in Theorem 3.2 with an R-R diagram of the form shown in Figure 6.

Suppose {m1,,ml}={s,s+ϵ}\{m_{1},\dots,m_{l}\}=\{s,s+\epsilon\}. If there are only two bands of connections in the BB-handle, then β\beta must have R-R diagram of the form shown in Figure 12 and by Lemma 3.7, β\beta is not a proper power curve.

If there are three bands of connections in the BB-handle, then β\beta has two types of R-R diagrams as shown in Figure 8. By Lemma 3.8, the R-R diagram of β\beta in Figure 8a cannot be a proper power curve. Therefore we have the R-R diagram of β\beta in Figure 8b, which gives Type V of a proper power curve in Theorem 3.2.

Thus we complete the proof of Theorem 3.2

Lemma 3.5.

Suppose a simple closed curve β\beta has an R-R diagram with the form shown in Figure 9 where s>0,ϵ=±1s>0,\epsilon=\pm 1 with min{s,s+ϵ}>0\{s,s+\epsilon\}>0, and a,b>0a,b>0. Then β\beta is a primitive curve.

Refer to caption
Figure 9. An R-R diagram of β\beta where s>0,ϵ=±1s>0,\epsilon=\pm 1, and a,b>0a,b>0.
Proof.

Note that since β\beta is a simple closed curve, gcd(a,b)=1(a,b)=1. Let b=ρa+ηb=\rho a+\eta, where ρ0\rho\geq 0 and 0η<a0\leq\eta<a (if η=0\eta=0, then ρ>0\rho>0 and a=1a=1). Now we record the curve β\beta algebraically by starting the aa parallel arcs, i.e., the band of width aa, entering into the (s+ϵ)(s+\epsilon)-connection in the BB-handle. It follows that β\beta is the product of two subwords ABs+ϵ(ABs)ρAB^{s+\epsilon}(AB^{s})^{\rho} and ABs+ϵ(ABs)ρ+1AB^{s+\epsilon}(AB^{s})^{\rho+1} with |ABs+ϵ(ABs)ρ|=aη|AB^{s+\epsilon}(AB^{s})^{\rho}|=a-\eta and |ABs+ϵ(ABs)ρ+1|=η|AB^{s+\epsilon}(AB^{s})^{\rho+1}|=\eta. Here, for example, |ABs+ϵ(ABs)ρ||AB^{s+\epsilon}(AB^{s})^{\rho}| denotes the total number of appearances of ABs+ϵ(ABs)ρAB^{s+\epsilon}(AB^{s})^{\rho} in the word of β\beta in π1(H)=F(A,B)\pi_{1}(H)=F(A,B).

There is a change of cutting disks of the handlebody HH, which induces an automorphism of π1(H)\pi_{1}(H) that takes AABsA\mapsto AB^{-s} and leaves BB fixed. Then by this change of cutting disks, ABs+ϵ(ABs)ρAB^{s+\epsilon}(AB^{s})^{\rho} and ABs+ϵ(ABs)ρ+1AB^{s+\epsilon}(AB^{s})^{\rho+1} are sent to Aρ+1BϵA^{\rho+1}B^{\epsilon} and Aρ+2BϵA^{\rho+2}B^{\epsilon} respectively. Therefore the resulting Heegaard diagram of β\beta realizes a new R-R diagram of the form in Figure 10, where the positions of the AA and BB-handles are switched. The new R-R diagram has the same form as that in Figure 9 with less number of arcs. One can continue inductively until one of the labels of parallel arcs is 0, in which case [β]=ABj[\beta]=AB^{j} for some j>0j>0 up to replacement of AA with A1A^{-1}, BB with B1B^{-1}, or exchange of AA and BB. Such a curve β\beta is a primitive curve. ∎

Refer to caption
Figure 10. The R-R diagram after change of cutting disks of the handlebody HH inducing an automorphism of π1(H)\pi_{1}(H) that takes AABsA\mapsto AB^{-s} and leaves BB fixed.
Lemma 3.6.

Suppose a simple closed curve β\beta has an R-R diagram with the form shown in Figure 11 where a,b,c>0a,b,c>0. Then β\beta is not a proper power curve.

Refer to caption
Figure 11. An R-R diagram of β\beta where a,b,c>0a,b,c>0.
Proof.

First, note that gcd(a+b,b+c)=1(a+b,b+c)=1 in order for β\beta to be a simple closed curve. Since ABAB and AB2AB^{2} appear in [β][\beta], by considering {AB,AB2}\{AB,AB^{2}\} as a generating set of F(A,B)F(A,B) and by Theorem 3.4 one of the two must appear with exponent 11. However, from the R-R diagram one can see that if one reads β\beta from the right-hand edge of the band of width aa entering the 1-connection in the BB-handle, then ABAB appears twice consecutively. Therefore AB2AB^{2} must have only exponent 11.

If b=1b=1, then AB2AB^{2} appears only once and thus β\beta is a primitive curve. Suppose b>1b>1. Consider following the band of width bb around the R-R diagram. Since β\beta is a simple closed curve, this band must split into two subbands at the endpoint of the (1)(-1)-connection in the AA-handle. If the band splits into two subbands belonging to the bands of width aa and bb respectively, then by tracing arcs of this band we see that β\beta has two subwords AB2(AB)nAB^{2}(AB)^{n} and AB2(AB)n+eAB^{2}(AB)^{n+e} for some positive integers nn and ee with e>1e>1. It follows from Theorem 3.4 that β\beta cannot be a proper power curve. Similarly for the case where the band splits into two subbands belonging to the bands of width bb and cc respectively. ∎

Lemma 3.7.

Suppose a simple closed curve β\beta has an R-R diagram with the form shown in Figure 12 where s>0,ϵ=±1s>0,\epsilon=\pm 1 with min{s,s+ϵ}>0\{s,s+\epsilon\}>0, and a,b,c>0a,b,c>0. Then β\beta is not a proper power curve.

Refer to caption
Figure 12. An R-R diagram of β\beta where s>0,ϵ=±1s>0,\epsilon=\pm 1, and a,b,c>0a,b,c>0.
Refer to caption
Figure 13. The hybrid diagram of β\beta and change of cutting disks of HH.
Proof.

We use the argument of hybrid diagrams which are introduced in [K20]. Consider the corresponding hybrid diagram as shown in Figure 13a. Then we drag the vertex A- together with the edges meeting the vertex A- over the ss-connection in the BB-handle. This corresponds to the change of the cutting disks of HH inducing an automorphism of π1(H)\pi_{1}(H) which takes AABsA\mapsto AB^{-s} and leaves BB fixed. The resulting hybrid diagram is shown in Figure 13b.

Transforming the hybrid diagram in Figure 13b back into an R-R diagram, there are three possible cases to consider as follows:

  1. (1)

    There is only one band of connections in the AA-handle;

  2. (2)

    There are only two bands of connections in the AA-handle;

  3. (3)

    There are three bands of connections in the AA-handle.

Note from Figure 13b that since b+c>0b+c>0, all of the labels of bands of connections in the AA-handle are greater than 0 and at least one of the bands of connections has label greater than 11.

(1) Suppose that there is only one band of connections in the AA-handle. Then aa must be 1 and β=BϵAb+c+1\beta=B^{\epsilon}A^{b+c+1}, which is primitive.

(2) Suppose that there are only two bands of connections in the AA-handle. Let pp and qq be the labels of the two bands of connections. Since b+c>0b+c>0 and thus at least one of the bands of connections has label greater than 11, pqp\neq q. Theorem 3.4 forces |pq||p-q| to be 11. However, by Lemma 3.5 such a curve β\beta is a primitive curve.

(3) Suppose that there are three bands of connections in the AA-handle. Let p,q,p,q, and rr be the labels of the three bands of connections with q=p+rq=p+r. In order for β\beta to be a proper power, (p,q,r)=(1,2,1)(p,q,r)=(1,2,1). However, by Lemma 3.6, β\beta is not a proper power curve. ∎

Lemma 3.8.

Suppose a simple closed curve β\beta has an R-R diagram with the form shown in Figure 14 where a,b,c,d>0a,b,c,d>0. Then β\beta is not a proper power curve.

Refer to caption
Figure 14. An R-R diagram of β\beta with a,b,c,d>0a,b,c,d>0.
Proof.

Consider the corresponding hybrid diagram as shown in Figure 15a.

Then we drag the vertex A- together with the edges meeting the vertex A- over the 22-connection in the BB-handle. This corresponds to the change of the cutting disks of HH inducing an automorphism of π1(H)\pi_{1}(H) which takes AAB2A\mapsto AB^{-2} and leaves BB fixed. The resulting hybrid diagram is shown in Figure 15b.

Transforming the hybrid diagram in Figure 15b back into an R-R diagram, there are three possible cases to consider as follows:

  1. (1)

    There is only one band of connections in the AA-handle;

  2. (2)

    There are only two bands of connections in the AA-handle;

  3. (3)

    There are three bands of connections in the AA-handle.

Note from Figure 15b that since b>0b>0, all of the labels of bands of connections in the AA-handle are greater than 0 and at least one of the bands of connections has label greater than 11.

(1) Suppose that there is only one band of connections in the AA-handle. Let pp be the label of the band of connections. Since there are the bb edges connecting A+ and A-, p>1p>1. On the other hand, in Figure 14, consider chasing back and forth the outermost arc in the band of width aa entering the 1-connection in the AA-handle. This represents ABABABAB\cdots, which is carried into AB1AB1AB^{-1}AB^{-1}\cdots by the automorphism AAB2A\mapsto AB^{-2}. This implies that β\beta has a 1-connection in the AA-handle, a contradiction.

Refer to caption
Figure 15. The hybrid diagram of β\beta and the change of cutting disks of HH.

(2) Suppose that there are only two bands of connections in the AA-handle. Let pp and qq be the labels of the two bands of connections. By the similar argument in the proof of Lemma 3.7, |pq|=1|p-q|=1. However, by Lemma 3.7 such a curve β\beta cannot be a proper power.

(3) Suppose that there are three bands of connections in the AA-handle. Let p,q,p,q, and rr be the labels of the three bands of connections with q=p+rq=p+r. In order for β\beta to be a proper power, (p,q,r)=(1,2,1)(p,q,r)=(1,2,1). Then by switching the AA- and BB-handles, and the signs of the labels, the R-R diagram of β\beta has the same form as in Figure 14 with less number of the arcs connecting the AA- and BB-handles. So if we can continue to perform the change of cutting disks of HH, then since the number of the edges is strictly decreasing under the change of cutting disks, this case must eventually belong to the cases (1) and (2). ∎

4. R-R diagrams of Seifert-d curves

In this section, we classify the R-R diagrams of a simple closed curve which is Seifert-d in a genus two handlebody. If α\alpha is a Seifert-d curve in a genus two handlebody HH, then by its definition H[α]H[\alpha] is a Seifert-fibered space over D2D^{2} with two exceptional fibers. In order to compute the type of exceptional fibers in H[α]H[\alpha], we need the following notation and lemma.

Notation. If U=(a,b)U=(a,b) is an element of \mathbb{Z}\oplus\mathbb{Z}, let UU^{\perp} denote the element (b,a)(-b,a) of \mathbb{Z}\oplus\mathbb{Z}, and let ‘\circ’ denote the usual inner product or dot product of vectors.

Lemma 4.1.

Let U=(a,b)U=(a,b), V=(c,d)V=(c,d) and W=(e,f)W=(e,f) be three elements of \mathbb{Z}\oplus\mathbb{Z} such that adbc=±1ad-bc=\pm 1. If WW is expressed as a linear combination of UU and VV, say W=xU+yVW=xU+yV, then y=±(UW)y=\pm(U^{\perp}\circ W).

Proof.

We take an inner product by UU^{\perp} on both sides of W=xU+yVW=xU+yV . Then since UU=0U^{\perp}\circ U=0 and UV=±1U^{\perp}\circ V=\pm 1, the result follows. ∎

The main result of this section is the following theorem.

Theorem 4.2.

If α\alpha is a nonseparating simple closed curve in the boundary of a genus two handlebody HH such that H[α]H[\alpha] is Seifert-fibered over D2D^{2} with two exceptional fibers, then α\alpha has an R-R diagram with the form of Figure 16a with n,s>1n,s>1, or 16b with n,s>1n,s>1, a,b>0a,b>0, and gcd(a,b)=1\gcd(a,b)=1.

Conversely, if α\alpha has an R-R diagram with the form of Figure 16a with n,s>1n,s>1, or Figure 16b with n,s>1n,s>1, a,b>0a,b>0, and gcd(a,b)=1\gcd(a,b)=1, then H[α]H[\alpha] is Seifert-fibered over D2D^{2} with two exceptional fibers of indexes nn and ss in Figure 16a or indexes n(a+b)+bn(a+b)+b and ss in Figure 16b.

In addition, a curve β\beta shown in Figure 17, which is an augmentation of Figure 16, is a regular fiber of H[α]H[\alpha].

Refer to caption
Figure 16. If α\alpha is a nonseparating simple closed curve in the boundary of a genus two handlebody HH such that H[α]H[\alpha] is Seifert-fibered over D2D^{2} with two exceptional fibers, then α\alpha has an R-R diagram with the form of one of these figures with n,s>1n,s>1, a,b>1a,b>1, and gcd(a,b)=1\gcd(a,b)=1. The converse also holds. (See Figure 17.)
Proof.

We start by showing that if H[α]H[\alpha] is Seifert-fibered over D2D^{2} with two exceptional fibers, then α\alpha has an R-R diagram of the claimed form.

The key idea is that H[α]H[\alpha], which is defined to be the manifold obtained by adding a 2-handle to HH along α\alpha, induces a genus two Heegaard decomposition of a Seifert-fibered space over D2D^{2} with two exceptional fibers. However Heegaard decompositions of a Seifert-fibered space over D2D^{2} are well understood. For instance, Theorem 4.4, of Boileau, Rost and Zieschang, completely describes the genus two Heegaard diagrams of a Seifert-fibered space over D2D^{2} with two exceptional fibers. Using this result, Theorem 4.5 shows how the Heegaard diagrams described in Theorem 4.4 translate into R-R diagrams. And then Lemma 4.6 adds a finishing detail to the proof of this direction by showing that it is always possible to assume that n>1n>1 in Figure 16b.

With the proof of one direction of Theorem 4.2 finished, it remains to show that if α\alpha has an R-R diagram with the form of Figure 16a with n,s>1n,s>1, or Figure 16b with n,s>1n,s>1, a,b>0a,b>0, and gcd(a,b)=1\gcd(a,b)=1, then H[α]H[\alpha] is Seifert-fibered over D2D^{2} with two exceptional fibers.

To see that this is the case, consider Figure 17 in which each of the R-R diagrams of Figure 16 has been augmented with a simple closed curve β\beta disjoint from α\alpha. Then, in each diagram of Figure 17, two parallel copies of β\beta bound an essential separating annulus 𝒜\mathcal{A} in HH. (Figure 18 illustrates the situation when s=2s=2, and DAD_{A} and DBD_{B} are cutting disks of HH underlying the A-handle and B-handle of the R-R diagram of α\alpha.)

Cutting HH apart along 𝒜\mathcal{A} yields a genus two handlebody WW and a solid torus VV. Note that α\alpha lies in W\partial W as a primitive curve in WW implying that W[α]W[\alpha] is a solid torus, because any component of DBWD_{B}\cap W is a cutting disk DCD_{C} of WW such that an R-R diagram of α\alpha with respect to {DA,DC}\{D_{A},D_{C}\} of WW has the form in Figure 16 with ss replaced by 11, in which case α\alpha in Figure 16a intersects DCD_{C} only once and thus is primitive, and α\alpha in Figure16b is primitive by Lemma 3.5. It follows that H[α]H[\alpha] is obtained by gluing the two solid tori W[α]W[\alpha] and VV together along 𝒜\mathcal{A}. So H[α]H[\alpha] is Seifert-fibered over D2D^{2} with β\beta as a regular fiber and the cores of W[α]W[\alpha] and VV as exceptional fibers.

The last step is to compute the indexes of the two exceptional fibers of H[α]H[\alpha]. It is clear that the annulus 𝒜\mathcal{A} wraps around the solid torus VV ss times longitudinally, so the core of VV is an exceptional fiber of index s>1s>1. For the other index, it follows by computing π1((W[α])[β])(=π1((W[β])[α])=n)\pi_{1}((W[\alpha])[\beta])(=\pi_{1}((W[\beta])[\alpha])=\mathbb{Z}_{n}) that if α\alpha has an R-R diagram with the form of Figure 16a, then the core of W[α]W[\alpha] is an exceptional fiber of index nn.

Finally, suppose the R-R diagram of α\alpha has the form of Figure 16b. In this case, Lemma 4.1 can be used to compute the index of the second exceptional fiber formed by the core of W[α]W[\alpha].

Abelianizing π1(W)\pi_{1}(W), we have: [α][\vec{\alpha}] = (n(a+b)+b,a+b)(n(a+b)+b,a+b) and [β][\vec{\beta}\,] = (0,1)(0,1). By Lemma 4.1:

[β]=±([α][β])=±(((a+b),n(a+b)+b)(0,1))=±(n(a+b)+b)[\vec{\beta}\,]=\pm([\vec{\alpha}]^{\perp}\circ[\vec{\beta}\,])=\pm((-(a+b),n(a+b)+b)\circ(0,1))=\pm(n(a+b)+b)

in H1(W[α])H_{1}(W[\alpha]). Thus the regular fiber β\beta wraps around the solid torus W[α]W[\alpha] longitudinally n(a+b)+b>1n(a+b)+b>1 times, so the core of W[α]W[\alpha] is an exceptional fiber of index n(a+b)+bn(a+b)+b. ∎

Remark 4.3.

(1) If α\alpha has an R-R diagram of the form shown in Figure 16a(Figure 16b, resp.), α\alpha is said to be a Seifert-d curve of rectangular(non-rectangular, resp.) form.

(2) The regular fiber β\beta is a proper power curve representing BsB^{s} in π1(H)=F(A,B)\pi_{1}(H)=F(A,B), where AA and BB are dual to the cutting disks DAD_{A} and DBD_{B} underlying the AA-handle and BB-handle respectively of the R-R diagram. In addition, since the R-R diagram of Figure 16a is symmetric, the argument of the fifth paragraph in the proof above shows that that the proper power curve representing AnA^{n} in the R-R diagram of Figure 16a is also a regular fiber of H[α]H[\alpha].

Refer to caption
Figure 17. In this figure, each of the R-R diagrams of Figure 16 has been augmented with a simple closed curve β\beta disjoint from α\alpha. Then, in each case, two parallel copies of β\beta bound an essential separating annulus 𝒜\mathcal{A} in HH. Cutting HH apart along 𝒜\mathcal{A} yields a genus two handlebody WW and solid torus VV. Then α\alpha lies in W\partial W and W[α]W[\alpha] is a solid torus. It follows that H[α]H[\alpha] is obtained by gluing two solid tori together along 𝒜\mathcal{A}. So H[α]H[\alpha] is Seifert-fibered over D2D^{2} with β\beta as regular fiber and the cores of W[α]W[\alpha] and VV as exceptional fibers. As for the indexes of the exceptional fibers: If the R-R diagram of α\alpha has the form of Figure 16a, the indexes are nn and ss respectively. If the R-R diagram of α\alpha has the form of Figure 16b, the indexes are n(a+b)+bn(a+b)+b and ss respectively.
Refer to caption
Figure 18. If HH is a handlebody of genus two with cutting disks DAD_{A} and DBD_{B} and β\beta is a nonseparating simple closed curve in H\partial H such that |βDA|=0|\beta\cap\partial D_{A}|=0, and |βDB|=s>1|\beta\cap\partial D_{B}|=s>1, then two parallel copies of β\beta bound an essential separating annulus 𝒜\mathcal{A} in HH. This figure illustrates DAD_{A}, DBD_{B}, and 𝒜\mathcal{A} in the special case s=2s=2.

Turning to the description of Heegaard decompositions of orientable Seifert-fibered spaces over D2D^{2} with two exceptional fibers, let S(ν/p,ω/q)S(\nu/p,\omega/q) denote an orientable Seifert-fibered space over the disk D2D^{2} which has two exceptional fibers of types ν/p\nu/p and ω/q\omega/q with 0<ν<p0<\nu<p and 0<ω<q0<\omega<q.

Also let Wm,n(x,y)W_{m,n}(x,y) be the unique primitive word up to conjugacy in the free group F(x,y)F(x,y) which has (m,n)(m,n) as its abelianization. Then, if vv and ww are words in xx and yy, Wm,n(v,w)W_{m,n}(v,w) is the word obtained from Wm,n(x,y)W_{m,n}(x,y) by substituting vv for xx and ww for yy in Wm,n(x,y)W_{m,n}(x,y).

Then the following theorem of Boileau, Rost and Zieschang(Theorem 5.4 in [BRZ88]) completely describes the genus two Heegaard diagrams of S(ν/p,ω/q)S(\nu/p,\omega/q). For the notations in Theorem 4.4, see Sections 2, 4, and 5 in [BRZ88].

Theorem 4.4.

[BRZ88] The manifold S(ν/p,ω/q)S(\nu/p,\omega/q) admits three genus two Heegaard decompositions HD0HD_{0}, HDSHD_{S}, and HDTHD_{T}, represented by the following Heegaard diagrams:

HD0(sptq;λ,μ),HDS(Wp,ν(u1,tq);,μ),HDT(Wq,ω(v1,sp);λ,).HD_{0}\leftrightarrow(s^{p}t^{-q};\lambda,\mu),\>HD_{S}\leftrightarrow(W_{p,\nu}(u^{-1},t^{q});-,\mu),\>HD_{T}\leftrightarrow(W_{q,\omega}(v^{-1},s^{p});\lambda,-).

Here νλ1modp\nu\lambda\equiv 1\mod p and ωμ1modq\omega\mu\equiv 1\mod q. Any Heegaard decomposition of genus two of S(ν/p,ω/q)S(\nu/p,\omega/q) is homeomorphic to one of these. Moreover:

  1. (1)

    HD0HD_{0} is homeomorphic to HDTHD_{T} (or HDSHD_{S}) if and only if ω±1modq\omega\equiv\pm 1\mod q ( or ν±1modp\nu\equiv\pm 1\mod p, respectively).

  2. (2)

    If ω±1modq\omega\equiv\pm 1\mod q and ν±1modp\nu\equiv\pm 1\mod p, then HD0HD_{0}, HDSHD_{S}, and HDTHD_{T} are all homeomorphic.

  3. (3)

    HDSHD_{S} and HDTHD_{T} are homeomorphic if and only if either case (2) occurs or ν/p±ω/qmod1\nu/p\equiv\pm\omega/q\mod 1 (that is, p=qp=q and νωmodp\nu\equiv\omega\mod p).

Theorem 4.5.

The R-R diagrams in Figures 19, 20, and 21 correspond to the Heegaard diagrams HD0HD_{0}, HDSHD_{S}, and HDTHD_{T} of Theorem 4.4 respectively.

Proof.

First, observe that the curves α\alpha in Figures 19, 20, and 21 represent sptqs^{p}t^{-q}, Wp,ν(u1,tq)W_{p,\nu}(u^{-1},t^{q}), and Wq,ω(v1,sp)W_{q,\omega}(v^{-1},s^{p}) respectively in π1(H)\pi_{1}(H).

Next, consider the diagram of Figure 19, and let CAC_{A} and CBC_{B} be cores of the A-handle and B-handle of HH. Then CAC_{A} and CBC_{B} are the exceptional fibers of the Seifert-fibration of H[α]H[\alpha]. Let N(CA)N(C_{A}) and N(CB)N(C_{B}) be closed regular neighborhoods of CAC_{A} and CBC_{B} respectively in HH, and let MAM_{A} and MBM_{B} be meridional disks of N(CA)N(C_{A}) and N(CB)N(C_{B}) respectively.

Observe that the pair of dotted curves β\beta, γs\gamma_{s} on the A-handle of Figure 19 can be considered to lie on N(CA)\partial N(C_{A}), while the pair of dotted curves β\beta^{\prime}, γt\gamma_{t} on the B-handle of Figure 19 can be considered to lie on N(CB)\partial N(C_{B}). Also observe that as indicated in Remark 4.3, the curves β\beta and β\beta^{\prime} represent regular fibers of the Seifert-fibration of H[α]H[\alpha]. Then MA\partial M_{A} = (βνγsp)±1(\beta^{\nu}\gamma_{s}^{p})^{\pm 1} in π1(N(CA))\pi_{1}(\partial N(C_{A})), while MB\partial M_{B} = (βωγtq)±1(\beta^{\prime\omega}\gamma_{t}^{q})^{\pm 1} in π1(N(CB))\pi_{1}(\partial N(C_{B})). So CAC_{A} and CBC_{B} are exceptional fibers of types ν/p\nu/p and ω/q\omega/q in the Seifert-fibration of H[α]H[\alpha].

Refer to caption
Figure 19. Suppose ν\nu, ω\omega, pp, and qq are positive integers such that 0<ν<p0<\nu<p, 0<ω<q0<\omega<q, gcd(ν,p)=gcd(ω,q)=1\gcd(\nu,p)=\gcd(\omega,q)=1, and HH is a genus two handlebody. Then the manifold H[α]H[\alpha], obtained by adding a 2-handle to H\partial H along a simple closed curve α\alpha in H\partial H that has an R-R diagram with the form of this figure, is a Seifert-fibered space over D2D^{2} with exceptional fibers of types ν/p\nu/p and ω/q\omega/q.
Refer to caption
Figure 20. Suppose ν\nu, ω\omega, pp, and qq are positive integers such that 1<ν<p1<\nu<p, 0<ω<q0<\omega<q, and gcd(ν,p)=gcd(ω,q)=1\gcd(\nu,p)=\gcd(\omega,q)=1. In addition, suppose aa, bb, and nn are positive integers such that a+b=νa+b=\nu, nν+a=pn\nu+a=p, and HH is a genus two handlebody. Then the manifold H[α]H[\alpha], obtained by adding a 2-handle to H\partial H along a simple closed curve α\alpha in H\partial H that has an R-R diagram with the form of this figure, is a Seifert-fibered space over D2D^{2} with exceptional fibers of types ν/p\nu/p and ω/q\omega/q.
Refer to caption
Figure 21. Suppose ν\nu, ω\omega, pp, and qq are positive integers such that 0<ν<p0<\nu<p, 1<ω<q1<\omega<q, and gcd(ν,p)=gcd(ω,q)=1\gcd(\nu,p)=\gcd(\omega,q)=1. In addition, suppose aa, bb, and nn are positive integers such that a+b=ωa+b=\omega, nω+a=qn\omega+a=q, and HH is a genus two handlebody. Then the manifold H[α]H[\alpha], obtained by adding a 2-handle to H\partial H along a simple closed curve α\alpha in H\partial H that has an R-R diagram with the form of this figure, is a Seifert-fibered space over D2D^{2} with exceptional fibers of types ν/p\nu/p and ω/q\omega/q.

This leaves the diagrams of Figures 20 and 21. Since these diagrams are similar, we will only consider Figure 20 in detail. To start, note that the configuration of the curves β\beta and γt\gamma_{t} on the B-handle of Figure 20 is identical to that of β\beta^{\prime} and γt\gamma_{t} on the B-handle of Figure 19. Since β\beta is again a regular fiber in the Seifert-fibration of H[α]H[\alpha] when α\alpha has an R-R diagram on H\partial H with the form of Figure 20, the core of the B-handle CBC_{B} of HH is again an exceptional fiber of type ω/q\omega/q. The other exceptional fiber that exists in H[α]H[\alpha] when α\alpha has an R-R diagram on H\partial H with the form of Figure 20, arises in a slightly different way.

As in Figure 18, two parallel copies of the regular fiber β\beta in Figure 20 bound an essential separating annulus 𝒜\mathcal{A} in HH. Cutting HH open along 𝒜\mathcal{A} cuts HH into a genus two handlebody WW and a solid torus VV which has CBC_{B} as its core. The curve α\alpha lies on W\partial W, and the R-R diagram of α\alpha on W\partial W appears in Figure 22. Since α\alpha is primitive in WW, W[α]W[\alpha] is a solid torus VV^{\prime}. Let CVC_{V^{\prime}} be the core of VV^{\prime}. Then CVC_{V^{\prime}} is the second exceptional fiber of the Seifert-fibration of H[α]H[\alpha].

Let MM be the meridional disk of VV^{\prime}, and note that the curves β\beta and γu\gamma_{u} of Figure 22 lie on W\partial W and β\beta and γu\gamma_{u} form a basis for π1(W[α])\pi_{1}(\partial W[\alpha]). The next step is to obtain an expression for M\partial M in π1(W[α])\pi_{1}(\partial W[\alpha]) in terms of the basis β\beta, γu\gamma_{u} of π1(W[α])\pi_{1}(\partial W[\alpha]).

We can do this by using Lemma 4.1. Abelianizing π1(W)\pi_{1}(W), we have: [α][\vec{\alpha}] = (n(a+b)a,a+b)(-n(a+b)-a,a+b) = (p,ν)(-p,\nu), [γu][\vec{\gamma}_{u}] = (1,0)(-1,0), and [β][\vec{\beta}\,] = (0,1)(0,1). By Lemma 4.1:

[γu]=δ([α][γu])=δ((ν,p)(1,0))=δν,and[\vec{\gamma}_{u}]=\delta([\vec{\alpha}]^{\perp}\circ[\vec{\gamma}_{u}])=\delta((\nu,p)\circ(-1,0))=-\delta\nu,\quad\text{and}
[β]=δ([α][β])=δ((ν,p)(0,1))=δp[\vec{\beta}\,]=\delta([\vec{\alpha}]^{\perp}\circ[\vec{\beta}\,])=\delta((\nu,p)\circ(0,1))=\delta p

in H1(W[α])H_{1}(W[\alpha]), where δ=±1\delta=\pm 1. It follows that M\partial M = (βνγup)±1(\beta^{\nu}\gamma_{u}^{p})^{\pm 1} in π1(W[R])\pi_{1}(\partial W[R]). So CVC_{V^{\prime}} is an exceptional fiber of type ν/p\nu/p in the Seifert-fibration of H[α]H[\alpha] when α\alpha has an R-R diagram with the form of Figure 20.

Similarly, one sees that if α\alpha has an R-R diagram with the form of Figure 21, then H[α]H[\alpha] is also Seifert-fibered over D2D^{2} with two exceptional fibers of types ν/p\nu/p and ω/q\omega/q. ∎

It will be convenient to be able to assume that n>1n>1 in Figure 16b. The following lemma shows this can always be done.

Refer to caption
Figure 22. Suppose α\alpha has an R-R diagram on H\partial H with the form of Figure 20. Let WW be the genus two handlebody obtained when HH is cut open along the essential separating annulus in HH bounded by two parallel copies of the regular fiber β\beta in Figure 20. Then α\alpha lies on W\partial W, and α\alpha has an R-R diagram on W\partial W with the form of this figure. Then W[α]W[\alpha] is a solid torus, and the curves β\beta and γu\gamma_{u} are a basis for W[α]\partial W[\alpha]. By using Lemma 4.1 to compute the images of β\beta and γu\gamma_{u} in H1(W[α])H_{1}(W[\alpha]), it is possible to see that the core of W[α]W[\alpha] is an exceptional fiber of type ν/p\nu/p in the Seifert-fibration of H[α]H[\alpha].
Refer to caption
Figure 23. Suppose α\alpha has an R-R diagram on H\partial H with the form of Figure 21. Let WW be the genus two handlebody obtained when HH is cut open along the essential separating annulus in HH bounded by two parallel copies of the regular fiber β\beta in Figure 21. Then α\alpha lies on W\partial W, and α\alpha has an R-R diagram on W\partial W with the form of this figure. Then W[α]W[\alpha] is a solid torus, and the curves β\beta and γv\gamma_{v} are a basis for W[α]\partial W[\alpha]. By using Lemma 4.1 to compute the images of β\beta and γu\gamma_{u} in H1(W[α])H_{1}(W[\alpha]), it is possible to see that the core of W[α]W[\alpha] is an exceptional fiber of type ω/q\omega/q in the Seifert-fibration of H[α]H[\alpha].
Lemma 4.6.

Suppose S(ν/p,ω/q)S(\nu/p,\omega/q) has an R-R diagram with the form of Figure 20 or 21 with n=1n=1. Then S(ν/p,ω/q)S(\nu/p,\omega/q) also has another R-R diagram with the form of Figure 20 or 21 in which n>1n>1.

Proof.

Suppose a simple closed curve α\alpha in the boundary of a genus two handlebody HH has an R-R diagram with the form of Figure 20 with n=1n=1. For the form of Figure 21, the similar argument can apply. Note that since n=1n=1, p=2a+bp=2a+b and ν=a+b\nu=a+b. It is easy to see that the underlying Heegaard diagram of α\alpha on H\partial H does not have minimal complexity. Thus in order to have minimal complexity, we can perform a change of cutting disks of HH, i.e., replace the cutting disk DBD_{B} of HH which underlies the BB-handle with a new cutting disk DBD^{\prime}_{B} by bandsumming DBD_{B} with the cutting disk DAD_{A} in the AA-handle along the arc of α\alpha.

Specifically, for the weights aa and bb in the R-R diagram, since gcd(a,b)=1(a,b)=1, we can let b=ρa+rb=\rho a+r, where ρ0\rho\geq 0 and 0r<a0\leq r<a. However we may assume r>0r>0, otherwise a=1a=1 and thus p=2+bp=2+b and ν=1+b\nu=1+b, which implies that ν1modp\nu\equiv-1\mod p and it follows from Theorem 4.4 that this Heegaard decomposition is homeomorphic to HD0HD_{0}.

Now we record α\alpha by starting the aa parallel arcs entering into the 2-2-connection in the AA-handle. It follows from the R-R diagram that α\alpha is the product of two subwords A2Bq(A1Bq)ρA^{-2}B^{q}(A^{-1}B^{q})^{\rho} and A2Bq(A1Bq)ρ+1A^{-2}B^{q}(A^{-1}B^{q})^{\rho+1} with |A2Bq(A1Bq)ρ|=ar|A^{-2}B^{q}(A^{-1}B^{q})^{\rho}|=a-r and |A2Bq(A1Bq)ρ+1|=r|A^{-2}B^{q}(A^{-1}B^{q})^{\rho+1}|=r. We perform a change of cutting disks of the handlebody HH which induces an automorphism of π1(H)\pi_{1}(H) that takes A1A1BqA^{-1}\mapsto A^{-1}B^{-q}. Then by this change of cutting disks, A2Bq(A1Bq)ρA^{-2}B^{q}(A^{-1}B^{q})^{\rho} and A2Bq(A1Bq)ρ+1A^{-2}B^{q}(A^{-1}B^{q})^{\rho+1} are sent to A1Bq(A1)ρ+1A^{-1}B^{-q}(A^{-1})^{\rho+1} and A1Bq(A1)ρ+2A^{-1}B^{-q}(A^{-1})^{\rho+2}. Therefore the resulting Heegaard diagram of α\alpha has minimal complexity and realizes a new R-R diagram of the form in Figure 24, which is the same form as in Figure 20 with n>1n>1. ∎

Refer to caption
Figure 24. A new R-R diagram of α\alpha obtained by performing a change of cutting disks of the handlebody HH which induces an automorphism of π1(H)\pi_{1}(H) that takes A1A1BqA^{-1}\mapsto A^{-1}B^{-q}.
Remark 4.7.

The following observations are relevant for Lemma 4.6.

  1. (1)

    The change of cutting disks of HH inducing the automorphism A1A1BqA^{-1}\mapsto A^{-1}B^{-q} in π1(H)\pi_{1}(H) corresponds to the change of the cutting disk DBD_{B} to a new cutting disk DBD_{B}^{\prime} which is obtained by bandsumming DBD_{B} with the cutting disk DAD_{A} along the arcs of α\alpha qq times.

  2. (2)

    The diagram of Figure 24 corresponds to the Heegaard decomposition of S((pν)/p,ω/q)S((p-\nu)/p,-\omega/q), which is homeomorphic to S(ν/p,ω/q)S(\nu/p,\omega/q) as desired.

5. R-R diagrams of Seifert-m curves

If α\alpha is a Seifert-m curve on a genus two handlebody HH, then by its definition H[α]H[\alpha] is a Sefiert-fibered space over the Möbius band with at most one exceptional fiber. The main result of this section is the following theorem.

Theorem 5.1.

If α\alpha is a nonseparating simple closed curve in the boundary of a genus two handlebody HH such that H[α]H[\alpha] is a Seifert-fibered space MM over the Möbius band, then α\alpha has an R-R diagram of the form shown in Figure 25, and α\alpha represents ABsA1BsAB^{s}A^{-1}B^{s} in π1(H)\pi_{1}(H). There is no loss in taking s>0s>0, and then MM has an exceptional fiber if and only if s>1s>1, in which case, ss equals the index of the exceptional fiber of MM.

Conversely, if α\alpha has an R-R diagram of the form shown in Figure 25, then H[α]H[\alpha] is Seifert-fibered over the Möbius band with one exceptional fiber of index ss provided that s>1s>1. If s=1s=1 in Figure 25, then H[α]H[\alpha] is Seifert-fibered space over the Möbius band with no exceptional fibers.

In addition, a curve β\beta shown in Figure 26, which is an augmentation of Figure 26, is a regular fiber of H[α]H[\alpha].

Refer to caption
Figure 25. If attaching a 2-handle to a genus two handlebody HH along a simple closed curve α\alpha yields a Seifert-fibered space over the Möbius band with one exceptional fiber of index s>1s>1, then α\alpha has an R-R diagram with the form of this figure in which α=ABsA1Bs\alpha=AB^{s}A^{-1}B^{s} in π1(H)\pi_{1}(H).
Proof.

Suppose H[α]H[\alpha] is homeomorphic to a Seifert-fibered space MM over the Möbius band. Then MM contains an essential nonseparating annulus which is vertical in the Seifert fibration of MM. (Such an annulus can be easily obtained by starting with a nonseparating arc in the Möbius band—taking care to choose an arc which misses any exceptional fiber of MM—and then saturating that arc in the Seifert-fibration of MM.)

Since MM is Seifert-fibered, and not a solid torus, H[α]H[\alpha] is \partial-irreducible. This implies α\alpha intersects every cutting disk of HH. Now a theorem of Eudave-Muñoz applies. It is shown in [EM94] that if H[α]H[\alpha] contains an essential nonseparating annulus, then there exists an essential nonseparating annulus 𝒜\mathcal{A} in HH, with 𝒜\partial\mathcal{A} and α\alpha disjoint, such that 𝒜\mathcal{A} is essential in H[α]H[\alpha]. (Note that in [EM94] the definition of an essential annulus which is properly embedded in a 3-manifold MM is that it is incompressible and not \partial-parallel.) Furthermore, it follows from Lemmas 1.10, 1.11 and the argument following them in [H07] that 𝒜\mathcal{A} is vertical in the Seifert-fibration of H[α]H[\alpha].

This suggests looking for all possible R-R diagrams of α\alpha by starting with an R-R diagram 𝒟\mathcal{D} of the boundary components of a nonseparating annulus 𝒜\mathcal{A} in H\partial H. Then any R-R diagram of α\alpha must be obtained by adding α\alpha to 𝒟\mathcal{D} so that α\alpha is disjoint from the curves β\beta and β^\hat{\beta} of 𝒜\partial\mathcal{A}, and α\alpha intersects every cutting disk of HH.

Lemma 5.3 carries out the first step of this scheme by showing that if 𝒜\mathcal{A} is a nonseparating essential annulus in a genus two handlebody HH, and β\beta and β^\hat{\beta} are the components of 𝒜\partial\mathcal{A} in H\partial H, then β\beta and β^\hat{\beta} have an R-R diagram 𝒟\mathcal{D} of the form shown in Figure 28.

Next, Lemma 5.5 shows that if 𝒟\mathcal{D} is an R-R diagram with the form of Figure 28, and a simple closed curve α\alpha is added to 𝒟\mathcal{D} so that α\alpha is disjoint from β\beta and β^\hat{\beta} in 𝒟\mathcal{D}, and α\alpha intersects every cutting disk of HH, then the resulting R-R diagram must have the form shown in Figure 31.

Refer to caption
Figure 26. Suppose ss and ω\omega are positive integers such that 0<2ωs0<2\omega\leq s, gcd(s,ω)=1\gcd(s,\omega)=1, and HH is a genus two handlebody. Then the manifold H[α]H[\alpha], obtained by adding a 2-handle to H\partial H along a simple closed curve α\alpha in H\partial H that has an R-R diagram with the form of this figure, is a Seifert-fibered space over the Möbius band with one exceptional fiber of type ω/s\omega/s, whose regular fiber is the curve β\beta.

Lemmas 5.6 and 5.7 finish the argument by showing that if H[α]H[\alpha] is Seifert-fibered over the Möbius band, then (a,b)=(0,1)(a,b)=(0,1) in Figure 31, so Figure 31 reduces to Figure 25. In addition, these lemmas show that H[α]H[\alpha] has an exceptional fiber if and only if s>1s>1, and when s>1s>1, ss equals the index of the exceptional fiber of H[α]H[\alpha].

Now we prove the second statement of the theorem. First, suppose α\alpha and β\beta have an R-R diagram of the form shown in Figure 26 with ss replaced by 11, which is an augmentation of the R-R diagram of α\alpha in Figure 25. We will show that H[α]H[\alpha] is a Seifert-fibered space over the Möbius band with no exceptional fibers whose regular fiber is represented by the curve β\beta. Consider a properly embedded nonseparating annulus 𝒜\mathcal{A} in HH whose boundary consists of the curves β\beta and β^\hat{\beta}, where β^\hat{\beta} is a curve illustrated in Figure 28. Figure 27a shows the genus two handlebody HH, the simple closed curve α\alpha on H\partial H, and the annulus 𝒜\mathcal{A} with 𝒜=ββ^\partial\mathcal{A}=\beta\cup\hat{\beta} oriented, which realize the R-R diagram of α\alpha and β\beta in Figure 26 with s=1s=1 and the R-R diagram of β^\hat{\beta} in Figure 28. Let H/𝒜H/\mathcal{A} be the manifold obtained by cutting HH along 𝒜\mathcal{A}. Let 𝒜1\mathcal{A}_{1} and 𝒜2\mathcal{A}_{2}, β1\beta_{1} and β2\beta_{2}, and β^1\hat{\beta}_{1} and β^1\hat{\beta}_{1} be the copies of 𝒜\mathcal{A}, β\beta, and β^\hat{\beta} in H/𝒜H/\mathcal{A}. Then it is easy to see that H/𝒜H/\mathcal{A} is a genus two handlebody such that 𝒜1\mathcal{A}_{1} and 𝒜2\mathcal{A}_{2} together with 𝒜1=β1β^1\partial{\mathcal{A}_{1}}=\beta_{1}\cup\hat{\beta}_{1} and 𝒜2=β2β^2\partial{\mathcal{A}_{2}}=\beta_{2}\cup\hat{\beta}_{2} lie in (H/𝒜)\partial(H/\mathcal{A}) as shown in Figure 27b.

Refer to caption
Figure 27. The genus two handlebody HH, the simple closed curve α\alpha on H\partial H, and the annulus 𝒜\mathcal{A} with 𝒜=ββ^\partial\mathcal{A}=\beta\cup\hat{\beta} oriented, which realize the R-R diagram of α\alpha in Figure 25 with s=1s=1 and the R-R diagram of β\beta and β^\hat{\beta} in Figure 28 in a), and the manifold H/𝒜H/\mathcal{A} obtained by cutting HH along 𝒜\mathcal{A} in b), which is a genus two handlebody.

Since α\alpha is disjoint from 𝒜\mathcal{A}, H[α]H[\alpha] is obtained from (H/𝒜)[α](H/\mathcal{A})[\alpha] by gluing the two copies 𝒜1\mathcal{A}_{1} and 𝒜2\mathcal{A}_{2} such that the orientations of their boundaries β1β^1\beta_{1}\cup\hat{\beta}_{1} and β2β^2\beta_{2}\cup\hat{\beta}_{2} match. However, we can observe from Figure 27b that α\alpha is primitive in the genus two handlebody H/𝒜H/\mathcal{A} and thus (H/𝒜)[α](H/\mathcal{A})[\alpha] is a solid torus. Thus gluing 𝒜1\mathcal{A}_{1} and 𝒜2\mathcal{A}_{2} in the boundary of the solid torus (H/𝒜)[α](H/\mathcal{A})[\alpha] yields Seifert-fibered over the Möbius band with no exceptional fibers such that β\beta is a regular fiber. Therefore H[α]H[\alpha] is a Seifert-fibered space over the Möbius band with no exceptional fibers whose regular fiber is represented by the curve β\beta.

Now we suppose that α\alpha and β\beta have an R-R diagram of the form shown in Figure 26 with s>1s>1. Similarly as in the proof of Theorem 4.2, the two parallel copies of β\beta bound an essential separating annulus 𝒜\mathcal{A}^{\prime} in HH as shown in Figure 18, which cuts HH apart into a genus two handlebody WW and a solid torus VV. Note that α\alpha lies in the boundary of WW. To complete the proof, it suffices to show that W[α]W[\alpha] is a Seifert-fibered space over the Möbius band with no exceptional fibers and the curve β\beta is a regular fiber of W[α]W[\alpha].

A component of DBWD_{B}\cap W is a cutting disk DCD_{C} of WW such that α\alpha and β\beta intersect DCD_{C} transversely once. This implies that the R-R diagram of α\alpha and β\beta with respect to {DA,DC}\{D_{A},D_{C}\} has the form in Figure 26 with ss replaced by 11. Therefore by the argument above in the case that s=1s=1, we see that W[α]W[\alpha] is a Seifert-fibered space over the Möbius band with no exceptional fibers and the curve β\beta is a regular fiber of W[α]W[\alpha], as desired. ∎

Remark 5.2.

In Theorem 5.1, if s=1s=1, then H[α]H[\alpha] is a Seifert-fibered space MM over the Möbius band with no exceptional fibers. Then MM admits another Seifert-fibered structure, i.e., MM is a Seifert-fibered space over D2D^{2} with two exceptional fibers of both indexes 2. Therefore we can regard a Seifert-m curve α\alpha in this case as a Seifert-d curve, and thus we may assume s>1s>1 in the R-R diagram of a Seifert-m curve α\alpha.

(2) Figure 26 shows an R-R diagram of α\alpha such that H[α]H[\alpha] is a Seifert-fibered space over the Möbius band with one exceptional fiber of type ω/s\omega/s, whose regular fiber is the curve β\beta.

Lemma 5.3.

Suppose 𝒜\mathcal{A} is an essential nonseparating annulus properly embedded in a genus two handlebody HH. Let β\beta and β^\hat{\beta} be the two curves in H\partial H that form 𝒜\partial\mathcal{A}. Then the pair β\beta, β^\hat{\beta} have an R-R diagram that appears in Figure 28.

Refer to caption
Figure 28. If 𝒜\mathcal{A} is a nonseparating essential annulus in a genus two handlebody HH with 𝒜\partial\mathcal{A} = ββ^\beta\cup\hat{\beta}, then there exists s>0s>0 such that β\beta and β^\hat{\beta} have an R-R diagram with the form of this figure.
Proof.

Given 𝒜\mathcal{A} and its boundary components β\beta and β^\hat{\beta}, we claim the following.

Claim 5.4.

There exists a complete set of cutting disks {DA,DB}\{D_{A},D_{B}\} of HH such that one of the cutting disks, say DAD_{A}, is disjoint from 𝒜\mathcal{A} and 𝒜DB\mathcal{A}\cap D_{B} consists of a set of s>0s>0 essential spanning arcs in 𝒜\mathcal{A}.

Refer to caption
Figure 29. Gluing two copies DD^{\prime} and D′′D^{\prime\prime} of DD to the disk DD^{*} of 𝒜\mathcal{A} along σ\sigma^{\prime} and σ′′\sigma^{\prime\prime} yields a disk DCD_{C}.
Proof.

First, we show that there exists a cutting disk of HH disjoint from 𝒜\mathcal{A}. Let DAD_{A} be a cutting disk of HH which intersects 𝒜\mathcal{A} minimally.

Suppose DA𝒜D_{A}\cap\mathcal{A}\neq\varnothing. Then we may assume that DAD_{A} intersects 𝒜\mathcal{A} essentially and DA𝒜D_{A}\cap\mathcal{A} consists of properly embedded disjoint arcs and disjoint circles. However by the incompressibility of 𝒜\mathcal{A}, irreducibility of HH, and the minimality condition rule out circle intersections. Suppose γ\gamma is an outermost arc of DA𝒜D_{A}\cap\mathcal{A} which cuts a disk DD of 𝒜\mathcal{A}. Then γ\gamma also cuts DAD_{A} into two subdisks D1D_{1} and D2D_{2} of DAD_{A}. Consider two disks DγD1D\cup_{\gamma}D_{1} and DγD2D\cup_{\gamma}D_{2}, which are obtained by gluing DD and D1D_{1}, and DD and D2D_{2} respectively along γ\gamma. Then since DAD_{A} is a cutting disk which means that it is nonseparating, at least one of DγD1D\cup_{\gamma}D_{1} and DγD2D\cup_{\gamma}D_{2} is nonseparting and thus is a cutting disk. But this cutting disk intersects 𝒜\mathcal{A} less than DAD_{A}. This is a contradiction to the minimality.

Suppose σ\sigma is a spanning arc of DA𝒜D_{A}\cap\mathcal{A} in 𝒜\mathcal{A} which is outermost in DAD_{A}. Let DD^{\prime} and D′′D^{\prime\prime} be two copies of DD, and let σ=D𝒜\sigma^{\prime}=D^{\prime}\cap\mathcal{A} and σ′′=D′′𝒜\sigma^{\prime\prime}=D^{\prime\prime}\cap\mathcal{A} as shown in Figure 29. Also let DD^{*} be a disk 𝒜N(σ)\mathcal{A}-N(\sigma) such that σ\sigma^{\prime} and σ′′\sigma^{\prime\prime} are included in D\partial D^{*}. Gluing the two copies DD^{\prime} and D′′D^{\prime\prime} to the disk DD^{*} along σ\sigma^{\prime} and σ′′\sigma^{\prime\prime} yields a disk DCD_{C}. Since 𝒜\mathcal{A} is nonseparating, from the construction DCD_{C} is nonseparating (and thus is a cutting disk) and also does not intersect 𝒜\mathcal{A}, a contradiction. Therefore there exists a cutting disk DAD_{A} of HH disjoint from 𝒜\mathcal{A}.

Now let DBD_{B} be a cutting disk of HH chosen so that {DA,DB}\{D_{A},D_{B}\} is a complete set of cutting disks of HH and DBD_{B} intersects 𝒜\mathcal{A} minimally. By applying the same argument above, we can show that DB𝒜D_{B}\cap\mathcal{A} consists of spanning arcs of 𝒜\mathcal{A}. Note that |DB𝒜|>0|D_{B}\cap\mathcal{A}|>0, otherwise 𝒜\mathcal{A} would embed properly in a 3-ball and thus not be essential. This completes the proof of the claim. ∎

By the claim, there exists a complete set of cutting disks {DA,DB}\{D_{A},D_{B}\} of HH such that DAD_{A} is disjoint from 𝒜\mathcal{A} and 𝒜DB\mathcal{A}\cap D_{B} consists of a set of s>0s>0 essential spanning arcs in 𝒜\mathcal{A}. Now consider the solid torus VV obtained by cutting HH open along DAD_{A}. Then V\partial V contains two disks DA+D_{A}^{+} and DAD_{A}^{-}, which are copies of DAD_{A}. The simple closed curves β\beta and β^\hat{\beta} also lie in V\partial V and cut V\partial V into two annuli, say A+A^{+} and AA^{-}. Since β\beta and β^\hat{\beta} are not isotopic in H\partial H, DA+D_{A}^{+} lies in the interior of one of these annuli, and DAD_{A}^{-} lies in the interior of the other. It is easy to see that there exists an arc τ\tau in V\partial V connecting DA+D_{A}^{+} and DAD_{A}^{-} which |τ(ββ^)|=1|\tau\cap(\beta\cup\hat{\beta})|=1. Furthermore since a regular neighborhood of DA+DAτD_{A}^{+}\cup D_{A}^{-}\cup\tau is a disk in V\partial V, we can isotope DA+DAτD_{A}^{+}\cup D_{A}^{-}\cup\tau keeping |τ(ββ^)|=1|\tau\cap(\beta\cup\hat{\beta})|=1 so that DA+DAτD_{A}^{+}\cup D_{A}^{-}\cup\tau is disjoint from the meridional disk DBD_{B} of VV. Since |τ(ββ^)|=1|\tau\cap(\beta\cup\hat{\beta})|=1, now we may assume that |τβ^|=1|\tau\cap\hat{\beta}|=1.

Next, let TAT_{A} be a once-punctured torus in H\partial H which is a regular neighborhood in H\partial H of τDA\tau\cup\partial D_{A}, with TAT_{A} chosen so that TAT_{A} is disjoint from DB\partial D_{B}, and so that TAT_{A} intersects ββ^\beta\cup\hat{\beta} minimally. Then, if 𝒟\mathcal{D} is the R-R diagram of β\beta and β^\hat{\beta}, whose A-handle corresponds to TAT_{A}, then β^\hat{\beta} crosses the A-handle of 𝒟\mathcal{D} in a single connection without intersecting DAD_{A}, while β\beta lies completely in the B-handle of 𝒟\mathcal{D}. So 𝒟\mathcal{D} has the form of the diagram in Figure 28. ∎

Lemma 5.5.

If 𝒟\mathcal{D} is an R-R diagram with the form of Figure 28, and a simple closed curve α\alpha is added to 𝒟\mathcal{D} so that α\alpha is disjoint from β\beta and β^\hat{\beta} in 𝒟\mathcal{D}, and α\alpha intersects every cutting disk of HH, then the resulting R-R diagram must have the form shown in Figure 31.

Refer to caption
Figure 30. This figure shows that a simple closed curve α\alpha on H,\partial H, which is disjoint from the curves β\beta and β^\hat{\beta} forming 𝒜\partial\mathcal{A}, can not contain both a 0-connection and a 1-connection on the A-handle of this diagram. (Otherwise α\alpha is forced to spiral endlessly.)
Refer to caption
Figure 31. If a 2-handle is added to the genus two handlebody HH of Figure 28 along a simple closed curve α\alpha, disjoint from the curves β\beta and β^\hat{\beta} forming 𝒜\partial\mathcal{A}, the annulus 𝒜\mathcal{A} of Figure 28 is essential in H[α]H[\alpha], and α\alpha intersects every cutting disk of HH, then there are nonnegative integers aa and bb such that α\alpha, β\beta and β^\hat{\beta} have an R-R diagram with the form of this figure.
Proof.

Since α\alpha must intersect every cutting disk of HH, it must traverse both the A-handle and the B-handle of 𝒟\mathcal{D}. In particular, α\alpha must have 1-connections on the A-handle of 𝒟\mathcal{D} and s-connections on the B-handle of 𝒟\mathcal{D}. Now Figure 30 shows that α\alpha can’t have both 1-connections and 0-connections on the A-handle of 𝒟\mathcal{D}, otherwise α\alpha is forced to spiral endlessly. So α\alpha has only 1-connections on the A-handle of 𝒟\mathcal{D}. Now it is not hard to see that there must exist nonnegative integers aa and bb such that the diagram of α\alpha, β\beta, and β^\hat{\beta} has the form of Figure 31. ∎

Lemma 5.6.

Suppose α\alpha has an R-R diagram with the form of Figure 31, H[α]H[\alpha] is Seifert-fibered over the Möbius band, and s>1s>1 in Figure 31. Then (a,b)=(0,1)(a,b)=(0,1), Figure 31 reduces to Figure 25, and α=ABsA1Bs\alpha=AB^{s}A^{-1}B^{s} in π1(H)\pi_{1}(H). Furthermore, the core of the B-handle of HH is an exceptional fiber of index ss in the Seifert-fibration of H[α]H[\alpha].

Proof.

Since the annulus 𝒜\mathcal{A} is vertical in the Seifert-fibration of H[α]H[\alpha], its boundary components β\beta and β^\hat{\beta} are regular fibers. Then as in Figure 18 two parallel copies of β\beta bound an essential separating annulus 𝒜\mathcal{A}^{\prime} in HH such that 𝒜\mathcal{A}^{\prime} is saturated and vertical in the Seifert-fibration of H[α]H[\alpha], and 𝒜\mathcal{A}^{\prime} cuts HH into a solid torus VV and a genus two handlebody WW, with α\alpha lying on W\partial W.

Let λ\lambda be the core of VV. (Note that λ\lambda is also a core of the B-handle of HH.) Then, since s>1s>1, λ\lambda is an exceptional fiber of index ss in the Seifert-fibration of H[α]H[\alpha]. And, since the Seifert-fibration of H[α]H[\alpha] can have at most one exceptional fiber, λ\lambda is the only exceptional fiber in H[α]H[\alpha]. It follows that the manifold W[α]W[\alpha] is Seifert-fibered over the Möbius band with no exceptional fibers. Then, using well-known formulas for presentations of Seifert-fibered spaces, one gets π1(W[α])\pi_{1}(W[\alpha]) = x,y|x2y2\langle x,y\,|\,x^{2}y^{2}\rangle.

By a result of Zieschang in [Z77], π1(W[α])\pi_{1}(W[\alpha]) has only one Nielsen equivalence class of generators. It follows that if x,y|\langle x,y\,|\,\mathcal{R}\rangle is a one-relator presentation of π1(W[α])\pi_{1}(W[\alpha]), then there is an automorphism of the free group F(x,y)F(x,y) which carries \mathcal{R} onto a cyclic conjugate of x2y2x^{2}y^{2} or its inverse.

Now it is not hard to see that one obtains a one-relator presentation of π1(W[α])\pi_{1}(W[\alpha]) from the one-relator presentation A,B|α\langle A,B\,|\,\alpha\rangle of π1(H[α])\pi_{1}(H[\alpha]) by setting s=1s=1 in Figure 31. But if s=1s=1 in Figure 31, then the Heegaard diagram underlying the R-R diagram in Figure 31 has a graph with the form of Figure 32. This graph shows α\alpha has minimal length under automorphisms of F(A,B)F(A,B). It follows that if α\alpha is a cyclic conjugate of A2B2A^{2}B^{2} or its inverse in F(A,B)F(A,B), then a+b=1a+b=1. So (a,b)=(1,0)(a,b)=(1,0) or (a,b)=(0,1)(a,b)=(0,1). If (a,b)=(1,0)(a,b)=(1,0), then α=ABA1B1\alpha=ABA^{-1}B^{-1}, which is a commutator, α\alpha separates H\partial H, and α\alpha is not an automorph of (A2B2)±1(A^{2}B^{2})^{\pm 1} in F(A,B)F(A,B). The only remaining possibility is (a,b)=(0,1)(a,b)=(0,1). In this case, Figure 31 reduces to Figure 25, and α=ABsA1Bs\alpha=AB^{s}A^{-1}B^{s} in π1(H)\pi_{1}(H), as desired. ∎

Lemma 5.7.

Suppose α\alpha has an R-R diagram with the form of Figure 31, H[α]H[\alpha] is Seifert-fibered over the Möbius band, and s=1s=1 in Figure 31. Then (a,b)=(0,1)(a,b)=(0,1), Figure 31 reduces to Figure 25, and α=ABA1B\alpha=ABA^{-1}B in π1(H)\pi_{1}(H). Furthermore, the Seifert-fibration of H[α]H[\alpha] has no exceptional fibers.

Proof.

As in Lemma 5.6, the annulus 𝒜\mathcal{A} is vertical in the Seifert-fibration of H[α]H[\alpha], and its boundary components β\beta and β^\hat{\beta} are regular fibers. Then two parallel copies of β\beta bound a separating annulus 𝒜\mathcal{A}^{\prime} in HH such that 𝒜\mathcal{A}^{\prime} is saturated and vertical in the Seifert-fibration of H[α]H[\alpha], and 𝒜\mathcal{A}^{\prime} cuts HH into a solid torus VV and a genus two handlebody WW, with α\alpha lying on W\partial W. Let λ\lambda be the core of VV. (Note that like the s>1s>1 case of Lemma 5.6, λ\lambda is also a core of the B-handle of HH, but unlike the s>1s>1 case, here 𝒜\mathcal{A}^{\prime} is parallel into H\partial H.) Then, since s=1s=1, λ\lambda is a regular fiber in the Seifert-fibration of H[α]H[\alpha]. By performing Dehn surgery on λ\lambda, one can change HH into another genus two handlebody HH^{\prime} such that H[α]H^{\prime}[\alpha] is Seifert-fibered over the Möbius band with λ\lambda as an exceptional fiber of index s>1s^{\prime}>1. And then, since H[α]H^{\prime}[\alpha] is Seifert-fibered over the Möbius band, λ\lambda must be its only exceptional fiber. This implies that the Seifert-fibration of H[α]H[\alpha] had no exceptional fibers.

To finish, notice that the R-R diagram of α\alpha on H\partial H^{\prime} is obtained from the diagram of α\alpha on H\partial H by replacing ss in Figure 31 with ss^{\prime}. Now the argument used in Lemma 5.6 applies and shows that, as before, (a,b)=(0,1)(a,b)=(0,1) and the diagram of Figure 31 reduces to that of Figure 25. ∎

Refer to caption
Figure 32. If s=1s=1 in Figure 31, the Heegaard diagram underlying the R-R diagram in Figure 31 has a graph with this form.

References

  • [B90] Berge, J., Some Knots with Surgeries Yielding Lens Spaces, Unpublished manuscript. Univ. of Texas at Austin, 1990.
  • [B18] Berge, J., Some Knots with Surgeries Yielding Lens Spaces, arXiv preprint arXiv:1802.09722, 2018. original unpublished work from 1990.
  • [B08] Berge, J., The simple closed curves in genus two Heegaard surface of S3S^{3} which are double-primitives, Unpublished manuscript.
  • [BK20] Berge, J. and Kang, S. The hyperbolic primitive/Seifert knots in S3\mathrm{S^{3}}, preprint.
  • [BRZ88] Boileau, M., Rost, M., and Zieschang, H., On Heegaard decompositions of torus knot exteriors and related Seifert fibre spaces. Math. Ann. 279 (1988), 553–581.
  • [CMZ81] Cohen, M., Metzler, W. and Zimmerman, A., What Does a Basis of F(a,b) Look Like?, Math. Ann. 257 (1981), 435–445.
  • [D03] Dean, J., Small Seifert-fibered Dehn surgery on hyperbolic knots, Algebraic and Geometric Topology 3 (2003), 435–472.
  • [EM94] Eudave-Muñoz, M., On non-simple 3-manifolds and 2-handle addition, Topology Appl. 55 no. 2, (1994), 131–152.
  • [G87] Gordon, C. McA., On non-simple 3-manifolds and 2-handle addition, Topology Appl. 27 (1987), 285–299.
  • [G13] Green, J. E., The lens space realization problem, Ann. of Math. (2) 177 (2013), 449–511.
  • [H07] Hatcher, A., Notes on Basic 3-Manifold Topology, Unpublished, available on author’s webpage, http://www.math.cornell.edu/ hatcher.
  • [K20] Kang, S. Tunnel-number-one knot exteriors in S3S^{3} disjoint from proper power curves, preprint.
  • [W36] Whitehead, J. H. C., On equivalent sets of elements in a free group, Ann. of Math. 37 no. 2 (1936), 782–800.
  • [Z77] Zieschang, H., Generators of the Free Product with Amalgamation of two Infinite Cyclic Groups, Math. Ann. 227 no. 3 (1977), 195–221.
  • [Z70] Zieschang, H., On simple systems of paths on complete pretzels, Aemr. Math. Soc. Transl. 92 no. 2 (1970), 127–137.