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

Generalized torsions in once punctured torus bundles

Nozomu Sekino
Abstract

A generalized torsion in a group, an non-trivial element such that some products of its conjugates is the identity. This is an obstruction for a group being bi-orderable. Though it is known that there is a non bi-orderable group without generalized torsions, it is conjectured that 3-manifold groups without generalized torsions are bi-orderable. In this paper, we find generalized torsions in the fundamental groups of once punctured torus bundles which are not bi-orderable. Our result contains a generalized torsion in a tunnel number two hyperbolic once punctured torus bundle.

1 Introduction

A group Γ\Gamma is called a bi-orderable group if it admits total order << which is invariant under multiplication to both sides, i.e. if g<gg<g^{\prime} then fgh<fghfgh<fg^{\prime}h for every f,hΓf,h\in\Gamma. As a convention, the trivial group is regarded as a bi-orderable group. There is one obstruction for groups being bi-orderable, called generalized torsions. An non-trivial element gΓg\in\Gamma is called generalized torsion if some product of conjugates of gg is the identity of Γ\Gamma. In general, there exist groups which are not bi-orderable and have no generalized torsions [2], [3], [10]. However, it is conjectured that bi-orderability and having no generalized torsions is equivalent for the class of 3-manifold groups [8]. We have many affirmative answers for this conjecture, for some geometric manifolds, for some manifolds with non-trivial geometric decompositions, for many link complements and their Dehn fillings [5],[6],[8],[9],[11],[14],[15].

For the complements of fibered links (not necessarily in S3S^{3}), the following important sufficient condition and necessary condition (not necessary and sufficient condition) for the bi-orderability are known.:

Theorem 1.1.

(A consequence of Theorem 2.7 of [13])
Suppose that the isomorphism on the first homology group of the fiber surface induced by the monodromy of a fibered knot has only real positive eigenvalues. Then the fundamental group of the complement of this fibered knot is bi-orderable.

Theorem 1.2.

(A consequence of Theorem 1.4 of [4])
Suppose that the fundamental group of the complement of a fibered knot is bi-orderable. Then the isomorphism on the first homology group of the fiber surface induced by the monodromy of the fibered knot has at least one real positive eigenvalue.

In this paper, we consider once punctured torus bundles, surface-bundles over circles each of whose fiber is torus with connected boundary. In this case, the conditions above are necessary and sufficient condition for the bi-orderability. We will show that if the fundamental group of a once punctured torus bundle is not bi-orderable then it admits a generalized torsion. This adds the class of once punctured torus bundles to the list of 3-manifolds where the conjecture holds.

The rest of this paper is organized as follows. In Section 2, we represent the fundamental groups as oriented based paths and the conjugation classes of the fundamental groups as oriented free loops. We review the relations between theses and some operations. In Section 3, we review that generalized torsions in the fundamental group of fiber-bundles over circles are in that of the fibers and satisfy some relation in terms of monodromies. In Section 4, we assign words and cyclic words to oriented based paths and oriented free loops in once punctured torus, respectively. After these preparations, we prove the main result in Section 5. Thanks to Theorem 1.1 and Theorem 1.2, we know when the fundamental group of a once punctured torus bundle is not bi-ordearble. We will find a generalized torsion in this case. At last, as a remark, we show that our result contains a generalized torsion in the fundamental group of a tunnel number two hyperbolic once punctured torus bundle.

In this paper, we do not distinguish the homotopy classes of paths or loops with their representatives. The words in groups are read from left to right. For a group Γ\Gamma, C(Γ)C(\Gamma) denotes the set of conjugacy classes of Γ\Gamma.

Acknowledgements

The author would like to thank professor Motegi for introducing the area of generalized torsions to him, and giving him many helpful comments.

2 Based paths and free loops

For a topological space XX and a base point X*\in X, the fundamental group π1(X,)\pi_{1}(X,*) is defined as the set of the homotopy types of oriented based paths ([0,1],{0,1})(X,)([0,1],\{0,1\})\longrightarrow(X,*) with multiplications as concatenations. Let F:π1(X,)C(π1(X,))F:\pi_{1}(X,*)\longrightarrow C\left(\pi_{1}(X,*)\right) be a map which sends an element to its conjugacy class. Topologically, C(π1(X,))C\left(\pi_{1}(X,*)\right) is regarded as the set of the homotopy types of oriented free loops S1XS^{1}\longrightarrow X and FF as the operation which identifies the endpoints of paths and forgets the identified special point.

Definition 2.1.

For an oriented free loop ll and a point pp on ll, l{p}l\{p\} denotes the oriented path starting at pp, following ll and ending at pp. Moreover, for another point qq on ll, which is not pp, l{p,q}l\{p,q\} denotes the subpath of ll whose initial endpoint is pp and terminal point is qq. Note that l{p,q}l\{p,q\} and l{q,p}l\{q,p\} are different.

Definition 2.2.

For an oriented path or an oriented free loop II, I¯\bar{I} denotes the oriented path or the oriented free loop obtained by reversing the orientation of II.

Definition 2.3.

For two oriented paths I1I_{1} and I2I_{2} such that the terminal point of I1I_{1} and the initial point of I2I_{2} are identical, I1I2I_{1}-I_{2} denotes the concatenation of I1I_{1} and I2I_{2}.

Definition 2.4.

Let l1,l2C(π1(X,))l_{1},l_{2}\in C\left(\pi_{1}(X,*)\right) be two free loops, and α\alpha an oriented path in XX whose initial endpoint α(0)\alpha(0) is on l1l_{1} and terminal endpoint α(1)\alpha(1) is on l2l_{2}. Then l(l1,α,l2)l\left(l_{1},\alpha,l_{2}\right) denotes a new free loop obtained by forgetting the endpoints of l1{α(0)}αl2{α(1)}α¯l_{1}\{\alpha(0)\}-\alpha-l_{2}\{\alpha(1)\}-\bar{\alpha}. We call l(l1,α,l2)l\left(l_{1},\alpha,l_{2}\right) the loop obtained by connecting l1l_{1} and l2l_{2} using α\alpha.

Lemma 2.1.

Let l1,l2C(π1(X,))l_{1},l_{2}\in C\left(\pi_{1}(X,*)\right) be two free loops, and α\alpha an oriented path in XX whose initial endpoint α(0)\alpha(0) is on l1l_{1} and terminal endpoint α(1)\alpha(1) is on l2l_{2}. Let I1,I2π1(X,)I_{1},I_{2}\in\pi_{1}(X,*) be elements satisfying F(Ii)=liF(I_{i})=l_{i} for i=1,2i=1,2. Then there exists an element gπ1(X,)g\in\pi_{1}(X,*) such that F(I1gI2g1)=l(l1,α,l2)F(I_{1}gI_{2}g^{-1})=l\left(l_{1},\alpha,l_{2}\right).

Proof.

Note that there exist oriented paths βi\beta_{i} starting from * and ending on lil_{i}, denoted by βi(1)\beta_{i}(1), such that βili{βi(1)}βi¯=Ii\beta_{i}-l_{i}\{\beta_{i}(1)\}-\bar{\beta_{i}}=I_{i} for i=1,2i=1,2 (under homotopies). Let gπ1(X,)g\in\pi_{1}(X,*) be β1¯l1¯{β1(1),α(0)}αl2{β2(1),α(1)}¯β2¯\bar{\beta_{1}}-\bar{l_{1}}\{\beta_{1}(1),\alpha(0)\}-\alpha-\overline{l_{2}\{\beta_{2}(1),\alpha(1)\}}-\bar{\beta_{2}}. Then we have I1gI2g1=β1l(l1,α,l2){β1(1)}β1¯I_{1}gI_{2}g^{-1}=\beta_{1}-l\left(l_{1},\alpha,l_{2}\right)\{\beta_{1}(1)\}-\bar{\beta_{1}} and F(I1gI2g1)=l(l1,α,l2)F(I_{1}gI_{2}g^{-1})=l\left(l_{1},\alpha,l_{2}\right). See Figure 1.

Refer to caption
Figure 1: A schematic of the based path gg.

\Box

3 Generalized torsions in a fibered space

In this section, we show that generalized torsions in the fundamental group of a fiber bundle over a circle are same as the elements in the fundamental group of the fiber space satisfying some relations (Proposition 3.1).

Definition 3.1.

(Generalized torsion)
Let Γ\Gamma be a group. A non-trivial element gg is called a generalized torsion if there exist positive integer nn and elements w1,,wnΓw_{1},\dots,w_{n}\in\Gamma such that (w1gw11)(wngwn1)=1Γ(w_{1}g{w_{1}}^{-1})\cdots(w_{n}g{w_{n}^{-1}})=1_{\Gamma}, the identity of Γ\Gamma.

Let (X,)(X,*) be a based topological space and ϕ\phi a self-homeomorphism of XX fixing *. Let (Y,Y)(Y,*_{Y}) be a topological space obtained from X×[0,1]X\times[0,1] by identifying (x,1)(x,1) and (ϕ(x),0)(\phi(x),0) for xXx\in X, and Y*_{Y} comes from (,0)(*,0). In this situation, (Y,Y)(Y,*_{Y}) is called the (X,)(X,*)-bundle over a circle using ϕ\phi and ϕ\phi is called monodromy. Let τπ1(Y,Y)\tau\in\pi_{1}(Y,*_{Y}) be the based path coming from [0,1]X×[0,1];t(,t)[0,1]\longrightarrow X\times[0,1];t\longmapsto(*,t) and let TT be a subgroup of π1(Y,Y)\pi_{1}(Y,*_{Y}) generated by τ\tau. Then it is known that TT is isomorphic to \mathbb{Z}, and that π1(X,)\pi_{1}(X,*) injects in π1(Y,Y)\pi_{1}(Y,*_{Y}). Under these notation, π1(Y,Y)\pi_{1}(Y,*_{Y}) is isomorphic to π1(X,)T\pi_{1}(X,*)\rtimes T, where the multiplication is defined as (g1,τn1)(g2,τn2)=(g1ϕ~n1(g2),τn1+n2)(g_{1},\tau^{n_{1}})\cdot(g_{2},\tau^{n_{2}})=(g_{1}\tilde{\phi}^{n_{1}}(g_{2}),\tau^{n_{1}+n_{2}}), where ϕ~\tilde{\phi} is the induced map between π1(X,)\pi_{1}(X,*) by ϕ\phi. Abusing notations, (g,1T)(g,1_{T}) for gπ1(X,)g\in\pi_{1}(X,*) and (1π1(X,),τn)(1_{\pi_{1}(X,*)},\tau^{n}) are written as gg and τn\tau^{n}, respectively. Note that for gπ1(X,)g\in\pi_{1}(X,*), we have τgτ1=ϕ~(g)\tau g\tau^{-1}=\tilde{\phi}(g) and τ1gτ=ϕ~1(g)\tau^{-1}g\tau=\tilde{\phi}^{-1}(g) in π1(Y,Y)\pi_{1}(Y,*_{Y}).

Suppose that gπ1(Y,Y)g\in\pi_{1}(Y,*_{Y}) is a generalized torsion. Take positive integer nn and w1,,wnπ1(Y,Y)w^{\prime}_{1},\dots,w^{\prime}_{n}\in\pi_{1}(Y,*_{Y}) such that (w1gw11)(wngwn1)=1π1(Y,Y)(w^{\prime}_{1}g{w^{\prime}_{1}}^{-1})\cdots(w^{\prime}_{n}g{w^{\prime}_{n}}^{-1})=1_{\pi_{1}(Y,*_{Y})}. By considering the projection π1(Y,Y)T\pi_{1}(Y,*_{Y})\longrightarrow T, we have gπ1(X,)g\in\pi_{1}(X,*). Represent wiw^{\prime}_{i} as gi1τni1gikτnikg_{i_{1}}\tau^{n_{i_{1}}}\cdots g_{i_{k}}\tau^{n_{i_{k}}}, where gijπ1(X,)g_{i_{j}}\in\pi_{1}(X,*) and nijn_{i_{j}}\in\mathbb{Z} for j=1,kj=1,\dots k. Then (wigwi1)=(j=1kϕ~(l=1jnil)(gij))ϕ~(l=1knil)(g)(j=1kϕ~(l=1jnil)(gij))1(w^{\prime}_{i}g{w^{\prime}_{i}}^{-1})=\left(\prod^{k}_{j=1}\tilde{\phi}^{\left(\sum^{j}_{l=1}n_{i_{l}}\right)}(g_{i_{j}})\right)\tilde{\phi}^{\left(\sum^{k}_{l=1}n_{i_{l}}\right)}(g)\left(\prod^{k}_{j=1}\tilde{\phi}^{\left(\sum^{j}_{l=1}n_{i_{l}}\right)}(g_{i_{j}})\right)^{-1} in π1(Y,Y)\pi_{1}(Y,*_{Y}). Set NiN_{i} to be l=1knil\sum^{k}_{l=1}n_{i_{l}} and wiπ1(X,)w_{i}\in\pi_{1}(X,*) to be (j=1kϕ~(l=1jnil)(gij))\left(\prod^{k}_{j=1}\tilde{\phi}^{\left(\sum^{j}_{l=1}n_{i_{l}}\right)}(g_{i_{j}})\right). Then we have an equation (w1ϕ~N1(g)w11)(wnϕ~Nn(g)wn1)=1π1(Y,Y)\left(w_{1}\tilde{\phi}^{N_{1}}(g){w_{1}}^{-1}\right)\cdots\left(w_{n}\tilde{\phi}^{N_{n}}(g){w_{n}}^{-1}\right)=1_{\pi_{1}(Y,*_{Y})} in π1(Y,Y)\pi_{1}(Y,*_{Y}). Since the left-hand side is a composition of elements in π1(X,)\pi_{1}(X,*) and π1(X,)\pi_{1}(X,*) injects in π1(Y,Y)\pi_{1}(Y,*_{Y}), we have an equation (w1ϕ~N1(g)w11)(wnϕ~Nn(g)wn1)=1π1(X,)\left(w_{1}\tilde{\phi}^{N_{1}}(g){w_{1}}^{-1}\right)\cdots\left(w_{n}\tilde{\phi}^{N_{n}}(g){w_{n}}^{-1}\right)=1_{\pi_{1}(X,*)} in π1(X,)\pi_{1}(X,*).

Conversely, suppose that there exist gπ1(X,)g\in\pi_{1}(X,*), positive integer nn, integers N1,,NnN_{1},\dots,N_{n} and elements w1,wnπ1(X,)w_{1}\dots,w_{n}\in\pi_{1}(X,*) such that (w1ϕ~N1(g)w11)(wnϕ~Nn(g)wn1)=1π1(X,)\left(w_{1}\tilde{\phi}^{N_{1}}(g){w_{1}}^{-1}\right)\cdots\left(w_{n}\tilde{\phi}^{N_{n}}(g){w_{n}}^{-1}\right)=1_{\pi_{1}(X,*)} holds in π1(X,)\pi_{1}(X,*). Then (w1τN1gτN1w11)(wnτNngτNnwn1)=1π1(Y,Y)\left(w_{1}\tau^{N_{1}}g\tau^{-N_{1}}{w_{1}}^{-1}\right)\cdots\left(w_{n}\tau^{N_{n}}g\tau^{-N_{n}}{w_{n}}^{-1}\right)=1_{\pi_{1}(Y,*_{Y})} holds in π1(Y,Y)\pi_{1}(Y,*_{Y}). This implies that gg is a generalized torsion.

We state the above argument as a Proposition:

Proposition 3.1.

Let (Y,Y)(Y,*_{Y}) be the (X,)(X,*)-bundle over a circle using ϕ\phi as the monodromy. An element gπ1(Y,Y)g\in\pi_{1}(Y,*_{Y}) is a generalized torsion if and only if gπ1(X,)g\in\pi_{1}(X,*) and there exist positive integer nn, integers N1,,NnN_{1},\dots,N_{n} and elements w1,,wnπ1(X,)w_{1},\dots,w_{n}\in\pi_{1}(X,*) such that (w1ϕ~N1(g)w11)(wnϕ~Nn(g)wn1)=1π1(X,)\left(w_{1}\tilde{\phi}^{N_{1}}(g){w_{1}}^{-1}\right)\cdots\left(w_{n}\tilde{\phi}^{N_{n}}(g){w_{n}}^{-1}\right)=1_{\pi_{1}(X,*)} holds in π1(X,)\pi_{1}(X,*).

4 Paths and loops in a once punctured torus

Let (Σ,)(\Sigma,*) be a torus with connected boundary with a base point on the boundary. In this section, we review the way to represent oriented based paths, elements of π1(Σ,)\pi_{1}(\Sigma,*), and oriented free loops, elements of C(π1(Σ,))C\left(\pi_{1}(\Sigma,*)\right). Take two disjoint properly embedded oriented arcs IxI_{x} and IyI_{y} such that they are disjoint from * and cut Σ\Sigma into a disk as in Figure 2.

Refer to caption
Figure 2: Oriented arcs IxI_{x}, IyI_{y} on Σ\Sigma.

4.1 Based paths on Σ\Sigma

For an oriented based path Iπ1(Σ,)I\in\pi_{1}(\Sigma,*), we give a word w(I)w(I), which is an element of the free group GG of rank 22 generated by the alphabets {x,y}\{x,y\}, as follows: Take the empty word at first, and follow II. If II hits IxI_{x} from the left (or right) of IxI_{x}, then we add xx (or x1x^{-1}, respectively) to the right of the word we have, and if II hits IyI_{y} from the left (or right) of IyI_{y}, then we add yy (or y1y^{-1}, respectively) to the right of the word we have. Repeat this operation until we reach the terminal point of II. Note that this w()w(\cdot) gives the isomorphism between π1(Σ,)\pi_{1}(\Sigma,*) and GG.

4.2 Free loops on Σ\Sigma

For an oriented free loop lC(π1(Σ,))l\in C\left(\pi_{1}(\Sigma,*)\right), we give a cyclic word W(l)W(l), which is an element of C(G)C(G^{\prime}) the set of the conjugacy classes of the free group GG^{\prime} of rank 22 generated by the alphabets {X,Y}\{X,Y\}, as follows: Take the empty word at first, and take a point l*_{l} on ll and follow l{l}l\{*_{l}\}. If l{l}l\{*_{l}\} hits IxI_{x} from the left (or right) of IxI_{x}, then we add XX (or X1X^{-1}, respectively) to the right of the word we have, and if l{l}l\{*_{l}\} hits IyI_{y} from the left (or right) of IyI_{y}, then we add YY (or Y1Y^{-1}, respectively) to the right of the word we have. Repeat this operation until we reach the terminal point of II. Then connect the last of the word to the first of the word so that we get a cyclic word. Note that this W()W(\cdot) gives the isomorphism as sets between C(π1(Σ,))C\left(\pi_{1}(\Sigma,*)\right) and C(G)C(G^{\prime}). Note also that for an oriented based path Iπ1(Σ,)I\in\pi_{1}(\Sigma,*), W(F(I))W\left(F(I)\right) is obtained from w(I)w(I) by changing x±1x^{\pm 1} and y±1y^{\pm 1} to X±1X^{\pm 1} and Y±1Y^{\pm 1} respectively, and connecting the last and the first of the word. Due to space limitation, we represent cyclic words in lines, and they should be regarded as cyclic.

We give oriented simple loops on Σ\Sigma names as follows, where “simple” means having no self-intersections. Take two oriented simple loops α\alpha and β\beta as in Figure 3.

Refer to caption
Figure 3: Loops α\alpha and β\beta.
Definition 4.1.

Let pp and qq are relative prime integers. Write |p||p| parallel (disjoint) α\alpha loops and give them the same orientation as α\alpha if p0p\geq 0, and the opposite orientations if p<0p<0. And write |q||q| parallel (disjoint) β\beta loops and give them the same orientation as β\beta if q0q\geq 0, and the opposite orientations if q<0q<0. Then resolve the intersection points so that the orientations are compatible. Since pp and qq are relative prime, we get an oriented simple loop. We call this loop L(p,q)L(p,q). See Figure 4 for example. It is well-known that every oriented simple free loop on Σ\Sigma is homotopic to L(p,q)L(p,q) for the unique relative prime integers pp and qq.

Refer to caption
Figure 4: L(3,-2)

5 Once punctured torus bundles

Let (Σ,)(\Sigma,*) be a torus with connected boundary with a base point on the boundary, and take the same arcs IxI_{x} and IyI_{y} as in Figure 2 and the same loops α\alpha and β\beta as in Figure 3. Let (Mϕ,Mϕ)(M_{\phi},*_{M_{\phi}}) be the Σ\Sigma-bundle over a circle with orientation preserving and boundary fixing monodromy ϕ\phi. The 3-manifolds admitting such structures are called once punctured torus bundles, (OPTB for short). The representation matrix under the basis {[α],[β]}\{[\alpha],[\beta]\}, which are the homology classes of α\alpha and β\beta, of the induced isomorphism of H1(Σ;)H_{1}(\Sigma;\mathbb{Z}) by ϕ\phi is denoted by AϕSL2()A_{\phi}\in SL_{2}(\mathbb{Z}). Then the following Fact are known:

Fact 5.1.

(Lemma 3 of [7])
Two Σ\Sigma-bundles over a circle (Mϕ,Mϕ)(M_{\phi},*_{M_{\phi}}) and (Mψ,Mψ)(M_{\psi},*_{M_{\psi}}) are homeomorphic (possibly reversing the orientation) if and only if AϕA_{\phi} is conjugate to AψA_{\psi} or Aψ1A^{-1}_{\psi} in GL2()GL_{2}(\mathbb{Z}).

Since the eigenvalues of ASL2()A\in SL_{2}(\mathbb{Z}) is the roots of x2(trA)x+1=0x^{2}-(trA)x+1=0, Theorem 1.1 implies that π1(Mϕ,Mϕ)\pi_{1}(M_{\phi},*_{M_{\phi}}) is bi-orderable (and thus admits no generalized torsions) if trAϕ2trA_{\phi}\geq 2. Moreover, Theorem 1.2 implies that π1(Mϕ,Mϕ)\pi_{1}(M_{\phi},*_{M_{\phi}}) is not bi-orderable if trAϕ<2trA_{\phi}<2. We want to show that π1(Mϕ,Mϕ)\pi_{1}(M_{\phi},*_{M_{\phi}}) admits a generalized torsion if trAϕ<2trA_{\phi}<2.

Under Fact 5.1, we assume that representing matrix AϕA_{\phi} is of the form of the following:

Fact 5.2.

(Lemma 4 of [7])
Every element of SL2()SL_{2}(\mathbb{Z}) is conjugate in GL2()GL_{2}(\mathbb{Z}) to the matrix AA such that A1,1A2,20A_{1,1}\cdot A_{2,2}\geq 0 and |A1,1||A2,2||A_{1,1}|\geq|A_{2,2}|, where Ai,jA_{i,j} is the (i,j)(i,j)-entry of AA.

Hereafter, we assume that Aϕ=(abcd)A_{\phi}=\left(\begin{array}[]{cc}a&b\\ c&d\\ \end{array}\right) with ad0ad\geq 0 and |a||d||a|\geq|d|. Note that α(=L(1,0))\alpha\left(=L(1,0)\right) is mapped to L(a,c)L(a,c) by ϕ\phi.

Remark 5.1.

For the case of closed torus bundles over circles, the homeomorphism types are classified by the conjugation classes of matrices similarly. Though they are not the complements of fibered knots, almost same arguments for Theorem 1.1 and Theorem 1.2 imply that a closed torus bundle over a circle has bi-orderable fundamental group if and only if the trace of the monodromy matrix is grater than or equal to 22. In the other cases, we can easily find generalized torsions since the fundamental group of the torus is abelian:

  • When a+d=1a+d=1, we assume that Aϕ=(1110)(=(1001)(1110)(1001))A_{\phi}=\left(\begin{array}[]{cc}1&-1\\ 1&0\\ \end{array}\right)\left(=\left(\begin{array}[]{cc}1&0\\ 0&-1\\ \end{array}\right)\left(\begin{array}[]{cc}1&1\\ -1&0\\ \end{array}\right)\left(\begin{array}[]{cc}1&0\\ 0&-1\\ \end{array}\right)\right). Then Aϕ3+E=O{A_{\phi}}^{3}+E=O, where EE and OO are the identity matrix and zero-matrix, and thus every element of the fundamental group of the closed torus is a generalized torsion.

  • When a+d0a+d\leq 0, we have Aϕ2+(ad)Aϕ+E=O{A_{\phi}}^{2}+(-a-d)A_{\phi}+E=O by the Cayley-Hamilton’s equation. Thus every element of the fundamental group of the closed torus is a generalized torsion.

However, in the case of once punctured torus bundles, we should look more carefully since π1(Σ,)\pi_{1}(\Sigma,*) is not abelian.

5.1 The case where trAϕ=1trA_{\phi}=1

We assume that Aϕ=(1110)A_{\phi}=\left(\begin{array}[]{cc}1&-1\\ 1&0\\ \end{array}\right). Then α(=L(1,0))\alpha\left(=L(1,0)\right) is mapped to L(1,0)L(-1,0), α\alpha with the opposite orientation, by ϕ3\phi^{3}. We can connect α\alpha and ϕ3(α)\phi^{3}(\alpha) by an arc γ\gamma as in Figure 5. Note that W(l(α,γ,ϕ3(α)))W\left(l\left(\alpha,\gamma,\phi^{3}(\alpha)\right)\right) is the empty word. Fix an oriented based path Iαπ1(Σ,)I_{\alpha}\in\pi_{1}(\Sigma,*) such that F(Iα)=αF(I_{\alpha})=\alpha. Then W(ϕ3(Iα))=ϕ3(α)W\left(\phi^{3}(I_{\alpha})\right)=\phi^{3}(\alpha) also holds. By Lemma 2.1, there exists an element gπ1(Σ,)g\in\pi_{1}(\Sigma,*) such that W(F(Iαgϕ3(Iα)g1))W\left(F\left(I_{\alpha}g\phi^{3}(I_{\alpha})g^{-1}\right)\right) is the empty word in C(G)C(G^{\prime}), and this implies that w(Iαgϕ3(Iα)g1)w\left(I_{\alpha}g\phi^{3}(I_{\alpha})g^{-1}\right) is the empty word in GG. Thus Iαgϕ3(Iα)g1=1π1(Σ,)I_{\alpha}g\phi^{3}(I_{\alpha})g^{-1}=1_{\pi_{1}(\Sigma,*)} holds, and we see that IαI_{\alpha} is a generalized torsion.

Refer to caption
Figure 5: γ\gamma and l(α,γ,ϕ3(α))l\left(\alpha,\gamma,\phi^{3}(\alpha)\right).

5.2 The case where trAϕ0trA_{\phi}\leq 0

In this case, α(=L(1,0))\alpha\left(=L(1,0)\right) is mapped to L(a,c)L(a,c) by ϕ\phi and to L(d,c)L(d,-c) by ϕ1\phi^{-1}. Note that ad0a\leq d\leq 0 since ad0ad\geq 0 and |a||d||a|\geq|d|. If a=0a=0, then ϕ2\phi^{2} reverses the orientation of α\alpha. Thus we get a generalized torsion as in Subsection 5.1. In the following, we assume that a<0a<0. W(L(a,c))W\left(L(a,c)\right) consists of |a||a|-times X1X^{-1} and |c||c|-times Ysign(c)Y^{sign(c)}, W(α)W(\alpha) is XX, and W(L(d,c))W\left(L(d,-c)\right) consists of |d||d|-times X1X^{-1} and |c||c|-times Ysign(c)Y^{-sign(c)}, where sign(c)sign(c) is +1+1 if c0c\geq 0 and 1-1 otherwise. Along L(a,c)L(a,c), take |a||a| points 1,,|a|*_{1},\dots,*_{|a|} just after intersection points with IxI_{x}. Take |a||a|-copies α1,,α|a|\alpha_{1},\dots,\alpha_{|a|} of α(=L(1,0))\alpha\left(=L(1,0)\right) and arcs γ1,,γ|a|\gamma_{1},\dots,\gamma_{|a|} which are disjoint from IxI_{x} and IyI_{y} so that γi\gamma_{i} connects i*_{i} and γi\gamma_{i} for each i=1,,|a|i=1,\dots,|a|. See Figure 6. Fix an oriented based path Iαπ1(Σ,)I_{\alpha}\in\pi_{1}(\Sigma,*) such that F(Iα)=αF(I_{\alpha})=\alpha. By using Lemma 2.1 repeatedly, we see that there exist elements g1,,g|a|π1(Σ,)g_{1},\dots,g_{|a|}\in\pi_{1}(\Sigma,*) such that W(F(ϕ(Iα)(g1Iαg11))(g|a|Iαg|a|1)))=YcW\left(F\left(\phi(I_{\alpha})\cdot(g_{1}I_{\alpha}{g_{1}}^{-1}))\cdot\cdots\cdot(g_{|a|}I_{\alpha}{g_{|a|}}^{-1})\right)\right)=Y^{c}.

Refer to caption
Figure 6: 1,,|a|*_{1},\dots,*_{|a|} and α1,,α|a|\alpha_{1},\dots,\alpha_{|a|} and γ1,,γ|a|\gamma_{1},\dots,\gamma_{|a|}.

If d=0d=0, W(ϕ1(α))=YcW\left(\phi^{-1}(\alpha)\right)=Y^{-c}. Then by using Lemma 2.1, choosing connecting arc which is disjoint from IxI_{x} and IyI_{y}, we see that there exists an element gπ1(Σ,)g\in\pi_{1}(\Sigma,*) such that
W(F(ϕ(Iα)(g1Iαg11))(g|a|Iαg|a|1)(gϕ1(Iα)g1)))W\left(F\left(\phi(I_{\alpha})\cdot(g_{1}I_{\alpha}{g_{1}}^{-1}))\cdot\cdots\cdot(g_{|a|}I_{\alpha}{g_{|a|}}^{-1})(g\phi^{-1}(I_{\alpha})g^{-1})\right)\right) is the empty word in C(G)C(G^{\prime}), and this implies that w((ϕ(Iα)(g1Iαg11))(g|a|Iαg|a|1)(gϕ1(Iα)g1)))w\left(\left(\phi(I_{\alpha})\cdot(g_{1}I_{\alpha}{g_{1}}^{-1}))\cdot\cdots\cdot(g_{|a|}I_{\alpha}{g_{|a|}}^{-1})(g\phi^{-1}(I_{\alpha})g^{-1})\right)\right) is the empty word in GG. Thus (ϕ(Iα)(g1Iαg11))(g|a|Iαg|a|1)(gϕ1(Iα)g1))=1π1(Σ,)\left(\phi(I_{\alpha})\cdot(g_{1}I_{\alpha}{g_{1}}^{-1}))\cdot\cdots\cdot(g_{|a|}I_{\alpha}{g_{|a|}}^{-1})(g\phi^{-1}(I_{\alpha})g^{-1})\right)=1_{\pi_{1}(\Sigma,*)} holds, and we see that IαI_{\alpha} is a generalized torsion.
If d<0d<0, like ϕ(α)\phi(\alpha), there exist elements g1,,g|d|π1(Σ,)g^{\prime}_{1},\dots,g^{\prime}_{|d|}\in\pi_{1}(\Sigma,*) such that
W(F(ϕ1(Iα)(g1Iαg11))(g|d|Iαg|d|1)))=YcW\left(F\left(\phi^{-1}(I_{\alpha})\cdot(g^{\prime}_{1}I_{\alpha}{g^{\prime}_{1}}^{-1}))\cdot\cdots\cdot(g^{\prime}_{|d|}I_{\alpha}{g^{\prime}_{|d|}}^{-1})\right)\right)=Y^{-c}. Then by Lemma 2.1, choosing connecting arc which is disjoint from IxI_{x} and IyI_{y}, we see that there exists an element g~π1(Σ,)\tilde{g}\in\pi_{1}(\Sigma,*) such that
W(F(ϕ(Iα)(g1Iαg11)(g|a|Iαg|a|1)g~(ϕ1(Iα)(g1Iαg11)(g|d|Iαg|d|1))g~1))W\left(F\left(\phi(I_{\alpha})\cdot(g_{1}I_{\alpha}{g_{1}}^{-1})\cdot\cdots\cdot(g_{|a|}I_{\alpha}{g_{|a|}}^{-1})\cdot\tilde{g}\left(\phi^{-1}(I_{\alpha})\cdot(g^{\prime}_{1}I_{\alpha}{g^{\prime}_{1}}^{-1})\cdot\cdots\cdot(g^{\prime}_{|d|}I_{\alpha}{g^{\prime}_{|d|}}^{-1})\right)\tilde{g}^{-1}\right)\right) is the empty word in C(G)C(G^{\prime}), and this implies that
w(ϕ(Iα)(g1Iαg11)(g|a|Iαg|a|1)g~(ϕ1(Iα)(g1Iαg11)(g|d|Iαg|d|1))g~1)w\left(\phi(I_{\alpha})\cdot(g_{1}I_{\alpha}{g_{1}}^{-1})\cdot\cdots\cdot(g_{|a|}I_{\alpha}{g_{|a|}}^{-1})\cdot\tilde{g}\left(\phi^{-1}(I_{\alpha})\cdot(g^{\prime}_{1}I_{\alpha}{g^{\prime}_{1}}^{-1})\cdot\cdots\cdot(g^{\prime}_{|d|}I_{\alpha}{g^{\prime}_{|d|}}^{-1})\right)\tilde{g}^{-1}\right) is the empty word in GG.
Thus ϕ(Iα)(g1Iαg11)(g|a|Iαg|a|1)g~(ϕ1(Iα)(g1Iαg11)(g|d|Iαg|d|1))g~1=1π1(Σ,)\phi(I_{\alpha})\cdot(g_{1}I_{\alpha}{g_{1}}^{-1})\cdot\cdots\cdot(g_{|a|}I_{\alpha}{g_{|a|}}^{-1})\cdot\tilde{g}\left(\phi^{-1}(I_{\alpha})\cdot(g^{\prime}_{1}I_{\alpha}{g^{\prime}_{1}}^{-1})\cdot\cdots\cdot(g^{\prime}_{|d|}I_{\alpha}{g^{\prime}_{|d|}}^{-1})\right)\tilde{g}^{-1}=1_{\pi_{1}(\Sigma,*)} holds, and we see that IαI_{\alpha} is a generalized torsion.

Remark 5.2.

(This remark is suggested by professor Motegi)
In general, the tunnel number of every once punctured torus bundle is less than or equal to 22, where the tunnel number is the minimal number of properly embedded arcs such that the complement is the handlebody. We can choose arcs IxI_{x} and IyI_{y} in one fiber for the complement being a handlebody, for example. In [1], tunnel number one once punctured torus bundles are completely determined. They said that the tunnel number of a once punctured torus bundle is one if and only if there is a simple free loop on a fiber such that this loop and the image of this loop under the monodromy intersect once. The monodromy matrix of such a once punctured torus bundle is (m110)\left(\begin{array}[]{cc}m&1\\ -1&0\\ \end{array}\right) for some mm\in\mathbb{Z} under appropriate basis. In our result, we can find a tunnel number two once punctured torus bundle whose fundamental group has a generalized torsion: For example, consider a once punctured torus bundle whose monodromy matrix is A=(4n12n21)A=\left(\begin{array}[]{cc}4n-1&-2n\\ 2&-1\\ \end{array}\right) for some n0n\leq 0. This has a generalized torsion since the trace is 4n214n-2\leq 1. Suppose for a contradiction that the tunnel number of this is one. Then AA is conjugate in GL2()GL_{2}(\mathbb{Z}) to B=(4n2110)B=\left(\begin{array}[]{cc}4n-2&1\\ -1&0\\ \end{array}\right) or B1B^{-1}. Then modulo 22, every element conjugate to AA in GL2()GL_{2}(\mathbb{Z}) is congruent to (1001)\left(\begin{array}[]{cc}1&0\\ 0&1\\ \end{array}\right). However, BB and B1B^{-1} are congruent to (0110)\left(\begin{array}[]{cc}0&1\\ 1&0\\ \end{array}\right) modulo 22, this leads a contradiction.

Moreover, if we suppose n1n\leq-1 in addition, then the monodromy is pseudo-Anosov (not reducible and not periodic). Then by [12] [16], corresponding once punctured torus bundle admits a complete hyperbolic structure of finite volume.

References

  • [1] K. L. Baker, J. E. Johnson, and E. A. Klodginski, Tunnel number one, genus one fibered knots, arXiv:math/0606379v2.
  • [2] V. V. Bludov, An example of an unordered group with strictly isolated identity element, Algebra and Logic 11 (1972), 341–349.
  • [3] V. V. Bludov, and E. S. Lapshina, On ordering groups with a nilpotent commutant (Russian), Sibirsk. Mat. Zh. 44 (2003), no.3, 513–520; translation in Siberian Math. J. 44 (2003), no.3, 405–410.
  • [4] A. Clay, and D. Rolfsen, Ordered groups, eigenvalues, knots, surgery and L-spaces, Math. Proc. Camb. Phil. Soc. 152 (2012), 115–129.
  • [5] T. Ito, K. Motegi, and M. Teragaito, Generalized torsion and decomposition of 3-manifolds, Proc. Amer. Math. Soc. 147 (2019) 4999–5008.
  • [6] T. Ito, K. Motegi, and M. Teragaito, Generalized torsion and Dehn filling, Topology and its applications. Vol. 301 (2021), 107515.
  • [7] K. Morimoto, Genus one fibered knots in lens spaces. J. Math. Soc. Japan, 41(1):81–96, 1989.
  • [8] K. Motegi, and M. Teragaito, Generalized torsion elements and bi-orderability of 3-manifold groups, Canad. Math. Bull. 60 (2017), 830–844.
  • [9] K. Motegi, and M. Teragaito, Generalized torsion for knots with arbitrarily high genus, arXiv:2106.14449.
  • [10] R. Mura, and A. Rhemtulla, Orderable groups, Lecture Notes in Pure and Applied Mathematics, 27. Marcel Dekker, Inc., New York-Basel, 1977.
  • [11] G. Naylor, and D. Rolfsen, Generalized torsion in knot groups, Canad. Math. Bull. 59 (2016), 182–189.
  • [12] J. P. Otal, The hyperbolization theorem for fibered 3-manifolds, Translated from the 1996 French original by Leslie D. Kay. SMF/AMS Texts and Monographs, vol 7, American Mathematical Society, 2001.
  • [13] B. Perron, and D. Rolfsen, On orderability of fibered knot groups, Math. Proc. Camb. Phil. Soc. 135 (2003), 147–153.
  • [14] M. Teragaito, Generalized torsion elements in the knot groups of twist knots, Proc. Amer. Math. Soc. 144 (2016), no.6, 2677–2682.
  • [15] M. Teragaito, Generalized torsion elements and hyperbolic links, J. Knot Theory Ramifications 29 (2020), no.11, 2050079 (9 pages).
  • [16] W. P. Thurston, Hyperbolic Structures on 3-manifolds, Topology Appl. 52 (1993), 127–149.

GRADUATE SCHOOL OF MATHEMATICAL SCIENCES, THE UNIVERSITY OF TOKYO, 3-8-1 KOMABA, MEGURO–KU, TOKYO, 153-8914, JAPAN
  E-mail address: [email protected]