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

S-stable foliations on flow-spines with transverse Reeb flow

Shin Handa Mathematical Institute, Tohoku University, Sendai, 980-8578, Japan [email protected]  and  Masaharu Ishikawa Department of Mathematics, Hiyoshi Campus, Keio University, 4-1-1 Hiyoshi, Kohoku, Yokohama 223-8521, Japan [email protected]
Abstract.

The notion of S-stability of foliations on branched simple polyhedrons is introduced by R. Benedetti and C. Petronio in the study of characteristic foliations of contact structures on 33-manifolds. We additionally assume that the 11-form β\beta defining a foliation on a branched simple polyhedron PP satisfies dβ>0d\beta>0, which means that the foliation is a characteristic foliation of a contact form whose Reeb flow is transverse to PP. In this paper, we show that if there exists a 11-form β\beta on PP with dβ>0d\beta>0 then we can find a 11-form with the same property and additionally being S-stable. We then prove that the number of simple tangency points of an S-stable foliation on a positive or negative flow-spine is at least 22 and give a recipe for constructing a characteristic foliation of a 11-form β\beta with dβ>0d\beta>0 on the abalone.

1. Introduction

A flow-spine PP is a branched simple spine embedded in an oriented, closed, smooth 3-manifold MM such that there exists a non-singular flow in MM that is transverse to PP and “constant” in the complement MPM\setminus P. This notion was introduced by I. Ishii in [8]. In that paper, he proved that any non-singular flow in MM has a flow-spine. Therefore, regarding a Reeb flow on a contact 3-manifold as a flow of its flow-spine, we may use it for studying contact structures on 3-manifolds. This setting is analogous to the setting of the correspondence between contact 33-manifolds and open book decompositions in [11, 4]. One of the advantages of this setting is that the contact structure in the complement of a flow-spine is always tight since the Reeb flow is “constant”. Thus the study of contact structures via flow-spines divides into two parts, one is the study of contact structures in small neighborhoods of flow-spines and the other is to see what happens by gluing a tight 3-ball to the neighborhoods.

A characteristic foliation on a branched polyhedron embedded in a contact 3-manifold had been studied by Benedetti and Petronio in [1], following the work of Giroux on characteristic foliations on surfaces [3]. Setting the branched polyhedron in a general position, we may assume that the foliation is non-singular on the singular set S(P)S(P) of PP. We further assume that the indices of simple tangency points of the foliation to S(P)S(P) are always +1+1 and away from vertices of S(P)S(P). See Figure 1 for the definition of the index of a simple tangency point. A foliation that satisfies the above conditions is called an S-stable foliation. In [1], they proved several statements. For instance, it is proved that if a characteristic foliation on a branched polyhedron PP in a contact 33-manifold MM is S-stable then the contact structure with this foliation is unique in a small neighborhood of PP up to contactomorphism. It is also proved that if that contact structure is tight in a neighborhood of PP then it extends to a tight contact structure on MM and the extended contact structure on MM is unique up to contactomorphism. Remark that the Reeb flows of these contact structures may not be transverse to PP. In this paper, we always assume that a contact structure is positive, that is, its contact form α\alpha satisfies αdα>0\alpha\land d\alpha>0.

Refer to caption
Figure 1. The index of simple tangency points.

Suppose that there exists a contact form α\alpha on MM whose Reeb flow is transverse to a flow-spine PP. We choose the orientations of the regions of PP so that their intersections with the Reeb flow are positive. In this setting, the form β=α|P\beta=\alpha|_{P} on PP satisfies dβ>0d\beta>0. To advance the study in this setting, we need to study a 11-form β\beta on PP with dβ>0d\beta>0 and whose characteristic foliation on PP is S-stable. Note that there are many branched simple polyhedrons that admit a 11-form β\beta with dβ>0d\beta>0, see Remark 3.2.

The aim of this paper is to understand if there is a 11-form β\beta on PP with dβ>0d\beta>0 and whose kernel gives an S-stable foliation on PP and if there exists a constraint for positions of leaves of S-stable foliations on PP. The following theorem answers the first question.

Theorem 1.1.

Let PP be a branched simple polyhedron. If there exists a 11-form β\beta on PP with dβ>0d\beta>0 then there exists a 11-form β\beta^{\prime} on PP such that dβ>0d\beta^{\prime}>0 and the foliation defined by β=0\beta^{\prime}=0 on PP is S-stable.

Note that it is proved in [1] that any characteristic foliation on a branched simple polyhedron PP in a contact 33-manifold MM can be made to be S-stable by C0C^{0}-perturbation of PP in MM. In our claim, there is no direct relation between β\beta and β\beta^{\prime}.

Our construction of an S-stable foliation is very efficient in the sense that the number of simple tangency points is very small (at most twice the number of triple lines). On the other hand, it is difficult to find an S-stable foliation defined by a 11-form β\beta with dβ>0d\beta>0 and without simple tangency points. Actually, in Theorem 1.2 below, we show that such a foliation does not exist if a branched simple polyhedron is a flow-spine and satisfies a certain condition. A region of a branched simple polyhedron is called a preferred region if the orientations of all edges and circles on its boundary are opposite to the one induced from the orientation of the region defined by the branching of PP. Note that the number of simple tangency points is always even, see Lemma 4.2.

Theorem 1.2.

If a flow-spine PP has a preferred region then any foliation on PP defined by a 11-form β\beta with dβ>0d\beta>0 has at least two simple tangency points.

A point on a simple polyhedron that has a neighborhood shown in Figure 2 (iii) is called a vertex. A branched simple polyhedron has two kinds of vertices: the vertex shown on the middle in Figure 3 is called a vertex of \ell-type and the one on the right is called of rr-type. A flow-spine PP is said to be positive if it has at least one vertex and all vertices are of \ell-type. In [9] it is shown that the map sending a positive flow-spine PP to the contact structure whose Reeb flow is a flow of PP gives a surjection from the set of positive flow-spines to the set of contact 33-manifolds up to contactomorphism. It is also proved that we cannot expect the same result without restricting the source to the set of positive flow-spines. Thus, the positivity is important when we study contact 33-manifolds using flow-spines. We say that a flow-spine is negative if it has at least one vertex and all vertices are of rr-type.

If a flow-spine is either positive or negative then it always has a preferred region. Hence the following corollary holds.

Corollary 1.3.

Let PP be a positive or negative flow-spine of a closed, oriented, smooth 33-manifold MM. If the Reeb flow of a contact form α\alpha on MM is a flow of PP then the characteristic foliation of kerα\ker\alpha on PP has at least two simple tangency points.

Remark that if PP is a positive flow-spine then there exists a contact form α\alpha on MM whose Reeb flow is a flow of PP and such a contact structure is unique up to contactomorphism, which is proved in [9, Theorem 1.1]. On the other hand, if PP is a negative flow-spine, we do not know if there exists such a contact form or not.

As we mentioned, our second aim is to understand if there is a constraint for S-stable characteristic foliations in our setting, and the above corollary gives some insight into this question. Furthermore, in Section 6.2, we will give an example of an S-stable foliation given by a 11-form β\beta with dβ>0d\beta>0 on the abalone explicitly, in which we can see that there exists a constraint for the positions of leaves of S-stable characteristic foliations, see Remark 6.2. We will also give a branched standard spine that admits an S-stable foliation defined by β=0\beta=0 with dβ>0d\beta>0 and without simple tangency points, see Section 6.3. We do not know if there exists a flow-spine that admits an S-stable foliation defined by a 11-form β\beta with dβ>0d\beta>0 and without simple tangency points.

This paper is organized as follows. In Section 2, we recall some terminologies of polyhedrons that we use in this paper. Section 3 and Section 4 are devoted to the proofs of Theorems 1.1 and 1.2, respectively. In Section 5, we shortly introduce the DS-diagram of a flow-spine and give the proof of Corollary 1.3. In Section 6, after giving a Poincaré-Hopf lemma for flow-spines, we give an example of an S-stable foliation given by a 11-form β\beta with dβ>0d\beta>0 on the abalone explicitly. We also give an example of a branched standard spine that admits an S-stable foliation given by a 11-form β\beta with dβ>0d\beta>0 and without simple tangency points.

We would like to thank Ippei Ishii for precious comments and especially for telling us the 33-manifold of the spine in Figure 13. We are also grateful to Yuya Koda and Hironobu Naoe for useful conversation. Finally, we thank the anonymous referee for insightful comments on improving the paper. This work is partially supported by JSPS KAKENHI Grant Number JP17H06128. The second author is supported by JSPS KAKENHI Grant Numbers JP19K03499 and Keio University Academic Development Funds for Individual Research.

2. Preliminaries

In this section, we recall terminologies of polyhedrons used in this paper.

2.1. Simple polyhedron

A polyhedron PP is said to be simple if every point on PP has a neighborhood represented by one of the models shown in Figure 2. Each connected component of the set of points with the model (i) is called a region, that with the model (ii) is called a triple line and that with the model (iii) is called a true vertex, which we call a vertex for short. Let S(P)S(P) denote the union of triple lines and vertices of PP, which is called the singular set of PP. A triple line is either an open arc, called an edge, or a circle. A simple polyhedron PP is said to be standard111It is also called a special polyhedron. if every connected component of PS(P)P\setminus S(P) is homeomorphic to an open disk and every triple line is an open arc.

Refer to caption
Figure 2. The local models of a simple polyhedron.

Let PP be a simple polyhedron embedded in an oriented, 33-manifold MM and assume that each region of PP is orientable. An assignment of orientations to the regions of PP such that for any triple line the three orientations induced from those of the adjacent regions do not coincide is called a branching. A simple polyhedron equipped with a branching is called a branched polyhedron. If a branched polyhedron is standard then it is called a branched standard polyhedron. For a branched polyhedron PP, we define the orientation of each triple line of S(P)S(P) by the orientation induced from the two adjacent regions, see Figure 3.

Refer to caption
Figure 3. The orientation of a triple line induced from the branching.

A region of a branched simple polyhedron is called a preferred region if the orientations of all edges on its boundary are opposite to the one induced from the orientation of the region defined by the branching of PP.

2.2. Spine and flow-spine

Let MM be a closed, connected, oriented 33-manifold and PP be a simple polyhedron embedded in MM. The polyhedron PP is called a spine of MM if MInt BM\setminus\text{Int\,}B collapses to PP, where BB is a 33-ball in MPM\setminus P. If a spine is standard (resp. branched) then it is called a standard (resp. branched) spine.

The singular set S(P)S(P) of a branched simple polyhedron PP can be regarded as the image of an immersion of a finite number of circles. The immersion has only normal crossings as shown in Figure 3. A flow-spine is defined from a non-singular flow in a closed, connected, oriented, smooth 33-manifold and a disk DD intersecting all orbits of the flow transversely by floating the boundary of DD smoothly until it arrives in the disk DD itself. See [8] for the precise definition (cf. [9]). By the construction, the flow is positively transverse to the flow-spine. We can easily see that a branched simple polyhedron PP is a flow-spine if and only if S(P)S(P) is the image of an immersion of one circle.

2.3. Admissibility condition

Let PP be a branched simple polyhedron. To have a 11-form β\beta on PP with dβ>0d\beta>0, PP need satisfy the following admissibility condition. Let R1,,RnR_{1},\ldots,R_{n} be the regions of PP. Regions and triple lines of PP are oriented by the branching. Let R¯i\bar{R}_{i} be the metric completion of RiR_{i} with the path metric inherited from a Riemannian metric on RiR_{i}. Let κi:R¯iP\kappa_{i}:\bar{R}_{i}\to P be the natural extension of the inclusion RiPR_{i}\to P.

Definition.

A branched simple polyhedron PP is said to be admissible if there exists an assignment of real numbers x1,,xmx_{1},\ldots,x_{m} to the triple lines e1,,eme_{1},\ldots,e_{m}, respectively, of PP such that for any i{1,,n}i\in\{1,\ldots,n\}

(2.1) e~jR¯iεijxj>0,\sum_{\tilde{e}_{j}\subset\partial\bar{R}_{i}}\varepsilon_{ij}x_{j}>0,

where e~j\tilde{e}_{j} is an open arc or a circle on R¯i\partial\bar{R}_{i} such that κi|e~j:e~jej\kappa_{i}|_{\tilde{e}_{j}}:\tilde{e}_{j}\to e_{j} is a homeomorphism, and εij=1\varepsilon_{ij}=1 if the orientation of eje_{j} coincides with that of κi(e~j)\kappa_{i}(\tilde{e}_{j}) induced from the orientation of RiR_{i} and εij=1\varepsilon_{ij}=-1 otherwise.

3. Proof of Theorem 1.1

Theorem 1.1 will follow from the following proposition.

Proposition 3.1.

Let PP be a branched simple polyhedron. Then, there exists a 11-form β\beta on PP with dβ>0d\beta>0 such that the foliation on PP defined by β=0\beta=0 is S-stable if and only if PP is admissible.

Remark 3.2.

It is known that many flow-spines satisfy the admissibility condition, see [9, Section 4.1]. For instance, a branched standard spine in a rational homology 33-sphere is admissible. Such branched simple polyhedrons always satisfy the condition in Theorem 1.1.

Let PP be a branched simple polyhedron, QQ be a small compact neighborhood of S(P)S(P) in PP, R1,,RnR_{1}^{\prime},\ldots,R_{n}^{\prime} be connected components of PInt QP\setminus\text{Int\,}Q and e1,,eme_{1},\ldots,e_{m} be triple lines of PP. The orientation of RiR_{i}^{\prime} is defined from the branching of PP and that of eje_{j} is defined as explained in Section 2.1.

Lemma 3.3.

Suppose that PP is admissible. Then there exists a 11-form β0\beta_{0} on QQ such that

  • (1)

    the foliation on QQ defined by β0=0\beta_{0}=0 is S-stable,

  • (2)

    Riβ0>0\int_{\partial R_{i}^{\prime}}\beta_{0}>0 for i=1,,ni=1,\ldots,n, and

  • (3)

    dβ0>0d\beta_{0}>0 on QQ.

Proof.

Let v1,,vnvv_{1},\ldots,v_{n_{v}} be the vertices of PP and NvjN_{v_{j}} be a small compact neighborhood of vjv_{j} in PP. For each j=1,,nvj=1,\ldots,n_{v}, we define a projection prvj\mathrm{pr}_{v_{j}} from NvjN_{v_{j}} to 2\mathbb{R}^{2} such that

  • (i)

    the orientation of each region defined by the branching coincides with that of 2\mathbb{R}^{2},

  • (ii)

    the image is included in a small open disk centered at the point (cos2πjnv,sin2πjnv)\left(\cos\frac{2\pi j}{n_{v}},\sin\frac{2\pi j}{n_{v}}\right), and

  • (iii)

    the orientations of the images of the triple lines in NvjN_{v_{j}} are counter-clockwise.

See Figure 4. The arrowed edges in the figure are the images of triple lines of PP, which are oriented according to the branching of PP as shown in Figure 3.

Refer to caption
Figure 4. The images of Nv1,,NvnvN_{v_{1}},\ldots,N_{v_{n_{v}}} on 2\mathbb{R}^{2}.

Let (r,θ)(r,\theta) be the polar coordinates of 2\mathbb{R}^{2}, set a 11-form on 2\mathbb{R}^{2} as r2dθr^{2}d\theta and define the 11-form β0\beta_{0} on NvjN_{v_{j}} by prvj(r2dθ)\mathrm{pr}_{v_{j}}^{*}(r^{2}d\theta). Note that dβ0>0d\beta_{0}>0 on NviN_{v_{i}} since d(r2dθ)=2rdrdθ>0d(r^{2}d\theta)=2rdr\land d\theta>0.

Next, we define the 11-form β0\beta_{0} on a neighborhood of each triple line of PP. Let e1,,eme_{1},\ldots,e_{m} be the triple lines of PP and NejN_{e_{j}}, j=1,,mj=1,\ldots,m, be a small compact neighborhood of eje_{j} in PP and set Nej=NejInt(Nv(ej)Nv(ej))N_{e_{j}}^{\prime}=N_{e_{j}}\setminus\mathrm{Int}\,(N_{v(e_{j})}\cup N_{v^{\prime}(e_{j})}), where v(ej)v(e_{j}) and v(ej)v^{\prime}(e_{j}) are the two vertices at the endpoints of eje_{j}. Since PP is assumed to be admissible, we can choose an mm-tuple of real numbers (E1,,Em)(E_{1},\ldots,E_{m}) in C(P)C(P).

Suppose that eje_{j} is an edge. If Ej>0E_{j}>0, we choose a projection prej\mathrm{pr}_{e_{j}} from NejN^{\prime}_{e_{j}} to 2\mathbb{R}^{2} as shown on the left in Figure 5 and define the 11-form β0\beta_{0} on NejN^{\prime}_{e_{j}} by prej(r2dθ)\mathrm{pr}_{e_{j}}^{*}(r^{2}d\theta) as before. Set ej=ejNeje_{j}^{\prime}=e_{j}\cap N^{\prime}_{e_{j}}. Let aja_{j} and aja^{\prime}_{j} be the endpoints of prej(ej)\mathrm{pr}_{e_{j}}(e_{j}^{\prime}) and let ej\ell_{e_{j}} and ej\ell_{e_{j}}^{\prime} be the line segments on 2\mathbb{R}^{2} connecting the origin and aja_{j} and the origin and aja_{j}^{\prime}, respectively. Since these segments are in the kernel of r2dθr^{2}d\theta, the integrations of r2θr^{2}\theta along ej\ell_{e_{j}} and ej\ell_{e_{j}}^{\prime} are 0. By Stokes’ theorem, the absolute value of the integration of β0\beta_{0} along eje_{j}^{\prime} coincides with the area of the region bounded by prej(ej)\mathrm{pr}_{e_{j}}(e^{\prime}_{j}), ej\ell_{e_{j}} and ej\ell_{e_{j}}^{\prime}. We may choose prej\mathrm{pr}_{e_{j}} so that the area becomes the given real number Ej>0E_{j}>0, which means that the integration of β0\beta_{0} along eje^{\prime}_{j} coincides with EjE_{j}. Note that the foliation defined by β0=0\beta_{0}=0 has no simple tangency point on NejN_{e_{j}}^{\prime}.

Refer to caption
Figure 5. The images of NejN^{\prime}_{e_{j}} on 2\mathbb{R}^{2} in the case Ej>0E_{j}>0 and Ej<0E_{j}<0.

If Ej<0E_{j}<0, we choose a projection prej\mathrm{pr}_{e_{j}} from NejN^{\prime}_{e_{j}} to \mathbb{R} as shown on the right in Figure 5 and define the 11-form β0\beta_{0} on NejN^{\prime}_{e_{j}} by prej(r2dθ)\mathrm{pr}_{e_{j}}^{*}(r^{2}d\theta) as before. By the same observation as in the case Ej>0E_{j}>0, we may choose prej\mathrm{pr}_{e_{j}} so that the integration of β0\beta_{0} along eje^{\prime}_{j} coincides with EiE_{i}. For each edge eje_{j} with Ej<0E_{j}<0, the foliation defined by β0=0\beta_{0}=0 has exactly two simple tangency points and their indices are +1+1.

If Ej=0E_{j}=0, we choose the projection prej\mathrm{pr}_{e_{j}} as shown in Figure 6. In the figure, prej\mathrm{pr}_{e_{j}} is chosen in such a way that the area of the region Dej+D_{e_{j}}^{+} bounded by prej(ej)\mathrm{pr}_{e_{j}}(e^{\prime}_{j}), ej\ell_{e_{j}} and ej\ell_{e_{j}}^{\prime} is equal to the area of the region DejD_{e_{j}}^{-} bounded by prej(ej)\mathrm{pr}_{e_{j}}(e_{j}^{\prime}). Then we have

Ej=prej(ej)r2𝑑θ=Dej+2r𝑑rdθDej2r𝑑rdθ=0.E_{j}=\int_{\mathrm{pr}_{e_{j}}(e^{\prime}_{j})}r^{2}d\theta=\int_{D_{e_{j}}^{+}}2rdr\land d\theta-\int_{D_{e_{j}}^{-}}2rdr\land d\theta=0.

Two simple tangency points of index +1+1 appear in this case.

Refer to caption
Figure 6. The image of NejN^{\prime}_{e_{j}} on 2\mathbb{R}^{2} in the case Ej=0E_{j}=0.

If eje_{j} is a circle and Ej0E_{j}\neq 0 then we embed NejN_{e_{j}} into 2\mathbb{R}^{2} along a circle centered at the origin and define β0\beta_{0} by prej(r2dθ)\mathrm{pr}_{e_{j}}^{*}(r^{2}d\theta) as before. We choose the embedding such that the orientation of the triple line in the image is counter-clockwise if Ej>0E_{j}>0 and clockwise if Ej<0E_{j}<0. Choosing the radius of the circle suitably, we have ejprej(r2dθ)=Ej\int_{e_{j}}\mathrm{pr}_{e_{j}}^{*}(r^{2}d\theta)=E_{j}. It has no simple tangency point.

If eje_{j} is a circle and Ej=0E_{j}=0 then we embed NejN_{e_{j}} into 2\mathbb{R}^{2} along an “88”-shaped immersed curve so that the origin is in the middle of the region whose boundary is oriented counter-clockwise. Using the same trick as in Figure 6, we may have an embedding of NejN_{e_{j}} such that ejprej(r2dθ)=0\int_{e_{j}}\mathrm{pr}_{e_{j}}^{*}(r^{2}d\theta)=0. Note that the foliation defined by r2dθ=0r^{2}d\theta=0 has two simple tangency points of index +1+1.

Finally, we check if the 11-form β0\beta_{0} obtained above satisfies the required conditions. It satisfies the condition (1) by the construction. The condition (2) is satisfied by choosing QQ to be sufficiently narrow so that Riβ0\int_{\partial R_{i}^{\prime}}\beta_{0} is sufficiently close to j=1mεijEj\sum_{j=1}^{m}\varepsilon_{ij}E_{j}, where εij\varepsilon_{ij} is the coefficient in inequality (2.1). Since (E1,,Em)C(P)(E_{1},\ldots,E_{m})\in C(P), the sum j=1mεijEj\sum_{j=1}^{m}\varepsilon_{ij}E_{j} is positive. Hence Riβ0\int_{\partial R_{i}^{\prime}}\beta_{0} is also positive. The condition (3) is obviously satisfied since d(r2dθ)>0d(r^{2}d\theta)>0. This completes the proof. ∎

Now we extend the 11-form β0\beta_{0} on QQ to PP by applying the following lemma used in the paper of Thurston and Winkelnkemper [11]. For a compact surface Σ\Sigma with boundary Σ\partial\Sigma, let Nbd(Σ;Σ)\mathrm{Nbd}(\partial\Sigma;\Sigma) denote a small compact neighborhood of Σ\partial\Sigma in Σ\Sigma.

Lemma 3.4 (Thurston-Winkelnkemper).

Let Σ\Sigma be a compact, oriented surface with boundary and β0\beta_{0} be a 11-form on Nbd(Σ;Σ)\mathrm{Nbd}(\partial\Sigma;\Sigma) such that Σβ0>0\int_{\partial\Sigma}\beta_{0}>0 and dβ0>0d\beta_{0}>0. Then there exists a 11-form β\beta on Σ\Sigma such that

  • β=β0\beta=\beta_{0} on Nbd(Σ;Σ)\mathrm{Nbd}(\partial\Sigma;\Sigma), and

  • dβ>0d\beta>0 on Σ\Sigma.

Proof of Proposition 3.1.

Suppose that a branched simple polyhedron PP is admissible. The conditions (2) and (3) for β0\beta_{0} in Lemma 3.3 allow us to use Lemma 3.4 for each region RiR_{i} of PP. Then the 11-form β\beta obtained by this lemma satisfies the required conditions in the assertion.

Conversely, if there exists a 11-form β\beta that satisfies the required conditions then the mm-tuple (e1β,,emβ)\left(\int_{e_{1}}\beta,\ldots,\int_{e_{m}}\beta\right) is an element in C(P)C(P) by Stokes’ theorem. Therefore PP is admissible. This completes the proof. ∎

Proof of Theorem 1.1.

As mentioned in the end of the above proof, if there exists a 11-form β\beta on PP with dβ>0d\beta>0 then PP is admissible. Thus Theorem 1.1 follows from Proposition 3.1. ∎

Remark 3.5.

The number of simple tangency points of the foliation defined by β\beta constructed in the above proof is at most 2m2m, where mm is the number of triple lines of PP.

4. Proof of Theorem 1.2

In this section, we prove Theorem 1.2

Lemma 4.1.

Let PP be a branched simple polyhedron. If PP has a preferred region then

{(x1,,xm)C(P)xi>0,i=1,,m}\{(x_{1},\ldots,x_{m})\in C(P)\mid x_{i}>0,\;\,i=1,\ldots,m\}

is empty.

Proof.

Let R¯\bar{R} be the metric completion of a preferred region RR and κ:R¯P\kappa:\bar{R}\to P be the natural extension of the inclusion RPR\to P. Let {l1,,ls}\{l_{1},\cdots,l_{s}\} be the set of edges and circles on the boundary R¯\partial\bar{R} of R¯\bar{R} such that κ(lk)\kappa(l_{k}) is a triple line of PP.

Assume that there exists a point (E1,,Em)(E_{1},\ldots,E_{m}) in C(P)C(P) with Ei>0E_{i}>0 for i=1,,mi=1,\ldots,m. For k=1,,sk=1,\ldots,s, let LkL_{k} be the real number EjkE_{j_{k}} assigned to the triple line ejke_{j_{k}} of PP with κ(lk)=ejk\kappa(l_{k})=e_{j_{k}}. Since the orientation of lkl_{k} is opposite to the orientation induced from that of RR, the sum of the assigned real numbers along R¯\partial\bar{R} becomes L1Ls<0-L_{1}-\cdots-L_{s}<0. This contradicts the assumption that (E1,,Em)C(P)(E_{1},\ldots,E_{m})\in C(P). ∎

Let \mathcal{F} be an S-stable foliation on a branched simple polyhedron PP defined by a 11-form β\beta on PP. Since \mathcal{F} is S-stable, it is transverse to S(P)S(P) in a small neighborhood of a vertex of PP.

Lemma 4.2.

Let PP be a branched simple polyhedron and \mathcal{F} be an S-stable foliation on PP defined by a 11-form β\beta. Then the number of simple tangency points of \mathcal{F} is even.

Proof.

The assertion follows from the fact that S(P)S(P) consists of the image of an immersion of a finite number of circles and \mathcal{F} is co-oriented. ∎

Since \mathcal{F} is transverse to S(P)S(P) in small neighborhoods of vertices, there are four kinds of projections as shown in Figure 7. In the figure, type 11 is the case where both of the oriented edges of S(P)S(P) intersect the leaves of \mathcal{F} in the same direction and type 22 is the case where they intersect the leaves in opposite directions. The arrows with symbol β\beta represent the direction along which the integration of β\beta becomes positive. The sign of type is ++ if the direction of the arrow coincides with the orientation of the edge passing from the left-bottom to the right-top and the sign is - otherwise. Each sign ++ or - written near the endpoints of the edges of S(P)S(P) represents the sign of the value obtained by inserting a vector tangent to the edge whose direction is consistent with the orientation of the edge into the 11-form β\beta. In each figure, only the painted region can be a part of a preferred region in both of \ell-type and rr-type vertex cases. We call the small neighborhoods of vertices in Figure 7 with labels 1+1_{+}, 11_{-}, 2+2_{+} and 22_{-} the H-pieces of type 1+1_{+}, 11_{-}, 2+2_{+} and 22_{-}, respectively.

Refer to caption
Figure 7. H-pieces.
Lemma 4.3.

Let PP be a flow-spine of a closed, oriented, smooth 33-manifold and \mathcal{F} be an S-stable foliation on PP defined by a 11-form. If there exists a vertex of PP whose H-piece is of type 22 then \mathcal{F} has a simple tangency point.

Recall that S(P)S(P) consists of the image of an immersion of a finite number of circles. In the following proofs, a circuit means the image of an immersion of one of these circles. If a circuit passes all triple lines then it is called an Euler circuit. A branched simple spine is a flow-spine if and only if it has an Euler circuit.

Proof.

Assume that there exists a vertex of PP whose H-piece is of type 22 and \mathcal{F} has no tangency point. The latter condition implies that if two edges of S(P)S(P) are connected in a circuit then the integrations of β\beta along these edges are either both positive or both negative. Since there is an H-piece of type 22, S(P)S(P) has at least two circuits. This contradicts the assumption that PP is a flow-spine. ∎

Proof of Theorem 1.2.

Assume that there exists an S-stable foliation \mathcal{F} defined by a 11-form β\beta with dβ>0d\beta>0 and without simple tangency points. If PP has no vertex then it does not admit a 11-form β\beta with dβ>0d\beta>0. Suppose PP has at least one vertex. By Lemma 4.3, the H-pieces of all the vertices are of type 11. Furthermore, all vertices must have the same sign, otherwise there exists an edge of PP which connects two vertices with opposite signs. If all vertices are of type 1+1_{+} then ejβ>0\int_{e_{j}}\beta>0 for all edges e1,,eme_{1},\ldots,e_{m}. Since PP has a preferred region, this is impossible by Lemma 4.1. If all vertices are of type 11_{-} then ejβ<0\int_{e_{j}}\beta<0 for all edges e1,,eme_{1},\ldots,e_{m}. Now we take the sum of the integrations of β\beta along the boundaries of all the regions of PP. In this calculation, ejβ\int_{e_{j}}\beta appears twice and ejβ\int_{-e_{j}}\beta once for each j=1,,mj=1,\ldots,m. Since we have ejβ<0\int_{e_{j}}\beta<0 for all j=1,,mj=1,\ldots,m, the sum of the integrations is negative. This contradicts the assumption that dβ>0d\beta>0 on PP. ∎

5. DS-diagram and Proof of Corollary 1.3

To explain the proof of Corollary 1.3, we first shortly introduce the DS-diagram of a flow-spine, see [10, 9] for details (cf. [6]). Let PP be a flow-spine of a closed, oriented, smooth 33-manifold MM. The complement MPM\setminus P is an open ball. The singular set S(P)S(P) of PP induces a trivalent graph on the boundary of the closed ball B3B^{3} obtained as the geometric completion of MPM\setminus P, which is called the DS-diagram of PP. The flow of PP induces a flow on B3B^{3}, and the set of points on the boundary B3\partial B^{3} at which the flow is tangent to B3\partial B^{3} constitutes a simple closed curve. This curve is called the E-cycle. The E-cycle separates B3\partial B^{3} into two open disks S+S^{+} and SS^{-}, where the flow is positively transverse to S+S^{+} and negatively transverse to SS^{-}. Here the orientation of B3\partial B^{3} is induced from that of B3B^{3}, and those of S+S^{+} and SS^{-} are induced by the inclusion S+,SBS^{+},S^{-}\subset\partial B. The E-cycle is oriented as the boundary of S+S^{+}. Observing the geometric completion of B3B^{3} in MM, we may see that each region of PP corresponds to a region on S+S^{+} bounded by the DS-diagram and each triple line of PP corresponds to an edge or circle in S+S^{+}. The same is true for SS^{-}. Each triple line of PP also corresponds to an edge or circle in the E-cycle. The orientation of S+S^{+} (resp. SS^{-}) is consistent with (resp. opposite to) the orientations of the regions of PP. We assign orientations to the edges and circles of the DS-diagram so that they coincide with the orientations of the triple lines of PP.

We call a region on S+S^{+} not adjacent to the E-cycle an internal region. If a DS-diagram has a circle component then the DS-diagram consists of three parallel circles one of which is the E-cycle, one is in S+S^{+} and the last one is in SS^{-}. The 33-manifold of the flow-spine given by this DS-diagram is S1×S2S^{1}\times S^{2}. See [2, Remark 1.2] (cf. [9, Example 3]).

Lemma 5.1.

The DS-diagram of a flow-spine has an internal region on S+S^{+} homeomorphic to a disk.

Proof.

Let nvn_{v} be the number of vertices of PP. Observing the geometric completion of MPM\setminus P, we can verify that the DS-diagram has nvn_{v} vertices on S+S^{+} and 2nv2n_{v} vertices on the E-cycle. If the DS-diagram has a circle component then the assertion follows. We assume that it has no circle component. Let GG be the subgraph of the DS-diagram consisting of edges lying in S+S^{+} and vertices adjacent to these edges. Note that the edges on the E-cycle are not contained in GG. We denote by G¯\bar{G} the union of GG and the regions on S+S^{+} bounded by GG. Let e(G¯)e(\bar{G}) and f(G¯)f(\bar{G}) be the numbers of edges and regions of G¯\bar{G}, respectively. Since GG has nvn_{v} univalent vertices and nvn_{v} trivalent vertices, we have nv+3nv=2e(G¯)n_{v}+3n_{v}=2e(\bar{G}). On the other hand, the Euler characteristic χ(G¯)\chi(\bar{G}) of G¯\bar{G} is given as

χ(G¯)=2nve(G¯)+f(G¯).\chi(\bar{G})=2n_{v}-e(\bar{G})+f(\bar{G}).

These equalities imply that f(G¯)=χ(G¯)f(\bar{G})=\chi(\bar{G}). Since χ(G¯)>0\chi(\bar{G})>0, we have f(G¯)>0f(\bar{G})>0. Hence there exists an internal region homeomorphic to a disk. ∎

Proof of Corollary 1.3.

By Lemma 5.1, the DS-diagram of a flow-spine has an internal region homeomorphic to a disk. It is proved in [9, Lemma 4.6] that if a flow-spine is positive then an internal region is always a preferred region, and the proof written there works for negative flow-spines also. Thus, the assertion follows from Theorem 1.2. ∎

6. Examples

In this section, we give an S-stable characteristic foliation given by a 11-form β\beta with dβ>0d\beta>0 on the abalone explicitly. We also give an example of branched standard spine that has an S-stable foliation without simple tangency points.

6.1. Poincaré-Hopf

We introduce a lemma that is useful for determining singularities of foliations on flow-spines. Let PP be a flow-spine of a closed, oriented, smooth 33-manifold MM and \mathcal{F} be a foliation on PP with only elliptic and hyperbolic singularities. Assume that \mathcal{F} is tangent to S(P)S(P) at a finite number of points in the interior of triple lines and the singularities do not lie on S(P)S(P). Let ee be the number of elliptic singularities, hh be the number of hyperbolic singularities and t+t_{+} and tt_{-} be the numbers of simple tangency points of \mathcal{F} with index +1+1 and 1-1, respectively.

Lemma 6.1.

Suppose that the foliation \mathcal{F} is in the above setting. Then the following equality holds:

eh=1+t+t2.e-h=1+\frac{t_{+}-t_{-}}{2}.

Note that if \mathcal{F} is S-stable then t=0t_{-}=0.

Proof.

According to the definition of DS-diagram, the regions and triple lines of PP can be described on the boundary of the closed 33-ball B3B^{3} obtained as the geometric completion of the complement MPM\setminus P. Thus, we may describe \mathcal{F} on the boundary B3\partial B^{3} of B3B^{3}, which has 2e2e elliptic singularities and 2h2h hyperbolic singularities. Each simple tangency point of index +1+1 becomes as shown on the top-middle in Figure 8 and it can be regarded as a hyperbolic singularity as shown on the right. Each simple tangency point of index 1-1 becomes as shown on the bottom in the figure and it can be regarded as an elliptic singularity. By Poincaré-Hopf formula on the sphere B3\partial B^{3}, we have (2e+t)(2h+t+)=2(2e+t_{-})-(2h+t_{+})=2. Thus the assertion follows. ∎

Refer to caption
Figure 8. Simple tangency points and corresponding singularities on B3\partial B^{3}.

6.2. Foliation on the abalone

The branched polyhedron AA obtained from the neighborhood of the singular set shown in Figure 9 by attaching two disks D1D_{1} and D2D_{2} is called the abalone. This is a flow-spine of S3S^{3}. We can easily see that (E1,E2)C(A)(E_{1},E_{2})\in C(A) if and only if

{E1>02E1+E2>0.\left\{\begin{split}&-E_{1}>0\\ &2E_{1}+E_{2}>0.\end{split}\right.

This system of inequalities has a solution, for example (1,13)2(-1,13)\in\mathbb{R}^{2}. Hence AA is admissible.

Refer to caption
Figure 9. The abalone AA.

We give an S-stable foliation on AA defined by a 11-form β\beta with dβ>0d\beta>0 explicitly. Set (E1,E2)=(1,13)C(A)(E_{1},E_{2})=(-1,13)\in C(A). According to the construction in the proof of Theorem 1.1, one can construct an S-stable foliation on Nbd(S(A);A)\mathrm{Nbd}(S(A);A) defined by a 11-form β0\beta_{0} with dβ0>0d\beta_{0}>0 as the left two figures in Figure 10. Let R1R_{1} and R2R_{2} be the regions of AA containing D1D_{1} and D2D_{2}, respectively. We describe four leaves on R2R_{2} as shown on the right in Figure 10. We denote these leaves by 1,,4\ell_{1},\ldots,\ell_{4}. For each i\ell_{i}, we can find a 11-form βi\beta_{\ell_{i}} on Nbd(i;R2)\mathrm{Nbd}(\ell_{i};R_{2}) so that βi=β0\beta_{\ell_{i}}=\beta_{0} on Nbd(i;R2)Nbd(R2;R2)\mathrm{Nbd}(\ell_{i};R_{2})\cap\mathrm{Nbd}(\partial R_{2};R_{2}) and i\ell_{i} is a leaf of the foliation defined by βi=0\beta_{\ell_{i}}=0.

Refer to caption
Figure 10. An S-stable foliation on Nbd(S(A);A)\mathrm{Nbd}(S(A);A) (left and middle) and the leaves 1,,4\ell_{1},\ldots,\ell_{4} on R2R_{2} (right).

We decompose R2\partial R_{2} into 1515 arcs by the endpoints of the leaves 1,,4\ell_{1},\ldots,\ell_{4}, the tangent points of these leaves with R2\partial R_{2} and the vertices on R2\partial R_{2}. We label these arcs as a1,,a15a_{1},\ldots,a_{15} as shown on the right in Figure 10. For the arcs a1,,a15a_{1},\ldots,a_{15}, we assign real numbers A1,,A15A_{1},\ldots,A_{15} so that

A1+A2+A3=E2,A4+A5+A6=E1,A7+A8+A9=E2,A10+A11+A12=E1,A13+A14+A15=E2\begin{split}&A_{1}+A_{2}+A_{3}=E_{2},\hskip 27.0301ptA_{4}+A_{5}+A_{6}=E_{1},\hskip 19.91692ptA_{7}+A_{8}+A_{9}=-E_{2},\\ &A_{10}+A_{11}+A_{12}=E_{1},\hskip 14.22636ptA_{13}+A_{14}+A_{15}=E_{2}\end{split}

and

A15+A1+A8>0,A2+A5>0,A3>0,A4>0,A6+A7>0,A9+A10>0,A11+A14>0,A12>0,A13>0,\begin{split}&A_{15}+A_{1}+A_{8}>0,\hskip 14.22636ptA_{2}+A_{5}>0,\hskip 21.9086ptA_{3}>0,\hskip 19.91692ptA_{4}>0,\hskip 19.91692ptA_{6}+A_{7}>0,\\ &A_{9}+A_{10}>0,\hskip 40.4029ptA_{11}+A_{14}>0,\hskip 14.22636ptA_{12}>0,\hskip 15.649ptA_{13}>0,\end{split}

and (A4,A5,A6)=(A10,A11,A12)(A_{4},A_{5},A_{6})=(A_{10},A_{11},A_{12}). The five equalities mean that the assignment of real numbers to a1,,a15a_{1},\ldots,a_{15} is a refinement of the assignment of those to e1e_{1} and e2e_{2}. The nine inequalities mean that, for each region bounded by R2i=14i\partial R_{2}\cup\bigcup_{i=1}^{4}\ell_{i}, the sum of the real numbers along its boundary is positive, which is necessary for applying Lemma 3.4. Note that we set the real number assigned to i\ell_{i} to be 0 since i\ell_{i} will be a leaf of the foliation obtained by a 11-form. The last equality is needed since, for each i=4,5,6i=4,5,6, the two edges aia_{i} and ai+6a_{i+6} correspond to the same arc on the edge e1e_{1} divided by the two simple tangency points. This system of equalities and inequalities has a solution, for example as

(A1,,A15)=(6,6,1,2,5,2,1,11,1,2,5,2,1,6,6).(A_{1},\ldots,A_{15})=(6,6,1,2,-5,2,-1,-11,-1,2,-5,2,1,6,6).

Let β\beta^{\prime} be a 11-form defined on a small neighborhood NN of S(P)(i=14i)S(P)\cup(\cup_{i=1}^{4}\ell_{i}) in PP so that

  • dβ>0d\beta^{\prime}>0 on NN,

  • the characteristic foliation on NN has leaves shown on the right in Figure 10,

  • ajβ=Aj\int_{a_{j}}\beta^{\prime}=A_{j} for j=1,,15j=1,\ldots,15.

Such a 11-form can be easily constructed as in the proof of Theorem 1.1 (cf. [9, the proof of Lemma 6.3]). We may extend β\beta^{\prime} to the whole PP by applying Lemma 3.4.

Remark 6.2.

The arcs 1,,4\ell_{1},\ldots,\ell_{4} are Legendrian if we consider a contact form α\alpha on Nbd(A;M)\mathrm{Nbd}(A;M) whose characteristic foliation coincides with the foliation defined by β=0\beta^{\prime}=0. When we describe a characteristic foliation of a flow-spine with a transverse Reeb flow, we need to choose the positions of these leaves so that the integration of β\beta^{\prime} along the boundary of each region separated by these leaves is positive. This gives a strong restriction for positions of leaves, which is different from the case of characteristic foliations on surfaces.

Consider a foliation on AA shown in Figure 11, which coincides with the one defined by β\beta^{\prime} on NN. It has two elliptic singularities with positive divergence and no hyperbolic singularities (cf. Lemma 6.1). It is not difficult to find a 11-form β\beta on PP with dβ>0d\beta>0 that defines the foliation in the figure. For example, let RR be the region on R2R_{2} bounded by the leaves 3\ell_{3} and 4\ell_{4} and the edges a11a_{11} and a14a_{14} and regard RR as a rectangle with coordinates (x,y)(x,y) so that the foliation is parallel to the xx-axis. On Nbd(R;R)\mathrm{Nbd}(\partial R;R), β\beta is given in the form φ(x,y)dy\varphi(x,y)dy, where φ(x,y)\varphi(x,y) is a smooth function on Nbd(R;R)\mathrm{Nbd}(\partial R;R) with φx(x,y)>0\frac{\partial\varphi}{\partial x}(x,y)>0. Remark that a11φ(x,y)𝑑y=A11\int_{a_{11}}\varphi(x,y)dy=A_{11}, a14φ(x,y)𝑑y=A14\int_{a_{14}}\varphi(x,y)dy=A_{14} and A11+A14>0A_{11}+A_{14}>0. We observe the graph of the function φ(x,y)\varphi(x,y) on Nbd(R;R)\mathrm{Nbd}(\partial R;R), and extend it to the whole RR so that φx(x,y)>0\frac{\partial\varphi}{\partial x}(x,y)>0. The 11-form φ(x,y)dy\varphi(x,y)dy defines the foliation on RR and satisfies d(φ(x,y)dy)>0d(\varphi(x,y)dy)>0.

The region RR^{\prime} bounded by the leaf 4\ell_{4} and the edges a12a_{12} and a13a_{13} has one elliptic singularity with positive divergence. For any region TT bounded by an arc on a12a13a_{12}\cup a_{13} and two leaves connecting the elliptic singularity and the endpoints of the arc, the 11-form β\beta that we are going to make should satisfy T𝑑β>0\int_{T}d\beta>0. Therefore, the assignment to the edges a12a_{12} and a13a_{13} must satisfy the inequalities A12>0A_{12}>0 and A13>0A_{13}>0. Under this setting, the 11-form on RR^{\prime} with required property can be found by the same way as before.

Applying the same construction for other regions bounded by the E-cycles and the leaves 1,,4\ell_{1},\ldots,\ell_{4}, we can obtain a 11-form β\beta on AA that defines the foliation in Figure 11 and satisfies dβ>0d\beta>0. The DS-diagram with this S-stable foliation is described in Figure 12.

Refer to caption
Figure 11. An S-stable foliation on AA with dβ>0d\beta>0 and with minimal number of simple tangency points.
Refer to caption
Figure 12. The DS-diagram with the S-stable foliation in Figure 11.

The foliation in Figure 11 has two simple tangency points on e1e_{1}. Since R1R_{1} is a preferred region, due to Theorem 1.2, we can conclude that there is no S-stable foliation defined by a 11-form β\beta with dβ>0d\beta>0 and with less simple tangency points.

A characteristic foliation of the flow-spine of the Poincaré homology 33-sphere whose DS-diagram is the dodecahedron is observed by the first author in [5] by the same manner.

6.3. An example without simple tangency point

Let PP be the branched simple polyhedron obtained from the branched polyhedron with boundary described in Figure 13 by attaching four disks along the four connected components of the boundary of PP. It has three vertices, six edges and four disk regions. The orientations of the edges e1,,e6e_{1},\ldots,e_{6} in the figure are those induced from the adjacent regions by the rule in Figure 3. The Euler characteristic of PP is 36+4=13-6+4=1, and hence the Euler characteristic of the boundary of a thickening of PP is 22. We may easily verify that the boundary of a thickening of PP is connected and hence it is S2S^{2}. This means that PP is a branched standard spine of a closed, oriented 33-manifold. We can check that the DS-diagram of this spine, forgetting the branching, is the diagram (1-10) in [7] and conclude that the 33-manifold is S3S^{3}.

Refer to caption
Figure 13. A branched standard spine that admits an S-stable foliation defined by a 11-form β\beta with dβ>0d\beta>0 and without simple tangency points.

The region containing the disk D0D_{0} in the figure is a preferred region. The defining inequalities of C(P)C(P) are E6>0E_{6}>0, E2E3>0-E_{2}-E_{3}>0, E1+E2E6>0E_{1}+E_{2}-E_{6}>0 and E2+2E3+E4+E5+E6>0E_{2}+2E_{3}+E_{4}+E_{5}+E_{6}>0. This system of inequalities has a solution, for example as (6,1,2,1,1,6)6(6,1,-2,-1,-1,6)\in\mathbb{R}^{6}. Hence PP is admissible. We choose the projections prvi\mathrm{pr}_{v_{i}} for each of vertices v1v_{1}, v2v_{2} and v3v_{3} so that their H-pieces become as shown in Figure 14. Under this setting, we can obtain an SS-stable foliation defined by a 11-form β\beta with dβ>0d\beta>0 and without simple tangency points.

Refer to caption
Figure 14. The H-pieces of v1v_{1}, v2v_{2} and v3v_{3}.

We here give an S-stable foliation on PP defined by a 11-form β\beta with dβ>0d\beta>0 and without simple tangency points on PP explicitly. Set

(E1,E2,E3,E4,E5,E6)=(6,1,2,1,1,6)C(P).(E_{1},E_{2},E_{3},E_{4},E_{5},E_{6})=(6,1,-2,-1,-1,6)\in C(P).

Let R1,R2,R3,R4R_{1},R_{2},R_{3},R_{4} be the regions of PP bounded by the sequences of edges {e6}\{e_{6}\}, {e2,e3}\{-e_{2},-e_{3}\}, {e1,e2,e5,e6,e5}\{e_{1},e_{2},-e_{5},-e_{6},e_{5}\} and {e1,e4,e5,e3,e1,e3,e4,e6,e4,e2}\{e_{1},e_{4},e_{5},e_{3},-e_{1},e_{3},e_{4},e_{6},-e_{4},e_{2}\}, respectively, see Figure 15. Draw four leaves 1,2,3\ell_{1},\ell_{2},\ell_{3} and 4\ell_{4} as shown in the figure. The edge e1e_{1} on the boundary of R3R_{3} divides into two oriented arcs by an endpoint of 1\ell_{1}, named a1a_{1} and a2a_{2} as in the figure. Similarly, we name a3,a4,a5,a6,a7a_{3},a_{4},a_{5},a_{6},a_{7} and a8a_{8} as in the figure. For these arcs a1,,a8a_{1},\ldots,a_{8}, we assign real numbers A1,,A8A_{1},\ldots,A_{8} as

(A1,A2,A3,A4,A5,A6,A7,A8)=(1.6,4.4,0.8,5.2,1.5,4.5,3.2,2.8)(A_{1},A_{2},A_{3},A_{4},A_{5},A_{6},A_{7},A_{8})=(1.6,4.4,0.8,5.2,1.5,4.5,3.2,2.8)

so that they satisfy A1+A2=A5+A6=E1A_{1}+A_{2}=A_{5}+A_{6}=E_{1}, A3+A4=A7+A8=E6A_{3}+A_{4}=A_{7}+A_{8}=E_{6} and

E5+A1>0,A2+E2A4>0,E5A3>0,E1+E4+E5+E3A5>0,A6+A8E4+E2>0,E3+E4+A7>0.\begin{split}&E_{5}+A_{1}>0,\hskip 91.04881ptA_{2}+E_{2}-A_{4}>0,\hskip 45.5244pt-E_{5}-A_{3}>0,\\ &E_{1}+E_{4}+E_{5}+E_{3}-A_{5}>0,\hskip 11.38109pt-A_{6}+A_{8}-E_{4}+E_{2}>0,\hskip 11.38109ptE_{3}+E_{4}+A_{7}>0.\end{split}

Thus, this assignment satisfies the required conditions for obtaining a 11-form β\beta with dβ>0d\beta>0 on PP. We may draw a foliation in the regions R1,,R4R_{1},\ldots,R_{4} described in Figure 15 so that R1R_{1} has only one elliptic singularity in the middle and the other regions have no singularity, and there exists a 11-form β\beta with dβ>0d\beta>0 on PP which gives this foliation as in the case of abalone. There are two “tangency points” on the boundary of R3R_{3} described in Figure 15, one is at the intersection of the edges e5e_{5} and e1e_{1} and the other is of the edges e5e_{5} and e6e_{6}, and the boundary of R4R_{4} also has two “tangency points”. The “tangency point” at the intersection of e5e_{5} and e1e_{1} appears in a neighborhood of the vertex v1v_{1} shown on the left in Figure 14, where the leaves of the foliation given by β=0\beta=0 is horizontal. The other three “tangency points” appear by the same reason. Therefore, these leaves are transverse to the singular set of PP and the foliation given in Figure 15 has no simple tangency points. The DS-diagram with this S-stable foliation is described in Figure 12.

Refer to caption
Figure 15. An S-stable foliation on PP with dβ>0d\beta>0 and without simple tangency points.
Refer to caption
Figure 16. The DS-diagram with the S-stable foliation in Figure 15.

References

  • [1] R. Benedetti, C. Petronio, Branched spines and contact structures on 3-manifolds, Ann. Mat. Pure Appl. (4) 178 (2000), 81–102.
  • [2] Endoh, M., Ishii, I., A new complexity for 3-manifolds, Japan. J. Math. (N.S.) 31 (2005), 131–156.
  • [3] Giroux, E., Convexité en topologie de contact, Comm. Math. Helv. 66 (1991), 637–677.
  • [4] Giroux, E., Géométrie de contact: de la dimension trois vers les dimensions supérieures, Proceedings of the International Congress of Mathematicians, Vol. II (Beijing, 2002), 405–414, Higher Ed. Press, Beijing, 2002.
  • [5] S. Handa, Construction of S-stable foliations on branched standard spines, Master’s Thesis, Tohoku University, 2018 (in Japanese).
  • [6] Ikeda, H., Inoue, Y., Invitation to DS-diagram, Kobe J. Math. 2 (1985) 169–186.
  • [7] Ishii, I., 12 Chōten E-cycle tsuki DS-diagrams, Hakone Seminar 1 (1985), 25–60 (in Japanese).
  • [8] Ishii, I., Flows and spines, Tokyo J. Math. 9 (1986), 505–525.
  • [9] Ishii, I., Ishikawa, M., Koda, Y., Naoe, H., Positive flow-spines and contact 33-manifolds, arXiv:1912.05774 [math.GT]
  • [10] Ishii, I., Moves for flow-spines and topological invariants of 33-manifolds, Tokyo J. Math. 15 (1992), 297–312.
  • [11] Thurston, W., Winkelnkemper, H., On the existence of contact forms, Proc. Amer. Math. Soc. 52 (1975), 345–347.