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

Geometry and topology of closed geodesics complements in the 3-torus

José Andrés Rodríguez Migueles Centro de Investigación en Matemáticas, GTO 36023, Mexico jose.migueles@cimat.mx
Abstract.

We show that for at most three closed geodesics with linearly independent directions, the homeomorphism type of its complement in the 33-torus is determine by the orbit of their direction vectors subspaces under the action of PSL3().\operatorname{PSL}_{3}(\mathbb{Z}). Moreover, we provide asymptotically sharp volume bounds for a family of closed geodesics complements. The bounds depend only on the distance in the Farey graph.

1. Introduction

Let 𝕋3\mathbb{T}^{3} be the 33-torus, this three-dimensional manifold admits an Euclidean structure obtained as the orbit space of 3,\mathbb{R}^{3}, under the discrete group 3\mathbb{Z}^{3} acting as integer translations, the quotient map is denoted by 𝒫\mathcal{P} . Every closed geodesic RR is an embedding of 𝕊1\mathbb{S}^{1} into 𝕋3,\mathbb{T}^{3}, so it can then be considered as a knot in 𝕋3.\mathbb{T}^{3}.

The aim of this paper, is to study the geometry and topology of the complement of a finite collection a closed geodesics inside 𝕋3.\mathbb{T}^{3}. Hui and Purcell gave in [4] and [3] an explicit characterization of the JSJ decomposition of this link complement, under the name of rod complements, by using Thurston’s geometrization theorem. A particular consequence is that each link complement with three or more linearly independent closed geodesics is hyperbolic or has a unique hyperbolic piece in its JSJ decomposition. In the case when the link complement is hyperbolic, such metric is unique by the Mostow’s Rigidity Theorem, meaning that any geometric invariant is a topological invariant.

In a recent paper [2] Do, Hui and Purcell, obtained the first volume bounds for the hyperbolic closed geodesics complement in terms of rod parameters (see [2, Theorem 3.2]) and they showed that these bounds may be loose in general. But in [2, Theorem 5.7] they gave an asymptotically sharp volume bounds for a family of closed geodesics complement, in terms of the lengths of the continued fractions formed from the rod parameters.

In this paper, we give asymptotically sharp volume bounds for a bigger family than in [2, Theorem 5.7] in terms of the Farey distance between pair of slopes coming from the projection of two components to two simple closed geodesics on a once-punctured 22-torus with a non-complete flat metric..

Theorem 1.1.

There exist a universal constant K11,K_{1}\geq 1, such that for at least three disjoint closed geodesics R1,R2,,RnR_{1},R_{2},...,R_{n} in the 33-torus where RnR_{n} has a direction vector (0,0,1)(0,0,1), for i<n,i<n, RiR_{i} is the image under 𝒫\mathcal{P} of the geodesic {x¯i+(tpi,tqi,si)|t}3\{\bar{x}_{i}+(tp_{i},tq_{i},s_{i})|t\in\mathbb{R}\}\subset\mathbb{R}^{3} with 0<si<sj<10<s_{i}<s_{j}<1 if i<j,i<j, [pi:qi][pi+1:qi+1][{p_{i}}:{q_{i}}]\neq[{p_{i+1}}:{q_{i+1}}], for i=1,2,,n2,i=1,2,...,n-2, and [p1:q1][pn1:qn1][{p_{1}}:{q_{1}}]\neq[{p_{n-1}}:{q_{n-1}}]. Then

12K1i=1n1d([pi:qi],[pi+1:qi+1])Vol(𝕋3{Ri}i=1n)v8i=1n1d([pi:qi],[pi+1:qi+1]),\frac{1}{2K_{1}}\sum_{i=1}^{n-1}d_{\mathcal{F}}\left([{p_{i}}:{q_{i}}],[{p_{i+1}}:{q_{i+1}}]\right)\leq\operatorname{Vol}(\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{n})\leq v_{8}\sum_{i=1}^{n-1}d_{\mathcal{F}}\left([{p_{i}}:{q_{i}}],[{p_{i+1}}:{q_{i+1}}]\right),

where v8v_{8} is the volume of a regular ideal octahedra and [pn:qn]:=[p1:q1][{p_{n}}:{q_{n}}]:=[{p_{1}}:{q_{1}}].

Although the upper bound is explicit in general, and sharp when the closed geodesis R1,,Rn1R_{1},...,R_{n-1} is a collection of Farey neighbors (see [5, Section 4] and [5, Corollary 7.3]) our lower bound depends on a non-explicit universal positive constant. This makes difficult to compare the bounds in [2, Theorem 5.7] with respect Theorem 1.1. However we can relate the invariant of the lengths of the continued fractions of the slope of the closed geodesic with respect the Farey distance between infinity and the corresponding slope, because they are the same. For that reason, we can conclude by triangle inequality that our upper bound is sharper in general and equal for some special cases.

Furthermore, we also investigate the question of the characterization on the topological type of the closed geodesics complement in the following new result which is the three-dimensional and at most three linearly independent component version of [2, Proposition 2.5]:

Theorem 1.2.

Let {Ri}i=1k\{R_{i}\}_{i=1}^{k} and {Ri}i=1k\{R^{\prime}_{i}\}_{i=1}^{k} be a pair of k3k\leq 3 disjoint linearly independent closed geodesics in the 3-torus. Then

𝕋3{Ri}i=1k𝕋3{Ri}i=1k,\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{k}\cong\mathbb{T}^{3}{\smallsetminus}\{R^{\prime}_{i}\}_{i=1}^{k},

if and only if there exist APSL3()A\in\operatorname{PSL}_{3}(\mathbb{Z}) such that maps the subspace generated by the direction vector of RiR_{i} to the subspace generated by RiR^{\prime}_{i} for all i.i.

1.1. Acknowledgements

I was greatly benefited with conversations from Araceli Guzmán-Tristán, Connie On Hui, Jessica Purcell and Jesús Rodríguez-Viorato. I also thank CIMAT and my first semester students in the Geometry class for creating an attractive mathematical environment.

2. Preliminaries

Let 𝕋3\mathbb{T}^{3} be the 33-torus obtained by gluing identically the opposite faces of the unit cube [0,1]×[0,1]×[0,1][0,1]\times[0,1]\times[0,1] in the 3-dimensional Euclidean space. Equivalently, we can see 𝕋3\mathbb{T}^{3} as the orbit space of 3,\mathbb{R}^{3}, under the discrete group 3\mathbb{Z}^{3} acting as integer translations. This quotient map 𝒫:3𝕋3\mathcal{P}\colon\mathbb{R}^{3}\to\mathbb{T}^{3} is also a Universal Riemannian covering map (see Definition 2.1 in [4] for an explicit map). Then 𝕋3\mathbb{T}^{3} inherits the Euclidean metric from 3.\mathbb{R}^{3}.

In this section, we fix a parametrization for closed geodesics and totally geodesic 22-torus inside the 33-torus, following ideas in [4] and [3], but instead of calling our links rods we called them closed geodesics.

2.1. Closed geodesics in the 33-torus

Recall that a geodesic in the three-dimensional Euclidean space can be given by a parametric expression of the form:

{p¯+td|t}\{\bar{p}+t\vec{d}\,|\,t\in\mathbb{R}\}

where p¯3\bar{p}\in\mathbb{R}^{3} and d3{(0,0,0)}\vec{d}\in\mathbb{R}^{3}{\smallsetminus}\{(0,0,0)\} is the direction vector of the geodesic.

Notice that this parametrization is not done by arc-length. And in many cases does not projects injectivelly under the universal covering map to 𝕋3.\mathbb{T}^{3}. But as we will be only concern about the closed geodesic as a knot, we do not inquire in a precise injective arc-length parametrization.

Recall a version of Lemma 2.2 in [4] that characterizes a closed geodesic in 𝕋3,\mathbb{T}^{3}, without precising the parametrization of the geodesic.

Lemma 2.1.

A geodesic in three-dimensional Euclidean space projects under 𝒫\mathcal{P} to a closed geodesic in 𝕋3,\mathbb{T}^{3}, if and only if, the coordinates of its direction vector generate a one dimensional vector space over \mathbb{Q}.

2.2. Totally geodesic 22-torus inside the 33-torus

Recall that a Euclidean plane in the three-dimensional Euclidean space can be given by a parametric expression of the form:

{p¯+td1+sd2|t,s}\{\bar{p}+t\vec{d_{1}}+s\vec{d_{2}}\,|\,t,s\in\mathbb{R}\}

where p¯3\bar{p}\in\mathbb{R}^{3} and d1,d23{(0,0,0)}\vec{d_{1}},\vec{d_{2}}\in\mathbb{R}^{3}{\smallsetminus}\{(0,0,0)\} are linearly independent direction vectors of the plane.

Recall the Lemma 2.6 in [3] characterizes a totally geodesic 22-torus in 𝕋3.\mathbb{T}^{3}.

Lemma 2.2.

A plane in three-dimensional Euclidean space projects under 𝒫\mathcal{P} to a totally geodesic 22-torus in 𝕋3,\mathbb{T}^{3}, if and only if, the plane has two linearly independent director vectors such that the coordinates of each one generate a one dimensional vector space over \mathbb{Q}.

Notice that the difference in the statement of Lemma 2.6 in [3] rely on the flexibility of the way we do not precise the parametrization of the plane.

3. Homeomorphism type of three closed geodesics complement

In this section we characterize the homomorphism type of the complement of at most three closed geodesics that are linearly independent in terms of the action of PSL3()\operatorname{PSL}_{3}(\mathbb{Z}) in at most three non-colinear points in the real projective plane 2\mathbb{P}^{2} (or the subspaces generated by the corresponding vector directions).

Theorem 1.2.

Let {Ri}i=1k\{R_{i}\}_{i=1}^{k} and {Ri}i=1k\{R^{\prime}_{i}\}_{i=1}^{k} be a pair of k3k\leq 3 disjoint linearly independent closed geodesics in the 3-torus. Then

𝕋3{Ri}i=1k𝕋3{Ri}i=1k,\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{k}\cong\mathbb{T}^{3}{\smallsetminus}\{R^{\prime}_{i}\}_{i=1}^{k},

if and only if there exist APSL3()A\in\operatorname{PSL}_{3}(\mathbb{Z}) such that maps the subspace generated by the direction vector of RiR_{i} to the subspace generated by RiR^{\prime}_{i} for all i.i.

Proof.

On one side, suppose there exist APSL3()A\in\operatorname{PSL}_{3}(\mathbb{Z}) such that sends the direction vector of RiR_{i} to RiR^{\prime}_{i} for all i.i. Because APSL3()A\in\operatorname{PSL}_{3}(\mathbb{Z}) then the linear map associated in 3\mathbb{R}^{3} passes to an homeomorphism A¯\bar{A} of the quotient space 𝕋3.\mathbb{T}^{3}.

If A¯(R1)R1,\bar{A}(R_{1})\neq R^{\prime}_{1}, then there is a translation τ\tau in 𝕋3\mathbb{T}^{3} that make them coincide, and we have the case 𝕋3R1𝕋3R1\mathbb{T}^{3}{\smallsetminus}R_{1}\cong\mathbb{T}^{3}{\smallsetminus}R^{\prime}_{1} . In the case k=2,k=2, if we have that τA¯(R2)\tau\circ\bar{A}(R_{2}) and R2,R^{\prime}_{2}, are linearly isotopic in 𝕋3R1,\mathbb{T}^{3}{\smallsetminus}R^{\prime}_{1}, so 𝕋3{Ri}i=12𝕋3{Ri}i=12.\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{2}\cong\mathbb{T}^{3}{\smallsetminus}\{R^{\prime}_{i}\}_{i=1}^{2}.

For the case k=3k=3, if τA¯(Ri)\tau\circ\bar{A}(R_{i}) is not linear isotopic to RiR^{\prime}_{i} for i=1,2i=1,2 in the complement of τA¯(Rj)Rj\tau\circ\bar{A}(R_{j})\cup R^{\prime}_{j} with iji\neq j and j=2,3,j=2,3, then there is a translation τ\tau^{\prime} in 𝕋3\mathbb{T}^{3} with the same vector direction as τA¯(R1)\tau\circ\bar{A}(R_{1}) such that makes them linear isotopic in the corresponding complements.

Finally, up to disjoint local linear isotopies between ττA¯(Ri)\tau^{\prime}\circ\tau\circ\bar{A}(R_{i}) and RjR^{\prime}_{j} for i=1,2i=1,2 we found an homeomorphism of 𝕋3\mathbb{T}^{3} sending (R1,R2,R3)(R_{1},R_{2},R_{3}) to (R1,R2,R3),(R^{\prime}_{1},R^{\prime}_{2},R^{\prime}_{3}), giving the wanted homeomorphism between the corresponding complements.

On the other side, is enough to show that the homeomorphism hh between 𝕋3{Ri}i=1k\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{k} and 𝕋3{Ri}i=1k\mathbb{T}^{3}{\smallsetminus}\{R^{\prime}_{i}\}_{i=1}^{k} can be extended to a self-homeomorphism h^\widehat{h} of 𝕋3\mathbb{T}^{3} with h^(Ri)=Ri.\widehat{h}(R_{i})=R^{\prime}_{i}. Notice that if hh maps mi,m_{i}, the meridian of Ri,R_{i}, to the meridian of Ri,R^{\prime}_{i}, then we can extend hh to the disks that they bound inside 𝕋3\mathbb{T}^{3}. Moreover, we can extend this homeomorphism along the core of 𝒩Ri,\mathcal{N}_{R_{i}}, the normal neigborhood of Ri,R_{i}, to 𝒩Ri.\mathcal{N}_{R^{\prime}_{i}}. As the morphism in homology induced by the restriction of hh on 𝒩Ri\partial\mathcal{N}_{R_{i}} is an isomorphism, then H1(h𝒩Ri)(Ker(H1(iRi)))=Ker(H1(iRi)),H_{1}(h_{\mid\partial\mathcal{N}_{R_{i}}})(\operatorname{Ker}(H_{1}(i_{R_{i}})))=\operatorname{Ker}(H_{1}(i_{R^{\prime}_{i}})), where iRii_{R_{i}} is the inclusion of 𝒩Ri\partial\mathcal{N}_{R_{i}} into 𝕋3{Ri}i=1k,\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{k}, and analogous for iRi.i_{R^{\prime}_{i}}. As the image of a essential simple closed curve that represents the class of the meridian under the homeomorphism h𝒩Rih_{\mid\partial\mathcal{N}_{R_{i}}} has to be a essential simple closed curve inside 𝒩Ri,\partial\mathcal{N}_{R^{\prime}_{i}}, then we just need to prove the following claim:

Claim 3.1.

Ker(H1(iRi))\operatorname{Ker}(H_{1}(i_{R_{i}})) is generated by a multiple of [mRi][m_{R_{i}}] in H1(𝒩Ri;).H_{1}(\partial\mathcal{N}_{R_{i}};\mathbb{R}).

To proof the Claim 3.1 notice that the image of the homology class of a longitude in 𝒩Ri\partial\mathcal{N}_{R_{i}} under H1(iRi)H_{1}(i_{R_{i}}) is not trivial in H1(𝕋3{Ri}i=1k;)H_{1}(\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{k};\mathbb{R}) because [Ri][R_{i}] is not trivial in H1(𝕋3;).H_{1}(\mathbb{T}^{3};\mathbb{R}). Then by considering the Mayer-Vietoris sequence in homology for the triad (𝕋3,𝕋3{Ri}i=1k,i=1k𝒩Ri),(\mathbb{T}^{3},\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{k},\bigcup_{i=1}^{k}\mathcal{N}_{R_{i}}), we only need to show that there is an element in H2(𝕋3)H_{2}(\mathbb{T}^{3}) whose image under δ^\widehat{\delta}, the connecting morphism of the sequence, is a non trivial element of H1(𝒩Ri).H_{1}(\partial\mathcal{N}_{R_{i}}). Consider TiT_{i} the torus in 𝕋3\mathbb{T}^{3} parallel to the plane that contains the direction vectors of the RkR_{k}’s but not the vector direction of RiR_{i}. Then

δ^([Ti])=[(Ti𝒩Ri)]=n[mRi],withn0.\widehat{\delta}([T_{i}])=[\partial(T_{i}\cap\mathcal{N}_{R_{i}})]=n[m_{R_{i}}],\hskip 5.69046pt\mbox{with}\hskip 5.69046ptn\neq 0.

Now we give an application of this result to a special family of links coming from periodic orbits of the geodesic flow on surfaces. We recall that the projective tangent bundle over 𝕋2\mathbb{T}_{*}^{2} a once-punctured torus with a flat metric, denoted by PT(𝕋2),PT(\mathbb{T}_{*}^{2}), is homeomorphic to the complement of a closed geodesic in the 33-torus, which is the oriented trivial circle bundle over 𝕋2.\mathbb{T}_{*}^{2}. To every closed geodesics γ\gamma on 𝕋2,\mathbb{T}_{*}^{2}, we associate the complement in PT(𝕋2)PT(\mathbb{T}_{*}^{2}) of the periodic orbit of the geodesic flow associated to γ\gamma which we denoted by γ^.\widehat{\gamma}. Then we obtained a criterium to decide when a couple of two periodic orbits of the geodesic flow have not homeomorphic complement, which is an analog version of Theorem 1.6 in [6], for a two link component case.

Corollary 3.2.

Let {γ1,γ2}\{\gamma_{1},\gamma_{2}\} and {η1,η2}\{\eta_{1},\eta_{2}\} a couple of two non-homotopic simple closed geodesics in 𝕋2\mathbb{T}_{*}^{2} relative to a metric with nowhere positive curvature. Then there is an homeomorphism, preserving the cusp coming from the puncture, between PT(𝕋2){γ^1,γ^2}PT(\mathbb{T}_{*}^{2}){\smallsetminus}\{\widehat{\gamma}_{1},\widehat{\gamma}_{2}\} and PT(𝕋2){η^1,η^2}PT(\mathbb{T}_{*}^{2}){\smallsetminus}\{\widehat{\eta}_{1},\widehat{\eta}_{2}\} if and only if there is a diffeomorphism ϕ\phi on 𝕋2\mathbb{T}_{*}^{2} such that ϕ(γi)\phi(\gamma_{i}) is homotopic to ηi\eta_{i} for i=1,2.i=1,2.

Proof.

Suppose there exist a diffeomorphism ϕ\phi on 𝕋2\mathbb{T}_{*}^{2} such that ϕ(γi)\phi(\gamma_{i}) is homotopic to ηi\eta_{i} for i=1,2.i=1,2. Then we can adapt a flat metric on 𝕋2\mathbb{T}_{*}^{2} such that the original simple closed geodesics are transversaly homotopic to the the ones in the flat metric. Notice that the linear isotopy class is determine by their slopes [γi][\gamma_{i}] and [ηi],[\eta_{i}], and that a transveral homotopy lifts to an isotopy of the corresponding periodic orbits (see [6, Subsection 2.1]).

Moreover, the existence of ϕ\phi induces an element APSL2()A\in\operatorname{PSL}_{2}(\mathbb{Z}) such that A[γi]=[ηi].A[\gamma_{i}]=[\eta_{i}]. Then naturally we can extend AA to A¯PSL3()\bar{A}\in\operatorname{PSL}_{3}(\mathbb{Z}) which fixes the last coordinate, and this induces an homeomorphism ϕ^\widehat{\phi} on 𝕋3\mathbb{T}^{3} which fixes the closed geodesic with direction vector (0,0,1).(0,0,1). Up to composing with a translation parallel to the direction (0,0,1)(0,0,1) we get that ϕ^(γ^i)\widehat{\phi}(\widehat{\gamma}_{i}) and η^i\widehat{\eta}_{i} are linearly isotopic in the complement and leaves invariant the cusp of PT(𝕋2)PT(\mathbb{T}_{*}^{2}) coming from the puncture of 𝕋2\mathbb{T}_{*}^{2} . This gives the wanted homeomorphism between the corresponding complements.

If we suppose that there is an homeomorphism

h:PT(𝕋2){γ^1,γ^2}PT(𝕋2){η^1,η^2},h\colon PT(\mathbb{T}_{*}^{2}){\smallsetminus}\{\widehat{\gamma}_{1},\widehat{\gamma}_{2}\}\to PT(\mathbb{T}_{*}^{2}){\smallsetminus}\{\widehat{\eta}_{1},\widehat{\eta}_{2}\},

preserving the cusp coming from the puncture, then by Theorem 1.2 this extends to an homeomorphism of the 33-torus which, up to isotopy, lifts to a linear map A¯PSL3()\bar{A}\in\operatorname{PSL}_{3}(\mathbb{Z}) which fixes the last coordinate because it preserves cusp coming from the fibers in PT(𝕋2).PT(\mathbb{T}_{*}^{2}). Then by forgetting the last coordinate on the linear map A¯\bar{A} induces a linear APSL2(),A\in\operatorname{PSL}_{2}(\mathbb{Z}), which induces an diffeomorphism of 𝕋2\mathbb{T}_{*}^{2} sending γi\gamma_{i} to a simple closed geodesic linearly isotopic to ηi.\eta_{i}.

4. Volume bounds for stratified closed geodesics complements

4.1. Farey graph

Let 𝕋2\mathbb{T}^{2} be the 22-torus, obtained by gluing identically the opposite sides of the unit square [0,1]×[0,1][0,1]\times[0,1] in the 2-dimensional Euclidean plane, and 𝕋2\mathbb{T}_{*}^{2} the once-punctured 22-torus with a non-complete flat metric. The Farey graph \mathcal{F} is the metric graph obtained by taking the set of isotopy classes of simple closed geodesics as vertices (which is in correspondence with the rational points of the projective line 1).\mathbb{P}^{1}). We say that [p:q],[r:s]1[{p}:{q}],[{r}:{s}]\in\mathbb{P}^{1} two vertices are connected by an edge if the corresponding closed geodesics intersect only once (this does not depend on the representant of the class). For more facts about the Farey graph see Subsection 2.3 in [5].

Theorem 1.1.

There exist a universal constant K11,K_{1}\geq 1, such that for at least three disjoint closed geodesics R1,R2,,RnR_{1},R_{2},...,R_{n} in the 33-torus where RnR_{n} has a direction vector (0,0,1)(0,0,1), for i<n,i<n, RiR_{i} is the image under 𝒫\mathcal{P} of the geodesic {x¯i+(tpi,tqi,si)|t}3\{\bar{x}_{i}+(tp_{i},tq_{i},s_{i})|t\in\mathbb{R}\}\subset\mathbb{R}^{3} with 0<si<sj<10<s_{i}<s_{j}<1 if i<j,i<j, [pi:qi][pi+1:qi+1][{p_{i}}:{q_{i}}]\neq[{p_{i+1}}:{q_{i+1}}], for i=1,2,,n2,i=1,2,...,n-2, and [p1:q1][pn1:qn1][{p_{1}}:{q_{1}}]\neq[{p_{n-1}}:{q_{n-1}}]. Then

12K1i=1n1d([pi:qi],[pi+1:qi+1])Vol(𝕋3{Ri}i=1n)v8i=1n1d([pi:qi],[pi+1:qi+1]),\frac{1}{2K_{1}}\sum_{i=1}^{n-1}d_{\mathcal{F}}\left([{p_{i}}:{q_{i}}],[{p_{i+1}}:{q_{i+1}}]\right)\leq\operatorname{Vol}(\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{n})\leq v_{8}\sum_{i=1}^{n-1}d_{\mathcal{F}}\left([{p_{i}}:{q_{i}}],[{p_{i+1}}:{q_{i+1}}]\right),

where v8v_{8} is the volume of a regular ideal octahedra and [pn:qn]:=[p1:q1][{p_{n}}:{q_{n}}]:=[{p_{1}}:{q_{1}}].

Proof.

The volume upper bound is obtained by removing in between the RiR_{i} and the Ri+1R_{i+1} closed geodesic a finite ordered sequence of perpendicular closed geodesics to the vector direction (0,0,1)(0,0,1) whose corresponding vector direction corresponds to the vertices of the geodesic in Farey graph joining [pi:qi][{p_{i}}:{q_{i}}] and [pi+1:qi+1][{p_{i+1}}:{q_{i+1}}]. Since Dehn filling does not increase volume by Thurston [7, Proposition 6.5.2], we have an upper bound by estimating the volume of this new drilled manifold. By [5, Theorem 4.2] the resulting manifold has a complete hyperbolic structure obtained by gluing i=1n1d([pi:qi],[pi+1:qi+1])\sum_{i=1}^{n-1}d_{\mathcal{F}}\left([{p_{i}}:{q_{i}}],[{p_{i+1}}:{q_{i+1}}]\right) regular ideal octahedra.

For the volume lower bound, we used the fact that the image under 𝒫\mathcal{P} of {(Ri,z=si)}i=1n1\{(R_{i},z=s_{i})\}_{i=1}^{n-1} minus the puncture coming from RnR_{n} is a stratification of RiR_{i} with i=1,,n1i=1,...,n-1 inside 𝕋3Rn\mathbb{T}^{3}{\smallsetminus}R_{n} (see [1, Definition 1] ) and [1, Theorem E] there is a positive constant K11K_{1}\geq 1 depending only on the topology of the once-punctured torus, such that:

12K1(i=1n1d(PXi,PYi+1))Vol(𝕋3{Ri}i=1n)\frac{1}{2K_{1}}\left(\sum_{i=1}^{n-1}d_{\mathcal{F}}(P_{X_{i}},P_{Y_{i+1}})\right)\leq Vol(\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{n})

where {(PXi,PYi+1)}i=1n1\{(P_{X_{i}},P_{Y_{i+1}})\}_{i=1}^{n-1} are the Bers pants decomposition coming from the conformal structures of the stratifying surfaces, which is the projection of z=si,z=s_{i}, minus the point (0,0,si),(0,0,s_{i}), under 𝒫\mathcal{P} .

Since any such pants decomposition contains only one loop the pants distance is the same as the Farey graph distance. Moreover, this loop is forced to be the rank one cusp induced by a simple closed curve of slope [pi:qi][{p_{i}}:{q_{i}}]. Therefore, we see that:

d(PXi,PYi+1)=d([pi:qi],[pi+1:qi+1])d_{\mathcal{F}}(P_{X_{i}},P_{Y_{i+1}})=d_{\mathcal{F}}\left([{p_{i}}:{q_{i}}],[{p_{i+1}}:{q_{i+1}}]\right)

Corollary 4.1.

There exist universal constant K11,K_{1}\geq 1, such that for R1,R2,R3R_{1},R_{2},R_{3} linear independent closed geodesics such that de direction vector of R3R_{3} is orthogonal to the plane generated by the direction vectors of R2R_{2} and R1R_{1}. Then

12K1d([p1:q1],[p2:q2])Vol(𝕋3{Ri}i=13)2v8d([p1:q1],[p2:q2]),\frac{1}{2K_{1}}d_{\mathcal{F}}\left([{p_{1}}:{q_{1}}],[{p_{2}}:{q_{2}}]\right)\leq\operatorname{Vol}(\mathbb{T}^{3}{\smallsetminus}\{R_{i}\}_{i=1}^{3})\leq 2v_{8}d_{\mathcal{F}}\left([{p_{1}}:{q_{1}}],[{p_{2}}:{q_{2}}]\right),

where [pi:qi][{p_{i}}:{q_{i}}] is the direction vector of RiR_{i}, form i=1,2i=1,2 relative to an orthogonal basis of the plane generated by the direction vectors of R2R_{2} and R1R_{1} and v8v_{8} is the volume of a regular ideal octahedra.

Proof.

By Theorem 1.2 we can apply an an ortogonal matrix in PSL3()\operatorname{PSL}_{3}(\mathbb{Z}) such that sends the direction vector subespaces of R3R_{3} to (0,0,1)(0,0,1) and the other direction vectors subespaces corresponding to R2R_{2} and R1R_{1} to two direction vectors with last coordinate equal to zero.

Then after choosing an ortogonal basis for the plane generated by the direction vectors corresponding to R2R_{2} and R1,R_{1}, we obtain the slope of each direction vector of RiR_{i} as [pi:qi][{p_{i}}:{q_{i}}].

Finally, by Theorem 1.1 we get the wanted result for the case with three closed geodesics. ∎

As in the previous section, we give an application to a special family of links coming from periodic orbits of the geodesic flow on the once-punctured torus 𝕋2\mathbb{T}_{*}^{2} with a flat metric. Notice that a closed geodesic γi\gamma_{i} in this case is given by the parametrization {(xi,yi)+t(pi,qi)|t},\{(x_{i},y_{i})+t(p_{i},q_{i})\,|\,t\in\mathbb{R}\}, then the periodic orbit γ^\widehat{\gamma} is given by the parametrization {(xi,yi,arctan(piqi))+t(pi,qi,0)|t}.\{(x_{i},y_{i},\arctan\left(\frac{p_{i}}{q_{i}}\right))+t(p_{i},q_{i},0)\,|\,t\in\mathbb{R}\}. Consequently, given Γ\Gamma any finite family of nonparallel closed geodesics on the once-punctured torus 𝕋2\mathbb{T}_{*}^{2} with a flat metric, induces an ordering by calculating π2arctan(piqi)<π2,\frac{-\pi}{2}\leq\arctan\left(\frac{p_{i}}{q_{i}}\right)<\frac{\pi}{2}, and then we explicit and improve the right inequality of Theorem B in [1] as follows:

Corollary 4.2.

Let Γ\Gamma is a filling collection of nonparallel essential simple closed curves in minimal position on 𝕋2.\mathbb{T}_{*}^{2}. Then, there exists an ordering in Γ={γi}i=1n\Gamma=\{\gamma_{i}\}^{n}_{i=1} and a universal constant K11,K_{1}\geq 1, such that:

12K1(i=1nd([pi:qi],[pi+1:qi+1]))Vol(PT(𝕋2)Γ^)2v8(i=1nd([pi:qi],[pi+1:qi+1])),\frac{1}{2K_{1}}\left(\sum_{i=1}^{n}d_{\mathcal{F}}\left([{p_{i}}:{q_{i}}],[{p_{i+1}}:{q_{i+1}}]\right)\right)\leq\operatorname{Vol}\left(PT(\mathbb{T}_{*}^{2}){\smallsetminus}\widehat{\Gamma}\right)\leq 2v_{8}\left(\sum_{i=1}^{n}d_{\mathcal{F}}\left([{p_{i}}:{q_{i}}],[{p_{i+1}}:{q_{i+1}}]\right)\right),

where [pi:qi][{p_{i}}:{q_{i}}] is the slope associated to γi,\gamma_{i}, v8v_{8} is the volume of a regular ideal octahedra and γn+1:=γ1\gamma_{n+1}:=\gamma_{1}.

5. Further questions

Our results on the geometry and topology of closed geodesics complements suggest various questions worthy of further exploration, such as the following.

Question 5.1.

Does there exist a more general characterization of the topological type of closed geodesics complement in 𝕋3\mathbb{T}^{3}?

A problem appears, for example when there are two closed geodesics whose directions vectors are linearly dependent and there is another closed geodesic intersecting the annuli between them, because one can obtain an homeomorphism coming by doing annular Dehn filling (see Section 4 in [2]). Other possible direction of generalization is for the complements of closed geodesics in 𝕋n\mathbb{T}^{n} as is stated for the one component case in [2, Proposition 2.5]

Question 5.2.

Do there exist explicit sharp volume bounds for three linear independent closed geodesics complements, such that any pair is orthogonal in between, in terms of same combinatorial invariant of the closed geodesics?

Even though Do, Hui and Purcell, gave in [2] explicit volume bounds for all closed geodesics complement, these are loose in general and are not in terms of the same combinatorial invariant of the closed geodesics.

Question 5.3.

For the lower bound in [2, Theorem 5.7] and our Theorem 1.1, which is sharper?

References

  • [1] Tommaso Cremaschi, José Andrés Rodríguez-Migueles, and Andrew Yarmola, On volumes and filling collections of multicurves, J. Topol. 15 (2022), no. 3, 1107–1153.
  • [2] Norman Do, Jessica S. Purcell, and Connie On Yu Hui, Volume bounds for hyperbolic rod complements in the 3-torus, 2409.02357v2, 2022.
  • [3] Connie On Yu Hui, A geometric classification of rod complements in the 3-torus, Proc. Amer. Math. Soc. 56 (2024), no. 4, 1291–1309.
  • [4] Connie On Yu Hui and Jessica S. Purcell, On the geometry of rod packings in the 3-torus, Bulletin of the London Mathematical Society 56 (2024), no. 4, 1291–1309.
  • [5] Tali Pinsky, Jessica Purcell, and José Andrés Rodríguez-Migueles, Arithmetic modular links, Pacific Journal of Mathematics 327 (2023), no. 2, 337–358.
  • [6] J. Andres Rodriguez-Migueles, A lower bound for the volumes of complements of periodic geodesics, J. London Math. Soc. 2 (2020), no. 3, 1–27.
  • [7] William P. Thurston, The geometry and topology of three-manifolds, Princeton Univ. Math. Dept. Notes, Available at https://www.msri.org/communications/books/gt3m, 1979.