The Big Dehn Surgery Graph and the link of
Dedicated to Bill Thurston
Abstract
In a talk at the Cornell Topology Festival in 2004, W. Thurston discussed a graph which we call “The Big Dehn Surgery Graph”, Here we explore this graph, particularly the link of . and prove facts about the geometry and topology of , We also investigate some interesting subgraphs and pose what we believe are important questions about . .
1. Introduction
In unpublished work, W. Thurston described a graph that had a vertex for each closed, orientable, 3-manifold and an edge between two distinct vertices and if there exists a Dehn surgery between , and That is, there is a knot . and is obtained by non-trivial Dehn surgery along in The edges are unoriented since . is also obtained from via Dehn surgery. Roughly following W. Thurston, we will call this graph the Big Dehn Surgery Graph, denoted by We will sometimes denote the vertex . by If . and are obtained from one another via Dehn surgery along two distinct knots, we do not make two distinct edges, although this would also make an interesting graph. We first record some basic properties of These follow from just some of the extensive work that has been done in the field of Dehn surgery. .
The graph is connected by the beautiful work of Lickorish Reference 21 and Wallace Reference 35 who independently showed that all closed, orientable can be obtained by surgery along a link in -manifolds That every vertex . in has infinite valence can be seen, amongst other ways, by constructing a hyperbolic knot in via the work of Myers in Reference 26. Then by work of Thurston Reference 33 all but finitely many fillings are hyperbolic, and the volumes of the filled manifolds approach the volume of the cusped manifold. The graph has infinite diameter since the rank of can change by at most one via drilling and filling, and there are 3-manifolds with arbitrarily high rank.
The Lickorish proof explicitly constructs a link, and therefore allows us to describe a natural notion of distance. A shortest path from to in counts the minimum number of components needed for a link in to admit as a surgery. We will refer to the number of edges in a shortest edge path between and as the Lickorish path length and denote this function by For example, if . denotes the Poincaré homology sphere, then See § .3 for more on Lickorish path length appears in the literature as surgery distance (see .Reference 3, Reference 18). This gives us a metric on which we assume throughout the paper. ,
The Big Dehn Surgery Graph is very big. In order to get a handle on it, we will study some useful subgraphs. We denote the subgraph of a graph generated by the vertices by The link of a vertex . is the subgraph If there is an automorphism of . taking a vertex to a vertex then the links of , and are isomorphic as graphs. We study the links of vertices and a possible characterization of the link of in 2. Associated to any knot in a manifold is a the complete graph on infinitely many vertices (to distinguish between knots and complete graphs we will use , and to denote complete graphs). When we want to describe a subgraph of we will use , See .2 for notation conventions used in this paper.
Interestingly, not every arises this way. We prove this in 4 and make some further observations about these subgraphs. In 6 we study the subgraph The vertices of the subgraph . are closed hyperbolic 3-manifolds and there is an edge between two vertices and if there is a cusped hyperbolic 3-manifold with two fillings homeomorphic to and We also study the geometry of . and showing that neither is , in -hyperbolic7. In 7 we also construct flats of arbitrarily large dimension in An infinite family of hyperbolic 3-manifolds with weight one fundamental group which are not obtained via surgery on a knot in . is given in 3. This shows that a characterization of the vertices in the link of remains open. Bounded subgraphs whose vertices correspond to other geometries are detailed in 5.
2. The link of
We now set notation which we will use for the remainder of the paper. A slope on the boundary of a 3-manifold is an isotopy class of unoriented, simple closed curves on We denote the result of Dehn surgery on . along a knot with filling slope by We denote Dehn filling along a link . by or with a dash denoting an unfilled component. Thus the exterior of , in is denoted and the complement is denoted by We will say that . or is hyperbolic if its interior admits a complete hyperbolic metric of finite volume. For knot and link exteriors in we will frame the boundary tori homologically, unless otherwise noted.
Here we study the links of vertices in particularly the link of , As above, the link of a vertex in . is the subgraph If . is associated to the manifold the vertices in this subgraph correspond to distinct manifolds which can be obtained via Dehn surgery on knots in , We refer to this subgraph as the link of . in , or just the link of .
The link of in is connected. There are several proofs of this fact. Perhaps the most intuitive is to use that a crossing change on a knot in can be realized as a Dehn surgery along an unknotted circle; see Reference 31. One must be careful to ensure that none of these surgeries results in .
The proof we give here arose from conversations with Luisa Pauoluzzi, and the path shows that the link of has bounded diameter.
One might hope to distinguish the links of vertices combinatorially in For example, is the link of any vertex in . connected? of bounded diameter? A negative answer would lead to an obstruction to automorphisms of the graph that do not fix More generally, an answer to the following question would lead to a better understanding of how the Dehn surgery structure of a manifold relates to the homeomorphism type. .
Given our results below in §3, we do not know of a conjectured answer to the following problem, which amounts to characterizing manifolds obtained via surgery on a knot in .
3. Hyperbolic examples with weight one fundamental groups
A group is weight if it can be normally generated by elements and no normal generating set with fewer elements exists. Recall that all knot groups are weight one and hence all manifolds obtained by surgery along a knot in have weight one fundamental groups. It is a folklore question if a manifold which admits a geometric structure and has a weight one fundamental group can always be realized as surgery along a knot in (see Reference 1, Question 9.23). The restriction to geometric manifolds is necessary since the fundamental group of is weight one, where is the Poincaré homology sphere. This cannot be surgery along a knot in since if a reducible manifold is surgery along a non-trivial knot in one of the factors is a lens space ,Reference 11.
In Theorem 3.4 we show that there are infinitely many hyperbolic 3-manifolds whose fundamental groups are normally generated by one element but which are not in the link of (Theorem 3.4). Our technique is a generalization to the hyperbolic setting of a method of Margaret Doig, who in Reference 8 first came up with examples that could not be obtained via surgery on a knot in using the Boyer and Lines -invariant.Reference 5 exhibited a different set of small Seifert fibered spaces which are weight one but not surgery along a knot in .
Before describing the hyperbolic examples, we make a few remarks regarding the weight one condition. We have the following obstruction to surgery due to James Howie:
This implies, for example, that is not obtained via surgery on a knot in , since its fundamental group is not weight one. However, when the , are pairwise relatively prime, its homology is cyclic.
The following proposition extends this consequence of Howie’s result to hyperbolic manifolds.
Over the two papers Reference 2Reference 3, Dave Auckly exhibited hyperbolic integral homology spheres that could not be surgery along a knot in However, it is unknown if these examples have weight one since his construction involves a 4-dimensional cobordism that preserves homology, but not necessarily group weight. .
Margaret Doig has recently exhibited examples of manifolds admitting a Thurston geometry, but which cannot be obtained by surgery along a knot in As part of a larger result, she shows: .
Although not explicitly stated in her result, for a finite group the weight of , is determined by the weight (see Reference 20), and so the above elliptic manifolds have weight one fundamental groups.
Using similar techniques and the work of Greene and Watson in Reference 13, we are able to exhibit hyperbolic manifolds that have weight one fundamental groups but are never surgery along a knot in As in Greene and Watson, our examples are the double branched covers of the knots . (see Figure 1a) where , and , We denote these knots by . and their corresponding double branched covers by The techniques of the proof may require us to omit finitely many of these double branched covers from the statement of the theorem. We will use . to denote the manifolds in this (possibly) pared down set.
In the following proof, we require two standard definitions from Heegaard-Floer homology (see Reference 28, Reference 8). First, a rational homology sphere is an L-space if the hat version of its Heegaard-Floer homology is as simple as possible, namely for each Spin structure of the hat version of , has a single generator and no cancellation. The d-invariant, is an invariant assigned to each ,Spin structure of which records the minimal degree of any non-torsion class of coming from see section 4 of ;Reference 28. Crudely, the can be thought of as a way of measuring how far from -invariant a manifold is. This mentality is motivated by the argument in the proof below.
For all , and for all but at most finitely many , .
Since we can produce the unknot by switching two crossing regions of the diagram for as in Figure 2, the Montesinos trick shows that can be obtained from surgery along a two component link in Hence, we see the upper bound . and is established for all but at most finitely many by Theorem 3.4.
■In Reference 22, Marengon extends the techniques given here to exhibit an infinite four parameter family of double branched covers of knots given by a Kaneobu like construction.
4. Complete infinite subgraphs
Here we discuss an interesting property which may allow one to “see” knots in the graph We also want to employ the notion of the set of neighbors of a vertex in a graph. More formally, for a graph . and a subset of the vertices of let , be the subgraph induced by these vertices. That is, the vertices of are and , is an edge of exactly when is an edge of Then, as in the introduction, we define the link a vertex . in to be for all , which are path length one from .
If is a knot in a 3-manifold then , where , is the set of 3-manifolds obtained from via Dehn surgery on .
For any closed -manifold and knot , is a .
That every vertex in is connected to every other one is a consequence of the definition. We just need to observe that there are infinitely many different manifolds in this subgraph. If admits a hyperbolic structure, then all but finitely many fillings are hyperbolic. Furthermore, the volumes approach the volume of and so there are infinitely many different homeomorphism types. If is Seifert-fibered (including the unknot complement in it is Seifert-fibered over an orbifold ), with boundary. The fillings can be chosen so that the result is Seifert-fibered over an orbifold where the boundary component of is replaced with a cone point of arbitrarily high order, so the Seifert-fibered spaces are not homeomorphic. If admits a decomposition along incompressible tori, then infinitely many fillings have this same decomposition Reference 12. Then the boundary of is in either a hyperbolic piece or a Seifert-fibered piece, and the above arguments apply. Finally, if is reducible, then there exists a separating such that where is irreducible. In this case, the previous arguments can be applied to yield the desired result.
■Note that, sometimes, the intersection of two subgraphs arising from fillings on knot complements may intersect in a For example, let . be the unknot and a torus knot. Let be the associated to and , be the associated to Then . is a where each vertex is a lens space (see Reference 24). However, this phenomena cannot happen for hyperbolic manifolds.
If and are hyperbolic and not homeomorphic, then the subgraphs and have at most finitely many vertices in common.
Assume that and have infinitely many vertices in common. Then infinitely many of these are hyperbolic. Denote this set by Choose a basepoint in the thick part of each . Then the geometric limit of the . is and it is also so they must be homeomorphic. See ,Reference 14 for background on geometric limits.
■4.1. Subgraphs which do not arise from filling
There is a of small Seifert-fibered spaces that does not come from surgery along a one cusped manifold.
We will construct a family , of Seifert-fibered spaces over an orbifold with base space , and negative Euler characteristic. We follow notation in Reference 15. In particular, we denote a closed Seifert-fibered space by where is the underlying space of the base orbifold. The cone points of the base orbifold will have multiplicities The Seifert-fibered invariants of the exceptional fibers are . which are allowed to take values in , Two Seifert fiberings . and are isomorphic by a fiber preserving diffeomorphism if and only if after possibly permuting indices, and, if is closed, Reference 15, Proposition 2.1.
Now let be four distinct rational numbers, such that , and are relatively prime. Let be the Seifert-fibered space over with three exceptional fibers labeled by We can define . , , similarly. The condition ensures that each manifold will be Seifert-fibered over a hyperbolic orbifold.
Note that each manifold has exactly two common exceptional fibers with the others mod 1, and manifolds with fibrations over hyperbolic base orbifolds have unique Seifert-fibered structures Reference 32, Theorem 3.8.
The set of manifolds form a in .
We will now construct a which consists of infinitely many of these Note that if we add . to each Seifert invariant of each exceptional fiber above, we get another Each new manifold is distinct from the manifolds in the previous . since the sum of its Seifert invariants is not equal to any vertex in the original. Each vertex in the new is connected to each vertex of the previous as, for example, Dehn surgery along one of the exceptional fibers can result in any manifold which is a vertex of the original . Continuing this way, we have a . parametrized by , where , and Specifically, . is as follows:
Assume this comes from filling a one cusped manifold First, . must be irreducible. Indeed, if it were reducible, there would be a two-sphere that did not bound a ball in If the sphere is non-separating it will remain non-separating in any filling. If it is separating, there is at most one filling of a knot in a ball which will make it a ball .Reference 11.
Next we observe that each is a small Seifert-fibered space and in particular non-Haken. We claim that we may assume does not contain an essential torus. Indeed, if does contain an essential torus then that torus compresses in infinitely many fillings. Infinitely many fillings cannot be pairwise distance 1 and thus by ,Reference 7, Theorem 2.01, and cobound a cable space, Surgeries on cable spaces are well understood. As in .Reference 4, p 179, the filling of the cable space is either reducible (only along the cabling slope), a solid torus, or a manifold with an incompressible boundary torus. Since is compressible in the fillings, bounds a solid torus in each of the filled manifolds. Therefore, we can replace filling along by filling along the torus boundary of and get the same set of manifolds.
Thus we may assume does not contain an essential torus and since it is irreducible is geometric.
If is not hyperbolic, it is a Seifert-fibered space containing no essential tori. Thus must be Seifert-fibered over the disk with at most two exceptional fibers. Each admits a unique Seifert fibration (see Reference 32, Theorem 3.8). Any choice of two elements to label the exceptional fibers of will disagree with two exceptional fibers in one of the which is a contradiction to the existence of such an , Thus . must be hyperbolic. However, there are infinitely many Seifert-fibered manifolds that come from surgery on This contradicts Thurston’s Hyperbolic Dehn Surgery Theorem .Reference 33, Theorem 5.8.2.
■5. Seifert-fibered spaces and Solv manifolds
Before we discuss the hyperbolic part of the graph, we briefly discuss other geometries and Seifert-fibered spaces. By compiling work of Montesinos and Dunbar, we can obtain upper bounds for any non-hyperbolic geometric manifold. The general idea is that “simpler” manifolds lie close to We begin with a theorem of Montesinos .Reference 23. is the Euler characteristic of .
Let be a closed, orientable Seifert-fibered space over the surface with exceptional fibers.
- (1)
If is orientable, then .
- (2)
If is non-orientable, then .
This theorem follows from the discussion in Reference 23, Chapter 4 (see specifically Figure 12 in that chapter). Since each link in that figure has a component labeled by mild Kirby Calculus can be applied to the links in that figure to obtain a link with one fewer component. ,
To understand the geometric non-hyperbolic manifolds, it remains to investigate Solv manifolds, which are either torus bundles over or the union of twisted bundles over the Klein bottle (see Reference 32, Theorem 4.17). Work of Dunbar provides an orbifold analog to this statement, namely if is an orientable Solv orbifold, is either a manifold as above or an orbifold with fiber over or the union of twisted bundles with fiber (see Reference 9, Propostion 1.1). Using these two results, we can obtain the following:
(1) If is a Solv torus bundle, then .
(2) If is the union of twisted bundles over the Klein bottle admitting an orientable Solv structure, then .
(1) By Reference 9, admits a 2-fold quotient such that the base space of is and the singular locus is a four strand braid Although that paper is careful to classify such braids, the details will not be relevant to this argument. Using the Montesinos trick, we have a sequence of tangle replacements to get from . to the trivial two strand braid in The first two replacements of this sequence are shown in Figure .3a. The resulting link is two-bridge and therefore a single rational tangle replacement yields the unknot. The trivial two strand braid can be obtained from a single rational tangle replacement on the unknot. Hence, and .
(2) Let be the union of twisted bundles over the Klein bottle admitting an orientable Solv structure. Then is the 2-fold quotient of a Solv torus bundle. Moreover, is the index 2 subgroup of elements that preserve the orientation of every fiber of and we may consider where is the composition of a translation in a fiber and a symmetry of Solv taking the form or .
Denote by the 2-fold quotient of by The base space of . is and the singular set is isotopic to the link picture in Figure 3b. The rational tangle replacements in that figure yield a two-bridge link and so the double branched cover of the resulting link is a lens space. A lens space is path length one from completing the proof. ,
■Immediately from this section, we have that if is a closed orientable 3-manifold which admits a Nil, , , or Solv geometry, then However, these upper bounds are not in general known to be sharp. The reader is referred to Margaret Doig’s work .Reference 8 for a more comprehensive treatment of which manifolds admitting an elliptic geometric structure can be obtained from surgery along a knot in .
6. The subgraph for hyperbolic manifolds
Let be the subgraph of such that the vertices correspond to closed hyperbolic 3-manifolds, and there is an edge between two vertices and exactly when there is a one-cusped hyperbolic 3-manifold with two fillings homeomorphic to and .
Note that there is not necessarily a hyperbolic Dehn surgery between and in our definition. For example, the surgery knot might not be represented by embedded geodesics in and .
As mentioned above in §4, this part of the graph has the nice property that if two different graphs that arise as and intersect, they must do so in finitely many vertices. We conjecture that the combinatorics of this subgraph may reveal more of geometry and topology than in the full graph. For the same reasons as , is infinite valence and infinite diameter. We show here that it is connected, using the work of Myers. Let be a compact orientable 3-manifold, possibly with boundary. Following Myers, we say that is excellent if it is irreducible and boundary irreducible, not a 3-ball, every properly embedded incompressible surface of zero Euler characteristic is isotopic into the boundary, and it contains a two-sided properly embedded incompressible surface. These manifolds are known by Thurston Reference 34, Theorem 1.2 to admit hyperbolic structures. By slight abuse of notation, if a properly embedded 1-manifold has an excellent exterior, we will call excellent.
Let be a compact connected whose boundary does not contain -manifold or projective planes. Let -spheres be a compact properly embedded in -manifold Then . is homotopic rel to an excellent -manifold .
For the following lemma, we observe Myers’ notation in stating the following technical lemma. Namely, let be a 3-manifold and be a compact, (possibly disconnected) properly embedded two-sided surface in Let . be the manifold obtained by cutting along and let be such that identifying and yields .
If each component of is excellent, and are incompressible in and each component of , has negative Euler characteristic, then is excellent.
Suppose and are closed hyperbolic such that the associated vertices -manifolds and are connected via a path of length in Then . and are connected via a path of length in .
Observe that, under this hypothesis, there is an link, -component in and closed manifolds such that
We will find a knot in and a slope such that the closed manifolds
are hyperbolic. Each is obtained from via Dehn surgery on with slope The knot . and slope will also have the property that the one-cusped manifolds
and
are hyperbolic. We will use Myers’ Theorem 6.2 and Lemma 6.3, stated above. We will also use the fact, proven in Lemma 6.5, that, given a and two slopes and on there is an arc , in with endpoints on such that the exterior of in is excellent. Furthermore, the results of Dehn filling along the slopes and are excellent.
Now we prove the existence of a knot in the exterior of the link in with the desired properties. First fix a homeomorphism of a neighborhood of each with For each component . we construct an arc , in such that: (i) (ii) the exterior of ; in is excellent; and (iii) the results of filling the exterior of along the slopes and on are excellent. This is done in Lemma 6.5 below. Now by Myers’ Theorem, stated as Theorem 6.2 above, there is an excellent collection of arcs in such that connects an endpoint of to one of Then we claim the following: .
- (1)
is an excellent knot in .
- (2)
The result of filling along any choice of or for any subset of the is excellent.
The fact that the union of arcs in (1) above is a knot follows from the recipe. The fact that the exterior in (1) is excellent follows from Myers’ Lemma 6.3 above and the fact that each is excellent and that the exterior of the union of the is excellent. Similarly, since each filled along or is excellent, Myers’ Gluing Lemma 6.3 yields that filling any subset of the along or is excellent. Thus, in particular, and above are hyperbolic.
Let be a knot in having property (1) above. Choose a slope on such that lies outside of the finite set of slopes that makes any one of the closed manifolds or the cusped manifolds not hyperbolic.
Then the path is a path in connecting and Here the . and are closed hyperbolic manifolds (represented by vertices in and the ) and are cusped hyperbolic manifolds (represented by edges in ).
■Given and two isotopy classes of curves and on there is an arc , with endpoints on such that:
- (1)
is excellent.
- (2)
The results of filling along the slopes and are excellent.
By Myers’ Theorem 6.2, there exists an arc in with endpoints on such that the exterior is excellent. The arc we will use is wrapped around enough to make filling along two specified slopes , and hyperbolic. We detail this wrapping around below.
Fix up to isotopy. Let be an oriented slope on Let . be an essential annulus bounded by and a curve on Let . be a slope that has intersection number 1 with There are homeomorphisms . obtained by cutting along and twisting once, and then gluing back by the identity on this annulus. We twist so that an oriented , and in the original isotopy class of , Furthermore, given an . and an oriented slope there is an , which is a composition of , and such that the oriented intersection of and and and is larger than .
Now let be the exterior of in There is a subsurface . of the boundary such that it and its complement are incompressible in Thus we may apply Myers’ Gluing Lemma (Lemma .6.3) to the double along , and conclude that it is excellent, hence hyperbolic. The manifold is the exterior of a knot in We say that filling along the components . and such that the filling is the double along of a filling of is a symmetric filling. Then, by Thurston’s Hyperbolic Dehn Surgery Theorem Reference 33, Theorem 5.8.2, all but finitely many symmetric fillings of are hyperbolic. (Note that the filling curves have the same length in the complete structure on The maps .) and extend naturally to (by restriction to and doubling) and take symmetric slopes to symmetric slopes. Thus there is a map which can be taken to be of the form , such that filling , symmetrically along and is hyperbolic. Then the arc in has the property that filling along and is hyperbolic. Indeed, doubling the exterior of results in which is hyperbolic when symmetrically Dehn filled along and .
■7. Obstructions to hyperbolicity
We recall the following definitions. A geodesic metric space is if every geodesic triangle is -hyperbolic“ thin”, that is, every side is contained in a of the union of the other two sides. Two metric spaces -neighborhood with metrics are quasi-isometric if there exists a function and such that (1) for all , and (2) every point of lies in the of -neighborhood A k-quasi-flat in a metric space . is a subset of that is quasi-isometric to In this section we will construct quasi-flats in . and showing that these spaces are not , -hyperbolic.
We will need to compute the exact distance in some simple examples. To do so, we first give a method for a lower bound on the distance.
Let and be closed orientable and let -manifolds and a prime. If and but then ,
Let be a knot in a closed manifold and let , be a word in We claim that if . is a surjection, then induces a surjection from to or Indeed, the image of . under is either trivial or non-trivial. If it is trivial, then induces a surjection to If . is non-trivial, then it is order in since every element is order , Then there is a minimal generating set of . where is a basis element. Then is a surjection. This proves the claim.
We note that if surjects then , does as well, since there is a presentation of the two groups which differs only by the addition of a relation. Then the claim implies that the maximum such that surjects can change by at most 1 under the operation of Dehn surgery along a knot in and the lemma follows. ,
■contains a Hence -quasi-flat. is not -hyperbolic.
For each let , be the unlink in with components with the natural homological framing. Then we will consider the manifolds:
This represents a in -quasi-flat In particular, the distance between the manifolds in the figure can be determined using the edges in the figure. .
In other words, the surgeries on the first components are either or with surgery on the first , components being Let . and be distinct primes. The surgeries on the second components are either or with the first , being The number of non-trivial fillings of . is of , is and of , is Thus we can take . as large as needed and these are still well defined.
Then:
Then, by repeated use of Lemma 7.1, since every map to an abelian group factors through the homology, the distances between these manifolds are as in the diagram.
■Using the same methods, we can show the following.
has a based at -quasi-flat Furthermore, . has a based at each vertex -quasi-flat .
By choosing four distinct primes and the graph , can be seen to exhibit a large quasi-flat based at The vertices of such a quasi-flat are: .
In fact, if the manifolds and are connect summed with a given closed orientable manifold then by the same homology argument as above, there is a quasi-flat based at , .
■This construction can be adapted to construct for arbitrarily large -quasi-flats .
The behavior of homology under Dehn filling is a key property of the component split link in the argument above. The pairwise linking number of the components of that link is 0. The rest of this section will be devoted to finding a component link that has similar behavior. First, we construct a hyperbolic link where each one of the pairwise linking numbers is This is accomplished as follows: .
There is a knot in the complement of the component split link such that is hyperbolic and all components have pairwise linking number .
The proof is similar to the methods in §6. Consider the link exterior and label each boundary component by Let . be a neighborhood of each in By .Reference 26, we can drill out a set of excellent arcs from such that connects a neighborhood of the component with the th component and st connects the last component to the first. Furthermore, orient each such that is based on a neighborhood of the component. For convenience denote the last arc by th and If . is the disk in with as a boundary, the union of the the arcs in will have oriented intersection number with .
Let With a slight abuse of notation we consider . as homeomorphic to with the marked annulus embedded in it. Again by Reference 26, we can drill out an excellent arc from in any homotopy class, and hence with any intersection number with connecting the endpoints of , and Here, we choose . to be this intersection number.
Let be an oriented knot in For each component . of the disk , is also a Seifert surface for Thus, the pairwise linking number of . and is the oriented intersection number of with , .
■In the above proof, there is a special component of the link, such that drilling out , from is hyperbolic. We call this component of the link the Myers component. For a general link an -component must be specified to determine the Myers component. -component
is not -hyperbolic.
As above, we construct a quasi-flat. Using the link as in Lemma 7.5, we have that is hyperbolic and each pair of components has linking number This condition implies that an embedded curve, is null homologous in ,
since the homology class of is determined by the sum of the oriented mod intersection number with the Seifert surface of the component of th One can observe this directly by consideration of . as a curve in .
Thus,
and so we can choose surgery coefficients such that the homology of the fillings behaves analogously to the manifolds , and , as in the proof of Theorem 7.2.
Finally, we remark that choosing sufficiently large choices of primes and and a large choice of the manifolds obtained by filling the first , components of by either or is hyperbolic by Thurston’s Hyperbolic Dehn Surgery Theorem Reference 33, Theorem 5.8.2.
■Acknowledgements
Both authors have benefited from many conversations with colleagues. We would particularly like to thank Margaret Doig for suggesting that we look at the manifolds in Reference 13. We are also grateful to Steven Boyer, Nathan Dunfield, Marc Lackenby, Tao Li, Luisa Paoluzzi, and Richard Webb. The first author was partially supported by ARC Discovery Grant DP130103694 and the Max Planck Institute for Mathematics. The second author was partially supported through NSF Grant 1207644 and a Tufts University Faculty Research Award. We particularly thank the referee for many helpful comments.