Generalized Fibonacci groups H(r,n,s) that are connected Labelled Oriented Graph groups
Gerald Williams

TL;DR
This paper classifies when generalized Fibonacci groups H(r,n,s) are connected Labelled Oriented Graph groups, linking them to knot groups and using circulant matrices to analyze their properties.
Contribution
It provides an almost complete classification of H(r,n,s) groups as connected LOG groups, identifying all torus knot groups and the infinite cyclic group within this class.
Findings
All torus knot groups are connected LOG groups.
H(r,n,s) is a 2-generator knot group if and only if it is a torus knot group.
Conjecture: only torus knot groups and the infinite cyclic group are connected LOG groups.
Abstract
The class of connected LOG (Labelled Oriented Graph) groups coincides with the class of fundamental groups of complements of closed, orientable 2-manifolds embedded in S^4, and so contains all knot groups. We investigate when Campbell and Robertson's generalized Fibonacci groups H(r,n,s) are connected LOG groups. In doing so, we use the theory of circulant matrices to calculate the Betti numbers of their abelianizations. We give an almost complete classification of the groups H(r,n,s) that are connected LOG groups. All torus knot groups and the infinite cyclic group arise and we conjecture that these are the only possibilities. As a corollary we show that H(r,n,s) is a 2-generator knot group if and only if it is a torus knot group.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsGeometric and Algebraic Topology · Homotopy and Cohomology in Algebraic Topology · Advanced Combinatorial Mathematics
Generalized Fibonacci groups that are connected Labelled Oriented Graph groups
Gerald Williams
Abstract
The class of connected LOG (Labelled Oriented Graph) groups coincides with the class of fundamental groups of complements of closed, orientable 2-manifolds embedded in , and so contains all knot groups. We investigate when Campbell and Robertson’s generalized Fibonacci groups are connected LOG groups. In doing so, we use the theory of circulant matrices to calculate the Betti numbers of their abelianizations. We give an almost complete classification of the groups that are connected LOG groups. All torus knot groups and the infinite cyclic group arise and we conjecture that these are the only possibilities. As a corollary we show that is a 2-generator knot group if and only if it is a torus knot group.
1 Introduction
A Labelled Oriented Graph (LOG) consists of a finite graph (possibly with loops and multiple edges) with vertex set and edge set together with three maps called the initial vertex map, terminal vertex map, and labelling map, respectively. The LOG determines a corresponding LOG presentation
[TABLE]
A group with a LOG presentation is called a LOG group [18]. When the underlying graph is connected we have a *connected LOG, a connected LOG presentation, and a connected LOG group. A -knot group *() is the fundamental group of the complement of an -sphere in ; we refer to a -knot group as a knot group. The Wirtinger presentation of a knot group is a connected LOG presentation and so all knot groups are connected LOG groups; in particular, the infinite cyclic group is a connected LOG group. Further examples of LOG groups include right angled Artin groups and braid groups. Clearly the abelianisation of a LOG group is torsion-free, and the abelianisation of a connected LOG groups is the infinite cyclic group.
As pointed out in [14],[13], by [28] a group is a connected LOG group if and only if it is the fundamental group of the complement of a closed, orientable 2-manifold embedded in . It follows that every -knot group is a connected LOG group ([16]). A particular instance of [16, Theorem 1.1] is that, given a finite presentation of a group, there is no algorithm that can decide if that group is a connected LOG group. Another is that, given a finite presentation of a group, there is no algorithm that can decide if that group is a knot group (see also [23, Theorem 9.2.1],[29]).
A *cyclically presented group *is a group defined by a presentation of the form
[TABLE]
where is some fixed element of the free group . Connections between HNN extensions of cyclically presented groups and LOG groups have been investigated in [14],[31],[19]. Asphericity of certain cyclic presentations that are (connected) *word labelled oriented graph (WLOG) presentations *is established in [17, Section 3]. In this paper we investigate a particular family of cyclically presented groups and aim to classify when they are connected LOG groups or when they are knot groups. Namely, we investigate the generalized Fibonacci groups
[TABLE]
where , , and subscripts are taken mod , that were introduced in [6]. Setting we get the Fibonacci groups introduced in [9].
These groups have been considered from algebraic and topological perspectives. Finite groups have been obtained in [6],[5],[3, Corollary E],[4, Corollary 11]; conditions under which is infinite are given in [6],[7]; and conditions under which is large, SQ-universal, or contains a non-abelian free subgroup can be extracted from [32]. Asphericity of the presentations is considered in [26, Theorem 3]. A class of groups that are fundamental groups of closed 3-manifolds was obtained in [30, Proposition 3]. In the opposite direction conditions under which is not the fundamental group of a hyperbolic 3-dimensional orbifold of finite volume are given in [8, Corollary 3.2]. Corollary 5.5 of [8] gives that, for , the natural HNN extension of the group is a 3-knot group if and only if . Note that by inverting the relators, replacing each generator by its inverse, and negating the subscripts we have that .
Our main result is the following (recall that a group is *perfect *if ).
Theorem A**.**
Let , . If is a connected LOG group then one of the following holds:
- (a)
* and in which case , the fundamental group of the complement of the -torus knot in ;*
- (b)
, and either mod or mod , in which case ;
- (c)
, , , , mod , mod , and the group is perfect.
We conjecture that condition (c) cannot hold.
Conjecture 1**.**
Let , , mod , mod . Then .
Knots (i.e. complements of in ) for which the minimum number of generators required to generate the corresponding knot group is equal to two are called *2-generator knots *and the corresponding knot group is a *2-generator knot group. *Since the only knot for which the corresponding group is cyclic is the unknot, 2-generator knots are, from one perspective, the ‘simplest’ non-trivial knots. All tunnel number one knots (in particular all torus knots) are 2-generator knots, and it has been conjectured that all 2-generator knots are tunnel number one knots (in [1] this is attributed to Scharlemann [27], who attributes it to Casson; see also [25, Conjecture 3.9]). The conjecture has been shown to hold for cable knots [1] and the satellite knots that have a two-generator presentation in which at least one generator is represented by a meridian for the knot are classified in [2]. As a corollary to Theorem A we classify when is a 2-generator knot group.
Corollary B**.**
Let , . Then is a 2-generator knot group if and only if and , in which case , the fundamental group of the complement of the -torus knot in .
Since connected LOG groups abelianize to the infinite cyclic group , the abelianisation of is of interest to us. Any finitely generated abelian group is isomorphic to a group of the form where is a finite abelian group and . The number is called the *Betti number *(or torsion-free rank) of , and we write to denote the minimum number of generators of . Clearly is infinite if and only if and if is a connected LOG group then . Theorem 1 of [6] asserts that for we have that if and only if the greatest common divisor . In Theorem C we generalize this to give the value of in all cases.
Theorem C**.**
Let , .
- (a)
If then ;
- (b)
if then .
In support of Conjecture 1 we have the following.
Corollary 2**.**
Let (resp. ). Then there are at most finitely many values of , (resp. ) such that .
Remark 3**.**
It is not hard to prove that for all , so if there is a choice of , , mod , mod , such that then there are infinitely many such choices of such that . Therefore Corollary 2 does not imply that there are at most finitely many , such that .
The proofs of Theorem C and Corollary 2 use the theory of circulant matrices. The circulant matrix is the matrix whose first row is and where each subsequent row is a cyclic shift of its predecessor by one column. Thus if, for each , the exponent sum of in is then the relation matrix of is the circulant matrix . The *representer polynomial *of is the polynomial
[TABLE]
and we define . It is well known that
[TABLE]
and so this is the order when it is non-zero, and is infinite otherwise. This fact has long been used in the theory of cyclically presented groups (see [21]) and, in particular, it was used to obtain [6, Theorem 1]. The rank of can also be expressed in terms of the polynomials ; specifically
[TABLE]
where denotes the degree (see [20, Proposition 1.1] or [24, Theorem 1]) and so
[TABLE]
and this is the engine of the proof of Theorem C. While the formula (2) is old, we believe that it has not been applied to cyclically presented groups before. Moreover, we expect it to be of independent interest in studying other classes of cyclically presented groups, and with wider applications than that considered here. The proof of Corollary 2 uses a result from [11] for determining when ; that is, when is a unimodular matrix.
2 A class of groups that are knot groups
If is odd then
[TABLE]
The relations () imply the ’th relation , so this redundant relation may be eliminated to give the generator, relation presentation
[TABLE]
of . This is precisely the Dehn presentation for the -torus knot (see, for example, [15, page 155], where the case is illustrated).
Furthermore, if is odd, then the following sequence of relations are implied by the relations of (4)
[TABLE]
and, in particular, . Conversely, the relations () imply the sequence of relations
[TABLE]
and, in particular, . Thus the group is also given by the presentation
[TABLE]
which is a (cyclic) Wirtinger presentation for the -torus knot ([15, pages 151–153]) arising from a LOG where the underlying graph is a cycle.
Thus if is odd we have that is the fundamental group of the complement of the -torus knot in . More generally we have the following.
Lemma 4**.**
Let , . If then , the fundamental group of the complement of the -torus knot in .
Proof.
Since there exist such that . Let , then
[TABLE]
Now
[TABLE]
and
[TABLE]
so we may add the relations and to get
[TABLE]
The relations imply so . Therefore
[TABLE]
so the relation is equivalent to which is equivalent to , i.e. to , so the set of relations is equivalent to the set of relations (). Similarly we have
[TABLE]
so the relation is equivalent to , which is equivalent to , i.e. to . Thus
[TABLE]
Now
[TABLE]
so we may add the relation and then eliminate the redundant relations to get , as required. ∎
3 Betti numbers and perfect groups
In this section we prove Theorem C and Corollary 2.
Proof of Theorem C.
Let , , , . The representer polynomial of is
[TABLE]
Since we may assume . If then
[TABLE]
That is,
[TABLE]
where
[TABLE]
when and
[TABLE]
when . By (3) we must find the degree of the highest common factor of and . Observe that
[TABLE]
where
[TABLE]
Therefore .
Suppose then is as given at (6). Now, writing to denote the th cyclotomic polynomial, we have
[TABLE]
since . Therefore so , which is of degree , as required.
Suppose then that so is as given at (7). We must show . If , then is not a root of (for otherwise we get a contradiction to ). Therefore is a root of and of if and only if it is a root of and of . Assume is such a root. Then, after simplifying, the equations , imply
[TABLE]
Since we have for some , and so . Taking the complex conjugate of the first equation then gives
[TABLE]
Multiplying the equations and and simplifying gives , so . Similarly, multiplying the equations and and simplifying gives , so and hence . Therefore we have , or equivalently . But so , contradicting the fact that is a root of . Therefore have no common roots so , as required. ∎
Proof of Corollary 2.
Let . By (1) if for some then (where is as given at (5)) and so and in particular and there are at most two possible values of , namely . If has a cyclotomic factor , then this is also a factor of so . But so Theorem C(a) (or [6, Theorem 1]) gives that and so by (3) we have , a contradiction. Therefore has no cyclotomic factors. Since also , Theorem 1 of [11] implies that the set of integers such that the relation matrix of the presentation has determinant equal to or is finite. Therefore the set of integers for which is finite. ∎
4 Minimum number of generators for
Lemma 5**.**
The minimum number of generators
[TABLE]
Hence if is a connected LOG group then if is even and if is odd.
Proof.
Using an idea from the proof of [6, Lemma 4], we see that the group maps onto
[TABLE]
The remainder of the proof is similar to that of [33, Theorem C]. Let . Then there exist such that so mod . The relation implies so and hence . But so we have for each . Eliminating generators gives
[TABLE]
where if is even and otherwise. Hence maps onto so . If is a connected LOG group then , and the result follows. ∎
In connection with Conjecture 1 we record the following:
Corollary 6**.**
If is perfect then and .
Proof.
If is perfect then , where is the representer polynomial for given at (5); that is . By Lemma 5 if then is not perfect. ∎
Remark 7**.**
When and , computer experiments using GAP [12] indicate that the order is often a product of large primes so straightforward quotient methods, such as those employed in the proof of Lemma 5, are unlikely to suffice for proving Conjecture 1 in general. Since is increasing in (e.g. [10]), one might hope to be able to prove Conjecture 1 by showing that for any the order is increasing in ; however, this is not the case since, for example, and .
We also note the following corollary to Lemma 5 which generalizes [6, Lemma 4] (which deals with the case mod ). It follows immediately from Theorem 9(i) of [22] which states that if a group defined by a balanced presentation is finite then .
Corollary 8**.**
If either
- (a)
* is even and ; or*
- (b)
* is odd and *
then is infinite.
Lemma 9**.**
Let , and suppose that . If then , and hence is not a connected LOG group.
Proof.
The abelianisation maps onto
[TABLE]
which is non-cyclic if , and hence . ∎
Lemma 10**.**
Let , and suppose that and .
- (a)
If then .
- (b)
The group is not a 2-generator knot group.
Proof.
Since we may assume , and so . Let and let . Then
[TABLE]
Therefore maps onto , so if this is non-cyclic so , proving part (a). For part (b), suppose for contradiction that is a 2-generator knot group; then in particular so by part (a) we have that . If is non-trivial then it is non-cyclic so , and hence , is not a 2-generator group, a contradiction. Therefore and it follows from (36) that , a contradiction. ∎
5 Proof of Theorem A and Corollary B
First observe the following:
Lemma 11**.**
Let , and suppose mod or mod . Then
[TABLE]
Hence is a connected LOG group if and only and , in which case .
Proof.
Since we may assume that mod . Then the set of relations of are () which, after cancelling, become the set of relations . That is, which by [33, Theorem C] is isomorphic to
[TABLE]
If is a connected LOG group then which happens if and only if and (equivalently and , since mod ), in which case . ∎
Proof of Theorem A.
If is a connected LOG group then so by Theorem C we may assume that either and or and .
In the first case the result follows from Lemma 4 so assume and . If we have that mod or mod then the result follows from Lemma 11, so we may assume further that mod and mod . If then is not a connected LOG group by Lemma 9, and if and then is not a connected LOG group, by Lemma 10. Now mod so is odd and by Lemma 5 we may assume so . If then and it is well known (see, for example, [10]) that so we may assume further that . Thus we have that mod , , , , and , as in part (c). ∎
Proof of Corollary B.
If is a 2-generator knot group then is a connected LOG group so one of the conclusions (a),(b),(c) of Theorem A hold. In (b) we have that , a contradiction, and in (c) is not a 2-generator knot group by Lemma 10. ∎
Acknowledgements
I thank Bill Bogley for helpful comments on a draft of this article.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] Steven A. Bleiler. Two-generator cable knots are tunnel one. Proc. Am. Math. Soc. , 122(4):1285–1287, 1994.
- 2[2] Steven A. Bleiler and Amelia C. Jones. On two generator satellite knots. Geom. Dedicata , 104:1–14, 2004.
- 3[3] William A. Bogley and Gerald Williams. Efficient finite groups arising in the study of relative asphericity. Math. Z. , 284(1):507–535, 2016.
- 4[4] William A. Bogley and Gerald Williams. Coherence, subgroup separability, and metacyclic structures for a class of cyclically presented groups. J.Algebra , 480:266–297, 2017.
- 5[5] A.M. Brunner. On groups of Fibonacci type. Proc. Edinb. Math. Soc., II. Ser. , 20:211–213, 1977.
- 6[6] C.M. Campbell and E.F. Robertson. On a class of finitely presented groups of Fibonacci type. J. Lond. Math. Soc., II. Ser. , 11:249–255, 1975.
- 7[7] Colin M. Campbell and Richard M. Thomas. On infinite groups of Fibonacci type. Proc. Edinb. Math. Soc., II. Ser. , 29:225–232, 1986.
- 8[8] Alberto Cavicchioli, Dušan Repovš, and Fulvia Spaggiari. Topological properties of cyclically presented groups. J. Knot Theory Ramifications , 12(2):243–268, 2003.
