betaFeedback

WorksMathematics

Meridional ranks of iterated Whitehead doubles

Abstract

We prove that every iterated untwisted Whitehead double of a nontrivial two-bridge knot satisfies the meridional rank conjecture. More precisely, for such a knot KK and every N≥0N\geq0, the augmentation rank, meridional rank, and bridge number of Wh⁡N(K)\operatorname{Wh}^N(K) are all 2N+12^{N+1}. We use microlocal sheaf theory and a doubling construction to produce a lower bound for the meridional rank of the Whitehead double, assuming the peripheral data satisfy suitable conditions. For two-bridge knots, the required curves arise from ideal points detecting nonzero boundary slopes. In contrast, the ordinary group rank of an NN-fold Whitehead double is at most the rank of the companion group plus NN. We also give a non-example when the doubling construction does not work – the first Whitehead double of the square knot has augmentation rank at most 5, whereas its meridional rank and bridge number are 6.

Introduction

Let K⊂S3K \subset S^3 be a knot and let πK=π1(S3∖K)\pi_K = \pi_1(S^3 \setminus K). The meridional rank mr⁡(K)\operatorname{mr}(K) is the minimum number of meridians needed to generate πK\pi_K. A bridge presentation gives a meridional generating set, so mr⁡(K)≤b(K)\operatorname{mr}(K) \le b(K), where b(K)b(K) denotes the bridge number. The meridional rank conjecture of Cappell–Shaneson asks whether equality holds for every knot. It appears in Kirby’s problem list [27, Problem 1.11] and, in its link formulation. See also [3, Problem 1.18(a)].

The conjecture is known for several classes of knots and links. Boileau–Zieschang proved it for Montesinos links by using their twofold branched covers [10], and Rost–Zieschang proved it for torus links [31]. The characterization of knots whose groups can be generated by two meridians is due to Boileau–Zimmermann [11, Corollary 3.3], together with the subsequent orbifold geometrization theorem [9].

Diagrammatic descriptions of bridge number clarify a difficulty in the conjecture. Blair–Kjuchukova–Velazquez–Villanueva [6] proved that the bridge number equals the Wirtinger number, the minimum number of seed strands needed for successive Wirtinger generation over all diagrams. Blair–Kjuchukova–Pfaff allow relations witnessed by certain embedded two-spheres; their plain sphere number also equals the bridge number [5]. To deduce the conjecture from these descriptions, one would still need to pass from an arbitrary meridional generating set to such a system without increasing its size.

For lower bounds, the conjugacy class of a meridian is essential. Ordinary group rank gr⁡(K)\operatorname{gr}(K) allows arbitrary generators and can be much smaller: a nontrivial torus-knot group has rank two, whereas its meridional rank can be arbitrarily large [31]. A quotient retaining the meridian class can give a stronger bound. Baader–Blair–Kjuchukova use meridian-to-reflection Coxeter quotients to prove the conjecture for twisted links and arborescent links associated to bipartite trees with nonzero even weights [1]. Baader–Kjuchukova [2, Section 2] use symmetric-group quotients to prove the meridional rank conjecture for a family of twisted Whitehead doubles of connected sums of T(2,5)T(2,5), with arbitrarily large bridge number. Recent work of Blair–Kjuchukova–Pfaff extends this approach to fishnet families with any prescribed bridge number at least three and arbitrarily large bridge distance relative to that number [4]. The question of what finite quotients can detect is also recorded in [3, Problem 1.18(b)].

Knot contact homology gives a different lower bound. Cornwell [15, Theorems 1.2 and 1.5] relates its augmentations to representations of the knot group and proves

ar⁡(K)≤mr⁡(K)≤b(K),(1)\operatorname{ar}(K) \le\operatorname{mr}(K) \le b(K), \tag*{(1)}

where ar⁡(K)\operatorname{ar}(K) denotes augmentation rank. In the corresponding microlocal rank one representations, a meridian acts by a diagonalizable rank-one perturbation of the identity. For an irreducible microlocal rank one representation ρ\rho, the images of ρ(mi)−I\rho(m_i)-I for any meridional generating set {mi}\{m_i\} span the representation space. Its dimension, therefore, bounds the meridional rank from below. We will vary these representations and their peripheral eigenvalues along an algebraic curve to choose data for successive Whitehead doubles. We recall the definitions in Section 2.1.

We study these ranks under untwisted Whitehead doubling. Two satellite results are particularly relevant. For satellites with braid patterns, Cornwell–Hemminger prove a maximal augmentation-rank multiplication theorem and apply it to the iterated torus knots satisfying their hypotheses [16, Theorem 1.2 and Corollary 1.3]. The Whitehead pattern has winding number zero, so that theorem does not apply directly. Using the action on the tree associated to a satellite decomposition, Boileau–Dutra–Jang–Weidmann control the peripheral intersections of meridional subgroups. They prove the conjecture for the class generated from meridionally tame knots by connected sums and satellites with braid patterns [7, Theorem 1]. This includes every knot whose exterior is a graph manifold [7, Corollary 1]. However, Whitehead doubles of nontrivial knots are not meridionally tame [7, Introduction and Section 6], so this theorem does not directly include Whitehead iteration.

Dutra’s preprint [18, Theorem 1.1] states the meridional rank conjecture for (once) Whitehead doubles of prime algebraically tame knots. Even with this conclusion, iteration would require a new argument: the first Whitehead double is no longer algebraically tame [18, Example 2.8].1 Our contribution is a sheaf-level doubling construction whose iteration is controlled on peripheral curves realized by the companion. By keeping track of both peripheral eigenvalues, we can choose representations for which every required doubling step is defined.

Throughout, Wh⁡(K)\operatorname{Wh}(K) denotes the untwisted Whitehead double with the pattern and peripheral conventions fixed in § 2.2. We set Wh⁡0(K)=K\operatorname{Wh}^{0}(K)=K and Wh⁡N+1(K)=Wh⁡(Wh⁡N(K))\operatorname{Wh}^{N+1}(K)=\operatorname{Wh}(\operatorname{Wh}^{N}(K)). For a nontrivial companion,

b(Wh⁡(K))=2b(K);(2)b(\operatorname{Wh}(K))=2b(K); \tag*{(2)}

see [18, Section 4]. Thus, an augmentation rank lower bound that doubles at every step proves the meridional rank conjecture for the iterates, provided the initial augmentation rank equals b(K)b(K).

Main results

Our first main result applies to every nontrivial two-bridge knot.

Theorem 1.1. Let KK be a nontrivial two-bridge knot. For every integer N≥0N \ge0,

ar⁡(Wh⁡N(K))=mr⁡(Wh⁡N(K))=b(Wh⁡N(K))=2N+1.\operatorname{ar}(\operatorname{Wh}^{N}(K))=\operatorname{mr}(\operatorname{Wh}^{N}(K))=b(\operatorname{Wh}^{N}(K))=2^{N+1}.

For N=1N=1, the meridional-rank equality already follows from the lower bound for prime satellite knots in [8, Corollary 1.6]; see also [7, Introduction]. Theorem 1.1 also provides maximal-rank augmentations and propagates both equalities through every finite number of Whitehead doubles.

Given an irreducible microlocal rank one representation ρ\rho on VV, we extend ρ⊕ρ∗\rho\oplus\rho^{*} from the companion group to the satellite group, where (V∗,ρ∗)(V^{*},\rho^{*}) is the dual representation. The two distinguished peripheral eigenlines span a two-dimensional subspace of V⊕V∗V \oplus V^{*}, on which we solve the Whitehead-pattern relations explicitly. The remaining subspace is fixed by the two pattern meridians. Under the nonvanishing conditions of Theorem 2.12, this gives an irreducible 2dim⁡V2\dim V-dimensional representation whose meridian still has microlocal rank 1, doubling the resulting lower bound for meridional rank.

To iterate, we compute the output longitude eigenvalue as well, so that the output of one step can be tested against the hypotheses of the next. This gives the algebraic correspondence of § 2.4. We must then find peripheral values realized by a fixed companion that admits any prescribed finite number of steps.

The difficulty is that the peripheral values realized by a fixed knot might all lie in the exceptional locus. An open dense set in the full peripheral torus does not rule this out. We address this issue using a branch at infinity of the curve. The branch is called Whitehead-iterable branch if the meridian eigenvalue has a pole and the squared longitude eigenvalue either has a pole or has a zero of order greater than that pole order; see Definition 4.1. The key observation is that one branch of the doubling correspondence again satisfies this condition. By Theorem 4.2, for each N≥0N \ge0 there is a nonempty Zariski-open subset of the curve on which NN doubling steps can be performed. For two-bridge knots, the ideal-point argument in § 4.2 supplies a curve with an iterable branch whose peripheral values are realized by rank-two augmentations on a Zariski-dense subset. Combining these facts with (1.1) and (1.2) proves Theorem 1.1.

The same construction provides a comparison with ordinary group rank. We call an irreducible curve in the peripheral torus an actual rank-dd augmentation peripheral curve if a Zariski-dense subset of its points is realized by rank-dd augmentations with ϵ(μ)≠1\epsilon(\mu) \ne1. The coordinates are (μ,λ2)(\mu,\lambda^2), where λ\lambda is the preferred longitude eigenvalue; see Definition 4.4.

Theorem 1.2. For every knot KK and every integer N≥0N \ge0,

gr⁡(Wh⁡N(K))≤gr⁡(K)+N.\operatorname{gr}(\operatorname{Wh}^{N}(K)) \le\operatorname{gr}(K) + N.

If KK has an actual rank-dd augmentation peripheral curve with an iterable branch, then

ar⁡(Wh⁡N(K))≥2Nd\operatorname{ar}(\operatorname{Wh}^{N}(K)) \ge2^{N}d

for every N≥0N \ge0. In particular,

gr⁡(Wh⁡N(K))≤ar⁡(Wh⁡N(K))whenevergr⁡(K)+N≤2Nd.\operatorname{gr}(\operatorname{Wh}^{N}(K)) \le\operatorname{ar}(\operatorname{Wh}^{N}(K)) \quad\text{whenever} \quad\operatorname{gr}(K) + N \le2^{N}d.

For a nontrivial two-bridge knot, this gives

gr⁡(Wh⁡N(K))≤N+2≤ar⁡(Wh⁡N(K))=2N+1.\operatorname{gr}(\operatorname{Wh}^{N}(K)) \le N + 2 \le\operatorname{ar}(\operatorname{Wh}^{N}(K)) = 2^{N+1}.

The proof is given in § 5. A complementary lower bound for ordinary group rank was obtained by Gu–Wang–Zou [21, Proposition 1.3]. For every nontrivial knot KK and N≥1N \ge1, their estimate and Theorem 1.2 give

N+1≤gr⁡(Wh⁡N(K))≤N+gr⁡(K).N + 1 \le\operatorname{gr}(\operatorname{Wh}^{N}(K)) \le N + \operatorname{gr}(K).

Thus ordinary group rank grows linearly, whereas the augmentation and meridional ranks in Theorem 1.1 grow exponentially. The lower bound in [21] uses the hyperbolic pieces of the JSJ decomposition and holds without an augmentation hypothesis. That work also determines ordinary group rank exactly for hyperbolic two-bridge companions; see § 5.

The augmentation lower bound is not always sharp; this limitation is already present in Cornwell’s work [15]. Our last main result exhibits a further obstruction to iteration: even when the augmentation rank agrees with the bridge number for the companion, it may fail to double. The example nevertheless satisfies the meridional rank conjecture.

Theorem 1.3. Let S=T(2,3)#T(2,3)S = T(2,3)\mathbin{\#}T(2,3) be the square knot. Then

ar⁡(S)=mr⁡(S)=b(S)=3,\operatorname{ar}(S) = \operatorname{mr}(S) = b(S) = 3,

whereas

ar⁡(Wh⁡(S))≤5<6=mr⁡(Wh⁡(S))=b(Wh⁡(S)).\operatorname{ar}(\operatorname{Wh}(S)) \le5 < 6 = \operatorname{mr}(\operatorname{Wh}(S)) = b(\operatorname{Wh}(S)).

The augmentation obstruction uses the two trefoil factors of the companion group and their preferred longitudes. The meridional lower bound uses a different representation: an explicit epimorphism πWh⁡(S)↠S14\pi_{\operatorname{Wh}(S)} \twoheadrightarrow S_{14} and its permutation module F214\mathbb{F}_{2}^{14}. The first cohomology of πWh(S)\pi_{\mathrm{Wh}}(S) with coefficients in this module has dimension 14, while the meridian defect has rank 3. A cocycle estimate then rules out generation by five meridians; see § 6.2 for the representation and exact linear calculation. Here the finite quotient supplies the coefficient module for the cohomology of the knot group. This gives a strict gap between augmentation rank and meridional rank. The exact augmentation rank of Wh(S)\mathrm{Wh}(S) and the ranks of further Whitehead doubles of SS remain undetermined in this paper.

Conventions and organization

All knots are oriented. A meridian is positively oriented, and elements represented by meridians are understood up to conjugacy. We use a preferred zero-linking longitude. Sheaves and augmentations may be defined over any field k\mathbb{k}, but in the algebraic-curve applications, we take k=C\mathbb{k} = \mathbb{C}. The finite-dimensional permutation module in § 6.2 is explicitly over F2\mathbb{F}_2.

We use πK:=π1(S3∖K)\pi_K := \pi_1(S^3 \setminus K) for the knot group throughout. The commutator convention is [a,b]=aba−1b−1[a,b] = aba^{-1}b^{-1}. Products of permutations are composed from right to left, and permutation matrices act on column vectors by gei=eg(i)ge_i = e_{g(i)}. The embedding ϕK\phi_K of the Whitehead solid torus into a tubular neighborhood of KK respects the preferred zero framing. A “two-bridge knot” in the statements below is nontrivial.

§ 2 recalls the representations and augmentations and constructs the doubling correspondence. § 2.5 gives the meridional upper bounds. § 3 recalls the algebraic curves, valuations, and character varieties used to find peripheral data to which the iteration applies. § 4.1 proves the iterable-branch criterion, and § 4.2 applies slope detection to prove Theorem 1.1. § 5 compares ordinary group rank with augmentation rank. Finally, § 6 proves Theorem 1.3 for the first Whitehead double of the square knot.

Microlocal rank-1 representations and Whitehead doubling

The meridian and the preferred longitude of a knot KK are denoted by mm and ℓ\ell, respectively. We use the specialization of knot contact homology in which the extra variable UU is set equal to one. For an embedded submanifold A⊂S3A \subset S^3, we denote by Op⁡(A)\operatorname{Op}(A) a fixed sufficiently small open tubular neighborhood of AA.

Microlocal rank-1 representations and augmentations

We first introduce the notation for the microlocal rank of a representation of the knot group.

Definition 2.1. For a representation ρ:πK→GL⁡(V)\rho: \pi_K \to\operatorname{GL}(V), define the microlocal rank to be

μrk⁡K(ρ):=rank⁡(ρ(m)−I),\mu\operatorname{rk}_K(\rho) := \operatorname{rank}(\rho(m)-I),

where II is the identity matrix of size dim⁡V\dim V. This number is independent of the chosen meridian, since all meridians are conjugate.

We call ρ\rho a μrk⁡\mu\operatorname{rk}-one representation if μrk⁡K(ρ)=1\mu\operatorname{rk}_K(\rho)=1.

A μrk⁡\mu\operatorname{rk}-one representation is diagonalizable if there is a basis in which

ρ(m)∼diag⁡(μ,1,…,1),μ≠1.\rho(m) \sim\operatorname{diag}(\mu,1,\ldots,1), \qquad\mu\ne1.

A μrk⁡\mu\operatorname{rk}-one representation is unipotent if there is a basis in which

ρ(m)∼(1101)⊕Idim⁡V−2.\rho(m) \sim\begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix} \oplus I_{\dim V-2}.

We recall the sheaf-theoretic reason for this terminology [20, Section 3.2]. Let j:S3∖K↪S3j:S^3\setminus K\hookrightarrow S^3 be the open embedding. A representation ρ:πK→GL⁡(V)\rho:\pi_K\to\operatorname{GL}(V) determines a local system Eρ\mathcal{E}_{\rho} on the knot complement. For the underived direct image, the stalk of j∗Eρj_*\mathcal{E}_{\rho} on KK is the meridian-invariant subspace Vρ(m)V^{\rho(m)}. The local gluing map is the inclusion

T:Vρ(m)↪V.T:V^{\rho(m)}\hookrightarrow V.

Its mapping cone is the microlocal stalk, up to a degree shift, in the sense of Kashiwara–Schapira, along the knot conormal torus ΛK\Lambda_K inside the contact cosphere bundle, i.e. ΛK⊂ST∗S3\Lambda_K\subset ST^*S^3 is a Legendrian surface. Consequently,

dim⁡H∙(Cone⁡(T))=dim⁡(V/Vρ(m))=rank⁡(ρ(m)−I).\dim H^\bullet(\operatorname{Cone}(T))=\dim(V/V^{\rho(m)})=\operatorname{rank}(\rho(m)-I).

Thus a μ\mu rk-one representation induces a microlocal rank one sheaf microsupported along the conormal. Conversely, up to the equivalence modulo local systems on S3S^3 in [20, Definition 3.8 and Proposition 3.11], a microlocal rank one abelian sheaf comes from one of the three forms: (1) an irreducible diagonalizable μ\mu rk-1 representation via the above construction; (2) an irreducible unipotent μ\mu rk-1 representation via the above construction; and (3) a local system of rank 1 that is supported on KK [20]. The first type is generic and the main focus of this paper.

A diagonalizable μ\mu rk-one representation is also called a KCH representation by Cornwell [15], where KCH refers to knot contact homology. Next, we explain its background related to knot contact homology and augmentations.

The (framed) knot contact homology is a knot invariant defined via counting holomorphic disks for ΛK\Lambda_K, whose formula was first discovered by Ng [29]. We use the specialization U=1U=1 of the Legendrian dg algebra A(ΛK)\mathcal{A}(\Lambda_K) associated to ΛK\Lambda_K. The dg algebra A(ΛK)\mathcal{A}(\Lambda_K) is semi-freely generated by Reeb chords over Z[H1(ΛK)]≅Z[μ±1,λ±1]\mathbb{Z}[H_1(\Lambda_K)]\cong\mathbb{Z}[\mu^{\pm1},\lambda^{\pm1}]. An augmentation of the dg algebra is a morphism ϵ:A(ΛK)→k\epsilon:\mathcal{A}(\Lambda_K)\to\mathbb{k} which preserves grading, differential, and the unit. The case of the knot conormal has the special property that its generators are concentrated in non-negative degrees. Hence an augmentation descends to the degree 0 homology ϵ:H0(A(ΛK))→k\epsilon:H_0(\mathcal{A}(\Lambda_K))\to\mathbb{k}.

The algebra H0(A(ΛK))H_0(\mathcal{A}(\Lambda_K)) admits a combinatorial model by the framed cord algebra.

References

  1. [28]T. W. Mattman, G. Maybrun, and K. Robinson. 2-bridge knot boundary slopes: diameter and genus. Osaka J. Math. 45 (2008), no. 2, 471–489.DOI
  2. [29]L. Ng. Framed knot contact homology. Duke Math. J. 141 (2008), no. 2, 365–406.arxiv.org/abs/math/0407071
  3. [30]T. Ohtsuki. Ideal points and incompressible surfaces in two-bridge knot complements. J. Math. Soc. Japan 46 (1994), no. 1, 51–87.
  4. [31]M. Rost and H. Zieschang. Meridional generators and plat presentations of torus links. J. London Math. Soc. (2) 35 (1987), no. 3, 551–562.DOI
  5. [32]J. Schultens. Additivity of bridge numbers of knots. Math. Proc. Cambridge Philos. Soc. 135 (2003), no. 3, 539–544.DOI
  6. [33]I. R. Shafarevich. Basic Algebraic Geometry 1: Varieties in Projective Space. 3rd ed. Translated from the Russian by Miles Reid. Berlin, Heidelberg: Springer-Verlag, 2013.
  7. [35]A. S. Sikora. Generating sets for coordinate rings of character varieties. J. Pure Appl. Algebra 217 (2013), no. 11, 2076–2087.arxiv.org/abs/1106.4837
  8. [36]C. Tuffley. Generalised knot groups distinguish the square and granny knots. J. Knot Theory Ramifications 18 (2009), no. 8, 1129–1157. With an appendix by David Savitt.DOI

Paper details

vlnc:W_wzvvfrgz
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

OpenAI’s ChatGPT

View statement
Looking for related papers…
Contents