2000 Mathematics Subject Classification. Primary 57M27; Secondary 11S05, 37B40. The author was supported in part by JSPS fellowship for young scientists.
<ph f="ptmb7t">Zeros of the Alexander polynomial of knot</ph>

### Akio Noguchi

Department of Mathematics, Tokyo Institute of Technology, Oh-okayama, Meguro-ku, Tokyo 152-8551, Japan E-mail address: akio@math.titech.ac.jp
• Abstract. The leading coefficient of the Alexander polynomial of a knot is the most informative element in this invariant, and the growth of orders of the first homology of cyclic branched covering spaces is also a familiar subject.
Accordingly, there are a lot of investigations into each subject. However, there is no study which deal with the both subjects in a same context. In this paper, we show that the two subjects are closely related in $p$  -adic number theory and dynamical systems.

1 Introduction

The leading coefficient of the Alexander polynomial ${\Delta }_{K}\left(t\right)$  of a knot $K$  is a well-known invariant for detecting fibered knots. The Alexander polynomial of a fibered knot is always monic [21. The converse is not always true, but it holds for many knots, for example, alternating knots [20. Moreover the monic condition characterizes fibered knots in a sense of realization [3, 22.
The leading coefficient of the Alexander polynomial of a knot is also related to the commutator subgroup ${G}_{K}^{\prime }$  of the knot group ${G}_{K}={\pi }_{1}\left({S}^{3}\K\right)$  . The abelianaization of ${G}_{K}^{\prime }$  is finitely generated if and only if the leading coefficient is $±1$  [5, 24.
The $r$  -fold cyclic covering branched over a knot $K$  , which is denoted by ${X}_{r}$  , is a fundamental object in the knot theory, since topological invariants of it are also invariants of the knot. In [9, Gordon studied the growth of the order of ${H}_{1}\left({X}_{r};\mathbb{Z}\right)$  with respect to $r$  and asked whether the growth is exponential in case same zero of ${\Delta }_{K}\left(t\right)$  are not a root of unity. More than 15 years later, this question was affirmatively answered by Riley [25and González-Acun͂a and Short [8independently.
But it still remains improvable. We completely express the growth by the zeros of the Alexander polynomial.
The entropy is an invariant of a self-map for measuring complexity of the map.
However, we study the entropy of the meridian action on the Alexander module (Theorem  1 ) and regard it as an invariant to measure complexity of the Alexander module, for the meridian action is canonical for every Alexander module. Precisely speaking, it should be called the entropy of the dual action of the dual group of the Alexander module. But we confuse them by their duality (see Section 2.1).
With interpretation of the entropy, we obtain the following results. For more precise statements, see Section 4.
Results. Let ${\Delta }_{K}\left(t\right)={\sum }_{i=0}^{n}{a}_{i}{t}^{i}\left({a}_{0}{a}_{n}\ne 0\right)$  be the Alexander polynomial of a knot $K$  and ${\alpha }_{i}$  the zeros (counted with multiplicity) of ${\Delta }_{K}\left(t\right)$  . Then,
• (1) the leading coefficient of ${\Delta }_{K}\left(t\right)$  is
 $\begin{array}{cc}log|{a}_{n}|={\sum }_{p<\infty }{\sum }_{|{\alpha }_{i}{|}_{p}>1}log|{\alpha }_{i}{|}_{p}& \end{array}$
 $\begin{array}{}\end{array}$
(Corollary  4 ), and
• (2) the growth of order of the first homology of the $r$  -fold cyclic covering branched over $K$  is
 $\begin{array}{cc}{lim}_{\stackrel{r\to \infty }{|{H}_{1}\left(\cdot \right)|\ne 0}}\frac{log|{H}_{1}\left({X}_{r};\mathbb{Z}\right)|}{r}={\sum }_{p\le \infty }{\sum }_{|{\alpha }_{i}{|}_{p}>1}log|{\alpha }_{i}{|}_{p}& \end{array}$
 $\begin{array}{}\end{array}$
(Corollary  1 ).
Here, $|\cdot {|}_{p}$  are the $p$  -adic norms and $|\cdot {|}_{\infty }$  is the standard norm. (we assume that the embeddings $\overline{\mathbb{Q}}\to \overline{{\mathbb{Q}}_{p}}$  are fixed.)
In our study, we establish the followings.
• $\bullet$  The leading coefficient of the Alexander polynomial can be recovered from its zeros. Furthermore, the distribution of the zeros measures a certain distance of the Alexander module from being finitely generated as $\mathbb{Z}$  -module and a new interpretation of the leading coefficient is given. (Section 4.2)
• $\bullet$  Since the primary interests of Gordon [9was to investigate the periodicity of ${H}_{1}\left({X}_{r};\mathbb{Z}\right)$  , he studied the growth of orders to determine the non-periodic case. However, the growth also measures complexity of the Alexander module. (Section 4.1)
Here, we have a few comments on this study. It might make our study a little more attractive.
The Alexander polynomial of a knot is defined as a greatest common divisor of the initial Fitting ideal (elementary ideal) of the Alexander module ${H}_{1}\left({X}_{\infty };\mathbb{Z}\right)$  as $\mathbb{Z}\left[{t}^{±}\right]$  -module. Here, the indeterminate $t$  is identified with the meridian action on ${H}_{1}\left({X}_{\infty };\mathbb{Z}\right)$  and ${X}_{\infty }$  is the infinite cyclic cover of $X={S}^{3}\K$  . Then, by tensoring with the rational numbers $\mathbb{Q}$  , the Alexander polynomial is also generator of the Fitting ideal of the module ${H}_{1}\left({X}_{\infty };\mathbb{Q}\right)$  as $\mathbb{Q}\left[{t}^{±}\right]$  -module, and hence it is the characteristic polynomial of the meridian action on ${H}_{1}\left({X}_{\infty };\mathbb{Q}\right)$  , up to units (see Theorem 6.17 in [15).
Although the rational homology ${H}_{1}\left({X}_{\infty };\mathbb{Q}\right)$  gives a nice explanation of the Alexander polynomial, the leading coefficient ${a}_{n}$  is lost in ${H}_{1}\left({X}_{\infty };\mathbb{Q}\right)$  because ${a}_{n}$  is a unit in $\mathbb{Q}\left[{t}^{±}\right]$  . On the other hand, the entropy has an advantage over the Fitting ideal because we can replace ${H}_{1}\left({X}_{\infty };\mathbb{Z}\right)$  with ${H}_{1}\left({X}_{\infty };\mathbb{Q}\right)$  with preserving the entropy (cf. Step 1 in the proof of Proposition  9 ). This is why the zeros of the Alexander polynomial keep information about ${H}_{1}\left({X}_{\infty };\mathbb{Z}\right)$  .
As Gordon mentioned, the difficulty in computing the growth of orders arises from the case in which all zeros belong to the unit circle but some are not a root of unity (e.g. ${\Delta }_{{5}_{2}}\left(t\right)=2{t}^{2}-3t+2$  ). In this case, the standard norm is useless. Reliy [25managed the difficulty by the $p$  -adic analysis. On the other hand, González-Acun͂a and Short [8managed it by showing that the growth is equal to the Mahler measure of the Alexander polynomial.
Our feature is the interpretation of the growth as the entropy of the meridian action on the Alexander module, which can be obtained by combining the result of González-Acun͂a and Short [8and Einseidler and Ward [6. As a result, the growth turn out to be an invariant to measure the complexity of the Alexander module.
Because the Alexander module is not always finite generated, toral automorphisms are not enough to investigate it. But solenoidal automorphisms are enough for the Alexander module (Lemma  1 ). So, we can apply the works of Lind and Ward [17:
the Haar measure on solenoid can be lifted to the Haar measure on adele rings with preserving the entropy and the entropy is the sum of the entropies for the $p$  -adic directions. That is to say, they established a kind of Hasse principle for dynamical systems. Finally, the growth of orders is expressed by the $p$  -adic norms of the zeros of the Alexander polynomial, for expansions (or entropy) in adele rings can be computed by the $p$  -adic norms (see Example  1 , 2 ). Although our approach is different from Riley, the $p$  -adic method is useful again.
In [25, Riley also proved the other results, which are on the $p$  -part of $|{H}_{1}\left({X}_{r};\mathbb{Z}\right)|$  .
He obtained the upper bounds for the $p$  -parts: $|{H}_{1}\left({X}_{r};\mathbb{Z}\right){|}^{\left(p\right)}  , where $A,H,E,n$  are constants depending on a knot (Theorem 2 in [25). This result implies that the $p$  -parts have trivial growth with respect to $r$  for any primes (as against ${e}^{r}$  ). Silver and Williams [29re-proved this trivial growth and generalized it under mild hypotheses, and their argument is helpful for us. As Silver and Williams mentioned, this fact implies that if $|{H}_{1}\left({X}_{r};\mathbb{Z}\right)|$  dose not have the trivial growth then the sequence of them displays infinitely many prime numbers in the factorization of its terms. Furthermore, Riley also confirmed that almost all $p$  -parts actually increase and his upper bounds are best possible (except for the constant multiplier) with respect to $r$  and also $p$  (Theorem 3 in [25).
Acknowledgment. I would like to thank Dr. Kazuo Masuda for his helpful advice and valuable discussion, and also thank Dr. Sadayoshi Kojima, Dr. Hitoshi Murakami, Dr. Masanori Morishita and Dr. Gregor Masbaum for their kind reading my rough draft and giving helpful comments. And, I am grateful to Dr. Kunio Murasugi for his encouragement in this study and kind hospitality during my visiting University of Toronto, and also Dr. Miho Aoki for her helpful conversation on number theory. I also appreciate the seminor with Dr. Shoichi Nakajima, Dr. Shin Nakano and Dr. Mikami Hirasawa, which was helpful to meke this paper readable.

2 Preliminaries

2.1 Fourier analysis on number fields

The classical Fourier analysis is based on the Pontryagin duality between the integers $\mathbb{Z}$  and the torus $\mathbb{T}=\mathbb{R}/\mathbb{Z}$  . The modern Fourier analysis is extended over more general Pontryagin dualities. It is called harmonic analysis sometimes. In this section, we review the Pontryagin duality for the rational number field $\mathbb{Q}$  and related topics.

2.1.1 Fourier analysis on LCA groups

Let $G$  be a locally compact group. A collection $\stackrel{^}{G}$  of all continuous homomorphisms $\chi :G\to \mathbb{C}$  is called a dual group or character group if $|\chi \left(g\right)|=1$  for all $g\in G$  with pointwise multiplication and compact-open topology. The dual group $\stackrel{^}{G}$  has the following properties. For more details, Rudin's book [27is a standard exposition.
Proposition 1. A locally compact abelian group $G$  and its dual group $\stackrel{^}{G}$  have the following properties.
• (1) $\stackrel{^}{G}$  is also a locally compact abelian group.
• (2) $G$  is compact if and only if $\stackrel{^}{G}$  is discrete.
• (3) $\stackrel{^}{\stackrel{^}{G}}$  is naturally isomorphic to $G$  , which is called the Pontryagin duality.
• (4) Let $〈g,\chi 〉=\chi \left(g\right)$  for every $g\in G,\chi \in \stackrel{^}{G}$  . Then, for any continuous homomorphism $\phi :G\to H$  , there exists a continuous homomorphism $\stackrel{^}{\phi }:\stackrel{^}{H}\to \stackrel{^}{G}$  such that $〈\phi \left(g\right),{\chi }_{H}〉=〈g,\stackrel{^}{\phi }\left({\chi }_{H}\right)〉$  and $\stackrel{^}{\stackrel{^}{\phi }}=\phi$  .
• (5) If $G$  and $H$  are both either compact or discrete, $\phi$  is surjective if and only if $\stackrel{^}{\phi }$  is injective.
By the above proposition, an automorphism of a discrete abelian group can be transformed to an automorphism of a compact abelian group, and this transformation is reversible.
The duality translates topological structures to algebraic structures.
Proposition 2. Let $G$  be a compact abelian group.
• (1) $G$  is connected if and only if $\stackrel{^}{G}$  is torsion free.
• (2) $G$  is finite dimension $d$  if and only if $\stackrel{^}{G}$  is finite rank $d$  .
• (3) $G$  is metrizable if and only if $\stackrel{^}{G}$  is countable.

2.1.2 Fourier analysis on the rational number field

We now deal with the Pontryagin duality for the rational number field $\mathbb{Q}$  with the discrete topology. To do that, we need $p$  -adic number fields and the adele ring of $\mathbb{Q}$  . See [23, 32for more details.
Every non-zero rational number $a\in \mathbb{Q}$  can be presented as $a={p}^{m}\frac{u}{v}$  for each rational prime number $p$  . Then, the $p$  -adic norm of $a$  is defined by $|a{|}_{p}=\left\{\begin{array}{cc}{p}^{-m}& a\ne 0\\ 0& a=0,\end{array}$  and defines a metric on $\mathbb{Q}$  by ${d}_{p}\left(a,b\right)=|a-b{|}_{p}$  , which is called the $p$  -adic metric.
The $p$  -adic number field ${\mathbb{Q}}_{p}$  is defined as the completion of $\mathbb{Q}$  with respect to the $p$  -adic metric ${d}_{p}$  . An algebraic closure $\overline{{\mathbb{Q}}_{p}}$  has a unique norm which extends from the $p$  -adic norm on ${\mathbb{Q}}_{p}$  .
These spaces do not seem natural. However, they are natural under the following concept.
Definition 1. Let $F$  be a locally compact field and $\mu$  an additive Haar measure on $F$  . Then for every $a\in F$  , $\mu \left(a\cdot \right)$  is another Haar measure on $F$  . Therefore, for every $a\in F$  , there exists the module $|a{|}_{F}$  such that $\mu \left(a\cdot \right)=|a{|}_{F}\mu \left(\cdot \right)$  .
Example 1. When $F$  is the real number field $\mathbb{R}$  , the module $|a{|}_{\mathbb{R}}$  is the ordinary norm $|a|$  . When $F$  is the $p$  -adic number field ${\mathbb{Q}}_{p}$  , the module $|a{|}_{{\mathbb{Q}}_{p}}$  is the $p$  -adic norm $|a{|}_{p}$  .
The $p$  -adic norms are natural concept with respect to an additive Haar measure on ${\mathbb{Q}}_{p}$  , and consequently $p$  -adic norms are useful for computing an entropy (see Example  2 ).
Definition 2. The adele ring of the rational numbers is defined as a restricted direct product space of ${\mathbb{Q}}_{p}$  , that is ${\mathbb{A}}_{\mathbb{Q}}=\left\{x=\left({x}_{p}\right)\in {\prod }_{p\le \infty }{\mathbb{Q}}_{p}||{x}_{p}{|}_{p}\le 1\text{for almost all}p\right\}.$
The adele group (with addition) is a locally compact abelian group, and hence has a Haar measure, which is a kind of product measure.
The following proposition means that the adele ring of the rational numbers is parallels to the real numbers $\mathbb{R}$  in the duality $\stackrel{^}{\mathbb{Z}}\sim =\mathbb{R}/\mathbb{Z}$  .
Proposition 3. $\mathbb{Q}$  is uniform lattice in ${\mathbb{A}}_{\mathbb{Q}}$  , i.e. discrete co-compact subgroup of ${\mathbb{A}}_{\mathbb{Q}}$  . Moreover, ${\mathbb{A}}_{\mathbb{Q}}/\mathbb{Q}\sim =\stackrel{^}{\mathbb{Q}}.$

2.2 Entropy

An entropy is a quantity to measure a complexity of a self-map. There are several definitions of entropies from various aspect. Kolmogorov [13and Sinai [30introduced the measure theoretic entropy ${h}_{\mu }\left(T\right)$  for a measure preserving map $T$  on a probability space $\left(X,\mathfrak{B},\mu \right)$  . Alder, Konheim and McAndrew [1introduced the topological entropy $h\left(T\right)$  for a continuous map $T$  on a compact space. Bowen [2defined a variant of topological entropy ${h}_{d}\left(T\right)$  , which is called Bowen's topological entropy, for a uniformally continuous map $T$  on a metric space $\left(X,d\right)$  . Each entropy has its own property and is related to the others. For more details, see Walters [31.
Bowen's topological entropy is equal to the original topological entropy for compact metrizable space.
Proposition 4. When $X$  is compact, ${h}_{d}\left(T\right)$  is not depend on a metric $d$  and
 $\begin{array}{cc}{h}_{d}\left(T\right)=h\left(T\right).& \end{array}$
 $\begin{array}{}\end{array}$
The following proposition shows that the topological entropy is a supremum of measure theoretic entropy, and we usually call it the variational principle.
Proposition 5. Let $T$  be a continuous map of a compact metric space $X$  and $M\left(X,T\right)$  the all probability measures which are preserved by $T$  . Then
 $\begin{array}{cc}h\left(T\right)=sup\left\{{h}_{\mu }\left(T\right)|\mu \in M\left(X,T\right)\right\}& \end{array}$
 $\begin{array}{}\end{array}$
Our main interest is an automorphism of a locally compact abelian group and especially a compact abelian group. Since any surjective endomorphism of a compact group preserves the Haar measure, an automorphism of a compact group is a simple example for which both topological entropy and measure theoretic entropy can be defined. Moreover the both entropies are identical in this case.
Proposition 6. Suppose $G$  is a compact metrizable group, $T$  a surjective endomorphism of $G$  , and $\mu$  a normalized Haar measure on $G$  . Then
 $\begin{array}{cc}{h}_{\mu }\left(T\right)={h}_{d}\left(T\right)=h\left(T\right).& \end{array}$
 $\begin{array}{}\end{array}$
From Proposition  6 , we can choose any definition for studying the entropy in this condition. Bowen [2also introduce yet another topological entropy, which is computable and inherits its habit from ${h}_{d}\left(T\right)$  and ${h}_{\mu }\left(T\right)$  . In our condition, this entropy can be identified with ${h}_{d}\left(T\right)$  and computed by the following formula.
Proposition 7. Let $G$  be a locally compact metrizable abelian group with a Haar measure $\mu$  and $T$  a surjective endomorphism, then the entropy can be computed by the formula
 $\begin{array}{cc}{h}_{d}\left(T\right)={lim}_{\varepsilon \to 0}{limsup}_{n\to \infty }\left[-\frac{1}{n}log\mu \left({\cap }_{k=0}^{n-1}{T}^{-k}B\left(e,\varepsilon \right)\right)\right],& \end{array}$
 $\begin{array}{}\end{array}$
where $B\left(e,\varepsilon \right)$  is a open $\varepsilon$  -ball of the identity element with respect to an invariant metric $d$  .
Example 2. Let ${T}_{a}$  be an automorphism of ${\mathbb{Q}}_{p}$  defined by a multiplication by $a\in {\mathbb{Q}}_{p}^{*}$  . Then it follows from Example  1 that
 $\begin{array}{cc}{h}_{{d}_{p}}\left({T}_{a}\right)=\left\{\begin{array}{cc}log|a{|}_{p}& |a{|}_{p}>1,\\ 0& |a{|}_{p}\le 1.\end{array}& \end{array}$
 $\begin{array}{}\end{array}$
Bowen's topological entropy is compatible with coverings.
Proposition 8 (Bowen [2). Let $G$  be a locally compact metrizable abelian group with an invariant metric $d$  . Suppose $\Gamma$  is a uniform lattice of $G$  , that is, a discrete and co-compact subgroup of $G$  . Let $\stackrel{~}{T}$  and $T$  be endomorphisms of $G$  and $G/\Gamma$  such that $\pi \circ \stackrel{~}{T}=T\circ \pi$  , where $\pi :G\to G/\Gamma$  is the projection. Then
 $\begin{array}{cc}h\left(T\right)={h}_{d}\left(\stackrel{~}{T}\right).& \end{array}$
 $\begin{array}{}\end{array}$

3 Solenoidal entropy and Alexander polynomial

A solenoid ${\Sigma }^{d}$  is, by definition, a compact connected finite-dimensional abelian group and arose from a generalization of the torus ${\mathbb{T}}^{d}$  . The following theorem, which was given by Lind and Ward [17, plays a key role for our results.
Proposition 9 (Lind and Ward [17). Let $A$  be an automorphism of $d$  -dimensional solenoid ${\Sigma }^{d}$  . Then,
• (1) the entropy of $A$  is the sum of the entropies of the automorphisms of ${\mathbb{Q}}_{p}^{d}$  induced by $A$
 $\begin{array}{cc}h\left(A;{\Sigma }^{d}\right)={\sum }_{p\le \infty }h\left(A;{\mathbb{Q}}_{p}^{d}\right),& \end{array}$
 $\begin{array}{}\end{array}$
and
• (2) the $p$  -adic entropy is computed by the eigenvalues ${\lambda }_{1},...,{\lambda }_{d}$  of the induced automorphism in $GL\left(d,{\mathbb{Q}}_{p}\right)$  as follows:
 $\begin{array}{cc}h\left(A;{\mathbb{Q}}_{p}^{d}\right)={\sum }_{|{\lambda }_{i}{|}_{p}>1}log|{\lambda }_{i}{|}_{p}.& \end{array}$
 $\begin{array}{}\end{array}$
In [17, Lind and Ward computed the entropy of solenoidal automorphisms with intrinsic arguments. The proof is really helpful for our applications later. Although we review an outline of the proof, referring to the original paper is strongly recommend.
• Outline of the proof.
• Step 1: Because the dual group of ${\Sigma }^{d}$  can be embed into ${\mathbb{Q}}^{d}$  , $lim-\to {\Gamma }_{n}\sim ={\mathbb{Q}}^{d}$  , where ${\Gamma }_{n}=\frac{1}{n!}{\stackrel{^}{\Sigma }}^{d}$  . Hence ${\stackrel{^}{\mathbb{Q}}}^{d}\sim =lim←-{\stackrel{^}{\Gamma }}_{n}$  and ${\stackrel{^}{\Gamma }}_{n}\sim ={\stackrel{^}{\mathbb{Q}}}^{d}/{K}_{n}$  . By addition formula [10, $h\left(A;{\stackrel{^}{\mathbb{Q}}}^{d}\right)=h\left(A;{\stackrel{^}{\Gamma }}_{n}\right)+h\left(A;{K}_{n}\right)$  . Because $h\left(A;{\stackrel{^}{\Gamma }}_{n}\right)=h\left(A;{\Sigma }^{d}\right)$  for any $n$  and $h\left(A;{K}_{n}\right)\to 0$  as $n\to \infty$  ,
 $\begin{array}{cc}h\left(A;{\Sigma }^{d}\right)=h\left(A;{\stackrel{^}{\mathbb{Q}}}^{d}\right).& \end{array}$
 $\begin{array}{}\end{array}$
• Step 2: By Proposition  3 and Proposition  8 , the entropy on the full solenoid ${\stackrel{^}{\mathbb{Q}}}^{d}$  can be lifted to the entropy on the adele ring ${\mathbb{A}}_{\mathbb{Q}}$  ,
 $\begin{array}{cc}h\left(A;{\stackrel{^}{\mathbb{Q}}}^{d}\right)=h\left(A;{\mathbb{A}}_{\mathbb{Q}}^{d}\right).& \end{array}$
 $\begin{array}{}\end{array}$
• Step 3: Since the adele ring is a restricted direct product space, the entropy on it can be decomposed into the entropies of each direction
 $\begin{array}{cc}h\left(A;{\mathbb{A}}_{\mathbb{Q}}^{d}\right)={\sum }_{p\le \infty }h\left(A;{\mathbb{Q}}_{p}^{d}\right).& \end{array}$
 $\begin{array}{}\end{array}$
This additivity follows from that ${\mathbb{A}}_{\mathbb{Q}}^{d}$  is almost the direct product space.
But it needs delicate arguments.
• Step 4: Finally, using the formula in Proposition  7 , we have
 $\begin{array}{cc}h\left(A;{\mathbb{Q}}_{p}^{d}\right)={\sum }_{|{\lambda }_{i}{|}_{p}>1}log|{\lambda }_{i}{|}_{p}.& \end{array}$
 $\begin{array}{}\end{array}$
The computation is essentially similar to Example  2 .
To connect the Alexander polynomial with a solenoidal automorphism, we need the following lemma.
Lemma 1. For any knot, the dual group of the first homology group of an infinite cyclic cover ${H}_{1}\left({X}_{\infty };\mathbb{Z}\right)$  is a $n$  -dimensional solenoid. Here, $n$  is the degree of the Alexander polynomial of the knot.
• Proof. From Proposition  1 and  2 , it is sufficient to prove that ${H}_{1}\left({X}_{\infty };\mathbb{Z}\right)$  is a discrete torsion-free abelian group which has finite-rank $n$  . Rapaport [24and Crowell [5proved that ${H}_{1}\left({X}_{\infty };\mathbb{Z}\right)$  is torsion-free and has finite rank $n$  . (Here, the rank of $A$  means the cardinality of any maximal set of $\mathbb{Z}$  -linearly independent elements of $A$  .)
Because the Alexander polynomial is equal to the characteristic polynomial of the meridian action on ${H}_{1}\left({X}_{\infty };\mathbb{Q}\right)$  , up to multiplication by a unit, the following theorem follows from Proposition  9 .
Theorem 1. Let ${\alpha }_{i}$  be the zeros (counted with multiplicity) of the Alexander polynomial of a knot. Then,
• (1) the entropy of the meridian action on the $p$  -adic Alexander module; ${t}_{p}:{H}_{1}\left({X}_{\infty };{\mathbb{Q}}_{p}\right)\to {H}_{1}\left({X}_{\infty };{\mathbb{Q}}_{p}\right)$  is
 $\begin{array}{cc}h\left({t}_{p}\right)={\sum }_{|{\alpha }_{i}{|}_{p}>1}log|{\alpha }_{i}{|}_{p},& \end{array}$
 $\begin{array}{}\end{array}$
where $|\cdot {|}_{p}$  is the $p$  -adic norms, and
• (2) the entropy of dual action of meridian $\stackrel{^}{t}:\stackrel{^}{{H}_{1}\left({X}_{\infty };\mathbb{Z}\right)}\to \stackrel{^}{{H}_{1}\left({X}_{\infty };\mathbb{Z}\right)}$  is $h\left(\stackrel{^}{t}\right)={\sum }_{p\le \infty }h\left({t}_{p}\right)$  , that is
 $\begin{array}{cc}h\left(\stackrel{^}{t}\right)={\sum }_{p\le \infty }{\sum }_{|{\alpha }_{i}{|}_{p}>1}log|{\alpha }_{i}{|}_{p}.& \end{array}$
 $\begin{array}{}\end{array}$
Here ${\mathbb{Q}}_{\infty }=\mathbb{R}$  by convention.

4 Applications

4.1 Growth of order of homology of branched cyclic covering space

In this section, we study the relation between the $p$  -adic zeros of ${\Delta }_{K}$  and the growth of orders of the first homology groups of the $r$  -fold cyclic covering of ${S}^{3}$  branched over $K$  . Roughly speaking, the $r$  -fold cyclic covering branched over $K$  is a compact space which associated with a homomorphism
 $\begin{array}{cc}G={\pi }_{1}\left({S}^{3}\K\right)\stackrel{}{⟶}\mathbb{Z}/r\mathbb{Z},& \end{array}$
 $\begin{array}{}\end{array}$
(for precise definition, see [4, 15, 26).
The order of the first homology group of this space can be computed by the following formula.
Proposition 10 (Fox [7). Let ${X}_{r}$  be the $r$  -fold cyclic covering of ${S}^{3}$  branched over $K$  . Then, the order of the first homology group of ${X}_{r}$  is given by $|{H}_{1}\left({X}_{r};\mathbb{Z}\right)|=|{}^{r-1}{\prod }_{d=1}{\Delta }_{K}\left(exp\left(2d\pi \sqrt{-1}/r\right)\right)|.$  By convention, $|{H}_{1}\left({X}_{r};\mathbb{Z}\right)|=0$  means that ${H}_{1}\left({X}_{r};\mathbb{Z}\right)$  is an infinite group.
Definition 3 (logarithmic Mahler measure [18). For non-zero Laurent polynomial $f\left(x\right)$  with integral coefficients, the logarithmic Mahler measure of $f$  is defined by
 $\begin{array}{cc}m\left(f\right)={\int }_{0}^{1}log|f\left(exp\left(2\pi t\sqrt{-1}\right)\right)|dt.& \end{array}$
 $\begin{array}{}\end{array}$
The growth of orders $|{H}_{1}\left({X}_{r};\mathbb{Z}\right)|$  is expressed by the logarithmic Mahler measure of the Alexander polynomial. It has been already proved by González-Acun͂a and Short [8, but we show a proof because we base on the other definition from [8.
 $\begin{array}{cc}{lim}_{\stackrel{r\to \infty }{|{H}_{1}\left(\cdot \right)|\ne 0}}\frac{log|{H}_{1}\left({X}_{r};\mathbb{Z}\right)|}{r}& ={lim}_{\stackrel{r\to \infty }{{\Delta }_{K}\left(\cdot \right)\ne 0}}\frac{1}{r}log|{}^{r-1}{\prod }_{d=0}{\Delta }_{K}\left(exp\left(2d\pi \sqrt{-1}/r\right)\right)|\end{array}$
 $\begin{array}{cc}& ={lim}_{\stackrel{r\to \infty }{{\Delta }_{K}\left(\cdot \right)\ne 0}}{\sum }_{d=0}^{r-1}\frac{1}{r}log|{\Delta }_{K}\left(exp\left(2\pi \sqrt{-1}\frac{d}{r}\right)\right)|\end{array}$
 $\begin{array}{cc}& =m\left({\Delta }_{K}\right).\end{array}$
 $\begin{array}{}\end{array}$
Mahler measure is deeply related to the entropy of an algebraic dynamical system, which is found in [16for example. In this paper, we use a more suitable result which was proved by Einseidler and Ward [6.
Proposition 11 (Einseidler and Ward [6). Let $0\stackrel{}{⟶}{F}_{n}\stackrel{}{⟶}...\stackrel{}{⟶}{F}_{1}\stackrel{{\phi }_{1}}{⟶}{F}_{0}\stackrel{}{⟶}M\stackrel{}{⟶}0$  be a finite free resolution of the $\mathbb{Z}\left[{t}^{±}\right]$  -module $M$  and $J\left({\phi }_{1}\right)$  the initial Fitting ideal.
Let ${\alpha }_{t}$  a natural automorphism which is induced by the shift of the indeterminate $t$  . Then the entropy of ${\alpha }_{t}$  is $h\left(\stackrel{^}{{\alpha }_{t}}\right)=m\left(gcd\left(J\left({\phi }_{1}\right)\right)\right).$
By combining Proposition  10 and  11 , we can see that the growth of orders gives another method to compute the entropy in Theorem  1 . Therefore, we can obtain the following corollary.
Corollary 1.
 $\begin{array}{cc}{lim}_{\stackrel{r\to \infty }{|{H}_{1}\left(\cdot \right)|\ne 0}}\frac{log|{H}_{1}\left({X}_{r};\mathbb{Z}\right)|}{r}={\sum }_{p\le \infty }{\sum }_{|{\alpha }_{i}{|}_{p}>1}log|{\alpha }_{i}{|}_{p},& \end{array}$
 $\begin{array}{}\end{array}$
where ${\alpha }_{i}$  are the zeros of the Alexander polynomial ${\Delta }_{K}\left(t\right)$  .
In a special case of Corollary  1 , we obtain that the growth is zero if and only if the zeros of ${\Delta }_{K}\left(t\right)$  are roots of unity. That is, we have the following.
Corollary 2 (Riley [25, González-Acun͂a and Short [8). Let ${X}_{r}$  be the $r$  -fold cyclic covering branched over $K$  . Then, if the Alexander polynomial ${\Delta }_{K}\left(t\right)$  have zeros which are not roots of unity, the finite values of the order of the first homology group $|{H}_{1}\left({X}_{r};\mathbb{Z}\right)|$  grows exponentially with respect to $r$  .
• Proof. (Indirect proof) Because all ${\alpha }_{i}$  belong to the valuation ring ${\mathcal{O}}_{p}=\left\{x\in \overline{{\mathbb{Q}}_{p}}||x{|}_{p}\le 1\right\}$  , $f\left(t\right)={\prod }_{i}\left(t-{\alpha }_{i}\right)$  , which is $\Delta \left(t\right)$  up to constant multiples, belongs to ${\mathbb{Z}}_{p}\left[t\right]\cap \mathbb{Q}\left[t\right]$  , where ${\mathbb{Z}}_{p}=\left\{x\in {\mathbb{Q}}_{p}||x{|}_{p}\le 1\right\}$  . This holds for any prime $p$  . Hence $f\left(t\right)\in \mathbb{Z}\left[t\right]$  . (Another way to see this is to prove Corollary  4 before. But in this proof, the condition $\Delta \left(1\right)=±1$  is not necessary.) Consequently, the zeros of the Alexander polynomial must be roots of unity from $|{\alpha }_{i}|\le 1$  and the Kronecker's theorem [14.
Remark 1. In [28, Silver and Williams generalized the result of González-Acun͂a and Short [8to links with dynamical systems viewpoint. And our study is motivated by their approach.

4.2 Leading coefficient of Alexander polynomial

In this section, we apply Theorem 1 to a criterion for being finitely generated as $\mathbb{Z}$  -module. Formerly, we have utilized the leading coefficient of the Alexander polynomial for the criterion. But it only determines whether the Alexander module is finitely generated or not. Now, we reveal that the leading coefficient of the Alexander polynomial is not just the criterion for being finitely generated as $\mathbb{Z}$  -module.
From the following corollary, we can regard the entropies $h\left({t}_{p}\right)$  for all primes $p<\infty$  as obstructions for being finitely generated.
Corollary 3. Let $h\left({t}_{p}\right)$  be the entropy of the meridian action on the $p$  -adic Alexander module ${H}_{1}\left({X}_{\infty };{\mathbb{Q}}_{p}\right)$  . If the Alexander module ${H}_{1}\left({X}_{\infty };\mathbb{Z}\right)$  is finitely generated as $\mathbb{Z}$  -module, then all the entropies $h\left({t}_{p}\right)$  are equal to zero for finite prime $p<\infty$  .
• Proof. The Alexander module ${H}_{1}\left({X}_{\infty };\mathbb{Z}\right)$  is finitely generated if and only if $\stackrel{^}{{H}_{1}\left({X}_{\infty };\mathbb{Z}\right)}$  is isomorphic to the $n$  dimensional torus. Then, it is covered by ${H}_{1}\left({X}_{\infty };\mathbb{R}\right)$  . By the well-known result for toral automorphisms, the entropy of the meridian action on $\stackrel{^}{{H}_{1}\left({X}_{\infty };\mathbb{Z}\right)}$  is
 $\begin{array}{cc}h\left(t\right)={\sum }_{|{\alpha }_{i}|>1}log|{\alpha }_{i}|,& \end{array}$
 $\begin{array}{}\end{array}$
where ${\alpha }_{i}$  are eigenvalues of the meridian action on ${H}_{1}\left({X}_{\infty };\mathbb{R}\right)$  .
These entropy must be equal to the entropy in Theorem 1-(2). Therefore, the entropies of the meridian action on ${H}_{1}\left({X}_{\infty };{\mathbb{Q}}_{p}\right)$  are zero for any $p<\infty$  , that is $h\left({t}_{p}\right)={\sum }_{|{\alpha }_{i}{|}_{p}>1}log|{\alpha }_{i}{|}_{p}=0\text{for}p<\infty .$
This obstructions give a new interpretation of the leading coefficient of the Alexander polynomial. In fact, the following Corollary means that the entropies $h\left({t}_{p}\right)$  are fine factors of the leading coefficient of the Alexander polynomial.
Corollary 4. Let ${\alpha }_{i}$  be the zeros of ${\Delta }_{K}\left(t\right)={\sum }_{i=0}^{n}{a}_{i}{t}^{i}$  . Then the leading coefficient of ${\Delta }_{K}\left(t\right)$  is the sum of the entropies of the meridian action on the $p$  -adic Alexander module ${H}_{1}\left({X}_{\infty };{\mathbb{Q}}_{p}\right)$  for the finite primes $p<\infty$  , that is
 $\begin{array}{cc}log|{a}_{n}|& ={\sum }_{p<\infty }{\sum }_{|{\alpha }_{i}{|}_{p}>1}log|{\alpha }_{i}{|}_{p}.\end{array}$
 $\begin{array}{}\end{array}$
• Proof. Let $f\left(t\right)={\Delta }_{K}\left(t\right)/{a}_{n}=\prod \left(t-{\alpha }_{i}\right)$  and $s$  the least common multiple of the denominators of the coefficients of $f\left(t\right)$  . Then,
 $\begin{array}{cc}{\sum }_{p<\infty }h\left({t}_{p}\right)={\sum }_{p<\infty }{\sum }_{|{\alpha }_{i}{|}_{p}>1}log|{\alpha }_{i}{|}_{p}=logs.& \end{array}$
 $\begin{array}{}\end{array}$
Because ${\Delta }_{K}\left(1\right)=±1$  , the coefficients are relatively prime $\left({a}_{n},...,{a}_{1}\right)=1$  .
Hence $s=|{a}_{n}|$  . (The above argument is essentially found in the proofs of Theorem 3 in [17and Theorem 2 in [25.)
Since ${\Delta }_{K}\left(1\right)=±1$  for any knot, the Alexander polynomial can be completely determined (up to $±1$  ) by the zeros. Hence the leading coefficient is also determined.
On the other hand, Corollary  4 shows furthermore that the distribution of the zeros measures how much different from being finitely generated $\mathbb{Z}$  -module the Alexander module is.

4.3 Final remarks

4.3.1 Determining knots by cyclic branched covers

In [12, Kojima showed that prime knots are determined by their cyclic branched covers. So, there might be a method to determine the Alexander module by the data of cyclic branched covers.
The growth of orders $|{H}_{1}\left({X}_{r};\mathbb{Z}\right)|$  does not determine completely the Alexander module ${H}_{1}\left({X}_{\infty };\mathbb{Z}\right)$  , but it has information of the Alexander module. In fact, the growth measures complexity of the Alexander module and is similar invariant with the leading coefficient of the Alexander polynomial.
In addition, it is still remain open whether infinitely many branched covers are necessary for determining the knot.

4.3.2 Volume conjecture

The volume (or Kashaev) conjecture [11, 19expects that the asymptotic behavior of the Kashaev invariant (= the specialization of the colored Jones polynomial) implies the hyperbolic volume of the complement of the knot (hyperbolic case).
In general, a topological entropy picks up natural measures, which are called measures of the maximal entropy, from the variational principle (see Proposition  5 ).
Then, the topological entropy is equal to the measure theoretic entropy with respect to a measure of the maximal entropy. In our case, the Haar measure on the Alexander module (precisely, the Plancherel measure on its dual group) is a measure of the maximal entropy, and this measure is lifted to adele rings with preserving the entropy.
Consequently the asymptotic behavior (growth of orders) can be translated into expansions of volumes with respect to the Haar measure on the adele ring because the growth is interpreted as the entropy of the meridian action.
Our point is that, when an asymptotic behavior is interpreted as an entropy, it can be related with a natural measure theory. Does this strategy work out for the volume conjecture?
References

1. R. L. Adler, A. G. Konheim, and M. H. McAndrew. Topological entropy. Trans. Amer. Math. Soc., 114:309–319, 1965.
2. R. Bowen. Entropy for group endomorphisms and homogeneous spaces. Trans. Amer. Math. Soc., 153:401–414, 1971.
3. G. Burde. Alexanderpolynome Neuwirthscher Knoten. Topology, 5:321–330, 1966.
4. G. Burde and H. Zieschang. Knots, volume 5 of de Gruyter Studies in Mathematics. Walter de Gruyter & Co., Berlin, 1985.
5. R. H. Crowell. The group ${G}^{\prime }/{G}^{\prime \prime }$  of a knot group $G$  . Duke Math. J., 30:349–354, 1963.
6. M. Einsiedler and T. Ward. Fitting ideals for finitely presented algebraic dynamical systems. Aequationes Math., 60(1-2):57–71, 2000.
7. R. H. Fox. Free differential calculus. III. Subgroups. Ann. of Math. (2), 64:407–419, 1956.
8. F. González-Acun͂a and H. Short. Cyclic branched coverings of knots and homology spheres. Rev. Mat. Univ. Complut. Madrid, 4(1):97–120, 1991.
9. C. M. Gordon. Knots whose branched cyclic coverings have periodic homology. Trans. Amer. Math. Soc., 168:357–370, 1972.
10. S. A. Juzvinskiĭ. Metric properties of the endomorphisms of compact groups. Izv. Akad. Nauk SSSR Ser. Mat., 29:1295–1328, 1965. English transl. in Ams. Math. Soc. Transl. 66 (1968), 63-98.
11. R. M. Kashaev. The hyperbolic volume of knots from the quantum dilogarithm. Lett. Math. Phys., 39(3):269–275, 1997.
12. S. Kojima. Determining knots by branched covers. In Low-dimensional topology and Kleinian groups (Coventry/Durham, 1984), volume 112 of London Math. Soc. Lecture Note Ser., pages 193–207. Cambridge Univ. Press, Cambridge, 1986.
13. A. N. Kolmogorov. A new metric invariant of transient dynamical systems and automorphisms in Lebesgue spaces. Dokl. Akad. Nauk SSSR (N.S.), 119:861–864, 1958.
14. L. Kronecker. Zwei Sätze ueber Gleichungen mit ganzzahligen Coeffichienten. J. Reine Angew. Math., 53:173–175, 1857.
15. W. B. R. Lickorish. An introduction to knot theory, volume 175 of Graduate Texts in Mathematics. Springer-Verlag, New York, 1997.
16. D. Lind, K. Schmidt, and T. Ward. Mahler measure and entropy for commuting automorphisms of compact groups. Invent. Math., 101(3):593–629, 1990.
17. D. A. Lind and T. Ward. Automorphisms of solenoids and $p$  -adic entropy. Ergodic Theory Dynam. Systems, 8(3):411–419, 1988.
18. K. Mahler. An application of Jensen's formula to polynomials. Mathematika, 7:98–100, 1960.
19. H. Murakami and J. Murakami. The colored Jones polynomials and the simplicial volume of a knot. Acta Math., 186(1):85–104, 2001.
20. K. Murasugi. The commutator subgroups of the alternating knot groups. Proc. Amer. Math. Soc., 28:237–241, 1971.
21. L. P. Neuwirth. Knot groups. Annals of Mathematics Studies, No. 56. Princeton University Press, Princeton, N.J., 1965.
22. C. V. Quach. Polynôme d'Alexander des noeuds fibrés. C. R. Acad. Sci. Paris Sér. A-B, 289(6):A375–A377, 1979.
23. D. Ramakrishnan and R. J. Valenza. Fourier analysis on number fields, volume 186 of Graduate Texts in Mathematics. Springer-Verlag, New York, 1999.
24. E. S. Rapaport. On the commutator subgroup of a knot group. Ann. of Math. (2), 71:157–162, 1960.
25. R. Riley. Growth of order of homology of cyclic branched covers of knots. Bull. London Math. Soc., 22(3):287–297, 1990.
26. D. Rolfsen. Knots and links, volume 7 of Mathematics Lecture Series. Publish or Perish Inc., Houston, TX, 1990. Corrected reprint of the 1976 original.
27. W. Rudin. Fourier analysis on groups. Wiley Classics Library. John Wiley & Sons Inc., New York, 1990. Reprint of the 1962 original, A Wiley-Interscience Publication.
28. D. S. Silver and S. G. Williams. Mahler measure, links and homology growth. Topology, 41(5):979–991, 2002.
29. D. S. Silver and S. G. Williams. Torsion numbers of augmented groups with applications to knots and links. Enseign. Math. (2), 48(3-4):317–343, 2002.
30. J. Sinaĭ. On the concept of entropy for a dynamic system. Dokl. Akad. Nauk SSSR, 124:768–771, 1959.
31. P. Walters. An introduction to ergodic theory, volume 79 of Graduate Texts in Mathematics. Springer-Verlag, New York, 1982.
32. A. Weil. Basic number theory. Springer-Verlag, New York, third edition, 1974. Die Grundlehren der Mathematischen Wissenschaften, Band 144.

Department of Mathematics, Tokyo Institute of Technology, Oh-okayama, Meguro-ku, Tokyo 152-8551, Japan E-mail address: akio@math.titech.ac.jp