We show exponential growth of torsion numbers for links whose first nonzero Alexander polynomial has positive logarithmic Mahler measure. This extends a theorem of Silver and Williams to the case of a null first Alexander polynomial and provides a partial solution for a conjecture of theirs.
57M10; 57M25, 57Q10
IntroductionLet M be a compact three-manifold; the homology groups H i .M / can be written as the direct sums H i .M / tors˚Hi .M / free of a finite abelian group with a finite-rank free abelian group. The torsion summand is nontrivial only for i D 1: H 0 and H 3 are Z or 0, and the universal coefficients theorem for cohomology implies that H 1 .M; @M / is free, and by Poincaré duality it follows that H 2 .M / is also torsion-free. On the other hand, the torsion in H 1 can be arbitrarily large (for example, for lens spaces; see below for hyperbolic examples) and it is believed that "most" 3-manifolds should have a rather large torsion. For example E Kowalski shows in [9, Proposition 7.19] that the first homology group of a "Dunfield-Thurston random 3-manifold" typically has a large torsion subgroup. This paper is concerned with the growth rate of the order of H 1 .M N / tors in a sequence of finite coverings M N of a manifold M . The least precise question that can be asked is whether it is exponential in the degree or not, that is, whether the sequencehas a positive limit (or limit superior). This shall be partially answered here in the case where the M N are abelian coverings converging to a free abelian covering of M . The main motivation to study this question was to provide a partial result towards a conjecture of Silver and Williams on the growth rate of torsion numbers of abelian coverings of complements of links (see Conjecture 6.1] or (0-1) below).Historically, the first context where the growth of torsion in the homology of coverings has been studied is that of cyclic coverings of a knot complement. Let K be an open knotted solid torus in the three-sphere and M D S 3 K ; then M is a compact threemanifold with H 1 .M / D Z. Thus we can consider the infinite cyclic covering M of M and its finite quotients M N , which are the finite coverings of M corresponding to