Thurston's pullback map on the augmented Teichmüller space and applications (Q443972)

In a major innovation, the author shows that the Thurston's pullback map on the Teichmüller space induced by a postcritically finite branched self-covering \(f: S^2 \to S^2\) of the 2-sphere extends to the augmented Teichmüller space. He then applies this analysis to show that the first return map of each hyperbolic part in the canonical decomposition of \(f\) (in the sense of [the reviewer, Adv. Math. 158, No. 2, 154--168 (2001; Zbl 1193.57002)]) is equivalent to a rational function, answering a conjecture of the reviewer from [Combinations of complex dynamical systems. Berlin: Springer (2003; Zbl 1045.37028)]. In more detail: let \(f: S^2 \to S^2\) be an orientation-preserving branched covering map of the two-sphere to itself of degree \(\geq 2\), and suppose the postcritical set \(P_f\) (that is, the union of the forward orbits of the set of branch points) is finite. For brevity we call \(f\) a \textit{Thurston map}. Two such maps \(f, g\) are called \textit{combinatorially equivalent} if they are conjugate up to isotopy through homeomorphisms agreeing on \(P_f\). Thurston (see [\textit{A. Douady} and \textit{J. H. Hubbard}, Acta Math. 171, No. 2, 263--297 (1993; Zbl 0806.30027)]) provided a complete topological characterization of those Thurston maps which are combinatorially equivalent to rational functions acting on the Riemann sphere. Apart from very special cases, the rational function is unique up to Möbius conjugacy. Let \(\mathcal{T}_f\) denote the Teichmüller space modelled on \((S^2, P_f)\). The proof of this theorem shows that a Thurston map \(f\) is equivalent to a rational map if and only if a certain pullback map \(\sigma_f: \mathcal{T}_f \to \mathcal{T}_f\) has a fixed point. A simple calculation with quadratic differentials shows that \(\sigma_f\) cannot increase distances and, in fact, its second iterate always decreases distances (though not uniformly) unless \(f\) is a member of the ubiquitous family of so-called Lattès exceptions. The author considers a slightly more general setup where \(P_f\) is replaced with a forward invariant set \(P \supset P_f\). He observes that essentially the same calculation, together with the Schwarz-Pick lemma, shows that the corresponding pullback map \(\sigma_f\) is \(\sqrt{\deg(f)}\)-Lipschitz with respect to the Weil-Petersson (WP) metric on \(\mathcal{T}_f\), now modelled on \((S^2, P)\). The WP completion is, by a result of \textit{H. Masur} [Duke Math. J. 43, 623--635 (1976; Zbl 0358.32017)], naturally the augmented Teichmüller space \(\overline{T}_f\), and so \(\sigma_f\) extends to a map \(\overline{\sigma}_f: \overline{\mathcal{T}}_f \to \overline{\mathcal{T}}_f\). The augmented Teichmüller space is similarly defined using noded Riemann surfaces, and is a stratified space whose strata \(\mathcal{T}^\Gamma_f\) are in bijective correspondence with (possibly empty) multicurves \(\Gamma \subset S^2 - P\). The corresponding mapping class group acts isometrically, and its quotient is a compactification of moduli space. The author shows that \(\overline{\sigma}_f(\mathcal{T}_f^\Gamma) \subset \mathcal{T}_f^{f^{-1}(\Gamma)}\), and hence that (totally) invariant multicurves correspond to invariant strata. From this, it follows that unlike the dynamics on \(\mathcal{T}_f\) itself, the dynamics of \(\overline{\sigma}_f\) can be interesting -- it can have periodic points, for example. By projecting an orbit of \(\sigma_f\) to the moduli space, one can find an accumulation point on a stratum of smallest dimension. The author then shows (i) if this accumulation point lies in the moduli space itself (and not its boundary), then \(f\) is equivalent to a rational map, and (ii) in the absence of so-called Thurston obstructions, this must be the case. This gives a new, slightly different proof of Thurston's fundamental characterization theorem. In the presence of Thurston obstructions, the author shows that there is a canonical such obstruction \(\Gamma_f\) and that orbits of \(\sigma_f\) accumulate on the corresponding stratum; this reproves the main results of the reviewer from [Adv. Math. 158, No. 2, 154-168 (2001; Zbl 1193.57002)]. If in addition all parts of the canonical decomposition have hyperbolic orbifold, this orbit accumulates at a unique fixed-point in this stratum. The author's results raise fundamental finiteness questions, many still open at the time of the writing of this review, regarding the dynamics of \(\overline{\sigma}_f\). For example, if \(f\) is rational, can there exist strata which wander under iteration of \(\overline{\sigma}_f\)?