Fractal representation of the attractive lamination of an automorphism of the free group. (Q2372772)

Let \(\mathcal A=\{a_1,a_2,\dots,a_N\}\) be a finite alphabet, \(\mathcal A^*\) be the free monoid of finite words on \(\mathcal A\) with the empty word as trivial element and \(\mathcal A^\mathbb{N}\), \(\mathcal A^\mathbb{Z}\) be the sets of one-sided and two-sided (infinite) sequences on \(\mathcal A\). For a word \(w\in\mathcal A^*\) and \(a\in\mathcal A\), \(|w|_a\) denotes the number of occurrences of \(a\) in \(w\). So it is defined the Abelization map \(\ell\colon w\in\mathcal A^*\to(|w|_a)_{a\in\mathcal A}\in\mathbb{N}^N\). A substitution \(\sigma\) is an endomorphism of the free monoid \(\mathcal A^*\) such that the image of each letter of \(\mathcal A\) is non empty (without loss of generality it is always supposed that for at least one letter, say \(a\), the length of the successive iteration \(\sigma^n(a)\) tends to infinity). With every substitution \(\sigma\) it is associated an \(N\times N\) matrix \(M_\sigma\) (the incidence or transition matrix) induced by Abelization, i.e. \(\ell(\sigma(w))=M_\sigma\ell(w)\), \(w\in\mathcal A^*\). A substitution \(\sigma\) is primitive if there exists an integer \(k\) (independent of the letters) such that, for each pair \((a,b)\in\mathcal A^*\), the word \(\sigma^k(a)\) contains at least one occurrence of the letter \(b\). Let \(\sigma\) be a primitive substitution of \(\mathcal A^*\), \(S\) be the shift map on the set \(\mathcal A^\mathbb{N}\) (or \(\mathcal A^\mathbb{Z}\)) defined by \(S(u_i)=u_{i+1}\) for all \(i\) and for any sequence \(w=(u_i)_{i\in I}\) (\(I=\mathbb{N}\) or \(\mathbb{Z}\)) and \(X_w\) be the closure of the orbit of \(w\) under the action of \(S\): \(X_w=\overline{\{S^n(w)\mid n\in\mathbb{N}\}}\). For a periodic element \(w\) (\(\sigma^n(w)=w\) for some \(n>0\)) the pair \((X_w,S)\) is called the `symbolic dynamical system generated by \(\sigma\)'. Since \(\sigma\) is primitive and \(w\) periodic, \(X_w\) does not depend on \(w\), so it is denoted \((X_\sigma,S)\). An extensive study of the symbolic dynamical systems generated by a substitution \(\sigma\) has been done (see the many references in the paper). A substitution \(\sigma\) defined over the free monoid \(\mathcal A^*\) naturally extends to the free group on \(\mathcal A\) by defining \(\sigma(a^{-1})=(\sigma(a))^{-1}\), \(a\in\mathcal A\). Therefore it is natural to try to extend the previous construction to a general endomorphism of the free group \(F_N\) on \(N\) generators. This extension is not so easy for every endomorphism of a free group. Applying an endomorphism of a free group to a word it may appear cancellations, this implies that the construction of a fixed infinite word by iterating the image of a letter does not apply in general. Also a free group can be considered as the fundamental group of different graphs, this implies that many automorphisms of a free group are induced by a continuous map of the graph to itself, so it is difficult to understand what is intrinsic in the construction of a symbolic dynamical system generated by an automorphism. In the paper under consideration the authors focus only on special cases of automorphisms of the free groups and exhibit a connection between the geometric representation of symbolic dynamical systems and the theory of free groups automorphisms. To obtain this they invoke the substantial work on automorphisms of free groups by \textit{M. Bestvina, M. Feighn} and \textit{M. Handel} [in Geom. Funct. Anal. 7, No. 2, 215-244 (1997); erratum ibid. 7, 1143 (1997; Zbl 0884.57002), Ann. Math. (2) 151, No. 2, 517-623 (2000; Zbl 0984.20025)] and by \textit{M. Bestvina} and \textit{M. Handel} [in Topology 34, No. 1, 109-140 (1995; Zbl 0837.57010), Ann. Math. (2) 135, No. 1, 1-51 (1992; Zbl 0757.57004)]. Recalling the terminology and the results established there, let \(R_N\) be the rose with \(N\) petals and \(G\) be a topological graph with \(1-\mathcal X(G)=N\) (\(\mathcal X(G)\) is the Euler characteristic of \(G\)) and \(\tau\colon R_N\to G\) be a (marking) homotopy equivalence which makes \(G\) a marked graph and induces an isomorphism between the (free) fundamental groups \(\pi_1(R_N,*)\simeq\pi_1(G,\tau(*))\simeq F_N\). Suppose that \(\Phi\in\text{Out}(F_N)\) is an outer automorphism, a `topological representative' of \(\Phi\) is a map \(f\colon G\to G\), where \(G\) is a marked graph, such that: \(\bullet\) the image of a vertex is a vertex, \(\bullet\) the image of an edge is a finite path of \(G\), \(\bullet\) \(f\) induces \(\Phi\) on \(F_N\simeq\pi_1(G,\tau(*))\). A `train-track map' is a topological representative \(f\colon G\to G\) of an outer free group automorphism, such that: \(\bullet\) \(G\) has no vertex of valence 1 or 2, \(\bullet\) for all edges \(e\in\mathcal E(G)\) and all \(n>0\), \(f^n(e)\) is a path in \(G\). Let \(f\colon G\to G\) be a topological representative of an outer automorphism \(\Phi\), if \(e_1,e_1^{-1},\dots,e_k,e_k^{-1}\) denote the edges of \(G\), the transition matrix \(M_f\) associated with \(f\) is a \(k\times k\) matrix with \((i,j)\)-entry the number of times \(f(e_i)\) crosses \(e_j\) or \(e_j^{-1}\). If \(M_f\) is primitive then it has unique dominant real eigenvalue \(\lambda_f>1\), \(\lambda_f\) depends on the outer automorphism \(\Phi\), so it is called dilation coefficient of \(\Phi\) and denoted by \(\lambda_\Phi\). An automorphism \(\varphi\in\Aut(F_N)\) is called `iwip', that is, irreducible with irreducible powers, if no proper free factor \(F\) of \(F_N\) is mapped by any positive power of \(\varphi\) to a conjugate of \(F\). An outer automorphism \(\Phi\) is iwip if one automorphism \(\varphi\in\Aut(F_N)\) in \(\Phi\) is iwip. An iwip automorphism is said to be unit Pisot if its dilation coefficient is a unit Pisot number. Let \(f\colon G\to G\) be a train-track map. Let \(\mathcal A=\{e_1,e_1^{-1},\dots,e_k,e_{k}^{-1}\}\) and \(h\) the natural map from the set of paths in \(G\) to \(\mathcal A^*\) which, with any finite path, associates its name, namely the word given by the names of successive edges contained in the path. The `double substitution' \(\sigma_f\) associated with \(f\) is defined by \(\sigma_f(e)=h(f(e))\) for every \(e\in\mathcal A\). A double substitution \(\sigma_f\) is orientable if there is an orientation of \(G\) (a subset \(\mathcal E\) of \(\mathcal A\) such that \(\mathcal E\cup\mathcal E^{-1}=\mathcal A\) and \(\mathcal E\cap\mathcal E^{-1}=\emptyset\)) such that \(\sigma_f(\mathcal E)\subseteq\mathcal E\) or \(\sigma_f(\mathcal E)\subseteq\mathcal E^{-1}\). Otherwise \(\sigma_f\) is nonorientable. For a train-track map \(f\) and the associated double substitution \(\sigma_f\) the `attractive symbolic lamination' is defined to be the set \(X_{\sigma_f}\) of the symbolic dynamical system \((X_{\sigma_f},S)\) in case where \(\sigma_f\) is nonorientable and it is denoted by \(L_f\). A primitive substitution is said to be unit Pisot if its dominant eigenvalue is a unit Pisot number. A primitive substitution is said to be irreducible if the characteristic polynomial of its incidence matrix is irreducible. For a primitive unit Pisot substitution \(\sigma\) with dominant eigenvalue \(\beta\) the `central tile' is defined to be the projection on the \(\beta\)-contracting plane of the discrete line associated with any periodic point \(u=(u_i)_{i\in\mathbb{N}}\) of \(\sigma\): \[ \mathcal T=\overline{\{\pi_c(\ell(u_0\cdots u_{i-1}));\;i\in\mathbb{N}\}}. \] Subtiles of the central tile \(\mathcal T\) are naturally defined, depending on the letter associated with the vertex of the discrete line that is projected, namely \[ \mathcal T_\sigma(a_j)=\overline{\{\pi_c(\ell(u_0\cdots u_{i-1}));\;i\in\mathbb{N},\;u_i=a_j\}}. \] Let \(\Phi\) be a unit Pisot outer automorphism and let \(f\colon G\to G\) be a train-track map representing \(\Phi\). If the double substitution \(\sigma_f\) is nonorientable, then the central tile of \(f\) is defined to be the central tile associated with its double substitution \(\sigma_f\). It is denoted by \(\mathcal T_f\) and the subtiles by \(\mathcal T_f(e)\). Now we are in the position to quote some of the results in the paper: Theorem 4.11. Let \(f\colon G\to G\) be a nonorientable train-track for a unit Pisot iwip outer automorphism. Let \(2k\) be the number of edges of \(G\). Let \(d\) be the degree of the dilation coefficient of \(f\). The central tile of \(f\) is a compact subset with nonempty interior, hence non-zero measure, of a \((d-1)\)-dimensional subspace of \(\mathbb{R}^{2k}\). It is divided into \(2k\) subtiles. (A detailed description of the subtiles is given in the Theorem 4.14.) Theorem 4.18. With the assumptions above, assume, furthermore, that \(\sigma_f\) satisfies the strong coincidence condition. Then the domain exchange \(E_f\) is defined almost everywhere on the central tile \(\mathcal T_f\). The attractive symbolic lamination \(\mathcal L_f\) provided with the shift map \(S\) is measure-theoretically isomorphic to \((\mathcal T_f,E_f)\), i.e., there exists a map \(\mu\colon L_f\to\mathcal T_f\) that is continuous, onto and one-to-one almost everywhere, and that satisfies \(\mu\circ S=E_\sigma\circ\mu\). The paper concludes with some examples and a discussion on the stability of the representation under conjugation.