Sigma-invariants and tropical varieties (Q2049956)

This paper is about the relationship obtained through tropical geometry between the Bieri-Neumann-Strebel-Renz invariants [\textit{R. Bieri} and \textit{B. Renz}, Comment. Math. Helv. 63, No. 3, 464--497 (1988; Zbl 0654.20029)] and the characteristic varieties of a CW-complex. We start this review by defining the Bieri-Neumann-Strebel-Renz invariants \(\Sigma^q(X,\mathbb{Z})\) of a connected CW-complex \(X\). Let \(\widetilde{X}\) be the universal cover of \(X\), which is endowed with a CW-complex structure by lifting the cells of \(X\). Let \(G\) be the fundamental group of \(X\). The set \[S(X)=(\mathrm{Hom}(G,\mathbb{R})\setminus\{0\})/\mathbb{R}^+\] is called the \textit{character sphere}, that is, the set of nonzero homomorphisms \(G\to\mathbb{R}\) modulo homotethy. Note that \(\mathrm{Hom}(G,\mathbb{R})\cong \mathrm{Hom}(H_1(X,\mathbb{R}),\mathbb{R})=H^1(X,\mathbb{R})\), so the character sphere can be thought of as the unit sphere in the vector space \(H^1(X,\mathbb{R})\). For every character \(\chi:G\rightarrow \mathbb{R}\), let \(G_\chi\) be the submonoid of \(G\) given by \[ G_\chi=\{g\in G\mid \chi(g)\geq 0\}. \] The cellular chain complex \(C_*(\widetilde{X},\mathbb{Z})\) is endowed with a (free) \(\mathbb{Z}G\)-module structure, where \(G\) acts on \(C_*(\widetilde{X},\mathbb{Z})\) by deck transformations. Consequently, it's also endowed with a \(\mathbb{Z}G_\chi\)-module structure. We say that a chain complex \(C\) of modules over a ring \(R\) is of finite \(q\)-type over \(R\) if there exists a chain complex \(C'\) of finitely generated, projective (left) \(R\)-modules and a chain map \(C'\rightarrow C\) inducing isomorphisms in homology for degrees \(i<q\) and an epimorphism in degree \(q\). Definition: [\textit{M. Farber} et al., Russ. Math. Surv. 65, No. 1, 173--174 (2010; Zbl 1245.57005)] Let \(X\) be a connected CW-complex \(X\) with finite \(1\)-skeleton. The \textit{Bieri-Neumann-Strebel-Renz invariants} \(\Sigma^q(X,\mathbb{Z})\) of \(X\) are \[ \Sigma^q(X,\mathbb{Z})=\{\chi\in S(X)\mid C_*(\widetilde{X},\mathbb{Z})\text{ is of finite }q\text{-type over }\mathbb{Z}G_\chi\}. \] These invariants only depend on the homotopy type of \(X\), and are a generalization (up to sign convention) of the Bieri-Neumann-Strebel (BNS) invariant \(\Sigma^1(G)\) of a finitely generated group \(G\) [\textit{R. Bieri} et al., Invent. Math. 90, 451--477 (1987; Zbl 0642.57002)], the generalization being given by associating to \(G\) the CW complex \(K(G,1)\). The actual computation of these invariants is very complicated, and the goal of this paper is to find a computable approximation which strengthens the bound given by \textit{S. Papadima} and \textit{A. Suciu} [Proc. Lond. Math. Soc. (3) 100, No. 3, 795--834 (2010; Zbl 1273.55003)]. Definition: Let \(X\) be a connected CW-complex \(X\) with finite \(q\)-skeleton, for some \(q\geq 1\). The \textit{characteristic varieties} of \(X\) (in degree \(i\leq q\)) are \[ \mathcal{V}^i(X)=\{\rho\in H^1(X,\mathbb{C}^{\times})\mid H_i(X,\mathbb{C}_\rho)\neq 0\}, \] where \(\mathbb{C}_\rho\) is the rank \(1\) \(\mathbb{C}\)-local system corresponding to the character \(\rho:G\rightarrow\mathbb{C}^{\times}\). Before we state the main result of this paper, we define what the tropicalization of the characteristic varieties is. Let \(\mathbb{K}^{\times}=\mathbb{C}\{\{t\}\}\) be the field of Puiseux series with complex coefficients, and let \(\mathrm{val}\) be the valuation given by \[ \begin{array}{cccc} \mathrm{val}: & \mathbb{K} & \longrightarrow & \mathbb{Q}\subset\mathbb{R}\\ \ & c_1t^{a_1}+c_2t^{a_2}+\cdots & \longmapsto & a_1 \end{array} \] where \(c_i\in\mathbb{C}^{\times}\) for all \(i\geq 1\) and \(a_1<a_2<\cdots\) are rational numbers that have a common denominator. The map \[ \nu_X:H^1(X,\mathbb{K}^\times)\longrightarrow H^1(X,\mathbb{R}) \] is defined as the coefficient homomorphism induced by \(\mathrm{val}\). Definition: Let \(W\) be a subvariety of \(H^1(X,\mathbb{C}^\times)\). The \textit{tropicalization} of \(W\) is given by \[ \mathrm{Trop}(W)=\overline{\nu_X(W\times_{\mathbb{C}}\mathbb{K})}, \] where the closure is in the Euclidean topology of \(H^1(X,\mathbb{R})\). If \(V\) is a subset of \(H^1(X,\mathbb{R})\), we denote by \(S(V)\) the set \(S(X)\cap V\), where the intersection is in \(H^1(X,\mathbb{R})\). The main result of this paper is the following theorem. Theorem: Let \(X\) be a connected CW-complex with finite \(q\)-skeleton for some \(q\geq 1\), and let \(\mathcal{V}^{\leq q}(X)=\cup_{i\leq q}\mathcal{V}^i(X)\). Then, \[ \Sigma^q(X,\mathbb{Z})\subseteq S\left(\mathrm{Trop}(\mathcal{V}^{\leq q}(X))\right)^c, \] where the upper script \(c\) denotes the complement in \(S(X)\). This result improves on the previous bound from \textit{S. Papadima} and \textit{A. Suciu} [Proc. Lond. Math. Soc. (3) 100, No. 3, 795--834 (2010; Zbl 1273.55003)], namely \[\Sigma^q(X,\mathbb{Z})\subseteq S\left(\tau_1^\mathbb{R}(\mathcal{V}^{\leq q}(X))\right)^c.\] Here, \(\tau_1^\mathbb{R}(W)\) is the \textit{real exponential tangent cone of \(W\) at \(1\)} for a given a subvariety \(W\) of \(H^1(X,\mathbb{C}^\times)\). More precisely, \[ \tau_1^\mathbb{R}(W)=\{z\in H^1(X,\mathbb{C})\mid \exp(\lambda z)\in W,\text{ for all }\lambda\in\mathbb{C}\}\cap H^1(X,\mathbb{R}) \] where \(\exp:H^1(X,\mathbb{C})\to H^1(X,\mathbb{C}^\times)\) denotes the coefficient homomorphism induced by the exponential map. Indeed, the main result (tropical bound) improves on the previous result (exponential tangent cone bound) because, as it is proved in this paper, \(\tau_1^\mathbb{R}(W)\subseteq \mathrm{Trop}(W)\) for every algebraic subvariety \(W\subset H^1(X,\mathbb{C}^\times)\). The paper contains many examples and applications of the main result. There are explicit examples in which the tropical bound given by the main result of this theorem for \(q=1\) is better than the exponential tangent cone bound. These examples can arise in the world of compact connected orientable \(3\)-manifolds, where the author uses the tropical bound to recover a known upper bound for the BNS invariant in terms of the Alexander norm, see [\textit{N. M. Dunfield}, Pac. J. Math. 200, No. 1, 43--58 (2001; Zbl 1049.57012)], which in turn uses a result relating the Thurston and Alexander norm unit balls from \textit{C. T. McMullen} [Ann. Sci. Éc. Norm. Supér. (4) 35, No. 2, 153--171 (2002; Zbl 1009.57021)]. There are also examples of this behavior in the world of essential central hyperplane arrangement complements in \(\mathbb{C}^n\), where the open question of whether the tropical bound for \(q=1\) is always an equality in this case is posed. The tropical bound is in fact an equality in some of the examples that appear on this paper. For example, the tropical bound allows the author to show that certain \(\Sigma^1\) invariants are empty for some orbifold Riemann surfaces and some Seifert manifolds. The bound is also an equality for compact Kähler manifolds by a result of \textit{T. Delzant} [Math. Ann. 348, No. 1, 119--125 (2010; Zbl 1293.32030)]. Some explicit examples where equality is achieved (but not for the exponential tangent cone bound) are also given in the case when \(G\) is a one-relator group, a case in which the BNS invariant can be computed.