Roots of the characteristic polynomials of hyperplane arrangements and their restrictions and localizations (Q2139837)

Let \(\mathcal{A} \subset V:=\mathbb{K}^{l}\) be a central arrangement of hyperplanes and take \(H \in \mathcal{A}\). Consider a triple \((\mathcal{A},\mathcal{A}',\mathcal{A}^{H})\) which is defined by \(\mathcal{A}' := \mathcal{A} \setminus \{H\}\) and \(\mathcal{A}^{H} :=\{L \cap H \, : L \in \mathcal{A}'\}\). Let us denote by \(\chi(\mathcal{A},t)\) a characteristic polynomial of \(\mathcal{A}\) and recall that if \(\mathbb{K} = \mathbb{C}\), then \(\chi(\mathcal{A},t) = \mathrm{Poin}(V \setminus \cup_{L \in \mathcal{A}}L,t)\). It is well known that \(\chi(\mathcal{A},t)\) is not easy to compute in general, and in that context the most useful method is the so-called deletion-restriction theorem which tells us that \[ \chi(\mathcal{A},t) = \chi(\mathcal{A}',t) - \chi(\mathcal{A}^{H},t). \] For an arrangement \(\mathcal{A}\) we define the logarithmic derivation module \(D(\mathcal{A})\) as \[ D(\mathcal{A}) := \{\theta \in\mathrm{Der}(S): \theta(\alpha_{H}) \in S \cdot \alpha_{H}, \,\,\, \forall \, H \in \mathcal{A}\}, \] where \(\alpha_{H}\) is a defining linear form for \(H \in \mathcal{A}\). We say that \(\mathcal{A}\) is free with the exponents \(\exp(\mathcal{A}) = (d_{1}, ..., d_{l})\) if \(D(\mathcal{A})\) is a free \(S\)-module with a homogeneous basis \(\theta_{1}, ..., \theta_{l}\) of degree \(\deg \theta_{i} = d_{i}\) for \(i =1, ..., l\). By Terao's factorization theorem we know that if \(\mathcal{A}\) is free with the exponents \(\exp(\mathcal{A}) = (d_{1}, ...,d_{l})\), then \[ \chi(\mathcal{A},t) = \prod_{i=1}^{l}(t-d_{i}). \] We say that \(\mathcal{A}\) is integer rooted if \(\chi(\mathcal{A},t)\) has \(l\)-integer roots, and is real rooted if \(\chi(\mathcal{A},t)\) has \(l\)-real roots. The main problem that is considered in the present note can be summed up as follows. Problem: Describe the roots of \(\chi(\mathcal{A}_{X},t)\) and \(\chi(\mathcal{A}^{H},t)\) when they are both integer, or real rooted and \(\mathcal{A}\) is integer, or real rooted, where \(\mathcal{A}_{X}\) is the localization of \(\mathcal{A}\) with respect to \(X \in L(\mathcal{A})\). In order to present the main results of the paper under review, we need the following definition. Definition: For non-negative integers \(a_{1}, ..., a_{l}\), let \[ A(a_{1}, ..., a_{l}):= \sum_{1\leq i < j \leq l}|a_{i}-a_{j}|^{2}, \] \[ B(a_{1}, ..., a_{l}):=\sum_{1\leq i<j\leq l}a_{i}a_{j}. \] Theorem A. Let \(\mathcal{A}\) be a real rooted \(l\)-arrangement with \(\chi_{0}(\mathcal{A},t) = \prod_{i=2}^{l}(t-d_{i})\), \(d_{i} \in \mathbb{R}\). For \(H \in \mathcal{A}\) assume that \(\mathcal{A}^{H}\) is real rooted with \(\chi_{0}(\mathcal{A}^{H},t) = \prod_{i=2}^{l-1}(t-e_{i})\). Then \[ A(d_{2}, ..., d_{l}) \leq A(e_{2}, ..., e_{l-1}, |\mathcal{A}| - |\mathcal{A}^{H}|). \] Let \(\mathcal{A}\) be an essential \(l\)-arrangement and \(X \in L_{l-1}(\mathcal{A})\), we denote by \(\mathcal{A}_{X}^{e}\) the essentialization of \(\mathcal{A}_{X}\), i.e., \(\mathcal{A}_{X} = \mathcal{A}_{X}^{e} \oplus \emptyset_{1}\). Theorem B. Let \(\mathcal{A}\) be an essential arrangement with \(\chi_{0}(\mathcal{A},t) = \prod_{i=2}^{l}(t-d_{i})\), \(d_{i} \in \mathbb{R}\). Assume that \(d_{l-1}+1\leq d_{l}-1\). Then for \(e_{i} = d_{i}\) with \(i\leq l-2\), \(d_{l-1} := d_{l-1}+1\), \(e_{l} :=d_{l}-1\), there are no \(X \in L_{l-1}(\mathcal{A})\) such that \[ \chi_{0}(\mathcal{A}_{X}^{e},t) = \prod_{i=2}^{l}(t-e_{i})/(t-e_{j}). \] Theorem C. Let \(\mathcal{A}\) be an arrangement in \(\mathbb{K}^{3}\), \(H \in \mathcal{A}\) and \(\mathcal{A}':=\mathcal{A}\setminus \{H\}\). Assume that \(\mathcal{A}\) and \(\mathcal{A}'\) are both integer rooted. a) Assume that \(\mathcal{A}\) is free with \(\exp(\mathcal{A}) = (1,a,b)\), \(a \leq b\), and \(\chi_{0}(\mathcal{A}',t) = (t-c)(t-d)\) with \(c\leq d\). Then \(a-1 \leq c \leq d \leq b\). b) Assume that \(\mathcal{A}'\) is free with \(\exp(\mathcal{A}) = (1,a,b)\), \(a\leq b\), and \(\chi_{0}(\mathcal{A},t) = (t-c)(t-d)\) with \(c\leq d\). Then \(a\leq c \leq d\leq b+1\).