Line and rational curve arrangements, and Walther's inequality (Q2335825)

In this interesting paper, the authors study two important invariants that one can associate with a line arragnement \(C\), namely the freeness defect \(\nu(C)\) and an upper bound for it, denoted by \(\nu'(C)\), called Walther's bound. Let \(S := \mathbb{C}[x,y,z]\) be the graded polynomial ring and \(C: f=0\) be a reduced curve of degree \(d\) in the complex projective plane \(\mathbb{P}^{2} := \mathbb{P}^{2}_{\mathbb{C}}\). The minimal degree \(\mathrm{mdr}(f)\) of the Jacobian relation for \(f\) is the smallest integer \(m\geq 0 \) such that there is a non-trivial relation \[a f_{x} + b f_{y} + c f_{z} = 0\] with \(a,b,c \in S_{m}\). Denote by \(\tau(C)\) the total Tjurina number which is equal to the sum of the Tjurina numbers of the singular points of \(C\). If \(J_{f}\) is the Jacobian ideal of \(f\) and \(\mathfrak{m} = \langle x,y,z \rangle\), then we consider the local cohomology group \[N(f) = I_{f} / J_{f} = H^{0}_{\mathfrak{m}}(M(f)),\] where \(I_{f}\) is the saturation of \(J_{f}\) with respect to \(\mathfrak{m}\) and \(M(f)\) is the Milnor algebra of \(f\). The defect of the freeness is defined as \[\nu(C) =\max_{j}\{n(f)_{j}\},\] where \(n(f)_{j}=\dim\, N(f)_{j}\) for any integer \(j\). The famous Terao's conjecture tells us that the freeness of line arrangements is combinatorially determined. One can formulate a stronger version of Terao's conjecture in the language of the defect. Conjecture 1. Let \(C : f=0\) be a line arrangement in \(\mathbb{P}^{2}\). Then the invariant \(\nu(C)\) is combinatorially determined. Let us denote by \(T\langle C \rangle \) the vector bundle of the logarithmic vector fields along \(C\) and denote by \((d_{1}^{\ell}, d_{2}^{\ell})\) the splitting type of the vector bundle \(T \langle C \rangle (-1)\) along a generic line \(\ell\) in \(\mathbb{P}^{2}\). Now we can restate the above conjecture in the language of the splitting type. Conjecture 2. Let \(C: f=0\) be a line arrangement in \(\mathbb{P}^{2}_{\mathbb{C}}\), then the generic splitting type \((d_{1}^{\ell}, d_{2}^{\ell})\) of the vector bundle \(T\langle C \rangle (-1)\) is combinatorially determined. It is worth emphasizing that there is only one known upper bound on \(\nu(C)\) obtained by Walther. This upper bound \(\nu'(C)\) is obtained by using the monodromy of the Milnor fiber \(F : f=1\) associated to a given curve \(C : f=0\) and the theory of Mixed Hodge Structures. A close inspection allowed the authors of the paper to formulate the following unexpected conjecture. Conjecture 3. Let \(C : f=0\) be an essential arrangement of \(d\geq 3\) lines in \(\mathbb{P}^{2}\) and let \(m(C)\) be the maximal multiplicity of an intersection point of \(C\). Then \(\nu(C) = \nu(C)'\) if and only if either \(m(C) = d-1\) or \(m(C) \leq 3\). The first main result of the paper is an application of Walther's result to the case of rational curve arrangements. We denote by \(st(f)\) the stability threshold \[st(f) =\min\{q:\dim M(f)_{k} = \tau(C) \text{ for all } k \geq q\},\] where \(C : f=0\) is a reduced plane curve. Theorem A. Let \(C : f=0\) be a plane curve of degree \(d\) such that each irreducible component of \(C\) is a rational curve. Then \(n(f) _{k} = 0\) for \(k\leq d - 3\) or \(k \geq 2d-3\). Moreover, \(st(F) \leq 2d-4\) with the equality when \(C\) is nodal. The main result of the paper can be formulated as follows. Theorem B. Let \(C : f=0\) be a line arrangement of \(d\geq 3\) lines in \(\mathbb{P}^{2}\). i) If \(C\) has only double and triple intersection points, then \(\mathrm{mdr}(f) \geq (d-2)/2\) -- Conjectures 1 and 2 hold. ii) For a line arrangement \(C : f=0\) satisfying \(\mathrm{mdr}(f) \geq (d-2)/2\) one has \(\nu(C) = \nu'(C)\) if and only if the line arrangement \(C\) has only double and triple points. In the context of Conjecture 3, the authors prove the following result. Theorem C. Let \(C : f=0\) be an arrangement of \(d\) lines in \(\mathbb{P}^{2}\). If \(3\leq d \leq 10\), then Conjectures 1, 2, and 3 hold for \(C\).