Logarithmic sheaves attached to arrangements of hyperplanes (Q2474978)

Any divisor \(D\) on a nonsingular variety \(X\) defines a sheaf of logarithmic differential forms \(\Omega^1_X(\log D)\); various definitions and useful properties were discussed some years ago in a paper of \textit{K. Saito} [J. Fac. Sci., Univ. Tokyo, Sci. 1 A 27, 265--291 (1980; Zbl 0496.32007)]. This sheaf is locally free when \(D\) is a strictly normal crossing variety (the union of smooth divisors intersecting transversally). In the special case where \(D\) is a generic arrangement of hyperplanes in \({\mathbb{P}}^n\), many properties of \(\Omega^1_X(\log D)\) were studied by the author and \textit{M. Kapranov} [Duke Math. J. 71, No. 3, 633--664 (1993; Zbl 0804.14007)]. These properties included a Torelli type theorem asserting that two arrangements with isomorphic \(\Omega^1_X(\log D)\) coincide provided neither osculates a rational normal curve. The aim of the present paper is to introduce and study a certain subsheaf \(\widetilde{\Omega}^1_X(\log D)\) of \(\Omega^1_X(\log D)\), which has some good properties not shared by \(\Omega^1_X(\log D)\) in general. In particular there is always a residue exact sequence \(0\to\Omega^1_X\to\widetilde{\Omega}^1_X(\log D)\to\nu_*{\mathcal O}_{D'}\to0\), where \(\nu:D'\to D\) is a resolution of singularities of \(D\). In the special case where \(D={\mathcal A}\) is an arrangement of \(m\) hyperplanes in \({\mathbb{P}}^n\), we write \(\widetilde{\Omega}^1({\mathcal A})\) for \(\widetilde{\Omega}^1_X(\log D)\). Provided \(m\geq n+2\), there is then a projective resolution \(0\to{\mathcal O}_{{\mathbb{P}}^n}(-1)^{m-n-1}\to{\mathcal O}_{{\mathbb{P}}^n}^{m-1}\to\widetilde{\Omega}^1({\mathcal A})\to0\), so the Chern polynomial of \(\widetilde{\Omega}^1({\mathcal A})\) does not depend on the combinatorics of the arrangement. One can therefore introduce a notion of stability for arrangements and define a map \(\log:\text{ Ar}^{ss}_{n,m}\to{\mathcal S}_{n,m}\) from the variety of semistable arrangements to a connected component \({\mathcal S}_{n,m}\) of the Maruyama moduli space \({\mathcal M}_{{\mathbb{P}}^n}(n,(1-t)^{n-m+1})\). Let \(W(\widetilde{\Omega}^1({\mathcal A}))\) denote the set of unstable hyperplanes for \(\widetilde{\Omega}^1({\mathcal A})\), that is \(L\in\widetilde{\Omega}^1({\mathcal A})\) if and only if \(H^{n-1}(L,\widetilde{\Omega}^1({\mathcal A})(-n)| L)\neq0\). (In the case \(n=2\) the unstable lines are analogous to the jumping lines of \textit{W. Barth} [Invent. Math. 42, 63--91 (1977; Zbl 0386.14005)], but the two concepts do not coincide.) The hyperplanes of \({\mathcal A}\) belong to \(W(\widetilde{\Omega}^1({\mathcal A}))\); if these are the only hyperplanes in \(W(\widetilde{\Omega}^1({\mathcal A}))\), the arrangement \({\mathcal A}\) is called a Torelli arrangement. The main theorem of the paper (Theorem 5.1) asserts that the subset of \(\text{ Ar}^{ss}_{n,m}\) consisting of Torelli arrangements is Zariski open and the map \(\log\) is injective on this subset. The Torelli arrangements include generic arrangements with \(m\geq n+2\) which do not osculate a r ational normal curve. A natural conjecture is that a semistable arrangement of \(m\geq n+2\) hyperplanes in \({\mathbb{P}}^n\) is Torelli unless the corresponding points in the dual space \(\check{\mathbb{P}}^n\) lie on a stable rational normal curve. (A stable rational normal curve in \(\check{\mathbb{P}}^n\) is a non-degenerate connected reduced curve of arithmetic genus \(0\) and degree \(n\).) The layout of the paper is as follows. In section 2, the sheaf \(\widetilde{\Omega}^1_X({\log D})\) is defined and discussed. Section 3 is concerned with \(\widetilde{\Omega}^1({\mathcal A})\), the main result (Theorem 3.1) being the existence of a projective resolution of \(\widetilde{\Omega}^1({\mathcal A})\) as described above. Section 4 involves stability properties, leading up to the definition of stability for arrangements. Section 5 begins with the construction of the map \(\log\) and the definition of an unstable hyperplane and leads up to the main theorem. Finally, in section 6, the author discusses the case \(n=2\) in more detail and verifies the conjecture when \(n=2\), \(m\leq 6\).