On the difference between real and complex arrangements (Q1319293)

Let \({\mathcal B}\) be an arrangement of complex hyperplanes in \(\mathbb{C}^ d\). Then a classical result by \textit{P. Orlik} and \textit{L. Solomon} [Invent. Math. 56, 167-189 (1980; Zbl 0432.14016)] asserts that the cohomology algebra \(H^*(\mathbb{C}^ d \setminus \bigcup{\mathcal B};\mathbb{Z})\) of the complement can be constructed from the combinatorial data that are given by the intersection lattice. If \({\mathcal B}'\) is, more generally, a 2-arrangement in \(\mathbb{R}^{2d}\) (an arrangement of real subspaces of codimension 2 with only even-dimensional intersections), then the intersection lattice still determines the cohomology groups \(H^ i(\mathbb{R}^{2d}\setminus \bigcup{\mathcal B}';\mathbb{Z})\) of the complement, as was shown by \textit{M. Goresky} and \textit{R. MacPherson} [Stratified Morse theory (1988; Zbl 0639.14012)]. We prove, however, that for 2-arrangements the cohomology algebra is not determined by the intersection lattice. It encodes extra information on sign patterns, which can be computed from determinants of linear relations or, equivalently, from linking coefficients in the sense of knot theory. This also allows us to identify arrangements with the same lattice but with different fundamental groups (in the case \(d=2\), or with non-homeomorphic links (in the case \(d=3\)).