Hyperplane arrangements and their cells

An affine hyperplane in real d-dimensional space is a set where a nonconstant affine linear form equals zero. A finite arrangement consists of distinct such hyperplanes. Away from their union, each point assigns a positive or negative sign to every defining form. A nonempty sign pattern defines an intersection of open halfspaces, hence a convex connected region. Some formal sign patterns are infeasible, so counting all binary strings generally overcounts the geometry.

The closure of a full-dimensional region is a convex polyhedron; it may be unbounded, even though the arrangement is finite. Lower-dimensional cells occur where one or more signs become zero. Our region count concerns the full-dimensional components of the complement, not every cell of every dimension. This convention matters when a software display labels intersection points and segments separately from the two-dimensional regions between lines.

In affine general position, every collection of k hyperplanes with k at most d intersects in dimension d − k, and no d + 1 hyperplanes have a common point. In the plane this means no parallel pair and no triple concurrence. The previous hyperplanes restrict to a general-position arrangement within the new hyperplane. Every region of that induced arrangement separates one previously existing region into two, producing the recurrence displayed above.

Pascal’s identity solves the recurrence: the number of regions is the sum of binomial coefficients from zero through d, using zero when the lower index exceeds N. In the plane this is one plus N plus the number of pairs of lines. In three dimensions an additional term counts triples of planes. The formula is polynomial in N for fixed d, which is far smaller than the two-to-the-N possible sign strings once N grows.

If all planar lines pass through one common point, N distinct lines create only 2N regions for N positive. If they are all parallel, they create N + 1 regions. The general-position formula therefore cannot be used without checking its assumptions. Small perturbations can separate a concurrence into multiple intersection points and increase the region count; that perturbation changes the exact arrangement, even if a low-resolution drawing looks nearly identical.

The recurrence survives beyond general position in the more careful deletion-and-restriction form: remove a chosen hyperplane and consider the distinct induced hyperplanes inside it, discarding empty restrictions and repetitions. Their intersection structure determines the extra regions. Stanley’s characteristic-polynomial theory packages this data through the intersection poset, but one should first understand the geometric recurrence. An affine bijection preserves all cells and their incidences, whereas changing a line offset may genuinely change the combinatorics.