Configuration (geometry)

From Wikipedia, the free encyclopedia
  (Redirected from Projective configuration)
Jump to: navigation, search
Configurations (4362) (a complete quadrangle, at left) and (6243) (a complete quadrilateral, at right).

In mathematics, specifically projective geometry, a configuration in the plane consists of a finite set of points, and a finite arrangement of lines, such that each point is incident to the same number of lines and each line is incident to the same number of points.[1]

Although certain specific configurations had been studied earlier (for instance by Thomas Kirkman in 1849), the formal study of configurations was first introduced by Theodor Reye in 1876, in the second edition of his book Geometrie der Lage, in the context of a discussion of Desargues' theorem. Ernst Steinitz wrote his dissertation on the subject in 1894, and they were popularized by Hilbert and Cohn-Vossen's 1932 book Anschauliche Geometrie, reprinted in English (Hilbert & Cohn-Vossen 1952).

Configurations may be studied either as concrete sets of points and lines in a specific geometry, such as the Euclidean or projective planes (these are said to be realizable in that geometry), or as a type of abstract incidence geometry. In the latter case they are closely related to regular hypergraphs and biregular bipartite graphs, but with some additional restrictions: every two points of the incidence structure can be associated with at most one line, and every two lines can be associated with at most one point. That is, the girth of the corresponding bipartite graph (the Levi graph of the configuration) must be at least six.


A configuration in the plane is denoted by (pγ π), where p is the number of points, the number of lines, γ the number of lines per point, and π the number of points per line. These numbers necessarily satisfy the equation

as this product is the number of point-line incidences (flags).

Configurations having the same symbol, say (pγ π), need not be isomorphic as incidence structures. For instance, there exist three different (93 93) configurations: the Pappus configuration and two less notable configurations.

In some configurations, p = and consequently, γ = π. These are called symmetric or balanced (Grünbaum 2009) configurations and the notation is often condensed to avoid repetition. For example, (93 93) abbreviates to (93).


A (103) configuration that is not incidence-isomorphic to a Desargues configuration

Notable projective configurations include the following:

Duality of configurations[edit]

The projective dual of a configuration (pγ lπ) is a (lπ pγ) configuration in which the roles of "point" and "line" are exchanged. Types of configurations therefore come in dual pairs, except when taking the dual results in an isomorphic configuration. These exceptions are called self-dual configurations and in such cases p = l.[2]

The number of (n3) configurations[edit]

The number of nonisomorphic configurations of type (n3), starting at n = 7, is given by the sequence

1, 1, 3, 10, 31, 229, 2036, 21399, 245342, ... (sequence A001403 in the OEIS)

These numbers count configurations as abstract incidence structures, regardless of realizability (Betten, Brinkmann & Pisanski 2000). As Gropp (1997) discusses, nine of the ten (103) configurations, and all of the (113) and (123) configurations, are realizable in the Euclidean plane, but for each n ≥ 16 there is at least one nonrealizable (n3) configuration. Gropp also points out a long-lasting error in this sequence: an 1895 paper attempted to list all (123) configurations, and found 228 of them, but the 229th configuration was not discovered until 1988.

Constructions of symmetric configurations[edit]

There are several techniques for constructing configurations, generally starting from known configurations. Some of the simplest of these techniques construct symmetric (pγ) configurations.

Any finite projective plane of order n is an ((n2 + n + 1)n + 1) configuration. Let Π be a projective plane of order n. Remove from Π a point P and all the lines of Π which pass through P (but not the points which lie on those lines except for P) and remove a line l not passing through P and all the points that are on line l. The result is a configuration of type ((n2 - 1)n). If, in this construction, the line l is chosen to be a line which does pass through P, then the construction results in a configuration of type ((n2)n). Since projective planes are known to exist for all orders n which are powers of primes, these constructions provide infinite families of symmetric configurations.

Not all configurations are realizable, for instance, a (437) configuration does not exist.[3] However, Gropp (1990) has provided a construction which shows that for k ≥ 3, a (pk) configuration exists for all p ≥ 2 lk + 1, where lk is the length of an optimal Golomb ruler of order k.

Higher dimensions[edit]

The concept of a configuration may be generalized to higher dimensions, for instance to points and lines or planes in space. In such cases, the restrictions that no two points belong to more than one line may be relaxed, because it is possible for two points to belong to more than one plane.

Notable three-dimensional configurations are the Möbius configuration, consisting of two mutually inscribed tetrahedra, Reye's configuration, consisting of twelve points and twelve planes, with six points per plane and six planes per point, the Gray configuration consisting of a 3×3×3 grid of 27 points and the 27 orthogonal lines through them, and the Schläfli double six, a configuration with 30 points, 12 lines, two lines per point, and five points per line.

See also[edit]

  • Perles configuration, a set of 9 points and 9 lines which do not all have equal numbers of incidences to each other


  1. ^ In the literature, the terms projective configuration (Hilbert & Cohn-Vossen 1952) and tactical configuration of type (1,1) (Dembowski 1968) are also used to describe configurations as defined here.
  2. ^ Coxeter 1999, pp. 106-149
  3. ^ This configuration would be a projective plane of order 6 which does not exist by the Bruck-Ryser theorem.


External links[edit]