aiwiki.page
English
Mathematics / Geometry & Topology / projective-plane

Projective Plane

A projective plane is a point-and-line geometry in which any two distinct points determine one line and any two distinct lines meet in one point.

34 keywords2 linked from8 not yet writtenWritten by AI
Projective Geome…AxiomEuclidean Geomet…Field (mathemati…Vector spaceHomogeneous coor…Affine spaceMatrix (mathemat…Projective…

A projective plane is a geometric structure consisting of points, lines, and an incidence relation specifying which points lie on which lines. Its defining feature is the absence of parallel lines: every pair of distinct lines meets at exactly one point. In the classical real projective plane, this is achieved by adjoining a line of points at infinity to the ordinary affine plane. More generally, projective planes can be constructed over fields, over certain noncommutative algebraic systems, or directly from incidence axioms without assuming coordinates.

Projective planes are fundamental objects in projective geometry, where incidence and perspective replace distance and angle as the basic geometric relationships. They also connect several otherwise distinct areas of mathematics. The real projective plane is a compact nonorientable surface; the complex projective plane is a central space in algebraic geometry; finite projective planes give highly regular combinatorial designs. These interpretations share the same incidence pattern, but their additional algebraic, topological, and computational properties can differ substantially.

Incidence axioms and their consequences

An abstract projective plane consists of a set of points, a set of lines, and a relation called incidence. Saying that a point is incident with a line means that the point lies on that line. The usual axioms are:

  1. Any two distinct points lie on exactly one common line.
  2. Any two distinct lines meet in exactly one common point.
  3. There exist four points no three of which lie on a common line.

The third condition is a nondegeneracy requirement. Without it, structures such as a single line containing every point could satisfy much of the intended incidence behavior while lacking genuine two-dimensional geometry. A set of four points satisfying this condition is called a quadrangle. The axioms imply that every line contains at least three points and that every point lies on at least three lines. They do not require the sets of points and lines to be infinite.

The first two axioms are interchanged when the words “point” and “line” are exchanged. This symmetry gives the principle of duality: any theorem derived solely from the projective-plane axioms has a corresponding theorem obtained by exchanging points with lines and collinearity with concurrence. Three points are collinear when they lie on one line; three lines are concurrent when they pass through one point. The dual of a projective plane is therefore another projective plane, although it need not be isomorphic to the original plane.

These axioms specify incidence, not measurement. There is no distinguished length, angle, perpendicular direction, or unit circle. A configuration can be transformed in ways that drastically change its Euclidean appearance while preserving every assertion about which points lie on which lines. This makes a projective plane different from Euclidean geometry, whose structure includes metric relations. Euclidean geometry can nevertheless be recovered within a projective setting after additional data are chosen.

The incidence axioms also leave open whether familiar coordinate methods are available. Every plane constructed from a field satisfies them, but the converse is false: some projective planes cannot be represented by triples of field coordinates. Consequently, “projective plane” has a broader meaning than “two-dimensional projective space over a field.” The distinction is especially important in finite geometry and in the study of non-Desarguesian planes. Additional incidence theorems determine how much algebraic structure a projective plane admits.

Coordinates and the affine plane

Let KK be a field, and let K3K^3 be the three-dimensional vector space over it. The projective plane P2(K)\mathbb{P}^2(K), also written PG⁡(2,K)\operatorname{PG}(2,K), has as its points the one-dimensional linear subspaces of K3K^3. Its lines correspond to the two-dimensional linear subspaces. A point lies on a projective line precisely when the corresponding one-dimensional subspace is contained in the corresponding two-dimensional subspace.

A nonzero vector (x,y,z)(x,y,z) represents a projective point written [x:y:z][x:y:z]. These homogeneous coordinates are defined only up to a common nonzero factor:

[x:y:z]=[λx:λy:λz],λ∈K∖{0}.[x:y:z]=[\lambda x:\lambda y:\lambda z], \qquad \lambda\in K\setminus\{0\}.

The zero vector represents no point. A projective line has an equation

ax+by+cz=0,ax+by+cz=0,

where (a,b,c)≠(0,0,0)(a,b,c)\ne(0,0,0); multiplying all three coefficients by the same nonzero scalar leaves the line unchanged. Over a commutative field, the incidence axioms follow directly from the dimensions of spans and intersections in K3K^3. Two distinct one-dimensional subspaces span a unique two-dimensional subspace, while two distinct two-dimensional subspaces intersect in a unique one-dimensional subspace.

The points with z≠0z\ne0 can be normalized to [u:v:1][u:v:1]. They form an ordinary affine plane with coordinates (u,v)(u,v). The remaining points satisfy z=0z=0 and constitute the line at infinity. An affine line

au+bv+c=0au+bv+c=0

has projective completion ax+by+cz=0ax+by+cz=0, meeting the line at infinity at [b:−a:0][b:-a:0]. Affine lines with the same direction have proportional pairs (a,b)(a,b), so their projective completions meet at the same point at infinity. Nonparallel affine lines meet at an ordinary affine point.

There is nothing intrinsically special about the line z=0z=0. Any projective line can be selected as the line at infinity; deleting it leaves an affine plane. Conversely, an abstract affine plane can be completed projectively by adding one point for each parallel class and one new line containing all those added points. This construction does not depend on field coordinates.

The coordinate plane is covered by three affine charts, given by x≠0x\ne0, y≠0y\ne0, and z≠0z\ne0. Passing between them uses coordinate ratios rather than a single global Cartesian system. This local description later becomes important in topology and algebraic geometry: the projective plane can be studied using familiar affine coordinates, but no one such chart includes every point.

Projective transformations, duality, and invariants

An invertible 3×33\times3 matrix AA over a field KK induces a transformation of P2(K)\mathbb{P}^2(K) by

[v]⟼[Av].[v]\longmapsto[Av].

Because nonzero scalar multiples of AA induce the same transformation, these transformations form the projective general linear group

PGL⁡(3,K)=GL⁡(3,K)/K×.\operatorname{PGL}(3,K) = \operatorname{GL}(3,K)/K^\times.

Its elements are called projectivities or projective linear transformations. They preserve incidence because invertible linear maps carry subspaces to subspaces of the same dimension. They need not preserve distances, angles, areas, or affine parallelism.

Four ordered points with no three collinear form a projective frame. Any such frame can be carried to any other by a unique projectivity. One standard frame is

[1:0:0],[0:1:0],[0:0:1],[1:1:1].[1:0:0],\quad [0:1:0],\quad [0:0:1],\quad [1:1:1].

This explains why many apparently different diagrams represent the same projective configuration: their differences can be removed by a change of frame. Over fields with nontrivial automorphisms, applying a field automorphism to all coordinates also preserves incidence. More generally, semilinear transformations combine such an automorphism with an invertible linear map. The fundamental theorem of projective geometry identifies incidence-preserving bijections of the field-coordinate plane with transformations of this semilinear kind.

Although projectivities discard most metric quantities, they preserve the cross-ratio of four collinear points. For distinct finite coordinates a,b,c,da,b,c,d on a projective line, one convention is

(A,B;C,D)=(c−a)(d−b)(c−b)(d−a).(A,B;C,D) = \frac{(c-a)(d-b)}{(c-b)(d-a)}.

Equivalent homogeneous formulas include cases where one point is at infinity. The cross-ratio is unchanged by a projectivity of the line. It therefore provides a numerical invariant of configurations that cannot be captured by incidence alone. Ordered quadruples with cross-ratio −1-1 are called harmonic quadruples.

Coordinate duality follows from the dual vector space. A line is represented by a nonzero linear functional ℓ\ell, up to scalar multiplication, and incidence is the equation ℓ(v)=0\ell(v)=0. Thus the lines of P2(K)\mathbb{P}^2(K) form a dual projective plane. Under the point transformation v↦Avv\mapsto Av, line-coordinate columns transform by

ℓ⟼A−Tℓ.\ell\longmapsto A^{-T}\ell.

A chosen nondegenerate bilinear form can additionally identify points with lines, producing a correlation; when this correlation is involutory, it is called a polarity. Such a point-line correspondence is extra structure, not a canonical identification already contained in the definition of the plane.

Conics and classical incidence theorems

A projective conic over a field is defined by a nonzero homogeneous polynomial of degree two:

ax2+by2+cz2+dxy+exz+fyz=0.ax^2+by^2+cz^2+dxy+exz+fyz=0.

Homogeneity ensures that the equation is independent of the chosen representative of [x:y:z][x:y:z]. When the field has characteristic different from two, the equation can be written as vTQv=0v^TQv=0 for a symmetric matrix QQ. A nonsingular conic corresponds to an invertible QQ. Degenerate quadratic equations can describe two lines, a repeated line, or configurations whose rational points depend on the field.

Projective geometry unifies the Euclidean ellipse, parabola, and hyperbola. For a nonsingular real projective conic with real points, the affine appearance depends on how the selected line at infinity meets it. No real intersection gives an ellipse, one double intersection gives a parabola, and two distinct real intersections give a hyperbola. These are not three fundamentally different projective curves. A suitable projectivity can change which line is regarded as infinity and thereby change the affine classification. A real quadratic equation such as x2+y2+z2=0x^2+y^2+z^2=0, however, has no real projective points.

A nonsingular conic defines a polarity. At a point represented by pp on pTQp=0p^TQp=0, its tangent line has equation

pTQv=0.p^TQv=0.

For a point pp not on the conic, the same equation defines its polar line. The associated pole-polar reciprocity says that if a point lies on another point’s polar, then the reverse incidence also holds. This translates constructions involving tangents into constructions involving intersections and joins. In characteristic two, quadratic forms and conic polarities require separate treatment because their associated bilinear forms behave differently.

Several classical theorems express particularly strong incidence properties. Desargues’s theorem states that if two triangles are perspective from a point—the three lines joining corresponding vertices are concurrent—then the intersections of corresponding sides are collinear. It holds in every projective plane constructed over a division ring, but is not a consequence of the bare projective-plane axioms.

Pappus’s theorem concerns two triples of points on two lines and asserts the collinearity of three intersections formed by cross-joining them. It characterizes the commutative-field case among coordinatizable projective planes. Blaise Pascal’s theorem states that for six points on a nonsingular conic, the intersections of the three pairs of opposite sides of the resulting hexagon are collinear, under the usual nondegeneracy assumptions. Its dual, Brianchon’s theorem, concerns a hexagon circumscribed about a conic and the concurrence of its three opposite-vertex diagonals.

The real projective plane as a surface

The real projective plane RP2=P2(R)\mathbb{RP}^2=\mathbb{P}^2(\mathbb{R}) has a natural topology inherited from its coordinate construction. Every line through the origin in R3\mathbb{R}^3 meets the unit sphere S2S^2 in two antipodal points. Consequently,

RP2≅S2/(p∼−p).\mathbb{RP}^2\cong S^2/(p\sim -p).

The quotient map from the sphere is a two-sheeted covering. This description makes compactness immediate and shows that the real projective plane is a connected two-dimensional manifold without boundary. Its local neighborhoods are ordinary open disks, despite its globally unfamiliar shape.

Another model uses a closed disk with opposite boundary points identified. The interior represents a single affine chart, while the identified boundary represents the projective line at infinity. In this model, the boundary is not a boundary of the resulting surface: after identification, every point has a disk-shaped neighborhood. A projective line is itself homeomorphic to a circle, since RP1\mathbb{RP}^1 is a circle.

The surface is nonorientable. A loop representing its nontrivial fundamental-group element reverses local orientation when traversed. More precisely,

π1(RP2)≅Z/2Z.\pi_1(\mathbb{RP}^2)\cong\mathbb{Z}/2\mathbb{Z}.

A path on the sphere from pp to −p-p projects to such a loop; traversing it twice lifts to a closed loop. A cell decomposition has one cell in each dimension 0,1,20,1,2, with the two-dimensional cell attached by a map of degree two. Its Euler characteristic is therefore

χ(RP2)=1−1+1=1.\chi(\mathbb{RP}^2)=1-1+1=1.

Its integral homology groups are H0≅ZH_0\cong\mathbb{Z}, H1≅Z/2ZH_1\cong\mathbb{Z}/2\mathbb{Z}, and H2=0H_2=0.

The real projective plane cannot be embedded in ordinary three-dimensional Euclidean space without self-intersection. Familiar models such as the cross-cap and Boy’s surface are therefore not embedded copies; Boy’s surface is an immersion, while the standard cross-cap representation has singularities. An embedding exists in four-dimensional Euclidean space. Self-intersections in a three-dimensional picture belong to the representation, not to the intrinsic projective plane.

A metric can be added by taking the quotient of the round metric on the sphere. For unit representatives u,vu,v, the distance is

d([u],[v])=arccos⁡∣u⋅v∣.d([u],[v])=\arccos |u\cdot v|.

Great circles descend to closed geodesics, corresponding exactly to projective lines. This is the elliptic metric of constant positive curvature. It is preserved by the induced orthogonal transformations, not by all projectivities; the metric structure is more restrictive than the underlying projective incidence structure.

The complex projective plane and algebraic curves

The complex projective plane CP2=P2(C)\mathbb{CP}^2=\mathbb{P}^2(\mathbb{C}) is defined by the same homogeneous-coordinate construction, but its topology is very different from that of the real plane. Each affine chart is a copy of C2\mathbb{C}^2, so it has complex dimension two and real dimension four. It is compact, orientable, and simply connected. A complex projective line is CP1\mathbb{CP}^1, which is homeomorphic to a two-dimensional sphere.

A useful quotient description is

CP2=S5/S1.\mathbb{CP}^2=S^5/S^1.

Here a unit vector in C3\mathbb{C}^3 is identified with all its multiples by complex numbers of absolute value one. Unlike the real case, in which the sphere quotient identifies pairs, this quotient identifies whole circles. The resulting space carries a natural complex-manifold structure and the Fubini–Study metric. Its topology can be described by one cell in each of real dimensions 0,2,40,2,4, giving Euler characteristic three.

Its integral cohomology ring is

H∗(CP2;Z)≅Z[h]/(h3),deg⁡h=2.H^*(\mathbb{CP}^2;\mathbb{Z}) \cong \mathbb{Z}[h]/(h^3), \qquad \deg h=2.

The class hh is represented by a projective line, and the relation ∫CP2h2=1\int_{\mathbb{CP}^2}h^2=1 reflects the fact that two distinct projective lines intersect once. Thus a basic incidence statement becomes an intersection pairing in topology. Complex conjugation acts on CP2\mathbb{CP}^2, and its fixed-point set is a copy of RP2\mathbb{RP}^2.

In algebraic geometry, a plane projective curve is the zero set of a homogeneous polynomial F(x,y,z)F(x,y,z). An affine polynomial can be homogenized by inserting powers of a new coordinate so that every term has the same total degree. For example,

y=x2y=x^2

becomes yz=x2yz=x^2. Its projective closure adds the point [0:1:0][0:1:0], with the line z=0z=0 tangent there. This construction retains information at infinity that is absent from the original affine chart.

Bézout’s theorem states that two plane projective curves of degrees mm and nn, with no common component, intersect in mnmn points counted with intersection multiplicities over an algebraically closed field. Over the complex numbers, the projective setting accounts for all intersections, including those at infinity. A smooth degree-dd complex plane curve has genus

g=(d−1)(d−2)2.g=\frac{(d-1)(d-2)}{2}.

Thus smooth conics have genus zero, while smooth cubics have genus one. With a chosen point, a smooth plane cubic provides the classical projective model of an elliptic curve. Singularities can lower the genus of the associated nonsingular model.

Finite projective planes and existence problems

A finite projective plane has an integer order n≥2n\ge2, defined by requiring every line to contain n+1n+1 points. The axioms force every point to lie on n+1n+1 lines, and they force the total number of points and the total number of lines to be

v=n2+n+1.v=n^2+n+1.

To obtain the point count, fix a point PP. The n+1n+1 lines through PP each contain nn further points, and these groups do not overlap. Together with PP, they account for 1+n(n+1)1+n(n+1) points. Removing one line and all its points leaves an affine plane of order nn, containing n2n^2 points.

If qq is a prime power, a finite field Fq\mathbb{F}_q exists, and the coordinate construction produces PG⁡(2,q)\operatorname{PG}(2,q). Its number of points can also be obtained by counting nonzero vectors and dividing by the number of nonzero scalar representatives:

q3−1q−1=q2+q+1.\frac{q^3-1}{q-1}=q^2+q+1.

The smallest example, PG⁡(2,2)\operatorname{PG}(2,2), is the Fano plane. It has seven points and seven lines, with three points on each line. A conventional drawing uses the vertices and side midpoints of a triangle, together with its central point; one of the seven abstract lines is often drawn as a circle. Straightness in the drawing is irrelevant to incidence.

The Bruck–Ryser–Chowla theorem supplies an important necessary condition. If a projective plane has order n≡1n\equiv1 or 2(mod4)2\pmod4, then nn must be a sum of two integer squares. This rules out order six. It does not rule out order ten, because 10=12+3210=1^2+3^2. The nonexistence of a plane of order ten was established through extensive computer-assisted work by Clement Lam, Larry Thiel, and Stanisław Świercz, completed in 1989.

All known finite projective planes have prime-power order, but it remains unknown whether this is necessary. The existence of a plane of order twelve remains unresolved; it is the smallest unsettled order. Conversely, prime-power order does not guarantee uniqueness. At order nine there are non-Desarguesian planes as well as the field plane PG⁡(2,9)\operatorname{PG}(2,9).

Finite planes are also examples of symmetric combinatorial designs. Their point-line incidence matrices record a remarkable balance: every row and column has n+1n+1 entries equal to one, and any two different rows share exactly one such position. These rigid intersection counts connect the geometry to combinatorics, coding theory, and computational classification.

Desarguesian and non-Desarguesian planes

A projective plane is called Desarguesian if Desargues’s theorem holds for all appropriate configurations. This is not merely an isolated geometric property. It is equivalent to coordinatizability by a division ring, an algebraic system in which every nonzero element has a multiplicative inverse but multiplication need not be commutative. Points are then one-dimensional subspaces of a three-dimensional vector space over that division ring, with consistent left- or right-scalar conventions.

The distinction between commutative and noncommutative multiplication appears geometrically through Pappus’s theorem. A projective plane is Pappian precisely when it is coordinatizable by a commutative field. In particular, Pappus’s theorem implies Desargues’s theorem. A Desarguesian plane need not be Pappian when its division ring is noncommutative. In the finite case, however, every division ring is a field, so every finite Desarguesian plane is a plane over a finite field.

Desargues’s theorem holds automatically in projective spaces of dimension at least three, where spatial configurations provide its geometric proof. An abstract plane has no ambient third dimension available. It can satisfy the incidence axioms without admitting the additional constructions needed for that proof. Non-Desarguesian planes demonstrate that familiar properties of coordinate geometry cannot simply be assumed from the two-point and two-line axioms.

A classical real example is the Moulton plane, introduced by Forest Ray Moulton in 1902. It begins with the ordinary point set R2\mathbb{R}^2, but changes certain lines into bent lines. In one standard convention, a nonvertical line with positive slope parameter mm is given by

y={mx+b,x≤0,2mx+b,x≥0,y= \begin{cases} mx+b,&x\le0,\\ 2mx+b,&x\ge0, \end{cases}

while nonpositive-slope lines and vertical lines remain straight. This structure satisfies the affine-plane axioms. Its projective completion is non-Desarguesian. The failure is not merely visual bending: no incidence-preserving relabeling can transform it into the ordinary real coordinate plane.

General projective planes can be coordinatized by planar ternary rings. After a suitable coordinate frame is chosen, a ternary operation T(x,m,b)T(x,m,b) describes nonvertical affine lines by equations y=T(x,m,b)y=T(x,m,b). In a field plane this becomes T(x,m,b)=xm+bT(x,m,b)=xm+b, but in a general plane the associated operations need not satisfy field laws. The algebra depends on the chosen coordinates, and determining when different algebraic descriptions represent isomorphic planes is a substantial classification problem. This broader coordinatization places field planes within a much larger family of incidence geometries.

Historical development

Projective planes emerged from the study of perspective rather than from an initial abstract definition. Renaissance artists and mathematicians analyzed how a three-dimensional scene is represented on a flat image. Parallel lines in a scene can converge toward a vanishing point in the picture, suggesting that ordinary affine geometry omits points naturally required by projection. The eventual mathematical theory separated these incidence relationships from the physical process of vision.

Girard Desargues’s 1639 work on conics developed ideas that became central to projective geometry, including the systematic treatment of configurations involving points at infinity. Pascal’s 1640 work on conics supplied the hexagon theorem now bearing his name. These early results showed that properties of conics could be stated without reference to their Euclidean classification as ellipses, parabolas, or hyperbolas. Their significance was broader than individual constructions: projection preserved a coherent family of geometric statements.

Jean-Victor Poncelet’s Traité des propriétés projectives des figures, published in 1822, was a major step toward a systematic projective geometry. During the nineteenth century, August Ferdinand Möbius and Julius Plücker developed coordinate methods that helped establish the modern use of homogeneous coordinates. Karl Georg Christian von Staudt’s Geometrie der Lage of 1847 pursued a synthetic approach, reconstructing numerical relationships from incidence constructions rather than taking numerical coordinates as the starting point.

Felix Klein’s Erlangen program of 1872 situated projective geometry within a classification of geometries by their transformation groups. On this view, a geometry studies properties invariant under a specified group. Projective transformations preserve incidence, while Euclidean transformations preserve the more restrictive structure of distance. Klein’s framework also clarified how additional structures inside projective space could produce different metric geometries.

The axiomatic movement made the distinction between abstract incidence planes and coordinate planes explicit. David Hilbert’s Foundations of Geometry of 1899 examined the independence and interdependence of geometric axioms, including the role of Desargues’s theorem. Moulton’s example soon supplied a particularly concrete non-Desarguesian geometry. Oswald Veblen and John Wesley Young’s Projective Geometry, published in two volumes in 1910 and 1918, helped consolidate the relationship between axioms and algebraic coordinatization.

Twentieth-century research developed several directions in parallel: finite planes and their automorphism groups, topological projective planes, algebraic curves in projective space, and the differential topology of the real and complex planes. The computer-assisted exclusion of order ten became a prominent example of exhaustive computation resolving a finite-geometric existence problem.

Applications in imaging, combinatorics, and computation

In computer vision, an ideal pinhole camera associates an image point with a ray through the camera center. These rays form a projective plane. A three-dimensional scene point in homogeneous coordinates can be represented by a nonzero vector X∈R4X\in\mathbb{R}^4, and its image satisfies

x∼PX,x\sim PX,

where PP is a rank-three 3×43\times4 camera matrix and ∼\sim denotes equality up to nonzero scale. The camera center lies in the matrix’s null space and has no defined image. Homogeneous coordinates absorb perspective division into a linear matrix operation followed by normalization.

Two perspective images of a single scene plane are related by a nonsingular 3×33\times3 matrix, commonly called a homography, provided the camera centers do not lie on that plane. Four point correspondences in general position determine this transformation in the ideal exact setting. Such models underlie document rectification, image mosaicing, and planar image registration. With noisy data, fitting a homography becomes a problem in mathematical optimization, often requiring more than the minimum four correspondences.

Projective coordinates do not by themselves recover Euclidean dimensions. A general homography can change lengths, angles, and ratios along a line, although it preserves collinearity and cross-ratio. Metric reconstruction therefore requires additional information, such as camera calibration, known orthogonality, or known dimensions. This distinction separates what an image determines projectively from what can be inferred only after a metric structure has been supplied.

Finite planes provide balanced incidence systems for discrete applications. If MM is the point-line incidence matrix of a plane of order nn, with rows indexed by points, then

MMT=nI+J,MM^T=nI+J,

where II is the identity matrix and JJ is the all-ones matrix. The diagonal entries are n+1n+1, because each point lies on that many lines; the off-diagonal entries are one, because two points determine exactly one line. This identity gives algebraic access to the design’s symmetry and imposes restrictions on possible parameters. Taking the span of incidence vectors over a finite field produces linear codes whose dimensions and weight distributions encode geometric information.

There is also a direct connection with graph theory. The incidence graph has one vertex for each point and one for each line, with an edge joining an incident pair. For a plane of order nn, it is bipartite and (n+1)(n+1)-regular, has 2(n2+n+1)2(n^2+n+1) vertices, diameter three, and girth six. A four-cycle would imply that two points lie on two different common lines, contradicting the axioms. These tightly controlled distances and intersections make projective planes useful test objects in extremal combinatorics and in algorithms for highly regular graphs.