In mathematics, projective space \mathbb{P}^n(F) over a field F is defined as the set of one-dimensional subspaces (lines through the origin) of the vector space F^{n+1}, or equivalently, the quotient space of F^{n+1} \setminus \{0\} by the equivalence relation of scalar multiplication by nonzero elements of F.[1] This construction generalizes affine space by incorporating points at infinity, allowing parallel lines to intersect and eliminating the distinction between Euclidean and affine transformations in favor of projective ones.[2]The concept of projective space emerged from early efforts in perspective drawing during the Italian Renaissance, where architect Filippo Brunelleschi developed linear perspective around 1425 to represent three-dimensional scenes on two-dimensional surfaces, leading to the observation that parallel lines converge at a vanishing point.[3] In the 17th century, Gérard Desargues formalized these ideas in his 1639 work Brouillon projet d'une atteinte aux événements des rencontres du cônne avec un plan, introducing projective methods for conic sections, while Blaise Pascal contributed theorems on hexagons inscribed in conics.[4] The 19th century saw projective geometry mature as an independent field, with key developments by Jean-Victor Poncelet, who axiomatized it in 1822, and August Ferdinand Möbius, emphasizing invariants like the cross-ratio under projections.[4]In algebraic geometry, projective space serves as a foundational ambient space for studying varieties, constructed as the Proj of a polynomial ring in n+1 variables, enabling the use of homogeneous coordinates to compactify affine varieties and handle phenomena at infinity.[5] Points in \mathbb{P}^n are represented by homogeneous coordinates [x_0 : x_1 : \dots : x_n], where scaling does not change the point, and it admits a natural scheme structure that supports morphisms and embeddings of algebraic sets.[5] This framework is essential for theorems like Bézout's on intersections and for classifying curves and surfaces up to projective equivalence.[6]Beyond pure mathematics, projective space finds applications in computer vision, where it models perspective projections from three-dimensional scenes onto image planes, facilitating tasks like camera calibration, homography estimation, and structure-from-motion reconstruction.[2] In this context, the projective plane \mathbb{P}^2 represents image coordinates, allowing invariantproperties to handle distortions and multiple viewpoints.[7] It also appears in robotics for pose estimation and in computer graphics for rendering transformations, bridging geometric theory with practical computation.[8]
Motivation
Geometric Intuition
In the projective plane, denoted \mathbb{RP}^2, parallel lines from Euclidean geometry are understood to intersect at points at infinity, forming a line at infinity that completes the space and resolves the exception where parallels do not meet. This intuition arises from considering the plane as the set of directions from a viewpoint, where all lines in a parallel class converge to a single ideal point on the horizon, eliminating the need to treat parallelism as a special case. For visualization, imagine a diagram of \mathbb{RP}^2 as a disk with opposite boundary points identified: finite points lie inside, while the boundary represents the line at infinity, and lines appear as great circles crossing the boundary at their infinite endpoints, showing how Euclidean parallels meet there.[9][10][2]This geometric idea has historical roots in Renaissance art, where artists like Filippo Brunelleschi and Masaccio employed perspective techniques to create illusions of depth, converging parallel lines—such as architectural edges or receding floors—to vanishing points on the horizon line. In works like Masaccio's The Holy Trinity (c. 1427), these vanishing points simulate the meeting of parallels at infinity, prefiguring the mathematical formalization of projective geometry by figures like Girard Desargues in the 17th century, who recognized such projections as invariant properties of space. This artistic innovation thus provided an intuitive bridge to projective concepts, treating the canvas as a projective plane where visual convergence mirrors geometric completion.[11][12]A further intuition comes from conic sections, where slicing a cone with planes parallel to its side yields a parabola tangent to the line at infinity, while hyperbolas cross it at two points and ellipses avoid it entirely; in projective space, these distinctions dissolve as the space completes the affine plane by adding the infinite line, allowing all non-degenerate conics to be projectively equivalent without degeneration exceptions at infinity. Projective space thus serves as a natural compactification of affine space, incorporating these infinite behaviors seamlessly.[2]Overall, projective space unifies points, lines, and planes by treating infinite elements as ordinary points, ensuring that any two lines intersect at exactly one point and any two planes intersect in a line, without special rules for directions at infinity.[10][2]
Algebraic Perspective
In algebraic geometry, projective space arises naturally as a quotient construction from linear algebra. Specifically, the real projective space \mathbb{RP}^n is the quotient of the vector space \mathbb{R}^{n+1} minus the origin by the equivalence relation of scalar multiplication by nonzero reals, where two nonzero vectors v, w \in \mathbb{R}^{n+1} are identified if w = \lambda v for some \lambda \in \mathbb{R} \setminus \{0\}.[13] Each equivalence class represents a one-dimensional subspace, or line through the origin, of \mathbb{R}^{n+1}.[14] For instance, \mathbb{RP}^2 is obtained as \mathbb{R}^3 \setminus \{0\} / \sim, parameterizing lines through the origin in three-dimensional Euclidean space.[13]This algebraic structure provides significant benefits for studying polynomialequations by enabling homogenization. An affine polynomial f(x_1, \dots, x_n) of degree d is homogenized to a projective polynomial F(X_0, \dots, X_n) by introducing a new variable X_0 and multiplying each term of degree less than d by appropriate powers of X_0 to make all terms homogeneous of degree d.[15] For example, the affine equation x^2 + y^2 = 1 homogenizes to X^2 + Y^2 - Z^2 = 0 in \mathbb{P}^2, where the original affine plane corresponds to the patch Z \neq 0 via dehomogenization x = X/Z, y = Y/Z.[15] This process defines the projective closure of the affine variety, incorporating points at infinity and ensuring the equation is invariant under scaling.[15]Projective space thus resolves coordinate singularities inherent in affine descriptions by covering the space with multiple affine patches without global inconsistencies.[13] In the circle example, the affine unit circle is compact but lacks a uniform treatment at "infinity" in affine coordinates; its projective closure as the conic X^2 + Y^2 - Z^2 = 0 embeds it smoothly into \mathbb{P}^2, adding ideal points at infinity [1 : \pm i : 0] over the complexes and providing a closed, proper variety free from affine boundary artifacts.[15]
Definitions
Homogeneous Coordinates
The n-dimensional projective space over a field K, denoted \mathbb{P}^n(K), is defined as the set of equivalence classes of (n+1)-tuples (x_0, x_1, \dots, x_n) \in K^{n+1} \setminus \{0\}, where two tuples (x_0, \dots, x_n) and (y_0, \dots, y_n) are equivalent if there exists a nonzero \lambda \in K such that y_i = \lambda x_i for all i = 0, \dots, n.[2][16] This construction identifies points that differ by scalar multiplication, capturing lines through the origin in the vector space K^{n+1}.[2]Points in \mathbb{P}^n(K) are typically denoted by square brackets [x_0 : x_1 : \dots : x_n], with the understanding that [x_0 : \dots : x_n] = [\lambda x_0 : \dots : \lambda x_n] for any \lambda \in K \setminus \{0\}.[16] To recover affine coordinates, standard affine charts cover \mathbb{P}^n(K): for each i = 0, \dots, n, define the open set U_i = \{ [x_0 : \dots : x_n] \mid x_i \neq 0 \}, which is isomorphic to the affine space K^n via dehomogenization, mapping [x_0 : \dots : x_n] to (x_0 / x_i, \dots, \hat{x}_i / x_i, \dots, x_n / x_i), where the hat indicates omission of the i-th term.[2] These charts provide a coordinate atlas, allowing local affine descriptions of projective geometry.[16]A concrete example is the real projective line \mathbb{RP}^1, which consists of points [x : y] with (x, y) \in \mathbb{R}^2 \setminus \{(0,0)\} up to scaling by nonzero \lambda \in \mathbb{R}.[2] Topologically, \mathbb{RP}^1 is homeomorphic to the circle S^1, obtained by identifying antipodal points on the unit circle \{ (x,y) \in \mathbb{R}^2 \mid x^2 + y^2 = 1 \}, such as [0:1] representing the origin and [1:0] representing the point at infinity.[2][17]Homogeneous coordinates facilitate the embedding of affine n-space K^n into \mathbb{P}^n(K) as the chart U_n = \{ [x_0 : \dots : x_{n-1} : 1] \mid x_0, \dots, x_{n-1} \in K \}, with the complementary hyperplane at infinity \{ [x_0 : \dots : x_{n-1} : 0] \mid (x_0, \dots, x_{n-1}) \neq (0, \dots, 0) \} consisting of directions of parallel lines in the affine space.[16][2] This structure resolves issues like parallel lines intersecting at infinity, providing a unified framework for affine and projective geometries.[16]
Quotient Construction
The projective space associated to a vector space V over a field K, denoted P(V), is formally defined as the quotient set (V \setminus \{0\}) / K^\times, where K^\times = K \setminus \{0\} acts on the nonzero vectors by scalar multiplication.[2] Two nonzero vectors v, w \in V are equivalent, written v \sim w, if there exists a nonzero scalar \lambda \in K^\times such that w = \lambda v; each equivalence class $$ thus corresponds to a one-dimensional subspace (or ray) through the origin in V.[2]This quotient construction satisfies a universal property: any function f: V \setminus \{0\} \to X to another set X that is constant on equivalence classes (i.e., f(\lambda v) = f(v) for all \lambda \in K^\times) factors uniquely through the projection map \pi: V \setminus \{0\} \to P(V), yielding a well-defined map \overline{f}: P(V) \to X such that f = \overline{f} \circ \pi.[18]When V is an (n+1)-dimensional vector space over K, the resulting projective space P^n(K) = P(V) has dimension n, since each point in P^n(K) represents a one-dimensional subspace of V.[2]A concrete realization of this construction uses homogeneous coordinates, where points are represented by equivalence classes of (n+1)-tuples in K^{n+1} \setminus \{0\}.[2] For example, over the complex numbers, the one-dimensional projective space \mathbb{CP}^1 = P(\mathbb{C}^2) is isomorphic to the Riemann sphere \mathbb{C} \cup \{\infty\}, with the isomorphism given by stereographic projection from the unit sphere in \mathbb{R}^3.[19]
Basic Structures
Subspaces and Dimensions
In projective geometry, a subspace of a projective space \mathbb{P}(V), where V is a vector space over a field K of dimension n+1, is defined as the image under the quotient map \pi: V \setminus \{0\} \to \mathbb{P}(V) of a linear subspace W \subseteq V with \dim W \geq 1.[2] When \dim W = 1, \mathbb{P}(W) is a point, the 0-dimensional subspace. Such a projective subspace, denoted \mathbb{P}(W), consists of all lines through the origin in W.[16]The dimension of a projective subspace \mathbb{P}(W) is given by \dim \mathbb{P}(W) = \dim W - 1.[20] Equivalently, a k-dimensional projective subspace corresponds to a (k+1)-dimensional vector subspace of V.[21] This relation establishes the linear algebraic foundation for the geometry of projective spaces, linking their subspaces directly to those of the underlying vector space.For instance, in the real projective space \mathbb{RP}^3, which arises from a 4-dimensional vector space, a projective line is \mathbb{P}(L) where L \subseteq \mathbb{R}^4 is a 2-dimensional subspace, and a projective plane is \mathbb{P}(M) where M \subseteq \mathbb{R}^4 is a 3-dimensional subspace.[2]The entire projective space \mathbb{P}^n has dimension n, while its points, which are the 1-dimensional subspaces of V, form the 0-dimensional subspaces.[16]
Lines and Incidence
In projective space \mathbb{P}^n(K) over a field K, a projective line is defined as a one-dimensional projective subspace, corresponding to a two-dimensional vector subspace of the underlying vector space V^{n+1}.[22] Such a line can be parametrized by any two distinct points on it, which generate the line as their span.[22]Incidence between a point and a line in \mathbb{P}^n(K) is determined by the corresponding vector subspaces: a point P, represented by a one-dimensional subspace U \subseteq V, lies on a line L, represented by a two-dimensional subspace W \subseteq V, if and only if U \cap W \neq \{0\}.[22] This relation satisfies the axioms of projective geometry, ensuring that any two distinct points determine a unique line. In higher dimensions, two lines intersect at a point if and only if they are contained in a common plane.The join of a set of points is the smallest projective subspace containing them, obtained as the projectivization of the span of their representing vectors.[22] Dually, the meet of a set of subspaces is their intersection, provided it is nonempty and of the appropriate dimension; for two lines, if they intersect, their meet is the unique point of intersection.[22]In the real projective plane \mathbb{RP}^2, any two distinct points determine a unique line, and there are no parallel lines—all lines intersect at exactly one point, reflecting the absence of parallelism in projective geometry.[21] This incidence structure implies theorems such as Desargues' theorem, which concerns the collinearity of intersection points of corresponding sides of two perspective triangles.[23]
Topology
Topological Properties
The real projective space \mathbb{RP}^n is equipped with the quotient topology induced from the projection map p: \mathbb{R}^{n+1} \setminus \{0\} \to \mathbb{RP}^n, where points are identified under scalar multiplication by nonzero reals, or equivalently from the double covering S^n \to \mathbb{RP}^n identifying antipodal points.[24] This topology renders \mathbb{RP}^n compact, as it is the continuous image of the compact sphere S^n, and Hausdorff, since the equivalence relation is closed and the projection is open.[25][24]As a topological space, \mathbb{RP}^n admits the structure of a smooth n-dimensional manifold, with an atlas derived from affine charts corresponding to hyperplanes not containing the origin in \mathbb{R}^{n+1}.[24] Regarding orientability, \mathbb{RP}^n is orientable if and only if n is odd, since the antipodal map on S^n preserves orientation precisely when n+1 is even; for even n, it is non-orientable.[26] For low dimensions, \mathbb{RP}^1 is homeomorphic to the circle S^1, which is a compact orientable 1-manifold, while \mathbb{RP}^2 can be realized as the disk D^2 with antipodal points on the boundary \partial D^2 identified, yielding a compact non-orientable surface diffeomorphic to the real projective plane.[24][25]Key homotopy and homology invariants further characterize \mathbb{RP}^n. The fundamental group is \pi_1(\mathbb{RP}^n) \cong \mathbb{Z}/2\mathbb{Z} for n \geq 2, arising from the two-sheeted covering S^n \to \mathbb{RP}^n and the action of the antipodal map.[24] With \mathbb{Z}/2\mathbb{Z}-coefficients, the homology groups are H_k(\mathbb{RP}^n; \mathbb{Z}/2\mathbb{Z}) \cong \mathbb{Z}/2\mathbb{Z} for $0 \leq k \leq n and zero otherwise, computed via cellular homology of the CW structure with one cell per dimension up to n.[24]The complex projective space \mathbb{CP}^n inherits its topology from the quotient map \mathbb{C}^{n+1} \setminus \{0\} \to \mathbb{CP}^n under multiplication by nonzero complex scalars, or equivalently from the Hopf fibration S^{2n+1} \to \mathbb{CP}^n with S^1-fibers.[27][24] This endows \mathbb{CP}^n with the structure of a compact Hausdorff space and a smooth $2n-dimensional real manifold (or n-dimensional complex manifold), which is always orientable due to its compatible complex structure.[27][24] Unlike \mathbb{RP}^n, \mathbb{CP}^n is simply connected, with trivial fundamental group.[24]
CW Complex Decomposition
The real projective space \mathbb{RP}^n admits a CW complex structure consisting of one open cell e^k in each dimension k from 0 to n.[24] This construction arises from viewing \mathbb{RP}^n as the quotient of the n-sphere S^n by the antipodal map, which identifies each point x with -x.[24] Equivalently, \mathbb{RP}^n can be built inductively by taking the (k-1)-skeleton to be \mathbb{RP}^{k-1} and attaching a k-cell via the quotient map q: S^{k-1} \to \mathbb{RP}^{k-1}, which is the projection induced by the antipodal identification and serves as a degree-2 covering map.[24][28]In this attaching process, the boundary of the k-disk D^k is mapped to \mathbb{RP}^{k-1} by collapsing antipodal points on S^{k-1}, ensuring the CW structure aligns with the topological quotient.[24] The resulting cellular chain complex has \mathbb{Z} in each degree from 0 to n, with boundary maps d_k: C_k(\mathbb{RP}^n) \to C_{k-1}(\mathbb{RP}^n) given by multiplication by 2 if k is even and by 0 if k is odd.[24][28]For the example of \mathbb{RP}^2, the CW structure includes one 0-cell e^0 (a point), one 1-cell e^1 (forming a circle after attachment), and one 2-cell e^2 attached via the degree-2 map S^1 \to \mathbb{RP}^1 \cong S^1.[24] The cellular chain complex is then $0 \to \mathbb{Z} \xrightarrow{d_2 = \cdot 2} \mathbb{Z} \xrightarrow{d_1 = 0} \mathbb{Z} \to 0, which yields the homology groups H_0(\mathbb{RP}^2; \mathbb{Z}) \cong \mathbb{Z}, H_1(\mathbb{RP}^2; \mathbb{Z}) \cong \mathbb{Z}/2\mathbb{Z}, and H_2(\mathbb{RP}^2; \mathbb{Z}) = 0.[28] This decomposition facilitates algebraic topology computations, such as those involving homology via cellular chains.[24]
Algebraic Aspects
Over Finite Fields
Projective spaces over finite fields \mathbb{F}_q, where q is a prime power, provide discrete analogs of classical projective geometry with finite point sets and well-defined combinatorial structures. The n-dimensional projective space \mathbb{P}^n(\mathbb{F}_q), often denoted PG(n, q), consists of the 1-dimensional subspaces (lines through the origin) of the (n+1)-dimensional vector space \mathbb{F}_q^{n+1}. Each point in this space corresponds to an equivalence class of nonzero vectors under scalar multiplication by nonzero elements of \mathbb{F}_q. The total number of points is given by the formula \frac{q^{n+1} - 1}{q - 1}, which counts the distinct directions in the vector space.[29]A key combinatorial feature is the enumeration of subspaces. The number of k-dimensional projective subspaces in PG(n, q) equals the Gaussian binomial coefficient \dbinom{n+1}{k+1}_q, which generalizes the ordinary binomial coefficient and arises from counting the (k+1)-dimensional vector subspaces of \mathbb{F}_q^{n+1}. This coefficient is defined as\dbinom{m}{r}_q = \prod_{i=0}^{r-1} \frac{q^{m-i} - 1}{q^{r-i} - 1},for m = n+1 and r = k+1, and it satisfies recursive properties analogous to those of binomial coefficients. These counts highlight the symmetric and balanced nature of the geometry, where subspaces intersect in controlled ways.[30]For the specific case of the projective plane PG(2, q), the formula simplifies to q^2 + q + 1 points, with each line also containing exactly q + 1 points and the same number of lines in total. Every pair of points determines a unique line, and every pair of lines intersects in a unique point, embodying the fundamental incidence axioms. This structure exemplifies the order q of the plane, where lines have q + 1 points.[31]Finite projective spaces exhibit rich combinatorial properties, including their interpretation as partial geometries. In particular, the point-line incidence structure of PG(2, q) forms a partial geometry pg(q, q, 1), where every two non-collinear points have exactly one common neighbor, satisfying the defining axioms of regularity and intersection control. Higher-dimensional spaces extend these properties, serving as frameworks for partial linear spaces with uniform line sizes. These geometries find applications in coding theory, notably in the construction of projective Reed-Muller codes, which evaluate polynomials on the points of PG(n, q) to yield error-correcting codes with parameters tied to the space's dimension and field size. For instance, these codes generalize classical Reed-Muller codes and achieve minimum distances determined by the geometry's intersection properties.[32][33]
Projective Modules
In the context of commutative algebra, the notion of projective space can be generalized beyond vector spaces over fields to the projectivization of projective modules over a commutative ring R. Specifically, for a projective R-module P of constant rank n+1, the projectivization \mathbb{P}(P) is the scheme \operatorname{Proj}_R(\operatorname{Sym}(P^\vee)), where \operatorname{Sym}(P^\vee) is the symmetric algebra on the dual module P^\vee, equipped with a structure morphism to \operatorname{Spec} R.[34] This construction parameterizes the rank-1 locally free quotients of P (or dually, line submodules), assuming P is locally free of constant rank, which holds when R is such that projective modules of constant rank are locally free, as over domains or local rings. When P is free, \mathbb{P}(P) recovers the classical projective space \mathbb{P}^n_R.[35]A key property enabling this geometric interpretation is the projectivity of P, which implies that P is locally free over R. That is, there exists a covering of \operatorname{Spec} R by basic open sets D(f_i) such that the localization P_{f_i} is a free R_{f_i}-module of rank n+1. This local freeness ensures that \mathbb{P}(P) behaves geometrically like a projective space locally on \operatorname{Spec} R, bridging the algebraic structure of modules with geometric intuition.Over polynomial rings, such as R = k[x_1, \dots, x_m] for a field k, projective modules play a central role in relating this construction to homogeneous ideals in graded rings. By the Quillen-Suslin theorem, every finitely generated projective module over a polynomial ring is free, so \mathbb{P}(P) aligns with the standard projective space; more broadly, it connects to the homogeneous spectrum of the symmetric algebra on P, providing a foundation for algebraic geometry over such rings.This module-theoretic projective space also relates to vector bundles: the classical projective space \mathbb{P}^n_R can be viewed as the projectivization of the trivial vector bundle of rank n+1 over \operatorname{Spec} R, where points correspond to line subbundles (rank 1 projective submodules). In general, for a projective module P, \mathbb{P}(P) parameterizes the line subbundles of the associated vector bundle, emphasizing the correspondence between projective modules and locally free sheaves.[36]
Synthetic Geometry
Axiomatic Foundations
Projective geometry can be developed synthetically through axiomatic systems that emphasize incidence relations between points and lines, eschewing coordinates to define spaces purely in terms of these primitives. Hilbert's axiomatic approach to Euclidean geometry influenced subsequent foundations for projective geometry, where the parallel postulate (Group III) is replaced by the axiom that any two lines in a plane intersect at exactly one point. This is achieved by incorporating points at infinity, ensuring all lines meet. Hilbert's incidence axioms (Group I) include: two distinct points determine a unique line (Axiom I, 1), any three non-collinear points determine a unique plane (Axiom I, 3), and every line contains at least two points while every plane contains at least three non-collinear points (Axiom I, 7). The order axioms (Group II) define betweenness on lines, the congruence axioms (Group IV) handle segment and angle equivalences, and continuity axioms (Group V) ensure completeness via Dedekind cuts.[37]A more direct synthetic framework for projective spaces appears in the work of Oswald Veblen and John Wesley Young, who define a projective space as an incidence structure satisfying specific axioms on points and lines. Their incidence axioms include: any two distinct points lie on exactly one line (Axiom A1), and the Veblen axiom (A2 or V): given a triangle ABC, if a line intersects sides AB and AC at points not A, B, or C, then it intersects BC (Axiom A2). Extension assumptions guarantee non-degeneracy: every line has at least three points (E0), there exists at least one line (E1), not all points are collinear (E2), and not all points are coplanar (E3). For higher dimensions, additional axioms require that any two planes intersect in a unique line, and that the space has dimension three or higher, with points existing outside any given plane. These axioms distinguish projective spaces from affine ones by eliminating parallels.In such axiomatic projective planes, classical theorems emerge as consequences of the incidence structure. For instance, Pappus' theorem states that if two lines are each intersected by three parallel lines (or in projective terms, if a hexagon is inscribed in two lines with alternate vertices on each), then the intersection points of opposite sides are collinear; this holds in any Desarguesian projective plane satisfying the basic incidence axioms (P1–P4) and Desargues' axiom (P7), where P1 asserts any two points determine a unique line, P2 any two lines intersect in a unique point, P3 ensures non-collinearity of all points, and P4 guarantees a line through a point not on a given line.[38]A pivotal result in these systems is the coordinatization theorem, which demonstrates that the axioms independently imply the existence of a coordinate structure without presupposing one. Specifically, for projective spaces of dimension at least three satisfying the Veblen-Young axioms, the space is isomorphic to the projective space over a vector space defined by a division ring, allowing homogeneous coordinates to be constructed algebraically from the incidence relations alone; this independence underscores the synthetic approach's power in deriving analytic models from pure incidence.[39]
Finite Projective Spaces
Finite projective spaces are incidence structures satisfying the axioms of projective geometry, restricted to finite point and line sets. For dimensions greater than or equal to 3, the Veblen-Young theorem establishes that all such spaces are Desarguesian, meaning they are isomorphic to the projective geometry PG(n, q) constructed from a vector space over a finite field \mathbb{F}_q, where q is a prime power and n ≥ 3. This classification follows from the fact that any projective space satisfying Desargues' axiom in dimension at least 3 can be coordinatized by a division ring, and by Wedderburn's little theorem, all finite division rings are commutative fields. Thus, no non-Desarguesian examples exist in these higher dimensions.In dimension 2, finite projective planes exhibit greater variety. Desarguesian planes of order q (a prime power) are precisely the PG(2, q) over \mathbb{F}_q, but non-Desarguesian planes also exist, constructed via alternative coordinatizations such as ternary rings that fail to yield division rings. Notable examples include the Hughes planes of order p^{2k} for odd prime p and integer k ≥ 1, which are obtained by replacing the field multiplication in the Desarguesian plane with a non-associative operation derived from a near-field. These planes satisfy the projective plane axioms but violate Desargues' theorem, distinguishing them from their field-based counterparts.The smallest finite projective plane is the Fano plane, denoted PG(2, 2), which has order 2 and consists of 7 points and 7 lines, each line containing 3 points. It serves as the foundational example of a Desarguesian plane over the field with 2 elements and illustrates the basic incidence structure of projective geometry. All known finite projective planes have order q, where q is a prime power, and it is conjectured that this holds for all such planes, though this remains an open problem.Existence conditions for finite projective planes are constrained by the Bruck-Ryser-Chowla theorem, which provides a necessary criterion: if a projective plane of order n exists and n ≡ 1 or 2 (mod 4), then n must be expressible as the sum of two integer squares. This theorem rules out planes of certain orders, such as n=6, and has been instrumental in computational searches for non-existent planes, like the unresolved case of order 12.
Transformations
Projective Linear Groups
The projective linear group associated to a vector space V over a field K, denoted \mathrm{PGL}(V), is defined as the quotient group \mathrm{GL}(V)/Z, where \mathrm{GL}(V) is the general linear group of invertible linear endomorphisms of V and Z is the center consisting of scalar multiples of the identity map.[2] This construction identifies linear transformations that differ by a scalar multiple, reflecting the projective nature of the space. The group \mathrm{PGL}(V) acts faithfully on the projective space \mathrm{P}(V) by sending equivalence classes $$ (lines through the origin spanned by x \in V \setminus \{0\}) to [f(x)], where f \in \mathrm{GL}(V).[2]In matrix terms, if V = K^{n+1}, elements of \mathrm{PGL}(n+1, K) can be represented by invertible (n+1) \times (n+1) matrices modulo scalar multiples, acting via[A]() = [A x],where denotes the projective point corresponding to the column vector $x \neq 0$, and the action preserves collinearity: if and $$ lie on a projective line (i.e., x, y are linearly dependent), then so do [A x] and [A y].[2] This preservation extends to higher-dimensional subspaces, mapping projective subspaces to projective subspaces of the same dimension.[2]A concrete example arises in the real projective plane \mathbb{RP}^2, where \mathrm{PGL}(3, \mathbb{R}) is the group of all projective transformations, including perspectivities—central projections from a point outside a line or plane—and more general maps that generate harmonic divisions on lines, such as the inversion of two points in a harmonic set of four collinear points.[2] These transformations maintain the cross-ratio and thus preserve harmonic properties, which are fundamental invariants in projective geometry.[2]The fundamental theorem of projective geometry characterizes these groups as the full automorphism groups of projective spaces in many cases: any bijective map between projective spaces of dimension at least 2 over a division ring that preserves incidence of points and lines (i.e., a collineation) is induced by a semilinear transformation on the underlying vector space.[40] Over fields, this implies that automorphisms of \mathrm{P}^n(K) (for n \geq 2) arise from elements of the projective semilinear group \mathrm{P\Gamma L}(n+1, K), with \mathrm{PGL}(n+1, K) as the linear subgroup when the field automorphism is trivial.[40] This result, tracing back to foundational work by von Staudt and later formalized by Veblen and Young, underscores the rigid structure imposed by projective incidence.
Morphisms Between Spaces
In algebraic geometry, a morphism between projective spaces f: \mathbb{P}^m \to \mathbb{P}^n over an algebraically closed field k is defined by n+1 homogeneous polynomials F_0, \dots, F_n \in k[x_0, \dots, x_m] of the same degree d > 0 such that F_0, \dots, F_n have no common zeros except the origin in k^{m+1}, ensuring the map is well-defined and regular everywhere.[41] This construction respects the projective equivalence, mapping [x_0 : \dots : x_m] to [F_0(x) : \dots : F_n(x)], and extends the notion of rational maps by avoiding indeterminacy loci.[42]A special case arises when d=1, where the polynomials are linear forms, inducing f via a linear map \phi: k^{m+1} \to k^{n+1} between the ambient vector spaces, such that f() = [\phi(v)]. If \phi is invertible, f is an isomorphism; more generally, such maps induce embeddings of linear subspaces if \phi is injective.[43] Collineations, which are these degree-1 isomorphisms between projective spaces of the same dimension, are precisely those induced by semilinear isomorphisms of the underlying vector spaces, as established by the fundamental theorem of projective geometry.[44]The Veronese embedding provides a canonical example of a non-linear morphism, mapping \mathbb{P}^n into \mathbb{P}^N with N = \binom{n+d}{d} - 1 via all monomials of degree d in the coordinates: \nu_d([x_0 : \dots : x_n]) = [ \dots : x_0^{i_0} \cdots x_n^{i_n} : \dots ], where the indices satisfy i_0 + \dots + i_n = d. This embedding is an isomorphism onto its image, a projective variety known as the Veronese variety, and highlights how higher-degree morphisms embed lower-dimensional spaces into higher ones while preserving projective structure.[45]An important birational example is the projection from a point O \in \mathbb{P}^n (not on a target hyperplane H \cong \mathbb{P}^{n-1}) to H, defined rationally by sending a point P to the intersection of the line OP with H; this map is birational, with an inverse given by lines through O from points on H, and is undefined only at O.[46] Such projections illustrate how rational maps between projective spaces can be resolved to morphisms via blow-ups, maintaining birational equivalence.Since projective spaces are complete varieties over k, any morphism f: \mathbb{P}^m \to \mathbb{P}^n is proper, meaning it is separated, of finite type, and universally closed, a property inherited from the properness of the structure morphism to \operatorname{Spec} k.[47] This ensures that images of closed sets remain closed, facilitating compactness-like behavior in algebraic settings.[48]
Dualities and Generalizations
Dual Projective Space
In projective geometry, the dual projective space (\mathbb{P}^n)^*, also denoted \mathbb{P}^n_*, of a projective space \mathbb{P}^n over a field K is constructed such that its points correspond to the hyperplanes of \mathbb{P}^n, and its k-dimensional subspaces correspond to the sets of hyperplanes in \mathbb{P}^n that contain a fixed subspace of codimension k+1.[2] This duality arises from the vector space perspective: if \mathbb{P}^n = \mathbb{P}(V) for a vector space V of dimension n+1, then (\mathbb{P}^n)^* = \mathbb{P}(V^*), where V^* is the dual vector space of linear functionals on V, and each point \in \mathbb{P}(V^*) represents the hyperplane \ker f \subset \mathbb{P}(V).[2] Lines in the dual space consist of pencils of hyperplanes, namely the hyperplanes containing a fixed codimension-2 subspace of \mathbb{P}^n.[2]The duality principle reverses incidence relations: a point p \in \mathbb{P}^n lies on a hyperplane H \subset \mathbb{P}^n if and only if the hyperplane corresponding to p in the dual contains the point corresponding to H in (\mathbb{P}^n)^*.[2] This reversal enables the dualization of geometric theorems by interchanging points with hyperplanes and incidence with containment, preserving the logical structure.[2] For instance, Desargues' theorem, which asserts that two triangles in perspective from a point are in perspective from a line, has as its dual the converse statement that two triangles in perspective from a line are in perspective from a point.[2][49]In the real projective plane \mathbb{RP}^2, the dual space interchanges points and lines, transforming configurations accordingly.[2] A notable example is conic duality: a conic section, viewed as a curve of points in \mathbb{RP}^2, dualizes to the envelope of its tangent lines, which forms another conic in the dual plane.[50]Over a field K, the projective space \mathbb{P}^n(K) is naturally isomorphic to its dual (\mathbb{P}^n(K))^* via a nondegenerate bilinear form on the underlying vector space V, which induces an isomorphism V \to V^* by v \mapsto \langle \cdot, v \rangle, preserving the projective structure.[2] This isomorphism identifies points and hyperplanes symmetrically through the pairing.[2]
Abstract Generalizations
In infinite-dimensional settings, the notion of projective space extends beyond finite-dimensional vector spaces to spaces like Hilbert and Banach spaces, where it plays a crucial role in functional analysis and quantum mechanics. For a complex Hilbert space H, the projective space \mathbb{P}(H) consists of one-dimensional subspaces (rays) of H, equipped with a natural Kähler structure induced by the inner product on H. This structure arises from the quotient of the unit sphere in H by the action of the unit circle, providing a manifold model for the space of pure quantum states, where each ray corresponds to an equivalence class of state vectors differing by a phase factor.[51] In quantum mechanics, \mathbb{P}(H) serves as the state space, with observables represented as functions on this space and dynamics governed by a symplectic form derived from the Fubini-Study metric, enabling a classical Hamiltonian formulation of quantum evolution.[52]A further generalization appears in the context of Banach spaces, where the projective space \mathbb{P}(V) for an infinite-dimensional complex Banach space V is defined similarly as the set of one-dimensional subspaces, metrized via an analog of the Fubini-Study metric adapted to the norm on V. This metric ensures \mathbb{P}(V) is paracompact and metrizable, facilitating the study of infinitesimal neighborhoods and submanifold structures in infinite-dimensional geometry.[53] Such constructions are essential in functional analysis for examining properties like injectivity and extensions of operators, though the absence of an inner product complicates the Kähler aspects compared to the Hilbert case.In categorical terms, projective spaces generalize to projective objects within abelian categories, which abstract the notion of subspaces that "project" onto quotients via lifts of homomorphisms. An object P in an abelian category \mathcal{A} is projective if, for every epimorphism A \twoheadrightarrow B and morphism P \to B, there exists a lift P \to A making the diagram commute; this property ensures that \operatorname{Hom}(P, -) preserves epimorphisms.[54] Projective objects thus generalize direct summands of free objects, providing a framework for resolutions and derived categories that extends the subspace structure of classical projective spaces to arbitrary abelian settings, such as modules over rings or sheaves on schemes.[55]The Grassmannian \operatorname{Gr}(k, V), parametrizing k-dimensional subspaces of a vector space V, serves as a higher-dimensional analog of projective space, embedding as a projective variety via the Plücker embedding into \mathbb{P}(\bigwedge^k V). For finite k and \dim V = n, \operatorname{Gr}(k, n) generalizes the lines in \mathbb{P}^{n-1} (where k=1) to k-planes, inheriting a rich geometry including Schubert cycles and cohomology rings that mirror aspects of projective space but in higher rank.[56] This structure is foundational in algebraic geometry and representation theory, where Grassmannians classify partial flags and support generalizations of projective duality to multi-dimensional subspaces.