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

Projective Space

A projective space parametrizes lines through the origin of a vector space, providing a setting for incidence geometry, algebraic varieties, and geometric transformations.

47 keywords4 linked from10 not yet writtenWritten by AI
Vector spaceField (mathemati…Projective Geome…Affine spaceAlgebraic Geomet…Equivalence Rela…Homogeneous coor…Basis (linear al…Projective…

A projective space is a mathematical space whose points represent one-dimensional linear subspaces of a vector space. If VV has dimension n+1n+1 over a field KK, its projective space, denoted P(V)\mathbb P(V), has projective dimension nn. After choosing coordinates, it is written Pn(K)\mathbb P^n(K). A projective point is therefore a nonzero vector considered only up to multiplication by a nonzero scalar.

Projective spaces provide the fundamental setting for projective geometry. They extend affine spaces by incorporating points at infinity, so that parallel lines become intersecting lines without singling out a preferred direction. This makes incidence relations more uniform and allows geometric constructions to be expressed by homogeneous equations. Over the real and complex numbers, projective spaces are also important compact manifolds; over finite fields, they are finite incidence structures.

Their significance extends beyond classical geometry. Projective spaces are basic ambient spaces in algebraic geometry, state spaces in quantum theory, and coordinate models for perspective imaging. Their different roles arise from the same construction, but depend strongly on the underlying field and on whether the structure being studied is algebraic, topological, differential, or purely incidence-theoretic.

Definition and homogeneous coordinates

Let VV be a vector space over KK. On V∖{0}V\setminus\{0\}, define an equivalence relation by

v∼w⟺w=λvfor some λ∈K×,v\sim w \quad\Longleftrightarrow\quad w=\lambda v \quad\text{for some }\lambda\in K^\times,

where K×K^\times denotes the nonzero elements of KK. The projective space P(V)\mathbb P(V) is the set of equivalence classes under this relation. Each class consists of all nonzero vectors on a particular line through the origin. The zero vector must be excluded because it does not determine a unique one-dimensional subspace.

When V=Kn+1V=K^{n+1}, a point is written

[x0:x1:⋯:xn],[x_0:x_1:\cdots:x_n],

where the coordinates are not all zero and

[x0:⋯:xn]=[λx0:⋯:λxn](λ≠0).[x_0:\cdots:x_n]=[\lambda x_0:\cdots:\lambda x_n] \qquad(\lambda\ne0).

These are homogeneous coordinates. For example, [1:2:3][1:2:3], [2:4:6][2:4:6], and [−1:−2:−3][-1:-2:-3] represent the same point of P2(R)\mathbb P^2(\mathbb R). There is no generally meaningful individual value of xix_i; ratios such as xi/xjx_i/x_j, wherever the denominator is nonzero, are meaningful.

The projective dimension is one less than the vector-space dimension because common scaling removes one independent parameter. Thus P0(K)\mathbb P^0(K) is a single point, P1(K)\mathbb P^1(K) is a projective line, and P2(K)\mathbb P^2(K) is a projective plane. A projective line has many points: it is the collection of one-dimensional subspaces of a two-dimensional vector space, not an individual vector line through the origin. This distinction between a projective point and its representing vector subspace is essential in coordinate calculations.

Every choice of basis for VV gives homogeneous coordinates, but P(V)\mathbb P(V) itself does not require a basis. An invertible change of basis changes the coordinate representation without changing the underlying space. By contrast, arbitrary vector addition does not descend to projective points. The expression [v+w][v+w] generally depends on the chosen representatives vv and ww, since replacing vv by 2v2v may change the result. Projective space therefore is not naturally a vector space and has no distinguished origin.

For real and complex vector spaces, the quotient construction also gives a natural quotient topology. Over an arbitrary field, the same set-theoretic construction remains valid, while additional geometric structures must be specified separately. In particular, an algebraic projective space and the set of its points over a field are closely related but are not identical objects in modern algebraic geometry.

Affine charts and points at infinity

Projective space is covered by n+1n+1 standard coordinate charts

Ui={[x0:⋯:xn]:xi≠0}.U_i=\{[x_0:\cdots:x_n]:x_i\ne0\}.

On UiU_i, scaling makes the ii-th coordinate equal to 11, and the remaining coordinates give a bijection with KnK^n:

[x0:⋯:xn]⟼(x0xi,…,xixi^,…,xnxi).[x_0:\cdots:x_n]\longmapsto \left(\frac{x_0}{x_i},\ldots,\widehat{\frac{x_i}{x_i}},\ldots, \frac{x_n}{x_i}\right).

The hat indicates that the ii-th entry is omitted. Over R\mathbb R or C\mathbb C, these are smooth or holomorphic coordinate charts. In algebraic geometry, each is an affine open subset.

Choosing U0U_0 identifies an affine space with the subset

(a1,…,an)⟼[1:a1:⋯:an].(a_1,\ldots,a_n)\longmapsto[1:a_1:\cdots:a_n].

Its complement,

H∞={x0=0},H_\infty=\{x_0=0\},

is a projective hyperplane isomorphic to Pn−1(K)\mathbb P^{n-1}(K). This is the hyperplane at infinity for the chosen affine chart. It is not intrinsically distinguished: another choice of chart makes a different hyperplane the hyperplane at infinity. Thus “finite” and “infinite” describe positions relative to an affine presentation, not two invariant kinds of projective point.

An affine line with parametrization a+tva+tv, where v≠0v\ne0, has projective closure

[s:sa+tv],[s:t]∈P1(K).[s:sa+tv],\qquad [s:t]\in\mathbb P^1(K).

At s=0s=0, this line contains the point [0:v][0:v]. Two affine lines with proportional direction vectors therefore acquire the same point at infinity. In the projective plane, this turns the distinction between parallel and nonparallel lines into a distinction caused only by the selected affine chart: any two distinct projective lines meet in exactly one point. In dimensions three and higher, however, skew lines still need not meet.

The projective line has a particularly simple description:

P1(K)=K∪{∞}.\mathbb P^1(K)=K\cup\{\infty\}.

The affine coordinate z=x1/x0z=x_1/x_0 covers every point except [0:1][0:1], and that remaining point is denoted ∞\infty. In higher dimensions, adding just one point is insufficient: an entire projective space of directions must be added. Consequently, Pn(R)\mathbb P^n(\mathbb R) is not the one-point compactification of Rn\mathbb R^n when n>1n>1.

The same construction explains homogenization of equations. If a polynomial f(x1,…,xn)f(x_1,\ldots,x_n) has total degree dd, its homogenization is

F(x0,…,xn)=x0df(x1x0,…,xnx0).F(x_0,\ldots,x_n)=x_0^d f\left(\frac{x_1}{x_0},\ldots,\frac{x_n}{x_0}\right).

The result is a homogeneous polynomial of degree dd. For example, the affine parabola y=x2y=x^2 becomes

X2=YZX^2=YZ

in coordinates [X:Y:Z][X:Y:Z], with affine chart Z=1Z=1. Its unique point at infinity is [0:1:0][0:1:0]. For varieties defined by several equations, computing the projective closure requires homogenizing the defining ideal appropriately; homogenizing an arbitrary list of generators can introduce unwanted components at infinity.

Linear subspaces, incidence, and duality

A vector subspace W⊆VW\subseteq V determines a projective linear subspace P(W)⊆P(V)\mathbb P(W)\subseteq\mathbb P(V). If dim⁡W=r+1\dim W=r+1, then P(W)\mathbb P(W) is a projective rr-plane. Projective points, lines, and hyperplanes thus correspond respectively to vector subspaces of dimensions 11, 22, and dim⁡V−1\dim V-1. This correspondence translates incidence questions into linear algebra: containment of projective subspaces is exactly containment of their representing vector subspaces.

The projective span of points [v1],…,[vk][v_1],\ldots,[v_k] is the projectivization of their linear span. Two distinct points determine a unique projective line because their representatives span a two-dimensional vector space. More generally, points represented by linearly independent vectors determine a projective subspace of dimension one less than the number of points. A projective frame in Pn(K)\mathbb P^n(K) consists of n+2n+2 points such that every n+1n+1 of them have linearly independent representatives. Frames serve as reference configurations for projective coordinates.

For vector subspaces A,B⊆VA,B\subseteq V,

dim⁡(A∩B)=dim⁡A+dim⁡B−dim⁡(A+B).\dim(A\cap B)=\dim A+\dim B-\dim(A+B).

If projective subspaces of dimensions rr and ss lie in Pn(K)\mathbb P^n(K), this implies

dim⁡(P(A)∩P(B))≥r+s−n,\dim\bigl(\mathbb P(A)\cap\mathbb P(B)\bigr)\ge r+s-n,

using the convention that the empty intersection has dimension −1-1. Whenever r+s≥nr+s\ge n, the intersection is nonempty. Two lines in a projective plane must therefore meet; two lines in projective three-space may be disjoint.

A hyperplane is the zero set of a nonzero linear functional ℓ:V→K\ell:V\to K. Multiplying ℓ\ell by a nonzero scalar leaves its kernel unchanged, so the hyperplanes of P(V)\mathbb P(V) are parametrized by P(V∗)\mathbb P(V^*), where V∗V^* is the dual space. This is the dual projective space. For a subspace WW, its annihilator

W∘={ℓ∈V∗:ℓ(w)=0 for all w∈W}W^\circ=\{\ell\in V^*:\ell(w)=0\text{ for all }w\in W\}

reverses containment. A projective rr-plane in Pn\mathbb P^n corresponds under this duality to a projective (n−r−1)(n-r-1)-plane.

In the projective plane, duality exchanges points with lines and collinearity with concurrence. The statement that two points determine a line is dual to the statement that two lines determine a point. This symmetry is a structural feature of incidence geometry, not a rule about distances or angles. A particular identification of points with hyperplanes requires extra information, such as a nondegenerate bilinear form. Such a choice produces a polarity; it is not supplied by projective space alone.

Projective transformations and geometric invariants

An invertible linear map A:V→VA:V\to V induces a projective transformation

[v]⟼[Av].[v]\longmapsto[Av].

Two invertible maps differing by a nonzero scalar induce the same transformation. Their projective transformation group is therefore

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

In coordinates, its elements are represented by nonsingular (n+1)×(n+1)(n+1)\times(n+1) matrices, considered up to a common scalar multiple. These transformations preserve projective linear subspaces, incidence, and projective dimension.

In an affine chart, a projective transformation generally has fractional linear form:

x⟼Bx+bcTx+d,x\longmapsto\frac{Bx+b}{c^{\mathsf T}x+d},

where division by the denominator applies to all components. The expression fails to be finite where cTx+d=0c^{\mathsf T}x+d=0, but the transformation remains defined there as a map of projective space. An affine transformation is the special case that preserves the chosen hyperplane at infinity. Projective transformations need not preserve parallelism in the selected chart, nor do they generally preserve length, angle, area, or ratios of affine lengths.

On a projective line, the basic numerical invariant of four ordered distinct points is the cross-ratio. In an affine coordinate, 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)}.

Expressions involving infinity are interpreted by cancellation or limits. The same invariant can be written with determinants of homogeneous coordinate vectors, avoiding exceptional affine cases. Different orderings of the four points can produce six values related by transformations such as r↦1/rr\mapsto1/r and r↦1−rr\mapsto1-r, so the ordering convention matters.

Every ordered triple of distinct points on a projective line can be carried to any other such triple by a unique projective transformation. In particular, three points can be normalized to 00, 11, and ∞\infty; the position of a fourth is then recorded by a cross-ratio. In dimension nn, a projective transformation is likewise uniquely determined by the images of the n+2n+2 points of a projective frame. These facts explain why projective geometry has invariants of configurations but no preferred coordinates for individual points.

The fundamental theorem of projective geometry describes a broader class of incidence-preserving maps. For projective spaces of dimension at least two over fields, suitable bijections preserving lines arise from invertible semilinear maps. A semilinear map is additive but may transform scalars through a field automorphism:

T(av)=σ(a)T(v).T(av)=\sigma(a)T(v).

This distinction matters over fields with nontrivial automorphisms. Projective linear transformations correspond to σ=id\sigma=\mathrm{id}; general collineations may also include field automorphisms. The projective line requires separate treatment because preserving its sole full-dimensional line imposes no useful incidence restriction.

Real projective space and its topology

For K=RK=\mathbb R, every nonzero vector can be normalized to unit length. A real line through the origin meets the unit sphere Sn⊂Rn+1S^n\subset\mathbb R^{n+1} in exactly two antipodal points. Consequently,

RPn≅Sn/(x∼−x).\mathbb{RP}^n\cong S^n/(x\sim -x).

This identifies real projective space as a compact smooth manifold of real dimension nn. The antipodal action is free, so the quotient has no boundary or singular points. Compactness follows from compactness of the sphere.

The first cases have distinctive forms. The space RP0\mathbb{RP}^0 is a point, and RP1\mathbb{RP}^1 is homeomorphic to a circle. The real projective plane RP2\mathbb{RP}^2 can be represented by a closed disk whose opposite boundary points are identified. It is a compact nonorientable surface. It cannot be embedded in three-dimensional Euclidean space, although familiar pictures represent immersions with self-intersections. Such intersections belong to the representation, not to the abstract projective plane.

For n≥2n\ge2, the quotient map Sn→RPnS^n\to\mathbb{RP}^n is a two-sheeted universal covering, giving

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

The case n=1n=1 differs: its fundamental group is Z\mathbb Z, because RP1\mathbb{RP}^1 is a circle and S1S^1 is not simply connected. Real projective space is orientable precisely when nn is odd, apart from the zero-dimensional case. This follows from the antipodal map on SnS^n, whose degree is (−1)n+1(-1)^{n+1}: it preserves orientation exactly for odd nn.

A useful cell decomposition has one cell in every dimension from 00 through nn. With integer coefficients, the cellular boundary map is multiplication by 22 in positive even dimensions and zero in odd dimensions. Hence the intermediate homology groups are Z/2Z\mathbb Z/2\mathbb Z in positive odd degrees and zero in positive even degrees. The top homology is Z\mathbb Z when nn is odd and zero when n>0n>0 is even. The Euler characteristic is therefore 11 for even nn and 00 for odd nn.

Using coefficients in the two-element field produces a particularly concise cohomology ring:

H∗(RPn;F2)≅F2[a]/(an+1),deg⁡a=1.H^*(\mathbb{RP}^n;\mathbb F_2) \cong\mathbb F_2[a]/(a^{n+1}),\qquad \deg a=1.

The generator aa is the first Stiefel–Whitney class of the tautological real line bundle. This connects projective topology with the study of how vector spaces vary continuously over a base space. Infinite-dimensional real projective space, formed by including each RPn\mathbb{RP}^n into the next, classifies real line bundles over spaces of the usual homotopy-theoretic kind, such as CW complexes.

Complex projective space and its differential geometry

Complex projective space CPn=Pn(C)\mathbb{CP}^n=\mathbb P^n(\mathbb C) has complex dimension nn and real dimension 2n2n. Normalizing vectors gives

CPn≅S2n+1/S1,\mathbb{CP}^n\cong S^{2n+1}/S^1,

because the unit vectors representing one complex line differ by multiplication by a complex number of absolute value 11. Unlike the real case, the set being collapsed over each point is a circle rather than a pair of points. The quotient is a compact complex manifold, with holomorphic coordinate changes furnished by the standard affine charts.

The case CP1\mathbb{CP}^1 is the Riemann sphere. Its affine chart is the complex plane, and adding one point gives a space homeomorphic to S2S^2. For n=1n=1, the map S3→CP1≅S2S^3\to\mathbb{CP}^1\cong S^2 is the Hopf fibration. Its circular fibers exhibit a relationship between the topology of spheres and complex lines that has no analogue in the naive picture of a sphere as a product of a circle and another sphere.

A Hermitian inner product on Cn+1\mathbb C^{n+1} induces the Fubini–Study metric on projective space. In one common normalization, the distance between points represented by nonzero vectors zz and ww is

d([z],[w])=arccos⁡(∣⟨z,w⟩∣∥z∥ ∥w∥).d([z],[w])= \arccos\left(\frac{|\langle z,w\rangle|} {\|z\|\,\|w\|}\right).

The absolute value makes the expression independent of the phases of the representatives, and the denominator removes their magnitudes. Other conventions multiply the metric by a constant. For the displayed normalization, CP1\mathbb{CP}^1 is a round sphere of radius 1/21/2, and the holomorphic sectional curvature is 44.

This metric is Kähler: its complex structure, Riemannian metric, and associated closed two-form are compatible. On an affine chart with coordinates z1,…,znz_1,\ldots,z_n, its Kähler form is proportional to

i ∂∂ˉlog⁡(1+∑j=1n∣zj∣2),\mathrm{i}\,\partial\bar\partial \log\left(1+\sum_{j=1}^n|z_j|^2\right),

with the constant determined by normalization. The full projective linear group does not preserve this metric. The projective unitary group does, reflecting the difference between a projective structure and the additional geometry selected by an inner product.

Topologically, CPn\mathbb{CP}^n has one cell in every even dimension 0,2,…,2n0,2,\ldots,2n, and no odd-dimensional cells. Its integral cohomology ring is

H∗(CPn;Z)≅Z[h]/(hn+1),deg⁡h=2.H^*(\mathbb{CP}^n;\mathbb Z) \cong\mathbb Z[h]/(h^{n+1}),\qquad \deg h=2.

Here hh is the hyperplane class, and ∫CPnhn=1\int_{\mathbb{CP}^n}h^n=1. In particular, the Euler characteristic is n+1n+1, and CPn\mathbb{CP}^n is simply connected for n≥1n\ge1. These topological properties make complex projective space a basic testing ground for differential geometry and the topology of complex manifolds.

Projective algebraic geometry

A projective algebraic set is defined by homogeneous polynomial equations

Fα(x0,…,xn)=0F_\alpha(x_0,\ldots,x_n)=0

inside projective space. Homogeneity ensures that the equations are independent of the chosen representative, since a degree-dd polynomial satisfies

F(λx)=λdF(x).F(\lambda x)=\lambda^dF(x).

The coordinates themselves do not define global scalar-valued functions on projective space. Homogeneous expressions instead naturally describe sections of line bundles, while ratios of homogeneous polynomials of equal degree define rational functions wherever the denominator does not vanish.

The modern construction uses a graded polynomial ring:

PKn=Proj⁡K[x0,…,xn].\mathbb P^n_K=\operatorname{Proj}K[x_0,\ldots,x_n].

As a scheme, this object contains more information than the set Pn(K)\mathbb P^n(K): it includes non-rational points and local algebraic data. For example, a projective variety defined over R\mathbb R may have complex points even if it has no real points. Statements about intersections therefore require attention to the field and to whether points are being counted geometrically, with multiplicity, or only among KK-rational points.

Projective space supplies a natural setting for intersection theory. Bézout’s theorem states, in its classical plane form, that two projective plane curves of degrees dd and ee, over an algebraically closed field and with no common component, have dede intersection points counted with intersection multiplicity. Two conics thus have four intersections with multiplicity, even when some are non-real or occur at infinity. The qualifications are indispensable: tangencies contribute multiplicity, and curves sharing a component do not have a finite intersection to which this elementary count applies.

The tautological line bundle O(−1)\mathcal O(-1) on P(V)\mathbb P(V) has fiber equal to the line represented by each point. Its dual is the hyperplane bundle O(1)\mathcal O(1), and its tensor powers are denoted O(d)\mathcal O(d). For d≥0d\ge0, global sections of O(d)\mathcal O(d) correspond to degree-dd homogeneous polynomials, forming a vector space of dimension

(n+dd).\binom{n+d}{d}.

Their monomials define the degree-dd Veronese embedding into projective space of dimension (n+dd)−1\binom{n+d}{d}-1. This construction turns degree-dd equations into linear hyperplane equations in a larger ambient space.

Projectivity also provides an algebraic counterpart of compactness. Projective space over a field is proper, a condition that controls behavior under families and base change rather than merely referring to an ordinary topology. Images of closed sets under projective morphisms are closed. A regular function on an integral projective variety over an algebraically closed field is constant, despite the existence of many nonconstant rational functions. These properties explain why passing from affine varieties to projective closures is central to studying limits, intersections, and algebraic families.

Finite fields and axiomatic projective spaces

If the underlying field is the finite field Fq\mathbb F_q, then projective space is a finite set with a highly regular incidence structure. There are qn+1−1q^{n+1}-1 nonzero vectors in Fqn+1\mathbb F_q^{n+1}, and every one-dimensional subspace contains q−1q-1 nonzero vectors. Therefore

∣Pn(Fq)∣=qn+1−1q−1=1+q+⋯+qn.|\mathbb P^n(\mathbb F_q)| =\frac{q^{n+1}-1}{q-1} =1+q+\cdots+q^n.

Each projective line contains q+1q+1 points. In a projective plane, there are q2+q+1q^2+q+1 points and the same number of lines; every point lies on q+1q+1 lines.

The smallest example is P2(F2)\mathbb P^2(\mathbb F_2), the Fano plane. It has seven points and seven lines, with three points on each line and three lines through each point. Since the only nonzero scalar in F2\mathbb F_2 is 11, its projective points can be represented uniquely by the seven nonzero vectors of F23\mathbb F_2^3. The line through distinct vectors uu and vv consists of the three points represented by uu, vv, and u+vu+v. Common drawings use a circle for one of the lines; the drawing records incidence rather than Euclidean straightness.

The number of projective rr-planes in Pn(Fq)\mathbb P^n(\mathbb F_q) is a Gaussian binomial coefficient:

[n+1r+1]q=∏i=0rqn+1−i−1qr+1−i−1.{n+1\brack r+1}_q = \prod_{i=0}^{r} \frac{q^{n+1-i}-1}{q^{r+1-i}-1}.

This formula counts (r+1)(r+1)-dimensional vector subspaces, with the denominator correcting for the number of ordered bases of each subspace. Its symmetry under r↦n−r−1r\mapsto n-r-1 reflects point–hyperplane duality. Such counts make finite projective spaces important examples in combinatorics and in the construction of incidence designs.

Projective geometry can also be defined axiomatically, starting with points and lines rather than coordinates. A projective plane satisfies the requirements that any two distinct points lie on exactly one line, any two distinct lines meet in exactly one point, and a nondegenerate quadrangle exists. Not every plane satisfying these axioms comes from a vector space over a field. Non-Desarguesian projective planes exist, so “projective plane” in incidence geometry is a broader notion than P2(K)\mathbb P^2(K).

Desargues’s theorem is central to the distinction. It relates two triangles whose corresponding vertices are joined by concurrent lines to the collinearity of intersections of corresponding sides. Planes coordinatized by division rings satisfy it, and Desarguesian planes can be coordinatized by division rings. In projective dimensions at least three, the standard incidence axioms force the corresponding coordinatization phenomenon. For finite division rings, Wedderburn’s theorem implies commutativity, so finite Desarguesian projective planes are obtained from finite fields.

Historical development

Projective space emerged from the study of perspective and incidence rather than from an initial abstract definition. Renaissance investigations of painting and architecture examined how three-dimensional scenes could be represented on a plane. A perspective image depends on lines through a viewpoint, anticipating the idea that a geometric point can represent a direction or a line of sight. These practical studies did not yet provide the homogeneous-coordinate construction, but they established the importance of properties preserved by projection.

Gérard Desargues’s 1639 work on conic sections developed a systematic approach to projection and points at infinity. His triangle theorem became a foundational result of incidence geometry. In 1640, Blaise Pascal formulated the theorem now bearing his name: for six points on a conic, the intersections of three pairs of opposite sides lie on a line, with the usual treatment of degenerate configurations. Such results shifted attention from lengths and angles toward relationships preserved by projection.

During the early nineteenth century, Jean-Victor Poncelet made projective methods into a sustained geometric program. His Traité des propriétés projectives des figures, published in 1822, emphasized properties invariant under projection and the role of ideal and imaginary elements. Joseph Gergonne and Poncelet developed the principle of duality, which connected statements about points with corresponding statements about lines or planes. These developments helped replace the exceptional treatment of parallel lines with a uniform incidence language.

August Ferdinand Möbius introduced barycentric coordinates in 1827, while Julius Plücker developed homogeneous analytic methods in works beginning in 1828. Coordinate descriptions made projective constructions accessible to algebra and helped clarify how affine geometry sits inside projective geometry. Karl von Staudt’s Geometrie der Lage of 1847 pursued a synthetic approach, reconstructing arithmetic and projective invariants from incidence constructions rather than assuming a metric background.

In 1872, Felix Klein placed projective geometry within his Erlangen program, classifying geometries by transformation groups and their invariants. In this framework, affine and Euclidean geometries arise by preserving additional structures inside projective geometry. During the twentieth century, projective spaces became standard objects in algebraic geometry, topology, and differential geometry. The development of sheaf and scheme methods gave precise algebraic formulations of projective families, while the study of characteristic classes and fiber bundles made real and complex projective spaces fundamental examples in topology.

Perspective imaging, quantum states, and parameter spaces

In perspective imaging, a spatial point and a camera center determine a line of sight. The image records that line rather than the distance of the point along it, making projective coordinates a natural model. An ideal pinhole camera can be represented by a rank-three matrix

P:R4⟶R3,[X]⟼[PX].P:\mathbb R^4\longrightarrow\mathbb R^3, \qquad [X]\longmapsto[PX].

This defines a map from P3(R)\mathbb P^3(\mathbb R) to P2(R)\mathbb P^2(\mathbb R) except at the projective point corresponding to ker⁡P\ker P, the camera center. All spatial points on a line through that center produce the same image point.

A plane not containing the camera center maps projectively to the image plane. Correspondences between two images of the same plane can therefore be described by a 3×33\times3 nonsingular matrix up to scale, usually called a planar homography. It has eight independent parameters. Four point correspondences in general position determine such a transformation in the ideal exact setting. In actual images, estimation must account for measurement error, lens distortion, and degeneracies. Projective geometry supplies the geometric model; calibration introduces additional metric information about the camera.

In quantum mechanics, a pure state of a complex Hilbert space is represented by a ray rather than by a particular nonzero vector. Multiplying a state vector by a nonzero complex scalar does not change its normalized physical predictions. For a finite-dimensional Hilbert space of dimension n+1n+1, the pure-state space is therefore CPn\mathbb{CP}^n. For normalized representatives ψ\psi and ϕ\phi, the transition probability is

∣⟨ψ,ϕ⟩∣2,|\langle\psi,\phi\rangle|^2,

which is independent of their phases and is directly related to their Fubini–Study distance.

For a two-dimensional complex Hilbert space, the pure-state space is CP1\mathbb{CP}^1, represented by the Bloch sphere. This does not mean that every quantum state lies on that sphere: mixed states occupy its interior when represented by Bloch vectors. Likewise, a nonzero vector in a tensor product need not be a product vector. For two finite-dimensional systems, the product pure states form a Segre-embedded product of projective spaces, while other points of the composite projective state space represent entangled pure states.

Projective spaces also parametrize equations and geometric objects whose coefficients are meaningful only up to scale. A plane conic has an equation

aX2+bXY+cY2+dXZ+eYZ+fZ2=0,aX^2+bXY+cY^2+dXZ+eYZ+fZ^2=0,

so nonzero coefficient lists define points of P5\mathbb P^5. Singular conics occupy an algebraic subset determined by the vanishing of the determinant of the associated symmetric matrix, over fields of characteristic different from 22. More generally, the set of kk-dimensional subspaces of a vector space forms a Grassmannian, embedded into a projective space by the Plücker embedding. Its homogeneous coordinates satisfy polynomial relations, illustrating how projective space serves not only as a space of geometric points but also as an ambient space for families of subspaces and equations.