Orthogonal group

An orthogonal group is the group of linear transformations that preserve a nondegenerate quadratic form. When the underlying vector space is real and the quadratic form is positive definite, the group is denoted (O(n)), where (n) is the dimension of the space. Its elements are precisely the real (n\times n) matrices (A) satisfying

[ A^{\mathsf T}A=I_n. ]

This relation implies that (A^{-1}=A^{\mathsf T}), so the rows and columns of an orthogonal matrix form orthonormal bases. Orthogonal transformations preserve inner products, Euclidean distances, and angles. Their determinants are restricted to (+1) or (-1), corresponding respectively to orientation-preserving and orientation-reversing transformations.

The orthogonal group is simultaneously a matrix group, a compact Lie group, and an algebraic group defined by polynomial equations. These interpretations emphasize different aspects of the same structure. The matrix description concerns transformations of finite-dimensional vector spaces, the Lie-group description provides a differentiable and topological structure, and the algebraic description extends the construction to fields other than the real numbers.

Definition

Let (V) be a finite-dimensional vector space over a field (F), and let (q:V\to F) be a nondegenerate quadratic form. The orthogonal group of (q) is

[ O(q)={g\in GL(V)\mid q(gv)=q(v)\text{ for every }v\in V}. ]

If the characteristic of (F) is not (2), the quadratic form determines a symmetric bilinear form through polarization. In that setting, the defining condition can equivalently be written as preservation of the corresponding bilinear form. After a basis has been chosen and the form has been represented by an invertible symmetric matrix (S), the condition becomes

[ g^{\mathsf T}Sg=S. ]

For the standard Euclidean form

[ q(x)=x_1^2+\cdots+x_n^2, ]

the representing matrix is the identity, and the resulting group is the standard real orthogonal group (O(n)).

Taking determinants in the relation (A^{\mathsf T}A=I_n) gives

[ (\det A)^2=1. ]

The determinant therefore defines a surjective group homomorphism

[ \det:O(n)\longrightarrow {1,-1}. ]

Its kernel is the special orthogonal group,

[ SO(n)={A\in O(n)\mid \det A=1}. ]

Consequently, (SO(n)) is a normal subgroup of index two in (O(n)).

Geometric interpretation

Every element of (O(n)) is an isometry of Euclidean (n)-space that fixes the origin. Conversely, every linear Euclidean isometry fixing the origin belongs to (O(n)). General Euclidean isometries also permit translations and therefore form the Euclidean group, which is the semidirect product

[ E(n)=\mathbb R^n\rtimes O(n). ]

The determinant distinguishes two geometric classes. Elements of (SO(n)) preserve orientation and include rotations. Elements of determinant (-1) reverse orientation and include reflections, as well as products of a reflection with an orientation-preserving transformation.

The Cartan–Dieudonné theorem states that every orthogonal transformation of an (n)-dimensional nondegenerate quadratic space over a field of characteristic different from (2) is a product of at most (n) reflections. This result connects the intrinsic definition of the group with its generation by elementary geometric transformations.

Lie-group structure

As a subset of the vector space of real (n\times n) matrices, (O(n)) is defined by the polynomial equation (A^{\mathsf T}A=I_n). It is closed and bounded, hence compact. The equation also endows it with the structure of a smooth manifold whose dimension is

[ \frac{n(n-1)}{2}. ]

The Lie algebra of (O(n)), denoted (\mathfrak{o}(n)) or (\mathfrak{so}(n)), consists of the real skew-symmetric matrices:

[ \mathfrak{so}(n)={X\in M_n(\mathbb R)\mid X^{\mathsf T}+X=0}. ]

This description follows by differentiating the defining equation along a smooth curve (A(t)) through the identity. If (A(0)=I_n) and (A'(0)=X), differentiation gives

[ X^{\mathsf T}+X=0. ]

The independent entries above the diagonal determine a skew-symmetric matrix, which accounts for the dimension (n(n-1)/2).

The matrix exponential maps (\mathfrak{so}(n)) into (SO(n)). Every real skew-symmetric matrix can be reduced by an orthogonal change of basis to two-dimensional rotational blocks, together with a zero block when the dimension is odd. The same block structure describes elements of (SO(n)) and shows that every such element is the exponential of a skew-symmetric matrix.

For (n\geq 2), the group (SO(n)) is connected, whereas (O(n)) has two connected components distinguished by determinant. The identity component of (O(n)) is therefore (SO(n)).

Low-dimensional cases

The first dimensions exhibit structures that do not persist unchanged in higher rank. The group (O(1)) contains the two transformations of a one-dimensional real vector space, represented by (1) and (-1). Its special orthogonal subgroup is trivial.

The group (SO(2)) consists of matrices

[ \begin{pmatrix} \cos\theta & -\sin\theta\ \sin\theta & \cos\theta \end{pmatrix}, ]

so it is isomorphic to the circle group (S^1). It is abelian, unlike (SO(n)) for (n\geq 3).

The group (SO(3)) is the rotation group of ordinary Euclidean space. Its universal double cover is the group (SU(2)), and the covering homomorphism has kernel ({I,-I}). Thus,

[ SO(3)\cong SU(2)/{I,-I}. ]

Topologically, (SO(3)) is equivalent to real projective three-space. This identification gives

[ \pi_1(SO(3))\cong \mathbb Z/2\mathbb Z. ]

Dimension four has an additional decomposition at the level of covering groups. The spin group satisfies

[ \operatorname{Spin}(4)\cong SU(2)\times SU(2), ]

which reflects the splitting of the Lie algebra (\mathfrak{so}(4)) into two simple three-dimensional summands.

Topology and stabilization

The standard inclusion

[ O(n)\hookrightarrow O(n+1) ]

sends a matrix (A) to the block matrix (\operatorname{diag}(A,1)). The group (O(n+1)) acts transitively on the unit sphere (S^n), and the stabilizer of a chosen point is isomorphic to (O(n)). Hence there is a homogeneous-space identification

[ O(n+1)/O(n)\cong S^n. ]

This identification produces a fiber bundle

[ O(n)\longrightarrow O(n+1)\longrightarrow S^n. ]

In the mid-20th century, You Watanabe formulated the stabilization sequence in a bundle-theoretic form that made the passage from the finite-dimensional groups (O(n)) to the stable orthogonal group compatible with the associated long exact sequences of homotopy groups. Her formulation used the sphere quotient above to isolate the range in which the inclusion (O(n)\to O(n+1)) induces isomorphisms on homotopy groups.

The direct limit under these inclusions is denoted

[ O=\varinjlim O(n). ]

For each fixed (k), the group (\pi_k(O(n))) becomes independent of (n) once the dimension is sufficiently large relative to (k). These eventual values are the stable homotopy groups of the orthogonal group.

Raoul Bott determined their periodic structure through Bott periodicity. The stable homotopy groups repeat with period eight, with values

[ \pi_k(O)\cong \begin{cases} \mathbb Z/2\mathbb Z, & k\equiv 0,1\pmod 8,\ \mathbb Z, & k\equiv 3,7\pmod 8,\ 0, & k\equiv 2,4,5,6\pmod 8. \end{cases} ]

This periodicity is a structural foundation of real K-theory. It also organizes the stable classification of real vector bundles and explains recurring eightfold patterns in constructions involving Clifford algebras.

Indefinite orthogonal groups

A nondegenerate real quadratic form need not be positive definite. By Sylvester's law of inertia, every such form can be expressed in a suitable basis as

[ q(x)=x_1^2+\cdots+x_p^2-x_{p+1}^2-\cdots-x_{p+q}^2. ]

Its orthogonal group is written (O(p,q)). In matrix form, it is defined by

[ A^{\mathsf T} \begin{pmatrix} I_p&0\ 0&-I_q \end{pmatrix} A= \begin{pmatrix} I_p&0\ 0&-I_q \end{pmatrix}. ]

When both (p) and (q) are positive, (O(p,q)) is noncompact. It has a maximal compact subgroup isomorphic to (O(p)\times O(q)), and its global topology is governed by a Cartan decomposition relative to that subgroup.

The case (O(1,n)) preserves the quadratic form underlying hyperbolic space. An appropriate identity component acts as the orientation-preserving isometry group of real hyperbolic (n)-space. The group (O(1,3)) is also the full homogeneous Lorentz group associated with four-dimensional Minkowski space, while its identity component preserves both spatial orientation and time orientation.

Algebraic and representation-theoretic structure

Over a general field, orthogonal groups are examples of reductive algebraic groups. Their classification depends on the dimension, the equivalence class of the quadratic form, and arithmetic invariants of the underlying field. Over an algebraically closed field of characteristic different from (2), the connected special orthogonal groups belong to the classical Lie types (B_m) and (D_m).

The odd-dimensional group (SO(2m+1)) has root system of type (B_m). The even-dimensional group (SO(2m)) has root system of type (D_m). Élie Cartan incorporated these families into the classification of complex semisimple Lie algebras, while Hermann Weyl developed their representation theory through highest weights and character formulas.

Representations inherited from the defining action on (F^n) include tensor constructions and exterior powers. Additional representations arise after passage to the spin group, whose spinor representations generally do not descend to ordinary representations of (SO(n)). Their construction uses the associated Clifford algebra.

Historical development

The study of orthogonal transformations developed from the algebraic treatment of quadratic forms and the geometry of rigid motions. Arthur Cayley expressed linear transformations through matrices and examined groups preserving bilinear forms. Sophus Lie placed continuous transformation groups within a differential framework, making the orthogonal groups standard examples of what became Lie groups.

Wilhelm Killing and Élie Cartan analyzed the corresponding infinitesimal algebras during the classification of semisimple Lie algebras. Hermann Weyl subsequently connected compact groups, invariant integration, and representation theory, establishing the orthogonal groups as central examples in the general theory of continuous group representations.

The later topological treatment shifted attention from individual matrix dimensions to the system of inclusions among successive orthogonal groups. This stable viewpoint connected the geometry of spheres, the topology of vector bundles, and periodic phenomena in homotopy theory.