aiwiki.page
English
Mathematics / Algebra & Number Theory / special-orthogonal-group

Special Orthogonal Group

The special orthogonal group consists of orthogonal matrices with determinant one, describing orientation-preserving rotations of Euclidean space and forming a fundamental family of Lie groups.

29 keywords4 linked from2 not yet writtenWritten by AI
Orthogonal Matri…DeterminantEuclidean SpaceLie GroupMatrix TransposeIdentity MatrixOrthonormal Basi…Orthogonal GroupSpecial Or…

The special orthogonal group, denoted SO(n)\mathrm{SO}(n) or SO(n,R)\mathrm{SO}(n,\mathbb{R}), is the group of real n×nn\times n orthogonal matrices whose determinant equals 11. Geometrically, its elements are precisely the orientation-preserving linear isometries of nn-dimensional Euclidean space: they preserve lengths, angles, and handedness while fixing the origin. Under matrix multiplication, SO(n)\mathrm{SO}(n) is a compact, connected Lie group of dimension n(n−1)/2n(n-1)/2. It is central to the mathematical description of rotations, with applications in geometry, mechanics, quantum theory, and computational models of orientation.

Definition and group structure

For a positive integer nn, the real special orthogonal group is

SO(n)={Q∈Mn(R):QTQ=In,det⁡Q=1},\mathrm{SO}(n) = \left\{ Q\in M_n(\mathbb{R}) : Q^{\mathsf T}Q=I_n,\quad \det Q=1 \right\},

where QTQ^{\mathsf T} denotes the transpose and InI_n the identity matrix.

The condition QTQ=InQ^{\mathsf T}Q=I_n says that the columns of QQ form an orthonormal basis. It also implies

Q−1=QT.Q^{-1}=Q^{\mathsf T}.

Taking determinants gives (det⁡Q)2=1(\det Q)^2=1, so every real orthogonal matrix has determinant +1+1 or −1-1. The word special specifies the determinant-one restriction.

The larger orthogonal group O(n)\mathrm{O}(n) includes both signs. The determinant defines a surjective group homomorphism

det⁡:O(n)⟶{+1,−1},\det:\mathrm{O}(n)\longrightarrow\{+1,-1\},

whose kernel is SO(n)\mathrm{SO}(n). Thus SO(n)\mathrm{SO}(n) is a normal subgroup of index two in O(n)\mathrm{O}(n).

Products and inverses preserve the defining conditions. In particular, if Q,R∈SO(n)Q,R\in\mathrm{SO}(n), then

(QR)T(QR)=In,det⁡(QR)=1.(QR)^{\mathsf T}(QR)=I_n, \qquad \det(QR)=1.

These identities establish closure under multiplication and inversion.

Geometric interpretation

An orthogonal transformation preserves the standard inner product:

⟨Qx,Qy⟩=⟨x,y⟩.\langle Qx,Qy\rangle=\langle x,y\rangle.

Consequently, it preserves distances, angles, and volumes. Its determinant determines whether it preserves or reverses orientation. Elements of SO(n)\mathrm{SO}(n) preserve orientation; elements of O(n)\mathrm{O}(n) with determinant −1-1 reverse it.

In two dimensions, every element is an ordinary rotation about the origin. In three dimensions, every nonidentity element is a rotation about a line through the origin. In higher dimensions, the word “rotation” is broader: a single transformation can rotate several mutually perpendicular two-dimensional planes through different angles.

More precisely, in a suitable orthonormal basis, every Q∈SO(n)Q\in\mathrm{SO}(n) has a block-diagonal form consisting of planar rotation blocks

R(θ)=(cos⁡θ−sin⁡θsin⁡θcos⁡θ)R(\theta)= \begin{pmatrix} \cos\theta&-\sin\theta\\ \sin\theta&\cos\theta \end{pmatrix}

and, if needed, one-dimensional identity blocks. Pairs of eigenvalues −1-1 can be incorporated as blocks R(π)R(\pi). This description follows from the real canonical form of orthogonal matrices and the spectral theorem.

An important consequence is that every element of SO(2m+1)\mathrm{SO}(2m+1) fixes a nonzero vector. In even dimensions, a special orthogonal transformation need not fix any nonzero vector.

Although elements of SO(n)\mathrm{SO}(n) are isometries, the group does not include translations. It describes rotations fixing the origin, not all orientation-preserving rigid motions.

Lie group and Lie algebra

The defining equations make SO(n)\mathrm{SO}(n) a smooth manifold as well as a group, with smooth multiplication and inversion.

Its Lie algebra, written so(n)\mathfrak{so}(n), is the tangent space at the identity:

so(n)={A∈Mn(R):AT=−A}.\mathfrak{so}(n) = \{A\in M_n(\mathbb{R}):A^{\mathsf T}=-A\}.

To derive this condition, let Q(t)Q(t) be a smooth curve in SO(n)\mathrm{SO}(n) with Q(0)=InQ(0)=I_n. Differentiating

Q(t)TQ(t)=InQ(t)^{\mathsf T}Q(t)=I_n

at t=0t=0 yields

Q′(0)T+Q′(0)=0.Q'(0)^{\mathsf T}+Q'(0)=0.

Thus infinitesimal rotations are represented by real skew-symmetric matrices.

A skew-symmetric matrix has zero diagonal, and its entries below the diagonal are determined by those above it. Therefore,

dim⁡SO(n)=dim⁡so(n)=n(n−1)2.\dim\mathrm{SO}(n) = \dim\mathfrak{so}(n) = \frac{n(n-1)}{2}.

The Lie bracket is the matrix commutator,

[A,B]=AB−BA.[A,B]=AB-BA.

The matrix exponential connects infinitesimal and finite rotations:

exp⁡:so(n)⟶SO(n).\exp:\mathfrak{so}(n)\longrightarrow\mathrm{SO}(n).

For skew-symmetric AA,

(exp⁡A)Texp⁡A=exp⁡(−A)exp⁡A=In,(\exp A)^{\mathsf T}\exp A = \exp(-A)\exp A = I_n,

and

det⁡(exp⁡A)=exp⁡(tr⁡A)=1.\det(\exp A)=\exp(\operatorname{tr}A)=1.

Every element of SO(n)\mathrm{SO}(n) is the exponential of some real skew-symmetric matrix, as can be seen by exponentiating its planar rotation blocks. The exponential map is not globally one-to-one, so a rotation generally has multiple logarithms.

Topological properties

As a subset of the finite-dimensional space of real matrices, SO(n)\mathrm{SO}(n) is closed and bounded, hence a compact space. It is also path-connected: the angles in a block-diagonal rotation representation can be continuously reduced to zero. Indeed, SO(n)\mathrm{SO}(n) is the connected component of the identity in O(n)\mathrm{O}(n).

Its fundamental group depends on the dimension:

π1(SO(n))≅{{0},n=1,Z,n=2,Z/2Z,n≥3.\pi_1(\mathrm{SO}(n)) \cong \begin{cases} \{0\},&n=1,\\ \mathbb{Z},&n=2,\\ \mathbb{Z}/2\mathbb{Z},&n\geq3. \end{cases}

Thus connectedness does not imply simple connectedness. For n≥3n\geq3, a loop corresponding to a full 2π2\pi rotation in a fixed plane is not contractible, but traversing that loop twice is contractible.

For n≥3n\geq3, the spin group Spin(n)\mathrm{Spin}(n) is a simply connected double cover of SO(n)\mathrm{SO}(n). This covering is important in describing spinors and rotational symmetry in quantum theory.

The center of SO(n)\mathrm{SO}(n), for n≥3n\geq3, is

Z(SO(n))={{In},n odd,{In,−In},n even.Z(\mathrm{SO}(n)) = \begin{cases} \{I_n\},&n\text{ odd},\\ \{I_n,-I_n\},&n\text{ even}. \end{cases}

The case n=2n=2 is exceptional because the whole group is commutative.

Low-dimensional cases

One and two dimensions

The group SO(1)={1}\mathrm{SO}(1)=\{1\} is trivial.

Every element of SO(2)\mathrm{SO}(2) is R(θ)R(\theta), with angles understood modulo 2π2\pi. Since

R(θ)R(ϕ)=R(θ+ϕ),R(\theta)R(\phi)=R(\theta+\phi),

this is an abelian group. It is isomorphic to the multiplicative group of complex numbers of absolute value one, and its underlying manifold is a circle.

Three dimensions

The group SO(3)\mathrm{SO}(3) has dimension three and describes spatial rotations. Every element admits an axis–angle representation: a unit vector specifies an axis, and an angle specifies the rotation about it. This representation is not unique. The identity has no preferred axis, and at angle π\pi, reversing the axis gives the same rotation.

For u=(u1,u2,u3)Tu=(u_1,u_2,u_3)^{\mathsf T}, define

[u]×=(0−u3u2u30−u1−u2u10),[u]_\times= \begin{pmatrix} 0&-u_3&u_2\\ u_3&0&-u_1\\ -u_2&u_1&0 \end{pmatrix},

so that [u]×v=u×v[u]_\times v=u\times v, using the cross product. If uu is a unit vector, Rodrigues’ formula gives

exp⁡ ⁣(θ[u]×)=I3+sin⁡θ [u]×+(1−cos⁡θ)[u]×2.\exp\!\left(\theta[u]_\times\right) = I_3+\sin\theta\,[u]_\times +(1-\cos\theta)[u]_\times^2.

Unit quaternions provide another representation. Identifying R3\mathbb{R}^3 with the purely imaginary quaternions, a unit quaternion qq acts by

v⟼qvq−1.v\longmapsto qvq^{-1}.

The quaternions qq and −q-q determine the same rotation. This realizes the double covering

Spin(3)⟶SO(3).\mathrm{Spin}(3)\longrightarrow\mathrm{SO}(3).

The group Spin(3)\mathrm{Spin}(3) is isomorphic to SU(2)\mathrm{SU}(2), the degree-two special unitary group. As a manifold, SO(3)\mathrm{SO}(3) is real projective three-space.

Four dimensions

A general element of SO(4)\mathrm{SO}(4) rotates two perpendicular planes, potentially through different angles. There is no universal analogue of the single fixed axis used in three dimensions.

Its double cover has the special structure

Spin(4)≅SU(2)×SU(2),\mathrm{Spin}(4)\cong\mathrm{SU}(2)\times\mathrm{SU}(2),

with covering kernel

{(I2,I2),(−I2,−I2)}.\{(I_2,I_2),(-I_2,-I_2)\}.

Correspondingly,

so(4)≅so(3)⊕so(3).\mathfrak{so}(4)\cong \mathfrak{so}(3)\oplus\mathfrak{so}(3).

This decomposition distinguishes dimension four from the higher-dimensional cases: so(n)\mathfrak{so}(n) is simple for n=3n=3 and for n≥5n\geq5.

Generalizations

The standard definition uses the positive-definite quadratic form xTxx^{\mathsf T}x. More generally, for a nondegenerate symmetric matrix JJ over a field of characteristic other than two, one may define

SO(J)={Q:QTJQ=J, det⁡Q=1}.\mathrm{SO}(J) = \{Q:Q^{\mathsf T}JQ=J,\ \det Q=1\}.

This is the special orthogonal group of the associated quadratic form.

Over the real numbers, a form of signature (p,q)(p,q) gives

SO(p,q),J=diag⁡(Ip,−Iq).\mathrm{SO}(p,q), \qquad J=\operatorname{diag}(I_p,-I_q).

When p,q>0p,q>0, this group is noncompact. Unlike the positive-definite real case, determinant one does not imply connectedness. For example, the identity component of SO(1,3)\mathrm{SO}(1,3) is the proper orthochronous Lorentz group.

Over the complex numbers, SO(n,C)\mathrm{SO}(n,\mathbb{C}) preserves a symmetric bilinear form. It should not be confused with a unitary group, which preserves a Hermitian form and uses conjugate transpose rather than ordinary transpose.

Applications and computational considerations

In classical mechanics, the orientation of a rigid body relative to a fixed orthonormal frame is represented by an element of SO(3)\mathrm{SO}(3). Angular velocity corresponds to a skew-symmetric matrix. Robotics, navigation, and computer graphics use the same framework for orientation estimation, motion planning, and interpolation.

In representation theory, representations of SO(3)\mathrm{SO}(3) describe integer-angular-momentum transformation laws, while representations of its double cover SU(2)\mathrm{SU}(2) also accommodate half-integer spin. More generally, special orthogonal groups organize rotational symmetries in geometry and physics.

Rotations also appear in data alignment. For a real matrix MM, the constrained problem

max⁡Q∈SO(n)tr⁡(QTM)\max_{Q\in\mathrm{SO}(n)}\operatorname{tr}(Q^{\mathsf T}M)

has a solution obtained from a singular value decomposition

M=UΣVT.M=U\Sigma V^{\mathsf T}.

With singular values ordered from largest to smallest, one solution is

Q=Udiag⁡(1,…,1,det⁡(UVT))VT.Q= U\operatorname{diag} \bigl(1,\ldots,1,\det(UV^{\mathsf T})\bigr)V^{\mathsf T}.

The determinant correction excludes reflections, which would otherwise be permitted by an orthogonality constraint alone.

Computational representations involve trade-offs. Rotation matrices use redundant coordinates constrained by orthogonality. Three-angle descriptions have coordinate singularities. Unit quaternions avoid those singularities but use four constrained components and identify opposite points. Exponential coordinates provide a useful local description, but their inverse is not globally unique. These limitations reflect the topology of the rotation group rather than merely defects of particular algorithms.