_

SO(3)SO(3) is a manifold

Identity as a regular value proof

To show that SO(3)SO(3) is a manifold, we consider it as a subspace of the Euclidean space of matrices M(3,)9M(3, \mathbb{R}) \cong \mathbb{R}^9.

1. The Mapping

Define a smooth map f:M(3,)Sym(3,)f: M(3, \mathbb{R}) \to \text{Sym}(3, \mathbb{R}) by: f(A)=ATAf(A) = A^T A where Sym(3,)6\text{Sym}(3, \mathbb{R}) \cong \mathbb{R}^6 is the space of 3×33 \times 3 symmetric matrices. The orthogonal group O(3)O(3) is the preimage of the identity matrix: O(3)=f1(I)O(3) = f^{-1}(I)

2. The Derivative

To apply the Regular Value Theorem, we calculate the derivative DfADf_A at a point AO(3)A \in O(3) in the direction of a tangent vector HM(3,)H \in M(3, \mathbb{R}): DfA(H)=ddt|t=0f(A+tH)Df_A(H) = \left. \frac{d}{dt} \right|_{t=0} f(A + tH) DfA(H)=ddt|t=0(A+tH)T(A+tH)Df_A(H) = \left. \frac{d}{dt} \right|_{t=0} (A + tH)^T(A + tH) DfA(H)=ATH+HTADf_A(H) = A^T H + H^T A

3. Surjectivity Proof

For II to be a regular value, DfADf_A must be surjective onto Sym(3,)\text{Sym}(3, \mathbb{R}). Let SSym(3,)S \in \text{Sym}(3, \mathbb{R}) be an arbitrary symmetric matrix. We must find an HH such that DfA(H)=SDf_A(H) = S.

Choose H=12ASH = \frac{1}{2} AS. Substituting this into the derivative: DfA(12AS)=AT(12AS)+(12AS)TADf_A\left(\frac{1}{2} AS\right) = A^T\left(\frac{1}{2} AS\right) + \left(\frac{1}{2} AS\right)^T A Since ATA=IA^T A = I and ST=SS^T = S: DfA(12AS)=12(ATA)S+12ST(ATA)Df_A\left(\frac{1}{2} AS\right) = \frac{1}{2}(A^T A)S + \frac{1}{2}S^T(A^T A) DfA(12AS)=12IS+12SI=SDf_A\left(\frac{1}{2} AS\right) = \frac{1}{2}IS + \frac{1}{2}SI = S

4. Conclusion

Since DfADf_A is surjective for all Af1(I)A \in f^{-1}(I), the Regular Value Theorem implies that O(3)O(3) is a smooth manifold of dimension: dim(M(3,))dim(Sym(3,))=96=3\dim(M(3, \mathbb{R})) - \dim(\text{Sym}(3, \mathbb{R})) = 9 - 6 = 3 Because SO(3)SO(3) is the subgroup where det(A)=1\det(A) = 1 (an open condition in O(3)O(3)), it is also a smooth 33-dimensional manifold.


Proof using the Exponential Map

To show SO(3)SO(3) is a manifold using the exponential map, we demonstrate that the group is locally homeomorphic to a linear vector space.

1. The Lie Algebra 𝔰𝔬(3)\mathfrak{so}(3)

The tangent space at the identity, denoted 𝔰𝔬(3)\mathfrak{so}(3), consists of 3×33 \times 3 skew-symmetric matrices: 𝔰𝔬(3)={XM(3,):XT=X}\mathfrak{so}(3) = \{ X \in M(3, \mathbb{R}) : X^T = -X \} An arbitrary element X𝔰𝔬(3)X \in \mathfrak{so}(3) has the form: X=(0zyz0xyx0)X = \begin{pmatrix} 0 & -z & y \\ z & 0 & -x \\ -y & x & 0 \end{pmatrix} This is a linear subspace of 9\mathbb{R}^9 with dimension 33.

2. The Exponential Map

The matrix exponential exp:M(3,)M(3,)\exp: M(3, \mathbb{R}) \to M(3, \mathbb{R}) is defined by the power series: exp(X)=k=0Xkk!\exp(X) = \sum_{k=0}^{\infty} \frac{X^k}{k!} If XX is skew-symmetric, then R=exp(X)R = \exp(X) is orthogonal (exp(X)T=exp(XT)=exp(X)=exp(X)1\exp(X)^T = \exp(X^T) = \exp(-X) = \exp(X)^{-1}) and has det(R)=1\det(R) = 1. Thus, exp\exp maps 𝔰𝔬(3)\mathfrak{so}(3) into SO(3)SO(3).

3. Local Diffeomorphism

The derivative of the exponential map at the origin 0𝔰𝔬(3)0 \in \mathfrak{so}(3) is the identity map: D(exp)0=ID(\exp)_0 = I By the Inverse Function Theorem, there exists a neighborhood UU of 00 in 𝔰𝔬(3)\mathfrak{so}(3) and a neighborhood VV of II in SO(3)SO(3) such that exp:UV\exp: U \to V is a diffeomorphism.

4. Manifold Structure

This provides a coordinate chart (V,exp1)(V, \exp^{-1}) around the identity matrix II. For any other point RSO(3)R \in SO(3), we can define a chart around RR by shifting the identity chart: ϕR(A)=exp1(R1A)\phi_R(A) = \exp^{-1}(R^{-1}A) Because we can cover the entire group with these smooth charts, SO(3)SO(3) is a smooth manifold of dimension 33.


Homeomorphism between SO(3)SO(3) and P3\mathbb{R}P^3

To show SO(3)P3SO(3) \cong \mathbb{R}P^3, we utilize the axis-angle representation of rotations.

1. The Ball Model

Consider the closed ball of radius π\pi in 3\mathbb{R}^3: Dπ3={v3:vπ}D^3_{\pi} = \{ \vec{v} \in \mathbb{R}^3 : \|\vec{v}\| \le \pi \} For any vDπ3\vec{v} \in D^3_{\pi}, let θ=v\theta = \|\vec{v}\| and u=v/θ\vec{u} = \vec{v}/\theta. We associate v\vec{v} with the rotation Ru,θSO(3)R_{\vec{u}, \theta} \in SO(3).

2. Boundary Identification

This map is a bijection on the interior of the ball. However, on the boundary where v=π\|\vec{v}\| = \pi, the rotation Ru,πR_{\vec{u}, \pi} is identical to Ru,πR_{-\vec{u}, \pi} because a rotation by 180180^\circ is orientation-equivalent to a rotation by 180-180^\circ (or 180180^\circ in the opposite direction). Thus, we define an equivalence relation \sim on Dπ3D^3_{\pi}: vvv=π\vec{v} \sim -\vec{v} \iff \|\vec{v}\| = \pi The resulting quotient space is Dπ3/D^3_{\pi} / \sim.

3. Homeomorphism to P3\mathbb{R}P^3

By definition, the real projective space P3\mathbb{R}P^3 can be constructed by taking a 33-disk D3D^3 and identifying antipodal points on its boundary sphere S2S^2. Therefore: SO(3)Dπ3/P3SO(3) \cong D^3_{\pi} / \sim \;\cong \mathbb{R}P^3


Killing form

To compute the signature of the Killing form B(X,Y)B(X, Y) for 𝔰𝔬(3)\mathfrak{so}(3), we use the adjoint representation.

1. The Basis and Commutation Relations

A standard basis for the Lie algebra 𝔰𝔬(3)\mathfrak{so}(3) consists of the skew-symmetric matrices {L1,L2,L3}\{L_1, L_2, L_3\}: L1=(000001010),L2=(001000100),L3=(010100000)L_1 = \begin{pmatrix} 0 & 0 & 0 \\ 0 & 0 & -1 \\ 0 & 1 & 0 \end{pmatrix}, \quad L_2 = \begin{pmatrix} 0 & 0 & 1 \\ 0 & 0 & 0 \\ -1 & 0 & 0 \end{pmatrix}, \quad L_3 = \begin{pmatrix} 0 & -1 & 0 \\ 1 & 0 & 0 \\ 0 & 0 & 0 \end{pmatrix} The commutation relations are given by [Li,Lj]=k=13ϵijkLk[L_i, L_j] = \sum_{k=1}^3 \epsilon_{ijk} L_k, or specifically: [L1,L2]=L3,[L2,L3]=L1,[L3,L1]=L2[L_1, L_2] = L_3, \quad [L_2, L_3] = L_1, \quad [L_3, L_1] = L_2

2. The Adjoint Representation

The Killing form is defined as B(X,Y)=tr(adXadY)B(X, Y) = \text{tr}(\text{ad}_X \text{ad}_Y). We first find the matrix representation of adL1\text{ad}_{L_1} acting on the basis {L1,L2,L3}\{L_1, L_2, L_3\}: * adL1(L1)=[L1,L1]=0\text{ad}_{L_1}(L_1) = [L_1, L_1] = 0 * adL1(L2)=[L1,L2]=L3\text{ad}_{L_1}(L_2) = [L_1, L_2] = L_3 * adL1(L3)=[L1,L3]=L2\text{ad}_{L_1}(L_3) = [L_1, L_3] = -L_2

In matrix form: adL1=(000001010)\text{ad}_{L_1} = \begin{pmatrix} 0 & 0 & 0 \\ 0 & 0 & -1 \\ 0 & 1 & 0 \end{pmatrix}

3. Calculating the Trace

To find B(L1,L1)B(L_1, L_1), we compute (adL1)2(\text{ad}_{L_1})^2: (adL1)2=(000010001)(\text{ad}_{L_1})^2 = \begin{pmatrix} 0 & 0 & 0 \\ 0 & -1 & 0 \\ 0 & 0 & -1 \end{pmatrix} The trace is: B(L1,L1)=tr((adL1)2)=011=2B(L_1, L_1) = \text{tr}((\text{ad}_{L_1})^2) = 0 - 1 - 1 = -2 By symmetry of the commutation relations, it follows that B(L2,L2)=2B(L_2, L_2) = -2 and B(L3,L3)=2B(L_3, L_3) = -2. For iji \neq j, the off-diagonal terms B(Li,Lj)B(L_i, L_j) vanish.

4. The Killing Matrix and Signature

The matrix representing the Killing form in this basis is: 𝐁=(200020002)\mathbf{B} = \begin{pmatrix} -2 & 0 & 0 \\ 0 & -2 & 0 \\ 0 & 0 & -2 \end{pmatrix} The eigenvalues are {2,2,2}\{-2, -2, -2\}. The signature is (0,3)(0, 3), meaning the form is negative definite.