Advanced Mathematical Concepts
Angles via Dot Product and Inner Products
In vector spaces, particularly Euclidean space Rn\mathbb{R}^nRn, the angle between two vectors u\mathbf{u}u and v\mathbf{v}v is defined using the dot product, which provides a scalar measure of their alignment. The dot product is given by u⋅v=∑i=1nuivi\mathbf{u} \cdot \mathbf{v} = \sum_{i=1}^n u_i v_iu⋅v=∑i=1nuivi, and it relates to the angle θ\thetaθ between the vectors through the formula u⋅v=∥u∥∥v∥cosθ\mathbf{u} \cdot \mathbf{v} = |\mathbf{u}| |\mathbf{v}| \cos \thetau⋅v=∥u∥∥v∥cosθ, where ∥u∥=u⋅u|\mathbf{u}| = \sqrt{\mathbf{u} \cdot \mathbf{u}}∥u∥=u⋅u is the Euclidean norm.[64][65] Solving for the angle yields θ=arccos(u⋅v∥u∥∥v∥)\theta = \arccos \left( \frac{\mathbf{u} \cdot \mathbf{v}}{|\mathbf{u}| |\mathbf{v}|} \right)θ=arccos(∥u∥∥v∥u⋅v), which assumes 0≤θ≤π0 \leq \theta \leq \pi0≤θ≤π to ensure the cosine is non-negative in the relevant range.[66][67]
This formulation implies that vectors are orthogonal if θ=90∘\theta = 90^\circθ=90∘, where cosθ=0\cos \theta = 0cosθ=0, so u⋅v=0\mathbf{u} \cdot \mathbf{v} = 0u⋅v=0.[64][65] In R2\mathbb{R}^2R2 and R3\mathbb{R}^3R3, the dot product geometrically interprets the angle as the smaller angle between the directions of the vectors, aligning with intuitive notions from plane and space geometry.[68][69] For instance, consider the vectors ⟨1,0⟩\langle 1, 0 \rangle⟨1,0⟩ and ⟨1,1⟩\langle 1, 1 \rangle⟨1,1⟩ in R2\mathbb{R}^2R2: their dot product is 1⋅1+0⋅1=11 \cdot 1 + 0 \cdot 1 = 11⋅1+0⋅1=1, with norms ∥⟨1,0⟩∥=1|\langle 1, 0 \rangle| = 1∥⟨1,0⟩∥=1 and ∥⟨1,1⟩∥=2|\langle 1, 1 \rangle| = \sqrt{2}∥⟨1,1⟩∥=2, so cosθ=11⋅2=12\cos \theta = \frac{1}{1 \cdot \sqrt{2}} = \frac{1}{\sqrt{2}}cosθ=1⋅21=21 and θ=45∘\theta = 45^\circθ=45∘.[64][70]
The Cauchy-Schwarz inequality underpins the validity of this angle definition by stating that ∣u⋅v∣≤∥u∥∥v∥|\mathbf{u} \cdot \mathbf{v}| \leq |\mathbf{u}| |\mathbf{v}|∣u⋅v∣≤∥u∥∥v∥, which implies ∣cosθ∣≤1|\cos \theta| \leq 1∣cosθ∣≤1, ensuring θ\thetaθ is well-defined in [0,π][0, \pi][0,π].[71][72] Equality holds when the vectors are linearly dependent.[73]
This concept generalizes to inner product spaces, where an inner product ⟨u,v⟩\langle \mathbf{u}, \mathbf{v} \rangle⟨u,v⟩ replaces the dot product, satisfying properties like linearity, symmetry, and positive-definiteness. In a Hilbert space—a complete inner product space—the angle is defined analogously by ⟨u,v⟩=∥u∥∥v∥cosθ\langle \mathbf{u}, \mathbf{v} \rangle = |\mathbf{u}| |\mathbf{v}| \cos \theta⟨u,v⟩=∥u∥∥v∥cosθ, with ∥u∥=⟨u,u⟩|\mathbf{u}| = \sqrt{\langle \mathbf{u}, \mathbf{u} \rangle}∥u∥=⟨u,u⟩, extending the Euclidean case to infinite-dimensional settings like function spaces.[74][75] Orthogonality follows similarly as ⟨u,v⟩=0\langle \mathbf{u}, \mathbf{v} \rangle = 0⟨u,v⟩=0.[70][76]
Angles Between Subspaces
In linear algebra, the concept of angles between subspaces generalizes the angle between individual vectors to describe the relative orientation of two linear subspaces UUU and VVV in a Euclidean space Rn\mathbb{R}^nRn. The principal angles θ1≤θ2≤⋯≤θk∈[0,π/2]\theta_1 \leq \theta_2 \leq \cdots \leq \theta_k \in [0, \pi/2]θ1≤θ2≤⋯≤θk∈[0,π/2], where k=min(dimU,dimV)k = \min(\dim U, \dim V)k=min(dimU,dimV), provide a complete characterization of this orientation, with cosθi\cos \theta_icosθi representing the singular values (in decreasing order) of the matrix QUTQVQ_U^T Q_VQUTQV, where QUQ_UQU and QVQ_VQV are matrices whose columns form orthonormal bases for UUU and VVV, respectively.[77][78] This definition originates from the work of Camille Jordan in 1875, who introduced principal angles and vectors recursively as cosθk=max{∣⟨x,y⟩∣:x∈U,y∈V,∥x∥=∥y∥=1,x⊥u1,…,uk−1,y⊥v1,…,vk−1}\cos \theta_k = \max { |\langle x, y \rangle| : x \in U, y \in V, |x| = |y| = 1, x \perp u_1, \dots, u_{k-1}, y \perp v_1, \dots, v_{k-1} }cosθk=max{∣⟨x,y⟩∣:x∈U,y∈V,∥x∥=∥y∥=1,x⊥u1,…,uk−1,y⊥v1,…,vk−1}, where uiu_iui and viv_ivi are the corresponding principal vectors achieving the maxima for previous angles.[79]
A key special case occurs when UUU and VVV are orthogonal, meaning U∩V={0}U \cap V = {0}U∩V={0} and ⟨u,v⟩=0\langle u, v \rangle = 0⟨u,v⟩=0 for all u∈Uu \in Uu∈U, v∈Vv \in Vv∈V; in this situation, QUTQV=0Q_U^T Q_V = 0QUTQV=0, so all singular values are zero, and thus all principal angles are θi=π/2\theta_i = \pi/2θi=π/2.[77] For one-dimensional subspaces (lines through the origin), the construction reduces to a single principal angle θ1\theta_1θ1, which is precisely the angle between their direction vectors as defined via the dot product: cosθ1=∣⟨u,v⟩∣\cos \theta_1 = |\langle u, v \rangle|cosθ1=∣⟨u,v⟩∣ for unit vectors u,vu, vu,v spanning the lines.[78]
An illustrative example in R3\mathbb{R}^3R3 is the angle between two planes, which are two-dimensional subspaces. If the planes intersect along a line, the principal angles consist of θ1=0\theta_1 = 0θ1=0 (corresponding to unit vectors along the intersection direction) and θ2=ϕ\theta_2 = \phiθ2=ϕ, where ϕ\phiϕ is the dihedral angle between the planes, computed as the angle between their normals.[80] More generally, in the Grassmannian Gr(k,n)\mathrm{Gr}(k, n)Gr(k,n)—the manifold parameterizing all kkk-dimensional subspaces of Rn\mathbb{R}^nRn—the principal angles between two points (subspaces) serve as coordinates for their relative orientation, enabling metrics such as the two-norm of the principal angle vector to quantify geodesic distances on the manifold.[81]
Angles in Non-Euclidean Geometries
In Riemannian geometry, the angle θ between two tangent vectors u and v at a point on a manifold is defined using the metric tensor g via the relation
cosθ=g(u,v)g(u,u)g(v,v).\cos \theta = \frac{g(u,v)}{\sqrt{g(u,u)} \sqrt{g(v,v)}}.cosθ=g(u,u)g(v,v)g(u,v).
This formula generalizes the inner product in Euclidean space, enabling the measurement of angles in curved spaces where the metric varies across the manifold. The positive definiteness of g ensures that θ lies between 0 and π radians, preserving the standard interpretation of angles while accounting for local curvature.
In hyperbolic geometry, angles between geodesics are defined analogously through the hyperbolic metric, but a distinctive parameter is the rapidity φ, which serves as a hyperbolic analogue to the ordinary angle θ. Unlike the Euclidean case where arc length equals radius times θ, in the hyperbolic plane, distances along geodesics correspond directly to φ in units of the curvature radius, facilitating additive compositions in transformations.[82] A fundamental property is the angle sum of triangles: in elliptic geometries like the sphere, it exceeds π radians, while in hyperbolic geometry, it is less than π radians, with the defect or excess proportional to the enclosed area and inversely to the curvature.[83]
Illustrative examples highlight these properties. On a sphere, an elliptic space, angles form between great circles—geodesics that are intersections of the sphere with planes through its center; for instance, the angle between two meridians at the pole equals their longitudinal separation.[84] The Gauss-Bonnet theorem further connects angles to curvature: for a geodesic triangle, the integral of Gaussian curvature K over its interior equals the angle excess (sum of interior angles minus π), linking local geometry to global topology.[85]
Hyperbolic trigonometry provides tools for computation, as in the hyperbolic law of cosines:
coshc=coshacoshb−sinhasinhbcosC,\cosh c = \cosh a \cosh b - \sinh a \sinh b \cos C,coshc=coshacoshb−sinhasinhbcosC,
which relates side lengths a, b, c to the opposite angle C, adapting Euclidean formulas to negative curvature and yielding results like larger possible angles for given sides compared to the plane.[86] This law underscores how angles in hyperbolic triangles can be arbitrarily small while maintaining the sub-π sum, reflecting the expansive nature of the space.[86]