logo

Gauss Curvature and Geodesic Curvature 📂Geometry

Gauss Curvature and Geodesic Curvature

Buildup1

{T(s),N(s),B(s),κ(s),τ(s)} \left\{ T(s), N(s), B(s), \kappa (s), \tau (s) \right\}

Recall how we used the Frenet-Serret apparatus when analyzing curves. When studying surfaces, we will consider similar concepts. When α\boldsymbol{\alpha} is the unit speed curve, the curvature of the curve was defined as the magnitude of acceleration κ=T=α\kappa = \left| T^{\prime} \right| = \left| \boldsymbol{\alpha}^{\prime \prime} \right|. It is natural to think about how curved a surface is by looking at how curved the curves on the surface are.

Consider the surface given as x:UR2M\mathbf{x} : U\subset \R^{2} \to M. Let α(s)\boldsymbol{\alpha}(s) be the unit speed curve on the simple surface x\mathbf{x}. Then let’s denote the Frenet-Serret apparatus for α\boldsymbol{\alpha} as follows.

{T,N,B,κ,τ} \left\{ \mathbf{T}, \mathbf{N}, \mathbf{B}, \kappa, \tau \right\}

Let’s call the unit normal at point pMp \in M as n\mathbf{n}. Let’s define the set of all vectors that are point pp as NpMN_{p}M.

NpM:={rn:rR}={all vectors perpendicular to M at p} N_{p}M := \left\{ r \mathbf{n} : r \in \R \right\} = \left\{ \text{all vectors perpendicular to } M \text{ at } p \right\}

Then by definition of the tangent plane, TpMT_{p}M is the orthogonal complement of NpMN_{p}M.

NpM=TpM N_{p}M ^{\perp} = T_{p}M

Therefore, R3\R^{3} is orthogonally decomposed as follows, and α\boldsymbol{\alpha}^{\prime \prime} can be expressed as the linear combination of vectors of the two spaces.

R3=NpMTpMandα(s)=n1n(s)+n2n(s)(nNpM, nTpM) \R^{3} = N_{p}M \oplus T_{p}M \quad \text{and} \quad \boldsymbol{\alpha}^{\prime \prime}(s) = n_{1}\mathbf{n}(s) + n_{2}\mathbf{n}^{\perp}(s)\quad (\mathbf{n}\in N_{p}M,\ \mathbf{n}^{\perp}\in T_{p}M)

Let’s call T=α\mathbf{T} = \boldsymbol{\alpha}^{\prime} the tangent vector. Since α\boldsymbol{\alpha} is the unit speed vector, the following equation holds.

α(s)2=T(s)2=T,T=1 \left| \boldsymbol{\alpha}^{\prime}(s) \right|^{2} = \left| \mathbf{T}(s) \right|^{2} = \left\langle \mathbf{T}, \mathbf{T} \right\rangle = 1

Differentiating both sides gives the following result by the derivative of dot product.

T,T= 0    T,T= 0    α,T= 0 \begin{align*} && \left\langle \mathbf{T}, \mathbf{T} \right\rangle^{\prime} =&\ 0 \\ \implies && \left\langle \mathbf{T}^{\prime}, \mathbf{T} \right\rangle =&\ 0 \\ \implies && \left\langle \boldsymbol{\alpha}^{\prime \prime}, \mathbf{T} \right\rangle =&\ 0 \end{align*}

Therefore, α\boldsymbol{\alpha}^{\prime \prime} is perpendicular to T\mathbf{T}. Separating α\boldsymbol{\alpha}^{\prime \prime}, since n\mathbf{n} and T\mathbf{T} are perpendicular to each other, we obtain the following.

α,T= 0    n1n+n2n,T= 0    n1n,T+n2n,T= 0    n2n,T= 0 \begin{align*} && \left\langle \boldsymbol{\alpha}^{\prime \prime}, \mathbf{T} \right\rangle =&\ 0 \\ \implies && \left\langle n_{1}\mathbf{n} + n_{2}\mathbf{n}^{\perp}, \mathbf{T} \right\rangle =&\ 0 \\ \implies && \left\langle n_{1}\mathbf{n}, \mathbf{T} \right\rangle + \left\langle n_{2}\mathbf{n}^{\perp}, \mathbf{T} \right\rangle =&\ 0 \\ \implies && \left\langle n_{2}\mathbf{n}^{\perp}, \mathbf{T} \right\rangle =&\ 0 \end{align*}

Therefore, it can be known that n\mathbf{n}^{\perp} is a vector perpendicular to both n\mathbf{n} and T\mathbf{T}. Let’s define the vector S\mathbf{S} as follows.

S:=n×Tandα=n1n+sS \mathbf{S} := \mathbf{n}\times \mathbf{T} \quad \text{and} \quad \boldsymbol{\alpha}^{\prime \prime} = n_{1}\mathbf{n} + s\mathbf{S}

S\mathbf{S} is called the intrinsic normal of α\boldsymbol{\alpha}.

Definition

The component n1n_{1} of n\mathbf{n} is called the normal curvature of the unit speed curve α\boldsymbol{\alpha}, denoted as κn\kappa_{n}.

κn:=α,n \kappa_{n} := \left\langle \boldsymbol{\alpha}^{\prime \prime}, \mathbf{n} \right\rangle

The component ss of S\mathbf{S} is called the geodesic curvature of the unit speed curve α\boldsymbol{\alpha}, denoted as κg\kappa_{g}.

κg:=α,S \kappa_{g} := \left\langle \boldsymbol{\alpha}^{\prime \prime}, \mathbf{S} \right\rangle

Therefore, the following equation holds.

κ(s)N(s)=T(s)=α(s)=κn(s)n(s)+κg(s)S(s) \kappa (s) \mathbf{N}(s) = \mathbf{T}^{\prime}(s) = \boldsymbol{\alpha}^{\prime \prime}(s) = \kappa_{n}(s)\mathbf{n}(s) + \kappa_{g}(s)\mathbf{S}(s)

Explanation

The normal curvature κn\kappa_{n} is used to measure how much the surface MM is curved at R3\R^{3}. The geodesic curvature κg\kappa_{g} is used to measure how much the curve α\boldsymbol{\alpha} is curved on the surface MM. For example, a curve with the geodesic curvature κg\kappa_{g} equal to 00 implies a straight line on the surface, i.e., a geodesic.

Since n,S\mathbf{n}, \mathbf{S} is a unit vector, the following equation holds according to the definition above.

κ2=κn2+κg2 \kappa^{2} = \kappa_{n}^{2} + \kappa_{g}^{2}


  1. Richard S. Millman and George D. parker, Elements of Differential Geometry (1977), p102-104 ↩︎