Chapter 4 › 4.1 Geodesic Curvature › 4.1.2 Geodesic curvature
Subsection

4.1.2  Geodesic curvature

Assume for simplicity that \(\gamma\) is a curve on a two-dimensional surface \(\Sigma^2\subset\mathbb{R}^3\) parametrized by arc-length. Then \(\gamma\) is a geodesic if and only if \(\nabla_{\gamma'}\gamma'\equiv 0\). Therefore, the quantity \(|\nabla_{\gamma'}\gamma'|\) somehow measures the deviation from being a geodesic. This quantity, more or less, is called the geodesic curvature — however, we hope to ensure this quantity is differentiable too. Similar to the definition of signed curvature of plane curves, we define the geodesic curvature using the dot product instead of taking norm. The detail is as follows:

Definition 4.5 (Geodesic curvature)

Let \(\Sigma^2\subset\mathbb{R}^3\) be a regular surface, and \(\gamma(s):[0,L]\to\Sigma\) be an arc-length parametrized curve on \(\Sigma\). We denote and define the geodesic curvature of \(\gamma\) on \(\Sigma\) at point \(\gamma(s)\) as: \[k_g(s):=\langle\gamma''(s),\nu(\gamma(s))\times\gamma'(s)\rangle,\] or in short, \(k_g=\langle\gamma'',\nu\times\gamma'\rangle\).

Checkpoint
The geodesic curvature of an arc-length parametrized curve \(\gamma(s)\) on \(\Sigma^2\subset\mathbb{R}^3\) is defined in Definition 4.5 as \(k_g=\langle\gamma'',\,\nu\times\gamma'\rangle\). Which of the following best explains what \(k_g\) measures?
WhyThe vector \(\nu\times\gamma'\) is the unit tangent vector of \(\Sigma\) perpendicular to \(\gamma'\) (it lies in the tangent plane and is orthogonal to the direction of travel). The inner product \(\langle\gamma'',\nu\times\gamma'\rangle\) extracts the component of curvature that causes the curve to turn laterally within the surface, as opposed to bending off the surface.

Here is the explanation why \(k_g\) is related to \(|\nabla_{\gamma'}\gamma'|\). Note that \(\gamma'(s)\) is unit since \(\gamma(s)\) is arc-length parametrized. It is a standard fact that then \(\gamma''(s)\) is orthogonal to \(\gamma'(s)\). Denote \(\nu\) to be the Gauss map of \(\Sigma\); then \(\gamma'(s)\) being a tangent vector to \(\Sigma\) implies \(\gamma'(s)\) and \(\nu(\gamma(s))\) are orthogonal. Therefore, for any \(s\), the vectors \(\{\gamma'(s),\nu(\gamma(s)),\nu(\gamma(s))\times\gamma'(s)\}\) form an orthonormal basis of \(\mathbb{R}^3\).

For simplicity write \(\gamma',\gamma''\) and \(\nu\) for \(\gamma'(s),\gamma''(s)\) and \(\nu(\gamma(s))\). Expressing \(\gamma''\) in this basis and recalling \(\gamma''\perp\gamma'\), we have \[\gamma''=\langle\gamma'',\nu\rangle\nu+\langle\gamma'',\nu\times\gamma'\rangle(\nu\times\gamma').\] Interpreting \(\gamma''=D_{\gamma'}\gamma'\) and projecting onto the tangent plane gives \[\nabla_{\gamma'}\gamma'=\langle\gamma'',\nu\times\gamma'\rangle(\nu\times\gamma').\] This shows \(|\nabla_{\gamma'}\gamma'|=|k_g|\) since \(\nu\times\gamma'\) is unit. Therefore \(k_g\) is a quantity whose absolute value measures how much the curve deviates from being a geodesic. By this observation, we have proved:

Proposition 4.6

Let \(\gamma(s)\) be an arc-length parametrized curve on a regular surface \(\Sigma^2\subset\mathbb{R}^3\). Then \(\gamma\) is a geodesic on \(\Sigma\) if and only if its geodesic curvature \(k_g\) on \(\Sigma\) is identically zero.

Checkpoint
Proposition 4.6 gives a clean characterization of geodesics. An arc-length parametrized curve \(\gamma(s)\) on \(\Sigma^2\subset\mathbb{R}^3\) is a geodesic if and only if:
WhyBy Proposition 4.6, the geodesic condition \(\nabla_{\gamma'}(\gamma')=0\) (for unit-speed curves) is equivalent to \(k_g\equiv 0\). A curve can have nonzero ambient curvature \(\kappa\) (e.g. great circles on a sphere) while still being geodesic, so vanishing \(\kappa\) is not required.
Remark 4.7

\(k_g\) depends on the direction of the curve \(\gamma\) and the choice of Gauss map. If \(\gamma\) and \(\tilde{\gamma}\) are the same curve but in opposite directions, say \(\tilde{\gamma}(s)=\gamma(L-s)\), then \(\tilde{\gamma}'(s)=-\gamma'(L-s)\) and \(\tilde{\gamma}''(s)=\gamma''(L-s)\). The tangent reverses sign, so \(k_g\) changes sign when the direction of the curve is reversed (with \(\nu\) fixed).

Some textbooks may define the geodesic curvature as \(\langle\gamma'',\gamma'\times\nu\rangle\), which is the negative of our definition. It is just a matter of taste, but then the integral of the geodesic curvature in the Gauss-Bonnet's Theorem should bear an opposite sign from our version.

Checkpoint
Remark 4.7 discusses the sign behavior of \(k_g\). If \(\tilde{\gamma}(s)=\gamma(L-s)\) is \(\gamma\) traversed in the opposite direction (with \(\nu\) fixed), what happens to \(k_g\)?
WhyReversing orientation replaces \(\gamma'\) by \(-\gamma'\), so \(\nu\times\gamma'\) changes sign. Meanwhile \(\gamma''\) is unchanged (second derivative of the reparametrized curve with respect to arc-length picks up a sign squared, but the remark shows directly \(\tilde{\gamma}''(s)=\gamma''(L-s)\)), so \(k_g\mapsto -k_g\). This sign-dependence is important for the oriented boundary integral in the Gauss-Bonnet theorem.
Example 4.8

Consider a longitude \(\gamma(\theta)=(\cos\theta\sin\varphi_0,\sin\theta\sin\varphi_0,\cos\varphi_0)\) on a unit sphere, with fixed \(\varphi_0\in(0,\pi)\). We want to find the geodesic curvature of \(\gamma\) in terms of \(\varphi_0\).

Since \(|\gamma'(\theta)|=\sin\varphi_0\), we reparametrize: \[\gamma(s)=\left(\cos\!\left(\frac{s}{\sin\varphi_0}\right)\sin\varphi_0,\;\sin\!\left(\frac{s}{\sin\varphi_0}\right)\sin\varphi_0,\;\cos\varphi_0\right).\] The unit normal \(\nu(\gamma(s))\) of the unit sphere at \(\gamma(s)\) is the position vector \(\gamma(s)\) itself. Therefore \[k_g(s)=\langle\gamma''(s),\gamma(s)\times\gamma'(s)\rangle.\] As \(\gamma(s)\) and \(\gamma'(s)\) are both unit with angle \(\frac{\pi}{2}\) between them, \(\gamma(s)\times\gamma'(s)\) is a unit vector tangent to a latitude. A simple sketch shows \(\gamma''(s)\) makes angle \(\varphi_0\) with \(\gamma(s)\times\gamma'(s)\), so \[k_g(s)=|\gamma''(s)|\cos\varphi_0=\frac{\cos\varphi_0}{\sin\varphi_0}.\] Here we use that \(\gamma(s)\) is a circle of radius \(\sin\varphi_0\), so \(|\gamma''(s)|=\frac{1}{\sin\varphi_0}\).

Checkpoint
In Example 4.8, the geodesic curvature of a latitude circle at colatitude \(\varphi_0\) on the unit sphere is computed to be \(k_g=\frac{\cos\varphi_0}{\sin\varphi_0}\). For which value of \(\varphi_0\) is the latitude circle a geodesic?
WhyThe geodesic curvature \(k_g=\cos\varphi_0/\sin\varphi_0\) vanishes precisely when \(\cos\varphi_0=0\), i.e., \(\varphi_0=\pi/2\). The equator is the unique latitude circle that is a geodesic on the sphere (consistent with Example 4.4: it is the circle where the radius function \(x(t)=\sin t\) has its maximum).
Exercise 4.4

Consider the surface of revolution \(\Sigma\) parametrized by \[F(t,\theta)=(x(t)\cos\theta,\,x(t)\sin\theta,\,z(t)),\] and consider the curve \(\gamma(\theta)=F(t_0,\theta)\) where \(t_0\) is fixed. Find the geodesic curvature of \(\gamma\).

Need help?