I started a Mathematica package to performs computations of Riemannian geometry. As an exercise in using it, I present here the computation of the curvature of a 2-sphere embedded in the euclidean space. The package is in development and its content is not meant to be exhaustive. This suffices however to perform most common computations on (pseudo-)Riemannian real manifolds. It is available from its own GitHub repository along with the notebook for the computation.

For something more general, I recommend you take a look at SageManifolds.

Start by defining a background 3-dimensional euclidean space by providing a set of coordinates ${x,y,z}$ and a metric in the form of a squared line-element: \begin{equation} ds^2=dx^2+dy^2+dz^2~. \end{equation} \begin{align} g=\left( \begin{array}{ccc} 1 & 0 & 0 \\
0 & 1 & 0 \\
0 & 0 & 1 \\
\end{array} \right)~. \end{align}

One way to embed in the 2-sphere of radius $R$ is by imposing that \begin{equation} \Phi \equiv x^2+y^2+z^2 - R^2 = 0~, \end{equation} with $R>0$. One parametrisation satisfying this constraint is \begin{align} x&=R \sin{\theta}\cos{\phi}~,\\
y&=R \sin{\theta}\sin{\phi}~,\\
z&=R \cos{\theta}~, \label{eq:mapSigma} \end{align} with ${\theta, \phi}$ taking up the role of local ordinates on the 2-sphere thus described which we will denote $\Sigma$ for simplicity.

The components of the one-form normal to $\Sigma$ are obtained as follows \begin{equation} n_\alpha = \frac{\partial \Phi}{\partial x^\alpha}~. \end{equation} Once it has been normalised, this turns out to be

The general expression of the projector onto $\Sigma$ is \begin{equation} P_{\alpha\beta}=g_{\alpha\beta}-n_\alpha n_\beta~. \end{equation} Here this gives:

\begin{align} \left( \begin{array}{ccc} \frac{y^2+z^2}{x^2+y^2+z^2} & -\frac{x y}{x^2+y^2+z^2} & -\frac{x z}{x^2+y^2+z^2} \\
-\frac{x y}{x^2+y^2+z^2} & \frac{x^2+z^2}{x^2+y^2+z^2} & -\frac{y z}{x^2+y^2+z^2} \\
-\frac{x z}{x^2+y^2+z^2} & -\frac{y z}{x^2+y^2+z^2} & \frac{x^2+y^2}{x^2+y^2+z^2} \\
\end{array} \right) \end{align}

Finally, the extrinsic curvature is the Lie derivative of this tensor along the normal vector defined above:

\begin{equation} K_{\alpha\beta}=-\frac{1}{2}L_n P_{\alpha\beta}~. \end{equation} Here this gives:

\begin{align} \left( \begin{array}{ccc} -\frac{y^2+z^2}{\left(x^2+y^2+z^2\right)^{3/2}} & \frac{x y}{\left(x^2+y^2+z^2\right)^{3/2}} & \frac{x z}{\left(x^2+y^2+z^2\right)^{3/2}} \\
\frac{x y}{\left(x^2+y^2+z^2\right)^{3/2}} & -\frac{x^2+z^2}{\left(x^2+y^2+z^2\right)^{3/2}} & \frac{y z}{\left(x^2+y^2+z^2\right)^{3/2}} \\
\frac{x z}{\left(x^2+y^2+z^2\right)^{3/2}} & \frac{y z}{\left(x^2+y^2+z^2\right)^{3/2}} & -\frac{x^2+y^2}{\left(x^2+y^2+z^2\right)^{3/2}} \\
\end{array} \right) \end{align}

The mean curvature is given by the trace of this tensor and here is found to be \begin{equation} \text{Tr}~K=-\frac{2}{R} \label{eq:trK} \end{equation}

All the above was done using only vectors (and one-forms) belonging to the (co-)tangent bundle of the surrounding $\mathbb{R}^3$ without any explicit reference to the local coordinates on $S^2$. The extrinsic curvature also has a representation on $S^2$.

From (\ref{eq:mapSigma}), one constructs the vielbein: \begin{equation} e^\alpha_a=\frac{\partial x^\alpha}{\partial y^a}~, \end{equation} where $x^\alpha$ denotes the coordinates on $\mathbb{R}^3$, and $y^a$ are the coordinates on $S^2$, . In the matrix form, this reads on our case: \begin{align} \left( \begin{array}{cc} R \cos (\theta ) \cos (\phi ) & -R \sin (\theta ) \sin (\phi ) \\
R \cos (\theta ) \sin (\phi ) & R \cos (\phi ) \sin (\theta ) \\
-R \sin (\theta ) & 0 \\
\end{array} \right) \end{align}

The induced metric on $\Sigma$ can then be computed with \begin{equation} \gamma_{ab} = e^\alpha_ae^\beta_bg_{\alpha\beta}~, \label{eq:inducedgamma} \end{equation} which here gives: \begin{align} \left( \begin{array}{cc} R^2 & 0 \\
0 & R^2 \sin ^2(\theta ) \\
\end{array} \right) \end{align} Applying a transformation similar to (\ref{eq:inducedgamma}) to the extrinsic curvature gives \begin{align} \left( \begin{array}{cc} -R & 0 \\
0 & -R \sin ^2(\theta ) \\
\end{array} \right) \end{align} and so we recover (\ref{eq:trK}) by taking the trace ($\gamma^{ab}K_{ab}$).


Leave a Comment