Riemannian metrics on a coordinate chart

A single chart carries a metric tensor $g(x,y)$ — a positive‑definite symmetric matrix at every point. The teal ellipses are the (scaled) infinitesimal unit balls it defines; drag the two handles to compute the geodesic between them. drag the endpoints · hover to probe the metric.

drag the two endpoints to move the geodesic

Metric at cursor

move the cursor over the chart

Geodesic vs. straight segment

drag an endpoint

Controls

unit ball $\{v: g_p(v,v)\le\varepsilon^2\}$ numerical geodesic analytic (check) point $p$ point $q$

The mathematics on display

A Riemannian metric assigns to each point $p$ of the chart a positive‑definite symmetric bilinear form $g_p$ on the tangent space. In the coordinates $(x,y)$ it is the matrix $$g(x,y)=\begin{pmatrix} E & F\\ F & G\end{pmatrix},\qquad E>0,\quad EG-F^2>0,$$ and the length of an infinitesimal displacement is the line element $$ds^2 = E\,dx^2 + 2F\,dx\,dy + G\,dy^2 = \sum_{i,j} g_{ij}\,dx^i\,dx^j.$$ The norm of a tangent vector $v$ at $p$ is $\lVert v\rVert_g=\sqrt{v^\top g(p)\,v}$.

The infinitesimal unit ball (Tissot indicatrix)

At a point $p$ the metric ball of radius $\varepsilon$ is the set of tangent vectors $$B_\varepsilon(p)=\bigl\{\,v\in T_pM : g_p(v,v)\le\varepsilon^2\,\bigr\},$$ which is an ellipse because $g_p$ is positive definite. Diagonalising $g_p = \lambda_1\,u_1u_1^\top+\lambda_2\,u_2u_2^\top$ with orthonormal eigenvectors $u_1,u_2$, the ellipse has semi‑axes $$a_i=\frac{\varepsilon}{\sqrt{\lambda_i}}\quad\text{along }u_i.$$ Where the metric is large the ball is small: a short coordinate step already costs a lot of metric length. Each ellipse is drawn centred on its grid point and enlarged by a display factor for visibility — it is a picture of the tangent space, not a region of the chart.

Geodesics

The length and energy of a curve $\gamma:[0,1]\to M$ are $$L[\gamma]=\int_0^1\!\sqrt{g_{\gamma}(\dot\gamma,\dot\gamma)}\,dt,\qquad E[\gamma]=\tfrac12\int_0^1\! g_{\gamma}(\dot\gamma,\dot\gamma)\,dt.$$ A geodesic is a critical point of $E$; equivalently it satisfies $$\ddot x^{\,k}+\Gamma^{k}_{ij}\,\dot x^{\,i}\dot x^{\,j}=0,\qquad \Gamma^{k}_{ij}=\tfrac12\,g^{kl}\!\left(\partial_i g_{jl}+\partial_j g_{il}-\partial_l g_{ij}\right).$$ Energy minimisers are automatically parametrised at constant speed. This page solves the boundary‑value problem $\gamma(0)=p,\ \gamma(1)=q$ directly: the curve is discretised into $N$ segments and the discrete energy $E=\tfrac{N}{2}\sum_i\Delta_i^\top g(m_i)\Delta_i$ (with midpoints $m_i$) is minimised with fixed endpoints. The stiff second‑difference part is handled implicitly through a block‑tridiagonal solve while the metric‑gradient force $\tfrac12(\partial g)(\dot\gamma,\dot\gamma)$ is applied explicitly, with a line search guaranteeing the energy decreases. For the classical charts the result is checked live against the known closed‑form geodesic (dashed).

Gaussian curvature $K$ is computed from $g$ by the Brioschi formula (numerical derivatives). The classical charts below have constant $K\in\{0,-1,+1\}$; the surface‑graph chart has varying $K$.