Chapter 4 › 4.3 Gauss-Bonnet's Theorem: an intrinsic proof › 4.3.2 Global version of Gauss-Bonnet’s Theorem
Subsection

4.3.2  Global version of Gauss-Bonnet’s Theorem

Now we can use the local version of the Gauss-Bonnet's Theorem to prove its global version, which is a beautiful formula to end this course.

Theorem 4.18 (Gauss-Bonnet's Theorem: global version)

Suppose \(\Sigma^2\subset\mathbb{R}^3\) is a compact, orientable, regular surface without boundary. Then \[\oint_\Sigma K\,dS=2\pi\chi(\Sigma),\] where \(\chi(\Sigma)\) is the Euler's characteristic of \(\Sigma\), which equals \(V-E+F\) when we triangulate \(\Sigma^2\) using curved triangles.

Checkpoint
Theorem 4.18 (Gauss-Bonnet: global version) states \(\oint_\Sigma K\,dS=2\pi\chi(\Sigma)\) for a compact orientable regular surface without boundary. What does this theorem imply about a compact surface homeomorphic to a torus (\(\chi=0\))?
WhySince \(\chi(\text{torus})=0\), the Gauss-Bonnet theorem forces \(\oint K\,dS=0\). This does not mean \(K\equiv 0\) — a flat torus has \(K=0\) everywhere, but the standard \"donut\" torus embedded in \(\mathbb{R}^3\) has regions of positive and negative Gauss curvature that cancel in total. This is a striking topological constraint on the geometry.

The theorem says that even though the Gauss curvature at every point changes when you smoothly deform the surface \(\Sigma\), its total Gauss curvature does not change! Also, the LHS of the Gauss-Bonnet's formula is purely geometric, involving quantities such as Gauss curvature and surface area, whereas the RHS is purely topological. The formula bridges the two subjects together. In my opinion, it is much more beautiful and intriguing than the identity \(e^{\pi i}+1=0\).

Proof — revealed step by step

We invoke the well-known fact in topology that any such surface can be triangulated. Let \(\{\Delta_i\}_{i=1}^F\) be a collection of curved triangles on \(\Sigma^2\) such that \(\Sigma^2=\cup_i\Delta_i\), with \(\Delta_i\)'s intersecting each other only on edges, and edges intersecting only at vertices. Denote \(F\) = number of triangles. For each triangle \(\Delta_i\), denote its three exterior angles \(\{\alpha_{ij}\}_{j=1}^3\) and edges \(\{\gamma_{ij}\}_{j=1}^3\). By the local Gauss-Bonnet: \[\int_{\Delta_i}K\,dS+\sum_{j=1}^3\int_{\gamma_{ij}}k_g\,ds=2\pi-\sum_{j=1}^3\alpha_{ij}.\]

Summing over all \(i\): the left-hand sides give \(\int_{\Sigma^2}K\,dS\) (for the curvature part) and for the right-hand side \(\sum_{i=1}^F\left(2\pi-\sum_{j=1}^3\alpha_{ij}\right)=2\pi F-\sum_{i,j}\alpha_{ij}\). The geodesic curvature integrals along each edge cancel pairwise (each edge is shared by two faces with opposite orientations).

Each exterior angle \(\alpha_{ij}=\pi-\beta_{ij}\) where \(\beta_{ij}\) is the interior angle. Grouping interior angles at each vertex, their sum is \(2\pi V\). Therefore \[\sum_{i=1}^F\sum_{j=1}^3\alpha_{ij}=3\pi F-2\pi V\implies\sum_{i=1}^F\left(2\pi-\sum_{j=1}^3\alpha_{ij}\right)=2\pi V-\pi F.\] Thus \(\int_\Sigma K\,dS=2\pi V-\pi F\). Finally, each edge is adjacent to two faces and each face has three edges, giving \(3F=2E\). Therefore \[\int_\Sigma K\,dS=2\pi V-2\pi E+2\pi F=2\pi\chi(\Sigma).\quad\square\]