The Gauss-Bonnet Theorem

How could a purely local quantity — how the surface bends at each point — know about a purely global property like the number of handles? That is the miracle of the Gauss-Bonnet theorem. Integrate the Gaussian curvature $K$ over any compact orientable surface without boundary, and the answer is $2\pi$ times the Euler characteristic — an integer that you can read off from any triangulation. Local data, global truth. This single equation is the prototype of modern index theory and the gateway from differential geometry into topology.

The theorem

Let $M$ be a smooth, compact, oriented $2$-dimensional Riemannian manifold without boundary, and let $K$ be its Gaussian curvature. Then

where $\chi(M)$ is the Euler characteristic. For any triangulation with $V$ vertices, $E$ edges, and $F$ faces: $\chi = V - E + F$. This value is a topological invariant — every triangulation gives the same answer, and the value is unchanged under any continuous deformation.

Euler characteristic by genus

For an orientable closed surface of genus $g$ (an $g$-holed torus): $\chi = 2 - 2g$. The cases that come up everywhere:

• Sphere ($g = 0$): $\chi = 2$, so $\int K, dA = 4\pi$.
• Torus ($g = 1$): $\chi = 0$, so $\int K, dA = 0$.
• Double torus ($g = 2$): $\chi = -2$, so $\int K, dA = -4\pi$.
• Genus-$g$ surface: $\int K, dA = 2\pi(2 - 2g) = 4\pi(1 - g)$.

Two consistency checks

Sphere of radius $R$. Constant curvature $K = 1/R^2$, area $4\pi R^2$. So $\int K, dA = (1/R^2)(4\pi R^2) = 4\pi = 2\pi \cdot 2 = 2\pi\chi$. Independent of $R$ — you can squash, stretch, or deform the sphere; the integral stays at $4\pi$.

Tetrahedron. A triangulation with $V = 4, E = 6, F = 4$. So $\chi = 4 - 6 + 4 = 2$ — matching the sphere, as expected because a tetrahedron is topologically a sphere.

Why this is astonishing

The integrand $K$ depends on the metric — on how distances are measured locally. Bend the sphere into an ellipsoid and $K$ varies from point to point; the metric is completely different. Yet the integral over the whole surface stays exactly the same, locked to a topological invariant. This is the first sighting of a vast principle: a curvature integral computes an index. The Atiyah-Singer index theorem, Chern-Gauss-Bonnet in arbitrary dimension, and the modern theory of characteristic classes are all descendants of this one equation.

The local version (with boundary)

If you allow boundary, the theorem reads $\int_M K, dA + \int_{\partial M} \kappa_g, ds + \sum \theta_i = 2\pi\chi(M)$, where $\kappa_g$ is the geodesic curvature along the boundary and $\theta_i$ are exterior angles at corners. Specialising to a geodesic triangle on a surface of constant curvature gives the classical fact that the angle sum equals $\pi + K\cdot\text{Area}$ — the formula behind every triangle-angle anomaly in spherical and hyperbolic geometry.

Pause and think: Bend a sphere so that one hemisphere becomes very flat ($K\approx 0$). What must happen to the curvature on the other hemisphere? Why does Gauss-Bonnet force the answer, even though the integrand is purely local?

Try it

• Predict first: what is $\int K,dA$ for a triple torus ($g = 3$)? Verify using $\chi = 2 - 2g$.
• For an icosahedron (a triangulation of the sphere with $V = 12, E = 30, F = 20$): compute $\chi$ and confirm it matches the topological sphere.
• True or false: a compact orientable surface admitting a metric with $K > 0$ everywhere must be homeomorphic to a sphere. Justify with Gauss-Bonnet.
• Use the local version to compute the angle sum of a spherical triangle on the unit sphere with side lengths given by three great-circle arcs forming a triangle of area $A$. The answer should be $\pi + A$.

A trap to watch for

Gauss-Bonnet computes total curvature, not signed integrals split over regions. A torus has $\int K,dA = 0$, but this does not mean $K = 0$ everywhere on the torus — in fact $K > 0$ on the outer half and $K < 0$ on the inner half, and the two regions cancel exactly. Students sometimes try to argue "since the integral is zero, the surface is flat," which is wrong. The integral being zero is a topological constraint on how positive and negative curvature must balance; it does not banish curvature point-by-point.

What you now know

You can state the Gauss-Bonnet theorem precisely, compute Euler characteristics from triangulations or genus, and use the theorem to constrain the total curvature of any closed orientable surface. The chapter's thread — that local curvature accumulates into global topology — is now closed. The next chapter steps back from differential geometry to the parent question: which axiomatic geometries are even possible, and how does curvature classify them?

References

• Garrity, T. (2002). All the Mathematics You Missed: But Need to Know for Graduate School. Cambridge University Press, ch. 7.
• do Carmo, M. P. (2016). Differential Geometry of Curves and Surfaces (2nd ed.). Dover, ch. 4.
• Pressley, A. (2010). Elementary Differential Geometry (2nd ed.). Springer, ch. 13.
• Spivak, M. (1999). A Comprehensive Introduction to Differential Geometry (3rd ed., Vol. 3). Publish or Perish, ch. 6.
• Misner, C. W., Thorne, K. S., Wheeler, J. A. (2017). Gravitation. Princeton University Press, ch. 11 (Gauss-Bonnet generalisations to higher dimensions).