the Gauss-Bonnet theorem
/ GOWSS bo-NAY /
Here is a fact that should feel almost too good to be true. Add up all the curvature spread across a closed surface like a sphere or a doughnut, and the grand total does not depend on the surface's exact shape at all — it depends only on how many holes the surface has. Dent the sphere, stretch it, make it lumpy: the total curvature stays fixed. The Gauss-Bonnet theorem is the precise law tying local curvature (a metric, geometric quantity) to global topology (a hole-counting, shape-blind quantity).
There are two levels. The local (regional) version: for a region R on a surface bounded by a piecewise-smooth curve, the integral of the Gaussian curvature K over R, plus the integral of the geodesic curvature k_g along the boundary, plus the sum of the exterior turning angles at any corners, equals 2*pi. In symbols: (integral of K over R) + (integral of k_g along the boundary) + (sum of exterior angles) = 2*pi. The global version: for a CLOSED orientable surface S without boundary, the boundary terms vanish and you get the integral of K over all of S = 2*pi*chi(S), where chi = V - E + F is the Euler characteristic — a pure topological invariant (chi = 2 for a sphere, 0 for a torus, 2 - 2g for a surface with g holes). So the total Gaussian curvature is quantized by topology.
This is one of the most beautiful theorems in mathematics because it forces geometry and topology to agree. It generalises the schoolbook fact that a triangle's angles sum to 180 degrees: on a curved surface the angle sum of a geodesic triangle exceeds (or falls short of) pi by exactly the total curvature inside it — angle excess on a sphere (K > 0), angle deficit on a saddle (K < 0). It explains why you cannot give a sphere a flat metric (that would need total curvature 0, but topology demands 2*pi*2 = 4*pi), and why a torus CAN be made flat (chi = 0 allows total curvature 0 — hence flat-torus video-game worlds). One honest scope note: the theorem needs the surface to be compact and (for the clean global form) without boundary and orientable; and chi counts holes, not size — a tiny torus and a vast one have the same total curvature 0, because topology, not geometry, sets the sum.
Take a geodesic triangle on a sphere of radius R, with all three sides arcs of great circles. Gauss-Bonnet (with k_g = 0 along geodesic sides) reduces to: (sum of the three angles) - pi = (integral of K over the triangle) = (1/R^2)(area). So the angles overshoot 180 degrees by exactly area/R^2. A triangle covering one octant of the sphere (one-eighth of its surface) has three right angles, summing to 270 degrees — a 90-degree excess — vividly impossible on a flat plane. For the whole sphere, integral of K = (1/R^2)(4*pi*R^2) = 4*pi = 2*pi*chi with chi = 2.
A geodesic triangle's angle excess equals its enclosed total curvature.
The total Gaussian curvature of a closed surface depends ONLY on its topology (number of holes), not its size or shape: a golf-ball-bumpy sphere and a perfect sphere both integrate to 4*pi. Bending or denting moves curvature around but can never change the total — the Euler characteristic fixes it.