Gaussian Curvature: The Number a Surface Cannot Fake

Curvature that a 2D bug could measure

Gaussian curvature K is the single number that answers a deceptively hard question: how curved is a surface at this exact point, measured entirely from within the surface, with no reference to the 3D space it happens to sit in? At any point on a smooth surface you can find two special perpendicular directions — the directions of maximum and minimum bending — with associated principal curvatures κ₁ and κ₂ (positive if the surface curves toward its own normal, negative if it curves away). Gaussian curvature is simply their product: K = κ₁κ₂.

Sphere, plane, saddle: the three signs

A sphere of radius r curves the same way in every direction, κ₁ = κ₂ = 1/r, so K = 1/r² > 0 everywhere — small spheres are 'more curved' than large ones, which matches intuition. A flat plane has κ₁ = κ₂ = 0 everywhere, so K = 0. A saddle (a hyperbolic paraboloid, like a Pringle chip) curves up in one direction and down in the perpendicular direction, so κ₁ and κ₂ have opposite signs and K = κ₁κ₂ < 0 — negative curvature, the signature shape of hyperbolic geometry. These three signs — positive, zero, negative — are the entire classification of local surface geometry, and they are exactly why a sphere, a plane and a saddle each host their own distinct non-Euclidean geometry with their own version of the parallel postulate.

The cylinder puzzle: bent, yet flat

Roll a flat sheet of paper into a cylinder and it looks obviously curved, yet its Gaussian curvature is exactly K = 0 everywhere, same as the flat sheet it came from. The resolution is that a cylinder bends in only one principal direction — around its circumference, κ₁ = 1/r — while the direction along its axis stays perfectly straight, κ₂ = 0. Their product is zero regardless of r. This is not a quirk of the formula; it is the precise statement of why you can roll a flat piece of paper into a tube or a cone without tearing or stretching it, but you can never flatten a sphere (an orange peel, a globe) onto a table without ripping it — a sphere's K ≠ 0, and Gaussian curvature is provably preserved by any bending that does not stretch the surface.

Theorema Egregium: curvature is intrinsic

Gauss proved this preservation fact in 1827 and was proud enough of it to call it the Theorema Egregium — the 'remarkable theorem'. It says K depends only on lengths and angles measurable within the surface itself (the metric), never on how the surface happens to be embedded or bent in the surrounding 3D space. Bend a surface without stretching, tearing or compressing it — an isometry — and K is unchanged at every point, even though the surface's shape in 3D looks completely different. This single fact is why every flat map of the Earth distorts something: the globe has K = 1/r² > 0 everywhere, a flat map has K = 0, and Theorema Egregium guarantees no bending-only deformation can turn one into the other — cartography's oldest and most fundamental limitation is a direct corollary.

The torus: both signs on one surface

A torus is the richest of the five shapes here because its curvature is not constant — it changes sign as you move around it. On the outer rim, both principal directions curve the same way (like a sphere), so K > 0. On the inner rim, near the hole, one direction curves toward the centre while the other curves away, giving K < 0, saddle-like. Exactly on the top and bottom circles where the surface transitions between the two regimes, K = 0. Integrate K over the whole torus and by the Gauss–Bonnet theorem the positive and negative regions cancel exactly, giving a total of 0 — matching the torus's Euler characteristic χ = 0 from the companion article on Euler's formula, and revealing that Gaussian curvature and the purely combinatorial V − E + F count are, remarkably, two views of the same topological fact.

Why this matters beyond the demo

Gaussian curvature is not a mathematical curiosity kept in a drawer — it governs why geodesic domes need panels of more than one shape, why a pizza slice held flat droops but curls up (adding curvature in one direction stiffens it in the other, at fixed K), why sheet-metal car body panels can only be stamped into certain double-curved shapes without wrinkling, and why general relativity, which describes gravity as the curvature of 4D spacetime, generalises exactly this K to four dimensions via the Riemann tensor.

Frequently asked questions

Try it live

Everything above runs in your browser — open Gaussian Curvature Explorer and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.