Curvature and the Three Geometries

Chapter 8: Models of Geometry

Learning objectives

Classify the three classical model geometries by the sign of Gaussian curvature
Use the angular-defect/excess formula to determine curvature from a single geodesic triangle
State the uniformization theorem for compact orientable surfaces
Preview Thurston's geometrization in dimension three

One number — the sign of $K$ — partitions geometry into three universes. Sections 8.1-8.3 introduced them separately; this section ties them together with the curvature classification and previews how the same logic extends to higher dimensions. The deeper point is that geometry and topology are not independent: the genus of a closed surface forces a sign on the Gaussian curvature it can admit, and vice versa. This is the entry point to Thurston's geometrization program in dimension three — the geometric heart of twenty-first-century low-dimensional topology.

(Live curvature-classification widgets and uniformization animations are in development. The classifications below are paper-friendly given the previous three sections.)

The three-way classification

Each constant-curvature model surface is the unique simply-connected complete surface with the given $K$ :

$K = 0$ — Euclidean plane $\mathbb{R}^2$ . Angle sum of every triangle is $\pi$ . Exactly one parallel through a given external point. Distance formula $d = \sqrt{(\Delta x)^2 + (\Delta y)^2}$ . Similar triangles exist at every scale.
$K > 0$ — sphere (or projective plane). Angle sum strictly more than $\pi$ ; excess equals $\text{Area}/R^2$ . No parallel lines — every pair of geodesics meets. Triangles bounded above in size.
$K < 0$ — hyperbolic plane (Poincare disk or upper half-plane). Angle sum strictly less than $\pi$ ; defect equals $\text{Area}$ on $K = -1$ . Infinitely many parallels. Triangles bounded above in area but unbounded in side length.

One formula to detect curvature

Measure the interior angles of any small geodesic triangle on a constant-curvature surface and compute

K = \dfrac{(\alpha + \beta + \gamma) - \pi}{\text{Area}}

The sign and magnitude of the right-hand side are exactly the Gaussian curvature. On a variable-curvature surface, the same formula in the limit (triangle shrinking to a point) gives the local $K$ — the local Gauss-Bonnet theorem.

The uniformization theorem

Globally, the trichotomy is enforced by the uniformization theorem (Poincare, Koebe, 1907): every compact orientable surface admits a metric of constant Gaussian curvature, and the sign is determined by the Euler characteristic:

$\chi > 0$ (sphere, $g = 0$ ): admits $K > 0$ . Geometry is spherical / elliptic.
$\chi = 0$ (torus, $g = 1$ ): admits $K = 0$ . Geometry is flat Euclidean.
$\chi < 0$ (genus $g \geq 2$ ): admits $K < 0$ . Geometry is hyperbolic.

This is a remarkable result: topology (a discrete invariant) selects a continuous geometry up to scale. Every closed surface lives in exactly one of three geometric universes, and the topology of the surface decides which one.

Higher dimensions: Thurston's geometrization

In dimension three the story branches further. Thurston's geometrization conjecture, proved by Grigori Perelman in 2002-2003 using Ricci flow, says every closed orientable $3$ -manifold can be cut into pieces along spheres and tori, and each piece admits one of eight model geometries: the constant-curvature $S^3, \mathbb{R}^3, \mathbb{H}^3$ analogues plus five mixed geometries ( $S^2 \times \mathbb{R}$ , $\mathbb{H}^2 \times \mathbb{R}$ , $\widetilde{SL_2(\mathbb{R})}$ , Nil, Sol). The hyperbolic geometry $\mathbb{H}^3$ is the generic case — most closed $3$ -manifolds turn out to be hyperbolic. This is the geometric machinery behind Perelman's proof of the Poincare conjecture: any closed simply-connected $3$ -manifold is necessarily $S^3$ .

Why is the trichotomy so robust?

It is forced by the structure of the orthogonal group and the way the metric tensor interacts with parallel transport. In any dimension, the curvature tensor sits in a representation of the orthogonal group, and the constant-curvature case corresponds to a one-dimensional sub-representation. The sign of that single scalar — positive, zero, negative — classifies the geometry. The same logic gives the trichotomy in every dimension $n$ , with $\mathbb{H}^n, \mathbb{R}^n, S^n$ as the three model geometries of constant curvature.

Where this shows up

- **Cosmology and the geometry of the universe:** The large-scale geometry of the universe is one of the three constant-curvature cases — closed (

K > 0

), flat (

K = 0

), or open (

K < 0

). The Planck mission (2018) measured the spatial curvature parameter

\Omega_K = -0.001 \pm 0.002

: the universe is observationally consistent with flat geometry, but the error bars still allow either slight curvature direction. - **The Poincare conjecture and Perelman:** Perelman proved that every closed simply-connected

3

-manifold is the

3

-sphere. The proof used Hamilton's Ricci flow to evolve any starting metric toward one of the eight geometric models, with surgeries handling the singularities. The geometrization conjecture — the geometric heart of the result — is exactly the curvature-classification story extended to dimension three. - **Neural networks on manifolds:** Riemannian deep learning chooses an embedding space — flat (standard neural networks), spherical (rotation-invariant tasks like 3D point cloud classification), or hyperbolic (tree-structured data like word hierarchies). The choice of curvature affects the geometry of optimization: gradient descent on

\mathbb{H}^n

uses different update rules than on

\mathbb{R}^n

. - **String theory compactifications:** Six "extra" dimensions of string theory must curl up on a compact manifold. Calabi-Yau manifolds — complex three-folds with vanishing Ricci curvature — are the standard choice, and their classification has been deeply influenced by Thurston-style geometric ideas in dimension six. - **Architectural shells and form-finding:** Engineers designing tensile shell roofs (e.g., the Munich Olympic Stadium, Frei Otto 1972) choose surfaces with prescribed curvature distribution to handle wind and snow loads optimally. The trichotomy — flat, spherical, or saddle — determines structural behaviour: flat panels are sensitive to local buckling; spherical caps and saddles distribute loads more uniformly.

Pause and think: A compact orientable surface admits a constant-curvature metric whose sign matches the sign of $\chi$ . Why does that imply, immediately, that no compact surface with $\chi = 0$ (i.e. a torus) can have a metric of constant non-zero curvature? (Hint: integrate the curvature and use Gauss-Bonnet.)

Try it

Predict first: a hiker on an unknown surface measures a geodesic triangle and finds angles $65^\circ, 70^\circ, 50^\circ$ . Is the local curvature positive, zero, or negative? Justify.
A compact orientable surface has genus $g = 5$ . Apply the uniformization theorem to identify the sign of the curvature on its constant-curvature metric, then use Gauss-Bonnet to compute the total curvature.
True or false: a flat torus (a torus with $K = 0$ everywhere) can be isometrically embedded in $\mathbb{R}^3$ . (This was an open problem for decades and was finally resolved by Nash (1954) and Kuiper (1955) for the existence proof; the Hévéa project (Borrelli, Jabrane, Lazarus, Thibert, 2012) for the explicit visualization.)
List the eight Thurston model geometries in dimension three, classifying each as constant-curvature, product, or twisted product.

A trap to watch for

The uniformization theorem says every compact orientable surface admits a constant-curvature metric; it does not say every metric on the surface is constant-curvature. A typical metric on the genus- $2$ surface has variable $K$ ; the theorem only guarantees the existence of a single special metric, the so-called uniformizing metric, in which $K$ is constant. Equivalent confusion arises with the Earth: the Earth's surface has variable curvature (mountains, valleys), but it admits a constant-curvature metric — the ideal sphere — that is the canonical reference for cartography.

What you now know

You can classify the three model geometries by the sign of Gaussian curvature, compute $K$ from the angular defect of a geodesic triangle, state the uniformization theorem connecting topology to constant-curvature geometry, and preview Thurston's geometrization in dimension three. The full chapter is now in your toolkit: from local bending of plane curves all the way to the eight model geometries that classify $3$ -manifolds. The next chapter (Complex Analysis) pivots from differential geometry of real surfaces to the analytic geometry of $\mathbb{C}$ to a different flavour of mathematics — you will see, in chapter 9 onward, that combinatorial structures and geometric ones share more than initially meets the eye.

References

Garrity, T. (2002). All the Mathematics You Missed: But Need to Know for Graduate School. Cambridge University Press, ch. 8.
Greenberg, M. J. (2008). Euclidean and Non-Euclidean Geometries (4th ed.). W. H. Freeman, ch. 12 (the trichotomy and uniformization).
Stillwell, J. (2008). Mathematics and Its History (3rd ed.). Springer, ch. 19 and 25 (Riemann's geometry and Thurston's programme).
Coxeter, H. S. M. (1969). Introduction to Geometry (2nd ed.). Wiley, ch. 15 (the three classical geometries).
Penrose, R. (2004). The Road to Reality. Knopf, ch. 27-28 (cosmological topology and dimension three).