Klein's Erlangen Program

By 1872 there were many geometries — Euclidean, hyperbolic, elliptic, projective, affine, conformal — and no organizing principle relating them. Felix Klein's Erlangen Program (his 1872 inaugural address at Erlangen) supplied one, with a single reframing definition:

A geometry is the study of the properties of a space that are invariant under a chosen group of transformations.

Geometry is group theory in disguise. This retroactively unifies the entire spine — each geometry of the isometry-group page is now recognized as "the invariant theory of its group" — and orders all geometries into a single hierarchy by inclusion of their groups. It is the conceptual capstone of Part E, and the reason group theory sits at the foundation of modern geometry.

References: Klein, Vergleichende Betrachtungen über neuere geometrische Forschungen (1872); Yaglom, Felix Klein and Sophus Lie; Coxeter, Introduction to Geometry; Sharpe, Differential Geometry: Cartan's Generalization of Klein's Erlangen Program.

1. The Core Idea

Fix a set $X$ (the "space") and a group $G$ of transformations of $X$. A geometric property or quantity is one preserved by every transformation in $G$ — an invariant. Two figures are "the same" in this geometry precisely when some $g\in G$ carries one to the other (they lie in the same $G$-orbit). Thus:

• The group $G$ determines the geometry. Enlarging $G$ coarsens the geometry (fewer invariants survive, more figures become "equivalent"); shrinking $G$ refines it (more invariants, finer distinctions).
• Congruence in the synthetic sense is exactly "same $G$-orbit" for $G$ the isometry group.

This is the definition that makes "what is a geometry?" a precise mathematical question with a precise answer: a group acting on a space.

2. The Hierarchy of Geometries

Different groups acting on (essentially) the plane yield a nested tower, from the most permissive (projective, largest group, fewest invariants) to the most rigid (Euclidean, smallest group, most invariants):

Each group contains the ones below it, so each geometry is a specialization (subgeometry) of the ones above:

$$\text{Euclidean}\subset \text{Similarity}\subset \text{Affine}\subset \text{Projective}.$$

Reading upward, one forgets structure: the projective plane cannot tell a circle from an ellipse (both are conics), the affine plane cannot measure angles, the similarity geometry has no absolute unit of length, and only Euclidean geometry has all of it. The parallel postulate reappears here as the affine-level fact that parallelism is a well-defined, group-invariant relation — which it is not in projective geometry, where parallels meet at points at infinity.

3. The Non-Euclidean Geometries as Subgroups

The Erlangen viewpoint slots the non-Euclidean geometries into the same scheme — they are the invariant theories of other subgroups of the projective group, singled out by a preserved conic (the Cayley–Klein construction):

So the three constant-curvature geometries are precisely the three subgroups of $PGL$ preserving a conic of each signature — imaginary, degenerate, real — exactly matching the isometry groups $O(n+1)$, $E(n)$, $O(n,1)$. The Beltrami–Klein model is nothing but hyperbolic geometry realized as the projective geometry that fixes a real conic — Klein's own construction, and the historical source of this synthesis.

4. Klein's Program and the Rest of Mathematics

The Erlangen principle reaches far beyond the classical geometries:

• Topology is the "geometry" of the group of all homeomorphisms — the coarsest useful group, whose invariants are exactly the topological properties (connectedness, genus, the Euler characteristic of Gauss–Bonnet).
• Conformal geometry is the invariant theory of angle-preserving maps — the natural home of the Poincaré models and of complex analysis.
• Special relativity is, in this exact spirit, the geometry whose group is the Lorentz/Poincaré group: physics as the invariant theory of spacetime symmetries, with the invariant interval $t^2-|\mathbf{x}{|}^2$ playing the role of the preserved structure. Klein's slogan is why the group-theoretic view of physical law is so powerful.

5. Limits: Klein and Then Cartan

The Erlangen program captures homogeneous geometries — spaces with a transitive symmetry group, where every point looks alike. But general Riemannian geometry, where curvature varies from point to point, has in general no such global group: a generic curved surface admits only the identity isometry. Klein's picture does not, by itself, cover it.

The reconciliation is Élie Cartan's theory of connections (Cartan geometries), which makes a Klein geometry the "tangent model" attached to each point of a curved space — the flat homogeneous $G/H$ that best approximates the manifold locally, glued together by a connection. Constant-curvature geometry is the case where the local Klein model is the same everywhere and the gluing is flat; general relativity's curved spacetime is a Cartan geometry modeled on Minkowski space. Thus:

$$\underset{\text{homogeneous, global group}}{\underbrace{\text{Klein (Erlangen)}}}\subset \underset{\text{local Klein model}+\text{connection}}{\underbrace{\text{Cartan}}}\supset \underset{\text{variable curvature}}{\underbrace{\text{Riemann}}}.$$

6. The Program's Verdict on the Spine

Seen through the Erlangen lens, the whole spine tells one story: the parallel postulate is a choice of which group acts on space, and the three classical geometries are the invariant theories of the three maximal groups preserving a conic. Group theory is therefore not an application of geometry but its foundation — the answer to "what makes two figures geometrically the same." The final page develops the projective geometry that sits atop this hierarchy and, via the Cayley–Klein metric, generates all three classical geometries from a single projective source.