Hyperbolic Geometry

Hyperbolic geometry is the geometry obtained from the neutral axioms by denying the parallel postulate in the one way neutral geometry permits: through a point off a line there are infinitely many parallels, not one. It is the acute-angle case of the Saccheri hypotheses, the geometry of constant negative curvature $K<0$, and the one Bolyai and Lobachevsky founded (history). This page develops its synthetic content — parallels, defect, area, trigonometry — treating it as an abstract axiomatic system. That the system is consistent is proved separately by the models; here we take consistency for granted and explore what the hyperbolic plane is like.

References: Greenberg, Euclidean and Non-Euclidean Geometries, chs. 6–7; Stillwell, Sources of Hyperbolic Geometry; Anderson, Hyperbolic Geometry; Thurston, Three-Dimensional Geometry and Topology, ch. 2.

1. The Hyperbolic Axiom

Replace Playfair's axiom with its negation, keeping every other neutral axiom:

Hyperbolic Parallel Axiom. There exist a line $\ell$ and a point $P\notin >\ell$ such that at least two distinct lines through $P$ fail to meet $\ell$.

By the homogeneity of the plane this "at least two somewhere" propagates to "in fact infinitely many, everywhere." Because all the neutral theorems still hold, hyperbolic geometry inherits the Saccheri–Legendre bound: every triangle has angle sum strictly less than $\pi$.

2. Parallels, Ultraparallels, and the Angle of Parallelism

The single Euclidean notion "parallel" splits into three distinct relations.

Fix a point $P$ at perpendicular distance $p$ from a line $\ell$, with foot $Q$. Rotate a ray at $P$ from the perpendicular $PQ$ outward. There is a critical angle $\Pi (p)$ at which the ray just stops meeting $\ell$:

• rays making an angle $<\Pi (p)$ with $PQ$ cross $\ell$;
• the two rays at exactly $\Pi (p)$ are the limiting (asymptotic) parallels — they approach $\ell$ asymptotically without meeting;
• rays making an angle $>\Pi (p)$ are ultraparallel — they diverge from $\ell$ and share a common perpendicular.

The function $\Pi (p)$ is Lobachevsky's angle of parallelism, given by the beautiful formula

$$\tan \frac{\Pi (p)}{2}=e^{-p/R},$$

where $R$ is the curvature radius ($K=-1/R^2$). Read off its limits:

• As $p\to 0$ (or $R\to \infty$), $\Pi (p)\to \pi /2$: the parallel becomes unique and perpendicular — the Euclidean limit is recovered.
• As $p\to \infty$, $\Pi (p)\to 0$: distant lines have a vast fan of parallels.

An absolute unit of length. Because $\Pi$ is a specific function of distance, an angle determines a length and vice versa: the equation $\tan (\Pi /2)=e^{-p/R}$ lets you convert between them. Hyperbolic geometry therefore has a natural length scale $R$ built in — unlike Euclidean geometry, which is scale-invariant. This is the deep reason there are no similar-but-unequal figures (§4).

3. Defect, Area, and the Absence of Rectangles

The angle defect $\delta (△)=\pi -(\alpha +\beta +\gamma )$ is now strictly positive, and it measures area exactly.

Defect–Area Theorem (Gauss, Bolyai, Lobachevsky). In the hyperbolic plane of curvature $K=-1/R^2$, $$\mathrm{Area}(△)=R^2\delta (△)=R^2(\pi -\alpha -\beta -\gamma ).$$

Consequences, each sharply non-Euclidean:

• Bounded area. Since $\delta <\pi$, every triangle has area less than $\pi R^2$ — there is a maximum possible triangle area, no matter how long the sides. Triangles cannot be arbitrarily large in area.
• Ideal triangles. Letting all three vertices recede to infinity (all angles $\to 0$) gives an ideal triangle of maximal area $\pi R^2$, with three asymptotically-parallel sides and zero angles.
• No rectangles. A quadrilateral with four right angles would have angle sum $2\pi$, i.e. defect $-\pi <0$, impossible. The existence of rectangles is a parallel-postulate equivalent; their non-existence is forced.
• The Saccheri quadrilateral's summit angles are acute, confirming the acute hypothesis.

4. No Similar Triangles: AAA is Congruence

AAA Congruence. In hyperbolic geometry, two triangles with equal corresponding angles are congruent.

Why. Equal angles mean equal defect, hence (by §3) equal area; combined with the angle data this pins down the sides. Contrast Euclidean AAA similarity, where equal angles leave a free scale factor. In the hyperbolic plane the scale factor is not free — size is determined by shape. There are no scale models, no blueprints, no magnification. This is the geometric face of the natural length scale $R$ noted in §2.

5. Curves of Constant Distance

Hyperbolic geometry has three kinds of "circle-like" curve, where Euclidean geometry has only one:

• Circles — loci equidistant from a point (a genuine centre). Circumference grows exponentially with radius: $C(r)=2\pi R\sinh (r/R)$, not $2\pi r$.
• Horocycles — limits of circles whose centres recede to infinity along a pencil of asymptotic parallels; "circles of infinite radius." Their centre is an ideal point on the boundary at infinity.
• Hypercycles (equidistant curves) — loci at a fixed distance from a line, on one side. Crucially, a hypercycle is not a line — this is the failure of parallel-postulate equivalent #7 (equidistance).

The exponential growth $C(r)=2\pi R\sinh (r/R)$ expands, for small $r$, as $2\pi r(1+\frac{r^2}{6R^2}+\cdots )$: circumferences exceed the Euclidean $2\pi r$, the signature of negative curvature (the plane has "more room" as you go out — the reason hyperbolic space so naturally embeds infinite trees and drives the exponential expansion familiar from tilings by the models).

6. Hyperbolic Trigonometry

The metric relations of a right triangle (legs $a,b$; hypotenuse $c$; opposite angles $\alpha ,\beta$) replace the circular functions of Euclidean trig with hyperbolic ones (setting $R=1$ for brevity):

$$\cosh c=\cosh a\cosh b(\text{hyperbolic Pythagoras}),$$ $$\sinh a=\sinh c\sin \alpha ,\tanh b=\tanh c\cos \alpha ,\cosh c=\cot \alpha \cot \beta .$$

The general hyperbolic law of cosines reads

$$\cosh c=\cosh a\cosh b-\sinh a\sinh b\cos \gamma .$$

Expanding $\cosh x=1+\frac{x^2}{2}+\cdots$ and $\sinh x=x+\cdots$ to second order collapses every formula to its Euclidean counterpart — e.g. hyperbolic Pythagoras $\to a^2+b^2=c^2$. Reinstating $R$, the corrections are $O(1/R^2)=O(K)$: Euclidean geometry is the infinitesimal / flat limit of the hyperbolic plane, exactly as anticipated in euclidean-results.md §1.3.

7. What Hyperbolic Geometry Is (and Where It Lives)

Synthetically, hyperbolic geometry is neutral geometry $+$ the hyperbolic parallel axiom: an internally coherent world of infinitely-many parallels, positive defect, bounded-area triangles, and no similar figures. Two questions remain, answered elsewhere in the spine:

1. Is it consistent? Yes — the Beltrami–Klein, Poincaré, and hyperboloid models realize it inside Euclidean/real-number geometry, proving relative consistency and the independence of the parallel postulate.
2. What is its curvature, precisely? It is the constant-negative-curvature case $K=-1/R^2$ of Riemann's metric geometry; the defect–area law of §3 is the two-dimensional Gauss–Bonnet theorem, and its classification among the space forms is in constant-curvature.md.

Its isometry group is $O(2,1)$ (equivalently $PSL(2,\mathbb{R})$ acting on the half-plane model), a group that reappears in physics as the Lorentz group — see isometry-groups.md and SR/group/lorentz-poincare.md. The relativistic velocity space is literally a hyperbolic space, and the rapidity that adds linearly under boosts is hyperbolic arc length.