Why a curved geometry needs a flat map
In the last three guides you watched the parallel postulate fall, met the hyperbolic plane where a point off a line has infinitely many parallels, and measured triangles whose angle defect is genuinely positive. All of that was argued from the axioms. But a fair and stubborn doubt remains: how do you know such a world is not secretly self-contradictory? You cannot draw the true hyperbolic plane on paper, because paper is flat and the hyperbolic plane is not. The answer is to build a model — a concrete object, made of familiar Euclidean parts, in which every hyperbolic axiom comes true.
A model is a translation dictionary. You declare what the words 'point', 'line', and 'distance' will mean inside some Euclidean shape, then check that, under that translation, the hyperbolic axioms hold. If they do, then any contradiction in hyperbolic geometry would echo back as a contradiction in ordinary Euclidean geometry. So the model proves a relative consistency: hyperbolic geometry is exactly as trustworthy as the flat geometry you already believe. Crucially, there is not one such dictionary but several — and comparing them is the whole fun of this guide.
The Poincare disk: a world inside a circle
The most beloved model is the Poincare disk. Take the open interior of a circle — the boundary circle itself is excluded. Those interior points are your 'points'. Your 'lines' are two kinds of curve: any diameter through the center, and any arc of a circle that meets the boundary at a right angle (a perpendicular crossing). Distance is rigged so that the boundary sits infinitely far away: as you walk toward the rim, your steps look shorter and shorter to a Euclidean eye, and you never arrive. An infinite plane has been folded into a finite disk.
Now the failure of uniqueness is something you can see. Pick a hyperbolic line — say a boundary-perpendicular arc — and a point P not on it. Through P you can draw not one but infinitely many other boundary-perpendicular arcs that never touch the first one inside the disk. They are all parallels. Two of them are special: they share an endpoint with the original line on the boundary circle, just grazing it at infinity. Those are the limiting parallels from guide 2, and the angle they make with the perpendicular from P is exactly the angle of parallelism — now a thing you can point at on a picture.
The Poincare disk pays for its finiteness with a beautiful honesty about angles. It is a conformal model: the hyperbolic angle between two curves equals the ordinary Euclidean angle you measure between them with a protractor where they cross. So shapes look locally correct — a small hyperbolic square really does look square-ish. What is sacrificed is distance and size: a hyperbolic triangle drawn near the rim looks tiny to us but is, in true hyperbolic length, just as large as a fat one near the center. The disk keeps angles honest and lets size lie.
The upper half-plane: the same world, slid open
Cut the disk open and unroll it and you get the second great picture, the upper half-plane. Here the 'plane' is everything strictly above a horizontal line (the real axis), the boundary line itself excluded. The 'lines' are again two families: vertical rays shooting straight up, and semicircles whose center sits on the boundary line, so they meet it at right angles. As in the disk, distance is warped — points high up are 'close' in hyperbolic terms while a step taken just above the boundary line is hyperbolically enormous. The boundary line, plus a single point at infinity, plays the role the boundary circle did before.
Why keep two conformal models that say the same thing? Because each makes different work easy, and a clean bridge connects them. The bridge is a Mobius transformation — a map of the form z maps to (a z + b) / (c z + d) on the plane of complex numbers. There is a specific Mobius map that carries the upper half-plane exactly onto the Poincare disk, sending hyperbolic lines to hyperbolic lines and preserving every angle and every hyperbolic distance. The two models are not rivals; they are the same hyperbolic plane wearing two coats, with a dictionary translating perfectly between them.
POINCARE DISK UPPER HALF-PLANE
unit circle is the boundary the real axis is the boundary
(boundary at infinity) (boundary + one point at infinity)
lines = diameters lines = vertical rays
+ arcs _|_ boundary + semicircles centered on axis
CONFORMAL (angles true) CONFORMAL (angles true)
linked by a Mobius map: z --> (z - i) / (z + i)
upper half-plane ----> Poincare disk (angles & distances preserved)The Klein disk and the hyperboloid: straightness vs. true shape
The conformal models bend lines into arcs to keep angles true. The Klein-Beltrami disk makes the opposite bargain. Its 'points' are again the interior of a circle, but now its 'lines' are ordinary straight Euclidean chords — the plain line segments joining two boundary points. That is its gift: a hyperbolic line looks perfectly straight, so the failure of the parallel postulate becomes very direct. Through a point inside the disk, all the chords that miss a given chord (sharing no interior point with it) are exactly the parallels, and you can count the infinitely many of them at a glance.
Nothing is free. The price the Klein disk pays for straight lines is that it is not conformal: the angle you measure with a protractor between two chords is almost never the true hyperbolic angle. A hyperbolic right angle can look acute or obtuse on the page. So the Klein model is wonderful for questions about incidence and which lines cross which, and treacherous for anything involving angles. Compare the trade honestly: the disk and half-plane keep angles and bend lines; the Klein disk keeps lines and bends angles. Same world, different lie.
The fourth model refuses to flatten anything, and is the most honest of all: the hyperboloid model, also called the Minkowski or Lorentz model. The hyperbolic plane is drawn as the upper sheet of a two-sheeted hyperboloid sitting in 3D space — the surface x^2 + y^2 - z^2 = -1 with z > 0 — and a 'line' is the curve where that surface meets a flat plane through the origin. Distance is measured not with the usual dot product u . v = u_1 v_1 + u_2 v_2 + u_3 v_3 but with a sign-flipped version, u_1 v_1 + u_2 v_2 - u_3 v_3. With that one minus sign, the surface becomes a perfect, undistorted hyperbolic plane: every distance and angle is exactly right, paid for by living in three dimensions instead of two.
What the models prove, and what they don't
Step back and collect the payoff. Because all four models are constructed entirely out of ordinary Euclidean ingredients — circles, chords, complex numbers, a quadric surface — if Euclidean geometry is free of contradiction, then so is hyperbolic geometry. This is the decisive answer to the doubt we opened with, and it is why we can finally say plainly that non-Euclidean geometry never broke Euclid. It revealed a genuine choice: the parallel postulate is independent, and both branches are faithful geometries, each true to its own axioms.
Be honest about what a model does not do, though. Drawing the Poincare disk does not magically prove that the hyperbolic 'lines' really satisfy every congruence and continuity axiom — verifying that takes genuine work, measuring those warped distances with the tools of Mobius transformations and, ultimately, the calculus and metric geometry of later rungs. We have shown you the idea faithfully and given you something you can point at, but the full check that the dictionary is airtight belongs to courses past this one. The picture earns belief; it does not replace proof.
- Need angles to look right and a finite picture? Reach for the Poincare disk — conformal, bounded, lines are boundary-perpendicular arcs.
- Working with complex-number maps or symmetries? The upper half-plane and its Mobius transformations make the algebra cleanest.
- Asking which lines cross and which are parallel? The Klein disk draws lines dead straight, so incidence is obvious — just never trust its angles.
- Want distances and angles both exactly true, with no flat-page distortion at all? Climb into the hyperboloid model and accept the extra dimension.