Pairs of Pants



One of the striking features of hyperbolic geometry is the possibility to construct shapes that are impossible in Euclidean geometry, and one of the most important such shapes is the right angled hyperbolic hexagon. Above is a tiling of the hyperbolic disk by copies of one of them.


They are useful to make pairs of pants by sewing two congruent such hexagons together to a single pair of pants, and then to continue sewing several such pants to eventually create coordinates of Teichmüller space. For that and other things, one needs to know that they actually exist, and the basic theorem here is that for any a,b,c>0 there is a unique right angled hyperbolic hexagon that has these numbers as the lengths of its odd edges.


Above is a short version of a proof of that theorem which I learned from Hermann Karcher. It employs parallel curves to geodesics in the hyperbolic plane. To see what they are, let’s switch to the upper half plane model. Geodesics (green) are vertical lines or circles orthogonal to the real axis (black).


A parallel curve (red) at distance d to such a geodesic is obtained by taking a point distance d away from it and applying hyperbolic translations along that geodesic to this point. These hyperbolic translations are simplest for the vertical geodesics, where they are just scalings with center at the end point of the geodesic. Therefore these orbits are also lines, or, in the case of circular geodesics, other circles ending at the same points as the geodesic (as follows using Möbius transformations)



Now let’s come back to the existence of right angled hexagons. Let’s start with a (blue) geodesic segment 𝛂 of length a. Two adjacent edges 𝛆 and 𝛇 begin at the end points of that segment and are orthogonal to it, but we don’t know yet how long they are. We construct parallel curves 𝛆’ and 𝛇’ to these geodesics, distance c and b away from them. Now take a geodesic 𝛅 that touches 𝛆’ and 𝛇’ in points B and C. The geodesic segment together with the perpendiculars from B to 𝛇 and C to 𝛆 completes the right angled hexagon with the prescribed edge lengths.

Exist 1

To see the existence of the touching geodesic 𝛅, it is useful to switch again to the upper half plane, where we can assume that 𝛆 and 𝛆’ are straight lines. Then the existence of 𝛅 can be seen by looking at geodesic half circles 𝛅 near the end point and touching 𝛆’ and away from  𝛇’. Increasing their radius , monotonicity implies that there is a unique such half circle that also touches 𝛇’ (from the correct side). This then also establishes the uniqueness of the right angled hexagon, up to a hyperbolic isometry.

Hyperbolic Cubes (Spheres X)

The hyperbolic plane is a geometry whose points are those of a disk, and the lines are circular arcs that meet the boundary circle of the disk at a right angle. Like so:

Squares60 01

We see that the arcs in this figure cut out regions, and the central one looks like a deflated square. However, to the hyperbolic eyes the circles are actually straight, and all regions would be considered congruent squares. There is yet another difference to Euclidean geometry: These squares have corners with 60 degree angles, instead of the traditional 90 degrees. In fact, hyperbolic people can make squares with any angle less than 90 degrees at the corners, all the way down to 0 degrees. Then the vertices lie on the boundary of the disk, and we are getting what is called an ideal square.

Ideal4 01

Of course there is also hyperbolic space, represented by a round ball. The planes are spherical shells meeting the boundary of the ball at a right angle. We can make cubes in hyperbolic space, with dihedral angles less than the traditional 90 degrees. This time, when the cube becomes ideal (i.e. when the corners lie on the boundary of the ball), the dihedral angles of the hyperbolic cube have shrunk down to 60 degrees.


Because six such cubes will fit around an edge very much like six hyperbolic 60 degree squares fir around a corner, we are able to tile all of hyperbolic space with ideal cubes. Visualizing this is a challenge, as if we just draw all the cubes, we will just see a ball, and all the effort was in vain.

But we can leave out some of the cubes. If we start with the central ideal cube, and just take those cubes that can be obtained by rotating about the edges by 180 degrees, we get after a few steps the following object, shown as a stereo pair for cross-eyed viewing.


The final image is obtained by applying more such rotations, and making the material relective.

Cubes2d left