Beltrami–Klein model

This model made its first appearance for hyperbolic geometry in two memoirs of Eugenio Beltrami published in 1868, first for dimension n = 2 and then for general n, these essays proved the equiconsistency of hyperbolic geometry with ordinary Euclidean geometry.[1][2][3]

The papers of Beltrami remained little noticed until recently and the model was named after Klein ("The Klein disk model"). This happened as follows. In 1859 Arthur Cayley used the cross-ratio definition of angle due to Laguerre to show how Euclidean geometry could be defined using projective geometry.[4] His definition of distance later became known as the Cayley metric.

In 1869, the young (twenty-year-old) Felix Klein became acquainted with Cayley's work. He recalled that in 1870 he gave a talk on the work of Cayley at the seminar of Weierstrass and he wrote:

"I finished with a question whether there might exist a connection between the ideas of Cayley and Lobachevsky. I was given the answer that these two systems were conceptually widely separated."[5]

Later, Felix Klein realized that Cayley's ideas give rise to a projective model of the non-Euclidean plane.[6]

As Klein puts it, "I allowed myself to be convinced by these objections and put aside this already mature idea." However, in 1871, he returned to this idea, formulated it mathematically, and published it.[7]

The distance function for the Beltrami–Klein model is a Cayley–Klein metric. Given two distinct points p and q in the open unit ball, the unique straight line connecting them intersects the boundary at two ideal points, a and b, label them so that the points are, in order, a, p, q, b , so that |aq| > |ap| and |pb| > |qb|.

The hyperbolic distance between p and q is then: ${\displaystyle d(p,q)={\frac {1}{2}}\ln {\frac {\left|aq\right|\,\left|pb\right|}{\left|ap\right|\,\left|qb\right|}}}$

The vertical bars indicate Euclidean distances between the points in the model, ln is the natural logarithm and the factor of one half is needed to give the model the standard curvature of −1.

When one of the points is the origin and Euclidean distance between the points is r then the hyperbolic distance is:

${\displaystyle {\frac {1}{2}}\ln \left({\frac {1+r}{1-r}}\right)=\operatorname {artanh} r,}$

where artanh is the inverse hyperbolic function of the hyperbolic tangent.

Lines in the projective model of the hyperbolic plane

In two dimensions the Beltrami–Klein model is called the Klein disk model. It is a disk and the inside of the disk is a model of the entire hyperbolic plane. Lines in this model are represented by chords of the boundary circle (also called the absolute). The points on the boundary circle are called ideal points; although well defined, they do not belong to the hyperbolic plane. Neither do points outside the disk, which are sometimes called ultra ideal points.

The model is not conformal, meaning that angles are distorted, and circles on the hyperbolic plane are in general not circular in the model. Only circles that have their centre at the centre of the boundary circle are not distorted. All other circles are distorted, as are horocycles and hypercycles

Circles, hypercycles and horocycles

Circles in the Klein-Beltrami model of hyperbolic geometry.

While lines in the hyperbolic plane are easy to draw in the Klein disk model, it is not the same with circles, hypercycles and horocycles.

Circles (the set of all points in a plane that are at a given distance from a given point, its center) in the model become ellipses increasingly flattened as they are nearer to the edge. Also angles in the Klein disk model are deformed.

For constructions in the hyperbolic plane that contain circles, hypercycles, horocycles or non right angles it is better to use the Poincaré disk model or the Poincaré half-plane model.

Relation to the Poincaré disk model

Main article: Hyperbolic geometry § Connection between the models

Combined projections from the Klein disk model (yellow) to the Poincaré disk model (red) via the hemisphere model (blue)

The Beltrami–Klein model (K in the picture) is an orthographic projection from the hemispherical model and a gnomonic projection of the hyperboloid model (Hy) with the center of the hyperboloid (O) as its center.

Both the Poincaré disk model and the Klein disk model are models of the hyperbolic plane. An advantage of the Poincaré disk model is that it is conformal (circles and angles are not distorted); a disadvantage is that lines of the geometry are circular arcs orthogonal to the boundary circle of the disk.

The two models are related through a projection on or from the hemisphere model. The Klein model is an orthographic projection to the hemisphere model while the Poincaré disk model is a stereographic projection.

When projecting the same lines in both models on one disk both lines go through the same two ideal points. (the ideal points remain on the same spot) also the pole of the chord is the centre of the circle that contains the arc.

If P is a point a distance ${\displaystyle s}$ from the centre of the unit circle in the Beltrami–Klein model, then the corresponding point on the Poincaré disk model a distance of u on the same radius:

${\displaystyle u={\frac {s}{1+{\sqrt {1-s^{2}}}}}={\frac {\left(1-{\sqrt {1-s^{2}}}\right)}{s}}.}$

Conversely, If P is a point a distance ${\displaystyle u}$ from the centre of the unit circle in the Poincaré disk model, then the corresponding point of the Beltrami–Klein model is a distance of s on the same radius:

${\displaystyle s={\frac {2u}{1+u^{2}}}.}$

The regular hyperbolic dodecahedral honeycomb, {5,3,4}

Given two distinct points U and V in the open unit ball of the model in Euclidean space, the unique straight line connecting them intersects the unit sphere at two ideal points A and B, labeled so that the points are, in order along the line, A, U, V, B. Taking the centre of the unit ball of the model as the origin, and assigning position vectors u, v, a, b respectively to the points U, V, A, B, we have that that ‖a − v‖ > ‖a − u‖ and ‖u − b‖ > ‖v − b‖, where ‖ · ‖ denotes the Euclidean norm. Then the distance between U and V in the modelled hyperbolic space is expressed as

${\displaystyle d(\mathbf {u} ,\mathbf {v} )={\frac {1}{2}}\ln {\frac {\left\|\mathbf {v} -\mathbf {a} \right\|\,\left\|\mathbf {b} -\mathbf {u} \right\|}{\left\|\mathbf {u} -\mathbf {a} \right\|\,\left\|\mathbf {b} -\mathbf {v} \right\|}},}$

where the factor of one half is needed to make the curvature −1.

The associated metric tensor is given by[12][13]

${\displaystyle ds^{2}=g(\mathbf {x} ,d\mathbf {x} )={\frac {\left\|d\mathbf {x} \right\|^{2}}{1-\left\|\mathbf {x} \right\|^{2}}}+{\frac {(\mathbf {x} \cdot d\mathbf {x} )^{2}}{{\bigl (}1-\left\|\mathbf {x} \right\|^{2}{\bigr )}^{2}}}={\frac {(1-\left\|\mathbf {x} \right\|^{2})\left\|d\mathbf {x} \right\|^{2}+(\mathbf {x} \cdot d\mathbf {x} )^{2}}{{\bigl (}1-\left\|\mathbf {x} \right\|^{2}{\bigr )}^{2}}}}$

Partial {7,3} hyperbolic tiling of the hyperboloid as seen in Beltrami-Klein perspective.

The hyperboloid model is a model of hyperbolic geometry within (n + 1)-dimensional Minkowski space. The Minkowski inner product is given by

${\displaystyle \mathbf {x} \cdot \mathbf {y} =x_{0}y_{0}-x_{1}y_{1}-\cdots -x_{n}y_{n}\,}$

and the norm by ${\displaystyle \left\|\mathbf {x} \right\|={\sqrt {\mathbf {x} \cdot \mathbf {x} }}}$. The hyperbolic plane is embedded in this space as the vectors x with ‖x‖ = 1 and x₀ (the "timelike component") positive. The intrinsic distance (in the embedding) between points u and v is then given by

${\displaystyle d(\mathbf {u} ,\mathbf {v} )=\operatorname {arcosh} (\mathbf {u} \cdot \mathbf {v} ).}$

This may also be written in the homogeneous form

${\displaystyle d(\mathbf {u} ,\mathbf {v} )=\operatorname {arcosh} \left({\frac {\mathbf {u} }{\left\|\mathbf {u} \right\|}}\cdot {\frac {\mathbf {v} }{\left\|\mathbf {v} \right\|}}\right),}$

which allows the vectors to be rescaled for convenience.

The Beltrami–Klein model is obtained from the hyperboloid model by rescaling all vectors so that the timelike component is 1, that is, by projecting the hyperboloid embedding through the origin onto the plane x₀ = 1. The distance function, in its homogeneous form, is unchanged. Since the intrinsic lines (geodesics) of the hyperboloid model are the intersection of the embedding with planes through the Minkowski origin, the intrinsic lines of the Beltrami–Klein model are the chords of the sphere.

Both the Poincaré ball model and the Beltrami–Klein model are models of the n-dimensional hyperbolic space in the n-dimensional unit ball in Rⁿ. If ${\displaystyle u}$ is a vector of norm less than one representing a point of the Poincaré disk model, then the corresponding point of the Beltrami–Klein model is given by

${\displaystyle s={\frac {2u}{1+u\cdot u}}.}$

Conversely, from a vector ${\displaystyle s}$ of norm less than one representing a point of the Beltrami–Klein model, the corresponding point of the Poincaré disk model is given by

${\displaystyle u={\frac {s}{1+{\sqrt {1-s\cdot s}}}}={\frac {\left(1-{\sqrt {1-s\cdot s}}\right)s}{s\cdot s}}.}$

Given two points on the boundary of the unit disk, which are traditionally called ideal points, the straight line connecting them in the Beltrami–Klein model is the chord between them, while in the corresponding Poincaré model the line is a circular arc on the two-dimensional subspace generated by the two boundary point vectors, meeting the boundary of the ball at right angles. The two models are related through a projection from the center of the disk; a ray from the center passing through a point of one model line passes through the corresponding point of the line in the other model.

Wikimedia Commons has media related to Beltrami–Klein models.

• Poincaré half-plane model
• Poincaré disk model
• Poincaré metric
• Inversive geometry

• Luis Santaló (1961), Geometrias no Euclidianas, EUDEBA.
• Stahl, Saul (2007), A Gateway to Modern Geometry: The Poincare Half-Plane (2nd ed.), Jones & Bartlett Learning, ISBN 978-0-7637-5381-8
• Nielsen, Frank; Nock, Richard (2009), "Hyperbolic Voronoi diagrams made easy", 2010 International Conference on Computational Science and Its Applications, pp. 74–80, arXiv:0903.3287, doi:10.1109/ICCSA.2010.37, ISBN 978-1-4244-6461-6, S2CID 14129082