Poincaré half-plane model

The metric of the model on the half-plane, ${\displaystyle \{\langle x,y\rangle \mid y>0\},}$ is:

${\displaystyle (ds)^{2}={\frac {(dx)^{2}+(dy)^{2}}{y^{2}}}}$

where s measures the length along a (possibly curved) line. The straight lines in the hyperbolic plane (geodesics for this metric tensor, i.e., curves which minimize the distance) are represented in this model by circular arcs perpendicular to the x-axis (half-circles whose centers are on the x-axis) and straight vertical rays perpendicular to the x-axis.

Distance calculation

The distance between two points in the half-plane model can be computed in terms of Euclidean distances in an isosceles trapezoid formed by the points and their reflection across the x-axis: a "side length" s, a "diagonal" d, and two "heights" h₁ and h₂. It is the logarithm dist(p₁, p₂) = log((s + d)²/h₁h₂)

Distance between two points can alternately be computed using ratios of Euclidean distances to the ideal points at the ends of the hyperbolic line.

Distance from the apex of a semicircle to another point on it is the inverse Gudermannian function of the central angle.

If ${\textstyle p_{1}=\langle x_{1},y_{1}\rangle }$ and ${\textstyle p_{2}=\langle x_{2},y_{2}\rangle }$ are two points in the half-plane ${\textstyle y>0}$ and ${\textstyle {\tilde {p}}_{1}=\langle x_{1},-y_{1}\rangle }$ is the reflection of ${\textstyle p_{1}}$ across the x-axis into the lower half plane, the distance between the two points under the hyperbolic-plane metric is:

$${\displaystyle {\begin{aligned}\operatorname {dist} (p_{1},p_{2})&=2\operatorname {arsinh} {\frac {\|p_{2}-p_{1}\|}{2{\sqrt {y_{1}y_{2}}}}}\\[10mu]&=2\operatorname {artanh} {\frac {\|p_{2}-p_{1}\|}{\|p_{2}-{\tilde {p}}_{1}\|}}\\[10mu]&=2\ln {\frac {\|p_{2}-p_{1}\|+\|p_{2}-{\tilde {p}}_{1}\|}{2{\sqrt {y_{1}y_{2}}}}},\end{aligned}}}$$

where ${\textstyle \|p_{2}-p_{1}\|={\sqrt {(x_{2}-x_{1})^{2}+(y_{2}-y_{1})^{2}}}}$ is the Euclidean distance between points ${\textstyle p_{1}}$ and ${\textstyle p_{2},}$ ${\textstyle \operatorname {arsinh} x=\ln {\bigl (}x+{\sqrt {x^{2}+1}}{\bigr )}}$ is the inverse hyperbolic sine, and ${\textstyle \operatorname {artanh} x={\frac {1}{2}}\ln \left((1+x)/(1-x)\right)}$ is the inverse hyperbolic tangent. This ${\textstyle 2\operatorname {arsinh} }$ formula can be thought of as coming from the chord length in the Minkowski metric between points in the hyperboloid model, ${\textstyle \operatorname {chord} (p_{1},p_{2})=2\sinh {\tfrac {1}{2}}\operatorname {dist} (p_{1},p_{2}),}$ analogous to finding arclength on a sphere in terms of chord length. This ${\textstyle 2\operatorname {artanh} }$ formula can be thought of as coming from Euclidean distance in the Poincaré disk model with one point at the origin, analogous to finding arclength on the sphere by taking a stereographic projection centered on one point and measuring the Euclidean distance in the plane from the origin to the other point.

If the two points ${\textstyle p_{1}}$ and ${\textstyle p_{2}}$ are on a hyperbolic line (Euclidean half-circle) which intersects the x-axis at the ideal points ${\textstyle p_{0}=\langle x_{0},0\rangle }$ and ${\textstyle p_{3}=\langle x_{3},0\rangle ,}$ the distance from ${\textstyle p_{1}}$ to ${\textstyle p_{2}}$ is:

$${\displaystyle \operatorname {dist} (p_{1},p_{2})=\left|\ln {\frac {\|p_{2}-p_{0}\|\|p_{1}-p_{3}\|}{\|p_{1}-p_{0}\|\|p_{2}-p_{3}\|}}\right|.}$$

Cf. Cross-ratio.

Some special cases can be simplified. Two points with the same ${\textstyle x}$ coordinate:[1]

$${\displaystyle \operatorname {dist} (\langle x,y_{1}\rangle ,\langle x,y_{2}\rangle )=\left|\ln {\frac {y_{2}}{y_{1}}}\right|=\left|\ln(y_{2})-\ln(y_{1})\right|.}$$

Two points with the same ${\textstyle y}$ coordinate:

$${\displaystyle \operatorname {dist} \left(\langle x_{1},y\rangle ,\langle x_{2},y\rangle \right)=2\operatorname {arsinh} {\frac {|x_{2}-x_{1}|}{2y}}.}$$

One point ${\textstyle \langle x_{1},r\rangle }$ at the apex of the semicircle ${\textstyle (x-x_{1})^{2}+y^{2}=r^{2},}$ and another point at a central angle of ${\textstyle \phi .}$

$${\displaystyle \operatorname {dist} (\langle x_{1},r\rangle ,\langle x_{1}\pm r\sin \phi ,r\cos \phi \rangle )={2\operatorname {artanh} }{\bigl (}{\tan {\tfrac {1}{2}}\phi }{\bigr )}=\operatorname {gd} ^{-1}\phi ,}$$

where ${\textstyle \operatorname {gd} ^{-1}}$ is the inverse Gudermannian function, and ${\textstyle \operatorname {artanh} x={\tfrac {1}{2}}\ln {\dfrac {1+x}{1-x}}}$ is the inverse hyperbolic tangent.

• Ideal points (points at infinity) in the Poincaré half-plane model are of two kinds:

• the points on the x-axis, and
• one imaginary point at ${\displaystyle y=\infty }$ which is the ideal point to which all lines orthogonal to the x-axis converge.

• Straight lines, geodesics (the shortest path between the points contained within it) are modeled by either:

• half-circles whose origin is on the x-axis
• straight vertical rays orthogonal to the x-axis

• A circle (curves equidistant from a central point) with center ${\displaystyle (x,y)}$ and radius ${\displaystyle r}$ is modeled by:

a circle with center ${\displaystyle (x,y\cosh(r))}$ and radius ${\displaystyle y\sinh(r)}$

• A hypercycle (a curve equidistant from a straight line, its axis) is modeled by either:

• a circular arc which intersects the x-axis at the same two ideal points as the half-circle which models its axis but at an acute or obtuse angle
• a straight line which intersects the x-axis at the same point as the vertical line which models its axis, but at an acute or obtuse angle.

• A horocycle (a curve whose normals all converge asymptotically in the same direction, its center) is modeled by either:

• a circle tangent to the x-axis (but excluding the ideal point of intersection, which is its center)
• a line parallel to the x-axis, in this case the center is the ideal point at ${\displaystyle y=\infty }$.

Here is how one can use compass and straightedge constructions in the model to achieve the effect of the basic constructions in the hyperbolic plane.[2] For example, how to construct the half-circle in the Euclidean half-plane which models a line on the hyperbolic plane through two given points.

Stellated regular heptagonal tiling of the model

The projective linear group PGL(2,C) acts on the Riemann sphere by the Möbius transformations. The subgroup that maps the upper half-plane, H, onto itself is PSL(2,R), the transforms with real coefficients, and these act transitively and isometrically on the upper half-plane, making it a homogeneous space.

There are four closely related Lie groups that act on the upper half-plane by fractional linear transformations and preserve the hyperbolic distance.

• The special linear group SL(2,R) which consists of the set of 2×2 matrices with real entries whose determinant equals +1. Note that many texts (including Wikipedia) often say SL(2,R) when they really mean PSL(2,R).
• The group S*L(2,R) consisting of the set of 2×2 matrices with real entries whose determinant equals +1 or −1. Note that SL(2,R) is a subgroup of this group.
• The projective special linear group PSL(2,R) = SL(2,R)/{±I}, consisting of the matrices in SL(2,R) modulo plus or minus the identity matrix.
• The group PS^*L(2,R) = S^*L(2,R)/{±I}=PGL(2,R) is again a projective group, and again, modulo plus or minus the identity matrix. PSL(2,R) is contained as an index-two normal subgroup, the other coset being the set of 2×2 matrices with real entries whose determinant equals −1, modulo plus or minus the identity.

The relationship of these groups to the Poincaré model is as follows:

• The group of all isometries of H, sometimes denoted as Isom(H), is isomorphic to PS^*L(2,R). This includes both the orientation preserving and the orientation-reversing isometries. The orientation-reversing map (the mirror map) is ${\displaystyle z\rightarrow -{\overline {z}}}$.
• The group of orientation-preserving isometries of H, sometimes denoted as Isom⁺(H), is isomorphic to PSL(2,R).

Important subgroups of the isometry group are the Fuchsian groups.

One also frequently sees the modular group SL(2,Z). This group is important in two ways. First, it is a symmetry group of the square 2x2 lattice of points. Thus, functions that are periodic on a square grid, such as modular forms and elliptic functions, will thus inherit an SL(2,Z) symmetry from the grid. Second, SL(2,Z) is of course a subgroup of SL(2,R), and thus has a hyperbolic behavior embedded in it. In particular, SL(2,Z) can be used to tessellate the hyperbolic plane into cells of equal (Poincaré) area.

The group action of the projective special linear group ${\displaystyle {\rm {PSL}}(2,\mathbb {R} )}$ on ${\displaystyle \mathbb {H} }$ is defined by

${\displaystyle {\begin{pmatrix}a&b\\c&d\\\end{pmatrix}}\cdot z={\frac {az+b}{cz+d}}={\frac {(ac|z|^{2}+bd+(ad+bc)\Re (z))+i(ad-bc)\Im (z)}{|cz+d|^{2}}}.}$

Note that the action is transitive: for any ${\displaystyle z_{1},z_{2}\in \mathbb {H} }$, there exists a ${\displaystyle g\in {\rm {PSL}}(2,\mathbb {R} )}$ such that ${\displaystyle gz_{1}=z_{2}}$. It is also faithful, in that if ${\displaystyle gz=z}$ for all ${\displaystyle z\in \mathbb {H} ,}$ then g = e.

The stabilizer or isotropy subgroup of an element ${\displaystyle z\in \mathbb {H} }$ is the set of ${\displaystyle g\in {\rm {PSL}}(2,\mathbb {R} )}$ which leave z unchanged: gz = z. The stabilizer of i is the rotation group

${\displaystyle {\rm {SO}}(2)=\left.\left\{{\begin{pmatrix}\cos \theta &\sin \theta \\-\sin \theta &\cos \theta \\\end{pmatrix}}\right|\theta \in \mathbb {R} \right\}.}$

Since any element ${\displaystyle z\in \mathbb {H} }$ is mapped to i by some element of ${\displaystyle {\rm {PSL}}(2,\mathbb {R} )}$, this means that the isotropy subgroup of any z is isomorphic to SO(2). Thus, ${\displaystyle \mathbb {H} ={\rm {PSL}}(2,\mathbb {R} )/{\rm {SO}}(2)}$. Alternatively, the bundle of unit-length tangent vectors on the upper half-plane, called the unit tangent bundle, is isomorphic to ${\displaystyle {\rm {PSL}}(2,\mathbb {R} )}$.

The upper half-plane is tessellated into free regular sets by the modular group ${\displaystyle {\rm {SL}}(2,\mathbb {Z} ).}$

The geodesics for this metric tensor are circular arcs perpendicular to the real axis (half-circles whose origin is on the real axis) and straight vertical lines ending on the real axis.

The unit-speed geodesic going up vertically, through the point i is given by

${\displaystyle \gamma (t)={\begin{pmatrix}e^{t/2}&0\\0&e^{-t/2}\\\end{pmatrix}}\cdot i=ie^{t}.}$

Because PSL(2,R) acts transitively by isometries of the upper half-plane, this geodesic is mapped into the other geodesics through the action of PSL(2,R). Thus, the general unit-speed geodesic is given by

${\displaystyle \gamma (t)={\begin{pmatrix}a&b\\c&d\\\end{pmatrix}}{\begin{pmatrix}e^{t/2}&0\\0&e^{-t/2}\\\end{pmatrix}}\cdot i={\frac {aie^{t}+b}{cie^{t}+d}}.}$

This provides a basic description of the geodesic flow on the unit-length tangent bundle (complex line bundle) on the upper half-plane. Starting with this model, one can obtain the flow on arbitrary Riemann surfaces, as described in the article on the Anosov flow.

The metric of the model on the half- space ${\textstyle \{\langle x,y,z\rangle \mid z>0\}}$ is given by

$${\displaystyle (ds)^{2}={\frac {(dx)^{2}+(dy)^{2}+(dz)^{2}}{z^{2}}}\,}$$

where s measures length along a possibly curved line. The straight lines in the hyperbolic space (geodesics for this metric tensor, i.e. curves which minimize the distance) are represented in this model by circular arcs normal to the z = 0-plane (half-circles whose origin is on the z = 0-plane) and straight vertical rays normal to the z = 0-plane.

The distance between two points ${\textstyle p_{1}=\langle x_{1},y_{1},z_{1}\rangle }$ and ${\textstyle p_{2}=\langle x_{2},y_{2},z_{2}\rangle }$ measured in this metric along such a geodesic is:

$${\displaystyle \operatorname {dist} (p_{1},p_{2})=2\operatorname {arsinh} {\frac {\|p_{2}-p_{1}\|}{2{\sqrt {z_{1}z_{2}}}}}.}$$

This model can be generalized to model an ${\displaystyle n+1}$ dimensional hyperbolic space by replacing the real number x by a vector in an n dimensional Euclidean vector space.

Notes

Sources