An Investigation into the Properties of Hyperbolic
Circles
Anton Bubna-Litic *
I Introduction
Geometry has always been a well-studied field. Significant progress
was made by the Ancient Greeks—specifically by Euclid, who
postulated that two parallel lines will never meet. Many mathematicians
attempted to prove this postulate and failed. But, through their work,
hyperbolic geometry was born in the 18th century.
There are now three types of geometry: Euclidean, Spherical and
Hyperbolic. Euclidean geometry works on a space with no curvature,
x=0, and parallel lines never meet. Spherical geometry works on a space
with positive curvature, x > 0, and all infinite lines meet. Finally,
hyperbolic geometry works on a space with negative curvature, x < 0,
and infinitely non-intersecting lines may be drawn in the Hyperbolic
plane.
This essay focuses on the hyperbolic plane. There are many ways of
modelling the hyperbolic plane, but this essay will only use two models.
Both models have a line, which is defined to be infinity, and so any line
that meets this `infinity line' is infinitely long. What makes hyperbolic
geometry so interesting is that if you move a line segment around the
hyperbolic plane but do not change its Euclidean length, the hyperbolic
length of that segment changes depending on its position on the plane.
What’s more, the closer it is to the `infinity line', the larger it is.
The first of these models is the Poincarѐ Disc Model ( ). This model is
a unit circle with the circumference being infinity, and lines meet the
circumference at right angles. The second model is the Upper Half
Plane model, which is an upper half plane with the bottom defined to be
infinity, and the lines are semicircles, which intersect the bottom at right
angles as well as vertical lines.
Before I begin this essay I will briefly outline some background
knowledge. Firstly, a metric is a function on a set used to measure
*
Anton Bubna-Litic is in his second year of a Bachelor of Philosophy (Science) degree at
the Australian National University. He is a current resident of Bruce Hall.
2 Cross-sections | Volume V 2009
distances, and it is denoted by d. For example, in the Euclidean plane
the metric is given by the distance between two points:
A metric space is a space with a metric. Now an isometry of a metric
space is a continuous function, which is bijective, has a continuous
inverse, and preserves distances. A special group of isometries in the
hyperbolic plane is Möb( ) which takes Hyperbolic lines and circles to
Hyperbolic lines and circles.
Finally we need to know about what the distances in the hyperbolic
plane are. To calculate this, we cut up a curve into infinitesimally small
pieces and then sum them. Doing this we get the hyperbolic length of a
piecewise C1 path to be given by the integral:
[1]
Furthermore, the distance between two points X and Y in where X <
Y < 1 can be taken as the distance from the centre to Y minus the
distance from the centre to X, or:
Where:
II Definition of the hyperbolic circle
In Euclidean Geometry a circle is defined as the locus of all points a
fixed distance from a fixed point. For this essay we shall take this
definition to the hyperbolic plane. We will be working mainly with the
Poincarѐ Disc Model, and sometimes use the Upper Half Plane Model of
hyperbolic geometry for clarification.
Let us begin with a definition of a hyperbolic circle in taken from [1]. A
hyperbolic circle in is a set in of the form:
Hyperbolic Circles | Anton Bubna-Litic 3
Where and s > 0 are fixed. We will refer to x as the hyperbolic
centre of C and s as the hyperbolic radius of C.
Later on, we will prove the properties of a Hyperbolic circle in depth,
but for now we will take a trivial example of x = 0 in the disk model in
order to visualise what a hyperbolic circle really is. Since the hyperbolic
distance from a point to the centre of the Poincarѐ disc is related to the
Euclidean distance then this gives us the image of a Euclidean circle
with its centre at x = 0 as seen in figure 1 below:
Figure 1: A hyperbolic circle in the Poincarѐ disk model
Now we know that the above circle is preserved under Möb( ) but we
cannot be certain that the position of the centre will be preserved under
Möb( ). In fact, hyperbolic centres are not preserved at all.
Let us rigorously explore the properties of a hyperbolic circle adapted
from [1]. We will begin by looking at the set of hyperbolic circles in
whose centres are at 0. Let us take an s > 0 and r = tanh s, such that
. Now Möb( ) contains so then p is an isometry
4 Cross-sections | Volume V 2009
of , and so every point in satisfies as
well. Now we have a Euclidean circle with Euclidean centre 0 and
Euclidean radius r, and a hyperbolic circle with hyperbolic centre 0 and
hyperbolic radius s, where s and r are related by r = tanh s. Therefore,
it is clear that a hyperbolic circle in takes the same form as a
Euclidean circle and although they have the same centres their radii are
different.
Now, let C be the hyperbolic circle in with hyperbolic centre c and
hyperbolic radius s. Let m be an element in Möb( ) taking c to 0. Then
we have m(C) as the hyperbolic circle in with hyperbolic centre 0 and
hyperbolic radius s. In particular, m(C) is a Euclidean circle. Since the
elements of Möb( ) take circles in t circles in
to , and since no
element in Möb( ) maps a point of to ∞, we see that
is also a Euclidean circle in . Therefore all
hyperbolic circles in have the form of a Euclidean circle.
Now, let us take a Euclidean circle C in with Euclidean centre of C
not being 0. Let us take a Euclidean line L passing through 0 and the
centre of the Euclidean circle. Note that L is also a hyperbolic line in .
Now L cuts C at X and Y. It is clear from the above definition of
hyperbolic distance in that the hyperbolic midpoint c of X and Y is
not at the centre of the Euclidean circle. Choosing an element m of Möb(
), taking c to 0 and taking L to , and as the X and Y are equidistant
from c and lie on L, and m takes them to the two points s and –s for
some , then m(C) is a Euclidean circle in perpendicular to ,
passing through s and -s. In particular, m(C) is a Euclidean circle in
centred at 0, and so it is also a hyperbolic circle in centred at 0.
Therefore, C is a hyperbolic circle in .
So, summarising what was proved above, we get that a hyperbolic circle
in is a Euclidean circle in and vice versa, although the hyperbolic
and Euclidean centres, and the hyperbolic and Euclidean radii, will, in
general, be different.
Hyperbolic Circles | Anton Bubna-Litic 5
Figure 2: Hyperbolic circles in the Poincarѐ disk
III Circumference of a hyperbolic circle
Now we shall prove some basic properties of hyperbolic circles starting
with the circumference represented in terms of its radius. Now let us
take a hyperbolic circle C with s > 0 in with hyperbolic centre 0 and
hyperbolic radius s. Now as the hyperbolic circle C has a radius of s,
and hence:
Rearranging to make r the subject we get r = tanh s, and so our
hyperbolic circle is also a Euclidean circle with radius r. Now let us
6 Cross-sections | Volume V 2009
impose some parameters on the circumference of C by f : [0,2π] →
where f (t) = reit so:
Therefore the circumference of a hyperbolic circle with radius s is 2π
sinh s.
IV Area of a hyperbolic circle
Let us now find that the area of a hyperbolic circle with radius s can be
calculated by summing all of the circumferences from 0 to s. So, we will
simply integrate the circumferences from 0 to s:
Hyperbolic Circles | Anton Bubna-Litic 7
V Comparison with Euclidean Geometry
Now, comparing the above two results with their Euclidean equivalent
we notice that sinh s > s, so 2πsinh s > 2π s, and so, the hyperbolic
circumference of a circle is always greater than the Euclidean
circumference of the same circle for s > 0. Now rearranging the
equation for hyperbolic area we get:
(1)
Now taking the power series for sinh x / x, we get:
For all positive x we notice that:
8 Cross-sections | Volume V 2009
So therefore:
Thus, the hyperbolic area of a circle is always greater than the Euclidean
area of the same circle for s > 0. Now, rearranging the equation for
hyperbolic circumference we get:
(2)
Now we know that sinh x / x approaches 1 as x approaches 0, so we
notice in equations (1) and (2) that for very small values of s, our
hyperbolic circumference and area are approximately Euclidean which
shows that hyperbolic geometry is approximately Euclidean geometry
in small regions as both lines and areas approximate each other.
Now in Euclidean geometry it is true that we can draw a circle of any
Euclidean centre and any Euclidean centre, and any three non-collinear
points determine a circle. In hyperbolic geometry, it can be seen above
that we can have a circle in of any hyperbolic centre and hyperbolic
radius. However, it is not the case that any three non-collinear points
determine a circle. This is simply because of the limited Euclidean size
of . Specifically, let us take the three points z1 = ½, z2 = 0 and z3 = -1/2
+ i in for sufficiently small . Now we know that a hyperbolic
circle through z1, z2 and z3, if it exists, is also a Euclidean circle.
However, for small values of the Euclidean radius of the circle
through z1, z2 and z3 is so large that it passes outside the Poincarѐ disc
and thus the circle cannot exist.
Hyperbolic Circles | Anton Bubna-Litic 9
Figure 3: A Euclidean circle that cannot fit in the Poincarѐ disk
VI Circle centers and their radii
Now we shall look at direct calculations of hyperbolic radii and centres
from Euclidean circles in and . First let us take a Euclidean circle
A in with Euclidean centre c = a + ib and Euclidean radius r where
|c| > r. Note that the entire circle must lie within and so 0 < r < s < 1.
Therefore, from the formula for hyperbolic distance given above, we
get:
10 Cross-sections | Volume V 2009
Now let us write the Euclidean centre as with
|c| > r. We see that A intersects the hyperbolic ray
in two points. As reflection in R takes
A to itself and is a hyperbolic isometry, we see that the hyperbolic
centre of A also lies on R. The two points of intersection of A and R are
and . Calculating the hyperbolic
distant between these two points we get:
Therefore, the hyperbolic radius of A is simply - . Now let us call
the hyperbolic centre of our circle in to be for (|c|- r) < α < (|c|+
r). Now the distance from the centre of the hyperbolic circle to the
points X and Y are equal so:
Hyperbolic Circles | Anton Bubna-Litic 11
Where the hyperbolic centre of A is .
Now, working in we shall do the same as above. We will take a
Euclidean circle A in with Euclidean centre c = a + ib and Euclidean
radius r. Now as vertical Euclidean lines in are both Euclidean and
hyperbolic lines, and as the reflection in the line
takes A to itself, then the hyperbolic centre of A lies on l. Now A
intersects l twice at X = a + (b – r)i and Y = a + (b + r)i, and so we can
now find the hyperbolic radius R:
Rearranging we get:
12 Cross-sections | Volume V 2009
Now the hyperbolic centre is the point a + si where:
Therefore, the hyperbolic centre in is
It is clear that when comparing Euclidean circle radii and centres to the
hyperbolic equivalents it is much easier to work in than .
Hyperbolic Circles | Anton Bubna-Litic 13
VII Comparison to Euclidean theory
We shall begin this final section with the Euclidean theory that the angle
formed at the centre of the circle by lines originating from two points on
the circle's circumference is double the angle formed on the
circumference of the circle by lines originating from the same points.
Figure 4: A hyperbolic circle with hyperbolic lines in
Now let us take the Euclidean circle C shown in Figure 4, with
Euclidean centre at 5i and Euclidean radius 4. Thus its hyperbolic
centre and radius is O = 3i and R = ln 3. We shall take the vertical line
through which cuts C at A = 9i and B=i, and the point
on the circumference of the circle. Now to find the
angle α we need to find the lengths of AX, OA, and XO. Using the
formula given in [1] we have:
14 Cross-sections | Volume V 2009
Where:
Therefore we get:
Now as OA is a vertical line we can calculate it to be:
Hyperbolic Circles | Anton Bubna-Litic 15
Now for any compact hyperbolic triangle in with angles α, β and γ
which are opposite the sides of hyperbolic length a, b and c respectively,
we can obtain the hyperbolic law of cosines taken from [1] which is:
Or:
Subbing our hyperbolic lengths into the above equation we get:
Similarly to calculate β we need to find the lengths of BX, OB and XO:
16 Cross-sections | Volume V 2009
Now as OB is a vertical line we can calculate it to be:
Now using the law of cosines we get:
It can be seen that β . In fact β > 2α, so therefore the Euclidean
theory does not hold for hyperbolic geometry. Now the statement that
β > 2α can be easily proved as the hyperbolic triangle AXO is isosceles
(as AO and XO are equal radii of the hyperbolic circle C. Now we know
that in hyperbolic geometry the angle sum of a hyperbolic triangle is
smaller than π so:
Hyperbolic Circles | Anton Bubna-Litic 17
But since supplementary angles add to π then we know that:
But, as the hyperbolic triangle AXO is isosceles, then the base angles are
equal. To prove this, we shall use a compact hyperbolic triangle in
with angles φa , φb , and γc , which are opposite the sides of hyperbolic
length a, b and c respectively, where b = a. Now using the law of cosines
we get:
And:
But b = a, so:
So since the base angles of an isosceles hyperbolic triangle are equal
then , so therefore β > 2α.
In summary hyperbolic circles in and are Euclidean circles and
vice versa, yet their centres, radii, circumferences and area will differ
except on very small scales where hyperbolic geometry approximates
Euclidean geometry. Furthermore, a hyperbolic circle differs from a
Euclidean circle since a hyperbolic circle cannot be drawn through
every three non-collinear points. Finally the properties of Euclidean
18 Cross-sections | Volume V 2009
circle geometry do not always hold with hyperbolic circles but there
may be some that do hold. Hyperbolic circle geometry may have its
own theorems similar to the Euclidean circle geometry theorems, but
these will need to be proved another time.