Through the Looking Glass:

A glimpse of Euclid’s twin geometry:

The Minkowski geometry

Supplementary Notes
ICME-10 Copenhagen 2004
Bjørn Felsager
Haslev Gymnasium & HF, Denmark

Preliminary remarks (about the history of Minkowski geometry) p2

Introduction (motivation and background for the notes) p3
An overview of possible 2-dimensional geometries p4
Common ground: The affine geometry p5
Following Alice through the Looking Glass: Adopting the
rectangular hyperbola as the most symmetrical figure p5
Right angles in Minkowski geometry p6
Correspondences between the Euclidean and the Minkowski geometry p8
Dialogues between Alice and the Gryphon concerning
the Pythagorean theorem p11
Angles in Minkowski geometry p14
Trigonometry in Minkowski geometry p16

1
Preliminary remarks: Let us first remind ourselves a little about the history of the
Minkowski geometry. It is a fairly new discovery going back to the beginning of
the previous century, where Einstein in 1905 published his famous paper ‘Zur
Elektrodynamik bewegter Körper’ which later became known as his introduction
to the special relativity. In this paper he constructed a theory of space and time,
where space and time no longer were considered absolute concepts, but mingled
with each other in a way that could be precisely described by the so called Lo-
rentz transformations.

The theory of relativity made an enormous impact on the contemporary scientists

including many famous mathematicians. In 1907 Minkowski gave a most influ-
ential lecture, where he showed that the theory of special relativity could be cast
into a purely geometrical theory of space and time with an invariant based upon a
variant of the Pythagorean theorem:
dτ2 = dt2 – ds2/c2
The square of the distance between two neighboring space-time events (measured
in the proper time τ) is the same as the difference between the square of distance
measured in the inertial (laboratory) time t and the square of the Euclidean dis-
tance s. (The velocity of light c takes care of the conversion between time units
and space units). Minkowski concluded his lecture with the famous and dramatic
prophecy:

‚Von Stund an sollen Raum für sich und Zeit für sich völlig zu Schatten herab-
sinken, und nur noch eine art Union der beiden soll Selbständigkeit bewahren’

Shortly afterwards in 1910 Klein gave an important lecture on the Minkowski

geometry of space-time, where he showed how it fitted into the scheme of the
Erlangen Programme: A particular kind of geometry was to be characterized by a
group of symmetry transformations. The study of the particular geometry could
then be considered a study of the properties of geometrical figures left invariant
by the group of symmetry transformations. In the case of Minkowski geometry the
group of symmetry transformations consisted of the Lorentz transformations or
rather the extended group of Poincare transformations, which also included dis-
placements.

So this is the official line of history behind the Minkowski geometry and because
of it’s mingling of space and time it is usually considered to be a more abstract
theory than both the usual Euclidean geometry and its extensions, the spherical
geometry as well as the hyperbolic geometry.

These preliminary remarks obviously raise the following question: Why should
we be interested in Minkowski geometry in this setting? I hope to be able to
produce a satisfactory answer to this question in the following discussion!

Remark: These notes have been used on several occasions as a general introduc-
tion to Minkowski geometry. In one case – at a teachers college – I was also al-
lowed to include a workshop. I have reproduced this workshop in an accompany-
ing paper, to show how one can supplement a general introduction with activities,
that allows students to get their hands dirty in Minkowski geometry as well as
letting them gain some experience themselves.

2
Introduction: In a high school as well as in a teachers college the courses in
Euclidean geometry can be supplemented with a short excursion into non-
Euclidean geometries for several reasons:
1) It is by itself a dramatic experience to realize that the Euclidean geometry is
not the only possible geometry – and the discovery of the non-Euclidean
geometries is certainly a very important part of the cultural history of
mathematics with lots of philosophical implications.
2) Gaining experiences with non-Euclidean geometries puts Euclidean geome-
try itself in a new fresh perspective. You can no longer rely on your intui-
tion and many subtleties that are easily overlooked in Euclidean geometry
suddenly bring themselves to your attention.
It is for these and other reasons that short excursions into non-Euclidean geome-
tries can be very rewarding in traditional mathematics courses at least beginning
with the high school level.
But what possibilities do we then have for bringing such concepts to the stu-
dent’s minds? Traditionally elementary books on geometry focus exclusively upon
the spherical geometries and the hyperbolic geometries. And very often a histori-
cal line of arguments is used to motivate the new geometries beginning with a
discussion of the parallel axiom.
But these are not our only options! Not only do we know much more now about
geometry than was known two centuries ago, when non-Euclidean geometries
were first discovered – and hence we have alternative routes into non-Euclidean
geometries. But in particular we now have available dynamical geometry pro-
grams, which – suitably modified – allows us to experiment and thus gain first
hand experiences with non-Euclidean geometries. This has been emphasized for
some time in relationship to the hyperbolic geometry, where various standard
models – such as the Poincare disk model – has been successfully implemented.
But is seems much less known, that it is much easier to implement tools for the
Minkowski geometry than for the full hyperbolic geometry and because the Min-
kowski geometry is much closer related to the Euclidean geometry, it is in fact
much easier to introduce.
In this lecture I will therefore outline a possible introduction to Minkowski ge-
ometry based upon the following principles:
1) The use of a dynamical geometry program such as Cabrii or SketchPad to
make geometrical constructions in the Minkowski geometry immediately
available to students.
2) The similarity between the usual Euclidean geometry and the Minkowski
geometry is emphasized – in particular there is no mention of the space-
time structure in the beginning. In stead their common ground (the affine
geometry) is being exploited.
3) A dramatic setting based upon the well-known tales of Lewis Carroll – ‘Alice
in wonderland’ and ‘Through the Looking Glass’ – is used to capture the
imagination of the students.
Remark: Although the following content is well known, to the best of my knowl-
edge the setting is original. In fact I don’t think it will be easy to find the ideas
explicitly revealed in the literature. They seem to belong to the Mathematical folk-
lore, which are of course well known by the experts, but some how no one got the
time to write them down!

3
So we begin with the following important questions:
Is Euclidean geometry the only possible geometry?
Can you justify the answer in a different way from the historical
line of argument (which invokes the axiom of parallel lines)?
Let’s recall the axiom of parallel lines in the following version, which is particu-
larly simple and in particular it does not involve any advanced concepts such as
circles or angles:
Playfair’s axiom:
Through a point not on a given line there is one and only one line parallel to
the given line (i.e. that does not intersect the given line).

We begin by revealing that there are four simple geometrical structures you can
put upon a two-dimensional space, two of which obeys the parallel axiom (i.e. the
corresponding space is flat) and two of which fails to obey the parallel axiom (i.e.
the corresponding space is curved):

Euclidean geometry (2d): Minkowski geometry (2d):

Geometry of straight lines and Geometry of straight lines and
circles. rectangular hyperbolas.
Circular trigonometry: Hyperbolic trigonometry:
sine and cosine exp and ln

Spherical geometry: Hyperbolic geometry:

Geometry of a spherical surface in Geometry of a hyperbolic surface in

Euclidean space. Minkowski space.

Spherical trigonometry. Hyperbolic trigonometry.

4
So the Euclidean geometry does posses a twin, the Minkowski geometry, which
avoids curvature and hence satisfies the axiom of parallel lines. The two geome-
tries share the affine structure of plane, i.e. they share:

• Points
• Straight lines
• Parallel lines

• Ratios
• Midpoints
• Medians

• Translations
• Multiplications
• Similar figures

But is it also possible to give an elementary introduction to the Minkowski

geometry avoiding abstract concepts such as the geometrical structure of
space-time?

We suggest the following strategy: A Euclidean geometry is based upon a view on

symmetry that makes the circle the most symmetrical figure – since ancient times
considered the most perfect figure. This view e.g. dominated astronomy for sev-
eral thousands years. But is it possible to imagine another view of the world,
where it is not the circle, but the rectangular hyperbola, that is considered the
most symmetrical figure? To make the transition to this alternative view more po-
tent, we imagine that we follow Alice through the looking glass, and that she pre-
cisely discovers it is the alternative view, that prevails behind the looking glass:

Following Alice through the Looking Glass:

5
 1
1

The most symmetrical figure as seen

in front of the looking glass – behind the looking glass

The adoption of the rectangular hyperbola as the most symmetrical figure re-
quires the introduction of new structures in Minkowski geometry:

• Two asymptotic directions: Vertical/ horizontal

• A different concept of right angles: We need to replace the traditional
right angles, because of their intimate relationship with the geometry of the
circle!

Right angles in the Minkowski geometry: To motivate the introduction of hy-

perbolic right angles we need some characteristic properties of a right angle,
which links the right angle to a circle. Which property does not matter! We chose
the following one, since this is a very basic one, easy to understand also on an
intuitive level.

• The tangent is perpendicular to the radius

To understand how tangents of a hyperbola behave we will use some analytical
geometry. You may find this a little disturbing: Why not use elementary geometry
all the way through. There are two reasons. The first one is that analytical geome-
try is in fact much simpler in relation to the Minkowski geometry than the
Euclidean geometry. The second is that we lack intuition about the structure of
figures in Minkowski geometry. For these reasons simple proofs in Minkowski ge-
ometry tends to be easier to follow using a little analytical geometry!

As a starting point we therefore take the following observation about the slope of
a secant. By the way this is the only detailed argument based upon analytical ge-
ometry I will present in this paper. There will thus be lots of opportunities for the
reader verifying results analytically later on your own!

The theorem of the secant: To determine the slope of a secant we perform the
following standard calculation:

1 1 x1 − x 2
−
x x1 x ⋅x 1
α sekant = 2 = 2 1 =−
x 2 − x1 x 2 − x1 x 2 ⋅ x1

(see the figure attached!)

6
 y =1/x

1/x1

1/x2

x1 x0 x2

Notice in particular that the slope of the secant only depends upon the product
of the abscissa x! This has deep implications for the addition of hyperbolic angles!

To compute the slope of the tangent, we now simply let the two endpoints of the
secant coincide. We thereby find the following well-known result:

1
α tangent = − 2
x0

Notice that we have managed to derive this elementary result without appealing
to calculus! Next we compare the slope of the tangent to the slope of the radius:
1
1 x 1
α tangent = − 2 and α radius = 0 = 2
x0 x0 x0

y =1/x

1/x0

x0

Conclusion: In Minkowski geometry two lines are perpendicular precisely

when they have opposite slopes!

This observation is the main result. In particular it makes it easy to investigate

simple properties of right angles!

7
Recall the following theorems linking ordinary right angles to circles:

Example 1 – First theorem of circles: Example 2 – Second theorem of circles:

Thales’ theorem: The theorem of Chords:
An angle is a right angle precisely The perpendicular bisector of a chord
when it spans the diameter. passes through the center.
Slope AC = -1,158 (Slope AC) (Slope CB) = -1,000
B
Slope CB = 0,864
C

A
A

O O

Example 3 – Third theorem of circles: The circumscribed circle of a triangle:

The perpendicular bisectors of a triangle pass through the same point, the
center of the circumscribed circle (i.e. the circumcentre O).

M BC M AB

C
M CA

In each of the above cases it is elementary to verify the corresponding theorem in

Minkowski geometry experimentally using a Dynamic Geometry program capable
of drawing rectangular hyperbolas (a hyperbolic compass!). We leave these verifi-
cations as exercises (see the accompanying workshop).

8
Other more non-trivial examples of correspondences between the Euclidean ge-
ometry and the Minkowski geometry involve the Euler line and the nine-point cir-
cle:

The Euler line:

a) The Euler line in Euclidean b) The Euler line in the Minkowski

geometry: geometry:

The nine-point configuration:

a) The nine-point circle: b) The nine-point hyperbola:

Notice that both cases are special cases of a theorem in affine geometry, which
says that the heights in the triangle can be replaced by any set of three lines from
the vertices passing through the same point. This gives rise to nine points
through which a unique conic passes, the nine-point conic.

• In the case of an ellipse, you may consider the configuration to be a parallel

projection of the nine-point circle.
• In the case of a hyperbola, you may similarly consider the configuration to
be a parallel projection of the nine-point hyperbola.

9
In both cases the configuration is a shadow of a corresponding simpler configura-
tion in Euclidean respectively Minkowski geometry. The above configuration with
a nine-point conic is thus one of the places, where two separate theorems from
Euclidean and Minkowski geometry are unified in affine geometry!

As a final example of such a correspondence we will look at the kissing circles.

In Euclidean geometry any triple of points admits circles, which kiss each other
(tangentially) e.g. as shown on the following figure:

In Minkowski geometry it is slightly more complicated, but in many cases you can
still find kissing hyperbolas as shown on the following figure:

10
Important remark: At this point you may perhaps think that all results in Euclid-
ean geometries involving circles have a Minkowski counterpart. But that is not
the case: The symmetry structure of the two geometries also has characteristic
differences. E.g. rotations in Euclidean space have a repetitive periodic structure
unlike the rotations in Minkowski space, where the asymptotic directions break
the periodicity. As a consequence the Minkowski geometry lack regular polygons.
And thus the regular polygons constitute an example of an important concept in
Euclidean geometry, which has no correspondence in Minkowski geometry. But
for pedagogical reasons we have emphasized the striking similarities rather than
the (also important!) differences between the two twin geometries.

At this point you should now have obtained some feeling for the Minkowski ge-
ometry and we proceed with a discussion of the most basic theorem in Minkowski
geometry – the analogue of the Pythagorean theorem, which controls all distance
calculations!

We present the derivation in the form a dialogue between Alice, the Mock Turtle
and the Gryphon starting with a famous dialogue written by Lewis Carroll for ‘Al-
ice in Wonderland’:

'I couldn't afford to learn it,' said the Mock Turtle with a sigh. 'I only took the
regular course.'
'What was that?' inquired Alice.
'Reeling and Writhing, of course, to begin with,' the Mock Turtle replied; 'and
then the different branches of Arithmetic – Ambition, Distraction, Uglification,
and Derision.'
'I never heard of "Uglification",' Alice ventured to say. 'What is it?'
The Gryphon lifted up both its paws in surprise. 'Never heard of uglifying!’ it
exclaimed. 'You know what to beautify is, I suppose?'

11
A fictitious dialogue between Alice and the Gryphon about geometry:

‘I do suppose you know what a square is?’ the Gryphon exclaimed.

‘Of course’ Alice replied. ‘ It’s a totally symmetrical quadrilateral with four right an-
gles’.
And to prove that she really understood what she was talking about, she made a
sketch of a square:

Alice makes a drawing of a square: The Gryphon makes a drawing of a

square:

Alice claims the Gryphon’s square is a

diamond (i.e. a rhombus with horizontal
and vertical diagonals)!

‘Oh no’, the Gryphon said in surprise: ‘That’s not a square – It’s just some silly
parallelogram! This is how a square looks like!’
To Alice surprise the Gryphon made a sketch of a diamond figure. ‘Is that what a
square looks like?’ she exclaimed.
‘Of course! Every child knows that a square has four right angles and is totally
symmetrical! Don’t you learn anything in your schools? Didn’t they ever tell you
about the Pythagorean theorem?’
‘Yes they did’, Alice replied cautiously, ‘ The Square of the hypotenuse is the sum of
the Squares of the legs’.
‘What are you talking about’, the Gryphon replied, not believing what it just heard:
‘Every child knows that the square of the hypotenuse is the difference between
the square of the legs’.
‘But I thought I had a proof?’ Alice dared to say.
‘Proof’, snorted the Gryphon. ‘You don’t even know what a square is!’

And the conversation continues with Alice demonstrating her proof and the Gry-
phon demonstrating his proof. Both use the same simple argument: First they
decompose a square according to the formula for the square of a binomial:

(a+b)2 = a2 + b2 + 2ab

Next they rearrange the figure suitably and the Pythagorean theorem follows im-
mediately by comparing the two figures obtained in this way and ignoring the
common right angled triangles:

12
 Alice explains the Pythagorean theorem:

a a2

c2

b2 c
b b

b a a

c 2 =a 2 +b 2

The Gryphon explains the Pythagorean theorem:

b2 b2
a
c
c c2
b
a2

b2

b 2 + a 2 = 2b 2 + c 2 ⇒ c 2 =a 2 –b 2

Remark: Once we have established the Pythagorean theorem for the Minkowski
geometry we can make some important observations. There exist a Euclidean
square, which is also a Minkowski square, namely the square with slopes ±1:

It is then obvious to assign them this common square the same length of the side
in the two geometries. Consequently the common square also gets the same area
in the two geometries. Since collages of such special squares can approximate
any simple figure, it follows that the area of any simple figure must in fact be the
same in the two geometries. Thus they also have areas in common!

It is now easy to show that a rectangular hyperbola can be characterized as a set

of points with constant hyperbolic distance to a center etc.

13
Once we control distances we can also introduce trigonometry. In fact the so-
called hyperbolic trigonometry is precisely the trigonometry associated with Min-
kowski geometry! As a preparation for trigonometry we must first introduce a
measure for the hyperbolic angles (as opposed to the usual circular measure of
angles). The starting point is a very important remark concerning addition of an-
gles:

On addition of angles:
a) Adding circular angles: P0Pu+v is parallel to PuPv:

Q=Pu+v Pv

Pu

0.2
P0
0.2

b) Adding hyperbolic angles: P0Pu+v is parallel to PuPv:

P0

Pu
Pv
Q=Pu+v

1 x1 x2 x3

The theorem of secants has the following important consequence:

1 1
The slope of P0Q: − The slope of PuPv: −
1 ⋅ x3 x1 ⋅ x 2

Conclusion: x3 = x1 ⋅ x 2

14
The hyperbolic angle is thus a logarithmic function of the associated abscissa
since we get the following identity for the measure hyp(x) of the hyperbolic angle:
u + v = hyp(x1) + hyp(x2) = hyp(x3) = hyp(x1· x2)
This makes it obvious to identify the measure of hyperbolic angle with the area of
the sector OP0Pu, which makes sence, since on the one hand it is a well known
fact – a fact that is elementary to verify! – that the area is also a logarithm
function of the associated abscissa. On the other hand the circular angle is
represented by an area in Euclidean geometry:

Angles in Euclidean geometry (notice that the circle goes through the ‘unit-
point’ (1,1) and that the circle has the total area 2π, since the radius is √2!):

P0

Pu

0.2

0.2 1 x1

Angles in Minkowski geometry: u = hyp(x) = Area(sector OP0Pu).

P0

Pu

0.2

0.2

Conclusion: Hyperbolic angles generate natural logarithms: u = ln(x)

15
Notice that the hyperbolic measure of angles leads to a very simple canonical
parametrization of the rectangular unit hyperbola, x·y = 1, in terms of the
anglular measure (hyperbolic ‘radians’). Since u = ln(x), we immediately get the
abscissa expressed through an inverse natural logarithm, i.e. a natural exponen-
tial function: x = exp(u). The ordinate is the reciprocal value, i.e. y = exp(–u). In
contrast to ordinary trigonometry, where it is customary to introduce two trigono-
metric functions cosine and sine, we thus need only one basic trigonometric
function for for the hyperbolic trigonometry: exp.

Canonical parametrizations
in Euclidean geometry and Minkowski geometry:

a) The unit circle: b) The unit hyperbola:

( x, y ) = ( cos(u), sin(u) ) ( x, y ) = ( exp(u), exp(-u) )

This makes it possible to introduce hyperbolic trigonometry in precisely the same

way you introduce circular trigonometry using right angled triangles:

√(2)· cosh(u)
√(2)· sinh(u)

(exp(u),
exp(–u))

√(2)

opposite leg exp(u) − exp( −u )

sinh(v) = =
hypotenuse 2
adjacent leg exp(u) + exp( −u)
cosh(v) = =
hypotenuse 2
opposite leg exp(u) − exp( −u)
tanh(v) = =
adjacent leg exp(u) + exp( −u)

We leave the details as an exercise!

16