Mastermath, Geometry, Lectures 8–11

Bas Edixhoven

2013/11/15

1 Planning
There will be 4 lectures, on December 13 an “overview and questions” session, and on December
20 the 2nd partial exam. If necessary, there will be a resit on January 24 (for both Jeroen’s and
my parts of this course).
These notes, and the geogebra files and other files that they refer to, can be downloaded from
http://www.math.leidenuniv.nl/˜edix/teaching/2013-2014/geometry/
at least shortly after each lecture.
I prefer to write some notes like this and using chalk and blackboard over making “power-
point presentations”, because I feel that just clicking to the next pause is somehow cheating in
showing how to construct some mathematics. It is more honest to “create” the material in real-
time, by thinking, speaking and writing. Of course, this is only a personal opinion of a university
professor who has never had any education in teaching. It is not meant to convince you to do the
same. And I will use the video-projector to show pictures.
These notes are not intended to be complete. Their purpose is to make it possible for the
lecturer to skip some parts if there is not enough time, and to have an overview of what was
treated. Students should make their own detailed notes during the lectures.

2 Lecture 8
2.1 From last week
• Go through Jeroen’s slides.

• Explain slide 10 if necessary (see Kindt’s “Lessen in de projectieve meetkunde”, Epsilon

uitgaven, Utrecht, page 23, for example).

1
 • Union Jack construction: show it again (union-jack-constr.ggb), and show by
moving that the result only depends on the point of intersection with the horizon.

• Then the homework given on Jeroen’s slides 15–17. I am very pleasantly surprised by this,
I have learned a lot from it myself (quadrangle-square.ggb).

• He also gave homework Stilwell 5.1–5.3. So let us also go through that.

Then it is now time for new stuff.

2.2 Axioms for projective geometry

Think of Hilbert’s approach: just some data satisfying some conditions. A projective plane is a
triple (Π, Λ, I) where Π is a set, called the set of Points, Λ is a set called the set of Lines, and I
is a subset of Π × Λ, called the incidence relation ((P, L) ∈ I means: P is contained in L). The
axioms are then:

P1 : every two distinct Points are contained in a unique Line,

P2 : every two distinct Lines contain a unique Point,

P3 : there are 4 distinct Points of which no 3 are collinear (that is, lie on one Line).

2.3 A model for P2 (R)

We put the “all-seeing eye” where it belongs, at the origin in R3 :

O = (0, 0, 0).

See figure 5.10 in Stillwell.

The idea is now to consider the set of all lines through O and not to make a choice of a
plane not containing O to project to. I see two good reasons for this. The first reason is simply
that it is better to avoid choices; choices imply loss of symmetry, and we want to profit from all
the symmetry that projective geometry offers. The second reason is that for every plane H not
through O we get trouble with the plane H 0 through O that is parallel to H: points in H 0 have no
projection to H.
For Π, the set of Points, we take the set of lines in R3 that contain O, and for Λ, the set of
Lines, we take the set of planes in R3 containing O. And for I we take the set of (P, L) with
P ⊂ L.

2
 Then this triple (Π, Λ, I) satisfies the axioms P1–P3 (apply some linear algebra). We denote
this projective plane by P2 (R).
In fact, we could have used any field F instead of R. That gives P2 (F ).

2.4 Embedding of R2 in P2 (R)

For this we must choose a plane H not containing O. We choose H = {(x, y, z) ∈ R3 : z = 1}.
A point (x, y) in R2 gives a point (x, y, 1) in H, and that point gives us the Point
{t·(x, y, 1) : t ∈ R}, also denoted R·(x, y, z), in P2 (R). Note that this map (x, y) 7→ R·(x, y, 1)
is injective: if R·(x, y, 1) = R·(x0 , y 0 , 1), then indeed (x, y) = (x0 , y 0 ).
Hence we have extended the euclidean plane R2 to the projective plane P2 (R).

2.5 Homogeneous coordinates

We introduce a notation called “homogeneous coordinates” that makes it easier to work with
P2 (R).
For (x, y, z) ∈ R3 − {O} we put:

(x : y : z) := R·(x, y, z) (the notation suggests ratios!).

So, then, (x : y : z) is a Point. And for every nonzero t in R we have:

(tx : ty : tz) = (x : y : z).

Formally speaking, the map q : R3 − {O} → P2 (R) that sends (x, y, z) to (x : y : z) is the
quotient map for the equivalence relation:

(x, y, z) ∼ (x0 , y 0 , z 0 ) if and only if ∃t ∈ R∗ : (tx, ty, tz) = (x0 , y 0 , z 0 ).

We also introduce similar notation for Lines. For (a, b, c) ∈ R3 − {O} we put:

(a : b : c)⊥ := {(x, y, z) ∈ R3 : ax + by + cz = 0}.

Then, (a : b : c)⊥ is a Line.

The exterior product (a, b, c) × (d, e, f ) = (bf − ce, −(af − cd), ae − bd) gives the Line
through two distinct Points, and also the intersection Point of two Lines.
Do exercises 5.4.1—5.4.3 of Stillwell.

3
2.6 Projective spaces
For every natural number n ≥ 1 we have the set of lines through 0 in Rn+1 , the set of Points
of Pn (R). The set of Lines are the planes in Rn+1 through 0, the Planes are the 3-dimensional
subvectorspaces of Rn+1 , etc. This also works with R replaced by any field F .
Linear algebra in F 4 gives: for any 3 distinct Points in P3 (F ) there is a unique Plane that
contains them, and the intersection of 2 distinct Planes is a Line, etc. Even: it is sometimes good
to replace F n+1 by any F -vectorspace V , and denote the result by P(V ). This occurs naturally
if we consider a Line V in P2 (R): recall that V is a plane in R3 with 0 ∈ V , that is, V is a
2-dimensional subspace of R3 , but it does not come with a preferred basis. Choosing a basis
of V gives an identification of V with P1 (R). But another choice of basis of V gives another
identification of V with P1 (R), leading to the group of projective transformations of P1 (R) that
we will study next week.
So, the case n = 1 is also important to us, it gives us the projective line P1 (R). In terms of
homogeneous coordinates, the embedding above of R2 into P2 (R) is the map:
R2 → P2 (R), (x, y) 7→ (x : y : 1)
The points of P2 (R) that are not in the image of this map are the (x : y : 0). The set of these is in
bijection with the set of Points (x : y) of the projective line P1 (R). This is the set of “directions”
in R2 , and indeed these directions correspond to the points on the “horizon” in P2 (R).

2.7 An important principle

This is not contained in Stillwell. It leads to a very simple proof of some incidence theorem in
euclidean geometry.

2.7.1 Theorem. Let P , Q, R and S be in P2 (R), not any three of them on a Line. Then,
after a suitable choice of basis (v1 , v2 , v3 ) of R3 , the homogenous coordinates are as follows:
P = (1 : 0 : 0), Q = (0 : 1 : 0), R = (0 : 0 : 1) en S = (1 : 1 : 1). The basis (v1 , v2 , v3 ) is
unique up to simulaneous scaling (v10 , v20 , v30 ) = (tv1 , tv2 , tv3 ) with t ∈ R − {0}.
Proof. We must take v1 on P , v2 on Q, and v3 on R, all three non-zero. As P , Q and R are
not contained in one Line, v1 , v2 and v3 are linearly independent, hence a basis of R3 . Then
S = a1 v1 + a2 v2 + a3 v3 with all ai 6= 0 (because S does not lie on any of the lines containing
two of P , Q and R). Now replace v1 by a1 v1 , etc. Done. 
2.7.2 Remark. The choice of such P , Q, R and S involves 4·2 = 8 degrees of freedom, whereas
a choice of origin in R2 plus a choice of basis has only 6 degrees of freedom (affine geometry). A
choice of origin and an orthonormal basis has only 3 degrees of freedom (euclidean geometry).

4
2.8 A funny construction
2.8.1 Theorem. Let A and B in P2 (R) be distinct. Let L1 and L2 be distinct lines through A,
not containing B, and M1 , M2 en M3 distinct lines through B not containing A. Let Pi,j be the
intersection point of Li with Mj . Let Qi,j be the intesection point of the line containing Pi,i and
Qj,j with the line containing Pi,j and Qj,i . Then A is on the line containing Q1,2 and Q2,3 .

Proof. We choose a basis of R3 such that A = (0 : 1 : 0), B = (1 : 0 : 0) and P1,3 = (0 : 0 : 1).

Then the line containing A and B is the line with equation z = 0, the horizon of the euclidean
plane R2 embedded by (x, y) 7→ (x : y : 1), L1 is the line with equation x = 0 and L2 the line
with equation y = 0. In the euclidean plane, our configuration consists of two rectangles with a
common side, and the line containing Q1,2 and Q2,3 is horizontal, hence meets A on the horizon.


This proof illustrates well how one can use the extra freedom in projective geometry to reduce
the proof of some incidence results in the projective plane to simpler cases in euclidean geometry.
It also gives something back to euclidean geometry, namely, such incicende results.

2.9 Homework
1. Use Theorem 2.8.1 to show that in the euclidean plane one can draw the segment contain-
ing two distinct points, even if the distance between them is larger than the length of the
ruler you have.

2. Read Stillwell §5.5-5.6 and do the exercises of those two sections.

5
3. Let P = (Π, Λ, I) be a projective plane (that is, satisfying the three axioms P1, P2 and
P3). Assume that Π is a finite set. Then show that all lines in P have the same number
of points, that through every point the same number of lines passes, and that there are as
many points as lines. Hint: use projections, show that they give bijections.

4. Let (Π, Λ, I) be a projective plane. Show that (Λ, Π, I t ), with

I t = {(l, p) ∈ Λ × Π : (p, l) ∈ I}

is also a projective plane.

6
3 Lecture 9
3.1 Remarks, and homework from lecture 8
3.1.1 Remark. Last week it was hard to understand that a Point of P2 (R) is a line in R3
through 0, even though the projection from 0 of such a line should of course be a point. The
problem is that we have to define that projection. That projection should be a map from R3 − {0}
to some set, but the question is then what that set should be. The answer to that question is:
it should be the projective plane. So we are turning around in a circle, until we realise that
nowadays we live in Cantor’s paradise of set theory1 , and we can just consider the set whose
elements are the lines in R3 through 0. In set theoretical notation:

P2 (R) = {P : P is a line in R3 through 0}.

One attempt to make the concept of Points more concrete was to use homogeneous coordinates
for them:

for (x, y, z) ∈ R3 − {0}: (x : y : z) = R·(x, y, z) = {(tx, ty, tz) : t ∈ R}.

Then we have, for all (x, y, z) ∈ R3 − {0} and for all t in R − {0}:

(tx : ty : tz) = (x : y : z),

and this is consistent with the suggestion that the colons instead of commas indicate ratios.
I have tried, during last week, to find a simpler presentation of P2 (R), and have not succeeded.
The best that I can come up with, is to go down by 1 in dimension: P1 (R) is the set of lines in
R2 passing through 0. The advantage in this case is that the homogeneous coordinates are of the
form (x : y), and that this we are used to identify with the real number x/y, unless y = 0, in
which case we can call it ∞. Indeed, the embedding R → P1 (R), x 7→ (x : 1), misses only one
Point: (1 : 0), and that point we call “the point at infinity”. In the higher dimensional case of
P2 (R), we have a “line at infinity”. I hope that this helps. Time will also help (at least if you
spend some of your time thinking about it, and working with it). Kindt’s book, section 30 gives
some more details, but no better ideas.

3.1.2 Homework

1. The exercises of Stillwell, §5.5–5.6, questions?

1
a famous quote of Hilbert: “No one shall expel us from the Paradise that Cantor has created.”

7
 2. Who wants to construct the segment AB with a ruler that is too short?

3. The exercise with the finite projective plane: questions?

Here is a sketch of a solution. Let l and l0 be in Λ. Because of the axioms there is a
P 6∈ l ∪ l0 (fill in the details!). We get a map f : l → l0 by sending Q ∈ l to P Q ∩ l0 , the
unique intersection point of the line through P and Q with l0 . This map is bijective (that is,
injective and surjective) because of axioms P1 and P2. The inverse of f is the projection
g : l0 → l from P . Conclusion: all lines have the same number of points. Let us denote it
by n + 1.
Let now P be a point. We take a line l with P 6∈ l (see exercise 5.3.4 in Stillwell). Then
we get a map from the set of points on l to the set of lines through P by sending Q on l to
the line P Q. This map is bijective. The inverse map sends a line m through P to m ∩ l.
Conclusion: through every point pass exactly n + 1 lines.
How many points are there in P? Choose a point P , and consider the lines
through P . There are n + 1 of these, and they intersect only at P . Hence there are
1 + (n + 1)n = n2 + n + 1 points.
The number of lines. Well, do the “dual argument” (exchange the notions of lines and
points). Let l be a line. Every line m 6= l meets l in one point. Through every point on l
pass n + 1 lines, among which l itself. That gives 1 + (n + 1)n lines.

3.1.3 Example. Let p be a prime number. Then #P2 (Fp ) = p2 + p + 1. In homogeneous

coordinates: there are p2 of the form (x : y : 1), p of the form (x : 1 : 0) and there is
(1 : 0 : 0).

3.2 Today’s program

We treat Stillwell, 5.5, 5.6 and 7.3: projective transformations. The main points are:
• projective transformations are fractional linear in inhomogeneous coordinates,

• projective transformations are linear in homogeneous coordinates.

3.3 Affine transformations of lines

Let L be a line in R2 that does not contain 0. Then L has many parametrisations f : R → L ⊂ R2
(choice of a vector in L (steunvector in Dutch) and choice of a directional vector (richtings-
vector)). For f1 and f2 parametrisations of L: f2−1 ◦ f1 : R → R is an affine transformation, that
is, of the form x 7→ ax + b, with a and b in R and a 6= 0.

8
 If we want that the parametrisation is an isometry, then the directional vector must have
length 1. That means that there are exactly two directional vectors that differ by a factor −1.
Then we find as f2−1 ◦ f1 the isometries of R: x 7→ ax + b met a2 = 1.
We can also do this in higher dimension. Let V be a plane in R3 . A parametrisation then
corresponds to the choice of a vector v0 in V , and a basis (v1 , v2 ) of the plane V − v0 parallel to
V that passes through 0. Two parametrisations then differ by an affine map of R2 : x 7→ Ax + b,
met A ∈ GL2 (R) en b ∈ R2 . The matrix A is the matrix that expresses one basis in terms of the
other. These transformations form a group, Aff 2 (R).
The principle here is: groups occur as differences between parametrisations (isomorphisms)
that preserve a certain “structure”. For isometries it is “distance”. Sometimes it is not so easy to
say what that structure exactly is, for example for affine transformations.
If we consider planes in R3 containing 0, then we preserve the vector space structure, and
we find the group GL2 (R) of automorphisms of the R-vectorspace R2 . If we want moreover to
preserve distances, then. . .

3.4 Projective transformations of P1 (R)

We have already seen that in P2 (R) projections between lines give bijections. Now we are
interested in the formulas that we obtain in that way.
There are at least two ways to do this: from R2 , with inhomogeneous coordinates, as in
Stillwell, §5.5, and with P2 (R) itself, using homogeneous coordinates, as in, §7.3.
The first way. Consider the lines L1 en L2 given by the equations x = 1 en y = 1, and the
projection p : L1 − {(1, 0)} → L2 from 0. Then p(1, t) = (1/t, 1). If we parametrise L1 and L2
by f1 : R → L1 and f2 : R → L2 , by choosing vectors and directions vectors, then we find a map
(f1 (x) = (1, cx + d) and f2−1 (y, 1) = ay + b):

R → R, x 7→ (1, cx + d) 7→ (1/(cx + d), 1) 7→ a/(cx + d) + b

that is, a fractional linear transformation as in Stillwell 5.5 and 5.6, a map of the form:
ax + b
x 7→ , met ad − bc 6= 0.
cx + d
These transformations are invertible, and closed under composition: they form a group, the group
of projective transformations of P1 (R). But there is a problem: we do not want to divide by 0, so
we have to specify each time the domain of our map. The second approach solves this problem
(it makes it disappear).

9
 Second way. We use that P1 (R) is the set of lines through 0 in R2 . For M = ( ac db ) in GL2 (R)
we have the map:
R2 → R2 , x 7→ M x
that sends lines through 0 to lines through 0, and hence gives us a map from P1 (R) to P1 (R).
Written out in detail:
(x0 : x1 ) 7→ (ax0 + bx1 : cx0 + dx1 ).
In inhomogeneous coordinates this is:

x = (x : 1) 7→ (ax + b : cx + d),

where we can write:

ax + b
x = (x : 1) 7→ (ax + b : cx + d) = , if cx + d 6= 0.
cx + d
We see: these are the same transformations as the fractional linear transformations from R to R,
but in homogeneous coordinates they are linear.
We also see that M and M 0 in GL2 (R) give the same projective transformation on P1 (R)
precisely when there is a k in R∗ with M 0 = kM . So the group of projective transformations of
P1 (R) is the same as the quotient group PGL2 (R) of GL2 (R) by the subgroup of scalar matrices
( k0 k0 ) (this is just for information to those who know about groups and subgroups and quotients).
Now we can write the transformations as fractional linear, and do computations with them, if
we define: x = (x : 1), for x ∈ R, ∞ = (1 : 0), and, for ad − bc 6= 0, (a∞ + b)/(c∞ + d) = a/c
if c 6= 0 and ∞ if c = 0.
Finally: every x 7→ (ax + b)/(cx + d) can be obtained by composition of transformations of
the form

• x 7→ x + b, translation;

• x 7→ ax, scaling;

• x 7→ 1/x, inversion.

3.5 Stillwell, 5.7–5.9, the cross ratio

Main points:

• the cross ratio is invariant under projective transformations of P1 (R);

• a permutation σ of P1 (R) is a projective transformation if and only if σ preserves the cross

ratio;

10
 • every invariant of 4 points on P1 (R) is a function of the cross ratio.
Compare this to the role of distance in the context of isometries of R2 .
Stillwell defines the cross ratio by a formula, without much motivation (like a magician, he
pulls it out of a hat). Then he deduces properties of it, and then uses those for example for the “3
points map theorem”. We are going to do this is the opposite order.
Still, let us first look at the definition. For p, q, r, s in P1 (R), distinct:
r−p s−q (r − p)/(r − q)
[p, q; r, s] := · = ∈ R − {0, 1}
r−q s−p (s − p)/(s − q)
This is a ratio of ratios: that explains the name “dubbelverhouding” in Dutch. Notice: one of
p, q, r, s may be ∞.

3.5.1 Theorem. (3 point map theorem) Let p, q, r in P1 (R) be distinct. Then there is a unique
projective transformation f of P1 (R) with f (p) = ∞, f (q) = 0 and f (r) = 1.

Proof. We follow the proof of the theorem that every isometry in R2 is the composition of at
most three line-reflections.
First the existence of f . If p = ∞, then let f1 = Id. Otherwise, let f1 : x 7→ 1/(x − p).
Then f1 (p) = ∞. Now the problem is reduced to an affine transformation! Let q1 = f1 (q) and
r1 = f1 (r). Let f2 : x 7→ x − q1 . Then, under f2 ◦ f1 : (p, q, r) 7→ (∞, 0, r2 ), with r2 = f2 (r1 ).
Let f3 : x 7→ x/r2 , en f = f3 ◦ f2 ◦ f1 . This one works.
Now the uniqueness: . . . 

3.5.2 Remark. We can also prove Theorem 3.5.1 in homogeneous coordinates: there is a basis
(v1 , v2 ) of R2 , unique up to a common scaling factor, such that the homogenous coordinates of
p, q, r with respect to this basis are (1 : 0), (0 : 1) en (1 : 1).

3.5.3 Definition. Let p, q, r, s in P1 (R) be distinct. Let f be the unique projective transformation
of P1 (R) that sends (p, q, r) to (∞, 0, 1). Then we define:

[p, q; r, s] = f (s).

Note that [p, q; r, s] is an element of P1 (R) − {∞, 0, 1}, hence we can write it as (x : 1) for a
unique x in R − {0, 1}.

We can now compute [p, q; r, s], by following the proof of Theorem 3.5.1:

x 7→ 1/(x − p) : (p, q, r, s) 7→ (∞, 1/(q − p), 1/(r − p), 1/(s − p)),

x 7→ x−1/(q−p) : (∞, 1/(q−p), 1/(r−p), 1/(s−p)) 7→ (∞, 0, 1/(r−p)−1/(q−p), 1/(s−p)−1/(q−p)),

and do the last step yourself.

11
3.5.4 Theorem. (invariance) Let g be a projective transformation of P1 (R), and p, q, r, s in
P1 (R), distinct. Then:
[g(p), g(q); g(r), g(s)] = [p, q; r, s].
In words: projective transformations leave the cross ratio invariant.

Proof. Let f be the unique projective transformation with (g(p), g(q), g(r)) 7→ (∞, 0, 1).
Then. . . . 

3.5.5 Theorem. (4th point determination) Let p, q, r in P1 (R) be distinct, and y ∈ R − {0, 1}.
Then there is a unique s in P1 (R) with [p, q; r, s] = y.

Proof. We can do this by solving the equation:

r−p x−q
· = y.
r−q x−p
But then we have to distinguish cases (for example, what if q = ∞?).
So I prefer the following way. Let f be the unique projective transformation with
(p, q, r) 7→ (∞, 0, 1). Then the equation is: f (s) = y. . . 

3.5.6 Theorem. Let f : P1 (R) → P1 (R) be a bijection (this is rather wild: no continuity, not
any given form for f !) with the property: for all p, q, r, s in P1 (R) distinct we have:

[f (p), f (q); f (r), f (s)] = [p, q, r, s].

Then f is a projective transformation.

Proof. Let g be the unique projective transformation that sends (f (∞), f (0), f (1)) to (∞, 0, 1).
Then:
g ◦ f : (∞, 0, 1) 7→ (∞, 0, 1)
and g ◦ f preserves cross rations, because. . . . 

Conclusion. The role played by the notion of distance for isometries is played by the cross ratio
for projective transformations of P1 (R).

3.5.7 Theorem. (fundamental invariant) Let X be a set, and

I : {(p, q, r, s) : p, q, r, s in P1 (R) distinct} → X

be a function that is invariant under projective transformations. Then there is a unique function
I : R − {0, 1} → X such that for all (p, q, r, s):

I(p, q, r, s) = I([p, q; r, s]).

12
Proof. Let p, q, r, s in P1 (R) be distinct. Let g be the unique projective transformation with
(p, q, r) 7→ (∞, 0, 1). Then, by definition, g(s) = [p, q; r, s]. Hence (invariance of I):

I(p, q, r, s) = I(g(p), g(q), g(r), g(s)) = I(∞, 0, 1, [p, q; r, s]).

Then I : R−{0, 1}, x 7→ I(∞, 0, 1, x) has the required property. The uniqueness follows from
the fact that each x in R − {0, 1} is a cross ratio. 

3.6 Homework
1. Read 5.7–5.9 and 7.3 in Stillwell. Do exercise 5.7.2.

2. Make a geogebra file with a tiled floor and let it compute some cross ratios in it (there is a
command in geogebra for cross ratio).

3. For each permutation σ of {∞, 0, 1} (points on P1 (R)) give the unique projective trans-
formation f such that f (∞) = σ(∞), f (0) = σ(0) en f (1) = σ(1). Example: x 7→ 1/x
exchanges ∞ and 0 and fixes 1.

4. Do the exercises in Stillwell §5.8, in 2 ways. By computing with the expression of the
cross ratio, and by applying the definition of the cross ratio of this lecture.

5. Give an invariant for 3 points on R for affine transformations. Is every bijection f : R → R

that preserves this invariant an affine transformation?

6. Somewhat deeper (testing your understanding): for how many points in P2 (R) does it
make sense to define an invariant under projective transformations, and how can one then
do this?

13
14
4 Lecture 10
4.1 Odds and ends from last week
1. Treat the remaining material of last week, starting at Theorem 3.5.5.

2. Go over the homework.

4.2 Today’s program

We discuss Chapter 7 of Stillwell. Because of time limitations, we restrict ourselves to §7.1–7.3,
but we give some real definitions and theorems, whereas Stillwell does not.
The subject is Felix Klein’s “Erlanger Program”. Erlangen is the place where he became
professor, in 1872 at age 23. In his inaugural lecture (“oratie” in Dutch), he gave a new view on
geometry, in terms of symmetry and the theory of groups. Well, this point of view was not really
new, it had been “in the air”, but he gave a precise formulation (well, precise. . . , nowadays, the
language of set theory makes that easier).
To put the whole idea in a nutshell:

1. each geometry has its transformations that preserve its defining data, we call them the
“automorphisms” of the geometry;

2. the set of these transformations is closed under composition and taking inverses (by defin-
ition, automorphisms have inverses), and contains the identity map, so they form a group;

3. the geometry in question can then be considered as the study of properties that are invari-
ant under the group of automorphisms.

Let us give some examples. Before we go into detail in the next sections, les us have a look at
the introduction of: http://robotics.stanford.edu/˜birch/projective/

4.3 Projective geometry

We consider the real projective plane Π = P2 (R), with its collection Λ of subsets called Lines.

4.3.1 Definition. An automorphism of P2 (R) is a bijective map g : P2 (R) → P2 (R) such that
for all subsets L of P2 (R):

L is a Line if and only if g(L) := {g(P ) : P ∈ L} is a Line.

We let Aut(P2 (R)) denote the set of automorphisms of P2 (R).

15
4.3.2 Remark. In other words, an automorphism of P2 (R) is a bijection g : P2 (R) → P2 (R) that
preserves the collection Λ of Lines.

4.3.3 Proposition. The identity map idP2 (R) is in Aut(P2 (R)). For all g1 and g2 in Aut(P2 (R)),
g1 ◦ g2 is in Aut(P2 (R)). For all g in Aut(P2 (R)), g −1 is in Aut(P2 (R)).

Proof. Let us write out a proof of the last statement. So let g be in Aut(P2 (R)). As g is bijective,
we have the inverse map g −1 : P2 (R) → P2 (R). Let L be a subset of P2 (R). If L is a Line, then
we must show that g −1 (L) is a Line, and if g −1 (L) is a line, then we must show that L is a line.
Note that we have:

g(g −1 (L)) = g({g −1 P : P ∈ L}) = {g(g −1 P ) : P ∈ L} = {P : P ∈ L} = L.

Now suppose that L is a line. Then g −1 L is a Line because of the “if” in Definition 4.3.1, applied
to g and g −1 (L). Now suppose that g −1 (L) is a Line. Then L is a line because of the “only if” in
Definition 4.3.1, applied to g and g −1 (L). 

4.3.4 Remark. I should admit that I manipulate the term “if and only if” without constantly
realising what the meanings of the “if” and especially of the “only if” are. I find it easier to think
in terms of implications (“if A then B”, “if B then A”) written given by arrows A ⇒ B, A ⇐ B,
and of equivalence A ⇔ B.

4.3.5 Remark. One could wonder if we get the same notion of automorphism if we demand
only one implication: if L is a Line then g(L) is a Line. We will see that this is indeed the case.
But the reason for stating Definition 4.3.1 as we do is to guarantee that for all automorphisms g,
g −1 is also an automorphism.

We have many examples of automorphisms of P2 (R): the projective transformations given

by multiplication by an invertible 3 by 3 matrix M with real coefficients (notation: GL3 (R)). We
write out explicitly what these transformations are. We write
 
a b c
M = d e f 
 

g h i

Then M gives the map M ·, multiplication by M :

       
x a b c x ax + by + cz
M · : R3 → R3 , y  7→ d e f  · y  = dx + ey + f z  .
       

z g h i z gx + hy + iz

16
Such a map M · maps lines in R3 through 0 to lines in R3 through 0. The inverse of M · is M −1 ·,
hence the inverse also has this property. So, M · gives us a bijection from P2 (R) to itself. In
homogeneous coordinates it is:

M · : P2 (R) → P2 (R), (x : y : z) 7→ (ax + by + cz : dx + ey + f z : gx + hy + iz).

In inhomogeneous coordinates it is:

 
ax + by + c dx + ey + f
(x, y) = (x : y : 1) 7→ (ax+by+c : dx+ey+f : gx+hy+i)“=”. , ,
gx + hy + i gx + hy + i
where we should be careful with the last “=” because of the possible division by 0. Especially
so because in this case we have not just one point at infinity (as in the case of projective trans-
formations of P1 (R)), but a whole Line.
As M · and its inverse map planes in R3 containing 0 to planes in R3 containing 0, these maps
M · : P2 (R) → P2 (R) are in Aut(P2 (R)). The following theorem shows that these are in fact all
automorphisms. But we will see that this depends on a special property of R, and that with R
replaced by C the outcome is different.

4.3.6 Theorem. Let g be in Aut(P2 (R)). Then there is an M in GL3 (R) such that g is the
projective transformation given by M .

Proof. Consider the standard 4 Points P1 = (1 : 0 : 0), P2 = (0 : 1 : 0), P3 = (0 : 0 : 1)

and P4 = (1 : 1 : 1). Then g(P1 ), g(P2 ), g(P3 ) and g(P4 ) are 4 Points of which no 3 are on a
Line. Therefore, there exists an M , even unique up to scaling (M 0 = kM with k ∈ R∗ ), such
that M ·g(P1 ) = P1 , M ·g(P2 ) = P2 , M ·g(P3 ) = P3 and M ·g(P4 ) = P4 .
Let f : P2 (R) → P2 (R) be the map P 7→ M ·g(P ). Then f ∈ Aut(P2 (R) (why?), and f fixes
the points P1 , . . . , P4 .
Note that (1 : 0 : 0) and (0 : 0 : 1) lie on the Line (0 : 1 : 0)⊥ . Hence f fixes the line
(0 : 1 : 0)⊥ , and on that line it fixes (0 : 0 : 1), (1 : 0 : 0), and also (1 : 0 : 1) (draw a picture!).
Therefore, for every x in R, there is a unique y in R such that f (x : 0 : 1) = (y : 0 : 1). This
defines a function σ : R → R, bijective.
We will now show that σ is an automorphism of the field R: σ is bijective and for all x and
y in R we have σ(x + y) = σ(x) + σ(y) and σ(xy) = σ(x)σ(y). The proof of this is a simple
application of the geometric constructions of addition and multiplication as in Stillwell §1.4.
These constructions are given in the geogebra files addition_projective.ggb,
addition_euclidean.ggb, multiplication_projective.ggb and
multiplication_euclidean.ggb. These files probably also help you becoming
more familiar with going back and forth between the euclidean and projective perspectives.

17
 How do these constructions show that σ(x + y) = σ(x) + σ(y) and σ(xy) = σ(x)σ(y)? Well,
apply f to all the Points and Lines in the constructions, and use the definitions. Here are some
details. By definition, f (x : 0 : 1) = (σ(x) : 0 : 1), and f (y : 0 : 1) = (σ(y) : 0 : 1). We see
that f (0 : x : 1) = (0 : σ(x) : 1) and that f (y : x : 1) = (σ(y) : σ(x) : 1), and finally that
(σ(x + y) : 0 : 1) = f (x + y) : 0 : 1) is the intersection of the Line through (1 : −1 : 0) and
(σ(y) : σ(x) : 1) and the line (0 : 1 : 0)⊥ , which is the Point (σ(y) + σ(x) : 0 : 1).
In the previous paragraph we have also seen that for all x and y in R we have
f (x : y : 1) = (σ(x) : σ(y) : 1). It is a good exercise to show that we have, for all (x : y : z) in
P2 (R), that f (x : y : z) = (σ(x) : σ(y) : σ(z))
Finally, we conclude by applying the following theorem. It gives us that f = idP2 (R) , hence
that M · is the inverse of g. But then g = M −1 ·. 

4.3.7 Theorem. The identity is the only automorphism of the field R.

Proof. Let σ be an automorphism of the field R. Then σ(0) = σ(0 + 0) = σ(0) + σ(0),
hence (subtract σ(0) from both sides) σ(0) = 0. We have σ(1) = σ(1·1) = σ(1)·σ(1),
hence (subtract σ(1) from both sides): σ(1)·(σ(1) − 1) = 0, hence σ(1) = 0 or
σ(1) = 1. As σ is bijective, and already σ(0) = 0, we conclude that σ(1) = 1. Then,
σ(2) = σ(1 + 1) = σ(1) + σ(1) = 1 + 1 = 2. By induction, we get σ(n) = n for all
n ∈ Z≥0 . For all x in R we have 0 = σ(0) = σ(x + (−x)) = σ(x) + σ(−x), hence
σ(−x) = −σ(x). Hence we have, for all n ∈ Z, σ(n) = n. For all x in R with x 6= 0 we
have 1 = σ(1) = σ(x·x−1 ) = σ(x)·σ(x−1 ), hence: σ(x−1 ) = σ(x)−1 . For all a and b in Z with
b 6= 0 we find: σ(a/b) = σ(a·b−1 ) = σ(a)·σ(b)−1 ) = a/b. So σ fixes all elements of Q.
Let us now show that σ preserves the ordering of R. So, let x and y be in R, with
x ≤ y. Then y − x ≥ 0, hence there is a z in R with y − x = z 2 . Then we have
σ(y) − σ(x) = σ(y − x) = σ(z 2 ) = σ(z)2 , hence we see that σ(x) ≤ σ(y). So, if x ≤ y,
then σ(x) ≤ σ(y). As for each x and y in R precisely one of the three statements x < y, x = y,
x > y holds, we have σ(x) ≤ σ(y) ⇔ x ≤ y.
Now let x ∈ R such that x 6∈ Q. Then x divides Q into two subsets: {y ∈ Q : y < x}
and {y ∈ Q : y > x}, and x is the unique real number that is greater than each element of
{y ∈ Q : y < x} and smaller than each element of {y ∈ Q : y > x}. It follows that σ(x) = x.


4.3.8 Remark. The field C does have nontrivial automorphisms, for example the complex con-
jugation z = a + bi 7→ z = a − bi (a and b are in R), and many more, but for showing that one
needs more algebra. In the homework you will show that this leads to automorphisms of P2 (C)
that are not given by an element of GL3 (C).

18
4.3.9 Klein’s program for P2 (R), cross ratio for 5 points in P2 (R)

Let us continue with Klein’s program for P2 (R). Its group of automorphisms is the group of
projective transformations given by GL3 (R). The next step is then the study of properties that
are preserved by the group of projective transformations.
For 2 distinct points P1 and P2 in P2 (R), there is a unique Line L containing them, and
therefore, for any projective transformation g of P2 (R), the Line through g(P1 ) and g(P2 )
(distincts because g is bijective!), is g(L). There is a projective transformation g such that
g(P1 ) = (1 : 0 : 0) and g(P2 ) = (0 : 1 : 0), so there is no R-valued invariant in this case
(except constant ones).
For 3 distinct points P1 , P2 , P3 , there are two possibilities: P3 is on the unique Line con-
taining P1 and P2 (we say that P1 , P2 , P3 are collinear), or not. Collinearity is preserved by
projective transformations. If they are collinear, then there is a projective transformation g such
that g(P1 ) = (1 : 0 : 0), g(P2 ) = (0 : 1 : 0), and g(P3 ) = (1 : 1 : 0). If they are not collinear,
then there is a projective transformation such that f (P1 ) = (1 : 0 : 0), f (P2 ) = (0 : 1 : 0),
f (P3 ) = (0 : 0 : 1). Again, no interesting invariants.
For 4 distinct points P1 , . . . , P4 , no 3 of which collinear, there is a unique projective trans-
formation such that f (P1 ) = (1 : 0 : 0), f (P2 ) = (0 : 1 : 0), f (P3 ) = (0 : 0 : 1) and
f (P4 ) = (1 : 1 : 1). No interesting invariant in this case. However, if they are collinear, then
there is a projective transformation g such that g(P1 ) = (1 : 0 : 0), g(P2 ) = (0 : 1 : 0),
g(P3 ) = (1 : 1 : 0), and a unique x in R such that g(P4 ) = (x : 1 : 0), and we have seen last
week that this x is the cross ratio of (P1 , . . . , P4 ). Let us still denote it by [P1 , P2 ; P3 , P4 ]. As we
see now, this is the first interesting R-valued invariant of 4 collinear points in P2 (R).
As suggested in the previous homework assignment, we can attach a “super cross ratio” to 5
points P1 , . . . , P5 in P2 (R), of which no 3 are collinear. Let f : P2 (R) → P2 (R) be the unique
projective transformation such that f (P1 ) = (1 : 0 : 0), f (P2 ) = (0 : 1 : 0), f (P3 ) = (0 : 0 : 1)
and f (P4 ) = (1 : 1 : 1). Then there is a unique (x, y) in R2 with f (P5 ) = (x : y : 1). We call
this (x, y) the cross ratio of (P1 , . . . , P5 ). It has the property that x and y are in R − {0, 1}, and
x 6= y (draw a picture!). By its very definition, it is preserved by all automorphisms of P2 (R). I
should admit here that I came up with this definition myself, and so it should be considered as
not more than my own hobby.

4.3.10 Klein’s program for P2 (F ) with F any field

This section is not part of the material for the 2nd partial exam. It is included for the sake of
completeness.
Let F be a field. Then we have the projective plane P2 (F ), because we have the F -vector

19
space F 3 , with its one-dimensional subspaces (Points) and two-dimensional subspaces (Lines).
Let Aut(F ) be the group of automorphisms of the field F , and let PGL3 (F ) denote the group of
projective transformations of P2 (F ). Then the group Aut(P2 (F )) of automorphisms of P2 (F )
contains both PGL3 (F ) and Aut(F ), where for σ in Aut(F ) and (x : y : z) in P2 (F ) we put
σ(x : y : z) = (σ(x) : σ(y) : σ(z)). Our proof of Theorem 4.3.6 shows that for each element
f of Aut(P2 (F )) there are unique σ in Aut(F ) and g in PGL3 (F ) such that for all P in P2 (F )
we have f (P ) = g(σ(P )). In technical terms, Aut(P2 (F )) is the semi-direct product of Aut(F )
by PGL3 (F ) (just as isometries are uniquely the composition of an orthogonal linear map and a
translation).
The cross ratio of 4 distinct collinear points P1 , . . . , P4 is then an invariant for PGL3 (F ), but
for σ in Aut(F ) we have:

[σ(P1 ), σ(P2 ); σ(P3 ), σ(P4 )] = σ([P1 , P2 ; P3 , P4 ]).

We conclude that PGL3 (F ) is the group of those transformations in Aut(P2 (F )) that preserve
the cross ratio. This is nice, because now we understand more about the group of projective
transformations of P2 (F ).

4.4 Euclidean geometry

We consider the plane R2 with all the extra data that make it into a Hilbert plane: the collection
of subsets called lines, the notion of betweenness, congruence of line segments, congruence of
angles. We consider bijective maps g : R2 → R2 that preserve all this structure. In particular,
such a g has to preserve distances because it preserves congruence of line segments. Hence, such
a g is an isometry: for all x and y in R2 , we have kg(x) − g(y)k = kx − yk (here, for x = (x1 , x2 )
p
in R2 , kxk = x21 + x22 denotes the length of x). We have seen in Stillwell, §3.7, the “three
reflections theorem”, that g is then a composition of at most 3 reflections about lines in R2 . But
then it follows that g preserves all the other structures (lines, betweenness, angles) as well. So
we have proved the following theorem.

4.4.1 Theorem. The automorphism group of the Hilbert plane R2 is the group Isom(R2 ) of
isometries of R2 .

In the light of Klein’s program, this makes us wonder if the notions of lines, betweenness and
angles in the Hilbert plane R2 can be expressed in terms of distance. This is one of the homework
exercises.

20
4.5 Affine geometry
Here we are in an interesting situation, where we have already the notion of “affine transforma-
tions”, but not yet the notion of a corresponding geometry.
An affine transformation of R is a map f : R → R of the form x 7→ ax + b, where a and b
are in R, with a 6= 0. Note that such maps are invertible, hence bijective. Note also that a and b
are uniquely determined by f .
Of course, one cannot hope to do geometry just in dimension one, we have to consider a
plane. An affine transformation of R2 is a map f : R2 → R2 such that there exist an element A
of GL2 (R) and a b in R2 such that for all x in R2 :
f (x) = Ax + b.
In the homework, you will show that such an f is invertible, and that the inverses and composi-
tions of such f ’s are again affine transformations. Hence they constitute a group of transforma-
tions. We call it the affine group of R2 and denote it by Aff(R2 ).
Now what is the geometry? One notion that is preserved is that of lines. Indeed, let L be
a line in R2 , and let f be as above. Then there are x0 and d in R2 , with d 6= 0, such that
L = {x0 + td : t ∈ R}. Then
f (L) = {f (x0 + td) : t ∈ R} = {A(x0 + td) + b : t ∈ R}
= {Ax0 + tAd + b : t ∈ R} = {t(Ad) + (Ax0 + b) : t ∈ R},
hence f (L) is the line with “steunvector” Ax0 + b and direction vector Ad. Just as in the case of
P2 (R), it is a remarkable fact that this notion just by itself defines the notion of affine geometry
on R2 .

4.5.1 Theorem. Let g : R2 → R2 be a bijective map that preserves the collection of lines in R2 .
Then g is an affine transformation.

Proof. We argue as in the proof of Theorem 4.3.6. Let g be as in the statement. Consider the
3 points P1 = (0, 0), P2 = (1, 0) and P3 = (0, 1). There is a unique affine transformation
x 7→ Ax + b that maps f (P1 ) to P1 , f (P2 ) to P2 and f (P3 ) to P3 : first translate by −f (P1 ) to
move f (P1 ) to P1 . Then the images of f (P2 ) and f (P3 ) form a basis of R2 , and that gives us a
unique A that finishes the job.
Let f : R2 → R2 be the composition of g with this affine transformation. It fixes (0, 0), (1, 0)
and (0, 1). For all x in R, there is a unique y in R such that f (x, 0) = (y, 0). This defines a
map σ : R → R that is an automorphism of R as a field. One shows that for all (x, y) in R2 ,
f (x, y) = (σ(x), σ(y)). As σ is the identity automorphism of R (Theorem 4.3.7), f = idR2 and
g is an affine transformation. 

21
4.5.2 Affine transformations of F 2 with F any field

This section is not part of the material for the 2nd partial exam. It is included for completeness.
Let F be a field. Then we have the affine plane F 2 , with its collection of lines. Let Aut(F ) be
the group of automorphisms of F . Then the group of bijections of F 2 that preserve collinearity
contains the affine transformations but also the automorphisms of F . We want to describe the
group of affine transformations as a group that leaves a property invariant. What can that property
be? In the case of P2 (F ) we found the cross ratio. In this case we simply find the ratio of three
distinct points on a line, as we will now show.
Let P1 , P2 and P3 be three distinct points in F 2 , that are collinear. Then there is an af-
fine transformation g such that g(P1 ) = (0, 0), g(P2 ) = (1, 0), and a unique x in F such that
g(P3 ) = (0, x). This x does not depend on the choice of g, and we call it the ratio of P1 , P2 and
P3 . The group of affine transformations of F 2 is then the group of bijections of F 2 that preserve
collinearity and the ratio.
# »
Note: the ratio of P1 , P2 and P3 is the element λ of F such that the vector P1 P3 is λ times the
# »
vector P1 P2 . This is why the group of affine transformations is so closely related to the structure
of F -vector space of F 2 .

4.5.3 Comparison of some geometries and their groups

Let us now compare the automorphism groups of euclidean geometry, affine geometry and pro-
jective geometry, describe a few more groups, and find what geometries they correspond to.
The group of automorphisms of the euclidean plane R2 is:

Isom(R2 ) = {x 7→ Ax + b : A ∈ GL2 (R) orthogonal, and b ∈ R2 }.

Let us denote by GL2 (R)+ the group of g in GL2 (R) with det(g) > 0. These are precisely the g
that preserve the standard orientation of R2 . Then Isom(R2 ) has the subgroup:

Isom(R2 )+ = {x 7→ Ax + b : A ∈ GL2 (R)+ orthogonal, and b ∈ R2 }.

which is know as the group of direct isometries of R2 . Its elements are of the form x 7→ Ax + b
with A a rotation.
The automorphism group of the affine plane R2 is:

Aff(R2 ) = {x 7→ Ax + b : A ∈ GL2 (R), and b ∈ R2 }.

This group contains Isom(R2 ), and this corresponds to the fact that affine geometry is obtained
from euclidean geometry by forgetting everything except collinearity. In between is the group

22
Sim(R2 ) of similarities, that is, affine transformations that preserve angles: the x 7→ Ax + b with
A of the form kB with k ∈ R∗ and B orthogonal. This group is generated by translations, re-
flections about lines through 0, and scaling (with respect to 0). We have the subgroups Aff(R2 )+
and Sim(R2 )+ .
We do not want to consider groups smaller than Isom(R2 )+ because we feel that for geometry
at least all points in R2 should be equivalent, and also all directions, so we want at least all
translations and all rotations.
It is interesting to connect the affine transformations of R2 with the projective transformations
of P2 (R). Recall that our standard way to view R2 as a subset of P2 (R) is by sending (x, y) in
R2 to (x : y : 1) in P2 (R). The complement of R2 in P2 (R) is then the line at infinity, the Line
L∞ = {(x : y : 0) : (x, y) ∈ R2 − {0}}. Each projective transformation g of P2 (R) with
g(L∞ ) = L∞ gives an automorphism of R2 as affine plane. Let us compute what this looks like
in terms of matrices. Recall that projective tansformations of P2 (R) are of the form:

M · : P2 (R) → P2 (R), (x : y : z) 7→ (ax + by + cz : dx + ey + f z : gx + hy + iz).

For (x : y : 0) in L∞ we have:

M · : P2 (R) → P2 (R), (x : y : 0) 7→ (ax + by : dx + ey : gx + hy).

We see that g preserves L∞ if and only if g = h = 0. Then i 6= 0 and as M and i−1 M give the
same projective transformation, we may assume that i = 1. Then we have:
 
a b c
M = d e f 
 

0 0 1

with ae − db 6= 0. We find, for all (x, y) ∈ R2 :

     
a b c x ax + by + c
d e f  · y  = dx + ey + f  .
     

0 0 1 1 1

We can write this as:

! ! ! !
ax + by + c a b x c
= · + .
dx + ey + f d e y f

So we see that the projective transformations that preserve L∞ correspond exactly to the af-
fine transformations of R2 , and we have found that the affine transformation x 7→ Ax + b

23
of R2 can be written as a matrix multiplication in dimension 3. We see that Aff(R) is con-
tained in Aut(P2 (R)) in a very precise way. The description of affine transformations by 3
by 3 real matrices is used in many places, for example in the postscript language that most
printers use. See for example §4.3.3 of the PostScript Language Reference book, 3rd edi-
tion: http://www.adobe.com/products/postscript/pdfs/PLRM.pdf. There,
the matrix M is the transpose of what we have, because they consider the transformation ·M that
multiplies row vectors from the right.

4.6 Homework
1. Read Stillwell §7.1–7.3, and do the exercises.

2. Show that the map g : P2 (C) → P2 (C), (x : y : z) →7 (x : y : z) is bijective and

2
preserves Lines, hence is an automorphism of P (C). Show that it is not given by an
element of GL3 (C).

3. We consider the Hilbert plane R2 . Can one express the the notions of lines, betweenness
and angles in the Hilbert plane R2 in terms of distance?

4. Show that inverses and compositions of affine transformations of R2 are affine transform-
ations.

24
5 Lecture 11
5.1 Remaining material from last week
1. Treat the remaining material of last week, starting at Theorem 4.3.7.

2. Go over the homework.

5.2 Today’s program

We finally get to hyperbolic geometry, a natural geometry that satisfies all of Hilbert’s axioms
(pages 43–45 of Stillwell) except I4 on the existence of a unique parallel; for this reason, it is
called a non-euclidean geometry. The existence of this geometry implies that axiom I4 cannot
be derived from the others.
This material is treated in Chapter 8 of Stillwell, but we do not follow his exposition so
closely. We first study the complex projective line P1 (C) and its projective transformations and
the complex conjugation, and only then restrict to automorphisms coming from the real projective
line, and to the upper half plane.

5.3 The complex projective line, and its Moebius transformations

We have already defined P1 (F ) for any field F , so a special case of this is then P1 (C): the set
of one-dimensional subspaces in the C-vector space C2 . It is not easy to draw pictures of what
happens in C2 . But we have homogeneous coordinates:

P1 (C) = {(x0 : x1 ) : (x0 , x1 ) ∈ C2 − {(0, 0)}} = {(z : 1) : z ∈ C} ∪ {(1 : 0)} = C ∪ {∞}.

Very often, P1 (C) is called the Riemann sphere, because topologically (one has to specify the
topology first), it is homeomorphic to the 2-dimensional sphere S 2 . It would be nice to treat more
details here, but there is no time. A nice reference for this is Niels uit de Bos’s bachelor thesis:
http://www.math.leidenuniv.nl/nl/theses/312/.
It is important to recall that whereas the real projective plane P2 (R) is obtained from R2 by
adding a whole horizon (a real projective Line), P1 (C) is obtained from C by adding the point ∞.
Each invertible complex 2 by 2 matrix A = ( ac db ) in GL2 (C) gives a projective transforma-
tion:
az + b
A· : P1 (C) → P1 (C), (x0 : x1 ) 7→ (ax0 + bx1 : cx0 + dx1 ), z 7→ ,
cz + d

25
with the usual interpretation of the last fraction regarding ∞. And we also have the complex
conjugation:
(x0 : x1 ) 7→ (x0 : x1 ), z 7→ z, ∞ 7→ ∞.
Composing complex conjugation with A· (in this order, first complex conjugation) gives:
az + b
(x0 : x1 ) 7→ (ax0 + bx1 : cx0 + dx1 ), z 7→ .
cz + d
It is a homework exercise for to show that composing in the other order also gives a transform-
ation of this form. The group of Moebius transformations of P1 (C) is defined to be the set of
transformations of P1 (C) that are of one of the forms z 7→ (az + b)/(cz + d) or (az + b)/(cz + d).
We denote it by M (P1 (C)).
As a complex vector space, C has dimension one, and there is not really any geometry as
intersecting of lines possible. But as a real vector space, it is of dimension two (the complex
plane), and we have geometry of lines and circles.

5.3.1 Definition. A generalised circle in P1 (C) is a circle in C in C, or a line in C together with

∞.

5.3.2 Theorem. The Moebius transformations of P1 (C) preserve the collection of generalised
circles.

Proof. This is proved (well, in a special case) in §8.4 of Stillwell. So we give just a sketch. We
know that the group of Moebius transformations is generated by:

1. the translations (z 7→ z + b),

2. the multiplications (scaling, dilations) (z 7→ az),

3. the complex conjugation (z 7→ z),

4. and the inversion (z 7→ 1/z).

So if we show that all these transformations preserve the collection of generalised circles, then
we are done. The first three types of transformations clearly do what we claim, and for the last
one it is a computation that settles it (Stillwell, page 185).
As the proof has an interesting ingredient: inversion in a circle, we say something about
it. The transformation z 7→ 1/z is called inversion in the unit circle. For z 6= 0 we have:
1/z = z/zz = z/|z|2 , hence the argument is unchanged, but the absolute value is inverted. The
z with |z| = 1 are all fixed.

26
 What we have to show is that this inversion preserves the collection of generalised circles.
So let f denote this map, and let C be a circle in C. We may assume that the center a of C is in
R (why?) and that a ≥ 0 (why?). Let r be the radius. Then C = {z ∈ C : |z − a| = r}. Note
that f 2 = id, hence f = f −1 . Therefore,
 
−1
1 
f (C) = f C = {z ∈ C :  − a = r}
z

where we are sloppy with what happens with ∞. Then. . . . What is important is that f (C)
is a circle if 0 6∈ C, and f (C) is a line if 0 ∈ C, and then it is the line given by the equation
<(z) = 1/2r. Note that this tells us that for L a line in C with 0 6∈ L, f (L) is a circle containing 0.


5.3.3 Remark. Inversion in a circle is a great tool to simplify problems about circles, as some
circles can be made into lines.

Moebius transformations preserve angles. This is defined and proved in §8.5 of Stillwell. The
difficulty here is that we take an infinitesimal perpective on it, as we must not only talk about
angles between lines, but about angles between generalised circles (not at ∞, luckily for us).
This property is in fact true for all maps f : C → C that have a non-zero complex derivative.
Such maps are called conformal. They scale in a non-uniform way, but preserve angles at all
points. For example, see Escher’s picture of the portrait gallery and what is written about it and
very nice animations at http://escherdroste.math.leidenuniv.nl.
Let us do it for our Moebius transformations. Translations, multiplications and complex
conjugation preserve (unoriented) angles (complex conjugation is a reflection, it changes the
orientation). So, we only need to show that z 7→ −1/z preserves angles. Let f denote this map,
and let z0 6= 0 be in C. Here is the computation (and implicitly the definition) that f preserves
infinitesimally angles at z0 . For h in C with |h| small:
−1 1
f (z0 + h) = f (z0 ) + f (z0 + h) − f (z0 ) = f (z0 ) + + =
z0 + h z0
h h h h
= f (z0 ) + = f (z0 ) + 2 + − 2
z0 (z0 + h) z0 z0 (z0 + h) z0
1 1
= f (z0 ) + 2 ·h − 2 ·h2
z0 z0 (z0 + h)

Note that the first term is f (z0 ), the image of z0 , the second term is linear in h, and the third term
is neglegible compared to h if |h| tends to zero. The complex derivative at z0 of our map f is
1/z02 , as is to be expected from f (z) = −1/z. We have defined the notion of angle and proved
the following theorem.

27
5.3.4 Theorem. The Moebius transformations of P1 (C) preserve angles at all points of C that
are not mapped to ∞.

This is nice, but for euclidean (well, Hilbertian) geometry, we need the concept of Lines, Line
segments, and congruence of Line segments, and for that last thing, it is good to have the notion
of length of a line segment. We want these notions to be preserved by the transformations of the
geometry. But if we have scaling transformations z 7→ az with |a| < 1, then no sensible notion
of length can be preserved. So we have to restrict from the group of Moebius transformations
to a smaller group. And we also have to restrict the space P1 (C) to something smaller, because
circles can have no “betweenness” notion.

5.4 The upper half plane with its Moebius transformations

Inside P1 (C) we have the subset P1 (R). This is maybe hard to see from the definition as complex
lines in C2 through zero, but in terms of inhomogeneous coordinates it is a triviality: R ∪ {∞} is
indeed a subset of C ∪ {∞}. The complement of R ∪ {∞} in C ∪ {∞} is the union of the upper
half plane H = {z ∈ C : =(z) > 0} and the lower half plane H = {z ∈ C : =(z) < 0}. We
will concentrate on H; it is the set of Points for our non-euclidean geometry. Before we discuss
its Lines, we decide what should be its transformations, à la Klein.
In the homework you will show that the Moebius transformations of P1 (C) that preserve
P1 (R) are precisely those of the form z 7→ (az + b)/(cz + d) with a, b c and d in R, or of the
form z 7→ (az + b)/(cz + d) with a, b c and d in R. Such Moebius tranformations either preserve
H and H, or exchange them. The ones preserving H are precisely the z 7→ (az + b)/(cz + d)
with a, b c and d in R and ad − bc > 0, together with the z 7→ (az + b)/(cz + d) with a, b c and d
in R with ad − bc < 0. We denote the group of all Moebius transformations of H by M (H), and
the subgroup of all z 7→ (az + b)/(cz + d) with ad − bc > 0 by M (H)+ , and the complement of
M (H)+ in M (H) by M (H)− .
The fact that ∞ is not in H is nice, we do not need to worry about it anymore, our transform-
ations of H do not divide by zero. The next proposition shows that our transformations in M (H)
make H “everywhere the same”, just like the isometries in R2 for euclidean geometry.

5.4.1 Proposition. Let z be in H. Then there is a g in M (H)+ such that g(z) = i.

Proof. Write z = x + yi with x and y in R. Then z 7→ z − x is in M (H)+ and it maps z to yi.

Then the map z 7→ y −1 z maps yi to i. 

The next proposition about the elements g in M (H) such that g(i) = i will show that we have a
good notion of congruence of Line Segments defined in terms of M (H).

28
5.4.2 Proposition. For g in M (H) we have that g(i) = i is equivalent to g being of the form
z 7→ (az + b)/(−bz + a) or z 7→ (−az + b)/(bz + a) with a and b in R and a2 + b2 6= 0.
For g in M (H) we have that g(i) = i and g(R>0 i) = R>0 i is equivalent to g being of one of
the form z 7→ z, z 7→ −1/z, z 7→ −z, z 7→ 1/z.

Proof. For the first statement: a simple computation. For the understanding of the second
a b 2
statement, it is useful to note that ( −b a ) represents a rotation in R , followed by a scaling by
√
a2 + b2 (it is really multiplication by a − bi in C, with C viewed as R2 ). The derivative of
z 7→ (az + b)/(−bz + a) at z = i is (computation. . . ) equal to (a + bi)/(a − bi), a complex
number of absolute value 1, and with argument twice that of a + bi. So z 7→ (az + b)/(−bz + a)
acts on the set of infinitesimal directions at i by rotation by twice the argument of a + bi. It is
clear that one can prove the 2nd statement by a computation. 

5.5 The Lines of hyperbolic geometry

It is time to decide on the Lines of hyperbolic geometry. We present it now without much
motivation, but when we define hyperbolic length later (if there is time for it), we can prove that
these Lines are really the lines in the sense of distance and “triangle equalities”.
We decide that the vertical line R>0 i = {yi : y ∈ R>0 } is a Line. Then, because all g in
M (H) are postulated to be transformations, all images of R>0 i are Lines. The g of the form
z 7→ z + x (with x in R) show that all vertical lines {x + yi : y ∈ R>0 } are Lines. Note that the
generalised circle g(Ri) that they are half of (up to ∞) intersect R perpendicularly. This means
that there are no more lines that are Lines. Then what circles do we get as images? They are of
the form g(Ri) ∩ H, and g(Ri). It is not suprising that these are precisely the circles that intersect
R perpendicularly. Equivalently: the circles with center on R. See the pictures in Stillwell. You
can also watch http://www.youtube.com/watch?v=eAn6NHpBn2c. I do not know
how much of hyperbolic geometry is availabe in geogebra.
To summarise: the Lines in H are the vertical lines {x + yi : y ∈ R> 0} for x in R, and the
half circles {z ∈ H : |z − a| = r} for a in R and r in R>0 .

5.5.1 Theorem. Let z and w be in H, distinct. Then there is a unique Line containing them.

Proof. Let us first show existence. If <(z) = <(w), then there is a vertical line containing
them. Now suppose that <(z) 6= <(w). Then the bisector of z and w (middellloodlijn) is not
parallel to R, hence intersects R, and that gives a point a in R that has equal distances to z and
w, hence a Line containing z and w.

29
 Now uniqueness. If <(z) = <(w) then there is no a in R with |z − a| = |w − a|, hence
the only Lines containing z and w are vertical lines, and there is exactly one of them. If
<(z) 6= <(w). Then there is no vertical line containing them, and there is a unique a in R
such that |z − a| = |w − a|. 

5.6 Back to Hilbert

Let us now equip H with all the structure that is needed to make it make it into a Hilbert plane,
and let us show that all axioms except I4 are satisfied.

1. We have H the set of Points.

2. We have the collection of subsets called Lines. The axioms I1, I2 and I3 are satisfied, but
I4 is not.

3. On each Line we have a betweenness notion: each Line is a homeomorphic image of R.

The axioms B1, B2, B3 and B4 are satisfied (why?). So we have the notions of ray AB for
Points A 6= B and of angle ∠ABC for distinct Points A, B and C with C not on the Line
containing B and A.

4. We have the notion of congruence of angles ∠ABC ∼ ∠A0 B 0 C 0 because our angles have
a value in the interval (0, π). This is consistent with our transformations (they preserve
these values). Axioms C4 and C5 are satisfied.

5. Congruence of Line Segments. We use the transformations, but it would be nicer to have
a value for the length of a segment; we remedy this in the net section. We say that
AB ∼ = A0 B 0 if there is a g in M (H) such that g(A) = A0 and g(B) = B 0 . Then axiom
C2 is satisfied because M (H) is a group. It follows from Proposition 5.4.2 that axioms C1
and C3 are satisfied. Axiom C6 (SAS) is satisfied because there is a g in M (H) that sends
(A, B, C) to (D, E, F ).

6. Axioms E, A and D are also satisfied.

As a consequence we have the following theorem.

5.6.1 Theorem. Axiom I4 cannot be derived from the other axioms in Hilbert’s list of axioms.

30
5.7 Hyperbolic distance
We just imagine that we drive our car on H and that we have to pay toll (rekeningrijden!), and
that the toll per meter depends on the position where we are, but not on the direction in which
we go: the toll per meter is a function T : H → R, let us assume that T is continuous. So if our
trip is described by a function f : [0, 1] → H, say f (t) = (f1 (t), f2 (t)) = f1 (t) + f2 (t)i, with f1
and f2 differentiable with continuous derivative, then the toll we pay is given by:
Z 1 p Z 1
Toll(f ) = 0 2 0 2
T (f (t))· f1 (t) + f2 (t) dt = T (f (t))·|f 0 (t)| dt,
t=0 t=0
p
because our speed vector is the vector f 0 (t) = (f10 (t), f20 (t)) and so our speed is f10 (t)2 + f20 (t)2 .
We want that for any g in M (H), that the trip g ◦ f costs the same as f . This is achieved if the
integrands for Toll(f ) and Toll(g ◦ f ) are equal. Note that (g ◦ f )0 (t) = g 0 (f (t))·f 0 (t) (chain
rule):
Z 1 Z 1
0
Toll(g ◦ f ) = T (g(f (t)))·|(g ◦ f ) (t)| dt, = T (g(f (t)))·|g 0 (f (t))·f 0 (t)| dt.
t=0 t=0

And that means that we want, for all g in M (H), and all z in H, that:

T (g(z))·|g 0 (z)| = T (z).

So, if we decree that T (i) = 1, then there is at most one such a function T by Proposition 5.4.1,
and exactly one by (the proof of) Proposition 5.4.2. Using z 7→ yz and z 7→ z + x, one sees that
T (x + yi) = 1/y.
We define the hyperbolic distance d(z, w) between z and w in H as the minimum toll required
to drive from z to w. By its construction, this is a distance function (triangle inequality), and for
each g in M (H) we have d(g(z), g(w)) = d(z, w). It is also clear that the toll we pay depends
only on the route we take, not on our speed (changing the speed gives another function f ◦s, with
s : [0, 1] → [0, 1], computation left to th reader). To determine d, it suffices to determine d(i, yi),
say for y > 1. It is easy to show that the shortest route is via the imaginary axis: f : [1, y] → H,
f (t) = it. Then we have:
Z y Z y
1
d(i, yi) = T (f (t))·|i| dt = dt = ln(y).
t=1 t=0 t

I stop here, there is still a lot to say, but there is no time now. . .

5.8 Homework
1. Prove that doing first z 7→ (az + b)/(cz + d) and then complex conjugation is also a
Moebius transformation. Is the group of Moebius transformations commutative?

31
 2. Let g be a Moebius transformation of P1 (C). Show that it preserves P1 (R) if and only
if it is of the form z 7→ (az + b)/(cz + d) with a, b, c and d in R, or of the form
z 7→ (az + b)/(cz + d) with a, b, c and d in R.

3. Use the invariance of the hyperbolic distance under M (H) to prove give a general formula
for the hyperbolic distance d(p, q), for p and q in H, distinct. Here is what you will find: if
z and w are the boundary points in P1 (R) of the Line containing p and q, and the ordering
on the Line is (z, p, q, w), then:

d(p, q) = | ln[w, z; p, q]|.

5.9 A few exercises for practicing for the 2nd partial exam
1. (a) Explain how to draw the perspective view of a tiled floor with only a straightedge, if
one tile is already given and both pairs of its opposite sides are not parallel.
(b) And what if exactly one pair of opposite sides that is parallel? (This one is maybe
too hard.)
(c) And what if both pairs of opposite sides are parallel? (This one should be easy.)

2. Let P = (0 : 0 : 1), Q = (6 : 0 : 1), R = (6 : 4 : 1) and S = (0 : 4 : 1) be in P2 (R).

(a) Compute the equations ax+by+cz = 0 (or the homogeneous coordinates (a : b : c)⊥
if you prefer) of the Lines P R and QS, (these Lines are the diagonals of the quadri-
lateral P QRS).
(b) Compute the intersection Point of the two diagonals.
(c) Now interpret the quadrilateral in R2 embedded in P2 (R) by (x, y) 7→ (x : y : 1),
and give a second computationn of the intersection point of the diagonals.

3. The cross ratio of 4 ordered distinct Points p, q, r and s on P1 (R) is defined to be g(s),
where g is the unique projective transformation of P1 (R) that sends (p, q, r) to (∞, 0, 1).
Derive the usual formula
r−p s−q
[p, q; r, s] = ·
r−q s−p
for the cross ratio from this.

4. If you have a photograph with an image of three equally spaced points on a line, how do
you know where the image of the fourth equally spaced point in this sequence must be?
Apply this to the case where the images of the three given points are at positions 0, 3, 5 on
a line.

32
 5. The affine transformations of R2 are the maps x 7→ Ax + b with A in GL2 (R) and b in R2 .
The set of them is denoted by Aff(R2 ).

(a) Show that Aff(R2 ) is a group: contains idR2 , is closed under composition and in-
verses.
(b) Let G be the subset of Aff(R2 ) consisting of those affine transformations that preserve
oriented angles. What are then the conditions on A and b for the affine x 7→ Ax + b
to be in G?
(c) Show that G is a group.
(d) Try to give, a la Hilbert, a set of notions and axioms such that R2 with the appropriate
structure is a model with G as group of automorphisms.2

6. In P1 (C), we have the generalised circles P1 (R) and the unit circle C in C. We want to
find a g in M (P1 (C))+ such that g(P1 (R)) = C.

(a) Show that for three distinct points in C, there is a unique generalised circle that
contains them.
(b) Find the unique g in M (P1 (C))+ that maps (0, 1, ∞) to (−i, 1, i).
(c) Show that g maps P1 (R) to C, and maps i to 0.
(d) Show that g(H) is the open unit disk {z ∈ C : |z| < 1}.

7. Explain in a few (at most 10, say) lines how we know that the parallel axiom cannot be
deduced from the other axioms of Hilbert’s list for euclidean geometry.

2
This is probably a bit hard. If such a questions is in the partial exam, then you will get a list of Hilberts notions
and axioms.

33
 2nd partial exam “geometry”, mastermath, Fall 2013
December 20, 14:00–16:45, Bas Edixhoven
You may write your solutions in Dutch or English.

1. (a) Draw, with only a straightedge, a perspective view of a floor that is tiled with squares.
At least 4 tiles with one common corner should be drawn. Explain the steps in your
construction in the “algebra view style” of geogebra.
(b) Draw, with only a straightedge, a perspective view of a floor that is tiled with reg-
ular 6-gons; draw at least 2 adjacent 6-gons. Hint: use the contruction in (a), and
subdivide some tiles into two triangles.

2. Let P = (1 : −1 : 1), Q = (1 : 1 : 1), R = (−1 : 1 : 1) and S = (−1 : −1 : 1) in P2 (R).

(a) Give the points P 0 , Q0 , R0 and S 0 in R2 that correspond to P , Q, R and S if we embed

R2 in P2 (R) via:
f : R2 → P2 (R), (x, y) 7→ (x : y : 1).
And draw them. They are the corners of the square P 0 Q0 R0 S 0 .
(b) Give the points P 00 , Q00 , R00 and S 00 in R2 that correspond to P , Q, R and S if we
embed R2 in P2 (R) via:

g : R2 → P2 (R), (u, v) 7→ (1 : u : v).

And draw them. They are, in this order, the corners of the square P 00 Q00 S 00 R00 . Note
that in the order P 00 Q00 R00 S 00 we do not get a square.
(c) Figure out what is happening here by computing the images of the sides of the square
P 0 Q0 R0 S 0 under the map

g −1 ◦ f : (x, y) 7→ (x : y : 1) = (1 : y/x : 1/x) 7→ (y/x, 1/x)

that is defined on the set of (x, y) with x 6= 0. Draw these images. What is the image
of the interior of P 0 Q0 R0 S 0 ? Can you understand this by looking at what happens
in R3 ?

3. (a) The cross ratio of 4 ordered distinct Points p, q, r and s on P1 (R) is defined to be
g(s), where g is the unique projective transformation of P1 (R) that sends (p, q, r) to
(∞, 0, 1). Derive the usual formula
r−p s−q
[p, q; r, s] = ·
r−q s−p
for the cross ratio from this.

34
 (b) Use the definition of the cross ratio in (a) to prove that it is invariant under all pro-
jective transformations of P1 (R).
(c) Suppose you have a photograph of a scene in which there are 4 equidistant points P1 ,
P2 , P3 and P4 on a line. Let Q1 , Q2 , Q3 and Q4 be the images of these points in the
photograph. Explain that these 4 points Qi are on a line L.
(d) Suppose that, for some coordinate, Q1 , Q2 and Q4 are at positions 0, 1 and 4 on L.
Then what is the position of Q3 ?

4. Consider the following Lines in the hyperbolic plane H: the lines {z ∈ H : <(z) = 0},
{z ∈ H : <(z) = 1/2}, and the arcs {z ∈ H : |z| = 1} and {z ∈ H : |z| = 2}.

(a) Show that there is a 4-gon in H of which the sum of the angles is less than 360 degrees
(you do not need to compute the sum).
(b) Deduce from this that there are triangles of which the sum of the angles is less than
180 degrees.

5. Explain in a few (at most 10, say) lines why the parallel axiom cannot be deduced from
the other axioms of Hilbert’s list for euclidean geometry.

6. Recall the 3 axioms for projective geometry:

P1 every two distinct Points are contained in a unique Line,

P2 every two distinct Lines contain a unique common Point,
P3 there are 4 distinct Points of which no 3 are collinear (that is, lie on one Line).

Explain why these axioms do not imply that there are at least 8 Points.

Please fill out the evaluation form on the mastermath website. This feedback is important for
the mastermath organisation, and for the teachers.

35
 2e deeltoets “geometry”, mastermath, najaar 2013
20 december, 14:00–16:45, Bas Edixhoven
De uitwerkingen mogen in het nederlands of engels gedaan worden.

1. (a) Teken, met alleen een latje, een plaatje in perspectief van een vloer die betegeld is
met vierkanten. Tenminste 4 tegels met een gemeenschappelijk hoekpunt moeten
getekend zijn. Licht de stappen van je constructie toe in de “algebra view style” van
geogebra.
(b) Teken, met alleen een latje, een plaatje in perspectief van een vloer die betegeld is
met gelijkzijdige 6-hoeken; teken tenminste twee 6-hoeken met een gemeenschap-
pelijke zijde. Hint: gebruik je constructie in (a), en verdeel een aantal tegels in twee
driehoeken.

2. Laat P = (1 : −1 : 1), Q = (1 : 1 : 1), R = (−1 : 1 : 1) en S = (−1 : −1 : 1) in P2 (R).

(a) Geef de punten P 0 , Q0 , R0 en S 0 in R2 die corresponderen met P , Q, R en S als we

R2 inbedden in P2 (R) via:

f : R2 → P2 (R), (x, y) 7→ (x : y : 1).

En teken ze. Ze zijn de hoekpunten van het vierkant P 0 Q0 R0 S 0 .

(b) Geef de punten P 00 , Q00 , R00 en S 00 in R2 die corresponderen met P , Q, R en S als we
R2 inbedden in P2 (R) via:

g : R2 → P2 (R), (u, v) 7→ (1 : u : v).

En teken ze. Ze zijn, in deze volgorde, de hoekpunten van het vierkant P 00 Q00 S 00 R00 .
Merk op dat we in de volgorde P 00 Q00 R00 S 00 we geen vierkant krijgen.
(c) Vind uit wat er hier gebeurt door de beelden te berekenen van de zijden van het
vierkant P 0 Q0 R0 S 0 onder de afbeelding

g −1 ◦ f : (x, y) 7→ (x : y : 1) = (1 : y/x : 1/x) 7→ (y/x, 1/x)

die is gedefinieerd op de verzameling van (x, y) met x 6= 0. Teken deze beelden. Wat
is het beeld van het inwendige van P 0 Q0 R0 S 0 ? Kun je dit begrijpen door te kijken
naar wat er gebeurt in R3 ?

36
3. (a) De dubbelverhouding van 4 geordende verschillende Punten p, q, r en s op P1 (R) is
gedefinieerd als g(s), waarbij g de unieke projectieve transformatie van P1 (R) is die
(p, q, r) naar (∞, 0, 1) stuurt. Leid hieruit de bekende formule
r−p s−q
[p, q; r, s] = ·
r−q s−p
voor de dubbelverhouding af.
(b) Gebruik de definitie van de dubbelverhouding in (a) om te bewijzen dat ze invariant
is onder alle projectieve transformaties van P1 (R).
(c) Stel je hebt een foto van een tafereel waarin 4 punten P1 , P2 , P3 en P4 op gelijke
afstanden op een lijn liggen. Laat Q1 , Q2 , Q3 en Q4 de beelden van deze punten in
de foto zijn. Leg uit dat deze 4 punten Qi op een lijn L liggen.
(d) Stel dat, voor een coordinaat op L, Q1 , Q2 and Q4 de punten 0, 1 en 4 op L zijn. Wat
is dan het punt Q3 ?

4. Beschouw de volgende Lijnen in het hyperbolisch vlak H: de lijnen {z ∈ H : <(z) = 0},

{z ∈ H : <(z) = 1/2}, en de cirkelbogen {z ∈ H : |z| = 1} en {z ∈ H : |z| = 2}.

(a) Laat zien dat er een 4-hoek in H is waarvan de som van de hoeken kleiner is dan 360
graden (je hoeft de som niet uit te rekenen).
(b) Leid hieruit af dat er driehoeken zijn waarvan de som van de hoeken kleiner is dan
180 graden.

5. Leg uit in een paar (ten hoogste 10, zeg) regels waarom het parallel-axioma niet kan
worden afgeleid uit de rest van Hilbert’s axioma’s voor euclidische meetkunde.

6. Hier zijn de 3 axioma’s voor projectieve meetkunde:

P1 ieder tweetal verschillende Punten is bevat in een unieke Lijn,

P2 ieder tweetal verschillende Lijnen snijdt in een uniek punt,
P3 er zijn 4 verschillende Punten waarvan geen 3 collineair zijn (d.w.z., op één Lijn lig-
gen).

Leg uit waarom deze axioma’s niet impliceren dat er tenminste 8 Punten zijn.

Vul alsjeblieft het evaluatieformulier op de mastermath website in. Deze informatie is

belangrijkvoor de mastermath organisatie, en voor de docenten.

37