MA243 November 28, 2022 77

24. Lecture 24: Projective transformations and projective frames of reference

Definition 44. The projective general linear group of a vector space V over a field k is the group
of all invertible linear maps from T : V æ V upto scalar multiplication, that is, we consider T1 ≥ T2
if there is some non-zero element ⁄ œ k with T2 = ⁄T1 .
We write P GL(n) = P GL(Rn ). In this case, the elements of P GL(n) are invertible matrices up to
scalar muliplication, that is
A ≥ B ≈∆ A = ⁄B for some ⁄ œ Rú
Example, in P GL(2), we have
3 4 3 4
1 2 2 4
≥
3 4 6 8
Definition 45. If A is an invertible (n + 1) ◊ (n + 1) matrix, then the map
T A : Pn æ P n

[v] ‘æ [Av]
is called a projective transformation, or a projectivity or a projective linear map.
More generally, for any invertable linear map T between any vector spaces V and W , the induced
map T : [v] ‘æ [T (v)] is called a projective linear map. I.e., the definition is basis independent, and
doesn’t whether the field is the reals.
TA is well defined, since for x, y œ Rn+1 , if x ≥ y, that is, if [x] = [y], then x = ⁄y for some ⁄ œ R,
so Ax = A⁄y = ⁄Ay, so [Ax] = [Ay].
Also, if A ≥ B, then TA = TB , since if A = ⁄B, then [Ax] = [⁄Ax] = [Bx]. So, projective linear
maps of Pn correspond to elements of P GL(n + 1).
Definition 46. A projective frame of reference for Pn is a set of n+2 points, P0 , P1 , . . . , Pn+1 œ Pn ,
any n + 1 of which are linearly independent (and so span Pn )
Definition 47. The standard frame of reference for Pn is given by [e1 ], . . . , [en+1 ] together with
[e1 + · · · + en+1 ], where ei are the standard basis for Rn+1 .

semi-circle of representatives of P1

[e2 ]
Standard frame of reference points for P1
[e1 + e2 ]

[e1 ]

Example: Any three distinct points P1 , P2 , P3 in P1 is a frame of reference for P1 , since being
distinct means that PÊ1 , P
Ê2 , P
Ê3 are pairwise linearly independent.
Motivation: Why do we need n + 2 points? This is because n + 2 points is enough to determine
a projecive linear map of Pn , but n + 1 is not. This might be suprising, since the images of n points
is enough to determine a linear map on Rn .
Example: There are infinitely many projective linear maps P1 æ P1 with
(1 : 0) ‘æ (1 : 0)
(0 : 1) ‘æ (0 : 1)
MA243 November 28, 2022 78

namely, any diagonal map, of the form Example: The radial projec-
4 4
3
– 0
4 tion, with centre 0, from the
A= line x = 1 to the line x = 3
0 — 3 3 takes a point with y = 1 to
However, these maps are not all the a point with y = 3. We can
same map, since we have 2 2 paramaterise points on these
A : (1 : 0) ‘æ (1 : 0) lines by their y-coordinates.
A : (0 : 1) ‘æ (0 : 1) 1 1 If we complete these lines to
form copies of P1 , with point
A : (1 : 1) ‘æ (– : —) with parameter y maping to
0 0
If we fix the image of (1 : 1), then we (y :31), then the map is given

x=1

x=3
4
uniquely determine the projective linear 3 0
≠1 ≠1 by : P1 æ P1
map. 0 1
1
Example: The map t ‘æ t from R to R given in Lecture 22 can be considered as a map on P1 .
On R µ P1 , given by 53 46 53 4 3 46 53 46
t 0 1 t 1
‘æ = ,
1 1 0 1 t
extended to a map on P1 by allowing the map to apply to (1 : 0). This map has the effect of
switching 0 and Œ, thought of as the points (0 : 1) and (1 : 0).
4
3 4
0 –
Any map : P1 æ P1 , for –, — ”= 0 also
— 0 ≠1 0
3
1 2 3 4
switches (0 : 1) and (1 : 0), and corresponds to y=3
–
t ‘æ —t . red line, æ blue line,
2
By specifying the image of (1 : 1), the map is {x = 1} æ {y = 3}
uniquely determined. 0 ‘æ Œ
1
The diagram on the right shows a projection from ΠԾ 0
{x
3
= 1}
4
to {y = 3}, which corresponds to the map 1 ‘æ 3
0 3 0
t ‘æ 3t
: P1 æ P1 sending (1 : 1) ‘æ (3 : 1)
x=1

1 0
≠1
4 Example: In the figure on the right,
≠1 we project from the line {x = 1} to the
y=
≠1 0 3
line {y = ≠ 12 x + 3}.
2x+ 1 We parameterise {x = 1} by the y co-
red 3
æ blue 2 ordinate, and {y = ≠ 12 x + 3} by the x
line line 2
coordinate, though we could choose dif-
0 ‘æ 6 3
ΠԾ 0 ferent parameterisations.
1 4
1 ‘æ 2 With respect to this parameterisation,
5
as a map from P1 to 1
3 P ,4this map is
≠ 12 ‘æ Œ 6
6 0 0 6
given by the matrix , and corre-
x=1

t ‘æ 2t+1
2 1
6
≠1 sponds to t ‘æ 2t+1 as a map on R

The maps in these examples have been written as maps from P1 to P1 , but drawn as maps from
R æ R. To extend from R to P1 , we should embed each copy of R in a copy of P1 , which can
be acheived by replacing R2 in these figures by P2 . Then these are called projective perspectives,
which we will come to next lecture.
MA243 November 28, 2022 79

Theorem 14. There is a bijection between projective transformations of Pn and projective frames
of reference of Pn , defined as follows:
„ : P GL(n + 1) æ {projective frames of reference}
T ‘æ {T ([e1 ]), T ([e2 ]), . . . , T ([en+1 ]), T ([e1 + · · · + en+1 ])}
Proof. We need to show that this map is well defined, injective and surjective.
Well defined: The image of the standard projective frame of reference under T must also be a
projective frame of reference, because T is invertible, so maps any basis of Rn+1 to another basis.
Surjectivity: Suppose we are given a projective frame of reference, P0 , P1 , . . . , Pn+1 . By definition,
any choice of n + 1 of these these are linearly independent in Pn , which means that a choice of lifts,
a0 = PÊ0 , a1 = P
Ê1 , . . . an = P
Ên is linearly independent in Rn+1 , and so for any nonzero –0 , . . . , –n ,
the matrix A with columns –i ai , is invertible.
Q R
| | ... |
A = a–0 a0 –1 a1 . . . –n an b
| | ... |
We know that Aei is the ith column of A, and so for i = 0, . . . , n,
[TA (ei )] = [Aei ] = [–i ai ] = [ai ] = Pi
We want C A n BD
ÿ
TA ei = Pn+1 .
i=1
q q q
Let an+1 = P ] so Pn+1 = [an+1 ]. We have [TA ( ni=1 ei )] = [A( ni=1 ei )] = [ ni=1 –i ai ], so it’s
n+1 , q
sufficient to solve ni=1 –i ai = an+1 , i.e., A(–0 , –1 , . . . , –n )T = an+1 . Since A is invertible, set
Q R
–0
c –1 d
c d
c . d = A≠1 an+1 .
a .. b
–n
So „(TA ) = {P0 , . . . , Pn+1 }. Thus „ is surjective.
Injectivity:
Suppose that „(TA ) = „(TB ) for matrices A, B œ P GL(n). Then for i = 0, . . . , n, we have
TA (ei ) = TB (ei ),
so the columns of A are equal to those of B up to some scalar multiples, e.g. up to b = (—0 , . . . , —n ),
and we have Q R
—0
c —1 d
c d
B = Ac . .. d
a b
—n
q q
Since „(TA ) = „(TB ), we have [TA ( ei )] = [TB ( ei )] so
Ë ÿ È Ë ÿ È Ë ÿ È
A ei = B ei = A —i ei .
q q
We know that A is invertible,
q
so we can’t have A ei = 0, since then we would have ei =
A≠1 0 = 0, but we know that ei ”= 0. So for some scalar ⁄,
ÿ ÿ
⁄A ei = A —i ei
q q
rearranging, A (⁄ ≠ —i )ei = 0 and since A is invertible, (⁄ ≠ —i )ei = 0 which is only possible
if ⁄ = —i for each i = 0, . . . , n, which implies B = ⁄A, and so TA and TB are equal as elements of
P GL(n). ⇤
MA243 November 28, 2022 80

Corollary 14. Any three distinct points P1 , P2 , P3 in P1 can be projected onto any other three
distinct points Q1 , Q2 , Q3 . This is called three transitivity of the action of P GL(2) on P1 .

Proof. By Theorem 14 there is some A maping the standard frame of reference to P1 , P2 , P3 , and B
maps the standard frame of reference to Q1 , Q2 , Q3 , then BA≠1 maps P1 , P2 , P3 to Q1 , Q2 , Q3 . ⇤

≠1 0 1 2 3 4 5
Example:
OK, so how do I project
between these set of
points???
≠1 0 1 2 3 4 5

I am going to take R1 P1
via t ‘æ (1 : t).

x=1
µ 2 both lines parametrized
To project [e1 ], [e2 ], [e1 + e2 ] (think “0, Œ, 1”) to 2 red æ blue
by y-coordinate.
x projection is from 0
(1 : 0), (13: 1),4(1 : 2), (think “0,1,2”) we first + line line
1 1 y 1 0 ‘æ 0
take A = , which gives TA ([e1 ]) = (1 : 0), = 1
0 1 1 ΠԾ 1
TA ([e2 ]) = (1 : 1), but does not map (1 : 1) to 0
0
1 ‘æ 12
(1 : 2).
≠1
≠1
3
3
Now take + 2
x=1

3 4 3 4 3
1 1 ≠1
43 4 3 4
1 2
2 =x
–1
= A≠1 = =
≠1
, 2y red æ blue
–2 2 0 1 2 2 both lines parametrized line line
1
1 by y-coordinate. 0 ‘æ 0
so the required element of P GL(2) is
3 4
projection is from 0 ΠԾ 1
≠1 2 0 1 ‘æ 2
0
0 2

≠1

red æ blue To project [e1 ], [e2 ], [e1 + e2 ] to

line line (1 : 1), (1 : 4
4), (1 : 5), we first take
3y + 5 0 ‘æ ≠1 3
1 1
x= A= , which gives TA ([e1 ]) = (1 : 1),
4 ΠԾ 4 1 4
3
4 1 ‘æ 5 TA ([e2 ]) = (1 : 4). Take the column multiplier:
3 4 3 4 3 43 4 3 4
1 4 ≠1 1 1
x=1

–1
3 = A≠1 = = ,
2 –2 5 ≠1 1 5 ≠4
{x = 1} parametrized
3 4
by y-coordinate.
1 ≠4
1
{3y + x = 4} so the required element of P GL(2) is
parametrized 1 ≠16
by 1 + y-coordinate.
projection is from 0 Note that we have ignored the determinant factor
0 when computing the inverse of A, since elements
of P GL(2) are only defined up to multiplcation
≠1 by non-zero scalar.
Now to map (0 : 1), (1 : 1), (1 : 2) to (1 : 1), (1 : 4), (1 : 5), we take
3 43 4≠1 3 43 4 3 4 3 4
1 ≠4 ≠1 2 1 ≠4 2 ≠2 2 2 1 1
A= = = =
1 ≠16 0 2 1 ≠16 0 ≠1 2 14 1 7
MA243 November 28, 2022 81

Again, this is a computation in P GL(2), so any

scale factors of the whole matrix are ignored. lines assumed suitably
linearly parameterised
We can check the result, e.g., 0 ‘æ 1

x=1
53
1 1
4 3 46
1
53
3
46 1 ‘æ 4

x
TA ((2 : 1)) = = = (1 : 5)

+
1 7 2 15 2 ‘æ 5

y
=
(1 : t) ‘æ (1 + t : 1 + 7t)

6
Note, although the element of P GL(2) is unique,
the picture is not unique, because we could choose t ‘æ 1+7t
1+t
different parameterisations. Finding a suitable
geometric interpretation is a little more work.
Example: Find an element of P GL(4) mapping the standard frame of reference of P3 to (1 : 2 :
3 : 4), (0 : 1 : 0 : 1), (1 : 0 : 1 : 0), (1 : 1 : 1 : 0), (1 : 0 : 0 : 1)
Solution: The required matrix must have the form
Q R
a 0 c d
c2a b 0 dd
A=c
a3a
d
0 c db
4a b 0 0
We need to solve [A(1, 1, 1, 1)T ] = (1 : 0 : 0 : 1), which we could solve by finding an inverse of a
certain 4 ◊ 4 matrix, or by row reduction:
Q R Q R Q R Q R
1 0 1 1 1 ≠2 0 0 0 1 ≠2 0 0 0 1 1 0 0 0 ≠ 12
c 2 1 0 1 0 d c ≠2 0 0 1 ≠1 d c 0 0 0 1 ≠2 d c 0 0 0 1 ≠2 d
c d∆c d∆c d∆c d
a 3 0 1 1 0 b a 3 0 1 1 0 b a 3 0 1 1 0 b a 0 0 1 0 72 b
4 1 0 0 1 4 1 0 0 1 0 1 0 0 3 0 1 0 0 3
This has solution (a : b : c : d) = (≠ 12 : 3 : 7
2 : ≠2) = (≠1 : 6 : 7 : ≠4). Hence the required matrix is
Q R
≠1 0 7 ≠4
c≠2 6 0 ≠4d
A=c d
a≠3 0 7 ≠4b
≠4 6 0 0
Any scalar multiple of this is also correct, as it will be equal to this matrix in P GL(4).

Example: Considering R as a subset of P1 , included as {(t : 1) : t œ R} µ P1 , we can ask how to

extend a linear map on R to a projective linear map of P1 . For example, t æ t + a extends to:
53 4 3 46 53 46
1 a x x + ay
(x : y) æ = = (x + ay : y)
0 1 y y
If y = 1 and x = t, this becomes (t : 1) æ (t + a : 1).
Projective linear maps of P1 become ratios of linear maps when restricted to R1 .
Example: Consider R2 as a subset of P2 , and the map
SQ R Q RT
1 2 3 x
(x : y : z) æ Ua1 0 1b ay bV = (x + 2y + 3z : x + z : x + y)
1 1 0 z
The restriction to the copy of R2 given by setting z = 1 is a map
3 4
x + 2y + 3 x + 1
(x, y) æ ,
x+y x+y
Example: Theorem 14 implies that we can project any quadrilateral on a plane in R2 to any other
quadrilateral on some other plane in R2 , provided they both are actual quadrilaterals, not triangles
or lines or points, and that we can move the planes to suitable positions.
MA243 November 28, 2022 82

25. Lecture 25: The cross ratio and projective perspectives

The projection maps we have been discussing are called “perspectivities”.
Definition 48. A perpsectivity f is a map between two distinct hyperplanes (linear subspaces of
dimension n ≠ 1), 1 and 2 in Pn , given by projection from a point O ”œ 1 fi 2 . That is, if
P œ 1 , then f (P ) œ 2 is the point such that 0, P, f (P ) all lie on the same line.
f (P )
L2

f (Q) R

f (R)

Q
P

L1 O

Lemma 36. Perspectivities are well defined.

Proof. We require in this definition that f (P ) exists and is unique. This follows from Theorem 12,
with E = L := ÈO, P Í and F = 2 . We have dim(L) = 1 and dim( 2 ) = n ≠ 1. Since O ”œ 2 , we
must have dim(ÈL, 2 Í) = n, so by the theorem,
dim(L fl 2) = 1 + (n ≠ 1) ≠ n = 0
and so L fl 2 is a single point, which we define to be f (P ). ⇤
Corolary 14 shows that we can map any three distinct points on a projective line to any other three
disctinct points, by a projective linear map. However, this is not true for four points on a line. We
will show that the cross ratio of four points has to be the same for it to be possible to map a set
of four points on a line to another set of four points on a line. So the cross ratio is an invariant of
projective linear maps.
RÕ
Definition 49. Let P, Q, R, S be four distinct
points on a line in Pn . Choose an approrpiate
basis such that P = (1 : 0) and Q = (0 : 1) (fol-
lowed by n ≠ 1 zeros, which we can ignore). Since Q=(0:1) R=(1:1) SÕ
PÂ and QÂ span the line that R and S lie on, and
since R, S ”= P , there are ⁄, µ œ R such that we
can write with respect to this basis: S=(1: 12 )

R = (1 : ⁄), S = (1 : µ) P =(1:0) PÕ
Then the cross ratio is
⁄ in this example {P, Q; R, S} = 2,
{P, Q; R, S} = which is also the case for the pro-
µ jection to {P Õ , Q; RÕ , S Õ }.

Warning: The cross ratio will change when the points are permuted, so it depends on the order
P, Q, R, S are given in.
MA243 November 28, 2022 83

Note: We could also map P, Q, R to the standard projective frame of reference for P1 , then the
cross ratio is the image of S.
It is convenient not to have to make a change of basis in order to compute the cross ratio, so the
following result is useful:
Proposition 20. If P, Q, R, S are points in P1 , then they can be considered as lines in R2 . Let L
be any line in R2 not through the origin. Let p, q, r, s be the vector coordinates of the intersections
of P, Q, R, S with L. Then 3 43 4
p≠r q≠s
{P, Q; R, S} = ,
p≠s q≠r
where the ratios of vectors make sense, because they all in the direction of L, and we define ⁄vv = ⁄.

P
p
≠
r

R
q
≠
p
≠

r
s

q
≠
s Q

Proof. Suppose T is an invertible linear map. Then if v is a non-zero vector, and ⁄ a non-zero
scalar, the by definition of linearity, T (⁄v) = ⁄T (v), so
T (⁄v) ⁄v
=
T (v) v
and so when we apply a change of basis transformation T , with
T (P ) = (1 : 0), T (Q) = (0 : 1), T (R) = (1 : ⁄), T (S) = (1 : µ)
By definition,
⁄
{P, Q; R, S} = .
µ
Now for some non-zero scalars a, b, c, d, we have
T (p) = (a, 0), T (q) = (0, b), T (r) = (c, c⁄), T (s) = (d, d⁄)
So we have
3 43 4 3 43 4 3 43 4
p≠r q≠s (a, 0) ≠ (c, c⁄) (0, b) ≠ (d, dµ) (c ≠ a, c⁄) (d, dµ ≠ b)
= = .
p≠s q≠r (a, 0) ≠ (d, dµ) (0, b) ≠ (c, c⁄) (d ≠ a, dµ) (c, c⁄ ≠ b)
It is given that the points p, q, r, s all lie on a line, and so their images under T also lie on a line,
since T is linear. So, the ratio of the vectors in this expression is given by either taking the ratio
of the x or y components, since the result will be the same, since ⁄(x, y) = (⁄x, ⁄y), so we have
3 43 4
(c ≠ a, c⁄) (d, dµ ≠ b) c⁄ d ⁄
= = .
(d ≠ a, dµ) (c, c⁄ ≠ b) dµ c µ
⇤
MA243 November 28, 2022 84

Lemma 37. A perspectivity f : 0 æ 1 is a projective linear transformation.

Proof. We want to show that the perspectivity defined by projection from a point O can be induced
by a linear map from Ê0 to Ê1 .
If 0 = 1 , then f is the identity, and the result is trivially true, so we may assume 0 ”= 1 .
We have dim( 0 ) = dim( 1 ) = n ≠ 1, and since these planes are not idential, their span must be
all of Pn , and so by Theorem 12, dim( 0 fl 1 ) = 2(n ≠ 1) ≠ n = n ≠ 1. Let v2 , . . . , vn , vn+1 be
a basis of ^ 1 fl 2 , and let v0 and v1 be vectors extending to a basis for
 0 and  1 respectively.
The point O is in neither 1 nor 2 , and therefore O  can’t be written in terms of the basis for  1
or  2 and so O, v0 , v1 , v2 , . . . , vn+1 is a projective frame of reference. So by Theorem 14 there is a
projective linear map T with T ( 0 ) = {x0 = 0}, T ( 1 ) = {x2 = 0}, and T (O) = (1 : 1 : 1 : · · · : 1).
A composition of linear maps is linear, thus by the change of basis T , we may assume that 0 =
{x0 = 0} and 1 = {x1 = 0}, and O = (1 : 1 : 1 : 1 : · · · : 1).
Now given a point [x] = (0 : x1 : x2 : · · · : xn ) œ 0 , then the line through O and [x] consists of
points of the form
(⁄ : µx1 + ⁄ : µx2 + ⁄ : · · · : µxn + ⁄).
This line intersects the hyperplane 1 when ⁄ = ≠µx1 , so
f ([x]) = (≠x1 : 0 : x2 ≠ x1 : · · · : xn ≠ x1 ),
which is clearly linear, and so f is a projective linear transformation. ⇤
Proposition 21. If P1 , Q1 , R1 , S1 and P2 , Q2 , R2 , S2 are two sets of distinct points lying on lines
L1 and L2 , then if there is a perspectivity f taking the first set of points to the second, then we
have equality of cross ratios:
{P1 , Q1 , R1 , S1 } = {P2 , Q2 , R2 , S2 }
Proof. In the proof of Proposition 20, we show that the cross ratio is invariant under projective linear
transformations, and in Lemma 37, we show that perspectivity is a projective linear transformation.
⇤
Example: Does there exist a projective linear map taking points P1 , Q1 , R1 , S1 = 0, 1, 2, 3 to points
P2 , Q2 , R2 , S2 = 0, 1, 7, 8, where the point t œ R is interpreted as a point (t : 1) in P1 ?
Solution: The cross ratios are
3 43 4 3 43 4
0≠2 1≠3 4 0≠7 1≠8 7◊7
{P1 , Q1 ; R1 , S1 } = = ”= {P2 , Q2 ; R2 , S2 } = =
0≠3 1≠2 3 0≠8 1≠7 8◊6
Since these are not equal, there is no such projective linear map.
Example: Does there exist a projective linear map taking points P1 , Q1 , R1 , S1 = 0, 1, 2, 5 to points
P2 , Q2 , R2 , S2 = 1, 4, 5, 6, where the point t œ R is interpreted as a point (t : 1) in P1 ?
Solution: The cross ratios are
3 43 4 3 43 4
0≠2 1≠5 8 1≠5 4≠6 8
{P1 , Q1 ; R1 , S1 } = = = {P2 , Q2 ; R2 , S2 } = =
0≠5 1≠2 5 1≠6 4≠5 5
So, there is such a map, and in fact this is the map given in Example 1 in Lecture 24, pages 80 and
81. Note that to compute the map, we only need to consider the images of the first three points.
Given that the cross ratios are equal, the fourth point of one set will automatically get mapped to
the fourth point of the other set.
Remark (suggestion for further reading): We do not have time to prove the converse of
Proposition 21. We have also not shown that if a map between projective spaces preserves lines
and the cross ratio, then it’s a perspectivity, and thus given by an element of P GL(n).
Note that the preservation of the cross ratio is a kind of generalisation of the fact that for 3 points
P, Q, R in a line in R2 , the ratio of the distance between P Q and QR is preserved by isometries of
R2 , Proposition 2.
MA243 November 28, 2022 85

26. Lecture 26: Desargues Theorem

Desargues’s theorem is a classical result in Projective geometry, known since the 17th century.
Definition 50. A triangle P QR in Pn is a set triangles in perspective from a point
of three distinct points, P, Q, R in Pn , and the PÕ
“sides” of the triangle, which are the three lines P
spanned by the three pairs of points. O
Q QÕ
Definition 51. Two triangles P QR
and P Õ QÕ RÕ in Pn are said to be
R
in perspective from a point O if the lines ÈP, P Õ Í,
ÈQ, QÕ Í ÈR, RÕ Í all intersect in a common point
RÕ
O.
The “dual” concept is ‘perspective from a line’ – Ltriangles in perspective from a line
see [RS 5.11] for more on the principal of duality.
Definition 52. Two triangles P QR and
rÕ
P Õ QÕ RÕ , with sides p, q, r and pÕ , q Õ , rÕ in Pn r
are said to be in perspective from a line L if the q
points p fl pÕ , q fl q Õ , r fl rÕ , all lie on a common p qÕ pÕ
line L.
Lemma 38. The projection fi of P3 to P2 = {(x : y : z : 0) µ P3 given by projection from a point
M in P2 µ P2 to P2 is given by [v] æ [T (v)] for some linear map T .
Proof. By a projective linear map, we map assume that M = (0 : 0 : 0 : 1). Then P = (x : y : z :
w) œ P3 lies on the line ÈP, M Í = (⁄x : ⁄y : ⁄z : ⁄w + µ), which intersects P2 at (x : y : z : 0), by
taking µ = ≠⁄w, and this is clearly a linear map. ⇤
Theorem 15. (Desargues’ Theorem) If P QR and P Õ QÕ RÕ are two distinct triangles in Pn which
are in perspective from a point, then they are also in perspective from a line.
Proof. Notation: Let p = ÈQ, RÍ, q = ÈP, RÍ, r = ÈQ, P Í, and pÕ = ÈQÕ , RÕ Í, q Õ = ÈP Õ , RÕ Í, rÕ =
ÈQÕ , P Õ Í. Let 1 be the plane spanned by P, Q, R and 2 be the plane spanned by P Õ , QÕ , RÕ . Let
O be the point from which P QR and P Õ QÕ RÕ are in perspective.

Claim: p fl pÕ , q fl q Õ , r fl rÕ are non empty: The

In this Geogebra picture,
plane spanned by O, QÕ , RÕ also contains Q and O is the origin.
R, since these are on the lines OQÕ and ORÕ . So, The vertices of the triangles
lie on the coordinate axes.
the lines p and pÕ must lie in this plane, and so by L is the line through I, J, K.
Theorem 13 they intersect in a point, unless they https://www.geogebra.org/3d/
njambarp
are equal. Similarly for q fl q Õ and rÕ fl rÕ .
Since O, QÕ , RÕ , Q and R are all contained in a
single plane, and P is in the same plane as OP ,
the space spanned by 1 and 2 is at most 3
dimensional.
Case 1: Suppose that dim(È 1 , 2 Í) = 3, so
1 ”= 2 . Let L = 1 fl 2 . By Theorem 12
dim(L) = 1. Since p µ 1 and pÕ µ 2 , we must
have p fl pÕ µ L, and similarly for q fl q Õ and r fl rÕ ,
and so the result is true.

Case 2: Suppose that 1 = 2. By a change of basis we can identify 1 with P2 = {(x : y : z :

0)} µ P3 .
MA243 November 28, 2022 86

We construct two triangles which are not coplanar, and which project to P QR and P Õ QÕ RÕ .

We may assume R ”= RÕ , since the triangles are construction of S and S Õ

distinct, so at least one point differs. Let M be
a point in P3 \ 1 . Let S be any point on the
line M R which is not in 1 , and not equal to M .
Since R, RÕ O are on a line, R, RÕ , O, M, S are in a
plane, so by Theorem 13 the lines M RÕ and OS
intersect in a point, S Õ .
Now the triangles P QS and P Õ QÕ S Õ can’t lie
in a plane, since S and S Õ are not in the same
plane as P, Q, P Õ , QÕ .

By construction, the triangles P QS and P QS Õ

construction of triangles P QS and P Õ QÕ S Õ
project from M to the triangles P QR and
P Õ QÕ RÕ in the plane 1 . Let fi be the projection
map from M to 1 .
By Case 1, the pairs of edges of P QS and
P Õ QÕ S Õ have intersection points all lying on a
line, L. By Lemma 38, fi is described by a linear
map, so the image of L is a line (or if a point,
it is contained in a line). Since fi is a projective
linear map, it maps the lines spanned by edges
of P QS and P Õ QÕ S Õ to the lines spanned by
edges of P QR and P QRÕ . So the points pflpÕ ,
q fl q Õ , r fl rÕ all lie on the line fi(L).
(See https://www.geogebra.org/3d/thzujqpa, https://www.geogebra.org/3d/whpsdqxa.) ⇤

Remark: This result is true in Pn . It is not al-

ways true in Rn , since intersections of edges might
be at infinity, or the line L might end up being
the line at infinity, which happens if pairs of sides
of the two triangles are parallel. Working in pro-
jective space generally results in more uniform re-
sults without special exceptions. (Here “infinity”
means any point in Pn \ Rn , where we can include
Rn µ Pn as the subset with xn+1 = 1.)
Desmos version: https://www.desmos.com/
calculator/uq2okdn9ku
MA243 November 28, 2022 87

27. Lecture 27: Pappus’ Theorem

Pappus’ Theorem has been known since the ancient Greeks. It is closely related to Desargues
Theorem.
Theorem 16 (Pappus’ Theorem). Let L, LÕ µ P2 be distinct projective lines.
Let P, Q, R œ L \ LÕ , P Õ , QÕ , RÕ œ LÕ \ L be distinct points.
Then the intersection points
A := ÈP, QÕ Í fl ÈP Õ , QÍ, B := ÈP, RÕ Í fl ÈP Õ , RÍ, C := ÈQ, RÕ Í fl ÈQÕ , RÍ
ine
R
are colinear.
o nl
Q
o mm
n ac
o
lie
C
B,
P A,
C
L
B
A

LÕ
PÕ QÕ RÕ

Proof. Since P, Q, P Õ , QÕ are distinct and on two distinct lines, these points must be a projective
frame of reference, and so by Theorem 14 there is a projective transformation which takes these
points to the standard frame of reference for P2 .
Since a projective linear map preserves lines and intersection points, we can now assume
P = (1 : 0 : 0), Q = (0 : 1 : 0), P Õ = (0 : 0 : 1), QÕ = (1 : 1 : 1).
The picture, from https://www.geogebra.org/3d/zhwkackg shows these points on the projective
plane projected to a sphere. The line through P, Q, R is the “line at infinity”, where the sphere
intersects the z = 0 plane.
Since R œ ÈP, QÍ, R, being a linear combination of
P and Q, has the form R = (1 : – : 0), where we
may assume the first coordinate is 1, since it can
not be zero, since R ”= Q. If the first coordinate is
non-zero, by scaling we can set it equal to 1. Since
RÕ is on ÈP Õ , QÕ Í, RÕ has the form RÕ = (1 : 1 : —),
where again, the first coordinates can not be zero,
since RÕ ”= P Õ .
Now compute the lines and intersection points:
<
ÈP, QÕ Í = {(⁄ + µ : µ : µ)}
A = (0 : 1 : 1)
ÈP Õ , QÍ = {(0 : ⁄2 : µ2 )}
<
ÈQ, RÕ Í = {(µ2 : µ2 + ⁄2 : µ2 —)}
B = (1 : – + — ≠ –— : —)
ÈR, QÕ Í = {(µ + ⁄ : µ– + ⁄ : ⁄)}
<
ÈP, RÕ Í = {(µ2 + ⁄2 : µ2 : µ2 —)}
C = (1 : – : –—)
ÈR, P Õ Í = {(µ : µ– : ⁄)}
Now check that –AÂ + CÂ = B, Â where A Â =
(0, 1, 1), C = (1, –, –—), B = (µ + ⁄, µ– + ⁄, ⁄),
 Â
which shows that A, B, C are on a line. ⇤