Geo Cross Ratio
Geo Cross Ratio
Geo Cross Ratio
Contents
1 Projective Geometry and Cross ratios
1.1 Cross Ratios . . . . . . . . . . . . . .
1.2 Cross Ratios on a Conic Section . . .
1.3 Cross Ratios on the Inversive Plane . .
1.4 Invertible functions on the line . . . .
1.5 Angles and the circle points . . . . . .
1.6 Polar maps . . . . . . . . . . . . . . .
1.7 Coharmonic points . . . . . . . . . . .
1.8 Symmetries of the plane . . . . . . . .
1.9 The Cross Cross Ratio . . . . . . . . .
1.10 A few miscellaneous exercises . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
2
6
9
10
11
13
16
20
23
26
Definition 1. The projective plane P2 is the set of lines through an observation point O in three
dimensional space. A projective line l is a plane passing through O, and a projective point P is
a line passing through O. If the line defining P is contained in the plane defining l, we say that
P l.
If A2 is an ordinary plane which does not pass through O, then we can identify most projective
points of P2 with ordinary points on A2 by taking the intersection of the line defining the projective
point with A2 . The projective line which is defined by a plane passing through O and parallel to A2
is called the line at infinity, or the horizon line. Projective points contained in the line at infinity
are called infinite points.
If we take O = (0, 0, 0), then we can put coordinates on the projective plane as follows. Every
projective point P is a line through O and some other point (p, q, r). Then every point on the line
defining P is of the form (p, q, r) for some . We write P = [p : q : r], where the colons indicate
that we only care about the ratios of the coordinates. If A2 is the plane z = 1, then the ordinary
point on A2 corresponding to P is ( pr , qr , 1), or if we ignore the z-coordinate it is just ( pr , qr ). If
r = 0, then P is an infinite point with slope pq .
We can define projective coordinates for projective lines as well. A projective line l is defined
by a single linear equation
dx + ey + f z = 0,
with not all of d, e, f equal to 0. Furthermore, this equation defines the same line if all of d, e, f are
rescaled by the same nonzero . Thus we say that l = (d : e : f ). If P = [p : q : r], then we have
P l if and only if
dp + eq + f r = 0.
The intersection of l with the ordinary plane A2 defined by z = 1 is just the line dx + ey + f = 0.
The line at infinity has coordinates (0 : 0 : 1).
The coordinate system described above can be called cartesian projective coordinates. There are
other projective coordinate systems, one of the most useful of which is the barycentric coordinate
system. In the barycentric coordinate system, a triangle ABC in A2 is fixed and the coordinates of
three dimensional space are chosen such that A = (1, 0, 0), B = (0, 1, 0), C = (0, 0, 1) - so the plane
A2 is now defined by the equation x + y + z = 1. If P is an ordinary point in A2 , then the projective
coordinates [p : q : r] of P are defined to be any three numbers p, q, r, not all zero, proportional
to the three directed areas [P BC], [AP C], [ABP ]. In the barycentric coordinate system, a line
l = (d : e : f ) is the set of points P such that
d[P BC] + e[AP C] + f [ABP ] = 0.
The line at infinity has barycentric coordinates (1 : 1 : 1).
1.1
Cross Ratios
AC
,
BC
where the ratio is taken to be positive if the rays AC and BC point in the same direction, and
negative otherwise. If l1 , l2 , l3 are three directed lines passing through a point, not all equal, then
their ratio is defined by
sin l1 l3
(l1 , l2 ; l3 ) =
,
sin l2 l3
where the angles are oriented in the counterclockwise sense.
Exercise 1. (a) Show that if A 6= B then there is a bijection between points C on the line AB
and ratios (A, B; C). Thus we can use the ratio as a coordinate on the line AB.
(b) Show that the ratio (l1 , l2 ; l3 ) does not depend on the orientation of line l3 . Show that if
l1 6= l2 we can use the ratio (l1 , l2 ; l3 ) as a coordinate on the set of lines through the point
l1 l2 .
Exercise 2. Suppose that points A, B, C, not all equal, are on a line, and that point P is not on
that line. Show that
(A, B; C)
|P A|
=
.
(P A, P B; P C)
|P B|
2
Definition 3. If A, B, C, D are four points on a line, no three of them equal, then we define their
cross ratio to be
AC AD
(A, B; C)
=
.
(A, B; C, D) =
(A, B; D)
CB DB
If l1 , l2 , l3 , l4 are four lines passing through a point, no three of them equal, then their cross ratio
is defined by picking an orientation for each line, and then setting
(l1 , l2 ; l3 )
sin l1 l3 sin l1 l4
(l1 , l2 ; l3 , l4 ) =
.
=
(l1 , l2 ; l4 )
sin l3 l2 sin l4 l2
Theorem 1 (The fundamental theorem of cross ratios). If A, B, C, D are on a line, no three of
them equal, and if E is a point not on that line, then
(EA, EB; EC, ED) = (A, B; C, D).
We would like to extend the above definitions to any four points or lines in the projective
plane. One way to do this is to make special definitions if one of A, B, C, D is an infinite point: for
instance, if is the infinite point on line AB, then we have
(A, B; C, ) = (A, B; C) =
AC
.
CB
Similarly, if all of A, B, C, D are infinite points with slopes a, b, c, d, then their cross ratio is
ca da
.
(a, b; c, d) =
bc bd
However, the best way to do this is to simply change perspectives to get a coordinate system where
none of A, B, C, D is an infinite point. In other words, we find a new plane A0 2 not passing through
the observation point O, which intersects the four lines corresponding to the projective points
OA, OB, OC, OD at some new points A0 , B 0 , C 0 , D0 . Then for finite points A, B, C, D we have
(A, B; C, D) = (OA, OB; OC, OD) = (A0 , B 0 ; C 0 , D0 ),
so the cross ratio in the new coordinate system will be the same as the original cross ratio. If one
of A, B, C, D is an infinite point we use this formula as the definition of the cross ratio.
To check your understanding, calculate the cross ratio of four parallel lines in terms of the
distances between them (parallel lines intersect at the infinite point corresponding to their common
slope).
Exercise 3. Let ABC be a triangle, let M be the midpoint of AC, and let N be a point on line
BM such that AN is parallel to BC. Let P be any point on line AC, and let Q be the intersection
of line BP with line AN . Use cross ratios to prove that
AQ
1 AP
=
.
QN
2 PM
Exercise 4.
1
(A,B;C,D) .
1
1(A,B;C,D) .
B
E
C
X
D
F
Proof 2, using cross ratios. Let P be the intersection of the diagonals AC and BD. We have
(E, F ; X, Y ) = (AE, AF ; AX, AY ) = (B, D; P, Y ) = (CB, CD; CP, CY ) = (F, E; X, Y ).
Since E 6= F and X 6= Y , we conclude that (E, F ; X, Y ) = 1.
If EA, EB, EC, ED intersect a line l at points A0 , B 0 , C 0 , D0 , it often saves space to abbreviate
the inference
(A, B; C, D) = (EA, EB; EC, ED) = (A0 , B 0 ; C 0 , D0 )
by just writing
E
(A, B; C, D) = (A0 , B 0 ; C 0 , D0 ).
Now lets use this notation to give a compact proof of Desargues Theorem:
Theorem 3 (Desargues Theorem). Suppose that triangles ABC and XY Z are perspective from
a point, that is, suppose that the lines AX, BY, CZ all meet at a point P . Then the triangles ABC
and XY Z are perspective from a line, that is, the intersections AB XY , BC Y Z, CA ZX
all lie on a line.
Proof. Let U = BC Y Z, V = CA ZX, W = AB XY . We want to show that U, V, W lie on
a line, so we may as well suppose that V 6= W . Let Q, M, N be the intersections of line BY with
the lines W V , AC, XZ, respectively. Then we have
B
P
A
M C
B
Q W
Y
Z
N
X
Figure 2: Desargues Theorem
Z
H
(E, F ; G, H) = (P, Y ; Q, CH XY ). Thus (A, B; C, D) = (E, F ; G, H) if and only if CH, DG, and
XY meet at a point.
1.2
Proposition 1. Suppose that A, C, B, D are on circle , and that the (directed) arcs AC, CB, BD, DA
of have central angles 2, 2, 2, 2. Let E be any other point on . Then
sin sin
(EA, EB; EC, ED) =
.
sin sin
In particular, we have
|(EA, EB; EC, ED)| =
|AC||BD|
.
|AD||BC|
Corollary 1. Let be any conic section, that is, any intersection of a cone C through the observation point O with the plane A2 . If A, B, C, D, E, F are any six points on , then we have
(EA, EB; EC, ED) = (F A, F B; F C, F D).
(b) Given points A, B, C, D, E construct, using only a straightedge, the line l which is tangent to
the conic through the points A, B, C, D, E at A.
Exercise 8. Suppose points A, B, C, D, E, F, G, H lie on a conic . Let X = AF BE, Y =
BG CF, Z = CH DG. Show that (A, B; C, D) = (E, F ; G, H) if and only if X, Y, Z are on a
line.
Another easy application is a short proof of the butterfly theorem.
A
P
Figure 6: Exercise 12
1.3
Just as we used three projective coordinates for the projective plane, we use two projective coordinates to describe a projective line. Specifically, the projective point [s : t] will correspond to the
ordinary point with coordinate z = st if t 6= 0, and to the point if t = 0. When we allow s, t to be
complex numbers, we get what is sometimes called the complex projective line CP1 , the inversive
plane, or the Riemann sphere.
We define cross ratios on the inversive plane the same way we define cross ratios on a line:
ca da
(a, b; c, d) =
,
bc bd
where now a, b, c, d are complex numbers corresponding to ordinary points A, B, C, D in the inversive
plane.
Proposition 2. The points A, B, C, D corresponding to the complex numbers a, b, c, d are on a
circle or a line if and only if (a, b; c, d) is a real number. If they are on a line, we have (a, b; c, d) =
(A, B; C, D), and if they are on a circle , we have (a, b; c, d) = (A, B; C, D) .
Proof. Left as an exercise.
Inversion around the unit circle is given by the simple formula z 7 z1 in the inversive plane.
We have
1 1
1
1
1 1 1 1
ac ad
, ; , = 1c a1 d1 a1 =
= (a, b; c, d),
a
b c d
cb db
b c
b d
so inversion takes cross ratios to their complex conjugates. As a consequence, we see that inversion
takes circles and lines to circles and lines, and furthermore it takes harmonic quadrilaterals to
harmonic quadrilaterals.
a b
Definition 7. To every two by two matrix M =
with determinant ad bc not equal to
c d
zero, we associate a transformation fM of the inversive plane as follows. In projective coordinates
9
[s : t], we write
fM ([s : t]) = [as + bt : cs + dt].
In ordinary coordinates z = st , we write
fM (z) =
az + b
.
cz + d
1.4
Suppose a projective line P1 has coordinate z, and we have defined an invertible map f : P1 P1
via some geometric procedure that has no configuration issues (so for instance, taking the leftmost
intersection of a circle with a line would not count). Since any geometrically defined map can be
described algebraically by writing every point out in coordinates, our function f may be written
as an algebraic function of z, and if there are no configuration issues, then f must be a rational
function, i.e. a ratio of two polynomials:
f (z) =
p(z)
.
q(z)
Since f is invertible, the equation f (z) = w should have exactly one solution, so the polynomial
p(z) wq(z)
should have degree 1 for every constant w. Thus p and q are both linear polynomials, and we can
write
az + b
f (z) =
.
cz + d
Thus, f is in fact a M
obius transformation, and so f preserves the cross ratio. We record this as
an informal theorem.
Theorem 7. If f is an invertible function from a line to a line that is defined by a geometric
procedure that has no configuration issues, then f preserves the cross ratio. Furthermore, in this
case f is a M
obius transformation.
10
Exercise 16. Prove the converse: if f : P1 P1 is any function that preserves the cross ratio, prove
that f is a M
obius transformation, and find a geometric construction of the function f .
As an application, we consider the harmonic conjugation map. For any points A, B on P1 , we
define
hA,B (C) = D if (A, B; C, D) = 1.
We can construct D geometrically using the Quadrilateral Theorem, and hA,B is clearly invertible,
so by the above discussion hA,B is a Mobius transformation. In coordinates, if A has coordinate a
and B has coordinate b, we have
ha,b (z) =
(a + b)z 2ab
.
2z a b
Harmonic conjugation has the property that hA,B (hA,B (C)) = C - in other words, harmonic conjugation is always an involution. In fact, this property characterizes harmonic conjugation.
Theorem 8. If f is a M
obius transformation with the further property that f is an involution,
i.e. f (f (C)) = C for all points C, then f is either the identity map or there is a pair of (possibly
imaginary) points A, B such that f = hA,B .
Proof. In coordinates, the equation f (z) = z becomes a quadratic after clearing the denominator.
If f is not the identity map, this quadratic will have two solutions, corresponding to two distinct
points A, B. For any point C, write D = f (C). Since f preserves the cross ratio, we have
(A, B; C, D) = (f (A), f (B); f (C), f (D)) = (A, B; D, C),
so the points A, B, C, D are harmonic.
1.5
Two special points in the projective plane allow us to talk about angles using cross ratios. These
points are both infinite and imaginary, but we can treat them the same way we treat any other
points in projective geometry. This allows us to solve many problems that are traditionally thought
to be out of the scope of projective geometry.
Definition 8. The circle points are the points = [i : 1 : 0] and
= [1 : i : 0]. These are the
points at infinity of slope i and i.
Theorem 9 (Angle Theorem). If lines l, m intersect the line at infinity in points L, M , then
(L, M ; , )
= e2ilm .
In particular, lines l and m are orthogonal if and only if points L, M, ,
are harmonic.
Proof. Let s be the slope of line l and let t be the slope of line m. By the tangent subtraction
formula, we have
ts
tan(lm) =
.
1 + st
11
We have
= (s, t; i, i)
(L, M ; , )
si
s+i
=
i t i t
(s + i)2 (t i)2
= 2
(s + 1)(t2 + 1)
(st + 1)2 (t s)2 + 2i(t s)(st + 1)
=
(st + 1)2 + (t s)2
1 tan2 (lm)
2 tan(lm)
=
+i
1 + tan2 (lm)
1 + tan2 (lm)
= cos(2lm) + i sin(2lm)
= e2ilm .
Theorem 10. A conic is a circle if and only if it passes through the two circle points.
Proof. First, suppose is a circle with center (a, b) and radius r. In projective coordinates, is
the set of points [x : y : z] such that
(x az)2 + (y bz)2 = (rz)2 .
Plugging in, we can check that [x : y : z] = [i : 1 : 0] and [x : y : z] = [1 : i : 0] satisfy the equation
defining .
Now suppose is any conic passing through , .
Let A, B, C, D be any four points on . Then
we have
e2iCAD = (AC, AD; A, A
) = (C, D; , )
= (BC, BD; B , B )
= e2iCBD ,
so the directed angles CAD and CBD are congruent modulo . Thus A, B, C, D are concyclic.
Corollary 2. Let A, B be two points on a circle with center O. Then
(A, B; , )
= eiAOB .
In particular, if A, B are diametrically opposite then A, B, ,
are harmonic.
Exercise 17. Say that four distinct points A, B, C, D on a line are melodic if we have
(A, B; C, D) = (A, D; B, C),
and make a similar definition for four points on a conic. Let ABCDEF be a regular hexagon
inscribed in a circle . Prove that the four points A, B, ,
are melodic with respect to .
12
1.6
Polar maps
Definition 9. We say that two points P, Q are harmonic conjugates with respect to a conic if
P, Q, X, Y are harmonic, where X, Y are the (possibly imaginary) points of intersection of and
P Q.
Theorem 11. Let P be a point and a conic. Then the locus p of harmonic conjugates of P with
respect to is a line.
Proof 1, using tangents. Let U, V be the feet of the two tangents from P to . We will show that
every point Q on the line U V is a harmonic conjugate of P with respect to . Let the line P Q
meet at X, Y .
U
P
U
B
E
P
V
C Y
R
Theorem 12. Let be a circle with center O. Let P 6= O be a finite point, and let P 0 be its inverse
with respect to the circle . Then the polar of P passes through P 0 and is perpendicular to the line
OP .
Proof. Let meet OP in the points X, Y . When we restrict inversion to the line OP , we see that it
is a nontrivial involution fixing X and Y , so it must be harmonic conjugation with respect to X, Y .
Thus X, Y, P, P 0 are harmonic conjugates (this can also be checked using coordinates, or alternatively by drawing tangents and using facts we have already proven about harmonic quadrilaterals).
Now let OP meet the line at infinity in the point L, and let M be the harmonic conjugate of
L with respect to the circle points , .
Let l, m, o, p denote the polars of L, M, O, P , respectively.
Since the circle points are the intersection of with the line at infinity, L and M are conjugate
with respect to , so M = o l and thus m = OL. Since P is on m, M must be on p, so p = M P 0 .
By the angle theorem, M P 0 is perpendicular to OP , so we are done.
Theorem 13. Let be a conic, and let points A, B, C, D on a line l have polars a, b, c, d. Then
we have
(A, B; C, D) = (a, b; c, d).
Proof. Note that all four lines a, b, c, d pass through L, the pole of l. First suppose that l is
not tangent to . Let intersect l in points X, Y , and let a, b, c, d intersect l at the points
A0 , B 0 , C 0 , D0 . Then by the definition of the polar, the points A0 , B 0 , C 0 , D0 are the harmonic conjugates of A, B, C, D with respect to X, Y . Thus if hXY denotes harmonic conjugation with respect
to X, Y , we have
L
(A, B; C, D) = (hXY (A), hXY (B); hXY (C), hXY (D)) = (A0 , B 0 ; C 0 , D0 ) = (a, b; c, d).
Now suppose the line l is tangent to . Let M be any point not on l or , and let m be its
polar with respect to . Then by the previous case applied to the line m,
(A, B; C, D) = (M A, M B; M C, M D) = (m a, m b; m c; , m d) = (a, b; c, d).
Exercise 18. Give a direct proof of Theorem 13 in the case that the line l is tangent to . (Hint:
consider the map from l to taking a point P on l to the foot of the second tangent from P to .
Prove that this map preserves the cross ratio.)
15
Exercise 19. (a) Let ABCD be a quadrilateral inscribed in a conic . Let E = AB CD, F =
AD BC be the intersections of the opposite sides, and let G = AC BD be the intersection
of the diagonals. Prove that the triangle EF G is self-polar with respect to , that is, that
the polars of E, F, G are F G, GE, EF , respectively.
(b) Let ABC be a self-polar triangle with respect to a conic , and let X, Y, Z be points on
such that Z, A, Y are collinear and X, B, Z are collinear. Prove that Y, C, X are collinear.
Exercise 20. If a, b, c, d, e are lines tangent to a conic , define the cross ratio of a, b, c, d with
respect to by
(a, b; c, d) = (a e, b e; c e, d e).
(a) Show that (a, b; c, d) is independent of the choice of e.
(b) If a, b, c, d meet at A, B, C, D, show that
(a, b; c, d) = (A, B; C, D) .
Exercise 21 (Anders Kaseorg). Let , be distinct circles, and let , be the polar maps with
respect to , . Show that the composite map is the identity if and only
if the circles , have equal radii and intersect in 60 arcs.
1.7
Coharmonic points
For any two pairs of distinct points {A, X} and {B, Y } on a line, we can find a Mobius transformation f satisfying f (A) = X, f (X) = A, f (B) = Y (since Mobius transformations have three
independent parameters). Since f preserves the cross ratio, for any other point C we must have
(X, A; f (C), C) = (A, X; C, f (C)) = (f (A), f (X); f (C), f (f (C))) = (X, A; f (C), f (f (C))),
so C = f (f (C)) and f is a harmonic conjugation in a pair of points {M, N }. Motivated by this
fact, we make the following definition.
Definition 11. Three pairs of points {A, X}, {B, Y }, {C, Z} on the same line are called coharmonic
if there is another pair of (possibly imaginary) points {M, N } such that
(M, N ; A, X) = (M, N ; B, Y ) = (M, N ; C, Z) = 1.
Theorem 14 (Main theorem of coharmonic points). Let A, B, C, X, Y, Z be on a line, no three the
same, and suppose A 6= X. The following are equivalent:
(a) The three pairs of points {A, X}, {B, Y }, {C, Z} are coharmonic.
(b) There is a M
obius transformation f satisfying f (A) = X, f (B) = Y, f (C) = Z which is an
involution.
(c) (A, X; B, C) = (X, A; Y, Z).
(d)
AY CX BZ
Y C XB ZA
= 1.
16
Proof. By the above discussion, (a) and (b) are clearly equivalent. To see the equivalence of (b)
and (c), let f be the M
obius function satisfying f (A) = X, f (X) = A, f (B) = Y . Then since f
preserves the cross ratio, we have
(A, X; B, C) = (f (A), f (X); f (B), f (C)) = (X, A; Y, f (C)),
so f (C) = Z if and only if (A, X; B, C) = (X, A; Y, Z).
Now we show that (b) implies (d). We start by making the definition
(A, B, C; X, Y, Z) =
AY CX BZ
.
Y C XB ZA
c+a
b+c
a+b
= 1.
b c a b c a
Finally, to see that (d) implies (b), note that for any A, B, C, X, Y there is a unique Z such
that (A, B, C; X, Y, Z) = 1, and if f is a Mobius involution taking A to X and B to Y , then
(A, B, C; X, Y, f (C)) = 1 by the above.
Theorem 15 (Three Conic Law). Let A, B, C, D be any four points, no three on a line. Let l be a
line passing through at most one of A, B, C, D. Let 1 , 2 , 3 be three (possibly degenerate) conics
passing through A, B, C, D. For each i = 1, 2, 3, let Xi , Yi be the two points of intersection of conic
i with line l. Then the three pairs {X1 , Y1 }, {X2 , Y2 }, {X3 , Y3 } are coharmonic.
Proof. Consider the following map f from the line l to itself. For any point P on l, let P be
the conic passing through the points A, B, C, D, P , and define f (P ) to be the second point of
intersection of P with the line l. By Theorem 7, or more concretely by the solution to Exercise
7, f is a M
obius transformation. Since f is clearly also an involution satisfying f (Xi ) = Yi for
i = 1, 2, 3, the main theorem of coharmonic points shows that {X1 , Y1 }, {X2 , Y2 }, {X3 , Y3 } are
coharmonic.
Exercise 22. (a) Let be a conic, and let P be a point not on , and let A, B, C be three points
on . Let X, Y, Z be the second intersections of the lines P A, P B, P C with . Show that
the three pairs {A, X}, {B, Y }, {C, Z} are coharmonic with respect to the conic . (Hint: see
Exercise 5.)
17
M
N
L
C
D
B
G
P
M
L
S
F
A
Y
O
(J, P ; L, N ) = (D, F ; U, V ),
so by Theorem 14 its enough to show that {D, F }, {U, T }, {S, V } are coharmonic.
Applying Three Conic Law to the points M, L, J, O, the line DF , the conic and the degenerate
conics M L JO, M J LO, we see that {D, F }, {X, Y }, {U, T } are coharmonic. Similarly, applying
19
the Three Conic Law to the points M, N, K, P , the line DF , the conic and the conics M N
KP, M P N K, we see that {D, F }, {X, Y }, {S, V } are coharmonic.
Thus the harmonic conjugation map that exchanges D with F and exchanges X with Y also
exchanges U with T and S with V , so {D, F }, {U, T }, {S, V } are coharmonic and we are done.
Proof 2 (from [2], using Pascals Theorem). Again, we assume that I, J, K, L, M, N, O, P lie on a
conic . Its enough to show that G, H, A, B, C, D lie on a conic, since then by symmetry we have
H, A, B, C, D, E on a conic, etc.
Let X be the intersection of lines KP and IJ. Applying Pascals Theorem to the hexagon
M P KN IJ inscribed in the conic , we see that D, G, X lie on a line. From this we see that I, J, X
are the intersections of the opposite sides of the hexagon GHABCD, so by the converse to Pascals
Theorem GHABCD is also inscribed in a conic.
M
N
C
D
B
G
P
1.8
a b c
Definition 12. Let M = d e f be a three by three matrix with nonzero determinant.
g h i
2
2
The map fM : P P defined by f ([x : y : z]) = [ax + by + cz : dx + ey + f z : gx + hy + iz] is
called a projective transformation of the plane.
Exercise 28. (a) Show that every projective transformation sends straight lines to straight lines,
sends conics to conics, and preserves cross ratios.
(b) Show that if M, N are three by three matrices with nonzero determinants, then fM fN =
fM N .
20
(c) Show that if A, B, C, D are any four points with no three on a line, and E, F, G, H are
any four points with no three on a line, then there is a projective transformation f with
f (A) = E, f (B) = F, f (C) = G, f (D) = H.
Definition 13. A bijection f : P2 P2 is a collineation if it takes straight lines to straight lines.
Exercise 29. (a) Let A, B, C, D, E, F be six distinct points on a line. Show that {A, B}, {C, D}, {E, F }
are coharmonic if and only if
(A, B; C, D) = (A, B; C, E)(A, B; C, F ).
(b) Given distinct points A, B, C, D, E on a line, construct points F and G on the same line such
that
(A, B; C, F ) = (A, B; C, D)(A, B; C, E)
and
(A, B; C, G) = (A, B; C, D) + (A, B; C, E)
using only a straightedge.
Exercise 30. Let f : R R be a function such that f (1) = 1 and such that for any x, y R we
have f (xy) = f (x)f (y) and f (x + y) = f (x) + f (y). Show that f (x) = x for all x R.
Theorem 17 (Fundamental theorem of projective geometry). A bijection f : RP2 RP2 is a
collineation if and only if it is a projective transformation (here we write RP2 for the real points of
the projective plane).
Proof. We start by showing that if f is a collineation then it must preserve cross ratios. If A, B, C, D
are distinct points on a line and E, F, G, H are distinct points on another line, then by Theorem 4 we
can check whether (A, B; C, D) = (E, F ; G, H) by checking whether the points X = AF BE, Y =
BG CF, Z = CH DG lie on a line. Since f is a collineation, we have f (X) = f (AF ) f (BE) =
f (A)f (F ) f (B)f (E) and so on, and f (X), f (Y ), f (Z) lie on a line if and only if X, Y, Z lie on a
line, so
(A, B; C, D) = (E, F ; G, H) (f (A), f (B); f (C), f (D)) = (f (E), f (F ); f (G), f (H)).
Thus we get a well-defined bijection f : R {} R {} by taking
f((A, B; C, D)) = (f (A), f (B); f (C), f (D)).
This bijection automatically satisfies f(0) = 0, f(1) = 1, f() = . By Exercise 29 we have
f(xy) = f(x)f(y) and f(x + y) = f(x) + f(y) for any real x, y, and thus by Exercise 30 we must
have f(x) = x for all real x. Thus f preserves cross ratios.
To finish, note that by Exercise 28 we may assume without loss of generality that f fixes some
collection of four points A, B, C, D such that no three are on a line. Letting P = AB CD, we see
that f (P ) = P , and thus for any point X on AB we have
(A, B; P, X) = (f (A), f (B); f (P ), f (X)) = (A, B; P, f (X)),
so f (X) = X. Thus if l is any line through C, and X = l AB, then f (l) = f (C)f (X) = CX = l,
so every line through C is sent to itself. Similarly, every line through A or B is sent to itself. Since
any point E is determined by the three lines AE, BE, CE, every point E must be sent to itself,
and we are done.
21
Remark 1. A collineation of CP2 might not preserve cross ratios: for instance, the map [x : y :
z] 7 [
x : y : z] taking every point to its complex conjugate sends every cross ratio to its complex
conjugate. More generally, if f : C C satisfies f(1) = 1, f(xy) = f(x)f(y), f(x+y) = f(x)+ f(y),
then the map [x : y : z] 7 [f(x) : f(y) : f(z)] is called an automorphic collineation, and sends a set
of four points on a line with cross ratio c to a set of four points with cross ratio f(c).
The same argument as above can be used to show that every collineation of CP2 can be written
as the composition of an automorphic collineation and a projective transformation.
Definition 14. Let P be a point and l be a line not passing through P . Define the projective
reflection rP,l by sending a point Q 6= P to the harmonic conjugate of Q with respect to P, P Q l
along the line P Q, and sending P to P .
Example 2. (a) Let l intersect the line at infinity at L. If P is on the line at infinity with L, P, ,
harmonic, then rP,l is (ordinary) reflection across the line l. As a consequence, (ordinary)
reflections always interchange the two circle points.
(b) If l is the line at infinity, then rP,l is a 180 rotation around P (sometimes called a reflection
through the point P ).
Theorem 18. For any point P and any line l not passing through P , the projective reflection rP,l
is a projective transformation.
Proof. We just need to show that rP,l sends lines to lines and preserves cross ratios. We leave this
as an easy exercise to the reader.
Definition 15. If A, B, C, D are four points with no three on a line and : {A, B, C, D}
{A, B, C, D} is a permutation, define r to be the projective transformation taking A to (A), B
to (B), etc. We will often write using its cycle decomposition, including the cycles of length
1, so that for instance r(A)(B)(C D) is the projective transformation taking A and B to themselves,
and swapping C and D.
Exercise 31. Suppose A, B, C, D are four points with no three on a line.
(a) If P = AB CD and l is the line connecting AC BD to AD BC, show that r(A B)(C D)
is the projective reflection rP,l .
(b) Show that if is a conic passing through A, B, C, D then r(A B)(C D) () = .
Exercise 32. (a) Show that for every permutation : {A, B, C, D} {A, B, C, D} we can write
r as a composition of two projective reflections.
(b) Show that a projective transformation defined by a three by three matrix M can be written
as a composition of two projective reflections if and only if the eigenvalues of M are in a
geometric progression.
Exercise 33. Let f (p, q, r), g(p, q, r), h(p, q, r) be homogeneous polynomials of the same degree having no common factor. The map [p : q : r] 7 [f (p, q, r) : g(p, q, r) : h(p, q, r)] is called biregular if it
is defined everywhere (i.e. f, g, h are never simultaneously 0 unless p, q, r are all 0) and is a bijection of the complex points of the projective plane. Prove that every biregular map is a projective
transformation.
22
1.9
Since any four points (no three on a line) can be sent to any other four points (no three on a line) by
a projective transformation, there are no interesting invariants of four general points in the plane.
If we have five general points A, B, C, D, E, then we can form the cross ratio (EA, EB; EC, ED).
Going one step further, we have the following natural definition.
Definition 16. Let A, B, C, D, E, F be six points in the plane, such that either none of ACE, ADF ,
BCF, BDE are lines or none of ACF, ADE, BCE, BDF are lines. Define their cross cross ratio
to be
(EA, EB; EC, ED)
(A, B; C, D; E, F ) =
.
(F A, F B; F C, F D)
First we will prove that this definition is more symmetric than it seems.
E F
D
C
X M
A B
Y N
(A, B; C, D; E, F ) =
Proposition 6. Let two circles , 0 intersect at points A, B, and let C be a point on , D a point
on 0 . Let be the (directed) angle of intersection between the circles , 0 at A. Then we have
(A, B; , ;
C, D) = e2i .
In particular, and 0 are orthogonal if and only if (A, B; , ;
C, D) = 1.
23
Proof.
(A, B; , ;
C, D) =
(A, B; , )
= e2i(ACBADB) = e2i .
(A, B; , )
0
(A, B; C, D)
.
(A, B; C, D)0
If (A, B; C, D; , 0 ) = 1, we say that the conics , 0 are projectively orthogonal with respect to
the partition {A, B}, {C, D} of their intersection points.
(A, B; C, D)
(E, F ; P, X)
(E, F ; P, Q)
=
=
= (E, F ; P, Q).
(A, B; C, D)
(E, F ; Q, Y )
(E, F ; X, Y )
A
E
Q P
Y
D
25
1.10
Exercise 36. (a) Let ABC be a triangle, let D be a point on BC, let E be a point on CA, and
let F be a point of AB. Show that the lines AD, BE, CF meet in a point if and only if there
is a conic which is tangent to BC at D, tangent to CA at E, and tangent to AB at F .
(b) Let ABC be a triangle and let P be a point not lying on any edge of ABC. Let U, X be
points on BC with X = r(A)(P )(B C) (U ), let V, Y be points on CA with Y = r(B)(P )(A C) (V ),
and let W, Z be points on AB with Z = r(C)(P )(A B) (W ). Show that U, V, W, X, Y, Z lie on a
conic.
27
2.1
Let A, B, C, D be four points in the projective plane, no three on a line. Choose projective coordinates such that A = [1 : 0 : 0], B = [0 : 1 : 0], C = [0 : 0 : 1], D = [1 : 1 : 1] (one way to do this is
to start with barycentric coordinates on the triangle A, B, C, and then rescale the coordinates to
make D = [1 : 1 : 1]). For future reference, let E = [1 : 1 : 1], F = [1 : 1 : 1], G = [1 : 1 : 1] in
this coordinate system.
Exercise 40. Show that E, F, G satisfy DE F G = A, DF EG = B, DG EF = C, and that
they are uniquely determined by these conditions. Show that E is the harmonic conjugate of D
with respect to A, AD BC.
One of the simplest nonlinear functions we can write down is the Cremona involution: if p, q, r
are all nonzero, it takes the point P = [p : q : r] in the above coordinate system to the point
fABCD (P ) = [ p1 : 1q : 1r ]. We would like to extend this to an involution of the plane. Clearing
denominators, we get fABCD (P ) = [qr : pr : pq], and this lets us define fABCD (P ) as long as no
two of p, q, r are 0, i.e. as long as P is not equal to one of A, B, C. If P is on line BC, then p = 0,
so fABCD (P ) = [qr : 0 : 0] = A, and fABCD is not injective. We can fix these problems by blowing
up the points A, B, C.
Definition 18. If A, B, C are three distinct points in the projective plane, we set
BlABC P2 = {(P, lA , lB , lC ) | P P2 , {A, P } lA , {B, P } lB , {C, P } lC }.
If (P, lA , lB , lC ) BlABC P2 has P 6= A, B, C, then lA = AP, lB = BP, lC = CP , and we write P
as shorthand for (P, lA , lB , lC ). Let eA be the set of points (P, lA , lB , lC ) in BlABC P2 with P = A,
that is,
eA = {(A, l, AB, AC) | A l},
and define eB , eC similarly. If (A, l, AB, AC) eA , we write (A, l) as shorthand for it. If P = (A, l),
we write AP as shorthand for l. The three lines eA , eB , eC are called the exceptional lines above
A, B, C. We say that a curve passing through A intersects the exceptional line eA in the point
(A, lA ) if line lA is tangent to at A.
In coordinates, we have
BlABC P2 = {([p : q : r], (0 : a : b), (c : 0 : d), (e : f : 0)) | aq + br = cp + dr = ep + f q = 0}.
Proposition 7. The map [p : q : r] 7 [ p1 : 1q : 1r ], defined for p, q, r 6= 0, extends to an involution
fABCD : BlABC P2 BlABC P2 . The extended involution fABCD takes eA (resp. eB , eC ) bijectively
to BC (resp. AC, AB). The fixed points of fABCD are D, E, F, G, and we have fABCD = fABCE =
fABCF = fABCG .
Proof. In coordinates, if P = ([p : q : r], (0 : a : b), (c : 0 : d), (e : f : 0)) we set
Remark 2. More generally, for any three homogeneous polynomials f (p, q, r), g(p, q, r), h(p, q, r) of
the same degree having no common factor we can define a map [p : q : r] [f (p, q, r) : g(p, q, r) :
h(p, q, r)], which is well-defined whenever f, g, h are not simultaneously zero. Such a map is called
a rational map. It is called birational if it is usually one-to-one - in this case you can write down a
rational function which inverts it whenever both are defined. Noether and Castelnuovo have proved
that every birational map P2 P2 can be built out of projective transformations and Cremona
involutions.
Proposition 8. Let A, B, C, D, E, F, G be such that A = DE F G, B = DF EG, C = DG EF ,
and suppose that fABCD (P ) = Q. Then we have
(AP, AQ; DE, F G) = (BP, BQ; DF, EG) = (CP, CQ; DG, EF ) = 1.
In other words, AQ is the harmonic conjugate of AP with respect to AD, AF , and similarly for
BQ, CQ.
Proof. By symmetry, its enough to show that (AP, AQ; DE, F G) = 1. In the coordinate system
described above, suppose that AP = (0 : a : b). We then have DE = (0 : 1 : 1), F G = (0 : 1 :
1), AQ = (0 : b : a), so
(AP, AQ; DE, F G) = (a/b, b/a; 1, 1) = 1.
Example 3. Let G be the centroid of triangle ABC, and suppose fABCG (P ) = Q. Let F ED have
parallel sides to ABC, such that A is the midpoint of DE, B is the midpoint of DF , and C is the
midpoint of EF . Let M = AG BC, X = AP BC, Y = AQ BC, = AD BC. Then
(X, Y ; , M ) = (AP, AQ; DE, F G) = 1,
so X is the reflection of Y across M , the midpoint of BC. Similarly, BP AC is the reflection of
BQ AC across the midpoint of AC, etc. The point Q is called the isotomic conjugate of P .
Corollary 3. Let f = fABCD . For any four points P, Q, R, S BlABC P2 we have
(AP, AQ; AR, AS) = (Af (P ), Af (Q); Af (R), Af (S)).
Proof. Harmonic conjugation preserves the cross ratio.
Theorem 21. Let l be a line which does not pass through any of A, B, C. Then fABCD (l) is a
circumconic, that is, a conic passing through all three of A, B, C. Conversely, if is a circumconic
then fABCD () is a line which does not pass through any of A, B, C.
Proof. Write f = fABCD , and let P, Q, R be any three points on l. Let S = l BC, so that
f (S) eA . By the Corollary, we have
(Bf (P ), Bf (Q); Bf (R), BA) = (P, Q; R, S) = (Cf (P ), Cf (Q); Cf (R), CA),
so f (R) lies on the conic through A, B, C, f (P ), f (Q). The converse is left as an exercise.
Exercise 41. Let I be the incenter of triangle ABC. The map fABCI is called isogonal conjugation.
(a) Show that fABCI () = .
29
(b) Let be the circumcircle of triangle ABC. Show that fABCI () is the line at infinity.
(c) Let m be the median through A. Show that fABCI (m) passes through the pole of BC with
respect to . (Hint: show that the intersections of m, BC, AB, AC with the line at infinity
are harmonic, then apply fABCI .)
(d) Let be the circumcircle of triangle BCI. Show that fABCI () = .
Exercise 42. Write f = fABCD , let l be a line which doesnt pass through any of A, B, C, let
= f (l), and let P, Q, R, S be any four points on l. Show that
(P, Q; R, S) = (f (P ), f (Q); f (R), f (S)) .
Theorem 22. If is a conic which passes through B and C but not A, then fABCD () is also a
conic passing through B and C but not A. We have fABCD () = if and only if either passes
through D and E or passes through F and G.
Proof. Write f = fABCD , and let P, Q, R, S be any four points on . By Corollary 3, we have
B
Definition 20. If has type (m, p, q, r), then the self-intersection number of is defined to be
m2 p2 q 2 r2 .
Proposition 9. If has type (m, p, q, r) then fABCD () has type (2m + p + q + r, m q r, m
p r, m p q).
Exercise 43.
(b) Using Proposition 9 and Theorem 24, check that the number of intersection points between
and is the same as the number of intersection points between fABCD () and fABCD ().
In particular, the self-intersection number of is the same as the self-intersection number of
fABCD ().
(c) Use Proposition 9 to give another proof of Theorem 21.
(d) Find all curves in BlABC P2 which have self-intersection number at most 0.
References
[1] H. S. M. Coxeter and S. L. Greitzer. Geometry revisited, volume 19 of New Mathematical
Library. Random House, Inc., New York, 1967.
[2] Lawrence S. Evans and John F. Rigby. Octagrammum mysticum and the golden cross-ratio.
The Mathematical Gazette, 86(505):pp. 3543, 2002.
31