COLLINEATIONS OF PROJECTIVE PLANES

TIMOTHY VIS

1. Fixed Structures
The study of geometry has three main streams:
(1) Incidence Geometry: the study of what points are incident with what lines
and what can be derived about a geometry.
(2) Analytic Geometry: the study of coordinates of a geometry and the algebra
derived, as well as geometric properties involving this algebra.
(3) Transformational Geometry: the study of collineations of a geometry and
the way structures can be moved around.
All three areas have been covered so far in class. Incidence Geometry was used
to give axiomatic descriptions of projective planes and projective spaces, and to
prove results such as the fact that all lines in a finite projective space are incident
with a constant number of points. Analytic Geometry was used in our study of field
planes and in the development of coordinates when we coordinatized an arbitrary
plane. Transformational Geometry was used when we developed the Fundamental
Theorem of Field Planes. We are now going to study Transformational Geometry
in an arbitrary plane and see what we can say about collineations in general.
Definition 1. A collineation of a projective plane π is a bijective map on the points
such that the images of collinear points are collinear. That is, P , Q, R are collinear
if and only if P σ , Qσ , Rσ are collinear.
Given a line l and points P , Q on l, we define lσ to be the line P σ Qσ .
One of the most fundamental properties studied about collineations is the prop-
erty of which points and lines are mapped to themselves.
Definition 2. If σ is a collineation of a projective plane π,
(1) A point P is a fixed point if and only if P σ = P .
(2) A line l is a fixed line if and only if lσ = l.
(3) A line l is fixed pointwise if and only if P σ = P for all points P ∈ l.
(4) A point P is fixed linewise if and only if lσ = l for all lines l containing P .
Obviously any line that is fixed pointwise is a fixed line and any point that is
fixed linewise is a fixed point, but the converse is not necessarily true.
Example 1. Consider the collineation of the Fano plane shown in Figure 1. Notice
that the points A, F , and G are fixed and that the lines ABC, ADE, and AF G
are fixed. Since all points on the line AF G are fixed, AF G is fixed pointwise, and
since all lines through A are fixed, A is fixed linewise.

Date: April 23, 2009.

1
2 TIMOTHY VIS

A A

B F C F
D E

C G B G
E D

Figure 1

We’ll focus somewhat on the applications of these ideas to field planes; however
they apply equally well to arbitrary projective planes. Recall the Fundamental
Theorem of Field Planes, which states that every collineation of a field plane is
the product of an automorphic collineation and a homography. That is, if σ is a
collineation, we have
σ : X 7→ AX α
where α is an automorphism of the underlying field. We also have for lines
σ : L 7→ Lα A−1 .
Suppose that α is the identity collineation. We then have X σ = AX and Lσ =
T
T
LA−1 or (Lσ ) = A−1 LT . If a point X is fixed by σ, we then have AX = λX,

T
while if a line L is fixed by σ, we have A−1 LT = µLT . Thus the eigenvectors

T
of A are the fixed points of σ, while the eigenvectors of A−1 are the fixed lines
of σ.

Example 2. Consider the homography given by

 
a 0 0
A= 0 1 0 
0 0 1
where a ∈ F \ {0, 1}. The fixed points are the eigenvectors of A, while the fixed
lines are the eigenvectors of
 1 
a 0 0

T
A−1 =  0 1 0 .
0 0 1
To find the eigenvectors of A, we determine the characteristic polynomial of A.
 
 λ−a 0 0 

|λI − A| =  0 λ−1 0 
 0 0 λ−1 
2
= (λ − a) (λ − 1) .
 COLLINEATIONS OF PROJECTIVE PLANES 3

Thus A has eigenvalues a and 1. Corresponding to a are eigenvectors determined

as vectors in the null space of aI − A. That is,
      
0 0 0 x 0 0
 0 a−1 0   y  =  (a − 1) y  =  0  .
0 0 a−1 z (a − 1) z 0
Now since a 6= 1, we must have both y = 0 and z = 0. Thus, the only fixed point
corresponding to the eigenvalue a is (1, 0, 0).
Corresponding to 1 are eigenvectors determined as vectors in the null space of
I − A. That is,
      
1−a 0 0 x (1 − a) x 0
 0 0 0  y  =  0  =  0 .
0 0 0 z 0 0
Again, since a 6= 1, we must have x = 0. Thus, the fixed points corresponding to
the eigenvalue 1 are all points of the form (0, y, z).

T
On the other hand, notice that A−1 , being diagonal, is self-transpose. So A−1
has the same eigenvectors as A−1 , which are precisely the eigenvectors of A. Thus,
the fixed lines have the same coordinates as the fixed points: [1, 0, 0] and [0, y, z].
Now notice that every line through (1, 0, 0) is of the form [0, y, z], and each of
these lines are fixed, so that (1, 0, 0) is fixed linewise. Similarly, every point on
[1, 0, 0] is of the form (0, y, z), and each of these points are fixed, so that [1, 0, 0] is
fixed pointwise.
We now prove some results regarding what possibilities exist for fixed points and
lines. We shall assume the following easily verified facts (which you prove in your
homework):
Proposition 3. Given a collineation σ,
(1) If P and Q are distinct fixed points of σ then P Q is a fixed line of σ.
(2) If l and m are distinct fixed lines of σ then l ∩ m is a fixed point of σ.
These results allow us to prove several other results.
Proposition 4. Let σ be a collineation of π fixing l pointwise. If distinct points P
and Q in π \ l are fixed by σ, then σ is the identity collineation (σ fixes all points
of π.
Proof. Suppose R is some point of π that does not lie on either l or P Q. So P R
and QR are distinct lines through R meeting l in points P ′ and Q′ respectively.
But P R = P P ′ and since both P and P ′ are fixed, P R is fixed. Similarly, QR is
fixed. But then P R ∩ QR = R is fixed.
Now let S be a point of P Q other than P and P Q ∩ l. Let R be any point on
neither l nor P Q. Then P , R, S are non-collinear points, and both P and R are
fixed. By the last argument then, S is also fixed. So every point of π is fixed, and
σ is the identity collineation. 
Collineations that fix a line pointwise or a point linewise play a central role in
the study of projective planes. One of the most fundamental results regarding such
collineations is the following theorem that states that any collineation fixing a line
pointwise fixes a point linewise. The converse of this statement is, of course, also
true by the principle of duality.
4 TIMOTHY VIS

Theorem 5. If σ is a collineation of π fixing l pointwise, σ fixes some point V

linewise.

Proof. If σ fixes a point V not on l let m be a line through V and let m ∩ l = W .

Then m = V W , and since both V and W are fixed, m is also fixed. So V is fixed
linewise.
Suppose then that no point of π \ l is fixed by σ. Let P be a point of π \ l and
σ
consider the line P P σ . This line intersects l in a point V . So P V = P σ V = (P V )
and is fixed by σ. Let Q be any point of π \ (l ∪ P V ). In the same manner as for
P , QQσ is a fixed line. So P P σ ∩ QQσ is a fixed point and must then lie on l. So
QQσ ∩ l = V as well.
Now consider any line m through V and let R be any point on this line. Then
RRσ is a fixed line through V , so that RRσ = m and m is fixed. So V is fixed
linewise. 

Corollary 6. If σ is a collineation of π fixing V linewise, σ fixes some line l

pointwise.

Proof. This is the dual to Theorem 5. 

With these results in hand, we are ready to make several definitions.

Definition 7. A collineation σ fixing a point V linewise and a line l pointwise is
called a (V, l)-perspectivity or central collineation. The point V is called the center
of σ and the line l is called the axis of σ. If V ∈ l, σ is called an elation, while if
V 6∈ l, σ is called a homology.
Example 3. Two of the most familiar motions in the Euclidean plane give us
examples of central collineations of the Extended Euclidean Plane. Consider a
translation of the Euclidean plane. No points of the Euclidean plane are fixed,
but a translation fixes the slopes of all lines in the Euclidean plane. These slopes
correspond exactly to the points at infinity, so that every point at infinity is fixed
by a translation, and thus, l∞ is the axis of every translation. Furthermore, the
lines parallel to the direction of the translation stay in place, so that these lines are
fixed. But then every line through the point at infinity corresponding to this slope
is fixed, so that this point is the center of the translation. Since this point is on l∞ ,
a translation is an elation with axis l∞ .
Now consider a reflection over the line l of the Euclidean plane. Every point on
l is fixed by this reflection, so l is the axis of the reflection. The only lines that
are fixed by a reflection, however, are the lines perpendicular to l, so the point at
infinity corresponding to the slope perpendicular to l is the center of the reflection.
So a reflection is a homology of the Extended Euclidean Plane.
Example 4. The homography determined by the matrix
 
a 0 0
 0 1 0 
0 0 1
was shown to have the point (1, 0, 0) fixed linewise and the line [1, 0, 0] fixed point-
wise. Since (1, 0, 0) does not lie on the line [1, 0, 0], this homography defines a
homology.
 COLLINEATIONS OF PROJECTIVE PLANES 5

Example 5. Consider the homography determined by the matrix A and its inverse

T
transpose A−1 as given.
   
1 0 1 1 0 0

T
A= 0 1 1  A−1 =  0 1 0 .
0 0 1 −1 −1 1

The characteristic polynomial of both matrices is (λ − 1)3 and the only eigenvalue

T
of either is 1. A has eigenvectors (x, y, 0) and A−1 has eigenvectors [x, −x, z]
Thus, the fixed points are the points of the form (x, y, 0), while the fixed lines are
the lines of the form [x, −x, z]. Notice that the fixed points all lie on the line [0, 0, 1]
so that this line is fixed pointwise. Similarly, the fixed lines all contain the point
(1, 1, 0). Since (1, 1, 0) ∈ [0, 0, 1], this homography defines an elation.
A significant property of central collineations is that they are very easily deter-
mined.
Theorem 8. A central collineation is completely and uniquely determined by its
axis l, center V , and the image P σ of any point P distinct from V not lying on l.

Proof. Let R be any point of π \ (l ∪ P V ) and let RP ∩ l = W and RV = m. Then

R = P W ∩ m so that
σ
Rσ = (P W ∩ m)
= P σ W σ ∩ mσ
= P σ W ∩ m.
Thus, Rσ is completely and uniquely determined.
Now let S ∈ P V and let R ∈ π \ (l ∪ P V ). So S ∈ π \ (l ∪ RV ) and since Rσ
is completely and uniquely determined, S σ is completely and uniquely determined
using V , l, and the image Rσ of R as argued when determining Rσ . 

A word of warning here: this theorem applies only when a central collineation
actually exists with the appropriate property. There may not be any collineations
(other than the identity) with a particular axis, with a particular center, with a
particular axis, center pair, or mapping a given point to any other. This theorem
does not state that such a collineation exists; it only states that when one exists, it
is unique. In order to show that one exists, we need particular properties of a plane.
For example, in a field plane, every possible central collineation exists, although we
will not prove this.
Proposition 9. If σ is a (V, l)-perspectivity and τ a collineation, then τ −1 στ is a
(V τ , lτ )-perspectivity.

Proof. Let m be a line through V τ . Then mτ necessarily contains V and is fixed

−1

by σ.
 −1 σ τ
mτ στ = mτ
−1

 −1 τ
= mτ
= m.
6 TIMOTHY VIS

So m is fixed by τ −1 στ and V τ is a center of τ −1 στ . Similarly, if P is a point on

lτ , P τ lies on l and is fixed by σ.
−1

 −1 σ τ
P τ στ = P τ
−1

 −1 τ
= Pτ
= P.
So P is fixed by τ −1 στ and lτ is an axis of τ −1 στ . So τ −1 στ is a (V τ , lτ )-
perspectivity. 
Example 6. Suppose in a field plane there is a ((1, 0, 0) , [1, 0, 0])-perspectivity σ
σ
such that (1, 1, 1) = (1, b, b). What is the equation of this perspectivity? We know
that as a collineation, σ must be the product of an automorphic collineation and a
homography. We also know that every automorphic collineation fixes every point
of the fundamental quadrangle.
Notice now that the points of the triangle of reference are all fixed by σ, and
thus by the associated homography A ((1, 0, 0) is the center and both (0, 1, 0) and
(0, 0, 1) are on the axis). Thus,
    
a00 a01 a02 1 0 0 ρx 0 0
 a10 a11 a12   0 1 0  =  0 ρy 0  .
a20 a21 a22 0 0 1 0 0 ρz
Since (1, 1, 1) is fixed by automorphic collineations, we must have (up to scalar
multiples) ρx = 1, ρy = ρz = b. So the homography of this collineation is
 
1 0 0
A =  0 b 0 .
0 0 b
Now notice that for any point X on [1, 0, 0], X = (0, y, z). But AX = bX then,
so that X is fixed by A. Further, notice that for any line L through (1, 0, 0),
L = [0, y, z] and LA−1 = 1b L, so that L is fixed by A. So A is the desired
((1, 0, 0) , [1, 0, 0])-perspectivity.
In the last example, we did not need any automorphic collineation to obtain the
appropriate perspectivity. In fact, in a finite field plane, every (V, l)-perspectivity
is a homography; that is, every (V, l)-perspectivity has the identity as associated
automorphic collineation.
Theorem 10. Every (V, l)-perspectivity of a finite field plane is a homography.
That is, every (V, l)-perspectivity of a finite field plane has the identity as the asso-
ciated automorphic collineation.
Proof. By Proposition 9 and the transitivity of homographies on ordered quadran-
gles in P G (2, q), we need only show the result holds for all perspectivities with
a given axis. Other axes are then obtained by conjugation by an appropriate
homography. Suppose then that l = [1, 0, 0]. Let α be the associated automor-
phic collineation and A = (aij ) be the associated homography, and let σ be the
collineation in question.
Now since both (0, 1, 0) and (0, 0, 1) are fixed both by σ and α, we know that
σ
a01 = a02 = a12 = a21 = 0. Let a11 = ρy and let a22 = ρz . Now consider (0, 1, u) .
 COLLINEATIONS OF PROJECTIVE PLANES 7

σ
Since this point lies on the axis, it must be fixed. But (0, 1, u) = A (0, 1, uα) =
(0, ρy , uα ρz ). It follows that
ρz
u = uα
ρy
α ρz
0=u −u
ρy
∗
for all values of u in GF (q) . If q = ph , we need ph − 1 zeroes to a polynomial of
degree α. But α ≤ ph−1 , a contradiction unless uα ρρyz − u is the zero polynomial.
But this can only be the zero polynomial if α = 1, so we necessarily have α = 1,
and σ is a homography.