arXiv:math/0612499v1 [math.

MG] 18 Dec 2006

Affine and Projective Universal Geometry

N J Wildberger
School of Mathematics and Statistics
UNSW Sydney 2052 Australia
February 2, 2008
Abstract
By recasting metrical geometry in a purely algebraic setting, both
Euclidean and non-Euclidean geometries can be studied over a general
field with an arbitrary quadratic form. Both an affine and a projective
version of this new theory are introduced here, and the main formulas
extend those of rational trigonometry in the plane. This gives a unified,
computational model of both spherical and hyperbolic geometries, allows
the extension of many results of Euclidean geometry to the relativistic
setting, and provides a new metrical approach to algebraic geometry.

Introduction
Universal Geometry extends Euclidean and non-Euclidean geometries to general fields and quadratic forms. This new development is a natural outgrowth
of Rational Trigonometry as described in the elementary text [5]. It was there
developed in the planar context with an emphasis on the applications to Euclidean geometry. In this paper the subject is built up in two very general
settingsthe affine one in an n dimensional vector space over a general field
with a metrical structure given by an arbitrary, but fixed, symmetric bilinear
form, and the associated projective one involving the space of lines through the
origin of a vector space with a symmetric bilinear form.
This allows us to dramatically simplify the usual trigonometric relations
for both Euclidean and non-Euclidean geometries, to extend them to general
bilinear forms, and to reveal the rich geometrical structure of projective space,
with interesting implications for algebraic geometry.
It is pleasant that the main laws of planar rational trigonometry have affine
and projective versions which turn out to hold simultaneously in elliptic geometry, in hyperbolic geometry, and indeed in any metrical geometry based on a
symmetric bilinear form. The usual dichotomy between spherical and hyperbolic
trigonometry deserves re-evaluation.
The first section introduces and motivates the new approach in a particularly
simple but important special casethat of two dimensional hyperbolic geometry. The second section establishes the basic trigonometric laws in a general
1

affine setting, using n dimensional space with an arbitrary symmetric bilinear

form over a general field. The Triple quad formula, Pythagoras theorem, the
Spread law, Thales theorem, the Cross law and the Triple spread formula include algebraic analogs of the familiar Sine law and Cosine law, along with the
fact that the sum of angles in a triangle is 180 . Then we derive the projective
versions of these laws, which are seen to be deformations of the affine ones,
along with formulas for projective right, isosceles and equilateral triangles, including general versions of Napiers rules for solving right triangles in spherical
trigonometry. We briefly mention the important Spread polynomials, which are
universal analogs of the Chebyshev polynomials of the first kind, but have an
interpretation over any field. Two examples are shown, one affine and the other
projective, one over the rationals and the other over a finite field. Finally, the
Lambert quadrilateral from hyperbolic geometry is shown to be a general feature
in the projective setting.
Because of the novelty of the approach, some remarks of a subjective nature
may be useful to orient the reader. Metrical geometry is here presented as fundamentally an algebraic subject rather than an analytic one, and the main division
in the subject is not between Euclidean and non-Euclidean, but rather between
affine and projective. Elliptic and hyperbolic geometries should be considered
as projective theories. Their natural home is the projective space of a vector
space, with metrical structurenot a metric in the usual sensedetermined by
a bilinear or quadratic form. Over arbitrary fields the familiar close relation between spheres or hyperboloids and projective space largely disappears, and the
projective space is almost always more basic. The fundamental formulas and
theorems of metrical geometry are those which hold over a general field and are
independent of the choice of bilinear form. Many results of Euclidean geometry
extend to the relativistic setting, and beyond, once you have understood them
in a universal framework.
This paper lays out the basic tools to begin a dramatic extension of Euclidean
and non-Euclidean geometries, and to once again investigate those aspects of
algebraic geometry concerned with the metrical properties of curves and varieties, in the spirit of Archimedes, Apollonius and the seventeenth and eighteenth
century mathematicians.

A motivating example from the hyperbolic plane

Hyperbolic geometry is usually regarded as either a synthetic theory or an analytic theory. With the synthetic approach you replace Euclids fifth axiom with
an axiom that allows more than one line through a point parallel to a given
line, and follow Bolyai, Gauss and Lobachevsky as described in [3]. With the
more modern analytic approach, you introduce a Riemannian metric in the upper half plane or Poincare disk (or sometimes the hyperboloid of two sheets in
three dimensional space), calculate the geodesics, derive formulas for hyperbolic
distancesoften employing the group P SL2 of isometries, and then develop hyperbolic trigonometry. This is described in many places, for example [1] and [4].

The initial interest is in the regular tesselations of the hyperbolic plane, complex analysis and Riemann surfaces, and the connections with number theory
via quadratic forms and automorphic forms, although nowadays the applications
extend much further.
The new approach to be described here is entirely algebraic and elementary,
and allows us to formulate two dimensional hyperbolic geometry as a projective
theory over a general field. There are numerous computational, pedagogical and
conceptual advantages.
Begin with three dimensional space, with typical vector [x, y, z] and bilinear
form
[x1 , y1 , z1 ] [x2 , y2 , z2 ] = x1 x2 + y1 y2 z1 z2 .
Define the projective point a = [x : y : z] to be the line through the origin O =
[0, 0, 0] and the non-zero vector [x, y, z] . The projective quadrance between
projective points a1 = [x1 : y1 : z1 ] and a2 = [x2 : y2 : z2 ] is the number
2

q (a1 , a2 ) =

(x2 y1 x1 y2 ) (y1 z2 z1 y2 ) (z1 x2 x1 z2 )

(x21 + y12 z12 ) (x22 + y22 z22 )
2

=1

(x1 x2 + y1 y2 z1 z2 )
.
(x21 + y12 z12 ) (x22 + y22 z22 )

Define the projective line L = (l : m : n) to be the plane through the origin

(in three dimensional space) with equation
lx + my nz = 0.
The projective spread between projective lines L1 = (l1 : m1 : n1 ) and L2 =
(l2 : m2 : n2 ) is the number
S (L1 , L2 ) =

(l1 m2 l2 m1 )2 (m1 n2 m2 n1 )2 (n1 l2 n2 l1 )2

.
(l12 + m21 n21 ) (l22 + m22 n22 )

To give a concrete example, consider the projective triangle a1 a2 a3 with

projective points
a1 = [1 : 0 : 2]

a2 = [1 : 1 : 3]

a3 = [2 : 1 : 5] .

Over the real numbers these lines would correspond to (pairs of) points on the
usual hyperboloid of two sheets inside the null cone x2 + y 2 z 2 = 0. The
projective lines determined by these projective points are
L1 = a2 a3 = (8 : 1 : 3)

L2 = a1 a3 = (2 : 1 : 1)

L3 = a1 a2 = (2 : 1 : 1) .

The projective quadrances of the projective triangle a1 a2 a3 are then

q1 = 2/5

q2 = 1/15

q3 = 4/21

while the projective spreads are

S1 = 3/4

S2 = 1/8
3

S3 = 5/14.

The analog of the hyperbolic Sine law is

3/4
1/8
5/14
=
=
2/5
1/15
4/21
and there are also analogs of both types of hyperbolic Cosine law.
Over the real numbers, to convert this information into the familiar Poincare
model, we intersect the projective point a = [x0 : y0 : 1] with the plane z = 1/2,
yielding the point A = [x0 /2, y0 /2, 1/2] . If a lies inside the null cone x2 +
y 2 z 2 = 0 then A lies in the open disk x2 + y 2 < 1/4, z = 1/2, which is
the equatorial disk of the sphere x2 + y 2 + (z 1/2)2 = 1/4 with north pole
N = [0, 0, 1] . Project A orthogonally onto the lower hemisphere of this sphere,
giving the point
#
"
p
1 x20 y02
x0 y0 1

, ,
A =
2 2 2
2
and then stereographically project A from N to the Poincare disk x2 + y 2 < 1,
z = 0 (viewed as the open unit disk in the complex plane) to get the point




p
p
y0 1 1 x20 y02
x0 1 1 x20 y02
+i
.
zA =
x20 + y02
x20 + y02
Also project the point [x0 , y0 , 1] in the plane z = 1 orthogonally onto the
Poincare disk to get wA = [x0 , y0 , 0]. Then zA = (a) is the corresponding point
in the Poincare model to a, and 0 is the corresponding point to o = [0 : 0 : 1].
If is the usual hyperbolic distance in the Poincare disk then it is well-known
(see for example [1, Chapter 7]) that
p
1 + x20 + y02
1
1
p
(0, zA ) = (0, wA ) = log
2
2
1 x20 + y02
and

sinh2 ( (0, zA )) =

x20 + y02
.
1 x20 y02

Note that the latter is just the negative of q (o, a) .

Now returning to our example, you may use the above formulas to find the
corresponding points in the Poincare model to be

z3 = 2 45 5 + i 1 52 5 .
z1 = 2 3
z2 = 32 21 7 + i 23 + 21 7
The standard formula

(z, w) = log

|1 zw| + |z w|
|1 zw| |z w|

gives (approximately) the hyperbolic distances

1 0.596 455 365

2 0.255 412 812

4

3 0.423 648 930.

The corresponding angles in the hyperbolic triangle z1 z2 z3 may then be calculated using the hyperbolic Cosine Rule
cosh 3 = cosh 1 cosh 2 sinh 1 sinh 2 cos 3
to be, in radians, (approximately)
1 2. 094 395 102 2/3

2 0.361 367 126

3 0.640 522 314.

To check correctness, you can verify (approximately) the hyperbolic Sine Law
sin 2. 094 395 1
sin 0.361 367
sin 0.640 522

1. 36931.
sinh 0.596 455 3
sinh 0.255 412
sinh 0.423 648
To relate the two approaches, the projective quadrance in the projective rational model is the negative of the square of the hyperbolic sine of the hyperbolic distance between the corresponding points in the Poincare model, and the
projective spread is the square of the sine of the angle between corresponding
geodesics in the Poincare model.
The advantages of the projective rational model of the hyperbolic plane
include a cleaner derivation of the theory, simpler and more precise calculations,
with no approximations to transcendental functions required, a more complete
symmetry between rational formulations of the two hyperbolic Cosine laws,
a view of the usual hyperbolic plane as part of a larger picture involving all
projective points, thus accessing the line at infinity (the null cone) as well as
the exterior hyperbolic plane corresponding to the hyperboloid of one sheet,
and the existence of a beautiful duality between projective points and projective
lines that greatly simplifies hyperbolic geometry in two dimensions. And as
stated previously the theory generalizes to higher dimensions, to arbitrary fields,
and to general symmetric bilinear forms, and so unifies elliptic and hyperbolic
trigonometry.
A more complete account of the two dimensional case, with emphasis on
duality, isometries and applications to tesselations, will be given elsewhere. Now
we turn to develop the general affine theory, and after that the projective theory
which generalizes the above situation.

Affine rational trigonometry

Trigonometry studies the measurement of triangles. We work in an n dimensional vector space over a field, not of characteristic two. Elements of the field
are referred to as numbers. Elements of the vector space are called points
or vectors (the two terms will be used interchangeably) and are denoted by
U, V, W and so on. The zero vector or point is denoted O. The unique line l
through distinct points U and V is denoted U V. For a non-zero point U the line
OU is also denoted [U ] . The unique plane through non-collinear points U, V
and W is denoted U V W. The plane OU V is also denoted [U, V ]. The unique
5

affine 3-flat (translate of a three dimensional subspace) through non-planar

points U, V, W and Z is denoted U V W Z, and so on.
Fix a symmetric bilinear form and represent it by
U V.
In terms of this form, the line U V is perpendicular to the line W Z precisely
when
(V U ) (Z W ) = 0.

A point U is a null point or null vector precisely when U U = 0. The origin

O is always a null point, but in general there are others as well. A line U V is
a null line precisely when the vector U V is a null vector. Null points and
lines are themselves well worth studying, but in this paper we concentrate on
non-null points and lines. Some definitions will be empty when applied to null
points or null lines.
For any vector V the number V V will be denoted aV while more generally
for any vectors U and V the number U V will be denoted bUV . Thus bV V = aV .
Given vectors V and U the projection of U onto the line [V ] is
U V
V.
V V

Then a line perpendicular to [V ] in the plane [U, V ] is



U V
U
V .
V V
Motivated by this, we define the intersecting planes [U, W ] and [V, W ] to be
perpendicular precisely when
 


V W
U W
W V
W =0
U
W W
W W
which is the same as
(W W ) (U V ) (U W ) (V W ) = 0

(1)

or
aW bUV bUW bV W = 0.

A set {U, V, W } of three distinct non-collinear points is a triangle and is

denoted U V W . The lines of the triangle U V W are U V, V W and U W. Such a
triangle is non-null precisely when each of its lines is non-null.
The quadrance between the points U and V is the number
Q (U, V ) = (V U ) (V U ) .
Then clearly
Q (U, V ) = Q (V, U ) = V V 2U V + U U
= aV 2bUV + aU .
6

The line U V is a null line precisely when Q (U, V ) = 0, or equivalently when it

is perpendicular to itself.
In ordinary Euclidean geometry distance along a line is additive, assuming
you know what the correct order of points is. With universal geometry, the
important relation between the quadrances formed by three collinear points
is described by a quadratic symmetric function which goes back essentially to
Archimedes discovery of what is usually called Herons formula.
Theorem 1 (Triple quad formula) If U, V and W are collinear then the
quadrances QW = Q (U, V ), QU = Q (V, W ) and QV = Q (U, W ) satisfy


(QU + QV + QW )2 = 2 Q2U + Q2V + Q2W .

Proof. First note that the equation can be rewritten as

2

(QU + QV QW ) = 4QU QV .
Assume U, V and W are collinear and say U and V are distinct so that W U =
(V U ) for some number . In this case
QV = Q (U, W ) = (W U ) (W U ) = 2 QW

QU = Q (V, W ) = (W V ) (W V ) = ( 1) QW .
If U V is a null line then the result is automatic, as both sides are zero, and
otherwise the equation amounts to the identity


( 1) + 2 1

2

= 42 ( 1) .

The next theorem is a restatement and generalization of the most important

theorem in mathematics.
Theorem 2 (Pythagoras theorem) If U, V and W are three distinct points
then U W is perpendicular to V W precisely when the quadrances QW = Q (U, V ),
QU = Q (V, W ) and QV = Q (U, W ) satisfy
QW = QU + QV .
Proof. The condition U W perpendicular to V W means that
(W U ) (W V ) = 0
or
aW bUW bV W + bUV = 0.
The condition QW = QU + QV is
aV 2bUV + aU = (aV 2bV W + aW ) + (aU 2bUW + aW ) .
7

(2)

After a division by two, these conditions are seen to be the same.

In Euclidean geometry, the separation of lines is traditionally measured by
the transcendental notion of angle. The difficulties in defining an angle precisely,
and in extending the concept, are eliminated in rational trigonometry by using
instead the notion of spread in Euclidean geometry the square of the sine of
the angle between two rays lying on those lines. Fortunately we can formulate
the concept completely algebraically.
The spread between the non-null lines U W and V Z is the number
2

s (U W, V Z) = 1

((W U ) (Z V ))
.
Q (U, W ) Q (V, Z)

This depends only on the two lines, not the choice of points lying on them. The
spread between two non-null lines is 1 precisely when they are perpendicular.
If W is a non-null point then the spread between the two planes OU W and
OV W may be defined to be the spread between the lines




UW
V W
UW
and
V W
W W
W W .

The next result is an algebraic generalization of the Sine law in planar

trigonometry.
Theorem 3 (Spread law) Suppose the non-null triangle U V W has quadrances
QW = Q (U, V ), QU = Q (V, W ) and QV = Q (U, W ), and spreads sU =
s (U V, U W ), sV = s (V W, V U ) and sW = s (W U, W V ). Then
sU
sV
sW
=
=
.
QU
QV
QW

Proof. Some straightforward simplification shows that the spread sW = s (W U, W V )

is
(aW bV W bUW + bUV )2
1
(aW 2bUW + aU ) (aW 2bV W + aV )


2
aU aV + aU aW + a
V aW + (bUV + bUW + bV W )
2 b2UV + b2UW + b2V W 2aU bV W 2aV bUW 2aW bUV
=
.
(aW 2bUW + aU ) (aW 2bV W + aV )

(3)

(4)

The numerator in the expression (4) is symmetric in the points U, V and W,

while the denominator is QU QV . It follows that by dividing by QW you get an
expression which is symmetric in the three points.
Thales theorem is the basis of similar triangles in universal geometry.
Theorem 4 (Thales theorem) If the non-null triangle U V W has spread sW =
1 then
QU
sU =
.
QW

Proof. Immediate from the Spread law.

The next result is an algebraic generalization of the Cosine law.
Theorem 5 (Cross law) Suppose the non-null triangle U V W has quadrances
QW = Q (U, V ), QU = Q (V, W ) and QV = Q (U, W ), and spreads sU =
s (U V, U W ), sV = s (V W, V U ) and sW = s (W U, W V ). Then
2

(QU + QV QW ) = 4QU QV (1 sW ) .
Proof. From (2) we have the expression
QU + QV QW = (aV 2bV W + aW ) + (aU 2bUW + aW ) (aV 2bUV + aU )
= 2 (aW + bUV bUW bV W )

while from (3) we have

2

1 sW =

(aW bV W bUW + bUV )

.
QU QV

The result follows.

Note that the Cross law includes as a special case both the Triple quad
formula and Pythagoras theorem. The next result is an algebraic analog of the
fact that the sum of the angles in a triangle is 180 .
Theorem 6 (Triple spread formula) Suppose the non-null triangle U V W
has spreads sU = s (U V, U W ), sV = s (V W, V U ) and sW = s (W U, W V ).
Then


2
(sU + sV + sW ) = 2 s2U + s2V + s2W + 4sU sV sW .
Proof. Assume the quadrances of U V W are QW = Q (U, V ), QU = Q (V, W )
and QV = Q (U, W ). Write the Spread law as
sV
sW
1
sU
=
=
=
QU
QV
QW
D
for some non-zero number D. Now substitute QU = DsU , QV = DsV and
QW = DsW into the Cross law, and cancel a factor of D2 , yielding
2

(sU + sV sW ) = 4sU sV (1 sW ) .
Rearrange this to get


2
(sU + sV + sW ) = 2 s2U + s2V + s2W + 4sU sV sW .
The Triple spread formula can be reinterpreted as a statement about three
non-parallel coplanar lines. If three lines lie in a (two dimensional) plane then
the spread between any two of them is unaffected if one or more of the lines
9

is translated. In particular we can arrange that the three lines are concurrent,
and so the Triple spread formula still applies.
Another useful observation is that if say SW = 1, then the Triple spread
formula becomes
2
(sU + sV 1) = 0
so that
sU + sV = 1.
Secondary results of planar rational trigonometry, some developed in [5], are
consequences of the main laws of this section, and so still hold in this general
setting.

An affine example over the rational numbers

Here is an example of trigonometry in four dimensional space over the rational
numbers (the most important field) with bilinear form
U V = UMV T
where

0
1
M =
0
3

1
1
2
1

0
2
1
0

Consider the triangle U V W where

U = [1, 2, 4, 3/2]

3
1
.
0
1

V = [1, 0, 1/2, 3]

W = [2, 2, 1, 5] .

Then the quadrances are

QU =

177
4

QV =

71
4

QW = 38

and the spreads are

sU =

10 263
10 792

sV =

3421
8968

sW =

3421
4189 .

Then you may verify the Spread law

10 263/10 792
3421/8968
3421/4189
3421
=
=
=
177/4
71/4
38
159182
(one of the forms of) the Cross law


177 71
+
38
4
4

2



3421
177 71
.

1
=4
4
4
4189

10

The Triple spread law becomes

2
10 263 3421 3421
29 258 102 500
=
+
+
10 792 8968 4189
6334 727 281

2 
2 
2 !
10 263
10 263 3421 3421
3421
3421
=2
+4
+
+

.
10 792
8968
4189
10 792 8968 4189

Geometry in such a setting has many familiar features. Here are the circumcenter C and circumquadrance K of the triangle


144 3789 18 773 46 709
C=
,
,
,
311 3421 13 684 13 684
79 591
.
K=
6842
The orthocenter of the triangle is


334 6106 9429 9145
,
,
,
O=
311 3421 3421 3421
the centroid is

2 4 11 19
G=
, , ,
3 3 6 6

and the nine-point center is



239 9895 56 489 83 289
N=
.
,
,
,
311 6842 27 368 27 368
All four points C, G, N and O are collinear, lying on the Euler line of the
triangle, and as expected
2
C+
3
1
N = C+
2
G=

1
O
3
1
O.
2

Many other aspects of Euclidean geometry may be explored.

Projective rational trigonometry

Fix an (n + 1) dimensional vector space over a field with a symmetric bilinear
form U V as in the previous section. A line through the origin O will now be
called a projective point and denoted by a small letter such as u. The space
of such projective points is called n dimensional projective space. This is
a natural domain for algebraic geometry, and the metrical structure we will
introduce gives new directions for this subject.

11

If V is a non-zero vector (or point) in the vector space, then v = [V ] will

denote the projective point OV. A projective point is a null projective point
precisely when some non-zero null point lies on it. Two projective points u = [U ]
and v = [V ] are perpendicular precisely when they are perpendicular as lines.
This is equivalent to the condition U V = 0.
A plane through the origin (two dimensional subspace) will be called a projective line and denoted by a capital letter such as L. A three dimensional
subspace will be called a projective plane. If V and W are independent vectors then L = [V, W ] will denote the projective line OV W. If V, W and Z are
independent vectors then = [V, W, Z] will denote the projective plane OV W Z.
A projective point u lies on a projective line L (or equivalently L passes
through u) precisely when the line u lies on the plane L. Similar terminology applies to projective points lying on projective planes, or projective lines
lying on projective planes etc. There is a unique projective line L = uv which
passes through any two distinct projective points u and v. Three or more projective points which lie on a single projective line are collinear. Three or more
projective lines which all pass through a single projective point are concurrent.
Two intersecting projective lines uw and vw are perpendicular precisely
when they are perpendicular as intersecting planes. If u = [U ], v = [V ] and
w = [W ] then this is equivalent to the condition (1)
(W W ) (U V ) (U W ) (V W ) = 0.
A set {u, v, w} of three distinct non-collinear projective points is a projective triangle, and is denoted uvw. The projective triangle uvw is null if one
or more of its projective points is null. The projective lines of the projective
triangle uvw are uv, vw and uw.
The rich metrical structure of projective space arises with the correct notion
of the separation of two projective points u and v. Since each is a line through
the origin in the ambient vector space, we may apply the notion of spread
between these two lines, as developed in the previous section.
The projective quadrance between the non-null projective points u = [U ]
and v = [V ] is the number
q (u, v) = 1

(U V )2
(U U ) (V V )

This is the same as the spread s (OU, OV ), and has the value 1 precisely when
the projective points are perpendicular. In terms of the as and bs of the
previous section
aU aV b2UV
.
q (u, v) =
aU aV
Note the use of the small letter q for a projective quadrance to suggest that it
is really a spread, and to distinguish it from a quadrance Q.
Theorem 7 (Projective triple quad formula) If u = [U ] , v = [V ] and
w = [W ] are collinear projective points then the projective quadrances qw =
12

q (u, v), qu = q (v, w) and qv = q (u, w) satisfy

2
+ 4qu qv qw .
(qu + qv + qw )2 = 2 qu2 + qv2 + qw

Proof. If u = [U ] , v = [V ] and w = [W ] are collinear then the three vectors

U, V and W are dependent and the lines OU, OV and OW are coplanar. Since
the projective quadrance between projective points is just the spread between
these lines, the theorem is an immediate consequence of the (affine) Triple spread
formula of the previous section.
Here is the projective version of the most important theorem in mathematics.
Not surprisingly, it has consequences in many directions.
Theorem 8 (Projective Pythagoras theorem) Suppose that u, v and w
are three distinct non-null projective points and that uw is perpendicular to vw.
Then the projective quadrances qw = q (u, v), qu = q (v, w) and qv = q (u, w)
satisfy
qw = qu + qv qu qv .
Proof. If u = [U ], v = [V ] and w = [W ] for some vectors U, V and W, then
qw qu qv + qu qv

aV aU b2UV
aV aW b2V W
aW aU b2UW

aU aV
aV aW
aU aW




2
2
aV aW bV W aW aU bUW
+
aV aW
aU aW
(aW bUV + bUW bV W ) (bUW bV W aW bUV )
=
.
a2W aV aU
=

But we have seen above that the condition that uw is perpendicular to vw is

equivalent to
aW bUV bUW bV W = 0.
Thus this implies that
qw qu qv + qu qv = 0.
Note that the converse does not in general follow.
The projective spread between the intersecting projective lines wu =
[W, U ] and wv = [W, V ] is defined to be the spread between these intersecting planes, namely the number
S (wu, wv)
=1

UW
W W

W U

UW
W W W


UW
W W W

13

2
V W
W W W

V W
W
W W

V
V

V W
W W

.

It is undefined if one of the denominators is zero, and is 1 precisely when the

two projective lines are perpendicular. In terms of the as and bs
2

S (wu, wv) = 1

(aW bUV bUW bV W )

.
(aU aW b2UW ) (aV aW b2V W )

(5)

The projective form of the spread law has the same form as the affine one.
Theorem 9 (Projective spread law) Suppose the non-null projective triangle uvw has projective quadrances qu = q (v, w), qv = q (u, w) and qw = q (u, v),
and projective spreads Su = S (uv, uw), Sv = S (vw, vu) and Sw = S (wu, wv).
Then
Su
Sv
Sw
=
=
.
qu
qv
qw
Proof. Assume that u = [U ] , v = [V ] and w = [W ]. After expansion and
simplification,


aW aU aW aV aV b2UW aU b2V W aW b2UV + 2bUW bUV bV W
S (wu, wv) =
(aU aW b2UW ) (aV aW b2V W )
(6)
Together with
aV aU b2UV
qw = q (u, v) =
aU aV
(6) shows that the quotient Sw /qw is actually symmetric in the three variables
U, V and W.
The next result is a simple but surprising consequence of the Projective
spread law. Even in the simple context of two dimensional elliptic trigonometry,
it reveals that there is an aspect of similar triangles in spherical geometry. This
interesting point helps explain why the spread ratio (opposite quadrance over
hypotenuse quadrance) is so important in rational trigonometry.
Theorem 10 (Projective Thales theorem) If the projective triangle uvw
has projective spread Sw = 1 then
Su =

qu
.
qw

Proof. This follows from the previous theorem.

Here is the projective version of the Cross law. Unlike the affine result, it is
a quadratic equation in the projective spread Sw .
Theorem 11 (Projective cross law) Suppose the non-null projective triangle uvw has projective quadrances qu = q (v, w), qv = q (u, w) and qw = q (u, v),
and projective spreads Su = S (uv, uw), Sv = S (vw, vu) and Sw = S (wu, wv).
Then
(Sw qu qv qu qv qw + 2)2 = 4 (1 qu ) (1 qv ) (1 qw ) .
14

Proof. If u = [U ], v = [V ] and w = [W ] for some vectors U, V and W, then

b2UV b2V W b2UW
aU aV aV aW aU aW
4b2 b2UW b2UV
= VW
.
a2W a2U a2V

4 (1 qu ) (1 qv ) (1 qw ) = 4

Also
aW aU aW aV aV b2UW aU b2V W aW b2UV + 2bUW bUV bV W
Sw qu qv =
(aU aW b2UW ) (aV aW b2V W )




aV aW b2V W aU aW b2UW

aV aW
aU aW
2
aU aW aV aV bUW aU b2V W aW b2UV + 2bUW bUV bV W
=
.
aU aV aW

Thus
Sw qu qv qu qv qw + 2

aU aW aV aV b2UW aU b2V W aW b2UV + 2bUW bUV bV W

aU aV aW
2
aV aW b V W
aU aW b2UW
aV aU b2UV

+2
aV aW
aU aW
aV aU
2bV W bUW bUV
=
.
aW aU aV
=

The result follows.

Note that the projective quadrea A =Sw qu qv = Su qv qw = Sv qu qw is
symmetric in U, V and W, also because of the Projective spread law.
Theorem 12 (Dual projective cross law) Suppose the non-null projective
triangle uvw has projective quadrances qu = q (v, w), qv = q (u, w) and qw =
q (u, v), and projective spreads Su = S (uv, uw), Sv = S (vw, vu) and Sw =
S (wu, wv). Then
2

(qw Su Sv Su Sv Sw + 2) = 4 (1 Su ) (1 Sv ) (1 Sw ) .
Proof. If Cu = 1 Su , Cv = 1 Sv and Cw = 1 Sw then the required identity
can be rewritten as
2

(qw (1 Cu ) (1 Cv ) + Cu + Cv + Cw 1) = 4Cu Cv Cw .

(7)

If u = [U ], v = [V ] and w = [W ] for some vectors U, V and W, then using

(5) the left hand side of (7) is the square of an expression which simplifies in a
pleasant fashion to
2 (aW bUV bUW bV W ) (aV bUW bUV bV W ) (aU bV W bUV bUW )
.
(aV aW b2V W ) (aU aW b2UW ) (aU aV b2UV )
The square of this is exactly the right hand side of (7).
15

Theorem 13 (Dual projective Pythagoras theorem) If qw = 1 in the

notation of the previous proof, then
Sw = Su + Sv Su Sv .
Proof. This follows from the previous result and the polynomial identity
(Su Sv Su Sv Sw + 2)2 4 (1 Su ) (1 Sv ) (1 Sw )
2

= (Sw Su Sv + Su Sv ) .
The proof of the next result utilized a computer, although it could be checked
by hand.
Theorem 14 (Projective triple spread formula) Suppose that u, v, w and
z are coplanar projective points with projective spreads Ru = S (zv, zw), Rv =
S (zu, zw) and Rw = S (zu, zv). Then


2
2
+ 4Ru Rv Rw .
(Ru + Rv + Rw ) = 2 Ru2 + Rv2 + Rw

Proof. (Using a computer) Suppose that u = [U ], v = [V ], w = [W ] and

z = [Z] for vectors U, V, W and Z. Then


aZ aV aW aZ aZ b2V W aV b2W Z aW b2V Z + 2bV W bV Z bW Z
Ru =
(aV aZ b2V Z ) (aZ aW b2W Z )


aZ aU aW aZ aZ b2UW aU b2W Z aW b2UZ + 2bUW bUZ bW Z
Rv =
(aU aZ b2UZ ) (aZ aW b2W Z )


aZ aU aV aZ aZ b2UV aU b2V Z aV b2UZ + 2bUV bUZ bV Z
Rw =
(aU aZ b2UZ ) (aZ aV b2V Z )
A computer calculation shows that


2
2
(Ru + Rv + Rw ) 2 Ru2 + Rv2 + Rw
4Ru Rv Rw

has a factor which is the determinant


 aU
bUV

 bUV
aV

bUW bV W

 bUZ bV Z

bUW
bV W
aW
bW Z


bUZ 
bV Z 
.
bW Z 
aZ 

But if u, v, w and z are coplanar then U, V, W and Z are linearly dependent, so

this determinant is zero.
It is worth pointing out that for the case of a non-degenerate bilinear form
and n = 3 there is a duality between projective points and projective lines,
so the previous three theorems can be deduced from the corresponding earlier
results. However in general this duality is not available.
16

Special projective triangles

The following theorem gives universal analogs of Napiers rules in spherical
trigonometry. The various formulas in this proof are fundamental for projective
trigonometry, and they can all be derived easily enough from the basic equations
of the theorem.
Theorem 15 (Napiers rules) Suppose the projective triangle uvw has projective quadrances qu = q (v, w), qv = q (u, w) and qw = q (u, v) and projective
spreads Su = S (uv, uw), Sv = S (vu, vw) and Sw = S (wu, wv). If Sw = 1 then
any two of the five quantities [qu , qv , qw , Su , Sv ] determine the other three, solely
through the three basic equations
qw = qu + qv qu qv

Su = qu /qw

Sv = qv /qw .

Proof. Two of the projective quadrances allow you to determine the third via
the Projective Pythagoras theorem qw = qu + qv qu qv , and then the other two
Projective Thales equations Su = qu /qw and Sv = qv /qw give the projective
spreads.
Given the two projective spreads Su and Sv , use the Projective Pythagoras
theorem and the Thales equations Su = qu /qw and Sv = qv /qw to obtain
1 = Su + Sv Su Sv qw .
Thus
Su + Sv 1
Sv
Su + Sv 1
qv = Sv qw =
Su

qu = Su qw =

and
qw =

Su + Sv 1
.
Su Sv

If you know a projective spread, say Su , and a projective quadrance, then there
are three possibilities. If the projective quadrance is qw , then qu = Su qw ,
qv =

qw qu
qw Su qw
=
1 qu
1 Su qw

and
Sv =

qv
1 Su
=
.
qw
1 Su qw

If the projective quadrance is qu , then qw = qu /Su ,

qv =

qu (1 Su )
qw qu
=
1 qu
Su (1 qu )

17

and
Sv =

qv
1 Su
=
.
qw
1 qu

If the projective quadrance is qv , then substitute qu = Su qw into the Projective

Pythagoras equation to get
qw = Su qw + qv Su qv qw .
So
qv
1 Su (1 qv )
Su qv
qu =
1 Su (1 qv )

qw =

and
Sv =

qv
= 1 Su (1 qv ) .
qw

A projective triangle is isosceles precisely when at least two of its projective

quadrances are equal.
Theorem 16 (Pons Asinorum) Suppose a non-null projective triangle uvw
has projective quadrances qu , qv and qw , and projective spreads Su , Sv and Sw .
Then qu = qv precisely when Su = Sv .
Proof. Follows immediately from the Projective spread law.
It follows from the Projective Pythagoras theorem that if Su = Sv = 1 then
qw = 0 or qu = qv = 1.
Theorem 17 (Isosceles projective triangle) Suppose an isosceles projective
triangle has non-zero projective quadrances qu = qv = q and qw , and projective
spreads Su = Sv = S and Sw . Then
qw =
Sw =

4q (1 S) (1 q)

(1 Sq)2
4S (1 S) (1 q)
2

(1 Sq)

Proof. Use the Projective spread law in the form

Sw =

Sqw
q

to replace Sw in the Projective cross law

2
2
q 2 Sw (2q + qw 2) = 4 (1 q) (1 qw ) .
18

This yields a quadratic equation in qw with solutions qw = 0, which is impossible

by assumption, and
4q (1 S) (1 q)
qw =
.
2
(1 Sq)
Thus

Sw =

Sqw
4S (1 S) (1 q)
.
=
2
q
(1 Sq)

A projective triangle is equilateral precisely when all its quadrances are

equal. The following formula appeared in the Euclidean spherical case as Exercise 24.1 in [5].
Theorem 18 (Equilateral projective triangles) Suppose that a projective
triangle is equilateral with common non-zero projective quadrance q1 = q2 =
q3 = q, and with common projective spread S1 = S2 = S3 = S. Then
2

(1 Sq) = 4 (1 S) (1 q) .
Proof. From the Isosceles projective triangle theorem
q=

4q (1 S) (1 q)
(1 Sq)2

Since q 6= 0 this yields

2

(1 Sq) = 4 (1 S) (1 q) .
The above result is symmetric in S and q. Note that if S = 3/4 then q =
8/9. This value is important in chemistryit is the tetrahedral spread found
for example in the methane molecule, and corresponds to an angle which is
approximately 109. 47. As I will show elsewhere, rational trigonometry provides
a much more refined analysis of the geometry of the Platonic solids, but some
basic results in this direction can be found in [5].

Spread polynomials
We have seen that both affine and projective trigonometry involve the Triple
spread formula


2
(a + b + c) = 2 a2 + b2 + c2 + 4abc.

If a = b = s then c turns out to be either 0 or 4s (1 s). If a = 4s (1 s) and

b = s then c turns out to be either s or s (3 4s)2 . There is then a sequence
of polynomials Sn (s) for n = 0, 1, 2, with the property that Sn1 (s) , s
and Sn (s) always satisfy the Triple spread formula. They play a role in all
metrical geometries, independent of the nature of the symmetric bilinear form,
19

and are defined over the integers. These are rational analogs of the Chebyshev
polynomials of the first kind, and they have many remarkable properties.
The spread polynomial Sn (s) is defined recursively by S0 (s) = 0, S1 (s) =
s and the rule
Sn (s) = 2 (1 2s) Sn1 (s) Sn2 (s) + 2s.
The coefficient of sn in Sn (s) is a power of four, so the degree of the polynomial Sn (s) is n in any field of characteristic not two. It turns out that in the
decimal number field
1 Tn (1 2s)
Sn (s) =
2
where Tn is the n-th Chebyshev polynomial of the first kind. The first few
spread polynomials are S0 (s) = 0, S1 (s) = s, S2 (s) = 4s 4s2 , S3 (s) =
9s 24s2 + 16s3 , S4 (s) = 16s 80s2 + 128s3 64s4 and S5 (s) = 25s 200s2 +
560s3 640s4 + 256s5 . Note that S2 (s) is the logistic map.
As shown in [5], Sn Sm = Snm for n, m 1, and the spread polynomials
have interesting orthogonality properties over finite fields. S. Goh [2] observed
that there is a sequence of spread-cyclotomic polynomials k (s) of degree (k)
with integer coefficients such that for any n = 1, 2, 3,
Y
k (s) .
Sn (s) =
k|n

A projective example over F11

Consider a five dimensional vector space over the field F11 with bilinear form
U V = UMV T
where

1 10 1
10 2 5

M =
1 5 1
0 2 4
0 0 3

0
2
4
7
2

0
0

3
.
2
8

The geometry of the associated four dimensional projective space is pleasantly

accessible. The use of a finite field simplifies calculations and provides an ideal
laboratory for geometrical explorations, even if you are interested in other fields.
Consider the triangle uvw where
u = [1 : 4 : 2 : 6 : 1]

v = [1 : 2 : 3 : 4 : 1]

w = [0 : 8 : 8 : 3 : 1] .

The projective quadrances are then qu = 9, qv = 8 and qw = 1, and the

projective spreads are Su = 2, Sv = 3 and Sw = 10. Since qw = 1, uvw is a
dual right triangle, so that Sw = Su + Sv Su Sv .
20

The Projective spread law is verified to be

2
3
10
= =
9
8
1
while the Projective cross law takes the form
2

(10 9 8 9 8 1 + 2) = 0 = 4 (1 9) (1 8) (1 1)
and the Dual projective cross law takes the form
2

(1 2 3 2 3 10 + 2) = 5 = 4 (1 2) (1 3) (1 10) .
Note that the squares in F11 are 0, 1, 3, 4, 5, 9, so Sv is the only projective spread
of the triangle uvw which is a square. As in the discussion in [5], this implies
that only the vertex at v has a bisector, and there are two such. The points
b1 = [3 : 0 : 5 : 8 : 7] and b2 = [3 : 2 : 7 : 6 : 10] lie on uw and
S (vu, vb1 ) = S (vw, vb1 ) = 10

and

S (vu, vb2 ) = S (vw, vb2 ) = 2.

While bisectors and medians are dependent on number theoretic considerations,

the orthocenter of the projective triangle turns out not to be. You can check
that here it is
O = [9 : 1 : 0 : 4 : 1] .

Lambert quadrilaterals
Here are two (of many) results from hyperbolic geometry (see [1, Chapter 7])
that hold more generally.
Theorem 19 (Lambert quadrilateral) Suppose the projective points u, v, w
and z are coplanar and form projective spreads
S (uv, uz) = S (vu, vw) = S (wv, wz) = 1
and projective quadrances q (u, v) = q and q (v, w) = p. Then
q (w, z) = y = q (1 p) / (1 qp)
q (u, w) = s = q + p qp

q (u, z) = x = p (1 q) / (1 qp)
q (v, z) = r = (q + p 2qp) / (1 qp)

and
S (vu, vz) = x/r
S (uw, uv) = p/s

S (vw, vz) = y/r

S (uw, uz) = q (1 p) /s

and
S (zu, zw) = S = 1 pq.

21

S (wv, wu) = q/s

S (wu, wz) = p (1 q) /s

Proof. The fact that the four points are coplanar implies that the Projective
triple spread theorem applies to any three projective lines of the projective
quadrilateral uvwz meeting at a projective point. Furthermore it implies that
where three projective lines meet and one of the spreads is 1, the other two
spreads must sum to 1.
The expressions for S (vu, vz), S (vw, vz), S (wv, wu) and S (uw, uv) follow
from the Projective Thales theorem. The expression for s follows from the
Projective Pythagoras theorem applied to uvw. The same theorem applied to
uvz and vwz gives the equations
r = q + x qx
r = p + y py

and since S (vu, vw) = 1

x y
+ = 1.
r
r
These three equations can then be solved to yield the stated values for r, x and
y. Also the equations
S (uw, uv) + S (uw, uz) = 1 = S (wv, wu) + S (wu, wz)
can be used to solve for S (uw, uz) and S (wu, wz). Finally use the Projective
spread law in uwz to get
S (zu, zw) = S = 1 pq.

Theorem 20 (Right hexagon) Suppose a projective hexagon a1 a2 a3 a4 a5 a6

is planar, meaning that all the projective points lie in some fixed projective
plane, and that all successive projective spreads are equal to 1, that is
S (a1 a2 , a1 a6 ) = S (a2 a3 , a2 a1 ) = = S (a6 a1 , a6 a5 ) = 1.
Then

q (a1 , a2 )
q (a2 , a3 )
q (a3 , a4 )
=
=
.
q (a4 , a5 )
q (a5 , a6 )
q (a1 , a6 )

Proof. Follows by repeated use of the last formula from the previous theorem.

References
[1] A F Beardon, The Geometry of Discrete Groups, GTM 91, Springer-Verlag,
New York, 1983.
[2] S. Goh, Chebyshev polynomials and spread polynomials, Honours Thesis,
School of Mathematics, UNSW, 2005.
22

[3] M. J. Greenberg, Euclidean and Non-Euclidean Geometries: Development

and History, W. H. Freeman and Co., San Francisco, 1972.
[4] S. Katok, Fuchsian Groups, Chicago Lectures in Mathematics, University of
Chicago Press, Chicago, 1992.
[5] N. J. Wildberger, Divine Proportions: Rational Trigonometry to Universal
Geometry, Wild Egg Books (http://wildegg.com), Sydney, 2005.

23