You are on page 1of 28

Dorst Chapter 14, New Primitives for Euclidean Geometry (15/4/2020)

14.1 Rounds
14.1.1 Dual Rounds
Given a general, not necessarily null, vector x in R3 +1,1,

( ) ( ) ( ( ) ( )
2 2
1 1 1 x 1 x 1 x
x=αe + x + β e=α o− ∞ + x + β o + ∞ =( α+ β ) o+ x + ( β−α ) ∞=( α + β ) o+ + ∞− ∞+
2 2 2 α +β 2 α+β 2 α+ β

where y=o+
x
+
1 x 2
α+β 2 α+ β ( )
∞ is a null vector.

Now consider any vector x of the form y +α ∞ where y is a unit weight null vector. We know
that every unit weight vector can be put in such a form. Suppose z is a null vector and z ∙ x=0 .
Then

( 1 2
)(
0=z ∙ x= o+ z + z ∞ ∙ o+ y +
2
y2
2 (
+α ∞ =
− y2
2 )) 1 2 −1
−α + z ∙ y − z =
2 2
2
( z − y ) −α

So ( z− y )2=−2 α . So x dually represent a sphere whose center is at y and whose radius squared
is −2 α . If α =0 so x is a null point, then x dually represent the sphere which is just the point y
2
−y
with zero radius. If α = , so x=o + y . Then x dually represent the sphere centered at y whose
2
radius squared is y 2. If the location of y is the origin, so x=o +α ∞, then x dually represent a
sphere whose center is at o and whose radius squared is −2 α , namely the sphere with the
−1 2
equation z 2=−2 α . This is the sphere Dorst introduces here, where he takes α = ρ so the
2
y 1
radius is ρ . So both o+ ∧e=o− ∞ dually represent spheres with radius one, the first
| y| 2
y 1
centered at and the second at the origin. It is curious that e=o+ ∞ dually represents an
| y| 2
imaginary sphere centered at the origin whose radius squared is −1.
Note that there is a natural bijection between the Euclidean vectors in R3 and the unit weight null
1 2
2 (
1 2
)
vectors in R3 +1,1, y ⟺ o+ y + y ∞. The null vector α o+ y + y ∞ where α ≠ 1 does not
2
correspond to any Euclidean vector but it does represent a point whose location is y . The vector
1 2
αo + y + y ∞ where α ≠ 1, is not null and does not corresponds to any Euclidean vector and does
2
y
not represent a point but (if α ≠ 0 ) dually a sphere centered at and whose radius squared is
α

()
2
( 1−α ) y . There is also the Euclidean point space, which is not a vector space, which R3 +1,1
α
1 2
( )
represents. The points represented by αy=α o + y + y ∞ and βy where α ≠ β , differ in weight
2
but not in location, which is y .
1 2
Let σ =o+ y+ y ∞ +α ∞ be the dual representation of a sphere centered at y and radius squared
2
-2α . Let Σ be its direct representation, or Σ=σe e1 e 2 e 3 e. Now

( 1
) 1
( )
I 5=e e1 e 2 e 3 e= o− ∞ e1 e 2 e 3 o+ ∞ =∞ ∧ o ∧e 1 ∧ e2 ∧e 3=o ∧e 1 ∧ e2 ∧e 3 ∧ ∞
2 2
Note the difference between 5D and 3D with respect to reverse,
−1 ~ −1 ~
I 5 =− I 5=−I 5 , I 3 = I 3=−I 3

( 1 2
)
Σ=σe e1 e 2 e 3 e= o+ y 1 e1 + y 2 e 2+ y 3 e 3+ y ∞ +α ∞ ∞ ∧ o ∧e 1 ∧ e 2 ∧ e3=−o e 1 e 2 e3 + y 1 ( ∞ ∧ o ) e 2 e 3 + y 2 ( ∞ ∧ o ) e3 e
2

I checked that indeed −Σe e1 e 2 e 3 e=−Σ ( ∞ ∧ o ∧e 1 ∧ e2 ∧e 3 )=σ . While the direct representation
~
of a plane with unit normal n and distance δ from the origin is Π= (−δ−o n ) I 3 ∞ =( δ + o n ) I 3 ∞
and its dual representation is n+ δ ∞ . Note that the direct conformal representations of spheres
and planes are blades since the dual of a blade is a blade since the contraction of blades is a
blade. The dual representation of a plane through the origin with unit normal n is just n , while
the dual representation of a sphere centered at the origin with radius squared −2 α is
1 2
o+ α ∞ ≡o− ρ ∞, so the radius is ρ . The direct representation of a plane through the origin is
2
the 4-blade o ∧ ¿ and the direct representation of a sphere centered at the origin and radius
~ ~
squares −2 α is the 4-blade ( o−α ∞ ) ∧ I 3=o I 3 +α I 3 ∞.
The intersection of the plane and the sphere is a great circle. In Figure 14.1a, this is the equator
of the north and south poles. Each point on the sphere can be considered as a pole, and infinitely
many great circles go through it, each one corresponding to a plane which intersects the sphere
and goes through the sphere’s center. The unique unit normal to the sphere at that pole-point, is
perpendicular to every tangent to the sphere at that pole-point and lies in every plane which
intersect the sphere through its center and contains that pole-point. So in that sense one can say
that any plane which goes through the center of the sphere, is perpendicular to the sphere
because it contains all the normals to the sphere at the points where the plane intersects the
1 2
surface of the sphere. This observation can be related to the fact that n ∙ o− ρ ∞ =0. By
2 ( )
covariance Dorst claims that this zero inner product will be true for any sphere and plane rotated
and translated away from the origin, as long as the plane goes through the center of the sphere. If
we check that for the case where the center of the sphere is at y and the plane goes through the
center, than

( y − 12 ρ ∞ )∙ ( n+ δ ∞ )=(o+ y+ 12 y ∞− 12 ρ ∞) ∙ ( n+δ ∞ )=−δ + y ∙ n=0


2 2 2

Let’s calculate the inner product using the direct representations

[
Σ∙ Π= −o+ ( ∞∧ o ) y + ( 12 y +α ) ∞ ] e e e ∙ [ (−δ −o n ) e e e ∞ ]=−δ [−o+( ∞ ∧o ) y +( 12 y +
2
1 2 3 1 2 3
2

Now
( δo e 1 e 2 e3 ) ∙ ( e 1 e 2 e3 ∞ )=δ
( ∞ ∧o ) ( y 1 e 2 e 3 ) ∙ ( e1 e 2 e 3 ∞ ) =( ∞ ∧o ) ( y 2 e 3 e1 ) ∙ ( e1 e 2 e 3 ∞) =( ∞ ∧o ) ( y 3 e 1 e 2 ) ∙ ( e1 e 2 e 3 ∞ ) =0

( ∞ e 1 e 2 e3 ) ∙ ( e 1 e 2 e3 ∞ )=0
( o e1 e 2 e 3 ) ∙ ( n1 o e2 e3 ∞ ) =( o e 1 e2 e 3 ) ∙ ( n2 o e 3 e1 ∞ )=( o e 1 e 2 e3 ) ∙ ( n3 o e 1 e 2 ∞ )=0
−( ∞ ∧o ) ( y1 e 2 e 3 ) ∙ ( n1 o e2 e 3 ∞ ) =− y 1 n1
( ∞ ∧o ) ( y 1 e 2 e 3 ) ∙ ( n2 o e3 e 1 ∞ ) =( ∞ ∧ o ) ( y 1 e 2 e 3 ) ∙ ( n 3 o e1 e 2 ∞ ) =0
( ∞ ∧o ) ( y 2 e 3 e1 ) ∙ ( n1 o e2 e 3 ∞ ) =( ∞ ∧ o ) ( y 2 e 3 e 1 ) ∙ ( n 3 o e1 e 2 ∞ ) =0
−( ∞ ∧o ) ( y2 e 3 e1 ) ∙ ( n2 o e3 e 1 ∞ ) =− y 2 n2
( ∞ ∧o ) ( y 3 e 1 e 2 ) ∙ ( n1 o e2 e 3 ∞ ) =( ∞ ∧ o ) ( y 3 e 1 e 2 ) ∙ ( n 2 o e3 e 1 ∞ ) =0
−( ∞ ∧o ) ( y3 e 1 e2 ) ∙ ( n3 o e1 e 2 ∞ ) =− y 3 n3

( ∞ e 1 e 2 e3 ) ∙ ( n1 o e 2 e 3 ∞ )=( ∞e 1 e2 e3 ) ∙ ( n2 o e 3 e1 ∞ )=( ∞e 1 e2 e 3 ) ∙ ( n3 o e 1 e2 ∞ )=0


Altogether we have δ− y 1 n1− y 2 n2− y 3 n3=δ − y ∙ n=0.
Let Σ=σ I 5 , Π =π I 5 be the direct representation of the sphere and the plane whose dual
representation is σ ∧π . Since the contraction of blades is a blade and the dual of a blade is a
blade, Σ∧Π are 4-blades. Then σ ∧ π =¿. Note that the join of Σ∧Π , namely the smallest blade
which contains the sphere’s blade and the plane’s blade, is a pseudoscalar. Now, if we join
everything outside of Σ to everything outside of Π then we get everything outside the
intersection of Σ∧Π , which is the dual of the intersection, namely π ∧ σ =( Π I 5 ) ∧ ( Σ I 5 ) =¿,
−1 −1

see Dorst page 129 equations (5.7-8). Note the reversal of the factors. Now the intersection
Σ∨ Π is the direct representation of the intersection which is a great circle, and its dual is a dual
representation of the great circle. Now by definition of Dorst page 129 equation (5.6), Σ∨ Π=¿ ,
which is the intersection, so the dual representation is π ∧ σ =¿. Recall that I −1
5 =−I 5 and
~
I 5= I 5=∞ ∧o ∧ e1 ∧e 2 ∧ e3 .

5 =[ ( −δ−o n ) e1 e2 e 3 ∞ ] (−∞ ∧ o∧ e 1 ∧ e2 ∧e 3 )=δ e1 e 2 e 3 ∞ ( ∞∧ o ∧e 1 ∧ e2 ∧e 3 ) +n1 o e 2 e3 ∞ ( ∞ ∧ o ∧e 1 ∧ e2 ∧e 3


Π I −1
¿
−¿
¿
¿
¿

δ ( 12 y +α ) ¿
2

−¿
−¿
−¿
¿
¿
¿
¿
¿
¿
¿
¿
¿

( 12 y +α ) n ¿
2
1

( 12 y +α ) n ¿
2
2
( 1 2
2 )
y +α n3 ¿

Altogether we have
¿
¿
−δ ¿
¿
¿
−¿

( 12 y +α ) ¿
2

To check, calculate the 2-blade π ∧ σ directly

(
( n+ δ ∞ ) ∧ ( y+ α ∞ )= ( n+δ ∞ ) ∧ o+ y+ ( 12 y +α ) ∞)=n o+ n∧ y +( 12 y +α ) n ∞+δ ∞∧ o+δ ∞ y
2 2

When the sphere is centered at the origin, then the direct representation of the great circle
centered at the origin and radius squared −2 α , made by the intersection of a plane through the
origin with unit normal n with the sphere, is the 3-blade
Σ∨ Π=¿
while the dual representation of that great circle is
σ ∧ π =( o+ α ∞ ) ∧ n
Actually note that σ ∧ π is the dual representation of Π ∨ Σ, while π ∧ σ is the dual
representation of Σ∨ Π . Just like σ ∧ π =−π ∧ σ so Π ∨ Σ=(−1 )( 5−4) (5 −4) Σ ∨ Π =−Σ∨ Π (see
section 5.6 page 133). That means that ¿.
~
Note that according to the above ( o+ α ∞ ) ∧n=n I 3 ( o+ α ∞ ) I −1
5 . It’s a not a trivial computation to

check that. We already did the checking in the general, center away from the origin, case.
1 2
Suppose a point x=o + x+ x ∞ is on the great circle, so ¿.
2
0=¿
1 2
So α + x =0 or x 2=−2 α . This shows that the point is on the sphere. Also x ∙ n=0. This shows
2
that the point is in the plane. Together they show that the point is on the great circle of
intersection.
Intersect the sphere with another plane π ' =n ' through the center of a sphere centered at the
origin, such that n ∙ n' =0 . There are infinitely many such planes. See one in Figure 14.1(c). The
intersection of two planes and the sphere is the intersection of two great circles, which is a pair
of antipodal points on the two great circles. If we look at the earth as a sphere, than the
intersection of the equator plane with a plane which goes through the north and south poles, and
the interior of the sphere, is a diameter between two antipodal points on the equator, and these
antipodal points are the intersection of the two planes and the sphere, as well as the intersection
of two orthogonal great circles, like the equator with any longitude great circle through the north
and south poles. In Figure 14.1(c) the two points are right most and left most points on the
equator.
What is the direct representation of this double intersection?
( Π ' ∨ Π ) ∨ Σ=¿
which is a 2-blade (4 + ( 4 + 4−5 )−5=2). It’s strange that the two points, which is the double
intersection, is a 2-blade. The dual of this direct is a 3-blade, just omit the I 3
[ ( Π ' ∨ Π ) ∨ Σ ] I−1
5 =Σ I 5 ∧ ( Π ∨ Π ) I 5 =Σ I 5 ∧ ( Π I 5 ∧ Π I 5 ) =σ ∧ π ∧ π = ( o+α ∞ ) n ∧n '
−1 ' −1 −1 −1 ' −1 '

The intersection of the two planes is the 3-blade (4 + 4−5 )


'
Π ∨ Π=¿
The dual conformal representation of the intersection of the planes is
¿
In Euclidean 3D, the intersection of two orthogonal planes through the origin is a line through
the origin orthogonal to the normals to the planes, namely n ' × n, but in the conformal
representation, o∧∞ are added and in the homogeneous representation, e 0 is added, see below.
In the homogeneous representation, the direct representation of a plane is δ I 3 +e 0 ( n I 3 ), where n
is the unit normal to the plane and δ is the distance of the plane from the origin. To see that, let
the Euclidean shortest vector from the origin to the plane be δ n, and let a , b be two Euclidean
unit orthogonal vectors in the plane. Then nab=n ∧a ∧ b=I 3. The equation of the plane is
( e 0 +δ n ) ∧a ∧b=e 0 ab+ δ I 3 =e 0 n I 3+ δ I 3. The dual representation is
( δ I 3 +e 0 n I 3 ) I −1 −1 −1 −1
4 = ( δ I 3 +e 0 n I 3 ) I 3 e 0 =δ e 0 −n

(see Dorst page 289 section 11.6.2). So the intersection of two orthogonal planes through the
origin (n ' ⋅ n=0) is the 2-blade (3+3−4 )
( δ ' I 3 +e 0 n ' I 3 ) ∨ ( δ I 3 + e0 n I 3 ) =( e0 n' I 3 ) ∨ ( e0 n I 3 ) =−¿
Its dual is also a 2-blade,

4 =e 0 ( n' ∧ n ) I 3 I 4 =e 0 ( n' ∧ n ) e3 e 2 e1 e0 e 1 e 2 e3 =−e 0 ( n ∧ n ) e0 =−¿


e 0 ( n' × n ) I −1 −1 −1 '

which is the negative of the dual of the intersection in the conformal case.
Coming back to Dorst. Had we intersected next the sphere with a third plane Π ' ' , perpendicular
to the first two, then Π ' ' ∨ Π ' ∨ Π ∨ Σ=0 . In Figure 14.1, the plane in (b) separates the bottom
half from the top half. The plane in (c) separate the front half from the back half. The third plane
would separate the left half from the right half. This third plane has an empty intersection with
the two points which are the intersection of the sphere with the first two planes. In Rn +1,1 it would
take n intersecting planes to get an empty intersection (don’t confuse between n the normal
vector, and n the dimension of the space).
Notice that in the intersections above we didn’t need the condition n ∙ n' =0 . As long as n ≠ n ' , the
results do not change. The first intersection gives a great circle. The second intersection gives
two antipodal points on the great circle. The third intersection is empty provided n ≠ n' ≠ n ' ' ≠ n.
So in Rn +1,1 the intersection of the hyper-sphere with k < n distinct hyper-planes, has a dual
conformal representation σ ∧ π 1 ∧⋯ ∧ π k =( o+ α ∞ ) n1 ∧ ⋯ ∧ nk ≡ ( o+ α ∞ ) E k which is equation
(14.1). So Ek is a blade whose vectors are Euclidean vectors. When the ni are mutually
orthogonal, the wedge product in Ek can be replaced by the geometric product, but not otherwise.
These objects are called dual rounds. When k =0 it is a hyper-sphere, and then a hyper-great
circle, and so on.
We didn’t assume so far whether the radius is real or imaginary, only that the radius squared is
real, −2 α . If α <0, the radius is real, and if α >0, the radius is imaginary.
If the sphere and the planes are translated so that the center of the sphere is at a point c (a null
vector), then the dual representation of the translated sphere is c +α ∞. The blade Ek , which is a
dual representation of the wedge product of planes through the origin, after translation to be

planes which go through the point c , has the equation 1− ( c∞


2 ) (
Ek 1+
c∞
2 )
=E k + ∞ ¿ , (see

equation (13.7) page 372). So we got equation (14.2).


Check the case k =1
−¿
which is exactly the dual representation of a plane which goes through the point c and has unit
normal n .
Check the case k =2, the dual representation of the intersection of two planes, which is a line, is
the wedge product of the dual representation of the planes,
( n ' + c ∙ n ' ∞ ) ∧ ( n+c ∙n ∞ ) =n' ∧ n+ ( c ∙ n ) n' ∞ +∞ ( c ∙ n' ) n' =n ' ∧n+ ∞ ( ( c ∙ n' ) n' −( c ∙ n ) n' )
And using the formula,
−¿
In homogeneous model, a line which goes through the point e 0 +c and has a direction vector
'
n ×n , has a direct representation

( e 0 +c ) ∧ ( n ' × n )=e0 ∧ ( n ' ×n )+ c ∧ ( n' × n ) =e 0 ∧ ( n' × n ) + [ c × ( n' × n ) ] I 3 =e 0 ∧ ( n' × n ) +[ ( c ⋅n ) n' −( c ⋅ n' ) n ] I 3


In conformal model, the direct representation of that line is

[ n ∧n+ ∞ ( ( c ⋅ n ) n−( c ⋅n ) n )] (( ∞ ∧ o ∧ I ) )=∞ ∧ [ o ∧ ( n× n ) + ( c ⋅n ) ( n I ) −( c ⋅n ) ( n I ) ]=[ o ∧ ( n × n ) +(( c ⋅ n ) n− ( c ⋅ n


' ' '
3
' '
3
'
3
' '

which is analogous to the homogeneous case, but with a sign change.


14.1.2 Direct rounds
1 2
Above we calculated the direct representation of the sphere with center at point c=o+c + c ∞
2

and radius squared −2 α , to be the 4-blade Σ= −o+ ( ∞∧ o ) c+ [ ( 1


2 ]
c +α ) ∞ I , while its dual
2
3

=c+ α ∞=o+c +( c + α ) ∞. So −o I I =o and


−1 1 2 −1
representation is the vector σ =Σ I 5 3 5
2
( ∞ ∧o ) c I 3 I −1
5 =c . The long way to show that is
( ∞ ∧o ) ( c1 e 2 e 3 +c 2 e3 e 1+ c3 e 1 e 2 ) ( o ∧ ∞∧ e1 ∧e 2 ∧ e3 ) which indeed yields c=c1 e 1+ c 2 e 2+ c 3 e 3. The

short way is first to calculate I 3 I −1


5 =∞ ∧ o. Then ( ∞ ∧o ) c ( ∞ ∧o )= ( ∞ ∧o )( c ∧∞ ∧o ) =¿. And

( 12 c +α ) ∞ I I =( 12 c +α ) ∞.
2 −1
3 5
2

Dorst want to find another expression for Σ using the wedge product of four points on the
sphere, p ∧q ∧ r ∧ s= p ∧ ( q− p ) ∧ ( r− p ) ∧ ( s− p ). The dual of that 4-blade is equation (14.3)
¿
We have to show that this expression for the dual is indeed a scalar multiple of c +α ∞. Now the
inner product of all four points with c +α ∞ must be zero, so 0= p∙ c +αp ∙ ∞=p ∙ c−α . So α = p ∙ c
. Now p ∙ ∞=−1. So we can write c +α ∞=c + ( p ∙c ) ∞=¿ (equation 14.4)). So we need to find
show that ¿ is a scalar multiple of c ∧∞ .
Note that c ∧∞ is a flat point. Recall that a k -flat in Rn +1,1 where k ≤ n, is defined to be directly
represented by p ∧ Ak ∧∞=( o+ p ) ∧ A k ∧ ∞ where Ak is a k -blade of Euclidean vectors. The flat
1 2
contains the point p and the point ∞ and any point q=o +q+ q ∞ such that ( q− p ) ∧ Ak =0. This
2
flat can also be represented by the wedge product p ∧q 1 ∧ ⋯ ∧q k ∧ ∞ where q 1 , ⋯ ,q k are points,
such that p , q1 , ⋯ , q k are linearly independent and ( q i− p ) ∧ A k =0.
When k =n=3, the flat is a 5-blade which is a pseudoscalar p ∧q 1 ∧ q2 ∧ q3 ∧∞ .
When k =2 , n=3 , the flat is a 4-blade representing a plane p ∧q 1 ∧ q2 ∧ ∞ through the point p. In
the previous section we calculated the direct representation of such a plane to be the 4-blade
(−δ−o n ) I 3 ∞ , where δ is the distance of the plane from the origin, and n is the unit normal to the
1 2
plane. Here p=o+δ n+ δ ∞ . Now we know that there exists points q 1 , q 2 such that the
2
representation of that plane is also p ∧q 1 ∧ q2 ∧ ∞.
When k =1 , n=3 the flat is a 3-blade representing a line p ∧q 1 ∧ ∞ through the point p. In the
previous section we calculated the direct representation of such a line to be the 3-blade

[ o ∧ ( n × n )+ ( ( p ⋅ n ) n−( p ⋅n ) n ) I ] ∞ as the intersection of two planes with unit normals n , n ' ,


' ' '
3

which go through the point p. It was not obvious that this is in fact a blade, we only knew that
because the dual of a blade is a blade. So now we know that it can be written as p ∧q 1 ∧ ∞ for
some point q 1.
When k =0 , n=3, we get a flat point p ∧∞ , to be distinguished from the point p which is a null
vector but a flat point is not a null vector. The intersection of a 2-flat, a plane, and a 1-flat, a line,
or the intersection of three 2-flats, or 3 planes, is a 0-flat, a flat point, not a null vector.
Note that Dorst wants to directly represent a sphere by a 4-blade of finite points, which is not a
flat, since it doesn’t contain ∞ .
So we want to compare ¿ to ¿. c ∧∞ is a direct representation of a flat point. And the vectors
q− p , r− p , s− p are not Euclidean vectors, so they are not the dual representations of planes
through the origin. We’ll show that these vectors represent dually planes through the center of
the sphere c . And their wedge, which is a 3-blade, dually represent the intersection of these
planes. And then the un-dual of the wedge, which is a 2-blade, is the direct representation of the
flat point c ∧∞ . To show that q− p represents dually a plane, let x ∙ ( q− p )=0. So x ∙ p=x ∙ q. But
−1 −1
( p−x )2=x ∙ p=x ∙ q= ( q−x )2. So the Euclidean distance between x and p equals the
2 2
Euclidean distance between x andq . So q− p dually represent the plane whose normal is q− p
and which goes through the point half way between p∧q, which corresponds to the Euclidean
1 1
vector t= p+ ( q− p )= ( p+q ). this plane contains every point in space whose distance to p∧q
2 2
is the same, which includes c (equation 14.5).
Recall that the dual representation of a plane is n+ δ ∞ . So q− p must be a scalar multiple of
q−p 1
2
q− p 1 q − p
2
n+ δ ∞ . Indeed, n= and δ=t ∙ n= ( p+ q ) ∙ = . So
|q−p| 2 |q− p| 2 |q− p|

n+ δ ∞=
1
|q− p|( 1
q− p+ ( q2− p2 ) =
2 )
1
|q− p|
( q−p )

Each two of the three planes q− p, r −p , s− p , intersect along a line which goes through the
center of the sphere, and all three obviously intersect at the center. I want to prove that
algebraically.
1
−1
(( Π s ∨ Π r ) ∨ Π q ) I 5 =π q ∧ π r ∧ π s=( q− p ) ∧ ( r− p ) ∧ ( s− p )=− p ∧ q ∧r + p ∧q ∧ s− p ∧r ∧ s+q ∧ r ∧s + ( s2−r 2)
2
Now take the dual of the wedge. Later I realized that below we should calculate ¿, not ¿, since
the wedge is a dual representation of an intersection, so all the signs below should be changed.
¿
¿
¿
¿
1 2 2
( s −r ) ¿
2
−1 2 2
( s −q ) ¿
2
1 2 2
( r −q ) ¿
2
−1 2 2
( p −s ) ¿
2
−1 2
( r − p 2) ¿
2
1 2 2
(q − p ) ¿
2
So
¿
So it must be that

[
[ ( p ⋅ ( q × r ) )−( p ⋅ ( q × s ) ) + ( p ⋅ ( r ×s ) )− ( q ⋅ ( r × s ) ) ] o + −1
2
( s2−r 2 ) ( p ×q ) + 1 ( s 2−q 2 ) ( p × r )− 1 ( r 2−q2 ) ( p × s ) + 1 ( p 2−
2 2 2
Is a scalar multiple of o+ c .The next step is to write p=c+ p ' , q=c +q ' , r =c+ r ' , s=c+ s ' , where
2 2 2 2
p ' =q ' =r ' =s ' . The expression becomes (we write p instead of p ' etc)

[ ( ( c+ p ) ∙ ( q × r )+ p ∙ ( c × ( r −q ) ) )−(( c+ p ) ∙ ( q × s ) + p ∙ ( c × ( s−q ) ) )+ ( ( c+ p ) ∙ ( r × s ) + p ∙ ( c × ( s−r ) ) )−(( c +q ) ∙ ( r × s) +q ∙ ( c


We have to show that the second term is the coefficient of o in the first term times c . This is too
messy to do by hand, so I used Mathematica, and indeed it is true. I simply wrote c , p , q ,r , s
(recall that p here is really p ' where p=c+ p ' , and the same for q ,r , s , so p2=q2=r 2 =s 2=−2 α )
with their e 1 ,e 2 , e3 components, and expanded all the terms. Mathematica was clever enough to
write the second term as γ c , where γ is the coefficient of o in the first term, and
γ= p1 q 2 r 3− p 1 q 3 r 2 −p 1 q 2 s3 + p1 q3 s 2 + p1 r 2 s 3− p1 r 3 s 2− p2 q1 r 3+ p 2 q 3 r 1 + p2 q1 s 3− p2 q3 s1− p2 r 1 s 3+ p2 r 3 s 1+ p 3 q 1 r
Because of the mistake above of doing dualization instead of un-dualization, the signs here
should be changed. So we found an expression for the point at the center of a sphere in terms of
4 points on the sphere. We also found above an expression for the radius squared, and we’ll
derive another below.
Now let’s look at Dorst. The 4-blade p ∧q ∧ r ∧ s directly represents the sphere. It’s dual is a
−1 2
scalar multiple of c +α ∞ (note that Dorst uses α for the scalar multiple, while we use α = ρ ).
2
So
γ ( c+ α ∞ )=¿
where γ is the weight of the vector ¿.
γ=−¿
We calculated γ above in terms of the coordinates of the points p , q , r , s relative to the center of
the sphere. Next, the radius squared of the sphere is −2 α . It is straightforward to prove that

−2 α= o+c +
( ( 12 c + α ) ∞)( o+c +( 12 c + α ) ∞)
2 2
So
−2 α =¿ ¿
which is another expression for the radius squared, besides the one above α = p ∙ c .
Dorst claims that ¿ ¿. To prove that recall that for two blades A , B of grade k ,
A∗B= A ⌋ B=A ⌊ B= ⟨ AB ⟩0. Also, AA=¿. Now

( A I −1 )∗( B I −1 )= ⟨ A I −1 B I −1 ⟩= ⟨ A I −1 ( B ⌋ I −1 ) ⟩ = ⟨ A I −1 (−1 )k (n+1 ) ( I −1 ⌊ B ) ⟩ =(−1 )k (n +1) ⟨ A I −1 I −1 B ⟩ =(−1 )k (n +1) I 2 ⟨ AB ⟩=


So up to a sign, the scalar product of two blades of equal grade is the same as the scalar product
of their duals. In our case, k =4 , n+1=6 , A=B , I 2=−1, so I got the opposite sign from Dorst.
To actually calculate the two sides is very messy. I calculated ¿.
¿
And this vector still needs to be squared. Also

| |
( p−s )2 ( p−r )2 ( p−q )2 0
2 2
−1 ( q−s ) ( q−r ) 0 ( p−q )2
( p ∧q ∧ r ∧ s )2=
2 ( r−s )2 0 ( q−r )2 ( p−r )2
0 ( r−s )2 ( q−s )2 ( p−s )2

I calculated it using Mathematica, and it’s a big mess, naturally.


Show that if t is a fifth point on the sphere, then p ∧q ∧ r ∧ s ∧ t=0. Sufficient to show that
( q− p ) ∧ ( r− p ) ∧ ( s− p ) ∧ ( t− p )=0 .
' 1 1
q− p=o+c +q + ( c+ q ) ∞−o−c− p − ( c+ p ) ∞=( q − p ) + ( c ∙ ( q −p ) ) ∞≡ a+ ( c ∙ a ) ∞
' 2 ' ' 2 ' ' ' '
2 2
So
( q− p ) ∧ ( r− p ) ∧ ( s− p ) ∧ ( t− p )=[ a+ ( c ∙ a ) ∞ ] ∧ [ b+ ( c ∙ b ) ∞ ] ∧ [ d+ ( c ∙ d ) ∞ ] ∧ [ α a+ β b+δ d+ ( c ∙ ( α a+ β b +δ d ) ) ∞ ] =[
Now Dorst on page 402 claims that just like in 3D, a sphere can be directly represented in
conformal model using 4 points on the sphere, so in 2D, a circle can be directly represented
using 3 points on the circle. Now in 1D, the corresponding “sphere” is just two points, where the
center of the “sphere” is the point half way between them, and the radius is half the distance
between the points. So a 1D “sphere” can be represented as the wedge of two points p ∧q . This
point pair p ∧q is to be distinguished from a flat point p ∧∞ , which we get when we intersect a
plane and a line in space, while a point pair does not includes ∞ . Also, this point pair p ∧q is to
be distinguished from a line in 3D considered as a 2-flat, which has a direct representation
p ∧q ∧ ∞, while a plane in 3D has representation p ∧q ∧ r ∧ ∞ and the whole 3D has
representation p ∧q ∧ r ∧ s ∧ ∞, which is a pseudoscalar. The relation of a point to a point pair, is
to be likened to the relation of a flat point to a 2-flat. When above we intersected a sphere
centered at the origin with two distinct planes through the origin, geometrically, the intersection
is two antipodal points on a great circle, and its direct representation is a 2-blade ( o+ α ∞ ) ( n ' ×n ),
where −2 α is the radius squared of the sphere, and n , n ' are the unit normals to the planes. This
is neither a point pair nor a flat point.
Exercise (4) in section 14.8.2 page 427. At first this question doesn’t make sense. How can a
point pair be given except in the form of a wedge of two points. But then I thought of a sphere in

3D. It can be represented directly in the form −o+ ( ∞ ∧ o ) c+ [ ( 12 c +α ) ∞ ] I , while its dual
2
3

−1
representation is the vector σ =Σ I 5 =c+ α ∞=o+c + ( 12 c + α ) ∞. This dual form is correct also in
2

a plane and in a line. First calculate the unit pseudoscalars in the plane and line.
~
( ∞
) ∞
( )
I 4= I 4 =e e1 e2 e= o− ∧ e1 ∧e 2 ∧ o+ =o∧ ∞ ∧ e 1 ∧ e2=o ∧ I 2 ∧ ∞
2 2
~ −1
−I 4=−I 4 =I 4 =∞ ∧o ∧ I 2=∞∧ I 2 ∧o

I 3=I −1 ~
3 =− I 3=e ∧e ∧ e= o−

2 ( ∞
)
∧ e ∧ o+ =o ∧e ∧∞=o ∧ I 1 ∧∞
2 ( )
In contrast,
~ −1
I 5= I 5=o ∧ I 3 ∧∞=∞∧ o ∧ I 3 ∧I 5 =o ∧ ∞ ∧ I 3
So the direct representation of a circle in a plane where the origin is also in the plane, is

[ (
o+ c+
1 2
2 ) ] [ 1
( ) ] 1
(
c + α ∞ I 4 = o+c 1 e1 +c 2 e2 + c 2+ α ∞ ( o ∧∞ ∧e 1 ∧ e2 ) =o I 2+ o ∧∞ ( c I 2 )− c 2 +α ∞ I 2= o+ ( o ∧∞ )
2 2 ) [
And the direct representation of a point pair in a line where the origin is in the line, is
¿

[
Check that p ∧q ∧ r is a scalar multiple of o+ ( o ∧∞ ) c− ( 12 c + α ) ∞] I . To make life easy, I’ll
2
2

assume that c=o, and p , q , r are points on a circle centered about the origin whose radius
squared is −2 α , so p2=q2=r 2 =−2 α . So p=o+ p1 e1 + p2 e 2−α ∞ , and similarly for q∧r . And I
expect that p ∧q ∧ r=γ ( o−α ∞ ) e 1 e 2. Indeed, straight calculation gives
p ∧q ∧ r=( ( q1− p 1) r 2 −( q2 −p 2 ) r 1+ p1 q 2−p 2 q 1 ) ( o−α ∞ ) e 1 e 2
Above we proved that p ∧q ∧ r ∧ s is a scalar multiple of −o+ ( ∞ ∧ o ) c+ [ ( 12 c +α ) ∞ ] I , by
2
3

showing that their duals are scalar multiples.


Now assume that a “sphere” is in 1D. Denote by ρ ≡ √ −2 α the radius and let c=β e . Need to
show that

[ −o+ β ( ∞ ∧o ) e + ( 12 β +α ) ∞ ] e=γ ( o+( β−ρ ) e + 12 ( β−ρ ) ∞) ∧( o+( β + ρ ) e+ 12 ( β + ρ ) ∞)


2 2 2

The calculation shows that

( )( ) ( ( ) ) ( (
2 2
1 2 1 2 β ρ
γ o+ ( β−ρ ) e+ ( β −ρ ) ∞ ∧ o+ ( β + ρ ) e+ ( β + ρ ) ∞ =2 γρ o e+ βo∧ ∞+ − e ∞ =−2 γρ −o e + β ∞ ∧o +
2 2 2 2
−1
So γ= .

Now let’s look at the exercise. Suppose that a point pair is given as a bivector like

( )
2
β
P=e o+ β ∞ ∧o+ +α ∞ e . Since we know that this represents a point pair, a 1-dimensional
2
“sphere” whose center is at β e and whose radius is ρ=√−2α , it must be that for some scalar
coefficient γ ,

(
γ e o+ β ∞ ∧ o+ ( β2
2 ) )( 1
2
2
)( 1
2
2
+α ∞ e = o+ ( β−ρ ) e+ ( β− ρ ) ∞ ∧ o+ ( β+ ρ ) e+ ( β + ρ ) ∞ )
Indeed it is true where γ=−2 ρ .

Dorst claims that these two points are P−√ P and P+ √ P . In the format above, −¿ , and also ¿ ¿
2 2

−¿ ¿ −¿ ¿
.

2
(
P = e o+ β ∞ ∧ o+ ( β2
2 ) )(
+ α ∞ e ∙ e o+ β ∞ ∧o+
β2
2 (
+α ∞ e =−
2) ) (
β2 2
+α + β −
β2
2 ) (
+α =−2 α )
So √ P2=± √ −2 α is the radius ± ρ. Now in our form
P¿¿
1 2 1 2
So we get the two points o+ ( β−ρ ) e+ ( β− ρ ) ∞∧o+ ( β + ρ ) e+ ( β + ρ ) ∞ as above. If P is given
2 2
in the form of a arbitrary bivector in R1+1,1 .
τo e +φo ∧ ∞+ ω ∞ e
Compare that to
( )
2 2
β ρ
e o+ β ∞∧ o+ − ∞e
2 2
where we know that β e is the location of the center of the “sphere” and the radius is ρ . Equating
φ
τ
ω
(
the coefficients, τo e +φo ∧ ∞+ ω ∞ e has to be rewritten as τ o e + o ∧ ∞+ ∞ e . Then τ can be
τ )
φ
√( )
2
φ ω
any non-zero scalar, e is the location of the center of the “sphere” β e, and −2 is the
τ τ τ
radius of the “sphere” ρ . To check for consistency, consider the wedge of two arbitrary points

( o+γ e+ 12 γ ∞) ∧ (o+ δ e+ 12 δ ∞)=( γ −δ ) [ e o+ 12 ( γ + δ ) ∞ ∧o+ 12 γδ ∞e ]


2 2

( )
2 2
β ρ
Equating the coefficients to the coefficients of e o+ β ∞ ∧ o+ − ∞ e , we get either
2 2
γ=β + ρ∧δ=β−ρ or γ=β −ρ∧δ=β + ρ , which is consistent.
So given any bivector in τo e +φo ∧ ∞+ ω ∞ e in R1+1,1 , converts the coefficients to look like

( )
2 2
β ρ
e o+ β ∞∧ o+ − ∞ e , and then the solutions of Dorst checks out. But working with
2 2
arbitrary coefficients τ , φ , ω is a mess.
14.1.3 Oriented Rounds
2
ρ
Recall the dual representation of a sphere centered at the point c and radius ρ is σ =c− ∞. And
2
the dual representation of a plane with unit normal n through the point c is π=n+ ( c ∙ n ) ∞ . Note
that σ ∙ π =0 which Dorst interprets that the plane is perpendicular to the sphere. At each point on
the intersection of the plane and the sphere, there are 3 mutually orthogonal unit vectors. The
normal to the plane, n . The normal to the sphere which is parallel to the plane. The tangent to the
sphere in the direction of the great circle which the plane cut in the sphere, which is also in the
plane. Above we calculated the direct representation of a circle in a plane where the origin is in

[
the plane to be o+ ( o ∧∞ ) c− ( )
c 2− ρ2
2 ]
∞ I 2, which is a trivector in R2+ 1,1. But if we want the

calculation to be at R3 +1,1, then the direct representation of the great circle of intersection is
Π ∨ Σ=¿
which is a trivector in R3 +1,1. Its dual representation is a bivector in R3 +1,1,
( Π ∨ Σ ) I −1 −1 −1 1 2
(
5 =Σ I 5 ∧ Π I 5 =σ ∧ π = c− ρ ∞ ∧ ( n+ ( c ∙ n ) ∞ )
2 )
1 2
The center of that great circle is at c . A point on the circle is p=o+ ( c + ρ x )+ ( c+ ρ x ) ∞ for any
2
unit vector x such that x ∙ n=0. For example, n=n1 e1 +n 2 e 2+ n3 e3 , where 1=n21+ n22 +n23 and
−n2 e 1+ n1 e 2
x= . So x is a unit normal to the sphere at p and the unit tangent at p is t=n ∧ x . A
√1−n 2
3

direct representation of the great circle is the wedge of 3 points on it, p ∧q ∧ r , for example the
n3 e1 −n1 e3
point p above, the point q which corresponds to x= , and the point r which
√ 1−n 2
2

−n3 e 2+ n2 e3
corresponds to x= .
√1−n 2
1

We know from section 13.3 page 370, that the direct representation of the carrier plane of the
circle, namely the plane whose intersection with the sphere produces the great circle, is
p ∧q ∧ r ∧ ∞, which must be equal to
o∧¿
I calculated

[( ) ( ) ][ ( ) ( ) ][
2 2
−n2 e 1 +n1 e2 1 −n e +n e n e −n e 1 n e −n e
p ∧q ∧ r ∧ ∞= o+ c+ + c + 2 1 21 2 ∞ ∧ o+ c + 3 1 12 3 + c+ 3 1 12 3 ∞ ∧ o+
√ 1−n 2
3
2 √1−n3 √1−n2 2 √ 1−n 2
for ρ=1 using Mathematica, and it came out that it is equal to
−n3 o ∧ e1 ∧e 2 ∧ ∞−n2 o ∧ e3 ∧e 1 ∧ ∞−n1 o ∧ e1 ∧e 2 ∧ ∞−( c ∙n ) e 1 ∧ e 2 ∧e 3 ∧∞
times
n1 √1−n1+ n2 √1−n2 +n3 √1−n3
2 2 2

√ 1−n √ 1−n √ 1−n


2
1
2
2
2
3

Now let’s look at Dorst. Up to this point, we started with the dual representation of a sphere in
1 2
R
3 +1,1
or a circle in R2+ 1,1 or a point pair in R1+1,1 , namely o− ρ ∞ if the center is at the origin,
2
and then got the direct representation using un-dualization. For example, through un-dualization
1 2
[
1 2
( ) ]
of σ =c− ρ ∞ , we got Σ= −o+ ( ∞ ∧ o ) c+ c +α ∞ I 3 . And if the sphere is centered at the
2 2
1 2
2
1 2
( )
origin, then un-dualization of o− ρ ∞ gives − o+ ρ ∞ I 3. Here Dorst want to start with the
2
direct representation and get the dual representation. He starts with the space Rn +1,1 and asks what

( o+ 12 ρ ∞) A represents directly, where A is an Euclidean blade, and k ≤ n (equation (14.6))


2
k k

For example, when k =n=3, we have a 4 -vector in R3 +1,1, which is a weighted sphere centered at
the origin with radius ρ , and o ∧ A 3 ∧ ∞ is the whole R3 +1,1 space determined by 4 points on the
sphere or the origin plus three points on the sphere.
And if k =n−1=2, then we have a trivector in R3 +1,1, which is a circle centered at the origin
parallel to the plane o ∧ A 2 ∧ ∞ which is a 4 -vector in R3 +1,1, which goes through the origin,
determined by 3 points on the circle or the origin plus two points on the circle.
And if k =n−2=1, then we have a bivector in R3 +1,1, which is point pair centered at the origin on
the line o ∧ A 1 ∧ ∞ which is a trivector in R3 +1,1, which goes through the origin, determined by 2
points, the pair points, or the origin plus one of the pair points.
1 2
( )
Now let’s calculate the dual of Σ= o+ ρ ∞ A k , whose carrier flat through the origin is
2
Π=o A k ∞ . Now the unit pseudoscalar of the space is

(
I n+2=e I n e=(−1 )n e e I n=(−1 )n o−

2 )(∞
)
o+ I n= (−1 )n ( o ∧ ∞ ) I n=o ∧ I n ∧ ∞
2
−1 ~
(n+1
I n+2=−e I n e= (−1 ) o+

2
∞ ~
)( n +1
) ~ ~
o− I n=(−1 ) ( ∞ ∧ o ) I n=o ∧ I n ∧ ∞
2
¿
This is equation (14.7). When k =n=3, Ak⋆ is a scalar, so we get the weighted dual of a sphere
centered at the origin we’ve seen before, which is a vector.
But when k =n−1=2, the dual is a bivector. It is the dual representation of a circle in space, in
the plane which goes through the origin, and contains the vectors x , y where A2=x ∧ y . Now

A2 =x × y , which is a normal n to the plane. So the dual representation is a weighted

( o− 12 ρ ∞) n which is a bivector, and we got that expression above for the case when the center
2

is not at the origin.


And when k =n−2=1, the dual is a trivector. It is a dual representation of a point pair in space,
in a line which goes through the origin, which contains a vector A1=x . So

A1 =x 1 e2 e3 + x 2 e3 e 1+ x 3 e 1 e 2.
So we can look at a point pair in R1+1,1 and then its direct representation is a bivector and its dual
is a vector, while in R3 +1,1 the direct representation of a point pair is bivector and its dual is a
trivector. The bivector in the direct representation is essentially the same when the center of the
1 2
( ) 1 2
point pair is at the origin, o+ ρ ∞ I 1 as opposed to o+ ρ ∞ x .
2 2 ( )
So In R3 +1,1 a sphere corresponds to k =3, a circle to k =2 and a point pair to k =1, but to conform
to tradition, a sphere in 3D is 2-sphere, a circle in a plane is 1-sphere and a point pair on a line is
1 2
(
0 -sphere. When the radius is imaginary, the direct representation is o− ρ ∞ A k and the dual is
2 )
( o+ 12 ρ ∞) A .
2 ⋆
k

To translate the origin we use the operator T c. My understanding is that c ∧ Ak =0. Take the case
k =3.

(1− c2∞ )( o A + 12 ρ ∞ A )(1+ c2∞ )=[ o A + 12 ρ ∞ A − 12 ( c ∞ ) ( o A )− 14 ρ ( c ∞ ) ( ∞ A ) ](1+ c2∞ )=o A + 12 ρ ∞ A


3
2
3 3
2
3 3
2
3 3
2
3

1 −1 1
( 3 ) − ( c ∞ ) ( o A3 ) = c ( −A 3 +∞ ∧o ∧ A 3 )= ¿
2 2 2
1
( 4 )− ρ2 ( c ∞ ) ( ∞ A3 ) =0
4
1 1 1
( 5 ) ( o A3 ) ( c ∞ )= ( o A 3 ⌊ c ¿ ¿ ( ∞ )= (− A 3 ⌊ c +o ( A3 ⌊ c ¿ ¿ ∞ ¿ ¿
2 2 2
1
( 6 ) ρ2 ( ∞ A 3 ) ( c ∞ )=0
4
1 −1 −1
( 7 ) − ( c ∞ ) ( o A 3) ( c ∞ ) = c (− A3 + ∞∧ o ∧ A 3 ) ( c ∞ )= ¿
4 4 4
1
( 8 ) − ρ2 ( c ∞ ) ( ∞ A 3 ) ( c ∞ ) =0
8

(1− c2∞ )( o A + 12 ρ ∞ A )(1+ c2∞ )=o A + 12 ρ ∞ A + 12 c ⌋ A − 12 ∞ ∧o ∧ ( c ⌋ A )− 12 ( A ⌊ c )+ 12 o ∧ ( A ⌊ c )∧ ∞−


3
2
3 3
2
3 3 3 3 3

Dorst formula (14.9) for the translation for the case k =3 is


( 1
c+ ρ2 ∞ ∧¿
2 )
When expanded gives the same result. Dorst formula is simply

[( 1 2
) ] (1 2
)
T c o+ ρ ∞ ∧ A 3 =T c o + ρ ∞ ∧T c ( A3 ) and then using equation (13.7) page 372. In
2 2
equation (14.10) Dorst followed the same outermorphism, employing equations (13.7) & (14.7).
T c ( (−1 )n A ⋆k ) =(−1 )n T c ( A ⋆k ) =(−1 )n ¿

A k ) and (−1 ) A ⋆k = ^

Note the difference between ( (−1 ) A k ) =( ^
k ⋆ n−k ⋆
Ak.
I shall now check the case k =4 too.
Using Dorst formula one gets

( o− 12 ( c − ρ ) ∞+ o ∧∞ ∧c ) A
2 2
4

The long way,

(1− c2∞ )( o A + 12 ρ ∞ A )(1+ c2∞ )=o A + 12 ρ ∞ A − 12 ( c ∞ ) ( o A )− 14 ρ ( c ∞ ) (∞ A )+ 12 ( o A ) ( c ∞ )+ 14 ρ ( ∞ A ) (


4
2
4 4
2
4 4
2
4 4
2
4

1 −1 1
( 3 ) − ( c ∞ ) ( o A 4 )= c (− A 4 +∞ ∧o ∧ A 4 ) = ¿
2 2 2
1
( 4 )− ρ2 ( c ∞ ) ( ∞ A 4 )=0
4
1 1 1
( 5 ) ( o A 4 ) ( c ∞ )= ( o A 4 ⌊ c ¿ ¿ ( ∞ ) = A 4 ⌊ c ¿ ¿ ∞
2 2 2
1
( 6 ) ρ2 ( ∞ A 4 ) ( c ∞ )=0
4
1 1 1
( 7 ) − ( c ∞ ) ( o A 4 ) ( c ∞ )= c ( A 4−∞ ∧ o ∧ A 4 ) ( c ∞ ) = ¿
4 4 4
1
( 8 ) − ρ2 ( c ∞ ) ( ∞ A 4 ) ( c ∞ )=0
8

(1− c2∞ )( o A + 12 ρ ∞ A )(1+ c2∞ )=o A + 12 ρ ∞ A + 12 c ⌋ A − 12 ∞ ∧ o ∧( c ⌋ A )+ 12 A ⌊ c + 12 o ∧ ( A ⌊ c )∧ ∞−


4
2
4 4
2
4 4 4 4 4

Same as Dorst. I also checked the case k =2 and it comes out ( o− ( c − ρ ) ∞+ o ∧∞ ∧c ) A .


1 2 2
2
2
So it seems that the general formula for the direct representation of a ( k −1 )-sphere whose center
is at the point c and whose radius is ρ and whose carrier flat is c ∧ Ak ∧∞ is
( 1
)
o− ( c2− ρ2 ) ∞+ o ∧∞ ∧c A k. Note that such a sphere can also be directly represented by a
2
wedge of k +1 finite points on the sphere. The direct representation of the carrier flat is the
wedge of k + 2 points - the center of the sphere, k of the k +1 finite points on the sphere, and the
point of infinity.
14.2 Tangents as Intersections of Touching Rounds
We’ve just shown that the direct representation of a ( k −1 )-sphere whose center is at the point c

(
1 2 2
and whose radius is ρ and whose carrier flat is c ∧ Ak ∧∞ is o− ( c − ρ ) ∞+ o ∧∞ ∧c A k, or
2 )
the wedge of k +1 points on the sphere. However, when ρ=0, there are no k +1 distinct points on
the sphere. In the case k =n=3, the direct representation, equation (14.9), becomes

( o− 12 c ∞+ o ∧ ∞ ∧c ) I , and the dual representation, equation (14.10), becomes


2
3

c ∧¿
namely minus the center point, a null vector. When the radius is nonzero, equation (14.10)
1 2
becomes − c− ρ ∞ .
2 ( )
Consider two spheres as in Figure 14.2, of equal radius ρ , whose centers are
1 1
o−e 1 + ∞∧o+ e1 + ∞ , where e1 points upward. Their dual representations are
2 2
1 1 2 1 1 2
σ 1=o−e 1+ ∞− ρ ∞ and σ 2=o+ e1 + ∞− ρ ∞ (we omit the meaningless minus signs). The
2 2 2 2
dual representation of their circular intersection, is equation (14.11)

( 1
2 )( 1
2 ) (1
)
σ 1 ∧ σ 2= o−e 1+ ( 1−ρ2 ) ∞ ∧ o+ e1 + ( 1−ρ 2) ∞ =2 o ∧ e1 + ( 1− ρ2 ) ∞ ∧ e1= o− ( ρ2−1 ) ∞ ∧ 2 e1
2
Now π=2 e 1 is the dual representation of the e 2 e3 horizontal plane through the origin whose non-
1 2
unit normal is 2 e1 . And σ 3=o− ( ρ −1 ) ∞ , if ρ2 >1, is the dual representation of a sphere whose
2
center is at the origin, and whose radius squared is ρ2−1 . So σ 1 ∧ σ 2=σ 3 ∧ π , namely, the
intersection of the two spheres is exactly the same as the intersection of the third sphere and the
plane, this intersection being a circle centered at the origin with radius squared ρ 2−1 . We can
also derive the dual representation of the intersection circle using Equation (14.10) when
k =n−1=2.
−¿
So using equation (14.10), the dual representation of the circle centered at the origin and radius
1 2
( )
squared ρ2−1 whose carrier plane has normal e 1 is o− ( ρ −1 ) ∞ ∧−e 1 which is the same as
2
the expression we derived above directly, up to a scalar coefficient.
Now suppose ρ2 <1. So the spheres do not intersect physically, since their centers are at ± e 1 and
1 2
their radius squared is ρ2 <1. The calculation of the intersection gives o− ( ρ −1 ) ∞ ∧2 e1
2 ( )
which is the dual representation of the circle which is the intersection of a sphere with center at
the origin and the radius squared ρ2−1 which is negative, with the horizontal plane. So the
mathematical expression of no intersection of the two sphere is the intersection of a sphere with
imaginary radius and the horizontal plane. So the resultant circle has imaginary radius.
Finally, let ρ2=1 . So the intersection of the two sphere with centers at ± e 1 and radius 1, is

( o− 12 ( ρ −1) ∞) ∧2 e =o ∧ e , which is a 2-blade. The physical intersection is the point of origin,


2
1 1

but the mathematical intersection is not a null vector, but a null 2-blade. The direct
representation of the intersection is −o ∧ e2 ∧e 3, which is a null 3-blade. This is not a flat, since
it has no ∞ , nor a circle, since a circle is the wedge of three points on the circle while e 2 ,e 3 are
not points. This is a new form which we haven’t seen yet, a tangent, directly represented by a
null blade, o ∧ A k .
If the two sphere in Figure 14.2(b) are moved anywhere in space, not necessarily in the e 2 e3
plane, their touch, say at point c , would still be a tangent blade, which is not c ∧ Ak ! It is
T c ( o∧ A k ) =T c ( o ) ∧T c ( A k ).

(
T c ( o )= 1−
c∞
2 )(
o 1+
c∞
2
= o−
2 )(
c∞ o
1+
c∞
2
=o+ )(
2

2
− )
oc ∞ c∞o c ∞ oc ∞
4
1
2
1
=o+ ( c−o∧ ∞∧ c ) + ( c+ o ∧∞ ∧
2
And by equation (13.7), T c ( A k )=−¿ . So T c ( o∧ A k ) =c ∧ ¿. Since ( o ∧ A k ) ( o ∧ A k )=0, also
0=T c ( 0 )=T c ( ( o ∧ A k ) ( o ∧ A k ) ) =T c ( o ∧ A k ) T c ( o ∧ A k ), see page 194. Let’s prove it directly.
c ∧ Ak + c ∧¿
¿
¿¿
¿ [ o A k ]∗[ o A k ]+ [ o A k ]∗[ c ∧ A k ] + [ o Ak ]∗ [ 1 2
2 ]
c ∞ Ak + [ o A k ]∗¿

+ [ c ∧ A k ]∗[ o A k ]+ [ c ∧ A k ]∗[ c ∧ Ak ] + [ c ∧ A k ]∗
[ 1 2
2 ]
c ∞ A k + [ c ∧ A k ]∗¿

+
[ 1 2
2 ] 1
2 [ ]1
2
1
2 [ 1
c ∞ A k ∗[ o A k ] + c 2 ∞ A k ∗[ c ∧ A k ]+ c 2 ∞ A k ∗ c 2 ∞ Ak + c 2 ∞ A k ∗¿
2 ][ ][ ]
+¿
+¿
Note that
2
Ak ∗A k =( det ( A k ) ) (−1 )k (k −1 )/ 2
k+1 2
[ o A k ]∗[ ∞ A k ]= [ ∞ A k ]∗[ o A k ]=(−1 ) A k ∗A k =−( det ( A k ) ) (−1 )k (k +1) /2

On the RHS are two answers, the first if c ∧ Ak =0, the second if ¿.
( 1 ) [ o A k ]∗[ o A k ]=0,0
( 2 ) [ o A k ]∗[ c ∧ A k ]=0,0

[ ]
k ( k+1 ) k ( k+1)
( 3 ) [ o Ak ]∗ 1 c 2 ∞ A k = 1 c 2 [ o A k ]∗[ ∞ Ak ]=¿− 1 (−1 ) 1 2 2
c ( det ( A k ) ) ,− (−1 ) c ( det ( A k ) )
2 2 2 2
2 2 2 2
( 4 ) [ o Ak ]∗¿
( 5 ) [ o Ak ]∗¿
( 6 ) [ c ∧ Ak ] ∗[ o Ak ] =0,0
( k +1) k
2
( 7 ) [ c ∧ A k ]∗[ c ∧ A k ] =0 , (−1 ) 2
c 2 ( det ⁡( A k ) )

( 8 ) [ c ∧ A k ]∗ [ 1 2
2
c ∞ A k =0,0
]
( 9 ) [ c ∧ A k ]∗¿
( 10 ) [ c ∧ A k ]∗¿

[ ]
k ( k+1 )
( 11 ) 1 c 2 ∞ A k ∗[ o Ak ] = 1 c 2 [ ∞ A k ]∗[ o A k ]= −1 (−1 ) 1
2 ( ) 2
2
c 2 ( det ( A k ) ) ,− (−1 )k k+1 /2 c 2 ( det ( A k ) )
2 2 2 2

( 12 ) [ 1 2
2 ]
c ∞ A k ∗[ c ∧ A k ]=0,0

( 13 ) [ 1 2
2
1
][
c ∞ Ak ∗ c 2 ∞ A k =0,0
2 ]
( 14 ) [ 1 2
2
c ∞ A k ∗¿]
)[ ]
1 2
( 15 c ∞ Ak ∗¿
2
( 16 ) ¿
( 17 ) ¿
( 18 ) ¿
( 19 ) ¿
( 20 ) ¿
( 21 ) ¿
( 22 ) ¿
( 23 ) ¿
( 24 ) ¿
( 25 ) ¿
Adding up gives zero in both cases.
14.2.1 Euclid’s Elements
I shall make a summary. The space is Rn +1,1 with orthonormal basis e 1 ,⋯ , en , e , e , where all base
1
vectors square to 1 except e which square to minus 1. Define o= ( e +e ), ∞=e−e . oo=∞ ∞=0 .
2
~
o ∞=−1+o ∧ ∞. I n+2=e e 1 ⋯ en e=o e1 ⋯ e n ∞=o I n ∞. I −1 −1
n+1=o I n ∞=o I n ∞ . Every null vector is

called a point. Every multivector A in the subspace Rn whose basis is e 1 ,⋯ , en , called Euclidean.
1 2
A Euclidean vector x corresponds to a null vector in Rn +1,1, x=o + x+ x ∞. x is called the
2
location of x . The zero vector in Rn corresponds to o . The inner product of two null vectors x , y
−1
is x ∙ y= ( x− y )2 . The weight of a vector x is the coefficient of o , which is −∞ ∙ x . The
2
1 2 x
location of a null vector αo + x + x ∞ is . Every non-null vector can be expressed as
2 α
σ =x +α ∞ where x is null, and α is a scalar. The sets of points p such that p ∙ σ=0 satisfy
( p−x )2=−2α ≡ ρ2 which is a sphere centered at point x and radius squared −2 α . So σ is the
dual representation of a sphere. The direct representation of that sphere is an n+1 blade,
[ 1
]
Σ=σ I n+ 2= o+ ( o ∧ ∞ ) x− ( x 2−ρ2 ) ∞ I n . Another direct representation is the wedge product of
2
n+1 points on the sphere. A point is a dual representation of a zero radius sphere, centered at the
point. A vector with weight zero, π=n+ δ ∞ is the dual representation of an hyper plane whose
unit normal is n and whose distance from the origin o is δ . The direct representation of the hyper
plane is the n+1 blade Π= (−δ−o n ) I n ∞. Another direct representation of an hyper plane is the
wedge of n points on the hyper plane and ∞ . This is called an n−1 dimensional flat even though
it is represented by an n+1 blade. The direct representation of a k dimensional flat, with k ≤ n, is
a k + 2 blade, the wedge of k +1 points and ∞ , p ∧q 1 ∧ ⋯ ∧q k ∧ ∞. It can also be written as the
wedge of a point and k Euclidean vectors and ∞ , p ∧ ( q 1− p ) ∧⋯ ∧ ( q k − p ) ∧∞ =p ∧ A k ∧ ∞ . A 0
dimensional flat, Ak =1, called a flat point, is a 2-blade, not a vector. The dual representation of
p ∧ Ak ∧∞ is (−1 )n−k ¿ which is n−k blade. The direct representation of a direction is Ak ∧ ∞ a
k +1 blade. The direction of a flat point is ∞ . The dual representation of a direction is −¿, an
n−k +1 blade. The direction is unaffected by translation.
A 2-sphere, which is a sphere in 3D with center at point c , can be directly represented by the
wedge of 4 finite points p ∧q ∧ r ∧ s on the surface of the sphere. A plane in 3D is the wedge of
3 finite points and infinity, so it is a 2 dimensional flat. If it intersects the sphere through the
center point c , it is directly represented by c ∧ p ∧q ∧ ∞. So the plane does not contain the
sphere. The whole 3D space is a 3 dimensional flat, the wedge of 4 finite points and infinity,
c ∧ p ∧q ∧ r ∧ ∞. So it’s a wedge of a sphere and infinity. So it contains the sphere and all
planes, so it’s called the carrier flat of the sphere. The intersection of a sphere with a plane,
which is a 2 dimensional flat, through the center of the sphere, is a great circle, which is a 1-
sphere in 3D, directly represented by the wedge of 3 finite points, p ∧q ∧ r , and the plane, the 2
dimensional flat, can be directly represented as c ∧ p ∧q ∧ ∞, and is the carrier flat of the great
circle. The intersection of a line, a 1-dimensional flat, through the center of the great circle, is a
pair point, a 0-sphere, directly represented by the wedge of 2 finite points, p ∧q . The intersecting
line is a 1 dimensional flat, the carrier flat of the 0-sphere, directly represented by the wedge
c ∧ p ∧∞ .
The general rule is. A k −1 sphere has a carrier flat of dimension k . The k −1 sphere is directly
represented by the wedge of k +1 finite points on the sphere, and the carrier flat is directly
represented by the wedge of k +1 finite points, being the center of the sphere and k finite points
on the sphere and infinity, and the k finite points can be replaced by k Euclidean vectors.
Another direct representation of the k dimensional flat is c ∧ Ak ∧∞ , which is a carrier flat for
the k −1 sphere whose direct representation is

( o− 12 ( c − ρ ) ∞+ o ∧∞ ∧c ) A =( c+ 12 ρ ∞ )∧ ¿
2 2
k
2

The dual representation of the k −1 sphere is equation (14.10)

( c− 12 ρ ∞ )∧ ¿
2

And the dual representation of the k dimensional flat through the center of the sphere is equation
(13.9)
(−1 )n−k ¿
The last basic blade we discuss is tangents, which are neither flats nor sphere nor directions, but
~
are null blades, c ∧ Ak + c ∧¿ . It’s dual at c=o is (−1 ) o ∧ ( A k I n ). A 2-sphere in R3 +1,1 is a 4-blade
n

whose dual is a vector. The intersection of two 2-spheres is a 1-sphere, a 4 + 4−5=3 blade
whose dual is a 2-blade. This is also true when the intersection is a point say at the origin. So the
direct representation of the point intersection at the origin is o ∧ A 2, which is the tangent.
Let’s work an example of the intersection of two circles in R3 +1,1. A circle of radius one centered
1
( )
at the origin has direct representation Σ= o+ ∞ A 2. It’s orientation in space depends on A2.
2
1 2
Let’s take A2=e1 e 2. Suppose x=o + x 1 e 1+ x2 e 2+ x 3 e 3+ x ∞ is on the circle, or x ∧ Σ=0. So
2

( 1 2
2 )( 1
2 ) 1
2
1
( )
0= o+ x 1 e1 + x 2 e 2 + x3 e 3 + x ∞ ∧ o+ ∞ ∧e 1 ∧ e2 =−x3 o+ ∞ I 3 + ( 1−x ) ( o ∧ e1 ∧ e 2 ∧ ∞ )
2
2

So x 3=0 , and x 21+ x22 =1, which makes perfect sense. It could also be represented as

( o+e + 12 ∞) ∧( o+e + 12 ∞) ∧ (o−e + 12 ∞)=2(o+ 12 ∞) e e


1 2 1 1 2

The dual is

( o+ 12 ∞) e e I =( o+ 12 ∞) e e o e e e ∞=(o+ 12 ∞) o e ∞=¿
−1
1 2 5 1 2 3 2 1 3

where e 3 is normal to the carrier flat of the circle. This fits exactly equation (14.7) for the case
1
k =n−3=2. Now let’s translate this circle to create two circles, one centered at c 1=o+e 1+ ∞
2
1
and the other at c 2=o−e1 + ∞ , so they touch at the origin. So we get two circles whose direct
2
representations are, according to equation (14.9),
Σ 1=( o+e 1+ ∞ ) ∧ ¿
Check it by wedging three points on the circle.
o ∧ ( o+2 e 1+ 2 ∞ ) ∧ ( o+ e1 +e 2+ ∞ )=2 ( o e1 e2 −o e 2 ∞ )
Σ 2=( o−e1 +∞ ) ∧¿
Check it by wedging three points on the circle.
o ∧ ( o−2 e1 +2 ∞ ) ∧ ( o−e 1 +e 2+ ∞ )=−2 ( o e1 e2 + o e2 ∞ )

The intersection of two 3-blades is Σ 1 ∨ Σ 2 =¿. Here J is the 2 dimensional flat o ∧ e1 ∧e 2 ∧ ∞


which is the carrier flat for both 1-spheres Σ 1∧Σ 2.
¿
¿
Which is a null 2-blade. This is the direct representation of the tangent. It’s direction is
perpendicular to the line between the centers of the circles.
It’s dual representation is not relative to J but relative to the whole space.
−2 ¿
which is a null 3-blade.
Let’s work an example of the intersection of two point pairs in R3 +1,1. A point pair of radius one
1
( )
centered at the origin has direct representation Σ= o+ ∞ A 1. It’s orientation in space depends
2
on A1. Let’s take A1=e1.
It could also be represented as

( o+e + 12 ∞) ∧( o−e + 12 ∞ )=−2( o+ 12 ∞ ) e


1 1 1

The dual is

( o+ 12 ∞) e I =(o+ 12 ∞) e o e e e ∞=−( o+ 12 ∞ ) o e e ∞=−¿


−1
1 5 1 3 2 1 3 2
1
Now let’s translate this point pair to create two point pairs, one centered at c 1=o+e 1+ ∞ and
2
1
the other at c 2=o−e1 + ∞ , so they touch at the origin. So we get two point pairs whose direct
2
representations are, according to equation (14.9),
Σ 1=( o+e 1+ ∞ ) ∧ ¿
Check it by wedging two points on the point pair.
o ∧ ( o+2 e 1+ 2 ∞ ) =2 ( o e 1+ o ∧ ∞ )
Σ 2=( o−e1 +∞ ) ∧¿
Check it by wedging two points on the point pair.
o ∧ ( o−2 e1 +2 ∞ )=−2 ( o e1 −o ∧ ∞ )

The intersection of two 2-blades is Σ 1 ∨ Σ 2 =¿. Here J is the 1 dimensional flat o ∧ e1 ∧∞ which
is the carrier flat for both 0-spheres Σ 1∧Σ 2.
¿
¿
Which is a null 1-blade. This is the direct representation of the tangent. It has no direction.
It’s dual representation is not relative to J but relative to the whole space.
−2 ¿
which is a null 4-blade.
So the touch intersection of 2-spheres in 3D space or two 1-spheres in 2D space, or two 0-
spheres in 1D space, is a tangent, o A 2 , o A1 , o A0 , respectively. But note that, if the two spheres
are not in one 3D space, or if the two circles are not in one plane, or if the two point pairs are not
in one line, even though they just touch, the intersection does not have a form of a tangent, which
is a null blade, and when the touch is at the origin, a tangent has the form of a wedge of the
origin with an Euclidean blade without infinity. In the case of two spheres in one 3D who touch
at c , the direct representation of the tangent is c ∧ A2 +c ∧¿. Let’s work it out for a specific
simple case. Let A2=e2 e 3 and let c=o+e 1 +e 2+ ∞. Then the direct representation of the tangent is
( o +e 1+ e2 +∞ ) ∧ e2 ∧e 3+ ( o+ e1 +e 2+ ∞ ) ∧¿

| || |
( o +e 1) ∙ e 3 ( o+e 1 ) ∙ ( e2 +∞ ) ( o+e 1 ) ∙ ( o+e 1 ) 0 −1 1
Its square is ( e2 +∞ ) ∙ e3 ( e 2+ ∞ ) ∙ ( e 2 +∞ ) ( e2 + ∞ ) ∙ ( o+ e1 ) = 0 1 −1 =0
e3∙ e3 e3 ∙ ( e2 +∞ ) e 3 ∙ ( o+e 1 ) 1 0 0
Note that the tangent away from the origin may involve the infinity point, unlike at the origin.
15.5.20 Enough for now Conformal Model.

You might also like