3 D Geometric Algebra 1

Chapter 1
Three-dimensional geometric
algebra
1.1 Scalars
Scalars are quantities, which can be specified by one number. For example,
temperature, mass, and work are scalars. Often, scalars are identified with real
numbers (which, we assume, reader is already familiar with). However, we define
scalars implicitly by their arithmetic properties, which are relevant in the present
context. For any scalars α and β there exists an addition operation “+” which
produces another scalar α + β, and there exists a multiplication operation (for
which no special symbol is reserved), which creates another scalar αβ. Scalars α,
β and γ obey following rules:
1. The sum α + β + γ can be calculated in any order,
α + β + γ = (α + β) + γ = α + (β + γ) = (α + γ) + β (1.1)
i.e. it is both associative, distributive, and commutative1 .

1
Not all the sums containing infinite number of terms can be rearranged at will. Typical
example is the alternating harmonic series S = 1 − 12 + 13 − 14 + 15 − ..., which converges to
R2
ln 2. To justify this, note that ln 2 = 1 (1/x)dx. With the substitution x = 1 + t we obtain
R1
ln 2 = 0 1/(1 + t)dt. By expanding 1/(1 + t) into the Taylor series 1 − t + t2 − t3 + −... and
integrating term by term, the result ln 2 = S follows. However, if we regroup the terms in S
11
12 CHAPTER 1. THREE-DIMENSIONAL GEOMETRIC ALGEBRA
2. The product αβγ is associative i.e.
αβγ = (αβ)γ = α(βγ) (1.2)
3. The product is commutative, i.e. αβ = βα
4. The multiplication is associative, distributive and commutative with respect

to addition, i.e.
α(β + γ) = αβ + αγ = βα + γα = (β + γ)α
The advantage in defining scalars implicitly via addition and multiplication

operations is the possibility to extend the concept of scalars to include other
numbers as well besides the familiar reals. Such quantities are, for example, the
hyperreal and surreal numbers (see Appendix A). Of course, the answer to the
question “which numbers can be considered as scalars” depends now crucially
on the definition of addition and multiplication operations. The operations must
be chosen so that if α, β, and γ are real numbers these operations are identical
to the familiar addition and multiplication operations. Just as real numbers,
the extended set of scalars possesses an additive identity element zero: For all
scalars α + 0 = 0 + α = α. Every scalar α also possesses an additive inverse
−α such that α + (−α) = −α + α = 0. As a notable difference to conventional
treatments, a scalar α need not possess inverse, i.e. α−1 need not be defined.
Consequently, scalars form a ring, but not a field. Also, we do not require that
the scalars α and β necessarily satisfy the trichotomy law. In other words, the
statement (α < β) or (α > β) or (α = β) is not necessarily always true. The
relaxing of these conditions allow us to extend the concept of scalars to include
smooth infinitesimals (see Appendix A.1) as scalars.
Scalars possess a very important property, which comes into action only when
we consider quantities which are not scalars, namely
1
as (1 + 3 + 15 ) − 1
2 + ( 17 + 1
9 + 1
11 + 1
13 + 1
15 ) − 1
4 + ... the series S would appear to converge to
3
2. Also, if we would multiply S by 2, and rearrange the resulting series by adding together the
terms with the common denominator, we would obtain a series looking like the original one,
and we would be tempted to write S = 2S, where it would appear that S = 0, or S = ±∞.
1.2. VECTORS 13
Definition 1 (Commutation property) Scalars commute with all the elements

of geometric algebra.
The precise meaning and significance of this property becomes evident later
on. Because we prefer to identify scalars with numbers without direction, we also
require that
Definition 2 (Square of scalar is positive) The square of any scalar is pos-

itive or zero, i.e. α2 ≥ 0.
This definition serves to exclude complex numbers, to which a direction can be

assigned (for example, via the Argand diagram), from scalars. However, it does
not mean that quantities analogous to the common imaginary numbers would
not exist in geometric algebra. On the contrary: we shall learn that there exist
more than one type of unit imaginaries, i.e. unit quantities with square −1.
In addition, these elements and their combinations possess a definite geometric
interpretation.
1.2 Vectors
Vectors are quantities, whose specification requires a direction as well as a nu-
merical value. For example, displacement and velocity are vectors. A vector can
be pictured by a segment of a straight line (i.e. by an arrow) regarded as having
a definite length and direction. Vectors are not tied to any specific point. Any
two vectors with the same length and direction represent the same vector (see
Fig. 1.1). To emphasize this inherent vectorial property, one occasionally uses
the term free vector. The simplest interpretation of a vector is the displacement
from a point coinciding the tail of the vector to the point coinciding with the
head of the vector. We introduce the notation Γp (a) to signify that the vector
a is transported along the line passing through the point p and containing the
tail of the vector a without changing its orientation, until the tail of the vector a
coincides with the point p. The independency of vectors from any specific point
implies that the parallel transportation is independent of path, i.e.
Γp (a) = a (1.3)
Figure 1.1: Arrows representing the same vector.
for any point p. This, in turn, implies flat space. Furthermore, we can interpret
Eq. (1.3) describing vectors as an equivalence class of rays (i.e. point bound
vectors). If a vector is multiplied by the scalar λ, its length is dilated by the
factor |λ|. The direction also changes to opposite, if the scalar is negative.
1.3 Addition of vectors
The sum a + b of the vectors a and b is found by joining the tip of vector a to
the tail of b and drawing an arrow labelled a + b that connects the remaining
tail and tip. The triangle construction is depicted in Fig. 1.3. The subtraction
a − b of vectors is equal to the addition a + (−b). Because the same arrow
can be obtained by joining the tip of vector b to the tail of a, and drawing an
arrow labelled a + b that connects the remaining tail and tip, it follows that
a + b = b + a. Thus, the sum a + b can be represented as the diagonal of the
parallelogram, which has vectors a and b as its sides. This is depicted in Fig.
1.4. As Fig. 1.5 shows, a + b + c = (a + b) + c = a + (b + c). Thus, vector
addition is associative, distributive, and commutative.
1.3. ADDITION OF VECTORS 15
Figure 1.2: a and −a.
Figure 1.3: Vector addition through triangle construction.

Figure 1.4: Vector addition through parallelogram construction.
Figure 1.5: Associativity of vector addition.

1.4. PRODUCTS OF VECTORS 17
Through the vector addition rule,
a + b = Γp (a + b) (1.4)
wee see that in a flat space the parallelogram formed with four vectors in its sides
remains closed under transportation. In other words, the flat space is torsion
free. Of course, it is possible to define the concept of directed displacement also
in the curved spaces. However, the algebra of curvilinear displacements would
differ from the algebra of vectors. Consider the two-dimensional space consisting
of a surface of a sphere. It is possible to define spherical vectors as directed
segments of great circles. However, the addition of two spherical vectors is not
commutative. For further details, see Sect. 3.6.4.
1.4 Products of vectors

There is no single and obvious way to define multiplication of two vectors. In the
contemporary books of vector analysis, the student is introduced to the dot and
cross products, which are presented separately and seemingly are two different
entities (the result of the dot product a · b is a scalar and the result of the cross
product a × b is a vector). However, as we shall see, both of these products can
be derived from a more significant product, namely from the geometric product
ab.
The geometric product ab of two arbitrary vectors a and b possesses the
properties
1. It is distributive over addition, i.e.
a(b + c) = ab + ac (1.5)
2. It is associative, i.e.
abc = a(bc) = (ab)c (1.6)
3. In Euclidean space, the square of the vector a is equal to the square of its
length2 , i.e.
a2 = |a|2 ≥ 0 (1.7)
2
In other spaces the length of a vector is not necessarily well defined.
Figure 1.6: Inner product a · b.
Incidentally, these three properties are sufficient to define a unique product,

which fully specifies the relative orientation of the vector factors in the product.
The dot and cross products (or rather, a related product, the outer product) can
also be derived from the geometric product. We can reverse the problem, and
use these derived products to obtain the fundamental geometric product. In the
course of defining products, we will encounter objects, namely blades which are
higher order generalization of vectors.
1.4.1 Inner product between two vectors

By expanding the square of the sum of the two vectors we obtain
(a + b)2 = |a|2 + |b|2 + ab + ba (1.8)
where, by the Pythagorean theorem, we can write
|a + b|2 = |a|2 + |b|2 + 2a · b (1.9)
The a · b is the inner product (the “dot” product) of a and b. By inspecting

Eq. (1.9) it is evident that the product a · b is scalar valued and it is equal to
the projection of a onto b multiplied by the length of b (Fig. 1.6). We can get
outside of the realms of the conventional vector analysis by using Eqs. (1.8) and
(1.9) to define the dot product a · b of any two vectors a and b in terms of the
yet unknown product ab as
ab + ba
a·b= (1.10)
2
where it follows that the symmetrized combination of geometric product is scalar
valued.
1.4.2 Outer product of two vectors

By the analogy to the definition of the dot product in Eq. (1.10), we may define
the outer product a ∧ b as an antisymmetric part of the yet unknown product ab
as
ab − ba
a∧b= = −b ∧ a (1.11)
2
The entity a ∧ b cannot be a scalar, because it anticommutes with the vector a:
ab − ba |a|2 b − aba b |a|2 − aba ba2 − aba

a (a ∧ b) = a = = =
2 2 2 2
= (b ∧ a) a = − (a ∧ b) a (1.12)
Nor is a ∧ b a vector, because its square is negative, as seen by

2
(ab)2 − 2 |a|2 |b|2 + (ba)2

2 ab − ba
(a ∧ b) = =
2 4
(a · b + a ∧ b)2 − 2 |a|2 |b|2 + (a · b − a ∧ b)2
=
4
(a ∧ b) + (a · b) − |a|2 |b|2
2 2
= (1.13)
2
(because (a · b)2 ≤ |a|2 |b|2 , where the equality holds only for a which is collinear
with b), and its direction does not change when its vector factors a and b are
both multiplied by −1. It is a bivector, a new kind of entity. It can be pictured as
an oriented parallelogram with sides a and b (See Fig. 1.7). In Fig. 1.7, we have
labelled the vertices of the parallelogram representing the bivector a ∧ b. The
direction of the bivector is the direction of the path going through the vertices
ABCD (in this order), and it is indicated by the red arrow in the bivector in the
upper left corner in the figure. The same bivector could as well be pictured as
Figure 1.7: Bivector a ∧ b.
any other planar object with the same orientation and area: the particular shape
is unimportant (see Fig. 1.8). In other words, a bivector is an equivalence class
of planes, in the similar manner as a vector is an equivalence class of rays. If the
bivector is multiplied by the scalar λ, its area is dilated by the factor |λ|. The
orientation also changes to opposite, if the scalar is negative (or in other words,
−a ∧ b = b ∧ a). The sense of the bivector a ∧ b can be determined graphically
as the direction of the arc traversed by the tip of vector a, when the tail of a is
placed to the tail of b, and a is turned through the smaller of the two possible
sectors so that it coincides with b (see Fig. 1.9). We also say that the grade of
a bivector a ∧ b is two, since it is an outer product of two vector factors.
1.4.3 Geometric product of two vectors

We can combine the inner a · b and outer product a ∧ b for arbitrary vectors a
and b to a single geometric product
ab = a · b + a ∧ b (1.14)
We can as well regard a · b as the scalar part of the geometric product ab, i.e.
a · b = habi0̄ (1.15)
Figure 1.8: Different representations of a bivector B.
Figure 1.9: The direction of a bivector u ∧ v, indicated by the directed solid arc
(now u and v are unit vectors).
and
a ∧ b = habi2̄ (1.16)
as the bivector part of the geometric product ab. The notation hAik̄ stands for
the k-grade part of A, where an object of the grade k consistst of sums of outer
products of k vectors. Occasionally, they are referred as k-vectors.
In Eq. (1.14), a scalar a · b is added to a bivector a ∧ b, which at a first
sight may seem absurd. You are supposed to add only “like things”, what ever
that means. Of course, you are free to ask “does the sum in Eq. (1.14) exist in
reality?” However, as commented by an anonymous reviewer in www.amazon.com,
This is an entirely misguided question that will only confuse the

reader. All mathematical concepts are real within mathematics, and
do not exist (except as approximations) in the real world. It is a
worthwhile topic in the philosophy of mathematics, but it has noth-
ing whatsoever to do with mathematics per se.
This comment refers to the existence of fractals, but it applies also perfectly
to the present problem. The justification of Eq. (1.14) is its usefulness, which will
be amply demonstrated in what follows. Complex numbers encountered similar
objections when they were first introduced. It seemed absurd to add something
with a negative square to ordinary real number. However, this is hardly the
opinion of the mathematicians today.
Generally, the product ab is neither commutative nor anticommutative. Only
if a is perpendicular to b, then a · b = 03 , and it follows from Eq. (1.11) that for
any two perpendicular vectors ab = −ba. Also, if the vectors are perpendicular,
one can replace the outer product a ∧ b by the geometric product ab (or vice
versa). This often considerably simplifies the manipulation of formulas. Analo-
gously, only if a is parallel (or antiparallel) to b, then a ∧ b = 0, and it follows
from Eq. (1.10) that for any two parallel (or antiparallel) vectors ab = ba. Also,
if the vectors are parallel (or antiparallel), one can replace the dot product a · b
by the geometric product ab (or vice versa).
3
Unless a is a null vector, in which case aa = a · a = 0. As we shall see, null vectors play an
important role in geometry and calculus.
Figure 1.10: Invertibility of the geometric product
Example 1 Let a = u1 + 2u2 and b = u1 − u2 + u3 /2, where {u1 , u2 , u3 } is a

set of orthonormal vectors, i.e. u2i = 1 and ui uj = ui ∧ uj for i 6= j. Then
ab = (u1 + 2u2 )(u1 − u2 + u3 /2)

= 1 − u1 ∧ u2 + u1 ∧ u3 /2 + 2u2 ∧ u1 − 2 + u2 ∧ u3
= −1 − 3u1 ∧ u2 + u1 ∧ u3 /2 + u2 ∧ u3
Unlike the dot or the outer product, the geometric product ab is invertible,
i.e. any vector a possesses a unique inverse
a
a−1 = (1.17)
|a|2
so that aa−1 = a−1 a = 1. The invertibility of the geometric product and the un-
invertibility of the dot product is illustrated in Fig 1.10. Notice that while there
is infinite number of vectors with the same projection to the unit vector u, there
is only one bivector, which corresponds with the bivector part of the product
au. It picks up the vector a for any given projection. One can similarly show
that starting from any given bivector part with u as a factor, there corresponds
a unique projection a · uu of the vector a to the unit vector u.
1.4.4 Trivectors
We define the outer product of the vector a and the bivector b∧c as the symmetric
part of the geometric product a (b ∧ c) as
ab ∧ c + b ∧ ca
a ∧ (b ∧ c) = (1.18)
2
Since trivector is formed as an outer product of three vector factors, its grade
is three. As we shall see, the grade of the antisymmetric part of the geometric
product between a vector and a bivector is one, not three, which rules out the
possibility of using it to define the outer product between a vector and a bivec-
tor. The notation in Eq. (1.18) is unambiguous, since we use the precedence
convention, where the outer product has priority over geometric product.
Because three vector factors are included in the outer product, it is called a
trivector. The same trivector is also obtained, if we take the outer product of the
bivector a ∧ b with the vector c. Thus, the outer product is associative in each
of its vector factors, i.e.
a ∧ (b ∧ c) = (a ∧ b) ∧ c (1.19)
so the notation a∧b∧c is unambiguous. By using the associativity, we also easily

see that the trivector a ∧ b ∧ c is antisymmetric with respect to the interchange
of any its two adjacent vector factors, i.e.
a ∧ b ∧ c = −b ∧ a ∧ c = −a ∧ c ∧ b (1.20)
and invariant with respect to symmetric permutations, i.e.,
a∧b∧c=b∧c∧a=c∧a∧b (1.21)
The trivector a∧(b ∧ c) as the grade-3 part of the geometric product a (b ∧ c),
i.e,
a ∧ b ∧ c = ha (b ∧ c)i3̄ (1.22)
or equally, due to the associativity of outer product,
a ∧ b ∧ c = h(a ∧ b) ci3̄ (1.23)

Figure 1.11: Trivector a ∧ b ∧ c.
where we have used the parenthesis to clearly emphasize the factors involved in
the geometric product.
The trivector a ∧ b ∧ c can be pictured as an oriented parallelepiped with

sides a, b and c (see Fig. 1.11). If the trivector is multiplied by the scalar λ, its
volume is dilated by the factor |λ|. The orientation also changes to opposite, if
the scalar is negative. The outer product a1 ∧ a2 ∧ . . . ∧ ak is zero for k > 3 in
the three-dimensional Euclidean space E(3). In this space, any trivector can be
expressed as a multiple λi of a unit trivector i, so in the three-dimensional space
E(3) all trivectors are related by a scalar factor. Since the geometric product
of trivector with scalars, vectors, bivectors and other trivectors is commutative,
trivectors are called pseudoscalars of E(3). The unit trivector i = u1 ∧ u2 ∧ u3 is
of the unit magnitude, i.e the cube with u1 , u2 and u3 as sides has a unit volume
(now, as always, {u1 , u2 , u3 } is an orthonormal vector triplet). More formally
(as we shall see in Sect. 1.7) the magnitude of a unit trivector |i| refers to the
expression i† i = 1 = |i|, where the superscript dagger signifies the change of the
order of the vector factors. By further taking the account of the commutativity
of i with the other elements of E(3) (over geometric product) and the fact that
i2 = −1, in some sense i plays the role of a unit imaginary in E(3).
1.4.5 Contractions
So far we have been introduced to the dot products a·b, and to the outer products
b ∧ c, and b ∧ c ∧ d. This leads to the question if it is possible to define mixed
products, such as a·b∧c. As we shall see this is indeed possible. This generalizes
straightforwardly to the inner products between arbitrary multivectors A and B
P P
by the simple substitution A = k hAik̄ and B = l hBil̄ , where the notation
hAik̄ stands for the k-vector part of A as before. However, as a price to pay for
this generalization, we must somewhat modify the concept of the inner product.
Contraction of bivector by vector
What is the inner product between a vector a and a bivector B = b∧c? First, we
notice that a · b is the zero-grade part of the geometric product, i.e. a · b = habi0̄ .
Its grade is the difference of the grades of a and b. By extrapolating this property,
the inner product between a and B should be the part of the geometric product
aB, whose grade is the difference of the grades of a and B. However, the grade
of a is one, and the grade of B = b ∧ c is two, so their difference is either 1 or
−1. Because the grade is always ≥ 0, it is best to define two inner products,
namely the left contraction (or contraction onto) and the right contraction (or
contraction by). The left factor lowers the grade of the right factor in the left
contraction so that the grade of the result is equal to the grade of the blade on
the right of the contraction sign minus the grade of the blade on the left of the
contraction sign. Similarly, the right factor lowers the grade of the left factor
in the right contraction so that the grade of the result is equal to the grade of
the blade on the left of the contraction sign minus the grade of the blade on the
right of the contraction sign. The left contraction is the antisymmetric part of
the geometric product aB, i.e.
1
acB = haBi1̄ = (aB − Ba) (1.24)
2
and the right contraction is is the antisymmetric part of Ba, i.e.
1
Bba = hBai1̄ = (Ba − aB) (1.25)
2
Naturally, Bca = hBai−1 = 0 and abB = haBi−1 = 0, because the grade −1
does not exist. Notice that the tip of the contraction sign shows the blade that
contracts (or lowers the grade) of the blade on the other side of the sign. By
comparing Eqs. (1.24) and (1.25) we see that
acB = −Bba (1.26)
Furthermore, by taking the reverse of Eq. (1.24) it follows that (acB)† =

1
2
(Ba − aB) = Bba.
In order to evaluate the left contraction acB of a bivector B = b ∧ c with the
vector a in practice, we must express the contraction in terms of the products
which we already know how to evaluate. The dot products a · b, a · c and b · c
are already familiar to us, so we transform Eq. (1.24) to a form, which includes
only dot products between the vector factors. To meet this end, we first write the
contraction and outer products in terms of the geometric product. This results
in
bc − cb abc − bca − acb + cba
acB = ac = (1.27)
2 4
By adding
bac − cab − bac + cab
0= (1.28)
4
to this, we obtain
(ab + ba)c − (ac + ca)b − b(ac + ca) + c(ba + ab)
acB =
4
(a · b)c − (a · c)b − b(a · c) + c(a · b)
= (1.29)
2
where we obtain straightforwardly the formula
acb ∧ c = a · bc − a · cb (1.30)
This notation is unambiguous, since we follow the precedence convention, where

outer product has priority over contractions, which in turn have priority over
geometric product, so acb ∧ c ≡ ac(b ∧ c) and a · bc ≡ (a · b)c. Obviously,
acb ∧ c is a vector, which lies in the b ∧ c plane and is orthogonal to a. The
k k
bivector b ∧ c can be written as b ∧ c = bcub u⊥
b , where ub is the is a unit vector
in the direction of b and u⊥
b is the unit vector orthogonal to b in the plane b ∧ c.
Consequently, Eq. (1.30) can be written as
k k k k
acb ∧ c = bc(a · ub u⊥ ⊥ ⊥ ⊥
b − a · ub ub ) = bc(a · ub ub + a · ub ub )i (1.31)
k
where i = ub u⊥
b is the unit bivector in the plane b ∧ c. To understand the
geometric significance of the right-most form in Eq. (1.31), we note that if the
k
vectors ub and u⊥
b are post multiplied by i, the result
k
ub i = u⊥
b (1.32)
k
u⊥
b i = −ub (1.33)
This clearly indicates that the product xi of any vector x in the plane i is equal
to the vector obtained by rotating x by π/2 in the plane i. The direction of
rotation is given by the positive orientation of4 i. Thus it follows that acb ∧ c is
a vector, which can be obtained by rotating the orthogonal projection of a in the
plane b ∧ c by 1/4 turn (see Fig. 1.12) and dilating the result by bc.
By taking the advantage of Eqs. (1.26) and (1.30), we can express the right
contraction of b ∧ c by a as
b ∧ cba = a · cb − a · bc (1.34)
Geometrically, b ∧ cba is obtained by rotating the projection of a dilated by the

factor bc in the plane b ∧ c by 1/4 turn (in the plane) to the direction opposite
to that in acb ∧ c.
Contraction of trivector by vector
The left contraction of a unit trivector i by a vector a is obtained as a symmetric

part of the geometric product ai and the right contraction of i by a as the
symmetric part of ia, i.e.
1
aci = haii2̄ = (ai + ia) (1.35)
2
(and ica = 0 = abi). The product of this contraction is a bivector, as implied by
the first equality. Since the contraction is defined as the symmetric part of the
geometric product, it follows necessarily that
aci = iba (1.36)

4
As shown later, it is no coincidence that i acts as a generator of rotations in the same way
as the unit imaginary in the conventional complex analysis.
k
Figure 1.12: Contraction acb ∧ c. The term P̂i (a) stands for the part of a in the
plane i.
By a similar derivation to that of the contraction of a bivector by vector we can

prove that the left and right contraction of a trivector αi = b ∧ c ∧ d (where α
is some scalar) by a vector a can be written as
acb ∧ c ∧ d = a · bc ∧ d − a · cb ∧ d + a · db ∧ c = b ∧ c ∧ dba (1.37)
and these represent the bivector (plane) perpendicular to the vector a
Contraction of bivector by bivector
By further generalizing the contraction products we define contraction of bivector

B1 by a bivector B2 as the scalar part of their geometric product, since the
difference of their grades is zero, i.e.
B1 · B2 = hB1 B2 i0̄ (1.38)
The notation in Eq. (1.38) is indeed well-defined, since the left contraction of
a bivector with another bivector is equal to their right contraction. In this case
we favor dot as the symbol for contraction. By a simple derivation (which is
presented for more general case in Sect. 2.2.1) we can show that the contraction
of the bivector a ∧ b by the bivector c ∧ d can be expressed in terms of the inner

products between the vector factors as
a ∧ b · c ∧ d = a · db · c − a · cb · d (1.39)
Evidently, fthe inner product B1 · B2 of the bivectors B1 and B2 is commutative,

i.e.
B1 · B2 = B2 · B1 (1.40)
Contraction of trivector by bivector
We define the left contraction of a unit trivector i by a bivector B as the vector

part of their geometric product, i.e.
Bci = hBii0̄ (1.41)
since the the grade of a trivector minus the grade of a bivector is one. By a simple
derivation (which is presented for more general case in Sect. 2.2.1) we can show
that the left contraction of the bivector a ∧ b by the trivector c ∧ d ∧ e can be
expressed in terms of the inner products between the vector factors as
a ∧ bcc ∧ d ∧ e = (a ∧ b) · (c ∧ d)e − (a ∧ b) · (c ∧ e)d + (a ∧ b) · (d ∧ e)c (1.42)
It turns out that the right contraction produces identical result, i.e.
Bci = ibB (1.43)
Contraction of trivector by trivector
The contraction of a trivector αi by a trivector βi is defined as
(αi) · (βi) = hαiβii0̄ = −αβ (1.44)
The notation in Eq. (1.39) is indeed well-defined, since the left contraction of a
trivector with another trivector is equal to their right contraction. In this case
we favor dot as the symbol for contraction. By a simple derivation (which is
presented for more general case in Sect. 2.2.1) we can show that the contraction
of the trivectors a ∧ b ∧ c and d ∧ e ∧ f can be expressed in terms of the inner
products between the vector factors as
a∧b∧c·d∧e∧f = c·d(a∧b)·(e∧f )−c·e(a∧b)·(d∧f )+c·f (a∧b)·(d∧e) (1.45)

On the notation
As we have seen, the both the left and the right contractions of an r-vector by
another r-vector are equal in E(3). Thus in that case we favor the dot as a symbol
for contraction. Thus, for example, we denote a · b rather than abb or acb.
1.4.6 Geometric products and some special results

The reader has no doubt noticed that the geometric product between a vector a
and any other quantity A has so far been written as
aA = acA + a ∧ A (1.46)
Aa = Aba + A ∧ a (1.47)
where A can be a vector b, a bivector B or a trivector βi. What if A is a scalar

α? In that case the result should simply be a vector αa, so we can immediately
conclude that
αa = a ∧ α = α ∧ a (1.48)
What if we would contract some vector a by the scalar α? By following our line
of reasoning we choose αca = hαai1−0 = hαai1̄ = haαi1̄ . But then it follows that
αca ≡ αa = a ∧ α = α ∧ a (1.49)
which proves that the left contraction of a vector by a scalar equals to their outer
product, which in turn gives the usual result of the multiplication of a vector by
a scalar. Along a similar line of reasoning we can conclude that
αβ = αcβ = αbβ = α ∧ β = βcα = βbα = β ∧ α = βα (1.50)
for any two scalars α and β. The reader should note, that this also proves that
generally αA is not equal to αcA + α ∧ A. The same conclusion applies as well to
the geometric product AB where A is a bivector: the geometric product of two
bivectors A and B is not related by the formula analogous to Eq. (1.46), but it
can be written as
1
AB = A · B + [A, B] + A ∧ B (1.51)
2
of two bivectors A and B. The grade of A·B is zero and the grade of A∧B is four
(if it exist, i.e. if the dimension of space is at least 4). It turns out that the grade
of the commutator of A and B, namely the middle term [A, B] = AB − BA is
two (see Sect. 1.6).
Example 2 Let A = u1 ∧ u2 = u1 u2 and B = u3 ∧ u1 = u3 u1 . Then, AB

1
= u1 u2 u3 u1 = u2 u3 = u2 ∧ u3 . Because hABi0̄ = AcB, hABi2̄ = 2
[A, B],
and hABi4̄ = A ∧ B (= 0 in the three dimensional space), we may deduce (u1 ∧
u2 )c(u3 ∧ u1 ) = 0, [u1 ∧ u2 , u3 ∧ u1 ] = 2u2 ∧ u3 , and u1 ∧ u2 ∧ u3 ∧ u1 = 0.
(Of course, the geometric product between the vector a and the bivector B
can be written as aB = acB + a ∧ B and Ba = Bba + B ∧ a).
1.5 Angles and negative Euclidean metrics

The inner product a · b is directly related the concept of angle θ between a and
b. Let u and v be two unit vectors positioned so that their tails coincide. The
included angle θ is defined here (as in most of the textbooks in vector analysis and
geometry) as the arc length in radians (or twice the area) of the smaller circular
sector with u and v as sides (see Fig. 1.13). By this definition, θ is always in the
range 0 ≤ θ ≤ π. The inner product a · b between any a and b can be written
as
a · b = |a| |b| cos θ (1.52)
Similarly, a bivector a ∧ b can be written as
a ∧ b = i |a| |b| sin θ (1.53)
where i = a ∧ b/ |a ∧ b| is the unit bivector in the a ∧ b-plane (i.e. |i| = 1).

The unit bivector i determines the direction of the bivector, and the area (or
magnitude) is given by |a| |b| sin θ (note that 0 ≤ θ < π, so sin θ ≥ 0). By
combining the scalar and bivector parts, the geometric product in Eq. (1.14) can
be written as
ab = ab cos θ + iab sin θ (1.54)
To be precise, an included angle θ between vectors a and b does not refer
to any specific pair of vectors a and b, but it defines an equivalence class of
1.5. ANGLES AND NEGATIVE EUCLIDEAN METRICS 33
vector pairs through the relation |a| |b| cos θ = (Dc a) · (Dc b), where Dc x signifies
rotation of the vector x followed by a translation c. Thus, an angle forms an
equivalence class of circular sectors. For any given sector there are infinite many
other sectors, which are obtained by rigidly displacing the given sector, and they
correspond to the same included angle.
The angle convention followed in the present treatment is not the only one.
Let u and v be two unit vectors, whose tails coincide. An angle, denoted now
by t, could be defined as the signed length (in radians) of the circular arc from
the head of u to the head of v. The sign is positive, if the direction of the arc
is the same as the sense of some given bivector j, which lies in the plane of the
angle and is used to define the positive sense of the angle t (so i = u ∧ v/|u ∧ v|
is either parallel or antiparallel compared to j, i.e. i = ±j). Otherwise the angle
t is negative (note that by this definition, j sin t = i sin θ and cos t = cos θ; See
Exercise 36). If |t| ≤ π, |t| is the included angle θ. If |t| > π, it is related to the
included angle θ as θ = 2π − |t|. In an influential textbook on the application of
geometric algebra to classical mechanics (Ref. [32]) a further step is taken and
an angle is defined as a bivector tj.
Even by sticking with the more conventional choice of the angle as an included
angle there are four different ways to define an included angle θ between vectors
u and v as a equivalence class of sectors. Usually, it is taken for granted that the
angle between vectors u and v is measured about the vertex, where the tail of
vector u and the tail of vector v coincide. This is the TT sector depicted in Fig.
1.14 (where for simplicity it is assumed that u and v are unit vectors). Also, the
direction of the unit bivector i = u ∧ v/ sin θ can be defined as the direction of
the sector, which is the direction where u turns when the angle is closed (and
depicted as a red arrow in the figure). However, as shown in the picture, there
are three other sectors, which could be used to define the included angle, namely
head-to-head (HH), head-to-tail (HT) and tail-to-head (TH) sectors. The angle
defined via HH sector does not differ from the angle defined as TT sector, nor
does the direction of the bivector i. However, in HT and TH sectors the included
angle is supplementary to the included angle defined as TT or HH sector, i.e.
θ(HT or TH) = π − θ(TT or HH) (1.55)

Figure 1.13: Two possible angle conventions. In Fig. (a), the angle between u
and v is defined as the included angle θ. In Fig. (b), the angle between u and v
is defined as the directed angle t. The positive direction of t is indicated by the
the arrow in the arc. It is opposite to the direction of u ∧ v.
The formal rules for adding and multiplying vectors do not depend on the angle
convention. Because sin (π − θ) = sin θ and cos (π − θ) = − cos θ, the geometric
product of any two vectors a and b reads as
ab = |a| |b| (η cos θ + sin θi) (1.56)
where the signature η is +1 for TT or HH angle convention, and −1 for HT or

TH angle convention. By choosing a = b or θ = 0, we see that
a2 = − |a|2 (1.57)
This implies negative Euclidean metrics. The sense of the bivector i is in TH

or HT angle convention is opposite to the sense of the bivector i in TT or HH
convention. Depending on the sector convention used, there are four different
1.5. ANGLES AND NEGATIVE EUCLIDEAN METRICS 35
Figure 1.14: Angles as sectors.
rules for addition operation, namely
−→ −−→ −−→ −−→ −→ −−→

∠(AB, CB)HH + ∠(CB, DB)HH = ∠(AB, DB)HH (1.58)
−→ −−→ −−→ −−→ −→ −−→
∠(AB, BC)HT + ∠(CB, BD)HT = ∠(AB, BD)HT (1.59)
−→ −→ −→ −−→ −→ −−→
∠(AB, CA)TH + ∠(AC, DA)TH = ∠(AB, DA)TH (1.60)
−→ −→ −→ −−→ −→ −−→
∠(AB, AC)TT + ∠(AC, AD)TT = ∠(AB, AD)TT (1.61)
−→
where AB stands for the (point bound) vector, whose tail is at the point A and
−→ −−→ −→
head at B, and ∠(AB, CB) stands for the sector spanned by the vectors AB and
−−→
CB. The subscripts imply how the sides of the sectors are joined together. For
−→ −−→ −→ −−→
example, ∠(AB, CB)HT means that the head of AB is joined to the tail of CB.
Because the addition of simple bivectors can be identified with the addition of
sectors (any simple bivector can be decomposed as outer product of two vector
factors, which can be parallelly transported to any suitable point), it follows that
there exists also four possibilities to add bivectors. The particular one chosen
depends on the sector convention followed. Unless stated otherwise, we follow
exclusively the TT convention.
Figure 1.15: Bivector a ∧ b and the cross product a × b.
1.6 Duality and cross product

The familiar vector cross product a × b is related to the bivector a ∧ b as
a × b = −i (a ∧ b) (1.62)
where a × b is a vector perpendicular to the plane a ∧ b (see Fig. 1.15). To derive

the above identity, note that the unit trivector i can be factored as i = ua×b ∧ a ∧
b/ (ab sin θ) = ua×b a∧b/ (ab sin θ), where ua×b is the unit vector in the direction
of a×b, and θ is the angle between a and b. By using this decomposition of i, the
right hand side of Eq. (1.62) reads as −i (a ∧ b) = −ua×b a ∧ b (a ∧ b) / (ab sin θ)
= ab sin θua×b , where the left hand side follows.
The unit trivector can also be used to switch between the wedge and the
contraction symbol as
i(a ∧ b) = ac(ib) = b × a (1.63)
This equality can be extracted by separating the vector part from the simple
identity
iab = aib (1.64)
1.6. DUALITY AND CROSS PRODUCT 37
From the trivector part of Eq. (1.64), we obtain another useful identity relating
dot and outer products, namely
ia · b = a ∧ (ib) (1.65)
With the help of this duality, where a quantity is obtained from another
quantity by multiplying it with the unit trivector i (this is generalized to N -
dimensional spaces in Sect. 2.5), it is simple to obtain identities including cross
products of several vector factors. For example, by using Eq. (1.63), it is possible
to write the triple cross product as
a × (b × c) = −ia ∧ (b × c) = −ac(ib × c) = −ac(b ∧ c) (1.66)
By further applying Eq. (1.30), the familiar identity
a × (b × c) = a · cb − a · bc (1.67)
follows. By using a × (b × c) = −(b × c) × a, and cyclically interchanging the

vectors (i.e. by substituting b = a, c = b and a = c) we see that
(a × b) × c = a · cb − b · ca (1.68)
which clearly is not always equal to a × (b × c). Thus, we see that the cross-
product is not associative. More identities can be created from Eq. (1.67) by
substituting cross products to one (or more) of its vector factors. For example,
the identity
(a × b) × (c × d) = (a × b) · dc − (a × b) · cd (1.69)
can be obtained straightforwardly by replacing a by a × b, b by c and c by d.

It is also possible to relate the box product a · b × c to the trivector a ∧ b ∧ c by
writing a ∧ (b ∧ c) = −a ∧ (ib × c) and then using Eq. (1.65) to switch between
the wedge and dot. This produces the valuable identity
a ∧ b ∧ c = −ia · b × c (1.70)
The grade of any blade in a three-dimensional space is equal or smaller than

3. By using the duality of vectors and bivectors, scalars and trivectors, any
multivector A in a three-dimensional space can be written as
A = α + iβ + a + ib (1.71)
where α = hAi0̄ is the scalar part of A, a = hAi1̄ is the vector part of A, ib = hAi2̄
is the bivector part of A, and iβ = hAi3̄ is the trivector part of A. Thus, the
geometric product A = A1 A2 of two multivectors A1 = α1 + iβ1 + a1 + ib1 and
A2 = α2 + iβ2 + a2 + ib2 contains the scalar part
α = α1 α2 − β1 β2 + a1 · a2 − b1 · b2 (1.72)
the vector part
a = α1 a2 + α2 a1 − β1 b2 − β2 b1 + i (a1 ∧ b2 − a2 ∧ b1 ) (1.73)
the bivector part
ib = a1 ∧ a2 + i(α1 b2 + α2 b1 ) − b1 ∧ b2 (1.74)
and the trivector part
iβ = i(α1 β2 + α2 β1 − a1 · b2 − b1 · a2 ) (1.75)
The last term −b1 ∧ b2 in the bivector part [Eq. (1.74)] is the bivector part of
the geometric product (ib1 )(ib2 ), i.e. −b1 ∧ b2 = h(ib1 )(ib2 )i2̄ . This also shows
that the commutator of any two bivectors is also a bivector, because [ib1 , ib2 ] =
−[b1 , b2 ]/2 = −b1 ∧ b2 .
1.6.1 Axial vectors

It is often claimed that the cross product a × b is invariant with respect to the
reversion of all directions in the space, which maps every vector a to
a∗ = − a (1.76)
Consequently, it is stated that there exist two kinds of vectors, namely ordinary
“polar” vectors, which obey Eq. (1.76), and “axial” vectors which are cross-
products of polar vectors obeying (a × b)∗ = a∗ × b∗ = (−a) × (−b) = a × b.
However, this confusing admittance of two kinds of vectors is unnecessary, since
an axial vector is nothing more than a bivector disguised as an vector. We see
at once that (a × b)∗ = (−ia ∧ b)∗ = − (i)∗ a∗ ∧ b∗ = −a × b, which clearly
shows that a × b is an ordinary vector. As a side note, even in the conventional
1.7. INDEMPOTENTS, INVERSES AND PITFALLS 39
textbooks the inconsistency in the behavior of cross product under reversion of

directions can be remedied by simply defining (a × b)∗ = b∗ ×a∗ (which amounts
in writing f in terms of geometric products before reversing the directions to get
f ∗ ). This removes the illogical practice of using “axial” cross products as basis
vectors in expanding “polar” vectors, and it is in harmony with the fact that the
right hand triplets {a, b, c} change to the left hand triplets {−a, −b, −c} under
the reversion of all directions in space (now, we can as well let c = a × b).
1.7 Indempotents, inverses and pitfalls

Although geometric product reminds the product between real numbers in being
associative, i.e.
ABC = (AB)C = A(BC) (1.77)
and distributive, i.e.
A(B + C) = AB + AC (1.78)
for any A, B and C, one should not assume that the zero and cancellation rules of
real numbers apply for arbitrary multivectors, no more than one should assume
commutativity of the geometric product. On the contrary, the equation
AB = AC (1.79)
does not necessarily imply that B = C (see the Exercise 2). The invalidity of the
zero rule is obtained as a special case of the above: the equation
AB = 0 (1.80)
does not necessarily imply that either A or B is zero (see the Exercise 3). One
can, though, conclude that A is zero, if AB is zero for an arbitrary B.
Because the cancellation rule (AB = AC ⇔ B = C) is not always valid,
it follows that not all multivectors possess inverse. Especially if A2 = A, the
multivector A does not possess an inverse. This can be seen by writing C = AB
so AC = A2 B = AB. If AA−1 = 1, then B = C, which is a contradiction if B
differs from C. Also, because geometric product is not commutative, the the left
multiplicative inverse L−1 of the multivector A defined via
L−1 A = 1 (1.81)
is not necessarily equal to the right multiplicative inverse R−1 , which is defined
via
AR−1 = 1 (1.82)
Luckily, the inverse A−1

k̄
exists for a simple blade5 Ak̄ = a1 ∧ a2 ∧ ... ∧ ak . The
inverse is simultaneously left and right inverse, i.e.
A−1
k̄
Ak̄ = Ak̄ A−1
k̄
=1 (1.83)
and it is obtained as
A†k̄
A−1 = (1.84)
k̄
|Ak̄ |2
where A†k̄ is the reverse of Ak̄ , i.e.,
A†k̄ = (a1 ∧ a2 ∧ . . . ∧ ak )† = ak ∧ ak−1 ∧ . . . ∧ a1 = (−1)k(k−1)/2 Ak̄ (1.85)
and |Ak̄ |2 = A†k̄ ·Ak̄ is the square of the magnitude of a k-blade Ak̄ . The magnitude
of a multivector A is analogously given as the scalar part of the product of A
with its reverse, i.e.
N
X
|A|2 = A† A 0̄ = |hAik̄ |2

(1.86)
k=0
Example 3 Let A = 1 + u1 + 2u2 + 7u1 ∧ u2 + 5i, where u1 and u2 are two

orthonormal vectors (i.e. u21 = u22 = 1 and u1 · u2 = 0). Then
|A|2 = 12 + 12 + 22 + 72 + 52 = 80
1.8 Vector equations

Using geometric algebra, we may solve vector equations directly instead solving
the set of scalar equations for the components of the vectors, as done in the
conventional vector analysis. Consider the following simple equation
αx + ax · b = e (1.87)
5
An r-blade is a multivector, which can be factored as a outer product of r vectors. Note
however that generally the r-vector part hAir̄ of siome multivector A can be a sum of r-blades.
For example, in four-dimensional Euclidean space the bivector u1 ∧ u2 + u3 ∧ u4 can not be
written as a (simple) blade.
1.8. VECTOR EQUATIONS 41
where the unknown x and the parameters a, b and e are vectors, and the pa-
rameter α is a scalar. If both sides of Eq. (1.87) are dotted with vector b,
e·b
x·b= (1.88)
α+a·b
follows. By inserting this to Eq. (1.87), the unknown vector x can be easily
solved as
e e · ba
x= − (1.89)
α α (α + a · b)
This solution can be easily generalized for more complicated cases, such as
αx + a1 x · b1 + a2 x · b2 = c (1.90)
(where a1 , b1 , a2 , b2 and c are vectors). If both sides of Eq. (1.90) are dotted
with vector b2 , the equation
(α + a2 · b2 ) x · b2 + a1 · b2 x · b1 = c · b2 (1.91)
follows, and x · b2 can be solved as

c · b2 − a1 · b2 x · b1
x · b2 = (1.92)
α + a2 · b2
By inserting this to Eq. (1.90),

a1 · b2 c · b2
αx + a1 − a2 x · b1 = c − a2 (1.93)
α + a2 · b2 α + a2 · b2
follows. But this is identical to Eq. (1.87) with
a1 · b2 c · b2
a = a1 − a2 b = b1 e = c − a2 (1.94)
α + a2 · b2 α + a2 · b2
and the solution is given by Eq. (1.89). This produces

c·b2 c·b2 a1 ·b2
c − a2 α+a 2 ·b2
(c · b1 − a2 · b1 α+a 2 ·b2
)(a1 − a2 α+a )
x= − 2 ·b2 (1.95)
α α α + a1 · b1 − a2 · b1 α+a a1 ·b2
2 ·b2
By a very similar procedure, we may solve the equation
αx + x × d = c (1.96)
where the unknown x and the parameters c and d are vectors, and the parameter
α is a scalar. The solution to Eq. (1.96) is obtained by first multiplying the both
sides of Eq. (1.96) with the unit trivector i, which produces
αix + x ∧ d = ic (1.97)
and then taking the outer product of them (from right) by d, which gives α(ix) ∧
d = (ic) ∧ d or
c·d
x·d= (1.98)
α
where Eq. (1.63) was used to switch between dot and wedge. By using this and
x ∧ d = xd − x · d, Eq. (1.96) can be written as
c·d
x(αi + d) = + ic (1.99)
α
where the solution
c · dd + αd × c + α2 c
x= (1.100)
α(α2 + d2 )
follows by multiplying both sides (from right) with (αi+d)−1 = (d−αi)/(α2 +d2 ).
The interested reader can easily verify the correctness of the inverse by performing
the multiplication (αi + d)(αi + d)−1 (or equaivalently, (αi + d)−1 (αi + d)).
1.9 Exponential function

The exponential function exp(A) of a multivector argument A is defined via the
usual infinite power series
∞
A2 X Ak
exp(A) = 1 + A + + ... = (1.101)
2 k=1
k!
It is easy to show that this series is absolutely convergent for all multivectors A,
which obey A2 = λA or A2 = λ, where λ is a finite scalar. It is more complicated
to show that the result also converges for any finite multivector A (i.e. for a
multivector A with finite magnitude |A|). From now on we simply regard Eq.
(1.101) as an algebraic definition for the exponential function.
Consider next the product exp(A) exp(B). If A and B commute, then
exp(A) exp(B) = exp(A + B) (1.102)

1.9. EXPONENTIAL FUNCTION 43
The left-hand side of Eq. (1.102) is referred as the disentangled, and the right-
hand side of Eq. (1.102) is referred as the entangled form of the product of
exponentials. The reader can easily verify Eq. (1.102) by using the binomial
expansion formula
l l
l
X l! X l!
(A + B) = Al−k B k = Ak B l−k (1.103)
k=1
k!(l − k)! k=1
k!(l − k)!
which is valid for mutually commuting A and B. Contrary to the common belief,
the mutual commutativity is sufficient but not necessary condition for the validity
of Eq. (1.102). For example, let A = αiu and B = αiv, where the angle between
the unit vectors u and v is 2π/3. If α is any multiple of 2π, then Eq. (1.102) is
satisfied.
If A and B do not commute, it is more difficult to obtain formulas analogous
to that in Eq. (1.102). Some of them, however, can be found with relative little
effort. Consider the case, where the commutator
[A, B] ≡ AB − BA (1.104)
commutes with both A and B. The direct expansion of exp(A + B) produces

(A + B)(A + B)
exp(A + B) = 1 + (A + B) + + ... (1.105)
2
By writing BA = AB − [A, B], the expansion can be re-written as
A2 + 2AB − [A, B] + B 2
exp(A + B) = 1 + A + B + + ... (1.106)
2
which, in turn, implies the disentanglement formula
[A, B]
exp(A + B) = exp(A) exp(B) exp(− ) (1.107)
2
This result can be proven rigorously via parametric differentiation:
Proof. Define
f (λ) = exp(λA) exp(λB)
where λ is a scalar parameter. Then,

∂f (λ)
= A exp(λA) exp(λB) + exp(λA)B exp(λB)
∂λ
= [A + exp(λA)B exp(−λA)]f (λ)
Since now [B, Ak ] = kAk−1 [B, A] (see Exercise 23), it follows that
∞ ∞
X λk [B, A]k X λk−1 Ak−1 [B, A]
[B, exp(−λA)] = (−1)k = −λ (−1)k−1
k=1
k! k=1
(k − 1)!
= −λ exp(−λA)[B, A]
Thus, exp(λA)B exp(−λA) = B − λ[B, A] and

∂f (λ)
= (A + B + λ[A, B])f (λ)
∂λ
The solution to this differential equation subjected to the initial condition f (0) =
1 is f (λ) = exp[λ(A + B)] exp(λ2 [A, B]/2). By setting λ = 1, the result in Eq.
(1.107) follows.
Sometimes one can guess the functional form f and g in the entangled expo-
nential function U = exp(f (A) + g(B)) and the functional form f 0 and g 0 in the
disentangled exponential function U = exp(f 0 (A)) exp(g 0 (B)). Then the disen-
tanglement formulas can be found (i.e. unknown parameters fixed in f , g, f 0 and
g 0 ) from the pair of equations involving A and B:
U AU −1 = V AV −1 (1.108)
U BU −1 = V BV −1 (1.109)
The similarity transformations can be evaluated via the Baker-Hausdorff lemma

∞
X [A, B] n
exp(A)B exp(−A) = (1.110)
n=0
n!
where the repeated commutator is defined as

[A, B]n = A, [A, B]n−1 (1.111)
and [A, B]0 = B. This method is illustrated in the next example:
Example 4 Let [A, B] = B. We guess that the disentangled exponential corre-

sponding to the entangled form U = exp(αA+βB) is of the form V = exp(γA) exp(κB),
where we need to determine the unknown scalar coefficients γ and κ as the func-
tions of known scalar coefficients α and β (or vice versa). Because
[αA + βB, γA + κB]0 = γA + κB

[αA + βB, γA + κB]n = αn−1 (ακ − βγ)B for n = 1, 2, ...
it follows using Eq. (1.110) that

∞
βB X αn β βB β
U AU −1 = A − =A+ B− exp α = A + (1 − exp α)B
α n=1 n! α α α
∞
−1
X αn
U BU = B = B exp α
n=1
n!
and
V AV −1 = exp(γA) exp(κB)A exp(−κB) exp(−γA) = exp(γA)(A − κB) exp(−γA)

∞
X γn
= A − exp(γA)κB exp(−γA) = A − κB = A − κB exp γ
n=0
n!
V BV −1 = exp(γA) exp(κB)B exp(−κB) exp(−γA) = exp(γA)B exp(−γA)

∞
X γ nB
= = B exp γ
n=0
n!
By equating the two sets of transformations, we obtain

β
(1 − exp α) = −κ exp γ
α
exp α = exp γ
where we can easily solve γ = α, κ = β(1 − exp α)/(α exp α).
The hyperbolic sine and cosine functions are also defined via the usual series
expansion
∞
A2 A4 X A2k
cosh A = 1 + + = (1.112)
2 4! k=0
(2k)!
∞
A3 A5 X A2k+1
sinh A = A + + = (1.113)
3! 5! k=0
(2k + 1)!
The hyperbolic cosine cosh A is the even, and sinh A is the odd part of exp A, i.e.
exp A = cosh A + sinh A (1.114)
By utilizing the parity of the hyperbolic functions, it is evident that they can
be obtained as symmetric and antisymmetric combinations of the exponential
function, i.e.
exp A + exp(−A)
cosh A = (1.115)
2
exp A − exp(−A)
sinh A = (1.116)
2
The trigonometric functions cosine and sine are defined (as usual)
∞
A2 A4 A 4 X A2k
cos A = 1 − + − = (−1)k (1.117)
2 4! 6! k=0
(2k)!
∞
A 3 A5 A7 X A2k+1
sin A = A − + − = (−1)k (1.118)
3! 5! 7! k=0
(2k + 1)!
If I is a multivector with I 2 = −1 and IA = AI, we can easily prove using the

series expansions in Eqs. (1.112), (1.113), (1.117) and (1.118) that
cosh(IA) = cos A (1.119)

sinh(IA) = I sin A (1.120)
so
exp(IA) = cos A + I sin A (1.121)
By comparing Eq. (1.121) to Eq. (1.54) we see that a spinor (i.e. geometric
product of two vector factors) z = ab can be written in an exponential form
z = ab = |a| |b| exp(iθ) = exp(λ) exp(iθ) = exp(λ + iθ) (1.122)
where θ is the angle between the vectors a and b, i is the unit bivector in the
k k
plane a ∧ b (i.e. i = a ∧ b/ |a ∧ b| ≡ ua u⊥
a where ua is a unit vector in the
direction of a, and u⊥
a is a unit vector orthogonal to a in the plane a ∧ b), and
λ is a positive scalar given by exp λ = |a| |b|. Since the effect of postmultiplying
the vector a by z is
az = |a|2 b (1.123)
we see that z rotates a by an angle θ to the positive direction of i and dilates its
length by the factor of |a|2 . And vice versa, given the rotation plane i, the angle
θ from the unit vector u to another unit vector v, we can rotate the unit vector
u to another unit vector v by post multiplying it with exp(iθ), i.e.
v = u exp(iθ) (1.124)
Figure 1.16: Rotation of vector u in the plane i by an angle θ as an infinite sum

of vectors.
where u ∧ v/ |u ∧ v|. Incidentally, since the postmultiplication of any vector in

the plane i rotates it in the plane by π/2, we see by expanding the right-hand
side of Eq. (1.124) that the infinite sum

θ θ θ
u + θui + (θui) i + (θui) i i + ... (1.125)
2 3 2
consisting of the vectors, from which the nth vector is always perpendicular to the
(n − 1)th vector and its length is obtained by multiplying the length of (n − 1)th
vector by the factor θ/n, always converges to the vector v (see Fig. 1.16).
By writing spinors S1 = ab and S2 = cd as S1 = |a| |b| (α1 + iβ 1 ) and
S2 = |c| |d| (α2 + iβ 2 ), where αk = cos θk is a scalar and β k = −ijk sin θk a vector
(jk is a unit bivector of the plane hSk i2̄ ), we can easily verify the combination
rule
S = S1 S2 = |a| |b| |c| |d| (α + iβ) (1.126)
where
α = α1 α2 − β 1 · β 2 (1.127)
β = α1 β 2 + α2 β 1 + β 2 × β 1 (1.128)
By taking the advantage of Eq. (1.122), this implies the exponential disentangle-
ment formula
exp(λ1 + j1 θ1 ) exp(λ2 + j2 θ2 ) = exp(λ1 + λ2 + iθuβ ) (1.129)
Now, the scalars λ1 , λ2 are positive, uβ is the unit vector in the direction of
the vector cos θ1 v2 + cos θ2 v1 + sin θ1 sin θ2 v1 × v2 and the angle θ is given by
cos θ = cos θ1 cos θ2 − sin θ1 sin θ2 v1 · v2 , where vk = −ijk .
1.10 Dihedral angles

The sine of the angle φijk between a unit vector vi and a plane vj ∧ vk is given
by
sin φijk = −ivi ∧ ijk (1.130)
where ijk is the unit bivector in the plane vj ∧ vk . This is part of a more
fundamental relation
vi ijk = uφijk cos φijk + i sin φijk = uφijk exp iφijk (1.131)
where uφijk = φijk /φijk = vi cijk / cos φijk is the unit vector in the plane ijk , which
k
is orthogonal both to the vector vi and its projection P̂ijk (vi ) in the plane ijk
(see the text after Eq. (1.30) in Sect. 1.4.5). The vector uφijk specifies the axis
of rotation, about which the vector vi can be rotated to the plane ijk . The angle
φijk (or magnitude of the vector φ) is uniquely determined by Eq. (1.131), and
it lies in the range 0 ≤ φijk ≤ 2π. The sign of sin φijk is positive (negative) if
vi ∧ ijk is right-handed (left-handed). It should be emphasized that Eq. (1.130)
is not sufficient to determine φijk uniquely (because the angle φijk varies in the
interval of the size 2π, but the function arcsin x returns values in the interval of
the size π).
The cosine of the dihedral angle τijk between planes vi ∧ vj and vj ∧ vk is
given by
cos τijk = iij · ijk (1.132)
Note that the orientation of the plane iij is opposite of the orientation of ijk ,
when the two planes coincide. By using Laplace’s rule in Eq. (2.24), it is evident
1.10. DIHEDRAL ANGLES 49
Figure 1.17: Angle φijk between vector vi and plane ijk .
that the cosine of the dihedral angle can be written in terms of the angles θij
between the vectors vi and vj as
cos θik − cos θjk cos θij
cos τijk = (1.133)
sin θij sin θjk
The bivectors iij and ijk can be written as iij = vi⊥ ∧ vj = vi⊥ vj and ijk =
vj ∧ vk⊥ = vj vk⊥ , where vi⊥ (vk⊥ ) is the component of vi (vk ) orthogonal to vj .
Thus, the product of the bivectors iij and ijk can be written in terms of the
dihedral angle τijk as
iij ijk = vi⊥ vk⊥ = cos τijk − ivj sin τijk (1.134)
where the minus sign in follows from the choice of vk⊥ , vi⊥ , vj as right-handed

triplet (see Fig. 1.18). Obviously, Eq. (1.134) can as well be written in the
exponential form as
iij ijk = exp (ivj τijk ) (1.135)
Eq. (1.135) is the definition of the dihedral angle: it determines uniquely both
the magnitude τijk = |ivj τijk | (0 ≤ τijk ≤ 2π) and the orientation of the plane
ivj . It is specifically emphasized that this does not apply to Eq. (1.132). It
can not be used to determine the dihedral angle, because the function arccos x is
restricted to the range 0 ≤ x ≤ π.
Figure 1.18: Dihedral angle τijk .
Let vi , vj , and vk be three unit vectors, and θij the angle between vi and
vj , θjk the angle between vj and vk , and θik the angle between vk and vi . Let
furthermore τijk be the dihedral angle between the planes vi ∧ vj and vj ∧ vk , τjik
the dihedral angle between the planes vj ∧ vi and vi ∧ vk , and τikj the dihedral
angle between the planes vi ∧ vk and vk ∧ vj .
By taking the geometric products of the unit vectors vi , vj , and vk as
(vi vj ) (vj vk ) (vk vi ) = 1 (1.136)
we obtain
exp (θij iij ) exp (θjk ijk ) exp (θki iki ) = 1 (1.137)
where many (scalar) relations of spherical trigonometry can be read off. For
example, by solving
− exp (θki iki ) = exp (θij iij ) exp (θjk ijk ) (1.138)
and rewriting the above equality in Euler form, we get
cos θik − iki sin θik = (cos θij + iij sin θij ) (cos θjk + ijk sin θjk )
= cos θij cos θjk + iij ijk sin θij sin θjk
+ijk sin θjk cos θij + iij cos θjk sin θij (1.139)
1.11. BASIS REPRESENTATION 51
By Eq. (1.134) this reads as
cos θik − iki sin θik = cos θij cos θjk + cos τijk sin θij sin θjk
+ijk sin θjk cos θij + iij cos θjk sin θij
−ivj sin τijk sin θij sin θjk (1.140)
where we may extract from the scalar part a standard relation of spherical
trigonometry:
cos θik = cos θij cos θjk + cos τijk sin θij sin θjk (1.141)
Other similar relations are easily found by cyclically permuting the indices i, j, k.
From the bivector part we obtain
iki sin θik = ivj sin τijk sin θij sin θjk
−ijk sin θjk cos θij − iij cos θjk sin θij (1.142)
which has no counter part in the conventional spherical trigonometry. By using

Laplace’s expansion rule in Eq. (2.24), one obtains
|vi ∧ vj ∧ vk |2 = 1 − cos2 θij − cos2 θjk − cos2 θik + 2 cos θij cos θjk cos θik (1.143)
One can use dihedral angles τijk to relate

vi ∧ vj ∧ vk
= sin θjk sin θik sin τikj = sin θjk sin θij sin τijk
i
= sin θik sin θij sin τjik (1.144)
and one can use angles φijk between vectors and planes to obtain
vi ∧ vj ∧ vk
= sin φijk sin θjk = sin φjki sin θik = sin φkij sin θij (1.145)
i
1.11 Basis representation
1.11.1 Orthonormal frame

There are only three linearly independent orthonormal vectors {u1 , u2 , u3 }. Any
vector a can be expanded in this orthonormal vector basis as
a = a1 u1 + a2 u2 + a3 u3 (1.146)
where ai = a · ui is the scalar component of the bivector a in ui . Similarly, there

are only three linearly independent bivectors {i1 , i2 , i3 }. These bivectors can be
represented as duals of {u1 , u2 , u3 } as
i1 = u2 u3 = iu1 (1.147)
i2 = u3 u1 = iu2 (1.148)
i3 = u1 u2 = iu3 (1.149)
where the set {i1 , i2 , i3 } is right-handed. Any bivector B can be expanded in this
bivector basis as
B = B1 i1 + B2 i2 + B3 i3 (1.150)
where Bi = B · i†i is the scalar component of the bivector B in ii . By using the

duality B = ib, it is evident that the components of B and b are the same, i.e.
Bj = bj . By including the remaining basis elements, namely the scalar unit 1,
and the unit trivector i, we may present the multiplication table of the elements
of a three-dimensional geometric algebra as
1 u1 u2 u3 i1 i2 i3 i
1 1 u1 u2 u3 i1 i2 i3 i
u1 u1 1 i3 −i2 i −u3 u2 i1
u2 u2 −i3 1 i1 u3 i −u1 i2
u3 u3 i2 −i1 1 −u2 u1 i i3 (1.151)
i1 i1 i −u3 u2 −1 −i3 i2 −u1
i2 i2 u3 i −u1 i3 −1 −i1 −u2
i3 i3 −u2 u1 i −i2 i1 −1 −u3
i i i1 i2 i3 −u1 −u2 −u3 −1
1.11.2 Oblique frame

Generally, any vector a can be represented as a linear combination
a = a(1) g1 + a(2) g2 + a(3) g3 (1.152)
where
a(i) = a · g(i) (1.153)
1.11. BASIS REPRESENTATION 53
is the contravariant component of the vector a in the frame {g1 , g2 , g3 }. The

reciprocal frame {g(1) , g(2) , g(3) } is obtained as
gj ∧ gk gj × gk
g(i) = = (1.154)
g1 ∧ g2 ∧ g3 g
where g = −ig1 ∧g2 ∧g3 = g1 ·g2 ×g3 is the determinant of the frame {g1 , g2 , g3 }.
If the determinant is positive (negative) the frame is right (left) handed. The basis
vectors are related to their reciprocals as
g(i) · gj = δij (1.155)
Any bivector may analogously be represented in the bivector basis {ig1 , ig2 , ig3 }
as
B = B (1) ig1 + B (2) ig2 + B (3) ig3 (1.156)
where
B (i) = B · [ig(i) ]† = −B · [ig(i) ] (1.157)
is the contravariant component of the bivector B. Similarly, one can use the
covariant components ai = a · gi and Bi = −B · (igi ) to represent the vector a as
a = a1 g(1) + a2 g(2) + a3 g(3) (1.158)
and the bivector B as
B = B1 ig(1) + B2 ig(2) + B3 ig(3) (1.159)
The covariant basis vectors can be expressed as

3
X
gi = gij g(j) (1.160)
j=1
Vice versa, the reciprocal (or contravariant) basis vectors can be written as
3
X
(i)
g = g (ij) gj (1.161)
j=1
The components of covariant Gram tensor gij are given by
gij = gi · gi (1.162)
and the components of contravariant Gram tensor g (ij) are given by
g (ij) = g(i) · g(j) (1.163)
By using the reciprocality of the co- and contravariant basis vectors (Eq. (1.155))
together with the definitions in Eqs. (1.160) and (1.161), we obtain
3
X 3
X 3
X
(i) (ik) (l) (ik)
δij = g · gj = g gjl gk · g = g gjl δkl = g (ik) gjk (1.164)
kl=1 kl=1 k=1
where it follows that the contravariant Gram tensor is the inverse of the covariant
one (and vice versa), i.e.
3
X
g (ik) gkj = δij (1.165)
k=1
Example 5 Let g1 = u1 − u2 /2 + 3u3 , g2 = 2u1 + u2 , and g3 = u3 . Then

g2 × g3 = (2u1 + u2 ) × u3 = −2u2 + u1 , and the determinant of the frame is
g = g1 · g2 × g3 = (u1 − u2 /2 + 3u3 ) · (−2u2 + u1 ) = 2
The reciprocal basis vectors are

g2 × g3 u1
g(1) = = − u2
g 2
g3 × g1 u3 × (u1 − u2 /2 + 3u3 )
g(2) = =
g 2
u2 + u1 /2 1 1
= = u1 + u2
2 4 2
g1 × g2 (u1 − u2 /2 + 3u3 ) × (2u1 + u2 )
g(3) = =
g 2
u3 + u3 + 6u2 − 3u1 3
= = − u1 + 3u2 + u3
2 2
The use of reciprocal basis also offers an elegant way of expanding cross prod-
ucts of basis vectors as combinations of basis vectors, or vice versa.
Example 6 The cross product g1 × g2 can be written as

1
g1 × g2 = [(g1 × g2 ) · (g2 × g3 ) g1 + (g1 × g2 ) · (g3 × g1 ) g2
g
+ (g1 × g2 ) · (g1 × g2 ) g3 ]
1
= [g1 g22 g3 (cos θ12 cos θ23 − cos θ13 ) g1
g
+g12 g2 g3 (cos θ13 cos θ12 − cos θ23 ) g2 + g12 g22 g3 ]
1.12. QUATERNIONS 55
where θij is the angle between gi and gj .
Example 7 Vector g1 can be written as
g1 = [g1 · g1 (g2 × g3 ) + g1 · g2 (g3 × g1 ) + g1 · g3 (g1 × g2 )]/g

= [g12 g2 × g3 + g1 g2 cos θ12 g3 × g1 + g1 g3 cos θ13 g1 × g2 ]/g
1.11.3 Versor representation

Any multivector
3
X (j)
a + iB (j) gj

M = α + iβ + (1.166)
j=1
can also be uniquely expressed in a versor form as
M = c0 + c(1) g1 + c(2) g2 + c(3) g3 + c(12) g1 g2 + c(23) g2 g3

+c(31) g3 g1 + c3 g1 g2 g3 (1.167)
for any frame {g1 , g2 , g3 }. By comparing the coefficients in Eqs. (1.166) and
(1.167), it is evident that
β = c3 g1 · (g2 × g3 ) (1.168)
B (3) = c(12) B (1) = c(23) B (2) = c(31) (1.169)
a(1) = c(1) + c3 g1 g2 · g3 a(2) = c(2) − c3 g1 · g3 g2
a(3) = c(3) + c3 g1 · g2 g3 (1.170)
α = c0 + c(12) g1 · g2 + c(23) g2 · g3 + c(31) g3 · g1 (1.171)
where the second term in the vectorial part (for example c3 g1 g2 · g3 in a(1) ) comes
from c3 hg1 g2 g3 i1̄ = c3 [g1 c (g2 ∧ g3 ) + g1 (g2 · g3 )].
1.12 Quaternions
A comparison of the products of the unit bivectors {i1 , i2 , i3 } in Eq. (1.151) with
the quaternionic multiplication table
i2 = j2 = k2 = −1 (1.172)
ij = k = −ji jk = i = −kj ki = j = −ik (1.173)
ijk = −1 (1.174)
reveals that Hamilton’s unit quaternions {i, j, k} are a set of left-handed orthonor-
mal unit bivectors (i.e. i = −i1 , j = −i2 , and k = −i3 ), and any quaternion
Q = Q(0) + Q(1) i + Q(2) j + Q(3) k (1.175)
(where Q(0) , Q(1) , Q(2) , and Q(3) are scalars) is an even-graded multivector or
spinor (i.e. a multivector, which possesses a scalar plus a bivector part), and the
product between two quaternions is equal to their geometric product. Quater-
nions are closed under geometric product, i.e. the product
Q1 Q2 = αβ − a · b + i(αb + βa) − a ∧ b (1.176)
of two quaternions Q1 = α + ia and Q2 = β + ib is another quaternion. The

magnitude |Q| is given as the positive square root of the scalar |Q|2 = Q† Q.
Example 8 Find the magnitude of the quaternion
Q = u1 (u1 + u2 )
Solution:
Q† = (u1 + u2 ) u1
so
Q† Q = (u1 + u2 ) u1 u1 (u1 + u2 ) = (u1 + u2 ) · (u1 + u2 ) = 2
so
√
|Q| = 2
Since quaternion Q is a spinor, it can be written in an exponential form using

Eq. (1.122) as
Q = |Q| exp(iθ) (1.177)
where i is the unit bivector of the bivector part of the quaternion. The scalar θ
can be solved straightforwardly from the Euler representation of the exponential
function, as the next Example illustrates:
Example 9 Find the exponential form of the quaternion
Q = u1 (u1 + u2 )
1.12. QUATERNIONS 57
Solution:
√
Q = u1 (u1 + u2 ) = 1 + u1 ∧ u2 = |Q| (cos θ + i3 sin θ) = 2 (cos θ + i3 sin θ)
√
(where the magnitude |Q| = 2 was calculated in Example 8), which implies that
√ √
cos θ = 1/ 2 and sin θ = 1/ 2 or equivalently, cot θ = 1, where it follows that
θ = π/4 and
√ π
Q= 2 exp i3
4
By using the exponential representation, it is easy to show that the inverse of

the quaternion Q is given by
Q†
Q−1 = (1.178)
|Q|2
Example 10 Find the inverse to the quaternion
Q = u1 (u1 + u2 )
Solution: By using the magnitude |Q| derived in Example 8, we see that
Q† 1 1 π
Q−1 = = (u 1 + u 2 ) u 1 = √ exp − i3
|Q|2 2 2 4
where the last equality follows using the result of Example 9.
√
k
We obtain the kth root of a quaternion Q as
p p (θ + 2mπ)i
k
Q = k |Q| exp (1.179)
k
p
where the argument θ is confined to the interval 0 ≤ θ ≤ π, the factor k
|Q|
stands for the real and positive kth root of the magnitude of Q, and m =
0, 1, ..., (k − 1). Since a quaternion Q can be written as Q = exp λ exp(iθ) =
exp(λ + iθ), where the scalar λ is related to |Q| as exp λ = |Q|, we obtain a
definition for the (main branch of the) logarithm of a quaternion:
ln Q = λ + iθ (1.180)
It is again assumed that the argument θ is confined to the interval 0 ≤ θ ≤ π.

The quaternions can be represented in terms of the unit matrix

" #
1 0
1= (1.181)
0 1
and the three self-adjoint Pauli matrices
" # " # " #
0 1 0 −i 1 0
σ1 = σ2 = σ3 = (1.182)
1 0 i 0 0 −1
as
Q = Q(0) 1 + Q(1) iσ1 + Q(2) iσ2 + Q(3) iσ3 (1.183)
or " #
Q(0) + iQ(3) iQ(1) + Q(2)
Q= (1.184)
iQ(1) − Q(2) Q(0) − iQ(3)
Note that the i in σ2 is assumed to be the unit trivector in contrast to the conven-
tional treatments, where it stands for the uninterpreted scalar “complex unit”.
Also, we can easily infer that the Pauli matrices form a “complex” representation
of a orthonormal vector triad {u1 ,u2 ,u3 }.
1.13 Complex numbers

Complex numbers are a natural part of geometric algebra. They can be created
by taking the geometric product between u1 and x = xu1 + yu2 where {u1 , u2 }
are two orthonormal vectors. We obtain
z = u1 x = u1 (xu1 + yu2 ) = x + yi (1.185)
The complex numbers z1 = x1 + y1 i and z2 = x2 + y2 i are added as z1 + z2 =

x1 + x2 + (y1 + y2 )i and multiplied as z1 z2 = x1 x2 − y1 y2 + (x1 y2 − x2 y1 )i. The
reversion (i.e. reversing the order of the vector factors) of z corresponds to the
conventional complex conjugation, i.e. z † = (u1 x)† = xu1 = x − yi. However,
the bivector i has geometric and algebraic properties beyond those traditionally
accorded to the “imaginary unit”.
Apart from the quartic and cubic equations, complex numbers arise naturally,
for example, in the power series expansion
X∞
H(x) = (−1)j x2j (1.186)
j=1
1.14. FURTHER READING 59
of the real valued function

1
h(x) = (1.187)
1 + x2
of a real variable x. It is not obvious that the power series should have the interval
of convergence for −1 < x < 1, unless one writes
1
h(z) = (1.188)
(z + i) (z − i)
and notices that h(z) has poles at z = i and z = −i (so the radius of convergence
for the power series representation of h(z) is 1, and it is centered in the origin
z = 0).
1.14 Further reading

Clifford’s seminal article on the unification of quaternions and blades [13], which
started the development of geometric algebra can be obtained in pdf-format in the
home page http://www.jstor.org/journals/00029327.html of American Journal of
Mathematics.
The textbook Geometric algebra for physicists by Doran and Lasenby (Ref.
[18]) offers a treatment of geometric algebra at the present level. The axiomatic
development of geometric algebra is given in the book Clifford algebra to geomet-
ric calculus by Hestenes and Sobczyk [33]. This book is very compact, and it
contains no figures and almost no examples, which makes it rather heavy-going
for a beginner in geometric algebra. Many tutorial introductions to geometric
algebra can be found on-line (such as [31]), along with more advanced material.
Most of Hestenes’ articles can be found in pdf-format in his geometric algebra
page http://modelingnts.la.asu.edu.
Geometric algebra is used to formulate classical mechanics in Hestenes’ in-
fluential book New foundations for classical mechanics [32]. It also includes a
hundred (or so) page length intro to geometric algebra at the present level.
Dorst compares the merits of contraction and scalar products in [20].
An analysis on the angle convention to signature of the geometric algebra is
given in an article by Joyce and Butler [47]. Note that the authors use the term
r-vector to denote a1 a2 ...ar , which contradicts the present terminology.
Hamilton’s collected works can be found on-line in http://www.emis.de/classics/

Hamilton/. The theory of quaternions is developed in [30].
The relation of (polar and axial) vectors to quaternions is discussed in a
recent article [65] by Silva and de Andrade Martins. Curiously enough, Clifford’s
contribution is omitted.
An interesting approach to the conventional vector analysis is offered in a book
Vector and tensor analysis by Brand [10], particularly since it includes chapters
on motors, quaternions and dyads, topics which are almost never found in the
more recent textbooks.
The article by Wilcox [72] is a standard reference on the use of paramet-
ric differentiation in obtaining disentanglement formulas. DasGupta [17] uses
Baker-Hausdorff lemma to derive disentanglement formulas for double photon
and related algebras. Leblond [51] gives examples of non-commuting A and B,
for which the disentanglement formula in Eq. (1.107) still applies.
In his book Visual complex analysis Needham offers an exceptional geometric
approach to complex analysis, and shows how much more intuitive this approach
is compared to the more conventional algebraic approach. Many of the arguments
presented in the book are quite new, and the range of topics covered is way
beyond any traditional textbooks. The interested reader should try to re-express
Needham’s arguments in the language of geometric algebra.
Several free-ware geometric algebra computer programs can be found on-line.
For example, the Clifford Library and Utilities (CLU) is available in http://www.
perwass.de/cbup/clu.html, and the an multi-purpose program (GAViewer) for
performing geometric algebra computations and visualizing geometric algebra can
be obtained from the web page http://www.science.uva.nl/ga/software/index.html.
1.15 Exercises
Exercise 1 Let a = u1 + 5u2 , b = 2u2 + u3 and c = u1 + u2 − u3 . Calculate
a) a∧b b) a×b c) ab d) a∧b∧c e) (a ∧ b)·(b ∧ c) f ) a·(b ∧ c) g) [a · (b ∧ c)]·a.
Exercise 2 Let a = u2 − u1 ∧ u2 , b = u1 + u2 and c = 1 + u2 . Calculate a) ab

b) ac. What do you observe?
1.15. EXERCISES 61
Exercise 3 Let a = u2 + u1 ∧ u2 , b = (1 + u1 ) /2. Calculate a) ab b) ba. What

do you observe?
Exercise 4 Show that |a + b| ≤ |a| + |b| for any vectors a and b.
Exercise 5 Show that |b| a + |a| b and |b| a − |a| b are orthogonal.
Exercise 6 Show that |b| a + |a| b bisects the angle between a and b.
Exercise 7 Let a = u1 + 5u2 , b = 2u2 + u3 . Calculate the undirected included

angle θ between a and b.
Exercise 8 Let a = u1 +5u2 , b = 2u2 +u3 . Calculate the directed angle t between
a and b, when the direction of the plane of the angle is a) in the direction of a ∧ b
b) opposite to that of a ∧ b.
Exercise 9 Derive the law of sines for plane triangles. In other words, find out
how the ratios (sine of an angle)/(length of the opposite side) are related for
the three angles and sides of the triangle (hint: wedge both sides of the equation
a + b + c = 0 by a and b in turn).
Exercise 10 Does changing the included angle convention from θ to t (see Sect.
1.4.2) affect the results in spherical trigonometry derived in this chapter?
Exercise 11 Derive Eq. (1.144).
Exercise 12 Derive Eq. (1.145).
Exercise 13 Let z = x + yi. Show that x = (z + z † )/2 and y = (z − z † )/(2i).
Exercise 14 Prove by the direct substitution that A2 = A for A = (1 + u)/2,

where u is a unit vector.
Exercise 15 Prove that the equation αA = A implies A2 = A. Must α have to

be scalar? If it is not, does the equation Aα = A also imply A2 = A?
Exercise 16 Determine the scalar coefficients α, β and the vectors a, b for the
most general indempotent A = α + iβ + a + ib (i.e. for A, which obeys A2 = A).
Exercise 17 Is the outer product A ∧ A always zero for an arbitrary multivector

A?
Exercise 18 Which of the following are always true and which are not always
true
√
a) |A|2 = A† A 0̄

b) |a| = a·a
c) a × b = b × a d) a × (b + c) = a × b + a × c
e) (a × b) × (c × a) = λa for some scalar λ
f ) a · (λb) = λa · b for some scalar λ
g) a · (ib) = ia · b for unit trivector i
Exercise 19 Which of the following bivectors a ∧ b, c ∧ d and e ∧ f has the same
orientation as the bivector B?
Exercise 20 Which algebraic laws (commutative, associative, distributive) are

satisfied, and which are not satisfied by a) the cross product b) the outer product
Exercise 21 Prove that
a × b = a · (b × u1 )u1 + a · (b × u2 )u2 + a · (b × u3 )u3
Exercise 22 Prove that for any vector a
u1 · (a × u1 ) + u2 · (a × u2 ) + u3 · (a × u3 ) = 2a
Exercise 23 Prove by induction that [B, Ak ] = kAk−1 [B, A], if both A and B
commute with [A, B] = AB − BA.
1.15. EXERCISES 63
Exercise 24 Let A and B commute with [A, B] = AB − BA. Use Eq. (1.107)
to derive a binomial expansion for (A + B)k .
Exercise 25 Derive the Baker-Hausdorff formula in Eq. (1.110).
Exercise 26 Does the entaglement formula in Eq. (1.129) hold if either λ1 or

λ2 is negative?
Exercise 27 Assume the quaternionic multiplication rules ij = k = −ji, jk =

i = −kj and ki = j = −ik. Show that the product i (i + j) j is associative, i.e.
[i (i + j)] j = i [(i + j) j], only if i2 = j2 = −1.
Exercise 28 Repeat the previous exercise by treating the quaternions i, j and k

as orthonormal unit vectors, and by replacing the products ii, jj and ij by
a) geometric products ii = 1, jj = 1 and ij = i ∧ j
b) dot products i · i = 1, j · j = 1 and i · j = 0
c) cross products i × i = 0, j × j = 0 and i × j = k
respectively. What do you observe?
Exercise 29 If unit quaternions are interpreted as unit vectors, is it possible to

generalize the quaternion products of the type ij = k to higher dimensional spaces?
Exercise 30 Prove that the Pauli matrices {σ1 ,σ2 ,σ3 } satisfy σi σj = δij 1 +
iijk σk , where the Kronecker’s delta δij is one if i = j and zero otherwise, and
the permutation symbol ijk is zero, if any two indices are the same, ijk is +1 if
{i, j, k} is a cyclic permutation of {1, 2, 3} (i.e. {i, j, k} = {1, 2, 3}, {2, 3, 1}, {3, 1, 2})
and ijk is −1 if {i, j, k} is an anticyclic permutation of {1, 2, 3} (i.e. {i, j, k} =
{2, 1, 3}, {3, 2, 1}, {1, 3, 2}).
Exercise 31 Verify that the orthonormal vectors {u1 ,u2 ,u3 } can be represented
in terms of the Pauli matrices {σ1 ,±σ2 ,σ3 }, i.e. the sign of Pauli matrix σ2 is
not uniquely determined by the product ui uj = δij + iijk uk .
Exercise 32 Is it generally possible to infere from the matrix representation of

a vector if the frame used to expand it is right- or left-handed?
Exercise 33 Start from the given representation u3 ↔ σ3 . Derive the form of

the remaining self-adjoint Pauli matrices {σ2 ,σ3 }. Hint: write
" #
α β
u1 ↔
β ∗ −α
where α and β are odd multivectors (i.e. can contain scalar and trivector parts)
and asterisk signifies changing the sign of the trivector part. Use u1 u3 +u3 u1 = 0
to determine the value of β, and the relation u21 = 1 ↔ 1 to fix the value of β.
Similarly, show that the identities u2 u3 + u3 u2 = 0 and u22 = 1 ↔ 1 imply that
" #
0 γ
u2 ↔
γ∗ 0
with γγ ∗ = 1, which , in turn, implies that γ = ±i.
Exercise 34 Solve the vector equation
αx + a1 x · b1 + a2 x · b2 + x × d = c (1.189)
Hint: Combine the solutions to the vector equations in Eqs. (1.90) and (1.96).
Exercise 35 Prove the expansion rule in Eq. (2.23).
Exercise 36 Prove that the representation of the geometric product ab in terms

of the lengths a, b, unit bivector i = a ∧ b/ |a ∧ b|, and the included angle is
independent wether one uses the θ or t for the included angle. Hint: to prove
j sin t = i sin θ take into account the change in the positive orientation of i for
t > π.
−→ −→
Exercise 37 Prove that the sum of the bivector faces B1 = AB ∧ AC/2, B2 =
−→ −−→ −−→ −→ −−→ −−→
AC ∧ AD/2, B3 = AD ∧ AB/2 and B4 = BD ∧ BC/2 of the tetrahedron ABCD
is zero. Does this result also generalize to the arbitrary polyhedras (why)?
Exercise 38 Prove that it is possible to interchange the dot and cross in a·b × c.
Exercise 39 Establish the “Jacobi identity” for vectors: a · (b ∧ c) + b · (c ∧ a) +

c · (a ∧ b) = 0.
1.15. EXERCISES 65
Exercise 40 Prove that the bivector a ∧ b + c ∧ d can be factored into a product

of vectors if and only if a ∧ b ∧ c ∧ d = 0.
Exercise 41 Vectors a, b and c are linearly dependent if there exists scalars α,

β and γ, which are not all zero, so that αa+βb+γc = 0. Prove that a∧b∧c = 0
only if a, b and c are linearly dependent.
Exercise 42 Let a = a(1) g1 + a(2) g2 . Determine the contravariant coefficients

a(1) and a(2) . Hint: form the outer products a ∧ g2 and a ∧ g1 .
Exercise 43 Expand a = u1 + 3u2 in an oblique basis {u1 , u1 − 5u2 }.
Exercise 44 Suppose that vectors a and b are not parallel and a = c + d, where
c is parallel to b and d is orthogonal to b. Express c and d in terms of a and b.
Exercise 45 Expand a = u1 +3u2 +5u3 in an oblique basis {u1 , u1 − 5u2 , u1 + u2 + u3 }.


3 D Geometric Algebra 1

Uploaded by

Copyright:

Available Formats

3 D Geometric Algebra 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

3 D Geometric Algebra 1

Uploaded by

Copyright:

Available Formats

Chapter 1

1. The sum α + β + γ can be calculated in any order,

i.e. it is both associative, distributive, and commutative1 .

2. The product αβγ is associative i.e.

αβγ = (αβ)γ = α(βγ) (1.2)

3. The product is commutative, i.e. αβ = βα

4. The multiplication is associative, distributive and commutative with respect

The advantage in defining scalars implicitly via addition and multiplication

Definition 1 (Commutation property) Scalars commute with all the elements

Definition 2 (Square of scalar is positive) The square of any scalar is pos-

This definition serves to exclude complex numbers, to which a direction can be

Figure 1.1: Arrows representing the same vector.

1.3 Addition of vectors

Figure 1.2: a and −a.

Figure 1.3: Vector addition through triangle construction.

Figure 1.4: Vector addition through parallelogram construction.

Figure 1.5: Associativity of vector addition.

Through the vector addition rule,

1.4 Products of vectors

1. It is distributive over addition, i.e.

Figure 1.6: Inner product a · b.

Incidentally, these three properties are sufficient to define a unique product,

1.4.1 Inner product between two vectors

(a + b)2 = |a|2 + |b|2 + ab + ba (1.8)

where, by the Pythagorean theorem, we can write

|a + b|2 = |a|2 + |b|2 + 2a · b (1.9)

The a · b is the inner product (the “dot” product) of a and b. By inspecting

1.4.2 Outer product of two vectors

ab − ba |a|2 b − aba b |a|2 − aba ba2 − aba

Nor is a ∧ b a vector, because its square is negative, as seen by

Figure 1.7: Bivector a ∧ b.

1.4.3 Geometric product of two vectors

Figure 1.8: Different representations of a bivector B.

This is an entirely misguided question that will only confuse the

Figure 1.10: Invertibility of the geometric product

Example 1 Let a = u1 + 2u2 and b = u1 − u2 + u3 /2, where {u1 , u2 , u3 } is a

ab = (u1 + 2u2 )(u1 − u2 + u3 /2)

so the notation a∧b∧c is unambiguous. By using the associativity, we also easily

and invariant with respect to symmetric permutations, i.e.,

or equally, due to the associativity of outer product,

a ∧ b ∧ c = h(a ∧ b) ci3̄ (1.23)

Figure 1.11: Trivector a ∧ b ∧ c.

The trivector a ∧ b ∧ c can be pictured as an oriented parallelepiped with

Contraction of bivector by vector

acB = −Bba (1.26)

Furthermore, by taking the reverse of Eq. (1.24) it follows that (acB)† =

This notation is unambiguous, since we follow the precedence convention, where

Geometrically, b ∧ cba is obtained by rotating the projection of a dilated by the

Contraction of trivector by vector

The left contraction of a unit trivector i by a vector a is obtained as a symmetric

aci = iba (1.36)

By a similar derivation to that of the contraction of a bivector by vector we can

acb ∧ c ∧ d = a · bc ∧ d − a · cb ∧ d + a · db ∧ c = b ∧ c ∧ dba (1.37)

and these represent the bivector (plane) perpendicular to the vector a

Contraction of bivector by bivector

By further generalizing the contraction products we define contraction of bivector

B1 · B2 = hB1 B2 i0̄ (1.38)

of the bivector a ∧ b by the bivector c ∧ d can be expressed in terms of the inner

Evidently, fthe inner product B1 · B2 of the bivectors B1 and B2 is commutative,