2

Traditional Vector Theory

The earliest denition of a vector usually encountered is that a vector is a thing possessing
length and direction. This is the arrow in space view with length naturally being the
length of the arrow and direction being the direction the arrow is pointing. We will denote
vectors by writing their name in boldface (e.g., v and a ) or by drawing a little arrow above their
name (e.g.,

v and

a ).
Some examples of such vectors include the velocity of an object at a particular time and the
acceleration of an object at a particular time. (For now, we are not considering vector elds; that
is, our vectors will not be functions of time or position.)
Perhaps the most fundamental of vectors are those describing relative position or dis-
placement: If A and B are two points in space (i.e., positions in space), then the vector from
A to B , denoted

AB , is simply the arrow
1
starting at point A and ending at point B .

AB
gives you the direction and distance to move from position A to position B (hence the term
relative position).
We will use displacement vectors as the basic model for traditional vectors. Almost all
other traditional vectors in physics velocities, accelerations, forces, etc. are derived from
displacement vectors. For example, the velocity v of an object at a given position p is the limit
v = lim
t 0

pq
t
where q is the position of the object t time units after being at p . It is through these derivations
that the properties we derive for displacement (relative position) vectors can be shown to hold
for all the traditional vectors in physics.
By the way, throughout this discussion, we are assuming that the points in space are points
in a Euclidean space. We will discuss exactly what this means later (in another chapter). For
now, assume high school geometry. In particular:
1. Any two points can be connected by a straight line segment (totally contained in the
space) whose length is the distance between the two points.
2. The angles of each triangle add up to (or 180 degrees).
3. The laws of similar triangles hold.
4. Parallelograms are well dened.
1
ofcially,

AB is a directed line segment.
11/28/2011
Traditional Vector Theory Chapter & Page: 22
Keep in mind that there are different Euclidean spaces. Two different at planes, for example,
are two different Euclidean spaces.
?

Exercise 2.1: Give several reasons why a sphere is not a Euclidean space.
2
2.1 Fundamental Dening Concepts
Fundamental Geometrically Dened Concepts
The Two Fundamental Measurable Quantities
Our goal is to develop the fundamental theory of vectors as things that are completely dened by
length and direction, using the set of vectors describing relative position in some Euclidean
space as a basic model. Keep in mind the requirement that length and direction completely
denes a vector here. So if we have two arrows, both pointing in the same direction and of the
same length, then they are the same vector.
?

Exercise 2.2: Let A , B , C and D be points in space. How does the concept that
Parallelograms are well dened allow us to decide when

AB =

CD ?
With traditional vectors, there are two fundamental measurable quantities to work with:
Length: The length (also called the magnitude or norm) of a vector v is denoted by v or
|v| or v , depending on the whim of the author.
Angle Between Two Vectors: While it is intuitively clear what direction means, it is
difcult to quantify. What can be measured (physically, in theory) is the smallest angle
between two vectors u and v visualized with their starting points touching. We will typically
denote this angle by or (u, v) , and we will use radians or degrees to measure this angle
as seems convenient. Do note that
(u, v) = (v, u) .
It should be noted that measurable means that we can (conceptually, at least) nd these
quantities by directly measuring the length and the angle with fairly basic real-world tools.
Consequently, we automatically have
v 0 and 0 .
(We may later relax this restriction on .)
Units for Length and Types of Vectors
In many basic developments of traditional vector theory, the length of a vector, v , is treated as
a simple, unadorned real number. In fact, however, any vector length measurement is inherently
2
Terminology: A sphere is just the surface of a ball. It does not contain the inside of the ball.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 23
in terms of some measurement unit such as feet, meters, miles/hour, meters/second
2
, newtons,
etc. Even if we dont explicitly state the unit of length for a vector we sketch on paper, we
must have some notion of what one unit of length is, and that unnamed quantity is our basic
unit of length give it a name if you wish.
Whats more, we need to realize that there are different types of vectors according to what
is being measured. For any vector describing displacement fromone position to another, length
really means distance in the normal sense, and can be measured in feet, meters, furlongs, and
so forth. This is one type of vector. On the other hand, the length of a vector describing the
velocity of an object at a given time is really the speed of that object, and can be measured in
feet/second, meters/hour, furlongs/fortnight, and so on. The set of velocity vectors make up
another type of vector. And we also have acceleration vectors, force vectors, ux vectors,
. . . .
Subsequent Geometric Concepts
All of our traditional vector computations resulting in scalar values will be based on length
and angle . Note that, right off, we can use the length to dene one very special vector and a
set of vectors that we may nd useful:
1. The zero vector, denoted by 0 or

0 , is the vector of length zero. One can argue that

there are innitely many zero vectors, each corresponding to a different direction, but
that is being silly. The zero vector, in our model of relative position, corresponds to just
not moving from a point. For that and other practical reasons, we take the position that
there is only one zero vector in each set of vectors
2. A unit vector is any vector

v of length one (i.e.,
_
_

v
_
_
= 1 ). If it is signicant that
this is a unit vector, then the vector may be denoted by v or v , instead of

v . We will
nd various uses for unit vectors.
And with the concept of angles between vectors, we can introduce terminology for when
vectors point in special directions relative to each other. Most of this terminology is what we use
everyday.
Parallelness: Two vectors u and v are said to be parallel if either they point in the same
direction or they point in opposite directions. Obviously,
u and v point in the same direction (u, v) = 0
and
u and v point in opposite directions (u, v) = .
Orthogonality:
(a) A pair of vectors {u, v} is said to be orthogonal
Either (u, v) =

2
or u = 0 or v = 0 .
(Thus, weve dened orthogonality so that 0 is automatically orthogonal to every
vector. The reasons we include this in our denition is simply that it will simplify
statements and formulas later.)
On occasion, we will indicate that {u, v} is orthogonal by writing u v .
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 24
u
u
v
v
u
+
v
Figure 2.1: The vector sum, based on the well-dened parallelogram having u and v as
two sides.
(b) A set of vectors {v
1
, v
2
, . . . , v
K
} is said to be orthogonal
Every distinct pair in the set is orthogonal.
With both length and angle we can dene
Orthonormality: A set of vectors {v
1
, v
2
, . . . , v
K
} is said to be orthonormal
It is an orthogonal set of unit vectors.
Using orthogonal or orthonormal sets of vectors will greatly simplify life.
Fundamental Geometrically Dened Algebraic Operations
There are two (to begin with):
Scalar Multiplication of a Vector: Let be a scalar
3
and let v be a vector. The product of
with v , denoted v (or v ), is dened as follows:
If > 0 , then v is the vector of length v pointing in the same direction as v .
If < 0 , then v is the vector of length || v pointing in the direction opposite of v .
If = 0 , then v = 0
Along these lines, we automatically have that the division of a vector v by a nonzero scalar
is given by
v

=
1

v .
Vector Addition: Within our model of vectors representing relative position, the addition
of

AB to

BC naturally corresponds to go from position A to position B and then go
from position B to position C . The result, of course, is that youve gone from position A
to position C . So we dene the sum of

AB and

BC to be

AC , and write

AB +

BC =

AC .
Geometrically, this is exactly the same as dening the sum of two vectors u and v using a
parallelogram as in gure 2.1.
3
Until otherwise announced, assume all scalars are real numbers.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 25
From the gure (and our assumptions about Euclideaness), it should be obvious that
u + v = v + u .
Other properties of addition and scalar multiplication should also be obvious or easily veried.
?

Exercise 2.3: Using just the denitions, assumptions, pictures, and high school geometry,
convince yourself that
(u +v) = u + v and ( +)v = v + v
for arbitrary vectors u and v , and arbitrary scalars and .
4
?

Exercise 2.4: Convince yourself that vector addition is associative, i.e., that, for any three
vectors u , v and w,
(u +v) + w = u + (v +w) .
at least when the vectors are dened in terms of relative position, as above.
Keep in mind that vector addition requires that the vectors be of the same type. Adding a
velocity vector to an acceleration vector makes no sense.
Decomposing Vectors, Part I (Length and Direction)
Let v be a nonzero (traditional) vector. Corresponding to v is the corresponding unit vector
u in the direction of v is given by
u =
v
v
.
Note that
u =
_
_
_
_
v
v
_
_
_
_
=
_
_
_
_
1
v
v
_
_
_
_
=
1
v
v = 1 ,
so, as the terminology suggests, it really is a unit vector. Also,
v = 1 v =
v
v
v = v
v
v
= v u .
Thus, we can decompose (or resolve) a nonzero vector v into the product of its length times
the unit vector in the direction of v . This will be useful in a number of situations.
2.2 Traditional Vector Spaces
Using the four geometric concepts discussed above (length, angle between vectors, scalar multi-
plication and vector addition), we can now develop the rest of the theory for traditional vectors.
And as any good mathematician can tell you, half the work is coming up with the right denitions.
By the way, please note two things:
4
Really convince yourself. Dont just say, Oh yeah, I know its true. Be able to convince skeptics.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 26
1. At this point, our theory is completely component free we are viewing traditional
vectors the way God does. All those component formulas youve mucked about with for
years have yet to be derived.
2. Much of the terminology and development that follows will apply to more general, non-
traditional sets of vectors. We will use this fact when we generalize our results.
I should also mention that K , N and M will denote arbitrary xed positive integers
throughout the rest of this chapter (unless otherwise indicated).
Linear Combinations and Related Notions
Linear Combinations and Span
A linear combination of vectors v
1
, v
2
, . . . and v
N
is any expression of the form

1
v
1
+
2
v
2
+ +
N
v
N
where
1
,
2
, and
n
are scalars. It is normally assumed that the number of terms in a linear
combination ( N in the above) is nite. However, when we get to more general vector spaces,
especially the vector spaces that will be useful in solving partial differential equations, we will
start allowing linear combinations with innitely many terms (we will then also have worry about
convergence issues).
Observe that the subtraction of one vector from another is just a particularly simple linear
combination,
u v = (1)u + (1)v .
The following should be obvious:
u u = 0
and
w = u v w + v = u
?

Exercise 2.5: Draw two vectors u and v with their starting points touching. (Make
them of different lengths and pointing in different directions!) On this gure, sketch
u + v , u v , v u and 2u + 4v .
Some more terminology: Given a set of vectors
{v
1
, v
2
, v
3
, . . .}
(not necessarily nite), the span of this set is the set of all (nite) linear combinations of these
v
k
s .
Linear (In)dependence
We will be interested in minimal sets of vectors for generating particular collections of linear
combinations. To label when such a set
{ v
1
, v
2
, v
3
, . . . , v
N
}
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 27
is or is not minimal in this sense, we have the following terminology:
{ v
1
, v
2
, v
3
, . . . , v
N
} is linearly dependent
One of the v
k
s can be expressed as a linear combination of the others.
and
{ v
1
, v
2
, v
3
, . . . , v
N
} is linearly independent
{ v
1
, v
2
, v
3
, . . . , v
N
} is not linearly dependent
None of the v
k
s can be expressed as a linear combination of the others.
Our interest will often be in linearly independent sets. They will be the minimal sets for
constructing linear combinations.
?

Exercise 2.6: Why can the zero vector, 0 , never be in a linearly independent set of vectors.
Alternative denitions for linear dependence and independence (which should really be
considered as tests for linear dependence and independence) are described in the next exercises.
?

Exercise 2.7: Show that

a: { v
1
, v
2
, . . . , v
N
} is linearly dependent (as dened above) if and only if there are scalars

1
,
2
,
3
, . . . and
N
not all zero such that

1
v
1
+
2
v
2
+ +
N
v
N
= 0 .
b: { v
1
, v
2
, . . . , v
N
} is linearly independent (as dened above) if and only if the only choice
for scalars
1
,
2
,
3
, . . . and
N
such that

1
v
1
+
2
v
2
+ +
N
v
N
= 0
is
1
=
2
= =
N
= 0 .
?

Exercise 2.8: Come up with at least one alternative denition/test for the linear indepen-
dence of { v
1
, v
2
, . . . , v
N
} based on the span of this set and subsets of this set.
Vector Spaces
Let V be some (big) set of traditional vectors (all are the same type so they can be added
together). Then this set is called a vector space (of traditional vectors) if and only if V is
closed with respect to linear combinations. And what does V is closed with respect to
linear combinations mean? It means that every linear combination of vectors in V is another
vector in V . That is, a set V of vectors is a vector space if and only if, for each
N , v
1
, v
2
, . . . , v
N
V and
1
, ,
2
, . . . ,
N
,
we have

1
v
1
+
2
v
2
+ +
N
v
N
V .
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 28
For example, all the displacement vectors generated from points in a single plane is a vector
space.
Note that the span of a set of vectors is automatically a vector space. It is often helpful to
nd minimal sets of vectors that span a given vector space V . Such a set is called a basis for
V . That is, a basis for a vector space V is any (nite) set of vectors
B = { b
1
, b
2
, . . . , b
N
}
(all from V ) that satises both of the following:
1. {b
1
, b
2
, . . . , b
N
} is linearly independent .
2. Each vector in V is a linear combination of vectors from B. That is, if v is any vector
in V , then there is a corresponding ordered set of N scalars (v
1
, v
2
, . . . , v
N
) such that
v = v
1
b
1
+ v
2
b
2
+ + v
N
b
N
.
The v
k
s just describedare calledthe components
5
of v withrespect tothe basis {b
1
, b
2
, . . . , b
N
} .
The following three statements are all easily veried. In them
B = { b
1
, b
2
, . . . , b
N
}
is a basis for some vector space V , v is some vector in V , and (v
1
, v
2
, . . . , v
N
) are the
components of v with respect to B.
1. The v
k
s are unique. That is, given any set of N scalars {
1
,
2
, . . . ,
N
} , then

1
b
1
+
2
b
2
+ +
N
b
N
= v = v
1
b
1
+ v
2
b
2
+ + v
N
b
N
if and only if

1
= v
1
,
2
= v
2
, . . . and
N
= v
N
.
2. Any set of more than N vectors in V cannot be linearly independent.
3. Any set of fewer than N vectors in V cannot span all of V (i.e., there will be vectors
in V that cannot be expressd as linear combinations of vectors from that set).
The last two facts tell us that a set of vectors can be a basis for V only if it is a linearly independent
set of exactly N vectors. On the other hand, if we have a linearly independent set of N vectors
{v
1
, v
2
, . . . , v
N
} , and w is another vector in V , then w must be a linear combination of the
v
k
s otherwise {w, v
1
, v
2
, . . . , v
N
} would be a linearly independent set of more than N
vectors, contrary to fact 2, above. Consequently, we must have
4. A set of vectors in V is a basis for V if and only if it is a linearly independent set of
exactly N vectors.
The number N of vectors in any basis of a vector space V is called the dimension of that vector
space. In applications with traditional vectors (relative positions, velocities, etc.), the dimension
is nite, usually 3 (or 2 if we want to simplify things).
It is often convenient to be able to switch from one basis to another. We will discuss this in
great detail later. For now:
5
NOT the coordinates of v ! Vectors do NOT have coordinates.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 29
?

Exercise 2.9: Let V be a two-dimensional vector space with basis B = {b

1
, b
2
} , and let
c
1
= 2b
1
+ 3b
2
and c
2
= b
1
b
2
.
a: Is C = {c
1
, c
2
} a basis for V . (Give good reasons for your answer, based on the above
facts.)
b: Suppose v = 5b
1
+ 4b
2
.
Find v in terms of c
1
and c
2
. In other words, nd scalars a and b so that
v = ac
1
+ bc
2
.
What are the components of v with respect to B?
What are the components of v with respect to C ?
c: Find b
1
and b
2
in terms of c
1
and c
2
.
d: Suppose w = 5c
1
+ 4c
2
.
Find the components of w with respect to B.
Find the components of w with respect to C .
Let (v
1
, v
2
, . . . , v
N
) be the N-tuple of components of a vector v with respect to some basis
B = {b
1
, b
2
, . . . , b
N
} . In computations, we often substitute (v
1
, v
2
, . . . , v
N
) for v , sometimes
even writing
v = (v
1
, v
2
, . . . , v
N
) .
This is sloppy. Strictly speaking, it is even wrong. v is a vector (relative position, velocity, etc.)
while (v
1
, v
2
, . . . , v
N
) is an element of
N
. The two are not the same thing, and identifying
the two is safe only under the simplest of circumstances. It is certainly not a good idea when
there are two different bases at hand. So dont write things like
v = (v
1
, v
2
, . . . , v
N
)
unless you keep reminding yourself that this is really shorthand for
v = v
1
b
2
+ v
2
b
2
+ + v
N
b
N
.
In other words, (v
1
, v
2
, . . . , v
N
) should be viewed as a convenient description of v with respect
to a given basis B. Change the basis, and the corresponding representation for v can change
dramatically. (So (v
1
, v
2
, . . . , v
N
) is a basis-dependent description of v .)
Still, this basis-dependent description can be useful, and is used in basis-independent for-
mulas. This is illustrated in the next exercise, in which you conrm your favorite formulas for
vector addition and scalar multiplication.
?

Exercise 2.10: Let v and w be vectors ina vector space havingbasis B = {b

1
, b
2
, . . . , b
N
} .
Let and be any two scalars and conrm that the components of v +w are
( v
1
+w
1
, v
2
+w
2
, . . . , v
N
+w
N
)
where
(v
1
, v
2
, . . . , v
N
) and (w
1
, w
2
, . . . , w
N
)
are, respectively, the components of v and w with respect to B.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 210
2.3 The Dot Product
Geometric Denition
Now we combine the fundamental geometric notions of the length v of a vector v and the
angle (v, w) between vectors v and w to dene the classic dot product. Throughout this
section, all vectors are assumed to be from a single traditional vector space V .
The dot (or scalar or inner) product of two vectors v and w, denoted by either
v w or v | w ,
is the scalar given by
v w = v | w = v w cos() where = (v, w) .
The dot notation, v w is more traditional, but the bracket notation v | w corresponds to
notation we will later use for generalizations of the dot product.
6
Of course, we must observe that
v w = w v , v v = v
2
and cos() =
v w
v w
.
(Note: If you ever see v
2
, it usually means v v , i.e., v
2
.)
?

Exercise 2.11: Let be any scalar, and v and w any two vectors. Show that factors
out of the dot product, i.e., that
(v) w = (v w) .
Be sure to consider all three cases: > 0 , = 0 and < 0 (a picture may help with the
< 0 case). Do NOT use the component formula for the dot product!
?

Exercise 2.12: Using the result of the last exercise, show that two scalars and must
be the same if
v = v
for some nonzero vector v .
We can now express our denitions for orthogonality and orthonormality in terms of the dot
product:
Orthogonality:
(a) A pair of vectors {v, w} is said to be orthogonal
either v = 0 or w = 0 or the angle between v and w is

/
2
v w = 0 .
6
In the language of tensor analysis, the dot product is also called the metric on the vector space V . This confuses
some mathematicians because, for them, the metric on the vector space V means something else; namely, the
function given by (v, u) = v u .
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 211
(b) A set of vectors {v
1
, v
2
, . . . , v
K
} is said to be orthogonal
every distinct pair in the set is orthogonal
v
i
v
j
=
_
0 if i = j
v
i

2
if i = j
.
Orthonormality: A set of vectors {v
1
, v
2
, . . . , v
K
} is said to be orthonormal
it is an orthogonal set of unit vectors
v
i
v
j
=
_
0 if i = j
1 if i = j
.
Using orthogonal or orthonormal bases will greatly simplify life.
At this point, let me introduce a commonly used symbol in mathematical physics: For any
two integers i and j , the corresponding Kronecker delta,
i j
, is dened by

i j
=
_
0 if i = j
1 if i = j
.
Using the Kronecker delta, we can dene set of vectors {v
1
, v
2
, . . . , v
K
} to be orthonormal if
and only if
v
i
v
j
=
i j
.
(Okay, inventing this notation just to simplify the denition of orthonormality was silly. But the
Kronecker delta will simplify notation elsewhere, too.)
?

Exercise 2.13: Let S = {v

1
, v
2
, . . . , v
K
} be any nite set of nonzero vectors. Show that
S is orthogonal. S is linearly independent.
Note that it then follows that
S is linearly dependent. S is not orthogonal.
Is it true that a linearly independent set must be orthogonal? If so, explain why; if not, give
an example of a linearly independent set that is not orthogonal.
Projections and Linearity of the Dot Product
Decomposing Vectors, Part II (Projections)
Let v and a be two vectors, with a , at least, being nonzero. By elementary geometry, we can
write v as the sum of two vectors
v = A + B
where A is parallel to a , and B is orthogonal to a (see gure 2.2). The vector parallel to a
will be denoted by

pr
a
(v) (instead of A), and will be called either the projection of v onto a
or the (vector) component of v in the direction of a (depending on the whim of the speaker).
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 212
a a

v
v
A =

pr
a
(v)
A =

pr
a
(v)
B =

or
a
(v)
(a) (b)
Figure 2.2: The projection of v onto a , and the projection of v orthogonal to a .
This is the piece of v we will usually be most interested in. The other vector component, B,
doesnt have a name or notation generally agreed upon. We will denote it by

or
a
(v) , and will
call it either the projection of v orthogonal to a or the (vector) component of v orthogonal to a
(depending on the whim of the speaker). So, using the direction given by a as a base direction,
we have now decomposed v into a pair of vectors
v =

pr
a
(v) +

or
a
(v)
where

pr
a
(v) is parallel to a and

or
a
(v) is orthogonal to a
To get a useful formula for

pr
a
(v) , it helps to introduce the scalar component of v in the
direction of a , which, for convenience, we will briey denote by
a
(v) and dene by

a
(v) =
_ _
_

pr
a
(v)
_
_
if

pr
a
(v) points in the direction of a

_
_

pr
a
(v)
_
_
if

pr
a
(v) points in the direction opposite of a
.
Thus,

pr
a
(v) =
a
(v) the unit vector in the direction of a .
Using elementary trigonometry, you can easily derive a simple formula for this scalar component
in terms of , and then rewrite that formula in terms v a and a . Combining that with the
last equation above then leads to the standard dot product formula for the vector projection. The
details are left to you:
?

Exercise 2.14: Let a and v be as above. Show, using trigonometry and the geometric
denition of the dot product from page 210, that

a
(v) =
v a
a
,
and then that

pr
a
(v) =
v a
a
2
a .
(Be sure to consider both cases.)
Linearity and the Component Formulas for the Dot Product
To derive the standard component formula for the dot product, you start by verifying some simple
linearity equations for projections and the dot product in two dimensions:
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 213
?

Exercise 2.15: Let v , w and a be three vectors in a plane, with at least a being
nonzero, and do the following (without using the well-known component formula for the dot
product):
a: Draw a gure with a and v starting at the same point. Add w starting a the end of v
and draw in the sum v +w from the start of v to the end of w. Using this gure, convince
yourself that

pr
a
(v +w) =

pr
a
(v) +

pr
a
(w) .
b: Now, using what you just found (along with the results from exercise 2.14), show that
(v +w) a = v a + w a .
With a little ability to visualize vectors in three dimensions, you should be able to see that
the rst part holds when a , v and w are not all in one plane (no matter what dimension our
space of traditional vectors is). Consequently, the result of the second part,
(v +w) a = v a + w a ,
holds for any three vectors a , v and w (in the same vector space, of course). By the commu-
tativity of the dot product, we also then have
a (v +w) = a v + a w ,
from which it follows that, with four vectors,
(a +b) (v +w) = a (v +w) + b (v +w)
= a v + a w + b v + b w .
And, of course, similar expansions can be done with even more vectors. This now allows
you to derive the various component formulas for the dot product, including the one most of you
so love.
?

Exercise 2.16: For the following, assume

B = { b
1
, b
2
, . . . , b
N
}
is a basis for a traditional vector space, and let
( v
1
, v
2
, . . . , v
N
) and ( w
1
, w
2
, . . . , w
N
)
be the components with respect to B of vectors v and w, respectively. (So,
v = v
1
b
1
+ v
2
b
2
+ . . . + v
N
b
N
=
N

j =1
v
j
b
j
and
w = w
1
b
1
+ w
2
b
2
+ . . . + w
N
b
N
=
N

k=1
w
k
b
k
.)
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 214
a: Show that
v w =
N

j =1
N

k=1

j k
v
j
w
k
where each
j k
is the constant b
j
b
k
.
b: To what does the formula for v w reduce when the basis B is orthogonal?
c: To what does the formula for v w reduce when the basis B is orthonormal?
The answer to the very last part of the last exercise is, of course,
v w =
N

j =1
v
j
w
j
.
Remember this only holds when the basis being used is orthonormal.
Because of the relation between the dot product and norms of vectors, you should be able
to quickly do the corresponding exercise for obtaining the component formula(s) for v :
?

Exercise 2.17: For the following, assume

B = { b
1
, b
2
, . . . , b
N
}
is a basis for a traditional vector space, and let
( v
1
, v
2
, . . . , v
N
)
be the components with respect to B of a vector v in this space.
a: Show that
v =

_
N

j =1
N

k=1

j k
v
j
v
k
where the
j k
s are constants computable fromthe basis vectors. Be sure to give the formula
for computing these constants.
b: To what does the formula for v reduce when the basis B is orthogonal?
c: To what does the formula for v reduce when the basis B is orthonormal?
And, of course, the answer to the last part of the last exercise is
v =

_
N

j =1
v
j
2
.
At this point, we can also verify a simple dot product test for equality, which, oddly enough,
will be useful when discussing the cross product in the next section.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 215
?

Exercise 2.18 (dot product test for equality):

a: Suppose D is a vector such that
D c = 0 for every c .
Show that D = 0 . (Hint: What if we choose c = D?)
b: Suppose A and B are two vectors such that
A c = B c for every c .
Show that A = B. (Hint: What about D = A B?)
Components and the Dot Product
We need to be a little careful about using the word components. If we have a vector v and a
basis
B = { b
1
, b
2
, . . . , b
N
} ,
then v will have components (v
1
, v
2
, . . . , v
N
) with respect to the basis. These are the scalars
such that
v = v
1
b
1
+ v
2
b
2
+ + v
N
b
N
.
Vector v will also have scalar and vector components with respect to each b
j
given by

b
j
(v) =
v b
j
_
_
b
j
_
_
and

pr
b
j
(v) =
v b
j
b
j
2
b
j
,
respectively. As you can easily verify, these yield slightly different things when the basis is not
orthonormal.
?

Exercise 2.19: Let (v

1
, v
2
, . . . , v
N
) be the components of a vector v with respect to some
arbitrary basis
B = { b
1
, b
2
, . . . , b
N
} .
a i: Find, in terms of the v
j
s , the formulas for the vector and scalar components of v with
respect to each b
k
.
ii: How, then, would you nd the v
j
s if you knew the vector and scalar components of
v with respect to each b
k
.
b: Show that, if B is orthogonal, then
v
j
=
v b
j
_
_
b
j
_
_
2
.
(A more general version of this which will look almost the same will be important in
solving many partial differential equation problems later.)
c: Show that, if B is orthonormal, then
v
j
= v b
j
.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 216
2.4 The Cross Product
The dot product can be dened whatever the dimension of our traditional vector space is, and can
(and will) be generalized to nontraditional vector spaces. On the other hand, the cross product is
essentially an operation requiring that we have a three-dimensional space of traditional vectors.
So, in the following, assume that the vector space is just such a space.
Geometric Denition
Given two vectors v and w, the cross product vw is dened to be the vector with the following
length and direction:
Length: v w = v w sin() where = (v, w) .
Direction: If v w = 0 , then we automatically have vw = 0 . Otherwise the direction
of v w is the one direction orthogonal to both v and w given by the classic right-hand
rule (which I will not attempt to sketch for you here look in your old Intro to Physics or
old Calculus text.)
There are a number of properties that are relatively easily veried from this denition:
1. The cross product is anticommutative. That is,
w v = v w .
2. Assuming v and w are nonzero, then
v w = 0 v and w are parallel .
In particular, we always have
v v = 0 .
3. If {v, w} is an orthonormal pair of vectors, then {v, w, v w} is an orthnormal set.
4. Scalars factor out. That is, for any scalar ,
( v) w = (v w) = v ( w) .
5. The cross product is not associative. That is, in general,
(u v) w = u (v w) .
6. The cross product is related to the dot product via
v w
2
= v
2
w
2
(v w)
2
.
(This follows from the denitions and an elementary trigonometric identity.)
?

Exercise 2.20: Explain (to yourself, at least) why each of the previous statements is true.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 217
a a

b
b
c
a b (or a b)

h
h
(a) (b)
Figure 2.3: (a) The parallelogram generated by a and b. (b) The parallelepiped generated
by a , b and c .
Area and Volume and the Cross Product
When we talk about the parallelogram or parallelepiped generated by two or three vectors a ,
b and c , we are talking about the two- or three-dimensional objects with at/straight sides
sketched in gure 2.3.
Area of a Parallelogram
Look at the parallelogram generated by vectors a and b in gure 2.3a. Clearly, the length of
the base of this rectangle is a , and, from basic trigonometry, we know the height h of this
parallelogram is
h = b sin() .
So, applying a little elementary geometry, we have
area of the parallelogram =
_
length of base
_ _
height
_
= a b sin()
= a b .
(Aside from the obvious uses of this fact to nd areas of parallelograms oating in space, this
fact can be used in nding the formula for the differential element of surface area using any
coordinate system.)
By the way, the vector a b is sometimes called the vector area of the parallelogram.
(The concept of vector area can be relevant in computations of certain surface integrals.)
Volume of a Parallelepiped
Now look at the parallelepiped generated by vectors a , b and c in gure 2.3b (the squished
box with every side being a parallelogram). Recall (from high school) that the volume of this
box is the area of its base times its height h as measured from bottom to top along a line
perpendicular to the bottom plane. From above we have that
area of the base = area of parallelogram generated by a and b = a b .
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 218
Since a b is orthogonal to both a and b, it is perpendicular to the plane containing a and b,
and, so, gives the direction for a line perpendicular to the bottom. Consequently, (see the gure)
the height h is simply the magnitude of the projection of vector c onto a b. Using either the
projection formulas from before or simple trigonometry, we have
h = c cos()
where is the angle between a b and c . With this and the geometric denition of the dot
product, we then get
volume of parallelepiped =
_
area of base
__
height
_
= a b c cos()
= |(a b) c| .
The expression (ab) c is called a triple product because it has three factors (even though
there are only two multiplies). Other triple products will be briey discussed later. For now,
however, observe that we have a certain symmetry in play here: It makes no difference which two
of the three vectors a , b and c are used to generate the bottom. Consequently, we automatically
have the identity
volume of parallelepiped = |(a b) c| = |a (b c)| = |(a c) b| .
In the next subsection, we will expand upon this observation, and then use it to derive another
fundamental identity which, in turn, will allow us to derive the component formula for the cross
product.
(By the way, we may later use the formulas just derived for area and volume to develop the
formulas for the integral elements of area and volume, d A and dV , using any coordinate system
inanyEuclideanor nonEuclideanspace. Withluck, we will evendiscuss ndinghyper-volumes
of arbitrary N-dimensional hyper-parallelepipeds, and use those formulas in multidimensional
integration in arbitrary coordinate systems. All that, however, goes beyond our traditional vector
theory.)
Linearity of the Cross Product
We start with an exercise:
?

Exercise 2.21: We just saw that, given three vectors a , b and c , then
|(a b) c| = |a (b c)| .
Thus, either
(a b) c = a (b c) or (a b) c = a (b c) . (2.1)
a: Nowassume that a , b and c are oriented so that the angle between each pair is no greater
than

/
2
( 90

), and that c is in the general direction of a b (as in gure 2.3b). Convince

yourself that, of the two equations in (2.1), it is
(a b) c = a (b c)
which holds.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 219
b: Now try different orientations for a , b and c and convince yourself that, no matter how
they are oriented with respect to each other,
(a b) c = a (b c) .
(See if you can come up with every possible orientation! Then you have a proof of this
identity!)
This last exercise demonstrates that, given any three vectors a , b and c (in a three-
dimensional traditional vector space), then
(a b) c = a (b c) . (2.2)
?

Exercise 2.22: Let v , w, b and c be any four vectors in in three-dimensional space

of traditional vectors. Using identity (2.2) and a result concerning the dot product (look at
exercise 2.15), show that
_
(v +w) b
_
c =
_
(v b) + (w b)
_
c .
(Do NOT assume (v +w) b = (v b) + (wb) ; thats what we are working to derive!)
Keep in mind that the equation derived in the last exercise holds for every possible c . If you
think about it (or recall exercise 2.18 on page 215), you will realize that could only mean that
(v +w) b = (v b) + (w b)
holds (this is why you did exercise 2.18). By the anticommutativity of the cross product, we also
have
b (v +w) = (v +w) b =
_
(v b) + (w b)
_
= (b v) + (b w) .
And fromthese identities (the distributive properties of the cross product), along with a fewmore
basic properties, you can derive a most general component formula for the cross product.
?

Exercise 2.23: Let B = {b

1
, b
2
, b
3
} be a basis for a three-dimensional space of traditional
vectors, and let
( v
1
, v
2
, v
3
) and ( w
1
, w
2
, w
3
)
be the components with respect to B of two vectors v and w. Show that
v w = A(b
1
b
2
) + B(b
1
b
3
) + C(b
2
b
3
)
where A , B and C are formulas of the v
j
s and w
k
s . Find those formulas.
A clever choice of basis, of course, can lead to a nicer component formula for the cross
product.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 220
Right-Handed Bases and the Classical Formula for the Cross
Product
A right-handed basis is an (ordered) orthogonal basis B = {b
1
, b
2
, b
3
} for a three-dimensional
traditional vector space in which b
3
points in the same direction as b
1
b
2
. In practice, we
typically restrict ourselves to orthonormal right-handed bases, in which case the requirement that
b
3
points in the same direction as b
1
b
2
simplies to
b
3
= b
1
b
2
.
Typically, also, orthonormal right-handed bases are denoted by either
{ e
1
, e
2
, e
3
} or { i, j, k } or { x, y, z } ,
or
{ e
1
, e
2
, e
3
} or {

i ,

j ,

k } or { x, y, z } ,
or even
{ e
1
, e
2
, e
3
} or {

i,

j,

k } or { x, y, z, } .
In the future, if a basis is referred to as a standard basis for a three-dimensional vector space
of traditional vectors, assume it is a right-handed orthonormal basis.
So assume { i, j, k } is a standard basis. By the above denition, k = i j . With a little
thought and a few pictures, you can easily conrm that
j k = i and k i = j .
Combining this with the formula you derived for v w (in exercise 2.23) yields the standard
component formula
v w =
_
v
2
w
3
v
3
w
2
_
i
_
v
1
w
3
v
3
w
1
_
j +
_
v
1
w
2
v
2
w
1
_
k
where, of course, (v
1
, v
2
, v
3
) and (w
1
, w
2
, w
3
) are the components of v and w with respect
to the given standard basis. Fortunately for those who dislike memorizing formulas with indices
in which order is important, this formula can also be written as the determinant of a more easily
remembered matrix,
v w = det
_
_
i j k
v
1
v
2
v
3
w
1
w
2
w
3
_
_
.
2.5 Triple Products
Any product of three vectors is called a triple product. If you think about it, you will realize that
the only legitimate triple products are of the form
(a b) c , (a b) c or a (b c) .
The rst one has already been discussed because of its relation to the volume of a parallelepiped
and a distributive law for the cross product. Since we already know that the cross product is not
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 221
associative, we known that the other two triple products above are not equal. However, we do
have
(a b) c = c (a b) ,
so any identity for one can easily be converted to an identity for the other.
One famous identity is the bac-cab rule,
a (b c) = b(a c) c(a b) .
This is signicant for at least three reasons:
1. It explicitly shows that a (b c) is a linear combination of b and c . (In fact, a good
derivation of the rule starts with the geometric derivation of this fact.)
2. It can be used to reduce the rather tedious computations of the triple cross product to a
much simpler set of computations with two dot products.
3. It has a neat name.
Arkin and Webers text has a reasonably good derivation of the bac-cab rule.
7
?

Exercise 2.24: What would be the cab-bac rule?

?

Exercise 2.25: Three vectors A, B and C are given in problem 1.5.10 on page 31 of our
text. Using these vectors:
a: Compute A(B C) just using the standard formula for the cross product to compute
both cross products.
b: Compute A (B C) using the bac-cab rule (or, in this case, the BAC-CAB rule).
c: Which way of computing A (B C) was easier?
2.6 Reciprocal Bases

Assume V is a traditional nite-dimensional vector space with basis

B = { b
1
, b
2
, b
3
, . . . , b
N
} .
The corresponding reciprocal basis is the set of vectors
B

= { b

1
, b

2
, b

3
, . . . , b

N
}
such that
b

j
b
k
=
j k
for all j and k .
7
In the sixth edition, this derivation starts at the bottom of page 27 and goes to the middle of page 29.

In traditional vector theory, the idea of reciprocal bases is mainly of just academic interest. However, this idea
will re-emerge when we develop tensor theory. In fact, it is this idea that actually leads to the distinction between
covariance and contravariance in tensor descriptions.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 222
In other words, for each index j , b

j
is the vector orthogonal to all the b
k
s except b
j
that also
satises
b

j
b
j
= 1 .
It is not hard to show that B

, as dened above, exists, is unique, and really is a basis for V .

It is also easy to verify that the reciprocal basis of B

is B. In fact, it would be more appropriate

to refer to B and B

as a reciprocal pair of bases.

As the following exercises show, there isnt much to nding the reciprocal basis B

if the
original basis B is orthogonal or even orthonormal.
?

Exercise 2.26: Let B and B

be as above. Convince yourself of the following:

a: If B is orthonormal, then
b

k
= b
k
for each k .
b: If B is orthogonal, then
b

k
=
b
k
b
k

2
for each k .
Finding the reciprocal basis corresponding to a nonorthogonal basis is a little more chal-
lenging.
?

Exercise 2.27: Let V be a two-dimensional traditional vector space with standard basis
{ i, j } , and let
B = { b
1
, b
2
}
be the basis with
b
1
= 3i and b
2
= i + 2j .
Find the reciprocal basis
B

= { b

1
, b

2
}
and sketch both the vectors in B and B

. (You may want to do the sketching rst.)

?

Exercise 2.28: Let V be a three-dimensional traditional vector space with standard basis
{ i, j, k } , and let
B = { b
1
, b
2
, b
3
}
be the basis with
b
1
= i + j , b
2
= 2j and b
3
= j 2k .
Find the reciprocal basis
B

= { b

1
, b

2
, b

3
}
and sketch both the vectors in B and B

. (You may nd the cross product useful in nding

the b

k
s .)
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 223
So what? Well, let v and w be two vectors in V . These vectors will have components
(v
1
, v
2
, . . . , v
N
) and (w
1
, w
2
, . . . , w
N
)
with respect to B, along with components
(v

1
, v

2
, . . . , v

N
) and (w

1
, w

2
, . . . , w

N
)
with respect to B

. That is,
v =
N

i =1
v
i
b
i
=
N

j =1
v

j
b

j
and
w =
N

m=1
w
m
b
m
=
N

n=1
w

n
b

n
.
From this, you can easily show that
v w =
N

j =1
v

j
w
j
and v =

_
N

j =1
v

j
v
j
even when B is not orthogonal.
?

Exercise 2.29: Verify the above formulas for v w and v .

That is the importance of the reciprocal basis. It lets us have a simple component formula
for the inner product and the norm, provided we use the two different bases for the components.
Using the above, you can also discover that the reciprocal basis gives an easy way to nd
components of vectors with respect to an arbitrary basis (provided youve found that reciprocal
basis).
?

Exercise 2.30: Let B and B

be a reciprocal pair of bases for an N-dimensional vector

space, and let v be a vector in this space. Show that each component of v with respect to
one of these bases can be expressed as an inner product of v with the corresponding element
of the other basis. More precisely, show that
v =
N

k=1
v
k
b
k
where v
k
= b

k
v
and
v =
N

k=1
v

k
b

k
where v

k
= b
k
v .
(Later, compare this with the approach described in the next chapter.)
Now, use these results in the next exercises.
version: 11/28/2011
Traditional Vector Theory Chapter & Page: 224
?

Exercise 2.31: Let B and B

be the reciprocal pair of basis from exercise 2.27, and nd

the components with respect to each of these bases for each of the following vectors:
i , j and 3i 5j .
?

Exercise 2.32: Let B and B

be the reciprocal pair of basis from exercise 2.28, and nd

the components with respect to each of these bases for each of the following vectors:
i , j and 3i 5j + 2k .
version: 11/28/2011