Physics I Course Material 1

1 Scalars and Vectors

Contents: Scalars and pseudo-scalars; Vectors and pseudo-vectors; dot-product
and the Einstein summation convention; cross-product and the Levi-Civita ten-
sor; applications.
In physics, we are interested in obtaining laws (in the form of mathematical
equations) which govern the behaviour of dierent systems. These laws should
hold at dierent places as well as times. They should also be independent of the
observer testing them. We thus infer that the physical laws and values of exper-
imentally measurable quantities must be frame independent, i.e., independent of
the choice of location of the origin of coordinates and the coordinate axes that
we choose. In order to check this, we need to be able to convert results in one
frame to equivalent results in another frame.
Suppose we are able to write the physical laws, i.e., the mathematical equa-
tions which describe them, in terms of objects that transform in a denite manner
under a change of frame, the rule for conversion is manifest. Examples of such
objects are scalars, vectors and more generally, tensors. When the equations are
written in terms of these objects, we say that the equations are form invariant.
For instance, an equation such as Newtons second law,

F = ma, has vectors
on both sides of the equation. So while two dierent observers have dierent
numerical values for each component of their vectors, they will both agree that
the equation holds.
However, the elementary or high-school denition of a vector (a quantity
with both magnitude and direction) is rather unsatisfactory. In this unit, we
shall rst provide a better denition of a vector through its properties under a
change in coordinate axes (change in basis vectors).
1.1 What is a vector?
The standard example of a vector in three dimensions is the position vector
r = x e
x
+ y e
y
+ z e
z
.
The three numbers (x, y, z) specify the vector fully. However, their actual nu-
merical values depend on the particular coordinate axes chosen: In other words,
a set ( e
x
, e
y
, e
z
) of mutually perpendicular unit vectors is chosen, and the values
(x, y, z) are with reference to this set
1
. This choice is not unique. A dierent
orthonormal basis will give rise to a dierent set of numbers, say, (x

, y

, z

) asso-
ciated with the same position vector. How are the two sets related? It is useful
to relabel the coordinates as
x
1
= x , x
2
= y , x
3
= z .
1
This is called an orthonormal basis. Each vector has unit magnitude, and they are mutually
orthogonal (perpendicular): e
x
e
y
= 0 etc.
2 V. Balakrishan & S. Govindarajan
The relation between the two sets of coordinates can be written as
x

i
=
3

j=1
R
ij
x
j
, (1.1)
where the nine numbers R
ij
are the coecients showing how the two sets of axes
are related. This can be rewritten in matrix form if we represent the coordinates
as a column vector and R
ij
as the entries in a matrix R with i as the row label
and j as the column label.

1
x

2
x

R
11
R
12
R
13
R
21
R
22
R
23
R
31
R
32
R
33

x
1
x
2
x
3

. (1.2)
We can show that the matrix R satises the conditions (R
T
is the transpose of
the matrix R)
R
T
= R
1
or R
T
R = R R
T
= I (1.3)
and det R = 1 , (1.4)
where I is the (33) unit matrix. The rst condition ensures that the magnitude
of vectors remain unchanged in all bases
2
. The second condition preserves ori-
entation, i.e., the basis vectors always form a right-handed triad. Such matrices
are called rotation matrices because two dierent choices of basis vectors are
related by rotations. For instance, consider the case when two bases have one
common basis vector, e

z
= e
z
, i.e., they have the same z-axis. The coordinates
are related by

cos sin 0
sin cos 0
0 0 1

x
y
z

(1.5)
i.e., the two x- and y-axes are related by a rotation by an angle about the
z-axis.
We can now generalise this to arbitrary rotations. Arbitrary rotations are
specied by an axis of rotation n and an angle of rotation about the direction
of n. Let us denote the corresponding rotation matrix by R( n, ). Thus, the
rotation matrix given above corresponds to R( e
z
, ). One can give an explicit
form for the general rotation matrix but we shall not do so now.
We can now provide the denition of a vector. A vector is a set of three
quantities that transform, under rotations of the coordinate axes, ex-
actly as the set of coordinates itself transform.. We usually represent the
2
Actually it also ensures that the new basis vectors are unit vectors and orthogonal to each
other.
Physics I Course Material 3
three quantities as a column vector, v =

v
1
v
2
v
3

. Under the change of coordinates

given by Eqn.(1.1), the components transform as
v

i
=
3

j=1
R
ij
v
j
. (1.6)
Exercise Given a vector v, nd a rotation matrix R (or equivalently, a change
of basis) such that v

= (0, 0, v)
T
. Is R unique? Explain.
1.2 The Einstein summation convention
This convention is very useful in writing in a compact and uncluttered form,
formulae involving several summations. The convention is simply that any index
that occurs twice is summed over. The summation always runs from 1 to 3 since
we are in three dimensions. Further, no index can occur more than twice in any
equation. Thus, we can rewrite Eqn. (1.6) as
v

i
= R
ij
v
j
(1.7)
where the summation of j is implicitly assumed under the Einstein summation
convention and i is a free (i.e., unsummed) index. The index that is summed
over is referred to as a dummy index. Care should be taken to see that dierent
letters are used to indicate dierent summations. Further, a simple check is
to see that the same free indices occur on either side of any equation.
1.3 The dot-product: Scalars from vectors
A scalar is dened as an object that is unchanged or invariant under change of
basis/rotations.
Given two vectors v = (v
1
, v
2
, v
3
)
T
and w = (w
1
, w
2
, w
3
)
T
, we can construct a
scalar from them. It is called the dot product, and is dened by
v w v
i

ij
w
j
, (1.8)
where

ij
=

0 i = j
1 i = j
is the Kronecker delta. When it is written as a 3 3 matrix, it is nothing but
the identity matrix I. Thus, it is not hard to see that
v w = w v = v
i
w
i
(1.9)
4 V. Balakrishan & S. Govindarajan
(Note the summation over i as per our convention.) The invariance of the dot-
product follows from the condition R
T
R = I satised by rotation matrices. Ex-
plicitly, this condition is
R
T
R = I R
ji
R
jk
=
ik
.
Here i and k are free indices and the dummy index j is summed over from 1 to
3. The following steps show how the dot product is unchanged when going from
one basis to another
v

j
w

j
= R
ji
v
i
R
jk
w
k
= R
ji
R
jk
v
i
w
k
=
ik
v
i
w
k
= v
i
w
i
The dot-product when we choose v = w gives the square of the magnitude of
a vector. That is
|v|
2
= v
i
v
i
= v
2
1
+ v
2
2
+ v
2
3
, (1.10)
which is clearly positive denite. Note that |v| = 0 implies v = 0. For any two
vectors u and v, we can show that
u v = |u||v| cos ,
where is the angle between the two vectors u and v. Recall that in elementary
vector analysis this formula was used to dene the scalar product. We now see
that it follows as a consequence of the denition u v = u
i
v
i
.
1.4 The cross-product: New vectors from old
Given two vectors v and w, one can construct a new vector u by the cross
product dened below: the components of the cross product are
u
1
= v
2
w
3
v
3
w
2
u
2
= v
3
w
1
v
1
w
3
(1.11)
u
3
= v
1
w
2
v
2
w
1
,
written symbolically as u = v w. Eqs. (1.12) can be written in compact form
as
u
i
=
ijk
v
j
w
k
(v w)
i
(1.12)
where
ijk
is the Levi-Civita tensor (also called the permutation symbol)
dened as: the conditions:

123
= 1

ijk
=
jik
=
ikj
=
kji
The second condition implies that the Levi-Civita tensor is totally antisymmetric
and is non-vanishing if and only if all three indices are distinct. Thus,
123
=
231 =
312
= 1 and
132
=
321
=
213
= 1. All other components are vanishing.
Physics I Course Material 5
For any two vectors u and v, one can show that
|u v| = |u||v| sin ,
where is the angle between the two vectors u and v. Clearly, u u = 0.
1.5 More complicated objects: Tensors
As mentioned earlier, scalars and vectors are not the only kinds of objects that
one encounters in physical situations. A simple example of a more complicated
object - a tensor, is given by A
ij
= v
i
w
j
, where v and w are any two vectors.
Under a change of basis, one can see that
A

ij
= v

i
w

j
= R
ik
R
jl
v
k
w
l
= R
ik
R
jl
A
kl
.
A tensor like A
ij
can be represented as a 3 3 matrix A. Then, the above
transformation law can be written as
A

= R
T
A R (1.13)
Objects which transform in this fashion are called tensors of rank two (because
its components are specied by two indices i and j). Not all tensors of rank two
can be written as v
i
w
j
and thus one can use the above equation as the dening
condition. An example of such an object is given by the moment of inertia of a
rigid body. Other examples include the stress and strain tensors, the dielectric
tensor and so on.
Clearly, a tensor of second rank is the simplest generalisation of a vector.
Generalising further, tensors of rank r are objects which have r-indices. Thus,
scalars are tensors of rank zero and vectors are tensors of rank one. The
Levi-Civita tensor has rank three and the Kronecker delta has rank two. The
elastic constants that relate stress and strain in linear materials is a tensor of
rank four.
Exercise Using the transformation law for a tensor of rank two, show that the
Kronecker delta is invariant under rotations of the coordinate axes.
Exercise Using the transformation law for a tensor of rank three and the anti-
symmetry of the Levi-Civita tensor, show that

ijk
= (detR)
ijk
.
Thus, the Levi-Civita tensor is invariant under rotations since detR = 1.
6 V. Balakrishan & S. Govindarajan
1.6 An important identity
As we just saw, the Kronecker delta and Levi-Civita tensor are invariant under
change of basis. They are related by the following important identity which
enables one to derive just about all vector identities:

ijk

ilm
= (
jl

km

jm

kl
) (1.14)
Here we must remember that the index i is repeated, and so a summation over
i from i = 1 to i = 3 is implicit. The most straightforward proof is to explicitly
verify that this is true for all values of the free indices (j, k, l, m). This is left
to the student. However, some simple checks are: both the LHS and RHS are
antisymmetric under the exchanges: j k and k l.
As an application of the identity (1.14), consider the multiple cross-product:
u (v w). The i-th component of this is
(u (v w))
i
=
ijk
u
j
(v w)
k
=
ijk
u
j

klm
v
l
w
m
=
kij

klm
u
j
v
l
w
m
Using the identity (1.14), we obtain
(u (v w))
i
= (
il

jm

im

jl
) u
j
v
l
w
m
= u
m
v
i
w
m
u
j
v
j
w
i
which can be rewritten as
(u (v w)) = (u w)v (u v) w (1.15)
One can further show that
(u v) w = u (v w) ,
i.e., the order of the cross-product is important. The student may explicitly verify
this.
1.7 Reections and Parity
We saw that rotations are given by matrices that satisfy R
T
= R
1
and detR = 1.
Are there transformations which satisfy only the rst of these two conditions?
Taking the determinant on both sides, we nd that detR
2
= 1, which suggests
that one consider transformations such that detR = 1. It is not hard to see
that reection about any plane is such an operation. Under reections about the
xy-plane, one obtains

x
y
z

.
Physics I Course Material 7
The matrix R = diag(1, 1, 1) clearly has determinant 1. Another example, is
the parity operation under which all coordinates change sign:
P :

x
y
z

x
y
z

(1.16)
One can show that any matrix satisfying R
T
= R
1
with determinant 1 can be
obtained by following the parity operation with a rotation. Thus, it suces to
consider parity alone.
A scalar is invariant under both rotations and parity while a pseudo-scalar
is one that is invariant under rotations but changes sign under parity.
A vector is one that transforms identical to the position vector under both
rotations as well as parity while a pseudo-vector transforms like the position
vector under rotations but is invariant under parity.
The momentum p of a particle is a vector, while its angular momentum

L is a
pseudo-vector. This follows from the fact that both r and p are vectors and hence
change sign under parity. Thus

L = r p does not change sign, and hence is a
pseudo-vector. In general, the cross-product of two vectors gives rise to a pseudo-
vector, while the cross-product of a vector with a pseudo-vector is a vector. The
dot product of two vectors is a scalar (e.g. the kinetic energy of a particle), while
the dot product of a vector and a pseudo-vector is a pseudo-scalar. Vectors are
sometimes called polar vectors, to distinguish them from axial vectors which
is another name for pseudovectors.
The Lorentz force law

F = q(

E +v

B)
enables us to x the behaviour of the electric and magnetic elds under parity.
Both

F and v are vectors and q is a scalar, it follows that the electric eld is a
vector while the magnetic eld is a pseudo-vector. We usually decide if an
object is a pseudo-vector or pseudo-scalar by considering simple known equations.
Never equate a vector to a pseudo-vector or a scalar to a pseudo-scalar.
1.8 Form invariance of physical laws
The full power of working with objects which transform nicely under change of
basis/rotations is that the equations which describe the physical motion
of particles can be written in manifestly invariant manner i.e., they retain
the same form in dierent bases. For instance, Newtons law

F = ma
is the same in all bases. The LHS is a vector and the RHS is the product of a
scalar and a vector. Hence the RHS is a vector as well.
8 V. Balakrishan & S. Govindarajan
In more complicated situations, the equations need not be relations between
vectors but those between tensors of the same rank. That way, invariance is
guaranteed since both sides change in an identical fashion under change of bases
and reections. The simple way to verify this is to check that free indices are
the same on both sides of any equation. One must also be cautious and check
that the dummy (non-free) indices occur only twice. If any index occurs three or
more timee, theres been a mistake somewhere.
1.9 Geometric meaning of the dot product and cross prod-
uct
1.9.1 Area element as a vector
Given two vectors u and v, one can construct a parallelogram with two edges u
and v(see Fig. 1.1). By considerations of elementary geometry, the area of the
!

! #$%
&
Figure 1.1: Area of a parallelogram
parallelogram is
Area = |u||v| sin ,
where is the angle between the two vectors u and v. Thus, one can see that the
area can be written as Area = |u v|.
It turns out that it is better to think of the area of a parallelogram (a planar
gure) as the vector (u v). Clearly, the direction of this vector is normal to
the surface of the parallelogram. We thus dene the area vector to be

A
= (u v) . (1.17)
Note that there is an ambiguity of an overall sign in the denition, for we could
also have chosen it to be (v u). We will return to this point when discussing
the unit vector perpendicular to an area element.
1.9.2 Volume as a scalar
Just as two vectors can dene a parallelogram, three (non co-planar) vectors u,
v and w can dene a parallelepiped which has the three vectors as three of its
Physics I Course Material 9
&
!
'
Figure 1.2: Volume of a parallelepiped
edges (see Fig. 1.2). Again by means of elementary geometry, one can show that
the volume of the parallelepiped is given by
Volume = u (v w) . (1.18)
The above expression seems to treat the three vectors dierently. but this is not
so. We can re-express the above formula by using the Levi-Civita tensor
Volume =
ijk
u
i
v
j
w
k
= u (v w) = v ( w u) = w (u v) (1.19)
We note again that there is a sign ambiguity in the above denition. This can
be removed by taking the magnitude in the RHS of the above equation, since we
require the volume to be non-negative. Note that if u, v, w are co-planar, the
volume of the parallelepiped vanishes automatically, as it must.