Introduction to Tensors

Sumeet Khatri
Table of Contents
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
List of Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
1 The Index Notation and Einstein Summation Convention . . . . . . . . . . . 1
2 Covariant and Contravariant Vectors . . . . . . . . . . . . . . . . . . . . . . . . . 7
3 Introducing Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1 The Inner Product and the First Tensor . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Creating Tensors from Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4 Tensor Definition and Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.1 Symmetry and Anti-Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2 Contraction of Indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
i
List of Figures
ii
List of Tables
iii
List of Theorems
iv
Introduction
Tensors are geometric objects that describe linear relations between vectors, scalars, and other
tensors. Elementary examples of such relations include the dot product, the cross product, and
linear mappings. We will see that in fact vectors and scalars are also tensors.
Tensors are important in physics because they provide a concise mathematical framework for
formulating and solving physics problems in areas such as elasticity, fluid mechanics, and special
and general relativity.
v
1 The Index Notation and Einstein Summation Convention
Let us first introduct a new notation for vectors and matrices and their algebraic manipulations,
called the index notation.
Let us take a manifold with dimension n. We will denote the components of a vector v with the
numbers v
1
, v
2
, . . . , v
n
in some basis {e
1
, e
2
, . . . , e
n
}. If one modifies the vector basis in which the
components of v are expressed, then these components will also change. Such a transformation
is often called a “change-of-basis matrix”, say A, in which the columns are the old basis vectors
e
1
, e
2
, . . . , e
n
expressed in the new basis, say {e

1
, e

2
, . . . , e

n
}. So we have
_
_
_
_
_
v

1
v

2
.
.
.
v

n
_
_
_
_
_
=
_
_
_
A
11
. . . A
1n
.
.
.
.
.
.
A
n1
. . . A
nn
_
_
_
_
_
_
_
_
v
1
v
2
.
.
.
v
n
_
_
_
_
_
, (1.1)
taking note of the fact that the first index denotes the row of the matrix A and the second index
the column.
According to the rules of matrix multiplication, the above matrix equation is the system of
equations
v

1
= A
11
v
1
+ A
12
v
2
+· · · + A
1n
v
n
,
.
.
.
v

n
= A
n1
v
1
+ A
n2
v
2
+· · · + A
nn
v
n
,
(1.2)
or equivalently,
v

1
=
n

ν=1
A

v
ν
,
.
.
.
v

n
=
n

ν=1
A

v
ν
,
(1.3)
or even more succinctly,
v

µ
=
n

ν=1
A
µν
v
ν
(∀µ ∈ N, 1 ≤ µ ≤ n) . (1.4)
1
Introduction to Tensors The Index Notation and Einstein Summation Convention
Each of the three systems above are written in the index notation. In (1.4), we call ν a dummy
index and µ a running index or a free index. Keep in mind that µ and ν are merely
labels—we could have equally well called them whatever else we like, say α and β.
Usually, the condition for µ in (1.4) are not explicitly stated because they should be obvious
from the context. We therefore have
v = y ⇔ v
µ
= y
µ
v = Ay ⇔ v
µ
=
n

ν=1
A
µν
y
ν
.
(1.5)
The index notation is also applicable to operation such as the dot product (and indeed inner
products in general), so that if v and w are any two vectors, then
v • w = v
1
w
1
+ v
2
w
2
+· · · + v
n
w
n
=
n

µ=1
v
µ
w
µ
. (1.6)
We also have
C = A + B ⇔ C
µν
= A
µν
+ B
µν
z = v +w ⇔ z
µ
= v
µ
+ w
µ
.
(1.7)
Example 1.0.1: Working with index notation.
1. A, B, and C are matrices of appropriate dimensions. Assume that A = BC. Write
out this matrix multiplication using index notation.
2. A and B are matrices and x is a vector. Show that
n

ν=1
A
µν
_
n

α=1
B
να
x
α
_
=
n

ν=1
n

α=1
(A
µν
B
να
x
α
)
=
n

α=1
n

ν=1
(A
µν
B
να
x
α
)
=
n

α=1
_
n

ν=1
(A
µν
B
να
) x
α
_
.
3. Which of the following statements is true?
(a) The summation signs in an expression can always be moved to the far left without
changing the meaning of the expression.
(b) If all summation signs are on the far left of an expression, you can exchange their
order without changing the meaning of the expression.
2
Introduction to Tensors The Index Notation and Einstein Summation Convention
(c) If all summation signs are on the far left of an expression, you cannot just change
the order of the variables in the expression because this changes the order in which
matrices are multiplied, and generally AB = BA for any two arbitrary matrices
A and B of appropriate dimensions.
(d) A
µν
=
_
A
T
_
νµ
.
(e) A
µν
=
_
A
T
_
µν
.
Solution:
1. Let
B =
_
_
_
B
11
. . . B
1n
.
.
.
.
.
.
B
m1
. . . B
mn
_
_
_ and C =
_
_
_
C
11
. . . C
1m
.
.
.
.
.
.
C
n1
. . . C
nm
_
_
_.
Then,
A =
_
_
_
B
11
C
11
+· · · + B
1n
C
n1
B
11
C
12
+· · · + B
1n
C
n2
. . . B
11
C
1m
+· · · + B
1m
C
nm
.
.
.
.
.
.
B
m1
C
11
+· · · + B
mn
C
n1
B
m1
C
12
+· · · + B
mn
C
n2
. . . B
m1
C
1m
+· · · + B
mn
C
nm
_
_
_,
so that
A
µν
=
n

i=1
B
µi
C

.
2. We have
n

ν=1
A
µν
_
n

α=1
B
να
x
α
_
= A
µ1
_
n

α=1
B

x
α
_
+· · · + A
µn
_
n

α=1
B

x
α
_
=
n

α=1
A
µ1
B

x
α
+
n

α=1
A
µ2
B

x
α
+· · · +
n

α=1
A
µn
B

x
α
=
n

ν=1
n

α=1
A
µν
B
να
x
α
,
proving the first equality. The second equality follows from the commutativity of
addition (i.e., elements can be added in different orders without altering the result),
and
n

α=1
_
n

ν=1
(A
µν
B
να
) x
α
_
=
n

α=1
n

ν=1
(A
µν
B
να
) x
α
=
n

α=1
n

ν=1
A
µν
B
να
x
α
.
3.
3
Introduction to Tensors The Index Notation and Einstein Summation Convention
(a) True, since doing so simply changes the order in which elements are multiplied,
and since addition of numbers is commutative, this is no problem.
(b) True, since this also merely changes the order of addition.
(c) False as long as the summand contains expression involving commutative opera-
tors, like multiplicaiton. As for the matrix multiplication mentioned, it is not the
order of the multiplication/addition of elements that makes matrix multiplication
non-commutative but rather the definition of matrix multiplication.
(d) True
(e) False
We have seen in the above example that the summation symbol can always be put at the start of
any expression and that if there is more than one summation sign then their order is irrelevant.
It is therefore convenient to omit the summation signs as long as we make it clear in advance
which index is being summed over, for instance, by putting it beside the formula as shown below:
n

ν=1
A
µν
v
ν
→ A
µν
v
ν
{ν}
n

β=1
n

γ=1
A
αβ
B
βγ
C
γδ
→ A
αβ
B
βγ
C
γδ
{β, γ} .
(1.8)
The above example also indicates to us the following:
• It appears that if an index only appears once in a summand then this index is not summed;
• It appears that if an index appears at least twice in a summand then that index is summed.
After making routine usage of index notation, however, indicating the summation index every
time might become irritating. Our two observations above will eventually allow us to easily
determine the summation index (or indices), so that indicating it no longer becomes necessary.
This leads to the Einstein Summation Convention:
Einstein Summation Convention
In a summation over one or more indices, the summation sign and the
index of summation may be omitted with the following conventions:
• A summation is assumed over all indices that appear twice in a
summand; and
• No summation is assumed over indices that appear only once in the
summand.
4
Introduction to Tensors The Index Notation and Einstein Summation Convention
We will use index notation with Einstein summation convention from now on. So we will write
n

ν=1
A
µν
v
ν
→ A
µν
v
ν
n

β=1
n

γ=1
A
αβ
B
βγ
C
γδ
→ A
αβ
B
βγ
C
γδ
.
(1.9)
Also,
n

i=1
A
i
A
i
→ A
i
A
i
n

i=1
n

j=1
A
ijk
B
ij
→ A
ijk
B
ij
,
(1.10)
and by extension we shall also understand summation in such expressions as
∂u
i
∂x
i
,
∂q
∂x
i
dx
i
dt
, etc. (1.11)
Example 1.0.2: Working with the Einstein summation convention.
1. Write as matrix multiplication:
(a) D
αβ
= A
αµ
B
µν
C
βν
;
(b) D
αβ
= A
αµ
B
βγ
C
µγ
;
(c) D
αβ
= A
αγ
(B
γβ
+ C
γβ
).
2. Consider a vector field in an n-dimensional space, F(x). We perform a coordinate
transformation x

= Ax, where A is a n×n change-of-basis matrix. Show that F

= AF.
3. For a change of basis, we have x

= Ax. This corresponds to x

µ
=

n
ν=1
A
µν
x
ν
. Can
you understand the expression

n
ν=1
x
ν
A
µν
, and how can you construct the matrix
multiplication equivalent of

n
µ=1
x
µ
A
µν
?
Solution:
1. (a) We have by the summation convention
D
αβ
=
n

ν=1
n

µ=1
A
αµ
B
µν
C
βν
=
n

ν=1
(AB)
αν
(previous example)
=
n

ν=1
(AB)
αν
_
C
T
_
νβ
(previous example),
so that D = ABC
T
.
5
Introduction to Tensors The Index Notation and Einstein Summation Convention
(b) By the summation convention, and since the order of elements in a summand is
irrelevant (as multiplication is commutative), we have
D
αβ
=
n

γ=1
n

µ=1
A
αµ
C
µγ
B
βγ
=
n

γ=1
(AC)
αγ
B
βγ
=
n

γ=1
(AC)
αγ
_
B
T
_
γβ
,
so that D = ACB
T
.
(c) We have
D
αβ
= A
αγ
B
γβ
+ A
αγ
C
γβ
= (AB)
αβ
+ (AC)
αβ
,
so that D = AB + AC = A(B + C).
2.
3. Because we can change the order of elements in the summand, we have simply
n

ν=1
x
ν
A
µν
=
n

ν=1
A
µν
x
ν
= x

µ
.
And if we let
x =
_
_
_
_
_
x
1
x
2
.
.
.
x
n
_
_
_
_
_
=
_
x
1
x
2
. . . x
n
_
T
,
then
x
T
A =
_
x
1
x
2
. . . x
n
_
_
_
_
A
11
. . . A
1n
.
.
.
.
.
.
A
n1
. . . A
nn
_
_
_ =
_
_
_
_
_
x
1
A
11
+ x
2
A
21
+· · · + x
n
A
n1
x
1
A
12
+ x
2
A
22
+· · · + x
n
A
n2
.
.
.
x
1
A
1n
+ x
2
A
2n
+· · · + x
n
A
nn
_
_
_
_
_
,
so that the matrix multiplication equivalent of

n
µ=1
x
µ
A
µν
is x
T
A.
6
2 Covariant and Contravariant Vectors
In this chapter we will describe how vectors change under a coordinate transformation, i.e.,
under a change of basis. Doing this will allow us to make a distinction between two types of
vectors, which we will call contravariant vectors and covariant vectors (sometimes shortened to
“covector”).
In physics, a vector typically arises as the outcome of a measurement or series of measurements
and is represented as a tuple of numbers, such as (v
1
, v
2
, v
3
). This tuple of numbers, each of
which is called a coordinate, depends on the choice of coordinate system. Let us assume that we
use a linear coordinate system, so that we can use linear algebra to describe it. The position of
physical objects, for example, can be specified using a Cartesian coordinate system and are often
represented as an arrow from its origin. We can then use a chosen set of basis vectors belonging to
the coordinate system, for example, the standard basis {i = (1, 0, 0) , j = (0, 1, 0) , k = (0, 0, 1)}
in R
3
, specify the location of the object as r
1
i + r
2
j + r
3
k, so that (r
1
, r
2
, r
3
) is the 3-tuple of
the coordinates of the object.
In such a description of objects with coordinates, we must be fully aware that the coordinates
themselves have no meaning. Only with the corresponding basis vectors do these numbers
acquire meaning. Remember that the object one describes is independent of the coordinate
system (and hence the set of the basis vectors) chosen. We are thus interested in how the
coordinates of an object transform when the original coordinate system is changed to a new
one.
Now, suppose we have two bases in three-dimensional vector space V (indeed, we could generalise
this to n dimensions, but this makes the subsequent equations less cumbersome to write down),
{e
1
, e
2
, e
3
} and {e

1
, e

2
, e

3
}. Suppose every basis vector in the primed basis can be written as a
linear combination of the basis vectors of the unprimed basis, i.e.,
e

1
= a
11
e
1
+ a
12
e
2
+ a
13
e
3
,
e

2
= a
21
e
1
+ a
22
e
2
+ a
23
e
3
,
e

3
= a
31
e
1
+ a
32
e
2
+ a
33
e
3
,
(2.1)
or
_
_
e

1
e

2
e

3
_
_
= Λ
_
_
e
1
e
2
e
3
_
_
, where Λ =
_
_
a
11
a
12
a
13
a
21
a
22
a
23
a
31
a
32
a
33
_
_
, (2.2)
where we assume that Λ is non-singular (and hence invertible).
Now, let us take a vector v = (v
1
, v
2
, v
3
) in the unprimed basis and see how its coordinates
transform when written in terms of the primed basis. From linear algebra, we know that the
change-of-basis matrix is constructed by letting the columns be the old basis vectors expressed
7
Introduction to Tensors Covariant and Contravariant Vectors
in terms of the new ones, i.e., the columns will be unprimed basis vectors expressed in terms of
the primed basis vectors. Now, since Λ above is invertible, we have
_
_
e
1
e
2
e
3
_
_
= Λ
−1
_
_
e

1
e

2
e

3
_
_
,
and hence the columns of
_
Λ
−1
_
T
will contain unprimed basis vectors in terms of the primed
basis vectors.
Remark: To see why we must take the transpose, note that
Λ
T
=


a11 a21 a31
a12 a22 a32
a13 a23 a33


,
so that the columns of Λ
T
are the primed basis vectors expressed in terms of the unprimed basis vectors. In
the same way, Λ
−1
will have as it rows the coordinates of the unprimed basis vectors in terms of the primed
basis vectors, so that its transpose will contain these coordinates as its columns, as required.
Therefore,
v

=
_
Λ
−1
_
T
v ⇒ v

µ
=
_
Λ
−1
_
T
v
µ
. (2.3)
Vectors that transform in this manner are called contravariant vectors and the transformation
_
Λ
−1
_
T
represents a contravariant transformation. These are the vectors that we typically
deal with, which is why we almost always simply call them “vectors”. Observe that the coor-
dinates of v transform in the opposite way (i.e., “contrary”) as the basis vectors that describe
it. This means that the vector itself does not change, i.e., if we use an arrow to indicate our
vector, then physically this arrow will be unchanged, as we require. Instead, the components of
the vector make a change that cancels the change in the basis vectors, resulting in a change of
coordinates. In other words, if the basis vectors were rotated in one direction, the component
representation of the vector would rotate in exactly the same way (with the effect seen in the
values of the coordinates). Similarly, if the basis vectors were stretched in one direction, the
components of the vector, like the coordinates, would reduce in an exactly compensating way.
Let us now look at vectors that transform in the same way as the basis vectors. Consider an
n-dimensional manifold V with coordinates x
1
, x
2
, . . . , x
n
. Let f be some scalar function. Then
the gradient of f (x
1
, x
2
, . . . , x
n
) is
_

∇f
_
µ
=
∂f
∂x
µ
, i.e.,

∇f =
∂f
∂x
1
e
1
+
∂f
∂x
2
e
2
+· · · +
∂f
∂x
n
e
n
. (2.4)
Suppose we have a vector field defined on this manifold V , V = V(x). Let us perform a
homogeneous linear transformation of the coordinates:
x

µ
= A
µν
x
ν
. (2.5)
8
Introduction to Tensors Covariant and Contravariant Vectors
As we saw in the previous example, we thus have a corresponding change in the vector field V:
V

µ
(x) = A
µν
V
ν
(x) , (2.6)
where A is the same matrix as in (2.5). Note that this matrix describes the transformation of
the vector components, while previously our matrix Λ described the transformation of the basis
vectors, so that A =
_
Λ
−1
_
T
.
Now, take the function f (x
1
, x
2
, . . . , x
n
) and the gradient w
α
at a point P as so,
w
α
=
∂f
∂x
α
; (2.7)
and in the new coordinate system,
w

α
=
∂f
∂x

α
. (2.8)
(i.e., the w
α
are the components of the gradient vector) Then, by the chain rule,
∂f
∂x

1
=
∂f
∂x
1
∂x
1
∂x

1
+
∂f
∂x
2
∂x
2
∂x

1
+· · · +
∂f
∂x
n
∂x
n
∂x

1
,
that is,
∂f
∂x

µ
= w

µ
=
∂f
∂x
ν
∂x
ν
∂x

µ
= w
ν
∂x
ν
∂x

µ
⇒ w

µ
=
_
∂x
ν
∂x

µ
_
w
ν
.
(2.9)
Now, take (2.5) and rewrite it as
x
µ
=
_
A
−1
_
µν
x

ν
.
Then,
∂x
µ
∂x

α
=

_
_
A
−1
_
µν
x

ν
_
∂x

α
=
_
A
−1
_
µν
∂x

ν
∂x

α
+

_
A
−1
_
µν
∂x

α
x

ν
. (2.10)
Because in this case A does not depend on x

α
, the last term on the right-hand side of the above
equation vanishes. Also,
∂x

ν
∂x

α
= δ
να
, δ
να
=
_
1 when ν = α,
0 when ν = α.
(2.11)
Therefore, what remains is
∂x
µ
∂x

α
=
_
A
−1
_
µν
δ
να
=
_
A
−1
_
µα
. (2.12)
9
Introduction to Tensors Covariant and Contravariant Vectors
Finally, in combination with (2.9), we get the following transformation of the components of the
gradient:
w

µ
=
_
A
−1
_
T
µν
w
ν

_

∇f
_

=
_
A
−1
_
T

∇f. (2.13)
But remember that A =
_
Λ
−1
_
T
, so that
_
A
−1
_
T
= Λ. Therefore,
_

∇f
_

= Λ
_

∇f
_
, (2.14)
i.e., we have shown that the components of the gradient vector transform in exactly the same
way as the basis vectors. Vectors that transform in this manner are called covariant vectors
or simply covectors, and the matrix Λ represents a covariant transformation.
To distinguish contravariant vectors from covariant vectors, we will write the indices of con-
travariant vectors as a superscript and the indices of covariant vectors as subscripts.
y
α
: contravariant vector
w
α
: covariant vector
In addition, we will denote contravariant vectors in boldface (v) and write then explicitly as
column matrices, as we have been doing all along,
v =
_
_
_
_
_
v
1
v
2
.
.
.
v
n
_
_
_
_
_
(contravariant vector), (2.15)
and we will denote covariant vectors in boldface with a tilde (˜ v) and write them explicitly as
row matrices,
˜ v =
_
v
1
v
2
. . . v
n
_
(covariant vector). (2.16)
(Note the position of the index in both cases, following the convention in the box above.) We
also introduce a similar notation convention for matrices, which we will regard as an extension
of the Einstein summation convention. Instead of the usual index notation A
mn
used to refer to
the mth row and nth column of a matrix A, we will write A
m
n
, which means that the transpose
is
_
A
T
_
m
n
= A
n
m
. Then the transformation rules for contravariant vectors and covariant vectors,
respectively, are
v

µ
= A
µ
ν
v
ν
(contravariant vectors),
w

µ
=
_
A
−1
_
µ
ν
w
ν
(covariant vectors).
(2.17)
As for matrix multiplication, we get
(AB)
i
k
= A
i
j
B
j
k
. (2.18)
10
Introduction to Tensors Covariant and Contravariant Vectors
This new notation will be useful later because it will indicate that such matrices have mixed
contravariant and covariant transformation properties.
11
3 Introducing Tensors
3.1 The Inner Product and the First Tensor
The dot product is very important in physics. In classical mechanics, for example, we have that
the work that is done when an object is moved equals the dot product of the force F acting on
the object and the displacement vector x of the object: W = F • x. As we know from linear
algebra, the dot product is just a special case of the inner product (the dot product is often
called the standard inner product on R
n
), so we might also write W = F, x. The work must
of course be independent of the coordinate system in which the vectors F and x are expressed.
However, the dot product
s = a, b = a
µ
b
µ
does not in general have this invariance property for arbitrary vectors a and b and arbitrary
linear transformations a

µ
= A
µ
α
a
α
and b

µ
= A
µ
β
b
β
:
s

=
¸
a

, b

_
= A
µ
α
a
α
A
µ
β
b
β
=
_
A
T
_
α
µ
A
µ
β
a
α
b
β
.
So we see that s

= s if and only if A
−1
= A, i.e., if and only if we are dealing with orthog-
onal transformation (i.e., A is an orthogonal matrix). However, we would like s = s

for any
transformation matrix A. To try to accomplish this, notice that the dot product between a
(contravariant) vector x and a covector y, s = x
µ
y
µ
, is invariant under all transformations since
for all transformation matrices A
s

= x

µ
y

µ
= A
µ
α
x
α
_
A
−1
_
β
µ
y
β
=
_
A
−1
_
β
µ
A
µ
α
x
α
y
β
= δ
β
α
x
α
y
β
= s.
With the help of this dot product, we can introduce a new standard inner product between two
contravariant vectors that also has the invariance property. Let us define the inner product as
s = g
µν
x
µ
y
ν
, (3.1)
where, in R
3
,
g =
_
_
g
11
g
12
g
13
g
21
g
22
g
23
g
31
g
32
g
33
_
_
. (3.2)
Now, we must make sure that this object g is chosen so that our new inner product reproduces
the old one if we choose an orthonormal coordinate system. So, in R
3
, we should get
s = g
µν
x
µ
y
ν
=
_
x
1
x
2
x
3
_
_
_
g
11
g
12
g
13
g
21
g
22
g
23
g
31
g
32
g
33
_
_
_
_
y
1
y
2
y
3
_
_
= x
1
y
1
+ x
2
y
2
+ x
3
y
3
(in an orthonormal system).
12
Introduction to Tensors Introducing Tensors
This implies that
g
µν
=
_
_
1 0 0
0 1 0
0 0 1
_
_
in an orthonormal coordinate system. (3.3)
Note, however, that g
µν
does not have the transformation properties of an ordinary matrix.
Remember that the matrix A of the previous chapter had one index up and one index down,
like A
µ
ν
, indicating that it has mixed contravariant and covariant transformation properties.
This new object g
µν
, however, has been written with both indices down, so it transforms in a
covariant. This object, which looks like a matrix but does not transform like one, is an example
of a tensor. A matrix is also a tensor, as are vectors and covectors. Matrices, vectors, and
covectors are special cases of the more general class of objects called “tensors”. The object g
µν
is a kind of tensor that is neither a matrix nor a vector nor a covector. It is a new kind of object
for which only tensor mathematics has a proper description. It is called a metric tensor or
simply a metric.
3.2 Creating Tensors from Vectors
We have seen that the inner product of a vector with a covector is
s = x
µ
y
µ
.
In this case the indices are paired, indicating by the Einstein convention a summation over all
possible values of the index. We can also multiply vectors and covectors without pairing the
indices, and therefore without summation. For example, in three dimensions, we get
s
µ
ν
= x
µ
y
ν
=
_
_
x
1
y
1
x
1
y
2
x
1
y
3
x
2
y
1
x
2
y
2
x
2
y
3
x
3
y
1
x
3
y
2
x
3
y
3
_
_
.
This object still looks very much like a matrix, since a matrix is also nothing more or less than an
array of numbers labelled with two indices. To check if this is a true matrix, or something else, we
need to see how it transforms. From linear algebra, we know that if A is a matrix representing a
linear mapping, and S is a change-of-basis matrix (from unprimed to prime coordinate systems),
then A

= SAS
−1
, where A

represents the matrix A in the primed coordinate system. Now,
_
s

_
α
β
= x

α
y

β
= A
α
µ
x
µ
_
A
−1
_
ν
β
y
ν
= A
α
(x
µ
y
ν
)
_
A
−1
_
ν
β
= A
α
µ
s
µ
ν
_
A
−1
_
ν
β
,
so that s
α
β
transforms like an ordinary matrix, which means that s
α
β
is indeed an ordinary
matrix. But if we instead use two covectors,
t
µν
= x
µ
y
ν
=
_
_
x
1
y
1
x
1
y
2
x
1
y
3
x
2
y
1
x
2
y
2
x
2
y
3
x
3
y
1
x
3
y
2
x
3
y
3
_
_
,
then we get a tensor with different transformation properties,
t

αβ
= x

α
y

β
=
_
A
−1
_
µ
α
x
µ
_
A
−1
_
ν
β
y
ν
=
_
A
−1
_
µ
α
(x
µ
y
ν
)
_
A
−1
_
ν
β
=
_
A
−1
_
ν
α
t
µν
_
A
−1
_
ν
β
.
13
Introduction to Tensors Introducing Tensors
The difference here lies in the first matrix of the transformation equation. For s it is the
transformation matrix for contravariant vectors, while for t it is the transformation for covariant
vectors. The tensor t is clearly not a matrix, so we indeed created something new here. The
metric tensor g of the previous section is of the same type as t.
The beauty of tensors is that they can have an arbitrary number of indices. One can also
product, for instance, a tensor with three indices,
A
αβγ
= x
α
y
β
z
γ
.
In three dimensions, this gives an ordered array of 27 elements, a kind of “super matrix”.
Let us know introduce some terminology.
• The tensor A
αβγ
is a rank 3 tensor. Tensors of rank 0 are scalars, tensors of rank 1
are vectors and covectors, and tensors of rank 2 are matrices and other types of tensors
(such as the metric tensor).
In general, in n-dimensional space, a tensor of rank r has n
r
elements.
• We can distinguish between the contravariant rank and covariant rank of a tensor. A
αβγ
is a tensor of covariant rank 3 and contravariant rank 0. Its total rank is 3. One can also
produce tensors of, for instance, contravariant rank 2 and covariant rank 3, B
αβ
µνφ
, with
total rank 5.
Typically, when tensor mathematics is applied, the meaning of each index has been defined
beforehand: the first index means this, the secon means that, etc. As long as this is well-
defined, then one can have covariant and contravariant indices in any order.
Remark: Although a multiplication (without summation) of m vectors and m covectors produces a tensor
of rank m+n, not every tensor of rank m+n can be constructued as such a product. Tensors are much more
general than these simple products of vectors and covectors. It is therefore important to step away from this
picture of combining vectors and covects into a tensor, and consider this construction as nothing more than a
simple example.
Remark: We have said that tensors of rank 2 are matrices. It is not true, however, that all matrices are
tensors (of rank 2), as we have seen already with the object s
µ
ν
.
14
4 Tensor Definition and Properties
Let us now formally define a tensor.
Definition of Tensor
A (n, m) tensor t
µ
1
···µn
ν
1
···νm
at a given point in space can be described by an
array of numbers with n + m indices that transforms, upon coordinate
transformation by a given matrix A, in the following way:
t

α
1
···αn
β
1
···βm
= A
α
1
µ
1
· · · A
αn
µn
_
A
−1
_
ν
1
β
1
· · ·
_
A
−1
_
νm
βm
t
µ
1
···µn
ν
1
···νm
.
An (n, m) tensor in a k-dimensional manifold therefore has k
n+m
ele-
ments. It is contravariant in n components and covariant in m compo-
nents.
4.1 Symmetry and Anti-Symmetry
In practice is often happens that tensors display a certain amount of symmetry, like what we
know from matrices. Such symmetries have a strong effect on the properties of tensors. Often,
many of these properties or even tensors equations can be derived solely on the basis of these
symmetries.
A tensor t is called symmetric in the indices µ and ν if the elements are equal upon exchange
of the index values. So, for a second-rand contravariant tensor,
t
µν
= t
νµ
(symmetric (2,0) tensor). (4.1)
A tensor t is called anti-symmetric in the indices µ and ν if the elements are equal in absolute
value but opposite in sign upon exchange of the index values. So, for a second-rank contravariant
tensor,
t
µν
= −t
νµ
(anti-symmetric (2,0) tensor). (4.2)
It is not useful to speak of symmetry in a pair of indices that are not of the same type (either
covariant to contravariant), i.e., we can only consider symmetry of a tensor with respect to two
indices that are either both covariant or both contravariant. The reason for this is that the
properties of symmetry only remain invariant upon basis transformatoin if the indices are of the
same type.
15
Introduction to Tensors Tensor Definition and Properties
4.2 Contraction of Indices
With tensors of at least on covariant and at least
16

Sign up to vote on this title
UsefulNot useful