Preface

This book is prepared from personal notes and tutorials about tensor calculus at an
introductory level. The language and method used in presenting the ideas and techniques of
tensor calculus make it very suitable for learning this subject by the beginners who have
not been exposed previously to this elegant discipline of mathematics. Yes, some general
background in arithmetic, elementary algebra, calculus and linear algebra is needed to
understand the book and follow the development of ideas and techniques of tensors.
However, we made considerable efforts to reduce this dependency on foreign literature by
summarizing the main items needed in this regard to make the book self-contained. The
book also contains a number of graphic illustrations to aid the readers and students in their
effort to visualize the ideas and understand the abstract concepts. In addition to the graphic
illustrations, we used illustrative techniques such as highlighting key terms by boldface
fonts.
The book also contains extensive sets of clearly explained exercises which cover most of
the materials presented in the book where each set is given in an orderly manner in the end
of each chapter. These exercises are designed to provide thorough revisions of the supplied
materials and hence they make an essential component of the book and its learning
objectives. Therefore, they should not be considered as a decorative accessory to the book.
We also populated the text with hyperlinks, for the ebook users, to facilitate referencing
and connecting related objects so that the reader can go forth and back with minimum effort
and time and without compromising the continuity of reading by interrupting the chain of
thoughts.
In view of all the above factors, the present text can be used as a textbook or as a reference
for an introductory course on tensor algebra and calculus or as a guide for self-studying
and learning. I tried to be as clear as possible and to highlight the key issues of the subject
at an introductory level in a concise form. I hope I have achieved some success in reaching
these objectives for the majority of my target audience.
Finally, I should make a short statement about credits in making this book following the
tradition in writing book prefaces. In fact everything in the book is made by the author
including all the graphic illustrations, front and back covers, indexing, typesetting, and
overall design. However, I should acknowledge the use of the LaTeX typesetting package
and the LaTeX based document preparation package LyX for facilitating many things in
typesetting and design which cannot be done easily or at all without their versatile and
powerful capabilities. I also used the Ipe extensible drawing editor program for making all
the graphic illustrations in the book as well as the front and back covers.
Taha Sochi
London, November 2016
Table of Contents
Preface
Nomenclature
1: Preliminaries
1.1: Historical Overview of Development & Use of Tensor Calculus
1.2: General Conventions
1.3: General Mathematical Background
1.3.1: Coordinate Systems
1.3.2: Vector Algebra and Calculus
1.3.3: Matrix Algebra
1.4: Exercises
2: Tensors
2.1: General Background about Tensors
2.2: General Terms and Concepts
2.3: General Rules
2.4: Examples of Tensors of Different Ranks
2.5: Applications of Tensors
2.6: Types of Tensor
2.6.1: Covariant and Contravariant Tensors
2.6.2: True and Pseudo Tensors
2.6.3: Absolute and Relative Tensors
2.6.4: Isotropic and Anisotropic Tensors
2.6.5: Symmetric and Anti-symmetric Tensors
2.7: Exercises
3: Tensor Operations
3.1: Addition and Subtraction
3.2: Multiplication of Tensor by Scalar
3.3: Tensor Multiplication
3.4: Contraction
3.5: Inner Product
3.6: Permutation
3.7: Tensor Test: Quotient Rule
3.8: Exercises
4: delta and epsilon Tensors
4.1: Kronecker delta
4.2: Permutation epsilon
4.3: Useful Identities Involving delta or/and epsilon
4.3.1: Identities Involving delta
4.3.2: Identities Involving epsilon
 4.3.3: Identities Involving delta and epsilon
4.4: Generalized Kronecker delta
4.5: Exercises
5: Applications of Tensor Notation and Techniques
5.1: Common Definitions in Tensor Notation
5.2: Scalar Invariants of Tensors
5.3: Common Differential Operations in Tensor Notation
5.3.1: Cartesian Coordinate System
5.3.2: Cylindrical Coordinate System
5.3.3: Spherical Coordinate System
5.3.4: General Orthogonal Coordinate System
5.4: Common Identities in Vector and Tensor Notation
5.5: Integral Theorems in Tensor Notation
5.6: Examples of Using Tensor Techniques to Prove Identities
5.7: Exercises
6: Metric Tensor
6.1: Exercises
7: Covariant Differentiation
7.1: Exercises
References
Footnotes

Nomenclature
In the following table, we define some of the common symbols, notations and abbreviations
which are used in the book to avoid ambiguity and confusion.
∇ nabla differential operator
∇f gradient of scalar f
∇⋅A divergence of vector A
∇×A curl of vector A
2
∇ or Δ Laplacian operator
⊥ perpendicular to
2D, 3D, nD two-dimensional, three-dimensional, n-dimensional
det determinant of matrix
Ei ith covariant basis vector
Ei ith contravariant basis vector
Eq./Eqs. Equation/Equations
hi scale factor for ith coordinate in general orthogonal system
iff if and only if
r, θ, φ coordinates of spherical system in 3D space
tr trace of matrix
u1, u2, u3 coordinates of general orthogonal system in 3D space
x1, x2, x3 labels of coordinate axes of Cartesian system in 3D space
X1, X2, X3 same as the previous entry
x1 , x2 , x3 coordinates of general curvilinear system in 3D space
x, y, z coordinates of points in Cartesian system in 3D space
ρ, φ, z coordinates of cylindrical system in 3D space

Chapter 1

Preliminaries
In this introductory chapter, we provide the reader with a general overview about the
historical development of tensor calculus and its role in modern mathematics, science and
engineering. We also provide a general set of notes about the notations and conventions
which are generally followed in the writing of this book. A general mathematical
background about coordinate systems, vector algebra and calculus and matrix algebra is
also presented to make the book, to some extent, self-sufficient. Although the general
mathematical background section is not comprehensive, it contains essential mathematical
terminology and concepts which are needed in the development of the ideas and methods of
tensor calculus in the subsequent chapters of the book.

1.1 Historical Overview of Development & Use of

Tensor Calculus
First, we should remark that some of the following historical statements represent
approximate rather than exact historical facts due to the reality that many of the first hand
historical records are missing or not available to the author. Moreover, many ideas,
terminology, notation and techniques of tensor calculus, like any other field of knowledge
and practice, have been developed gradually over long periods of time although the credit
is usually attributed to a few individuals due to their prominence and fame or because they
played crucial and distinctive roles in the creation and development of the subject.
It is believed that the word “tensor” was coined by Hamilton but he used it in a rather
different meaning to what is being used for in modern mathematics and science. The credit
for attaching this term to its modern technical meaning, approximately in the late nineteenth
century, is usually given to Voigt. Apparently, the term “tensor” was originally derived
from the Latin word “tensus” which means tension or stress since one of the first uses of
tensors (in whatever meaning) was related to the mathematical description of mechanical
stress.
The names “tensor calculus” or “tensor analysis” have been used to label this subject in
its modern form rather recently, probably in the second quarter of the twenties century. The
early forms of this subject have been called “Ricci calculus” or “absolute differential
calculus”. The latter names may still be found in the modern literature of tensor calculus.
Many mathematicians and scientists have contributed to the development of tensor calculus
directly or indirectly. However, numerous components of the modern tensor calculus were
not developed as such and for the purpose of the theory of tensors but as parts or
byproducts of other disciplines, notably the differential geometry of curves and surfaces.
This generally applies prior to the official launch of tensor calculus as an independent
branch of mathematics by Ricci and Levi-Civita who are commonly recognized as the
founding fathers of this discipline.
Several concepts and techniques of tensor calculus have been developed by Gauss and
Riemann in the nineteenth century, mainly as part of their efforts to develop and formulate
the theory of differential geometry of curves and surfaces. Their contributions are
highlighted by the fact that many concepts, methods and equations, which are related
directly to tensor calculus or to subjects with a close affinity to tensor calculus, bear their
names, e.g. Gaussian coordinates, Gauss curvature tensor, Gauss-Codazzi equations,
Riemannian metric tensor, Riemannian manifold, and Riemann-Christoffel curvature tensor.
A major player in the development of tensor calculus is Christoffel whose contribution is
documented, for instance, in the Christoffel symbols of the first and second kind which
infuse throughout the whole subject of tensor calculus and play very essential roles. Also,
the above mentioned Riemann-Christoffel curvature tensor is another acknowledgment of
his achievements in this respect. Bianchi has also contributed a number of important ideas
and techniques, such as Bianchi identities, which played an important role in the
subsequent development of this subject.
As indicated above, the major role in the creation and development of tensor calculus in its
modern style is attributed to Ricci and Levi-Civita in the end of the nineteenth century and
the beginning of the twentieth century where these mathematicians extracted and merged the
previous collection of tensor calculus notations and techniques and introduced a number of
new ideas, notations and methods and hence they gave the subject its modern face and
contemporary style. For this reason, they are usually accredited for the creation of tensor
calculus as a standalone mathematical discipline although major components of this subject
have been invented by their predecessors as indicated above. The role of Ricci and Levi-
Civita in the creation and renovation of tensor calculus is documented in terms like Ricci
calculus, Ricci curvature tensor, Levi-Civita symbol, and Levi-Civita identity.
Many other mathematicians and scientists have made valuable indirect contributions to the
subject of tensor calculus by the creation of mathematical ideas which subsequently played
important roles in the development of tensor calculus although they are not related directly
to tensors. A prominent example of this category is Kronecker and his famous Kronecker
delta symbol which is embraced in tensor calculus and adapted for extensive use in tensor
notation and techniques.
The widespread use of tensor calculus in science has begun with the rise of the general
theory of relativity which is formulated in the language, methods and techniques of tensor
calculus. Tensor calculus has also found essential roles to play in a number of other
disciplines with a particular significance in differential geometry, continuum mechanics
and fluid dynamics.
Nowadays, tensor algebra and calculus propagate throughout many branches of
mathematics, physical sciences and engineering. The success of tensor calculus is credited
to its elegance, power and concision as well as the clarity and eloquence of its notation.
Indeed, tensor calculus is not just a collection of mathematical methods and techniques but
it is a compact and effective language for communicating ideas and expressing concepts of
varying degrees of complexity with rigor and reasonable simplicity.

1.2 General Conventions

In this book, we largely follow certain conventions and general notations; most of which
are commonly used in the mathematical literature although they may not be universally
approved. In this section, we outline the most important of these conventions and notations.
We also give some initial definitions of the most basic terms and concepts in tensor
calculus; more thorough technical definitions will follow, if needed, in the forthcoming
sections and chapters. It should be remarked that tensor calculus is riddled with conflicting
conventions and terminology. In this book we use what we believe to be the most common,
clear and useful of all of these.
Scalars are algebraic objects which are uniquely identified by their magnitude (absolute
value) and sign (±), while vectors are broadly geometric objects which are uniquely
identified by their magnitude (length) and direction in a presumed underlying space. At this
early stage in this book, we generically define tensor as an organized array of
mathematical objects such as numbers or functions.
In generic terms, the rank of a tensor signifies the complexity of its structure. Rank-0
tensors are called scalars while rank-1 tensors are called vectors. Rank-2 tensors may be
called dyads although dyad, in common use, may be designated to the outer product (see §
3.3↓) of two vectors and hence it is a special case of rank-2 tensors assuming it meets the
requirements of a tensor and hence transforms as a tensor. Like rank-2 tensors, rank-3
tensors may be called triads. Similar labels, which are much less common in use, may be
attached to higher rank tensors; however, none of these will be used in the present book.
More generic names for higher rank tensors, such as polyad, are also in use.
We may use the term “tensor” to mean tensors of all ranks including scalars (rank-0) and
vectors (rank-1). We may also use this term as opposite to scalar and vector, i.e. tensor of
rank-n where n > 1. In almost all cases, the meaning should be obvious from the context. It
should be remarked that in the present book all tensors of all ranks and types are assumed
to be real quantities, i.e. they have real rather than complex components.
Due to its introductory nature, the present book is largely based on assuming an underlying
orthonormal Cartesian coordinate system. However, certain parts of the book are based
on other types of coordinate system; in these cases this is stated explicitly or made clear by
the notation and context. As will be outlined later (see Orthonormal Cartesian
Coordinate System in § 1.3.1↓), we mean by “orthonormal” a system with mutually
perpendicular and uniformly scaled axes with a unit basis vector set. We may also use
“rectangular” for a similar meaning or to exclude oblique coordinate systems (see
Footnote 1 in § 8↓) which may also be labeled by some as Cartesian. Some of the
statements in this book in which these terms are used may not strictly require these
conditions but we add these terms to focus the attention on the type of the coordinate system
which these statements are made about.
To label scalars, we use non-indexed lower case light face italic Latin letters (e.g. f and
h), while for vectors, we use non-indexed lower or upper case bold face non-italic Latin
letters (e.g. a and A). The exception to this is the basis vectors where indexed bold face
lower or upper case non-italic symbols (e.g. e 1 and Ei) are used. However, there should
be no confusion or ambiguity about the meaning of any one of these symbols. As for
tensors of rank > 1, non-indexed upper case bold face non-italic Latin letters (e.g. A and
B) are used (see Footnote 2 in § 8↓).
Indexed light face italic Latin symbols (e.g. ai and Bjki) are used in this book to denote
tensors of rank > 0 in their explicit tensor form, i.e. index notation. Such symbols may
also be used to denote the components of these tensors. The meaning is usually transparent
and can be identified from the context if it is not declared explicitly. Tensor indices in this
book are lower case Latin letters, usually taken from the middle of the Latin alphabet like
(i, j, k). We also use numbered indices like (i1, i2, …, ik) when the number of tensor
indices is variable. Numbers are also used as indices (e.g. ε12) in some occasions for
obvious purposes such as making statements about particular components.
Mathematical identities and definitions are generally denoted in the mathematical
literature by using the symbol “ ≡ ”. However, for simplicity we use the equality sign “=”
to mark identities and mathematical definitions as well as normal equalities.
We use vertical bars (i.e. |⋅|) to symbolize determinants and square brackets (i.e. [⋅]) to
symbolize matrices. This applies where these symbols contain arrays of objects; otherwise
they have their normal meaning according to the context, e.g. bars embracing vectors mean
modulus of the vector. We normally use indexed square brackets (e.g. [A]i and [∇f]i) to
denote the ith component of vectors in their symbolic or vector notation. For tensors of
higher rank, more than one index is needed to denote their components, e.g. [A]ij for the ijth
component of a rank-2 tensor.
Partial derivative symbol with a subscript index (e.g. ∂i) is frequently used to denote the
ith component of the gradient operator nabla ∇ (see Footnote 3 in § 8↓):
(1)

A comma preceding a subscript index (e.g. , i) is also used to denote partial

differentiation with respect to the ith spatial coordinate, mainly in Cartesian systems, e.g.
(2)

Partial derivative symbol with a spatial subscript, rather than an index, is used to denote
partial differentiation with respect to that spatial variable. For instance:
(3)

is used for the partial derivative with respect to the radial coordinate r in spherical
coordinate systems identified by the spatial variables (r, θ, φ) (see Footnote 4 in § 8↓).
Partial derivative symbol with repeated double index is used to denote the Laplacian
operator:
2
(4) ∂ii = ∂i∂i = ∇ = Δ
The notation is not affected by using repeated double index other than i (e.g. ∂jj or ∂kk).
The following notations: ∂2ii, ∂2 and ∂i∂i are also used in the literature of tensor calculus
to symbolize the Laplacian operator. However, these notations will not be used in the
present book.
We follow the common convention of using a subscript semicolon preceding a subscript
index (e.g. Akl;i) to symbolize the operation of covariant differentiation with respect to
the ith coordinate (see § 7↓). The semicolon notation may also be attached to the normal
differential operators for the same purpose, e.g. ∇;i or ∂;i to indicate covariant
differentiation with respect to the variable indexed by i.
Finally, all transformation equations in the present book are assumed to be continuous and
real, and all derivatives are continuous in their domain of variables. Based on the
continuity condition of the differentiable quantities, the individual differential operators in
the second (and higher) order partial derivatives with respect to different indices are
commutative, that is:
(5) ∂i∂j = ∂j∂i

1.3 General Mathematical Background

In this section, we provide a general mathematical background which is largely required to
understand the forthcoming materials about tensors and their algebra and calculus. This
mathematical background is intended to provide the essential amount needed for the book
to be self-contained, to some degree, but it is not comprehensive. Also some concepts and
techniques discussed in this section require more elementary materials which should be
obtained from textbooks at more basic levels.

1.3.1 Coordinate Systems

In generic terms, a coordinate system is a mathematical device used to identify the location
of points in a given space. In tensor calculus, a coordinate system is needed to define non-
scalar tensors in a specific form and identify their components in reference to the basis set
of the system. Hence, non-scalar tensors require a predefined coordinate system to be fully
identified (see Footnote 5 in § 8↓).
There are many types of coordinate system, the most common ones are: the orthonormal
Cartesian, the cylindrical and the spherical. A 2D version of the cylindrical system is the
plane polar system. The most general type of coordinate system is the general curvilinear.
A subset of the latter is the orthogonal curvilinear. These types of coordinate system are
briefly investigated in the following subsections.
A. Orthonormal Cartesian Coordinate System
This is the simplest and the most commonly used coordinate system. It consists, in its
simplest form, of three mutually orthogonal straight axes that meet at a common point
called the origin of coordinates O. The three axes, assuming a 3D space, are scaled
uniformly and hence they all have the same unit length. Each axis has a unit vector oriented
along the positive direction of that axis (see Footnote 6 in § 8↓). These three unit vectors
are called the basis vectors or the bases of the system. These basis vectors are constant in
magnitude and direction throughout the system (see Footnote 7 in § 8↓). This system with
its basis vectors (e 1, e 2 and e 3) is depicted in Figure 1↓.
The three axes, as well as the basis vectors, are usually labeled according to the right
hand rule, that is if the index finger of the right hand is pointing in the positive direction of
the first axis and its middle finger is pointing in the positive direction of the second axis
then the thumb will be pointing in the positive direction of the third axis.

Figure 1 Orthonormal right-handed Cartesian coordinate system and its basis vectors
e 1, e 2 and e 3 in a 3D space (left frame) with the components of a vector v in this system
(right frame).

B. Cylindrical Coordinate System

The cylindrical coordinate system is defined by three parameters: ρ, φ and z which range
over: 0 ≤ ρ < ∞, 0 ≤ φ < 2π and − ∞ < z < ∞. These parameters identify the coordinates of
a point P in a 3D space where ρ represents the perpendicular distance from the point to
the x3-axis of a corresponding rectangular Cartesian system, φ represents the angle
between the x1-axis and the line connecting the origin of coordinates O to the
perpendicular projection of the point on the x1-x2 plane of the corresponding Cartesian
system, and z is the same as the third coordinate of the point in the reference Cartesian
system. The sense of the angle φ is given by the right hand twist rule, that is if the fingers of
the right hand curl in the sense of rotation from the x1-axis towards the line of projection,
then the thumb will be pointing in the positive direction of the x3-axis.
The basis vectors of the cylindrical system (e ρ, e φ and e z) are pointing in the direction of
increasing ρ, φ and z respectively. Hence, while e z is constant in magnitude and direction
throughout the system, e ρ and e φ are coordinate-dependent as they vary in direction from
point to point. All these basis vectors are mutually perpendicular and they are defined to
be of unit length. Figure 2↓ is a graphical illustration of the cylindrical coordinate system
and its basis vectors with a corresponding reference Cartesian system in a standard
position.
Figure 2 Cylindrical coordinate system, superimposed on a rectangular Cartesian system
in a standard position, and its basis vectors e ρ, e φ and e z in a 3D space. The point P in the
figure is identified simultaneously by (x, y, z) coordinates in the Cartesian system and by
(ρ, φ, z) coordinates in the cylindrical system where these coordinates are related through
the two sets of Eqs. 6↓ and 7↓.

The transformation from the Cartesian coordinates (x, y, z) of a particular point in the
space to the cylindrical coordinates (ρ, φ, z) of that point, where the two systems are in a
standard position, is performed through the following equations (see Footnote 8 in § 8↓):
(6)
ρ = √(x2 + y2)
φ = arctan(y ⁄ x)
z=z
while the opposite transformation from the cylindrical to the Cartesian coordinates is
performed by the following equations:
(7)
x = ρ cosφ
y = ρ sinφ
z=z

C. Plane Polar Coordinate System

The plane polar coordinate system is the same as the cylindrical coordinate system with the
absence of the z coordinate, and hence all the equations of the polar system are obtained by
setting z = 0 in the equations of the cylindrical coordinate system, that is:
(8)
ρ = √(x2 + y2)
φ = arctan(y ⁄ x)
and
(9)
x = ρ cosφ
y = ρ sinφ
This system is illustrated graphically in Figure 3↓ where the point P is located by (ρ, φ) in
the polar system and by (x, y) in the corresponding 2D Cartesian system.
Figure 3 Plane polar coordinate system, superimposed on a 2D rectangular Cartesian
system in a standard position, and their basis vectors e ρ and e φ for the polar system and e 1
and e 2 for the Cartesian system. The point P in the figure is identified simultaneously by
(x, y) coordinates in the Cartesian system and by (ρ, φ) coordinates in the polar system
where these coordinates are related through the two sets of Eqs. 8↑ and 9↑.

D. Spherical Coordinate System

The spherical coordinate system is defined by three parameters: r, θ and φ which range
over: 0 ≤ r < ∞, 0 ≤ θ ≤ π and 0 ≤ φ < 2π. These parameters identify the coordinates of a
point P in a 3D space where r represents the distance from the origin of coordinates O to
P, θ is the angle from the positive x3-axis of the corresponding Cartesian system to the line
connecting the origin of coordinates O to P, and φ is the same as in the cylindrical
coordinate system.
The basis vectors of the spherical system (e r, e θ and e φ) are pointing in the direction of
increasing r, θ and φ respectively. Hence, all these basis vectors are coordinate-
dependent as they vary in direction from point to point. All these basis vectors are
mutually perpendicular and they are defined to be of unit length. Figure 4↓ is a graphical
illustration of the spherical coordinate system and its basis vectors with a corresponding
reference rectangular Cartesian system in a standard position.
Figure 4 Spherical coordinate system, superimposed on a rectangular Cartesian system in
a standard position, and its basis vectors e r, e θ and e φ in a 3D space. The point P in the
figure is identified simultaneously by (x, y, z) coordinates in the Cartesian system and by
(r, θ, φ) coordinates in the spherical system where these coordinates are related through
the two sets of Eqs. 10↓ and 11↓.

The transformation from the Cartesian coordinates (x, y, z) of a particular point in the
space to the spherical coordinates (r, θ, φ) of that point, where the two systems are in a
standard position, is performed by the following equations (see Footnote 9 in § 8↓):
(10)
r = √(x2 + y2 + z2)
θ = arccos(z ⁄ √(x2 + y2 + z2))
φ = arctan(y ⁄ x)
while the opposite transformation from the spherical to the Cartesian coordinates is
performed by the following equations:
(11)
x = r sinθ cosφ
y = r sinθ sinφ
z = r cosθ

E. General Curvilinear Coordinate System

The general curvilinear coordinate system is characterized by having coordinate axes
which are curved in general. Also, its basis vectors are generally position-dependent and
hence they are variable in magnitude and direction throughout the system. Consequently, the
basis vectors are not necessarily of unit length or mutually orthogonal. A graphic
demonstration of the general curvilinear coordinate system at a particular point in the space
with its covariant basis vectors (see § 2.6.1↓) is shown in Figure 5↓.
Figure 5 General curvilinear coordinate system and its covariant basis vectors E1, E2
and E3 (see § 2.6.1↓) in a 3D space, where x1, x2 and x3 are the labels of the coordinates.

F. General Orthogonal Curvilinear Coordinate System

The general orthogonal curvilinear coordinate system is a subset of the general curvilinear
coordinate system as described above. It is distinguished from the other subsets of the
general curvilinear system by having coordinate axes and basis vectors which are
mutually orthogonal throughout the system. The cylindrical and spherical systems are
examples of orthogonal curvilinear coordinate systems.

1.3.2 Vector Algebra and Calculus

This subsection provides a short introduction to vector algebra and calculus, a subject that
is closely related to tensor calculus. In fact many ideas and methods of tensor calculus find
their precursors and roots in vector algebra and calculus and hence many concepts,
techniques and notations of the former can be viewed as extensions and generalizations of
their counterparts in the latter.
A. Dot Product of Vectors
The dot product, or scalar product, of two vectors is a scalar quantity which has two
interpretations: geometric and algebraic. Geometrically, the dot product of two vectors a
and b can be interpreted as the projection of a onto b times the length of b, or as the
projection of b onto a times the length of a, as demonstrated in Figure 6↓. In both cases, the
dot product is obtained by taking the product of the length of the two vectors involved
times the cosine of the angle between them when their tails are made to coincide, that is:
(12) a⋅b = |a| |b| cosθ
where the dot between a and b on the left hand side of the equation stands for the dot
product operation, 0 ≤ θ ≤ π is the angle between the two vectors and the bars notation
means the modulus or the length of the vector.

Figure 6 Demonstration of the geometric interpretation of the dot product of two vectors
a and b (left frame) as the projection of a onto b times the length of b (middle frame) or as
the projection of b onto a times the length of a (right frame).

Algebraically, the dot product is the sum of the products of the corresponding components
of the two vectors, that is:
(13) a⋅b = a1b1 + a2b2 + a3b3
where ai and bj (i, j = 1, 2, 3) are the components of a and b respectively. Here, we are
assuming an orthonormal Cartesian system in a 3D space; the formula can be easily
extended to an nD space, that is:
(14) a⋅b = Σni = 1 aibi
From Eq. 12↑, it is obvious that the dot product is positive when 0 ≤ θ < π ⁄ 2, zero when
θ = π ⁄ 2 (i.e. the two vectors are orthogonal), and negative when π ⁄ 2 < θ ≤ π. The
magnitude of the dot product is equal to the product of the lengths of the two vectors when
they have the same orientation (i.e. parallel or anti-parallel). Based on the above given
facts, the dot product is commutative, that is:
(15) a⋅b = b⋅a

B. Cross Product of Vectors

Geometrically, the cross product, or vector product, of two vectors, a and b, is a vector
whose length is equal to the area of the parallelogram defined by the two vectors as its
two main sides when their tails coincide and whose orientation is perpendicular to the
plane of the parallelogram with a direction defined by the right hand rule as demonstrated
in Figure 7↓. Hence the cross product of two vectors a and b is given by:
(16) a × b = (|a| |b| sinθ) n
where 0 ≤ θ ≤ π is the angle between the two vectors when their tails coincide and n is a
unit vector perpendicular to the plane containing a and b and is directed according to the
right hand rule.
From Eq. 16↑, it can be seen that the cross product vector is zero when θ = 0 or θ = π, i.e.
when the two vectors are parallel or anti-parallel respectively. Also, the length of the cross
product vector is equal to the product of the lengths of the two vectors when the two
vectors are orthogonal (i.e. θ = π ⁄ 2).

Figure 7 Graphical demonstration of the cross product of two vectors a and b (left frame)
with the right hand rule (right frame).
Algebraically, the cross product of two vectors a and b is expressed by the following
determinant (see Determinant of Matrix in § 1.3.3↓) where the determinant is expanded
along its first row, that is:
(17)

Here, we are assuming an orthonormal Cartesian coordinate system in a 3D space with a

basis vector set e 1, e 2 and e 3.
Based on the above given facts, since the direction is determined by the right hand rule the
cross product is anti-commutative, that is:
(18) a × b = − b × a

C. Scalar Triple Product of Vectors

The scalar triple product of three vectors (a, b and c) is a scalar quantity defined by the
expression:
(19) a⋅(b × c)
where the dot and multiplication symbols stand respectively for the dot and cross product
operations of two vectors as defined above. Hence, the scalar triple product is defined
geometrically by:
(20) a⋅(b × c) = |a| |b| |c| sinφ cosθ
where φ is the angle between b and c while θ is the angle between a and b × c. The scalar
triple product is illustrated graphically in Figure 8↓. As there is no meaning of a cross
product operation between a scalar and a vector, the parentheses in the above equation
are redundant although they provide a clearer and more clean notation.
Figure 8 Graphic illustration of the scalar triple product of three vectors a, b and c. The
magnitude of this product is equal to the volume of the seen parallelepiped.

Now, since |b × c|( = |b| |c| sinφ) is equal to the area of the parallelogram whose two main
sides are b and c, while |a|cosθ represents the projection of a onto the orientation of b × c
and hence it is equal to the height of the parallelepiped (refer to Figure 8↑), the magnitude
of the scalar triple product is equal to the volume of the parallelepiped whose three main
sides are a, b and c while its sign is positive or negative depending, respectively, on
whether the vectors a, b and c form a right-handed or left-handed system.
The scalar triple product is invariant to a cyclic permutation of the symbols of the three
vectors involved, that is:
(21) a⋅(b × c) = c⋅(a × b) = b⋅(c × a)
It is also invariant to an exchange of the dot and cross product symbols, that is:
(22) a⋅(b × c) = (a × b)⋅c
Hence, from the three possibilities of the first invariance with the two possibilities of the
second invariance, we have six equal expressions for the scalar triple product of three
vectors (see Footnote 10 in § 8↓). The other six possibilities of the scalar triple product
of three vectors, which are obtained from the first six possibilities with the opposite cyclic
permutations, are also equal to each other for the same reason. However, they are equal in
magnitude to the first six possibilities but are different in sign (see Footnote 11 in § 8↓).
From the above interpretation of the scalar triple product as the signed volume of the
parallelepiped formed by the three vectors, it is obvious that this product is zero when the
three vectors are coplanar. This, of course, includes the possibility of being collinear.
The scalar triple product of three vectors is also defined algebraically as the determinant
(refer to Determinant of Matrix in § 1.3.3↓) of the matrix formed by the components of the
three vectors as its rows or columns in the given order, that is:
(23)

D. Vector Triple Product of Vectors

The vector triple product of three vectors (a, b and c) is a vector quantity defined by the
following expressions:
(24) a × (b × c) or (a × b) × c
where the multiplication symbols stand for the cross product operation of two vectors as
defined previously. Vector triple product is not associative, so in general we have (see
Footnote 12 in § 8↓):
(25) a × (b × c) ≠ (a × b) × c

E. Differential Operations of nabla Operator (see Footnote 13 in § 8↓)

There are many differential operations that can be performed on scalar and vector fields
(see Footnote 14 in § 8↓). Here, we define the most important and widely used of these
operations that involve the nabla ∇ differential operator, that is the gradient, divergence
and curl. We also define the Laplacian differential operator which is based on a
combination of the gradient and divergence operations. The definitions here are given in
Cartesian coordinates only. More definitions of these operations and operators will be
given in § 5.3↓ in terms of tensor notation for the Cartesian, as well as the equivalent
definitions for some non-Cartesian coordinate systems.
The differential vector operator nabla ∇ is defined in rectangular Cartesian coordinate
systems by the following expression:
(26)

where i, j and k are the unit vectors in the x, y and z directions respectively.
The gradient of a scalar field f(x, y, z) is a vector defined, in Cartesian coordinate
systems, by:
(27)

Geometrically, the gradient of a scalar field f(x, y, z), at any point in the space where the
field is defined, is a vector normal to the surface f(x, y, z) = constant (refer to Figure 9↓)
pointing in the direction of the fastest increase in the field at that point.
Figure 9 The gradient of a scalar field f(x, y, z) as a vector normal to the surface
f(x, y, z) = constant pointing in the direction of the fastest increase in the field at that point.

The gradient operation is distributive but not commutative or associative, that is:
(28) ∇(f + h) = ∇f + ∇h
(29) ∇f ≠ f∇
(30) (∇f)h ≠ ∇(fh)
where f and h are differentiable scalar functions of position.
The divergence of a vector field v(x, y, z) is a scalar quantity defined as the dot product
of the nabla operator with the vector. Hence, in Cartesian coordinate systems it is given by:
(31)
where vx, vy and vz are the components of v in the x, y and z directions respectively. In
broad terms, the physical significance of the divergence of a vector field is that it is a
measure of how much the field diverges or converges at a particular point in the space
where the field is defined. When the divergence of a vector field is identically zero, the
field is called solenoidal.
The divergence operation is distributive but not commutative or associative, that is:
(32) ∇⋅(A + B) = ∇⋅A + ∇⋅B
(33) ∇⋅A ≠ A⋅∇
(34) ∇⋅(fA) ≠ ∇f⋅A
where A and B are differentiable vector functions of position.
The curl of a vector field v(x, y, z) is a vector defined as the cross product of the nabla
operator with the vector. Hence, in Cartesian coordinate systems it is given by (refer to
Cross Product of Vectors in § 1.3.2↑):
(35)

Broadly speaking, the curl of a vector field is a quantitative measure of the circulation or
rotation of the field at a given point in the space where the field is defined. When the curl
of a vector field vanishes identically, the field is called irrotational.
The curl operation is distributive but not commutative or associative, that is:
(36) ∇ × (A + B) = ∇ × A + ∇ × B
(37) ∇ × A ≠ A × ∇
(38) ∇ × (A × B) ≠ (∇ × A) × B
where A and B are differentiable vector functions of position.
2
The Laplacian (see Footnote 15 in § 8↓) scalar operator ∇ is defined as the divergence
of the gradient operator and hence it is given, in Cartesian coordinates, by:
(39)
The Laplacian can act on scalar, vector and tensor fields of higher rank. When the
Laplacian operates on a tensor (in its general sense which includes scalar and vector) it
produces a tensor of the same rank; hence the Laplacian of a scalar is a scalar, the
Laplacian of a vector is a vector, the Laplacian of a rank-2 tensor is a rank-2 tensor, and so
on.

F. Divergence Theorem
The divergence theorem, which is also known as Gauss theorem, is a mathematical
statement of the intuitive idea that the integral of the divergence of a vector field over a
given volume is equal to the total flux of the vector field out of the surface enclosing the
volume. Symbolically, the divergence theorem states that:
(40) ∭V∇⋅A dτ = ∬SA⋅n dσ
where A is a differentiable vector field, V is a bounded volume in an nD space enclosed
by a surface S, dτ and dσ are volume and surface elements respectively, and n is a variable
unit vector normal to the surface.
The divergence theorem is useful for converting volume integrals into surface integrals
and vice versa. In many cases, this can result in a considerable simplification of the
required mathematical work when one of these integrals is easier to manipulate and
evaluate than the other, or even overcoming a mathematical hurdle when one of the
integrals cannot be evaluated analytically. Moreover, the divergence theorem plays a
crucial role in many mathematical proofs and theoretical arguments in mathematical and
physical theories.
G. Stokes Theorem
Stokes theorem is a mathematical statement that the integral of the curl of a vector field
over an open surface is equal to the line integral of the field around the perimeter
surrounding the surface, that is:
(41) ∬S(∇ × A)⋅n dσ = ∮CA⋅dr
where A is a differentiable vector field, C symbolizes the perimeter of the surface S, dr is
a vector element tangent to the perimeter, and the other symbols are as defined in the
divergence theorem. The perimeter should be traversed in a sense related to the direction
of the normal vector n by the right hand twist rule, that is when the fingers of the right
hand twist in the sense of traversing the perimeter the thumb will point approximately in the
direction of n, as seen in Figure 10↓.

Figure 10 Illustration of Stokes integral theorem (left frame) with the right hand twist rule
(right frame).

Similar to the divergence theorem, Stokes theorem is useful for converting surface
integrals into line integrals and vice versa, which is useful in many cases for reducing the
amount of mathematical work or overcoming technical and mathematical difficulties.
Stokes theorem is also crucial in the development of many proofs and theoretical arguments
in mathematics and science.

1.3.3 Matrix Algebra

There is a close relation between rank-2 tensors and square matrices where the latter
usually represent the former. Hence, there are many ideas, techniques and notations which
are common or similar between the two subjects. We therefore provide in this subsection a
set of short introductory notes about matrix algebra to supply the reader with the essential
terminology and methods of matrix algebra which are needed in the development of the
forthcoming chapters about tensor calculus.
A. Definition of Matrix
A matrix is a rectangular array of mathematical objects (mainly numbers or functions)
which is subject to certain rules in its manipulation and over which certain mathematical
operations are defined. Hence, two indices are needed to define a matrix unambiguously
where the first index labels the rows while the second index labels the columns. A matrix
which consists of m rows and n columns is said to be an m × n matrix.
The elements or entries of a matrix A are usually labeled with light-face symbols similar to
the symbol used to label the matrix where each element is suffixed with two indices: the
first refers to the row number of the entry and the second refers to its column number.
Hence, for a matrix A the entry in its second row and fifth column is labeled A25.
The two indices of a matrix are not required to have the same range since a matrix can have
different number of rows and columns. When the two indices have the same range, the
matrix is described as square matrix. Examples of matrices are:
(42)

When the range of the row/column index has only one value (i.e. 1) while the range of the
other index has multiple values the matrix is described as row/column matrix. Vectors may
be represented by row or column matrices (see Footnote 16 in § 8↓). Scalars may be
regarded as a trivial case of matrices.
For a square matrix, the entries with equal values of row and column indices are called the
main diagonal of the matrix. For example, the entries of the main diagonal of the third
matrix in Eq. 42↑ are C11 and C22. The elements of the other diagonal running from the top
right corner to the bottom left corner form the trailing or anti-diagonal, i.e. C12 and C21 in
the previous example.
B. Special Matrices
The zero matrix is a matrix whose all entries are 0. The identity or unit or unity matrix
is a square matrix whose all entries are 0 except those on its main diagonal which are 1. A
matrix is described as singular iff its determinant is zero (see Determinant of Matrix in §
1.3.3↓). A singular matrix has no inverse (see Inverse of Matrix in § 1.3.3↓). A square
matrix is called diagonal if all of its elements which are not on the main diagonal are zero.
The transpose of a matrix is a matrix obtained by exchanging the rows and columns of the
original matrix. For example, if A is a 3 × 3 matrix and AT is its transpose then (see
Footnote 17 in § 8↓):
(43)
The transposition operation is defined even for non-square matrices. For square matrices,
transposition represents a reflection of the matrix elements in the main diagonal of the
matrix.
C. Matrix Multiplication
The multiplication of two matrices, A of m × k dimensions and B of k × n dimensions, is
defined as an operation that produces a matrix C of m × n dimensions whose Cij entry is
the dot product of the ith row of the first matrix A and the jth column of the second matrix
B. Hence, if A is a 3 × 2 matrix and B is a 2 × 2 matrix, then their product AB is a 3 × 2
matrix which is given by:
(44)

From the above, it can be seen that matrix multiplication is defined only when the number
of columns of the first matrix is equal to the number of rows of the second matrix. Matrix
multiplication is associative and distributive over a sum of compatible matrices, but it is
not commutative in general even if both forms of the product are defined, that is:
(45) (AB)C = A(BC)
(46) A(B + C) = AB + AC
(47) AB ≠ BA
As seen above, no symbol is used to indicate the operation of matrix multiplication
according to the notation of matrix algebra, i.e. the two matrices are put side by side with
no symbol in between. However, in tensor symbolic notation such an operation is usually
represented by a dot between the symbols of the two matrices, as will be discussed later in
the book (see Footnote 18 in § 8↓).
D. Trace of Matrix
The trace of a matrix is the sum of its diagonal elements, therefore if a matrix A is given
by:
(48)
then its trace is given by:
(49) tr(A) = A11 + A22 + A33
From its definition, it is obvious that the trace of a matrix is a scalar and it is defined only
for square matrices.
E. Determinant of Matrix
The determinant is a scalar quantity associated with a square matrix. There are several
definitions for the determinant of a matrix; the most direct one is that the determinant of a 2
× 2 matrix is the product of the elements of its main diagonal minus the product of the
elements of its trailing diagonal, that is:
(50)

The determinant of an n × n (n > 2) matrix is then defined, recursively, as the sum of the
products of each entry of any one of its rows or columns times the cofactor of that entry
where the cofactor of an entry is defined as the determinant obtained from eliminating the
row and column of that entry from the parent matrix with a sign given by ( − 1)i + j with i
and j being the indices of the row and column of that entry (see Footnote 19 in § 8↓). For
example:
(51)
where the determinant is evaluated along the first row. It should be remarked that the
determinant of a matrix and the determinant of its transpose are equal, that is:
(52) det(A) = det(AT)
where A is a square matrix and T stands for the transposition operation. Another remark is
that the determinant of a diagonal matrix is the product of its main diagonal elements.
F. Inverse of Matrix
The inverse of a square matrix A is a square matrix A − 1 where:
(53) AA − 1 = A − 1 A = I
with I being the identity matrix (see Special Matrices in § 1.3.3↑) of the same dimensions
as A. The inverse of a square matrix is formed by transposing the matrix of cofactors of the
original matrix with dividing each element of the transposed matrix of cofactors by the
determinant of the original matrix (see Footnote 20 in § 8↓). From this definition, it is
obvious that a matrix possesses an inverse only if its determinant is not zero, i.e. it must be
non-singular.
It should be remarked that this definition includes the 2 × 2 matrices where the cofactor of
an entry is a single entry with the designated sign, that is:
(54)

Another remark is that the inverse of an invertible diagonal matrix is a diagonal matrix
obtained by taking the reciprocal of the corresponding diagonal elements of the original
matrix.

1.4 Exercises
Exercise 1. Name three mathematicians accredited for the development of tensor calculus.
For each one of these mathematicians, give a mathematical technical term that bears his
name.
Exercise 2. What are the main scientific disciplines that employ the language and
techniques of tensor calculus?
Exercise 3. Mention one cause for the widespread use of tensor calculus in science.
Exercise 4. Describe some of the distinctive features of tensor calculus which contributed
to its success and extensive use in mathematics, science and engineering.
Exercise 5. Give preliminary definitions of the following terms: scalar, vector, tensor, rank
of tensor, and dyad.
Exercise 6. What is the meaning of the following mathematical symbols?
∂i
∂ii
∇
A, i
Δ
Ai;k
Exercise 7. Is the following equality correct? If so, is there any condition for this to hold?
∂k ∂l = ∂l ∂k
Exercise 8. Describe, briefly, the following six coordinate systems outlining their main
features: orthonormal Cartesian, cylindrical, plane polar, spherical, general curvilinear,
and general orthogonal.
Exercise 9. Which of the six coordinate systems in the previous exercise are orthogonal?
Exercise 10. What “basis vectors” of a coordinate system means and what purpose they
serve?
Exercise 11. Which of the six coordinate systems mentioned in the previous exercises have
constant basis vectors (i.e. some or all of their basis vectors are constant both in magnitude
and in direction)?
Exercise 12. Which of the above six coordinate systems have unit basis vectors by
definition or convention?
Exercise 13. Explain the meaning of the coordinates in the cylindrical and spherical
systems (i.e. ρ, φ and z for the cylindrical, and r, θ and φ for the spherical).
Exercise 14. What is the relation between the cylindrical and plane polar coordinate
systems?
Exercise 15. Is there any common coordinates between the above six coordinate systems?
If so, what? Investigate this thoroughly by comparing each pair of these systems.
Exercise 16. Write the transformation equations between the following coordinate systems
in both directions: Cartesian and cylindrical, and Cartesian and spherical.
Exercise 17. Make a sketch representing a spherical coordinate system, with its basis
vectors, superimposed on a rectangular Cartesian system in a standard position.
Exercise 18. What are the geometric and algebraic definitions of the dot product of two
vectors? What is the interpretation of the geometric definition?
Exercise 19. What are the geometric and algebraic definitions of the cross product of two
vectors? What is the interpretation of the geometric definition?
Exercise 20. What is the dot product of the vectors A and B if A = (1.9, − 6.3, 0) and B =
( − 4, − 0.34, 11.9)?
Exercise 21. What is the cross product of the vectors in the previous exercise? Write this
cross product in its determinantal form and expand it.
Exercise 22. Define the scalar triple product operation of three vectors geometrically and
algebraically.
Exercise 23. What is the geometric interpretation of the scalar triple product? What is the
condition for this product to be zero?
Exercise 24. Is it necessary to use parentheses in the writing of scalar triple products and
why? Is it possible to interchange the dot and cross symbols in the product?
Exercise 25. Calculate the following scalar triple products:
a⋅(b × c)
a⋅(d × c)
d⋅(c × b)
a⋅(c × b)
(a × b)⋅c
where a = (7, − 0.4, 9.5), b = ( − 12.9, − 11.7, 3.1), c = (2.4, 22.7, − 6.9) and d = ( −
56.4, 29.5, 33.8). Note that some of these products may be found directly from other
products with no need for detailed calculations.
Exercise 26. Write the twelve possibilities of the scalar triple product a⋅(b × c) and
divide them into two sets where the entries in each set are equal. What is the relation
between the two sets?
Exercise 27. What is the vector triple product of three vectors in mathematical terms? Is it
scalar or vector? Is it associative?
Exercise 28. Give the mathematical expression for the nabla ∇ differential operator in
Cartesian systems.
Exercise 29. State the mathematical definition of the gradient of a scalar field f in
Cartesian coordinates. Is it scalar or vector?
Exercise 30. Is the gradient operation commutative, associative or distributive? Express
these properties mathematically.
Exercise 31. What is the relation between the gradient of a scalar field f and the surfaces
of constant f? Make a simple sketch to illustrate this relation.
Exercise 32. Define, mathematically, the divergence of a vector field V in Cartesian
coordinates. Is it scalar or vector?
Exercise 33. Define “solenoidal” vector field descriptively and mathematically.
Exercise 34. What is the physical significance of the divergence of a vector field?
Exercise 35. Is the divergence operation commutative, associative or distributive? Give
your answer in words and in mathematical forms.
Exercise 36. Define the curl of a vector field V in Cartesian coordinates using the
determinantal and the expanded forms with full explanation of all the symbols involved. Is
the curl scalar or vector?
Exercise 37. What is the physical significance of the curl of a vector field?
Exercise 38. What is the technical term used to describe a vector field whose curl
vanishes identically?
Exercise 39. Is the curl operation commutative, associative or distributive? Express these
properties symbolically.
2
Exercise 40. Describe, in words, the Laplacian operator ∇ and how it is obtained. What
are the other symbols used to denote it?
Exercise 41. Give the mathematical expression of the Laplacian operator in Cartesian
systems. Using this mathematical expression, explain why the Laplacian is a scalar rather
than a vector operator?
Exercise 42. Can the Laplacian operator act on rank-0, rank-1 and rank-n (n > 1) tensor
fields? If so, what is the rank of the resulting field in each case?
Exercise 43. Write down the mathematical expression for the divergence theorem, defining
all the symbols involved, and explain the meaning of this theorem in words.
Exercise 44. What are the main uses of the divergence theorem in mathematics and
science? Explain why this theorem is very useful theoretically and practically.
Exercise 45. If a vector field is given in the Cartesian coordinates by A = ( − 0.5, 9.3,
6.5), verify the divergence theorem for a cube defined by the plane surfaces x1 = − 1, x2 =
1, y1 = − 1, y2 = 1, z1 = − 1, and z2 = 1.
Exercise 46. Write down the mathematical expression for Stokes theorem with the
definition of all the symbols involved and explain its meaning in words. What is this
theorem useful for? Why it is very useful?
Exercise 47. If a vector field is given in the Cartesian coordinates by A = (2y, − 3x, 1.5z),
verify Stokes theorem for a hemispherical surface x2 + y2 + z2 = 9 for z ≥ 0.
Exercise 48. Make a simple sketch to demonstrate Stokes theorem with sufficient
explanations and definitions of the symbols involved.
Exercise 49. Give concise definitions for the following terms related to matrices: matrix,
square matrix, main diagonal, trailing diagonal, transpose, identity matrix, unit matrix,
singular, trace, determinant, cofactor, and inverse.
Exercise 50. Explain the way by which matrices are indexed.
Exercise 51. How many indices are needed in indexing a 2 × 3 matrix, an n × n matrix, and
an m × k matrix? Explain, in each case, why.
Exercise 52. Does the order of the matrix indices matter? If so, what is the meaning of
changing this order?
Exercise 53. Is it possible to write a vector as a matrix? If so, what is the condition that
should be imposed on the indices and how many forms a vector can have when it is written
as a matrix?
Exercise 54. Write down the following matrices in a standard rectangular array form
(similar to the examples in Eq. 42↑) using conventional symbols for their entries with a
proper indexing: 3 × 4 matrix A, 1 × 5 matrix B, 2 × 2 matrix C, and 3 × 1 matrix D.
Exercise 55. Give detailed mathematical definitions of the determinant, trace and inverse
of matrix, explaining any symbol or technical term involved in these definitions.
Exercise 56. Find the following matrix multiplications: AB, BC, and CB where:

Exercise 57. Referring to the matrices A, B and C in the previous exercise, find all the
permutations (repetitive and non-repetitive) involving two of these three matrices, and
classify them into two groups: those which do represent possible matrix multiplication and
those which do not.
Exercise 58. Is matrix multiplication associative? commutative? distributive over matrix
addition?
Exercise 59. Calculate the trace, the determinant, and the inverse (if the inverse exists) of
the following matrices:

Exercise 60. Which, if any, of the matrices D and E in the previous exercise is singular?

Chapter 2

Tensors
In this chapter, we present the essential terms and definitions related to tensors, the
conventions and notations which are used in their representation, the general rules that
govern their manipulation, and their main types and classifications. We also provide some
illuminating examples of tensors of various complexity as well as an overview of their use
in mathematics, science and engineering.

2.1 General Background about Tensors

A tensor is an array of mathematical objects (usually numbers or functions) which
transforms according to certain rules under coordinates change. In a d-dimensional space, a
tensor of rank-n has dn components which may be specified with reference to a given
coordinate system. Accordingly, a scalar, such as temperature, is a rank-0 tensor with
(assuming a 3D space) 30 = 1 component, a vector, such as force, is a rank-1 tensor with
31 = 3 components, and stress is a rank-2 tensor with 32 = 9 components. Figure 11↓
graphically illustrates the structure of a rank-3 tensor in a 3D space.
Figure 11 Graphical illustration of a rank-3 tensor Aijk in a 3D space, i.e. each one of
i, j, k ranges over 1, 2, 3.

The dn components of a tensor are identified by n distinct integer indices (e.g. i, j, k)

which are attached, according to the commonly-employed tensor notation, as superscripts
or subscripts or a mix of these to the right side of the symbol utilized to label the tensor,
e.g. Aijk, Aijk and Ai jk. Each tensor index takes all the values over a predefined range of
dimensions such as 1 to d in the above example of a d-dimensional space. In general, all
tensor indices have the same range, i.e. they are uniformly dimensioned (see Footnote 21
in § 8↓). When the range of tensor indices is not stated explicitly, it is usually assumed to
have the values 1, 2, 3. However, the range must be stated explicitly or implicitly to avoid
ambiguity.
The characteristic property of tensors is that they satisfy the principle of invariance under
certain coordinate transformations. Therefore, formulating the fundamental physical laws in
a tensor form ensures that they are form-invariant; hence they are objectively-representing
the physical reality and do not depend on the observer. Having the same form in different
coordinate systems may be labeled as being covariant although this word is also used for a
different meaning in tensor calculus, as will be fully explained in § 2.6.1↓.
While tensors of rank-0 are generally represented in a common form of light face non-
indexed italic symbols like f and g, tensors of rank ≥ 1 are represented in several forms
and notations, the main ones are the index-free notation, which may also be called the
direct or symbolic or Gibbs notation, and the indicial notation which is also called the
index or component or tensor notation. The first is a geometrically oriented notation
with no reference to a particular reference frame and hence it is intrinsically invariant to
the choice of coordinate systems, whereas the second takes an algebraic form based on
components identified by indices and hence the notation is suggestive of an underlying
coordinate system, although being a tensor makes it form-invariant under certain coordinate
transformations and therefore it possesses certain invariant properties. The index-free
notation is usually identified by using bold face non-italic symbols, like a and B, while the
indicial notation is identified by using light face indexed italic symbols such as ai and Bij.
It is noteworthy that although rank-0 and rank-1 tensors are, respectively, scalars and
vectors, not all scalars and vectors (in their generic sense) are tensors of these ranks.
Similarly, rank-2 tensors are normally represented by square matrices but not all square
matrices represent rank-2 tensors.

2.2 General Terms and Concepts

In the following, we introduce and define a number of essential concepts and terms which
form a principal part of the technical and conceptual structure of tensor calculus. These
concepts and terms are needed in the development of the forthcoming sections and
chapters.
Tensor term is a product of tensors including scalars and vectors. Tensor expression is an
algebraic sum (or more generally a linear combination) of tensor terms which may be a
trivial sum in the case of a single term. Tensor equality (which is symbolized by “ = ”) is
an equality of two tensor terms and/or expressions. A special case of this is tensor identity
which is an equality of general validity (see Footnote 22 in § 8↓).
An index that occurs once in a tensor term is a free index while an index that occurs twice
in a tensor term is a dummy or bound index. The order of a tensor is identified by the
number of its indices (e.g. Aijk is a tensor of order 3) which normally identifies the tensor
rank as well. However, when contraction (see § 3.4↓) of indices operation takes place
once or more, the order of the tensor is not affected but its rank is reduced by two for each
contraction operation (see Footnote 23 in § 8↓). Hence, the order of a tensor is equal to the
number of all of its indices including the dummy indices, while the rank is equal to the
number of its free indices only.
Tensors with subscript indices, like Aij, are called covariant, while tensors with
superscript indices, like Ak, are called contravariant. Tensors with both types of indices,
like Alk lmn, are called mixed type. More details about this classification will follow in §
2.6.1↓. Subscript indices, rather than subscripted tensors, are also dubbed covariant and
superscript indices are dubbed contravariant.
The Zero tensor is a tensor whose all components are zero. The Unit tensor or unity
tensor, which is usually defined for rank-2 tensors, is a tensor whose all elements are zero
except the ones with identical values of all indices which are assigned the value 1.
In general terms, a transformation from an nD space to another nD space is a correlation
that maps a point from the first space (original) to a point in the second space
(transformed) where each point in the original and transformed spaces is identified by n
independent variables or coordinates. To distinguish between the two sets of coordinates
in the two spaces, the coordinates of the points in the transformed space may be notated
with barred symbols like (x̃1, x̃2, …, x̃n) or (x̃1, x̃2, …, x̃n) where the superscripts and
subscripts are indices, while the coordinates of the points in the original space are notated
with unbarred symbols like (x1, x2, …, xn) or (x1, x2, …, xn). Under certain conditions,
such a transformation is unique and hence an inverse transformation from the transformed
to the original space is also defined.
Mathematically, each one of the direct and inverse transformations can be regarded as a
mathematical correlation expressed by a set of equations in which each coordinate in one
space is considered as a function of the coordinates in the other space. Hence, the
transformations between the two sets of coordinates in the two spaces can by expressed
mathematically by the following two sets of independent relations:
(55) x̃i = x̃i(x1, x2, …, xn)
(56) xi = xi(x̃1, x̃2, …, x̃n)
where i = 1, 2, …, n.
An alternative to the latter view of considering the transformation as a mapping between
two different spaces is to view it as a correlation relating the same point in the same space
but observed from two different coordinate systems which are subject to a similar
transformation. The following will be largely based on the latter view.
Coordinate transformations are described as proper when they preserve the handedness
(right- or left-handed) of the coordinate system and improper when they reverse the
handedness. Improper transformations involve an odd number of coordinate axes
inversions in the origin of coordinates. Inversion of axes may be called improper rotation
while ordinary rotation is described as proper rotation. Figure 12↓ illustrates proper and
improper coordinate transformations of a rectangular Cartesian system.

Figure 12 Proper and improper transformations of a rectangular Cartesian coordinate

system where the former is achieved by a rotation of the coordinate system while the latter
is achieved by a rotation followed by a reflection of the first axis in the origin of
coordinates. The transformed systems are shown as dashed while the original system is
shown as solid.

Transformations can be active, when they change the state of the observed object (e.g.
translating the object in space), or passive when they are based on keeping the state of the
object and changing the state of the coordinate system from which the object is observed.
Such distinction is based on an implicit assumption of a more general frame of reference in
the background.
A permutation of a set of objects, which are normally numbers like (1, 2, …, n) or
symbols like (i, j, k), is a particular ordering or arrangement of these objects. An even
permutation is a permutation resulting from an even number of single-step exchanges (also
known as transpositions) of neighboring objects starting from a presumed original
permutation of these objects. Similarly, an odd permutation is a permutation resulting from
an odd number of such exchanges. It has been shown that when a transformation from one
permutation to another can be done in different ways, possibly with different numbers of
exchanges, the parity of all these possible transformations is the same, i.e. all are even or
all are odd, and hence there is no ambiguity in characterizing the transformation from one
permutation to another by the parity alone.

2.3 General Rules

In the following, we present some very general rules that apply to the mathematical
expressions and relations in tensor calculus. No index is allowed to occur more than
twice in a legitimate tensor term (see Footnote 24 in § 8↓). A free index should be
understood to vary over its range (e.g. 1, …, n) and hence it can be interpreted as saying
“for all components represented by the index”. Therefore, a free index represents a number
of terms or expressions or equalities equal to the number of allowed values of its range.
For example, when i and j can vary over the range 1, …, n the following expression:
Ai + Bi
represents n separate expressions while the following equation:
Aji = Bji
represents n × n separate equations.
According to the summation convention, which is widely used in the literature of tensor
calculus including in the present book, dummy indices imply summation over their range,
e.g. for an nD space we have:
(57) AiBi = Σni = 1 AiBi = A1B1 + A2B2 + … + AnBn
(58) δijAij = Σni = 1 Σnj = 1 δijAij
(59) εijkAijBk = Σni = 1 Σnj = 1 Σnk = 1 εijkAijBk
When dummy indices do not imply summation, the situation must be clarified by enclosing
such indices in parentheses or by underscoring or by using upper case letters (with
declaration of these conventions) or by adding a clarifying comment like “no summation
on repeated indices” (see Footnote 25 in § 8↓).
Each tensor index should conform to one of the forthcoming variance transformation rules
as given by Eqs. 65↓ and 66↓, i.e. it is either covariant or contravariant. For orthonormal
Cartesian coordinate systems, the two variance types (i.e. covariant and contravariant) do
not differ because the metric tensor is given by the Kronecker delta (refer to § 4.1↓ and
6↓) and hence any index can be upper or lower although it is common to use lower indices
in such cases.
For tensor invariance, a pair of dummy indices should in general be complementary in
their variance type, i.e. one covariant and the other contravariant. However, for
orthonormal Cartesian systems the two are the same and hence when both dummy indices
are covariant or both are contravariant it should be understood as an indication that the
underlying coordinate system is orthonormal Cartesian if the possibility of an error is
excluded.
As indicated earlier, tensor order is equal to the number of its indices while tensor rank is
equal to the number of its free indices; hence vectors (terms, expressions and equalities)
are represented by a single free index and rank-2 tensors are represented by two free
indices. The dimension of a tensor is determined by the range taken by its indices.
The rank of all terms in legitimate tensor expressions and equalities must be the same.
Moreover, each term in valid tensor expressions and equalities must have the same set of
free indices (e.g. i, j, k). Also, a free index should keep its variance type in every term in
valid tensor expressions and equations, i.e. it must be covariant in all terms or
contravariant in all terms.
While free indices should be named uniformly in all terms of tensor expressions and
equalities, dummy indices can be named in each term independently, e.g.
Aiik + Bjjk + Clmlmk
Dji = Ejkik + Fjmim
A free index in an expression or equality can be renamed uniformly using a different
symbol, as long as this symbol is not already in use, assuming that both symbols vary over
the same range, i.e. have the same dimension.
Examples of legitimate tensor terms, expressions and equalities are:
Aijij
Aimm + Binknk
Cij = Aij − Bji
a = Bjj
while examples of illegitimate tensor terms, expressions and equalities are:
Biii
Ai + Bij
Ai + Bj
Ai − Bi
Aii = Bi
Indexing is generally distributive over the terms of tensor expressions and equalities, for
example:
(60) [A + B]i = [A]i + [B]i
and
(61) [A = B]i [A]i = [B]i
Unlike scalars and tensor components, which are essentially scalars in a generic sense,
operators cannot in general be freely reordered in tensor terms, therefore we have:
(62) fh = hf
(63) AiBi = BiAi
but:
(64) ∂iAi ≠ Ai∂i
It should be remarked that the order of the indices (see Footnote 26 in § 8↓) of a given
tensor is important and hence it should be observed and clarified, because two tensors with
the same set of indices and with the same indicial structure but with different indicial order
are not equal in general. For example, Aijk is not equal to Ajik unless A is symmetric (refer
to § 2.6.5↓) with respect to the indices i and j. Similarly, Bmln is not equal to Blmn unless B
is symmetric in its indices l and m.
The confusion about the order of indices occurs, in particular, in the case of mixed type
tensors such as Aijk. Spaces are usually used in this case to indicate the order, e.g. the latter
tensor is symbolized as Aj i k if the order of the indices is j, i, k while it is symbolized as
Ai jk if the order of the indices is i, j, k (see Footnote 27 in § 8↓). Dots may also be used in
the case of mixed type tensors to indicate, more explicitly, the order of the indices and
remove any ambiguity. For example, if the indices i, j, k of the tensor A, which is covariant
in i and k and contravariant in j, are of that order, then A may be symbolized as Ai j. k
where the dot between i and k indicates that j is in the middle (see Footnote 28 in § 8↓).
Finally, many of the identities in the present book which are given in a covariant or a
contravariant or a mixed form are similarly valid for the other forms and hence they can be
obtained with a minimal effort. The objective of reporting in only one form is conciseness
and to avoid unnecessary repetition. Moreover, in the case of orthonormal Cartesian
systems the variance type of indices is irrelevant.

2.4 Examples of Tensors of Different Ranks

Examples of rank-0 tensors (scalars) are energy, mass, temperature, volume and density.
These are totally identified by a single number regardless of any coordinate system and
hence they are invariant under coordinate transformations (see Footnote 29 in § 8↓).
Examples of rank-1 tensors (vectors) are displacement, force, electric field, velocity and
acceleration. These require for their complete identification a number, representing their
magnitude, and a direction representing their geometric orientation within their space.
Alternatively, they can be uniquely identified by a set of numbers, equal to the number of
dimensions of the underlying space, in reference to a particular coordinate system and
hence this identification is system-dependent although they still have system-invariant
properties such as length.
Examples of rank-2 tensors are the Kronecker delta (see § 4.1↓), stress, strain, rate of
strain and inertia tensors. These require for their full identification a set of numbers each of
which is associated with two directions. These double directions are usually identified by
a set of unit dyads. Figure 13↓ is a graphic illustration of the nine unit dyads which are
associated with the double directions of rank-2 tensors in a 3D space with a rectangular
Cartesian coordinate system.
Examples of rank-3 tensors are the Levi-Civita tensor (see § 4.2↓) in 3D spaces and the
tensor of piezoelectric moduli. Examples of rank-4 tensors are the elasticity or stiffness
tensor, the compliance tensor and the fourth-order moment of inertia tensor. It is noteworthy
that tensors of high ranks are relatively rare in science and engineering.
Figure 13 The nine unit dyads associated with the double directions of rank-2 tensors in a
3D space with a rectangular Cartesian coordinate system. The vectors e i and e j (i, j = 1, 2,
3) in the dyad e ie j are unit vectors in the directions of coordinate axes where the first
indexed e represents the first vector of the dyad while the second indexed e represents the
second vector of the dyad. In these nine frames, the first vector is fixed along each row
while the second vector is fixed along each column.

2.5 Applications of Tensors

Tensor calculus is a very powerful mathematical tool; hence tensors are commonplace in
science and engineering where they are used to represent physical and synthetic objects
and ideas in mathematical invariant forms. Tensor notation and techniques are used in many
branches of mathematics, science and engineering such as differential geometry, fluid
mechanics, continuum mechanics, general relativity and structural engineering. Tensor
calculus is used for elegant and compact formulation and presentation of equations and
identities in mathematics, science and engineering. It is also used for algebraic
manipulation of mathematical expressions and proving identities in a neat and succinct way
(refer to § 5.6↓). As indicated earlier, the invariance of tensor forms serves a theoretically
and practically important role by allowing the formulation of physical laws in coordinate-
free forms.

2.6 Types of Tensor

In the following subsections we introduce a number of tensor types and categories and
highlight their main characteristics and differences. These types and categories are not
mutually exclusive and hence they overlap in general; moreover they may not be
exhaustive in their classes as some tensors may not instantiate any one of a complementary
set of types such as being symmetric or anti-symmetric.

2.6.1 Covariant and Contravariant Tensors

These are the main types of tensor with regard to the rules of their transformation between
different coordinate systems. Covariant tensors are notated with subscript indices (e.g. Ai)
while contravariant tensors are notated with superscript indices (e.g. Aij). A covariant
tensor is transformed according to the following rule:
(65)

while a contravariant tensor is transformed according to the following rule:

(66)

where the barred and unbarred symbols represent the same mathematical object (tensor or
coordinate) in the transformed and original coordinate systems respectively.
An example of covariant tensors is the gradient of a scalar field while an example of
contravariant tensors is the displacement vector. Some tensors of rank > 1 have mixed
variance type, i.e. they are covariant in some indices and contravariant in others. In this
case the covariant variables are indexed with subscripts while the contravariant variables
are indexed with superscripts, e.g. Ai j which is covariant in i and contravariant in j. A
mixed type tensor transforms covariantly in its covariant indices and contravariantly in its
contravariant indices, e.g.
(67)

To clarify the pattern of mathematical transformation of tensors, we explain step-by-

step the practical rules to follow in writing tensor transformation equations between two
coordinate systems, unbarred and barred. Since there are three types of tensors: covariant,
contravariant and mixed, we use three equations in each step.
In this demonstration we use rank-4 tensors as examples since this is sufficiently general
and hence adequate to elucidate the rules for transforming tensors of any rank. The
demonstration is based on the assumption that the transformation is taking place from the
unbarred system to the barred system; the same rules should apply for the opposite
transformation from the barred system to the unbarred system. We use the sign “≗” for the
equality in the transitional steps to indicate that the equalities are under construction and
are not complete.
We start with the very generic equations between the barred tensor à and the unbarred
tensor A for the three types:
(68)

We assume that the barred tensor and its coordinates are indexed with ijkl and the unbarred
are indexed with npqr, so we add these indices in their presumed order and position
(lower or upper) paying particular attention to the order in the mixed type:
(69)
Since the barred and unbarred tensors are of the same type, as they represent the same
tensor in two coordinate systems (see Footnote 30 in § 8↓), the indices on the two sides of
the equalities should match in their position and order. We then insert a number of partial
differential operators on the right hand side of the equations equal to the rank of these
tensors, which is 4 in our example. These operators represent the transformation rules for
each pair of corresponding coordinates, one from the barred and one from the unbarred:
(70)

Now we insert the coordinates of the barred system into the partial differential operators
noting that (i) the positions of any index on the two sides should match, i.e. both upper or
both lower, since they are free indices in different terms of tensor equalities, (ii) a
superscript index in the denominator of a partial derivative is in lieu of a covariant index
in the numerator (see Footnote 31 in § 8↓), and (iii) the order of the coordinates should
match the order of the indices in the tensor, that is:
(71)
For consistency, these coordinates should be barred as they belong to the barred tensor;
hence we add bars:
(72)

Finally, we insert the coordinates of the unbarred system into the partial differential
operators noting that (i) the positions of the repeated indices on the same side should be
opposite, i.e. one upper and one lower, since they are dummy indices and hence the
position of the index of the unbarred coordinate should be opposite to its position in the
unbarred tensor, (ii) an upper index in the denominator is in lieu of a lower index in the
numerator, and (iii) the order of the coordinates should match the order of the indices in
the tensor:
(73)
We also replaced the “≗” sign in the final set of equations with the strict equality sign “=”
as the equations now are complete.
The covariant and contravariant types of a tensor are linked through the metric tensor, as
will be detailed later in the book (refer to § 6↓). As indicated before, for orthonormal
Cartesian systems there is no difference between covariant and contravariant tensors, and
hence the indices can be upper or lower although it is common to use lower indices in this
case.
A tensor of m contravariant indices and n covariant indices may be called type (m, n)
tensor. When one or both variance types are absent, zero is used to refer to the absent
variance type in this notation. Accordingly, Aij k is a type (1, 2) tensor, Bik is a type (2, 0)
tensor, Cm is a type (0, 1) tensor, and Dpqr st is a type (2, 3) tensor.
The vectors providing the basis set for a coordinate system are of covariant type when
they are tangent to the coordinate axes, and they are of contravariant type when they are
perpendicular to the local surfaces of constant coordinates. These two sets, like the
tensors themselves, are identical for orthonormal Cartesian systems.
Formally, the covariant and contravariant basis vectors are given respectively by:
(74)

where r = xie i is the position vector in Cartesian coordinates and xi is a general curvilinear
coordinate. As before, a superscript in the denominator of partial derivatives is
equivalent to a subscript in the numerator. It should be remarked that in general the basis
vectors (whether covariant or contravariant ) are not necessarily of unit length and/or
mutually orthogonal although they may be so (see Footnote 32 in § 8↓).
The two sets of covariant and contravariant basis vectors are reciprocal systems and hence
they satisfy the following reciprocity relation:
(75) Ei⋅Ej = δi j
where δi j is the Kronecker delta (refer to § 4.1↓) which can be represented by the unity
matrix (see Special Matrices in § 1.3.3↑). The reciprocity of these two sets of basis
vectors is illustrated schematically in Figure 14↓ for the case of a 2D space.

Figure 14 The reciprocity relation between the covariant and contravariant basis vectors
in a 2D space where E1⊥ E2, E1⊥ E2, and |E1|| E1|cosφ = |E2|| E2|cosφ = 1.

A vector can be represented either by covariant components with contravariant

coordinate basis vectors or by contravariant components with covariant coordinate basis
vectors. For example, a vector A can be expressed as:
(76) A = AiEi or A = AiEi
where Ei and Ei are the contravariant and covariant basis vectors respectively. This is
illustrated graphically in Figure 15↓ for a vector A in a 2D space. The use of the covariant
or contravariant form of the vector representation is a matter of choice and convenience
since these two representations are equivalent as they represent and correctly describe the
same object.
Figure 15 The representation of a vector A in covariant and contravariant basis vector
sets in a 2D space where the components shown at the four points are with reference to unit
vectors in the given directions, e.g. A1| E1| is a component with reference to a unit vector in
the direction of E1.

More generally, a tensor of any rank ( ≥ 1) can be represented covariantly using

contravariant basis tensors of that rank, or contravariantly using covariant basis tensors,
or in a mixed form using a mixed basis of opposite type. For example, a rank-2 tensor A
can be written as:
(77) A = AijEiEj = AijEiEj = Ai jEiEj = Ai jEiEj
where EiEj, EiEj, EiEj and EiEj are dyadic products of the basis vectors of the presumed
system (refer to § 2.4↑ and 3.3↓).

2.6.2 True and Pseudo Tensors

These are also called polar and axial tensors respectively although it is more common to
use these terms for vectors. Pseudo tensors may also be called tensor densities (see
Footnote 33 in § 8↓). True tensors are proper or ordinary tensors and hence they are
invariant under coordinate transformations, while pseudo tensors are not proper tensors
since they do not transform invariantly as they acquire a minus sign under improper
orthogonal transformations which involve inversion of coordinate axes through the origin
of coordinates with a change of system handedness.
Figure 16↓ demonstrates the behavior of a true vector v and a pseudo vector p where the
former keeps its direction following a reflection of the coordinate system through the
origin of coordinates while the latter reverses its direction following this operation.

Figure 16 The behavior of a true vector (v and V) and a pseudo vector (p and P) on
reflecting the coordinate system in the origin of coordinates. The lower case symbols stand
for the objects in the original system while the upper case symbols stand for the same
objects in the reflected system.

Because true and pseudo tensors have different mathematical properties and represent
different types of physical entities, the terms of consistent tensor expressions and
equations should be uniform in their true and pseudo type, i.e. all terms are true or all are
pseudo.
The direct product (refer to § 3.3↓) of true tensors is a true tensor. The direct product of
even number of pseudo tensors is a true tensor, while the direct product of odd number of
pseudo tensors is a pseudo tensor. The direct product of a mix of true and pseudo tensors is
a true or pseudo tensor depending on the number of pseudo tensors involved in the product
as being even or odd respectively.
Similar rules to those of the direct product apply to the cross product, including the curl
operation, involving tensors (which are usually of rank-1) with the addition of a pseudo
factor for each cross product operation. This factor is contributed by the permutation
tensor ε which is implicit in the definition of the cross product (see Eqs. 173↓ and 192↓).
As we will see in § 4.2↓, the permutation tensor is a pseudo tensor.
In summary, what determines the tensor type (true or pseudo) of the tensor terms involving
direct (see Footnote 34 in § 8↓) and cross products is the parity of the multiplicative
factors of pseudo type plus the number of cross product operations involved since each
cross product operation contributes an ε factor.
Examples of true scalars are temperature, mass and the dot product of two polar or two
axial vectors, while examples of pseudo scalars are the dot product of an axial vector and
a polar vector and the scalar triple product of polar vectors. Examples of polar vectors
are displacement and acceleration, while examples of axial vectors are angular velocity
and cross product of polar vectors in general, including the curl operation on polar vectors,
due to the involvement of the permutation symbol ε which is a pseudo tensor as stated
already. As indicated before, the essence of the distinction between true (i.e. polar) and
pseudo (i.e. axial) vectors is that the direction of a pseudo vector depends on the observer
choice of the handedness of the coordinate system whereas the direction of a true vector is
independent of such a choice.
Examples of true tensors of rank-2 are stress and rate of strain tensors, while examples
of pseudo tensors of rank-2 are direct products of two vectors: one polar and one axial.
Examples of true tensors of higher ranks are piezoelectric moduli tensor (rank-3) and
elasticity tensor (rank-4), while examples of pseudo tensors of higher ranks are the
permutation tensor of these ranks.

2.6.3 Absolute and Relative Tensors

Considering an arbitrary transformation from a general coordinate system to another, a

tensor of weight w is defined by the following general tensor transformation:
(78)

where |∂x ⁄ ∂x̃| is the Jacobian of the transformation between the two systems (see
Footnote 35 in § 8↓). When w = 0 the tensor is described as an absolute or true tensor, and
when w ≠ 0 the tensor is described as a relative tensor. When w = − 1 the tensor may be
described as a pseudo tensor, while when w = 1 the tensor may be described as a tensor
density (see Footnote 36 in § 8↓). As indicated earlier, a tensor of m contravariant indices
and n covariant indices may be described as a tensor of type (m, n). This may be extended
to include the weight w as a third entry and hence the type of the tensor is identified by
(m, n, w).
Relative tensors can be added and subtracted (see § 3.1↓) if they are of the same variance
type and have the same weight (see Footnote 37 in § 8↓); the result is a tensor of the same
type and weight. Also, relative tensors can be equated if they are of the same type and
weight. Multiplication of relative tensors produces a relative tensor whose weight is the
sum of the weights of the original tensors. Hence, if the weights are added up to a non-zero
value the result is a relative tensor of that weight; otherwise it is an absolute tensor.

2.6.4 Isotropic and Anisotropic Tensors

Isotropic tensors are characterized by the property that the values of their components are
invariant under coordinate transformation by proper rotation of axes. In contrast, the
values of the components of anisotropic tensors are dependent on the orientation of the
coordinate axes. Notable examples of isotropic tensors are scalars (rank-0), the vector 0
(rank-1), Kronecker delta δij (rank-2) and Levi-Civita tensor εijk (rank-3). Many tensors
describing physical properties of materials, such as stress and magnetic susceptibility, are
anisotropic.
Direct and inner products (see § 3.3↓ and 3.5↓) of isotropic tensors are isotropic tensors.
The zero tensor of any rank is isotropic; therefore if the components of a tensor vanish in a
particular coordinate system they will vanish in all properly and improperly rotated
coordinate systems (see Footnote 38 in § 8↓). Consequently, if the components of two
tensors are identical in a particular coordinate system they are identical in all transformed
coordinate systems. This means that tensor equalities and identities are invariant under
coordinate transformations, which is one of the main motivations for the use of tensors in
mathematics and science. As indicated, all rank-0 tensors (scalars) are isotropic. Also, the
zero vector, 0, of any dimension is isotropic; in fact it is the only rank-1 isotropic tensor.

2.6.5 Symmetric and Anti-symmetric Tensors

These types of tensor apply to high ranks only (rank ≥ 2) (see Footnote 39 in § 8↓).
Moreover, these types are not exhaustive, even for tensors of rank ≥ 2, as there are high-
rank tensors which are neither symmetric nor anti-symmetric. A rank-2 tensor Aij is
symmetric iff for all i and j the following condition is satisfied:
(79) Aji = Aij
and anti-symmetric or skew-symmetric iff for all i and j the following condition is
satisfied:
(80) Aji = − Aij
Similar conditions apply to contravariant type tensors (refer also to the following).
A rank-n tensor Ai1…in is symmetric in its two indices ij and il iff the following condition
applies identically:
(81) Ai1…il…ij…in = Ai1…ij…il…in
and anti-symmetric in its two indices ij and il iff the following condition applies
identically:
(82) Ai1…il…ij…in = − Ai1…ij…il…in
Any rank-2 tensor Aij can be synthesized from (or decomposed into) a symmetric part
A(ij), which is marked with round brackets enclosing the indices, and an anti-symmetric
part A[ij], which is marked with square brackets, where the following relations apply:
(83) Aij = A(ij) + A[ij]
(84) A(ij) = (Aij + Aji) ⁄ 2
(85) A[ij] = (Aij − Aji) ⁄ 2
Similarly, a rank-3 tensor Aijk can be symmetrized by the following relation:
(86) A(ijk) = (Aijk + Akij + Ajki + Aikj + Ajik + Akji) ⁄ 3!
and anti-symmetrized by the following relation:
(87) A[ijk] = (Aijk + Akij + Ajki − Aikj − Ajik − Akji) ⁄ 3!
More generally, a rank-n tensor Ai1…in can be symmetrized by:
(88) A(i1…in) = (sum of all even &odd permutations of indices i's) ⁄ n!
and anti-symmetrized by:
(89) A[i1…in] = (sum of all even permutations minus sum of all odd permutations) ⁄ n!
A tensor of high rank ( > 2) may be symmetrized or anti-symmetrized with respect to only
some of its indices instead of all of its indices. For example, in the following the tensor A
is symmetrized and anti-symmetrized only with respect to its first two indices:
(90) A(ij)k = (Aijk + Ajik) ⁄ 2
(91) A[ij]k = (Aijk − Ajik) ⁄ 2
A tensor is described as totally symmetric iff it is symmetric with respect to all of its
indices, that is:
(92) Ai1…in = A(i1…in)
and totally anti-symmetric iff it is anti-symmetric in all of its indices, that is:
(93) Ai1…in = A[i1…in]
For a totally anti-symmetric tensor, non-zero entries can occur only when all the indices
are different.
It should be remarked that the indices whose exchange defines the symmetry and anti-
symmetry relations should be of the same variance type, i.e. both upper or both lower.
Another important remark is that the symmetry and anti-symmetry characteristic of a tensor
is invariant under coordinate transformations. Hence, a symmetric/anti-symmetric tensor in
one coordinate system is symmetric/anti-symmetric in all other coordinate systems.
Similarly, a tensor which is neither symmetric nor anti-symmetric in one coordinate system
remains so in all other coordinate systems (see Footnote 40 in § 8↓).
Finally, for a symmetric tensor Aij and an anti-symmetric tensor Bij (or the other way
around) we have the following useful and widely used identity:
(94) AijBij = 0
This is because an exchange of indices will change the sign of one tensor only and this will
change the sign of the term in the summation resulting in having a sum of terms which is
identically zero due to the fact that each term in the sum has its own negation.

2.7 Exercises
Exercise 1. Make a sketch of a rank-2 tensor Aij in a 4D space similar to Figure 11↑. What
this tensor looks like?
Exercise 2. What are the two main types of notation used for labeling tensors? State two
names for each.
Exercise 3. Make a detailed comparison between the two types of notation in the previous
question stating any advantages or disadvantages in using one of these notations or the
other. In which context each one of these notations is more appropriate to use than the
other?
Exercise 4. What is the principle of invariance of tensors and why it is one of the main
reasons for the use of tensors in science?
Exercise 5. What are the two different meanings of the term “covariant” in tensor calculus?
Exercise 6. State the type of each one of the following tensors considering the number and
position of indices (i.e. covariant, contravariant, rank, scalar, vector, etc.):
ai
Bi jk
f
bk
Cji
Exercise 7. Define the following technical terms which are related to tensors: term,
expression, equality, order, rank, zero tensor, unit tensor, free index, dummy index,
covariant, contravariant, and mixed.
Exercise 8. Which of the following is a scalar, vector or rank-2 tensor: temperature, stress,
cross product of two vectors, dot product of two vectors, and rate of strain?
Exercise 9. What is the number of entries of a rank-0 tensor in a 2D space and in a 5D
space? What is the number of entries of a rank-1 tensor in these spaces?
Exercise 10. What is the difference between the order and rank of a tensor considering the
different conventions in this regard?
Exercise 11. What is the number of entries of a rank-3 tensor in a 4D space? What is the
number of entries of a rank-4 tensor in a 3D space?
Exercise 12. Describe direct and inverse coordinate transformations between spaces and
write the generic equations for these transformations.
Exercise 13. What are proper and improper transformations? Draw a simple sketch to
demonstrate them.
Exercise 14. Define the following terms related to permutation of indices: permutation,
even, odd, parity, and transposition.
Exercise 15. Find all the permutations of the following four letters assuming no repetition:
(i, j, k, l).
Exercise 16. Give three even permutations and three odd permutations of the symbols (α,
β, γ, δ) in the stated order.
Exercise 17. Discuss all the similarities and differences between free and dummy indices.
Exercise 18. What is the maximum number of repetitive indices that can occur in each term
of a legitimate tensor expression?
Exercise 19. How many components are represented by each one of the following
assuming a 4D space?
Ai jk
f+g
Cmn − Dnm
5Dk + 4Ak = Bk
Exercise 20. What is the “summation convention”? To what type of indices this convention
applies?
Exercise 21. Is it always the case that the summation convention applies when an index is
repeated? If not, what precaution should be taken to avoid ambiguity and confusion?
Exercise 22. In which cases a pair of dummy indices should be of different variance type
(i.e. one upper and one lower)? In what type of coordinate systems these repeated indices
can be of the same variance type and why?
Exercise 23. What are the rules that the free indices should obey when they occur in the
terms of tensor expressions and equalities?
Exercise 24. What is illegitimate about the following tensor expressions and equalities
considering in your answer all the possible violations?
Aij + Bij k
Cn − Dn = Bm
Aij = Aji
Aj = f
Exercise 25. Which of the following tensor expressions and equalities is legitimate and
which is illegitimate?
Bi + Cijj
Ai − Bki
Cm + Dm = Bmmm
Bik = Aik
State in each illegitimate case all the reasons for illegitimacy.
Exercise 26. Which is right and which is wrong of the following tensor equalities?
∂nAn = An∂n
[B]k + [D]k = [B + D]k
ab = ba
AijMkl = MklAji
Explain in each case why the equality is right or wrong.
Exercise 27. Give at least two examples of tensors used in mathematics, science and
engineering for each one of the following ranks: 0, 1, 2 and 3.
Exercise 28. State the special names given to the rank-0 and rank-1 tensors.
Exercise 29. What is the difference, if any, between rank-2 tensors and matrices?
Exercise 30. Is the following statement correct? If not, re-write it correctly: “all rank-0
tensors are vectors and vice versa, and all rank-1 tensors are scalars and vice versa”.
Exercise 31. Give clear and detailed definitions of scalars and vectors and compare them.
What is common and what is different between the two?
Exercise 32. Make a simple sketch of the nine unit dyads associated with the double
directions of rank-2 tensors in a 3D space.
Exercise 33. Name three of the scientific disciplines that heavily rely on tensor calculus
notation and techniques.
Exercise 34. What are the main features of tensor calculus that make it very useful and
successful in mathematical, scientific and engineering applications.
Exercise 35. Why tensor calculus is used in the formulation and presentation of the laws of
physics?
Exercise 36. Give concise definitions for the covariant and contravariant types of tensor.
Exercise 37. Describe how the covariant and contravariant types are notated and how they
differ in their transformation between coordinate systems.
Exercise 38. Give examples of tensors used in mathematics and science which are
covariant and other examples which are contravariant.
Exercise 39. Write the mathematical transformation rules of the following tensors: Aijk to
Ãrst and Bmn to B̃ pq.
Exercise 40. Explain how mixed type tensors are defined and notated in tensor calculus.
Exercise 41. Write the mathematical rule for transforming the mixed type tensor Dij klm to
D̃ pq rst.
Exercise 42. Express the following tensors in indicial notation: a rank-3 covariant tensor
A, a rank-4 contravariant tensor B, a rank-5 mixed type tensor C which is covariant in ij
indices and contravariant in kmn indices where the indices are ordered as ikmnj.
Exercise 43. Write step-by-step, similar to the detailed example given in § 2.6.1↑, the
mathematical transformations of the following tensors: Aij to Ãrs, Blmn to B̃ pqr, Cij mn to
C̃ pq rs and Dm kl to D̃ r st.
Exercise 44. What is the relation between the rank and the (m, n) type of a tensor?
Exercise 45. Write, in indicial notation, the following tensors: A of type (0, 4), B of type
(3, 1), C of type (0, 0), D of type (3, 4), E of type (2, 0) and F of type (1, 1).
Exercise 46. What is the rank of each one of the tensors in the previous question? Are
there tensors among them which may not have been notated properly?
Exercise 47. Which tensor provides the link between the covariant and contravariant types
of a given tensor D?
Exercise 48. What coordinate system(s) in which the covariant and contravariant types of
a tensor do not differ? What is the usual tensor notation used in this case?
Exercise 49. Define in detail, qualitatively and mathematically, the covariant and
contravariant types of the basis vectors of a general coordinate system explaining all the
symbols used in your definition.
Exercise 50. Is it necessary that the basis vectors of the previous exercise are mutually
orthogonal and/or of unit length?
Exercise 51. Is the following statement correct? “A superscript in the denominator of
partial derivatives is equivalent to a superscript in the numerator”. Explain why.
Exercise 52. What is the reciprocity relation that links the covariant and contravariant
basis vectors? Express this relation mathematically.
Exercise 53. What is the interpretation of the reciprocity relation (refer to Figure 14↑ in
your explanation)?
Exercise 54. Are the covariant and contravariant forms of a specific tensor A represent the
same mathematical object? If so, in what sense they are equal from the perspective of
different coordinate systems?
Exercise 55. Correct, if necessary, the following statement: “A tensor of any rank ( ≥ 1)
can be represented covariantly using contravariant basis tensors of that rank, or
contravariantly using contravariant basis tensors, or in a mixed form using a mixed basis of
the same type”.
Exercise 56. Make corrections, if needed, to the following equations assuming a general
curvilinear coordinate system where, in each case, all the possible ways of correction
should be considered:
B = BiEi
M = MijEi
D = DiEiEj
C = CiEj
F = FnEn
T = TrsEsEr
Exercise 57. What is the technical term used to label the following objects: EiEj, EiEj,
EiEj and EiEj? What they mean?
Exercise 58. What sort of tensor components that the objects in the previous question
should be associated with?
Exercise 59. What is the difference between true and pseudo vectors? Which of these is
called axial and which is called polar?
Exercise 60. Make a sketch demonstrating the behavior of true and pseudo vectors.
Exercise 61. Is the following statement correct? “The terms of tensor expressions and
equations should be uniform in their true and pseudo type”. Explain why.
Exercise 62. There are four possibilities for the direct product of two tensors of true and
pseudo types. Discuss all these possibilities with respect to the type of the tensor produced
by this operation and if it is true or pseudo. Also discuss in detail the cross product and
curl operations from this perspective.
Exercise 63. Give examples for the true and pseudo types of scalars, vectors and rank-2
tensors.
Exercise 64. Explain, in words and equations, the meaning of absolute and relative
tensors. Do these intersect in some cases with true and pseudo tensors (at least according
to some conventions)?
Exercise 65. What “Jacobian” and “weight” mean in the context of absolute and relative
tensors?
Exercise 66. Someone stated: “A is a tensor of type (2, 4, − 1)”. What these three numbers
refer to?
Exercise 67. What is the type of the tensor in the previous exercise from the perspectives
of lower and upper indices and absolute and relative tensors? What is the rank of this
tensor?
Exercise 68. What is the weight of a tensor A produced from multiplying a tensor of
weight − 1 by a tensor of weight 2? Is A relative or absolute? Is it true or not?
Exercise 69. Define isotropic and anisotropic tensors and give examples for each using
tensors of different ranks.
Exercise 70. What is the state of the inner and outer products of two isotropic tensors?
Exercise 71. Why if a tensor equation is valid in a particular coordinate system it should
also be valid in all other coordinate systems under admissible coordinate transformations?
Use the isotropy of the zero tensor in your explanation.
Exercise 72. Define “symmetric” and “anti-symmetric” tensors and write the mathematical
condition that applies to each assuming a rank-2 tensor.
Exercise 73. Do we have symmetric/anti-symmetric scalars or vectors? If not, why?
Exercise 74. Is it the case that any tensor of rank > 1 should be either symmetric or anti-
symmetric?
Exercise 75. Give an example, writing all the components in numbers or symbols, of a
symmetric tensor of rank-2 in a 3D space. Do the same for an anti-symmetric tensor of the
same rank.
Exercise 76. Give, if possible, an example of a rank-2 tensor which is neither symmetric
nor anti-symmetric assuming a 4D space.
Exercise 77. Is it true that any rank-2 tensor can be decomposed into a symmetric part and
an anti-symmetric part? If so, write down the mathematical expressions representing these
parts in terms of the original tensor. Is this also true for a general rank-n tensor?
Exercise 78. What is the meaning of the round and square brackets which are used to
contain indices in the indexed symbol of a tensor (e.g. A(ij) and B[km]n)?
Exercise 79. Can the indices of symmetry/anti-symmetry be of different variance type?
Exercise 80. Is it possible that a rank-n (n > 2) tensor is symmetric/anti-symmetric with
respect to some, but not all, of its indices? If so, give an example of a rank-3 tensor which
is symmetric or anti-symmetric with respect to only two of its indices.
Exercise 81. For a rank-3 covariant tensor Aijk, how many possibilities of symmetry and
anti-symmetry do we have? Consider in your answer total, as well as partial, symmetry and
anti-symmetry. Is there another possibility (i.e. the tensor in neither symmetric nor anti-
symmetric with respect to any pair of its indices)?
Exercise 82. Can a tensor be symmetric with respect to some combinations of its indices
and anti-symmetric with respect to other combinations? If so, can you give a simple
example of such a tensor?
Exercise 83. Repeat the previous exercise considering the additional possibility that the
tensor is neither symmetric nor anti-symmetric with respect to another set of indices, i.e. it
is symmetric, anti-symmetric and neither with respect to different sets of indices (see
Footnote 41 in § 8↓).
Exercise 84. A is a rank-3 totally symmetric tensor and B is a rank-3 totally anti-
symmetric tensor. Write all the mathematical conditions that these tensors satisfy.
Exercise 85. Justify the following statement: “For a totally anti-symmetric tensor, non-zero
entries can occur only when all the indices are different”. Use mathematical, as well as
descriptive, language in your answer.
Exercise 86. For a totally anti-symmetric tensor Bijk in a 3D space, write all the elements
of this tensor which are identically zero. Consider the possibility that it may be easier to
find first the elements which are not identically zero, then exclude the rest (see Footnote
42 in § 8↓).

Chapter 3

Tensor Operations
There are various operations that can be performed on tensors to produce other tensors in
general. Examples of these operations are addition/subtraction, multiplication by a scalar
(rank-0 tensor), multiplication of tensors (each of rank > 0), contraction and permutation.
Some of these operations, such as addition and multiplication, involve more than one
tensor while others, such as contraction and permutation, are performed on a single
tensor. In this chapter we provide a glimpse on the main elementary tensor operations of
algebraic nature that permeate tensor algebra and calculus.
First, we should remark that the last section of this chapter, which is about the quotient rule
for tensor test, is added to this chapter because it is the most appropriate place for it in the
present book considering the dependency of the definition of this rule on other tensor
operations; otherwise the section is not about a tensor operation in the same sense as the
operations presented in the other sections of this chapter. Another remark is that in tensor
algebra division is allowed only for scalars, hence if the components of an indexed tensor
should appear in a denominator, the tensor should be redefined to avoid this, e.g. Bi =
1 ⁄ Ai.

3.1 Addition and Subtraction

Tensors of the same rank and type (see Footnote 43 in § 8↓) can be added algebraically
to produce a tensor of the same rank and type, e.g.
a=b+c
Ai = Bi − Ci
Aij = Bij + Cij
The added/subtracted terms should have the same indicial structure with regard to their
free indices, as explained in § 2.3↑; hence Aijk and Bjik cannot be added or subtracted
although they are of the same rank and type, but Amimjk and Bijk can be added and
subtracted. Addition of tensors is associative and commutative, that is (see Footnote 44
in § 8↓):
(95) (A + B) + C = A + (B + C)
(96) A + B = B + A

3.2 Multiplication of Tensor by Scalar

A tensor can be multiplied by a scalar, which generally should not be zero, to produce a
tensor of the same variance type, rank and indicial structure, e.g.
(97) Aik j = aBik j
where a is a non-zero scalar. As indicated by the equation, multiplying a tensor by a scalar
means multiplying each component of the tensor by that scalar. Multiplication by a scalar
is commutative, and associative when more than two factors are involved.

3.3 Tensor Multiplication

This may also be called outer or exterior or direct or dyadic multiplication, although
some of these names may be reserved for operations on vectors. On multiplying each
component of a tensor of rank r by each component of a tensor of rank k, both of dimension
m, a tensor of rank (r + k) with mr + k components is obtained where the variance type of
each index (covariant or contravariant) is preserved.
For example, if A and B are covariant tensors of rank-1, then on multiplying A by B we
obtain a covariant tensor C of rank-2 where the components of C are given by:
(98) Cij = AiBj
while on multiplying B by A we obtain a covariant tensor D of rank-2 where the
components of D are given by:
(99) Dij = BiAj
Similarly, if A is a contravariant tensor of rank-2 and B is a covariant tensor of rank-2,
then on multiplying A by B we obtain a mixed tensor C of rank-4 where the components of
C are given by:
(100) Cij kl = AijBkl
while on multiplying B by A we obtain a mixed tensor D of rank-4 where the components
of D are given by:
(101) Dij kl = BijAkl
In the outer product operation, it is generally understood that all the indices of the involved
tensors have the same range although this may not always be the case (see Footnote 45 in §
8↓).
In general, the outer product of tensors yields a tensor. The outer product of a tensor of
type (m, n) by a tensor of type (p, q) results in a tensor of type (m + p, n + q). This means
that the tensor rank in the outer product operation is additive and the operation conserves
the variance type of each index of the tensors involved.
The direct multiplication of tensors may be marked by the symbol:

mostly when using symbolic notation for tensors, e.g.

However, in the present book no symbol is being used for the operation of direct
multiplication and hence the operation is symbolized by putting the symbols of the tensors
side by side, e.g. AB where A and B are non-scalar tensors. In this regard, the reader
should be vigilant to avoid confusion with the operation of matrix multiplication which,
according to the notation of matrix algebra, is also symbolized as AB where A and B are
matrices of compatible dimensions, since matrix multiplication is an inner product, rather
than an outer product, operation.
The direct multiplication of tensors is not commutative in general as indicated above;
however it is distributive with respect to the algebraic sum of tensors, that is (see
Footnote 46 in § 8↓):
(102) AB ≠ BA
(103) A(B±C) = AB±AC and (B±C)A = BA±CA
As indicated before, the rank-2 tensor constructed by the direct multiplication of two
vectors is commonly called dyad. Tensors may be expressed as an outer product of vectors
where the rank of the resultant product is equal to the number of the vectors involved, e.g. 2
for dyads and 3 for triads. However, not every tensor can be synthesized as a product of
lower rank tensors. Multiplication of a tensor by a scalar (refer to § 3.2↑) may be regarded
as a special case of direct multiplication.

3.4 Contraction
The contraction operation of a tensor of rank > 1 is to make two free indices identical, by
unifying their symbols, and perform summation over these repeated indices, e.g.
(104) Aji (contraction) Aii
(105) Ajk il (contraction on jl) Amk im
Contraction results in a reduction of the rank by 2 since it implies the annihilation of two
free indices. Therefore, the contraction of a rank-2 tensor is a scalar, the contraction of a
rank-3 tensor is a vector, the contraction of a rank-4 tensor is a rank-2 tensor, and so on.
For general non-Cartesian coordinate systems, the pair of contracted indices should be
different in their variance type, i.e. one upper and one lower. Hence, contraction of a
mixed tensor of type (m, n) will, in general, produce a tensor of type (m − 1, n − 1). A
tensor of type (p, q) can, therefore, have p × q possible contractions, i.e. one contraction
for each combination of lower and upper indices.
A common example of contraction is the dot product operation on vectors (see Dot
Product of Vectors in § 1.3.2↑) which can be regarded as a direct multiplication (refer to
§ 3.3↑) of the two vectors, which results in a rank-2 tensor, followed by a contraction.
Also, in matrix algebra, taking the trace of a square matrix, by summing its diagonal
elements, can be considered as a contraction operation on the rank-2 tensor represented by
the matrix, and hence it yields the trace which is a scalar.
Conducting a contraction operation on a tensor results into a tensor. Similarly, the
application of a contraction operation on a relative tensor (see § 2.6.3↑) produces a
relative tensor of the same weight as the original tensor.

3.5 Inner Product

On taking the outer product (refer to § 3.3↑) of two tensors of rank ≥ 1 followed by a
contraction (refer to § 3.4↑) on two indices of the product, an inner product of the two
tensors is formed. Hence, if one of the original tensors is of rank-m and the other is of rank-
n, the inner product will be of rank-(m + n − 2). In the symbolic notation of tensor
calculus, the inner product operation is usually symbolized by a single dot between the two
tensors, e.g. A⋅B, to indicate the contraction operation which follows the outer
multiplication.
In general, the inner product is not commutative. When one (see Footnote 47 in § 8↓) or
both of the tensors involved in the inner product are of rank > 1 then the order of the
multiplicands does matter in general, that is (see Footnote 48 in § 8↓):
(106) A⋅B ≠ B⋅A
However, the inner product operation is distributive with respect to the algebraic sum of
tensors, that is:
(107) A⋅(B±C) = A⋅B±A⋅C and (B±C)⋅A = B⋅A±C⋅A
As indicated before (see § 3.4↑), the dot product of two vectors is an example of the inner
product of tensors, i.e. it is an inner product of two rank-1 tensors to produce a rank-0
tensor. For example, if a is a covariant vector and b is a contravariant vector, then their dot
product can be depicted as follow:
(108) [ab]i j = aibj (contraction) a⋅b = aibi
Another common example, from linear algebra, of inner product is the multiplication of a
matrix representing a rank-2 tensor, by a vector, which is a rank-1 tensor, to produce a
vector. For example, if A is a rank-2 covariant tensor and b is a contravariant vector, then
their inner product can be depicted, according to tensor calculus, as follow (see Footnote
49 in § 8↓):
(109) [Ab]ij k = Aijbk (contraction on jk) [A⋅b]i = Aijbj
This operation is equivalent to the above mentioned operation of multiplying a matrix by a
vector as defined in linear algebra.
The multiplication of two n × n matrices, as defined in linear algebra, to produce another
n × n matrix is another example of inner product (see Eq. 171↓). In this operation, each one
of the matrices involved in the multiplication, as well as the product itself, can represent a
rank-2 tensor.
For tensors whose outer product produces a tensor of rank > 2 and type (m, n) where m, n
> 0, various contraction operations between different pairs of indices of opposite variance
type can occur and hence more than one inner product, which are different in general, can
be defined. Moreover, when the outer product produces a tensor of rank > 3 and type
(m, n) where m, n > 1, more than one contraction can take place simultaneously (see
Footnote 50 in § 8↓).
There are more specialized types of inner product; some of these may be defined
differently by different authors. For example, a double inner product of two rank-2 tensors,
A and B, may be defined and denoted by double vertically- or horizontally-aligned dots
(e.g. A:B or A⋅⋅B) to indicate double contraction taking place between different pairs of
indices. An instance of these types is the inner product with double contraction of two
dyads which is commonly defined by (see Footnote 51 in § 8↓):
(110) ab:cd = (a⋅c)(b⋅d)
The single dots in the right hand side of this equation symbolize the conventional dot
product of two vectors. The result of this operation is obviously a scalar since it is the
product of two scalars, as seen from the right hand side of the equation.
Some authors may define a different type of double contraction inner product of two dyads,
symbolized by two horizontally-aligned dots, which may be called a transposed
contraction, and is given by:
(111) ab⋅⋅cd = ab:dc = (a⋅d)(b⋅c)
where the result is also a scalar. However, different authors may have different
conventions and hence one should be on the lookout for such differences.
For two rank-2 tensors, the aforementioned double contraction inner products are
similarly defined as in the case of two dyads, that is:
(112) A:B = AijBij
(113) A⋅⋅B = AijBji
Inner products with higher multiplicities of contraction are similarly defined, and hence
they can be regarded as trivial extensions of the inner products with lower contraction
multiplicities.
The inner product of tensors produces a tensor because the inner product is an outer
product operation followed by a contraction operation and both of these operations on
tensors produce tensors, as stated before.

3.6 Permutation
A tensor may be obtained by exchanging the indices of another tensor. For example, Ai kj
is a permutation of the tensor Ai jk. A common example of the permutation operation of
tensors is the transposition of a matrix (refer to Special Matrices in § 1.3.3↑) representing
a rank-2 tensor since the first and second indices, which represent the rows and columns of
the matrix, are exchanged in this operation.
It is obvious that tensor permutation applies only to tensors of rank > 1 since no exchange
of indices can occur on a scalar with no index or on a vector with a single index. The
collection of tensors obtained by permuting the indices of a reference tensor may be called
isomers.

3.7 Tensor Test: Quotient Rule

Sometimes a tensor-like object may be suspected for being a tensor; in such cases a test
based on what is called the “quotient rule” (see Footnote 52 in § 8↓) can be used to clarify
the situation. According to this rule, if the inner product of a suspected tensor with a known
tensor is a tensor then the suspect is a tensor. In more formal terms, if it is not known if A is
a tensor but it is known that B and C are tensors; moreover it is known that the following
relation holds true in all rotated (i.e. properly-transformed) coordinate frames:
(114) Apq…k…mBij…k…n = Cpq…mij…n
then A is a tensor. Here, A, B and C are respectively of ranks m, n and (m + n − 2), where
the rank of C is reduced by 2 due to the contraction on k which can be any index of A and B
independently (see Footnote 53 in § 8↓).
Testing for being a tensor can also be done by applying the first principles through direct
substitution in the transformation equations of tensors to see if the alleged tensor satisfies
the transformation rules or not. However, using the quotient rule is generally more
convenient and requires less work. It is noteworthy that the quotient rule may be
considered by some authors as a replacement for the division operation which is not
defined for tensors.

3.8 Exercises
Exercise 1. Give preliminary definitions of the following tensor operations: addition,
multiplication by a scalar, tensor multiplication, contraction, inner product and
permutation. Which of these operations involve a single tensor?
Exercise 2. Give typical examples of addition/subtraction for rank-n (0 ≤ n ≤ 3) tensors.
Exercise 3. Is it possible to add two tensors of different ranks or different variance types?
Is addition of tensors associative or commutative?
Exercise 4. Discuss, in detail, the operation of multiplication of a tensor by a scalar and
compare it to the operation of tensor multiplication. Can we regard multiplying two scalars
as an example of multiplying a tensor by a scalar?
Exercise 5. What is the meaning of the term “outer product” and what are the other terms
used to label this operation?
Exercise 6. C is a tensor of rank-3 and D is a tensor of rank-2, what is the rank of their
outer product CD? What is the rank of CD if it is subjected subsequently to a double
contraction operation?
Exercise 7. A is a tensor of type (m, n) and B is a tensor of type (s, t), what is the type of
their direct product AB?
Exercise 8. Discuss the operations of dot and cross product of two vectors (see Dot
Product of Vectors in § 1.3.2↑ and Cross Product of Vectors in 1.3.2↑) from the
perspective of the outer product operation of tensors.
Exercise 9. Are the following two statements correct (make corrections if necessary)?
“The outer multiplication of tensors is commutative but not distributive over sum of
tensors” and “The outer multiplication of two tensors may produce a scalar”.
Exercise 10. What is contraction of tensor? How many free indices are consumed in a
single contraction operation?
Exercise 11. Is it possible that the contracted indices are of the same variance type? If so,
what is the condition that should be satisfied for this to happen?
Exercise 12. A is a tensor of type (m, n) where m, n > 1, what is its type after two
contraction operations assuming a general coordinate system?
Exercise 13. Does the contraction operation change the weight of a relative tensor?
Exercise 14. Explain how the operation of multiplication of two matrices, as defined in
linear algebra, involves a contraction operation. What is the rank of each matrix and what
is the rank of the product? Is this consistent with the rule of reduction of rank by
contraction?
Exercise 15. Explain, in detail, the operation of inner product of two tensors and how it is
related to the operations of contraction and outer product of tensors.
Exercise 16. What is the rank and type of a tensor resulting from an inner product
operation of a tensor of type (m, n) with a tensor of type (s, t)? How many possibilities do
we have for this inner product considering the different possibilities of the embedded
contraction operation?
Exercise 17. Give an example of a commutative inner product of two tensors and another
example of a non-commutative inner product.
Exercise 18. Is the inner product operation distributive over algebraic addition of tensors?
Exercise 19. Give an example from matrix algebra of inner product of tensors explaining
in detail how the two are related.
Exercise 20. Discuss specialized types of inner product operations that involve more than
one contraction operation focusing in particular on the operations A:B and A⋅⋅B where A
and B are two tensors of rank > 1.
Exercise 21. A double inner product operation is conducted on a tensor of type (1, 1) with
a tensor of type (1, 2). How many possibilities do we have for this operation? What is the
rank and type of the resulting tensor? Is it covariant, contravariant or mixed?
Exercise 22. Assess the following statement considering the two meanings of the word
“tensor” related to the rank: “Inner product operation of two tensors does not necessarily
produce a tensor”. Can this statement be correct in a sense and wrong in another?
Exercise 23. What is the operation of tensor permutation and how it is related to the
operation of transposition of matrices?
Exercise 24. Is it possible to permute scalars or vectors and why?
Exercise 25. What is the meaning of the term “isomers”?
Exercise 26. Describe in detail the quotient rule and how it is used as a test for tensors.
Exercise 27. Why the quotient rule is used instead of the standard transformation equations
of tensors?
Chapter 4

delta and epsilon Tensors

In this chapter, we conduct a preliminary investigation about the δ and ε tensors and their
properties and functions as well as the relation between them. These tensors are of
particular importance in tensor calculus due to their distinctive properties and unique
transformation attributes. They are numerical tensors with fixed components in all
coordinate systems. The first is called Kronecker delta or unit tensor, while the second
is called Levi-Civita (see Footnote 54 in § 8↓), permutation, anti-symmetric and
alternating tensor.

4.1 Kronecker delta

The Kronecker δ is a rank-2 tensor in all dimensions. It is defined as:
(115)

where n is the space dimension, and hence it can be considered as the identity matrix. For
example, in a 3D space the Kronecker δ tensor is given by:
(116)
The components of the covariant, contravariant and mixed types of this tensor are the same,
that is:
(117) δij = δij = δi j = δi j
The Kronecker δ tensor is symmetric, that is:
(118) δij = δji
(119) δij = δji
where i, j = 1, 2, …, n. Moreover, it is conserved (see Footnote 55 in § 8↓) under all
proper and improper coordinate transformations. Since it is conserved under proper
transformations, it is an isotropic tensor (see Footnote 56 in § 8↓).

4.2 Permutation epsilon

The permutation tensor ε has a rank equal to the number of dimensions, and hence a rank-n
permutation tensor has nn components. The rank-2 permutation tensor εij is defined by:
(120) ε12 = 1 ε21 = − 1 ε11 = ε22 = 0
Similarly, the rank-3 permutation tensor εijk is defined by:
(121)

Figure 17↓ is a graphical illustration of the rank-3 permutation tensor εijk while Figure 18↓,
which may be used as a mnemonic device, demonstrates the cyclic nature of the three even
permutations of the indices of the rank-3 permutation tensor and the three odd permutations
of these indices assuming no repetition in indices. The three permutations in each case are
obtained by starting from a given number in the cycle and rotating in the given direction to
obtain the other two numbers in the permutation.
Figure 17 Graphical illustration of the rank-3 permutation tensor εijk where circular
nodes represent 0, square nodes represent 1 and triangular nodes represent − 1.
Figure 18 Graphical demonstration of the cyclic nature of the even and odd permutations
of the indices of the rank-3 permutation tensor assuming no repetition in indices.

The definition of the rank-n permutation tensor (i.e. εi1i2…in) is similar to the definition of
the rank-3 permutation tensor with regard to the repetition in its indices (i1, i2, ⋯, in) and
being even or odd permutations in their correspondence to (1, 2, ⋯, n), that is:
(122)

As well as the inductive definition of the permutation tensor (as given by Eqs. 120↑, 121↑
and 122↑), the permutation tensor of any rank can also be defined analytically where the
entries of the tensor are calculated from closed form formulae. The entries of the rank-2
permutation tensor can be calculated from the following closed form equation:
(123) εij = (j − i)
Similarly, for the rank-3 permutation tensor we have:
(124) εijk = (1 ⁄ 2)(j − i)(k − i)(k − j)
while for the rank-4 permutation tensor we have:
(125) εijkl = (1 ⁄ 12)(j − i)(k − i)(l − i)(k − j)(l − j)(l − k)
More generally, the entries of the rank-n permutation tensor can be obtained from the
following identity:
(126)

where S(n − 1) is the super factorial function of the argument (n − 1) which is defined by:
(127)

A simpler formula for calculating the entries of the rank-n permutation tensor can be
obtained from the previous one by dropping the magnitude of the multiplication factors and
taking their signs only, that is:
(128)

where sgn(k) is the sign function of the argument k which is defined by:
(129)

The permutation tensor is totally anti-symmetric (see § 2.6.5↑) in each pair of its indices,
i.e. it changes sign on swapping any two of its indices, that is:
(130) εi1…ik …il…in = − εi1…il…ik …in
The reason is that any exchange of two indices requires an even/odd number of single-step
shifts to the right of the first index plus an odd/even number of single-step shifts to the left
of the second index, so the total number of shifts is odd and hence it is an odd permutation
of the original arrangement.
The permutation tensor is a pseudo tensor since it acquires a minus sign under an
improper orthogonal transformation of coordinates, i.e. inversion of axes with possible
superposition of rotation (see § 2.2↑). However, it is an isotropic tensor since it is
conserved under proper coordinate transformations.
The permutation tensor may be considered as a contravariant relative tensor of weight +
1 or a covariant relative tensor of weight − 1. Hence, in 2D, 3D and nD spaces we have
the following identities for the components of the permutation tensor (see Footnote 57 in §
8↓):
(131) εij = εij
(132) εijk = εijk
(133) εi1i2…in = εi1i2…in

4.3 Useful Identities Involving delta or/and epsilon

In the following subsections we introduce and discuss a number of common identities
which involve the Kronecker and permutations tensors. Some of these identities involve
only one of these tensors while others involve both.

4.3.1 Identities Involving delta

When an index of the Kronecker delta is involved in a contraction operation by repeating

an index in another tensor in its own term, the effect of this is to replace the shared index in
the other tensor by the other index of the Kronecker delta, that is:
(134) δij Aj = Ai
In such cases the Kronecker delta is described as an index replacement or substitution
operator. Hence, we have:
(135) δij δjk = δik
Similarly:
(136) δij δjk δki = δik δki = δii = n
where n is the space dimension. The last part of this equation (i.e. δii = n) can be easily
justified by the fact that δii is the trace of the identity tensor considering the summation
convention.
Due to the fact that the coordinates are independent of each other (see § 2.2↑), we also
have the following identity (see Footnote 58 in § 8↓):
(137)

Hence, in an nD space we obtain the following identity from the last two identities:
(138) ∂ixi = δii = n
Based on the above identities and facts, the following identity can be shown to apply in
orthonormal Cartesian coordinate systems:
(139)

This identity is based on the two facts that the coordinates are independent, and the
covariant and contravariant types are the same in orthonormal Cartesian coordinate
systems.
Similarly, for a coordinate system with a set of orthonormal (see Footnote 59 in § 8↓)
basis vectors, such as the orthonormal Cartesian system, the following identity can be
easily proved:
(140) e i⋅e j = δij
where the indexed e are the basis vectors. This identity is no more than a mathematical
statement of the fact that the basis vectors in orthonormal systems are mutually orthogonal
and of unit length.
Finally, the double inner product of two dyads (see § 3.5↑) formed by an orthonormal set of
basis vectors of a given coordinate system satisfies the following identity:
(141) e ie j : e ke l = δik δjl
which is a combination of Eq. 110↑ and Eq. 140↑.

4.3.2 Identities Involving epsilon

From the definition of the rank-3 permutation tensor, we have the following identity which
demonstrates the sense of cyclic order of the non-repetitive permutations of this tensor:
(142) εijk = εkij = εjki = − εikj = − εjik = − εkji
This identity is also a demonstration of the fact that the rank-3 permutation tensor is totally
anti-symmetric in all of its indices since a shift of any two indices reverses its sign (see
Footnote 60 in § 8↓). Moreover, it reflects the fact that this tensor has only one
independent non-zero component since any one of the non-zero entries, all of which are
given by Eq. 142↑, can be obtained from any other one of these entries.
We also have the following identity for the rank-n permutation tensor (see Footnote 61 in §
8↓):
(143) εi1i2⋯in εi1i2⋯in = n!
This identity is based on the fact that the left hand side is actually the sum of the squares
of εi1i2⋯in over all the n! non-repetitive permutations of n different indices where the value
of ε of each one of these permutations is either + 1 or − 1 and hence in both cases their
square is 1.
The double inner product of the rank-3 permutation tensor and a symmetric tensor Ajk is
given by the following identity:
(144) εijk Ajk = 0
This is because an exchange of the two indices of Ajk does not affect its value due to the
symmetry of Ajk whereas a similar exchange in these indices in εijk results in a sign change;
hence each term in the sum has its own negative and therefore the total sum is identically
zero.
Another identity with a trivial outcome that involves the rank-3 permutation tensor and a
vector A is the following:
(145) εijk Ai Aj = εijk Ai Ak = εijk Aj Ak = 0
This can be explained by the fact that, due to the commutativity of ordinary multiplication,
an exchange of the indices in A’s will not affect the value but a similar exchange in the
corresponding indices of εijk will cause a change in sign; hence each term in the sum has its
own negative and therefore the total sum will be zero.
Finally, for a set of orthonormal basis vectors in a 3D space with a right-handed
coordinate system, the following identities are satisfied:
(146) e i × e j = εijk e k
(147) e i⋅(e j × e k) = εijk
These identities are based, respectively, on the forthcoming definitions of the cross product
(see Eq. 173↓) and the scalar triple product (see Eq. 174↓) in tensor notation plus the fact
that these vectors are unit vectors.

4.3.3 Identities Involving delta and epsilon

For the rank-2 permutation tensor, we have the following identity which involves the
Kronecker delta in 2D:
(148)
This identity can simply be proved inductively by building a table for the values on the left
and right hand sides as the indices are varied. The pattern of the indices in the determinant
of this identity is simple, that is the indices of the first ε provide the indices for the rows
while the indices of the second ε provide the indices for the columns (see Footnote 62 in §
8↓).
Another useful identity involving the rank-2 permutation tensor with the Kronecker delta in
2D is the following:
(149) εil εkl = δik
This can be obtained from the previous identity by replacing j with l followed by a
minimal algebraic manipulation using tensor calculus rules (see Footnote 63 in § 8↓).
Similarly, we have the following identity which correlates the rank-3 permutation tensor to
the Kronecker delta in 3D:
(150)

Again, the indices in the determinant of this identity follow the same pattern as that of Eq.
148↑.
Another useful identity in this category is the following:
(151)
This identity can be obtained from the identity of Eq. 150↑ by replacing n with k (see
Footnote 64 in § 8↓). The pattern of the indices in this identity is as before if we exclude
the repetitive indices.
More generally, the determinantal form of Eqs. 148↑ and 150↑, which link the rank-2 and
rank-3 permutation tensors to the Kronecker tensors in 2D and 3D spaces, can be extended
to link the rank-n permutation tensor to the Kronecker tensor in an nD space, that is:
(152)

Again, the pattern of the indices in the determinant of this identity in their relation to the
indices of the two epsilons follow the same rules as those of Eqs. 148↑ and 150↑.
The identity of Eq. 151↑, which may be called the epsilon-delta identity, the contracted
epsilon identity or the Levi-Civita identity, is very useful in manipulating and simplifying
tensor expressions and proving vector and tensor identities; examples of which will be
seen in § 5.6↓. The sequence of indices of the δ’s in the expanded form on the right hand
side of this identity can be easily memorized using the following mnemonic expression
(see Footnote 65 in § 8↓):
(153) (FF × SS) − (FS × SF)
where the first and second F stand respectively for the first index in the first and second ε
while the first and second S stand respectively for the second index in the first and second
ε, as illustrated graphically in Figure 19↓. The mnemonic device of Eq. 153↑ can also be
used to memorize the sequence of indices in Eq. 148↑.

Figure 19 Graphical illustration of the mnemonic device of Eq. 153↑ which is used to
remember the sequence of indices in the epsilon-delta identity of Eq. 151↑.

Other common identities in this category are:

(154) εijk εljk = 2δil
(155) εijk εijk = 2δii = 2 × 3 = 3! = 6
The first of these identities can be obtained from Eq. 151↑ with the replacement of m with j
followed by some basic tensor manipulation (see Footnote 66 in § 8↓), while the second
can be obtained from the first by replacing l with i and applying the summation convention
in 3D. The second identity is, in fact, an instance of Eq. 143↑ for a 3D space. Its restriction
to the 3D space is justified by the fact that the rank and dimension of the permutation tensor
are the same, which is 3 in this case. As indicated previously, δii is the trace of the
identity tensor, and hence in a 3D space it is equal to 3.
Another one of the common identities involving the rank-3 permutation tensor with the
Kronecker delta in 3D is the following:
(156) εijk δ1i δ2j δ3k = ε123 = 1
This identity is based on the use of the Kronecker delta as an index replacement operator
where each one of the deltas replaces an index in the permutation tensor.
Finally, the following identity can be obtained from the definition of the rank-3 permutation
tensor (Eq. 121↑) and the use of the Kronecker delta as an index replacement operator (Eq.
134↑):
(157) εijk δij = εijk δik = εijk δjk = 0

4.4 Generalized Kronecker delta

The generalized Kronecker delta is defined inductively by:
(158)
It can also be defined analytically by the following n × n determinant:
(159)

where the δij entries in the determinant are the ordinary Kronecker deltas as defined
previously. In this equation, the pattern of the indices in the generalized Kronecker delta
symbol in connection to the indices in the determinant is similar to the previous patterns,
that is the upper indices in the symbol provide the upper indices in the ordinary deltas by
indexing the rows of the determinant, while the lower indices in the symbol provide the
lower indices in the ordinary deltas by indexing the columns of the determinant.
From the above given identities, it can be shown that:
(160)
Now, on comparing the last equation with the definition of the generalized Kronecker delta,
i.e. Eq. 159↑, we conclude that:
(161)

As an instance of Eq. 161↑, the relation between the rank-n permutation tensor in its
covariant and contravariant forms and the generalized Kronecker delta in an nD space is
given by:
(162)

where the first of these equations can be obtained from Eq. 161↑ by substituting (1…n) for
(i1…in) in the two sides with relabeling j with i and noting that ε1 … n = 1, while the
second equation can be obtained from Eq. 161↑ by substituting (1…n) for (j1…jn) and
noting that ε1 … n = 1.
Hence, the permutation tensor ε can be considered as an instance of the generalized
Kronecker delta. Consequently, the rank-n permutation tensor can be written as an n × n
determinant consisting of the ordinary Kronecker deltas. Moreover, Eq. 162↑ can provide
another definition for the permutation tensor in its covariant and contravariant forms, in
addition to the previous inductive and analytic definitions of this tensor as given by Eqs.
122↑ and 126↑.
Returning to the widely used epsilon-delta identity of Eq. 151↑, if we define (see
Footnote 67 in § 8↓):
(163)
and consider the above identities which correlate the permutation tensor, the generalized
Kronecker tensor and the ordinary Kronecker tensor, then an identity equivalent to Eq.
151↑ that involves only the generalized and ordinary Kronecker deltas can be obtained,
that is:
(164)

The mnemonic device of Eq. 153↑ can also be used with this form of the identity with
minimal adjustments to the meaning of the symbols involved.
Other identities involving the permutation tensor and the ordinary Kronecker delta can also
be formulated in terms of the generalized Kronecker delta.

4.5 Exercises
Exercise 1. What “numerical tensor” means in connection with the Kronecker δ and the
permutation ε tensors?
Exercise 2. State all the names used to label the Kronecker and permutation tensors.
Exercise 3. What is the meaning of “conserved under coordinate transformations” in
relation to the Kronecker and permutation tensors?
Exercise 4. State the mathematical definition of the Kronecker δ tensor.
Exercise 5. What is the rank of the Kronecker δ tensor in an nD space?
Exercise 6. Write down the matrix representing the Kronecker δ tensor in a 3D space.
Exercise 7. Is there any difference between the components of the covariant, contravariant
and mixed types of the Kronecker δ tensor?
Exercise 8. Explain how the Kronecker δ acts as an index replacement operator giving an
example in a mathematical form.
Exercise 9. How many mathematical definitions of the rank-n permutation tensor we have?
State one of these definitions explaining all the symbols involved.
Exercise 10. What is the rank of the permutation tensor in an nD space?
Exercise 11. Make a graphical illustration of the array representing the rank-2 and rank-3
permutation tensors.
Exercise 12. Is there any difference between the components of the covariant and
contravariant types of the permutation tensor?
Exercise 13. How the covariant and contravariant types of the permutation tensor are
related to the concept of relative tensor?
Exercise 14. State the distinctive properties of the permutation tensor.
Exercise 15. How many entries the rank-3 permutation tensor has? How many non-zero
entries it has? How many independent entries it has?
Exercise 16. Is the permutation tensor true or pseudo and why?
Exercise 17. State, in words, the cyclic property of the even and odd non-repetitive
permutations of the rank-3 permutation tensor with a simple sketch to illustrate this
property.
Exercise 18. Correct the following equations:
δij Aj = Aj
δij δjk = δjk
δij δjk δki = n!
xi, j = δii
Exercise 19. In what type of coordinate system the following equation applies?
∂i xj = ∂j xi
Exercise 20. Complete the following equation assuming a 4D space:
∂i xi = ?
Exercise 21. Complete the following equations where the indexed e are orthonormal basis
vectors of a particular coordinate system:
e i⋅e j = ?
e ie j : e k e l = ?
Exercise 22. Write down the equations representing the cyclic order of the rank-3
permutation tensor. What is the conclusion from these equations with regard to the
symmetry or anti-symmetry of this tensor and the number of its independent non-zero
components?
Exercise 23. Write the analytical expressions of the rank-3 and rank-4 permutation tensors.
Exercise 24. Correct, if necessary, the following equations:
εi1⋯in εi1⋯in = n
εijk Cj Ck = 0
εijk Djk = 0
e i × e j = εijk e j
(e i × e j)⋅e k = εijk
where C is a vector, D is a symmetric rank-2 tensor, and the indexed e are orthonormal
basis vectors in a 3D space with a right-handed coordinate system.
Exercise 25. What is wrong with the following equations?
εijk δ1i δ2j δ3k = − 1
εij εkl = δik δjl + δil δjk
εil εkl = δil
εij εij = 3!
Exercise 26. Write the following in their determinantal form describing the general pattern
of the relation between the indices of ε and δ and the indices of the rows and columns of
the determinant:
εijk εlmk
εijk εlmn
εi1⋯in εj1⋯jn
Exercise 27. Give two mnemonic devices used to memorize the widely used epsilon-delta
identity and make a simple graphic illustration for one of these.
Exercise 28. Correct, if necessary, the following equations:
εrst εrst = 3!
εpst εqst = 2 δpq
εrst δrt = εrst δst
Exercise 29. State the mathematical definition of the generalized Kronecker delta.
Exercise 30. Write each one of εi1…in and εi1…in in terms of the generalized Kronecker δ.
Exercise 31. Write the mathematical relation that links the covariant permutation tensor,
the contravariant permutation tensor, and the generalized Kronecker delta.
Exercise 32. State the widely used epsilon-delta identity in terms of the generalized and
ordinary Kronecker deltas.

Chapter 5

Applications of Tensor Notation and

Techniques
In this chapter, we provide common definitions in the language of tensor calculus for some
basic concepts and operations from matrix and vector algebra. We also provide a
preliminary investigation of the nabla based differential operators and operations using, in
part, tensor notation. Common identities in vector calculus as well as the integral theorems
of Gauss and Stokes are also presented from this perspective. Finally, we provide a rather
extensive set of detailed examples about the use of tensor language and techniques in
proving common mathematical identities from vector calculus.
5.1 Common Definitions in Tensor Notation
In this section, we give some common definitions of concepts and operations from vector
and matrix algebra in tensor notation.
The trace of a matrix A representing a rank-2 tensor in an nD space is given by:
(165) tr(A) = Aii (i = 1, …, n)
For a 3 × 3 matrix representing a rank-2 tensor in a 3D space, the determinant is given by:
(166)

where the last two equalities represent the expansion of the determinant by row and by
column. Alternatively, the determinant of a 3 × 3 matrix can be given by:
(167) det(A) = (1 ⁄ 3!) εijk εlmn Ail Ajm Akn
More generally, for an n × n matrix representing a rank-2 tensor in an nD space, the
determinant is given by:
(168) det(A) = εi1⋯in A1i1…Anin = εi1⋯in Ai11…Ainn = (1 ⁄ n!)εi1⋯in εj1⋯jn Ai1j1…Ainjn
The inverse of a matrix A representing a rank-2 tensor is given by:
(169)

The multiplication of a matrix A by a vector b, as defined in linear algebra, is given by:

(170) [Ab]i = Aij bj
It should be remarked that in the writing of Ab we are using matrix notation. According to
the symbolic notation of tensors, the multiplication operation should be denoted by a dot
between the symbols of the tensor and the vector, i.e. A⋅b (see Footnote 68 in § 8↓).
The multiplication of two compatible matrices A and B, as defined in linear algebra, is
given by:
(171) [AB]ik = Aij Bjk
Again, we are using here matrix notation in the writing of AB; otherwise a dot should be
inserted between the symbols of the two matrices.
The dot product of two vectors of the same dimension is given by:
(172) A⋅B = δijAi Bj = Ai Bi
The readers are referred to § 3.5↑ for a more general definition of this type of product that
includes higher rank tensors. Similarly, the cross product of two vectors in a 3D space is
given by:
(173) [A × B]i = εijk Aj Bk
The scalar triple product of three vectors in a 3D space is given by:
(174)

while the vector triple product of three vectors in a 3D space is given by:
(175) [A × (B × C)]i = εijk εklm Aj Bl Cm
The expression of the other principal form of the vector triple product [i.e. (A × B) × C]
can be obtained from the above form by changing the order of the factors in the external
cross product and reversing the sign; other operations, like relabeling the indices and
exchanging some of the indices of the epsilons with a shift in sign, can then follow to obtain
a more organized form. The expressions of the subsidiary forms of the vector triple product
[e.g. B × (A × C) or (A × C) × B] can be obtained from the above with relabeling the
vectors in the indicial form according to their order in the symbolic form.

5.2 Scalar Invariants of Tensors

In the following, we list and write in tensor notation a number of invariants of low rank
tensors which have special importance due to their widespread applications in vector and
tensor calculus. All these invariants are scalars.
The value of a scalar (rank-0 tensor), which consists of a magnitude and a sign, is invariant
under coordinate transformations. An invariant of a vector (rank-1 tensor) under
coordinate transformations is its magnitude, i.e. length (see Footnote 69 in § 8↓). The main
three independent scalar invariants of a rank-2 tensor A under a change of basis are:
(176) I = tr(A) = Aii
(177) II = tr(A2) = Aij Aji
(178) III = tr(A3) = Aij Ajk Aki
Different forms of the three invariants of a rank-2 tensor A, which are also widely used,
are the following:
(179) I1 = I = Aii
(180) I2 = (1 ⁄ 2)(I2 − II) = (1 ⁄ 2)(Aii Ajj − Aij Aji)
(181) I3 = det(A) = (1 ⁄ 3!)(I3 − 3I II + 2III) = (1 ⁄ 3!) εijk εpqr Aip Ajq Akr
where I, II and III are as defined in the previous set of equations.
The invariants I, II and III can similarly be defined in terms of the invariants I1, I2 and I3
as follow:
(182) I = I1
(183) II = (I1)2 − 2I2
(184) III = (I1)3 − 3I1I2 + 3I3
Since the determinant of a matrix representing a rank-2 tensor is invariant, as given
above, then if the determinant vanishes in one coordinate system, it will vanish in all
transformed coordinate systems, and if not it will not. Consequently, if a rank-2 tensor is
invertible in a particular coordinate system, it will be invertible in all coordinate systems,
and if not it will not.
Ten joint invariants between two rank-2 tensors, A and B, can be formed; these are: tr(A),
tr(B), tr(A2), tr(B2), tr(A3), tr(B3), tr(A⋅B), tr(A2 ⋅B), tr(A⋅B2) and tr(A2 ⋅B2).

5.3 Common Differential Operations in Tensor

Notation
In the first part of this section, we present the most common nabla based differential
operations in Cartesian coordinates as defined by tensor notation. These operations are
based on the various types of interaction between the vector differential operator nabla ∇
with tensors of different ranks where some of these interactions involve the dot and cross
product operations. We then define in the subsequent parts of this section some of these
differential operations in the most commonly used non-Cartesian coordinate systems,
namely cylindrical and spherical, as well as in the general orthogonal coordinate
systems.
Regarding the cylindrical and spherical systems, we could have used indexed generalized
coordinates like x1, x2 and x3 to represent the cylindrical coordinates (ρ, φ, z) and the
spherical coordinates (r, θ, φ) and hence we express these operations in tensor notation as
we did for the Cartesian system. However, for more clarity at this level and to follow the
more conventional practice, we use the coordinates of these systems as suffixes in place of
the usual indices used in the tensor notation (see Footnote 70 in § 8↓). Also, for clarity and
simplicity of some formulae, we use numbers to index the coordinates of the general
orthogonal system, where a 3D space is assumed, although this is not a compact way of
representation since the formulae are given in their expanded form.

5.3.1 Cartesian Coordinate System

The nabla operator ∇ is a spatial partial differential operator which is defined in

Cartesian coordinate systems by:
(185)

The gradient of a differentiable scalar function of position f is a vector obtained by

applying the nabla operator on f and hence it is defined by:
(186)

Similarly, the gradient of a differentiable vector function of position A is the outer product
(refer to § 3.3↑) between the ∇ operator and the vector and hence it is a rank-2 tensor
given by:
(187) [∇A]ij = ∂iAj
The divergence of a differentiable vector A is the dot product of the nabla operator and
the vector A and hence it is a scalar given by:
(188)

The divergence operation can also be viewed as taking the gradient of the vector followed
by a contraction. Hence, the divergence of a vector is invariant because it is the trace of a
rank-2 tensor (see § 5.2↑; also see Footnote 71 in § 8↓).
Similarly, the divergence of a differentiable rank-2 tensor A is a vector defined in one of
its forms by:
(189) [∇⋅A]i = ∂jAji
and in another form by:
(190) [∇⋅A]j = ∂iAji
These two different forms can be given, respectively, in symbolic notation by:
(191) ∇⋅A and ∇⋅AT
where AT is the transpose of A.
More generally, the divergence of a tensor of rank n ≥ 2, which is a tensor of rank-(n − 1),
can be defined in several forms, which are different in general, depending on the
combination of the contracted indices.
The curl of a differentiable vector A is the cross product of the nabla operator and the
vector A and hence it is a vector defined by:
(192)

The curl operation may be generalized to tensors of rank > 1, and hence the curl of a
differentiable rank-2 tensor A can be defined as a rank-2 tensor given by:
(193) [∇ × A]ij = εimn ∂mAnj
The Laplacian scalar operator acting on a differentiable scalar f is given by:
(194)

The Laplacian operator acting on a differentiable vector A is defined for each component
of the vector in a similar manner to the definition of the Laplacian acting on a scalar, that
is:
2 2
(195) [∇ A]i = ∇ [ A]i = ∂jjAi
The following scalar differential operator is commonly used in science (e.g. in fluid
dynamics):
(196)

where A is a vector. As indicated earlier, the order of Ai and ∂i should be respected. The
following vector differential operator also has common applications in science:
(197) [A × ∇]i = εijk Aj ∂k
where, again, the order should be respected.
It should be remarked that the differentiation of a tensor increases its rank by one, by
introducing an extra covariant index, unless it implies a contraction in which case it
reduces the rank by one. Therefore the gradient of a scalar is a vector and the gradient of a
vector is a rank-2 tensor (∂iAj), while the divergence of a vector is a scalar and the
divergence of a rank-2 tensor is a vector (∂jAji or ∂iAji). This may be justified by the fact
that nabla ∇ is a vector operator. On the other hand the Laplacian operator does not
change the rank since it is a scalar operator; hence the Laplacian of a scalar is a scalar
and the Laplacian of a vector is a vector.

5.3.2 Cylindrical Coordinate System

For the cylindrical system identified by the coordinates (ρ, φ, z) with a set of orthonormal
basis vectors e ρ, e φ and e z we have the following definitions for the nabla based
operators and operations (see Footnote 72 in § 8↓).
The nabla operator ∇ is given by:
(198)

The Laplacian operator is given by:

(199)

The gradient of a differentiable scalar f is given by:

(200)

The divergence of a differentiable vector A is given by:

(201)

The curl of a differentiable vector A is given by:

(202)
For the plane polar coordinate system, these operators and operations can be obtained by
dropping the z components or terms from the cylindrical form of the above operators and
operations.

5.3.3 Spherical Coordinate System

For the spherical system identified by the coordinates (r, θ, φ) with a set of orthonormal
basis vectors e r, e θ and e φ we have the following definitions for the nabla based
operators and operations (see Footnote 73 in § 8↓).
The nabla operator ∇ is given by:
(203)

The Laplacian operator is given by:

(204)

The gradient of a differentiable scalar f is given by:

(205)

The divergence of a differentiable vector A is given by:

(206)
The curl of a differentiable vector A is given by:
(207)

5.3.4 General Orthogonal Coordinate System

For general orthogonal systems in a 3D space identified by the coordinates (u1, u2, u3)
with a set of unit basis vectors u1, u2 and u3 and scale factors h1, h2 and h3 where hi
= |∂r ⁄ ∂ui| and r = xie i is the position vector, we have the following definitions for the
nabla based operators and operations.
The nabla operator ∇ is given by:
(208)

The Laplacian operator is given by:

(209)

The gradient of a differentiable scalar f is given by:

(210)
The divergence of a differentiable vector A is given by:
(211)

The curl of a differentiable vector A is given by:

(212)

5.4 Common Identities in Vector and Tensor Notation

In this section, we present some of the widely used identities of vector calculus using the
traditional vector notation as well as its equivalent tensor notation. In the following bullet
points, f and h are differentiable scalar fields; A, B, C and D are differentiable vector
fields; and r = xie i is the position vector. In each bullet point, the first equation represents
the identity in vector notation while the second equation represents the identity in tensor
notation.
•
(213) ∇⋅r = n
(214) ∂ixi = n
where n is the space dimension.
•
(215) ∇ × r = 0
(216) εijk ∂jxk = 0
•
(217) ∇(a⋅r) = a
(218) ∂i(ajxj) = ai
where a is a constant vector.
•
2
(219) ∇⋅(∇f) = ∇ f
(220) ∂i(∂if) = ∂iif
•
(221) ∇⋅(∇ × A) = 0
(222) εijk ∂i∂jAk = 0
•
(223) ∇ × (∇f) = 0
(224) εijk ∂j∂kf = 0
•
(225) ∇(fh) = f∇h + h∇f
(226) ∂i(fh) = f∂ih + h∂if
•
(227) ∇⋅(fA) = f∇⋅A + A⋅∇f
(228) ∂i(fAi) = f∂iAi + Ai∂if
•
(229) ∇ × (fA) = f∇ × A + ∇f × A
(230) εijk ∂j(fAk) = f εijk ∂jAk + εijk (∂jf) Ak
•
(231) A⋅(B × C) = C⋅(A × B) = B⋅(C × A)
(232) εijk AiBjCk = εkij CkAiBj = εjki BjCkAi
•
(233) A × (B × C) = B(A⋅C) − C(A⋅B)
(234) εijk Aj εklm BlCm = Bi(AmCm) − Ci(AlBl)
•
(235) A × (∇ × B) = (∇B)⋅A − A⋅∇B
(236) εijk εklm Aj∂lBm = (∂iBm)Am − Al(∂lBi)
•
2
(237) ∇ × (∇ × A) = ∇(∇⋅A) − ∇ A
(238) εijk εklm ∂j∂lAm = ∂i(∂mAm) − ∂llAi
•
(239) ∇(A⋅B) = A × (∇ × B) + B × (∇ × A) + (A⋅∇)B + (B⋅∇)A
(240) ∂i(AmBm) = εijk Aj(εklm ∂lBm) + εijk Bj(εklm ∂lAm) + (Al∂l)Bi + (Bl∂l)Ai
•
(241) ∇⋅(A × B) = B⋅(∇ × A) − A⋅(∇ × B)
(242) ∂i(εijk AjBk) = Bk(εkij ∂iAj) − Aj(εjik ∂iBk)
•
(243) ∇ × (A × B) = (B⋅∇)A + (∇⋅B)A − (∇⋅A)B − (A⋅∇)B
(244) εijk εklm ∂j(AlBm) = (Bm∂m)Ai + (∂mBm)Ai − (∂jAj)Bi − (Aj∂j)Bi
•
(245)

(246) εijk Aj Bk εilm Cl Dm = (AlCl)(BmDm) − (AmDm)(BlCl)

•
(247) (A × B) × (C × D) = [D⋅(A × B)]C − [C⋅(A × B)]D
(248) εijk εjmn AmBn εkpq CpDq = (εqmn DqAmBn)Ci − (εpmn CpAmBn)Di
• In vector and tensor notations, the condition for a vector field A to be solenoidal is:
(249) ∇⋅A = 0
(250) ∂iAi = 0
• In vector and tensor notations, the condition for a vector field A to be irrotational is:
(251) ∇ × A = 0
(252) εijk∂jAk = 0

5.5 Integral Theorems in Tensor Notation

The divergence theorem for a differentiable vector field A in vector and tensor notations
is given by:
(253) ∭V∇⋅A dτ = ∬SA⋅n dσ
(254) ∭V∂iAidτ = ∬SAinidσ
where V is a bounded region in an nD space enclosed by a generalized surface S, dτ and dσ
are generalized volume and surface elements respectively, n and ni are unit vector normal
to the surface and its ith component respectively, and the index i ranges over 1, …, n.
Similarly, the divergence theorem for a differentiable rank-2 tensor field A in tensor
notation for the first index is given by:
(255) ∭V∂iAildτ = ∬SAilnidσ
while the divergence theorem for differentiable tensor fields of higher rank A in tensor
notation for the index k is given by:
(256) ∭V∂kAij…k…mdτ = ∬SAij…k…mnkdσ
Stokes theorem for a differentiable vector field A in vector and tensor notations is given
by:
(257) ∬S(∇ × A)⋅n dσ = ∮CA⋅dr
(258) ∬Sεijk∂jAknidσ = ∮CAidxi
where C stands for the perimeter of the surface S, and dr is a vector element tangent to the
perimeter while the other symbols are as defined before.
Similarly, Stokes theorem for a differentiable rank-2 tensor field A in tensor notation for
the first index is given by:
(259) ∬Sεijk∂jAklnidσ = ∮CAildxi
while Stokes theorem for differentiable tensor fields of higher rank A in tensor notation
for the index k is given by:
(260) ∬Sεijk∂jAlm…k…nnidσ = ∮CAlm…k…ndxk

5.6 Examples of Using Tensor Techniques to Prove

Identities
In this section we provide some examples for using tensor techniques to prove vector and
tensor identities. These examples, which are based on the identities given in § 5.4↑,
demonstrate the elegance, efficiency and clarity of the methods and notation of tensor
calculus. In each example, we justify each step of the proof by a comment contained within
parentheses which follows that step.
• ∇⋅r = n:
∇⋅r = ∂ixi
(Eq. 188↑)
= δii
(Eq. 138↑)
=n
(Eq. 138↑).
• ∇ × r = 0:
[∇ × r]i = εijk∂jxk
(Eq. 192↑)
= εijkδkj
(Eq. 137↑)
= εijj
(Eq. 134↑)
=0
(Eq. 121↑).
Since i is a free index, the identity is proved for all components.
• ∇(a⋅r) = a:
[∇(a⋅r)]i = ∂i(ajxj)
(Eqs. 186↑ & 172↑)
= aj∂ixj + xj∂iaj
(product rule)
= aj∂ixj
(aj is constant)
= ajδji
(Eq. 137↑)
= ai
(Eq. 134↑)
= [a]i
(definition of index).
Since i is a free index, the identity is proved for all components.
2
• ∇⋅(∇f) = ∇ f:
∇⋅(∇f) = ∂i[∇f]i
(Eq. 188↑)
= ∂i(∂if)
(Eq. 186↑)
= ∂i ∂i f
(rules of differentiation)
= ∂iif
(definition of 2nd derivative)
2
=∇ f
(Eq. 194↑).
• ∇⋅(∇ × A) = 0:
∇⋅(∇ × A) = ∂i[∇ × A]i
(Eq. 188↑)
= ∂i(εijk∂jAk)
(Eq. 192↑)
= εijk∂i∂jAk
(∂ not acting on ε)
= εijk∂j∂iAk
(continuity condition)
= − εjik∂j∂iAk
(Eq. 142↑)
= − εijk∂i∂jAk
(relabeling dummy indices i and j)
=0
(since εijk∂i∂jAk = − εijk∂i∂jAk).
This can also be concluded from line three by arguing that: since by the continuity
condition ∂i and ∂j can change their order with no change in the value of the term while a
corresponding change of the order of i and j in εijk results in a sign change, we see that
each term in the sum has its own negative and hence the terms add up to zero (see Eq.
145↑).
• ∇ × (∇f) = 0:
[∇ × (∇f)]i = εijk∂j[∇f]k
(Eq. 192↑)
= εijk∂j(∂kf)
(Eq. 186↑)
= εijk∂j∂kf
(rules of differentiation)
= εijk∂k∂jf
(continuity condition)
= − εikj∂k∂jf
(Eq. 142↑)
= − εijk∂j∂kf
(relabeling dummy indices j and k)
=0
(since εijk∂j∂kf = − εijk∂j∂kf).
This can also be concluded from line three by a similar argument to the one given in the
previous point. Because [∇ × (∇f)]i is an arbitrary component, then each component is
zero.
• ∇(fh) = f∇h + h∇f:
[∇(fh)]i = ∂i(fh)
(Eq. 186↑)
= f∂ih + h∂if
(product rule)
= [f∇h]i + [h∇f]i
(Eq. 186↑)
= [f∇h + h∇f]i
(Eq. 60↑).
Because i is a free index, the identity is proved for all components.
• ∇⋅(fA) = f∇⋅A + A⋅∇f:
∇⋅(fA) = ∂i[fA]i
(Eq. 188↑)
= ∂i(fAi)
(definition of index)
= f∂iAi + Ai∂if
(product rule)
= f∇⋅A + A⋅∇f
(Eqs. 188↑ & 196↑).
• ∇ × (fA) = f∇ × A + ∇f × A:
[∇ × (fA)]i = εijk∂j[fA]k
(Eq. 192↑)
= εijk∂j(fAk)
(definition of index)
= fεijk∂jAk + εijk(∂jf)Ak
(product rule & commutativity)
= fεijk∂jAk + εijk[∇f]jAk
(Eq. 186↑)
= [f∇ × A]i + [∇f × A]i
(Eqs. 192↑ & 173↑)
= [f∇ × A + ∇f × A]i
(Eq. 60↑).
Because i is a free index, the identity is proved for all components.
• A⋅(B × C) = C⋅(A × B) = B⋅(C × A):
A⋅(B × C) = εijkAiBjCk
(Eq. 174↑)
= εkijAiBjCk
(Eq. 142↑)
= εkijCkAiBj
(commutativity)
= C⋅(A × B)
(Eq. 174↑)
= εjkiAiBjCk
(Eq. 142↑)
= εjkiBjCkAi
(commutativity)
= B⋅(C × A)
(Eq. 174↑).
The negative permutations of this identity can be similarly obtained and proved by
changing the order of the vectors in the cross products which results in a sign change.
• A × (B × C) = B(A⋅C) − C(A⋅B):
[A × (B × C)]i = εijkAj[B × C]k
(Eq. 173↑)
= εijkAjεklmBlCm
(Eq. 173↑)
= εijkεklmAjBlCm
(commutativity)
= εijkεlmkAjBlCm
(Eq. 142↑)
= (δilδjm − δimδjl)AjBlCm
(Eq. 151↑)
= δilδjmAjBlCm − δimδjlAjBlCm
(distributivity)
= (δilBl)(δjmAjCm) − (δimCm)(δjlAjBl)
(commutativity & grouping)
= Bi(AmCm) − Ci(AlBl)
(Eq. 134↑)
= Bi(A⋅C) − Ci(A⋅B)
(Eq. 172↑)
= [B(A⋅C)]i − [C(A⋅B)]i
(definition of index)
= [B(A⋅C) − C(A⋅B)]i
(Eq. 60↑).
Because i is a free index, the identity is proved for all components.
Other variants of this identity [e.g. (A × B) × C] can be obtained and proved similarly by
changing the order of the factors in the external cross product with adding a minus sign.
• A × (∇ × B) = (∇B)⋅A − A⋅∇B:
[A × (∇ × B)]i = εijkAj[∇ × B]k
(Eq. 173↑)
= εijkAjεklm∂lBm
(Eq. 192↑)
= εijkεklmAj∂lBm
(commutativity)
= εijkεlmkAj∂lBm
(Eq. 142↑)
= (δilδjm − δimδjl)Aj∂lBm
(Eq. 151↑)
= δilδjmAj∂lBm − δimδjlAj∂lBm
(distributivity)
= Am∂iBm − Al∂lBi
(Eq. 134↑)
= (∂iBm)Am − Al(∂lBi)
(commutativity & grouping)
= [(∇B)⋅A]i − [A⋅∇B]i
(Eq. 187↑ & § 3.5↑)
= [(∇B)⋅A − A⋅∇B]i
(Eq. 60↑).
Because i is a free index, the identity is proved for all components.
2
• ∇ × (∇ × A) = ∇(∇⋅A) − ∇ A:
[∇ × (∇ × A)]i = εijk∂j[∇ × A]k
(Eq. 192↑)
= εijk∂j(εklm∂lAm)
(Eq. 192↑)
= εijkεklm∂j(∂lAm)
(∂ not acting on ε)
= εijkεlmk∂j∂lAm
(Eq. 142↑ & definition of derivative)
= (δilδjm − δimδjl)∂j∂lAm
(Eq. 151↑)
= δilδjm∂j∂lAm − δimδjl∂j∂lAm
(distributivity)
= ∂m∂iAm − ∂l∂lAi
(Eq. 134↑)
= ∂i(∂mAm) − ∂llAi
(∂ shift, grouping & Eq. 4↑)
2
= [∇(∇⋅A)]i − [∇ A]i
(Eqs. 188↑, 186↑ & 195↑)
2
= [∇(∇⋅A) − ∇ A]i
(Eq. 60↑).
Because i is a free index, the identity is proved for all components.
This identity can also be considered as an instance of the identity before the last one,
observing that in the second term on the right hand side the Laplacian should precede the
vector, and hence no independent proof is required.
• ∇(A⋅B) = A × (∇ × B) + B × (∇ × A) + (A⋅∇)B + (B⋅∇)A:
We start from the right hand side and end with the left hand side:
[A × (∇ × B) + B × (∇ × A) + (A⋅∇)B + (B⋅∇)A]i
= [A × (∇ × B)]i + [B × (∇ × A)]i + [(A⋅∇)B]i + [(B⋅∇)A]i
(Eq. 60↑)
= εijkAj[∇ × B]k + εijkBj[∇ × A]k + (Al∂l)Bi + (Bl∂l)Ai
(Eqs. 173↑, 188↑ & indexing)
= εijkAj(εklm∂lBm) + εijkBj(εklm∂lAm) + (Al∂l)Bi + (Bl∂l)Ai
(Eq. 192↑)
= εijkεklmAj∂lBm + εijkεklmBj∂lAm + (Al∂l)Bi + (Bl∂l)Ai
(commutativity)
= εijkεlmkAj∂lBm + εijkεlmkBj∂lAm + (Al∂l)Bi + (Bl∂l)Ai
(Eq. 142↑)
= (δilδjm − δimδjl)Aj∂lBm + (δilδjm − δimδjl)Bj∂lAm + (Al∂l)Bi + (Bl∂l)Ai
(Eq. 151↑)
= (δilδjmAj∂lBm − δimδjlAj∂lBm) + (δilδjmBj∂lAm − δimδjlBj∂lAm) + (Al∂l)Bi + (Bl∂l)Ai
(distributivity)
= δilδjmAj∂lBm − Al∂lBi + δilδjmBj∂lAm − Bl∂lAi + (Al∂l)Bi + (Bl∂l)Ai
(Eq. 134↑)
= δilδjmAj∂lBm − (Al∂l)Bi + δilδjmBj∂lAm − (Bl∂l)Ai + (Al∂l)Bi + (Bl∂l)Ai
(grouping)
= δilδjmAj∂lBm + δilδjmBj∂lAm
(cancellation)
= Am∂iBm + Bm∂iAm
(Eq. 134↑)
= ∂i(AmBm)
(product rule)
= [∇(A⋅B)]i
(Eqs. 186↑ & 188↑).
Because i is a free index, the identity is proved for all components.
• ∇⋅(A × B) = B⋅(∇ × A) − A⋅(∇ × B):
∇⋅(A × B) = ∂i[A × B]i
(Eq. 188↑)
= ∂i(εijkAjBk)
(Eq. 173↑)
= εijk∂i(AjBk)
(∂ not acting on ε)
= εijk(Bk∂iAj + Aj∂iBk)
(product rule)
= εijkBk∂iAj + εijkAj∂iBk
(distributivity)
= εkijBk∂iAj − εjikAj∂iBk
(Eq. 142↑)
= Bk(εkij∂iAj) − Aj(εjik∂iBk)
(commutativity & grouping)
= Bk[∇ × A]k − Aj[∇ × B]j
(Eq. 192↑)
= B⋅(∇ × A) − A⋅(∇ × B)
(Eq. 172↑).
• ∇ × (A × B) = (B⋅∇)A + (∇⋅B)A − (∇⋅A)B − (A⋅∇)B:
[∇ × (A × B)]i = εijk∂j[A × B]k
(Eq. 192↑)
= εijk∂j(εklmAlBm)
(Eq. 173↑)
= εijkεklm∂j(AlBm)
(∂ not acting on ε)
= εijkεklm(Bm∂jAl + Al∂jBm)
(product rule)
= εijkεlmk(Bm∂jAl + Al∂jBm)
(Eq. 142↑)
= (δilδjm − δimδjl)(Bm∂jAl + Al∂jBm)
(Eq. 151↑)
= δilδjmBm∂jAl + δilδjmAl∂jBm − δimδjlBm∂jAl − δimδjlAl∂jBm
(distributivity)
= Bm∂mAi + Ai∂mBm − Bi∂jAj − Aj∂jBi
(Eq. 134↑)
= (Bm∂m)Ai + (∂mBm)Ai − (∂jAj)Bi − (Aj∂j)Bi
(grouping)
= [(B⋅∇)A]i + [(∇⋅B)A]i − [(∇⋅A)B]i − [(A⋅∇)B]i
(Eqs. 196↑ & 188↑)
= [(B⋅∇)A + (∇⋅B)A − (∇⋅A)B − (A⋅∇)B]i
(Eq. 60↑).
Because i is a free index, the identity is proved for all components.
•
(A × B)⋅(C × D) = [A × B]i[C × D]i
(Eq. 172↑)
= εijkAjBkεilmClDm
(Eq. 173↑)
= εijkεilmAjBkClDm
(commutativity)
= (δjlδkm − δjmδkl)AjBkClDm
(Eqs. 142↑ & 151↑)
= δjlδkmAjBkClDm − δjmδklAjBkClDm
(distributivity)
= (δjlAjCl)(δkmBkDm) − (δjmAjDm)(δklBkCl)
(commutativity)
= (AlCl)(BmDm) − (AmDm)(BlCl)
(Eq. 134↑)
= (A⋅C)(B⋅D) − (A⋅D)(B⋅C)
(Eq. 172↑)

(definition).
• (A × B) × (C × D) = [D⋅(A × B)]C − [C⋅(A × B)]D:
[(A × B) × (C × D)]i = εijk[A × B]j[C × D]k
(Eq. 173↑)
= εijkεjmnAmBnεkpqCpDq
(Eq. 173↑)
= εijkεkpqεjmnAmBnCpDq
(commutativity)
= εijkεpqkεjmnAmBnCpDq
(Eq. 142↑)
= (δipδjq − δiqδjp)εjmnAmBnCpDq
(Eq. 151↑)
= (δipδjqεjmn − δiqδjpεjmn)AmBnCpDq
(distributivity)
= (δipεqmn − δiqεpmn)AmBnCpDq
(Eq. 134↑)
= δipεqmnAmBnCpDq − δiqεpmnAmBnCpDq
(distributivity)
= εqmnAmBnCiDq − εpmnAmBnCpDi
(Eq. 134↑)
= εqmnDqAmBnCi − εpmnCpAmBnDi
(commutativity)
= (εqmnDqAmBn)Ci − (εpmnCpAmBn)Di
(grouping)
= [D⋅(A × B)]Ci − [C⋅(A × B)]Di
(Eq. 174↑)
= [[D⋅(A × B)]C]i − [[C⋅(A × B)]D]i
(index definition)
= [[D⋅(A × B)]C − [C⋅(A × B)]D]i
(Eq. 60↑).
Because i is a free index, the identity is proved for all components.

5.7 Exercises
Exercise 1. Write, in tensor notation, the mathematical expression for the trace,
determinant and inverse of an n × n matrix.
Exercise 2. Repeat the previous exercise for the multiplication of a matrix by a vector and
the multiplication of two n × n matrices.
Exercise 3. Define mathematically the dot and cross product operations of two vectors
using tensor notation.
Exercise 4. Repeat the previous exercise for the scalar triple product and vector triple
product operations of three vectors.
Exercise 5. Define mathematically, using tensor notation, the three scalar invariants of a
rank-2 tensor: I, II and III.
Exercise 6. Express the three scalar invariants I, II and III in terms of the other three
invariants I1, I2 and I3 and vice versa.
Exercise 7. Explain why the three invariants I1, I2 and I3 are scalars using in your
argument the fact that the three main invariants I, II and III are traces?
Exercise 8. Justify, giving a detailed explanation, the following statement: “If a rank-2
tensor is invertible in a particular coordinate system it is invertible in all other coordinate
systems, and if it is singular in a particular coordinate system it is singular in all other
coordinate systems”. Use in your explanation the fact that the determinant is invariant under
admissible coordinate transformations.
Exercise 9. What are the ten joint invariants between two rank-2 tensors?
Exercise 10. Provide a concise mathematical definition of the nabla differential operator
∇ in Cartesian coordinate systems using tensor notation.
Exercise 11. What is the rank and variance type of the gradient of a differentiable scalar
field in general curvilinear coordinate systems?
Exercise 12. State, in tensor notation, the mathematical expression for the gradient of a
differentiable scalar field in a Cartesian system.
Exercise 13. What is the gradient of the following scalar functions of position f, g and h
where x1, x2 and x3 are the Cartesian coordinates and a, b and c are constants?
f = 1.3x1 − 2.6ex2 + 19.8x3
g = ax3 + bex2
h = a(x1)3 − sinx3 + c(x3)2
Exercise 14. State, in tensor notation, the mathematical expression for the gradient of a
differentiable vector field in a Cartesian system.
Exercise 15. What is the gradient of the following vector where x1, x2 and x3 are the
Cartesian coordinates?
V = (2x1 − 1.2x2, x1 + x3, x2x3)
What is the rank of this gradient?
Exercise 16. Explain, in detail, why the divergence of a vector is invariant.
Exercise 17. What is the rank of the divergence of a rank-n (n > 0) tensor and why?
Exercise 18. State, using vector and tensor notations, the mathematical definition of the
divergence operation of a vector in a Cartesian coordinate system.
Exercise 19. Discuss in detail the following statement: “The divergence of a vector is a
gradient operation followed by a contraction”. How this is related to the trace of a rank-2
tensor?
Exercise 20. Write down the mathematical expression of the two forms of the divergence
of a rank-2 tensor.
Exercise 21. How many forms do we have for the divergence of a rank-n (n > 0) tensor
and why? Assume in your answer that the divergence operation can be conducted with
respect to any one of the tensor indices.
Exercise 22. Find the divergence of the following vectors U and V where x1, x2 and x3 are
the Cartesian coordinates:
U = (9.3x1, 6.3cosx2, 3.6x1e − 1.2x3)
V = (x2sinx1, 5(x2)3, 16.3x3)
Exercise 23. State, in tensor notation, the mathematical expression for the curl of a vector
and of a rank-2 tensor assuming a Cartesian coordinate system.
Exercise 24. Define, in tensor notation, the Laplacian operator acting on a differentiable
scalar field in a Cartesian coordinate system.
Exercise 25. Is the Laplacian a scalar or a vector operator?
Exercise 26. What is the meaning of the Laplacian operator acting on a differentiable
vector field?
Exercise 27. What is the rank of a rank-n tensor acted upon by the Laplacian operator?
Exercise 28. Define mathematically the following operators assuming a Cartesian
coordinate system:
A⋅∇
A×∇
What is the rank of each one of these operators?
Exercise 29. Make a general statement about how differentiation of tensors affects their
rank discussing in detail from this perspective the gradient and divergence operations.
Exercise 30. State the mathematical expressions for the following operators and
operations assuming a cylindrical coordinate system: nabla operator, Laplacian operator,
gradient of a scalar, divergence of a vector, and curl of a vector.
Exercise 31. Explain how the expressions for the operators and operations in the previous
exercise can be obtained for the plane polar coordinate system from the expressions of the
cylindrical system.
Exercise 32. State the mathematical expressions for the following operators and
operations assuming a spherical coordinate system: nabla operator, Laplacian operator,
gradient of a scalar, divergence of a vector, and curl of a vector.
Exercise 33. Repeat the previous exercise for the general orthogonal coordinate system.
Exercise 34. Express, in tensor notation, the mathematical condition for a vector field to
be solenoidal.
Exercise 35. Express, in tensor notation, the mathematical condition for a vector field to
be irrotational.
Exercise 36. Express, in tensor notation, the divergence theorem for a differentiable vector
field explaining all the symbols involved. Repeat the exercise for a differentiable tensor
field of an arbitrary rank ( > 0).
Exercise 37. Express, in tensor notation, Stokes theorem for a differentiable vector field
explaining all the symbols involved. Repeat the exercise for a differentiable tensor field of
an arbitrary rank ( > 0).
Exercise 38. Express the following identities in tensor notation:
∇⋅r = n
∇(a⋅r) = a
∇⋅(∇ × A) = 0
∇(fh) = f∇h + h∇f
∇ × (fA) = f∇ × A + ∇f × A
A × (B × C) = B(A⋅C) − C(A⋅B)
2
∇ × (∇ × A) = ∇(∇⋅A) − ∇ A
∇⋅(A × B) = B⋅(∇ × A) − A⋅(∇ × B)
Exercise 39. Prove the following identities using the language and techniques of tensor
calculus:
∇×r=0
2
∇⋅(∇f) = ∇ f
∇ × (∇f) = 0
∇⋅(fA) = f∇⋅A + A⋅∇f
A⋅(B × C) = C⋅(A × B) = B⋅(C × A)
A × (∇ × B) = (∇B)⋅A − A⋅∇B
∇(A⋅B) = A × (∇ × B) + B × (∇ × A) + (A⋅∇)B + (B⋅∇)A
∇ × (A × B) = (B⋅∇)A + (∇⋅B)A − (∇⋅A)B − (A⋅∇)B
(A × B) × (C × D) = [D⋅(A × B)]C − [C⋅(A × B)]D

Chapter 6

Metric Tensor
The subject of the present chapter is the metric tensor which is one of the most important
special tensors, if not the most important of all, in tensor calculus. Its versatile usage and
functionalities permeate the whole discipline of tensor calculus.
The metric tensor is a rank-2 tensor which may also be called the fundamental tensor.
The main purpose of the metric tensor is to generalize the concept of distance to general
curvilinear coordinate frames and maintain the invariance of distance in different
coordinate systems.
In orthonormal Cartesian coordinate systems the distance element squared, (ds)2, between
two infinitesimally neighboring points in space, one with coordinates xi and the other with
coordinates xi + dxi, is given by:
(261) (ds)2 = dxidxi = δijdxidxj
This definition of distance is the key to introducing a rank-2 tensor, gij, called the metric
tensor which, for a general coordinate system, is defined by:
(262) (ds)2 = gijdxidxj
The above defined metric tensor is of covariant type. The metric tensor has also a
contravariant type which is usually notated with gij.
The components of the covariant and contravariant metric tensor are given by:
(263) gij = Ei⋅Ej and gij = Ei⋅Ej
where the indexed E are the covariant and contravariant basis vectors as defined in §
2.6.1↑.
The metric tensor has also a mixed type which is given by:
(264) gi j = Ei⋅Ej = δi j and gi j = Ei⋅Ej = δi j
and hence it is the same as the unity tensor.
For a coordinate system in which the metric tensor can be cast in a diagonal form where
the diagonal elements are ±1 the metric is called flat. For Cartesian coordinate systems,
which are orthonormal flat-space systems, we have:
(265) gij = δij = gij = δij
The metric tensor is symmetric in its two indices, that is:
(266) gij = gji and gij = gji
This can be easily explained by the commutativity of the dot product of vectors in reference
to the above equations involving the dot product of the basis vectors.
The contravariant metric tensor is used for raising covariant indices of covariant and
mixed tensors, e.g.
(267) Ai = gikAk and Aij = gikAk j
Similarly, the covariant metric tensor is used for lowering contravariant indices of
contravariant and mixed tensors, e.g.
(268) Ai = gikAk and Aij = gikAk j
In these raising and lowering operations the metric tensor acts, like a Kronecker delta, as
an index replacement operator as well as shifting the position of the index.
Because it is possible to shift the index position of a tensor by using the covariant and
contravariant types of the metric tensor, a given tensor can be cast into a covariant or a
contravariant form, as well as a mixed form in the case of tensors of rank > 1. However, it
should be emphasized that the order of the indices must be respected in this process,
because two tensors with the same indicial structure but with different indicial order are
not equal in general, as stated before. For example:
(269) Ai j = gjkAik ≠ Aj i = gjkAki
Some authors insert dots (e.g. Ai⋅ j and A⋅j i) to remove any ambiguity about the order of
the indices.
The covariant and contravariant metric tensors are inverses of each other, that is:
(270) [gij] = [gij] − 1 and [gij] = [gij] − 1
Hence:
(271) gikgkj = δi j and gikgkj = δi j
It is common to reserve the term metric tensor to the covariant form and call the
contravariant form, which is its inverse, the associate or conjugate or reciprocal metric
tensor. As a tensor, the metric has a significance regardless of any coordinate system
although it requires a coordinate system to be represented in a specific form. For
orthogonal coordinate systems the metric tensor is diagonal, i.e. gij = gij = 0 for i ≠ j.
As indicated before, for orthonormal Cartesian coordinate systems in a 3D space, the
metric tensor is given in its covariant and contravariant forms by the 3 × 3 unit matrix, that
is:
(272)

For cylindrical coordinate systems with coordinates (ρ, φ, z), the metric tensor is given in
its covariant and contravariant forms by:
(273)

while for spherical coordinate systems with coordinates (r, θ, φ), the metric tensor is
given in its covariant and contravariant forms by:
(274)
As seen, all these metric tensors are diagonal since all these coordinate systems are
orthogonal. We also notice that all the corresponding diagonal elements of the covariant
and contravariant types are reciprocals of each other. This can be easily explained by the
fact that these two types are inverses of each other, plus the fact that the inverse of an
invertible diagonal matrix is a diagonal matrix obtained by taking the reciprocal of the
corresponding diagonal elements of the original matrix, as stated in Inverse of Matrix in §
1.3.3↑.

6.1 Exercises
Exercise 1. Describe in details, using mathematical tensor language when necessary, the
metric tensor discussing its rank, purpose, designations, variance types, symmetry, its role
in the definition of distance, and its relation to the covariant and contravariant basis
vectors.
Exercise 2. What is the relation between the covariant and contravariant types of the
metric tensor? Express this relation mathematically. Also define mathematically the mixed
type metric tensor.
Exercise 3. Correct, if necessary, the following equations:
gi j = δi i
gij = Ei⋅Ej
(ds) = gijdxidxj
gij = Ei⋅Ej
Ei⋅Ej = δj i
Exercise 4. What “flat metric” means? Give an example of a coordinate system with a flat
metric.
Exercise 5. Describe the index-shifting (raising/lowering) operators and their relation to
the metric tensor. How these operators facilitate the transformation between the covariant,
contravariant and mixed types of a given tensor?
Exercise 6. What is wrong with the following equations?
Ci = gijCj
Di = gijDj
Ai = δijAj
Make the necessary corrections considering all the possibilities in each case.
Exercise 7. Is it necessary to keep the order of the indices which are shifted by the index-
shifting operators and why?
Exercise 8. How and why dots may be inserted to avoid confusion about the order of the
indices following an index-shifting operation?
Exercise 9. Express, mathematically, the fact that the contravariant and covariant metric
tensors are inverses of each other.
Exercise 10. Correct, if necessary, the following statement: “The term metric tensor is
usually used to label the covariant form of the metric, while the contravariant form of the
metric is called the conjugate or associate or reciprocal metric tensor”.
Exercise 11. Write, in matrix form, the covariant and contravariant types of the metric
tensor of the Cartesian, cylindrical and spherical coordinate systems of a 3D flat space.
Exercise 12. Regarding the previous question, what do you notice about the corresponding
diagonal elements of the covariant and contravariant types of the metric tensor in these
systems? Does this relate to the fact that these types are inverses of each other?

Chapter 7

Covariant Differentiation
The focus of this chapter is the operation of covariant differentiation of tensors which, in a
sense, is a generalization of the ordinary differentiation. The ordinary derivative of a
tensor is not a tensor in general. The objective of covariant differentiation is to ensure the
invariance of derivative (i.e. being a tensor) in general coordinate systems, and this results
in applying more sophisticated rules using Christoffel symbols where different
differentiation rules for covariant and contravariant indices apply. The resulting covariant
derivative is a tensor which is one rank higher than the differentiated tensor.
The Christoffel symbol of the second kind is defined by:
(275)
where the indexed g is the metric tensor in its contravariant and covariant forms with
implied summation over l. It is noteworthy that Christoffel symbols are not tensors. The
Christoffel symbols of the second kind are symmetric in their two lower indices, that is:
(276) {kij} = {kji}
For Cartesian coordinate systems, the Christoffel symbols are zero for all values of the
indices. For cylindrical coordinate systems, marked with the coordinates (ρ, φ, z), the
Christoffel symbols are zero for all values of the indices except:
(277) {122} = − ρ
(278) {212} = {221} = 1 ⁄ ρ
where (1, 2, 3) stand for (ρ, φ, z).
For spherical coordinate systems, marked with the coordinates (r, θ, φ), the Christoffel
symbols are zero for all values of the indices except:
(279) {122} = − r
(280) {133} = − r sin2θ
(281) {212} = {221} = 1 ⁄ r
(282) {233} = − sinθ cosθ
(283) {313} = {331} = 1 ⁄ r
(284) {323} = {332} = cotθ
where (1, 2, 3) stand for (r, θ, φ).
For a differentiable scalar f, the covariant derivative is the same as the ordinary partial
derivative, that is:
(285) f;i = f, i = ∂if
This is justified by the fact that the covariant derivative is different from the ordinary
partial derivative because the basis vectors in general coordinate systems are dependent on
their spatial position, and since a scalar is independent of the basis vectors the covariant
and partial derivatives are identical.
For a differentiable vector A, the covariant derivative of the covariant and contravariant
forms of the vector is given by:
(286) Aj;i = ∂iAj − {kji}Ak (covariant)
(287) Aj ;i = ∂iAj + {jki}Ak (contravariant)
For a differentiable rank-2 tensor A, the covariant derivative of the covariant,
contravariant and mixed forms of the tensor is given by:
(288) Ajk;i = ∂iAjk − {lji}Alk − {lki}Ajl (covariant)
(289) Ajk ;i = ∂iAjk + {jli}Alk + {kli}Ajl (contravariant)
(290) Ak j;i = ∂iAkj + {kli}Alj − {lji}Akl (mixed)
More generally, for a differentiable rank-n tensor A, the covariant derivative is given by:
(291)

From the last equations, a pattern for the covariant differentiation operation emerges, that
is it starts with an ordinary partial derivative term, then for each tensor index an extra
Christoffel symbol term is added, positive for superscripts and negative for subscripts,
where the differentiation index is the second of the lower indices in the Christoffel symbol.
Since the Christoffel symbols are identically zero in Cartesian coordinate systems, the
covariant derivative is the same as the ordinary partial derivative for all tensor ranks.
Another important fact about covariant differentiation is that the covariant derivative of the
metric tensor in its covariant, contravariant and mixed forms is zero in all coordinate
systems and hence it is treated like a constant in covariant differentiation.
Several rules of ordinary differentiation similarly apply to covariant differentiation. For
example, covariant differentiation is a linear operation with respect to algebraic sums of
tensor terms, that is:
(292) ∂;i(aA±bB) = a ∂;iA±b ∂;iB
where a and b are scalar constants and A and B are differentiable tensor fields.
The product rule of ordinary differentiation also applies to covariant differentiation of
tensor multiplication, that is:
(293) ∂;i(AB) = (∂;iA)B + A∂;iB
However, as seen in this equation, the order of the tensors should be observed since tensor
multiplication, unlike ordinary algebraic multiplication, is not commutative. The product
rule is also valid for the inner product of tensors because the inner product is an outer
product operation followed by a contraction of indices, and covariant differentiation and
contraction of indices do commute.
Since the covariant derivative of the metric tensor is identically zero, as stated above, the
covariant derivative operator bypasses the index raising/lowering operator, that is:
(294) ∂;m(gijAj) = gij∂;mAj
and hence the metric tensor behaves like a constant with respect to the covariant
differential operator.
A principal difference between the partial differentiation and the covariant differentiation
is that for successive differential operations with respect to different indices the partial
derivative operators do commute with each other, assuming certain continuity conditions,
but the covariant differential operators do not commute, that is:
(295) ∂i∂j = ∂j∂i but ∂;i∂;j ≠ ∂;j∂;i
Higher order covariant derivatives are similarly defined as derivatives of derivatives;
however the order of differentiation, in the case of differentiating with respect to different
indices, should be respected, as explained above.

7.1 Exercises
Exercise 1. Is the ordinary derivative of a tensor necessarily a tensor or not? Can the
ordinary derivative of a tensor be a tensor? If so, give an example.
Exercise 2. Explain the purpose of the covariant derivative and how it is related to the
invariance property of tensors.
Exercise 3. Is the covariant derivative of a tensor necessarily a tensor? If so, what is the
rank of the covariant derivative of a rank-n tensor?
Exercise 4. How is the Christoffel symbol of the second kind symbolized? Describe the
arrangement of its indices.
Exercise 5. State the mathematical definition of the Christoffel symbol of the second kind
in terms of the metric tensor defining all the symbols involved.
Exercise 6. The Christoffel symbols of the second kind are symmetric in which of their
indices?
Exercise 7. Why the Christoffel symbols of the second kind are identically zero in the
Cartesian coordinate systems? Use in your explanation the mathematical definition of the
Christoffel symbols.
Exercise 8. Give the Christoffel symbols of the second kind for the cylindrical and
spherical coordinate systems explaining the meaning of the indices used.
Exercise 9. What is the meaning, within the context of tensor differentiation, of the comma
“,” and semicolon “;” when used as subscripts preceding a tensor index?
Exercise 10. Why the covariant derivative of a differentiable scalar is the same as the
ordinary partial derivative and how is this related to the basis vectors of coordinate
systems?
Exercise 11. Differentiate the following tensors covariantly:
As
Bt
Cji
Dpq
Emn
Alm…pij…k
Exercise 12. Explain the mathematical pattern followed in the operation of covariant
differentiation of tensors. Does this pattern also apply to rank-0 tensors?
Exercise 13. The covariant derivative in Cartesian coordinate systems is the same as the
ordinary partial derivative for all tensor ranks. Explain why.
Exercise 14. What is the covariant derivative of the covariant and contravariant forms of
the metric tensor for an arbitrary type of coordinate system? How is this related to the fact
that the covariant derivative operator bypasses the index-shifting operator?
Exercise 15. Which rules of ordinary differentiation apply equally to covariant
differentiation and which do not? Make mathematical statements about all these rules with
sufficient explanation of the symbols and operations involved.
Exercise 16. Make corrections, where necessary, in the following equations explaining in
each case why the equation should or should not be amended:
(C±D);i = ∂;iC±∂;iD
∂;i(AB) = B(∂;iA) + A∂;iB
(gijAj);m = gij(Aj);m
∂;i∂;j = ∂;j∂;i
∂i ∂j = ∂j ∂i
Exercise 17. How do you define the second and higher order covariant derivatives of
tensors? Do these derivatives follow the same rules as the ordinary partial derivatives of
the same order in the case of different differentiation indices?

References
• G.B. Arfken; H.J. Weber; F.E. Harris. Mathematical Methods for Physicists A
Comprehensive Guide. Elsevier Academic Press, seventh edition, 2013.
• R.B. Bird; R.C. Armstrong; O. Hassager. Dynamics of Polymeric Liquids, volume 1. John
Wiley & Sons, second edition, 1987.
• R.B. Bird; W.E. Stewart; E.N. Lightfoot. Transport Phenomena. John Wiley & Sons,
second edition, 2002.
• M.L. Boas. Mathematical Methods in the Physical Sciences. John Wiley & Sons Inc.,
third edition, 2006.
• C.F. Chan Man Fong; D. De Kee; P.N. Kaloni. Advanced Mathematics for Engineering
and Science. World Scientific Publishing Co. Pte. Ltd., first edition, 2003.
• T.L. Chow. Mathematical Methods for Physicists: A concise introduction. Cambridge
University Press, first edition, 2003.
• J.H. Heinbockel. Introduction to Tensor Calculus and Continuum Mechanics. 1996.
• D.C. Kay. Schaum’s Outline of Theory and Problems of Tensor Calculus. McGraw-Hill,
first edition, 1988.
• K.F. Riley; M.P. Hobson; S.J. Bence. Mathematical Methods for Physics and Engineering.
Cambridge University Press, third edition, 2006.
• D. Zwillinger, editor. CRC Standard Mathematical Tables and Formulae. CRC Press,
32nd edition, 2012.

Footnotes
Footnote 1. Oblique coordinate systems are usually characterized by similar features as
orthonormal Cartesian systems but with non-perpendicular axes.
Footnote 2. Since matrices in this book are supposed to represent rank-2 tensors, they also
follow the rules of labeling tensors symbolically by using non-indexed upper case bold
face non-italic Latin letters.
Footnote 3. This mainly applies in this book to Cartesian coordinate systems.
Footnote 4. As indicated, in notations like ∂r the subscript is used as a label rather than an
index.
Footnote 5. There are special tensors of numerical nature, such as the Kronecker and
permutation tensors, which do not require a particular coordinate system for their full and
unambiguous identification since their components are invariant under coordinate
transformations (the reader is referred to Chapter 4↑ for details). However, this issue may
be debated.
Footnote 6. As indicated before, these features are what qualify a Cartesian system to be
described as orthonormal.
Footnote 7. In fact, the basis vectors are constant only in rectangular Cartesian and oblique
Cartesian coordinate systems. As indicated before, the oblique Cartesian systems are the
same as the rectangular Cartesian but with the exception that their axes are not mutually
orthogonal. Also, labeling the oblique as Cartesian may be controversial.
Footnote 8. In the second equation, arctan(y ⁄ x) should be selected consistent with the
signs of x and y to be in the right quadrant.
Footnote 9. Again, arctan(y ⁄ x) in the third equation should be selected consistent with the
signs of x and y.
Footnote 10. This equality can be explained by the fact that the magnitude of the scalar
triple product is equal to the volume of the parallelepiped whereas the cyclic permutation
preserves the sign of these six possibilities.
Footnote 11. The stated facts about the other six possibilities can be explained by the two
invariances (as explained in the previous footnote) plus the fact that the cross product
operation is anti-commutative.
Footnote 12. Associativity here should be understood in its context as stated by the
inequality.
Footnote 13. Although we generally use (x, y, z) for the Cartesian coordinates of a
particular point in the space while we use (x1, x2, x3) to label the axes and the coordinates
of the Cartesian system in general, in this subsection we use (x, y, z), instead of
(x1, x2, x3), because it is more commonly used in vector algebra and calculus and is
notationally clearer, especially at this level and at this stage in the book. We also label the
basis vectors of the Cartesian system with (i, j, k), instead of (e 1, e 2, e 3), for similar
reasons.
Footnote 14. As we will see later in the book, similar differential operations can also be
defined and performed on higher rank tensor fields.
Footnote 15. This operator is also known as the harmonic operator.
Footnote 16. In this book, only square matrices are of primary interest as they are
qualified to represent uniformly dimensioned rank-2 tensors. Row and column matrices are
also qualified to represent employed vectors.
Footnote 17. The indexing of the entries of AT is not standard; the purpose of this is to
demonstrate the exchange of rows and columns.
Footnote 18. In brief, AB represents an inner product of A and B according to the matrix
notation, and an outer product of A and B according to the symbolic notation of tensors,
while A⋅B represents an inner product of A and B according to the symbolic notation of
tensors. Hence, AB in the matrix notation is equivalent to A⋅B in the symbolic notation of
tensors.
Footnote 19. In fact, this rule for the determinant of an n × n matrix applies even to the 2 ×
2 matrix if the cofactor of an entry in this case is taken as a single entry with the designated
sign. However, we separated the 2 × 2 matrix case in the definition to be more clear and to
avoid possible confusion.
Footnote 20. The matrix of cofactors (or cofactor matrix) is made of the cofactors of its
elements taking the same positions as the positions of these elements. The transposed
matrix of cofactors may be called the adjugate or adjoint matrix although the terminology
may differ between the authors.
Footnote 21. This assertion, in fact, applies to the common cases of tensor applications;
however, there are instances, for example in the differential geometry of curves and
surfaces, of tensors which are not uniformly dimensioned because the tensor is related to
two spaces with different dimensions.
Footnote 22. As indicated previously, the symbol “ ≡ ” may be used for identity as well as
for definition.
Footnote 23. In the literature of tensor calculus, rank and order of tensors are generally
used interchangeably; however some authors differentiate between the two as they assign
order to the total number of indices, including contracted indices, while they reserve rank
to the number of free indices. We think the latter is better and hence in the present book we
embrace this terminology.
Footnote 24. We adopt this assertion, which is common in the literature of tensor calculus,
as we think it is suitable for this level. However, there are many instances in the literature
of tensor calculus where indices are repeated more than twice in a single term. The bottom
line is that as long as the tensor expression makes sense and the intention is clear, such
repetitions should be allowed with no need in our view to take special precaution like
using parentheses. In particular, the forthcoming summation convention will not apply
automatically in such cases although summation on such indices, if needed, can be carried
out explicitly, by using the summation symbol ∑ or by a special declaration of such
intention similar to the summation convention. Anyway, in the present book we will not use
indices repeated more than twice in a single term.
Footnote 25. These precautions are obviously needed if the summation convention is
adopted in general but it does not apply in some exceptional cases where repeated indices
are needed in the notation with no intention of summation.
Footnote 26. This should not be confused with the order of tensor as defined above in the
same context as tensor rank. It should also be noticed that the order of indices in legitimate
tensor expressions and equalities is not required to be the same in all terms; the point of the
remark about the order of indices is to highlight the importance of this aspect of the indicial
structure so that it is observed and clarified so that the tensor becomes well defined and
correctly notated and used.
Footnote 27. We exaggerate the spacing here for clarity.
Footnote 28. In many places in this book (like other books) and for the convenience in
typesetting, the order of the indices is not clarified by spacing or inserting dots in the case
of mixed type tensors. This commonly occurs where the order of the indices is irrelevant in
the given context or the order is clear. Sometimes, the order of the indices may be indicated
implicitly by the alphabetical order of the selected indices.
Footnote 29. The focus of this section is on providing examples of tensors of different
ranks. As we will see later in this chapter (refer to § 2.6.2↑), there are true and pseudo
scalars, vectors and tensors, and hence some of the statements and examples given here
may qualify for certain restrictions and conditions.
Footnote 30. Similar basis vectors are assumed.
Footnote 31. The use of upper indices in the symbols of general coordinates is to indicate
the fact that the coordinates and their differentials transform contravariantly.
Footnote 32. In fact there are standard mathematical procedures to orthonormalize the
basis set if it is not and if this is needed.
Footnote 33. The terminology in this part, like many other parts, is not universal.
Footnote 34. Inner product (see § 3.5↑) is the result of a direct product (see § 3.3↑)
operation followed by a contraction (see § 3.4↑) and hence it is like a direct product in this
context.
Footnote 35. The Jacobian J is the determinant of the Jacobian matrix J of the
transformation between the unbarred and barred systems, that is:
(296)
For more details, the reader is advised to consult more advanced textbooks on this subject.
Footnote 36. Some of these labels are used differently by different authors as the
terminology of tensor calculus is not universally approved and hence the conventions of
each author should be checked. Also, there is an obvious overlap between this
classification (i.e. absolute and relative) and the previous classification (i.e. true and
pseudo) at least according to some conventions.
Footnote 37. This statement should be generalized by including w = 0 which corresponds
to absolute tensors and hence “relative” in this statement is more general than being
opposite to “absolute”. Accordingly, and from the perspective of relative tensors (i.e.
assuming that other qualifications such as matching in the indicial structure are met), two
absolute tensors can be added/subtracted but an absolute and a relative tensor (i.e. with w
≠ 0) cannot since they are “relative” tensors with different weights.
Footnote 38. For improper rotation, this is more general than being isotropic.
Footnote 39. Symmetry and anti-symmetry of tensors require in their definition two free
indices at least; hence a scalar with no index and a vector with a single index do not
qualify to be symmetric or anti-symmetric.
Footnote 40. In this context, like many other contexts in this book, there are certain
restrictions on the type and conditions of the coordinate transformations under which such
statements are valid. However, these details cannot be discussed here due to the
elementary level of this book.
Footnote 41. The best way to tackle this sort of exercises is to build a table or array of
appropriate dimensions where the indexed components in symbolic or numeric formats are
considered and a trial and error approach is used to investigate the possibility of creating
such a tensor.
Footnote 42. The concepts of repetitive and non-repetitive permutations may be useful in
tackling this question.
Footnote 43. Here, “type” refers to variance type (covariant/contravariant/mixed) and
true/pseudo type as well as other qualifications to which the tensors participating in an
addition or subtraction operation should match such as having the same weight if they are
relative tensors, as outlined previously (refer for example to § 2.6.3↑).
Footnote 44. Associativity and commutativity can include subtraction if the minus sign is
absorbed in the subtracted tensor; in which case the operation is converted to addition.
Footnote 45. As indicated before, there are cases of tensors which are not uniformly
dimensioned, and in some cases these tensors can be regarded as the result of an outer
product of lower rank tensors.
Footnote 46. Regarding the associativity of direct multiplication, there seems to be cases
in which this operation is not associative. The interested reader is advised to refer to the
research literature on this subject.
Footnote 47. The non-commutativity of the inner product in the case of only one of the
involved tensors is of rank > 1 my not be obvious; however, a simple example is
multiplying Ai and Bj kl with a contraction of j and k or j and l.
Footnote 48. In fact, this statement is rather vague and rudimentary and may not apply in
some cases. There are many details related to the issue of commutativity of inner product
of tensors which cannot be discussed here due to the level of this book. In general, several
issues should be considered in this regard such as the order of the indices in the outer
product of the two tensors involved in the inner product, the (m, n) type of the outer
product and whether the contracted indices are contributed by the same tensor or the two
tensors involved in the product assuming that the first case is conventionally an inner
product operation. Another important issue to be considered is that the contracted indices
must, in general, be of opposite variance type. Many of these details can be worked out
rather easily by the vigilant reader from first principles if they cannot be obtained from the
textbooks of tensor calculus.
Footnote 49. It should be emphasized that we are using the symbolic notation of tensor
calculus, rather than the matrix notation, in writing Ab and A⋅b to represent, respectively,
the outer and inner products. In matrix notation, Ab is used to represent the product of a
matrix by a vector which is an inner product according to the terminology of tensor
calculus.
Footnote 50. In these statements, we assume that the contracted indices can be contributed
by the same tensor in the product as well as by the two tensors (i.e. one index from each
tensor); otherwise more details are required. We are also assuming a general coordinate
system where the contracted indices should be opposite in their variance type.
Footnote 51. This is also defined differently by some authors.
Footnote 52. This should not be confused with the quotient rule of differentiation.
Footnote 53. We assume, of course, that the rules of contraction of indices, such as being
of opposite variance type in the case of non-Cartesian coordinates, are satisfied in this
operation.
Footnote 54. This name is usually used for the rank-3 tensor. Also some authors
distinguish between the permutation tensor and the Levi-Civita tensor even for rank-3.
Moreover, some of the common labels and descriptions of ε are more specific to rank-3.
Footnote 55. “Conserved” means that the tensor keeps the values of its components
following a coordinate transformation.
Footnote 56. Here, being conserved under all transformations is stronger than being
isotropic as the former applies even under improper coordinate transformations while
isotropy is restricted to proper transformations.
Footnote 57. Considering the difference in weight, the difference in the variance type on
the two sides of the equations should not be considered as a violation to the rules of
indices as stated in § 2.3↑. Similarly, considering the difference in the variance type, the
difference in weight should not be considered as a violation to the rules of equating
relative tensors as stated in § 2.6.3↑.
Footnote 58. This identity, like many other identities in this chapter and in the book in
general, is valid even for general coordinate systems although we use Cartesian notation to
avoid unnecessary distraction at this level. The alert reader should be able to notate such
identities in their general forms.
Footnote 59. As explained previously, “orthonormal” means that the vectors in the set are
mutually orthogonal and each one of them is of unit length.
Footnote 60. This also applies to the zero entries of this tensor which correspond to the
permutations with repetitive indices.
Footnote 61. This is a product of the rank-n permutation tensor by itself entry-by-entry
with the application of the summation convention and hence it can be seen as a multi-
contraction inner product of the permutation tensor by itself.
Footnote 62. The role of these indices in indexing the rows and columns can be shifted.
This can be explained by the fact that the positions of the two epsilons can be exchanged,
since ordinary multiplication is commutative, and hence the role of the epsilons in
providing the indices for the rows and columns will be shifted. This can also be done by
taking the transposition of the array of the determinant, which does not change the value of
the determinant since det(A) = det(AT), with an exchange of the indices of the Kronecker
symbols since the Kronecker symbol is symmetric in its two indices.
Footnote 63. That is:
εil εkl = δikδll − δilδlk = 2δik − δilδlk = 2δik − δik = δik
Footnote 64. That is:
εijk εlmk = δilδjmδkk + δimδjkδkl + δikδjlδkm − δilδjkδkm − δimδjlδkk − δikδjmδkl
= 3δilδjm + δimδjl + δimδjl − δilδjm − 3δimδjl − δilδjm
= δilδjm − δimδjl
Footnote 65. In fact, the determinantal form given by Eq. 151↑ can also be considered as a
mnemonic device where the first and second indices of the first ε index the rows while the
first and second indices of the second ε index the columns, as given above. However, the
mnemonic device of Eq. 153↑ is more economic in terms of the required work and more
convenient in writing. It should also be remarked that the determinantal form in all the
above equations is in fact a mnemonic device for these equations where the expanded form,
if needed, can be easily obtained from the determinant which can be easily built following
the simple pattern of indices, as explained above.
Footnote 66. That is:
εijk εljk = δilδjj − δijδjl = 3δil − δil = 2δil
Footnote 67. In fact this can be obtained from the determinantal form of δijklmn by
relabeling n with k.
Footnote 68. Matrix multiplication in matrix algebra is equivalent to inner product in
tensor algebra.
Footnote 69. The direction is also invariant but it is not a scalar! In fact the magnitude
alone is invariant under coordinate transformations even for pseudo vectors because it is a
true scalar.
Footnote 70. There is another reason that is the components given in the cylindrical and
spherical coordinates are physical, not covariant or contravariant, and hence suffixing with
coordinates looks more appropriate. The interested reader should consult on this issue
more advanced textbooks of tensor calculus.
Footnote 71. It may also be argued more simply that the divergence of a vector is a scalar
and hence it is invariant.
Footnote 72. It should be obvious that since ρ, φ and z are labels for specific coordinates
and not variable indices, the summation convention does not apply.
Footnote 73. Again, the summation convention does not apply to r, θ and φ.