Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 47

CHAPTER ONE

Fundamental Concept of Tensors


A tensor is a kind of mathematical construct or entity that takes numbers
(“scalars”) and vectors and generalize them to phenomena involving multiple
directions all at once. In fact, scalars (involving no direction) and vectors
(involving one direction) represent the most elementary kinds of tensors. Just like
numbers and vectors, tensors can represent many kinds of different things.
To some extent tensors can be thought of in terms of certain arrays or matrices of
numbers. The elements of tensor arrays are called the “components” or
“coordinates” of the tensors. In three dimensions a vector, or first rank tensor, can
be thought of as a kind of 3× 1 array of numbers. Likewise, a second rank tensor
can be thought of as a kind of 3× 3 array, a 3rd-rank tensor a 3× 3× 3 array.
There are equivalent approaches to visualizing and working with tensor :that the
content is actually the same may only become apparent with some familiarity with
the materials.

The classical approach :


The classical approach defines a tensor to a collection of multidimensional
arrays such that one array is associated to each possible coordinate system of any
fixed vector space. This notion generalizes scalars , vectors ,matrices etc.
To represent a vector x as a tensor one can simply let the array associated to any
basis B be the vector of coordinates of x with respect to B.
However to count as a tensor the arrays need to obey a relation that precisely
corresponds to how vectors ,matrices, linear functionals etc transforms when one
pass from one coordinate system to another.

The modern approach :


The modern approach views tensors initially as abstract objects expressing some
definite type of multilinear concept. Their well known properties can be
derived from their definitions as linear maps and the rules for the manipulations
of tensors arises as an extension of linear algebra to multilinear algebra. This
treatment has attempted to replace the component based treatment for advance
study , in the way that the more component free treatment of vectors replaces the
traditional component based treatment after the component based treatment has
been used to provide an elementary motivation for the concept of a vector. Never
the less a component free treatment has not been become fully popular ,owing to
difficulties involved with giving a geometrical interpretation to higher rank
tensors.

1
1.1 Fields, Directionality, and Tensors
A concept central to working with tensors is the idea of the field. A field is a
function that is defined at each position throughout some region of volume or
space. Most fields (at least all of those that we shall encounter) are well-behaved
in the sense that they are continuous and differentiable. As an example of a field it
is easy, for instance, to think about the temperature of an object as being a
function of position inside that object. The temperature can be uniform, in which
case the function is a constant, or the temperature can be non-uniform. It is easy
to see that the mass density of an object is, in general, a function of position within
the object.

The mass density and temperature inside objects are examples of quantities that
we can describe using the concept of a scalar field. For a field to be scalar, it
should be possible to express the value of the function at each point in space using
a single number. Using T to represent the temperature and ρ to represent the
density, then using rectangular Cartesian coordinates to represent location, we may
write that T and ρ depend on position according to

T = T ( x, y , z ) and ρ = ρ( x, y, z ) .
(1.1.1)

When treating the temperature or density as mathematical objects it is


commonplace to recognize T and ρ depend on x, y, and z, without writing this
down. Thus, we write T and ρ without explicitly stating that they depend on
position. It is understood that they depend on position without this having to be
written down.

In addition to scalar fields there are also such things as vector fields. Now, you
should already know that to uniquely specify a vector, it is necessary to specify
both a direction and a length. Alternatively, in 3 dimensions, a vector has 3
coordinates or components. An example of a vector field is the velocity field
describing liquid flow through a pipe. At any instant in time, each point in the
fluid is moving in some direction at such-and-such speed. The electric field is
another example of a vector field. In Fick's first law, concentration gradient and

diffusion flux are vector fields. We write electric

field as E and concentration
gradient and diffusion flux as ∇ C and J , respectively. As vectors, each of
these has three components (in 3-D space), and to recognize this, we write
  
E =(E x , E y , E z ) , ∇ C =(∂C / ∂x, ∂C / ∂y , ∂C / ∂z ) , and J =( J x , J y , J z ) .

(1.1.2)

2
  
It is to be understood that each coordinate (or component) of E , ∇ c, and J is a
function of position. Thus, for instance,

and E z = E z ( x, y, z ) .
E x = E x ( x, y , z ) , E y = E y ( x, y , z ) , (1.1.3)
It turns out that scalar and vector fields are two examples of what are known as
tensor fields. A scalar field is a tensor of rank 0, and requires one component,
which is a function of position, in order for the scalar to be uniquely specified.
Temperature and density are examples. Scalar fields have magnitude but are
without direction. Thus, it doesn't make sense to say that the temperature of an
object is 0°C pointing in the z-direction or that the density is 3 gm/cm3 in the x-
direction. The terminology "rank zero" or "0 rank" has to do with the fact that no
direction is involved when this kind of mathematical object is being used.
Vectors, on the other hand, are examples of 1st rank tensors. They have both
magnitude and direction, and are specified in 3-D space using 3 coordinates or
components. Beyond 0-th and 1st rank tensors there are 2nd-rank, 3rd-rank, 4th-
rank tensors, and so on. We won’t go beyond 4-th rank tensors. For a 2nd-rank
tensor field, each point in space is associated with 2 directions. The tensor field is
described by 9 components, each of which is a function of position. 3rd rank
tensors are associated with 3 directions and 27 components, and 4th-rank tensors
are associated with 4 directions and 81 components. The number of components
of a n-th rank tensor in 3-D space is 3 n .

1.2 Coordinate Transformations


The coordinates of points in the manifold may be assigned in a number of different
ways. If we select two different sets of coordinates,   x 1, x2 .....x n and
1 2 n
  x ′ , x ′ , ..... x ′ , there will obviously be a connection between them of the form

r r 1 2 n
  x ′ = f (x , x ....x ) r = 1, 2........n. (1.2.1)

where the f's are assumed here to be well behaved functions. Another way of
expressing the same relationship is
r r 1 2 n
  x ′ = x ′ (x , x ....x ) r = 1, 2........n. (1.2.2)

r 1 2 n
where   x ′r (x1 ,x 2 ....xn ) denotes the n functions   f (x ,x ....x ) , r = 1, 2......n.

Recall that if a variable z is a function of two variables x and y, i.e. z = f (x, y),
then the connection between the differentials dx, dy and dz is

3
∂f ∂f
dz = dx + dy . (1.2.3)
   ∂ x ∂ y

Extending this to several variables therefore, for each one of the new coordinates
we have

. r=1, 2........n. (1.2.4)

The transformation of the differentials of the coordinates is therefore linear and


homogeneous, which is not necessarily the case for the transformation of the
coordinates themselves.

1.3 Range and Summation Conventions.


Equations such as (1.2.4) may be simplified by the use of two conventions [2]:

Range Convention: When a suffix is unrepeated in a term, it is understood to take


all values in the range 1, 2, 3.....n.

Summation Convention: When a suffix is repeated in a term, summation with


respect to that suffix is understood, the range of summation being 1, 2, 3.....n.

With these two conventions applying, equation (1.2.4) may be written as

∂x ′r s
dx ′r = s dx . (1.3.1)
   ∂ x
Note that a repeated suffix is a "dummy" suffix, and can be replaced by any
convenient alternative. For example, equation (9) could have been written as

∂ x ′ r
dx ′r = m dxm . (1.3.2)
   ∂x
where the summation with respect to s has been replaced by the summation with
respect to m.

1.4 Contravariant Vectors and Tensors:


Consider two neighboring points P and rQ in the manifold whose coordinates are xr
and xr + dxr respectively. The vector   P Q is then described by the quantities dxr

4
which are the components r of the vector in this coordinate system. In the dashed
coordinates, the vector   P Q is described by the components   dx ′r which are related
to dxr by equation (1.3.1), the differential coefficients being evaluated at P. The
infinitesimal displacement represented by dxr or   dx ′r is an example of a
contravariant vector [2] .
Definition: A set of n quantities T r associated with a point P are said to be the
components of a contravariant vector if they transform, on change of coordinates,
according to the equation

∂ x ′ r s
T ′ r = s T . (1.4.1)
   ∂ x
where the partial derivatives are evaluated at the point P. (Note that there is no
requirement that the components of a contravariant tensor should be infinitesimal.)
A set of n 2 quantities T rs associated with a point P are said to be the components
of a contravariant tensor of the second order if they transform, on change of
coordinates, according to the equation

rs ∂x ′r ∂ x ′ s mn
T ′ = m n T . (1.4.2)
   ∂ x ∂x

Obviously the definition can be extended to tensors of higher order. A


contravariant vector is the same as a contravariant tensor of first order.
A contravariant tensor of zero order transforms, on change of coordinates,
according to the equation

  T ′ = T , (1.4.3)

i.e. it is an invariant whose value is independent of the coordinate system used.

1.5 Covariant Vectors and Tensors.


Let φ be an invariant function of the coordinates, i.e. its value may depend on
position P in the manifold but is independent of the coordinate system used. Then
the partial derivatives of φ transform according to

5
∂φ ∂φ ∂ xs
= (1.5.1)
r ∂ xs ∂ x ′ r
  ∂ x ′
Here the transformation is similar to equation (1.2.1) except that the partial
derivative involving the two sets of coordinates is the other way up. The partial
derivatives of an invariant function provide an example of the components of a
covariant vector [2] .
Definition :.A set of n quantities   T r associated with a point P are said to be the
components of a covariant vector if they transform, on change of coordinates,
according to the equation

∂x s
T ′r = T . (1.5.2)
   ∂ x ′ r s
By convention, suffices indicating contravariant character are placed as
superscripts, and those indicating covariant character as subscripts. Hence the
reason for writing the coordinates as xr. (Note however that it is only the
differentials of the coordinates, not the coordinates themselves, that always have
tensor character. The latter may be tensors, but this is not always the case.)
Extending the definition as before, a covariant tensor of the second order is
defined by the transformation
∂ xm ∂x n
T ′rs = r ∂x ′ s T mn (1.5.3)
   ∂ ′

and similarly for higher orders.

1.6 Mixed Tensors


These are tensors with at least one covariant suffix and one contravariant suffix.
r
An example is the third order tensor   T st which transforms according to

p
r ∂x ′r ∂x n ∂x m
T ′st = m Tnp (1.6.1)
s t
   ∂ x ∂x ′ ∂ x ′

Another example is the Kronecker delta defined by

r
  δ s = 1, r = s
  = 0, r ≠ s (1.6.2)

6
mn.. t
It is a tensor of the type indicated because (a) in an expression such as   Bpq.. δ m ,
which involves summation with respect to m, there is only one non-zero
mn.. t tn..
contribution from the Kronecker delta, that for which m = t, and so Bpq.. δ m = B pq..
  
; (b) the coordinates in any coordinate system are necessarily independent of each
∂ xr r ∂ x ′ r
other, so that s = δ s and = δ s′r ; so these two properties taken together
s
  ∂ x   ∂ x ′
imply that

∂ x ′ r ∂ xn m
δ s′r = m s δn . (1.6.3)
   ∂ x ∂ ′

1.7 Symmetric and Anti-symmetric Tensors


rs is a symmetric contravariant tensor if Ars = Asr and anti-symmetric if
  A   
rs sr
  A = − A . Similarly for covariant tensors. Symmetry properties are conserved
under transformation of coordinates, e.g. if   Ars = Asr , then

∂x ′m ∂ x ′ n rs ∂ x ′ m ∂x ′n sr


A ′mn = r s A = r s A = A ′nm . (1.7.1)
   ∂ x ∂x ∂x ∂ x

s r
Note however that for a mixed tensor, a relation such as   Ar = As does not
transform to give the equivalent relation in the dashed coordinates. The concept of
symmetry (with respect to a pair of suffices which are either both subscripts or
both superscripts) can obviously be extended to tensors of higher order.

Any covariant or contravariant tensor of second order may be expressed as the


sum of a symmetric tensor and an anti-symmetric tensor, e.g.

1 1
Ars = (Ars + Asr ) + (Ars − A sr ) . (1.7.2)
   2 2

1.8 Tests for Tensor character.


The direct way of testing whether a set of quantities form the components of a
tensor is to see whether they obey the appropriate tensor transformation law when
the coordinates are changed. There is also an indirect method however, two
examples of which will now be given [2] :

7
Theorem 1. Let   X r be the components of an arbitrary contravariant vector. Let
r
  Ar be another set of quantities. If   ArX is an invariant, then   Ar form the
components of a covariant vector .
Proof: Since   X r is a tensor, it obeys the tensor transformation law. Invariance of
r
  ArX means that
∂ x ′ s r
ArX r = A ′sX ′ s = A s′ X
   ∂ xr
∂x ′ s r
and so (Ar − A s′ )X = 0 .
   ∂ xr

Hence, since   X r is an arbitrary tensor,

∂x ′s
Ar = A s′ .
   ∂ xr

As an extension of this theorem, it is easy to show that any set of functions of the
coordinates, whose inner product with an arbitrary covariant or contravariant
vector is a tensor, are themselves the components of a tensor. For example, if
rs r rs
  A Xs is a tensor   B , then   A is a second order contravariant tensor.

r s
Theorem 2. If   arsX X is invariant,   X r being an arbitrary contravariant vector
and   ars being symmetric in all coordinate systems, then   ars are the components of
a covariant tensor of second order.
r s
Proof: From our assumption about the invariance of   arsX X ,

m n
′ X ′r X ′ s
  amnX X = a rs
∂x ′r ∂x ′s m n
= a rs
′ X X
   ∂ x m ∂x n

∂x ′r ∂x ′ s m n
Hence bmn X mXn ≡ (a mn − a ′rs m )X X = 0 .
   ∂x ∂x n

Since   X m is arbitrary and the total coefficient of   X mX n is   bmn + bnm , we deduce
that   bmn + bnm = 0 , i.e.

8
∂ x ′ r ∂ x ′ s ∂x ′r ∂ x ′ s
amn + anm = a rs
′ + ′

rs
   ∂ xm ∂ xn ∂x n ∂x m
∂x ′r ∂x ′s
= (a rs
′ + a ′sr ) m
   ∂x ∂x n
on interchanging the summation variables r and s in the second term. But
  amn = anm in all coordinate systems, hence

∂ x ′r ∂ x ′ s
amn = a rs
′ .
   ∂x m ∂ xn

1.9 The Metric Tensor


Consider first the familiar Euclidean space in three dimensions, i.e. a space in
which one can define Cartesian coordinates x, y and z so that the distance   dl
between two neighboring points   x, y, z and  x + dx , y + dy , z + dz is given by

2 2 2 2
  dl = (dx) + (dy) + (dz) . (1.9.1)
1 2 3
If we choose any other coordinates   x , x ,x to identify points in this space, the
original coordinates will be functions of these new coordinates, and their
differentials will be linear combinations of the differentials of the new coordinates.
Thus in terms of the latter coordinates,
2 m n
  dl = amn dx dx (1.9.2)
where the   amn will be functions of   x m . (For example in spherical polar
1 2 3 2 2 2
coordinates   x = r, x = θ , x = φ we have   a11 = 1, a 22 = r , a 33 = r sin θ and all
other a's are zero.)
We now show that   amn is a covariant tensor of second order. The proof goes as
follows:
(a)   amn may be taken to be symmetric since each   apq occurs only in the
combination   apq + a qp on the RHS of (1.9.2).
2 m n
(b)   dl = amn dx dx is invariant, since the distance between two points does not
depend on the coordinates used to evaluate it.
(c) By keeping one point fixed and letting the second point vary in the
neighborhood of the first,   dx r may be considered an arbitrary contravariant tensor.
Hence, using the theorem above,   amn is a covariant tensor of second order. It is
called the metric tensor for the Euclidean 3-space. A similar tensor obviously
exists in the case of a two dimensional Euclidean space.

9
1.10 The Conjugate Metric Tensor.
From the covariant metric tensor   gmn we can construct a contravariant tensor
mn
  g defined by

g mn gnp = δ pm .
  
mn
To show that   g is a tensor, we note that, for any contravariant vector   V p ,
g mn gnpV p = δ pmV p = V m . This means that the inner product of   g mn with the
  
p mn
arbitrary covariant vector   gnpV is a tensor,   Vm , and so we deduce that   g is
indeed a tensor of the type indicated. It is said to be conjugate to   gmn . It is easily
shown that when the metric tensor is diagonal, i.e. when   gmn = 0, m ≠ n , the
conjugate tensor is also diagonal, with each diagonal element satisfying
nn
  g = 1/ gnn .
The following theorem can be proved, but will just be quoted here: if g is the
determinant of the matrix   gmn (i.e. choosing to write the components of the tensor
  gmn in the form of a matrix array), then
∂ ∂
g mn r gmn = r ln g .
   ∂x ∂x
1.11 Riemannian space.
A manifold is said to be Riemannian if there exists within it a covariant tensor of
the second order which is symmetric. This tensor is called the metric tensor and
normally denoted by   gmn . Its significance is that it can be used to define the
analogue of "distance" between points, and the lengths of vectors. We will assume
that all manifolds that we will be dealing with from now on are Riemannian.
Definition:. The interval ds between the neighboring points   x r and   x r + dxr is
given by

2 m n
  ds = gmn dx dx .

This is of course invariant. In the familiar Euclidean space where   gmn is just the
2 2
  amn above,   ds = dl ≥ 0 , being zero only when the two points coincide. In other
cases however, e.g. in space time in relativity theory,   ds2 may take on negative
values, so that   ds itself is not necessarily real. If ds = 0 for   dx r not all zero, the
displacement   dx r is called a null displacement. Note that there is no requirement
that ds should necessarily have the physical dimensions of length.

10
1.12 Raising and Lowering Suffices.
m
Given a tensor   T rs , we may form another tensor   T mrs defined by
T nrs = gnmT mrs (1.12.1)
  

Note that g mnTnrs = g mn gntT trs = δtmT trs = T mrs . (1.12.2)


  

The tensor   T nrs may therefore be regarded as possessing a special relationship
m
with the original tensor   T rs in that either of them may be found from the other by
the operation of forming the inner product of the first with the metric tensor or its
conjugate. For this reason, the same symbol is used (T in this instance), and we
describe the above processes by saying that in (1.12.1)we have "lowered the suffix
m", and that in (1.12.2) we have "raised the suffix n". The process of raising or
lowering suffices can be extended to cover all the indices of a tensor. For example
we can raise one or both of the suffices in the tensor   T mn , generating the
m m mn
corresponding tensors   T n ,   T n and   T . Notice the distinction between the
two forms of the mixed tensor, effected by leaving appropriate gaps in the set of
indices. When the tensor is symmetric however this distinction disappears and we
m
simply write either of these as   T n .

1.13 Tensors as Transformations (mapping):


A linear transformation from rank one tensor space V into itself defines a rank two
tensor.
Mapping :Tb=c
Linear mapping:
T(a+b)=Ta+Tb
T( α a)= α Ta
T (α a + β b) = α Ta + β Tb ∀ a, b ∈V α , β ∈ R
Problem:(1) Ta=3a then is T Tensor? Yes.
T(a+b)=3(a+b)=3a+3b
Ta+Tb=3a+3b
Thus T(a+b)=Ta+Tb
T (α a ) = 3(α a ) = 3α a
α Ta= α (3a)=3 α a
thus T( α a)= α Ta

11
Hence it is tensor.

(2) if Ta=3a+2 e1 ,then is T tensor? No.


T(a+b)=3(a+b)+2 e1 =3a+3b+2 e1
Ta+Tb=(3a+2 e1 )+(3b+2 e1 )
So,T(a+b) ≠ Ta+Tb
Thus T is not a tensor.

Specific Statements on Tensors [7]:


1. All scalar are not tensor ,although all tensors of rank 0 is scalar.
2. All vectors are not tensor but all tensor of rank 1 are vectors.
3. All matrices are not tensors although all tensor of rank 2 are matrices.
4. Tensors can be multiplied by another tensors to form new tensor
5. The product of a tensor and a scalar(tensor of rank 0) is commutative.
6. Tensor multiplication in general is not commutative.
7. The rank of a new tensor formed by the product of two other tensors is the sum
of their individual ranks.
8. A tensor of rank n in three dimensional space has 3n components.

12
CHAPTER TWO

THE CHRISTOFEL SYMBOLS AND DERIVATIVE OF TENSORS

2.1 Christoffel Symbols


The Christoffel Symbols [11] are tensor like objects derived from a Riemann
metric g. They are used to study the geometry of the metric and appear for
example in the geodesic equation .there are two closely related kinds of
christoffel ,the first kind Γijk and the second kind Γ i , j .Christoffel Symbols of
k

second kind are also known as affine connections or as connection coefficients.


Christoffel Symbols of first kind are variously denoted [ i j, k],  i j
 Γ abc or
k
,
{ ab, c} .
The Christoffel Symbols of first kind defined by
1 ∂g jk ∂g ik ∂gij
[ i j, k]= [k, i j]= ( + j − k ) (2.1.1)
2 ∂xi ∂x ∂x
Christoffel Symbols of second kind is defined by

{ik j } = g kl [ij , l ] (2.1.2)

2.2 Transformations of Christoffel Symbols [1]:


We now find the relation between the Christoffel Symbols of the second kind in
two coordinate systems. As the fundamental tensor gij being covariant, transform
according to the equation

∂x i ∂x j
g p,q = gij p q (i, j , p, q = 1, 2,3) (2.2.1)
∂x ∂x

And obtain by differentiating with respect to x r ,

∂g pq ∂gij ∂x i ∂x j ∂x k  ∂xi ∂ 2 x j ∂x j ∂ 2 xi 
= + g ij  p q r + q p r  (2.2.2)
∂x r ∂x k ∂x p ∂x q ∂x r  ∂x ∂x ∂x ∂x ∂x ∂x 

By suitable changes of free and dummy indices in the above equation, we have the
following equation

∂g rq ∂g kj ∂x i ∂x j ∂x k  ∂xi ∂ 2 x j ∂x j ∂ 2 xi 
= + g ij  +  (2.2.3)
∂x p ∂x i ∂x p ∂x q ∂x r  ∂x ∂x ∂x
r q p
∂x q ∂x p ∂x r 

13
∂g pr ∂g ik ∂x i ∂x j ∂x k  ∂xi ∂ 2 x j ∂x j ∂ 2 xi 
and = + g ij  +  (2.2.4)
∂x q ∂x j ∂x p ∂x q ∂x r  ∂x ∂x ∂x
p q r
∂x r ∂x p ∂x q 

Now substituting (2.2.2) from the sum of the above equations and dividing by 2, in
consequence of (2.1.1) we have
∂x i ∂x j ∂x k ∂xi ∂ 2 x j
 pq, r  = [ ij , k ] p q r + g ij r (2.2.5)
  ∂x ∂x ∂x ∂x ∂x p ∂x q

where,  pq, r  is formed with respect to tensor g ij .

Now the transformation law of the contravariant fundamental tensor is

∂x h hl ∂x
r
g rs
=g (2.2.6)
∂x s ∂x l

If we multiply the left and right hand members of equation (2.2.5) by left and right
hand members of (2.2.6) respectively and sum with respect to r, in consequence of
(2.1.1) we have

 s  ∂x h hl ∂x ∂x
i i
∂2 x j
  s [ = ] δ + δ
k hl i
ij , k g l g ij g l
 pq  ∂x ∂x p ∂x q ∂x p ∂x q

∂2 xh  h  ∂xi ∂x j  s  ∂xh
which reduces to , +   =  (2.2.7)
∂x p ∂x q ij  ∂x p ∂x q  pq  ∂x s

Hence, Equations (2.2.5) and (2.2.7) are known as the transformation laws of the
h 
Christoffel Symbols and clearly neither [ ij , k ] nor   are components of a tensor.
ij 

2.3 Covariant differentiation:


In mathematics the covariant derivatives is a way of specifying a derivative along
tangent vectors of a manifold. Alternatively the covariant derivative is away of
introducing and working with a connection of a manifold by means of differential
operator, to be contrasted with the approach via a connection form.
The covariant derivative is generalization of the directional derivative from vector
calculus. Covariant derivative obey the following laws:

14
1.The covariant derivative of the sum of two tensors is the sum of their covariant
derivatives.
2.The covariant derivative of an outer(or inner ) or inner product of two tensors is
equal to the sum of two terms obtained by outer (or inner multiplication) of each
tensor with the covariant derivative of the other tensor.
ij i
3. The tensors gij , g , g j are constants with respect to covariant differentiation.

Covariant Derivatives:
∂xα
If we write the tensor equation Bi = Aα and take the partial derivative using
∂y i

the operator then we get[4],
∂y j
∂Bi ∂xα ∂x β ∂Aα ∂ 2 xα
= + Aα (2.3.1)
∂y j ∂y j ∂y j ∂x β ∂y j dyi
∂ 2 xα
to evaluate the term j i in above equation we write directly
∂y dy

{ }
α
∂ 2 xα  γ  ∂x α ∂xγ ∂xβ
= i j  γ
− γ β j .thus we write the result
∂y j dy i   y ∂y x ∂y ∂y
i

∂Bi ∂xα ∂x β ∂Aα  γ  ∂xα


= + i
 j
∂y j ∂y j ∂y j ∂x β   y ∂yγ
Aα − γ
α
β { } x
∂xγ ∂xβ
∂yi ∂y j

α
∂x
now substituting the value Bγ = Aα in above equation yields the result:
∂y γ
∂Bi  γ  ∂Aα  γ  ∂xα ∂x β
− i j
  γB = [ − α β  γA ] (2.3.2)
∂y j   y ∂x β  x ∂yi ∂y j
The term in the brackets of the equation 2.4.9 is called the covariant derivative of
a covariant tensor, rank 1 . it is a covariant tensor rank 2. similarly the covariant
derivative of a contravariant tensor ,rank 1 can be written:
A,i j =
∂Ai
∂x j
+{ }
α
i
j Aα
i
the covariant derivative of a mixed tensor , A j , rank 2 can be expressed in tensor

notation by the equation : A


i
j ,k
=
∂Aij
∂x k { } { }
− jk
α
i
i
α
Aα + α k Aj (2.3.3)

equation (2.3.3) is a mixed tensor ,rank 3 .it is covariant rank 2 and contravariant
rank 1 .

15
2.4 Intrinsic Derivative
The intrinsic or absolute derivative of a covariant vector Ai taken along a curve
x i = xi ( t ) , i = 1, 2,...N is defined as the inner product of the covariant derivative
with the tangent vector to the curve. The intrinsic derivative is represented [6] ,
δ Ai ∂x j
= Ai , j
δt ∂t
δ Ai  ∂Ai
{}
=  j − Aα i j 
δ t  ∂x
α
 ∂x j
 ∂t
δ Ai dAi
δt
=
dt {} α
− Aα i j
dx j
dt
similarly the intrinsic or absolute derivative of contravariant tensor Ai is given by:
δ Ai
δt
= Ai, j
dx j dAi
dt
=
dt { }
i
+ j k Ak
dx j
dt
the intrinsic or absolute derivative is used to differentiate sums and products in the
sane manner as used in ordinary differentiation as used. also if the coordinate
system is Cartesian the intrinsic derivative becomes an ordinary derivative.
The intrinsic derivative of higher order tensors is similarly defined as an inner
product of the covariant derivative with the tangent vector to the given curve. For
example
ij p
δ A klm = ij dx
δt A klm, p dt
ij
is the intrinsic derivative of fifth order mixed tensor A klm .

2.5 Riemann Christoffel Tensor


If we differentiate equation (2.2.7) with respect to x r and subtract the resulting
equation from the one we obtained by interchanging the indices q and r, we obtain

 ∂ h ∂  h   ∂x i ∂x j ∂x k  h   ∂ 2 xi ∂x j ∂ 2 xi ∂x j 
 j  − k   p q r   p q r + − 
 ∂x ik  ∂x ij   ∂x ∂x ∂x ij  ∂x ∂x ∂x ∂x p ∂x r ∂xq 

 ∂  s  ∂  s   ∂x h  s  ∂ 2 x h  s  ∂ 2 xh
= q  − r  +
  s   s q   s r −
 ∂x  pq  ∂x  pq   ∂x  pq  ∂x ∂x  pq  ∂x ∂x

16
On substituting for the second derivatives in this equation their expressions from
the equations of the form (2.2.7) we obtain

∂xi ∂x j ∂x k ∂xk
Rijkh = R s
pqr (2.5.1)
∂x p ∂x q ∂x r ∂x s

where,
∂  h  ∂  h   l  h   l  h 
Rijkh =  −   +    −    [l is a dummy index] (2.5.2)
∂x j ik  ∂x k ij  ik lj  ij lk 

s
and R pqr is the similar expression in the Christoffel Symbols formed with respect
g ∂x t
to ij . If the equation (2.5.1) be multiplied by and summed with respect to h,
∂x h
we have

∂xi ∂x j ∂x k ∂x t
t
R pqr = Rijkh (2.5.3)
∂x p ∂x q ∂x r ∂xh
h
From the form of these equation it follows that Rijk , which are the components of
a contravariant tensor of the rank one and covariant tensor of the rank three. It is
Known as Riemann Christoffel tensor or Riemann tensor or Riemann symbols of
second kind. From (2.5.2), it is clear that Riemann tensor is skew symmetric in the
index j and k.

The quantities Rlijk are called Riemann symbols of the first kind, which are defined
by –

Rlijk = glh Rijkh and Rijkh = g lh Rlijk (2.5.4)

Covariant form of Rlijk also can be expressed in either of the following form:

1  ∂ 2 g kl ∂ 2 gik ∂ 2 g jl ∂ 2 gij 
 + g ( [ ij , β ] [ kl , α ] − [ ik , β ] [ jl , α ] ) Or
αβ
Rlijk =  i j− j l − i k − k l
2  ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x 
∂ ∂ s s
Rlijk = j [ ik , l ] − k [ ij , l ] + [ kl , s ]   − [ jl , s ]   [s is dummy index]
∂x ∂x ij  ik 

17
From this form we find that , Rlijk is skew-symmetric in the first two indices and
the last two indices as well as being symmetric in the interchange of the first pair
and the last pairs of indices and consequently

Rlijk = − Riljk ; Rlijk = − Rlikj ; Rlijk = Rjkli

2.6 Ricci Tensor


Ricci tensor or Ricci curvature tensor denoted by Rij , is defined by

Rij = Rijhh = g hk Rhijk (2.6.1)

h
where, Rijh is the Riemann Christoffel tensor or Riemann symbols of second kind
and Rhijk is the Riemann symbols of the first kind.

Applying contraction of h and k in (2.6.1),

∂ log g  i 
A little manipulation with Christoffel Symbols leads to =   using the
∂x k ik 
relation , from the definition of Riemann Christoffel tensor [equation(2.5.2)] we
have –

∂ h ∂  h   l   h   l  h 
h
Rijh =  − k   +     −   
∂x j ih  ∂x ij  ih  lj  ij lh 
∂  ∂ log g  ∂  h   l   h   l  ∂ log g
=  −   +    −  
∂x j  ∂xi  ∂xh ij  ih  lj  ij  ∂x
l

∂ 2 log g ∂  h   l   h   l  ∂ log g
⇒ Rij = R =h
ijh − h   +    −   (2.6.2)
∂x ∂x
i j
∂x ij  ih  lj  ij  ∂x
l

Where g is defined as cofactor of gij in the determinant g ≡ g ij . Again from the


above equation it is clear that Rij is symmetric.

18
2.7 Bianchi’s Identity
i 
Consider any space as a general coordinate system xi and   denote the values
 jk 0
of the Christoffel Symbols for this coordinate system at the point x0i , and
coordinates x i are defined by,
1 i 
x i = x0i + x i −   x j x k (2.7.1)
2  jk 0

Then the point x i = 0 is the point x0i and at this point

∂xi ∂ 2 xi  i 
= δ pi ; = −  
∂x p
∂x p ∂x q  pq 0

 i 
Substituting these values in (2.2.7), we find that   = 0 at the point x i = 0 .
 jk 0
Now the covariant derivative of a tensor at origin, i.e., x i = 0 is the ordinary
derivative. In particular, we have –
∂R pqr
l
∂2 l  ∂2 l 
R l
pqr , s = =   −   [by (2.5.2)] (2.7.2)
∂x s
∂x q ∂x s  pr  ∂x ∂x
r s
 pr 

From these equations we have

, s + R prs , q + R psq , r = 0
l l l
R pqr

The left hand members of these equations are components of a tensor. Since they
are equal to zero at the point in one coordinate system, they are equal to zero at
this point in any coordinate system. Moreover, this result can be be obtained at
each point of the space, and consequently for any space and in any coordinate we
have
R ijkl , m + Rijl m , k + Rijmk , l = 0 (2.7.3)

Since the covariant derivatives of g ik are equal to zero, we have

R ijkl , m = ( g ih Rhjkl ) = g ih Rhjkl , m


,m

Since the determinant of g ik is not equal to zero, from this result and equation
(2.5.9) it follows (on changing indices),

19
Rijkl , m + Rijl m, k + Rijmk , l = 0 (2.7.4)
The equation (2.7.3) and (2.7.4) are called the Bianchi’s identity. They hold in any
space of any dimensionality and any coordinate system.

2.8 Einstein Tensor


The Einstein tensor [10] expresses space-time curvature in the Einstein field
equations for gravitation in the theory of general relativity. It is sometimes called
the trace-reversed Ricci tensor. The Einstein tensor denoted by G is a tensor of
rank two defined over Riemannian manifolds, which is defined as –

1
Gµv = Rµ v − gµ v R
2

where is the Gµv Ricci tensor, Rµ v is the metric tensor and R is the scalar
curvature. The Ricci tensor depends only on the metric tensor, so the Einstein
tensor can be defined directly with just the metric tensor. However, this expression
is complex and rarely quoted in textbooks. The complexity of this expression can
be shown using the formula for the Ricci tensor in terms of Christoffel Symbols:

1
Gαβ = Rαβ − gαβ
2

1
= Rαβ − gαβ g γζ Rγζ
2

 γ ζ 1 γζ 
=  δα δ β − gαβ g  Rγζ
 2 

 γ ζ 1 γζ 
2
(
=  δα δ β − gαβ g  Γγζ ,ε − Γγε ,ζ + Γεσ − Γζσ Γεγ
ε ε ε ε σ
)
 
α
where δ Γ βγ
α
β
is the Kronecker tensor and the Christoffel Symbols is defined as

1
Γαβγ = g αε ( g βε ,γ + gγε ,β − gβγ ,ε )
2

Before cancellations, this formula results in 2 × ( 6 + 6 + 9 + 9 ) = 60 individual


terms. Cancellations bring this number down somewhat. In the special case of a
locally inertial reference frame near a point, the first derivatives of the metric
tensor vanish and the component form of the Einstein tensor is considerably
simplified:

20
 1 
Gαβ = g γµ  gγ [β ,µ ]α + gα [µ ,β ]γ − gαβ gεσ ( gε [µ ,σ ]γ + gγ [σ ,µ ]ε )  =
 2 
 1 
g γµ  δ ε α δ βα − g εσ gαβ  ( gε [µ ,σ ]γ + gγ [σ ,µ ]ε ) ,
 2 

where square brackets conventionally denote antisymmetrization over bracketed


indices, i.e.

gα [ β ,γ ]ε =
1
(
g −g
2 αβ ,γε αγ ,βε ).

21
CHAPTER 3
APPLICATION OF TENSORS IN MATHEMATICS
3.1 Introduction to Differential Geometry
In this section we will examine some fundamental properties of curves and
surfaces. In particular, at each point of a space curve we can construct a moving
coordinate system consisting of a tangent vector, a normal vector and a binormal
vector which is perpendicular to both the tangent and normal vectors. The
curvature is a measure of how the tangent vector to the curve is changing and the
torsion is a measure of the twisting of the curve out of a plane.
In a similar fashion, associated with every smooth surface there are two coordinate
surface curves and a normal surface vector through each point on the surface. The
coordinate surface curves have tangent vectors which together with the normal
surface vectors create a set of basis vectors. These vectors can be used to define
such things as a two dimensional surface metric and a second order curvature
tensor. The coordinate curves have tangent vectors which together with the surface
normal form a coordinate system at each point of the surface. How these surface
vectors change brings into consideration two different curvatures. A normal
curvature and a tangential curvature (geodesic curvature). How these curvatures
are related to the curvature tensor and to the Riemann Christoffel tensor,
introduced in the last chapter, as well as other interesting relationships between the
various surface vectors and curvatures, is the subject area of differential geometry.

(a) Space Curves and Curvature


For x i = x i (s),i = 1, 2, 3, a 3-dimensional space curve in a Riemannian space Vn
g dx i
with metric tenso i j , and arc length parameter s, the vector T =
i
represents a
ds
tangent vector to the curve at a point P on the curve. The vector T i is a unit
vector because
dxi dx j
gijT iT j = g ij =1
ds ds
Differentiate intrinsically, with respect to arc length,
δT j δTi j
gijT i + gij T =0
δs δs
δT j
which implies that g ijT i =0
δs
δT i
Hence, the vector is perpendicular to the tangent vector Ti.
δs

22
Define the unit normal vector N i to the space curve to be in the same direction as
δT i
the vector and write
δs
1 δT i
Ni =
κ δs

where κ is a scale factor, called the curvature, and is selected such that
g ij N i N j = 1

δT i δT j
which implies that g ij =κ2
δs δs
The reciprocal of curvature is called the radius of curvature. The curvature
measures the rate of change of the tangent vector to the curve as the arc length
varies. By differentiating intrinsically, with respect to arc length s, the relation
gjjTiNj = 0, we find that
δN j δT i j
gijT i + g ij N =0
δs δs
Consequently, the curvature κ can be determined from the relation
δNj δ Ti j
gijTi
= − gij N = − gij κ Ni N j = − κ
δs δs
which defines the sign of the curvature.
In a similar fashion we differentiate the relation and find that
δN j
g ij N i =0
δs
δNj
This later equation indicates that the vector is perpendicular to the unit
δs
normal Ni. The equation indicates that Ti is also perpendicular to Ni and hence any
linear combination of these vectors will also be perpendicular to Ni. The unit
binormal vector is defined by selecting the linear combination
δNj
+ κT j
δs
and then scaling it into a unit vector by defining
1δN j 
Bj =  +κ T j 
τ  δs 
where τ is a scalar called the torsion.

23
The sign of τ is selected such that the vectors Ti , Ni and Bi form a right handed
system with ε ijk T N B = 1 and the magnitude of τ is selected such that Bi is a unit
i j k

vector satisfying
g ij B i B j = 1

The triad of vectors Ti, Ni, Bi at a point on the curve form three planes. The plane
containing Ti and Bi is called the rectifying plane. The plane containing Ni and Bi is
called the normal plane. The plane containing Ti and Ni is called the osculating
plane. The reciprocal of the torsion is called the radius of torsion. The torsion
measures the rate of change of the osculating plane. The vectors Ti, Ni and Bi form
a right-handed orthogonal system at a point on the space curve and satisfy the
relation
B i = ε ijk T j N k

By using the equation it can be shown that Bi is perpendicular to both the vectors
Ti and Ni since
gij B iT j = 0 = gij B i N j
It is left as an exercise to show that the binormal vector Bi satisfies the relation
δ Bi
= −τ N i
δs
The three relations
δT i 
= κNi 
δs 
δ Ni 
= τ Bi − κ T i 
δs 
δ Bi 
= −τ N i 
δs 
are known as the Frenet-Serret formulas of differential geometry.

Problem .
Find the number of independent components associated with the Riemann
Christoffel tensor Rijkm ; (i,j,k,m= 1,2,..., N). There are N4 components to examine
in an N-dimensional space. Many of these components are zero and many of the
nonzero components are related to one another by symmetries or the cyclic
properties. Verify the following cases:

24
CASE I: We examine components of the form Rinin , i ≠ n with no summation of i
or n. The first index can be chosen in N ways and therefore with i ≠ n the second
index can be chosen in (N-1) ways. Observe that Rinin = Rnini ,(no summation on i or
n) and so one half of the total combinations are repeated. This leaves
1
M 1 = N ( N − 1) components of the form Rinin. The quantity M1 can also be
2
thought of as the number of distinct pairs of indices (i,n).
CASE II: We next examine components of the form Rinji , i ≠ n ≠ j where there is
no summation on the index i. We have previously shown that the first pair of
indices can be chosen in M1 ways. Therefore, the third index can be selected in
1
(N-2) ways and consequently there are M 2 = N ( N − 1)( N − 2) distinct
2
components of the form Rinji with i ≠ n ≠ j .
CASE III: Next examine components of the form Rinjk where i ≠ n ≠ j ≠ k , From
CASE-I, the first pairs of indices (i, n) can be chosen in M1 ways. Taking into
account symmetries, it can be shown that the second pair of indices can be chosen
1 1
in ( N − 2)( N − 3) ways. This implies that there are N ( N − 1)( N − 2)( N − 3)
2 4
ways of choosing the indices i,n,j and k with i ≠ n ≠ j ≠ k . By symmetry the
pairs (i, n) and (j, k) can be interchanged and therefore only one half of these
combinations are distinct.
1
This leaves N ( N − 1)( N − 2)( N − 3)
8
distinct pairs of indices. Also from the cyclic relations we find that only two thirds
of the above components are distinct. This produces
1
M3 = N ( N − 1)( N − 2)( N − 3)
12
distinct components of the form Rinjk with i ≠ n ≠ j ≠ k
Adding the above components from each case we find there are distinct and
independent components
1 2 2
M 4 = M1 + M 2 + M 3 = N ( N − 1)
12

25
Using the above rule, we have the following results according to the value of N
i.e., dimensions of space as in the table-2 [6] :
Table 2: Number of Independent components of R ijkm for different values of N

Dimension of space N 1 2 3 4 5
Number of components N4 1 16 81 256 625
M4 = Independent components of Rijkm 0 1 6 20 50

Note 1: A one dimensional space can not be curved and all one dimensional
spaces are Euclidean.
i.e. if we have an element of arc length squared given by ds 2 = f(x)(dx) 2 ,
we can make the coordinate transformation f ( x)dx = du and reduce the arc
length squared to the form ds 2 = du 2 .
Note 2: In a two dimensional space, the indices can only take on the values 1 and
2. In this special case there are 16 possible components. It can be shown that the
only non vanishing components are:
R1212 = -R 1221 = -R 2112 = R 2121
For these non-vanishing components only one independent component exists. By
convention, the component -R1212 is selected as the single independent component
and all other nonzero components are expressed in terms of this component.
Hence we can Find all the non-vanishing independent components Rijki for
i,j,k,l=1, 2, 3, 4 and show that
R1212 R3434 R2142 R4124
R1313 R1231 R2342 R4314
R2323 R1421 R3213 R4234
R1414 R1341 R3243 R1324
R2424 R2132 R3143 R1432
can be selected as the twenty independent components.

3.2 Fundamental Forms of Surface


In a Riemannian space V N with metric tensor g i j and curvilinear coordinates x i ,
i = 1,2,3, the equations of a surface can be written in the parametric form
x i = x i (u 1 ,u 2 ), where u  ,  = 1 , 2 are called the curvilinear coordinates of the
surface.

26
Since
∂x i α
dx i = du
∂uα
then a small change du  on the surface results in change dx i in the space
coordinates. Hence an element of arc length on the surface can be represented in
terms of the curvilinear coordinates of the surface. This same element of arc length
can also be represented in terms of the curvilinear coordinates of the space. Thus,
an element of arc length squared in terms of the surface coordinates is represented
ds 2 = aαβ duα du β
where a  is the metric of the surface. This same element when viewed as a
spatial element is represented
ds 2 = gij dx i dx j
By equating the equations and we find that
dx i dx j α β
gij dx dx = gij α β du du = aαβ duα du β
i j

du du
The equation shows that the surface metric is related to the spatial metric [3] and
can be calculated from the relation
∂x i ∂x j
aαβ = gij α β .
∂u ∂u
This equation reduces to the equation to the equation of the surface metric
associated with the two dimensional surface, the special case of Cartesian
coordinates. In the surface coordinates we define the quadratic form A = a 
du  du  as the first fundamental form of the surface. The tangent vectors to the
∂x i
coordinate curves defining the surface are given by and can be viewed as
∂uα
either a covariant surface vector or a contravariant spatial vector. We define this
vector as
∂x i
xi
α ; i = 1,2,3,  = 1,2
∂uα
Any vector which is a linear combination of the tangent vectors to the coordinate
curves is called a surface vector. A surface vector A  can also be viewed as a
spatial vector A i . The relation between the spatial representation and surface
representation is Ai = Aα xαi .The surface representation A  ,  = 1 , 2 and the

27
spatial representation A 1 , i = 1 , 2 , 3 define the same direction and magnitude
since
g ij Ai A j = gij Aα xαi Aβ xβj = gij xαi xβj Aα Aβ = aαβ Aα Aβ
Consider any two surface vectors A a and B a and their spatial representations A 1
and B l where
Ai = Aα xαi and B i = Bα xαi
These vectors are tangent to the surface and so a unit normal vector to the surface
can be defined from the cross product relation
ni AB sin θ = ε ijk A j B k
where A, B are the magnitudes of Ai, Bi and θ is the angle between the vectors
when their origins are made to coincide. Substituting equations into the equation ,
we find
ni AB sin θ = ε ijk Aα xαj B β xβk
In terms of the surface metric, we have
AB sin θ = ε αβ Aα B β
so that equation can be written in the form
( nε
i αβ − ε ijk xαj xβk ) Aα B β = 0

which for arbitrary surface vectors implies


niε αβ = ε ijk xαj xβk
1
or ni = ε αβ ε ijk xαj xβk
2
The equation defines a unit normal vector to the surface in terms of the tangent
vectors to the coordinate curves. This unit normal vector is related to the covariant
derivative of the surface tangents as is now demonstrated.
We know that the differential equation of the geodesic curve on a surface is given
by the equation
d 2u α  α  du β duγ
+  = 0; α , β , γ = 1, 2
ds 2  βγ a ds ds
Using the tensor derivative of equation with respect to the surface coordinates,
produces

28
∂ 2 xi i  σ 
xαi , β = +   xαp xβq −   xσi
αβ a
α β
∂u ∂u  pq  g
where the subscripts on the Christoffel symbols refer to the metric from which
they are calculated.
Consider a mixed tensor of form
Tαi ......βj Ai ... Aj Bα ...B β

where Ai ,..., Aj and Bα ,..., B β are parallel vector fields along the given curve, say
C.
Now the derivative of the equation produces the result
g ij Aα xαi , γ xβj + gij xαi xβi , γ = aαβ , γ = 0

Interchanging the indices α , β , γ cyclically in the equation one can verify that
g ij xαi , β xγj = 0

The equation indicates that in terms of the space coordinates, the vector xαi , β is
i
perpendicular to the surface tangent vector xγ and so must have the same direction
as the unit surface normal ni. Therefore, there must exist a second order tensor bα β
such that
bαβ ni = xαi , β

By using the relation gij ni n j = 1 we can transform equation to the form


1
bαβ = g ij n j xαi , β = ε γδ ε ijk xαi , β xαj xβk .
2
The second order symmetric tensor bαβ is called the curvature tensor and the
quadratic form
B = bαβ duα du β
is called the second fundamental form of the surface.
Consider also the tensor derivative with respect to the surface coordinates of the
unit normal vector to the surface. This derivative is
∂ni  i 
n,iα = α +   n j xαk
∂u  jk  g
Taking the tensor derivative of g ij n i n j = 1 with respect to the surface coordinates
produces the result g ij n n, α = 0 which shows that the vector n, α is perpendicular to
i j j

29
n i and must lie in the tangent plane to the surface. It can therefore be expressed as
i
a linear combination of the surface tangent vectors x α and written in the form

n, jα = ηαβ xβi

where the coefficients ηαβ can be written in terms of the surface metric
components aαβ and the curvature components bαβ as follows. The unit vector ni is
normal to the surface so that
g ij ni xαj = 0
The tensor derivative of this equation with respect to the surface coordinates gives
g ij nβi xαj + gij ni xαj ,β = 0
Substitute into equation the relations from equations , and to
show that
bαβ = − aαγ ηβγ
γ
Solving the equation for the coefficients η β we find
η βγ = −aαγ bαβ
Now substituting equation into the equation produces the Weingarten formula
n,iα = −aγβ bγα xβi
This is a relation for the derivative of the unit normal in terms of the surface
metric, curvature tensor and surface tangents.
A third fundamental form of the surface is given by the quadratic form
C = cαβ duα du β where
cα β is defined as the symmetric surface tensor
C = gij n,iα n,βj
By using the Weingarten formula [9] from in the equation one can verify that
cαβ = a γβ bαγ bβδ

3.3 Dynamics
Dynamics is concerned with studying the motion of particles and rigid bodies. By
studying the motion of a single hypothetical particle, one can discern the motion of
a system of particles. This in turn leads to the study of the motion of individual
points in a continuous deformable medium.

30
Particle Movement
The trajectory of a particle in a generalized coordinate system is described by the
parametric equations
x i = xi (t ), i = 1,..., N
where t is a time parameter. If the coordinates are changed to a barred system by
introducing a coordinate transformation
x i = x i ( x1 , x 2 ,..., x N ), i = 1,..., N
then the trajectory of the particle in the barred system of coordinates is
x i = x i ( x1 (t ), x 2 (t ),..., x N (t )), i = 1,..., N
The generalized velocity of the particle in the unbarred system is defined by
dx i
vi = , i = 1,..., N
dt
By the chain rule differentiation of the transformation equations one can verify
that the velocity in the barred system is
dx r ∂x r dx j ∂x r j
vr = = j = j v , r = 1,..., N ,
dt dx dt dx
Consequently, the generalized velocity v i is a first order contravariant tensor. The
speed of the particle is obtained from the magnitude of the velocity and is
v 2 = gij v i v j

The generalized acceleration f i of the particle is defined as the intrinsic derivative


of the generalized velocity. The generalized acceleration has the form
δ vi dx n
f =i
= v ,n
i

δt dt
dv  i  m n
i
= +  v v
dt ij 

d 2 x i  i  dxm dxn
= + 
dt 2 ij  dt dt

and the magnitude of the acceleration is f = gij f f


2 i j

3.4 Serret-Frenet Formulas


The parametric equations describe a curve in our generalized space. With
reference to the figure 2 we wish to define at each point P of the curve the
following orthogonal unit vectors:

31
T i = unit tangent vector at each point P.
N i = unit normal vector at each point P.
B i = unit bi-normal vector at each point P.
These vectors define the osculating, normal and rectifying planes illustrated in the
Figure 2.

Figure 2: Tangent, normal and binormal to point P on curve.

In the generalized coordinates the arc length squared is


ds 2 = gij dxi dx j

Define dx i
Ti =
ds
as the tangent vector to the parametric curve defined by equation .
This vector is a unit tangent vector because if we write the element of arc length
squared in the form
dx i dx j
1 = gij = gij T iT j
ds ds
we obtain the generalized dot product for Ti .This generalized dot product implies
that the tangent vector is a unit vector. Differentiating the equation intrinsically
with respect to arc length s along the curve produces
δT m n δT n
g mn T + gmn T m =0
δs δs
δT m
which simplifies to g mnT n =0
δs

32
The equation is a statement that the vector δ T is orthogonal to the vector T m .
m

δs
The unit normal vector is defined as
1 δT i 1 δ Ti
Ni = or Ni =
κ δs κ δs
where κ is a scalar called the curvature and is chosen such that the magnitude of
1
N i is unity. The reciprocal of the curvature is R = which is called the radius of
κ
curvature.
The curvature of a straight line is zero while the curvature of a circle is a constant.
The curvature measures the rate of change of the tangent vector as the arc length
varies.
The equation can be expressed in the form
j
gijT i N = 0

Taking the intrinsic derivative of equation with respect to the arc length s
produces
δNj i δT i j
gij T + gij N =0
δs δs
δNj δT i j
or gijT i
= − gij N = −κ gij N i N j = −κ
δs δs
The generalized dot product can be written
gijT iT j = 1

and consequently we can express equation in the form


δ Ni
gijT i = −κ gij T i T j
δs
δN j
or gijT i ( + κT j ) = 0
δs
Consequently the vector
δNj
+ κT j
δs
is orthogonal to Ti .
In a similar manner, we can use the relation g i j N i N j =1 and differentiate
intrinsically with respect to the arc length s to show that

33
δN j
gij N i =0
δs
This in turn can be expressed in the for
δN j
gij N i ( + κT j ) = 0
δs
This form of the equation implies that the vector represented in equation is also
orthogonal to the unit normal Ni. We define the unit bi-normal vector as
1 δ Ni 1 δNi
B = (
i
+ κT i ) or Bi = ( + κT i )
τ δs τ δs
where τ is a scalar called the torsion. The torsion is chosen such that the bi-
normal vector is a unit vector. The torsion measures the rate of change of the
osculating plane and consequently, the torsion τ is a measure of the twisting of the
curve out of a plane.
The value τ = 0 corresponds to a plane curve. The vectors T i ,N i ,B i , i=1,2,3
satisfy the cross product relation
B i = ε ijk T j N k

If we differentiate this relation intrinsically with respect to arc length s


we find
δ Bi  δ Nk δ Tj 
= ε ijk  T j + Nk 
δs  δs δS 
= ε  Τj (τ Βk − κ Tk ) + κ Ν j Ν k 
ijk

= τ ε ijk T j Bk

δ Bi
Therefore = −τ ε ijk Bk T j = −τ N i
δs
The relations , and are now summarized and written
δT i 
= κ Ni 
δs 
δ Ni 
= τ Bi − κT i 
δs 
δ Bi 
= −τ N i 
δs 
These equations are known as the Frenet-Serret formulas [6] of differential
geometry.

34
3.5 Differential Operators in Generalized Coordinates
We define the quantities grad ϕ , div A, ∆ϕ and curl A in a system of generalized
coordinates x1 , x 2 , x3

(a) Gradient
By the gradient of a scalar field ϕ = ϕ ( x1, x 2 , x 3 ) we mean the vector with
∂ϕ
covariant components
∂x i

Thus, introducing the “del” operator ∇ = ei
∂xi
we have
∂ϕ
grad ϕ = ∇ϕ = e i
∂xi
the physical components of ∇ϕ are
1 ∂ϕ
(∇ϕ )i* = i (no summation over i)
gii ∂x
In the case of orthogonal coordinates, becomes
1 ∂ϕ
(∇ ϕ )i* =
hi ∂xi
in terms of the metric coefficients hi . All the properties of the gradient continue to
hold in generalized coordinates. In particular, the directional derivative of ϕ in
the direction l equals

= l.∇ϕ
dl
where , l = l i ei

(b) Divergence
The divergence of a vector field A = A(x1, x2, x3) is defined as the contraction of
the (mixed) covariant derivative of A, i.e.
∂Ai i  j
div A = A , i =
i
+  A
∂xi ij 
i 
The sum   can be expressed in terms of the metric tensor.
ij 

35
In fact, using symmetry of gij, we have
i 
  = g [k , ij ] + [i, kj ]
ij

ij
 
∂gij
where, [k , ij ] + [i, kj ] =
∂xi
Expanding the determinant G = det gik with respect to elements of the i th row,
we obtain
G = g ikG ik (no summation over i )

where G ik is the cofactor of g ik in the determinant G. But Gik is independent of gik


, and hence
∂G
= G ik
∂gik
G ik 1 ∂G
Therefore g =
ik
=
G G ∂gik
It follows from , , that
 i  1 ∂G ∂g ik 1 ∂G 1 ∂ G
 = = =
ij  2G ∂gik ∂x 2G ∂x G ∂x
j j j

Thus, finally,
∂Ai Ai ∂ G 1 ∂ i
div A = i + = ( A G)
∂x G ∂x
i
G ∂x
j

1 ∂
= ( g ik Ak G )
G ∂x
i

1 3 ∂ A*i G
= ∑ i
G i=1 ∂x gii

where A*i the are the physical components of the vector A. In particular,
1 3 ∂ Ai* h1 h2 h3
div A = ∑
h1h2 h3 i =1 ∂xi hi
in orthogonal coordinates, where Ai* can be replaced by Ai if the local basis is
orthonormal.

(c) Laplacian

36
By the Laplacian of a scalar field ϕ we meant the quantity
∆ϕ = div grad ϕ
Combining and , we have
1 ∂ 3
 ∂ϕ 
∆ϕ = ∑  g ik G 
G ∂x ∂x k 
i
i =1 

In orthogonal coordinates, becomes


1 ∂  h1h2 h3 ∂ϕ 
∆ϕ =  
h1h2 h3 ∂xi  hi ∂xi 
(d) Curl
In generalized coordinates, the curl of a vector field A is defined as the vector
product of the operator ∇ and the vector A.
∂ ∂ A
Thus, curl A= ∇×A =e j
×jA
= e j = j e× je Ak k ,j
∂x ∂ x
∂A
since, gek = Ak , j
∂x j
But
 ei
 G if i, j, k is a cyclic permutation of 1,2,3

 e
e × e =  − i if i , j ,k is a cyclic permutation of 2,1,3
j k

 G
 0 otherwise


where G = det g ik . Therefore takes the form
ei
curl A = ∑ ( Ak , j − A j.k )
i, j ,k G
where the indices i,j,k are a cyclic permutation of the numbers 1, 2, 3.
Moreover,
∂Ak  l 
Ak , j = −   Al
∂x j kj 
∂Aj l 
and A j, k = −   Al
∂x k  jk 

37
∂Ak ∂A j
Hence Ak , i − A j , k = −
∂x j ∂x k
 l  l 
since  = 
 jk  kj 
3
ei  ∂Ak ∂A j 
Thus, finally, curl A = ∑  j − ∂x k 
G  ∂x
i , j ,k 
where i,j,k is a cyclic permutation of 1, 2, 3.
Formula leads to the following expressions for the contravariant and physical
components of curl A (there is no summation over j and k):
1  ∂Ak ∂Aj 
( curl A )
i
=  j − k
G  ∂x ∂x 
gii  ∂Ak ∂Aj 
( curl A )
*i
=  j − k
G  ∂x ∂x 

g ii  ∂Ak* g kk ∂Aj g jj 
*

=  − 
G  ∂x ∂x k 
j

In the case of orthogonal coordinates with an orthonormal local basis, reduces to
hi  ∂ ( Ak hk ) ∂ ( Aj h j ) 
( curl A ) i =  −  (no summation over j and k)
h1h2 h3  ∂x j ∂x k 

Here, we introduced the unit pseudo tensor ε jkl [5], defined as ε jkl = i j × i k gi l ( )
in a system of rectangular coordinates with orthonormal basis i1 , i2 , i3 . In a system
of generalized coordinates with local basis e1 , e2 , e3 we replace this definition by
ε jkl = ( e j × e k ) g e l
so that
 G if j, k , l is a cyclic permutation of 1,2,3

ε jkl = − G if j, k , l is a cyclic permutation of 2,1,3
 0 otherwise

The contravariant components ε jki
then turn out to be [5]

38
 1
 G if j, k , l is a cyclic permutation of 1,2,3

 1
ε jkl = − if j, k , l is a cyclic permutation of 2,1,3
 G
 0 otherwise


In terms of the unit pseudo tensor [5], the vector product C = A × B has
covariant and contravariant components
C i = ε ijk A j Bk and Ci = ε ijk Aj Bk
while curl A has components

( curl A ) = ε jki Ak , j
i

( curl A ) i = gil ( curl A) = gil ε jkl Ak , j


l

39
Table 3: Comaparision of gradiant, divergent, curl, Laplacian of vectors, general
tensor and Cartesian vector [6] .

Vector General Tensor Cartesian Tensor


r
A Ai or Ai Ai

Ai Bi = gij A B = Ai B
i j i
r r
A.B ij
Ai Bi
= g Ai B j

r r r 1 ijk
C=A× B Ci = e Aj Bk Ci = eijk Aj Bk
g
∂φ
∇ φ =grad φ g mnφ , m φ ,i =
∂x i
r r 1 ∂ ∂Ai
∇ . A =Div A g mn Am,n = Ar ,r = ( g Ar ) Ai ,i =
g ∂x
r
∂x i

r r r ∂Ak
C i = ε Ak , j Ci = ε ijk
ijk
∇ × A = C =Curl A
∂x j
1 ∂ ∂U ∂ ∂U
∇2 U g mnU ,mn = ( g g ij i ) ( )
g ∂x ∂x
j
∂x i ∂xi
r r r ∂Bi
C =( A . ∇ .) B C i = Ai B j ,m Ci = Am
∂x m
r r r ∂Bm
C = A (∇ . B ) C i = Ai B j , j Ci = Am
∂x m
r r ∂ ∂Ai
C= ∇ .A
2
C i = g im Ai ,mj or Ci = g jm Ai ,mj Ci = ( )
∂x m ∂x m
r
( A . ∇ )φ g im Ai φ ,m Aiφ ,i

r ∂ 2 Ar
∇ (∇ . A ) g im ( Ar , r ) m
∂xi ∂xr

r ∂ 2 Aj ∂ 2 Ai
∇ ×(∇ × A ) ε ijk g (ε At ,s ),m
jm kst

∂x j ∂xi ∂x j ∂x j

40
3.6 Applications to Fluid Dynamics
Equations of fluid motion
Given a moving fluid (liquid or gas) described by a velocity field v = v(r, t),
Let V be a ‘material volume’ i.e., a volume moving with the fluid and hence
always consisting of the same fluid particles. Then V and its surface S are in
general functions of time. The total momentum of the volume V is

∫∫∫ ρ v dV ,
V

While the total body force acting on V is

∫∫∫ f dV
V

where f is the body force per unit mass. Besides in , V is also subject to internal
forces acting across its surface S.
Let Pn be the stress acting on an element of area dS with unit exterior normal n.
Then the total force acting on V due to the stress on S is

∫∫ P
S
n dS

It follows from Newton’s second law that


d
dt ∫∫∫
ρ v dV = ∫∫∫ ρ f dV + ∫∫ Pn dS
V V S

To transform further, we note that


d dv
∫∫∫
dt V
ρ v dV = ∫∫∫ ρ dV .
V
dt
Let V0 be the volume occupied by V = V(t) at time t = 0, and let r = r( ξ , t) be the
position at time t of the fluid particle which had radius vector ξ at time t = 0. Then
d d
∫∫∫ ρ v dV = ∫∫∫ ρ ( r, t ) v ( r, t ) dV
dt V dt V
d dV
( ) ( )
dt ∫∫∫
= ρ 
 r ξ , t , t 
 v 
 r ξ , t , t 
 dV0
V0
dV0

d dV d  dV  
= ∫∫∫  ( ρ v ) + ρv    dV0
V0 
dt dV0 dt  dV0  
d  dV
= ∫∫∫  ( ρ v ) + ρ v div v  dV0
V0   dV0
dt

41
where we use the fact that
d  dV 
  = div v
dt  dV0 
Therefore
d d 
∫∫∫ ρ v dV = ∫∫∫  ( ρ v ) + ρ v div v  dV
V  
dt V dt
 dv  dρ 
= ∫∫∫  ρ + v + div v   dV
V 
dt  dV0 
dv
= ∫∫∫ ρ dV
V
dt
in accordance with , because of the equation of continuity
dρ ∂ρ
+ ρ div v = + v.∇ρ + ρ div v
dt ∂t
∂ρ
= + div ( ρ v )
∂t
=0
Using , we can write
dv
∫∫∫ ρ
v
dt
dV = ∫∫∫ ρ f dV + ∫∫ Pn ds
v s

= ∫∫∫ ρ f dV + ∫∫ Pk nk ds
v s

In component form,
dvi
∫∫∫ ρ
V
dt
dV = ∫∫∫ ρ f i dV + ∫∫ Pik nk dS
V S

where pik is the stress tensor. To obtain a differential equation describing the
motion of the fluid, we use Gauss’ theorem [3] in the form to transform the surface
integral into a volume integral:
∂pik
∫∫ p
S
ik nk ds = ∫∫∫
V
∂xk
dV

Then becomes
 ∂vi ∂pik 
∫∫∫
V 
 ρ
∂t
− ρ f i −
∂xk
 dV = 0

42
Since the volume V is arbitrary, we have
∂vi ∂p
ρ = ρ f i + ik
∂t ∂xk
dvi
assuming that the integrand is continuous. Here is the total derivative, which
dt
can be written in the form
dvi ∂vi ∂v
= + vk i
dt ∂t ∂xk
Thus finally, combining and , we find that the motion of the fluid is described by
the differential equation
∂vi ∂v ∂p
ρ + ρ vk i = ρ fi + ik ( i = 1, 2,3)
∂t ∂xk ∂xk
Next we make the usual hydro dynamical assumption that the viscous stress tensor
pik is a linear function of the rate of deformation tensor vik . Then we have
pik = − pδ ik + 2 µ vik + µ ′δik vll
where p is the hydrostatic pressure, μ and μ’ are constants of proportionality
(called viscosity coefficients), and
∂vl
vll = = div v
∂xl
Under certain circumstances, it can be assumed that
2
µ′ + µ = 0
3
and then becomes
2
pik = − pδ ik + 2 µ vik − µδik vll .
3
If the fluid is incompressible, div v = 0 and hence
pik = − pδ ik + 2 µ vik

If the fluid is at rest or if the fluid is “ideal” with no viscosity (μ = 0), we have just
pik = − pδ ik

43
Substituting , , in turn into , we deduce the equations of motion for three kinds of
fluids:
1) The ideal fluid (μ = 0). In this ease, becomes
∂vi ∂v ∂p
ρ + ρ vk i = ρ fi −
∂t ∂xk ∂xi
∂v
Or ρ + ρ vk (v.∇) v = ρ f − ∇p [in vector notation]
∂t
2) The viscous incompressible fluid (μ=const ≠ 0,div v=0). We now have the
Navier-Stokes equation
∂vi ∂vi ∂p ∂ 2 vi
ρ + ρ vk = ρ fi − +µ
∂t ∂xk ∂xi ∂xk ∂xk
∂ ∂ ∂vi ∂vk
Since (2 µ vik ) = µ ( + )
∂xk ∂xk ∂xk ∂xi

∂ 2 vi ∂  ∂vk 
=µ +µ  
∂xk ∂xk ∂xk  ∂xk 
∂ 2 vi

∂xk ∂xk
The vector form of the Navier-Stokes equation is
∂v
ρ + ρ vk (∇ ⋅ v ) v = ρ f − ∇p + µ ∆v
∂t
3) The viscous compressible fluid (μ=const ≠ 0,div v≠0).
In this case becomes
∂vi ∂v ∂p ∂ 2 vi 1 ∂ 2 vk
ρ + ρ vk i = ρ fi − +µ + µ
∂t ∂xk ∂xi ∂xk ∂xk 3 ∂xk ∂xk
∂v 1
Or ρ + ρ (v.∇) v = ρ f − ∇p + µ ∆v + µ grad div v
∂t 3

Example (Archimedes’ law)


The force F exerted by a fluid on a body of volume V and surface S immersed in
the fluid is
F = ∫∫ p n ds =∫∫ pn nk dS
S S

44
with components
Fi = ∫∫ pik nk ds
S

If the fluid is at rest (v= 0), then, according to and


pik = − pδ ik 

∇p = ρ f = ρ g 
where g is the acceleration due to gravity . It follows that
Fi = − ∫∫ pni dS
S

Or F = − ∫∫ pn ds = −∫∫∫ ∇p dv
S V

where g has been used in the last step Substituting from , we find that
F = − ∫∫∫ ρ g dv = −g∫∫∫ ρ dv = −gm
V V

where m is the mass of fluid displaced by the body. Thus we have proved
Archimedes’ law, i.e., the force exerted by a fluid on a body immersed in the fluid
equals the weight m|g| of the displaced fluid and points in the direction opposite to
the force of gravity.

3.7 The Fundamental Theorem of Vector Analysis


Statement [3]: Let A = A(r) be a continuous vector field with continuous
1
divergence and curl, such that |A| falls off at infinity like 1+ε while |div A| and |
r
1
curl A| fall off at infinity like 2+ε where ε > 0. Then A has a unique
r
representation (to within constant vectors) as a sum of a potential field A1 = A1(r)
and a solenoidal field A2 = A2(r), i.e.,
A = A1 + A2
where,
curl A1 = 0, div A2 = 0.
Proof. First we construct A1. It follows from that
div A = div A1, curl A = curl A2,
so that curl A1 = 0 , div A1 = div A.
The first of these equations implies
A1= ∇ φ + c1

45
where c1 is a constant vector and φ is a single-valued potential (we assume that the
region occupied by A and hence by A1 is simply connected). The second of the
equations then gives
div (∇ φ + c1) = div A,
or ∇ φ = div A
Thus the constant vector c1 plays no role in determining the potential φ. To solve
for φ we use formula
1 1 1  ∂  1  1 ∂ϕ 
ϕ (M o ) =
4π ∫∫∫ r ∆ϕ dv − 4π ∫∫ ϕ  ∂n  r  − r ∂n  dS ,
V S
∂ϕ
where V is all of space (in the limit). The functions φ and both approach zero
∂n
∂ϕ 1
at infinity, with going to zero faster than . It follows that the surface
∂n r
integral
 ∂  1  1 ∂ϕ 
∫∫S ϕ ∂n  r  − r ∂n  dS
vanishes in the limit as V becomes all of space, so that reduces to
1 div A
ϕ =−
4π ∫∫∫
V
r
dV

Here div A is regarded as known. It follows


1 div A
A 1 = ∇ϕ + c1 = − grad ∫∫∫ dV + c1
4π V
r
Next we construct A2. According to and ,
div A2 = 0, curl A2 = curl A.
The first of these equations
A2 = curl Ф + c2,
where c2 is a constant vector and Ф is a vector potential. Since Ф is determined
only to within the gradient of an arbitrary function f, Ф can be subjected to the
extra condition
div Ф = 0
Substituting into the second of the equations , we obtain
curl (curl Ф + c2) = curl A

46
The constant vector c2 plays no role in determining the vector potential Φ here.
Thus
curl curl Ф = curl A,
which implies
ΔФ = − curl A ,
Just as we used to solve for φ, similarly we can solve Φ such that
1 curl A
Φ=−
4π ∫∫∫
V
r
dV

And hence
1 curl A
A2 = curl ∫∫∫ dV + c2
4π V
r
We finally have the representation
A = A1 + A 2
1 div A 1 curl A
=− grad ∫∫∫ dV + curl ∫∫∫ dV
4π V
r 4π V
r
where the constants c1 and c2 have been dropped since A is assumed to vanish at
infinity. By construction, A1 is a potential field (curl A1 = 0), while A2 is a
solenoidal field (div A2 = 0).
We must still prove the uniqueness of the representation . Suppose A has another
representation A= A1’+ A2’; where curl A1’= 0, div A2’ =0.
Since div A = div A1 = div A1’,
we have div (A1 - A1’) = 0 and curl (A1 - A1’) = 0.
Similarly curl A = curl A2 = curl A2’
and hence div (A2 – A2’) = 0 and curl (A2 - A2’) = 0
But then the same argument leading from and to and shows that
A1 – A1’= c1 and A2 – A2’= c2
where c1 and c2 are constant vectors.
Thus the representation A = A1 + A2 is unique to within constant vectors
Hence, the proof of the theorem is complete . 

47

You might also like