You are on page 1of 119

Tensor Calculus

Differentials & Directional Derivatives

The Gateaux Differential

We are presently concerned with Inner Product spaces in our treatment of the Mechanics of Continua. Consider a map, : V W This maps from the domainV to W both of which are Euclidean vector spaces. The concepts of limit and continuity carries naturally from the real space to any Euclidean vector space.
Dept of Systems Engineering, University of Lagos 2 oafak@unilag.edu.ng 12/27/2012

The Gateaux Differential

Let 0 V and 0 W, as usual we can say that the limit lim = 0


0

if for any pre-assigned real number > 0, no matter how small, we can always find a real number > 0 such that 0 whenever 0 < . The function is said to be continuous at 0 if 0 exists and 0 = 0
Dept of Systems Engineering, University of Lagos 3 oafak@unilag.edu.ng 12/27/2012

The Gateaux Differential

Specifically, for let this map be: + , lim = + 0

=0

We focus attention on the second variable while we allow the dependency on to be as general as possible. We shall show that while the above function can be any given function of (linear or nonlinear), the above map is always linear in irrespective of what kind of Euclidean space we are mapping from or into. It is called the Gateaux Differential.
Dept of Systems Engineering, University of Lagos 4 oafak@unilag.edu.ng 12/27/2012

The Gateaux Differential


Let us make the Gateaux differential a little more familiar in real space in two steps: First, we move to the real space and allow and we obtain, + (, ) = lim = + 0 =0 And let , the middle term becomes, + lim = 0 from which it is obvious that the Gateaux derivative is a generalization of the well-known differential from elementary calculus. The Gateaux differential helps to compute a local linear approximation of any function (linear or nonlinear).
Dept of Systems Engineering, University of Lagos 5 oafak@unilag.edu.ng 12/27/2012

The Gateaux Differential

It is easily shown that the Gateaux differential is linear in its second argument, ie, for R , = , Furthermore, , + = , + , and that for , R , + = , + ,

Dept of Systems Engineering, University of Lagos

oafak@unilag.edu.ng 12/27/2012

The Frecht Derivative

The Frecht derivative or gradient of a differentiable function is the linear operator, grad () such that, for V, grad , Obviously, grad () is a tensor because it is a linear transformation from one Euclidean vector space to another. Its linearity derives from the second argument of the RHS.
Dept of Systems Engineering, University of Lagos 7 oafak@unilag.edu.ng 12/27/2012

Differentiable Real-Valued Function


The Gradient of a Differentiable Real-Valued Function For a real vector function, we have the map: : V R The Gateaux differential in this case takes two vector arguments and maps to a real value: : V V R This is the classical definition of the inner product so that we can define the differential as, () , This quantity, () defined by the above equation, is clearly a vector. oafak@unilag.edu.ng 12/27/2012 Dept of Systems Engineering, University of Lagos 8

Real-Valued Function
It is the gradient of the scalar valued function of the vector variable. We now proceed to find its components in general coordinates. To do this, we choose a basis V. On such a basis, the function, = and, = we may also express the independent vector = on this basis.
Dept of Systems Engineering, University of Lagos 9 oafak@unilag.edu.ng 12/27/2012

Real-Valued Function
The Gateaux differential in the direction of the basis vector is, + + , = lim = lim 0 0 + = lim 0 As the function does not depend on the vector basis, we can substitute the vector function by the real function of the components 1 , 2 , 3 so that, = 1 , 2 , 3 in which case, the above differential, similar to the real case discussed earlier becomes, 1 , 2 , 3 , =
Dept of Systems Engineering, University of Lagos 10 oafak@unilag.edu.ng 12/27/2012

Of course, it is now obvious that the gradient of the scalar valued vector function, expressed in its dual basis must be, () 1 , 2 , 3 = so that we can recover, 1 , 2 , 3 , = = 1 , 2 , 3 1 , 2 , 3 = = Hence we obtain the well-known result that () 1 , 2 , 3 grad = = which defines the Frecht derivative of a scalar valued function with respect to its vector argument.
Dept of Systems Engineering, University of Lagos 11 oafak@unilag.edu.ng 12/27/2012

()

Vector valued Function


The Gradient of a Differentiable Vector valued Function The definition of the Frecht clearly shows that the gradient of a vector-valued function is itself a second order tensor. We now find the components of this tensor in general coordinates. To do this, we choose a basis V. On such a basis, the function, = The functional dependency on the basis vectors are ignorable on account of the fact that the components themselves are fixed with respect to the basis. We can therefore write, = 1 , 2 , 3
Dept of Systems Engineering, University of Lagos 12 oafak@unilag.edu.ng 12/27/2012

Vector Derivatives
, = , = , = , Again, upon noting that the functions , , = 1,2,3 are not functions of the vector basis, they can be written as functions of the scalar components alone so that we have, as before, , = , = ,

Dept of Systems Engineering, University of Lagos

13

oafak@unilag.edu.ng 12/27/2012

Which we can compare to the earlier case of scalar valued function and easily obtain, 1 , 2 , 3 , = 1 , 2 , 3 = 1 , 2 , 3 = = grad Clearly, the tensor gradient of the vector-valued vector function is, 1 , 2 , 3 = grad = Where due attention should be paid to the covariance of the quotient indices in contrast to the contravariance of the non quotient indices. Dept of Systems Engineering, University of Lagos 14 oafak@unilag.edu.ng 12/27/2012

The Trace & Divergence


The trace of this gradient is called the divergence of the function: 1 , 2 , 3 div = tr 1 , 2 , 3 = 1 , 2 , 3 = 1 , 2 , 3 =
Dept of Systems Engineering, University of Lagos 15 oafak@unilag.edu.ng 12/27/2012

Tensor-Valued Vector Function


Consider the mapping, : V T Which maps from an inner product space to a tensor space. The Gateaux differential in this case now takes two vectors and produces a tensor: : V V T With the usual notation, we may write, , = , = , = grad = , = , Each of these nine functions look like the real differential of several variables we considered earlier.
Dept of Systems Engineering, University of Lagos 16 oafak@unilag.edu.ng 12/27/2012

The independence of the basis vectors imply, as usual, that 1 , 2 , 3 , = so that, , = , 1 , 2 , 3 = 1 , 2 , 3 = 1 , 2 , 3 = = grad Which defines the third-order tensor, 1 , 2 , 3 = grad = and with no further ado, we can see that a third-order tensor transforms a vector into a second order tensor.
Dept of Systems Engineering, University of Lagos 17 oafak@unilag.edu.ng 12/27/2012

The Divergence of a Tensor Function


The divergence operation can be defined in several ways. Most common is achieved by the contraction of the last two basis so that, 1 , 2 , 3 div = 1 , 2 , 3 = 1 , 2 , 3 1 , 2 , 3 = = It can easily be shown that this is the particular vector that gives, for all constant vectors , div div T
Dept of Systems Engineering, University of Lagos 18 oafak@unilag.edu.ng 12/27/2012

Real-Valued Tensor Functions


Two other kinds of functions are critically important in our study. These are real valued functions of tensors and tensor-valued functions of tensors. Examples in the first case are the invariants of the tensor function that we have already seen. We can express stress in terms of strains and vice versa. These are tensor-valued functions of tensors. The derivatives of such real and tensor functions arise in our analysis of continua. In this section these are shown to result from the appropriate Gateaux differentials. The gradients or Frecht derivatives will be extracted once we can obtain the Gateaux differentials.
Dept of Systems Engineering, University of Lagos 19 oafak@unilag.edu.ng 12/27/2012

Frecht Derivative
Consider the Map: : T R The Gateaux differential in this case takes two vector arguments and maps to a real value: : T T R Which is, as usual, linear in the second argument. The Frecht derivative can be expressed as the first component of the following scalar product: () , = : which, we recall, is the trace of the contraction of one tensor with the transpose of the other second-order tensors. This is a scalar quantity.
Dept of Systems Engineering, University of Lagos 20 oafak@unilag.edu.ng 12/27/2012

Guided by the previous result of the gradient of the scalar valued function of a vector, it is not difficult to see that, () (11 , 12 , , 33 ) = In the dual to the same basis, we can write, = Clearly, () (11 , 12 , , 33 ) : = : (11 , 12 , , 33 ) = Again, note the covariance of the quotient indices.

Dept of Systems Engineering, University of Lagos

21

oafak@unilag.edu.ng 12/27/2012

Examples
Let be a symmetric and positive definite tensor and let 1 , 2 and 3 be the three principal invariants of show that (a)
2 1 1

= the identity tensor, (b)


3

= 1 and (c)

= 3 1

can be written in the invariant component form


1 1 =
22 oafak@unilag.edu.ng 12/27/2012

as,

Dept of Systems Engineering, University of Lagos

(a) Continued
Recall that 1 = tr = hence 1 1 = = = = =

which is the identity tensor as expected.


Dept of Systems Engineering, University of Lagos 23 oafak@unilag.edu.ng 12/27/2012

2 (b)
2

= 1

in a similar way can be written in the invariant component form

as,

2 1 1 = 2 1 where we have utilized the fact that 2 = 2 tr 2 tr( 2 ) . Consequently, 2 1 = 2 1 = + 2 1 = + 2 = = 1


Dept of Systems Engineering, University of Lagos 24 oafak@unilag.edu.ng 12/27/2012

the cofactor of . Clearly = det 1 = 3 1 as required. Details of this for the contravariant components of a tensor is presented below. Let 1 = = 3! Differentiating wrt , we obtain, 1 = + + 3! 1 = + + 3! 1 = + + 3! 1 = c 2! Which is the cofactor of
Dept of Systems Engineering, University of Lagos 25 oafak@unilag.edu.ng 12/27/2012

det

Real Tensor Functions

Consider the function () = tr where , and is a tensor. = tr = :

To be specific, let = 3.

Dept of Systems Engineering, University of Lagos

26

oafak@unilag.edu.ng 12/27/2012

Real Tensor Functions


The Gateaux differential in this case, , = + = tr + 3 =0 =0 = tr + + + =0 = tr + + + =0 = tr + + + + + + + + = tr + + = 3 2 T :
Dept of Systems Engineering, University of Lagos 27 oafak@unilag.edu.ng 12/27/2012

=0

Real Tensor Functions


With the last equality coming from the definition of the Inner product while noting that a circular permutation does not alter the value of the trace. It is easy to establish inductively that in the most general case, for > 0, we have,

, =
Clearly,

1 T

:
T

tr = 1

Dept of Systems Engineering, University of Lagos

28

oafak@unilag.edu.ng 12/27/2012

Real Tensor Functions


When = 1,
, = + =0 = tr + =0 = tr = :

Or that,

tr = .
29 oafak@unilag.edu.ng 12/27/2012

Dept of Systems Engineering, University of Lagos

Real Tensor Functions

Derivatives of the other two invariants of the tensor can be found as follows: 2 () 1 = tr 2 tr 2 2 1 = 2tr 2 T = tr T 2 .
Dept of Systems Engineering, University of Lagos 30 oafak@unilag.edu.ng 12/27/2012

Real Tensor Functions


To Determine the derivative of the third invariant, we begin with the trace of the Cayley-Hamilton for : tr 3 1 2 + 2 3 = tr(3 ) 1 tr 2 + 2 tr 33 = 0 Therefore, 33 = tr(3 ) 1 tr 2 + 2 tr 1 2 2 = tr tr 2 2
.
Dept of Systems Engineering, University of Lagos 31 oafak@unilag.edu.ng 12/27/2012

Differentiating the Invariants

We can therefore write, 33 = tr(3 ) tr tr 2 1 2 + tr tr 2 tr 2 so that, in terms of traces only, 1 3 3 = tr 3tr tr 2 + 2tr(3 ) 6

Dept of Systems Engineering, University of Lagos

32

oafak@unilag.edu.ng 12/27/2012

Real Tensor Functions

Clearly, 3 () 1 = 3tr 2 3tr 2 3tr 2 + 2 3 2 6 = 2 tr + 2

Dept of Systems Engineering, University of Lagos

33

oafak@unilag.edu.ng 12/27/2012

Real Tensor Functions

Dept of Systems Engineering, University of Lagos

34

oafak@unilag.edu.ng 12/27/2012

The Euclidean Point Space

In Classical Theory, the world is a Euclidean Point Space of dimension three. We shall define this concept now and consequently give specific meanings to related concepts such as Frames of Reference, Coordinate Systems and Global Charts

Dept of Systems Engineering, University of Lagos

35

oafak@unilag.edu.ng 12/27/2012

The Euclidean Point Space

Scalars, Vectors and Tensors we will deal with in general vary from point to point in the material. They are therefore to be regarded as functions of position in the physical space occupied. Such functions, associated with positions in the Euclidean point space, are called fields. We will therefore be dealing with scalar, vector and tensor fields.
Dept of Systems Engineering, University of Lagos 36 oafak@unilag.edu.ng 12/27/2012

The Euclidean Point Space


The Euclidean Point Space is a set of elements called points. For each pair of points , , , E with the following two properties: 1. , = , + , , , 2. , = , = Based on these two, we proceed to show that, , = And that, , = ,
Dept of Systems Engineering, University of Lagos 37 oafak@unilag.edu.ng 12/27/2012

The Euclidean Point Space

From property 1, let it is clear that, , = , + , And it we further allow , we find that, , = , + , = 2 , Which clearly shows that , = the zero vector. Similarly, from the above, we find that, , = = , + , So that , = ,
Dept of Systems Engineering, University of Lagos 38 oafak@unilag.edu.ng 12/27/2012

Dept of Systems Engineering, University of Lagos

39

oafak@unilag.edu.ng 12/27/2012

Coordinate System

Dept of Systems Engineering, University of Lagos

40

oafak@unilag.edu.ng 12/27/2012

Coordinate System

Dept of Systems Engineering, University of Lagos

41

oafak@unilag.edu.ng 12/27/2012

Position Vectors
Note that is NOT a vector space. In our discussion, the vector space E to which belongs, is associated as we have shown. It is customary to oscillate between these two spaces. When we are talking about the vectors, they are in E while the points are in . We adopt the convention that , referring to the vector . If therefore we choose an arbitrarily fixed point , we are associating , and respectively with , , , and , . These are vectors based on the points , and with reference to the origin chosen. To emphasize the association with both points of as well as members of E they are called Position Vectors.
Dept of Systems Engineering, University of Lagos 42 oafak@unilag.edu.ng 12/27/2012

Length in the Point Space


Recall that by property 1, , = , + , Furthermore, we have deduced that , = (, ) We may therefore write that, = () Which, when there is no ambiguity concerning the chosen origin, we can write as, = And the distance between the two is, = =
Dept of Systems Engineering, University of Lagos 43 oafak@unilag.edu.ng 12/27/2012

Metric Properties
The norm of a vector is the square root of the inner product of the vector with itself. If the coordinates of and on a set of independent vectors are and , then the distance we seek is, = = ( )( )

The more familiar Pythagorean form occurring only when = 1.


Dept of Systems Engineering, University of Lagos 44 oafak@unilag.edu.ng 12/27/2012

Coordinate Transformations
Consider a Cartesian coordinate system , , or 1 , 2 and 3 with an orthogonal basis. Let us now have the possibility of transforming to another coordinate system of an arbitrary nature: 1 , 2 , 3 . We can represent the transformation and its inverse in the equations: = 1 , 2 , 3 , = (1 , 2 , 3 ) And if the Jacobian of transformation,
Dept of Systems Engineering, University of Lagos 45 oafak@unilag.edu.ng 12/27/2012

Jacobian Determinant
1 1 1 1 2 3 1 , 2 , 3 2 2 2 1 , 2 , 3 1 2 3 3 3 3 1 2 3 does not vanish, then the inverse transformation will exist. So that, = (1 , 2 , 3 )
Dept of Systems Engineering, University of Lagos 46 oafak@unilag.edu.ng 12/27/2012

Given a position vector = ( 1 , 2 , 3 ) In the new coordinate system, we can form a set of three vectors, = , = 1,2,3 Which represent vectors along the tangents to the coordinate lines. (This is easily established for the Cartesian system and it is true in all systems. = 1 , 2 , 3 = 1 + 2 + 3 So that = 1 , = 2 and = 3 The fact that this is true in the general case is easily seen when we consider that along those lines, only the variable we are differentiating with respect to varies. Consider the general case where, = 1 , 2 , 3 The total differential of is simply,
Dept of Systems Engineering, University of Lagos 47 oafak@unilag.edu.ng 12/27/2012

Natural Dual Bases


1+ 2+ = 1 3 2 3 1 1 + 2 2 + 3 3 With 1 , 2 , 3 , = 1,2,3 now depending in general on 1 , 2 and 3 now forming a basis on which we can describe other vectors in the coordinate system. We have no guarantees that this vectors are unit in length nor that they are orthogonal to one another. In the Cartesian case, , = 1,2,3 are constants, normalized and orthogonal. They are our familiar = 1 , = 2 and = 3 .
Dept of Systems Engineering, University of Lagos 48 oafak@unilag.edu.ng 12/27/2012

We now proceed to show that the set of basis vectors, is reciprocal to . The total differential can be written in terms of partial differentials as, = Which when expressed in component form yields, 1 1 1 2 1 3 1 = 1 + 2 + 3 2 1 2 2 2 3 2 = 1 + 2 + 3 3 1 3 2 3 3 3 = 1 + 2 + 3 Or, more compactly that, = =
Dept of Systems Engineering, University of Lagos 49 oafak@unilag.edu.ng 12/27/2012

In Cartesian coordinates, , = 1, . . , 3 for any scalar field (1 , 2 , 3 ), = + + = grad We are treating each curvilinear coordinate as a scalar field, = (1 , 2 , 3 ). = = = . The last equality arises from the fact that this is the only way one component of the coordinate differential can equal another. = Which recovers for us the reciprocity relationship and shows that and are reciprocal systems.
Dept of Systems Engineering, University of Lagos 50 oafak@unilag.edu.ng 12/27/2012

The position vector in Cartesian coordinates is = . Show that (a) = , (b) = , (c) = , and (d) = and (e) = grad = , = = div = , = = = 3. = = grad = , div = , + , = + = + = 4 = 4 curl = , = , + , = + = + = =
Dept of Systems Engineering, University of Lagos 51 oafak@unilag.edu.ng 12/27/2012

For a scalar field and a tensor field show that = + . Also show that = + . grad = , = , + , = grad + grad Furthermore, we can contract the last two bases and obtain, div = , + , = , + , = , + , = grad + div
Dept of Systems Engineering, University of Lagos 52 oafak@unilag.edu.ng 12/27/2012

For two arbitrary vectors, and , show that =


grad = , = , + , = , + , = grad + grad

Dept of Systems Engineering, University of Lagos

53

oafak@unilag.edu.ng 12/27/2012

For any two tensor fields and , Show that, : grad = curl : grad = : , = , = , = , = curl

Dept of Systems Engineering, University of Lagos

54

oafak@unilag.edu.ng 12/27/2012

For a vector field , show that is a third ranked tensor. Hence or otherwise show that = . The secondorder tensor is defined as . Taking the covariant derivative with an independent base, we have grad = , This gives a third order tensor as we have seen. Contracting on the last two bases, div = , = , = , = curl

Dept of Systems Engineering, University of Lagos

55

oafak@unilag.edu.ng 12/27/2012

Show that = grad Note that = , and that curl = , so that, curl = , = , = , = grad Show that curl = div grad = curl = , so that curl = , = , = , , = div grad
56 oafak@unilag.edu.ng 12/27/2012

Dept of Systems Engineering, University of Lagos

Show that = div = , Noting that the tensor behaves as a constant under a covariant differentiation, we can write, div = , = , + , = curl curl Given a scalar point function and a vector field , show that = curl + grad . curl = , = , + , = , + , = grad + curl
Dept of Systems Engineering, University of Lagos 57 oafak@unilag.edu.ng 12/27/2012

Given that = () , Show that =

2 :
Now,

2 = 2 = : + : = 2: as inner product is commutative. We can therefore write that = : = : as required.


Dept of Systems Engineering, University of Lagos 58 oafak@unilag.edu.ng 12/27/2012

Given a tensor field , obtain the vector and show that its divergence is : +

The divergence of is the scalar sum , , . Expanding the product covariant derivative we obtain, div T = , = , + , = div + tr T grad = div + : grad Recall that scalar product of two vectors is commutative so that div T = : grad + div

Dept of Systems Engineering, University of Lagos

59

oafak@unilag.edu.ng 12/27/2012

For a second-order tensor define curl , show that for any constant vector , curl = curl Express vector in the invariant form with contravariant components as = . It follows that curl = , = , = , = , = ,
The last equality resulting from the fact that vector is a constant vector. Clearly, curl = curl
Dept of Systems Engineering, University of Lagos 60 oafak@unilag.edu.ng 12/27/2012

For any two vectors and , show that curl = grad + curl where is the skew tensor .
Recall that the curl of a tensor is defined by curl , . Clearly therefore, curl = , = , + , = , + , = , + , = , + , = grad + curl = grad + curl upon noting that the vector cross is a skew tensor.

Dept of Systems Engineering, University of Lagos

61

oafak@unilag.edu.ng 12/27/2012

Show that curl = div The vector = = and curl = , . Therefore, curl = , = ,
= =

, + ,

= , + , , + , = , , = div

since div = , = , .
Dept of Systems Engineering, University of Lagos 62 oafak@unilag.edu.ng 12/27/2012

Given a scalar point function and a second-order tensor field , show that = + grad where grad is the skew tensor , curl , = , + , = , + , = , + , = curl + grad
Dept of Systems Engineering, University of Lagos 63 oafak@unilag.edu.ng 12/27/2012

For a second-order tensor field , show that div curl = curl div

Define the second order tensor as curl , = .


The gradient of is . , = , Clearly, div curl = , = , = , = curl div T
Dept of Systems Engineering, University of Lagos 64 oafak@unilag.edu.ng 12/27/2012

If the volume is enclosed by the surface , the position vector = and is the external unit normal to each surface element, show that equals the volume contained in . = = By the Divergence Theorem, =

, + ,

+ , =

2 ,

2 = 6

65 oafak@unilag.edu.ng 12/27/2012

Dept of Systems Engineering, University of Lagos

For any Euclidean coordinate system, show that div = curl curl Given the contravariant vector and with their associated vectors and , the contravariant component of the above cross product is .The required divergence is simply the contraction of the covariant derivative of this quantity: = , + , where we have treated the tensor as a constant under the covariant derivative. Cyclically rearranging the RHS we obtain, = , + , = , + ,
, ,

where we have used the anti-symmetric property of the tensor . The last expression shows clearly that div = curl curl as required.

Dept of Systems Engineering, University of Lagos

66

oafak@unilag.edu.ng 12/27/2012

Liouville Formula
For a scalar variable , if the tensor = () and , Show that det = det tr( ) Proof: Let 1 so that, = . In component form, we have = . Therefore, 1 2 3 3 3 3 det = = 1 2 + 1 2 + 1 2 3 3 = 1 2 + 1 2 + 1 2 3 =
3 1 1 + 1 2 + 1 3 2 1 2 3 3 3 2 3 1 2 + 1 1 1 + 2 2 + 2 3 + 1 2 1 + 3 + 3 2 3 2 3

All the boxed terms in the above equation vanish on account of the contraction of a symmetric tensor with an antisymmetric one.

Dept of Systems Engineering, University of Lagos

67

oafak@unilag.edu.ng 12/27/2012

Liouville Theorem Contd


3 (For example, the first boxed term yields, 1 2 2 2

Which is symmetric as well as antisymmetric in and . It therefore vanishes. The same is true for all other such terms.)
3 3 3 det = 1 1 2 + 1 2 2 + 1 2 3 1 2 3 3 = 1 2
= tr 1 det

as required.

Dept of Systems Engineering, University of Lagos

68

oafak@unilag.edu.ng 12/27/2012

For a general tensor field show that, = + + curl = , = . curl = , . so that curl = curl curl = , = , + = , + = , .. , + , , .. . . . + . , . , . = 2 tr div div + grad div + grad div 2 T
Dept of Systems Engineering, University of Lagos 69

grad grad tr
oafak@unilag.edu.ng 12/27/2012

When is symmetric, show that tr(curl ) vanishes. curl = , tr curl = , = , = , which obviously vanishes on account of the symmetry and antisymmetry in and . In this case, curl curl = 2 tr div div grad grad tr + 2 grad div 2

as grad div = grad div if the order of differentiation is immaterial and T is symmetric.
Dept of Systems Engineering, University of Lagos 70 oafak@unilag.edu.ng 12/27/2012

For a scalar function and a vector show that the divergence of the vector is equal to, grad + div , = , + ,i Hence the result.

Dept of Systems Engineering, University of Lagos

71

oafak@unilag.edu.ng 12/27/2012

Show that = + div div Taking the associated (covariant) vector of the expression for the cross product in the last example, it is straightforward to see that the LHS in indicial notation is,
,

Expanding in the usual way, noting the relation between the alternating tensors and the Kronecker deltas, = , ,
, = , , = , , + , , = , , + , , Which is the result we seek in indicial notation.
Dept of Systems Engineering, University of Lagos 72

= , , =

, ,

oafak@unilag.edu.ng 12/27/2012

Differentiation of Fields
Given a vector point function 1 , 2 , 3 and its covariant components, 1 , 2 , 3 , = 1,2,3, with as the reciprocal basis vectors, Let = , then = = + Clearly, = + And the projection of this quantity on the direction is, = + = + = +
Dept of Systems Engineering, University of Lagos 73 oafak@unilag.edu.ng 12/27/2012

Christoffel Symbols
Now, = so that + = 0. = . This important quantity, necessary to quantify the derivative of a tensor in general coordinates, is called the Christoffel Symbol of the second kind.

Dept of Systems Engineering, University of Lagos

74

oafak@unilag.edu.ng 12/27/2012

Derivatives in General Coordinates


Using this, we can now write that, = + = The quantity on the RHS is the component of the derivative of vector along the direction using covariant components. It is the covariant derivative of . Using contravariant components, we could write, = + = + So that, = + The components of this in the direction of can be obtained by taking a dot product as before: = = = + + +

Dept of Systems Engineering, University of Lagos

75

oafak@unilag.edu.ng 12/27/2012

Covariant Derivatives

The two results above are represented symbolically as, , = and , = + Which are the components of the covariant derivatives in terms of the covariant and contravariant components respectively. It now becomes useful to establish the fact that our definition of the Christoffel symbols here conforms to the definition you find in the books using the transformation rules to define the tensor quantities.
Dept of Systems Engineering, University of Lagos 76 oafak@unilag.edu.ng 12/27/2012

Christoffel Symbols
We observe that the derivative of the covariant basis, = , 2 2 = = = Taking the dot product with , 1 = + 2 1 = + 2 1 = + 2 1 = + 2 1 = + 2 Dept of Systems Engineering, University of Lagos 77 oafak@unilag.edu.ng 12/27/2012

The First Kind


Which is the quantity defined as the Christoffel symbols of the first kind in the textbooks. It is therefore possible for us to write, 1 , = = + 2 It should be emphasized that the Christoffel symbols, even though the play a critical role in several tensor relationships, are themselves NOT tensor quantities. (Prove this). However, notice their symmetry in the and . The extension of this definition to the Christoffel symbols of the second kind is immediate:
Dept of Systems Engineering, University of Lagos 78 oafak@unilag.edu.ng 12/27/2012

The Second Kind


Contract the above equation with the conjugate metric tensor, we have, , = = = = Which connects the common definition of the second Christoffel symbol with the one defined in the above derivation. The relationship, , = apart from defining the relationship between the Christoffel symbols of the first kind and that second kind, also highlights, once more, the index-raising property of the conjugate metric tensor. Dept of Systems Engineering, University of Lagos 79 oafak@unilag.edu.ng 12/27/2012

Two Christoffel Symbols Related


We contract the above equation with and obtain, , = , = , = so that, = , Showing that the metric tensor can be used to lower the contravariant index of the Christoffel symbol of the second kind to obtain the Christoffel symbol of the first kind.
Dept of Systems Engineering, University of Lagos 80 oafak@unilag.edu.ng 12/27/2012

Higher Order Tensors


We are now in a position to express the derivatives of higher order tensor fields in terms of the Christoffel symbols. For a second-order tensor , we can express the components in dyadic form along the product basis as follows: = = = . = . This is perfectly analogous to our expanding vectors in terms of basis and reciprocal bases. Derivatives of the tensor may therefore be expressible in any of these product bases. As an example, take the product covariant bases.
Dept of Systems Engineering, University of Lagos 81 oafak@unilag.edu.ng 12/27/2012

Higher Order Tensors


We have: = + + Recall that, = . It follows therefore that, = = = 0. Clearly, = (Obviously since is a basis vector it cannot vanish)
Dept of Systems Engineering, University of Lagos 82 oafak@unilag.edu.ng 12/27/2012

Higher Order Tensors


= + + = + + = + + = + + = , Where = + + or + + are the components of the covariant derivative of the tensor in terms of contravariant components on the product covariant bases as shown.
,
Dept of Systems Engineering, University of Lagos 83 oafak@unilag.edu.ng 12/27/2012

Higher Order Tensors


In the same way, by taking the tensor expression in the dyadic form of its contravariant product bases, we can write, = + + + + = Again, notice from previous derivation above, = so that, = = Therefore, + + = = + = = , So that , =
Dept of Systems Engineering, University of Lagos 84 oafak@unilag.edu.ng 12/27/2012

Higher Order Tensors


Two other expressions can be found for the covariant derivative components in terms of the mixed tensor components using the mixed product bases defined above. It is a good exercise to derive these. The formula for covariant differentiation of higher order tensors follow the same kind of logic as the above definitions. Each covariant index will produce an additional term similar to that in 3 with a dummy index supplanting the appropriate covariant index. In the same way, each contravariant index produces an additional term like that in 3 with a dummy index supplanting an appropriate contravariant index.
Dept of Systems Engineering, University of Lagos 85 oafak@unilag.edu.ng 12/27/2012

Higher Order Mixed Tensors


1 2 The covariant derivative of the mixed tensor, 1 ,2 ,, is the most general case for the covariant derivative of an absolute tensor:
1 2 1 ,2 ,,

, ,,

, ,,

1 2 1 ,2 ,,

, ,,

1 ,2 ,, 1 ,2 ,, ,2 ,, 1 ,,, 1 2 , ,, 1 ,,, + 1 1 ,2 ,, + 2 1 ,2 ,, + + 2

1 ,2 ,, 1 ,2 ,, 1 ,2,, 1 ,2 ,,

Dept of Systems Engineering, University of Lagos

86

oafak@unilag.edu.ng 12/27/2012

Riccis Theorem

The metric tensor components behave like constants under a covariant differentiation. The proof that this is so is due to Ricci: , = = , , 1 1 = + + 2 2 = 0.

Dept of Systems Engineering, University of Lagos

87

oafak@unilag.edu.ng 12/27/2012

Riccis Theorem
The conjugate metric tensor behaves the same way as can be seen from the relationship, = The above can be differentiated covariantly with respect to to obtain , + , = , , = 0 + + , = 0 + =0 The contraction of with , vanishes. Since we know that the metric tensor cannot vanish in general, we can only conclude that , = 0
Dept of Systems Engineering, University of Lagos 88 oafak@unilag.edu.ng 12/27/2012

Which is the second part of the Ricci theorem. With these two, we can treat the metric tensor as well as its conjugate as constants in covariant differentiation. Notice that along with this proof, we also obtained the result that the Kronecker Delta vanishes under a partial derivative (an obvious fact since it is a constant), as well as under the covariant derivative. We summarize these results as follows: The metric tensors and as well as the alternating tensors , are all constants under a covariant differention. The Kronecker delta is a constant under both covariant as well as the regular partial differentiation
Dept of Systems Engineering, University of Lagos 89 oafak@unilag.edu.ng 12/27/2012

Integral Theorems

The divergence theorem is central to several other results in Continuum Mechanics. We present here a generalized form [Ogden] which states that, Gauss Divergence Theorem For a tensor field , The volume integral in the region , grad =

where is the outward drawn normal to the boundary of .


Dept of Systems Engineering, University of Lagos 90 oafak@unilag.edu.ng 12/27/2012

Special Cases: Vector Field

Vector field. Replacing the tensor with the vector field and contracting, we have, div =

Which is the usual form of the Gauss theorem.

Dept of Systems Engineering, University of Lagos

91

oafak@unilag.edu.ng 12/27/2012

Scalar Field

For a scalar field , the divergence becomes a gradient and the scalar product on the RHS becomes a simple multiplication. Hence the divergence theorem becomes, grad =

The procedure here is valid and will become obvious if we write, = where is an arbitrary constant vector.
Dept of Systems Engineering, University of Lagos 92 oafak@unilag.edu.ng 12/27/2012

div =

For the LHS, note that, div = tr grad grad = , = , The trace of which is, , = , = , = grad For the arbitrary constant vector , we therefore have that, div =

grad =

grad =

Dept of Systems Engineering, University of Lagos 93

oafak@unilag.edu.ng 12/27/2012

Second-Order Tensor field


For a second-order tensor , the Gauss Theorem becomes,
div =

The original outer product under the integral can be expressed in dyadic form: grad =


94 oafak@unilag.edu.ng 12/27/2012

Dept of Systems Engineering, University of Lagos

Second-Order Tensor field


Or

, =
, =

Contracting, we have

, =

, =

Which is the same as,

div =

Dept of Systems Engineering, University of Lagos 95

oafak@unilag.edu.ng 12/27/2012

Physical Components of Tensors


The components of a vector or tensor in a Cartesian system are projections of the vector on directions that have no dimensions and a value of unity. These components therefore have the same units as the vectors themselves. It is natural therefore to expect that the components of a tensor have the same dimensions. In general, this is not so. In curvilinear coordinates, components of tensors do not necessarily have a direct physical meaning. This comes from the fact that base vectors are not guaranteed to have unit values ( 1 in general).
Dept of Systems Engineering, University of Lagos 96 oafak@unilag.edu.ng 12/27/2012

Physical Components
They may not be dimensionless. For example, in orthogonal spherical polar, the base vectors are 1 , 2 and 3 . These can be expressed in terms of dimensionless unit vectors as, , sin , and since the magnitudes of the basis vectors are , sin , and 1 or 11 , 22 , 33 respectively. As an example consider a force with the contravariant components 1 , 2 and 3 , = 1 1 + 2 2 + 3 3 = 1 + sin 2 + 3
Dept of Systems Engineering, University of Lagos 97 oafak@unilag.edu.ng 12/27/2012

Physical Components
Which may also be expressed in terms of physical components, = + + . While these physical components { , , } have the dimensions of force, for the contravariant components normalized by the in terms of the unit vectors along these axes to be consistent, {1 , sin 2 , 3 } must each be in the units of a force. Hence, 1 , 2 and 3 may not themselves be in force units. The consistency requirement implies, = 1 , = sin 2 , and = 3
Dept of Systems Engineering, University of Lagos 98 oafak@unilag.edu.ng 12/27/2012

Physical Components
For the same reasons, if we had used covariant components, the relationships, 1 2 = , = , and = 3 sin The magnitudes of the reciprocal base vectors are 1 1 , , and 1. While the physical components have the
sin

dimensions of force, 1 = and 2 = sin have the dimensions of moment, while 1 =


and 2 =

sin

are

in dimensions of force per unit length. Only the third components in both cases are given in the dimensions of force.
Dept of Systems Engineering, University of Lagos 99 oafak@unilag.edu.ng 12/27/2012

The physical components of a vector or tensor are components that have physically meaningful units and magnitudes. Often it is convenient to derive the governing equations for a problem in terms of the tensor components but to solve the problem in physical components. In an orthogonal system of coordinates, to obtain a physical component from a tensor component we must divide by the magnitude of the relevant coordinate for each covariant index and multiply by each contravariant index. To illustrate this point, consider the evaluation of the physical components of the symmetric tensor components or or in spherical polar coordinates. Here, as we have seen, the magnitudes of the base vectors 11 , 22 , 33 or 1 , 2 and 3 are , sin and 1. Using the rule specified above, the table below computes the physical components from the three associated tensors as follows: Dept of Systems Engineering, University of Lagos 100 oafak@unilag.edu.ng 12/27/2012

Physical Component ()

Contravariant 11 1 1 = 11 2

Covariant 11 11 = 2 1 1 12 12 = 2 1 1 sin 22 22 = 2 2 2 2 sin 23 23 = 2 3 sin

Mixed
1 1 1 1 = 1 1 1 1 2 2 = 2 1 sin 2 2 2 2 = 2 2 2 3 2 2 = 3 sin 3

()

12 1 2 = 12 2 sin

()

22 2 2 = 22 2 sin2

()

23 2 3 = 23 sin

()

33 3 3 = 33

33 = 33 3 3
31 31 = 3 1
101

3 3 3 3 = 3 3
3 3 1 1 = 1 3
oafak@unilag.edu.ng 12/27/2012

()

31 3 1 = 31

Dept of Systems Engineering, University of Lagos

17. The transformation equations from the Cartesian to the oblate spheroidal coordinates , and are: = , = , and = , where f is a constant representing the half the distance between the foci of a family of confocal ellipses. Find the components of the metric tensor in this system. The metric tensor components are: 2 2 2 = + + 1 2 2 2 2 2 2 2 2 = sin + + cos = 2 1 2 1 =
2

+
2

+
2

=
2

2 2 1 2

= + + = 2 = + + 2 1 = sin sin 1 2 1 2 2 + cos cos 1

= 0 = =

Dept of Systems Engineering, University of Lagos

102

oafak@unilag.edu.ng 12/27/2012

Find an expression for the divergence of a vector in orthogonal curvilinear coordinates. The gradient of a vector = is = = , . The divergence is the contraction of the gradient. While we may use this to evaluate the divergence directly it is often easier to use the computation formula in equation Ex 15: 1 = , = , = 1 1 2 = 1 2 3 + 2 1 2 3 + 3 1 2 3 3 1 1 2 3 Recall that the physical (components having the same units as the tensor in question) components of a contravariant tensor are not equal to the tensor components unless the coordinate system is Cartesian. The physical component () = (no sum on i). In terms of the physical components therefore, the divergence becomes, 1 = 2 3 (1) + 2 1 3 (2) + 3 1 2 (3) 1 2 3 1

Dept of Systems Engineering, University of Lagos

103

oafak@unilag.edu.ng 12/27/2012

Find an expression for the Laplacian operator in Orthogonal coordinates. For the contravariant component of a vector, , 1 = , . Now the contravariant component of gradient = , . Using this in place of the vector , we can write, 1 , , = given scalar , the Laplacian 2 is defined as, , so that, 1 2 = , = When coordinates are orthogonal, = = 0 whenever . Expanding the computation formula therefore, we can write, 1 1 2 3 1 2 3 1 2 3 2 = + 2 + 3 1 2 3 1 1 1 2 2 3 3 1 = 2 3 1 + 2 1 3 2 + 3 1 2 3 1 2 3 1

Dept of Systems Engineering, University of Lagos

104

oafak@unilag.edu.ng 12/27/2012

Show that the oblate spheroidal coordinate systems are orthogonal. Find an expression for the Laplacian of a scalar function in this system. Example above shows that = = = 0 . This is the required proof of orthogonality. Using the computation formula in example 11, we may write for the oblate spheroidal coordinates that, 2 = 2 1 1 2 3 2 2 2 2 1 1 2

1 2 + 2 1 +
Dept of Systems Engineering, University of Lagos

2 2 2 1 1 2
105


oafak@unilag.edu.ng 12/27/2012

For a vector field , show that grad is a third ranked tensor. Hence or otherwise show that div = curl . The secondorder tensor is defined as . Taking the covariant derivative with an independent base, we have grad = , This gives a third order tensor as we have seen. Contracting on the last two bases, div = , = , = , = curl
Dept of Systems Engineering, University of Lagos 106 oafak@unilag.edu.ng 12/27/2012

Coordinate transformations
Begin with our familiar Cartesian system of coordinates. We can represent the position of a point (position vector) with three coordinates , , ( R) such that, = + + That is, the choice of any three scalars can be used to locate a point. We now introduce a transformation (called a polar transformation) of , {, } such that, = cos , and = sin . Note also that this transformation is invertible: = 2 + 2 ,and 1 = tan

Dept of Systems Engineering, University of Lagos

107

oafak@unilag.edu.ng 12/27/2012

Curvilinear Coordinates
With such a transformation, we can locate any point in the 3-D space with three scalars {, , } instead of our previous set , , . Our position vector is now, = cos + sin + = + where we define cos + sin , is no different from . In order to complete our triad of basis vectors, we need a third vector, . In selecting , we want it to be such that , , can form an orthonormal basis.

Dept of Systems Engineering, University of Lagos

108

oafak@unilag.edu.ng 12/27/2012

Cylindrical Polar coordinate system


Let

= + To satisfy our conditions,


= 0, = 0, and 2 + 2 = 1. It is easy to see that sin + cos satisfies these requirements. , , forms an orthonormal (that is, each member has unit magnitude and they are pairwise orthogonal) triad just like , , . The transformation we have just described can be given a geometric interpretation. In either case, it is the definition of the Cylindrical Polar coordinate system.
Dept of Systems Engineering, University of Lagos 109 oafak@unilag.edu.ng 12/27/2012

Unlike our Cartesian system, we note that (), (), as the first two of these are not constants but vary with angular orientation. remains a constant vector as in the Cartesian case.

Dept of Systems Engineering, University of Lagos

110

oafak@unilag.edu.ng 12/27/2012

Spherical Polar
Continuing further with our transformation, we may again introduce two new scalars such that , {, } in such a way that the position vector, = + = sin + cos Here, = sin , = cos . As before, we can use three scalars, , , instead of {, , }. In comparison to the original Cartesian system we began with, we have that, = + + = sin + cos = sin cos + sin + cos = sin cos + sin sin + cos from which it is clear that the unit vector sin cos + sin sin + cos .
111

Dept of Systems Engineering, University of Lagos

oafak@unilag.edu.ng 12/27/2012

Spherical Polar
Again, we introduce the unit vector, cos cos + cos sin sin and retain = sin + cos as before. It is easy to demonstrate the fact that these vectors constitute another orthonormal set. Combining the two transformations, we can move from , , system of coordinates to , , directly by the transformation equations, = sin cos , = sin sin and = cos . The orthonormal set of basis for the , , system is , , , , . This is the Spherical Polar Coordinate System.
Dept of Systems Engineering, University of Lagos 112 oafak@unilag.edu.ng 12/27/2012

Variable Bases

There two main points to note in transforming from Cartesian to Cylindrical or Spherical polar coordinate systems. The latter two (also called curvilinear systems) have unit basis vector sets that are dependent on location. Explicitly, we may write, , , = cos + sin (, , ) = sin + cos , , =
Dept of Systems Engineering, University of Lagos 113 oafak@unilag.edu.ng 12/27/2012

For Cylindrical Polar, and for Spherical Polar, (, , ) = sin cos + sin sin + cos , , = cos cos + cos sin sin , , = sin + cos . Of course, there are specific cases in which some of these basis vectors are constants as we can see. It is instructive to note that curvilinear (so called because coordinate lines are now curves rather than straight lines) coordinates generally have basis vectors that depend on the coordinate variables. Unlike the Cartesian system, we will no longer be able to assume that the derivatives of the basis vectors vanish.
Dept of Systems Engineering, University of Lagos 114 oafak@unilag.edu.ng 12/27/2012

The second point to note is more subtle. While it is true that a vector referred to a curvilinear basis will have coordinate projections in each of the basis so that, a typical vector = 1 1 + 2 2 + 3 3 when referred to the basis vectors {1 , 2 , 3 }, for position vectors, in order that the transformation refers to the same position vector we started with in the Cartesian system
Dept of Systems Engineering, University of Lagos 115 oafak@unilag.edu.ng 12/27/2012

We have for Cylindrical Polar, the vector, (, , ) = () + and for Spherical Polar, (, , ) = (, )

Expressing position vectors in Curvilinear coordinates must be done carefully. We do well to take into consideration the fact that in curvilinear systems, the basis vectors are not fully specified until they are expressly specified with their locational dependencies.
Dept of Systems Engineering, University of Lagos 116 oafak@unilag.edu.ng 12/27/2012

It is only in this situation that we can express the position vector of a specific location Spherical and Cylindrical Polar coordinates are two more commonly used curvilinear systems. Others will be introduced as necessary in due course.

Dept of Systems Engineering, University of Lagos

117

oafak@unilag.edu.ng 12/27/2012

Natural Bases
Beginning from the position vector, given a system with coordinate variables 1 , 2 , 3 , it is easy to prove that the set of vectors, 1 , 2 , 3 1 , 2 , 3 1 , 2 , 3 , , 1 1 1 Which we can write more compactly as , = 1,2,3 constitute an independent set. This set, with its dual set of vectors, such that = , constitute the Natural bases for the coordinate system.
Dept of Systems Engineering, University of Lagos 118 oafak@unilag.edu.ng 12/27/2012

Orthonormality of Natural Bases


We can only guarantee the independence of the bases for any arbitrary system. They may NOT be normalized neither are they guaranteed of mutual orthogonality. HW. Show that the natural bases for Spherical and Cylindrical polar systems are orthogonal but not normalized. And that the dual natural bases for the Cartesian system coincide.

Dept of Systems Engineering, University of Lagos

119

oafak@unilag.edu.ng 12/27/2012