Mechanics of Solids - Review

Index notation and Cartesian Tensors

• Index notation, Summation convention

• Orthogonal Transformation
• Scalars, Vectors, Tensors
• Tensor Properties and Operations
• Gauss theorem

From the book: Mechanics of Continuous Media: an Introduction

J Botsis and M Deville, PPUR 2018.
Solutions: https://www.epflpress.org/produit/908/9782889152810/mechanics-of-continuous-media
 Continuum mechanics review (material length scales)
Length Material Length Material
Scale (mm) Morphology Scale (mm) Morphology
-10
10  -5
10 Piles of Dislocations
Atomic Level - Interphases

- Interfaces
...
... .

-9 
10 
.
.
...... ...
-4
Amorphous 10
Crystal Structure Structure . ....
-8    
10           -3 Grain Boundaries
      
    10 Composites
        

     
-7
10 Subgrain Domains Porous Structure -2
10
Dislocations
    
-6 
 
 Continuum Level
10 -1
 
  10

-5
10 0
10
 Continuum mechanics review
Experimental Evidence (Fluid)

Mass Density
h
Mass Continuum
Density

M
ρh = 3
h
We ignore the discrete aspect of matter and we consider that
properties such as density, viscosity, modulus of elasticity, etc.,
assigned to a midpoint continuous are continuous functions
of space variables.
 Continuum mechanics review

Experimental Evidence (Solid)

Mass Density
h
h
Mass Continuum
Density

M
ρh = 3
h
We ignore the discrete aspect of matter and we consider that
properties such as density, viscosity, modulus of elasticity, etc.,
assigned to a midpoint continuous are continuous functions
of space variables.
Continuum mechanics review-Vector Algebra

In mechanics of continuous media, motion and the associated physical quantities are
described in Euclidean space R3 (physical space) with which a three-dimensional vector
space E3 is associated.

The elements of R3 and E3 are called the points and vectors, respectively.

The scalars, vectors, and tensors that describe the physical quantities are also attached to a
space (typically R3) and form what are called scalar, vector, or tensor fields.
Continuum mechanics review-Vector Algebra
A vector space is defined uniquely from the properties of operations on its elements and assumes
the existence of an arbitrary field (typically the field of real numbers R) whose elements are called
scalars. The vector space E3 is then the set of elements denoted u, v, w , . . . such that:

for every u, v, w ∈ E3 and a, b ∈ R.

By providing E3 with a scalar product, to be able to calculate lengths and angles,

it takes the name Euclidean space. The scalar product associates
with every pair of vectors u, v ∈ E3 a scalar denoted u.v with the following properties:

·
for every u, v, w ∈ E3 and α, β ∈ R. Note: numbers refer to the equations
in the book by Botsis & Deville
 Continuum mechanics review-Vector Algebra
The scalar product is consequently an application of E3 x E3 in R that is linear with respect to each of
its arguments. It is also called a positive definite bilinear form (See (1.2) in the previous slide).
The definition of the vector norm u, denoted is given by the relation:

The vector u is called a unit vector when

·
= 1, and two vectors u and v are orthogonal if and only if u.v = 0.

Every vector in E3 can be decomposed uniquely according to a basis formed of three linearly independent vectors of
E3. The choice of a basis is arbitrary but generally one uses the canonical basis (e1, e2, e3) defined by:

The basis is called orthogonal when the basis vectors are not unit vectors but are still orthogonal.
·
Continuum mechanics review-Vector Algebra
Physical quantities in continuum mechanics
Scalars: Quantities for which only one value can be associated. For example, the mass density of a material. We
denote it ρ and it has for SI units kg/m3 or dimensions ML−3 where M is the mass and L a length. This density
is practically constant and, in addition, there is no direction associated with its value.
Vectors: quantities have not only a value but also a direction.
A force of one Newton is that which, applied to a point, gives it an acceleration of 1 m s−2 per kg.
Since this force has a direction, it is a vector. In a given coordinate system, this vector is specified by its components.
Going from one set of axes to another, the vector remains invariant and only the components of the vector
change by a transformation rule.
·

Tensors: We introduce the concept of a tensor in a simplistic way as follows. Consider, a stress is a force
per unit surface: (a) force is a vector and (b) an element of a surface is also a vector (must specify both
its size and its orientation).
If f describes the force vector and s the normal vector of the surface S, then we might think that the
stress T could be expressed by f /s. But, as the division of two vectors is an undefined operation, we get
around the difficulty by saying that given s, we can find f by multiplying s by a new entity T such that:
This mathematical object is a tensor which yields the stress at a point and is a tensor
of order 2. It is associated with two spatial directions and can be represented by a
matrix with two indices, each index· corresponding to one direction in Euclidean space.
It is thus an entity with nine components.
Continuum mechanics review-Orthogonal Transformation
In the physical Euclidean space R3, let there be a Cartesian orthonormal co- ordinate system (O, e1, e2, e3),
that we denote also as Oxi (i = 1, 2, 3), with origin at O and the unit vectors ei (i = 1, 2, 3) directed along
the axes Oxi (fig. 1.3).

Another system O xi’ (i = 1, 2, 3) with unit vectors ei’ defines a Cartesian coordinate system with the same origin O. The
direction cosines of the axes xi’ with respect to the axes xi, denoted by cpi, are given by the scalar products of the basis
vectors.

Similarly, the direction cosines of the first system with respect to the
second are given by:

Two Cartesian coordinate systems

with the same origin
 Continuum mechanics review-Orthogonal Transformation

Let P be a point with coordinates xi in the first system and xi’ in the second.

. From equation (1.6), the coordinates xi’ are related to those of xi and xi’

Two Cartesian
Using equation (1.6)
coordinate systems

Or

For the inverse

transformation we have:

Or
Summation convention
We can suppress the symbol Σ by adopting, from here on, the Einstein summation convention for
repeated indices to write for the coordinates of point P and agree to the following:

When an index appears twice in a product, a sum with respect to that index is implied by taking
successively all its possible values (in this case, i = 1, 2, 3). In this way equations (1.9) and (1.11),

are written in the compact form:

Summation convention
Illustrations of summation convention

(the index i is fixed and has a value among 1, 2, 3. It’s called free index.
We sum over the repeated index).

(No free index. We sum over the two repeated indices i and j =1, 2, 3)

ui designated the set of the 31 quantities

u1, u2, u3 (3 for space and 1 for the free index).
Lij signifies the set of the 32 quantities
L11, L12, L13, L21, L22, L23, L31, L32, L33
(3 for space and 2 for the free indices).
Lii, we have 30 = 1 quantity.
Summation convention and Kronecker delta
Illustrations of summation convention
Note that the index over which we sum is a dummy index;
we can change the notation of this index without changing the significance of the sum:

Very important in operations with complex equations

A dummy index does not appear

more than twice in a product.
Using this property and

(1.12)
(2) (1)
we can wrirte: We can introduce the Kronecker delta to reflect such a property.
(3)
from which: The coefficients in the last equations should be equal to one
when i=q and 0 for i not q. This is also true for the other equation.
Summation convention, Kronecker delta
Orthogonality conditions: From the Kronecker delta The Kronecker delta can be used to change
the index of a component:
(very useful property)

and

we can wrirte the orthogonality condition:

The components form an orthogonal matrix such that:

Note also the following relations usefull in

deriving relations in mechanics:
with
∂Aik
; = δ ijδ kl
∂Ajl

((+) indicates a direct rotation and (-) a reflection)

Scalars
Recall: In mechanics of continuous media, we work with scalars, vectors, and tensors that describe the physical
quantities that are also attached to a space (typically R3) and form what are called scalar, vector, or tensor fields.
What is important in the context of continuous media is how we can describe these quantities in different
coordinate systems.
Consider a point P in a continuous medium and a real value continuous function F (P) at P.
If the value F (P) does not depend on the coordinate system, then F is called a scalar function, or scalar,
or a tensor of order 0. This is the case, for example, for temperature, pressure, kinetic energy, etc.

Suppose that P has coordinates xi and if F (P) has a value f (xi),

Two Cartesian the change of coordinate systems:
coordinate systems

for the scalar F (P) leads to

Note that the values remains the same but the

form of the function can varie in the new coordinate system.
 Vectors
Let v = P Q be a vector having its origin at point P and its extremity at point Q (fig. 1.4). This vector has a direction
and three components vi. The vector itself is independent of the coordinate system.

Here xi, yi are the coordinates of the points P and Q in the first coordinate system and
xi’, yi’ in the second. The components of v in the two systems are written as:

(1.19)
(1.20)

In a Cartesian coordinate system the are Independent of the coordinates of P.

(1.19)
Vector in two Cartesian (1.21)
coordinate systems
The object v, characterized by the three components vi in a
Cartesian coordinate system, is a vector or a tensor of order 1 if
(1.20), (1.21)
its components are transformed according to this rule during a
coordinate system change that is an orthogonal transformation
(1.22) satisfying (1.16).
Vectors
Using index notation we can express the algebra of vectors and their components:

With a being a scalar we can write, using (1.22): Scalar product of two vectors b:
In symbolic form :
In index form :
The scalar product of two vectors is independent of
the vector addition results in a vector with components: an orthogonal change of coordinates.
Using (1.20), we have:

or in vector notation :

With and being the norms of the

two vectors we have :

is the angle between the two vectors).

For a single vector:
Permutation symbol and vector product
The permutation symbol is defined as:

or

- 1 +

3 2
Permutation symbol and vector product

With the Kronecker symbol

and the permutation symbol we have

an important relation used to easily demonstrate vector identities
and express vector products in index form.

In an orthonormal basis of R3, the vector product of two vectors w = u × v,

sometimes denoted u ∧ v, is defined by the equality:

The norm of the vector is given by the equality

(with angle between u and v):
Permutation symbol and vector product

Example 1.1
Permutation symbol and vector product
Example 1.2: Verify the following identity:
(1.36)

(1.30)
Tensor Algebra: definition of a tensor
Definition of a tensor of order 2:

Let E3 be the Euclidean vector space of vectors associated with R3, and L a linear mapping on E3
that transforms a vector to another:

such that:

If L transforms the two arbitrary vectors as

and has the properties

(1.38)

where u1 and u2 are two arbitrary vectors of E3 and α ∈ R, then we say that L is a linear transformation.
It is also a tensor of order 2, or simply, a tensor.

The unit tensor I and the zero tensor O are defined by the relations u = Iu and 0 = Ou.
Tensor Algebra: definition of a second order tensor
For a vector u, the vector v is given by: (1.39)
If we express the components of v as: (1.40)

(1.41)
(1.42)
The components of the tensor are:

or (1.43)
matrix associated
which is the transformation of the two vectors with the tensor
in index form. In a matrix from, it is.
determinant associated
with the tensor

From (1.42) we see that the columns

of the matrix are the components of vectors
and depend on the coordinate system. However,
the operator does not depend on the base vector.
Tensor Algebra: Dyadic (or tensor) product of two vectors
The tensor product or dyadic product a ⊗ b of two vectors a and b is defined as the tensor which,
for any vector v defines the following transformation :

(1.48)

For every vector v and w and for α, β ∈ R, we have

The matrix form of a dyadic product is:

The definition of the dyadic product and the last relation

demonstrates that the product is a tensor. Its components are:
(1.42)
Tensor Algebra: Dyadic (or tensor) product of two vectors

For two vectors u and v, (1.43)

Note that :

With the help of (1.43) we obtain:

Note two important relations:
(1.43)

and
Tensor Algebra: transformation rule for Cartesian tensor
The representation in the Cartesian coordinate system xi of the linear operator L,

which is invariant, is given by its components Lij (1.42)

In the coordinate system x’i the components of are expressed as

We can easily evaluate the relation between the components Lij and L’ij. Using (1.20),

becomes (1.52)

Recalling, (1.21)

(1.53)

In matrix notation, equation (1.52) is written as:

Tensor Algebra: transformation rule for Cartesian tensor
Definition of second order tensor: a matrix [L] with nine components corresponds to a 2nd
order tensor if its components are transformed according to (1.53):

(1.53) These transformation rules guarantee

the invariance of L with respect to
during a coordinate change that obeys (1.20): the choice of coordinates.

(1.20)
and that is an orthogonal transformation according to (1.16):

(1.16)

By generalizing these transformation rules, we can define a tensor of order n.

By definition, T is a tensor of order n if, during a coordinate transformation,
its components aretransformed according to the rule: For a tensor of order 1 (vector)
and order 2 we easily see:

(1.55)
Tensor Algebra
Important rules of nth order tensors
Multiplication by a Scalar: the multiplication of a tensor of order n by a scalar is carried out
by multiplying each component of the tensor by the scalar. The result is a tensor of order n.

Linear Combination: the linear combination of two tensors of order n is by linear combination of the corresponding
components. A tensor of the same order is obtained.

Zero Tensor: it is the tensor for which all the components are equal to zero.

Equivalent Tensors: when the components of two tensors of the same order are equal term by term in a
coordinate system, then they are equal in every other system; the tensors are equivalent.
Consequently, if a tensor relation is verified in one coordinate system, it is true in all coordinate systems.

Exterior Product of Tensors: Consider Ai1 ···in and Bj1 ···jm as the respective components of a tensor of order
n and a tensor of order m in a coordinate system. The 3n+m quantities obtained by:
C i1 ···inj1 ···jm = Ai1 ···in Bj1 ···jm form a tensor C of order n + m.
An example, is the dyadic product of two vectors (i.e. tensors of order 1) yields a tensor of order 2.
Tensor Algebra
Important rules of n order tensors
Tensor Contraction: Consider a tensor A of order n whose components in a coordinate system are Ai1 ···in .
The contraction consists of setting equal two indices of the tensor, i.e. the jth and the kth with j and k ≤ n, and
summing over these indices (j, k = 1, 2, 3) to form a tensor of order n-2 thus having 3n−2 components.
We say that this tensor is obtained by contraction of the indices j and k.
Example, Lii is the only contraction possible of Lij which is a scalar (tensor of order 0).

Consider two tensors S and T of order 2. Their exterior product results in a tensor R of order 4 with components:

. The components obtained by contraction of the second and third indices of R are:

. .. We can show that this is a 2nd order 2.

From transformation (1.55), we can write:

contraction

i.e., from a 4th order

to a 2nd order tensor
Tensor Algebra
(we consider tensors of order 2)

Sum of Tensors
Consider two tesnors L and T. Their sum (L+T) is such that for every vector a,

with components:

Interior Product of Two Tensors In matrix form, the matrix of the interior
product is equal to the product of the
Consider two tesnors L and T. For or every vector a, their products LT and TL
matrices of the two tensors:
are given by the equations:

and
and
The components are:
Tensor Algebra
Interior Product of Two Tensors Useful relations of interior product:

Note that (not communicative).

For three tensors , L, T and S, we can write:

Example 1.5

and The first one is easily demonstrated as follows:

(it is associative)

When L= T

TT =T2, TT2 =T3…..

Note also:
Tensor Properties
Transpose of a Tensor: Inverse of a Tensor:

It is obtained by exchanging two indices. For a tesnor with det L ≠ 0, there exist a unique tensor
The transpose of Lij is Lji . called inverse tensor and denoted L-1 of L and satisfies:

The transpose of the tesnor L is denoted as LT

and We can easiliy show that:

Also we can easily show that:

Note that: For two tensors:

Tensor Properties
Symmetric Tensor: Trace of a Tensor:
It is defined when the tensor is equal to its The trace of a 2nd order tensor L, is the sum (scalar)
transpose: of its diagonal elements and denoted by ‘tr’

Or Lij = Lji
Note that a symmetric tensor has 6
independent components.
For a dyad of vectors a and b, it is their scalar product
antisymmetric Tensor:

A tensor L is said to be antisymmetric if

Some properties of the trace (α ∈ R )
Or Lij = −Lji

It can be proven that all 2nd tensors can be

uniquely decomposed into the sum of
symmetric LS and antisymmetric LA tensors:
Tensor Properties
Deviatoric Tensor: Orthogonal Tensor:

A tensor L can be decomposed into a spherical For a tesnor Q, that satisfies the condition:
tensor LS and a tensor with zero trace Ld, called
deviatoiric tensor, so that for every u and v.

Using relation
where we obtain:

An orthogonal tensor satisfies

Note that since u.v is preserved, the angle between the
Note that the components of the deviatoric vectors as well as their norms are preserved.
tensor are not independent.
Note that the matrix of Q is equal to the matrix [C] of a
rotation of the basis vectors.
Tensor Properties
Scalar Product of two tensors: Example 1.6:

For two tensors S, T of odrer two the following Verify the last identity. Following the definition we obtain:
scalar is defined as scalar product:

It is a double tensor contraction. Example 1.7:

Consider two tensors A, B such that Aij =Aji (symmetric)
The norm of a tensor L is defined as: and Bij = − B ji (antisymmetric). Show that their
scalar product is zero (use the properties of symmetric and
antisymmetric tensors).

Properties of scalar product:

Dual Vector of a 2nd order tensor
The dual vector components di of a tensot L The dual vector has zero components if L is symmetric:
are defined by the product:
( L =L )
ij ji

This is shown as follows. From the definition of the vector

and spliting the tensor in two parts we can write:

Explicitly they are:

The first term in the parentehsis is zero (product of
symmetric and antisymmetric tensors) and the antisymmetric
part is zero due to symmetry.

For any tensor L (using the definiton) the dual vector depends on
the antisymmetric part:

Further it can be shown that:

Eigenvalues and Eigenvectors of a tensor
For a tensot L if u is a vector that when L is applied The last equation is called characteristic equation of the
is transformed into avector parralell to itself: tesnor L. Its solution gives the eigenvalues and eigenvectors.

( )
For a symmetric tensor Lij =L ji the following theorem is
Then u is an eigenvector of L and λ is the holds for its matrix (from linear algebra):
corresponding eigenvalue. Conventionally the
eigenvectors are normalized to vectors n Theorem: The eigenvalues of a real nxn symmetric matrix
of unit length (unit eigenvector): are all real. The corresponding eigenvectors are orthogonal.

The solution of the characteristic equations results in the

Or with eigenvalues λ1, λ2, λ3 & corresponding eigenvectors n1, n2, n3 .

Note that
1: If λ1 ≠ λ2 then n1 . n2 = 0 and n1, n2 are orthogonal.
2: If λ1 = λ2 ≠ λ3 we have n1 . n3 = n2 . n3 =0. In such case
directions n1, n2, are chose mutually orthogonal and normal to
n3 .
3: λ1 = λ2 = λ3 the directions n1, n2, n3 are chosen mutually
orthogonal and without restriction.
Eigenvalues and Eigenvectors of a tensor
The characteristic equation:
Note that any independent combination of these
invariants results in another invariant.
is a third order polynomial:
Example 1.9

The expression
Its solution gives the eigenvalues and eigenvectors:
is an invariant of the tensor L.
The following three parameters are called invariants
Use the identity
of the tensor L:

to modify the expression as follows:

Eigenvalues and Eigenvectors of a tensor
Positive definite tensor Spectral decomposition of a tensor or
spectral representation of a tensor
Such a tensor satifies the following relation
For a tensor L with eigenvalues λ1, λ2, λ3, and
corresponding eigenvectors n1, n2, n3 .
It can be shown that the eigenvalues of a positive The orthogonal eigenvectors form a basis for the spectral
Definite tensor are all positive: decompostion writen as follows:

For the tensor L with one of its eigenvalue λ and

and corresponding eigenvector n, we can easily
see that since
We can show it as follows:

use and
3
L=L(ni ⊗ ni=
) ( Lni ) ⊗ n=
i (λi ni ) ⊗ n=
i ∑ λ (n ⊗ n )
1
i i i

(1.65) (1.109)
Square root of a tensor & Polar decomposition
Theorem (square root)
Theorem (polar decomposition)
For a symmetric ,positiove definite tensor C with
eigenvalues λi2 and corresponding eigenvectors nι , For a tensor F with with determinant det F >0 there
there is a symmetric positive definite tensor U exist symmetric positive definite tensors U and V
such that: and a rotation (an orthogonal tensor with a positive
Determinant equal to 1) R such that:
U2 = C
F = RU = VR
and denote it as C =U
These decompositions are unique and we have:
These two tensors have the following spectral forms:
U = F T F and V = FF T

Representation F = RU is called right decomposition.

Representation F = VR is called left decomposition.
Functions of a tensor
Isotropic tensor function of a symmetric tensor Scalar function of a tensor
The function is defined as a scalar funtion
By definition an tensor isotropic function f, for which of the tensor T and yield a scalar. When T is symmetric
the variable is a 2nd order symmetric tensor T , satisfies and the condition:
the identity:

is satisfied, then is isotropic of T and is

for any orthogonal tensor Q . represenetd by:
For a symmetric tensor L the following relation is true:

where the parameters

Rivlin-Ericksen Theorem are the invariants of T. This is also equivalent to:
The last expression can be written in the form

where λ1, λ2, λ3, are the eirenvalues of T.

It can be shown that for the isotropic function
its derivative with respct to T is:
Tensor Analysis
Notation: Derivatives:
For a tensor the derivative with respect to a
scalar parameter (i.e., time) is a tensor of the same order
Scalar field
F(xi, t) F(x1,x2 ,x3 ,t)

Vector field In terms of its components, it is given by

vi (xm ,t) : it covers all three components

v1( x1,x2 ,x3 ,t), v2 (x1,x2 ,x3 ,t), v3 (x1,x2 ,x3 ,t) .
For a vector the first and second time derivatives are
Tensor field
Lij (xm ,t) : it covers all nine components and

L11 (x1,x2 ,x3 ,t), L12 (x1,x2 ,x3 ,t), …., L33 (x1,x2 ,x3 ,t),
Tensor Analysis
Derivatives: Derivatives

The following identites are establied easily Demonstration of

From the defintion

Tensor Analysis
Gradient of a scalar field Gradient of a vector field

Associated with a scalar field is a vector field With a vector field , we associate a tensor,
called the gradient of . called the gradient of , and denote it It is a
tensor of order 2 which, applied to , gives the
It is denoted or and is such that the scalar difference of between and .
product with gives the difference between the
values of evaluated at and at . We have
we obtain
With ( )

With ( )) We obtain

in direction 1

In Cartesian coordinates we have:

Tensor Analysis
Gradient of a scalar valued tensor function Gradient of a tensor valued tensor function

For a regular, scalar valued, smooth function For a regular, tensor valued, smooth function of
of a tensor of order 2, the first two terms of a Taylor a tensor of order 2, the first two terms of a Taylor
series expansion around are: expansion around are:

Where is the remainder of the expansion which When we have:

tends to zero as → 0, as expressed in the relation

The total differential is expressed as follows Or

The tensor is of order 4 and is the

gradient of in .

Or

2nd order tensor defined as gradient of in

Tensor Analysis
Divergence of vectors and tensors Curl of a vector field

Let be a vector field. The divergence of is For a vector field the curl is defined as a vector:
the scalar obtained by a contraction:

or In index notation it is

When the divergence of a vector field is zero, Using the property of the permutation symbol we get
that is, the field its three components:
is called a solenoidal field.

For a tensor we have its divergence defined

as a vector:

Or

If the curl of the field is zero, that is,

The field is called irrotational.

Tensor Analysis
Laplacian of a scalar field Laplacian of a vector field

We also encounter second order derivatives in We can also treat a vector function in the same way.
expressions of physical quantities in mechanics. The divergence of the gradient of a vector is written
The divergence of the gradient of a scalar as:
function is an example:

The result of these operations is a vector. We also

which is also the Laplacian of denoted denote the operation as , that is,
or :

When the function is said harmonic

known as Laplace’s equation.

When where f is scalar it is called

Poisson’s equation.
Tensor Analysis
Definition of the flux Gauss theorem

Consider a body in a 3D space of volume and surface The Gauss theorem or divergence theorem, written
: as follows for an arbitrary component T jk ..... ( xi ) :

∂T jk .....
∫
ω
∂xi
dv = ∫ nT
∂ω
i jk ..... ds

This theorem transforms the volume integral of the

divergence of a property of a continuous medium
into a surface integral and plays an important role in
Surface and volume elements mechanics of continuous media.
In Gauss’s theorem

The flux of a property through the surface of the

body is (v is interpreted as the velocity field):
Curvilinear coordinates
Consider a point P with Cartesian coordinate xi . Cylindrical coordinates
In a curvilinear coordinate system
(cylindrical or spherical coordinates), the position
of the point P is given by the three numbers that
represent the coordinate curves
passing through P that is, by the curves on which two
of the three coordinates are constant.
The curvilinear coordinates can be considered
as functions of Cartesian coordinates:

Spherical coordinates

If the jacobian:

is not zero the

transformation
Is invertible.
Curvilinear coordinates
Cylindrical coordinates

All relevant parameters and operators

are found in:

Appendix A: Cylindrical coordinates

Appendix B: Spherical Coordinates

In the book: Mechanics of Continuous Media

(Botsis and Deville PPUR 2018)
Spherical coordinates