You are on page 1of 6

Lecture 9 69

Vectors & Tensors

9 . 3 Einstein Summation Convention

The above equation is written in a form that may not be familiar to you because it uses a simple
mathematical shorthand notation. We need the shorthand because that equation actually represents a set
of three linear equations which are somewhat cumbersome to deal with and write down all the time.
There are nine coefficients, εij , which correspond to the values of the strain tensor with respect to whatever
coordinate system you happen to be using. Those three equations are:

p1 = σ 11l1 + σ 12l2 + σ 13l3 ,

p2 = σ 21l1 + σ 22l2 + σ 23l3 , (9.2)

p3 = σ 31l1 + σ 32l2 + σ 33l3 .

We could write the same in matrix notation:

 p1  σ 11 σ 12 σ 13   l1 
 p  = σ   
 2   21 σ 22 σ 23  l2  , (9.3)
 p3  σ 31 σ 32 σ 33  l3 

but this is still awkward, so we use the notation above, known as dummy suffix notation, or Einstein
Summation Convention. Equations 8-2 can be written more efficiently:

3
p1 = σ 11l1 + σ 12l2 + σ 13l3 = ∑ σ 1 j l j ,
j =1

3
p2 = σ 21l1 + σ 22l2 + σ 23l3 = ∑ σ 2 j l j ,
j =1

3
p3 = σ 31l1 + σ 32l2 + σ 33l3 = ∑ σ 3 j l j .
j =1

From here, it is just a short step to equation 9.1:

pi = σ ij l j , where i and j both can have values of 1, 2, or 3.

p1, p2, and p3 are the tractions on the plane parallel to the three axes of the coordinate system, X1, X 2, and
X3, and l1, l2, and l3 are equal to cosα, cosβ, and cosγ, respectively.

Economic Geo. _ JH
draft date: 20 Jan, 1999
Lecture 9 70
Vectors & Tensors

In equation 9.1, because the “j” suffix occurs twice on the right hand side, it is the dummy suffix,
and the summation occurs with respect to that suffix. The suffix, “i”, on the other hand is the free suffix;
it must occur once on each side of each equation.

You can think of the Einstein summation convention in terms of a nested do-loop in any
programming language. In a FORTRAN type language, one would write the above equations as follows:

Do i = 1 to 3
p(i) = 0
Do j = 1 to 3
p(i) = sigma(i,j)*l(j) + p(i)
repeat
repeat

9.4 Coordinate Systems and Tensor Transformations

The specific values attached to both vectors and tensors -- that is the three numbers that represent
a vector or the nine numbers that represent a tensor -- depend on the coordinate system that you choose.
The physical property that is represented by the tensor (or vector) is independent of the coordinate
system. In other words, I can describe it with any coordinate system I want and the fundamental nature
of the thing does not change. As you can see in the diagram, below, for vectors:

X3 X 3'

V'
V

V3 V 2' X 2'
V 3'
V2
V1 X2
V 1'

(note that the length and relative orientation


X1
of V on the page has not changed; only the
axes have changed) X 1'

Economic Geo. _ JH
draft date: 20 Jan, 1999
Lecture 9 71
Vectors & Tensors

The same is true of tensors; a strain ellipse has the same dimensions regardless of whether I take
a coordinate system parallel to geographic axes or a different one. In the earth, we can use a variety of
different coordinate systems; the one most commonly used when we’re talking about vectors and tensors
is the Cartesian system with direction cosines described earlier:

• north, east, down .

There are times when we want to look at a problem a different way: For example, we are
studying a fault and we want to make the axes of the coordinate system parallel to the pole to the fault
and the slip direction;

There is a simple way to switch between geographic and fault coordinates: Coordinate
transformation, and the related transformations of vectors and tensors.

We’re not going to go into the mathematics of transformations (although they are reasonably
simple). Just remember that the difference between a tensor and any old random matrix of nine numbers
is that you can transform the tensor without changing its fundamental nature.

The nine numbers that represent an infinitesimal strain tensor, or any other tensor, can be
represented as a matrix, but not all matrices are tensors. The specific values of the components change
when you change the coordinate system, the fundamental nature does not. If I happen to choose my
coordinates so that they are parallel to the principal axes of stress, then the form of the tensor looks like:

σ 1 0 0 
σ ij =  0 σ 2 0 
 0 0 σ 3 

9.5 Symmetric, Asymmetric, & Antisymmetric Tensors

Coming back to our original problem of describing the changes of vectors during deformation,
the tensor that relates all those vectors in a circle to their position is known at the displacement gradient
tensor.

Economic Geo. _ JH
draft date: 20 Jan, 1999
Lecture 9 72
Vectors & Tensors

The displacement gradient tensor, in general, is an asymmetric tensor. What that means is that it
has nine independent components, or, if you look at it in matrix form:

e11 e12 e13 


eij = e21 e22 e23  , where e12 ≠ e21 , e13 ≠ e31, and
e32 ≠ e23 .
e31 e32 e33 

If eij were a symmetric tensor, then e12 = e21 , e13 = e31, and e32 = e23 , and it would have only 6
independent components.

It turns out that any asymmetric tensor can be broken down into a symmetric tensor and an
antisymmetric tensor. So, for the displacement gradient tensor, we can break it down like:

eij =
(eij + e ji ) + (e
ij − e ji )
2 2

 (e12 + e21 ) (e13 + e31 )   (e12 − e21 ) (e13 − e31 ) 


 e11   0 
 2 2   2 2
(e + e )
=  21 12 e22
(e23 + e32 )  +  (e21 − e12 ) 0
(e23 − e32 ) 
 2 2   2 2 
 (e31 + e13 ) (e32 + e23 )   (e31 − e13 ) (e32 −−e23 ) 
 e33   0 
 2 2   2 2 

Writing the same equation in a more compact form:

eij = ε ij + ω ij ,
where

ε ij =
(eij + e ji ) and ω ij =
(e
ij − e ji ) .
2 2

The symmetric part is the infinitesimal strain tensor and the antisymmetric part is the rotation tensor.
Written in words, this equation says:

“the displacement gradient tensor = strain tensor + rotation tensor”.

Note that the infinitesimal strain tensor is always symmetric. Thus, you can think of pure shear as ωi j = 0

Economic Geo. _ JH
draft date: 20 Jan, 1999
Lecture 9 73
Mohrs Circle for Stress

and simple shear as ωi j ≠ 0.

9.6 Finding the Principal Axes of a Symmetric Tensor

The principal axes of a second order tensor can be found by solving an equation known as the
“Characteristic” or “secular” equation. This equation is a cubic, with the following general form:

λ3 − Ιλ2 − ΙΙλ − ΙΙΙ = 0

The three solutions for λ are called the eigenvalues; they are the magnitudes of the three principal axes.
Knowing those, you can calculate the eigenvectors, which give the orientations of the principal axes. The
calculation is generally done numerically using a procedure known as a Jacobi transformation. The
coefficients, Ι, ΙΙ and ΙΙΙ are known as the invariants of the tensor because they have the same values
regardless of the orientation of the coordinate system. The first invariant, Ι, is particularly useful because
it is just the sum of the principal diagonal of the tensor. Thus, for the infinitesimal strain tensor, it is
always true that:

σ 1 + σ 2 + σ 3 = σ 11 + σ 22 + σ 33 .

This is particularly useful when we get to stress and something known as hydrostatic pressure.

Economic Geo. _ JH
draft date: 20 Jan, 1999
Lecture 1 0 74
Mohrs Circle for Stress

LECTURE 10—STRESS II: MOHR ’S CIRCLE

10.1 Stresses on a Plane of Any Orientation from Cauchy’s law

We would like to be able to calculate the stress on any plane in a body. To do this, we will use
Cauchy’s Law, which we derived last time.

X3

p3N
γ
p3
α
p3S p1
γ α

γ p1S p1N

α
X1

We will assume that we know the orientations of the principal stresses and that we have chosen our
coordinate system so that the axes are parallel to those stresses. This gives us the following matrix for the
stress tensor:

σ 1 0 0 
σ ij =  0 σ 2 0  (10.1)
 0 0 σ 3 

The general form of Cauchy’s Law is:


pi = σ ij l j (10.2)

which, if we expand it out for the case shown above will be:

p1 = σ 1l1 = σ 1 cos α
p3 = σ 3l3 = σ 3 cos γ = σ 3 cos(90 − α ) = σ 3 sin α

Economic Geo. _ JH
draft date: 20 Jan, 1999

You might also like