Properties of Tensors II

MEG 324 SSG 321 Introduction to Continuum Mechanics Instructors: OA Fakinlede & O Adewumi
[Link] [Link] oafak@[Link]
Last Week: Outstanding Issues 2

• It is expected that you should have read up on the notes and slides
from last week. I will not ordinarily repeat them in this week’s
matter but there are two issues that can prevent you from coping
if, you have not thoroughly mastered them:
• Determinant of a Product
• Trace of a Composition, and
• Scalar Product of Two Tensors

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

Determinant of a Product 3
• For the tensors 𝐀 and 𝐁, we use the definition of the determinant
to show that det 𝐀𝐁 = det 𝐀 × det 𝐁:
• Select linearly independent vectors 𝐚, 𝐛 and 𝐜. If 𝐁 is non-singular,
it is easy to show that 𝐮 = 𝐁𝐚 , 𝐯 = 𝐁𝐛 and 𝐰 = 𝐁𝐜 are also 𝐀𝐮, 𝐀𝐯, 𝐀𝐰
linearly independent. 𝐮, 𝐯, 𝐰
𝐀𝐁𝐚, 𝐀𝐁𝐛, 𝐀𝐁𝐜 𝐀𝐁𝐚, 𝐀𝐁𝐛, 𝐀𝐁𝐜 𝐁𝐚, 𝐁𝐛, 𝐁𝐜
det 𝐀𝐁 = =
𝐚, 𝐛, 𝐜 𝐁𝐚, 𝐁𝐛, 𝐁𝐜 𝐚, 𝐛, 𝐜
𝐀𝐮, 𝐀𝐯, 𝐀𝐰 𝐁𝐚, 𝐁𝐛, 𝐁𝐜
= = det 𝐀 × det 𝐁
𝐮, 𝐯, 𝐰 𝐚, 𝐛, 𝐜

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

The Inner Product Tensors 4

• The inner product of tensors 𝐒 and 𝐓 is the trace

𝐒: 𝐓 ≡ tr 𝐒 T 𝐓 = tr(𝐒𝐓 T )
• For the tensor, 𝐓 = 𝑇𝛼𝛽 𝐞𝛼 ⊗ 𝐞𝛽 , this definition leads to a simple
formula for the components of tensors as we shall show:
• Consider the composition, 𝐓 𝐞𝑗 ⊗ 𝐞𝑖 . We take the trace of this
composition and obtain,
tr 𝐓 𝐞𝑗 ⊗ 𝐞𝑖 = tr 𝑇𝛼𝛽 𝐞𝛼 ⊗ 𝐞𝛽 𝐞𝑗 ⊗ 𝐞𝑖

= tr 𝑇𝛼𝛽 𝐞𝛼 ⊗ 𝐞𝑖 𝛿𝛽𝑗

= 𝑇𝛼𝛽 𝛿𝛼𝑖 𝛿𝛽𝑗 = 𝑇𝑖𝑗

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019
The Inner Product Tensors 5

• From the above definition, the scalar components of tensor 𝐓 on

the dyad bases 𝐞𝑖 ⊗ 𝐞𝑗 is given by,
𝑇𝑖𝑗 = tr [𝐓 𝐞𝑗 ⊗ 𝐞𝑖 ]
T
= tr[𝐓 𝐞𝑖 ⊗ 𝐞𝑗 ]
= 𝐓: 𝐞𝑖 ⊗ 𝐞𝑗
• The definition of the inner product of tensors leads to a simple
way to compute the components. Just like vectors, it is by simply
taking the inner product with the product base.

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

Properties Covered Here Property

1. Members of a Euclidean Vector Space

• In continuation of our closer
examination of the properties of
tensors; we shall cover the 2. Additive Decompositions (4)
properties listed here.
• The Mathematics may look 3. The Cofactor Tensor
cumbersome, the strategy is
simple: Focus on definitions and
principles, all that will be left is the 4. Orthogonal Tensors
repeated applications of simple
rules.
• You will surprise yourself on what 5. Vector Cross: Other side of the Coin
you will know at the end! Axial Vector

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019 6

 One consequence of the foregoing is
the important fact that the set of
tensors, as we have defined it,

The Tensor: A constitutes a Euclidean Vector

Space.

Euclidean
Vector Space With this approach, the definition of
a tensor as a linear transformation
of vectors, applies to tensors of
higher orders!

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019 7

• In the same way as vectors, the inner product of tensors induces the
concept of magnitude and direction to tensors.
• Inspired by the fact that 𝐓: 𝐓 is a scalar, we define the magnitude of a tensor
𝐓 = 𝐓: 𝐓
• The angle between two tensors can be computed from,
−1
𝐒: 𝐓
𝜃 = cos .
𝐒 𝐓
• Unlike vectors, these values do not have the familiar geometric
interpretation of directed lines and included angles. The algebra
trumps the geometry.

Tensor Magnitude & Direction 8

Monday, September 9, 2019 [Link]; [Link]; oafak@[Link]
The Tensor Set as a Vector Space 9

A second-order tensor fulfils all the stipulations necessary to be a

Euclidean Vector Space 𝕃:
• Addition operation is defined, it is commutative and associative
under 𝕃: that is, 𝐓 + 𝐒 ∈ 𝕃, 𝐒 + 𝐓 = 𝐓 + 𝐒, 𝐓 + 𝐒 + 𝐕 = 𝐓 + 𝐒 +
𝐕, ∀ 𝐓, 𝐒, 𝐕 ∈ 𝕃. Furthermore,𝕃 is closed under addition: That is,
given that 𝐓, 𝐒 ∈ 𝕃, then 𝐕 = 𝐓 + 𝐒 = 𝐒 + 𝐓, ⇒ 𝐕 ∈ 𝕃.

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

The Tensor Set as a Vector Space 10

• 𝕃 contains a zero element 𝐎 such that 𝐓 + 𝐎 = 𝐓 ∀ 𝐓 ∈ 𝕃. For

every 𝐓 ∈ 𝕃, ∃ − 𝐓: 𝐓 + −𝐓 = 𝐎.
• Multiplication by a scalar. For 𝛼, 𝛽 ∈ ℝ and 𝐓, 𝐒 ∈ 𝕃, 𝛼𝐓 ∈ 𝕃 , 1𝐓 =
𝐓, 𝛼 𝛽𝐓 = 𝛼𝛽 𝐓, 𝛼 + 𝛽 𝐓 = 𝛼𝐓 + 𝛽𝐓, 𝛼 𝐓 + 𝐒 = 𝛼𝐓 + 𝛼𝐒.

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

 • The Zero element in the Second Axiom is obviously
the Annihilator Tensor.

Axioms for • The remaining two axioms are easily established

by recalling that dyads can be added

Second-Order • We break the tensor into its components as

follows:

Tensors 𝛼𝐒 + 𝛽𝐓 𝐮 = 𝛼𝑆𝑖𝑗 (𝐞𝑖 ⊗ 𝐞𝑗 ) + 𝛽𝑇𝑖𝑗 (𝐞𝑖 ⊗ 𝐞𝑗 ) 𝑢𝑙 𝐞𝑙

= 𝛼𝑆𝑖𝑗 + 𝛽𝑇𝑖𝑗 𝑢𝑙 𝐞𝑖 ⊗ 𝐞𝑗 𝐞𝑙
= 𝛼𝑆𝑖𝑗 + 𝛽𝑇𝑖𝑗 𝑢𝑙 𝐞𝑖 𝛿𝑗𝑙 = 𝛼𝑆𝑖𝑗 𝑢𝑗 𝐞𝑖
= 𝛼𝐒𝐮 + 𝛽𝐓𝐮

Monday, September 9, 2019 [Link]; [Link]; oafak@[Link] 11

Decomposing Tensors 12

• We break vectors to their components in

only one way. Tensors can be usefully
broken down in several different ways.
• Some of these decompositions are, like
the component form, additive. Others
are multiplicative.
• Additive decompositions include
Skew & Symmetric Forms, Spherical
and Deviatoric Forms, Spectral
Representation
• The Important Multiplicative is the
Polar Decomposition.
[Link]; [Link]; oafak@[Link] Monday, September 9, 2019
 Component Form: Components are
the weights. This
Weighted sum of turns the tensor
into a weighted
Dyads adds up to sum of up to nine
the Tensor dyad bases.

Decomposing
Symmetric &
Three other Additive Skew Parts, and
Spherical &
Decompositions are
Tensors
Deviatoric Parts
possible. Spectral Form of
eigenbases

There are also Multiplicative

Decompositions of Tensors

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019 13

Spherical & Deviatoric Parts 14

• Every tensor can be decomposed into Spherical and Deviatoric

parts. The Spherical Part of a tensor is obtained by dividing its
trace by three and using the result to scale an identity tensor. For
a tensor 𝐒,
1
• Spherical Part, the real multiplier of the Identity Tensor, 𝛾 = tr 𝐒 so that,
3
1 1
sph 𝐒 = tr 𝐒 𝐈 = 𝑆𝑘𝑘 𝛿𝑖𝑗 𝐞𝑖 ⊗ 𝐞𝑗
3 3
• Deviatoric Part is what remains after removing the Spherical Part:
1 1
dev 𝐒 = 𝐒 − tr 𝐒 𝐈 = 𝑆𝑖𝑗 − 𝑆𝑘𝑘 𝛿𝑖𝑗 𝐞𝑖 ⊗ 𝐞𝑗
3 3
[Link]; [Link]; oafak@[Link] Monday, September 9, 2019
Traces of Spherical & Deviatoric Parts 15

• For the Spherical Part, the trace,

1 1
tr sph 𝐒 = tr 𝐒 tr 𝐈 = tr 𝐒 3 = tr 𝐒
3 3
equals the trace of the original, undecomposed tensor.
• For the Deviatoric Part, the trace,
1
tr dev 𝐒 = tr 𝐒 − tr 𝐒 tr 𝐈 = tr 𝐒 − tr 𝐒 = 0.
3
• The deviatoric component has zero trace. A tensor
with zero trace is said to be “traceless”.

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

Skew & Symmetric Parts 16

• We can also decompose a tensor 𝐒, into Symmetrical and Anti-

Symmetrical parts.
• An Anti-Symmetric tensor, also called a Skew Tensor is defined as that which
is the negative of its transpose.
• The Symmetric Part,
1 T
1
sym 𝐒 = 𝐒 + 𝐒 = 𝑆𝑖𝑗 + 𝑆𝑗𝑖 𝐞𝑖 ⊗ 𝐞𝑗
2 2
• And the Skew Part:
1 T
1
skw 𝐒 = 𝐒 − 𝐒 = 𝑆𝑖𝑗 − 𝑆𝑗𝑖 𝐞𝑖 ⊗ 𝐞𝑗
2 2
[Link]; [Link]; oafak@[Link] Monday, September 9, 2019
Skew & Symmetric Parts 17

• The transposes,
T
T 1
sym 𝐒 = 𝐒 + 𝐒T Transpose of a sum equals the sum
2
1 T of the transposes:
= 𝐒T + 𝐒T = sym 𝐒 𝐚 ⋅ 𝐒 + 𝐓 T𝐛 = 𝐛 ⋅ 𝐒 + 𝐓 𝐚
2
• T
= 𝐛 ⋅ 𝐒𝐚 + 𝐛 ⋅ 𝐓𝐚
T 1 = 𝐚 ⋅ 𝐒T𝐛 + 𝐚 ⋅ 𝐓T𝐛
skw 𝐒 = 𝐒 − 𝐒T And, further, transpose of a
2
1 T transpose is the original vector
T
= 𝐒 − 𝐒 T = −skw 𝐒
2

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

Component Method 18

• The Symmetric Part,

T
1
sym 𝐒 = 𝑆𝑖𝑗 + 𝑆𝑗𝑖 𝐞𝑗 ⊗ 𝐞𝑖 = sym 𝐒
2
• And the Skew Part:
1 T
skw 𝐒 = 𝑆𝑖𝑗 − 𝑆𝑗𝑖 𝐞𝑗 ⊗ 𝐞𝑖
2
1
= − 𝑆𝑖𝑗 − 𝑆𝑗𝑖 𝐞𝑖 ⊗ 𝐞𝑗 = −skw 𝐒
2
• It is often more instructive to prove these in the direct form
whenever possible.
[Link]; [Link]; oafak@[Link] Monday, September 9, 2019
 Traces of Skew & Symmetric Parts 19
tr 𝐒 T = 𝑆𝑖𝑗 tr 𝐞𝑗 ⊗ 𝐞𝑖
= 𝑆𝑖𝑗 𝐞𝑗 ⋅ 𝐞𝑖 1 T
1
= 𝑆𝑖𝑗 𝐞𝑖 ⋅ 𝐞𝑗 tr sym 𝐒 = tr 𝐒 + 𝐒 = tr 𝐒 + tr 𝐒
= 𝑆𝑖𝑗 tr 𝐞𝑖 ⊗ 𝐞𝑗 = tr 𝐒 2 2
• commutative property of the scalar product makes the trace of
a transpose the same as the trace of the tensor from which the
transpose is obtained.
• For the same reason, a Skew tensor is traceless:
1 T
1
tr skw 𝐒 = tr 𝐒 − 𝐒 = tr 𝐒 − tr 𝐒 = 0.
2 2
• The Spherical Part of a tensor is a diagonal tensor, and
therefore, always symmetric.
[Link]; [Link]; oafak@[Link] Monday, September 9, 2019
Symmetry of Deviatoric Parts

• No judgement can be made on the symmetry or

skewness of a deviatoric tensor, however. Its
symmetry wholly depends on the original tensor
from which the deviatoric part is taken.
• If the latter is symmetric, so will the deviatoric
part. If skew, so also will the deviatoric part.
• It is quite possible that the deviatoric tensor is
neither symmetric nor skew.

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019 20

 In Continuum Mechanics, we are
concerned with changes in shape of
objects that have been subjected to
Transformation mechanical loads.

of Line
Elements We shall see later that
these creates
We are interested
in what happens
to line elements
in materials
important tensors such where such strain
and deformation
as stress, strain and tensors have been
deformation tensors. created by
external loads.

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019 21

• It will be our business to demonstrate (Chapter 4, Kinematics) that
small line elements are transformed by applying the deformation
gradient tensor to the original, vectors of undeformed line elements.
• What can we say about vectors of small areas in the deformed state?
• We will show that when a tensor transforms lines, the corresponding
vector areas are transformed by its Cofactor.

Cofactor Tensor & Area Vector 22

Monday, September 9, 2019 [Link]; [Link]; oafak@[Link]
The Cofactor & Area
Vector
• Vectors 𝐮 and 𝐯 have both been transformed by
the same tensor 𝐓 to 𝐓𝐮 and 𝐓𝐯.
• How is the vector area 𝐓𝐮 × 𝐓𝐯 created by
these two vectors in the transformed state
related to the original vector area 𝐮 × 𝐯 of
the parallelogram as shown?
• The tensor transforming the area of
parallelogram sides that have been
transformed by tensor 𝐓 is called the cofactor,
𝐓 c of 𝐓.
• This is the physical meaning of the
cofactor. Any other definition can be
obtained from this physical definition.

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019 23

The Cofactor is a Bilinear Transformation 24

• For example, given 𝛼, 𝛽 ∈ ℝ, and 𝐮, 𝐯, 𝐰 ∈ 𝔼, the linearity of tensor

𝐓⇒
𝐓 𝛼𝐮 + 𝛽𝐯 × 𝐓𝐰 = 𝛼𝐓𝐮 + 𝛽𝐓𝐯 × 𝐓𝐰
= 𝛼𝐓𝐮 × 𝐓𝐰 + 𝛽𝐓𝐯 × 𝐓𝐰
= 𝛼𝐓 c 𝐮 × 𝐰 + 𝛽𝐓 c 𝐮 × 𝐰
= 𝐓 c 𝛼𝐮 + 𝛽𝐯 × 𝐰
• The last equality coming from the linearity of cofactor tensor
𝐓 c , and the distributive property of the vector product over
addition.

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

Components of the Cofactor 25

• For a tensor 𝐓, let the cofactor,

cof 𝐓 = 𝐓 𝐜 = 𝑇𝑖𝑗𝑐 𝐞𝑖 ⊗ 𝐞𝑗
• Clearly, the component,
𝑇𝑖𝑗𝑐 = 𝐓 𝐜 : 𝐞𝑖 ⊗ 𝐞𝑗 = 𝐞𝑖 ⋅ 𝐓 𝐜 𝐞𝑗
𝐜
1
= 𝐞𝑖 ⋅ 𝐓 𝑒𝑗𝑚𝑛 𝐞𝑚 × 𝐞𝑛
2
1
= 𝑒𝑗𝑚𝑛 𝐞𝑖 ⋅ 𝐓 𝐜 𝐞𝑚 × 𝐞𝑛
2

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

Components of the Cofactor 26

• In the expression, 𝐞𝑖 ⋅ 𝐓𝐞𝑚 × 𝐓𝐞𝑛 we seek the 𝑖 𝑡ℎ component of For any vector 𝐮,
𝐞𝑖 ⋅ 𝐮 = 𝑢𝑖
𝐓𝐞𝑚 × 𝐓𝐞𝑛 . To get this, we remember that the 𝛼 and 𝛽 components of The 𝑖 component of 𝐮
𝑡ℎ

each operand are 𝐞𝛼 ⋅ 𝐓𝐞𝑚 and 𝐞𝛽 ⋅ 𝐓𝐞𝑛 respectively. Consequently we

seek, 𝑒𝑖𝛼𝛽 𝐞𝛼 ⋅ 𝐓𝐞𝑚 𝐞𝛽 ⋅ 𝐓𝐞𝑛 so that,
𝑐 1
𝑇𝑖𝑗 = 𝑒𝑗𝑚𝑛 𝑒𝑖𝛼𝛽 𝐞𝛼 ⋅ 𝐓𝐞𝑚 𝐞𝛽 ⋅ 𝐓𝐞𝑛
2
1
= 𝑒 𝑒 𝑇 𝑇
2 𝑖𝛼𝛽 𝑗𝑚𝑛 𝛼𝑚 𝛽𝑛
• The Cofactor Tensor, in component form,
c
1
𝐓 = 𝑒𝑖𝛼𝛽 𝑒𝑗𝑚𝑛 𝑇𝛼𝑚 𝑇𝛽𝑛 𝐞𝑖 ⊗ 𝐞𝑗
2
[Link]; [Link]; oafak@[Link] Monday, September 9, 2019
Components of the Cofactor 27

• Second principal invariant of 𝐓 is the trace of its cofactor, tr 𝐓 c

1
𝐼2 𝐓 = 𝑒𝑖𝛼𝛽 𝑒𝑗𝑚𝑛 𝑇𝛼𝑚 𝑇𝛽𝑛 𝐞𝑖 ⋅ 𝐞𝑗
2
1
= 𝛿𝛼𝑚 𝛿𝛽𝑛 − 𝛿𝛼𝑛 𝛿𝛽𝑚 𝑇𝛼𝑚 𝑇𝛽𝑛
2
1
= 𝑇𝑚𝑚 𝑇𝑛𝑛 − 𝑇𝑚𝑛 𝑇𝑛𝑚
2
1 2
= tr 𝐓 − tr 𝐓 2
2
• This is exactly same expression we obtained from the definition of
the second Principal Invariant.
[Link]; [Link]; oafak@[Link] Monday, September 9, 2019
Orthogonal Tensors 28

• In our study of tensor properties, we

encountered some interesting tensors. CLASS ONE CLASS TWO
The table here shows some of these
tensors as we classify them into two
groups as shown. Projection 1
2 Identity 𝐈
𝐏𝐱 ≡ (𝐱 ⊗ 𝐱)
𝐱
• Tensors in the first group change the
magnitude of the vectors they
transform; the second group are Vector Cross 𝐯 × ≡ −𝑒𝑖𝑗𝑘 𝑣𝑘 𝐞𝑖 ⊗ 𝐞𝑗 Coordinate 𝛏𝑖 ⊗ 𝐞𝑖
tensors that create transformed Rotation
vectors with the same magnitudes as
the input vector. Spherical 𝛾𝐈
• The second group are just two
members of a distinguished class of Annihilator 𝐎
tensors we will need to spend a little
more time to know better. They are in
a class called “Orthogonal tensors”. Dyad 𝐚⊗𝐛

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

CLASS ONE CLASS TWO

Projection 1
2 Identity 𝐈
𝐏𝐱 ≡ (𝐱 ⊗ 𝐱)
𝐱

Vector Cross 𝐯 × ≡ −𝑒𝑖𝑗𝑘 𝑣𝑘 𝐞𝑖 ⊗ 𝐞𝑗 Coordinate Rotation 𝛏𝑖 ⊗ 𝐞 𝑖

Spherical 𝛾𝐈

Annihilator 𝐎

Dyad 𝐚⊗𝐛

Orthogonal Tensors 29
[Link]; [Link]; oafak@[Link] Monday, September 9, 2019
• Given a pair of vectors 𝐚 and 𝐛, an orthogonal tensor 𝐐 is said to be
orthogonal if,
𝐐𝐚 ⋅ 𝐐𝐛 = 𝐚 ⋅ 𝐛
• Specifically, we can allow 𝐚 = 𝐛, so that
𝐐𝐚 ⋅ 𝐐𝐚 = 𝐚 ⋅ 𝐚
• Or
𝐐𝐚 = 𝐚
• As we can see, one important attribute of the orthogonal tensor is
that the magnitude of the input as well as that of the result remain
the same: A magnitude-preserving transformation.

Definition: What is an
Orthogonal Tensor? 30
Monday, September 9, 2019 [Link]; [Link]; oafak@[Link]
• Let 𝐪 = 𝐐𝐚
𝐐𝐚 ⋅ 𝐐𝐛 = 𝐪 ⋅ 𝐐𝐛 = 𝐚 ⋅ 𝐛 = 𝐛 ⋅ 𝐚
• Recall the definition of the transpose; we have that,
𝐪 ⋅ 𝐐𝐛 = 𝐛 ⋅ 𝐐𝐓 𝐪 = 𝐛 ⋅ 𝐐𝐓 𝐐𝐚 = 𝐛 ⋅ 𝐚
• Clearly, 𝐐𝐓 𝐐 = 𝐈. A condition necessary and sufficient for a tensor 𝐐
to be orthogonal is that 𝐐 be invertible and its inverse equal to its
transpose.
• Every orthogonal tensor possesses an inverse;
• This inverse is simply its transpose!

Definition: What is an
Orthogonal Tensor? 31
Monday, September 9, 2019 [Link]; [Link]; oafak@[Link]
The Determinant of an
Orthogonal Tensor

• Upon noting that the determinant of a product is

the product of the determinants and that
transposition does not alter a determinant, it is easy
to conclude that,
det 𝐐𝐓 𝐐 = det 𝐐𝐓 det 𝐐 = det 𝐐 2 = 1
• Which clearly shows that
det 𝐐 = ±1
• When the determinant of an orthogonal tensor is
strictly positive, it is called “proper orthogonal”. In
particular, a rotation is a proper orthogonal tensor
while a reflection is not.

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019 32

Proper Orthogonal Tensors are Self-Cofactor

For any pair of vectors 𝐮, 𝐯 we show that 𝐐 𝐮 × 𝐯 = (𝐐𝐮) × (𝐐𝐯) 33

• This question is the same as showing that the cofactor of 𝐐 is 𝐐 itself.
That is that a rotation is self cofactor. We can write that
𝐐𝐜 𝐮 × 𝐯 = (𝐐𝐮) × (𝐐𝐯)
• where
𝐓 = cof 𝐐 = det 𝐐 𝐐−T
• Now that 𝐐 is a rotation, det 𝐐 = 1, and
𝐐−T = (𝐐−1 )T = (𝐐T )T = 𝐐
• This implies that 𝐓 = 𝐐 and consequently,
𝐐 𝐮 × 𝐯 = (𝐐𝐮) × (𝐐𝐯)

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

One Eigenvalue of a Proper Orthogonal is +1 34

This means that there is always a solution for the equation,

𝐐𝐮 = 𝐮
• For any invertible tensor,
𝐒c = det 𝐒 𝐒−T
• For a proper orthogonal tensor 𝐐, det 𝐐 = 1. It therefore follows that,
𝐐c = det 𝐐 𝐐−T = 𝐐−T = 𝐐
• Characteristic equation for 𝐐 is,
det 𝐐 − 𝜆𝐈 = 𝜆3 − 𝜆2 𝑄1 + 𝜆𝑄2 − 𝑄3 = 0
• Or,
𝜆3 − 𝜆2 𝑄1 + 𝜆𝑄1 − 1 = 0
• Which is obviously satisfied by 𝜆 = 1.
[Link]; [Link]; oafak@[Link] Monday, September 9, 2019
The Axial Vector 35

• The Vector Cross is a Skew Tensor.

• This fact can be seen by transposing and showing that the transpose is its
negative.
• Like all other skew tensors, it is a traceless tensor.
• Now the question: Given a Skew tensor, can we find a vector that the
tensor is a vector cross of?
• When this happens, the vector found in this way is called the axial vector of the
tensor.
• Note that, in either case, there are only three components. This is obvious for
the vector.
• For the skew tensor, once three independent components are found, the
remaining are simply the negatives of the independent components.

[Link]; [Link]; oafak@[Link] Monday, September 9, 2019

The Axial Vector 36

• From Week Six (Slide 26) the vector cross was defined as:
𝛀 = 𝐯 × ≡ −𝑒𝑖𝑗𝑘 𝑣𝑘 𝐞𝑖 ⊗ 𝐞𝑗 = 𝛺𝑖𝑗 𝐞𝑖 ⊗ 𝐞𝑗
• Given 𝛀 = 𝛺𝑖𝑗 𝐞𝑖 ⊗ 𝐞𝑗 , can we solve the above equation for the
components 𝑣𝑘 ?
• Sure, we can! Begin from,
−𝑒𝑖𝑗𝑘 𝑣𝑘 = 𝛺𝑖𝑗
Multiply both sides by 𝑒𝛼𝑖𝑗 and we have:
−𝑒𝑖𝑗𝑘 𝑒𝛼𝑖𝑗 𝑣𝑘 = −2𝛿𝑘𝛼 𝑣𝑘 = 𝑒𝛼𝑖𝑗 𝛺𝑖𝑗
1
• The components of the axial vector are 𝑣𝑖 = − 𝑒𝑖𝑗𝑘 𝛺𝑗𝑘
2
[Link]; [Link]; oafak@[Link] Monday, September 9, 2019