Robert Andrew Martin

Linear Algebra the vectors of S are linearly independent if no

nontrivial linearcombination of the vectors is zero
Vector Spaces i li Xi O x xn O
A basis is a set of linearly independent vectors that
A matrix can be thought of as a linear relationship between the
spans space
two vectors all bases of V have the same number of
Scalars are elements of a number field e.g Rl or elements the dimension of the space
A field is a set of elements on which addition and any vector x EV can be written uniquely as eix
multiplication are defined and are commutative associative we can convert between bases with a transformation matrix
and distributive e
di Rig di ki ee og Ei Rijog
closed under addlmutt
includes identity elements 0 for add 1 for multi Linearnvm
operators
includes inverses for every element except aero A linear operator A acts on a vector space to produce
Vectors are elements of a vectorspace defined over other elements of V
some number field A Ca Et P y a A E t B A
y
vector addition and inner product are defined A matrix is an array of numbers that can represent a
closed under these linear operator It contains the components of the operator
ops
includes identity element for addition with respect to a certain basis
because A is linear knowledge of its action on a basis
Let f et ez ene be a subset of some is sufficient to know its action on any vector in thespace
vector space V The of S is the set of all As of Ei Ai og
span
vectors that are linear combinations of S
 Robert Andrew Martin

we can rewriteA in a different basis as follows the Cauchy Schwarz inequality states
we
require ei Ai Xj Ei Ai od
by using Eg Gi Ri and relabelling indices
I Loe It TE LEI E Lyla
Evi Rai Ai og e n Airy od
RA CR od A E
or
I Get g I e 1 1141
A RAN equality when E and y are linearly dependent
canbe proven by considering Ge ay Iz ay
lnnerproductsm then later setting lat hey 171
The inner product GE l y is a scalar function we can use Cauchy Schwarz to define cos 0 in R
of two vectors It must cos Q Lofty
be bilinear i.e linear in the second argument 111141
and antilinear in the first
og lay x
Ely and Casely a CE ly Hermitianmatries
have Hermitian symmetry The Hermitian conjugate of a matrix is the complex
yl x Cos I y conjugate of its transpose
be positive definite At At At ij A
LEI E o c equality iff og it obeys similar rules to the transpose
t
distributive in the first argument Att t A Ao ftp.t
Get 4 I El z t C F lZ the Hermitian conjugate of a scalar is just the
t
In Rn LE ly xiyi conjugate e.g LE ly Cody
In L El 4 xi
y a matrix is Hermitian if A At
 Robert Andrew Martin

knowing Leite for basis vectors is sufficient to Eigenvalues and Eigenvectors

know Ely because of bilinearity If Cei IE Gig
Lofty Lei xi IE yj Kitty Gi An eigenvector of a linear operator is a nonzero

Gig are the metric coefficients rector of such that A E X E We can find
G is Hermitian since Gi's GTi the eigenvalues and eigenvectors by solving the
a basis is orthonormal if characteristic equation det CA XI o
Lei led Sig
if the n roots are distinct there are n linearly
The adjoint of a linear operator Ar with respect to independent eigenvectors unique to a constant factor
some inner product is another linearoperator At such that if an eigenvalue is degenerate and occurs m times
Atx ly LEI Ay there may be between 1 and m linearly independent
for given basis the components of Ant
a are eigenvectors for that eigenvalue spanning the eigenspace
the entries in the matrix At In general we can prove eigenvalue Eigenvector
properties as follows using the example of Hermitian matrices
Consider two eigenvalueHector
pairs
Matrix Symmetry Equation Ax Xx A y my
AT A
symmetric
antisymmetric At A
Take Hermitian conjugate of
then use Hermitian property
i
ytAt µ yt
ytA µ yt
orthogonal AAt ATA I Cainmaphokgues
apply yt to to get two expressions for ytAx
Hermitian At A 7 M ytx o
anti Hermitian At A suppose x and y are the same eigenvector and 2m
unitary At A these are all X ft x Tx 0
normal A At AtA c normal six to a
p X
so
eigenvalues of Hermitian matrix are real
 Robert Andrew Martin

if x and are different eigenvectors A matrix is diagonalisable if it is similar to a diagonal

y
CT pi ytx O matrix i e A S AS
for X F M To diagonalise we form 5 from
ytx
e
o
hence eigenvectors orthogonal for different eigenvalues the eigenvectors of A The entries
of 1 are then the corresponding
g Ey
If
The eigenvectors of normal matrices corresponding to eigenvalues
distinct eigenvalues are orthogonal l Asi AI
5 AS 5 A of s
if A is Hermitian IA is anti Hermitian vice versa

l
if A is Hermitian exp CIA is unitary A
an eigenbasis can ALWAYS be constructed for a
s GE En't C
normal matrix even if there are degenerate eigenvalues
YointeeartyhatindepenananghsingtiesdaiaginimumitionTTi
Symmetry Eigenvalues Interpretation only if A has n linearly independent eigenvectors
Hermitian X T Real Thus normal matrices are diagonals'sable and the
eigenvectors can be chosen to be orthonormal
reofafie.ba
anti Hermitian
Unitary
X
x 4x
7 Imaginary
Unit modulus yi Intuitively diagonalisation is the process of expressing
a matrix in its eigenbasis the simplest form Hence
Diagonalisation of a matrix the similarity matrix is unitary and A U n ut
E RE wtf for a transformation between basis vectors Diagonalisation is useful because some operations are
a linear operator can be transformed via A's RAR mucheasier to carry out on the diagonalised repr A Srs t
Two
square
matrices are similar if 8 5 AS where Am SMS
S is some invertible similarity matrix del A deth For any matrix detA III i

tr A tr A tr A E Xi
 Robert Andrew Martin

2
Thetransformation between orthonormal bases is described by Given X od t Iz y't 732 K
a unitary matrix As have same sign ellipsoid
a real symmetric matrix can be diagonalised by Ts have mixed sign hyperboloid
a real orthogonal transformation 7 22 7 sphere
7 72 surface of revolution about 2 axis
Quadratic forms 13 0 translation of conic section along 2 axis

wmThe
quadratic form associated with a real symmetric matrix
hencethe name Hermitian
A is Q x at Aoc Aigociog 2
forms
Q is a homogeneous quadratic function of oc i e The Hermitian form is a complex extension of the quadratic
QCxD NQ Loc Any homogeneous quadratic is the form Hbd xtAx real scalar quantity
quadratic form of some symmetric matrix Hermitian matrices can be diagonals'ed with unitary
Because real symmetric matrices can be diagonalised by transformations Had x'thx Ti l Kil
orthogonal transformations
The Rayleigh quotient associated with a Hermitian matrix
QCod SEA x x'TA x oc Sx is the normalised Hermitian form
yay xtAxxtx
the eigenvectors of A are the principal axes
in the eigenbasis the quadratic form is just a if x is aneigenvector of A X is an eigenvalue
sum of squares Q Xi Xi easily verified by substitution
Quadratic forms represent quadratic surfaces
can the Rayleigh Ritz variational principle considers
QCH K f constant 87 Xcx tsx 764 and shows that the
hence representingQ in its eigenbasis allows us eigenvectors of A are the stationary points of Xxl

to easily identify the shape

 Robert Andrew Martin

Cartesian Tensors Tensors

nm

A tensor of order rank n transforms between two

In Cartesian basis vectors are independent of position orthonormal basis sets as described by the transformation law
To transform from basis vectors Ej to Ei i
Ti in Liis Linn Tj jn
k v E Vi Ei vi Ei y Ei E v

The order of atensor is

equal to
the number of indices
vi L Vj with Lij EI E
needed to tablet it Scalars are order zero vectors are
rotates the frame
is the transformation matrix order 1 matrices are order 2
the
argument applies when interchanging v and
same Pseudotensors are defined with an additional Det L
v So LTL L LT I L is orthogonal factor changing the sign during reflection
A Cartesian vector v is defined as a set of coefficients The Kronecker delta is a second order tensor

vi with respect to orthonormal basis Eni such that

an fig Lipljqfpq lipljp fi.TL is orthogonal
an orthogonal transformation transforms to another orthonormal basis The Levi Civita symbol is a third order pseudotensor This
Eli with coefficients vi can be shown by verifying that one of the nonzero terms

Orthogonal matrices have determinant II stays constant under a transformation

det L it is a proper rotation E izz dett Liphq r Epqr Get LT
det L l is an improper rotation ie rotation creflection The inertia tensor relates the angular momentum I to the
if L and L are proper rotations theircomposition angular velocity y DI dm Ex Ceyx E dm hehe HE
is also a
proper rotation Vi Vij L'Yu Vk Ji Ii w with Ii Ju plkkxkxr.fi Xix DV
A Cartesian pseudovector transforms via ai del L Liga
i gains a sign change under
e
l Susceptibility tensors 2ndorder relate the polarisation to
q 1 f Vector
the applied E field Pi
any reflection change of handedness EX Ej
cross products are T l k pseudo
always pseudovectors 1 Vector
 Robert Andrew Martin

Elastic deformation is described by thestrain tensor Ei If Sijk is symmetric in i j and Apgar is antisymmetric
Ei IN t
3 in pg then the contraction Sijk Ai r 0

the associated stress tensor oi defines the j th componen

of force on a plane perpendicular to i Secondrumrunner
order tensors
they are related by a fourth order stiffness tensor 2ndorder tensors can be represented as matrices and thus
have matrix properties
Propertiesoftensoism An antisymmetric second order tensor is equivalent to a
If A and B are order then so is any linear
n tensors certain pseudovector the dual vector
combination of them Proof transform C L A t
pB
C i in NA i in t B 8 i in
Ais Eijkak fg.IE af
L Lij Lin'sn Aj j n
P Lij Linjn Bj n
t
j Any symmetric second order tensor can be uniquely
Lig
Lin'sn x
Aj in P Oji jn written as the sum of a symmetric traceless tensor
The tensor product of tensors of order n and m is and a scalar multiple of the identity tensor
a tensor of order n m also called outerproduct S S I tr s I I tr s I
A 8 Ci in in Ai in B int
in m in m traceless
a general tensor can be written
as T Ti
in Ei Ox Ein Symmetric second order tensors can be diagonals's ed
tensor pseudo tensor pseudotensor
A tensor contraction sets two indices equal and sums Isotropictensorim
over returning a tensor of order n 2 Isotropic tensors are in variant with respect to the frame
A tensor is symmetric in a and thus have no preferred direction
pair of indices if
T i T j i and antisymmetric if 0th order all scalars isotropic transformation land
n
j are

T i j T The Cantilsymmetry of a 1st order only the zero vector is isotropic

j i
tensor is in variant under a change of coordinates 2ndorder A fig for scalar 7
 Robert Andrew Martin

3rdorder X Eijk for scalar X

4th order Xf Sm t Mfikf tv fits a for scalar X M V

Isotropy may be used to evaluate integrals when

the integration region is symmetric r er du du

Xi freak pCrYdV freaRipe Pcr dV RijXj Xi

Xi RigX for general Rig means Xi is isotropic

the only isotropic vector is the zero vector so E Q
E g for a second order tensor integral

kij frieaxilx.jp
CmdV RikRj ki Kii
Ki X fi with 2 I Tr k
Ki Sr Ea ErpCr dV 8ijJ
Tensor
c
fields
in some
A tensor field assigns tensor to every position I
a domain
e g a conductivity field 2ndorder tensor field
The divergence of a vector field is scalar the contraction
of the tensor product of two vector fields Oi and Fj
Tx E is a pseudovector field
the contraction of the
tensor product of pseudotensor Eijk and vectors 2C Fm
The derivative of a secondorder tensor field is a third order
tensor field 2 i ogk