Part IA — Vectors and Matrices

Based on lectures by N. Peake

Notes taken by Dexter Chua

Michaelmas 2014

These notes are not endorsed by the lecturers, and I have modified them (often
significantly) after lectures. They are nowhere near accurate representations of what
was actually lectured, and in particular, all errors are almost surely mine.

Complex numbers
Review of complex numbers, including complex conjugate, inverse, modulus, argument
and Argand diagram. Informal treatment of complex logarithm, n-th roots and complex
powers. de Moivre’s theorem. [2]

Vectors
Review of elementary algebra of vectors in R3 , including scalar product. Brief discussion
of vectors in Rn and Cn ; scalar product and the Cauchy-Schwarz inequality. Concepts
of linear span, linear independence, subspaces, basis and dimension.
Suffix notation: including summation convention, δij and εijk . Vector product and
triple product: definition and geometrical interpretation. Solution of linear vector
equations. Applications of vectors to geometry, including equations of lines, planes and
spheres. [5]

Matrices
Elementary algebra of 3 × 3 matrices, including determinants. Extension to n × n
complex matrices. Trace, determinant, non-singular matrices and inverses. Matrices as
linear transformations; examples of geometrical actions including rotations, reflections,
dilations, shears; kernel and image. [4]
Simultaneous linear equations: matrix formulation; existence and uniqueness of solu-
tions, geometric interpretation; Gaussian elimination. [3]
Symmetric, anti-symmetric, orthogonal, hermitian and unitary matrices. Decomposition
of a general matrix into isotropic, symmetric trace-free and antisymmetric parts. [1]

Eigenvalues and Eigenvectors

Eigenvalues and eigenvectors; geometric significance. [2]
Proof that eigenvalues of hermitian matrix are real, and that distinct eigenvalues give
an orthogonal basis of eigenvectors. The effect of a general change of basis (similarity
transformations). Diagonalization of general matrices: sufficient conditions; examples
of matrices that cannot be diagonalized. Canonical forms for 2 × 2 matrices. [5]
Discussion of quadratic forms, including change of basis. Classification of conics,
cartesian and polar forms. [1]
Rotation matrices and Lorentz transformations as transformation groups. [1]
Contents
0 Introduction

1 Complex numbers
1.1 Basic properties
1.2 Complex exponential function
1.3 Roots of unity
1.4 Complex logarithm and power
1.5 De Moivre’s theorem
1.6 Lines and circles in C

2 Vectors
2.1 Definition and basic properties
2.2 Scalar product
2.2.1 Geometric picture (R2 and R3 only)
2.2.2 General algebraic definition
2.3 Cauchy-Schwarz inequality
2.4 Vector product
2.5 Scalar triple product
2.6 Spanning sets and bases
2.6.1 2D space
2.6.2 3D space
2.6.3 Rn space
2.6.4 Cn space
2.7 Vector subspaces
2.8 Suffix notation
2.9 Geometry
2.9.1 Lines
2.9.2 Plane
2.10 Vector equations

3 Linear maps
3.1 Examples
3.1.1 Rotation in R3
3.1.2 Reflection in R3
3.2 Linear Maps
3.3 Rank and nullity
3.4 Matrices
3.4.1 Examples
3.4.2 Matrix Algebra
3.4.3 Decomposition of an n × n matrix
3.4.4 Matrix inverse
3.5 Determinants
3.5.1 Permutations
3.5.2 Properties of determinants
3.5.3 Minors and Cofactors

4 Matrices and linear equations

4.1 Simple example, 2 × 2
4.2 Inverse of an n × n matrix
4.3 Homogeneous and inhomogeneous equations
4.3.1 Gaussian elimination
4.4 Matrix rank
4.5 Homogeneous problem Ax = 0
4.5.1 Geometrical interpretation
4.5.2 Linear mapping view of Ax = 0
4.6 General solution of Ax = d

5 Eigenvalues and eigenvectors

5.1 Preliminaries and definitions
5.2 Linearly independent eigenvectors
5.3 Transformation matrices
5.3.1 Transformation law for vectors
5.3.2 Transformation law for matrix
5.4 Similar matrices
5.5 Diagonalizable matrices
5.6 Canonical (Jordan normal) form
5.7 Cayley-Hamilton Theorem
5.8 Eigenvalues and eigenvectors of a Hermitian matrix
5.8.1 Eigenvalues and eigenvectors
5.8.2 Gram-Schmidt orthogonalization (non-examinable)
5.8.3 Unitary transformation
5.8.4 Diagonalization of n × n Hermitian matrices
5.8.5 Normal matrices

6 Quadratic forms and conics

6.1 Quadrics and conics
6.1.1 Quadrics
6.1.2 Conic sections (n = 2)
6.2 Focus-directrix property

7 Transformation groups
7.1 Groups of orthogonal matrices
7.2 Length preserving matrices
7.3 Lorentz transformations
0 Introduction
Vectors and matrices is the language in which a lot of mathematics is written
in. In physics, many variables such as position and momentum are expressed as
vectors. Heisenberg also formulated quantum mechanics in terms of vectors and
matrices. In statistics, one might pack all the results of all experiments into a
single vector, and work with a large vector instead of many small quantities. In
group theory, matrices are used to represent the symmetries of space (as well as
many other groups).
So what is a vector? Vectors are very general objects, and can in theory
represent very complex objects. However, in this course, our focus is on vectors
in Rn or Cn . We can think of each of these as an array of n real or complex
numbers. For example, (1, 6, 4) is a vector in R3 . These vectors are added in the
obvious way. For example, (1, 6, 4) + (3, 5, 2) = (4, 11, 6). We can also multiply
vectors by numbers, say 2(1, 6, 4) = (2, 12, 8). Often, these vectors represent
points in an n-dimensional space.
Matrices, on the other hand, represent functions between vectors, i.e. a
function that takes in a vector and outputs another vector. These, however, are
not arbitrary functions. Instead matrices represent linear functions. These are
functions that satisfy the equality f (λx + µy) = λf (x) + µf (y) for arbitrary
numbers λ, µ and vectors x, y. It is important to note that the function x 7→ x+c
for some constant vector c is not linear according to this definition, even though
it might look linear.
It turns out that for each linear function from Rn to Rm , we can represent
the function uniquely by an m × n array of numbers, which is what we call the
matrix. Expressing a linear function as a matrix allows us to conveniently study
many of its properties, which is why we usually talk about matrices instead of
the function itself.
1 Complex numbers
In R, not every polynomial equation has a solution. For example, there does
not exist any x such that x2 + 1 = 0, since for any x, x2 is non-negative, and
x2 + 1 can never be 0. To solve this problem, we introduce the “number” i that
satisfies i2 = −1. Then i is a solution to the equation x2 + 1 = 0. Similarly, −i
is also a solution to the equation.
We can add and multiply numbers with i. For example, we can obtain
numbers 3 + i or 1 + 3i. These numbers are known as complex numbers. It turns
out that by adding this single number i, every polynomial equation will have a
root. In fact, for an nth order polynomial equation, we will later see that there
will always be n roots, if we account for multiplicity. We will go into details in
Chapter 5.
Apart from solving equations, complex numbers have a lot of rather important
applications. For example, they are used in electronics to represent alternating
currents, and form an integral part in the formulation of quantum mechanics.
1.1 Basic properties
Definition (Complex number). A complex number is a number z ∈ C of the
form z = a + ib with a, b ∈ R, where i2 = −1. We write a = Re(z) and b = Im(z).
We have

z1 ± z2 = (a1 + ib1 ) ± (a2 + ib2 )

= (a1 ± a2 ) + i(b1 ± b2 )
z1 z2 = (a1 + ib1 )(a2 + ib2 )
= (a1 a2 − b1 b2 ) + i(b1 a2 + a1 b2 )
1
z −1 =
a + ib
a − ib
= 2
a + b2
Definition (Complex conjugate). The complex conjugate of z = a + ib is a − ib.
It is written as z̄ or z ∗ .
It is often helpful to visualize complex numbers in a diagram:
Definition (Argand diagram). An Argand diagram is a diagram   in which a
x
complex number z = x + iy is represented by a vector p = . Addition of
y
vectors corresponds to vector addition and z̄ is the reflection of z in the x-axis.

Im
z1 + z2
z1
z2

Re
z̄2

Definition (Moduluspand argument of complex number). The modulus of

z = x + iy is r = |z| = x2 + y 2 . The argument is θ = arg z = tan−1 (y/x). The
modulus is the length of the vector in the Argand diagram, and the argument is
the angle between z and the real axis. We have

z = r(cos θ + i sin θ)

Clearly the pair (r, θ) uniquely describes a complex number z, but each complex
number z ∈ C can be described by many different θ since sin(2π + θ) = sin θ
and cos(2π + θ) = cos θ. Often we take the principle value θ ∈ (−π, π].
When writing zi = ri (cos θi + i sin θi ), we have

z1 z2 = r1 r2 [(cos θ1 cos θ2 − sin θ1 sin θ2 ) + i(sin θ1 cos θ2 + sin θ2 cos θ1 )]

= r1 r2 [cos(θ1 + θ2 ) + i sin(θ1 + θ2 )]

In other words, when multiplying complex numbers, the moduli multiply and
the arguments add.
Proposition. z z̄ = a2 + b2 = |z|2 .

Proposition. z −1 = z̄/|z|2 .
Theorem (Triangle inequality). For all z1 , z2 ∈ C, we have

|z1 + z2 | ≤ |z1 | + |z2 |.

Alternatively, we have |z1 − z2 | ≥ ||z1 | − |z2 ||.

1.2 Complex exponential function
Exponentiation was originally defined for integer powers as repeated multiplica-
tion. This is then extended to rational powers using roots. We can also extend
this to any real number since real numbers can be approximated arbitrarily
accurately by rational numbers. However, what does it mean to take an exponent
of a complex number?
To do so, we use the Taylor series definition of the exponential function:
Definition (Exponential function). The exponential function is defined as
∞
z2 z3 X zn
exp(z) = ez = 1 + z + + + ··· = .
2! 3! n=0
n!

This automatically allows taking exponents of arbitrary complex numbers.

Having defined exponentiation this way, we want to check that it satisfies the
usual properties, such as exp(z + w) = exp(z) exp(w). To prove this, we will
first need a helpful lemma.
Lemma.
∞ X
X ∞ ∞ X
X r
amn = ar−m,m
n=0 m=0 r=0 m=0

Proof.
∞ X
X ∞
amn = a00 + a01 + a02 + · · ·
n=0 m=0
+ a10 + a11 + a12 + · · ·
+ a20 + a21 + a22 + · · ·
= (a00 ) + (a10 + a01 ) + (a20 + a11 + a02 ) + · · ·
X∞ X r
= ar−m,m
r=0 m=0

This is not exactly a rigorous proof, since we should not hand-wave about
infinite sums so casually. But in fact, we did not even show that the definition of
exp(z) is well defined for all numbers z, since the sum might diverge. All these
will be done in that IA Analysis I course.
Theorem. exp(z1 ) exp(z2 ) = exp(z1 + z2 )
Proof.
∞ X ∞
X z1m z2n
exp(z1 ) exp(z2 ) =
n=0 m=0
m! n!
∞ X
r
X z1r−m z2m
=
r=0 m=0
(r − m)! m!
∞ r
X 1 X r!
= z1r−m z2m
r=0
r! m=0
(r − m)!m!
∞
X (z1 + z2 )r
=
r=0
r!

Again, to define the sine and cosine functions, instead of referring to “angles”
(since it doesn’t make much sense to refer to complex “angles”), we again use a
series definition.
Definition (Sine and cosine functions). Define, for all z ∈ C,
∞
X (−1)n 2n+1 1 1
sin z = z = z − z3 + z5 + · · ·
n=0
(2n + 1)! 3! 5!
∞
X (−1)n 2n 1 2 1
cos z = z =1− z + z4 + · · ·
n=0
(2n)! 2! 4!

One very important result is the relationship between exp, sin and cos.
Theorem. eiz = cos z + i sin z.
Alternatively, since sin(−z) = − sin z and cos(−z) = cos z, we have

eiz + e−iz
cos z = ,
2
eiz − e−iz
sin z = .
2i
Proof.
∞ n
X i n
eiz = z
n=0
n!
∞ ∞
X i2n 2n X i2n+1 2n+1
= z + z
n=0
(2n)! n=0
(2n + 1)!
∞ ∞
X (−1)n 2n X (−1)n 2n+1
= z +i z
n=0
(2n)! n=0
(2n + 1)!
= cos z + i sin z

Thus we can write z = r(cos θ + i sin θ) = reiθ .

1.3 Roots of unity
Definition (Roots of unity). The nth roots of unity are the roots to the equation
z n = 1 for n ∈ N. Since this is a polynomial of order n, there are n roots of
unity. In fact, the nth roots of unity are exp 2πi nk for k = 0, 1, 2, 3 · · · n − 1.

Proposition. If ω = exp 2πi

2 n−1
n , then 1 + ω + ω + · · · + ω =0
Proof. Two proofs are provided:

(i) Consider the equation z n = 1. The coefficient of z n−1 is the sum of

all roots. Since the coefficient of z n−1 is 0, then the sum of all roots
= 1 + ω + ω 2 + · · · + ω n−1 = 0.
(ii) Since ω n − 1 = (ω − 1)(1 + ω + · · · + ω n−1 ) and ω 6= 1, dividing by (ω − 1),
we have 1 + ω + · · · + ω n−1 = (ω n − 1)/(ω − 1) = 0.
1.4 Complex logarithm and power
Definition (Complex logarithm). The complex logarithm w = log z is a solution
to eω = z, i.e. ω = log z. Writing z = reiθ , we have log z = log(reiθ ) = log r + iθ.
This can be multi-valued for different values of θ and, as above, we should select
the θ that satisfies −π < θ ≤ π.
Example. log 2i = log 2 + i π2

Definition (Complex power). The complex power z α for z, α ∈ C is defined as

z α = eα log z . This, again, can be multi-valued, as z α = eα log |z| eiαθ e2inπα (there
are finitely many values if α ∈ Q, infinitely many otherwise). Nevertheless, we
make z α single-valued by insisting −π < θ ≤ π.
1.5 De Moivre’s theorem
Theorem (De Moivre’s theorem).

cos nθ + i sin nθ = (cos θ + i sin θ)n .

Proof. First prove for the n ≥ 0 case by induction. The n = 0 case is true since
it merely reads 1 = 1. We then have

(cos θ + i sin θ)n+1 = (cos θ + i sin θ)n (cos θ + i sin θ)

= (cos nθ + i sin nθ)(cos θ + i sin θ)
= cos(n + 1)θ + i sin(n + 1)θ

If n < 0, let m = −n. Then m > 0 and

(cosθ + i sin θ)−m = (cos mθ + i sin mθ)−1

cos mθ − i sin mθ
=
(cos mθ + i sin mθ)(cos mθ − i sin mθ)
cos(−mθ) + i sin(−mθ)
=
cos2 mθ + sin2 mθ
= cos(−mθ) + i sin(−mθ)
= cos nθ + i sin nθ

Note that “cos nθ + i sin nθ = einθ = (eiθ )n = (cos θ + i sin θ)n ” is not a valid
proof of De Moivre’s theorem, since we do not know yet that einθ = (eiθ )n . In
fact, De Moivre’s theorem tells us that this is a valid rule to apply.
Example. We have cos 5θ + i sin 5θ = (cos θ + i sin θ)5 . By binomial expansion
of the RHS and taking real and imaginary parts, we have

cos 5θ = 5 cos θ − 20 cos3 θ + 16 cos5 θ

sin 5θ = 5 sin θ − 20 sin3 θ + 16 sin5 θ
1.6 Lines and circles in C
Since complex numbers can be regarded as points on the 2D plane, we can often
use complex numbers to represent two dimensional objects.
Suppose that we want to represent a straight line through z0 ∈ C parallel to
w ∈ C. The obvious way to do so is to let z = z0 + λw where λ can take any
real value. However, this is not an optimal way of doing so, since we are not
using the power of complex numbers fully. This is just the same as the vector
equation for straight lines, which you may or may not know from your A levels.
Instead, we arrange the equation to give λ = z−z w . We take the complex
0

z̄−z¯0
conjugate of this expression to obtain λ̄ = w̄ . The trick here is to realize that
λ is a real number. So we must have λ = λ̄. This means that we must have
z − z0 z̄ − z¯0
=
w w̄
z w̄ − z̄w = z0 w̄ − z̄0 w.

Theorem (Equation of straight line). The equation of a straight line through

z0 and parallel to w is given by

z w̄ − z̄w = z0 w̄ − z̄0 w.

The equation of a circle, on the other hand, is rather straightforward. Suppose

that we want a circle with center c ∈ C and radius ρ ∈ R+ . By definition of a
circle, a point z is on the circle iff its distance to c is ρ, i.e. |z − c| = ρ. Recalling
that |z|2 = z z̄, we obtain,

|z − c| = ρ
|z − c|2 = ρ2
(z − c)(z̄ − c̄) = ρ2
z z̄ − c̄z − cz̄ = ρ2 − cc̄

Theorem. The general equation of a circle with center c ∈ C and radius ρ ∈ R+

can be given by
z z̄ − c̄z − cz̄ = ρ2 − cc̄.
2 Vectors
We might have first learned vectors as arrays of numbers, and then defined
addition and multiplication in terms of the individual numbers in the vector.
This however, is not what we are going to do here. The array of numbers is just
a representation of the vector, instead of the vector itself.
Here, we will define vectors in terms of what they are, and then the various
operations are defined axiomatically according to their properties.
2.1 Definition and basic properties
Definition (Vector). A vector space over R or C is a collection of vectors v ∈ V ,
together with two operations: addition of two vectors and multiplication of a
vector with a scalar (i.e. a number from R or C, respectively).
Vector addition has to satisfy the following axioms:
(i) a + b = b + a (commutativity)
(ii) (a + b) + c = a + (b + c) (associativity)

(iii) There is a vector 0 such that a + 0 = a. (identity)

(iv) For all vectors a, there is a vector (−a) such that a + (−a) = 0 (inverse)
Scalar multiplication has to satisfy the following axioms:

(i) λ(a + b) = λa + λb.

(ii) (λ + µ)a = λa + µa.
(iii) λ(µa) = (λµ)a.
(iv) 1a = a.

Often, vectors have a length and direction. The length is denoted by |v|. In
this case, we can think of a vector as an “arrow” in space. Note that λa is either
parallel (λ ≥ 0) to or anti-parallel (λ ≤ 0) to a.
Definition (Unit vector). A unit vector is a vector with length 1. We write a
unit vector as v̂.

Example. Rn is a vector space with component-wise addition and scalar mul-

tiplication. Note that the vector space R is a line, but not all lines are vector
spaces. For example, x + y = 1 is not a vector space since it does not contain 0.
2.2 Scalar product
In a vector space, we can define the scalar product of two vectors, which returns
a scalar (i.e. a real or complex number). We will first look at the usual scalar
product defined for Rn , and then define the scalar product axiomatically.

2.2.1 Geometric picture (R2 and R3 only)

Definition (Scalar/dot product). a · b = |a||b| cos θ, where θ is the angle
between a and b. It satisfies the following properties:
(i) a · b = b · a
2
(ii) a · a = |a| ≥ 0
(iii) a · a = 0 iff a = 0
(iv) If a · b = 0 and a, b 6= 0, then a and b are perpendicular.

Intuitively, this is the product of the parts of a and b that are parallel.
a

|a|

b
|a| cos θ

Using the dot product, we can write the projection of b onto a as (|b| cos θ)â =
(â · b)â.
The cosine rule can be derived as follows:
−−→ −−→ −→
|BC|2 = |AB + AC|2
−−→ −→ −−→ −→
= (AB + AC) · (AB + AC)
−−→ −→ −−→ −→
= |AB|2 + |AC|2 − 2|AB||AC| cos θ

We will later come up with a convenient algebraic way to evaluate this scalar
product.

2.2.2 General algebraic definition

Definition (Inner/scalar product). In a real vector space V , an inner product
or scalar product is a map V × V → R that satisfies the following axioms. It is
written as x · y or hx | yi.
(i) x · y = y · x (symmetry)

(ii) x · (λy + µz) = λx · y + µx · z (linearity in 2nd argument)

(iii) x · x ≥ 0 with equality iff x = 0 (positive definite)
Note that this is a definition only for real vector spaces, where the scalars
are real. We will have a different set of definitions for complex vector spaces.
In particular, here we can use (i) and (ii) together to show linearity in 1st
argument. However, this is generally not true for complex vector spaces.
Definition. The norm of a vector, written as |a| or kak, is defined as
√
|a| = a · a.

Example. Instead of the usual Rn vector space, we can consider the set of all
real (integrable) functions as a vector space. We can define the following inner
product: Z 1
hf | gi = f (x)g(x) dx.
0
2.3 Cauchy-Schwarz inequality
Theorem (Cauchy-Schwarz inequality). For all x, y ∈ Rn ,

|x · y| ≤ |x||y|.

Proof. Consider the expression |x − λy|2 . We must have

|x − λy|2 ≥ 0
(x − λy) · (x − λy) ≥ 0
λ |y| − λ(2x · y) + |x|2 ≥ 0.
2 2

Viewing this as a quadratic in λ, we see that the quadratic is non-negative and

thus cannot have 2 real roots. Thus the discriminant ∆ ≤ 0. So

4(x · y)2 ≤ 4|y|2 |x|2

(x · y)2 ≤ |x|2 |y|2
|x · y| ≤ |x||y|.

Note that we proved this using the axioms of the scalar product. So this
result holds for all possible scalar products on any (real) vector space.
Example. Let x = (α, β, γ) and y = (1, 1, 1). Then by the Cauchy-Schwarz
inequality, we have
√ p
α + β + γ ≤ 3 α2 + β 2 + γ 2
α2 + β 2 + γ 2 ≥ αβ + βγ + γα,

with equality if α = β = γ.
Corollary (Triangle inequality).

|x + y| ≤ |x| + |y|.

Proof.

|x + y|2 = (x + y) · (x + y)
= |x|2 + 2x · y + |y|2
≤ |x|2 + 2|x||y| + |y|2
= (|x| + |y|)2 .

So

|x + y| ≤ |x| + |y|.
2.4 Vector product
Apart from the scalar product, we can also define the vector product. However,
this is defined only for R3 space, but not spaces in general.

Definition (Vector/cross product). Consider a, b ∈ R3 . Define the vector

product
a × b = |a||b| sin θn̂,
where n̂ is a unit vector perpendicular to both a and b. Since there are two
(opposite) unit vectors that are perpendicular to both of them, we pick n̂ to be
the one that is perpendicular to a, b in a right-handed sense.

a×b

The vector product satisfies the following properties:

(i) a × b = −b × a.
(ii) a × a = 0.
(iii) a × b = 0 ⇒ a = λb for some λ ∈ R (or b = 0).

(iv) a × (λb) = λ(a × b).

(v) a × (b + c) = a × b + a × c.
−→ −−→ −→ −−→
If we have a triangle OAB, its area is given by 12 |OA||OB| sin θ = 12 |OA× OB|.
−→ −−→
We define the vector area as 12 OA × OB, which is often a helpful notion when
we want to do calculus with surfaces.
There is a convenient way of calculating vector products:
Proposition.

a × b = (a1 î + a2 ĵ + a3 k̂) × (b1 î + b2 ĵ + b3 k̂)

= (a2 b3 − a3 b2 )î + · · ·
 
 î ĵ k̂ 

= a1 a2 a3 
 b1 b2 b3 
2.5 Scalar triple product
Definition (Scalar triple product). The scalar triple product is defined as

[a, b, c] = a · (b × c).

Proposition. If a parallelepiped has sides represented by vectors a, b, c that

form a right-handed system, then the volume of the parallelepiped is given by
[a, b, c].

Proof. The area of the base of the parallelepiped is given by |b||c| sin θ = |b × c|.
Thus the volume= |b × c||a| cos φ = |a · (b × c)|, where φ is the angle between
a and the normal to b and c. However, since a, b, c form a right-handed system,
we have a · (b × c) ≥ 0. Therefore the volume is a · (b × c).

Since the order of a, b, c doesn’t affect the volume, we know that

[a, b, c] = [b, c, a] = [c, a, b] = −[b, a, c] = −[a, c, b] = −[c, b, a].

Theorem. a × (b + c) = a × b + a × c.

Proof. Let d = a × (b + c) − a × b − a × c. We have

d · d = d · [a × (b + c)] − d · (a × b) − d · (a × c)
= (b + c) · (d × a) − b · (d × a) − c · (d × a)
=0

Thus d = 0.
2.6 Spanning sets and bases
2.6.1 2D space
Definition (Spanning set). A set of vectors {a, b} spans R2 if for all vectors
r ∈ R2 , there exist some λ, µ ∈ R such that r = λa + µb.
In R2 , two vectors span the space if a × b 6= 0.

Theorem. The coefficients λ, µ are unique.

Proof. Suppose that r = λa + µb = λ0 a + µ0 b. Take the vector product with a
on both sides to get (µ − µ0 )a × b = 0. Since a × b 6= 0, then µ = µ0 . Similarly,
λ = λ0 .
Definition (Linearly independent vectors in R2 ). Two vectors a and b are
linearly independent if for α, β ∈ R, αa + βb = 0 iff α = β = 0. In R2 , a and b
are linearly independent if a × b 6= 0.
Definition (Basis of R2 ). A set of vectors is a basis of R2 if it spans R2 and
are linearly independent.

Example. {î, ĵ} = {(1, 0), (0, 1)} is a basis of R2 . They are the standard basis
of R2 .

2.6.2 3D space
We can extend the above definitions of spanning set and linear independent set
to R3 . Here we have
Theorem. If a, b, c ∈ R3 are non-coplanar, i.e. a · (b × c) 6= 0, then they form
a basis of R3 .
Proof. For any r, write r = λa + µb + νc. Performing the scalar product
with b × c on both sides, one obtains r · (b × c) = λa · (b × c) + µb · (b × c) +
νc · (b × c) = λ[a, b, c]. Thus λ = [r, b, c]/[a, b, c]. The values of µ and ν can
be found similarly. Thus each r can be written as a linear combination of a, b
and c.
By the formula derived above, it follows that if αa + βb + γc = 0, then
α = β = γ = 0. Thus they are linearly independent.
Note that while we came up with formulas for λ, µ and ν, we did not actually
prove that these coefficients indeed work. This is rather unsatisfactory. We
could, of course, expand everything out and show that this indeed works, but
in IB Linear Algebra, we will prove a much more general result, saying that if
we have an n-dimensional space and a set of n linear independent vectors, then
they form a basis.
In R3 , the standard basis is î, ĵ, k̂, or (1, 0, 0), (0, 1, 0) and (0, 0, 1).

2.6.3 Rn space
In general, we can define

Definition (Linearly independent vectors). A set of vectors {v1 , v2 , v3 · · · vm }

is linearly independent if
m
X
λi vi = 0 ⇒ (∀i) λi = 0.
i=1

Definition (Spanning set). A set of vectors {u1 , u2 , u3 · · · um } ⊆ Rn is a

spanning set of Rn if
n
X
(∀x ∈ Rn )(∃λi ) λi ui = x
i=1

Definition (Basis vectors). A basis of Rn is a linearly independent spanning

set. The standard basis of Rn is e1 = (1, 0, 0, · · · 0), e2 = (0, 1, 0, · · · 0), · · · en =
(0, 0, 0, · · · , 1).

Definition (Orthonormal basis). A basis {ei } is orthonormal if ei · ej = 0 if

i 6= j and ei · ei = 1 for all i, j.
Using the Kronecker Delta symbol, which we will define later, we can write
this condition as ei · ej = δij .
Definition (Dimension of vector space). The dimension of a vector space is
the number of vectors in its basis. (Exercise: show that this is well-defined)
We usually denote the components of a vector x by xi . So we have x =
(x1 , x2 , · · · , xn ).
n
P (Scalar product). The scalar product of x, y ∈ R is defined as
Definition
x·y = xi yi .
The reader should check that this definition coincides with the |x||y| cos θ
definition in the case of R2 and R3 .

2.6.4 Cn space
Cn is very similar to Rn , except that we have complex numbers. As a result, P we
need a different definition of the scalar product. If we still defined u · v = ui vi ,
then if we let u = (0, i), then u · u = −1 < 0. This would be bad if we want to
use the scalar product to define a norm.
Definition (Cn ). Cn = {(z1 , z2 , · · · , zn ) : zi ∈ C}. It has the same standard
n n
P ∗ as R but the scalar product is defined differently. For u, v ∈ C , u · v =
basis
ui vi . The scalar product has the following properties:

(i) u · v = (v · u)∗
(ii) u · (λv + µw) = λ(u · v) + µ(u · w)
(iii) u · u ≥ 0 and u · u = 0 iff u = 0

Instead of linearity in the first argument, here we have (λu + µv) · w =

λ∗ u · w + µ∗ v · w.
Example.
4
X
(−i)k |x + ik y|2
k=1
X
= (−i)k hx + ik y | x + ik yi
X
= (−i)k (hx + ik y | xi + ik hx + ik y | yi)
X
= (−i)k (hx | xi + (−i)k hy | xi + ik hx | yi + ik (−i)k hy | yi)
X
= (−i)k [(|x|2 + |y|2 ) + (−1)k hy | xi + hx | yi]
X X X
= (|x|2 + |y|2 ) (−i)k + hy | xi (−1)k + hx | yi 1
= 4hx | yi.

We can prove the Cauchy-Schwarz inequality for complex vector spaces using
the same proof as the real case, except that this time we have to first multiply y
by some eiθ so that x · (eiθ y) is a real number. The factor of eiθ will drop off at
the end when we take the modulus signs.
2.7 Vector subspaces
Definition (Vector subspace). A vector subspace of a vector space V is a subset
of V that is also a vector space under the same operations. Both V and {0} are
subspaces of V . All others are proper subspaces.
A useful criterion is that a subset U ⊆ V is a subspace iff

(i) x, y ∈ U ⇒ (x + y) ∈ U .
(ii) x ∈ U ⇒ λx ∈ U for all scalars λ.
(iii) 0 ∈ U .

This can be more concisely written as “U is non-empty and for all x, y ∈ U ,

(λx + µy) ∈ U ”.
Example.
(i) If {a, b, c} is a basis of R3 , then {a + c, b + c} is a basis of a 2D subspace.
Suppose x, y ∈ span{a + c, b + c}. Let

x = α1 (a + c) + β1 (b + c);
y = α2 (a + c) + β2 (b + c).

Then

λx + µy = (λα1 + µα2 )(a + c) + (λβ1 + µβ2 )(b + c) ∈ span{a + c, b + c}.

Thus this is a subspace of R3 .

Now check that a + c, b + c is a basis. We only need to check linear
independence. If α(a + c) + β(b + c) = 0, then αa + βb + (α + β)c = 0.
Since {a, b, c} is a basis of R3 , therefore a, b, c are linearly independent
and α = β = 0. Therefore a + c, b + c is a basis and the subspace has
dimension 2.
Pn
(ii) Given a set of numbers αi , let U = {x ∈ Rn : i=1 αi xi = 0}. We show
is a vector subspace ofPRn : Take x,
that this P Py ∈ U , then consider λx + µy.
We have αi (λxi + µyi ) = λ αi xi + µ αi yi = 0. Thus λx + µy ∈ U .
The dimension of the subspace is n − 1 as we can freely choose xi for
i = 1, · · · , n − 1 and then xn is uniquely determined by the previous xi ’s.
(iii) Let W = {x ∈ Rn :
P P
αi xi = 1}. Then 6 1.
αi (λxi + µyi ) = λ + µ =
Therefore W is not a vector subspace.
2.8 Suffix notation
Here we are going to introduce a powerful notation that can help us simplify a
lot of things.
First of all, let v ∈ R3 . We can write v = v1 e1 + v2 e2 + v3 e3 = (v1 , v2 , v3 ).
So in general, the ith component of v is written as vi . We can thus write
vector equations in component form. For example, a = b → ai = bi or
c = αa + βb → ci = αai + βbi . A vector has one free suffix, i, while a scalar
has none.
P
Notation (Einstein’s summation convention). ConsiderP a sum x · y = xi yi .
The summation convention says that we can drop the symbol and simply
write x · y = xi yi . If suffixes are repeated once, summation is understood.
Note that i is a dummy suffix and doesn’t matter what it’s called, i.e.
xi yi = xj yj = xk yk etc.
The rules of this convention are:
(i) Suffix appears once in a term: free suffix

(ii) Suffix appears twice in a term: dummy suffix and is summed over
(iii) Suffix appears three times or more: WRONG!
Example. [(a · b)c − (a · c)b]i = aj bj ci − aj cj bi summing over j understood.

It is possible for an item to have more than one index. These objects are
known as tensors, which will be studied in depth in the IA Vector Calculus
course.
Here we will define two important tensors:
Definition (Kronecker delta).
(
1 i=j
δij = .
0 i 6= j

We have    
δ11 δ12 δ13 1 0 0
δ21 δ22 δ23  = 0 1 0 = I.
δ31 δ32 δ33 0 0 1
So the Kronecker delta represents an identity matrix.
Example.

(i) ai δi1 = a1 . In general, ai δij = aj (i is dummy, j is free).

(ii) δij δjk = δik
(iii) δii = n if we are in Rn .
(iv) ap δpq bq = ap bp with p, q both dummy suffices and summed over.

Definition (Alternating symbol εijk ). Consider rearrangements of 1, 2, 3. We

can divide them into even and odd permutations. Even permutations include
(1, 2, 3), (2, 3, 1) and (3, 1, 2). These are permutations obtained by performing
two (or no) swaps of the elements of (1, 2, 3). (Alternatively, it is any “rotation”
of (1, 2, 3))
The odd permutations are (2, 1, 3), (1, 3, 2) and (3, 2, 1). They are the
permutations obtained by one swap only.
Define 
+1 ijk is even permutation

εijk = −1 ijk is odd permutation

0 otherwise (i.e. repeated suffices)


εijk has 3 free suffices.

We have ε123 = ε231 = ε312 = +1 and ε213 = ε132 = ε321 = −1. ε112 =
ε111 = · · · = 0.

We have
(i) εijk δjk = εijj = 0
(ii) If ajk = akj (i.e. aij is symmetric), then εijk ajk = εijk akj = −εikj akj .
Since εijk ajk = εikj akj (we simply renamed dummy suffices), we have
εijk ajk = 0.

Proposition. (a × b)i = εijk aj bk

Proof. By expansion of formula
Theorem. εijk εipq = δjp δkq − δjq δkp

Proof. Proof by exhaustion:


+1
 if j = p and k = q
RHS = −1 if j = q and k = p

0 otherwise


LHS: Summing over i, the only non-zero terms are when j, k = 6 i and p, q 6= i.
If j = p and k = q, LHS is (−1)2 or (+1)2 = 1. If j = q and k = p, LHS is
(+1)(−1) or (−1)(+1) = −1. All other possibilities result in 0.
Equally, we have εijk εpqk = δip δjq − δjp δiq and εijk εpjq = δip δkq − δiq δkp .
Proposition.
a · (b × c) = b · (c × a)

Proof. In suffix notation, we have

a · (b × c) = ai (b × c)i = εijk bj ck ai = εjki bj ck ai = b · (c × a).

Theorem (Vector triple product).

a × (b × c) = (a · c)b − (a · b)c.

Proof.

[a × (b × c)]i = εijk aj (b × c)k

= εijk εkpq aj bp cq
= εijk εpqk aj bp cq
= (δip δjq − δiq δjp )aj bp cq
= aj bi cj − aj ci bj
= (a · c)bi − (a · b)ci

Similarly, (a × b) × c = (a · c)b − (b · c)a.

Spherical trigonometry
Proposition. (a × b) · (b × c) = (a · a)(b · c) − (a · b)(a · c).
Proof.

LHS = (a × b)i (a × c)i

= εijk aj bk εipq ap cq
= (δjp δkq − δjq δkp )aj bk ap cq
= aj bk aj ck − aj bk ak cj
= (a · a)(b · c) − (a · b)(a · c)

Consider the unit sphere, center O, with a, b, c on the surface.

δ(A, B) α

B C

Suppose we are living on the surface of the sphere. So the distance from A to B is
the arc length on the sphere. We can imagine this to be along the circumference
of the circle through A and B with center O. So the distance is ∠AOB, which we
shall denote by δ(A, B). So a · b = cos ∠AOB = cos δ(A, B). We obtain similar
expressions for other dot products. Similarly, we get |a × b| = sin δ(A, B).

(a × b) · (a × c)
cos α =
|a × b||a × c|
b · c − (a · b)(a · c)
=
|a × b||a × c|

Putting in our expressions for the dot and cross products, we obtain

cos α sin δ(A, B) sin δ(A, C) = cos δ(B, C) − cos δ(A, B) cos δ(A, C).

This is the spherical cosine rule that applies when we live on the surface of a
sphere. What does this spherical geometry look like?
Consider a spherical equilateral triangle. Using the spherical cosine rule,

cos δ − cos2 δ 1
cos α = =1− .
sin2 δ 1 + cos δ

Since cos δ ≤ 1, we have cos α ≤ 12 and α ≥ 60◦ . Equality holds iff δ = 0, i.e. the
triangle is simply a point. So on a sphere, each angle of an equilateral triangle is
greater than 60◦ , and the angle sum of a triangle is greater than 180◦ .
2.9 Geometry
2.9.1 Lines
Any line through a and parallel to t can be written as

x = a + λt.

By crossing both sides of the equation with t, we have

Theorem. The equation of a straight line through a and parallel to t is

(x − a) × t = 0 or x × t = a × t.

2.9.2 Plane
To define a plane Π, we need a normal n to the plane and a fixed point b. For
any x ∈ Π, the vector x − b is contained in the plane and is thus normal to n,
i.e. (x − b) · n = 0.
Theorem. The equation of a plane through b with normal n is given by

x · n = b · n.

If n = n̂ is a unit normal, then d = x · n̂ = b · n̂ is the perpendicular distance

from the origin to Π.
Alternatively, if a, b, c lie in the plane, then the equation of the plane is

(x − a) · [(b − a) × (c − a)] = 0.

Example.
(i) Consider the intersection between a line x × t = a × t with the plane
x · n = b · n. Cross n on the right with the line equation to obtain

(x · n)t − (t · n)x = (a × t) × n

Eliminate x · n using x · n = b · n

(t · n)x = (b · n)t − (a × t) × n

Provided t · n is non-zero, the point of intersection is

(b · n)t − (a × t) × n
x= .
t·n
Exercise: what if t · n = 0?
(ii) Shortest distance between two lines. Let L1 be (x − a1 ) × t1 = 0 and L2
be (x − a2 ) × t2 = 0.
The distance of closest approach s is along a line perpendicular to both L1
and L2 , i.e. the line of closest approach is perpendicular to both lines and
thus parallel to t1 × t2 . The distance
 s can then be found by projecting
a1 − a2 onto t1 × t2 . Thus s = (a1 − a2 ) · |tt11 ×t2 

×t2 | .
2.10 Vector equations
Example. x − (x × a) × b = c. Strategy: take the dot or cross of the equation
with suitable vectors. The equation can be expanded to form

x − (x · b)a + (a · b)x = c.

Dot this with b to obtain

x · b − (x · b)(a · b) + (a · b)(x · b) = c · b
x · b = c · b.

Substituting this into the original equation, we have

x(1 + a · b) = c + (c · b)a

If (1 + a · b) is non-zero, then

c + (c · b)a
x=
1+a·b
Otherwise, when (1 + a · b) = 0, if c + (c · b)a 6= 0, then a contradiction is
reached. Otherwise, x · b = c · b is the most general solution, which is a plane
of solutions.
3 Linear maps
A linear map is a special type of function between vector spaces. In fact, most
of the time, these are the only functions we actually care about. They are maps
that satisfy the property f (λa + µb) = λf (a) + µf (b).
We will first look at two important examples of linear maps — rotations and
reflections, and then study their properties formally.
3.1 Examples
3.1.1 Rotation in R3
In R3 , first consider the simple cases where we rotate about the z axis by θ. We
call this rotation R and write x0 = R(x).
Suppose that initially, x = (x, y, z) = (r cos φ, r sin φ, z). Then after a
rotation by θ, we get

x0 = (r cos(φ + θ), r sin(φ + θ), z)

= (r cos φ cos θ − r sin φ sin θ, r sin φ cos θ + r cos φ sin θ, z)
= (x cos θ − y sin θ, x sin θ + y cos θ, z).

We can represent this by a matrix R such that x0i = Rij xj . Using our formula
above, we obtain  
cos θ − sin θ 0
R =  sin θ cos θ 0
0 0 1
Now consider the general case where we rotate by θ about n̂.

n̂
A0
A0

θ
B A B A
x0 C C

O
−−→ −−→ −−→
We have x0 = OB + BC + CA0 . We know that
−−→
OB = (n̂ · x)n̂
−−→ −−→
BC = BA cos θ
−−→ −→
= (BO + OA) cos θ
= (−(n̂ · x)n̂ + x) cos θ
−→ −−→ −−→ −−→
Finally, to get CA, we know that |CA0 | = |BA0 | sin θ = |BA| sin θ = |n̂ × x| sin θ.
−−→ −−→
Also, CA0 is parallel to n̂ × x. So we must have CA0 = (n̂ × x) sin θ.
Thus x0 = x cos θ + (1 − cos θ)(n̂ · x)n̂ + n̂ × x sin θ. In components,

x0i = xi cos θ + (1 − cos θ)nj xj ni − εijk xj nk sin θ.

We want to find an R such that x0i = Rij xj . So

Rij = δij cos θ + (1 − cos θ)ni nj − εijk nk sin θ.

3.1.2 Reflection in R3
Suppose we want to reflect through a plane through O with normal n̂. First of
all the projection of x onto n̂ is given by (x · n̂)n̂. So we get x0 = x − 2(x · n̂)n̂.
In suffix notation, we have x0i = xi − 2xj nj ni . So our reflection matrix is
Rij = δij − 2ni nj .

n̂ x

x0
3.2 Linear Maps
Definition (Domain, codomain and image of map). Consider sets A and B
and mapping T : A → B such that each x ∈ A is mapped into a unique
x0 = T (x) ∈ B. A is the domain of T and B is the co-domain of T . Typically,
we have T : Rn → Rm or T : Cn → Cm .

Definition (Linear map). Let V, W be real (or complex) vector spaces, and
T : V → W . Then T is a linear map if
(i) T (a + b) = T (a) + T (b) for all a, b ∈ V .
(ii) T (λa) = λT (a) for all λ ∈ R (or C).

Equivalently, we have T (λa + µb) = λT (a) + µT (b).

Example.
(i) Consider a translation T : R3 → R3 with T (x) = x + a for some fixed,
given a. This is not a linear map since T (λx + µy) 6= λx + µy + (λ + µ)a.

(ii) Rotation, reflection and projection are linear transformations.

Definition (Image and kernel of map). The image of a map f : U → V is the
subset of V {f (u) : u ∈ U }. The kernel is the subset of U {u ∈ U : f (u) = 0}.
Example.

(i) Consider S : R3 → R2 with S(x, y, z) = (x + y, 2x − z). Simple yet

tedious algebra shows that this is linear. Now consider the effect of S on
the standard basis. S(1, 0, 0) = (1, 2), S(0, 1, 0) = (1, 0) and S(0, 0, 1) =
(0, −1). Clearly these are linearly dependent, but they do span the whole
of R2 . We can say S(R3 ) = R2 . So the image is R2 .
Now solve S(x, y, z) = 0. We need x + y = 0 and 2x − z = 0. Thus x =
(x, −x, 2x), i.e. it is parallel to (1, −1, 2). So the set {λ(1, −1, 2) : λ ∈ R}
is the kernel of S.
(ii) Consider a rotation in R3 . The kernel is the zero vector and the image is
R3 .

(iii) Consider a projection of x onto a plane with normal n̂. The image is the
plane itself, and the kernel is any vector parallel to n̂
Theorem. Consider a linear map f : U → V , where U, V are vector spaces.
Then im(f ) is a subspace of V , and ker(f ) is a subspace of U .

Proof. Both are non-empty since f (0) = 0.

If x, y ∈ im(f ), then ∃a, b ∈ U such that x = f (a), y = f (b). Then
λx + µy = λf (a) + µf (b) = f (λa + µb). Now λa + µb ∈ U since U is a vector
space, so there is an element in U that maps to λx + µy. So λx + µy ∈ im(f )
and im(f ) is a subspace of V .
Suppose x, y ∈ ker(f ), i.e. f (x) = f (y) = 0. Then f (λx + µy) = λf (x) +
µf (y) = λ0 + µ0 = 0. Therefore λx + µy ∈ ker(f ).
3.3 Rank and nullity
Definition (Rank of linear map). The rank of a linear map f : U → V , denoted
by r(f ), is the dimension of the image of f .

Definition (Nullity of linear map). The nullity of f , denoted n(f ) is the

dimension of the kernel of f .
Example. For the projection onto a plane in R3 , the image is the whole plane
and the rank is 2. The kernel is a line so the nullity is 1.

Theorem (Rank-nullity theorem). For a linear map f : U → V ,

r(f ) + n(f ) = dim(U ).

Proof. (Non-examinable) Write dim(U ) = n and n(f ) = m. If m = n, then f is

the zero map, and the proof is trivial, since r(f ) = 0. Otherwise, assume m < n.
Suppose {e1 , e2 , · · · , em } is a basis of ker f , Extend this to a basis of the
whole of U to get {e1 , e2 , · · · , em , em+1 , · · · , en }. To prove the theorem, we
need to prove that {f (em+1 ), f (em+2 ), · · · f (en )} is a basis of im(f ).
(i) First show that it spans im(f ). Take y ∈ im(f ). Thus ∃x ∈ U such that
y = f (x). Then

y = f (α1 e1 + α2 e2 + · · · + αn en ),

since e1 , · · · en is a basis of U . Thus

y = α1 f (e1 ) + α2 f (e2 ) + · · · + αm f (em ) + αm+1 f (em+1 ) + · · · + αn f (en ).

The first m terms map to 0, since e1 , · · · em is the basis of the kernel of f .

Thus
y = αm+1 f (em+1 ) + · · · + αn f (en ).

(ii) To show that they are linearly independent, suppose

αm+1 f (em+1 ) + · · · + αn f (en ) = 0.

Then
f (αm+1 em+1 + · · · + αn en ) = 0.
Thus αm+1 em+1 + · · · + αn en ∈ ker(f ). Since {e1 , · · · , em } span ker(f ),
there exist some α1 , α2 , · · · αm such that

αm+1 em+1 + · · · + αn en = α1 e1 + · · · + αm em .

But e1 · · · en is a basis of U and are linearly independent. So αi = 0 for all i.

Then the only solution to the equation αm+1 f (em+1 ) + · · · + αn f (en ) = 0
is αi = 0, and they are linearly independent by definition.

Example. Calculate the kernel and image of f : R3 → R3 , defined by

f (x, y, z) = (x + y + z, 2x − y + 5z, x + 2z).
First find the kernel: we’ve got the system of equations:

x+y+z =0
2x − y + 5z = 0
x + 2z = 0

Note that the first and second equation add to give 3x+6z = 0, which is identical
to the third. Then using the first and third equation, we have y = −x − z = z.
So the kernel is any vector in the form (−2z, z, z) and is the span of (−2, 1, 1).
To find the image, extend the basis of ker(f ) to a basis of the whole of R3 :
{(−2, 1, 1), (0, 1, 0), (0, 0, 1)}. Apply f to this basis to obtain (0, 0, 0), (1, −1, 0)
and (1, 5, 2). From the proof of the rank-nullity theorem, we know that f (0, 0, 1)
and f (0, 0, 1) is a basis of the image.
To get the standard form of the image, we know that the normal to the plane
is parallel to (1, −1, 0) × (1, 5, 2) k (1, 1, −3). Since 0 ∈ im(f ), the equation of
the plane is x + y − 3z = 0.
3.4 Matrices
In the examples above, we have represented our linear maps by some object R
such that x0i = Rij xj . We call R the matrix for the linear map. In general, let
α : Rn → Rm be a linear map, and x0 = α(x).
Let {ei } be a basis of Rn . Then x = xj ej for some xj . Then we get

x0 = α(xj ej ) = xj α(ej ).

So we get that
x0i = [α(ej )]i xj .
We now define Aij = [α(ej )]i . Then x0i = Aij xj . We write
 
A11 · · · A1n
A = {Aij } =  ... .. 

Aij . 
Am1 · · · Amn

Here Aij is the entry in the ith row of the jth column. We say that A is an
m × n matrix, and write x0 = Ax.
We see that the columns of the matrix are the images of the standard basis
vectors under the mapping α.
Example.

3.4.1 Examples
(i) In R2 , consider a reflection in a line with an angle θ to the x axis. We
know thatî 7→ cos 2θî + sin2θĵ , with ĵ 7→ − cos 2θĵ + sin 2θî. Then the
cos 2θ sin 2θ
matrix is .
sin 2θ − cos 2θ

(ii) In R3 , as we’ve previously seen, a rotation by θ about the z axis is given

by  
cos θ − sin θ 0
R =  sin θ cos θ 0
0 0 1

(iii) In R3 , a reflection in plane with normal n̂ is given by Rij = δij − 2n̂i n̂j .
Written as a matrix, we have

1 − 2n̂21 −2n̂1 n̂2

 
−2n̂1 n̂3
−2n̂2 n̂1 1 − 2n̂22 −2n̂2 n̂3 
−2n̂3 n̂1 −2n̂3 n̂2 1 − 2n̂23

(iv) Dilation (“stretching”) α : R3 → R3 is given by a map (x, y, z) 7→

(λx, µy, νz) for some λ, µ, ν. The matrix is
 
λ 0 0
 0 µ 0
0 0 ν

(v) Shear: Consider S : R3 → R3 that sheers in the x direction:

sheer in x direction
x x0

We have (x, y, z) 7→ (x + λy, y, z). Then

 
1 λ 0
S = 0 1 0
0 0 1

3.4.2 Matrix Algebra

This part is mostly on a whole lot of definitions, saying what we can do with
matrices and classifying them into different types.
Definition (Addition of matrices). Consider two linear maps α, β : Rn → Rm .
The sum of α and β is defined by

(α + β)(x) = α(x) + β(x)

In terms of the matrix, we have

(A + B)ij xj = Aij xj + Bij xj ,

or

(A + B)ij = Aij + Bij .

Definition (Scalar multiplication of matrices). Define (λα)x = λ[α(x)]. So

(λA)ij = λAij .

Definition (Matrix multiplication). Consider maps α : R` → Rn and β :

Rn → R m . The composition is βα : R` → Rm . Take x ∈ R` 7→ x00 ∈ Rm .
Then x00 = (BA)x = Bx0 , where x0 = Ax. Using suffix notation, we have
x00i = (Bx0 )i = bik x0k = Bik Akj xj . But x00i = (BA)ij xj . So

(BA)ij = Bik Akj .

Generally, an m × n matrix multiplied by an n × ` matrix gives an m × ` matrix.

(BA)ij is given by the ith row of B dotted with the jth column of A.
Note that the number of columns of B has to be equal to the number of rows
of A for multiplication to be defined. If ` = m as well, then both BA and AB
make sense, but AB 6= BA in general. In fact, they don’t even have to have the
same dimensions.
Also, since function composition is associative, we get A(BC) = (AB)C.
Definition (Transpose of matrix). If A is an m × n matrix, the transpose AT
is an n × m matrix defined by (AT )ij = Aji .

Proposition.
(i) (AT )T = A.


x1
 x2 
(ii) If x is a column vector . , xT is a row vector (x1 x2 · · · xn ).
 
 .. 
xn

(iii) (AB)T = B T AT since (AB)Tij = (AB)ji = Ajk Bki = Bki Ajk

= (B T )ik (AT )kj = (B T AT )ij .
Definition (Hermitian conjugate). Define A† = (AT )∗ . Similarly, (AB)† =
B † A† .

Definition (Symmetric matrix). A matrix is symmetric if AT = A.

Definition (Hermitian matrix). A matrix is Hermitian if A† = A. (The diagonal
of a Hermitian matrix must be real).
Definition (Anti/skew symmetric matrix). A matrix is anti-symmetric or skew
symmetric if AT = −A. The diagonals are all zero.

Definition (Skew-Hermitian matrix). A matrix is skew-Hermitian if A† = −A.

The diagonals are pure imaginary.
Definition (Trace of matrix). The trace of an n × n matrix A is the sum of the
diagonal. tr(A) = Aii .

Example. Consider the reflection matrix Rij = δij − 2n̂i n̂j . We have tr(A) =
Rii = 3 − 2n̂ · n̂ = 3 − 2 = 1.
Proposition. tr(BC) = tr(CB)
Proof. tr(BC) = Bik Cki = Cki Bik = (CB)kk = tr(CB)

Definition (Identity matrix). I = δij .

3.4.3 Decomposition of an n × n matrix

Any n × n matrix B can be split as a sum of symmetric and antisymmetric parts.
Write
1 1
Bij = (Bij + Bji ) + (Bij − Bji ) .
|2 {z } |2 {z }
Sij Aij

We have Sij = Sji , so S is symmetric, while Aji = −Aij , and A is antisymmetric.

So B = S + A.
Furthermore , we can decompose S into an isotropic part (a scalar multiple
of the identity) plus a trace-less part (i.e. sum of diagonal = 0). Write
1 1
Sij = tr(S)δij + (Sij − tr(S)δij ) .
|n {z } | n{z }
isotropic part Tij

We have tr(T ) = Tii = Sii − n1 tr(S)δii = tr(S) − n1 tr(S)(n) = 0.

Putting all these together,
 
1 1 1 1
B = tr(B)I + (B + B T ) − tr(B)I + (B − B T ).
n 2 n 2

In three dimensions, we can write the antisymmetric part A in terms of a single

vector: we have  
0 a −b
A = −a 0 c
b −c 0
and we can consider
 
0 ω3 −ω2
εijk ωk = −ω3 0 ω1 
ω2 −ω1 0

So if we have ω = (c, b, a), then Aij = εijk ωk .

This decomposition can be useful in certain physical applications. For
example, if the matrix represents the stress of a system, different parts of the
decomposition will correspond to different types of stresses.

3.4.4 Matrix inverse

Definition (Inverse of matrix). Consider an m × n matrix A and n × m matrices
B and C. If BA = I, then we say B is the left inverse of A. If AC = I, then
we say C is the right inverse of A. If A is square (n × n), then B = B(AC) =
(BA)C = C, i.e. the left and right inverses coincide. Both are denoted by A−1 ,
the inverse of A. Therefore we have

AA−1 = A−1 A = I.

Note that not all square matrices have inverses. For example, the zero matrix
clearly has no inverse.
Definition (Invertible matrix). If A has an inverse, then A is invertible.
Proposition. (AB)−1 = B −1 A−1

Proof. (B −1 A−1 )(AB) = B −1 (A−1 A)B = B −1 B = I.

Definition (Orthogonal and unitary matrices). A real n×n matrix is orthogonal
if AT A = AAT = I, i.e. AT = A−1 . A complex n × n matrix is unitary if
U † U = U U † = I, i.e. U † = U −1 .

Note that an orthogonal matrix A satisfies Aik (ATkj ) = δij , i.e. Aik Ajk = δij .
We can see this as saying “the scalar product of two distinct rows is 0, and the
scalar product of a row with itself is 1”. Alternatively, the rows (and columns —
by considering AT ) of an orthogonal matrix form an orthonormal set.
†
Similarly, for a unitary matrix, Uik Ukj = δij , i.e. uik u∗jk = u∗ik ujk = δij . i.e.
the rows are orthonormal, using the definition of complex scalar product.
Example.
(i) The reflection in a plane is an orthogonal matrix. Since Rij = δij − 2ni nj ,
We have

Rik Rjk = (δik − 2ni nk )(δjk − 2nj nk )

= δik δjk − 2δjk ni nk − 2δik nj nk + 2ni nk nj nk
= δij − 2ni nj − 2nj ni + 4ni nj (nk nk )
= δij

(ii) The rotation is an orthogonal matrix. We could multiply out using suffix
notation, but it would be cumbersome to do so. Alternatively, denote
rotation matrix by θ about n̂ as R(θ, n̂). Clearly, R(θ, n̂)−1 = R(−θ, n̂).
We have

Rij (−θ, n̂) = (cos θ)δij + ni nj (1 − cos θ) + εijk nk sin θ

= (cos θ)δji + nj ni (1 − cos θ) − εjik nk sin θ
= Rji (θ, n̂)

In other words, R(−θ, n̂) = R(θ, n̂)T . So R(θ, n̂)−1 = R(θ, n̂)T .
3.5 Determinants
Consider a linear map α : R3 → R3 . The standard basis e1 , e2 , e3 is mapped to
e01 , e02 , e03 with e0i = Aei . Thus the unit cube formed by e1 , e2 , e3 is mapped to
the parallelepiped with volume

[e01 , e02 , e03 ] = εijk (e01 )i (e02 )j (e03 )k

= εijk Ai` (e1 )` Ajm (e2 )m Akn (e3 )n
| {z } | {z } | {z }
δ1` δ2m δ3n

= εijk Ai1 Aj2 Ak3

We call this the determinant and write as

 
A11 A12 A13 

det(A) = A21 A22 A23 
A31 A32 A33 

3.5.1 Permutations
To define the determinant for square matrices of arbitrary size, we first have to
consider permutations.

Definition (Permutation). A permutation of a set S is a bijection ε : S → S.

Notation. Consider the set Sn of all permutations of 1, 2, 3, · · · , n. Sn contains
n! elements. Consider ρ ∈ Sn with i 7→ ρ(i). We write
 
1 2 ··· n
ρ= .
ρ(1) ρ(2) · · · ρ(n)

Definition (Fixed
  point). A fixed point of ρ is a k such that ρ(k) = k. e.g. in
1 2 3 4
, 3 is the fixed point. By convention, we can omit the fixed point
4 1 3 2 
1 2 4
and write as .
4 1 2

Definition (Disjoint permutation). Two permutations aredisjoint if numbers 

1 2 4 5 6
moved by one are fixed by the other, and vice versa. e.g. =
   5 6 1 4 2
2 6 1 4 5
, and the two cycles on the right hand side are disjoint.
6 2 5 1 4
Disjoint permutations commute, but in general non-disjoint permutations do
not.
 
2 6
Definition (Transposition and k-cycle). is a 2-cycle or a transposition,
6 2
 
1 4 5
and we can simply write (2 6). is a 3-cycle, and we can simply write
5 1 4
(1 5 4). (1 is mapped to 5; 5 is mapped to 4; 4 is mapped to 1)
Proposition. Any q-cycle can be written as a product of 2-cycles.

Proof. (1 2 3 · · · n) = (1 2)(2 3)(3 4) · · · (n − 1 n).

Definition (Sign of permutation). The sign of a permutation ε(ρ) is (−1)r ,
where r is the number of 2-cycles when ρ is written as a product of 2-cycles. If
ε(ρ) = +1, it is an even permutation. Otherwise, it is an odd permutation. Note
that ε(ρσ) = ε(ρ)ε(σ) and ε(ρ−1 ) = ε(ρ).

The proof that this is well-defined can be found in IA Groups.

Definition (Levi-Civita symbol). The Levi-Civita symbol is defined by

+1 if j1 j2 j3 · · · jn is an even permutation of 1, 2, · · · n

εj1 j2 ···jn = −1 if it is an odd permutation

0 if any 2 of them are equal


Clearly, ερ(1)ρ(2)···ρ(n) = ε(ρ).

Definition (Determinant). The determinant of an n × n matrix A is defined as:
X
det(A) = ε(σ)Aσ(1)1 Aσ(2)2 · · · Aσ(n)n ,
σ∈Sn

or equivalently,
det(A) = εj1 j2 ···jn Aj1 1 Aj2 2 · · · Ajn n .
Proposition.  
a b 

c = ad − bc
d

3.5.2 Properties of determinants

Proposition. det(A) = det(AT ).
Proof. Take a single term Aσ(1)1 Aσ(2)2 · · · Aσ(n)n and let ρ be another permuta-
tion in Sn . We have

Aσ(1)1 Aσ(2)2 · · · Aσ(n)n = Aσ(ρ(1))ρ(1) Aσ(ρ(2))ρ(2) · · · Aσ(ρ(n))ρ(n)

since the right hand side is just re-ordering the order of multiplication. Choose
ρ = σ −1 and note that ε(σ) = ε(ρ). Then
X
det(A) = ε(ρ)A1σ(1) A2σ(2) · · · Anσ(n) = det(AT ).
ρ∈Sn

Proposition. If matrix B is formed by multiplying every element in a single row

of A by a scalar λ, then det(B) = λ det(A). Consequently, det(λA) = λn det(A).
Proof. Each term in the sum is multiplied by λ, so the whole sum is multiplied
by λn .

Proposition. If 2 rows (or 2 columns) of A are identical, the determinant is 0.

Proof. wlog, suppose columns 1 and 2 are the same. Then
X
det(A) = ε(σ)Aσ(1)1 Aσ(2)2 · · · Aσ(n)n .
σ∈Sn

Now write an arbitrary σ in the form σ = ρ(1 2). Then ε(σ) = ε(ρ)ε((1 2)) =
−ε(ρ). So X
det(A) = −ε(ρ)Aρ(2)1 Aρ(1)2 Aρ(3)3 · · · Aρ(n)n .
ρ∈Sn

But columns 1 and 2 are identical, so Aρ(2)1 = Aρ(2)2 and Aρ(1)2 = Aρ(1)1 . So
det(A) = − det(A) and det(A) = 0.
Proposition. If 2 rows or 2 columns of a matrix are linearly dependent, then
the determinant is zero.

Proof. Suppose in A, (column r) + λ(column s) = 0. Define

(
Aij j=6 r
Bij = .
Aij + λAis j = r

Then det(B) = det(A) + λ det(matrix with column r = column s) = det(A).

Then we can see that the rth column of B is all zeroes. So each term in the sum
contains one zero and det(A) = det(B) = 0.
Even if we don’t have linearly dependent rows or columns, we can still run
the exact same proof as above, and still get that det(B) = det(A). Linear
dependence is only required to show that det(B) = 0. So in general, we can add
a linear multiple of a column (or row) onto another column (or row) without
changing the determinant.
Proposition. Given a matrix A, if B is a matrix obtained by adding a multiple
of a column (or row) of A to another column (or row) of A, then det A = det B.
Corollary. Swapping two rows or columns of a matrix negates the determinant.

Proof. We do the column case only. Let A = (a1 · · · ai · · · aj · · · an ). Then

det(a1 · · · ai · · · aj · · · an ) = det(a1 · · · ai + aj · · · aj · · · an )
= det(a1 · · · ai + aj · · · aj − (ai + aj ) · · · an )
= det(a1 · · · ai + aj · · · − ai · · · an )
= det(a1 · · · aj · · · − ai · · · an )
= − det(a1 · · · aj · · · ai · · · an )

Alternatively, we can prove this from the definition directly, using the fact that
the sign of a transposition is −1 (and that the sign is multiplicative).

Proposition. det(AB) = det(A) det(B).

P
Proof. First note that σ ε(σ)Aσ(1)ρ(1) Aσ(2)ρ(2) = ε(ρ) det(A), i.e. swapping
columns (or rows) an even/odd number of times gives a factor ±1 respectively.
We can prove this by writing σ = µρ.
Now
X
det AB = ε(σ)(AB)σ(1)1 (AB)σ(2)2 · · · (AB)σ(n)n
σ
X n
X
= ε(σ) Aσ(1)k1 Bk1 1 · · · Aσ(n)kn Bkn n
σ k1 ,k2 ,··· ,kn
X X
= Bk1 1 · · · Bkn n ε(σ)Aσ(1)k1 Aσ(2)k2 · · · Aσ(n)kn
k1 ,··· ,kn σ
| {z }
S

Now consider the many different S’s. If in S, two of k1 and kn are equal, then S
is a determinant of a matrix with two columns the same, i.e. S = 0. So we only
have to consider the sum over distinct ki s. Thus the ki s are are a permutation
of 1, · · · n, say ki = ρ(i). Then we can write
X X
det AB = Bρ(1)1 · · · Bρ(n)n ε(σ)Aσ(1)ρ(1) · · · Aσ(n)ρ(n)
ρ σ
X
= Bρ(1)1 · · · Bρ(n)n (ε(ρ) det A)
ρ
X
= det A ε(ρ)Bρ(1)1 · · · Bρ(n)n
ρ

= det A det B

Corollary. If A is orthogonal, det A = ±1.

Proof.

AAT = I
det AAT = det I
det A det AT = 1
(det A)2 = 1
det A = ±1

Corollary. If U is unitary, | det U | = 1.

Proof. We have det U † = (det U T )∗ = det(U )∗ . Since U U † = I, we have

det(U ) det(U )∗ = 1.
Proposition. In R3 , orthogonal matrices represent either a rotation (det = 1)
or a reflection (det = −1).

3.5.3 Minors and Cofactors

Definition (Minor and cofactor). For an n × n matrix A, define Aij to be the
(n − 1) × (n − 1) matrix in which row i and column j of A have been removed.
The minor of the ijth element of A is Mij = det Aij
The cofactor of the ijth element of A is ∆ij = (−1)i+j Mij .
Notation. We use ¯ to denote a symbol which has been missed out of a natural
sequence.

Example. 1, 2, 3, 5 = 1, 2, 3, 4̄, 5.
The significance of these definitions is that we can use them to provide a
systematic way of evaluating determinants. We will also use them to find inverses
of matrices.

Theorem (Laplace expansion formula). For any particular fixed i,

n
X
det A = Aji ∆ji .
j=1

Proof.
n
X n
X
det A = Aj i i εj1 j2 ···jn Aj1 1 Aj2 2 · · · Aji i · · · Ajn n
ji =1 j1 ,··· ,ji ,···jn

Let σ ∈ Sn be the permutation which moves ji to the ith position, and leave
everything else in its natural order, i.e.
 
1 · · · i i + 1 i + 2 · · · ji − 1 ji ji + 1 · · · n
σ=
1 · · · ji i i + 1 · · · ji − 2 ji − 1 ji + 1 · · · n

if ji > i, and similarly for other cases. To perform this permutation, |i − ji |

transpositions are made. So ε(σ) = (−1)i−ji .
Now consider the permutation ρ ∈ Sn

1 · · · · · · j¯i · · · n
 
ρ=
j1 · · · j¯i · · · · · · jn

The composition ρσ reorders (1, · · · , n) to (j1 , j2 , · · · , jn ). So ε(ρσ) = εj1 ···jn =

ε(ρ)ε(σ) = (−1)i−ji εj1 ···j̄i ···jn . Hence the original equation becomes
n
X X
det A = Aj i i (−1)i−ji εj1 ···j̄i ···jn Aj1 1 · · · Aji i · · · Ajn n
ji =1 j1 ···j̄i ···jn
n
X
= Aji i (−1)i−ji Mji i
ji =1
X n
= Aj i i ∆ j i i
ji =1
X n
= Aji ∆ji
j=1
 
2 4
2

Example. det A = 3 2
1. We can pick the first row and have
2 10
     
2 1
 − 4 3 1 + 2 3 2
   
det A = 2 
0 1  2 1   2 0
= 2(2 − 0) − 4(3 − 2) + 2(0 − 4)
= −8.

Alternatively, we can pick the second column and have

     
3 1 2 2  2 2
det A = −4   + 2
  − 0
 
2 1 2 1 3 1
= −4(3 − 2) + 2(2 − 4) − 0
= −8.

In practical terms, we use a combination of properties of determinants with

a sensible choice of i to evaluate det(A).
 
1 a a2 
 
Example. Consider 1 b b2 . Row 1 - row 2 gives
1 c c2 
   
0
 a−b a2 − b2  0 1
 a + b
2
1
 b b 
 = (a − b) 1 b b2  .
1 c c2  1 c c2 

Do row 2 - row 3. We obtain

 
0 1 a + b

(a − b)(b − c) 0 1 b + c  .
1 c c2 

Row 1 - row 2 gives

 
0 0 1 

(a − b)(b − c)(a − c) 0 1 b + c = (a − b)(b − c)(a − c).
1 c c2 
4 Matrices and linear equations
4.1 Simple example, 2 × 2
Consider the system of equations

A11 x1 + A12 x2 = d1 (a)

A21 x1 + A22 x2 = d2 . (b)

We can write this as

Ax = d.

If we do (a)×A22 −(b)×A12 and similarly the other way round, we obtain

(A11 A22 − A12 A21 )x1 = A22 d1 − A12 d2

(A11 A22 − A12 A21 ) x2 = A11 d2 − A21 d1
| {z }
det A

Dividing by det A and writing in matrix form, we have

    
x1 1 A22 −A12 d1
=
x2 det A −A21 A11 d2

On the other hand, given the equation Ax = d, if A−1 exists, then by multiplying
both sides on the left by A−1 , we obtain x = A−1 d.
Hence, we have constructed A−1 in the 2 × 2 case, and shown that the
condition for its existence is det A 6= 0, with
 
1 A22 −A12
A−1 =
det A −A21 A11
4.2 Inverse of an n × n matrix
For larger matrices, the formula for the inverse is similar, but slightly more
complicated (and costly to evaluate). The key to finding the inverse is the
following:
P
Lemma. Aik ∆jk = δij det A.
Proof. If i 6= j, then consider an n × n matrix B, which is identical to A except
the jth row is replaced by the ith row of A. So ∆jk of B = ∆jk of A, since ∆jk
does not depend on the elements in row j. Since B has a duplicate row, we know
that
Xn Xn
0 = det B = Bjk ∆jk = Aik ∆jk .
k=1 k=1

If i = j, then the expression is det A by the Laplace expansion formula.

Theorem. If det A 6= 0, then A−1 exists and is given by
∆ji
(A−1 )ij = .
det A
Proof.
∆ki δij det A
(A−1 )ik Akj = Akj = = δij .
det A det A
So A−1 A = I.
The other direction is easy to prove. If det A = 0, then it has no inverse,
since for any matrix B, det AB = 0, and hence AB cannot be the identity.
 
1 λ 0
Example. Consider the shear matrix Sλ = 0 1 0. We have det Sλ = 1.
0 0 1
The cofactors are
∆11 = 1 ∆12 = 0 ∆13 = 0
∆21 − λ ∆22 = 1 ∆23 = 0
∆31 = 0 ∆32 = 0 ∆33 = 1
 
1 −λ 0
So Sλ−1 = 0 1 0 .
0 0 1
How many arithmetic operations are involved in calculating the inverse of an
n × n matrix? We just count multiplication operations since they are the most
time-consuming. Suppose that calculating det A takes fn multiplications. This
involves n (n − 1) × (n − 1) determinants, and you need n more multiplications to
put them together. So fn = nfn−1 + n. So fn = O(n!) (in fact fn ≈ (1 + e)n!).
To find the inverse, we need to calculate n2 cofactors. Each is a n − 1
determinant, and each takes O((n−1)!). So the time complexity is O(n2 (n−1)!) =
O(n · n!).
This is incredibly slow. Hence while it is theoretically possible to solve
systems of linear equations by inverting a matrix, sane people do not do so
in general. Instead, we develop certain better methods to solve the equations.
In fact, the “usual” method people use to solve equations by hand only has
complexity O(n3 ), which is a much better complexity.
4.3 Homogeneous and inhomogeneous equations
Consider Ax = b where A is an n × n matrix, x and b are n × 1 column vectors.

Definition (Homogeneous equation). If b = 0, then the system is homogeneous.

Otherwise, it’s inhomogeneous.
Suppose det A 6= 0. Then there is a unique solution x = A−1 b (x = 0 for
homogeneous).
How can we understand this result? Recall that det A 6= 0 means that the
columns of A are linearly independent. The columns are the images of the stan-
dard basis, e0i = Aei . So det A 6= 0 means that e0i are linearly independent and
form a basis of Rn . Therefore the image is the whole of Rn . This automatically
ensures that b is in the image, i.e. there is a solution.
To show that there is exactly one solution, suppose x and x0 are both solutions.
Then Ax = Ax0 = b. So A(x − x0 ) = 0. So x − x0 is in the kernel of A. But
since the rank of A is n, by the rank-nullity theorem, the nullity is 0. So the
kernel is trivial. So x − x0 = 0, i.e. x = x0 .

4.3.1 Gaussian elimination

Consider a general solution

A11 x1 + A12 x2 + · · · + A1n xn = d1

A21 x1 + A22 x2 + · · · + A2n xn = d2
..
.
Am1 x1 + Am2 x2 + · · · + Amn xn = dm

So we have m equations and n unknowns.

Assume A11 6= 0 (if not, we can re-order the equations). We can use the
first equation to eliminate x1 from the remaining (m − 1) equations. Then use
the second equation to eliminate x2 from the remaining (m − 2) equations (if
anything goes wrong, just re-order until things work). Repeat.
We are left with

A11 x1 + A12 x2 + A13 x3 + · · · + A1n xn = d1

(2) (2) (2)
A22 x2 + A23 x3 + · · · + A2n xn = d2
..
.
A(r) (r)
rr xr + · · · + Arn xn = dr
(r)
0 = dr+1
..
.
0 = d(r)
m

(i)
Here Aii 6= 0 (which we can achieve by re-ordering), and the superfix (i) refers
(2)
to the “version number” of the coefficient, e.g. A22 is the second version of the
coefficient of x2 in the second row.
Let’s consider the different possibilities:
(r) (r)
(i) r < m and at least one of dr+1 , · · · dm =
6 0. Then a contradiction is
reached. The system is inconsistent and has no solution. We say it is
overdetermined.
Example. Consider the system

3x1 + 2x2 + x3 = 3
6x1 + 3x2 + 3x3 = 0
6x1 + 2x2 + 4x3 = 6

This becomes

3x1 + 2x2 + x3 = 3
0 − x2 + x3 = −6
0 − 2x2 + 2x3 = 0

And then

3x1 + 2x2 + x3 = 3
0 − x2 + x3 = −6
0 = 12
(3)
We have d3 = 12 = 0 and there is no solution.

(r)
(ii) If r = n ≤ m, and all dr+i = 0. Then from the nth equation, there
(n) (n)
is a unique solution for xn = dn /Ann , and hence for all xi by back
substitution. This system is determined.
Example.

2x1 + 5x2 = 2
4x1 + 3x2 = 11

This becomes

2x1 + 5x2 = 2
−7x2 = 7

So x2 = −1 and thus x1 = 7/2.

(r)
(iii) If r < n and dr+i = 0, then xr+1 , · · · xn can be freely chosen, and there
are infinitely many solutions. System is under-determined. e.g.

x1 + x2 = 1
2x1 + 2x2 = 2

Which gives

x1 + x2 = 1
0=0

So x1 = 1 − x2 is a solution for any x2 .

In the n = m case, there are O(n3 ) operations involved, which is much less than
inverting the matrix. So this is an efficient way of solving equations.
This is also be related to the determinant. Consider the case where m = n
and A is square. Since row operations do not change the determinant and
swapping rows give a factor of (−1). So
 
A11 A12 · · · · · · · · · A1n 
 
(2) (n)
 0
 A22 · · · · · · · · · A2n 
 . .. .. .. .. .. 
 . .
. . . . . 
det A = (−1)k 

(r) (n) 
 0
 0 · · · Arr · · · Arn 
 0 0 ··· 0 0 · · · 
 .. .. .. .. .. .. 

 . . . . . . 

This determinant is an upper triangular one (all elements below diagonal are 0)
and the determinant is the product of its diagonal elements.
(r)
Hence if r < n (and di = 0 for i > r), then we have case (ii) and the
(2) (n)
det A = 0. If r = n, then det A = (−1)k A11 A22 · · · Ann 6= 0.
4.4 Matrix rank
Consider a linear map α : Rn → Rm . Recall the rank r(α) is the dimension of
the image. Suppose that the matrix A is associated with the linear map. We
also call r(A) the rank of A.
Recall that if the standard basis is e1 , · · · en , then Ae1 , · · · , Aen span the
image (but not necessarily linearly independent).
Further, Ae1 , · · · , Aen are the columns of the matrix A. Hence r(A) is the
number of linearly independent columns.
Definition (Column and row rank of linear map). The column rank of a matrix
is the maximum number of linearly independent columns.
The row rank of a matrix is the maximum number of linearly independent
rows.

Theorem. The column rank and row rank are equal for any m × n matrix.
Proof. Let r be the row rank of A. Write the biggest set of linearly independent
rows as v1T , v2T , · · · vrT or in component form vkT = (vk1 , vk2 , · · · , vkn ) for k =
1, 2, · · · , r.
Now denote the ith row of A as rTi = (Ai1 , Ai2 , · · · Ain ).
Note that every row of A can be written as a linear combination of the v’s.
(If ri cannot be written as a linear combination of the v’s, then it is independent
of the v’s and v is not the maximum collection of linearly independent rows)
Write
r
X
rTi = Cik vkT .
k=1

For some coefficients Cik with 1 ≤ i ≤ m and 1 ≤ k ≤ r.

Now the elements of A are
r
X
Aij = (ri )Tj = Cik (vk )j ,
k=1

or    
A1j C1k
 A2j  Xr  C2k 
 ..  = vkj  . 
   
 .   .. 
k=1
Amj Cmk
So every column of A can be written as a linear combination of the r column
vectors ck . Then the column rank of A ≤ r, the row rank of A.
Apply the same argument to AT to see that the row rank is ≤ the column
rank.
4.5 Homogeneous problem Ax = 0
We restrict our attention to the square case, i.e. number of unknowns = number
of equations. Here A is an n × n matrix. We want to solve Ax = 0.
First of all, if det A 6= 0, then A−1 exists and x−1 = A−1 0 = 0, which is the
unique solution. Hence if Ax = 0 with x 6= 0, then det A = 0.

4.5.1 Geometrical interpretation

We consider a 3 × 3 matrix  T
r1
A = rT2 
rT3
Ax = 0 means that ri · x = 0 for all i. Each equation ri · x = 0 represents a
plane through the origin. So the solution is the intersection of the three planes.
There are three possibilities:
(i) If det A = [r1 , r2 , r3 ] 6= 0, span{r1 , r2 , r3 } = R3 and thus r(A) = 3. By
the rank-nullity theorem, n(A) = 0 and the kernel is {0}. So x = 0 is the
unique solution.
(ii) If det A = 0, then dim(span{r1 , r2 , r3 }) = 1 or 2.

(a) If rank = 2, wlog assume r1 , r2 are linearly independent. So x lies

on the intersection of two planes x · r1 = 0 and x · r2 = 0, which is
the line {x ∈ R3 : x = λr1 × r2 } (Since x lies on the intersection of
the two planes, it has to be normal to the normals of both planes).
All such points on this line also satisfy x · r3 = 0 since r3 is a linear
combination of r1 and r2 . The kernel is a line, n(A) = 1.
(b) If rank = 1, then r1 , r2 , r3 are parallel. So x·r1 = 0 ⇒ x·r2 = x·r3 = 0.
So all x that satisfy x · r1 = 0 are in the kernel, and the kernel now is
a plane. n(A) = 2.
(We also have the trivial case where r(A) = 0, we have the zero mapping and
the kernel is R3 )

4.5.2 Linear mapping view of Ax = 0

In the general case, consider a linear map α : Rn → Rn x 7→ x0 = Ax. The
kernel k(A) = {x ∈ Rn : Ax = 0} has dimension n(A).

(i) If n(A) = 0, then A(e1 ), A(e2 ), · · · , A(en ) is a linearly independent set,

and r(A) = n.
(ii) If n(A) > 0, then the image is not the whole of Rn . Let {ui }, i =
1, · · · , n(A) be a basis of the kernel, i.e. so given any solution to Ax = 0,
n(A)
X
x= λi ui for some λi . Extend {ui } to be a basis of Rn by introducing
i=1
extra vectors ui for i = n(A) + 1, · · · , n. The vectors A(ui ) for i =
n(A) + 1, · · · , n form a basis of the image.
4.6 General solution of Ax = d
Finally consider the general equation Ax = d, where A is an n × n matrix and
x, d are n × 1 column vectors. We can separate into two main cases.

(i) det(A) 6= 0. So A−1 exists and n(A) = 0, r(A) = n. Then for any d ∈ Rn ,
a unique solution must exists and it is x = A−1 d.
(ii) det(A) = 0. Then A−1 does not exist, and n(A) > 0, r(A) < n. So the
image of A is not the whole of Rn .
(a) If d 6∈ im A, then there is no solution (by definition of the image)
(b) If d ∈ im A, then by definition there exists at least one x such that
Ax = d. The general solution of Ax = d can be written as x = x0 +y,
where x0 is a particular solution (i.e. Ax0 = d), and y is any vector
in ker A (i.e. Ay = 0). (cf. Isomorphism theorem)
If n(A) = 0, then y = 0 only, and then the solution is unique (i.e.
case (i)). If n(A) > 0 , then {ui }, i = 1, · · · , n(A) is a basis of the
kernel. Hence
n(A)
X
y= µj uj ,
j=1
so
n(A)
X
x = x0 + µj uj
j=1

for any µj , i.e. there are infinitely many solutions.

Example.     
1 1 x1 1
=
a 1 x2 b
We have det A = 1 − a. If a 6= 1, then A−1 exists and
 
1 1 1 −1
A−1 = = .
1−a 1 − a −a 1

Then  
1 1−b
x= .
1−a −a + b
If a = 1, then    
x1 + x2 1
Ax = = (x1 + x2 ) .
x1 + x2 1
     
1 1 1
So im A = span and ker A = span 6 1, then
. If b = 6∈ im A
1 −1 b
 
1
and there is no solution. If b = 1, then ∈ im A.
 b
1
We find a particular solution of . So The general solution is
0
   
1 1
x= +λ .
0 −1

Example. Find the general solution of

    
a a b x 1
 b a a y  =  c 
a b a z 1

We have det A = (a − b)2 (2a + b). If a 6= b and b 6= −2a, then the inverse exists
and there is a unique solution for any c. Otherwise, the possible cases are
(i) a = b,b6= −2a.
  So a6=0. The kernel is the plane x + y + z = 0which
 is
 −1 −1  1
span  1  ,  0  We extend this basis to R3 by adding 0.
0 1 0
 
     
a 1 1
So the image is the span of a = 1. Hence if c 6= 1, then  c  is not
a 1 1
in 
theimage and there is no solution. If c = 1, then a particular solution
1
a
is  0  and the general solution is
0
1    
a −1 −1
x = 0 + λ  1  + µ  0 
0 0 1

(ii) If a 6= b and b = −2a, then a 6= 0. The kernel satisfies

x + y − 2z = 0
−2x + y + z = 0
x − 2y + z = 0
 
 1 
This can be solved to give x = y = z, and the kernel is span 1 . We
1
 
   
1 0
add 0 and 0 to form a basis of R3 . So the image is the span of
0 1
   
1 −2
−2 ,  1 .
1 1
 
1
If  c  is in the image, then
1
     
1 1 −2
 c  = λ −2 + µ  1  .
1 1 1

Then the only solution is µ = 0, λ = 1, c = −2. Thus 

there
 is no solution if
1
a
c 6= −2, and when c = −2, pick a particular solution  0  and the general
0
solution is 1  
a 1
x =  0  + λ 1
0 1

(iii) If a = b and b = −2a, then a = b = 0 and ker A = R3 . So there is no

solution for any c.
5 Eigenvalues and eigenvectors
Given a matrix A, an eigenvector is a vector x that satisfies Ax = λx for some
λ. We call λ the associated eigenvalue. In some sense, these vectors are not
modified by the matrix, and are just scaled up by the matrix. We will look
at the properties of eigenvectors and eigenvalues, and see their importance in
diagonalizing matrices.
5.1 Preliminaries and definitions
Theorem (Fundamental theorem of algebra). Let p(z) be a polynomial of degree
m ≥ 1, i.e.
Xm
p(z) = cj z j ,
j=0

where cj ∈ C and cm 6= 0.
Then p(z) = 0 has precisely m (not necessarily distinct) roots in the complex
plane, accounting for multiplicity.
Note that we have the disclaimer “accounting for multiplicity”. For example,
x2 − 2x + 1 = 0 has only one distinct root, 1, but we say that this root has
multiplicity 2, and is thus counted twice. Formally, multiplicity is defined as
follows:
Definition (Multiplicity of root). The root z = ω has multiplicity k if (z − ω)k
is a factor of p(z) but (z − ω)k+1 is not.

Example. Let p(z) = z 3 − z 2 − z + 1 = (z − 1)2 (z + 1). So p(z) = 0 has roots

1, 1, −1, where z = 1 has multiplicity 2.
Definition (Eigenvector and eigenvalue). Let α : Cn → Cn be a linear map
with associated matrix A. Then x 6= 0 is an eigenvector of A if

Ax = λx

for some λ. λ is the associated eigenvalue. This means that the direction of the
eigenvector is preserved by the mapping, but is scaled up by λ.
There is a rather easy way of finding eigenvalues:

Theorem. λ is an eigenvalue of A iff

det(A − λI) = 0.

Proof. (⇒) Suppose that λ is an eigenvalue and x is the associated eigenvector.

We can rearrange the equation in the definition above to

(A − λI)x = 0

and thus
x ∈ ker(A − λI)
But x 6= 0. So ker(A − λI) is non-trivial and det(A − λI) = 0. The (⇐) direction
is similar.
Definition (Characteristic equation of matrix). The characteristic equation of
A is
det(A − λI) = 0.
Definition (Characteristic polynomial of matrix). The characteristic polynomial
of A is
pA (λ) = det(A − λI).
From the definition of the determinant,

pA (λ) = det(A − λI)

= εj1 j2 ···jn (Aj1 1 − λδj1 1 ) · · · (Ajn n − λδjn n )
= c0 + c1 λ + · · · + cn λn

for some constants c0 , · · · , cn . From this, we see that

(i) pA (λ) has degree n and has n roots. So an n × n matrix has n eigenvalues
(accounting for multiplicity).

(ii) If A is real, then all ci ∈ R. So eigenvalues are either real or come in

complex conjugate pairs.
(iii) cn = (−1)n and cn−1 = (−1)n−1 (A11 + A22 + · · · + Ann ) = (−1)n−1 tr(A).
But cn−1 is the sum of roots, i.e. cn−1 = (−1)n−1 (λ1 + λ2 + · · · λn ), so

tr(A) = λ1 + λ2 + · · · + λn .

Finally, c0 = pA (0) = det(A). Also c0 is the product of all roots, i.e.

c0 = λ1 λ2 · · · λn . So
det A = λ1 λ2 · · · λn .

The kernel of the matrix A − λI is the set {x : Ax = λx}. This is a vector

subspace because the kernel of any map is always a subspace.

Definition (Eigenspace). The eigenspace denoted by Eλ is the kernel of the

matrix A − λI, i.e. the set of eigenvectors with eigenvalue λ.
Definition (Algebraic multiplicity of eigenvalue). The algebraic multiplicity
M (λ) or Mλ of an eigenvalue λ is the multiplicity of λ in pA (λ) = 0. By the
fundamental theorem of algebra,
X
M (λ) = n.
λ

If M (λ) > 1, then the eigenvalue is degenerate.

Definition (Geometric multiplicity of eigenvalue). The geometric multiplicity

m(λ) or mλ of an eigenvalue λ is the dimension of the eigenspace, i.e. the
maximum number of linearly independent eigenvectors with eigenvalue λ.
Definition (Defect of eigenvalue). The defect ∆λ of eigenvalue λ is

∆λ = M (λ) − m(λ).

It can be proven that ∆λ ≥ 0, i.e. the geometric multiplicity is never greater

than the algebraic multiplicity.
5.2 Linearly independent eigenvectors
Theorem. Suppose n×n matrix A has distinct eigenvalues λ1 , λ2 , · · · , λn . Then
the corresponding eigenvectors x1 , x2 , · · · , xn are linearly independent.

Proof. Proof by contradiction: Suppose x1 , x2 , · · · , xn are linearly dependent.

Then we can find non-zero constants di for i = 1, 2, · · · , r, such that

d1 x1 + d2 x2 + · · · + dr xr = 0.

Suppose that this is the shortest non-trivial linear combination that gives 0 (we
may need to re-order xi ).
Now apply (A − λ1 I) to the whole equation to obtain

d1 (λ1 − λ1 )x1 + d2 (λ2 − λ1 )x2 + · · · + dr (λr − λ1 )xr = 0.

We know that the first term is 0, while the others are not (since we assumed
λi 6= λj for i 6= j). So

d2 (λ2 − λ1 )x2 + · · · + dr (λr − λ1 )xr = 0,

and we have found a shorter linear combination that gives 0. Contradiction.

Example.
 
0 1
(i) A = . Then pA (λ) = λ2 + 1 = 0. So λ1 = i and λ1 = −i.
−1 0
To solve (A − λ1 I)x = 0, we obtain
  
−i 1 x1
= 0.
−1 −i x2

So we obtain    
x1 1
=
x2 i
 
1
to be an eigenvector. Clearly any scalar multiple of is also a solution,
  i
1
but still in the same eigenspace Ei = span
i
Solving (A − λ2 I)x = 0 gives
   
x1 1
= .
x2 −i
 
1
So E−i = span .
−i
Note that M (±i) = m(±i) = 1, so ∆±i = 0. Also note that the two
eigenvectors are linearly independent and form a basis of C2 .
(ii) Consider  
−2 2 −3
A= 2 1 −6
−1 −2 0
Then det(A − λI) = 0 gives 45 + 21λ − λ2 − λ3 . So λ1 = 5, λ2 = λ3 = −3.
The eigenvector with eigenvalue 5 is
 
1
x= 2 
−1

We can find that the eigenvectors with eigenvalue −3 are

 
−2x2 + 3x3
x= x2 
x3

for any
  x2 ,
x3 . This gives two linearly independent eigenvectors, say
−2 3
 1  , 0.
0 1
So M (5) = m(5) = 1 and M (−3) = m(−3) = 2, and there is no defect for
both of them. Note that these three eigenvectors form a basis of C3 .
(iii) Let  
−3 −1 1
A = −1 −3 1
−2 −2 0
Then 0 = pA (λ) = −(λ + 2)4 . So λ = −2, −2, −2. To find the eigenvectors,
we have   
−1 −1 1 x1
(A + 2I)x = −1 −1 1 x2  = 0
−2 −2 2 x3
The general  is thus x1 + x2 − x3 = 0, and 
 solution the
 general
   solution
 is
x1  1 0 
thus x =  x2 . The eigenspace E−2 = span 0 , 1 .
x1 + x2 1 1
 

Hence M (−2) = 3 and m(−2) = 2. Thus the defect ∆−2 = 1. So the

eigenvectors do not form a basis of C3 .
(iv) Consider the reflection R in the plane with normal n. Clearly Rn = −n.
The eigenvalue is −1 and the eigenvector is n. Then E1 = span{n}. So
M (−1) = m(−1) = 1.
If p is any vector in the plane, Rp = p. So this has an eigenvalue of 1 and
eigenvectors being any vector in the plane. So M (1) = m(1) = 2.
So the eigenvectors form a basis of R3 .
(v) Consider a rotation R by θ about n. Since Rn = n, we have an eigenvalue
of 1 and eigenspace E1 = span{n}.
We know that there are no other real eigenvalues since rotation changes
the direction of any other vector. The other eigenvalues turn out to be
e±iθ . If θ 6= 0, there are 3 distinct eigenvalues and the eigenvectors form a
basis of C3 .

(vi) Consider a shear  

1 µ
A=
0 1
The characteristic equation is(1− λ)2 = 0 and λ = 1. The eigenvectors
1
corresponding to λ = 1 is x = . We have M (1) = 2 and m(1) = 1. So
0
∆1 = 1.
If n × n matrix A has n distinct eigenvalues, and hence has n linearly
independent eigenvectors v1 , v2 , · · · vn , then with respect to this eigenvector
basis, A is diagonal.
In this basis, v1 = (1, 0, · · · , 0) etc. We know that Avi = λi vi (no summation).
So the image of the ith basis vector is λi times the ith basis. Since the columns
of A are simply the images of the basis,
 
λ1 0 · · · 0
 0 λ2 · · · 0 
 
 .. .. . . .. 
. . . . 
0 0 ··· λn

The fact that A can be diagonalized by changing the basis is an important

observation. We will now look at how we can change bases and see how we can
make use of this.
5.3 Transformation matrices
How do the components of a vector or a matrix change when we change the
basis?
Let {e1 , e2 , · · · , en } and {ẽ1 , ẽ2 , · · · , ẽn } be 2 different bases of Rn or Cn .
Then we can write
X n
e˜j = Pij ei
i=1

i.e. Pij is the ith component of e˜j with respect to the basis {e1 , e2 , · · · , en }.
Note that the sum is made as Pij ei , not Pij ej . This is different from the formula
for matrix multiplication.
Matrix P has as its columns the vectors e˜j relative to {e1 , e2 , · · · , en }. So
P = (e˜1 e˜2 · · · e˜n ) and
P (ei ) = e˜i
Similarly, we can write
n
X
ei = Qki e˜k
k=1

with Q = (e1 e2 · · · en ).
Substituting this into the equation for e˜j , we have
n n
!
X X
e˜j = Qki e˜k Pij
i=1 k=1
n n
!
X X
= e˜k Qki Pij
k=1 i=1

But ẽ1 , ẽ2 , · · · , ẽn are linearly independent, so this is only possible if
n
X
Qki Pij = δkj ,
i=1

which is just a fancy way of saying QP = I, or Q = P −1 .

5.3.1 Transformation law for vectors

Pn
Pn respect to basis {ei }, u = i=1 ui ei . With respect to basis {e˜i }, u =
With
i=1 ũi e˜i . Note that this is the same vector u but has different components
with respect to different bases. Using the transformation matrix above for the
basis, we have
n
X n
X
u= u˜j Pij ei
j=1 i=1
 
n
X Xn
=  Pij u˜j  ei
i=1 j=1

By comparison, we know that

n
X
ui = Pij u˜j
j=1

Theorem. Denote vector as u with respect to {ei } and ũ with respect to {e˜i }.
Then
u = P ũ and ũ = P −1 u

Example. Take the first basis as {e1 = (1, 0), e2 = (0, 1)} and the second as
{e˜1 = (1, 1), e˜2 = (−1, 1)}.
So e˜1 = e1 + e2 and e˜2 = −e1 + e2 . We have
 
1 −1
P= .
1 1

Then for an arbitrary vector u, we have

u = u1 e1 + u2 e2
1 1
= u1 (e˜1 − e˜2 ) + u2 (e˜1 + e˜2 )
2 2
1 1
= (u1 + u2 )e˜1 + (−u1 + u2 )e˜2 .
2 2
Alternatively, using the formula above, we obtain

ũ = P −1 u
  
1 1 1 u1
=
2 −1 1 u2
 1 
2 (u1 + u2 )
= 1
2 (−u1 + u2 )

Which agrees with the above direct expansion.

5.3.2 Transformation law for matrix

Consider a linear map α : Cn → Cn with associated n × n matrix A. We have

u0 = α(u) = Au.

Denote u and u0 as being with respect to basis {ei } (i.e. same basis in both
spaces), and ũ, ũ0 with respect to {e˜i }.
Using what we’ve got above, we have

u0 = Au
P ũ0 = AP ũ
ũ0 = P −1 AP ũ
= Ãũ

So
Theorem.
à = P −1 AP.
 
1 λ 0
Example. Consider the shear Sλ = 0 1 0 with respect to the standard
0 0 1
basis. Choose a new set of basis vectors by rotating by θ about the e3 axis:

e˜1 = cos θe1 + sin θe2

e˜2 = − sin θe1 + cos θe2
e˜3 = e3

So we have
   
cos θ − sin θ 0 cos θ sin θ 0
P =  sin θ cos θ 0 , P −1 = − sin θ cos θ 0
0 0 1 0 0 1

Now use the basis transformation laws to obtain

λ cos2 θ
 
1 + λ sin θ cos θ 0
2
S̃λ =  −λ sin θ 1 − λ sin θ cos θ 0
0 0 1

Clearly this is much more complicated than our original basis. This shows that
choosing a sensible basis is important.
More generally, given α : Cm → Cn , given x ∈ Cm , x0 ∈ Cn with x0 = Ax.
We know that A is an n × m matrix.
Suppose Cm has a basis {ei } and Cn has a basis {fi }. Now change bases to
{e˜i } and {f˜i }.
We know that x = P x̃ with P being an m × m matrix, with x0 = Rx̃0 with
R being an n × n matrix.
Combining both of these, we have

Rx̃0 = AP x̃
x̃0 = R−1 AP x̃

Therefore à = R−1 AP .
Example. Consider α : R3 → R2 , with respect to the standard bases in both
spaces,  
2 3 4
A=
1 6 3
   
2 1
Use a new basis , in R2 and keep the standard basis in R3 . The basis
1 5
change matrix in R3 is simply I, while
   
2 1 1 5 −1
R= , R−1 =
1 5 9 −1 2

is the transformation matrix for R2 . So

  
2 1 2 3 4
à = I
1 5 1 6 3
  
1 5 −1 2 3 4
=
9 −1 2 1 6 3
 
1 1 17/9
=
0 1 2/9
   
2 ˜ 1
We can alternatively do it this way: we know that f˜1 = , f2 = Then
1 5
we know that

e˜1 = e1 7→ 2f1 + f2 = f1
e˜2 = e2 7→ 3f1 + 6f2 = f˜1 + f˜2
17 ˜ 2 ˜
e˜3 = e3 7→ 4f1 + 3f2 = f1 + f2
9 9
and we can construct the matrix correspondingly.
5.4 Similar matrices
Definition (Similar matrices). Two n × n matrices A and B are similar if there
exists an invertible matrix P such that

B = P −1 AP,

i.e. they represent the same map under different bases. Alternatively, using the
language from IA Groups, we say that they are in the same conjugacy class.
Proposition. Similar matrices have the following properties:

(i) Similar matrices have the same determinant.

(ii) Similar matrices have the same trace.
(iii) Similar matrices have the same characteristic polynomial.

Note that (iii) implies (i) and (ii) since the determinant and trace are the
coefficients of the characteristic polynomial
Proof. They are proven as follows:
(i) det B = det(P −1 AP ) = (det A)(det P )−1 (det P ) = det A

(ii)

tr B = Bii
= Pij−1 Ajk Pki
= Ajk Pki Pij−1
= Ajk (P P −1 )kj
= Ajk δkj
= Ajj
= tr A

(iii)

pB (λ) = det(B − λI)

= det(P −1 AP − λI)
= det(P −1 AP − λP −1 IP )
= det(P −1 (A − λI)P )
= det(A − λI)
= pA (λ)
5.5 Diagonalizable matrices
Definition (Diagonalizable matrices). An n × n matrix A is diagonalizable if
it is similar to a diagonal matrix. We showed above that this is equivalent to
saying the eigenvectors form a basis of Cn .
The requirement that matrix A has n distinct eigenvalues is a sufficient
condition for diagonalizability as shown above. However, it is not necessary.
Consider the second example in Section 5.2,
 
−2 2 −3
A= 2 1 −6
−1 −2 0

We found three linear eigenvectors

     
1 −2 3
e˜1 = 2 , e˜2 =  1  , e˜3 = 0
1 0 1

If we let    
1 −2 3 1 2 −3
1
P = 2 1 0 , P −1 = −2 4 6 ,
8
1 0 1 1 2 5
then  
5 0 0
à = P −1 AP = 0 −3 0 ,
0 0 −3
so A is diagonalizable.
Theorem. Let λ1 , λ2 , · · · , λr , with r ≤ n be the distinct eigenvalues of A. Let
B1 , B2 , · · · Br be the bases of the eigenspaces Eλ1 , Eλ2 , · · · , Eλr correspondingly.
[r
Then the set B = Bi is linearly independent.
i=1

This is similar to the proof we had for the case where the eigenvalues are
distinct. However, we are going to do it much concisely, and the actual meat of
the proof is actually just a single line.
(1) (1) (1)
Proof. Write B1 = {x1 , x2 , · · · xm(λ1 ) }. Then m(λ1 ) = dim(Eλ1 ), and simi-
larly for all Bi .
Consider the following general linear combination of all elements in B. Con-
sider the equation
r m(λ
Xi ) (i)
X
αij xj = 0.
i=1 j=1

The first sum is summing over all eigenspaces, and the second sum sums over
the basis vectors in Bi . Now apply the matrix
Y
(A − λk I)
k=1,2,··· ,K̄,··· ,r

to the above sum, for some arbitrary K. We obtain

 
m(λK )
(K)
X Y
αKj  (λK − λk ) xj = 0.
j=1 k=1,2,··· ,K̄,··· ,r

(K)
Since the xj are linearly independent (BK is a basis), αKj = 0 for all j. Since
K was arbitrary, all αij must be zero. So B is linearly independent.
Proposition. A is diagonalizable iff all its eigenvalues have zero defect.
5.6 Canonical (Jordan normal) form
Given a matrix A, if its eigenvalues all have non-zero defect, then we can find
a basis in which it is diagonal. However, if some eigenvalue does have defect,
we can still put it into an almost-diagonal form. This is known as the Jordan
normal form.

Theorem. Any 2 × 2 complex matrix A is similar to exactly one of

     
λ1 0 λ 0 λ 1
, ,
0 λ2 0 λ 0 λ

Proof. For each case:

(i) If A has two distinct eigenvalues, then eigenvectors are linearly independent.
Then we can use P formed from eigenvectors as its columns

(ii) If λ1 = λ2 = λ and dim Eλ = 2, then write Eλ = span{u, v}, with

u, v linearly independent.
  Now use {u, v} as a new basis of C2 and
λ 0
à = P −1 AP = = λI
0 λ
Note that since P −1 AP = λI, we have A = P (λI)P −1 = λI. So A is
isotropic, i.e. the same with respect to any basis.

(iii) If λ1 = λ2 = λ and dim(Eλ ) = 1, then Eλ = span{v}. Now choose basis

of C2 as {v, w}, where w ∈ C2 \ Eλ .
We know that Aw ∈ C2 . So Aw
 = αv
 + βw. Hence, if we change basis to
−1 λ α
{v, w}, then à = P AP = .
0 β
However, A and à both have eigenvalue λ with algebraic multiplicity 2.
So we must have β = λ. To make α = 1, let u = (Ã − λI)w. We know
u 6= 0 since w is not in the eigenspace. Then
  
2 0 α 0 α
(Ã − λI)u = (Ã − λI) w = w = 0.
0 0 0 0

So u is an eigenvector of à with eigenvalue λ.

We have u = Ãw − λw. So Ãw = u + λw.
 
λ 1
Change basis to {u, w}. Then A with respect to this basis is .
0 λ
This is a two-stage process: P sends basis to {v, w} and then matrix Q
sends to basis {u, w}. So the similarity transformation is Q−1 (P −1 AP )Q =
(P Q)−1 A(P Q).
Proposition. (Without proof) The canonical form, or Jordan normal form,
exists for any n × n matrix A. Specifically, there exists a similarity transform
such that A is similar to a matrix to à that satisfies the following properties:
(i) Ãαα = λα , i.e. the diagonal composes of the eigenvalues.
(ii) Ãα,α+1 = 0 or 1.

(iii) Ãij = 0 otherwise.

The actual theorem is actually stronger than this, and the Jordan normal
form satisfies some additional properties in addition to the above. However, we
shall not go into details, and this is left for the IB Linear Algebra course.
Example. Let  
−3 −1 1
A = −1 −3 1
−2 −2 0
   
−1 1
The eigenvalues are −2, −2, −2 and the eigenvectors are  1  , 0. Pick
0 1
      
1 −1 −1 1 1 −1
w = 0. Write u = (A − λI)w = −1 −1 1 0 = −1. Note that
0 −2 −2 2 0 −2
Au = −2u. We also have Aw = u − 2w. Form a basis   {u, w, v}, where v is
1
another eigenvector linearly independent from u, say 0.
1
 
−1 1 1
Now change to this basis with P = −1 0 0. Then the Jordan normal
−2 0 1
 
−2 1 0
form is P −1 AP =  0 −2 0 
0 0 −2
5.7 Cayley-Hamilton Theorem
Theorem (Cayley-Hamilton theorem). Every n × n complex matrix satisfies
its own characteristic equation.
Proof. We will only prove for diagonalizable matrices here. So suppose for our
matrix A, there is some P such that D = diag(λ1 , λ2 , · · · , λn ) = P −1 AP . Note
that
Di = (P −1 AP )(P −1 AP ) · · · (P −1 AP ) = P −1 Ai P.
Hence
pD (D) = pD (P −1 AP ) = P −1 [pD (A)]P.
Since similar matrices have the same characteristic polynomial. So

pA (D) = P −1 [pA (A)]P.

However, we also know that Di = diag(λi1 , λi2 , · · · λin ). So

pA (D) = diag(pA (λ1 ), pA (λ2 ), · · · , pA (λn )) = diag(0, 0, · · · , 0)

since the eigenvalues are roots of pA (λ) = 0. So 0 = pA (D) = P −1 pA (A)P and

thus pA (A) = 0.

There are a few things to note.

(i) If A−1 exists, then A−1 pA (A) = A−1 (c0 + c1 A + c2 A2 + · · · + cn An ) = 0.
So c0 A−1 + c1 + c2 A + · · · + cn An−1 . Since A−1 exists, c0 = ± det A 6= 0.
So
−1
A−1 = (c1 + c2 A + · · · + cn An−1 ).
c0
So we can calculate A−1 from positive powers of A.
(ii) We can define matrix exponentiation by
1 2 1
eA = I + A + A + · · · + An + · · · .
2! n!
It is a fact that this always converges.
If A is diagonalizable with P with D = P −1 AP = diag(λ1 , λ2 , · · · , λn ),
then
1 −1 2
P −1 eA P = P −1 IP + P −1 AP + P A P + ···
2!
1 2
=I +D+ D + ···
2!
= diag(eλ1 , eλ2 , · · · eλn )

So
eA = P [diag(eλ1 , eλ2 , · · · , eλn )]P −1 .
 
λ 1
(iii) For 2 × 2 matrices which are similar to B = We see that the
0 λ
characteristic polynomial pB (z) = det(B − zI) = (λ − z)2 . Then pB (B) =
 2  
2 0 −1 0 0
(λI − B) = = .
0 0 0 0
Since we have proved for the diagonalizable matrices above, we now know
that any 2 × 2 matrix satisfies Cayley-Hamilton theorem.
In IB Linear Algebra, we will prove the Cayley Hamilton theorem properly for
all matrices without assuming diagonalizability.
5.8 Eigenvalues and eigenvectors of a Hermitian matrix
5.8.1 Eigenvalues and eigenvectors
Theorem. The eigenvalues of a Hermitian matrix H are real.
Proof. Suppose that H has eigenvalue λ with eigenvector v 6= 0. Then

Hv = λv.

We pre-multiply by v† , a 1 × n row vector, to obtain

v† Hv = λv† v (∗)

We take the Hermitian conjugate of both sides. The left hand side is

(v† Hv)† = v† H † v = v† Hv

since H is Hermitian. The right hand side is

(λv† v)† = λ∗ v† v

So we have
v† Hv = λ∗ v† v.
From (∗), we know that λv† v = λ∗ v† v. Since v 6= 0, we know that v† v =
v · v 6= 0. So λ = λ∗ and λ is real.

Theorem. The eigenvectors of a Hermitian matrix H corresponding to distinct

eigenvalues are orthogonal.
Proof. Let

Hvi = λi vi (i)
Hvj = λj vj . (ii)

Pre-multiply (i) by vj† to obtain

vj† Hvi = λi vj† vi . (iii)

Pre-multiply (ii) by vi† and take the Hermitian conjugate to obtain

vj† Hvi = λj vj† vi . (iv)

Equating (iii) and (iv) yields

λi vj† vi = λj vj† vi .

Since λi 6= λj , we must have vj† vi = 0. So their inner product is zero and are
orthogonal.
So we know that if a Hermitian matrix has n distinct eigenvalues, then
the eigenvectors form an orthonormal basis. However, if there are degenerate
eigenvalues, it is more difficult, and requires the Gram-Schmidt process.

5.8.2 Gram-Schmidt orthogonalization (non-examinable)

Suppose we have a set B = {w1 , w2 , · · · , wr } of linearly independent vectors.
We want to find an orthogonal set B̃ = {v1 , v2 , · · · , vr }.
Define the projection of w onto v by Pv (w) = hv|wi hv|vi v. Now construct B̃
iteratively:
(i) v1 = w1
(ii) v2 = w2 − Pv1 (w)
 
hv1 |w2 i
Then we get that hv1 | v2 i = hv1 | w2 i − hv1 |v1 i hv1 | v1 i = 0

(iii) v3 = w3 − Pv1 (w3 ) − Pv2 (w3 )

.
(iv) ..
r−1
X
(v) vr = wr − Pvj (wr )
j=1

At each step, we subtract out the components of vi that belong to the space
of {v1 , · · · , vk−1 }. This ensures that all the vectors are orthogonal. Finally, we
normalize each basis vector individually to obtain an orthonormal basis.

5.8.3 Unitary transformation

Suppose U is the transformation between one orthonormal basis and a new
orthonormal basis {u1 , u2 , · · · , un }, i.e. hui | uj i = δij . Then
 
(u1 )1 (u2 )1 · · · (un )1
 (u1 )2 (u2 )2 · · · (un )2 
U = .
 
.. .. .. 
 .. . . . 
(u1 )n (u2 )n ··· (un )n

Then

(U † U )ij = (U † )ik Ukj

∗
= Uki Ukj
= (ui )∗k (uj )k
= hui | uj i
= δij

So U is a unitary matrix.

5.8.4 Diagonalization of n × n Hermitian matrices

Theorem. An n × n Hermitian matrix has precisely n orthogonal eigenvectors.
Proof. (Non-examinable) Let λ1 , λ2 , · · · , λr be the distinct eigenvalues of H (r ≤
n), with a set of corresponding orthonormal eigenvectors B = {v1 , v2 , · · · , vr }.
Extend to a basis of the whole of Cn

B 0 = {v1 , v2 , · · · , vr , w1 , w2 , · · · , wn−r }

Now use Gram-Schmidt to create an orthonormal basis

B̃ = {v1 , v2 , · · · , vr , u1 , u2 , · · · , un−r }.

Now write  
↑ ↑ ↑ ↑ ↑
P = v1 v2 ··· vr u1 ··· un−r 
↓ ↓ ↓ ↓ ↓
We have shown above that this is a unitary matrix, i.e. P −1 = P † . So if we
change basis, we have

P −1 HP = P † HP
 
λ1 0 ··· 0 0 0 ··· 0
 0 λ2 ··· 0 0 0 ··· 0 
 .. .. .. .. ..
 
.. .. 
.
 . . . . . . 0 

0 0 ··· λr 0 0 ··· 0 
= 
0
 0 ··· 0 c11 c12 ··· c1,n−r 

0
 0 ··· 0 c21 c22 ··· c2,n−r 

. .. .. .. .. .. .. ..
.

. . . . . . . . 
0 0 ··· 0 cn−r,1 cn−r,2 ··· cn−r,n−r

Here C is an (n − r) × (n − r) Hermitian matrix. The eigenvalues of C are also

eigenvalues of H because det(H − λI) = det(P † HP − λI) = (λ1 − λ) · · · (λr −
λ) det(C − λI). So the eigenvalues of C are the eigenvalues of H.
We can keep repeating the process on C until we finish all rows. For example,
if the eigenvalues of C are all distinct, there are n − r orthonormal eigenvectors
wj (for j = r + 1, · · · , n) of C. Let
 
1

 1 


 . .. 

Q=
 
 1 


 ↑ ↑ ↑ 

 wr+1 wr+2 · · · wn 
↓ ↓ ↓

with other entries 0. (where we have a r × r identity matrix block on the top
left corner and a (n − r) × (n − r) with columns formed by wj )
Since the columns of Q are orthonormal, Q is unitary. So Q† P † HP Q =
diag(λ1 , λ2 , · · · , λr , λr+1 , · · · , λn ), where the first r λs are distinct and the re-
maining ones are copies of previous ones.
The n linearly-independent eigenvectors are the columns of P Q.

So it now follows that H is diagonalizable via transformation U (= P Q). U

is a unitary matrix because P and Q are. We have

D = U † HU
H = U DU †

Note that a real symmetric matrix S is a special case of Hermitian matrices. So

we have

D = QT SQ
S = QDQT

Example. Find the orthogonal

  matrix which diagonalizes the following real
1 β
symmetric matrix: S = with β 6= 0 ∈ R.
β 1
We find the eigenvalues by solving the characteristic equation: det(S−λI) = 0,
and obtain λ = 1 ± β.
The
 corresponding
 eigenvectors satisfy (S − λI)x = 0, which gives x =
1 1
√
2 ±1    
1 1 1 1
We change the basis from the standard basis to √ ,√ (which
2 1 2 −1
is just a rotation by π/4).  √ √ 
1/√2 1/ √2
The transformation matrix is Q = . Then we know that
1/ 2 −1/ 2
S = QDQT with D = diag(1, −1)

5.8.5 Normal matrices

We have seen that the eigenvalues and eigenvectors of Hermitian matrices satisfy
some nice properties. More generally, we can define the following:

Definition (Normal matrix). A normal matrix as a matrix that commutes with

its own Hermitian conjugate, i.e.

N N † = N †N

Hermitian, real symmetric, skew-Hermitian, real anti-symmetric, orthogonal,

unitary matrices are all special cases of normal matrices.
It can be shown that:
Proposition.
(i) If λ is an eigenvalue of N , then λ∗ is an eigenvalue of N † .

(ii) The eigenvectors of distinct eigenvalues are orthogonal.

(iii) A normal matrix can always be diagonalized with an orthonormal basis of
eigenvectors.
6 Quadratic forms and conics
We want to study quantities like x21 + x22 and 3x21 + 2x1 x2 + 4x22 . For example,
conic sections generally take this form. The common characteristic of these is
that each term has degree 2. Consequently, we can write it in the form x† Ax
for some matrix A.

Definition (Sesquilinear, Hermitian and quadratic forms). A sesquilinear form

is a quantity F = x† Ax = x∗i Aij xj . If A is Hermitian, then F is a Hermitian
form. If A is real symmetric, then F is a quadratic form.
Theorem. Hermitian forms are real.

Proof. (x† Hx)∗ = (x† Hx)† = x† H † x = x† Hx. So (x† Hx)∗ = x† Hx and it is

real.
We know that any Hermitian matrix can be diagonalized with a unitary
transformation. So F (x) = x† Hx = x† U DU † x. Write x0 = U † x. So F =
(x0 )† Dx0 , where D = diag(λ1 , · · · , λn ).
We know that x0 is the vector x relative to the eigenvector basis. So
n
X
F (x) = λi |x0i |2
i=1

The eigenvectors are known as the principal axes.

 
x
Example. Take F = 2x2 − 4xy + 5y 2 = xT Sx, where x = and S =
y
 
2 −2
.
−2 5
Note that we can always choose the matrix to be symmetric. This is since
for any antisymmetric A, we have x† Ax = 0. So we can just take the symmetric
part.    
1 2 1 1
The eigenvalues are 1, 6 with corresponding eigenvectors √ ,√ .
5 1 5 −2
Now change basis with  
1 2 1
Q= √
5 1 −2
 
2x + y
Then x0 = QT x = √15 . Then F = (x0 )2 + 6(y 0 )2 .
x − 2y
So F = c is an ellipse.
6.1 Quadrics and conics
6.1.1 Quadrics
Definition (Quadric). A quadric is an n-dimensional surface defined by the
zero of a real quadratic polynomial, i.e.

xT Ax + bT x + c = 0,

where A is a real n × n matrix, x, b are n-dimensional column vectors and c is a

constant scalar.
As noted in example, anti-symmetric matrix has xT Ax = 0, so for any A,
we can split it into symmetric and anti-symmetric parts, and just retain the
symmetric part S = (A + AT )/2. So we can have

xT Sx + bT x + c = 0

with S symmetric.
Since S is real and symmetric, we can diagonalize it using S = QDQT with
D diagonal. We write x0 = QT x and b0 = QT b. So we have

(x0 )T Dx0 + (b0 )T x0 + c = 0.

If S is invertible, i.e. with no zero eigenvalues, then write x00 = x0 + 12 D−1 b0

which shifts the origin to eliminate the linear term (b0 )T x0 and finally have
(dropping the prime superfixes)

xT Dx = k.

So through two transformations, we have ended up with a simple quadratic form.

6.1.2 Conic sections (n = 2)

From the equation above, we obtain

λ1 x21 + λ2 x22 = k.

We have the following cases:

(i) λ1 λ2 > 0: we have ellipses with axes coinciding with eigenvectors of S.
(We require sgn(k) = sgn(λ1 , λ2 ), or else we would have no solutions at all)

(ii) λ1 λ2 < 0: say λ1 = k/a2 > 0, λ2 = −k/b2 < 0. So we obtain

x21 x2
2
− 22 = 1,
a b
which is a hyperbola.
(iii) λ1 λ2 = 0: Say λ2 = 0, λ1 = 6 0. Note that in this case, our symmetric
matrix S is not invertible and we cannot shift our origin using as above.
From our initial equation, we have

λ1 (x01 )2 + b01 x01 + b02 x02 + c = 0.

We perform the coordinate transform (which is simply completing the

square!)

b01
x001 = x01 +
2λ1
c (b0 )2
x002 = x02 + 0 − 1 0
b2 4λ1 b2

to remove the x01 and constant term. Dropping the primes, we have

λ1 x21 + b2 x2 = 0,

which is a parabola.
Note that above we assumed b02 6= 0. If b02 = 0, we have λ1 (x01 )2 +b01 x01 +c =
0. If we solve this quadratic for x01 , we obtain 0, 1 or 2 solutions for x1
(and x2 can be any value). So we have 0, 1 or 2 straight lines.
These are known as conic sections. As you will see in IA Dynamics and Relativity,
this are the trajectories of planets under the influence of gravity.
6.2 Focus-directrix property
Conic sections can be defined in a different way, in terms of
Definition (Conic sections). The eccentricity and scale are properties of a conic
section that satisfy the following:
Let the foci of a conic section be (±ae, 0) and the directrices be x = ±a/e.
A conic section is the set of points whose distance from focus is e× distance
from directrix which is closer to that of focus (unless e = 1, where we take the
distance to the other directrix).
Now consider the different cases of e:

(i) e < 1. By definition,

y
x = a/e

(x, y)

ae
x
O

p a 
(x − ae)2 + y 2 = e −x
e
x2 y2
+ =1
a2 a2 (1 − e2 )
√
Which is an ellipse with semi-major axis a and semi-minor axis a 1 − e2 .
(if e = 0, then we have a circle)
(ii) e > 1. So

y
x = a/e

(x, y)

ae
x
O

p  a
(x − ae)2 + y 2 = e x −
e
x2 y2
− 2 2 =1
a2 a (e − 1)

and we have a hyperbola.

(iii) e = 1: Then

y
x=a

(x, y)

a
x
O

p
(x − a)2 + y 2 = (x + 1)
y 2 = 4ax

and we have a parabola.

Conics also work in polar coordinates. We introduce a new parameter l such
that l/e is the distance from the focus to the directrix. So

l = a|1 − e2 |.

We use polar coordinates (r, θ) centered on a focus. So the focus-directrix

property is
 
l
r=e − r cos θ
e
l
r=
1 + e cos θ
We see that r → ∞ if θ → cos−1 (−1/e), which is only possible if e ≥ 1, i.e.
hyperbola or parabola. But ellipses have e < 1. So r is bounded, as expected.
7 Transformation groups
We have previously seen that orthogonal matrices are used to transform between
orthonormal bases. Alternatively, we can see them as transformations of space
itself that preserve distances, which is something we will prove shortly.
Using this as the definition of an orthogonal matrix, we see that our definition
of orthogonal matrices is dependent on our choice of the notion of distance, or
metric. In special relativity, we will need to use a different metric, which will
lead to the Lorentz matrices, the matrices that conserve distances in special
relativity. We will have a brief look at these as well.
7.1 Groups of orthogonal matrices
Proposition. The set of all n × n orthogonal matrices P forms a group under
matrix multiplication.
Proof.

0. If P, Q are orthogonal, then consider R = P Q. RRT = (P Q)(P Q)T =

P (QQT )P T = P P T = I. So R is orthogonal.
1. I satisfies II T = I. So I is orthogonal and is an identity of the group.
2. Inverse: if P is orthogonal, then P −1 = P T by definition, which is also
orthogonal.

3. Matrix multiplication is associative since function composition is associative.

Definition (Orthogonal group). The orthogonal group O(n) is the group of

orthogonal matrices.

Definition (Special orthogonal group). The special orthogonal group is the

subgroup of O(n) that consists of all orthogonal matrices with determinant 1.
In general, we can show that any matrix in O(2) is of the form
   
cos θ − sin θ cos θ sin θ
or
sin θ cos θ sin θ − cos θ
7.2 Length preserving matrices
Theorem. Let P ∈ O(n). Then the following are equivalent:
(i) P is orthogonal
(ii) |P x| = |x|
(iii) (P x)T (P y) = xT y, i.e. (P x) · (P y) = x · y.

(iv) If (v1 , v2 , · · · , vn ) are orthonormal, so are (P v1 , P v2 , · · · , P vn )

(v) The columns of P are orthonormal.
Proof. We do them one by one:

(i) ⇒ (ii): |P x|2 = (P x)T (P x) = xT P T P x = xT x = |x|2

(ii) ⇒ (iii): |P (x + y)|2 = |x + y|2 . The right hand side is

(xT + yT )(x + y) = xT x + y T y + yT x + xT y = |x|2 + |y|2 + 2xT y.

Similarly, the left hand side is

|P x + P y|2 = |P x|2 + |P y| + 2(P x)T P y = |x|2 + |y|2 + 2(P x)T P y.

So (P x)T P y = xT y.
(iii) ⇒ (iv): (P vi )T P vj = viT vj = δij . So P vi ’s are also orthonormal.

(iv) ⇒ (v): Take the vi ’s to be the standard basis. So the columns of P , being
P ei , are orthonormal.
(v) ⇒ (i): The columns of P are orthonormal. Then (P P T )ij = Pik Pjk =
(Pi ) · (Pj ) = δij , viewing Pi as the ith column of P . So P P T = I.

Therefore the set of length-preserving matrices is precisely O(n).

7.3 Lorentz transformations
Consider the Minkowski 1 + 1 dimension spacetime (i.e. 1 space dimension and
1 time dimension)
Definition (Minkowski inner product). The Minkowski inner product of 2
vectors x and y is
hx | yi = xT Jy,
where  
1 0
J=
0 −1
Then hx | yi = x1 y1 − x2 y2 .
This is to be compared to the usual Euclidean inner product of x, y ∈ R2 ,
given by
hx | yi = xT y = xT Iy = x1 y1 + x2 y2 .
Definition (Preservation of inner product). A transformation matrix M pre-
serves the Minkowski inner product if

hx|yi = hM x|M yi

for all x, y.
We know that xT Jy = (M x)T JM y = xT M T JM y. Since this has to be
true for all x and y, we must have

J = M T JM.

We can show that M takes the form of

   
cosh α sinh α cosh α − sinh α
Hα = or Kα/2 =
sinh α cosh α sinh α − cosh α

where Hα is a hyperbolic rotation, and Kα/2 is a hyperbolic reflection.

This is technically all matrices that preserve the metric, since these only
include matrices with M11 > 0. In physics, these are the matrices we want, since
M11 < 0 corresponds to inverting time, which is frowned upon.
Definition (Lorentz matrix). A Lorentz matrix or a Lorentz boost is a matrix
in the form  
1 1 v
Bv = √ .
1 − v2 v 1
Here |v| < 1, where we have chosen units in which the speed of light is equal to
1. We have Bv = Htanh−1 v

Definition (Lorentz group). The Lorentz group is a group of all Lorentz matrices
under matrix multiplication.
It is easy to prove that this is a group. For the closure axiom, we have
Bv1 Bv2 = Bv3 , where
v1 + v2
v3 = tanh(tanh−1 v1 + tanh−1 v2 ) =
1 + v1 v 2
The set of all Bv is a group of transformations which preserve the Minkowski
inner product.