Download as pdf or txt
Download as pdf or txt
You are on page 1of 36

UNIT 5:

DIAGONALIZATION

Polytechnic School of Mieres

University of Oviedo
Index

1 Introduction.

2 Diagonalization of matrices and endomorphisms.

3 Conditions for diagonalization.

4 Systems of linear ordinary differential equations.


Introduction.
Linear ordinary differential equations (ODEs).
Systems of first-order, linear, homogeneous ODEs.
Contents

1 Introduction.

2 Diagonalization of matrices and endomorphisms.

3 Conditions for diagonalization.

4 Systems of linear ordinary differential equations.


Introduction.
Linear ordinary differential equations (ODEs).
Systems of first-order, linear, homogeneous ODEs.
Endomorphisms

This unit deals with endomorphisms.


Recall that, if V is a vector space over a commutative field K, then an
endomorphism of V is a linear map T : V → V . In particular, an
endomorphism T : V → V which is bijective is called automorphism of V .

Remark 5.1
Let T be an endomorphism of V .
To study T , we choose the same basis of V as initial and final space.
Both Ker(T ) and Im(T ) are subspaces of V , so that their sum and
their intersection are meaningful.
T ◦T = T 2 (or, in general, T◦T
| ◦·
p
{z· ·◦T} =T ) is an endomorphism of V .
p times
If M is the matrix associated to T w.r.t. a basis of V , then M p is the
matrix associated to T p with respect to such basis.
Contents

1 Introduction.

2 Diagonalization of matrices and endomorphisms.

3 Conditions for diagonalization.

4 Systems of linear ordinary differential equations.


Introduction.
Linear ordinary differential equations (ODEs).
Systems of first-order, linear, homogeneous ODEs.
Diagonalizable endomorphisms

Definition 5.1
Let V be a finite-dimensional vector space over a commutative field K, and
T : V → V an endomorphism. Then T is diagonalizable if there exists a
basis of V with respect to which the matrix associated to T is diagonal.

Remark 5.2
Some endomorphisms are not diagonalizable.
Diagonalization

Diagonalization is the process of finding a basis {~


e1 , . . . , e~n } for which
the matrix associated to T is diagonal:
 
d1 0 · · · 0
 0 d2 · · · 0 
D= . ..  .
 
..
 .. . . 
0 0 · · · dn

In that case,

e1 ) = d1 e~1 , T (~
T (~ e2 ) = d2 e~2 , . . . , T (~
en ) = dn e~n .
Eigenvalues and eigenvectors

Definition 5.2
Let V be a vector space on a commutative field K, and T : V → V an
endomorphism on V . A vector ~v 6= ~0 is called an eigenvector of T ,
associated to an eigenvalue λ ∈ K of T , if T (~v ) = λ~v .

The set of eigenvalues of an endomorphism T is called the spectrum of T ,


n o
σ(T ) = λ ∈ K ; ∃~v 6= ~0 such that T (~v ) = λ~v .
Eigenspaces

Theorem 5.1
Let T : V → V be an endomorphism of V , and λ ∈ K an eigenvalue of T .
The set
{~v ∈ V ; T (~v ) = λ~v } = Ker(T − λI )
is a vector subspace of V , called the eigenspace associated to λ and it is
denoted by ST (λ) or simply Sλ or S(λ).
How to diagonalize an endomorphism

Diagonalization of a diagonalizable endomorphism T : V → V :


1 Choose a basis of V and determine the matrix A associated to T .
2 Compute the characteristic polynomial of T :

p(λ) = det(A − λI ) = (−1)n λn + (−1)n−1 tr(A)λn−1 + · · · + det(A).

3 Find the eigenvalues of T , that are just the roots of the characteristic
equation p(λ) = 0.
4 For each eigenvalue λ ∈ σ(T ),
determine the eigenspace Sλ solving the linear system (A − λI )X = [0];
choose a basis Bλ of the eigenspace Sλ .
5 The union of such bases, ∪λ∈σ(T ) Bλ , defines a basis of V for which
the matrix associated to T is diagonal.
Diagonalizable matrices

A matrix A ∈ Mn (K) is diagonalizable if there exists a diagonalizable


endomorphism whose matrix associated for some basis is A.

Remark 5.3
The definition does not depend on the choice of the endomorphism:
A ∈ Mn (K) is diagonalizable if and only if any endomorphism, with
associated matrix A for some basis, is diagonalizable.
Tipically, when K = R or C, one chooses the endomorphism
T : Kn → Kn whose associated matrix w.r.t. the standard basis is A.
In practice, no endomorphism is needed.
How to diagonalize a matrix

Diagonalization of a diagonalizable matrix A ∈ Mn (K):


Compute the characteristic polynomial of A:

p(λ) = det(A − λI ) = (−1)n λn + (−1)n−1 tr(A)λn−1 + · · · + det(A).

Solve the characteristic equation p(λ) = 0.


For each λ root of the charactersitic polynomial,
solve the linear system (A − λI )X = [0];
choose a basis Bλ of the subspace {X ∈ Kn ; (A − λI )X = [0]}.
Using the column vectors ∪λ Bλ = {X1 , . . . , Xn }, define the matrix
P = [X1 X2 · · · Xn ]. Then A = PDP −1 , being D the diagonal matrix
such that AXi = dii Xi .
Examples

Example 5.1
Let T be the endomorphism of R3 [x] defined by
T (ax 3 + bx 2 + cx + d) = dx 3 + cx 2 + bx + a.
a) Calculate the matrix associated to T w.r.t. the basis {1, x, x 2 , x 3 }.
b) Find a basis for which the matrix associated to T is diagonal, and
compute this matrix.

Example 5.2
Calculate the n-th power of the following matrix:
 
0 1 1
A= 1 0 1 .
1 1 0
Contents

1 Introduction.

2 Diagonalization of matrices and endomorphisms.

3 Conditions for diagonalization.

4 Systems of linear ordinary differential equations.


Introduction.
Linear ordinary differential equations (ODEs).
Systems of first-order, linear, homogeneous ODEs.
Necessary and sufficient conditions for diagonalization
Theorem 5.2
Let T : V → V be an endomorphism of V and λ ∈ σ(T ) a root of
multiplicity m of the characteristic polynomial of T . Then

1 ≤ dim ST (λ) ≤ m.

Theorem 5.3
Let T : V → V be an endomorphism of V . Then, T is diagonalizable over
K if and only if there exist λ1 , λ2 , . . . , λp ∈ K such that
(i) p(λ) = (λ1 − λ)n1 (λ2 − λ)n2 · · · (λp − λ)np , and
(ii) dim S(λi ) = ni ∀i = 1, . . . , p.

Remark 5.4
If a polynomial with real coefficients, p ∈ Rn [x], has a complex root λ,
then also λ is a root of p.
Example

Example 5.3
Consider the endomorphism Tα : R3 → R3 whose associated matrix w.r.t.
the standard basis is
 
2α − 1 0 2α − 2
A= 1 α 2 .
1−α 0 2−α

Find the values of α ∈ R for which Tα is diagonalizable over R.


Examples

Example 5.4
Find out if the matrix  
0 −1
A=
1 0
is diagonalizable over R or C. Diagonalize it whenever possible.

Example 5.5
Diagonalize the following matrix:
 i i 

A= 2 2 .

i i

2 2
Contents

1 Introduction.

2 Diagonalization of matrices and endomorphisms.

3 Conditions for diagonalization.

4 Systems of linear ordinary differential equations.


Introduction.
Linear ordinary differential equations (ODEs).
Systems of first-order, linear, homogeneous ODEs.
Differential equations

Definition 5.3
A differential equation is any equation containing some unknown
function of one or several variables, as well as some of its derivatives with
respect to such variables.

Example 5.6
The following are differential equations:

dy d 4x d 2x ∂2T ∂2T
= 2y 2 , + 5 + 3x = sin(t), + = 0.
dx dt 4 dt 2 ∂x 2 ∂y 2
Order of a differential equation

Definition 5.4
The order of a differential equation is that of the highest derivative of the
unknown function that appears in the equation. We say that a differential
equation of order n is a nth-order differential equation.

Example 5.7
dy
= 2y 2 is a differential equation of order 1;
dx
d 4x d 2x
4
+ 5 2 + 3x = sin(t) is a fourth-order differential equation;
dt dt
∂2T ∂2T
+ = 0 has order 2.
∂x 2 ∂y 2
Ordinary differential equations

Definition 5.5
If the unknown of a differential equation is a funcion of a single variable,
the differential equation is an ordinary differential equation (ODE). On
the contrary, if the unknown is a function of multiple variables and the
equation involves its partial derivatives, the equation is a partial
differential equation (PDE).

Example 5.8
dy
= 2y 2 ODE;
dx
d 4x d 2x
+ 5 + 3x = sin(t) ODE;
dt 4 dt 2
2
∂ T 2
∂ T
2
+ = 0 PDE.
∂x ∂y 2
Solution of an ODE: examples

Example 5.9
Which of the following functions solve the ODE y 00 (x) − 4y (x) = 0?

(i) f (x) = e x ; (ii) g (x) = e 2x ; (iii) h(x) = e −2x ;


(iv ) l(x) = c1 g (x) + c2 h(x) (c1 , c2 ∈ R).

Example 5.10
Consider the ODE
y 0 (x) = 2x.
Solve it integrating on both hands of the ODE. Among all the solutions,
select the particular solution whose value at x = 1 is y (1) = 7.
Linear ordinary differential equations

Definition 5.6
An ordinary differential equation of order n is linear if it is of the form

d ny d n−1 y dy
a0 (x) n
+ a 1 (x) n−1
+ · · · + an−1 (x) + an (x)y = F (x),
dx dx dx
being F (x) and ai (x) (i = 0, 1, . . . , n) given real-valued functions, and a0
not identically zero.

Example 5.11
dy
= 2y 2 is not linear,
dx
d 4x d 2x
+ 5 + 3x = sin(t) is linear.
dt 4 dt 2
System of first-order linear ordinary differential equations

Definition 5.7
A system of first-order linear ordinary differential equations consists
of n first-order linear ODEs that relate n unknown functions y1 (x), y2 (x),
. . . , yn (x) as follows:
 0
y (x) = a11 (x) y1 (x) + a12 (x) y2 (x) + · · · + a1n (x) yn (x) + f1 (x)
 10


y2 (x) = a21 (x) y1 (x) + a22 (x) y2 (x) + · · · + a2n (x) yn (x) + f2 (x)
(1)

 ...
 0
yn (x) = an1 (x) y1 (x) + an2 (x) y2 (x) + · · · + ann (x) yn (x) + fn (x)

Remark 5.5
Every function involved in (1) is a real-valued function of a real variable.
Matrix form

Denote
     
y1 (x) f1 (x) a11 (x) a12 (x) · · · a1n (x)
 y2 (x)   f2 (x)   a21 (x) a22 (x) · · · a2n (x) 
Y = ,F =  ,A =  .
     
.. .. .. .. .. ..
 .   .   . . . . 
yn (x) fn (x) an1 (x) an2 (x) · · · ann (x)

Then, the system of first-order linear ODEs (1) can be rewritten as

Y 0 (x) = A(x)Y (x) + F (x) (2)


Solution
Definition 5.8
Given a system of first-order linear ODEs in matrix form,
Y 0 = A(x)Y + F (x), and an interval I ⊆ R, a vector of functions
Φ : I → Rn is a solution of the system in I if it is differentiable in I and

Φ0 (x) = A(x)Φ(x) + F (x) ∀x ∈ I .

Example 5.12
y10 = xy1 − xy2 − x + 1

The system can be rewritten as
y20 = y1 − y2
   
0 x −x −x + 1
Y = Y+ .
1 −1 0
 
x +1
A solution of the system is Φ(x) = .
x
Initial value problem

Definition 5.9
An initial value problem (or Cauchy problem) is an ODE together with a
specified value (called the initial condition) of the unknown function at a
given point:  0
Y = A(x)Y + F (x),
Y (x0 ) = c.

Theorem 5.4 (Existence and uniqueness of solution)


Consider an initial value problem
 0
Y = A(x)Y + F (x),
Y (x0 ) = c,

where the functions A and F are continuous in an interval I ⊆ R that


contains x0 . Then, there exists a unique solution of the problem in I .
Example

Example 5.13
 
x +1
We have already seen that Φ(x) = is a solution of the system
x
   
x −x −x + 1
Y 0 (x) = Y (x) + .
1 −1 0
 
2
Since Φ(1) = , we deduce that Φ is the unique solution of the
1
initial value problem
    
0 x −x −x + 1
 Y (x) = Y (x) + ,


1 −1 0

 
 2
 Y (1) = .


1
Homogeneous system

Definition 5.10
A system of first-order linear ODEs is called homogeneous if it can be
written in matrix form as
Y 0 = A(x)Y .

General solution:
The general solution of a non-homogeneous system

Y 0 = A(x)Y + F (x)

is any particular solution plus the general solution of the homogeneous


system Y 0 = A(x)Y .
Thus, it is of great importance to know how to solve homogeneous
systems.
Solution set

The set of all possible solutions of a system in an interval I ⊆ R,

{Φ : I → Rn ; Φ solution of the system in I } ,

is called the solution set in I .

Theorem 5.5
The solution set in I of a linear homogeneous system Y 0 = A(x) Y is a
vector space over R with dimension n.
Fundamental set of solutions
Definition 5.11
Let {Φ1 , Φ2 , . . . , Φn } be a basis of the solution set in I of the
homogeneous system
Y 0 = A(x)Y .
Then {Φ1 , Φ2 , . . . , Φn } is called fundamental set of solutions of the
system in I .

If {Φ1 , Φ2 , . . . , Φn } is a fundamental set of solutions of the system


Y 0 = A(x)Y in I , then its solution set in I is
{c1Φ1 + c2 Φ2 + · · · + cn Φn ; c1 , c2 , . . . , cn ∈ R} ,
and its general solution in I is
Y = c1Φ1 + c2 Φ2 + · · · + cn Φn , with c1 , c2 , . . . , cn ∈ R.
Thus, solving the system is equivalent to finding n linearly independent
solutions.
Constant coefficients

Particular case:
Functions aij (x) are constant.

Remark 5.6
Since constant functions are continuous in any interval I ⊆ R, all that we
have studied until now applies.

Simplification:
We assume that the matrix A is diagonalizable over C.
Case 1: real eigenvalues

Theorem 5.6
Let {~v1 , ~v2 , . . . ~vn } be a set of linearly independent eigenvectors of A
associated to the real eigenvalues λ1 , λ2 , . . . , λn , respectively. Then,
n o
~v1 e λ1 x , ~v2 e λ2 x , . . . , ~vn e λn x

is a fundamental set of solutions of the homogeneous system Y 0= AY in R.

Example 5.14
Solve the following system:
 0
 y1 = y1 + y2 − y3 ,
y 0 = −y1 − y2 + y3 ,
 20
y3 = 2y1 + 2y2 − 2y3 .
Case 2: a complex eigenvalue

Let us now consider a matrix A ∈ Mn (R) with a complex eigenvalue


λ ∈ C \ R and an eigenvector w~ associated to λ.

Theorem 5.7
~ is an eigenvector associated to λ.
λ is an eigenvalue of A and w

Lemma 5.1
The complex-valued vector functions Φ(x) = e λx w
~ and Φ(x) = e λx w
~ are
0
solutions of the system Y = A Y in R.

Lemma 5.2
Denoting as before Φ(x) = e λx w
~ , the real-valued vector functions
Re(Φ(x)) and Im(Φ(x)) are solutions of Y 0 = A Y in R.
Case 2: a complex eigenvalue
Given A ∈ Mn (R), let ~v1 , ~v2 , . . . , ~vk , ~vk+1 , . . . , ~vn be linearly independent
|{z} |{z}
~
w ~
w
eigenvectors associated to the eigenvalues λ1 , λ2 , . . . , λk , λk+1 , . . . , λn .
|{z} | {z }
a+bi a−bi
Then
{e λ1 x ~v1 , e λ2 x ~v2 , . . . , e (a+bi)x w ~ , . . . , e λn x ~vn }
~ , e (a−bi)x w
is a fundamental set of complex-valued solutions of Y 0 = A Y , and

{e λ1 x ~v1 , e λ2 x ~v2 , . . . , Re(Φ(x)), Im(Φ(x)), . . . , e λn x ~vn }

is a fundamental set of real-valued solutions of Y 0 = A Y .

Remark 5.7
When there are more than two complex eigenvalues, we proceed as before
for each couple of complex eigenvalues (λk , λk ) to obtain a fundamental
set of real-valued solutions.
Example

Example 5.15
Solve the following system of linear ODEs:
 0
y = −y1 + y2 − y3 ,
 1

y20 = y3 ,

 0
y3 = y1 − y2 + y3 .

You might also like