Chapter 3

Reduction of Endomorphisms

Throughout the rest, K denotes a commutative field (often R or C) and E a K-vector space of
dimension n ∈ N∗ .

Problem

Let f ∈ End(E). Can we find a basis B of E such that MB (f ) is diagonal?

3.1 Generalities
Definition 3.1. Let f ∈ End(E). We say that f is diagonalizable if there exists a basis B of
E such that the matrix MB (f ) is diagonal.

Suppose that f ∈ End(E) is diagonalizable and B = (v1 , . . . , vn ) is a basis of E such that:

 
λ1 0 ··· 0
 
0 λ2 · · · 0

MB (f ) =

,
 . .. . . .. 

 .. . . .
 
 
0 0 ··· λn
where λi ∈ K, for 1 ≤ i ≤ n.

Definition 3.2. Let f ∈ End(E). A vector v ∈ E is called an eigenvector of f if f (v) is

collinear to v, i.e., ∃λ ∈ K such that f (v) = λv.

26
3. Reduction of Endomorphisms 27

Remarks:

1. Since f (0) = 0, the zero vector is always an eigenvector, for any scalar λ.

2. If v is a non-zero eigenvector of f , then the scalar λ is unique. Indeed, if λ1 and λ2 are

two scalars such that f (v) = λ1 v = λ2 v, then (λ1 − λ2 )v = 0, hence λ1 = λ2 .

Definition 3.3. A scalar λ ∈ K is called an eigenvalue of f if there exists a non-zero vector

v ∈ E such that f (v) = λv. In this case, v is an eigenvector associated with the eigenvalue λ,
or equivalently, λ is the eigenvalue associated with the eigenvector v.

Proposition 3.1. Let f ∈ End(E). Then, f is diagonalizable if and only if there exists a basis
of E formed by eigenvectors of f .

Definition 3.4. Let f ∈ End(E) and λ ∈ K. The set of vectors v ∈ E satisfying f (v) = λv is
denoted by Eλ , i.e.,

Eλ = {v ∈ E | f (v) = λv} = {v ∈ E | (f − λIdE )(v) = 0} = ker(f − λIdE ).

Proposition 3.2. With the above notations:

1. Eλ is a subspace of E.

2. λ is an eigenvalue of f if and only if Eλ 6= {0}.

3. If λ is an eigenvalue of f , Eλ is the set of all eigenvectors associated with λ and is called

the eigenspace of f associated with λ. Moreover, dim Eλ ≥ 1.

Definition 3.5. The spectrum of f , denoted by spec(f ), is the set of eigenvalues of f .

II. Eigenvalues and Eigenvectors

II-1 Finding Eigenvalues

Proposition 3.3. Let f ∈ End(E) and λ ∈ K. The following assertions are equivalent:
3. Reduction of Endomorphisms 28

1) λ is an eigenvalue of f .

2) Eλ 6= {0}.

3) f − λIdE is not injective.

4) f − λIdE is not surjective.

5) det(f − λIdE ) = 0.

Remark 3.1. In practice, property (5) is most often used to determine the eigenvalues of f ,
reducing the problem to calculating the determinant of a matrix.

Definition 3.6. Let A ∈ Mn (K) (n ∈ N∗ ). An eigenvector of A is any column matrix X such

that there exists λ ∈ K satisfying

A·X =λ·X

.
An eigenvalue of A is any scalar λ ∈ K such that there exists a non-zero column matrix
X verifying A · X = λ · X. The set of eigenvalues of A is called the spectrum of A, denoted
spec(A).

Remark 3.2. 1. If f ∈ End(E) and B is any basis of E, then any eigenvalue of f is also
an eigenvalue of MB (f ).

2. If v is an eigenvector of f , then the column matrix X of the coordinates of v in B is an

eigenvector of MB (f ).

Definition 3.7. Let A ∈ Mn (K). A is said to be diagonalizable if it is similar to a diagonal

matrix.

Proposition 3.4. Let A ∈ Mn (K) and λ ∈ K. The following statements are equivalent:

i) λ is an eigenvalue of A.

ii) A − λIn is not invertible.

3. Reduction of Endomorphisms 29

iii) det(A − λIn ) = 0.

Definition 3.8. Let A ∈ Mn (K). The characteristic polynomial of A, denoted PA (X), is

defined by:
PA (X) = det(A − XIn ),

where A − XIn is considered a matrix with coefficients in the field K(X) of rational fractions
in X.

Proposition 3.5. Let A ∈ Mn (K). Then, PA (X) is a polynomial of degree n.

Proposition 3.6. Let A ∈ Mn (K) and λ ∈ K. Then:

λ is an eigenvalue of A ⇐⇒ λ is a root of PA (X).

Remark 3.3. The eigenvalues of A ∈ Mn (K) are the roots of its characteristic polynomial
PA (X) of degree n. Therefore, the number of eigenvalues of A is at most n (exactly n if
K = C).

Definition 3.9. Let λ be an eigenvalue of A ∈ Mn (K). The multiplicity, denoted m(λ), of the
eigenvalue λ is the order of multiplicity of λ as a root of the characteristic polynomial PA (X).
Specifically:

• If m(λ) = 1, λ is a simple eigenvalue.

• If m(λ) ≥ 2, λ is a multiple eigenvalue.

• If m(λ) = 2, λ is a double eigenvalue.

• If m(λ) = 3, λ is a triple eigenvalue.

Proposition 3.7. Two similar matrices have the same characteristic polynomial.
3. Reduction of Endomorphisms 30

II-2 Finding Eigenvectors

Proposition 3.8. Let f ∈ End(E), B be a basis of E, and A = MB (f ). To find the column
matrices X representing the coordinates (in basis B) of the eigenvectors of f , proceed as follows:
Let λ ∈ Spec(A). A column matrix X is an eigenvector of A associated with λ if and only
if:
det(A − λI)X = 0.

Note that this system is not of Cramer type.

Example 25. Given the following matrices:

 
2 0 4
 
1. A = 3

−4 12 , 
 
1 −2 5
 

 
2 0 1
 
2. B = 0 3 1 .
 
 
0 0 2
 

Proposition 3.9. Let A ∈ Mn (K) and λ ∈ Spec(A). Then:

1. dim(Eλ ) = n − rank(A − λIn ),

2. 1 ≤ dim(Eλ ) ≤ m(λ).

III. Diagonalization
Theorem 3.1. Let f ∈ End(E), and let λ1 , . . . , λp be p distinct eigenvalues of f (p ∈ N∗ ).
Then, the sum Eλ1 + · · · + Eλp of the corresponding eigenspaces is direct.

Example 26. Let the matrix:  

2 0 0
 
A = 0 3 0 .
 
 
0 0 4
 
3. Reduction of Endomorphisms 31

The eigenvalues of A are the diagonal entries: λ1 = 2, λ2 = 3, λ3 = 4. For each eigenvalue λ,

the eigenspace Eλ is:
Eλ1 = h{(1, 0, 0)}i.

For λ2 = 3: The eigenspace is:

Eλ2 = h{(0, 1, 0)}i.

For λ3 = 4: The eigenspace is:

Eλ3 = h{(0, 0, 1)}i.

Their sum is:

Eλ1 + Eλ2 + Eλ3 = R3 .

Since the basis vectors of these eigenspaces are linearly independent, their sum is direct:

Eλ1 ⊕ Eλ2 ⊕ Eλ3 = R3 .

Remark 3.4. From the above result, non-zero eigenvectors associated with pairwise distinct
eigenvalues of f ∈ End(E) are linearly independent.

Corollary 3.1. Let f ∈ End(E). If f has n distinct eigenvalues, then f is diagonalizable.

Example 27. Consider:  

2 0 4
 
A= 3 −4 12 .
 
 
1 −2 5
 

Note: This corollary provides a sufficient condition for diagonalization.

Proposition 3.10. Let f ∈ End(E) and λ1 , . . . , λp be distinct eigenvalues of f . Then:

f is diagonalizable if and only if E = Eλ1 ⊕ · · · ⊕ Eλp .

Corollary 3.2. Let f ∈ End(E) and λ1 , . . . , λp be distinct eigenvalues of f . Then:

f is diagonalizable if and only if dim(E) = dim(Eλ1 ) + · · · + dim(Eλp ).

3. Reduction of Endomorphisms 32

Proposition 3.11. Let f ∈ End(E). Then:


P A (X) has n roots (counted with multiplicities),


f is diagonalizable (over K) if and only if
dim(E
λ) = m(λ) for every eigenvalue λ of A.



Example 28. Consider:  

2 0 4
 
A= 3 −4 12 .
 
 
1 −2 5
 

This matrix demonstrates diagonalizability under the given conditions.