Linear and Bilinear Algebra

Lebanese university

Ali Ayad
Assistant professor at the lebanese university - Section 1
[email protected]

Course & proposed exercises

Chapter 1

Recall of some notions in linear

algebra

In this chapter, we recall some notions from linear algebra which are interesting for us in the
rest of this book. Most of the results of this chapter are given without demonstrations (see
any course of linear algebra).

In all this chapter, K denotes a commutative field, E and F are two K-vector spaces of
finite dimensions n ∈ N∗ and m ∈ N∗ respectively.

1.1 Direct sum

Proposition 1.1.1 Let F1 , . . . , Fr be subspaces of E. The set
n o
F1 + · · · + Fr = x1 + · · · + xr , such that x1 ∈ F1 , . . . , xr ∈ Fr

is a subspace of E. We call it the sum of F1 , . . . , Fr .

Definition 1.1.1 Let F1 , . . . , Fr be subspaces of E. We say that the sum F1 + · · · + Fr is

direct if, for all x1 ∈ F1 , . . . , xr ∈ Fr ,

x1 + · · · + xr = 0 ⇒ x1 = · · · = xr = 0.

In this case, we write

F1 + · · · + Fr = F1 ⊕ · · · ⊕ Fr .

Proposition 1.1.2 Let F1 , . . . , Fr be subspaces of E such that

Fi = vect(vi ) where vi ∈ E with vi 6= 0, ∀ 1 ≤ i ≤ r.

Then F1 + · · · + Fr = F1 ⊕ · · · ⊕ Fr if and only if the vectors v1 , . . . , vr are linearly

independent over K.

2
Chapter 1 Recall of some notions in linear algebra

Proposition 1.1.3 Let F1 , . . . , Fr be subspaces of E. Then F1 + · · · + Fr = F1 ⊕ · · · ⊕ Fr

if and only if, for all 1 ≤ i ≤ r,
 
r
X 
Fi ∩ 
  = {0}.
Fj 
j=1
j6=i

Corollary 1.1.1 Let F1 and F2 be two subspaces of E. Then F1 + F2 = F1 ⊕ F2 if and

only if
F1 ∩ F2 = {0}.

Remark 1.1.1 Let F1 , . . . , Fr be subspaces of E. If F1 + · · · + Fr = F1 ⊕ · · · ⊕ Fr , then

Fi ∩ Fj = {0} for all i 6= j. But the converse is not necessarily true.

Proposition 1.1.4 Let F1 , . . . , Fr be subspaces of E such that

F1 + · · · + Fr = F1 ⊕ · · · ⊕ Fr .

Then, every element x ∈ F1 ⊕ · · · ⊕ Fr can be written, in a unique way, in the form:

x = x1 + · · · + xr , where x1 ∈ F1 , . . . , xr ∈ Fr .

Theorem 1.1.1 Let F1 , . . . , Fr be subspaces of E such that

F1 + · · · + Fr = F1 ⊕ · · · ⊕ Fr .
r
[
If, for every 1 ≤ i ≤ r, bi is a basis of Fi over K, then the set b = bi is a basis of
i=1
F1 ⊕ · · · ⊕ Fr over K.

Corollary 1.1.2 Let F1 , . . . , Fr be subspaces of E. Then

dim(F1 ⊕ · · · ⊕ Fr ) = dim F1 + · · · + dim Fr .

Definition 1.1.2 Let F1 , . . . , Fr be subspaces of E. We say that E is the direct sum of

F1 , . . . , Fr if E = F1 ⊕ · · · ⊕ Fr .

Proposition 1.1.5 Let F1 , . . . , Fr be subspaces of E.

Then E = F1 ⊕ · · · ⊕ Fr if and only if the following two conditions are satisfied:
i) dim E = dim F1 + · · · + dim Fr .

ii) F1 + · · · + Fr = F1 ⊕ · · · ⊕ Fr .

c Ali Ayad page 3

Chapter 1 Recall of some notions in linear algebra

1.2 Representation of a vector in a basis

Proposition 1.2.1 Let b = {u1 , . . . , un } be a basis of E over K. Every element x of E
can be written, in a unique way, in the form:

x = x1 u1 + · · · + xn un , where x1 , . . . , xn ∈ K.
 
x1
The column matrix X = M (x, b) =  ...  is called the column vector of coordinates (or
 

xn
the representative matrix) of x in the basis b (or relatively to the basis b).

Example 1.2.1 The column vector of coordinates of v = (1, 2, 3) ∈ R3 in the canonical

basis C = {e1 , e2 , e3 } of R3 is:
 
1
X = M (v, C) =  2  .
3

Proposition 1.2.2 Let x, y ∈ E, α ∈ K and b be a basis of E over K. Then

1)
(x = y) ⇔ M (x, b) = M (y, b).

2)
M (x + y, b) = M (x, b) + M (y, b).

3)
M (αx, b) = α M (x, b).

Proposition 1.2.3 Let b = {u1 , . . . , un } be a basis of E over K and b0 = {v1 , . . . , vn } be

a subset of E. Let P ∈ Mn (K) be the square matrix whose columns
are the column vectors
of coordinates of v1 , . . . , vn in the basis b. In other words, P = pij ∈ Mn (K) where, for
any 1 ≤ j ≤ n,
vj = p1j u1 + · · · + pnj un .
Then b0 is a basis of E over K if and only if P is invertible. Moreover, if b0 is a basis of E
over K, then the invertible matrix P ∈ Mn (K) is called the transition matrix from the basis
b to the basis b0 .

Example 1.2.2 Consider the following vectors in R3 :

v1 = (2, 0, −1), v2 = (1, −1, 0), v3 = (1, 0, −1).

c Ali Ayad page 4

Chapter 1 Recall of some notions in linear algebra

In order to prove that the set B = {v1 , v2 , v3 } is a basis of R3 over R, we consider the
matrix P whose columns are the column vectors of coodinates of v1 , v2 , v3 in the canonical
basis C = {e1 , e2 , e3 } of R3 , i.e.,
 
2 1 1
P =  0 −1 0  .
−1 0 −1

As |P | = 1 6= 0, then P is invertible, and therefore B is a basis of R3 over R. Consequently,

P is the transition matrix from the basis C to the basis B.

Remark 1.2.1
1) If P is the transition matrix from a basis b to a basis b0 of E over K, then P −1 is the
transition matrix from the basis b0 to the basis b.

2) Every invertible matrix is a transition matrix. More precisely, if P is an invertible

matrix and if b is a basis of E over K, then there exists a basis 0
0

b of E over K such that
P is the transition matrix from b to b . Indeed, if P = pij ∈ Mn (K) is an invertible
matrix and b = {u1 , . . . , un } is a basis of E over K, then, for every 1 ≤ j ≤ n, put:

vj = p1j u1 + · · · + pnj un .

As P is invertible, then the set b0 = {v1 , . . . , vn } is a basis of E over K by the

proposition 1.2.3, and therefore P is the transition matrix from the basis b to the basis
b0 . Moreover, as P −1 is the transition matrix from the basis b0 to the basis b, then, in
order to determine P −1 , it is sufficient to calculate u1 , . . . , un as linear combinations
over K of v1 , . . . , vn .

Example 1.2.3 Let P be the transition matrix from the canonical basis C to the basis B of
R3 defined in the example 1.2.2. The matrix P −1 is the transition matrix from B to C. As
 
 v1 = 2e1 − e3  e1 = v1 − v3
v2 = e1 − e2 , then e2 = v1 − v2 − v3
v3 = e1 − e3 e3 = v1 − 2v3
 

So  
1 1 1
P −1 =  0 −1 0  .
−1 −1 −2

Proposition 1.2.4 Let b and b0 be two bases of E over K and P be the transition matrix
from b to b0 . Let x be an element of E and X (resp. X 0 ) be the column vector of coordinates
of x in b (resp. b0 ). Then

X = P X0 and X 0 = P −1 X.

c Ali Ayad page 5

Chapter 1 Recall of some notions in linear algebra

Example 1.2.4 Let B = {v1 , v2 , v3 } be the basis of R3 over R given in the exam-
ple 1.2.2 Let P be the transition matrix from the canonical basis C to the basis B. By
the proposition 1.2.4 and by using the example 1.2.3, the column vector of coordinates of
v = (1, 2, 3) ∈ R3 in the basis B is:
    
1 1 1 1 6
X 0 = P −1 X =  0 −1 0   2  =  −2  .
−1 −1 −2 3 −9

Hence v = 6v1 − 2v2 − 9v3 .

1.3 Representation of an endomorphism in a basis

Definition 1.3.1 A mapping f from E to F is said to be a homomorphism from E to F if
it satisfies the following two conditions:

i) f (x + y) = f (x) + f (y) for all x, y ∈ E.

ii) f (αx) = αf (x) for all α ∈ K and x ∈ E.

A homomorphism from E to F is also called a K-linear mapping from E to F . We denote

by L(E, F ) the set of homomorphisms from E to F .

Definition 1.3.2

1) An isomorphism from E onto F is every bijective homomorphism from E onto F .

2) We say that E is isomorphic to F , and we write E ' F , if there exists an isomorphism

from E onto F .

3) An endomorphism of E is every homomorphism from E to E. We denote by L(E) or

End(E) the set of endomorphisms of E, i.e., L(E) = L(E, E).

4) An automorphism of E is every bijective endomorphism of E, i.e., every isomorphism

from E onto E. We denote by Aut(E) the set of automorphisms of E.

Remark 1.3.1 A homomorphism from E to F is completely defined by its action on a basis

of E over K. In other words, if b = {u1 , . . . , un } is a basis of E over K and if f and g are
two homomorphisms from E to F such that

(∗) f (ui ) = g(ui ), ∀ 1 ≤ i ≤ n,

then f = g. Indeed, let x ∈ E, then there exist x1 , . . . , xn ∈ K such that

x = x1 u1 + · · · + xn un .

c Ali Ayad page 6

Chapter 1 Recall of some notions in linear algebra

So

f (x) = f (x1 u1 + · · · + xn un )
= x1 f (u1 ) + · · · + xn f (un ) (since f is a homomorphism)
= x1 g(u1 ) + · · · + xn g(un ) (by (∗))
= g(x1 u1 + · · · + xn un ) (since g is a homomorphism)
= g(x).

Hence f = g.

Example 1.3.1 Let f be the endomorphism of R2 defined by:

f (e1 ) = (1, 2), f (e2 ) = (3, 4),

where C = {e1 , e2 } is the canonical basis of R2 . In order to determine the expression of

f (x) for all x = (x1 , x2 ) ∈ R2 , we write:

f (x) = f (x1 e1 + x2 e2 )
= x1 f (e1 ) + x2 f (e2 ) (since f is an endomorphism of E)
= x1 (1, 2) + x2 (3, 4) = (x1 + 3x2 , 2x1 + 4x2 ).

For example, f (−1, 1) = (2, 2).

Definition 1.3.3 Let f be an endomorphism of E. We say that a subspace V of E is closed

under f if f (V ) ⊆ V .

Proposition 1.3.1 Let f be an endomorphism of E. If V is a subspace of E closed under

f , then the restriction f/V of f on V is an endomorphism of V .

Proof Let g = f/V be the restriction of f on V . Since V is closed under f , then

f (V ) ⊆ V , and therefore g is a mapping from V to V . Moreover, for all x, y ∈ V
and α ∈ K,
g(x + y) = f (x + y) = f (x) + f (y) = g(x) + g(y)
and
g(αx) = f (αx) = αf (x) = αg(x).
So g is an endomorphism of V . 2

Definition 1.3.4 Let f : E 7−→ F be a homomorphism.

1) We call kernel of f , denoted by ker f or ker(f ), the inverse image of the singleton
{0} by f , i.e.,
  n o
ker f = f −1 {0} = x ∈ E such that f (x) = 0 .

c Ali Ayad page 7

Chapter 1 Recall of some notions in linear algebra

2) We call image of f , denoted by Im f or Im(f ), the direct image of E by f , i.e.,

n o
Im f = f (E) = f (x), x ∈ E .

Proposition 1.3.2 Let f : E 7−→ F be a homomorphism. Then

1) ker f is a subspace of E.
2) Im f is a subspace of F .
3) f is injective if and only if ker f = {0}.

Definition 1.3.5 Let f : E 7−→ F be a homomorphism. We call rank of f , and we denote

it by rk(f ), the dimension of the image of f , i.e.,

rk(f ) = dim(Im f ).

Theorem 1.3.1 (Rank theorem)

If f : E 7−→ F is a homomorphism, then

rk(f ) + dim(ker f ) = dim E.

Proposition 1.3.3 Let f be an endomorphism of E. Then the following properties are

equivalent:
1) f is injective.
2) f is surjective.
3) f is bijective.

Proposition 1.3.4 Let f be an endomorphism of E and b = {u1 , . . . , un } be a basis of E

over K. If f is injective, then the set
n o
f (b) = f (u1 ), . . . , f (un )

is a basis of E over K.

Definition 1.3.6 Let f : E 7−→ F be a homomorphism. Let b = {u1 , . . . , un } be a basis

of E over K and b0 = {v1 , . . . , vm } a basis of F over K. Let A ∈ Mm,n (K) be the matrix
whose columns are the
column vectors of coordinates of f (u1 ), . . . , f (un ) in the basis b0 . In
other words, A = aij ∈ Mm,n (K) where, for any 1 ≤ j ≤ n,

f (uj ) = a1j v1 + · · · + amj vm .

The matrix A is called the representative matrix of f in the bases b and b0 (or relatively to
the bases b and b0 ), it is denoted by A = M (f, b, b0 ).

c Ali Ayad page 8

Chapter 1 Recall of some notions in linear algebra

Example 1.3.2 Let f : R3 7−→ R2 be the homomorphism defined by:

f (x1 , x2 , x3 ) = (2x1 + x3 , x1 + x2 − x3 ).

The representative matrix of f in the canonical bases C = {e1 , e2 , e3 } and C 0 = {e01 , e02 }
of R3 and R2 respectively is:
 
2 0 1
A = M (f, C, C 0 ) = ∈ M2,3 (K),
1 1 −1

by using the fact that

 f (e1 ) = (2, 1) = 2e01 + e02


f (e2 ) = (0, 1) = e02

f (e3 ) = (1, −1) = e01 − e02


Definition 1.3.7 Let f be an endomorphism of E and b be a basis of E over K. The

representative matrix of f in the basis b (or relatively to the basis b), denoted by M (f, b),
is the representative matrix of f in the bases b and b, i.e.,

M (f, b) = M (f, b, b) ∈ Mn (K).

Example 1.3.3 Let f be the endomorphism of R3 defined by:

f (x1 , x2 , x3 ) = (−x2 , x1 + x2 + x3 , x1 − x2 − x3 ).

The representative matrix of f in the canonical basis C = {e1 , e2 , e3 } of R3 is:

 
0 −1 0
A = M (f, C) =  1 1 1  ∈ M3 (R),
1 −1 −1

by using the fact that


 f (e1 ) = (0, 1, 1) = e2 + e3
f (e2 ) = (−1, 1, −1) = −e1 + e2 − e3
f (e3 ) = (0, 1, −1) = e2 − e3


Proposition 1.3.5 Let f and g be two endomorphisms of E, α ∈ K and b a basis of E

over K. Then
1) M (idE , b) = In , where idE is the identity endomorphism of E.

2) (f = g) ⇔ M (f, b) = M (g, b).

3) M (f + g, b) = M (f, b) + M (g, b).

4) M (αf, b) = α M (f, b).

5) M (f ◦ g, b) = M (f, b) M (g, b).

c Ali Ayad page 9

Chapter 1 Recall of some notions in linear algebra

Proposition 1.3.6 Let f : E 7−→ F be a homomorphism, b a basis of E over K and b0 a

basis of F over K. Let A = M (f, b, b0 ), then, for any x ∈ E, we have:
 
M f (x), b0 = A M (x, b).

By abuse of notation, sometimes we write f (x) = Ax.

Example 1.3.4 Let f be an endomorphism of R3 such that

 
0 −1 0
M (f, C) = A =  1 1 1 .
1 −1 −1
In order to determine the expression of f (x) for all x = (x1 , x2 , x3 ) ∈ R3 , we use the
proposition 1.3.6:
    
  0 −1 0 x1 −x2
M f (x), C = A M (x, C) =  1 1 1   x2  =  x1 + x2 + x3  .
1 −1 −1 x3 x1 − x2 − x3
So
f (x) = (−x2 , x1 + x2 + x3 , x1 − x2 − x3 ).

Corollary 1.3.1 Let f be an endomorphism of E, b a basis of E over K and A = M (f, b).

Then n o
ker f = x ∈ E, A M (x, b) = 0 .
In other words, the column vectors of coordinates of the elements of ker f in the basis b are
the solutions in Kn of the homogeneous linear system AX = 0. By abuse of notation, we
write ker f = ker A.

Example 1.3.5 Let f be the endomorphism of R3 defined in the example 1.3.3 and C =
{e1 , e2 , e3 } be the canonical basis of R3 . In order to determine ker f , we solve the homoge-
neous linear system AX = 0 where
   
0 −1 0 x1
A = M (f, C) =  1 1 1  and X =  x2  .
1 −1 −1 x3

x = (x1 , x2 , x3 ) ∈ ker f ⇔ AX = 0

 −x2 = 0
⇔ x1 + x2 + x3 = 0
x1 − x2 − x3 = 0

⇔ x1 = x2 = x3 = 0
⇔ x = (0, 0, 0).
So ker f = {0}, and therefore f is injective. Hence f is bijective.

c Ali Ayad page 10

Chapter 1 Recall of some notions in linear algebra

Proposition 1.3.7 For every matrix A ∈ Mn (K), there exists one and only one endomor-
phism fA of Kn such that A = M (fA , C) where C is the canonical basis of Kn . We call it
the endomorphism of Kn canonically associated to A.
Proof Let C = {e1 , . . . , en } be the canonical basis of Kn . For every x = (x1 , . . . , xn ) ∈
Kn , put
fA (x) = y1 e1 + · · · + yn en ,
where    
y1 x1
 ..  .. 
 
 .  = A i.e., M fA (x), C = A M (x, C).

. 
yn xn
fA is an endomorphism of Kn : Let x, x0 ∈ Kn , then
  h i
M fA (x + x0 ), C = A M (x + x0 , C) = A M (x, C) + M (x0 , C)
= A M (x, C) + A M (x0 , C)
   
= M fA (x), C + M fA (x0 ), C
 
= M fA (x) + fA (x0 ), C .

So
fA (x + x0 ) = fA (x) + fA (x0 ).
Let α ∈ K, then
  h i h i
M fA (αx), C = A M (αx, C) = A αM (x, C) = α A M (x, C)
h  i  
= α M fA (x), C = M αfA (x), C .

So
fA (αx) = αfA (x).
Hence fA is an endomorphism of

Kn .
A = M (fA , C): Put A = aij . For any 1 ≤ j ≤ n,


a1j
M fA (ej ), C = A M (ej , C) =  ...  .
 
 

anj
So A = M (fA , C).
Uniqueness of fA : If there exists an endomorphism g of Kn such that A = M (g, C), then
M (g, C) = M (fA , C), and therefore g = fA by the proposition 1.3.5. 2

Proposition 1.3.8 Let f : E 7−→ F be a homomorphism, b and B two bases of E over K

and b0 and B0 two bases of F over K. If A = M (f, b, b0 ) and B = M (f, B, B0 ), then
B = Q−1 AP,
where P is the transition matrix from the basis b to the basis B and Q is the transition
matrix from the basis b0 to the basis B0 .

c Ali Ayad page 11

Chapter 1 Recall of some notions in linear algebra

Remark 1.3.2 Let f : E 7−→ F be a homomorphism. Let b = {u1 , . . . , un } be a basis of

E over K, b0 a basis of F over K and A = M (f, b, b0 ). Then
 
rk(A) = dim vect f (u1 ), . . . , f (un ) = dim Im f = rk(f ).

So the rank of f is equal to the rank of the representative matrix of f in a basis of E over
K and a basis of F over K. This is independent of the choice of the bases. Indeed, if B is
another basis of E over K and B0 is another basis of F over K, and if B = M (f, B, B0 ),
then B = Q−1 AP by the proposition 1.3.8. So

rk(B) = rk(Q−1 AP ) = rk(AP ) = rk(A).

by using the fact that if M ∈ Mm,n (K), R ∈ Mm (K) and S ∈ Mn (K) such that R and
S are invertible, then
rk(RM ) = rk(M S) = rk(M ).

Corollary 1.3.2 Let f be an endomorphism of E and b and B two bases of E over K. If

A = M (f, b) and B = M (f, B), then

B = P −1 AP,

where P is the transition matrix from the basis b to the basis B.

Example 1.3.6 Let f be the endomorphism of R3 defined in the example 1.3.3 and B =
{v1 , v2 , v3 } the basis of R3 over R given in the example 1.2.2.
1) In order to determine the representative matrix B = M (f, B) of f in B, we distinguish
the following two methods:
1st method: We use the definition. We have:

 f (v1 ) = (0, 1, 3) = e2 + 3e3
f (v2 ) = (1, 0, 2) = e1 + 2e3
f (v3 ) = (0, 0, 2) = 2e3 .


In order to determine the column vectors of coordinates of f (v1 ), f (v2 ), f (v3 ) in the
basis B, we use the proposition 1.2.4 and the example 1.2.3:
    
    1 1 1 0 4
M f (v1 ), B = P −1 M f (v1 ), C =  0 −1 0   1  =  −1 
−1 −1 −2 3 −7
    
    1 1 1 1 3
M f (v2 ), B = P −1 M f (v2 ), C =  0 −1 0   0  =  0 
−1 −1 −2 2 −5
    
    1 1 1 0 2
M f (v3 ), B = P −1 M f (v3 ), C =  0 −1 0   0  =  0  .
−1 −1 −2 2 −4

c Ali Ayad page 12

Chapter 1 Recall of some notions in linear algebra

So 
 f (v1 ) = (0, 1, 3) = 4v1 − v2 − 7v3
f (v2 ) = (1, 0, 2) = 3v1 − 5v3
f (v3 ) = (0, 0, 2) = 2v1 − 4v3 .


Hence  
4 3 2
B =  −1 0 0 .
−7 −5 −4
2nd method: Let P be the transition matrix from C to B. By the corollary 1.3.2 and by
using the examples 1.2.2, 1.2.3 and 1.3.3, we obtain:
   
1 1 1 0 −1 0 2 1 1
B = P −1 AP =  0 −1 0   1 1 1   0 −1 0 
−1 −1 −2 1 −1 −1 −1 0 −1
 
4 3 2
=  −1 0 0 .
−7 −5 −4

2) In order to determine the expression of f (x) in terms of the coordinates x01 , x02 , x03
of x in the basis B of R3 (i.e., for x = x01 v1 + x02 v2 + x03 v3 ∈ R3 ), we use the
proposition 1.3.6:
  0 
  4 3 2 x1
M f (x), B = B M (x, B) =  −1 0 0   x02 
−7 −5 −4 x03
4x01 + 3x02 + 2x03
 

=  −x01 .
0 0
−7x1 − 5x2 − 4x3 0

So
f (x) = (4x01 + 3x02 + 2x03 )v1 − x01 v2 + (−7x01 − 5x02 − 4x03 )v3 .

Proposition 1.3.9 Let f be an endomorphism of E and b a basis of E over K. Let

A = M (f, b) be the representative matrix of f in the basis b. Then f is bijective (i.e., f is
an automorphism of E) if and only if A is invertible.

Example 1.3.7 The endomorphism f of R3 defined in the example 1.3.3 is an automor-

phism of R3 since the matrix A = M (f, C) is invertible (from the fact that |A| = −2 6= 0).

1.4 Similar matrices

Definition 1.4.1 Let A, B ∈ Mn (K). We say that A is similar to B, and we write
A ' B, if there exists an invertible matrix P ∈ Mn (K) such that B = P −1 AP .

c Ali Ayad page 13

Chapter 1 Recall of some notions in linear algebra

Proposition 1.4.1 The similarity relation ' is an equivalence relation on Mn (K).

Proof Let A, B, C ∈ Mn (K).

• As In is invertible and A = In−1 AIn , then A ' A. So ' is reflexive.

• If A ' B, then there exists P ∈ Mn (K) invertible such that B = P −1 AP . So

A = P BP −1 = (P −1 )−1 BP −1 = Q−1 BQ

where Q = P −1 ∈ Mn (K) is invertible. Therefore B ' A. Hence ' is symmetric.

• If A ' B and B ' C, then there exist P, Q ∈ Mn (K) invertible such that

B = P −1 AP and C = Q−1 BQ.

So

C = Q−1 (P −1 AP )Q = (Q−1 P −1 )A(P Q) = (P Q)−1 A(P Q) = R−1 AR,

where R = P Q ∈ Mn (K) is invertible. Therefore A ' C. Hence ' is transitive.

Thus ' is an equivalence relation on Mn (K). 2

Example 1.4.1
1) The only matrix similar to the identity matrix In is In itself. Indeed, if A is similar
to In , then In is similar to A (since ' is symmetric), so there exists P ∈ Mn (K)
invertible such that
A = P −1 In P = P −1 P = In .
Conversely, In is similar to In (since ' is reflexive).

2) The only matrix similar to the zero matrix 0 is 0 itself.

Proposition 1.4.2 Two similar matrices have the same determinant, same trace and same
rank.

Proof Let A, B ∈ Mn (K) be two similar matrices, then there exists P ∈ Mn (K) invertible
such that B = P −1 AP . So
1
|B| = |P −1 AP | = |P −1 ||A||P | =|A||P | = |A|
|P |
   
tr(B) = tr (P −1 A)P = tr P (P −1 A) = tr(In A) = tr(A)
   
rk(B) = rg (P −1 A)P = rg P −1 A = rk(A),
by using the fact that if M, N, Q ∈ Mn (K) such that Q is invertible, then

tr(M N ) = tr(N M ) and rk(M Q) = rk(QM ) = rk(M ). 2

c Ali Ayad page 14

Chapter 1 Recall of some notions in linear algebra

Remark 1.4.1

1) Since ' is symmetric, then, if A ' B, then we say that A and B are similar.

2) The contrapositive of the proposition 1.4.2 is useful to prove that two matrices are not
similar, i.e., if two matrices have not the same determinant or not the same trace or
not the same rank, then they are not similar. For example, the two matrices
   
1 3 4 1 0 8
A =  −1 0 0  and B =  1 2 0 
−6 −8 −4 −4 −6 7

are not similar since tr(A) = −3 6= 10 = tr(B).

3) The converse of the proposition 1.4.2 is not necessarily true. Indeed, the two matrices
   
1 0 1 1
A= and B =
0 1 0 1

have the same determinant, same trace and same rank. On the other hand, A and
B are not similar since A = I2 and the only matrix similar to I2 is I2 itself by the
example 1.4.1.

Theorem 1.4.1 Two matrices in Mn (K) are similar if and only if they are the representative
matrices of an endomorphism f of Kn in two bases of Kn over K. In other words, if A, B ∈
Mn (K), then A and B are similar if and only if there exists an endomorphism f of Kn and
two bases b and b0 of Kn over K such that

A = M (f, b) and B = M (f, b0 ).

Proof
N.C. Since A and B are similar, then there exists P ∈ Mn (K) invertible such that B =
P −1 AP . Let f = fA be the endomorphism of Kn canonically associated to A, i.e., A =
M (fA , C) where C is the canonical basis of Kn (proposition 1.3.7). Since P is invertible,
then, by the remark 1.2.1, there exists a basis B of Kn such that P is the transition matrix
from C to B. By the corollary 1.3.2, we have:

M (f, B) = P −1 M (fA , C)P = P −1 AP = B.

S.C. Let P be the transition matrix from the basis b to the basis b0 , then P is invertible
and B = P −1 AP by the corollary 1.3.2. So A and B are similar. 2

Example 1.4.2 The two matrices

   
1 2 4 2 1 4
A= 1 2 4  and B= 2 1 4 
1 3 3 3 1 3

c Ali Ayad page 15

Chapter 1 Recall of some notions in linear algebra

are similar. Indeed, let f = fA be the endomorphism of R3 canonically associated to A, i.e.,

A = M (f, C) where C = {e1 , e2 , e3 } is the canonical basis of R3 (proposition 1.3.7). So

 f (e1 ) = e1 + e2 + e3
f (e2 ) = 2e1 + 2e2 + 3e3
f (e3 ) = 4e1 + 4e2 + 3e3


Let’s find a basis B = {v1 , v2 , v3 } of R3 over R such that B = M (f, B). As B is obtained
from A by interchanging the first two columns, then we put

v1 = e2 , v2 = e1 , v3 = e3 .

So 
 f (v1 ) = f (e2 ) = 2v1 + 2v2 + 3v3
f (v2 ) = f (e1 ) = v1 + v2 + v3
f (v3 ) = f (e3 ) = 4v1 + 4v2 + 3v3


It is obvious that B is a basis of R3 over R and B = M (f, B). Hence A and B are similar.

Remark 1.4.2 Let f be an endomorphism of E. We call determinant (resp. trace) of

f , and we denote it by det(f ) (resp. tr(f )), the determinant (resp. the trace) of the
representative matrix of f in a basis of E over K. This definition is independent of the
choice of the basis. Indeed, if b and b0 are two bases of E over K, then the two matrices
A = M (f, b) and B = M (f, b0 ) are similar by the theorem 1.4.1, and therefore A and B
have the same determinant and same trace by the proposition 1.4.2. Recall that the rank of
f is already defined in the remark 1.3.2.

Example 1.4.3 Let f be the endomorphism of R3 defined in the example 1.3.3. The repre-
sentative matrix of f in the canonical basis C = {e1 , e2 , e3 } of R3 is:
 
0 −1 0
A = M (f, C) =  1 1 1 .
1 −1 −1

So
det(f ) = |A| = −2, tr(f ) = tr(A) = 0 and rk(f ) = rk(A) = 3.

Proposition 1.4.3 Let λ ∈ K, f and g be two endomorphisms of E.

1) f is bijective (i.e., invertible) if and only if det(f ) 6= 0. Moreover, if f is invertible,

then
1
det(f −1 ) = .
det(f )

2) We have:
rk(f ) = dim(Im f ) = dim E − dim(ker f ).

c Ali Ayad page 16

Chapter 1 Recall of some notions in linear algebra

3) We have:
det(f ◦ g) = det(f ) det(g), tr(f ◦ g) = tr(g ◦ f ),
tr(f + g) = tr(f ) + tr(g)
and
det(λf ) = λn det(f ), tr(λf ) = λ tr(f ).

c Ali Ayad page 17

Chapter 1 Recall of some notions in linear algebra

1.5 Exercises
Exercise 1.1
We call projector of E every idempotent endomorphism of E, i.e., every endomorphism p of
E such that p ◦ p = p.

1) Let p be a projector of E.

(a) Show that the endomorphism q = idE − p is a projector of E, we call it the

projector of E associated to p.
(b) Show that p ◦ q = q ◦ p = 0.
(c) Show that ker q = Im p and Im q = ker p.

2) Let F and G be two supplementary subspaces of E, i.e., E = F ⊕ G. For every

x = u + v ∈ E (with u ∈ F and v ∈ G), put pF (x) = u.

(a) Show that pF is an endomorphism of E.

(b) Show that pF is a projector of E. We call it the projection of E on F parallel to
G.
(c) Deduce that ker pF = Im(idE − pF ) and Im pF = ker(idE − pF ).
(d) Show that ker pF = G and Im pF = F .

Exercise 1.2
We call symmetry of E every involutive endomorphism of E, i.e., every endomorphism s of
E such that s ◦ s = idE .

1) Show that the transposition of matrices is a vector symmetry of Mn (K).

2) Show that an endomorphism s of E is a vector symmetry of E if and only if s is bijective

and s−1 = s.

3) Prove that the composite s1 ◦ s2 of two vector symmetries of E is a vector symmetry

of E if and only if s1 ◦ s2 = s2 ◦ s1 .

4) Let F and G be two supplementary subspaces of E. For every x = u + v ∈ E (with

u ∈ F and v ∈ G), put sF (x) = u − v.

(a) Show that sF is an endomorphism of E.

(b) Show that sF is a vector symmetry of E. We call it the symmetry with respect to
F parallel to G.
(c) Show that sF = 2pF − idE , where pF is the projection of E on F parallel to G.
(d) Show that
F = ker(sF − idE ) = Im(sF + idE )
and
G = Im(sF − idE ) = ker(sF + idE ).

c Ali Ayad page 18

Chapter 1 Recall of some notions in linear algebra

Exercise 1.3
The following questions are independent.

1) Show that, for any λ ∈ K, λIn is the only matrix similar to λIn .

2) Does the similarity preserve the sum ? Justify your answer.

3) Let A, B ∈ Mn (K). Show that if A is invertible, then AB is similar to BA.

 
1 2 −1
4) Let A =  0 1 0  ∈ M3 (R). Find infinity many matrices in M3 (R) which
−1 3 1
are similar to A.

5) Let    
1 −1 −1 0
A= and B=
−2 0 −5 2
Find all the invertible matrices P ∈ M2 (R) such that B = P −1 AP .

Exercise 1.4
Let    
0 2 1 0
A= and B= .
−1 3 0 2

1) Show that A and B are similar.

2) Deduce An for all n ∈ N.

Exercise 1.5
Prove that the two matrices A and B are not similar in each of the following cases:

1)    
1 2 4 −2 −2 4
A= 0 π 8  and B =  2 −1 7 
−1 3 4 0 1 π

2)    
5 3 −6 4 2 1
A= 1 4 0  and B= 0 5 6 
0 0 2 0 −3 2

3)    
2 0 0 2 −2 1
A= 0 2 0  and B= 0 2 3 
0 0 2 0 0 2

c Ali Ayad page 19

Chapter 1 Recall of some notions in linear algebra

Exercise 1.6
Let A, B ∈ Mn (K) such that A is similar to B. Prove that:
1) λA is similar to λB for all λ ∈ K.
2) Ak is similar to B k for all k ∈ N.
3) tA is similar to t B.
4) If A is invertible, then B is invertible and A−1 is similar to B −1 .

Exercise 1.7
In each of the following cases, show that the two matrices A and B are similar and determine
an invertible matrix P ∈ M3 (R) such that B = P −1 AP .
1)   
1 1 0 1 1 1
A= 0 1 2  and B= 0 1 1 
0 0 1 0 0 1

2)    
0 1 1 1 1 0
A =  −1 1 1  and B= 0 1 1 
−1 1 2 0 0 1

3)    
1 2 3 4 1 1 0 0
 0 1 2 3   0 1 1 0 
A=
 0
 and B= 
0 1 2   0 0 1 1 
0 0 0 1 0 0 0 1

Exercise 1.8
Let   
0 1 0 0 0 1
A= 0 0 1  and B =  1 0 0 .
1 0 0 0 1 0
1) Show that A and B are similar and determine an invertible matrix P ∈ M3 (R) such
that B = P −1 AP .
2) Let    
1 2 3 1 3 2
A0 =  3 1 2  and B0 =  2 1 3  .
2 3 1 3 2 1

(a) Verify that B = A2 and A = B 2 , and that

A0 = 3A2 + 2A + I3 and B 0 = 3B 2 + 2B + I3 .

(b) Deduce that A0 and B 0 are similar.

c Ali Ayad page 20

Chapter 1 Recall of some notions in linear algebra

Exercise 1.9
Let A, B ∈ Mn (K) and λ ∈ K.

1) Show that if A is similar to B, then A − λIn is similar to B − λIn .

2) Show that if A − λIn is similar to B, then A is similar to B + λIn .

Exercise 1.10
Let    
1 2 3 3 1 2
A= 0 1 2  and B =  2 0 1 .
0 0 1 1 0 0

1) Verify that A and B have the same determinant, same trace and same rank.

2) Show that B − I3 is invertible.

3) Deduce that A and B are not similar.

c Ali Ayad page 21