Chapitre 3 : Applications linéaires

Noyau, image, théorème du rang et changement de base

Auteur·rice

Ibrahim BENKHERFELLAH

Modifié

27 mai 2026

Applications linéaires

Définitions et exemples

Définition 3.1 - Application linéaire

Soient \(E\) et \(F\) deux espaces vectoriels réels. Une application \(f : E \to F\) est dite linéaire (ou homomorphisme d’espaces vectoriels) si :

  1. \(\forall x, y \in E, \quad f(x+y) = f(x) + f(y)\) (additivité)
  2. \(\forall \lambda \in \mathbb{R}, \forall x \in E, \quad f(\lambda x) = \lambda f(x)\) (homogénéité)

Remarque

Les deux conditions peuvent être regroupées en une seule : \[ \forall \lambda, \mu \in \mathbb{R}, \forall x, y \in E, \quad f(\lambda x + \mu y) = \lambda f(x) + \mu f(y) \]

C’est-à-dire que \(f\) respecte les combinaisons linéaires.

Proposition 3.2

Si \(f : E \to F\) est linéaire, alors :

  1. \(f(0_E) = 0_F\)
  2. \(\forall x \in E, \quad f(-x) = -f(x)\)

Démonstration

  1. \(f(0_E) = f(0 \cdot 0_E) = 0 \cdot f(0_E) = 0_F\)
  2. \(f(-x) = f((-1) \cdot x) = (-1) \cdot f(x) = -f(x)\)

Exemple 3.1.1 - L’application identité

L’application \(\text{id}_E : E \to E\), \(x \mapsto x\) est linéaire.

Exemple 3.1.2 - L’application nulle

L’application \(f : E \to F\), \(x \mapsto 0_F\) est linéaire.

Exemple 3.1.3 - Homothétie

Pour \(\lambda \in \mathbb{R}\) fixé, l’application \(h_\lambda : E \to E\), \(x \mapsto \lambda x\) est linéaire.

Exemple 3.1.4 - Dérivation

L’application \(D : \mathbb{R}[X] \to \mathbb{R}[X]\), \(P \mapsto P'\) est linéaire car : \[ (P+Q)' = P' + Q' \quad \text{et} \quad (\lambda P)' = \lambda P' \]

Exemple 3.1.5 - Intégration

L’application \(I : C([0,1], \mathbb{R}) \to \mathbb{R}\), \(f \mapsto \int_0^1 f(t) \, dt\) est linéaire car : \[ \int_0^1 (f+g) = \int_0^1 f + \int_0^1 g \quad \text{et} \quad \int_0^1 \lambda f = \lambda \int_0^1 f \]

Exemple 3.1.6 - Produit matriciel

Pour \(A \in \mathcal{M}_{m,n}(\mathbb{R})\) fixée, l’application \(f : \mathcal{M}_{n,p}(\mathbb{R}) \to \mathcal{M}_{m,p}(\mathbb{R})\), \(X \mapsto AX\) est linéaire.

Exercice

L’application \(f : \mathbb{R}^2 \to \mathbb{R}^2\), \((x,y) \mapsto (x+y, xy)\) est-elle linéaire ?

Exercice

Montrer que l’application \(f : \mathbb{R}^3 \to \mathbb{R}^2\), \((x,y,z) \mapsto (x+2y, y-z)\) est linéaire.

Définition 3.3 - Espace des applications linéaires

L’ensemble des applications linéaires de \(E\) dans \(F\) est noté \(\mathcal{L}(E, F)\) ou \(\text{Hom}(E, F)\).

Quand \(E = F\), on note \(\mathcal{L}(E) = \mathcal{L}(E, E)\) l’ensemble des endomorphismes de \(E\).

Proposition 3.4

L’ensemble \(\mathcal{L}(E, F)\) muni de l’addition et de la multiplication par un scalaire est un espace vectoriel réel.

L’addition est définie par : \((f+g)(x) = f(x) + g(x)\)

La multiplication par un scalaire est définie par : \((\lambda f)(x) = \lambda f(x)\)

Noyau et image

Définition 3.5 - Noyau et image

Soit \(f : E \to F\) une application linéaire.

Le noyau de \(f\), noté \(\text{Ker}(f)\), est l’ensemble : \[ \text{Ker}(f) = \{x \in E \mid f(x) = 0_F\} = f^{-1}(\{0_F\}) \]

L’image de \(f\), notée \(\text{Im}(f)\), est l’ensemble : \[ \text{Im}(f) = \{y \in F \mid \exists x \in E, f(x) = y\} = f(E) \]

Proposition 3.6

Soit \(f : E \to F\) une application linéaire. Alors :

  1. \(\text{Ker}(f)\) est un sous-espace vectoriel de \(E\)
  2. \(\text{Im}(f)\) est un sous-espace vectoriel de \(F\)

Démonstration

Pour le noyau :

  • \(f(0_E) = 0_F\) donc \(0_E \in \text{Ker}(f)\)
  • Si \(x, y \in \text{Ker}(f)\) et \(\lambda, \mu \in \mathbb{R}\), alors : \[ f(\lambda x + \mu y) = \lambda f(x) + \mu f(y) = \lambda \cdot 0_F + \mu \cdot 0_F = 0_F \] donc \(\lambda x + \mu y \in \text{Ker}(f)\)

Pour l’image :

  • \(0_F = f(0_E) \in \text{Im}(f)\)
  • Si \(y_1, y_2 \in \text{Im}(f)\) et \(\lambda, \mu \in \mathbb{R}\), il existe \(x_1, x_2 \in E\) tels que \(f(x_1) = y_1\) et \(f(x_2) = y_2\). Alors : \[ \lambda y_1 + \mu y_2 = \lambda f(x_1) + \mu f(x_2) = f(\lambda x_1 + \mu x_2) \in \text{Im}(f) \]

Exemple 3.2.1

Soit \(f : \mathbb{R}^3 \to \mathbb{R}^2\), \((x,y,z) \mapsto (x+y, y+z)\).

Noyau : \((x,y,z) \in \text{Ker}(f)\) si et seulement si \(x+y = 0\) et \(y+z = 0\).

Donc \(y = -x\) et \(z = -y = x\).

Ainsi \(\text{Ker}(f) = \{(x, -x, x) \mid x \in \mathbb{R}\} = \text{Vect}((1, -1, 1))\) et \(\dim(\text{Ker}(f)) = 1\).

Image : Soit \((a, b) \in \mathbb{R}^2\). Cherchons \((x,y,z)\) tel que \(f(x,y,z) = (a,b)\).

On doit avoir \(x+y = a\) et \(y+z = b\).

Prenons par exemple \(y = 0\) : alors \(x = a\) et \(z = b\), donc \((a,b) = f(a, 0, b)\).

Ainsi tout vecteur de \(\mathbb{R}^2\) est dans l’image, donc \(\text{Im}(f) = \mathbb{R}^2\) et \(\dim(\text{Im}(f)) = 2\).

Proposition 3.7

Soit \(f : E \to F\) une application linéaire. Alors :

  1. \(f\) est injective si et seulement si \(\text{Ker}(f) = \{0_E\}\)
  2. \(f\) est surjective si et seulement si \(\text{Im}(f) = F\)

Démonstration

Pour l’injectivité :

\((\Rightarrow)\) Si \(f\) est injective et \(x \in \text{Ker}(f)\), alors \(f(x) = 0_F = f(0_E)\), donc \(x = 0_E\).

\((\Leftarrow)\) Si \(\text{Ker}(f) = \{0_E\}\) et \(f(x) = f(y)\), alors \(f(x-y) = f(x) - f(y) = 0_F\), donc \(x-y \in \text{Ker}(f) = \{0_E\}\), d’où \(x = y\).

Définition 3.8 - Isomorphisme

Si \(f : E \to F\) est linéaire, injective et surjective, on dit que \(f\) est un isomorphisme.

On dit alors que \(E\) et \(F\) sont isomorphes, noté \(E \cong F\).

Proposition 3.9

Deux espaces vectoriels de dimension finie sont isomorphes si et seulement s’ils ont la même dimension.

En particulier, tout espace vectoriel réel de dimension \(n\) est isomorphe à \(\mathbb{R}^n\).

Exercice

Déterminer le noyau et l’image de l’application linéaire \(f : \mathbb{R}^3 \to \mathbb{R}^2\), \((x,y,z) \mapsto (x-y, x+y+z)\).

Théorème du rang

Théorème 3.10 - Théorème du rang

Soit \(f : E \to F\) une application linéaire avec \(E\) de dimension finie. Alors : \[ \dim(E) = \dim(\text{Ker}(f)) + \dim(\text{Im}(f)) \]

La dimension de \(\text{Im}(f)\) est appelée le rang de \(f\), noté \(\text{rg}(f)\).

Démonstration

Posons \(n = \dim(E)\) et \(k = \dim(\text{Ker}(f))\). Soit \(\{e_1, \ldots, e_k\}\) une base de \(\text{Ker}(f)\). Par le théorème de la base incomplète, on la complète en une base \(\{e_1, \ldots, e_k, e_{k+1}, \ldots, e_n\}\) de \(E\).

Montrons que \(\{f(e_{k+1}), \ldots, f(e_n)\}\) est une base de \(\text{Im}(f)\).

Famille génératrice de \(\text{Im}(f)\) : Soit \(y \in \text{Im}(f)\). Il existe \(x \in E\) tel que \(y = f(x)\). On écrit \(x = \lambda_1 e_1 + \cdots + \lambda_k e_k + \lambda_{k+1} e_{k+1} + \cdots + \lambda_n e_n\). Par linéarité : \[ y = f(x) = \lambda_1 \underbrace{f(e_1)}_{= 0_F} + \cdots + \lambda_k \underbrace{f(e_k)}_{= 0_F} + \lambda_{k+1} f(e_{k+1}) + \cdots + \lambda_n f(e_n) \]

Donc \(y = \lambda_{k+1} f(e_{k+1}) + \cdots + \lambda_n f(e_n)\), ce qui montre que la famille est génératrice.

Famille libre : Supposons que \(\mu_{k+1} f(e_{k+1}) + \cdots + \mu_n f(e_n) = 0_F\). Par linéarité, cela donne \(f(\mu_{k+1} e_{k+1} + \cdots + \mu_n e_n) = 0_F\), donc le vecteur \(v = \mu_{k+1} e_{k+1} + \cdots + \mu_n e_n\) appartient à \(\text{Ker}(f)\).

Comme \(v \in \text{Ker}(f)\), on peut l’écrire \(v = \alpha_1 e_1 + \cdots + \alpha_k e_k\), d’où : \[ \mu_{k+1} e_{k+1} + \cdots + \mu_n e_n - \alpha_1 e_1 - \cdots - \alpha_k e_k = 0_E \]

Par liberté de la base \(\{e_1, \ldots, e_n\}\), tous les coefficients sont nuls, en particulier \(\mu_{k+1} = \cdots = \mu_n = 0\).

Au final, \(\{f(e_{k+1}), \ldots, f(e_n)\}\) est une base de \(\text{Im}(f)\), de cardinal \(n - k\). D’où : \[ \dim(E) = n = k + (n-k) = \dim(\text{Ker}(f)) + \dim(\text{Im}(f)) \]

Corollaire 3.11

Soit \(f : E \to F\) une application linéaire avec \(\dim(E) = \dim(F) = n\). Alors les assertions suivantes sont équivalentes :

  1. \(f\) est injective
  2. \(f\) est surjective
  3. \(f\) est bijective (isomorphisme)
  4. \(\text{Ker}(f) = \{0_E\}\)
  5. \(\text{Im}(f) = F\)
  6. \(\text{rg}(f) = n\)

Exemple 3.3.1

Reprenons \(f : \mathbb{R}^3 \to \mathbb{R}^2\), \((x,y,z) \mapsto (x+y, y+z)\) de l’exemple précédent.

On avait : \(\dim(\text{Ker}(f)) = 1\) et \(\dim(\text{Im}(f)) = 2\).

Vérifions le théorème du rang : \(\dim(\mathbb{R}^3) = 3 = 1 + 2 = \dim(\text{Ker}(f)) + \dim(\text{Im}(f))\).

Exercice

Soit \(f : \mathbb{R}^4 \to \mathbb{R}^3\) une application linéaire de rang 2. Quelle est la dimension de \(\text{Ker}(f)\) ?

Proposition 3.12

Soit \(f : E \to F\) et \(g : F \to G\) deux applications linéaires. Alors :

  1. \(g \circ f : E \to G\) est linéaire
  2. \(\text{Ker}(f) \subset \text{Ker}(g \circ f)\)
  3. \(\text{Im}(g \circ f) \subset \text{Im}(g)\)
  4. \(\text{rg}(g \circ f) \leq \min(\text{rg}(f), \text{rg}(g))\)

Démonstration

Propriété (1) : Soient \(x, y \in E\) et \(\lambda, \mu \in \mathbb{R}\). Alors : \[ (g \circ f)(\lambda x + \mu y) = g(f(\lambda x + \mu y)) = g(\lambda f(x) + \mu f(y)) = \lambda g(f(x)) + \mu g(f(y)) \]

Propriété (2) : Si \(x \in \text{Ker}(f)\), alors \(f(x) = 0_F\), donc \((g \circ f)(x) = g(0_F) = 0_G\), d’où \(x \in \text{Ker}(g \circ f)\).

Propriété (3) : Si \(z \in \text{Im}(g \circ f)\), il existe \(x \in E\) tel que \(z = g(f(x))\). En posant \(y = f(x) \in F\), on a \(z = g(y) \in \text{Im}(g)\).

Propriété (4) : De (3), \(\text{Im}(g \circ f) \subset \text{Im}(g)\), donc \(\text{rg}(g \circ f) \leq \text{rg}(g)\). Par ailleurs, \(g\) restreinte à \(\text{Im}(f)\) est linéaire, et \(\text{Im}(g \circ f) = g(\text{Im}(f))\), donc par le théorème du rang appliqué à cette restriction, \(\dim(\text{Im}(g \circ f)) \leq \dim(\text{Im}(f)) = \text{rg}(f)\).

Exemple : la dérivation dans \(\mathbb{R}[X]\)

Considérons l’application de dérivation \(D : \mathbb{R}_n[X] \to \mathbb{R}_{n-1}[X]\), \(P \mapsto P'\).

Proposition 3.13

L’application de dérivation \(D : \mathbb{R}_n[X] \to \mathbb{R}_{n-1}[X]\) est une application linéaire surjective.

On a :

  • \(\text{Ker}(D) = \mathbb{R}_0[X]\) (les polynômes constants)
  • \(\text{Im}(D) = \mathbb{R}_{n-1}[X]\)
  • \(\dim(\text{Ker}(D)) = 1\) et \(\dim(\text{Im}(D)) = n\)

Vérification du théorème du rang : \(\dim(\mathbb{R}_n[X]) = n+1 = 1 + n = \dim(\text{Ker}(D)) + \dim(\text{Im}(D))\).

Exercice

Considérer l’application \(I : \mathbb{R}_n[X] \to \mathbb{R}_{n+1}[X]\) qui à \(P\) associe sa primitive s’annulant en 0, c’est-à-dire : \[ I(P)(X) = \int_0^X P(t) \, dt \]

Montrer que \(I\) est linéaire injective et calculer son noyau et son image.

Matrice d’une application linéaire

Cas général

Définition 3.14 - Matrice d’une application linéaire

Soient \(E\) et \(F\) deux espaces vectoriels de dimension finie, \(\mathcal{B}_E = \{e_1, \ldots, e_n\}\) une base de \(E\) et \(\mathcal{B}_F = \{f_1, \ldots, f_m\}\) une base de \(F\).

Soit \(u : E \to F\) une application linéaire.

La matrice de \(u\) dans les bases \(\mathcal{B}_E\) et \(\mathcal{B}_F\), notée \(\text{Mat}_{\mathcal{B}_E, \mathcal{B}_F}(u)\) ou \([u]_{\mathcal{B}_E}^{\mathcal{B}_F}\), est la matrice \(A \in \mathcal{M}_{m,n}(\mathbb{R})\) dont la \(j\)-ème colonne contient les coordonnées de \(u(e_j)\) dans la base \(\mathcal{B}_F\).

Plus précisément, si : \[ u(e_j) = a_{1j} f_1 + a_{2j} f_2 + \cdots + a_{mj} f_m \]

alors : \[ A = \begin{pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & \vdots & \ddots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \end{pmatrix} \]

Proposition 3.15

Avec les notations ci-dessus, si \(x = x_1 e_1 + \cdots + x_n e_n \in E\) et \(y = u(x)\), alors les coordonnées de \(y\) dans \(\mathcal{B}_F\) sont données par : \[ \begin{pmatrix} y_1 \\ \vdots \\ y_m \end{pmatrix} = A \begin{pmatrix} x_1 \\ \vdots \\ x_n \end{pmatrix} \]

\(A = \text{Mat}_{\mathcal{B}_E, \mathcal{B}_F}(u)\).

Exemple 3.5.1

Soit \(u : \mathbb{R}^2 \to \mathbb{R}^3\), \((x,y) \mapsto (x+y, x-y, 2x)\).

Calculons la matrice de \(u\) dans les bases canoniques \(\mathcal{B}_E = \{e_1, e_2\}\) de \(\mathbb{R}^2\) et \(\mathcal{B}_F = \{f_1, f_2, f_3\}\) de \(\mathbb{R}^3\).

On a : \[ \begin{aligned} u(e_1) = u(1,0) &= (1, 1, 2) = 1 \cdot f_1 + 1 \cdot f_2 + 2 \cdot f_3 \\ u(e_2) = u(0,1) &= (1, -1, 0) = 1 \cdot f_1 + (-1) \cdot f_2 + 0 \cdot f_3 \end{aligned} \]

Donc : \[ \text{Mat}_{\mathcal{B}_E, \mathcal{B}_F}(u) = \begin{pmatrix} 1 & 1 \\ 1 & -1 \\ 2 & 0 \end{pmatrix} \]

Exercice

Soit \(u : \mathbb{R}_2[X] \to \mathbb{R}_1[X]\), \(P \mapsto P'\) la dérivation.

Calculer la matrice de \(u\) dans les bases canoniques \(\{1, X, X^2\}\) et \(\{1, X\}\).

Propriétés

Proposition 3.16

L’application \(\text{Mat}_{\mathcal{B}_E, \mathcal{B}_F} : \mathcal{L}(E, F) \to \mathcal{M}_{m,n}(\mathbb{R})\) est un isomorphisme d’espaces vectoriels.

En particulier, \(\dim(\mathcal{L}(E, F)) = mn\) si \(\dim(E) = n\) et \(\dim(F) = m\).

Proposition 3.17

Soient \(u : E \to F\) et \(v : F \to G\) deux applications linéaires. Alors : \[ \text{Mat}_{\mathcal{B}_E, \mathcal{B}_G}(v \circ u) = \text{Mat}_{\mathcal{B}_F, \mathcal{B}_G}(v) \cdot \text{Mat}_{\mathcal{B}_E, \mathcal{B}_F}(u) \]

La matrice de la composée est le produit des matrices.

Proposition 3.18

Soit \(u : E \to F\) une application linéaire. Alors : \[ \text{rg}(u) = \text{rg}(\text{Mat}_{\mathcal{B}_E, \mathcal{B}_F}(u)) \]

Le rang de \(u\) est égal au rang de sa matrice (dans n’importe quelle base).

Changement de base

Matrice de passage

Définition 3.19 - Matrice de passage

Soient \(\mathcal{B} = \{e_1, \ldots, e_n\}\) et \(\mathcal{B}' = \{e'_1, \ldots, e'_n\}\) deux bases d’un espace vectoriel \(E\).

La matrice de passage de \(\mathcal{B}\) à \(\mathcal{B}'\), notée \(P_{\mathcal{B} \to \mathcal{B}'}\), est la matrice dont la \(j\)-ème colonne contient les coordonnées de \(e'_j\) dans la base \(\mathcal{B}\).

C’est aussi la matrice de l’identité \(\text{id}_E\) dans les bases \(\mathcal{B}'\) (au départ) et \(\mathcal{B}\) (à l’arrivée).

Proposition 3.20

La matrice de passage \(P = P_{\mathcal{B} \to \mathcal{B}'}\) est inversible et : \[ P^{-1} = P_{\mathcal{B}' \to \mathcal{B}} \]

Proposition 3.21 - Formule de changement de coordonnées

Si \(x \in E\) a pour coordonnées \((x_1, \ldots, x_n)\) dans \(\mathcal{B}\) et \((x'_1, \ldots, x'_n)\) dans \(\mathcal{B}'\), alors : \[ \begin{pmatrix} x_1 \\ \vdots \\ x_n \end{pmatrix} = P_{\mathcal{B} \to \mathcal{B}'} \begin{pmatrix} x'_1 \\ \vdots \\ x'_n \end{pmatrix} \]

ou de manière équivalente : \[ \begin{pmatrix} x'_1 \\ \vdots \\ x'_n \end{pmatrix} = P_{\mathcal{B}' \to \mathcal{B}} \begin{pmatrix} x_1 \\ \vdots \\ x_n \end{pmatrix} \]

Exemple 3.6.1

Dans \(\mathbb{R}^2\), soit \(\mathcal{B} = \{e_1, e_2\}\) la base canonique et \(\mathcal{B}' = \{e'_1, e'_2\}\) avec \(e'_1 = (1,1)\) et \(e'_2 = (1,-1)\).

On a : \[ \begin{aligned} e'_1 &= 1 \cdot e_1 + 1 \cdot e_2 \\ e'_2 &= 1 \cdot e_1 + (-1) \cdot e_2 \end{aligned} \]

Donc : \[ P_{\mathcal{B} \to \mathcal{B}'} = \begin{pmatrix} 1 & 1 \\ 1 & -1 \end{pmatrix} \]

Pour trouver \(P_{\mathcal{B}' \to \mathcal{B}}\), on inverse : \[ P_{\mathcal{B}' \to \mathcal{B}} = \begin{pmatrix} 1 & 1 \\ 1 & -1 \end{pmatrix}^{-1} = \frac{1}{-2} \begin{pmatrix} -1 & -1 \\ -1 & 1 \end{pmatrix} = \begin{pmatrix} 1/2 & 1/2 \\ 1/2 & -1/2 \end{pmatrix} \]

Formule de changement de base pour les endomorphismes

Proposition 3.22

Soit \(u : E \to E\) un endomorphisme et soient \(\mathcal{B}\) et \(\mathcal{B}'\) deux bases de \(E\).

Si \(A = \text{Mat}_{\mathcal{B}}(u)\) et \(A' = \text{Mat}_{\mathcal{B}'}(u)\), alors : \[ A' = P^{-1} A P \]

\(P = P_{\mathcal{B} \to \mathcal{B}'}\) est la matrice de passage.

Démonstration

L’endomorphisme \(u\) peut s’écrire comme la composée \(\text{id}_E \circ u \circ \text{id}_E\). En passant aux matrices, on utilise la propriété “la matrice de la composée est le produit des matrices”. Plus précisément :

La matrice de \(u\) dans la base \(\mathcal{B}'\) revient à exprimer \(u\) en “passant” par \(\mathcal{B}\) : \[ \text{Mat}_{\mathcal{B}'}(u) = \text{Mat}_{\mathcal{B}, \mathcal{B}'}(\text{id}) \cdot \text{Mat}_{\mathcal{B}}(u) \cdot \text{Mat}_{\mathcal{B}', \mathcal{B}}(\text{id}) \]

Or \(\text{Mat}_{\mathcal{B}', \mathcal{B}}(\text{id}) = P\) (la matrice de passage de \(\mathcal{B}\) à \(\mathcal{B}'\)) et \(\text{Mat}_{\mathcal{B}, \mathcal{B}'}(\text{id}) = P^{-1}\).

D’où \(A' = P^{-1} A P\).

Définition 3.23 - Matrices semblables

Deux matrices \(A\) et \(B\) de \(\mathcal{M}_n(\mathbb{R})\) sont dites semblables s’il existe une matrice inversible \(P\) telle que : \[ B = P^{-1} A P \]

Remarque

La similitude est une relation d’équivalence sur \(\mathcal{M}_n(\mathbb{R})\).

Deux matrices semblables représentent le même endomorphisme dans des bases différentes.

Proposition 3.24

Deux matrices semblables ont le même rang, la même trace et le même déterminant.

Démonstration

Supposons \(B = P^{-1}AP\) avec \(P\) inversible.

Rang : On a \(\text{rg}(B) = \text{rg}(P^{-1}AP) = \text{rg}(A)\) car la multiplication par des matrices inversibles ne change pas le rang (chapitre 2).

Trace : On utilise la propriété \(\text{tr}(MN) = \text{tr}(NM)\) : \[ \text{tr}(B) = \text{tr}(P^{-1}AP) = \text{tr}(APP^{-1}) = \text{tr}(A) \]

Déterminant : On admet pour l’instant que \(\det(MN) = \det(M)\det(N)\) (ce sera démontré au chapitre 4). Alors : \[ \det(B) = \det(P^{-1}AP) = \det(P^{-1})\det(A)\det(P) = \frac{1}{\det(P)}\det(A)\det(P) = \det(A) \]

Remarque

Cet exercice illustre un point subtil : trace et déterminant égaux ne suffisent pas à garantir la similitude. Ce sont des conditions nécessaires mais pas suffisantes. Pour déterminer si deux matrices sont semblables, il faut chercher explicitement la matrice \(P\) ou utiliser des invariants plus fins (comme les valeurs propres et la structure de Jordan, qu’on verra au chapitre 4).

Exercice

Montrer que les matrices \(A = \begin{pmatrix} 2 & 1 \\ 0 & 2 \end{pmatrix}\) et \(B = \begin{pmatrix} 2 & 0 \\ 1 & 2 \end{pmatrix}\) sont semblables. Trouver une matrice de passage \(P\) telle que \(B = P^{-1}AP\).

Exercice

Montrer que les matrices \(A = \begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix}\) et \(B = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix} = I_2\) ne sont pas semblables, bien qu’elles aient la même trace et le même déterminant.