3.1 Systèmes: formulation matricielle

Dans ce chapitre, nous allons reformuler ce qui a été dit à propos des systèmes en utilisant le langage vectoriel de l'algèbre linéaire. Ceci aura plusieurs avantages, et mènera en particulier à une compréhension plus profonde des divers aspects liés à la recherche des solutions d'un système linéaire.

On peut voir un système \(m\times n\) général, de la forme \[ (*) \left\{ \begin{array}{ccccccccc} a_{11}x_1&+&a_{12}x_2&+&\cdots&+&a_{1n}x_n&=&b_1\\ a_{21}x_1&+&a_{22}x_2&+&\cdots&+&a_{2n}x_n&=&b_2\\ \vdots&&&&&&&\vdots&\\ a_{m1}x_1&+&a_{m2}x_2&+&\cdots&+&a_{mn}x_n&=&b_m \end{array} \right. \,, \] comme une égalité entre deux vecteurs de \(\mathbb{R}^m\): \[ \begin{pmatrix} a_{11}x_1&+&a_{12}x_2&+&\cdots&+&a_{1n}x_n\\ a_{21}x_1&+&a_{22}x_2&+&\cdots&+&a_{2n}x_n\\ &&\vdots&&&&\\ a_{m1}x_1&+&a_{m2}x_2&+&\cdots&+&a_{mn}x_n \end{pmatrix} = \begin{pmatrix} b_1\\ b_2\\ \vdots\\ b_m \end{pmatrix}\,. \] Or on peut récrire cette dernière comme suit: \[ x_1 \begin{pmatrix} a_{11}\\ a_{21}\\ \vdots\\ a_{m1} \end{pmatrix} + x_2 \begin{pmatrix} a_{12}\\ a_{22}\\ \vdots\\ a_{m2} \end{pmatrix} +\cdots+ x_n \begin{pmatrix} a_{1n}\\ a_{2n}\\ \vdots\\ a_{mn} \end{pmatrix} = \begin{pmatrix} b_1\\ b_2\\ \vdots\\ b_m \end{pmatrix}\,, \] dans laquelle on reconnaît maintenant, dans le membre de gauche, une combinaison linéaire des colonnes de la matrice associée au système, qui est donnée, rappelons-le, par \[ A= \begin{pmatrix} a_{11}&a_{12}&\cdots&a_{1n}\\ a_{21}&a_{22}&\cdots&a_{2n}\\ \vdots&\vdots&&\vdots\\ a_{m1}&a_{m2}&\cdots&a_{mn} \end{pmatrix}\,. \] Récrivons la même chose de façon plus compacte, en commençant par définir le vecteur associé au second membre, \[ \boldsymbol{b}:= \begin{pmatrix} b_1\\ b_2\\ \vdots\\ b_m \end{pmatrix}\,, \] et récrivons la matrice du système comme une famille de colonnes, \[ A= \begin{bmatrix} \,\boldsymbol{a}_1&\boldsymbol{a}_2&\cdots&\boldsymbol{a}_n\, \end{bmatrix}\,, \] où, la \(k\)-ème colonne est le vecteur de \(\mathbb{R}^m\) donné par \[ \boldsymbol{a}_k := \begin{pmatrix} a_{1k}\\ a_{2k}\\ \vdots\\ a_{mk} \end{pmatrix}\,. \] Donc la recherche de solutions \((x_1,x_2,\dots,x_n)\) au système \((*)\) est équivalente à demander si le membre de droite \(\boldsymbol{b}\) appartient à la partie de \(\mathbb{R}^m\) engendrée par les colonnes de \(A\), c'est-à-dire si on peut écrire \(\boldsymbol{b}\) comme une combinaison linéaire des colonnes de \(A\): \[ x_1\boldsymbol{a}_1+x_2\boldsymbol{a}_2+\cdots +x_n\boldsymbol{a}_n=\boldsymbol{b}\,. \] Dans cette formulation, les inconnues \(x_1,\dots,x_n\) jouent le rôle de coefficients de la combinaison linéaire.

Une dernière définition permettra de faire encore un pas dans la description du système \((*)\).

Soit \(A\) une matrice \(m\times n\), dont la \(k\)-ème colonne est notée \(\boldsymbol{a}_k\in \mathbb{R}^m\), \[A= \begin{bmatrix} \,\boldsymbol{a}_1&\boldsymbol{a}_2&\cdots&\boldsymbol{a}_n\, \end{bmatrix}\,, \] et soit \(\boldsymbol{x}\) un vecteur de \(\mathbb{R}^n\), \[ \boldsymbol{x}= \begin{pmatrix} x_1\\ \vdots\\ x_n \end{pmatrix}\,. \] Le produit de \(A\) par \(\boldsymbol{x}\) est le vecteur \(A\boldsymbol{x}\in \mathbb{R}^m\) défini par la combinaison linéaire \[ A\boldsymbol{x}:= x_1\boldsymbol{a}_1+\dots+x_n\boldsymbol{a}_n\,. \]

Le produit d'une matrice \(A\) (\(m\times n\)) par un vecteur \(\boldsymbol{x}\in \mathbb{R}^n\) crée donc un vecteur \(A\boldsymbol{x}\in \mathbb{R}^m\). Cette transformation est l'exemple standard de ce que l'on appellera plus tard une transformation linéaire, puisqu'elle satisfait aux deux propriétés suivantes:

Lemme: Soit \(A\) une matrice \(m\times n\). Alors pour tous \(\boldsymbol{x},\boldsymbol{y}\in \mathbb{R}^n\) et pour tout scalaire \(\lambda\in \mathbb{R}\),

  1. \(A(\boldsymbol{x}+\boldsymbol{y})=A\boldsymbol{x}+A\boldsymbol{y}\,\),
  2. \(A(\lambda \boldsymbol{x})=\lambda\, A\boldsymbol{x}\).
Ensemble, ces deux propriétés constituent la linéarité.

Notons la matrice \(A=[\boldsymbol{a}_1\cdots\boldsymbol{a}_n]\), et les vecteurs \[ \boldsymbol{x}= \begin{pmatrix} x_1\\ x_2\\ \vdots\\ x_n \end{pmatrix} \,, \qquad \boldsymbol{y}= \begin{pmatrix} y_1\\ y_2\\ \vdots\\ y_n \end{pmatrix}\,. \] Pour la première propriété, \[\begin{aligned} A(\boldsymbol{x}+\boldsymbol{y})&= \begin{bmatrix} \,\boldsymbol{a}_1&\boldsymbol{a}_2&\cdots&\boldsymbol{a}_n\, \end{bmatrix} \begin{pmatrix} x_1+y_1\\ x_2+y_2\\ \vdots\\ x_n+y_n \end{pmatrix}\\ &=(x_1+y_1)\boldsymbol{a}_1+\cdots+(x_n+y_n)\boldsymbol{a}_n\\ &=\bigl(x_1\boldsymbol{a}_1+\cdots+x_n\boldsymbol{a}_n\bigr)+ \bigl(y_1\boldsymbol{a}_1+\cdots+y_n\boldsymbol{a}_n\bigr)\\ &=A\boldsymbol{x}+A\boldsymbol{y}\,. \end{aligned}\] Pour la deuxième, \[\begin{aligned} A(\lambda\boldsymbol{x})&= \begin{bmatrix} \,\boldsymbol{a}_1&\boldsymbol{a}_2&\cdots&\boldsymbol{a}_n\, \end{bmatrix} \begin{pmatrix} \lambda x_1\\ \lambda x_2\\ \vdots\\ \lambda x_n \end{pmatrix}\\ &=(\lambda x_1)\boldsymbol{a}_1+\cdots+(\lambda x_n)\boldsymbol{a}_n\\ &=\lambda (x_1\boldsymbol{a}_1)+\cdots+\lambda (x_n\boldsymbol{a}_n)\\ &=\lambda \bigl(x_1\boldsymbol{a}_1+\cdots+x_n\boldsymbol{a}_n\bigr)\\ &=\lambda\, A\boldsymbol{x}\,. \end{aligned}\]

Avec les notations introduites, on peut maintenant écrire le système \((*)\) sous une forme purement vectorielle: \[ (*):\quad A\boldsymbol{x}=\boldsymbol{b} \] L'existence d'une solution \(\boldsymbol{x}\), pour ce système, signifie donc qu'il existe au moins une façon d'écrire le membre de droite \(\boldsymbol{b}\) comme combinaison linéaire des colonnes de \(A\).

Quiz 3.1-1 : Vrai ou faux?
  1. Si \(A=[\boldsymbol{a}_1\cdots\boldsymbol{a}_n]\) est une matrice \(m\times n\), alors \(A\boldsymbol{x}\in\mathrm{Vect}\{\boldsymbol{a}_1,\dots,\boldsymbol{a}_n\}\), quel que soit \(\boldsymbol{x}\in\mathbb{R}^n\).
  2. Si \(A\boldsymbol{x}\in\mathrm{Vect}\{\boldsymbol{v}_1,\dots,\boldsymbol{v}_n\}\) pour tout \(\boldsymbol{x}\in\mathbb{R}^n\), alors \(A=[\boldsymbol{v}_1\cdots\boldsymbol{v}_n]\).
  3. Il existe une matrice \(A\), \(m\times n\), telle que \(A\boldsymbol{0}\neq \boldsymbol{0}\).