📚 الرئيسية

مسابقة دكتوراه 2022Université Ferhat Abbas - Sétif 1 — الموضوع 04

مسابقة عامة · الرياضيات

Concours national d'accès au Doctorat 2022-2023 — Faculté des Sciences, Université Ferhat Abbas Sétif 1 — Épreuve d'Algèbre Linéaire (épreuve commune, toutes spécialités)

التمرين 1

Endomorphismes commutants et décomposition en somme directe

#algèbre linéaire#endomorphismes#noyau et image#somme directe

Soit EE un espace vectoriel et u,v,f,gu, v, f, g quatre endomorphismes de EE qui commutent deux à deux, et qui vérifient uf+vg=idEu \circ f + v \circ g = \mathrm{id}_E.

  1. Montrer que KerfKerg={0}\mathrm{Ker}\, f \cap \mathrm{Ker}\, g = \{0\} et E=Imf+ImgE = \mathrm{Im}\, f + \mathrm{Im}\, g.
  2. Montrer que Ker(fg)=KerfKerg\mathrm{Ker}(f\circ g) = \mathrm{Ker}\, f \oplus \mathrm{Ker}\, g et Im(fg)=ImfImg\mathrm{Im}(f\circ g) = \mathrm{Im}\, f \cap \mathrm{Im}\, g.
  3. On suppose de plus que fg=0f \circ g = 0. Montrer que E=KerfKerg=ImfImgE = \mathrm{Ker}\, f \oplus \mathrm{Ker}\, g = \mathrm{Im}\, f \oplus \mathrm{Im}\, g, Kerf=Img\mathrm{Ker}\, f = \mathrm{Im}\, g, Kerg=Imf\mathrm{Ker}\, g = \mathrm{Im}\, f.

Idée clé : l'identité x=u(f(x))+v(g(x))=f(u(x))+g(v(x))x=u(f(x))+v(g(x))=f(u(x))+g(v(x)) permet de tout démontrer par substitution directe, sans jamais utiliser la dimension finie — les résultats restent donc valables en dimension infinie.

الحل

On utilise systématiquement l'identité, valable pour tout xEx\in E (obtenue en appliquant uf+vg=iduf+vg=\mathrm{id} et les commutations uf=fuuf=fu, vg=gvvg=gv) : x=u(f(x))+v(g(x))=f(u(x))+g(v(x)).x = u(f(x)) + v(g(x)) = f(u(x)) + g(v(x)).

1) Si xKerfKergx\in \mathrm{Ker} f\cap \mathrm{Ker} g, l'identité donne x=u(0)+v(0)=0x = u(0)+v(0) = 0, donc KerfKerg={0}\mathrm{Ker} f\cap \mathrm{Ker} g=\{0\}. Pour tout xEx\in E, x=f(u(x))+g(v(x))Imf+Imgx = f(u(x)) + g(v(x)) \in \mathrm{Im} f + \mathrm{Im} g, donc E=Imf+ImgE=\mathrm{Im} f+\mathrm{Im} g.

2) Comme fg=gffg=gf: si xKergx\in\mathrm{Ker} g, (fg)(x)=f(0)=0(fg)(x)=f(0)=0; si xKerfx\in\mathrm{Ker} f, (fg)(x)=(gf)(x)=g(0)=0(fg)(x)=(gf)(x)=g(0)=0. Donc Kerf+KergKer(fg)\mathrm{Ker} f+\mathrm{Ker} g\subset \mathrm{Ker}(fg). Réciproquement soit xKer(fg)=Ker(gf)x\in\mathrm{Ker}(fg)=\mathrm{Ker}(gf), i.e. f(g(x))=g(f(x))=0f(g(x))=g(f(x))=0. Écrivons x=f(u(x))+g(v(x))x=f(u(x))+g(v(x)). On vérifie g(f(u(x)))=f(u(g(x)))=0g(f(u(x)))=f(u(g(x)))=0 donc f(u(x))Kergf(u(x))\in\mathrm{Ker} g, et f(g(v(x)))=g(v(f(x)))=0f(g(v(x)))=g(v(f(x)))=0 donc g(v(x))Kerfg(v(x))\in\mathrm{Ker} f. Ainsi xKerg+Kerfx\in \mathrm{Ker} g+\mathrm{Ker} f. D'où Ker(fg)=Kerf+Kerg\mathrm{Ker}(fg)=\mathrm{Ker} f+\mathrm{Ker} g, somme directe d'après 1). Pour les images: Im(fg)Imf\mathrm{Im}(fg)\subset \mathrm{Im} f et Im(fg)=Im(gf)Img\mathrm{Im}(fg)=\mathrm{Im}(gf)\subset\mathrm{Im} g, donc Im(fg)ImfImg\mathrm{Im}(fg)\subset \mathrm{Im} f\cap\mathrm{Im} g. Réciproquement si y=f(a)=g(b)ImfImgy=f(a)=g(b)\in\mathrm{Im} f\cap \mathrm{Im} g, l'identité donne y=f(u(y))+g(v(y))=f(g(u(b)))+g(f(v(a)))=(fg)(u(b)+v(a))Im(fg)y=f(u(y))+g(v(y)) = f(g(u(b))) + g(f(v(a))) = (fg)(u(b)+v(a)) \in \mathrm{Im}(fg). D'où l'égalité.

3) Si fg=0fg=0 alors gf=fg=0gf=fg=0, donc Ker(fg)=E\mathrm{Ker}(fg)=E, et par 2) E=KerfKergE=\mathrm{Ker} f\oplus \mathrm{Ker} g. De même Im(fg)={0}=ImfImg\mathrm{Im}(fg)=\{0\}=\mathrm{Im} f\cap \mathrm{Im} g, et comme E=Imf+ImgE=\mathrm{Im} f+\mathrm{Im} g, la somme est directe: E=ImfImgE=\mathrm{Im} f\oplus \mathrm{Im} g. Puisque fg=0fg=0, ImgKerf\mathrm{Im} g\subset \mathrm{Ker} f. Soit xKerfx\in\mathrm{Ker} f: x=f(u(x))+g(v(x))x=f(u(x))+g(v(x)) et f(u(x))=u(f(x))=0f(u(x))=u(f(x))=0, donc x=g(v(x))Imgx=g(v(x))\in\mathrm{Im} g. Ainsi Kerf=Img\mathrm{Ker} f=\mathrm{Im} g. Par symétrie, Kerg=Imf\mathrm{Ker} g=\mathrm{Im} f.

التمرين 2

Rang, valeurs propres et diagonalisation d'une matrice alternée

#algèbre linéaire#valeurs propres#diagonalisation#rang d'une matrice

Soient a,bRa,b\in\mathbb{R} tels que ab|a|\neq|b|. On considère la matrice carrée de taille 2n2n dont les lignes impaires valent (a,b,a,b,)(a,b,a,b,\dots) et les lignes paires (b,a,b,a,)(b,a,b,a,\dots).

  1. Calculer le rang de AA. En déduire que si n>1n>1, alors 00 est valeur propre de AA, et déterminer la dimension du sous-espace propre associé.
  2. Déterminer deux vecteurs propres associés à deux autres valeurs propres. En déduire que AA est diagonalisable.

Astuce : remarquer d'emblée qu'il n'y a que deux lignes distinctes réduit immédiatement le rang à 2\le 2. Les vecteurs (1,,1)(1,\dots,1) et (1,1,1,1,)(1,-1,1,-1,\dots) sont les candidats naturels pour une matrice à structure alternée.

الحل

AA n'a que deux lignes distinctes : r1=(a,b,a,b,)r_1=(a,b,a,b,\dots) (répétée nn fois) et r2=(b,a,b,a,)r_2=(b,a,b,a,\dots) (répétée nn fois), donc rg(A)2\mathrm{rg}(A)\le 2. Ces deux vecteurs sont proportionnels seulement si a=ba=b ou a=ba=-b, c'est-à-dire a=b|a|=|b|; comme ab|a|\ne|b|, r1,r2r_1,r_2 sont indépendants et rg(A)=2\mathrm{rg}(A)=2. Par le théorème du rang, dimKerA=2n2\dim\mathrm{Ker} A = 2n-2. Pour n>1n>1 cette dimension est strictement positive, donc 00 est valeur propre de AA, de sous-espace propre de dimension 2n22n-2.

2) Soit v1=(1,1,,1)v_1=(1,1,\dots,1). Chaque ligne contient nn fois aa et nn fois bb, donc Av1=n(a+b)v1Av_1=n(a+b)v_1: v1v_1 est vecteur propre pour λ1=n(a+b)\lambda_1=n(a+b). Soit v2=(1,1,1,1,)v_2=(1,-1,1,-1,\dots). Le produit scalaire d'une ligne impaire avec v2v_2 vaut n(ab)n(a-b), celui d'une ligne paire vaut n(ab)-n(a-b), ce qui correspond exactement à n(ab)v2n(a-b)\,v_2. Donc Av2=n(ab)v2Av_2=n(a-b)v_2: v2v_2 est vecteur propre pour λ2=n(ab)\lambda_2=n(a-b). Comme ab|a|\ne|b|, a+b0a+b\ne 0 et ab0a-b\ne0, donc λ1,λ20\lambda_1,\lambda_2\ne 0; v1,v2KerAv_1,v_2\notin\mathrm{Ker} A. La somme des dimensions des sous-espaces propres vaut (2n2)+1+1=2n=dimE(2n-2)+1+1=2n=\dim E, donc AA est diagonalisable.

التمرين 3

Matrices de dérivation et multiplication par X sur les polynômes

#algèbre linéaire#polynômes#matrice d'application linéaire#valeurs propres

Notons R[X]d\mathbb{R}[X]_d l'espace vectoriel des applications polynomiales de degré d\le d, muni de sa base canonique BdB_d.

  1. Soit f:R[X]4R[X]3f:\mathbb{R}[X]_4\to \mathbb{R}[X]_3, f(P)=Pf(P)=P'. a. Montrer que ff est linéaire. b. Donner la matrice de ff dans les bases B4B_4 et B3B_3.
  2. Soit g:R[X]3R[X]4g:\mathbb{R}[X]_3\to\mathbb{R}[X]_4 définie par g(P)=XPg(P)=XP. Donner la matrice de gg dans les bases B3B_3 et B4B_4.
  3. Donner la matrice de gfg\circ f dans la base B4B_4.
  4. Soit λR\lambda\in\mathbb{R}. Discuter suivant λ\lambda la dimension de Ker(gfλidR[X]4)\mathrm{Ker}(g\circ f-\lambda\,\mathrm{id}_{\mathbb{R}[X]_4}).

XkX^k est automatiquement vecteur propre de gfg\circ f pour la valeur propre kk, ce qui rend la question 4 immédiate sans nouveau calcul.

الحل

1a. La dérivation est linéaire : (P+Q)=P+Q(P+Q)'=P'+Q' et (μP)=μP(\mu P)'=\mu P'.

1b. Avec B4=(1,X,X2,X3,X4)B_4=(1,X,X^2,X^3,X^4) et B3=(1,X,X2,X3)B_3=(1,X,X^2,X^3) : f(1)=0, f(X)=1, f(X2)=2X, f(X3)=3X2, f(X4)=4X3f(1)=0,\ f(X)=1,\ f(X^2)=2X,\ f(X^3)=3X^2,\ f(X^4)=4X^3, donc Mat(f)=(01000002000003000004)\mathrm{Mat}(f)=\begin{pmatrix}0&1&0&0&0\\0&0&2&0&0\\0&0&0&3&0\\0&0&0&0&4\end{pmatrix}.

2. g(1)=X, g(X)=X2, g(X2)=X3, g(X3)=X4g(1)=X,\ g(X)=X^2,\ g(X^2)=X^3,\ g(X^3)=X^4, donc Mat(g)=(00001000010000100001)\mathrm{Mat}(g)=\begin{pmatrix}0&0&0&0\\1&0&0&0\\0&1&0&0\\0&0&1&0\\0&0&0&1\end{pmatrix}.

3. (gf)(P)=XP(g\circ f)(P)=X P'. Sur B4B_4: (gf)(Xk)=kXk(g\circ f)(X^k)=kX^k pour k=0,,4k=0,\dots,4, donc Mat(gf)=diag(0,1,2,3,4)\mathrm{Mat}(g\circ f)=\mathrm{diag}(0,1,2,3,4).

4. Mat(gfλid)=diag(λ,1λ,2λ,3λ,4λ)\mathrm{Mat}(g\circ f-\lambda\,\mathrm{id})=\mathrm{diag}(-\lambda,1-\lambda,2-\lambda,3-\lambda,4-\lambda), donc dimKer\dim\mathrm{Ker} est le nombre d'entiers k{0,1,2,3,4}k\in\{0,1,2,3,4\} tels que k=λk=\lambda : si λ{0,1,2,3,4}\lambda\notin\{0,1,2,3,4\}, dimKer=0\dim\mathrm{Ker}=0; si λ{0,1,2,3,4}\lambda\in\{0,1,2,3,4\}, dimKer=1\dim\mathrm{Ker}=1, engendré par XλX^\lambda.