MAA 201 2022-2023

Reduction of endomorphisms

Exercises - Week VII - Minimal Polynomial and Eigenvalues

In the following, E will denote a finite dimensional vector space over a field K.
We let L(E) be the set of endomorphisms of E.

Exercise 1.
Let f be the endomorphism of R[X] given by f (P ) = XP (0) + 2P 0 . Compute the
eigenvalues end eigenvectors of f .

Solution: If λ is an eigenvalue of f and P an associated eigenvector. Then XP (0) + 2P 0 = λP .

If λ 6= 0 the degree of P is less than or equal to one. Writing P as P (X) = aX + b with √
(a, b) ∈ R2 \ {(0, 0)} we get bX + 2a = λ(aX + b), therefore b = λa and 2a = λb, then λ = ± 2
and P (X) = X + λ.√If λ = 0 then XP (0) + 2P 0 = 0 ⇐⇒ P = P (0) − X 2 P (0)/4.√To conclude the
eigenvalues are {± 2, 0} and the respective eigenvectors are P (X) = a(X ± 2) and P (X) =
a(1 − X 2 /4) for a ∈ R∗ .

Exercise 2.
For t ∈ R, we let  
cos(t) − sin(t)
R(t) = .
sin(t) cos(t)
By interpreting geometrically R(t), determine the values of t for which R(t) has a
real eigenvalue.

Solution: R(t) is a rotation of angle t. If R(t) for t 6= 0 has a real eigenvalue then the associated
eigenspace is a vector line of R2 (it can not be the whole space as R(t) 6= λId for t 6= 0). But the
rotation R(t) let invariant a line only if t = π. Therefore R(t) has a real eigenvalue λ if and only
if t = 0 (then λ = 1) or t = π (then λ = −1).

Exercise 3.
Let J = Mn (R) with only 10 s.
1. What are the eigenvalues of J ?
2. Compute the minimal polynomial of J.
3. Deduce J k for any k > 2.
Solution:
P P
1. Let v = (v1 , · · · , vn ) 6= 0 and λP∈ K with Jv = λv. We have Jv = ( i vi , · · · , i vi ).
Therefore λv1P = λv2 = λvn = i vi . Either λ 6= 0, then v1 = v2 = · · · = vn and λ = n
or λ = 0 and i vi = 0. Conversely we check that λ = n and λ = 0 are eigenvalues with
(1, 1, · · · 1) and (−1, 1, 0 · · · , 0) being associated eigenvectors.
2. we compute J 2 = nJ. Therefore PJ divide X 2 − nX : either PJ = X 2 − nX, = X − n.
But J 6= nId, therefore PJ = X 2 − nX.
3. Consider the euclidean division of X k by X 2 − nX : we get

X k = Q(X 2 − nX) + R (1)

with degree of R less than or equal to 1, i.e. R(X) = aX + b. We may identify the
coefficients a and b by letting firstly X = 0 in (1) : 0 = b, then by letting X = n : nk = na,
i.e. a = nk−1 . Therefore by letting X = J in (1) we get finally J k = nk−1 J.

Exercise 4. 1. Show the roots of the minimal polynomial of an endomorphism

u ∈ L(E) are exactly the eigenvalues of u.
2. Deduce that the set of eigenvalues of u ∈ L(E) is finite.
3. Assume K = C. Show the set of eigenvalues is non-empty.

Solution:
1. λ is a root of Pu
⇐⇒ (i.) Pu (λ) = 0
⇐⇒ (ii.) Pu (X) = (X − λId)Q with Q ∈ K[X] \ {0}
⇐⇒ (iii.) u − λId is not invertible
⇐⇒ (iv.) u − λId is not injective
⇐⇒ (v.) λ is an eigenvalue of u.

Let us detail the proofs of (ii.) =⇒ (iii.) and (v.) =⇒ (i.) :

— ii =⇒ iii. If Pu (X) = (X − λId)Q and u − λId is invertible then by composing on
the left the equality (u − λId)Q(u) = 0 by (u − λId)−1 we get Q(u) = 0, therefore Pu
divides Q. But it is impossible as the degree of Q is less than the degree of Pu .
— v =⇒ i. Assume λ is an eigenvalue. Let v (6= 0) be an associated eigenvector. Then
Pu (v) = P (λ)v = 0, therefore P (λ) = 0.
2. This follows from (1) and the fact that a polynomial in K[X] has at most deg(P ) distinct
roots.
3. This follows from (1) and the fundamental theorem of algebra, which says that any non
constant polynomial in C[X] has a root.

Exercise 5.
Let u be an invertible endomorphism of E.

Page 2
 1. Show 0 is not an eigenvalue of u.
2. Prove that if λ ∈ K is an eigenvalue of u then 1
λ
is an eigenvalue of u−1 .

Solution:
1. 0 is an eigenvalue ⇐⇒ u is not injective ⇐⇒ u is not invertible.
2. Let λ 6= 0 be an eigenvalue of u and let 0 6= x ∈ Eλ (u). We have u(x) = λx, therefore
x = u−1 (λx) = λu−1 (x) and finally λ1 x = u−1 (x). Therefore λ1 is an eigenvalue of u and
x ∈ E λ1 (u−1 ).

Exercise 6.
Assume K = C. Let u, v ∈ L(E) with u ◦ v = v ◦ u. Show that u and v have a
common eigenvector.

Solution: By item 4 of exercise 4, u has an eigenvalue λ ∈ C. Let Eλ (u) be the associated

eigenspace. If x ∈ Eλ (u) we have v(u(x)) = λv(x) = u(v(x)). Therefore v(x) belongs also to
Eλ (u). In other terms Eλ (u) is stable by v. The restriction v|Eλ (u) of v to Eλ (u) admits an
eigenvectors y by applying again item 4 of Exercise 4. Then y is an eigenvector of both u and v.

Exercise 7.
In this exercise we show by induction on the dimension that the minimal polyno-
mial of an endomorphism u ∈ L(E) has degree less than or equal to the dimension
of E.
1. Let v ∈ E \ {0}. Let m be the largest integer such that v, u(v), · · · , um−1 (v)
are linearly independent. Let W be the m dimensional vector space generated
by v, u(v), · · · , um−1 (v). Show there is a monic P ∈ K[X] with degree m such
that P (u)W = {0}.
2. By using the rank nullity theorem, show that the vector space P (u)E has
dimension less than or equal to n − m < n.
3. Conclude the proof by applying the induction hypothesis to u restricted to
P (u)E.

Solution:
By definition of m, the vectors um v is a linear combination
1. P of v, · · · , um−1 v : um v =
m−1 k m
P k
k=0 −ck u v with ck ∈ K for all k. If we let P = X + k ck X then P (u)(v) = 0, but
k k k
also for any k we have P (u)(u (v)) = u (P (u)(v)) = u (0) = 0. Therefore P (u)W = {0}.
2. By the rank nullity theorem, dim(P (u)E) + dim(Ker(P (u))) = dim(E) = n. But W ⊂
Ker(P (u)) and dim(W ) = m. Therefore dim(P (u)E) = n − dim(Ker(P (u))) 6 n −
dim(W ) = n − m.

Page 3
 3. Clearly u(P (u)E) ⊂ P (u)E. Let u0 be the restriction of u to P (u)E. By induction hy-
pothesis, its minimal polynomial Pu0 has degree less than dim(P (u)E) 6 n − m. Then
observe that (Pu0 P )(u) = 0. Indeed for all v ∈ E we have (Pu0 P )(u)(v) = Pu0 (u)(P (u)(v)).
But P (u)(v) ∈ P (u)E and Pu0 is the minimal polynomial of u on P (u)(E), therefore
Pu0 (u)(P (u)(v)) = 0 and (Pu0 P )(u) = 0. In other terms Pu0 P lies in the kernel of the
evaluation map evu , therefore is divided by the minimal polynomial Pu of u. In particular
deg(Pu ) 6 deg(Pu0 P ) = deg(Pu0 ) + deg(P ) 6 n − m + m = n.

Exercise 8.
For c0 , · · · , cn−1 ∈ K, compute the minimal polynomial of the matrix A ∈ Mn (K)
given by  
0 0 · · · 0 −c0
1
 0 · · · 0 −c1  
A = 0
 1 · · · 0 −c2  .
 .. .. . . . .. .. 
. . . . 
0 0 · · · 1 −cn−1

Page 4