MATH2040A/B Homework 8 Solution

   
1 1 0  1−λ 1 0 
 
1. (f) For A =  0 1 2  , det(A − λI) =  0 1−1 2  = (1 − λ)2 (3 − λ).

0 0 3  0 0 3−λ 
Hence the eigenvalues of A are λ = 1 or 3. For λ = 1,
   
0 1 0 0 1 0
RREF
A − λI =  0 0 2  −→  0 0 1 
0 0 2 0 0 0

then Eλ = span{(1, 0, 0)T }. Since γA (1) = 1, µA (1) = 2, A is not diagonalizable.

2. Denote β 0 = 1, x, x2 as the standard ordered basis of P2 (R). Then we have


 
     0 0 1
T x2 β 0 = x2 β 0 [x]β 0 [1]β 0 =  0 1 0 


[T ]β 0 = [T (1)]β 0 [T (x)]β 0
1 0 0
We proceeds to find the characteristic function of T , given by
 
−t 0 1  
−t 1
fT (t) = det ([T ]β 0 − tI3 ) = det  0 1−t 0  = (1−t) det = −(1−t)2 (t+1)
1 −t
1 0 −t
The characteristic polynomial of T splits. It follows that the root is given by λ1 = 1
and λ2 = −2, and their algebraic multiplicity is given by 2 and 1 respectively. We then
find the eigenspaces one by one.
− (λ1 = 1) The eigenspace associated with the eigenvalue λ = λ1 = 1 is given by
       
−1 0 1  1 0 
Eλ1 = N ([T ]β 0 − I3 ) = N  0 0 0  = t  0  + s  1  : t, s ∈ R
1 0 −1 1 0
 

Therefore, µT (1) = 2 = dim (Eλ1 ) =: γT (1)

− (λ2 = −1) The eigenspace associated with the eigenvalue λ = λ2 = −1 is given by
     
1 0 1  1 
Eλ2 = N ([T ]β 0 + I3 ) = M  0 2 0  = t  0  : t ∈ R
1 0 1 −1
 

Therefore, µT (−1) = 1 = dim (Eλ2 ) =: γT (−1)

As µT (1) = γT (1) and µT(−1) = γT (−1), it follows that T is diagonalizable.
As T is
diagonalizable, also β1 := (1, 0, 1)T , (0, 1, 0)T and β2 := (1,0, −1)T are the ordered
basis of Eλ1 and Eλ2 respectively. It follows that β := β1 ∪β2 = (1, 0, 1)T , (0, 1, 0)T , (1, 0, −1)T
is an ordered basis for V consisting of eigenvectors and hence
 
1 0 0
[T ]β =  0 1 0 
0 0 −1
   
5 0 1 2
3. Diagonalize the matrix A by Q−1 AQ = D with D = and Q = .
0 −1 1 −1
So we know that  n 
n −1 5 0
n
A = QD Q = Q Q−1 .
0 (−1)n

1
 4. As [T ]β is upper triangular, denote the i -th diagonal entries as δi , then we have
n
Y
fT (t) = det ([T ]β − tI) = (δi − t) = 0
i=1

have the roots δ1 , · · · , δn , where n = dim(V ). As proven in the previous assignment,

[T ]β and T has same set of eigenvalues. Therefore, δ1 , · · · , δn is the set of eigenvalues
for T and hence it must be the case that λi appears mi times exactly within δ1 , · · · , δn .
The statement then holds.
5. (a) We may pick one basis α such that both [T ]α and [U ]α are diagonal. Let Q = [I]βα .
And we will find out that
[T ]α = Q−1 [T ]β Q
and
[U ]α = Q−1 [U ]β Q.
(b)Let Q be the invertible matrix who makes A and B simultaneously diagonalizable.
Say β be the basis consisting of the column vectors of Q. And let α be the standard
basis. Now we know that
−1
[T ]β = [I]βα [T ]α [I]α
β =Q AQ
and
−1
[U ]β = [I]βα [U ]α [I]α
β =Q BQ.
 
1 0
6. (e) No. For ∈ W, we have
0 2
 
0 2
T (A) = ∈
/ W.
1 0

7. Let {Wi }i∈I be the collection of T -invatirant subspaces and W be the intersection of
them. For every v ∈ W, we have T (v) ∈ Wi for every i ∈ I, since v is an element is each
Wi . This means T (v) is also an element in W .
8. (a) Let w be an element in W . We may express w to be
k
X
w= ai T i (v)
i=0

And thus we have

k
X
T (w) = ai T i+1 (v) ∈ W.
i=0

(b) Let U be a T -invariant subspace of V containing v. since it’s T invariant, we know

that T (v) is an element in U . Inductively, we know that T k (v) ∈ U for all nonnegative
integer k. By Theorem 1.5 we know that U must contain W .
Remark: Theorem 1.5: The span of any subset S of a vector space V is a subspace
of V. Moreover, any subspace of V that contains S must also contain the span of S
9. If w is an element in W, it’s a linear combination of
v, T (v), T 2 (v), . . .


So w = g(T )(v) for some polynomial g. Conversely, if w = g(T )(v) for some polynomial
g, this means w is a linear combination of the same set. Hence w is an element in W .
10. Define
   
0 0 ··· 0 −a0 −t 0 · · · 0 −a0

 1 0 ··· 0 −a1 


 1 −t · · · 0 −a1 

 0 1 ··· 0 −a2   0 1 ··· 0 −a2 
Ak =  ⇒ Ak −tIk = 
   
.. .. .. ..  .. .. .. .. 

 . . . . 


 . . . . 

 0 0 ··· 0 −ak−2   0 0 ··· −t −ak−2 
0 0 ··· 1 −ak−1 0 0 ··· 1 −ak−1 − t

2
Let P (n) be the statement that the characteristic polynomial of An is given by

(−1)n a0 + · · · + an−1 tn−1 + tn .

For n = 1, notice that A1 = −a0 − t = (−1)1 a0 + t1 . P (1) is true. Suppose P (k) is
true for some k ∈ N, i.e.
 
−t 0 · · · 0 −a0
 1 −t · · · 0 −a1 
 
 0 1 ··· 0 −a2 
 = (−1)k a0 + · · · + ak−1 tk−1 + tk


det (Ak − tIk ) = det  .
 
. .
. .
. .
.
 . . . . 
 
 0 0 · · · −t −ak−2 
0 0 · · · 1 −ak−1 − t

It follows that
 
−t 0 · · · 0 0 −a0

 1 −t · · · 0 0 −a1 


 0 1 ··· 0 0 −a2 

det (Ak+1 − tIk+1 ) = det 
 .. .. .. .. .. 
 . . . . . 


 0 0 ··· −t 0 −ak−2 

 0 0 ··· 1 −t −ak−1 
0 0 ··· 0 1 −ak − t
 ,
−t 0 · · · 0 0 −a0

 1 −t · · · 0 0 −a1


 0 1 ··· 0 0 −a2

= det 
 .. .. .. .. ..
 . . . . .


 0 0 ··· −t 0 −ak−2 

 0 0 ··· 1 −t −ak−1 − t 
0 0 ··· 0 1 −ak − t + 1

where the last equality follows from adding the second last column to the last column.
Consider expanding the above determinant along the last row, we obtained that
 
−t 0 · · · 0 0
 1 −t · · · 0 0 
 
 0 1 · · · 0 0 
det (Ak+1 − tIk+1 ) = − det (Ak − tIk ) + (−ak − t + 1) det  .
 
.. .. .. 
 .. . ··· . . 
 
 0 0 · · · −t 0 .
0 0 · · · 1 −t
k k−1
+ t + (−ak − t + 1) (−t)k
k


= −(−1) a0 + · · · + ak−1 t
= (−1)k+1 a0 + · · · + ak−1 tk−1 + tk + (−1)k+1 ak tk + tk+1 − tk

= (−1)k+1 a0 + · · · + ak−1 tk−1 + ak tk + tk+1

Therefore, P (k + 1) also holds. It follows by principle of mathematical induction

that
the characteristic polynomial of A is (−1)k a0 + a1 t + · · · + ak−1 tk−1 + tk .