EE580, Fall 2023

Homework two
Due on October 3, 2023, 4:30PM. Upload your solution to Canvas.
Please write down name, student ID and course code on the first page of the solution.
   
0 1 0 1 2 −3 4
Problem 1: (20 points) Consider A1 =  0 0 0  , A2 =  0 −1 2 2  .
0 0 −1 0 0 0 1
(P1) Find the ranks and nullities of A1 and A2 ;
(P2) Find bases of the range spaces and null spaces of A1 and A2 .    
1 0
Solution: (P1) ρ(A1 ) = 2. By the rank-nullity theorem, ν(A1 ) = 3 − 2 = 1. The vectors  0  and  0 
  0 −1
1
consist of a basis of R(A1 ) and  0  is a basis of N (A1 ).
0
(P2) We perform elementary transformations on A2 and have
     
1 2 −3 4 1 0 0 0 1 0 0 0
A2 =  0 −1 2 2  →  0 1 2 2  →  0 1 0 0  .
0 0 0 1 0 0 0 1 0 0 0 1
     
1 2 4
So ρ(A2 ) = 3 and thus ν(A2 ) = 4 − 3 = 1. The vectors  0 ,  −1  and  2  consist of a basis of R(A2 ).
0 0 1
In order to find a basis of N (A2 ), we need to solve
x1 + 2x2 − 3x3 + 4x4 = 0
−x2 + 2x3 + 2x4 = 0
x4 = 0.
Since
 x4 = 0, we let x3 = 1 in the second equation and have x2 = 2. Thus x1 = −1 from the first equation. So
−1
 2 
 1  is a basis of N (A2 ).
 

0
 
1 1 0
Problem 2: (15 points) Consider A =  0 0 1  . Compute A103 and eAt .
0 0 1
Solution: A is a triangular matrix and has eigenvalues 0, 1 and 1.
Let h(λ) = β0 + β1 λ + β2 λ2 and f (λ) = λ103 . We have
f (0) = h(0) ⇒ β0 = 0
f (1) = h(1) ⇒ β1 + β2 = 1
f 0 (1) = h0 (1) ⇒ β1 + 2β2 = 103.
 
1 1 102
So β1 = −101 and β2 = 102. Then A103 = −101A + 102A2 =  0 0 1  .
0 0 1
Let h(λ) = β0 + β1 λ + β2 λ2 and f (λ) = eλt . We have
f (0) = h(0) ⇒ β0 = 1
f (1) = h(1) ⇒ β0 + β1 + β2 = et
f 0 (1) = h0 (1) ⇒ β1 + 2β2 = tet .
 et et − 1 tet − et + 1
 

So β1 = 2et − 2 − t and β2 = 1 + (t − 1)et . Then eAt = 0 1 et − 1 

Problem 3: (15 points) Consider a polynomial function f . Show that if λ is an eigenvalue of A with eigenvector
x, then f (λ) is an eigenvalue of f (A) with the same eigenvector x.
Solution: Note that Ax = λx. Multiply A at both sides and we have A2 x = Aλx = λ2 x. Assume Am x = λm x for
some positive integer m. Then Am+1 x = Aλm x = λm+1 x. By induction, we have Am x = λm x for any positive
integer.
Then we have f (A)x = (β0 I + β1 A + · · · + βn−1 An−1 )x = (β0 + β1 λ + · · · + βn−1 λn−1 )x = f (λ)x. Therefore,
f (λ) is an eigenvalue of f (A) with eigenvector x.

Problem 4: (15 points) Consider Ax = y where A is m × n and has rank m. Is (AT A)−1 AT y a solution? If not,
under what condition will it be a solution? Is AT (AAT )−1 y a solution? Explain why.
Solution: Since A has rank m, then n ≥ m.
If n > m, then AT A is n × n and singular. So (AT A)−1 is not defined and (AT A)−1 AT y is not a solution.
AAT is m × m and non-singular. Note that AAT (AAT )−1 y = y . Thus AT (AAT )−1 y is a solution.
If n = m, then A is non-singular. (AT A)−1 AT = AT (AAT )−1 = A−1 and both are solutions.

Problem 5: (20 points) Use the Cayley-Hamilton theorem to show that, for every n × n matrix A, there are n
n−1
X
scalar functions α0 (t), α1 (t), · · · , αn−1 (t) such that eAt = αi (t)Ai for all t ∈ R.
i=0
Solution: By the Cayley-Hamilton theorem, An + a1 An−1 + · · · + an−1 A + an I = 0 where ai are the coefficients
of the characteristic polynomial of A. Therefore, An = −a1 An−1 − · · · − an−1 A − an I . Using this, we have
An+1 = −a1 An − a2 An−1 − · · · − an−1 A2 − an A = (a21 − a2 )An−1 + (a1 a2 − a3 )An−2 + · · · + (a1 an−1 − an )A +
a1 a2 I − · · · − an−1 A2 . Therefore, An+1 can also be written as a linear combination of An−1 , An−2 , · · · , A, I .
Applying the same procedure for increasing powers of A, we have that for every k ≥ 0, Ak can be written as
Ak = ān−1 (k)An−1 + ān−2 (k)An−2 + · · · + ā1 (k)A + ā0 (k)I for appropriate coefficients āi (k). Therefore, it
establishes
∞ k ∞ n−1 n−1 ∞ k
At
X t k X tk X i
XX t āi (k) i
e = A = āi (k)A = ( )A .
k! k! k!
k=0 k=0 i=0 i=0 k=0

Problem 6: (15 points) Consider A ∈ Rm×n , and let R(A) be the range space of A, and N (A) be the null space
of A. Show that
(P1) R(A) is orthogonal to N (AT );
(P2) R(A) ∩ N (AT ) = {0};
(P3) The union of a basis of R(A) and a basis of N (AT ) qualifies as a basis of Rm ;
Hint: Two subspaces V, W of Rn are orthogonal provided v T w = 0 for any v ∈ V and w ∈ W . The set of all
vectors in Rn which are orthogonal to V is called the orthogonal complement of V .
Solution: (P1) Choose y1 ∈ R(A) and y2 ∈ N (AT ). Then we have y1 = Ax for some x ∈ Rn . So y1T y2 =
(Ax)T y2 = xT (AT y2 ) = 0. This implies that R(A) is orthogonal to N (AT ).
(P2) Choose a non-zero y ∈ N (AT ). Then AT y = 0 and y T Ax = 0 for all x ∈ Rn . This implies that y 6= Ax
for any x ∈ Rn . So, y ∈ / R(A). Hence, R(A) ∩ N (AT ) = {0}.
(P3) Since rank(A) = rank(AT ), dim(R(A)) + dim(N (AT )) = dim(R(AT )) + dim(N (AT )) = m. So, if
{v1 , · · · , vk } is a basis of R(A) and {w1 , · · · , wl } is a basis of N (AT ), then k + l = m. Since vi and wj are
orthogonal to each other, v1 , · · · , vk , w1 , · · · , wl are linearly independent. Hence, {v1 , · · · , vk , w1 , · · · , wl } is a
basis of Rm .