MASSACHUSETTS INSTITUTE OF TECHNOLOGY

Department of Electrical Engineering and Computer Science

6.241: Dynamic Systems—Fall 2003

Homework 1 Solutions

Exercise 1.1 a)Given A1 and A4 are square matrices, we know that A is square:



A1 A2 I 0
A= = .
0 A4 0 A4

Note that

A1 A2
det = det(I)det(A4 ) = det(A4 ),
0 I
which can be verified by recursively computing the principal minors. Also, by the elementary
operations of rows, we have



A1 A2 A1 0
det = = det = det(A1 ).
0 I 0 I

Thus, finally using the fact that det(AB) = det(A) = det(B), we have

det(A) = det(A1 )det(A4 ).

b) Assume A−1 −1
1 and A4 exist. Then




−1 A1 A2 B1 B2 I 0
AA = = ,
0 A4 B3 B4 0 I

which yields four matrix equations:

1. A1 B1 + A2 B3 = I,

2. A1 B2 + A2 B4 = 0,

3. A4 B3 = 0,

4. A4 B4 = I.

From Eqn (4), B4 = A−1 −1 −1

4 , with which Eqn (2) yields B2 = −A1 A2 A4 . Also, from Eqn (3) B3 = 0,
−1
with which from Eqn (1) B1 = A1 . Therefore,

−1 
−1 A1 −A1−1 A2 A4−1
A = .
0 A−1
4

Exercise 1.2 a)




0 I A1 A2 A3 A4
=
I 0 A3 A4 A1 A2

1
b) Let us find


B1 B2
B =
B3 B4

such that


A1 A2
BA =
0 A4 − A3 A−1
1 A2

The above equation implies four equations for submatrices

1. B1 A1 + B2 A3 = A1 ,

2. B1 A2 + B2 A4 = A2 ,

3. B3 A1 + B4 A3 = 0,

4. B3 A2 + B4 A4 = A4 − A3 A−1
1 A2 .

First two equations yield B1 = I and B2 = 0. Express B3 from the third equation as B3 =
−B4 A3 A−1 −1
1 and plug it into the fourth. After gathering the terms we get B4 A4 − A3 A1 A2 =
−1
A4 − A3 A1 A2 , which turns into identity if we set B4 = I. Therefore


I 0
B =
−A3 A−11 I

c) Using linear operations

 on rows we see that det (B)
 = 1. Then, det(A) = det(B)det(A) =
det (BA) = det (A1 ) det A4 − A3 A−1
1 A2 . Note that A4 − A3 A1−1 A2 does not have to be invert­
ible for the proof.

Exercise 1.3 We have to prove that det(I − AB) = det(I − BA).

Proof: Since I and I − BA are square,


I 0
det(I − BA) = det
B I − BA




I A I −A
= det
B I 0 I



I A I −A
= det ,
B I 0 I

yet, from Exercise 1.1, we have


I −A
det = det(I)det(I) = 1.
0 I

Thus,

I A
det(I − BA) = det .
B I
Now,


I A I − AB 0
det = det = det(I − AB).
B I B I

2
Therefore
det(I − BA)det(I − AB).
Note that (I − BA) is a q × q matrix while (I − AB) is a p × p matrix. Thus, one wants to compute
the determinant of (I − AB) or (I − BA), s/he compare p and q to pick a smaller size of product
either AB or BA.
b) We have to show that (I − AB)−1 A = A(I − BA)−1 .
Proof: Assume that (I − BA)−1 and (I − AB)−1 exist. Then,

A = A · I = A(I − BA)(I − BA)−1

= (A − BAB)(I − BA)−1
= (I − AB)A(I − BA)−1
→ (I − AB)−1 A = A(I − BA)−1 .

This completes the proof.

c) We have to show that (A + BCD)−1 = A−1 − A−1 B(C −1 + DA−1 B)−1 DA−1 .
Proof: Suppose (A + BCD)−1 = A−1 − A−1 B(C −1 + DA−1 B)−1 DA−1 , then

(A + BCD)−1 (A + BCD) = I = (A−1 − A−1 B(C −1 + DA−1 B)−1 DA−1 )(A + BCD)

= (I + A−1 BCD) − A−1 B(C −1 + DA−1 B)−1 D(I + A−1 BCD)

= (I + A−1 BCD) − A−1 B(C −1 (I + CDA−1 B))−1 D(I + A−1 BCD)
= (I + A−1 BCD) − A−1 B(I + CDA−1 B)−1 CD(I + A−1 BCD)
= (I + A−1 BCD) − A−1 BCD(I + A−1 BCD)−1 (I + A−1 BCD)
= I ∵ part b.

where we have used from part (b) that (I + CDA−1 B)−1 CD = CD(I + A−1 BCD)−1 . Then we
have to show that (I − abT )−1 only requires inversion of A scalar.
Proof: From the lemma proved above, let

A = I, B = −a, C = I, D = bT ,

then

(I − abT )−1 = I + a(I − bT a)−1 bT

abT
= I+ .
(1 − bT a)

Note that bT a is a scalar.

Exercise 1.4 a) First define all the spaces:

R(A) = {y ∈ Cn | Ax = y, ∀x ∈ Cn }
R⊥ (A) = {z ∈ Cm | y  z = z  y = 0, ∀y ∈ R(A)}
R(A ) = {p ∈ Cn | A v = p, ∀v ∈ Cm }
N (A) = {x ∈ Cn | Ax = 0}
N (A ) = {q ∈ Cm | A q = 0}

3
i) Prove that R⊥ (A) = N (A ).
Proof: Let
z ∈ R⊥ (A) → y  z = 0 ∀y ∈ R(A)
→ x A z = 0 ∀x ∈ Cn
→ A z = 0 → z ∈ N (A )
→ R⊥ (A) ⊂ N (A ).
Now let
q ∈ N (A ) → A q = 0
→ x A q = 0 ∀x ∈ Cn
→ y  q = 0 ∀y ∈ R(A)
→ q ∈ R⊥ (A)
→ N (A ) ⊂ R⊥ (A).
Therefore
R⊥ (A) = N (A ).
ii) Prove that N ⊥ (A) = R(A ).
Proof: From i) we know that N (A) = R⊥ (A ) by switching A with A . That implies that
N ⊥ (A) = {R⊥ (A )}⊥ = R(A ).
b) Show that rank(A) + rank(B) − n ≤ rank(AB) ≤ min{rank(A), rank(B)}.
Proof: i) Show that rank(AB) ≤ min{rank(A), rank(B)}. It can be proved as follows:
Each column of AB is a combination of the columns of A, which implies that R(AB) ⊆ R(A) →
rank(AB) ≤ rank(A).
Each row of AB is a combination of the rows of B → rowspace (AB) ⊆ rowspace (B), but the
dimension of rowspace = dimension of column space = rank, so that rank(AB) ≤ rank(B).
Therefore,

rank(AB) ≤ min{rank(A), rank(B)}.

ii) Show that rank(A) + rank(B) − n ≤ rank(AB).
Let

rB = rank(B)
rA = rank(A)
KB = dim(N (B  )) = nullity of B  = n − rB
KA = nullity of A = n − rA ,
where A ∈ Cm×n , B ∈ Bn×p .
Now, let {v1 , · · · , vrB } be a basis set of R(B), and add n − rB linearly independent vectors
{w1 , · · · , wn−rB } to the basis to span all of Cn , {v1 , v2 , · · · , vn , w1 , · · · , wn−rB }.
Let
   
M= v1 | v2 · · · vrB | w1 | · · · wn−rB = V W .
Suppose x ∈ Cn , then x = M α for some α ∈ Cn .

4
 1. R(A) = R(AM ) = R([AV |AW ]).
Proof: i) Let x ∈ R(A) → Ay = x for some y ∈ Cn . Yet , y can be written as a linear
combination of the basis vectors of Cn , so y = M α for some α ∈ Cn .
Then, Ay = AM α = x → x ∈ R(AM ) → R(A) ⊂ R(AM ).
ii) Let x ∈ R(AM ) → AM y = x for some y ∈ C n , but M y = z ∈ Cn → Az = x → x ∈
R(A) → R(AM ) ⊂ R(A).
Therefore, R(A) = R(AM ) = R([AV |AW ]).
2. R(AB) = R(AV ).
Proof: i) Let x ∈ R(AV ) → AV y = x for some y ∈ CrB . Yet, V y = Bα for some α ∈ Cp
since the columns of V and B span the same space. That implies that AV y = ABα = x →
x ∈ R(AB) → R(AV ) ⊂ R(AB).
ii) Let x ∈ R(AB) → (AB)y = x for some y ∈ Cp . Yet, again By = V θ for some θ ∈ CrB →
ABy = AV θ = x → x ∈ R(AV ) → R(AB) ⊂ R(AV ).
Therefore, R(AV ) = R(AB).
Using the fact 1, we see that the number of linearly independent columns of A is less than or
equal to the number of linearly independent columns of AV + the number of linearly independent
columns of AW . , which means that
rank(A) ≤ rank(AV ) + rank(AW ).
Using the fact 2, we see that
rank(AV ) = rank(AB) → rank(A) ≤ rank(AB) + rank(AW ),
yet, there re only n − rB columns in AW . Thus,

→ rank(AW ) ≤ n − rB
→ rank(A) − rank(AB) ≤ rank(AW ) ≤ n − rB
→ rA − (n − rB ) ≤ rAB .
This completes the proof.

Exercise 1.8 Let X = {g(x) = α0 + α1 x + α2 x2 + · · · + αM xM | αi ∈ C}.

a) We have to show that the set B = {1, x, · · · , xM } is a basis for X.
Proof :
1. First, let’s show that elements in B are linearly independent. It is clear that each element in
B can not be written as a linear combination of each other. More formally,
c1 (1) + c1 (x) + · · · + cM (xM ) = 0 ↔ ∀i ci = 0.
Thus, elements of B are linearly independent.

2. Then, let’s show that elements in B span the space X. Every polynomial of order less than
or equal to M look like

M
p(x) = αi xi
i=0
for some set of αi ’s.
Therefore, {1, x1 , · · · , xM } span X.

5
b) T : X → X and T (g(x)) = d
dx g(x).

1. Show that T is linear.

Proof:
d
T (ag1 (x) + bg2 (x)) = (ag1 (x) + bg2 (x))
dx
d d
= a g1 + b g2
dx dx
= aT (g1 ) + bT (g2 ).

Thus, T is linear.

2. g(x) = α0 + α1 x + α2 x2 + · · · + αM xM , so

T (g(x)) = α1 + 2α2 x + · · · + M αM xM −1 .

Thus it can be written as follows:

    
0 1 0 0 ··· 0 α0 α1
 0 0 2 0 ··· 0  α1   2α2 
    
 0 0 0 3 ··· 0  α2   3α3 
    
 .. .. .. .. ..  α3 = .. .
 . . . . . 
0    . 
  ..   
 0 0 0 0 ··· M  .   M αM 
0 0 0 0 ··· 0 αM 0
The big matrix, M , is a matrix representation of T with respect to basis B. The column
vector in the left is a representation of g(x) with respect to B. The column vector in the
right is T (g) with respect to basis B.

3. Since the matrix M is upper triangular with zeros along diagonal (in fact M is Hessenberg),
the eigenvalues are all 0;
λi = 0 ∀i = 1, · · · , M + 1.

4. One eigenvector of M for λ1 = 0 must satisfy M V1 = λ1 V1 = 0

 
1
 0 
 
V1 =  . 
 .. 
0

is one eigenvector. Since λi ’s are not distinct, the eigenvectors are not necessarily inde­
pendent. Thus in order to computer the M others, ones uses the generalized eigenvector
formula.