Answers/Solutions of Exercise 2 (Version: August 24, 2024)

 
−4 2 5 8
−6 6 −6
 
 1 −5 −5 −8 
 
1. (a) −6 6 −6 (b) (c) Not possible
−1 2 2 8 
 
−6 6 −6
1 −2 −5 −11
 
−3 3 −4
3 −6 −15 −24
   
1 3 6
 3 −3 4
 
(d) 0 4 10 (e) (f) 8 −4 −16 −28
−3 3 −4
 
0 0 9 9 0 −9 −18
3 −3 4
−38
   
0 0
(g) Not possible (h) 0 0  (i) −28
0 0 −18
   
16 0 0 0 0 0
(j) Not possible (k) 22 2 (l) 0 0 0 0
6 0 0 0 0 0
 
  1 −1 3 2
0 1 1
−1 1 −3 −2
 
(m)  −1 0 2 (n) Not possible (o) 
 3 −3 9 6

−1 −2 0
2 −2 6 4
(p) 15

2. a = 0, b = −1, c = 2, d = −4.

3. (a) (i) (3,4)-entry of AB (ii) (4,1)-entry of AB

(iii) (3,2)-entry of BA (iv) (2,5)-entry of BA
Pn Pm
(b) (i) j=1 a3j bj2 (ii) i=1 b4i ai1

p
X
4. (a) ai1 b1j + ai2 b2j + ai3 b3j + · · · + aip bpj = aik bkj
k=1
p
X
(b) ci1 c1j + ci2 c2j + ci3 c3j + · · · + cip cpj = cik ckj
k=1
p
X
(c) ai1 cj1 + ai2 cj2 + ai3 cj3 + · · · + aip cjp = aik cjk
k=1

1
  
0 1 1
5. For example, A = −1 0 1.
−1 −1 0
The general form of the matrix A = (aij )3×3 is aii = 0 for i = 1, 2, 3 and
aij = −aji for all other values of 1 ≤ i, j ≤ 3.
   
1 0 0 0 0 0
6. (a) For example, A = 0 0 0 and B = 0 0 1.
0 0 0 0 0 0
   
1 0 0 0 0 1
(b) For example, A = 0 0 0 and B =  0 0 0.
0 0 0 0 0 0
     
1 0 0 0 0 1 0 0 2
(c) For example, A = 0
 0 0 ,B= 0 0
  0 and C = 0 0 0.
0 0 0 0 0 0 0 0 0
     
2 3 −1 2 3 −1 2 3 −1
7. The matrix A can be , , , etc.
0 0 0 2 3 −1 4 6 −2

8. (a) S is a straight line joining (1, 0, 3) and (0, −1, 3).

   
1 −1 0 1
(b) For example, A = and b = .
0 0 1 3
The linear system consists of two planes which intersect at the line S.

9. Suppose x = v is a solution to Ax = 0, i.e. Av = 0. If Ax = b has a solution

x = u, then
A(u + v) = Au + Av = b + 0 = b
and hence u + v is also a solution to Ax = b.
As Ax = 0 has infinitely many solutions (why?), if Ax = b has a solutions, it
has inifinitely many solutions. We conclude that Ax = b has either no solutions
or infinitely many solutions.

10. (a) Let x = u be any solution to the system Bx = 0. Then ABu = A0 = 0.

The system ABx = 0 has at least as many solutions as the system Bx =
0. Thus it has infinitely many solutions.
   
1 0 1 0
(b) No. For example, let B = and consider two cases (i) A =
0 1 0 1
 
1 0
and (ii) A = . Note that Bx = 0 has only the trivial solution. For
0 0

2
 (i), ABx = 0 has only the trivial solution while for (ii), ABx = 0 has
infinitely many solutions.

11. (a) (i) 2; (ii) −6; (iii) 16.

(b) tr(A + B) = (a11 + b11 ) + (a22 + b22 ) + · · · + (ann + bnn )
= (a11 + a22 + · · · + ann ) + (b11 + b22 + · · · + bnn ) = tr(A) + tr(B).
(c) tr(cA) = ca11 + ... + cann = c(a11 + ... + ann ) = c tr(A).
(d) The (i, i)-entry of CD = ci1 d1i + ci2 d2i + ... + cin dni . Thus,
m
X n
X
tr(CD) = (ci1 d1i +ci2 d2i +...+cin dni ) = (c1j dj1 +c2j dj2 +...+cmj djm ).
i=1 j=1

But the (i, i)-entry of DC = di1 c1i + di2 c2i + ... + dim cmi . So the trace of
DC is precisely the term on the right hand side above.
(e) By (d), tr(AB) = tr(BA). Then by (b) and (c), tr(AB − BA) =
tr(AB) − tr(BA) = 0. However, tr(I)=n. It is impossible to have square
matrices A and B such that AB − BA = I.

12. (a) (i) is not orthogonal while (ii) is orthogonal.

(b) (AB)(AB)T = ABB T AT = AIAT = I and (AB)T (AB) = B T AT AB =
BIB T = I since both A and B are orthogonal. Thus AB is orthogonal.
         
0 1 0 1 0 0 0 0 0 0 0 0
13. (a) = and = .
0 0 0 0 0 0 1 0 1 0 0 0

(b) Since AB = BA, (AB)k = Ak B k (you need to prove it by using the

mathematical induction). Since A is nilpotent, Ak = 0 for some positive
integer k. Thus (AB)k = Ak B k = 0 and AB is nilpotent.
   
0 1 0 0
(c) No. For example, let A = and B = . Note that A is
0 0 1 0
     
1 0 0 0 k 1 0
nilpotent and AB = ̸= = BA. But (AB) =
0 0 0 1 0 0
for all k and hence AB is not nilpotent. (For this case, (AB) ̸= Ak B k .)
k

 
1 1
14. (a) All except satisfy (⋆).
1 0

(b) Since P , Q ∈ B, AP = P A and AQ = QA. Then,

A(P + Q) = AP + AQ = P A + QA = (P + Q)A.

3
 Hence P + Q satisfies (⋆).
Likewise, A(P Q) = AP Q = P AQ = P QA = (P Q)A and henceP Q
satisfies (⋆).
       
p q p q p+r q+s p p+q
(c) A = A ⇔ =
r s r s r s r r+s
Thus the conditions are r = 0 and s = p.

15. (a) The statement is clearly true when k = 1. Assume that statement is true
when k = n, i.e.
 n 
a 0 0
n
D = 0  bn 0  .
0 0 cn
Then D n+1 = DD n ie.
  n   n+1 
a 0 0 a 0 0 a 0 0
D n+1 = 0 b 0  0 bn 0  =  0 bn+1 0 .
0 0 c 0 0 cn 0 0 cn+1
Thus the statement is true when k = n+1. By the mathematical induction
the statement is true for all positive intergers k.
−1 0 0
 

(b) A =  0 1 0 .
0 0 −1
±1 0
 
0
(c) There are 8 such diagonal matrices B:  0 ±2 0 .
0 0 ±3
     
1 0 1 1 1 0
16. (a) No. For example, A = ,B= and C = .
0 1 0 0 1 0
(b) ABC = BAC = BCA and ACB = CAB = CBA.

17. (a) x1 = z0 is the number of babies in next year; y1 = 0.5x0 is the number
of one-year-old cubs in next year; and z1 = 0.6y0 + 0.7z0 is the number of
adults in next year.
(b) xn , yn and zn are the numbers of babies, one-year-old cubs and adults,
respectively, after n years.
   3    
x3 0 0 1 0 49
(c)  y3  = 0.5 0 0   0  =  35 .
z3 0 0.6 0.7 100 64.3
Thus the total population three years later is x3 + y3 + z3 ≈ 148.

4
18. Let A = (aij )m×n and B = (bij )m×n . Since all matrices in this question are of
the same size, we only need to check the (i, j)-entries.

(a) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,

the (i, j)-entry of A + B = aij + bij
= bij + aij (by a property of real numbers)
= the (i, j)-entry of B + A.
Thus A + B = B + A.
(b) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,
the (i, j)-entry of c(A + B) = c(the (i, j)-entry of A + B)
= c(aij + bij )
= caij + cbij (by a property of real numbers)
= the (i, j)-entry of cA + the (i, j)-entry of cB
= the (i, j)-entry of cA + cB.
Thus c(A + B) = cA + cB.
(c) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,
the (i, j)-entry of (c + d)A = (c + d)aij
= caij + daij (by a property of real numbers)
= the (i, j)-entry of cA + the (i, j)-entry of dA
= the (i, j)-entry of cA + dA.
Thus (c + d)A = cA + dA.
(d) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,
the (i, j)-entry of c(dA) = c(the (i, j)-entry of dA)
= c(daij )
= (cd)aij (by a property of real numbers)
= the (i, j)-entry of (cd)A.
Thus c(dA) = (cd)A and hence d(cA) = (dc)A = (cd)A (where the last
equality follows by dc = cd which is a property of real number).
(e) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,
the (i, j)-entry of 0 + A = 0 + aij
= aij (by a property of real numbers)
= the (i, j)-entry of A.
an by (a), A + 0 = 0 + A = A.

5
 (f) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,

the (i, j)-entry of A − A = aij − aij

= 0 (by a property of real numbers)
= the (i, j)-entry of 0.

Thus A − A = 0.

(g) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,

the (i, j)-entry of 0A = 0 · aij

= 0 (by a property of real numbers)
= the (i, j)-entry of 0.

Thus 0A = 0.

P
19. It is easier to use the summation notation to do this question.

(a) Let A = (aij )m×p , B = (bij )p×q and C = (cij )q×n .

(i) The size of BC is p × n and hence the size of A(BC) is m × n. On
the other hand, the size of AB is m × q and hence the size of (AB)C
is m × n. So the sizes of A(BC) and (AB)C are the same.
(ii) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,
the (i, j)-entry of A(BC)
p
X
= aik (the (k, j)-entry of BC)
k=1
p
X
= aik (bk1 c1j + bk2 c2j + · · · + bkq cqj )
k=1
p
X
= (aik bk1 c1j + aik bk2 c2j + · · · + aik bkq cqj )
k=1
p q
X X
= aik bkr crj .
k=1 r=1

On the other hand,

the (i, j)-entry of (AB)C
q
X
= (the (i, r)-entry of AB)cr,j
r=1
q
X
= (ai1 b1r + ai2 b2r + · · · + aiq bqr )cr,j
r=1

6
 q
X
= (ai1 b1r crj + ai2 b2r crj + · · · + aiq bqr crj )
r=1
q
p p q
XX X X
= aik bkr crj = aik bkr crj .
r=1 k=1 k=1 r=1

Thus the (i, j)-entries of A(BC) and (AB)C are the same.
By (i) and (ii), A(BC) = (AB)C.

(b) Let A = (aij )p×n , C1 = (cij )m×p and C2 = (dij )m×p .

(i) The size of C1 + C2 is m × p and hence the size of (C1 + C2 )A is
m × n. On the other hand, the sizes of both C1 A and C2 A are m × n
and hence the size of C1 A+C2 A is m×n. So the sizes of (C1 +C2 )A
and C1 A + C2 A are the same.

(ii) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,

the (i, j)-entry of (C1 + C2 )A
p
X
= (the (i, k)-entry of C1 + C2 )akj
k=1
p
X
= (cik + dik )akj
k=1
p
X
= (cik akj + dik akj )
k=1
p p
X X
= cik akj + dik akj
k=1 k=1
= (the (i, j)-entry of C1 A) + (the (i, j)-entry of C2 A).
By (i) and (ii), (C1 + C2 )A = C1 A + C2 A.

(c) Let A = (aij )m×p and B = (bij )p×n .

(i) The sizes of all the three matrices are m × n.

(ii) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,

p p
X X
the (i, j)-entry of c(AB) = c aik bkj = caik bkj ,
k=1 k=1
p p
X X
the (i, j)-entry of (cA)B = (the (i, k)-entry of cA)bkj = (caik )bkj ,
k=1 k=1
p p
X X
the (i, j)-entry of A(cB) = aik (the (k, j)-entry of cB) = aik (cbkj ).
k=1 k=1

7
 Thus the (i, j)-entries of all the three matrices are the same.
By (i) and (ii), c(AB) = (cA)B = A(cB).

0 if i = j
(d) Let A = (aij )m×p and let δij =
1 if i ̸= j.
(i) The size of A0n×q is m × q which is equal to the size of 0m×q ; the size
of 0p×m A is p × n which is equal to the size of 0p×n ; and finally, all
three matrices AI n , I m A and A are m × n.

(ii) For i = 1, 2, . . . , m and j = 1, 2, . . . , q,

n
X
the (i, j)-entry of A0n×q = aik 0 = 0 = the (i, j)-entry of 0m×q .
k=1

For i = 1, 2, . . . , p and j = 1, 2, . . . , n,
X m
the (i, j)-entry of 0p×m A = 0akj = 0 = the (i, j)-entry of 0p×n .
k=1

For i = 1, 2, . . . , m and j = 1, 2, . . . , n,
Xn
the (i, j)-entry of AI n = aik δkj = aij = the (i, j)-entry of A.
k=1

For i = 1, 2, . . . , m and j = 1, 2, . . . , n,
Xm
the (i, j)-entry of I m A = δik akj = aij = the (i, j)-entry of A.
k=1
Thus A0n×q = 0m×q , 0p×m A = 0p×n and AI n = I m A = A.

20. (a) (i) The size of AT is n × m and hence the size of (AT )T is m × n which
is equal to the size of A.

(ii) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,

the (i, j)-entry of (AT )T = the (j, i)-entry of AT = the (i, j)-entry of A.
By (i) and (ii), (AT )T = A.

(b) Let A = (aij )m×n and B = (bij )m×n .

(i) The sizes of the two matrices are n × m.

(ii) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,

the (i, j)-entry of (A + B)T
= the (j, i)-entry of A + B

8
 = aji + bji
= the (i, j)-entry of AT + the (i, j)-entry of B T
= the (i, j)-entry of AT + B T .
By (i) and (ii), (A + B)T = AT + B T .

(c) (i) The sizes of the two matrices are n × m.

(ii) For i = 1, 2, . . . , m and j = 1, 2, . . . , n,
the (i, j)-entry of (cA)T
= the (j, i)-entry of cA
= c (the (j, i)-entry of A)
= c (the (i, j)-entry of AT )
= the (i, j)-entry of cAT .
By (i) and (ii), (cA)T = cAT .
 
1 0 4
21. X = 0 0 0.
0 0 3

22. It suffice to show that if the linear system has more than one solution, it must
has infinitely many solutions.
Suppose Ax = b has two different solutions u and v, i.e. Au = b, Av = b and
u ̸= v. Then for all t ∈ R,

A(tu + (1 − t)v) = tAu + (1 − t)Av = tb + (1 − t)b = b

and hence tu + (1 − t)v is a solution of Ax = b. Since t1 u + (1 − t1 )v ̸=

t2 u + (1 − t2 )v whenever t1 =
̸ t2 , there are infinitely many solutions.



23. (a) Let B1 = b1 · · · bp and B2 = c1 · · · cq where b1 , . . . , bp are
columns of B1 and c1 , . . . , cq are columns of B2 . Then



B1 B2 = b 1 · · · b p c 1 · · · c q .

By (2) of Notation 2.2.15, we have

AB1 = Ab1 · · · Abp ,


AB2 = Ac1 · · · Acq ,



A B1 B2 = Ab1 · · · Abp Ac1 · · · Acq .



Hence A B1 B2 = AB1 AB2 .

9


(b) No. The size of C1 C2 is r × 2m and hence we cannot pre-multiply the
matrix to A.
   
d1 f1
 ..   .. 
(c) Let D1 =  .  and D2 =  .  where d1 , . . . , ds are rows of D1 and
ds ft
f1 , . . . , ft are rows of D2 . Then
 
d1
 .. 
.
   
D1  ds 
=
 f1  .

D2  
 .. 
.
ft
By (3) of Notation 2.2.15, we have

d1 A
 .. 
     . 
d1 A f1 A    
D1 ds A
D1 A =  ...  , D2 A =  ...  , A=
 f1 A  .
    
D2
ds A ft A
 
 .. 
 . 
ft A
   
D1 D1 A
Hence A= .
D2 D2 A

24. (a) True. Let A = (aij )n×n and B = (bij )n×n . Since aij = bij = 0 for i ̸= j,
the (i, j)-entry of AB is equal to

aii bii if i = j
ai1 b1j + ai2 b2j + · · · + ain bnj =
0 if i ̸= j.
Likewise, the (i, j)-entry of BA is equal to

bii aii if i = j
bi1 a1j + bi2 a2j + · · · + bin anj =
0 if i ̸= j.
Thus AB = BA.

(b) True. Let D = 12 (A + AT ).

T
D T = 12 (A + AT ) = 12 (A + AT )T = 12 (AT + (AT )T ) = 21 (AT + A) = D.


Thus D is symmetric.

10
    
0 1 0 0
(c) False. For example, let A = and B = .
0 0 1 0
(Note that (A + B)2 = A2 + AB + BA + B 2 .)

(d) True. Since A and B are symmetric, (A − B)T = AT − B T = A − B.

   
1 0 0 −1
(e) False. For example, let A = and B = .
0 0 −1 0
 
0 1
(f) False. For example, let A = .
0 0

(g) True. The (i, i)-entry of AAT is equal to

n
X
ai1 ai1 + ai2 ai2 + · · · + ain ain = a2ik .
i=1

So AAT = 0 implies that aik = 0 for all i and k, i.e. A = 0.

4 −6 −6 −12 6
     
6 8 0 0
25. (a) A2 = 0 10 6 , −6A =  0 −18 −6 , 8I = 0 8 0.
0 6 10 0 −6 −18 0 0 8
It is easy to be checked that A2 − 6A + 8I = 0.
(b) By (a), A2 = 6A − 8I. Since

= 8 A(6I − A) = 81 (6A − A2 ) = 18 (6A − 6A + 8I) = I,

1  1
A 8
(6I − A)

A−1 = 18 (6I − A).

26. (a) Since (I − A)(I + A) = I − A2 = I, I − A is invertible and (I − A)−1 =

I + A.
(b) Since (I − A)(I + A + A2 ) = I − A3 = I, I − A is invertible and
(I − A)−1 = I + A + A2 .
(c) Yes. In general, we have (I − A)(I + A + · · · + An−1 ) = I − An . So if
An = 0, then I − A is invertible and its inverse is I + A + · · · + An−1 .
     
0 0 1 0 1 0
27. (a) For example, , and .
0 0 0 1 0 0
(b) Since (I + A) 12 (2I − A) = 21 (I + A)(2I − A) = 21 (2I + A − A2 ) = I,
 

I + A is invertible and its inverse is 21 (2I − A).

11
    
1 0 −1 0
28. (a) False. For example, let A = and B = .
0 1 0 −1
   
1 0 0 0
(b) False. For example, let A = and B = .
0 0 0 1

29. Since we cannot assume that A−1 + B −1 is invertible at the beginning, we

cannot prove (A + B)−1 = A−1 (A−1 + B −1 )−1 B −1 directly. Instead, we first
prove the equivalent form

−1
A(A + B)−1 B = A−1 + B −1 .

Since A, B and A + B are invertible, A(A + B)−1 B is invertible and

−1
A(A + B)−1 B = B −1 (A + B)A−1 = (B −1 A + I)A−1 = A−1 + B −1 .

−1
Hence A−1 + B −1 is invertible and A(A + B)−1 B = A−1 + B −1 which
implies (A + B)−1 = A−1 (A−1 + B −1 )−1 B −1 .

30. (a) (cA)( 1c A−1 ) = (c 1c )AA−1 = I and ( 1c A−1 )(cA) = ( 1c c)A−1 A = I.

So cA is invertible and (cA)−1 = 1c A−1 .

(b) A−1 A = I and AA−1 = I.

So A−1 is invertible and (A−1 )−1 = A.

(c) (AB)(B −1 A−1 ) = ABB −1 A−1 = AIA−1 = AA−1 = I and

(B −1 A−1 )(AB) = B −1 A−1 AB = B −1 IB = B −1 B = I.
So AB is invertible and (AB)−1 = B −1 A−1 .

31. (a) Ak = (P DP −1 )(P DP −1 ) · · · (P DP −1 ) = P D k P −1 .

| {z }
k times
 −1  
1 1 2 −1
(b) =
1 2 −1 1
   
1 1 −2 0 2 −1
It is easy to be checked that A = .
1 2 0 3 −1 1
Hence

(−2)10 0
  11
2 − 310 310 − 210
   
10 1 1 2 −1
A = = .
1 2 0 310 −1 1 211 − 2 · 310 2 · 310 − 210

12
 5 −2 6 0 R2 + 52 R1 5 −2 6 0 R1 + 10R2 5 0 60 10
     
32. A =
−2 1 3 1 −→ 0 15 275
1 −→ 0 51 27
5
1
1
   
5 R1 1 0 12 2 5R2 1 0 12 2
=R
−→ 0 15 27 5
1 −→ 0 1 27 5
  1      
1 0 5
0 1 10 1 0 1 0 12 2
So 2 A= and hence
0 5 0 1 0 1 5
1 0 1 27 5
     
1 0 1 −10 5 0 1 0 1 0 12 2
A= .
− 25 1 0 1 0 1 0 51 0 1 27 5
    
1 0 0 0 1 0 0 0 1 −1 0 0 1 0 0 0
 1
0 1 0 0 0 0 0 1 0 1 0 0 0 2 0 0
   
33. (a) A =  .
3 0 1 0 0 0 1 0 0 0 1 0 0 0 1 0
  
0 0 0 1 0 1 0 0 0 0 0 1 0 0 0 1
    
1 0 0 0 1 1 0 0 1 0 0 0 1 0 0 0
0 2 0 0 0 1 0 0 0 0 0 1  0 1 0 0
    
(b) A−1 = .
0 0 1 0 0 0 1 0 0 0 1 0 −3 0 1 0
  
0 0 0 1 0 0 0 1 0 1 0 0 0 0 0 1

R1 − R3 R1 ↔ R3 R3 + 2R2 2R3
34. (a) B −→ −→ −→ −→ A

(b) Yes. Since B = E4 −1 E3 −1 E2 −1 E1 −1 A, if A is invertible, B is invertible.

35. Since E1 E2 A = E3 E4 B, we have E4 −1 E3 −1 E1 E2 A = B. Thus B can be

obtained from A by the following elementary row operations.

R2 − R1 2R3 R1 − 2R3 R1 ↔ R4
A −→ −→ −→ −→ B

36. (a) Since ac ̸= 0, we have a ̸= 0 and c ̸= 0.

1  1 R2  b  R1 − ab R2
a R1
1 ab c

1 a
A −→ −→ −→ I 2
0 c 0 1
   b
a 0 1 0 1 a
So A = .
0 1 0 c 0 1

13
 R3 + R1 R3 − R2 R2 − 3R3 R1 − 2R2
(b) B −→ −→ −→ −→ I3
    
1 0 0 1 0 0 1 0 0 1 2 0
So B =  0 1 0   0 1 0 0 1 3 0 1 0.
−1 0 1 0 1 1 0 0 1 0 0 1

1 1 1
 
0 1 1 1 0 0 Gauss-Jordan 1 0 0 − 2 2
 
2
37. (a)  1 0 1 0 1 0  −→  0 1 0 12 − 12 21 
 
1 1 0 0 0 1 Elimination 0 0 1 12 1
− 12
2
 1 1 1

−2 2 2
Hence the matrix is invertible and its inverse is  12 − 12 12 .
 
1 1
2 2
− 12

−1 3 −4 1 0 0 −1 3 −4 1 0 0
   
Gaussian
(b)  2 4 1 0 1 0  −→  0 10 −7 2 1 0 
−4 2 −9 0 0 1 Elimination 0 0 0 −2 1 1
Hence thematrix is not invertible.
1 1 1
 
1 1 0 0 Gauss-Jordan 1 0 0 2 − 6 − 3
 
2 0
(c) 0 2 −1 0 1 0  −→  0 1 0 0 32 1 
 
3 
0 −1 2 0 0 1 Elimination 0 0 1 0 31 2
3
1 1 1

2
−6 −3
Hence the matrix is invertible and its inverse is  0 32 1 
.

3 
1 2
0 3 3
 1

1 0 0 0 1 0 − 14
 
1 0 0 1 1 0 0 0 4
 Gauss-Jordan   0 1 0 0 − 12 12 − 81 18 

 1 2 0 0 0 1 0 0

(d)  −→
  
 1 2 3 0 0 0 1

0  0 0 1 0 0 −1 1 0 
Elimination  3 3 
1 2 3 4 0 0 0 1 0 0 0 1 0 0 −4 4 1 1

 1 1

1 0 4
− 4
 1 1
− 2 2 − 18 18 

Hence the matrix is invertible and its inverse is 
 0 −1 1
.
 3 3
0 

1 1
0 0 −4 4
   
1 1 1 1 1 0 0 0 1 1 1 1 1 0 0 0
 Gaussian 
 −1 2 6 3 0 1 0 0   0 3 7 4 1 1 0 0 
 
(e)  −→
 1 −2 −6 −4 0 0 1 0   0 0 0 −1 0 1 1 0 
  
Elimination
1 1 1 0 0 0 0 1 0 0 0 0 −1 −1 −1 1

Hence the matrix is not invertible.

14
    
1 3 1 1 1 0 0 0 1 0 0 0 −4 3 0 −1
 Gauss-Jordan 
2 5 2 2 0 1 0 0 0 1 0 0 2 −1 0 0
 
(f)  −→
   
1 3 8 9 0 0 1 0 0 0 1 0 −7 0 −1 8
  
   
Elimination
1 3 2 2 0 0 0 1 0 0 0 1 6 0 1 −7
 
−4 3 0 −1
2 −1 0 0
 
Hence the matrix is invertible and its inverse is  .
−7 0 −1 8 
6 0 1 −7

4 −1 −1
   
2 1 1
1
38. The inverse of 0 1 2 is −2 −3 4 . So
7
1 3 2 1 5 −2

4 −1 −1 5 11 12 −5
   
2 3 4 1
1 1
X = −2 −3 4  1 0 3 7 = 1 −2 −13 −15 .
7 7
1 5 −2 2 1 1 2 3 1 17 32

39. (a) Let x1 , x2 , x3 denote the number of chairs of type A, B, C manufactured

respectively. We have the linear system

 4x1 + 4x2 + 3x3 = 260
x2 + 2x3 = 60

2x1 + 4x2 + 5x3 = 240,
or     
4 4 3 x1 260
0 1 2 x2  =  60  .
2 4 5 x3 240
 3
4 − 52

2
The inverse of the data matrix is −2 −7 4  and hence
1 4 −2
   3
4 − 52
   
x1 2
260 30
x2  = −2 −7 4   60  = 20 .
x3 1 4 −2 240 20

That is, 30 chairs of type A, 20 chairs of type B and 20 chairs of type C

should be manufactured.
(b) Since 10 × (the (3,1)-entry of the inverse of the data matrix) = 10, the
number of chairs of type C is increased by 10.

15
  
1 0 a
 
1 0 a
R3 − aR1 R2 ↔ R3 R3 − aR2  2 
40.  0 a 1   0 1 −a 
−→ −→ −→
a 1 0 0 0 1 + a3
1 −a a2
 
1 
The matrix is invertible if and only if a ̸= −1. The inverse is −a a2 1 .
1 + a3
a2 1 −a

   
1 1 1 R2 − aR1 1 1 1 R3 − (b + a)R2
41. (a)  a b c  −→  0 b−a c−a  −→
2 2 2 2 2 2 2 2
a b c R3 − a R1 0 b −a c −a
 
1 1 1
 0 b−a c−a 
0 0 (c − a)(c − b)
The homogeneous linear system has nontrivial solution if and only if (b −
a) = 0 or (c − a)(c − b) = 0, i.e. a = b or a = c or b = c.
(b) The matrix is invertible if and only if the homogeneous system in (a) has
only the trivial solution, i.e. a ̸= b and a ̸= c and b ̸= c.

42. Assume AB is invertible. Let C be the inverse of AB. Then (AB)C = I and
hence A(BC) = I. By Theorem 2.4.12, A is invertible which contradicts that
A is singular.
Assume BA is invertible. Let D be the inverse of BA. Then D(BA) = I and
hence (DB)A = I. By Theorem 2.4.12, A is invertible which contradicts that
A is singular.
 
R
43. Suppose A = Ek · · · E1 0 · · · 0 for some elementary matrices E1 , . . . , Ek . Let
 
0
 .. 
b = Ek · · · E1  . . (This is only an example of many possible choices of b.)
 
0
1
Then  
0
   .. 
R
Ax = b ⇔ x = .
 
0 ··· 0 0
1
which is inconsistent, see Remark 1.4.8.1.

16
  
44. (a) R
A is row equivalent to 0 · · · 0
 
R
⇒ A = Ek · · · E1 0 · · · 0 for some elementary matrices E1 , . . . , Ek .
 
R
⇒ AB = Ek · · · E1 0 · · · 0 B for some elementary matrices E1 , . . . , Ek .
     
R RB RB
⇒ AB is row equivalent to 0 · · · 0 B = = 0 ··· 0 .
(0 · · · 0)B
By Remark 2.4.10, AB is singular.
(b) Since a row-echelon form of A can have at most n non-zero rows and
m > n, a row-echelon form of A must have a zero row. By (a), AB
cannot be invertible.
 
  1 0  
1 0 0 1 0
(c) For example, let A = and B = 0 1. Then AB =
0 1 0 0 1
0 0
is invertible.

45. For i = 1, 2, . . . , n, let Ei be the elementary matrix associated with the row
operation Ri (and the column operation Ci ). Since A is reduced to I by the
row operations R1 , R2 , . . . . Rn , we have

En · · · E2 E1 A = I.

By Theorem 2.4.12, A is invertible and A−1 = En · · · E2 E1 . So

AEn · · · E2 E1 = I.
Cn Cn−1 · · · C1
Thus A −→ −→ · · · −→ I .

46. If B = EA where E is an elementary matrix, then B −1 = A−1 E −1 . Note that

E −1 is also an elementary matrix, see Discussion 2.4.2. By Discussion 2.4.15,
post-multiplying an elementary matrix to a matrix A is equivalent to do an
elementary column operation on A.

(a) If B is obtained from A by multiplying a constant c to the ith row, then

B −1 can be obtained from A−1 by multiplying 1c to the ith column.
(b) If B is obtained from A by interchanging the ith row and the jth row,
then B −1 can be obtained from A−1 by interchanging the ith column and
the jth column.
(c) If B is obtained from A by adding c times of the ith row to the jth row,
then B −1 can be obtained from A−1 by adding −c times of the jth column
to the ith column.

17
 1 1 1 0
47. (a) (i) 0 − + =2
1 0 1 1
   
0 1 1 1 0 1
R1 ↔ R2 R3 − R1 R3 − R2 
(ii) 1 0 1 0 1 1
−→ −→ −→
1 1 0 0 0 −2
0 1 1 1 0 1
So 1 0 1 = − 0 1 1 = 2.
1 1 0 0 0 −2
 T
0 1 1 1 1 0
 1 0 −
 1 0 1 1 

−1
 
  1 1
1 1 1 0 1  = 1  1 −1 1 
0 1
(iii)  − −
2 1 0 1 0 1 1
 2
1 1 −1
 
 1 1 0 1 0 1 
−
0 1 1 1 1 0

4 1 2 1 2 4
(b) (i) (−1) −3 + (−4) =0
2 −9 −4 −9 −4 2
−1 3 −4 −1 3 −4
   
R2 + 2R1 R3 − 4R1 R3 + R2 
(ii)  2 4 1  0 10 −7
−→ −→ −→
−4 2 −9 0 0 0
−1 3 −4 −1 3 −4
So 2 4 1 = 0 10 −7 = 0.
−4 2 −9 0 0 0
(iii) The matrix is not invertible.

2 −1
(c) (i) 2 −0+0=6
−1 2
   
2 0 1 1 2 0 1
R3 + 2 R2 
(ii) 0 2 −1 0 2 −1
−→
0 −1 2 0 0 32
2 0 1 2 0 1
So 0 2 −1 = 0 2 −1 = 6.
0 −1 2 0 0 32

18
 T
2 −1 0 −1

0 2
 −1 2 −
 0 2 0 −1 
−1 −2
 
  3
1 0 1 2 1 2 0  = 1 0 4
(iii)  − − 2
6  −1 2
 0 2 0 −1  6


 0 2 4
 0 1 2 1 2 0 
−
2 −1 0 −1 0 2

2 0 0 1 2 0    
3 0 2 3 1 2
(d) (i) 2 3 0 −0+0− 1 2 3 = 2 −0+0 − −2 +0
3 4 2 3 1 2
2 3 4 1 2 3
= 24
   
1 0 0 1 R2 − R1 1 0 0 1
1 2 0 0 R3 − R1 R3 − R2 R4 − R3 0 2 0 −1
   
(ii) 
1 2 3 0 −→ −→ −→ 0 0 3 0
  
1 2 3 4 R4 − R1 R4 − R2 0 0 0 4
1 0 0 1 1 0 0 1
1 2 0 0 0 2 0 −1
So = = 24.
1 2 3 0 0 0 3 0
1 2 3 4 0 0 0 4
 T
2 0 0 1 0 0 1 2 0 1 2 0

 2 3 0 − 1 3 0 1 2 0 − 1 2 3

 2 3 4 1 3 4 1 2 4 1 2 3
 
 0 0 1 1 0 1 1 0 1 1 0 0 
 
− 2 3 0 1 3 0 − 1 2 0 1 2 3 
 
2 3 4 1 3 4 1 2 4 1 2 3 
 
1 
(iii)  
24  0 0

1 1 0 1 1 0 1 1 0 0

− 1 0 0 − 1 2 0
 
 2 0 0 1 2 0
 
 2 3 4 1 3 4 1 2 4 1 2 3
 
 
 0 0 1 1 0 1 1 0 1 1 0 0 
 
− 2 0 0 1 0 0 − 1 2 0 1 2 0 
2 3 0 1 3 0 1 2 0 1 2 3
 
24 0 6 −6
1 −12

12 −3 3 

=
24  0 −8 8 0
 
0 0 −6 6

48. (a) x = 1, y = −1.

19
 (b) x = 21 , y = 21 , z = 32 .
(c) x = 1, y = 0, z = −2.
(d) w = 0, x = 0, y = 0, z = −1.

49. (a) abc

(b) A is invertible if and only if a ̸= 0, b ̸= 0 and c ̸= 0.
1
− a1 0

a
A−1 =  0 1b − 1b .
1
0 0 c

50. (a) det(C) = 0.

(b) Since C is singular, by Theorem 2.4.14, AC is singular. Thus by Theorem
2.4.7, ACx = 0 has infinitely many solutions.

51. (a) Since det(A) = (λ − 2)(λ + 4) + 5 = (λ + 3)(λ − 1), det(A) = 0 if and only
if λ = −3 or 1.
(b) Since det(A) = (λ − 1)(λ2 − λ − 6) = (λ − 4)(λ − 3)(λ + 2), det(A) = 0 if
and only if λ = 4, 3 or −2.
   
1 λ λ λ 1 λ λ λ
 2 λ λ λ  R2 − R1  1 0 0 0
   
(c) 
λ + 1 1 2 0  −→ λ + 1 1 2 0 
  
4 0 1 2λ 4 0 1 2λ
1 λ λ λ
λ λ λ
1 0 0 0
det(A) = = − 1 2 0 = 2λ2 + λ.
λ+1 1 2 0
0 1 2λ
4 0 1 2
Hence det(A) = 0 if and only if λ = 0 or − 21 .
   
1 1 2 3 1 1 2 3
2
1 2 − λ 2 3  R2 − R1 R4 − R3 0 1 − λ2 0 0 
   
(d) 
2 3 1 5  −→ −→ 2 3 1 5 
  
2
2 3 1 9−λ 0 0 0 4 − λ2
1 1 2 3
1 2 3
0 1 − λ2 0 0 2
det(A) = = (1 − λ ) 2 1 5
2 3 1 5
0 0 4 − λ2
0 0 0 4 − λ2
1 2
= (1 − λ2 )(4 − λ2 ) = −3(1 − λ2 )(4 − λ2 ).
2 1

20
 Hence det(A) = 0 if and only if λ = ±1 or ±2.

1 a a2 R2 − R1 1 a2
   
a
52. 1 b b2  −→ 0 b − a b2 − a2 
1 c c2 R3 − R1 0 c − a c2 − a2
1 a a2 1 a a2
So 1 b b2 = 0 b − a b2 − a2 = (b − a)(c2 − a2 ) − (c − a)(b2 − a2 )
1 c c2 0 c − a c 2 − a2
= (b − a)(c − a)(c − b).
1 1 1
53. (a) 34 · 9 = 729 (b) 9
(c) 34 · 9
=9 (d) 729

R4 + R2 R2 ↔ R3 R1 − R2 3R2 R3 + 2R1
54. (a) B −→ −→ −→ −→ −→ A
(b) det(A) = 1 · 2 · 3 · (−1) = −6 and hence det(B) = (−1) · 13 · det(A) = 2.
1
R1 − 2R3 R3 − 3R2 2 R2 R1 ↔ R2
55. (a) B −→ −→ −→ −→ A
(b) det(B) = (−1) · 2 · det(A) = −8

56. det(A) = aei + bf g + cdh − af h − bdi − ceg.

If all a, b, c, d, e, f, g, h, i are 1, then det(A) = 0.
Suppose at least one of a, b, c, d, e, f, g, h, i is 0, say a = 0 (other cases are
similar). Then det(A) = bf g + cdh − bdi − ceg. As b, c, d, e, f, g, h, i can only
be 0 and 1, −2 ≤ det(A) ≤ 2.
1 1 0 0 1 1
Note that 0 1 1 = 2 and 1 1 0 = −2.
1 0 1 1 0 1
The maximum possible value of det(A) is 2 and the minimum is −2.

57. (a) Since AAT = I and det(A) = det(AT ), we have det(A)2 = det(I) = 1.
Thus det(A) = ±1.
 
a b
(b) Let A = . Since A is orthogonal, AT = A−1 , i.e.
c d
     
a c 1 d −b d −b
= = .
b d det(A) −c a −c a
So a = d and b = −c. Furthermore, det(A) = 1 implies a2 + c2 = ad − bc =
1. Let a = cos(θ) and c = sin(θ). Then
   
a −c cos(θ) − sin(θ)
A= = .
c a sin(θ) cos(θ)

21
 (c) Similar to (b) except now a = −d and b = c.
 
a b
58. (a) Let A be a 2 × 2 matrix with two identical rows, say, A = . Then
a b
det(A) = ab − ab = 0.
Assume that the determinant of any k ×k matrices with two identical rows
is zero where k ≥ 2.
Let A be a (k + 1) × (k + 1) matrices with two identical rows, say, the
ith and jth row of A are identical. Take m = 1, 2, . . . , k + 1 such that
m ̸= i, j. Then by Theorem 2.5.6,

det(A) = am1 Am1 + am2 Am2 + · · · + ai,k+1 Am,k+1

where Amr = (−1)m+r det(Mmr ). Each Mmr is a k × k matrix obtained

from A by deleting the mth row and the rth column of A. Since the
ith and jth row of A are identical, the corresponding rows of Mmr are
identical. By the inductive assumption, det(Mmr ) = 0, i.e. Amr = 0, for
every r. This means det(A) = 0.
By mathematical induction, the determinant of any square matrix with
two identical row is zero.

(b) If A is a square matrix with two identical columns, then AT has two
identical rows. By (a), det(AT ) = 0. So det(A) = det(AT ) = 0.

59. Since Theorem 2.5.15.3 has been proved, we can use it in the following proofs.

(a) Let A = (aij )n×n . Suppose B is obtained from A by multiplying the mth
row of A by k. Observe that for all j, the (m, j)-cofactor of B is the equal
to the (m, j)-cofactor of A; and the (m, j)-entry of B is kamj . Thus by
Theorem 2.5.6,

det(B) = kai1 Ai1 + kai2 Ai2 + · · · + kai,n Ai,n

= k(ai1 Ai1 + ai2 Ai2 + · · · + ai,n Ai,n ) = k det(A).

(b) Suppose B is obtained from A by interchanging the ith and jth rows of
A. Observe that

Ri + Rj Rj − Ri Ri + Rj −Rj
A −→ −→ −→ −→ B.

By (a) and Theorem 2.5.15.3, det(B) = − det(A).

22
 (c) (i) Suppose E is the elementary matrix defined in Discussion 2.4.2.1. Note
that det(E) = k. Since EA can be obtained from A by multiplying
the ith row by k, by (a), det(EA) = k det(A) = det(E) det(A).
(ii) Suppose E is the elementary matrix defined in Discussion 2.4.2.2.
Note that E can be obtained form I by interchanging the ith and
jth rows of I. By (b), det(E) = − det(I) = −1. Since EA can be
obtained from A by interchanging the ith and jth rows of A, by (b)
again, det(EA) = − det(A) = det(E) det(A).
(iii) Suppose E is the elementary matrix defined in Discussion 2.4.2.3.
Note that det(E) = 1. Since EA can be obtained from A by adding
k times of the ith row of A to the jth row, by Theorem 2.5.15.3,
det(EA) = det(A) = det(E) det(A).

h i h i
1 1
60. (a) A det(A)
adj(A) =I ⇒ det(A)
A adj(A) = I
So adj(A) is invertible.
(b) det(adj(A)) = det(A)n−1 and adj(A)−1 = 1
det(A)
A.
(c) adj(A)−1 = 1
det(adj(A))
adj(adj(A)) ⇒ adj(adj(A)) = det(A)n−2 A
If det(A) = 1, then adj(adj(A)) = A.

61. (a) False. For example, let A = I 2 and B = −I 2 .

(b) True. det(A + I) = det((A + I)T ) = det(AT + I).
(c) True. Since det(A) = det(P ) det(B) det(P −1 ) and det(P ) det(P −1 ) = 1,
det(A) = det(B)
(d) False. For example, let A = I 2 and B = C = −I 2 .

23