MATH 0047: Advanced Linear Algebra

Isidoros Strouthos University College London 2023-2024

Sample Questions: Solutions

1. (a) A system of real linear equations is a finite set of, m say, real linear equations in a finite
number of unknowns, n say:

a11 x1 + · · · + a1n xn = b1
.. .
. = ..
am1 x1 + · · · + amn xn = bm

where, for each 1 ≤ i ≤ m, 1 ≤ j ≤ n: aij is a real number and bi is a real number, and where
x1 , · · · , xn are the n unknowns of the system of real linear equations.

(b) A system of real linear equations has no solutions, or one (unique) solution, or infinitely many
solutions.

(c) Set a1 = 4, a2 = 0, a3 = −3 and b1 = 0, b2 = 2, b3 = 5.

Then, for any real number λ, we can find a solution of the form

x1 = λa1 + (1 − λ)b1 , x2 = λa2 + (1 − λ)b2 , x3 = λa3 + (1 − λ)b3

To find one solution, we could, for example, set λ = 12 , to obtain:

−3 5
x1 = 21 a1 + 21 b1 = 2+0 = 2 , x2 = 12 a2 + 12 b2 = 0+1 = 1 , x3 = 12 a3 + 21 b3 = 2
+2 =1

For another solution, we could, for example, set λ = 2, to obtain:

x1 = 2a1 −b1 = 8−0 = 8 , x2 = 2a2 −b2 = 0−2 = −2 , x3 = 2a3 −b3 = −6−5 = −11

So, we may, for example, give the following two solutions to the original system of real linear
equations:

x1 = 2 , x2 = 1 , x3 = 1 and x1 = 8 , x2 = −2 , x3 = −11

1
(d) The system of equations may be written as the following matrix equation:
    
1 −2 0 1 x1 7
1 0 2 0 x2  −2
3 2 3 1 x3  =  10 
    

0 1 1 1 x4 3

The augmented matrix for this system of equations is

 
1 −2 0 1 | 7
1 0 2 0 | −2
 
3 2 3 1 | 10 
0 1 1 1 | 3

2
Let us first use elementary row operations to find the reduced row echelon form of this matrix:
   
1 −2 0 1 | 7 1 −2 0 1 | 7
1 0 2 0 | −2 R2 →R2 −R1 0 2 2 −1 | −9

3 2 3 1 | 10 
 −− −− − −−→ 
3 2 3 1 | 10 


0 1 1 1 | 3 0 1 1 1 | 3
 
1 −2 0 1 | 7
R3 →R3 −3R1 0 2 2 −1 | −9 
−− −−−−−→ 
0 8 3 −2 | −11


0 1 1 1 | 3
 
1 −2 0 1 | 7
R ↔R4 0 1 1 1 | 3 
−−2−−→ 
0 8 3 −2 | −11


0 2 2 −1 | −9
 
1 0 2 3 | 13
R1 →R1 +2R2 0 1 1 1 | 3 
−− −−−−−→ 
0 8 3 −2 | −11


0 2 2 −1 | −9
 
1 0 2 3 | 13
R3 →R3 −8R2 0 1 1 1 | 3 
−− −−−−−→ 
0 0 −5 −10 | −35


0 2 2 −1 | −9
 
1 0 2 3 | 13
R4 →R4 −2R2 0 1 1 1 | 3 
−− −−−−−→ 
0 0 −5 −10 | −35


0 0 0 −3 | −15
 
1 0 2 3 | 13
R3 → −1
5
R3 0 1 1 1 | 3 
−−−−− −→ 
0 0 1 2 | 7 


0 0 0 −3 | −15
 
1 0 0 −1 | −1
R1 →R1 −2R3 0 1 1 1 | 3 
−− −−−−−→ 
0 0 1 2 | 7 


0 0 0 −3 | −15
 
1 0 0 −1 | −1
R →R −R3 0 1 0 −1 | −4 
−−2−−−2−−→ 
0 0 1 2 | 7 


0 0 0 −3 | −15
 
1 0 0 −1 | −1
R4 → −1
3
R4 0 1 0 −1 | −4
−−−−− −→ 
0 0 1 2 | 7 

3
0 0 0 1 | 5
  
1 0 0 0 | 4
R →R +R 0 1 0 −1 | −4
−−1−−−1−−→
4  
0 0 1 2 | 7
0 0 0 1 | 5
 
1 0 0 0 | 4
R →R +R 0 1 0 0 | 1
−−2−−−2−−→
4  
0 0 1 2 | 7
0 0 0 1 | 5
 
1 0 0 0 | 4
R →R −2R
3 3 4
0 1 0 0 | 1
−− −−− −−→ 
0 0 1 0 |

−3
0 0 0 1 | 5

So, the original system of real linear equations is equivalent to the following system:

x1 =4
x2 =1
x3 = −3
x4 =5

and this is the unique solution to the given system of real linear equations.

4
2. (a) A system of real linear equations is a finite set of, m say, real linear equations in a finite
number of unknowns, n say:

a11 x1 + · · · + a1n xn = b1
.. .
. = ..
am1 x1 + · · · + amn xn = bm

where, for each 1 ≤ i ≤ m, 1 ≤ j ≤ n: aij is a real number and bi is a real number, and where
x1 , · · · , xn are the n unknowns of the system of real linear equations.

(b) A system of real linear equations is inconsistent if no (real) solution to the system of real linear
equations exists.

(c) (i) The following is an example of a system, of two real linear equations in two unknowns,
which is inconsistent:

2x + y = 1
10x + 5y = 3

(ii) The following is an example of a system, of two real linear equations in two unknowns,
which has a unique solution:

2x + y = 1
2x − y = 7

(iii) The following is an example of a system, of two real linear equations in two unknowns,
which has infinitely many solutions:

2x + y = 1
10x + 5y = 5

(d) The system of equations may be written as the following matrix equation:
    
1 −2 0 1 x1 0
1 0 2 1  x2  6
3 2 8 −2 x3  = −1
    

0 4 4 −5 x4 2

5
The augmented matrix for this system of equations is
 
1 −2 0 1 | 0
1 0 2 1 | 6 
 
3 2 8 −2 | −1
0 4 4 −5 | 2

Let us first use elementary row operations to find the reduced row echelon form of this matrix:
   
1 −2 0 1 | 0 1 −2 0 1 | 0
1 0 2 1 | 6  R2 →R2 −R1 0 2 2 0 | 6 

3 2 8 −2 | −1
 − −− − − −−→ 
3 2 8 −2 | −1


0 4 4 −5 | 2 0 4 4 −5 | 2
 
1 −2 0 1 | 0
R3 →R3 −3R1 0 2 2 0 | 6 
−− −−−−−→ 
0 8 8 −5 | −1


0 4 4 −5 | 2
 
1 −2 0 1 | 0
R2 → 1 R2 0 1 1 0 | 3 
−−−−2−→ 
0 8 8 −5 | −1


0 4 4 −5 | 2
 
1 0 2 1 | 6
R1 →R1 +2R2 0 1 1 0 | 3 
−− −−−−−→ 
0 8 8 −5 | −1


0 4 4 −5 | 2
 
1 0 2 1 | 6
R3 →R3 −8R2 0 1 1 0 | 3 
−− −−−−−→ 
0 0 0 −5 | −25


0 4 4 −5 | 2
 
1 0 2 1 | 6
R4 →R4 −4R2 0 1 1 0 | 3 
−− −−−−−→ 
0 0 0 −5 | −25


0 0 0 −5 | −10
 
1 0 2 1 | 6
R3 → −1
5
R3 0 1 1 0 | 3 
−−−−− −→ 
0 0 0 1 | 5 


0 0 0 −5 | −10

6
  
1 0 2 0 | 1
R →R −R 0 1 1 0 | 3 
−−1−−−1−−→
3  
0 0 0 1 | 5 
0 0 0 −5 | −10
 
1 0 2 0 | 1
R →R +5R
4 4 3
0 1 1 0 | 3
−− −−− −−→  
0 0 0 1 | 5
0 0 0 0 | 15
 
1 0 2 0 | 1
R4 → 1
R4 0 1 1 0 | 3
−−−−15
−−→  
0 0 0 1 | 5
0 0 0 0 | 1
 
1 0 2 0 | 0
R →R −R 0 1 1 0 | 3
−−1−−−1−−→
4  
0 0 0 1 | 5
0 0 0 0 | 1
 
1 0 2 0 | 0
R →R −3R
2 2 4
0 1 1 0 | 0
−− −−− −−→  
0 0 0 1 | 5
0 0 0 0 | 1
 
1 0 2 0 | 0
R →R −5R
3 3 4
0 1 1 0 | 0
−− −−− −−→  
0 0 0 1 | 0
0 0 0 0 | 1

So, the original system of real linear equations is equivalent to the following system:

x1 + 2x3 = 0
x2 + x3 = 0
x4 = 0
0=1

The fourth equation in this system can “never” hold, so the given system is inconsistent.

7
3. (a) The (three) types of elementary row operations that may be applied to a real matrix are:
1) Multiply (each entry of) a row of the matrix by a non-zero real number.
(If we multiply row i of a matrix by a non-zero real number λ, we write: Ri → λRi .)
2) Add or subtract a multiple of one row of the matrix to or from another row of the matrix.
(If, to row i of a matrix, we add λ times row j of the matrix, we write: Ri → Ri + λRj .)
3) Permute (or interchange) two rows of the matrix.
(If we permute rows i and j of a matrix, we write: Ri ↔ Rj .)

(b) A real matrix is in reduced row echelon form if it satisfies the following four conditions:
1) If there are any zero rows (rows consisting only of zero entries), then they appear at the
bottom of the matrix.
2) For each non-zero row, the first (leftmost) non-zero entry is a 1 (known as the leading 1
of that row).
3) If two successive rows are non-zero, then the lower row starts with more zeros than the
row above, i.e. the leading 1 of the lower row is (below and) to the right of the leading 1
of the row above.
4) Each column that contains a leading 1 has zero entries everywhere else.

(c) The system of equations may be written as the following matrix equation:
    
1 −2 0 1 x1 0
 1 0 2 0   x2  1
3 2 8 −2 x3  = −1
    

0 1 1 1 x4 8

The augmented matrix for this system of equations is

 
1 −2 0 1 | 0
1 0 2 0 | 1 
 
3 2 8 −2 | −1
0 1 1 1 | 8

8
Let us first use elementary row operations to find the reduced row echelon form of this matrix:
   
1 −2 0 1 | 0 1 −2 0 1 | 0
1 0 2 0 | 1  R2 →R2 −R1 0 2 2 −1 | 1 

3 2 8 −2 | −1
 −− − −− − −
→ 
3 2 8 −2 | −1


0 1 1 1 | 8 0 1 1 1 | 8
 
1 −2 0 1 | 0
R3 →R3 −3R1 0 2 2 −1 | 1 
−− −−−−−→ 
0 8 8 −5 | −1


0 1 1 1 | 8
 
1 −2 0 1 | 0
R ↔R4 0 1 1 1 | 8 
−−2−−→ 
0 8 8 −5 | −1


0 2 2 −1 | 1
 
1 0 2 3 | 16
R1 →R1 +2R2 0 1 1 1 | 8 
−− −−−−−→ 
0 8 8 −5 | −1


0 2 2 −1 | 1
 
1 0 2 3 | 16
R3 →R3 −8R2 0 1 1 1 | 8 
−− −−−−−→ 
0 0 0 −13 | −65


0 2 2 −1 | 1
 
1 0 2 3 | 16
R4 →R4 −2R2 0 1 1 1 | 8 
−− −−−−−→ 
0 0 0 −13 | −65


0 0 0 −3 | −15
 
1 0 2 3 | 16
R3 → −1 R3 0 1 1 1 | 8 
−−−−13 −−→ 
0 0 0 1 | 5 


0 0 0 −3 | −15
 
1 0 2 0 | 1
R1 →R1 −3R3 0 1 1 1 | 8 
−− −−−−−→ 
0 0 0 1 | 5 


0 0 0 −3 | −15
 
1 0 2 0 | 1
R →R −R3 0 1 1 0 | 3 
−−2−−−2−−→ 
0 0 0 1 | 5 


0 0 0 −3 | −15
 
1 0 2 0 | 1
R4 →R4 +3R3 0 1 1 0 | 3
−− −−−−−→ 
0 0 0 1 | 5

9
0 0 0 0 | 0
So, the original linear system of equations is equivalent to the following system:

x1 + 2x3 = 1
x2 + x 3 = 3
x4 = 5
0=0

The fourth equation “always” holds, so we may safely ignore it, and rearrange the first two
equations in order to obtain x1 and x2 in terms of x3 :

x1 = 1 − 2x3
x2 = 3 − x3

There are infinitely many solutions to these equations; we may choose any (real) value for x3 .
Renaming x3 as λ allows us to deduce that the general solution to this system of real linear
equations may be given by:

x1 = 1 − 2λ
x2 =3−λ
x3 =λ
x4 =5

for λ ∈ R.

10
4. (a) Let A be a square matrix, of size n × n say. Then, the matrix A is invertible if there exists an
n × n matrix B such that:
AB = In and BA = In
where In denotes the n × n identity matrix.

(b) (i) We assume that A and B are invertible. So:

There exists a matrix A−1 such that AA−1 = I and A−1 A = I.
There exists a matrix B −1 such that BB −1 = I and B −1 B = I.
Now, consider (AB)(B −1 A−1 ):
(AB)(B −1 A−1 ) =A(BB −1 )A−1
=AA−1 since we assume that BB −1 = I
=I since we assume that AA−1 = I
Similarly:
(B −1 A−1 )(AB) =B −1 (A−1 A)B
=B −1 B since we assume that A−1 A = I
=I since we assume that B −1 B = I
So: (AB)(B −1 A−1 ) = I and (B −1 A−1 )(AB) = I.
Hence, the matrix AB is invertible, with inverse B −1 A−1 , i.e.
(AB)−1 = B −1 A−1

(ii) We seek a matrix C such that (A + I)C = I and C(A + I) = I.

Consider (A + I)(A2 − A + I):
(A + I)(A2 − A + I) =A3 + IA2 − A2 − IA + AI + I 2
=A3 + A2 − A2 − A + A + I
=A3 + I
=I since we assume that A3 = 0
Similarly:
(A2 − A + I)(A + I) =A3 − A2 + A + A2 − A + I
=A3 + I
=I since we assume that A3 = 0
So, we have shown that (A + I)(A2 − A + I) = I and that (A2 − A + I)(A + I) = I.
Hence, the matrix A + I is invertible, with inverse A2 − A + I.

11
(c) We row reduce the augmented matrix ( A | I3 ); at each step, we give the elementary row
operation being performed, as well as the elementary matrix corresponding to it:

   
2 1 1 | 1 0 0 1 0 −3 | 0 1 0
R ↔R
1 0 −3 | 0 1 0 −−1−−→
2
2 1 1 | 1 0 0
P (1,2)
3 2 6 | 0 0 1 3 2 6 | 0 0 1
 
1 0 −3 | 0 1 0
R →R −2R
2 2 1
−− −−− −−→ 0 1 7 | 1 −2 0
E(2,1;−2)
3 2 6 | 0 0 1
 
1 0 −3 | 0 1 0
R →R −3R
3 3 1
−− −−− −−→ 0 1 7 | 1 −2 0
E(3,1;−3)
0 2 15 | 0 −3 1
 
1 0 −3 | 0 1 0
R →R −2R
3 3 2
−− −−− −−→ 0 1 7 | 1 −2 0
E(3,2;−2)
0 0 1 | −2 1 1
 
1 0 0 | −6 4 3
R →R +3R
1 1 3
−− −−− −−→ 0 1 7 | 1 −2 0
E(1,3;3)
0 0 1 | −2 1 1
 
1 0 0 | −6 4 3
R →R −7R
2 2 3
−− −−− −−→ 0 1 0 | 15 −9 −7
E(2,3;−7)
0 0 1 | −2 1 1

(i) At the end of the row reduction, we obtain the identity matrix on the “left half” of the
augmented matrix (as the reduced row echelon form of A). We deduce that the matrix A
is invertible. Its inverse is the matrix on the “right half” of the final augmented matrix:
 
−6 4 3
A−1 =  15 −9 −7
−2 1 1

(ii) We may express A−1 as a product of elementary matrices by writing down the elementary
matrices corresponding to the operations performed during the row reduction:
A−1 = E(2, 3; −7)E(1, 3; 3)E(3, 2; −2)E(3, 1; −3)E(2, 1; −2)P (1, 2)

(iii) We may express A as a product of elementary matrices, by writing down a matrix product
consisting of the inverses of the elementary matrices from part (c)(ii), in the reverse order
of the order in which they appear in the matrix product for A−1 :
A = P (1, 2)E(2, 1; 2)E(3, 1; 3)E(3, 2; 2)E(1, 3; −3)E(2, 3; 7)

12
5. (a) Let A be a square matrix, of size n × n say. Then, the matrix A is invertible if there exists an
n × n matrix B such that:

AB = In and BA = In

where In denotes the n × n identity matrix.

(b) We assume that B and C are both inverse matrices for A. By the definition of an inverse:
AB = I and BA = I (since B is an inverse of A)
AC = I and CA = I (since C is an inverse of A).
Then, we can show that B = C by using the fact that (BA)C = B(AC) (i.e. by using the fact
that matrix multiplication is associative):
We note that:
(BA)C = IC = C using BA = I
Also, B(AC) = BI = B using AC = I
So, using the fact that (BA)C = B(AC), we conclude that B = C, as required.

(c) (i) We row reduce the augmented matrix ( A | I3 ):

   
−3 0 2 | 1 0 0 1 1 1 | 0 0 1
R ↔R
 2 3 4 | 0 1 0 −−1−−→
3
 2 3 4 | 0 1 0
1 1 1 | 0 0 1 −3 0 2 | 1 0 0
 
1 1 1 | 0 0 1
R →R −2R
2 2 1
−− −−− −−→  0 1 2 | 0 1 −2
−3 0 2 | 1 0 0
 
1 1 1 | 0 0 1
R →R +3R
3 3 1
−− −−− −−→ 0 1 2 | 0 1 −2
0 3 5 | 1 0 3
 
1 0 −1 | 0 −1 3
R →R −R
−−1−−−1−−→
2
0 1 2 | 0 1 −2
0 3 5 | 1 0 3
 
1 0 −1 | 0 −1 3
R →R −3R
3 3 2
−− −−− −−→ 0 1 2 | 0 1 −2
0 0 −1 | 1 −3 9
 
1 0 −1 | 0 −1 3
R →−R
−−3−−−→
3
0 1 2 | 0 1 −2
0 0 1 | −1 3 −9

13
  
1 0 0 | −1 2 −6
R →R +R
−−1−−−1−−→
3
0 1 2 | 0 1 −2
0 0 1 | −1 3 −9
 
1 0 0 | −1 2 −6
R →R −2R
2 2 3
−− −−− −−→ 0 1 0 | 2 −5 16 
0 0 1 | −1 3 −9

At the end of the row reduction, we obtain the identity matrix on the “left half” of the
augmented matrix (i.e. the reduced row echelon form of A is the identity matrix). So, we
deduce that the matrix A is invertible. Its inverse is the matrix on the “right half” of the
final augmented matrix:  
−1 2 −6
A−1 =  2 −5 16 
−1 3 −9

(ii) Since we have shown that A is an invertible matrix, we may note that, if Ax = b, then
A−1 (Ax) = A−1 b, i.e. (A−1 A) x = A−1 b, so that:
x = A−1 b
Therefore, the given equation has the unique solution
      
x1 −1 2 −6 −1 3
−1
x = x2 = A b =
   2 −5 16   7  = −5

x3 −1 3 −9 2 4

(d) We seek a matrix C such that (A − I)C = I and C(A − I) = I.

Consider (A − I)(−A2 − A − I):
(A − I)(−A2 − A − I) = − A3 + IA2 − A2 + IA − AI + I 2
= − A3 + A2 − A2 + A − A + I
= − A3 + I
=I since we assume that A3 = 0
Similarly:
(−A2 − A − I)(A − I) = − A3 − A2 − A + A2 + A + I
= − A3 + I
=I since we assume that A3 = 0
So, we have shown that (A − I)(−A2 − A − I) = I and that (−A2 − A − I)(A − I) = I.
Hence, the matrix A − I is invertible, with inverse −A2 − A − I.

14
6. (a) (i) A permutation of the set of numbers {1, · · · , n} is an arrangement of the numbers in some
order, without omissions or repetitions.

(ii) The sign of a permutation σ, written sign(σ) or sgn(σ), is defined as:

sign(σ) = +1 if σ contains an even number of inversions
sign(σ) = −1 if σ contains an odd number of inversions
Equivalently, if σ contains k inversions in total:
sign(σ) = (−1)k

(iii) An elementary product from a square n × n matrix A is a product of n entries of A, such

that there is exactly one entry from each row and each column (i.e. such that each row
and column of A is “represented” exactly once).

(iv) If A is a square n × n matrix, then the determinant of A, written det(A) or |A|, is the
sum of all signed elementary products from A, i.e. the sum of products of the form
sign(σ) A1σ(1) · · · Anσ(n) , taken over all possible permutations of n numbers:
X
det(A) = sign(σ)A1σ(1) · · · Anσ(n)
σ∈Sn

(b) As stated in the question, we may assume that:

(1) every real invertible matrix of size n×n may be expressed as a product of real elementary
matrices of size n × n;
(2) if E is a real elementary matrix of size n × n, and M is a real matrix of size n × n, then
det(EM ) = det(E)det(M ).
Since A is an invertible matrix, we may use assumption (1), from above, in order to deduce
that it is possible to express A as a product of elementary matrices, A = E1 · · · Ek say.
In that case:
det(AB) = det (E1 · · · Ek B)
= det(E1 )det(E2 · · · Ek−1 Ek B) using assumption (2) from above
.. ..
= . .
= det(E1 )det(E2 ) · · · det(Ek−1 )det(Ek )det(B) using assumption (2) from above
= det(E1 )det(E2 ) · · · det(Ek−1 Ek )det(B) using assumption (2) from above
.. ..
= . .
= det(E1 · · · Ek )det(B) using assumption (2) from above
= det(A)det(B)

15
 So, we have shown that det(AB) = det(A)det(B), as required.

(c) (i) Using the definition of the determinant, for a general 3 × 3 matrix A:

A11 A12 A13

A21 A22 A23 = A11 A22 A33 −A12 A21 A33 −A13 A22 A31 −A11 A23 A32 +A12 A23 A31 +A13 A21 A32
A31 A32 A33

we obtain:
−1 2 −2
det(A) = 1 5 1
2 0 3
= (−1)(5)(3) − (2)(1)(3) − (−2)(5)(2) − (−1)(1)(0) + (2)(1)(2) + (−2)(1)(0)
= 3

(ii) We use the entries and corresponding cofactors of the second row of the matrix to find the
determinant of B, as required:

5 −1 2 X3
det(B) = 0 4 3 = (−1)(2+j) B2j |B(2̂, ĵ)|
3 5 4 j=1

−1 2 5 2 5 −1
= −0 +4 −3
5 4 3 4 3 5
= −(0)(−14) + (4)(14) − (3)(28)
= −28

(iii) We will use the following result:

For all square n × n matrices A and B: det(AB) = det(A)det(B).
Using this result, we may deduce that

det(AB) = det(A)det(B) = (3)(−28) = −84

(iv) We will use the following result:

For all square n × n matrices A and B: det(AB) = det(A)det(B).
Using this result, we may deduce that

det(A4 ) = det(A)det(A)det(A)det(A) = det(A)4 = 34 = 81

16
7. (a) (i) An elementary product from a square n × n matrix A is a product of n entries of A, such
that there is exactly one entry from each row and each column (i.e. such that each row
and column of A is “represented” exactly once).

(ii) If A is a square n × n matrix, then the determinant of A, written det(A), or |A|, is the
sum of all signed elementary products from A, i.e. the sum of products of the form
sign(σ) A1σ(1) · · · Anσ(n) , taken over all possible permutations of n numbers:
X
det(A) = sign(σ)A1σ(1) · · · Anσ(n)
σ∈Sn

(b) (i) Each (signed) elementary product from the matrix A must contain one entry from each
row and column, by definition. So, if there is a zero row or column, then each (signed)
elementary product from A must contain a zero, and so must be equal to zero. Therefore,
the determinant of the matrix A, which is the sum of all the signed elementary products
from A, must also be equal to zero, i.e. det(A) = 0, as required.

(ii) Let us show that A11 A22 · · · Ann is the only possibly non-zero elementary product present
in the determinant of A. We try to construct a possibly non-zero elementary product from
A, by selecting a suitable entry from each column of the upper triangular matrix A (one
column at a time):  
A11 A12 A13 · · · A1n
 0 A22 A23 · · · A2n 
 
A =  0
 0 A33 · · · A3n  
 .. .. . . . . .. 
 . . . . . 
0 0 · · · 0 Ann
Firstly, the only possibly non-zero entry we can choose from column 1 is A11 .
Then, the only possibly non-zero entry we can choose from column 2 is the entry A22
(we cannot choose A12 , since we are constructing an elementary product and have
already chosen an entry from row 1, namely A11 ).
Then, similarly, the only possibly non-zero entry we can choose from column 3 is
A33 .
..
.
Finally, the only possibly non-zero entry we can choose from column n is Ann .
So, the only possibly non-zero elementary product from the matrix A is A11 A22 · · · Ann .
This corresponds to the permutation (1, 2, · · · , n), which has 0 inversions and so has
sign +1. Therefore, the elementary product A11 A22 · · · Ann occurs with sign +1 in the
determinant, as the signed elementary product +A11 A22 · · · Ann .

17
 This is the only possibly non-zero signed elementary product, so the determinant of A is
precisely A11 A22 · · · Ann , as required.

(c) (i) We may use the cofactor method, for example, to find the determinant of A. We use the
entries and corresponding cofactors of the first row of the matrix to find det(A):

3 5 2 X3
det(A) = 6 0 3 = (−1)(1+j) B1j |B(1̂, ĵ)|
5 1 2 j=1

0 3 6 3 6 0
= 3 −5 +2
1 2 5 2 5 1
= (3)(−3) − (5)(−3) + (2)(6)
= 18

To find the determinant of B, we will use the result from part (b)(i).
The third row of matrix B is a zero row, so we may deduce that

det(B) = 0

To find the determinant of C, we will use the result from part (b)(ii).
The matrix C is an upper triangular matrix, so that

det(C) = C11 C22 C33 C44 C55 = (1)(3)(−1)(5)(2) = −30

(ii) We will use the following result:

A square matrix is invertible if and only if it has non-zero determinant.
Using this result, and the determinants computed in part (c), we obtain:
det(A) = 18, so that A is invertible.
det(B) = 0, so that B is not invertible.
det(C) = −30, so that C is invertible.
Therefore, matrices A and C are invertible, whereas B is not.

18
8. (a) (i) An elementary product from a square n × n matrix A is a product of n entries of A, such
that there is exactly one entry from each row and each column (i.e. such that each row
and column of A is “represented” exactly once).

(ii) If A is a square n × n matrix, then the determinant of A, written det(A), or |A|, is the
sum of all signed elementary products from A, i.e. the sum of products of the form
sign(σ) A1σ(1) · · · Anσ(n) , taken over all possible permutations of n numbers:
X
det(A) = sign(σ)A1σ(1) · · · Anσ(n)
σ∈Sn

(b) For M a matrix of size n × n and E an elementary matrix of size n × n:

det(EM ) = det(E)det(M )

(c) Let us show that A11 A22 · · · Ann is the only possibly non-zero elementary product present in
the determinant of A. We try to construct a possibly non-zero elementary product from A, by
selecting a suitable entry from each column of the upper triangular matrix A (one at a time):
 
A11 A12 A13 · · · A1n
 0 A22 A23 · · · A2n 
 
A = 
 0 0 A 33 · · · A 3n


 .. .. . . . . .. 
 . . . . . 
0 0 · · · 0 Ann

Firstly, the only possibly non-zero entry we can choose from column 1 is A11 .
Then, the only possibly non-zero entry we can choose from column 2 is the entry A22 (we
cannot choose A12 , since we are constructing an elementary product and have already
chosen an entry from row 1, namely A11 ).
Then, similarly, the only possibly non-zero entry we can choose from column 3 is A33 .
..
.
Finally, the only possibly non-zero entry we can choose from column n is Ann .
So, the only possibly non-zero elementary product from the matrix A is A11 A22 · · · Ann . This
corresponds to the permutation (1, 2, · · · , n), which has 0 inversions and so has sign +1.
Hence, the elementary product A11 A22 · · · Ann occurs with sign +1 in the determinant, as
the signed elementary product +A11 A22 · · · Ann . This is the only possibly non-zero signed
elementary product, so the determinant of A is precisely A11 A22 · · · Ann , as required.

19
(d) (i) Let us use elementary operations to find the required determinant:

1 0 2 3 1 0 2 3
−1 1 2 7 0 1 4 10
= using R2 → R2 + R1
0 2 4 8 0 2 4 8
3 2 2 5 3 2 2 5
1 0 2 3
0 1 4 10
= using R4 → R4 − 3R1
0 2 4 8
0 2 −4 −4
1 0 2 3
0 1 4 10
= using R3 → R3 − 2R2
0 0 −4 −12
0 2 −4 −4
1 0 2 3
0 1 4 10
= using R4 → R4 − 2R2
0 0 −4 −12
0 0 −12 −24
1 0 2 3
0 1 4 10 −1
= −4 × using R3 → 4
R3
0 0 1 3
0 0 −12 −24
1 0 2 3
0 1 4 10
= −4 × using R4 → R4 + 12R3
0 0 1 3
0 0 0 12
= −4 × (1 × 1 × 1 × 12) using the result from part (c)
 
1 0 2 3
−1 1 2 7
So det   = −48
0 2 4 8
3 2 2 5

(ii) We will use the following result:

A square matrix is invertible if and only if it has non-zero determinant.
Using this result, we may deduce that the given matrix is invertible, since its determinant
is equal to −48 i.e. since the given matrix has non-zero determinant.

20
9. (a) (i) A set V is a real subspace of Rn if it is a subset of Rn satisfying the following conditions:
(1) The zero vector belongs to V : 0 ∈ V .
(2) If vectors belong to V , then so does their sum: if a, b ∈ V , then a + b ∈ V .
(3) If a vector belongs to V , then so does any (real) scalar multiple of that vector: if
a ∈ V and λ ∈ R, then λa ∈ V .

(ii) Vectors v1 , · · · , vm in Rn span a subspace V of Rn if v1 , · · · , vm are vectors in V , and

each vector in V can be written as a linear combination of the vectors v1 , · · · , vm i.e. each
vector in V can be written in the form
λ1 v1 + · · · + λm vm
for real numbers λ1 , · · · , λm .

(b) (i) Let us verify that the three defining conditions for a real subspace hold for S:
(1) The zero vector is of the required form, i.e. if
   
x1 0
 x2  0
  =  
 x3  0
x4 0
then the condition of being in S is satisfied:
2x1 + x3 − x4 = 2(0) + (0) − (0) = 0
so that 0 ∈ S.
(2) Suppose that we start with vectors x and y in S, say
   
x1 y1
 x2  y 2 
x =  
 x3  , y =  
y 3 
x4 y4
Then, we may assume that the following equations hold:
2x1 + x3 − x4 = 0
2y1 + y3 − y4 = 0
Consider the sum of these vectors:
     
x1 y1 x1 + y 1
x2  y2   x2 + y 2 
x+y =  x3  + y3  = x3 + y3 
    

x4 y4 x4 + y 4

21
 Let us verify that x + y also satisfies the defining equation of S:

2(x1 + y1 ) + (x3 + y3 ) − (x4 + y4 ) = 2x1 + 2y1 + x3 + y3 − x4 − y4

= (2x1 + x3 − x4 ) + (2y1 + y3 − y4 )
= 0+0 using the fact that x and y are in S
= 0

So, as required, x + y ∈ S.
(3) Suppose that we are given any real number λ and a vector x in S, say
 
x1
 x2 
x =  
 x3 
x4

Then, we may assume that the following equation holds:

2x1 + x3 − x4 = 0

Consider the scalar product λx:

 
λx1
λx2 
λx = 
λx3 


λx4

Let us verify that λx also satisfies the defining equation of S:

2(λx1 ) + (λx3 ) − (λx4 ) = 2λx1 + λx3 − λx4

= λ(2x1 + x3 − x4 )
= λ×0 using the fact that x is in S
= 0

So, as required, λx ∈ S.
Since we have shown that all three conditions hold, we have shown that the set S is a real
subspace of R4 , as required.

22
 (ii) Every vector in S must satisfy the equation 2x1 + x3 − x4 = 0. Rearranging this defining
equation of S, we see that a general vector in S satisfies

x4 = 2x1 + x3

So, by substituting 2x1 + x3 for x4 , we deduce that a general element of S has the form
 
x1
 x2 

 x3 
 for x1 , x2 , x3 ∈ R
2x1 + x3

and we can decompose this general element into a linear combination of three vectors:
       
x1 1 0 0
 x2  0 1 0
 x3  = x1 0 + x2 0 + x3 1
        for x1 , x2 , x3 ∈ R
2x1 + x3 2 0 1

The above equation shows that S is spanned by the following set of vectors in R4 :
      
 1
 0 0  
 0 1 0
 , , 
 0 0 1
 
2 0 1
 

(c) (i) The given set S is not a real subspace of R4 .

For example, the zero vector does not satisfy the equation 2x1 + x3 − x4 = 5.
Explicitly, if we set    
x1 0
x2  0
  =  
x3  0
x4 0
then
2x1 + x3 − x4 = 2(0) + (0) − (0) = 0 ̸= 5
So, since 2x1 + x3 − x4 ̸= 5 for the zero vector, we deduce that 0 ∈ / S.
Therefore, condition (1) in the definition of a real subspace fails to hold, and so we may
deduce that S is not a real subspace of R4 .

23
(ii) The given set S is a real subspace of R4 .
The only possible (real) values of x1 , x2 , x3 , x4 for which x21 + x22 + x23 + x24 = 0 is
satisfied are x1 = 0, x2 = 0, x3 = 0, x4 = 0. So, S contains only one vector, the zero
vector:  

 0 
 
0

S =    or, equivalently, S = {0}

 0 
0
 

We may now verify that the three defining conditions of a subspace hold:
(1) The zero vector is contained in the set S: 0 ∈ S.
(2) Suppose that we are given vectors x, y in S. Since S consists only of the vector 0, we
must have x = 0 and y = 0. Then, the sum x + y = 0 is also in S, as required.
(3) Suppose that we are given any real number λ and a vector x in S, i.e. the vector
x = 0. Then, the product λx = λ × 0 = 0 is also in S, as required.
Since we have shown that all three conditions hold, we have shown that the set S is a real
subspace of R4 .

(iii) The given set S is not a real subspace of R4 .

It fails to satisfy condition (2) in the definition of a real subspace.
The following two vectors are in S:
   
1 1
0 0
v =  0
 , w =  0


1 −1

since v12 − v42 = (1)2 − (1)2 = 0 , w12 − w42 = (1)2 − (−1)2 = 0

Let us consider the sum of these vectors:
 
2
0
v+w = 
0


This sum is not in S, since in this case

(v + w)21 − (v + w)24 = (2)2 − (0)2 = 4 ̸= 0

Therefore, we have found vectors v, w, which are in S, but such that their sum, v + w, is
not in S. This shows that condition (2) in the definition of a subspace does not hold for
S, and so allows us to deduce that S is not a real subspace of R4 .

24
10. (a) (i) The standard real inner product of vectors x, y in Rn , denoted by ⟨x, y⟩, is the real number
defined by:
X n
⟨x, y⟩ = xi y i
i=1
   
* x1 y1 +
 ..   .. 
⟨x, y⟩ =  .  ,  .  = x1 y1 + · · · + xn yn
xn yn

(ii) The associated norm of a vector x in Rn , denoted by ||x||, is the real number defined by:
v
u n
p uX
||x|| = ⟨x, x⟩ = t x2i
i=1

(iii) Vectors x and y, in Rn , are orthogonal with respect to the standard real inner product if
they are non-zero vectors in Rn satisfying ⟨x, y⟩ = 0 i.e. satisfying

⟨x, y⟩ = x1 y1 + · · · + xn yn = 0

(b) (i) Suppose that x, y are (non-zero) vectors in Rn . Then, the following holds:

| ⟨x, y⟩ | ≤ ||x||||y||

(ii) Let us expand, and then simplify, the square of ||x + y||, noting that, by the Cauchy-
Schwarz inequality, | ⟨x, y⟩ | ≤ ||x||||y||, and, therefore, that ⟨x, y⟩ ≤ ||x||||y||:

||x + y||2 = ⟨x + y, x + y⟩ by definition of norm

= ⟨x, x⟩ + ⟨x, y⟩ + ⟨y, x⟩ + ⟨y, y⟩ by assumption
= ⟨x, x⟩ + 2 ⟨x, y⟩ + ⟨y, y⟩ by assumption
=||x||2 + 2 ⟨x, y⟩ + ||y||2 by definition of norm
≤||x||2 + 2||x||||y|| + ||y||2 by the Cauchy-Schwarz inequality
= (||x|| + ||y||)2 by factorising

Therefore, noting also that ||x + y|| and ||x|| + ||y|| cannot be negative, we may deduce
that, for vectors x, y in Rn , ||x + y|| ≤ ||x|| + ||y||, as required.

25
(c) (i) The Cauchy-Schwarz inequality states, in this case:

| ⟨a, b⟩ | ≤ ||a|| ||b||

In order to verify this inequality, let us compute the various inner products and norms
present in the inequality:
* 1   2 +
⟨a, b⟩ =  0  ,  1  = (1)(2) + (0)(1) + (−1)(−1) = 3
−1 −1
* 1   1 +
√
⟨a, a⟩ =  0  ,  0  = (1)(1) + (0)(0) + (−1)(−1) = 2 so that ||a|| = 2
−1 −1
* 2   2 +
√
⟨b, b⟩ =  1  ,  1  = (2)(2) + (1)(1) + (−1)(−1) = 6 so that ||b|| = 6
−1 −1

Then, we can confirm that the Cauchy-Schwarz

√ inequality holds
√ in this case. We have
shown that √ | ⟨a,√b⟩ | =√|3| = 3, √ ||a||
√ while √ = 2 and ||b|| = √6, from which we obtain
||a|| ||b|| = 2 6 = 12 = 4 3 = 2 3. Then, since 3 < 2 3, the inequality holds:

| ⟨a, b⟩ | ≤ ||a|| ||b||

(ii) For vectors, a and b say, the angle between the vectors, θ, may be found using:

⟨a, b⟩
cos θ =
||a|| ||b||

So, in this case, we may substitute the norms and inner product computed in part (c)(i),
to obtain: √
3 3
cos θ = √ =
2 3 2
The smallest positive angle with this cosine is 30◦ , or π6 , so this is the smallest positive
angle between the two vectors:
θ = 30◦

26
(iii) Let us start by computing a − ⟨a,a⟩
⟨a,b⟩
b (by substituting the values of the inner products
involved, which have already been computed in part (c)(i)):
      4  −1 
1 2 1
⟨a, a⟩ 2 3 3
a− b =  0  −  1  =  0  −  23  =  −2 3

⟨a, b⟩ 3 −2 −1
−1 −1 −1 3 3

We now confirm that this new vector is orthogonal to the vector a, by verifying that the
inner product of a and a − ⟨a,a⟩
⟨a,b⟩
b is equal to zero:

  * 1   −1 +
⟨a, a⟩ 3
a, a − b =  0  ,  −2
3
 = (1)( −1 )+(0)( −2 )+(−1)( −1 ) = − 1 +0+ 1 = 0
3 3 3 3 3
⟨a, b⟩ −1
−1 3

This completes the verification that the vector a is orthogonal to the vector a − ⟨a,a⟩
⟨a,b⟩
b with
respect to the standard real inner product; we have shown that
 
⟨a, a⟩
a, a − b =0
⟨a, b⟩

27
11. (a) (i) We may determine the norm of 3x, ||3x||, by first determining ||3x||2 = ⟨3x, 3x⟩:
Z 2
⟨3x, 3x⟩ = (3x)(3x) dx
0
Z 2
= 9x2 dx
0
 3 2
= 3x 0
= 3(23 ) − 3(03 )
= 24
√ √ √ √
so that the norm of 3x is ||3x|| = 24 = 4 6, i.e. ||3x|| = 2 6.

(ii) We start by setting u1 = 3x, u2 = x − 1, and now try to obtain u′1 , u′2 that are orthogonal.
We do not change u1 here: u′1 = u1 = 3x. We find a u′2 that is orthogonal to u′1 using:
⟨u2 , u′1 ⟩ ′ ⟨x − 1, 3x⟩
u′2 = u2 − ′ ′
u1 = (x − 1) − · (3x)
⟨u1 , u1 ⟩ ⟨3x, 3x⟩
The inner product ⟨3x, 3x⟩ was determined in the solution to part (a)(i): ⟨3x, 3x⟩ = 24.
Let us determine the other inner product involved above:
Z 2
⟨x − 1, 3x⟩ = (x − 1)(3x) dx
0
Z 2
= 3x2 − 3x dx
0
 3 3 2 2
= x − 2x 0
= (23 ) − 23 (22 ) − (03 ) − 32 (02 )

= 2

We may now use these values in order to determine u′2 :

⟨x − 1, 3x⟩
u′2 = (x − 1) − · (3x)
⟨3x, 3x⟩
2
= (x − 1) − · (3x)
24
1
= (x − 1) − · (3x)
12
i.e. u′2 = 34 x − 1

To complete the process, we compute the norm of each of the vectors u′1 and u′2 and then
divide each vector by its norm, to obtain an orthonormal set of vectors, {e1 , e2 } say:

28
 √
Starting with u′1 , we may use the fact that that ||u′1 || = ||3x|| = 2 6, from part (a)(i).
This leads to
u′1
e1 = = 2√1 6 u′1
||u′1 ||
√
i.e. e1 = 2√3 6 x = 2√32 x.
Proceeding to u′2 , we may compute the norm of this polynomial by first computing:
Z 2
3 3
4
x − 1, 4 x − 1 = ( 34 x − 1)( 34 x − 1) dx
0
Z 2
9 2
= 16
x − 64 x + 1 dx
0
3 3 3 2 2
= 16 x − 4 x + x 0
3
(23 ) − 34 (22 ) + (2) − 16
3
(03 ) − 34 (02 ) + (0)



= 16
= 12
From this, we may deduce that ||u′2 || = √1 .
2
This leads to
u′2 √ ′
e2 = = 2 u2
||u′2 ||
√
i.e. e2 = 2√3 2 x − 2.
So, the Gram-Schmidt process has led to the orthonormal set of vectors {e1 , e2 }, i.e. to:
n√ √ o
√3 x , √3
x− 2
2 2 2 2

(iii) Here, we are given that x2 is orthogonal to 4x + a, i.e. that ⟨x2 , 4x + a⟩ = 0.

We may use this information to determine the value of a, by first computing ⟨x2 , 4x + a⟩:
Z 2
2
x , 4x + a = (x2 )(4x + a) dx
Z0 2
= 4x3 + ax2 dx
0
 4 a 3 2
= x + 3x 0
= (24 ) + a3 (23 ) − (04 ) + a3 (03 )

= 16 + 8a 3

so that ⟨x2 , 4x + a⟩ = 16 + 8a
3
.
8a
As a result, since we are given that ⟨x2 , 4x + a⟩ = 0, we may deduce that 16 + 3
= 0,
i.e. that 8a
3
= −16, from which we may determine the required value of a:
a = −6

29
(b) In order to determine a Fourier series for f (x), we determine c, a1 , b1 , a2 , b2 , · · · , such that
f (x) = c + a1 cos(x) + b1 sin(x) + a2 cos(2x) + b2 sin(2x) + a3 cos(3x) + b3 sin(3x) + · · ·
To do so, we use the general rules:
Z π
1 π π
Z Z
1 1
c = f (x) dx , an = f (x)cos(nx) dx , bn = f (x)sin(nx) dx
2π −π π −π π −π

Note that, in this case, given that f (x) is defined differently on three different “parts” of
[−π, +π), we will compute each of these integrals by separating it into three distinct integrals:
Let us start with the “constant coefficient”, c:
Z π
1
c = f (x) dx
2π −π
Z −π Z π Z π
1 2 1 2 1
= 0 dx + 1 dx + 0 dx
2π −π 2π − π2 2π π2
1 π
= [x]−2 π
2π 2

1  π   π 
= − −
2π 2 2
1
=
2
Before computing bn , we note that sin(nx) describes an odd function: for each real number
x, sin(−nx) = −sin(nx). As a result, for any natural number n:
Z 0 Z π
2
sin(nx) dx = − sin(nx) dx
− π2 0

Let us now use this to compute the general coefficient bn :

1 π
Z
bn = f (x)sin(nx) dx
π −π
Z π Z π
1 −2 1 π
Z
1 2
= 0 · sin(nx) dx + 1 · sin(nx) dx + 0 · sin(nx) dx
π −π π − π2 π π2
Z π
1 2
= sin(nx) dx
π − π2
Z π
1 0
Z
1 2
= sin(nx) dx + sin(nx) dx
π − π2 π 0
Z π Z π
1 2 1 2
= − sin(nx) dx + sin(nx) dx
π 0 π 0
= 0

30
Similarly, before computing an , we note that cos(nx) describes an even function: for each real
number x, cos(−nx) = cos(nx). As a result, for any natural number n:
Z 0 Z π
2
cos(nx) dx = cos(nx) dx
− π2 0

Let us now use this to compute the general coefficient an :

1 π
Z
an = f (x)cos(nx) dx
π −π
Z π Z π
1 −2 1 π
Z
1 2
= 0 · cos(nx) dx + 1 · cos(nx) dx + 0 · cos(nx) dx
π −π π − π2 π π2
Z π
1 2
= cos(nx) dx
π − π2
Z π
1 0
Z
1 2
= cos(nx) dx + cos(nx) dx
π − π2 π 0
Z π
2 2
= cos(nx) dx
π 0
  π2
2 1
= sin(nx) dx
π n 0
 
2 1  nπ  1
= sin − sin(0)
π n 2 n
2  nπ 
= sin
nπ 2
We now note that
 nπ   nπ  n−1
sin = 0 if n is even , sin = (−1) 2 if n is odd
2 2
As a result, overall, we conclude that:
1 n−1 2
c = , an = 0 if n is even , an = (−1) 2 if n is odd , bn = 0
2 nπ
Substituting these values into the general formula for the Fourier series of a function:
∞
X
c + a1 cos(x) + b1 sin(x) + a2 cos(2x) + b2 sin(2x) + · · · = c + (an cos(nx) + bn sin(nx))
n=1

we conclude that, for the function f (x) described here:

 
1 2 2 1 2 1
f (x) = + cos(x) − cos(3x) + · · · = + cos(x) − cos(3x) + · · ·
2 π 3π 2 π 3

31
12. (a) (i) The vectors v1 , · · · , vm ∈ Rn are linearly independent if, whenever
λ1 v1 + · · · + λm vm = 0 for λ1 , · · · , λm ∈ R
then λ1 = 0 , · · · , λm = 0
(i.e. if the only solution of the above equation is the zero solution).

(ii) The vectors v1 , · · · , vm ∈ Rn span a subspace V of Rn if v1 , · · · , vm are vectors in V ,

and each vector in V can be written as a linear combination of the vectors v1 , · · · , vm i.e.
each vector in V can be written in the form
λ1 v1 + · · · + λm vm
for real numbers λ1 , · · · , λm .

(iii) The vectors v1 , · · · , vm ∈ Rn form a basis for a subspace V of Rn if:

1. The vectors v1 , · · · , vm span V , i.e. V = span{v1 , · · · , vm }.
2. The vectors v1 , · · · , vm are linearly independent.

(iv) The dimension of a subspace V of Rn , written dimV , or dim(V ), is the number of vectors
in a basis for V .

(b) Let us verify the two defining conditions of a basis (for R3 ) for the following set:
      
 1 0 1 
0 , 2 , 4
2 1 3
 

by setting the following as a general element of R3 :

 
x1
x =  x2  for real numbers x1 , x2 , x3
x3
and subsequently considering the matrix equation which expresses this element as a (real)
linear combination of the vectors v1 , v2 , v3 :
       
x1 1 0 1
x2  = α 0 + β 2 + γ 4
x3 2 1 3
We write this in the form of the matrix equation:
    
1 0 1 α x1
0 2 4 β  = x2 
2 1 3 γ x3

32
and we may use row reduction in order to solve this equation, by row reducing
 
1 0 1 | x1
 0 2 4 | x2 
2 1 3 | x3
   
1 0 1 | x1 1 0 1 | x1
R →R −2R
3 3 1
 0 2 4 | x2  −− −−− −−→ 0 2 4 | x2 
2 1 3 | x3 0 1 1 | −2x1 + x3
 
R2 → 1 R2
1 0 1 | x1
1
−−−−2−→ 0 1 2 | x
2 2

0 1 1 | −2x1 + x3
 
1 0 1 | x1
R →R −R 1
−−3−−−3−−→
2
0 1 2 | x
2 2

1
0 0 −1 | −2x1 − 2 x2 + x3
 
1 0 1 | x1
R →−R 1
−−3−−−→
3
0 1 2 | x
2 2

1
0 0 1 | 2x1 + 2 x2 − x3
0 0 | −x1 − 21 x2 + x3
 
1
R →R −R 1
−−1−−−1−−→
3
0 1 2 | x
2 2

1
0 0 1 | 2x1 + 2 x2 − x3
0 | −x1 − 12 x2 + x3
 
1 0
R →R −2R
2
−− 2
−−− 3
−−→ 0 1 0 | −4x1 − 12 x2 + 2x3 
0 0 1 | 2x1 + 12 x2 − x3

Therefore, returning to the original (matrix) equation, we may deduce that

α = −x1 − 12 x2 + x3 , β = −4x1 − 21 x2 + 2x3 , γ = 2x1 + 12 x2 − x3
This gives a way of expressing each vector in R3 as a linear combination of v1 , v2 , v3 :
       
x1 1 0 1
x2  = (−x1 − 1 x2 + x3 ) 0 + (−4x1 − 1 x2 + 2x3 ) 2 + (2x1 + 1 x2 − x3 ) 4
2 2 2
x3 2 1 3

So, we have shown that the given vectors span R3 , as required.

Furthermore, if we set x1 = 0 , x2 = 0 , x3 = 0 in the above row reduction, then we are able
to show that the only solution to the (real) vector equation
       
1 0 1 0
α 0 + β 2 + γ 4 = 0
      
2 1 3 0

33
 is

α = −x1 − 21 x2 + x3 = −(0) − 12 (0) + (0) = 0

β = −4x1 − 12 x2 + 2x3 = −4(0) − 21 (0) + 2(0) = 0
γ = 2x1 + 12 x2 − x3 = 2(0) + 21 (0) − (0) = 0

i.e.
α=0 , β=0 , γ=0
So, we may deduce that the vectors v1 , v2 , v3 are linearly independent (also).
Since we have now verified that the vectors v1 , v2 , v3 span R3 and that they are linearly
independent, we have concluded our verification that the set {v1 , v2 , v3 } is a basis for R3 .

(c) (i) We may rearrange the two ‘defining equations’ for S, in order to obtain, for example:

x1 = x4 , x2 = −2x3

We may then substitute the expressions for x1 and x2 into the general element of R4 , to
deduce that a general element of S has the form
     
x4 0 1
−2x3  −2 0
 x3  = x3  1  + x4 0 for real numbers x3 , x4
     

x4 0 1

In this case, we have shown that each element of S is a linear combination of the two
vectors given above. The two vectors are also linearly independent. So, we may deduce
that the following two vectors form a basis for the real subspace S of R4 :
   
0 1
−2 0
 , 
 1  0
0 1

(ii) Since a basis for S contains two vectors, we may deduce that the subspace S of R4 is
two-dimensional:
dim(S) = 2

34
13. (a) (i) The vectors v1 , · · · , vm ∈ Rn are linearly independent if, whenever

λ1 v1 + · · · + λm vm = 0 for λ1 , · · · , λm ∈ R

then
λ1 = 0 , · · · , λ m = 0
i.e. if the only solution of the above equation is the zero solution.

(ii) The vectors v1 , · · · , vm ∈ Rn span a subspace V of Rn if v1 , · · · , vm are vectors in V ,

and each vector in V can be written as a linear combination of the vectors v1 , · · · , vm i.e.
each vector in V can be written in the form

λ1 v1 + · · · + λm vm

for real numbers λ1 , · · · , λm .

(iii) The vectors v1 , · · · , vm ∈ Rn form a basis for a subspace V of Rn if:

1. The vectors v1 , · · · , vm span V , i.e. V = span{v1 , · · · , vm }.
2. The vectors v1 , · · · , vm are linearly independent.

(b) (i) Let us verify the two defining conditions of a basis (for R3 ) for the set {v1 , v2 , v3 }, by
setting the following as a general element of R3 :
 
x1
x = x2 
 for real numbers x1 , x2 , x3
x3

and subsequently considering the matrix equation which expresses this element as a (real)
linear combination of the vectors v1 , v2 , v3 : x = αv1 + βv2 + γv3 . We write this in the
form of the matrix equation:
    
1 1 0 α x1
2 1 1   β  =  x2 
−3 −1 0 γ x3

and may solve this equation by row reducing the augmented matrix
 
1 1 0 | x1
2 1 1 | x2 
−3 −1 0 | x3

35
    
1 1 0 | x1 11 0 | x1
R →R −2R
2 2 1
2 1 1 | x2  −− −−− −−→ 0−1 1 | −2x1 + x2 
−3 −1 0 | x3 −3
−1 0 | x3
 
1 0 |
1 x1
R →R +3R
3 3 1
−− −−− −−→ −1 1 | −2x1 + x2 
0
2 0 | 3x1 + x3
0
 
1 1 0 | x1
R →−R
−−2−−−→
2
0 1 −1 | 2x1 − x2 
0 2 0 | 3x1 + x3
 
1 0 1 | −x1 + x2
R →R −R
−−1−−−1−−→
2
0 1 −1 | 2x1 − x2 
0 2 0 | 3x1 + x3
 
1 0 1 | −x1 + x2
R →R −2R
3 3 2
−− −−− −−→ 0 1 −1 | 2x1 − x2 
0 0 2 | −x1 + 2x2 + x3
 
R3 → 1 R3
1 0 1 | −x1 + x2
−−−−2−→ 0 1 −1 | 2x1 − x2 
1 1
0 0 1 | − 2 x1 + x2 + 2 x3
− 12 x1 − 12 x3
 
1 0 0 |
R →R −R
−−1−−−1−−→
3
0 1 −1 | 2x1 − x2 
1 1
0 0 1 | − 2 x1 + x2 + 2 x3
− 12 x1 − 12 x3
 
1 0 0 |
R →R +R 3
−−2−−−2−−→
3
0 1 0 | x + 12 x3
2 1

1 1
0 0 1 | − 2 x1 + x2 + 2 x3

Therefore, returning to the original (matrix) equation, we may deduce that

α = − 12 x1 − 21 x3 , β = 32 x1 + 12 x3 , γ = − 12 x1 + x2 + 12 x3

This gives a way of expressing each vector in R3 as a linear combination of the vectors
v1 , v2 , v3 :
       
x1 1 1 0
x2  = (− 1 x1 − 1 x3 )  2  + ( 3 x1 + 1 x3 )  1  + (− 1 x1 + x2 + 1 x3 ) 1
2 2 2 2 2 2
x3 −3 −1 0

So, we have shown that the set {v1 , v2 , v3 } spans R3 , as required.

36
 Furthermore, if we set x1 = 0 , x2 = 0 , x3 = 0 in the above row reduction, then we are
able to show that the only solution to the (real) vector equation
       
1 1 0 0
α  2  + β  1  + γ 1 = 0
−3 −1 0 0
is
α = − 21 x1 − 21 x3 = − 12 (0) − 12 (0) = 0
β = 23 x1 + 12 x3 = 32 (0) + 12 (0) = 0
γ = − 21 x1 + x2 + 12 x3 = − 12 (0) + (0) + 21 (0) = 0
i.e. α = 0 , β = 0 , γ = 0.
So, we may deduce that the vectors v1 , v2 , v3 are linearly independent (also).
Since we have now verified that the vectors v1 , v2 , v3 span R3 and that they are linearly
independent, we have concluded our verification that the set {v1 , v2 , v3 } is a basis for R3 .

(ii) We may substitute w into the equation obtained in part (b)(i), which expresses any vector
in R3 as a linear combination of v1 , v2 , v3 .
For the vector w: x1 = 8, x2 = 9, x3 = 10, so we substitute these values into the equation
from part (b)(i), and simplify to obtain the values of α, β, γ, in this case, leading to:
       
8 1 1 0
 9  = −9  2  + 17  1  + 10 1
10 −3 −1 0
So, as required, we have expressed w as a linear combination of the vectors v1 , v2 , v3 :
w = −9v1 + 17v2 + 10v3

(iii) Suppose that a linear combination of the given vectors is equal to zero, say
αv1 + βv2 + γv4 = 0
We may rewrite the above as a matrix equation, to obtain:
    
1 1 −1 α 0
2 1 1   β = 0
 
−3 −1 −3 γ 0
and may use row reduction to solve this equation, by row reducing the augmented matrix
 
1 1 −1 | 0
2 1 1 | 0
−3 −1 −3 | 0

37
    
1 1 −1 | 0 1 1 −1 | 0
R →R −2R
2 2 1
2 1 1 | 0 −− −−− −−→ 0 −1 3 | 0
−3 −1 −3 | 0 −3 −1 −3 | 0
 
1 1 −1 | 0
R →R +3R
3 3 1
−− −−− −−→ 0 −1 3 | 0
0 2 −6 | 0
 
1 1 −1 | 0
R →−R
−−2−−−→
2
0 1 −3 | 0
0 2 −6 | 0
 
1 0 2 | 0
R →R −R
−−1−−−1−−→
2
0 1 −3 | 0
0 2 −6 | 0
 
1 0 2 | 0
R →R −2R
3 3 2
−− −−− −−→ 0 1 −3 | 0
0 0 0 | 0

So, the original (matrix) equation reduces to the following equation:

    
1 0 2 α 0
0 1 −3 β  = 0
0 0 0 γ 0
Therefore, we deduce that
α + 2γ = 0 , β − 3γ = 0 , 0=0
Hence, we have one free variable, for example γ in this case, using which we may obtain:
α = −2γ , β = 3γ , γ = γ
So, we have shown that it is not necessarily the case that α = 0, β = 0, γ = 0.
Therefore, we deduce that the set of vectors {v1 , v2 , v4 } is not linearly independent; it is
a linearly dependent set.
To give an explicit linear dependency between the vectors, we may choose a “compatible”
set of values for α, β, γ. For example, if we set γ = 1, we obtain:
α = −2 , β = 3 , γ =1
and, substituting these values into the ‘linear combination equation’ we started with leads
to an explicit linear dependency between the vectors, as required:
       
1 1 −1 0
−2 2 + 3 1 + 1 1
      = 0
 or − 2v1 + 3v2 + v4 = 0
−3 −1 −3 0

38
14. (a) A real linear map T from Rn to Rm is a function T : Rn → Rm satisfying the following three
conditions:
(1) T (0) = 0.
(2) If a, b ∈ Rn , then T (a + b) = T (a) + T (b).
(3) If a ∈ Rn and λ ∈ R, then T (λa) = λ T (a).

(b) (i) For a real linear map T , from Rn to Rm , the kernel of T , written as kernel(T ) or Ker(T ),
is the set
{x ∈ Rn : T (x) = 0}
i.e. it is the set of vectors in Rn that the map T sends to the zero vector in Rm .

(ii) For a real linear map T , from Rn to Rm , the image of T , written as image(T ) or Im(T ),
is the set
{y ∈ Rm : T (x) = y for some x ∈ Rn }
i.e. it is the set of vectors in Rm that are attained by the map T .

(c) (i) We may obtain the required answer by applying the real linear map to the standard basis
vectors, and then writing down the resulting answers as the columns of MT , in order.
For example, in this case:
         
1   0   0   0   0  
0 1  1 2  0 1  0 3  0  5
         
T 0
 
 = 1 , T 0 =  1  , T 1 =  0  , T 0 = 1 , T 0 = 3
       
0 2 0 −1 0 −3 1 1 0 5
0 0 0 0 1

so that:  
1 2 1 3 5
MT = 1 1 0 1 3
2 −1 −3 1 5

39
(ii) We wish to find a basis for the kernel of T , i.e. for the set of all vectors
 
x1
 x2 
 
 x3 
 
 x4 
x5

such that  
x1  
x2  0
 
T x3  = 0
  
x4  0
x5
Since multiplying on the left by MT is “the same as” applying the linear map T , we may
rewrite the above equation as
 
  x1  
1 2 1 3 5  x
  2
 0
1 1 0 1 3  x = 0
 3
 
2 −1 −3 1 5  x4  0
x5

and this is now a (homogeneous) matrix equation that we may solve by row reducing the
augmented matrix  
1 2 1 3 5 | 0
1 1 0 1 3 | 0
2 −1 −3 1 5 | 0

40
    
1 2 1 3 5 | 0 1 2 1 3 5 | 0
R2 →R2 −R1
1 1 0 1 3 | 0 −−−−−−−→ 0 −1 −1 −2 −2 | 0
2 −1 −3 1 5 | 0 2 −1 −3 1 5 | 0
 
1 2 1 3 5 | 0
R3 →R3 −2R1
−−−−−−−→ 0 −1 −1 −2 −2 | 0
0 −5 −5 −5 −5 | 0
 
1 2 1 3 5 | 0
R2 →−R2
−−−−−→ 0 1 1 2 2 | 0
0 −5 −5 −5 −5 | 0
 
1 0 −1 −1 1 | 0
R1 →R1 −2R2
−−−−−−−→ 0 1 1 2 2 | 0
0 −5 −5 −5 −5 | 0
 
1 0 −1 −1 1 | 0
R →R +5R
3 3 2
−− −−− −−→ 0 1 1 2 2 | 0
0 0 0 5 5 | 0
 
R3 → 1 R3
1 0 −1 −1 1 | 0
−−−−5−→ 0 1 1 2 2 | 0
0 0 0 1 1 | 0
 
1 0 −1 0 2 | 0
R →R +R
−−1−−−1−−→
3
0 1 1 2 2 | 0
0 0 0 1 1 | 0
 
1 0 −1 0 2 | 0
R →R −2R
2 2 3
−− −−− −−→ 0 1 1 0 0 | 0
0 0 0 1 1 | 0

So, the original matrix equation reduces to the following equation:

 
  x1  
1 0 −1 0 2   x2 
 0
0 1 1 0 0 x3  = 0
 
0 0 0 1 1  x4  0
x5

Therefore, we deduce that

x1 − x3 + 2x5 = 0 , x2 + x3 = 0 , x4 + x5 = 0

i.e. we may set x3 and x5 as free variables, and set x1 = x3 − 2x5 , x2 = −x3 , x4 = −x5 .

41
 Substituting these equations into the starting vector, composed of x1 , x2 , x3 , x4 , x5 , we
deduce that a general element satisfying the above equation (i.e. a general element of the
kernel of the map) has the form:
     
x3 − 2x5 1 −2
 −x3  −1 0
     
 x3  = x3  1  + x5  0  for x3 , x5 in R
     
 −x5  0 −1
x5 0 1
In this case, we may say that each element of the kernel of T is a linear combination of
the two vectors given above. The two vectors are also linearly independent, so that we
may deduce that the following set is a basis for the kernel of T :
    

 1 −2  
−1  0 

    


1,0
   


  0  −1 
 
0 1
 

(iii) The row reduction performed in part (c)(ii) allows us to write down a basis for the image
of the linear map. The leading ones of the (final) row reduced matrix from part (c)(ii)
occur in columns 1, 2 and 4.
It follows that if we choose columns 1, 2, 4, of the original matrix, MT , we will have a
basis for the image of T .
Since  
1 2 1 3 5
MT = 1 1 0 1 3
2 −1 −3 1 5
we deduce that the following set is a basis for the image of T :
     
 1 2 3 
1 ,  1  , 1
2 −1 1
 

(iv) As determined in part (c)(iii), a basis for the image of the linear map T consists of three
vectors, i.e. the image of T is three-dimensional. So, the rank of T is 3:
rank(T ) = 3
Also, as determined in part (c)(ii), a basis for the kernel of the linear map T consists of
two vectors, i.e. the kernel of T is two-dimensional. So, the nullity of T is 2:
nullity(T ) = 2

42
15. (a) A real linear map T from Rn to Rm is a function T : Rn → Rm satisfying the following three
conditions:
(1) T (0) = 0.
(2) If a, b ∈ Rn , then T (a + b) = T (a) + T (b).
(3) If a ∈ Rn and λ ∈ R, then T (λa) = λ T (a).

(b) (i) For a real linear map T : Rn → Rm , the kernel of T , written as kernel(T ) or Ker(T ), is
the set
{x ∈ Rn : T (x) = 0}
i.e. it is the set of vectors in Rn that the map T sends to the zero vector in Rm .

(ii) For a real linear map T : Rn → Rm , the image of T , written as image(T ) or Im(T ), is
the set
{y ∈ Rm : T (x) = y for some x ∈ Rn }
i.e. it is the set of vectors in Rm that are attained by the map T .

(c) Let us use the defining conditions of a real linear map, given in part (a), in order to verify that
the defining conditions of a real subspace hold for Ker(T ):
(1) By definition of a linear map, the zero vector maps to the zero vector, i.e. T (0) = 0. So,
the zero vector lies in the kernel of T : 0 ∈ Ker(T ), as required.
(2) Suppose that we start with vectors a and b in Rn , which lie in the kernel of T , i.e. such
that T (a) = 0 and T (b) = 0. Let us try to show that the sum a + b also lies in the kernel
of T :

T (a + b) = T (a) + T (b) since T is a linear map

= 0+0 since a, b ∈ Ker(T )
= 0

Therefore, since T (a + b) = 0, we deduce that a + b ∈ Ker(T ), as required.

(3) Suppose that we start with a real number λ, and a vector, a, in Rn , that is in the kernel of
T , i.e. such that T (a) = 0. Let us try to show that the scalar multiple λa also lies in the
kernel of T :

T (λa) = λT (a) since T is a linear map

= λ×0 since a ∈ Ker(T )
= 0

Therefore, since T (λa) = 0, we deduce that λa ∈ Ker(T ), as required.

43
 Since we have shown that all three conditions hold, we have proven that Ker(T ), the kernel
of the real linear map T : Rn → Rm , is a real subspace of Rn , as required.

(d) (i) We are asked to verify that T is a real linear map. To this end, let us verify that the
defining conditions of a linear map, given in part (a), hold for T :
(1) Let us apply our map to the zero vector, i.e the vector with x1 = 0, x2 = 0:
   
  2(0) + 0 0
0
T =  0−0  = 0

0
3(0) 0
Hence T (0) = 0, as required.
(2) Consider general vectors in R2 , on which to apply the map, say
   
a1 b1
a = , b =
a2 b2
Then:
   
a1 b1
T (a + b) = T +
a2 b2
 
a1 + b 1
= T
a2 + b 2
 
2(a1 + b1 ) + (a2 + b2 )
=  (a1 + b1 ) − (a2 + b2 ) 
3(a1 + b1 )
 
(2a1 + a2 ) + (2b1 + b2 )
=  (a1 − a2 ) + (b1 − b2 ) 
(3a1 ) + (3b1 )
   
2a1 + a2 2b1 + b2
=  a1 − a2  +  b 1 − b 2 
3a1 3b1
   
a1 b
= T +T 1
a2 b2
= T (a) + T (b)
Hence, if a, b ∈ R3 , then T (a + b) = T (a) + T (b), as required.
(3) Suppose that λ is a given real number and a is a general vector in R2 , say
 
a1
a =
a2

44
 Then, we are required to show that T (λa) is equal to λ T (a):
  
a
T (λa) = T λ 1
a2
 
λa1
= T
λa2
 
2(λa1 ) + (λa2 )
=  (λa1 ) − (λa2 ) 
3(λa1 )
 
λ(2a1 + a2 )
=  λ(a1 − a2 ) 
λ(3a1 )
 
2a1 + a2
= λ  a1 − a2 
3a1
 
a1
= λT
a2
= λ T (a)

Hence, if a ∈ R2 and λ ∈ R, then T (λa) = λ T (a), as required.

Having verified the three defining properties of a linear map, we may deduce that T is a
real linear map, as required.

(ii) We wish to find a basis for the kernel of T , i.e. for the set of all vectors
 
x1
x2
such that  
  0
x1
T = 0

x2
0
We may rewrite the above equation as
   
2 1   0
1 −1 x1 = 0
x2
3 0 0
and this is now a (homogeneous) matrix equation that we may solve directly; it reduces
to the following equations:

2x1 + x2 = 0 , x1 − x2 = 0 , 3x1 = 0

45
 From the third equation, we may deduce that x1 = 0, and, by substituting this into either
of the first two equations, we may deduce that x2 = 0.
Therefore, we deduce that
x1 = 0 , x2 = 0
Substituting these into the starting vector, composed of x1 , x2 , we deduce that a general
element satisfying the above equation (i.e. a general element of the kernel of the map)
has the form:  
0
0
In this case, we may say that the kernel of T consists of a single vector, the zero vector.
Therefore, the kernel of T has no elements in its basis, i.e. a basis for the kernel of T is
the empty set, ∅.

(iii) We note that a general element of the image of T has the form
     
2x1 + x2 2 1
 x1 − x2  = x1 1 + x2 −1 for real numbers x1 , x2
3x1 3 0

The above decomposition allows us to identify two vectors (one corresponding to each
free variable, i.e. one corresponding to each real number we are allowed to choose) whose
span is the image of the linear map T . The two vectors are also linearly independent
(something that we may, but are not required to, verify directly). As a result, we deduce
that the following set is a basis for the image of T :
   
 2 1 
1 , −1
3 0
 

46
16. (a) An eigenvector of a complex n × n matrix A is a non-zero vector v in Cn such that

Av = λv for some complex number λ

The complex number λ is then the corresponding eigenvalue of A.

(b) We assume that v is an eigenvector of A, corresponding to an eigenvalue λ, so that Av = λv.

Consider An v:

An v = An−1 (Av)
= An−1 (λv) using Av = λv
= λ(An−1 v) since A corresponds to a linear map
= λ An−2 (Av)

= λ An−2 (λv)


using Av = λv
= λ2 (An−2 v) since A corresponds to a linear map
..
.
= λn v

Therefore, we have shown that An v = λn v, i.e. that v is an eigenvector of An , corresponding

to the eigenvalue λn , as required.

(c) (i) We first determine the eigenvalues of A by using the characteristic polynomial of A:

2−λ 0 0
det(A − λI) = 0 ⇔ 0 5−λ 4 = 0
0 3 6−λ
⇔ (2 − λ) ((5 − λ)(6 − λ) − 12) = 0
(2 − λ) λ2 − 11λ + 30 − 12


⇔ = 0
(2 − λ) λ2 − 11λ + 18


⇔ = 0
⇔ (2 − λ)2 (9 − λ) = 0

So, the matrix A has two eigenvalues:

λ = 2 with algebraic multiplicity 2

λ = 9 with algebraic multiplicity 1

Let us determine the eigenvectors corresponding to each eigenvalue. Each eigenvector x

corresponding to an eigenvalue λ must satisfy the equation (A − λI)x = 0.

47
So, let us substitute each of the eigenvalues of A into this equation and then solve the
resulting matrix equation.
For the eigenvalue λ = 2, we solve the equation (A − 2I)x = 0:
    
0 0 0 x1 0
0 3 4 x2  = 0
0 3 4 x3 0
by row reducing the augmented matrix
 
0 0 0 | 0
0 3 4 | 0
0 3 4 | 0
   
0 0 0 | 0 0 3 4 | 0
R ↔R
0 3 4 | 0 −−1−−→
3
0 3 4 | 0
0 3 4 | 0 0 0 0 | 0
1 43 | 0
 
R1 → 1 R1
0
−−−−3−→ 0 3 4 | 0
0 0 0 | 0
1 43 | 0
 
0
2 R →R −3R
2 1
−− −−− −−→ 0 0 0 | 0
0 0 0 | 0

So, the original matrix equation reduces to the equations:

x2 + 43 x3 = 0 , 0=0 , 0=0
−4
Therefore, setting x2 = x,
3 3
we see that a general solution to the above matrix equation
is of the form:  
x1
 −4 x3 
3
x3
In other words, a general eigenvector of A corresponding to the eigenvalue 2 is of the
form:
     
x1 1 0
 −4 x3  = x1 0 + x3  −4  for x1 , x3 ∈ R , x1 ̸= 0 and/or x3 ̸= 0
3 3
x3 0 1
Equivalently, the set of eigenvectors corresponding to the eigenvalue 2 is
  
 x1 
 −4 x3  : x1 ∈ R , x3 ∈ R , x1 ̸= 0 and/or x3 ̸= 0
 3
x3


48
For the eigenvalue λ = 9, we solve the equation (A − 9I)x = 0, by row reducing the
augmented matrix  
−7 0 0 | 0
 0 −4 4 | 0
0 3 −3 | 0

   
−7 0 0 | 0 R1 → −1 R1
1 0 0 | 0
7
 0 −4 4 | 0 −−−−−−→ 0 −4 4 | 0
0 3 −3 | 0 0 3 −3 | 0
 
R2 → −1 R2
1 0 0 | 0
4
−−−−− −→ 0 1 −1 | 0
0 3 −3 | 0
 
1 0 0 | 0
R →R −3R
3 3 2
−− −−− −−→ 0 1 −1 | 0
0 0 0 | 0

So, using the same notation as above for a(n) (eigen)vector x, the original matrix equation
reduces to the equations:

x1 = 0 , x2 − x3 = 0 , 0=0

Therefore, after setting x1 = 0 and x2 = x3 , we see that a general eigenvector of A

corresponding to the eigenvalue 9 is of the form:
   
0 0
x3  = x3 1 for x3 ∈ R , x3 ̸= 0
x3 1

Equivalently, the set of eigenvectors corresponding to the eigenvalue 9 is

  
 0 
x3  : x3 ∈ R , x3 ̸= 0
x3
 

49
(ii) In order to find a suitable invertible matrix P , which diagonalises A, we are required to
find three linearly independent eigenvectors for the matrix A. Three such vectors were
determined in part (c)(i) (two corresponding to the eigenvalue 2, and one corresponding
to the eigenvalue 9):      
1 0 0
0 ,  −4  , 1
3
0 1 1
So, we may deduce that A is diagonalisable.
For example, we may choose  
1 0 0
−4
P = 0 
3
1
0 1 1
In this case, we have chosen to place two eigenvectors corresponding to the eigenvalue 2
in columns 1 and 2, and an eigenvector corresponding to the eigenvalue 9 in column 3, so
the product P −1 AP (which is the required diagonal matrix) will be equal to
 
2 0 0
D = 0 2 0
0 0 9

50
17. (a) (i) An eigenvector of a complex n × n matrix A is a non-zero vector v in Cn such that
Av = λv for some complex number λ
The complex number λ is then the corresponding eigenvalue of A.

(ii) The characteristic equation of a complex n × n matrix A is the equation

det(A − λIn ) = 0
where In is the identity matrix of size n × n, and det denotes the determinant of a matrix.

(b) By definition, a complex number λ is an eigenvalue of A if and only if there exists a non-zero
vector v ∈ Cn such that Av = λv. But, then:
Av = λv ⇔ Av − λv = 0
⇔ Av − λIv = 0
⇔ (A − λI)v = 0
Since we are assuming that v is a non-zero vector, we may deduce, using one of the assumed
results, that (A − λI)v = 0 holds (for a non-zero vector v) if and only if the matrix A − λI is
not invertible.
Finally, we may now use the other assumed result to conclude that the matrix A − λI is not
invertible if and only if the determinant of this matrix is 0, i.e. if and only if det(A − λI) = 0.
Therefore, overall, we deduce that a complex number is an eigenvalue of a complex square
matrix A if and only if it is a root of the characteristic equation of A, as required.

(c) We first determine the eigenvalues of A, by solving the characteristic equation of A:

7−λ 5 1
det(A − λI) = 0 ⇔ 0 4 − λ −1 = 0
0 1 2−λ
⇔ (7 − λ) ((4 − λ)(2 − λ) + 1) = 0
⇔ (7 − λ)(λ2 − 6λ + 8 + 1) = 0
⇔ (7 − λ)(λ2 − 6λ + 9) = 0
⇔ (7 − λ)(λ − 3)2 = 0
So, the matrix A has two eigenvalues:
λ = 3 with algebraic multiplicity 2
λ = 7 with algebraic multiplicity 1
Let us now determine the eigenvectors corresponding to each eigenvalue. Each eigenvector x
corresponding to an eigenvalue λ must satisfy the equation (A − λI)x = 0.

51
So, let us substitute each of the eigenvalues of A into this equation and then solve the resulting
matrix equation.
For the eigenvalue λ = 3, we solve the equation (A − 3I)x = 0:
    
4 5 1 x1 0
0 1 −1 x2  = 0
0 1 −1 x3 0

by row reducing the augmented matrix

 
4 5 1 | 0
0 1 −1 | 0
0 1 −1 | 0

5 1
   
4 5 1 | 0 R1 → 41 R1
1 4 4
| 0
0 1 −1 | 0 −−−−−→ 0 1 −1 | 0
0 1 −1 | 0 0 1 −1 | 0
0 64 | 0
 
R1 →R1 − 5 R2
1
4
−−−−−−− −→ 0 1 −1 | 0
0 1 −1 | 0
0 64 | 0
 
1
R →R −R
−−3−−−3−−→
2
0 1 −1 | 0
0 0 0 | 0

So, the original matrix equation reduces to the equations:

x1 + 64 x3 = 0 , x2 − x3 = 0 , 0=0

−3
Therefore, setting x1 = x
2 3
and x2 = x3 , we see that a general solution to the above matrix
equation is of the form:  −3 
x3
2
 x3 
x3
In other words, a general eigenvector of A corresponding to the eigenvalue 3 is of the form:
 −3   −3 
2
x 3 2
 x3  = x3  1  for x3 ∈ R , x3 ̸= 0
x3 1

52
Equivalently, the set of eigenvectors corresponding to the eigenvalue 3 is
 −3  
 2 x3 
 x3  : x3 ∈ R , x3 ̸= 0
x3
 

For the eigenvalue λ = 7, we solve the equation (A − 7I)x = 0, by row reducing

 
0 5 1 | 0
0 −3 −1 | 0
0 1 −5 | 0

   
0 5 1 | 0 0 1 −5 | 0
R ↔R
0 −3 −1 | 0 −−1−−→
3
0 −3 −1 | 0
0 1 −5 | 0 0 5 1 | 0
 
0 1 −5 | 0
R →R +3R
2 2 1
−− −−− −−→ 0 0 −16 | 0
0 5 1 | 0
 
0 1 −5 | 0
R →R −5R
3 3 1
−− −−− −−→ 0 0 −16 | 0
0 0 26 | 0
 
R2 → −1 R2
0 1 −5 | 0
−−−−16
−−→ 0 0 1 | 0
0 0 26 | 0
 
0 1 0 | 0
R →R +5R
1 1 2
−− −−− −−→ 0 0 1 | 0
0 0 26 | 0
 
0 1 0 | 0
R →R −26R
−−3−−−
3 2
−−−→ 0 0 1 | 0
0 0 0 | 0

So, using the same notation as above for a(n) (eigen)vector x, the original matrix equation
reduces to the equations:
x2 = 0 , x3 = 0 , 0 = 0
Therefore, setting x2 = 0 and x3 = 0, we see that a general eigenvector of A corresponding to
the eigenvalue 7 is of the form:
   
x1 1
 0  = x1 0 for x1 ∈ R , x1 ̸= 0
0 0

53
Equivalently, the set of eigenvectors corresponding to the eigenvalue 7 is
  
 x1 
 0  : x1 ∈ R , x1 ̸= 0
0
 

In this case, overall, we have not found enough (i.e. three) linearly independent eigenvectors
for the matrix A in order to be able to form a suitable matrix P that diagonalises A.
(For the eigenvalue 3, the algebraic multiplicity is 2 but the geometric multiplicity is 1, so it is
not possible to find two linearly independent eigenvectors corresponding to this eigenvalue.)
Our solution led to sets consisting of only two linearly independent eigenvectors overall, e.g.:
 −3   
2
1
 1  , 0
1 0

So, we may deduce that A is not diagonalisable.

54
18. (a) A square matrix D is a diagonal matrix if all of its entries above and below the main diagonal
are zero entries, i.e. if Dij = 0 for i ̸= j.

(b) We assume that P −1 AP = D, and may “cancel out” P and P −1 from this equation in order
to obtain an expression for A:
P −1 AP = D ⇔ P (P −1 AP ) = P D
⇔ AP = P D
−1
⇔ (AP )P = (P D)P −1
⇔ A = P DP −1

Let us now substitute this expression for A into (the expression) An and simplify:
An = (P DP −1 )n by substituting A = P DP −1
= (P DP −1 )(P DP −1 ) · · · (P DP −1 )
= P D(P −1 P )D(P −1 P )D · · · D(P −1 P )DP −1
= P DIDID · · · DIDP −1 by using P −1 P = I
= P Dn P −1

So, we have shown that, if P −1 AP = D, then An = P Dn P −1 , as required.

(c) (i) The Spectral Theorem for Hermitian matrices states that, for a complex square matrix M ,
of size n × n say, if M is a Hermitian matrix (i.e. if M T = M , where M T denotes the
transpose of M and M denotes the conjugate of M ), then:
· The matrix M is diagonalisable.
· Each eigenvalue of M is a real number.
· Eigenvectors of M corresponding to distinct eigenvalues are orthogonal with respect
to the standard complex inner product.
· There exists an orthonormal set of n eigenvectors for M (orthonormal with respect to
the standard complex inner product).

(ii) We will first determine the eigenvalues of A by solving the characteristic equation of A:
1−λ i
det(A − λI) = 0 ⇔ = 0
−i 1 − λ
⇔ (1 − λ)(1 − λ) − (i)(−i) = 0
⇔ λ2 − 2λ + 1 − 1 = 0
⇔ λ2 − 2λ = 0
⇔ λ(λ − 2) = 0

55
So, the matrix A has two eigenvalues:
λ = 0 with algebraic multiplicity 1
λ = 2 with algebraic multiplicity 1
Let us find the eigenvectors corresponding to each eigenvalue, by substituting each of the
eigenvalues of A into the equation
(A − λI)x = 0
and then solving the resulting matrix equation.
For the eigenvalue λ = 0, we solve the equation (A − 0I)x = 0, i.e. Ax = 0:
    
1 i x1 0
=
−i 1 x2 0
Since this already gives a relatively simple system of equations, we shall not row reduce
the augmented matrix. Each row is a multiple of the other row (e.g. if we multiply row 1
through by −i, we obtain row 2), so we essentially have a single equation: x1 + ix2 = 0.
So, setting x1 = −ix2 above, we see that a general solution to the above equation is of
the form:  
−ix2
x2
In other words, a general eigenvector of A corresponding to the eigenvalue 0 is of the
form:    
−ix2 −i
= x2 for x2 ∈ C , x2 ̸= 0
x2 1
For the eigenvalue λ = 2, we solve the equation (A − 2I)x = 0:
    
−1 i x1 0
=
−i −1 x2 0
Once again, since this already gives a relatively simple system of equations, we shall not
row reduce the augmented matrix. Each row is a multiple of the other row (e.g. if we
multiply row 1 through by i, we obtain row 2), so we essentially have a single equation:
−x1 + ix2 = 0.
So, setting x1 = ix2 above, we see that a general solution to the above equation is of the
form:  
ix2
x2
In other words, a general eigenvector of A corresponding to the eigenvalue 2 is of the
form:    
ix2 i
= x2 for x2 ∈ C , x2 ̸= 0
x2 1

56
(iii) Using the solution to part (c)(ii), we may find two linearly independent eigenvectors for
the matrix A (one corresponding to each eigenvalue). Therefore, we deduce that the
matrix A is diagonalisable.
For example, we may choose  
−i i
P =
1 1
In this case, since we have chosen to place an eigenvector corresponding to the eigenvalue
0 in column 1 of P and an eigenvector corresponding to the eigenvalue 2 in column 2 of
P , the product P −1 AP (which is the required diagonal matrix) will be equal to
 
0 0
D =
0 2

We will use the diagonalisation of A and the result from part (b) to find a formula for An .
Starting from the matrix D, Dn is simple to compute (since D is a diagonal matrix):
   
0 0 n 0 0
D = ⇒ D =
0 2 0 2n

Also, we may use P to find P −1 , for example by row reducing the augmented matrix
(P |I), as we have seen before:

   
−i i | 1 0 R1 ↔R2 1 1 | 0 1
−−−−→
1 1 | 0 1 −i i | 1 0
 
R2 →R2 +iR1 1 1 | 0 1
−−−−−−−→
0 2i | 1 i
1  
R2 → R
2i 2 1 1 | 0 1
−−−−−−→
0 1 | 2i1 12
1 0 | − 2i1 1
 
R1 →R1 −R2
−−−−−−−→ 2
0 1 | 2i1 1
2

The identity matrix now forms the “left half” of the reduced matrix, so the “right half”
must be the inverse of P :
 1 1  i 1
−1 − 2i 2 2 2
P = 1 1 = i 1
2i 2
− 2 2

57
We may now determine the required formula, for An , using the result from part (b):

An = P Dn P −1
   i 1

−i i 0 0 2 2
=
1 1 0 2n − 2i 1
2
   i 1
0 2n i 2 2
=
0 2n − 2i 12
 n−1 
2 2n−1 i
=
−2n−1 i 2n−1

Therefore:  
n 2n−1 2n−1 i
A =
−2n−1 i 2n−1

58