Imperial College London

Department of Mathematics

Solutions to Question Sheet 2 - Pr. Class week 4

MATH40003 Linear Algebra and Groups Term 2, 2021/22

This is the problem sheet for the problem classes in week 4. All questions can be at-
tempted with the material in lectures 1–5. Solutions will be released after the classes on
Monday of week 4.

Question 1 Let n ∈ N, n ≥ 2. Suppose D : Mn (R) → R is a function on which

elementary row operations have the same effect as they do for det (for example, if B
is obtained from A ∈ Mn (R) by interchanging two rows, then D(B) = −D(A), etc.).
Suppose also that D(In ) = 1. Prove that D(C) = det(C) for all C ∈ Mn (R).
Harder: What if we replace R by an arbitrary field F ?

Solution: Suppose first that C is row equivalent to the identity matrix In . So there
is a sequence of elementary row operations which turns In into C. Then, by induction
on the number of operations used, det(C) = D(C) (the base step, where there are no
operations used and C = In is the assumption D(In ) = 1).
If C is not row equivalent to In , then there is a sequence of elementary row operations
which turns C into a matrix with a row of zeros. Multiplying this row by a non-zero
constant α ̸= 1 we get (α − 1)D(C) = 0.
For an arbitrary field, the same argument will work unless F = {0, 1}: then we cannot
find a non-zero constant α ̸= 1. So the argument will fail in F2 , the field of integers modulo
2. The result also breaks down in this case: for example in M2 (F2 ) we can let D(C) = 1
for all rank 1 matrices C and D(C) = det(C) in other cases, and the required properties
hold.

Question 2 For each of the following linear maps T : V → V , choose a basis B for V
and compute [T ]B . Hence, or otherwise, compute det(T ).
(i) T : R3 → R3 defined by T (x1 , x2 , x3 ) = (−x1 + x2 − x3 , −4x2 + 6x3 , −3x2 + 5x3 ).
(ii) V is the
 vector space of all 2 × 2 matrices over R, and T (A) = M A for all A ∈ V ,
1 −2
where M = .
1 4
(iii) V is the vector space of polynomials over R of degree at most 2, and T (p(x)) =
x(2p(x + 1) − p(x) − p(x − 1)) for all p(x) ∈ V .
 
−1 1 −1
Solution: (i) The matrix of T with respect to the standard basis is  0 −4 6 .
0 −3 5
So the determinant of T is 2.  
        1 −2 0 0
1 0 0 0 0 1 0 0 1 4 0 0 
(ii) Matrix of T w.r.t. basis , , , is A = 
0 0 1 −2.

0 0 1 0 0 0 0 1
0 0 1 4
2
So the determinant is det(M ) = 36.

1
 2 2 2
 (iii) Tsends 1 7→ 0, x 7→ 3x, x 7→ x + 6x , so matrix of T wrt basis 1, x, x is
0 0 0
0 3 1. Thus det(T ) = 0 (which we can also see from the fact that T is singular).
0 0 6

Question 3 Suppose n ≥ 2 and A ∈ Mn (F ). The adjugate matrix adj(A) is the

transpose of the matrix of cofactors of A and we showed that adj(A)A = det(A)In . Give
an expression for adj(adj(A)) in the case where A is invertible.

Solution: Using the given equation, we have adj(adj(A))adj(A) = det(adj(A))In and

(taking determinants) det(adj(A))det(A) = det(A)n . So if det(A) ̸= 0 we obtain that
adj(A) is invertible and

adj(adj(A)) = det(A)n−1 (adj(A))−1 = det(A)n−2 A.

Question 4 Suppose F is a field. Let n ∈ N and a0 , ..., an−1 ∈ F , not all zero. Using
the Vandermonde determinant, prove that the polynomial

f (x) = a0 + a1 x + · · · + an−1 xn−1

has at most n − 1 distinct roots in F , i.e. there are at most n − 1 distinct α ∈ F such
that f (α) = 0.

Solution: Suppose x1 , ..., xn ∈ F are roots, so f (xi ) = a0 + a1 xi + · · · + an−1 xin−1 = 0,

for i = 1, ..., n. Then
     n−1   
1 x1 x1 0
1  x2  xn−1   0 
 2   
a0  . .  + a1  . .  + · · · + an−1  . .  =  . .  .
   
 .  .  .   .
1 xn xn−1
n 0

Hence the columns of the Vandermonde determinant are linearly dependent so the deter-
minant is 0. Hence xi = xj for some i ̸= j.

Question 5 Suppose U, V, W are vector spaces over a field F and T : U → V and

S : V → W are linear transformations. Show that the composition S ◦ T : U → W is a
linear transformation. If U, V, W are finite dimensional with bases B, C, D, prove that

D [S ◦ T ]B = D [S]C C [T ]B .

Solution: To check that S ◦ T is a linear transformation, just use the definition.

For the next part, take a deep breath and repeatedly use a result 4.3.4(?) from last
term:
Let v ∈ V . Then:

D [S ◦ T ]B [v]B = [S(T (v))]D = D [S]C [T (v)]C = D [S]C C [T ]B [v]B .

As the vector [v]B is arbitrary, we obtain the required matrix equality.

2
Question 6 Let V be a vector space over a field F and T : V → V be a linear
transformation. Suppose that λ ∈ F is an eigenvalue of T . Let m ≥ 1 be an integer
and denote by T m the composition T ◦ . . . ◦ T (m times). Note that this is a linear
transformation V → V .
i) Show that λm is an eigenvalue of T m .

ii) If a0 , . . . , am ∈ F are such that a0 Id +a1 T + a2 T 2 + . . . + am T m = 0, show that λ

is a root of the polynomial p(x) = a0 + a1 x + . . . + am xm .

Solution:
i) Suppose 0 ̸= v ∈ V is an eigenvector with eigenvalue λ. Then T (v) = λ(v).
Furthermore (see by induction or otherwise) T m (v) = λm (v). So v is an eigenvector
of T m with eigenvalue λm .

ii) Again, let v be an eigenvector of T with eigenvalue λ. Then

0 = 0v = (a0 Id +a1 T +a2 T 2 +. . .+am T m )v = a0 v+a1 λv+a2 λ2 v+. . .+am λm v = p(λ)v.

As v ̸= 0 this implies that p(λ) = 0.

Question 7 Suppose that T : V → V is a linear map with the property that T (T (v)) =
T (v) for all v ∈ V .
(i) Show that

V = ker(T ) + im(T ) and ker(T ) ∩ im(T ) = {0}.

Hint: Note that if v ∈ V then v = (v − T (v)) + T (v).

(ii) Deduce that if V is of dimension n, then there is a basis B of V such that
 
Is 0s×(n−s)
[T ]B = ,
0(n−s)×s 0(n−s)×(n−s)

where s = dim(im(T )).

Solution: (i) We need to show that ker(T )∩im(T ) = {0} and that V = ker(T )+im(T ).
Suppose that v ∈ ker(T ) ∩ im(T ). Then there exists w ∈ V such that v = T (w). Then

0 = T (v) = (T (T (w))) = T (w) = v

so that ker(T ) ∩ im(T ) = {0}. Now suppose v ∈ V , so that v = (v − T (v)) + T (v). Note
that T (v) ∈ im(T ). Also T (v − T (v)) = T (v) − T 2 (v) = 0 so v − T (v) ∈ ker(T ). So
v ∈ ker(T ) + im(T ).
(ii) Note that all (non-zero) vectors in ker(T ) are eigenvectors with eigenvalue 0.
Moreover, if v ∈ im(T ) there is w ∈ V with T (w) = v, so T (v) = T (T (w)) = T (w) = v.
So all non-zero vectors in im(T ) are eigenvectors with eigenvalue 1. Let v1 , . . . , vt be a
basis for ker(T ) and w1 , . . . , ws a basis for im(T ). By rank + nullity dim(V ) = t + s and
by (i), w1 , . . . , ws , v1 , . . . , vt span V . So these s + t vectors form a basis B for V . The
matrix [T ]B is of the required form.
[Remark: Such a T is called a projection: can you see why?]