Lecture 2.

Kernel and Image of a Linear Transformation

September 2023

Similarity
A particular case of Change-of-Basis Theorem (see Lecture 1) is so important that it is
worth stating separately. It corresponds to the case in which m = n and B 0 = B.
COROLLARY 2.1 (Similarity Theorem) Let f : V → V be a linear operator on
V , and E, B be two bases of V . If A is a matrix representing f with respect to basis E,
then, with respect to the basis B, the operator f is represented by the matrix

à = P −1 AP,

where P is a transition matrix from E to B.

The relationship between the matrices à and A is a central one in the theory of linear
algebra. The matrix à performs the same linear transformation as the matrix A, but Ã
describes it in terms of the basis B. This effect inspires the following definition.
DEFINITION 2.2 (Similarity) We say that the square matrix C is similar to the
matrix A if there is an invertible matrix P such that C = P −1 AP .
Similarity defines an equivalence relation on matrices. Recall that an equivalence relation
satisfies three properties; it is reflexive, symmetric and transitive. For similarity, this means:
• a matrix A is similar to itself (reflexive),
• if C is similar to A, then A is similar to C (symmetric), and
• if D is similar to C, and C to A, then D is similar to A (transitive).
Because the relationship is symmetric, we usually just say that A and C are similar
matrices, meaning one is similar to the other, and we can express this either as C =
P −1 AP or A = Q−1 CQ for invertible matrices P and Q (in which case Q = P −1 ).
Similar matrices share many properties, some of which are collected in the next theorem.
THEOREM 2.3 If A and C are similar n × n-matrices, then A and C have the same
determinant, rank, and trace.
Proof. 1) Let C = P −1 AP for some invertible matrix P . Then we have

det C = det(P −1 ) det A det P = (1/ det P ) det A det P = det A.

2) Since rank(AB) 6 rank A, it follows that rank C 6 rank A and rank A 6 rank C, so
they are equal.
3)PWrite A = [aij ] and B = [bij ]. For each i, the (i, i)-entry of the matrix AB is
di = j aij bji . Hence
!
X X
tr(AB) = d1 + d2 + · · · + dn = aij bji = tr(BA).
i j
Therefore, tr C = tr(P −1 AP ) = tr(AP P −1 ) = tr A.
EXAMPLE 2.4 By computing similarity invariants (trace, determinant, and rank)
show that A and B are not similar in each case.
       
1 2 1 1 2 1 3 0
(a) A = , B= ; (b) A = , B= .
2 1 −1 1 1 −1 1 −1
Solution. (a) tr A = tr B = 2, rank A = rank B = 2, but det A 6= det B, so A and B
are not similar.
(b) det A = det B = −3, rank A = rank B = 2, but tr A 6= tr B, so A and B are not
similar.
Definition of Kernel and Image
Now we will introduce two important subspaces associated with a linear transformation
f : V → W.
DEFINITION 2.5 The kernel of f (denoted Ker f ) and the image of f (denoted
Im f or f (V )) are defined by
Ker f = {v ∈ V | f (v) = 0}, Im f = f (V ) = {w ∈ V | ∃v ∈ V : w = f (v)}.
The kernel of f is often called the null space of f . It consists of all vectors v in V
satisfying the condition f (v) = 0. The image of f is often called the range of f and
consists of all vectors w in W of the form w = f (v) for some v in V .
THEOREM 2.6 Let f : V → W be a linear transformation.
1. Ker f is a subspace of V .
2. Im f is a subspace of W .
Proof. The fact that f (0) = 0 shows that Ker f and Im f contain the zero vector of V
and W respectively.
1. If v and v1 lie in Ker f , then f (v) = 0 = f (v1 ), so
f (v + v1 ) = f (v) + f (v1 ) = 0 + 0 = 0
f (kv) = kf (v) = k · 0 = 0 for all k ∈ R.
Hence v+v1 and kv lie in Ker f (they satisfy the required condition), so Ker f is a subspace
of V since it is closed under addition and scalar multiplication.
2. If w and w1 lie in Im f , write w = f (v) and w1 = f (v1 ) where v, v1 ∈ V . Then
w + w1 = f (v) + f (v1 ) = f (v + v1 )
kw = kf (v) = f (kv) for all k ∈ R.
Hence w + w1 and kw both lie in Im f (they have the required form), so Im f is a subspace
of W .
EXAMPLE 2.7 Let Pn denote the vector space of polynomials of degree 6 n. Prove
that differentiation D : Pn → Pn that transforms a polynomial p(x) into the polynomial
p0 (x) is a linear operator. Find the kernel and the image of D. Find the matrix of D with
respect to the following bases:
(a) 1, x, x2 , . . . , xn ;
(b) 1, x − x0 , . . . , (x − x0 )n , where x0 is some fixed point;
x xn
(c) 1, x, , . . . , .
2! n!
Solution. The kernel consists of constant polynomials (of degree 0), since only constant
polynomials vanish under differentiation. The image is Pn−1 , as differentiation reduces the
degree of any polynomial by 1.
To find the matrices of D in different bases, we keep in mind that the columns of these
matrices are images of basis vectors under D. So if we use basis (a), we find: D(1) = 0
yields the zero first column, D(x) = 1 yields the second column (1, 0, . . . , 0)T , D(x2 ) = 2x
yields the third column (0, 2, 0, . . . , 0)T , etc. As a result, we obtain the answer:
     
0 1 0 ... 0 0 1 0 ... 0 0 1 0 ... 0
0 0 2 . . . 0  0 0 2 . . . 0  0 0 1 . . . 0
 .. .. .. . . ..   .. .. .. . . ..   .. .. .. . . .. 
     
(a)  . . . . .  ; (b)  . . . . .  ; (c)  . . . . . .
     
0 0 0 . . . n  0 0 0 . . . n  0 0 0 . . . 1
0 0 0 ... 0 0 0 0 ... 0 0 0 0 ... 0

DEFINITION 2.8 Given a linear transformation f : V → W :

• dim(Ker f ) is called the nullity of f and denoted as nullity f
• dim(Im f ) is called the rank of f and denoted as rank f
The rank of a matrix A was defined earlier to be the dimension of col A, the column
space of A. The two usages of the word rank are consistent in the following sense.
THEOREM 2.9 Let fA be a matrix transformation induced by an m × n matrix A.
Then
1. Im fA = col A;
2. rank fA = rank A.
Proof. Since fA (x) = Ax, then

Im fA = {Ax | x ∈ Rn } = {x1 c1 + · · · + xn cn | xi ∈ R} = col A.

Therefore,
rank A = dim(col A) = dim(row A).
One-to-One and Onto Transformations
DEFINITION 2.10 Let f : V → W be a linear transformation.
• f is said to be onto (or surjective) if Im f = W .
• f is said to be one-to-one (or injective) if f (v) = f (v1 ) implies v = v1 .
 EXAMPLE 2.11 The identity operator 1V : V → V is both one-to-one and onto for
any vector space V .
A vector w in W is said to be hit by f if w = f (v) for some v in V . Then f is onto if
every vector in W is hit at least once, and f is one-to-one if no element of W gets hit twice.
Clearly the onto transformations f are those for which Im f = W is as large a subspace of
W as possible. By contrast, the next theorem shows that the one-to-one transformations f
are the ones with Ker f as small a subspace of V as possible.
THEOREM 2.12 If f : V → W is a linear transformation, then f is one-to-one if and
only if Ker f = {0}.
Proof. If f is one-to-one, let v be any vector in Ker f . Then f (v) = 0, so f (v) = f (0).
Hence v = 0 because f is one-to-one. Hence Ker f = {0}.
Conversely, assume that Ker f = {0} and let f (v) = f (v1 ) with v and v1 in V . Then
f (v − v1 ) = f (v) − f (v1 ) = 0, so v − v1 lies in Ker f = {0}. This means that v − v1 = 0,
so v = v1 , proving that f is one-to-one.
EXAMPLE 2.13 Consider the linear transformations
• f : R3 → R2 given by f (x, y, z) = (x + y + z, x − y)
• g : R2 → R3 given by g(x, y) = (x, x + y, x − y)
Show that f is onto but not one-to-one, whereas g is one-to-one but not onto.
Solution. The verification that f and g are linear is omitted.
Transformation f is not one-to-one because (1, 1, −2) lies in Ker f . But every element
(s, t) in R2 lies in Im f because (s, t) = (x + y + z, x − y) = f (x, y, z) for some x, y, and
z (for example, x = 12 (s + t), y = 12 (s − t), and z = 0). Hence f is onto.
Transformation g is one-to-one because
Ker g = {(x, y)| x + y = x − y = x = 0} = {(0, 0)}.
However, it is not onto. For example (1, 0, 0) does not lie in Im g because if (1, 0, 0) = (x, x+
y, x − y) for some x and y, then x + y = 0 = x − y and x = 1, which is contradictory.
EXAMPLE 2.14 Let U be an invertible m × m matrix and define f : Mm×n → Mm×n
by f (X) = U X for all X ∈ Mm×n . Show that f is a linear transformation that is both
one-to-one and onto.
Solution. The verification that f is linear is left to the reader. To see that f is one-
to-one, let f (X) = 0. Then U X = 0, so left-multiplication by U −1 gives X = 0. Hence
Ker f = {0}, so f is one-to-one. Finally, if Y is any member of Mm×n , then U −1 Y lies in
Mm×n too, and f (U −1 Y ) = U (U −1 Y ) = Y . This shows that f is onto.
EXAMPLE 2.15 Find the matrix of projection of R3 onto L = {(x, y, z) ∈ R3 | x = 0}
along M = {(x, y, z) ∈ R3 | 2x = 2y = −z}.
Solution. Since R3 = L ⊕ M , any vector x ∈ R3 can be uniquely represented as
x = l + m, where l ∈ L, m ∈ M , so that f (x) = l. Thus we need to find the formula of
computing l from x. We see that M = span{(1, 1, −2)T }. Therefore, we have
     
x1 0 1
x = x2 = l2 + k 1  .
    
x3 l3 −2
It is clear that k = x1 , l2 = x2 − x1 , l3 = x3 + 2x1 . Thus we obtain the answer:
        
x1 0 0 0 0 x1 0 0 0
f (x) = f x2  = −x1 + x2  = −1 1 0 x2  =⇒ A = −1 1 0 .
x3 2x1 + x3 2 0 1 x3 2 0 1

Exercises

1. Two bases B = {b1 , b2 } and E = {e1 , e2 } are related

 by e1 = 2b1 + b2 , e2 = 3b1 + b2 .
1 2
A linear operator f has the matrix AB = with respect to the basis B. Find the
−4 2
matrix of this operator
 with respect
 tothe
 basis E. 
−1 3 1 2 2 3 −22 −35
Answer. AE = = .
1 −2 −4 2 1 1 16 25
2. Two bases B = {b1 , b2 } and E = {e1 ,  e2 } are related
 by b1 = e1 + e2 , b2 = 3e1 + 4e2 .
−20 −69
A linear operator f has the matrix AB = with respect to the basis B. Find
5 17
the matrix of this operator
 with respectto the basis E.  
1 3 −20 −69 4 −3 −2 −3
Answer. AE = = .
1 4 5 17 −1 1 1 −1
3. Suppose f is linear. Find a basis of Ker f and a basis of Im f .
(a) f : P 2 → R2 ;
f (p(x)) = (p(0), p(1))
(b) f : P 2 → R2 ;
f (a + bx + cx2 ) = (a, b)
(c) f : R3 → R3 ; f (x, y, z) = (x + y, x + y, 0)
(d) f : R3 → R4 ; f (x, y,
 z) =(x, x,
 y, y) 
a b a+b b+c
(e) f : M22 → M22 ; T =
c d c+d d+a
 
a b
(f) f : M22 → R; T =a+d
c d
 
0 1
(i) f : M22 → M22 ; f (X) = XA − AX, where A =
1 0
Answer. (a) {x2 − x}, {(1, 0), (0, 1)}, (b) {x2 }, {(1, 0), (0, 1)}
(c) {(0, (1, −1,
 0, 1),    1,0)},  (d) {(0,
0)}; {(1, 0, 1)}; {(1, 1, 0, 0), (0, 0, 1, 1)}
1 −1 1 0 1 1 0 1
(e) , , ,
1 −1 0 1 0 0 1 0
     
1 0 0 1 0 0
(f) , , , {1}
0 −1 0 0 1 0
       
0 1 1 0 1 0 0 1
(i) , , ,
1 0 0 1 0 −1 −1 0