MAT3701 - Linear Algebra III

Linear Transformations, Null Spaces, and Ranges

In this section, we consider a number of examples of linear transformations. Many of these transformations are studied
in more detail in later sections.

Recall from Appendix B that a function T with domain V and codomain W is denoted by
T : V → W.
Definition (Linear Operator)

Let V and W be vector spaces (over F). We call a function T : V → W a linear transformation from V to W if,
for all x, y ∈ V and c ∈ F, we have

(a) T (x + y) = T (x) + T (y) and

(b) T (cx) = cT (x).

If the underlying field F is the field of rational numbers, then (a) implies (b) (see Exercise 37), but, in general (a) and
(b) are logically independent. See Exercises 38 and 39.

We often simply call T linear.

Exercises

Verify the following properties of a function T : V → W. (See Exercise 7.)

1) If T is linear, then T (0) = 0.

2) T is linear if and only if T(cx + y) = cT (x) + T (y) for all x, y ∈ V and c ∈ F.

3) If T is linear, then T (x − y) = T (x) − T (y) for all x, y ∈ V.

4) T is linear if and only if, for x1 , x2 , . . . , xn ∈ V and a1 , a2 , . . . , an ∈ F, we have

n
! n
X X
T ai xi = ai T (xi ).
i=1 i=1

We generally use property 2 to prove that a given transformation is linear.

Example 1

Define
T : R2 → R2 by T (a1 , a2 ) = (2a1 + a2 , a1 ).
To show that T is linear, let c ∈ R and x, y ∈ R2 , where x = (b1 , b2 ) and y = (d1 , d2 ). Since
cx + y = (cb1 + d1 , cb2 + d2 ),
we have
T (cx + y) = (2 (cb1 + d1 ) + cb2 + d2 , cb1 + d1 ).
Also

cT (x) + T (y) = c (2b1 + b2 , b1 ) + (2d1 + d2 , d1 )

= (2cb1 + cb2 + 2d1 + d2 , cb1 + d1 )
= (2 (cb1 + d1 ) + cb2 + d2 , cb1 + d1 ) .

1
So T is linear. □

As we will see in Chapter 6, the applications of linear algebra to geometry are wide and varied. The main reason for
this is that most of the important geometrical transformations are linear. Three particular transformations that we
now consider are rotation, reflection, and projection. We leave the proofs of linearity to the reader.

Example 2

For any angle θ, define Tθ : R2 → R2 by the rule: Tθ (a1 , a2 ) is the vector obtained by rotating (a1 , a2 ) counterclockwise
by θ if (a1 , a2 ) ̸= (0, 0), and Tθ (0, 0) = (0, 0). Then Tθ : R2 → R2 is a linear transformation that is called the rotation
by θ.

We determine an explicit formula for Tθ .

2
p vector (a1 , a2 ) ∈ R . Let α be the angle that (a1 , a2 ) makes with the positive x-axis (see Figure 2.1(a)),
Fix a nonzero
2 2
and let r = a1 + a2 .

Then a1 = r cos α and a2 = r sin α.

Also, Tθ (a1 , a2 ) has length r and makes an angle α + θ with the positive x-axis. It follows that

Tθ (a1 , a2 ) = (r cos(α + θ), r sin(α + θ))

= (r cos α cos θ − r sin α sin θ, r cos α sin θ + r sin α cos θ)
= (a1 cos θ − a2 sin θ, a1 sin θ + a2 cos θ) .

Finally, observe that this same formula is valid for (a1 , a2 ) = (0, 0).

It is now easy to show, as in Example 1, that Tθ is linear. □

Example 3

Define T : R2 → R2 by T (a1 , a2 ) = (a1 , −a2 ).

T is called the reflection about the x-axis. (See Figure 2.1(b).) □

Example 4

Define T : R2 → R2 by T (a1 , a2 ) = (a1 , 0).

T is called the projection on the x-axis. (See Figure 2.1(c).) □

We now look at some additional examples of linear transformations.

Example 5

Define T : Mm×n (F) → Mn×m (F) by T (A) = At , where At is the transpose of A, defined in Section 1.3.

Then T is a linear transformation by Exercise 3 of Section 1.3. □

Example 6

Define T : Pn (R) → Pn−1 (R) by T (f (x)) = f ′ (x), where f ′ (x) denotes the derivative of f (x).

2
To show that T is linear, let g(x), h(x) ∈ Pn (R) and a ∈ R.

Now
T (ag(x) + h(x)) = (ag(x) + h(x))′ = ag ′ (x) + h′ (x) = aT (g(x)) + T (h(x)).
So by property 2 above, T is linear. □

Example 7

Let V = C(R), the vector space of continuous real-valued functions on R. Let a, b ∈ R, a < b. Define T : V → R by
Z b
T (f ) = f (t)dt
a

for all f ∈ V. Then T is a linear transformation because the definite integral of a linear combination of functions is
the same as the linear combination of the definite integrals of the functions. □

Two very important examples of linear transformations that appear frequently in the remainder of the book, and
therefore deserve their own notation, are the identity and zero transformations.

For vector spaces V and W (over F), we define

the identity transformation

IV : V → V by IV (x) = x for all x ∈ V
and the zero transformation

T0 : V → W by T0 (x) = 0 for all x ∈ V.

It is clear that both of these transformations are linear. We often write I instead of IV .

We now turn our attention to two very important sets associated with linear transformations: the range and null
space.

The determination of these sets allows us to examine more closely the intrinsic properties of a linear transformation.

Definitions

Let V and W be vector spaces, and let T : V → W be linear. We define the null space (or kernel) N(T ) of T to be
the set of all vectors x in V such that T (x) = 0; that is,
N(T ) = {x ∈ V : T (x) = 0}

We define the range (or image) R(T ) of T to be the subset of W consisting of all images (under T ) of vectors in V;
that is,
R(T ) = {T (x) : x ∈ V}.
□

Example 8

Let V and W be vector spaces, and let I : V → V and T0 : V → W be the identity and zero transformations,
respectively. Then
N(I) = {0} and R(I) = V,
and

N(T0 ) = V and R(T0 ) = {0}.

□

3
Example 9

Let T : R3 → R2 be the linear transformation defined by

T (a1 , a2 , a3 ) = (a1 − a2 , 2a3 ).
It is left as an exercise to verify that
N(T ) = {(a, a, 0) : a ∈ R} and R(T ) = R2 .
□

In Examples 8 and 9, we see that the range and null space of each of the linear transformations is a subspace.

The next result shows that this is true in general.

Remark

Up to now we have been careful to distinguish the zero in the field from the zero in the vector space by using different
notations - 0 for the zero in the field and 0 for the zero in the vector space. From now we will use a single notation, 0,
to represent both. In each occurrence of the symbol, context will make it clear which of the two is being referred to.

Theorem 2.1

Let V and W be vector spaces and T : V → W be linear. Then

(a) N(T ) is a subspace of V and

(b) R(T ) is a subspace of W.

Proof

To clarify the notation, we use the symbols 0V and 0W to denote the zero vectors of V and W, respectively.

We use the subspace theorem, Theorem 1.3, to prove the two parts of the theorem, as follows.

Since T (0V ) = 0W , we have that 0V ∈ N(T ).

Next, let x, y ∈ N(T ) and c ∈ F.

Then

T (x + y) = T (x) + T (y) (T is linear)

= 0W + 0W (x, y ∈ N(T ))
= 0W ,

and

T (cx) = c T (x) (T is linear)

= c 0W (x ∈ N(T ))
= 0W .

Hence x + y ∈ N(T ) and cx ∈ N(T ), so that N(T ) is a subspace of V.

Next we prove the second part of the theorem, again using the subspace theorem.

Because T (0V ) = 0W , we have that 0W ∈ R(T ).

Now let x, y ∈ R(T ) and c ∈ F. Then there exist v and w in V such that
T (v) = x and T (w) = y.
So
T (v + w) = T (v) + T (w) = x + y, and

4
 T (cv) = cT (v) = cx.

Thus x + y ∈ R(T ) and cx ∈ R(T ), so R(T ) is a subspace of W. □

The next theorem provides a method for finding a spanning set for the range of a linear transformation. With this
accomplished, a basis for the range is easy to discover using the technique of Example 6 of Section 1.6.

Theorem 2.2

Let V and W be vector spaces, and let T : V → W be linear. If β = {v1 , v2 , . . . , vn } is a basis for V, then
R(T ) = span(T (β)) = span ({T (v1 ) , T (v2 ) , . . . , T (vn )}).
Proof

Clearly T (vi ) ∈ R(T ) for each i. Because R(T ) is a subspace,

R(T ) contains span ({T (v1 ) , T (v2 ) , . . . , T (vn )}) = span(T (β))
by Theorem 1.5. This proves that span(T (β)) ⊆ R(T ). Next we prove that R(T ) ⊆ span(T (β)).

Now suppose that w ∈ R(T ). Then w = T (v) for some v ∈ V. Because β is a basis for V, we have
n
X
v= ai v i for some a1 , a2 , . . . , an ∈ F.
i=1

Since T is linear, it follows that

n
X
w = T (v) = ai T (vi ) ∈ span(T (β)).
i=1

So R(T ) is contained in span(T (β)). Hence R(T ) ⊆ span(T (β)).

Since span(T (β)) ⊆ R(T ) and R(T ) ⊆ span(T (β)), we have R(T ) = span(T (β)) □

It should be noted that Theorem 2.2 is true if β is infinite, that is, R(T ) = span({T (v) : v ∈ β}). (See Exercise 33.)

The next example illustrates the usefulness of Theorem 2.2.

Example 10

Define the linear transformation T : P2 (R) → M2×2 (R) by

 
f (1) − f (2) 0
T (f (x)) = .
0 f (0)

Since β = 1, x, x2 is a basis for P2 (R), we have

R(T ) = span(T (β)) = span T (1), T (x), T x2




     
0 0 −1 0 −3 0
= span , ,
0 1 0 0 0 0
   
0 0 −1 0
= span , .
0 1 0 0

Thus we have found a basis for R(T ), and so dim(R(T )) = 2. □

5
As in Chapter 1, we measure the ”size” of a subspace by its dimension. The null space and range are so important
that we attach special names to their respective dimensions.

Definitions

Let V and W be vector spaces, and let T : V → W be linear.

If N(T ) and R(T ) are finite-dimensional, then we define the nullity of T as follows
nullity(T ) = dim N(T )

and we define the rank of T as

rank(T ) = dim R(T ).
□ Reflecting on the action of a linear transformation, we see intuitively that the larger the nullity, the smaller the
rank. In other words, the more vectors that are carried into 0 , the smaller the range. The same heuristic reasoning
tells us that the larger the rank, the smaller the nullity. This balance between rank and nullity is made precise in the
next theorem, appropriately called the dimension theorem.

Theorem 2.3 (Dimension Theorem)

Let V and W be vector spaces, and let T : V → W be linear. If V is finite-dimensional, then

nullity(T ) + rank(T ) = dim(V).

Proof

Suppose that dim(V) = n, dim(N(T )) = k, and {v1 , v2 , . . . , vk } is a basis for N(T ). By the corollary to Theorem 1.11, we
may extend {v1 , v2 , . . . , vk } to a basis β = {v1 , v2 , . . . , vn } for V. We claim that S = {T (vk+1 ) , T (vk+2 ) , . . . , T (vn )}
is a basis for R(T ).

First we prove that S generates R(T ). Using Theorem 2.2 and the fact that T (vi ) = 0 for 1 ≤ i ≤ k, we have

R(T ) = span ({T (v1 ) , T (v2 ) , . . . , T (vn )}

= span ({T (vk+1 ) , T (vk+2 ) , . . . , T (vn )} = span(S).

Now we prove that S is linearly independent. Suppose that

n
X
bi T (vi ) = 0 for bk+1 , bk+2 , . . . , bn ∈ F.
i=k+1

Using the fact that T is linear, we have

n
!
X
T bi vi = 0.
i=k+1

So
n
X
bi vi ∈ N(T ).
i=k+1

Hence there exist c1 , c2 , . . . , ck ∈ F such that

n
X k
X k
X n
X
bi v i = ci vi or (−ci ) vi + bi vi = 0.
i=k+1 i=1 i=1 i=k+1

Since β is a basis for V, we have bi = 0 for all i. Hence S is linearly independent. Notice that this argument also
shows that T (vk+1 ) , T (vk+2 ) , . . . , T (vn ) are distinct; therefore rank(T ) = n − k. □

If we apply the dimension theorem to the linear transformation T in Example 9 , we have that nullity(T ) + 2 = 3, so
nullity(T ) = 1.

6
The reader should review the concepts of ”one-to-one” and ”onto” presented in Appendix B. Interestingly, for a linear
transformation, both of these concepts are intimately connected to the rank and nullity of the transformation. This
is demonstrated in the next two theorems.

Theorem 2.4

Let V and W be vector spaces, and let T : V → W be linear. Then T is one-to-one if and only if N(T ) = {0}.

Proof

We will show:
If T is one to one then N(T ) = {0}. (1)
and
If N(T ) = {0} then T is one to one. (2)
To prove that (1) is the case suppose that T is one to one. We will show that N(T ) = {0}.

We’ll do that by showing that

{0} ⊆ N(T ) (∗).

and
N(T ) ⊆ {0} (∗∗)

To show (∗), from the exercises above, since T is linear we have that T (0) = 0. This proves (∗).

To see (∗∗), let x ∈ N(T ). Then T (x) = 0. Since T is linear, we have that T (0) = 0 (same exercise as before in (∗)).
Hence T (x) = T (0). Since T is one to one we have that x = 0. Hence we’ve shown that N(T ) ⊆ {0}. From (∗) and
(∗∗) we have (1).

Now assume that N(T ) = {0}, and suppose that T (x) = T (y). Then 0 = T (x) − T (y) = T (x − y) (exercises above)
since T is linear. Therefore x − y ∈ N(T ) = {0}. So x − y = 0, or x = y. This means that T is one-to-one. □

The reader should observe that Theorem 2.4 allows us to conclude that the transformation defined in Example 9 is
not one-to-one.

Surprisingly, the conditions of one-to-one and onto are equivalent in an important special case.

Theorem 2.5

Let V and W be vector spaces of equal (finite) dimension, and let T : V → W be linear. Then the following are
equivalent.

(a) T is one-to-one.

(b) T is onto.

(c) rank(T ) = dim(V).

Proof

From the dimension theorem, we have

nullity(T ) + rank(T ) = dim(V).

Now, with the use of Theorem 2.4, we have that T is one-to-one if and only if N(T ) = {0}, if and only if nullity(T ) = 0, if
and only if rank(T ) = dim(V), if and only if rank(T ) = dim(W), and if and only if dim(R(T )) = dim(W). By Theorem
1.11, this equality is equivalent to R(T ) = W, the definition of T being onto. □

We note that if V is not finite-dimensional and T : V → V is linear, then it does not follow that one-to-one and onto
are equivalent. (See Exercises 15, 16, and 21.)

The linearity of T in Theorems 2.4 and 2.5 is essential, for it is easy to construct examples of functions from R into
R that are not one-to-one, but are onto, and vice versa.

7
The next two examples make use of the preceding theorems in determining whether a given linear transformation is
one-to-one or onto.

Example 11

Let T : P2 (R) → P3 (R) be the linear transformation defined by

Z x
T (f (x)) = 2f ′ (x) + 3f (t) dt.
0

Now
3x, 2 + 23 x2 , 4x + x3




R(T ) = span T (1), T (x), T x2 = span .
Since 3x, 2 + 23 x2 , 4x + x3 is linearly independent, rank(T ) = 3.


Since dim (P3 (R)) = 4, T is not onto.

From the dimension theorem, nullity(T )+ 3 = 3. So nullity(T ) = 0, and therefore, N(T ) = {0}. We conclude from
Theorem 2.4 that T is one-to-one. □

Example 12

Let T : F2 → F2 be the linear transformation defined by

T (a1 , a2 ) = (a1 + a2 , a1 ).
It is easy to see that N(T ) = {0}. So T is one-to-one. Hence Theorem 2.5 tells us that T must be onto. □

In Exercise 14, it is stated that if T is linear and one-to-one, then a subset S is linearly independent if and only if
T (S) is linearly independent. Example 13 illustrates the use of this result.

Example 13

Let T : P2 (R) → R3 be the linear transformation defined by

T a0 + a1 x + a2 x2 = (a0 , a1 , a2 ).

Clearly T is linear and one-to-one. Let S = 2 − x + 3x2 , x + x2 , 1 − 2x2 . Then S is linearly independent in P2 (R)
because
T (S) = {(2, −1, 3), (0, 1, 1), (1, 0, −2)}
is linearly independent in R3 . □

In Example 13, we transferred a property from the vector space of polynomials to a property in the vector space of
3-tuples. This technique is exploited more fully later.

One of the most important properties of a linear transformation is that it is completely determined by its action on a
basis. This result, which follows from the next theorem and corollary, is used frequently throughout the book.

Theorem 2.6

Let V and W be vector spaces over F, and suppose that {v1 , v2 , . . . , vn } is a basis for V. For w1 , w2 , . . . , wn in W,
there exists exactly one linear transformation T : V → W such that T (vi ) = wi for i = 1, 2, . . . , n.

Proof

Let x ∈ V. Then
n
X
x= ai vi ,
i=1

where a1 , a2 , . . . , an are unique scalars. Define

n
X
T : V → W by T (x) = ai wi .
i=1

8
(a) T is linear: Suppose that u, v ∈ V and d ∈ F. Then we may write
n
X n
X
u= bi vi and v = ci vi
i=1 i=1

for some scalars b1 , b2 , . . . , bn , c1 , c2 , . . . , cn . Thus

n
X
du + v = (dbi + ci ) vi
i=1

So
n
X n
X n
X
T (du + v) = (dbi + ci ) wi = d bi w i + ci wi = dT (u) + T (v).
i=1 i=1 i=1

(b) Clearly
T (vi ) = wi for i = 1, 2, . . . , n.
(c) T is unique: Suppose that U : V → W is linear and U (vi ) = wi for i = 1, 2, . . . , n. Then for x ∈ V with
n
X
x= ai vi
i=1

we have
n
X n
X
U(x) = ai U (vi ) = ai wi = T (x).
i=1 i=1

Hence U = T . □

Corollary

Let V and W be vector spaces, and suppose that V has a finite basis {v1 , v2 , . . . , vn }. If U, T : V → W are linear and
U (vi ) = T (vi ) for i = 1, 2, . . . , n, then U = T . □

Example 14

Let T : R2 → R2 be the linear transformation defined by

T (a1 , a2 ) = (2a2 − a1 , 3a1 ),

and suppose that U : R2 → R2 is linear. If we know that U(1, 2) = (3, 3) and U(1, 1) = (1, 3), then U = T . This
follows from the corollary and from the fact that {(1, 2), (1, 1)} is a basis for R2 . □

Exercises

Attempt Exercises:

1 a - h, 2, 3, 13, 16, 21.