An introduction to Linear Algebra, January, 2025

Department of Mathematics and Statistics Indian Institute of Technology - Kanpur

General properties of linear transformations

We shall now focus our attention to the study of linear transformations. These are maps between
vector spaces, over a given field, that preserve the linear structure of a vector space. More precisely:

Definition 0.1. Let (V, F ) and (W, F ) be two vector spaces over. A function L : V −→ W is called
a linear transformation/linear map if

• L(v1 + v2 ) = L(v1 ) + L(v2 ) for all v1 , v2 ∈ V .

• For all α ∈ F and for all v ∈ V , L(αu) = αL(u).

Given V, W ,vector spaces over F , we denote by L(V, W ) the set of all linear transformations
from V into W . The map 0 ∈ L(V, W ), where 0(v) = 0, for all v ∈ V , is called the zero
map/transformation. The set L(V, W ) is a vector space over F under the following addition and
scalar multiplication:


L1 + L2 (v) := L1 (v) + L2 (v) ∀v ∈ V, L1 , L2 ∈ L(V, W ).


αL (v) := αL(v) ∀α ∈ F, v ∈ V.

Examples.

1. Consider V = W = F , then any linear transformation L : F −→ F is given by L(v) = αv for

all v ∈ V where α ∈ F is a fixed scalar.

2. Given a matrix A ∈ Mm,n (F ), consider the map L : F n −→ F m defined by L(x) = Ax for all
x ∈ F n . Note that Ax denoteds the matrix multiplication of A with x where we consider x
as column matrix (x1 , . . . , xn )T . Clearly L(x + y) = A(x + y) = Ax + Ay = L(x) + L(y) and
L(αx) = αx for all α ∈ F and x, y ∈ F n . This shows that L is a linear transformation from
F n to F m . Note that if we write y = L(x), and y = (y1 , . . . , ym ) then
n
X
yi = aij xj for all 1 ≤ i ≤ m
j=1

3. Let V be the space of all real-valued differentiable functions on R and W be the space of all
real-valued functions on R. Note that both are vector spaces over R under the point-wise
addition of functions and pointwise scalar multiplication of a function. Consider the map
D : V −→ W defined by D(f ) = f ′ = the derivative of f . The linearity of D follows from
the fact that if f and g are differentiable functions then so is f + g and αf for any α ∈ R.
 2

4. Recall R[x; n] the vector space of polynomials

h with coefficients in R of degree less than or
iT 
equal to n. Define L : Rn+1 → R[x; n] by L a1 . . . an+1 = a1 + a2 x + · · · + an+1 xn ,
h iT
for each a1 . . . an+1 ∈ Rn+1 . Then L is a linear transformation.

5. Is the map T : R[x; n] → R[x; n + 1] defined by T (f (x)) = xf (x), for all f (x) ∈ R[x; n] a
linear transformation?

Definition 0.2. Let L : V −→ W be a linear transformation between two vector spaces V, W

over F . The kernel of L is the set Ker(L) := {x ∈ V : L(x) = 0}. The image of L is the set
Img(L) = {L(x) : x ∈ V }. Ker(L) and Img(L) are also called the null space of L and the range
space of T respectively.

Ker(L) and Img(L) are subspaces of V and W respectively. When V is finite-dimensional, the
dimension of Ker(L) is called the nullity of L and is denoted by N ullity(L); the dimension of the
image of L or range of L is called the rank of L and is denoted by rank(L). We shall now restrict
our attention to linear transformations between finite-dimensional vector space.

Lemma 0.3. Let V and W be finite-dimensional vector spaces over F and let L1 , L2 ∈ L(V, W ).
Let B be any basis of V such that

L1 (v) = L2 (v) ∀v ∈ B.

Then L1 = L2 , i.e., a linear transformation L is determined by the images of L on a basis of V .

Proof. Let B be a basis of V over F . Then, for each v ∈ V , there exist vectors u1 , . . . , uk in B and
k
P
c1 , . . . , ck ∈ F such that v = ci ui . Thus
i=1

k k k k
!
X X X X
L1 (v) = L1 ci ui = L1 (ci ui ) = ci L1 (ui ) = ci L2 (ui ) = L2 (v) ∀v ∈ V,
i=1 i=1 i=1 i=1

and we are done.

The proof of the above lemma implies the following:

Lemma 0.4. Let V be a finite dimensional vector space over F and W be any vector spaces over F .
If T : V → W is a linear transformation and B is a basis of V then Img(T ) = LS({T (x)|x ∈ B}).

Fix a ∈ Rn and write a = (a1 , . . . , an )and consider f (x) = aT x = a1 x1 + a2 x2 + · · · + an xn , for all

x ∈ Rn . Then by Example (2) above, f ∈ L(Rn , R) is a linear transformation. We now show that
these are the only ones.

Corollary 0.5. Let T ∈ L(Rn , R). Then, there exists a ∈ Rn such that T (x) = aT x.
 3

Proof. By the first lemma, T is determined by {T (e1 ), . . . , T (en )}, for the standard basis {e1 , . . . , en }
h iT
of Rn . So, for 1 ≤ i ≤ n, let T (ei ) = ai , for some ai ∈ R. Now set a := a1 · · · an .
h iT
Then for all x := x1 · · · xn ∈ Rn , we have

n n n
!
X X X
T (x) = T xi ei = xi T (ei ) = xi ai = aT x for all x ∈ Rn .
i=1 i=1 i=1

Thus, the required result follows.

We now prove a very important theorem, namely, the rank-nullity theorem.

Theorem 0.6 (Rank-Nullity Theorem). Let V and W be vector spaces over F and assume that
dimF V is finite. If T ∈ L(V, W ) then

Rank(T ) + Nullity(T ) = dimF Img(T ) + dimF Ker(T ) = dimF V.

Proof. Since Ker(T ) is a subspace of V and V is finite dimensional, Ker(T ) is finite dimensional
and dim(Ker(T )) ≤ dim(V ). Let B = {v1 , . . . , vm } be a basis of Ker(T ). Extend this basis to a set
B1 = {v1 , . . . , vm , w1 , . . . , wn } such that B1 is a basis of V . Clearly dim(V ) = dim(Ker(T )) + n.
Claim. n = dim(Image(T )).
To establish the claim, we shall prove that the set {T (w1 ), . . . , T (wn )} is a basis of Image(T ). Note
that if w ∈ Image(T ) then w = T (v) for some v ∈ V and since B1 = {v1 , . . . , vm , w1 , . . . , wn } is a
basis of V , v = α1 v1 + · · · + αm vm + β1 w1 + · · · + βm wm . Therefore

w = T (v) = T (α1 v1 + · · · + αm vm + β1 w1 + · · · + βm wm )
= α1 T (v1 ) + · · · αm T (vm ) + β1 T (w1 ) + · · · + βm T (wm )
= β1 T (w1 ) + · · · + βm T (wm ).

Since w is arbitrary, we get Image(T ) = LS({T (w1 ), . . . , T (wn )}). Now we show that {T (w1 ), . . . , T (wn )}
is linearly independent. For that let us consider
n
X n
X n
X


βj T (wj ) = 0 =⇒ T βj wj = 0 =⇒ βj wj ∈ Ker(T ). (0.1)
j=1 j=1 j=1

As B = {v1 , . . . , vm } is a basis of Ker(T ), there exist αj ’s such that

n
X m
X
β j wj = αi vi .
j=1 i=1

Since B1 = {v1 , . . . , vm , w1 , . . . , wn } is linearly independent, we get αi = 0 = βj for all i, j. Hence

the set {T (w1 ), . . . , T (wn )} is linearly independent whence is a basis for Image(T ). Therefore,
dim(Image(T)) = n and we are done.
 4

Given vector spaces U, V, W over the field F , we consider L(U, V ) and L(V, W ). Given L1 ∈ L(U, V )
and L2 ∈ L(V, W ), we consider the product L2 ◦ L1 defined by



L2 ◦ L1 (u) := L2 L1 (u) ∀u ∈ U.

Check that L2 ◦ L1 is a linear transformation from U into W , i.e., L2 ◦ L1 L(U, W ). The product of
linear transformation satisfies the distributive properties over the addition and scalar multiplication.

Definition 0.7. Let T ∈ L(V, W ), where V and W are vector spaces over F . We say T is invertible
if there exists a linear transformation S ∈ L(W, V ) such that S ◦ T = IV and T ◦ S = IW , where
IV , IW are the identity maps on V and W , respectively, which map a vector to itself.

Definition 0.8. Let V and W be two finite dimensional vector spaces over F . Then V is said to
∼ W , if there exists an invertible linear transformation T from V
be isomorphic to W , denoted V =
into W .

We record the following lemma:

Lemma 0.9. Let V and W be vector spaces over F . A linear transformation T ∈ L(V, W ) is a
vector space isomorphism if and only if the linear transformation T ∈ L(V, W ) is one-one and onto.

Proof. Note that T ◦ S = IdW implies that T is onto. If T (u) = T (v), then u = ST (u) = ST (v) = v
showing T is one-one.
On the other hand, if T is one-one and onto, for every v ∈ V , then there exists a unique u ∈ V
such that T (u) = v. Thus the inverse function S : V → V by S(v) = T −1 {v}, the pre-image of v
under the function T , is well defined. Moreover, by construction, S ◦ T = T ◦ S = IV (check that
S is a linear transformation!).
h iT
Example 0.10. Let V = { x y z ∈ R3 : z = 0}. Then, any element of V is of the form
h iT h iT h iT
x y 0 . So, we can define a map f : R2 → V by f ( x y ) = x y 0 . Then, it can
be easily verified that f is a linear transformation which is one-one and onto. Hence, f gives an
isomorphism of V and R2 .

In general, we have the following result.

Theorem 0.11. Let V be an n-dimensional vector space over F . Then V ∼

= F n.

Proof. Let {v1 , . . . , vn } be a basisof V and {e1 , . . . , en }, the standard basis of F n . Now define
n n

P P
T (vi ) = ei , for 1 ≤ i ≤ n and T αi vi = αi ei , for α1 , . . . , αn ∈ F . Then, it is easy to
i=1 i=1
observe that T ∈ L(V, F n ), T is one-one and onto. Hence, T is an isomorphism.

Exercise 0.12. Prove that the vector space Mn,m (R) is isomorphic to Rmn .
 5

We now describe various equivalent conditions under which a linear operator on a finite dimensional
vector space is an isomorphism. The proof basically uses the rank-nullity theorem.

Theorem 0.13. Let V be a vector space over F with dimF V = n. Then the following statements
are equivalent for T ∈ L(V ).
1. T is one-one.
2. ker(T ) = {0}.
3. Rank(T ) = n = dimF Img(T ).
4. T is onto.
5. T : V → V is an isomorphism.
6. For any linearly independent set {v1 , . . . , vm } in V , {T (v1 ), . . . , T (vm )} is a linearly indepen-
dent set in V .

Proof. 1 =⇒ 2 : Let T be one-one. Suppose u ∈ ker(T ). Then, T (u) = 0 = T (0). Thus, one-one
implies u = 0.
2 =⇒ 3 : Let ker(T ) = {0}. Rank-nullity theorem gives Rank(T ) = n − nullity(T ) = n.
3 =⇒ 4 : Let n = dimF Img(T ) ≤ dimF V = n. Thus, Img(T ) = V .
4 =⇒ 1 : Let T be onto. Then, dimF Img(T ) = n. The Rank-Nullity theorem gives nullity(T ) = 0
implying Ker(T ) = 0. Hence T (u) = T (v) implies T (u − v) = 0. Thus, u − v ∈ Ker(T ) = {0}
implying u = v.
Thus, we have shown that Parts 1 to 4 are equivalent. Now, let us proceed with the rest of the
equivalent conditions.
1, 2, 3, 4 ⇔ 5 : So, let T be one-one. Then, T is one-one and onto. Thus T gives an isomorphism.
Further, T is an isomorphism implies T is one-one and onto.
1, 2, 3, 4 ⇔ 6 : Let ker(T ) = {0} and {vi , 1 ≤ i ≤ m} be a linearly independent set in V . Consider
the relation
X
xi T (vi ) = 0, xi ∈ F. (0.2)
1≤i≤m
P
This implies T ( 1≤i≤m xi vi ) = T (0) = 0 and hence by our hypothesis on ker(T ), we obtain
P
1≤i≤m xi vi = 0. Thus from the linear independence of vi , 1 ≤ i ≤ m, we deduce that xi = 0 ∀ i
whence the set {T (vi ), 1 ≤ i ≤ m} is linearly independent. If Condition 6 holds then a basis of V
is sent onto a basis of V . Thus, T is onto.