Algebra and Matrix Representations of Linear

Transformations

Atanu Kumar

Assistant Professor
Department of Physics
Chandernagore College

August 24, 2020

Contents

I Operations on Linear Transformations

I Operations on Linear Operators
I Algebra of Linear Transformations
I Representation of Linear Transformations by Matrices.
Operations on Linear Transformations
I Let f : U → V and g : U → V are two linear transformations
and α ∈ F is a scalar. then
1. f = g iff f (u) = g (u), ∀u ∈ U.
2. The “addition” (f + g ) : U → V is also an linear
transformation defined by (f + g )(u) = f (u) + g (u), ∀u ∈ U.
3. The “scalar multiplication” (αf ) : U → V is also an linear
transformation defined by (αf )(u) = αf (u), ∀u ∈ U.
I Theorem: Let U and V are two vetor spaces over same field
F . Collection of all linear transformations from U to V form a
vector space Hom(U, V) over F under the operations
“addition” and “scalar multiplications”. Zero Vector O is
defined by O(u) = φ, ∀u ∈ U.
I Let U, V , W are three vector spaces over F . f : U → V and
g : V → W are linear operators. Then their Composition
g ◦ f : U → W is defined by w = g ◦ f (u) = g (v ) ∈ W , for
u ∈ U and v = f (u) ∈ V . g ◦ f is also a linear transformation.
Figure: Composition of Linear Transformations
Solved Problems

Let f : R3 → R2 and g : R3 → R2 be defined by

f (x, y , z) = (y , x + z) and g (x, y ) = (2z, x − y ). Find formulas
defining the mappings f + g and 3f − 2g .

(f + g )(x, y , z) = f (x, y , z) + g (x, y , z)

= (y , x + z) + (2z, x − y )
= (y + 2z, 2x − y + z),
(3f − 2g )(x, y , z) = 3f (x, y , z) − 2g (x, y , z)
= (3y , 3x + 3z) − (4z, 2x − 2y )
= (3y − 4z, x + 2y + 3z).
Let h : R2 → R is defined by h(x, y ) = (x + y ). Using maps f and
g in previous problem, find the formulas defining the mapping
h ◦ f , h ◦ g , and h ◦ (f + g ). Verify that h ◦ (f + g ) = h ◦ f + h ◦ g .
We have f (x, y , z) = (y , x + z), g (x, y ) = (2z, x − y ) and
(f + g )(x, y , z) = (y + 2z, 2x − y + z). So

h ◦ f (x, y , z) = h(y , x + z) = y + x + z,
h ◦ g (x, y , z) = h(2z, x − y ) = 2z + x − y ,
h ◦ (f + g )(x, y , z) = h(y + 2z, 2x − y + z) = 2x + 3z.

Then

h ◦ f (x, y , z) + h ◦ g (x, y , z) = y + x + z + 2z + x − y
= 2x + 3z
= h ◦ (f + g )(x, y , z).
Show that f , g , h ∈Hom(R2 , R2 ), defined by f (x, y ) = (x, 2y ),
g (x, y ) = (y , x + y ), h(x, y ) = (0, x), are linearly independent.
Let us consider the linear combination equating to the zero
operator: af + bg + ch = O:

af (x, y ) + bg (x, y ) + ch(x, y ) = O(x, y )

⇒ a(x, 2y ) + b(y , x + y ) + c(0, x) = (0, 0)
⇒ (ax + by , 2ay + bx + by + cx) = (0, 0),

which leads to two algebric equations:

ax + by = 0, (b + c)x + (2a + b)y = 0.

Putting x = 1, y = 0, we get a = 0 and b + c = 0. Then putting

x = 0, y = 1, we get b = 0, which implies c = 0 also. So
af + bg + ch = O implies a = 0, b = 0, c = 0. So the operators
are linearly independent.
Operations on Linear Operators
Let us consider the linear operators on a vector space V .
I Two operator  and B̂ are equal iff Âv = B̂v , ∀v ∈ V .
I For any two linear operators  and B̂ in the vector space V ,
the sum of the operators  + B̂ is defined as
(Â + B̂)v = Âv + B̂v , ∀v ∈ V .
I The operator (−Â) is defined as (−Â)v = Â(−v ), ∀v ∈ V ,
−v being the additive inverse of v . Then subtraction of the
operators is given by  − B̂ =  + (−B̂).
I The Null Operator (Ô) maps each vector v to the null
vector φ: Ôv = φ, ∀v ∈ V . It can be easily shown that
 −  = Ô, ∀Â.
I The Identity Operator (Iˆ) is defined by Iˆv = v , ∀v ∈ V .
I For any two linear operators  and B̂ in the vector space V ,
the product of the operators ÂB̂ is defied as
ÂB̂v = Â(B̂v ), ∀v ∈ V . If ÂB̂v = B̂ Âv , ∀v ∈ V , then  is
said to commute with B̂. The operator [Â, B̂] = ÂB̂ − B̂ Â is
called the Commutator of  and B̂.
Let us consider two operators  and B̂ in R2 , defined by
       
x x +y x x −y
 = , B̂ = .
y x −y y x +y

Sum  + B̂ is
       
x x +y x −y 2x
(Â + B̂) = + = .
y x −y x +y 2x

Products of the operators ÂB̂ and B̂ Â will be given by

     
x x −y 2x
ÂB̂ = Â = ,
y x +y −2y
     
x x +y 2y
B̂ Â = Â = .
y x −y 2x

So the operators  and B̂ do not commute.

Algebra of Linear Operators

I Space of all linear operators on V form an algebraic structure

A(V ), which is wider than the vector space structure.
I If Â, B̂, Ĉ are linear operators on V on in other words
“vectors” in the algebra A(V ), then their “addition” Â + B̂,
“product”ÂB̂ and “scalar multiplication” αA, exist and
belong to A(V ).
I The operators in A(V ) obey the following rules:
1. Â(B̂ + Ĉ ) = ÂB̂ + ÂĈ , ∀Â, B̂, Ĉ ∈ A(V )
2. (Â + B̂)Ĉ = ÂĈ + B̂ Ĉ , ∀Â, B̂, Ĉ ∈ A(V )
3. α(ÂB̂) = (αÂ)B̂ + Â(αB̂), ∀Â, B̂, ∈ A(V ) and ∀α ∈ F .
I If in addition Â(B̂ Ĉ ) = (ÂB̂)Ĉ holds then A(V ) will be an
associative algebra.
I If dim(V )=n, then dim(A(V ))=n2 .
Let us consider the vector space R2 . General form of a linear
operator  : R2 → R2 is
      
x ax + by a b x
 = = .
y cx + dy c d y

Thus any linear operator on R2 can be represented as 2 × 2 real

matrices. So A(R2 ) is the space of all 2 × 2 real matrices and the
basic operations of A(R2 ) are usual matrix addition, matrix
multiplication and scalar multiplication. The zero operator is
 
0 0
.
0 0

Shortly we will see that linear operators on any n dimensional

vector space Vn can be represented as n × n matrices.
Solved Problems
The operators  and B̂, on the space of functions F, are defined by
Âu(x) = du ˆ
dx and B̂u(x) = xu(x). Show that [Â, B̂] = I and
[Â2 , B̂] = 2Â.

d du du
ÂB̂u = (xu) = u + x , B̂ Âu = x
h i dx x x
ˆ
⇒ Â, B̂ u = ÂB̂u − B̂ Âu = u = I u.

d 2u
Now Â2 u = dx 2
. Then

d2 d 2u
 
2 d du du
 B̂u = (xu) = u + x = 2 + x
dx 2 dx dx dx dx 2
2
d u
B̂ Â2 u = x 2
dx
h i du
⇒ Â2 , B̂ u = 2 = 2Âu.
dx
Matrix Representation of Linear Operators

Let us consider an n dimensional real vector space V and an linear

operator  : V → V . Choose a set of basis vectors
S={e1 , e2 , ..., en } in V . Then any vector u ∈ V can be written as
a linear combination of ei s:
n
X
u = u1 e1 + u2 e2 + ...un en = ui ei ,
i=1

where the scalars ui s are components of u in the basis {ei }.

v = Â(u), the image of u, is similarly expressed in terms of bases
and components:
n
X
v = v1 e1 + v2 e2 + ...vn en = vi ei .
i=1
Since V is isomorphic to Rn , u and v can be represented as
column matrices

   
u1 v1
 u2   v2 
   
. .
[u]S = 
 . ,
 [v ]S = 
 . .

   
. .
un vn

Now since ei s are vectors in V , so are their images Â(ei )s. Hence
each Â(ei ) can be written as a linear combination of ei s:
n
X
Â(ei ) = a1i e1 + a2i e2 + ...ani en = aji ej .
j=1

Let us consider aij s as elements of an n × n matrix [Â]S . To see

that this is the matrix representation of the operator Â, let us
evaluate v = Â(u).
  
n n n n
!
X X X X
Â(u) =  ui ei = ui  (ei ) = ui  aji ej 
i=1 i=1 i=1 j=1
n n
!
X X
= uj aij ei (interchanging i and j)
j=1 i=1
 
n
X n
X
=  aij uj  ei (interchanging the order of summations).
i=1 j=1

Pn
Comparing with v = i=1 vi ei :

n
X
vi = aij uj ,
j=1

which is formula of matrix multiplication. Thus we have

[v ]S = [Â]S [u]S and identify [Â]S as the matrix representation of Â.
If the basis set S = {ei } is orthonormal one, i.e. hei , ej i = δij then
the matrix elements of [Â]S are evaluated as
n
X n
X n
X
hei , Â(ej )i = hei , akj ek i = akj hei , ek i = akj δik = aij .
k=1 k=1 k=1

d
Example: Consider the differential operator dx on P2 . For an
2
arbitrary polynomial a(x) = a0 + a1 x + a2 x ∈ P2 ,

d
b(x) = a(x) = a1 + 2a2 x.
dx
Comparing with b(x) = b0 + b1 x + b2 x 2 , we have b0 = a1 ,
d
b1 = 2a2 , b2 = 0. So the matrix representation of dx in the basis
2
set {1, x, x } is

0 1 0
  !
d
= 0 0 2 .
dx 0 0 0
Solved Problems
Consider the linear operator  on R2 and basis S:
Â(x, y ) = (2x − 7y , 4x + 3y ) and S = {e1 , e2 } = {(1, 0), (1, −1)}.
Find the matrix representation [Â]S . Verify [Â]S [u]S for the vector
u = (4, −3) in R2 .
First let us find the components of an arbitrary vector in the basis
v = (x, y ) in S:
     
x 1 1 x =a+b a=x +y
=a +b , ⇒ , ⇒ .
y 0 −1 y = −b b = −y
 
x +y
⇒ v = (x + y )e1 − ye2 , or [v ]S = .
−y

Then operating  on the basis vectors we get

   
1 2
Âe1 = Â = Â = 6e1 − 4e2 ,
0 4
   
1 9
Âe2 = Â = Â = 10e1 − e2 .
−1 1
So the matrix representation of  in the basis S is
 
h i 6 10
 = .
S −4 −1

Now the matrix representation of u = (4, −3) and Âu are found as
   
4 1
u= = (4 − 3)e1 + 3e2 = e1 + 3e2 ⇒ [u]S = ,
−3 3
     
8 + 21 29 h i 36
Âu = = = 36e1 − 7e2 ⇒ Âu = .
16 − 9 7 S −7

Then calculating [Â]S [u]S ,

    
h i 6 10 1 36
 [u]S = = ,
S −4 −1 3 −7

it is verified that [Â]S [u]S = [Âu]S .

References

1. S P Kuila, “Vector Analysis, Tensor Analysis and Linear

Vector Space”
2. B.S. Vatssa, “Elements of Modern Algebra”
3. Arfken and Weber, ”Mathematical Methods for Physicists”
4. S Lipschutz, “Schaum’s Outlines: Linear Algebra”