Algebra-III

Ashok Bingi

Department of Mathematics
St. Xavier’s College-Autonomous, Mumbai

15 and 20 June 2022

Ashok Bingi SMAT0302 15 and 20 June 2022 1/6

Unit 1: Linear Transformations and Matrices

Linear transformation: Let U and V be vector spaces over R.

A function T:U−→ V is said to be a linear map/transformation if
(1) T(u1 + u2 )=T(u1 )+ T(u2 ) ∀u1 , u2 ∈ U
(2) T(αu)=α T(u) ∀α ∈ R and ∀u ∈ U
OR T(α1 u1 +α2 u2 )=α1 T(u1 ) + α2 T(u2 ) ∀α1 , α2 ∈ R and ∀u1 , u2 ∈ U
OR T(α1 u1 + u2 )=α1 T(u1 ) + T(u2 ) ∀α1 ∈ R and ∀u1 , u2 ∈ U
Kernel of T=Ker(T)= {u ∈ U|T (u) = 0v } It is subspace(null space)of U
Range of T=R(T)= {T (u) ∈ V |u ∈ U} It is subspace (range space) of V
If Ker(T) is finite dimensional then dim(Ker(T))=nullity of T = n(T)
If R(T) is finite dimensional then dim(R(T))=rank of T = r(T)
Rank-Nullity Theorem: Let T:U−→ V be linear map and
U be finite dimensional. Then r(T) + n(T)= dim(U)

Ashok Bingi SMAT0302 15 and 20 June 2022 2/6

Proof of Rank-Nullity Theorem
Obviously, ker(T) is itself a finite dimensional vector space(subspace of U).
Let dim(ker(T))=n and dim(U)= p. Then p ≥ n
Let B={u1 , · · · , un } be basis for ker(T). Clearly, B is LI in U
By extension theorem, this LI set B can be extended to form basis for U
Let B1 = {u1 , · · · , un , un+1 , · · · , up } be basis for U.
Consider the set A = {T (un+1 ), · · · , T (up )}. We show A is basis for R(T)
Claim-1: A spans R(T) i.e;[A]=R(T) Clearly, [A] ⊆ R(T )
Let v ∈ R(T ).Then v= T(u) for some u∈ U =[u1 , · · · , un , un+1 , · · · , up ].
So u= a1 u1 + a2 u2 + · · · + an un + an+1 un+1 · · · + ap up
for some scalars a1 , a2 , · · · , an , an+1 , · · · , ap .
Hence v = T (u) = T (a1 u1 + a2 u2 + · · · + an un + an+1 un+1 + · · · + ap up )
v = a1 T (u1 ) + a2 T (u2 ) + · · · + an T (un ) + an+1 T (un+1 ) + · · · + ap T (up )
v = 0 + an+1 T (un+1 ) + · · · + ap T (up ) as T (u1 ) = · · · = T (un ) = 0
v ∈ [T (un+1 ), · · · , T (up )] i.e. v ∈ [A]
Thus v ∈ R(T ) =⇒ v ∈ [A]
So R(T ) ⊆ [A] and therefore R(T)=[A]
Ashok Bingi SMAT0302 15 and 20 June 2022 3/6
Claim-2: A = {T (un+1 ), · · · , T (up )} is LI
an+1 T (un+1 ) + · · · + ap T (up ) = 0V for some scalars an+1 , · · · , ap
T (an+1 un+1 + · · · + ap up ) = 0V as T is linear
=⇒ an+1 un+1 + · · · + ap up = x ∈ Ker (T ) = [B] = [u1 , · · · , un ]
So an+1 un+1 + · · · + ap up = b1 u1 + · · · + bn un
for some scalars b1 , b2 , · · · , bn .
=⇒ b1 u1 + · · · + bn un + (−an+1 )un+1 + · · · + (−ap )up = 0U
b1 = · · · = bn = an+1 = · · · = ap = 0
as B1 = {u1 , · · · , un , un+1 , · · · , up } is LI
So an+1 T (un+1 ) + · · · + ap T (up ) = 0V =⇒ an+1 = · · · = ap = 0
A = {T (un+1 ), · · · , T (up )} is LI
Thus A = {T (un+1 ), · · · , T (up )} is basis for R(T)
Therefore dim(R(T))=Number of elements in its basis set A
=p−n
= dim(U) − dim(ker (T ))
i.e; dim(ker(T))+dim(R(T))=dim(U)
i.e; n(T) + r(T) = dim(U)

Ashok Bingi SMAT0302 15 and 20 June 2022 4/6

Result Linear map T:U−→ V is one-one ⇐⇒ Ker(T)={0U }
Result For Linear map T:U−→ V,
U=[u1 , u2 , · · · , un ] =⇒ R(T)=[T (u1 ), T (u2 ), · · · , T (un )]
Proof: Obviously, [T (u1 ), T (u2 ), · · · , T (un )] ⊆ R(T )
Conversely, let v ∈ R(T ). Then v= T(u) for some
u∈ U=[u1 , u2 , · · · , un ].
So u=a1 u1 + a2 u2 + · · · + an un for some scalars a1 , a2 , · · · , an .
Hence v = T (u) = T (a1 u1 + a2 u2 + · · · + an un )
v = a1 T (u1 ) + a2 T (u2 ) + · · · + an T (un ) as T is linear
v ∈ [T (u1 ), T (u2 ), · · · , T (un )]
Thus v ∈ R(T )] =⇒ v ∈ [T (u1 ), T (u2 ), · · · , T (un )]
So R(T ) ⊆ [T (u1 ), T (u2 ), · · · , T (un )]
Therefore R(T)=[T (u1 ), T (u2 ), · · · , T (un )]

Ashok Bingi SMAT0302 15 and 20 June 2022 5/6

Result For Linear map T:U−→ V,
T is 1-1 and u1 , · · · , un are LI in U =⇒ T (u1 ), T (u2 ), · · · , T (un ) are LI
Proof: a1 T (u1 ) + · · · + an T (un ) = 0V for some scalars a1 , a2 , · · · , an
T (a1 u1 + · · · + an un ) = T (0U ) as T is linear and T (0U ) = 0V
a1 u1 + · · · + an un = 0U since T is 1-1
a1 = · · · = an = 0 as u1 , · · · , un are LI
Therefore T (u1 ), T (u2 ), · · · , T (un ) are LI
Result For Linear map T:U−→ V, {v1 , · · · , vn }are LI in R(T) and
u1 , · · · , un are vectors in U such that T (u1 ) = v1 , · · · , T (un ) = vn
=⇒ u1 , u2 , · · · , un are LI
Proof: a1 u1 + · · · + an un = 0U for some scalars a1 , a2 , · · · , an
T (a1 u1 + · · · + an un ) = T (0U )
a1 T (u1 ) + · · · + an T (un ) = 0V as T is linear and T (0U ) = 0V
a1 v1 + · · · + an vn = 0V
a1 = · · · = an = 0 as v1 , · · · , vn are LI
Therefore u1 , u2 , · · · , un are LI

Ashok Bingi SMAT0302 15 and 20 June 2022 6/6

 Algebra-III

Ashok Bingi

Department of Mathematics
St. Xavier’s College-Autonomous, Mumbai

21 June 2022

Ashok Bingi SMAT0302 21 June 2022 1/8

Recall one-one and onto function

the function f:A−→B is onto if and only if

∀b ∈ B(codomian), there exists a ∈ A(domain) such that b=f(a)
A function is onto iff its range=codomain

the function f:A−→B is one-one if and only if

f (a1 ) = f (a2 ) =⇒ a1 = a2

f −1 : B −→A is exists if and only if f is one-one and onto

in such a case, f(a)=b ⇐⇒ a=f −1 (b)
Let f and g be 2 functions with same domain A.Then
f=g iff f(x) =g(x) ∀x ∈ A

Ashok Bingi SMAT0302 21 June 2022 2/8

Linear Transformations
Non-singular/Isomorphism: Let U and V be vector spaces over R.
A linear map T:U−→ V is said to be Non-singular/Isomorphism if
T is one-one and T is onto.
In such case, we write U ∼
= V and say U and V are isomorphic.
Here, the map T −1 : V −→ U exists and say T is invertible
Note: T(x)=y ⇐⇒ x=T −1 (y )
Result: If T:U−→ V is a non-singular linear map then the map
T −1 : V −→ U is a linear, one-one and onto map
Proof: Let v1 , v2 ∈ V . As T:U−→ V is one-one and onto,
there exists u1 , u2 ∈ U such that T (u1 ) = v1 and T (u2 ) = v2
As the map T −1 : V −→ U exists, we have T −1 (v1 ) = u1 , T −1 (v2 ) = u2
To show T −1 : V −→ U is linear: For any scalar a ∈ R, T being linear,
T (a u1 + u2 ) = a T (u1 ) + T (u2 ) = a v1 + v2
By definition of T −1 , we have
T −1 (a v1 + v2 ) = a u1 + u2 = a T −1 (v1 ) + T −1 (v2 )
Therefore, the map T −1 : V −→ U is linear.
Ashok Bingi SMAT0302 21 June 2022 3/8
proof is continued

To show T −1 : V −→ U is onto: Let u ∈ U

w.r.t. function T:U−→ V, we have, T (u) = v ∈ V and hence u = T −1 (v )
Thus ∀u ∈ U, there exists v ∈ V such that T −1 (v ) = u
T −1 : V −→ U is onto
To show T −1 : V −→ U is one-one: For v1 , v2 ∈ V
T −1 (v1 ) = u1 ∈ U,T −1 (v2 ) = u2 ∈ U. Hence T (u1 ) = v1 , T (u2 ) = v2
consider T −1 (v1 ) = T −1 (v2 )
=⇒ u1 = u2
=⇒ T (u1 ) = T (u2 )
=⇒ v1 = v2
T −1 : V −→ U is one-one
Thus, the map T −1 : V −→ U is a linear, one-one and onto map

Ashok Bingi SMAT0302 21 June 2022 4/8

one-one ⇐⇒ onto
if vector spaces are finite dimensional with same dimension
For finite-dimensional vector spaces U, V of same dimension, a linear map
T:U−→ V is one-one ⇐⇒ T:U−→ V is onto.
Proof: Given dim(U)=dim(V)=n(say)
Claim If dim(R(T))=dim(V)=n where R(T) is subspace of V then R(T)=V
Let the set B with n elements be basis for R(T). Then R(T)=[B].
Also then this set B (with n elements)is LI in V. Hence B forms basis for V
because in n-dim. vector space V, any set of n LI vectors forms basis for V
So V=[B] Thus R(T)=V. Hence the claim.
Consider T is 1-1
⇐⇒ ker(T)={0U }
⇐⇒ dim(ker(T))=0
⇐⇒ 0+dim(R(T))=dim(U) by rank-nullity theorem
⇐⇒ dim(R(T))=dim(V)=n where R(T) is subspace of V
⇐⇒ R(T)=V
⇐⇒ T is onto
Ashok Bingi SMAT0302 21 June 2022 5/8
This result fails for infinite dimensional vector spaces

U=set of all (infinite) sequences {x1 , x2 , · · · , xn , · · · } of real numbers.

For x={x1 , x2 , · · · , xn , · · · } and y={y1 , y2 , · · · , yn , · · · } in U,
Define vector addition as x+y={x1 + y1 , x2 + y2 , · · · , xn + yn , · · · }
and scalar multiplication as αx={αx1 , αx2 , · · · , αxn , · · · } where α ∈ R
Also x=y ⇐⇒ x1 = y1 , · · · , xn = yn , · · ·
Then U is infinite dimensional vector space over R with standard basis
B={e1 = {1, 0, 0, · · · , 0, · · · }, e2 = {0, 1, 0, · · · , 0, · · · }, · · · ,
en = {0, 0, 0, · · · , 0, 1, 0, · · · }, · · · }
Define a function T:U−→U as T({x1 , x2 , · · · , xn , · · · })={x2 , · · · , xn , · · · }
Then T is linear. (H.W.) Show that T(αx + y ) = αT (x) + T (y )
clearly, given {x1 , x2 , · · · , xn , · · · } ∈ U(codomain), there exists
{z, x1 , x2 , · · · , xn , · · · } ∈ U(domain) such that
T ({z, x1 , x2 , · · · , xn , · · · }) = {x1 , x2 , · · · , xn , · · · }
Therefore T is onto

Ashok Bingi SMAT0302 21 June 2022 6/8

example continued
Now, ker(T)={x ∈ U(domain) | T (x) = 0U }
ker(T)={{x1 , x2 , · · · , xn , · · · } ∈ U | T ({x1 , x2 , · · · , xn , · · · }) = 0U }
ker(T)={{x1 , x2 , · · · , xn , · · · } ∈ U | {x2 , · · · , xn , · · · } = {0, · · · , 0, · · · }}
ker(T)={{x1 , x2 , · · · , xn , · · · } ∈ U(domain) | x2 = 0, · · · , xn = 0 · · · }
ker(T)={{x1 , 0, · · · , 0, · · · }}
ker(T)̸= {0U }
Hence T is not one-one (though T is onto)
This is because vector spaces under consideration are not finite
dimensional though dim(domain space)=dim(codomain space)
Def: Coordinate vector of v ∈ V w.r.t. ordered basis B={v1 , · · · , vn }:-
Clearly, v=a1 v1 + · · · + an vn for some scalars a1 , · · · , an in R
Then the n-tuple (a1 , · · · , an ) ∈ Rn is called coordinate vector of v∈ V
relative to ordered basis B.
If basis B changes then coordinate vector of same vector v also changes.
Note: Vector space over R is called real vector space
Vector space over C is called complex vector space
Ashok Bingi SMAT0302 21 June 2022 7/8
(dim n)V ∼
= Rn
Every real vector space V of dimension n is isomorphic to Rn
Proof: Let B={v1 , · · · , vn } be (ordered) basis for V.
For v ∈ V ,we have, v=a1 v1 + · · · + an vn for some scalars a1 , · · · , an in R.
Then the n-tuple (a1 , · · · , an ) ∈ Rn is coordinate vector of v∈ V relative
to ordered basis B.
Define a function T:V−→ Rn as
T(v)=(a1 , · · · , an ) ∀a1 v1 + · · · + an vn = v ∈ V
Then T is linear. (H.W.) Show that T(αu + v ) = αT (u) + T (v )
Now, ker(T)={v ∈ V | T (v ) = 0(codomain)}
ker(T)={v = a1 v1 + · · · + an vn ∈ V | (a1 , · · · , an ) = (0, · · · , 0)}
ker(T)={v = a1 v1 + · · · + an vn ∈ V | a1 = 0, · · · , an = 0}
ker(T)={0V }
Hence T is one-one
As dim(V)=dim(Rn ), T is one-one =⇒ T is onto
therefore, T:V−→ Rn is linear, 1-1, onto
T is an isomorphism and thus V ∼ = Rn
Ashok Bingi SMAT0302 21 June 2022 8/8
 Algebra-III

Ashok Bingi

Department of Mathematics
St. Xavier’s College-Autonomous, Mumbai

22 June 2022

Ashok Bingi SMAT0302 22 June 2022 1/6

Consequences of rank nullity theorem

Let T : U −→ V be a linear map and dim(U)=dim(V)=p. Then the

following statements are equivalent:
(a) T is non-singular(an isomorphism)
(b) T is 1-1
(c) T transforms LI subsets of U into LI subsets of V
(d) T transforms every basis for U into a basis for V
(e) T is onto
(f) r(T)=p
(g) n(T)=0
(h) T−1 exists

Ashok Bingi SMAT0302 22 June 2022 2/6

Proof
To show (a)=⇒(b): Let T be non-singular.
Then by its definition, T is 1-1 and we are done.
To show (b)=⇒(c): Let T be 1-1
Consider a LI set {u1 , · · · , un } in U. Then n ≤ p
Claim: {T (u1 ), T (u2 ), · · · , T (un )} are LI set in V
Let a1 T (u1 ) + · · · + an T (un ) = 0V for some scalars a1 , a2 , · · · , an
T (a1 u1 + · · · + an un ) = T (0U ) as T is linear and T (0U ) = 0V
a1 u1 + · · · + an un = 0U since T is 1-1
a1 = · · · = an = 0 as u1 , · · · , un are LI
Therefore T (u1 ), T (u2 ), · · · , T (un ) are LI
To show (c)=⇒(d): Assume (c) holds. Consider basis {u1 , · · · , up } for U.
Claim: {T (u1 ), T (u2 ), · · · , T (up )} forms basis for V
As {u1 , · · · , up } is LI in U, by (c), we have, {T (u1 ), T (u2 ), · · · , T (up )} is
LI in V. We know, in p-dim. vector space V, any set of p LI vectors forms
basis for V.
So {T (u1 ), T (u2 ), · · · , T (up )} forms basis for V
Ashok Bingi SMAT0302 22 June 2022 3/6
Proof(continued)

To show (d)=⇒(e): Assume (d) holds. Consider basis {u1 , · · · , up } for U.

Then by (d), {T (u1 ), T (u2 ), · · · , T (up )} forms basis for V
So V=[T (u1 ), T (u2 ), · · · , T (up )]. Also, clearly U=[u1 , · · · , up ]
Claim: R(T)=[T (u1 ), T (u2 ), · · · , T (up )]
Obviously, [T (u1 ), T (u2 ), · · · , T (up )] ⊆ R(T )
Now, let v ∈ R(T ). Then v= T(u) for some u∈ U =[u1 , u2 , · · · , up ].
So u=a1 u1 + a2 u2 + · · · + ap up for some scalars a1 , a2 , · · · , ap .
Hence v = T (u) = T (a1 u1 + a2 u2 + · · · + ap up )
v = a1 T (u1 ) + a2 T (u2 ) + · · · + ap T (up ) as T is linear
v ∈ [T (u1 ), T (u2 ), · · · , T (up )]
Thus v ∈ R(T )] =⇒ v ∈ [T (u1 ), T (u2 ), · · · , T (up )]
So R(T ) ⊆ [T (u1 ), T (u2 ), · · · , T (up )]
Therefore R(T)=[T (u1 ), T (u2 ), · · · , T (up )].
Hence, R(T)=V. So T is onto.

Ashok Bingi SMAT0302 22 June 2022 4/6

Proof(continued)

To show (e)=⇒(f): Assume (e) holds. i.e. T is onto.

R(T)=V. So dim(R(T))=dim(V). Hence r(T)=p
To show (f)=⇒(g): Assume (f) holds. i.e, r(T)=p
Then rank-nullity theorem, r(T)+n(T)=dim(U) gives n(T)=0
To show (g)=⇒(h): Assume (g) holds. i.e, n(T)=0
dim(ker(T))=0. So T is 1-1.
=⇒ T is onto since dim(U)=dim(V)
Thus, T−1 exists
To show (h)=⇒(a): Assume (h) holds. i.e.T−1 exists
Then T is invertible
. i.e. T is linear, 1-1, onto
T is non-singular(isomorphism)
Thus we proved
(a) =⇒ (b) =⇒ (c) =⇒ (d) =⇒ (e) =⇒ (f ) =⇒ (g ) =⇒ (h) =⇒ (a)
Ashok Bingi SMAT0302 22 June 2022 5/6
Composition of linear maps

If T : U −→ V and S : V −→ W are linear maps then their composition

S ◦ T : U −→ W is also linear map where (S ◦ T )(u) = S(T (u)) ∀u ∈ U
For convinience, we write S ◦ T as ST (product instead of composition)
Identity linear map f : U −→ U defined as f(u)=u ∀u ∈ U is denoted as IU
For the linear map T : U −→ U, the composition TT is denoted as T 2 .
Also T k = T k−1 T for k ∈ N and T 0 = IU
Clearly, IV T = T and T IU = T {f = g ⇐⇒ f (x) = g (x) for all x}
TT−1 = IV and T−1 T = IU
ST is 1-1 =⇒ T is 1-1 because
u ∈ ker (T ) =⇒ T (u) = 0V =⇒ (ST )(u) = 0W = (ST )(0U ) =⇒ u = 0U
ST is onto =⇒ S is onto because
for w ∈ W , there exists u ∈ U s.t. (ST)(u)=w as ST is onto
So, if v = T (u) ∈ V then w=(ST)(u)=S(T(u))=S(v). Hence S is onto

Ashok Bingi SMAT0302 22 June 2022 6/6

 Algebra-III

Ashok Bingi

Department of Mathematics
St. Xavier’s College-Autonomous, Mumbai

27 and 28 June 2022

Ashok Bingi SMAT0302 27 and 28 June 2022 1/5

Characterisation of non-singular map

A linear map T : U −→ V is non-singular iff there exists a linear map

S : V −→ U such that TS=IV and ST=IU . Here S=T−1 and T=S−1
Proof: A linear map T : U −→ V is non-singular then the linear map
T −1 : V −→ U exists such that TT−1 = IV and T−1 T = IU .
Renaming S=T−1 , we are done
Conversely, let linear maps S and T exist such that TS=IV and ST=IU .
Let u ∈ ker (T ). Then T(u)=0. So u = IU (u) = (ST )(u) = S(T (u)) = 0.
Thus ker(T)={0} and hence T is 1-1
Now let v ∈ V . If S(v)=u then v = IV (v ) = (TS)(v ) = T (S(v )) = T (u)
Thus T is onto
Therefore, T is non-singular. So T−1 exists
−1
From TT = IV , we have
T−1 = T −1 IV = T −1 (TS) = (T −1 T )S = IU S = S
similarly, one can show T=S−1

Ashok Bingi SMAT0302 27 and 28 June 2022 2/5

The space L(U,V)

If T : U −→ V and S : U −→ V are linear maps then their sum

S + T : U −→ V defined as (S+T)(u)=S(u)+T(u) is also a linear map
If S : U −→ V is linear map then scalar multiplication aS : U −→ V
defined as (aS)(u)=a(S(u)) ∀u ∈ U is also a linear map where a∈ R
L(U,V)=set of all linear maps from U to V={T : U −→ V | T is linear }
Result: The set L(U,V) forms a vector space with respect to above vector
addition and scalar multiplication forms a vector space over R
Here zero vector is the zero linear map O : U −→ V defined as
O(u)=0V ∀u ∈ U
a linear map T : U −→ U is called linear operator
a linear map T : U −→ R is called linear functional
If U and V are finite dimensional vector spaces then
dim(L(U,V))=(dim(U))(dim(V))
Ashok Bingi SMAT0302 27 and 28 June 2022 3/5
dual space L(V,R)

The vector space L(V,R) is called dual space and denoted as V ∗

V ∗ =L(V,R)=set of all linear maps from V to R
= {f ∗ : V −→ R | f is linear }
If V is finite dimensional (say, dim(V)=m) then L(V,R) is also finite
dimensional given by
dim(V ∗ )=dim(L(V,R)=(dim(V))(dim(R))=dim(V).1=dim(V)=m
So if B={v1 , · · · , vm } is basis for V then B ∗ = {f1∗ , · · · , fm∗ } is basis for V ∗
B ∗ is called dual basis of basis B for V
For v ∈ V ,we have, v=a1 v1 + · · · + am vm for some scalars a1 , · · · , am in R.
For fi ∗ : V −→ R, Define fi ∗ (v ) = ai ∀a1 v1 + · · · + am vm = v ∈ V
∗
From basis point of view, define fi (vk ) =0 or 1 depending on i ̸= k or
i = k respectively where k=1,2,...,m

Ashok Bingi SMAT0302 27 and 28 June 2022 4/5

Result
For any v̸= 0 in a finite dimensional vector space V, there exists functional
f in V ∗ such that f(v)̸= 0 ∈ R
Proof: Let dim(V)=m. Obviously, {v } is LI set in V.
By extension theorem, this LI set can be extended to form basis for V.
So there exists a basis B={v1 , · · · , vm } for V with v1 = v ̸= 0
Its corresponding dual basis (for V ∗ ) is B ∗ = {f1∗ , · · · , fm∗ }
From basis point of view, define fi ∗ (vk ) =0 or 1 depending on i ̸= k or
i = k respectively where k=1,2,...,m.
consider f1∗ (v ) = f1∗ (v1 ) = 1 ̸= 0
Thus there exists f1∗ = f (say) in V ∗ such that f1∗ (v ) ̸= 0
Corollary: In a finite dimensional vector space V, if v1 ̸= v2 then there
exists functional f in V ∗ such that f (v1 ) ̸= f (v2 ).
Proof: Given v = v1 − v2 ̸= 0 in V
Then by above result there exists functional f in V ∗ such that f (v ) ̸= 0
i.e;f (v1 − v2 ) ̸= 0
As f is linear, this gives f (v1 ) − f (v2 ) ̸= 0 and hence f (v1 ) ̸= f (v2 ).
Ashok Bingi SMAT0302 27 and 28 June 2022 5/5
 Algebra-III

Ashok Bingi

Department of Mathematics
St. Xavier’s College-Autonomous, Mumbai

30 June 2022

Ashok Bingi SMAT0302 30 June 2022 1/3

some more results
Let f be non-zero linear functional on V with f (v1 ) ̸= 0 for some v1 ∈ V .
Then for each v2 ∈ V , there exists a unique scalar c and a unique vector
v ∈ ker (f ) such that v2 = c v1 + v
Proof: Denote f (v1 ) = y ∈ R. Then y ̸= 0. So y −1 y = 1 where y −1 ∈ R
Consider v2 ∈ V and denote f (v2 ) = x ∈ R.
Then f (v2 ) = x 1 = x (y −1 y ) = (xy −1 ) y = c y where c = xy −1 ∈ R
f (v2 ) = c f (v1 ) = f (c v1 ) as f is linear
f (v2 ) − f (c v1 ) = 0 and hence f (v2 − (c v1 )) = 0. as f is linear
Thus (v2 − (c v1 )) ∈ ker (f ). Denote v2 − (c v1 ) = v .
Then v ∈ ker (f ) such that v2 = c v1 + v and we are done.
Uniqueness: Assume v2 = c ′ v1 + v ′ for some c ′ ∈ R , v ′ ∈ ker (f )
Then c v1 + v = c ′ v1 + v ′ =⇒ (c − c ′ )v1 + (v − v ′ ) = 0
=⇒ ′ ′
f ((c − c )v1 + (v − v )) = f (0)
=⇒ (c − c ′ ) f (v1 ) + f (v − v ′ ) = 0 ∈ R as f is linear
=⇒ ′
(c − c ) f (v1 ) = 0 as f (v − v ′ ) = 0
=⇒ ′
(c − c ) = 0 as f (v1 ) ̸= 0
=⇒ c =c ′
Ashok Bingi ′ ′ ′
SMAT0302 30 June 2022 2/3
Recall: (1) T is zero linear map ⇐⇒ aT is zero linear map where a ∈ R
(2) T is zero linear map ⇐⇒ Kernel of linear map T= domain of T
Result:If f , g ∈ V ∗ such that f(v)=0 =⇒ g(v)=0 ∀v ∈ V then g=af (a∈ R)
Proof: Clearly ker (f ) ⊆ ker (g ) because
w ∈ ker (f ) = {v ∈ V | f (v ) = 0} =⇒ w ∈ {v ∈ V | g (v ) = 0} = ker (g )
case-I f be zero linear functional on V
Then ker(f)= V and af is also zero linear functional on V where a ∈ R
Clearly, ker(g)⊆V= ker(f).
Thus ker(g)=ker(f)= V and hence g is zero linear functional on V
Therefore g(v)=(af)(v) ∀v ∈ V and hence g=af
case-II f be non-zero linear functional on V
Then there exists v1 ∈ V such that f (v1 ) ̸= 0
g (v1 )
Define a = . Let v2 ∈ V .
f (v1 )
Then there exists a unique scalar c and a unique vector v ∈ ker (f ) such
that v2 = c v1 + v . So f (v ) = 0 and hence g(v)=0
Clearly, g (v2 ) = g (c v1 + v ) = c g (v1 ) + g (v ) = c g (v1 )
(af )(v2 ) = a {f (c v1 + v )} = a {c f (v1 ) + f (v )} = c {a f (v1 )} = c g (v1 )
Therefore g (v2 ) = (af )(v2 ) ∀v2 ∈ V and hence g=af
Ashok Bingi SMAT0302 30 June 2022 3/3
 Algebra-III

Ashok Bingi

Department of Mathematics
St. Xavier’s College-Autonomous, Mumbai

4 July 2022

Ashok Bingi SMAT0302 4 July 2022 1/4

Annihilator: For subspace W of vector space V, Annihilator of W=A(W)
={f ∈ V ∗ | f (w ) = 0 ∀w ∈ W }
Note: (1) W is zero subspace of V =⇒ A(W)= V ∗
(2) W=V =⇒ A(W)= zero subspace of V ∗
(3) For subspaces U and W of V, U ⊆ W =⇒ A(W ) ⊆ A(U)
Result: A(W) is subspace of V ∗ where W is subspace of V
Proof: For zero linear functional O : V −→ R, we have O(v ) = 0 ∀v ∈ V
and hence O(v ) = 0 ∀v ∈ W as W ⊆ V .
So O ∈ A(W ). Thus A(W)̸= ∅
Let f , g ∈ A(W ) and a ∈ R
Claim: af + g ∈ A(W )
∀w ∈ W , consider (af+g)(w)
=(af)(w)+g(w) vector addition in V ∗
=a f(w)+g(w) scalar multiplication in V ∗
=a (0)+(0) as f , g ∈ A(W )
=0
Thus (af+g)(w)=0 ∀w ∈ W
therefore af + g ∈ A(W ) ∀f , g ∈ A(W ) and ∀a ∈ R
A(W) is subspace of V ∗
Ashok Bingi SMAT0302 4 July 2022 2/4
If W is subspace of finite dimensional vector space V then
dim(A(W )) = dim(V ) − dim(W )
Proof:Obviously, W is itself finite dimensional vector space(subspace of V).
Let dim(W)=n and dim(V)= p. Then p ≥ n
Let {u1 , · · · , un } be basis for W. Clearly, this basis set is LI in V
By extension theorem, this LI set can be extended to form basis for V
Let B = {u1 , · · · , un , un+1 , · · · , up } be basis for V.
Then B ∗ = {f1∗ , · · · , fn∗ , fn+1
∗ , · · · , f ∗ } is dual basis (of B) for V ∗ .
p
Here dim(V)=dim(V )=p and fi ∗ (uj ) = δij = 0 or 1 if i̸=j or i=j resp.
∗

So ∀ j = 1, · · · , n and ∀ i = n + 1, · · · , p, we have, fi ∗ (uj ) = 0

Consider the set A = {fn+1 ∗ , · · · , f ∗ }. We show A is basis for A(W)
p
Claim-1: A spans A(W) i.e;[A]=A(W)
Let f ∗ ∈ A(W ). Then as A(W ) ⊆ V ∗ we have f ∗ ∈ V ∗ whose basis is B ∗
f ∗ = a1 f1∗ + · · · + an fn∗ + an+1 fn+1∗ + · · · + a f ∗ for some scalars a , · · · , a
p p 1 p
∗
Now f ∈ A(W ) =⇒ f (w ) = 0 ∗ ∀ w ∈W
=⇒ f ∗ (uj ) = 0 ∀ j = 1, · · · , n
=⇒ (a1 f1∗ + · · · + an fn∗ + an+1 fn+1 ∗ + · · · + a f ∗ ) (u ) = 0 ∀ j = 1, · · · , n
p p j
=⇒ a1 (f1∗ (uj )) + · · · + an (fn∗ (uj )) + an+1 (fn+1 ∗ (u )) + · · · + a (f ∗ (u )) = 0
j p p j
∀ j = 1, · · · , n by vector addition and scalar multi. in V ∗
Ashok Bingi SMAT0302 4 July 2022 3/4
 =⇒ a1 = 0, · · · , an = 0 as fi ∗ (uj )= 0 or 1 if i̸=j or i=j resp.
∗ ∗ ∗
So f = a1 f1 + · · · + an fn + an+1 fn+1 ∗ + ··· + a f ∗ gives
p p
∗ ∗ ∗
f = an+1 fn+1 + · · · + ap fp = L.C. of elements of A
Thus f ∗ ∈ [A] and therefore A(W)⊆[A]
Now to show [A]⊆A(W), consider g ∗ ∈ [A]
Then g ∗ = bn+1 fn+1
∗ + ··· + b f ∗
p p for some scalars bn+1 , · · · , bp
Consider w ∈ W . Then w = c1 u1 + · · · + cn un for some scalars c1 , · · · , cn
So now g ∗ (w ) = g ∗ (c1 u1 + · · · + cn un )
=c1 g ∗ (u1 ) + · · · + cn g ∗ (un ) as g ∗ is linear
=c1 (bn+1 fn+1 + · · · + bp fp )(u1 ) + · · · + cn (bn+1 fn+1 + · · · + bp fp∗ )(un )
∗ ∗ ∗
∗ (u )) + · · · + b (f ∗ (u ))} + · · · +
=c1 {bn+1 (fn+1 1 p p 1
cn {bn+1 (fn+1 (un )) + · · · + bp (fp∗ (un ))} by vector addition, scalar multi.V ∗
∗

=c1 {0} + · · · + cn {0} as fi ∗ (uj )= 0 or 1 if i̸=j or i=j resp.

∗
Thus g (w ) = 0 ∀w ∈ W
So g ∗ ∈ A(W ) and therefore [A]⊆A(W)
Hence [A]=A(W) which implies the set A spans A(W)
We know, subset of LI set is LI. So as A⊆ B ∗ (LI being basis), we have,
A is LI. Thus the set A = {fn+1 ∗ , · · · , f ∗ } is basis for A(W)
p
dim(A(W))=no. of elements in its basis set A =p − n=dim(V ) − dim(W )
Ashok Bingi SMAT0302 4 July 2022 4/4
 Algebra-III

Ashok Bingi

Department of Mathematics
St. Xavier’s College-Autonomous, Mumbai

5 July 2022

Ashok Bingi SMAT0302 5 July 2022 1/6

Recall

Coordinate vector of v ∈ V w.r.t. ordered basis B={v1 , · · · , vp }:-

Clearly, v=a1 v1 + · · · + ap vp for some scalars a1 , · · · , ap in R
Then the p-tuple (a1 , · · · , ap ) ∈ Rp is called coordinate vector of v∈ V
relative to ordered basis B.  
a1
 .. 
Henceforth, we represent this coordinate vector by column vector  . 
ap
Matrices: (aij )m×n + (bij )m×n = (aij + bij )m×n
α(aij )m×n = (αaij )m×n
Equality of 2 matrices ?

Ashok Bingi SMAT0302 5 July 2022 2/6

Matrix representation of a linear map
definition: Let U be vector space of dimension n with ordered basis
B1 = {u1 , · · · , un }. Let V be vector space of dimension m with ordered
basis B2 = {v1 , · · · , vm }
Consider a linear map T:U−→V defined as follows:
T (u1 ) = a11 v1 + a21 v2 + · · · + am1 vm
T (u2 ) = a12 v1 + a22 v2 + · · · + am2 vm
..
.
T (un ) = a1n v1 + a2n v2 + · · · + amn vm
i.e, T (uj ) = a1j v1 + a2j v2 + · · · + amj vm = m
P
i=1 aij vi ∀ j = 1, 2, · · · , n
Then the matrix of T relative to ordered bases B1 and B2 is obtained by
writing the coordinate vectors of T (u1 ), T (u2 ), · · · , T (un ) successively as
column
 vectors in the form
 of rectangular array as follows:
a11 a12 · · · a1n
 a21 a22 · · · a2n 
 =(aij )m×n = {m(T )}B1 ,B2
 
 ..
 . 
am1 am2 · · · amn
Ashok Bingi SMAT0302 5 July 2022 3/6
Example
Let T : R2 −→ R3 is a linear map defined as
T (x1 , x2 ) = (x1 + 2x2 , −x1 , 0).
Find matrix of T w.r.t.bases B1 = {u1 = (1, 3), u2 = (−2, 4)} of R2 and
B2 = {v1 = (1, 1, 1), v2 = (2, 2, 0), v3 = (3, 0, 0)} of R3
Ans: T (u1 ) = a11 v1 + a21 v2 + a31 v3
(7, −1, 0) = a11 (1, 1, 1) + a21 (2, 2, 0) + a31 (3, 0, 0)
a11 + 2a21 + 3a31 = 7
a11 + 2a21 = −1
a11 = 0
which on solving gives a21 = −1 2 and a31 = 38
Now, similarly T (u2 ) = a12 v1 + a22 v2 + a32 v3 gives
a12 = 0, a22 = 1, a32 = 43 (check)
Required
 matrix={m(T
  )}B1 ,B2 = (T(u1 ) T (u2 )) = (aij )3×2
a11 a12 0 0
= a21
 a22 = −1/2
  1 
a31 a32 8/3 4/3
Ashok Bingi SMAT0302 5 July 2022 4/6
Recall an element in the product of 2 matrices

Let A=(aij )m×n and B=(bjk )n×p .

Then product of A and B is the matrix C=(cik )m×p .
To get (i, k)th element cik of matrix C,  
b1k
b2k 
we take i th row of A i.e;(ai1 ai2 · · · ain ) and k th column of B i.e; . 
 
 .. 
bnk
multiply them elementwise and take the sum of all these products.
n
X
So cik = ai1 b1k + ai2 b2k + · · · + ain bnk = aij bjk
j=1

Ashok Bingi SMAT0302 5 July 2022 5/6

To prove {m(ST )}B = {m(S)}B {m(T )}B
Let S and T be linear operators on V of dimension m. i.e,S:V−→V and
T:V−→V. P Let its ordered basis be B = {v1 , · · · , vm }
If S(vj ) = Pm i=1 aij vi ∀ j = 1, 2, · · · , m then {m(S)}B =(aij )m×m
If T (vk ) = m b
j=1 jk j v ∀ k = 1, 2, · · · , m then {m(T )}B =(bjk )m×m
 
Pm
= m
P
Consider (ST )(vk ) = S(T (vk )) = S j=1 bjk vj j=1 bjk S(vj )
   
Pm Pm Pm Pm
(ST )(vk ) = j=1 bjk i=1 aij vi = i=1 j=1 (aij bjk ) vi
Pm Pm
(ST )(vk ) = i=1 cik vi where cik = j=1 (aij bjk )
{m(ST )}B = (cik )m×m
Now (i,P k)th element of {m(S)}B {m(T )}B
= m j=1 aij bjk −→ by definition of product of 2 matrices
=cik = (i, k)th element of {m(ST )}B
Hence {m(ST )}B = {m(S)}B {m(T )}B
H.W.: {m(S + T )}B = {m(S)}B + {m(T )}B and
{m(αT )}B = α {m(T )}B
Ashok Bingi SMAT0302 5 July 2022 6/6
 Algebra-III

Ashok Bingi

Department of Mathematics
St. Xavier’s College-Autonomous, Mumbai

6 July 2022

Ashok Bingi SMAT0302 6 July 2022 1/5

L(U,V)∼
= Mm×n (R) , hence dim{L(U, V )}=(dimU)(dimV)

Let U be vector space of dimension n with ordered basis

B1 = {u1 , · · · , un }. Let V be vector space of dimension m with ordered
basis B2 = {v1 , · · · , vm }
Consider P
a linear map T∈ L(U, V ) defined as follows:
T (uj ) = m i=1 aij vi ∀ j = 1, 2, · · · , n
Then the matrix of T relative to ordered bases B1 and B2 is denoted by
{m(T )}B1 ,B2 and defined as (aij )m×n
′ ′
Now, for T ∈ L(U, V ) defined as T (uj ) = m
P
i=1 bij vi ∀ j = 1, 2, · · · , n
′
, we have {m(T )}B1 ,B2 = (bij )m×n
Define a function f: L(U,V)−→ Mm×n (R) as Pm
f(T)=(aij )m×n if T (uj ) = i=1 aij vi
To show f is linear, 1-1, onto

Ashok Bingi SMAT0302 6 July 2022 2/5

 ′ ′ ′

P, mT ∈ L(U, V P
For T ) and α ∈ R, P (αT + T )(uj ) = α(T (uj )) + T (uj )
m m
=α( i=1 aij vi ) + i=1 bij vi = i=1 {(αaij ) + bij }vi
′
So by definition of f, we get, f (αT + T ) = ((αaij ) + bij )m×n
′
=(αaij )m×n + (bij )m×n = α(aij )m×n + (bij )m×n = αf (T ) + f (T )
′ ′ ′
f (αT + T ) = αf (T ) + f (T ) ∀T , T ∈ L(U, V ) and ∀α ∈ R
thus f is linear
′
Now Consider f (T ) = f (T )
=⇒ (aP ij )m×n = (bij )m×n
Pm=⇒ aij = bij ∀i, j
′
=⇒ ( m a v
i=1 ′ ij i ) = b v
i=1 ij i =⇒ T (uj ) = T (uj ) ∀j = 1, · · · , n
=⇒ T = T
Thus f is 1-1
Now consider P(aij )m×n ∈ Mm×n (R). Then there exists S ∈ L(U, V ) such
that S(uj ) = m i=1 aij vi ∀ j = 1, 2, · · · , n
So f (S) = (aij )m×n and thus f is onto
f: L(U,V)−→ Mm×n (R) is linear, 1-1, onto
f is an isomorphism and thus L(U,V)∼
= Mm×n (R)
Hence dim{L(U, V )}=dim{Mm×n (R)} = mn = nm = (dimU)(dimV )
Ashok Bingi SMAT0302 6 July 2022 3/5
linear map T : Rn −→ Rm as multiplication by a matrix
Treat the vectors in Rn and Rm as column vectors. B1 = {e1 , · · · , en } is
′ ′
standard basis for Rn . B2 = {e1 , · · · , em } is standard basis for Rm .
′ ′ ′
If T (e1 ) = a11 e1 + a21 e2 + · · · + am1 em
′ ′ ′
T (e2 ) = a12 e1 + a22 e2 + · · · + am2 em
..
.
′ ′ ′
T (en ) = a1n e1 + a2n e2 + · · · + amn em
 {m(T )}B1 ,B2 = (T (e
then 1 ) T (e2 ) · · · T (en ))) = (aij )m×n
a11 a12 · · · a1n
 a21 a22 · · · a2n 
= .  = Am×n (say)
 
 .. 
am1 am2 · · · amn
Note that w.r.t. standard basis,
 coordinate
 vector of a vector is itself.
x1
 x2 
for the (column) vector X= .  in Rn , X=x1 e1 + x2 e2 + · · · + xn en .
 
 .. 
xn n×1
Ashok Bingi SMAT0302 6 July 2022 4/5
continued

Then as T is linear,
T(X)=x1 T(e1 ) + 2 ) + ·
x2 T (e · · + xn T (en
). 
a11 a12 a1n
 a21   a22   a2n 
T(X)=x1  . +x2  . +· · · +xn  . 
     
 . . .
 .  .
 . 
am1 am2 amn
     
a11 x1 + a12 x2 + · · · + a1n xn a11 a12 ··· a1n x1
 a21 x1 + a22 x2 + · · · + a2n xn
  a21 a22 ··· a2n   x2 
= = .
     
..  .. 
 ..
 
 .   .
am1 x1 + am2 x2 + · · · + amn xn am1 am2 · · · amn xn
T (X ) = Am×n X
T is called multiplication by matrix A and denoted as TA (referred as
matrix transformation)

Ashok Bingi SMAT0302 6 July 2022 5/5

 Algebra-III

Ashok Bingi

Department of Mathematics
St. Xavier’s College-Autonomous, Mumbai

7 July 2022

Ashok Bingi SMAT0302 7 July 2022 1/9

Nullity of matrix
For a fixed matrix Am×n , the function F : Rn −→ Rm defined as
F (X ) = A X is a linear map
′
Ans: For (column )vectors X , X in Rn , by algebra of matrices,
′ ′ ′ ′
F (αX + X ) = A (αX + X ) = αAX + AX = αF (X ) + F (X )
F is a linear map
Note: F = TA (matrix transformation/multiplication by a matrix)
Definition: For a matrix Am×n , Null space of A
={X ∈ Rn |AX = O}
=set of all solutions of homogeneous system of equations AX=O
=Nul(A)
HW Prove Nul(A) is subspace of Rn
Dimension of Null space of A
=nullity of A=n(A)
=number of free(non-leading variables) variables in the solution of AX=O
If TA : Rn −→ Rm is multiplication by a matrix Am×n then
ker (TA ) = Nul(A) and nullity(T)=nullity(A)
Ashok Bingi SMAT0302 7 July 2022 2/9
row space and column space of A

 
a11 a12 · · · a1n
Let A=(aij ) =  ...  be m × n matrix.
 

a a · · · amn
 m1  m2
R1
We write A= ...  where Ri = (ai1 , · · · , ain ) is the i th row
 

Rm
Treat Ri as a row vector in Rn
The vector subspace spanned by Ri′ s is called Row space of A=Row(A)
The non-zero rows of row echelon form of A form basis for row
space of A
(∗)Dimension of row space of A=row rank of A

Ashok Bingi SMAT0302 7 July 2022 3/9

column space of A

 
a11 a12 · · · a1n
 ..
Let A=(aij ) =  .  be m × n matrix.


am1 am2 · · · amn

 
a1j
Also, We write A=(C1 , · · · , Cn ) where Cj =  ...  is the j th Column
 

amj
Treat Cj as a column vector in Rm
The vector subspace spanned by Cj′ s is called Column space of A=Col(A)
The pivot columns (columns with leading ones) of A form basis for
column space of A
(∗)Dimension of column space of A=column rank of A

Ashok Bingi SMAT0302 7 July 2022 4/9

Find basis for Row(A), Col(A), Nul(A)
 
−2 −5 8 0 − 17
1 3 −5 1 5 
A= 
3 11 −19 7 1 
1 7 − 13 5 −3
Ans We convert A to its row-echelon form
Perform
 R12 
1 3 −5 1 5
−2 −5 8 0 − 17
A∼  Perform
3 11 − 19 7 1 
1 7 − 13 5 −3
2R1 , R3 − 3R1 , R4 − R1
R2 + 
1 3 −5 1 5
0 1 −2 2 −7
A∼ 
0 2 −4 4 − 14
0 4 −8 4 −8

Ashok Bingi SMAT0302 7 July 2022 5/9

continued

 R3 − 2R2 , R4 − 4R2
Perform 
1 3 −5 1 5
0 1 −2 2 −7 
A∼ 0

0 0 0 0
0 0 0 −4 20
Perform −1 R Then interchange 3rd row and 4th row
 4 4 
1 3 −5 1 5
0 1 −2 2 − 7
A∼  which is in row echelon form
0 0 0 1 − 5
0 0 0 0 0
Basis for row space of A={(1, 3, −5, 1, 5), (0, 1, −2, 2, −7), (0, 0, 0, 1, −5)}
     
 −2 −5 0 
 1   3  1
Basis for column space of A=   3  ,  11  , 7
    

1 7 5
Ashok Bingi SMAT0302 7 July 2022 6/9
continued

Null space of A4×5

={X ∈ R5 |AX = O}
Solving AX=O gives  
  x1  
1 3 −5 1 5 x2  0
0
 1 − 2 2 − 7
 
x3 =
0
 
0 0 0 1 − 5  
x4 
0
0 0 0 0 0 0
x5
Clearly x1 , x2 , x4 are leading variables
x3 , x5 are free (non-leading) variables
Working backwards and rewriting in terms of free (non-leading) variables,
we get, x4 = 5x5
x2 − 2x3 + 2x4 − 7x5 = 0 =⇒ x2 = 2x3 − 3x5
x1 + 3x2 − 5x3 + x4 + 5x5 = 0 =⇒ x1 = −x3 − x5

Ashok Bingi SMAT0302 7 July 2022 7/9

continued

Null space of A4×5 ={X ∈ R5 |AX = O}

       
x1 −x3 − x5 −1 −1
x2  2x3 − 3x5   2  −3
       
x3  =  x3
where X =     =  1  x3 +  0  x5
    
x4   5x5   0  5
x5 x5 0 1
   
−1 −1
  2  −3 
   
Basis for Null space of A=   1 , 0 
  
0 5
0 1

Ashok Bingi SMAT0302 7 July 2022 8/9

 Algebra-III

Ashok Bingi

Department of Mathematics
St. Xavier’s College-Autonomous, Mumbai

12 July 2022

Ashok Bingi SMAT0302 12 July 2022 1/3

Conclusions from this A4×5
Basis row space of A4×5 ={(1, 3, −5, 1, 5), (0, 1, −2, 2, −7), (0, 0, 0, 1, −5)}
   
      −1 −1
 −2 −5 0    2  −3 
 1   3  1    
Basis col.=   ,   ,   . Basis nul.=  1  ,  0 
 3   11  7    
0 5
1 7 5
0 1
So we have following conclusions:
1 row rank of A=dimension of row space of A= 3

2 column rank of A= dimension of column space of A= 3

3 row rank of A=column rank of A. This common value =rank of A

4 nullity of A=dimension of null space of A= 2

5 column space of A is subspace of R4

6 row space of A is subspace of R5

7 null space of A is subspace of R5

5 4
8 rank-nullity theorem for matrix A
4×5 of linear map T:R −→ R is
rank(A)+nullity(A)=Number of columns of A (3 +2 =5)
Ashok Bingi SMAT0302 12 July 2022 2/3
Find ker(TA ) and range(TA )
if linear map TA : Rn −→ Rm is multiplication by matrix Am×n
Recall Here TA (X ) = AX
ker(TA ) = {X ∈ Rn |TA (X ) = O} = {X ∈ Rn |AX = O} = Nul(A)
range(TA ) = {b ∈ Rm | there exists X ∈ Rn such that TA (X ) = b}
= {b ∈ Rm | there exists X ∈ Rn such that AX = b}
=Col(A)
 
a1j
 .. 
Note: Let Am×n =(C1 , · · · , Cn ) where Cj =  .  is the j th Column
a
 mj
x1
 .. 
In view of this, AX=b ⇐⇒ (C1 , · · · , Cn )  .  = b
xn
AX=b ⇐⇒ x1 C1 + · · · + xn Cn = b
System AX=b is consistent ⇐⇒ b is L.C. of columns of A ⇐⇒ b∈ Col(A)
Columns of A span Rm ⇐⇒ every vector b ∈ Rm is L.C. of columns of A
Ashok Bingi SMAT0302 12 July 2022 3/3