Module 3:

Properties of Subspaces

Row and column spaces:

Let A  aij  be an m x n matrix i.e.,
mn

Row Vectors:

Note: There are m Row vectors r1 , r2 , r3 , ..... rm  of A and each Row vector belongs to n
.

Column Vectors:
Dr. PARVEZ ALAM

Note: There are n Column vectors c1 , c2 , c3 , ..... cn  of A and each column vector belongs to m
.
 Row space: Let A  aij 
mn
be an m x n matrix, then the Row Space of A is the

subspace in n
is spanned by Row Vectors r1 , r2 , r3 , ..... rm  denoted by R(A).

Column space: Let A  aij 

mn
be an m x n matrix, then the Column Space of A is

the subspace in m
is spanned by Row Vectors c1 , c2 , c3 , ..... cn  denoted by C(A).

Null space: Let A  aij 

mn
be an m x n matrix, then the solution set of homogeneous

 x1 
x 
equation AX  0 is called Null Space of A for any column vector X    .
2

:
 
 xn  n1

It is denoted by N(A); N ( A)  X   x1 , x2 , x3 ,.... xn  
n
; such that AX  0  .
Remarks:
1. Dimension of any space is equal to the cardinality of its basis. (see module 2 note)
2. Dim[R(A)] is called Row rank of matrix A.
3. Dim[C(A)] is called Column rank of matrix A.
4. Dim[N(A)] is called Nullity of matrix A.
Nullity of A = Dim[N(A)] = No. of free variables in the solution of AX=0.
5. Since the Row vectors of A are just Column vectors of AT and Column vectors of A are
just Row vectors of AT. Then Row space of A is Column space of AT and Column space
of A is Row space of AT i.e.
R(A)=C(AT) and C(A)=R(AT)
6. Row rank of A is always less or equal to n, i.e. R(A)  n .
7. Column rank of A is always less or equal to m, i.e. C(A)  m .

Theorem 4.1: (Fundamental Theorem)

If A is an m x n matrix, then the row space and the column space of A have the same
dimension, i.e.,
Dr. PARVEZ ALAM

Dim[R(A)] = Dim[C(A)]

Rank of Matrix A: The dimension of the row (or column) space of a matrix A is called the
rank of A and is denoted by Rank (A). We also use symbol  for the Rank of A.

Dim[R(A)] = Dim[C(A)]= Rank (A) = 

Remark: 7. By result 4 and Theorem 4.1; Rank (AT) = Rank (A) = 

 Note: If we do elementary row operations in a given matrix A, then the new obtained matrix is
called equivalent (~) matrix of A.
Example:
Dr. PARVEZ ALAM

*Properties of equivalent matrices are same.

 Theorem 4.2: (Row equivalent matrices have the same row space)
If an (m x n) matrix A is row equivalent to an m x n matrix B, then the row space of A is equal
to the row space of B.

Theorem 4.3: (Basis for the row space of a matrix)

If a matrix A is row equivalent to a matrix B in Row-Echelon or Reduced Row-Echelon form,
then the nonzero row vectors of B form a basis for the row space of A.

Theorem 4.4: If A is an m x n matrix, then the set of all solutions (solution set) of the
homogeneous system of linear equations Ax=0 is a subspace of Rn called the null space of A.
 
Proof: By definition N ( A)  X   x1 , x2 , x3 ,.... xn   n ; such that AX  0

Clearly  N ( A)  n
; N ( A)    A0  0; where 0 is 0 vector in 
n
Dr. PARVEZ ALAM
  Finding Basis for the Row Space of a matrix A:
First convert the given matrix into Row-Echelon or Reduced Row-Echelon form.
Collect the non-zero rows; it gives basis for the Row Space.
Cardinality of the basis will be its dimension and called Row Rank.

Example 1:

Solution: First we find Reduced Row Echelon form of A.

Basis for R(A)= u1 , u2 , u3   1, 3, 1, 3 ,  0, 1, 1, 0  ,  0, 0, 0, 1

or 1 3 1 3 , 0 1 1 0, 0 0 0 1 .
Dim R(A)= Row Rank of A=3.

Observation: We can see that A is matrix of 5 x 4 order; R( A)  4

.

 3 6 1 1 7 

Example 2: Find a basis for the row space of matrix A  1 2 2 3 1 .

 
 2 4 5 8 4 
 3 6 1 1 7 

Solution: Given matrix A  1 2 2 3 1
 , we find Reduced Row Echelon form of A.
 
 2 4 5 8 4  35

 3 6 1 1 7  1 2 0 1 3   u1
A   1 2 2 3 1 Reduced Row Echelon Form B  0 0 1 2 2   u2
 
 2 4 5 8 4  0 0 0 0 0   u3
Basis for R(A)= u1 , u2   1, 2, 0, 1, 3 ,  0, 0, 1, 2, 2 
Dr. PARVEZ ALAM

or 1 2 0 1 3 , 0 0 1 2 2 .
Dim R(A)=Row Rank of A=2.

Observation: We can see that A is matrix of 3 x 5 order; R( A)  5

.
  Finding Basis for the Column Space of a matrix A:
First convert the given matrix into Reduced Row Echelon form.
Collect the corresponding column of ‘Leading 1’ column; it gives basis for the Column
Space.
Leading 1 means only single non-zero entry is ‘1’ in column and rest entries are zero.
Cardinality of the basis will be its dimension and called Column Rank.
 3 6 1 1 7 

Example 1: Find a basis for the row space of matrix A  1 2 2 3 1 .

 
 2 4 5 8 4 
 3 6 1 1 7 
Solution: Given matrix A   1 2 2 3 1 , we find Reduced Row Echelon form of A.
 
 2 4 5 8 4  35

 3 6 1 1 7  1 2 0 1 3 
A   1 2 2 3 1 Reduced Row Echelon Form B  0 0 1 2 2 
 2 4 5 8 4  0 0 0 0 0 
         
v1 v2 v3 v 4 v5 v1 v2 v3 v4 v5
  3   1 
  2 
 
Basis for C(A)= v1 , v3   3 1 2  ,  1 2 5  or   1  ,  
 2   5  
 
Dim C(A)=Column Rank of A=2.

Observation:
1. We can see that A is matrix of 3 x 5 order; C ( A)  3 .
2. Why we are collecting v1 and v3 only? As they have leading 1 only. But the
mechanism/reason behind is that rest columns are L.D. So not choosing them for Basis
element.
 2 1 
(1). v 2 is multiple of v1  v 2  2v1   0   2 0  ;
 
 0  0
so v 2 is L.D., hence not including in basis. (By Remark 6 in Module-2)
1  0  1
(2). v 4 is Linear Combination of v1 & v2   v1  2v2 = v 4   0  2 1    2  ;
 
Dr. PARVEZ ALAM

0 0  0 
so v 4 is L.D., hence not including in basis. (By a Theorem in Module-2)
1  0  3 
(3). v5 is Linear Combination of v1 & v 2  3v1  2v 2 = v5  3 0   2 1    2 ;
0 0  0 
so v5 is L.D., hence not including in basis. (By a Theorem in Module-2)
 Remark 6 and Theorem in Module 2:

Example 2:

Solution: First we find Reduced Row Echelon form of A.

  1  3   3  
      
  0  1   0  
 
 
Basis for C(A)= v1 , v 2 , v 4  1,0,  3, 3, 2  ,  3, 1, 0, 4, 0  ,  3, 0,  1, 1, 2  or   3 , 0  ,  1  ,
  3   4 1  
      
Dr. PARVEZ ALAM

  2  0   2  
Dim C(A)=Column Rank of A=3.

Observation:
1. We can see that A is matrix of 5 x 4 order; C ( A)  5 .
2. Why we are collecting v1 and v3 only? Because, v3 is L.D. by Theorem as -2v1+v2=v3.
 Alternate: Here we have an alternate method to find Basis for Column space:
First convert the given matrix A into Transpose
Then we find, Row Echelon or Reduced Row Echelon form of A T .
Collect the non-zero rows of A T ; it gives the basis for the Column Space.
Cardinality of the basis will be its dimension and called Column Rank.
Note: We follow same rule as we used for finding basis for Row spaces, but here for A T .
 3 6 1 1 7 

Example 1: Find a basis for the row space of matrix A  1 2 2 3 1

 2 4 5 8 4 
 3 1 2  1 1 / 3 2 / 3  w1
 6 2 4  0 1 13 / 5   w2
  
A   1 2 5  Row Echelon Form B  0
T
0 0  .
   
1 3 8 0 0 0 
 7 1 4  0 0 0 

Basis for R( A T )= v1 , v2   1, 1 / 3, 2 / 3 ,  0, 1, 13 / 5 

or 1, 1/ 3, 2 / 3 , 0, 1, 13 / 5 ,
 1   0 
  1  ,
So, Basis for C(A)= v1 , v2   1, 1 / 3, 2 / 3 ,  0, 1, 13 / 5  or   1 / 3  ,  
  2 / 3 13 / 5 
 
Dim C(A)=2.
Dr. PARVEZ ALAM

Note: In Exam you have to show steps of finding Row Echelon and Reduced Row Echelon
forms. In my note, directly written, because already you are learned finding the Row Echelon
and Reduced Row Echelon forms in module 1.
  Finding Null Space & Basis for the Null Space of a matrix A:
Let A is an m x n matrix.
First we solve AX  0 for any X   x1 , x2 , x3 ,.... xn  
n
.
The solution set of No-trivial solution gives Null Space.
We can collect vectors after taking free variables common from the vectors. These
vectors forms basis for the Null Space.
Cardinality of the basis will be its dimension and called Nullity.

Soln: To find the null space of matrix A, we need to find solution of

homogeneous linear system of eqn. AX=0 for X=

Here, n   =4-3=1, so assuming one free variable x3  c . Using the method of back
1 3
substitution, we obtain x1   c, x2  c, x3  c, x4  c .
2 2
Dr. PARVEZ ALAM

 1 3  
Then the solution Space=N(A)=   c, c, c, c  c   .
 2 2  
 1 3   1 3   1 3 
  c, c, c, c   c   , , 1, 1  , So, Basis of N(A)= BN(A) =   , , 1, 1 
 2 2   2 2   2 2 

Nullity=Dim[N(A)]=1.  n    1
Dr. PARVEZ ALAM
 1 2 2 1 0  1 2 2 1 0 
 x  2x  2x  x  0
 
 A | 0  3 6 5 4 0 Row Echelon Form 0 0 1 1 0  1 2 x3  x4  0
1 2 0 3 0  0 0 0 0 0  3 4

  2, n  4, n    4  2  2, two free variables. Let x4  t  x3  t & x2  s  x1  2s  3t

Then the solution Space=N(A)=  2s  3t , s,  t, t  s, t   .

 2s  3t , s,  t, t    2s , s, 0, 0    3t , 0,  t, t 
,
 s  2 , 1, 0, 0   t  3 , 0,  1, 1
So, Basis of N(A)= BN(A) =  2 , 1, 0, 0  ,  3 , 0,  1, 1

Nullity=Dim[N(A)]=2.  n    2
Dr. PARVEZ ALAM
Dr. PARVEZ ALAM
 Theorem: (Rank-Nullity Theorem)
Let A  aij  be an m x n matrix then
mn

1. Dim R(A) + Dim N(A) = n  Rank (A) + Nullity (A) = n

2. Dim C(A) + Dim N(AT) = m  Rank (A) + Nullity (AT) = m

Dim[R(A)] = Dim[C(A)]= Rank (A) = 

Note: If AX=0 has only trivial solution then solution Space=N(A)= 0
Then Dim N(A) = 0  Rank (A)=n. (see last page Q4)
It means all the Rows of matrix A are L.I. (non-zero in RE/RRE form).

Question: Find the Rank of the matrix A and using the Rank-nullity theorem
find the dim of Null space of A and AT
Dr. PARVEZ ALAM

Note: If it asking to find Rank of A, just find Dim R(A) or Dim C(A) (any one).
 Theorem: (Rank-Nullity Theorem)
Let A  aij  be an m x n matrix then
mn

1. Dim R(A) + Dim N(A) = n  Rank (A) + Nullity (A) = n

2. Dim C(A) + Dim N(AT) = m  Rank (A) + Nullity (AT) = m

Proof 1:
Case 1: If Rank (A) = n
Then, it means that A has exactly n non-zero rows in Row echelon form (i.e. exactly n L.I.
rows in matrix A, so Rank (A) =  = n.
It also means that the homogeneous system AX=0 has no free variables because n-  = n-n=0,
so AX=0 will have only trivial solution. So the Null space N(A)={0}, Nullity of A is 0.
So the statement is satisfying as
Rank (A) + 0 = n
 Rank (A) + Nullity (A) = n.

Case 2: If Rank (A) = r < n

Then, it means that A has exactly r non-zero rows in Row echelon form (i.e. exactly n L.I. rows
in matrix A, so Rank (A) =  = r.
It also means that the homogeneous system AX=0 has n-  = n-r free variables.
So, Nullity of A is n-  = n-r.
Nullity (A) = n-r
 r + Nullity (A) = n.
 Rank (A) + Nullity (A) = n.
(as assumed case 2 Rank (A) = r)

Proved.

In the same way, we can prove for second statement Rank(A) + Nullity(AT) = m, for matrix AT
Dr. PARVEZ ALAM
 Dimension theorem: Suppose U & W are subspaces of vector space V. If U+V is linear
sum, then dimension theorem states: dim U  W   dim U  dim W  dim U  W .
Dr. PARVEZ ALAM
Dr. PARVEZ ALAM
Dr. PARVEZ ALAM
Dr. PARVEZ ALAM