Properties of Subspaces

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 .

Remark: If U  V is direct sum, then U  W={0}  dim(U  W)=0, then dimension theorem
states:
dim U  W   dim U  dim W .

Example 1: Let U and W are the subspaces of R4 generated by {(1,1,0, 1), (1,2, 3,0), (2,3,3, 1)}
and {(1,2,2, 2), (2,3,2, 3), (1,3,4, 3)} respectively. Find the Basis & dimension of U, W.

Solution:

(1). Basis & Dim of U:

U is generated by set {(1,1,0, 1), (1,2, 3,0), (2,3,3, 1)} , but this set is L.D., to find basis, we
remove L.D. vector from this as

1 1 0  1 1 1 0  1
1 2 3 0  converting in Row Echelon form
 0 1 3 1 
  (You have to show proper setps in exam)  
 2 3 3  1 0 0 0 0 

Basis of subspace U is BU  {(1,1,0, 1), (0,1,3,1)}  Dim(W)  2

(2). Basis & Dim of W:

W is generated by set {(1,2,2, 2), (2,3,2, 3), (1,3,4, 3)} , but this set is L.D., to find basis, we
remove L.D. vector from this as

1 2 2  2  1 2 2  2
 2 3 2  3 converting in Row Echelon form
 0  1  2 1 
  (You have to show proper setps in exam)  
1 3 4  3  0 0 0 0 

Basis of subspace W is BW  {(1,2,2, 2), (0, 1, 2,1)}  Dim(W)  2.

Dr. PARVEZ ALAM
 Finding basis of linear sum of two subspaces:
Tow find basis of linear sum set U+W subspace of V, we first add basises of U and W, It gives
spanning set of U+W, find row echelon form of matrix arranged by spanned vectors.
Corresponding non-zero row vectors will give basis of U+W.

Example 2: Let U and W are the subspaces of R4 generated by {(1,1,0, 1), (1,2, 3,0), (2,3,3, 1)}
and {(1,2,2, 2), (2,3,2, 3), (1,3,4, 3)} respectively. Find the dimension of U, W, U+W and
U  W.

Solution: Basis & Dim for U & W are computed in Example 1. Dim(U)  2 & Dim(W)  2

BU  {(1,1,0, 1), (0,1,3,1)} is basis of U, so definitely it spans U and also L.I.

BW  {(0, 1, 2,1), (1, 2,2, 2)} is basis of W, so definitely it spans W and also L.I.

Therefore, combined set {(1,1,0, 1), (0,1,3,1),(0, 1, 2,1), (1,2,2, 2)} definitely span U+W, but to
be a basis, it should be L.I. also (It is L.D.)

Now we remove L.D. vector by the following process.

Collecting the corresponding non-zero vectors; {(1,1,0, 1), (0,1,3,1),(0, 1, 2,1)} forming a
basis for U+W, as it spanning U+W and L.I. also. Dim(U+W)=3.

Then by Dimension Theorem: dim U  W   dim U  dim W  dim U  W 

 dim U  W   dim U  dim W  dim U  W 
 dim U  W   2  2  3
 dim U  W   1
Dr. PARVEZ ALAM
 Finding basis of intersection of two subspaces:
See the example to understand;

Example 3: If basis of subspace U is BU  {(1,1,0, 1), (0,1,3,1)} and basis of subspace W is

BW  {(0, 1, 2,1), (1, 2,2, 2)} , then find the basis of U  W.

Solution: Let vU  W, so vU and vW.

It means v can be generated by the basis of U as vU  v  1 (1,1,0, 1)  2 (0,1,3,1) ......(1)
Also, v can be generated by the basis of W as vW  v  1 (0, 1, 2,1)  2 (1,2,2, 2) ......(2)

By, equation (1) and (2),

v  v  1 (1,1,0, 1)  2 (0,1,3,1)  1 (0, 1, 2,1)  2 (1,2,2, 2)
 1 (1,1,0, 1)   2 (0,1,3,1)  1 (0, 1, 2,1)  2 (1,2,2, 2)  (0,0,0,0)
 (1  2 , 1   2  1  22 ,3 2  21  22 , 1   2  1  22 )  (0,0,0,0)
1  2  0  1 0 0 1 0
 
1   2  1  22  0   1 1 1 2 0  Solving by
   1  1,  2  0, 1  1, 2  1
3 2  21  22  0  0 3 2 2 0  G.J.E.method

1   2  1  22  0   1 1 1 2 0

So, by putting values of 1  1, 2  0 in eqn (1) OR by putting values of 1  1, 2  1 in eqn (2),

we get solution (t,t,0,−t)= t(1,1, 0,  1)
So the basis of U  W is BU W  {(1,1, 0,  1)}
Dim(U  W)=1

Example 4: If basis of subspace U is BU  {(1, 1, 4), (1, 1,  2)} and basis of subspace W is
BW  {(1, 2,  4), (3,  1, 1)} , then find the basis of U  W.

Solution: Let vU  W, so vU and vW.

It means v can be generated by the basis of U
as vU  v  1 (1, 1, 4)  2 (1, 1,  2) ...........................(1)
Also, v can be generated by the basis of W as vW  v  1 (1, 2,  4)  2 (3,  1, 1) ......(2)

By, equation (1) and (2),

v  v  1 (1, 1, 4)  2 (1, 1,  2)  1 (1, 2,  4)  2 (3,  1, 1)
Dr. PARVEZ ALAM

 1 (1, 1, 4)   2 (1, 1,  2)  1 (1, 2,  4)  2 (3,  1, 1)  (0,0,0,0)

1   2  1  32  0


It gives a linear system of eqn  1   2  21  2  0 
4 2  2 2  41  2  0 
 1 3
So, by putting values of 1   c,  2  c in eqn (1) OR by putting values of 1  c, 2  c in
2 2
eqn (2), we get solution  2c, c,  5c  .
So the basis of U  W is BU W  { 2, 1,  5} by taking scalars out
Dim(U  W)=1

Very Important Note: Questions related to the Basis of vector subspaces are just same as
finding Questions related to the Basis of Vector Spaces, because vector subspaces subset of
any V.S. and also forming V.S. with same operations. See questions of Basis from Module 2.
Dr. PARVEZ ALAM
Dr. PARVEZ ALAM
 Invertibiliy:
For m x n order matrices

Remark:

1. A matrix A of order m x n has a right inverse if and only if Rank (A)=m.

2. A matrix A of order m x n has a left inverse if and only if Rank (A)=n.
3. It means if m  n , then matrix A cannot have both left and right inverses.
4. If Rank (A)=m, then its right inverse is given by A  AA 
T T 1
.

If Rank (A)=n, then its left inverse is given by  A A A .

T 1 T
5.
Dr. PARVEZ ALAM
Dr. PARVEZ ALAM
 For Square matrices

If m = n then matrix A become square matrix and it will have both right and left inverses if
and only if Rank (A) = m = n. And matrix is called invertible (Non-Singular).

Summary of equivalent conditions for square matrices:

For a square matrix A of order n, the following statements are equivalent.

We can check above results by following examples:

Note: Invertible means non-singular matrix.

Examples 1: Given matrix (of 4 x 4 order) is invertible
and Rank the matrix is n=4

Row Echelon form 

Dr. PARVEZ ALAM

So Rank A =Dim R(A) = 4.

Also, no. of non zero Rows are 4.
Non-zero Rows are basis of R(A).
Leading 1 columns are basis of C(A) (in RRE form of A)
 Examples 2: Given matrix (of 4 x 4 order) is not invertible and Rank the matrix is 3

Row Echelon form 

Here, Rank A =Dim R(A) = 3  4

Also, no. of non zero Rows are 3.
Non-zero Rows are basis of R(A).
Leading 1 columns are basis of C(A) (in RRE form of A)
Dr. PARVEZ ALAM