Linear Algebra - MA2034

5 Matrix form of Linear Transformations,

Column and Row Space of a Matrix

5.1 Matrix form of a Linear Transformation

In this section we will show that a general linear transformation from any 𝑛-
dimensional vector space 𝑉 to any 𝑚-dimensional vector space 𝑊 can be
performed using an appropriate matrix transformation from ℝ𝑛 to ℝ𝑚 . This idea
is used in computer computations since computers are well suited for performing
matrix computations.

Matrices of Linear Transformations

Suppose that 𝑉 is an 𝑛-dimensional vector space, that 𝑊 is an 𝑚-dimensional
vector space, and that 𝑇 ∶ 𝑉 → 𝑊 is a linear transformation. Suppose further that
𝐵 is a basis for 𝑉, that 𝐵′ is a basis for 𝑊, and that for each vector 𝑥 in 𝑉, the
coordinate vectors for 𝑥 and 𝑇(𝑥) are [𝑥]𝐵 and [𝑇(𝑥)]𝐵′ respectively.

It will be our goal to find an 𝑚 × 𝑛 matrix 𝐴 such that multiplication by 𝐴 maps the
vector [𝑥]𝐵 into the vector [𝑇(𝑥)]𝐵′ for each 𝑥 in 𝑉.

1
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

If we can do so, we will be able to execute the linear transformation 𝑇 by using

matrix multiplication and the following indirect procedure:
Step 1. Compute the coordinate vector [𝑥]𝐵 .

Step 2. Multiply [𝑥]𝐵 on the left by 𝐴 to produce [𝑇(𝑥)]𝐵′ .

Step 3. Reconstruct 𝑇(𝑥) from its coordinate vector [𝑇(𝑥)]𝐵′ .

The key to executing this plan is to find an 𝑚 × 𝑛 matrix 𝐴 with the property
𝐴[ 𝑥 ]𝐵 = [ 𝑇( 𝑥 )]𝐵′ ; ∀ 𝑥 ∈ 𝑉 (1)

For this purpose, let 𝐵 = {𝑢1 , 𝑢2 , . . . , 𝑢𝑛 } be a basis for the 𝑛-dimensional space 𝑉 and
𝐵′ = {𝑣1 , 𝑣2 , . . . , 𝑣𝑚 } be a basis for the 𝑚-dimensional space 𝑊. Since Equation (1)
must hold for all vectors in 𝑉, it must hold, in particular, for the basis vectors in 𝐵; that
is,

𝐴[𝑢1 ]𝐵 = [𝑇(𝑢1 )]𝐵′ , 𝐴[𝑢2 ]𝐵 = [𝑇(𝑢2 )]𝐵′ , … , 𝐴[𝑢𝑛 ]𝐵 = [𝑇(𝑢𝑛 )]𝐵′ (2)

But

2
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

1 0 0
0 1 0
[𝑢1 ]𝐵 = [ ], [𝑢2 ]𝐵 = [ ] , … , [𝑢𝑛 ]𝐵 = [ ]
⋮ ⋮ ⋮
0 0 1
so

𝐴[𝑢1 ]𝐵 =

𝐴[𝑢2 ]𝐵 =

⋮ ⋮ ⋮

𝐴[𝑢𝑛 ]𝐵 =

Substituting these results into (2) yields

𝑎11 𝑎12 𝑎1𝑛
𝑎21 𝑎22 𝑎2𝑛
[ ⋮ ] = [𝑇(𝑢1 )]𝐵′ , [ ⋮ ] = [𝑇(𝑢2 )]𝐵′ ,… , [ ⋮ ] = [𝑇(𝑢𝑛 )]𝐵′
𝑎𝑚1 𝑎𝑚2 𝑎𝑚𝑛
which shows that the successive columns of 𝐴 are the coordinate vectors of
𝑇(𝑢1 ), 𝑇(𝑢2 ), . . . , 𝑇(𝑢𝑛 ) with respect to the basis 𝐵′ . Thus,
𝐴 = [[𝑇(𝑢1 )]𝐵′ | [𝑇(𝑢2 )]𝐵′ |… | [𝑇(𝑢𝑛 )]𝐵′ ] (3)

3
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

We will call this the matrix for 𝑇 relative to the bases 𝐵 and 𝐵′ and will denote it by
the symbol [𝑇]𝐵′ ,𝐵 .

Using this notation, Formula (3) can be written as

[𝑇]𝐵′,𝐵 = [[𝑇(𝑢1 )]𝐵′ | [𝑇(𝑢2 )]𝐵′ |… | [𝑇(𝑢𝑛 )]𝐵′ ] (4)

and from (1), this matrix has the property [𝑇]𝐵′,𝐵 [𝑥]𝐵 = [𝑇(𝑥)]𝐵′ (5)

We leave it as an exercise to show that in the special case where 𝑇𝐶 : ℝ𝑛 → ℝ𝑚 is

multiplication by 𝐶, and where 𝐵 and 𝐵′ are the standard bases for ℝ𝑛 and ℝ𝑚 ,
respectively, then
[𝑇]𝐵′,𝐵 = 𝐶 (6)

Example 1. Matrix for a Linear Transformation

Let 𝑇 ∶ 𝑃1 → 𝑃2 be the linear transformation defined by 𝑇(𝑝(𝑥)) = 𝑥𝑝(𝑥).
Find the matrix for 𝑇 with respect to the standard bases 𝐵 = {𝑢1 , 𝑢2 } and 𝐵′ =
{𝑣1 , 𝑣2 , 𝑣3 } where 𝑢1 = 1, 𝑢2 = 𝑥; 𝑣1 = 1, 𝑣2 = 𝑥, 𝑣3 = 𝑥 2

Solution.
From the given formula for 𝑇 we obtain
𝑇(𝑢1 ) = 𝑇(1) = (𝑥)(1) = 𝑥
𝑇(𝑢2 ) = 𝑇(𝑥) = (𝑥)(𝑥) = 𝑥 2

By inspection, the coordinate vectors for 𝑇(𝑢1 ) and 𝑇(𝑢2 ) relative to 𝐵′ are
0 0
[𝑇(𝑢1 )]𝐵′ = [1] , [𝑇(𝑢2 )]𝐵′ = [0]
0 1
Thus, the matrix for 𝑇 with respect to 𝐵 and 𝐵′ is
0 0
[𝑇]𝐵′ ,𝐵 = [[𝑇(𝑢1 )]𝐵′ | [𝑇(𝑢2 )]𝐵′ ] = [1 0]
0 1

Example 2. The Three-Step Procedure

Let 𝑇 ∶ 𝑃1 → 𝑃2 be the linear transformation in Example 1, and use the three-step
procedure described in the following figure to perform the computation 𝑇(𝑎 + 𝑏𝑥) =
𝑥(𝑎 + 𝑏𝑥) = 𝑎𝑥 + 𝑏𝑥 2 .

4
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

Solution.
Step 1. The coordinate matrix for 𝑥 = 𝑎 + 𝑏𝑥 relative to the basis 𝐵 = {1, 𝑥} is
𝑎
[𝑥]𝐵 = [ ]
𝑏
Step 2. Multiplying [𝑥]𝐵 by the matrix found in Example 1 we obtain
0 0 𝑎 0
[𝑇]𝐵′ ,𝐵 [𝑥]𝐵 = [1 0] [ ] = [𝑎] = [𝑇(𝑥)]𝐵′
𝑏
0 1 𝑏
Step 3. Reconstructing 𝑇(𝑥) = 𝑇(𝑎 + 𝑏𝑥) from [𝑇(𝑥)]𝐵′ we obtain 𝑇(𝑎 + 𝑏𝑥) =
0 + 𝑎𝑥 + 𝑏𝑥 2 = 𝑎𝑥 + 𝑏𝑥 2 .

Example 3. Matrix for a Linear Transformation

Let 𝑇 ∶ ℝ2 → ℝ3 be the linear transformation defined by

Find the matrix for the transformation 𝑇 with respect to the bases 𝐵 = {𝑢1 , 𝑢2 } for ℝ2
and 𝐵′ = {𝑣1 , 𝑣2 , 𝑣3 } for ℝ3 , where
1 −1 0
3 5
𝑢1 = [ ] , 𝑢2 = [ ] ; 𝑣1 = [ 0 ] , 𝑣2 = [ 2 ] , 𝑣3 = [1]
1 2
−1 2 2
Solution:
From the definition of 𝑇,

1 2
𝑇(𝑢1 ) = [−2] , 𝑇(𝑢2 ) = [ 1 ].
−5 −3

Expressing these vectors as linear combinations of the vectors in the basis 𝐵′,
𝑇(𝑢1 ) = 𝑣1 − 2𝑣3 and 𝑇(𝑢2 ) = 3𝑣1 + 𝑣2 − 𝑣3.

5
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

So,
1 3
[𝑇(𝑢1 )]𝐵′ = [ 0 ] , [𝑇(𝑢2 )]𝐵′ = [ 1 ].
−2 −1

Therefore,
1 3
[𝑇]𝐵′,𝐵 = [[𝑇(𝑢1 )]𝐵′ | [𝑇(𝑢2 )]𝐵′ ] = [ 0 1 ].
−2 −1

Matrices of Linear Operators

In the special case where 𝑉 = 𝑊 (so that 𝑇 ∶ 𝑉 → 𝑉 is a linear operator), it is usual to
take 𝐵 = 𝐵′ when constructing a matrix for 𝑇. In this case the resulting matrix is called
the matrix for 𝑇 relative to the basis 𝐵 and is usually denoted by [𝑇]𝐵 rather
than [𝑇]𝐵,𝐵 . If 𝐵 = {𝑢1 , 𝑢2 , … , 𝑢𝑛 }, then Formulas (4) and (5) become

[𝑇]𝐵 = [[𝑇(𝑢1 )]𝐵 |[𝑇(𝑢2 )]𝐵 | ··· |[𝑇(𝑢𝑛 )]𝐵 ] (7)

[𝑇]𝐵 [𝑥]𝐵 = [𝑇(𝑥)] 𝐵 (8)

In the special case where 𝑇 ∶ ℝ𝑛 → ℝ𝑛 is a matrix operator, say multiplication by 𝐴,

and 𝐵 is the standard basis for ℝ𝑛 , then Formula (7) simplifies to [𝑇]𝐵 = 𝐴 (9)

Matrices of Identity Operators

Recall that the identity operator 𝐼: 𝑉 → 𝑉 maps every vector in 𝑉 into itself, that is,
𝐼(𝑥) = 𝑥 for every vector 𝑥 in 𝑉. The following example shows that if 𝑉 is 𝑛-
dimensional, then the matrix for 𝐼 relative to any basis 𝐵 for 𝑉 is the 𝑛 × 𝑛 identity
matrix.

6
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

Example 4. Matrices of Identity Operators

If 𝐵 = {𝑢1 , 𝑢2 , … , 𝑢𝑛 } is a basis for a finite-dimensional vector space 𝑉, and if 𝐼: 𝑉 →
𝑉 is the identity operator on 𝑉, then
𝐼(𝑢1 ) = 𝑢1 , 𝐼(𝑢2 ) = 𝑢2 , . . . , 𝐼(𝑢𝑛 ) = 𝑢𝑛

[𝐼(𝑢1 )]𝐵 [𝐼(𝑢2 )]𝐵 [𝐼(𝑢𝑛 )]𝐵

Therefore,
1 0 … 0
0 1 … 0
[𝐼]𝐵 = 0 0 … 0 =𝐼
⋮ ⋮ ⋮
[0 0 … 1]
Example 5. Linear Operator on 𝑃2

Solution.
(a) From the formula for 𝑇,
𝑇(1) = 1, 𝑇(𝑥) = 3𝑥 − 5 = −5 + 3𝑥, 𝑇(𝑥 2 ) = (3𝑥 − 5)2 = 25 − 30𝑥 + 9𝑥 2 .
So,
1 −5 25
[𝑇(1)]𝐵 = [0], [𝑇(𝑥)]𝐵 = [ 3 ], [𝑇(𝑥 2 )]𝐵 = [−30].
0 0 9

Therefore,

7
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

1 −5 25
[𝑇]𝐵 = [0 3 −30].
0 0 9

(b)
1
[1 + 2𝑥 + 3𝑥 2 ]𝐵 = [2]
3

1 −5 25 1 66
[𝑇(1 + 2𝑥 + 3𝑥 2 )]𝐵 = [𝑇]𝐵 [1 + 2𝑥 + 3𝑥 2 ]𝐵 = [0 3 −30] [2] = [−84]
0 0 9 3 27

Therefore, 𝑇(1 + 2𝑥 + 3𝑥 2 ) = 66 − 84𝑥 + 27𝑥 2.

(c)
By direct computation,
𝑇(1 + 2𝑥 + 3𝑥 2 ) = 1 + 2(3𝑥 − 5) + 3(3𝑥 − 5)2
= 1 + 6𝑥 − 10 + 27𝑥 2 − 90𝑥 + 75
= 66 − 84𝑥 + 27𝑥 2 .

Matrices of Compositions
Theorem 1.
If 𝑇1 : 𝑈 → 𝑉 and 𝑇2 : 𝑉 → 𝑊 are linear transformations, and if 𝐵, 𝐵′′ and 𝐵′ are bases
for 𝑈, 𝑉, and 𝑊, respectively, then [𝑇2 ∘ 𝑇1 ]𝐵′,𝐵 = [𝑇2 ]𝐵′,𝐵′′ [𝑇1 ]𝐵′′ ,𝐵 (10)

Remark.
In (10), observe how the interior subscript 𝐵′′ (the basis for the intermediate space 𝑉)
seems to “cancel out,” leaving only the bases for the domain and image space of the

8
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

composition as subscripts. This “cancellation” of interior subscripts suggests the

following extension of Formula (10) to compositions of three linear transformations:
[𝑇3 ∘ 𝑇2 ∘ 𝑇1 ]𝐵′,𝐵 = [𝑇3 ]𝐵′,𝐵′′′ [𝑇2 ]𝐵′′′,𝐵′′ [𝑇1 ]𝐵′′ ,𝐵 (11)

The following example illustrates Theorem 1.

Example 6. Composition

In this example, 𝑃1 plays the role of 𝑈 in Theorem 1, and 𝑃2 plays the roles of both 𝑉
and 𝑊; thus we can take 𝐵′ = 𝐵′′ in (10) so that the formula simplifies to

9
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

−5 25 𝑐
0
[(𝑇2 ∘ 𝑇1 )(𝑐0 + 𝑐1 𝑥)]𝐵′ = [𝑇2 ∘ 𝑇1 ]𝐵′,𝐵 [𝑐0 + 𝑐1 𝑥]𝐵 = [ 3 −30] [𝑐 ]
1
0 9
(−5𝑐0 + 25𝑐1 )
= [ (3𝑐0 − 30𝑐1 ) ].
9𝑐1
Therefore,
(𝑇2 ∘ 𝑇1 )(𝑐0 + 𝑐1 𝑥) = (−5𝑐0 + 25𝑐1 )1 + (3𝑐0 − 30𝑐1 )𝑥 + 9𝑐1 𝑥 2
= 𝑐0 (3𝑥 − 5) + 𝑐1 (9𝑥 2 − 30𝑥 + 25) = 𝑐0 (3𝑥 − 5) + 𝑐1 (3𝑥 − 5)2

10
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

5.2 Column and Row Space of a Matrix

Definition 3.
For an 𝑚 × 𝑛 matrix
𝑎11 𝑎12 … 𝑎1𝑛
𝑎21 𝑎22 … 𝑎2𝑛
𝐴 =[ ⋮ ⋮ … ⋮ ]
𝑎𝑚1 𝑎𝑚2 … 𝑎𝑚𝑛
the vectors
𝑟1 = [𝑎11 𝑎12 · · · 𝑎1𝑛 ]
𝑟2 = [𝑎21 𝑎22 · · · 𝑎2𝑛 ]
⋮ ⋮
𝑟𝑚 = [𝑎𝑚1 𝑎𝑚2 · · · 𝑎𝑚𝑛 ]
in ℝn that are formed from the rows of 𝐴 are called the row vectors of 𝐴, and the vectors
𝑎11 𝑎11 𝑎1𝑛
𝑎21 𝑎22 𝑎2𝑛
𝑐1 = [ ⋮ ] , 𝑐2 = [ ⋮ ] , … , 𝑐𝑛 = [ ⋮ ]
𝑎𝑚1 𝑎𝑚2 𝑎𝑚𝑛
in ℝm formed from the columns of 𝐴 are called the column vectors of 𝐴.

Example 1. Row and Column Vectors of a 2 × 3 Matrix

Let
2 1 0
𝐴 =[ ]
3 −1 4
The row vectors of 𝐴 are
𝑟1 = [2 1 0] and 𝑟2 = [3 − 1 4]
and the column vectors of 𝐴 are
2 1 0
𝑐1 = [ ], 𝑐2 = [ ] and 𝑐3 = [ ]
3 −1 4

Definition 4.
If 𝐴 is an 𝑚 × 𝑛 matrix, then the subspace of ℝ𝑛 spanned by the row vectors of 𝐴 is
called the row space of 𝐴, and the subspace of ℝ𝑚 spanned by the column vectors of 𝐴
is called the column space of 𝐴.

• We will denote the row space of 𝐴 and the column space of 𝐴 by row(𝐴) and
col(𝐴) respectively.

11
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

Theorem 3. Elementary row operations do not change the row space of a matrix.
The proof is omitted.

Theorem 3 might tempt you into incorrectly believing that elementary row operations
do not change the column space of a matrix. To see why this is not true, compare the
matrices
1 3 1 3
𝐴 =[ ] and 𝐵 = [ ]
2 6 0 0
The matrix 𝐵 can be obtained from 𝐴 by adding −2 times the first row to the second.
However, this operation has changed the column space of 𝐴, since that column space
consists of all scalar multiples of
1
[ ]
2
whereas the column space of 𝐵 consists of all scalar multiples of
1
[ ]
0
and the two are different spaces.

Theorem 4. If a matrix 𝑅 is in row echelon form, then the row vectors with the leading
1’s (the nonzero row vectors) form a basis for the row space of 𝑅, and the column
vectors with the leading 1’s of the row vectors form a basis for the column space of 𝑅.
The proof is omitted.

Example 1. Bases for the Row and Column Spaces of a Matrix in Row Echelon
Form
Find bases for the row and column spaces of the matrix
1 −2 5 0 3
0 1 3 0 0
𝑅=[ ]
0 0 0 1 0
0 0 0 0 0
Solution. Since the matrix 𝑅 is in row echelon form, it follows from Theorem 4, that
the vectors
𝑟1 = [1 −2 5 0 3]
𝑟2 = [0 1 3 0 0]
𝑟3 = [0 0 0 1 0]

12
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

form a basis for the row space of 𝑅, and the vectors

1 −2 0
0 1 0
𝑐1 = [ ] , 𝑐2 = [ ] , 𝑐4 = [ ]
0 0 1
0 0 0
form a basis for the column space of 𝑅.

Example 2. Basis for a Row Space by Row Reduction

Find a basis for the row space of the matrix
1 −3 4 −2 5 4
2 −6 9 −1 8 2
𝐴=[ ]
2 −6 9 −1 9 7
−1 3 −4 2 −5 −4
Solution. Since elementary row operations do not change the row space of a matrix, we
can find a basis for the row space of 𝐴 by finding a basis for the row space of any row
echelon form of 𝐴. Reducing 𝐴 to row echelon form, we obtain (verify)
1 −3 4 −2 5 4
0 0 1 3 −2 −6
𝑅=[ ]
0 0 0 0 1 5
0 0 0 0 0 0
By Theorem 4, the nonzero row vectors of 𝑅 form a basis for the row space of 𝑅 and
hence form a basis for the row space of 𝐴. These basis vectors are
𝑟1 = [1 −3 4 −2 5 4]
𝑟2 = [0 0 1 3 −2 −6]
𝑟3 = [0 0 0 0 1 5]

Theorem 5. If 𝐴 and 𝐵 are row equivalent matrices, then:

(a) A given set of column vectors of 𝐴 is linearly independent if and only if the
corresponding column vectors of 𝐵 are linearly independent.
(b) A given set of column vectors of 𝐴 forms a basis for the column space of 𝐴 if and
only if the corresponding column vectors of 𝐵 form a basis for the column space of 𝐵.
Proof is omitted.

Although elementary row operations can change the column space of a matrix, it
follows from Theorem 5 that they do not change the dimension of its column space.

13
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

Example 3. Basis for a Column Space by Row Reduction

Find a basis for the column space of the matrix that consists of column vectors of
1 −3 4 −2 5 4
2 −6 9 −1 8 2
𝐴=[ ]
2 −6 9 −1 9 7
−1 3 −4 2 −5 −4
Solution.
We observed in Example 2 that the matrix
1 −3 4 −2 5 4
0 0 1 3 −2 −6
𝑅=[ ]
0 0 0 0 1 5
0 0 0 0 0 0

is a row echelon form of 𝐴. Keeping in mind that 𝐴 and 𝑅 can have different column
spaces, we cannot find a basis for the column space of 𝐴 directly from the column
vectors of 𝑅. However, it follows from Theorem 5 that if we can find a set of column
vectors of 𝑅 that forms a basis for the column space of 𝑅, then the corresponding
column vectors of 𝐴 will form a basis for the column space of 𝐴.
Since the first, third, and fifth columns of 𝑅 contain the leading 1’s of the row vectors,
the vectors
1 4 5
0 1 −2
𝑘1 = [ ] , 𝑘2 = [ ] , 𝑘4 = [ ]
0 0 1
0 0 0
form a basis for the column space of 𝑅. Thus, the corresponding column vectors of 𝐴,
which are
1 4 5
2 9 8
𝑐1 = [ ] , 𝑐2 = [ ] , 𝑐4 = [ ]
2 9 9
−1 −4 −5

form a basis for the column space of 𝐴.

Example 4. Basis for the Space Spanned by a Set of Vectors

The following vectors span a subspace of ℝ4 . Find a subset of these vectors that forms
a basis of this subspace.
𝑣1 = (1,2,2, −1), 𝑣2 = (−3, −6, −6,3),

14
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

𝑣3 = (4,9,9, −4), 𝑣4 = (−2, −1, −1,2),

𝑣5 = (5,8,9, −5), 𝑣6 = (4,2,7, −4)

Solution.
If we rewrite these vectors in column form and construct the matrix that has those
vectors as its successive columns, then we obtain the matrix 𝐴 in Example 3 (verify).
Thus,
span{𝑣1 , 𝑣2 , 𝑣3 , 𝑣4 , 𝑣5 , 𝑣6 } = col(𝐴)

Proceeding as in that example (and adjusting the notation appropriately), we see that
the vectors 𝑣1 , 𝑣3 , and 𝑣5 form a basis for
span{𝑣1 , 𝑣2 , 𝑣3 , 𝑣4 , 𝑣5 , 𝑣6 }

In Example 2, we found a basis for the row space of a matrix by reducing that matrix to
row echelon form. However, the basis vectors produced by that method were not all
row vectors of the original matrix. The following adaptation of the technique used in
Example 3 shows how to find a basis for the row space of a matrix that consists entirely
of row vectors of that matrix.

Example 5. Basis for the Row Space of a Matrix

Find a basis for the row space of
1 −2 0 0 3
2 −5 −3 −2 6
𝐴=[ ]
0 5 15 10 0
2 6 18 8 6
consisting entirely of row vectors from 𝐴.
Solution.
We will transpose 𝐴, thereby converting the row space of 𝐴 into the column space of
𝐴𝑇 ; then we will use the method of Example 3 to find a basis for the column space of
𝐴𝑇 ; and then we will transpose again to convert column vectors back to row vectors.
Transposing 𝐴 yields

and then reducing this matrix to row echelon form we obtain

15
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

The first, second, and fourth columns contain the leading 1’s, so the corresponding
column vectors in 𝐴𝑇 form a basis for the column space of 𝐴𝑇 ; these are
1 2 2
−2 −5 6
𝑐1 = 0 , 𝑐2 = −3 , 𝑐4 = 18
0 −2 8
[3] [6] [6]
Transposing again and adjusting the notation appropriately yields the basis vectors
𝑟1 = [1 − 2 0 0 3], 𝑟2 = [2 − 5 − 3 − 2 6],
𝑟4 = [2 6 18 8 6]
for the row space of 𝐴.

Problem. Given a set of vectors 𝑆 = {𝑣1 , 𝑣2 , . . . , 𝑣𝑛 } in ℝn , find a subset of these

vectors that forms a basis for span(𝑆), and express each vector that is not in that basis
as a linear combination of the basis vectors.

Example 6. Basis and Linear Combinations

(a) Find a subset of the vectors
𝑣1 = (1, −2, 0, 3), 𝑣2 = (2, −5, −3, 6),
𝑣3 = (0, 1, 3, 0), 𝑣4 = (2, −1, 4, −7), 𝑣5 = (5, −8, 1, 2)
that forms a basis for the subspace of ℝ4 spanned by these vectors.

(b) Express each vector not in the basis as a linear combination of the basis vectors.

Solution (a).
We begin by constructing a matrix that has 𝑣1 , 𝑣2 , . . . , 𝑣5 as its column vectors:

16
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

(1)

𝑣1 𝑣2 𝑣3 𝑣4 𝑣5
The first part of our problem can be solved by finding a basis for the column space of
this matrix. Reducing the matrix to reduced row echelon form and denoting the column
vectors of the resulting matrix by 𝑤1 , 𝑤2 , 𝑤3 , 𝑤4, and 𝑤5 yields

(2)

𝑤1 𝑤2 𝑤3 𝑤4 𝑤5

The leading 1’s occur in columns 1, 2, and 4, so by Theorem 4,

{𝑤1 , 𝑤2 , 𝑤4 }
is a basis for the column space of (2), and consequently,
{𝑣1 , 𝑣2 ,𝑣4 }
is a basis for the column space of (1).

Solution (b).
We will start by expressing 𝑤3 and 𝑤5 as linear combinations of the basis vectors 𝑤1 ,
𝑤2 , 𝑤4. The simplest way of doing this is to express 𝑤3 and 𝑤5 in terms of basis vectors
with smaller subscripts. Accordingly, we will express 𝑤3 as a linear combination of 𝑤1
and 𝑤2 , and we will express 𝑤5 as a linear combination of 𝑤1 , 𝑤2 , and 𝑤4 . By inspection
of (2), these linear combinations are
𝑤3 = 2𝑤1 − 𝑤2
𝑤5 = 𝑤1 + 𝑤2 + 𝑤4
We call these the dependency equations. The corresponding relationships in (1) are
𝑣3 = 2𝑣1 − 𝑣2
𝑣5 = 𝑣1 + 𝑣2 + 𝑣4

17
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

Basis for the Space Spanned by a Set of Vectors (a summary of the steps of above
example)
Step 1. Form the matrix 𝐴 whose columns are the vectors in the set 𝑆 = {𝑣1 , 𝑣2 , . . . , 𝑣𝑘 }.
Step 2. Reduce the matrix 𝐴 to reduced row echelon form 𝑅.
Step 3. Denote the column vectors of 𝑅 by 𝑤1 , 𝑤2 ,…, 𝑤𝑘 .
Step 4. Identify the columns of 𝑅 that contain the leading 1’s. The corresponding
column vectors of 𝐴 form a basis for span(𝑆).
This completes the first part of the problem.
Step 5. Obtain a set of dependency equations for the column vectors 𝑤1 , 𝑤2 ,…, 𝑤𝑘 of
𝑅 by successively expressing each 𝑤𝑖 that does not contain a leading 1 of 𝑅 as a linear
combination of predecessors that do.
Step 6. In each dependency equation obtained in Step 5, replace the vector 𝑤𝑖 by the
vector 𝑣𝑖 for 𝑖 = 1, 2, . . . , 𝑘.
This completes the second part of the problem.

Rank

Theorem 6. The row space and the column space of a matrix 𝐴 have the same
dimension.
Proof is omitted.

Definition 5.
The common dimension of the row space and column space of a matrix 𝐴 is called the
rank of 𝐴 and is denoted by rank(𝐴).

Example 1. Rank of a 4 × 6 Matrix

Find the rank of the matrix

Example 2. Maximum Value for Rank

18
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa
 Linear Algebra - MA2034

What is the maximum possible rank of an 𝑚 × 𝑛 matrix 𝐴 that is not square?

Solution.
Since the row vectors of 𝐴 lie in ℝ𝑛 and the column vectors in ℝ𝑚 , the row space of 𝐴
is at most 𝑛-dimensional and the column space is at most 𝑚-dimensional. Since the
rank of 𝐴 is the common dimension of its row and column space, it follows that the rank
is at most the smaller of 𝑚 and 𝑛. We denote this by writing
rank(𝐴) ≤min(𝑚, 𝑛)
in which 𝑚𝑖𝑛(𝑚, 𝑛) is the minimum of 𝑚 and 𝑛.

Theorem 7. If A is any matrix, then rank(𝐴) = rank(𝐴𝑇 ).

Proof. rank(𝐴) = dim(row space of 𝐴) = dim(column space of 𝐴𝑇 ) = rank(𝐴𝑇 ).

19
Dr. Supem Samarasiri, Department of Mathematics, University of Moratuwa