Row Echelon Form and Reduced Row Echelon Form

These notes closely follow the presentation of the material given in David C. Lay’s
textbook Linear Algebra and its Applications (3rd edition). These notes are intended
primarily for in-class presentation and should not be regarded as a substitute for
thoroughly reading the textbook itself and working through the exercises therein.
Row Echelon Form and Reduced Row Echelon Form
A non–zero row of a matrix is defined to be a row that does not contain all zeros.
The leading entry of a non–zero row of a matrix is defined to be the leftmost
non–zero entry in the row.
For example, if we have the matrix
0 0 4 1 0
0 0 0
0
0
0 0 0
0
3
,
then the first row is a non–zero row with leading entry 4, the second row is a zero row,
and the third row is a non–zero row with leading entry 3.
Definition A matrix is said to have echelon form (or row echelon form) if it has the
following properties:
1.
2.
3.
4.
5.
All non–zero rows are above any zero rows.
Each leading entry of a each non–zero row is in a column to the right of the
leading entry of the row above it.
In each column that contains a leading entry, each entry below the leading
entry is 0.
If a matrix has row echelon form and also satisfies the following two
conditions, then the matrix is said to have reduced echelon form (or
reduced row echelon form):
The leading entry in each non–zero row is 1.
Each leading 1 is the only non–zero entry in its column.
Quiz
Decide whether or not each of the following matrices has row echelon form. For each
that does have row echelon form, decide whether or not it also has reduced row
echelon form.
1.
0 0 4 1 0
0 0 0
0
0
0 0 0
0
3
2.
1
1 1 0 1
0 0 1 1
0 0 0 0
3.
1 1 0 0
0 1 1 0
0 0 1 1
4.
1 0 0 0
1 1 0 0
0 1 1 0
0 0 1 1
5.
0 1 1 1 1
0 0 2 2 2
0 0 0 0 3
0 0 0 0 0
Given any matrix, we can always perform a sequence of elementary row
operations to arrive at an equivalent matrix that has row echelon form. In fact, we can
always perform a sequence of row operations to arrive at an equivalent matrix that has
reduced row echelon form. For any non–zero matrix, there are infinitely many
equivalent matrices that have row echelon form. However, there is only one equivalent
matrix that has reduced row echelon form. (This is proved in Appendix A of the
textbook, but we will not prove it in this course. We will just accept it to be true.)
Theorem (1) Every matrix is equivalent to exactly one matrix that has reduced row echelon
form.
Example Find infinitely many different matrices that have row echelon form and that are
equivalent to the matrix
0 0 4 1 0
0 0 0 0 0
.
0 0 0 0 3
2
Then find the unique matrix that has reduced row echelon form and that is
equivalent to this matrix.
Solution By performing an interchange operation, we obtain
0 0 4 1 0
0 0 4 1 0

0 0 0 0 0
.
0 0 0 0 3
0 0 0 0 3
0 0 0 0 0
The matrix on the right is equivalent to the matrix on the left and has row echelon
form. If we like, we can now scale the top row by 3 to obtain the matrix
0 0 12 3 0
0 0 0
0 3
0 0 0
0 0
which also has row echelon form and which is also equivalent to the original
matrix.
Clearly, we can obtain an infinite number of such matrices by continuing to
scale the first or second rows by whatever non–zero number we like.
To find the unique reduced row echelon matrix that is equivalent to the original
matrix, we continue the row operations started above as follows:
0 0 4 1 0
0 0 0 0 0
0 0 4 1 0

0 0 0 0 3
0 0 1  14 0

0 0 0 0 3
0 0 0 0 0
0 0 0
0
3
0 0 0
0
0

0 0 1  14 0
0 0 0
0
1
0 0 0
0
0
The matrix on the far right has reduced row echelon form and is equivalent to the
original matrix. No other such reduced echelon matrix can be found. (This is
guaranteed by Theorem 1.)
Notation For any matrix A, we will use the notation rrefA to denote the unique matrix
having reduced row echelon form and equivalent to A. (This notation is not used in
the textbook we are using but it is done in many other books.)
For example, if
0 0 4 1 0
A
0 0 0
0
0
0 0 0
0
3
,
then
rrefA 
0 0 1  14
0
0 0 0
0
1
0 0 0
0
0
3
.
Pivot Positions and Pivot Columns
Definition A pivot position in a matrix, A, is a location in A that corresponds to a leading
1 in rrefA. A pivot column in A is a column of A that contains a pivot position.
For example, look at the matrices A and rrefA in the example done above. The pivot
positions in A are the positions indicated below:
0 0
1
4
pivot position
A
0 0
0
0
0
0
pivot position
0 0
0
0
3
and the pivot columns of A are the third and fifth columns. Note that we can only tell
what the pivot positions and pivot columns of a matrix, A, are after we have found
rrefA.
Example Find the pivot positions and pivot columns of the matrix
A
0 3 6 4
9
1 2 1 3
1
2 3 0
1
4
.
3 1
5 9 7
Solution Since
1 0 3 0 5
rrefA 
0 1 2 0 3
0 0 0 1 0
,
0 0 0 0 0
we see that the pivot positions of A are as indicated below:
3
6
4
9
2
1
3
1
2
3
0
3
1
1
4
5
0
pivot position
A
1
pivot position
pivot position
9
.
7
The pivot columns of A are the first, second, and fourth columns.
4
The Row Reduction Algorithm
Here is an algorithm that always works for finding rrefA for any matrix A.
1. Begin with the leftmost non–zero column. This is a pivot column. The pivot
position is at the top.
2. If the top entry in this column is 0, then interchange the top row with some
other row that has a non–zero entry as its first entry. Then scale the top row
to make the leading entry be a 1.
3. Use replacement operations to create zeros in every position in this column
below the pivot position.
4. Cover (or ignore) the row containing the pivot position and cover (or ignore)
any rows above it. Repeat steps 1, 2, and 3 for the submatrix that remains.
Repeat the process until there is no non–zero column in the submatrix that
remains.
5. Beginning with the rightmost pivot position and working upward and to the
left, create zeros above each pivot position by using replacement operations.
Example For the matrix
0 1 1 0
A
,
2 0 5 5
4 4 5 0
use the reduction algorithm to find rrefA.
Answer:
1 0 0  452
rrefA 
0 1 0
10
0 0 1
10
.
Solutions of Linear Systems
It is easy to find the solution set of a linear system whose augmented matrix has
reduced row echelon form. Also, if A is the augmented matrix of a system, then the
solution set of this system is the same as the solution set of the system whose
augmented matrix is rrefA (since the matrices A and rrefA are equivalent). Thus,
finding rrefA allows us to solve any given linear system. This is illustrated in the three
examples that follow. In the first example, it turns out that the system is inconsistent.
In the second example, the system has infinitely many solutions. In the third example,
the system has a unique solution.
5
Example Find the solution set of the linear system
3x 1  4x 2  9
2x 1  4x 2  x 3  0
10x 1  2x 3  4.
Solution The augmented matrix of this system is
3 4 0
A
2
9
4 1 0
10 0 2 4
and
1 0  15
rrefA 
0
0 1  203 0
0 0
0
.
1
Since rrefA is the augmented matrix of the linear system
x1  1 x3  0
5
3
x2 
x 0
20 3
01
which obviously has no solution (because of the equation 0  1), we conclude that
the original system that was given is also inconsistent. (Its solution set is the empty
set.)
As an important remark, note that what causes this system to be inconsistent is
the fact that the last column of its augmented matrix is a pivot column.
Example Find the solution set of the linear system
3x 1  4x 2  9
2x 1  4x 2  x 3  0
10x 1  2x 3  18.
Solution The augmented matrix of this system is
3 4 0
A
2
9
4 1 0
10 0 2 18
and
6
1 0  15
rrefA 
0 1  203
0 0
9
5
 109
0
.
0
The matrix rrefA is the augmented matrix for the linear system
x1  1 x3  9
5
5
3
x2 
x  9
10
20 3
00
This system is consistent and has infinitely many solutions. To obtain a solution,
we can let x 3 be any number that we like and then let
x2   9  3 x3,
10
20
and then let
x1  9  1 x3.
5
5
Since x 3 can be any number that we like (for which reason we say that x 3 is a free
variable), we see that the system under consideration has infinitely many
solutions. As an example of a particular solution, suppose we let x 3  10. Then
x 2   9  3 10  3
5
10
20
and
x 1  9  1 10  19 .
5
5
5
Thus, the ordered triple 195 , 35 , 10 is a solution of the original system. Let us
check:
3 19  4 3  9
5
5
19
3
4
2
 10  0
5
5
10 19  210  18.
5
As an important remark, note that was causes this system to have infinitely
many solutions is the fact the last column of its augmented matrix is not a pivot
column (thus making the system consistent) together with the fact that not every
column of its coefficient matrix is a pivot column (allowing the system to have free
variables).
Example Find the solution set of the linear system
3x 1  4x 2  9
2x 1  4x 2  x 3  0
x 1  x 3  0.
7
Solution The augmented matrix of this system is
3 4 0 9
A
2 4 1 0
1 0 1 0
and
1 0 0
rrefA 
0 1 0
0 0 1
9
4
 169
9
4
.
Since rrefA is the augmented matrix of the linear system
x1  9
4
x2   9
16
x3  9 ,
4
we conclude that the only solution of the original system is
9
4
,  169 ,
9
4
.
Note that the reason that this system has a unique solution is that its last
column is not a pivot column (meaning that the system is consistent) together with
the fact that every column of its coefficient matrix is a pivot column (meaning that
the system has no free variables).
Answering the “Existence and Uniqueness”
Questions
Suppose that we have a linear system whose coefficient matrix is A and whose
augmented matrix is B. Then B is the same as A except for the fact that B has an
extra column on the right. Because of the way that elementary row operations work,
the matrix rrefB is also the same as rrefA except, once again, for the fact that
rrefB has an extra column on the right. Inspired by the preceding three examples, we
arrive at the following criteria for answering the questions:
1. Does our system have a solution? (Existence)
2. If so, then does it have just one solution or infinitely many? (Uniqueness)
Theorem (2) For a linear system with coefficient matrix A and augmented matrix B :
1.
2.
The system is consistent if and only if the last column of B is not a pivot
column.
If the system is consistent, then the system has a unique solution if and only
if every column of A is a pivot column.
8
Example Answer the existence and uniqueness questions for the system
x 1  2x 2  5x 3  6x 4  5
x 2  6x 3  3x 4  2
x 4  5x 5  5.
Solution The coefficient matrix for this system is
1 2 5 6 0
A
0 1 6 3 0
0 0 0
1 5
and the augmented matrix is
1 2 5 6 0 5
B
0 1 6 3 0
0 0 0
.
2
1 5 5
Also,
1 0 7 0
rrefB 
9
0
0 1 6 0 15 13
0 0 0 1 5
5
from which we can immediately conclude that the system is consistent (because the
last column of B is not a pivot column).
In addition,
1 0 7 0
rrefA 
0
0 1 6 0 15
,
0 0 0 1 5
from which we conclude that the system has infinitely many solutions (rather than
a unique solution) because not every column of A is a pivot column.
As an important remark, we note that it is not necessary to compute rrefA
separately once rrefB has been computed. This is because rrefA will always be
the same as rrefB with the last column deleted.
Example For the system given in the above example (which was found to have infinitely
many solutions), give a parametric description of its solution set.
Solution Looking at the matrix rrefB in the above example, we see that the system in
that example is equivalent to the system
9
x 1  7x 3  9
x 2  6x 3  15x 5  13
x 4  5x 5  5.
The free variables for this system (corresponding to the non–pivot columns in
rrefA) are x 3 and x 5 . The general solution of the system (described in parametric
form) is
x 1  9  7x 3
x 2  13  6x 3  15x 5
x 3  free
x 4  5  5x 5
x 5  free.
10