Section 2.2.
Elementary Row Operations
Example 2.2.1 Find all solutions of the following system :
x
3x
x
+
+
+
2y
y
4y
z
- 2z
+ 2z
=
=
=
5
9
0
In other (perhaps simpler) examples we were able to find solutions by simplifying
the system (perhaps by eliminating certain variables) through operations of the
following types :
1. We could multiply one equation by a non-zero constant.
2. We could add one equation to another (for example in the hope of eliminating
a variable from the result).
A similar approach will work for Example 2.2.1 but with this and other harder
examples it may not always be clear how to proceed. We now develop a new technique both for describing our system and for applying operations of the above types
more systematically and with greater clarity.
Back to Example 2.2.1 We associate a matrix to our system of equations (a matrix
is an array of numbers).
x
3x
x
+
+
+
2y
y
4y
z
- 2z
+ 2z
=
=
=
5
9
0
1 2 1 5
Eqn 1
3 1 2 9 Eqn 2
1 4
2 0
Eqn 3
Note that the first row of this matrix contains as its four entries the coefficients of
the variables x, y, z, and the number appearing on the right-hand-side of Equation
1 of the system. Rows 2 and 3 correspond similarly to Equations 2 and 3. The
columns of the matrix correspond (from left to right) to the variables x, y, z and
the right hand side of our system of equations.
Definition 2.2.2 The above matrix is called the augmented matrix of the system of
equations in Example 2.2.1 .
In solving systems of equations we are allowed to perform operations of the
following types :
1
1. Multiply an equation by a non-zero constant.
2. Add one equation (or a non-zero constant multiple of one equation) to another
equation.
These correspond to the following operations on the augmented matrix :
1. Multiply a row by a non-zero constant.
2. Add a multiple of one row to another row.
3. We also allow operations of the following type : Interchange two rows in the
matrix (this only amounts to writing down the equations of the system in a
different order).
Definition 2.2.3 Operations of these three types are called Elementary Row Operations (EROs) on a matrix.
We now describe how EROs on the augmented matrix can be used to solve the
system of Example 2.2.1 . The following table describes how an ERO is performed
at each step to produce a new augmented matrix corresponding to a new (hopefully
simpler) system.
ERO
Matrix
System
1 2 1 5
3 1 2 9
1 4
2 0
x + 2y z = 5
3x + y 2z = 9
x + 4y + 2z = 0
1.
R3 R3 + R1
1 2 1 5
3 1 2 9
0 6
1 5
1
2 1
5
1 6
2. R2 R2 3R1 0 5
0
6
1
5
3.
R2 R2 + R3
4. R3 R3 6R2
5.
1
R3 11
1 2 1
5
0 1
2 1
0 6
1
5
1 2 1
5
0 1
2 1
0 0 11 11
1 2 1
5
0 1
2 1
0 0
1 1
x + 2y z = 5
3x + y 2z = 9
6y + z = 5
x + 2y z =
5
5y + z = 6
6y + z =
5
x + 2y z =
5
y + 2z = 1
6y + z =
5
x + 2y
y +
z =
5
2z = 1
11z = 11
x + 2y z =
5 (A)
y + 2z = 1 (B)
z = 1 (C)
We have produced a new system of equations. This is easily solved :
(C) z = 1
(B) y = 1 2z
= y = 1 2(1) = 1
Backsubstitution
(A) x = 5 2y + z = x = 5 2(1) + (1) = 2
Solution : x = 2, y = 1, z = 1
You should check that this is a solution of the original system. It is the only
solution both of the final system and of the original one (and every intermediate
one).
Note : The matrix obtained in Step 5 above is in Row-Echelon Form. This means
:
1. The first non-zero entry in each row is a 1 (called a Leading 1 ).
2. If a column contains a leading 1, then every entry of the column below the
leading 1 is a zero.
3
3. As we move downwards through the rows of the matrix, the leading 1s move
from left to right.
4. Any rows consisting entirely of zeroes are grouped together at the bottom of
the matrix.
Note : The process by which the augmented matrix of a system of equations is
reduced to row-echelon form is called Gaussian Elimination. In Example 2.2.1 the
solution of the system was found by Gaussian elimination with Backsubstitution.
General Strategy to Obtain a Row-Echelon Form
1. Get a 1 as the top left entry of the matrix.
2. Use this first leading 1 to clear out the rest of the first column, by adding
suitable multiples of Row 1 to subsequent rows.
3. If column 2 contains non-zero entries (other than in the first row), use EROs
to get a 1 as the second entry of Row 2. After this step the matrix will look
like the following (where the entries represented by stars may be anything):
1 ... ...
0 1 ... ... ...
0 ... ... ...
0 ... ... ...
.. ..
..
. .
.
0 ... ... ...
4. Now use this second leading 1 to clear out the rest of column 2 (below Row
2) by adding suitable multiples of Row 2 to subsequent rows. After this step
the matrix will look like the following :
1 ... ...
0 1 ... ...
0 0 ... ...
0 0 ... ...
.. .. .. ..
..
. . . .
.
0 0 ... ...
5. Now go to column 3. If it has non-zero entries (other than in the first two
rows) get a 1 as the third entry of Row 3. Use this third leading 1 to clear out
the rest of Column 3, then proceed to column 4. Continue until a row-echelon
form is obtained.
4
Example 2.2.4 (Summer 2001, Q6 (a))
Let A be the matrix
1 1 1 2 0
2
1 1 2 8
1 3
2 7 2
Reduce A to row-echelon form.
Solution:
1. Get a 1 as the first entry of Row 1. Done.
2. Use this first leading 1 to clear out column 1 as follows :
1 1 1
2 0
0
R2 R2 2R1
3
1 2 8
R3 R3 R1
0 2
3
5 2
3. Get a leading 1 as the second entry of Row 2, for example as follows :
1 1 1 2 0
0
R2 R2 + R3
1
4 3 10
0 2
3 5 2
4. Use this leading 1 to clear out whatever appears below it in Column 2 :
1 1 1 2 0
0
R3 R3 + 2R2
1
4 3 10
0
0 11 11 22
5. Get a leading 1 in Row 3 :
R3
1
11
1 1 1 2 0
0
1
4 3 10
0
0
1 1 2
This matrix is now in row-echelon form.
Definition 2.2.5 Let A be a matrix. The rank of A, denoted rank(A) is the number
of leading 1s in a row-echelon form obtained from A by Gaussian elimination as
above.
Remarks :
1. The rank of the matrix A in Example 2.2.4 is 3, since the row-echelon form
obtained had 3 leading 1s (one in each row).
2. The rank of any matrix can be at most equal to the number of rows, since
each row in a REF (row-echelon form) can contain at most one leading 1. If a
REF obtained from some matrix contains rows full of zeroes, the rank of this
matrix will be less than the number of rows.
3. Starting with a particular matrix, different sequences of EROs can lead to
different row-echelon forms. However, all have the same rank.
