Chapter 4 Gauss-Jordan Elimination

Systems of Linear Equations; ƒ Any linear system must have exactly one solution,
solution no
Matrices solution, or an infinite number of solutions.
ƒ Previously we considered the 2 × 2 case, in which the term
consistent is used to describe a system with a unique
Section 3 solution, inconsistent is used to describe a system with no
solution, and dependent is used for a system with an
Gauss-Jordan Elimination infinite number of solutions. In this section we will
consider larger systems with more variables and more
equations, but the same three terms are used to describe
them.

Carl Friedrich Gauss Matrix Representations of Consistent,

1777-1855 Inconsistent and Dependent Systems

At the age of seven, Carl

Friedrich Gauss started ƒ The following matrix representations of three linear
elementary school, and his equations in three unknowns illustrate the three different
potential was noticed almost cases:
immediately. His teacher,
Büttner, and his assistant, ƒ Case I: consistent ƒ From this matrix
Martin Bartels, were amazed representation, you can
when Gauss summed the ⎛1 0 0 3⎞ determine that
⎜ ⎟
integers from 1 to 100 x=3
instantly by spotting that the
⎜0 1 0 4⎟
y=4
⎜ 5 ⎟⎠
sum was 50 pairs of numbers, ⎝0 0 1 z=5
each pair summing to 101.

3 4
 Matrix Representations Matrix Representations
(continued) (continued)

ƒ Case 2: inconsistent ƒ From the second row of ƒ Case 3: dependent ƒ When there are fewer nonnon-
the matrix, we find that zero rows of a system than
⎛1 2 3 4⎞ 0x + 0y +0z =6 ⎛1 2 3 4⎞ there are variables, there
⎜ ⎟ ⎜ ⎟ will be infinitely many
or
⎜0 0 0 6⎟ ⎜0 0 0 0⎟ solutions, and therefore
⎜0 0 0 0 ⎟⎠
0 = 6, ⎜0 0 0 0 ⎟⎠ the system is dependent.
⎝ an impossible equation. ⎝
From this, we conclude
that there are no solutions
to the linear system.

5 6

Examples of Reduced
Reduced Row Echelon Form
Row Echelon Form

ƒ A matrix is said to be in reduced row echelon form or, or ⎛1 2 3 4⎞ ⎛1 0 0 3⎞

⎜ ⎟ ⎜ ⎟
more simply, in reduced form, if
⎜0 0 0 6⎟ ⎜0 1 0 4⎟
• Each row consisting entirely of zeros is below any row ⎜0 0 0 0 ⎟⎠ ⎜0 0 1
having at least one non-zero element. ⎝ ⎝ 5 ⎟⎠
• The leftmost nonzero element in each row is 1.
• All other elements in the column containing the
⎛1 3 0 0 − 2⎞
l ft
leftmostt 1 off a given
i row are zeros. ⎜ ⎟
• The leftmost 1 in any row is to the right of the leftmost ⎜0 0 1 0 7 ⎟
1 in the row above. ⎜0 0 0 1 8 ⎟⎠
⎝
7 8
 Solving a System Solving a System
Using Gauss-Jordan Elimination Using Gauss-Jordan Elimination

ƒ Example: Solve ƒ Example: Solve

x + y – z = –2 x + y – z = –2
2x – y + z = 5 2x – y + z = 5
–x + 2y + 2z = 1 –x + 2y + 2z = 1
ƒ Solution:
We begin by writing the system as an augmented matrix

9 10

Example Example
(continued) (continued)

We already have a 1 in the

⎡ 1 1 −1 −2 ⎤ The second 0 can be obtained
diagonal position of first column. ⎢ 2 −1 1 5 ⎥ ⇒ by adding row 1 to row 3:
Now we want 0’s below the 1. ⎢ ⎥
The first 0 can be obtained by ƒ Row 1 is unchanged
⎢⎣ −1 2 2 1 ⎥⎦
multiplying row 1 by –2 and ƒ Row 2 is unchanged
adding the results to row 2: ƒ Row 1 is added to Row 3
ƒ Row 1 is unchanged
ƒ (–2) times Row 1 is added to
Row 2
ƒ Row 3 is unchanged

11 12
 Example Example
(continued) (continued)

Moving to the second To obtain a 0 below the 1 ,

column, we want a 1 in the we multiply row 2 by –3 and
diagonal position (where add it to the third row:
there was a –3). We get this ƒ Row 1 is unchanged
by dividing every element in ƒ Row 2 is unchanged
row 2 by -3:
ƒ (–3) times row 2 is added
ƒ Row 1 is unchanged t row 3
to
ƒ Row 2 is divided by –3
ƒ Row 3 is unchanged

13 14

Example Example
(continued) (continued)

We can now work upwards to get

To obtain a 1 in the third zeros in
i the
th third
thi d column,
l above
b
position of the third row, we the 1 in the third row.
divide that row by 4. Rows 1
ƒ Add R3 to R2 and replace R2
and 2 do not change.
with that sum
ƒ Add R3 to R1 and replace R1
with the sum.
⎛1 1 0 0⎞
Row 3 will not be changed. ⎜ ⎟
All that remains to obtain reduced ⎜0 1 0 − 1⎟
row echelon form is to eliminate
⎜0 0 1 2 ⎟⎠
the 1 in the first row, 2nd position.
⎝
15 16
 Example
(continued)
Final Result

To get a zero in the first row

⎛1 1 0 0⎞
We can now “read”
read our ⎛1 0 0 1⎞
and second position, we ⎜ ⎟ solution from this last matrix. ⎜ ⎟
multiply row 2 by –1 and add
the result to row 1 and
⎜0 1 0 − 1⎟ We have
⎜0 1 0 − 1⎟
⎜0 0 1 2 ⎟⎠ ⎜0 0 1 2 ⎟⎠
x = 1,
replace row 1 by that result. ⎝ y = –1
Rows 2 and 3 remain z = 2. ⎝
unaffected.
⎛1 0 0 1⎞ Written as an ordered triple
triple,
⎜ ⎟ we have (1, –1, 2). This is a
⎜0 1 0 − 1⎟ consistent system with a
⎜0 0 1 2 ⎟⎠
unique solution.
⎝
17 18

Example 2 Example 2

Example: Solve the system Example: Solve the system

3x – 4y + 4z = 7 3x – 4y + 4z = 7
x – y – 2z = 2 x – y – 2z = 2
2x – 3y + 6z = 5 2x – 3y + 6z = 5
Solution:
Begin by representing the
system as an augmented matrix:
⎛ 3 −4 4 7 ⎞
⎜ ⎟
⎜ 1 − 1 −2 2 ⎟
⎜ 2 −3 6 5 ⎟
⎝ ⎠
19 20
 Example 2 Example 2
(continued) (continued)

ƒ Since the first number in ⎛ 3 −4 4 7 ⎞ In this step,

step we will get ⎛ 1 −1 −2 2⎞
the second row is a 1, we ⎜ ⎟ zeros in the entries beneath ⎜ ⎟
interchange rows 1 and 2 ⎜ 1 −1 −2 2⎟ the 1 in the first column: ⎜ 3 −4 4 7⎟
and leave row 3 ⎜ 2 −3 6 5 ⎟ ƒ Multiply row 1 by –3 , add ⎜ 5 ⎟⎠
unchanged: ⎝ ⎠ to row 2 and replace row 2: ⎝ 2 −3 6
–3R1 + R2 → R2.
⎛1 −1 −2 2⎞ ƒ Multiply row 1 by –2, 2 add ⎛ 1 −1 −2 2⎞
⎜ ⎟ to row 3 and replace row 3: ⎜ ⎟
⎜3 −4 4 7⎟ ⎜ 0 −1 10 1 ⎟
–2R1+R3 → R3.
⎜ 5 ⎟⎠ ⎜ ⎟
⎝2 −3 6 ⎝ 0 −1 10 1 ⎠

21 22

Representation of a Solution of a
Final Result
Dependent System

⎛ 1 −1 −2 2⎞
Too get a zero
e o in the
t e third
t d ⎛ 1 −1 −2 2⎞ ⎜ ⎟ Next we can express
p the variable x
row, second entry we ⎜ ⎟ ⎜ 0 −1 10 1⎟ in terms of t as follows: From the
multiply row 2 by –1 and add ⎜ 0 −1 10 1 ⎟ ⎜0 0 0
⎝ 0 ⎟⎠ first row of the matrix, we have x –
the result to R3 and replace ⎜ ⎟ y -2z = 2.
R3 by that sum: Notice this ⎝ 0 −1 10 1 ⎠ ƒ We can interpret the
second row of this matrix If z = t and y = 10t – 1, we have
operations “wipes out” row 3 x – (10t – 1) – 2t = 2 or x = 12t + 1.
so row 3 consists entirely of ⎛ 1 −1 −2 2⎞ as
–y + 10z = 1, or Our general solution can now be
zeros. ⎜ ⎟
⎜ 0 −1 10 1⎟ 10z – 1 = y expressed in terms of t:
Any time you have fewer
ƒ So, if we let z = t (12t + 1,10t – 1, t),
non-zero rows than variables ⎜0 0 0 0 ⎟⎠
you will have a dependent ⎝ (arbitrary real number,) where t is an arbitrary real number.
system. then in terms of t,
y = 10t – 1.
23 24
 Procedure for
Applications
Gauss-Jordan Elimination
Step 1. Choose the leftmost nonzero column and use appropriate
Systems of linear equations provide an excellent opportunity to
row operations to get a 1 at the top. discuss
di mathematical
th ti l modeling.
d li The
Th process off using
i mathematics
th ti
Step 2. Use multiples of the row containing the 1 from step 1 to to solve real-world problems can be broken down into three steps:
get zeros in all remaining places in the column containing this 1.
Step 1. Construct a mathematical model whose solution will
Step 3. Repeat step 1 with the submatrix formed by (mentally)
provide information about the real-world problem.
deleting the row used in step and all rows above this row.
Step 4. Repeat step 2 with the entire matrix, including the rows Real-world
problem
d l t d mentally.
deleted t ll Continue
C ti this
thi process until
til the
th entire
ti matrix
t i is
i 3. Interpret 1. Construct
in reduced form.
Note: If at any point in this process we obtain a row with all Mathematical Mathematical
zeros to the left of the vertical line and a nonzero number to the Solution 2. Solve Model
right, we can stop because we will have a contradiction.
25 26

Applications Solution
(continued)

Step 2. Solve the mathematical model. Step 1. The question in this example indicates that the relevant
variables are the number of each type of truck.
truck
Step 3. Interpret the solution to the mathematical model in
terms of the original real-world problem. x = number of 10-foot trucks

Example: Purchasing. A company that rents small moving y = number of 14-foot trucks
trucks wants to purchase 25 trucks with a combined capacity z = number of 24-foot trucks
of 28,000 cubic feet. Three different types of trucks are
We form the mathematical model:
available:
il bl a 10-foot
10 f t truck
t k withith a capacity
it off 350 cubic
bi feet,
f t a
14-foot truck with a capacity of 700 cubic feet, and a 24-foot x + y + z = 25 (Total number of trucks)
truck with a capacity of 1,400 cubic feet. How many of each
350x + 700y + 1,400z = 28,000 (Total capacity)
type of truck should the company purchase?

27 28
 Solution Solution
(continued) (continued)

Step 2. Now we form the augmented coefficient matrix of Let z = t. Then for t any real number
th system
the t andd solve
l by
b using
i Gauss-Jordan
G J d elimination:
li i ti
x = 2t – 30
⎡ 1 1 1 25 ⎤
⎢350 700 1, 400 28, 000 ⎥ (1/350)R2 R2 y = –3t + 55
⎣ ⎦ z=t
⎡1 1 1 25⎤ –R1 + R2 R2 is a solution to our mathematical model.
⎢1 2 4 80 ⎥
⎣ ⎦ Step 3. We must interpret this solution in terms of the original
⎡1 1 1 25⎤ –R2 + R1 R1 problem. Since the variables x, y, and z represent numbers of trucks,
⎢0 1 3 55 ⎥ they must be nonnegative. And since we can’t purchase a fractional
⎣ ⎦
Matrix is in reduced form. number of trucks, each must be a nonnegative whole number.
⎡1 0 −2 −30 ⎤ x – 2z = –30 or x = 2z – 30,
⎢ 0 1 3 55 ⎥
⎣ ⎦ y + 3z = 55 or y = –3z + 55.
29 30

Solution
(continued)
Since t = z, it follows that t must also
be a nonnegative
g whole number. The 10-ft 14-ft. 24-ft
first and second equations in the truck truck truck
model place additional restrictions on t x y z
the values t can assume:
15 0 10 15
x = 2t – 30 > 0 implies that t > 15
y = –3t + 55 > 0 implies that t < 55/3 16 2 7 16
Thus the only possible values of t that 17 4 4 17
will produce meaningful solutions to
the original problem are 15, 16, 17, 18 6 1 18
and 18. A table is a convenient way to
display these solutions.
31