Vous êtes sur la page 1sur 5

7.

3 Solving Systems
of Equations
In Section 7.2 we learned how to solve a few systems of linear equations. Examples 2.20 and 2.24 were
especially illuminating. In this section we will find a
systematic way to solve systems of equations, a way that always works and that we
could easily program on a computer. Sophistication is not important here. We are
only interested in finding a way that always works. Furthermore, we want a method
that finds all solutions when there is more than one.
As we already indicated in Section 7.2, the method will use the augmented
matrix. We will eliminate coefficients to put the system into an equivalent form that
can be easily solved by the method of back-solving. Our first task will be to describe
the form of equations that can be solved easily in this way. We will then describe
in more detail how to go about the process of elimination. Finally, we will discover
some more useful facts about process of back-solving.
Row echelon form of a matrix — the goal of elimination
Look back at the augmented matrices of the systems which we solved by backsolving in Section 7.2.
These are in equations (2.12), (2.22), and (2.28). To describe how these matrices are different from the
others, we will introduce some
terminology.
DEFINITION 3.1 The pivot of a row vector in a matrix is the first nonzero element of that row.
To give us a specific example to talk about, consider the system
2x2 + 4x3 = 2
x1 - 2x3 = -1
-2x1 + 2x2 + 8x3 = 4.

(3.2)
This system can be written as Ax = b, where
A = ⎛ ⎝ -0 2 4 1 0 2 2 8 -2 ⎞ ⎠ , x = ⎛ ⎝ x x x1 2 3 ⎞ ⎠ , and b = ⎛ ⎝ -2 1 4 ⎞ ⎠ . (3.3)
The augmented matrix for the system is
M = [A, b] = ⎛ ⎝ -0 12 2 02 8 4 -4 2 2 -1 ⎞ ⎠ . (3.4)
In this matrix the pivots of each row are printed in blue.
7.3 Solving Systems of Equations 293
DEFINITION 3.5 A matrix is in row echelon form (or just echelon form) if, in each row that
contains a pivot, the pivot lies to the right of the pivot in the preceding row. Any
rows that contain only zeros must be at the bottom of the matrix.
The matrices in equations (2.12), (2.22), and (2.28) are in row echelon form.
Other matrices in Section 7.2 are not in row echelon form. The matrix M in (3.4) is
not in row echelon form, since, for example, the pivot in row three lies below and
not to the right of the pivot in row two. The matrix
⎛⎜⎜⎜⎜⎜⎜⎜⎝
P∗∗∗∗∗∗∗∗∗
0P∗∗∗∗∗∗∗∗
000P∗∗∗∗∗∗
000000P∗∗∗
0000000P∗∗
0000000000
0000000000
⎞⎟⎟⎟⎟⎟⎟⎟⎠
, (3.6)
where the Ps refer to pivots and must therefore be nonzero numbers and the asterisks refer to arbitrary
numbers, is in row echelon form. This matrix illustrates some
of the possibilities of row echelon form that we have not seen before. Matrices in
echelon form are set up for solution by the process of back-solving, so the goal of
elimination is to put the augmented matrix into row echelon form.
Row operations and elimination
In all of our examples to this point we have achieved row echelon form by the
operation of adding a multiple of one row to another. There are other permissible
operations. One of them is illustrated by the matrix M in (3.4). For a matrix that has
a nonzero entry in the first column to be in row echelon form, it is necessary that the
entry in the upper left-hand corner be a pivot—a nonzero number. The easiest way
to achieve this for M is to interchange rows 1 and 2 to obtain
⎛⎝
1 0 -2 -1
0242
-2 2 8 4
⎞⎠
. (3.7)
Notice that this is the augmented matrix for the system
x1 - 2x3 = -1
2x2 + 4x3 = 2
-2x1 + 2x2 + 8x3 = 4.

(3.8)
Thus the operation on the rows of M is equivalent to interchanging the first two
equations in the system in (3.2). This does not change the solutions, so it is a
permissible operation.
Similarly, if we multiply an equation by a nonzero number, the solutions are
not affected. When dealing with the augmented matrix this is accomplished by
multiplying a row by a nonzero number.
To summarize, we have three allowed operations on the rows of a matrix (called
row operations):
R1: Add a multiple of one row to a different row.
R2: Interchange two rows.
R3: Multiply a row by a nonzero number.
294 CHAPTER 7 Matrix Algebra
E x a m p l e 3 . 9 Find the solution set for the system in (3.2).
We have already used R2 to reduce the problem to finding the solution set for
the equivalent system in (3.8), with augmented matrix given in (3.7). Notice that
one effect of this use of R2 was to raise the pivot from the second row to the first.
This is a typical use of R2.
The next step is to use R1 to eliminate the first entry in the third row. We add 2
times the first row to the third, getting the system and augmented matrix
x1 - 2x3 = -1
2x2 + 4x3 = 2
2x2 + 4x3 = 2
⎛⎝
1 0 -2 -1
0242
0242
⎞⎠
. (3.10)
By comparing the matrices in (3.4) and (3.10), we see that one of the effects of
eliminating the coefficient in the third row is to move the pivot in the third row
to the right. Notice that the last two equations are the same. We could stop here,
but to go all the way to row echelon form we should eliminate the last row of the
augmented matrix. We do this by adding -1 times the second row to the third to get
x1 - 2x3 = -1
2x2 + 4x3 = 2
0=0
⎛⎝
1 0 -2 -1
0242
0000
⎞⎠
. (3.11)
Notice that the third row in the new augmented matrix has a common factor of 2.
We could use R3 to remove this factor, but that is optional.
Because it is in row echelon form, we can back-solve the system in (3.11). The
last equation is always true, so we start with the second, 2x2 + 4x3 = 2. We can
assign an arbitrary value to one of the variables x2 and x3. We choose x3 and set
x3 = t to indicate that. Then, using the second equation, we solve for
x2 = (2 - 4x3)/2 = (2 - 4t)/2 = 1 - 2t.
Finally, we solve the first equation for x1 = -1 + 2x3 = -1 + 2t. The general
solution is
x=
⎛⎝
x1
x2
x3
⎞⎠
=
⎛⎝
-1 + 2t
1 - 2t
t
⎞⎠
=
⎛⎝
-1
10
⎞⎠
+ t ⎛ ⎝ -2 2 1 ⎞ ⎠ .
Looking back at Definition 2.6, we see that this is a parametric equation for a line
in R3. ●
Each of the row operations is reversible. Subtracting the same multiple reverses
R1. Interchanging the same two rows reverses the effect of R2. Multiplying by the
reciprocal of the nonzero number reverses R3. This reflects the fact that the row
operations replace one system of equations with another that has the same solution
set.
Implementing row operations, especially for large matrices, involves a lot of
arithmetic. It is very easy to make frustrating mistakes. This is just the type of
thing that computers are good at. There are a number of mathematical computer
programs, such as MATLABR , Maple, and Mathematica, in which it is easy to
manipulate matrices and to perform the row operations. After you have learned the
basics using hand calculations, we encourage you to use one of these to solve larger
systems to help you understand the fine points of how the solution method we have
described works.
7.3 Solving Systems of Equations 295
THEOREM 3.12 With operations R1, R2, and R3, any matrix can be transformed into row echelon
form. ■
Any attempt at a proof of this result will only confuse the issue. Working
through a few examples is more illuminating and also suggests how a proof might
be constructed.
For the application we have in mind, the original matrix will be the augmented
matrix associated with a system of linear equations. The key fact is that we can use
row operations to transform the matrix into a matrix in row echelon form, and this
new matrix is the augmented matrix associated with a system of equations that has
the same solution set as the original system. The advantage of the row echelon form
is that this system can be easily solved by back-solving.
Let’s look at a more complicated example.
Find the general solution to the system 2x5 = 1
x2 + 3x3 + 2x4 +
x1 + 2x2 + 3x3 + 5x4 + 7x5 = 8 (3.14)

E x a m p l e 3 . 1 3 2x1 + 4x2 + 6x3 + 9x4 + 15x5 = 2.


First we write this as a matrix equation
⎛⎝
01322
12357
2 4 6 9 15
⎞⎠
⎛⎜⎜⎜⎝
x1
x2
x3
x4
x5
⎞⎟⎟⎟⎠
=
⎛⎝
182
⎞⎠
.
The augmented matrix is
⎛⎝
013221
123578
2 4 6 9 15 2
⎞⎠
.
Again the pivots are printed in blue. The next step is to use row operations to put this
into row echelon form. We need a pivot in the upper left hand corner. To achieve
this we

Vous aimerez peut-être aussi