LECTURE 5
Solution of Nonlinear Equations
1. Introduction
In the next series of lectures we shall discuss the problem of identifying the roots of equations and systems
of equations..
The basic formulation of the problem in the simplest case is this: given a function f (x) of one variable, find
all x such that f (x) = 0.
All the algorithms we shall present are based on the following geometric idea. Suppose a continuous
function f (x) has a zero on the interval [a, b]. Then its graph must look something like
Well, not exactly like this, but we do know that the graphs has to cross the x-axis somewhere between the
points x = a and x = b. A graphical solution to the problem would then just amount to identifying at
which point(s) the graph of f (x) crossses the x-axis.
Unfortunately, this graphical method is not really practical for finding highly accurate solutions. The
reason for this is that once you start focussing in on the behavior of f (x) when f (x) is extremely close to
zero, is that floating point errors become more and more significant so that near a zero of f (x) a point plot
may look rather erratic.
1
2. BISECTION ALGORITHM 2
2. Bisection Algorithm
This method for finding roots is based on the following fundamental theorem from Calculus.
Theorem 5.1. (Intermediate Value Theorem). If f is a continuous function on the interval [a, b], and
then if ξ is a number between f (a) and f (b), then there exists a point c ∈ [a, b] such that f (c) = ξ .
This theorem is more than an mathematical abstraction of the graphical method mentioned above. Cer-
tainly, it says that if 0 is a number between f (a) and f (b), then there has to be a point c between a and b
such that f (c) = 0. But it also furnishes us with an algorithm for finding c.
Suppose
f (a) < 0 < f (b).
Then let c be the point half-way between a and b
a+b
c=
2
If we evaluate f (c) then either
• f (c) = 0, in which case we have found a zero of f
• f (c) < 0. In this case, the Intermediate Value Theorem tells us that there must be a zero between c
and b.
• f (c) > 0. In this case, the Intermediate Value Theorem tells us that there must be a zero between
a and c.
If the first case holds, then we are finished. If the second case holds, then we can set a1 = c and b1 = b and
a 1 + b1
c1 =
2
and fix the position of the zero to lie either at c1 , or between a1 and c1 , or between c1 and b1 . Similarly, if
the third case holds, we can fix the position of the zero between either c and (b + c) /2, or between (b + c)/2
and c.
Example 5.2. Starting with a = 0 and b = 2, use the bisection algorithm described above to find a zero of
f (x) = x2 − 2
• Now
f (a ) = f (0) = −2
f (b) = f (2) = 2
So we must have a zero of f somewhere between x = 0 and x = 2. If we evaluate f at the midpoint
c = 1 between 0 and 2, we find
f (1) = −1 < 0
So the zero must lie between x = 1 and x = 2. The midpoint of (1, 2) is 1.5, and there
f (1.5) = 0.25 > 0
Therefore the zero has to line within the interval (1, 1.5). Calculating f at the midpoint of that
interval yields
f (1.25) = −0.4375 < 0
so the zero must lie within (1.25, 1.50). Calculating f at the midpoint of that interval yields
f (1.375) = −0.109375 < 0
3. CODING THE BISECTION ALGORITHM 3
So now we know the zero lies within (1.375, 1.50). Iterating once more we find
1.375 + 1.50
f
2 = f (1.4375) = 0.066406250
so the zero lies within (1.375, 1.4375). √
Of course, the correct answer to this example is 2 = 1.41421356 . . . .
3. Coding the Bisection Algorithm
Let us now figure out how to tell a computer to carry out the bisection algorithm.
We’ll begin by first working out the logic of the algorithm.
1. We start with a function f and two numbers a and b such that f (a) < 0 < f (b).
2. We calculate the midpoint between a and b
a+b
c=
2
and
f (c)
3. We now have three possibilitites
(a) f (c) = 0. In this case we terminate the program since we’ve found a zeron.
(b) f (c) < 0. In this case, the zero mus lie between c and b, and so we loop back to step 2, with
the original values of a and b now being replaced, respectively, by c and b.
(c) f (c) > 0. In this case, the zero mus lie between a and c, and so we loop back to step 2, with
the original values of a and b now being replaced, respectively, by a and c.
In Maple, the problem in the example above would code as follows:
f := x -> x^2 -2;
a := 0.0;
b := 2.0;
c := (a+b)/2;
while (f(c) <> 0) do
if (f(c) < 0) then
a := c;
else
b := c;
fi;
od;
c;
Now if you try to run this code you’ll see find that Maple has gotten stuck in an infinite loop. The reason
for this, is that the condition we looking for, f (c) = 0, is never actually going to be met. In order to make
this algorithm palatable to Maple we are going to have to tell Maple when our answer is good enough.
Now there’s actually three natural ways to control the loop.
First, we can instruct Maple to cease calculating after a certain number of iterations, say 1000. To implement
this idea we would change the line
while (f(c) <>0) do
3. CODING THE BISECTION ALGORITHM 4
to
for i to 1000 do
Secondly, we can allow Maple to stop once f (c) is within a certain distance of 0, say within 0.00001. This
can be implemented by inserting a statement like
if |f(c)| < 0.00001 then break;
within the do loop
Finally, we can allow Maple to stop once c is determined to lie within a small enough interval. This can
be implemented via a statement like
if a-b
| | < 0.00001 then break;
Thus, a fully functioning Maple routine might be
f := x -> x^2 -2;
a := 0.0;
b := 2.0;
c := (a+b)/2;
for i from 1 to 1000 do
if (f(c) < 0) then
a := c;
else
b := c:
fi;
c := (a+b)/2:
if |f(c)| < 0.00001 then break fi:
if |a-b| < 0.00001 then break fi:
od:
c;
