MATH2033 (2023–2024)

Introduction to Scientific Computation

Nonlinear Equations I

Department of Mathematical Sciences

Nonlinear Equations I 2/17

The Bisection Method

Suppose that a ∈ R, that b ∈ R and that a < b.
Suppose that f is a function that is real-valued on [a, b].
We will consider the problem of finding p ∈ [a, b] such that f (p) = 0.
A solution p to f (p) = 0 is called a zero of f .
Note that for functions g1 and g2 that are real-valued on [a, b], the
problem of finding x such that g1 (x) = g2 (x) is equivalent to the problem
of finding x such that g3 (x) = 0 with g3 = g1 − g2 .
Also note that f (a)f (b) < 0 is equivalent to f (a) and f (b) having
opposite signs which is equivalent to

f (a) > 0 and f (b) < 0

or
f (a) < 0 and f (b) > 0.
Nonlinear Equations I 3/17

Theorem (The Intermediate Value Theorem)

Suppose that α ∈ R, that β ∈ R and that α < β. If g is a function that is
real-valued and continuous on [α, β] and K is any number between g(α)
and g(β) then there exists at least one c ∈ (α, β) for which g(c) = K.
Nonlinear Equations I 4/17

Theorem
Suppose that α ∈ R, that β ∈ R and that α < β. Suppose that f is a
function that is real-valued and continuous on [α, β] and that
f (α)f (β) < 0. Then there exists at least one p ∈ (α, β) for which
f (p) = 0.

Proof Since f is real-valued on [α, β] and f (α)f (β) < 0, either

f (α) < 0 < f (β)

or
f (β) < 0 < f (α).
Consequently, since f is also continuous on [α, β], by the Intermediate
Value Theorem there exists at least one p ∈ (α, β) for which f (p) = 0.
Nonlinear Equations I 5/17

Given a function f which is real-valued and continuous on [a1 , b1 ] and

such that f (a1 )f (b1 ) < 0 where a1 ∈ R, b1 ∈ R and a1 < b1 , we can use
the bisection method to find approximations to a zero of f that lies in the
interval [a1 , b1 ].
Starting from the initial interval [a1 , b1 ], the approximation obtained after
n iterations of the bisection method is
1
pn = (an + bn ).
2
If f (pn ) = 0 then pn is a zero of f and so we let an+1 = pn and
bn+1 = pn .
If f (an )f (pn ) < 0 then a zero of f lies in the interval (an , pn ) and
so we let an+1 = an and bn+1 = pn .
If f (pn )f (bn ) < 0 then a zero of f lies in the interval (pn , bn ) and so
we let an+1 = pn and bn+1 = bn .
1
Note that in the last two cases bn+1 − an+1 = (bn − an ).
2
Also note that, for all positive integers n, the an and bn generated by this
method are such that a zero of f lies in the interval [an , bn ].
Nonlinear Equations I 6/17

Problem 1
Use the bisection method to find the approximations p1 , p2 and p3 to the
zero of f that lies in the interval [1, 2] for f given by
f (x) = x3 + x2 − 2x − 2, where pn is the approximation obtained after
performing n iterations of the bisection method, starting from the initial
interval [1, 2].

We first compute that f (1) = −2 < 0 and f (2) = 6 > 0. We present the
results in a table:
n an bn pn f (pn )
1 1 2 1.5 >0
2 1 1.5 1.25 < 0
3 1.25 1.5 1.375
Nonlinear Equations I 7/17

After how many iterations should we stop?

Examples of stopping criteria are to stop once

∣f (pn )∣ < ,

∣pn+1 − pn ∣ < 
or
∣pn+1 − pn ∣
<
∣pn+1 ∣
for some tolerance .
We could also stop once a quantity that provides a bound on the error is
less than some tolerance.
Nonlinear Equations I 8/17

The error bound for the bisection method

Theorem
Suppose that a ∈ R, that b ∈ R and that a < b. Suppose that f is a
function that is real-valued and continuous on [a, b] and that
f (a)f (b) < 0. Let pn be the approximation to a zero of f obtained after
performing n iterations of the bisection method, starting from the initial
interval [a, b]. Then, for all positive integers n,

b−a
∣p − pn ∣ ≤
2n
for at least one p ∈ [a, b] satisfying f (p) = 0.

Proof We will use the notation introduced on page 5. Suppose that n is

a positive integer.
b−a
If f (pn ) = 0 then ∣p − pn ∣ = 0 ≤ n where p = pn .
2
Nonlinear Equations I 9/17

Let p be a zero of f that lies in the interval [an , bn ]. If f (pn ) ≠ 0 then

∣p − pn ∣ ≤ max ∣x − pn ∣
x∈[an ,bn ]
= max {∣an − pn ∣, ∣bn − pn ∣}
1 1
= max {∣an − (an + bn )∣ , ∣bn − (an + bn )∣}
2 2
1
= ∣ (bn − an )∣
2
1
= (bn − an )
2
1b−a
=
2 2n−1
1
since bi − ai = (bi−1 − ai−1 ) for all integers i such that 2 ≤ i ≤ n with
2
a1 = a and b1 = b.
Nonlinear Equations I 10/17

We can hence conclude that, for all positive integers n,

b−a
∣p − pn ∣ ≤
2n
for at least one p ∈ [a, b] satisfying f (p) = 0.
Nonlinear Equations I 11/17

Corollary
Suppose that a ∈ R, that b ∈ R and that a < b. Suppose that f is a
function that is real-valued and continuous on [a, b] and that
f (a)f (b) < 0. Let pn be the approximation to a zero of f obtained after
performing n iterations of the bisection method, starting from the initial
interval [a, b]. Then the sequence {pn }∞
n=1 converges to a zero p ∈ [a, b]
of f .
Nonlinear Equations I 12/17

Proof Given any  > 0, there exists a positive integer N () such that

b−a
<  for all integers n ≥ N ().
2n
Consequently, since
b−a
∣p − pn ∣ ≤ for all integers n ≥ 1,
2n
given any  > 0, there exists a positive integer N () such that

∣p − pn ∣ <  for all integers n ≥ N ()

and hence the corollary holds.

Nonlinear Equations I 13/17

Alternatively, we can conclude that the corollary holds because

b−a
∣p − pn ∣ ≤ for all integers n ≥ 1,
2n
and
b−a
lim = 0.
n→∞ 2n
Nonlinear Equations I 14/17

How to guarantee a given accuracy?

Problem 2
Let f be given by f (x) = x3 + x2 − 2x − 2. Use the error bound for the
bisection method to determine a number of iterations that will guarantee
that ∣p − pn ∣ ≤ 10−4 , where p is the zero of f that lies in [1, 2] and pn is
the approximation to p obtained after performing n iterations of the
bisection method, starting from the initial interval [1, 2].

We first compute that f (1) = −2 < 0 and f (2) = 6 > 0. Since 2 − 1 = 1,

1
∣p − pn ∣ ≤ n = 2−n . We want ∣p − pn ∣ ≤ 10−4 . This will be satisfied when
2
2−n ≤ 10−4 .
Nonlinear Equations I 15/17

Now,

2−n ≤ 10−4
⇔ log10 (2−n ) ≤ −4
⇔ −n log10 (2) ≤ −4
4
⇔ n≥ .
log10 (2)
4
Moreover, = 13.287 . . . and so ∣p − pn ∣ ≤ 10−4 if at least 14
log10 (2)
iterations are performed.
Nonlinear Equations I 16/17

Problem 3
Perform iterations of the bisection method, starting from the initial
interval [−1, 1], to find approximations to the zero of f that lies in the
interval [−1, 1] for f given by f (x) = (x − 0.1)(x − 2). Stop performing
iterations once the error bound for the bisection method guarantees that
the absolute error in the latest approximation is no greater than 0.1.

We first compute that f (−1) > 0 and f (1) < 0. We present the results in
a table where pn is the approximation obtained after performing n
iterations of the bisection method:
b1 − a1
n an bn pn f (pn ) ∣0.1 − pn ∣
2n
1 −1 1 0 >0 0.1 1
2 0 1 0.5 <0 0.4 0.5
3 0 0.5 0.25 <0 0.15 0.25
4 0 0.25 0.125 <0 0.025 0.125
5 0 0.125 0.0625 >0 0.0375 0.0625
Nonlinear Equations I 17/17

References
Burden, Faires & Burden, Numerical Analysis, 10E
Section 2.1