MATH2033 (2023–2024)

Introduction to Scientific Computation

Nonlinear Equations II

Department of Mathematical Sciences

Nonlinear Equations II 2/17

Fixed-Point Iteration
A fixed point of a function g is a point p for which g(p) = p.
The fixed points of a real-valued function g of a single real variable are
the values of x at which the curve y = g(x) and the line y = x intersect.
Suppose that a ∈ R, that b ∈ R and that a < b.
If f is a function that is real-valued on [a, b] then the root-finding
problem of finding p ∈ [a, b] such that f (p) = 0 is equivalent to the
fixed-point problem of finding p ∈ [a, b] such that g(p) = p with g given
by g(x) = x − f (x).
If g is a function that is real-valued on [a, b] then the fixed-point problem
of finding p ∈ [a, b] such that g(p) = p is equivalent to the root-finding
problem of finding p ∈ [a, b] such that f (p) = 0 with f given by
f (x) = x − g(x).
Nonlinear Equations II 3/17

Theorem
Suppose that a ∈ R, that b ∈ R and that a < b. Suppose that g is a
function that is real-valued and continuous on [a, b] and that g(x) ∈ [a, b]
for all x ∈ [a, b]. Then g has at least one fixed point in [a, b].

Proof If g(a) = a then a is a fixed point of g.

If g(b) = b then b is a fixed point of g.
We now consider the case when g(a) ≠ a and g(b) ≠ b in which case
g(a) > a and g(b) < b. Let the function h be defined by h(x) = g(x) − x.
Now, h is real-valued and continuous on [a, b]. Moreover,
h(a) = g(a) − a > 0 and h(b) = g(b) − b < 0 and so h(b) < 0 < h(a).
Consequently, by the Intermediate Value Theorem there exists at least
one p ∈ (a, b) for which h(p) = 0. This implies that there exists at least
one p ∈ (a, b) for which g(p) − p = 0, or equivalently, g(p) = p.
Therefore, by combining the conclusions reached for the three cases
considered, we can conclude that g has at least one fixed point in
[a, b].
Nonlinear Equations II 4/17

Theorem (The Mean 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 differentiable on (α, β) then
there exists c ∈ (α, β) for which

= g ′ (c).
g(β) − g(α)
β−α
Nonlinear Equations II 5/17

Theorem
Suppose that a ∈ R, that b ∈ R and that a < b. Suppose that g is a
function that is real-valued and continuous on [a, b] and that g(x) ∈ [a, b]
for all x ∈ [a, b]. Also suppose that g is differentiable on (a, b) and that
there exists a positive constant k < 1 which is such that

∣g ′ (x)∣ ≤ k for all x ∈ (a, b).

Then g has at exactly one fixed point in [a, b].

Proof We first note that, by the theorem on page 3, g has at least one
fixed point in [a, b].
Nonlinear Equations II 6/17

Suppose that p and q are distinct fixed points of g in [a, b] with p < q.
Then
∣q − p∣ = ∣g(q) − g(p)∣ = ∣g ′ (c)(q − p)∣
for some c ∈ (p, q), as by the Mean Value Theorem there exists c ∈ (p, q)
for which
= g ′ (c)
g(q) − g(p)
q−p
because g is real-valued and continuous on [p, q] and differentiable on
(p, q) since g is real-valued and continuous on [a, b] and differentiable on
(a, b). Since ∣g ′ (x)∣ ≤ k < 1 for all x ∈ (a, b), we arrive at

∣q − p∣ = ∣g ′ (c)∣∣q − p∣ < ∣q − p∣

which is a contradiction. We can hence conclude that g has exactly one

fixed point in [a, b].
Nonlinear Equations II 7/17

Problem 1
Let g be given by g(x) = 13 (x2 − 1). Does g have exactly one fixed point
in [−1, 1]?

We first note that g is real-valued and continuous on [−1, 1]. Moreover,

g ′ (x) = 23 x and so g is differentiable on (−1, 1) and g ′ (x) = 0 ⇔ x = 0.
Now, g(0) = − 13 , g(−1) = 0 and g(1) = 0. Consequently, g(x) ∈ [− 13 , 0]
for all x ∈ [−1, 1]. Therefore, since [− 13 , 0] ⊂ [−1, 1], g(x) ∈ [−1, 1] for all
x ∈ [−1, 1]. Moreover, ∣g ′ (x)∣ ≤ 23 < 1 for all x ∈ (−1, 1). Hence, by the
theorem on page 5, g has exactly one fixed point in [−1, 1].
Nonlinear Equations II 8/17

Problem 2
Let g be given by g(x) = e−3x . Does g have at least one fixed point in
[0, 1]?

We first note that g is real-valued and continuous on [0, 1]. Moreover,

g ′ (x) = −3e−3x and so g is differentiable on (0, 1) and there are no x for
which g ′ (x) = 0. Now, g(0) = 1 and g(1) = e−3 . Consequently,
g(x) ∈ [e−3 , 1] for all x ∈ [0, 1]. Therefore, since [e−3 , 1] ⊂ [0, 1],
g(x) ∈ [0, 1] for all x ∈ [0, 1] and hence, by the theorem on page 3, g has
at least one fixed point in [0, 1].
Now, ∣g ′ ( 13 )∣ = ∣ − 3e−1 ∣ = 3e ≥ 1 and hence we cannot use the theorem on
page 5 to determine whether or not g has exactly one fixed point in [0, 1].
Nonlinear Equations II 9/17

Fixed-Point Iteration
Given an appropriate function g we can use fixed-point iteration to find
approximations to a solution p of g(p) = p.
Starting from an initial approximation p0 , the approximation obtained
after n iterations of fixed-point iteration is

pn = g(pn−1 ).
Nonlinear Equations II 10/17

Problem 3
Let g be given by g(x) = 1 − 12 x2 . Use fixed-point iteration with an initial
approximation of 1 to compute approximations p1 , p2 , p3 , p4 , p5 , p6 , p7
and p8 to the fixed point of g that lies in the interval [0, 1], where pn is
the approximation obtained after performing n iterations of fixed-point
iteration.
Starting with p0 = 1 and using pn = g(pn−1 ) = 1 − 12 p2n−1 yields:
n pn
1 0.5
2 0.875
3 0.6171875
4 0.8095397 . . .
5 0.6723226 . . .
6 0.7739911 . . .
7 0.7004688 . . .
8 0.7546716 . . .
Nonlinear Equations II 11/17

Problem 4
Let g be given by g(x) = 2x − (1 − 12 x2 ). Use fixed-point iteration with an
initial approximation of 1 to compute approximations p1 , p2 , p3 and p4 to
the fixed point of g that lies in the interval [0, 1], where pn is the
approximation obtained after performing n iterations of fixed-point
iteration.
Starting with p0 = 1 and using pn = g(pn−1 ) = 2pn−1 − (1 − 12 p2n−1 ) yields:

n pn
1 1.5
2 3.125
3 10.1328125
4 70.6025695 . . .
Nonlinear Equations II 12/17

Theorem (The Fixed-Point Theorem)

Suppose that a ∈ R, that b ∈ R and that a < b. Suppose that g is a
function that is real-valued and continuous on [a, b] and that g(x) ∈ [a, b]
for all x ∈ [a, b]. Also suppose that g is differentiable on (a, b) and that
there exists a positive constant k < 1 which is such that

∣g ′ (x)∣ ≤ k for all x ∈ (a, b).

For all positive integers n, let

pn = g(pn−1 ).

Then, for any p0 ∈ [a, b], the sequence {pn }∞

n=0 converges to the unique
fixed point of g in [a, b].

Proof We first note that, by the theorem on page 5, there exists a

unique fixed point of g in [a, b]. Let p denote this fixed point.
Nonlinear Equations II 13/17

Suppose that p0 ∈ [a, b]. Then, since g(x) ∈ [a, b] for all x ∈ [a, b], we
have that pn ∈ [a, b] for all nonnegative integers n. For all positive
integers m we have that

∣p − pm ∣ = ∣g(p) − g(pm−1 )∣ = ∣g ′ (ξm )(p − pm−1 )∣

for some ξm ∈ (a, b) by the Mean Value Theorem and hence,

∣p − pm ∣ = ∣g ′ (ξm )∣∣p − pm−1 ∣ ≤ k∣p − pm−1 ∣

since ∣g ′ (x)∣ ≤ k for all x ∈ (a, b). Consequently, for all nonnegative
integers n,
∣p − pn ∣ ≤ k n ∣p − p0 ∣.
Now, since 0 < k < 1,
lim k n ∣p − p0 ∣ = 0.
n→∞

Therefore, the sequence {pn }∞

n=0 converges to p.
Nonlinear Equations II 14/17

The error bound for fixed-point iteration

Theorem
Suppose that a ∈ R, that b ∈ R and that a < b. Suppose that g is a
function that is real-valued and continuous on [a, b] and that g(x) ∈ [a, b]
for all x ∈ [a, b]. Also suppose that g is differentiable on (a, b) and that
there exists a positive constant k < 1 which is such that

∣g ′ (x)∣ ≤ k for all x ∈ (a, b).

Let p denote the unique fixed point of g in [a, b]. For all positive integers
n, let
pn = g(pn−1 )
where p0 ∈ [a, b]. Then, for all nonnegative integers n,

∣p − pn ∣ ≤ k n max{p0 − a, b − p0 }.
Nonlinear Equations II 15/17

Proof In the proof of the Fixed-Point Theorem we had shown that, for
all nonnegative integers n,

∣p − pn ∣ ≤ k n ∣p − p0 ∣

from which we can conclude that

∣p − pn ∣ ≤ k n max ∣x − p0 ∣
x∈[a,b]
n
= k max{∣a − p0 ∣, ∣b − p0 ∣}
= k n max{p0 − a, b − p0 }

since p ∈ [a, b] and p0 ∈ [a, b].

Nonlinear Equations II 16/17

Suppose that, for all positive integers n,

y(n) = ck n

where c and k are positive constants. Then, for all positive integers n,

log10 (y(n)) = log10 (ck n )

from which we can arrive at

log10 (y(n)) = log10 (k)n + log10 (c)

in which case the points (n, log10 (y(n))) lie on a straight line with slope
log10 (k).
Nonlinear Equations II 17/17

References
Burden, Faires & Burden, Numerical Analysis, 10E
Section 2.2