Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

Constructive Mathematics.

QS 3

1. Suppose f ∈ C 4 in a interval containing the root, α and that Newton’s method gives a
sequence of iterates xk , k = 0, 1, 2, . . . which converge to α. Show that Newton’s method
is at least quadratically convergent to α if f 0 (α) 6= 0.
If f 0 (α) = 0, then by using l’Hôpital’s Rule or otherwise, show that Newton’s method is
linearly convergent in both of the cases

(i)f 00 (α) 6= 0 (ii)f 00 (α) = 0, f 000 (α) 6= 0.

What is
|xk+1 − α|
lim
k→∞ |xk − α|

in each of these cases?


What can you say about the order of convergence of Newton’s method when f 0 (α) 6=
0, f 00 (α) = 0?

2. Deduce from the above that Newton’s Method has linear convergence to a double root.
Further show that in the case that the multiplicity of a particular root is known to be
m ∈ N (⇒ f (x) = (x − α)m h(x), h(α) 6= 0) then the modified Newton’s method

f (xk )
xk+1 = xk − m
f 0 (xk )

is at least quadratically convergent.

3. (M) Apply Newton’s method with x0 = 12 to find the root α = 0 of the following functions
(a) f (x) = sin x sinh x
(b) f (x) = x cos x
(c) f (x) = x − sin x
and relate the order of convergence observed in each case to question 1 above. (Note that
it is here easier to observe what |xk − α|/|xk−1 − α|p is, at least for p = 1 and possibly
p = 2, since α = 0.)
For (c) make the trivial modification to your matlab code to apply the appropriate
modified Newton’s method; observe the order of convergence.

4. Optional, but essentially easy


Division of two real numbers on a computer is much more difficult than addition, sub-
traction and even multiplication, so often Newton’s method is used even to divide 2 real
numbers. This question is about reciprocation: given a ∈ R − {0}, find 1/a. Clearly
reciprocation followed by multiplication gives the result of division.
For given a ∈ R − {0}, show that application of Newton’s method to f (x) = a − 1/x
yields an iterative formula which only requires addition, subtraction and multiplication.

1 Constructive Mathematics, TT 2018


5. Harder: strictly optional (Süli & Mayers, problem 1.10)
By replacing (approximating) f 0 (xk ) by the slope of the secant between two previous
points, one derives the secant method
 
xk − xk−1
xk+1 = xk − f (xk )
f (xk ) − f (xk−1 )

for finding roots of f .


Show that the secant iteration can be written in the form
xk f (xk−1 ) − xk−1 f (xk )
xk+1 = .
f (xk−1 ) − f (xk )

Supposing that f has a continuous second derivative in a neighbourhood of the root α


of f and that f 0 (α) > 0, f 00 (α) > 0, define
xk+1 − α
φ(xk , xk−1 ) = ,
(xk − α)(xk−1 − α)

where xk+1 has been expressed in terms of xk and xk−1 . Find an expression for

ψ(xk−1 ) = lim φ(xk , xk−1 ),


xk →α

and the determine limxk−1 →α ψ(xk−1 ). Deduce that

f 00 (α)
lim φ(xk , xk−1 ) = .
xk ,xk−1 →α 2f 0 (α)

Now assume that


|xk+1 − α|
lim = A.
k→∞ |xk − α|q

Show that q − 1 − 1/q = 0, and hence that q = 12 (1 + 5). Deduce finally that
q/(1+q)
f 00 (α)

|xk+1 − α|
lim =
k→∞ |xk − α|q 2f 0 (α)

so that secant iteration has order of convergence 12 (1 + 5) ≈ 1.618.
Note that two starting guesses, x0 and x1 , are required for the secant method.

6. Save this question for Prelims revision: For a given real number c > 0, find the order of
convergence of
c2
 
1 5c
xk+1 = 5xk + 2 − 5 = G(xk )
9 xk xk
to the real cube root of c.
Determine the other fixed point, d of x = G(x), and show that for no interval a < d <
b can convergence of the fixed point iteration to d be guaranteed by the Contraction
Mapping Theorem.

2 Constructive Mathematics, TT 2018

You might also like