Homework 3

Numerical Analysis (CMPS/MATH 305)

Hani Mehrpouyan

This homework is due by Thursday May 2nd 2013

Q UESTION 1 (5 POINTS )

Let the initial interval used in the bisection method have length b − a = 3. Find the number of midpoints cn that

must be calculated with the bisection method to obtain an approximate root within an error tolerance of 10−9 .

Q UESTION 2 (5 POINTS )

Let α be the smallest positive root of

f (x) = l − x + sin x = 0

Find an interval [a, b] containing α and for which the bisection method will converge to α. Then estimate the

number of iterates needed to find α within an accuracy of 5 × 10−8 .

 1

Q UESTION 3 (5 POINTS )

Imagine finding a root α satisfying 1 < α < 2. If you are using a binary computer with m binary digits in its

significand, what is the smallest error tolerance that makes sense in finding an approximation to α? If the original

interval [a, b] = [1, 2], how many interval halvings are needed to find an approximation c, to α with the maximum

accuracy possible for this computer?

Q UESTION 4 (5 POINTS )
√
Give Newton’s method for finding m
a with a > 0 and m a positive integer.

Hint: Solve xm − a = 0.

Q UESTION 5 (5 POINTS )

Consider applying Newton’s method to find the root α = 0 of sin(x) = 0. Find an interval [−r, r] for which the

Newton iterates will converge to α, for any choice of x0 in [−r, r]. Make r as large as possible.
 2

Hint: Draw a graph of y = sin(x) and graphically interpret the placement of x0 and x1 .

Q UESTION 6 (5 POINTS )

Write

−f 00 (ξn )
α − xn+l = (α − xn )(α − xn−l )
2f 0 (ζn )

as

α − xn+1 ≈ M (α − xn )(α − xn−1 )

where M is defined in the lectures. Then multiply both sides by M , obtaining

|M (a − xn+l ) ≈ |M (a − xn )||M (a − xn−l )|

Let Bn = |M (a − xn )|, n ≥ 0. To have xn converge to α, we must have Bn converge to 0. The above formula

yields

Bn+1 ≈ Bn Bn−1

For simplicity, assume B0 = B1 = γ .

a) Compute approximate values of B2 , B3 , B4 , B5 , B6 , B7 , in terms of γ .

b) If we write Bn = γ qn , n ≥ 0, give a formula for qn+l in terms of qn−1 and qn . What are q0 and q1 ?

c) Experimentally, confirm that

1 1
qn ≈ √ rn+1 = √ (1.618)n+1
5 5

for larger values of n, say, n ≥ 4. The number r was used in the lectures. The numbers {qn } are called Fibonacci

numbers. The number r is called the golden mean (some authors define the golden mean to be r − 1).
 3

d) Using (c), show that

Bn+1
Bnr

is approximately constant. Find the constant. This result can be used to construct a proof of

|α − xn+1 |  f 00 (α) r−1

 
lim = 0 ≡c
n→∞ |α − xn |r 2f (α) 

Q UESTION 7(5 POINTS )

Show that for any constants c and d, |d| < 1, the equation x = c + d cos(x) ≡ g(x) has a unique solution α. In

addition, show that the iteration xn+1 = c + d cos(xn ) will converge to α. Bound the rate of convergence.
 4

Q UESTION 8 (5 POINTS )

Convert the equation x2 − 5 = 0 to the fixed-point problem x = x + c(x2 − 5) ≡ g(x) with c a nonzero constant.
√
Determine the possible values of c to ensure convergence of xn+1 = xn + c(x2n − 5) to a = 5.

Q UESTION 9 (5 POINTS )
√ √
What are the solutions α , if any, of the equation x = 1 + x? Does the iteration xn+1 = xn + 1 converge to

any of these solutions (assuming x0 is chosen sufficiently close to α)?

 5

Q UESTION 10 (5 POINTS )

Does the following iteration converge to the indicated α, provided x0 is chosen sufficiently close to α? If it does

converge, determine the convergence order.

15x2n − 24xn + 13
xn+1 = , α = 1.
4xn

Q UESTION 11 (5 POINTS )

Consider the problem of solving x/(1 + x) − 0.99 = 0, calling its root α. Then, let α() be the solution of

x/(1 + x) − 0.99 +  = 0.

a) Using

g(α(0))
α() ≈ α(0) − 
f 0 (α(0))

, estimate α() − α.

b) Calculate α() directly, compute α() − α, and compare with (a). Comment on your results.
6