Stuart Dalziel, DAMTP
Numerical Methods
Natural Sciences Tripos 1B
Lent Term 1999
Problem Sheet 1
1. ROOT FINDING
1.1 Roots of a cubic
Consider the solution to
f
(
x
)
=
0.5
where
f
(
x
) =
x
3
. Choosing initial guesses of
x
a
= 0
and
x
b
= 1
,
(a)
Write down an expression to show how the error
ε
n
in the bisection method decreases
with subsequent iterations.
(b)
Using the bisection method, determine the solution to four decimal places. Does the
number of iterations this took agree with the predicted number?
(c)
Repeat this calculation using the Linear interpolation and secant methods. How do the
results compare after each iteration?
(d)
Repeat the calculations using the NewtonRaphson method, starting from (i)
x
0
=
1
, (ii)
x
0
=
0
and (iii)
x
0
=
3
. Comment on your results.
1.2 Direct iteration
(a)
Illustrate graphically the direct iteration method and show cases of monotonic
convergence, oscillatory convergence and divergence.
(b)
Derive the convergence criteria for the direct iteration method for any arbitrary function.
(c)
Using only the four basic operations (
+–*/
), write down a scheme which will solve
f
(
x
)
=
0.5
, where
f
(
x
)=sin
x
. Starting with
x
=90degrees=
π
/2 radians
, use this scheme to
compute the solution to six decimal places. Compare the convergence of your
calculation with your analysis of the convergence.
1.3 Direct iteration
Consider the roots of
f
(
x
)
=
0
for some continuous function
f
(
x
)
.
(a)
By rerranging and adding iteration labels (
i.e.
change
x
into
x
n
or
x
n
+1
), derive a direct
iteration equation in the form
x
n
+1
=
g
(
x
n
)
from
(
x

x
)
h
(
x
)
+
f
(
x
)
=
0
.
(b)
Prove that the quadratic convergence can be obtained with a particular choice of
h
(
x
)
and explain how this is related to the NewtonRaphson method.
2. LINEAR EQUATIONS
2.1 Gaussian elimination
(a)
Use Gaussian elimination to solve
1
2
3
2
3
4
3
4
5
6
9
12
1
2
3
=
x
x
x
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
NST1B Numerical Methods
Lent
1999
Stuart Dalziel, DAMTP
(b)
Explain what is meant by pivoting. Why are these techniques used? Outline the
differences between partial pivoting and full pivoting.
(c)
Evaluate the solution of
1
2
3
1
2
4
2
2
5
6
7
9
1
2
3
=
x
x
x
using both partial and full pivoting. Why is pivoting necessary?
2.2 Banded matrices
Prove that if a matrix contains only zeroes outside some band centred around the leading
diagonal, then performing Gaussian elimination (without pivoting) will maintain these zeros.
2.3 Tridiagonal matrices
Develop an algorithm for solving a tridiagonal system of equations.
2.4 Over determined systems*
(a)
Explain what is meant by the term “over determined system”.
This is the end of the preview.
Sign up
to
access the rest of the document.
 Spring '07
 Gallivan
 Numerical Analysis, Stuart Dalziel, NST1B Numerical Methods

Click to edit the document details