0% found this document useful (0 votes)
135 views6 pages

Assignment #4: Solutions

This document provides the solutions to 5 problems related to numerical integration. Problem 1 uses the trapezoid rule to approximate an integral and provides an error bound. Problem 2 uses Simpson's rule for another integral approximation with error analysis. Problem 3 determines the parameters needed for composite trapezoid and Simpson's rules to approximate an integral to within 10^-4. Problem 4 solves for the coefficients of a 3-point quadrature formula with precision 2. Problem 5 considers a 2-point formula and determines its highest precision and coefficients.

Uploaded by

sagrv
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
135 views6 pages

Assignment #4: Solutions

This document provides the solutions to 5 problems related to numerical integration. Problem 1 uses the trapezoid rule to approximate an integral and provides an error bound. Problem 2 uses Simpson's rule for another integral approximation with error analysis. Problem 3 determines the parameters needed for composite trapezoid and Simpson's rules to approximate an integral to within 10^-4. Problem 4 solves for the coefficients of a 3-point quadrature formula with precision 2. Problem 5 considers a 2-point formula and determines its highest precision and coefficients.

Uploaded by

sagrv
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

MATH 3650 Numerical Analysis

Assignment #4
Written solutions are due Oct. 20

SOLUTIONS
Z 1.5
1. Approximate the integral x2 ln x dx using the (non-composite) trapezoidal rule. Give a
1
rigorous error bound on this approximation.

With f (x) = x2 ln x and h = (1.5 1) = 0.5 the Trapezoid Rule gives


Z 1.5
h
x2 ln x dx [f (1) + f (1.5)]
1 2
0.5 2
= [1 ln 1 + 1.52 ln 1.5] 0.228
2

The absolute error is


h3 00
|E| = |f ()|
12
for some [1, 1.5]. Note that
|f 00 (x)| = 3 + 2 ln x
is increasing x > 1 so
|f 00 ()| |f 00 (1.5)| = 3 + 2 ln 1.5.
Thus
0.53 00 0.53
|E| = |f ()| (3 + 2 ln 1.5) 0.040 .
12 12
We can conclude:
Z 1.5
x2 ln x dx = 0.228 0.040
1

Z 0.5
2
2. Approximate the integral dx using the (non-composite) Simpsons rule. Give a
0 x4
rigorous error bound on this approximation.

With f (x) = 2/(x 4) and h = 0.5/2 = 0.25 Simpsons Rule gives


Z 0.5
h
f (x) dx [f (0) + 4f (0.25) + f (0.5)]
0 3
 
0.25 2 2 2
= +4 + 0.26706349
3 04 0.25 4 0.5 4

The absolute error is


h5 (4)
|E| = |f ()|
90
for some [0, 0.5]. Note that
48
|f (4) (x)| =
|x 4|5
is increasing on [0, 0.5] so
48
|f (4) ()| |f (4) (0.5)| = 48/|0.5 4|5 = .
(3.5)5
Thus
0.255 (4) 0.255 48
|E| = |f ()| 106 .
90 90 (3.5)5
We can conclude:
Z 0.5
2
dx = 0.267063 0.000001
0 x4

Z 2
3. Determine the values of n and h required to approximate e2x sin 3x dx to within 104
0
using the
(a) composite trapezoid rule.

The absolute error is


(b a) 2 00 2
|E| = h |f ()| = h2 |f 00 ()|
12 12
for some [0, 2]. Note that
|f 00 (x)| = |12e2x cos 3x 5e2x sin 3x|
|12e2x cos 3x| + |5e2x sin 3x| (triangle inequality)
2x 2x
= 12e | cos 3x| + 5e | sin 3x|
12e2x + 5e2x = 17e2x
which is increasing on [0, 2] so
|f 00 ()| 17e22 = 17e4 .
Thus
1 1
|E| = h2 |f 00 ()| h2 17e4 104
6 6
1/2
(6)(104 )

= h 8.04 104
17e4

2 2
= n = 2487.5
h 8.04 104

= h 8.04 104 , n 2488


(b) composite Simpsons rule.

The absolute error is


(b a) 4 (4) 2 4 (4)
|E| = h |f ()| = h |f ()|
180 180
for some [0, 2]. Note that

|f (4) (x)| = | 119e2x sin 3x 120e2x cos 3x|


|119e2x sin 3x| + |120e2x cos 3x| (triangle inequality)
= 119e2x | sin 3x| + 120e2x | cos 3x|
119e2x + 120e2x = 239e2x

which is increasing on [0, 2] so

|f (4) ()| 239e22 = 239e4 .

Thus
1 4 (4) 1
|E| = h |f ()| h4 239e4 104
90 90

1/4
(90)(104 )

= h 0.0288
239e4

2 2
= n = 69.4
h 0.0288

= h 0.0288, n 70

4. Suppose the quadrature formula


Z 1
f (x) dx = c0 f (1) + c1 f (0) + c2 f (1)
1

is exact whenever f (x) is a polynomial of degree less than or equal to 2 (i.e. the formula has
degree of precision 2). Determine c0 , c1 and c2 .

For f (x) = A + Bx + Cx2 we require:


Z 1
2
(A + Bx + Cx2 ) dx = 2A + C (exact value)
1 3
= c0 (A B + C) + c1 A + c2 (A + B + C) (quadrature rule)
To make this equation an identity (i.e., true for all values of A, B, C) we match coefficients
of A, B, C to get 3 (linear) equations in 3 unknowns:

A : 2 = c0 + c1 + c2
B : 0 = c0 + c2
C : 23 = c0 + c2

Solving this system (e.g. using linear algebra methods, or Wolfram Alpha, or. . . ) gives

1 4 1
c0 = , c1 = , c2 =
3 3 3

Notice: this reproduces Simpsons Rule! So the degree of precision is actually 3,


not 2.

5. Consider the quadrature formula


Z 1
f (x) dx = c0 f (0) + c1 f (x1 ).
0

(a) What is the largest integer n such that this formula has degree of precision n?

With 3 free parameters in the quadrature rule, we expect it can give the exact answer
only for a 3-parameter family of functions (e.g. the family of degree-2 polynomials).

(b) Determine c0 , c1 and x1 in this case.

For f (x) = A + Bx + Cx2 we require:


Z 1
1 1
(A + Bx + Cx2 ) dx = A + B + C (exact value)
0 2 3
= c0 A + c1 (A + Bx1 + Cx21 ) (quadrature rule)

To make this equation an identity, we match coefficients of A, B, C to get 3 (nonlinear)


equations in 3 unknowns:
A : 1 = c0 + c1
B : 12 = c1 x1
C : 13 = c1 x21

The last 2 equations give


1 2 1 3
= 2x1 = 3x21 = x1 (2 3x1 ) = 0 = x1 = = c1 = = .
c1 3 2x1 4
(the other root x1 = 0 is inconsistent with both equations). Subsituting this into the
first equation gives
3 1 2
c1 = , c0 = , x1 =
4 4 3
Note that if we apply this quadrature rule to f (x) = x3 we get

3 2 3 2
  Z 1
1 3 2 1
4 f (0) + 4 f ( 3 ) = 0 + 4 = 6= f (x) dx = .
3 9 0 4
Thus the degree of precision is indeed 2 (not 3).
Z 2
(c) Use your result to approximate the integral x ln x dx. Determine the absolute error
1
by evaluating the integral exactly.

We need to make a substitution to shift the interval:


Z 2 Z 1
u = x 1 = x ln x dx = (u + 1) ln(u + 1) du
1 0 | {z }
f (u)
1 3 2

4 f (0) + 4 f 3
1 3 2
= 4 (1) ln(1) + 4(3 + 1) ln( 23 + 1) 0.6385

The exact value is Z 2


3
x ln x dx = 2 ln 2 0.6363
1 4
so the absolute error is

|E| = |0.6385 0.6363| 0.0022

6. Determine constants a, b, c and d so that the quadrature formula


Z 1
f (x) dx = af (1) + bf (1) + cf 0 (1) + df 0 (1)
1

has degree of precision 3.

For f (x) = A + Bx + Cx2 + Dx3 we require:


Z 1
2
(A + Bx + Cx2 + Dx3 ) dx = 2A + C (exact value)
1 3
= a(A B + C D) + b(A + B + C + D)
+ c(B 2C + 3D) + d(B + 2C + 3D) (quadrature rule)

To make this equation an identity, we match coefficients of A, B, C, D to get 4 (linear)


equations in 4 unknowns:
A: 2=a+b

B : 0 = a + b + c + d


C : 23 = a + b 2c + 2d
D : 0 = a + b + 3c + 3d

Solving this linear system gives

1 1
a = 1, b = 1, c= , d=
3 3

You might also like