0% found this document useful (0 votes)
144 views41 pagesSolution - Manual For Numerical Analysis - VIEWER XDOCS
solution_manual for numerical analysis
Uploaded by
Hoàng Tố UyênCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
144 views41 pagesSolution - Manual For Numerical Analysis - VIEWER XDOCS
solution_manual for numerical analysis
Uploaded by
Hoàng Tố UyênCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 41
STUDENT STUDY GUIDE
FOR
Numerical Analysis
Eighth Edition
Richard L. Burden
J. Douglas Faires
YOUNGSTOWN STATE UNIVERSITY
COPYRIGHT © 2005 Richard L. Burden and J. Douglas Faires Contents
Preface
Mathematical Preliminaries
Exercise Set LL
Exercise Set 2
Exercise Set 1.3
Solutions of Equations of One Variable
Exercise Set 2.1
Exercise Set 22
Exercise Set 23,
Exercise Set 24
Exercise Set 2.5
Exercise Set 2.6
Interpolation and Polynomfal Approximation
Exercise Set3.1... . ‘
Exercise Set 3.2
Exercise Set 33
Exercise Set 34
Exercise Set 3.5
‘Numerical Differentiation and Integration,
Exercise Set 4.1
Exercise Set 4.2
Exercise Set 4.3
Exercise Set 44
Exercise Set 4.5
Exercise Set 4.6
Exercise Set 4.7
Exercise Set 48
Exercise Set 4.9
Initial-Value Problems for Ordinary Differential Equations
Exercise Set 5.1
Exercise Set 5.2
Exercise Set 5.3 ii CONTENTS
Exercise Set 4 50
Exercise Set 8.5 93
Exercise Set 5.6 95
Exercise Set 5.7 ake 25 99
ee vente . . 100
Exercise Set 59 . 102
Exercise Set 5.10 105
‘Direct Methods for Solving Linear Systems um
Exercise Set6.1 oo. see ccs eeeeeeeseeeee . . 1
Exercise Set 62 116
Exercise Set 6.3 120
Exercise Set64.. 2. . cee ee cece 125
Exercise Set 6.1 128
Exercise Set 6.6 . 134
Iterative Techniques in Matrix Algebra oy
Exercise Set 7.1 1
Exercise Set7.2.. 2.00.05 cece eens veces wees 2
Exercise Set 7.3 144
Exercise Set 7.4 146
Exercise Set7.5 147
Approximation Theory 158
Exercise Set 8.1 . 135
Exercise Set 82 137
Exercise Set8.30. 2. 000s ccs eveee esse eens . : 158
Exercise Set 84 : 159
Exercise Set 8.5 . 161
Exercise Set 8.6 let
Approximating Eigenvalues 167
Exercise Set 9.1 167
Exercise Set 9.2 169
Exercise Set 9.3 m
Exercise Set9.4 173,
‘Numerical Solutions of Nonlinear Systems of Equations 7
Exercise Set 10.1 seaeerteer son eaes 7
Exercise Set 10.2 179
Exercise Set 103... ee 2ay 182
Exercise Set 10.4 18s
Exercise Set 10.5, 188 CONTENTS
Boundary-Value Problems for Ord
Exercise Set 11.1
Exercise Set 11.2
Exercise Set 11.3,
Exercise Set 114.
Exercise Set 11.5
‘Numerical Solutions to Partial Differential Equations
Exercise Set 121... ¢
Exercise Set 122.
Exercise Set 123°
Exercise Set 12.4
ww
191
193
195
198
199
205
2u1
213 Preface Preface
‘This Stadeat Study Guide for Nnmerical Analysis, Eighth Edition, by Burden and Faires contains worked
‘out representative exercises forthe all the techniques discussed in the book. Although the answers tothe odd
‘exercises are also in the back of the text, the results listed in this Study Guide go well beyond those in the
‘book. The exercises that are solved in the Guide were chosen tobe those requiring insight into the methods
sliscussed in the text.
‘We have also added a number of exercises tothe text that involve the use of a computer algcbra system,
‘We chose Maple as our standard, bat any ofthese systems can be used. In our recent teaching ofthe course
‘we found that students understood the concepts beter when they worked through the algorithms step-by-step,
‘at let the computer algebra system do the tedious computation.
thas been our practice to include in our Numerical Analysis book structured algorithms of all the tech
niques discussed in the text. The algorithms are given ina form that can be coded in any appropriate pro-
gramming language, by students with even a minimal amount of programming expertise.
In the Fifth Edition of Numerical Analysis we included in the Student Study Guide a disk containing
FORTRAN and Pascal programs forthe algorithms inthe book. Ia the Sixth edition we placed the disk in the
text itself, and added C programs, as well as worksheets in Maple and Mathematica, forall the algorithms.
‘We continued ths practice for the Seventh Edition, have updated the added Maple programs to both versions
5.0 and 6.0, and added MATLAB programs as well.
For the Eighth Edition, we have added new Maple programs to reflect the linear algebra package change
‘fom the original Lina package to the more modem LinearAlgebra package. In addition, we now
also have the programs coded in Java.
‘You will not find a disk with this edition of the book. Instead, our reviewers suggested, and we agree,
‘that it is more useful to have the programs available for downloading from the web. At the website for the
book,
Ltp/Avww.as.ysu.edu/~faires/Numerical-Analysis/
‘you will find the programs. This site also contains additional information about the book and will be updated.
regularly o reflect any modifications that might be made. For example, we will place there any responses to
{questions from users ofthe book concerning interpretations of the exercises and appropriate applications of
the techniques, vi Preface
‘We hope our supplement package provides flexibility for instructors teaching Numerical Analysis. If
‘you have aay suggestions for impeovements that can be incorporated into future editions of the book or the
supplements, we would be most grateful to receive your comments. We can be most easily contacted by
‘electronic mal atthe addzesses listed below.
‘Youngstown State University Richard L. Burden
[email protected]
December 27, 2004 J.Douglas Faires
[email protected] Mathematical Preliminaries
Exercise Set 1.1, page 14
4. d. Show that the equation 2 ~ (In:r)* = 0 has a least one solution inthe iaterval [4,5]
SOLUTION: It is not possible to algebraically solve for the solution , but this is not required in the
problem, we must only show that a solution exists. Let
Sle) = 2— (nz) = x —exp(e(In(lnz))).
Since fis continuous on [4 5] with f(4) ~ 0.3006 and f(5) = —5.790, the Intermediate Value
Theorem implies that a number x must exist in (4,5) with O= f(x) =~ (In-z)*
2. ¢ Find intervals thet contain a solution tothe equation x* — 2s? — 4r-+3=0.
SOLUTION: Let f(x) = * 22% — 4 +3. The critical points of f occur when
= s'() =429 —4r—4= (Br + )(2-2);
that is, when r= ~3 and x = 2. Relative maxiomm and minimum valves of f ean occur only at
these Values, There are a most thee solutions to f(x) = 0, sine f(z) isa polynomial of degree
three. Since f(—2) = —6 and f (3) a 4.48; #(0) = 3 and f(1) = —2: and f(2) = —5 and
J(4) = 19; solotions lie in the intervals [-2, —2/3]. 0, 1), and [24
4. a Find maxocect [f(#)| when f(x) = (2 + 2x) /3.
SOLUTION: First note that /"(x) = (—e* + 2) /3 50 the only critical point of f occurs at r — In 2,
‘hich lies inthe interval (0, i], The maximum for |f(2}| must consequentiy be
‘max(|f(0)],|f(in2)|, |F(1)]} = max{1/3, (21n2)/3, (4 — 0)/3} = (21n2)/3.
9. Find the second Taylor polynomial for f() = e eos x about ro = 0.
SOLUTION: Since
S(2)=A(cosz—sine), f"(x)=—2e(sin-z), and f"(x) = —2*(sinz + e082),
we have f(0) = 1, /"(0) = 1. and f"(0) = 0. S0
2c
Po2)= 1+ and Role) == e086) pa
at Exercise Set 1.1
4-UseP(05) to apron (05), find an wpe bound for (05) — P05). an compare this
to the actual error,
SOLUTION: We have P(0.5) = 140.5 = 1.5 and
= Peking + 0086) (9 5
(1105) POS) < amas, =
i
< 1015)? ma. lef(cing +con
5 3009)" mar. le (eing + con8))
‘To maximize this quantity on [00.5], first note that Dge"(sin x + cost) = 2e* cos:r > 0, forall rin
[00.5] This implies that che maximum and minimum values of "(sin + eos:z) on [0,0.5] occur at
the endpoints of the interval, and
(sin 0 + e080) =
. Find a bound for the error | f(r) ~ Pa(x)}, for in (0, 1)
SOLUTION: A similar analysis to that in part (a) gives
[f(e) ~ Pa] < 2 (.0)%e4(in 1+ oos 1) = 1.282
¢. Approximate Jp f(r) de-using Jp Pa(s) de.
SOLUTION: :
“ 4 apis
ff terden [rvede= [ert] <5
4. Find an upper bound forthe enor ia pt (0
SOLUTION: From part),
[ecoiaes [Stout rsinttac~ [aie de ost
(cour bins) S(eost +sin1) ~ 5(1-+0) = 1.378,
the actual error is [1.378 — 1.5] ~ 0.12.
14, Use the error term of a Taylor polynomial to estimate the error involved in using sin x © 2 10
approximate sin 1°
SOLUTION: First we need to convert the degree measure forthe sine function to radians. We have
180° = rains, 30 1° = 45 radians. Since, f(z) — sin, f"(z) = cos, f"(2) = ~sin, and
J"2) = ~cosz, we have f(0) = 0, and f"(0) = 0. The approximation sine m2 is
aiven by f(x) = Py(z) and a(x) = —885r. Tf we use the bound [eos é| < 1, then
|=(iz)|
ae | 586 x 10-7
|=" si5 a y samen im . Use Maple to determine the third Maclaurin polynomial Ps(:r).
SOLUTION: Define f() by
(072) in (12
, (
xp (x¢/2) #sim (x/3) ;
‘Then find the first three tems of the Taylor series with
Extract the third Maclaurin polynomial with
>p3:=convert (g, polynem) +
a, Bp
2+ i+ oat
b. £02) and bound the eror|f(«) — P,()] 00 (0,1).
SOLUTION: Determine the fourth derivative.
af isdi ft (£,2,,%,2) 7
= 19 601720 gin (1
Sam —yogpelt/? sin (52) +
Find the fifth derivative
ofsisdift (fax):
f= 8 an (42) +e (40)
a at) * a
See ifthe fourth derivative has any critical points in (0, 1.
feolve(f5=0,x,0..2);
>»
p= 0047389076
The extreme values ofthe fourth derivative will occur at x = 0,1, or p.
ol = 00787176213
pe? :=evalf (subs (x=0, £4)) ;
2 = o92so2s0259
3c3 :seval£ (subs (:
LEAD):
4 = oneTa3463
‘The maximum absolute value of f(r) is e, and the error is given by
serror:=c1/24;
‘error == .004077990080 4
Exercise Set 1.2
26. Suppose that f is continuous on fab) thet ry andr are in [a,b], and that ey and e» are postive
constants. Show that a number € exists between ry and. with
cafes) + caf (2a)
ate
HO
SOLUTION: Let
and m < f(2) 2000,
>, How many terms are required to ensure the 10° accuracy needed for an approximation 10 x?
SOLUTION: In this case, we need
4
2n+)-1
Clearly, a more rapidly convergent method is needed for this approximation.
< 10 or 4000 < 2n +1
< 10 on > 20,000,000,000.
3. a, How many calculations are needed to determine a sum of the form,
Ld aws
SOLUTION: For each i, the ianer sum 5
total of
‘aby requires i multiplications and i — 1 additions, for
n(n +1)
7) smuhiptications
and
BOD additions
2
Once the m inner sums are computed, » ~ 1 additions are required forthe final sun.
‘The final total is:
‘multiplications and fittest) additions
», Re-express the series in a way that will reduce the number of calculations needed to determine this
SOLUTION: By rewriting the sum as
n(n 1)
2
‘ve can significantly reduce the amount of calculation, For each é, we now need i — 1 additions to sum
y's fora total of
n(n+0)
Fn additions
(Once the b's are summed, we need n multiplications by the a's followed by n — 1 additions of the
products,
The total additions by this method is still }(n+ 2)(1— 1), bat the numberof multiplications has
been reduced from }n(n +1) ton. Mathematical Preliminaries
10.
Devise an algorithm to compute the ral ots of a quadratic equation in the most efficient manner,
SOLUTION: The following algorithm uses the most effective formula for computing the roots of a
quadratic equation.
INPUT A, B,C.
OUTPUT 23, 22
Step LIF A= 0 then
if B= 0 then OUTPUT (‘NO SOLUTIONS’);
STOR,
else sen = —C/B:
OUTPUT (‘ONE SOLUTION’. 7)):
STOP,
Step 2 Set D = B? — 440.
Step 3 If D = Othen set:r; = ~B/(2A);
‘OUTPUT (‘MULTIPLE ROOTS", 1):
STOP.
Step 41F D < O then set
b= V=D/(2A);
a= —B/2A),
OUTPUT (‘COMPLEX CONJUGATE ROOTS’);
Step 6 OUTPUT (1,22):
STOP.
Suppose that as r approaches zero,
Fa) = 1140) and F(x) = 12 +0 (24)
Let ey and ep be nonzero constants, and define
Fe)= aha) +eF(2) and Cle) = F(a) + Fler).
‘Show that ify = minimum {a, 8), then as 2 approaches 2er0, 10 Exercise Set 1.3
a. Fle) =erla + enka +(e)
bG(x) = Ty +1a+0(07)
SOLUTION: Suppose for suficiently small |r| we have positive constants ky and ky independent of
sr for which
Re) Lal 0, and
define rq = Fy41/ Fr. Assuming that lity.» ty = converges, show that the limit isthe golden
ratio: 2 — (1+ V8) /2.
SOLUTION: Since
Jim te= imeem aa 4 2,
weave
r=1+4, which implies that
‘The only positive solution to this quadratic equation is x = (1 + v5) /2.
17, The Fbonaei sequence also satisfies the equation
on-al()-(
a Write a Maple procedure to calculate Foo,
SOLUTION:
Fy u
n=08
fal
aot
>for i from 1 ton do
>as
t9:f:s:8:sLiod;
1 :=218022005834555160026
Jf =135301852344706746049
1 :=218922005834555160026
1 = 854224848179261915075
', Use Maple with the default value of Digits followed by eval to calculate Fo.
SOLUTION:
F100
( (ange (5) /2)7200-((1-agre (5) /2°200) fog (511
ym)
rom = 3 ((j +398) - (3-44) \ws
o.3542208538 x 102
evalf (F100) ;
. Why is the result from pat (a) more accurate than the result from part (b)? R
Exercise Set 1.3
SOLUTION: The result in part (a) is computed using exact integer arithmetic, and the result in part
(b) is computed using ten-digit rounding atthmetic.
44. Why isthe result from part (b) obtained more rapidly than the result from part (a)?
SOLUTION: The result in part (a) required traversing a loop 98 times.
What results when you use the command simp1ify instead of eval f to compute Foi
SOLUTION: The result is the same as the result in part (a, Solutions of Equations of One Variable
Exercise Set 2.1, page 51
4. Use the Biseetion method to find ps for f(2) = y@ — cox on (0,1).
SOLUTION: Using the Bisccton method gives ox = 0 and b
4(b)) = 0.45970, We have p and f(p1) = -0.
f(p1) < 0, we assign az = p; LL. Thus, f(a2)
75. Since f(p2)
so f(a1) = —1 and
‘DHS <0, Since f(a) < Oand
0.17048 < 0,
13434 > 0, we have as = 0.5:
2. alLet f(2) = 3(e+1) (#— $) (2-1). Use the Bisection method on the interval [~2, 1.5] find
SOLUTION: Since
wwe have the following sign graph for f(2)
eH “OFTHE EEE HEHEHE
a4 Otte ee
rot ereererees
Jo) ~-----o444steo-oreeesee tee
weave f(ps) > 0.
1.125 and
0.6875.
Thus, a
Wea:
Sle.) <
with f(a) < 0, and by = 1.5, with f(y) > 0. Since py = —
‘ay = ~2, with (a2) <0, and bp = —3, with f(b) > 0. Thus, p
Heace, we assign a 1.195 and by = 0.25. Then ps
AL Let f(z) = (£ +2)(2 + I)2(# ~ 1)°(e —2). To which zero of f does the Bisection method
converge forthe following intervals?
SOLUTION: Since
Sle) = (2+ Met Yele- M22),
B “
2.
4.
1.
Exercise Set2.1
\ve have the following sign graph for f()
xe? Motttteteteteeet eet eet te
e-P (voreeereres
eon - ~nobtteee
fo - ote tobe
+0999 —_¢—_} >
3 2 9 6 f & 3 8
a[-3,25]
SOLUTION: The interval {—5,2.5] contains all § zeros of f. For ay = —8, with f(a) <0, and
by =2.5, with /(by) > 0, we have py = (~3-+ 2.5)/2 = ~0.25,s0 f(py) <0. Thus we asian
on =0.25, with faa) <0, and bp 5, with f(b1) > 0. Hence,
pa = (025 +2.5)/2 = 1.125 and f(g) < 0. Then we assign ay — 1.125, with f(ag) <0, and
by =2.5, with f(by) > 0. Since [1.125,2.5] contains only the zero 2, the method couverges to 2.
(1.75, 1.5)
SOLUTION: The interval {—1.75, 1.5] contains the zeros —1, 0,1. For = ~1.78, with f(a) > 0,
and by = 1.5, with f(b) < 0, We have p; = (~1.75 + 1.5)/2 = —0.125 and fim) < 0. Then we
assign ag = ay = ~1.75, with f(a) > 0. and by =p, = -0.125, with f(b) <0. Since
[-1.75, 0.125] contains only the 2er0 ~1, the method eoaverges to —1.
Use the Bisection Algorithm to find an approximation to V3 that is accurate to within 10~*
SOLUTION: The function defined by f(2) = 2* ~ has V3 its only positive root. Appling the
Bisection method to this function onthe interval [1,2] sives V-~ mq = 1.7320. Using smaller
stating interval would decrease the aumber of tertions tht ar required
Use Theorem 2.1 to find @ bound for the number of iterations needed to approximate a solution to the
equation 2 +r — 4 =0 on the interval [1,4] an accuracy of 10%
SOLUTION: First note thatthe particular equation plays no part in finding the bound: all that is
ceded isthe interval and the accuracy requirement, To find an approximation that is accurate to
within 10-8, we need to determine the number of iterations so that
that is,
3x10 <2"
‘Asa consequence, a bound for the number of iterations is n > 12. Applying the Bisection Algocitum
gives pi2 = 1.3787.
Define the sequence {p,} bY Pn = Ep.
sequence {pn} diverges.
4. Show that img sc Pa — Pat
0, even though the Solutions of Equations of One Variable 1s
SOLUTION: Since a — Pat = 1/mn, We have itn u(y — Pn) = O- However. the nth
partial sum ofthe divergent harmonic series. The harmonic series isthe classic example ofa series
whose tems go o zero, but not rapidly enough to produce a convergent series. There ae many proofs
of divergence ofthis series, ny calculus text shoud give at east two. One proof wil simply analyze
‘he partial suns ofthe series an another based on the Tateatl Test. The pont ofthe problem not
the fact that his particular sequence diverges itis that a est for an approximate solution ta oot
based on the condition that [fn — pu i amall should always be suspect. Consecutive terms of 2
sequence may be close to each other, but not sufficiently close to the actual solution you are seeking.
19, A tough of water of length J. = 10 feet has cross section inthe shape ofa semicscle with radius
foot. When filled with water to within a distance h of te top, the volume V = 12.4 "of the
water is given by the formula
12
10 [o.sx— resin kh (1 a2)"
Determine the depth ofthe water to within 0.01 fet.
SOLUTION: Applying the Bisection Algorithm on the interval (0, 1] tthe function
10) = 124~10 on ~asexin—H(1=12)"7]
o.t61
0.895 fet.
ives hss pis = 0.1617, so the depth sr — 1
Exercise Set 2.2, page 61
3. The following three methods are proposed to compute 21"/3. Rank them in order, based on their
apparent speed of couvergence, assuming py
SOLUTION:
a. Since
we have 16
2.
Exercise Set 2.2
«Since
we have :
and
_ Set — 6822 — de + Btr9 — 204 + 255
- (ay
21° +24 + 8Ar5
(= 21)"
Tha, g/ (214/4) = 5.706 > 1
@.Since
Thus, ¢ (2109)
‘The order of convergence should be (b), (d), (a). Choice (c) does not converge.
Use fixed-point iteration method to detemnine an approximation to V3 tat i accurate to within
wrt
SOLUTION: As always with fixed-point iteration, the trick is to choose the fixed-point problem that
will produce rapid convergence.
‘Recalling the solution to Exercise 10 in Section 21, we need to convert the root-finding problem
f(z) = x? ~ 3 into a fixed-point problem. One successful solution is to write
0
then add to both ses ofthe Inter equation and divide by 2. This gives g(x) = 0.5 (x + 2), and for
o= 1.0, we have V8 = py = 1.73205,
«. Determine a fixed-point funetion g and an appropriate interval that produces an approximation to a
positive solution ofr? — e* = 0 that is accurate to within 10-*
SOLUTION: There are numerous poss
[ye 00 [0 1] with 7p
1032? on [3,4] with po Solutions of Equations of One Variable v7
18. Show that (a) Theorem 22 is tue if |g/(x)| < kis replaced by the statement “y’(x) 0.
SOLUTION: First let g(x) = 1/2 + 1/.For x £ 0, we have g(x) = 1/2—1/a2.1Ex > v3, then
1/22 <1/2,0 (2) > 0. Also, g (V2) = v3.
Suppose, as is the assumption given in part (a) that ro > V3. Thea
a1 — V3 = alo) ~ 9 (V3) = 9116) (co - v2),
where V2 < £ and x1 > V2 Further,
yt
2a
and V3 < 21 < sro. By an inductive argument, we have
V2 < 2mt1 < tm <1 < 20.
“Thus, (>) is a decreasing sequence that has a lower bound and must therefore converge. Suppose
= lity, no Zu Then
Pp A
7
Thus, 1
=P 4)
pageh
‘which implies that
a =p 42,
sop= 42. Since ty > V2 or all m,
Jim_ tm = V3. 18
m
Exercise Set 2.3,
‘Now cousider the situation when 0 < «ry < /2, which is the situation in part (b). Then we have
0< (29 v3)" =23-200V/2-+2,
Qrov2 < 2h +2 and Va< Be
To complete the problem, we consider the thee possiblies ey > 0
Case I: xp > 2, which by pat (amples that ite = V2.
Case 2: xo = V2, which implies that tm = V2 for all m and that limtm_ao tm = v2.
Case 3: 0 < zo < v2, which implies that v2 < 21 by part (b). Thus,
O< myc VE camp Ste Sony and lim tm = v3.
In any situation, the sequence converges to V3, and rapidly, as we will discover in the Seetion 23.
Suppose thatthe function g has a fixed-point at p, that g € Ca, band that g exists in (a,b). Show
that if |g'(p)| > 1, then the fixed-point sequence will fil 10 converge for any initial choice of po,
except if pm = p for some value of 7,
SOLUTION: Since gis continous st p and l'(p)| > I. by letting e = |/(p)| — 1 there exists a
saumber 6 > 0 such that
We)-s@l<
whenever 0 < |x —p| < 6, Since
lg’) — 1,
Wee)-9'@l>
for any « satisfying 0 < [x — pl < 6, we have
lov) lah,
Lal) > La) — I) — 9) > W@)|- I I -D
Ifpo is chosen so that 0 < |p — pol <6, we have by the Mean Value Theorem that
ter — Pl = la(eo) — a(p)| = lo’(E) Ilo — a,
for some £ between po and p. Thus, 0 < |p ~ | <5 and
ter —Pl = 198) |lp0 — P| > leo — PI.
‘This means that when an approximation gets close top, but is not equal top, the succeeding terms of
the sequence move away from p. Solutions of Evatioas of One Variable 1»
Exercise Set 2.3, page 71
1. Let fw
~ 6 and py = 1. Use Newton's method to find p>.
SOLUTION: Let f(z) 1. Then f(x) = 2x, and Newton's method becomes
With pp = 1, we have
m=
and
~ Se 7200714,
3. Let f(x) =x? ~6, With pp = 3 and py ~ 2, find py for (a) the Secant method and (b) the method of
False Position.
SOLUTION: The formula for both the Secant method and the method of False Position is
— £00n1)\Pu-1 = Pua)
Pee Pet" lpn) — Fn-2)
The Secant method
With »p = and pr = 2, we have f(7») = 9-6 =3 and f(p1) =4—6 = 2. The Secaut method
sives
= Side
me Fox) — Fee)
and f(p2) = 24? 6 = 024, Then we have
py LeaNra— Pr) _ gg (0224-2) _g, -1.006 _ 5 yg
= Fon)— Fin) 4 Coma) 4 ae
The method of Flee Position
With pp = 3 and p1 = 2, we have f(pp) = 3 and J(p:) =—2. As in the Secant method (part (a)),
p= 24 and f(p2) = ~0.24. Since f(y) < Oand f(p2) <0 the method of False Postion requires,
‘reassignment of py. Then p, is changed to py $0 that py = 3, with f(p1) = 3,and pp = 2.4, with
‘S{gn) = 0.24. We calculate ps by
— 5, Slealm—m) 44 (0224-3) _, 0.
PsP "Fm)— Fn) 4 on A ag
€ Since V6 2.44949, the approximation (a) is better. 20 Exercise Set 2.3,
5. ¢. Apply Newion’s method to find a solution to ~ cos:r = 0 inthe interval [0, 7/2] that is aceurate
towithia 10-4
SOLUTION: With f¢
‘Newton's method is
coor, We have f(x) = 1 sin, and the sequence generated by
com Pp
Tsing
For po = 0, we have p; = 1, pp = 0.75036, py = 0.73911, and the suficiently accurate
Pa = 0.73909.
«. Apply the Seeant method fo finda solution to x — cos.x = 0 in the interval {0,/2| that is accurate
to within 10-4
SOLUTION: The Secant method approximations are generated by the sequence
Pa = Po —
1
Pa = Pra
(oa = ©057m-1) (Pa — 008 ma)
Using the endpoints ofthe intervals as yy and py, we have the entries inthe following tables.
Forthe Secant method: For the method of False Position:
n nm
oo oo
1 L.sro7o6s 1 Lsvo796a
2 ost101ss 2 06110155
3 0.7282005, 3 0,7232605
4 0.7395671 4 07372659
5 0,739084 5 O7S8S77S
6 0.7300851 6 0.7390615
7 0.739085
9. c. Apply the method of False Postion to find a solution to: —eos.r = O in the interval (0, x /2] that is
accurate to within 1
SOLUTION: The method of False Postion approximations are generated using tis same formula as
in Exercise 7, but incorporate the additional bracketing test
13. Apply Newton's method to find a solution, accurate to within 10~4, to the value of «that produces
the closest point on the graph of y = 1” to the point (1,0)
SOLUTION: The distance between an arbitrary point (2) on the graph of y ~ c* andthe point
(1,0) is
d(x) = ye 1)? + (2? —0) = Vt Be
‘Because a derivative is needed to find the critical points of dit is easier to work with the square of
this function,
Sa) = (dle)? = 428 2244, Solutions of Equations of One Variable a
16.
19,
2.
‘whose minimum will occu tthe same value of asthe minimum of d(x). To minimize f() we
need: so that
O= f(x) = 49 422-2
Applying Newton's method fo find he root of this equation with po ~ 1 gives ps
point on the graph of y = 2? that is closest to (1, 0) has the approximate coordinates:
(o.ssorss,o-trstn)
580755. The
‘Use Newton's method to solve for roots of
SOLUTION: Newton's method with py ~ § gives pis ~ 1.895488 and with py
‘9 = 1.895489. With py = 10x, the sequence does not converge in 200 iterations.
Sx gives
‘The results do not indicate the fast convergence usually associated with Newton's method because the
function and its derivative have the same roots. As we approach a root, we are dividing by numbers
‘ith small magnitude, which increases the round-off error.
Explain why the iteration equation forthe Secaut method should not be used inthe algebraically
esuivaleat form
$(0n-1)Pe=2~ H(Pn-2)Pn—a
Tl0m-1)~ Sn-2)
SOLUTION: This formula incorporates the subtraction of nearly equal numbers in both the
‘numerator and denominator when py and py are uearly equal. The form given in the Secant
Algorithm subtracts a correction from a result that should dominate the calculations. This is always
the preferred approach.
‘Use Maple to determine how many iterations of Newtoa’s method with pp
root of f(z) cto within 10-10,
15/Aare needed to find
SOLUTION: We first define f(r) and f(x) with
>£:=x-2e08 (x) “x:
fen ost) —
and >Ep:=x->(D) (£) (x);
Ip 2+ —sin(z)—1
‘We wish to use 100-digit rounding arithmetic so we set
spigits po
efor n from 1 to 7 do
> pl:sevalf (pO-£ (0) /£p (po) ); 2 Exercise Set 2.3,
Py = -T398085133321'516064165531208767387340401341175890075746406
1568063577312846548835475945903761069317065319,
hich is accurate to 10-1
23, The function defined by f(r) = In (2? + 1) —e cow x has an infinite aumber of zeros.
Approximate, to within 10-*,
(2) the only negative 210,
(@) the four smallest positive zeros, and
(@ the 25th smallest positive zero,
SOLUTION: The key to this problem i recognizing the behavior of ¢°4*, When is negative, this
term goes to ero, so f(z) is dominated by In (x2 + 1). However, when xis positive, "dominates
the calculations, and /(c) willbe zero approximately when tis term makes no contibuton; that is,
when eos re = 0. This oceurs when 1 = 1/2 for a positive integer n. Using this information to
determine initial approximations produces the following results
For part (a), we can use pp = —0.5 to find the sufficiently accurate ps = ~0.4341431
For past (b), we can use:
0.510 give py ~ 04506567; pp — 1.510 give py — L.TAA7381;
Po = 25 to give ps = 2.238319; and po = 35 to give pe = 3.709042.
In general, a reasonable initial approximation forthe nth postive rot is ~ 0.5. To salve part (4),
wwe let py = 24.5 to produce the suiciently aecurate approximation pz
Graphs for vatious parts ofthe region are shown below.
600
400
200
26, Determine the minimal annual interest rte at which an amount P = $1500 per month can be
invested to accumulate an amount A = $750, (00 at the end of 20 years based on the annuity due
equation
Fiasar- Solutions of Equations of One Variable 2B
28.
Let Fle
SOLUTION: This is simply a rootfinding problem where the function is given by
1500. y(12)(20)
eon — 280 Fr sry —]
‘Notice that » and i have been adjusted because the payments are made monthly rather than yearly.
‘The approximate solution to this equation can be found by any method in this section. Newton's
method is a bit cumbersome for this problem. since the derivative of f is complicated. The Secant
teh won bes iel choice The ial sal nr appoint 67%
fi
Pras
A drug administered toa patient produces a concentration in the blood stream given by
e(t) = Ate~‘/* mg/mL, t hours after A units have been administered, The maximum safe
concentration is! mg/mL,
8) What amount should be injected to reach this safe level, and when does this cur?
bb) When should an additional amount be administered, ift is administered when the level drops to
025 mg/mL?
©, Assuming 75% ofthe original amount is administered in the second injestion, when should a third
injection be given”
SOLUTION: The maximum concentration occurs when
w=a(-Je
This happens when t = 3 hours, and since the concentration at this time will be e(3) = Site, we
need to administer A — Le unit
For part (b) ofthe problem, we need to determine ¢ so that
0
oas-a=(ejie
This occurs when is 11 hours and 5 minutes; that is, when t = 11.083 hows
The solution to part (c) requires finding ¢ so that
Ge) te! 40.75 () (611.083) e~(t-1.08578,
This occurs after 21 hours and 14 minutes,
0.25 = e(t
gett 7. 52
a. Use the Maple commands solve and feoLve 0 try to find al roots off.
SOLUTION: First define the funetion by
pf pKco3" (Bextl)-74957 (2030 5
f
a 90H) _ 7g
psolve (£ (x)=0,x) 5
(3/7)
“in @T/25) Exercise Set 2.3,
>fsolve(E (x)=0,x)
feolve(g(*e+!) — 7 5128)
0.)
‘The procedure solve gives the ext ston, nd fool ve fis bose he neptive ai ian
asymplote forthe gapiof (2)
Plot f(t ind intial approximations to oot off.
SOLUTION: Using the Maple command >plot ({£ (x) },x=10.5..11.5) ; produces the
following graph.
«. Use Newton's method to find roots off to within 10.
SOLUTION: Define (x) using
>£psax->(D) (£) (x);
Sp 2+ 83%") In(a) — 145 In(5)
Digits = 18
wall
>for i from 1 to 5 do
>pl:zevalf (po-£ (po) /£p(po)) :
perr:sabs(pl-p0) ;
>p0:=pli
30a;
‘The results are given in the following table
Dy be peal
11,0007380401552508 .0007380401552503
111,0094380350662827 0002001041889676
‘11.0004386442681488 2016978339 10-*
TLOoONBSEAADSITIG «2772 10-2
11, 0094386442681716 0 Solutions of Equations of One Variable 2s
4. Find the exact solutions of f(x) = 0 algebraically.
SOLUTION: We have 34"! = 7.52%, Taking the natural logarithm of both sides gives
(r+1)In3=n7-+2rIn5,
‘Thus,
eln3 — 2rln5 =In7 —in3,
1?
2(3ln3—2In5) In,
and
n7/3 In7/8__ 3/7
5 In 27/25. ‘In 27/25"
This agrees with par),
Exercise Set 2.4, page 82
1, a. Use Newton’s method to find a solution accurate to within 10~° for x? — 2re~* + e~™* = 0,
where Oo <1
SOLUTION: Since .
2 are pe
Ste)
and
Sa) = Br — De + Dee — 2e™,
the iteration formula is
S(Pn-a)
FOr.
Pa =Pa-
Pat
Boga — 2ePat + Dpy pe Pont — Det
With pp = 0.5,
i= 0.5 ~ (0.01134878)/(-0.3422805) = 0.5331965,
[567135 is accurate to-within 10
Continuing inthis manner, pry =
3. a, Repest Exercise I(a) using the modified Newton-Raphson method described in Eq. (2.11). Is there
‘an improvement in speed or accuracy ver Exercise 1?
SOLUTION: Since
See
L@
and 10.
- a. Show that the sequence yn — 1/n converges linearly to p
Exercise Set 24
the iteration forma is
my flea Ons)
Pe PI Fas) Ln)" Ons)
With py = 0.5, we have f(po) = 0.011348781, f"(po) = ~0.342280542, f"(pp) = 5.201100744
and
(0.01134878)(—0.512280542)
(COsiaN06 12)? — (O.ULLSASTSI)(5.201T09744)
=0.5680137.
05 —
Continuing inthis manner, py = 0.567143 is accurate to within 10-*
‘and determine the number of tems
sequited t have [py — n| <5 10-2
SOLUTION: Since
tig Pm —al yg Wt py
8. pa ple nT
we have linear convergence. To have py ~ pl < 5 x 10-%, we need 1/n < 0.05, which implies that
n>,
3. Show that (a) the sequence pm = 10?" converges quadratically to zero, but that (b) pa = LO-"*
does not converge to 2er0 quadratically, regardless of the size of k > 1.
SOLUTION:
a. Since
bas -0| wo oe
a) ER, Gone Te 0
en = OF
the sequence is quadratically convergent.
b, However, for any k > 1,
tome rome
Boney A goa = ag me
diverges. So the sequence p,, = 10-"* does not converge quadratically for any positive value of k
Show that the fixed-point method
mf(x)
To
az)
has o/(p) = 0, fp isa zero of f of maltiplicity
SOLUTION: If f has a zero of multiplicity r at p, then a function q exists with
S(@)=(2—p)"a(2), where lim af) #0. Solutions of Equations of One Variable cy
2.
since
Si(2) = mle — pale) + (4 — pI (2),
wehave
is) =e 40) n(x ~ p) ale) .
Fe) ~*~ mle pyr-tqla) + (2 PPT)
which reduces to
m(x~ pale)
ingle) +(e pate)"
Differentiating this expression and evaluating at x = p gives
ae) =2—
ma(n)(matp)]
[nato)
If f'" is continous, Theorem 2.8 implies that this sequence produces quadratic convergence once We
are lose enough tothe solution p.
ge) =
‘Suppose that f has m continuous derivatives. Show that f has a zero of multiplicity m at p if and
only if
O= 0) =F) == FM), but FW) #0.
‘SOLUTION: If f has a zero of multiplicity m at p, then f can be written as
fle) =(e— pr)" ale),
for # p, where
Thus,
F(a) = mle — By" "ale) + (ep) (a)
and f"(p) = 0. Also,
ie
and J"(p) =0.
In general, for k o,
Thos,
Lam AYleg
So
Jen A CAMA HV
Since the powers of jen] must agree,
+1a and =
‘This number is called the Golden Ratio. It appears in mumerous situations in mathematics and in art.
Exercise Set 2.5, page 86
2. Apply Newton's method to approximate a oot of
f(z) =e + 3(ln2)%e™* — Inse"* — (In2)" = 0.
Generate tems until py+1 ~ pn| < 0.0002, and construct the Aitken’s A? sequence {fn}
SOLUTION: Applying Newton's method with py = 0 requires finding pyg = ~0.182888. For the
Aiken's A? sequence, we have sufficient accuracy with fig = ~0.183:387, Newton’s method fails to
converge quadratically because there is « multiple root. Solutions of Equations of One Variable 2»
‘cos(z ~ 1) and pf? = 2, Use Steffensen’s method to find p}”,
Let oe)
SOLUTION: With g() = coo ~1) and p(
2, we have
(06?) = eon 1) = cont
and
110) =9 (4) = en(ostoaoas — 1) = oso6ts
Tass,
(0) _(0))?
0 =”)
ee es es
1a — 2p” 308
(05103023 — 2)?
OpoIsGT — 2(0.5100023) +2
=2— 1.173578 = 0.826427.
‘5. Steffensen’s method is applied to a function g(r) using p{0”
What could "be?
SOLUTION: Steffensen's method uses the formula
» ey
= ee
PoP
(
i
Substnting for pl, p.®),and p62 gives
wr
3-2 +1
o
and 7 = 1.5009 =
11, b. Use Steffensen’s method to approximate the solution to within 10-* of = 0.5(sin-x + eos-x),
where gis the function in Exereise 11(f) of Section 2.2, thats, g(x) = 0.5(sin x + cos:). 30 Exercise Set2.5
SOLUTION: With (2) =0.5(sin r+ eon), we have
.5(sin 0.5 + e0s0.5) = 0.678503051,
(0 @)?
(ot — of
) wy = O.TTTOIATTA,
ry
ross,
rots,
(1)
Pa
raisin,
= 0.704872252,
=o.704812197,
(Ga = wy
To1812002,
=0.701812002,
Toas12197.
1 <9 (0?)
Since 7, p), and (all agree to within 10-, we accept p62 = 0.704812197 as an answer that is
accurate 0 within 10>
14, a. Show that a soquence {7a} tnt converges top with onder > 1 converges sperinsey top
SOLUTION: Since {nn} converges to pwith order a > 1, positive constant \ exists with
A= Tim Pats al
ww
Hence,
lnmss —p| e
_ pet =a-0
in Fe enol
and
fan Pett—P
Li
This implies that {p,} that converges supertinearly to.
by. Show that py = ;b converges suprlineary to zero, but doesnot converge of order a for any
a> Solutions of Equations of One Variable 31
SOLUTION: This sequence converges superlinearly to zero since
ae ene)
(Ga) ai
ya
=, (Geyer) att eo-&
However the sequence doesnot converge of order for ay a > 1 ice for a > 1, we have
tin WY ne
i Tae = GT
tn. (Ga) eer
1) ni
=i, (Gi) wel
17. Let Pa() be the nth Taylor polynomial for f(x) = e expanded about r» = 0.
a. For fixed x, show that rm = Py () satisfies the hypotheses of Theorem 2.13,
SOLUTION: Since py = Pq(t) =
—
b.Let.r = 1, and use Aitken’s A? method to generate the sequence fn, is. ++ sD
SOLUTION: The sequence has the terms shown inthe following tables.
no 1 2 3 4 5 6
mm 1 2 25 26 2.7083 _B71G 2.71805
fn 3 275 272 271875 2TISR 2.718287) 2.718283
nT 8 9 10
Py 27182539 2.7182787 2.7182815 2.7182818
Bq 27182818 27182818 32 Exercise Set 2.6
¢, Does Altken’s A? method accelerate the convergence in this situation?
SOLUTION: Aitken’s A? method gives quite an improvement for this problem. For example, is
accurate to within 5 x 10-7. We need po to have this accuracy.
Exercise Set 2.6, page 96
2. b. Use Newton's method to approximate, to within 10-5, the real zeros of
25 — 122" +160 — 40.
P(e)
‘Then reduce the polynomial to lower degree, and determine any complex zeros.
SOLUTION: Applying Newton's method with po =
‘pr = —8.548283, When py = 4, we find another zero to be ps,
ives the sufficiently seeurate approximation
‘381113. If we divide P(x) by
(e+ 3.548283)(o — 4.381113) =r? —0,8928800 — 15.5451
wwe find that
P(e) = (22 — 0.832880 ~ 15.5451) (2? — L.16TI2 + 2.57915)
‘The complex roots of the quadratic on the right can be found by the quadratic formula and are
approximately 0.58350 + 1494195,
4. b. Use Miller's method to find the real and complex zeros of
— 2x4 — 122" + 16x — 40,
Pls)
SOLUTION: The following table lists the initial approximation and the roots. The fst initial
approximation was used because f(0) = —40, f(1) = —87, and f(2) = —56 implies that there is a
rinimam in (0,2). Ths is confirmed by the complex roots tha are generated,
‘The second initial approximations are used to ind the real rot that is known tole between 4 and 5,
dive tothe fact tat f(4) = ~40 and f(5) = 115
‘The third inital approximations are used to find the ral root that is known fo lie between — and —
since f(-3) = 61 and f(~4) = 88.
Po PL P2—Approximated Roots Complex Conjugate Root
583560 — 14941881 0.582560 + 1494188
Pe = 4.381113
3.518233 Solutions of Equations of One Variable 33
S. by Find the 2eros and critical points of
S(e) = 0 ~20* 52? + 122-5,
‘and use this information to sketch the graph of f.
SOLUTION: There are at most four real zeros of f and f(0) < 0, f(1) > 0, and f(2) <0. This,
together with the fact that litig-.2o f(z) = oo and lite oo f(z) = co, implies that these zeros lie
inthe intervals (~2c,0), (0,1), (1,2), ad (2,00). Applyig Newton’ method oc varios intial
approximations in these intervals gives the approximate zeros: 0.5708, 1521, 2.282, and —2.492. To
find thecal points, we need the zeros of
Se) = 42" — 6x? — 108 + 12.
Since = 1 is quite easily seen to bea zero of /"(x) the cubic equation ean be reduced to a quadratic
to find the other two zeros: 2 and —1.5.
Since the quadratic form applied to
0= f"(@) = 122 — 122-10
ives « = 0.5 + (/39/6), we also have the point of inflection
AA sketch of te graph of f is given below,
7. Find a solution, accurate to within 10, o the problem
600." — 5500 + 2000? — 200 —
0, ford1
