673

Bisection Method
Solutions
Chapter 2
Rootfinding
2.1
Bisection Method
1. Verify that each of the following equations has a root on the interval (0, 1).
Next, perform the bisection method to determine p3 , the third approximation
to the location of the root, and to determine (a4 , b4 ), the next enclosing interval.
(a)
ln(1 + x) cos x = 0
(b)
x5 + 2x 1 = 0
x
(d)
cos x x = 0
(c)
e x=0
(a) Let f (x) = ln(1 + x) cos x. Because f is continuous on [0, 1] with f (0) =
ln 1 cos 1 = cos 1 < 0 and f (1) = ln 2 cos 2 1.109 > 0, the Intermediate Value Theorem guarantees that there exists a p (0, 1) such that
f (p) = 0. To start the bisection method, take (a1 , b1 ) = (0, 1). The midpoint
of this first interval, and our first approximation to the location of the root, is
p1 =
a1 + b1
0+1
=
= 0.5.
2
2
Note that f (p1 ) 0.472 < 0. Since f (a1 ) and f (p1 ) are of the same sign,
the Intermediate Value Theorem tells us that the root lies between p1 and b1 .
With (a2 , b2 ) = (p1 , b1 ) = (0.5, 1), our second approximation to the location
of the root is
a2 + b2
0.5 + 1
p2 =
=
= 0.75.
2
2
Now f (0.75) 0.172 < 0, which is of the same sign as f (a2 ). Hence, the
Intermediate Value Theorem guarantees that the root is between p2 and b2 ,
so we take (a3 , b3 ) = (p2 , b2 ) = (0.75, 1). For the third iteration, we calculate
p3 =
a3 + b3
0.75 + 1
=
= 0.875
2
2
and f (p3 ) 0.0124 < 0. Here, f (a3 ) and f (p3 ) are of the same sign,
which implies that the root lies between p3 and b3 . Finally, we set (a4 , b4 ) =
(0.875, 1).
Section 2.1
(b) Let f (x) = x5 + 2x 1. Because f is continuous on [0, 1] with f (0) =

1 < 0 and f (1) = 2 > 0, the Intermediate Value Theorem guarantees
that there exists a p (0, 1) such that f (p) = 0. To start the bisection
method, take (a1 , b1 ) = (0, 1). The midpoint of this first interval, and our
first approximation to the location of the root, is
p1 =
a1 + b1
0+1
=
= 0.5.
2
2
Note that f (p1 ) = 0.03125 > 0. Since f (a1 ) and f (p1 ) are of opposite sign,
the Intermediate Value Theorem tells us that the root lies between a1 and p1 .
With (a2 , b2 ) = (a1 , p1 ) = (0, 0.5), our second approximation to the location
of the root is
a2 + b2
0 + 0.5
p2 =
=
= 0.25.
2
2
Now f (p2 ) 0.499 < 0, which is of the same sign as f (a2 ). Hence, the
Intermediate Value Theorem guarantees that the root is between p2 and b2 , so
we take (a3 , b3 ) = (p2 , b2 ) = (0.25, 0.5). For the third iteration, we calculate
p3 =
a3 + b3
0.25 + 0.5
=
= 0.375
2
2
and f (p3 ) 0.243 < 0. Here, f (a3 ) and f (p3 ) are of the same sign,
which implies that the root lies between p3 and b3 . Finally, we set (a4 , b4 ) =
(0.375, 0.5).
(c) Let f (x) = ex x. Because f is continuous on [0, 1] with f (0) = 1 > 0 and
f (1) = e1 1 0.632 < 0, the Intermediate Value Theorem guarantees
p1 =
a1 + b1
0+1
=
= 0.5.
2
2
Note that f (p1 ) 0.107 > 0. Since f (a1 ) and f (p1 ) are of the same sign,
With (a2 , b2 ) = (a1 , p1 ) = (0.5, 1), our second approximation to the location
of the root is
a2 + b2
0.5 + 1
p2 =
=
= 0.75.
2
2
Now f (p2 ) 0.278 < 0, which is of the opposite from f (a2 ). Hence, the
Intermediate Value Theorem guarantees that the root is between a2 and p2 , so
we take (a3 , b3 ) = (a2 , p2 ) = (0.5, 0.75). For the third iteration, we calculate
p3 =
a3 + b3
0.5 + 0.75
=
= 0.625
2
2
and f (p3 ) 0.0897 < 0. Here, f (a3 ) and f (p3 ) are of opposite sign,
which implies that the root lies between a3 and p3 . Finally, we set (a4 , b4 ) =
(0.5, 0.625).
Bisection Method
(d) Let f (x) = cos x x. Because f is continuous on [0, 1] with f (0) = 1 > 0 and
f (1) = cos 1 1 0.460 < 0, the Intermediate Value Theorem guarantees
p1 =
a1 + b1
0+1
=
= 0.5.
2
2
Note that f (p1 ) 0.378 > 0. Since f (a1 ) and f (p1 ) are of the same sign,
With (a2 , b2 ) = (a1 , p1 ) = (0.5, 1), our second approximation to the location
of the root is
a2 + b2
0.5 + 1
p2 =
=
= 0.75.
2
2
Now f (p2 ) 0.0183 < 0, which is of the opposite from f (a2 ). Hence, the
Intermediate Value Theorem guarantees that the root is between a2 and p2 , so
we take (a3 , b3 ) = (a2 , p2 ) = (0.5, 0.75). For the third iteration, we calculate
p3 =
a3 + b3
0.5 + 0.75
=
= 0.625
2
2
and f (p3 ) 0.186 > 0. Here, f (a3 ) and f (p3 ) are of the same sign, which
implies that the root lies between p3 and b3 . Finally, we set (a4 , b4 ) =
(0.625, 0.75).
In Exercises 2 - 5, verify that the given function has a zero on the indicated
interval. Next, perform the first five (5) iterations of the bisection method
and verify that each approximation satisfies the theoretical error bound of the
bisection method, but that the actual errors do not steadily decrease. The exact
location of the zero is indicated by the value of p.
2.
f (x) = x3 + x2 3x 3,
(1, 2),
p=
Let f (x) = x3 + x2 3x 3. Because f is continuous on (1, 2) with f (1) = 4 < 0

and f (2) = 3 > 0, the Intermediate Value Theorem guarantees that there exists
a p (1, 2) such that f (p) = 0. The following table summarizes the first five
iterations of the bisection method starting from the interval (a1 , b1 ) = (1, 2).
n
1
2
3
4
5
pn
1.50000
1.75000
1.62500
1.68750
1.71875
|pn p|
0.23205
0.01794
0.10705
0.04455
0.01330
(b a)/2n
0.50000
0.25000
0.12500
0.06250
0.03125
4
3.
Section 2.1
f (x) = sin x,
(3, 4),
p=
Let f (x) = sin x. Because f is continuous on (3, 4) with f (3) = sin 3 0.141 > 0
and f (4) = sin 4 0.757 < 0, the Intermediate Value Theorem guarantees that
there exists a p (3, 4) such that f (p) = 0. The following table summarizes the first
five iterations of the bisection method starting from the interval (a1 , b1 ) = (3, 4).
n
1
2
3
4
5
4.
f (x) = 1 ln x,
pn
3.50000
3.25000
3.12500
3.18750
3.15625
(2, 3),
|pn p|
0.35841
0.10841
0.01659
0.04591
0.01466
(b a)/2n
0.50000
0.25000
0.12500
0.06250
0.03125
p=e
Let f (x) = 1 ln x. Because f is continuous on (2, 3) with f (2) = 1 ln 2

0.307 > 0 and f (3) = 1 ln 3 0.0986 < 0, the Intermediate Value Theorem
guarantees that there exists a p (2, 3) such that f (p) = 0. The following table
summarizes the first five iterations of the bisection method starting from the interval
(a1 , b1 ) = (2, 3).
5.
f (x) = x6 3,
n
1
2
3
4
5
pn
2.50000
2.75000
2.62500
2.68750
2.71875
(1, 2),
p=
|pn p|
0.21828
0.03172
0.09328
0.03078
0.00047
(b a)/2n
0.50000
0.25000
0.12500
0.06250
0.03125
Let f (x) = x6 3. Because f is continuous on (1, 2) with f (1) = 2 < 0 and

f (2) = 63 > 0, the Intermediate Value Theorem guarantees that there exists a
p (1, 2) such that f (p) = 0. The following table summarizes the first five
iterations of the bisection method starting from the interval (a1 , b1 ) = (1, 2).
n
1
2
3
4
5
pn
1.50000
1.25000
1.12500
1.18750
1.21875
|pn p|
0.29906
0.04906
0.07594
0.01344
0.01781
(b a)/2n
0.50000
0.25000
0.12500
0.06250
0.03125
6. Determine a formula which relates the number of iterations, n, required by the

bisection method to converge to within an absolute error tolerance of , starting
from the initial interval (a, b).
Bisection Method
Let p denote the location of the true root, and let pn denote the approximate
location of the root produced by the nth iteration of the bisection method. From
the proof of the bisection method convergence theorem, we know that
|pn p|
ba
.
2n
The bisection method sequence will therefore converge to within an absolute tolerance of provided
ba
< .
2n
Solving this last expression for n gives
n > log2
ba
.
7. Modify the algorithm for the bisection method as follows. Remove the input
Nmax, and calculate the number of iterations needed to achieve the specified
convergence tolerance using the results of Exercise 6.
Here is the modified bisection method algorithm:
GIVEN:
function whose zero is to be located, f

left endpoint of interval, a
right endpoint of interval, b
convergence tolerance,
STEP
STEP
STEP
STEP
STEP
STEP
set N max = 1 + int(log2 ((b a)/))

save sf a = sign(f (a))
for i from 1 to Nmax
p = a + (b a)/2
save sf p = sign(f (p))
if (sf a sf p < 0) then
assign the value of p to b
else
assign the value of p to a
assign the value of sfp to sfa
end
end
p
1:
2:
3:
4:
5:
6:
OUTPUT:
8. Suppose that an equation is known to have a root on the interval (0, 1). How
many iterations of the bisection method are needed to achieve full machine
precision in the approximation to the location of the root assuming calculations
are performed in IEEE standard double precision? What if the root were known
Section 2.1
to be contained in the interval (8,9)? (Hint: Consider the number of base 2

digits already known in the location of the root and how many base 2 digits are
available in the indicated floating point system.)
Recall that IEEE standard double precision has 53 binary digits of precision. If the
root is known to lie on the interval (0, 1), then no binary digits in the location of
the root are known from the outset. Each iteration of the bisection method adds
one binary digit to the approximation, so a total of 53 iterations will be needed to
approximate the root to full machine precision. If, on the other hand, the root is
known to lie on the interval (8, 9), then four binary digits (1000) are known from
the outset. 49 iterations of the bisection method will be needed to approximate the
location of the root to full machine precision.
9. By construction, the endpoints of the enclosing intervals produced by the bisection method satisfy a1 a2 a3 b3 b2 b1 . Prove that the
sequences {an } and {bn } converge and that
lim an = lim bn = lim pn = p.
The sequence {an } is increasing (an+1 an for all n) and bounded from above
(an b1 for all n); hence, the sequence must converge. Similarly, the sequence
{bn } is decreasing (bn+1 bn for all n) and bounded from below (bn a1 for all
n); hence, this sequence must also converge. Now, for all n,
an pn bn ,
so
lim an lim pn lim bn .
(1)
From the proof of the bisection method convergence theorem, we know that
bn an =
ba
,
2n
which implies that

lim (bn an ) = 0
or
lim an = lim bn .
(2)
Combining (1) and (2) yields

lim an = lim bn = lim pn = p.
10. It was noted that the function f (x) = x3 +2x2 3x1 has a zero on the interval
(3, 2) and another on the interval (1, 0). Approximate both of these zeroes
to within an absolute tolerance of 5 105 .
Bisection Method
The following table displays the results produced by the bisection method when
applied to the function f (x) = x3 + 2x2 3x 1 with the starting intervals
(a1 , b1 ) = (3, 2) and (a1 , b1 ) = (1, 0). In each case, a convergence tolerance
of = 5 105 is used.
n
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
(a1 , b1 ) = (3, 2)
2.5000000000
2.7500000000
2.8750000000
2.9375000000
2.9062500000
2.9218750000
2.9140625000
2.9101562500
2.9121093750
2.9130859375
2.9125976562
2.9123535156
2.9122314453
2.9121704102
2.9122009277
(a1 , b1 ) = (1, 0)
0.5000000000
0.2500000000
0.3750000000
0.3125000000
0.2812500000
0.2968750000
0.2890625000
0.2851562500
0.2871093750
0.2861328125
0.2866210938
0.2863769531
0.2864990234
0.2864379883
0.2864685059
We therefore estimate that f has roots at approximately x = 2.91220 and x =

0.28647. Both approximations are in error by at most 5 105 .
11. Approximate 3 13 to three decimal places by applying the bisection method to

the equation x3 13 = 0.
Let f (x) = x3 13. Since f (2) = 5 < 0 and f (3) = 14 > 0, we know there is a
root on the interval (a1 , b1 ) = (2, 3). Using a convergence tolerance of = 5104 ,
the bisection method yields
n
1
2
3
4
5
6
7
8
9
10
11
Thus,
Enclosing Interval
(2.000000,3.000000)
(2.000000,2.500000)
(2.250000,2.500000)
(2.250000,2.375000)
(2.312500,2.375000)
(2.343750,2.375000)
(2.343750,2.359375)
(2.343750,2.351562)
(2.347656,2.351562)
(2.349609,2.351562)
(2.350586,2.351562)
Approximation
2.5000000000
2.2500000000
2.3750000000
2.3125000000
2.3437500000
2.3593750000
2.3515625000
2.3476562500
2.3496093750
2.3505859375
2.3510742188
3
13 2.35107, with an error of at most 5 105 .
12. Approximate 1/37 to five decimal places by applying the bisection method to
Section 2.1
the equation 1/x 37 = 0.

Let f (x) =
1
x
37. Since
0.025 =
1
1
1
<
<
= 0.05,
40
37
20
1
1
we apply the bisection method with (a1 , b1 ) = ( 40
, 20
). Knowing that the first
decimal digit is a zero, to obtain five significant decimal digits, we take a convergence
tolerance of 5 107 . The bisection method then yields
n
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Thus,
1
37
Enclosing Interval
(0.025000,0.050000)
(0.025000,0.037500)
(0.025000,0.031250)
(0.025000,0.028125)
(0.026563,0.028125)
(0.026563,0.027344)
(0.026953,0.027344)
(0.026953,0.027148)
(0.026953,0.027051)
(0.027002,0.027051)
(0.027026,0.027051)
(0.027026,0.027039)
(0.027026,0.027032)
(0.027026,0.027029)
(0.027026,0.027028)
(0.027026,0.027027)
Approximation
0.0375000000
0.0312500000
0.0281250000
0.0265625000
0.0273437500
0.0269531250
0.0271484375
0.0270507813
0.0270019531
0.0270263672
0.0270385742
0.0270324707
0.0270294189
0.0270278931
0.0270271301
0.0270267487
0.0270267.
13. In one of the worked examples of this section, the smallest positive root of the
equation tan(x) x 6 = 0 was approximated. Graphically determine an
interval which contains the next smallest positive root of this equation, and
then approximate the root to within an absolute tolerance of 5 105 .
The function tan(x) is periodic with period 1. Since the smallest positive root is
found at x 0.45, it seems reasonable to expect the next smallest positive root
to be found at x 1.45. The graph below suggests that a starting interval of
(a1 , b1 ) = (1.45, 1.47) would be appropriate.
Bisection Method
With (a1 , b1 ) = (1.45, 1.47) and = 5 105 , the bisection method yields
n
1
2
3
4
5
6
7
8
9
Enclosing Interval
(1.450000,1.470000)
(1.450000,1.460000)
(1.455000,1.460000)
(1.457500,1.460000)
(1.457500,1.458750)
(1.457500,1.458125)
(1.457500,1.457812)
(1.457500,1.457656)
(1.457500,1.457578)
Approximation
1.4600000000
1.4550000000
1.4575000000
1.4587500000
1.4581250000
1.4578125000
1.4576562500
1.4575781250
1.4575390625
Thus, the second smallest positive root of the equation tan(x) x 6 = 0 is

approximately x = 1.457539.
14. The equation (x 0.5)(x + 1)3 (x 2) = 0 clearly has roots at x = 1, x =
0.5, and x = 2. Each of the intervals listed below encompasses all of these
roots. Determine to which root the bisection method converges when each of
the intervals below is used as the starting interval.
(a) (3, 3)
(b) (1.5, 3)
(c) (2, 4)
(d) (2, 3)
(e) (1.5, 2.2)
(f ) (7, 3)
(a) With (a1 , b1 ) = (3, 3), the bisection method converges toward x = 1.
(b) With (a1 , b1 ) = (1.5, 3), the bisection method converges toward x = 2.
(c) With (a1 , b1 ) = (2, 4), the bisection method converges toward x = 2.
(d) With (a1 , b1 ) = (2, 3), the bisection method converges toward x = 0.5.
(e) With (a1 , b1 ) = (1.5, 2.2), the bisection method converges toward x = 1.
(f ) With (a1 , b1 ) = (7, 3), the bisection method converges toward x = 0.5.
10
Section 2.1
15. It can be shown that the equation

3
1
x 6 sin(2x) = 0
2
2
has a unique real root.
(a) Find an interval on which this unique real root is guaranteed to exist.
(b) Using the interval found in part (a) and the bisection method, approximate
the root to within an absolute tolerance of 105 .
(a) The equation

3
1
x 6 sin(2x) = 0
2
2
is equivalent to
3
1
x 6 = sin(2x).
2
2
Because
1
1
1
sin(2x) ,
2
2
2
the root we are seeking must satisfy
1
3
1
x6
2
2
2
or
11
13
x
.
3
3
13
5
(b) With (a1 , b1 ) = ( 11
, the bisec3 , 3 ) and a convergence tolerance of = 10
tion method yields
n
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Enclosing Interval
(3.666667,4.333333)
(4.000000,4.333333)
(4.166667,4.333333)
(4.250000,4.333333)
(4.250000,4.291667)
(4.250000,4.270833)
(4.260417,4.270833)
(4.260417,4.265625)
(4.260417,4.263021)
(4.260417,4.261719)
(4.261068,4.261719)
(4.261393,4.261719)
(4.261393,4.261556)
(4.261475,4.261556)
(4.261475,4.261515)
(4.261475,4.261495)
(4.261475,4.261485)
Approximation
4.0000000000
4.1666666667
4.2500000000
4.2916666667
4.2708333333
4.2604166667
4.2656250000
4.2630208333
4.2617187500
4.2610677083
4.2613932292
4.2615559896
4.2614746094
4.2615152995
4.2614949544
4.2614847819
4.2614796956
Bisection Method
11
Thus, the unique root of the equation

3
1
x 6 sin(2x) = 0
2
2
is approximately x = 4.261482.
16. For each of the functions given below, use the bisection method to approximate
all real zeros. Use an absolute tolerance of 106 as a stopping criterion.
(a)
f (x) = ex + x2 x 4
(b)
f (x) = x3 x2 10x + 7
(c)
f (x) = 1.05 1.04x + ln x
(a) Let f (x) = ex + x2 x 4. Observe that the equation ex + x2 x 4 = 0 is

equivalent to the equation ex = x2 + x + 4. The figure below displays the
graphs of y = ex and y = x2 + x + 4.
The graphs appear to intersect over the intervals (2, 1) and (1, 2). Using
each of these intervals and a convergence tolerance of 106 , the bisection
method yields
12
Section 2.1
n
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
(a1 , b1 ) = (2, 1)
1.5000000000
1.7500000000
1.6250000000
1.5625000000
1.5312500000
1.5156250000
1.5078125000
1.5039062500
1.5058593750
1.5068359375
1.5073242188
1.5070800781
1.5072021484
1.5071411133
1.5071105957
1.5070953369
1.5071029663
1.5070991516
1.5071010590
1.5071001053
(a1 , b1 ) = (1, 2)
1.5000000000
1.2500000000
1.3750000000
1.3125000000
1.2812500000
1.2968750000
1.2890625000
1.2851562500
1.2871093750
1.2880859375
1.2885742188
1.2888183594
1.2886962891
1.2886352539
1.2886657715
1.2886810303
1.2886734009
1.2886772156
1.2886791229
1.2886781693
Thus, the zeros of f (x) = ex + x2 x 4 are approximately x = 1.5071001

and x = 1.2886782.
(b) Let f (x) = x3 x2 10x + 7. By trial and error, we find that f (4) < 0,
f (3) > 0, f (0) > 0, f (1) < 0, f (3) < 0 and f (4) > 0. Therefore, the
three real zeros of f lie on the intervals (4, 3), (0, 1) and (3, 4). Using each
of these intervals and a convergence tolerance of 106 , the bisection method
yields
Bisection Method
n
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
13
(a1 , b1 ) = (4, 3)
3.5000000000
3.2500000000
3.1250000000
3.0625000000
3.0312500000
3.0468750000
3.0390625000
3.0429687500
3.0410156250
3.0419921875
3.0424804688
3.0427246094
3.0426025391
3.0426635742
3.0426940918
3.0426788330
3.0426864624
3.0426826477
3.0426845551
3.0426836014
(a1 , b1 ) = (0, 1)
0.5000000000
0.7500000000
0.6250000000
0.6875000000
0.6562500000
0.6718750000
0.6796875000
0.6835937500
0.6855468750
0.6845703125
0.6850585938
0.6853027344
0.6851806641
0.6852416992
0.6852111816
0.6852264404
0.6852188110
0.6852226257
0.6852207184
0.6852197647
(a1 , b1 ) = (3, 4)
3.5000000000
3.2500000000
3.3750000000
3.3125000000
3.3437500000
3.3593750000
3.3515625000
3.3554687500
3.3574218750
3.3583984375
3.3579101562
3.3576660156
3.3575439453
3.3574829102
3.3574523926
3.3574676514
3.3574600220
3.3574638367
3.3574619293
3.3574628830
Thus, the zeros of f (x) = x3 x2 10x+7 are approximately x = 3.0426836,

x = 0.6852198 and x = 3.3574629.
(c) Let f (x) = 1.051.04x+ln x. Observe that the equation 1.051.04x+ln x =
0 is equivalent to the equation ln x = 1.04x 1.05. The figure below displays
the graphs of y = ln x and y = 1.04x 1.05.
The graphs appear to intersect over the intervals (0.80, 0.85) and (1.10, 1.15).
Using each of these intervals and a convergence tolerance of 106 , the bisection
method yields
14
Section 2.1
n
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
(a1 , b1 ) = (0.80, 0.85)

0.8250000000
0.8375000000
0.8312500000
0.8281250000
0.8265625000
0.8273437500
0.8269531250
0.8271484375
0.8272460938
0.8271972656
0.8271728516
0.8271850586
0.8271789551
0.8271820068
0.8271804810
0.8271812439
(a1 , b1 ) = (1.10, 1.15)

1.1250000000
1.1125000000
1.1062500000
1.1093750000
1.1109375000
1.1101562500
1.1097656250
1.1095703125
1.1096679688
1.1097167969
1.1096923828
1.1097045898
1.1097106934
1.1097137451
1.1097122192
1.1097129822
Thus, the zeros of f (x) = 1.051.04x+ln x are approximately x = 0.8271812

and x = 1.1097130.
17. Peters (Optimum Spring-Damper Design for Mass Impact, SIAM Review, 39
(1), pp. 118 - 122, 1997) models the impact of an object on a spring-damper
system. If the displacement of the object following impact is limited, then the
maximum force exerted on the object is minimized when the nondimensional
damping coefficient, , is the solution of the equation
h p
i
cos 4 1 2 = 1 + 8 2 8 4
on the interval 0 < < 1/2. The maximum (nondimensional) force is then
given by
Fm = exp [(f + m )] ,
where
p
f = cos1 / 1 2
is the time of the end of the stroke and

p
m = cos1 (3 4 2 ) / 1 2
is the time when the maximum force occurs. Determine to within an absolute
tolerance of 5 107 , and then calculate f , m and Fm .
h p
i
Let f () = cos 4 1 2 + 1 8 2 + 8 4 . With a starting interval of (a1 , b1 ) =
(0, 0.5) and a convergence tolerance of 5 107 , the bisection method yields
Bisection Method
15
n
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Enclosing Interval
(0.000000,0.500000)
(0.250000,0.500000)
(0.375000,0.500000)
(0.375000,0.437500)
(0.375000,0.406250)
(0.390625,0.406250)
(0.398438,0.406250)
(0.402344,0.406250)
(0.402344,0.404297)
(0.403320,0.404297)
(0.403809,0.404297)
(0.403809,0.404053)
(0.403931,0.404053)
(0.403931,0.403992)
(0.403961,0.403992)
(0.403961,0.403976)
(0.403969,0.403976)
(0.403973,0.403976)
(0.403973,0.403975)
(0.403973,0.403974)
Approximation
0.2500000000
0.3750000000
0.4375000000
0.4062500000
0.3906250000
0.3984375000
0.4023437500
0.4042968750
0.4033203125
0.4038085938
0.4040527344
0.4039306641
0.4039916992
0.4039611816
0.4039764404
0.4039688110
0.4039726257
0.4039745331
0.4039735794
0.4039731026
Thus, = 0.4039731. We then calculate f = 1.2625461, m = 0.3533436, and

Fm = 0.5205986.
18. DeSantis, Gironi and Marelli (Vector-liquid equilibrium from a hard-sphere
equation of state, Industrial and Engineering Chemistry Fundamentals, 15,
182-189, 1976) derive a relationship for the compressibility factor of real gases
of the form
z=
1 + y + y2 y3
,
(1 y)3
where y is related to the van der Waals volume correction factor. If z = 0.892,
what is the value of y?
Let
f (y) =
1 + y + y2 y3
0.892.
(1 y)3
By trial and error, we find

f (1.8) 1.298 < 0
and
f (2) = 0.108 > 0.
Applying the bisection method with (a1 , b1 ) = (1.8, 2) and a convergence tolerance
of 5 105 yields
16
Section 2.1
n
1
2
3
4
5
6
7
8
9
10
11
12
Enclosing Interval
(1.800000,2.000000)
(1.900000,2.000000)
(1.950000,2.000000)
(1.950000,1.975000)
(1.962500,1.975000)
(1.968750,1.975000)
(1.971875,1.975000)
(1.973438,1.975000)
(1.974219,1.975000)
(1.974609,1.975000)
(1.974609,1.974805)
(1.974609,1.974707)
Approximation
1.9000000000
1.9500000000
1.9750000000
1.9625000000
1.9687500000
1.9718750000
1.9734375000
1.9742187500
1.9746093750
1.9748046875
1.9747070312
1.9746582031
Thus, y 1.974658.
19. Reconsider the Saving for a Down Payment application problem. Which of
the following scenarios requires a smaller compounded monthly interest rate to
achieve a goal of $25,000 after three years:
(a) a $14,000 initial investment with $250 per month thereafter; or
(b) a $12,500 initial investment with $300 per month thereafter?
Let r denote the compounded monthly interest rate. Under scenario (a), the couple
will have saved

r 36

1 + 12
1
r 36
14000 1 +
+ 250
12
r/12
dollars by the end of three years. The couples goal is to save $25,000, so let

r 36

1 + 12
1
r 36
fa (r) = 14000 1 +
+ 250
25000.
12
r/12
Because fa (0.01) = 1441.32 < 0 and fa (0.05) = 948.95 > 0, the desired interest
rate lies somewhere between 1% and 5%. Under scenario (b), the couple will have
saved

r 36

1 + 12
1
r 36
12500 1 +
+ 300
12
r/12
dollars by the end of three years, so let

r 36

1 + 12
1
r 36
fb (r) = 12500 1 +
+ 300
25000.
12
r/12
With fb (0.01) = 1160.48 < 0 and fb (0.05) = 1144.40 > 0, it follows that the
desired interest rate is again somewhere between 1% and 5%.
The following table presents the results of the bisection method applied to the
functions fa and fb . In each case, a starting interval of (a1 , b1 ) = (0.01, 0.05) and
a convergence tolerance of = 5 106 were used.
Bisection Method
17
n
1
2
3
4
5
6
7
8
9
10
11
12
13
Scenario (a)
0.0300000000
0.0400000000
0.0350000000
0.0325000000
0.0337500000
0.0343750000
0.0346875000
0.0345312500
0.0346093750
0.0346484375
0.0346289063
0.0346191406
0.0346240234
Scenario (b)
0.0300000000
0.0400000000
0.0350000000
0.0325000000
0.0312500000
0.0306250000
0.0309375000
0.0307812500
0.0307031250
0.0306640625
0.0306445312
0.0306542969
0.0306591797
Thus, scenario (a) requires a minimum compounded monthly interest rate of 3.46%,
while scenario (b) requires a minimum compounded monthly interest rate of 3.07%.
Because scenario (b) requires a lower interest rate to achieve the $25,000 goal,
scenario (b) is the better investment option.
The Method of False Position
2.2
1. Each of the following equations has a root on the interval (0, 1). Perform the
method of false position to determine p3 , the third approximation to the location
of the root, and to determine (a4 , b4 ), the next enclosing interval.
(a)
ln(1 + x) cos x = 0
(b)
x5 + 2x 1 = 0
x
(d)
cos x x = 0
(c)
e x=0
(a) Let f (x) = ln(1 + x) cos x. For the first iteration, we have (a1 , b1 ) = (0, 1)
and we know that f (a1 ) = 1 < 0 and that f (b1 ) = ln 2 cos 1 0.153 > 0.
Our first approximation to the location of the root is
p1 = b1 f (b1 )
b1 a1
= 0.867419392.
f (b1 ) f (a1 )
To determine whether the root is contained on (a1 , p1 ) or on (p1 , b1 ), we

calculate f (p1 ) 0.0222 < 0. Since f (a1 ) and f (p1 ) are of the same sign,
the Intermediate Value Theorem tells us that root is between p1 and b1 . For
the next iteration, we therefore take (a2 , b2 ) = (p1 , b1 ) = (0.867419392, 1).
Our second approximation to the location of the root is
p2 = b2 f (b2 )
b2 a2
= 0.884259901.
f (b2 ) f (a2 )
Note that f (p2 ) 3.270 104 < 0, which is of the same sign as f (a2 ).
Hence, the Intermediate Value Theorem tells us the root is between p2 and
b2 , so we take (a3 , b3 ) = (p2 , b2 ) = (0.884259901, 1). In the third iteration,
we calculate
p3 = b3 f (b3 )
b3 a3
= 0.884506977
f (b3 ) f (a3 )
and f (p3 ) 4.746 106 < 0. Hence, we find that f (a3 ) and f (p3 )
are of the same sign, which implies that the root lies somewhere between p3
and b3 . For the fourth iteration, we will therefore take (a4 , b4 ) = (p3 , b3 ) =
(0.884506977, 1).
(b) Let f (x) = x5 +2x1. For the first iteration, we have (a1 , b1 ) = (0, 1) and we
know that f (a1 ) = 1 < 0 and that f (b1 ) = 2 > 0. Our first approximation
to the location of the root is
p1 = b1 f (b1 )
b1 a1
= 0.333333333.
f (b1 ) f (a1 )
Section 2.2
calculate f (p1 ) 0.329 < 0. Since f (a1 ) and f (p1 ) are of the same sign,
p2 = b2 f (b2 )
b2 a2
= 0.427561837.
f (b2 ) f (a2 )
Note that f (p2 ) 0.131 < 0, which is of the same sign as f (a2 ). Hence, the
Intermediate Value Theorem tells us the root is between p2 and b2 , so we take
(a3 , b3 ) = (p2 , b2 ) = (0.427561837, 1). In the third iteration, we calculate
p3 = b3 f (b3 )
b3 a3
= 0.462647607
f (b3 ) f (a3 )
and f (p3 ) 0.0535 < 0. Hence, we find that f (a3 ) and f (p3 ) are of
the same sign, which implies that the root lies somewhere between p3 and
b3 . For the fourth iteration, we will therefore take (a4 , b4 ) = (p3 , b3 ) =
(0.462647607, 1).
(c) Let f (x) = ex x. For the first iteration, we have (a1 , b1 ) = (0, 1) and we
know that f (a1 ) = 1 > 0 and that f (b1 ) = e1 1 0.632 < 0. Our first
approximation to the location of the root is
p1 = b1 f (b1 )
b1 a1
= 0.612699837.
f (b1 ) f (a1 )

calculate f (p1 ) 0.0708 < 0. Since f (a1 ) and f (p1 ) are of opposite sign,
the Intermediate Value Theorem tells us that root is between a1 and p1 . For
the next iteration, we therefore take (a2 , b2 ) = (a1 , p1 ) = (0, 0.612699837).
p2 = b2 f (b2 )
b2 a2
= 0.572181412.
f (b2 ) f (a2 )
Note that f (p2 ) 0.00789 < 0, which is of opposite sign from f (a2 ).
Hence, the Intermediate Value Theorem tells us the root is between a2 and
p2 , so we take (a3 , b3 ) = (a2 , p2 ) = (0, 0.572181412). In the third iteration,
we calculate
p3 = b3 f (b3 )
b3 a3
= 0.567703214
f (b3 ) f (a3 )
and f (p3 ) 8.774 104 < 0. Hence, we find that f (a3 ) and f (p3 )
are of opposite sign, which implies that the root lies somewhere between a3
and p3 . For the fourth iteration, we will therefore take (a4 , b4 ) = (a3 , p3 ) =
(0, 0.567703214).
(d) Let f (x) = cos x x. For the first iteration, we have (a1 , b1 ) = (0, 1) and we
know that f (a1 ) = 1 > 0 and that f (b1 ) = cos 1 1 0.460 < 0. Our first
approximation to the location of the root is
p1 = b1 f (b1 )
b1 a1
= 0.685073357.
f (b1 ) f (a1 )

calculate f (p1 ) 0.0893 > 0. Since f (a1 ) and f (p1 ) are of the same sign,
p2 = b2 f (b2 )
b2 a2
= 0.736298998.
f (b2 ) f (a2 )
Note that f (p2 ) 0.00466 > 0, which is of the same sign as f (a2 ). Hence,
the Intermediate Value Theorem tells us the root is between p2 and b2 , so
we take (a3 , b3 ) = (p2 , b2 ) = (0.736298998, 1). In the third iteration, we
calculate
b3 a3
p3 = b3 f (b3 )
= 0.738945356
f (b3 ) f (a3 )
and f (p3 ) 2.339 104 < 0. Hence, we find that f (a3 ) and f (p3 ) are
of the same sign, which implies that the root lies somewhere between p3
and b3 . For the fourth iteration, we will therefore take (a4 , b4 ) = (p3 , b3 ) =
(0.738945356, 1).
2. Construct an algorithm for the method of false position. Remember to save
function values which will be needed for later iterations and to implement a
stopping condition based on equations (6) and (7).
Here is an algorithm for the method of false position:
GIVEN:

left endpoint of interval, a
right endpoint of interval, b
convergence tolerance,
maximum number of iterations, Nmax
STEP
STEP
STEP
STEP
STEP
initialize polder = b and pold = b

save f a = f (a) and f b = f (b)
set p = b f b (b a)/(f b f a)
if ( i > 2 )
set = (p pold)/(pold polder)
set errest = |(p pold)/( 1)|
if ( errest < ) OUTPUT p
1:
2:
3:
4:
5:
Section 2.2
STEP 6:
STEP 7:
STEP 8:
OUTPUT:
end
save f p = f (p)
if (sign(f a) sign(f p) < 0)
assign the value of p to b
assign the value of fp to fb
else
assign the value of p to a
assign the value of fp to fa
end
assign the value of pold to polder
assign the value of p to pold
end
maximum number of iterations exceeded
3. Confirm that || < 1 for the remaining configurations in Figure 2-5.

Start with the configuration depicted in the upper right panel of Figure 2.5. Because
an is fixed, l = an p. Now, an p < 0 and f (p) > 0, so (an p)f (p) < 0. Since
f (p) is also less than zero, it follows that 2f (p) + (an p)f (p) < (an p)f (p)
and
(an p)f (p)
0<
= < 1.
2f (p) + (an p)f (p)
Hence, || < 1.
In the lower left panel of Figure 2.5, bn is fixed, so l = bn p. Now, bn p > 0 and
f (p) > 0, so (bn p)f (p) > 0. Since f (p) is also greater than zero, it follows
that 2f (p) + (bn p)f (p) > (bn p)f (p) and
0<
(bn p)f (p)

= < 1.
2f (p) + (bn p)f (p)
Hence, || < 1.
Finally, in the lower right panel of Figure 2.5, bn is fixed, so l = bn p. Now,
bn p > 0 and f (p) < 0, so (bn p)f (p) < 0. Since f (p) is also less than zero,
it follows that 2f (p) + (bn p)f (p) < (bn p)f (p) and
0<
(bn p)f (p)

= < 1.
2f (p) + (bn p)f (p)
Hence, || < 1.
In Exercises 4 - 7, an equation, an interval on which the equation has a root,
and the exact value of the root are specified.
(a) Perform the first five (5) iterations of the method of false position.
(b) Verify that the absolute error in the third, fourth and fifth approximations
satisfies the error estimate

|pn pn1 |.
|pn p|
1
(c) How does the error in the fifth false position approximation compare to the
maximum error which would result from six iterations of the bisection method?
4. The
equation x3 + x2 3x 3 = 0 has a root on the interval (1, 2), namely
x = 3.
The following table summarizes the results of five iterations of the method of false
position with f (x) = x3 + x2 3x 3 and (a1 , b1 ) = (1, 2).
n
1
2
3
4
5
pn
1.571429
1.705411
1.727883
1.731405
1.731951
|pn p|
0.160622
2.664 102
4.168 103
6.459 104
9.995 105
error
estimate
4.529 103
6.546 104
1.002 104
Note that the error in the fifth false position approximation, 9.995 105 , is substantially smaller than the maximum error which would result from six iterations of
the bisection method, (2 1)/26 = 0.015625.
5. The equation x7 = 3 has a root on the interval (1, 2), namely x =
7
3.
position with f (x) = x7 3 and (a1 , b1 ) = (1, 2).
n
1
2
3
4
5
pn
1.015748
1.030366
1.043882
1.056333
1.067761
|pn p|
0.154183
0.139565
0.126049
0.113598
0.102169
error
estimate
0.165874
0.145556
0.127730
Note that the error in the fifth false position approximation, 0.102169, is an order
of magnitude larger than the maximum error which would result from six iterations
of the bisection method, (2 1)/26 = 0.015625.
6. The equation x3 13 = 0 has a root on the interval (2, 3), namely
13.
position with f (x) = x3 13 and (a1 , b1 ) = (2, 3).
Section 2.2
pn
2.263158
2.330507
2.346490
2.350212
2.351075
n
1
2
3
4
5
|pn p|
0.088177
0.020827
4.844 103
1.123 103
2.600 104
error
estimate
4.973 103
1.130 103
2.604 104
Note that the error in the fifth false position approximation, 2.600 104 , is substantially smaller than the maximum error which would result from six iterations of
the bisection method, (3 2)/26 = 0.015625.
7. The equation 1/x37 = 0 has a zero on the interval (0.01, 0.1), namely x = 1/37.
position with f (x) = x1 37 and (a1 , b1 ) = (0.01, 0.1).
n
1
2
3
4
5
pn
0.073000
0.055990
0.045274
0.038522
0.034269
|pn p|
0.045973
0.028963
0.018247
0.011495
0.007242
error
estimate
0.018247
0.011495
0.007242
Note that the error in the fifth false position approximation, 0.007242, is nearly five
times larger than the maximum error which would result from six iterations of the
bisection method, (0.1 0.01)/26 = 0.00140625.
8. The function f (x) = sin x has a zero on the interval (3, 4), namely x = .
Perform three iterations of the method of false position to approximate this zero.
Determine the absolute error in each of the three computed approximations.
What is the apparent order of convergence? What explanation can you provide
for this behavior?
The following table summarizes the results of three iterations of the method of false
position with f (x) = sin x and (a1 , b1 ) = (3, 4).
n
1
2
3
pn
3.1571627924799466
3.1415462555891498
3.1415926554589646
|pn p|
1.557 102
4.640 105
1.869 109
Each error appears to be the square of the previous error, so the order of convergence
appears to be = 2; i.e., quadratic convergence. The order of convergence for this
specific problem is better than the expected linear convergence for the method of
false position because f () = sin = 0; thus,

|en |
lf ()
=
= 0,
n |en1 |
2f () + lf ()
lim
which implies that convergence is better than linear.

9. (a) Verify that the equation x4 18x2 +45 = 0 has a root on the interval (1, 2).
Next, perform three iterations of the method
of false position. Given that
the exact value of the root is x = 3, compute the absolute error in

the three approximations just obtained. What is the apparent order of
convergence? What explanation can you provide for this behavior?
(b) Verify that the equation x4 18x2 + 45 = 0 also has a root on the interval
(3, 4). Perform five iterations of the method of false position, and compute
the absolute
error in each approximation. The exact value of the root is
x = 15. What is the apparent order of convergence in this case?

(c) What explanation can you provide for the different convergence behavior
between parts (a) and (b)?
(a) Let f (x) = x4 18x2 + 45. Then f (1) = 28 > 0 and f (2) = 11 < 0, so the
Intermediate Value Theorem guarantees the existence of a root on the interval
(1, 2). With (a1 , b1 ) = (1, 2), the following table summarizes the results of
three iterations of the method of false position.
n
1
2
3
pn
1.717948717949
1.732218859330
1.732050802076
|pn p|
1.410 102
1.681 104
5.493 109
Convergence appears to be of order two (note that each error appears to be

the square of the previous error), which is better
than expected for the method
of false position. This happens because f ( 3) = 0; thus,
|en |
lf ( 3)
= 0,
lim
=
n |en1 |
2f ( 3) + lf ( 3)
which implies that convergence is better than linear.
(b) Let f (x) = x4 18x2 + 45. Then f (3) = 36 < 0 and f (4) = 13 > 0, so the
(3, 4). With (a1 , b1 ) = (3, 4), the following table summarizes the results of
five iterations of the method of false position.
n
1
2
3
4
5
pn
3.734693877551
3.859328133794
3.871720773366
3.872867347773
3.872972695145
|pn p|
1.383 101
1.366 102
1.263 103
1.160 104
1.065 105
Section 2.2
Each error is roughly one-tenth the previous error, so convergence is linear.
(c) In part (b), f ( 15) 6= 0, so the error analysis from the text holds, and the
method of false
position exhibits linear convergence. On the other hand, in
part (a), f ( 3) = 0 so convergence is faster than linear. We can expect this
to be true with the method of false position whenever f (p) = 0.
10. The function f (x) = x3 + 2x2 3x 1 has a zero on the interval (1, 0).
Approximate this zero to within an absolute tolerance of 5 105 .
With f (x) = x3 + 2x2 3x 1, (a1 , b1 ) = (1, 0) and = 5 105 , the method
of false position yields
n
1
2
3
4
Enclosing Interval
(1.0000000000, 0.0000000000)
(1.0000000000, 0.2500000000)
(1.0000000000, 0.2835820896)
(1.0000000000, 0.2862518550)
Approximation
0.2500000000
0.2835820896
0.2862518550
0.2864468212
Thus, the zero of f (x) = x3 + 2x2 3x 1 on the interval (1, 0) is approximately

x = 0.2864468. The estimate for the absolute error in this approximation is
1.536 105 .
11. For each of the functions given below, use the method of false position to approximate all real roots. Use an absolute tolerance of 106 as a stopping condition.
(a)
f (x) = ex + x2 x 4
(b)
f (x) = x3 x2 10x + 7
(c)
f (x) = 1.05 1.04x + ln x

each of these intervals and a convergence tolerance of 106 , the method of
false position yields
n
1
2
3
4
5
6
7
8
9
10
11
(a1 , b1 ) = (2, 1)
1.4332155776
1.4977012853
1.5059259021
1.5069532761
1.5070812744
1.5070972162
1.5070992016
(a1 , b1 ) = (1, 2)
1.1921393409
1.2578082629
1.2789515912
1.2856277625
1.2877228316
1.2883790162
1.2885844109
1.2886486901
1.2886688053
1.2886750999
1.2886770697

and x = 1.2886771.
f (3) > 0, f (0) > 0, f (1) < 0, f (3) < 0 and f (4) > 0. Therefore, the
three real zeros of f lie on the intervals (4, 3), (0, 1) and (3, 4). Using each
of these intervals and a convergence tolerance of 106 , the method of false
position yields
10
Section 2.2
n
1
2
3
4
5
6
7
8
9
10
(a1 , b1 ) = (4, 3)
3.0294117647
3.0385846418
3.0414199487
3.0422938990
3.0425630528
3.0426459232
3.0426714363
3.0426792907
3.0426817088
3.0426824532
(a1 , b1 ) = (0, 1)
0.7000000000
0.6856023506
0.6852297516
0.6852204836
(a1 , b1 ) = (3, 4)
3.2500000000
3.3277310924
3.3494379958
3.3553114501
3.3568869701
3.3573086084
3.3574213760
3.3574515307
3.3574595940
3.3574617500

x = 0.6852205 and x = 3.3574618.
Using each of these intervals and a convergence tolerance of 106 , the method
of false position yields
n
1
2
3
4
5
(a1 , b1 ) = (0.80, 0.85)

0.8298189963
0.8274663932
0.8272115742
0.8271841999
0.8271812617
(a1 , b1 ) = (1.10, 1.15)

1.1086787135
1.1096053349
1.1097012658
1.1097111652
1.1097121864

and x = 1.1097122.
11
12. In the literature, it is not uncommon to find the method of false position terminated when |pn pn1 | < . Comment on the accuracy of this stopping
condition. Consider the cases 0, 1/2 and 1.
Recall that

|pn pn1 |.
|pn p|
1
If 0, then |pn p| is much smaller than |pn pn1 |. In this case, using a
stopping condition based on |pn pn1 | < is overly pessimistic and will result in
more iterations being performed than are necessary to achieve the desired accuracy.
If 1/2, then /( 1) 1 and |pn p| |pn pn1 |. In this case, a stopping
condition based on |pn pn1 | < is acceptable. Finally, suppose 1. Then
/( 1) and |pn p| is much larger than |pn pn1 |. In this case, using a
stopping condition based on |pn pn1 | < is overly optimistic and will result in
termination before the desired accuracy has been achieved.
13. A storage tank is in the shape of a horizontal cylinder with length L and radius
r. The volume V of fluid in the tank is related to the depth h of the fluid by
the equation

p
rh
V = r2 cos1
(r h) 2rh h2 L.
r
If r = 1 meter, L = 3 meters and V = 7 cubic meters, determine h.
Because the radius of the tank is one meter, we are guaranteed that 0 h 2.
Applying the method of false position to the function

p
f (h) = 3 cos1 (1 h) (1 h) 2h h2 7
with a starting interval of (a1 , b1 ) = (0, 2) and a convergence tolerance of 5 105
yields
n
1
2
3
4
Enclosing Interval
(0.0000000000,2.0000000000)
(0.0000000000,1.4854461355)
(1.3852643791,1.4854461355)
(1.3852643791,1.3916656933)
Approximation
1.4854461355
1.3852643791
1.3916656933
1.3915144154
Thus, if the tank contains 7 cubic meters of fluid, it is filled to a depth of approximately h = 1.39 meters.
14. The equation x2 = 1 cos( 2x) + 2 sin( 2x) has two real roots. One of
them is at x = 0. Determine an interval which contains the other root, and
then approximate this root to three decimal places. This problem arises in the
calculation of the amplitude of the solution to a nonlinear third-order differential equation. See Gottlieb (Simple nonlinear jerk functions with periodic
solutions, American Journal of Physics, 66 (10), 903 - 906, 1998) for details.
12
Section 2.2
Let
f (x) = x2 1 + cos( 2x) 2 sin( 2x).
Because f (1) 1.241 < 0 and f (2) 1.613 > 0, we know the other root of f
lies on the interval (1, 2). Applying the method of false position to f with a starting
interval of (a1 , b1 ) = (1, 2) and a convergence tolerance of 5 104 yields
Enclosing Interval
Approximation
(1.0000000000,2.0000000000) 1.4348284887
(1.4348284887,2.0000000000) 1.5975178950
(1.5975178950,2.0000000000) 1.6369888058
(1.6369888058,2.0000000000) 1.6453392366
(1.6453392366,2.0000000000) 1.6470507507
Thus, the other root of x2 = 1 cos( 2x) + 2 sin( 2x) is approximately x =

1.647.
n
1
2
3
4
5
15. Rework the Depth of Submersion problem to determine the depth to which a
glass marble of radius 2 cm and density 0.040 g/cm3 sinks in water of density
0.998 g/cm3 .
When a spherical object of radius R and density o is placed on the surface of a
fluid of density f , it will sink to a depth h that is a root of the equation
f 3
4
h Rf h2 + R3 o = 0.
3
3
If a glass marble of radius 2 cm and density 0.040 g/cm3 is placed in water of
density 0.998 g/cm3 , then h is a root of the equation
0.998 3
1.28
h 1.996h2 +
= 0.
3
3
The method of false position with a starting interval of (0, 4) and a convergence
tolerance of 5 105 yields
n
1
2
3
4
5
6
7
8
9
10
11
12
Enclosing Interval
(0.0000000000,4.0000000000)
(0.1603206413,4.0000000000)
(0.2968460111,4.0000000000)
(0.3885525915,4.0000000000)
(0.4390256685,4.0000000000)
(0.4632877062,4.0000000000)
(0.4740993195,4.0000000000)
(0.4787428520,4.0000000000)
(0.4807045365,4.0000000000)
(0.4815273818,4.0000000000)
(0.4818714935,4.0000000000)
(0.4820152183,4.0000000000)
Approximation
0.1603206413
0.2968460111
0.3885525915
0.4390256685
0.4632877062
0.4740993195
0.4787428520
0.4807045365
0.4815273818
0.4818714935
0.4820152183
0.4820752160

Thus, the marble will sink to a depth of approximately h = 0.482 cm.
13
Fixed Point Iteration Schemes
2.3
1. Suppose the sequence {pn } is generated by the fixed point iteration scheme
pn = g(pn1 ). Further, suppose that the sequence converges linearly to the
fixed point p.
(a) Show that
g (p)
pn pn1
.
pn1 pn2
(b) Show that

g (p)

|pn pn1 |.
|en |
g (p) 1
Suppose the sequence {pn }, generated by the fixed point iteration scheme pn =
g(pn1 ), converges linearly to the fixed point p.
(a) By the Mean Value Theorem,
pn pn1
= g(pn1 ) g(pn2 )
= g ()(pn1 pn2 )
where is between pn1 and pn2 . As n increases, pn1 and pn2 both tend
toward p, so by the squeeze theorem, also tends toward p. Hence,
pn pn1
= g () g (p).
pn1 pn2
(b) Recall that when a fixed point iteration scheme converges linearly the asymptotic error constant is = g (p). Thus, en g (p)en1 , or en1 en /g (p).
Now,
en
= p pn
= pn pn1 + pn1 p
= pn pn1 + en1 .
Substituting en1 en /g (p) into this last expression, solving for en , and
taking absolute values yields

g (p)

|pn pn1 |.
|en |
g (p) 1
Section 2.3
2. Construct an algorithm for fixed point iteration when the order of convergence
is linear.
Because convergence is linear, we estimate the error using the formulas from Exercise
1. Here is the resulting algorithm. Note its similarity to the algorithm for the method
of false position in Exercise 2 of Section 2.2.
GIVEN:
iteration function g
starting approximation x0
convergence parameter
maximum number of iterations Nmax
STEP
STEP
STEP
STEP
initialize x1 = x0
set x2 = g(x1 )
if ( i > 2 )
set gp = (x2 x1 )/(x1 x0 )
set errest = |gp(x2 x1 )/(gp 1)|
if ( errest < ) OUTPUT x2
end
assign the value of x1 to x0
assign the value of x2 to x1
end
maximum number of iterations exceeded
1:
2:
3:
4:
STEP 5:
OUTPUT:
3. Construct an algorithm for fixed point iteration when the order of convergence
is superlinear.
Because convergence is superlinear, iteration is terminated when |pn pn1 | falls
below the specified convergence tolerance .
GIVEN:
STEP
STEP
STEP
STEP
iteration function g
1:
2:
3:
4:
for iter from 1 to Nmax

compute x1 = g(x0 )
if |x1 x0 | < , OUTPUT x1
copy the value of x1 to x0
end
OUTPUT: maximum number of iterations has been exceeded
4. In the literature, it is not uncommon to find fixed point iteration terminated
when |pn pn1 | < , even when convergence is only linear. Comment on the
accuracy of this stopping condition when convergence is linear. Consider the

cases g (p) 0, g (p) 1/2 and g (p) 1.
Recall that when convergence is linear

g (p)
|pn pn1 |.
|pn p|
g (p) 1
If g (p) 0, then |pn p| is much smaller than |pn pn1 |. In this case, using a
stopping condition based on |pn pn1 | < is overly pessimistic and will result in
more iterations being performed than are necessary to achieve the desired accuracy.
If g (p) 1/2, then g (p)/(g (p) 1) 1 and |pn p| |pn pn1 |. In this
case, a stopping condition based on |pn pn1 | < is acceptable. Finally, suppose
g (p) 1. Then g (p)/(g (p)1) and |pn p| is much larger than |pn pn1 |.
In this case, using a stopping condition based on |pn pn1 | < is overly optimistic
and will result in termination before the desired accuracy has been achieved.
5. Consider the function g(x) = cos x.
(a) Graphically verify that this function has a unique fixed point on the real
line.
(b) Can we prove that the fixed point is unique using the theorems of this
section? Why or why not?
(c) What order of convergence do we expect from the fixed point iteration
scheme pn = g(pn1 ) = cos(pn1 )? Why?
(d) Perform seven iterations starting from p0 = 0. Verify that the appropriate
error estimate is valid. To ten decimal places, the fixed point is x
0.7390851332.
(a) The figure below displays the graphs of y = cos x and y = x for 1.5 x
1.5. There is clearly one point of intersection between the graphs, and hence
one fixed point for cos x, between x = 0.5 and x = 1. Because | cos x| 1 for
all x, it follows there can be no points of intersection between y = cos x and
y = x for any x satisfying |x| > 1. Thus, the fixed point shown in the figure
is the unique fixed point of cos x on the entire real line.
Section 2.3
(b) Because there is no k < 1 such that |g (x)| k for all real numbers x, we
cannot use the theorems of this section to prove that the fixed point is unique.
(c) In the figure from part (a), we see that at the point of intersection between
the graphs of y = cos x and y = x, the tangent line to the graph of y = cos x
is not horizontal. In other words, if p denotes the fixed point of g(x) = cos x,
then g (p) 6= 0. Since g (p) 6= 0, convergence of the sequence generated by
the fixed point iteration scheme pn = g(pn1 ) will be linear.
(d) Because we expect linear convergence, the appropriate error estimate is

g (p)
|pn pn1 |
|en |
g (p) 1
where
g (p)
pn pn1
.
pn1 pn2
Here are the results of seven iterations of the fixed point scheme starting from
p0 = 0.
n
1
2
3
4
5
6
7
pn
1.0000000000
0.5403023059
0.8575532158
0.6542897905
0.7934803587
0.7013687736
0.7639596829
|pn p|
0.2609148668
0.1987828273
0.1184680826
0.0847953427
0.0543952255
0.0377163596
0.0248745497
6. Consider the function g(x) = 1 + x 18 x3 .
theoretical
error estimate
0.129542854379
0.079375374087
0.056574064402
0.036681647749
0.025323586019
(a) Analytically verify that this function has a unique fixed point on the real
line.
(b) Can we prove that the fixed point is unique using the theorems of this
section? Why or why not?
(c) What order of convergence do we expect from the fixed point iteration
scheme pn = g(pn1 )? Why?
(d) Perform seven iterations starting from p0 = 0. Verify that the appropriate
error estimate is valid.
(a) The equation 1+x 18 x3 = x is equivalent to x3 = 8. The only real solution of
this last equation is x = 2; therefore, the only fixed point of g(x) = 1+x 18 x3
on the real line is x = 2.
(b) Because there is no k < 1 such that |g (x)| k for all real numbers x, we
cannot use the theorems of this section to prove that the fixed point is unique.
(c) Note that

3 2
1
g (2) = 1 x
= 6= 0.
8
2
x=2
Since g (2) 6= 0, convergence of the sequence generated by the fixed point

iteration scheme pn = g(pn1 ) will be linear.
(d) Because we expect linear convergence, the appropriate error estimate is

g (p)
|pn pn1 |
|en |
g (p) 1
where
g (p)
pn pn1
.
pn1 pn2
Here are the results of seven iterations of the fixed point scheme starting from
p0 = 0.
n
1
2
3
4
5
6
7
pn
1.0000000000
1.8750000000
2.0510253906
1.9725180057
2.0131771467
1.9932809128
2.0033257219
|pn p|
1.0000000000
0.1250000000
0.0510253906
0.0274819943
0.0131771467
0.0067190872
0.0033257219
theoretical
error estimate
0.044329132602
0.024214600570
0.013872735932
0.006537159153
0.003369895664
7. Consider the function g(x) = 2x(1 x), which has fixed points at x = 0 and at
x = 1/2.
(a) Why should we expect that fixed point iteration, starting even with a value
very close to zero, will fail to converge toward x = 0?
Section 2.3
(b) Why should we expect that fixed point iteration, starting with p0 (0, 1)
will converge toward x = 1/2? What order of convergence should we
expect?
(c) Perform seven iterations starting from an arbitrary p0 (0, 1) and numerically confirm the order of convergence.
(a) Note that g (0) = 2 > 1. Thus, for x 0,
g(x) g(0) + g (0)x = 2x.
If we choose a starting value very close to 0, it follows that each new term in
the sequence pn = g(pn1 ) will be roughly twice the previous term. In other
words, the terms in the sequence will be moving further from 0, not converging
toward 0.
(b) Due to the symmetry of g about x = 1/2, we can restrict attention to p0
(0, 1/2). Since |g (x)| < 1 for 1/4 < x < 1/2, it follows from the theorems of
this section that the iteration will converge for any p0 (1/4, 1/2). Finally,
since g (x) 1 for 0 < x 1/4, it follows that for any p0 (0, 1/4), pn
will eventually lie in (1/4, 1/2) and the iteration will converge to 1/2. Since
g (1/2) = 0, but g (1/2) 6= 0, convergence will be quadratic.
(c) Here are the results of seven iterations of the fixed point iteration scheme
starting from p0 = 0.1. From the values in the fourth column of the table
below, we observe that the ratio
|pn p|
|pn1 p|2
approaches a constant, thereby confirming that convergence is quadratic.
n
0
1
2
3
4
5
6
7
pn
0.1000000000000000
0.1800000000000000
0.2952000000000000
0.4161139200000000
0.4859262511644672
0.4996038591874287
0.4999996861449132
0.4999999999998030
8. Verify that x =
|pn p|
4.000 101
3.800 101
2.048 101
8.389 102
1.407 102
3.961 104
3.139 107
1.970 1013
|pn p|/|pn1 p|2

2.375
1.418
2.000
1.999
2.000
2.001
1.999
a is a fixed point of the function

g(x) =
1
a
x+
.
2
x
Use the techniques of this section to determine the order of convergence

and the
asymptotic error constant of the sequence pn = g(pn1 ) toward x = a.
Let
g(x) =
and note that
1
g( a) =
2
1
a
x+
,
2
x
a
a+
a
1
( a + a) = a.
2
Thus, x = a is a fixed point of g. To determine the order of convergence and the

asymptotic error constant for the sequence
pn = g(pn1 ), we need to examine the
values of the derivatives of g at x = a. Now
1
a
1
a
g (x) =
1 2
so
g ( a) =
1
= 0.
2
x
2
a
The order of convergence is therefore at least 2. Because
g (x) =
a
x3
and
1
g ( a) = =
6 0,
a
the order of convergence is exactly 2. Furthermore, the asymptotic error constant

is
|g ( a)|
1
=
= .
2!
2 a
9. Verify that x =
a is a fixed point of the function

g(x) =
x3 + 3xa
.
3x2 + a
Use the techniques of the this section to determine the order of convergence
and
the asymptotic error constant of the sequence pn = g(pn1 ) toward x = a.
Let
g(x) =
x3 + 3xa
,
3x2 + a
and note that
a a + 3a a
4a a
g( a) =
=
= a.
3a + a
4a
Thus, x = a is a fixed point of g. To determine the order of convergence and the

asymptotic error constant for the sequence
pn = g(pn1 ), we need to examine the
values of the derivatives of g at x = a. Now

g (x) =
3(x4 2x2 a + a2 )
(3x2 + a)2
so
3(a2 2a2 + a2 )
g ( a) =
= 0.
(4a)2

g (x) =
48xa(x2 a)
(3x2 + a)3
and
48a a(a a)
g ( a) =
= 0,
(4a)3
Section 2.3
the order of convergence is at least 3. Finally,

g (x) =
and
48a(9x4 18x2 a + a2 )
(3x2 + a)4
48a(9a2 18a2 + a2 )
3
g ( a) =
=
,
(4a)4
2a
so the order of convergence is exactly 3. Furthermore, the asymptotic error constant

is
|g ( a)|
1
=
=
.
3!
4a
10. Verify that x = 1/a is a fixed point of the function g(x) = x(2 ax). Use the
techniques of the this section to determine the order of convergence and the
asymptotic error constant of the sequence pn = g(pn1 ) toward x = 1/a.
Let g(x) = x(2 ax), and note that

1
1
1
1
g
=
2a
= .
a
a
a
a
Thus, x = a1 is a fixed point of g. To determine the order of convergence and the
asymptotic error constant for the sequence pn = g(pn1 ), we need to examine the
values of the derivatives of g at x = a1 . Now
g (x) = 2 2ax
so

1
1
g
= 2 2a = 0.
a
a

1
g (x) = 2a
and
g
= 2a 6= 0,
a
the order of convergence is exactly 2. Furthermore, the asymptotic error constant
is

|g a1 |
=
= a.
2!
2
11. Consider the function g(x) = ex .

(a) Prove that g has a unique fixed point on the interval [0, 1].
(b) With a starting approximation of p0 = 0, use the iteration scheme pn =
2
epn1 to approximate the fixed point on [0, 1] to within 5 107 .
k
(c) Use the theoretical error bound |pn p| 1k
|p1 p0 | to obtain a theoretical bound on the number of iterations needed to approximate the fixed
point to within 5 107 . How does the number of iterations performed in
part (b) compare with the theoretical bound?
n
(a) Let g(x) = ex . We will proceed by showing that g is continuous on [0, 1],
maps [0, 1] to [0, 1] and there exists a k < 1 such that |g (x)| k for all
x [0, 1]. First, note that g is the composition of the functions ex and
2
x2 , both of which are continuous on [0, 1]. Consequently, g(x) = ex is
continuous on [0, 1]. Next, we see that
2
g (x) = 2xex < 0

for all x (0, 1). Hence, g is decreasing on (0, 1). Combining this fact
with g(0) = 1 and g(1) = e1 0.368, it then follows that for x [0, 1],
g(x) [e1 , 1] [0, 1]. Finally, we find
2
g (x) = (4x2 2)ex = 0
when x =
2/2. Because
g (0) = 0,
2
= 2e1/2 0.858,
2
and g (1) = 2e1 0.736,

we find that |g (x)| 2e1/2 for all x
1/2
[0, 1]. Thus, we take k = 2e
. Having established that g is continuous
on [0, 1], maps [0, 1] to [0, 1] and there exists a k < 1 such that |g (x)| k
for all x [0, 1], we conclude that g has a unique fixed point on the interval
[0, 1].
(b) With a starting approximation of p0 = 0 and a convergence tolerance of =
2
5 107 , fixed point iteration using g(x) = ex yields p84 = 0.6529181524
with an error estimate of 4.880 107 .
(c) In part (a) we found k = 2e1/2 . With p0 = 0, it follows that p1 = g(p0 ) =

e0 = 1. Solving the equation
kn
|p1 p0 | 5 107
1k
for n yields n 107.28, or, since n must be an integer, n 108. As these

calculations were carried out using
k = max |g (x)|,
x[0,1]
we see that the upper bound on the number of iterations needed to guarantee
an absolute error less than 5 107 is n = 108. In part (b), we found that
only 84 iterations were needed to achieve the prescribed level of accuracy,
confirming the theoretical upper bound.
10
Section 2.3
12. Repeat Exercise 11 for the function g(x) =
1
2
cos x.
(a) Let g(x) = 12 cos x. We will proceed by showing that g is continuous on [0, 1],
maps [0, 1] to [0, 1] and there exists a k < 1 such that |g (x)| k for all
x [0, 1]. First, note that g is a constant multiple of the function cos x,
which is continuous on [0, 1]. Consequently, g(x) = 12 cos x is continuous on
[0, 1]. Next, because (0, 1) (0, /2), we see that
1
g (x) = sin x < 0
2
for all x (0, 1). Hence, g is decreasing on (0, 1). Combining this fact with
g(0) = 12 and g(1) = 12 cos 1 0.270, it then follows that for x [0, 1],
g(x) [ 21 cos 1, 12 ] [0, 1]. Finally, we find
1
g (x) = cos x < 0
2
for all x [0, 1]. Because
g (0) = 0
and
1
g (1) = sin 1 0.421,
2
we find that |g (x)| 12 sin 1 for all x [0, 1]. Thus, we take k = 12 sin 1.
Having established that g is continuous on [0, 1], maps [0, 1] to [0, 1] and there
exists a k < 1 such that |g (x)| k for all x [0, 1], we conclude that g has
a unique fixed point on the interval [0, 1].
(b) With a starting approximation of p0 = 0 and a convergence tolerance of =
5 107 , fixed point iteration using g(x) = 21 cos x yields p9 = 0.4501838716
with an error estimate of 2.603 107 .
(c) In part (a) we found k = 12 sin 1. With p0 = 0, it follows that p1 = g(p0 ) =

1
1
2 cos 0 = 2 . Solving the equation
kn
|p1 p0 | 5 107
1k

k = max |g (x)|,
x[0,1]
an absolute error less than 5 107 is n = 17. In part (b), we found that only
9 iterations were needed to achieve the prescribed level of accuracy, confirming
the theoretical upper bound.
13. Repeat Exercise 11 for the function g(x) = 13 (2 ex + x2 ).
11
(a) Let g(x) = 13 (2 ex + x2 ). We will proceed by showing that g is continuous

on [0, 1], maps [0, 1] to [0, 1] and there exists a k < 1 such that |g (x)| k
for all x [0, 1]. First, note that g is a constant multiple of the sum of the
functions 2, ex , and x2 , all of which are continuous on [0, 1]. Consequently,
2
g(x) = ex is continuous on [0, 1]. Next, we see that
g (x) =
1
(2x ex ) < 0
3
for all x (0, 1). Hence, g is decreasing on (0, 1). Combining this fact with
g(0) = 13 and g(1) = 1 13 e 0.0939, it then follows that for x [0, 1],
g(x) [1 13 e, 13 ] [0, 1]. Finally, we find
g (x) =
1
(2 ex ) = 0
3
when x = ln 2. Because
1
g (0) = ,
3
g (ln 2) =
2
(ln 2 1) 0.205,
3
and g (1) = 13 (2 e) 0.239, we find that |g (x)| 13 for all x [0, 1].
Thus, we take k = 13 . Having established that g is continuous on [0, 1], maps
[0, 1] to [0, 1] and there exists a k < 1 such that |g (x)| k for all x [0, 1],
we conclude that g has a unique fixed point on the interval [0, 1].
(b) With a starting approximation of p0 = 0 and a convergence tolerance of
= 5 107 , fixed point iteration using g(x) = 13 (2 ex + x2 ) yields p10 =
0.2575298907 with an error estimate of 3.947 107 .
(c) In part (a) we found k = 13 . With p0 = 0, it follows that p1 = g(p0 ) =

1
1
0
3 (2 e + 0) = 3 . Solving the equation
kn
|p1 p0 | 5 107
1k
k = max |g (x)|,
x[0,1]
an absolute error less than 5 107 is n = 13. In part (b), we found that
only 10 iterations were needed to achieve the prescribed level of accuracy,
confirming the theoretical upper bound.
14. The function f (x) = ex + x2 x 4 has a unique zero on the interval (1, 2).
Create three different iteration functions corresponding to this function, and
compare their convergence properties for approximating the zero on (1, 2). Use
the same starting approximation, p0 , for each iteration function.
12
Section 2.3
Answers will of course vary. Here are two possibilities. If we rearrange the equation
ex + x2 x 4 = 0 as
ex = 4 + x x2
x = ln(4 + x x2 )
we may take g(x) = ln(4 + x x2 ). With p0 = 1 and a convergence tolerance
of 5 107 , fixed point iteration yields p16 = 1.2886775801. Alternately, if we
rearrange ex + x2 x 4 = 0 as
x2
x =
4 + x ex
4 + x ex
x
we may take g(x) = (4+xex )/x. Again using p0 = 1 and a convergence tolerance
of 5 107 , fixed point iteration fails to achieve convergence after 100 iterations.
15. Repeat Exercise 14 for the function f (x) = x3 x2 10x + 7 on the interval
(0, 1).
x3 x2 10x + 7 = 0 as
= x3 x2 + 7
x3 x2 + 7
x =
10
10x
we may take g(x) = (x3 x2 + 7)/10. With p0 = 1 and a convergence tolerance

rearrange the equation x3 x2 10x + 7 = 0 as
x3
= x2 + 10x 7
x2 + 10x77
x =
x2
we may take g(x) = (x2 + 10x 7)/x2 . Again using p0 = 1 and a convergence
tolerance of 5 107 , fixed point iteration yields p23 = 3.3574628337, which is
outside the interval (0, 1).
16. Repeat Exercise 14 for the function f (x) = 1.05 1.04x + ln x on the interval
(1, 2).
1.05 1.04x + ln x = 0 as
1.04x
x =
1.05 + ln x
1.05 + ln x
1.04
13
we may take g(x) = (1.05 + ln x)/1.04. With p0 = 1 and a convergence tolerance

rearrange the equation 1.05 1.04x + ln x = 0 as
ln x = 1.04x 1.05
x = e1.04x1.05
we may take g(x) = e1.04x1.05 . Again using p0 = 1 and a convergence tolerance
of 5 107 , fixed point iteration yields p92 = 0.8271813610, which is outside the
interval (1, 2).
Newtons Method
2.4
Newtons Method
1. Each of the following equations has a root on the interval (0, 1). Perform Newtons method to determine p4 , the fourth approximation to the location of the
root.
(a)
ln(1 + x) cos x = 0
(b)
x5 + 2x 1 = 0
x
(d)
cos x x = 0
(c)
e x=0
(a) Let f (x) = ln(1 + x) cos x. Then f (x) =
iterations of Newtons method yield
1
1+x
+ sin x. With p0 = 0, four
p1
= p0
ln(1 + p0 ) cos p0
= 1.0000000000;
1
1+p0 + sin p0
p2
= p1
ln(1 + p1 ) cos p1
= 0.8860617364;
1
1+p1 + sin p1
p3
= p2
ln(1 + p2 ) cos p2
= 0.8845109403; and
1
1+p2 + sin p2
p4
= p3
ln(1 + p3 ) cos p3
= 0.8845106162.
1
1+p3 + sin p3
(b) Let f (x) = x5 + 2x 1. Then f (x) = 5x4 + 2. With p0 = 0, four iterations

of Newtons method yield
p1
p2
p3
p4
p50 + 2p0 1
5p40 + 2
p5 + 2p1 1
= p1 1 4
5p1 + 2
5
p + 2p2 1
= p2 2 4
5p2 + 2
5
p + 2p3 1
= p3 3 4
5p3 + 2
= p0
= 0.5000000000;
= 0.4864864865;
= 0.4863890407; and
= 0.4863890359.
(c) Let f (x) = ex x. Then f (x) = ex 1. With p0 = 0, four iterations

p1
= p0
ep0 p0
= 0.5000000000;
ep0 1
Section 2.4
p2
p3
p4
ep1 p1
= 0.5663110032;
ep1 1
ep2 p2
= p2
= 0.5671431650; and
ep2 1
ep3 p3
= p3
= 0.5671432904.
ep3 1
= p1
(d) Let f (x) = cos x x. Then f (x) = sin x 1. With p0 = 0, four iterations
p1
p2
p3
p4
cos p0 p0
sin p0 1
cos p1 p1
= p1
sin p1 1
cos p2 p2
= p2
sin p2 1
cos p3 p3
= p3
sin p3 1
= p0
= 1.0000000000;
= 0.7503638678;
= 0.7391128909; and
= 0.7390851334.
2. Construct an algorithm for Newtons method. Is it necessary to save all calculated terms in the sequence {pn }?
Because convergence is quadratic, iteration is terminated when |pn pn1 | falls
below the specified convergence tolerance . Note that only the two most recent
terms in the sequence are needed.
GIVEN:
STEP
STEP
STEP
STEP

1:
2:
3:
4:

compute x1 = x0 f (x0 )/f (x0 )
end

(a) Perform five (5) iterations of Newtons method.
(b) For n 1, compare |pn pn1 | with |pn1 p| and |pn p|.
(c) For n 1, compute the ratio |pn p|/|pn1 p|2 and show that this value
approaches |f (p)/2f (p)|.
Newtons Method
3. The
x = 3.
Let f (x) = x3 + x2 3x 3. Then f (x) = 3x2 + 2x 3, f (x) = 6x + 2 and

f (3)
6 3+2

0.655.
=

2f ( 3) 12 + 4 3
With p0 = 1, the first five iterations of Newtons method yield
n
0
1
2
3
4
5
pn
1.00000000
3.00000000
2.20000000
1.83015075
1.73779545
1.73207229
|pn pn1 |
|pn p|
|pn1 p|
|pn p|/|pn1 p|2
2.00000000
0.80000000
0.36984925
0.09235530
0.00572316
1.26794919
0.46794919
0.09809995
0.00574465
0.00002148
0.73205081
1.26794919
0.46794919
0.09809995
0.00574465
2.366
0.291
0.448
0.597
0.651
Note that for n 3, |pn pn1 | provides an excellent estimate for |pn1 p| and is
substantially larger than |pn p|. Furthermore, the ratio |pn p|/|pn1 p|2 appears
to be approaching the value of |f (p)/2f (p)|, confirming quadratic convergence
of the sequence.
7
3.
Let f (x) = x7 3. Then f (x) = 7x6 , f (x) = 42x5 and

f (
7
3)
3

2.564.

=
7
7
2f ( 3)
3
With p0 = 1, the first five iterations of Newtons method yield

n
0
1
2
3
4
5
pn
1.00000000
1.28571429
1.19691682
1.17168905
1.16993871
1.16993081
|pn pn1 |
|pn p|
|pn1 p|
|pn p|/|pn1 p|2
0.28571429
0.08879746
0.02522777
0.00175034
0.00000790
0.11578347
0.02698601
0.00175824
0.00000790
1.598 1010
0.16993081
0.11578347
0.02698601
0.00175824
0.00000790
4.010
2.013
2.414
2.554
2.564
Note that for all n, |pn pn1 | provides a reasonable estimate for |pn1 p| and is
of the sequence.
13.
Section 2.4
Let f (x) = x3 13. Then f (x) = 3x2 , f (x) = 6x, and

f (
3
13)
1

0.425.
=

3
2f ( 3 13)
13
The following data was generated using MAPLE, with the Digits parameter set to
25.
n
0
1
2
3
4
5
pn
3.00000000
2.48148148
2.35804119
2.35135374
2.35133469
2.35133469
|pn pn1 |
|pn p|
|pn1 p|
|pn p|/|pn1 p|2
0.51851852
0.12344029
6.687 103
1.906 105
1.544 1010
0.13014679
6.707 103
1.906 105
1.544 1010
1.014 1020
0.64866531
0.13014679
6.707 103
1.906 105
1.544 1010
0.309
0.396
0.424
0.425
0.425
Note that for all n, |pn pn1 | provides a reasonable estimate for |pn1 p| and is
of the sequence.
Let f (x) =
1
x
37. Then f (x) = x2 , f (x) = 2x3 and

f (1/37)

2f (1/37) = 37.
With p0 = 0.01, the first five iterations of Newtons method yield

n
0
1
2
3
4
5
pn
0.01000000
0.01630000
0.02276947
0.02635634
0.02701038
0.02702701
|pn pn1 |
|pn p|
|pn1 p|
|pn p|/|pn1 p|2
0.00630000
0.00646947
0.00358687
0.00065405
0.00001663
0.010727027
0.004257557
0.000670691
0.000016644
1.025 108
0.017027027
0.010727027
0.004257557
0.000670691
0.000016644
37.000
37.000
37.000
37.000
37.000
Note that for all n, |pn pn1 | provides a reasonable estimate for |pn1 p| and
is larger than |pn p|. Furthermore, the ratio |pn p|/|pn1 p|2 appears to be
approaching the value of |f (p)/2f (p)|, confirming quadratic convergence of the
sequence.
7. Show that when Newtons method is applied to the equation x2 a = 0, the
resulting iteration function is g(x) = 12 x + xa .
Newtons Method
Let f (x) = x2 a. Then f (x) = 2x and the Newton method iteration function is
f (x)
f (x)
x2 a
= x
2x
x2 + a
1
a
=
=
x+
.
2x
2
x
g(x) = x
8. Show that when Newtons method is applied to the equation 1/x a = 0, the
resulting iteration function is g(x) = x(2 ax).
Let f (x) =
is
1
x
a. Then f (x) = x2 and the Newton method iteration function

f (x)
f (x)
1
a
= x x 2
x
= x + x ax2 = x(2 ax).
g(x) = x
Perform three iterations of Newtons method to approximate this zero, using
p0 = 4. Determine the absolute error in each of the computed approximations.
What is the apparent order of convergence? What explanation can you provide
for this behavior? (NOTE: If you have access to MAPLE, perform five iterations
with the Digits parameter set to at least 100.)
Let f (x) = sin x and p0 = 4. Using MAPLE, with the Digits parameter set to
100, Newtons method yields
n
1
2
3
4
5
|pn p|
2.994 101
9.280 103
2.664 107
6.304 1021
8.351 1062
|pn p|/|pn1 p|3

0.34577
0.33334
0.33344
0.33334
Because the ratio in the third column of the table appears to be approaching a
constant, convergence is of order three. The order of convergence for this specific
problem is better than the expected quadratic convergence for Newtons method
because f () = sin = 0; thus,
|en |
f ()
=
= 0,
2
n |en1 |
2f ()
lim
Section 2.4
which implies that convergence is better than quadratic.

10. (a) Verify that the equation x4 18x2 +45 = 0 has a root on the interval (1, 2).
Next, perform three iterations of Newtons
method, with p0 = 1. Given
that the exact value of the root is x = 3, compute the absolute error in the
approximations just obtained. What is the apparent order of convergence?
What explanation can you provide for this behavior? (NOTE: If you have
access to MAPLE, perform five iterations with the Digits parameter set
to at least 100.)
(3, 4). Perform five iterations of Newtons method, and compute the abso
lute error in each approximation. The exact value of the root is x = 15.
What is the apparent order of convergence in this case?
(c) What explanation can you provide for the different convergence behavior
between parts (a) and (b)?
(a) Let f (x) = x4 18x2 + 45. Then f (1) = 28 > 0 and f (2) = 11 < 0,
so the Intermediate Value Theorem guarantees the existence of a root on the
interval (1, 2). With p0 = 1 and using MAPLE, with the Digits parameter
set to 100, Newtons method yields
n
1
2
3
4
5
|pn p|
1.429 101
1.014 103
3.480 1010
1.404 1029
9.229 1088
|pn p|/|pn1 p|3

0.34730
0.33326
0.33333
0.33333
Because the ratio in the third column of the table appears to be approaching a
constant, convergence is of order three. The order of convergence for this specific problem is better
than the expected quadratic convergence for Newtons
method because f ( 3) = 0; thus,
|en |
f ( 3)
= 0,
lim
=
n |en1 |2
2f ( 3)
which implies that convergence is better than quadratic.
(b) Let f (x) = x4 18x2 + 45. Then f (3) = 36 < 0 and f (4) = 13 > 0,
so the Intermediate Value Theorem guarantees the existence of a root on the
interval (3, 4). With p0 = 3.5, the following table summarizes the results of
five iterations of Newtons method.
n
pn
|pn p|
|pn p|/|pn1 p|2
1
1 4.0590659341 1.861 10
2 3.8951971117 2.221 102
0.641521
3 3.8733563066 3.730 104
0.755820
4 3.8729834539 1.077 107
0.774274
5 3.8729833462 8.985 1015
0.774601
Newtons Method
Because the ratio in the fourth column of the table appears to be approaching
a constant, convergence is of order two, as expected.
(c) In part (b), f ( 15) 6= 0, so the error analysis from the text holds, and
Newtons
method exhibits quadratic convergence. On the other hand, in part
(a), f ( 3) = 0 so convergence is faster than quadratic. We can expect this
to be true with Newtons method whenever f (p) = 0.
11. The function f (x) = 27x4 + 162x3 180x2 + 62x 7 has a zero at x = 1/3.
Perform ten iterations of Newtons method on this function, starting with p0 =
0. What is the apparent order of convergence of the sequence of approximations?
What is the multiplicity of the zero at x = 1/3? Would the sequence generated
by the bisection method converge faster?
Let f (x) = 27x4 + 162x3 180x2 + 62x + 7 and p0 = 0. The following table
summarizes the results of ten iterations of Newtons method.
n
1
2
3
4
5
6
7
8
9
10
pn
0.1129032258
0.1871468695
0.2362083272
0.2687288261
0.2903276528
0.3046911326
0.3142510338
0.3206173081
0.3248585295
0.3276845680
|pn p|
0.2204301075
0.1461864638
0.0971250061
0.0646045072
0.0430056805
0.0286422007
0.0190822995
0.0127160252
0.0084748038
0.0056487653
|pn p|/|pn1 p|
0.663187
0.664391
0.665169
0.665676
0.666010
0.666230
0.666378
0.666466
0.666536
Convergence is clearly linear with an asymptotic error constant of = 2/3 = 11/3;

hence, the multiplicity of the zero at x = 1/3 is three. Because the bisection
method generates is linearly convergent with an asymptotic error constant of 1/2,
the sequence generated by the bisection method would converge faster.
12. Repeat Exercise 11 for the function
f (x) =
x
500
1 + x2
841

21x
1
,
125
which has a zero at x = 2.5. Start Newtons method with p0 = 2.

Let
f (x) =
x
500
2
1+x
841
21x
125
and p0 = 2. The following table summarizes the results of ten iterations of Newtons
method.
Section 2.4
n
1
2
3
4
5
6
7
8
9
10
pn
2.2600472810
2.3803798070
2.4401402930
2.4700436310
2.4850136160
2.4925046410
2.4962520120
2.4981261390
2.4990646340
2.4995334410
|pn p|
0.2399527190
0.1196201930
0.0598597070
0.0299563690
0.0149863840
0.0074953590
0.0037479880
0.0018738610
0.0009353660
0.0004665590
|pn p|/|pn1 p|
0.498516
0.500415
0.500443
0.500274
0.500145
0.500041
0.499965
0.499165
0.498798
Convergence is clearly linear with an asymptotic error constant of = 1/2 = 11/2;

hence, the multiplicity of the zero at x = 2.5 is two. Here, there is no comparison
with the bisection method because the bisection method cannot be used to locate
a root of even multiplicity.
13. The function f (x) = x3 + 2x2 3x 1 has a zero on the interval (1, 0).
Let f (x) = x3 + 2x2 3x 1. With an initial approximation of p0 = 0 and a
convergence tolerance of 5 105 , Newtons method yields
n
1
2
3
4
pn
0.3333333333
0.2870370370
0.2864621616
0.2864620650

x = 0.286462.
14. For each of the functions given below, use Newtons method to approximate all
real roots. Use an absolute tolerance of 106 as a stopping condition.
(a)
f (x) = ex + x2 x 4
(b)
f (x) = x3 x2 10x + 7
(c)
f (x) = 1.05 1.04x + ln x

Newtons Method
p0 = 2 and p0 = 1 and a convergence tolerance of 106 , Newtons method
yields
n
1
2
3
4
5
p0 = 2
1.5610519106
1.5079230514
1.5070996826
1.5070994841
p0 = 1
1.3447071068
1.2903157401
1.2886794153
1.2886779668
1.2886779668

and x = 1.2886780.
f (3) > 0, f (0) > 0, f (1) < 0, f (3) < 0 and f (4) > 0. Therefore, the three
real zeros of f lie on the intervals (4, 3), (0, 1) and (3, 4). Using p0 = 4,
p0 = 0, and p0 = 3 and a convergence tolerance of 106 , Newtons method
yields
n
1
2
3
4
5
p0 = 4
3.2826086957
3.0638181359
3.0428698316
3.0426827991
3.0426827843
p0 = 0
0.7000000000
0.6851963746
0.6852202473
0.6852202474
p0 = 3
3.4545454545
3.3620854211
3.3574738296
3.3574625370
3.3574625369

x = 0.6852202 and x = 3.3574625.
10
Section 2.4
Using p0 = 0.80 and p0 = 1.10 and a convergence tolerance of 106 , Newtons
method yields
n
1
2
3
4
p0 = 0.80
0.8244931015
0.8271502367
0.8271809044
0.8271809085
p0 = 1.10
1.1100083179
1.1097125596
1.1097123039

and x = 1.1097123.
15. An equation of state relates the volume V occupied by one mole of a gas to
the instantaneous pressure P and the Kelvin absolute temperature T of the gas.
The Redlich-Kwong equation of state is given by
P =
RT
a
,
V b V (V + b) T
where a and b are related to the critical temperature Tc and the critical pressure
Pc by the equations
!

5/2
R 2 Tc
RTc
a = 0.42747
and b = 0.08664
.
Pc
Pc
The coefficient R is a universal constant equal to 0.08206.
(a) Determine the volume of one mole of carbon dioxide at a temperature of
T = 323.15K and a pressure of one atmosphere. For carbon dioxide, Tc =
304.2K and Pc = 72.9 atmospheres.
(b) Determine the volume of one mole of ammonia at a temperature of T =
450K and a pressure of 56 atmospheres. For ammonia, Tc = 405.5K and
Pc = 111.3 atmospheres.
Newtons Method
11
Let
f (V ) =
Then
RT
a
P.
V b V (V + b) T
f (V ) =
RT
a(2V + b)
.
+
(V b)2
V 2 (V + b)2 T
(a) For carbon dioxide, Tc = 304.2 K and Pc = 72.9 atmospheres, so

a = 0.42747
0.082062 304.25/2
= 63.72930208
72.9
and
0.08206 304.2
= 0.029667546.
72.9
The initial approximation for the volume is taken from the ideal gas law:
b = 0.08664
V0 =
nRT
P
=
=
(1 mole)(0.08206 atm liter/mole K)(323.15 K)

1 atmosphere
26.517689 liters.
With a convergence tolerance of 5 107 , Newtons method yields

n
1
2
3
Vn
26.4130294622
26.4134392885
26.4134392948
Thus, one mole of carbon dioxide at a temperature of 323.15 K and a pressure

of one atmosphere occupies a volume of approximately 26.4134 liters.
(b) For ammonia, Tc = 405.5 K and Pc = 111.3 atmospheres, so
a = 0.42747
0.082062 405.55/2
= 85.634487113
111.3
and
0.08206 405.5
= 0.259027366.
111.3
The initial approximation for the volume is taken from the ideal gas law:
b = 0.08664
V0 =
nRT
P
=
=
(1 mole)(0.08206 atm liter/mole K)(450 K)

56 atmosphere
0.659410714 liters.
With a convergence tolerance of 5 107 , Newtons method yields

n
1
2
3
4
Vn
0.5578759882
0.5695932380
0.5698036385
0.5698037041
12
Section 2.4
Thus, one mole of ammonia at a temperature of 450 K and a pressure of 56
atmospheres occupies a volume of approximately 0.5698 liters.
16. In determining the minimum cushion pressure needed to break a given thickness
of ice using an air cushion vehicle, Muller (Ice Breaking with an Air Cushion
Vehicle, in Mathematical Modeling: Classroom Notes in Applied Mathematics,
M.S. Klamkin, editor, SIAM, 1987) derived the equation

2 3
h2
2 h4
h
3
2
2
2
p (1 ) + 0.4h 2 p +
p
= 0,
r
3r4
3r2
where p denotes the cushion pressure, h the thickness of the ice field, r the size
of the air cushion, the tensile strength of the ice, and is related to the width
of the ice wedge. Take = 0.5, r = 40 feet and = 150 pounds per square inch
(psi). Determine p for h = 0.6, 1.2, 1.8, 2.4, 3.0, 3.6 and 4.2 feet.
Let
h2
f (p) = (1 )p + 0.4h 2
r
2
Then
2 h4
p +
p
3r4
2
h2
3r2
3

h2
2 h4
f (p) = 3(1 2 )p2 + 2 0.4h 2 2 p +
.
r
3r4
With a starting approximation of p0 = 0 and a convergence tolerance of 5 107 ,

Newtons method yields
h (feet)
p (psi)
0.6
0.003050
1.2
0.015138
1.8
0.038278
2.4
0.073664
3.0
0.122199
3.6
0.184644
4.2
0.261681
17. A frame structure is composed of two vertical columns and one horizontal beam,
as shown below. The vertical columns are of length L and have modulus of elasticity E and moment of inertia I. The horizontal beam connecting the tops of
the columns is of length L1 with modulus of elasticity E and moment of inertia
I1 . The structure is pinned at the bottom and free to displace laterally at the
top. The buckling load, P, for the structure is given by
P
P = (kL)2
EI
,
L2
where kL is the smallest positive solution of

kL tan kL = 6
I1 L
.
IL1
Suppose E = 30 106 lb/in2 , I = 15.2 in4 ,

L = 144 in, I1 = 9.7 in4 and L1 = 120 in.
Determine the buckling load of the structure.
Newtons Method
13
Let x = kL and define the function

f (x) = x tan x 6
I1 L
.
IL1
Then f (x) = x sec2 x + tan x. With an initial approximation of x0 = 1.5 and a

convergence tolerance of 5 107 , Newtons method yields
n
1
2
3
4
5
6
7
xn
1.4472486988
1.3789408151
1.3208020657
1.2981292481
1.2959152790
1.2958973806
1.2958973794
Thus, kL 1.295897, and the buckling load of the structure is approximately

P = (1.295897)2
30 106 15.2
= 36.930 103 lb.
1442
Secant Method
2.5
Secant Method
1. Each of the following equations has a root on the interval (0, 1). Perform the
secant method to determine p4 , the fourth approximation to the location of the
root.
(a)
ln(1 + x) cos x = 0
(b)
x5 + 2x 1 = 0
x
(d)
cos x x = 0
(c)
e x=0
(a) Let f (x) = ln(1 + x) cos x. With p0 = 0 and p1 = 1, three iterations of the
secant method yield
p2
= p1 (ln(1 + p1 ) cos p1 )
=
p3
= p2 (ln(1 + p2 ) cos p2 )
=
p4
0.8674193918;
0.8842599009; and
= p3 (ln(1 + p3 ) cos p3 )
=
0.8845112004
p1 p0
(ln(1 + p1 ) cos p1 ) (ln(1 + p0 ) cos p0 )
p2 p1
(ln(1 + p2 ) cos p2 ) (ln(1 + p1 ) cos p1 )
p3 p2
(ln(1 + p3 ) cos p3 ) (ln(1 + p2 ) cos p2 )
(b) Let f (x) = x5 + 2x 1. With p0 = 0 and p1 = 1, three iterations of the

secant method yield
p2
= p1 (p51 + 2p1 1)
=
p3
= p2 (p52 + 2p2 1)
=
p4
0.3333333333;
(p52
p2 p1
+ 2p2 1) (p51 + 2p1 1)
(p53
p3 p2
+ 2p3 1) (p52 + 2p2 1)
0.4275618375; and
= p3 (p53 + 2p3 1)
=
p1 p0
(p51 + 2p1 1) (p50 + 2p0 1)
0.4895113751
(c) Let f (x) = ex x. With p0 = 0 and p1 = 1, three iterations of the secant

method yield
p2
= p1 (ep1 p1 )
p1 p0
(ep1 p1 ) (ep0 p0 )
Section 2.5
=
p3
p4
0.6126998368;
p2 p1
(ep2 p2 ) (ep1 p1 )
= 0.5638383892; and
p3 p2
= p3 (ep3 p3 ) p3
(e
p3 ) (ep2 p2 )
= 0.5671703584
= p2 (ep2 p2 )
(d) Let f (x) = cos x x. With p0 = 0 and p1 = 1, three iterations of the secant
method yield
p2
= p1 (cos p1 p1 )
=
p3
p4
0.6850733573;
p1 p0
(cos p1 p1 ) (cos p0 p0 )
p2 p1
= 0.7362989976; and
p3 p2
= p3 (cos p3 p3 )
= 0.7391193619
= p2 (cos p2 p2 )
2. Construct an algorithm for the secant method.

The stopping condition in STEP 3 is justified by the superlinear convergence (
1.618) of the secant method.
GIVEN:
STEP
STEP
STEP
STEP

starting approximations x0 and x1
1:
2:
3:
4:

compute x2 = x1 f (x1 )(x1 x0 )/(f (x1 ) f (x0 ))
end
3. Show that the equation for the secant method can be rewritten as
pn+1 =
f (pn )pn1 f (pn1 )pn

.
f (pn ) f (pn1 )
Secant Method
Explain why this formula is inferior to the one used in the text.
We proceed as follows:
pn+1
pn pn1
f (pn ) f (pn1 )
pn f (pn ) pn f (pn1 ) pn f (pn ) + pn1 f (pn )
f (pn ) f (pn1 )
.
f (pn ) f (pn1 )
= pn f (pn )
=
=
Note that both formulas for calculating pn+1 have the potential for cancellation
error. In the formula
pn pn1
pn+1 = pn f (pn )
,
f (pn ) f (pn1 )
cancellation error will primarily influence the second term, which is essentially just
a correction to pn . This can limit the overall effect of roundoff error on pn+1 . In
the formula
pn+1 =
,
f (pn ) f (pn1 )
however, cancellation error influences the entire calculation. Thus, the latter formula
is inferior to the former because it is more susceptible to roundoff error.
4. Fill in the missing details in the derivation of the error evolution equation
pn+1 p (pn p)(pn1 p)
2f (p)
f (p)
.
n + pn1 2p)
f (p)(p
First subtract the true root, p, from both sides of the recurrence formula for pn1 ,
yielding
pn pn1
pn+1 p = pn p f (pn )
.
(1)
f (pn ) f (pn1 )
Next, approximate the function values f (pn1 ) and f (pn ) by the second degree
Taylor polynomials
f (p)
(pn1 p)2 ,
2
f (p)
f (pn ) f (p)(pn p) +
(pn p)2 ,
2
f (pn1 ) f (p)(pn1 p) +
(2)
(3)
The term f (pn ) f (pn1 ) is then approximately

f (p)
f (pn ) f (pn1 ) f (p)(pn pn1 ) +
(pn p)2 (pn1 p)2
2

f (p)
= (pn pn1 ) f (p) +

(pn + pn1 2p) .
(4)
2
Section 2.5
Substituting (2), (3) and (4) into (1), factoring the term pn p and dividing out
the term pn pn1 yields
"
#
f (p) + f 2(p) (pn p)

pn+1 p (pn p) 1
f (p) + f 2(p) (pn + pn1 2p)

(pn p)(pn1 p)
2f (p)
f (p)
.
n + pn1 2p)
f (p)(p

(a) Perform seven (7) iterations of the secant method.
(b) For n 2, compare |pn pn1 | with |pn1 p| and |pn p|.
(c) For n 2, compute the ratio |pn p|/|pn1 p|1.618 and show that this
0.618
value approaches (|f (p)/2f (p)|)
.
5. The
x = 3.
Let f (x) = x3 + x2 3x 3. Then f (x) = 3x2 + 2x 3, f (x) = 6x + 2 and
!

f (3) 0.618

=

2f ( 3)
!0.618
6 3+2
0.770.
12 + 4 3
With p0 = 1 and p1 = 2, the first seven iterations of the secant method yield
n
0
1
2
3
4
5
6
7
pn
1.00000000
2.00000000
1.57142857
1.70541082
1.73513577
1.73199637
1.73205070
1.73205081
|pn pn1 |
|pn p|
0.42857143
0.13398225
0.02972495
0.00313940
0.00005433
0.00000011
1.606 101
2.664 102
3.085 103
5.444 105
1.098 107
3.193 1012
|pn1 p|
2.679
1.606
2.664
3.085
5.444
1.098
|pn p|/|pn1 p|1.618
101
101
102
103
105
107
1.353
0.513
1.088
0.628
0.871
0.713
Note that for n 3, |pn pn1 | provides an excellent estimate for |pn1 p| and
is substantially larger than |pn p|. Furthermore, the ratio |pn p|/|pn1 p|1.618
appears to be settling toward the value of (|f (p)/2f (p)|)0.618 , confirming the
sequence converges of order 1.618.
7
3.
Secant Method
Let f (x) = x7 3. Then f (x) = 7x6 , f (x) = 42x5 and

!0.618

0.618
f (
7
3)
3

1.790.
=

7
2f ( 7 3)
3
n
0
1
2
3
4
5
6
7
pn
1.00000000
2.00000000
1.01574803
1.03036560
1.25047859
1.13998478
1.16412646
1.17039516
|pn pn1 |
|pn p|
|pn1 p|
|pn p|/|pn1 p|1.618
0.98425197
0.01461756
0.22011299
0.11049380
0.02414168
0.00626869
1.542 101
1.396 101
8.055 102
2.995 102
5.804 103
4.643 104
8.301 101
1.542 101
1.396 101
8.055 102
2.995 102
5.804 103
0.208
2.874
1.949
1.763
1.694
1.928
13.
Let f (x) = x3 13. Then f (x) = 3x2 , f (x) = 6x, and

!0.618

0.618
f (
3
13)
1

= 3
0.589.
3

2f ( 13)
13
n
0
1
2
3
4
5
6
7
pn
2.00000000
3.00000000
2.26315789
2.33050735
2.35214053
2.35132751
2.35133469
2.35133469
|pn pn1 |
|pn p|
0.73684211
0.06734946
0.02163318
0.00081302
0.00000718
2.460 109
8.818 102
2.083 102
8.058 104
7.179 106
2.460 109
7.550 1015
|pn1 p|
6.487
8.818
2.083
8.058
7.179
2.460
101
102
102
104
106
109
|pn p|/|pn1 p|1.618

0.178
1.059
0.423
0.727
0.517
0.642
Section 2.5
Let f (x) =
1
x
37. Then f (x) = x2 , f (x) = 2x3 and

f (1/37) 0.618

= 370.618 9.314.
2f (1/37)
With p0 = 0.01 and p1 = 0.02, the first seven iterations of the secant method yield
n
0
1
2
3
4
5
6
7
pn
0.01000000
0.02000000
0.02260000
0.02587600
0.02683849
0.02701900
0.02702697
0.02702703
|pn pn1 |
|pn p|
|pn1 p|
|pn p|/|pn1 p|1.618
0.00260000
0.00327600
0.00096249
0.00018051
0.00000797
0.00000006
4.427 103
1.151 103
1.885 104
8.029 106
5.601 108
1.664 1011
7.027 103
4.427 103
1.151 103
1.885 104
8.029 106
5.601 108
13.491
7.408
10.729
8.533
9.829
9.004
Perform five iterations of the secant method to approximate this zero, using
p0 = 3 and p1 = 4. Determine the absolute error in each of the computed
approximations. What is the apparent order of convergence? What explanation
can you provide for this behavior? (NOTE: If you have access to MAPLE,
perform seven iterations with the Digits parameter set to at least 100.)
Let f (x) = sin x and take p0 = 3 and p1 = 4. Using MAPLE, with the Digits
parameter set to 100, the secant method yields
n
0
1
2
3
4
5
6
7
|pn p|
1.416 101
8.584 101
1.557 102
2.134 103
7.440 108
5.644 1014
5.206 1029
2.764 1056
|pn p|/|pn1 p|2

0.02113
8.80074
0.01634
10.19759
0.01634
10.19759
Because the ratio in the third column of the table appears to have settled into a
two-cycle, convergence appears to be of order two. The order of convergence for
Secant Method
this specific problem is better than the expected 1.618 for the secant method
because f () = sin = 0; thus,
|en |
=
n |en1 |1.618
lim
f ()
2f ()
0.618
= 0,
which implies that convergence is better than order 1.618.

10. (a) Verify that the equation x4 18x2 + 45 = 0 has a root on the interval
(1, 2). Next, perform five iterations of the secant method,
using p0 = 1
and p1 = 2. Given that the exact value of the root is x = 3, compute the
absolute error in the approximations just obtained. What is the apparent
order of convergence? What explanation can you provide for this behavior?
(NOTE: If you have access to MAPLE, perform seven iterations with the
Digits parameter set to at least 100.)
(3, 4). Perform seven iterations of the secant method, and computethe
absolute error in each approximation. The exact value of the root is 15.
What is the apparent order of convergence in this case? What explanation
can you provide for the different convergence behavior between parts (a)
and (b)?
(a) Let f (x) = x4 18x2 + 45. Then f (1) = 28 > 0 and f (2) = 11 < 0, so the
(1, 2). With p0 = 1, p1 = 2 and using MAPLE, with the Digits parameter
set to 100, the secant method yields
n
2
3
4
5
6
7
|pn p|
1.410 102
1.681 104
5.493 109
2.585 1017
1.300 1034
1.448 1068
|pn p|/|pn1 p|2

0.19642
0.84504
0.19449
0.85694
0.19449
0.85694
Because the ratio in the third column of the table appears to have settled into
a two-cycle, convergence appears to be of order two. The order of convergence
for this specific problem
is better than the expected 1.618 for the secant
method because f ( 3) = 0; thus,
|en |
lim
=
n |en1 |1.618
!0.618
f ( 3)
= 0,
2f ( 3)
which implies that convergence is better than order 1.618.
(b) Let f (x) = x4 18x2 + 45. Then f (3) = 36 < 0 and f (4) = 13 > 0, so the
Section 2.5
(3, 4). With p0 = 3 and p1 = 4, the following table summarizes the results of
seven iterations of the secant method.
n
pn
|pn p|
|pn p|/|pn1 p|1.618
1
2 3.73469388 1.383 10
3.897
3 3.85932813 1.366 102
0.335
4 3.87458114 1.598 103
1.662
5 3.87296633 1.701 105
0.570
6 3.87298333 2.104 108
1.095
7 3.87298335 2.771 1013
0.731
Because the ratio in the fourth column of the table appears to be approaching
a constant, convergence is of order 1.618, as expected.
(c) In part (b), f ( 15) 6= 0, so the error analysis from the text holds, and the
secant method exhibits
convergence of order 1.618. On the other hand,
in part (a), f ( 3) = 0 so convergence is faster. We can expect this to be

true with the secant method whenever f (p) = 0.
11. It was observed that Newtons method provides only linear convergence towards
roots of multiplicity greater than one. How does the secant method perform
under such circumstances? Each of the following functions has a zero at the
specified location. Perform ten iterations of the secant method to locate these
zeros. Does the sequence generated by the secant method converge with order
1.618 or has the order dropped to = 1?
(a) f (x) = x(1 cos x) has a zero at x = 0 use p0 = 1 and p1 = 2
(b) f (x) = 27x4 + 162x3 180x2 + 62x 7 has a zero at x = 1/3

x
500
21x
(c) f (x) = 1+x
has a zero at x = 2.5
2 841 1 125
(a) Let f (x) = x(1 cos x). With p0 = 1 and p1 = 2, ten iterations of the
secant method yield
n
0
1
2
3
4
5
6
7
8
9
10
pn
1.00000000
2.00000000
0.58107634
0.49699597
0.35355090
0.27156471
0.20308551
0.15362583
0.11577717
0.08741103
0.06596265
|pn p|
|pn p|/|pn1 p|
0.58107634
0.49699597
0.35355090
0.27156471
0.20308551
0.15362583
0.11577717
0.08741103
0.06596265
0.291
0.855
0.711
0.768
0.748
0.756
0.754
0.755
0.755
Because the ratio of errors in the fourth column approaches a constant, it is

clear that the order of convergence has dropped to linear.
Secant Method
(b) Let f (x) = 27x4 + 162x3 180x2 + 62x 7. With p0 = 0 and p1 = 1, ten
iterations of the secant method yield
n
0
1
2
3
4
5
6
7
8
9
10
pn
0.00000000
1.00000000
0.09859155
0.13220733
0.18992728
0.22316577
0.25096822
0.27106015
0.28642787
0.29793792
0.30663456
|pn p|
|pn p|/|pn1 p|
0.23474178
0.20112601
0.14340605
0.11016756
0.08236511
0.06227318
0.04690546
0.03539541
0.02669878
0.352
0.857
0.713
0.768
0.748
0.756
0.753
0.755
0.754

(c) Let
x
500
f (x) =
2
1+x
841
21x
1
125
With p0 = 2 and p1 = 3, ten iterations of the secant method yield

n
0
1
2
3
4
5
6
7
8
9
10
pn
2.00000000
3.00000000
46.00000000
2.94526305
2.90154959
2.70634662
2.63456117
2.58094899
2.55037793
2.53099767
2.51917084
|pn p|
|pn p|/|pn1 p|
43.50000000
0.44526305
0.40154959
0.20634662
0.13456117
0.08094899
0.05037793
0.03099767
0.01917084
87.000
0.010
0.902
0.514
0.652
0.602
0.622
0.615
0.618

12. Newtons method approximates the zero of f (x) = x3 + 2x2 3x 1 on the
interval (3, 2) to within 9.436 1011 in 3 iterations and 6 function evaluations. How many iterations and how many function evaluations are needed
by the secant method to approximate this zero to a similar accuracy? Take
p0 = 2 and p1 = 3.
To 15 decimal places, the zero of f (x) = x3 + 2x2 3x 1 on the interval (3, 2)
is 2.912229178484397. With p0 = 2 and p1 = 3, the secant method yields
10
Section 2.5
n
2
3
4
5
6
pn
2.833333333333333
2.907928388746803
2.912449640422373
2.912228585591192
2.912229178402829
The absolute error in p6 is approximately 8.157 1011 . Thus, the secant method
produces the same accuracy with 5 iterations and 6 function evaluations as Newtons
method produces with 3 iterations and 6 function evaluations.
In Exercises 13 - 16 we will investigate the influence of the starting approximations p0 and p1 on the performance of the secant method. In each exercise,
apply the secant method to the indicated function using the indicated values
for p0 and p1 . Iterate until |pn pn1 | < 5 107 . Record and compare the
final approximation and the number of iterations in each case.
13.
f (x) = x3 + 2x2 3x 1
(a) p0 = 3, p1 = 2 (b) p0 = 2, p1 = 3 (c) p0 = 4, p1 = 2
(d) p0 = 2, p1 = 4
Let f (x) = x3 + 2x2 3x 1. The results obtained from the secant method using
the indicated initial approximations and a convergence tolerance of 5 107 are
given in the following table. Convergence is to the same value in each case, though
the number of iterations varies between seven and ten.
n
0
1
2
3
4
5
6
7
8
9
10
14.
(a)
p0 = 3, p1 = 2
3.0000000000
2.0000000000
2.8333333333
2.9944751381
2.9081883745
2.9120292369
2.9122296837
2.9122291785
2.9122291785
(b)
p0 = 2, p1 = 3
2.0000000000
3.0000000000
2.8333333333
2.9079283887
2.9124496404
2.9122285856
2.9122291784
2.9122291785
f (x) = x3 + 2x2 3x 1
(a) p0 = 1, p1 = 2
(b) p0 = 2, p1 = 1
(d) p0 = 2, p1 = 3
(c)
p0 = 4, p1 = 2
4.0000000000
2.0000000000
2.3846153846
3.8612334802
2.6617916359
2.8033231878
2.9327517345
2.9107736764
2.9122107041
2.9122291952
2.9122291785
(d)
p0 = 2, p1 = 4
2.0000000000
4.0000000000
2.3846153846
2.6412710567
3.0558864262
2.8864811276
2.9100446268
2.9122647548
2.9122291299
2.9122291785
(c) p0 = 3, p1 = 2
Secant Method
11
Let f (x) = x3 + 2x2 3x 1. The results obtained from the secant method using
given in the following table. Convergence is to the same value in each case, though
the number of iterations varies between seven and nine.
n
0
1
2
3
4
5
6
7
8
9
15.
(a)
p0 = 1, p1 = 2
1.0000000000
2.0000000000
1.1000000000
1.1517436381
1.2034498609
1.1984799141
1.1986903248
1.1986912437
1.1986912435
(b)
p0 = 2, p1 = 1
2.0000000000
1.0000000000
1.1000000000
1.2217294900
1.1964853266
1.1986453684
1.1986913364
1.1986912435
(c)
p0 = 3, p1 = 2
3.0000000000
2.0000000000
1.6538461538
1.3728481600
1.2483178273
1.2054893661
1.1989880967
1.1986930836
1.1986912440
1.1986912435
(d)
p0 = 2, p1 = 3
2.0000000000
3.0000000000
1.6538461538
1.4785544338
1.2745475717
1.2140977241
1.1996947841
1.1987052452
1.1986912564
1.1986912435
f (x) = tan(x) x 6
(a) p0 = 0, p1 = 0.48
(b) p0 = 0.24, p1 = 0.48(c) p0 = 0.4, p1 = 0.48
Let f (x) = tan(x) x 6. The results obtained from the secant method using
given in the following table. Note that two of the sequences converge to a value
a significant distance from the initial approximations and the number of iterations
required to achieve convergence varies substantially.
12
Section 2.5
n
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
16.
(a)
p0 = 0, p1 = 0.48
0.0000000000
0.4800000000
0.1868365254
0.2952141654
1.2550766960
3.4808530993
3.5393929824
83.4009233756
16.8908104674
5.3255962724
7.2982933741
7.3532403371
7.2925249098
7.2908218330
7.2901175754
7.2901096775
7.2901096515
(b)
p0 = 0.24, p1 = 0.48
0.2400000000
0.4800000000
0.326454862258533
0.377422024517396
0.637104975140150
0.170321715296955
0.634270604597732
1.668744839753233
6.968913445451308
5.437274962629953
6.722758639606624
6.394525185882151
6.580968311054888
6.462020207206985
6.538177621602646
6.497919752622464
6.535973055981917
6.533775643314268
6.571643864647825
6.609177721885587
6.692826576413555
6.833226882570812
7.106285855780790
7.424498934155690
7.175895024540025
7.218392879503447
7.324774654336274
7.278768517197979
7.288289155805048
7.290204346622126
7.290108858335217
7.290109651133373
7.290109651479092
(c)
p0 = 0.4, p1 = 0.48
0.4000000000
0.4800000000
0.4208674108
0.4332027501
0.4620367140
0.4470431841
0.4501486990
0.4511207210
0.4510459110
0.4510472568
0.4510472588
f (x) = x3 2x 5
(a) p0 = 1, p1 = 3
(b) p0 = 1, p1 = 2
(c) p0 = 3, p1 = 2
Let f (x) = x3 2x 5. The results obtained from the secant method using the
indicated initial approximations and a convergence tolerance of 5 107 are given
in the following table. Convergence is to the same value in each case, though the
number of iterations varies by 50% from the minimum to the maximum.
Secant Method
n
0
1
2
3
4
5
6
7
8
9
13
(a)
p0 = 1, p1 = 3
1.0000000000
3.0000000000
1.5454545455
1.8591632292
2.2003500782
2.0797991805
2.0937042425
2.0945585626
2.0945514782
2.0945514815
(b)
p0 = 1, p1 = 2
1.0000000000
2.0000000000
2.2000000000
2.0889679715
2.0942329564
2.0945524852
2.0945514814
2.0945514815
(c)
p0 = 3, p1 = 2
3.0000000000
2.0000000000
2.0588235294
2.0965586368
2.0945105536
2.0945514353
2.0945514815
17. The function f (x) = x3 + 2x2 3x 1 has a simple zero on the interval (1, 0).
Let f (x) = x3 + 2x2 3x 1. With initial approximations of p0 = 1 and p1 = 0
and a convergence tolerance of 5 105 , the secant method yields
n
2
3
4
5
pn
0.2500000000
0.2909090909
0.2864128312
0.2864620013

x = 0.286462.
18. For each of the functions given below, use the secant method to approximate
all real roots. Use an absolute tolerance of 106 as a stopping condition.
(a)
f (x) = ex + x2 x 4
(b)
f (x) = x3 x2 10x + 7
(c)
f (x) = 1.05 1.04x + ln x

14
Section 2.5
each of these intervals and a convergence tolerance of 106 , the secant method
yields
n
2
3
4
5
6
7
p0 = 2, p1 = 1
1.4332155776
1.5206690376
1.5067999496
1.5070982988
1.5070994842
1.5070994841
p0 = 1, p1 = 2
1.1921393409
1.2578082629
1.2903593087
1.2886496391
1.2886779411
1.2886779668

and x = 1.2886780.
f (3) > 0, f (0) > 0, f (1) < 0, f (3) < 0 and f (4) > 0. Therefore, the three
real zeros of f lie on the intervals (4, 3), (0, 1) and (3, 4). Using each of
these intervals and a convergence tolerance of 106 , the secant method yields
n
2
3
4
5
6
7
p0 = 4, p1 = 3
3.0294117647
3.0429276988
3.0426813987
3.0426827842
3.0426827843
p0 = 0, p1 = 1
0.7000000000
0.6845425868
0.6852213246
0.6852202475
0.6852202474
p0 = 3, p1 = 4
3.2500000000
3.3277310924
3.3592609903
3.3574338361
3.3574625095
3.3574625369

x = 0.6852202 and x = 3.3574625.
Secant Method
15
Using each of these intervals and a convergence tolerance of 106 , the secant
method yields
n
2
3
4
5
6
p0 = 0.80, p1 = 0.85
0.8298189963
0.8268947961
0.8271842009
0.8271809126
0.8271809085
p0 = 1.10, p1 = 1.15
1.1086787135
1.1096053349
1.1097126285
1.1097123038

and x = 1.1097123.
19. Keller (Probability of a Shutout in Racquetball, SIAM Review, 26, 267-8,
1984) showed that the probability that Player A will shut out Player B in a
game of racquetball is given by

21
1+w
w
P =
,
2
1 w + w2
where w denotes the probability that Player A will win any specific rally, independent of the server. Determine the minimal value of w that will guarantee
that Player A will shut out Player B in at least one-quarter of the games they
play. Repeat your calculations for at least half the games being shutouts and
at least three-quarters of the games being shutouts.
Let
1+w
f (w) =
2
w
1 w + w2
21
P.
For various values of P, the following table displays the results obtained from the
secant method. In each case, w0 = 0.5, w1 = 1 and = 5 107 .
16
Section 2.5
n
2
3
4
5
6
7
8
9
10
P = 0.25
0.6249436049
0.7154534705
0.8988222452
0.7599522840
0.7749547792
0.7799794811
0.7795784401
0.7795855351
0.7795855457
P = 0.50
0.7499624033
0.8506640530
0.8406140860
0.8422953603
0.8423048039
0.8423047910
P = 0.75
0.8749812016
0.9116135075
0.8998519540
0.8991764038
0.8991989787
0.8991989425
Thus, for Player A to shut out Player B in at least one-quarter of the games they
play, Player A must win any specific rally with a probability of 77.96%. To shut out
Palyer B at least one-half of the time, the probability of winning any specific rally
must rise to 84.23%. Finally, to shut out Player B at least three-quarters of the
time, the probability of winning any specific rally must rise to 89.92%.
20. A couple wishes to open a money market account in which they will save the
down payment for purchasing a house. The couple has $13,000 from the sale of
some stock with which to open the account and plans to deposit an additional
$200 each month thereafter. By the end of three years, the couple hopes to
have saved $20,000. If the money market account pays an annual interest of
r%, compounded monthly, then at the end of three years, the balance of the
account will be

r 36

1 + 12
1
r 36
13000 1 +
+ 200
.
r
12
12
What is the lowest interest rate which will achieve the couples goal of saving
$20,000? What is the lowest interest rate if the couple can raise their monthly
deposit to $250?
To reach a goal of $20,000, the couple does not need to earn any interest. Lets
work the problem with a goal of $25,000. Let

r 3
1 + 12
61
r 36
f (r) = 13000 1 +
+ 200
25000.
12
r/12

Because f (0.05) 2150.19 and f (0.1) 882.73, we take r0 = 0.05 and r1 = 0.1.
With a convergence tolerance of 5 106 , the secant method yields
n
2
3
4
5
rn
0.08544756
0.08611831
0.08613105
0.08613104
Thus, the couple needs to find an account which yields at least 8.61%, compounded
Secant Method
17
monthly. If the couple can raise their monthly deposit to $250, then

r 3
1 + 12
61
r 36
f (r) = 13000 1 +
+ 250
25000.
12
r/12

With the same starting values and convergence tolerance, the secant method now
yields
n
2
3
4
5
rn
0.05333706
0.05353955
0.05355262
0.05355261
With the increased monthly deposit, the couple can reach their goal with an account
that pays 5.36%, compounded monthly.
21. Suppose it was discovered that Commissioner Gordon had the flu when he
died, and his core temperature at the time of his death was 103 F . With
k = 0.337114, solve the equation

1
1
72 + td + 18 +
ektd = 103,
k
k
to determine the time of death based on this new information. Does Doc Bs
alibi still hold?
Let k = 0.337114 and
f (td ) = 31 + td

1
1
+ 18 +
ektd .
k
k
With p0 = 2, p1 = 0 and = 5 107 , the secant method yields

n
2
3
4
5
6
7
pn
1.4301052119
1.6084936592
1.5636280702
1.5647100350
1.5647187910
1.5647187893
Thus, time of death is roughly 1.564719 hours before 8:00 PM, or roughly 6:26
PM. Doc Bs alibi no longer holds.
Accelerating Convergence
2.6
1. Show that the equation for Aitkens 2 -method can be rewritten as

pn =
pn pn2 p2n1
.
pn 2pn1 + pn2
Explain why this formula is inferior to the one used in the text.
Combining the terms on the right-hand side of the formula yields
pn
(pn pn1 )2
pn + pn2 2pn1
2
pn + pn pn2 2pn pn1 p2n + 2pn pn1 + p2n1
pn + pn2 2pn1
2
pn pn2 pn1
.
pn + pn2 2pn1
= pn
=
=
Note that both formulas for calculating pn have the potential for cancellation error.
In the formula
(pn pn1 )2
pn = pn
,
pn + pn2 2pn1
cancellation error will primarily influence the second term, which is essentially just
a correction to pn . This can limit the overall effect of roundoff error on pn . In the
formula
pn pn2 p2n1
pn =
,
pn + pn2 2pn1
however, cancellation error influences the entire calculation. Thus, the latter formula
is inferior to the former because it is more susceptible to roundoff error.
2. Should Aitkens 2 -method be applied to a sequence generated by the bisection
method? Explain.
Aitkens 2 -method is designed to accelerate the convergence of linearly convergent
sequences. Recall that though the general trend over several terms in the bisection
sequence is linear, the sequence does not technically converge linearly. Consequently,
Aitkens 2 -method should not be applied a sequence generated by the bisection
method.
Section 2.6
3. The sequence listed below was obtained from the method of false position applied to the function f (x) = tan(x) x 6 over the interval (0.40, 0.48).
1
2
3
4
5
6
7
0.420867411
0.433202750
0.440495739
0.444807925
0.447357748
0.448865516
0.449757107
(a) Apply Aitkens 2 -method to the given sequence.

(b) To nine digits, the zero of f on (0.40, 0.48) is x = 0.451047259. Use this
to show that both the original sequence and the output from Aitkens 2 method are linearly convergent and estimate the corresponding asymptotic
error constants. By how much has Aitkens 2 -method reduced the asymptotic error constant?
(a) Using the first three terms in the False Position sequence, we calculate
p3
= p3
(p3 p2 )2
p3 + p1 2p2
0.440495739
0.451043933.
(0.440495739 0.433202750)2
0.440495739 + 0.420867411 2(0.433202750)
Next, using p2 , p3 and p4 , we calculate

p4
= p4
(p4 p3 )2
p4 + p2 2p3
0.444807925
0.451046159.
(0.444807925 0.440495739)2
0.444807925 + 0.433202750 2(0.440495739)
Continuing in this manner, we find

p5 = 0.451046884, p6 = 0.451047132, and p7 = 0.451047214.
(b) The values in columns 3 and 5 in the table below confirm that both the original
False Position sequence and the accelerated seqeunce converge linearly. The
asymptotic error constant for the False Position sequence is 0.5913, while
the asymptotic error constant for the accelerated sequence is roughly 0.34.
Aitkens 2 -method has reduced the error constant by nearly 40%.

n
1
2
3
4
5
6
7
False Position
pn
|en |/|en1 |
0.420867411
0.433202750
0.5913
0.440495739
0.5913
0.444807925
0.5913
0.447357748
0.5913
0.448865516
0.5913
0.449757107
0.5913
Aitkens 2
pn
|en |/|en1 |
0.451043933
0.451046159
0.451046884
0.451047132
0.451047214
0.3307
0.3409
0.3387
0.3543
4. The sequence listed below was obtained from Newtons method applied to the
function f (x) = x(1 cos x) to approximate the zero at x = 0.
1
2
3
4
5
6
7
0.646703997
0.425971211
0.282530441
0.187933565
0.125165810
0.083407519
0.055594262

(b) Verify that both the original sequence and the output from Aitkens 2 method are linearly convergent and estimate the corresponding asymptotic
error constants. By how much has Aitkens 2 -method reduced the asymptotic error constant?
(a) Using the first three terms in the Newtons method sequence, we calculate
p3
= p3
(p3 p2 )2
p3 + p1 2p2
0.282530441
0.016328890.
(0.282530441 0.425971211)2
0.282530441 + 0.646703997 2(0.425971211)

p4
= p4
(p4 p3 )2
p4 + p2 2p3
0.187933565
0.004726040.
(0.187933565 0.282530441)2
0.187933565 + 0.425971211 2(0.282530441)
Section 2.6
p5 = 0.001386361, p6 = 0.000408978, and p7 = 0.000120947.
(b) The values in columns 3 and 5 in the table below confirm that both the
original Newtons Method sequence and the accelerated seqeunce converge
linearly. The asymptotic error constant for the Newtons method sequence is
roughly 23 , while the asymptotic error constant for the accelerated sequence
is roughly 0.3. Aitkens 2 -method has reduced the error constant by more
than 50%.
n
1
2
3
4
5
6
7
Newtons Method
pn
|en |/|en1 |
0.646703997
0.425971211
0.6587
0.282530441
0.6633
0.187933565
0.6652
0.125165810
0.6660
0.083407519
0.6664
0.055594262
0.6665
Aitkens 2
pn
|en |/|en1 |
0.016328890
0.004726040
0.001386361
0.000408978
0.000120947
0.2894
0.2933
0.2950
0.2957
5. The sequence listed p

below was obtained from fixed point iteration applied to
the function g(x) = 10/(2 + x), which has a unique fixed point.
1
2
3
4
5
6
7
2.236067977
1.536450382
1.681574897
1.648098560
1.655643081
1.653933739
1.654320556

(b) To ten digits, the fixed point of g is x = 1.654249158. Use this to show
that both the original sequence and the output from Aitkens 2 -method
are linearly convergent and estimate the corresponding asymptotic error
constant. By how much has Aitkens 2 -method reduced the asymptotic
error constant?
(a) Using the first three terms in the fixed point iteration sequence, we calculate
p3
= p3
(p3 p2 )2
p3 + p1 2p2
1.681574897
1.656642880.
(1.681574897 1.536450382)2
1.681574897 + 2.236067977 2(1.536450382)

p4
= p4
(p4 p3 )2
p4 + p2 2p3
1.648098560
1.654373251.
(1.648098560 1.681574897)2
1.648098560 + 1.536450382 2(1.681574897)

p5 = 1.654255499, p6 = 1.654249483, and p7 = 1.654249174.
(b) The values in columns 3 and 5 in the table below confirm that both the
original Fixed Point sequence and the accelerated seqeunce converge linearly.
The asymptotic error constant for the Fixed Point sequence is roughly 0.23,
while the asymptotic error constant for the accelerated sequence is roughly
0.05. Aitkens 2 -method has reduced the error constant by nearly 80%.
n
1
2
3
4
5
6
7
Fixed Point
pn
|en |/|en1 |
2.236067977
1.536450382
0.2025
1.681574897
0.2320
1.648098560
0.2251
1.655643081
0.2266
1.653933739
0.2263
1.654320556
0.2264
Aitkens 2
pn
|en |/|en1 |
1.656642880
1.654373251
1.654255499
1.654249483
1.654249174
0.0518
0.0511
0.0513
0.0502
6. Apply Steffensens method to the iteration function g(x) = 12 10 x3 using a

starting value of p0 = 1. Perform four iterations, compute the absolute error in
each approximation and confirm quadratic convergence. To twenty digits, the
fixed point of g nearest x = 1 is x = 1.3652300134140968458.
Let g(x) = 12 10 x3 and take p0 = 1. Then p1,0 = g(

p0 ) = 1.5, p2,0 = g(p1,0 ) =
1.2869537676233750395 and
p1 = p2,0
(p2,0 p1,0 )2
= 1.3506084018798266709.
p2,0 2p1,0 + p0
Reinitializing with p1 , we calculate p1,1 = g(

p1 ) = 1.3726158478774946479, p2,1 =
g(p1,1 ) = 1.3614229508242237619 and
p2 = p2,1
(p2,1 p1,1 )2
= 1.3651964342563950237.
p2,1 2p1,1 + p1
Continuing in this manner, we obtain the values in the second column of the table
below. The third and fourth columns demonstrate the quadratic convergence of the
sequence.
Section 2.6
n
1
2
3
4
pn
1.3506084018798266709
1.3651964342563950237
1.3652300132342780910
1.3652300134140968458
|en |
1.462 102
3.358 105
1.798 1010
5.157 1021
|en |/|en1 |2
0.15706
0.15948
0.15948
7. (a) Perform ten iterations to approximate the fixed point of g(x) = cos x using p0 = 0. Verify that the sequence converges linearly and estimate the
asymptotic error constant.
To 20 digits, the fixed point is
x = 0.73908513321516064166.
(b) Accelerate the convergence of the sequence obtained in part (a) using
Aitkens 2 -method. By how much has Aitkens 2 -method reduced the
asymptotic error constant?
(c) Apply Steffensens method to g(x) = cos x using the same starting approximation specified in part (a). Perform four iterations, and verify that
convergence is quadratic.
(a) Let g(x) = cos x and take p0 = 0. The sequence generated by fixed point
iteration is given in the second column below. The values in the third column confirm the linear convergence of the sequence with an asymptotic error
constant of roughly 0.67.
n
1
2
3
4
5
6
7
8
9
10
pn
1.000000000
0.540302306
0.857553216
0.654289790
0.793480359
0.701368774
0.763959683
0.722102425
0.750417762
0.731404042
|en |/|en1 |
0.7619
0.5960
0.7158
0.6415
0.6934
0.6595
0.6827
0.6673
0.6778
(b) Applying Aitkens 2 -method to the sequence obtained in part (a) produces
the sequence listed in the second column in the table below. The values
in the third column confirm the linear convergence of the sequence with an
asymptotic error constant of roughly 0.45, nearly 30% lower than the error
constant for the original sequence.
7
n
3
4
5
6
7
8
9
10
pn
0.728010361
0.733665165
0.736906294
0.738050421
0.738636097
0.738876583
0.738992243
0.739042511
|en |/|en1 |
0.4894
0.4020
0.4749
0.4340
0.4644
0.4454
0.4588
(c) Let g(x) = cos x and take p0 = 0. Steffensens method produces the sequence
given in the second column of the table below. The values in the third column
confirm quadratic convergence of the sequence.
n
1
2
3
4
pn
0.685073357326045
0.738660156167713
0.739085106356719
0.739085133215160
|en |/|en1 |2
0.1457
0.1487
0.1539
8. (a) Perform ten iterations to approximate the fixed point of g(x) = ln(4 +
x x2 ) using p0 = 2. Verify that the sequence converges linearly and
estimate the asymptotic error constant. To 20 digits, the fixed point is
x = 1.2886779668238684115.
(c) Apply Steffensens method to g(x) = ln(4 + x x2 ) using the same starting
approximation specified in part (a). Perform four iterations, and verify
that convergence is quadratic.
(a) Let g(x) = ln(4 + x x2 ) and take p0 = 2. The sequence generated by fixed
point iteration is given in the second column below. The values in the third
column confirm the linear convergence of the sequence with an asymptotic
error constant of roughly 0.44.
n
1
2
3
4
5
6
7
8
9
10
pn
0.6931471806
1.438102388
1.214902035
1.318795484
1.275243786
1.294452355
1.286155040
1.289772517
1.288201641
1.288884977
|en |/|en1 |
0.2509
0.4937
0.4082
0.4461
0.4298
0.4369
0.4338
0.4352
0.4346
Section 2.6
asymptotic error constant of roughly 0.19, more than 50% lower than the error
n
3
4
5
6
7
8
9
10
pn
1.266359052
1.285796238
1.288107895
1.288573331
1.288657975
1.288674207
1.288677255
1.288677832
|en |/|en1 |
0.1291
0.1978
0.1835
0.1911
0.1881
0.1894
0.1896
(c) Let g(x) = ln(4 + x x2 ) and take p0 = 2. Steffensens method produces the
sequence given in the second column of the table below. The values in the
third column confirm quadratic convergence of the sequence.
n
1
2
3
4
pn
1.16762958832039
1.28720918451180
1.28867772517834
1.28867796682386
|en |/|en1 |2
0.1002
0.1120
0.1122
9. (a) Perform ten iterations to approximate the fixed point of g(x) = (1.05 +
ln x)/1.04 using p0 = 1. Verify that the sequence converges linearly and
estimate the asymptotic error constant. To 20 digits, the fixed point is
x = 1.1097123038867133005.
(c) Apply Steffensens method to g(x) = (1.05 + ln x)/1.04 using the same
starting approximation specified in part (a). Perform five iterations, and
verify that convergence is quadratic.
(a) Let g(x) = (1.05 + ln x)/1.04 and take p0 = 1. The sequence generated
by fixed point iteration is given in the second column below. The values
asymptotic error constant of roughly 0.88.
9
n
1
2
3
4
5
6
7
8
9
10
pn
1.009615385
1.018816780
1.027540321
1.035738374
1.043379401
1.050446990
1.056938248
1.062861811
1.068235665
1.073084977
|en |/|en1 |
0.9081
0.9040
0.9002
0.8967
0.8935
0.8905
0.8878
0.8853
0.8831
asymptotic error constant of roughly 0.74, nearly 16% lower than the error
n
3
4
5
6
7
8
9
10
pn
1.186794449
1.163634612
1.148195630
1.137554512
1.130049871
1.124670628
1.120769521
1.117916138
|en |/|en1 |
0.6995
0.7137
0.7235
0.7305
0.7355
0.7392
0.7419
(c) Let g(x) = (1.05 + ln x)/1.04 and take p0 = 1. Steffensens method produces
the sequence given in the second column of the table below. The values in
the third column confirm quadratic convergence of the sequence.
n
1
2
3
4
5
pn
1.22332847076812
1.12748779369567
1.11042323656777
1.10971357798330
1.10971230389082
|en |/|en1 |2
1.3770
2.2500
2.5208
2.5312
10. The function f (x) = 27x4 + 162x3 180x2 + 62x 7 has a zero of multiplicity
3 at x = 1/3. Apply both techniques for restoring quadratic convergence to
Newtons method to this problem. Use p0 = 0, and verify that both resulting
sequences converge quadratically.
Let f (x) = 27x4 + 162x3 180x2 + 62x 7. Then
f (x) = 108x3 + 486x2 360x + 62
10
Section 2.6
and
f (x) = 324x2 + 972x 360.
In Approach #1, the iteration function
g(x) = x
f (x)f (x)
f (x)f (x)
[f (x)]2
is used, while in Approach #2, the iteration function

g(x) = x 3
f (x)
f (x)
is used. The values in the third and fifth columns confirm quadratic convergence
through the second iteration. Following the second iteration, roundoff error prevents
further improvement in the approximation to the root.
n
0
1
2
3
Approach #1
pn
|en |/|en1 |2
0.0000000000000000
0.3277945619335347
0.0498
0.3333319367687672
0.0045
0.3333313867886697
Approach #2
pn
|en |/|en1 |2
0.0000000000000000
0.3387096774193548
0.0484
0.3333346459174751
0.0454
0.3333463715379559

x
500
21x
11. The function f (x) = 1+x
has a zero of multiplicity 2 at x = 2.5.
2 841 1 125
Apply both techniques for restoring quadratic convergence to Newtons method
to this problem. Use p0 = 2, and verify that both resulting sequences converge
quadratically.
Let f (x) =
x
1+x2
500
841
21x
125
. Then
f (x) =
1 x2
84
+
2
2
(1 + x )
841
and
f (x) =
2x(x2 3)
.
(1 + x2 )3
In Approach #1, the iteration function

g(x) = x
f (x)f (x)
f (x)f (x)
[f (x)]2

g(x) = x 2
f (x)
f (x)
is used. The values in the third and fifth columns confirm quadratic convergence.
n
0
1
2
3
11
Approach #1
pn
|en |/|en1 |2
2.000000000000000
2.443476520429667
0.2261
2.500026470879555
0.0083
2.500000000017783
0.0254
Approach #2
pn
|en |/|en1 |2
2.000000000000000
2.520094562647761
0.0804
2.499990347448545
0.0239
2.499999999989277
0.1151
12. The function f (x) = x(1 cos x) has a zero of multiplicity 3 at x = 0. Apply
both techniques for restoring quadratic convergence to Newtons method to this
problem, using p0 = 1. You should observe that the resulting sequences appear
to converge faster than quadratically. What apparent order of convergence do
you observe? Why is convergence faster than quadratic for this problem?
Let f (x) = x(1 cos x). Then f (x) = x sin x + 1 cos x and f (x) = x cos x +
2 sin x. In Approach #1, the iteration function
g(x) = x
f (x)f (x)
[f (x)]2 f (x)f (x)

g(x) = x 2
f (x)
f (x)
is used. To explore the order of convergence fully, the values below were obtained
using Maple with the Digits parameter set to 100.
n
1
2
3
4
Approach #1
pn
1.0860 101
1.4229 104
3.2011 1013
3.6447 1039
Approach #2
pn
5.9888 102
1.1936 105
9.4473 1017
4.6843 1050
Recall the root of this function is x = 0. The values above suggest that both sequences are converging of order three, which is better than expected from Newtons
method. This better than expected performance occurs because in the modified
function, the second derivative at x = 0 is zero.
13. Suppose Newtons method is applied to a function with a zero of multiplicity
m > 1. Show that the multiplicity of the zero can be estimated as the integer
nearest to
1
m
pn pn1 .
1 pn1 pn2
Verify that this formula produces an accurate estimate when applied to the
sequence listed in Exercise 4 and when applied to the sequence generated when
Newtons method was applied to the function f (x) = 1 + ln x x in the text.
12
Section 2.6
It was established in Section 2.5 that when Newtons method is applied to a function with a root of multiplicity m > 1, the resulting sequence of approximations
converges linearly with asymptotic error constant
=1
1
.
m
Then, in this section, it was established that for linearly convergent sequences, the
asymptotic error constant can be approximated by
Thus,
1
pn pn1
.
pn1 pn2
1
pn pn1
m
pn1 pn2
or
1
1
pn pn1
pn1 pn2
For the sequence in Exercise 4 and the sequence in worked Example 2.16, the
following estimates for the multiplicity of the root are obtained. As the function
f (x) = x(1cos x) has a root of multiplicity 3 and the function f (x) = 1+ln xx
has a root of multiplicity 2, we see that the formula
m
1
1
pn pn1
pn1 pn2
produces a very accurate estimate.

n
1
2
3
4
5
6
7
x(1 cos x)
pn
m
0.646703997
0.425971211
0.282530441 2.8558
0.187933565 2.9367
0.125165810 2.9720
0.083407519 2.9876
0.055594262 2.9945
1 + ln x x
pn
m
1.386294361
1.172192189
1.081540403 1.7343
1.039705144 1.8570
1.019594918 1.9257
1.009734085 1.9621
1.004851327 1.9809
14. Each of the following functions has a zero of multiplicity greater than one at
the specified location. In each case, apply the secant method to the function
f (x)/f (x) to approximate the indicated zero. Has the order of convergence
been restored to 1.618?
(a) f (x) = 1 + ln x x has a zero at x = 1 use p0 = 1 and p1 = 2
(b) f (x) = 27x4 + 162x3 180x2 + 62x 7 has a zero at x = 1/3

x
500
21x
(c) f (x) = 1+x
has a zero at x = 2.5
2 841 1 125
13
(a) Let f (x) = 1 + ln x x. Then f (x) = x1 1. Applying the secant method

to the function
f (x)
1 + ln x x
g(x) =
=
f (x)
x1 1
with p0 = 12 and p1 = 2 yields the results summarized in the table below. The
values in the fourth column suggest that the order of convergence has been
restored to 1.618.
n
pn
|en |
|en |/|en1 |1.618
1
2 0.859075117368965 1.4092 10
3 0.971218269671556 2.8782 102
0.6856
4 1.001573976632949 1.5740 103
0.4899
5 0.999984548159100 1.5452 105
0.5300
6 0.999999991900572 8.0994 109
0.4928
(b) Let f (x) = 27x4 + 162x3 180x2 + 62x 7. Then f (x) = 108x3 + 486x2
360x + 62. Applying the secant method to the function
g(x) =
f (x)
f (x)
with p0 = 0 and p1 = 1 yields the results summarized in the table below.

Here, convergence appears to be erratic, likely due to roundoff error.
n
2
3
4
5
6
7
(c) Let f (x) =
pn
0.343046357615894
0.333063893889014
0.333333452237016
0.333082657194573
0.333098977769093
0.333333337349884
x
1+x2
500
841
21x
125
|en |
9.7310 103
2.6944 104
1.1890 107
2.5068 104
2.3436 104
4.0166 109
|en |/|en1 |1.618

0.4863
0.0709
4.0128 107
157.0762
0.0030

. Then
f (x) =
1 x2
84
+
.
2
2
(1 + x )
841
Applying the secant method to the function

g(x) =
f (x)
f (x)
with p0 = 2 and p1 = 3 yields the results summarized in the table below.

The values in the fourth column suggest that the order of convergence has
been restored to 1.618, though after p4 , roundoff error prevents further
improvement in the approximation to the root.
n
2
3
4
5
pn
2.502774664443287
2.500077686554038
2.500000004821072
2.500000004821072
|en |
2.7747 103
7.7687 105
4.8211 109
4.8211 109
|en |/|en1 |1.618

1.0647
0.0215
14
Section 2.6
15. Repeat Exercise 14, but this time replace the standard secant method formula
for pn+1 by the formula
pn+1 = pn mf (pn )
pn pn1
,
f (pn ) f (pn1 )
where m is the multiplicity of the zero being approximated. The functions in

(a) and (c) have m = 2, and the function in (b) has m = 3.
Let f (x) = 1 + ln x x. As x = 1 is a root of f (x) of multiplicity 2, we attempt
the iteration function
pn pn1
pn+1 = pn 2f (pn )
.
f (pn ) f (pn1 )
With p0 = 0.5 and p1 = 2, the following sequence is produced
p2
p3
p4
p5
p6
= 6.095978968
= 9.612752412 + 0.1648235604i
= 2.972344498 + 2.437342132i
= 8.476800312 + 0.241149696i
= 3.038729588 + 2.233167032i
etc. Clearly this sequence is not converging. Even with p0 = 0.9 and p1 = 1.1,
a divergent sequence is produced. Divergent sequences result for the other two
functions from Exercise 14 as well.
16. The method of false position and fixed point iteration generate linearly convergent sequences for which
pn p
lim
(1)
n pn1 p
exists. Note that this limit does not involve absolute values. Let denote
the value of this limit. This exercise will lead us through the proof that the
sequence produced by Aitkens 2 -method converges more rapidly than linearly
convergent sequences for which (6) exists.
(a) Let
n =
Show that n 0 as n .
(b) Show that
pn p
.
pn1 p

pn = (pn p) 1
1
n +
(c) Show that

2
pn
=
=

1
1
1
(pn p) 1
1
n + n +
n1 +

pn p
2
( 1) + n ,
(n + )(n1 + )
15
where n = n n1 + (n + n1 ) 2n1 . Further, show that n 0 as

n .
(d) Show that
hence, as n ,
pn p
n1 + (n + 1)2
=1
;
pn p
n + ( 1)2 + n
pn p
0.
pn p
(a) Let
n =
pn p
.
pn1 p
By the definition of , it follows that

pn p
lim n = lim
= = 0.
n
n pn1 p
(b) Solving the equation that defines n for pn1 p yields
pn1 p =
pn p
.
n +
Using this result,

pn
= pn pn1 = (pn p) (pn1 p)

pn p
= (pn p)
+
n

1
= (pn p) 1
.
n +
(c)
2 pn
=
=
=
=
pn pn1

(pn p) 1
1
n +
(pn1 p) 1
n1 +
pn p
[(n + 1)(n1 + ) (n1 + 1)]
(n + )(n1 + )

pn p
( 1)2 + n ,
(n + )(n1 + )
where n = n n1 + (n + n1 ) 2n1 . Next,

lim n
lim (n n1 + (n + n1 ) 2n1 )
0 0 + (0 + 0) 2(0) = 0.
16
Section 2.6
(d) Using the definition of pn and the results from parts (a) and (b),
pn p
= pn p
(pn )2
2 pn
= pn p (pn p)
n1 + (n + 1)2
.
n + ( 1)2 + n
Consequently,
pn p
n1 + (n + 1)2
=1
pn p
n + ( 1)2 + n
and
pn p
n pn p
lim

n1 + (n + 1)2
1
n
n + ( 1)2 + n
0 + (0 + 1)2
= 1
0 + ( 1)2 + 0
= 1 1 = 0.
=
lim
Roots of Polynomials
2.7
1. Use synthetic division to deflate the given polynomial by the indicated root.
(a)
p(x) = x4 2.25x3 25.75x2 + 28.5x + 126,
(b)
p(x) = x4 + 1.83x3 0.081x2 + 1.83x 1.081,
(c)
p(x) = x4 + 20.5x3 + 129.5x2 + 230x 150,
x = 3
x = 2.3
x = 0.5
(a) Let p(x) = x4 2.25x3 25.75x2 + 28.5x + 126. Applying synthetic division
to p with x = 3 yields
b3
b2
b1
= a4 = 1;
= a3 + b3 x = 2.25 + 1(3) = 0.75;
= a2 + b2 x = 25.75 + 0.75(3) = 23.5; and
b0
= a1 + b1 x = 28.5 + (23.5)(3) = 42.
Therefore p(x) = (x 3)q(x) where q(x) = x3 + 0.75x2 23.5x 42.

(b) Let p(x) = x4 + 1.83x3 0.081x2 + 1.83x 1.081. Applying synthetic division
to p with x = 2.3 yields
b3
b2
= a4 = 1;
= a3 + b3 x = 1.83 + 1(2.3) = 0.47;
b1
b0
= a2 + b2 x = 0.081 + (0.47)(2.3) = 1; and

= a1 + b1 x = 1.83 + 1(2.3) = 0.47.
Therefore p(x) = (x + 2.3)q(x) where q(x) = x3 0.47x2 + x 0.47.

(c) Let p(x) = x4 + 20.5x3 + 129.5x2 + 230x 150. Applying synthetic division
to p with x = 0.5 yields
b3
= a4 = 1;
b2
b1
b0
= a3 + b3 x = 20.5 + 1(0.5) = 21;

= a2 + b2 x = 129.5 + 21(0.5) = 140; and
= a1 + b1 x = 230 + 140(0.5) = 300.
Therefore p(x) = (x 0.5)q(x) where q(x) = x3 + 21x2 + 140x + 300.
Section 2.7
2. Apply Laguerres method to each of the following polynomials with a starting

approximation of x0 = 0 and a convergence tolerance of 5 1011 .
(a)
p(x) = x3 4x2 3x + 5
(b)
p(x) = x3 7x2 + 14x 6
(c)
p(x) = x4 + 20.5x3 + 129.5x2 + 230x 150
(d)
p(x) = x4 2x3 5x2 + 12x 5
(a) Let p(x) = x3 4x2 3x + 5 and take x0 = 0 as an initial approximation and

= 5 1011 as a convergence tolerance. Evaluating p, p and p at x = x0
yields
p(x0 ) = 5, p (x0 ) = 3 and p (x0 ) = 8.
With these values we now calculate
p (x0 )
3
= = 0.6; and
p(x0 )
5

2
p (x0 )
p (x0 )
=
= 1.96.
p(x0 )
p(x0 )
G =
H
Since G is negative, we choose the negative sign in front of the radical in the
denominator of the formula for a. With n = 3, we find
a=
therefore,
3
G
2(3H G2 )
= 0.76478919808068;
x1 = x0 a = 0.76478919808068.
Because |a| > , we perform another iteration. Evaluating p, p and p at
x = x1 yields
p(x1 ) = 0.8133494631,
p (x1 ) = 7.3636060321,
and
p (x1 ) = 3.4112648115.
These values then lead to
G = 9.0534344295,
H = 86.1587698399
and
a = 0.10774953361112.
Thus,
x2 = x1 a = 0.87253873169181.
Once again, |a| > , so we perform a third iteration. Evaluating p, p and p
at x = x2 yields
p(x1 ) = 1.3729879943 103 ,
p (x1 ) = 7.6963383386,
and
3
p (x1 ) = 2.7647676098.
From here, we calculate

G = 5.6055394299 103 ,
H = 3.1424085987 107
and
a = 1.7838924723 104 .
Thus,
x3 = x2 a = 0.87271712093904.
We need another iteration. Evaluating p, p and p at x = x3 yields
p(x1 ) = 6.0280669345 1012 ,
p (x1 ) = 7.6968314480,
and
p (x1 ) = 2.7636972744.
G = 1.2768324459 1012 ,
H = 1.6303010948 1024
and
a = 7.8318811777 1013 .
Thus,
x4 = x3 a = 0.87271712093982.
Because |a| < , we terminate the iteration.
(b) Let p(x) = x3 7x2 + 14x 6 and take x0 = 0 as an initial approximation
and = 5 1011 as a convergence tolerance. Evaluating p, p and p at
x = x0 yields
p (x0 ) = 14
p(x0 ) = 6,
and p (x0 ) = 14.

p (x0 )
14
=
= 2.3333333333; and
p(x0 )
6

2
p (x0 )
p (x0 )
=
= 3.1111111111.
p(x0 )
p(x0 )
G =
H
a=
therefore,
3
G
2(3H G2 )
= 0.58568582800318;
x1 = x0 a = 0.58568582800318.
Section 2.7
x = x1 yields
p(x1 ) = 6.8705855222 104 ,
p (x1 ) = 6.8294820753,
and
p (x1 ) = 10.4858850320.
G = 9.9401747539 103 ,
H = 9.8791812142 107
and
a = 1.0060962372 104 .
Thus,
x2 = x1 a = 0.58578643762690.
at x = x2 yields
p(x1 ) = 5.3290705182 1015 ,
p (x1 ) = 6.8284271247,
and
p (x1 ) = 10.4852813742.
G = 1.2813542439 1015 ,
H = 1.6418686985 1030
and
a = 7.8042430868 1016 .
Thus,
x3 = x2 a = 0.58578643762690.
(c) Let p(x) = x4 + 20.5x3 + 129.5x2 + 230x 150 and take x0 = 0 as an initial
approximation and = 5 1011 as a convergence tolerance. Evaluating p,
p and p at x = x0 yields
p (x0 ) = 230 and p (x0 ) = 259.
p(x0 ) = 150,

p (x0 )
230
=
= 1.5333333333; and
p(x0 )
150

2
p (x0 )
p (x0 )
=
= 4.0777777778.
p(x0 )
p(x0 )
G =
H
a=
4
G
3(4H G2 )
= 0.49969960506531;
therefore,
x1 = x0 a = 0.49969960506531.
x = x1 yields
p(x1 ) = 0.1127461534,
p (x1 ) = 375.2778283295,
and
p (x1 ) = 323.4594477667.
G = 3.3285200171 103 ,
H = 1.1081914422 107
and
a = 3.0039493464 104 .
Thus,
x2 = x1 a = 0.49999999999995.
at x = x2 yields
p(x1 ) = 1.9838353182 1011 ,
p (x1 ) = 375.3750000000,
and
p (x1 ) = 323.5000000000.
G = 1.8921681480 1013 ,
H = 3.580300300 1026
and
a = 5.2849425727 1014 .
Thus,
x3 = x2 a = 0.50000000000000.
(d) Let p(x) = x4 2x3 5x2 +12x5 and take x0 = 0 as an initial approximation
and = 5 1011 as a convergence tolerance. Evaluating p, p and p at
x = x0 yields
p(x0 ) = 5,
p (x0 ) = 12
and p (x0 ) = 10.

p (x0 )
12
=
= 2.4; and
p(x0 )
5

2
p (x0 )
p (x0 )
=
= 3.76.
p(x0 )
p(x0 )
G =
H
Section 2.7
a=
therefore,
4
G
3(4H G2 )
= 0.52108014190383;
x1 = x0 a = 0.52108014190383.
x = x1 yields
p(x1 ) = 0.3139073716,
p (x1 ) = 5.7259956250,
and
p (x1 ) = 12.9946675314.
G = 18.2410358694,
H = 291.3388863296
and
a = 0.05863414460675.
Thus,
x2 = x1 a = 0.57971428651058.
at x = x2 yields
p(x1 ) = 4.7729153080 104 ,
p (x1 ) = 4.9657404108,
and
p (x1 ) = 12.9237475903.
G = 1.0403998585 104 ,
H = 1.0821610929 108
and
a = 9.6128915628 105 .
Thus,
x3 = x2 a = 0.57981041542621.
We need another iteration. Evaluating p, p and p at x = x3 yields
p(x1 ) = 2.2062351945 1012 ,
p (x1 ) = 4.9644980738,
p (x1 ) = 12.9235635711.
G = 2.2502125277 1012 ,
H = 5.0634564199 1024
and
and
a = 4.4440246763 1013 .
Thus,
x4 = x3 a = 0.57981041542666.
3. Construct an algorithm to deflate the n-th degree polynomial
p(x) = an xn + an1 xn1 + an2 xn2 + + a1 x + a0
by the quadratic factor x2 + x + ; i.e., find the polynomial
q(x) = bn2 xn2 + bn3 xn3 + bn4 xn4 + + b1 x + b0
such that p(x) = (x2 + x + )q(x).
Expand the product
(x2 + x + )(bn2 xn2 + bn3 xn3 + + b1 x + b0 )
and equate coefficients of like powers of x with
an xn + an1 xn1 + an2 xn2 + + a1 x + a0 .
The resulting equations are:
bn2
= an
bn3
bk
= an1 bn2
= ak+2 bk+1 bk+2 ,
k = n 4, n 5, ..., 0
4. Determine all roots for each of the following polynomials. Use a convergence
tolerance of 5 1011 .
(a)
p(x) = 2x5 6x4 + 5x3 + x2 + 2
(b)
p(x) = 3x6 + x3 + 10x 1
(c)
p(x) = x6 + x5 9x4 8x3 + 29x2 4x + 4
(d)
p(x) = x4 + 5x3 + 7x2 + 1
(e)
p(x) = 16x4 40x3 + 5x2 + 20x + 6
(f )
p(x) = 10x3 8.3x2 + 2.295x 0.21141
For each polynomial, Laguerres method with an initial approximation of x0 = 0

and a convergence tolerance of 5 1011 is used in combination with polynomial
deflation to locate all roots.
Section 2.7
(a) Let p(x) = 2x5 6x4 + 5x3 + x2 + 2. The roots of p are:

0.139796972 0.670536257i, 0.630244368, 1.675325212 0.758447670i
Five iterations were needed to achieve convergence for the first complex conjugate pair, and four additional iterations were needed to achieve convergence
for the real root. The final complex conjugate pair was obtained from the
quadratic formula.
(b) Let p(x) = 3x6 + x3 + 10x 1. The roots of p are:
0.099900596, 0.338131167 1.188064392i,
1.035306860 0.785803708i, 1.294450789
Three iterations were needed to achieve convergence for the first root, six
iterations were needed to achieve convergence for the first complex conjugate
pair and five final iterations were needed to achieve convergence for the second
complex conjugate pair.
(c) Let p(x) = x6 + x5 9x4 8x3 + 29x2 4x + 4. The roots of p are:
0.047216133 0.365949402i, 1.999999554,
2.547216133 0.925537914i, 2.000000446
Four iterations were needed to achieve convergence for the first complex conjugate pair, thirteen iterations were needed to achieve convergence for the first
real root and six final iterations were needed to achieve convergence for the
second complex conjugate pair. This polynomial has a double root at x = 2;
note the difficulty the numerical routine has in finding that root.
(d) Let p(x) = x4 + 5x3 + 7x2 + 1. The roots of p are:
0.047216133 0.365949402i, 2.547216133 0.925537914i
Four iterations were needed to achieve convergence for the first complex conjugate pair. The second complex conjugate pair was obtained from the quadratic
formula.
(e) Let p(x) = 16x4 40x3 + 5x2 + 20x + 6. The roots of p are:
0.356061762 0.162758383i, 1.241677445, 1.970446079
Five iterations were needed to achieve convergence for the complex conjugate
pair. The pair of real roots was obtained from the quadratic formula.
(f ) Let p(x) = 10x3 8.3x2 + 2.295x 0.21141. The roots of p are:
0.269995453, 0.270004548, 0.289999999
Seven iterations were needed to achieve convergence for the first root. The
remaining real roots was obtained from the quadratic formula. This polynomial
has a double root at x = 0.27; note the difficulty the numerical routine has in
finding that root.
5. The Chebyshev polynomials, Ti (x), are a special class of functions. They satisfy
the two-term recurrence relation
Ti+1 (x) = 2xTi (x) Ti1 (x)
with T0 (x) = 1 and T1 (x) = x.
(a) Using the recurrence relation, determine the formula for T6 (x).
(b) Locate all roots of T6 (x).
(a) With T0 (x) = 1 and T1 (x) = x, we calculate

2x(x) 1 = 2x2 1;
2x(2x2 1) x = 4x3 3x;
T2 (x)
T3 (x)
=
=
T4 (x)
T5 (x)
T6 (x)
= 2x(4x3 3x) (2x2 1) = 8x4 8x2 + 1;

= 2x(8x4 8x2 + 1) (4x3 3x) = 16x5 20x3 + 5x; and
= 2x(16x5 20x3 + 5x) (8x4 8x2 + 1)
=
32x6 48x4 + 18x2 1.
(b) Using Laguerres method with an initial approximation of x0 = 0 and a convergence tolerance of 5 1011 combined with polynomial deflation, the roots
of T6 (x) were found to be:
0.258819045, 0.707106781, 0.965925826
6. The Hermite polynomials, Hi (x), are a special class of functions. They satisfy
Hi+1 (x) = 2xHi (x) 2iHi1 (x)
with H0 (x) = 1 and H1 (x) = x.
(a) Using the recurrence relation, determine the formula for H5 (x).
(b) Locate all roots of H5 (x).
(a) With H0 (x) = 1 and H1 (x) = 2x, we calculate

2x(2x) 2(1) = 4x2 2;
H2 (x)
H3 (x)
H4 (x)
= 2x(4x2 2) 4(2x) = 8x3 12x;

= 2x(8x3 12x) 6(4x2 2) = 16x4 48x2 + 12; and
H5 (x)
2x(16x4 48x2 + 12) 8(8x3 12x) = 32x5 160x3 + 120x.
10
Section 2.7
of H5 (x) were found to be:
0, 0.958572465, 2.020182870
7. The Laguerre polynomials, Li (x), are a special class of functions. They satisfy
Li+1 (x) = (1 + 2i x)Li (x) i2 Li1 (x)
with L0 (x) = 1 and L1 (x) = 1 x.
(a) Using the recurrence relation, determine the formula for L4 (x).
(b) Locate all roots of L4 (x).
(a) With L0 (x) = 1 and L1 = 1 x, we calculate
L2 (x) =
L3 (x) =
L4 (x) =
(3 x)(1 x) 1 = x2 4x + 2;
(5 x)(x2 4x + 2) 4(1 x) = x3 + 9x2 18x + 6; and
(7 x)(x3 + 9x2 18x + 6) 9(x2 4x + 2)
= x4 16x3 + 72x2 96x + 24.

of L4 (x) were found to be:
0.322547690, 1.745761101, 4.536620297, 9.395070912
8. The Legendre polynomials, Pi (x), are a special class of functions. They satisfy
Pi+1 (x) =
2i + 1
i
xPi (x)
Pi1 (x)
i+1
i+1
with P0 (x) = 1 and P1 (x) = x.

(a) Using the recurrence relation, determine the formula for P5 (x).
(b) Locate all roots of P5 (x).
(a) With P0 (x) = 1 and P1 (x) = x, we calculate
P2 (x) =
3
1
3
1
x(x) (1) = x2 ;
2
2
2
2
P3 (x)
11
=
=
P4 (x)
=
=
P5 (x)
=
=

5
3 2 1
2
x
x
(x)
3
2
2
3
5 3 3
x x;
2 2

7
5 3 3
3 3 2 1
x
x x
x
4
2
2
4 2
2
35 4 15 2 3
x x + ; and
8
4
8

9
35 4 15 2 3
4 5 3 3
x
x x +
x x
5
8
4
8
5 2
2
63 5 35 3 15
x x + x.
8
4
8
of P5 (x) were found to be:
0, 0.538469310, 0.906179846
9. The concentration, C, of a certain chemical in the bloodstream t hours after

injection into muscle tissue is given by
C=
3t2 + t
.
50 + t3
At what time is the concentration greatest?

To determine the time when the concentration is greatest, we need to solve C (t) =
0. Here,
C (t) =
(50 + t3 )(6t + 1) (3t2 + 3t)(3t2 )

3t4 2t3 + 300t + 50
=
.
(50 + t3 )2
(50 + t3 )2
Thus, C (t) = 0, when

3t4 2t3 + 300t + 50 = 0.
Using Laguerres method with an initial approximation of x0 = 0 and a convergence tolerance of 5 1011 combined with polynomial deflation, the roots of this
polynomial are found to be:
0.16669, 2.49315 4.00887i, 4.48632
We discard the negative root and the complex conjugate pair. Therefore, the concentration is greatest t = 4.48632 hours after injection (that is, roughly four and
one-half hours after injection).
12
Section 2.7
10. DeSanti (A Model for Predicting Aircraft Altitude Loss in a Pull-Up from
a Dive, SIAM Review, 30 (4), pp. 625 - 628, 1988) develops the following
relationship for the ratio between the final velocity, Vf , and the initial velocity,
V0 , for an aircraft executing a pull-up from a dive:
3
1 Vf
Vf
1
B
+ B cos 0 = 0.
3 V0
V0
3
0 is the initial flight path angle and B = g/(kV02 ), where g is the acceleration
due to gravity and k is related to the coefficient of lift. The altitude loss during
the pull-up can be determined from the ratio Vf /V0 using the equation
y =
1 (Vf /V0 )2
.
2kB
Determine the altitude loss associated with each of the following sets of system
parameters (take g = 9.8 m/s2 ):
(a) V0 = 100 m/s, 0 = 30 , k = 0.00196 m1
(b) V0 = 150 m/s, 0 = 10 , k = 0.00145 m1
(c) V0 = 200 m/s, 0 = 45 , k = 0.00128 m1
(d) V0 = 250 m/s, 0 = 30 , k = 0.00112 m1
(a) Let V0 = 100 m/s, 0 = 30 and k = 0.00196 m1 . Then
B=
g
1
= ,
kV02
2
and the ratio between the final velocity, Vf , and the initial velocity, V0 staisfies
1
3
Vf
V0
3
1 Vf
+ 0.09967936855889 = 0.
2 V0
Using Laguerres method with an initial approximation of x0 = 0 and a convergence tolerance of 5 1011 combined with polynomial deflation, the roots
of this polynomial are found to be:
0.20511, 1.10924, 1.31435
We discard the negative root and the positive root less than 1. Therefore,
Vf /V0 = 1.10924 and the altitude loss during the pull-up is
y =
1 (Vf /V0 )2
= 117.56 m.
2kB
(b) Let V0 = 150 m/s, 0 = 10 and k = 0.00145 m1 . Then

B=
g
= 0.30038314176245,
kV02
13
1
3
Vf
V0
3
0.30038314176245
Vf
0.03751368645151 = 0.
V0
0.12717, 0.87929, 1.00646
We discard the two negative roots. Therefore, Vf /V0 = 1.00646 and the
altitude loss during the pull-up is
y =
1 (Vf /V0 )2
= 14.89 m.
2kB
(c) Let V0 = 200 m/s, 0 = 45 and k = 0.00128 m1 . Then

B=
g
= 0.19140625000000,
kV02
1
3
Vf
V0
3
0.19140625000000
Vf
0.19798867599685 = 0.
V0
0.53207 0.52447i, 1.06414
We discard the complex conjugate pair. Therefore, Vf /V0 = 1.06414 and the
y =
1 (Vf /V0 )2
= 270.18 m.
2kB
(d) Let V0 = 250 m/s, 0 = 30 and k = 0.00112 m1 . Then

B=
g
= 0.14,
kV02
1
3
Vf
V0
3
0.14
Vf
0.21208977680351 = 0.
V0
14
Section 2.7
0.51064 0.60188i, 1.02128
We discard the complex conjugate pair. Therefore, Vf /V0 = 1.02128 and the
y =
1 (Vf /V0 )2
= 137.16 m.
2kB
11. In determining the minimum cushion pressure needed to break a given thickness
of ice using an air cushion vehicle, Muller (Ice Breaking with an Air Cushion
Vehicle, in Mathematical Modeling: Classroom Notes in Applied Mathematics,
M.S. Klamkin, editor, SIAM, 1987) derived the equation

2 3
h2
2 h4
h
p3 (1 2 ) + 0.4h 2 2 p2 +
p
= 0,
4
r
3r
3r2
where p denotes the cushion pressure, h the thickness of the ice field, r the size
of the air cushion, the tensile strength of the ice, and is related to the width
of the ice wedge. Taking = 0.5, r = 40 feet and = 150 pounds per square
inch (psi), determine the cushion pressure needed to break a sheet of ice 6 feet
thick.
Substituting = 0.5, r = 40, = 150 and h = 6 into the equation for p yields
0.75p3 2.775p2 + 3.796875p 1.423828125 = 0.
Using Laguerres method with an initial approximation of x0 = 0 and a convergence tolerance of 5 1011 combined with polynomial deflation, the roots of this
polynomial are found to be:
0.58665, 1.55667 0.90156i
We discard the complex conjugate pair. Therefore, a pressure of p = 0.58665
pounds per square inch is needed to break a sheet of ice 6 feet thick.
12. Determine the roots of the polynomials
P (x) = (x 1)(x 2)(x 3)(x 4)(x 5)(x 6)(x 7)(x 8)(x 9)(x 10)
and
P (x) = (x 1)(x 2)(x 3)(x 4)(x 5)(x 6)(x 7)(x 8)(x 9)(x 10) + x5
with Laguerres method as the central rootfinding scheme. Apply a convergence
tolerance of 5 1011 , and take 0 as the initial approximation.
15
Expanding the polynomial

(x 1)(x 2)(x 3)(x 4)(x 5)(x 6)(x 7)(x 8)(x 9)(x 10)
and then using Laguerres method with an initial approximation of x0 = 0 and a
convergence tolerance of 5 1011 combined with polynomial deflation, the roots
x1
1.00000000000000
x2
x3
=
=
2.00000000000009
2.99999999999851
x4
x5
x6
=
=
=
4.00000000001604
4.99999999991054
6.00000000026806
x7
x8
=
=
6.99999999954737
8.00000000043207
x9
x10
=
=
8.99999999978244
10.00000000004488
On the other hand, expanding the polynomial

(x 1)(x 2)(x 3)(x 4)(x 5)(x 6)(x 7)(x 8)(x 9)(x 10) + x5
and then using Laguerres method with an initial approximation of x0 = 0 and a
convergence tolerance of 5 1011 combined with polynomial deflation, the roots
x1
1.00000275579138
x2
x3
=
=
1.99920899238048
3.02591337402009
x4
x5
=
=
3.82274586537434
5.24675800922628 0.75148509914663i
x6
x7
=
=
5.24675800922628 + 0.75148509914663i
9.75659456161856 0.36838948049253i
x8
x9
x10
= 9.75659456161856 + 0.36838948049253i
= 7.57271193537201 1.11728346550533i
= 7.57271193537201 + 1.11728346550533i
13. One mole of H2 S is injected into a two liter reaction chamber, and the reversible
reaction
2H2 S
2H2 + S2
16
Section 2.7
is allowed to proceed to equilibrium. If the equilibrium constant for the indicated

reaction is k = 0.016, how much H2 and S2 are present at equilibrium?
Suppose that at equilibrium there are x moles/liter of S2 ; that is, [S2 ] = x. Since
the reaction chamber has a volume of two liters, this means that 2x moles of
S2 are present. From the reaction equation, it is seen that for every mole of S2
produced, two moles of H2 S are used. Thus, to produce 2x moles of S2 , 4x
moles of H2 S have reacted; hence, at equilibrium, 1 4x moles of H2 S remain.
In other words, [H2 S] = 14x
at equilibrium. By a similar argument, it follows that
2
[H2 ] = 2x at equilibrium. Substituting these concentrations into the equilibrium
constant expression yields
(2x)2 x
0.016 =
.
14x 2
2
This last equation can be rearranged into the form
16x3 0.256x2 + 0.128x 0.016 = 0.

The roots of this equation are:
0.03097 0.10895, 0.07795.
The complex conjugate pair is discarded. Therefore, at equilibrium, [S2 ] = 0.07795
and [H2 ] = 0.1559. Taking into account the volume of the reaction chamber, there
are 0.1559 moles of S2 and 0.3118 moles of H2 present at equilibrium.
14. The reversible reaction
2SO2 + O2
2SO3
is allowed to proceed to equilibrium in a one liter reaction chamber. If 0.012
moles of SO2 and 0.0076 moles of O2 are initially present and the equilibrium
constant for the indicated reaction is k = 44.643, how much SO3 is present at
equilibrium?
Suppose that at equilibrium there are x moles/liter of SO3 ; that is, [SO3 ] = x.
Since the reaction chamber has a volume of one liter, this means that x moles of
SO3 are present. From the reaction equation, it is seen that for every mole of SO3
produced, one mole of SO2 is used. Thus, to produce x moles of SO3 , x moles
of SO2 have reacted; hence, at equilibrium, 0.012 x moles of SO2 remain. In
other words, [SO2 ] = 0.012 x at equilibrium. By a similar argument, it follows
that [O2 ] = 0.0076 x2 at equilibrium. Substituting these concentrations into the
equilibrium constant expression yields
44.643 =
x2
.
(0.012 x)2 (0.0076 x2 )
This last equation can be rearranged into the form

0.0000488572992 0.0113571792x 0.1249972x2 22.3215x3 = 0.
17
The roots of this equation are:

0.00480 0.02289i, 0.00400.
The complex conjugate pair is discarded. Therefore, at equilibrium, [SO3 ] =
0.00400 and there are 0.00400 moles of SO3 present.

673

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

673

Uploaded by

Copyright:

Available Formats

Bisection Method

(b) Let f (x) = x5 + 2x 1. Because f is continuous on [0, 1] with f (0) =

Let f (x) = x3 + x2 3x 3. Because f is continuous on (1, 2) with f (1) = 4 < 0

Let f (x) = 1 ln x. Because f is continuous on (2, 3) with f (2) = 1 ln 2

Let f (x) = x6 3. Because f is continuous on (1, 2) with f (1) = 2 < 0 and

6. Determine a formula which relates the number of iterations, n, required by the

function whose zero is to be located, f

set N max = 1 + int(log2 ((b a)/))

to be contained in the interval (8,9)? (Hint: Consider the number of base 2

which implies that

Combining (1) and (2) yields

We therefore estimate that f has roots at approximately x = 2.91220 and x =

11. Approximate 3 13 to three decimal places by applying the bisection method to

the equation 1/x 37 = 0.

Thus, the second smallest positive root of the equation tan(x) x 6 = 0 is

15. It can be shown that the equation

(a) The equation

the root we are seeking must satisfy

Thus, the unique root of the equation

f (x) = 1.05 1.04x + ln x

(a) Let f (x) = ex + x2 x 4. Observe that the equation ex + x2 x 4 = 0 is

Thus, the zeros of f (x) = ex + x2 x 4 are approximately x = 1.5071001

Thus, the zeros of f (x) = x3 x2 10x+7 are approximately x = 3.0426836,

(a1 , b1 ) = (0.80, 0.85)

(a1 , b1 ) = (1.10, 1.15)

Thus, the zeros of f (x) = 1.051.04x+ln x are approximately x = 0.8271812

Thus, = 0.4039731. We then calculate f = 1.2625461, m = 0.3533436, and

By trial and error, we find

f (2) = 0.108 > 0.

The Method of False Position

The Method of False Position

To determine whether the root is contained on (a1 , p1 ) or on (p1 , b1 ), we

To determine whether the root is contained on (a1 , p1 ) or on (p1 , b1 ), we

The Method of False Position

To determine whether the root is contained on (a1 , p1 ) or on (p1 , b1 ), we

function whose zero is to be located, f

initialize polder = b and pold = b

3. Confirm that || < 1 for the remaining configurations in Figure 2-5.

(bn p)f (p)

(bn p)f (p)

The Method of False Position

The Method of False Position

false position because f () = sin = 0; thus,

which implies that convergence is better than linear.

the exact value of the root is x = 3, compute the absolute error in

x = 15. What is the apparent order of convergence in this case?

Convergence appears to be of order two (note that each error appears to be

Thus, the zero of f (x) = x3 + 2x2 3x 1 on the interval (1, 0) is approximately

f (x) = 1.05 1.04x + ln x

(a) Let f (x) = ex + x2 x 4. Observe that the equation ex + x2 x 4 = 0 is

The Method of False Position

Thus, the zeros of f (x) = ex + x2 x 4 are approximately x = 1.5070992

Thus, the zeros of f (x) = x3 x2 10x+7 are approximately x = 3.0426825,

(a1 , b1 ) = (0.80, 0.85)

(a1 , b1 ) = (1.10, 1.15)

Thus, the zeros of f (x) = 1.051.04x+ln x are approximately x = 0.8271813

The Method of False Position

f (x) = x2 1 + cos( 2x) 2 sin( 2x).

Thus, the other root of x2 = 1 cos( 2x) + 2 sin( 2x) is approximately x =

The Method of False Position

Fixed Point Iteration Schemes

Fixed Point Iteration Schemes

(b) Show that