MST121 Chapter B1 Modelling With Sequences
MST121 Chapter B1 Modelling With Sequences
MST121 Chapter B1 Modelling With Sequences
Using Mathematics
Chapter B1
This publication forms part of an Open University course. Details of this and
other Open University courses can be obtained from the Student Registration
and Enquiry Service, The Open University, PO Box 197, Milton Keynes
MK7 6BJ, United Kingdom: tel. +44 (0)845 300 6090, email
general-enquiries@open.ac.uk
Alternatively, you may visit the Open University website at
http://www.open.ac.uk where you can learn more about the wide range of
courses and packs offered at all levels by The Open University.
To purchase a selection of Open University course materials visit
http://www.ouw.co.uk, or contact Open University Worldwide, Walton Hall,
Milton Keynes MK7 6AA, United Kingdom, for a brochure: tel. +44 (0)1908
858793, fax +44 (0)1908 858787, email ouw-customer-services@open.ac.uk
First published 1997. Second edition 2002. Third edition 2008. Reprinted 2008.
Copyright
c 1997, 2002, 2008 The Open University
Introduction to Block B 4
Study guide 5
Introduction 6
Summary of Chapter B1 53
Learning outcomes 53
Solutions to Activities 55
Solutions to Exercises 58
Index 61
3
Introduction to Block B
4
Study guide
5
Introduction
6
1 Modelling car owner ship
Specify Create
purpose model
Do
mathematics
Evaluate Interpret
results
7
CHAPTER B1 MODELLING WITH SEQUENCES
Do not spend more than a Assume that you have decided which type (make and model) of car you
minute or two on this activity. wish to buy, but not its age nor how long you will keep it. Note down
some features that you might take into account in deciding your strategy
for car purchase.
A solution is given on page 55.
Specify Before creating a mathematical model, the purpose for which it is intended
stick to aspects of the situation that can be measured; hence aspects such
modelling purposes.
running costs, such as repairs and servicing, fuel, insurance and tax.
still too complex to cope with as it stands. We need to pick out the item
will show how to keep the cost of motoring as low as possible, in a sense yet
to be made precise.
depend upon? If the values of the inputs can be altered as we please, then
it may be possible to select them in order to achieve some ‘lowest cost’ for
the motorist.
8
SECTION 1 MODELLING CAR OWNERSHIP
Purpose of model
A car of given type is purchased when it is one year old. At what age
should it be sold in order to minimise the average annual cost of
owning and running the car?
9
CHAPTER B1 MODELLING WITH SEQUENCES
In the above list, n is the independent variable (the input which can be
varied at will) and an is the dependent variable (the output, which is to be
minimised). The other variables, i, vi and ci , are needed as intermediate
quantities, permitting us eventually to express the output, an , in terms of
the input, n.
Note that ci includes only the running costs of servicing and repairs, and
not items such as fuel, insurance and tax. For a given type of car with
normal mileage, these latter items are assumed to be the same each year,
so they will not affect the purpose stated for our model.
At this point we need to make assumptions about how both the value of
the car and its servicing and repair costs vary with age. The depreciation
Sometimes a combination of in the value of any article with time is usually dealt with in one of two
these two approaches is ways:
adopted.
by assuming that the article has a definite fixed life, and that its value
declines to zero linearly over that life;
by assuming that the value of the article declines by a fixed proportion
each year.
Where used cars are concerned, the second approach seems to be closer to
actual valuations than the first. This observation does not extend to the
initial year of a car’s life, because of the substantial additional premium
which is paid to acquire a brand-new vehicle. However, since we are
considering a car bought at one year old, depreciation by a fixed
proportion will fit the bill here. Hence we have
vi+1 = rvi (i = 0, 1, . . . , n),
Data are introduced here only where r is a constant such that 0 < r < 1. Data gathered for one particular
for illustrative purposes. type of car, for illustrative purposes within this model, give r = 0.85 and
They would not normally v0 = 16 000. This value of r is equivalent to a depreciation rate of 15% per
appear this early in the annum. The value of v0 can be regarded as the initial cost of the car minus
modelling cycle. the ‘brand-new’ premium referred to above (or as what you might expect
to receive if you bought a new car and then immediately tried to resell it).
Hence, for this type of car, vi is given by
v0 = 16 000, vi+1 = 0.85vi (i = 0, 1, . . . , n). (1.1)
Reliable data on the costs of servicing and repairs are less easy to come by.
It is generally acknowledged that these annual costs of running a car
This is consistent with the increase with the age of the vehicle, and the simplest approach consistent
principle of simplicity within with this observation is to assume that these costs increase linearly with
mathematical modelling, age; that is,
which states that, in the
absence of any other guide as ci+1 = ci + d (i = 0, 1, . . . , n − 1),
to how to model a where d is a positive constant. Some (admittedly rough and ready) data
relationship between two
for the type of car being considered here lead to the values c0 = 600 and
variables, you should choose
d = 250. Hence ci is given by
the simplest form of function
which takes into account the c0 = 600, ci+1 = ci + 250 (i = 0, 1, . . . , n − 1). (1.2)
known facts.
Equations (1.1) and (1.2) will permit vi and ci to be expressed in terms of
car age, i years. How can we then relate the dependent variable of the
model, an , to n, vi and ci ?
The net loss involved in buying the car when one year old and selling it
n years later is v1 − vn+1 . Also, the total cost of running the car between
purchase and sale is
c1 + c2 + · · · + cn .
10
SECTION 1 MODELLING CAR OWNERSHIP
The total cost of owning and running the car for n years is therefore
v1 − vn+1 + c1 + c2 + · · · + cn ,
and the corresponding average annual cost over these years is
1
an = (v1 − vn+1 + c1 + c2 + · · · + cn ) (n = 1, 2, . . .). (1.3)
n
The mathematical model is encapsulated in equations (1.1)–(1.3), together
with the list of definitions of the variables which appear in these equations.
Before moving on to solve the equations, it is timely to list some of the
assumptions which have been made, explicitly or implicitly, in the course
of constructing the model.
The value of the car at a given age is the same for both purchase and
resale (there is no ‘dealer margin’ involved).
The value of the car diminishes by a fixed proportion each year.
The annual servicing and repair costs increase linearly with age.
The car is sold after a whole number of years.
Both car value and annual running costs depend only on the age of the
car (and not, for example, on its mileage).
Any effects of inflation are to be ignored, as are the annual costs of
fuel, insurance and tax.
In terms of the modelling cycle, it is time now to ‘do the mathematics’, in
order to solve equations (1.1)–(1.3). Before embarking on this task, there
are some mathematical preliminaries to consider.
11
CHAPTER B1 MODELLING WITH SEQUENCES
2 n 1 − rn+1
a + ar + ar + · · · + ar = a (r
= 1).
1−r
Solution
n
�
(a) The sum is 1 + 2 + · · · + n = i.
i=1
9
�
(b) The sum is 33 + 43 + · · · + 93 = i3 .
i=3
1
Note that ar = ar and that (c) Using sigma notation on the left-hand side, we have
ar0 = a × 1 = a. Also, the n
� � �
(a) Use sigma notation to express the sum of the squares of the integers
between 5 and 13, inclusive.
(b) By first writing out the two sums involved, express
6
� 3
�
i
5 − 5i
i=1 i=1
12
SECTION 1 MODELLING CAR OWNERSHIP
In Example 1.1(c) you were reminded of the formula for the sum of a finite
geometric series. There is also a formula for the sum of the finite
arithmetic series 1 + 2 + · · · + n, that is, for the sum of the first n positive
integers. This formula is derived by writing down the sum in the normal
order and then with this order reversed:
n
�
i = 1 + 2 + 3 + · · · + (n − 1) + n,
i=1
�n
i = n + (n − 1) + (n − 2) + · · · + 2 + 1.
i=1
This result will be useful in the next subsection. Note the geometrical
interpretation of it, which is illustrated (for the case n = 5) in Figure 1.2.
The total shaded area is equal to the sum of the shaded column areas,
5
�
which is 1 + 2 + 3 + 4 + 5 = i, and also to half of the area of the whole
i=1
rectangle, which is 2
× 5 × 6. Figure 1.2 Shaded area
equals half of area of
rectangle
Activity 1.3 Summing positive integer s
Intellectual precocity
It is said that the great German mathematician Carl Friedrich Gauss
(1777–1855), at the age of ten, was asked along with the rest of his
school class to carry out the sum which you did in Activity 1.3, having
been shown no way of doing it beyond the addition of successive
numbers. Their teacher intended it to be a lengthy task, but Gauss
came up with the correct answer almost immediately, by applying the
approach used above to derive equation (1.6).
‘Gauss had not been shown the trick for doing such problems rapidly.
It is very ordinary once it is known, but for a boy of ten to find it
instantaneously by himself is not so ordinary. This opened the door
through which Gauss passed on to immortality.’
(E. T. Bell, Men of Mathematics, Volume 1 (Pelican, 1953))
13
CHAPTER B1 MODELLING WITH SEQUENCES
Do Do the mathematics
mathematics
Equations (1.1) define a geometric sequence, and equations (1.2) define an
arithmetic sequence. Once a closed form has been found for each of these,
The closed forms for an the resulting expressions for vi and ci can be used on the right-hand side of
arithmetic sequence and for equation (1.7). With an then expressed in terms of n, we can seek to
a geometric sequence were minimise the value of an with respect to changes in n.
given in Chapter A1,
Subsections 2.2 and 3.2,
respectively. It is common Activity 1.4 Finding the closed-form solutions
practice to refer to such
formulas as ‘closed-form (a) Find the closed-form solution of the recurrence system (1.1).
solutions’ of the underlying
recurrence system.
(b) Find the closed-form solution of the recurrence system (1.2).
(c) Use equations (1.4) and (1.6) to simplify as far as possible the
n
�
expression for ci that follows from your answer to part (b).
i=1
Solutions are given on page 55.
Comment
Any finite arithmetic series can be summed in a manner similar to that
seen in the solution to part (c). Alternatively, such a sum can be found
directly by application of the approach used to derive equation (1.6).
= 13 600(1 − (0.85)n ).
i=1
This graph indicates once more that the minimum value of an occurs for
n = 3, but it also shows that this minimum is barely below the value for
n = 4. This observation is substantiated by another look at the table. The
average annual cost for resale of the car after four years is less than £1
more than the minimum cost identified, and this difference is not enough
to decide whether to sell the car after three years or after four.
More generally, the variation of an around n = 3 is not huge. Perhaps you
would not object to selling the car after two years or after five, for the
average costs indicated. On the other hand, the increased costs involved in
selling after one year or beyond six years give pause for thought.
The model tells you that, on the basis of the various assumptions made
here, resale after anything between three and four years is the most
economical approach available, at an estimated average cost of about
£2850 per annum plus fuel, insurance and tax.
15
CHAPTER B1 MODELLING WITH SEQUENCES
We do not have sufficient information to make that check with reality here,
but it may well turn out (it is frequently the case) that the initial model is
judged as only partly successful once the comparison with reality has taken
place. If so, it may be necessary to try to improve the model, by taking
another trip around the five stages of the modelling cycle. An early step
would be to re-examine the assumptions which were made in constructing
the first model, in order to see whether any need amendment in the light
of experience gained.
For the model developed here, for example, the assumption of linearly
increasing annual costs for servicing and repairs may turn out to be too
simple. The data underlying this assumption are few, and more data on
annual running costs might suggest a suitable refinement.
This activity requires a (a) Find the closed-form solution of the recurrence relation (1.8).
generalisation of what you did
(b) Find the closed-form solution of the recurrence relation (1.9).
in Activity 1.4 and the main
text following that activity. (c) Use equations (1.4) and (1.6) to simplify as far as possible the
n
�
expression for ci which follows from your answer to part (b).
i=1
(d) Starting from equation (1.7) and your answers to parts (a) and (c),
express an as simply as possible in terms of the variable n and the
parameters v0 , r, c0 and d.
Solutions are given on page 55.
16
SECTION 1 MODELLING CAR OWNERSHIP
Comment
In the final formula obtained in part (d), parameter values can be
substituted for v0 , r, c0 and d to correspond to any given make and model
of car. The resulting specific version of the formula for an can then be used
to predict when such a car should be resold, after purchase at one year old,
to most economical effect.
Further generalisations of the model are possible, for example, to take into
account purchases at age other than one year. However, we shall not
pursue such extensions here.
Summar y of Section 1
This section has introduced:
the use of sigma notation to describe concisely sums of terms from a
sequence;
the word ‘series’ to describe a sum of consecutive terms of a sequence;
the formula 12 n(n + 1) for the sum of the first n positive integers;
the application of general modelling ideas to a specific problem
involving the economics of car ownership.
You will not be assessed directly on the modelling content of this section,
though the principles are important if you want to continue to further
studies in applied mathematics. The framework for the modelling cycle,
within the five key stages shown in Figure 1.1, was described concisely in
Chapter A1, Section 7. To some extent, at least, all of the aspects included
there were addressed in the course of the modelling done in this section.
( 12 )i .
i=m i=m
Exercise 1.2
(a) Find the sum of the integers from 51 to 100, inclusive.
(b) Hence find the sum of the numbers 517, 527, 537, . . . , 1007.
(Hint : 517 = 7 + 10i, where i = 51.)
17
2 Populations: exponential model
Purpose of model
Specify Describe how an animal population may change with time, in a way
purpose that is applicable over a long period of time and for a variety of
populations.
18
SECTION 2 POPULATIONS: EXPONENTIAL MODEL
19
CHAPTER B1 MODELLING WITH SEQUENCES
Solution
(a) During the (n + 1)th year, the ‘joiners’ will be the pheasants born
during the year, of which there are 2.6Pn , and the ‘leavers’ will be the
pheasants that die, of which there are 0.4Pn . The difference between
births and deaths gives the increase in the population size during the
year (since we are assuming that there is no migration), so
Pn+1 − Pn = 2.6Pn − 0.4Pn
= 2.2Pn .
(d) This model predicts that the population size will increase, and will go
on increasing more and more rapidly (see Figure 2.2).
Comment
This model predicts that the population size will increase very rapidly, and
this pheasant population did undergo rapid increase from 1937 to 1942
(as shown in Figure 2.1). However, the model predicts that this increase
will continue. For example, it predicts a population size on 1 April 1997 of
8(3.2)60 , which is about 1.6 × 1031 . Estimating the area of ground taken up
by a pheasant as 0.1 m2 , and using the fact that the island has area
1.6 × 106 m2 , this gives 1.6 × 1030 m2 of pheasants in an area of
1.6 × 106 m2 . This implies that the island will be covered about 1024 deep
in pheasants, which seems unlikely!
If the purpose of the model in this example is to predict the population
size in the long term, then it is certainly not reasonable. The population
cannot continue to grow indefinitely in the way predicted by the model.
20
SECTION 2 POPULATIONS: EXPONENTIAL MODEL
which is equivalent to
Pn+1 = (1 + b − c)Pn .
If 1 + b − c > 0 and P0 > 0,
then
This again describes a geometric sequence, for which the closed form is
Pn > 0 for n = 1, 2, 3, . . . .
Pn = (1 + b − c)n P0 .
If 1 + b − c = 0, then
Pn = 0 for n = 1, 2, 3, . . . .
The long-term behaviour of Pn will depend on the value of 1 + b − c.
The three cases identified in Activity 2.1 are illustrated in Figure 2.3.
Figure 2.3 Plots of Pn = (1 + b − c)n P0 for (a) b > c, (b) b = c, (c) b < c
The parameters in this model are the starting population size, P0 , the
proportionate birth rate, b, and the proportionate death rate, c. Since the
last two of these occur in the combination b − c, it is often convenient to
replace this combination by a single symbol, for which we shall choose r.
21
CHAPTER B1 MODELLING WITH SEQUENCES
This growth rate r gives the Thus r = b − c is the annual proportionate growth rate of the
net effect of births minus population. In terms of r, the recurrence relation for Pn is
deaths over each year, per
head of the population. Pn+1 = (1 + r)Pn ,
Although r is a growth rate, and the closed-form solution is
it can be negative.
Pn = (1 + r)n P0 .
The graphs in Figure 2.3 are The three cases b > c, b = c and b < c correspond respectively to r > 0,
those of Pn = (1 + r)n P0 in r = 0 and r < 0.
each of these cases.
The model just derived could be called a geometric model for population
variation, since it predicts that the population size rises or falls according
to a geometric sequence. However, it is more usual to describe it as the
exponential model, since the expression (1 + r)n P0 involves raising the base
number 1 + r to the exponent n.
Exponential model
The exponential model for population variation is based on the
assumption of a constant proportionate growth rate, r. The model is
described by either the recurrence relation
Pn+1 = (1 + r)Pn (n = 0, 1, 2, . . .), (2.1)
or its closed-form solution
Pn = (1 + r)n P0 (n = 0, 1, 2, . . .), (2.2)
where Pn is the population size at n years after some chosen starting
time.
22
SECTION 2 POPULATIONS: EXPONENTIAL MODEL
Solution
(a) Take n = 0 at the census date in 1889. Then n = 10 in 1899, so
P0 = 26 and P10 = 160. According to the exponential model
(equation (2.2)), we have Pn = (1 + r)n P0 , so, in particular,
P10 = (1 + r)10 P0 ; that is, 160 = (1 + r)10 × 26.
It remains to solve this equation for the proportionate growth rate, r.
On dividing through by 26 and then taking the 10th root, we obtain
� � �0.1
10 160 160
1+r = , so r = − 1 = 0.20 (to 2 s.f.).
26 26
Hence the proportionate growth rate is about 20% per year.
(b) The exponential model for the colony is therefore Pn = 26(1.20)n . As a check, the value of
26(1.20)10 should be close to
(c) The year 1914 corresponds to n = 25, for which the population size of
160, the given population size
the colony is predicted to be P25 = 26(1.20)25 2480.30. This must be
for 1899. In fact,
rounded to a whole number, to provide a meaningful prediction for a
population size. The prediction for the year 1914 is therefore 2480. 26(1.20)10 160.99.
(Note that even rounding to the nearest whole number may be We shall not pursue the
misleading. This prediction is based on a value of r which was rounded effects of such rounding. In
to two significant figures, so there is reason to question its accuracy. further discussion of this
If we use the exact value of r to calculate P25 , then the prediction model, we take r = 0.20.
becomes 2443 to the nearest integer.)
The (human) US population was about 4 million in 1790, when the first
national census was taken, and 63 million in 1890.
(a) Assuming that this population can be modelled exactly over the
intervening period by an exponential model, find to two significant
figures the value of the annual proportionate growth rate, r. (Take the
population to be measured in millions.)
(b) Write down the corresponding formula for the US population Pn
(in millions) at n years after the census date in 1790.
(c) Assuming that this model continues to hold for years after 1890, what
US population is predicted in the year 1950 ?
(d) In which year, according to the model, does the US population reach The approach needed to
500 million? answer this part is similar to
that for problems in
Solutions are given on page 56.
Chapter A3, Section 4: for
example, Activity 4.6(b) and
the main text before that
activity.
23
CHAPTER B1 MODELLING WITH SEQUENCES
If b < c (that is, r < 0), then the prediction is for a decreasing population
size. If a population has fewer births than deaths, then it can be expected
to decrease, and examples of real populations varying in this sort of way
can be given (see Figure 2.4). In this case, the mathematical solution
Pn = (1 + r)n P0 will eventually predict populations of size less than 1.
Such a prediction is not to be taken literally, of course. Values of Pn less
than 1 would suggest that the population has reached extinction. In
practice, a model implying such population decrease would be interpreted
as predicting extinction when the population size drops below some
minimum level from which there is no chance of recovery. There is nothing
obviously unrealistic about the general form of population change
predicted in this case.
Figure 2.4 Population (in breeding pairs) of red-backed shrikes in Great Britain
(1952–1989)
The graph in Figure 2.5(b) shows fairly steady population size over some
periods, interspersed with sudden drops and subsequent steady recoveries.
24
SECTION 2 POPULATIONS: EXPONENTIAL MODEL
Here we must recall that the exponential model relies on the assumption of
constant proportionate birth and death rates. This assumption will fail to
hold where there are marked variations in weather or other conditions
which could affect the population. It is noticeable that the largest sudden
drop for the grey heron population of Figure 2.5(b) coincides with the
particularly severe UK winter of 1963.
If b > c (that is, r > 0), then the prediction is for a population that
increases. This increase continues indefinitely, and becomes more and more
rapid. The prediction in this case is not reasonable as regards long-term
behaviour. Any population which is subject to unlimited increase will
eventually reach an unsustainable size. What constitutes an
‘unsustainable’ level, and how long the population may take to reach that,
depends on the particular population being considered.
Many human populations have shown exponential increase over quite long
periods of time, as for example in the case of the US population considered
in Activity 2.2. Figure 2.6(a) shows the US population for the period
1790–1890, together with the exponential model whose growth rate was
obtained in Activity 2.2(a). The fit between the model and the data
appears good here. However, Figure 2.6(b) looks at the US population on
a different time scale, and includes further census data up to 1990. It is
evident that the previous exponential model is far from accurate in its
predictions after 1900. For example, the prediction from Activity 2.2(c) of
a population size 332 million in the year 1950 is more than twice as large
as the actual figure for that year, 151 million.
25
CHAPTER B1 MODELLING WITH SEQUENCES
Other types of population variation are also found with some species:
examples are shown in Figure 2.8. (For illustrative purposes, it is assumed
that the flour beetles of Figure 2.8(a) breed in clearly separated
generations.) Clearly, the exponential model has little application in such
cases.
26
SECTION 2 POPULATIONS: EXPONENTIAL MODEL
Summar y of Section 2
This section has introduced:
the exponential (or geometric) model for population variation,
Pn+1 = (1 + r)Pn , which is based on the assumption of a constant
proportionate growth rate r and has the closed-form solution
Pn = (1 + r)n P0 , where Pn is the population size at n years after some
chosen starting time;
a method for estimating, from data about the population, a value for
the parameter r in the exponential model;
evidence of the invalidity of the exponential model for large population
sizes, in that it predicts unlimited growth.
Exercise 2.2
In fact, growth of the world population has been considerably faster since According to the UN, the
1950 than the model of Exercise 2.1 would suggest. Using the estimated world population reached
world population of 6.06 billion in the year 2000, repeat the task of 6 billion in October 1999.
Exercise 2.1(a) for the period 1950–2000.
27
3 Populations: logistic model
28
SECTION 3 POPULATIONS: LOGISTIC MODEL
Logistic model
The logistic model for population variation is based on the
assumption of a proportionate growth rate R(P ) of the form
� �
P
R(P ) = r 1 − , (3.2)
E
where r and E are positive parameters.
The model is described by the recurrence relation
� �
Pn
Pn+1 − Pn = rPn 1 − (n = 0, 1, 2, . . .), (3.3)
E
where Pn is the population size at n years after some chosen starting
time.
We have now completed the second stage in the modelling cycle for the
logistic model. It is useful at this point to review how far we have come,
being modelled?
29
CHAPTER B1 MODELLING WITH SEQUENCES
Variables
The time variable, n years, can be chosen to be zero at any convenient
point in time. The population size at that time is P0 , and Pn then denotes
Sometimes Pn may be taken the population size n years later. In fact, this can be generalised to other
to denote the number of time intervals. We usually want to look at population sizes at yearly
breeding pairs, rather than intervals, because of the pattern of population variation within each year
individuals, or to denote that was mentioned at the start of Subsection 2.2, but there are occasions
population density. when a time interval other than one year is appropriate. This is
particularly likely when considering laboratory experiments, which are not
affected by nature’s annual cycle. For the flour beetles in Figure 2.8(a), it
would be natural to take Pn to be the population in the nth generation
from the start of the experiment. The steps involved in creating the model
do not depend on the choice of time interval, so this may be altered as
appropriate to suit the circumstances of the population being studied.
Assumptions
As in the exponential model, we have assumed that migration is not a
significant factor for the populations considered here. In moving from the
exponential to the logistic model, we assumed that the proportionate
growth rate is a decreasing linear function of population size. Since the
growth rate is births minus deaths, the linear decline in proportionate
growth rate could arise in several ways:
proportionate birth rate decreases linearly with population size, while
proportionate death rate remains constant;
proportionate death rate increases linearly with population size, while
proportionate birth rate remains constant;
proportionate birth rate decreases linearly with population size, and
proportionate death rate increases linearly with population size.
In each case, it is possible to derive the appropriate form of the logistic
recurrence relation from suitable information on the birth and death rates
of a population, as in the following example and activity.
30
SECTION 3 POPULATIONS: LOGISTIC MODEL
Solution
(a) The proportionate growth rate R(P ) at population size P is the
proportionate birth rate minus the proportionate death rate; that is,
R(P ) = (0.31 − 1.5 × 10−5 P ) − 0.11 = 0.20 − 1.5 × 10−5 P.
Hence the population growth for the year which starts at time n is
R(Pn )Pn = (0.20 − 1.5 × 10−5 Pn )Pn .
Since this is also the increase from Pn to Pn+1 , we have the recurrence
system
P0 = 3200, Pn+1 − Pn = (0.20 − 1.5 × 10−5 Pn )Pn .
(b) To show that this recurrence relation is logistic, the right-hand side
must be written in the form rPn (1 − Pn /E):
� �
−5 1.5 × 10−5
(0.20 − 1.5 × 10 Pn )Pn = 0.20Pn 1 − Pn
0.20
� �
= 0.20Pn 1 − 7.5 × 10−5 Pn .
Parameter s
The parameter E may be interpreted from two different viewpoints. It
arose in Figure 3.1, as the intercept of the proportionate growth rate graph
(assumed linear) on the P -axis. Hence it is the population size at which the
proportionate growth rate is zero.
On the other hand, it can be seen directly from the logistic recurrence
relation (3.3) that if Pn attains the size E when n = m, then
� �
E
Pm+1 − E = rE 1 − = 0, so Pm+1 = E.
E
Hence Pm+2 = E, Pm+3 = E, . . . .
In other words, if Pn reaches size E, then it stays there.
31
CHAPTER B1 MODELLING WITH SEQUENCES
These two views are, of course, consistent. Zero growth rate implies a
constant size of population, and vice versa. The parameter E is called the
equilibrium population level (or, in some contexts, the carrying
capacity).
More generally, the constant sequence
Pn = E (n = 0, 1, 2, . . .)
is a solution of the logistic recurrence relation. In fact, this sequence is the
only non-zero constant sequence which satisfies the logistic recurrence
relation. For if Pn = c (n = 0, 1, 2, . . .), where c is a constant, then from
equation (3.3) we have
� �
c
The constant sequence c − c = rc 1 − , so c = 0 or c = E.
E
Pn = 0 (n = 0, 1, 2, . . .)
We do not yet know whether other sequences generated by the logistic
represents a total absence of recurrence relation will settle down near some particular positive value in
the population. the long term, but if they do so, then that value must be E. The
population is then ‘in equilibrium’.
Thus, according to the logistic model, the equilibrium population level for
the barnacle goose population of Example 3.1 is 13 300, while the
equilibrium population level for the ring-necked pheasants in Activity 3.1
is 1500.
The meaning of the parameter r is apparent from the way in which the
logistic model was constructed: it is the proportionate growth rate of the
population at small population sizes. We can use the interpretation of E
given above to be more precise about what ‘small’ means here. If Pn is
small compared with the equilibrium population level, E, then Pn /E is
small compared with 1, so that the logistic recurrence relation (3.3) is
approximated closely by
Pn+1 = (1 + r)Pn ,
which is the exponential model once again. We have demonstrated that
this more primitive model is a limiting case of the logistic model. At low
population levels, population growth can be modelled satisfactorily by the
exponential model formula, Pn = (1 + r)n P0 , but this continues only while
the population size remains well below its equilibrium level, E.
32
SECTION 3 POPULATIONS: LOGISTIC MODEL
For the population of gannets in Example 2.2 (page 22), it was shown that
the proportionate growth rate was r = 0.20 at relatively low population
levels. At higher population levels (later than 1919, see Figure 2.7(a)) this
growth rate is no longer appropriate.
The size of this population was 3200 in 1924, 3600 in 1929 and 4000 in
1934. The growth was therefore 400 in each of the 5-year periods before
and after 1929, when the population size was 3600. This suggests an
annual growth rate of about 80 for a population size of 3600.
Assuming that the behaviour of this population satisfies a logistic
recurrence relation, estimate the value of the equilibrium population
level, E.
Solution
The logistic recurrence relation is
� �
Pn
Pn+1 − Pn = rPn 1 − , (3.3)
E
where in this case r = 0.20 from consideration of the proportionate growth
rate at low population levels. Since the annual growth rate Pn+1 − Pn is
estimated to be 80 for the population size Pn = 3600, we have
� � � �
3600 3600
80 = 0.20 × 3600 1 − = 720 1 − .
E E
On solving this equation for E, we obtain
3600 80 8
=1− = , from which E = 4050.
E 720 9
This is the equilibrium level for the population.
For the US population (1790–1890), you found in Activity 2.2 (page 23)
that the annual proportionate growth rate was r = 0.028. At higher
population levels (later than 1900, see Figure 2.6(b)) this is no longer
appropriate.
The size of the US population was 76 million in 1900, 92 million in 1910
and 106 million in 1920. Hence there was a growth of 30 million over a
20-year period with midpoint 1910, when the population was 92 million.
This suggests an annual growth rate of about 1.5 million for a population
size of 92 million.
Assuming that the behaviour of this population satisfies a logistic
recurrence relation, estimate the value of the equilibrium population
level, E. (As before, take the population size Pn to be measured in
millions.)
A solution is given on page 56.
33
CHAPTER B1 MODELLING WITH SEQUENCES
The approach used in the last example and activity can be applied only if
the value of r is found first from data on the population growth at low
population levels. When such data are not available, a strategy of the
following type may be adopted instead.
34
SECTION 3 POPULATIONS: LOGISTIC MODEL
� �
Pn
Pn+1 − Pn = 2.9Pn 1 − .
250
If Pn = 150 for some n, then we have
� �
150
Pn+1 = 150 + 2.9 × 150 1 − = 324.
250
(Note that, in this case, the population size has moved in one year
from well below the equilibrium population level, 250, to well above it.)
35
CHAPTER B1 MODELLING WITH SEQUENCES
Graphical approach
First, note that there are several ways in which the information given by
Essentially, this first way equation (3.3) can be graphed. One way is to plot proportionate growth
gives the graph in Figure 3.1, rate, R(Pn ) = (Pn+1 − Pn )/Pn , against Pn , which we illustrated when
but with P = 0 excluded. deriving the logistic recurrence relation. More useful, for current purposes,
is a plot of growth rate, Pn+1 − Pn , against Pn , which is shown in
Figure 3.2(a). The graph is that of the function given by the right-hand
side of equation (3.3), which is a quadratic function of Pn . It cuts the
horizontal axis at Pn = 0 and Pn = E.
Note that population growth in a year means that Pn+1 − Pn > 0, while
population decline means that Pn+1 − Pn < 0. Hence there is population
growth where the graph of Figure 3.2(a) lies above the horizontal axis
(that is, where 0 < Pn < E) and population decline where the graph lies
below this axis (where Pn > E). The equilibrium population level, Pn = E,
appears as a point on the graph for which population growth is zero.
This maximum value is Also, the growth rate will be a maximum at the vertex of the parabola,
r× 1
− 1 1 where Pn = 12 E. The further that Pn is from 21 E, the smaller is the
2 E(1 2) = 4 rE,
corresponding growth rate.
as marked on Figure 3.2(a).
Another point to note is that altering the value of the parameter r has the
If the vertical axis were effect of scaling this parabola vertically, so that for a given value of Pn (and
labelled by y, then we would fixed E), the growth rate becomes larger in magnitude as r is increased.
call this a y-scaling.
The graph in Figure 3.2(b) is a plot of Pn+1 (next year’s population size)
against Pn (this year’s population size). This arises from rearranging
equation (3.3) as follows:
� �
Pn
Pn+1 = Pn + rPn 1 −
E
� � ��
Pn
= Pn 1 + r 1 − .
E
The right-hand side here is again a quadratic function of Pn , whose graph
passes through the point (E, E). The graph cuts the horizontal axis when
Pn+1 = 0, that is, when Pn = 0 and when 1 + r(1 − Pn /E) = 0. So the
intercepts on the horizontal axis are at Pn = 0 and Pn = E(1 + 1/r).
This latter value of Pn is of particular interest from the modelling point of
view since, according to the graph, if Pn = E(1 + 1/r) then Pn+1 = 0, and
if Pn > E(1 + 1/r) then Pn+1 < 0. Thus the population becomes extinct
within a year, according to the model, if its numbers ever rise as high as
E(1 + 1/r). As the parameter r is increased, this ‘ceiling on sustainability’
becomes closer to the equilibrium population level, E.
36
SECTION 3 POPULATIONS: LOGISTIC MODEL
Algebraic approach
We have already employed some algebra above, while investigating what
useful information about the logistic model can be gleaned from graphs.
The following argument, however, depends upon algebraic manipulation
alone.
Starting once more from the logistic recurrence relation (3.3), we divide
both sides of the equation by E, to obtain
� �
Pn+1 Pn Pn Pn
− =r 1− .
E E E E
Now we put xn = Pn /E, so that also xn+1 = Pn+1 /E. This results in a
recurrence relation for the new variable xn , namely,
xn+1 − xn = rxn (1 − xn ), (3.4)
which looks simpler than the original, since the parameter E no longer
appears. Mathematically, it is a version of the logistic recurrence
relation (3.3) with E = 1. In modelling terms, the quantity xn represents
the population size Pn as a proportion of its equilibrium level, E.
The derivation of equation (3.4) shows that the parameter E is just a
scaling factor for any logistic recurrence sequence. So, in order to calculate
terms of a sequence Pn which satisfies the logistic recurrence relation (3.3),
with given values for the parameters r, E and for P0 , we could proceed as
follows:
calculate x0 = P0 /E;
calculate the required number of terms of the corresponding sequence
xn which satisfies the recurrence relation (3.4);
calculate Pn = xn E for each relevant value of n.
The graph of Pn against n is just a vertically scaled version of the graph of
xn against n, with scaling factor E.
It follows that, in order to understand the long-term behaviour of
sequences Pn which satisfy equation (3.3), for given r and E, we need to
understand only the long-term behaviour of sequences xn which satisfy
equation (3.4). This observation saves considerable labour where computer
investigation of equation (3.3) is concerned, since it tells us that there is no
need to experiment with different values of the parameter E.
37
CHAPTER B1 MODELLING WITH SEQUENCES
Summar y of Section 3
This section has introduced:
the logistic model for population variation, with recurrence relation
� �
Pn
Pn+1 − Pn = rPn 1 − ,
E
which assumes that proportionate growth rate is a decreasing linear
function of population size;
methods for estimating, from data about a population, values for the
parameters r (proportionate growth rate at low population levels)
and E (equilibrium population level);
some graphical and algebraic analysis of the behaviour to be found
according to the logistic model.
Exercise 3.1
UN data suggest that the proportionate growth rate per decade of the
world population was 0.20 in 1960, when the population size was
3.02 billion, and 0.15 in 1990, when the population size was 5.27 billion.
(a) Assuming that the behaviour of the world population satisfies a
logistic model, estimate the values of the equilibrium population
level, E, and the parameter r.
(b) Using this model, and the world population estimate of 6.06 billion
for 2000, what population is forecast in the year 2010?
38
4 Logistic recur rence sequences on the computer
To study this section you will need access to your computer, together with
Computer Book B.
Are the features that are apparent here shared by sequences also generated
by the logistic recurrence relation but with other values of r? You can now
use your computer to examine this question.
39
CHAPTER B1 MODELLING WITH SEQUENCES
40
SECTION 4 LOGISTIC RECURRENCE SEQUENCES ON THE COMPUTER
Summar y of Section 4
In this section you investigated sequences generated by the logistic
recurrence relation, Pn+1 − Pn = rPn (1 − Pn /E), via the direct evaluation
of terms by computer. The choice of starting value P0 does not usually
affect the long-term behaviour of the sequences. This behaviour alters
qualitatively as the parameter r is increased, as summarised in the table
on page 40. Some sequences have 2-cycles, 4-cycles or chaotic behaviour in
the long term.
41
5 Sequences and limits
Figure 5.1 Graphs of logistic recurrence sequences with P0 = 3200 and E = 13 300 for (a) r = 0.20, (b) r = 2.9
42
SECTION 5 SEQUENCES AND LIMITS
You saw in Subsection 3.1 that if the logistic recurrence relation generates
a constant sequence, Pn = c say, then there are only two values that c can
take, namely 0 and E. If a sequence generated by the logistic recurrence
relation does converge, then for large enough n the terms of the sequence
are effectively equal. Thus in the long term, a convergent sequence is
effectively a constant sequence, so its limit must be one or other of the
values 0 or E. However, this says only that convergence to any other limit
is impossible. As the case above with r = 2.9 shows, there is no guarantee
of convergence even when potential limit values have been identified.
The above method of finding possible limit values applies to other types of
sequence generated by recurrence relations.
Suppose that a sequence xn is generated by the recurrence relation Each non-zero value of x0
� � here will give a different
1 2
xn+1 = xn + (n = 0, 1, 2, . . .). (5.1) sequence.
2 xn
To what limit values might such a sequence xn converge?
Solution
Suppose that xn = c is a constant sequence generated by the given
recurrence relation. Then we have xn+1 = xn = c, and so
c = 12 (c + 2/c).
This is equivalent to
2c = c + 2/c; that is, c = 2/c.
√
So we obtain c2 = 2 and hence c = ± 2. These are the two possible limit
values for any sequence generated by the recurrence relation (5.1).
43
CHAPTER B1 MODELLING WITH SEQUENCES
The modulus of a real In this subsection, the modulus of a real number is used to permit concise
number was defined in expression of certain types of inequality and to express the distance from
Chapter A3, Subsection 1.2. the origin to a point on the number line.
The distances from the origin to the points 5 and −4 on the number line
are 5 and 4, respectively. Since the modulus of a real number x is defined
by
�
x, if x ≥ 0,
|x| =
−x, if x < 0,
we have
|5| = 5 and |−4| = 4,
as shown in Figure 5.2.
More generally, the distance In general, the distance from the origin to a point a on the number line is
say that the distance from a to the origin is less than 1. Thus
44
SECTION 5 SEQUENCES AND LIMITS
45
CHAPTER B1 MODELLING WITH SEQUENCES
There are other closed forms for sequences an where you can see that an
becomes smaller and smaller (arbitrarily small, that is, arbitrarily close
to 0) as n becomes large. Examples of this are
1 1 1
bn = 2
, bn = n and bn = − .
n 2 5n
These sequences are convergent, with limit 0. In each case, bn is of the
form 1/an , where terms of the sequence an (given respectively by
equations (5.3)) become arbitrarily large as n increases. This highlights a
general rule to use when considering the convergence of sequences.
Reciprocal Rule
If the terms of a sequence bn are of the form 1/an , where terms of the
sequence an become arbitrarily large as n increases, then the
sequence bn is convergent, and lim bn = 0.
n→∞
Another useful result concerns what happens to a sequence in the long run
if each of its terms is multiplied by the same number. For example, it has
already been remarked that the sequence an = 1/n2 converges to 0, since
its values eventually become arbitrarily small as n increases. Multiplying
all terms an by 50 makes them all larger by this factor, but this does not
prevent the resulting sequence, bn = 50/n2 , becoming arbitrarily small in
the long run.
These two rules may be used in combination. For example, the sequence
3 + n becomes arbitrarily large as n increases, so (by the Reciprocal Rule)
the sequence 1/(3 + n) converges to the limit 0. Thus (by the Constant
Multiple Rule) the sequence 100/(3 + n) also converges to the limit 0.
In more complicated formulas, you can look for quantities that become
small as n becomes large. For example, consider the sequence
5000
an = n. (5.4)
1 + 24(0.5)
Since 0.5 < 1, the quantity (0.5)n becomes arbitrarily small as n becomes
large. The same is true of 24(0.5)n (by the Constant Multiple Rule), so
1 + 24(0.5)n tends to 1. In the long run, the values of an become
arbitrarily close to 5000. In other words, the sequence defined by
equation (5.4) converges to the limit 5000.
In arguments such as this, it is useful to recall the known behaviour for
large n of the ‘basic sequences’ r n and np .
46
SECTION 5 SEQUENCES AND LIMITS
(c) The given sequence can be written (dividing top and bottom of the
fraction by n) as
an = .
5/n + 10
Now 5/n converges to 0 for large n (1/n is of the form np with
p = −1), so 5/n + 10 converges to 10. Hence, for large n, the sequence
1
an will behave like 10 n, which is not convergent. We conclude that an
itself is not convergent either.
Comment
In part (c), since neither n nor n2 converges, it would appear that an is the
ratio of two sequences that do not converge. From this fact alone, we can
make no deductions about the convergence of an . To make progress, we
rewrite an in the form given, to produce a sequence in the denominator
that does converge.
47
CHAPTER B1 MODELLING WITH SEQUENCES
For each of the sequences below, decide whether or not it converges and, if
it does, to what limit. In each case n = 1, 2, 3, . . . .
1
(a) an =
2 + 3n
(b) an = 5 + n3
100
(c) an = n
4 + 20(0.6)
(d) an = 3 + (−1)n
60n
(e) an =
3 + 5n
Solutions are given on page 57.
1 − rn+1
ari = a (r
= 1). (5.5)
i=0
1−r
Now the terms ari (i = 0, 1, 2, . . .) form a geometric sequence. From this
sequence, it is possible to construct a new sequence sn , each term of which
is a sum of consecutive terms from the sequence ar i . Thus we have
s0 = a, s1 = a + ar, s2 = a + ar + ar2 ,
and so on. The nth term of the new sequence is given by
n
�
sn = ari (n = 0, 1, 2, . . .),
i=0
which is the sum of a finite geometric series. Does the new sequence sn
have a limit? If it does, then lim sn exists and we write
n→∞
∞
�
lim sn = ari .
n→∞
i=0
48
SECTION 5 SEQUENCES AND LIMITS
which is unbounded for large n. If r = −1, then the terms of sn take the
alternate values a and 0. Hence convergence can occur only if |r| < 1. In
this case, as recalled in Subsection 5.2, lim rn = 0. From equation (5.5), it
n→∞
follows that
�∞ � � The sequence rn+1 is the
1 − rn+1 a
ari = lim a = (|r| < 1). (5.6) sequence rn without the first
n→∞ 1−r 1−r
i=0 term r0 , so
lim rn+1 = lim rn .
n→∞ n→∞
Example 5.3 Sum of an infinite geometric series
We derived equation (5.6) by taking the limit for large n of equation (5.5),
for |r| < 1. An alternative approach to summing an infinite geometric
series comes in handy when the series concerned is in the form of a
recurring decimal. This approach permits us to find the fraction which is
equivalent to any recurring decimal, as the next example illustrates.
49
CHAPTER B1 MODELLING WITH SEQUENCES
Note that this decimal can be Find the fraction equivalent to 0.123 123 123 . . . .
considered as an infinite
a = 123 × 10−3 and r = 10−3 . Put s = 0.123 123 123 . . . . The repeating group, ‘123’, is 3 digits long, so
Recognition of these values, we multiply s by 103 , to obtain
followed by direct application
of equation (5.6), is not the 1000s = 123.123 123 123 . . . = 123 + s.
quickest way to find the Hence we have
equivalent fraction, though 123 41
you can check here that s= = .
equation (5.6) gives the same
999 333
result.
50
SECTION 5 SEQUENCES AND LIMITS
In Subsection 5.3 you saw how an infinite geometric series can be summed,
provided that the common ratio is less than 1 in magnitude. Some infinite
series which are not geometric can also be summed, and many fascinating
formulas arise in this way; for example,
∞
� (−1)i 1 1
=1− 3
+ 5
−
17 + · · ·
= 14 π,
i=0
2i + 1
∞
� (−1)i+1 1
=1− 2
+
13 − 1
4
+ · · ·
= ln 2.
i=1
i
∞
�
If ai (i = 1, 2, 3, . . .) is a sequence, then the sum ai can exist only if ai
i=1
becomes smaller and smaller in the long term, that is, only if lim ai = 0.
i→∞
However, this requirement of convergence to 0 for the sequence ai is not on
its own sufficient to guarantee the existence of a sum for the infinite series.
For example, the sequence 1, 12 ,
13 ,
14 , . . . converges to 0 but, as indicated
below, there is no sum for the infinite series 1 + 12 + 13 + 14 + · · · .
A non-existent sum
The following brief argument (by contradiction) shows that the infinite
series
1 1
1+ 2
+ 3
+
14 + · · ·
has no sum.
Suppose that the sum of the infinite series exists and is equal to s; that is,
1 1
s=1+ 2
+ 3
+
14 + · · · .
Multiply both sides by
12 , to obtain These manipulations of
1 1 1 infinite series can be shown to
2
s = 2
+ 4
+
16 + 1
8
+ ···. (5.8) be valid if the series have
The right-hand side is a subseries of the original. On subtracting it from sums.
the original, we have
1 1 1 1
2
s =1+ 3
+ 5
+ 7
+ ···. (5.9)
But now each term in series (5.9) is larger than its counterpart in
series (5.8): 1 > 21 ,
13 > 14 , and so on. Hence the sum of series (5.9) must be
greater than that of series (5.8) – but each is equal to 12 s! This
contradiction shows that the original series has no sum.
Summar y of Section 5
This section has introduced:
the concepts of convergence and limit for sequences, with the notation
lim an for the limit of a convergent sequence an ;
n→∞
for sequences defined by recurrence systems, the method of identifying
potential limit values by finding all constant sequences that satisfy the
recurrence relation;
for sequences an defined by closed-form formulas, methods of
reasoning based on how individual parts of the formula behave as n
becomes large;
the Reciprocal Rule and Constant Multiple Rule for sequences;
51
CHAPTER B1 MODELLING WITH SEQUENCES
Exercise 5.2
The following sequences are given by closed-form formulas. In each case,
decide whether or not the sequence converges and, if it does, find the limit.
In each case, n = 1, 2, 3, . . . .
16n + 5
(a) an = 4 + (−0.5)n (b) an = n2 + 1/n (c) an =
2n − 1
Exercise 5.3
The following sequences are given by recurrence systems. In each case, use
the method of Example 5.1 and Activity 5.1(b) to find all possible limit
values for the sequence. Then use methods from Chapter A1 to find a
closed form for the sequence, and use this to determine the long-term
behaviour of the sequence. In each case, n = 0, 1, 2, . . . .
(a) x0 = 3, xn+1 = 2xn
(b) x0 = 3, xn+1 = 0.2xn
(c) x0 = 3, xn+1 = 0.3xn + 140
Exercise 5.4
What is the sum of the following infinite series?
1+ 1
9
+ ( 19 )2 + ( 19 )3 + · · ·
Exercise 5.5
Find the fraction equivalent to 0.513 513 513 . . . .
52
Summar y of Chapter B1
In this chapter you have seen sequences put to use in the development of
models for the economics of car ownership and for variations in animal
populations. In the context of the logistic model you saw that, while
numerical investigation has its uses, prior reasoning may profitably reduce
the scope of the task.
Logistic recurrence sequences exhibit an interesting variety of long-term
behaviour. The topic of what happens to sequences in the long term was
studied further from the mathematical point of view, with reference to
convergence and limits. These concepts may be applied both to sequences
and to series. The sum of either a finite or an infinite series may be
expressed concisely using sigma notation.
Modelling skills
Appreciate the roles of the various stages of the modelling cycle, as
summarised in Figure 1.1.
Interpret mathematical information obtained from a model in terms of
the original purpose or situation.
Evaluate a model in the light of what it predicts and how this
compares with the real situation being modelled.
Be aware of the possible need to revise a model on the basis of
evaluation of it.
53
CHAPTER B1 MODELLING WITH SEQUENCES
Mathematical skills
Find the parameter values for the logistic model which correspond to
given information about the proportionate birth and death rates
(or growth rate) of a population.
Obtain information about a sequence generated by the logistic
recurrence relation, for given values of its parameters.
Recognise various possible forms of long-term behaviour for sequences,
in particular: convergence to a limit; cycling; chaotic behaviour; terms
which become arbitrarily large.
Apply and manipulate sigma notation, where appropriate.
n
�
Apply the formula i = 12 n(n + 1), where appropriate.
i=1
Apply, where appropriate, either of the formulas
n
� � �
i 1 − rn+1
ar = a 1),
(r =
i=0
1−r
∞
� a
ari = (|r| < 1).
i=0
1−r
Find the fraction which is equivalent to a given recurring decimal.
Mathcad skills
Calculate and graph terms of a logistic recurrence sequence, given a
starting value and values for the parameters.
Interpret the outcomes obtained from such calculations, in particular
as regards the long-term behaviour of the sequence.
Ideas to be aware of
The principle of simplicity (in mathematical modelling).
That the prior application of graphical or algebraic reasoning may
reduce the scope needed for a numerical investigation.
54
Solutions to Activities
= 600 × 4 + 250(1 + 2 + 3 + 4)
4
ci = (c0 + di)
� i=1 i=1
= 600 × 4 + 250 i. n
i=1
�
= c0 n + d i (equation (1.4))
i=1
Solution 1.3 = c0 n + 12 dn(n + 1) (equation (1.6))
� �
From equation (1.6), the required sum is = n c0 + 12 d(n + 1) .
100
� (d) From the result of part (a), we have
1
i= 2 × 100 × 101 = 5050.
vi = 16 000(0.85)i (i = 0, 1, . . . , n + 1).
55
CHAPTER B1 MODELLING WITH SEQUENCES
56
SOLUTIONS TO ACTIVITIES
(b) Since the proportionate growth rate has the Solution 5.2
form r(1 − P/E) for a population size P
(a) As n becomes large, 3n becomes arbitrarily
(equation (3.2)), we have
� � � � large, so this is also true of the sequence 2 + 3n.
100 400 By the Reciprocal Rule, therefore, the sequence
1.2 = r 1 − , −0.6 = r 1 − .
E E an = 1/(2 + 3n) converges, with limit 0; that is,
lim an = 0.
On dividing through each equation by r and n→∞
then rearranging, we obtain a pair of (b) As n becomes large, n3 and hence also 5 + n3
simultaneous linear equations in 1/r and 1/E: become arbitrarily large. The sequence
1.2 100 0.6 400 an = 5 + n3 does not converge.
+ = 1 and − + = 1.
r E r E
(c) The quantity (0.6)n becomes arbitrarily small as
Eliminating the 1/E term, by subtracting the
n becomes large. This is also true of 20(0.6)n , by
second equation from four times the first
the Constant Multiple Rule. Hence the sequence
equation, gives
4 + 20(0.6)n becomes close to 4 in the long term.
5.4 For large n, therefore, an = 100/(4 + 20(0.6)n )
= 3, that is, r = 1.8. becomes close to 100/4 = 25. In other words,
r
the sequence an converges, and lim an = 25.
On substituting this value for r into the first n→∞
equation, we obtain (d) The values given by this formula are (starting
1.2 100 with n = 1)
+ = 1;
1.8 E 2, 4, 2, 4, 2, . . . .
that is,
57
Solutions to Exercises
58
SOLUTIONS TO EXERCISES
Eliminating the 1/r term, by subtracting 3 times (i) If 0 < r < 1, then 0 < 1 − r < 1, so (1 − r)n
the first equation from 4 times the second, gives is positive and decreases towards 0 as n
21.08 − 9.06 increases. Since Qn is the amount by which Pn
= 1; that is, E = 12.02. is below E, this predicts that the values of Pn
E
are always less than E, but approach E more
On substituting this value for E into the first
and more closely as n increases.
equation, we obtain
(ii) If 1 < r < 2, then −1 < 1 − r < 0, so
0.20 3.02
+ = 1; (1 − r)n is alternately positive and negative,
r 12.02
59
CHAPTER B1 MODELLING WITH SEQUENCES
60
Index
basic sequences 46
carrying capacity 32
chaotic behaviour 40
convergent sequence 42
cycling behaviour 40
divergent sequence 42
exponential model 22
4-cycle 40
geometric model 22
limit of a sequence 42
limits of a sum 11
logistic model 29
long-term behaviour 40
Malthusian model 27
modelling cycle 7
population density 26
principle of simplicity 10
reciprocal rule 46
Ricker model 41
series 11
sigma notation 11
sum notation 11
2-cycle 40
61