Académique Documents
Professionnel Documents
Culture Documents
Mathematics
STUDY GUIDE
PART B
Contents
Part B
Finding eigenvalues 40
Leslie matrices 49
References 57
TOPIC 4
One of the most well-known models of population growth uses aspects of linear
algebra, which you studied last semester in your unit EMM302 Linear and
Abstract Algebra.
The Leslie model (often referred to as the Leslie matrix) was developed by P.H.
Leslie in 1945. It is used in population ecology to model the changes in
population over time of an age-structured population.
FIGURE 4.1
Special case when
Ax = λx Ax = λx
Av
x v
In Figure 4.1, it looks as though the vector x has been “stretched out” by a
factor of about three and a half so that it is longer but still lies on the same line.
[Note that for our purposes here we will only consider vector spaces over the
field of real numbers, so whenever we say that λ is a scalar we just mean that λ
is a real number.]
38 EMM400 TOPIC4
EXAMPLE 4.1 ⎡2 − 4 ⎤ ⎡1 ⎤ ⎡ − 6 ⎤ ⎡1 ⎤
⎢− 1 ⎥ ⎢ ⎥ = ⎢ ⎥ ≠ λ ⎢ ⎥ for any scalar λ, but
⎣ − 1 ⎦ ⎣ 2 ⎦ ⎣ − 3⎦ ⎣2⎦
⎡2 − 4⎤ ⎡ 4 ⎤ ⎡ 12 ⎤ ⎡4⎤
⎢− 1 ⎥ ⎢ ⎥ = ⎢ ⎥ = 3⎢ ⎥ .
⎣ − 1 ⎦ ⎣− 1⎦ ⎣− 3⎦ ⎣− 1⎦
⎡ 2 − 4⎤ ⎡1 ⎤
So the matrix ⎢ ⎥ moves the vector ⎢ ⎥ onto a different line in the
⎣− 1 − 1 ⎦ ⎣ 2⎦
⎡4⎤
plane but keeps the vector ⎢ ⎥ on the same line while changing only its
⎣−1⎦
length.
EIGENVALUES & If A is a square matrix and Ax = λx for some scalar λ and some non-zero
EIGENVECTORS vector x, then
▪ λ is called an eigenvalue of A; and
▪ x is called an eigenvector corresponding to λ.
The terms eigenvalue and eigenvector come from the German word “eigen”
meaning “own”. So we see that if Ax = λx, the vector x is moved onto its “own”
direction.
CHARACTERISTIC The eigenvalues associated with a matrix A are also called characteristic values
VALUES & or characteristic roots associated with A, and the eigenvectors are called the
CHARACTERISTIC characteristic vectors.
VECTORS
⎡4 − 1⎤ ⎡4 − 1⎤
EXAMPLE 4.2 Let A = ⎢ ⎥ . Then Ax = ⎢2 1 ⎥ x.
⎣2 1 ⎦ ⎣ ⎦
⎡1 ⎤ ⎡4 − 1⎤ ⎡1⎤ ⎡2⎤ ⎡1 ⎤
Also A ⎢ ⎥ = ⎢ ⎥ ⎢ ⎥ = ⎢ ⎥ =2 ⎢ 2⎥ .
⎣ 2⎦ ⎣ 2 1 ⎦ ⎣ 2⎦ ⎣ 4⎦ ⎣ ⎦
⎡1⎤ ⎡1 ⎤
So 3 and 2 are eigenvalues with corresponding eigenvectors ⎢ ⎥ and ⎢ ⎥ ,
⎣1⎦ ⎣ 2⎦
respectively.
⎡ 2⎤
Notice that if we took the vector ⎢ ⎥ , we would get a similar situation, namely
⎣ 2⎦
⎡ 2⎤ ⎡4 − 1⎤ ⎡2⎤ ⎡6⎤ ⎡ 2⎤
A ⎢ ⎥ = ⎢ ⎥ ⎢ ⎥ = ⎢ ⎥ = 3⎢ ⎥.
⎣ 2⎦ ⎣ 2 1 ⎦ ⎣ 2 ⎦ ⎣6 ⎦ ⎣ 2⎦
MODELLING POPULATION GROWTH 39
⎡ 2⎤
So ⎢ ⎥ is also an eigenvector corresponding to the eigenvalue 3.
⎣ 2⎦
⎡1⎤
(
then A α ⎢ ⎥ ) = α A ⎡⎢1⎤⎥ = 3α ⎡⎢1⎤⎥ .
⎣1⎦ ⎣1⎦ ⎣1⎦
⎡1⎤
So, for all α ≠ 0, α ⎢ ⎥ is an eigenvector corresponding to the eigenvalue 3.
⎣1⎦
⎡1 ⎤
Similarly, for all β≠ 0, β ⎢ ⎥ is also an eigenvector corresponding to the
⎣ 2⎦
eigenvalue 2.
⎡1⎤ ⎡1 ⎤
In Example 4.1, we verified the fact that ⎢ ⎥ and ⎢ ⎥ were eigenvectors by
⎣1⎦ ⎣ 2⎦
⎡1⎤ ⎡1⎤ ⎡1 ⎤ ⎡1 ⎤
showing that A ⎢ ⎥ = 3 ⎢ ⎥ and A ⎢ ⎥ = 2 ⎢ ⎥ . In the process of doing
⎣1⎦ ⎣1⎦ ⎣ 2⎦ ⎣ 2⎦
this, we also found the two corresponding eigenvalues 3 and 2.
In the next example, we will illustrate how we can go about verifying that a
particular scalar is an eigenvalue for a given matrix A , and find its
corresponding eigenvectors.
⎡2 − 5 5 ⎤
EXAMPLE 4.3 ⎢ ⎥
In order to show that 2 is an eigenvalue of A = 0 3 − 1 we will try to
⎢ ⎥
⎢⎣0 − 1 3 ⎥⎦
solve Ax = 2x.
⎡ 0 5 − 5⎤
⎢
Now (2I − A) = 0 − 1 1
⎥
⎢ ⎥
⎢⎣0 1 − 1⎥⎦
⎡0 5 − 5 0⎤
⎢
We can apply row operations to the augmented matrix 0 − 1 1 0⎥⎥
⎢
⎢⎣0 1 − 1 0⎥⎦
(see pp. 15–16 of your EMM302 Linear and Abstract Algebra Study Guide).
40 EMM400 TOPIC4
⎡0 5 − 5 0⎤ ⎡0 5 − 5 0⎤
⎢0 − 1 1 0⎥⎥ R3′ = R2 + R3
⎢0 − 1 1 0⎥⎥
⎢ ⎢
⎢⎣0 1 − 1 0⎥⎦ ⎢⎣0 0 0 0⎥⎦
⎡0 5 − 5 0 ⎤ ⎡0 1 − 1 0 ⎤
R2′ = R1 + 5R2
⎢0 0 0 0⎥⎥ R1′ = 1 R1
⎢0 0 0 0⎥⎥ .
⎢ 5
⎢
⎢⎣0 0 0 0⎥⎦ ⎢⎣0 0 0 0⎥⎦
⎡ x1 ⎤
⎢ ⎥
So the solutions for Ax = 2x are all the vectors of the form x = ⎢ x3 ⎥ ,
⎢x ⎥
⎣ 3⎦
⎡1 ⎤ ⎡0 ⎤
⎢ ⎥
which can also be written as x = x1 0 + x3
⎢1 ⎥ .
⎢ ⎥ ⎢ ⎥
⎢⎣0⎥⎦ ⎢⎣1 ⎥⎦
THEOREM 4.1 If λ is an eigenvalue of matrix A, then the set of all eigenvectors that
correspond to λ is a vector space.
PROOF The set of eigenvectors that correspond to λ is the null space of the matrix
(λI – A) − that is, the set of solutions of (λI – A)x = 0.
(See Topic 3, in your EMM302 Linear and Abstract Algebra Study Guide.)
EIGENSPACES The set of all eigenvectors corresponding to the eigenvalue λ is a vector space
called the eigenspace of λ.
Finding eigenvalues
In the previous two examples, we either started with some known eigenvectors
and found the corresponding eigenvalues (as in Example 4.2) or started with
some known eigenvalues and found the corresponding eigenvectors (as in
Example 4.3).
But what do we do if we know neither the eigenvectors nor the eigenvalues?
We will now show how to find the eigenvalues of a matrix when we don’t have
any other information.
MODELLING POPULATION GROWTH 41
⎡2 − 5 5 ⎤
EXAMPLE 4.4 ⎢
Let us try to find the eigenvalues of A = 0 3 − 1 .
⎥
⎢ ⎥
⎢⎣0 − 1 3 ⎥⎦
⎡λ − 2 5 −5 ⎤
⎢ λ −3 1 ⎥⎥ = (λ − 2){(λ − 3)(λ − 3) − 1}
det (λI − A) = det
⎢ 0
⎢⎣ 0 1 λ − 3⎥⎦
⎡− 2 − 5 5 0⎤ ⎡ 2 5 − 5 0⎤
⎢ ⎥ ⎢0 1 1 0⎥⎥ −
Applying row operations to
⎢ 0 − 1 − 1 0⎥ we get ⎢
⎢⎣ 0 − 1 − 1 0⎥⎦ ⎢⎣0 0 0 0⎥⎦
you should check this for yourselves.
⎡5 x3 ⎤ ⎡5 ⎤
⎢ ⎥ ⎢ ⎥
x = ⎢− x3 ⎥ = x3 − 1
⎢ ⎥
⎢x ⎥ ⎢⎣1 ⎥⎦
⎣ 3 ⎦
⎡5 ⎤
⎢ ⎥
Thus the eigenspace for λ = 4 is generated by the vector − 1 .
⎢ ⎥
⎢⎣1 ⎥⎦
42 EMM400 TOPIC4
Before trying some examples for yourselves, let us try another worked example
that is a bit more complicated in its calculations.
⎡− 2 3 1⎤
EXAMPLE 4.5 ⎢
Let A = 0 1 1 .
⎥
⎢ ⎥
⎢⎣ − 3 4 1⎥⎦
⎡λ + 2 − 3 −1 ⎤
⎢ λ − 1 − 1 ⎥⎥
So det (λI − A) = det
⎢ 0
⎢⎣ 3 − 4 λ − 1⎥⎦
⎡λ − 1 − 1 ⎤ ⎡ − 3 − 1⎤
= (λ + 2) det ⎢ ⎥ + 3 det ⎢ ⎥
⎣ − 4 λ − 1⎦ ⎣λ − 1 − 1⎦
= (λ + 2) (λ2 – 2λ +1 – 4) + 3(3 + λ – 1)
= λ3 – 7λ – 6 + 3λ + 6
= λ3 – 4λ
= λ(λ2 – 4) = λ(λ – 2) (λ + 2) .
So A has eigenvalues 0, 2, –2 .
(i) λ = 0
⎡λ + 2 − 3 − 1 0⎤ ⎡ 2 − 3 − 1 0⎤
⎢ 0 λ − 1 − 1 0⎥⎥ = ⎢⎢0 − 1 − 1 0⎥⎥
⎢
⎢⎣ 3 − 4 λ − 1 0⎥⎦ ⎢⎣3 − 4 − 1 0⎥⎦
⎡ 2 − 3 − 1 0⎤ ⎡ 2 − 3 − 1 0⎤
R3′ = 2R1 – 3R3
⎢0 − 1 − 1 0⎥ R3′ = R2 + R3
⎢0 − 1 − 1 0⎥ .
⎢ ⎥ ⎢ ⎥
⎢⎣0 1 1 0⎥⎦ ⎢⎣0 0 0 0⎥⎦
⎡− 1⎤
⎢ ⎥
So x = − 1 is an eigenvector for λ = 0.
⎢ ⎥
⎢⎣1 ⎥⎦
Notice that, as usual, once we have found one eigenvector for λ = 0, we know
⎡− 1⎤
⎢ ⎥
infinitely many eigenvectors – in this case all the scalar multiples of x = − 1
⎢ ⎥
⎢⎣1 ⎥⎦
⎡− 1⎤
⎢ ⎥
− that is, all vectors of the form x = α − 1 , where α is any scalar.
⎢ ⎥
⎢⎣1 ⎥⎦
(ii) λ = 2
⎡λ + 2 − 3 − 1 0⎤ ⎡ 4 − 3 − 1 0 ⎤
⎢ 0 λ − 1 − 1 0⎥⎥ = ⎢⎢0 1 − 1 0⎥⎥
⎢
⎢⎣ 3 − 4 λ − 1 0⎥⎦ ⎢⎣3 − 4 1 0⎥⎦
⎡ 4 − 3 − 1 0⎤ ⎡ 4 − 3 − 1 0⎤
R3′ =3R1 – 4R3
⎢0 1 − 1 0⎥ R3′ = 7R2 – R3
⎢0 1 − 1 0⎥
⎢ ⎥ ⎢ ⎥
⎢⎣0 7 − 7 0⎥⎦ ⎢⎣0 0 0 0⎥⎦
(iii) λ = −2
⎡λ + 2 − 3 − 1 0 ⎤ ⎡0 − 3 − 1 0 ⎤
⎢ 0 λ − 1 − 1 0⎥⎥ = ⎢⎢0 − 3 − 1 0⎥⎥
⎢
⎢⎣ 3 − 4 λ − 1 0⎥⎦ ⎢⎣3 − 4 − 3 0⎥⎦
⎡0 0 0 0⎤
R1′ =R1 –R2
⎢0 − 3 − 1 0⎥ etc.
⎢ ⎥
⎢⎣3 − 4 − 3 0⎥⎦
1
So we can make x3 arbitrary, x2 = − x3, and
3
1 1 4 5
x1 = (3x3 + 4 x2 ) = (3x3 − x3 ) = x3 .
3 3 3 9
⎡5 ⎤
⎢ ⎥
So x = − 3 is an eigenvector for λ = −2 and the set of all eigenvectors
⎢ ⎥
⎢⎣9 ⎥⎦
⎡5 ⎤
⎢ ⎥
corresponding to λ = −2 have the form x = α − 3 where α is any scalar.
⎢ ⎥
⎢⎣9 ⎥⎦
We should also note that not all matrices have eigenvalues, as can be seen in
the following example.
EXAMPLE 4.6
⎡ 0 3⎤
Let A = ⎢ ⎥.
⎣ − 2 0⎦
⎡ λ 3⎤
Then det (λI − A) = det ⎢ ⎥ = λ2 + 6 = 0 has no real solutions.
⎣− 2 λ ⎦
So the matrix A has no real eigenvalues.
ACTIVITY 4.1 Find the eigenvalues and corresponding eigenvectors for each of the
following matrices (when they exist).
⎡1 2 3 ⎤
⎢ ⎥ ⎡ 1 1⎤ ⎡0 1 ⎤ ⎡ 0 1⎤ ⎡ a 0⎤
a) 0 2 4 ; b) ⎢ ⎥ ; c) ⎢ ⎥ ; d) ⎢ ⎥ ; e) ⎢ ;
⎢ ⎥
⎣ − 1 3⎦ ⎣ 1 0⎦ ⎣ − 1 0⎦ ⎣ 0 b⎥⎦
⎢⎣0 0 2⎥⎦
⎡− 1 0 0 0⎤ ⎡0 0 0 1⎤
⎢0 ⎥ ⎢0 0⎥⎥
⎡2 0⎤ ⎢ 0 0 0⎥ 0 1 ⎡0 1⎤
f) ⎢ ⎥ ; g) ; h) ⎢ ; i) ⎢ ⎥ ;
⎣1 2 ⎦ ⎢0 0 0 1⎥ ⎢0 1 0 0⎥ ⎣− 1 − 1⎦
⎢ ⎥ ⎢ ⎥
⎣0 0 1 0⎦ ⎣1 0 0 0⎦
⎡ cosθ sin θ ⎤
j) ⎢ ⎥ .
⎣− sin θ cosθ ⎦
MODELLING POPULATION GROWTH 45
Do you think you would always get the same answer? Why or why not?
Notice that in Example 4.5 the characteristic polynomial of A, det (λI − A),
was a polynomial of degree 3.
This will always be the case whenever we have a 3 x 3 matrix A – can you see
why?
Finally, the constant term in the polynomial is the value of the polynomial
when λ = 0.
ACTIVITY 4.3 ⎡− 2 3 1⎤
⎢
In Example 4.5, we had A = 0 1 1 .
⎥
⎢ ⎥
⎢⎣ − 3 4 1⎥⎦
Check that the statements above about the co-efficients of λn, λn−1 , and the
constant term are correct.
THEOREM 4.3 The eigenvectors that correspond to distinct eigenvalues of a matrix are linearly
independent.
PROOF Suppose a matrix A has distinct eigenvalues λ1, λ2, …, λk , which have
eigenvectors v1, v2, …, vk respectively.
Suppose one of these, say v1, is a linear combination of the others.
Thus v1 = c2v2 + · · · + ckvk , for some scalars c2 , …, ck .
Then Av1 = Ac2v2 + · · · + Ackvk = c2λ2v2 + · · · + ckλkvk.
But λ1v1 = λ1c2v2 + · · · + λ1ckvk .
So 0 = c2(λ1 − λ2)v2 + · · · + ck(λ1 − λk)vk .
The eigenvalues are distinct, so none of the brackets are 0. Also not all of the
cis can be 0, otherwise v1 would be zero.
Therefore one of v2, …, vk is a linear combination of the others.
We can repeat this argument until only one eigenvector remains. Then this last
eigenvector must be zero, which is a contradiction.
Therefore all of the eigenvectors v1, v2, …, vk are lineraly independent.
THEOREM 4.4 The eigenvalues of a diagonal matrix are its diagonal entries.
PROOF ⎡ d1 0 ... 0⎤
⎢ ⎥
0 d2 ... 0 ⎥
Let D = ⎢ be a diagonal matrix.
⎢ ... ... ... ... ⎥
⎢ ⎥
⎣⎢ 0 0 ... d n ⎦⎥
MODELLING POPULATION GROWTH 47
⎡ λ − d1 0 ... 0 ⎤
⎢ ⎥
0 λ − d2 ... 0
= det ⎢ ⎥ = (λ − d1)(λ − d2) … (λ − dn).
⎢ ... ... ... ... ⎥
⎢ ⎥
⎣⎢ 0 0 ... λ − d n ⎦⎥
EXAMPLE 4.7 ⎡− 2 3 1⎤
⎢ ⎥
Let A = 0 1 1 be the matrix we used in Example 4.5.
⎢ ⎥
⎢⎣ − 3 4 1⎥⎦
⎡ 6 − 2 − 4⎤
1⎢
−1
You should find B = –
⎢ 3 − 7 − 4⎥⎥ .
8
⎢⎣− 1 1 0 ⎥⎦
⎡ − 1 1 5 ⎤ ⎡0 0 0 ⎤ ⎡ 6 − 2 − 4⎤
⎢ ⎥ ⎢ ⎥ 1 ⎢ ⎥
BDB = − 1 1 − 3 0 2 0 (– ) 3 − 7 − 4
−1
⎢ ⎥ ⎢ ⎥ 8 ⎢ ⎥
⎢⎣ 1 1 9 ⎥⎦ ⎢⎣0 0 − 2⎥⎦ ⎢⎣− 1 1 0 ⎥⎦
⎡0 2 − 10⎤ ⎡ 6 − 2 − 4⎤
⎢ 1 ⎢
= 0 2
⎢ 6 ⎥ (– ) ⎢ 3 − 7 − 4⎥⎥
⎥
8
⎢⎣0 2 − 18⎥⎦ ⎢⎣− 1 1 0 ⎥⎦
⎡16 − 24 − 8⎤ ⎡− 2 3 1⎤
1⎢
=–
⎢ 0 − 8 − 8⎥⎥ = ⎢⎢ 0 1 1⎥⎥ = A.
8
⎢⎣24 − 32 − 8⎥⎦ ⎢⎣ − 3 4 1⎥⎦
ACTIVITY 4.5 ⎡− 1 1 5 ⎤ ⎡0 0 0 ⎤
⎢ ⎥ ⎢ ⎥
Let B = − 1 1 − 3 and D = 0 2 0 , as in Example 4.7.
⎢ ⎥ ⎢ ⎥
⎢⎣ 1 1 9 ⎥⎦ ⎢⎣0 0 − 2⎥⎦
⎡2 −5 5⎤
ACTIVITY 4.6 In Examples 4.3 and 4.4 we found that the matrix A = 0
⎢ 3 − 1⎥⎥ has
⎢
⎢⎣0 − 1 3 ⎥⎦
⎡1 ⎤ ⎡0 ⎤ ⎡5 ⎤
⎢ ⎥
eigenvalues 2, 2, 4 with corresponding eigenvectors 0 ,
⎢1 ⎥ and ⎢− 1⎥ .
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢⎣0⎥⎦ ⎢⎣1 ⎥⎦ ⎢⎣1 ⎥⎦
⎡1 0 5 ⎤ ⎡2 0 0⎤
⎢ ⎥
Let B = 0 1 − 1 and D =
⎢0 2 0 ⎥ .
⎢ ⎥ ⎢ ⎥
⎢⎣0 1 1 ⎥⎦ ⎢⎣0 0 4⎥⎦
Leslie matrices
The Leslie model is one of the most used models in population ecology. It has
been used to describe the growth of populations of a wide variety of animals,
such as different species of fish, rabbits, beetles, and humans. In this section,
we will look at the model and its applications to a few examples, not all of
which are necessarily about living things.
The Leslie model uses a number of assumptions:
• The population can be divided into discrete categories;
• there is a known initial population distribution based on these categories;
• the population is time dependent and we can use discrete time intervals to
describe the population (for example, years);
• there is a known probability of movement from one category to another
from one time interval to the next; and
• (when dealing with animals, etc) there is a known probability of
reproduction for each age group.
As we have seen, matrix multiplication is a way of expressing how data may
change. For example, the entries in a vector xk might represent the population
of various groups of animals in an ecological system at time k, and a square
matrix A might represent how the populations change from one period to the
next.
Thus xk+1 = Axk for each k. If the initial population vector is x0 what
happens to xk in the long term?
EXAMPLE 4.8 In the year 1202, Fibonacci became interested in the reproduction of rabbits. He
FIBONACCI’S created an imaginary set of ideal conditions under which rabbits could breed,
RABBIT and posed the question, “How many pairs of rabbits will there be a year from
PROBLEM now?” The ideal set of conditions was a follows:
1. You begin with one male rabbit and one female rabbit. These rabbits have
just been born
2. A rabbit will reach sexual maturity after one month.
3. The gestation period of a rabbit is one month.
4. Once it has reached sexual maturity, a female rabbit will give birth every
month.
5. A female rabbit will always give birth to one male rabbit and one female
rabbit.
6. Rabbits never die.
So how many male/female rabbit pairs are there after one year (12 months)?
Month 1 – The pair of rabbits have mated but have not given birth. Therefore,
there is still only one pair of rabbits.
Month 2 – The first pair of have gives birth to another pair, making two pairs in
all.
Month 3 – The original pair gives birth again, and the second pair mate, but do
not give birth, making three pairs.
Month 4 – The original pair give birth, the pair born in month 2 give birth. The
pair born in month 3 mate, but do not give birth. This makes two new pair, for a
total of five pair.
Month 5 – Every pair that was alive two months ago gives birth. This makes
three new pair, for a total of eight.
The rabbit population creates the sequence: 1, 1, 2, 3, 5, ...
We could look at this as follows:
Let Fk = number of pairs of rabbits in the kth month
= (number of pairs of rabbits in the (k – 1) th month) +
(number of pairs of rabbits born in the (k – 1) th month)
= (number of pairs of rabbits in the (k – 1) th month) +
(number of pairs of rabbits in the (k – 2) th month)
= Fk–1 + Fk–2
This relation, Fk = Fk–1 + Fk–2 is known as Fibonacci’s relation.
The sequence satisfying the following conditions:
F0 = 0; F1 = 1; and, for all k > 1, Fk = Fk–1 + Fk–2 is known as the Fibonacci
sequence.
It’s first few terms are 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, ….
For k ≥ 1, the Fibonacci sequence describes the number of pairs of rabbits in the
problem above.
The Fibonacci sequence is very famous and has many fascinating properties – if
you are interested you can easily find out more about it on the World Wide
Web.
Here however, we will try to describe the population growth of the rabbits
using matrices.
In this problem, we can think of the rabbit population as having two age groups
– young and adult.
We can represent the number of rabbits at time k > 1 in each age group as a
⎡ Fk − 2 ⎤
vector xk = ⎢ ⎥ , where the first entry Fk–2 repesents the number of young
⎣ Fk −1 ⎦
rabbits at time k, and the second entry Fk–1 repesents the number of adult rabbits
⎡1⎤ ⎡0 ⎤
at time k, with the added condition that x0 = ⎢ ⎥ and x1 = ⎢ ⎥ .
⎣0 ⎦ ⎣1⎦
MODELLING POPULATION GROWTH 51
It is easy to see that we can no wrepresent the population growth over time as
follows:
⎡0 1⎤
xk+1 = Axk , where A = ⎢ ⎥ , so
⎣1 1⎦
⎡0 1⎤ ⎡ Fk − 2 ⎤ ⎡ Fk −1 ⎤
xk+1 = ⎢ ⎥⎢ ⎥ =⎢ ⎥ , which shows that at time k + 1, the
⎣1 1⎦ ⎣ Fk −1 ⎦ ⎣ Fk − 2 + Fk −1 ⎦
number of young rabbits is Fk–1, while the number of adult rabbits is Fk–2, as
expected.
Notice here the significance of each of the entries in the matrix A, which, for
⎡a b ⎤
convenience, we will rewrite here as A = ⎢ ⎥.
⎣c d ⎦
a = 0 represents the fact that none of the Fk–2 young rabbits at time k will have
reproduced yet;
b = 1 represents the fact that all of the Fk–1 adult rabbits at time k will have
reproduced;
c = 1 represents the fact that all of the Fk–2 young rabbits at time k will be
adults at time k; and
d = 1 represents the fact that all of the Fk–1 adult rabbits at time k will still be
adults at time k.
ACTIVITY 4.7 Use the matrix A and the other details from Example 4.8 to find x0, x1, x2,
x3, x4, x5, and in each case check that the total number of rabbits in each
month is as expected.
AGE In general, population problems can be represented through the use of an age
DISTRIBUTION distribution vector and a transition matrix.
VECTORS &
⎡ x1 ⎤ Number in the first age group
TRANSITION
⎢ ⎥
MATRICES ⎢ x2 ⎥ Number in the second age group
⎢.⎥ .
The age distribution vector x = ⎢ ⎥
⎢.⎥ .
⎢.⎥ .
⎢ ⎥
Number in the nth age group
⎣⎢ xn ⎦⎥
⎡b1 b2 b3 ... bn−1 bn ⎤
⎢ ⎥
⎢ s1 0 0 ... 0 0⎥
⎢0 s2 0 ... 0 0⎥
and the transition matrix A = ⎢ ⎥ .
⎢0 0 s3 ... 0 0⎥
⎢ ... ⎥
⎢ ⎥
⎣⎢ 0 0 0 ... sn−1 sn ⎦⎥
ACTIVITY 4.8 In Example 4.8, the matrix A represents the transition matrix for the
rabbit problem. Explain its entries in terms of the definition of the
transition matrix above.
EXAMPLE 4.9 Suppose a species of animals lives at most four years and the number of males
is a fixed proportion of the number of females.
Let xi = number of females of age (i–1) to age i years, for i = 1, 2, 3, 4.
For i = 1, 2, 3, 4, let bi = probability that a female in the age group i produces
a female offspring (i.e. the average number of offspring per female in this age
group).
For i = 1, 2, 3, let si = probability that a female in age group i survives to the
next age group. [Note that since these animals live at most four years, the
probablility that a female in age group 4 survives is zero – i.e. s4 = 0.]
The following data is given:
1 1 3 1
b1 = , b2 = , b3 = , b4 = ,
10 2 4 4
4 3 1
s1 = , s2 = , s3 = , s4 = 0 .
5 4 3
We want to predict the female population after one year, two years, etc.
⎡ x1 ⎤
⎢ ⎥
x2
Let x = ⎢ ⎥ be the current age distribution vector.
⎢ x3 ⎥
⎢ ⎥
⎢⎣ x4 ⎥⎦
After one year, we will have
1 1 3 1
x1′ = x1 + x2 + x3 + x4
10 2 4 4
4
x2′ = x1
5
3
x3′ = x2
4
1
x4′ = x3 .
3
MODELLING POPULATION GROWTH 53
⎡1 1 3 1⎤
⎢10 2 4 4⎥
⎢4 ⎥
So the transition matrix A = ⎢ 0 0 0 ⎥ and x′ = Ax.
⎢5 ⎥
⎢0 3
0 0⎥
⎢ 4 ⎥
⎢ 1 ⎥
⎢0 0 0⎥
⎣ 3 ⎦
That is, the population distribution after one year is given by Ax, after two years
by A(Ax) = A2x, after three years by A3x, etc.
⎡1200⎤
⎢1000⎥
Let the initial population be x0 = ⎢ ⎥.
⎢ 800 ⎥
⎢ ⎥
⎣ 100 ⎦
⎡1174⎤
⎢1160⎥
and x2 = A x1 = ⎢ ⎥.
⎢ 600 ⎥
⎢ ⎥
⎣ 250 ⎦
⎡1000⎤
⎢ 800 ⎥
However, notice that if instead we have x0 = ⎢ ⎥
⎢ 600 ⎥
⎢ ⎥
⎣ 200 ⎦
⎡1000⎤
⎢ 800 ⎥
and x1 = A x0 = ⎢ ⎥ = x0 .
⎢ 600 ⎥
⎢ ⎥
⎣ 200 ⎦
So in this case we have a stable population distribution.
This corresponds to A having an eigenvalue 1 with corresponding
⎡1000⎤
⎢ 800 ⎥
eigenvector ⎢ ⎥.
⎢ 600 ⎥
⎢ ⎥
⎣ 200 ⎦
54 EMM400 TOPIC4
We will finish this section with a slightly different example where there is
movement between two categories of “population” but no births.
EXAMPLE 4.10 Suppose in a small town in Malaysia every adult owns either a car or a
motorcycle, but not both and that this year 30% of the population own a car and
70% own a motorcycle.
⎡ 0.3⎤
We could write this as x0 = ⎢ ⎥.
⎣0.7⎦
Suppose also that 85% of the car owners remain car owners next year and 15%
change to motorcycles, while 20% of motorcycle owners change to cars while
80% remain motorcycle owners, and that this pattern continues each year.
⎡0.85 0.20⎤
We could write this as A = ⎢ ⎥.
⎣0.15 0.80⎦
To find out what percentage of adults will own cars and motorcycles in the
long-term, we can look for the eigenvalues and eigenvectors of A.
⎡λ − 0.85 − 0.20 ⎤
⎥ = λ − 1.65λ + 0.65
2
Let 0 = det (λI − A) = det ⎢
⎣ − 0.15 λ − 0.80 ⎦
= (λ − 1)(λ − 0.65) .
Thus λ = 0.65, 1.
⎡ 0.15 − 0.2 0⎤ ⎡ − 3 4 0⎤ ⎡ 4⎤
⎢− 0.15 ⎥ and get ⎢ ⎥ . Hence x = k⎢ ⎥ .
⎣ 0.2 0⎦ ⎣ 0 0 0⎦ ⎣ 3⎦
⎡ 4⎤ ⎡1⎤
Since ⎢ ⎥ and ⎢ ⎥ are linearly independent, x0 is a linear combination of
⎣ 3⎦ ⎣−1⎦
these two vectors.
MODELLING POPULATION GROWTH 55
⎡ 4⎤ ⎡1⎤ ⎡4 1 ⎤ ⎡ c1 ⎤ ⎡ 0.3⎤
Thus x0 = c1 ⎢ ⎥ + c2 ⎢ ⎥ or ⎢ ⎥⎢ ⎥ = ⎢ ⎥ .
⎣ 3⎦ ⎣−1⎦ ⎣3 − 1⎦ ⎣c 2 ⎦ ⎣0.7 ⎦
−1
⎡ c1 ⎤ ⎡4 1 ⎤ ⎡0.3⎤ 1 ⎡ − 1 − 1⎤ ⎡ 0.3⎤ 1 ⎡ 1 ⎤
Hence ⎢ ⎥ = ⎢ ⎥ ⎢ ⎥ = −7 ⎢− 3 = .
⎣c 2 ⎦ ⎣3 − 1⎦ ⎣0.7⎦ ⎣ 4 ⎥⎦ ⎢⎣0.7 ⎥⎦ 7 ⎢⎣− 1.9⎥⎦
⎡ 4⎤ ⎡1⎤
So x0 = 17 ⎢ ⎥ − 17.9 ⎢ ⎥ .
⎣ 3⎦ ⎣−1⎦
⎡ 4⎤ ⎡1⎤ ⎡ 4⎤ ⎡1⎤
Then x1 = Ax0 = 17 A ⎢ ⎥ − 17.9 A ⎢ ⎥ = 17 ⎢ 3⎥ −
1.9
0.65 ⎢ ⎥,
⎣−1⎦ ⎣−1⎦
7
⎣ 3⎦ ⎣ ⎦
⎡ 4⎤ ⎡1⎤ ⎡ 4⎤ ⎡1⎤
x2 = 17 A ⎢ ⎥ − 17.9 0.65A ⎢ ⎥ = 17 ⎢ ⎥ − 1.9
0.652 ⎢ ⎥,
⎣−1⎦ ⎣−1⎦
7
⎣ 3⎦ ⎣ 3⎦
⎡ 4⎤ ⎡1⎤
and so on to xk = 17 ⎢ ⎥ − 17.9 0.65k ⎢ ⎥ .
⎣ 3⎦ ⎣−1⎦
ACTIVITY 4.10 This problem is similar to Example 4.9, but with only three age groups.
⎡v1 ⎤
⎢ ⎥
Let v = ⎢v2 ⎥ give the number of members in each of the three age groups and
⎢ v3 ⎥
⎣ ⎦
let Av give the population distribution after a year has elapsed.
⎡1 5 1⎤
⎢4 6 4⎥
⎢3 ⎥
Let Av = ⎢ 0 0⎥ v .
⎢4 ⎥
⎢0 2
0⎥
⎢⎣ 3 ⎥⎦
d) Write down the matrix B = [x1 , x2 , x3] – that is the matrix with ith column
xi for i = 1, 2, 3.
e) Find B –1 .
56 EMM400 TOPIC4
⎡λ1 0 0⎤
⎢ ⎥
f) –1
Show that B AB = ⎢ 0 λ2 0 ⎥ and call this matrix D.
⎢0 0 λ3 ⎥⎦
⎣
⎡16 16 4⎤ ⎡ 20 − 20 − 10 ⎤
1 ⎢ ⎥ 1 ⎛ −1⎞ n ⎢
n
A =
⎢ 12 12 3⎥ + ⎜ ⎟ ⎢− 30 30 15 ⎥⎥
30 30 ⎝ 2 ⎠
⎢⎣ 8 8 2⎥⎦ ⎢⎣ 40 − 40 − 20⎥⎦
⎡ −6 4 6 ⎤
1 ⎛ −1⎞ n ⎢ ⎥
+ ⎜ ⎟ ⎢ 18 − 12 − 18⎥
30 ⎝ 4 ⎠
⎢⎣− 48 32 48 ⎥⎦
⎡v1 ⎤
⎢ ⎥
j) If the initial population is v = ⎢v2 ⎥ , show that the population distribution
⎢ v3 ⎥
⎣ ⎦
⎡ 4⎤
4v1 + 4v2 + v3 ⎢ ⎥
approaches
30 ⎢3⎥ after enough time has elapsed.
⎢⎣2⎥⎦
[In fact, after 10 years the population distribution os correct to within 1%
of the population.]
⎡1⎤ ⎡0 ⎤ ⎡0 ⎤
d) no real eigenvalues; e) a, ⎢ ⎥ , b, ⎢ ⎥ ; f) 2, ⎢ ⎥ ;
⎣0 ⎦ ⎣1⎦ ⎣1 ⎦
MODELLING POPULATION GROWTH 57
References
Reiner, I., Introduction to Matrix Theory and Linear Algebra, Holt, Rinehart
and Winston, New York, 1971. Chapter 11.