Chapter 6 (CONT’): Application: Powers of Matrices

and their Applications.

1 Powers of Matrices.
NOTE: The key to understanding this topic is to know how do diagonalize a
given matrix, which you did in previous chapter. If you have forgotten, please
review this topic on your own before going any further!

Here is an example of what you must know how to do: Suppose you are
given the matrix  
4 −2 1
A =  2 0 1 .
2 −2 3

You should know how to find the characteristic polynomial

−λ3 + 7λ2 − 16λ + 12 = −(λ − 3)(λ − 2)(λ − 2);
 
1
the eigenvalue λ1 = 3 and the associated eigenvector v1 =  1 , and the
1
repeated eigenvalue    
1 −1
λ2 = 2 and the associated eigenvectors v2 =  1  and v3 =  0 .
0 2
With the three eigenvectors v1 , v2 , v3 , we form the matrix
 
1 1 −1
P = 1  1 0 ,
1 0 2
and with the three eigenvalues 3, 2, 2 we form the diagonal matrix
 
3 0 0
D =  0 2 0 .
0 0 2

1
With the matrix P , we perform suitable elementary row operations to obtain
its inverse, P −1 . In the case here,
 
2 −2 1
P −1 =  −2 3 −1  .
−1 1 0

With the setup above, we have

A = P DP −1 .
This is what is known as the diagonalization of A.
NOTE:
1. Compute A5 directly is very tedious
2. Compute D5 is quick since D is a diagonal matrix.
3. Since A has been diagonalised, we can compute A5 by using the formula
A5 = P D5 P −1 .

In general, for any positive integer k and n × n matrix A, we have

Ak = P Dk P −1
where  
λk11 0 ··· 0
 0 λk22 ··· 0 
 
D =
k
.. . . . ...... 
 . 
0 0 · · · λknn

2 Application of Linear Algebra to Population Dynamics

and the Predator-Prey Models.
TRANSITION MATRICES
We want to apply this method of computing Ak to the analysis of a certain
type of physical system that can be described by means of the following kind
of mathematical model.

2
 Suppose that the sequence

x0 , x1 , x2 , . . . , xk , . . . (∗)

of n−vectors is defined by its initial vector x0 and an n × n transition matrix

A in the following manner:

xk+1 = Axk for each k ≥ 0. (∗∗)

One example of the physical system is such as a population with n specified
subpopulation that evolves through a sequence of successive states described
by the vector
x0 , x1 , x2 , . . . , xk , . . .
Then our goal is to calculate the k − th state vector xk .

But using (**) repeatedly we find that

x1 = Ax0 , x2 = Ax1 = A2 x0 , x3 = Ax2 = A3 x0 ,

and in general that

xk = Ak x0 .
Then our task is to calculate the k−th power Ak of the transition matrix A.

In the problems we are going to study now, we basically have four steps.

For example: The constant population of 1 million persons of a metropolitan

is divided between a city and its suburbs.

Let Ck denote the city population and Sk denote the suburban population
after k years.

Suppose that each year 15% of the people in the city move to the suburbs,
whereas 10% of the people in the suburbs move to the city.

STEP 1. Set up the equations to describe the physical situation:

Ck+1 = (0.85)Ck + (0.10)Sk
.
Sk+1 = (0.15)Ck + (0.90)Sk

3
 STEP 2. Put the system in matrix form:
[ ] [ ][ ]
Ck+1 0.85 0.10 Ck
= .
Sk+1 0.15 0.90 Sk

[ ]
0.85 0.10
The coefficient matrix A = is called the transition matrix now.
0.15 0.90
[ ] [ ]
Ck Ck+1
If we set xk = , so that xk+1 = , then our system can be
Sk Sk+1
written in the compact form
xk+1 = Axk
[ ]
k C0
and hence xk = A x0 , where x0 is the so-called initial vector, .
S0

STEP 3. Diagonalize A: (no doubt the messiest part!)

[ ] [ ] [ ]
2 −1 1 0 −1 1 1 1
P = ,D = , and P = .
3 1 0 43 5 −3 2
So A = P DP −1 and
[ ][ ]( )[ ]
k k −1 2 −1 1 0 1 1 1
A = PD P = .
3 1 3 k
0 (4) 5 −3 2

STEP 4. The so-called long-term behaviour of the system:

4
 Since xk = Ak x0 , we have
lim xk = lim Ak x0
k→ ∞ k→ ∞
([ ][ ]( )[ ]) [ ]
2 −1 1 0 1 1 1 C0
= lim
k→ ∞ 3 1 3 k
0 (4) 5 −3 2 S0
([ ][ ]( )[ ]) [ ]
2 −1 1 0 1 1 1 C0
= 3 k
3 1 0 lim ( 4 )
k→ ∞
5 −3 2 S0
([ ][ ]( )[ ]) [ ]
2 −1 1 0 1 1 1 C0
=
3 1 0 0 5 −3 2 S0
( )[ ][ ]
1 2 2 C0
=
5 3 3 S0
[ ]
2(C0 + S0 )/5
= .
3(C0 + S0 )/5

So, in the long run, we have

[ ] ( )[ ][ ] [ ]
Ck 1 2 2 C0 2(C0 + S0 )/5
≈ = .
Sk 5 3 3 S0 3(C0 + S0 )/5

Thus, in the long-run, the ratio of Ck to Sk , denoted by Ck : Sk , is

2(C0 + S0 ) 3(C0 + S0 )
: ,
5 5
or 2 : 3.

In terms of percentages, in the long run the population distribution will be

40% in the city,
[ and] 60% in the suburbs, irrespective of the initial population
C0
distribution . This is an example of a so-called stable limiting population.
S0

Example: We consider a predator-prey population consisting of the foxes

and rabbits living in a certain forest. Initially there are F0 foxes and R0 rabbits,
and after k months there are Fk foxes and Rk rabbits.

The rabbits eat plants in the forest and the foxes eat the rabbits.

5
 Suppose the population of foxes and rabbits from each month to the next is
governed by the equations

Fk+1 = (0.4)Fk + (0.3)Rk (∗)

. (∗∗)
Rk+1 = (−0.4)Fk + (1.2)Rk

The above equations constitute a mathematical model of the fox-rabbit pop-

ulation.
The term (0.4)Fk in (*) indicates that without rabbits to eat, only 40%
will survive each month, the term (0.3)Rk represents the growth in the fox
population due to the available food supply of rabbits.

The term (1.2)Rk in (**) indicates that in the absence of any foxes, the rabbit
population would increase by 20% each month; the term (−0.4)Fk represents
the decline in the rabbit population because of predation by foxes.
[ ]
Fk
In this case, xk = , the transition matrix is
Rk
[ ]
0.4 0.3
A= ,
−0.4 1.2

and xk = Ak x0 .
[ ] [ ] [ ]
1 3 1 0 −1 2 −3
Diagonalizing A, we get P = ,D= , and P = − 4
1
.
2 2 0 0.6 −2 1

6
 In the long run, we have

lim xk = lim Ak x0
k→ ∞ k→ ∞
([ ][ ]( )[ ]) [ ]
1 3 1 0 1 2 −3 F0
= lim −
k→ ∞ 2 2 0 (0.6) k
4 −2 1 R0
([ ] [ ] ( )[ ]) [ ]
1 3 1 0 1 2 −3 F0
= k −
2 2 0 lim (0.6)
k→ ∞
4 −2 1 R0
([ ][ ]( )[ ]) [ ]
1 3 1 0 1 2 −3 F0
= −
2 2 0 0 4 −2 1 R0
( )[ ][ ]
1 −2 3 F0
=
4 −4 6 R0
[ ]
(3R0 − 2F0 )/4
= .
(6R0 − 4F0 )/4

That is, [ ] [ ]
Fk 1
=α
Rk 2
where α = 14 (3R0 − 2F0 ).

So, in the long run, the ratio of Fk to Rk is 1 : 2 and the population of each
will remain roughly a constant, that is, there will be twice as many rabbits as
foxes eventually and the population of each will remain more or less at constant
level. This is another instance of a stable limiting population.

Example: Suppose now the population of foxes and rabbits are governed
by the equations

Fk+1 = (0.4)Fk + (0.3)Rk

.
Rk+1 = (−0.5)Fk + (1.2)Rk

7
 In the long run, we have
lim xk = lim Ak x0
k→ ∞ k→ ∞
([ ][ ]( )[ ]) [ ]
3 1 (0.9)k 0 1 1 −1 F0
= lim −
k→ ∞ 5 1 0 (0.7)k 2 −5 3 R0
([ [
] lim (0.9)k ] ( )[ ]) [ ]
3 1 0 1 1 −1 F
k→ ∞ 0
= −
5 1 0 lim (0.7)k
2 −5 3 R0
k→ ∞
([ ][ ]( )[ ]) [ ]
3 1 0 0 1 1 −1 F0
= −
5 1 0 0 2 −5 3 R0
[ ][ ]
0 0 F0
=
0 0 R0
[ ]
0
= .
0
So, eventually, both foxes and rabbits die out. This is called mutual extinc-
tion.
Example: Suppose now the population of foxes and rabbits are governed
by the equations

Fk+1 = (0.4)Fk + (0.3)Rk

.
Rk+1 = (−0.325)Fk + (1.2)Rk
In the long run, we have
lim xk = lim Ak x0
k→ ∞ k→ ∞
([ ][ ]( )[ ]) [ ]
6 2 (1.05)k 0 1 1 −2 F0
= lim −
k→ ∞ 13 1 0 (0.55)k 20 −13 6 R0
([ [
] lim (1.05)k ] ( )[ ]) [ ]
6 2 0 1 1 −2 F
k→ ∞ 0
= −
13 1 0 lim (0.55)k 20 −13 6 R0
k→ ∞
([ ][ ]( )[ ]) [ ]
k
6 2 lim (1.05) 0 1 1 −2 F 0
= k→ ∞ −
13 1 0 0 20 −13 6 R0
[ ][ ]
1 6 −12 F 0
= (− )( lim (1.05)k )
20 k→ ∞ 13 −26 R0
[ ]
1 6F 0 − 12R 0
= (− )( lim (1.05)k )
20 k→ ∞ 13F0 − 26R0

8
 So we get [ ] [ ]
k 6 ∞
lim xk = ( lim (1.05) )α =
k→ ∞ k→ ∞ 13 ∞
1
where α = ( 20 )(2R0 − F0 ).

So, this shows that the fox and rabbit populations both increases without
bound. This is called Population Explosion.