Difference Equations

National Seminar on
“Real-World Applications
of Mathematics and Statistics”

Dr. Seshadev Padhi

Professor
Department of Mathematics
Birla Institute of Technology, Mesra
Ranchi-835215, India
spadhi@[Link]; ses_2312@[Link]

Amity University Jharkhand, Ranchi

May 1, 2019

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Introduction

Introduction
I In many cases it is of interest to model the evolution of some
system over time.
I There are two distinct cases.
I One can think of time as a continuous variable, or one can think
of time as a discrete variable.
I The first case often leads to differential equations.
I If we consider a time period T and observe (or measure) the
system at times t = kT , k ∈ N0 , the result is a sequence
x0 , x1 , x2 , . . . .
I In some cases these values are obtained from a function f ,
which is defined for all t ≥ 0.
I In this case xk = f (kT ) and this method of obtaining the values
is called periodic sampling.
I One models the system using a difference equation, or what is
sometimes called a recurrence relation.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Introduction

Introduction
I Difference equations arises in many fields of science, for
example:
I In control engineering, the radar tracking devices receive
discrete pulses from the target which is being tracked.
I In electrical networks, the electrical signals are measured
in discrete time pulses
I Difference equations also arises in theory of probability,
statistical problems and many other fields.
I In fact, difference equations are essential for systems with
discrete or digital data.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Beginning Example

Example
Consider a plane that has lying in it k nonparallel lines. Into
how many separate compartments will the plane be divided if
not more than two lines intersect in the same point?

Solution. Let Nk be the number of compartments.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Beginning Example

Example
Consider a plane that has lying in it k nonparallel lines. Into
how many separate compartments will the plane be divided if
not more than two lines intersect in the same point?

Solution. Let Nk be the number of compartments.

When k = 0, there are no lines and hence the plane is

undivided and hence one compartment. and thus making total
two compartments.
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Beginning Example

Example
Consider a plane that has lying in it k nonparallel lines. Into
how many separate compartments will the plane be divided if
not more than two lines intersect in the same point?

Solution. Let Nk be the number of compartments.

When k = 1, there is one line, which divide the previous one

compartment into two compartments and thus making total two
compartments.
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Beginning Example

Example
Consider a plane that has lying in it k nonparallel lines. Into
how many separate compartments will the plane be divided if
not more than two lines intersect in the same point?

Solution. Let Nk be the number of compartments.

When k = 2, the second line, cut the previous one line at one
point and divide the previous two compartments into two
compartments and thus making total four compartments.
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Beginning Example

Example
Consider a plane that has lying in it k nonparallel lines. Into
how many separate compartments will the plane be divided if
not more than two lines intersect in the same point?

Solution. Let Nk be the number of compartments.

When k = 3, the third line, cut the previous two lines at two
points and divide the previous three compartments into two
compartments and thus making total seven compartments.
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Beginning Example

Example
Consider a plane that has lying in it k nonparallel lines. Into
how many separate compartments will the plane be divided if
not more than two lines intersect in the same point?

Solution. Let Nk be the number of compartments.

When k = 4, the fourth line, cut the previous three lines at three
points and divide the previous four compartments into two
compartments and thus making total eleven compartments.
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Beginning Example

Example
Consider a plane that has lying in it k nonparallel lines. Into
how many separate compartments will the plane be divided if
not more than two lines intersect in the same point?

Solution. Let Nk be the number of compartments.

So, generalizing, we have that the (k + 1)th line will be cut by k

previous lines in k points and, consequently, divides each of the
k + 1 prior existing compartments into two. This gives the
difference equation

Nk+1 = Nk + (k + 1).

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Difference Equation

Difference Equation
An ordinary difference equation is a relation, of the form

yk+n = F (k, yk+n−1 , yk+n−2 , . . . , yk ) (1)

between the differences of an unknown function at one or more

general values of the argument.

Order of a difference equation

Order of a difference equation is the difference between the
highest and the lowest indices that appear in the equation.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Difference Equation

Remark
1. The expression given by the equation (1) is an nth -order
difference equation if and only if the term yk appears in the
function F on the right-hand side.
2. Shifts in the labeling of the indices do not changed the
order of a difference equation. For example, for r integer,

yk+n+r = F (k + r , yk+n+r −1 , yk+n+r −2 , . . . , yk+r ) (2)

is the nth -order difference equation, which is equivalent to

equation (1).

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Linear Difference Equations

Linear Difference Equation

A difference equation is linear if it can be put in the form

yk+n +a1 (k )yk+n−1 +a2 (k )yk+n−2 +. . .+an−1 yk+1 +an (k )yk = Rk ,

(3)
where ai (k), i = 1, 2, . . . , n and Rk are given functions of k.

A difference equation is nonlinear if it is not linear.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples - First Order

Example (Drug Delivery)

A drug is administered once every four hours. Let D(k ) be the
amount of the drug in the blood system at the k th interval. The
body eliminates a certain fraction p of the drug during each
time interval. If the amount administered is D0 , find D(k).

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples - First Order

Example (Drug Delivery)

Solution: We first must create an equation to solve.
Since the amount of drug in the patient’s system at time (k + 1) is
equal to the amount at time k minus the fraction p that has been
eliminated from the body, plus the new dosage D0 , we arrive at the
following equation:

D(k + 1) = (1 − p)D(k ) + D0 .

We can solve the above equation, arriving at

 
D0 D0
D(k) = D0 − (1 − p)k + .
p p

Also,
D0
lim D(k) = .
k→∞ p

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples - First Order

Example (Applications to Economics)

I Here we study the pricing of a certain commodity. Let S(k) be the
number of units supplied in period k, D(k ) the number of units
demanded in period k, and p(k ) the price per unit in period k.
I For simplicity, we assume that D(k ) depends only linearly on p(k )
and is denoted by

D(k ) = −md p(k ) + bd , md > 0, bd > 0. (4)

I This equation is referred to as the price-demand curve. The

constant md represents the sensitivity of consumers towards the
price.
I The slop of the demand curve is negative because an increase of
one unit in price produces a decrease of md units in demand.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples - First Order

Example (Applications to Economics)

I We also assume that the price-supply curve relates the
supply in any period to the price one period before, i.e.,

S(k + 1) = ms p(k) + bs , ms > 0, bs > 0. (5)

I The constant ms is the sensitivity of suppliers to price.

I An increase of one unit in price causes an increase of ms
units in supply, thus creating a positive slope for that
price-supply curve.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples - First Order

Example (Applications to Economics)

I A third assumption we make here is that the market price is
the price at which the quantity demanded and the quantity
supplied are equal, that is, at which D(k + 1) = S(k + 1).
Thus
−md p(k + 1) + bd = ms p(k) + bs ,
or
p(k + 1) = Ap(k ) + B = f (p(k )), (6)
where
ms bd − bs
A=− , B= . (7)
md md
This equation is a first-order linear difference equation.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

 Difference Equations

Some Examples - First Order 16 1. Dynamics of First-Order Difference Equations

p(n+1)

Example (Applications to Economics)

An explicit solution of this difference equation with p(0) = p0 is
given by  
B B
p(k) = p0 − Ak +
p0
p(n)

1 − A
FIGURE 1.9. Asymptotically stable equilibrium price.
1 − A 1.3 Equilibrium Points 17

16 1. Dynamics of First-Order Difference Equations

p(n+1)
p(n+1)

p(n+1)

p0 p(n)

p0
p0
p(n) p(n)

FIGURE 1.9. Asymptotically stable equilibrium price. FIGURE 1.10. Stable equilibrium price. FIGURE 1.11. Unstable equilibrium price.

p(n+1)
(i) In case (a), prices alternate above and below
Anbut converge
explicit to the
solution of equi-
(1.3.7) with p(0) = p0 is given by
librium price p*. In economics lingo, the price p∗ is considered “stable”;
 
in mathematics, we refer to it as “asymptotically stable” (Figure
B 1.9). B
p(n) = p0 − An + (Exercises 1.3, Problem 9). (1.3.9)
1−A 1−A
(ii) In case (b), prices oscillate between two values only. If p(0) = p0 , then
Difference Equations p(1) = −p0 +B and p(2) = p0 . Hence the equilibrium point p∗ is stable
Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Some Examples - First Order

Example (Euler Scheme)

I Suppose the following differential equation:

dy
= f (y, t). (8)
dt
where f (y, t) is a given function of y and t, which cannot
be integrated in closed form in terms of the elementary
functions.
I We now proceed to construct the a numerical scheme to
determine the numerical solution.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples - First Order

Example (Euler Scheme)

I First, construct a lattice tk = (∆t)k, where ∆t is a fixed t
interval and k is the set of integers.
I Secondly, replace the derivative by the approximation,

dy (t) y(t + ∆t) − y(t) y − yk

≈ = k+1
dt ∆t ∆t
where yk is the approximation to the exact solution of the
equation at t = tk i.e., yk ≈ y(tk ).
I Also, the right-hand side of equation becomes
f (y, t) ≈ f (yk , (∆t)k).

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples - First Order

Example (Euler Scheme)

I Putting all of this together, we have

yk+1 − yk
= f (yk , (∆t)k).
∆t
I If y0 is specified, then yk for k = 1, 2, 3, . . . , can be
determined.
I This elementary method is called forward-Euler scheme.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples - Higher Order

Example (Power series solutions)

P
∞
I Let us determine a power-series solution y(x) = Ck x k
k=0
to the differential equation

d 2y dy
+ 3x + 3y = 0, (9)
dx 2 dx
where the coefficients, Ck , are to be found.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples - Higher Order

Example (Power series solutions)

I We have
X ∞ X ∞
dy
x =x kCk x k−1 = (k − 2)Ck−2 x k−2 (10)
dx
k=0 k=2

X ∞ X ∞
d 2y k−2
= k(k − 1)Ck x = k(k − 1)Ck x k−2 (11)
dx 2
k=0 k=2

and
∞
X ∞
X
y= Ck x k = Ck−2 x k−2 (12)
k=0 k=2

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples - Higher Order

Example (Power series solutions)

I Substituting the above equations in the given differential
equation, we obtain
∞
X
[k (k − 1)Ck + 3(k − 1)Ck−2 ] x k−2 = 0. (13)
k=2

I Equation each coefficient to zero gives the following

recursion relation, which the Ck must satisfy:

kCk + 3Ck−2 = 0. (14)

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

 pair of mature rabbits, if each pair of rabbits gives birth to a new pair
Difference Equations
each month starting when it reaches its maturity age of two months? (See
Figure 2.1.)
Fibonacci Sequence - The Famous Example
Table 2.2 shows the number of pairs of rabbits at the end of each month.
The first pair has offspring at the end of the first month, and thus we have
Example (Fibonacci Sequence - Rabbit Problem)
two pairs. At the end of the second month only the first pair has offspring,
and
This problem thus wefirst
have appeared
three pairs. Atinthe end of in
1202, theLiber
third month,
abaci,theafirst and
book
second pairs will have offspring, and hence we have five pairs. Continuing
aboutthis the abacus,
procedure, we written
arrive at by
Tablethe
[Link]
If F (n) isItalian mathematician
the number of pairs of
Leonardorabbits di
at the end of
Pisa, n months,
better then the
known asrecurrence
Fibonacci. relation
Thethatproblem
represents may
this model is given by the second-order linear difference equation
be stated as follows:
F (n + 2) = F (n + 1) + F (n), F (0) = 1, F (1) = 2, 0 ≤ n ≤ 10.

How This
manyexample
pairsis of
a special
rabbitscase will
of thethere
Fibonacci
be sequence,
after one given
yearby if starting

with one pair of mature rabbits, if each pair of rabbitsngives

F (n + 2) = F (n + 1) + F (n), F (0) = 0, F (1) = 1, ≥ 0. birth
(2.3.13)
to a new pair14each
The first month
terms are given starting
by 1, 2, 3, when
5, 8, 13,it21,
reaches
34, 55, 89,its maturity
144, 233,
andtwo
age of 377,months?
as already noted
(See in Figure
the rabbitbelow)
problem.

Month 0 Month 1 Month2

FIGURE 2.1.
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
 rabbits at the end of n months, then the recurrence relation that represents
Difference Equations
this model is given by the second-order linear difference equation
Fibonacci Sequence - The
F (n + 2) Famous
= F (n Example
+ 1) + F (n), F (0) = 1, F (1) = 2, 0 ≤ n ≤ 10.
This example is a special case of the Fibonacci sequence, given by
F (n + 2) = F (n + 1) + F (n), F (0) = 0, F (1) = 1, n ≥ 0.
Example (Fibonacci Sequence - Rabbit Problem)(2.3.13)
TableThe first 14 terms are given by 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233,
below shows the number of pairs of rabbits at the end of
and 377, as already noted in the rabbit problem.
each month.
The first pair has offspring at the end of the first month, and
thus we have two pairs.
At the end of the second month only the first pair has offspring,
and thus we have three pairs.
At the end ofMonth
the0 third month, the Month 1
first and second Month2
pairs will
have offspring, and hence we FIGURE
have [Link] pairs. Continuing this

procedure, we arrive at Table.

TABLE 2.2. Rabbits’ population size.
Month 0 1 2 3 4 5 6 7 8 9 10 11 12
Pairs 1 2 3 5 8 13 21 34 55 89 144 233 377

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Fibonacci Sequence - The Famous Example

Example (Fibonacci Sequence - Rabbit Problem)

If F (k) is the number of pairs of rabbits at the end of k months,
then the recurrence relation that represents this model is given
by the second-order linear difference equation

F (k +2) = F (k +1)+F (k ), F (0) = 1, F (1) = 2, 0 ≤ 10.

This example is a spacial case of the Fibonacci sequence,

given by

F (k+2) = F (k+1)+F (k ), F (0) = 0, F (1) = 1, n ≥ 0. (1)

The first 14 terms are given by

1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233, and 377, as already
noted in the rabbit problem.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Solution of a Difference Equation

Solution of a difference equation

I A solution of a difference equation is a function φ(k) that
reduces the equation to an identity.
I The general solution of a difference equation is the
solution in which the number of arbitrary constants is equal
to the order of the difference equation.
I A particular solution of a difference equation is that
solution which is obtained from the general solution by
giving particular values to the constants.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example

yk+1 − 3yk + yk−1 = e−k (second order, linear)

yk+1 = yk2 (first order, nonlinear)
k
yk +4 − yk = k2 (fourth order, linear)
yk +1 = yk − (1/100)yk2 (first order, nonlinear)
yk+3 = cos yk (third order, nonlinear)
k
yk+2 + (3k − 1)yk+1 − yk = 0 (second order, linear)
k +1

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

The first-order nonlinear equation

2
yk+1 − yk2 = 1
√
has the solution φ(k ) = k + c where c is a constant.
This statement can be checked by substituting φ(k ) into the
left-hand side of the difference equation to obtain
√ 2 √ 2
( k + 1 + c − k + c = (k + 1 − c) − (k + c) = 1,

which is equal to the expression on the right-hand side.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

The second-order linear difference equation

yk+1 − yk−1 = 0

has two solutions, φ1 (k ) = (−1)k and φ2 (k ) = 1.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Existence and Uniqueness Theorem

Thoerem (Existence and Uniqueness Theorem)

Let

yk+n = f (k , yk , yk+1 , . . . , yk+n−1 ), k = 0, 1, 2, 3, . . .

be an nth -order difference equation where f is defined for each

of its arguments. This equation has one and only one solution
corresponding to each arbitrary selection of the n initial values
y0 , y1 , . . . , yn−1 .

Proof.
If the values, y0 , y1 , . . . , yn−1 are given, then the difference
equation with k = 0 uniquely specifies, yn . Once yn is known,
the difference equation with k = 1 gives yn+1 . Proceeding in
this way, all yk , for k ≥ n, can be determined.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Operators ∆ and E

Operators ∆ and E
I In the theory of difference equations, more frequently, we
use the operators ∆ and E to denote the differences:
I The operator ∆ (called as (first) difference operator) is
defined as follows:
∆yk = yk+1 − yk .
I The second difference operator is defined as ∆2 = ∆ · ∆,
∆2 yk = ∆(∆(yk )) = ∆(yk +1 − yk ) = yk+2 − 2yk+1 + yk .
I In general,
n(n − 1)
∆n yk =yk+n − nyk+n−1 + yk+n−2 + . . .
2!
n(n − 1) . . . (n − i + 1)
+ (−1)i yk+n−i + . . . + (−1)n yk
i!
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Operators ∆ and E

Operators ∆ and E

I The operator, E, (called shift operator ) is defined as

E p = yk+p .
I From the definition of ∆ and E, we have

∆yk = (E − 1)yk

and that
∆ ≡ E − 1 or E ≡ ∆ + 1.
I Hence, we have
n  
X
n n n
yk +n = E yk = (1 + ∆) yk = ∆i yk .
i
i=0

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

General Theory of Linear Difference Equations

Definition
Let the functions a0 (k), a1 (k ), . . . , an (k ), and Rk be defined
over a set of integers, k1 ≤ k ≤ k2 , where k1 and k2 can be
either finite or unbounded in magnitude. An equation of the
form

a0 (k )yk+n + a1 (k)yk+n−1 + . . . + an (k)yk = Rk

is said to be linear. This equation is of order n if and only if

a0 (k ) 6= 0, for any k. In this case dividing by a0 (k) and
relabeling the ratio of coefficient functions, we can write the
general nth order linear difference equation as follows:

yk+n + a1 (k)yk+n−1 + . . . + an (k)yk = Rk . (15)

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

General Theory of Linear Difference Equations

Definition
The (15) is called homogeneous if Rk is identically zero for all k
i.e.
yk+n + a1 (k)yk+n−1 + . . . + an (k)yk = 0; (16)
otherwise, it is called an inhomogeneous equation.

Definition
If the functions a0 (k), a1 (k ), . . . an (k) are constant then the (15)
is said to Linear difference equations with constant coefficients.

Definition
A set of k linearly independent solutions of (16) is called a
fundamental set of solutions.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Casoratian - Linear Independent and Dependent Solutions

Definition
The Casoratian C(k ) of the solutions f1 (k ), f2 (k), ..., fn (k) is dined
as
 
 f (k ) f (k) . . . f (k ) 
 1 2 n 
 f1 (k + 1) f2 (k + 1) ... fn (k + 1) 

C(k) =  .. .. 
 . . 
 
f1 (k + n − 1) f2 (k + n − 1) . . . fn (k + n − 1)

Casoratian plays an important role in determining whether

particular set of functions are linearly independent or dependent.
Thoerem
The functions f1 (k), f2 (k), . . . , fn (k) are n linearly dependent
functions if and only if their Casoratian equals to for all k .

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Fundamental Theorems for Homogeneous Equations

Fundamental Theorems for Homogeneous Equations

Thoerem
Let the functions, a1 (k), a2 (k ), . . . , an (k ) be defined for all k; let an (k) be
nonzero for all k; then there exist n linearly independent solutions
y1 (k), y2 (k ), . . . , yn (k ) of (16).

Thoerem
An nth -order linear difference equation has n and only n linearly
independent solutions.

Thoerem
The general solution of equation (16) is given by

yk = c1 y1 (k ) + c2 y2 (k ) + . . . + ck yn (k ),

where ci , 1 ≤ i ≤ n, are n arbitrary constants and the yi (k), 1 ≤ i ≤ n ar a

fundamental set of solutions.
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Linear Difference Equations with Constant Coefficients

We consider the nth -order linear difference equations with

constant coefficients,

yk+n + a1 yk+n−1 + . . . + an yk = Rk , (17)

where ai are given set of n constants, with an 6= 0, and Rk is a

given function of k. If Rk = 0, then (17) is homogeneous:

yk+n + a1 yk+n−1 + . . . + an yk = 0; (18)

for Rk 6= 0, equation (17) is inhomogeneous.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Linear Difference Equations with Constant Coefficients

From, the previous slides, we have the homogeneous equation

(18) has a fundamental set of solutions that consists of n
(1) (2) (n)
linearly independent functions yk , yk , . . . , yk and that the
general solution is the linear combination
(H) (1) (2) (n)
yk = c1 yk + c2 yk + . . . + cn yk ,

where ci are n arbitrary constants.

The inhomogeneous equation (17) consists of a sum of

(H) (P)
homogeneous solution yk and a particular solution yk to
equation (17),
(H) (P)
yk = yk + yk .

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Linear Difference Equations with Constant Coefficients

Using the shift operator E, we can write the equation (17) as

f (E)yk = 0, (19)

where
f (E) = E n + a1 E n−1 + . . . + an−1 E + an . (20)

Definition
The characteristic equation associated with equation (17) or (19) is

f (r ) = r n + a1 r n−1 + . . . + an−1 r + an = 0. (21)

Thoerem
Let ri be any solution to the characteristic equation of (21); then

yk = rik (22)

is a solution to the homogeneous equation (18).

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Linear Difference Equations with Constant Coefficients

Thoerem
Assume the n roots of the characteristic equations are distinct;
(i)
then a fundamental set of solution is yk = rik , i = 1, 2, . . . , n
and that the general solution to the homogeneous equation
(18) is
(1) (2) (n)
yk = c1 yk + c2 yk + . . . + cn yk , (23)
where the n constants ci are arbitrary.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Linear Difference Equations with Constant Coefficients

Thoerem
Let the roots of the characteristic equation (21) of the
homogeneous difference equation (18) be ri with multiplicity mi ,
i = 1, 2, . . . , l, where m1 + m2 + . . . + ml = n. Then, the general
solution of (21) is:
(1) (1) (1)
yk =r1k (A1 + A2 k + . . . + Am1 k m1 −l )
(2) (2) (2)
+ r2k (A1 + A2 k + . . . + Am2 k m2 −l )
+ ...
(l) (l) (l)
+ rmk l (A1 + A2 k + . . . + Aml k ml −l ). (24)

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Linear Difference Equations with Constant Coefficients

Thoerem
If the linear difference equation (18) has real coefficients, then
any complex roots of the characteristic equation (21) must
occur in complex conjugates pairs. Moreover, the
corresponding fundamental solutions can be written in either of
the two equivalent forms:
(1) (2)∗
yk = yk = r1k

or
(1) (2)
ȳk = R k cos(kθ), ȳk = R k sin(kθ),
where the complex conjugate pair of roots are
p
r1 = r2∗ = a + ib = Reiθ ; R = a2 + b2 , tan θ = b/a.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

In particular, for a homogeneous linear difference equation of

order 2, i.e., yk+2 + a1 yk+1 + a2 yk = 0(a2 6= 0), we have the
following three situations.
Three Cases
Case 1: The characteristic roots r1 , r2 are real and distinct.
Then the general solution is

yk = c1 r1k + c2 r2k .

Case 2: The characteristic roots r1 , r2 are real and equal

(say r1 = r2 = r ). Then the general solution is

yk = (c1 + c2 k )r k .

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Three Cases
Case 3: The characteristic roots are complex conjugates,
√
say r1,2 = a ± ib = Re ±iθ , where R = a2 + b 2
 
b
and θ = tan−1 a . Then the general solution is

 k  k
yk = c1 Reiθ + c2 Re−iθ
= R k [c1 (cos k θ + i sin kθ) + c2 (cos kθ − i sin kθ)]
= R k [(c1 + c2 ) cos k θ + (ic1 − ic2 ) sin k θ]
= R k [A1 cos kθ + A2 sin kθ]

where A1 , A2 are arbitrary constants.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Example
Find the general solution of

yk+2 + 5yk+1 + 6yk = 0.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Solution
The characteristic equation for the given problem is:

r 2 + 5r + 6 = 0,

which has the roots r1 = −3 and r2 = −2. Therefore, the

general solution is

yk = c1 (−3)k + c2 (−2)k .

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Example
Find the general solution of

yk+2 − 2yk+1 + yk = 0.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Solution
The characteristic equation for the given problem is:

r 2 − 2r + 1 = 0 or (r − 1)2 = 0,

which has the roots r1 = r2 = 1. Therefore, the general solution

is
yk = (c1 + c2 k)(1)k = c1 + c2 k.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Example
Find the general solution of

yk+2 − 2yk+1 − 2yk = 0.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Solution
The characteristic equation for the given problem is:

r 2 − 2r + 2 = 0,

which√has the complex conjugate roots r1,2 = 1 ± i. Thus

R = 2 and θ = tan−1 1 = π4 .
Therefore, the general solution is
√ h  π  π i
yk = ( 2)k A1 cos k + A2 sin k .
4 4

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Example (Fibonacci Sequence - Revisited)

F (k+2) = F (k+1)+F (k ), F (0) = 0, F (1) = 1, k ≥ 0. (1)

The characteristic equation of (1) is

r 2 − r − 1 = 0.
√
1+ 5
Hence the characteristic roots are r1 = α = 2 and
√
1− 5
r2 = β = 2 .
The general solution of (1) is
√ !n √ !n
1+ 5 1− 5
F (k ) = a1 + a2 , k ≥ 1. (2)
2 2

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Example (Fibonacci Sequence - Revisited)

Using the initial values F (1) = 1 and F (2) = 1, one obtains

1 1
a1 = √ , a2 = − √ .
5 5
Consequently,
 √ !k √ !k 
1  1+ 5 1− 5  1
F (k) = √ − = √ (αk −β k ). (3)
5 2 2 5

F (k+1)
It is interesting to note that lim F (k) = α ≈ 1.618. This
k→∞
number is called the golden mean/ratio, which supposedly
represents the ratio of the sides of a rectangle that is most
pleasing to the eye.
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Example (The Transmission of Information)

I Suppose that a signaling system has two signals s1 and s2
such as dots and dashes in telegraphy.
I Messages are transmitted by first encoding them into a
string, or sequence, of these two signals.
I Let us suppose that s1 requires exactly n1 units of time,
and s2 requires exactly n2 units of time, to be transmitted.
I Let M(n) be the number of possible message sequence of
duration n and a signal of duration time n either ends with
an s1 signal or with an s2 signal.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Example (The Transmission of Information)

I Now, if the message ends with s1 , the last signal must start
at n − n1 .
I Hence there are M(n − n1 ) possible messages to which
the message s1 can be appended at the end.
I By a similar argument, there are M(n − n2 ) possible
messages to which the message s2 can be appended at
the end.
I Consequently, the total number of messages x(n) of
duration n may be given by

M(n) = M(n − n1 ) + M(n − n2 ).

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Example (The Transmission of Information)

I If n1 ≥ n2 then we obtain a difference equation of
n1 th-order

M(n + n1 ) − M(n + n1 − n2 ) − M(n) = 0.

I If n2 ≥ n1 then we obtain a difference equation of

n2 th-order

M(n + n2 ) − M(n + n2 − n1 ) − M(n) = 0.

I An interesting special case is that in which n1 = 1 and

n2 = 2. In this case we have

M(n + 2) − M(n + 1) − M(n) = 0

which is nothing but our Fibonacci Sequence.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Example (The Transmission of Information)

I The general solution is
√ !n √ !n
1+ 5 1− 5
M(n) = a1 + a2 , n = 0, 1, 2, . . .
2 2

I To find a1 , a2 let us take M(0) = 0, M(1) = 1, this yields

√ √
a1 = 1/ 5, a2 = −1/ 5.
I So, we have
√ !n √ !n
1 1+ 5 1 1− 5
M(n) = √ −√ , n = 0, 1, 2, . . .
5 2 5 2

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Theory of Second Order Linear Difference Equations with Constant Coefficients

Example (The Transmission of Information)

I In information theory, the capacity of C of the channel is
defined as
log2 M(n)
C = lim
n→∞ n
I So,
" √ !n √ !n #
log2 √15 1 1+ 5 1− 5
C = lim + lim log2 −
n→∞ n n→∞ n 2 2
√ !
1+ 5
= log2
2
≈ 0.7

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Inhomgeneous Equations: Method of Undetermined Coefficients

We now turn to a technique for obtaining solutions to the

nth -order linear inhomogeneous difference equations with
constant coefficients,

yk +n + a1 yk+n−1 + . . . + an yk = Rk , an 6= 0, (25)

where Rk is a linear combination of terms each having one of

the forms
ak , ebk , sin(ck ) cos(ck ), k l ,
where a, b, and c are constants and l is a non-negative integer.
We also include products of these forms; for example,

ak cos(ck ), k l ebk , ak k l cos(ck ), etc.

To proceed, we first need some definitions.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Inhomgeneous Equations: Method of Undetermined Coefficients

Definition
A family of a term Rk is the set of all functions of which Rk and
E m Rk , for m = 1, 2, 3, . . . , are linear combinations.

Definition
A finite family is a family that contains only a finite number of
functions.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Inhomgeneous Equations: Method of Undetermined Coefficients

For example, if Rk = ak , then E m ak = am ak , m = 1, 2, 3, . . . , and

the family ak contains only one member, namely, ak . We denote
this family by {ak }.

If Rk = k l , then E m k l = (k + m)l , which can be expressed as a

linear combination of 1, k, k 2 , . . . , k l ; thus, the family of E m k l is the
set {1, k, k 2 , . . . , k l }.

If Rk = cos(ck ) or sin(ck ), then the families are {cos(ck ), sin(ck )}.

Finally, note that for the case Rk is a product, the family consists of
all possible products of distinct members of the individual term
families. For example, the term Rk = k l ak has the finite family
{ak , kak , k 2 ak , . . . , k l ak }.
Likewise, the term Rk = k l cos(ck ) has the finite family
{cos(ck ), k cos(ck ), . . . , k l cos(ck ), sin(ck), k sin(ck), . . . , k l sin(ck )}.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Inhomgeneous Equations: Method of Undetermined Coefficients

Procedure for obtaining particular solutions to the

inhomogeneous equation (25)
(i) Construct the family of Rk .
(ii) If the family contains no terms of the homogeneous solution,
(P)
then write the particular solution yk as a linear combination of
the members of that family. Determine the constants of
combinations such that the inhomogeneous difference equation
is identically satisfied.
(iii) If the family contains terms of the homogeneous solution, then
multiply each member of the family by the smallest integral
power of k for which all such terms are removed. The particular
(P)
solution yk can then be written as a linear combination of the
members of this modified family. Again, determine the constants
of combination such that that the inhomogeneous difference
equation is identically satisfied.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (A)
The second-order difference equation

yk+2 − 5yk+1 + 6yk = 2 + 4k

has the characteristic equation r 2 − 5r + 6 = (r − 3)(r − 2) = 0, with roots

r1 = 3 and r2 = 2. Therefore, the homogeneous solution is
(H)
yk = c1 3k + c2 2k ,

where c1 and c2 are arbitrary constants. The right-hand side of the difference
equation is 2 + 4k . Note that, the 2 has the family that consists of only one
member {1}, while 4k has the family {1, k}. Therefore, the combined family is
{1, k}. Since, neither member of the combined family occurs in the
homogeneous solution, we write the particular solution as the following linear
combination:
(P)
yk = A + Bk ,
where constants A and B are to be determined.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (A)
Substituting the above into the given difference, we obtain

A + B(k + 2) − 5A − 5B(k + 1) + 6A + 6Bk = 2 + 4k .

Upon setting coefficients of the k 0 and k 1 terms equal to zero, we obtain

2A − 3B = 2, 2B = 4.

Therefore,
A = 4, B − 2,
and the particular solution is
(P)
yk = 4 + 2k.

The general solution to the given difference equation is:

yk = c1 3k + c2 2k + 4 + 2k.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (B)
Consider the difference equation

yk +2 − 6yk+1 + 8yk = 2 + 3k 2 − 5 · 3k .

The characteristic equation is: r 2 − 6r + 8 = (r − 2)(r − 4) = 0,

which leads to the following solution of the homogeneous
equation:
(H)
yk = c1 2k + c2 4k ,
where c1 and c2 are arbitrary constants. The families of the
terms in Rk are

2 → {1}; k 2 → {1, k, k 2 }; 3k → {3k };

So the combined family is {1, k, k 2 , 3k }.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (B)
No member of this family occur in the homogeneous solution.
Therefore, the particular solution takes the form
(P)
yk = A + Bk + Ck 2 + D3k ,

where A, B, C, and D are constants to be determined.

Substituting this in the given difference equation and simplifying
the resulting expression gives

(3A − 4B − 2C) + (3B − 8C)k + 3ck 2 − D3k = 2 + 3k 2 − 5 · 3k .

Equating the coefficients of the linearly independent terms on

both sides to zero gives

3A − 4B − 2C = 2; 3B − 8C = 0; 3C = 3; D = 5.
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Some Examples

Example (B)
Solving the above equations, we obtain

A = 44/9; B = 8/3; C = 1; D = 5,

and that the particular solution is

(P) 44 8
yk = + k + k 2 + 5 · 3k ,
9 3
and the general solution to the given difference is

44 8
yk = c1 2k + c2 4k + + k + k 2 + 5 · 3k .
9 3

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (C)
The equation
yk+2 − 4yk+1 + 3yk = k4k
has the homogeneous solution

ykH = c1 + c2 3k ,

where c1 and c2 are arbitrary constants. The family of Rk = k 4k

is {4k , k4k } and does not contain a term that appears in the
homogeneous solution. Therefore, the particular solution is of
the form
(P)
yk = (A + Bk )4k ,
where A and B can be determined by substituting this equation
in the given difference equation;

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (C)
doing this gives

(3A + 16B)4k + (3B)k4k = k 4k

and

3A + 16B = 0, 3B = 1; which implies A = −16/9, B = 1/3.

So, the particular solution is

(P) 16 k 1 k
yk =− 4 + k4
9 3
16 k
and the general solution is yl = c1 + c2 3k − 9 4 + 31 k4k .

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (D)
Consider the third-order difference equation

yk+3 − 7yk+2 + 16yk+1 − 12yk = k2k .

Its characteristic equation is:

r 3 − 7r 2 + 16r − 12 = (r − 2)2 (r − 3) = 0,

and the corresponding homogeneous solution is:

(H)
yk = (c1 + c2 k )2k + c3 3k ,

where c1 , c2 and c3 are arbitrary constants.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (D)
The family of Rk = k 2k is {2k , k2k } and both members of this
family occur in the homogeneous solution; therefore, we must
multiply the family by k 2 to obtain a new family that does not
contain any function that appear in the homogeneous solution.
The new family is {k 2 2k , k 3 2k }. Thus, the particular solution is
(P)
yk = (Ak 2 + Bk 3 )2k ,

where A and B are to be determined. The substitution of this

equation into the difference equation, gives

2k (−8A + 24B) + k 2k (−24B) = k2k ,

from which we obtain A = −1/8; B = −1/24.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (D)
Therefore, the particular solution is

(P) 1
yk =− (3 + k )k 2 2k ,
24
and the general solution is

1
yk = (c1 + c2 k)2k + c3 3k − (3 + k )k 2 2k .
24

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (E)
Consider the second order difference equation
 
k kπ
y(k + 2) + 4y(k) = 8(2 ) cos .
2

The characteristic equation of the homogeneous equation is

r 2 + 4 = 0.

The characteristic roots are r1 = 2i, r2 = −2i. Thus

R = 2, θ = π/2, and
    
(H) k kπ kπ
yk = 2 c1 cos + c2 sin .
2 2

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (E)  
Notice that the family of Rk = 8(2k ) cos cos k2π is
   
{(2k ) cos kπ
2 , (2k ) sin kπ }, both of the members belongs to
2
the homogeneous solution. So, we assume
   
(P) kπ
yk = 2k ak cos + bk sin (kπ2) .
2
(P)
Substituting yk into given difference equation gives
    
k +2 kπ kπ
2 a(k + 2) cos + π + b(k + 2) sin +π
2 2
      
kπ kπ kπ
+ (4)2k ak cos + bk sin = 8(2k ) cos .
2 2 2

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some Examples

Example (E)
Replacing cos ((k π)/2 + π) by − cos ((k π)/2), and
sin ((kπ)/2 + π) by − sin ((k π)/2) and then comparing the
coefficients of the cosine terms leads us to a = −1. Then by
comparing the coefficients of the sine terms, we realize that
b = 0.
(P)
By substituting these values back into yk , we have that
 
(P) k kπ
yk = −2 k cos ,
2

and the general solution is

      
k kπ kπ kπ
yk = 2 c1 cos + c2 sin − k cos .
2 2 2

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Nonlinear Difference Equations - Logistic Model

Example (The Logistic Equation)

I Let y(k ) be the size of a population at time k.
I If µ is the rate of growth of the population from one
generation to another, then we may consider a
mathematical model in the form

y (k + 1) = µy (k ), µ > 0. (26)

I If the initial population is given by y (0) = y0 , then by simple

iteration we find that

y (k) = µk y0

is the solution of (1) .

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Nonlinear Difference Equations - Logistic Model

Example (The Logistic Equation)

I If µ > 1, then y(k) increases indefinitely, and
lim y(k ) = ∞.
k→∞
I If µ = 1, then y(k) = y0 for all k > 0, which means that the
size of the population is constant for the indefinite future.
I However, for µ < 1, we have lim y (k ) = 0, and the
k→∞
population eventually becomes extinct.
I For most biological species, however, none of the above
cases is valid as the population increases until it reaches a
certain upper limit.
I Then, due to the limitations of available resources, the
creature will become testy and engage in competition for
those limited resources.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Nonlinear Difference Equations - Logistic Model

Example (The Logistic Equation)

I This competition is proportional to the number of
squabbles among them, given by y 2 (k ).
I A more reasonable model would allow b, the
proportionality constant, to be greater than 0,

y(k + 1) = µy(k) − by 2 (k). (27)

If in (27), we let x(k) = µb y (k ), we obtain

x(k + 1) = µx(k )(1 − x(k )) ≡ f (x(k)). (28)

I This equation is the simplest nonlinear first-order

difference equation, commonly referred to as the (discrete)
logistic equation.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Nonlinear Difference Equations - Logistic Model

Example (The Logistic Equation)

I However, a closed-form solution of (28) is not available
(except for certain values of µ).
I In spite of its simplicity, this equation exhibits rather rich
and complicated dynamics.
I To find the equilibrium points of (28) we let
f (x ∗ ) = µx ∗ (1 − x ∗ ) = x ∗ . Thus, we pinpoint two
equilibrium points:x ∗ = 0 and x ∗ = (µ − 1)/µ.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations
1.3 Equilibrium Points 15
Nonlinear Difference Equations - Logistic Model

x(k+1)

x(k)
x0 x*

FIGURE 1.8. Stair step diagram for µ = 2.5.

Stair step diagram of (x(k ), x(k + 1)) when µ = 2.5 and
x(0)in=price
[Link]
In this case, ofwe
an increase ms have
units intwo equilibrium
supply, points,
creating a positive one
slope
∗ for that curve. ∗
x = 0, which is unstable, and the other x = 0.6, which is
A third assumption we make here is that the market price is the price
asymptotically [Link] and the quantity supplied are equal, that
at which the quantity
is, at which D(n + 1) = S(n + 1).
Thus
−md p(n + 1) + bd = ms p(n) + bs ,
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
or
Difference Equations

Nonlinear Difference Equations - Logistic Model

1.7 The Logistic Equation and Bifurcation 45

x(k+1)

x(k)
x0 x*

FIGURE 1.33. µ > 3: x∗ is an unstable fixed point.

Stair step diagram of (x(k ), x(k + 1)) in which one interior
equilibrium x ∗ = 0 is unstable and the other interior equilibrium
(i) x∗ is an asymptotically stable fixed point for 1 < µ ≤ 3 (Figure 1.32).
is unstable.
(ii) x∗ is an unstable fixed point for µ > 3 (Figure 1.33).

1.7.2 2-Cycles
Difference Equations Dr. Seshadev Padhi, BIT, Mesra
Difference Equations

Nonlinear Difference Equations - Logistic Model

1.7 The Logistic Equation and Bifurcation 47

FIGURE 1.35. The bifurcation diagram of Fµ .

Bifurcation diagram of Logistic Equation w.r.t. parameter µ.
TABLE 1.4. Feigenbaum table.

Difference Equations µn − µn−1 Dr. Seshadev Padhi, BIT, Mesra

n µn µn − µn−1
Difference Equations

Relationship between Linear Differential and Difference Equations

Consider an nth-order linear homogeneous differential equation

having constant coefficients,

D n y (x) + a1 D n−1 y(x) + . . . + an y (x) = 0, (29)

where D ≡ d/dx is the differentiation operator, the ai ,

i = 1, 2, . . . , n, are given constants, and an 6= 0. Associated with
this differential equation is the following difference equation:

yk+1 + a1 yk+n−1 + . . . + an yk = 0. (30)

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Relationship between Linear Differential and Difference Equations

Thoerem
Let  
l
X i −1
nX n
X
y(x) =  ci,j+1 x j  eri x + cj erj x (31)
i=1 j=0 j=(n1 +...+nl )+1

be the general solution of equation(1), where ci,j+1 and cj are arbitrary

constants; ni ≥ 1,i = 1, 2, . . . , l, with n1 + n2 + . . . + nl ≤ n; and where the
characteristic equation

r n + a1 r n−1 + . . . + an = 0 (32)

has roots ri with multiplicity ni , i = 1, 2, . . . , l, and the simple roots rj .

Let yk be the general solution of equation (2). Then


yk = D k y (x) , (33)
x=0

and  
i
X i −1
nX n
X
ci1 + km  rik +
yk = γi , m cj rik , (34)
i=1 m=1 j=(n1 +...+nl )+1

where the γi,m are arbitrary constants.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Relationship between Linear Differential and Difference Equations

Example. The second-order differential equation

d 2y dy
−3 + 2y = 0 (35)
dx 2 dx
has the general solution
y (x) = c1 ex + c2 e2x , (36)
where c1 and c2 are arbitrary constants. The difference equation associated with this differential
equation is
yk+2 − 3yk+1 + 2yk = 0. (37)
Its general solution is
yk = A + B2k , (38)
r2
since the characteristic equation − 3r + 2 = 0 has roots r1 = 1 and r2 = 2; A and B are arbitrary
constants. We now show how the result given by equation (38) can be obtained from equation (36).
Let us calculate D k y(x);it is

dk
D k y(x) = (c1 ex + c2 e2x ) = c1 ex + c2 2k e2x . (39)
dx k
Therefore, 

yk = D k y (x) = c1 + c2 2k , (40)
x=0
which is the same as equation (38) except for the labeling of the arbitrary constants.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Relationship between Linear Differential and Difference Equations

Example. The differential equation

d 2y dy
−2 +y =0 (41)
dx 2 dx
has the general solution

y(x) = (c1 + c2 x)ex = c1 ex + c2 xex . (42)

The associated difference equation is

yk+2 − 2yk+1 + yk = 0. (43)

From equation (42) we obtain

dk
Dk y(x) = (c1 ex + c2 xex )
dx k
= c1 ex + c2 (xex + kex ), (44)

where the expression in parentheses on the right-hand side of equation (44) was obtained
by using the Leibnitz rule for the kth derivative of a product.
Therefore, 

yk ≡ D k y(x) = c1 + c2 k , (45)
x=0
which is easily shown to be the general solution of equation (43).

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some challenging questions to test the understanding of the students

Problem
Consider a set of k spheres so placed that each sphere intersects all the other
spheres. Let ck be the number of compartments into which space is divided. Show
that
ck+1 = ck + k 2 − k + 2.

Problem
Consider a collection of k boxes and k labels, with one label marked for each box.
Show that the number of ways, Nk , they can be mixed such that no box has its own
label is
Nk = (k − 1)Nk−1 + (k − 1)Nk−2 .

Problem
Let the single, self-interacting population model (logistic equation) be harvested;
i.e., a certain constant number of the population is removed at the end of the
interval tk = (∆t)k . What is the new population equation?

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some challenging questions to test the understanding of the students

Problem
A vacuum pump removes one third of the remaining air in a cylinder
with each stroke. Form an equation to represent this situation. After
how many strokes is just 1/1000000 of the initial air remaining?

Problem
A population is increasing at a rate of 25 per thousand per year. Define
a difference equation which describes this situation. Solve it and find
the population in 20 years’ time, assuming the population is now 500
million. How long will it take the population to reach 750 million?

Problem
Form and solve the difference equation defined by the sequence in
which the nth term is formed by adding the previous two terms and
then doubling the result, and in which the first two terms are both one.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra

Difference Equations

Some challenging questions to test the understanding of the students

Problem
In a new colony of geese there are 10 pairs of birds, none of which produce
eggs in their first year. In each subsequent year, pairs of birds which are in their
second or later year have, on average, 4 eggs (2 male and 2 female). Assuming
no deaths, show that the recurrence relation which describes the geese
population is
un+1 = un + 2un−1 , u1 = 10 and u2 = 10,
where un represents the geese population (in pairs) at the beginning of the nth
year.

Problem
The growth in number of neutrons in a nuclear reaction is modelled by the
recurrence relation
un+1 = 6un − 8un−1 ,
with initial values u1 = 2, u2 = 5, where un is the number at the beginning of the
time interval n (n = 1, 2, . . . ). Find the solution for un and hence, or otherwise,
determine the value of n for which the number of neutrons reaches 10000.

Difference Equations Dr. Seshadev Padhi, BIT, Mesra