You are on page 1of 21

BITS Pilani

Pilani Campus

MATH F213
Linear Recurrence Relations
Recall :

For non-negative integer k, letting 𝑐0 𝑛 , 𝑐1 𝑛 ,…, 𝑐𝑘 𝑛 , 𝑓(𝑛) be


functions of n for 𝑛 ≥ 𝑘. Then a recurrence relation of the form

𝑐0 𝑛 𝑎𝑛 + 𝑐1 𝑛 𝑎𝑛−1 + … + 𝑐𝑘 𝑛 , 𝑎𝑛−𝑘 = 𝑓 𝑛 , 𝑛 ≥ 𝑘

is called a linear recurrence relation of degree 𝑘 if 𝑐0 𝑛 , 𝑐𝑘 𝑛


are not identically 0. If 𝑐𝑖 𝑛 = 𝑐𝑖 is constant for all 0 ≤ 𝑖 ≤ 𝑛,
then we get the linear recurrence relation with constant
coefficients.
Further it is homogeneous if 𝑓 𝑛 = 0 for all n≥ 𝑘.
Generating function method can be used for solving such
recurrence relations.
BITS Pilani, Pilani Campus
BITS Pilani, Pilani Campus
BITS Pilani, Pilani Campus
BITS Pilani, Pilani Campus
Consequence of generating
functions

Homogeneous linear recurrence relation with constant


coefficients can be written as

𝑎𝑛 + 𝑐1 𝑎𝑛−1 + … + 𝑐𝑘 𝑎𝑛−𝑘 = 0 𝑛 ≥ 𝑘, 𝑐𝑘 ≠ 0.

Multiply by 𝑋 𝑛 and sum over 𝑛 ≥ 𝑘,


∞ ∞ ∞

෍ 𝑎𝑛 𝑋 𝑛 + 𝑋𝑐1 ෍ 𝑎𝑛−1 𝑋 𝑛−1 + 𝑋 2 𝑐2 ෍ 𝑎𝑛−2 𝑋 𝑛−2


𝑛=𝑘 𝑛=𝑘 𝑛=𝑘
+ ⋯ + 𝑋 𝑘 𝑐𝑘 σ∞ 𝑎
𝑛=𝑘 𝑛−𝑘 𝑋 𝑛−𝑘 = 0 -----------------(*)
BITS Pilani, Pilani Campus
Now by shifting of variables,

∞ ∞

෍ 𝑎𝑛−𝑟 𝑋 𝑛−𝑟 = ෍ 𝑎𝑛 𝑋 𝑛
𝑛=𝑘 𝑛=𝑘−𝑟
𝑘−𝑟−1

= 𝐴 𝑋 − ෍ 𝑎𝑛 𝑋 𝑛
𝑛=0

BITS Pilani, Pilani Campus


Substituting in (*) and solving for 𝐴 𝑋 , we get
𝑃(𝑋)
𝐴 𝑋 = , where
𝑄(𝑋)

𝑃 𝑋 = 𝑎0 + 𝑎1 + 𝑐1 𝑎0 𝑋 + ⋯ + 𝑎𝑘−1 + 𝑐1 𝑎𝑘−2 + ⋯ + 𝑐𝑘−1 𝑎0 𝑋 𝑘−1 ,


𝑄 𝑋 = 1 + 𝑐1 𝑋 + 𝑐2 𝑋 2 + ⋯ + 𝑐𝑘 𝑋 𝑘

RHS can be expanded as a power series in X by putting it into


partial fractions.
𝑎𝑛 = (coefficient of 𝑋 𝑛 in this power series) gives us a solution
to the recurrence relation.

BITS Pilani, Pilani Campus


Factorization of Q(X)

• The partial fractions method suggests the importance of


the role played by factorization of Q(X).
• Suppose 𝑄 𝑋 = 1 − 𝛼1 𝑋 𝑚1 … 1 − 𝛼𝑠 𝑋 𝑚𝑠 .
• After putting partial fractions, we use binomial
𝐴
expansions of its terms, of the type
1−𝛼𝑖 𝑋𝑙 , 𝑙 ≤ 𝑚𝑖 − 1.

The coefficient of 𝑋 𝑛 is 𝑓(𝑛)𝛼𝑖𝑛 , where 𝑓(𝑛) = 𝐶 (𝑙 − 1 +


𝑛, 𝑙 − 1)𝐴 is a polynomial in 𝑛 of degree at most 𝑙 − 1.
• Adding all the terms, we get 𝑎𝑛 = σ𝑠𝑖=1 𝑃𝑖 (𝑛)𝛼𝑖𝑛 ; where
𝑃𝑖 (𝑛) is a polynomial of degree at most 𝑚𝑖 − 1.
• The results can be summarized to give the form of the
solution.

BITS Pilani, Pilani Campus


Method of characteristic roots

To put the method in succinct form, we make some


definitions.

𝑘 1
If we let 𝐶 𝑡 = 𝑡 𝑄 , then 𝐶 𝑡 = 𝑡 𝑘 + 𝑐1 𝑡 𝑘−1 + ⋯ + 𝑐𝑘 .
𝑡
𝐶(𝑡) is called the characteristic polynomial of the given
recurrence relation. Roots of 𝐶 𝑡 are 𝛼𝑖 , 1 ≤ 𝑖 ≤ 𝑠; called
the characteristic roots of the given recurrence relation

BITS Pilani, Pilani Campus


Case I : Distinct roots

Theorem : Suppose characteristic polynomial 𝐶(𝑡) of the


homogeneous linear recurrence relation of degree k with
constant coefficients has k distinct roots 𝛼1 , 𝛼2 , … , 𝛼𝑘 .
Then the general form of its solution is given by
𝑎𝑛 = 𝐶1 𝛼1𝑛 + 𝐶2 𝛼2𝑛 + ⋯ + 𝐶𝑘 𝛼𝑘𝑛 , where 𝐶1 ,…, 𝐶𝑘 are
arbitrary constants.
• If 𝑘 initial conditions are given then these k arbitrary
constants can be found by solving the resulting system
of linear equations.
Remark : These theorems have analogues in theory of
homogeneous linear differential equations with constant
coefficients. Analogue of characteristic equation is called
auxiliary equation.
BITS Pilani, Pilani Campus
Find the general solution of the recurrence relation
𝑎𝑛 − 𝑎𝑛−1 − 𝑎𝑛−2 = 0, 𝑛 ≥ 2.
The characteristic polynomial 𝐶 𝑡 = 𝑡 2 − 𝑡 − 1 = 0 has distinct
1+ 5 1− 5
roots 𝛼1 = , 𝛼2 = .
2 2
Hence the general solution of the recurrence relation is
𝑛 𝑛
1+ 5 1− 5
𝑎𝑛 = 𝐶1 + 𝐶2
2 2
Where 𝐶1 , 𝐶2 are arbitrary constants.
Initial conditions 𝑎0 = 1, 𝑎1 = 1 give linear equations
1 1
1+ 5 1− 5
𝐶1 + 𝐶2 = 1, 𝐶1 + 𝐶2 =1 give
2 2
1
1+ 5 1 1− 5
𝐶1 = , 𝐶2 = −
5 2 5 2
BITS Pilani, Pilani Campus
BITS Pilani, Pilani Campus
Multiple Roots

Theorem : Assume that the characteristic polynomial 𝐶(𝑡)


for a homogeneous linear recurrence relation of degree
𝑘 with constant coefficients has roots 𝛼1 , 𝛼2 , … , 𝛼𝑠 , 𝑠 ≤ 𝑘
with multiplicities 𝑚1 , 𝑚2 , … , 𝑚𝑠 respectively. Then the
general solution of the recurrence relation is given by
𝑎𝑛 = 𝑈1 (𝑛) + 𝑈2 (𝑛) + ⋯ + 𝑈𝑠 (𝑛) where 𝑈𝑖 (𝑛) =
𝐷𝑖,0 + 𝐷𝑖,1 𝑛 + ⋯ + 𝐷𝑖,𝑚𝑖−1 𝑛𝑚𝑖−1 𝛼𝑖𝑛 , 𝐷𝑖,𝑗 are arbitrary
constants.
• Since the total number of arbitrary constants is 𝑚1 + ⋯ +
𝑚𝑠 = 𝑘, 𝑘 initial values can be used to determine these
constants. In particular, we can use the initial conditions
giving the values of 𝑎0 , 𝑎1 ,…, 𝑎𝑘−1 .

BITS Pilani, Pilani Campus


Find the general solution of the recurrence relations

BITS Pilani, Pilani Campus


Suppose the characteristic polynomial for a linear homogeneous
recurrence relation with constant coefficients is
(𝑡 − 2)3 (𝑡 − 3)2 (𝑡 − 4)3 . Then the recurrence relation has
general solution
𝑎𝑛 = 𝐷1 + 𝐷2 𝑛 + 𝐷3 𝑛2 2𝑛 + 𝐷4 + 𝐷5 𝑛 3𝑛
+ 𝐷6 + 𝐷7 𝑛 + 𝐷8 𝑛2 4𝑛 , 𝐷1 ,…,𝐷8 arbitrary constants.

Write the linear homogeneous recurrence relation from the


characteristic polynomial 𝐶 𝑡 = 𝑡 8 + 𝑐1 𝑡 7 + ⋯ + 𝑐8

𝑎𝑛 + 𝑐1 𝑎𝑛−1 + ⋯ + 𝑐8 𝑎𝑛−8 = 0.

BITS Pilani, Pilani Campus


Find the characteristic polynomial of the homogeneous linear
recurrence relation with constant coefficients whose general
solution has the form
(a)𝑎𝑛 = 𝐵1 + 𝐵2 𝑛.
Solution : 𝛼1𝑛 = 1𝑛 has coefficient a polynomial in 𝑛 of degree 1.
Hence 𝐶 𝑡 = (𝑡 − 1)2 .
(f) 𝑎𝑛 = 𝐵1 2𝑛 + 𝐵2 𝑛2𝑛 + 𝐵3 3𝑛 + 𝐵4 𝑛3𝑛 + 𝐵5 6𝑛 .
Solution : 𝛼1𝑛 = 2𝑛 has coefficient a polynomial in 𝑛 of degree 1.
Hence 2 is a double root of 𝐶 𝑡 .
𝛼2𝑛 = 3𝑛 has coefficient a polynomial in 𝑛 of degree 1. Hence 3 is
a double root of 𝐶 𝑡 .
Last term gives : 6 is a simple root of 𝐶 𝑡 .
∴𝐶 𝑡 = 𝑡−2 2 𝑡−3 2 𝑡−6 .
BITS Pilani, Pilani Campus
Solve the recurrence relation using the method of
characteristic roots :
𝑎𝑛 + 𝑎𝑛−1 − 5𝑎𝑛−2 + 3𝑎𝑛−3 = 0; 𝑎0 = 0, 𝑎1 = 1, 𝑎2 = 2.

Solution :
The characteristic equation is C t = 𝑡 3 + 𝑡 2 − 5𝑡 + 3 = 0.
Sum of the coefficients=0, hence 1 is a root.
Dividing by 𝑡 − 1, we get the quadratic 𝑡 2 + 2𝑡 − 3 = 0.
Thus characteristic roots are 1,1, −3.
Hence the general solution is 𝑎𝑛 = 𝐶1 + 𝐶2 𝑛 + 𝐶3 (−3)𝑛 , 𝑛 ≥ 0.

BITS Pilani, Pilani Campus


𝑎0 = 0 implies 𝐶1 + 𝐶3 = 0.
𝑎1 = 1 implies 𝐶1 + 𝐶2 − 3𝐶3 = 1.
𝑎2 = 2 implies 𝐶1 + 2𝐶2 + 9𝐶3 = 2.
Solving, we get 𝐶1 = 𝐶3 = 0, 𝐶2 = 1.
Hence the solution of the recurrence relation is 𝑎𝑛 = 𝑛, 𝑛 ≥ 0.

BITS Pilani, Pilani Campus


Remark

Suppose 𝑎𝑛 , 𝑏𝑛 are both solutions of the same linear


homogeneous recurrence relation with initial conditions
𝑎0 = 𝑐, 𝑎1 = 𝑑, and 𝑎0 = 𝑒 𝑎1 = 𝑓, and if 𝛼, 𝛽 are
constants then 𝛼𝑎𝑛 + 𝛽𝑏𝑛 is also a solution of the
same linear homogeneous recurrence relation with initial
conditions 𝑎0 = 𝛼𝑐 + 𝛽𝑒, 𝑎1 = 𝛼𝑑 + 𝛽𝑓. Thus solutions of
a homogeneous linear recurrence relation form a vector
space.

BITS Pilani, Pilani Campus


BITS Pilani, Pilani Campus

You might also like