GENERATING FUNCTIONS

Ankur 2/19/2020
Singh Ankur Singh 1
• Generating functions are used to represent sequences efficiently by coding the
terms of a sequence as coefficients of powers of a variable 𝑥 in a formal power
series.
• Generating functions can be used to solve many types of counting problems, such
as the number of ways to select or distribute objects of different kinds, subject to
a variety of constraints, and the number of ways to make change for a dollar
using coins of different denominations.
• Generating functions can be used to solve recurrence relations by translating a
recurrence relation for the terms of a sequence into an equation involving a
generating function. This equation can then be solved to find a closed form for the
generating function. From this closed form, the coefficients of the power series for
the generating function can be found, solving the original recurrence relation.
• Generating functions can also be used to prove combinatorial identities by taking
advantage of relatively simple relationships between functions that can be
translated into identities involving the terms of sequences.
2/19/2020 Ankur Singh 2
akxk.

DEFINITION 1
• The generating function for the sequence 𝑎0 , 𝑎1 , . . . , 𝑎𝑘 , . . . of real numbers is the
infinite series

𝐺 𝑥 = 𝑎0 + 𝑎1 𝑥 + 𝑎2 𝑥 2 + 𝑎3 𝑥 3 + ⋯ + 𝑎𝑘 𝑥 𝑘 + ⋯ = ෍ 𝑎𝑘 𝑥 𝑘
𝑘=0

Remark: The generating function for{𝑎𝑘 } given in Definition 1 is sometimes

called the ordinary generating function of {𝑎𝑘 } to distinguish it from other
types of generating functions for this sequence.

2/19/2020 Ankur Singh 3

Example:
Show that the generating function for the sequences {𝑎𝑘 } with 𝑎𝑘 = 3,
∞ ∞
𝑎𝑘 = 𝑘 + 1, and 𝑎𝑘 = 2 are ෌𝑘=0 3𝑥 , ෌𝑘=0 𝑘 + 1 𝑥 𝑘 , and
𝑘 𝑘
∞
෌𝑘=0 2𝑘 𝑥 𝑘 respectively.
Solution:
Since 𝑎𝑘 = 3𝑘 , implies 𝑎0 = 30 = 1; 𝑎1 = 31 = 3, 𝑎2 = 32 = 9
Therefore, by the generating function 𝐺 𝑥 = 𝑎0 + 𝑎1 𝑥 + 𝑎2 𝑥 2 + 𝑎3 𝑥 3 +
∞
⋯ + 𝑎𝑘 𝑥 + ⋯ = ෌𝑘=0 𝑎𝑘 𝑥 𝑘 , we have
𝑘

2 2 3 3 𝑘 𝑘 ∞
𝐺 𝑥 = 1 + 3 ∗ 𝑥 + 3 ∗ 𝑥 + 3 ∗ 𝑥 + ⋯+ 3 ∗ 𝑥 +. . = ෌𝑘=0 3𝑘 𝑥 𝑘
Similarly, we can get the other generating functions.
2/19/2020 Ankur Singh 4
• We can define generating functions for finite sequences of real numbers
by extending a finite sequence 𝑎0 , 𝑎1 , . . . , 𝑎𝑛 into an infinite sequence by
setting 𝑎𝑛+1 = 0, 𝑎𝑛+2 = 0, and so on.
• Thegenerating function 𝐺(𝑥) of this infinite sequence {𝑎𝑛 } is a
polynomial of degree 𝑛 because no terms of the form 𝑎𝑗 𝑥 𝑗 with 𝑗 > 𝑛
occur, that is,

𝐺 𝑥 = 𝑎0 + 𝑎1 𝑥 + · · · + 𝑎𝑛 𝑥 𝑛 .

2/19/2020 Ankur Singh 5

EXAMPLE 2
What is the generating function for the sequence 1, 1, 1, 1, 1, 1?
Solution:
The generating function of 1, 1, 1, 1, 1, 1 is 1 + 𝑥 + 𝑥 2 + 𝑥 3 + 𝑥 4 + 𝑥 5 .
By Theorem 1 of Section 2.4 we have
𝑥6 − 1
= 1 + 𝑥 + 𝑥 2 + 𝑥 3 + 𝑥 4 + 𝑥 5.
𝑥 − 1
when 𝑥 ≠ 1. Consequently, 𝐺(𝑥) = (𝑥 6 − 1)/(𝑥 − 1) is the generating function
of the sequence 1, 1, 1, 1, 1, 1.
Note: Because the powers of x are only place holders for the terms of the sequence
in a generating function, we do not need to worry that G(1) is undefined.]

2/19/2020 Ankur Singh 6

EXAMPLE 3
Let m be a positive integer. Let 𝑎𝑘 = 𝐶(𝑚, 𝑘), for 𝑘 = 0, 1, 2, . . . , 𝑚.
What is the generating function for the sequence 𝑎0 , 𝑎1 , . . . , 𝑎𝑚 ?

Solution:
The generating function for this sequence is

𝐺 𝑥 = 𝐶 𝑚, 0 + 𝐶 𝑚, 1 𝑥 + 𝐶 𝑚, 2 𝑥 2 + · · · + 𝐶 𝑚, 𝑚 𝑥 𝑚 .

The binomial theorem shows that 𝐺(𝑥) = 1 + 𝑥 𝑚

2/19/2020 Ankur Singh 7

 Useful Facts About Power Series
• When generating functions are used to solve counting problems, they are
usually considered to be formal power series. Questions about the
convergence of these series are ignored.
• However, to apply some results from calculus, it is sometimes important to
consider for which x the power series converges. The fact that a function
has a unique power series around 𝑥 = 0 will also be important.
• Generally, however, we will not be concerned with questions of
convergence or the uniqueness of power series in our discussions.
• Readers familiar with calculus can consult textbooks on this subject for
details about power series, including the convergence of the series we
consider here.
2/19/2020 Ankur Singh 8
• We now state some important facts about infinite series used when
working with generating functions. A discussion of these and related
results can be found in calculus texts.

EXAMPLE 4
1
The function 𝑓 𝑥 = is the generating function of the sequence
1−𝑥
1, 1, 1, 1, . . . , because

1
= 1 + 𝑥 + 𝑥2 + · · · for |𝑥| < 1.
1−𝑥

2/19/2020 Ankur Singh 9

EXAMPLE 5
1
The function 𝑓 𝑥 = is the generating function of the sequence
1 − 𝑎𝑥

1, 𝑎, 𝑎2 , 𝑎3 , . . . ,

1
Because = 1 + 𝑎 𝑥 + 𝑎2 𝑥 2 + ⋯ when |𝑎𝑥| < 1,
1 − 𝑎𝑥

or
1
equivalently, when 𝑥 < for 𝑎 ≠ 0.
𝑎

2/19/2020 Ankur Singh 10

 THEOREM 1
∞ ∞
• Let 𝑓 𝑥 = ෌𝑘=0 𝑎𝑘 𝑥 𝑘 and 𝑔(𝑥) = ෌𝑘=0 𝑏𝑘 𝑥 𝑘 . Then

∞
𝑓 𝑥 +𝑔 𝑥 = ෌𝑘=0 𝑎𝑘 + 𝑏𝑘 𝑥 𝑘 and
∞
𝑘
𝑓 𝑥 𝑔 𝑥 = ෎ ෍ 𝑎𝑗ሶ 𝑏𝑘−𝑗 𝑥 𝑘 .
𝑗=0
𝑘=0
• Remark: Theorem 1 is valid only for power series that converge in an interval, as
all series considered in this section do. However, the theory of generating
functions is not limited to such series. In the case of series that do not converge,
the statements in Theorem 1 can be taken as definitions of addition and
multiplication of generating functions.
2/19/2020 Ankur Singh 11
EXAMPLE 6
1
Let 𝑓 𝑥 = . Use Example 4 to find the coefficients 𝑎0 , 𝑎1 , 𝑎2 , . . . in the
1−𝑥 2
∞
expansion 𝑓 (𝑥) = ෌𝑘=0 𝑎𝑘 𝑥 𝑘
Solution:

From Example 4 we see that

1
= 1 + 𝑥 + 𝑥2 + 𝑥3 + · · · .
1−𝑥

Hence, from Theorem 1, we have

∞
1 ∞
=෍ σ𝑘𝑗=0 1 𝑘
𝑥 = ෌𝑘=0 𝑘 + 1 𝑥𝑘.
1−𝑥 2 𝑘=0
2/19/2020 Ankur Singh 12
DEFINITION 2
Let 𝑢 be a real number and 𝑘 a nonnegative integer. Then the extended binomial
𝑢
coefficient is defined by
𝑘

Note: Here, we are finding the binomial coefficients for the real number 𝑢
(need not be a positive integer always). If 𝑢 is a positive integer then, our
earlier binomial coefficients can be obtained as in earlier Section.

2/19/2020 Ankur Singh 13

2/19/2020 Ankur Singh 14
2/19/2020 Ankur Singh 15
THEOREM 2 THE EXTENDED BINOMIAL THEOREM
Let 𝑥 be a real number with |𝑥| < 1 and let 𝑢 be a real number. Then

Remark: When 𝑢 is a positive integer, the extended binomial theorem reduces

to the binomial theorem presented in Section 6.4, because in that case

2/19/2020 Ankur Singh 16

 EXAMPLE 9
• Find the generating functions for 1 + 𝑥 −𝑛 and 1 − 𝑥 −𝑛 , where 𝑛
is a positive integer, using the extended binomial theorem.
Solution: By the extended binomial theorem, it follows that

2/19/2020 Ankur Singh 17

2/19/2020 Ankur Singh 18
 Some useful Identities

2/19/2020 Ankur Singh 19

2/19/2020 Ankur Singh 20
2/19/2020 Ankur Singh 21
2/19/2020 Ankur Singh 22
 Counting Problems and Generating Functions
Generating functions can be used to solve a wide variety of counting problems.
In particular, they can be used to count the number of combinations of various
types.
In Chapter 6 we developed techniques to count the 𝑟 −combinations from a set
with 𝑛 elements when repetition is allowed and additional constraints may
exist.
Such problems are equivalent to counting the solutions to
equations of the form
𝑒1 + 𝑒2 + · · · + 𝑒𝑛 = 𝐶,
where 𝐶 is a constant and each 𝑒𝑖 is a nonnegative integer that may be subject
to a specified constraint.
Generating functions can also be used to solve counting problems of this type,
as Examples 10–12 show
2/19/2020 Ankur Singh 23
EXAMPLE 10
Find the number of solutions of 𝑒1 + 𝑒2 + 𝑒3 = 17,
where 𝑒1 , 𝑒2 , and 𝑒3 are nonnegative integers with
2 ≤ 𝑒1 ≤ 5, 3 ≤ 𝑒2 ≤ 6, and 4 ≤ 𝑒3 ≤ 7.
Solution:
The number of solutions with the indicated constraints is the coefficient of
𝑥 17 in the expansion of

(𝑥 2 + 𝑥 3 + 𝑥 4 + 𝑥 5 )(𝑥 3 + 𝑥 4 + 𝑥 5 + 𝑥 6 )(𝑥 4 + 𝑥 5 + 𝑥 6 + 𝑥 7 ).
This follows because we obtain a term equal to 𝑥 17 in the product by
picking a term in the first sum 𝑥 𝑒1 , a term in the second sum 𝑥 𝑒2 , and a
term in the third sum 𝑥 𝑒3 , where the exponents 𝑒1 , 𝑒2 , and 𝑒3 satisfy the
equation 𝑒1 + 𝑒2 + 𝑒3 = 17 and the given constraints. (Cont…)
2/19/2020 Ankur Singh 24
It is not hard to see that the coefficient of 𝑥 17 in this product is 3.
Hence, there are three solutions.

Note that the calculating of this coefficient involves about as

much work as enumerating all the solutions of the equation with
the given constraints. However, the method that this illustrates
often can be used to solve wide classes of counting problems with
special formulae, as we will see. Furthermore, a computer algebra
system can be used to do such computations.

2/19/2020 Ankur Singh 25

Determine the coefficient of 𝑥15 in
𝑓(𝑥) = (𝑥2 + 𝑥3 + 𝑥4 + ⋯ )4.

𝑥2
(𝑥2 + 𝑥3 + 𝑥4 + ⋯ = 𝑥2(1 + 𝑥 + 𝑥2 + ⋯ ) =
1−𝑥

𝑥8
𝑓(𝑥) = (𝑥2/(1 − 𝑥))4 = 1 − 𝑥 4

Hence the solution is the coefficient of 𝑥7 in 1 − 𝑥 −4 which is

𝐶(−4, 7)(−1)7 = 𝐶(10, 7).

2/19/2020 Ankur Singh 26

Find the number of ways to select R
balls from a pile of 2 red, 2 green
and 2 blue balls

𝑋 𝑒1 𝑋 𝑒2 𝑋 𝑒3

e1 + e2 + e3 = R
EXAMPLE 11
In how many different ways can eight identical cookies be distributed among three distinct
children if each child receives at least two cookies and no more than four cookies?
Solution: Because each child receives at least two but no more than four
cookies, for each child there is a factor equal to
(𝑥 2 + 𝑥 3 + 𝑥 4 )
in the generating function for the sequence {𝑐𝑛 }, where 𝑐𝑛 is the number of
ways to distribute 𝑛 cookies. Because there are three children, this generating
function is
𝑥2 + 𝑥3 + 𝑥4 3.
onstraints is the coefficient of 𝑥 17 in the expansion of

2/19/2020 Ankur Singh 28

EXAMPLE 12 (See Page no 543)
Use generating functions to determine the number of ways to insert
tokens worth $1, $2, and $5 into a vending machine to pay for an item that
costs 𝑟 dollars in both the cases when the order in which the tokens are
inserted does not matter and when the order does matter.
For example, there are two ways to pay for an item that costs $3 when the
order in which the tokens are inserted does not matter: inserting three $1
tokens or one $1 token and a $2 token. When the order matters, there are
three ways: inserting three $1 tokens, inserting a $1 token and then a $2
token, or inserting a $2 token and then a $1 token.

2/19/2020 Ankur Singh 29

EXAMPLE 13 (See Page no: 544)
Use generating functions to find the number of 𝑘 −combinations of a set
with 𝑛 elements. Assume that the binomial theorem has already been
established.
EXAMPLE 14 (See Page no: 545)
Use generating functions to find the number of 𝑟 −combinations from a
set with 𝑛 elements when repetition of elements is allowed.

EXAMPLE 15 (See Page no: 546)

Use generating functions to find the number of ways to select 𝑟 objects
of 𝑛 different kinds if we must select at least one object of each kind.

2/19/2020 Ankur Singh 30

Using Generating Functions to Solve Recurrence Relations
EXAMPLE 16
Solve the recurrence relation 𝑎𝑘 = 3𝑎𝑘−1 for 𝑘 = 1, 2, 3, . . . and initial
condition 𝑎0 = 2.
Solution: Let 𝐺(𝑥) be the generating function for the sequence {𝑎𝑘 }, that is,
∞
𝐺(𝑥) = ෌𝑘=0 𝑎𝑘 𝑥 𝑘 .

First note that

∞ ∞
𝑥𝐺 𝑥 = ෌𝑘=0 𝑎𝑘 𝑥 𝑘+1 = ෌𝑘=1 𝑎𝑘−1 𝑥 𝑘 .
Using the recurrence relation, we see that
𝐺 𝑥 −3𝑥𝐺 𝑥 =
∞ ∞ ∞
෌𝑘=0 𝑎𝑘 𝑥 𝑘 -3 ෌𝑘=1 𝑎𝑘−1 𝑥 𝑘 = 𝑎0 + ෌𝑘=1(𝑎𝑘 −3𝑎𝑘−1 )𝑥 𝑘
2/19/2020 Ankur Singh 31
Since 𝑎0 = 2 and 𝑎𝑘 = 3𝑎𝑘−1 .
Thus, 𝐺 𝑥 − 3 𝑥 𝐺 𝑥 = 2 + 0.
2
That is, 𝐺 𝑥 − 3 𝑥 𝐺 𝑥 = 1 − 3𝑥 𝐺 𝑥 = 2 ⇒ 𝐺 𝑥 = .
(1−3𝑥)

1 ∞
Using the identity = ෌𝑘=0 𝑎𝑘 𝑥 𝑘 , from Table 1, we have
1−𝑎𝑥

∞
∞

𝐺(𝑥) = 2 ා 3𝑘 𝑥 𝑘 = ෍ 2 ∗ 3𝑘 𝑥 𝑘 .
𝑘=0
𝑘=0
Consequently, 𝑎𝑘 = 2 ∗ 3𝑘
2/19/2020 Ankur Singh 32
 Example with another method/approach
The sequence of numbers 1,1,2,3,5,8,13, .... is known as the Fibonacci
sequence. We can describe the sequence in terms of a recurrence relation
as : 𝑎0 = 1, 𝑎1 = 1 and
𝑎𝑛 = 𝑎𝑛−1 + 𝑎𝑛−2 for 𝑛 ≥ 2.
How do we find a formula for 𝑎𝑛 ?
Let 𝐺(𝑥) = 𝑎0 + 𝑎1 𝑥 + 𝑎2 𝑥 2 +· · · +𝑎𝑛 𝑥 𝑛 +· · · be the generating
function for the sequence.

2/19/2020 Ankur Singh 33

Then 𝐺 𝑥 = 1 + 𝑥 + σ∞ 𝑎
𝑛=2 𝑛 𝑥 𝑛

∞
=1+𝑥+ ෌𝑛=2(𝑎𝑛−1 + 𝑎𝑛−2 )𝑥 𝑛
= 1 + 𝑥 + 𝑎1 𝑥 2 + 𝑎2 𝑥 3 + ⋯ + (𝑎0 𝑥 2 + 𝑎1 𝑥 3 + 𝑎2 𝑥 4 + ⋯ )
= 1 + 𝑥 + 𝑥 ∗ 𝑎1 𝑥 + 𝑎2 𝑥 2 + ⋯ . + 𝑥 2 (𝑎0 + 𝑎1 𝑥 + 𝑎2 𝑥 2 + ⋯ . )
= 1 + 𝑥 + 𝑥 𝐺 𝑥 − 1 + 𝑥 2 𝐺(𝑥)

Therefore, 𝐺 𝑥 = 1 + 𝑥 + 𝑥 𝐺 𝑥 − 1 + 𝑥 2 𝐺(𝑥).

1 1
From this, we obtain 𝐺 𝑥 = = − 2
1−𝑥−𝑥 2 𝑥 +𝑥−1

2/19/2020 Ankur Singh 34

2/19/2020 Ankur Singh 35
2/19/2020 Ankur Singh 36
2/19/2020 Ankur Singh 37
2/19/2020 Ankur Singh 38
2/19/2020 Ankur Singh 39
EXAMPLE 17
Suppose that a valid codeword is an 𝑛-digit number in decimal notation containing an
even number of 0s. Let 𝑎𝑛 denote the number of valid codewords of length 𝑛. In
Example 4 of Section 8.1 we showed that the sequence {𝑎𝑛 } satisfies the recurrence
relation
𝑎𝑛 = 8 𝑎𝑛−1 + 10𝑛−1
And the initial condition a1 = 9. Using generating functions to find an explicit formula
for 𝑎𝑛 .
Solution:
To make our work with generating functions simpler, we extend this sequence
by setting 𝑎0 = 1; when we assign this value to 𝑎0 and use the recurrence
relation, we have 𝑎1 = 8𝑎0 + 100 = 8 + 1 = 9, which is consistent with our
original initial condition.
(It also makes sense because there is one code word of length 0—the empty
string.)
2/19/2020
(Cont…) Ankur Singh 40
We multiply both sides of the recurrence relation by 𝑥 𝑛 to obtain
𝑎𝑛 𝑥 𝑛 = 8 𝑎𝑛−1 𝑥 𝑛 + 10𝑛−1 𝑥 𝑛 .
Let 𝐺(𝑥) = σ∞
𝑛=0 𝑎 𝑛 𝑥 𝑛 be the generating function of the sequence 𝑎 , 𝑎 , 𝑎 , . . . .
0 1 2

We sum both sides of the last equation starting with 𝑛 = 1, to find that
∞ ∞

𝐺 𝑥 − 1 = ෍ 𝑎𝑛 𝑥 𝑛 = ෍ 8𝑎𝑛−1 𝑥 𝑛 + 10𝑛−1 𝑥 𝑛
𝑛=1 𝑛=1

∞ ∞

= 8 ෍ 𝑎𝑛−1 𝑥 𝑛 + ෍ 10𝑛−1 𝑥 𝑛
𝑛=1 𝑛=1

∞ ∞
= 8𝑥 ෌𝑛=1 𝑎𝑛−1 𝑥 𝑛−1 + 𝑥 ෌𝑛=1 10𝑛−1 𝑥 𝑛−1 (Cont…)
2/19/2020 Ankur Singh 41
 ∞ ∞

= 8𝑥 ෍ 𝑎𝑛 𝑥 𝑛 + 𝑥 ෍ 10𝑛 𝑥 𝑛 = 8 𝑥 𝐺(𝑥) + 𝑥/(1 − 10 𝑥)

𝑛=0 𝑛=0

where we have used Example 5 to evaluate the second summation. Therefore, we

𝑥
have 𝐺 𝑥 − 1 = 8 𝑥 𝐺 𝑥 + .
1 − 10 𝑥

Solving for 𝐺 𝑥 we get,

1 − 9𝑥
𝐺 𝑥 =
(1 − 8𝑥) 1 − 10 𝑥

Expanding the right-hand side of this equation into partial fractions (as is done in
the integration of rational functions studied in calculus) gives

1 1 1
𝐺 𝑥 = + . (Cont…)
2 1−8 𝑥 1−10 𝑥
2/19/2020 42

Ankur Singh
Using Example 5 twice (once with 𝑎 = 8 and once with 𝑎 = 10) gives
∞
∞ ∞
1 1 𝑛
𝐺 𝑥 = ෍ 8 𝑥 + ෍ 10 𝑥 = ෎ 8 + 10𝑛 𝑥 𝑛
𝑛 𝑛 𝑛 𝑛
2 2
𝑛=0 𝑛=0
𝑛=0
Consequently, we have shown that
1 𝑛
𝑎𝑛 = 8 + 10𝑛 .
2

2/19/2020 Ankur Singh 43

 Proving Identities via Generating Functions
EXAMPLE 18
Use generating functions to show that
𝑛 2
෌𝑘=0 𝐶 𝑛, 𝑘 = 𝐶 2𝑛, 𝑛 whenever 𝑛 is a positive integer.
Solution:
First note that by the binomial theorem 𝐶(2𝑛, 𝑛) is the coefficient of 𝑥 𝑛 in
1 + 𝑥 2𝑛 .
However, we also have

1+𝑥 2𝑥 = 1+𝑥 𝑛 2

= 𝐶 𝑛, 0 + 𝐶 𝑛, 1 𝑥 + 𝐶 𝑛, 2 𝑥 2 + ⋯ + 𝐶 𝑛, 𝑛 − 1 𝑥 𝑛−1 + 𝐶 𝑛, 𝑛 𝑥 𝑛 2.

2/19/2020 Ankur Singh 44

The coefficient of 𝑥 𝑛 in this expression is

𝐶 𝑛, 0 𝐶 𝑛, 𝑛 + 𝐶 𝑛, 1 𝐶 𝑛, 𝑛 − 1 + 𝐶 𝑛, 2 𝐶 𝑛, 𝑛 − 2 + · · ·
+ 𝐶 𝑛, 𝑛 𝐶 𝑛, 0 .
𝑛
This equals ෌𝑘=0 𝐶 𝑛, 𝑘 2 , since 𝐶 𝑛, 𝑛 − 1 = 𝐶 𝑛, 𝑘 .

𝑛 2
Because bothe 𝐶(2𝑛, 𝑛) and ෌𝑘=0 𝐶 𝑛, 𝑘 arepresent the coefficeint of 𝑥 𝑛
2𝑛
in 1 + 𝑥 , they must be equal.

2/19/2020 Ankur Singh 45