Beruflich Dokumente
Kultur Dokumente
Jemma Davies
1 Introduction 4
1.1 Biography of Leonardo Fibonacci . . . . . . . . . . . . . . . . 4
1.2 The Fibonacci Sequence . . . . . . . . . . . . . . . . . . . . . 6
1.3 Lucas and The Lucas Sequence . . . . . . . . . . . . . . . . . 6
1.4 Relationships between Fibonacci and Lucas Sequences . . . . . 7
1.5 The Generalised Fibonacci Sequence . . . . . . . . . . . . . . 9
2 Golden Ratio 11
2.1 History of the Golden Ratio . . . . . . . . . . . . . . . . . . . 11
2.2 Fibonacci Sequence and the Golden Ratio . . . . . . . . . . . 12
2.3 Binet’s Formula . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4 Binet’s Formula and Generalised Fibonacci Numbers . . . . . 15
3 Polynomials 18
3.1 Fibonacci Polynomials . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Generalised Fibonacci Polynomials . . . . . . . . . . . . . . . 18
3.3 Binet’s Formula and Factorisations . . . . . . . . . . . . . . . 20
3.4 More Fibonacci Polynomial Expressions . . . . . . . . . . . . 23
4 Further Generalisations 28
4.1 Tribonacci Sequence . . . . . . . . . . . . . . . . . . . . . . . 28
4.2 Characteristic Equation . . . . . . . . . . . . . . . . . . . . . 30
4.3 Binet’s Formula . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5 Order K 32
5.1 Fibonacci Sequence of Order K . . . . . . . . . . . . . . . . . 32
5.2 Fibonacci Polynomials of Order K . . . . . . . . . . . . . . . . 35
5.3 Fibonacci-Type Polynomials of Order K . . . . . . . . . . . . 35
6 Applications 40
6.1 Arrangements . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2
6.2 Coin Tossing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
6.3 Bernoulli Trials . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.4 Geometric Distribution of Order K . . . . . . . . . . . . . . . 46
6.5 Longest Runs . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.6 Reliability of a Consecutive-k-out-of-n:F System . . . . . . . . 50
7 Conclusion 52
Bibliography 52
A Fundamentals 55
A.1 Binomial Coefficients . . . . . . . . . . . . . . . . . . . . . . . 55
A.2 Multinomial Coefficients . . . . . . . . . . . . . . . . . . . . . 55
A.3 Generating Functions . . . . . . . . . . . . . . . . . . . . . . . 56
3
Chapter 1
Introduction
On his return to Italy in the early 13th century, Fibonacci began writing
Liber Abaci (Book of the Abacus), a book based on the arithmetic and alge-
bra he had learned whilst travelling, which was published in 1202 and later
revised in 1228. It was this book which introduced the Hindu-Arabic system
of numbers to Western Europe, a notable event, as it is the same system used
today. He later went on to write Practica Geometriae (Practice of Geometry)
in 1220 and Flos (Flower) in 1225. Fibonacci’s most outstanding contribu-
tion to mathematics was his last publication Liber Quadratorum (Book of
4
Square Numbers) which, written in 1225, was a book based on number theory
that included techniques for finding Pythagorean triples.[MAC]
Fibonacci however, is most well known for the recurrence relation which
bears his name and is the solution to a small problem which he posed in
Liber Abaci about the population of rabbits.
The rabbit problem can be described by the following, which has been
adapted from [DUN].
‘A pair of newly born rabbits is placed in a confined space. This pair, and
every later pair, produces one pair of newly born rabbits every month, once
they reach their 2nd month of age. How many pairs will there be after 1,
2,... months assuming that no deaths occur?’
To solve the problem, let us first establish some useful notation. If we let
n represent the beginning of the nth month, then let Fn denote the number
of pairs of rabbits there are at the start of the nth month.
When n = 0, clearly we have no pairs of rabbits, so F0 = 0. At the start of
the first month we have 1 pair of rabbits, the ancestor pair, therefore F1 = 1.
There will still only be 1 pair of rabbits at the start of the second month
as the ancestor pair are only one month old and have not reached breeding
age yet, so F2 = 1 also. However, at the begining of the third month, the
ancestor pair have given birth to a pair of rabbits so we now have 2 pairs of
rabbits. When n = 4 we will have F4 = 3 as the ancestor pair would have
had a second pair of rabbits, but their first pair of offspring would not have
reached breeding age yet. Table 1.1 displays these results and the following
numbers of pairs of rabbits up to the beginning of the tenth month.
n 0 1 2 3 4 5 6 7 8 9 10
Fn 0 1 1 2 3 5 8 13 21 34 55
Table 1.1: Number of pairs of rabbits at the start of the nth month
In general there will be Fn pairs of rabbits at the start of the nth month,
and Fn+1 at the beginning of the (n+1)th month. At the start of the (n+2)th
month there will still be all the pairs from the (n + 1)th month, but there will
also be all the offspring of the pairs that were present in the nth month as
these will all be at breeding age. So for any n ≥ 0 the solution to the rabbit
problem is the recurrence relation Fn+2 = Fn+1 + Fn .
5
1.2 The Fibonacci Sequence
The Fibonacci sequence is one of the most interesting number sequences and
provides mathematicians of all ages and abilities with vast opportunities to
make conjectures and increase their knowledge in many areas of mathemat-
ics. The sequence has even led to the formation of The Fibonacci Association,
an organization of mathematicians who study Fibonacci and related integer
sequences. Founded in 1963 by Verner E.Hoggatt, Jr. (1921-1980), the orga-
nization produces articles related to these integer sequences in its publication
The Fibonacci Quarterly.[WIK]
Definition 1.2.1
We define the Fibonacci sequence by the recurrence relation
and setting the initial values as F0 = 0 and F1 = 1. This gives us the Fi-
bonacci numbers: 0, 1, 1, 2, 3, 5, 8, 13, 21, 34,...
The sequence did not appear in Fibonacci’s Liber Abaci, but was given
its name in 1876 by the French mathematician Francois Edouard Anatole
Lucas, who made many contributions to the study of the Fibonacci sequence,
as shall be seen later. It is widely believed that the Fibonacci sequence did
not originate from Fibonacci himself however, but was explained by him in
his book to make the Hindu-Arabic number system more understandable to
Western Europe. The sequence appears to have been studied prior to the
publication of Liber Abaci by Indian scholars, who were exploring rhythmic
patterns.[SUR]
6
Definition 1.3.1
We define the Lucas sequence by the recurrence relation
Ln+2 = Ln+1 + Ln n≥0 (1.2)
and setting the seed values as L0 = 2 and L1 = 1. This gives us the Lucas
numbers: 2, 1, 3, 4, 7, 11, 18, 29, 47, 76,...
Proof
Cassini’s Formula can be proved easily by induction.
When n = 1 in equation (1.3), we have F0 F2 − F12 = 0 · 1 − 12 = (−1)1 , so
the formula holds in this case.
We now assume the formula holds for any positive integer k, i.e. we assume
Fk−1 Fk+1 − Fk2 = (−1)k is true.
Then we find for n = k + 1, using the recurrence relation for the Fibonacci
numbers,
2 2
Fk Fk+2 − Fk+1 = (Fk+1 − Fk−1 )(Fk + Fk+1 ) − Fk+1
2 2
= Fk Fk+1 + Fk+1 − Fk Fk−1 − Fk−1 Fk+1 − Fk+1 .
7
So Cassini’s formula works for n = k + 1. Therefore by induction the formula
holds for all n ≥ 1.
Since Fn and Ln satisfy the same recurrence relation, it follows that there
would be many relations between the two sequences, one of which we shall
now look at in detail.[VAJ, pg 24]
Theorem 1.4.2
Ln+1 = Fn + Fn+2 n≥0 (1.4)
It can be seen by Table 1.2 that this relation holds, for example, L5 = 11 =
F4 + F6 . However, for a more rigorous proof we will use induction.
n 0 1 2 3 4 5 6 7 8 9 10
Fn 0 1 1 2 3 5 8 13 21 34 55
Ln 2 1 3 4 7 11 18 29 47 76 123
Proof
When n = 0 in equation (1.4) we have L1 = F0 + F2 = 0 + 1 = 1 so the
relation is true in this case.
We now assume the formula holds for any positive integer k, i.e. we assume
Lk+1 = Fk + Fk+2 is true.
Then we find for n = k + 1,
L(k+1)+1 = Lk+1 + Lk .
8
1.5 The Generalised Fibonacci Sequence
By looking at number sequences that follow the same recurrence relation as
the Fibonacci and Lucas sequences, but have arbitrary integer seed values, it
is possible to find common properties of the two sequences. These sequences
are called generalised Fibonacci sequences.[KOS, pg 109]
Definition 1.5.1
We define the generalised Fibonacci sequence as
The coefficients of each term in the generalised Fibonacci sequence are the
Fibonacci numbers, which leads us on to the following theorem.
Theorem 1.5.2
Let Gn denote the nth term of the generalised Fibonacci sequence. Then
Proof
The theorem can be proved by induction.
When n = 1, G3 = aF1 + bF2 = a + b, so the theorem holds in this case.
We shall now assume that the theorem holds for n = k, so Gk+2 = aFk +bFk+1
is assumed to be true.
Then we find for n = k + 1,
9
Using the induction hypothesis this equals
So the theorem holds for any n = k+1 and therefore the theorem is true for all
n ≥ 1.
Theorem 1.5.2 is a key relation that will be useful in the following chapters.
The main focus of this report is to expand upon the basics of the Fi-
bonacci sequence which have been discussed in this first chapter. The next
chapter will introduce the most well-known value associated with the Fibon-
naci sequence, the Golden Ratio. Progressing on to the third chapter, we
will study the Fibonacci polynomials, which will be followed by a chapter
concerning an expansion of the Fibonacci sequence, namely the Tribonacci
sequence. Chapter five will cover the Fibonacci sequence and polynomials of
order k, and finally we shall investigate their applications in probability and
statistics.
10
Chapter 2
Golden Ratio
There are many associations with the Golden Ratio, most common of which
is that of aesthetics. It is widely known that the Golden Ratio appears in
art, architecture and even nature, however this chapter will focus on the link
between the Golden Ratio and the Fibonacci sequence. It will show that the
Golden Ratio is vital to a key theorem for finding large Fibonacci numbers.
Definition 2.1.1
The Golden Ratio (τ ) is an irrational number which can be expressed as the
fraction √
1+ 5
τ= = 1.618033... (2.1)
2
11
2.2 Fibonacci Sequence and the Golden Ra-
tio
It was discovered by the German mathematician Johannes Kepler (1571-
1630) that the ratio of two consecutive Fibonacci numbers approaches a limit
as n becomes large. This is also true for the Lucas sequence and generalised
Fibonacci sequence. The following table 2.1 demonstrates this convergence.
Fn+1 Ln+1
n Fn Ln
1 3
1 1
= 1.000000 1
= 3.000000
2 4
2 1
= 2.000000 3
= 1.333333...
3 7
3 2
= 1.500000 4
= 1.750000
5 11
4 3
= 1.666666... 7
= 1.571428...
8 18
5 5
= 1.600000 11
= 1.636363...
89 199
10 55
= 1.618181... 123
= 1.617886...
987 2207
15 610
= 1.618032... 1364
= 1.618035...
10946 24476
20 6765
= 1.618033... 15127
= 1.618033...
Theorem 2.2.1
Fn+1 Gn+1
lim = τ = lim (2.2)
n→∞ Fn n→∞ Gn
We shall prove the theorem in the case of the generalised Fibonacci sequence,
so that it will also hold in the case of the Fibonacci and Lucas sequences.
12
Proof
Using Definition 1.5.1, the recurrence relation of the generalised Fibonacci
sequence, we see that
Gn+1 Gn−1 1
=1+ =1+ Gn
.
Gn Gn Gn−1
As n → ∞, lim GGn+1
n
= lim GGn−1
n
and so by setting x = lim GGn+1
n
, the
equation above becomes
1
x=1+ ⇐⇒ x2 = x + 1
x
⇐⇒ x2 − x − 1 = 0.
√
1+ 5
√
Solving this quadratic equation we get the solutions x = 2
and x =
1− 5
2
. As the limit is positive it follows that
√
Gn+1 1+ 5
lim = = τ.
n→∞ Gn 2
This formula demonstrates the link between the (generalised) Fibonacci num-
bers and the Golden Ratio. An important relation arises from this proof
which shall now be defined.
Definition 2.2.2
The characteristic equation for the recurrence relation for the (generalised)
Fibonacci numbers is
x2 − x − 1 = 0, (2.3)
√ √
1+ 5 1− 5
which has solutions τ = 2
and σ = 2
, the conjugate to τ .
13
pg 79]
Proof
Since τ is a root of the characteristic equation (2.3) it can be written as
τ 2 − τ − 1 = 0. (2.5)
τ2 =τ +1
τ3 = τ 2 + τ = 2τ + 1
τ4 = τ 3 + τ 2 = 3τ + 2
τ5 = τ 4 + τ 3 = 5τ + 3.
Here the coefficients are the Fibonacci numbers and we can form the equation
τ n = τ Fn + Fn−1 , (2.6)
τ n − τ Fn = σ n − σFn
τ n − σ n = (τ − σ)Fn
τ n − σn
= Fn .
τ −σ
Binet’s Formula can also be used to find large Lucas numbers.[KOS, pg 79]
Ln = τ n + σ n n≥0 (2.8)
14
Proof
Using Theorem 1.4.2 and Theorem 2.3.1 we find
Ln = Fn−1 + Fn+1
τ n−1 − σ n−1 τ n+1 − σ n+1
= +
τ −σ τ −σ
1 −1
( + τ) ( − σ)
= τn τ√ + σn σ √ .
5 5
Proof
We find by using Theorem 1.5.2 and Theorem 2.3.1 that
Gn = aFn−2 + bFn−1
a(τ n−2 − σ n−2 ) + b(τ n−1 − σ n−1 )
=
τ −σ
τ ( τ 2 + τ ) − σ n ( σa2 + σb )
n a b
= .
τ −σ
15
1 1
Since τ σ = −1 this gives us τ
= −σ and σ
= −τ , so the equation becomes
τ n (aσ 2 − bσ) − σ n (aτ 2 − bτ )
Gn = .
τ −σ
16
Since τ σ = −1 and by using Binet’s formula again, this gives us
From this we can see Catalan’s Formula is equal to Theorem 1.4.1 (Cassini’s
Formula) when k = 1.
As we have seen in this chapter, there are strong links between the Fi-
bonacci sequence and the Golden Ratio. The relations that we have found
will be important in the following chapters. In particular, Binet’s formula
will be useful and continually appears in proofs of theorems covered in the
next chapter, which examines the Fibonacci polynomials.
17
Chapter 3
Polynomials
Definition 3.1.1
The Fibonacci polynomials fn (x) are defined as
fn+2 (x) = xfn+1 (x) + fn (x) n ≥ 0, (3.1)
where the initial values are f0 (x) = 0 and f1 (x) = 1.
The first ten Fibonacci polynomials are shown in table 3.1.
At first glace, the link between the Fibonacci polynomials and the Fibonacci
sequence may be difficult to see, however if you sum the coefficients of each
term you will find they equal the Fibonacci sequence. Furthermore, you can
see we obtain the Fibonacci numbers by letting x = 1 so that fn (1) = Fn .
Definition 3.2.1
The generalised Fibonacci polynomials gn (x) are defined as
gn+2 (x) = xgn+1 (x) + gn (x) n ≥ 0, (3.2)
18
n fn (x)
0 0
1 1
2 x
3 x2 + 1
4 x3 + 2x
5 x4 + 3x2 + 1
6 x5 + 4x3 + 3x
7 x6 + 5x4 + 6x2 + 1
8 x7 + 6x5 + 10x3 + 4x
9 x8 + 7x6 + 15x4 + 10x2 + 1
10 x9 + 8x7 + 21x5 + 20x3 + 5x
Theorem 3.2.2
√
fn+1 (x) x + x2 + 4 gn+1 (x)
lim = τ (x) = = lim (3.3)
n→∞ fn (x) 2 n→∞ gn (x)
Again we shall prove the theorem in the case of the generalised Fibonacci
polynomials, so that it will also hold in the case of the Fibonacci polynomials.
Proof
Using Definition 3.2.1, the recurrence relation for the generalised Fibonacci
polynomials, we see that
19
As n → ∞, lim gn+1 (x)
gn (x)
gn (x)
= lim gn−1 (x)
and so by setting y = lim gn+1 (x)
gn (x)
, equation
(3.4) becomes
1
y =x+ ⇐⇒ y 2 = xy + 1 (3.5)
y
⇐⇒ y 2 − xy − 1 = 0. (3.6)
√
x+ x2 +4
Solving
√
the equation for y gives us the solutions y = 2
and y =
x− x2 +4
2
. Since the limit is positive it follows that
√
gn+1 (x) x + x2 + 4
lim = = τ (x).
n→∞ gn (x) 2
Proof
The proof of this theorem follows the same pattern as that of Theorem 2.3.1
(Binet’s formula for the Fibonacci numbers). Since τ (x) is a root of the
characteristic equation (3.6), it can be written as
τ 2 (x) − xτ (x) − 1 = 0.
20
Rearranging this and multiplying by τ (x) we find that
τ 2 (x) = xτ (x) + 1
τ 3 (x) = xτ 2 (x) + τ (x) = (x2 + 1)τ (x) + x
τ 4 (x) = xτ 3 (x) + τ 2 (x) = (x3 + 2x)τ (x) + x2 + 1
τ 5 (x) = xτ 4 (x) + τ 3 (x) = (x4 + 3x2 + 1)τ (x) + x3 + 2x.
We see that the coefficients are the Fibonacci polynomials and we can form
the equation
τ n (x) = τ (x)fn (x) + fn−1 (x). (3.8)
Similarly for σ(x) we find
Definition 3.3.2 √
If we let x = 2i cosh z, then x2 + 4 = 2 sinh z and we define
21
Using these values in Theorem 3.3.1 we can also define
(τ (x))n − (σ(x))n
fn (x) =
τ (x) − σ(x)
i e − in e−nz
n nz
e − e−nz
nz
n−1
= =i .
iez − ie−z ez − e−z
ez −e−z
Since sinh z = 2
this gives us
sinh nz
fn (x) = in−1 . (3.12)
sinh z
To find the zeros of the Fibonacci polynomials we need to look at the values of
z such that fn (x) = 0. This occurs if and only if sinh nz = 0 and sinh z 6= 0.
sinh nz = sinh n(x + iy)
en(x+iy) − e−n(x+iy)
= =0
2
This leads us to
e2n(x+iy) = 1
e2nx (cos 2ny + i sin 2ny) = 1.
Equating the real parts we find e2nx cos 2ny = 1. As e2nx ≥ 1 for nx ≥ 0, we
must have x = 0, which implies cos 2ny = 1. This is the case when y = ± kπ n
.
kπ
Equating the imaginary parts we find sin 2ny = 0, which implies y = ± 2n .
Combining these results we can see that sinh nz = 0 when x = 0 and y =
± kπ
n
, which implies z = ± kπ
n
i.
Using the identity i cosh iy = i cos y, we get x = 2i cosh z = 2i cos kπ
n
,
where 1 ≤ k ≤ n − 1 and find the zeros of the Fibonacci polynomials are
kπ
x = 2i cos 1 ≤ k ≤ n − 1. (3.13)
n
Definition 3.3.3
The Fibonacci polynomials can be factorised using the following expression
n−1
Y kπ
fn (x) = (x − 2i cos ). (3.14)
k=1
n
22
Setting x = 1 in the above equation we also find
n−1
Y kπ
Fn = (1 − 2i cos ). (3.15)
k=1
n
It can easily be seen by multiplying out the brackets that we have found the
factors of f3 (x) = x2 + 1.
Theorem 3.4.1
The generating function for the Fibonacci polynomials is given by
t
g(t) = . (3.16)
1 − xt − t2
[KOS, pg 447]
Proof
Let g(t) = ∞ n
P
n=0 fn (x)t be the generating function for the Fibonacci poly-
23
nomials. Using Definition 3.1.1 we find by putting in the initial values that
∞
X
g(t) = t + fn (x)tn
n=2
∞
X
=t+ (xfn−1 (x) + fn−2 (x))tn
n=2
X∞ ∞
X
r+1
=t+x fr (x)t + fs (x)ts+2
r=1 s=0
(τ n (x) − σ n (x))
fn (x) = ,
τ (x) − σ(x)
24
which is Binet’s formula for the Fibonacci polynomials. Therefore we have
proved equation (3.16) is the generting function for the Fibonacci polynomi-
als.
Theorem 3.4.2
The Fibonacci polynomials can be expressed explicitly in terms of binomial
coefficients by the formula
n−1
2
X n−j−1
fn (x) = xn−2j−1 n ≥ 0. (3.18)
j
j=0
[KOS, pg 448]
Before we can prove this theorem, it is necessary to define another type of
polynomial which has a recursive relation like the Fibonacci polynomials.
Definition 3.4.3
We define the Chebyshev polynomials of second kind Un (y) as
[PLA]
25
Proof Theorem (3.4.2)
This proof has been adapted from that found in [KOS, pg 448]. Using the
generating function for the Fibonacci polynomials
t
g(t) = ,
1 − xt − t2
we have ∞
t X
= fn (x)tn .
1 − xt − t2 n=0
Here we can see how similar this is to the generating function for the Fi-
bonacci polynomials, which we get by multiplying through by t, leading us
to ∞
t X
n
x
= i Un tn+1
1 − xt − t2 n=0
2i
∞
X ∞
X x
n
fn (x)t = in Un tn+1 .
n=0 n=0
2i
P∞ n
P∞
Since n=0 fn (x)t = n=0 fn+1 (x)tn+1 this gives
∞
X ∞
X x
n+1
fn+1 (x)t = in Un tn+1 .
n=0 n=0
2i
26
So we find
n−1
2
X n−j−1
fn (x) = xn−2j−1 .
j
j=0
By setting x = 1 in Theorem 3.4.2 we get the following result, which gives us
another way of finding the Fibonacci numbers, this time by using binomial
coefficients.
Corollary 3.4.4
n−1
2
X n−j−1
Fn = (3.21)
j
j=0
27
Chapter 4
Further Generalisations
Definition 4.1.1
The Tribonacci sequence satisfies the recurrence relation
n 0 1 2 3 4 5 6 7 8 9 10
Tn 0 1 1 2 4 7 13 24 44 81 149
28
Tn+1
n Tn
1 1.000000
2 2.000000
3 2.000000
4 1.750000
5 1.857142...
10 1.838926...
15 1.839285...
Definition 4.1.2
The Tribonacci polynomials tn (x) are defined by
with the initial values t0 (x) = 0, t1 (x) = 1 and t2 (x) = x2 . The first five
Tribonacci polynomials are shown in table 4.3.
n tn (x)
0 0
1 1
2 x2
3 x4 + x
4 x6 + 2x3 + 1
5 x8 + 3x5 + 3x2
29
4.2 Characteristic Equation
From the recurrence relation for the Tribonacci sequence we find that the
characteristic equation is
x3 − x2 − x − 1 = 0. (4.3)
To find the roots of the characteristic equation we need to solve the cubic
equation. The method to do this can be found in Appendix. B, which gives
the roots
√ √
q q
1 3 3
α = (1 + 19 + 3 33 + 19 − 3 33)
3
√ √
q q
1 1 3 3
β = − ( 19 + 3 33 + 19 − 3 33)
3 6 q
√ √
q
i 3 3
+ √ ( 19 + 3 33 − 19 − 3 33)
2 3
√ √
q q
1 1 3 3
γ = − ( 19 + 3 33 + 19 − 3 33)
3 6 q
√ √
q
i 3 3
− √ ( 19 + 3 33 − 19 − 3 33),
2 3
where β and γ are complex conjugates to the real root α, which is the Tri-
bonacci constant.
Proof
We can prove Binet’s formula for the Tribonacci sequence using induction
30
on n, in the same way that we proved Binet’s formula for the Fibonacci
sequence.
This chapter has shown that the Fibonacci sequence and Fibonacci poly-
nomials can be generalised from the sum of two terms to three. An obvious
progression from this is to find the Fibonacci sequence and Fibonacci poly-
nomials of order k, which will be investigated in the next chapter.
31
Chapter 5
Order K
Definition 5.1.1
(k)
We define the Fibonacci sequence of order k as Fn if
(k) (k)
F0 = 0, F1 = 1 and
(
(k) (k)
Fn−1 + ... + F0 2≤n≤k
Fn(k) = (k) (k) (5.1)
Fn−1 + ... + Fn−k n ≥ k + 1
(
(k)
2Fn−1 2≤n≤k
= (k) (k) (5.2)
2Fn−1 − Fn−1−k n ≥ k + 1.
(2) (2)
By setting k = 2 we get F0 = 0, F1
= 1 and
(
(2) (2)
F2−1 + F0 n=2
Fn(2) = (2) (2)
Fn−1 + Fn−2 n ≥ (2 + 1),
1
The notation I have used to define the Fibonacci sequence of order k differs from that
(k)
used in the literature. Specifically, the Fibonacci sequence of order k is denoted by fn ,
whilst I have used an upper case F for the sequence. By doing this, there is no change
from the earlier notation used in the definition of the Fibonacci sequence, so the material
is more comprehensible.
32
which is the Fibonacci sequence. Similarly it can be shown that when k = 3
we obtain the Tribonacci sequence.
Theorem 5.1.2
(k)
X n1 + ... + nk
Fn+1 = n ≥ 0, (5.3)
n1 , ..., nk
n1 ,...,nk
In order to prove this theorem, we will need to use two relations that are
discussed in detail and proved in chapter 6.
Proof
By equating Lemma 6.1.1 and equation (6.9) we find that the Fibonacci se-
quence of order k is equal to the number of arrangements of n elements (r
or s) where the last is r and there are no k adjacent that are all s, denoted
by akn (r). This gives
k
Fn+1 = akn (r),
with ak0 (r) = 1.
It is necessary to note that an arrangement of n elements (r or s) is of the
form akn (r) if and only if n1 of the elements are e1 = r, n2 are e2 = sr,..., and
nk are ek = |{z}
s...s r. In other words n1 + 2n2 + ... + knk = n.
k−1
33
where the summation is over all non-negative integers n1 , ..., nk such that
n1 + 2n2 + ... + knk = n.
From theorem 5.1.2 we get the following corollary
Corollary 5.1.3
n
2
X n−i
Fn+1 = n≥0 (5.4)
i
i=0
n n−2i
2 3
X X i + j n − i − 2j
Tn+1 = (5.5)
i i+j
i=0 j=0
To show how Corollary 5.1.3 follows from Theorem 5.1.2, we need to set
k = 2 in Theorem 5.1.2 to find equation (5.4). This gives
(2)
X n1 + n2
Fn+1 = .
n1 , n2
n1 ,n2
34
5.2 Fibonacci Polynomials of Order K
It is also possible to generalise Fibonacci polynomials up to order k in a
similar way to the Fibonacci sequence.
Definition 5.2.1
(k)
We define the Fibonacci polynomials of order k as fn (x) if
(k) (k)
f0 (x) = 0, f1 (x) = 1 and
(P
n k−i (k)
(k) i=1 x fn−i (x) 2 ≤ n ≤ k + 1
fn (x) = Pk k−i (k)
i=1 x fn−i (x) n ≥ k + 2
[PHI, pg 213].
Setting k = 2 and k = 3 gives us the Fibonacci and Tribonacci polynomials
respectively. Also, when x = 1 we obtain the Fibonacci sequence of order k,
however there is another generalisation of the Fibonacci polynomials which
is much more useful in applications, which we will now look at in depth.
(2) (2)
When we set k = 2 we get F0 (x) = 0, F1 (x) = 1 and
(
(2) (2)
x[F2−1 (x) + F0 (x)] n = 2
Fn(2) (x) = (2) (2)
x[Fn−1 (x) + Fn−2 (x)] n ≥ (2 + 1).
The first five Fibonacci-type polynomials of order 2 are shown in table 5.1.
As you can see, the Fibonacci-type polynomials of order 2 are not the same
as the Fibonacci polynomials as you might expect. However when x = 1 we
find that they behave in a similar way to the Fibonacci polynomials as
35
n Fn2 (x)
0 0
1 1
2 x
3 x2 + x
4 x3 + 2x2
5 x4 + 3x3 + x2
Theorem 5.3.2
(k)
X n1 + ... + nk
Fn+1 (x) = xn1 +...+nk n ≥ 0, (5.8)
n1 , ..., nk
n1 ,...,nk
In order to prove this theorem, we must first establish the formula for the
generating function for the Fibonacci-type polynomials of order k.
Lemma 5.3.3
(k)
The generating function of Fn (x), denoted by gk (s; x) is expressed by the
formula
s − s2 s
gk (s; x) = k+1
= , (5.9)
1 − (1 + x)s + xs 1 − xs − xs2 − ... − xsk
1
where |s| < 1+x
.
36
which leads us to
(k)
Fn(k) (x) = (1 + x)Fn−1 (x)
(k)
= (1 + x)2 Fn−2 (x) = ... = x(1 + x)n−2 2 ≤ n ≤ k + 1.
which leads us to
(k) (k)
Fn(k) (x) = (1 + x)Fn−1 (x) − xFn−1−k (x) n ≥ k + 2.
(k)
By induction on n, this implies Fn (x) ≤ x(1 + x)n−2 for n ≥ 2, which shows
1
the convergence of gk (s; x) for |s| < 1+x . Using equation (5.10) we find
∞
X
gk (s; x) = sn Fn(k) (x) (5.11)
n=0
k+1
X ∞
X
n n−2
=s+ s x(1 + x) + sn Fn(k) (x), (5.12)
n=2 n=k+2
37
Substituting this value into equation (5.12) we find
where the summation is over all non-negative integers n1 , ..., nk that satisfy
the condition n = n1 + n2 P
+ ... + nk .
If we replace n with n − ki=1 (i − 1)ni , i.e. n = n1 + 2n2 + ... + knk this
becomes
X∞ ∞
X X n1 + ... + nk
n (k) n
s Fn+1 (x) = s xn1 +...+nk n ≥ 0.
n1 , ..., nk
n=0 n=0 n1 ,...,nk
(k)
X n1 + ... + nk
Fn+1 (x) = xn1 +...+nk n ≥ 0,
n1 , ..., nk
n1 ,...,nk
38
where the summation is over all non-negative integers n1 , ..., nk which satisfy
the condition n = n1 + 2n2 + ... + knk .
Using the theorems stated and proved in this chapter, it is now possible
to explain the applications that the Fibonacci sequence and Fibonacci-type
polynomials of order k have with regards to probability and statistics.
39
Chapter 6
Applications
In this chapter we will cover the main applications of the Fibonacci sequence
of order k and the Fibonacci-type polynomials of order k which appear in
the probability and statistical areas of mathematics.
6.1 Arrangements
From our definition of the Fibonacci sequence of order k (5.1.1) we get
the following lemma, which shows how it can be used to find numbers of
arrangements.[MUW]
Lemma 6.1.1
(k)
If we let An denote the number of arrangements of (n + k) elements (r or
s) such that no k adjacent elements are all s except the last k. Then
(k)
A(k)
n = Fn+1 n≥0 (6.1)
To explain how this works, we will examine the simple case when k = 2 and
(2)
n = 4. So A4 is the number of arrangements of 6 elements (r or s) such
that no 2 elements are all s except the last 2. The arrangements of this form
are
rrrrss
srrrss
rsrrss
rrsrss
srsrss.
40
(2) (2)
This gives us A4 = 5 = F4+1 .
Before we can prove this lemma, it is necessary to make several definitions
to set up some notation.
Definition 6.1.2
The number of arrangements of n elements (r or s) where the last is r and
there is not a sequence of k adjacent elements that are all s is denoted by
a(k)
n (r) n ≥ 1, (6.2)
(k)
with a0 (r) = 1.
The number of arrangements of n elements (r or s) where the last is s and
there is not a sequence of k adjacent elements that are all s is denoted by
a(k)
n n ≥ 1, (6.4)
(k)
with a0 = 1.
To clarify these definitions, we will look at the simple example when k = 2
(2)
and n = 4. a4 (r) is the number of arrangements of 4 elements (r or s) where
2 s’s are not found together and the last element is r. So we have (rrrr, srrr,
(2)
rsrr, rrsr and srsr) 5 such sequences, i.e. a4 (r) = 5. Similarly we find
(2)
a4 (s) = 3 since the arrangements of this type are (rrrs, srrs and rsrs). As
(2)
the final element is unrestricted for a4 , we find it is equal to the sum of the
number of arrangements ending in r and the number of arrangements ending
(2)
in s, i.e. a4 = 5 + 3 = 8.
Also, we find
(k)
an+1 (r) = a(k)
n n ≥ 0, (6.6)
41
(k)
since an+1 (r) is the number of arrangements of n elements where no k adja-
(k)
cent elements are s with an extra r at the end. This is also true for an+1 (s)
when 0 ≤ n ≤ k − 2, however when n ≥ k − 1 we need to subtract the
(k)
number of arrangements of an which end in k − 1 s’s as these would give us
k consecutive s’s. So we find
(
(k)
(k) an 0≤n≤k−2
an+1 (s) = (k) (k) (6.7)
an − an+1−k (r) n ≥ k − 1.
(k) (k)
From the definition of An and an (r) we find that
a(k) (k)
n (r) = An n≥0 (6.9)
(k) (k) (k)
where A0 = A1 = 1, since An is the number of arrangements of n elements
r or s) where the last is r and there is not a sequence of k adjacent elements
that are all s, with an extra k s’s at the end.
This leads us to
1
n = 0, 1
(k) (k)
An = 2An−1 2≤n≤k
(k)
(k)
2An−1 − An−1−k n ≥ k + 1,
Theorem 6.2.1
42
The probability of waiting for n tosses of a coin before two consecutive heads
appear is
(2)
Fn+1 Fn+1
P [n + 2] = n+2 = n+2 n ≥ 0. (6.10)
2 2
Proof
In the case where n = 0 the probability of seeing two consecutive heads (HH)
is clearly 41 . To find a general formula for P [n + 2], it is necessary to study
the first of the (n + 2) tosses.
If the sequence begins with a head, H then a tail, T must be next so that
the sequence continues. In this case the sequence will be HT... followed by n
other terms. If however a T is the first to be flipped then there are no limits
on the second term. So the sequence will be T... followed by n + 1 other
terms.
Since these are the only possibilities for the first toss, and the probability
of H or T is 12 we find
1 1
P [n + 2] = P [n + 1] + P [n]. (6.11)
2 4
F2
In the case where n = 0, the formula holds since P [0+2] = 21 · 12 + 14 ·0 = 1
4
= 22
.
Using induction, we assume that the formula holds for n = k i.e.
1 1 Fk
P [k + 2] = P [k + 1] + P [k] = k+2
2 4 2
and look at the case where n = k + 1.
So we find
1 1
P [(k + 1) + 2] = P [(k + 1) + 1] + P [k + 1].
2 4
Using the induction hypothesis we find
1 Fk+1 1 Fk
P [(k + 1) + 2] = +
2 2k+2 4 2k+1
Fk+1 + Fk
=
2k+3
Fk+2
= k+3 .
2
So we find the formula works for n = k+1, therefore by induction the formula
holds for all n ≥ 0.
43
Similarly, the probability of waiting for n tosses before 3 consecutive heads
appear is
Tn+1
n ≥ 0. (6.12)
2n+3
Theorem 6.3.1
Proof
A typical element of the event [Nk = n + k] is an arrangement
x1 x2 · · · xk s· · · s},
| {z
k
such that n1 of the x’s are e1 = r, n2 of the x’s are e2 = sr,..., nk of the x’s
are ek = s· · · s} r and n1 + 2n2 + ... + knk = n.
| {z
k−1
Fixing n1 , ..., nk we find the number of the above arrangements is
n1 + ... + nk
,
n1 , ..., nk
44
each with the probability
If we set k = 2 in Theorem 6.3.1, in other words we want to find the proba-
bility of having to wait for n trials before we see two consecutive successes,
(6.13) reduces to the formula defined by the following corollary.
Corollary 6.3.2
When k = 2 in Theorem 6.3.1 we find that
n
2
X n−i
P [N2 = n + 2] = pi+2 (1 − p)n−i n≥0 (6.15)
i
i=0
n
2 n−i
n+2
X n−i q
=p n ≥ 0. (6.16)
i p
i=0
45
5.1.3.
Corollary 6.3.3
When p = 12 in Theorem 6.3.1 we find
(k)
Fn+1
P [Nk = n + k] = n+k n ≥ 0, (6.17)
2
which is the generalised form of the coin tossing example seen in section 6.2.
Theorem 6.3.4
The probability that we have to wait n trials before we see k consecutive
successes, with constant probability of success p is given by
n+k (k) q
P (Nk = n + k) = p Fn+1 n ≥ 0, (6.18)
p
where q = 1 − p.
Proof
q
It is clear to see that using the formula (6.14) and setting x = p
in equation
(5.8), that we get the result found in Theorem 6.3.4.
Definition 6.4.1
46
A random variable X has a geometric distribution of order k ≥ 1 with prob-
ability p, denoted Gk (x; p) if
X x1 + ... + xk q x1 +...+xk
P (X = x) = px x ≥ k, (6.19)
x1 , ..., xk p
x1 ,...,xk
where the summation is over all non-negative integers x1 , ..., xk such that
x1 + 2x2 + ... + kxk = x − k and q = 1 − p.
The distribution function of X is given by
!
1 − p[x]+1
P x 1 + ... + x k
x1 +...+xk
q
q x1 ,...,xk p
x≥k
P (X ≤ x) = x1 , ..., xk
0 otherwise
Where the summation is over all non-negative integers x1 , ..., xk such that
x1 + 2x2 + ... + kxk = x + 1.[GEN]
From this definition we can find the relation between the geometric distri-
bution of order k and the Fibonacci-type polynomials of order k, using the
random variable Nk as defined in section 6.3. The following material is based
on section 3 of [GEO].
Theorem 6.4.2
The distribution of a random variable X with geometric distribution of order
k can be expressed as
( [x]+1 (k)
1 − p q F[x]+2 ( pq ) x ≥ k
P (X ≤ x) = (6.20)
0 otherwise,
where q = 1 − p.
Before we can prove this theorem, it is necessary to establish the following
lemma.
Lemma 6.4.3
(k)
For the Fibonacci-type polynomials of order k, Fn (x),
m (k) (k)
X Fn+1 (x) Fm+k+2 (x)
=1− m ≥ 0. (6.21)
n=0
(1 + x)n+k x(1 + x)m+k
47
Proof
We can prove the lemma by using induction. For m = 0 we find both sides
(k)
of equation (6.21) equal (1 + x)−k as Fk+2 (x) = x(1 + x)k − x by equation
(5.10), so the relation holds in this case.
We now assume the formula holds for any positive integer m, i.e.
m (k) (k)
X Fn+1 (x) Fm+k+2 (x)
=1−
n=0
(1 + x)n+k x(1 + x)m+k
is true.
Then we find for n = m + 1
m+1
X Fn+1 (k) (k) m (k)
(x) Fm+2 (x) X Fn+1 (x)
= + .
n=0
(1 + x)n+k (1 + x)m+k+1 n=0 (1 + x)n+k
Using the induction hypothesis this equals
m+1 (k) (k) (k)
X Fn+1 (x) Fm+2 (x) F (x)
n+k
= m+k+1
+ 1 − m+k+2 m+k
n=0
(1 + x) (1 + x) x(1 + x)
(k) (k)
(1 + x)Fm+k+2 (x) − xFm+2 (x)
=1− .
x(1 + x)m+k+1
Using the definition of the Fibonacci-Type polynomials of order k given in
equation (5.10) gives
m+1 (k) (k)
X Fn+1 (x) Fm+k+3 (x)
n+k
=1− .
n=0
(1 + x) x(1 + x)m+k+1
So the relation holds for m + 1, therefore by induction, the relation holds for
all m ≥ 0.
48
1 z
If we let p = 1+z
and q = 1+z
this becomes
[x]−k
(k)
X Fn+1 (z)
P (X ≤ x) = .
n=0
(1 + z)n+k
p[x]+1 (k)
q
P (X ≤ x) = 1 − F[x]+2
q p
49
Theorem 6.5.1
The probability distribution of Sn is given by
pn+1 (k+1) q
P (Sn ≤ k) = F 0≤k≤n (6.24)
q n+2 p
Proof
P (Sn ≤ k) = 1 − P (Sn ≥ k + 1)
= 1 − P (Nk+1 ≤ n)
by the definition of Nk+1 found in section 6.3. Using equation (6.23) gives us
pn+1 (k+1) q
P (Sn ≤ k) = 1 − (1 − F )
q n+2 p
pn+1 (k+1) q
= F .
q n+2 p
It follows from this theorem, that if p = 12 , equation (6.24) reduces to
(k)
Fn+2
P (Sn ≥ k) = 1 − 0≤k≤n (6.25)
2n
50
the entire system to fail.
Theorem 6.6.1
Assuming the components are ordered linearly and function independently
with probability p, the reliability of a consecutive k-out-of-n: F system is
q n+1 (k) p
R(p; k, n) = F n ≥ k. (6.26)
p n+2 q
Proof
Let S̃n denote the length of the longust run of failures among n components.
Then, by the definition of R(p; k, n) we find
P (S̃n ≥ k) = 1 − P (S̃n ≤ k − 1)
q n+1 (k) p
=1− F n ≥ k,
p n+2 q
51
Chapter 7
Conclusion
52
Bibliography
[DUN] R A Dunlap, The Golden Ratio and Fibonacci Numbers, World Sci-
entific 1997.
[KOS] T Koshy, Fibonacci and Lucas Numbers with Applications, John Wi-
ley and Sons, Inc 2001.
53
[VAJ] S Vajda, Fibonacci and Lucas Numbers, and the Golden Section, Ellis
Horwood Limited 1989.
[WIK] Wikipedia,
en.wikipedia.org/wiki/The Fibonacci Association
54
Appendix A
Fundamentals
Binomial Theorem
n
X
n n
(a + b) = ar bn−r .
r
r=0
In particular,
n
X n
n
(1 + x) = xr ,
r
r=0
and ∞
−n
X n+r−1
(1 − x) = xr .
n−1
r=0
[MAR]
55
Multinomial Theorem
X n
(x1 + ... + xk ) =n
xn1 1 · · · xnk k ,
n1 , ..., nk
n1 ...,nk
where n1 + ... + nk = n.
In combinatorics, the multinomial coefficients are used for two main rea-
sons. Firstly, they are used to express the number of possible permutations
of n elements of which n1 are of one form, n2 of another form,..., and nk
are of the k th form. They can also be used to express the number of ways
of locating n different cells in which cell i contains ni elements, i = 1, ..., k
without taking the order of elements into account.[SPR]
56
Appendix B
z 3 + a2 z 2 + a1 z + a0 = 0, (B.1)
(x − λ)3 + a2 (x − λ)2 + a1 (x − λ) + a0 = 0
a22 2 2 −a3 a3 a1 a2
x3 + (−a2 + a2 )x2 + ( − a2 + a1 )x + ( 2 + 2 − + a0 ) = 0
3 3 27 9 3
a22 a1 a2 2
x3 + (a1 − )x − ( − a32 − a0 ) = 0
3 3 27
2
3a1 − a2 9a1 a2 − 27a0 − 2a32
x3 + 3 · x−2· = 0.
9 27
If we define
3a1 − a22 9a1 a2 − 27a0 − 2a32
p= and q= ,
3 54
then (B.1) can be written in the standard form
x3 + px = q. (B.2)
57
p
Making Vieta’s substitution, x = w − 3w
, reduces the cubic to
p3
w3 − − q = 0.
27w3
This can be turned in to a quadratic equation in w3 by multiplying through
by w3 , to get
1
(w3 )2 − q(w3 ) − p3 = 0.
27
Solving this using the quadratic formula we find
r
3 1 4
w = (q ± q 2 + p3 ) (B.3)
2 27
r
1 1 2 1
= q± q + p3 (B.4)
2 4 27
p
= R ± R2 + Q3 . (B.5)
x3 − B 3 = (x − B)(x2 + Bx + B 2 ). (B.7)
58
Matching the coefficients C and −(B 3 + BC) with those of (B.6) we find
C = 3Q and B 3 + BC = 2R,
which leads us to
B 3 + 3QB = 2R. (B.9)
Therefore, if we can find a value of B satisfying the above identity we have
factored a linear term from the cubic equation, reducing it to a quadratic
equation.
The solution accomplishing this is
p 1 p 1
B = [R + Q3 + R2 ] 3 + [R − Q3 + R2 ] 3 .
x2 + Bx + (B 2 + 3Q) = 0.
59
This can be simplified by defining
p 1 p 1
A = [R + Q3 + R2 ] 3 − [R − Q3 + R2 ] 3
p 2 1 p 2
A2 = [R + Q3 + R2 ] 3 − 2[R2 − (Q3 + R2 )] 3 + [R − Q3 + R2 ] 3
p 2 p 2
= [R + Q3 + R2 ] 3 + [R − Q3 + R2 ] 3 + 2Q
= B 2 + 4Q.
we find
A=S−T and B = S + T.
The roots of the cubic equation in z are therefore
1
z1 = − a2 + (S + T )
3
1 1 1 √
z2 = − a2 − (S + T ) + i 3(S − T )
3 2 2
1 1 1 √
z3 = − a2 − ((S + T ) − i 3(S − T ).
3 2 2
60
This piece of work is a result of my own work except where it forms an
assessment based on group project work. In the case of a group project, the
work has been prepared in collaboration with other members of the group.
Material from the work of others not involved in the project has been
acknowledged and quotations and paraphrases suitably indicated.
61