A BLAISDELL SCIENTIFIC PAPERBACK

N. N.Vorob’ev
POPULAR LECTURES IN MATHEMATICS SERIES

E ditors: I. N. Sneddon and M. Stark

Volume 2

FIBONACCI NUMBERS
TITLES IN THE POPULAR LECTURES IN MATHEMATICS SERIES

Vol. 1 The Method o f Mathematical Induction

By I. S. Sominskii
Vol. 2 Fibonacci Numbers
By N. N. Vorob'ev
Vol. 3 Some Applications o f Mechanics to Mathematics
By V. A. Uspenskd
Vol. 4 Geometrical Constructions using Compasses Only
By A. N. Kostovskii
Vol. 5 The Ruler in Geometrical Constructions
By A. S. Smogorzhevskii
Vol. 6 Inequalities
By P. P. K orovkin
 FIBONACCI NUMBERS

by
N. N. VOROB’EV

Translated from the Russian by

HALINA MOSS, B.Sc.

Translation Editor
IAN N. SNEDDON
Simson Professor of Mathematics
in the University of Glasgow

BLAISDELL PUBLISHING COMPANY

NEW YORK LONDON
A DI VI S I ON OF R A N D O M HOUS E
SOLE D I S T R I B U T O R S IN T HE U N I TE D STATES A N D C A N A D A
Blaisdell Publishing Company
22 East 51st Street, New York 22, N. Y.

Copyright © 1961
Pergamon Press L td.

A translation of the original volume

Chisla fibonachchi
(Moscow-Leningrad, Gostekhteoretizdat, 1951)

Library of Congress Card Number: 61-11536

Printed in Great Britain by Pergamon Printing and Art Services Limited, London
 CONTENTS

Foreword • • • • • • • a vii

Introduction . . . . . . . . I

I. The simplest properties of Fibonacci numbers . 6

II . Number-theoretic properties of Fibonacci num­
bers . . . . . . . . 25
II I . Fibonacci numbers and continued fractions . 36
IV. Fibonacci numbers and geometry . . . 55
V. Conclusion . . . . . . . 65

v
 Foreword

In elementary mathematics there are many d iff ic u lt and

in te re stin g problems not connected with the name of an
individual, but rather possessing the character of a kind
of "mathematical folklore". Such problems are scattered
throughout the wide lite ra tu re of popular (or, simply,
entertaining!) mathematics, and often i t is very d if­
f ic u lt to establish the source of a p a rticu lar problem.

These problems often circ u late in several versions.

Sometimes several such problems combine into a single,
more complex, one, sometimes the opposite happens and one
problem s p lits up into several simple ones: thus i t is
often d iff ic u lt to distinguish between the end of one
problem and the beginning of another. We should consider
th at in each of these problems we are dealing with l i t t l e
mathematical theories, each with i t s own history, i t s own
complex of problems and i t s own c h arac te ristic methods,
a ll, however, closely connected with the history and
methods of "great mathematics".

The theory of Fibonacci numbers is ju s t such a theory.

Derived from the famous “rabbit problem” , going back
nearly 750 years, Fibonacci numbers, even now, provide
one of the most fascinating chapters of elementary mathe­
matics. Problems connected with Fibonacci numbers occur
in many popular books on mathematics, are discussed at
meetings of school mathematical societies, and feature in
mathematical competitions.

The present booklet contains a set of problems which

were the themes of several meetings of the schoolchil­
dren’ s mathematical club of Leningrad State University in

vi 1
 Foreword

the academic year 1949-50. In accordance with the wishes

of those taking part, the questions discussed a t these
meetings were mostly number-theoretical, a theme which is
developed in greater detail here.

This book is designed to appeal basically to pupils of

16 or 17 years of age in a high school. The concept o a
lim it is met with only in examples 7 and 8 in chapter
II I. The reader who is not acquainted with th is concept
can omit these without prejudice to his undejstanding of
what follows. That applies also to binomial co efficien ts
(I, example 8) and to trigonometry (IV, examples 2 & 3).
The elements which are presented of the theory of d iv is i­
b ility and of the theory of continued fractions do not
presuppose any knowledge beyond the lim its of a school
course.

Those readers who develop an in te re st in the principle

of constructing recurrent series are recommended to read
the small but fu ll booklet of A.I. Markushevich, "Re­
current Sequences" (Vozvratnyye posledovatel’ nosti)
(Gostekhizdat, 1950). Those who become interested in
facts re la tin g to the theory of numbers are referred to
textbooks in th is subject*.

• Ehglish-speaking readers are referred to

H. Davenport, "The Higher Arithmetic" (London, Hutchinson,
1952);
Burton W. Jones, "The Theory of Numbers” (London, Constable,
1955).

V ll]
 INTRODUCTION

_1_. The ancient world was rich in outstanding mathema­

tic ia n s. Many achievements of ancient mathematics are
admired to th is day for the acuteness of mind of th e ir
authors, and the names of Euclid, Archimedes and Hero are
known to every educated person.

Things are d ifferen t as far as the mathematics of the Middle

Ages is concerned. Apart from Vieta, who lived as la te as
the sixteenth century, and mathematicians closer in time
to us, a school course of mathematics does not mention a
single name connected with the Middle Ages. This is, of
course, no accident. In th at epoch the science developed
extremely slowly, and mathematicians of real statu re were
few.

The greater then is the in te re st of the work L i b e r

Abacci ("a book about the abacus” ), w ritten by the
remarkable Ita lia n mathematician, Leonardo of Pisa, who
is b etter known by his nickname Fibonacci (an abbrevia­
tion of f i H u t B o n a c c i ) . This book, w ritten in 1202, has
survived in i t s second version, belonging to 1228.

is a voluminous work, containing nearly

L i b e r Abacci
a ll the arithm etical and algebraic knowledge of those
times. I t played a notable part in the development of
mathematics in Western Europe in subsequent centuries.
In p articu lar, i t was from th is book th at Europeans be­
came acquainted with the Hindu (Arabic) numerals.

The theory contained in L i b e r Abacci is illu s tra te d by

1
2 Fibonacci numbers

a great many examples, which make up a sig n ific a n t part

of the book.

Let us consider one of these examples, that which can

be found on pages 123-124 of the manuscript of 1228:

"How many pairs of rabbits are born of one p air in a y e a r? ”

This problem is stated in the form:-

"Someone placed a p air of rabbits in a certain place,

enclosed on all sides by a wall, to find out how many
pairs of rabbits w ill be born there in the course of one
year, i t being assumed that every month a p air of rabbits
produces another pair, and th at rabbits begin to bear
young two months a fte r th e ir own birth.

"As the f i r s t pair produces issue in the f i r s t month,

in th is month there w ill be 2 pairs. Of these, one pair,
namely the f i r s t one, gives birth in the following month,
so that in the second month there w ill be 3 pairs. Of
these, 2 pairs w ill produce issue in the following month,
so that in the third month 2 more pairs of rabbits will
be born, and the number of pairs of rabbits in that month
w ill reach 5; of which 3 pairs w ill produce issue in
the fourth month, so th at the number of pairs of rabbits
will then reach 8. Of these, 5 pairs will produce a
further 5 pairs, which, added to the 8 pairs, w ill give 13
pairs in the fifth month. Of these, 5 pairs do not
produce issue in that month but the other 8 do, so that
in the sixth month 21 pairs resu lt. Adding the 13 pairs
that will be born in the seventh month, 34 pairs are
obtained: added to the 21 pairs born in the eighth month
i t becomes 55 pairs in that month: th is, added to the 34
pairs born in the ninth month, becomes 89 pairs: and in ­
creased again by 55 pairs which are born in the tenth
month, makes 144 pairs in that month. Adding the 89
further pairs which are born in the eleventh month, we
get 233 pairs, to which we add, la stly , the 144 pairs
born in the final month. We thus obtain 377 pairs: th is
 Introduction 3

is the number of p airs procreated from the f i r s t p air by

the end of one year.

A p a ir
1
F irs t (Month)
2
Second
3
Third
5
Fourth
8
Fifth
13
Sixth
21
Seventh
34
Eighth
55
Ninth
89
Tenth
144
Eleventh
233
Twelfth
377

Fig. 1.

"From [Fig. l) * we see how we arriv e at it: we add to

the f i r s t number the second one i.e . 1 and 2; the second
one to the third; the th ird to the fourth; the fourth
to the fifth ; and in th is way, one a fte r another, until
we add together the tenth and the eleventh numbers (i.e .

Fibonacci does a ll calculation tables and diagrams in the

margin.
4 Fibonacci numbers

144 and 233) and obtain the total number of rabbits (i.e .
377); and i t is possible to do th is in th is order for an
in fin ite number of months”.

2. We now pass from rabbits to numbers and examine the

following numerical sequence

n* (1)

in which each term equals the sum of two preceding terms,

i. e. for any n ^ 2

un un + un- 2* ( 2)

Such sequences, in which each term is defined as some

function of the previous ones, are met with often in
mathematics, and are called recurrent sequences. The
process of successive definition of the elements of such
sequences is i t s e l f called the recurrence process, and
equation (2) is called a recurrence relation. The reader
can find the elements of the general theory of recurrent
sequences in the book by Markushevich mentioned above.

We note that we cannot calculate the terms of sequence

(1) by condition (2) above.

I t is possible to make up any number of d ifferen t

numerical sequences satisfying th is condition. For
exampl e

2, 5. 7. 12, 19, 31. 50.........

1. 3. 4. 7, 11. 18, 29.........
-1, -5, -6, -11, -17, . . . and so on.

This means th at for the unique construction of sequence

(1) the condition (2) is obviously inadequate, and we
must establish certain supplementary conditions. For
example, we can fix the f ir s t few terms of sequence (1).
 Introduction 5

How many of the f i r s t terms of sequence (1) must we fix

so that i t is possible to calculate a ll i t s following
terms, using only condition (2)?

We begin by pointing out that not every term of se ­

quence (1) can be obtained by (2) i f only because not all
terms of (1) have two preceding ones; for instance, the
f i r s t term of the sequence has no terms preceding i t , and
the second term is preceded by only one. This means that
in addition to condition (2) we must know the f i r s t two
terms of the sequence in order to define i t .

This is obviously su fficien t to enable us to calculate

any term of sequence (1). Indeed, ug can be calculated
as the sum of the prescribed Uj and u2; u4 as the sum of
u and the previously calculated ug;
2 as the sum of
the previously calculated ug and u4 and so on "in th is
order up to an in fin ite number of terms”.

Passing thus from two neighbouring terms to the one

immediately following them, we can reach the term with
any required suffix and calculate it.

3_. Let us now turn to the important p a rtic u la r case of

sequence (1), where = 1 and u2 = 1. As was pointed
out above, condition (2) enables us to calculate succes­
sively the terms of th is series. I t is easy to verify
that in th is case the f i r s t 13 terms are the numbers

1, 1, 2, 3. 5, 8. 13, 21. 34, 55. 89. 144, 233, 377,

which we already met in the rabbit problem. To honour

the author of the problem, sequence (1) when Uj - u - l
is called the Fibonacci sequence, and i t s terms are known
as Fibonacci numbers.

Fibonacci numbers possess a number of in te re stin g and

important properties, which are the subject of th is whole
booklet.
8 Fibonacci numbers

Now, le t us add u2n+l to b°th sides of (6)

U1 “ u 2 + u 3 - u 4 +

Combining (6) and (7) we get for the sum of Fibonacci

numbers with altern atin g signs:

U1 " u 2 + u 3 ~ u 4 + + ( - l ) n+\
( 8)
(- i ) n+1un_i + i

4. The formulae (3) and (4) were deduced by means of the

term by term addition of a whole series of obvious equa­
tions. A further example of the application of th is pro­
cedure is the proof of the formula for the sum of squares
of the f i r s t n Fibonacci numbers

(9)

We note that

ufc-iu* = V ufe+i - “*-!> - ul

Adding up the equations

= u,u
12
 Simplest properties 9

term by term, we obtain (9).

5. Many relationships between Fibonacci numbers are

conveniently proved with the aid of the method of induc­
tion.

The essence of the method of induction is as follows. In

order to prove that a certain proposition is correct for any
natural number i t is su ffic ie n t to establish:

(a) th at i t holds for the number 1;

(b) th at from the tru th of the proposition for an arb itra ry

natural number n follows i t s tru th for the number n + 1.

Any inductive proof of a proposition tru e for any natural

number consists, therefore, of two parts.

In the f i r s t part the tru th of the proposition being proved

is established for n = 1. The tru th of the proposition for
n = 1 is sometimes called the basis o f in d u c tio n .

In the second part of the proof the tru th of the proposition

is assumed for a certain a rb itra ry (but fixed) number n, and
from th is assumption, often called the inductive assumption,
the deduction is made that the proposition is also true for the
number n + l. The second part of the proof is called the induc­
t i v e t ransi tion.

The detailed presentation of the method of induction and

numerous examples of the application of d iffe re n t forms of th is
method can be found in I.S. Sominskii, "The Method of Mathe­
matical Induction” . • Thus, in p a rticu la r, the version of the
method of induction with the inductive tra n sitio n “from n and
n t 1 to n + 2" employed by us below is given in Sominskii’ s
book on page 9* and is illu s tra te d there on page 10* by problems
18 and 19.

English edition, Pergamon Press, 1961.

10 Fibonacci numbers

We prove by induction the following important formula:

un+» u ,u
n-l n
+ u u ,.
n n+1 ( 10 )

We shall carry out the proof of th is formula by induc­

tion on m. For m = 1 th is formula takes the form

n+1
un - l,u.1 + u u n_ 2= un - l, + u ,
T n’

which is obviously true. For m = 2 formula (10) is also

true, because

u‘n + 2 = Un - l U2 + Un

Thus the basis of the induction is proved. The induc­

tiv e tra n sitio n can be proved in th is form: supposing
formula (10) to be true for n = k and for m = k + 1, we
shall prove that i t also holds when m = k + 2 .

Thus, le t

un+k. = u n - l,u,k + u u, ,
^ nfc+1

and

un+fc, + 1, = un - l,u,«+ ,1 + u u.
T n fc+2

Adding the la s t two equations term by term we obtain

lin +4 + 2 = Un - l U* + 2 + Un Uk +V

and th is was the required resu lt.

Putting » = n in formula (10) we obtain

 Simplest properties 11

or

T un +,)
2n - un '(u n-1, + 1' . ( 11 )

From th is la s t equation i t is obvious th at u2n is d iv is­

ible by un. In the next chapter we shall prove a much
more general re su lt.

Since

u n = u n +, 1 - u n-1,,

formula (11) can be rew ritten thus:

U2n = (Un+1 - Un-l>(Un+l + Un-1>«

or

u 2, n = n+1n-1'

i . e . , the difference of the squares of two Fibonacci

numbers whose positions in the sequence d iffe r by two is
again a Fibonacci number.

Sim ilarly (taking m = 2n) i t can be shown, that

u 3n = u 3n + 1 + u n3 _ u n-1
3

6. The following formula will be found useful in what

follows:

2 un un+2„ + (-1)". ( 12)

+1

Let us prove i t by induction over n. For n - 1 , (12)

takes the form
12 Fibonacci numbers

U2 = U 1U3 - 1’

which is obvious.

We now suppose formula (12) proved for a certain n.

Adding un+iun+2 s *^es i t we obtain

2
u"+1 Un+lUn+2 = UnUn+2 + Ur,+lUn+2 + (-1)"

or

or

un+1,u n+3_ n+2 + (-1)".

or

“.’ ♦J = * t- 1*"*1

Thus, the inductive tran sitio n is established and formula

(12) is proved for any n.

7. In a sim ilar way, i t is possible to establish the

following properties of Fibonacci numbers:

U1U2 + U2U3 + U3U4 + • • • + U2 a - l U2n = U2n-

V 2 + U2U3 + U3U4 + + a 2nU2n+l = U2n+1 " 1 '

nux + (n - l) u 2 + (n - 2)u3 + . . . + 2ufi. 1 + =

= Un+4 - (n + 3,1

The proofs are le f t to the reader.

 Simplest properties 13

8. I t turns out that there is a connection between the

Fibonacci numbers and another set of remarkable numbers - the
binomial coefficients. Let us se t out the binomial c o e ffic i­
ents* in the following tria n g le , called Pascal’ s triangle:

c °o
c° c\
pO p i p2
z °2
3 ci c i <1

i. e. 1
1 1
1

w 1
/V s 4
1 5 IP 10 5 1
r 6 15 20 15 6 1

The stra ig h t lines drawn through the numbers of this triangle

a t an angle of 45 degrees to the rows we sh a ll c a ll “the risin g
diagonals” of Pascal’ s trian g le. For instance, the stra ig h t
lines passing through numbers 1, 4, 3, or 1, 5, 6, 1, are
risin g diagonals.

We shall show that the sum of numbers lying along a certain

risin g diagonal is a Fibonacci number.

Indeed, the f i r s t and topmost risin g diagonal of Pascal’ s

triangle is merely 1, the f ir s t Fibonacci number. The second
diagonal also consists of 1. To prove the general proposition,

* Expressions of the form C®, as used here and below in con­

formity with the o riginal, represent the DC„a of customary
English usage.
14 Fibonacci numbers

i t is su ffic ie n t to show that the sum of a ll numbers making up

the (n-2)th and the (n -l)th diagonal of Pascal’ s trian g le is
equal to the sum of the numbers making up the nth diagonal.

On the (n-2)th diagonal we have the numbers

n 0 nl r2
Un-3’ n-4' n -6’

and on the (n -l)th diagonal the numbers

C° „ C1 .. C2
n-2 n-3 n-4

The sum of a ll these numbers can be w ritten thus

^h-2 + ((^ - 3 + Cn-3} + (Cn-4 + Cn-4} + ” ■ (13)

But for binomial coefficients

C°
n -2 = f°
Ln-1 = 1

and

t l _ k(k - 1) . . . (k - i + 1)
1.2........... i
k(k - 1) . . . (k - i + 1)(* - i)
+ ------------------------------------------
1.2........... i . ( i + 1 )

k(k - 1) . . . (k - i + 1) / * - i\
1.2................ i \ i + 1/

_ k( k - 1) . . . (fc - i + 1) i + 1+ - i
1.2........... i i + 1

(k + 1)*(* - 1) . . . (k - i + 1)
- r'i+l
1.2........... i . ( i + 1) - V i-
 Simplest properties 15

Expression (13) therefore equals

Cn-1 + Cn-2 + Cn-3 +

i.e . the sum of the numbers lying on the nth diagonal of the
triangle.

Prom th is proof and formula (3) we immediately get: The sum

of a ll binomial coefficients lying above the nth risin g diagon­
al of Pascal’ s triangle (inclusive of that diagonal) equals
un+2 ‘ U

Making UBe of formulae (4). (5), (6) and sim ilar ones, the
reader can easily obtain further id e n titie s connecting
Fibonacci numbers with binomial coefficients.

9. So far, we have defined Fibonacci numbers by a

recurrence procedure, i.e . inductively, by th e ir su f­
fixes. I t turns out, however, th at any Fibonacci number
can also be defined d irectly , as a function of its su f­
fix.

To see th is, we investigate various sequences satisfy in g

the relationship (2). We shall call a ll such sequences
solutions of equation (2).

In future we shall denote the sequences

1' S' S' .• •f

1' S' S' •••t
1' S' vy • ••t
by V, V' and V" respectively.

To begin with we shall prove two simple lemmas.

16 Fibonacci numbers

Lemma 1. I f V is the solution o f equation (2) and c is

an arbitrary number, then' the sequence cV ( i . e . the
sequence cv., cv^, cv g, . . . ) is also a solution o f
equation (2).

Proof. Multiplying the relationship

= Vn-2 + Vn- 1

term by term by c, we get

cv n - CVn-2 +CVn - r

as was required.

Lemma 2. I f the sequences V and V* are solutions o f (2),

then their sum V' + V" ( i . e . the sequence + v'',
v2 + v 2‘ v 3 + v 3' ** also a solution o f (2).

Proof: Prom the conditions stated in the lemma we have

v n' = V'n-1. T
+ V'n-2„

and

Vn = V l + Vn - 2-

Adding these two equations term by term, we get

K + K = <K-1 + Vn-i> + (Vn-2 + Vn-2) *

Thus, the lemma is proved.

Now, le t V' and V" be two solutions of equation (2)

which are not proportional. We shall show that any se­
quence V which is a solution of equation (2) can be
written in the form
 Simplest properties 17

where Cj and c2 are constants. I t is therefore usual to

speak of (14) as the general solution of the equation
(2).

F irs t of a ll, we shall prove th a t i f solutions of (2)

V' and V" are not proportional, then

The proof of (15) is carried out by assuming the opposite.

For solutions V' and V" of (2) which are not propor­
tio n al, le t

(16)
vl 2

On w riting down the derived proportion we get

V1 + v 2 _ v2
+ V
2

or, taking into account th at V' and V" are solutions of

equation (2),

Sim ilarly, we convince ourselves (by induction!) that

Thus, i t follows from (16) th at the sequences V' and V"

are proportional, which contradicts the assumption. This
means th a t (15) is true.
18 Fibonacci numbers

Now, le t us take a certain sequence V, which is a solu­

tion of the equation (2). This sequence, as was pointed
out in section 2 of the Introduction, is fully defined if
its two f i r s t terms, and v2, are given.

Let us find such Cj and c 2, that

II
+ C V*

H-*
c i v i 2V1
(17)
+ = v 2
C 1V 2 C 2V 2

Then, on the basis of lemmas 1 and 2, c^V' + c ^ / " gives

us the sequence V.

In view of condition (15), the simultaneous equations

(17) are soluble with respect to Cj and c 2 no matter what
the numbers and v 2 are:

c = v l v 2 ~ V2V1 c = vl v2 ~ V2V1
1 v { v 2 ~ v "\v 2 ’ 2 u l v 2 ~ v l v2

[By the condition (15) the denominator does not equal

zero].

Substituting the values of Cj and c 2 thus calculated in

(14) we obtain the required representation of the se­
quence V.

This means that in order to describe a ll solutions of

equation (2) i t is su ffic ie n t to find any two solutions
of i t which are not proportional.

Let us look for these solutions among geometric pro­

gressions. In accordance with lemma _1^ i t is su ffic ie n t
to lim it ourselves to the consideration of only those
progressions whose f i r s t term is equal to unity. Thus,
le t us take the progression

1, <}, q2, •••

 Simplest properties 19

In order that th is progression should be a solution of

(2) i t is necessary that for any n the equality

should be fu lfille d . Or, dividing by qn~2,

1 + q - q2.
1 + VsT
The roots of th is quadratic equation, i.e . — ~ and
i -V s
— -----, w ill be the required common ra tio s of the pro­
2

gressions. We shall denote them by a and j8 respectively.

Note that a j3 = -1.

We have thus obtained two geometric progressions which

are solutions of (2). Therefore a ll sequences of the
form

C1 + c2' Cl a + C2 P -c la2+ c2fi2> ••• (18)

are solutions of (2). As the progressions found by us
have d ifferen t common ra tio s and are therefore not propor­
tional. formula (18) gives us all solutions of equation
( 2) .

In p articular, for certain values of and c2 formula

(18) should give us also the Fibonacci series. For this,
as was pointed out above, i t is necessary to find c 1 and
c 2 from the equations

and

cl a + c20= u2’
i.e . from the simultaneous equations

C1 + c 2 “
20 Fibonacci numbers

Having solved them, we get

1 +VT 1 -V 5

whence

i + ^ /w ( i + ^ n-1 i - V 5 / 1 - v r \n-1
2 V5 V 2 / 2 V5 \ 2 J
i«e

1 -V 5 \n
un (19)
vT

Formula (19) is called B i n e t ’s formula in honour of the

mathematician who f i r s t proved i t . Obviously, sim ilar
formulae can be derived for other solutions of (2). Hie
reader should do i t for the sequences introduced in
section 2 of the Introduction.

10. With the help of Binet’s formula i t is easy to find the

sums of many series connected with Fibonacci numbers.

For instance, we can find the sum

We have

a3 - p 3
a 6 - p/>6 a3n - p/}3n
a

- /--- ^ -- + • • • + ---
30 w v r v ir
 Simplest properties 21

F^=r(a3 + a 6 + ... + a 30 - £ 3 - jg6 - ... - /9 *).

or, haring summed the geometric progressions involved,

_ J _ (a** - a 3 g * * 3 - P 3\
U3 +U0 + --- + “ 3n = V T a3- 1 j53 - 1
But
a 3 - 1 = cr + a 2 - 1 = a + a + 1 - 1 = 2a,
and sim ilarly j3 3 - 1 = j .
2 3 Therefore

_ 1 / q ^ 3 - a 3 f l3" * 3 - £ 3'
U3 +U6 + - " +U3 n = V i-! 2a 2/3

or a fte r cancellations

3n t 2 _ a 2 _^fj3n+2 + o 2 '
U3 + U6 + ‘ ‘ + U 3 n = v f

1 g 2 - j82,
2 \ vH T -v^T

1 “ lnj.9 - 1
=7 (U3"+2 " U2}

11. As another example of the application of Binet’ s formula,

we shall calculate the sum of the cubes of the f ir s t n Fibonacci
numbers.

We note that

'«» -y3fc> ,3k 3 a 2* / 9 fc + 3 a kp 2k

>2
, V IT, 5 VT
22 Fibonacci numbers

Therefore:

or, using formula (8 ) and the resu lts of the preceding section,

- 1
+ 3 [1 + ( - l ) r,+1un_1]j =

u 3n+2 + + 5
10

12. I t is relevant to ask the question: how quickly do

Fibonacci numbers grow with increasing suffix? Binet’ s
formula gives us a su ffic ie n tly fu ll answer even to th is
question.

I t is not hard to prove the following theorem.

Theorem:
----- ■-------
The Fibonacci number un is the nearest whole
number to the nth term a o f the geometric progress ion

whose f i r s t term is a and whose common ratio equals a

Proof: Obviously i t is su ffic ie n t to establish that the

absolute value of the difference between un and an12 is
1
always less than — . But
2
 Simplest properties 23

on n n
_p a a a" - r 1/31"
v ir vt VT V5

As ^3 = -0 .6 1 8 ..., therefore |^ S |< 0 , and that means

th at for any n, | " <C1 and even more so (since v '5 ’ > 2)

—— \ — . The theorem is proved.

Yf 2
The reader who is acquainted with the theory of lim its
will be able to show by slig h tly a lte rin g the proof of
th is theorem that

lim un an = 0.
n —» > o o I

Using th is theorem i t is possible to calculate Fibonacci

numbers by means of logarithmic tables.

For instance, le t us calculate uJ4 (u14 is the answer

to the problem of Fibonacci about the rabbits):

V F = 2.2361. log V T = 0.34949;

1 +vT
a = ----------- =1.6180, log a = 0.20898;
2
14
log —---- = 14 X 0.20898 - 0.34949 = 2.5762,
Y!

376.9.
sTh
The nearest whole number to 376.9 is 377; th is is uJ4

When calculating Fibonacci numbers of very large su f­

fixes, we can no longer calculate a ll the figures of the
number by means of available tables of logarithms; we
can only indicate the f i r s t few figures of i t , so th at
the calculation turns out to be approximate.

As an exercise, the reader should prove that in the

24 Fibonacci numbers

decimal system, un for n ^ 17 has no more than — and no

« 4
fewer than — figures. And of how many figures will u 1Q(
consist?
 II
NUMBER-THEORETIC PROPERTIES OP
FIBONACCI NUMBERS

Before we continue the study of Fibonacci numbers, we

shall remind the reader of some of the sim plest facts
from the theory of numbers.

1^ F irst, we shall indicate the process of finding the

greatest common divisor of numbers a and 6.

Suppose we divide a by 6 with a quotient equal to qQ

and a remainder r-j. Obviously, a = bqQ + r j and
0 < 6 . Note that i f a < 6 , qQ = 0.

Let us further divide 6 by r 1 and le t us denote the

quotient by qx and the remainder by r^. Obviously
b = r i<li •+■r 2 < and 0 ^ r 2 < r j . Since Tj <b, therefore
q j / 0. Then, dividing Tj by r 2, we shall find q2 ^ 0
and r 3 such that r i = 92, 2 '+' r 3 and 0 ^r 3<" r 2‘ We
proceed in th is manner for as long as i t is possible to
continue the process.

Sooner or la te r our process must terminate, since all

the positive whole numbers r 1( r 2. r 3, . . . are d ifferen t,
and every one of them is smaller than 6. That means that
th e ir number does not exceed b, and the process should
terminate no la te r than at the 6th step. But i t can only
terminate when a certain division proves to be carried
out perfectly, i.e . the remainder turns out equal to
zero and i t will be impossible to divide anything by it.
The process thus described bears the name of Euclidean

25
26 Fibonacci numbers

Algorithm. As a result of its application we obtain the

following sequence of equations

a = bqQ + r y N

b = r l9l + r 2.

r l = r 2^2 + r 3-
r ( 20 )

rn-2 = rn -l9n-l + rn -

rn-l = rn % ‘

Let us examine the la s t non-zero remainder r n . Ob-

viously r n —.l is divisible by r n . Let us now take the
la s t but one equation in (20). On i t s right-hand side
both terms are div isib le by r n and therefore r n- 2„ is
d iv isib le by r . Similarly, we show step by step
(induction!) th at r n ■u_, r n* i. . . . . and fin a lly a and 6 are
d iv isib le by r h . Thus, r n is a common divisor of a and
b. Let us show th at r ft is the greatest common divisor of
a and b. In order to do th is, i t is su ffic ie n t to show
that any common divisor of a and b will also divide r .

Let d be a certain common divisor of a and b. From the

f i r s t equation of (20) we notice th at should be d i­
v isib le by d. But, in that case, on the basis of the
second equation of (20), r 2 is d iv isib le by d. Similarly
(induction!) we prove that d “goes into” r 3..........rn-1
and, fin a lly , r .

We have thus proved that the Euclidean algorithm when

applied to the natural numbers a and b does lead really
to th e ir greatest common divisor, This greatest common
divisor of the numbers a and b is denoted by (u, 6).

As an example, le t us find (u 20’ (115) = (67G5> 610):

 Number-theoretic properties 27

6765 =■ 610 X 11 + 55,

610 = 55 X 11 + 5,

55 = 5X1 1

Thus, ui ^ = 5 = Ug. The

5 fa c t th at the greatest
common divisor of two Fibonacci numbers turned out to be
again a Fibonacci number is not accidental. I t will be
shown la te r th a t that is always the case.

2. There is an analogy between Euclid’ s algorithm and a pro-

oess in geometry whereby the common measure of two commensur­
able segments is found.

Indeed, le t us examine two segments, one of length o, the

other of length b. Let us subtract the second segment from the
f ir s t as many times as i t is possible ( if b > a, obviously we
cannot do i t even once) and denote the length of the remainder
by r j. Obviously < b. Now, le t us subtract from the seg­
ment of length b the segment of length as many times as
possible, and le t us denote the newly obtained remainder by rg.
Carrying on in th is manner, we obtain a sequence of remainders
whose lengths, evidently, decrease. Up to th is point, the
resemblance to B ic lid 's algorithm is complete.

Subsequently, however, an important difference of the geo­

m etrical process from Euclid's algorithm for natural numbers is
revealed. The sequence of remainders obtained from the sub­
traction of segments might not terminate, as the process of
such subtraction can turn out to be capable of being continued
indefinitely. This w ill happen i f the chosen segments are in­
commensurable.

From the considerations in section 1, i t follows that two

segments whose lengths can be expressed by whole numbers are
always commensurable.

We now e s ta b lis h sev eral sim ple p ro p e rtie s of the

28 Fibonacci nunbers

greatest common divisor of two numbers.

_3_. (a, be) is div isib le by (a, 6). Indeed, 6, and

therefore 6c, is div isib le by (a, 6); a is d iv isib le by
(a, b) for obvious reasons. This means, according to the
proofs in section 1, that (a, 6c) is d iv isib le by (a, 6) also.

4. (ac, be) = (a, 6)c

Proof: Let the equations (20) describe the process of

finding (a, 6). Multiplying each of these equations by c
throughout, we shall, as is easily verified, obtain a set
of equations corresponding to the Euclidean algorithm as
applied to the numbers ac and 6c. The la st non-zero re ­
mainder here will be equal to rnc, i.e . (a, 6)c.

5. I f (a, c) = 1, then (a, 6c) = (a, 6). Indeed, (a, 6c)

divides (ab, be), according to section 3. But

(ab, be) = (a, c) 6 = 1 X6 — 6

in view of section 4. Thus 6 is d iv isible by (a, be).

On the other hand (a, be) divides a. By section 1 th is
means that (a, 6c) divides (a, 6) also. And since accord­
ing to section 3 (a, 6) divides (a, 6c) as well, then
(a, 6) = (a, 6c).

6^ a is divisible by 6 only i f (a, 6) = 6. This is

obvious.

7_. If c is divisible by 6, then (a, 6) = (a + c, 6).

Proof: Suppose that the application of the Euclidean

algorithm to the numbers a and 6 leads to the se t of
equations (20). Let us apply the algorithm to the
 Number-theoretic properties 29

numbers a + c and 6. Since c is d iv isib le by 6, as given,

we can put c = c^b. The f i r s t step of the algorithm
gives us the equation

a + c = (qQ + c i)6 + r r

The subsequent steps of th is algorithm w ill give us

consecutively the second, third, e t c ., equations of the
set (20). The la s t non-zero remainder is s t i l l r , and
th is means th at (a, 6) = (a + c, b).

A useful exercise for the reader would be to prove th is

theorem on the sole basis of the re su lts of sections 3-6,
i.e . without a repeated reference to the idea of the
Euclidean algorithm and to the set (20).

We now consider certain properties of Fibonacci numbers

concerning th e ir d iv is ib ility .

8. Theorem: I f n is d i v i s i b l e by m, then un is also d i ­

v i si b le byJ u_.
n
Proof: Let n be d iv isib le by m, i.e . le t n = We
sh all carry out the proof by induction over m^. For
= 1, n - m, so th at in th is case i t is obvious th a t un
is d iv isib le by un. We now suppose th a t u„ni is d iv is i­
ble by u„ and consider um(it +1). But um(m1+i) = unmi + „
and, according to (10), 1

u m .( « j + tl )x = u m + u mm j u m+1
,,.

The f i r s t term of the right-hand side of th is equation is

obviously d iv isib le by um. The second term contains
as a factor, i.e . is d iv isib le by um according to the in-
ductive assumption, Hence, th e ir sum, i.e . um,(mj+,i ) is
d iv isib le by u as well. Thus the theorem is proved.

9. The topic of the arithm etical nature of Fibonacci

30 Fibonacci numbers

numbers (i.e . the nature of th e ir divisors) is of great

in terest.

We prove th at for a compound n other than 4, un is a

compound number.

Indeed, for such an n we can w rite n = n jn 2. where

1 < nj < n, 1 < rc2 < n and e ith e r rij > 2 or n 2 > 2. To
be definite le t > 2. Then, according to the theorem
ju s t proved, u„ is div isib le by un i, while 1 < uni < un
and th is means that u„ is a compound number.

10. Theorem. Neighbouring Fibonacci numbers are prime

to each other.

Proof: Let u and u , have a certain common divisor

- ------ n "+1
a > 1, in contradiction of what the theorem sta te s. Thi

C T3
th e ir difference un+1 - should be d iv isib le by d. Ai
since un - un = un then un-1 should be d iv isib le by
d. Similarly, we prove (induction!) th a t un_2, un_3,
e tc ., and fin a lly uj, w ill be d iv isib le by d. But uj = 1
therefore i t cannot be divided by d > 1. The incompati­
b ility thus obtained proves the theorem,

11. Theorem. * For any m, n (u a. u_)

n - u,cfl p n /
Proof: To be d efin ite, we suppose m > n, and apply the
Euclidean algorithm to the numbers m and n:

m = nq0 + Tj, where 0 < r 1 < n,

n ~ r l (?l + r 2 ’ where 0 < r 2 < r i-

r l “ r 2<?2 + r 3’ w,iere 0 < r 3 < r 2.

r t _2 ~ + r t < where 0 < r t < r f_i.

 Number-theoretic properties 31

r t-l = r f9f

As we already know, r f is the greatest common divisor

of m and n.

Thus, m = nqQ + th is means th a t

(u
v «, un) = (unq^+r^, u n'),

or, by equation (10),

(u , u ) = (u ,u +u u ,, u ),
»’ n ' ri ~ r i +* "

or by sections 7, 8

v( u*’, u n'
) = ( u' "9Q-1, u , un),"

or by sections 10, 5
(u
' n*, un') = '(u , un').

Similarly, we prove th a t
(u . u ) - (u , u ).
1 " r2 1
(u . u ) = (u u ),
r2 rl r3 r2

(u . u ) - (u u ).
f-1 t-2 t t- 1
Combining a ll these equations, we get
(u , un )/ = v(u r f’, ur t - l ),
' n’

and since r (_1 is d iv isib le by r t , ur is also d iv is i-

t-1
ble by ur . Therefore (ur , ur ) = ur . Noting,
32 Fibonacci numbers

finally, th at r = (m, n), we obtain the required resu lt.

In particular, from the above proof we have a converse

of the theorem in section 8: i f un is div isib le by u^,
then n is d iv isible by m. For i f un is in fact d iv isib le
by u^, then, according to section 6,

<“„■ “■> = V <21>

But we have proved that

<V = “ („,■)• (22)

Combining (21) and (22) we get
un = u (n,«) ■

i.e . m = (n,n), which means that n is d iv isib le by m.

12. Combining the theorem in section 8 and the corollary

to the theorem in section 11 we have: un is d iv isib le by
um if, and only if, n is d iv isib le by m.

In view of th is, the d iv is ib ility of Fibonacci numbers

can be studied by studying the d iv is ib ility of th e ir
suffixes.

Let us find, for instance, some “signs of d iv is ib ility ”

of Fibonacci numbers. By "sign of d iv is ib ility ” we mean
a sign to show whether any p a rtic u la r Fibonacci number
is divisible by a certain given number.

A Fibonacci number is even if, and only if, its suffix

is divisible by 3.

A Fibonacci number is div isib le by 3 if, and only if,

its suffix is divisible by 4.

A Fibonacci number is div isib le by 4 if, and only

 Number-theoretic properties 33

if, i t s suffix is d iv isib le by 6.

A Fibonacci number is d iv isib le by 5 if , and only if,

i t s su ffix is div isib le by 5.

A Fibonacci number is d iv isib le by 7 if, and only if,

i t s suffix is div isib le by 8.

The proofs of a ll these signs of d iv is ib ility and all

sim ilar ones can be carried out easily by the reader,
with the help of the proposition put forward a t the
beginning of the section, and by considering the third,
fourth, sixth, fifth , eighth, etc. Fibonacci numbers
respectively.

At the same time, the reader should prove th a t no

Fibonacci number ex ists th at would give a remainder of 4
when divided by 8; also, th a t there are no odd Fibonacci
numbers div isib le by 17.

13. Let us now take a certain whole number m. If there

ex ists even one Fibonacci number un d iv isib le by m, i t is
possible to find as many such Fibonacci numbers as de­
sired. For example, such w ill be the numbers u2n. u3n,

I t would therefore be in te re stin g to discover whether

i t is possible to find a t le a s t one Fibonacci number
d iv isib le by a given number n. I t turns out th a t th is is
possible.

Let £ be the remainder of the division of k by m, and

le t us write down a sequence of p airs of such remainders:

<2, *2 S ' <n 2 ‘ “3>. < U4 > ’ <5 n+1 > .

(23)
I f we regard p airs < a 1( and <<* ' k2 ^> as equa*
2

when Oj = a2 and 6 = 62, the number of d ifferen t pairs

 34 Fibonacci numbers

of remainders of division by m equals m2. If, therefore,

we take the f i r s t m2 + 1 terms of the sequence (23) there
must be equal ones among them.

Let < u fe, > be the f i r s t p air that repeats its e lf

in the sequence (23). We shall show th at th is pair is
<1, 1 > . Indeed, le t us suppose the opposite, i.e . that
the f i r s t repeated p air is the p a ir K u k, u*+i > , where
k > 1. Let us find in (23) a p air < u j, > ( I > k)
equal to the pair < uk, ufc+1 > . Since uj_j = uJ+1 - uj
and = uk - u k , and u ,+1 = ufc+1 and if; = u h , the
remainders of division of Uj_j and uk J by m are equal,
i.e . U j.j = However, i t also follows th at
uk > = <. “ £_i, >, but the p air < , u* > is s itu a ­
ted in the sequence (23) e a rlie r than < u k, uk+1 > and
therefore < Uk, > is not the f i r s t p air th at repeated
its e lf , which contradicts our premise. This means that
the supposition k > 1 is wrong, and therefore k = 1.

Thus <1,1 > is the f i r s t p air that repeats i t s e l f in

(23). Let the repeated p air be in the tth place (in
accordance with what was established e a r lie r we can regard
1 < t < m2 -i-l), i.e . < u ( , u > = < 1. 1 > . This
means th at both u t and u f +1 when divided by m, give 1 as
a remainder. I t follows that th e ir difference is exactly
d iv isible by m. But

u t+i ~ u t ~ u t - r

so th at the (t - 1)th Fibonacci number is divisible by m.

We have thus proved the following theorem:

Theorem: H/ia(ei;er t h e w h o l e n u m b e r m, at least one

number d i v i s i b l e by m can be f o u n d among the f i r s t m^
F ibon acci numbers.

Note that this theorem does not sta te anything about

exactly which Fibonacci number w ill be d iv isib le by m.
 Number-theoretic properties 35

I t only te lls us that the f i r s t Fibonacci number d iv is i­

ble by m should not be p articu larly large.
 Ill

FIBONACCI NUMBERS AND CONTINUED FRACTIONS

1. We consider the expression

*0
9, +
9 o +-
(24)
93 +-

where q^, q 2> qn are whole po sit iv e numbers and qQ

is a whole non-negative number. Thus in contrast to the
numbers q^, q2..........qn, the number qQ can equal zero. We
shall keep th is somewhat special position of the number
q0 in mind, and not mention i t specially on each occasion.

The expression (24) is called a continued fraction and

the numbers g_,0 .............
1 q are called the pa rti al denomi-
n
nators of th is fraction.

Sometimes continued fractions are also known as chain

fractions. They are of use in a wide assortment of
mathematical problems. The reader who wants to study
them in greater detail is referred to A.Ya. Khinchin,
“Chain Fractions”*.

• Also to H.S. Wall, "Analytic Theory of Continued Fractions”

(Van Nostrand) - Translator.

36
 Continued fractions 37

The process of transformation of a certain number into a

continued fraction is called the development of th is
number into a continued fraction.

Let us see how we can find the p a rtia l denominators of

such an expansion of the ordinary fraction — .
b
We consider the Euclidean algorithm, as applied to the
numbers a and b.

a = b90 + r i-

b = r 1<7l + r 2.

rl = r 2<*2 + r 3’

r n - 2 = rn - l % - 1 + V

r n- 1 “ r n V

The f i r s t of these equations gives us

a _
b ^0 =%
r1

But i t follows from the second equation of se t (25)

that

b
— = <7 + _r 2
L = q. + —
1
rl r l rl

so that
36 Fibonacci numbers

a
~b

From the th ird equation of (25) we deduce

r
q2 + - * = q2 + 1
r2 r2

and therefore
ji
6 9n +
9i +
q2 + 1

Continuing th is process to the end (induction!) we

arrive, as is seen easily , a t the equation

a_

b +
+

'+
%

By the very sense of the Euclidean algorithm, qn > 1.

(If an were equal to unity then r rt • ,1 would equal r n and
r „« would have been d iv isib le by rn * 1, exactly, i.e . the
whole algorithm would have terminated one step e a r lie r .)
This means that in place of q we can consider the ex-
i ’n
pression (qn - 1) , i.e . consider (qn- l ) the la s t but
one p a rtia l denominator, and 1 the la s t. Such a conven­
tion turns out to be convenient for what follows.
The Euclidean algorithm as applied to a given pair of
 Continued fractions 39

natural numbers a and 6 is realized in a completely

d efin ite and unique way. Hie p a rtia l denominators of the
development of — into a continuous fraction are also
b
defined in a unique way by the system of equations des-
a
cribing th is algorithm. Any rational fraction ,
therefore, can be expanded into a continued fraction in
one and only one way.

2. Let
a) = +
(26)
91 +

be a certain continued fraction, and le t us consider the

following numbers

q° ‘ q° + ' g°

These numbers, w ritten down in the form of ordinary

simple fractions

0 <h_
So i

= 9n + ;

<?2 =*0
91 + —
?2
40 Fibonacci numbers

are called convergent fractions of the continued fraction

10

Pk P*+l
Note that the tra n sitio n from — to q---- is realized by
Qk v*+l
the replacement of the la s t of those p a rtia l denominators
which took part in the construction of th is convergent
1
fraction, i.e . q., by <j. + ----- .
**+l

3. The following lemma plays an important role in the

theory of continued fractions.

Lemma: For every continued fract ion (26) the following

relationships obtain:

Pfe+1 = Pfcgfe+1 + Pk-V (27)

^*+1 = ^ f c +l + Qfc-1* (28)

P*+l^fc “ Pk^k+1 = (29)

We prove a ll these equations simultaneously by indue-

tion over k.

We shall prove them f i r s t for k = 1.

P1 _ „ . 1 _ W l + 1
~n~ +
Qj <?! <7!

Since the numbers qQq^ + 1 and are prime to each

other, the fraction ^ 1 + 1 is reduced to its lowest

 Continued fractions 41

terms. The fraction £ i. is in i t s lowest terms accord-

91
ing to the definition, and equal fractions in th e ir
lowest terms have equal numerators and equal denominators.
This means th a t = q ^ j + 1 and = q^.

P2 _ 1 _ g0 ^ 1 g 2 + *> + g 2
(30)
Q2 9° + 9i + -±- gl g2 + 1
1 q2

The greatest common divisor of the numbers qQ( q + 1) +

+ q2 and q^q2 + 1 equals (q2, q^q2 + 1) on the basis of
section 7 of II, and on the basis of the same proposition
i t also equals (q2, 1), i.e . 1. This means th a t the
fraction on the right-hand side in (30) is in i ts lowest
terms, and therefore

P2 * q0(qlq2 + 1) + <72 = (909i + ^ q 2 + *0 = Plg2 + P0

and

Q2 = ql q2 + 1 = ^ 2 + <?()•
The equation

P2^1 ~ P1^2 “ 1
is easily verified.

The basis of the induction is thus proved.

Let us now suppose th a t the equations (27), (28) and

(29) are true and le t us consider the convergent fraction

P*+l _ Pkqk+i + P*-l

^*+1 + Qk-i

P P
The tra n sitio n from _ *±I to <i'f2 according to the
<fc+1 4+2
42 Fibonacci numbers

observation made above is brought about by the replacement

p j
of <fy+1 in the expression for ——~t- by qk ^ + ____ ; since
^+1 + qk +2
does not come into the expressions for Pk, Qk, P ^ . y
Qk _y then

pk « J k { i k + i + ^ 2 h pk .i

Qk+2 Qkfak+i + q ^ ) + ^ - i

or, remembering the inductive assumptions in (27) and

(28).

P* + 2 _ Pk + l q k+2 + P k
@k +2 ^ k + l q k+2 + <?A (3 1 )

We now prove th at the right-hand fractio n in (31) is in

its lowest terms. For th is i t is s u ffic ie n t to prove
that i t s numerator and denominator are mutually prime.

Let us suppose that the numbers ^ +1<7k+2 + a°d

@k+iqk + Qk have a certain common divisor d > 1 . The
+ 2

expression

(PA + l 9*+2 + P fe)(?* + 1 ~ ( Qk + l ^ * + 2 + ^ ) P k + l

should then be d iv isib le by d. But by the inductive

assumption (29) th is expression equals ( - l) * +1 and can­
not be divided by d.

Thus the right-hand side of (31) is in its lowest terms,

and (31) is therefore an equation between two fractions
reduced to th e ir lowest terms. This means that

P A+ 2 " Pk + l qk + 2 + Pk

and that

^fc+2 ~ ^fc+i^k+2 + ®k‘

 Continued fractions 43

To complete the proof of the inductive tra n sitio n i t

remains to show that

P k +2V k +l ~ P k H Qk +2 = < - V k+1 . (3 2 )

But in view of what was proved above,

P k+2®k+l ~ Pk + l ^ k +2 ~

- P k + i ^ k ^ k + i + p kQk+i - p k + i ^ k ^ k + i - p k+iQk'

and (32) follows d ire c tly from the inductive assumption

(29). In th is way the inductive tra n sitio n is established
and the whole lemma is proved.

Corollary.
P. . P. (-1)*
------ h- = ------- . (33)
^4+1 Qfik+1
Since p a rtia l denominators of continued fractions are
positive whole numbers, i t follows from the above lemma
that:

po < P, < P 2 < - - - .

Qo< Q i < e 2 < - -

This simple yet important observation w ill be made more
exact la te r in the book.

We now apply the lemma of section 3 to describe all

continued fractions with p a rtia l denominators equal to
unity. For such fractions we have the following in te re s t­
ing theorem.

Theorem: If a continued fr a c tio n has n p a r t i a l denom ina­

tors and each o f these p a r t i a l denom inators equals u n ity,
u 1
the f r a c t i o n equals n +l .
44 Fibonacci numbers

Proof. Let us denote the continued fraction with n unit

p a rtia l denominators by a n• Obviously

a i • a ............ a i
are consecutive convergent fractions of a.

Let

■* %
As
a , = l = _L

and
i + J - _ JL

therefore Pj = 1, P 2 = 2. Further, Pn+1 = P„q„+i + P „ - 1“

= P . - P n - 1- Therefore (compare I, section 8) P„ = un+1

Similarly, Qx = 1. Q2 = 1 and Q„+1 = Qn<?n+1 + =

= 0vi + Q —1,, so that 0n = un . This means that

_ “n+l (35)

The reader should compare th is re su lt with formulae

(12) and (29).

5. Suppose we are given two continued fractions to and to '■

to - qo + 1 . .= .<7o. +
to 1
1
<7l + 9l + •
<72 + <72 +
 Continued fractions 45

while

90 > % ’ q l > q V q 2 > q 2 ’ ••• (38)

Let us denote the convergent fractions of co ty

fo ^1_ ^2_
<?o Qi ' -
and the convergent fractions of ti/hy

p_o n p_2
<?;■
Prom the resu lts of the lemma of section 3 i t is easy to detect
th at in view of (36)

p ' > p Q, P [ > P V P'2 > P 2, ...

and

%>Qo- % > % •••

Obviously, the smallest value of any p a rtia l denominator is
unity. Ibis means that i f a ll the P a rtia l denominators of a
certain continued fraction are unity the numerators and denomi­
nators of i ts convergent fractions increase more slowly than
those of the convergent fractions of any other continued frac­
tion.

Let us estimate to what extent th is Increase is slowed down.

Obviously, discounting the continued fractions whose p a rtia l
denominators are unity, the slowest to increase are the numera­
tors and denominators of the convergent fractions of that con­
tinued fraction one of whose p a rtia l denominators is 2 and the
remaining ones unity. Such continued fractions are also con­
nected with Fibonacci numbers as shown ty the following lemma:

Lemma. I f the continued f r u c t i o n Ctt has as i t s p a r t i a l denomi­

nators the numbers qq, q^, q^, . . . . q^, while
46 Fibonacci numbers

% =<?! = <?2 = =<7i+l = =9n = 1' 9 i “ 2 (i * 0)

then

^ _ u t+1 . „ + u i Un-t+1
■ Mn-t+3 •,
. „ + u t-1
u tu n-t+3 . ,un - t•+ 1,

Proof of th is lemma is carried out by induction over i. If

i = l, then for any n

ui = 1 +
2+
n - 1 p a rtia l
A '+.—1
denominator 1

or, in view of what was proved a t the beginning of th is section.

1 ___ 1 1
<• >=! +- = 1+ V-1 = 1 +
2 +. 2un + un- 1-
2+
« n -l

U u „ + un
— n+ 2
= 1+ n
Un+ 2 Un+ 2

or, putting uQ = 0

u2 un + 2 + ulun
to =
ulun+2 + Y n
Thus the basis of induction has been proved.

Let us now suppose th at for any n

i p a rtial 1+
denominators 1+
+1+
2 +-
 Continued fractions 47

_ Ui+lUn-i+3 + Ui Un - i+ l (37)
" Ui Un- i + 3 + Ui - l Un -i+l '

Let us take the continued fraction

i + 1 partial . 1+
denominators 1 +,
'+ 1 +
2+
n-i-1
It can obviously be considered thus:

1 ±J_

i partial (38)
denominators •+ l + ■
2 +.
ln-i —
1
The continued fraction below the dotted line in (38) is, by
(37), equal to
u i+l
. ,un-1+2
■ r, + u.u
i n-1
i n-i+2 T i- l,un-i
u.u . „ + u.

The whole fraction (38) therefore equals

,1 ' i 4-
_ (u. ,)un - . „ + (u.i- l, + u.)u
T u.i+l' 1 + 2 i n-i_
1+
i i.+ ,u .
l n-i+2
„ + u.u
i n-i
u. ,u . ,„
i + l n —1 + 2
+ u.u
i n-i
U.U . ,_ + u.
i n —i+2 i - l n-i

U. ,_U . n + u. „u
_ i + 2 n - i + 2 T i+l n_i
u i+l
. n - i. + „2 + u i un-i

Thus the inductive transition has been proved and so has the
whole lemma.

Corollary: I f n o t a l l the p a r t i a l denominators o f the c o n t i n ­

ued f r a c t i o n oj are u n i t y , q / 0, and t here are no l e s s than n
o f t hese p a r t i a l denominat ors, then, on w r i t i n g (O in the form
48 Fibonacci numbers

o f an ordinary f r a c t i o n — . tee have

Q
P^U.,U . „ + U .U ■ , \ U • ,U . - + U U . , = u
i + l n-1+3 i n-i+1 i+l n-i+2 i n- i+1 n+2’

and s i m i l a r l y .

Q>
< Un+1
A substantial role is of course played here by the lemma of
section 3. on the basis of which we obtain only fractions in
th e ir lowest terms in the process of “contracting" a continued
fraction into a vulgar one. Therefore no diminution of
numerators and denominators of the fractions obtained due to
“cancelling" w ill take place.

6 . Theorem: For a cert ai n a the number o f steps in the

Euclidean algorithm applied to the numbers a and b equals n - 1
i f b - un, and f o r any a i t is l ess than n - 1 i f b < un -

Proof: The f i r s t part of the theorem can be proved quite simply.

I t is su fficie n t to take as a the Fibonacci number following 6 ,
i.e . un+ 1,. Then

The continued fraction an has n p a rtial denominators, i.e .

the number of steps of the FUclidean algorithm as applied to
the numbers a and 6 equals n - 1 .

Td prove the second part of the theorem, we suppose the con­

trary. i.e . that the number of steps of th is p articular algo­
rithm is not less than n - 1. Let us expand the ra tio — into
6
the continued fraction at. Obviously at will have no less than
n p a rtial denominators (in fact one more than the length of the
Euclidean algorithm). As 6 is not a Fibonacci number, not a ll
the p a rtial denominators of a>will be unity, and therefore,
according to the corollary of the lemma in section 5 . 6 > u ,
 Continued fractions 49

which contradicts the conditions of the theorem.

This theorem means th at the Euclidean algorithm as applied to

neighbouring Fibonacci numbers is in a sense the “longest”.

7. We shall call the expression

+ 1
«1 (39)
9a +
% +

an in fin ite continued fraction.

The definitions and re su lts of the preceding sections

can be extended quite naturally to in fin ite continued
fractions.

Let

be a sequence (obviously an in fin ite one) of the conver­

gent fractions of the fraction (39).

We shall show that th is sequence has a lim it.

With th is aim in mind, we examine separately the se­

quences

0 (41)

and
Pi 2n + l
(42)
Qi ^2n +l
50 Fibonacci numbers

From (33) and (34)

P2n+2 P2n P2n+2 P2n+1 P2n+l P2n

@2*42 $2 n $2n+2 ^n+l ^2n+l $2n

^2n+2^2n+l ^2n+1^2n °
This means th a t the sequence (41) is an increasing one.
In the same way. i t follows from

P2n+3 _ P2n+1 _ 1 1
^2n+3 *?2n+l ^+3^42 ^2n42^2n4l
th at the sequence (42) is a decreasing one.

Any term of the sequence (42) is greater than any term

of the sequence (41). Indeed, le t us examine the numbers
P P
and
^2n ^2*41
and le t us take the odd number k to be greater than 2n
and 2m + 1. I t follows from (33) that

*41
— > (43)
Ok Qk+1
and from the fact th at (41) increases and (42) decreases
i t follows that

(44)

and

*2*41 (45)
^2*41
Comparing (43), (44) and (45) we obtain
 Continued fractions 51

P2n , P2m+1
''
Q2n ^ V2«+l

From (33) and (34)

^n+l Pn
Qn+lQn ' n2 ’
and therefore, as n increases, the absolute value of the
difference of (n + l)th and nth convergent fractions tends
to zero.

From the above considerations i t is possible to conclude

th at the sequences (41) and (42) have the same lim it,
which is also, obviously, the lim it of (40). This lim it
is called the value o f the i n f i n i t e continued fraction
(39).

Let us prove now that any number can be thfe value of no more
than one continued fraction. / Let us take for th is purpose two
continued fractions to and to ( i t does not matter whether they
are fin ite or in fin ite ).

Let ijg, 9 ^, q^, . . . and q£, < 7 g2> ___ be th eir correspond­
ing p a rtia l denominators. We shall show th at i t follows from
the equation co=o>that g^ = g^, g^ = g ', g2 = q'y ••• and so

Indeed, g^ is the integral part of the number'at and g^ is the

integral part to', so th at g^ = g '. Further, the continued
fractions to and to can be represented respectively in the form

1 1
90 and g ' t —r
to1

where t o ^ and to ^ are again continued fractions. I t follows

from to = cu and gQ = g ' that a tj = at f also. This means that
the integral parts of the numbers at^ and at f are also equal,
i.e . gj = q y Continuing these arguments (induction!) we see
that gj = g2, 9 3 = q 3 . etc.
52 Fibonacci numbers

Since a rational number can always be expanded into a fin ite

continued fraction, i t follows from the foregoing proof th at i t
cannot be expanded into an in fin ite continued fraction. I t
follows th at the value of an in fin ite continued fraction must
necessarily be an irra tio n al number.

The theory of expansions of irra tio n a l numbers into continu­

ous fractions represents a branch of theory of numbers which is
rich in content and in terestin g in i ts re su lts. Ve sh all not
delve deeply into th is theory, but we shall consider only one
eiample connected with Fibonacci numbers.

6. Let us find the value of the in fin ite continued frac­

tion
1
1+
1+
1 + a• ■
As we have proved above, th is value is lim a n- Let us
calculate th is lim it. n 00

As has already been established in I, section 12, un is the

n
Cl
nearest whole number to —— ; th is means that
v5

‘ VT * »-•

where | 9 n | _L. whatever n is.

2
Therefore, in view of the resu lts of section 4,

a ^ 1
+ 9 n+1
1im a = 1im U"+1 = lim V i"
n oo n n-*oo _n
V T ^ n
 Continued fractions 53

a + 0 n+l nV 5
1ini a + @n+1 \
a n -» o ° a" /
= lim- 0nV5 1 *
n -» o o lim
1+ oo i +•
a"
But 0 n +1, Vjf is a bounded quantity ( its absolute value is
less them 2) and a continues to increase in d efin itely
as n tends to in fin ity (because a > l ) . This means

1im
n -fO o

Also, for the same reasons,

lim ----------= 0,
rwoo
a
and we obtain
l i ma = a .
n-too n
The theorem that has been proved means th at the ra tio
of neighbouring Fibonacci numbers approaches a as th e ir
suffixes increase. This re su lt can be used for the
approximate calculation of the number a . (Compare the
calculation of un in I, section 12.) This calculation
produces a very small error, even when small Fibonacci
numbers are taken. For example (correct to the fifth
decimal place)
u 55
10 _
1.6176,
u9 34

and a = 1.6180. As we see, the error is less than 0.1%.

Of the errors involved in the approximate calculation

of irratio n al numbers by convergent fractions i t turns
out that the number a represents the worst case. Any
other number is describable by means of its convergent
54 Fibonacci numbers

fractions in some sense more exactly than a . However,

we shall riot stop to consider th is circumstance, in te re s t­
ing though i t be.
 IV

FIBONACCI NUMBERS AND GEOMETRY

1. Let us divide the unit segment AB into two p arts in

such a way ( f ig .2) th a t the greater part is the mean
proportional of the smaller part and the whole segment.

Cg A Cf B

Fig. 2.

For th is purpose we denote the length of the greater

part of the segment by x. Obviously, the length of the
smaller part w ill be equal to 1 - x and the conditions of
our problem give us the proportion:

1 X (46)
X 1 — X

whence
x 2 = 1 - x. (47)

The positive root of (47) is 1 +v^~so th a t the ratio s

2
in proportion (46) are equal to
1 = 2 2(1 + V 5 ) m 1 + V5 , „
T -1 + v£ (-1 + V 5 ) ( 1 + V 5 ) 2

each. Such a division (at point C.) is called median

section. I t is also called the golden section.

55
56 Fibonacci numbers

If the negative value of the root of the equation (47)

is taken, the point of section C2 lie s outside the seg­
ment AB (th is kind of division is called external section
in geometry) as shown in fig. 2. I t is easily shown that
here, too, we are dealing with the golden section:
CJ3 AB
— ------------- = a .
AB C24

2. The golden section appears quite frequently in geo­

metry.

The side a 1Q of the regular decagon ( f ig .3) inscribed

in a c irc le of radius R is equal to

2R sin 360° .,
2 . 10

i.e . i t is 2R sin 18°.

We now calculate sin 18°. From well known formulae of

trigonometry we have
sin 36° = 2 sin 18° cos 18°,
cos 36° = 1 - 2 s in 218°,
 Geometry 57

so that
sin 72° = 4 sin 18°X cos 18° ( 1 - 2 s in 218°). (*)
Since
sin 72° = cos 18° # 0,
then i t follows from (*) that
1 = 4 sin 18° ( 1 - 2 s in 2 18°),
and therefore sin 18° is one of the roots of the equation
1 = 4x(l - 2x2),
or
8x 3 - 4x + 1 = 0.
Factorizing the left-hand side of the l a tte r equation we
obtain
(2x - l ) ( 4 x 2 + 2x - 1) = 0,

whence
x = 1 r _ 1-1 +V5
■' ■ . - -1 -V"5
1 T
A,
3 " — :-----

As sin 18° is a positive number, other than — , there-

2
fore sin 18° = ~ 1 .
4
Thus
a . n = 2B V T - 1 =
1U 4 2 a

In other words, a 1Q equals the larger part of the

radius of the c irc le , which has been divided by means of
the golden section.

In practice, in calculating a 1Q we can use the ra tio of

neighbouring Fibonacci numbers (I, section 12 or III,
section 8) instead of a and reckon approximately that
58 Fibonacci numbers

a 1n is _?L fl or even 2 . Ft.

10 13 8

3. Let us examine a regular pentagon. Its diagonals

form a regular pentagonal sta r.

Fig. 4.

The angle A F D equals 108°, and the angle ADF equals

36°. Therefore, according to the sine rule
-d£ = sin 108° a sin 72° = 2 cos 3gO = 2 1 + VT , a .
AF sin 36° sin 36° 4
Since i t is obvious that AF = AC, then
AD _ AD _
AF AC a ’

and the segment AD is divided a t C according to the

golden section.

But from the definition of the golden section

Noting that AH - ( D , we obtain

 Geometry 59

AC _ A B _
AB BC~ a '

Thus, of the segments

BC, AB, AC, AD

each is a times greater than the preceding one.

I t is le f t to the reader to prove that the equality

also holds.

4. Let us take a rectangle with sides a and b and le t us

proceed to inscribe in i t the la rg e st possible squares,
as shown in fig . 5.

Fig. 5.

The arguments in II,se c tio n 2 show th a t such a process

in the case of whole a and b corresponds to the Euclidean
algorithm as applied to these numbers. The numbers of
squares of equal size is in th is case (III, section 1)
equal to the corresponding p a rtia l denominators of the
expansion of into a continued fraction,
o
60 Fibonacci numbers

If a rectangle whose sides are to each other as neigh­

bouring Fibonacci numbers is divided into squares (fig.
6), then on the basis of II I , section 4 a ll squares
except the two sm allest ones are differen t.

A B

X
D F C
Fig. 6. Fig. 7.

Now, l e t the ra tio of the sides of a rectangle be equal to

a . (We shall call such rectangles "golden section re c t­
angles” for sh o rt.) We now prove that a fte r inscribing
the largest possible square into a golden section re c t­
angle (fig. 7) we again obtain a golden section rectangle.

Indeed,

also
AD = A E = EF,
since AEFD is a square.

This means that

EF _ AB - EB _ 2
EB EB ~ a ~
n
But a — 1 = a . so that
EF _
EB ~ a ’
 Geometry G1

I t is shown in fig. 8 how a golden section rectangle

can be “nearly completely*’ exhausted by means of squares
I , I I , I I I , ... Each successive time a square is inscribed,
the remaining figure is a golden section rectangle.

The reader should compare these arguments with sections

4 and 8 of the preceding chapter. We note that i f a
golden section rectangle I and squares I I and I I I are
inscribed in a square, as shown in fig. 9, the remaining
rectangle turns out to be a golden section rectangle
also. The proof of th is is le f t to the reader.

5. Golden section rectangles seem "proportional” and

are pleasant to look at. Things of th is shape are con­
venient in use. Therefore, many "rectangular” objects
of everyday use (books, matchboxes, suitcases and sim ilar
things) are given th is p a rticu lar form.

Fig. 9.
62 Fibonacci numbers

Various id e a list philosophers of ancient and mediaeval

times raised the outward beauty of golden section rec­
tangles and other figures which conform to the rules of
median section into an aesthetic and even a philosophic
principle. They trie d to explain natural and social
phenomena in terms of the golden section and certain other
number relationships, and they carried out a ll kinds of
mystic "operations” on the number a and its convergent
fractions. I t is clear th at such "theories” have nothing
in common with science.

Fig. 10. Fig. 11.

6^ We shall round off our presentation with a l i t t l e

geometrical joke. We shall demonstrate a "proof” that
64 = 65.

To do th is we take a square of side 8 and we cut i t up

into 4 parts as shown in fig. 10. We put the parts to ­
gether to form a rectangle ( f ig .11) of sides 13 and 5,
i.e . of area equal to 65.

The explanation of th is phenomenon, puzzling at f ir s t

sight, is easily found. The point is that the points A,
B, C and D in fig. 11 do not really lie on the same
stra ig h t line, but are the v ertices of a parallelogram,
whose area is exactly equal to the “extra” unit of area.

This plausible, but misleading "proof” of a statement

which is known beforehand to be incorrect (such "proofs”
are called sophisms) can be carried out even more "con­
vincingly” i f we take a square of side equal to some
 Geometry 63

Fibonacci number with a su ffic ie n tly large even suffix,

u2n, instead of a square with side 8. Let us cut up th is
square into parts (fig. 12) and le t us put these parts

U2n-2

U2n-1

Fig. 12.

together to form a rectangle (fig. 13). The “empty space1

in the form of a parallelogram stretched along the
diagonal of the rectangle is equal in area to unity,

Fig. 13.

according to I, section 6. I t is easily calculated th at

the greatest width of th is s l i t , i.e . the height of the
parallelogram, is equal to
1_______
/ 2 2
V u2n + u2n-2
If, therefore, we take a square with side 21 cm and
“convert” i t into a rectangle with sides 34 cm and 13 cm,
the greatest width of the s l i t is found to be
64 Fibonacci numbers

— cm,
V I + 82
2 2

i.e . about 0.4 mm, which is d iff ic u lt to detect by eye.

 V

CONCLUSION

Not a l l t h e p roblem s c o n n e c t e d w it h F i b o n a c c i numbers

can be s o l v e d a s e a s i l y a s t h e o n e s we h ave c o n s i d e r e d .
We shall* indicate several problems the answers to which
are e ith e r not known at a ll or can only be obtained by
quite complicated means with the application of much more
powerful methods of investigation.

J_. Let be div isib le by a certain prime number p,

while none of the Fibonacci numbers smaller than un is
d iv isib le by p. In th is case we shall c all the number
p "the proper divisor of un”. For example. 11 is the
proper divisor of u1Q, 17 is the proper divisor of u 9 and
so on.

I t turns out th at any Fibonacci number except u^, u2>

Ug and u 12 p o s s e s s e s a t l e a s t one p r o p e r d i v i s o r .

2. The natural question arises: What is the suffix n of

the Fibonacci number whose proper divisor is the given
prime number p?

From II, section 13 we know that n <^p2. I t is possible

to prove that n <^p + 1 . Furthermore, i t is possible to
establish th at i f p is of the form 5t + 1 then u . is
d iv isible by p, and i f p is of the form 5t + 2 then up+1
is d iv isible by p. However, we have no formula to
indicate the suffix of the term with the given proper
divisor p.

65
66 Fibonacci numbers

3. We have proved in II, section 9 that a ll Fibonacci

numbers with composite suffixes, except u^, are composite
themselves. The converse is not true, since, for example,
u ig = 4181 = 37 X 113, The question arises: is the
number of a ll prime Fibonacci numbers fin ite or in fin ite ,
in other words, is there among a ll the prime Fibonacci
numbers a greatest one? At th is moment th is question is
s t i l l far from being solved.
 F IB O IA C C I
IUMBERS
N. N.Vorob’ev
The 13th-century mathematician Leonardo of Pisa, whose surname
was Fibonacci, strove to illuminate the dark ages of Western mathe­
matics. Familiar with the mathematical literature of Arabia, India
and the ancient world, he introduced into Europe the concept of
zero, decimal notation, and algebra.
The numbers known by his name date back to a 750-year-old
problem concerning the number of descendants produced by a single
pair of rabbits in one year. The numbers of resulting pairs at the end
of each month form a Fibonacci sequence, which follows the formula
u n+1 — Un + u„_i •
In this book N. N. Vorob’ev presents the solution to the rabbit
problem, treats some of the remarkable properties of Fibonacci
numbers, and discusses their occurrence in number theory, continued
fractions and geometry. In his discussion of the “Golden Section”
rectangle (one in which the lengths of the sides can be expressed as
a ratio of two successive Fibonacci numbers), he cites the attempts
of ancient and medieval philosophers to base esthetic and philo­
sophical principles on the beauty of these figures. Written in a light
and entertaining style, this brief book should interest both the lay­
man and the serious student of mathematics.