Book I. Construction of numbers, length and area.

ISBN 978-952-94-0570-1
ISBN 978-952-94-0571-8
ISBN 978-952-94-0572-5
ISBN 978-952-94-0573-2

Copyright © Naumov Ivan, 2018. Illustrator, designer

Gudalova Tatiana

Finland 2018
 Preface

The 1-st book is done to provide a complete construction of numbers, length and area.
It is obviously the most reasonable way to start a course of mathematics. Almost everyone knows
that when we operate with numbers, we do it according to the certain rules.
Unfortunately, there is never any explanation of where these rules came from, despite the fact that
it is a very important question, to say the least.
Really, almost anyone knows the “sign rules” a  (b)  (a)  b  (a  b), (a)  (b)  a  b [A],

or a commutative law a  b  b  a [B], or an associative law (a  b)  c  a  (b  c) [C].

But it’s not typical of people to ask: “Why are these laws true?”.
If someone gets a “math education”, he will be told that the real numbers R form a “field” and even
a “complete field”, and that’s where all the properties like [A],[B],[C] came from. But it explains
nothing, because it immediately leads to the question: ”Why do the real numbers R form a field?”.
And the better person knows math, the easier he can understand that there must be some good
explanation of all these properties, as they aren’t obvious at all. So, there are “infinitely many” real
numbers. Why do all of them without any exceptions obey to the certain rules? Why can we take
some line and compare exactly one (appropriate) real number to every point of that line, such that
every real number will be compared to some point? I’ve never seen any students curriculum where
any of these fundamental questions are normally explained.
Is it so because it is something obvious or something insignificant? No, in reality, not at all.
Eventually, there exist only one reasonable way to answer all these questions - to provide a
complete and clear construction of the field R . We will start from the simpliest basic objects and
then, step by step, we will be building real numbers, and during the process we will check
straightly all their properties, without accepting any of them for granted. As far as I’m concerned,
the numbers construction     Q  R  C is one of the most amazing and the most
important part of mathematics. There exist several known approaches to such construction, and
some of them have serious flaws, because if we follow them, most of the properties of numbers will
stay unexplained. The main goal of this book is to provide the most convenient construction, where
definitions are clear and there is a strict logical structure, there are no “gaps ” and “uncertainties”,
and there are no inconvenient and protracted proofs at all. My main goal was to give only a good
quality information and to make the narrative consecutive and clear.

All The Best, Naumov Ivan

naumov8v@gmail.com
 Table of contents

Natural numbers …………………………………………………………………… 5

Groups, Rings, Fields……………………………………………………………….. 33

Matrixes………………………………………………………………………………. 46

Integer numbers…………………………………………………………………….. 53

Rational numbers…………………………………………………………………… 73

Sequences and limits……………………………………………………………….. 84

Real Numbers………………………………………………………………………... 93

Construction of length……………………………………………………………… 130

Construction of area………………………………………………………………… 149

Angles…………………………………………………………………………………. 165

Literature…...………………………………………………………………………... 171

I express my sincere gratitude to: Galochkin Alexandr Ivanovich

Drutsa Alexey Valerievich

Kozko Artem Ivanovich

Because their lectures and seminars made me interested in math

and formed my approach to this subject.

I also would like to thank Alexander Bebris, as his educational web site and the project
“English Galaxy” provided a great course of English grammar, which is essential for writing.
 Natural Numbers

Basic objects and notions.

Def: a set  is any collection of unique objects, every object of  must be understanded as
exceptional and different from any other object of  . Any separate objects in such collection
must be considered as different objects.
We use small letters a, b, c, d .... to denote objects of any set.
Every small letter always denotes one object of a set. It is allowable to use different letters a, b
to denote the same object. And if both letters a, b denote the same object, we must write a  b.
If letters a, b denote different objects, we must write a  b.
We will usually work with sets of symbols, like the set of natural numbers {1,2,3,4 ...}- notice that

every symbol of this set is unique. And we can use small letters a, b, c, d .... to denote some natural
numbers. If letters a, b both denote the same natural number, then we must write a  b ,
if a, b denote different natural numbers, we must write a  b.
Sets  and  are equal    if they consist of exactly the same elements.
In any other case we write    .
If  contains a , we write a   . If  doesn’t contain a , we write a   .
When a   we say “ a belongs to  ”, when a   we say “ a doesn’t belong to  ”.
The writing    means “  is a subset of  ”, it means that every element of  belongs to  .
For convenience we define an empty set  - the set without any elements.
And by definition, any set  contains an empty set  as a subset,    .

Exercise. If    and    , then    .

Example. Any set that consists of elements a, b can be written as {a, b} , or {b, a} .
A set that consists of elements a, b, c can be written as: {a, b, c} , or {a, c, b} , or {b, a, c} ,
or {b, c, a} , or {c, a, b} . The positioning of elements inside the brackets {...} is irrelevant.

Def. Let A and B are any sets. The intersection of A and B is the set
    {a | a   and a  } which consists of all the elements that belong to A and to B.
The union of A and B is the set     {a | a   or a  } which consists of all the elements
that belong to A or to B (Notice: if a is a common element of A and B, we do not include it twice
in   , only once). And finally, the difference of A and B is the set
 \   {a | a   but a  } which consists of all the elements that belong to A, but do not
belong to B.
Def. A and B are any sets. And f is any rule such that: for every element a   the rule f
compares exactly one element b   , then f is called a function (or a mapping) from A to B.
And we denote it like f :    . Let a   and f compares to a some element b   , then
we write 6 of a ” and “ a is a preimage of b ”.
f (a)  b . And we say that “ b is an image
 Natural Numbers

So, when we write f :    , we mean that for every a   some element b   is compared.

But there is no requirement [pict1] that every element b   is compared to some a   , some
elements of  may not have any preimage in  . The set  is called a domain of f , and we can
write D( f )   . The set E ( f )  {b   || a   : f (a)  b} - is a
subset of  , and E ( f ) is called an image of f . Notice that E ( f )
mustn’t coincide with  , it just belongs to  . In general, f may
transfer different elements a, c   into the same b   , so it can
“glue together” different elements of  [pict1].
In the case when f doesn’t “glue together” any elements of 
a  c   , then f (a)  f (c)   ),
(I.e., for any a, c   : if pict.1

we say “ f maps  into B”.

IfB is “completely covered” by f (for every b   there exist
some a   such that f (a)  b ), then we say “ f maps  onto  ”.

If f does not “glue together” elements of  and

“coversB completely” ( f is into and onto mapping),
then we say that “ f is one-to-one mapping from  to  ”.

Only when f is one-to-one, we can define the inverse mapping

f 1 :    [pict2], for any b   : if b  f (a) , then f 1 (b)  a .
pict.2
Exercise. If f is one-to-one, then f 1 is also one-to-one.
Also, for anya   we have f 1 ( f (a))  a and for any b   we have f ( f 1 (b))  b .
Let we have some mapping f :    . In order to show that f is one-to-one we need:
[Step1] To show that f covers  : for any b   there exist some a   such that f (a)  b .
[Step 2] To show that f doesn’t “glue together” elements of . Usually it can be done in the
next way: we fix any elements a  c   and we assume that f (a)  f (c) . From the last equality
" f (a)  f (c)" we try to derive (by using some properties of a concrete mapping f ) that a  c ,
it contradicts to our initial requirement a  c . And therefore, our assumption " f (a)  f (c)"
was false and f (a)  f (c) .

Def. A and B are any sets. If there exist any one-to-one mapping f :    , then we say
“A is equivalent to B”, and we write   .
If f :    is one to one, then f 1 :    is one-to-one, therefore        .
And we can just say “sets A and B are equivalent”.
If there is no any one-to-one mapping f :    , then we say “A is not equivalent to B”.
7
 Natural
General definition.  is a set of any kind. AndNumbers
"" is some condition that may be true for
some pairs a, b   . And for any a, b   exactly one of the next cases is true:
ab (the condition  is true for the pair a, b ) or a  b (the condition  is not true for the pair a, b ).
And also [A] a  a (for any a   ) (reflexivity), [B] If a  b , then b  a (symmetricity),
[C] If a  b and b  c , then a  c (transitivity). Then "" is called an equivalence relation on  .
Example.  is a set, and  is divided into some sets
 ||    [pict3] without common elements:
     if    . Let’s define that elements a, b are
equivalent a  b if and only if both a, b belong to
the same set  . Then "" is an equivalence relation on  .
Really, aa for any a   , if a  b , then both a, b belong to
some set  and therefore b  a . If a  b and b  c , then a, b, c pict.3
belong to the same set   and therefore a  c .
The main property of any equivalence relation.  is a set, then any equivalence relation ""
on  divides  into disjoint sets, which are classes of equivalent elements.
Proof. Let we have some equivalence relation "" on  . For every a   let’s consider the class
a  {all  ||   a}. The set  a includes all the elements    which are equivalent to a.
Let’s fix an arbitrary pair of sets  a and b . Let’s show that exactly one of the next cases is true:
a  b   , or a  b . If a  b   , then we have the first case. Let a  b   ,
then there exist some such that   a and   b . From the condition   a follows that
  a . From the condition   b follows that   b . Then a  b . Any element a  
belongs to some class  a , and any classes  a , b coincide completely, or do not have any common
elements at all, then  is divided into disjoint sets. Everything is proved.

Def. A set  is called an ordered set if there exist some condition " " that may be true for
some pairs a, b   . If " " is true for some pair a, b , we write a  b . And also:
[A] For any a, b   only one of the next cases is true: a  b , or a  b , or b  a .
[B] " " is transitive: if a  b and b  c , then a  c .
In such case we say that " " is an order relation on  . And we can write (, ) instead of  ,
in order to emphasize that there is an order on  . We must agree that in the case a  b we will
a ”. We also need to agree that the writings a  b
say “ a is less than b ”, or “ b is greater than
and b  a are equivalent. If a  b or a  b , we will write a  b and say “ a is not greater than b ”,
or “ b is not less than a ”.

Def. ( , ) and (, ~

) are some ordered sets. If exist one-to-one mapping f :    such that:
a  b (in )  f (a)  ~ f (b) (in ) , then we say that ( , ) is isomorphic to (, ~
) , and we write
~  . It’s very easy to show that:       8 and we can say “ordered sets ,  are isomorphic”.
 Natural Numbers

Natural numbers.
Def. N is a nonempty set. There exist the element e   , which is called “one”, and one-to-one
mapping n :    \ {e}. For any element a   , the element n(a)   is called “next to a ”.
And the induction axiom is true in  [pict4].

|Induction axiom| Let  is a subset of  ,

and [ A] , [B] are true.
[ A] e   [B] If some a belongs to  , then n(a)
also belongs to  .
Then there must be    . pict.4

Then  is called a set of natural numbers and elements of  are called natural numbers.
It’s easy to see that the set of natural numbers is not an ordinary set, it is a structure which consists
of the ( set )  (mapping )  (induction axiom) .

Let’s provide an example which satisfies to the given definition. It’s very easy to do, let’s take the
letter A and let’s consider the set of “words”   {A, AA, AAA , AAAA, AAAAA, AAAAAA, ...}.
A . We designate e  A .
Our set comprises all the possible “words” with only one letter
And we also define one-to-one mapping n :    \ { A}: for any word AA ... A we compare

the word on the right of it, so n( A ... A)  A ... AA . Then n is one-to-one mapping    \ { A} .
Let’s check that the induction axiom is true in  , really let’s assume that [ A] and [B] are true for
some set    . From [ A] follows that A   , from [B] follows that if some word belongs to  ,
then the word on the right of it also belongs to  . From [ A] we have A   , then from [B] follows
that AA   , then AAA  , then AAAA  and etc. And therefore,  must contain every word
A ..... A , then   . So the induction axiom is true. And our set  is a set of natural numbers.
Similarly  B  {B, BB, BBB, BBBB, ......} is also a set of natural numbers. As letters A and B
are in fact different symbols, then the sets  and  B are different sets. But it’s very easy to see
that these sets have the “same structure”. In math such sets are called isomorphic.

The main theorem. Any set of natural numbers N can be written as:
  {e, n(e), n(n(e)), n(n(n(e))), n(n(n(n(e)))) .........}.

Let’s assume that this theorem is true. From this theorem we see that if  is written like
  {e, n(e), n(n(e)), .......}(each next element is an image (after the mapping n ) of it’s left
neighbor), then for any a   the element n(a) is literally next to a .
Really, if a    a  n... n(e).., then n(a)  nn(... n(e)..) stays on the right of a and n(a)
is literally next to a.
9
 Natural Numbers

Let now   ,   are some sets of natural numbers. Then, according to the main theorem:
  {e , n(e ), n(n(e )), n(n(n(e ))), n(n(n(n(e )))) ...}

  {e , n(e ), n(n(e )), n(n(n(e ))), n(n(n(n(e )))) ...}

e n(e ) n(n(e )) n(n(n(e ))) n(n(n(n(e )))) ......

       .
e n(e ) n(n(e )) n(n(n(e ))) n(n(n(n(e )))) ......

From here we see that   ,   are almost identical, they have similar elements and the same
structure.
In order to prove the main theorem let’s prove at first several auxiliary lemmas.

Lemma1. a, b -are any natural numbers. Then: a  b  n(a)  n(b) and a  b  n(a)  n(b) .
Proof. By definition n :    \ {e} is one-to-one mapping (from here immediately follows
everything we need).
Def. Let a is any natural number.
Any element n(n(... n(a)..)) (the image of a
after several mappings n ) [pict5] is called a
descendant of a. pict.5

Lemma2. Any natural number a differs from any of it’s descendants.

Proof. Let  is the set of all natural numbers, every of which differs from any of it’s descendants.
Then: [A] e   . Really, let’s assume the contrary: e  n(n(... n(e)..)) . Then the mapping n
transfers the element (natural number) n(... n(e)..) into e . It is impossible, because
n :    \ {e} and there is no any element in  that is transferred into e .
[B] if some a   , then a differs from any of it’s descendants. Let’s assume that n(a)   ,
it means that n(a) is equal to some descendant of n(a), so n(a)  n(n(... nn(a) ...)) .
Here n(a) is the element which is next to a, and on the right side we have the element
n(n(... nn(a) ...)) which is next to n(... nn(a) ...) and these elements are equal.
According to the lemma1, there must be a  n(... nn(a) ...) , so a is a descendant of a , and
we have a contradiction. Therefore, our assumption n(a)   was false, then n(a)   .
By the induction axiom    and the lemma2 is proved.

Lemma3. Any natural number a  e is a descendant of one e .

Proof. Let Ω is the set of all descendants of e . Let’s consider the set     {e} .

[ A] e   , [B] If some a   , then a    {e} . If a  e , then n(a)  n(e) and n(a)

10
 Natural Numbers

is a descendant of e and n(a)    n(a)   . If a   , then a  n(n(... n(e)..)) and

n(a)  nn(n(... n(e)..)) is also a descendant of e , so n(a)   . According to the induction axiom,
   , then     {e} . According to the lemma2, the element e differs from any of it’s

descendants, then e does not belong to  , therefore     {e}     \ {e} , then any natural
number, except one, is a descendant of one.

Let’s prove now the main theorem. Let  is a set of natural numbers. Let’s take e   and
build the set  ord  {e, n(e), n(n(e)), n(n(n(e))), n(n(n(n(e)))) ........}it includes the element e
and all it’s descendants. Let’s show that  ord is really a set, we have to show that there are no
identical elements in the row: {e, n(e), n(n(e)), n(n(n(e))), n(n(n(n(e)))) ........}(in the other case

 ord contains twice the same natural number and it is not a set). Let’s assume the contrary:
~
there are some elements d , d in the row [R] {e, n(e), n(n(e)), n(n(n(e))), n(n(n(n(e)))) ........}
~
which are equal d  d and stay on different positions.

~ ~
Without loss of generality d is on the left of d in [R], then  ord  {e, n(e), n(n(e)) ... d .... d .....}.
~
Here every element, except e , is an image of it’s left neighbor, then d is a descendant of d , so
~ ~
n(... n(d )..)  d . As d  d , then n(... n(d )..)  d it contradicts to the lemma2, therefore, there
are no equal elements in [R]. So  ord is a set, each element of  ord is some natural number, then
 ord   .
Conversely    ord , really, according to the lemma3, any natural number a  e is a descendant
of one, so it belongs to  ord and also e   ord , therefore    ord . Then    ord .
The main theorem is proved.

Def.  is a set. And  is some rule that for every pair of elements a, b   compares some
element  a  b of  . Then  is called a binary operation (or just an operation) on .
And we write  :      .
 is called commutative if a  b  b  a for any a, b   .
 is called associative if (a  b)  c  a  (b  c) for any a, b, c   .
Associativity means: for any a, b, c   , the result of applying  to the pair (a  b), c is always
exactly the same as a result of applying  to the pair a, (b  c) .

Def. Addition "" of natural numbers is the binary operation on N such that
[A] a  e  n(a) a   [B] a  n(b)  n(a  b) a, b   .
Let’s show that this definition is correct. There exist the unique binary operation on N with the
properties [A] and [B].

11
 Natural Numbers

Existence. Let’s take one e and define the addition " e " on the set of all pairs {(e, b) || b  }
where e is fixed and b is an arbitrary natural number. We define: e e b  / by def /  n(b) b   .

Let’s check that " e " is an addition "" in the case a  e .

e e
 e

By definition e  e  n(e) (property [A]) and e  n(b)  nn(b)   n e  b (property [B]).

Let  is the set of all natural numbers a , for every of which there exist it’s own sum " a " , which
is defined on the set of all pairs {(a, b) || b  } and satisfies to [A] and [B]. By the induction axiom

we will show that    . [ A] e   (as were shown above) [B] If some a   , then it has it’s
own sum " a " , which is defined on the set of all pairs {(a, b) || b  } such that [A] and [B].
Let’s take the next element n(а) and define for it it’s own sum " n (a ) " on the set of all pairs
{(n(a), b) || b  } through the sum " a " . By definition: n(a)  n( a ) b  / by def /  n(a  a b) b   .

Then n(a)  e  n(a  a e)  n(n(a)) (property [A]) and

n( a)

     
n(a)  n( a ) n(b)  n a  a n(b)  n n(a  a b)  n n(a)  n( a ) b (property [B]) then n(a)   .

Therefore    . It means that for every natural number a there exist it’s own sum " a " , which
is defined on the set {(a, b) || b  } and satisfies [A] and [B]. Let’s finally define the operation ""

on every pair (a, b) . We define: a  b  / by def /  a  a b . Then "" satisfies to [A] and [B].
The existence is proved.

Uniqueness. Let’s show that the operation "" is unique. We assume that there exist some other

addition "" with the same properties [A] and [B]. If we show that for every pair (a, b) of natural

numbers a  b  a  b , then "" and "" are the same operation. Let’s fix an arbitrary a   .
Let M is the set of all natural numbers b such that a  b  a  b . We will show that M  N :
[ A] e   , because a  e  n(a) and a  e  n(a) (property [A]), then a  e  a  e .

[B] if some b   , then a  b  a  b , then a  n(b)  n(a  b) and a  n(b)  n(a  b)

(property [B]) as a  b  a  b , then according to the lemma1, n(a  b)  n(a  b) , then

a  n(b)  a  n(b)  n(b)   . Then a  b  a  b for any b   , but a is an arbitrary fixed

natural number, therefore a  b  a  b for any a, b   . a  n(b)  n(a  b)

12
 Natural Numbers

Properties of addition
Associativity. (a  b)  c  a  (b  c) for any a, b, c   .

Proof. Let’s fix arbitrary numbers a, b   . Let  is the set of all natural numbers e such that
(a  b)  c  a  (b  c) . [A] e   , because (a  b)  e  n(a  b)  a  n(b)  a  (b  e)
[B] if some c  M , then (a  b)  c  a  (b  c) . Let’s consider:
a  b  n(c)  n((a  b)  c)  n(a  (b  c))  a  n(b  c)  a  (b  n(c))  n(c)   .
And, according to the induction axiom,    . Then (a  b)  c  a  (b  c) for any a, b, c   .

Commutativity. a  b  b  a for any a, b   .

Step one. Commutativity for one: a  e  e  a for any a   .
 -is the set of all natural numbers a for which a  e  e  a .
[ A] e   , because e  e  e  e (left and right sides are exactly the same).
[B] if some a   , then a  e  e  a . Let’s consider:
n(a)  e  (a  e)  e  (e  a)  e  / associativ ity /  e  (a  e)  e  n(a) , then n(a)   .
Therefore    .

Step two. General commutativity a  b  b  a || a, b   . Let’s fix an arbitrary

natural number a . And  -is the set of all natural numbers b for which a  b  b  a .
[ A] e   because a  e  e  a (step one).
[B] if some b   , then a  b  b  a . Then
a  n(b)  n(a  b)  n(b  a)  b  n(a)  b  (a  e)  b  (e  a)  (b  e)  a  n(b)  a  n(b)  
So    and commutativity is proved.
From commutativity follows that the properties [A] and [B] of addition can be rewritten as:
[A] a  e  e  a  n(a) a   , [B] a  n(b)  n(b)  a  n(a  b) a, b   .

Order. Let’s define the order relation " " on  .

Lemma4 [for order]. For any pair of natural numbers a, b   exactly one of the next cases is true:
[A] a  b  c for some c   [B] b  a  c for some c   [C] a  b .
Proof. By using the induction axiom it’s very easy to prove the next auxiliary fact
Auxiliary1. There is no any natural number a , such that a  a  b for some b   .
From the auxiliary1 immediately follows that: if some pair a, b   one of the cases [A],[B],[C]
is true, then any other of these cases is not true fora, b . So we can reformulate our lemma in
the equivalent form. [Equivalent form] For any pair of natural numbers a, b   some one of
the next cases is true: [A] a  b  c for some c   [B] b  a  c for some c   [C] a  b .
Let’s fix an arbitrary a   . Let  is the set of all natural b, for every of which some one of
the cases [A],[B],[C] is true.

13
 Natural Numbers

[ A] b  e   . Really, if a  e , then the case [C] is true. Let a  e , then,

according to the lemma3, a is a descendant of e , so
a  n(... n(n(n(e)))..)  // n(e)  e  e //  n(... n(n(e  e))..)  // n(e  e)  (e  e)  e // 
 n(... n((e  e)  e)..)  // n((e  e)  e)  (e  e)  e  e //  ....  e  e  ....(e  e)  e  c  e  e  c ,
then a  e  c , and we have the case [A]. So e   .
[B] if some b   , then [A] a  b  c or [B] b  a  c or [C] a  b . Let’s consider n(b)  b  e .
Let we have [A] a  b  c . If c  e , then a  b  e  n(b) , then we have the case [C] for the pair
a, n(b) . If c  e , then (as we showed above) c  e   where    , then
a  b  c  a  b  (e  )  / associativ ity /  (b  e)    n(b)   , then we have the case [A]
for the pair a, n(b) . Let we have [B] b  a  c , then
n(b)  b  e  (a  c)  e  n(b)  a  (c  e)  n(b)  a  n(c) , so we have the case [B] for
the pair a, n(b) . Let we have [C] a  b , then n(b)  b  e  a  e  n(b)  a  e so we have
the case [B] for the pair a, n(b) . So, in any case: if b   , then n(b)   , therefore,
by the induction axiom,    . Everything is proved.
Def. Let a, b   are any natural numbers. In the case [A] a  b  c we will write a  b .
In the case [B] b  a  c we will write b  a .
Assertion1. " " is an order relation on  (and  is an ordered set).
Proof. Put simply, we need to show that any elements a, b   are comparable, and that " " is
transitive, then (by definition) " " is an order relation on  . From the lemma4 follows that for
any elements a, b   only one of the next cases is true: a  b , or a  b , or b  a .
And the relation " " is transitive: let a  b and b  c , then a    b and b    c ,
then (a  )    b    a  (   )  b    a  (   )  c  a  c .
During the proof of the lemma4 we have shown that if a  e , then a  e  c , so any natural number
a  e is greater than e . Therefore, e is the least natural number. And for any natural a obviously
a  n(a) . And numbers a, n(a) are neighbors (there is no any natural b such that a  b  n(a) ).
Really, if a  b  n(a) , then a  b , then a    b . As b  n(a)  a    n(a) . If   e , then we
have a  e  n(a)  n(a)  n(a) , therefore   e . As   e , then, as we showed above, there must
be   e   , then a    n(a)  a  (e   )  n(a)  (a  e)    n(a)  n(a)    n(a)
and we have a contradiction, because n(a)    n(a) . So, finally, there is no any b such that
a  b  n(a) and a, n(a) are neighbors.

Def. (, ) is some ordered set. Elements a, c   are called neighbors if there is no any b  Ω,
such that a  b  c , or c  b  a . The element a  is called the first if a  b for any other

b  Ω. The element a   is called the last if b  a for any other b  Ω.

Obviously 14 is no last number in  .

e is the first natural number. And there
 Natural Numbers

Subtraction ""
Def. Let’s consider the set  of all pairs (a, b) where a  b. Subtraction "" is the mapping

   , which for every pair (a, b)   compares the natural number c such that a  b  c .
We will denote c  a  b , and we will say that c is a “difference of a and b ”.
So, by definition, for any a  b we have a  b  (a  b) . Addition "" is commutative, therefore
we can write b  (a  b)  (a  b)  b  a for any a  b .

Assertion2. For any a  b , the difference a  b is uniquely defined.

Proof. Let’s notice that if c   , then b  c  b   . Really, c    c     , then
b  c  b  (  )  (b  )    b   . Let’s assume that a  b  c and a  b   , then,
according to the definition of "" , there must be a  b  c and a  b   , then b  c  b   .
If с   , then b  c  b   , if с   , then b  c  b   . Therefore c   .

Exercise. For any a, b, c, d   (where a  b and c  d ) the next formulas are true:
[A] a  a  b , [B] a  b  c  d  a  d  b  c , [C] (a  b)  (c  d )  (a  c)  (b  d ) ,
[D] (a  b)  (c  d )  (a  d )  (b  c) (in this property there is a requirement (a  b)  (c  d ) ),
[E] a  b  m  a  m  b (for any m  a, m  b )  a  c  b  c (for any c  a, c  b ),
[F] a  b  m  a  m  b (for any m  a, m  b )  a  c  b  c (for any c  a, c  b ).

Def. Let  is an ordered set and X is a subset of  . The element M max   is called the
greatest element of X if for any x   : x  M max . The element mmin   is called the least
element of X if for any x   : x  mmin .
Theorem1. Any nonempty subset    of natural numbers has the least number mmin .
If there exist some number M such that x  M || x   , then X also has the greatest
number M max .

Proof. Let’s fix an arbitrary subset  of natural numbers. If e   , then e is the least number
of . Let e   . Let’s consider the set L  {a || a  x, x  }, obviously e L .
There exist some natural number  such that   L , but n()  L (if there is no such number,
then, by the induction axiom, L   and  is empty).
Let’s show that n(∆) is the least number in L.
[A] Let’s show that n()  . We consider the set G  {a || a  x, x  }. Then the set of natural
numbers N is divided into the sets L, , G and these sets do not have any common elements.
We know that n()  L , then n()   , or n()  G .
15
 Natural Numbers

If G is an empty set, then n()   , and we go to the step [B].

If G is not an empty set and n()  G , then:  is less than any number of  and n() is greater
than any number of  . Then there exist some ~
x   such that   ~
x  n() which is impossible,
because , n() are neighbors. So, in any case n()  G , then there must be n()  .

[B] Let’s show that n()  x for any x   . If it’s not true, then there exist some ~
x   such that
~
x  n() . As  is less than any number of  , then again   ~
x  n() , which is impossible.
Therefore n()  x, x   . From [A] and [B] follows that n() is the least number of  .

Let there exist some natural number M such that x  M || x   .

Let’s consider again the set G  {a || a  x, x  } - this set is not empty, because it contains M .

Then G has the least number g min . Obviously g min  e (really, if g min  e , then  must be empty).

As g min  e , then g min  e and n( g min  e)  g min .

[C] Let’s show that ( gmin  e)   . As we have shown above, the set  is divided into the sets
L, , G without common elements. For sure ( g min  e)  G (because if ( g min  e)  G , then g min
is not the least number in G ). If L is an empty set, then ( gmin  e)   and we go to the step [D].
Let L is not empty, let’s assume that ( g min  e)  L . Then for every x   there must be
( g min  e)  x  gmin which is impossible, because ( g min  e) and g min are neighbors.
Therefore, in any case ( g min  e)  L  ( g min  e)   .
[D] Let’s show that ( g min  e)  x for any x   . If it is not true, then there exist ~
x   such that
( g  e)  ~
min x . Then ( g  e)  ~
min x  g which is impossible, because ( g  e) and g are
min min min

neighbors. Then ( g min  e)  x for any x   . From [C] and [D] follows that ( g min  e) is the
greatest number of  .
Any set of natural numbers can be written as
  {e, n(e), n(n(e)), n(n(n(e))), n(n(n(n(e)))) ......} , this writing is very voluminous,
that’s why we denote: e  1, n(e)  2, n(n(e))  3, .... n(n(n(n(n(n(n(n(n(e)))))))))  10 .
Then   {1, 2, 3, ...., 10, n(n(n(n(n(n(n(n(n(n(1)))))))))) .......} [N]. We will show later
the possibility of a decimal notation: any natural number can be written in a unique way as
a combination of symbols 0,1,2...9 . Now we do not have sufficient instruments to prove it.
For example, we do not have a zero number 0 in  , this number will appear only when we build
the ring of integer numbers. We could add the zero number now, but then we would have to go
through all the previous theory and check that all the properties and assertions are still true for
the new set   {0} . Such approach is very inconvenient. Also, the proof of the possibility of a
decimal notation requires some facts from the divisibility theory, and this theory can be build for
integer numbers. That’s why we will use the notation [N] of natural numbers for a while.
16
 Natural Numbers
[Ordinary induction].  is any assertion which makes a claim that is based on k, where k is
some natural number. If [ step A] and [ step B] are true, then  is true for every natural number k
[ step A]  is true in the case when k  1,
[ step B] If  is true for some number k, then  must be true for the next number n(k ) .

Proof.  is our assertion and both [ step A] and [ step B] are true.
Let’s denote the set   {all natural numbers k , for which  is true}. Let’s show that   ,
by using the induction axiom. From here will follow that  is true for every natural number k.

1  because of the [ step A] . If some natural number k belongs to , then  is true for this
number k. Because of the [ step B] , the assertion  must be true for the next natural number n(k )
and therefore n(k )   . Then: if some k   , then n(k )   .
According to the induction axiom,    .

Def. For any natural number k, the set {a   || 1  a  k} is called a segment of natural nubers,
and we denote it like [1, k ] .
The set of natural numbers can be written as
  {1, 2, 3, ...., 10, n(n(n(n(n(n(n(n(n(n(1)))))))))) .......} [N]. From [N] follows that all the
numbers of any segment [1, k ] go one after another in [N], starting from 1 and up to k .
And therefore, we can agree to use the similar notation for any segment [1, k ] . So,
[1,1]  {1} , [1,2]  {1,2} , [1,3]  {1,2,3} ... [1,8]  {1,2,3,4,5,6,7,8}.
Segments [1, k ] and [1, m] are equal (equal as sets)  k  m .
Reminder: set’s  and  are called equivalent    if there exist one-to-one mapping f :   

Def. A is a set of any kind. If there exist some segment of natural numbers [1, k ] such that
  [1, k ] , then we say “A is a finite set” and “there are k elements in A”.
And we write |  | k .
Assertion1.The given definition is correct. If some set  is equivalent to some segment of natural
numbers, then that segment is unique (there can’t be   [1, k ] and   [1, m] , where [1, k ]  [1, m] ).

Assertion  (auxiliary). For any k   : if [1, k ]  [1, m] , then [1, k ] is not equivalent to [1, m] .
Proof. By ordinary induction. [ step A] “  is true in the case when k  1”.

We have to show: if [1,1]  [1, m] , then these segments are not equivalent. As [1,1]  [1, m] ,
then m  1, then m  1 , then the set [1, m] contains 1 and 2 (and maybe some other numbers).
Let’s assume that the segments are equivalent, then [1,1]  [1, m] , then there exist one-to-one
mapping f : [1,1]  [1, m]  f : {1}  [1, m] . If f (1)  1, then 2 does not have any pre-image in {1} .

17
 Natural Numbers

If f (1)  2 , then 1 does not have any pre-image in {1} . If f (1)  1 and f (1)  2 , then both 1 and
2 do not have any pre-image in {1} . Therefore f is not one-to-one, and we have a contradiction.
So,  is true when k  1.
[ step B] if  is true for some natural number k. So, for [1, k ] we have: if [1, k ]  [1, m] ,
then these segments are not equivalent. We have to show that the same is true for n(k ) .
Let’s assume the contrary: [1, n(k )]  [1, w] , but these segments are equivalent. Then there exist
one-to-one mapping f : [1, n(k )]  [1, w] . Here w  1. Really, if w  1, then [1, n(k )]  [1,1] which is
impossible, because n(k )  1  [1, n(k )]  [1,1] , then these segments can’t be equivalent ( [step A] ).
So w  1.

f : [1, n(k )]  [1, w] is one-to-one

mapping. Then for some
a  [1, n(k )]  f (a)  w and for some
pict.6
b  [1, w]  f (n(k ))  b [pict6].
Then we will slightly redefine f : [earlier] f (a)  w and f (n(k ))  b [and now] f (a)  b and
f (n(k ))  w . And f is still the same on any other element of the set [1, n(k )] . So f is still
one-to-one mapping f : [1, n(k )]  [1, w] , but now it maps the end of the segment into the end of
the segment: f (n(k ))  w . Then we can discard the elements n(k ), w from [1, n(k )], [1, w] and
f still will be one-to-one mapping f : [1, k ]  [1, (w  1)]. Let’s notice that [1, k ]  [1, (w  1)] .
Really, if [1, k ]  [1, (w  1)], then [1, n(k )]  [1, w] , but we have assumed that [1, n(k )]  [1, w] .
Then [1, k ]  [1, (w  1)] and there is one-to-one mapping f : [1, k ]  [1, (w  1)], it contradicts to
our initial assumption (at the beginning of the [ step B] ). Therefore, if [1, n(k )]  [1, w] , then these
segments are not equivalent. And  is true for n(k ) .
The steps [ step A] and [ step B] of the ordinary induction are completed.
Then the assertion  is true for every natural number k.
Let’s prove now the assertion1. We assume the contrary:   [1, k ] and   [1, m] where
[1, k ]  [1, m] . Then, by transitivity, [1, k ]  [1, m] and [1, k ]  [1, m] which is impossible, according
to the assertion  . Therefore, if   [1, k ] , then the segment [1, k ] is unique, there is no any other
segment of natural numbers that is also equivalent to [1, k ] . Everything is proved.

Def. A is a set of any kind. If A is an empty set, then it is a finite set by definition.
If A is not an empty set and A is not equivalent to any segment [1, k ] , then we say
“A is an infinite set ” or “there are infinitely many elements in A”.

According to the previous definition, any set A is either finite or infinite and no other variants.
18
 Natural Numbers

Assertion2.  is a finite set and   k  1. Then for any a   we have  \ {a}  (k  1) .

Proof. As   k there exist one-to-one mapping f :   [1, k ] .

Let’s fix some a   , if f (a)  k , then we discard the elements a, k from ,[1, k ] , and f will
become one-to-one mapping f :  \ {a}  [1, (k  1)], then  \ {a}  (k  1) .
If f (a)  k , then we can slightly redefine f (as we did above) such that f (a)  k , and we will
get again  \ {a}  (k  1) .

Assertion3.  is a finite set, then any subset    ||    is also a finite set and    .

Proof. Let   k , we can reformulate our assertion:

~
(Assertion  )   k , then for any    ||    there must be   k . We prove it by induction.

[ step A] Let k  1, then   1. Then there is only one element a in  , so   {a} .

~
If    ||    , then    . Then  is a finite set (by definition). And  is true when k  1.
~
[ step ] Let  is true for some natural number k , i.e., if   k , then for any    ||    we

have   k . Let’s consider the case   n(k ) . According to the assertion2, for any element a  

there must be  \ a  k . Let’s fix now any subset    ||    . As    there exist the
element a   such that a   . Then    \ {a} .

If    \ {a}, then    \ {a}  k  n(k )    n(k ) .

If    \ {a} (and    \ {a}), then, according to our assumption,(at the beginning of the [ step ] )

there must be    \ {a}  k  n(k )    n(k ) . So the steps [ step A] and [ step ] are
~
completed, then our assertion  is true for every natural number
k . And the assertion3 is proved.
Consequence1.  is a set. If there exist some subset
   ||   ,    [pict7], then  is an infinite set.
Proof. Let’s assume the contrary:  is a finite set.
Let’s notice that    . Really, if    , then        pict.7
and therefore    , which contradicts to the requirement
   . Then there exist some segment [1, k ] such that   [1, k ] . As    , then   [1, k ] , then
    k . We also have    ||    , then (assertion2)    , but     k and we have
a contradiction. Therefore  is an infinite set.

Consequence2. The set of natural numbers  is an infinite set.

19
 Natural Numbers

Proof. n :    \ {e} is one-to-one mapping. And the set  is equivalent to it’s subset  \ {e} .
According to the consequence1,  is an infinite set.

[Advanced induction].  is any assertion which makes a claim that is based on k, where k is
some natural number. If [ step A] and [ step B] are true, then  is true for every natural number k
[ step A]  is true in the case when k  1,
[ step B] If  is true for some natural numbers 1,2 ... k , then  must be true for the number n(k ) .

Proof.  is our assertion and both [ step A] and [ step B] are true.
Let  is the set of all natural numbers k, for which  is not true. If X is an empty set, then  is
true for every natural number k. Let    , then according to the theorem1, the set X has the
least number xmin   . Notice that xmin  1 (really, if xmin  1, then  is not true for the number 1 ,
but according to the [ step A] , the assertion  is true for 1 ).

So xmin is a minimal number for which  is not true. Then  is true for any natural number
a  xmin . So  is true for 1,2 .... ( xmin  1) , then, according to the [ step B] , the assertion  must be
true for the number n( xmin  1)  xmin , then xmin   , and we have a contradiction. Therefore    ,
and  is true for every natural number k.

The rules of calculation of an expression with some arrangement of brackets.

"":      is any binary operation (for any elements a, b   it compares some element a  b
of  ). We will call it an “addition”. Everywhere later a, b, c, d , e ... m are any elements of  .
Quite often we have to consider sums of elements of  . For example, the expression (a  b)  (c  d )
denotes the sum of elements (a  b) and (c  d ) , the expression (a  b  c)  d denotes the sum of
elements (a  b  c) and d . Are these sums equal? The answer is yes, if the operation "" is
associative. We will show it soon. Let’s define the basic rules, according to which we must calculate
any expression.

 Def . Any sum a  b  c  d  ...  m without any brackets must be calculated in the next way:
we take a and add b to it from the right side, then we get (a  b) , then we add c to it from the
right side. Then we get ((a  b)  c) , then we add d to it from the right side. Then we get
(((a  b)  c)  d ) and etc. In the end we will have the element ...(((a  b)  c)  d )  ...)  m ,
by definition, this element is equal to the sum a  b  c  d  ...  m .
So a  b  c  ...  m  / by def /  ...(((a  b)  c)  d )  ...)  m - this arrangement of brackets is

called the standard arrangement of brackets. We say that brackets are arranged here in the
standard way.

20
 Natural Numbers

When there is only one element a in the sum, the standard arrangement of brackets has no
brackets at all. Even for the sum a  b the are no brackets at all in the standard arrangement.
Only when we have the sum like: a  b  c , then the standard arrangement of brackets is
(a  b)  c . And for any sum a  b  c  d , the standard arrangement is ((a  b)  c)  d and etc.

In general a sum of several elements with some arrangement of brackets is called an “expression”.
The “expression” is some collection of summands a, b, c, d ..., signs "" and pairs of brackets [],(),{} .
Summands a, b, c, d ... are always elements of some fixed set  , where the operation "" is defined.

Pairs of brackets must be arranged reasonably. Any pairs [],() may be arranged only like this
..[...]... (...).. or like this (...[...] ...). The arrangement like ..[...(... ]...).. is forbidden.
If some brackets [] contain only one element d , then we can discard these brackets: [d ]  d .
And similarly, if some brackets [] contain the whole initial expression, we can also discard them.
If some brackets [] contain some other brackets () inside and there are no elements between them
[()], then we can discard the internal brackets: [()] [] .

Example. The next arrangements of brackets is admissible: a  (b  c  {d  e  [....]...}...).... m .

Any summands a and b of any expression must be considered as different elements. And therefore,
every summand in any expression is unique, and we can speak about a set of all summands (of any
expression). Similarly, we can speak about a set of all pairs of brackets [],(),{} (of any expression).

The main restriction. From now on we consider only such expressions, in which a set of all
summands and a set of all pairs of brackets are both finite sets. Moreover, we consider only
expressions with admissible arrangements of brackets that do not have any redundant brackets
that can be discarded.

If the set of all summands (of some expression) is equivalent to some segment [1, k ] of natural
numbers, we say that “the expression contains k elements ” or “the expression consists of k
elements” (or “there are k elements in the expression”) or just that “we have a finite expression”.

Any expression denotes a sum of several elements of  , which must be calculated in a certain way
(the process is described in the [ Step 1 ] and [ Step 2 ] below). In any case, for any expression,
the result of our calculations is always some element v   .

[ Step 1 ] In order to calculate an expression with some arrangement of brackets we always

“move” from the leftmost element to the right. We will go through some elements and left brackets
a  (b  c  {d  e  [.. until we meet any right bracket on our way: ],}, ) .
For example, we met a square bracket ] , it means that we have already passed [ . And now we
are exactly between these square-brackets [... here ...] (notice that there is no any other pair
of brackets {} inside [] , because ] is the first right bracket on our way).

21
 Natural Numbers

Now we need to calculate the expression [expression] in these brackets, by using the  Def from
above, the result is some element S . Then all the internal expression [expression] , together with
it’s outer brackets [...] must be replaced by S , so [...]replaced
 by
S .

[Step 2 ] Now we have the new expression a  (b  c  {d  e  S...} ...).... m .

We have to return to it’s leftmost element. And we do exactly the same thing that we described in
the [ Step 1 ] (we move from the leftmost element to the right until we meet any right bracket }
on our way). Every time when we complete the [ Step 1 ] we discard one element from the set
of pairs of brackets. The set  of all pairs brackets is a finite set, let it contains n   elements.

Then after n performances of the [ Step 1 ] we will get “the final” expression without any brackets
at all. If the initial expression contained k elements, then the final expression (for sure) contains
t  k elements. If the final expression consists of one element, then that element must be taken as a
result of our calculations. If the final expression consists of several elements, then their sum must
be calculated according to the  Def , and the result must be taken as a result of our calculations.
So, in the result we will always get some element v   , then, by definition, our initial expression
is equal to v.

Def. Let we have some expression a  (b  c  {d  e  [....]...}...)...  m with some arrangement

of brackets. When we make actions like [A] and [B] we say that we “rearrange brackets in the
expression”.
[A] We discard any pairs of brackets from our expression.
[B] We place some new pairs of brackets in our expression (after [A] or before [A]).
And we say that we make the standard rearrangement of brackets if we discard all the brackets
from the expression and we place the new brackets in the standard way:
(...(((a  b)  c)  d )  ...)  m .
If "":      is an associative operation. Then we can rearrange brackets in any expression
in any way we want, a new expression will be always equal to an initial expression. This assertion
follows from the main theorem2, but at first, let’s consider the next example.

Let’s look at the expression (a  b)  (c  d ) , we can consider (a  b) as a sum of elements a and

b and the sum (c  d ) as a separate element, then, by associativity:
(a  b)  (c  d )  a  (b  (c  d )) . Now we can apply the associative law to the sum
b  (c  d )  (b  c)  d , then (a  b)  (c  d )  a  ((b  c)  d ) .

We can consider the last expression as a sum of elements a, (b  c), d . Then, by associativity:
(a  b)  (c  d )  (a  (b  c))  d . We have deduced that the expressions (a  b)  (c  d ) and
(a  (b  c))  d are equal, despite the fact that they have different arrangements of brackets.
22
 Natural Numbers

Assertion4. Let we have an expression with k elements with some arrangement of brackets,
then any brackets {} which do not contain the whole expression, contain an expression with m  k
elements. The proof is obvious here.

The main theorem2 (the main property of any associative operation) .

 is a set. And "":      is any binary associative operation on  .
Then: Any finite expression (which consists of k elements) will not change after the standard
rearrangement of brackets.

Proof. Let’s use an advanced induction. [ step A]  is obviously true in the case when k  1.

In such case an expression consists of only one element a and there are no any brackets.
(There can’t be brackets (a) or [(a)] , because we have agreed not to consider expressions that have
any redundant brackets). And the standard arrangement of brackets for the expression a also has
no brackets at all, so we already have the standard arrangement.

[ step B] if  is true for some numbers 1,2 ... k . Then any expression with 1,2 ... k elements will
not change after the standard rearrangement of brackets.
Let’s take an arbitrary expression with n(k ) elements, let an (k ) is it’s last summand.
[First case] If there are no brackets on the right of an (k ) , then our expression is:
the expression on the left of an( k )  an( k ) .
[A] If the whole expression on the left of an (k ) is already captured by some brackets, then the whole
expression is the sum: [ the expression on the left of an( k ) ]  an( k ) . The expression on the left
of an (k ) consists of k elements, therefore, it will not change if we rearrange brackets in it in the
standard way, then the initial expression is equal to (...(((a  b)  c)  d )  ...)  m  an( k ) -here
we have the standard arrangement.
[B] If the expression on the left of an (k ) is not captured by any brackets, then there can be the next
cases:
[1-st case] The expression on the left of an (k ) consists of one element b , then b can’t be captured
by any brackets, so the total expression is b  an (k ) and there are no any brackets in the standard
arrangement for such sum and we already have the standard arrangement.
[2-nd case] The expression on the left of an (k ) consists of more than one element.
Then, according to the rules of calculation of an expression with brackets ( [ Step 1 ] and
[Step 2 ] ), the expression on the left of an (k ) must be calculated at first in any case, regardless of
the arrangement of brackets in this expression.

Then we can enclose this expression in the new brackets, because in any case it must be calculated
at first, so: the expression on the left of an( k )  an( k )  [the expression on the left of an( k ) ]  an( k ) .

23
 Natural Numbers

And again, the expression inside the square brackets consists of k elements and therefore
it will not change if we rearrange brackets in it in the standard way. Then our initial expression
 
is equal to (...(((a  b)  c)  d )  ...)  m  an ( k ) -here we have the standard arrangement.

[Second case] Let there is one or several brackets ], ),} on the right of an (k ) .
Let’s take the rightmost bracket } . The brackets {} can’t contain the whole expression
(if {} contain the whole expression, then these are redundant brackets). Then the initial expression
is a sum of two expressions: the expression which stays on the left of {} and the expression which
stays inside {} . According to the assertion4 , the expression inside {} consists of m  n(k ) elements.
And similarly, the expression on the left of {} consists of h  n(k ) elements. Then, according to our
assumption, (at the beginning of the [ step B] ) we can rearrange brackets in each of these
expressions in the standard way, each of these expressions will not change.
Then the initial expression is a sum of two expressions with standard arrangements of brackets:
~
(...(((a  b)  c)  ...)  v)  {(...((((q  w)  x)  ...  u)  an( k ) }  (S )  {(S )  an( k ) } 
  
 / associativ ity /  S   S  an( k ) .
~

  
The expression S   S consists of k elements and (according to our assumption) it will not
~

change if we rearrange brackets in it in the standard way:

S    S~  ...(((a  b)  c)  ...)  ....) ...  u  . Then the initial expression is equal to
...(((a  b)  c)  ...)  ....) ...  u   an(k ) -here we have the standard arrangement of brackets.
So the [ step A] and [ step B] (of induction) are completed. Therefore, our assertion  is true for any

expression with k elements. Everything is proved.

Conclusion. From the main theorem2 follows that: if the operation "":      is associative,
then in any given expression we can rearrange brackets in any way we want (we can choose the
arrangement of brackets which is more convenient for calculations). A new expression is always
equal to an initial expression.

Def. Let we have some finite expression with some arrangement of brackets. And a,b are some
summands. We replace a by b and by
a , and we do not change any other summands,
b
or signs "" , or brackets [],(),{} . Then we say that “we permute elements a,b ”.

Assertion5. If the operation "":      is not only associative, but also commutative,
then any finite expression will not change after permutation of any of it’s elements.

Lemma5. If "":      is associative and commutative, then in any expression we can

permute elements which stay near by, an expression will not change.
Proof. Let’s take any expression and remember the initial arrangement of brackets in it, we will
change it for a while, but then we will return back to it. Let’s assume that the elements a,b which
we want to permute, are neighbors, so we can place them in the round brackets () . All the elements

24
 Natural Numbers

which are on the left of a and b (if there are such elements) we place in the square brackets [] ,
and all the elements which are on the right of a and b (if there are such elements) we place in {} .
Then:
The initial expression  [c  d  ...  m]  (a  b)  {h  u  ..  x}  / /commutativity a  b  b  a // 
 [c  d  ...  m]  (b  a)  {h  u  ..  x}  // Let' s return the initial arrangement of brackets // 
 The expression we need . The last expression is exactly the initial expression, where a and b
are permuted.
Let’s prove now the assertion5. Let we have some expression where we need to permute some
elements a, b . If these elements stay near by, then, by lemma5, we can permute them and the
expression will not change.
Let elements a, b are not near by. Without loss of generality a is on the left of b.
Then we can permute a with it’s “right neighbors” until a and b become neighbors.
Then we permute a and b . Then we permute b with it’s “left neighbors” until b stays on the
initial position of a . So, by making several permutations of elements that stood near by, we have
permuted a and b . Notice that during these permutations we haven’t changed any brackets, or
signs "" in our expression, we just permuted some neighbor elements. Now we have the “final
expression” which is exactly what we need. Any permutation of neighbor elements gives an equal
expression (lemma5). Then initial and final expressions are equal. Everything is proved.

Conclusion. If "":      is associative and commutative, then in any expression we can

rearrange brackets in any way we want, and we can permute elements how we want, it will not
change an expression.

We have already proved that the addition "" of natural numbers is commutative and associative,
therefore, in any sum of natural numbers with any arrangement of brackets we can rearrange
brackets how we want and permute numbers how we want, a sum will not change.
Therefore, if we have sums of natural numbers and these sums consist of the same numbers,
then these sums are equal.

For example: ((a  b)  c)  d  (e  m)  b  (c  a)  (m  d )  e for any natural numbers

a, b, c, d , e, m .

Reminder. Ordered sets ( , ) and (, ~

) are called isomorphic if there exist one-to-one f :   
such that a  b (in  )  f (a) ~ f (b) (in  ).

Theorem3. Any sets   and   of natural numbers are isomorphic (isomorphic as ordered sets).
Proof. Let   and   are our sets. We define 1 
f
1 and if a 
f
a , then
n(a ) 
f
n(a ) . In other words, by definition: f (1 )  1 and if f (a )  a , then
f n(a )  n(a ) . We can see that f is a rule that for some elements of   compares some
elements of  .

25
 Natural Numbers

[Step1] Let’s show that f is defined on all   , we need to show that for every a   the element
f (a ) exists, and it is some element of   .

Let   is the set of all elements a of   , for every of which the element f (a ) exists.
[A] 1    because f (1 )  1 , [B] if some a    , then f (a )  a for some a    , from
the equality f (a )  a and the definition of f follows that f n(a )  n(a ) , so for n(a ) the
element f n(a )  does exist. Then n(a )    . And      . And now we can say that f is
a mapping from   to   .

[Step2] Let’s show that f fully covers   . Let   is a subset of   , the set   consists of all
elements a  which have some pre-image in   . [ A] 1    , because f (1 )  1 .
[B] if some a    , then f (a )  a for some a    , from here (according to the definition
of f ) immediately follows that f n(a )   n(a ) , and it means that n(a )    . So      ,
then f fully covers   .

[Step3] Let’s show that f doesn’t “glue together” elements of   . Let f (a )  a and f (b )  b
and a  b , let’s show that a  b . Let’s assume the contrary: a  b , it is equivalent to
f (a )  f (b ) . As a  b , then without loss of generality a  b . According to the main
theorem, the set   can be written as:

   {1 , n(1 ), n(n(1 )), n(n(n(1 ))), n(n(n(n(1 )))) ...} and a must be on the left of b in
this “row” (if a is on the right of b , then b  a , which contradicts to a  b ), then
   {1 , n(1 ), n(n(1 )), n(n(n(1 ))), a .... b .....}, then b  n(n(... n(a ))) , then
f (b )  f n(n(... n(a ))). From the definition of f it’s easy to see that it has the next property:
f (n(a ))  n( f (a )) for any a    , then
f n(n(... n(a )))  n f (n(... n(a )))  nn f (... n(a ))  n(n(... n( f (a ))) , then
f (b )  n(n(... n( f (a ))), then f (b )  f (a ) . It contradicts to the assumption f (a )  f (b ) ,
then f (a )  f (b ) .
From the [Step1]+[Step2]+[Step3] follows that f :      is one-to-one mapping.
[Step4] Let’s show that   is isomorphic to   ,i.e., a  b (in   )  f (a )  f (b ) (in   ) [V].
The proof of [V] is done in the [Step3].
Comment. It’s very easy to show that from [V] immediately follows the converse one:
f (a )  f (b ) (in   )  a  b (in   ). Let’s fix any f (a )  f (b ) . As a , b    , then exactly
one of the next cases is true: a  b , or a  b , or a  b . If a  b , then (acc. to [V] ) there must be
f (a )  f (b ) , which is not true. If a  b , then there must be f (a )  f (b ) , which is not true.
Then the only possible variant is a  b . And similarly for any other ordered sets.

26
 Natural Numbers

Multiplication
Def. Multiplication "" of natural numbers is the binary operation      such that:
[A] a  1  a a   , [B] a  n(b)  a  b  a a, b   .
Note: [B] can be rewritten as a  (b  1)  a  b  a a, b   .

Let’s show that this definition is correct, i.e., there exist the unique operation with these properties.
Existence. Let’s take 1 and define the multiplication " 1" on the set of pairs {(1, b) || b  } ,
where 1 is fixed and b is an arbitrary natural number, so 1  b  / by def /  b .
1

We need to check that "  " is a multiplication in the case a  1. So 1  1  1 (property [A]) and
1 1

1 1 n(b)  n(b)  b  1  1 1 b  1 (property [B]).

Let  is the set of all natural numbers a , for every of which there exist it’s own multiplication
" " , which is defined on the set of all pairs {(a, b) || b  } and satisfies to [A] and [B].
a

We will show that    . [ A] 1  (as were shown above) [B] If some a   , then it has it’s own
multiplication "  " , which is defined on the set of all pairs {(a, b) || b  } such that [A] and [B].
a

Let’s take the next element n(a) and define for it it’s own multiplication " 
n (a )
" on the set
{(n(a), b) || b  }. We define: n(a) n ( a ) b  / by def /  a a b  b .
Then n(a)  1  a a 1  1  a  1  n(a) (property [A])
n(a)

n(a) n( a ) n(b)  a a n(b)  n(b)  (a a b  a)  n(b)  a a b  a  n(b) 

 a a b  n(a  b)  a a b  n(b  a)  a a b  (b  n(a))  (a a b  b)  n(a)  n(a) n( a ) b  n(a)
(property [B]). By the induction axiom    , so for any a there exists it’s own multiplication " a " .
Let’s define the ordinary multiplication "" on every pair ( a, b) .
By definition a  b  / by def /  a 
a
b . Then "" satisfies to [A] and [B]. The existence is proved.

Uniqueness. Let’s show that the operation "" is unique. We assume that there exist some other
binary operation :      with the properties [A] and [B]. We will show that
a  b  a  b a, b   , then "" "" . Let’s fix an arbitrary a   . And  -is the set of all b such
that a  b  a  b .
[ A] 1   . Really, according to [A], a 1  a and a 1  a , therefore a 1  a 1.
[B] If some b   , then a  b  a  b . Let’s consider
a  n(b)  a  b  b  // as a  b  a  b //  a  b  b  a  n(b) , therefore n(b)   .
By the induction axiom    and therefore a b  a b for any b   , where a is an arbitrary
fixed number, then a b  a b for any a, b   .
Before we prove the basic properties of multiplication we need an auxiliary lemma.

27
 Natural Numbers

Lemma6. Right distributivity (a  b)  c  a  c  b  c . Let’s fix arbitrary numbers a, b   .

And  is the set of all c , for which the formula (a  b)  c  a  c  b  c is true.
[ A] obviously 1   [B] If some с   , then (a  b)  c  a  c  b  c .
Let’s consider (a  b)  n(c)  // property [ B] // 
 (a  b)  c  (a  b)  (a  c  b  c)  (a  b)  (a  c  a)  (b  c  b)  // property [ B] // 
 a  n(c)  b  n(c) , then n(с)   , so    . As a, b are arbitrary fixed natural numbers, then
the right distributivity is true for any a, b, c   .

Commutativity. a  b  b  a for any a, b   .

Step one. Commutativity for one. Let b  1 . Let’s show that a  1  1  a for any a   . Let  is
the set of all numbers a , for which a 1  1 a . [A] 1  , because 1  1  1  1  1 (property [A]),
[B] If some a   , then a 1  1 a . Let’s consider
1  n(a)  1  a  1  a  1  1  / right distributivity /  (a  1)  1  n(a)  1, therefore n(a)   .
So    . And we are going to the step two.
Step Two. General commutativity. Let’s fix an arbitrary a   . Let  is the set of all b , for which
a  b  b  a . [A] 1   (step one), [B] If b   , then a  b  b  a . Let’s consider
a  n(b)  a  b  a  b  a  a  / right distributivity /  (b  1)  a  n(b)  a then n(b)   .
Then    . And everything is proved.
Consequences.
[A] General distributivity: (a  b)  c  c  (a  b)  a  c  b  c  c  a  c  b
(follows from commutativity and lemma6).

[B] Distributive law for subtraction: (a  b)  c  c  (a  b)  a  c  b  c  c  a  c  b for any

a, b, c   where a b.
Proof. Let a  b   , by the definition of subtraction it means that a  b   , then [a]
(a  b)  c    c and [b] a  c  b  c  (b  )  c  b  c  (b  c    c)  b  c    c .
From [a] and [b] we see that (a  b)  c  a  c  b  c . The rest follows from commutativity of "" .

Associativity. (a  b)  c  a  (b  c) for any a, b, c   . Let’s fix arbitrary a, b   . And  is

the set of allc , for which (a  b)  c  a  (b  c) . [ A] 1   . Really, (a  b) 1  a  b  a  (b 1) ,
[B] If c   , then (a  b)  c  a  (b  c) .

Let’s consider (a  b)  n(c)  / propery [ B] /  (a  b)  c  (a  b) 

 a  (b  c)  a  b  // left distributivity //  a  (b  c  b)  // left distributivity // 
 a  (b  (c  1))  a  (b  n(c)) . Then n(c)   . Then    .

28
 Natural Numbers

Exercise1. Prove the next basic properties (If there is a difference a  b or c  d , then a  b and
c  d ) [A] (a  b)  (c  d )  (a  c  a  d )  (b  c  b  d ) and
(a  b)  (c  d )  (a  c  a  d )  (b  c  b  d ) and (a  b)  (c  d )  (a  c  a  d )  (b  c  b  d )
ac  bc  a  b ac  bc  a  b
[B] and .
ac bc  a b ac bc  a b
[C] a  b, c  d  a  c  b  d , a  c  b  d .

Let’s show for example that (a  b)  (c  d )  (a  c  a  d )  (b  c  b  d ) . Let’s take the expression

(a  b)  (c  d ) . We can consider (a  b) as a sum of two numbers and (c  d ) as a separate

number. Then (a  b)  (c  d )  a  (c  d )  b  (c  d )  (a  c  a  d )  (b  c  b  d ) .

Let’s show (a  b)  (c  d )  (a  c  a  d )  (b  c  b  d ) . Let a  b   and c  d   , then

a  b   and c  d   . So the left part equals (a  b)  (c  d )     .

The right part is
(a  c  a  d )  (b  c  b  d )  a  (с  d )  b  (с  d )  a     b     (b  )     b    
 (b  )  b       . Then the left and right sides are equal.
Let’s show the last one: a  b, c  d  a  b  , c  d   .

Then a  c  (b  )  (d   )  (b  d  b   )  (  d     )  b  d  b    (  d     )  b  d .
The part a  c  b  d is proved. Let’s show now that a  c  b  d .
So a  c  (b  )  (d   )  (b  d )  (   )  (b  d ) .

Archimedes axiom. For any pair of natural numbers a, b   there exist n   such that
n  a  b (this property is called an Archimedes axiom).
Proof. Let’s fix an arbitrary pair of natural numbers a, b . Let’s take n  b  1 , then
(b  e)  a  b  a  a  / as a  1/  b  1  1  b  1  b .
Notice that when we say “axiom” we usually imply some fact which must be accepted without any
proofs. But for the natural numbers the Archimedes axiom is not an axiom at all, it is just one of the
properties of natural numbers. So now, the “Archimedes axiom” is just a name of the certain
property. Later, during the length construction, the Archimedes axiom will be the real axiom, which
will be accepted without any proofs.

29
 Natural Numbers

Division ":"

Def. Let’s consider the set  of all pairs (a, b) , where a can be represented as a  k  b for
some natural number k. Division ":" is the mapping    that for every pair (a, b)  
compares the natural number k such that a  k  b , for every such pair, k is called a quotient of
a
numbers a, b , and we write k  a / b  .
b
The given definition is correct: if k  a / b then there is no other number   k such that   a / b .
Really, let there exist some other  . Without loss of generality   k . As   a / b  a    b ,
as k  a / b  a  k  b , then   b  k  b , but   k , then there must be   b  k  b , we have a
contradiction. So   k .

Exercise2. Prove next simple properties: If there is a / b or c / d , then a  k  b (for some k   )

a c a c
and c    d (for some    ). If there is , then  m  (for some m   ).
b d b d
a c ad  bc a c ad bc a c ac a c ad
[A]   and   , [B]   and  ,
b d bd b d bd b d bd b d bc

Exercise3.
a c a b c c a
[A]   a  d  b  c , [B] a  b   and a  b   , [C] a  for any b   ,
b d c c a b b
a
if b  e , then a  .
b

Advanced part.
Notice that there exist different sets of natural numbers (it’s enough to change the designation of
any symbol in  , in order to turn it into a new set. For example, we can change the designation
1   and do not change any other numbers, then we will have the new set of natural numbers,
where the symbol  is an element “one”). Addition and multiplication of natural numbers are
always binary operations which are defined on a concrete set of natural numbers.
Let   is a set of natural numbers, we can write    (   ,,) , implying that we consider the set
  together with the addition and multiplication which are defined on it.
And similarly,    (   ,,) . Notice that we use different symbols to define operations on   and
  . Of course, the addition  on   has the same properties as the addition  on   , but  and
 are different operations, because they are defined on different sets.
The same is true for multiplications  ,  .

30
 Natural Numbers

Advanced Theorem3. For any sets (   ,,) and (   ,,) of natural numbers there exist the
unique isomorphism f :      (where f is an isomorphism of ordered sets).
Moreover, f has the next properties: [T]
a  b  c (in   )  f (a )  f (b )  f (c ) (in  ) and
a  b  d  (in   )  f (a )  f (b )  f (d  ) (in   ) .

Proof. The existence of f and even the explicit definition of f are provided in the theorem3.
Let’s assume that there exist some other isomorphism  :      (where  is also an
isomorphism of ordered sets, so there must be: [A]  is one-to-one and
[B] a  b (in   )  f (a)  f (b) (in  ) ). Let’s show that f (a)   (a) a    , then f  
and f is unique.

The main auxiliary lemma.  has exactly the same properties as f : it transfers one into one
 (1 )  1 and if  (a )  a , then  (n(a ))  n(a ) .
Proof. Let’s show that  (1 )  1 . We consider  (1 ) , let  (1 )  1 ,
then  (1 )  y ,where y is some element of   . We know that
 is one-to-one, so the element 1    must have some preimage
x    :  ( x )  1 . Then we have two equalities:  (1 )  y and
 (x )  1 [pict8]. The one 1 is the least number in   , then pict.8
1  x . Then  (1 )   ( x ) (look now at our equalities), then
y  1 and we have a contradiction. Then  (1 )  1 .

Let’s show that the condition “if  (a )  a , then  (n(a ))  n(a ) ” [V] is true for any a   
Let   is the set of all a , for which [V] is true.
[ A] 1   . We have  (1 )  1 , let’s show that  (2 )  2 . We will show that  (2 ) can be
equal only to 2 . Really, 1  2   (1 )   (2 )  1   (2 ) . If the value  (2 )  2 , then
[ A] is proved. Let’s assume that  (2 )  y  2 , then, anyway there
must be some element x    such that  (x )  2 and
x  1 , x  2  x  2 . Then we have:  (2 )  y || y  2 and
 ( x )  2 || x  2 [pict9], here we have a contradiction, really
x  2 , then there must be  ( x )   (2 )  2  y , but y  2 .
So  (2 ) can be equal only to 2 , and [ A] is proved. pict.9

31
 Natural Numbers

[B] If a    , it means that we have  (a )  a and  (n(a ))  n(a ) . We need to show that
the similar condition is true for n(a ) . We already have  (n(a ))  n(a ) , then we need to show
that  (n[n(a )])  n[n(a )] . It’s not difficult to do, by using an auxiliary picture (like pictures 8,9).
Perform it as a simple exercise. After it’s done, by the induction axiom, [V] is proved.

Let’s prove now the auxiliary lemma. Let   is the set of all natural numbers a , for which
f (a )   (a ) .
[ A] 1   . Because  (1 )  1 and f (1 )  1 . [B] If a    then f (a )   (a ) .
We want to show that f (n(a ))   (n(a )) . From the basic properties of these isomorphisms we
have f (n(a ))  n( f (a )) and (n(a ))  n( (a )) . As the elements f (a ) and  (a ) are
equal, then the elements n( f (a )) and n( (a )) are also equal, therefore f (n(a ))   (n(a )) .
And n(a )    . So      and f   on   .

Let’s prove the properties [T]. These properties are obviously equivalent to
[Tp]: f (a  b )  f (a )  f (b ) and f (a  b )  f (a )  f (b ) for any a , b    .
Let’s prove the first one. We fix an arbitrary a    . Let   is the set of all numbers b , for
which f (a  b )  f (a )  f (b ) .
[ A] 1    . Really f (a  1 )  f (n(a ))  n( f (a ))  f (a )  1  f (a )  f (1 ) , so 1   .
[B] If b   , then f (a  b )  f (a )  f (b ) . Let’s consider f (a  n(b ))  f (n(a  b )) 
 n( f (a  b ))  n( f (a )  f (b ))  f (a )  n( f (b ))  f (a )  f (n(b )) (we used the basic
properties of the addition  on   and  on   ), so n(b )    , then      and the
property f (a  b )  f (a )  f (b ) is proved.

Let’s prove the second one: f (a  b )  f (a )  f (b ) . We fix again an arbitrary a    and  
is the set of all numbers b , for which f (a  b )  f (a )  f (b ) .
[ A] 1   . Really f (a  1 )  f (a )  f (a )  1  f (a )  f (1 ) .
[B] If b   , then f (a  b )  f (a )  f (b ) . Let’s consider
f (a  n(b ))  f (a  b  a )  f (a  b )  f (a )  f (a )  f (b )  f (a )  // distributi bity // 
 f (a )   f (b )  1   f (a )  ( f (b )  f (1 ))  f (a )  f (b  1 )  f (a )  f (n(b )) ,
then n(b )    and      . Everything is proved.

32
 Groups, Rings, Fields

Multiplicative groups
Def. G is a set with a binary operation "": G  G  G “multiplication”.
And [A] Multiplication is associative: (a  b)  c  a  (b  c) for any a, b, c  G .
[B] There exist the element “one” e such that a  e  e  a for any a  G .
[C] For any element a  G there exist the inverse element a 1  G such that a  a 1  a 1  a  e .
Then G is called a multiplicative group, or just a group.
As "" is associative (The main theorem2, page 23), then in any expression with any arrangement of
brackets we can rearrange brackets in any way we want, it will not change an expression.
For example: (a  c)  (d  (v  m))  (a  c  d )  v  m .
Property. For any element a  G , an inverse element a 1  G is always unique.
Let’s fix any a  G . According to [C], there exist at least one inverse element a 1 . If there is some
1
other inverse element b  G , then a  a  e and a  b  e , then a  a 1  a  b . Let’s multiply both
sides by a
1
from the left side, then a 1  (a  a 1 )  a 1  (a  b)  [associativ ity ]  (a 1  a)  a 1 
 (a 1  a)  b  e  a 1  e  b  a 1  b , it proves the uniqueness.
Def: a subset   G is called a subgroup of G if H is closed under multiplication and H is
a group. When we say “  is closed under multiplication”, we mean that for any a, b   there
must be a, b   . And when we say “  is a group”, we mean that the properties [A],[B],[C] are
true for any elements of  .
Subgroup criterion.   G is a subgroup  for any a, b   there is a  b   and a 1   .
 if  is a subgroup of G, then obviously a b   and a 1   .
Proof.
 From a  b   follows that  is closed under multiplication. Let’s take any a   ,
then a   , then [C] is true for  . And a  a 1  e   , so [B] is true for  .
1

Then  is a subgroup of G . (The property [A] is always true for any elements of G ,
so there is nothing to check).
From this criterion immediately follows that any subgroup   G must contain an element one "e" ,
moreover, the subgroup which consists of only one element {e} is a minimal subgroup of G , and the
subgroup   G is a maximal subgroup of G .

A group G is called commutative if the operation "": G  G  G is commutative: a  b  b  a for

any a, b  G . In any commutative group we can speak about division " /" .

Def. G is a commutative multiplicative group.

Division" /" is the binary operation " /": G  G  G that for any pair a, b  G compares the
a
element x  (a / b)  such that a  x  b  b  x .
b

34
 Groups, Rings, Fields

The given definition is correct. For any a, b  G there exist the unique element x such that
a  x b  b x .
1
Let’s consider a  x  b as an equation, we multiply both sides by b from the right side:
a  b1  ( x  b)  b1  / associativity /  a  b1  x  (b  b1 )  x  a  b1 , and it’s easy to check
that x really satisfies to a  x  b  b  x . The existence is proved, the uniqueness is obvious.

Exercise1. G is an multiplicative commutative group, then the next properties are true for any
a e e a a b
elements of  e and a   e and a   and   e and a  b  c  b  a  c .
G [A]
a a b b b a
a c ac a с ad
[B]   and  .
b d bd b d bc

Additive groups (everything is very similar here).

Def. G is a set with a binary operation "": G  G  G “addition”.
And [A] Addition is associative: (a  b)  c  a  (b  c) for any a, b, c  G .
[B] There exist the element “zero” 0 such that a  0  0  a for any a  G .
[C] For any element a  G there exist the inverse element  a  G such that
a  ( a )  ( a )  a  0 .
Then G is called an additive group, or just a group.
Def: a subset H  G is called a subgroup of G if H is closed under addition and H is a group.
I.e.: for any a, b  H there must be (a  b)  H . And the properties [A],[B],[C] are true for any
elements of H.
Subgroup criterion. H  G is a subgroup  for any a, b  H there is a  b  H and  a  H .

In any commutative additive group G we can speak about subtraction "": G  G  G .

For any a, b  G , the element x  (a  b) is such element that a  (a  b)  b  b  (a  b) .
Exercise2: G is an additive commutative group, then the next properties are true for any elements
of G (compare with the exercise1 ):
[A] a  a  0 and a  (a)  0 and a  (b)  a  b and (a  b)  (b  a)  0
and a  b  c  b  a  c,
[B] (a  b)  (c  d )  (a  c)  (b  d ) and (a  b)  (c  d )  (a  d )  (b  c) .
It’s easy to notice that there is no essential difference between additive and multiplicative groups.
In general, a group G with a binary operation  is denoted like (G,)

Def. We say that (G,) is isomorphic to ( D,) if there exist some one-to-one mapping
f : G  D such that: if a  b  c (in G) , then f (a)  f (b)  f (c) (in D) .

35
 Groups, Rings, Fields

Property. If (G,) is isomorphic to ( D,) , then ( D,) is isomorphic to (G,) .

Proof. Let there exist one-to-one f : G  D with such property. Let’s consider the inverse mapping
f 1 : D  G . We need to show that if     (in D) , then f 1 ( )  f 1 ( )  f 1 (  ) (in G) .

Every element D can be uniquely represented as   f (a) for some a  G , then the equality
    is equivalent to f (a)  f (b)  f (c) , where a, b, c are some elements of G .
We want to show that f 1 ( )  f 1 ( )  f 1 ( ) and it is equivalent to
f 1  f (a)  f 1  f (b)  f 1  f (c)  a  b  c .

Let’s sum up: we need to show that from f (a)  f (b)  f (c) follows that ab  c .
Let’s assume the contrary a  b  c , then a  b  d (where d  c ). According to the main property
of f : if a  b  d , then f (a)  f (b)  f (d ) , therefore f (c)  f (d ) . As f is one-to-one, then
c  d , and we have a contradiction. Therefore, if f (a)  f (b)  f (c) , then a  b  c which is
equivalent to: if     (in D) , then f 1 ( )  f 1 ( )  f 1 ( ) (in D) . And f 1 is an
isomorphism.
From this property follows that if one group is isomorphic to the other, then we can just say
“groups are isomorphic”. Any two isomorphic groups are almost identical, there is one-to-one
correspondence between their elements, and if elements of one group are connected by some relation
(like one element is a product of others), then their images/preimages in the other group are
connected by exactly the same relation.

For practical purposes (when we want to show that two groups are isomorphic) it’s easier to use the
other, equivalent definition.

Def. (G,) is isomorphic to ( D,) if there exist one-to-one mapping f : G  D such that
f (a  b)  f (a)  f (b) for any a, b  G .
Show that this definition is equivalent to the initial one. From here follows (understand why) that
this definition is also symmetric: if (G,) is isomorphic to ( D,) (according to this definition),
then ( D,) is isomorphic to (G,) .

When groups are isomorphic we write (G,)  ( D,) , or just G  D .

Exercise3. f is an isomorphism of groups (G,) and ( D,) . Then “one always goes to one”
f (eG )  eD . And for any a  G we have f (a 1 )  f (a) 1 .
Solution. Any element d  D has the unique representation as d  f (a) || a  G , then for any
d  D we have d  f (eG )  f (a)  f (eG )  f (a  eG )  f (a)  d , so d  f (eG )  d for
any d  D and similarly f (eG )  d  d , then f (eG ) is a one of the group D , so f (eG )  eD .

Let’s fix now an arbitrary a  G . Let’s take f (a 1 ) and consider the product:
f (a 1 )  f (a)  f (a 1  a)  f (eG )  eD . By definition, f (a) 1 is such element that
36
 Groups, Rings, Fields

f (a 1 )  f (a)  eD , then f (a 1 )  f (a)  f (a)1  f (a) . Let’s multiply both sides of the last
equality by f (a) 1 from the right side, then f (a 1 )  f (a) 1 .

Def. G is a group. For any a  G and for any natural number n   let’s define the element:
na  a  a  ...  a (there are exactly n summands on the right side). If G is a multiplicative
group, then for any a  G and for any natural number n   we define a n  a  a  ... a
(there are exactly n factors on the right side).
Then in any commutative additive group, for any elements a, b and any natural numbers m, n :
[A] ma  na  (m  n)a [B] m(a  b)  ma  mb [C] mna  n(ma)  m(na) [D]  (ma)  m(a)
In any commutative multiplicative group, for any elements a, b and any natural numbers m, n :
[A] a m  a n  a m  n [B] (a  b) m  a m  b m [C] a mn  (a m ) n  (a n ) m [D] (a m ) 1  (a 1 ) m .

Def. R is a set with two binary operations on it: addition "" and multiplication "" .
[A] R is a commutative group with respect to addition "" ,
[B] Multiplication "" is associative. And there is a distributive law: (a  b)  c  a  c  b  c
and c  (a  b)  c  a  c  b for any a, b, c  R
Then R is called a ring.
Notice that multiplication mustn’t be commutative, there may no be the element “one” e  R such
that a  e  e  a  a (for any a  R ). If multiplication is commutative, we say “ R is a commutative
ring”. If there exist an element “one” e R , then we say “ R is a ring with one”.
And we can say for example: “ R is a commutative ring with one”. Any ring R is a commutative
group with respect to addition and therefore, subtraction is defined on R .

Property1. In any ring R the distributive law for subtraction is true:

a  (b  c)  a  b  a  c and (a  b)  c  a  c  b  c for any a, b, c  R .
Proof. Let’s show that a  (b  c)  a  b  a  c [a]. This equality is equivalent to
a  (b  c)  a  c  a  b , let’s simplify the left part:
a  (b  c)  a  c  / ordinary distributivity /  a  ((b  c)  c)  a  b and [a] is true.
The proof of the other formula is similar.
Consequence. In any ring R , for any element a  R we have a  0  0  a  0 .
Proof. Let’s fix an arbitrary a  R , then a  0  a  (a  a)  / property 1/  a  a  a  a  0
Exercise4. R is a ring. Then for any a, b, c, d  R :
[A] a  (b)  a  b and (a)  b  a  b and (a)  (b)  a  b ,
[B] (a  b)  (c  d )  (a  c  a  d )  (b  c  b  d ) and (a  b)  (c  d )  (a  c  a  d )  (b  c  b  d )
and (a  b)  (c  d )  (a  c  b  c)  (a  d  b  d ) and (a  b)  (c  d )  (a  c  b  d )  (a  d  b  c) ,

37
 Groups, Rings, Fields

[С] If R is a commutative ring, then the next well known formulas are true:
a 2  b 2  (a  b)  (a  b) and (a  b) 2  a 2  2  a  b  b 2 and
(a  b)3  a3  3  a  b2  3  b 2  a  b3 .

Def: a set   R is called a subring of R if  is closed with respect to addition "" and
multiplication "" and  is a ring.

When we say that  is closed with respect to "" and "" we mean that
a, b    a  b  , a  b   .
When we say that  is a ring, we mean that  satisfies to the ring-definition.
Subring criterion. R is a ring.
  R is a subring of R  a, b  H : a  b  , a  b  , a  b   .
Def: a ring F is called a field if all nonzero elements of F form a commutative group with respect
to multiplication "" .
Notice that any field is also a ring (by definition), but it is such ring, where all nonzero elements
form a commutative group with respect to multiplication.
Even if F is a commutative ring with one, we can’t assert that F is a field. There is no guarantee
1 1 1
that for any nonzero element a  F there exist an inverse element a  F : a  a  a  a  e .
The ring of integer numbers  (that we will build very soon) will be an exactly such type of ring:
a commutative ring with one. But  is not a field.

Def. F is a field. All elements of F form a commutative group with respect to addition "" , which
is called an additive group of F . All non-zero elements of F form a commutative group with
respect to multiplication "" , which is called a multiplicative group of F .
From this definition we see that multiplication is commutative on F \ {0} . But it’s very easy to
understand that multiplication is commutative on all F . Really, let’s consider the equality
a  b  b  a , if one of the elements a or b is zero, then both sides of the equality are zeroes and
therefore the equality is true for any a, b  F .

Def. Elements a, b of a ring/field are called zero divisors if a  b  0 , where a  0 and b  0 ,

or if b  a  0 , where b  0 and a  0 .
There is a very important fundamental difference between rings and fields. In general, there can be
some zero divisors in a ring, but there are no divisors in any field.
Let a  b  0 in some field F , where a  0 and b  0 . Let’s multiply both sides by a 1 from the left
side, so (a  b)  0  a 1  (a  b)  a 1  0  (a 1  a)  b  0  b  0 - it contradicts to b  0 .
So, there are no divisors of zero in any field F.
Before we start working with fields we have to make one very important restriction.
As non-zero elements of F form a commutative group with respect to multiplication, division ":"
38
 Groups, Rings, Fields

0
is defined on F \ {0}. Let’s “extend” this operation to F . For any a  0 we define  0 , it is
a
reasonable, because 0  a  0  a  0.
a
And our restriction is: for any a  F , the element is not defined. In other words:
0
a
It is forbidden to divide by zero. Really, let a  0 if we define  b , then it’s reasonable to require:
0
a  b  0  0  b  0 , then a  0 , but a  0 , and we have the contradiction. Therefore, we make an
a 0
agreement that is not defined for any a  F (even for a  0 , the element is not defined).
0 0
a c
Exercise5. R is a field. Then for any a, b, c, d  R (if there is or , then b  0 and d  0 ).
b d
a a a a a a a c ad  bc a c ad bc
[A]  and  and  [B]   and   .
b b b b b b b d bd b d bd
This theory will give us a great advantage in the future. Suppose that we deduced that some
structure (some set with "" "" ) is a ring, then all the properties from the exercises 2,4 are true for
the elements of that set. Similarly, if we deduced that some structure is a field, then all the
properties from the exercises 1,2,4,5 are true for the elements of that set.

Def: a set   F is called a subfield of F if  is closed with respect to addition ""

and multiplication "" and  is a field.

Subfield criterion. F is a field.   F is a subfield of F 

a
a, b   : a  b  , a  b  , a  b  ,  .
b
Notice that “addition” and “multiplication” are just conditional names of some binary operations
R  R  R . A ring/field R with two binary operations on it can be denoted as ( R , , ) .
For different rings/fields these operations may have absolutely different meaning.

Def. Let ( R , , ) and ( D , , ) are two rings (fields). We say that R is isomorphic to D
if there exist one-to-one mapping f : R  D such that:
[A] If a  b  c (in R) , then f (a)  f (b)  f (c) (in D) .
[B] If a  b  d (in R) , then f (a)  f (b)  f (d ) (in D) .

Property. If R is isomorphic to D, then D is isomorphic to R (the proof here is very similar to

the proof that we made for groups).
And again, from this definition follows that if one ring/field is isomorphic to the other, then we can
just say “these rings/fields are isomorphic”.

39
 Groups, Rings, Fields

In practice we will use the next equivalent definition.

Def. Let ( R , , ) and ( D , , ) are two rings (fields). R is isomorphic to D if there exist
one-to-one mapping f : R  D such that f (a  b)  f (a)  f (b) and f (a  b)  f (a)  f (b)
for any a, b  R .
Isomorphic rings/fields are almost identical, there is one-to-one correspondence between their
elements and if some elements in one ring are connected by some relation (like one element is
a sum/product of others), then their images/preimages in the other ring are connected by exactly
the same relation. When rings/fields are isomorphic, we write ( R , , )  ( D , , ) ,
or just R  D . The next theorem will be repeatedly used during the construction process.

Theorem4. ( R , , ) is a ring/field and [ D , , ] is some set with two (binary) operations on

it: “addition”   (there are just some binary operations on D, we do not
and “multiplication”
even need to know anything about their properties). If there exist one-to-one mapping f : R  D
such that f (a  b)  f (a)  f (b) and f (a  b)  f (a)  f (b) for any a, b  R , THEN
[ D , , ] is also a ring/field.
But at first we need to prove the next auxiliary lemma.

Lemma6. (G ,) is a multiplicative group and [ D ,] is some set with a binary operation
"": D  D  D on it. And there exist one-to-one mapping f : G  D such that
f (a  b)  f (a)  f (b) for any a, b  G . Then D is also a group.
And one of G goes into one of D: f (eG )  eD and for any a  G : f (a 1 )  f (a) 1 .
Moreover, if G is commutative then D is also commutative.
~, b~, c~ are any elements of D, each of these elements has can be uniquely represented
Proof. Let a
~  f (a), b~  f (b), c~  f (c) where a, b, c  G . Then
as a
~
(a~  b )  c~   f (a)  f (b)  f (c)  f (a  b)  f (c) 
~
 f (a  b)  c   f a  (b  c)  f (a)  f (b  c)  f (a)   f (b)  f (c)  a~  (b  c~) , then the
associative law is true in [ D ,] . Let’s show that f (e ) is a “one” of D. For any a ~  D let’s
G

consider a~  f (eG )  f (a)  f (eG )  f (a  eG )  f (a)  a~ and similarly f (eG )  a~  a~ ,

therefore f (eG )  eD . Let’s finally show that for any element of D there exist an inverse one.
We fix an arbitrary a ~  D , then a~  f (a) . Let’s show that the element f (a 1 ) is an inverse to a~ .
~  f (a 1 )  f (a)  f (a 1 )  f (a  a 1 )  f (e )  e and similarly f (a 1 )  a~  e .
So a G D D

Now we can say that D is a group. We have already shown that f (eG )  eD , and by the way,

40
 Groups, Rings, Fields

in the process we have also shown that f (a 1 )  f (a)1 . Really, the element f (a 1 ) is
an inverse to a~  f (a) , it means exactly that f (a 1 )  f (a) 1 .
~
Let now G is a commutative group. Let’s fix arbitrary elements a~, b  D , then
~ ~
a~  b  f (a)  f (b)  f (a  b)  f (b  a)  f (b)  f (a)  b  a~ , then D is also a commutative
group.

We can rewrite this lemma for additive groups: (G ,) is an additive group and [ D ,] is some set
with a binary operation "": D  D  D on it. If there exist one-to-one mapping f : G  D such
that f (a  b)  f (a)  f (b) for any a, b  G , then D is also a group.

And zero of G goes into zero of D : f (0G )  0D and for any a  G : f (a)   f (a) .
Moreover, if G is commutative, then D is also commutative.

Let’s prove now the theorem4: Let f : R  D is one-to-one correspondence with the properties
f (a  b)  f (a)  f (b) and f (a  b)  f (a)  f (b) for any a, b  R . Let R is a ring, then ( R,)
is a commutative group with respect to the addition "" , and ( D,) is a set with the binary
operation  , as f : R  D is one-to-one mapping and f (a  b)  f (a)  f (b) , then (by lemma6)
( D,) is a commutative group with respect to  . As f is one-to-one and f (a  b)  f (a)  f (b) ,
then f conserves the associative law for multiplication: a ~, b~, c~  D : (a~  b~)  c~  a~  (b~  c~)
(the proof is exactly the same as in the lemma6). And let’s finally show that the distributive law is
~
true in D , then we will be able to say that D is a ring. For any a~, b , c~  D :
~
(a~  b )  c~   f (a)  f (b)  f (c)  f (a  b)  f (c)  f ((a  b)  c)  f (a  c  b  c) 
~
 f (a  c)  f (b  c)   f (a)  f (c)    f (b)  f (c)   a~  c~  b  c~ .
And similarly c ~  (a~  b~)  c~  a~  c~  b~ . Therefore, if R is a ring, then D is a ring.

Let R is a field, the mapping f transfers zero 0 R into zero 0 D (because f is one-to-one mapping
of additive groups ( R,) 
f
( D,) ). Let’s discard zeroes from the sets R and D, then f is still
one-to-one mapping f : R \ {0 R }  D \ {0 D }. As ( R \ {0 R }),  is a commutative multiplicative
group and D \ {0 D },  is a set with a binary operation  on it, from the condition
f (a  b)  f (a)  f (b) , by lemma6 , we get that D \ {0 D },  is also a commutative multiplicative
group  D is a field.

Complement [for theorem4]. If ( R , , ) is a (commutative ring)/(ring with one), then [ D , , ]

is also a (commutative ring)/(ring with one).

41
 Groups, Rings, Fields

Ordered rings and fields

Def: a ring/field ( R,,) is called an ordered ring/field if there exist some subset   R of
elements, that we will call “positive”, (so a    a is “positive”) and
[A] for any a  R exactly one of the next cases is true: a is positive, or  a is positive, or a  0.
[B] If a, b are positive, then a  b and a  b are positive.

Notice. From [A] follows that 0   . In practice when we have some ring/field ( R,,) we need
to mark some set   R , which is closed with respect to addition and multiplication and does not
contain a zero element 0 , such that: for any a  R only one of the next cases is true:
a   , or  a   , or a  0.
Any ordered ring/field ( R,,) must be always turned into an ordered set, the order ""
must be defined in the next way.

Def. For any elements a, b  R we define: a  b if (a  b)   and b  a if (b  a)   .

Correctness. "" is an order relation on R.

Proof. We need to show that any elements a, b  R are comparable and also that "" is transitive,
then R is an ordered set by definition. Let’s fix any a, b  R and consider the element (a  b) .
According to [B], only one of the next cases is true: (a  b)  , or  (a  b)    (b  a)   ,
or (a  b)  0 it is equivalent to a  b , or b  a , or a  b , then any a, b  R are comparable.
We also have to show that the relation "" is transitive: ( if a  b and b  c , then a  c ).
Let a  b and b  c , then (a  b)   and (b  c)   . The set  is closed with respect to
addition, then (a  b)  (b  c)  (a  c)    a  c , then "" is transitive. So "" is an order
relation on M.
Def. Any element a  R such that  a   is called “negative”. So, any element a  R is
positive, or negative, or zero.

Properties of ordered rings/fields.

Property1. R is an ordered ring/field.
Then: a is positive  a  0 . And: a is negative  a  0.
Proof: Let a  a    (a  0)    a  0 .
is positive
Conversely: a  0  (a  0)    a   then a is positive.
Next: a is negative  a    (0  a)    0  a  a  0 .
And similarly: a  0  a is negative.

42
 Groups, Rings, Fields

Property2 [sign rules]. R is an ordered ring. Then for any a  0, b  0  a  b  0, a  b  0,

if R is a field, then also a / b  0 . In any ord. ring R : if a  0, b  0  a  b  0, a  b  0,

if R is also a field, then also a / b  0 . In any ord. ring R : if a  0, b  0 , or a  0, b  0 ,
then a  b  0 , if R is also a field, then also a / b  0 and b / a  0 .
Remember that when R is just a ring, division ":" is not defined on R , we can speak about
division only when R is a field.

Property3. Any ordered ring does not have any zero divisors.

Proof: R is an ordered ring. Let a b  0 , but a  0, b  0 .

If a  0, b  0  / property 2 /  a  b  0  a  b  0 .
If a  0, b  0  / property 2 /  a  b  0  a  b  0 . If a  0, b  0 or a  0, b  0 , then
a  b  0  a  b  0 . So in any case a b  0 and there are no zero divisors in R.

Absolute value
Def. R is an ordered ring/field and a  R . The absolute value of a is the element a  R such

that: if a  0 , then a  0 , if a  0 , then a  a , if a  0 , then a  a .

a  0 and a  0 .
In each case it’s enough to consider separately all the variants. In the case [A]
For [B],[C] we consider separately each case: a  0, b  0 and a  0, b  0 and a  0, b  0

and a  0, b  0 . Let’s notice that in the case [C] it’s enough to prove that a  b  a  b ,
then a  (a  b)  (b)  a  b   b  a  b  b . Then a  a  b  b  a  b  a  b .

Def. R is an ordered ring/field. We say that the Archimedes axiom is true in R if for any
positive elements a, b  R there exist the natural number n such that n  a  b .
Property4. R is an ordered ring/field.
[A] a  b  a  b [B] a  c  b  c (for some c  R )  a  b . And a  b  a  c  b  c
(for any c  R ).
[C] a  c  b  c (for some c  0 )  a  b . And a  b  a  c  b  c (for any c  0 ).
a  c  b  c (for some c  0 )  a  b . And a  b  a  c  b  c (for any c  0 ).
[D] a  b, c  d  a  c  b  d
Properties. [A] a   a for any a  R , [B] a  b  a  b for any a, b  R ,
Property5.
[C] a  b Rais an
b ordered
a  bfield
a,(use
bR [C]  property4).
fromathe
, [D] for some positive      a   ,


[E] ba d
[A]  0for  a
some apositive
, then:
c
 d ac  [B]a b
 b or d. 0 , then:   a  d  c  b
a c
b d b d

43
 Groups, Rings, Fields

Exercise: In any ordered ring/field a  0  a 2  0 .

Theorem5 [Transfer of order]. Suppose we have exactly the same conditions as in the theorem4.
Then: If ( R,,) is an ordered ring/field, then [ D , , ] is also an ordered ring/field.
And f is one-to-one correspondence between positive and negative elements of these rings/fields:
a is positive / negative (in R)  f (a) is positive / negative (in R) .
And f conserves the order relation: a  b (in R)  f (a)  f (b) (in D) .
Proof. Let  is a set of “positive elements” in R , i.e.,  is a set with the properties [A] and [B]
(from the definition above). Let’s show that f ()  { f (m) || m  } is the a with the same
properties in D . From [A] follows that R is divided into the next sets without common elements

R    {0R }   , as f : R  D is one-to-one, then D is divided into the next sets without

common elements: D  f ()  f (0 R )  f () . We know that f is an isomorphism of additive
groups ( R,)  ( D,) , then f (0 R )  0 D and also f (a)   f (a) for any a  R .
Then for any set   R : f ()   f ( ) and in particular f ()   f () .
Then D   f ()  {0D }  f () (the sets on the right side do not have any common elements).
Then the condition [A] is true for f ()  D , i.e., for any d  D exactly one of the next cases
is true: d  f (), d  0D , d   f ()  d  f () . Also f () is closed under
addition  /multiplication  , because  is closed under addition  /multiplication .
Then the condition [B] is true for f () . Therefore D is an ordered ring/field and the positive

elements f ()  D are exactly the images of the positive elements   R.

Then: a is positive in R  f (a) is positive in D. Similarly, the negative elements  f ()  D

are exactly the images of the negative elements    R .

Then: a is negative in R  f (a) is negative in D.

Next. Let ab in R, then (b  a) is positive in R , then f (b  a) is positive in D.

Let’s simplify: f (b  a)  f (b  (a))  f (b)  f (a)  f (b)  ( f (a))  f (b)  f (a) - this

element is positive in D, then f (a)  f (b) in D.

Conversely. Let f (a)  f (b) in D . Then f (b)  f (a)  f (b  a) is positive in D, then b  a is

positive in R, then b  a in R. Everything is proved.

44
 Groups, Rings, Fields

Exercise6. R is an ordered ring/field. Then any nonzero subring/subfield   R is also an

ordered ring/field. And a, b   : a  b (in )  a  b (in R) .

Notice, when we say “nonzero subring” we mean   {0} . Really, the set   {0} is a subring
of R which consists of only zero element. The word “nonzero” above refers to “subring”, because

  {0} is not a field, so there can’t be any zero subfield.

Solution. Let’s take the set of positive elements   R and consider the set    .
This set is not empty. Really, as   {0} , then there exist a   || a  0 . Then (according to the
properties of the set   R ) either a   , or  a   . In the same time both a  ,a  
(because  is a ring). Then    contains a or  a . So    is not empty.
Let’s define: the set    is the set of positive elements in  . Let’s check the main requirements:
[A] For any a   only one of the next cases is true: a     , or  a     , or a  0 .
Let’s take any a   . If a  0   , then a     and  a     (because 0   ).
Let a  0   . We will show that there can be only one of the next cases:
a     , or  a     . As a is an element of the ring R, then a   , or  a   .
[1-st variant] a   , then  a   . Also a   and  a   . Then a    
and  a     . [2-nd variant]  a   , then a   . Also  a   and a   .
Then  a     and a     .
[B] If a, b are positive (belong to    ), then a  b and a  b are positive (belong to    ).
Really, let a, b     , then a, b   (both a, b are positive in R ), then a  b, a  b   .
Next, a, b   , then (  is a ring) a  b, a  b   . So a  b, a  b     . We have proved that
any subring/subfield   R is also an ordered ring/field. And    are exactly the positive
elements of  , where  are the positive elements of R .
Let’s fix now any a, b   . Let a  b (in ) , then (b  a) is positive in H

 (b  a)      (b  a)   , then (b  a) is positive in R, so a  b (in R) .

Let now a, b   and a  b (in R) , then (b  a)   . Also (b  a)   (because H is a ring),

then (b  a)     , so (b  a) is positive in H, then a  b (in ) . Everything is proved.

45
 Matrixes

Matrixes
Everywhere later  is a set with addition and multiplication on it (binary operations on  )
which both associative, commutative and these operations are connected by the distributive law.
For example,    , or  is a commutative ring, or  is a field. Each of these variants is
appropriate.

Def. Any table with m rows and n columns, which is formed from some elements of  , is called
an m  n matrix (or just a matrix).

 1 3 1 1 3 4
For example,   and   and   are all 2  2 matrixes (here we use the set    ).
 2 1   3 8   5 8 
1 2 4
And 3 1 5 is 1 3 matrix,   is 2  3 matrix. By definition, any 1 1 matrix (a) is
3 5 5
exactly the element a , so (a)  a . When the number of rows m and the number of columns n are
equal: m  n , we say that we have “a square matrix”. When m  n we say that we have
“a rectangular matrix”.

The element of a matrix A which stays at the intersection of the row with number i and the column
with number j is denoted like aij .

1 3
For example: A    , then a11  1, a21  2, a12  3, a22  4 .
 2 4
 a11 a12 ... a1n 
 
 a21 a22 ... a2 n 
In general, m  n matrix A can be denoted as: A   .
... ... ... ... 
 
 am1 am 2 ... amn 

Matrixes A and B are equal ( A  B ) if they have exactly the same size ( mA  mB , nA  nB ) and
they consist of the same elements aij  bij i, j .

Def. Let A and B are both m  n matrixes (of the same size), then their sum is m  n matrix С
such that сij  aij  bij i, j . And we denote A  B  C .

 1 3  3 2  1  3 3  2   4 5 
For example,            and 1 2  2 3  1  2 2  3  3 5 .
 2 1 1 4   2  1 1  4   3 5
Addition of matrixes is commutative and associative: A  B  B  A and ( A  B)  C  ( A  B)  C
for any m  n matrixes A, B, C (because addition on  is commutative and associative).

47
 Matrixes

 b11 
 
b 
Def. The product of any 1  n matrix A  (a11, a12... a1n ) and any n  1 matrix B   21  is 1 1
...
 
 bn1 
matrix (the number) a11  b11  a12  b21  ...  a1n  bn1 . So, by definition:
(a11 a12 ... a1n )   b11 
 
 b21 
 ...   a11  b11  a12  b21  ...  a1n  bn1 .
 
 bn1 
3 1 1   1 
(1 2)   3   
For example,    1  3  2  1  5 and  3   3 1  1 3  1 2  8 .
1  2
 
We have defined how to multiply a row and a column (any 1 n row and any n  1 column).
We can also multiply matrixes of appropriate sizes (only if the number of columns in A is equal to

the number of rows in B , then the product (matrix) A  B is defined ).

Def. Let A is m  n and B is n  k matrix (notice that the number n of columns in A and the
number n of rows in B are equal). The product A  B is m  k matrix С such that: every element
сij  C (the element at the intersection of i – th row and j – th column ) is a product of i – th row
of A and j – th column of B .
Suppose that we need to calculate A  B . Let’s take the first row of A , we will multiply this row
consecutively by the 1-st,2-nd,3-rd… columns of B , the results of these multiplications must be
placed at the first row of A  B . Then we take the second row of A, and we multiply it consecutively

by the 1-st,2-nd, 3-rd… columns of B, the results must be placed at the second row of A  B and etc.

By definition, for any element сij  С  A  B we have:

сij  (i  th row of A)   j  th   (ai1 ai 2 ... ain )   b1 j   ai1  b1 j  ai 2  b2 j  ...  ain  bnj

   
 column   b2 j 
 of   ... 
   
b 
 B   nj 

48
 Matrixes

1
 1 2 3  
For example,   is 2  3 matrix and  2  is a 3 1 matrix, so they can be multiplied:
 3 2 1 1
 
 1 2 3   1  1  1  2  2  3  1  8 
        
 3 2 1   2   3  1  2  2  1  1  8 .
1
 

5 3 1 1
   
Let’s take 3 3 matrix  2 3 2  and 3 1 matrix  1  , these matrixes can be multiplied:
7 1 1  2
   
 5 3 1   1   5  1  3  1  1  2  10 
       
 2 3 2    1    2 1  3 1  2  2    9  .
 7 1 1   2   7  1  1  1  1  2  10 
       

1
Let’s take A    and B  1 3. We can multiply these matrixes, because their sizes are 2 1
 2
 1   1 3  1  1 1  3   1 3 
and 1 2 , so A  B          , here we can also calculate
 2  2 1 2  3  2 6 
1 3   1 
B A    1  1  3  2  7 .
 2
It’s always very easy to calculate the size of a product-matrix: if A is m  n and B is n  k , then
A  B is m k .
Multiplication of matrixes is NOT commutative. Really, let A is 3 1 and B is 1 2 matrix, then
A  B is defined and B  A is not even defined. Both matrixes A  B and B  A are defined only when
B and A have sizes m  n and n  m . And if m  n , then A  B and B  A are matrixes of different
sizes, so they can’t be equal. Only when m  n (when A and B are both square n  n matrixes)
matrixes A  B and B  A have the same size n  n . But even in this case we can’t guarantee that
these matrixes are equal.

Associativity. Multiplication of matrixes is associative: A  ( B  C )  ( A  B)  C for any matrixes

A, B, C .
Remark. In order the product A  ( B  С ) to be defined, the matrixes must have appropriate sizes:

A  m  n, B  n  k , C  k  h . In this way, both products A  ( B  C ) and ( A  B)  С are

defined. And obviously A  ( B  C ) and ( A  B)  С are both matrixes of the same size m  h .

49
 Matrixes

Proof. Let’s show that every element ij of A  ( B  С ) is equal to the element  ij of ( A  B)  С .
Let’s fix any element ij of A  ( B  С ) , then:

ij  (the row with number i of A)  (the column with number j of B  C )  (ai1, ai 2 ... ain )   p1 j 
 
 p2 j 
 ...  .
 
p 
 nj 
The column with number j of B  C consists of the elements:
 p1 j   (the row with number 1 of B)  (the column with number j of C )    b1s  csj 
    1 sk
 p2 j   (the row with number 2 of B)  (the column with number j of C )    b2 s  csj 
 ...       1 sk 
  
...
   ....

 p   (the row with number n of B)  (the column with number j of C ) 
 nj      bns  csj 
 1 sk 

  b1s  c sj 
 
 ij  ai1 , ai 2 ... ain   1sk
 b c 

 1 2s sj
  a   b  c   a   b  c   ...  a   b  c  

sk i1   1s sj  i2   2s sj  in   ns sj 
....   1 s  k   1 s  k   1 s  k 
 
  bns  csj 
 1sk 

 ai1  b11c1 j  b12c2 j  ..  b1k ckj   ai 2  b21c1 j  b22c2 j  ..  b2 k ckj   ...

...  ain  bn1c1 j  bn 2 c2 j  ..  bnk ckj 
.

Let’s regroup our summands, we want the numbers c1 j , c2 j ... ckj to be in front of the brackets:

 c1 j  ai1b11  ai 2b21  ..  ainbn1   c2 j  ai1b12  ai 2b22  ..  ainbn 2   ...

...  ckj  ai1c1k  ai 2b2k  ..  ainbnk   ij
[1].

Let’s calculate the element  ij of ( A  B)  C :

 ij  (the row with number i of   )  (the column with number j of C )  ( zi1 , zi 2 ... zik )   c1 j 
 
 c2 j 
 ... 
 
c 
 kj 
The row with number i of A  B consists of the elements:

50
 Matrixes

( zi1 , zi 2 ... zik ) 

 (i  th row of A )  (1  st column of B) .... (i  th row of A)  (k  th column of B)  

 (ai1 , ai 2 ... ain )   b11  (ai1 , ai 2 ... ain )   b12  (ai1 , ai 2 ... ain )   b1k  
       
  21 
b  22 
b  b2 k  
  ...   ...  ,.....,  ...  
       
  bn1   bn 2   bnk  
 

   
   ais  bs1 ,  ais  bs 2 .....  ais  bsk   then ij    ais  bs1 ,  ais  bs 2 .....  ais  bsk    с1 j 
 1 s  n   1 s  n   
 1 s  n 1 s  n   1 s  n 1 s  n   c2 j 
 ... 
 
c 
 kj 

     
 с1 j    ais  bs1   с2 j    ais  bs 2 ....  ckj    ais  bsk  
 1 s n   1 s n   1 s n 
 с1 j  ai1b11  ai 2b21  ...  ainbn1   с2 j  ai1b12  ai 2b22  ...  ainbn 2   ....
[2].
...  ckj  ai1b1k  ai 2b2 k  ...  ainbnk    ij

Let’s compare [1] and [2], we see that ij   ij , it proves that A  ( B  C )  ( A  B)  C.

As multiplication of matrixes is associative, then in any product of several matrixes we can

rearrange brackets in any way we want, it will not change the result of multiplication.
For example, ( A  B)  (C  D)  ( E  F )  A  ( B  C )  D  ( E  F )  .

It’s very easy to prove the distributive law for matrix-multiplication.

[A] ( A  B)  C  A  C  B  C (for any m  n matrixes A, B and any n  k matrix С ),
[B] C  ( A  B)  C  A  C  B (for any m  n matrix C and any n  k matrixes A and B ).

Def. For any element   F and any m  n matrix A . The matrix   A is such matrix C that
сij    aij i, j .

1 4  2 8 
For example: 2     and 2  3 2 4  6 4 8 .
3 5   6 10 

Let’s consider the set  of all m  n matrixes which elements are taken from some commutative
ring R . It’s easy to see that  is a commutative additive group. Really, addition of matrixes is
commutative and associative, the zero matrix O is m  n matrix which consists of zeroes.
For any m  n matrix A , an opposite matrix  A is the matrix which consists of opposite elements:
A  {aij }   A  {aij } . So  is a commutative group.
51
 Matrixes

Then we can speak about subtraction "" on  . For any matrixes A, B   , their difference is the
matrix С   such that A  B  С . We can consider the last equality as an equation A  B  C ,
where A, B are fixed matrixes, and we have to find the matrix C .
As A  B  C  aij  bij  cij || i, j  cij  aij  bij || i, j .
Then: for any m  n matrixes A and B , their difference ( A  B) is the matrix C such that
сij  aij  bij i, j .

3 9  2 3  1 6
For example,         and 5 5 5  3 2 1  2 3 4 .
 5 8   1 4   4 4 

If we consider the set 

n n
of all square n  n matrixes. Then  nn is a ring.
Really,  is obviously a commutative group under addition. Multiplication (on 
n n n n
) is
associative and distributive over addition. So  is a ring. Let’s notice that 
n n n n
mustn’t be
a field, at least because the multiplication of square matrixes (in general) is not commutative.

Def: a square n  n matrix  is called symmetric if it has a symmetry with respect to it’s main
diagonal a11, a22... ann . We can write it shortly:  is symmetric if aij  a ji i, j .

For such matrix any i  th row of  and i  th column of  consist of the same elements.

From here immediately follows the next simple fact: multiplication of symmetric matrixes is
commutative, i.e., for any symmetric matrixes ,  we have        (understand why).

52
 Integer Numbers

Integer numbers
Def.  is the ring that consists of natural numbers   {1,2,3,4,5...}, the zero number 0 , and all
the numbers that are opposite to naturals   {1,2,3,4,5...}, so     {0}   .
The addition and multiplication on  are extensions of the addition and multiplication on  , i.e.,
for any a, b     : a  b (addition by the rules of )  a  b (addition by the rules of ),
a  b (multiplication by the rules of )  a  b (multiplication by the rules of ) .
The ring Z is called a ring of integer numbers and elements of Z are called integer numbers.

Let’s build  . We fix any set N of natural numbers. Let’s consider the set  of all pairs (a, b)
where a, b are any natural numbers. Let’s define addition and multiplication on .
Def. (a, b)  (c, d )  / by def /  (a  c, b  d ) and (a, b)  (c, d )  / by def /  (a  c  b  d , a  d  b  c)
How to remember it? Imagine that any pair (a, b) denotes the difference a – b of elements a, b
in some ring, then (a, b)  (c, d )  (a  b)  (c  d )  (a  c)  (b  d )  (a  c, b  d )
and (a, b)  (c, d )  (a  b)  (c  d )  (a  c  b  d )  (a  d  b  c)  (a  c  b  d , a  d  b  c) .
For any pair (a, b) , it’s elements are natural numbers a, b . And we have already learned all
the basic properties of natural numbers (the addition and multiplication of natural numbers are
commutative and associative, there is also a distributive law and etc.).

Assertion1. The addition "" and multiplication "" (of pairs) are both commutative and
associative on  . There is also a distributive law:
(a, b)  (c, d )  (e, m)  (a, b)  (e, m)  (c, d )  (e, m) and
(e, m)  (a, b)  (c, d )  (e, m)  (a, b)  (e, m)  (c, d ).
Proof. We could check all these properties straightly. Let’s give a more interesting proof.
We consider the set X of all 2  2 symmetric matrixes of natural numbers.
 a b 
   A    || a, b    - this set is closed under the matrix addition and multiplication, really:
 b a 
a b  c d a c b d a b  c d  a  c  b  d a  d  b  c
        and      .
 b a  d c  b  d a  c   b a  d c   b  c  a  d b  d  a  c 
a b
Let’s define the mapping f :    : for any matrix   it’s image is the pair (a, b) ,
 b a 
a b
so f    (a, b) . Then f is one-to-one mapping, and f obviously has the next properties
b a 
[T] A, B   : f ( A  B)  f ( A)  f ( B), f ( A  B)  f ( A)  f ( B) (so f is a “sort of” isomorphism).
All the properties, that are listed in the assertion1, are true on X (because X consists of 2  2

54
 Integer Numbers

symmetric matrixes, and all the properties from the assertion1 are the standard properties of
square matrixes, except commutativity of multiplication, but it is true for any symmetric matrixes
of the same size). And we have one-to-one mapping f :    such that [T].
Then f transfers all the needed properties (associativity/commutativity/distributivity
of addition/multiplication) from  to  and the assertion1 is proved.

Comment: we have showed the transition of all these properties in the theorem4 (“Groups, rings,
fields”). The proof was done for rings/fields, and the Reader could notice that  is not a ring/field.
Now the situation is more simple, because  has just some properties of a ring (but not all of them),
and because of [T], f transfers all these properties from X to  .

Let’s define the equivalence relation "" on the set  of all pairs. By definition, pairs (a, b) and
(c, d ) are equivalent (a, b)  (c, d ) if a  d  b  c . It’s very easy to see that "" is really an
equivalence relation on  , because it is reflexive, symmetric and transitive. Therefore this relation
divides  into the classes A, B, C , D, E ... of equivalent pairs. All pairs in any class are equivalent
to each other. And the classes A, B, C, D, E ... do not have any common elements (common pairs).
The set of all classes we denote like  aux (aux=auxiliary).
Let’s define addition and multiplication on  aux in order to turn  aux into a ring.

Def. Let A,B are any classes from  aux . And (a, b)  A, (c, d )  B are any pairs from these
classes. The sum    is the class that contains the sum of pairs (a, b) and (c, d ) , i.e., the pair
(a  с, b  d )  (a, b)  (c, d ) . The product    is the class that contains the product of pairs
(a, b) and (c, d ) , i.e., the pair (a  c  b  d , a  d  b  c)  (a, b)  (c, d ) .
Notice: we use the same symbols  and  to denote the addition and multiplication of classes,
as we used for the addition and multiplication of pairs. But of course, the addition/multiplication of
pairs and classes are different operations which are defined on different sets. But there is no need
to complicate our designations, let’s just stick to  and  .

Assertion2. The given definition is correct. For any classes ,    aux , a sum    and
a product    are uniquely defined (it doesn’t matter which pairs (a, b) and (c, d ) are chosen
from A and B).

Proof. Let (a, b), (a , b )   and (c, d ), (c , d )   . We just need to show that
(a, b)  (с, d )  (a , b )  (с , d ) and (a, b)  (с, d )  (a , b )  (с , d ) , then the classes    and
   are uniquely defined. As (a, b), (a , b )   , then a  b  a  b , as (c, d )  (c , d ) , then
c  d  c  d . Let’s sum two last equalities:

(a  b )  (c  d )  (a  b)  (c  d )  (a  c)  (b  d )  (a  c )  (b  d ) [T].
55
 Integer Numbers

By definition, (a, b)  (с, d )  (a  c, b  d ) and (a , b )  (с , d )  (a  с , b  d ) .

From [T] we see that pairs (a  c, b  d ) and (a  c , b  d ) are equivalent, therefore:

(a, b)  (c, d )  (a , b )  (c , d ) .

The “straight proof ” here is inconvenient, let’s prove the auxiliary equivalence: if ( a, b)  ( a , b ) ,
then for any pair (c, d ) there must be: (a, b)  (c, d )  (a , b )  (c, d ) [aux].

We have the basic equality: a  b  a  b [D]. So (a, b)  (c, d )  (a  c  b  d , a  d  b  c) and

(a , b )  (с, d )  (a  с  b  d , a  d  b  c) . We need to show that

(a  c  b  d , a  d  b  c)  (a  с  b  d , a  d  b  c) it is equivalent to:
(a  c  b  d )  (a  d  b  c)  (a  с  b  d )  (a  d  b  c) and this equality is equivalent to:
(a  b )  c  (b  a )  d  (a  b)  c  (b  a)  d [V], let’s check that it is true by using [D].
We can substitute everywhere (in [V]) a  b instead of a  b . Then [V] looks like
(a  b )  c  (a  b )  d  (a  b )  c  (a  b )  d - which is a true equality, then [aux] is proved.
Let’s prove now the main one, as (a, b)  (a , b ) , then (a, b)  (c, d )  (a , b )  (c, d ) .
As (c, d )  (c , d ) , then (a , b )  (c, d )  (a , b )  (c , d ) and therefore (a, b)  (c, d )  (a , b )  (c , d ) .

Assertion3. The set ( aux,,  ) (of all classes) is a commutative ring with one.

Proof. The addition of pairs is commutative and associative, therefore the addition of classes is also
commutative and associative. The zero O is the class which contains (1,1) . Really, for any class A,
the class AO is the class which contains (a, b)  (1,1) || (a, b)  A, (1,1)  O .
Then AO contains (a  1, b  1) . But the pair (a  1, b  1) is equivalent to (a, b) , because
(a  1)  b  (b  1)  a therefore      and similarly      .
So O is a zero element of  aux . Next, for any class  which contains (a, b) , an opposite class  
is the class which contains (b, a) . Really, let’s fix an arbitrary (a, b)  A . The pair (b, a) belongs to
some class B. Then    contains the pair (a, b)  (b, a)  (a  b, b  a)  (1,1) . Therefore
     and similarly      , then, by definition,  is an opposite to  element,
so    .

Therefore  aux is a commutative group under addition.

The multiplication of pairs is associative, then the multiplication of classes is also associative.
And finally, there is a distributive law for pairs, from here immediately follows the distributive law
for classes. Then  aux is a ring. Moreover, the multiplication of pairs is commutative, then the

56
 Integer Numbers

multiplication of classes is also commutative, so  aux is a commutative ring. There is also the class

“one”  which contains the pair (2,1) . Really, for any class A the class A  I contains the pair

(a, b)  (2,1) || (a, b)  A, (2,1)  I . But (a, b)  (2,1)  (a  2  b 1, a 1  b  2) this pair is equivalent
to (a, b) , then      and similarly      . Then I is an element “one” of  aux .
Conclusion:  aux is a commutative ring with one.

Def: a class    aux is called positive if for any pair (a, b)   : a  b .

Correctness. The given definition is correct. If a  b for some pair (a, b) , then c  d for any pair
(c, d )  A .
Proof. Let (a, b)   and a  b , let’s take any other pair (с, d )   .
Then (a, b)  (c, d )  a  d  c  b .
If c  d , then from a  b follows that a  d  c  b . If c  d , then from a  b follows that
a  d  c  b . Then, the only possible variant is c  d .
Conclusion. If ab for some pair from  , then the same is true for any other pair from  .
It’s very easy to prove the similar assertion for the case a  b . And finally, if a  b for some pair
from  , then the same is true for any other pair and    .
We can sum up: for any class    aux only one of the next cases is true:
[E]  is positive, or   is positive, or    (remember that if (a, b)  A , then (b, a)   ).

Def. Let  POS is the set of all positive classes of  aux .

Assertion4.  aux is an ordered ring.
Proof. For any class    aux [E] is true. According to the basic definition, besides [E], we also need

to show that if ,  are positive, then both A  B and    are positive. Let’s take any
(a, b)   || a  b and (c, d )   || c  d , then obviously, the pair (a, b)  (c, d )  (a  b, c  d ) is
a positive pair, and similarly (a, b)  (c, d )  (a  c  b  d , a  d  b  c) is positive, let’s show the
last one: a ba b and c  d  c  d   , then a  c  b  d  (b  )  (d   )  b  d 
 b  d  b     d     b  d and a  d  b  c  (b  )  d  b  (d   ) 
 b  d    d  b  d  b   , we can compare two sums that we obtained, it’s easy to see that
a  c  b  d  a  d  b  c , and the pair (a, b)  (c, d ) is positive.
From here immediately follows that    and    are positive classes.
We have shown that  aux is an ordered ring.

According to the rules of any ordered ring, the class    aux is called negative if   is positive.
Let (a, b)   , then (b, a)   , if   is positive, then b  a . Then for any pair (a, b) from any
57
 Integer Numbers

negative class B we have a  b . Let  NEG is the set of all negative classes. As any class of the ring
 aux is positive, or negative, or zero, then aux   NEG     POS .
Assertion5.
For any positive class    POS there exist the unique natural number  such that
(a, b)   : a  b   .
And for any natural number  there exist the unique positive class  such that
(a, b)   : a  b   .
For any negative class    NEG there exist the unique natural number  such that
(a, b)   : a    b .
And for any natural number  there exist the unique negative class  such that
(a, b)   : a    b .
Proof. Existence. Let  is a positive class: (a, b), (c, d )   : a  d  c  b it is an equality of

natural numbers. Here a  b and c  d , then both sides of the equality a  d  c  b are greater
than (b  d ) , so we can subtract (b  d ) from both sides:
(a  d )  (b  d )  (c  b)  (b  d )  a  b  c  d .
So, for any pairs (a, b), (c, d )    a  b  c  d  / by def /   .

Then a  b   and c  d   and similarly for any other pair from A.

Uniqueness. Let for some natural number  we have (a, b)   : a  b   , and in the same

time for the other natural number   we also have: (a, b)   : a  b   ,

then b    b       .
Next. Existence. Let’s fix any natural  and consider the class  which contains the pair
(1  , 1) , then  is positive. For any (a, b)   we have (a, b)  (1  , 1) , then
a 1  1   b  a  b   .
Uniqueness. Let for some natural number  we have the classes ,  such that a  b   for
any (a, b)   and c  d   for any (c, d )   . Let’s sum the equalities we have in the next way:
a  (d  )  (b  )  c  a  d  b  c  (a, b)  (c, d )     .
For any negative class    NEG the proof is similar.

Let’s build now the set  of integer numbers. We take  aux and every class A from  POS with
a pair (a, b)   we replace by the natural number  such that a  b   (assertion5).
Then instead every positive class we will have exactly one natural number. And the set  POS will
58
 Integer Numbers

turn into the set  of natural numbers. All the other classes of  aux must stay untouched.
   NEG     POS   NEG       . The set    NEG     is called
a set of integer numbers. This set consists of natural numbers and classes of equivalent pairs
(of natural numbers). We need to define addition and multiplication on  (because  consists of
natural numbers and classes, and we don’t know how to add or multiply those).

Def. Any elements a, b   have unique representations as a  f ( ), b  f () , where , 

are some classes of the ring  aux . Then we define a  b  f ()  f ()  / by def /  f (  )

and a  b  f ()  f ()  / by def /  f (   ) .

 NEG   POS
Let’s define one-to-one mapping f :  aux   such that    .
 NEG  
For any class   (  NEG  ) we define f ()   . For any class    POS , which contains

a pair (a, b) : a  b   , we define f ()   . Then f is one-to-one mapping.

Then the addition "" and multiplication "" are binary operations on Z . And ( , ,  ) is a set

with two binary operations on it. We see that: f :  aux   is one-to-one mapping such that

f (  )  f ()  f () and f (  )  f ()  f () for any ,   aux .

From the theorem4 (page 40) immediately follows that ( , ,  ) is a ring.
From the complement for theorem4 follows that  is a commutative ring and a ring with one.
According to the theorem5 (Transfer of order) (page 44),  is an ordered ring and natural numbers
 are the only positive elements of  .

Property1. The addition and multiplication on  are extensions of the addition and
multiplication of natural numbers, i.e., for any a, b     we have:
a  b (addition by the rules of )  a  b (addition by the rules of ) ,
a  b (multiplication by the rules of )  a  b (multiplication by the rules of )
And similarly, the order relation "" on  is an extension of the order relation on  :
a  b ( by the rules of )  a  b (by the rules of ) .

Proof. For illustration purposes, let’s denote for a while (~

,~ ) - the addition and multiplication on
the ring Z and (, ) - the addition and multiplication on N. Let’s fix arbitrary natural numbers
a, b     . In order to find the sum a ~
 b (in ) we need to find the positive classes , 

(from  aux ) such that f ()  a, f ()  b , then, by definition, a ~

 b  f () ~
 f ()  f (  ) .

59
 Integer Numbers

So,  is the class with the pair (1  a, 1) and  is the class with the pair (1  b, 1) .
The class    contains the pair (1  a, 1)  (1  b, 1)  ((1  a)  (1  b), 1  1)  (2  (a  b), 2) ,
~
then f (   )  (a  b) and therefore a  b  a  b for any a, b   .

Next, a ~ b  f () ~ f ()  f (  ) and (  ) is the class which contains the pair

(1  a, 1)  (1  b, 1)  (1  a)  (1  b)  1  1, (1  a)  1  1  (1  b)  (2  a  b)  a  b, (2  a  b)

then f (  )  a  b therefore a ~ b  a  b a, b   .

And finally, a, b   and a ~

 b (in ) . The order relation on  was transferred from  aux ,
so,a~ b  f () ~
 f ()     (in  aux ) (look at the theorem5 (Transfer of order)).
 b     and the assertion a ~
So a ~  b (in )  a  b (in ) (that we want to prove) is
equivalent to    (in aux )  a  b (in ) . As a  f () , then  contains the pair (1  a, 1) ,
as b  f () , then  contains the pair (1  b, 1) . Let    (in aux ) , it means that the class
(  )    () is positive in  aux . The class   () contains the pair

(1  a, 1)  (1, 1  b)  (2  a, 2  b) as   () is positive, then 2  a  2  b  a  b (in  ).

Conversely. Let a  b (in ) , then the class (  )    () with the pair (2  a, 2  b) is
positive, then    (in  aux ).

Conclusion: as the order relation/ addition/multiplication on  are all extensions of the order
relation/ addition/multiplication on    , then it’s appropriate to use exactly the same signs
(  ) (as we used on  ) to denote these operations on .

Property2. The zero class    is a zero element of the ring Z, so   0 .

For any negative class    NEG there exist the unique natural number      such that
   . And finally:        .

Proof. Let’s fix any a , it can be uniquely represented as a  f () .

Then a    f ()  f ()  f (  )  f ()  a and similarly   a  a , then  is a zero

element of  . Next,
Existence. Let’s fix any    NEG . Then    aux . According to the assertion5,

there is the unique natural number  such that (a, b)   : a    b , and B contains

the pair (1, 1  ) . The class     aux is the class that contains the pair (1  , 1) .

60
 Integer Numbers

Then   ()   (in  aux ) , therefore f   ()  f ()  f ()  f ()  f () .
Here f ()   and f ()   and f ()    0 , then     0     .
Uniqueness. Let    and    , where ,    . Then         ,

and the uniqueness is proved.

Let’s finally show that        . We consider the set  NEG   and the set    .
Let’s will show that  NEG    . It’s enough to show that  NEG   and     NEG
(then     NEG ). Let’s fix any class    NEG . As we showed above, there exist the unique
natural number  such that          , then  NEG   .
Conversely, let’s fix now any element     , then    .
Let’s consider the class  with the pair (1,1   ) , as 1  1   , then    NEG and
    0 (in )       NEG     NEG , then     NEG .
As  NEG   and also   0 and (according to our construction)    NEG  0   ,
then     0   . The property2 is proved.
Let’s show that  is a minimal ring which contains  . Here we need to notice, that when we speak
about some ring R which contains  , we imply not only that   R , but also that addition and
multiplication on R are extensions of the addition and multiplication on  . If we do not make such
requirement, the addition and multiplication on  do not make any sense in R. And all the
properties of natural numbers, which are connected with addition/multiplication/order, mustn’t be
true inR. In such case we can’t say anything determined about R, except that it contains all the
elements (symbols) of .

Property3.  is a minimal ring which contains N, it means that:

any other ring R, which contains N, contains  as a subring.

Let some ring R contains  , this ring must also contain a zero element 0  R (and obviously 0
can’t be an element of  , because a   : a  b  b for any b   , so any a   can’t be a zero
element of R. Here we can see why we require from addition on R to be an extension of the
addition on  ). The ring R must also contain all the elements which are opposite to natural
numbers  (because R is an additive group). And any element  a , which is opposite to a natural
number a , can’t belong to   R .
Really,  a  a  0 , if we assume that  a   , then both a, a   and therefore 0  
(because a sum of natural numbers is a natural number), but 0   , so we have a contradiction.
We showed that {0}     and       . Let’s finally show that    {0}   , if it’s not
true, then 0   , then 0  a for some a   , then a  0 which is not true. So, any two of the
~ ~
sets  , {0},  do not intersect in R. Let’s denote      {0}   , any element of  is a
natural number, or an opposite to some natural, or zero.
61
 Integer Numbers

According to the rules that govern addition and multiplication in any ring,
(like a  b  (a  b), a  b  a  (b), (a)  (b)  a  b ), sum/difference/product of any
~ ~ ~
elements a,b   are again some elements of  , then  is a subring of R (subring criterion).
~
Then  is an independent ring, inside R . As addition/multiplication on R are extensions of
~
the addition/multiplication on   R , then addition/multiplication on  are extensions of
~ ~
the addition/multiplication on    . It’s easy to see that  is a ring of integer numbers
(according to the initial definition). Everything is proved.
Uniqueness. We have built the ring of integer numbers  , based on some set  of natural
numbers that we have fixed at the very beginning. Formally, there exist different sets of natural
numbers (of course, all of them are isomorphic as ordered sets), and formally, we can construct
different sets  . The main question that appears here: are these sets of integer numbers also
isomorphic? And the answer is “Yes”. All sets of integer numbers are isomorphic as rings.
And we will show even more.

Uniqueness theorem. Any rings   and   of integer numbers are isomorphic. Moreover,
the isomorphism f :     is unique, this isomorphism is an extension of the unique
isomorphism f :      (look at the theorem3/advanced theorem3 for natural numbers).
Proof: [Existence]. Let’s build the isomorphism f :      . We take the unique isomorphism
f :      . It’s explicit definition is f (1 ) 1 and if f (a )  a , then f n(a )  n(a ) .
As we showed above:   ()  0   , so
  (  )  0     {... 3 ,2 ,1 ,0 ,1 ,2 ,3......} and
  (  )  0     {... 3 ,2 ,1 ,0 ,1 ,2 ,3......}.
...  3  2  1 0  1 2  3 ...
And f is one-to-one correspondence:     .
...  3  2  1 0 1 2 3 ...

Let’s extend f . We define f (0 )  0 , and for any negative number  a    we define
f (a )   f (a ) .
...  3  2  1 0  1 2  3 ...
Then f becomes one-to-one mapping         .
...  3  2  1 0 1 2 3 ...

Really, we already know that f :      is one-to-one, also f (0 )  0 , then

f : {0  }     {0 }    is one-to-one. Let’s finally show that (  ) 
f
(  ) is
one-to-one (then f :      is one-to-one). By definition:  a     f (a )   f (a ) .
From here we see that f is defined on the set    , and the values of f on    lie in
the set    . So f is a mapping: (  ) 
f
(   ) .
62
 Integer Numbers

[A] f covers the set    . Let’s take any element  a     a    , then there exist
the unique a    such that f (a )  a . Let’s consider the element  a    , by definition
f (a )   f (a )  a .
[B] f does not “glue together” elements of    . Let f (a )  f (b ) || a  b , we can
rewrite it  f (a )   f (b ) || a  b . Then f (a )  f (b ) || a  b , then different natural
numbers a and b are transferred into the same (natural) number f (a )  f (b ) which is
impossible, because f :      is one-to-one, and we have a contradiction.
Then (   ) 
f
(  ) is one-to-one.
Now we can claim that f :      is really one-to-one. Let’s show that f :      is
a ring isomorphism. So we need to show that a , b : f (a  b )  f (a )  f (b ) and
f (a  b )  f (a )  f (b ) . Let’s consider several possibilities:

[1-st case] One of the numbers a , b is a zero number. Multiplication and addition are
commutative in both rings, and both equalities (that we want to prove) are symmetrical
with respect to a , b , without loss of generality we can assume that a  0 .

Then f (a  b )  f (0   b )  f (b )  0  f (b )  f (a )  f (b ) and

f (a  b )  f (0   b )  f (0  )  0  0  f (b )  f (0  )  f (b )  f (a )  f (b ) .

[2-nd case] Both a , b are not zero numbers. There can be exactly 3 variants[A],[B],[C].

[A] Both numbers a , b are natural. In this case the properties f (a  b )  f (a )  f (b )
and f (a  b )  f (a )  f (b ) have already been proved (advanced theorem3 for natural numbers).
~ ~
[B] Both numbers a , b are negative. Then a  a~ , b  b , where a~ and b are natural.
In the next actions we will use the sign rules which are true in any ring, like a  (b)  a  b and
 a  (b)  a  b  (a  b) and (a)  (b)  a  b and etc. We will also use the case [A]
(properties of f for natural numbers). So,
~ ~ ~ ~ ~
f (a  b )  f (a~  (b ))  f ((a~  b ))  //(a~  b )    //   f (a~  b )  ( f (a~ )  f (b )) 
~ ~ ~
  f (a~ )  f (b )   f (a~ )  ( f (b ))  f (a~ )  f (b )  f (a )  f (b ) .
And f (a  b )  f ((a ~ )  (b~ ))  f (a~  b~ )  // a~ , b~   //  f (a~ )  f (b~ ) 
        

  f (a   
~ )   f (b~ )  f (a~ )  f (b~ )  f (a )  (b ) .
    

[C] One of the numbers a , b is positive and the other one is negative. Similarly, as in the
[1-st case], we can assume that a  is negative and b is positive, then a  a~ where a~ is
 
natural and b is natural. Let’s prove the simple part:
f (a  b )  f ((a~ )  b )  f ((a~  b ))  //(a~  b )    //   f (a
~  b )   f (a~ )  f (b ) 
   
 f (a~ )  f (b )  f (a )  f (b ) .
Let’s consider now f (a  b )  f ((a ~ )  b )  f (b  a~ ) || b~ , a~   here can be several
       

63
 Integer Numbers
~ )
simple cases, in each case we will work with the expression f (b  a
(which is equal to f (a  b ) ).

[C1] b  a~ is natural, then b  a~  x     b  a~  x , then f (b  a~ )  f ( x ) .

Let’s consider now the sum
f (a )  f (b )  f (a~ )  f (a~  x )   f (a~ )  f (a~ )  f ( x )  f ( x ) . So, we have deduced
that f (a  b )  f ( x ) and f (a )  f (b )  f ( x ) .

[C2] b  a~  0 , then b  a~ . Then f (b  a~ )  f (0 )  0 .

~ )  f (a~ )   f (a~ )  f (a~ )  0 .
Let’s consider the sum f (a )  f (b )  f (a    
And we have shown that f (a  b )  0 and f (a )  f (b )  0 .

[C3] b  a~ is a negative number, then b  a~   x || x     b  x  a~ .

Then f (b  a ~ )  f ( x )  / x   /   f ( x ) . Let’s consider
    
~ ~
f (a )  f (b )  f (a )  f (b )   f (a )  f (b )   f (b  x )  f (b )  // b , x    // 
  f (b )  f ( x )  f (b )   f ( x ) . And we have shown that f (a  b )   f ( x ) and
f (a )  f (b )   f ( x ) .
Everything is proved. f :     is a ring isomorphism.

Uniqueness. We have already built f :     which is an extension of the unique isomorphism

f :      . Let’s show that if  :    is a ring isomorphism, then   f on   .

Let’s assume that there exist some other ring-isomorphism  :    . As any isomorphism,
 must transfer zero into zero and one into one:  (0 )  0 and  (1 )  1 . In the main theorem
(for natural numbers) we have shown that every natural number a    is a descendant of one
a  n(n(...(n(n(e ))))), then a can be represented as a “sum of ones”
a  n(n(...(n(n(e )))))  n(n(...(n(e  e ))))  n(n(...(e  e  e )))  ...  e  e  ...  e .
Then for any natural
a    :  (e  e  ...  e )  //  (a  b)   (a)   (b) //   (e )   (e )  ...   (e ) 
 e  e  ...  e   , from here follows that the values of  on the set   belong to the set   .
Then we can write  :      . Let’s show that  satisfies the conditions:  (1 )  1 and
if  (a )  a , then  (n(a ))  n(a ) .

If these conditions are true, then  coincides with f on   . As we mentioned above,  (1 )  1 .
And if  (a )  a , then  (n(a ))   (a  1 )   (a )   (1 )  a  e  n(a ) . So   f on   .
As we mentioned earlier  (0  )  0  , then   f on {0  }    . The mapping f is defined on
(  ) by the rule   a     f (a )   f (a ) . Let’s fix an arbitrary  a    .
According to the basic properties of any ring-isomorphism:  (a )   (a )   a    and
in particular for any  a    :  (a )   (a )  //  (a )  f (a ) //   f (a )  f (a ) ,
then   f on (  ) . Then   f on   .
64
 Integer Numbers

Let’s list the other simple, but very important properties of integer numbers.

Property4. The Archimedes axiom is true in  : for any positive a, b   there exist the natural

n   such that n  a  b.
Proof. Let’s fix arbitrary positive a, b   , then both a, b   . And we have already proved
the Archimedes axiom for .
Property5. Let  is a nonempty subset of  .
[A] If there exist some number M which is greater than any element of  , then  has
the greatest number M max .
[B] If there exist some number m which is less than any element of  , then  has
the least number mmin .
Proof. [A] If there are some positive numbers in
X, then these numbers are natural and there is
a nonempty subset     of natural numbers. If there exist some number M such that
x  M || x    x  M || x    then (as we proved earlier) there exist the greatest number
M max    , obviously M max is the greatest number in X.

If there are no positive numbers in  and0   , then 0 is the greatest number in  .

Let there are no positive numbers in X and 0   . Then there are only negative numbers in X,
therefore   consists only of natural numbers, and (as we proved earlier)   has the least
number mmin   || mmin   x ( x)   , then  mmin   || mmin  x x   and  mmin is
the greatest number in  .

The case [B]. There exist m  x || x   . We can simplify the proof if we consider the set   of
integer numbers, then  m  ( x) || ( x)   . So, the number  m is greater than any number
of   . Let’s use the result [A], the set   contains the greatest number M max , then  M max
is the least number of  .

Divisibility
Def. a, b   if there exist k   such that a  k  b , then we say “ a is divisible by b ” and we
write ab . Also when a  k  b , we can say “ b divides a ”, or “ b is a divisor of a ” and we can
write b | a.
Exercise: If a, b   and a is divisible by b , then a  b .
If a, b   || a  0 and a is divisible by b , then a  b .
Theorem1 [division with a remainder].
[1-st part] For any natural numbers a, b there exist the unique pair of integer numbers
k , r || k  0, 0  r  b such that a  k  b  r .

65
 Integer Numbers

Proof: [A] Existence of numbers k, r . Let’s fix arbitrary a, b   , according to the Archimedes
axiom, n   : a  nb . Let  is the set of all integer numbers m, for which m  b  a .
Then m   : m  n . Then there exist the greatest number mmax   , obviously mmax  0

and obviously mmax  b  a  (mmax  1)  b . Let’s designate k  mmax , so we can write:

k  b  a  (k  1)  b . Let’s define r  a  k  b , then obviously 0  r . Let’s show that r  b . If r  b ,
then a  k  b  b  a  (k  1)  b it contradicts to a  (k  1)  b , then r  b and finally 0  r  b .

[B] Uniqueness of numbers k, r . Let’s assume that there exist some other pair of integer numbers
k , r which satisfies to the same conditions. Then a  k  b  r and a  k  b  r , then
k  b  r  k  b  r . From the last equality follows that: if k  k , then r  r and the uniqueness is
proved. Let k  k , without loss of generality k  k . The equality k  b  r  k  b  r is equivalent
to (k  k )  b  r  r , here (k  k )    (k  k )  1  (k  k )  b  b . Also 0  r  b and
0  r  b  0  r  b  b  r  0 let’s sum the last equality with 0  r  b , then
 b  r  r  b . So, the left part of the equality (k  k )  b  r  r is greater than or equal to b ,
and the right part is less than b , therefore (k  k )  b  r  r , and we have a contradiction.

Then k  k and r  r .

[2-nd part] For any integer numbers a, b   || b  0 there exist the unique pair of integer numbers
k , r || 0  r  b such that a  k  b  r . And k is called a quotient of a, b .
r is called a remainder of a divided by b.
Proof. [Existence of representation]. If a  0 let’s take k  0, r  0 . Let a  0 , then can be the
next cases: [A] If a  0, b  0 , then a, b are natural, everything is proved in this case.
[B] a  0, b  0 . Let’s take the natural numbers a, (b) , then, according to the [1-st part]), there
exist the representation: a  k  (b)  r || k  0, 0  r  (b) . As b  0 , then (b)  b and we can

rewrite a  (k )  b  r || (k )  , 0  r  b . Now we just have to designate k  (k ) and rr
and we have the representation we need: a  k  b  r || k  , 0  r  b .

[C] a  0, b  0 , or [D] a  0, b  0 , both cases must be considered exactly as the case [B], by using
the needed representation for natural numbers we get the similar representation for integers.

[Uniqueness of representation]. Let a  k  b  r and a  k  b  r . Again, if k  k , then r  r

and the representation is unique. If k  k , we will get the exactly similar contradiction, as we got
earlier (in the uniqueness for natural numbers).

66
 Integer Numbers

Decimal notation
Reminder. We have introduced earlier the natural numbers 1, 2, 3, ...., 10 , and the set 
can be written like:   {1, 2, 3, ...., 10, n(n(n(n(n(n(n(n(n(n(1)))))))))) .......}.
Lemma1. For any natural number k the next equality is true:
9  10k  9  10k 1  ...  9  101  10  10k 1 .
Proof (by induction). When k  1 we need to check: 9  101  10  102 , let’s simplify the left part, by
using the distributive law: (9  1)  10  10  10  10 - it is a true equality.
2

If the equality is true for a natural number k , then let’s consider

9 10k 1  9 10k  9 10k 1  ...  9 101  10  9 10k 1  (9 10k  9 10k 1  ...  9 101  10) 
9 10k 1  10k 1  (9  1) 10k 1  10 10k 1  10k  2 and the equality is true for k  1  n(k ) ,
then the equality is true for every natural number k .

Lemma2. For any natural k   there exist some n   such that 10n  10  k  10n 1 .
Proof. Let k  1, then n  1 is appropriate, really 10  10  1 . Let’s check that 10  1  1011 - it
is equivalent to 1  10  10  1  10  (10  1)  1  10  9  1  1  10  9 , the last equality is true,
2

because 1  9, 1  10 . If our assertion is true for a natural number k , then let’s consider it for

(k  1) . As the assertion is true for k , then 10n  10  k  10n 1 for some n   . As 10  k  10n 1
there can be the next cases: [A] If 10  k  10
n 1
 1, then 10  (k  1)  10n 1 and
10n 1  10  (k  1)  10n  2  10n 1  10  (k  1)  10n  2 .
[B] If 10  k  10n 1  1, then 10  k  10n 1  1, then 10  (k  1)  10n 1 and also
10n  10  k  10  (k  1) , therefore 10n  10  (k  1)  10n 1 .
Our assertion is true for (k  1)   , therefore it is true for any natural number k .

The main theorem. For any a   || a  10 there exist the unique representation as the sum:
a   n10 n   n 110 n 1  ...  110   0 , where  n ,  n 1... 1,  0 are integer numbers:
1   n  9, 0   n 1  9, .... 0  0  9 .
[Existence of representation]. As a  10 , then a  10  k , where k   .
Assertion. For any natural number 10  k || k   there exist the representation from the main
theorem. Let k  1, then 10  1  110  1 || 1  1,  0  1. Let our assertion is true for some
numbers 1,2.... k . Let’s consider the number 10  (k  1) . The number 10  k can be represented as:
10  k   n10n   n 110n 1  ...  110   0 , then 10  (k  1)   n10n   n 110n 1  ...  110   0  1.

[A] If 0   0  8 , then 1  ( 0  1)  9 and we can designate 0  ( 0  1) , then

10  (k  1)   n10n   n 110n 1  ...  110  0 , it is the representation we need.

67
 Integer Numbers

[B] If  0  9 , then there can be the next cases:

[B1] All the other numbers 1,  2 ...  m ...  n (from the representation for 10  k ) are all equal to 9 ,
then 10  k  9  10n  9  10n 1  ...  9  10  9 , then
10  (k  1)  9  10n  9  10n 1  ...  9  10  9  1 
9  10n  9  10n 1  ...  9  10  10  / lemma 7 /  10n 1 .
Then 10  (k  1)  1  10n 1  0  10n  ...  0  10  0 ||  n 1  1,  n  0...  0  0 - it is the
representation we need for 10  (k  1) .
[B2]  0  9 and maybe several numbers 1,  2 ...  m (from the representation for (10  k ) ) are all
equal to 9 , but  m 1  9  0   m 1  8 .
Then (10  k )   n10n  ...   m 110m 1  9  10m  9  10m 1...  9  10  9 and therefore

10  (k  1)   n10n  ...   m 110m 1  9  10m  9  10m 1...  9  10  9  1  / lemma 7 / 

  n10n  ...   m 110m 1  10m 1   n10n  ... ( m 1  1)10m 1 , then
10  (k  1)   n10n  ...  ( m 1  1)10m 1 , where 1  ( m 1  1)  9 . Let’s designate
 m 1  ( m 1  1) , then 10  (k  1)   n10n  ...   m 110m 1 and the representation we need for
10  (k  1)   n10n  ...  m 110m 1  0 10m  ...  0 10  0 .

Then for any natural number a  10  k || k   there exist the representation from the main
theorem. The existence is proved.

[Uniqueness of representation]. For any natural number a the representation (from above):
a   n10n   n 110n 1  ...  110   0 is unique.

Lemma3. There is a sum b  c  d  ...  m . If every summand in the sum is divisible by k   ,

then all the sum is also divisible by k.

Proof. As every summand is divisible by k, then b  b  k , c  c  k ... m  m  k .

Then b  c  d  ...  m  b  k  c  k  ...  m  k  (b  c  ...  m )  k - the last product is

divisible by k , then the initial sum is also divisible by k .

Proof of the uniqueness. Let’s assume that some natural number a has two representations:
n 1
a   n10   n 110
n
 ...  110   0 and a   m10   m 110 m 1  ...  110  0 , then
m

 n10n   n 110n 1  ... 110   0   m10m   m 110m 1  ... 110  0 .

Let’s show that  0   0 . Let  0   0 , then

 10
n
n
  
  n 110n 1  ... 110   m10m   m 110m 1  ... 110  0   0 the left part here is
divisible by 10 (according to the lemma3), the right part  9  0   0  9 (because 0  0  9 and
0   0  9 ). Among the numbers {9,8,... 1, 0, 1, 2.... 8, 9} only 0 is divisible by 10 (because
68
 Integer Numbers

0  0 10 ). Then 0   0  0  0   0 . Let’s return to the initial equality

 n10n   n 110n 1  ... 110   0   m10m   m 110m 1  ... 110  0 here we can subtract 0   0
from both sides, then  n10n   n 110n 1  ...  110   m10m   m 110m 1  ...  110 , let’s use the
distributive law for each part: 10  ( n10n 1   n 110n  2  ...  1 )  10  ( m10m   m 110m 1  ...  1 )
here we have the equality like 10  b  10  c , if b  c , then 10  b  10  c and if b  c , then
10  b  10  c , then b  c . So  n10n 1   n 110n  2  ... 1   m10m   m 110m 1  ... 1 and we have
the situation which is exactly similar to the initial one, then 1  1 and etc. Then we will get
 2   2 ,  3  3 ... after each step we subtract equal numbers  2   2 ,  3  3 ... from both sides.
Obviously, this process will finish soon. Let’s assume that we came to the situation when there are
no numbers left on the left or on the right part of the equality. Without loss of generality, let there
are no numbers on the left part, then 0   m10m   m 110m 1  ...   k . As 1   m  9 and
0   m 1  9 ... 0   k  9 , the right part is a positive natural number, but there is a zero on the left
side, and the equality is not true. Then the only possible case is when there are no numbers
left on the both sides of the equality, and it looks like 0  0.
Then  0   0 , 1  1 ,  2   2 ...  n   m  m  n and the representations:
a   n10n   n 110n 1  ...  110   0 and a   m10m   m 110m 1  ...  110  0 are exactly the
same. The uniqueness is proved. And the main theorem is proved.

Def. Let and a   || a  10 and a   n10 n   n 110 n 1  ...  110   0 [E]- the representation
from the main theorem. The sum on the right side can be denoted as:
 n n 1... 1 0  / by def /   n10n   n 110n 1  ...  110   0 and this symbol is called a decimal
notation of a. As  n n 1... 1 0 defines the sum which is equal to a , we can write
a   n n 1... 1 0 .
Sometimes, in order to avoid ambiguity, we can draw the line above the symbol  n n 1... 1 0 ,
because  n n 1... 1 0 can be misunderstood as a product of numbers  n ,  n 1... 1,  0 .

Def [extension of the previous one]. Let a   is an integer number. If a   , then it’s decimal
notation is the symbol  n n 1... 1 0 (from above). If a  0 , then 0 is the decimal notation of a
If a is a negative number, then a   where    , then the decimal notation of  with the
minus sign in front of it is the decimal notation of a .
a   ||   ,    n n 1... 1 0  a  // by def //  n n 1... 1 0
Conclusion: Any integer number a   can be uniquely represented as a combination of symbols
0,1,2,3...9 with or without minus sign "" in front of this combination. Such representation is called
a decimal notation of a .

69
 Integer Numbers

Divisibility
Exercise. [A] a, b, c,  . If a  b  c , then a is divisible by both b and c . [B] If a is divisible
by b and b is divisible by c , then a is divisible by c . [C] Let a  b  c , if some two of these
numbers are divisible by k , then the third (the other number) is also divisible by k .

Def. If natural numbers a and b do not have any common divisors except 1 , then we write
(a, b)  1. And in this case we say that a and b are coprime numbers. In general, if some number
d is the greatest common divisor of natural numbers a and b , then we write (a, b)  d .
Theorem7. The given definition is correct, for any natural numbers a, b   there always exist
a unique greatest common divisor d , moreover, there exist the pair of integer numbers  ,  such
that   a    b  d .

Proof. Let’s consider the set   {m  a  n  b || m, n  } of integer numbers, here a and b are
fixed and m, n can be any integers. The set  is closed (as a subset of  ) under addition/subtraction
and multiplication by any k   . Also a   ( m  1, n  0 ) and b   ( m  0, n  1 ). As a, b are both
positive, there exist a nonempty subset     of natural numbers. There exist the least number
d    , this number is obviously the least positive number in  . Let’s show that any element
x   is divisible by d . We fix any x   .
According to the 2-nd part of the theorem1 [division with a remainder], there exist the unique pair
of integer numbers k , r || 0  r  d such that x  k  d  r , then r  x  k  d .
Notice that x, d    x, k  d    x  k  d   . Then we have r   and 0  r  d . If r  0
(positive), then we have a contradiction (because r   and r  d , but d is the least positive
number of  ). Therefore r  0 and x  k  d , so x is divisible by d .
Notice that x is an arbitrary number from  .
Let’s take a, b   , then a  m  d and b  n  d and also d   , so there exist some (concrete) pair
of integer numbers ,    a    b  d . From a  m  d and b  n  d follows that d is
such that
a divisor of a and b , from the equality   a    b  d follows that there is no any other common
divisor of a, b which is greater than d . Really, if a and b are both divisible by some number
~ ~ ~
d  d , then the left part of   a    b  d is also divisible by d , then d is divisible by d which
~
is impossible (because d  d ). So, d is not just a common divisor of numbers a, b , but it is also the
greatest common divisor of these numbers d  d . And we also got the representation we need
 a   b  d .
Consequence. (a, b)  1    a    b  1 for some integers  ,    .

Def. A natural number p  1 is called a prime number (or just a prime) if p is divisible only by 1
and p . Obviously: p is not a prime number  p has some divisor b such that 1  b  p .
The number 2 is the least prime number and the only prime number which is divisible by 2 .
70
 Integer Numbers

Assertion6. For any natural number a  1 there exist at least one prime divisor p .

Proof. Ifa is a prime, then it is it’s own prime divisor, because a  a 1. Let a is not a prime, then
it has some divisor b || 1  b  a . Then the set   {all b   || ab, b  1, b  a} of all divisors of
a , which aren’t equal to 1 or a , is not empty, therefore it contains the least number mmin   .
The number mmin must be a prime number. Really, if mmin is not a prime, then it has some divisor
d || 1  d  mmin , then d is also a divisor of a (exercise [B] above), then d   and d  mmin and
we have a contradiction. So mmin is a prime divisor of a.

Exercise. The set P of all prime numbers is an infinite set.

Assertion7. Any natural number a  1 can be represented as a product of several prime numbers.
Proof. Let’s fix an arbitrary a   . According to the assertion6, a has some prime divisor p ,
then a  p  a . If a  1, then a  p is the representation we need. If a  1, then, according to the
assertion6, a has some prime divisor p , then a  p  a and a  p  p  a . If a  1 , then a  p  p
is the representation we need, if not, then we make the similar step as above.
We must show that this process must end at some moment.
Auxiliary. For any natural number a there exist k   such that a  2k . (It’s very easy to prove
by induction). Let’s find and fix any k   such that a  2k . Then in the process, which is described
above, there can’t be more than k prime numbers on the right side of a  p  p  p  ... .
Really, 2 is the least prime number and if there is more than k prime numbers on the right side,
then the product on the right side is greater than 2 , it means that a  2 , which contradicts
k k

to a  2 . So, the process, which is described above, must end for sure in
k
n  k steps.

We can also prove that for any a  1 it’s representation as a product of primes is unique up to
the order of prime numbers.

Let we have two representations: a  p1  p2  ... pn and a  p1  p2  ... pm .

Notice! When we have some representation, we do not multiply equal prime numbers, for example,
a  2  2  2  3 is an appropriate representation, but we can’t write it as a  23  3 , because 23 is
not a prime number.
We need to prove the auxiliary assertion.
Auxiliary. The product b1  b2  ... bk is divisible by a prime number p , then at least one of the
numbers b1, b2 ...,bk is divisible by p.

Proof. Let k  1, then b1 is divisible by p and the assertion is true.

Let’s make an untypical step and prove the assertion for k  2 (we will need it later).
So b1  b2 is divisible by p . If b1 is divisible by p , then the assertion is true, if not, then (b1, p)  1.

71
 Integer Numbers

Really, as p is a prime, 1 and p are the only divisors of p , therefore 1 is the only common divisor
of numbers b1 , p . As (b1, p)  1, then there exist the representation   b1    p  1 , let’s multiply by
b2 both sides of the last equality, we will get   (b1  b2 )    p  b2  b2 . The left part of the last
equality is obviously divisible by p , then the right part (which is b2 ) is also divisible by p .
Everything is proved in the case k  2 .

If the assertion is true for a number k , let’s consider it for k  1 . So b1  b2  ... bk  bk 1 is divisible
by p . Let’s designate P  b1  b2  ... bk . The product of two numbers P  bk 1 is divisible by p ,
then one of these numbers is divisible by p . If bk 1  p , then everything is proved.
If P p , then the product b1  b2  ... bk is divisible by p and (according to our assumption) one of
the numbers b1, b2 ..,bk is divisible by p . The auxiliary assertion is proved.

Let a  p1  p2  ... pn and a  p1  p2  ... pm , then p1  p2  ... pn  p1  p2  ... pm .

Let’s list all the numbers from the left part and form the set A  { p1, p2 ...pn } , and we also list all
the numbers from the right part and form the set B  { p1, p2 ...pm} .
Let’s take any prime number p j from the left part, the product p1  p2  ... pn is divisible by p j ,
then the product p1  p2  ... pm must be also divisible by p j . According to our auxiliary assertion,
one of the numbers p1, p2 ...pm is divisible p j , so some pi is divisible by p j .
If some prime number pi is divisible by a prime number p j , then obviously pi  p j .
We have shown: p j    p j   , therefore    . In the exactly similar way we can prove
that  p j    p j   , therefore    . Then    and the representations of a are exactly
the same (of course, up to the order of prime factors).

Def. G is any commutative additive group. We have already defined na || n   .

Now we also define: 0  a  0 || a  G . And for any negative number (n)   || n   we define:
 na  (na) || a  G .
a, b  G and for any m, n  
Then all the previous basic properties are still true. For any elements
[A] ma  na  (m  n)a [B] m(a  b)  ma  mb [C] mna  n(ma )  m(na)
[D]  (ma )  m(a) . The proof of each property must be done by considering all the possible
variants: one of m, n is a zero number, both are natural, both negative, one is positive and the other
one is negative. The proof is very simple, but a bit voluminous.

Def (similarly). G is a commutative multiplicative group. We have already defined a n || n   .

Now we also define: a  e . For any negative number (n)   || n   we define
0
а  п  (а п ) 1 .

For any elements a, b  G and for any m, n   the next is true:

[A] a m  a n  a m  n [B] (a  b) m  a m  b m [C] a mn  (a m )n  (a n ) m [D] (a m ) 1  (a 1 ) m .

72
 Rational Numbers

Rational numbers
Def. Q is the field that contains the ring  of integer numbers. Any element q  Q can be
a
represented as a ratio of some integers q  . The addition and multiplication on Q are
b
extensions of the addition and multiplication on  , i.e., for any a, b    Q :
a  b (addition by the rules of Q)  a  b (addition by the rules of ) ,
a  b (multiplication by the rules of Q)  a  b (multiplication by the rules of ) .
The field Q is called a field of rational numbers and elements of Q are called rational numbers.

Let’s build Q . We fix any set  of integer numbers. Let’s consider the set  of all pairs (a, b)
where a  , b   || b  0 are any integer numbers. We define addition and multiplication on .

Def. (a, b)  (c, d )  (a  d  b  c, b  d ) and (a, b)  (c, d )  (a  c, b  d ) .

. a
How to remember it? Imagine that any pair (a, b) denotes the ratio of elements a, b in some field.
b
a с ad bc
Then (a, b)  (c, d )    (a  d  b  c, b  d ) and
b d bd
a с aс
(a, b)  (c, d )     (a  c, b  d ) .
b d bd
Moreover, we say that pairs (a, b) and (с, d ) are equivalent (a, b)  (c, d ) if ad  bc.
It’s very easy to see that  is really an equivalence relation, because it is reflexive, symmetric,
transitive.

Assertion1. The addition "" and multiplication "" of pairs are both commutative and associative
on  . And instead of distributivity we have the equivalence:
(a, b)  (c, d )  (e, m)  (a, b)  (e, m)  (c, d )  (e, m) and
(e, m)  (a, b)  (c, d )  (e, m)  (a, b)  (e, m)  (c, d ).
Proof. As the multiplication of integer numbers is associative and commutative, the multiplication
of pairs is also commutative and associative. Next:
(a, b)  (c, d )  (a  d  b  c, b  d ) and (c, d )  (a, b)  (с  b  d  a, d  b) , then
(a, b)  (c, d )  (c, d )  (a, b) (the addition of pairs is commutative).
And (a, b)  (c, d )  (e, m)  a  d  b  c, b  d   (e, m)  (a  d  b  c)  m  b  d  e, b  d  m and
(a, b)  (c, d )  (e, m)  (a, b)  (c  m  d  e, d  m)  a  d  m  b  (c  m  d  e), b  d  m .
It’s easy to compare the results and see that the addition of pairs  is associative.

  
Let’s finally consider (a, b)  (c, d )   (e, m)  a  d  b  c, b  d  (e, m)  (a  d  b  c)  e, b  d  m 
and also the sum

74
 Rational Numbers

(a, b)  (e, m)  (c, d )  (e, m)  (a  e, b  m)  (c  e, d  m)  (a  e  d  m  b  m  c  e, b  m  d  m) 

 m  (a  e  d  b  c  e), m  (b  d  m) -this pair is obviously equivalent to (a  d  b  c)  e, b  d  m.

So, we have showed that (a, b)  (c, d )  (e, m)  (a, b)  (e, m)  (c, d )  (e, m) [P].
The other distributivity-equivalence follows from [P] and commutativity of the pair multiplication  .
Everything is proved.

The equivalence relation of pairs divides  into the classes A, B, C, D, E ... of equivalent pairs.
The set of all classes we denote like Qaux . Let’s define addition and multiplication on Qaux
in order to turn Qaux into a field.

Def. Let ,  are any classes from Qaux . And (a, b)  , (c, d )   are any pairs from these
classes. The sum    is the class which contains the sum of pairs (a, b) and (c, d ) , i.e., the pair
(a  d  b  c, b  d )  (a, b)  (c, d ) . The product    is the class which contains the product of
pairs (a, b) and (c, d ) , i.e., the pair (a  c, b  d )  (a, b)  (c, d ) .

Exercise1. Show that the given definition is correct. For any classes ,    aux , a sum   
and a product    are uniquely defined.

Exercise2. The set (Qaux ,,  ) (of all classes) is a field.

Solution. The addition of pairs is associative and commutative, therefore the addition of classes is
also associative and commutative on Qaux . The “zero” class  is the class which contains the pair
(0,1) . The zero class obviously consists of the pairs (0, b) || b  , b  0 . For any class  with a pair
(a, b) , an opposite class   is the class with the pair (a, b) . So Qaux is a commutative group with
respect to addition. The multiplication of classes is associative, because the multiplication of pairs
is associative. And finally, there is a distributive law for classes of Qaux , it follows from the
“distributive equivalence” for pairs, that is proved in the assertion1. Then (Qaux,,  ) is a ring.
We have to show now that all nonzero classes of Qaux form a commutative group with respect to  .
[step1] Qaux \  is closed under multiplication (a very important requirement). Really, let’s take
any ,   Qaux \  , both these classes are non-zero classes, then for any (a, b)    a  0 and for
any (c, d )    c  0 . The class    contains the pair (a  c, b  d ) and a  c  0
(because integer numbers a, b are not zeroes), then    is not a zero class, so     Qaux \  .
As the multiplication of pairs is commutative, the multiplication of classes  is also commutative
on all Qaux , as Qaux \  is closed under multiplication, the multiplication of classes  is commutative
on Qaux \  . And similarly, the multiplication of classes is associative on Qaux \  .

[step2] The class “one”  is the class which contains the pair (1,1) and obviously   Qaux \ .
Also, for any nonzero class   Qaux \  with a pair (a, b) (here a  0 because    ) an
1 1
inverse class  is the class with the pair (b, a) and obviously   Qaux \  (because b  0 ).

75
 Rational Numbers

Then Qaux \  is a commutative multiplicative group, and (Qaux,,  ) is a field.

Def: a class   Qaux with a pair (a, b) is called positive if a b  0 .

Exercise3. Show that:

[1] The given definition is correct. If a  b  0 for some pair (a, b)   , then с  d  0 for any pair
(c, d )   .
[2] For any class   Qaux only one of the next cases is true:
 is positive, or   is positive, or   0 .
[3] If classes ,  are positive, then both    and    are positive.

From the exercise3 immediately follows that Qaux is an ordered field (the order "" on Qaux
appears “automatically” by the rule:    in Qaux if    is a positive class).

Exercise4. For any integer number k   there exist the unique class   Qaux , such that for
any (a, b)   we have a  k  b (let’s fix any k   and take the class  with the pair (k ,1) ,
for any other pair (a, b)   there must be (a, b)  (k ,1)  a  k  b ).

Let’s build the set Q . Any class   Qaux such that a  k  b (a, b)   (for some integer number

k   ) we call an integer class. Let Z classes is the set of all integer classes of Qaux .
Every integer class   {(k  b, b) || k   (is fixed), b  , b  0} we replace by k   ,
we will get the set Q    (classes of pairs) .
The set Q is called a set of rational numbers, it consists of integer numbers and classes of
equivalent pairs. Let’s define addition and multiplication on Q in order to turn Q into a field.
Z classes (classes of pairs)
We define the mapping f : Qaux  Q    .
Z (classes of pairs)
For any integer class   {( k  b, b) || k   (is fixed), b  , b  0} we define f ( )  k and
f ()   for any other class   Qaux . Then f is obviously one-to-one mapping Qaux  Q .

Any element a  Q can be uniquely represented as an image of some class a  f () ||   Qaux .
a  b  f ( )  f ()  / by definition /  f (   )  Q
Then we define a, b  Q  .
a  b  f ( )  f ()  / by definition /  f (   )  Q

From the theorem4 (“Groups, Rings, Fields”) immediately follows that (Q, ,  ) is a field.

According to the theorem5 (Transfer of order), Q is an ordered field. The field (Q, ,  ) is called
a field of rational numbers. Let’s explore the properties of this field.

76
 Rational Numbers

Property1. The addition and multiplication on Q are extensions of the addition and
multiplication on  , i.e., for any a, b    Q we have:
a  b (addition by the rules of Q)  a  b (addition by the rules of ) ,
a  b (multiplication by the rules of Q)  a  b (multiplication by the rules of ) .
And similarly, the order relation "" on Q is an extension of the order relation on  :
a  b (by the rules of Q)  a  b (by the rules of ) .

Proof. Let’s denote for a while (~

,~ ) - the addition and multiplication on the field Q and (, ) - the
addition and multiplication on the ring .
Let’s fix arbitrary integer numbers a, b    Q . In order to find the sum a ~
 b (in Q ) we need to
find the classes ,  (from Qaux ) such that f ()  a, f ()  b then, by definition,
a~
 b  f () ~
 f ()  f (  ) . Obviously, A is the class with the pair (a, 1) and B is the class
with the pair (b, 1) . Then    contains the pair (a, 1)  (b, 1)  (a  b, 1) , therefore    is
an integer class and f (   )  a  b . So a ~
 b  a  b for any a, b    Q .
Next, a ~ b  f ( ) ~ f ()  f (   ) and    is the class which contains the pair
(a, 1)  (b, 1)  (a  b, 1) , then    is an integer class and f (  )  a  b ,
so a ~ b  a  b for any a, b    Q . And finally, let a  b (in Q)  f ()  f () (in Q) 
 // f conserves the order " " //     (in Qaux ) , then the class (  )    () is
positive in Qaux . Here (a, 1)  , (b, 1)  ()  the class   () contains the
pair (a  b, 1) , as this class is positive, there must be (a  b) 1  0  a  b  0  a  b (in ) .
Conversely, let a  b (in ) . Let’s consider the class  such that (a, 1)   and the class  such
that (b, 1)   . The class    , which contains the pair (a  b,1) , is positive, then
   (in Qaux )  f ()  f () (in Q )  a  b (in Q) .

Conclusion. The order relation/ addition/multiplication on Q are all extensions of the order relation/
addition/multiplication on Z. Then it’s appropriate to use exactly the same signs (  ) (as we

used on  ) to denote these operations on Q . Remember that (  ) on Z are extensions of

(  ) on  . Then (  ) on Q are also extensions of (  ) on  . All natural numbers are
positive (greater than zero) in Q , because they are positive in Z.

As Q is a field, division ":" is defined on Q (the only restriction, we can’t divide by zero).

m
If a, m, n  Q and a  , then we say “ a is a ratio of numbers m and n ”.
n
77
 Rational Numbers

Property2. Every element of a  Q can be represented as a ratio of some integer numbers

m
a || m, n   (and such representation is not unique).
n
Proof. Let’s fix an arbitrary a  Q , it has the unique representation a  f ( ) ||   Qaux .

Let’s take any pair (m, n)   . Obviously (n,1)  (m, n)  (m,1) , then     C , where  is the
integer class with the pair (n,1) and C is the integer class with the pair (m,1) . Then
m
f (  )  f (C )  f ()  f ()  f (C )  n  a  m (in Q)  a  . This representation is not
n
m
unique, really if a || m, n   , then (according to the basic properties of any field)
n
qm k m
a for any q  Q || q  0 and in particular a for any k    Q || k  0 .
qn k n

Property3. Q is a minimal field that contains Z. It means that:

any other field F, which contains Z, contains Q as a subfield. (as earlier, when we say that some
field F contains Z, we mean not only that F contains the set Z, but also that addition and
multiplication on F are extensions of the addition and multiplication on Z).

Proof. Let some fieldF contains Z. Then F must contain all the ratios of elements from Z, i.e.,
~ m  ~ ~
all the elements Q   || m, n  , n  0 . So F contains the set Q . Obviously   Q ,
n 
a ~
because a   : a   Q .
1
~
The set Q consists of exactly the same elements as a field of rational numbers Q (property 2).

~
Let’s show that the set Q is a subfield of F. According to the basic properties of any field
(which we mentioned earlier), we have:
m k m p  nk m k mk m k m p
  ,   , :  . So, the sum/difference/product/ratio of any
n p n p n p n p n p nk
~ ~ ~
elements of Q belong to Q . Therefore Q is a subfield of F (subfield criterion).
~
Then Q is an independent field inside F. And finally, as addition/multiplication on F are
~
extensions of the addition/multiplication on Z, then addition/multiplication on Q   are
extensions of the addition/multiplication on Z.
~
Then, by definition, Q is a field of rational numbers and a subfield of F.

78
 Rational Numbers

Uniqueness. We have built the field of rational numbers Q , based on some set Z of integer
numbers which we have fixed at the very beginning. There exist different sets of integer numbers
(but all of them are isomorphic as rings). Let’s prove:

The uniqueness theorem. Any fields Q and Q of rational numbers are isomorphic.
Moreover, the isomorphism f : Q  Q is unique, this isomorphism is an extension of the unique
isomorphism f :     .

Proof. Existence. Let’s extend the unique isomorphism f :     up to f : Q  Q .

m f (m )
We define: for any a Q if a  || m , n    , then f (a )  || f (m ), f (n )   .
n f (n )
At first, we need to show that this definition is correct, for any element a Q the element
f (a ) Q is uniquely defined, it follows from [A] and [B]. Secondly, we need to show that the new
isomorphism is really an extension of the old one, it follows from [C].
m f (m )
[A] In any representation a  the number n  0 , then in f (a )  , the number
n f (n )
f (n )  0 (because f :    is a ring isomorphism, and it transfers zero into zero f (0 )  0 ).
From here follows that for any a Q the element(s) f (a ) Q exists.
m ~
m
[B] If a  and a  ~  are different representations of a Q , let’s show that
n n
f (m ) f (m ~ )

f ( a )   ~ (from here will follow that for any a Q the element f (a ) Q is
f (n )
 f (n )


m m ~
unique). So  ~  m  n~  n  m~ the last equality is an equality of integer numbers in

n n
  , and f is a ring isomorphism f :    , then
~ )  f (n  m
f (m  n ~ )  f (m )  f (n
~ )  f (n )  f (m
~ ) the last equality is an equality in Q ,
       
~
here f (n )  0 and f (n~ )  0 (because n  0 , n~  0 ), then f (m )  f (m ) .
     
f (n )  f (n~ )

After [A] and [B] we can say that f is a mapping Q  Q .

[C] The new mapping f coincides with the old mapping f on the set   Q . For the initial
isomorphism we had m     m 
f
f (m )   , according to the new rules we need to
m
represent m as a ratio of any integers, and then we can find it’s image. Let’s take m  ,
1
m f (m ) f (m )
then according to the new rules, m  
f
  f (m ) .
1 f (1 ) 1
So, the new mapping coincides with the initial mapping on   .

79
 Rational Numbers

Let’s show that f : Q  Q is one-to-one. By definition of f , we have

 m  f (m )
f     || m , n    .
 
n f ( n )
[Step1] f covers all Q . Let’s fix any a Q , there exist the representation

m
a  || m , n    . As f :    is one-to-one, then there exist unique
n

m f (m ) m 
m , n   || f (m )  m , f (n )  n , so a    f    and a is an image of
n f (n )  n 
m
the element  a  Q .
n
[Step2] f doesn’t glue together elements of Q . Let for some elements a , b Q we have:
a  b , but f (a )  f (b ) . Both elements a , b have representations
m k f (m ) f (k )
a  , b   || m , n , k , p   . Then f (a )  and f (b )  .
n p f (n ) f ( p )
f (m ) f (k )
As f (a )  f (b ) , then   f (m )  f ( p )  f (n )  f (k ) the last equality can
f (n ) f ( p )
be rewritten like f (m  p )  f (n  k ) (because f :     is a ring isomorphism), as f is
one-to-one, then m  p  n  k -it is an equality in   , in the same time it is an equality in Q ,
m k 
from which follows that  (notice that n  0 , p  0 , because f (n )  0 , f ( p )  0 )
n p
and the last equality means that a  b , and we have a contradiction.

So f : Q  Q is one-to-one. Let’s finally show that f is a field isomorphism:

a , b : f (a  b )  f (a )  f (b ) and f (a  b )  f (a )  f (b ) . Let’s fix arbitrary
m k
a , b Q , we fix any representations a  , b   . Then
n p

m k   m  p  n  k  f (m  p  n  k ) f (m  p )  f (n  k )

f (a  b )  f       f       
 n p   n  p  f (n  p ) f (n  p )
f (m )  f ( p )  f (n )  f (k ) f (m ) f (k ) m  k 
    f     f     f (a )  f (b ) .
f (n )  f ( p ) f (n ) f ( p )  n   p 
m k  m k  f (m  k ) f (m )  f (k ) f (m ) f (k )
And f (a  b )  f       f       
  
 
 

 
n p   
  
n p f ( n   p  ) f ( n  )  f ( p  ) f ( n  ) f ( p  )
m  k 
 f     f     f (a )  f (b ) . The existence of isomorphism f : Q  Q is proved.
 n   p 

80
 Rational Numbers

Uniqueness. We have already built f : Q  Q which is an extension of f :    .

Let  : Q  Q is a field isomorphism, we will show that   f on Q .

As  : Q  Q is a field isomorphism, then it transfers zero into zero and one into one:  (0 )  0
and  (1 )  1 . Let’s take    Q and   Q . Let’s show that  is an isomorphism of the
rings     .

At first we need to show that all the values of  on   belong to  .

We have:       0    and      0    .

So       0    . If a    , then a can be represented as a sum of ones,

a  1  1  ...  1 , then
 (a )   (1  1  ...  1 )   (1 )   (1 )  ...   (1 )  1  1  ...  1  a     .
If a  0 , then  (0 )  0   . If a    , then a  a~ || a~    , then
a  a~  (1  1  ...  1 ) , then
 (a )   (a~ )   ((1  1  ...  1 ))   (1  1  ...  1 )   (1 )   (1 )  ...   (1 ) 
 1  1  ...  1   a     . Only now we can write  :     and say that  is a
mapping from   to  .

[Step1] Let’s show that  :    covers all the set  . So     0    .

Let a   , there can be several variants. If a    , then a can be represented as a sum of

ones, so a  1  1  ...  1   (1 )   (1 )  ...   (1 )   (1  1  ...  1 )   (a ) || a   

and a is the image of a    .

If a  0 , then  (0 )  0 and 0 is the image of 0    .

~  (1  1  ...  1 ) 
If a    , then a  a   
 ( (1 )   (1 )  ...   (1 ))   (1  1  ...  1 )    (a ) || a      (a ) || a   
and a is the image of  a    . So, any element a   is an image of some element from   .
[Step2]  doesn’t glue together elements of   , because  doesn’t glue together elements of Q
(really,  is an isomorphism  : Q  Q , then it is one-to-one mapping).

From the [Step1] and [Step2] follows that  :    is one-to-one.

[Step3] For any a , b    we have  (a  b )   (a )   (b ) and  (a  b )   (a )   (b ) .
Really, these conditions are true for any a , b  Q , because  : Q  Q is an isomorphism,
and in particular these conditions are true for any a , b    .

From the [Step3] follows that  is an isomorphism of the rings     . And we have proved that
there exist only one isomorphism f :     , then   f on   . Let’s show   f on Q .

81
 Rational Numbers

m
Let’s fix an arbitrary a Q and any it’s representation as a ratio of integers a  .
n
f (m )
We know that f (a )  . Obviously f (m )   (m ) (because m    ) and similarly
f (n )
f (m )  (m )
f (n )   (n ) (because n    ), then f (a )   . Let’s show that  (a ) is also
f (n )  (n )
 (m )
equal to , then  (a )  f (a ) for any a Q and   f on Q .
 (n )
m
So a   a  n  m the last equality is an equality in Q and  : Q  Q is one-to-one
n
mapping, then  (a  n )   (m ) , as  is an isomorphism, then
 (a  n )   (m )   (a )   (n )   (m ) .
 (m )
Let’s notice that  (n )  0 because n  0 , then from the last equality we have  (a )  .
 (n )
Everything is proved.

Consequence. For any fields of integer

numbers Q , Q there exist the unique
isomorphism f : Q  Q which is an
extension of the unique isomorphism
f :    (where   Q ,   Q )
which, in it’s turn, is an extension of the
unique isomorphism f :      pict.1
(where       Q ,      Q )
[pict1].
Property4. The Archimedes axiom is true in Q .
Proof. Let’s fix any positive numbers a, b  Q . We want to show that there exist n    Q such

1 b b
that n  a  b . The element is positive, then the last inequality is equivalent to n  , where
a a a
is a positive element of Q . Let’s show that for any positive element q  Q there exist the natural
n   such that n  q (and everything will be proved). We fix an arbitrary positive q  0 , then
k k
q || k , p   . The fraction is positive, then k  p is positive, then k  p is a natural number.
p p
k
Then k  p  1, let’s multiply both sides of k  p  1 by 0 
p
k k k
 k  p   1  k 2   k 2  q || k 2   , everything is proved.
p p p
82
 Rational Numbers

Property5. For any rational number q  Q there exist the unique pair of integer numbers
k , k  1 such that k  q  k  1.

Existence. Let’s fix an arbitrary positive rational number q . Let’s consider the pair 1, q .
If q  1, then we have 0  q  1 and the pair of integers k , k  1  0,1 is appropriate.
If q  1, then 1 1  q and the set of natural numbers   {m || m  1  q} is not empty.
According to the Archimedes axiom, there exist the natural number n such that n 1  q .

The number n is greater than any number of the set  , therefore  has the greatest number
M max . And for M max we obviously have M max  1  q  (M max  1)  1  M max  q  (M max  1) so,
here we have the pair k , k  1  M max , ( M max  1) . Let now q  Q is negative, then  q is positive,
then, as we showed above, we can find the pair of integers k , k  1 such that
k  q  k  1  (k  1)  q  k . If q  k , then we have  (k  1)  q  k and  (k  1),k is
our pair of integers. If q  k , then we have  k  q  k  1 . And finally, let q  0 , then we take
k , k  1  0,1. The existence is proved.

Uniqueness. Let k  q  k  1[1] and m  q  m  1 [2] here k  m . Without loss of generality

k  m , then m  k is a positive (natural) number, so m  k   ||     m  k   .
Then from [2] we have k    q and from [1] we have q  k  1 . Then

k    q  k  1  k    k  1    1 and we have a contradiction, because    .

The uniqueness is proved.

Let’s name the property5 as “any rational number lies between two integers”.

83
 Sequences and limits

Sequences and limits

Def.  is any set and  is a set of natural numbers. Any function f :    is called a
sequence of  , or just a sequence. We will always write xn instead of f (n) (so, for any concrete
n   : f (n)  xn ). And we will write {xn } , or x1, x2 , x3 , x4 ..... instead of f .
For any concrete n   , the element xn is called an n-th term of the sequence {xn } .
From now on   F , where F is any ordered field which contains the field of rational numbers Q
as a subfield (in particular, there may be M  Q ).

Def. Let a is any element of F , for any   0 ||   F the set O (a)  {x  F || x  a   }

is called an  -neighborhood of a , or just a neighborhood of a . A positive element  is called
a radius of O (a) .
Def. {xn } is a sequence of F . If there exist some element a  F such that any  -neighborhood
O (a) (of a ) contains all the terms of the sequence {xn } starting from some number k, i.e.,
O (a) contains all the terms xk 1 , xk  2 , xk  3 ...., then we say “{xn } goes to a ”, or
“ {xn } has a limit a ”, or “ {xn } converges to a ”, or just “ {xn } converges”, and we write {xn }  a .
In other words, {xn }  a if for any   0 there exist some concrete natural number k
(which depends on  ) such that n  k  xn  O (a) . The last condition can be rewritten like
n  k  xn  a   .
The sequence may not have any limit at all, for example the sequence 1,0,1,0,1,0 ..... does not have

any limit in F . When {xn } has no limit, we say “ {xn } diverges”.

Assertion1. If {xn } has some limit a , then this limit is unique.

Proof. Let’s assume that it’s not true, then there exist some other limit b  a . Let’s fix any
neighborhoods of a and b that do not intersect. For example, the neighborhoods O (a), O (b)
with radiuses   b  a / 3 do not intersect (check it, by using the standard inequality
x  y  x  y x, y  F ). Then O (a) must contain all the terms of {xn } , starting from some

number k , and O (b) must contain all the terms of {xn } starting from some number p , then
we have a contradiction. Really, any n-th term xn , where n  k , n  p , must belong to O (a) and
to O (b) , but these neighborhoods do not intersect. So, any sequence {xn } may have only one limit.

Def. The sequence {xn } is called bounded if there exist some positive M  0 such that
xn  M || n . 85
 Sequences and limits

Assertion2. If {xn } converges (has some limit a ), then {xn } is bounded.

Proof. So {xn }  a . Let’s fix an arbitrary positive   0 , then there exist k   such that all the

terms xk 1 , xk  2 , xk  3 ... lie in O (a) . And the other terms x1 , x2 .. xk are not necessary from O (a) ,
let V  maxx1 , x2 .... xk , a    so V is the greatest among several non-negative elements which
are listed inside the brackets { ..... }. Obviously V  0 , because a    0 . Let’s show that for any
xm  {xn } we have xm  V  1. If xm is one of the elements x1 , x2 .. xk , for example xm  x2 , then
xm  x2  V  V  1 . Let xm is one of the elements xk 1, xk  2 , xk 3...., then xm belongs to O (a) ,
so xm  a   , from the standard inequality for absolute value we have:
xm  a  xm  a    xm  a    V  V  1.
So, for any element xm of our sequence, we have xm  V  1 . As V  0 , then V  1  0 .
Let’s denote M  V  1 , then xm  M || m and the sequence {xn } is bounded, everything is
proved.
The converse statement is not true. If {xn } is bounded it may not have any limit.
Example. 1,0,1,0,1,0 .....

Def. Any sequence { n } which goes to zero ({ n }  0 ) is called an infinitely small.
Auxiliary properties. [A] {xn }  a  the sequence { n }  {xn  a} ||  n  xn  a n is
infinitely small. And conversely, if some sequence {xn } has the representation xn  a   n || n ,
where { n } is an infinitely small sequence, then {xn }  a .
[B] {xn } is bounded (in particular {xn } converges) and { n } is infinitely small, then {xn   n } is
infinitely small.

[C] { n } and { n } are infinitely small, then { n   n } is infinitely small.

[D] If {xn }  a , then { xn }  a .

1 1
[E] {xn }  a and xn  0 n and a  0 , then   .
 n a
x
Proof. Let’s prove [A]. Let we have {xn }  a . Then, for any   0 there exist
k   : n  k  xn  a   , we can rewrite the last inequality
xn  a    ( xn  a)  0     n  0   .
So, we have: for any   0 there exist k   : n  k :  n  0   , by definition it means that
{ n }  0 .

86
 Sequences and limits

Consequence1. Let {xn }  a , then the sequence { n }  {xn  a} is infinitely small and there is
a simple representation: xn  a   n || n , where { n }  0 .
Let now we have some sequence {xn } and the representation xn  a   n || n , where { n }  0 .
As { n }  0 , then for any   0 there exist k   : n  k   n   . Let’s rewrite the last
inequality:  n    (a   n )  a    xn  a   . So, for any   0 there exist
k   : n  k  xn  a   , by definition it means that {xn }  a .

Let’s prove [B]. So,  M  0 such that xn  M || n . Let’s fix an arbitrary positive   0 .
Then for the positive element  /M (as for any other positive element) there exist
k   : n  k :  n  0   / M   n   / M , then n  k we have
xn   n  0  xn   n  xn   n  M   / M   . So, for any positive   0 there exist
k   : n  k  xn   n  0   , it means that {xn   n } is infinitely small.
Consequence2. For any element a  F and for any infinitely small { n } , the sequence {a   n }
is infinitely small. Really, the sequence {xn }  a, a, a, .... is a bounded sequence, then
{xn   n }  {a   n } is infinitely small. Also: if { n } and { n } are infinitely small, then { n   n }
is infinitely small.

Let’s prove [C]. { n }  0 and { n }  0 . Let’s fix an arbitrary positive   0 . For  / 2  0 there
exist such k : n  k we have  n  0   / 2   n   / 2 . In the same time for  / 2  0 there
exist p : n  p we have  n   / 2 . Let m  max( p, k ) , then for any n  m we have  n   / 2
and  n   / 2 , then ( n   n )  0   n   n   n   n   / 2   / 2   . We have deduced
that for any   0 there exist m , such that n  m we have ( n   n )  0   .
Then { n   n }  0 and { n   n } is infinitely small.

Consequence3. A sum of several infinitely small sequences is again an infinitely small sequence.

Let’s prove [D]. As {xn }  a , then for any

  0 k   : n  k : xn  a    ( xn )  (a)    { xn }  a .
 1 1
Let’s prove [E]. According to [A], it’s enough to prove that the sequence    is infinitely
 xn a 
 1 1   a  xn 
small. Let’s rewrite     . As {xn }  a , then {xn } is bounded,
 n
x a   n 
x  a
let’s fix any M  0 such that xn  M . We also have a  0 .

87
 Sequences and limits

Let a  0 , we take the positive a / 2  0 . Starting from some number k , all the terms of {xn }
lie in a / 2 - neighborhood of a , so k   : n  k : xn  a  a / 2 .
Then  a / 2  xn  a  a / 2  a / 2  xn  3a / 2 . Let’s fix now any positive   0 .
We have {xn }  a , then for the positive number   a 2 / 2 (as for any other positive number) there
exist m   : n  m : xn  a    a 2 / 2 . Let’s take now the maximal p  max( m, k ) , then for

any n  p both conditions a / 2  xn  3a / 2 and xn  a    a 2 / 2 are true.

For any concrete n  p , let’s estimate:

a  xn a  xn a  xn xn  a   a 2 / 2   a 2 / 2
0        . So, we had fixed an
xn  a xn  a xn  a xn  a xn  a a / 2a
arbitrary positive  0 and we deduced that there exist the number p , such that for any n  p
a  xn  1 1 1 1
we have  0   , it means that    is infinitely small, then   .
xn  a  xn a   xn  a

Let now a is negative, then a   a , where a is positive. We have {xn }  a  {xn }   a ,

 1  1  1 1
a  0 , there must be 
according to [D], we have { xn }  a , as       .
  xn  a  xn  a
1 1 1 1 1 1 1
And again, by using [D], we get        , so we have    .
 xn  a  xn   a a  xn  a
Everything is proved.

Basic properties of convergent sequences. If {xn }  a and { yn }  b , then

 xn  a
[1] {xn  yn }  a  b , [2] {xn  yn }  a  b , [3] if yn  0 n and b  0 , then   .
 n b
y
Proof: We will use the auxiliary properties from above. As{xn }  a and { yn }  b , then the next
representations are true: xn  a   n || n and yn  b   n || n , where { n } and { n } are
infinitely small sequences.

Let’s prove [1]. We have xn  yn  (a  b)  ( n   n ) || n . According to [C], the sequence

{ n   n } is infinitely small, then (according to [A]) {xn  yn }  a  b .
Let’s prove [2]. So xn  yn  (a   n )  (b   n )  a  b  (a   n   n  b   n   n ) .
According to [B], all the sequences {a   n }, { n  b}, { n   n } are infinitely small, then (acc. to

[C]), the sequence {a   n   n  b   n   n } is infinitely small, then (acc. to [A]) {xn  yn }  a  b .

88
 Sequences and limits

1 1
Let’s prove [3]. Asyn  0 n and b  0 and { yn }  b , then (acc. to [E])    .
 yn  b
 1 1  xn  a
We also have {xn }  a , then (acc. to [2]) we have  xn   a    .
 yn  b  yn  b
Consequence4. {xn }  a and { yn }  b , then { yn  xn }  b  a . Really, {xn }  a , then [D]
{ xn }  a , then (acc. to [1]) { yn  ( xn )}  b  (a)  { yn  xn }  b  a .
In particular, a difference of infinitely small sequences is an infinitely small sequence.

Consequence5. Let {xn }  a and { yn }  b . If the sequence { yn }  {xn }  { yn  xn } is

infinitely small, then a  b . Really, according to the consequence1,{ yn  xn }  b  a .
As { yn  xn } is infinitely small, then b  a  0  b  a .
Consequence6 (advanced6). { yn }  {xn }  { yn  xn } is infinitely small. And one of the sequences
{xn }, { yn } converges to a , then the other sequence also converges to a . Really, let for example
{xn }  a . Let’s denote { n }  { yn  xn } (for clarity), then for any n we have the representation
yn  xn   n . We know that {xn }  a and { n }  0 , then from [1] we have:
{xn   n }  a - it is equivalent to { yn }  a .
Consequence7. Let {xn }  a , then for any b  F we have {bxn }  ba . Really, let’s take the
sequence { yn }  b, b, b, b ..... Then (acc. to [2]) {xn  yn }  {xn  b}  a  b  {bxn }  ba .

Let’s sum up. We have deduced the basic properties of convergent sequences: {xn }  a and

{ yn }  b , then {xn  yn }  a  b and {xn  yn }  a  b and {xn / yn }  a / b (the last one if

yn  0 n and b  0 ). And we can also multiply any convergent sequence by any constant:
{xn }  a  b  F : {bxn }  ba .

Let’s consider the set  which consists of all infinitely small sequences { n } of the field F.
From the previous results follows that  is closed under addition/subtraction/multiplication/
multiplication by a constant, i.e., if { n } and { n } are infinitely small, then { n   n } and
{ n   n } and {b   n } b  F are infinitely small.
We will provide the other set of sequences  which has exactly the same properties, the set of
all fundamental sequences of F .

Def. A sequence {xn } is called fundamental if for any   0 there exist k   such that
m  k , n  k we have xm  xn   .

Def. A sequence {xn } is called a stationary sequence if {xn }  a, a, a, a .... for some a  F .

89
 Sequences and limits

Any stationary sequence is obviously fundamental.

Assertion3. {xn } and { yn } are fundamental sequences, then:

[1] {xn  yn } is fundamental [2]{xn  yn } is fundamental.

Proof. Let’s prove [1]. We fix an arbitrary   0 . For the positive element  / 2  0 there exist

k   such that m, n  k  xm  xn   / 2 . Similarly for  / 2  0 there exist p   such that

m, n  p  ym  yn   / 2 . Let’s take v  max( k , p) , then for any n  v we have:
xm  xn   / 2 and ym  yn   / 2 . Then for any n  v we have
( xm  ym )  ( xn  yn )  ( xm  xn )  ( ym  yn )  xm  xn  ym  yn   / 2   / 2   .
Then for any positive  0 we can find v   such that for any n v:
( xm  ym )  ( xn  yn )   it means exactly that {xn  yn } is fundamental.

Proof. Let’s prove [2]. Auxiliary. If {xn } is fundamental, then {xn } is bounded.
[step1] Let’s fix any   0 . Then k   m, n  k  xm  xn   .
Let’s fix any number p  k , then for any m  p we have xm  x p   (because anyway, both
numbers 
xm , x p are greater than k ). Let’s take V  max x1 , x2 .... x p , x p   . 
And let’s take any element xm  {xn }. If xm is one of the elements x1 , x2 .. x p ,
then obviously xm  V  V  1 . If xm is one of the elements x p , x p 1 , x p  2 ...., then xm  x p   .
Let’s use the standart inequality for absolute value: xm  x p  xm  x p   , then
xm  x p    xm  V  V  1 .

So, for any element xm  {xn } we have xm  V  1 . Obviously V  0  V  1  0 .

We can denote M  V  1, then xm  M || m and {xn } is a bounded sequence.

[step2] Let now {xn } and { yn } are fundamental, then they are both bounded:
M  0 : xm  M || m and T  0 : ym  T || m . Let’s take V  max( M , T ) , then xm  V
and ym  V for any m . Let’s fix an arbitrary positive   0 . As { yn } is fundamental, for  / V  0
there exist k   : m, n  k  ym  yn   / 2V . As {xn } is fundamental, for  / V  0 there
exist p   : m, n  p  xm  xn   / 2V . Let’s take the number v  max( k , p) , then for any
m, n  v we have ym  yn   / 2V and xm  xn   / 2V . Then for any m, n  v we have:
xm  ym  xn  yn  ( xm  xn )  ym  ( ym  yn )  xn  ( xm  xn )  ym  ( ym  yn )  xn 
 xm  xn  ym  ym  yn  xn  ( / 2V )  V  ( / 2V )  V   / 2   / 2   .
90
 Sequences and limits

Let’s sum up: we had started from an arbitrary positive   0 , and we showed that there exist some
number v , such that m, n  v we have xm  ym  xn  yn   , it means that {xn  yn } is
fundamental.

Consequence8. If {xn } is fundamental, then for any b F the sequence {b  xn } is fundamental.

Really, let’s take the stationary sequence { yn }  b, b, b, b .... (this sequence is fundamental), then
(according to [2]) {xn  yn }  {bxn } is fundamental.
Consequence9. If {xn } and { yn } are fundamental, then {xn }  { yn }  {xn  yn } is fundamental.
Really, as { yn } is fundamental, then (consequence8 b  1  Q  F ) the sequence { yn } is
fundamental. Then (according to [1]) {xn  ( yn )}  {xn  yn } is fundamental.
Let’s sum up. The set  of all fundamental sequences of F is closed under
addition/subtraction/multiplication/multiplication by a constant, i.e, if {xn } and { yn } are
fundamental, then {xn  yn },{xn  yn }, {bxn } b  F are fundamental.

Assertion4. If {xn } converges (to some limit a ), then {xn } is fundamental.

Proof. Let {xn }  a . We fix an arbitrary   0 , then for  / 2  0 there exist k   such that
n  k  xn  a   / 2 . Then for any m, n  k we have:
xm  xn  ( xm  a)  (a  xn )  xm  a  a  xn   / 2   / 2   , then {xn } is a fundamental
sequence.

Let’s remind. All the results, that we got above, are related to sequences of a field F .
Where F is any ordered field which contains the field of rational numbers Q as a subfield.
(In particular, all this theory is applicable for F  Q ). The existence of an ordered field
F  Q || F  Q is still under a question. We still haven’t built yet any example of such field F ,
but when we do, we will be able to use in F all these results.

Let’s look at the assertion4. The next question has a great importance in math:
“ If {xn } is fundamental in F, does it converge to some a  F ?” [Q].
The general answer is “no”, or better “not necessary”. In general, it depends on the field F and on
the concrete sequence {xn } of F. But we are not interested in such randomness, we want to work
only in such fields, where the answer on [Q] is “Yes”. And we will build the complete fields R, C ,
where the answer is always “Yes”.

Def. A field F is called a complete field if any fundamental sequence {xn } of F converges in F.
Notice, the field of rational numbers Q is not a complete field. There are sequences of rational
numbers which are fundamental, but do not have any limit in Q . It’s much more convenient to
provide examples of such sequences after construction of the field of real numbers R  Q .

91
 Sequences and limits

Squeeze Theorem for sequences. Let {xn }, { yn }, {zn } are some sequences of F .
Starting from some number k we have xn  yn  zn || n  k .
And both sequences {xn }, {zn } go to the same limit a , then { yn } also goes to a .

Auxiliary assertion. There are positive sequences { n }, { n } and 0   n   n || n .

If { n } is infinitely small, then { n } is infinitely small.
Proof.   0 k, n  k   n   . From 0   n   n immediately follows that n  n .

Then   0 k, n  k   n     n   , so { n } is infinitely small.

Comment. Let’s notice a very important moment. From any sequence {xn } we can discard any finite
amount of it’s first consecutive terms x1 , x2 , x3 ....xk , it does not affect fundamentality and
convergence. Let’s explain it, when we say that we discard x1 , x2 , x3 ....xk from {xn } , we mean that
we consider the new sequence {~
xn } || ~
x1  xk 1 , ~
x2  xk 2 , ~
x  x , .....
3 k 3

It’s very easy to understand that: {xn } is fundamental  {~

xn } is fundamental.
And:{xn } converges to a  {~
xn } converges to a . This great idea will be used several times
during the construction of real numbers.
Let’s prove now the squeeze theorem. We can obviously discard all the terms of the sequences with
numbers 1,2,3...k (it will not affect convergence), then we can assume that from the very beginning
we have xn  yn  zn || n . Let’s subtract xn from all sides, then 0  yn  xn  zn  xn .
Let ( zn  xn )   n and ( yn  xn )   n , then we have 0   n   n . As {xn }  a and {zn }  a ,
then (consequence4) {zn  xn }  0  { n }  0 , then (auxiliary assertion above)

{ n }  0  { yn  xn }  0 . As {xn }  a , then (consequence6) { yn }  a .

Def. F is an ordered field. If the Archimedes axiom is true in F, then we say

“F is an Archimedean ordered field”.

Comment. The next chapter is dedicated to the construction of real numbers. If we write {qn }  Q ,

we mean that we have a sequence of rational numbers {qn } . This writing is used because it’s
convenient. But normally, the symbol  is used to show that one set belongs to the other.
In our case Q is a set, but {qn } is not a set, it is a function! f :   Q (look at the definition of
a sequence, page 85). And numbers q1 , q2 , q3 , q4 .... are only values of that function.
So, if someone wants to avoid misunderstanding, it’s better to write {qn } || qn  Q, n , or to make
an explanation of a new writing, like we just did.

92
 Real Numbers

Pre-real numbers
As earlier, F is an ordered field which contains the field Q as a subfield. We will prove the
auxiliary theorem that will help us in the construction proсess.

Obviously, if some sequence {qn }  Q  F is fundamental in F , then {qn } is fundamental in Q .

Really: {qn } is fundamental in F , then   F ||   0 k : m, n  k  xm  xn   .
Then a fortiori   Q ||   0 k : m, n  k  xm  xn   (because Q  F ).

The converse assertion in general is not true, if {qn }  Q  F is fundamental in Q , it mustn’t

be fundamental in F.

{qn }  Q  F converges to some {qn }  Q as a sequence of F , then it converges

And similarly, if
to q as a sequence of Q . And again, the converse assertion is not true.
If{qn }  Q converges to some q  Q as a sequence of Q , it mustn’t converge to q as a
sequence of F (obviously, there can be only two variants, either {qn } does not converge in F ,
or {qn } converges in F , and in this case, of course, {qn } converges in F to the same q  Q ).
Theorem1. The Archimedes axiom is true in F . Then:
[A] Every sequence {qn }  Q that converges/fundamental in Q also converges/fundamental in F .
[B] Every element a  F can be represented as a limit of some rational sequence {qn }  Q .
{qn }  Q which is fundamental in Q . Let’s fix an arbitrary
Proof. Let’s fix an arbitrary sequence
positive   F . We want to show that k   : m, n  k  xm  xn   . Both elements 1, 1 / 
~
are positive, according to the Archimedes axiom, there exist k   such that
~ ~ ~
k  1  1/   k    1    1/ k  Q . As {qn } is fundamental in Q , for the rational number
~ ~
1/ k  0 , there exist k   : m, n  k  xm  xn  1/ k , then m, n  k  xm  xn   .
So {qn } is fundamental in F.
Let’s fix an arbitrary sequence {qn }  Q which converges in Q to some element q  Q .
Let’s fix an arbitrary positive   F . We want to show that k   : n  k  xn  q   .
~ ~
In the same way as above, we will find the positive rational number 1 / k such that 1/ k   .
~
As {qn }  q in Q , there exist k   : n  k  xn  q  1 / k , then n  k  xn  q   .
So {qn }  q in F . And [A] is proved.

Let’s prove [B]. The proof here will be a bit more voluminous than usually, but we will build here
a very important example which we will use in the future in the “Length construction”.
As this example is important for us, we name it an [Example L].
Let’s fix an arbitrary positive a  F , there exist n   such that 1 / n  a (we have shown it above,
there was an arbitrary positive  instead of a ).

94
 Real Numbers

The set  of natural numbers m such that m / n  a is not empty (it contains m  1), in the same
time, there exist the natural number k such that k  1/ n  k / n  a (the Archimedes axiom for 1/ n, a ).
Then  has the greatest number m0
max
  . Then m0max / n  a  (m0max  1) / n .
Let’s sum up: m0
max
 T0 is the greatest natural number such that T0 / n  a and m0max  1  M 0 is
T0 M
the least natural number such that a  M 0 / n and T0  1  M 0 . So  a  0 || L0  1  M 0 .
n 2n

 1 
Let’s consider all the possible fractions  || k    . Each of these fractions is less than the
2 n 
k

1
initial fraction (from above) and therefore each fraction is less than a. In exactly the same way
n
1
as above, for each fraction k there exist the pair of natural numbers Tk , M k such that
2 n
Tk Mk Tk Tk 1
 a  || T  1  M . Let’s notice: [At first] For any integer k  0 we have  .
2k n 2k 1 n
k k
2k n 2k n
T 2  Tk 2T
Really, kk   k 1k  a we know that Tk 1 is the greatest among all the natural numbers
2 n 22 n 2 n k

m
m , for which k 1  a , then of course 2Tk  Tk 1 and as a consequence,
2 n
2Tk Tk 1 Tk Tk 1
   .
2k 1 n 2k 1 n 2k n 2k 1 n

M k 1 M k M k 2  M k 2M k
[Secondly] For any integer k  0 we have  . Really,    a and
2k 1 n 2k n 2k n 2  2k n 2k 1 n
m
we know that M k 1 is the least among all the natural numbers m , for which k 1  a ,
2 n
2M M M M
then, of course, 2M k  M k 1 and as a consequence k 1k  k k11  k k  k k11 .
2 n 2 n 2 n 2 n

T0 T1 T T M M M M
So, we have:   22  33  ....  a  ....  3 3  2 2  1  0 which is equivalent to
n 2n 2 n 2 n 2 n 2 n 2n n
T0 T1 T T T  1 T  1 T  1 T0  1 T 
  22  33  ....  a  ....  3 3  2 2  1  [E]. The sequence  kk  is
n 2n 2 n 2 n 2 n 2 n 2n n 2 n 
obviously a sequence of rational numbers, let’s show that it converges to a . From [E] we have
Tk Tk  1
 a  || k  0 .
2k n 2k n

95
 Real Numbers

Tk Tk Tk  1 Tk
Let’s subtract from all sides, then 0  a     for any integer k  0
2k n 2k n 2k n 2k n
T 1
we have 0  a  kk  k . We can apply here the squeeze theorem for sequences.
2 n 2 n


Tk   1 
Really, let’s consider {xk }  0,0,0,0,0 ..... and { yk }  a 
k 
and {zk }   k  we have

2 n 2 n 
xk  yk  zk || k and obviously {xk }  0 (in F ). The rational sequence {zk } obviously converges
to zero in Q , then (according to [A]) it converges to zero in F  Q . So both {xk }  0, {zk }  0 ,
then ( the squeeze theorem for sequences) { yk }  0 .

 Tk   Tk   Tk 
So, the sequence a   is infinitely small, then  k   a , where  k   Q and a is
 2k n  2 n  2 n 
an arbitrary positive element of F . So, any positive a  F can be represented as a limit of some
rational sequence. Let a is negative, then a   a where a is positive, therefore there exist some
{qn }  Q || {qn }  a , then {qn }   a  {qn }  a and {qn }  Q .
And finally, let a  0 , then a can be represented as a limit of the rational sequence {0}  0,0,0,0....,

1  1 1 1
or as a limit of the rational sequence    1, , , ...., or as a limit of
n  2 3 4
 1 1 1 1
   1, , , .... . Everything is proved.
 n 2 3 4

The Reader could notice that for any element a  F it’s representation as a limit of some rational
sequence is not unique (we have shown it only for 0  F ). Really, let’s fix any a  F .
We shown above that there exist at least one rational sequence {qn }  a . Let’s take the sequence

1   1  1
   0 , then qn    a  0  a . So, the rational sequence qn   also goes to a
n   n  n
 1
(and it differs from {qn } ) and similarly qn    a .
 n

96
 Real Numbers

Real numbers
Def. The complete, Archimedean ordered field R, which contains the field Q of rational numbers
as a subfield, is called a field of real numbers. Elements of R are called real numbers.
The addition and multiplication on R are extensions of the addition and multiplication on Q ,
i.e., for any a, b  Q  R :
a  b (addition by the rules of R)  a  b (addition by the rules of Q) ,
a  b (multiplication by the rules of R)  a  b (multiplication by the rules of Q) .
 of all fundamental
Let’s build R . We fix any set Q of rational numbers. Let’s consider the set
sequences {qn } of rational numbers. We define addition and multiplication on  .

Def. {qn },{ pn }    {qn }  { pn }  {qn  pn } and {qn }  { pn }  {qn  pn } . We showed earlier
that the set  of all fundamental sequences is closed under addition and multiplication,
so if {qn },{ pn } are fundamental, then {qn  pn } and {qn  pn } are also fundamental, therefore
these sequences belong to .
We have defined the addition "" and multiplication "" on  , these operations are associative and
commutative (because the addition and multiplication of rational numbers are associative and
commutative). There is also a distributive law: { pn }  {qn }  {hn }  { pn }  {hn }  {qn }  {hn } and
{hn }  { pn }  {qn }  {hn }  { pn }  {hn }  {qn }[D] which immediately follows from the distributive law
for rational numbers.

Def. Sequences {qn },{ pn }   are equivalent {qn }  { pn } if the sequence {qn  pn } is infinitely
small. Such relation  on  is obviously reflexive, symmetric, transitive. So,  is an equivalence
relation on  . Then  divides  into the classes , , C , D .... of equivalent (fundamental)

sequences. The set of all classes we denote like Raux .

Let’s define addition and multiplication on Raux in order to turn Raux into a field.

Def. For any ,   Raux : if {qn }  and { pn }  , then    is the class which contains the
sequence {qn  pn } . And    is the class which contains the sequence {qn  pn }. As sum and
product of any fundamental sequences {qn },{ pn } are again fundamental sequences, the classes
   and    do exist.
Let’s show that these classes are uniquely defined. Let {qn },{qn }  and { pn },{ pn }  we need
to show that {qn }  { pn }  {qn }  { pn } and {qn }  { pn }  {qn }  { pn } . So, as {qn },{qn }  , then
{qn  qn }  { n } (an infinitely small sequence), then qn  qn   n n  qn  qn   n and similarly
{ pn  pn }  {n } (an infinitely small sequence), then pn  pn   n n  pn  pn   n .
The sequences {qn  pn } and {qn  pn } are equivalent if the sequence {(qn  pn )  (qn  pn )}

97
 Real Numbers

is infinitely small. So, {(qn  pn )  (qn  pn )}  {(qn   n  pn   n )  (qn  pn )}  { n   n } ,

the last sequence is infinitely small, because { n } and { n } are infinitely small.

Next, the sequences {qn  pn } and {qn  pn } are equivalent if the sequence {qn  pn  qn  pn } is
infinitely small. So, {qn  pn  qn  pn }  {(qn   n )  ( pn   n )  qn  pn }  {qn   n   n  pn   n   n }
the last sequence is infinitely small. Really, {qn } is fundamental and therefore it is bounded, the
product of any bounded and infinitely small sequence is an infinitely small sequence, then {qn   n }
is infinitely small. Similarly { n  pn } is infinitely small, and also { n   n } is infinitely small.
Then {qn   n   n  pn   n   n } is infinitely small, everything is proved.

Now we have the set of classes Raux with two operations ,  on it.

Assertion1. ( Raux,, ) is a field.

Proof. The addition of sequences is associative and commutative, therefore, the addition of classes is
associative and commutative on Raux . The zero class  is the class which contains the sequence
{0}  0,0,0,0......, it consists of all infinitely small sequences of rational numbers.
Really, let { n }   , then { n }  {0} , then { n  0} is infinitely small, then { n } is infinitely small.
Conversely, let { n } is infinitely small, then { n  0} is infinitely small, then { n }  {0} , then
{ n }   . Next, for any class  which contains a sequence {qn } , an opposite class   is the class
which contains the sequence {qn } .

The multiplication of sequences is associative, then the multiplication of classes is also associative.
And finally, there is a distributive law [D], then there is the same distributive law for classes.
So Raux is a ring.

Let’s show that Raux \  is commutative multiplicative group. We need to show that Raux \  is
closed under multiplication (or it’s not a multiplicative group). In order to do that, we need to derive
several important auxiliary facts.

Auxiliary1. For any class    only one of the next cases is true: [A] or [B].
[A] There exist the positive rational number  and the sequence {qn }   such that   qn || n .
[B] There exist the negative rational number  and the sequence {qn }   such that qn   || n .

Let’s fix an arbitrary sequence {qn }     , we will get the needed sequence from it, by using only
two conditions: {qn } is not an infinitely small and {qn } is fundamental.
Let’s explain. The zero class consists of all infinitely small sequences of rational numbers, as
{qn }   , then {qn } is not an infinitely small. If some sequence {xn } is infinitely small, it means
that   0 k, n  k  xn   . As {qn } is not an infinitely small, then there exist some concrete

  0 , for which the requirement k, n  k  qn   is not true, so for any natural number k

98
 Real Numbers

there exist at least one natural number n  k such that qn   .

Let’s sum up:   0 : k  n  k  qn   . We fix now this concrete positive  .

Let’s remember now that {qn } (which is fixed above) is fundamental, so for the number  / 2  0 we
can fix the number k such that for any m, n  k we have qm  qn   / 2 [F].
As we noticed above, we can fix some concrete n  k such that qn   . Then for any natural
number m  k we have m, n  k and therefore: qm  qn   / 2, qn    qn  qm   / 2, qn   .
Let’s use the inequality qn  qm  qn  qm , then
qn  qm   / 2, qn    qn   / 2  qm , qn       / 2  qm   / 2  qm for any m  k .
So, the absolute value of every term qk 1 , qk  2 , qk  3 .... is greater than  / 2.

Let’s show that only one of the next cases is true:

[First] Every term qk 1 , qk  2 , qk  3 .... is greater than  / 2.
[Second] Every term qk 1 , qk  2 , qk  3 .... is less than   / 2 .

Really, the absolute value of every term is greater than  / 2 . So, for any concrete term qm , where
m  k , we have the case [First], or the case [Second], because from  / 2  qm follows that either
 / 2  qm , or qm   / 2 . Let’s assume that for some m, n  k we have the next situation:
 / 2  qm and qn   / 2 [D]. Then qm  qn    qm  qn   which contradicts to [F].

So, for any m, n  k there can’t be the situation [D]. Then there can be only the case [First], or the
case [Second], and no other variants.

Let we have the case [First]. Then all the “initial” terms q1 , q2 .... qk of our sequence we replace by

the positive number  . Then we have the sequence {qn }   ,  ....  , qk 1 , qk  2 , qk 3 .....
This sequence is obviously equivalent to {qn } (really, the sequence {qn  qn } is infinitely small,
because all it’s terms, starting from the number k , are zeroes), then {qn }   . And we have the
sequence {qn }   || qn   / 2 n (it is exactly the case [A] of auxiliary1).

Let we have the case [Second]. Then we replace all the initial terms q1 , q2 .... qk by the negative

number   , we will get the sequence {qn }   , ....  , qk 1, qk  2 , qk  3 .... which is equivalent
to {qn } (and therefore {qn }   ). So {qn }   || qn   / 2 n (it is exactly the case [B] of auxiliary1).

Consequence1. We can unite the cases [A] and [B] in one. For any class    there exist
the positive rational number ∆ and the sequence {qn }   such that   qn || n .

99
 Real Numbers

Auxiliary2. If a rational sequence {qn } || qn  0 n is fundamental and   qn || n , then {1 / qn }

is also fundamental.

  0 . For the positive number   2 (as for any other positive

Proof. Let’s fix an arbitrary positive
number) there exist k : m, n  k  qm  qn     . Then for any m, n  k we have
2

1 1 q  qm qm  qn   2   2
  n      . Then {1 / qn } is fundamental.
qm qn qm  qn qm  qn qm  qn   

Let’s show now that Raux \  is closed under multiplication. We take any classes ,   Raux \  .
According to the consequence1, we can take the sequence {qn }   such that   qn || n for some
  0 . And similarly, we can take some { pn }   such that   pn || n . Then for the sequence
{qn  pn }     we have     qn  pn || n . Then {qn  pn } is not an infinitely small.
Then           Raux \  .

As the multiplication of sequences is associative and commutative, the multiplication of classes

is associative and commutative on all Raux . The set Raux \  is closed under multiplication,
then the multiplication of classes is associative and commutative on Raux \  .

A class “one”   Raux \  is obviously the class which contains the sequence {1}  1,1,1 .. .

Let’s fix any class   Raux \  , and we take the sequence {qn }   such that   qn || n for
some   0 . The sequence {1 / qn } is also fundamental (auxiliary2) and therefore, it belongs to
some class  . The class    contains the sequence {qn }  {1/ qn }  {1}  1,1,1..., then      and
similarly      , then, by definition  1 . Notice that we haven’t shown that 1 Raux \  ,
which is a very important requirement. Let’s assume the contrary: 1 Raux \  1  .
Then the sequence {1 / qn }1 is infinitely small, then   0 k : n  k  1/ qn   , the last
inequality is equivalent to 1 /   qn  1 /   qn . So, for any positive rational  0 we have

1/   qn , starting from some moment. From here immediately follows that {qn } is not bounded.
Really, let’s fix any M  0 , then for the positive   1/ M  0 , the absolute values of terms
of {qn } are greater than 1/   M (starting from some moment), then {qn } is not bounded.
As {qn } is not bounded, it can’t be fundamental, and we have a contradiction.
1
Then   Raux \  and Raux \  is a commutative group under multiplication. And Raux is a field.

Def: a class  is called positive if there exist some positive rational   0 and the sequence
{qn }   such that   qn || n .

100
 Real Numbers

Assertion2. Raux is an ordered field.

Proof. We need to show that for any   Raux only one of the next cases is true:  is positive,
  is positive,    . Let’s take an arbitrary   Raux . If    , then any sequence {qn }   is
infinitely small and there can’t be any positive   0 such that   qn || n , so  is not a positive
class. The class   in this case is also a zero class  , so it can’t be positive for the same reason.
If    , then (according to the auxiliary1) there can be exactly one case: [A] or [B].

[A] There exist the positive rational number  and the sequence {qn }   such that   qn || n .
In this case  is positive. [B] There exist the negative rational number  and the sequence
{qn }   such that qn   || n . Then the class   contains the sequence {qn } such that
 qn   || n , the number   is positive, then   is a positive class.

We also need to show that the set of positive classes is closed under addition and multiplication.
Let ,  are any positive classes, then, as earlier, we take {qn }   || qn    0 n and
{ pn }   || pn    0 n . The class    contains the sequence {qn  pn } || qn  pn      0 ,
so this class is positive. The class    contains the sequence {qn  pn } || qn  pn      0 and this
class is also positive. Everything is proved.

Assertion3. The Archimedes axiom is true in Raux .

Proof. Let’s fix an arbitrary pair of positive classes ,  . We need to show that there exist some

natural n such that n     . Notice,  1 is positive, because   1   , where  is positive.

If we assume that 1 is negative, then   1 must be negative (it is a basic property of ordered
rings/fields, all these properties were listed earlier).
1 1
Let’s multiply the equality n     from the right side by  , we will get n       .
1
So, the equalities n     and n       are obviously equivalent (if we multiply both sides
of the second one by  , we will get the first one). The class   1 is positive. If we show that for
any positive class  there exist some natural n such that n     , then everything will be proved.
Let’s fix an arbitrary positive class  and a positive sequence { pm}   || pm    0 , this sequence
is fundamental, therefore it is bounded, then there exist some rational M  0 such that
pm  M m  pm  M m . According to the property5 of rational numbers, the rational number
M lies between two integers k  M  k  1. The integer number k  1 is positive, then k  1 is
a natural number, let’s take n  k  1, then we have n  pm    0 m . Then (n  1)     .
Really, (n  1)   is the class which contains the sequence (n  1), (n  1), (n  1)..... and  contains
the sequence { pm }  p1, p2 , p3 .... The class (n  1)     contains the sequence

(n  1)  p1 , (n  1)  p2 , (n  1)  p3  ....., all the terms of this sequence are greater than 1
(because pm  n m ). So, the class (n  1)     is positive and (n  1)     . Everything is proved.
101
 Real Numbers

For any rational number q  Q ,the stationary sequence {q}  q, q, q, q, q.... is fundamental, and
therefore there exist the class   Raux which contains this sequence. For different rational
numbers p  q the stationary sequences {q}  q, q, q, q, q... and { p}  p, p, p, p, p... are not
equivalent, therefore they belong to different classes ,   Raux . A class   Raux may contain
only one stationary sequence (because different stationary sequences are not equivalent).
If  contains some stationary sequence, then we say “  is a stationary class”.

Def. Every stationary class   Raux , which contains some sequence {q}  q, q, q, q, q...., we
replace by the rational number q . The set Raux will turn into the set R which consists of rational
numbers and classes of equivalent fundamental sequences: R  Q  (Classes of Sequences ) .
The set R is called a set of real numbers and elements of R are called real numbers.
Let’s define addition and multiplication onR. At first we define the mapping: f : Raux  R .
For any stationary class   Raux which contains some {q}  q, q, q, q, q.... we define f ( )  q ,
and for any other class   Raux we define f ( )   . Then f is one-to-one mapping Raux  R ,

( Stationary classes ) (Other classes )

f : Raux  R :   . Any element a  R has the unique
Q (Other classes )
representation as an image of some class   Raux : a  f ( ) .

a  b  f ( )  f ()  / by definition /  f (   )  R
Then we define a, b  R  .
a  b  f ( )  f ()  / by definition /  f (   )  R

From the theorem4 (page 40) immediately follows that ( R, ,  ) is a field.

From the theorem5 (Transfer of order) follows that R is an ordered field.
The field ( R, ,  ) is called a field of real numbers.

As the Archimedes axiom is true in Raux , this axiom is true in R. Really, let’s fix arbitrary positive
a, b  R . There exist the unique representations a  f (), b  f () . And the classes ,  are
positive in Raux (because f conserves the order):
a  0 (in R)  f ()  f () (in R)     (in Raux )   is positive (in Raux ). And  is also
positive (in Raux ). As the Archimedes axiom is true in Raux , there exist n such that n   , then
f (n)  f ()  f (      .... )  f () 
 f ()  f ()  f ()  .... f ()  f ()  n  f ()  f ()  n  a  b . So, the Archimedes
axiom is true in R.

102
 Real Numbers

Assertion4. The addition and multiplication on R are extensions of the addition and
multiplication on Q , i.e., for any a, b  Q  R we have:
a  b (addition by the rules of R)  a  b (addition by the rules of Q) ,
a  b (multiplication by the rules of R)  a  b (multiplication by the rules of Q) .
And similarly, the order relation "" on R is an extension of the order relation on Q :
a  b (by the rules of R)  a  b (by the rules of Q) .
~
Proof. Let’s denote for a while (,~ ) - the addition and multiplication on the field R.

Let’s fix arbitrary a, b  Q  R . In order to find their sum/product we need to find their
representations as images of some classes ,   Raux , so a  f () and b  f () . As a, b  Q ,
the classes ,  are stationary classes, the class  contains the sequence {a}  a, a, a .... and 
~ ~
contains the sequence {b}  b, b, b ..... By definition a  b  f ( )  f ()  / by def /  f (   ) .

The class    contains the sequence {a}  {b}  a  b, a  b, a  b ...., therefore    is

~
a stationary class, then f (   )  a  b , then a  b  a  b.

Next, a ~ b  f () ~ f ()  / by def /  f (  ) . The class    contains the sequence

{a} {b}  a  b, a  b, a  b ...., therefore    is a stationary class, then f (  )  a  b , then
a ~ b  a  b . Next, let a  b  f ()  f () (in R), as f conserves the order, then   
(in Raux ), then       () is a positive class in Raux , this class contains the sequence
{a}  {b}  a  b, a  b, a  b .... So, the sequence a  b, a  b, a  b ... belongs to the positive class
   . According to auxiliary1, the positive class    contains some sequence
{qn } || qn    0 n . If a  b  0 , then every term of the sequence {qn  (a  b)} is greater than
  0 , so this sequence can’t be infinitely small. But {qn  (a  b)} must be infinitely small,
because the sequences {qn } and {a  b} belong to the same class. Then a  b  0  a  b .
We have deduced that from a  b (in R) follows that a  b (in Q ).
Conversely: a  b (in Q ). Let’s consider the stationary class   Raux which contains the sequence
{a}  a, a, a ... and the class   Raux which contains the sequence {b}  b, b, b ... .
The class       () contains the sequence a  b, a  b, a  b...., every term of this sequence
is greater than the positive number (a  b) / 2 , then    is a positive class, then    (in Raux ),
then (conservation of order) f ( )  f () (in R)  a  b (in R) . Everything is proved
Consequence2. Now we can say: R is an ordered field which contains the field of rational numbers
Q as a subfield. From here follows that all the results that we got earlier in the chapter
“Sequences and limits” can be applied to R. As the Archimedes axiom is true in R (assertion3), then,
according to the theorem1:
[A] Any sequence {qn }  Q which converges/fundamental in Q also converges/fundamental in R
[B] Any element a  R can be represented as a limit of some rational sequence {qn }  Q .
103
 Real Numbers

Theorem2. An element a  R is an image of a class   Raux (so a  f () ).

Let {qn }  q1 , q2 , q3 , q4 .... is any fundamental sequence from  . Then the sequence of rational
numbers {qn }  q1 , q2 , q3 , q4 ....  Q  R converges to a in the field R.
Auxiliary3. Let   Raux and {qn }  q1, q2 , q3 , q4 ... is any fundamental sequence from A.

Let’s consider the sequence of stationary classes St (q1 ), St (q2 ).... St (qn ).... of Raux . Where St (q1 ) is

a stationary class with the sequence q1 , q1 , q1 ... and St (q2 ) is a stationary class with the sequence

q2 , q2 , q2 ... and etc. Let’s show that for any positive class   Raux there exist k , such that n  k

we have   St (qn )   .

(Notice, here we would like to say that  is a limit of stationary classes St (q1 ), St (q2 ).... St (qn )....,
but we have defined what is a limit of a sequence for sequences of any ordered field which contains
the field of rational numbers, and in our case elements of Raux are the classes of equivalent
sequences, so there are no any rational numbers in Raux . And we can’t use the limit-notion in Raux .

[Step1] Let’s show that for any positive stationary class P  Raux there exist the natural number k
such that n  k we have   St (qn )  P . Let’s fix any positive stationary class P  Raux , it
contains some sequence p, p, p....|| p  0 . The sequence {qn }   is fundamental, then for p / 2  0
there exist the natural number k , such that n, n  k we have qn  qn  p / 2 [L]. We fix now
the numbers k , n , where n is an arbitrary natural number n  k .

The class   St (qn ) contains the sequence {qn  qn } . The absolute value of every term of the

sequence {qn  qn } with number n  k is less than p / 2 [L]. It may not be true for the terms with

numbers m  k , let’s then replace every term with number m  k by a constant p / 4 .

Then we have the new sequence, and now the absolute value of every term is less than p / 2 , and

the new sequence belongs to the same class   St (qn ) (because it is equivalent to the initial one).

So we have {vn }   St (qn ) such that vn  p / 2 n .

Let’s consider now the class P with the sequence p, p, p..... Let’s show that   St (qn )  P ,
this inequality is equivalent to  P    St (qn )  P . Let’s show at first that   St (qn )  P .
Really, the class P  (   St (qn )) contains the sequence { p}  {vn }  { p  vn } and we know that
vn  p / 2 n   p / 2  vn  p / 2 , then p  vn  p  p / 2  p / 2 so, every term of the sequence
{ p  vn } is greater than the positive number p / 2 , then the class P  (  St (qn ))
(which contains this sequence) is positive (by definition). Then   St (qn )  P .
Let’s show now that  P    St (qn ) . The class (   St (qn ))  ( P) contains the sequence
104
 Real Numbers

{vn }  { p}  {vn  p} and again, from the condition vn  p / 2 n   p / 2  vn  p / 2 we have

vn  p   p / 2  p  p / 2 so, every term of the sequence {vn  p} is greater than p / 2  0 .
Then the class (   St (qn ))  ( P) is positive, then  P    St (qn ) .

So, for any positive stationary class P there exist k , such that for any n  k we have
  St (qn )  P .

[Step2] Let’s fix an arbitrary positive class   Raux there exist some sequence
{ n }   ||  n    0 n . Let’s consider the positive stationary class P which contains the sequence
 / 2,  / 2,  / 2..... The class   P is positive, really, it contains the sequence { n   / 2} , and all
the terms of this sequence are greater than the positive number  / 2 . Then   P . For the class P
there exist k , such that for any n  k we have   St (qn )  P , then   St (qn )  P   .
So, for any positive class   Raux there exist k , such that for any n  k we have   St (qn )   .

Let’s prove now the theorem3. We fix an arbitrary element a  R and the unique   Raux such

that a  f () . Let’s take an arbitrary {qn }   , we want to show that {qn }  a in the field R .
Let’s fix an arbitrary positive element   R . The mapping f : Raux  R conserves the order,
if   f () , then  is positive in Raux , really,   0 (in R)  f ()  f () (in R)     (in Raux ) .
For the positive class  there exist k such that for any n  k we have
  St(qn )        St(qn )    / f consevesthe order/  f ()  f   St(qn )  f ()
Remember the basic properties of any ring/field isomorphism: f (a)   f (a) and
f (a  b)  f (a)  f (b) . Then we can rewrite  f ()  f ()  f (St(qn ))  f () . The image of the
class  is the element a , the image of St (qn ) is the rational number qn , so we can rewrite
   a  qn    a  qn   . We have deduced that: for any positive   R there exist k ,
such that for any n  k we have a  qn   . It means exactly that {qn }  a .
Theorem3. R is a complete field.
Proof. We will use the consequence2 and the theorem2. Let’s fix an arbitrary fundamental sequence
{ak }  R . Our goal is to show that it converges to some a  R . According to the consequence2 [B],
every element ak can be represented as a limit of some rational sequence q1 , q2 , q3 ....  {qn }  ak .
k k k k

Let’s fix now an infinitely small rational sequence {1 / k}  0 . This sequence is infinitely small in Q ,
and, according to the consequence2 [A], it is infinitely small in R. For every ak we can find and fix
one element from {qnk }  a (let’s denote it qkk ) such that ak  qkk  1 / k .
So, we have the sequence q1 , q2 , q3 ...  {qk } such that ak  qk  1 / k for any k .
1 2 3 k k

105
 Real Numbers
k
The sequence {qk } is a rational sequence. From the conditions: {ak } is fundamental and

ak  qkk  1 / k k follows that {qkk } is also fundamental. Really, let’s take the condition

ak  qkk  1/ k  1 / k  ak  qkk  1/ k . We can denote  k  ak  qkk , then  1/ k   k  1/ k ,

according to the squeeze theorem for sequences, { k } is infinitely small in R, then { k } is

fundamental in R. According to the consequence9, the difference of fundamental sequences is a
fundamental sequence. And we have  k  ak  qkk  qkk  ak   k  {qkk }  {ak }  { k } , then {qkk }
is a fundamental sequence of rational numbers. Then there exist the class   Raux which contains
k
the sequence {qk } . According to the theorem2, the image a  f ()  R is a limit of the rational
sequence {qk }  R . Then the fundamental sequence {qk } actually goes to
k k
a  R . Then the initial
sequence {ak } goes to the same limit a . Really, we have  k  ak  qkk  { k }  {ak }  {qkk } ,
so the sequence {ak }  {qk } is infinitely small and {qk }  a . Then, according to the consequence6 ,
k k

we have {ak }  a . So, the fundamental sequence {ak }  R converges to a  R . Everything is

proved.

Assertion5. For any real number a  R there exist the unique pair of integer numbers k , k  1
such that k  a  k  1. The proof is exactly the same as the proof of the similar assertion
(property5) for rational numbers.

Uniqueness of real numbers.

Def. {xn } converges, then it’s limit can be denoted as lim xn or lim n  xn .

All fields of rational numbers are isomorphic. But formally, there exist different fields of rational
numbers, and based on these fields we can construct formally different fields of real numbers.
Let’s show that all these fields are also isomorphic.

Uniqueness theorem. Any fields R and R of real numbers are isomorphic. Moreover, the
isomorphism f : R  R is unique, this isomorphism is an extension of the unique isomorphism
f : Q  Q where Q  R , Q  R .
Proof. Existence. Let’s extend the unique isomorphism f : Q  Q up to f : R  R .
We define: for any a  R if {qn }  a (in R ) || {qn }  Q , then
f (a )  / by def /  lim f (qn ) (in R ) . As we know, every element a  R can be represented as a
limit of some rational sequence {qn }  Q . Let’s outline the main steps of our proof. In [A] we will
{qn }  Q converges in R , then { f (qn )} converges in R . From here follows that for
show that if
any a  R the element(s) f (a )  R is defined.

106
 Real Numbers

In [B] we will show that for any a  R the element f (a ) is uniquely defined, i.e., for any
sequence {qn }  a the limitlim f (qn ) is always the same. After [A] and [B] we are able to say
that: f is a mapping from R to R .

In [C] we will show that f coincides with the unique isomorphism f : Q  Q on the set Q .
Then the new f is really an extension of the old f .

In [D] we will show that f : R  R is one-to-one mapping. And finally in [E] we will show that
f has the basic properties of any isomorphism: a , b  R || f (a  b )  f (a )  f (b ) and
f (a  b )  f (a )  f (b ) .

Let’s start. [A] Let {qn }  a (in R ) || {qn }  Q , then {qn } is fundamental in R , then {qn }
is fundamental in Q  R . Let’s show that { f (qn )} is fundamental in Q . Let’s fix an arbitrary

positive   Q , there exist the unique positive   Q such that f (  )    .

As {qn } is fundamental in Q , there exist

k   : m, n  k  xm  xn        xm  xn     / f : Q  Q conserves "" / 

 f (  )  f ( xm  xn )  f (  )   f (  )  f ( xm )  f ( xn )  f (  )     f ( xm )  f ( xn )   
 f ( xm )  f ( xn )    . So, for any positive   Q there exist
k   : m, n  k  f ( xm )  f ( xn )    and therefore { f (qn )} is fundamental in Q .
As the Archimedes axiom is true in R (theorem1), the sequence { f (qn )} is fundamental in R .

As R is a complete field, the sequence { f (qn )} converges to some element of R .

So, the limit lim f (qn )  f (a ) exists.

[A1] (During the proof of [A] we have shown that f : Q  Q transfers any fundamental rational
sequence into a fundamental rational sequence)
[B] Let {qn }  a and { pn }  a , then {qn  pn } is infinitely small in R , then {qn  pn } is

infinitely small in Q . Let’s show that { f (qn  pn )}  { f (qn )  f ( pn )} is infinitely small in Q .

Let’s fix any positive   Q , there exist the unique   Q such that f (  )    .

As {qn  pn } is infinitely small in Q , for  there exist k   : n  k  qn  pn    

    qn  pn     f (  )  f (qn  pn )  f (  )   f (  )  f (qn )  f ( pn )  f (  ) 

    f (qn )  f ( pn )     f (qn )  f ( pn )    , then { f (qn )  f ( pn )} is infinitely small in Q .
Then (theorem1) { f (qn )  f ( pn )} is infinitely small in R . We already know that { f (qn )}

converges to some limit in R , then (consequence6) { f ( pn )} converges to the same limit in R .

[B1] (During the proof of [B] we have shown that f : Q  Q transfers any infinitely small
rational sequence into an infinitely small rational sequence).
107
 Real Numbers

[C] Let’s fix an arbitrary q Q then q is a limit of the stationary sequence q , q , q .....
According to the new definition, the image of q is defined as a limit of the sequence
f (q ), f (q ), f (q ) ..., it is a stationary sequence of Q , and it obviously converges to f (q ) Q .
So, the new mapping is really an extension of the old one.

[D] [Step1] f fully covers the set R . Let’s fix any element a  R , it can be represented as
a limit {qn }  a . Every element qn  Q has the unique representation qn  f (qn ) || qn  Q .
As {qn } converges in R , then {qn } is fundamental in R .
1
Remember that f : Q  Q is a field isomorphism, then f : Q  Q is also a field
isomorphism, we have shown above ([A1]) that such isomorphism transfers any fundamental
1
rational sequence into a fundamental rational sequence, so f transfers {qn } which is
fundamental in R , into {qn } which is fundamental in R . Then {qn } converges in R to

some element a  R . And we have: {qn }  a (in R ) and also { f (qn )}  {qn }  a  R ,

so f (a )  a and any element a  R has at least one preimage in R .

[Step2] f doesn’t glue together elements of R . Let’s assume that a  b and f (a )  f (b ) .
Let {qn }  a and { pn }  b . As f (a )  f (b ) , then lim f (qn )  lim f ( pn ) , then the sequence

{ f (qn )  f ( pn )} is infinitely small in R . According to [B1], f 1 : Q  Q transfers any

1
infinitely small rational sequence into an infinitely small rational sequence. So f : Q  Q
transfers { f (qn )  f ( pn )}  Q into f 1
 f (qn )  f ( pn )  { f 1( f (qn ))  f 1( f ( pn ))}  {qn  pn}
-this sequence is infinitely small in R . Then both sequences {qn } and { pn } converge to the same
limit in R and a  b , and we have a contradiction. So, if a  b there must be f (a )  f (b ) .
From the [Step1] and [Step2] follows that f : R  R is one-to-one.

[E] Let’s fix arbitrary elements a , b  R , and let’s fix any sequences {qn },{ pn }  Q
such that {qn }  a and { pn }  b in R . Let’s find f (a )  f (b ) . As f (a )  lim f (qn ) and
f (b )  lim f ( pn ) , then f (a )  f (b )  lim f (qn )  lim f ( pn )  lim( f (qn )  f ( pn )) 
 lim( f (qn  pn )) [Y]- we have here the limit of the sequence { f (qn  pn )}.
As {qn }  a and { pn }  b , then {qn  pn }  a  b . Then, by definition of f , we have
f (a  b )  lim( f (qn  pn )) - let’s compare this result with [Y], then f (a )  f (b )  f (a  b ) .
Next. Let’s find f (a )  f (b ) . As f (a )  lim f (qn ) and f (b )  lim f ( pn ) , then
f (a )  f (b )  lim f (qn )  lim f ( pn )  lim( f (qn )  f ( pn ))  lim( f (qn  pn )) [U].
As {qn }  a and { pn }  b , then {qn  pn }  a  b . Then, by definition of f , we have
f (a  b )  lim qn  pn - let’s compare this result with [U], then f (a )  f (b )  f (a  b ) .

108
 Real Numbers

Uniqueness. Let we have some isomorphism  : R  R , as  is a field isomorphism, it

transfers one into one and zero into zero:  (0 )  0 ,  (1 )  1 we can show (in exactly the same
way as we did in the uniqueness theorem for rational numbers) that  is an isomorphism    
(where    R ,   R ) (it follows from the basic properties of isomorphism and conditions
 (0 )  0 ,  (1 )  1 ). And then we can show that  is an isomorphism Q  Q
(the proof is exactly the same as a proof for rational numbers).

So  : Q  Q || Q  R , Q  R is an isomorphism. But there exist only one isomorphism

f : Q  Q , then   f on Q . Let’s fix any element a  R , there exist {qn }  Q || {qn }  a ,
then f (a )  lim f (qn ) . Every element q n belongs to Q , then f (qn )   (qn ) , then the sequences
{ f (qn )} and { (qn )} are completely the same, then they have the same limit:
f (a )  lim  (qn )  lim f (qn ) [J]. From the condition {qn }  a follows that { (qn )}   (a )
(show it, by using that  : R  R is an isomorphism, remember that  conserves the order "" ,
it follows from the theorem “Transfer of order”). So we have { (qn )}   (a ) and { (qn )}  f (a ) ,
then f (a )   (a ) , then   f on R .

Consequence. For any fields of real numbers R , R there exist the unique isomorphism
f : R  R [pict1] which is an extension of the unique isomorphism f : Q  Q
(where Q  R , Q  R ) which is an extension of the unique isomorphism f :    
(where    Q  R ,   Q  R ) which, in it’s turn, is an extension of the unique
isomorphism f :      (where      Q  R ,      Q  R ).

pict.1

109
 Real Numbers

Main properties of real numbers

Def. a  b - are any real numbers. The set {x} of real numbers such that a  x  b is called a

segment of real numbers, and we denote it [a, b] . The set {x} of real numbers such that a  x  b

is called “an interval of real numbers”, and we denote it (a, b) . Similarly:

{x | a  x  b}  / by definition/  [a, b) and {x | a  x  b}  / by definition/  (a, b] - half-intervals

of real numbers.
Note. At this moment segments and intervals are just some sets in the field R, there can’t be any
representation of these segments/intervals like some sets on the line, because we haven’t built
one-to-one correspondence between the field R and some line L.
In order to do it we need to define the “length”, and it will be done in the next chapter.
Let’s explore now the most important properties of real numbers.

Def. A sequence {an } of real numbers is called monotonically increasing (or just increasing)
if an  an 1 n and monotonically decreasing (or just decreasing) if an  an 1 n .
A sequence {an } is called strictly increasing if an  an 1 n and strictly decreasing if an  an 1 n
Auxiliary1. If {an }   and the sequence {an } is monotonically increasing (in particular, strictly
increasing), then an   n . If {bn }   and {bn } is monotonically decreasing (in particular
strictly decreasing), then   bn n .

Solution. {an } is monotonically increasing. Let’s assume that there exist some concrete number k
such that   ak . As {an } is monotonically increasing, then ak  ak 1  ak  2  ak  3  ......

Let’s consider the neighborhood O () with the radius     ak / 2 . This neighborhood O ()
does not contain the term ak and therefore it does not contain the next terms

ak 1  ak  2  ak 3  ... (really, n  k    an  an    ak    ak      ak    ak / 2 ,
then   an    an  O () ). Therefore,  is not a limit of the sequence {an } ,
(because any neighborhood of  must contain all the terms, starting from some number, which is
not true in our case), we have a contradiction. Then the number k such that   ak , doesn’t exist.
Then an   n . The proof is similar when {bn } is monotonically decreasing.

The Nested Segment Theorem. For any sequence of segments [a1, b1 ]  [a2 , b2 ]  [a3 , b3 ]  ....
of real numbers such that bn  an   0 , there exist exactly one real number  which belongs to
every segment [an , bn ] .

Proof. Let’s consider the sequence of right ends {bn } and the sequence of left ends {an } .
From bn  an   0 follows that both sequences {an } and {bn } are fundamental. Really, lets fix

110
 Real Numbers

an arbitrary   0 , and we can find the number k such that n  k : bn  an   .

Let’s fix any pair of numbers m, n  k and let’s show that am  an   and bm  bn   .

If m  n , then am  an  0 and bm  bn  0 .
If m  n , then am  an  am  an  [as am  bn ]  bn  an   .
And bm  bn  bn  bm  [as an  bm ]  bn  an   . There is no need to consider the case n  m ,

because am  an  an  am and bm  bn  bn  bm .

Then for any   0 there exist k , such that m, n  k we have am  an   and bm  bn   ,

then both sequences {an } and {bn } are fundamental. As R is a complete field, both these sequences
must converge: {an }   and {bn }   . Let’s denote  n  bn  an , from the condition
bn  an   0 follows that  n  is an infinitely small sequence. Then (consequence5 in sequences
and limits) we have    . So, both sequences converge to the same limit: {an }   and {bn }   .
The sequence {an } is monotonically increasing, then (auxiliary1 above) we have an   n ,

the sequence {bn } is monotonically decreasing, then (auxiliary1 above)   bn n .

So an    bn n it means that  belongs to every segment [an , bn ] .

Uniqueness. Let’s assume that there exist some other number  [an , bn ] n , where    .
Without loss of generality    . So, we have an    bn n and an    bn n and
an      bn n , then every difference bn  an is greater than the fixed positive number
(  )  0 , then there can’t be bn  an   0 and we have a contradiction. Everything is proved.
~
Def1. A set  of real numbers is called bounded above if there exist some number M such that
~ ~
x   : x  M . The number M in such case is called an upper bound of the set X.
~ ~ ~ ~
~
An upper bound M is not unique. If M is an upper bound of  , then any M  M is also
an upper bound of M .

Def2. The number M is called “ the least upper bound of  ”, or just a “supremum of  ”,
~ ~
if [1] M is an upper bound of  [2] for any other upper bound M  M of  we have M  M .

Example. The number 1 is the least upper bound for the next sets:
[0,1) , [2,1] ,[1,1] , 1  1/ n || n  .

Exercise. From the def2 immediately follows that: if X has the least upper bound M, then M is
unique.

111
 Real Numbers

Upper bound criterion. M is the least upper bound of   M is greater than any element
of  and for any positive number   0 the set ( M   , M ] contains at least one element of  .
Proof.  Let M is the least upper bound of X. The requirement “ M is greater than any element
of  ” is automatically true. Let there exist some   0 such that ( M   , M ] does not contain any
elements of  . Then any number from the interval ( M   , M ) is greater than any element of 
and less than M , so M is not the least upper bound. We got a contradiction. Then   0 the
interval ( M   , M ] contains at least one element of M .

Conversely.  From the requirement “ M is greater than any element of  ” follows that M is
~
an upper bound of  . Let’s assume that there exist some other upper bound M  M of 
~ ~
such that M  M , then the half-interval ( M , M ] (it is an interval ( M   , M ] in the case
~ ~
  M  M ) does not contain any elements of  . And we have a contradiction. So, if M  M ,
~
then there must be M  M and M is the least upper bound (def2).
The next theorem is one of the most important in math and there will be a very simple proof of it.

Least upper bound property. Any non-empty set X of real numbers, which is bounded above,
has the least upper bound M.
~
Proof. Let’s fix an arbitrary set  and any upper bound M of  . Let’s also fix some number T
~
which is less than some (any one) element of  . Then the segment [T , M ] .
[1] Contains at least one element of the set  .
~
[2] The “right end” of this segment (now it is M ) is greater than any element of  .
~ ~
~  T M T  M ~ 
Let’s divide [T , M ] into two segments T ,   [a1 , b1 ] and  2 , M   [b1 , c1 ] .
 2   
It’s easy to check that exactly one of these
segments ( [a1 , b1 ] or [b1, c1 ] ) also satisfies
to the conditions [1] and [2] [pic2].

Without loss of generality, let now [a1 , b1 ] pict.2

satisfies to [1] and [2]. Then we divide
 a b 
[a1, b1 ] into two segments a1 , 1 1  Comment: we haven’t built yet one-to-one
 2  correspondence between real numbers and
 a1  b1  points on the line. Anyway, this picture can
and
 2 , b1  . And again, exactly one be used for convenience.

of these two segments must satisfy to [1]

and [2], we denote this segments as [a2 , b2 ] , and we continue the process. As a result we have
the sequence {[an , bn ]} of nested segments, it’s easy to see that for any n  1 we have

bn  an  (bn 1  an 1 ) / 2 and therefore, for any n  2 we have bn  an  (b1  a1 ) / 2n 1 , then

obviously {bn  an }  0 and (according to the nested segment theorem, that is proved above)
112
 Real Numbers

there exist the unique number M which belongs to every segment [an , bn ] . Let’s show that M
satisfies to the upper bound criterion (and therefore, M is the least upper bound of  ).

[A] M is greater than any element of  . Really, let’s assume that there exist ~ x   such that
~x  M . Let’s consider the interval (M , ~x ) . Starting from some number k all the “right ends” bn
of the segments [an , bn ] belong the interval ( M , ~
x ) . Really, we have an  M  bn , if bn  (M , ~x ) ,
then b  ~
nx and therefore b  a  ~n x  a  [as a  M ]  ~
n n n x  M  const  0 , so b  a  const ,
n n

it can’t be true for every n , because {bn  an }  0 . Then there exist k such that bk  ( M , ~
x),
then all the next right ends bk 1, bk  2 , bk  3.... ( M , ~
x ) . Then the right end of every segment
[ak , bk ],[ak 1, bk 1 ],[ak  2 , bk  2 ] .... is less than ~
x , it obviously contradicts to [2].
We have a contradiction, so [A] is proved.

[B] We have to check that for any   0 the set (M   , M ] contains at least one element of  .
Let’s assume that there exist some   0 such that ( M   , M ] does not contain any elements of  .
Anyway, starting from some number k all the “left ends” a n of the segments [an , bn ] must belong
to (M   , M ] and therefore, starting from some number k , every a n is greater than any element
of the set  , then every segment [ak , bk ],[ak 1, bk 1 ],[ak  2 , bk  2 ] .... does not contain any elements
of X at all, it obviously contradicts to [1]. So [B] is proved.

From [A] and [B] follows that M is the least upper bound of  .
Similarly we can define:

Def3: a set  of real numbers is bounded below if there exist some number m ~ such that
~ is called “a lower bound of the set  ”.
~  x . In such case the number m
x   : m

The definition of the greatest lower bound (or infinum) is similar, there is a similar criterion of
the greatest lower bound and the similar theorem: any non-empty set X of real numbers, which is
bounded below, has the greatest lower bound m . This theorem can be immediately derived from the
least upper bound property (which we have proved). We just need to use the simple idea:
m is a supremum of X   m is an infinum of − X.

Separability theorem.  and  are set’s of real numbers and any number of  is less than any
number of  , so for x   and for y   we have x  y . Then there exist the number M
such that x  M  y x  , y   .

Proof. The set  is bounded above (by any element of the set  ), then  has a supremum M.
Let’s show that x  M  y || x  , y   . The inequality x  M is obvious.

Let’s assume that there exist ~ y  M . The half-interval ( ~

y   such that ~ y , M ] must contain at
x   , then ~
least one element ~ y~
x and we have a contradiction. Then y   we have M  y .

113
 Real Numbers

Def. If some sequence is monotonically increasing or decreasing, then we say that the sequence is
monotonic. When some set  is bounded above and below, we say that  is bounded.

Theorem [limit of a monotonic sequence]. {xn } is monotonic and bounded, then {xn } has
a limit (then {xn } converges to some real number ).

Proof. Let {xn } is monotonically increasing. As {xn } is bounded, the set of real numbers

  {x j || j  } is bounded, then  has the least upper bound M . Let’s show that lim xn  M .
Let’s fix an arbitrary positive   0 , the half interval (M   , M ] contains at least one term xk ,
but xk  xk 1  xk  2  ...  M , then the set ( M   , M ] contains all the terms of the sequence {xn } ,

starting from the number k . Then a fortiori (M   , M   )  O (M ) contains all the terms of the
sequence {xn } , starting from the number k . So, for any positive   0 , the neighborhood O (M )
contains all the terms of {xn } , starting from some number, it means that lim xn  M .
And there is a similar proof for any monotonically decreasing sequence {xn } , but we have to

consider the set   {x j || j  } and the greatest lower bound m of  . And similarly (as above),
we can show that lim xn  m .

Def. The least upper bound of  (or a supremum) can be denoted as sup  .
The greatest lower bound of  (or an infinum) can be denoted as inf  .

114
 Real Numbers

Decimal notation of real numbers

Assertion1. Let {xn }  a and { yn }  b .
[1-st part] If there exist some   0 such that starting from some number k   we have
yk 1  xk 1   , yk  2  xk  2   , yk  3  xk  3   ..., then b  a .

[2-nd part] If b  a , then there exist some   0 such that starting from some number k   we
have yk 1  xk 1   , yk  2  xk  2   , yk  3  xk  3   ....

Proof. [1-st part] we have here yk 1  xk 1   , yk  2  xk  2   , yk  3  xk  3   ......

As {xn }  a , then for  / 2  0 , starting from some number p : n  p we have   / 2  xn  a   / 2 .

As { yn }  b , then for  / 2  0 , starting from some number m : n  m we have   / 2  yn  b   / 2 .
Let’s take j  max( p, m, k ) then for any n  j we have:
  / 2  xn  a   / 2 and   / 2  yn  b   / 2 and yn  xn   . Let’s rewrite it like:
[1] a   / 2  xn  a   / 2 and [2] b   / 2  yn  b   / 2 and [3] yn  xn   .
From [3] and [2] follows that xn    b   / 2  xn  (b   / 2) from this inequality and [1] we have
a   / 2  b   / 2 , then a  b .
[2-nd part] b  a , let’s take   (b  a) . As {xn }  a , then there exist p : n  p we have
  / 3  xn  a   / 3 . As { yn }  b , then there exist m : n  m we have   / 3  yn  b   / 3 .
Let’s take k  max( p, m) . Then for any n  k we have:   / 3  xn  a   / 3 and
  / 3  yn  b   / 3 .
Let’s rewrite these inequalities: [1] a   / 3  xn  a   / 3 and [2] b   / 3  yn  b   / 3 .
Let’s notice that a   / 3  b   / 3 , then we have: a   / 3  xn  a   / 3  b   / 3  yn  b   / 3
from here: yn  xn  (b   / 3)  (a   / 3)  (b  a)  2 / 3   / 3 . So, for any n  k we have
yn  xn   / 3  const  0 .

Let’s sum up. Ifb  a , then yn  xn is greater than some positive  (starting from some number k ).
And conversely, if yn  xn is greater than some positive  (starting from some number k ), then
b  a.
Consequence1. If xn  yn n and both sequences {xn } and { yn } converge. Then lim xn  lim yn .

Proof. Let’s assume that lim xn  lim yn , then we have xn  yn    0 starting from some number k ,
then xn  yn (for n  k ), it contradicts to the initial requirement xn  yn n .
Therefore lim xn  lim yn .

Notice, if xn  yn n we can’t guarantee that lim xn  lim yn (when both sequences {xn } , { yn }
converge) we can only guarantee that lim xn  lim yn .

115
 Real Numbers

1 1
Consider for example xn  1 and yn  1  , we have here xn  yn n and lim xn  lim yn  1 .
n 2n
In order to claim that lim xn  lim yn we need to have a condition like yn  xn    0
(starting from some k ).

Def. Let {an } is any sequence of real numbers. The symbol  an is called a series.
n 1

For any n   the sum Sn  a1  a2  ...  an is called a partial sum of the series  an .
n 1
 
We say that  an converges to S (or goes to S ) and we write  an  S if the sequence of partial
n 1 n 1
sums {Sn } goes to S .

The number S in such case is called a sum of the series  an .
n 1

So, when we look at some series  an we should perceive it as a sequence of it’s partial sums
n 1

(a1 ), (a1  a2 ), (a1  a2  a3 ), (a1  a2  a3  a4 ) ..... And the writing  an  S means that this
n 1
sequence of partial sums goes to S .

If the sequence of partial sums (a1 ), (a1  a2 ), (a1  a2  a3 ), (a1  a2  a3  a4 ) .... does not have

any limit, then we say that  an diverges.
n 1
  
When  an  S we can also write it as a1  a2  a3  ....  S . If we write  an   bn , we mean
n 1 n 1 n 1
  
that both serieses  an and  bn converge, and the sum of the series  an is less than the sum
n 1 n 1 n 1
  
of the series  bn . And similarly for  an   bn .
n 1 n 1 n 1

From now on we consider only non-negative serieses  an || an  0 n .
n 1

m 
Lemma1. A non-negative series  an converges  The set of it’s partial sums  an  is bounded
n 1 n 1 
above.

Proof.  Let  an converges.
n 1
Let’s denote: S1  a1 , S2  a1  a2 , S3  a1  a2  a3 , S4  a1  a2  a3  a4 , the sequence {Sn }

116
 Real Numbers

is monotonically increasing, and it has some limit S (because  an converges).
n 1
According to auxiliary1 (page 110), we have Sn  S n . Then the set of all partial sums is bounded
above by S .
Conversely.  Let the set of partial sums is bounded above, it means that the sequence {Sn } is
bounded. The sequence {Sn } is monotonically increasing (because we have a non-negative series).
Then (according to the theorem about a limit of a monotonic sequence) {Sn } converges,

it means that  an converges.
n 1
   
Lemma2.  bn converges and an  bn n , then [A]  an also converges and [B]  an   bn .
n 1 n 1 n 1 n 1
 
[C] If there exist just one number k   such that ak  bk , then  an   bn .
n 1 n 1

Notice. In [C] we claim that: if there exist just one (any one) number k , for which ak is strictly less
 
than bk , then the whole sum of the series  an is strictly less than the sum of the series  bn .
n 1 n 1


~
Proof. Let Sn  b1  ...  bn and Sn  a1  ...  an . As  bn converges (lemma1) the set of partial
n 1

~ ~
sums {Sn || n  } is bounded above. As Sn  Sn || n , the set of partial sums of  an is also
n 1
 
bounded above. Then  an converges (lemma1) (and [A] is proved). The sum S of the series  an
n 1 n 1

~ ~
is a limit of the sequence {Sn } and the sum S of the series  bn is a limit of the sequence {S n } ,
n 1
 
~ ~
we have Sn  Sn || n , then (consequence1) S  S it means exactly that  an   bn
n 1 n 1
(and [B] is proved).

Let now there exist k   such that ak  bk , then

~ b  ak
Sk  Sk  (b1  b2  ...  bk )  (a1  a2  ...  ak )  bk  ak  k    const . For any n  k
2
 
~ ~
we obviously have Sn  Sn    0 . Then (assertion1) lim Sn  lim S n   bn   an .
n 1 n 1
Everything is proved.

117
 Real Numbers

Lemma3 [sum of a geometric progression]. For any q  R || q  1 the next formula is true:
q n 1  1
1  q  q 2  ...  q n  .
q 1
n 1
Proof: The given formula is equivalent to (1  q  q  ...  q )  (q  1)  q  1 and this one is very
2 n

easy to check, we just need to simplify the expression (“expand the brackets”) on the left side.
a(q n 1  1)
Consequence2. For any a, q  R || q  1 we have a  aq  aq  ...  aq  2 n
.
q 1
Lemma3. For any q  R || | q | 1 we have lim n  q n  0 .

Proof. In order to prove this one we have to prove a very important auxiliary inequality, which we
will also use in the future.

Auxiliary [Bernoulli inequality]. For any n  , x  1 the next inequality is true:

(1  x)n  1  nx .
Proof. By induction: n  1, then the inequality is obviously true. If the inequality is true for some
k   , then (1  x)k  1  kx . Let’s consider
(1  x) k 1  (1  x) k  (1  x)  (1  kx)  (1  x)  1  x  kx  kx2  1  (k  1) x  kx2  1  (k  1) x .
So, the inequality is true for (k  1)   . Everything is proved.

Notice: the requirement x  1 is important, it guarantees that 1  x  0 which allows us to write:

" (1  x)k  (1  x)  (1  kx)  (1  x) " .
n
Let’s prove the lemma3. If we can show that lim n  q  0 , then lim n  q  0 . Really, for any n
n

n n
we have 0  q n  q and if { q }  0 , then, according to the squeeze theorem for sequences,
{q n }  0 .

1 1
Let’s represent | q | 1 as a ratio | q | (Where h   1  0 , so h  const  0 ).
1 h q
1  Bernoulli  1 1 1
Then | q | 
n
   . Then we have 0 | q |n
 and
(1  h) n inequality  1  nh nh nh
1
(squeeze theorem for sequences), from 0  0 and    0 follows that { q }  0 .
n

 nh 

Lemma4 [sum of a geometric series]. For any q  R || | q | 1 the next formula is true:
1
1  q  q 2  ...  q n  .....  .
1 q

118
 Real Numbers

Proof. Here we speak about the sum of the series, the partial sum here (lemma3) is
n 1 qn  1
Sn  1  q  ...  q  . We want to find
q 1
qn  1 1 1
lim n  Sn  lim n   lim n   (q n  1)  lim n   lim n  (q n  1) 
q 1 q 1 q 1
1 1
  (1)  .
q 1 1 q

a
Consequence3. For any a, q  R || | q | 1 we have a  aq  aq 2  ...  aq n  .....  .
1 q
Notice. We have used above lim an  bn  lim an  lim bn . Remember that we can do so only when both

limits lim an and lim bn do exist. Unless we aren’t sure that both limits lim an and lim bn do exist,
we can’t perform this action. Similarly, when we carry out any of the next actions,
lim( an  bn )  lim an  lim bn or lim( an / bn )  lim an / lim bn , we need to make sure at first that both
limits lim an and lim bn do exist (in the last case we also need lim bn  0 and bn  0 n ).

Def. A sequence of integer numbers d1, d2 , d3....|| 0  dn  9 n is called “allowable” if it is not

a zero sequence 0,0,0,0,...0... and there is no any k   such that dn  9 n  k .
Once again, there is a sequence which consists of integer numbers from 0 to 9 . This sequence is
called allowable if it is not a zero sequence, and it does not have a “ 9 -tail” like 9,9,9...9....

From now on we consider only allowable sequences.


d1 d 2 dm dn
Assertion2. For any allowable sequence d1 , d 2 , d 3 .... the series 1
10 10
 2
 ... 
10 m
 ....   n
n 1 10
converges, and the sum of such series is a number S  (0,1) .

Proof. Let’s fix any allowable sequence d1 , d 2 , d 3 .... and let’s take the sequence 9,9,9,9.... .
9 9 9 9
Let’s consider the series   ...   ....  // consequenc e 3 //   1.
101 102 10m 1  1 / 10
As d1, d 2 , d3.... is allowable, then d n  9 n and there exist k such that d k  9 .
dn 9 d 9
So we have: n
 n n and there exist k such that kk  k .
10 10 10 10

d1 d2 dm dn
Then any term (with number k ) of the series 1
10 10
 2
 ... 
10 m
 ....   10 n
is not greater than
n 1
9 9 9
any term (with number k ) of the series 1  2  ...  m  ...., and the k  th term of the first
10 10 10

119
 Real Numbers

series is strictly less than the k  th term of the second series. Then (lemma2) the first series

d
converges and it’s sum is less than the sum of the second series, which is equal to 1 . So  nn  1 .
n 1 10
0 0 0
Next: the sum of the zero series 1
 2  ...  m  .... is equal to zero. As d1, d 2 , d3.... is
10 10 10
allowable, then there exist m   such that d m  0 . Let’s compare the zero series with the series
d1 d2 dm
  ...   ..... We have 0  d n n and m   : 0  dm .
101 102 10m
   
0 dn dn d
Then (lemma2)  n   n  0   n . We have deduced that 0   nn  1 .
n 1 10 n 1 10 n 1 10 n 1 10

The main theorem. Let  is a set of all allowable sequences d1, d 2 , d3.... There exist
one-to-one mapping f :   (0,1) : for every allowable sequence d1, d 2 , d3.... it compares the

d
unique real number 1  (0,1) which is a sum of the series  10nn .
n 1

d
Proof. The mapping f is defined like f (d1 , d 2 , d3 ....)   10nn .
n 1
From the assertion2 follows that f is really a mapping from  to (0,1) .

[PART 1] Let’s show that f fully covers the set (0,1) . Let’s fix an arbitrary 1  (0,1) .
Let’s build the sequence of pairs of natural numbers (m1, k1 ), (m2 , k2 ), (m3 , k3 ) ... where 1  m j  9 j .
1
[Step1] For 1  (0,1) there exist the minimal natural number k1 such that  1
10k1
(it’s easy to show, by using the Archimedes axiom and induction). Let’s fix this number k1 .
1 1
As k1 is a minimal number with such property, then k1
 1  k 1 . Next, for the positive
10 10 1
1 m
numbers k1
, 1 there exist the maximal natural number m1 such that k1  1 (it follows from
10 10 1
m1 1
the Archimedes axiom). Then:    . There must be 1  m1  9 (really, the assumption
10k1 1
1
10k1
m 1
m1  10 , contradicts the inequality k11  1  k1 1 ).
10 10

m1 m1  1 m1
Let’s also notice that  1  [J] (because m1 is a maximal number such that  1 ).
10k1 10k1 10k1

m1  m1  1
Let’s subtract from all sides of [J], we will get 0   1    .
10k1  10k1  10k1

120
 Real Numbers

m1
If 0  1  , then we fix the pair (m1 , k1 ) and end the process.
10k1
m m 1
If 0  1  k1 , then we designate 2  1  k1 and we have 0  2  k  2  (0,1) and we
10 1 10 1 10 1
repeat the [Step1] for the number 2 .
1
We will find the pair of natural numbers (m2 , k2 ) . At first we will find k 2 such that  2 ,
10k 2
1 m2 1
notice that k1  k2 (because we already have 2  ). Then    from here follows
10k 2 1
2
10k1 10k 2

 m2  1
that 1  m2  9 , and finally 0   2  k2 
 k .
 10  10 2
m2
If 0  2  , then we fix the pairs (m1, k1 ), (m2 , k2 ) and end the process.
10k 2
m m
If 0  2  k2 , then we designate 3  2  k2 . And we repeat the [Step1] for the number 3 .
10 2 10 2
We will get the pair (m3 , k3 ) such that k1  k2  k3  .... and 1  m3  9 .

 mp  1
And so on. For any pair (m p , k p ) we have 0   p   k .
kp 
 10  10 p
mp  
[1-st case] If the process has ended in p steps, then we get 0   p   and we have fixed some
kp 
 10 
pairs (m1, k1 ), (m2 , k2 ) ... (m p , k p ) . Here everywhere 1  m j  9 and k1  k2  ...  k p   .
m p 1 m p 2 m p 3
Let’s remember that: p  p 1  k p 1
|| p 1  p  2  k p2
|| p  2  p 3  k p 3
|| .....
10 10 10
m2 m1 m1 m2 m3 m p 1
... || 3  2  , || 2  1  || , then p  1     ....  [F]
10 k2 10 k1 10k1 10k 2 10k3 k
10 p1
 mp 
in addition to it, we also have: 0   p   , then we have
k 
 10 p 
m1 m m m mp
1  k1
 k2  k3  ....  kp 1  k .
10 10 2 10 3 10 p1 10 p
[2-nd case] The process hasn’t ended in p steps. Then we have some sequence

(m1, k1 ), (m2 , k2 ) ... (m p , k p ) .... here everywhere 1  m j  9 and k1  k2  ...  k p  ....  .

 m
m1 m2 m3
Let’s show that the series k1
 k  k  ....   kj converges to 1 . We got above the
10 10 2 10 3 j 1 10
j

121
 Real Numbers

m1 m2 m3 m p 1
simple formula [F] p  1     ....  (for any p   ), then
10k1 10k 2 10k3 k
10 p1
m1 m2 m3 m p 1
   ....   1  p [T].
10k1 10k 2 10k3 10
k p 1

mp  mp  1
Let’s add kp
to both sides of [T] and use the inequality 0   p    k [E]
kp 
10  10  10 p
(we have such inequality for every p ). So:

m1 m2 m3 m p 1 mp mp
   ....    1  p  
10k1 10k 2 10k3 k
10 p1 10 p
k k
10 p
m m m m m  m 
 k1  k2  k3  .... kp 1  kp  1   p  kp  . Let’s regroup the summands
10 1 10 2 10 3 10 p1 10 p  10 p 
 m   m m m m m 
 p  kp   1   k1  k2  k3  ....  kp 1  kp  .
 10 p   10 1 10 2 10 3 10 p1 10 p 

 m1 m2 m3 m p 1 mp  1
According to [E]: 0  1   k
 k
 k
 ....  k
 k
  k [E].
 10 1 10 2 10 3 10 p1 10 p  10 p
  mj 
We have the estimation for 1   the partial sum of  10k with p summands  .
 j 
 j 1

 1 
Both sequences {x p }  {0}  0,0,0... and {z p }   k 
are infinitely small, then, according to the
10 p 
  m1 m2 m3 m p 1 mp 
squeeze theorem for sequences, 1   k
 k
 k
 ....  k
 k
 is infinitely small.
  10 1 10 2 10 3 10 p1 10 p 
 m
It means exactly that our series  10kj j
goes to 1 .
n 1

Let’s sum up: from the 1-st case and 2-nd case follows that any number 1  (0,1) can be represented
m1 m2 m3 m p 1 mp
as a finite or an infinite sum 1     ....    ...., where 1  m j  9 and
10k1 10k 2 10k3 k
10 p1 10 p
k

k1  k2  ... k p  ...  . Notice that k1  k2  ... k p  ... are not necessary consecutive natural
numbers.

Let’s show that: if 1 is represented as a sum of a series, the next situation is impossible:

from some moment all the numbers k p 1  k p  2  k p  3  ..... are consecutive natural numbers and
all the numbers m p 1 , m p  2 , m p  3 ..... are equal to 9 . Let’s assume the contrary, then we have the

122
 Real Numbers

m1 mp 9 9 9
representation: 1  k1
 ....  kp
 k 1  k  2  k  3  .... [V]. Let’s denote
10 10 10 p 10 p 10 p
m1 mp 1 1
Sp   ....  . From [E] we have 0  1  S p  kp
, then 1  S p  kp
  ||   0 [J].
10k1 10
kp
10 10
From [V] we see that 1 is a limit of the sequence

 9  9 9  9 9 9 
 S p  k p 1 ,  S p  k p 1  k p  2 ,  S p  k p 1  k p  2  k p  3  .... [P]. This sequence is
 10  10 10  10 10 10 
a sum of the sequences {an }  S p , S p , S p .... (here lim n an  S p ) and

 9  9 9  9 9 9  1
{bn }   k 1 ,  k 1  k  2 ,  k 1  k  2  k  3  .... (and here lim n   bn  k ,
 10 p
  10 p
10 p
  10 p
10 p
10 p
 10 p
because lim n  bn is a sum of the geometric series). The limit of the sequence [P] is a sum of limits
1 1
lim n  an  lim n  bn  S p  kp
. Then 1  S p  kp
, but from [J] we have
10 10
1
1  S p  kp
  ||   0 and we have a contradiction.
10

d
And finally, let’s define the series  10nn that “coincides” with the finite or infinite sum
n 1
m1 m2 m3
   ....  1 . We have concrete natural numbers k1  k2  ...  k p  ....  .
10k1 10k 2 10k3
For every natural number n which differs from any of these numbers: n  k j j we define d n  0 ,

dn
and the summand  0 in this case. And for every natural number n which is some k j , i.e.,
10n
d m
n  k j (for some j ), we define d n  m j , so the summand nn  kj in this case.
10 10 j

d m1 m2 m3
Then the series  10nn “coincides” with the finite or infinite sum  
10k1 10k 2 10k3
 .... and
n 1
 
dn d
therefore  n goes to 1 . So  nn  1 . The sequence d1, d 2 , d3..... is allowable.
n 1 10 n 1 10
Really, j  0  d j  9 . Let’s notice that d1, d 2 , d3..... is not a zero sequence, because all
m1 , m2 , m3 ... are natural. And also d1, d 2 , d3..... does not have a tail 9,9,9,9...
(because m1 , m2 , m3 .... does not have such tail, as we showed above).

123
 Real Numbers

So, for any 1  (0,1) there exist the allowable sequence d1, d 2 , d3.....  such that

dn
f (d1 , d 2 , d3....)   n
 1 , we have proved that the mapping f fully covers the set (0,1) .
n 1 10

[PART 2] Let’s show that f doesn’t glue together elements of Ω. Let’s assume that for different
allowable sequences d1 , d 2 , d3 ....  a1 , a2 , a3 .... we have

d n  an
f (d1, d 2 , d3....)  f (a1, a2 , a3....)   n   n . Let’s compare d1 and a1 , if d1  a1 , then we go
n 1 10 n 1 10

to d 2 and a2 . If d 2  a2 , then we compare d 3 and a3 and etc. As the sequences d1 , d 2 , d3 .... and
a1 , a2 , a3 .... are different, then there exist m   such that
d1  a1, d 2  a2 , ...... d m 1  am 1, d m  am (maybe m  1 and we have d1  a1 from the start ).
Without loss of generality d m  am  am  d m . In several consecutive steps we will show that

an  d n
 10n   10n . Really, as the sequence a1, a2 , a3.... does not have a 9,9,9.... tail, then (lemma2)
n 1 n 1

a a1 a 9 9 9 a a
 10nn  10
 ...  mm  m 1  m  2  m  3  ..... Let’s denote  m  1  ...  mm , then the
10 10 10 10 10 10
n 1
sum of the last series is a limit of the sequence
 9  9 9  9 9 9 
 m  m 1 ,  m  m 1  m  2 ,  m  m 1  m  2  m  3  , this sequence is a sum of
 10   10 10   10 10 10 
 9   9 9   9 9 9 
sequences {an }  m , m , m .... and {bn }   m 1 ,  m 1  m  2 ,  m 1  m  2  m  3 ...
 10   10 10   10 10 10 
1
and lim n bn is a sum of the geometric series, so lim n  bn  m and lim n  an  m , then
10
a1 a 9 9 9 1 a a a  1
 ...  mm  m 1  m  2  m 3  ....   m  m   1  ...  mm11  mm   m 
10 10 10 10 10 10  10 10 10  10
d d a  1 d d a 1
// as a1  d1 ... a m1  d m1 //   1  ...  mm11  mm   m  1  ...  mm11  m m 
 10 10 10  10 10 10 10
 a  d m   d1 d m 1 d m d1 d m 1 dm 0 0
 m 
 10  ...  m 1
   ...  m 1
  m 1
 m2
 ...
 a m  1  d m 10 10 m
10 10 10 m
10 10
.

d1 d m 1 dm d m 1 dm 2 dn
 // Lemma 2 //   ...  m 1  m  m 1  m  2  ...   n
10 10 10 10 10 n 1 10

an  d n
So, in several steps we have deduced that  n   n - it contradicts to our initial assumption
n 1 10 n 1 10

d n  an
 10n   10n . Then f doesn’t glue together elements of  .
n 1 n 1
After the [PART 1] and [PART 2] we can say that f is one-to-one mapping   (0,1) .

124
 Real Numbers

For every allowable sequence d1 , d 2 , d3 .... the mapping f compares the unique real number

dn
  (0,1) such that    n
. From here immediately follows the converse assertion: for any
n 1 10

dn
real number   (0,1) there exist the unique allowable sequence d1 , d 2 , d3 .... such that    n
.
n 1 10

dn
Def. For any real number   (0,1) , where    n
, the symbol 0, d1d 2 d3.... is called a decimal
n 1 10
notation of  .

Next, for any other real number a  R (assertion5) there exist the unique pair of integer numbers
k , k  1 such that k  a  k  1. As we proved earlier, any integer number k has a unique decimal
notation k   1 2 ....  n .

If a  k , then the decimal notation  1 2 .... n of k is called a decimal notation of a .

If a  k, then k  a  k  1  a  k  (a  k ) || (a  k )  (0,1) .
As we showed above, the number (a  k )  (0,1) can be uniquely represented as a sum of
 
dn d
a special series (a  k )   n , then a  k  (a  k )  a  k   nn and the symbol
n 1 10 n 1 10
 1 2 ....  n , d1d 2 d 3 .... is called a decimal notation of a .

Exponent
n
 1
Assertion3. The sequence {xn }  1   converges.
 n
Proof. We will show that {xn } is monotonically increasing and bounded above, from here will follow
that {xn } converges.
xn 1
[A] {xn } monotonically increasing. Let’s show that  1  xn 1  xn (for any n   )
xn
n 1 n 1 n 1
 1   1   1 
1   1   1
 1    n 1
xn1  n  1   1   n  1   1   n  2 n  1
  1    n 1
 1    n 1   1     :  
 n  1  n   1  1   n   n 1 n 
n
xn  1
1   1    n 
 n  n
n 1 n 1 n 1
 1   n  2n   1   (n  1)  1   1  1 
2 2
Bernoulli
 1      1      1    1  
2
 // // 
 n   (n  1)   n   (n  1)   n   (n  1) 
2 2
inequality
 1  1   1  1   1  n  xn 1
 1    1  (n  1)   1    1    1       1 . So  1.
 n  (n  1) 2   n   (n  1)   n   (n  1)  xn

125
 Real Numbers
n
 1
[B] In the exactly similar way we can show that the sequence { yn }  1   || n  2 is
 n
n n n
 1  1  1
monotonically increasing. And for any n   we have xn  yn  1    1    1  2   1 ,
 n  n  n 
1
then xn  . As { yn } is monotonically increasing, then yn  y2 || n  2 , then
yn
1 1
xn    4 || n  2 . Then {xn } is bounded above.
yn y2

n
 1
So {xn }  1   converges, the limit of this sequence (the number) is denoted like e and called
 n
“an exponent”. Practical calculations show that e  2,71828 (the sign  means “approximately
equal”).

Def. For any natural number n   we define n! 1  2  3  ... n and also 0! / by def /  1.
n!
Def. For any pair of integers n  k || n  , k  0 the next number Сnk  is called
k!(n  k )!
a binomial coefficient.

n!
Exercise1. For any pair of integers n  k || n  , k  0 , the binomial coefficient Сnk 
k!(n  k )!
is a natural number. And Сn is a number of different k -element subsets of the set {1,2,3....n} .
k

n(n  1)
Show that Сn  1, Сn  n, Сn  , Сnn  1 .
0 1 2
2
Exercise2 (Binomial theorem). For any a, b  R any n   the next formula is true
n
(a  b) n   Cnk  a k  b n  k  Cn0 a 0b n  Cn1 a1b n 1  Cn2 a 2b n  2  .... Cnn 1a n 1b1  Cnn a n .
k 0

Notice, the Binomial theorem can be proved by induction, or by a simple reasoning:

(a  b)n  (a  b)  (a  b)  .... (a  b) , we need to understand which summands will appear on
the right side after we simplify the expression (after we expand the brackets on the right side).

1 1 1 1
Assertion4. The series 1     ...   .... [S] converges to e . And for any natural
1! 2! 3! n!
1 1 1 1 
number n   there exist the real number   (0,1) such that e  1     ...   .
1! 2! 3! n! n!n

126
 Real Numbers
n
 1
Proof. We have proved above that the sequence {xn }  1   is monotonically increasing and
 n
goes to e. Let’s expand the brackets:
n n
 1 1  n
1 1 1 1 1
xn  1      1   Cnk k  Cn0  Cn1  Cn2 2  ...  Cnn1 n1  Cnn n 
 n n  k 0 n n n n n
 n!   n!  1  n!  1  n!  1  n!  1  n!  1
          2     3  ......     k  ......    n 
 0!n!   1!(n  1)!  n  2!(n  2)!  n  3!(n  3)!  n  k!(n  k )!  n  n!0!  n

 n  1  n(n  1)  1  n(n  1)(n  2)  1  n(n  1)....(n  (k  1))  1  n(n  1)  ...  (n  (n  1))  1

 1       2    3  ...     k  ...    n 
 1!  n  2!  n  3!  n  k!  n  n!  n
1 1  n(n  1)  1  n(n  1)(n  2)  1  n(n  1)....(n  (k  1))  1  n(n  1)  ...  (n  (n  1)) 
1       ....     ...   
1! 2!  n 2
 3!  n 3
 k!  n k
 n!  nn 
1 1   1  1   1   2  1   1  k 1   1   1  n 1  
 1   1 1     1 1    1     ....  1 1    ...  1     ...  1 1    ...  1   
1! 2!   n   3!   n   n   k!   n   n  n!   n   n  
1 1 1 1
 1     ...  . We have shown that xn  Sn || n [1-st result] where S n is a partial sum
1! 2! 3! n!
1 1 1 1
of the series: 1     ...   .... . Let’s also show that for any n we have Sn  e :
1! 2! 3! n!
We got the representation [T]:
1 1   1  1   1   2  1   1  k 1 
xn  1   1 1     1 1    1     ....  1 1    ... 1     ...
1! 2!   n   3!   n   n   k!   n   n 
1   1  n 1 
...  1 1    ... 1    . Let’s fix any k   || k  2 . And let’s consider all the terms of
n!   n   n 
the sequence {xn } with numbers n  k . In any term xn || n  k we are interested in the first k
summands from the representation [T].
1 1   1  1   1   2  1   1  k  1
1   1  1     1  1    1     ....  1  1    ... 1    
1! 2!   n   3!   n   n   k !  n   n 
the partial sum  ~
   ( xn )
 of x n || n  k 
~
Notice there are exactly k summands in any partial sum ( xn ) for any xn || n  k .
~
For any concrete n  k the sum ( xn ) is a concrete number, and we have the sequence
~ ~ ~
( xk 1 ), ( xk  2 ), ( xk 3 ).... Every term in this sequence is represented as a sum of k summands,
~ ~ ~
and it’s easy to see that the sequence ( xk 1 ), ( xk  2 ), ( xk 3 ).... is a sum of several other
sequences:

~  1   1   1   1   1   2   1   1   k  1 
{( xn )}  1       1       1    1  ..     1    ... 1   .
 1!  2!  n   3!  n   n   k !  n   n 

127
 Real Numbers

Some of these sequences go to 1 . Really,

 1   2   k  1 
1   n1,
  1   n
 
1 .... 1   n1 .
 
 n    n    n 

1  1 1 1 1  1 1 1
And the other sequences are stationary sequences, like    , , .....,    , , ......
 2! 2! 2! 2!  3! 3! 3! 3!


the partial sum    ~ 1 1 1 1
Then the sequence     {( xn )} n 1     ...   Sk

 of x n || n  k  
1! 2! 3! k !
~
(where S k is a partial sum of the series [S]). So {( xn )} n  Sk . From here immediately
 
~
follows that if we fix any number v  S k , then all the terms of sequence {( xn )} are greater than v ,
1 1 1 1
starting from some moment. Let’s fix S k 1  1     ...  S .
1! 2! 3! (k  1)! k
~
Then there exist some number n such that Sk 1  ( xn ) || n  n . Then
~ ~
Sk 1  ( xn )  xn || n  n (because every x n contains it’s partial sum ( xn ) ).

Every element xn , in it’s turn, is less than e , then Sk 1  e . We had started above from an arbitrary
natural number k  2 and we showed that Sk 1  e , from here follows that:
Sn  e || n   [2-nd result] and it is exactly what we need. From the [1-st result] and [2-nd result]
follows that xn  Sn  e || n , then from the squeeze theorem for sequences,{Sn }  e , it means that
1 1 1 1
the series 1     ...   .... goes to e .
1! 2! 3! n!
Next. Let’s show that for any n   there exist   (0,1) such that
1 1 1 1 
e 1    ...   ||   (0,1) . The sequence of partial sums {Sn } is monotonically
1! 2! 3! n! n!n
increasing and goes to e , then Sn  e || n . So,
 1 1 1 1   1 1 1 1
0  e  S n  1     ...   ....   1     ...   
 1! 2! 3! n!   1! 2! 3! n! 
1 1 1 1  1 1 1 
    ...       .... 
(n  1)! (n  2)! (n  3)! n!  (n  1) (n  1)(n  2) (n  1)(n  2)(n  3) 
1  1 1 1  1 1  1 1  geometric 
      ....    1  
 n! (n  1)  (n  1) (n  1) 2  ...   
n!  (n  1) (n  1) 2 (n  1) 3     series 

128
 Real Numbers

 
 
1
 
1
  1   1  1  (n  1)  1 . We have shown that 0  e  S  1 ,
n! (n  1)  1  1  n! (n  1)
n
n n!n n!n
 (n  1) 
 

then there exist some   (0,1) such that e  S n  , everything is proved.
n!n
Def. Any number a  R , which does not belong to Q  R , (it means that a can’t be represented
as a ratio of some integer numbers) is called an irrational number.

Assertion5. e is an irrational number.

m
Proof. Let’s assume the contrary e  || m, n   , we can assume that n   , if not, then
n
m (1)  m  m m
e   , where  n is already a natural number. So e  || m  , n   .
n (1)  n  n n
Let’s take the representation of e for this number n   .
1 1 1 1  m 1 1 1 1 
e  1     ...   ||   (0,1)   1     ...   ||   (0,1) .
1! 2! 3! n! n!n n 1! 2! 3! n! n!n
Let’s multiply both sides by n!n , then
 1 1 1 1  1 1 1 1
m  n! 1     ...    n!n      m  n!  1     ...    n!n .
 1! 2! 3! n!   1! 2! 3! n! 
The expression on the right side is obviously an integer number, but  on the left side is not
an integer,   (0,1) . We have a contradiction, therefore e is an irrational number.

Assertion6. The field of rational numbers Q is not a complete field ( Q is incomplete).

 1  1 1  1 1 1
Proof. The sequence {Sn }  1, 1  , 1   , 1     ... is obviously a sequence
 1!   1! 2!   1! 2! 3! 
of rational numbers, and it has a limit e in the field R. Therefore {Sn } is fundamental in R,
then {Sn } is fundamental in Q  R . Let’s assume that Q is a complete field.
Then any fundamental sequence in Q has a limit in Q . In particular, {Sn } must converge in Q to

some limit q  Q . As R is an Archimedean ordered field, which contains Q (look at the theorem1,

page 94), the sequence {Sn }  Q must also converge in R, and it must converge to exactly the same
limit q  Q . But we showed above that {Sn } converges to e  Q in R, we have a contradiction.
Then Q is not a complete field.

129
 Construction of Length and Area

Length construction
We must start from the elementary objects: a point, a segment, a line. These are our basic objects.
We also make several simple assumptions that are listed below.
We assume that: for any different points A, B there exist only one segment AB .
We agree that we understand the phrase:
“The point С lies between A and B ”
[pict1].
pict.1
Also, any segment AB can be laid along
any other segment CD [pict2].
Note, when AB is laid along CD ,
points A and С must coincide.
Suppose we laid AB along CD and:

[A] Points B, D coincide. Then we say

“ AB is equal to
CD ”
and we write AB  CD .
[B] D is between A and B .
pict.2
Then we write: CD  AB and

we say “ AB is greater than CD ”, or “ СD is less than AB ”.

[C] B is between A and D . Then we write: CD  AB and we say “ CD is greater than AB ”,
or “AB is less than CD ”. The segment equality "" is reflexive, symmetric and transitive on  ,
where  is the collection of all segments in the space. The relation "" is transitive on  .

Mid-point assumption. For any segment AB there exist the point С such that AC  CB .
The sum of any segments AB and CD
is the segment AD [pict3].
And we write AD  AB  CD .

For any natural number n and any

pict.3

segment AB: the segment nAB is

defined like nAB  AB  AB  ...  AB
(there are exactly n “summands” on the
right side) [pict4].

pict.4
131
 Construction of Length and Area

We also assume that the Archimedes axiom is true in  : for any segments AB and CD there exist
the natural number n such that nAB  CD .

And finally, the axiom of nested segments is

true [pict4.1]: for any sequence of nested
segments A1B1  A2 B2  A3 B3  .... there
exist at least one point
X which
belongs to every segment:   Ak Bk k . pict.4.1

This is it, we have finished with our assumptions.

Def. If nAB  СD then we say: “ CD is n times greater than AB” and

СD
“AB is n times less than CD “ we can also rewrite: nAB  СD  AB  [by def ]  .
n
From here the next question appears: CD is any segment. For which natural numbers n does
СD
the segment exist? (based on our assumptions). According to the mid-point assumption,
n
СD СD
the segment exists for n  2 , by using the mid-point assumption for the segment we will
n 2
СD СD
and etc. Finally, for any n  2 the segment p does exist.
p
get the segment
4 2
СD
At this moment we can’t claim that exists for any natural n , only for n which is a power of 2 .
n

Auxiliary1. For any segments AB and CD there exist the minimal natural number p such that:
СD
 AB .
2p
СD СD
Comment: as p is the minimal natural number with such property, then p
 AB  p 1 .
2 2
Proof. Let’s fix any AB and СD , according to the Archimedes axiom, there exist n such that
nAB  CD . There exist 2 p  n , then 2 p AB  nAB and therefore 2 p AB  CD [T].
СD СD СD
Let’s consider the segment p
, there must be p  AB . Really, if p  AB , then
2 2 2
 СD  СD
2 p   p   2 p  AB  СD  2 p  AB - it contradicts to [T]. So p  AB . From here follows
2  2
 СD 
that the set of natural numbers т   || т  AB is not empty, then it contains the minimal
 2 
number p , and p is exactly the number we need.

132
 Construction of Length and Area

Length.  is a collection of all segments (in the space). The length L is the correspondence
L :   R  . For every concrete segment    a unique positive real number L( ) is defined,
L( ) is called a length of  . And L :   R  has the next properties:
[1] There exist the segment e such that L(e)  1 (e is called “a scale”).
[2] If    , then L( )  L(  ) .
[3] L(   )  L( )  L( ) for any segments  ,  .
We will show:
[A] [Existence]. If the scale (the segment e ) is chosen, the correspondence L :   R  ,
such that [1],[2],[3] exists.
[B] [Uniqueness]. If the scale e is fixed, then such correspondence L is unique.

[A] Let’s take and fix an arbitrary segment e which we take as a scale. We must define L on every
segment. At first we define L(e)  1. Let’s fix an arbitrary segment  , the number L( ) is defined
as a result of the next process.

The process. [step1] For the pair of segments  , e we take

e
the minimal natural p such that   (auxiliary1) the
2p
e
segment can be laid along  at least once [pict5].
2p pict.5

Then, according to the Archimedes axiom, we can find the natural number T0 such that:
 e   e 
T0  p     (T0  1) p  [Y1].
2  2 
e
So, T0 is such number that the segment can be
2p
placed in  maximum T0 times, but (T0  1)
e
segments already fully cover the segment 
2p
[pict6]. And (T0  1) is a minimal number of
e
segments that we need to cover the segment 
2p
completely. Let’s fix the pair of (rational) numbers
T0 T0  1
and . (we had started here from the pict.6
2p 2p
e p
segment p
, and we got two fractions with denominators 2 ).
2

133
 Construction of Length and Area

e e
[step2] We consider the segment (which is a half of the initial segment ) and we perform
2 p 1 2p
e
the [step1] for the segments p 1
and  . So, there exist the natural T1 such that
2
 e   e 
T1 p 1     (T1  1) p 1  [Y2].
2  2 

e
Let’s show that [A] 2T0  T1 and [B] T1  1  2(T0  1) . The segment can be placed in  exactly
2p
e e
T0 times, every segment p
contains exactly two segments p 1 , then  contains (for sure) 2T0
2 2
e e
segments p 1 . But T1 is a maximal number of segments p 1 which we can place in  ,
2 2
then 2T0  T1 ([A] is done).

e e
Next, T0  1 segments p
cover  completely, each segment p contains exactly two segments
2 2
e e
, then 2 (T  1) segments cover  completely. But T1  1 is a minimal number of
2 p 1 2 p 1
0

e
segments p 1 that we need to cover the segment  , then T1  1  2(T0  1) ([B] is done).
2

p 1 T0 T T 1 T 1
Let’s divide both inequalities [A] and [B] by 2 , then p
 p11 and 1 p 1  0 p [E].
2 2 2 2
e
(we had started here from the segment p 1
, and we got two new fractions with denominators 2 p 1 )
2
e e
Next, we take the segment (which is a half of the previous one ). And we perform
2 p2 2 p 1
e  e   e 
the [step1] for p  2 , . We will find the number T2 such that T2  p 1     (T2  1) p 1  .
2 2  2 
By the similar reasoning we deduce that 2T1  T2 and T2  1  2(T1  1) , we divide these inequalities

p2 T1 T2 T2  1 T1  1 e
by 2 , then we get  and  . (we had started here from the segment ,
2 p 1 2 p  2 2 p2 2 p 1 2 2
p

and we got two new fractions with denominators 2 p  2 ).

T0 T1 T2 T2  1 T1  1 T0  1
Let’s compare it with [E], so   and  p 1  p .
2 p 2 p 1 2 p  2 2 p2 2 2

134
 Construction of Length and Area

e e
Now we take the segment (which is a half of the previous one ). And we make
2 p 3 2 p2
e
the [step1] for p 3
,  and etc.
2
T0 T1 T2 T2  1 T1  1 T0  1
As a result we get the pair of sequences    .......   p 1  p .
2 p 2 p 1 2 p  2 2 p2 2 2
 T  T  1
Let’s show that both sequences  pk k ,  k p  k  k  0,1,2 ... converge to the same limit.
2   2 
 Tk  T0  1
It’s easy to see that p  k  is monotonically increasing and bounded above (by for example)
2  2p
 T 
then, according to the theorem about a limit of a monotonic sequence,  pk k  converges to some
2 
 T  1
limit  . Similarly, the second sequence  k p k  is monotonically decreasing and bounded below
2 
T  T  1
(by 0p for example), then it converges to some limit  . As the difference of the sequences  k p k 
2 2 
 T   1 
and  pk k  is an infinitely small sequence  p  k  k  0,1,2 ... , there must be    .
2  2 
Let’s sum up. We had started from an arbitrary segment  and we described the concrete process
which always gives only one concrete real number    , then we define: L( )     , i.e., the
length of the segment  is a common limit of our sequences.
Comment: during the construction process there was a phrase “we will find”, because people usually
understand the process better in such formulation. But normally, there must be the phrase
“there exist” instead of it, it is more rigorous. Of course, no one really assumes that some person
will repeat “by hand” the [step1] again and again for pairs of segments, it is virtually impossible.
Anyway, we need only the fact of existence of such sequences with a common limit, and we have it
now. Some people may be puzzled, because there is an “infinite process”, so there are infinitely many
steps, and if we make a computer program with such algorithm, it will never stop working. But it is
a great process, according to mathematics. The similar processes are embedded in the fundamental
theorems of analysis, differential equations, linear algebra, abstract algebra and etc.
We just need to remember that mathematics is not programming, the greatest theorems in
mathematics are the theorems that prove the facts of existence of some value/solution/figure and etc.
And if we want to make a computer program, we will write a program with a needed process and we
will repeat it many times, in order to get a very good approximation of the value we need.
Ok, now L is defined on every segment    . Let’s show that L satisfies the requirements
[1],[2],[3] (from the initial definition). [2] is obvious. Really, let    , then we lay  along 
and they coincide. Then we perform the process, which is described above, for the segment  , in the
same time it is the process for the segment  , so on the one hand we will get the length L( ) ,
and on the other hand we will get L(  ) . And L( )  L(  ) .
135
 Construction of Length and Area

Let’s show that L(e)  1 (it is [1]). We must perform the process, which is described above, for

e
the segment e . At first we need to find the minimal p such that  e . Obviously p  1 ,
2p
e
because  e , then we will get T0  2, T1  4, T2  8 ... Tk  21k . And we have two sequences
2
2 4 8 8 1 4 1 2 1
1
 11  1 2  ....... 1 2  11  1 both these sequences converge to 1 , then L(e)  1.
2 2 2 2 2 2
Let’s finally show that L(   )  L(  )  L(  ) (which is [3]). We fix arbitrary segments  ,  .

Let’s add these segments:    .

For each segment  ,  , we will build it’s own sequence (to determine it’s length).
T0 T1 T2 T2  1 T1  1 T0  1
For  we build   .......  p 1  p .
2 p 2 p 1 2 p  2 2 p2 2 2
M M M M 1 M 1 M 1
For  we build d0  d 11  d 22  ....... d2  2  1d 1  0 d .
2 2 2 2 2 2
S S S S 1 S 1 S 1
And for  we build m0  m11  m2 2  ....... 2m  2  1m 1  0 m .
2 2 2 2 2 2
Let’s take h  max( p, d , m) , we want to have the sequences which all start from the fractions with

the same denominators 2 h , so we can discard several initial terms in each sequence, it does not
affect convergence and the limit value. There is no need to complicate our designations, and we can
assume that all our sequences (from the very beginning) start from the fractions with the same
h
denominators 2 .

T0 T T T 1 T 1 T 1
For  :
h
 h11  h2 2  ....... 2h  2  1h 1  0 h .
2 2 2 2 2 2
M M M M 1 M 1 M 1
For  : h0  h 11  h 22  ....... 2h  2  1h 1  0 h [V].
2 2 2 2 2 2
S S S S 1 S 1 S 1
For  : h0  h11  h 2 2  ....... 2h  2  1h 1  0 h .
2 2 2 2 2 2
e
Let’s remember the meaning of these numbers. So, T0 segments can be placed in the segment
2h
e e
 , and M 0 segments can be placed in the segment  , so with guarantee T0  M 0 segments
2h 2h
can be placed in the segment    (and maybe more). And S 0 is a maximal number of segments
e
, which we can place in      , then T0  M 0  S0 . And in the exactly similar way we can
2h
deduce Ln  M n  S n (for every number n ).

136
 Construction of Length and Area

e e
Next, T0  1 segments cover the segment  completely , and M 0  1 segments cover
2h 2h
e
the segment  completely. Then (T0  1)  ( M 0  1) segments h (with guarantee) cover
2
e
the segment    . Let’s remember that S0  1 is a minimal number of segments h that we need
2
to cover the segment      , then S0  1  (T0  1)  ( M 0  1) . And similarly we get:
Sn  1  (Tn  1)  (M n  1) n . So we have:
Tn  M n  Sn T  M n  Sn
n ||  n  Ln  M n  Sn  Ln  M n  1 ,
Sn  1  (Tn  1)  ( M n  1) Sn  Tn  M n  1
Tn Mn Sn Tn Mn 1
let’s divide all the sides by 2
h n
, then:      .
2hn 2hn 2hn 2hn 2hn 2hn

Let’s use the squeeze theorem for sequences (look at [V]):

 Tn   M n   T  M   1 
 h  n    h  n  n L( )  L(  ) and  hn n    h nn    h  n  n

 L( )  L(  )  0 ,

2  2  2  2  2 
 S   S 
then there must be  h n n   L( )  L(  ) . And we already have [V]  h n n   L( ) .
2  2 
Then L( )  L(  )  L( )  L(   ) , everything is proved.

We have proved the existence of the length. Let’s prove the uniqueness.


Let L is any correspondence   R such that [1],[2],[3], so L is not necessary the same
correspondence as we built above.

[Property1]. From [3] follows that for any segments 1,...  n we have
L(1  ...   n )  L(1 )  ...  L( n ) , and in particular
L(n )  L(  ...   )  L( )  ...  L( )  n  L( ) .
pk  e  1
From here follows that for any number 2 we have L pk 
 p  k , really
2  2
e e
pk
 ...  pk
 [2 p  k summands ]  e , then
2 2
 e e   e   e   e   e  1
L p  k  ...  p  k   L(e)  L p  k   ...  L p  k   1  L p  k   2 p  k  1  L p  k   p  k .
2 2  2  2  2  2  2

[Property2]. For any segments  ,  we have     L( )  L( ) .

Really, let          where  is some segment, then L( )  L( )  L( )  L( )  L( ) .
Conversely, let L( )  L(  ) . There are exactly three variants:    ,   ,    .

137
 Construction of Length and Area

If    , then L( )  L( ) , which is not true. If    , then L( )  L( ) , which is not true.
So, the last possibility is    . Everything is proved.
~
Uniqueness. Let L is some other “length” (the correspondence   R  such that [1],[2],[3]).
~
From [1] we have L (e)  L(e) . Let’s fix an arbitrary segment  . And let’s build two sequences as
T0 T T T 1 T 1 T 1
above p
 p11  p2 2  ..... 2p  2  1 p 1  0 p . According to the process, which is described
2 2 2 2 2 2
 e   e  ~
above, for any k  0,1,2,3.... we have Tk  p  k     (Tk  1) p  k  . Then for L ([property2]) we
2  2 
~  e   ~ ~  e  ~ e  ~ ~ e 
have L  Tk  pk  
 L ( )  L  (Tk  1) p  k    Tk  L  p  k   L ( )  (Tk  1)  L  p  k  
  2    2  2  2 
1 ~ 1 T ~ T 1  T 
([property1]) Tk  p  k  L ( )  (Tk  1)  p  k  pk k  L ( )  kp  k . Both sequences  pk k 
2 2 2 2 2 
Tk  1
and  pk 
converge to L( ) (by definition, we have defined L( ) as a common limit of these
 2 
T ~ T 1 ~
sequences). Then from pk k  L ( )  kp  k (squeeze theorem for sequences) we get L ( )  L( ) .
2 2
~ ~
Then the values L( ) and L ( ) are equal on every segment  , then L  L .


Let’s resolve now the question about the existence of the segment (where  is some segment and
n

n is a natural number). Now we know only that exists when n is a power of two: n  2p .
n
We need to prove the next property.

[Property3]. For any positive real number a there exist some segment  such that L( )  a.

Proof. Let’s remember the chapter “Pre-real numbers” there was an [Example L] (page 94), in which
we had taken an arbitrary positive element a (which belonged to some Archimedean ordered field
1
which contained Q , and now a  R  Q ). We had fixed any n   such that  a , and then we
n
 Tk 
built the rational sequence  k 
|| k  0,1,2..... such that
2 n 
T0 T1 T T T  1 T  1 T  1 T0  1 T 
  22  33  ....  a  ....  3 3  2 2  1  . And  kk   a .
n 2n 2 n 2 n 2n 2 n 2n n 2 n 
 T  1  T 
Let’s notice that the difference of the sequences  k k  ,  kk  is an infinitely small sequence
 2 n  2 n 
 1   T p  1
 k  , and therefore we also have  p   a .
2 n  2 n
138
 Construction of Length and Area

Let’s fix now any positive real number a  R , we can find the minimal natural number p such that
1
p
 a . Let’s take then n  2 p and build two sequences (from the [Example L]).
2
T T T T T 1 T 1 T 1 T 1
So we have 0p  p11  p2 2  p3 3  ....  a  ....  3p  3  2p  2  1 p 1  0 p [S].
2 2 2 2 2 2 2 2
e
Let’s take now the scale segment e , we know that for any k   the segment exists, then
2 pk
e
for any natural number Tk the segment Tk  pk
also exists.
2
Let’s use the sequences [S] in order to build the sequence of nested segments. We fix any point O ,
e e
and build the segment OA0  T0  p
, then we lay the segment OB0  (T0  1)  p along OA1 .
2 2
e
Then we build OA1  T1  p 1
and
2
e
OB1  (T1  1)  p 1
[pict7], then
2
e e pict.7
OA2  T2  p2
and OB2  (T2  1)  p2
2 2
and etc. The segments A0 B0 , A1B1 , A2 B2 ... are obviously the nested segments, then (according to
the axiom of nested segments) there exist some point  which belongs to every segment.
1
This point is unique, because the length L( Ak Bk )  pk
, and if we assume that there also exists
2
some other point  which belongs to every segment, then every segment Ak Bk contains the
1
segment  , then L( Ak Bk )  L( )  const k which contradicts to L( Ak Bk )  pk
k 0 .

2
Then the segment O is exactly the segment, for which L(O)  a . Really, according to our
 e  T
construction process we have: L(OAk )  L(O)  L(OBk ) || k and L(OAk )  L Tk    pk k
pk
 2  2
 e  (Tk  1) Tk (Tk  1)
and L(OBk )  L (Tk  1)    , then  L (O )  || k [F] and both
 2 pk  2 pk 2 pk 2 pk
 Tk   T  1
sequences  pk 
and  k p  k  converge to the number a (these sequences are built in such way).
2  2 
Then from [F] and from the squeeze theorem for sequences follows that L(O)  a .

Let’s fix now an arbitrary segment  and any natural number n . Let’s take the positive real
L( )
number . According to the [property3], there exist the segment  which’s length is
n
139
 Construction of Length and Area

L( )
L(  )  . Let’s show that if we lay  along  exactly n times we will get the segment n
n

which coincides with  (then n   and then  by definition).
n
Let’s lay the segment  exactly n times along  , we will get segment n which is laid along  .
If n coincides with  , then everything is proved. Suppose that n does not coincide with  .
Then there can be exactly two variants: [First] n   , then [property2] we have L(n )  L( ) 
L( ) L( )
[property1]  L(  )  ..  L(  )  L( )   ..   L( )  L( )  L( ) and we have a
n n
contradiction.
[Second] n   , we will get here a similar contradiction (as we got in the first case).
Then n must coincide with  and everything is proved.
Note. In practice, for convenience, we do not write the letter L to denote a length of a segment.
Instead of L( )  5 we write just   5 . Instead of L( AB)  10 we write just AB  10 .

Line coordinates. [pict8] Let’s fix any line l , we mark any point O on this line, the point O
divides the line in two rays, we compare the real number zero 0 to the point O .
We need to choose any segment e as a
scale, so e  1. Then we choose one of two
rays (any one), we will call it “a positive
ray”. For any point A  O on the positive
ray we compare a positive real number:
a length of a segment OA .
The other ray is called “a negative ray”. pict.8
And for any point B  O on the negative
ray we compare a negative real number: a length of a segment OB with the minus sign  OB .

Def. For any point A on l , the real number a , which is compared to A , is called a coordinate
of A . And l is called a coordinate line.
We have defined the mapping f : l  R . Let’s show that this mapping is one-to-one.
At first, it covers R . Really, it’s very easy to show by using the [property3]: for any positive real
number a there exist some segment  such that L( )  a . Secondly, we need to show that
f :l  R doesn’t “glue together” points of l , i.e., if A  B are different points, then numbers
f ( A), f ( B) are also different. It’s very easy to show just by considering all the possible cases:
[A] some point A or B coincides with O , and the other lies on the positive/negative ray.
[B] none of A, B coincide with O , then there may be: both A, B are on the positive ray,
both on the negative ray, one point on the positive ray and the other one on the negative ray.
In each case it’s very easy to understand that f ( A), f ( B) are different real numbers.
140
 Construction of Length and Area

Then f :l  R is one-to-one. From here follows that the inverse mapping f 1 : R  l is also
one-to-one, describe this mapping.

Let’s sum up: for every line l there exist one-to-one correspondence (the mapping)  :R l .
Because of that we say that a coordinate line l is a “geometrical interpretation”, or a “visual
interpretation” of real numbers. A visual interpretation l of real numbers has some amazing
properties which can simplify our actions (with real numbers). For example, for any aR
the absolute value a  L(OA) (where A is the point on l with a coordinate a ). Also, for any real
numbers a, b , the number a  b is a length of the segment AB , where A and B are the points
on l with coordinates a and b.
Also, a  b  c if and only if B lies between A and C . When we add positive numbers we can
represent it as we take one segment and put it next to the other segment. Addition of negative
numbers has a similar representation.
Any segment [a, b] /interval (a, b) / half-interval (a, b] or [a, b) of real numbers has a visual
interpretation as a segment AB /interval (AB) /half-interval (AB] or [ A, B) on the line.
Any  -neighborhood  (a) of a real number a has a visual interpretation as an interval with
the center at A .
So, the use of a visual representation is a very handy tool, it may help us to simplify our reasonings.
But it’s important to remember that when we perform actions like addition/subtraction/
multiplication/division of real numbers we “work” inside the field R , and we do it by the rules of R ,
these rules are precisely defined on R , and they don’t have any connection with any line.
Many properties of real numbers do not have any visual interpretation, the field R is an
independent structure which consists of numbers (which are symbols), and R includes precise rules,
how to operate with all these symbols. In fact, every number is just an abstract symbol, the symbol
from R . The connection between real numbers and points is extremely important. Plane coordinates
and space coordinates allow us to take purely geometrical questions and narrow them down to the
questions, concerning real numbers. By using the connection between points and numbers we can
precisely describe shapes of objects and their motions. Moreover, there is no other way to describe
precisely most of shapes and motions without the use of real numbers.
But the set R of real numbers is a purely abstract (imaginary) set. And we use this set as an
auxiliary tool for a huge variety of different purposes. We use real numbers to specify time and
dates, to estimate the distance we need to walk and the duration of our journey, to determine our
income, our age, how many calories we eat a day, to measure our weight and height, to estimate our
lifespan and etc.
But once again, the set of real numbers R is a purely mathematical (imaginary) structure which we
can use as a tool for many different purposes.

141
 Construction of Length and Area

Plane coordinates. [pict9]. We fix two perpendicular

intersecting lines, the point O of their intersection we call
“the origin”. Each of these two lines must become a
coordinate line (in the same way as it was described
above). The scale e for each coordinate line is the same
(but sometimes it’s more convenient to use different scales,
we will not meet such examples in this book and in the
next books). One coordinate line is called an Ox axis, and
the other one is called an Oy axis (there is a freedom to
choose which one is Ox , and which one is Oy ). For each
axis, the zero number 0 must be compared to the point O .
pict.9
An important moment. We need to have a standard,
according to which, we choose positive rays on the axises Ox and Oy . And this standard is our right
hand’s palm. Look at the palm of Your right hand, when Your thumb is perpendicular to the
forefinger. The hand must be put on the lines Ox, Oy : the thumb on Ox and the forefinger on Oy .
The ray of Ox , which lies under the thumb, must be taken as a positive ray (of Ox ), the ray of Oy ,
which lies under the forefinger, must be taken as a positive ray (of Oy ). (notice, positive rays on Ox
and Oy can be chosen in any other way, but it is not typical and no one really does it, at least
intentionally).

Let’s mark any point A on the plane, we can draw the dashed lines, through this point which
are perpendicular to Ox and Oy . By definition, coordinates of intersection points ( x A , y A ) (a pair of
real numbers) are the coordinates of A . Conversely, any ordered pair ( x A , y A ) of real numbers
uniquely defines exactly one point A on the plane (with coordinates ( x A , y A ) )
When we write A(5,3) , we mean that A has coordinates (5,3) .

Space coordinates. [pict10]. We fix three perpendicular intersecting lines (in the space).
The point O of their intersection
we call “the origin”. Each line must
become a coordinate line, and for
each line the zero number 0 must be
compared to O . There is a freedom to
choose which line is Ox , which is Oy
and which is Oz . And again, we need
to use our standard: we look at our
right hand’s palm, the thumb is
perpendicular to the forefinger, and
the middle finger is perpendicular to
the palm. Then, as earlier, we put the
thumb on Ox , the forefinger on Oy
pict.10
and the middle finger on Oz .

142
 Construction of Length and Area

Each finger lies on the ray which must be chosen as a positive ray of the axis.

Now we have the axises Ox, Oy, Oz . Let’s mark any point A in the space. We draw the planes
 x ,  y ,  z through the point A which are perpendicular to the axises Ox, Oy, Oz .

Then  x ,  y ,  z intersect Ox, Oy, Oz at some points with coordinates x A , y A , z A , these coordinates
must be taken as coordinates of A.
Conversely, any ordered set of real numbers ( x A , y A , z A ) defines only one point A in the space

(with coordinates ( x A , y A , z A ) ). In order to find the point A we need to mark the points with

coordinates x A , y A , z A on the axises Ox, Oy, Oz and to draw the planes  x ,  y ,  z through the

marked points, which are perpendicular to Ox, Oy, Oz , these planes will intersect at the point A
with coordinates ( x A , y A , z A ) . When we write A(1,3,5) , we mean that A has coordinates (1,3,5) .

143
 Construction of Length and Area

Square root and types of sets

Def. X is a set of real numbers. Any rule, that for every number x   compares exactly one real
number f ( x)  R , is called a function on X. Any number x   can be called “a point x ” and
f (x) can be called “a value of f at a point x ”.
Def. a   is any point (a number) of X. And f is a function which is defined on X.
If there exist some sequence {xn }   || xn  a (n) , {xn }  a , and for any such sequence
the sequence { f ( xn )} goes to f (a) , then we say “ f is continuous at a”.
If there is no any sequence {xn }   || xn  a (n) , {xn }  a , then, by definition,
we also say ” f is continuous at a ”.

Def: a function f is called continuous on X if f is defined on X and f is continuous at every

point a.
Assertion1. The power function f ( x)  x k || k   is continuous on R.
Proof. Let’s fix any point aR and let’s consider any sequence of real numbers
{xn }  a || xn  a (n) . According to the basic properties of convergent sequences,
( {xn }  a, { yn }  b  {xn  yn }  a  b ), we have {xn }  a,{xn }  a  {xn2 }  a 2 , then
{xn2 }  a 2 ,{xn }  a  {xn3}  a3 and etc.
Finally,{xnk }  a k , which is equivalent to { f ( xn )} goes to{ f (a)} , because f ( x)  x k , so f is
continuous at every point a  R , then f is continuous on R.

Theorem1 [A zero point of a continuous function]. f is continuous on [a, b] and the values of f
at the end points a, b have different signs: f (a)  0, f (b)  0 or f (a)  0, f (b)  0 , then
there exist some point с  (a, b) such that f (с)  0 .

Proof. Without loss of generality, let f (a)  0, f (b)  0 . We divide the segment [a, b] into two
 a  b a  b  ab
equal segments: a , and , b . If f reaches zero at the middle point , i.e.,
 2   2  2
ab
f   0 , then the zero point с is found. If it’s not the case, then one of the segments:
 2 
 a  b a  b 
or
 a ,
2   2 , b  has the same property as [a, b] , at the left end of that segment the value
of f is negative, and at the right end of that segment the value of f is positive.

144
 Construction of Length and Area

a  b   a1  b1 
Let it be
 2 , b  [a1 , b1 ] , then we divide it into two equal segments a1 , 2  and

 a1  b1  a  b1  a1  b1 
 2 , b1  . If at the middle point 1 : f    0 , then c is found, and if not, then one
2  2 
 a1  b1   a1  b1 
of the segments
a1 , 2  or  2 , b1  has the same property as [a1 , b1 ] , and we choose it as
a new segment [a2 , b2 ] . And again, we divide it into two equal segments and etc. If the process

ended after several steps, then f reaches zero at the middle point of some segment [ak , bk ] , and
the point c is found. If the process didn’t end after several steps, then we have the sequence
of nested segments [a, b]  [a1 , b1 ]  [a2 , b2 ]  .... The length of each next segment is a half of
a length of a previous one, therefore there exist the unique point c which belongs to every segment.
Obviously c [a, b], let’s show that f (c)  0 . We assume that f (c) is positive, then we can
choose exactly one end point xk of every segment [ak , bk ] ( ak or bk ) at which f  0 . The sequence
{xn } of ends of segments goes to c, and f is continuous at c, then { f ( xn )}  f (c) . But the
negative sequence { f ( xn )} can’t go to the positive number f (c) , we have a contradiction.
And similarly, if we assume that f (c) is negative, then we can choose exactly one end point xk of
every segment [ak , bk ] at which f  0 , and we will get the similar contradiction.
So f (c)  0 and c  [a, b] . According to the initial conditions, c  a and c  b , then c  (a, b) .

Consequence1 [intermediate values of a continuous function]. f is continuous on [a, b] .

Then f reaches all it’s intermediate values on this segment: for any number  between f (a) and
f (b) there exist some point с  (a, b) such that f (с)   .
Comment1. When we say that  is between f (a) and f (b) , we mean that f (a)    f (b) or
f (b)    f (a) .
Comment2. f is continuous on [a, b] , then for any constant   R the function   f is
continuous on [a, b] (It immediately follows from the basic properties of sequences).

Proof. Without loss of generality, let f (a)  f (b) . Let’s fix an arbitrary number f (a)    f (b) .
Then we consider the auxiliary function g ( x)    f ( x) || x  [a, b] , this function (comment2) is
continuous on [a, b] and the values of g (x) at the end points a, b have different signs:
g (a)    f (a)  0 and g (b)    f (b)  0 . Then, according to the theorem1, there exist some
point с  (a, b) such that g (с)  0    f (с)  0  f (с)   . The consequence1 is proved.

Def. The real number c  R is called a k -th root ( k   ) of a if c k  a . In such case we write
c k a. 145
 Construction of Length and Area

An integer number k is called even if k is divisible by 2 , so k is even if k  2 p for some p   .

An integer number k is called odd if k is not divisible by 2 , so k is odd if k  2 p  1 for some
p.
Any integer number is even or odd and no other variants, it follows from the
[theorem1,[part2],page66,(division with a remainder)] about divisibility of integer numbers.

Let’s take the pair of integers k ,2 , there exist the unique representation k  2  k  r || 0  r  2 ,
so the remainder r equals 0 or 1 , if r is zero, then k is even, if r is one, then k is odd.

Theorem2.
[1] a  R is a positive real number. For any even k   there exist exactly two different opposite
values of
k
a . For any odd k   there exist only one positive value of
k
a.
[2] a  R is a negative real number. For any even k   the value k
a does not exist.
For any odd k   there exist only one negative value of
k
a.

Comment: it’s easy to notice that this theorem doesn’t have a “symmetrical form”, so everything
depends on the pair a  R, k   , which is not a convenient thing. For complex numbers (which we
will build in the next book) the situation is more symmetrical. For any non-zero complex number z
there is always k different values of k
z and for z0 only one value
k
z  0.
Proof. [1.1] Let’s fix any positive a  R and any even k   .
Existence. Let’s consider the function f ( x)  x k || x  R . If k is even, then f ( x)  x k  0 x  R .
We consider any segment [0, d ] || d  1 which contains a, so 0  a  d , then
0  a  d  d k  0  a  d k  f (0)  a  f (d ) . Then a is an intermediate value of f ( x)  x k
on [0, d ] , and f (x) is continuous on [0, d ] (because f ( x)  x k is continuous on R).
Then f ( x)  x k reaches the value a at some point c  (0, d ) [consequence1], so f (с)  a  c k  a ,
it means that c is a k-th root of a . And c is a positive number. For any even number k we have

(c) k  c k , then (c) k  a and  c is also a k-th root of a.

Uniqueness. Let’s assume that there exist some other number d which is also a k-th root of a .
So d c and d  c . From the condition d k  a follows that (d ) k  a . One of the numbers
 d, d is positive, without loss of generality, let it be d . Let’s compare d and c , if d  c , then
d k  ck  a and we have a contradiction, if d  c , then d k  c k  a and we have a contradiction
again. Then d  c , which contradicts to our initial assumption d  c . The uniqueness is proved.

146
 Construction of Length and Area

[1.2] Let’s fix any positive a  R and any odd k   .

Existence. We consider again f ( x)  x k || x  R and any segment [0, d ] || d  1 which

contains a . And again, (as we showed above), there exist c  (0, d ) such that c k  a , but now the
number c is not appropriate, because (с) k is negative ( k is odd) and a is positive.

Uniqueness. Let there exist some other d  c such that d k  a . The number d can’t be negative,

because in this case d  0 , whereas

k
a is positive, so d is positive.
And again, if d  c  d k  c k  a and we have a contradiction, if d  c  d k  c k  a and we
have a contradiction again. So, there must be d  c which contradicts to our initial assumption
d  с . The uniqueness is proved.

[2.1] Let a is a negative real number, and k   is even. Let’s assume that there exist с  R such
that c k  a . Obviously c  0 , then с is positive/negative. The number c k is positive (because k is
even) whereas a is negative, so there can’t be ck  a .
[2.2] Existence. Let a is a negative real number, and k   is odd. Let’s consider the positive
number | a | 0 . We have a   | a | . According to [1.1], for | a | 0 there exist only one positive

cR such that c k | a | . Then  c is a k -th root of a, really, (c) k  [k is odd]  c k 

  | a | a .
Uniqueness. Let’s assume that there exist some other k-th root of a. So d  a   | a | .
k

Then (d ) k | a | where | a | is positive, then  d  c (from [2.2]), then d  c (from [2.2]).
The uniqueness is proved.

And finally, for a0 and for any natural k   we obviously have k
a  0.
Let’s make a very important agreement. For any positive a  R and any even k   , we agree that
k
the symbol a denotes the positive value of k -th root. If we want to denote the negative value

of k -th root, we will write k a.

Let’s consider any points ( x , y ) and ( x , y ) on the plane. The segment  (in general) is

a diagonal of a rectangle, which sides have lengths x  x and y  y .

  x  x  y  y . Similarly, for any points

2 2
From the Pythagorean theorem follows that

( x , y , z ) and ( x , y , z ) in the space, the segment  (in general) is a diagonal of

147
 Construction of Length and Area

a rectangular parallelepiped, which sides have lengths x  x and y  y and z  z .

  x  x  y  y  z  z .
2 2 2
By using the Pythagorean theorem two times we get:

For convenience we do not say “the segment  has a length a ”, but only
“the segment  is a ”. We can say for example, “the rectangle  with sides a, b ”.

Def. Let  is any point on the plane and   0 is a fixed positive

number. The  -neighborhood  () , or just a neighborhood of  ,
is the set  ( )  { ||    } . [pict1].

And for any point  in the space, there is a similar definition.

And from now on, any set of points  on the plane/in the space we
will call “a figure  ”.
pict.1

Def: a point  is called an internal point of  if  belongs to  together with some

neighborhood  () . So  is an internal point of   There exist  () such that
 ( )   [pict2].  is called a boundary point of  if any neighborhood  () contains one
point from  and one point not from  [pict3]. And finally,  is called an external point of 
if  doesn’t belong to  together with some neighborhood  () [pict4].

pict.2 pict.3 pict.4

Exercise. Let  is any fixed figure on the plane/in the space. Then for any point  exactly one of
the next cases is true: [A]  is an internal point of  . [B]  is a boundary point of  .
[C]  is an external point of .
Notice: any internal point of  always belongs to  , any boundary point of  may belong and may
not belong to  , any external point of  doesn’t belong to  .
Def: a set of all internal points of  is called an interior of  , and we denote it int , a set of all
boundary points of  is called a boarder of  , and we denote it  .
148
 Construction of Length and Area

Area construction
Let’s fix some plane П and any coordinate system xy on this plane. From now on, we work only
on the plane П.

Def1. A figure  is called bounded if there exist some

rectangle    which sides are parallel to coordinate
axises x, y [pict1]. From now on we work only
with bounded figures on the plane.

Def2. Figures ,  are called equal if there exist

one-to-one mapping f :    which conserves
distances and parallelism: ,    the segment 
is equal to the segment f ( ) f () and the line  is
pict.1
parallel to the line f ( ) f () .

Def3. The area S is the correspondence S :   R   {0} that is defined on the collection 
of plane-figures, these plane figures are called measurable. For every concrete figure    a
unique non-negative number S () is defined, S () is called an area of  .
And S :   R   {0} has the next properties:
[1] The area of any unit square, which sides are parallel to coordinate axises, is 1 , so S (11)  1
[2] If measurable figures ,  do not have any common internal points, then
S (  )  S ()  S () .
[3] If measurable figures ,  are equal, then S ()  S () .
Comment. The question about the existence of such
correspondence is very important .
In most of books the area or (measure) is defined in the
next way: by definition the area of any rectangle
(or parallelepiped) is defined as a product of it’s sides,
and after that the next theory appears. Such approach
has a huge flaw. If we assume that the area of any
rectangle (which sides are parallel to coordinate axises)
is a product of it’s sides, we automatically define the
area on the huge class of figures, i.e., on the class of
rectangles. How can we check now that the property [2]
(additivity of area) is still in power?
Really, there are a lot of intricate ways to divide any
rectangle into several rectangles without common pict.2
internal points [pict2], and it’s not obvious at all why
the area of a big rectangle is a sum of areas of smaller rectangles. And such approach implies that

149
 Construction of Length and Area

we just have to believe that the additivity [2] is still true. There will be given a normal constriction
of area which doesn’t have this flaw.

Construction process.

Let’s fix any coordinate system Oxy .

For any a  R , the line x  a is the set of points
with coordinates {( x, y) || x  a, y  R} [pict3],
and the line y  a is the set of points with
coordinates {( x, y) || x  R, y  a} .

For any natural number n   we consider the

pict.3
“quadratic net”, i.e., the set of all lines on the plane
n n
with equations x  k  10 , y  k  10 || k   .
Such net divides the plane into equal squares,
and the greater the number n, the “thicker” the net.

Let  is any bounded figure. Let’s fix any number

pict.4
n n
n   and the net x  k  10 , y  k  10 || k   .

Let  n is the figure that consists of all squares

int

which completely belong to  , so  n consists of

int
n
squares ( n is some non-negative integer number).

Next,  n is the figure that consists of all squares which

ext

have at least one common point with  , so  n

ext
consists
of  n squares (  n is some non-negative integer number).

n    n [pict4] and n   n .
n we obviously have int ext
For any natural

We will say that  n is an internal quadratic figure (for  ) and  n

int ext
is an external quadratic
2 n
figure (for  ). For any n we define the internal sum int
n  n 10 and the external sum
2 n
ext
n   n  10 . So, for any n we have  n   n . It’s very important to notice here: as  is
int ext

bounded (def1), then for any n   the internal and external sums are concrete real (even rational)
numbers. And therefore, we have two sequences { n } and { n } .
int ext

Assertion1.  is any figure on the plane. Then the sequence of internal sums { int
n } is

monotonically increasing (  n   n 1 || n ) and the sequence of external sums { n } is

int int ext

monotonically decreasing (  n 1   n || n ).
ext ext

Any internal sum is not greater than any external sum: n  m || m, n   .
int ext

150
 Construction of Length and Area

Proof. Let’s fix any bounded figure  and let’s build a quadratic net for some n   . This net
2 n
defines the internal figure  n and the internal sum  n  n 10
int int
, where n is the number
of net-squares in  int
n . Let’s consider the net for the number n  1. Every square of  n is divided
int

into 100 smaller squares, these squares are still completely belong to  , therefore n  n 1 .
int int

Each square, which is counted in n , gives 100 squares for n 1 , then 100  n  n 1 .
Then n 1  n 1  10
int 2( n 1)
 100  n   102( n 1)  n  102n  int
n , so  n 1   n || n .
int int

And similarly we can show  n 1   n || n .

ext ext

So, the sequence { n } is monotonically increasing and { n } is monotonically decreasing.

int ext

Next, let m  n , then the inequality  n   m is obvious. Let m  n , then n  m  m .

int ext int int ext

Let finally n  m , then  n   n   m .

int ext ext

From int
n   m || m, n   immediately follows that the sequence { n } is bounded above
ext int

(by any concrete term of the sequence of external sums) and { n } is bounded below (by any
ext

concrete term of the sequence of internal sums), then (the theorem about a limit of a monotonic
sequence) both these sequences converge { n }  Slow and { n }  Sup .
int ext

Def4. If Slow  Sup , then  is called measurable and the number S ()  Slow  Sup is called
an area of  .

The main theorem.

The area of any rectangle, which sides a, b are parallel to coordinate axises, equals a b .
The proof of this theorem can’t be done quickly, the reason is that everything depends on the
“position” of such rectangle. If it’s vertexes have “good” coordinates, the proof is quite simple.
But if we want to prove this theorem for any rectangle, we need to derive at first some auxiliary
properties and assertions.

Assertion2 [additivity of area]. Let ,  are measurable figures without common internal points,
then S (  )  S ()  S () .
Proof. Let’s fix any n   and build a quadratic net for n   . This net defines the internal and
external quadratic figures n  n , n  n . Let’s also consider the internal and external
int ext int ext

quadratic figures for    , so (  )n  (  ) n . Any square from (   ) n is a square

int ext ext

that has a common point with  or with  (or with  and with  ). Such square for sure belongs
to the union  n  n , then we have (  ) n  n  n . Then the number of squares in
ext ext ext ext ext

(  )ext
n is not greater than the number of squares in n  n , and the number of squares in
ext ext

ext
n  n obviously is not greater than the number of squares in  n plus the number of squares
ext ext

151
 Construction of Length and Area

in n . Then: [the number of squares in (  ) n ]  [the number of squares in n ] 

ext ext ext

 [the number of squares in next ]   n ( for   )   n ( for )   n ( for ) [T1].

Let’s consider now  n and n . Any square from  n is a square that belongs to  and
int int int

therefore it belongs to    , then it belongs to (  ) n , then all  n belongs to (  ) n .

int int int

And similarly, n belongs to (  ) n . Then the union n  n belongs to (  ) n .

int int int int int

Then the number of squares in n  n is not greater than the number of squares in (  ) n .
int int int

Let’s notice that figures n , n

int int
do not have any common squares, because if they have a common
square, then it’s center is a common internal point of ,  , and these figures do not have any
common internal points. Then:
[the number of squares in int
n  n ]  [the number of squares in n ] 
int int

[the number of squares in nint ] and, as we noticed above, the number of squares in int
n  n is
int

not greater than the number of squares in (   ) n , then:

int

[the number of squares in int

n ]  [the number of squares in n ] 
int

[T2].
 [the number of squares in (   )int
n ] n ( for )  n ( for  )  n ( for    )

And finally, for    (as for any other figure), the number of squares that belong to    is not
greater than the number of squares that have a common point with    , i.e.,
n ( for   )   n ( for   ) [T3]. From [T1],[T2],[T3] we have:
n ( for )  n ( for )  n ( for   )   n ( for   )   n ( for )   n ( for ) let’s multiply
2 n
all the sides by 10 , we will get:
int
n ( for )   n ( for )   n ( for   )   n ( for   )   n ( for )   n ( for ) [T4].
int int ext ext ext

We know that figures ,  are measurable, by definition it means that

lim n {int
n ( for )}  lim n { n ( for )}  // by def //  S () and
ext

lim n {int
n ( for )}  lim n  { n ( for )}  // by def //  S ( ) .
ext

Then in [T4] the outer sequences have equal limits:

int
n ( for )   n ( for ) n
int
 S ()  S () and


ext
n ( for )   n ( for  ) n
ext
 S ()  S () . Let’s apply the squeeze theorem for sequences:


both sequences { n ( for   )} and { n ( for   )} converge to S ()  S () , then
int ext

(by definition)    is measurable and it’s area is S ()  S () .

152
 Construction of Length and Area

Assertion3. There is a rectangle  on the plane with sides a, b , and each vertex of  has
 m n  ~
coordinates  k~ , k  , where m, n  , k , k   , then the area of such rectangle equals a  b .
 10 10 
Proof. Each coordinate of each vertex has a denominator 10 || k   . Let’s choose the maximal
k

power kmax  k among all 8 denominators. Then each coordinate of each vertex can be written as
k
a fraction with the denominator 10 . Let’s build the quadratic net for that number k.
h
As each coordinate of each vertex of  is a fraction || h   , then each side of  must lie on
10k
some line of the quadratic net for k. Let’s notice that if we build the quadratic net for the next
number k  1 all the lines of the net for k are still present in the net for k  1, but in addition we
will have multiple new lines. Then the sides of  lie on the lines of any quadratic net
which is built for n  k .
n  k and build the quadratic net for n. We choose the vertex
Let’s fix an arbitrary natural number
m n 
of  which x and y coordinates are both minimal. Let’s denote it  min , minn  , then the other
 10 10 
n

vertexes must have coordinates

 mmin n  m n  m n 
 n  b, minn ,  min , minn  a ,  min  b, minn  a  .
 10 10   10   10 
n n
10 10
Let’s find internal and external quadratic figures and internal and external sums for П.

[Internal sum]
The internal quadratic figure  n [pict5]
int

comprises all the squares which belong to П.

Let’s find the number of squares n in the
internal figure.
The number of net-squares in any row, which is
b
parallel to Ox , is .
10 n
The number of net-squares in any row, which is
a
parallel to Oy , is .
10 n pict.5
Then the number of net-squares in П is 0
a b a b
  .
10n 10n 102 n
Therefore, the internal sum for П is
 2n a b
int
n  (the total number of squares )  10  2n
 10 2n  a  b [V]
10

153
 Construction of Length and Area

[External sum]
The external quadratic figure  n [pict6]
ext

comprises all the squares which belong to П plus

all the squares which “stay around” П. There are
two lateral strips of squares, which are “parallel ”
b
to Ox (blue colour), each strip consists of
10 n
b
squares, so these strips give 2  squares.
10n
There are two lateral strips of squares which are
“parallel” to Oy (green colour), each strip consists
a a pict.6
of squares , and these strips give 2 
10 n
10 n
0
squares. And we also have exactly 4 squares (gray colour), each of these square has a common
b a
vertex with П. So there are exactly 2  n  2  n  4 squares “around” П.
10 10
The external sum  n here differs from the internal  n sum only by summands which appear
ext int

because of the surrounding squares. All the summands from the internal sum present in
the external sum. Then, according to [V]:
 b a 
 ext
n  a b  2  2  n  4   10 2 n  a  b  (2b  10 n  2a  10 n  4  10 n ) [V1].
 10 
n
10
The formulas [V] and [V1] are true for any n  k , and we are interested in the next two limits:
lim n int
n  //[V] //  lim n a  b  a  b (because a  b  const) and
n
lim n ext
n  //[V1] //  lim n (a  b  (2b  10  2a  10  n  4  10  n ))  lim n a  b 
lim n (2b  10  n  2a  10  n  4  10  n )  lim n a  b  0  a  b .
Then the sequence of external and internal sums for П both go to the same limit a  b , therefore
(by definition) П is measurable and S ()  a  b .
Consequence1. From the assertions2,3 follows that the area of any internal/external quadratic
figure n , n is equal to the sum of areas of the net squares that form n , n .
int ext int ext

And therefore, any internal/external  n ,  n sum is exactly the area of an internal/external

int ext

quadratic figure: int

n  S (n ),  n  S (n ) .
int ext ext

Let’s explain. Let’s fix any quadratic net n   , any net-square has coordinates of vertexes like in
n n
the assertion3. The sides of such square are both equal to 10 , then it’s area is 10  10 n  102n .
Let’s fix any figure  , then the figures n , n are defined, both these figures consist of some
int ext

net-squares that do not have any common internal points, then, by additivity (assertion2),
the area of each figure n , n is a sum of areas of all squares which form this figure.
int ext

154
 Construction of Length and Area
2 n 2 n
There are n squares in int
n with area 10 each, then the area of  n is n 10 - but this
int

number is exactly the internal sum  n .

int

And similarly, the area of  n is exactly the external sum  n .

ext ext

Then the definition of a measurable figure can be formulated now in the next way:

Def5: a figure  is called measurable if the sequence of areas of internal quadratic figures
n )} and the sequence of areas of external quadratic figures {S (n )} both go to the same
{S (int ext

limit, that limit must be taken as S () .

There is even a stronger assertion (assertion5) which has a great practical value.

Assertion4 [1-st criterion of measurability].

 is measurable  For any positive   0 there exist some measurable figures  int and  ext
such that 
int
   ext and S ( ext )  S (int )   .
Proof.  Let  is measurable, then [Def5] both sequences {S (n )} and {S ( n
ext ext
)} go to the same
limit S () . Let’s fix any   0 , according to the simpliest properties of limits,  / 2 neighborhood
of S () contains all the terms of both sequences {S (n )} and
ext
{S (ext
n )} , starting from some

number k. Let’s fix any m  k  S (int

n )   / 2 ( S ()) and S (m )   / 2 ( S ()) .
ext

The difference of any two numbers, which are taken from any  / 2 neighborhood, is not
greater than  . Then S (ext
m )  S (m )   . Then  m ,  m are the figures we need.
int ext int

Conversely.  Let’s fix an arbitrary   0 .

For the positive number  / 4 there exist some
measurable figures  and 
int ext
such that
int    ext and S ( ext )  S (int )   / 4 [E0].
As  is measurable, there exist n   and the
ext

external quadratic figure K n  

ext ext
such that
S ( Knext )  S (ext )   / 4 [pict7].
As 
int ~   and the
is measurable, there exist n
internal quadratic figure Vn~  
int int
such that
S (int )  S (Vn~int )   / 4 . We got two important pict.7
inequalities for not necessary equal n , n ~ , we fix any
number m  n , m  n ~ . Let’s notice that when we take the greater number which defines the net,
the area of any internal quadratic figure may only become greater and the area of any external
quadratic figure may only become less.

155
 Construction of Length and Area

Then for the number m we have K m  

ext ext
and S ( K m )  S (
ext ext
)   / 4 [E1]
(because S ( K n )  S ( K m )  S (
ext ext ext
) ) and also S (int )  S (Vmint )   / 4 [E2]
(because S (Vn )  S (Vm )  S ( ) ).
int int int

Let’s sum the inequalities [E0],[E1],[E2], then we get S ( K m )  S (Vm )  3 / 4 [E3].

ext int

Let’s notice that the external quadratic figure  m (for  ) must belong to K m and therefore
ext ext

m )  S ( K m ) and the internal quadratic figure Vm must belong to  m (for  ), then

S (ext ext int int

S (Vmint )  S (int
m ) . Let’s sum up: S (Vm )  S (m )  S (m )  S ( K m ) .
int int ext ext

And we also have [E3], from [E3] immediately follows S (m )  S (m )  3 / 4   .
ext int

We showed that for any (small) positive   0 we can find some natural m , such that
m )  S (m )   [U]. Let’s notice now that the sequence of areas of internal figures {S (n )}
S (ext int int

is monotonically increasing (because this sequence is exactly the sequence of internal sums { n }
int

and this sequence is monotonically increasing (assertion1)). And the sequence of areas of external
figures {S (n )} is monotonically decreasing. Both these sequences converge in any case, just
ext

because  is a bounded figure: {S (n )}  Slow and {S (n )}  Sup . We obviously have:
int ext

n )  Slow || n and Sup  S (n ) || n and Slow  Sup (because every term of the sequence
S (int ext

n )} is not greater than every term of the sequence S ( n ) ). Then for any n we have
{S (int ext

n )  S (n )  Sup  Slow  0 , the difference Sup  Slow is a constant and from [U] we see that
S (ext int

this constant can be less than any positive number  , then Sup  Slow  0 and Sup  Slow .
Then, by definition, the figure  is measurable and the number S ()  Sup  Slow is defined.

Auxiliary1. Both figures ,  are measurable and    , then S ()  S () .

n n   then for any n   the number
Proof. From    immediately follows that nint  int

of net-squares in n is not greater than the number of net squares in  n .

int int

As areas of n and  n are sums of areas of net-squares which form these figures (consequence1),
int int

then S (n )  S (n ) [J]. As figures ,  are measurable, then {S (n )}  S () and
int int int

n )  S (() from [J] and from the basic properties of limits follows that S ()  S () .
S (int

156
 Construction of Length and Area
Asserrtion5 [2-nd criterion of measurability] .
If there exist some sequence of measurable internal figures
{int
n } ||  n   n and some sequence of measurable
int

external figures { ext

n } ||    n n such that
ext

lim n  S (int
n )  lim n  S ( n ) [pict8], then  is
ext

measurable and S ()  lim n  S ( n )  lim n  S ( n )

int ext

And conversely. If  is measurable, then there exist some

sequence of measurable internal figures
{int
n } ||  n   n and some sequence of measurable
int

external figures { ext

n } ||    n n such that
ext

pict.8
S ()  lim n S (int
n )  lim n  S ( n ) .
ext

Proof.  The converse assertion is obvious, it immediately follows from the (def5), the sequences
of internal and external quadratic figures are the sequences we need:
{int
n }  { n } and {n }  { n } . Let’s prove the other one  .
int ext ext

We denote   lim n  S ( n )  lim n  S ( n ) . Let’s fix an arbitrary

int ext
  0 and consider
 / 2 -neighborhood  () of  . According to the basic properties of convergent sequences, ()
contains all the terms of the sequences {S ( n )} and {S ( n )} starting from some number k .
ext int

The difference of any two numbers which are taken from  / 2 () is less than  , then
n )  S ( n )   . For any concrete n  k we have two measurable figures  n ,  n such that
S ( ext int int ext

n     n and S ( n )  S ( n )   . And  can be fixed from the very beginning as an

int ext ext int

arbitrary small positive number, then (assertion4)  is measurable and the number S () is
defined. Next, from  int
n     n (n) (auxiliary1) follows that
ext

S (int
n )  S ()  S ( n ) (n) [T], both sequences {S ( n )} and {S ( n )} go to the same limit,
ext int int

we can consider the stationary sequence {xn }  S (), S (), S ().... and apply the squeeze theorem
for sequences in [T], then S () is equal to lim n  S (int
n )  lim n  S ( n ) .
ext

Assertion6. For any real number a there exist a monotonically increasing sequence of rational
 mn 
numbers  na (here n  1,2,3,4,... and m1 , m2 , m3 , m4 ... are integer numbers)
10 
m 
and a monotonically decreasing sequence of rational numbers a   nn  .
10 
Proof. Let’s take the decimal notation of a  1 2 ....  n , d1d 2 d3 .... Let a is not negative ( a  0 ):
a  1 2 ....  n , d1d 2 d3 .... We can take the sequence:
157
 Construction of Length and Area

q1  1 2 ....  n , d1 00000 ... || q2  1 2 ....  n , d1d 2 0000 ...

|| q3  1 2 ....  n , d1d 2 d3 000 ... || q4  1 2 ....  n , d1d 2 d3d 4 00 ... and etc.
This sequence is monotonically increasing and it goes to a. The term q1 can be represented as

 m1   m2 
 1  || m1   , the term q2 can be represented as  2  || m2   , the term q3 can be represented
10  10 
m  m 
as  33  || m3   and etc. So we have a monotonically increasing sequence  nn   a .
10  10 

Let’s build the second sequence. Again a  1 2 ....  n , d1d 2 d3 ... and we take
1 1
q1  1 2 ....  n , d1 00000...  || q2  1 2 ....  n , d1d 2 0000...  2 ||
10 10
1  1
| q3  1 2 ....  n , d1d 2 d 3 000...  3 .. || q4  1 2 ....  n , d1d 2 d 3 d 4 000...  4 and etc.
10 10
This sequence is monotonically decreasing and it goes to a. And here again, the term q1 can be

 m1   m2 
represented as  1  || m1   , the term q2 can be represented as  2  || m2   , the term q3 can
10  10 
m 
be represented as  33  || m3   and etc. So we have a monotonically decreasing sequence
10 
 mn 
 n   a . Let now  a is a negative real number (a  0) , so  a  1 2 ....  n , d1d 2 d3 ....
10 
Let’s take the positive number a  1 2 ....  n , d1d 2 d 3 ...(we reflect a over the zero) and we build
two sequences that we need, for the positive number a. Then we reflect everything back over the
zero (instead of each sequence we will have an opposite to it). A decreasing sequence (after reflection)
will become an increasing sequence which goes to  a , and an increasing sequence (after reflection)
will become a decreasing sequence which goes to  a .

 1   1 
And finally, when a  0 we can take the sequences  n ,  n  .
10  10 
Let’s prove now the main theorem.
The area of any rectangle, which sides a, b are parallel to coordinate axises, equals a b .

We fix any П on the plane with sides a, b . We can chose it’s vertex which x -coordinate and
y -coordinate are both minimal, let it be ( x0 , y0 ) .
Then ( x0  a, y0 ), ( x0 , y0  b), ( x0  a, y0  b) - coordinates of other vertexes.
For each coordinate of each vertex (assertion6) we build two monotonic sequences which converge
to that coordinate. For any concrete number n   we can mark the terms of these sequences which
belong to the segments [ x0 , x0  a], [ y0 , y0  b] , red points [pict9] (for any n   there are 4
158
 Construction of Length and Area

terms) these terms define an internal rectangle  n

int

(red) which coordinates are like in the assertion3, pict.9

then  n is measurable and we know it’s area.

int

Next, for any n   we can mark the terms of the

sequences which do not belong to the intervals
( x0 , x0  a), ( y0 , y0  b) , orange points [pict9]
(for any n   there are 4 terms). These terms define
an external rectangle  n
ext
(orange) which coordinates

are like in the assertion3, then  n

ext
is measurable and

we know it’s area. The sides of { n } , which are parallel to Ox , form a sequence that goes to
int
a.
Really, the sequence of the “right ends” goes to x0  a , the sequence of the “left ends” goes to x0 ,
then their difference (which is always a length of a side) goes to a. The sides of  nint , which are
parallel to Oy , form a sequence that goes to b . Then the sequence of areas {S ( nint )} goes to a  b .
And similarly, the sequence of external areas {S ( n )} goes to
ext
a  b . Then, according to the
assertion5,  is measurable and S ()  a  b . Everything is proved.

From the main theorem immediately follows the property [1] of area:
the area of any unit square 1 1, which sides are parallel to coordinate axises, is equal to 1 .

Let’s finally prove the [3]-rd property of area (equal measurable figures have equal areas).
Let  and  are equal figures, according to the def2, there exist one-to-one correspondence
f :    which conserves distances and parallelism. As  is measurable, we have
n )  lim n  S (n )  S () [B1]. And for any concrete n   we have
lim n  S (ext int

int
n    n . Then for the images of these figures we have
ext

f (int
n )  f ()  f (n )  // as f ()   //  f (n )    f (n ) [B2].
ext int ext

The figure  n consists of

int
n net squares, the sides of each square are parallel to coordinate axises
n
and both sides are equal to 10 . We know that f conserves distances and parallelism, so f
translates every net-square into the equal square, which sides are still parallel to coordinate axises,
n 2 n
and are still 10 each, then, by the theorem4, the area of such square is 10 (the same as it was).
The figure  n consists of
int
n squares without common internal points, then the image f (int
n )

also consists of n squares without common internal points. Then, by additivity [2],

S f (int 
n )  n  10
2 n 2 n
and S (n )  n  10 . So, for any n we have S (n )  S f (n ) and
int int int
 
ext
 ext

similarly S (n )  S f (n ) . Then, the sequences {S ( f (n ))} and {S ( f (n ))} coincide with
int ext

the sequences {S (n )} and {S (n )}, therefore they both go to the same limit
int ext
{S ()}.

159
 Construction of Length and Area

As we noticed above, for any n   we have f (int

n )  f ()  f ( n ) . We know that
ext

lim S ( f (int
n ))  lim S ( f ( n ))  S () , then from the assertion5 follows that f () is measurable
ext

and S ( f ())  S () . We have proved the properties [1],[3] of area. The additivity [2] was proved in
the assertion2. The existence of area is proved.
The proof of the uniqueness of area is very simple, but we are lack of just one additional auxiliary
fact: if figures  and  are measurable, then the figure  \  is also measurable. If we have this
fact, then the proof of the uniqueness can be performed very quickly. But now it’s not clear how to
obtain this result from the results that we have above. In order to prove this one we need to
introduce the last measurability criterion. By using it, we will be able to get the needed result and to
prove the uniqueness. Also the last [3-rd criterion of measurability] has a great importance in math.

Def: a figure X is called a zero area figure if it’s area is zero, S ()  0 .
It’s very easy to understand that for any zero area figure: any internal figure  k is an empty figure.
int

Really, if we assume that there is some net-square in  k then the area S (  k ) is a positive
int int

number. For any figure, the sequence of areas of internal figures is monotonically increasing:
S (int
k )  S (  k 1 )  S (  k  2 )  .... , then starting from the number k all the terms of the sequence
int int

{S ( int
n )} are greater than the fixed positive number S (  k ) , then {S (  n )} can’t converge to zero
int int

and S ( )  0 , and we have a contradiction. So, any internal figure  k is an empty figure,
int

2 n
it doesn’t contain any squares at all, i.e., n  0 n   , then n 10  S ( int
n )  0 n   .

And what about external figures  n ? These figures already contain some squares,
ext

and the sequence S (  n ) is monotonically decreasing (this sequence is decreasing for any bounded
ext

figure, not only for  ) and goes to zero: {S (  n )}  0 .

ext

Assertion7. A union    of any zero-area figures  and  is again a zero-area figure.

Proof. Let  and  are zero-area figures. For any n   we obviously have
(  )ext
n  n   n [L]. Then S ((  ) n )  S (n )  S (  n ) . Both sequences {S (n )} and
ext ext ext ext ext ext

n )} go to zero, then {S ((  ) n )} also goes to zero. So, the sequence of areas of external
{S ( ext ext

quadratic figures for    goes to zero, then S (  )  0 .

Assertion8. Any part of a zero-area figure is again a zero-area figure, i.e., S ()  0 and    ,
then S ( )  0 .

Proof. Obviously  n  n , then S (  n )  S (n ) . From {S (n )} n 0 follows that

ext ext ext ext ext


{S ( ext
n )} n 0 .


160
 Construction of Length and Area

Assertion9.  is a plane figure (or a space figure). Any segment  which connects an internal
point  (of  ) and an external point  (of  ) contains at least one boundary point of  .

Proof. Without loss of generality, let Ω is some plane figure. Let’s fix any internal point  and any
external point  . Let’s draw the coordinate line L through
 and  such that  is a point with a zero coordinate,
and the segment  lies on the positive ray of L.
Let’s consider all the internal points of  which lie on L,
this set is not empty (at least it contains  ), coordinates
~
of these points form some set R of real numbers, this set
~
is bounded above, by the coordinate b of  . Then R has pict.10
a supremum h which can be pictured as a point H, let’s
show that H is a boundary point of  [pict10]. If we assume that H is an internal/external point
of  , then it has the neighborhood O () which (completely belongs to  )/(does not have any
common points with  ), then the interval O ()  L on L (completely belongs to  )/(does not
have any common points with  ). Then there is an interval of real numbers with the center at h
~ ~
which (belongs to R )/(does not have any common points with R ), in both cases h is not
~
a supremum of R . Then H is a boundary point of  .
Consequence2. Any segment  which connects any point    and any point    , contains
at least one boundary point of  .
Proof.    then there can be only two variants: A is a boundary point of  , or A is an internal

point of  . If A is a boundary point, then AB contains the boundary point A. Let A is an internal
point (of  ). We know that    , then there are exactly two variants: B is a boundary point of 
(then AB contains the boundary point B), or B is an external point (of  ), then, according to
the assertion9,  contains some boundary point. Everything is proved.

Assertion10 [3-rd criterion of measurability].

 is measurable  the boundary  is a zero-area figure.
Proof.  Let  is measurable, we need to show that
S ()  0 . Let’s fix any n   and a quadratic net for n .
We fix any net-square which has a common point С with 
[pict11] (it is a square from ()ext
n ), this square has exactly

8 neighbor-squares around. All these 9 squares together form

a bigger square П. As С belongs to the center-square, we can pict.11
draw a small neighborhood O (C ) which belongs to П, this
161
 Construction of Length and Area

neighborhood must contain one point A   and one point B   (because C is a boundary point).
Then the big square П contains at least one point A from  and one point B not from  .
Let’s show that some small square (one of the 9 squares that formП) has the same property.
Here is a simple logic: any small square may [A] have a common point with  , but do not belong to
 completely [B] belong to  [C] do not have any common points with  .

If some small square is such that [A], then is has the same property as П (contains one point from
 and one not from  ). Let’s assume that there is no such small square, so every small square is
such that [B] or [C]. All 9 small squares can’t be such that [B], because then П does not contain
any point not from  . And similarly, all 9 small squares can’t be such that [C], because then П
does not contain any point from  . And what if some of these 9 squares completely belong to Ω,
and the others do not have any common points with  ? In such case there will be two small
neighbor-squares with a common side, such that one square completely belongs to  , and the other
one has no common points with  , so the common side of these squares belongs and doesn’t belong
to  at the same time, and we have a contradiction.
Then there exist some small square (among these 9 squares) such that [A]. This small square

contains some point A   and therefore it belongs to ext

n , and it also contains some point B   ,

then it does not belong to  n . So that square belongs to

int
ext
n \ n .
int

Let’s sum up: we had started from any square  from () n , and we deduced that the square  ,
ext

or one of 8 squares around  belongs to n \ n . Let’s estimate from above the number of
ext int

squares in () n . The most convenient way to do it is to build the correspondence

ext

f : ()ext
n  n \ n .
ext int

For any concrete square  from () n we define f ()  n \ n , where  is the square
  ext
ext int

from n \ n (here
ext int
  or  is one of 8 squares around  ).

Then some of the squares of the figure n \ n correspond to the squares of () n .
ext int ext

There are exactly  n ( for )  n ( for ) different squares in n \ n and any square may
ext int

correspond maximum to 9 squares (to itself and to every of 8 surrounding squares), so

 
9   n ( for )  n ( for ) is the maximal number of squares that can correspond to the squares

of n \ n , then there can’t be more than 9   n ( for )  n ( for ) squares in the figure
ext int

()ext 
n . Let’s write it:  n ( for )  9   n ( for )  n ( for ) let’s multiply both sides by 
102 n , we will get S (()ext  
n )  9  S (n )  S (n ) . When n   the right part of the last
ext int

equality goes to zero (because  is measurable), then S (()ext

n ) n
  0 then S ()  0 .


162
 Construction of Length and Area

Conversely.  (More simple). Let  is a zero-area figure. Let’s take any net-square  which has
a common point with  , but does not belong to  completely, it is a square from n \ n .
ext int

That  contains some point A   and some point B   , the segment AB belongs to  and
(consequence from assertion9) contains a boundary point С   . Then  belongs to () n .
ext

Therefore, any square from ext

n \  n is a square from () n , so we have n \ n  () n .
int ext ext int ext

Then the number of squares in n \ n is not greater than the number of squares in () n .
ext int ext

Let’s write it:  n ( for )  n ( for )   n ( for ) , we multiply both sides by 102n , then

n )  S (n )  S (() n ) . When n   the right part goes to zero, then the left part also
S (ext int ext

goes to zero and  is measurable.

Consequence3. If ,  are measurable figures, then  \  is also a measurable figure.

The main idea: As ,  are measurable, then their boundaries ,  are zero-area figures
(assertion10), then    is a zero area figure (assertion7).
We will show that the boundary ( \ ) of the figure  \  belongs to    , i.e.,
( \ )     , then (assertion8) the boundary ( \ ) is a zero-area figure.
Then (assertion10 )  \  is measurable.
So we just have to show that ( \ )     and the case is done.
С
Let’s define: for any figure  on the plane  , the figure  \    is called a complement of  .

[Fact 1] for any figures ,  we have  \      С (this one is very simple, we just need
to write what kind of points belong to  \  , and what kind of points belong to    .
С

We will see that these are exactly the same points).

[Fact 2] Any figure  and it’s complement С always have a common boundary, i.e.,   С
С
(just use the definition of a boundary point and show that   
С
and    ).

[Fact 3] for any figures ,  we have (  )    

(here we need to fix any boundary point A of    , and it’s very easy to deduce that this point
may be: a boundary point of  , or a boundary point of  , or a boundary point of  and a boundary
point of  , so in any case, A belongs to the union    , from here follows
(  )     )
С С
We are ready to prove [Y]: [ Fact 1]  \     ( )  ( \ )  (  ( )) 

 [ Fact 3] : ( \ )     С  [ Fact 2] : ( \ )     . And [T] is proved.

We are ready to prove the uniqueness of area.

~
Uniqueness of area. Let there exist some correspondence S which is defined on the collection
~
of all measurable figures (def4) and satisfies to [1],[2],[3], then S  S . The proof here is simple,
163
 Construction of Length and Area
~
let’s describe the main steps. For any 1 1 square: S (1  1)  1 (according to [1]). We can draw
2
several parallel lines in order to divide 1 1 square into n equal small squares with sides 1 / n each.
~
As [1] S (11)  1 , then by additivity, the area of any 1 / n 1 / n square is 1 / n (and it is true for
2

any n   ).

Let now  is any measurable figure.

~ ~
If    , then S ()  S () [monotonicity]. Really, let    are measurable figures, then  \ 
is measurable (consequence3). And we have    \    , where figures  and  \  do not have
~ ~ ~ ~ ~
any common internal points, then by additivity [2]: S ()  S ( \ )  S ()  S ()  S () .
~ ~
Next: as S coincides with S on every 1/ n  1/ n square, then (by additivity [2]), S coincides with S
on every figure  n and  n . As  is measurable, we have
int ext

lim n  S (int
n )  lim n  S (n )  S () , then we have:
ext

~ ~ ext ~ int ~ ext

[U] lim n  S (int
n )  lim n  S (n )  S () . So, the sequences {S ( n )} and {S ( n )} go to the

same limit S () . In the same time for any n we have int
n    n  [monotonicity] 
ext

~ ~ ~ ext ~ int
S (int
n )  S ()  S (n ) and here we can use the squeeze theorem for sequences {S (n )} and
~ ~ ~ ~ ~ ~
n )} and {S ()}  S (), S (), S (), S ()....
{S (ext

~ ~ ~
Both sequences {S (n )} and {S (n )} [U] go to the same limit
int ext
S () , then {S ()} goes to the
~ ~
same limit, therefore S ()  S () . So S coincides with S on every measurable figure  , it means
~
that S  S and the area is unique.

The definition of a volume V and it’s properties (including uniqueness) are exactly similar.
Moreover, the construction process and the theorems (including criterions of measurability 1,2,3)
are also exactly similar.

164
 Angles

Angles
When we speak about points, lines, rays, segments (on the plane or in the space) we do not define
these objects, because for us these are basic “elementary” objects. Both plane geometry and
stereometry have it’s own axioms and basic objects like “planes”, “lines” and etc. (in fact the plane
geometry is a part of stereometry). Basic geometrical axioms and theorems is an independent part
of mathematics, it does not depend on the theory of real numbers.

One of the basic geometrical figures is the angle BOA , it is a figure which consists of two rays OB
and OA and the part of the plane, which lies “between” these rays. In any particular case we have to
show which part of the plane between OB and OA is a part of our angle (because there is always
two variants). The point O is called a
“vertex of the angle”. Our basic angles are: pict.1
the right angle (right angle) ,
the straight angle ( straight angle)

and the full angle ( full angle) [pict1] .

We can compare angles [pict2] in the similar

way as we compare segments. The writings
   and    are equivalent.

pict.2
We can add any two angles  and  if they

are both are less than the straight angle [pict3],

the result angle is called a “sum of angles
 and  ” and we denote it (   ) .
For any angle  :
pict.3
  ...    [n summands ]  n , this
definition is correct only for such numbers n,
for which the sum   ...   is less than
the full angle. It’s easy to see that:
(straight angle)  2(right angle)
and ( full)  4(right angle) .

If we have n   , then we can write   ( / n) .

166
 Angles

Half angle assumption. For any angle 

there exist the angle  such that
  ( / 2) [pict4]. We also assume that
the right angle (right angle) can be divided
into 90 equal parts, so there exist “the unit
angle” e such that (right angle)  90e . pict.4

Archimedes axiom. For any angles  and

 (where   (straight angle) and
  (straight angle) ) there exist some
natural n such that n   . [pict5]. pict.5

Nested angles axiom. If we have some sequence { n }

of nested angles: 1   2  3  ...., with a common
vertex O , then there exist the ray OA , which belongs to every
angle  n || n [pict6].

pict.6

Def.   { ||   ( full angle)}. The “Angle value” is the correspondence L :   R  which
compares only one non-negative real number L( ) for every angle  . And also:
[1] L(e)  1 [2] for any angles  and  (where   (straight angle) and
  (straight angle) ) we have: L(   )  L( )  L( ) .
[3] If    , then L( )  L( ) .
For any angle  , the number L( ) is called a value of the angle  .
The construction of the angle value L is very similar to the length construction.
For any angle  the number L( ) is usually written with the “degree symbol”
o
right after it.

Like L(e)  1 , L(right angle)  90 , L(straight angle)  180 . In practice we never say
o o o

“the value of the angle  is one degree”, but only “  is one degree”.

Property.     L( )  L( ) . And for any real number a  [0o ,360 o ] there exist
the angle  such that L( )  a .

167
 Angles

Def1. Any angle, which is less than the right angle, is called an “acute angle”.
An angle  is acute if and only if L( )  [0o ,90o ) . Any angle, which is greater than the right
angle and less than the straight angle, is called an “obtuse angle”. An angle  is obtuse if and

only if L( )  (90o ,180o ) . In practice we neglect the letter L and the symbol  , instead
of L( )  50o we write just   50o , instead of L( )  30o we write   30o .
Trigonometric functions. In the course of plane geometry basic trigonometric functions
sin, cos,tg, ctg are defined for any acute angle  through the right triangle. We just need to draw
any right triangle with an angle  and calculate a ratio of relevant sides.
This definition is correct, because all the right triangles with the same angle  are similar, and
the ratio of relevant sides in any triangle is always the same. For any acute angle  the main
sin  cos
values are sin  and cos , because tg  , ctg  .
cos sin 
By using simple ideas from elementary geometry we can build the small table:

 / 30o 45o 60o

sin 1/ 2 1/ 2 3/2
cos 3 / 2 1/ 2 1/ 2
tg 1/ 3 1 3
ctg 3 1 1/ 3

Let’s extend the notion of angle and trigonometric functions.

We fix any coordinate system Oxy and draw the unit circle with
the center at the origin. We also mark the point A , at which
Ox intersects the circle.
Let’s take any angle   [0o , 360o ] . By moving along the circle
from the point A in the counterclockwise direction we build the
angle AOB   [pict7] , the point B has some coordinates
pict.7
xB and yB , then we define:
sin  cos
cos  // by definition //  xB || sin   yB || tg  || ctg  || .
cos sin 

It’s very important to notice that this definition of trigonometric functions sin, cos,tg, ctg extends
the old definition, where trigonometric functions are defined only for acute angles.

168
 Angles

Really, let’s take any acute angle  and build the angle
AOB   , then we can draw BH  Ox . We will get
the right triangle, which hypotenuse OB is 1 .
The numbers xB and y B are coordinates of the point B,
but in the same time these numbers are cathetuses
of the right triangle OBH [pict8].
So, according to the old definition:
yB x
 sin   yB and B  cos  xB .
1 1 pict.8
Then the new definition is really an extension of the old one.

Next, the function tg is not defined when cos  0 (when   90o or   270o ),
and ctg is not defined when sin   0 (when   0 or   180o ).
o

For any angle  , the values sin  and cos are numbers from the segment [1,1] , the only

restriction is: sin

2
  cos2   1. Really, any angle  [0o ,360o ] defines some point B on
the circle such that   AOB , and for any point B on the unit circle we obviously have
xB2  yB2  1  sin 2   cos2   1 it follows from the Pythagorean theorem for the right triangle
with sides xB , y B , 1. The values of tg and ctg can be any real numbers, the only restriction

is: ctg  1/ tg .

o o
We have defined sin, cos,tg, ctg for any angle from [0 , 360 ] . Let’s extend our definition.

At first we have to extend our domain, we have to consider the set of numbers (360 , ) .
o o

Let’s fix any number   360o . By definition, any angle   360o is the combination:
( k full counterclockwise rotations around the circle starting from the point A ) and (the angle  ),
where both numbers k ,  are taken from the representation

  360o  k   || k  ,  [0o ,360o ) . Such representation is obviously unique for any number
  360o .

For example, the angle   480o is one full counterclockwise rotation and the angle   120o .
The angle   980o is two full counterclockwise rotations and the angle   240o .

Any angle   360o defines some point B on the circle, with coordinates ( xB , yB ) . By definition:
sin  cos
cos  xB || sin   yB || tg  || ctg  || .
cos sin 
We have defined what is the angle  [0o , o ) and it’s trigonometric functions.
169
 Angles

Let’s fix any number   0o . The negative angle  is the angle  , which is built in the clockwise
direction from the point A.
In particular, any negative angle   360o is the combination of
( k full clockwise rotations around the circle starting from the point A ) and (some negative angle
   (360o , 0o ] ).
Properties. [A] For any angle  , the points on the circle, which correspond to the angles  and
  , are symmetric with respect to Ox , so these points have equal Ox -coordinates [pict9], then
cos  cos( ) . These points also have opposite Oy -coordinates, then sin    sin(  ) .
[B] For any angle  , the points on the circle, which correspond to angles  and 180o   are
symmetric with respect to Oy , therefore sin(180o   )  sin  and cos(180o   )   cos .

[C] Let’s fix any positive acute angle   [0o ,90o ) and mark the angles  and 90o   [pict10].
It’s easy to build two equal right triangles, from their equality follows that sin   cos(90o   )

and cos  sin(90   ) . In fact these formulas are true for any angle
o
.

pict.9 pict.10

Really, it’s easy to understand that for any angle  , the points on the circle, which correspond to
angles  and 90   , are always symmetric with respect to the angle bisector (red dashed line in
o

[pict10]). The symmetry with respect to this bisector changes the places of Ox and Oy , from here
follows the assertion we need.

From [A],[B],[C] we can deduce some trigonometric values. For example, we know that
sin 45o  cos 45o  1/ 2 , then from [A] follows that cos(45o )  cos 45o  1/ 2 and
sin( 45o )   sin 45o  1/ 2 and from [B] follows that cos(135o )   cos 45o  1/ 2 and
sin(135o )  sin 45o  1/ 2 .

In many cases it’s more convenient to use “radians” instead of angles.

Radians are defined in the next book.
170
 Literature

Literature

1. Encyclopedia of elementary mathematics. Volume1.

P.S Aleksandrov, A.I Markushevich,A. Hinchin. Leningrad, 1951.

2. Encyclopedia of elementary mathematics, Volume 3.

P.S Aleksandrov, A.I Markushevich,A. Hinchin. Leningrad, 1951.

3. Foundations of geometry, A.D Aleksandrov. Moscow, 1987.

4. Numbers and polynomials. I.V Proskuryakov. Moscow, 1965.

5. Course of differential and integral calculus,Volume 1. G.M. Fihtengolz. FizMatLit, 2001.

6. Mathematical Analysys. M.V Falalleev. Irkutsk 2013.

7. Mathematical Analysis. A.I.Kozko. MGU 2017

8. Lecture Course in Mathematical Analysis. V.I. Kolyada, A.A Korenovski. Odessa 2009.

9. Course of mathematical analysis. A.M. Ter-Krikorov, M.I. Shabunin. FizMatLit, 2001

171