SERNIAv1
SERNIAv1
SERNIAv1
Yaroslav D. Sergeyev∗
Abstract. In this survey, a recent computational methodology paying a special attention to the sep-
aration of mathematical objects from numeral systems involved in their representation is described.
It has been introduced with the intention to allow one to work with infinities and infinitesimals nu-
merically in a unique computational framework in all the situations requiring these notions. The
methodology does not contradict Cantor’s and non-standard analysis views and is based on the Eu-
clid’s Common Notion no. 5 ‘The whole is greater than the part’ applied to all quantities (finite,
infinite, and infinitesimal) and to all sets and processes (finite and infinite). The methodology uses a
computational device called the Infinity Computer (patented in USA and EU) working numerically
(recall that traditional theories work with infinities and infinitesimals only symbolically) with infi-
nite and infinitesimal numbers that can be written in a positional numeral system with an infinite
radix. It is argued that numeral systems involved in computations limit our capabilities to compute
and lead to ambiguities in theoretical assertions, as well. The introduced methodology gives the
possibility to use the same numeral system for measuring infinite sets, working with divergent se-
ries, probability, fractals, optimization problems, numerical differentiation, ODEs, etc. (recall that
traditionally different numerals ∞, ℵ0 , ω, etc. are used in different situations related to infinity).
Numerous numerical examples and theoretical illustrations are given. The accuracy of the achieved
results is continuously compared with those obtained by traditional tools used to work with infinities
and infinitesimals. In particular, it is shown that the new approach allows one to observe mathemat-
ical objects involved in the Hypotheses of Continuum and the Riemann zeta function with a higher
accuracy than it is done by traditional tools. It is stressed that the hardness of both problems is not
related to their nature but is a consequence of the weakness of traditional numeral systems used to
study them. It is shown that the introduced methodology and numeral system change our perception
of the mathematical objects studied in the two problems.
Mathematics Subject Classification (2010). 65-02, 03-02, 11-02, 40-02, 00A30, 03E65, 03E50.
Keywords. Numerical infinities and infinitesimals, numbers and numerals, grossone, Infinity Com-
puter, numerical analysis, infinite sets, divergent series, Continuum Hypothesis, Riemann zeta func-
tion, numerical differentiation, ODEs, optimization, probability, fractals.
∗ The author is very grateful to the reviewers for their careful and meticulous reading of the paper.
Yaroslav D. Sergeyev, University of Calabria, Via P. Bucci, Cubo 42-C, 87030 Rende, Italy; Lobachevsky
University, Nizhni Novgorod, Russia; Institute of High Performance Computing and Networking of the Na-
tional Research Council of Italy, Rende, Italy
E-mail: yaro@dimes.unical.it
2 Yaroslav D. Sergeyev
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Numeral systems and numbers they can express . . . . . . . . . . . . . . . . . 6
2.1 Differences among numeral systems used to express finite and infinite
quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Amazonian and Australian tribes prompt a new point of view on infinity . 7
3 The computational methodology . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.1 Methodological postulates . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Counting huge finite quantities in practice – a granary example . . . . . . 16
4 A new way of counting and the infinite unit of measure ¬ . . . . . . . . . . . . 17
4.1 The Infinite Unit Axiom and first examples . . . . . . . . . . . . . . . . . 17
4.2 Postulate 3 versus bijections of infinite sets: no contradiction . . . . . . . 24
4.3 Positional numeral system with the infinite radix ¬ . . . . . . . . . . . . 27
4.4 Arithmetical operations with ¬-based numerals . . . . . . . . . . . . . . 29
5 Numerical infinities and infinitesimals in applications . . . . . . . . . . . . . . 32
5.1 Higher order numerical differentiation on the Infinity Computer . . . . . . 32
5.2 Infinitesimal perturbations for solving systems of linear equations . . . . 34
5.3 Infinite penalty coefficients in constrained non-linear optimization . . . . 37
5.4 Lexicographic multi-objective linear programming . . . . . . . . . . . . 39
5.5 Numerical methods using infinitesimals for solving ODEs . . . . . . . . . 40
6 Traditional mathematical objects viewed from the new methodological positions 44
6.1 Infinite sequences and Turing machines . . . . . . . . . . . . . . . . . . 44
6.2 From divergent series to sums with a fixed infinite number of addends . . 47
6.3 Riemann series theorem and Ramanujan summation . . . . . . . . . . . . 51
6.4 From limits to expressions evaluated at actual infinite/infinitesimal points 55
6.5 Functions, continuity, derivatives, and integrals . . . . . . . . . . . . . . 59
7 Measuring infinite sets and the Continuum Hypothesis . . . . . . . . . . . . . . 65
7.1 Measuring infinite sets with elements defined by formulae . . . . . . . . . 65
7.2 Countable sets can be measured more accurately . . . . . . . . . . . . . . 67
7.3 The number of elements of sets having the cardinality of the continuum . 70
7.4 Relations to results of Georg Cantor . . . . . . . . . . . . . . . . . . . . 72
8 Repercussions on the Riemann zeta function . . . . . . . . . . . . . . . . . . . 76
8.1 Some Euler’s results revisited in the ¬-based framework . . . . . . . . . 78
8.2 Calculating ζ(s, n) for different infinite values of n at s = 0, −1, −2, etc . . 83
9 Grossone for studying probability and fractals . . . . . . . . . . . . . . . . . . 87
9.1 Probability and absence of nonempty sets of measure zero . . . . . . . . . 87
9.2 Koch snowflake and its quantitative analysis using ¬-based numerals . . . 89
10 A brief conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Numerical infinities and infinitesimals 3
1. Introduction
In different periods of human history, mathematicians and physicists developed mathe-
matical languages and philosophical positions leading to different approaches to the ideas
of infinite and infinitesimal (see [3, 6, 10, 16, 23, 25, 42, 44, 46, 49, 51, 53, 59, 63, 64,
72, 77, 82, 106, 111, 112, 113]) and references given therein). There is no need to explain
the importance of this subject for Mathematics. Let us just mention that the Continuum
Hypothesis related to infinity has been included by David Hilbert as the Problem Number
One in his famous list of 23 unsolved mathematical problems (see [53]) that have influ-
enced strongly development of Mathematics in the XX th century. The famous Riemann
Hypothesis is the problem number 8 in this list and it is also related to infinity since the
Riemann zeta function is introduced as an infinite series.
Many approaches describing manipulations with infinities and infinitesimals are rather
old: ancient Greeks following Aristotle distinguished the potential infinity from the actual
infinity; John Wallis (see [112]) credited as the person who introduced the infinity symbol,
∞, published his work Arithmetica infinitorum in 1655; the foundations of analysis we use
nowadays were developed more than 200 years ago with the goal to develop mathematical
tools allowing one to solve problems that were emerging in the world at that time; Georg
Cantor (see [16]) has introduced his cardinals and ordinals more than 140 years ago,
as well. As a result, mathematical languages that we use now to work with infinities
and infinitesimals do not reflect numerous achievements made by Physics of the XX th
century1 . Let us illustrate this observation by a couple of examples.
We know from the modern Physics that the same object can be viewed as either dis-
crete or continuous in dependence on the instrument used for the observation (we see a
table continuous when we look at it by eye and we see it discrete (consisting of molecules,
atoms, etc.) when we observe it through a microscope). In addition, physicists do not give
absolute results of their observations. Together with the result of the observation they al-
ways supply the accuracy of the instrument used for this observation.
In Mathematics, both facts are absent: each mathematical object (e.g., function) is
either discrete or continuous and nothing is said about the accuracy of the observation of
the mathematical objects and about tools used for these observations. The mathematical
notion of continuity itself comes from the XIX th century. Many of mathematical notions
have an absolute character and the ideas of relativity are rarely present in them. One of the
main ideas of Physics of the XX th century regarding ideas of the influence of the instru-
ment of an observation on the object of the observation are almost absent in Mathematics,
as well.
In some sense, there exists a gap between achievements in Physics made during the
last two hundred years (especially during the XX th century) and their mathematical mod-
1 Even the brilliant efforts of the creator of the non-standard analysis Robinson that were made in the middle
of the XX th century have been also directed to a reformulation of the classical analysis (i.e., analysis created
two hundred years before Robinson) in terms of infinitesimals and not to the creation of a new kind of analysis
that would incorporate new achievements of Physics. In fact, he wrote in paragraph 1.1 of his book [82]: ‘It is
shown in this book that Leibniz’s ideas can be fully vindicated and that they lead to a novel and fruitful approach
to classical analysis and to many other branches of Mathematics’ (italics mine). In fact, classical analysis
reformulated using the language of Robinson was then used to address such areas as topology, probability, etc.
4 Yaroslav D. Sergeyev
els that continue to be written using the mathematical language developed two centuries
ago on the basis of (among other things) physical ideas of that remote time that now are
absolutely outdated.
In particular, in relation to the concepts of infinite and infinitesimal we have an anal-
ogous situation. In fact, the point of view on infinity accepted nowadays (see [3, 23, 27,
51, 58, 59, 113]) takes its origins from the famous ideas of Cantor (see [16]) who in the
second half of the XIX th century has shown that there exist infinite sets having different
cardinalities. Infinitesimals have been developed even earlier when, in the early history
of calculus, arguments involving infinitesimals played a pivotal role in the differential
calculus developed by Leibniz and Newton (see [63, 77]). At that time the notion of an
infinitesimal, however, lacked a precise mathematical definition and in order to provide
a more rigorous foundation for the calculus infinitesimals were gradually replaced by the
d’Alembert-Cauchy concept of a limit (see [19, 28]).
The creation of a rigorous mathematical theory of infinitesimals on which it would
be possible to construct Calculus remained an open problem until the end of the 1950s
when, as was already mentioned, Robinson (see [82]) has introduced his non-standard
analysis approach (Levi-Civita has proposed his transfinite numbers in [64] much earlier
but they have not received so broad dissemination as the approach of Robinson). He has
shown that non-archimedean ordered field extensions of the reals contained numbers that
could serve as infinitesimals and their reciprocals could serve as infinitely large numbers.
Robinson then has derived the theory of limits, and more generally of calculus, and has
found a number of important applications of his ideas in many other fields of Mathematics
(see [82]). It is important to emphasize that in his approach Robinson used Cantor’s
mathematical tools and terminology (cardinal numbers, countable sets, continuum, one-
to-one correspondence, etc.) thus incorporating advantages and disadvantages of Cantor’s
approach into non-standard analysis.
After these few introductory words let us define the aim of the present text. The first
goal of this paper is to bridge the gap between modern Physics and Mathematics at least
partially, insofar as it concerns the separation of an object and a tool used to observe
it existing in Physics and often not present in Mathematics. The main attention will be
dedicated to infinities and infinitesimals and ways of their representation in comparison
with representation of finite numbers. Recall that the same numerals are used to represent
finite numbers in all the occasions we need them, in contrast, different infinities are used in
different situations. Traditionally, these various kinds of infinity are perceived by people
as distinct mathematical objects. The approach developed here shows that this is not the
case. It argues that the objects – infinite numbers – are the same, they are just viewed in
different ways by different theories. For example, in analysis and set theory, ∞ and ℵ0 are
not two different objects, they are two different images of infinite quantities that people
see through different ‘lenses’ provided by analysis and set theory.
Then, the second goal of the paper is to describe a numeral system expressing infinities
and infinitesimals that can be used in all situations where we need to work with them.
This is done in order to have a situation similar to what we have with finite numbers
where numerals expressing them can be used in all the occasions we need finite quantities.
Several examples of the usage of the new numerals expressing infinities and infinitesimals
are given to illustrate their versatility. In particular, numerous numerical applications and
Numerical infinities and infinitesimals 5
putations are the exact manipulations with mathematical expressions containing variables that have not any
given value and are thus manipulated as symbols.
6 Yaroslav D. Sergeyev
system gives a possibility to express finite, infinite, and infinitesimal numbers in a unique
framework and to execute numerical (not symbolic) operations with all of them. Sec-
tion 5 discusses a variety of numerical applications where the new methodology can be
useful. Traditional mathematical objects viewed from the new methodological positions
are studied in Section 6 whereas Section 7 establishes relations of the new methodology
to some of results of Cantor and the Continuum Hypothesis. Repercussions of the new
methodology on the Riemann zeta function are discussed in Section 8. Examples related
to probability and fractals are given in Section 9. Finally, Section 10 concludes the survey.
3 The last two numerals, = and Ĩ, are probably less known. The former belongs to the Maya numeral system
where one horizontal line indicates five and two lines one above the other indicate ten. Dots are added above
.
the lines to represent additional units. For instance, numerals = and = mean, respectively, ten and eleven in this
numeral system. The latter symbol, Ĩ, belongs to the Cyrillic numeral system derived from the Cyrillic script.
This numeral system was developed in the late X th century and was used by South and East Slavic peoples. The
system was used in Russia as late as the early XV III th century when it was replaced with Arabic numerals. To
distinguish numbers from text a special sign, ˜ , called titlo is drawn over the symbols showing so that this is a
numeral and, therefore, it represents a number and not just a character of text.
Numerical infinities and infinitesimals 7
computations with finite quantities are concrete and not generic. If we consider a finite
n then different values can be assigned to it, e.g., we can use the numeral 34 and write
n = 34 (clearly, any other known numeral used to express finite quantities and consisting
of a finite number of symbols can be taken for this purpose). Notice that the finiteness of
the number of symbols in the numeral is necessary for executing practical computations
since we should be able to write down (and/or to store) values we execute operations with.
In contrast, if we consider a non-standard infinite m then it is not clear which numerals
consisting of a finite number of symbols can be used to assign a concrete value to m since
non-standard analysis does not provide numeral systems that can be used for this purpose.
For a generic infinitesimal ε again it is not clear which numerals can be used to assign a
value to ε and to write ε = ... . Moreover, approaches of this kind leave unclear such
issues as, e.g., whether the infinite 1/ε is integer or not or whether 1/ε is the number
of elements of a concrete infinite set. If one wishes to consider two infinitesimals (or
infinities) h1 and h2 , where h2 is not expressed in terms of h1 , then it is not clear how
to compare them because numeral systems that can express different values of infinities
and infinitesimals are not provided by this kind of techniques. In contrast, when we work
with finite quantities, then we can compare n and k if they assume numerical values, e.g.,
k = 25 and n = 78 then, by using rules of the numeral system the symbols 25 and 78
belong to, we can compute that n > k.
Third, many arithmetics used to deal with infinities not only should be used for spe-
cific purposes only but in addition they are quite different with respect to the way we
execute computations with finite quantities. Let us give some examples:
– There exist undetermined operations (∞ − ∞, ∞
∞ , etc.) that are absent when we work
with finite numbers.
– Arithmetic with ordinals is very different from arithmetic with finite quantities.
For instance, addition and multiplication are not commutative (e.g., 1 + ω = ω and
ω + 1 > ω), there does not exist an ordinal γ such that γ + 1 = ω, etc.
– Arithmetic with ∞ and infinite cardinals does not satisfy Euclid’s Common No-
tion 5 saying ‘The whole is greater than the part’ that holds when we work with
finite quantities. In fact, it follows for these numerals, e.g., ∞ + 1 = ∞ and ℵ0 + 1 =
ℵ0 whereas clearly for any finite x it follows x + 1 > x.
Even though there exist codes allowing one to work symbolically with ∞ and other
symbols related to the concepts of infinity and infinitesimals, traditional computers work
numerically only with finite numbers and situations where the usage of infinite or in-
finitesimal quantities is required are studied mainly theoretically (see [6, 16, 25, 44, 49,
53, 63, 64, 82, 112] and references given therein). The fact that numerical computations
with infinities and infinitesimals have not been implemented so far on computers can be
explained by difficulties mentioned above. There exist also practical troubles that pre-
clude an implementation of numerical computations with infinity and infinitesimals. For
example, it is not clear how to store an infinite quantity in a finite computer memory.
2.2. Amazonian and Australian tribes prompt a new point of view on infinity. In
order to understand how one can change his/her view on infinity, let us consider some
8 Yaroslav D. Sergeyev
numeral systems used to express finite numbers. During the history of humanity numeral
systems (see, e.g., [25, 27]) became more and more sophisticated allowing people to
express more and more numbers. Since new numeral systems appear very rarely, in each
concrete historical period people tend to think that any number can be expressed by the
current numeral system and the importance of numeral systems for Mathematics is very
often underestimated (especially by pure mathematicians who often write phrases of the
kind ‘Let us take any x from the interval [1, 2]’ and do not specify how this operation of
taking can be executed). However, if we observe the situation with expression of numbers
in the historical prospective, we can immediately see limitations that various numeral
systems induce. It becomes clear that due to practical reasons different numeral systems
can express different limited sets of numbers and they can be more or less suitable for
executing arithmetical operations.
In fact, all numeral systems have practical limitations on the numbers they can √ ex-
press. Even the powerful positional system is not able to express, e.g., the number 2 by
a finite number of symbols (remind that the finiteness of symbols√in operands is essential
for executing numerical computations) and this special numeral, 2, is deliberately intro-
duced to express the desired quantity. Moreover, there exists another practical limitation
that prohibits us to work even with numbers having a finite number of digits in their posi-
tional records. It consists of the fact that nobody is able to write down very long numbers,
e.g., a numeral in the decimal positional system having 10100 digits cannot be written
due to physical limitations even by the fastest existing computers. As a consequence, no
practical operations can be done with these numbers, no theorems involving them can be
formulated, etc.
There exist many numeral systems that are weaker than the positional one. For in-
stance, Roman numeral system is not able to express zero and negative numbers and such
expressions as III – VIII or X – X are indeterminate forms in this numeral system. As
a result, before the appearance of positional systems and the invention of zero (by the
way, the second event was several hundred years later with respect to the first one) math-
ematicians were not able to create theorems involving zero and negative numbers and to
execute computations with them. The appearance of the positional numeral system not
only has allowed people to execute new operations but has led two new theoretical results,
as well. Thus, numeral systems not only limit us in practical computations, they induce
boundaries on theoretical results, as well.
Even though Roman numeral system is weaker than the positional one it is not the
weakest numeral system. There exist very weak numeral systems allowing their users to
express very few numbers and one of them is illuminating for our study. This numeral
system is used by a tribe, Pirahã, living in Amazonia nowadays. A study published in
Science in 2004 (see [45]) describes that these people use an extremely simple numeral
system for counting: one, two, many. For Pirahã, all quantities larger than two are just
‘many’ and such operations as 2+2 and 2+1 give the same result, i.e., ‘many’. Using their
weak numeral system Pirahã are not able to see, for instance, numbers 3, 4, and 5, to
execute arithmetical operations with them, and, in general, to say anything about these
numbers because in their language there are neither words nor concepts for that.
It is worthy of mention that the result ‘many’ is not wrong. It is just inaccurate.
Analogously, when we observe a garden with 517 trees, then both phrases: ‘There are
Numerical infinities and infinitesimals 9
517 trees in the garden’ and ‘There are many trees in the garden’ are correct. However,
the accuracy of the former phrase is higher than the accuracy of the latter one. Thus, the
introduction of a numeral system having numerals for expressing numbers 3 and 4 leads
to a higher accuracy of computations and allows one to distinguish results of operations
2+1 and 2+2.
The poverty of the numeral system of Pirahã leads also to the following results
that are crucial for changing our outlook on infinity. In fact, by changing in these relations
‘many’ with ∞ we get relations used to work with infinity in the traditional calculus
∞ + 1 = ∞, ∞ + 2 = ∞, ∞ + ∞ = ∞. (2.2)
ℵ0 + 1 = ℵ0 , ℵ0 + 2 = ℵ0 , ℵ0 + ℵ0 = ℵ0 , (2.3)
c + 1 = c, c + 2 = c, c + c = c. (2.4)
It should be mentioned that the astonishing numeral system of Pirahã is not an isolated
example of this way of counting. In fact, the same counting system, one, two, many, is
used by the Warlpiri people, aborigines living in the Northern Territory of Australia (see
[14]). The Pitjantjatjara people living in the Central Australian desert use numerals one,
two, three, big mob (see [62]) where ‘big mob’ works as ‘many’. It makes sense to remind
also another Amazonian tribe – Mundurukú (see [79]) who fail in exact arithmetic with
numbers larger than 5 but are able to compare and add large approximate numbers that
are far beyond their naming range. Particularly, they use the words ‘some, not many’
and ‘many, really many’ to distinguish two types of large numbers. Their arithmetic with
‘some, not many’ and ‘many, really many’ reminds one strongly of the rules Cantor uses
to work with ℵ0 and c, respectively. In fact, it is sufficient to compare
‘some, not many’+ ‘many, really many’ = ‘many, really many’ (2.5)
with
ℵ0 + c = c (2.6)
to see this similarity.
Let us compare now the weak numeral systems involved in (2.1), (2.5) and numeral
systems used to work with infinity. We have already seen that relations (2.1) are results of
the weakness of the numeral system employed. Moreover, the usage of a stronger numeral
system shows that it is possible to pass from records 1+2 = ‘many’ and 2+2 = ‘many’ pro-
viding for two different expressions the same result, i.e., ‘many’, to more precise answers
1+2 = 3 and 2+2 = 4 and to see that 3 ̸= 4. In these examples we have the same objects –
small finite numbers – but results of computations we execute are different in dependence
10 Yaroslav D. Sergeyev
of the instrument – numeral system – used to represent numbers. Substitution of the nu-
meral ‘many’ by a variety of numerals representing numbers 3, 4, etc. allows us both to
avoid relations of the type (2.1), (2.5) and to increase the accuracy of computations.
Relations (2.2)–(2.4), (2.6) manifest a complete analogy with (2.1), (2.5). Canoni-
cally, symbols ∞, ℵ0 , and c are identified with concrete mathematical objects and (2.2)–
(2.4), (2.6) are considered as intrinsic properties of these infinite objects (see e.g., [16,
22, 25, 53, 64, 82, 113]). However, the analogy with (2.1), (2.5) shows that relations
(2.2)–(2.4), (2.6) do not reflect the nature of infinite objects. They are just a result of
weak numeral systems used to express infinite quantities. As (2.1), (2.5) show the lack of
numerals in numeral systems of Pirahã, Warlpiri, Pitjantjatjara, and Mundurukú for ex-
pressing different finite quantities, relations (2.2)–(2.4), (2.6) show shortage of numerals
in mathematical analysis and in Cantor’s theories for expressing different infinite num-
bers. Another hint leading to the same conclusion is the situation with indeterminate
forms of the kind III-V in Roman numerals that have been excluded from the practice of
computations after introducing positional numeral systems.
Thus, the analysis made above allows us to formulate the following key observation
that changes our perception of infinity:
The way of reasoning where the object of the study is separated from the tool used
by the investigator is very common in natural sciences where researchers use tools to
describe the object of their study and the used instrument influences the results of the
observations and determine their accuracy. The same happens in Mathematics studying
natural phenomena, numbers, objects that can be constructed by using numbers, sets, etc.
Numeral systems used to express numbers are among the instruments of observation used
by mathematicians. As we have illustrated above, the usage of powerful numeral systems
gives the possibility to obtain more precise results in Mathematics in the same way as
usage of a good microscope gives the possibility of obtaining more precise results in
Physics. Traditional numeral systems have been developed to express finite quantities and
they simply have no sufficiently high number of numerals to express different infinities
(and infinitesimals).
In order to avoid situations of the type (2.2)–(2.4), (2.6), a numeral system allowing
one to express a variety of different infinities and infinitesimals has been introduced re-
cently in [86, 88, 94, 99, 101]. The respective computational methodology allows one
to execute numerical computations with infinities and infinitesimals on the Infinity Com-
puter (USA patent 7,860,914, EU patent 1728149). This numeral system uses the same
numerals for different purposes for dealing with infinities and infinitesimals as it happens
with finite numerals that can be used in all the occasions where we need to work with
finite quantities.
Numerical infinities and infinitesimals 11
to be introduced, this abstraction is not used and the following postulate is adopted.
Methodological Postulate 1. We postulate existence of infinite and infinitesimal objects
but accept that human beings and machines are able to execute only a finite number of
operations.
Thus, we accept that we shall never be able to give a complete (in any sense) de-
scription of infinite objects due to our finite capabilities. Particularly, this means that we
accept that we are able to write down only a finite number of symbols to express numbers.
However, we do not agre with finitists who deny infinite mathematical objects. We accept
their existence and shall try to study them using our finite capabilities.
The second postulate is adopted following the way of reasoning used in natural sci-
ences where researchers use tools to describe the object of their study and the used in-
strument influences the results of the observations. When a physicist uses a weak lens
A and sees two black dots in his/her microscope he/she does not say: The object of the
observation is two black dots. The physicist is obliged to say: the lens used in the micro-
scope allows us to see two black dots and it is not possible to say anything more about
the nature of the object of the observation until we replace the instrument - the lens or
the microscope itself - with a more precise one. Suppose that he/she changes the lens and
uses a stronger lens B and is able to observe that the object of the observation is viewed
as ten (smaller) black dots. Thus, we have two different answers: (i) the object is viewed
as two dots if the lens A is used; (ii) the object is viewed as ten dots by applying the lens
B. Which of the answers is correct? Both. Both answers are correct but with the different
accuracies that depend on the lens used for the observation. The answers are not in oppo-
sition one to another, they both describe the reality (or whatever is behind the microscope)
correctly with the precision of the used lens. In both cases our physicist discusses what
he/she observes and does not pretend to say what the object is.
The same happens in Mathematics studying natural phenomena, numbers, and objects
that can be constructed by using numbers. Numeral systems used to express numbers are
among the instruments of observations used by mathematicians. The usage of powerful
numeral systems gives the possibility to obtain more precise results in Mathematics in the
same way as usage of a good microscope gives the possibility of obtaining more precise
results in Physics. However, even for the best existing tool the capabilities of this tool
will be always limited due to Postulate 1 (we are able to write down only a finite number
of symbols when we wish to describe a mathematical object) and due to Postulate 2 we
shall never tell, what is, for example, a number but shall just observe it through numerals
expressible in a chosen numeral system.
Methodological Postulate 2. We shall not tell what are the mathematical objects we
deal with; we just shall construct more powerful tools that will allow us to improve our
capacities to observe and to describe properties of mathematical objects.
This Postulate is important for our study because it emphasizes that the triad – object
of a study, instrument of a the study, and a researcher executing the study – introduced in
Physics in XX th century exists in Mathematics, as well. With this Postulate we stress that
mathematical results are not absolute, they depend on mathematical languages used to
formulate them, i.e., there always exists an accuracy of the description of a mathematical
result, fact, object, etc. imposed by the mathematical language used to formulate this
Numerical infinities and infinitesimals 13
result. For instance, the result of Pirahã 2 + 2 = ‘many’ is not wrong, it is just inaccurate.
The introduction of a stronger tool (in this case, a numeral system that contains a numeral
for a representation of the number four) allows us to have a more precise answer.
The concept of the accuracy allows us to look at paradoxes in a new way: paradox
is a situation where the accuracy of the used language is not sufficient to describe the
phenomenon we are interested in. For instance, the answers of Pirahã 2 + 1 = ‘many’
and 2 + 2 = ‘many’ can be viewed as a paradox because from these two records one
could conclude that 2 + 1 = 2 + 2. This paradox shows us the borderline that separates
the zone where the language has the high precision from the region where the language
cannot be applied because it does not allow one to distinguish different objects within
‘many’. Analogously, the records ‘many’ + 1= ‘many’, ∞ + 1 = ∞, 1 + ω = ω ̸= ω + 1,
etc. (see (2.1)–(2.6)) can also be viewed as situations where the accuracy of the used
numeral systems is not sufficient.
It is necessary to comment upon another important aspect of the distinction between
a mathematical object and a mathematical tool used to observe this object. Postulates 1
and 2 impose us to think always about the possibility to execute a mathematical operation
by applying a numeral system. They tell us that there always exist situations where we are
not able to express the result of an operation. Let us consider, for example, the operation
of construction of the successor familiar from number and set theories. In the traditional
Mathematics, the question whether this operation can be executed or not is not taken
into consideration, it is supposed that it is always possible starting from any integer n to
execute the operation n + 1 and to assign the obtained result to k getting k = n + 1. Thus,
traditionally the problem of the distinction between the existence of the number k and the
possibility to execute the operation n + 1 and to express its result (i.e., to have a numeral
that can express k) is not debated.
Postulates 1 and 2 emphasize this distinction and tell us that: (i) in order to execute the
operation it is necessary to have a numeral system allowing one to express both numbers, n
and k; (ii) for any numeral system there always exists a number k∗ that cannot be expressed
in it. For instance, for Pirahã k∗ = 3, for Mundurukú k∗ = 6. Even for modern powerful
numeral systems there exist such a number k∗ , e.g., as was already mentioned, nobody is
able to write down a numeral in the decimal positional system having 10100 digits4 . This
means that theoretical considerations including the necessity to execute the operation n+1
should provide both a practical way to do this and the required numerals, otherwise an
ambiguity arises since it is supposed that this operation can be done but it actually cannot
be executed.
Another important issue related to Postulate 2 consists of its applications to axiomatic
systems. Traditionally, it is supposed that they define some mathematical objects. Due to
Postulate 2, they do not define mathematical objects but describe them. They determine
formal rules for operating with certain numerals reflecting some (not all) properties of the
studied mathematical objects using a certain mathematical language L. We are aware that
the chosen language L has its accuracy and there always can exist a richer language L̃
4 Suppose that one is able to write down one digit in one nanosecond. Then, it will take 1091 seconds to
record all 10100 digits. Since in one year there are 31 556 926 ≈ 3.2 · 107 seconds, 1091 seconds are approx-
imately 3.2 · 1083 years. This is a sufficiently long time since it is supposed that the age of the universe is
approximately 1.382 · 1010 years.
14 Yaroslav D. Sergeyev
that would allow us to describe the studied object better. As has already been discussed
above, any language has a limited expressibility, in particular, there always exist situa-
tions where the accuracy of the answers expressible in this language is not sufficient. As
a consequence, axiom systems can say about an object they describe only those things that
the used language allows. A richer language allows one to say more about the object and
a weaker language – less. Thus, development of the mathematical (and not only mathe-
matical) languages leads to a continuous necessity of a transcription and specification of
axiomatic systems. In fact, the history of Mathematics shows that important results are
under a continuous process of re-writing and improvement in order to increase the level
of their precision up to the current level of the development of Mathematics.
Numerals that we use to write down numbers, functions, etc. are among our tools of
investigation and, as a result, they strongly influence our capabilities to study mathemat-
ical objects. This separation (having an evident physical spirit) of mathematical objects
from the tools used for their description is crucial for our study but it is used rarely in
contemporary Mathematics. In fact, the idea of finding an adequate (absolutely the best)
set of axioms for one or another field of Mathematics continues to be among the most
attractive goals for many contemporary mathematicians whereas physicists perfectly un-
derstand the impossibility of any attempt to give a final description of physical objects due
to limitations of the instruments of study. In Philosophy and Linguistics, the relativity of
the language (the instrument) with respect to the world around us (the object of study) is
also a well known thing. In Linguistics, it is sufficient to remind the Sapir–Whorf thesis
(see [17, 84]), also known as the ‘linguistic relativity thesis’. As becomes clear from its
name, the thesis does not accept the idea of the universality of language and postulates
that the nature of a particular language influences the thought of its speakers. The thesis
challenges the possibility of perfectly representing the world with language, because it
implies that the mechanisms of any language condition the thoughts of its speakers.
Thus, due to Postulate 2, our point of view on axiomatic systems is significantly more
applied with respect to the modern mathematical fashion that tries to make all axiomatic
systems more and more precise (decreasing so degrees of freedom of the studied part of
Mathematics). We just define a set of general rules describing how practical computations
should be executed leaving as much space as possible for further changes and develop-
ments dictated by practice of the introduced mathematical language.
For example, from this applied point of view, axioms for real numbers do not define
numbers, they describe them. The axioms should be considered together with a particular
numeral system S used to write down numerals and are viewed as practical rules (asso-
ciative and commutative properties of multiplication and addition, distributive property
of multiplication over addition, etc.) describing operations with the numerals. The com-
√ property is interpreted as a possibility to extend S with additional symbols (e.g.,
pleteness
e, π, 2, etc.) taking care of the fact that the results of computations with these symbols
agree with the facts observed in practice. As a rule, assertions regarding numbers that
cannot be expressed in a numeral system are avoided5 . Assertions are made with respect
5 For instance, it is not supposed that real numbers form a field. In fact, in this case real numbers are our
objects of study and before saying something we should fix a tool of their observation. To follow Postulate 2 we
can discuss sets of real numbers expressible (visible) in a certain fixed numeral system that is our ‘lens’ in this
case.
Numerical infinities and infinitesimals 15
to what is visible at a mathematical ‘lens’ and not about the object located behind the
lens. This is done again by following Physics where observations are done using differ-
ent instruments and without specifying the instruments assertions have no meaning. For
instance, the question: ‘What do you see in this direction?’ is meaningless without indi-
cation a tool used for the observation. In fact, by eye the observer will see certain things,
by microscope other things, by telescope again other things, etc.
Instruments (numeral systems in this case) not only bound our practical capabilities to
compute, they can influence theoretical results in Mathematics, as well. We illustrate this
statement by considering the following simple phrase ‘Let us consider all x ∈ [1, 2]’. For
Pirahã, Warlpiri, Pitjantjatjara, and Mundurukú all numbers are just 1 and 2. For people
who do not know irrational numbers (or do not accept their existence) all numbers are
fractions qp where p and q can be expressed in a numeral system they know. If both p
and q can assume values 1 and 2 (as it happens for Pirahã), all numbers in this case are:
1, 1 + 12 , and 2. For persons knowing positional numeral systems all numbers are those
numbers that can be written in a positional system6 . Thus, in different historical periods
(or in different cultures) the phrase ‘Let us consider all x ∈ [1, 2]’ has different meanings.
As a result, without fixing the numeral system we use to express numbers we cannot fix
the numbers we deal with and an ambiguity holds.
Finally, before we switch our attention to Postulate 3, it should be noticed the key
difference distinguishing our approach from constructivism. Constructivists assert that
it is necessary to construct (in some sense) a mathematical object to prove that it exists.
Following the modern Physics, we do not discuss the questions of existence of mathemat-
ical objects at all. We discuss just what can be observed through our tools (languages,
numeral systems, etc.) and try to improve our tools of the observation.
Let us now introduce the last Postulate. We have already seen in the previous section
that situations of the kind (2.2)–(2.4), (2.6) are results of the weakness of numeral systems
used to express infinite quantities. Thus, we should treat infinite and infinitesimal numbers
in the same manner as we are used to deal with finite ones, i.e., by applying the Euclid’s
Common Notion no. 5 ‘The whole is greater than the part’. In our considerations, its
straight reformulation ‘The part is less than the whole’ will be used for practical reasons
that will become clear soon. This principle, in our opinion, very well reflects organization
of the world around us but is not incorporated in many traditional infinity theories where it
is true only for finite numbers. The reason of this traditional discrepancy (as the analysis
made in the previous section suggests) is related to the low accuracy of numeral systems
used to work with infinity.
Methodological Postulate 3. We adopt the principle ‘The part is less than the whole’
to all numbers (finite, infinite, and infinitesimal) and to all sets and processes (finite and
infinite).
Due to this Postulate, the traditional point of view on infinity accepting such results
as ∞ − 1 = ∞ should be substituted in a way that ∞ − 1 < ∞. One of the motivations
in favour of this substitution has already been discussed in detail in connection with the
6 Notice that Cantor’s cardinals do not allow us to distinguish the quantity of numbers written in the binary
and in the decimal systems providing the same answer: both sets have the cardinality of continuum. This
distinction for the new methodological positions will be discussed later.
16 Yaroslav D. Sergeyev
3.2. Counting huge finite quantities in practice – a granary example. Let us intro-
duce now a way of counting arising in practice where there is the necessity to operate
with extremely large finite quantities. This way of computing has existed for millennia
but strangely enough has not been studied in depth in Mathematics so far. This example
will help us to develop a new intuition required to compute in accordance with Postulates
1 – 3 and will be adopted hereinafter to work with infinite quantities.
Imagine that we are in a granary and the owner asks us to count how much grain he
has inside it. Obviously, it is possible to answer that there are many seeds in the granary.
This answer is correct but its accuracy is low. In order to obtain a more precise answer it
would be necessary to count the grain seed by seed but since the granary is huge, it is not
possible to do this due to practical reasons.
To overcome this difficulty and to obtain a more precise answer than ‘many’, people
take sacks, fill them with seeds, and count the number of sacks. In this situation, we
suppose that: (i) all the seeds have the same measure and all the sacks also; (ii) the
number of seeds in each sack is the same and is equal to K1 but the sack is so big that we
are not able to count how many seeds it contains and to establish the value of K1 ; (iii) in
any case the resulting number K1 would not be expressible by available numerals.
Then, if the granary is huge and it becomes difficult to count the sacks, then trucks
or even big train waggons are used. As it was for the sacks, we suppose that all trucks
contain the same number K2 of sacks, and all train waggons contain the same number K3
Numerical infinities and infinitesimals 17
7 Notice that nowadays not only positive integers but also zero is frequently included in N. However, since
historically zero has been invented significantly later than positive integers used for counting objects, zero is not
include in N in this article.
18 Yaroslav D. Sergeyev
can be considered as a sack and ¬ is the number of seeds in the sack. Following our
extrapolation, the introduction of ¬ allows us to write down the set of natural numbers as
N = {1, 2, 3, . . . , ¬ − 1, ¬}. Recall that the usage of a numeral indicating the totality of
the elements we deal with is not new in Mathematics. It is sufficient to mention the theory
of probability (axioms of Kolmogorov) where events can be defined in two ways. First,
as unions of elementary events; second, as a sample space, Ω, of all possible elementary
events (or its parts) from which some elementary events have been excluded (or added in
case of parts of Ω). Naturally, the latter way to define events becomes particularly useful
when the sample space consists of infinitely many elementary events.
Grossone is introduced by describing its properties postulated by the Infinite Unit
Axiom (IUA) consisting of three parts: Infinity, Identity, and Divisibility. Similarly, in
order to pass from natural to integer numbers a new element – zero – is introduced, a
numeral to express it is chosen, and its properties are described. The IUA is added to
axioms for real numbers (that are considered in sense of Postulate 2). Thus, it is postulated
that associative and commutative properties of multiplication and addition, distributive
property of multiplication over addition, existence of inverse elements with respect to
addition and multiplication hold for grossone as they do for finite numbers8 .
Let us introduce the axiom and then give some comments upon it. Notice that in the
IUA infinite sets will be described in the traditional form, i.e., without indicating the last
element. For instance, the set of natural numbers will be written as N = {1, 2, 3, . . .} in-
stead of N = {1, 2, 3, . . . , ¬ − 1, ¬}. We emphasize that in both cases we deal with the
same mathematical object – the set of natural numbers – that is observed through two dif-
ferent instruments. In the first case traditional numeral systems do not allow us to express
infinite numbers whereas the numeral system with grossone offers this possibility. Sim-
ilarly, Pirahã are not able to see finite natural numbers greater than 2 but these numbers
(e.g., 3 and 4) belong to N and are visible if one uses a more powerful numeral system.
A detailed discussion will follow shortly to show how to use grossone to calculate and
express the number of elements of certain infinite sets.
The Infinite Unit Axiom. The infinite unit of measure is introduced as the number
of elements of the set, N, of natural numbers. It is expressed by the numeral ¬ called
grossone and has the following properties:
Infinity. Any finite natural number n is less than grossone, i.e., n < ¬.
Identity. The following relations link ¬ to identity elements 0 and 1
¬
0 · ¬ = ¬ · 0 = 0, ¬ − ¬ = 0, = 1, ¬0 = 1, 1¬ = 1, 0¬ = 0. (4.1)
¬
Divisibility. For any finite natural number n sets Nk,n , 1 ≤ k ≤ n, being the nth parts of
the set, N, of natural numbers have the same number of elements indicated by the numeral
8 Remind that we speak about axioms of real numbers in sense of Postulate 2, i.e., axioms define formal
rules of operations with numerals in a given numeral system. Therefore, if we want to have a numeral system
including grossone, we should fix also a numeral system to express finite numbers. In order to concentrate our
attention on properties of grossone, this point will be investigated later.
Numerical infinities and infinitesimals 19
¬ where
n
∪
n
Nk,n = {k, k + n, k + 2n, k + 3n, . . .}, 1 ≤ k ≤ n, Nk,n = N. (4.2)
k=1
Let us comment upon this axiom. Its first part – Infinity – is quite clear. In fact, we
want to describe an infinite number, thus, it should be larger than any finite number. The
second part of the axiom – Identity – tells us that ¬ interacts with identity elements 0 and
1 as all other numbers do. In reality, we could even omit this part of the axiom because,
due to Postulate 3, all numbers should be treated in the same way and, therefore, at the
moment we have stated that grossone is a number, we have fixed the usual properties of
numbers, i.e., the properties described in Identity, associative and commutative properties
of multiplication and addition, distributive property of multiplication over addition, etc.
The third part of the axiom – Divisibility – is the most interesting, since it links infinite
numbers to infinite sets (in many traditional theories infinite numbers are introduced al-
gebraically, without any connection to infinite sets) and is based on Postulate 3. Let us
first illustrate it by an example.
Example 4.1. If we take n = 1, then it follows that N1,1 = N and Divisibility says that
the set, N, of natural numbers has ¬ elements. If n = 2, we have two sets N1,2 and N2,2 ,
where
N1,2 = {1, 3, 5, 7, . . . },
(4.3)
N2,2 = { 2, 4, 6, ... }
and they have ¬ 2 elements each. Notice that the sets N1,2 and N2,2 have the same number
of elements not because they are in a one-to-one correspondence but by the Divisibility
axiom. In fact, we are not able to count the number of elements of the sets N, N1,2 , and
N2,2 one by one because due to Postulate 1 we are able to execute only a finite number of
operations and all these sets are infinite. To define their number of elements we use Di-
visibility and implement Postulate 3 in practice by determine the number of the elements
of the parts using the whole.
Then, if n = 3, we have three sets
N1,3 = {1, 4, 7, . . . },
N2,3 = { 2, 5, . . . }, (4.4)
N3,3 = { 3, 6, ... }
and they have ¬3 elements each. Note that in formulae (4.3), (4.4) we have added extra
spaces writing down the elements of the sets N1,2 , N2,2 , N1,3 , N2,3 , N3,3 just to emphasize
Postulate 3 and to show visually that N1,2 ∪ N2,2 = N and N1,3 ∪ N2,3 ∪ N3,3 = N.
In general, to introduce ¬ n we do not try to count elements k, k + n, k + 2n, k + 3n, . . .
one by one in (4.2). In fact, we cannot do this due to Postulate 1. By using Postulate 3,
20 Yaroslav D. Sergeyev
we construct the sets Nk,n , 1 ≤ k ≤ n, by separating the whole, i.e., the set N, in n parts
and we affirm that the number of elements of the nth part of the set, i.e., ¬ , is n times n
less than the number of elements of the entire set, i.e., than ¬.
As was already mentioned, in terms of our granary example ¬ can be interpreted as
the number of seeds in the sack. In that example, the number K1 of seeds in each sack was
fixed and finite but it was impossible to express it in units u0 , i.e., seeds, by counting seed
by seed because we had supposed that sacks were very big and the corresponding number
would not be expressible by available numerals. In spite of the fact that K1 , K2 , and K3
were inexpressible and unknown, by using new units of measure (sacks, trucks, etc.) it
was possible to count more easily and to express the required quantities. Now our sack
has the infinite but again fixed number of seeds. It is fixed because it has a strong link to a
concrete set – it is the number of elements of the set of natural numbers. Since this number
is inexpressible by existing numeral systems with the same accuracy afforded to measure
finite small sets9 , we introduce a new numeral, ¬, to express the required quantity. Then,
we apply Postulate 3 and say that if the sack contains ¬ seeds, then, even though we are
not able to count the number of seeds of the nth part of the sack seed by seed, its nth part
contains n times less seeds than the entire sack, i.e., ¬n seeds. Notice that the numbers n
¬
are integer since they have been introduced as numbers of elements of sets Nk,n .
The new unit of measure allows us to express a variety of infinite numbers (including
those larger than ¬ that will be considered later) and calculate easily the number of ele-
ments of the union, intersection, difference, or product of sets of type Nk,n . Due to our
accepted methodology, we do it in the same way as these measurements are executed for
finite sets. Let us consider two simple examples showing how grossone can be used for
this purpose (a general rule for determining the number of elements of infinite sets having
a more complex structure will be given in Section 7).
Example 4.2. Let us determine the number of elements of the set Ak,n = Nk,n \{a},
a ∈ Nk,n , n ≥ 1. Due to the IUA, the set Nk,n has ¬
n elements. The set Ak,n has been
constructed by excluding one element from N . Thus, the set A has ¬ − 1 elements.
k,n k,n n
The granary interpretation can be also given for the number ¬
n − 1: the number of seeds
in the nth part of the sack minus one seed. For n = 1 we have ¬ − 1 interpreted as the
number of seeds in the sack minus one seed.
Divisibility and Example 4.2 show us that in addition to the usual way of counting,
i.e., by adding units, that has been formalized in the traditional Mathematics, there exists
also the way to count by taking parts of the whole and by subtracting units or parts of
the whole. The following example shows a slightly more complex situation (other more
9 First, this quantity is inexpressible by numerals used to count the number of elements of finite sets because
the set N is infinite. Second, traditional numerals existing to express infinite numbers do not have the required
high accuracy (recall that we would like to be able to register the variation of the number of elements of infinite
sets even when one element has been excluded or added). For example, by using Cantor’s alephs we say that
cardinality of the sets N and N \ {1} is the same – ℵ0 . This answer is correct but its accuracy is low – we are not
able to register the fact that one element was excluded from the set N even though we ourselves have executed
this exclusion. Analogously, we can say that both the sets have many elements. Again, the answer ‘many’ is
correct but its accuracy is low.
Numerical infinities and infinitesimals 21
sophisticated examples will be given later after the reader gets accustomed to the concept
of ¬).
Example 4.3. Let us consider the following two sets
B1 = {4, 9, 14, 19, 24, 29, 34, 39, 44, 49, 54, 59, 64, 69, 74, 79, . . .},
B2 = {3, 14, 25, 36, 47, 58, 69, 80, 91, 102, 113, 124, 135, . . .}
and determine the number of elements in the set B = (B1 ∩ B2 ) ∪ {3, 4, 5, 69}. It follows
immediately from the IUA that B1 = N4,5 and B2 = N3,11 . Their intersection
A = {1, 2, 3, 4, 5} (4.5)
is the largest element in this set. As was already mentioned, the numeral ¬ allows one to
write down the set, N, of natural numbers in the form
¬ ¬ ¬ ¬ ¬
N = {1, 2, . . . − 2, − 1, , + 1, + 2, . . . ¬ − 2, ¬ − 1, ¬}. (4.6)
2 2 2 2 2
and to observe infinite natural numbers
¬ ¬ ¬ ¬ ¬
... − 2, − 1, , + 1, + 2, . . . ¬ − 2, ¬ − 1, ¬ (4.7)
2 2 2 2 2
that are invisible if traditional numeral systems are used to observe N. Similarly, Pirahã
are not able to see finite natural numbers larger than 2.
This example illustrates also the fact that when we speak about sets (finite or infinite)
it is necessary to take care about tools used to describe a set (recall Postulate 2). In order
to introduce a set, it is necessary to have a language (e.g., a numeral system) allowing us
to describe both its elements and the number of the elements in the set. For instance, the
set A from (4.5) cannot be defined using the mathematical language of Pirahã.
Analogously, the words ‘the set of all finite numbers’ do not define a set completely
from our point of view. It is always necessary to specify which instruments are used
to describe (and to observe) the required set and, as a consequence, to speak about ‘the
set of all finite numbers expressible in a fixed numeral system’. For instance, for Pirahã
‘the set of all finite numbers’ is the set {1, 2} and for Mundurukú ‘the set of all finite
numbers’ is the set A from (4.5). As it happens in Physics, the instrument used for an
22 Yaroslav D. Sergeyev
observation bounds the possibility of the observation. It is not possible to say how we
shall see the object of our observation if we have not clarified which instruments will be
used to execute the observation (see discussion on page 15). Thus, in the moment one
chooses the instrument of the observation this person chooses also objects that can be
observed.
In order to be more acquainted with this methodology separating objects and tools of
their observation let us try to answer to the following question: Which natural numbers
can we express by using the numeral ¬? Analogously to the examples with the phrases
‘the set of all finite numbers’ and the question ‘What do you see in this direction?’ this
question is not sufficiently precise since it does not specify which numeral system is
used to express finite quantities. Suppose that we have fixed a numeral system, S , for
expressing finite natural numbers and it allows us to express KS numbers (not necessarily
consecutive) belonging to a set NS ⊂ N. Note that due to Postulate 1, KS is finite. Then,
addition of ¬ to this numeral system will allow us to express also infinite natural numbers
i¬ ¬ i¬
n ± k ≤ ¬ where 1 ≤ i ≤ n, k ∈ NS , n ∈ NS (notice that since n are integers, n
are integers too). Thus, the more powerful system S is used to express finite numbers,
the more infinite numbers can be expressed but their quantity is always finite, again due
to Postulate 1. The ¬-based numeral system allows us to express more numbers than
traditional numeral systems thanks to the introduced new numerals, however, as it happens
for all numeral systems, its abilities to express numbers are also limited.
Example 4.4. Let us consider the numeral system, P , of Pirahã able to express only
numbers 1 and 2 (the only difference will be in the usage of numerals ‘1’ and ‘2’ instead
of original numerals I and II used by Pirahã). If we add to P the new numeral ¬, we
obtain a new numeral system (we call it P b ) allowing us to express only ten numbers
represented by the following numerals
¬ ¬ ¬ ¬ ¬
1, 2 , ... − 2, − 1, , + 1, + 2, ... ¬ − 2, ¬ − 1, ¬ . (4.8)
|{z} |2 2 2 2
{z 2 } | {z }
finite infinite
infinite
The first two numbers in (4.8) are finite, the remaining eight are infinite. Traditionally
dots in a record of the type 1,2,3, ... show that we do not want but can proceed with
writing. Dots in (4.8) have a different meaning: they represent natural numbers that are
not expressible in the numeral system P b.
b
As a consequence, P does not allow us to execute such operation as 2 + 2 or to add
2 to ¬ b
2 + 2 because their results cannot be expressed in P . Of course, we do not say
that results of these operations are equal to ‘many’ (as Pirahã do for operations 2 + 2
and 2 + 1). We just say that the results are not expressible in P b and it is necessary to
take another, more powerful numeral system if we want to execute these operations. This
represents an important shift in the methodology applied in situations where one faces a
difficulty in counting. Instead of concluding that the problem is hard one says that the
used instrument is inadequate and, thefore, should be substituted with a more powerful
tool.
Note that crucial limitations discussed in Example 4.4 hold for sets, too. In fact, the
numeral system P allows us to define only the sets N1,2 and N2,2 among all possible sets of
Numerical infinities and infinitesimals 23
the form Nk,n from (4.2) because we have only two finite numerals, ‘1’ and ‘2’, in P . This
numeral system is too weak to define other sets of this type, for instance, N4,5 , because
numbers greater than 2 required for these definition are not expressible in P . Moreover
in each of the sets N1,2 and N2,2 the system P allows us to observe only one element
(1 in N1,2 and 2 in N2,2 ). These limitations have a general character and are related to
all questions requiring a numerical answer (i.e., an answer expressed only in numerals,
without variables). In order to obtain such an answer, it is necessary to know at least one
numeral system able to express numerals required to write down this answer.
We are now ready to state the following result being a direct consequence of the
accepted methodological postulates.
Theorem 4.5. The set N is not a monoid under addition.
Proof. Due to Postulate 3, the operation ¬ + 1 gives us as the result a number greater
than ¬. Thus, by definition of grossone, ¬ + 1 does not belong to N and, therefore, N is
not closed under addition and is not a monoid.
This result also means that adding the IUA to the axioms of natural numbers allows
b and including N as a
us to talk about the set of extended natural numbers indicated as N
proper subset
b = {1, 2, . . . , ¬ − 1, ¬, ¬ + 1, ¬ + 2, . . . , 2¬ − 1, 2¬, 2¬ + 1, . . .
N
| {z }
Natural numbers
C¬ = {(a1 , a2 , . . . , a¬−1 , a¬ ) : ai ∈ N, 1 ≤ i ≤ ¬}
having ¬¬ elements.
Note that we can also give the granary interpretation for the numbers of the type ¬m :
if we accept that the numbers Ki from page 17 are such that Ki = ¬, 1 ≤ i ≤ m − 1, then
¬2 can be viewed as the number of seeds in the truck, ¬3 as the number of seeds in the
train waggon, etc.
Example 4.7. Another important set having more than grossone elements is the set, Z, of
integers. Clearly, it has 2¬+1 elements (¬ positive numbers, ¬ negative, and zero).
b of extended integer numbers can be constructed from the set, Z, of inte-
The set, Z,
gers by a complete analogy and inverse elements with respect to addition are introduced
naturally. For example, 7¬ has its inverse with respect to addition equal to −7¬.
Let us emphasize that, due to Postulates 1 and 2, the new system of counting cannot
give answers to all questions regarding infinite sets. What can we say, for instance, about
the number of elements of the sets N b and Z?b Have they been described completely? The
introduced numeral system based on ¬ is too weak to give answers to these questions
since it does not allow us to express the number of elements of these sets. It is necessary
to introduce in a way a more powerful numeral system by defining new numerals (for
instance, , ®, etc).
We conclude this subsection by the following remark. The IUA describes a new num-
ber – the quantity of elements in the set of natural numbers – expressed by the numeral ¬.
However, other numerals and sets can be used to express the idea of the axiom. For ex-
ample, the numeral ¶ can be introduced as the number of elements of the set, E, of even
numbers and can be taken as the base of a numeral system. In this case, the IUA can
be reformulated using the numeral ¶ and numerals including it will be used to express
infinite numbers. For example, the numbers of elements |O| of the set, O, of odd numbers
and numbers of elements of the sets N and Z will be expressed as
We emphasize through this remark that infinite numbers (similarly to the finite ones) can
be expressed by various numerals and in different numeral systems.
The traditional conclusion from (4.10) is that both sets are countable and they have the
same cardinality ℵ0 .
Let us see now what we can say from the new methodological position, in particu-
lar, by using Postulate 2. The separation of the objects of study which are two infinite
sets from the instrument used to compare them, i.e., from the bijection, suggests that an-
other conclusion can be derived from (4.10): the accuracy of the used instrument is not
sufficiently high to see the difference between the sizes of the two sets.
We have already seen that when one executes the operation of counting, the accuracy
of the result depends on the numeral system used for counting. If one asked Pirahã to
measure sets consisting of four apples and five apples the answer would be that both sets
of apples have many elements. This answer is correct but its precision is low due to the
weakness of the numeral system used to measure the sets.
Thus, the introduction of the notion of accuracy for measuring sets is very important
and should be applied to infinite sets also. As was already discussed, since for cardinal
numbers it follows
ℵ0 + 1 = ℵ0 , ℵ0 + 2 = ℵ0 , ℵ0 + ℵ0 = ℵ0 ,
these relations suggest that the accuracy of the cardinal numeral system of Alephs is not
sufficiently high to see the difference with respect to the number of elements of the two
sets from (4.10).
In order to look at the record (4.10) using the new methodology let us remind that due
to the IUA the sets of even and odd numbers have ¬/2 elements each and, therefore, ¬
is even. It is also necessary to recall that numbers that are larger than ¬ are not natural,
they are extended natural numbers. For instance, ¬ + 1 is odd but not natural, it is the
extended natural, see (4.9). Thus, the last odd natural number is ¬-1. Since the number of
elements of the set of odd numbers is equal to ¬ 2 , we can write down not only the initial
(as it is usually done traditionally) but also the final part of (4.10)
1, 3, 5, 7, 9, 11, . . . ¬ − 5, ¬ − 3, ¬ − 1
↕ ↕ ↕ ↕ ↕ ↕ ↕ ↕ ↕ (4.11)
1, 2, 3, 4 5, 6, . . . ¬ − 2, ¬ − 1, ¬
2 2 2
concluding so (4.10) in a complete accordance with the principle ‘The part is less than the
whole’. Both records, (4.10) and (4.11), are correct but (4.11) is more accurate, since it
allows us to observe the final part of the correspondence that is invisible if (4.10) is used.
The accuracy of the ¬-based numeral system allows us to measure also, for instance,
such sets as O′ = O\{3} and O′′ = O\{1, ¬ − 1}. The set O′ is constructed by excluding
one element from O and the set O′′ by excluding from O two elements. Thus, O′ and
O′′ have ¬ ¬
2 − 1 and 2 − 2 elements, respectively. In case one wishes to establish the
corresponding bijections, starting with natural numbers 1, 2, 3, . . . we obtain for these two
sets
1, 5, 7, 9, 11, 13, . . . ¬ − 5, ¬ − 3, ¬ − 1
↕ ↕ ↕ ↕ ↕ ↕ ↕ ↕ ↕ (4.12)
1, 2, 3, 4 5, 6, . . . ¬ 2 − 3, ¬ − 2, ¬ − 1
2 2
26 Yaroslav D. Sergeyev
E2 = {2, 4, 6, . . . ¬ − 2, ¬, ¬ + 2, . . . 2¬ − 4, 2¬ − 2, 2¬}
2, 4, 6, . . . ¬ − 2, ¬, ¬ + 2, . . . 2¬ − 4, 2¬ − 2, 2¬
↕ ↕ ↕ ↕ ↕ ↕ ↕ ↕ ↕
1, 2, 3, . . . ¬ − 1, ¬, ¬ + 1, . . . ¬ − 2, ¬ − 1, ¬
2 2 2
room 1 to room 2, the guest occupying room 2 to room 3, etc. At the end of this procedure
the guest from the last room number ¬ should be moved to the room ¬+1. However, the
Hotel has only ¬ rooms and, therefore, the poor guy from the room ¬ will go out of
the Hotel (the situation taking place in hotels with a finite number of rooms if such a
procedure is implemented).
Notice, that there is no contradiction between the two ways to see the situation. The
traditional answer is that it is possible to put the newcomer in the first room. The ¬-way
of doing confirms this result but shows a thing that was invisible traditionally - the guest
from the last room should go out of the Hotel.
4.3. Positional numeral system with the infinite radix ¬. We have already started
to write down simple infinite numbers and to execute arithmetical operations with them
without concentrating our attention upon this question. Let us consider it systematically.
Different numeral systems have been developed to describe finite numbers. In posi-
tional numeral systems, fractional numbers are expressed by the record
where numerals ai , −q ≤ i ≤ n, are called digits, belong to the alphabet {0, 1, . . . , b − 1},
and the dot is used to separate the fractional part from the integer one. Thus, the numeral
(4.14) expresses the quantity obtained by summing up
Record (4.14) uses numerals consisting of one symbol each, i.e., digits ai ∈ {0, 1, . . . , b −
1}, to express how many finite units of the type bi belong to the number (4.15). Quantities
of finite units bi are counted separately for each exponent i and all symbols in the alphabet
{0, 1, . . . , b − 1} express finite numbers.
To express infinite and infinitesimal numbers we shall use records that are similar to
(4.14) and (4.15) but have some peculiarities. In order to construct a number C in the
numeral positional system with the base ¬, we subdivide C into groups corresponding to
powers of ¬:
represents the number C, where all numerals ci are called grossdigits. They are not equal
to zero and belong to a traditional numeral system. Grossdigits express finite positive or
negative numbers and show how many corresponding units ¬ pi should be added or sub-
tracted in order to form the number C. Grossdigits can be expressed by several symbols
using positional systems, can be written in the form Qq where Q and q are integers, or in
any other numeral system used to express finite quantities.
28 Yaroslav D. Sergeyev
Numbers pi in (4.17) called grosspowers can be finite, infinite, and infinitesimal (the
introduction of infinitesimal numbers will be given soon), they are sorted in the decreasing
order
pm > pm−1 > . . . > p1 > p0 > p−1 > . . . p−(k−1) > p−k
with p0 = 0.
In the traditional record (4.14), there exists a convention that a digit ai shows how
many powers bi are present in the number and the radix b is not written explicitly. In the
record (4.17), we write ¬ pi explicitly because in the ¬-based positional system the num-
ber i in general is not equal to the grosspower pi . This gives the possibility to write, for
example, such a number as 7.6¬244.5 34¬32 having grosspowers p2 = 244.5, p1 = 32 and
grossdigits c244.5 = 7.6, c32 = 34 without indicating grossdigits equal to zero correspond-
ing to grosspowers smaller than 244.5 and greater than 32. Note also that if a grossdigit
c pi = 1 then we often write ¬ pi instead of 1¬ pi .
The term having p0 = 0 represents the finite part of C because, due to (4.1), we have
c0 ¬0 = c0 . The terms having finite positive grosspowers represent the simplest infinite
parts of C. Analogously, terms having negative finite grosspowers represent the simplest
infinitesimal parts of C. For instance, the number ¬−1 = 1 is infinitesimal. It is the
¬
inverse element with respect to multiplication for ¬:
Note that all infinitesimals are different from zero. Particularly, 1 > 0 because it is
¬
a result of division of two positive numbers. It also has a clear granary interpretation.
Namely, if we have a sack containing ¬ seeds, then one sack divided by the number of
seeds in it is equal to one seed. Vice versa, one seed, i.e., 1 , multiplied by the number of
¬
seeds in the sack, ¬, gives one sack of seeds.
All of the numbers introduced above can be grosspowers, as well, giving so a pos-
sibility to have various combinations of quantities and to construct terms having a more
complex structure10 .
Example 4.9. The left-hand expression below shows how to write down numbers in the
new numeral system and the right-hand shows how the value of the number is calculated:
The number above has one infinite part having an infinite grosspower, one infinite part
having a finite grosspower, a finite part, and an infinitesimal part.
Finally, numbers having a finite and infinitesimal parts can be also expressed in the
new numeral system, for instance, the number −13.5¬0 -37¬−2 11¬−16¬+4.3 has a finite
and two infinitesimal parts, the second of them has the infinite negative grosspower equal
to −16¬ + 4.3. In case a finite number has no infinitesimal parts it is called purely finite.
and the result C = A + B is constructed by including in it all items aki ¬ki from A such that
ki ̸= m j , 1 ≤ j ≤ M, and all items bm j ¬m j from B such that m j ̸= ki , 1 ≤ i ≤ K. If in A and
B there are items such that ki = m j , for some i and j, then this grosspower ki is included
in C with the grossdigit bki + aki , i.e., as (bki + aki )¬ki .
Example 4.10. We consider two infinite numbers A and B, where
with the reminder Re3 = 0.28¬−6 . Naturally, it follows Ce = B · Ae3 + Re3 . The process
e
continues until a partial reminder Ri = 0 is found or when a required accuracy of the
result will be reached.
32 Yaroslav D. Sergeyev
5.1. Higher order numerical differentiation on the Infinity Computer. In many prac-
tical applications it is necessary to calculate derivatives of a function g(x) which is given
by a computer procedure calculating its approximation f (x). Since procedures for eval-
uating the exact values of f ′ (x) and higher derivatives are usually not available, either
some numerical approximations are used for this purpose or automatic differentiation
techniques are applied (see, e.g., [7, 9, 21, 26] and references given therein for a detailed
discussion). Since the latter approaches usually allow one to evaluate f ′ (x) only, they are
often symbolic and/or use specially developed procedures that re-write the code of f (x)
in order to obtain the code for f ′ (x), they are not considered in this text dedicated to
numerical computations.
The simplest formulae used on traditional computers to approximate f ′ (x) require the
evaluation of f (x) at two points and use forward and backward differences
f (x + h) − f (x) f (x) − f (x − h)
f ′ (x) ≈ , f ′ (x) ≈ . (5.1)
h h
However, due to the finiteness of digits in the mantissa of floating-point numbers, round-
off errors in these procedures dominate calculation when h → 0. Both f (x + h) and
f (x − h) tend to f (x), so that their difference tends to the difference of two almost equal
quantities and thus contains fewer and fewer significant digits that provokes an explosion
of the computational error. As an example, let us consider a computer procedure f (x)
x+1
implementing the function g(x) = x−1 and study errors provided by formulae (5.1) during
′
approximation of the value f (y) at the point y = 3 in dependence of the step h. It can
be seen from Fig. 5.1 that when h becomes sufficiently small the error of approximation
increases drastically.
Suppose now that we are in the ¬-based framework and a set of elementary functions
(sin(x), cos(x), ax etc.) is represented on the Infinity Computer by one of the usual ways
used in traditional computers (see, e.g., [76]) involving the argument x, finite constants,
and four arithmetical operations. A programmer writes a program P that should calculate
g(x) using the said implementations of elementary functions, the argument x, and finite
constants connected by four arithmetical operations. Obviously, P calculates a numerical
approximation f (x) of the function g(x). We suppose that f (x) approximates g(x) suf-
ficiently well with respect to some criteria and we shall not discuss the goodness of this
approximation here.
Then, as often happens in scientific computing, another person (a user) takes the pro-
gram P calculating f (x) and is interested to calculate f ′ (x) and higher derivatives numeri-
Numerical infinities and infinitesimals 33
−1
−2
−3
−4
log10Error −5
−6
−7
−8
−9
Forward difference
Backward difference
−10
0 1 2 3 4 5 6 7 8 9 10
Stepsize h x 10
−8
Figure 5.1. When h becomes sufficiently small the error of approximation increases drastically
cally by using P. Suppose that computer programs for calculating f ′ (x), f ′′ (x), . . . f (k) (x)
and their analytical formulae are unavailable and the internal structure of the program
calculating f (x) is unknown to the user, i.e., it is a black-box.
In this situation, our attention will be attracted to the problem of a numerical calcula-
tion of the derivatives f ′ (x), f ′′ (x), . . . f (k) (x) and to the information that can be obtained
from the computer procedure P calculating f (x) for this purpose when it is executed on
the Infinity Computer. The following theorem holds (for its proof and a detailed discus-
sion on this topic see [95]).
Theorem 5.1. Suppose that: (i) for a function f (x) calculated by a procedure imple-
mented at the Infinity Computer there exists an unknown Taylor expansion in a finite
neighborhood δ(y) of a purely finite point y; (ii) f (x), f ′ (x), f ′′ (x), . . . f (k) (x) assume
purely finite values or are equal to zero for purely finite points x ∈ δ(y); (iii) f (x) has
been evaluated at a point y + ¬−1 ∈ δ(y). Then the Infinity Computer returns the result of
this evaluation in the positional numeral system with the infinite radix ¬ in the following
form
f (y + ¬−1 ) = c0 ¬0 c−1 ¬−1 c−2 ¬−2 . . . c−(k−1) ¬−(k−1) c−k ¬−k , (5.2)
where
f (y) = c0 , f ′ (y) = c−1 , f ′′ (y) = 2! · c−2 , . . . f (k) (y) = k! · c−k . (5.3)
In order to illustrate the theorem let us consider the following example.
Example 5.2. Suppose that we have a computer procedure f (x) implementing the already
mentioned function g(x) = x+1 x−1 on the Infinity Computer and we want to evaluate the
′ ′′
values f (y), f (y), f (y), and f (3) (y) at the point y = 3. Suppose also that we do not know
34 Yaroslav D. Sergeyev
the analytical representation of g(x) and f (x) is given as a black-box, i.e., we supply an x
to the Infinity Computer and get f (x) without knowing how this result has been computed.
Instead of using formulae (5.1) that, as has been discussed above, can lead to errors,
it is proposed just to calculate f (x) at the point x = 3 + ¬−1 . The result provided by the
Infinity Computer is the following
i.e., the result is a finite number with several infinitesimal parts (their number can be fixed
a priori in dependence on the number of derivatives one wishes to calculate). From this
numeral, by applying the theorem (see (5.3)) we obtain
5.2. Infinitesimal perturbations for solving systems of linear equations. Very often
in computations, an algorithm performing calculations of f (x) at a point x = a can en-
counter a situation where division by zero occurs. Then, obviously, this operation cannot
be executed. If it is known that the problem under consideration has a finite solution f (a),
then a number of additional computational steps trying to avoid this division is performed.
For instance, the computation is repeated several times at points x = a − ε and x = a + ε
with different values of ε and trying to understand where the sequences of values f (a − ε)
and f (a + ε) converge. Notice that in the numbers f (a − ε) and f (a + ε) the contribution
of ε in the result cannot be separated from the contribution of a. For instance, suppose that
we have x = 2, ε = 0.001, and after evaluating f (2 + 0.001) we have got f (2.001) = 4.46.
This number, 4.46, does not show us where there are the contribution of x = 2 we are in-
terested in and the contribution of the perturbation ε = 0.001. These contributions are
merged in the result 4.46 and cannot be separated. Thus, by choosing different values of ε
and by repeating computations many times it is possible to get only an approximation of
the value f (a).
On the other hand, we know that the Infinity Computer works by collecting differ-
ent powers of ¬ separately. Thus, by taking ε = ¬−1 there is the possibility in certain
cases to obtain the exact finite result f (a) separated from the response provoked by the
infinitesimal perturbation ε = ¬−1 .
A nice example of this kind is solving linear systems of equations. Traditionally, the
operation of pivoting is used when one solves systems of linear equations by an algorithm
such as, e.g., Gauss-Jordan elimination. Pivoting is the interchanging of rows (or both
rows and columns) in order to avoid division by zero and to place a particularly ‘good’
element in the diagonal position prior to a particular operation.
Numerical infinities and infinitesimals 35
In case one decides to solve (5.4) by the method of Gauss-Jordan, the operation of pivoting
should be performed because the first pivotal element a11 = 0. Suppose now that we
would like to avoid pivoting and, in order to illustrate the discussion on perturbations
given above, introduce a finite perturbation ε = 0.01 in the element a11 = 0, i.e., put
a11 = 0.01. This gives us the possibility to perform Gauss transformations working with
the perturbed system
[ ] [ ] [ ] [ ]
0.01 1 2 1 100 200 1 100 200 1 0 −200
→ → → 198
2 2 2 0 −198 −398 0 1 398
198 0 1 398 198
[ ] [ ]
¬+2 2¬
1 0 2¬ − ¬ · −4
−2¬+2
1 0 −2¬+2 1 0 −1 + 1−1¬
−4¬+2
→ → .
0 1 −4¬+2 0 1 2 − 1−1¬
−2¬+2 0 1 −2¬+2
It follows immediately that, since the finite and infinitesimal parts can be separated on the
Infinity Computer, the solution to the system (5.4) is given by the finite parts of numbers
−1 + 1−1¬ and 2 − 1−1¬ , i.e., x1∗ = −1 and x2∗ = 2.
In this example, we have introduced the number ¬−1 once and, as a result, we have
obtained in the matrices expressions where the maximal power of grossone is one and
there are rational expressions depending on grossone, as well. It is possible to manage
these rational expressions in two ways: (i) to execute division in order to obtain its result in
the form (4.16), (4.17); (ii) without executing division. In the latter case, we just continue
to work with rational expressions. In the case (i), since we need finite numbers as final
results, in the result of division it is not necessary to store the parts c p ¬ p with p < −1.
These parts can be forgotten because in any way the result of their successive multipli-
cation with the numbers of the type c1 ¬1 (recall that 1 is the maximal exponent present
in the matrix under consideration) will give exponents less than zero, i.e., numbers with
36 Yaroslav D. Sergeyev
these exponents will be infinitesimals that are not interesting for us in this computational
context.
Thus, by using the positional numeral system (4.16), (4.17) with the radix grossone
we obtain [ ] [ ]
1 ¬ 2¬ 1 ¬ 2¬
¬+2 → →
0 1 −4 −2¬+2
0 1 2¬0 +1¬−1
[ ] [ ]
1 0 2¬ − ¬ · (2¬0 1¬−1 ) 1 0 −1¬0
→ .
0 1 2¬0 1¬−1 0 1 2¬0 1¬−1
The finite parts of numbers −1¬0 and 2¬0 1¬−1 then provide the required solution to the
system (5.4), i.e., x1∗ = −1 and x2∗ = 2. 2
Example 5.4. In this example, we consider the following system
0 0 1 x1 1
2 0 −1 x2 = 3 . (5.5)
1 2 3 x3 1
The coefficient matrix of this system has the first two leading principal minors equal to
zero. Consequently, the first two pivots, in the Gauss transformations, are zero. We solve
the system without pivoting by substituting the zero pivot by ¬−1 , where necessary.
Let us show how the exact computations are executed:
0 0 1 1 1 0 ¬ ¬
2 0 −1 3 → 0 0 −2¬ − 1 −2¬ + 3 →
1 2 3 1 0 2 −¬ + 3 −¬ + 1
1 0 ¬ ¬ 1 0 ¬ ¬
0 1 −2¬2 − ¬ −2¬2 + 3¬ → 0 1 −2¬2 − ¬ −2¬2 + 3¬ →
0 2 −¬ + 3 −¬ + 1 0 0 4¬2 + ¬ + 3 4¬2 − 7¬ + 1
8¬ +2¬
2
1 0 0
1 0 ¬ ¬ 4¬ +¬+3
2
0 1 −2¬2 − ¬ −2¬2 + 3¬ ¬ 2
+10¬
→ −8
0 1 0 4¬2 +¬+3 .
4¬ −7¬+1
2
0 0 1
0 0 1 4¬ 2−7¬+1
2
4¬ +¬+3
2
4¬ +¬+3
It is easy to see that the finite parts of the numbers
8¬2 + 2¬ 6
x̃1∗ = = 2− ,
4¬ + ¬ + 3
2
4¬ + ¬ + 3
2
provide the solution x1∗ = 2, x2∗ = −2, and x3∗ = 1 to the original unperturbed system (5.5).
In this procedure we have introduced the number ¬−1 two times. As a result, we
have obtained expressions where the maximal power of grossone is equal to 2 and there
are rational expressions depending on grossone, as well. By reasoning analogously to
Example 5.4, when we execute divisions, in the obtained results it is not necessary to
store the parts of the type c p ¬ p , p < −2, because in any way the result of their successive
multiplication with the numbers of the type c2 ¬2 will give finite exponents less than zero.
That is, numbers with these exponents will be infinitesimals that are not interesting for
us in this computational context. Thus, by using the positional numeral system (4.16),
(4.17), we obtain
1 0 ¬ ¬ 1 0 ¬ ¬
0 1 −2¬2 − ¬ −2¬2 + 3¬ 0 1
→ −2¬2 − ¬ −2¬2 + 3¬ .
4¬ −7¬+1
2
0 0 1 0 0 1 1¬0 -2¬−1
4¬ +¬+3
2
Note that the number 1¬0 -2¬−1 does not contain the part of the type c−2 ¬−2 because the
coefficient c−2 obtained after the executed division is such that c−2 = 0. Then we proceed
as follows
1 0 ¬ ¬ 1 0 0 2
0 1 0 −2 → 0 1 0 −2 .
0 0 1 1¬0 -2¬−1 0 0 1 1¬0 -2¬−1
The obtained solutions x1∗ = 2 and x2∗ = −2 have been obtained exactly without infinitesi-
mal parts and x3∗ = 1 is derived from the finite part of 1¬0 -2¬−1 . 2
We conclude this subsection by mentioning that zero pivots in the matrix are sub-
stituted dynamically by ¬−1 . Thus, the number of the introduced infinitesimals ¬−1
depends on the number of zero pivots.
Example 5.5. Let us consider the following quadratic 2-dimensional optimization prob-
lem with a single linear constraint proposed by Liuzzi in 2007:
1 2 1 2
minx 2 x1 + 6 x2 (5.6)
subject to x1 + x2 = 1
One of the traditional ways to solve this problem is to construct the corresponding uncon-
strained optimization problem using a finite penalty coefficient P as follows
1 2 1 2 P
min x + x + (1 − x1 − x2 )2 . (5.7)
x 2 1 6 2 2
Then, different values of P should be taken into consideration. For instance, let us take
P = 20 and write down the first order optimality conditions
{
x1 − 20(1 − x1 − x2 ) = 0,
3 x2 − 20(1 − x1 − x2 ) = 0.
1
The solution to this system of linear equations is the stationary point of the unconstrained
problem (5.7) with P = 20, namely, it is
20 60
x̃1∗ (20) = , x̃2∗ (20) = (5.8)
81 81
and it is not clear how to obtain from (5.8) the solution to the original constrained prob-
lem (5.6). Thus, another penalty coefficient P should be taken, the problem (5.7) should
be solved again in order to get a new approximation to the solution of (5.6). This proce-
dure should be repeated several times until a satisfactory approximation will be obtained.
In the Infinity Computer framework, the authors of [32] propose to take P = ¬. The
corresponding first order optimality conditions in this case are
{
x1 − ¬(1 − x1 − x2 ) = 0,
3 x2 − ¬(1 − x1 − x2 ) = 0.
1
1¬ 3¬
x̃1∗ = , x̃2∗ =
1+4¬ 1+4¬
1 1 1 3 3 3
x̃1∗ = − ¬−1 ( − ¬−1 + . . .), x̃2∗ = − ¬−1 ( − ¬−1 + . . .).
4 16 64 4 16 64
The authors of [32] have proved that the finite parts of x̃1∗ and x̃2∗ give the exact solution
x∗ = ( 14 , 34 ) to the original constrained problem (5.6). 2
Numerical infinities and infinitesimals 39
LexMax c1 · x, c2 · x, . . . , cr · x,
(5.9)
subject to Ax = b, x ≥ 0
max c̄ · x,
(5.10)
subject to Ax = b, x ≥ 0
40 Yaroslav D. Sergeyev
where
r
c̄ = ∑ ci M −i+1 . (5.11)
i=1
This is a powerful theoretical result. However, from the computational point of view find-
ing the value of M is not a trivial task. Finding an appropriate finite value of M and solving
the resulting LP problem can be more time consuming than solving the original problem
(5.9) following the preemptive approach. Indeed, the preemptive scheme requires solv-
ing r linear programming problems only in the worst case and, in addition, it does not
require the computation of M.
Let us describe now the nonpreemptive ¬-based scheme following [20] where the
problem
max ĉ · x,
(5.12)
subject to Ax = b, x ≥ 0
ĉ · x = (c11 x1 + ... + c1n xn )¬0 + (c21 x1 + ... + c2n xn )¬−1 + ... + (cr1 x1 + ... + crn xn )¬−r+1 .
The crucial advantage of (5.12) with respect to the traditional nonpreemptive scheme is
that ĉ does not involve any unknown. More precisely, it does not require the search of a
finite value M from (5.10), (5.11). In [20], a simplex-like algorithm called gross-simplex
was proposed. It is able to work with ¬-based numerals and, therefore, can solve (5.12)
numerically.
5.5. Numerical methods using infinitesimals for solving ODEs. Numerical solutions
to Ordinary Differential Equations (ODEs) are required very often in practical problems
and there exists a huge number of numerical methods proposed to solve ODEs on conven-
tional computers (see, e.g., [12, 13, 18, 47, 48, 52, 54, 73] and references given therein).
Let us consider the following initial value problem
and suppose that f (x, y) is a ‘black-box’ function, i.e., f (x, y) is given by a computer
procedure and the analytical representation of f (x, y) is unknown to the person who
solves (5.15).
Many numerical methods for solving ODEs require the computation of the derivative
of the unknown function y(x) at some specific points. In particular, this is the case with
Numerical infinities and infinitesimals 41
methods based on Taylor expansion. Usually derivatives are approximated by using auto-
matic differentiation or numerical approximations. Automatic differentiation techniques
make use of specific, often symbolic, tools based on the elementary functions involved
(see [5]). As was already mentioned, numerical differentiation, especially when higher
order derivatives are necessary, suffers of numerical instability and is not able to reach
a higher precision. Finite difference formulas are, in fact, ill-conditioned, moreover the
cancellation error produces a value of zero if h is small. As was shown in Section 5.1,
the use of the Infinity Computer allows to compute exact derivatives using the infinites-
imal perturbation equal to ¬−1 . This feature, besides preventing the above mentioned
ill-conditioning phenomenon, is particularly appealing in the case where the function to
be differentiated is a ‘black-box’. Let us show now how this fact can be used for solving
the problem (5.15).
The simplest algorithm to solve (5.15) is probably the explicit Euler Method (see, e.g.,
[13]). At each step it constructs a linear approximation of y(x) starting from the initial
point (x0 , y(x0 )). The (n + 1)th step of the Euler algorithm describes how to move from
the point xn to xn+1 = xn + h, n ≥ 0, and is executed as follows
yn+1 = yn + h f (xn , yn ). (5.16)
Let us denote by y(k) (xi ) an estimate of the k-th derivative of the solution y(x) at the point
xi . It has been shown in [98] that in order to calculate the k-th derivative of the unknown
solution y(x) at the point xi , k infinitesimals steps from the point xi using the Euler formula
with h = ¬−1 should be executed as follows
yi1 = yi + ¬−1 f (xi , yi ), yi2 = yi1 + ¬−1 f (xi + ¬−1 , yi1 ), . . .
yik = yi,k−1 + ¬−1 f (xi + (k − 1)¬−1 , yi,k−1 ).
Then, since approximations of the derivatives can be obtained by the forward differen-
ces ∆h , 1 ≤ j ≤ k, the value of h is taken equal to ¬−1 and it follows that
j
k ( )
j j
∆¬−1 = ∑ (−1)
k
y −1 , (5.17)
j=0 k xi +(k− j)¬
( j)
where k is a binomial coefficient and
∆k¬−1
y(k) (xi ) ≈ −k
+ O(¬−1 ). (5.18)
¬
Then, the main observation consists of the fact that, since the error in (5.18) has the order
∆k
equal to ¬−1 , the finite part of the value ¬−1 gives us the exact derivative y(k) (xi ) (for a
¬−k
more detailed description of the computation of derivatives for solving ODEs see [98]).
Since we are able to compute values of k derivatives of the function y(x) at a generic
point x0 , we can estimate y(x) in a neighborhood of the point x0 by its Taylor expansion
k
y(i) (x0 )
y(x) ≈ ŷ(x) = y0 + ∑ (x − x0 )i . (5.19)
i=1 i!
42 Yaroslav D. Sergeyev
In cases where the radius of convergence of the Taylor expansion ŷ(x) from (5.19)
covers the whole interval [a, b] of our interest, then there is no necessity to execute several
steps with a finite value of h. In fact, thanks to the exact derivatives calculated numerically
on the Infinity Computer the function y(x) can be approximated in the neighborhood of
the initial point x0 by its Taylor expansion ŷ(x) with an order k depending on the desired
accuracy and then ŷ(x) can be evaluated at any point x ∈ [a, b]. This method is called
Taylor for the Infinity Computer (TIC) hereinafter. This method does not assume the
execution of several iterations with a finite step h. The following example shows how this
procedure works.
Example 5.6. Let us consider the problem
Let us construct the Taylor expansion of y(x) at the initial point x0 by applying formula
(5.16) with h = ¬−1 to calculate y1 and y2 :
y1 = 1 + ¬−1 · (0 − 1) = 1 − ¬−1 ,
Thus, c−2 = 2. Let us now verify the obtained result and calculate the exact derivative
y′′ (0) using (5.21). Then we have y′′ (x) = 2e−x , and y′′ (0) = 2, i.e., c−2 = y′′ (0).
Let us proceed and calculate y3 as follows
Table 5.1. Results of a comparison on 12 test problems taken from the literature of the TIC method
with the Runge-Kutta method of the fourth order that executes 20 evaluations of f (x, y) to reach the
accuracy ε RK4
x3 x4
y(x) = x − 1 + 2e−x ≈ 1 − x + x2 − + . (5.22)
3 12
By a complete analogy it is possible to obtain additional terms in the expansion that
correspond to higher derivatives of y(x). 2
Table 5.5 presents results of numerical experiments from [105] executed on a class of
12 test functions taken from the literature. The method TIC is compared over the interval
[0, 0.2] with the Runge-Kutta method of the fourth order (RK4) with the integration step
h = 0.04, i.e., to obtain an approximation at the point x = 0.2 the method RK4 executes
5 steps and 20 evaluations of the function f (x, y) from (5.15). After the results for RK4
had been obtained, the TIC method was applied to each of 12 problems. The method
TIC stopped when the accuracy ε TIC at the point x = 0.2 was better than the accuracy
ε RK4 of the method RK4. The last column, N TIC, in Table 5.5 presents the number of
evaluations of f (x, y) executed by the TIC to reach the accuracy ε TIC. In other words, it
shows the number of infinitesimal steps executed by the TIC that is equal to the number
of exact derivatives calculated by this method. The respective solutions y RK4 and y TIC
are also shown in the table. For the considered problem the TIC method executes fewer
evaluations of f (x, y), in comparison with the Runge-Kutta method.
More sophisticated algorithms developing the idea of the usage of the infinitesimal
stepsize equal to ¬−1 can be found in [2, 74, 105]. In particular, it is shown in these
papers how to construct methods mixing finite and infinitesimal steps of integration.
44 Yaroslav D. Sergeyev
1, 2, 3, 4, . . . ¬ − 2, ¬ − 1, ¬, ¬ + 1, ¬ + 2, ¬ + 3, . . . (6.1)
| {z }
¬ elements
1, 2, 3, 4, . . . ¬ − 2, ¬ − 1, ¬, ¬ + 1, ¬ + 2, ¬ + 3, . . . (6.2)
| {z }
¬ elements
1, 2, 3, 4, . . . ¬ − 2, ¬ − 1, ¬, ¬ + 1, ¬ + 2, ¬ + 3, . . . (6.3)
| {z }
¬ elements
Of course, since we have postulated that our possibilities to express numerals are finite,
it depends on the chosen numeral system which numbers among ¬ members of these
processes we can observe.
Numerical infinities and infinitesimals 45
Suppose that we have a Turing machine with an infinite tape that contains an output
written using symbols {0, 1, . . . b − 2, b − 1} with a finite radix b. The traditional point of
view does not allow us to distinguish neither tapes having different infinite lengths nor
machines using different alphabets, i.e., {0, 1, . . . B − 2, B − 1} with B ̸= b. The question
of the possibility to have different infinite tapes is not discussed and it is supposed that
machines with any output alphabets have the same computational power if their tapes
are infinite. This happens because traditional numeral systems used to describe Turing
machines do not allow us to see these differences. The ¬-based numeral system offers
such a possibility thus providing a tool to describe Turing machines in a more precise
way.
In the ¬-based framework, it is not sufficient to say that the tape is infinite. It is
necessary to define the infinite length of the tape explicitly. As an example, let us consider
a Turing machine whose tape is a sequence of ¬ squares. Output sequences are written
on the tape using symbols from an output alphabet, let it be again {0, 1, . . . b − 2, b − 1}
with a finite b. The importance of the discussion of infinite sequences provided above for
Turing machines becomes now clear: the output sequences of symbols being sequences
cannot have more than ¬ elements.
Moreover, we can make a more accurate analysis and count the precise number of
infinite output sequences of symbols that the machine can produce. It is obvious that its
outputs can be viewed as numerals in the positional numeral system with the finite radix b
This means that we have ¬ positions that can be filled in with b symbols each, i.e., this
machine called hereinafter T1 can produce b¬ different outputs. Then, if we consider
another machine, T2 , having the tape with ¬-1 positions and outputs written using the
same base, b, the number of its outputs is b¬−1 < b¬ and each of them is one position
shorter than outputs of T1 . Moreover, if we consider the third machine, T3 , having the
tape with ¬ positions and outputs written using a base B > b, the number of its outputs
is B¬ > b¬ . In other words, the machine T3 is more powerful then the machine T1 that,
in its turn, is more powerful than the machine T2 . Let us give a couple of more detailed
illustrations.
Example 6.4. We start by considering a Turing machine T4 working with the alphabet
{0, 1, 2}, the tape with ¬/2 positions, and computing the following output
0, 1, 2, 0, 1, 2, 0, 1, 2, . . . 0, 1, 2, 0, 1, 2 . (6.7)
| {z }
¬/2 positions
Then a Turing machine T5 working with the output alphabet {0, 1} and the tape with
¬/2 positions cannot produce a sequence of symbols computing (6.7). In fact, since the
numeral 2 does not belong to the alphabet {0, 1} it should be coded by more than one
symbol. One of codifications using the minimal number of symbols in the alphabet {0, 1}
necessary to code numbers 0, 1, 2 is {00, 01, 10}. Then the output corresponding to (6.7)
and computed in this codification should be
00, 01, 10, 00, 01, 10, 00, 01, 10, . . . 00, 01, 10, 00, 01, 10. (6.8)
Numerical infinities and infinitesimals 47
Since the output (6.7) contains ¬/2 positions, the output (6.8) should contain ¬ positions.
However, by the definition of T5 it can produce outputs that have only ¬/2 positions.
Example 6.5. Let us now consider a Turing machine T6 working with the alphabet
{0, 1, 2} as T4 but whose infinite tape is one position longer than the tape of T4 , i.e., it
has ¬/2+1 positions, and T6 computes the following output
0, 1, 2, 0, 1, 2, 0, 1, 2, . . . 0, 1, 2, 0, 1, 2, 0 . (6.9)
| {z }
¬/2+1 positions
Then there is no Turing machine working with the output alphabet {0, 1} and coding the
numbers 0, 1, 2 as {00, 01, 10} such that it is able to compute the output corresponding
to (6.9) in this codification. The proof is very easy and is based on the fact that infinite
sequences cannot have more than ¬ elements. Since the output (6.9) contains ¬/2+1
positions, the output
00, 01, 10, 00, 01, 10, 00, 01, 10, . . . 00, 01, 10, 00, 01, 10, 00.
should contain ¬ + 2 positions. However, infinite sequences cannot have more than ¬
elements.
Significantly more sophisticated results for deterministic and non-deterministic Tur-
ing machines can be found in [103, 104].
6.2. From divergent series to sums with a fixed infinite number of addends. Suppose
that we consider two divergent series
∞ ∞
S1 = ∑ ai , S2 = ∑ bi . (6.10)
i=1 i=1
The traditional analysis can say a little about them and it does not allow us to execute
numerical computations with them, e.g., such operations as SS21 and S2 −S1 are not defined.
The ¬-based methodology allows us to express not only different finite numbers but
also different infinite numbers. Thus, such records as S1 = a1 +a2 +. . . or ∑∞
i=1 ai become
imprecise (to continue the analogy with Pirahã the record ∑∞ i=1 ai is viewed as a kind of
many
∑i=1 ai ). It is therefore necessary to indicate explicitly the infinite numbers, k and n,
being the number of summands in the respective sums S1 and S2 from (6.10). Thus, we
get more precise definitions of these sums where the infinite numbers of addends are
specified
Notice that this specification is exactly the same as is required in the finite case. In fact,
we know that when we have a sum with a finite number of addends it is not sufficient to
say that the number, l, of addends in the sum is finite, it is necessary to assign explicitly
a value to l applying for this purpose numerals available in a traditional numeral system
chosen to express finite numbers. For different finite values of l the result of summation
can be different.
48 Yaroslav D. Sergeyev
It is necessary to stress that (6.10) and (6.11) describe the same mathematical objects –
sums with an infinite number of addends – but with different accuracies. In the case
(6.10), the numeral ∞ is used and the number of addends is not fixed – it is just said that it
is infinite. The ¬-based lens allows us to observe that each divergent series is not a unique
object with a generic infinite number of addends expressed by the numerals ∞. Records
from (6.11) were visible like unique objects due to the low accuracy of the numeral ∞.
Now we are able to distinguish different sums S1 (k) and S2 (n) in (6.11) and for different
infinite values of k and n different results can be obtained. Analogously, Pirahã observing
sums 1+1+1+1 and 1+1+1+1+1 are not able to distinguish two sums and see them as the
same object having many addends whereas an observer able to count up to 5 will see two
different sums.
Notice also that it would not be correct to say that, e.g., the record S1 = ∑∞ i=1 ai from
(6.10) represents a ‘true infinite series that includes all possible addends ai ’ whereas
records (6.11) where the number of summands should be expressed by concrete infinite
¬-based numbers are some ‘partial infinite sums’. The sentence ‘all possible addends’
means that the numeral ∞ being the upper index of summation represents the largest pos-
sible integer (if it is not the largest one then we do not consider all possible addends).
Thus, the operation ∞ + 1 should be prohibited since we are not allowed to increase ∞
that represents the largest possible integer (notice that an interesting arithmetic where
there exists the largest integer K and the operation K + 1 is not allowed is given in [75])
and the operation ∞ − 1 should give a result different from ∞ since in this case we have
less addends in the sum. In other words, the assumption that S1 = ∑∞ i=1 ai includes all
possible addends ai leads by necessity to attribute to the symbol ∞ properties that are
different from its traditional usage, where relations ∞ + 1 = ∞ and ∞ − 1 = ∞, etc. hold.
Therefore, the meaning ‘sum of all possible addends ai ’ cannot be attributed to the record
S1 = ∑∞ i=1 ai if the numeral ∞ is used in its conventional sense. Thus, records (6.10)
and (6.11) describe the same mathematical objects – sums with an infinite number of
addends – but with different accuracies.
The following two comments should be made before we start to work with applica-
tions. First, notice that it is not possible to use ¬-based numerals in the same expression
where the numeral ∞ is present since they belong to two different numeral systems having
different accuracies. Such a record as ∞ + ¬ has no meaning because ∞ is not defined
within ¬-based numeral system and, vice versa, ¬ is not defined in traditional mathemat-
ical languages using ∞. Analogously, the record ‘many’+4 has no sense (Pirahã do not
understand what 4 is and people understanding this are not able to assign a concrete value
to ‘many’).
The second comment is required to link series to infinite sequences discussed in the
previous section. Notice that even though a sequence cannot have more than ¬ elements,
the number of items in a series can be greater than grossone because the process of sum-
ming up is not necessary executed by a sequential adding of summands.
Let us show how the ¬-based approach can be applied to calculate sums where an
infinite ¬-based number of summands has been specified.
Example 6.6. We consider two infinite series S1 = 7 + 7 + 7 + . . . and S2 = 3 + 3 + 3 + . . .
As was already mentioned for a general case, traditional analysis gives us a very poor
Numerical infinities and infinitesimals 49
information about them and such operations as, e.g., SS21 and S2 − S1 are not defined. In
order to use the ¬-based approach, it is necessary to indicate explicitly the number of
their summands.
Suppose that the sum S1 has k addends and S2 has n addends and, therefore, they are
defined as follows
S1 (k) = 7| + 7 + 7{z+ . . . + 7}, S2 (n) = 3| + 3 + 3{z+ . . . + 3} .
k addends n addends
Then S1 (k) = 7k and S2 (n) = 3n and by assigning different numerical values (finite or
infinite) to k and n we obtain different numerical values for the sums. Notice that the
new numerals for the chosen k and n allow us to calculate S2 (n) − S1 (k) (analogously,
(k)
the expression SS1 (n) can be calculated). If, for instance, k = 5¬ and n = ¬ we obtain
2
S1 (5¬) = 35¬, S2 (¬) = 3¬ and it follows
S2 (¬) − S1 (5¬) = 3¬ − 35¬ = −32¬ < 0.
If k = 3¬ and n = 7¬ + 2 we obtain S1 (3¬) = 21¬, S2 (¬) = 21¬ + 6 and it follows
S2 (7¬ + 2) − S1 (3¬) = 21¬ + 6 − 21¬ = 6.
It is also possible to calculate sums having an infinite number of infinite or infinitesimal
summands
S3 (l) = 2¬ + 2¬ + . . . + 2¬, S4 (m) = 4¬−1 + 4¬−1 + . . . + 4¬−1 .
| {z } | {z }
l addends m addends
Let us calculate now the following sum of infinitesimals where each item is ¬ times less
than the corresponding item of (6.12)
¬ ¬
¬−1 + 2¬−1 + . . . + (¬ − 1) · ¬−1 + ¬ · ¬−1 = ∑ i¬−1 = (¬−1 + 1) = 0.5¬1 + 0.5.
i=1 2
50 Yaroslav D. Sergeyev
Obviously, the obtained number, 0.5¬1 + 0.5 is ¬ times less than the sum in (6.12). This
example shows, in particular, that infinite numbers can also be obtained as the result of
summing up infinitesimals.
Let us consider now a geometric series ∑∞ i
i=0 q . Traditional analysis proves that it
converges to 1−q for q such that −1 < q < 1. We are able to give a more precise answer
1
for all values of q. To do this we should fix the number of items in the sum. If we suppose
that it contains n items, then
n
Qn = ∑ qi = 1 + q + q2 + . . . + qn . (6.13)
i=0
By multiplying the left-hand and the right-hand parts of this equality by q and by sub-
tracting the result from (6.13) we obtain
Qn − qQn = 1 − qn+1
holds for finite and infinite n. Thus, the possibility to express infinite and infinitesimal
numbers allows us to take into account infinite values of n and the infinitesimal value
qn+1 for finite q. Moreover, we can calculate Qn for infinite and finite values of n and
q = 1, because in this case we have just
Qn = 1 + 1 + 1 + . . . + 1 = n + 1.
| {z }
n+1 addends
To determine it, we should specify the number n of addends in it and, for example, for
n = ¬2 we obtain
¬2 1 − 3¬ +1
2
∑ 3 = 1 + 3 + 9 + . . . + 3 = 1 − 3 = 0.5(3¬ +1 − 1).
¬2 2
i
i=0
1 + 3 + 9 + . . . + 3¬ + 3¬ +1 = 0.5(3¬ +2 − 1).
2 2 2
¬ 1 1
∑ 2i = 1 − 2 ¬
,
i=1
6.3. Riemann series theorem and Ramanujan summation. Let us consider now se-
ries with alternating signs. Traditionally, convergence of series is studied and only for
unconditionally convergent series rearrangements of addends do not affect the sum. The
Riemann series theorem states that conditionally convergent series can be rearranged in
such a way that they either diverge or converge to an arbitrary real number. In its turn, the
¬-based methodology shows that Riemann’s result is a consequence of the weak numeral
system used traditionally to work with infinite series. In fact, the usage of ¬ allows us to
see that (as it happens in the case where the number of addends is finite) rearrangements
do not change the result for any sum with a fixed (finite or infinite) number of summands.
Example 6.10. Let us start from the famous series
S5 = 1 − 1 + 1 − 1 + 1 − 1 + . . .
In the literature there exist many approaches giving different answers regarding the value
of this series (see [60]). All of them use various notions of average (for instance, Cesàro
summation assigns the value 21 to S5 ). However, sum and its average of any kind are
different notions. In our approach we do not appeal to averages and calculate the required
sum directly. To do this, we should indicate explicitly the number of items, k, in the sum.
Then it follows that
{
0, if k = 2n,
S5 (k) = 1 − 1 + 1 − 1 + 1 − 1 + 1 − . . . = (6.15)
| {z } 1, if k = 2n + 1,
k addends
and it is not important whether k is finite or infinite. For example, S5 (¬) = 0 because the
number ¬ 2 being the result of division of ¬ by 2 has been introduced as the number of
elements of a set and, therefore, it is integer. As a consequence, ¬ is even. Analogously,
S5 (¬ − 1) = 1 because ¬ − 1 is odd.
As it happens in the cases where the number of addends in a sum is finite, the result
of summation does not depend on the way the summands are rearranged. In fact, if we
52 Yaroslav D. Sergeyev
know the exact infinite number of addends and the order the signs are alternated is clearly
defined, we know also the exact number of positive and negative addends in the sum.
Let us illustrate this point by supposing, for instance, that we want to rearrange ad-
dends in the sum S5 (2¬) in the following way
S5 (2¬) = 1 + 1 − 1 + 1 + 1 − 1 + 1 + 1 − 1 + . . .
The traditional tools give an impression that this rearrangement modifies the result. How-
ever, in the ¬-based framework we know that this is just a consequence of the weak lens
used to observe infinite numbers. In fact, thanks to ¬ we are able to fix an infinite num-
ber of summands. In our example the sum has 2¬ addends, the number 2¬ is even and,
therefore, it follows from (6.15) that S5 (2¬) = 0. This means also that in the sum there
are ¬ positive and ¬ negative items. As a result, addition of the groups 1+1–1 considered
above can continue only until the positive units present in the sum will not finish and then
there will be necessary to continue to add only negative summands. More precisely, we
have
S5 (2¬) = 1 + 1 − 1 + 1 + 1 − 1 + . . . + 1 + 1 − 1 −1 − 1 − . . . − 1 − 1 − 1 = 0, (6.16)
| {z }| {z }
¬ positive and ¬2 negative addends ¬
2 negative addends
S6 = 1 − 2 + 3 − 4 + . . .
It can be easily considered as the difference of two arithmetic progressions after we have
fixed the number of items, k, in the sum S6 (k). Suppose that it contains grossone items.
Then it follows
S6 (¬) = 1 − 2 + 3 − 4 + . . . − (¬ − 2) + (¬ − 1) − ¬ =
(1 + 3 + 5 + . . . + (¬ − 3) + (¬ − 1)) − (2 + 4 + 6 + . . . + (¬ − 2) + ¬) =
(1 + ¬ − 1)¬ (2 + ¬)¬ ¬2 − 2¬ − ¬2 ¬
− = =− .
4 4 4 2
Notice that some traditional summation techniques can be interpreted as certain weight-
ed averages on sums having ¬ addends (see [114, 115] for a detailed discussion). Here-
inafter we consider what happens in the ¬-based framework in situations where certain
summation techniques lead to negative sums of divergent series that contain infinitely
many positive integers.
Numerical infinities and infinitesimals 53
x = 1+2+4+8+... (6.17)
2x = 2 + 4 + 8 + . . . (6.18)
2x + 1 = 1 + 2 + 4 + 8 + . . . (6.19)
Thus, the right-hand side of (6.19) is equal to x from (6.17) and after the substitution we
get
2x + 1 = x, x = −1
and, as the final result, it follows
1 + 2 + 4 + 8 + . . . = −1. (6.20)
To define the value x from (6.17) using the ¬ methodology, it is necessary to indicate
the number of addends in this infinite sum. Suppose that this is an infinite number n and,
therefore, x should depend on n and be defined as
We proceed now in the manner illustrated above, i.e., we multiply both parts of this equal-
ity by 2, add 1 to both left-hand and right-hand parts of (6.21), and obtain
8 + . . . + 2n−1} +2n ,
2x(n) + 1 = |1 + 2 + 4 + {z (6.22)
x(n)
natural numbers has been already calculated in the ¬-based framework in (6.12). The
example below shows that by using Ramanujan’s approach and paying attention to the
infinite number of addends in sums involved in the consideration one arrives to the same
conclusion as in (6.12).
Example 6.13. Let us first recall how Ramanujan derives (6.24). He multiplies the left-
hand part of (6.24) by 4 and then subtracts the result from (6.24) as follows
c = 1 +2 +3 +4 +5 +6 + . . .
4c = 4 +8 +12 + . . . (6.25)
−3c = 1 −2 +3 −4 +5 −6 + . . .
Ramanujan then uses the result (considered in various forms by Euler, Cesàro and Hölder)
attributing to the alternating series 1−2+3−4+. . . the value 41 as the formal power series
1
expansion of the function (1+x) 2 for x = 1, i.e.,
1
1−2+3−4+... = . (6.26)
4
Thus, it follows from (6.25) and (6.26) that
1
−3c = 1 − 2 + 3 − 4 + 5 − 6 + . . . =
4
from where Ramanujan gets (6.24).
Let us use now the ¬ lens and observe Ramanujan’s procedure through it. First, it is
necessary to indicate an infinite number of addends, n, in the sum
c(n) = 1 + 2 + 3 + 4 + 5 + . . . + n. (6.27)
The ¬ methodology allows us to compute this sum for infinite values of n directly (see
(6.12)) and, therefore, multiplication by 4 leads to the following result
n
4c(n) = 4 + 8 + 12 + . . . + 4n = 4(1 + 2 + 3 + . . . + n) = 4 · (1 + n). (6.28)
2
Then, subtraction gives us the final answer where the left-hand part calculated by (6.12)
perfectly corresponds to the right-hand one
n
−3c(n) = −3 − 6 − 9 − . . . − 3n = −3(1 + 2 + 3 + . . . + n) = −3 · (1 + n). (6.29)
2
Notice also that Example 6.11 considers the series (6.26) providing the result valid for
infinite values of n and consistent with the respective result for its finite values
n
1 − 2 + 3 − 4 + . . . + (n − 1) − n = − ,
2
thus showing that (6.26) is just a consequence of the weakness of traditional tools used to
study series.
Numerical infinities and infinitesimals 55
However, the main trick in (6.25) consists of displacement of addends in its second
line. Without loss of generality let us see what happens with this shift in the case we have
n = ¬ addends in c(n). This means that both sums (6.27) and (6.28) have ¬ summands
each. However, the displacement of (6.25) proposes to sum up each number i from the
first line of (6.25) with the number 4 · 2i for all even i, while all odd i are summed up with
zeros, i.e., they are just moved directly to the third line of (6.25). The important novelty
offered by the ¬ methodology is that now we know that since n = ¬ in our case we have
1 ≤ i ≤ ¬. This means that to ¬ 2 even addends in (6.27) there will be added the first 2
¬
numbers from (6.28), the last addend ¬ from (6.27) corresponds to the number 4 · ¬ in 2
(6.28). As a consequence, in (6.28) there will remain ¬
2 more addends, namely, they are
( ) ( )
¬ ¬
4· +1 +4· + 2 + . . . + 4 · ¬. (6.30)
2 2
The record (6.25) does not allow us to observe them whereas thanks to ¬ these summands
are perfectly visible and, therefore, (6.25) can be re-written using (6.27) – (6.30) in the
following, more accurate than it is done traditionally, way
c(¬) = 1 + 2 +3 +4 +5 + . . . +¬ − 1+¬ ( )
4c(¬) = 4 +8 +... +4 ¬ +4 ¬ + 1 + . . . + 4¬
(6.31)
2 (2 )
−3c(¬) = 1 − 2 +3 −4 +5 + . . . +¬ − 1−¬ −4 ¬ 2 + 1 − . . . − 4¬
Let us compute now the right-hand part of the third line of (6.31) by using Example 6.11
for the first ¬ addends and the arithmetical progression formula for the remaining ¬ 2
summands
( ) ( )
¬ ¬
1−2+3−4+5+...+¬−1−¬−4 +1 −4 + 2 − . . . − 4¬ =
2 2
(( ) ( ) )
¬ ¬ ¬
− −4 +1 + +2 +...+¬ =
2 2 2
(( ) )
¬ ¬ ¬ ¬
− −4· + 1 + ¬ = −3 (¬ + 1) .
2 2 2 2
As was expected, the left-hand part of the third line of (6.31) computed using (6.29) for
n = ¬ provides the same result.
[82]), however, due to its symbolic character, this approach does not allow one to evaluate
expressions in actual infinite and infinitesimal points, it works with generic infinite and
infinitesimal variables only. The problem of assigning concrete values to these variables
is not considered in non-standard analysis.
The ¬-based methodology allows us to evaluate expressions at different actual in-
finite and infinitesimal points in the same way as we do it with finite constants. It be-
comes possible to substitute limits by precise infinite and infinitesimal numbers that can
be written explicitly and to calculate expressions even when limits do not exist. Thus,
limx→b f (x) = ∞ can be substituted by precise infinite numbers that can be different in de-
pendence on the value of x. In case b = ∞ one should choose a concrete infinite ¬-based
number x where to evaluate f (x), in case b = 0, the chosen x should be infinitesimal, and
in case b is finite, x should have a finite part and an infinitesimal one (positive or negative).
Analogously, limx→b f (x) = 0 can be substituted by precise infinitesimal numbers that can
be different in dependence on the value of x. Finally, in case there exists a finite value a
such that limx→b f (x) = a, in the ¬-based framework a is just a finite approximation of a
number A = a + α where α is an infinitesimal expressed in ¬-based numerals.
The following examples give some illustrations. The first one compares results pro-
vided by traditional analysis, non-standard analysis, and the ¬-based approach.
Example 6.14. If x is a fixed finite number then
(x + h)2 − x2
lim = 2x. (6.32)
h→0 h
(x + h)2 − x2
= 2x + h, (6.33)
h
where h is a generic infinitesimal. In its turn, the ¬-based approach allows us to work with
different values of h. If, for instance, h = ¬−1 , the answer is 2x¬0 ¬−1 , if h = 4.2¬−2
we obtain the value 2x¬0 4.2¬−2 , etc. Thus, the value of the limit (6.32), for a finite x,
is just the finite approximation of the number (6.33) having the finite part equal to 2x and
infinitesimal parts can vary in dependence on the value of h.
Example 6.15. Let us consider the following two limits
Both give us the same result, +∞, and it is not possible to execute numerically the opera-
tion
lim (5x3 − x2 + 1061 ) − lim (5x3 − x2 ).
x→+∞ x→+∞
that is an indeterminate form of the type ∞ − ∞ in spite of the fact that for any finite x it
follows
5x3 − x2 + 1061 − (5x3 − x2 ) = 1061 . (6.34)
Numerical infinities and infinitesimals 57
The new approach allows us to calculate exact values of both expressions, 5x3 − x2 + 1061
and 5x3 − x2 + 10, at any infinite (and infinitesimal) x expressible in the chosen numeral
system. For instance, the choice x = 3¬2 gives the value
for the first expression and 135¬6 -9¬4 for the second one. We can easily calculate the
difference of these two infinite numbers, thus obtaining the same result as we had for
finite values of x in (6.34):
Notice that the possibility of the direct evaluation of expressions is very important (in
particular, for automatic computations) because it eliminates indeterminate forms from
the practice of computations. For instance, in the traditional language if for a finite a,
limx→a f (x) = 0 and limy→∞ g(y) = ∞ then limx→a f (x) · limy→∞ g(y) is an indeterminate
form. In the ¬-based framework, this means that for any x = a +z where z is infinitesimal,
the value f (a + z) is also infinitesimal and for any infinite y it follows that g(y) is also
infinite. In order to be able to execute computations, we should behave ourselves as we
are used to do in the finite case. Namely, it is necessary to choose values of z and y and to
evaluate f (a + z) and g(y). After we have performed these operations it becomes possible
to execute multiplication f (a + z) · g(y) and to obtain the corresponding result that can
be infinite, finite or infinitesimal in dependence on the values of z and y and the form of
expressions f (x) and g(y).
It is possible also to execute other operations with infinitesimals and infinities making
questions with respect to f (a + z) and g(y) that could not even be formulated using the
traditional language using limits. For instance, we can ask about the result of the following
expression ( )
g(y1 )
f (a + z2 ) − 1.25g(y2 )3 (6.35)
f (a + z1 )
for two different infinitesimals z1 , z2 and two different infinite values y1 and y2 .
Example 6.16. Let us consider computation of the product f (a + z) · g(y). For the sake
of simplicity we take a = 0, g(y) = y, and
2x, x < 0,
f (x) = 1, x = 0,
3
x , x > 0.
We end this example by calculating the result of the expression (6.35) for z1 = −2¬−1 ,
z2 = −5¬−4 , y1 = ¬2 , and y2 = ¬. These values provide
( ) ( )
2
g(y1 ) g(¬ )
f (a + z2 ) − 1.25g(y2 )3 = f (−5¬−4 ) − 1.25g(¬)3 =
f (a + z1 ) f (−2¬−1 )
( )
¬2 ( )
−4
−10¬ · −1
− 1.25¬ = −10¬−4 · −0.25¬3 − 1.25¬3 = 15¬−1 . 2
3
−4¬
It is worthwhile to mention that expressions can be calculated even when their limits
do not exist. Thus, we obtain a very powerful tool for studying divergent processes.
Example 6.17. The following limn→+∞ f (n), f (n) = (−1)n n3 , does not exist. However,
we can easily calculate expression (−1)n n3 at different infinite points n. For instance, for
n = ¬ it follows f (¬) = ¬3 because grossone is even and for the odd n = 0.5¬ − 1 it
follows
Let us make a remark regarding irrational numbers. Among their properties, they are
characterized by the fact that we do not know any numeral system that would allow us to
express them by√a finite
√ number of symbols used to express other numbers. Thus, special
numerals (e, π, 2, 3, etc.) are introduced by describing their properties in some way.
These special symbols are then used in analytical transformations together with ordinary
numerals.
For example, it is possible to work directly with the symbol e in analytical transfor-
mations by applying suitable rules defining this number together with numerals taking
part in a chosen numeral system S . At the end of transformations, the obtained result will
be be expressed in numerals from S and, probably, in terms of e. If it is then required to
execute some numerical computations, this means that it is necessary to substitute e by a
numeral (or numerals) from S that will allow us to approximate e in some way.
The same situation takes place when one uses the ¬-based numeral system, i.e., while
we work analytically we use just the symbol e in our expressions and then, if we wish
to work numerically we should move to approximations. The ¬-based numeral system
opens a new perspective on the problem of the expression of irrational numbers. Let us
consider one of the possible ways to obtain an approximation of e, i.e., by using the limit
1
e = lim (1 + )n = 2.71828182845904 . . . (6.36)
n→+∞ n
In our numeral system the expression (1 + n1 )n can be written directly for finite and/or infi-
nite values of n. For n = ¬ we obtain the number e0 , so designated in order to distinguish
it from the record (6.36)
e0 = (1 + ¬−1 )¬ . (6.37)
Numerical infinities and infinitesimals 59
It becomes clear from this record why the number e cannot be expressed in a positional
numeral system with a finite base. Due to the definition of a sequence under the IUA,
such a system can have at maximum ¬ numerals – digits – to express fractional part of a
number (see sections 6.1, 7.3 for details) and, as it can be seen from (6.37), this quantity
is not sufficient for e because the item 1¬ is present in it.
¬
Naturally, it is also possible to construct more exotic e-type numbers by substituting
n in the expression (1 + 1n )n with any infinite number written in the ¬-based positional
system. For example, if n = ¬2 we obtain the number e1 = (1 + ¬−2 )¬ . The numbers
2
considered above take their origins in the limit (6.36). The same way of reasoning can be
used with respect to other irrational numbers, as well.
6.5. Functions, continuity, derivatives, and integrals. The goal of this subsection is to
discuss mathematical and physical definitions of continuity and to develop a new, Physics-
driven point of view on this notion using the ¬-based infinities and infinitesimals. The
new point of view is illustrated by a detailed consideration of one of the most fundamental
mathematical definitions – function. Then, the notion of continuity is studied in depth and
computations of derivatives and integrals are illustrated.
As was already mentioned in Introduction (see discussion on page 3) in modern
Physics the continuity of an object is relative. For example, if we observe a table by
eye, then we see it continuous. If we use a microscope for our observation, we see that
the table is discrete. This means that we decide how to regard the object, as a continuous
or as a discrete one, by choosing a suitable instrument of observation.
In traditional Mathematics any mathematical object is either continuous or discrete
(e.g., the same function cannot be both continuous and discrete). This contraposition of
discrete and continuous in traditional Mathematics does not reflect properly the modern
physical interpretations showing that the continuity of an object A depends on the accu-
racy of the instrument used for the observation of A. The ¬-based methodology gives us a
possibility to develop a theory of continuity that is closer to the physical world and better
reflects the new discoveries made by physicists. We start by introducing a definition of
the one-dimensional continuous set of points based on the above considerations and Pos-
tulate 2 and establish relations to such a fundamental notion as function using infinities
and infinitesimals.
Let us remind that traditionally a function f (x) is defined as a binary relation among
two sets X and Y (called the domain and the codomain of the relation) with the additional
property that to each element x ∈ X corresponds exactly one element f (x) ∈ Y . We con-
sider now a function f (x) defined over a one-dimensional interval [a, b]. It follows from
the previous sections that to define a function f (x) over an interval [a, b] it is not suffi-
cient to give a rule for evaluating f (x) and the values a and b because we are not able to
evaluate f (x) at any point x ∈ [a, b]. Only points x expressible in known numeral systems
can be used in computations. However, the traditional definition of function does not pay
attention to the problem of expressibility of points from the domain at which f (x) can be
evaluated. This fact in certain cases can lead to ambiguity.
Therefore, in order to be precise in the definition of a function, it is necessary to indi-
cate explicitly a numeral system, S , we intend to use to express points from the interval
60 Yaroslav D. Sergeyev
[a, b]. Thus, a function f (x) is defined when we know a rule allowing us to obtain f (x)
given x and its domain, i.e., the set [a, b]S of points x ∈ [a, b] expressible in the chosen
numeral system S . We suppose hereinafter that the system S is used to write down f (x)
(of course, the choice of S determines a class of formulae and/or procedures we are able
to express using S ) and it allows us to express any number
The number of points of the domain [a, b]S can be finite or infinite but the set [a, b]S
is always discrete. This means that for any point x ∈ [a, b]S it is possible to determine its
closest right and left neighbors, x+ and x− , respectively, as follows
x+ = min{z : z ∈ [a, b]S , z > x}, x− = max{z : z ∈ [a, b]S , z < x}. (6.38)
Apparently, the obtained discrete construction leads us to the necessity to abandon the
nice idea of continuity, which is a very useful notion used in different fields of Mathemat-
ics. But this is not the case. In contrast, the new approach allows us to introduce a new
definition of continuity very well reflecting the physical world.
Let us consider n + 1 points
and suppose that we have a numeral system S allowing us to calculate their coordinates
using a unit of measure µ (for example, meter, inch, etc.) and to construct the set X =
[a, b]S expressing these points.
The set X is called continuous in the unit of measure µ if for any x ∈ (a, b)S it fol-
lows that the differences x+ − x and x − x− from (6.38) expressed in units µ are equal to
infinitesimal numbers. In our numeral system with grossone this means that all the differ-
ences x+ − x and x − x− contain only negative (finite or infinite) grosspowers. Note that
it becomes possible to differentiate types of continuity by taking into account values of
grosspowers of infinitesimal numbers (continuity of order ¬−1 , continuity of order ¬−2 ,
etc.).
This definition emphasizes the physical principle that there does not exist an absolute
notion of continuity: it is relative with respect to the chosen instrument of observation
which in our case is represented by the unit of measure µ. Thus, the same set can be
viewed as continuous or discrete in dependence of the chosen unit of measure.
Example 6.18. Suppose that the set of six equidistant points
X1 = {a, x1 , x2 , x3 , x4 , x5 } (6.40)
from Fig. 6.1 has the distance d between the points equal to ¬−1 in a unit of measure µ
and, therefore, it is continuous in µ. Usage of a new unit of measure ν = ¬−3 µ implies
that d = ¬2 in ν and the set X1 is not continuous in ν.
Note that the introduced definition does not require that all the points from X are
equidistant. For instance, if in Fig. 6.1 for a unit measure µ the largest distance x6 − x5
over the set [a, b]S is infinitesimal then the whole set is continuous in µ.
Numerical infinities and infinitesimals 61
Figure 6.1. It is not possible to say whether this function is continuous or discrete until we have not
introduced a unit of measure and a numeral system to express distances between the points
The set X is called discrete in the unit of measure µ if for all points x ∈ (a, b)S it
follows that the differences x+ − x and x − x− from (6.38) expressed in units µ are not
infinitesimal numbers. In our numeral system with radix grossone this means that in all
the differences x+ − x and x − x− negative grosspowers cannot be the largest ones. For
instance, the set X1 from (6.40) is discrete in the unit of measure ν from Example 6.18.
Of course, it is also possible to consider intermediate cases where sets have continuous
and discrete parts.
The introduced notions allow us to give the following very simple definition of a
function continuous at a point. A function f (x) defined over a set [a, b]S continuous in
a unit of measure µ is called continuous in the unit of measure µ at a point x ∈ (a, b)S
if both differences f (x) − f (x+ ) and f (x) − f (x− ) are infinitesimal numbers in µ, where
x+ and x− are from (6.38). For continuity at points a, b it is sufficient that one of these
differences is infinitesimal. The notions of continuity from the left and from the right in a
unit of measure µ at a point are introduced naturally. Similarly, the notions of a function
discrete, discrete from the right, and discrete from the left can be defined.
The function f (x) is continuous in the unit of measure µ over the set [a, b]S if it is
continuous in µ at all points of [a, b]S . Again, it becomes possible to distinguish types of
continuity by taking into account values of grosspowers of infinitesimal numbers (conti-
nuity of order ¬−1 , continuity of order ¬−2 , etc.) and to consider functions in such units
of measure that they become continuous or discrete over certain subintervals of [a, b]. In
what follows, we shall often fix the unit of measure µ and write just ‘continuous func-
tion’ instead of ‘continuous function in the unit of measure µ’. Let us give three simple
examples illustrating the introduced definitions.
Example 6.19. We start by showing that the function f (x) = x2 is continuous over the
set X2 defined as the interval [0, 1] where numerals i , 0 ≤ i ≤ ¬, are used to express its
¬
points in units µ. First of all, note that the set X2 is continuous in µ because its points are
equidistant with the distance d = ¬−1 . Since this function is strictly increasing, to show
its continuity it is sufficient to check the difference f (x) − f (x− ) at the point x = 1. In this
62 Yaroslav D. Sergeyev
¬−1
x− = ¬ − 1 + = ¬ − ¬−1 ,
¬
This number is not infinitesimal because it contains the finite part 2 and, as a consequence,
f (x) = x2 is not continuous over the set X3 .
Example 6.21. Consider f (x) = x2 defined over the set X4 being the interval [¬ − 1, ¬]
where numerals ¬ − 1 + i 2 , 0 ≤ i ≤ ¬2 , are used to express its points in units µ. The set
¬
X4 is continuous and we check the difference f (x) − f (x− ) at the point x = ¬. We have
¬2 − 1
x− = ¬ − 1 + = ¬ − ¬−2 ,
¬2
i.e., in the neighborhood of the point x = 0 three different expressions are used to evalu-
ate f (x).
Numerical infinities and infinitesimals 63
where the number l is any number such that the same formula f1 (ξ) is used to define f (ξ)
at all points ξ such that x − l ≤ ξ < x. Analogously, the number r is any number such that
the same formula f3 (ξ) is used to define f (ξ) at all points ξ such that x < ξ ≤ x + r. Of
course, as a particular case it is possible that the same formula is used to define f (ξ) over
the interval [x − l, x + r], i.e.,
It is also possible that (6.44) does not hold but formulae f1 (ξ) and f3 (ξ) are defined at the
point x and are such that at this point they return the same value, i.e.,
If condition (6.45) holds, we say that function f (x) has continuous formulae at the point
x. Of course, in the general case, formulae f1 (ξ), f2 (ξ), and f3 (ξ) can be or cannot be
defined out of the respective intervals from (6.43). In cases where condition (6.45) is not
satisfied we say that function f (x) has discontinuous formulae at the point x. Definitions
of functions having formulae which are continuous or discontinuous from the left and
from the right are introduced naturally.
Example 6.23. Let us study the following function
{
2 −1
¬2 + xx−1 , x ̸= 1,
f (x) = (6.46)
a, x = 1,
at the point x = 1. By using designations (6.43) and the fact that for x ̸= 1 it follows
x2 −1
x−1 = x + 1 we have
f1 (ξ) = ¬2 + ξ + 1, ξ < 1,
f (ξ) = f2 (ξ) = a, ξ = 1,
f3 (ξ) = ¬2 + ξ + 1, ξ > 1,
Since
f1 (1) = f3 (1) = ¬2 + 2, f2 (1) = a,
we obtain that if a = ¬2 + 2, then the function (6.46) has continuous formulae11 at the
point x = 1. Analogously, the function (6.41) has continuous formulae at the point x = 0
from the left and discontinuous from the right.
11 Note, that even if a = ¬2 + 2 + ε, where ε is an infinitesimal number expressed in ¬-based numerals
(recall that all infinitesimals are different from zero), we are able to establish that the function has discontinuous
formulae.
64 Yaroslav D. Sergeyev
f (¬) = ¬ + ¬2 , f ′ (¬) = 2 + ¬.
For the infinitesimal x = ¬−1 we have the value f (¬−1 ) = 1 + ¬−3 having a finite and
an infinitesimal part and the first derivative is infinite having also an infinitesimal part,
namely, it follows f ′ (¬−1 ) = ¬ + 2¬−2 .
Analogously to the passage from series to sums with a fixed infinite number of items,
the introduction of infinite and infinitesimal numerals allows us (in fact, it imposes) to
substitute improper integrals of various kinds by integrals defined in a more precise way.
For example, let us consider the following improper integral
∫ ∞
x2 dx.
0
Now it is necessary to define its upper infinite limit of integration explicitly. Then, differ-
ent infinite numbers used instead of ∞ will lead to different results, as it happens in the
finite case where different finite limits of integration lead to different results.
Example 6.25. For instance, numbers ¬ and ¬2 give us two different integrals both
assuming infinite but different values
∫ ¬ ∫ ¬2
1 1
x2 dx = ¬3 , x2 dx = ¬6 .
0 3 0 3
Numerical infinities and infinitesimals 65
Moreover, it becomes possible to calculate integrals where both endpoints of the interval
of integration are infinite as in the following example
∫ ¬2
1 1
x2 dx = ¬6 − ¬3 .
¬ 3 3
It becomes possible to calculate integrals of functions assuming infinite and infinites-
imal values and the results of integration can be finite, infinite, and infinitesimal. For
instance, in the integral
∫ ¬+¬−2
1 1 1
x2 dx = (¬1 + ¬−2 )3 − ¬3 = 1¬0 + 1¬−3 + ¬−6
¬ 3 3 3
the result has a finite part and two infinitesimal parts and in the integral
∫ ¬+¬−2
1 1
x2 − x dx = 1¬−3 − ¬−4 + ¬−6
¬ 2 3
the result has three infinitesimal parts. The last two examples illustrate situations when
the integrand is infinite
∫ ¬2 ∫ ¬2
1 1
¬x dx = ¬
2
x2 dx = ¬7 − ¬4 ,
¬ ¬ 3 3
and infinitesimal
∫ ¬2 ∫ ¬2
−4 2 −4 1 1
¬ x dx = ¬ x2 dx = ¬2 − ¬−1 .
¬ ¬ 3 3
In the ¬-based terminology this question is meaningless because of the following reason.
In the finite case, to define a set it is not sufficient just to say that it is finite. It is
necessary to indicate its number of elements explicitly as, e.g., in this example
G1 = {g(i) : 1 ≤ i ≤ 5},
66 Yaroslav D. Sergeyev
or implicitly, by saying that elements satisfying a certain condition belong to the set, for
instance, as it is done below
where b is finite.
Thanks to the ¬-based numeral system we have mathematical tools to express the
number of elements for infinite sets, too. Thus, analogously to the finite case and due
to Postulate 3, it is not sufficient to say that a set has infinitely many elements. It is
necessary to indicate its number of elements explicitly or implicitly. For instance, the
number of elements of the set
G3 = {g(i) : 1 ≤ i ≤ ¬10 }
It follows from the IUA that A1 (k, n) = Nk,n from (4.2). By applying (7.2) we find for
A1 (k, n) its number of elements
⌊¬−k ⌋ ⌊¬−k⌋ ¬ ¬
J1 (k, n) = +1 = +1 = −1+1 = .
n n n n
Analogously, the set
has
⌊√ ¬ − k ⌋
j
J2 (k, n, j) =
n
elements. In particular, this means that the set, I 2 of square natural numbers traditionally
written as
I 2 = {x : x ∈ N, i ∈ N, x = i 2 } = {1, 4, 9, 16, 25, . . . } (7.3)
√
is A2 (0, 1, 2) and, therefore, it has ⌊ ¬⌋ elements. Notice that the same result has been
obtained in [66] using alternative instruments.
Numerical infinities and infinitesimals 67
As was already discussed in section 4.2, the new approach does not contradict the one-
to-one correspondence principle. The traditional record (famously mentioned by Galileo
Galilei) establishes bijection among the sets I 2 and N as follows
1, 22 , 32 , 42 , 52 , 62 , . . .
↕ ↕ ↕ ↕ ↕ ↕ (7.4)
1, 2, 3, 4 5, 6, . . .
but does not allow one to see that these two sets have different numbers of elements.
The answer that both sets are countable is correct but its accuracy is low. The ¬-based
methodology allows us to see the difference in their number of elements and to express
the final part of (7.4), as follows
1, 22 , 32 , 42 , 52 , . . . (⌊¬1/2 ⌋ − 2)2 , (⌊¬1/2 ⌋ − 1)2 , ⌊¬1/2 ⌋2
↕ ↕ ↕ ↕ ↕ ↕ ↕ ↕ (7.5)
1, 2, 3, 4 5, . . . ⌊¬ ⌋ − 2,
1/2
⌊¬ ⌋ − 1,
1/2
⌊¬1/2 ⌋
Infinite sets defined by other functions g(i) from (7.2) can be studied by a complete anal-
ogy.
7.2. Countable sets can be measured more accurately. Remind that Postulate 2 stres-
ses the difference between mathematical objects and the tools used to observe them. In
this section, infinite sets of numbers are our objects of observation and numeral systems
are our instruments. If we fix a numeral system S and a set A then the following three
situations can take place with respect to expressibility of the number of elements, J, of
the set A in S : (i) S is able to express J exactly, i.e., to the accuracy of one element;
(ii) S is able to express J with a low accuracy (Pirahã would say that the set of 5 apples
has many apples; this answer is correct but its accuracy is low); (iii) S is not able to
express J at all. The last option can occur for several reasons. For instance, because
J cannot be determined (e.g., the inequality from (7.2) cannot be solved or the formula
for elements of the set is not known). Another possibility is where the set A consists
of numbers not expressible in S (or A contains among its elements some numbers of
this kind) and, therefore, A itself cannot be observed through S . Finally, a numeral that is
required to express J may not belong to S . Thus, each numeral system can express exactly
the number of elements only of a limited number of sets. Let us consider some sets having
countable or continuum cardinalities and see what the ¬-based numeral system can say
with respect to the number of elements of these sets. The designation |A| will be used
hereinafter to indicate the number of elements of a set A expressed in ¬-based numerals.
We have already seen above that the set, N, of natural numbers has ¬ elements, the
set, Z, of integers has 2¬+1 elements (see example 4.7), and the sets, E and O, of even
and odd numbers have ¬ 2 elements each. Let us study the set of rational numbers now.
Traditionally, rational numbers are defined as ratios of two integers. The new approach
allows us to calculate the number of numerals in a fixed numeral system. Let us consider
a numeral system Q1 containing numerals of the form
p
, p ∈ Z, q ∈ Z, q ̸= 0. (7.6)
q
68 Yaroslav D. Sergeyev
Theorem 7.2. The number of elements of the set, Q1 , of rational numerals of the type
(7.6) is |Q1 | = 4¬2 + 2¬.
Proof. Since the set Z has 2¬+1 elements, the numerator of (7.6) can be filled in by 2¬ +
1 and the denominator by 2¬ numbers. Thus, the number of all possible combinations is
(2¬ + 1) · 2¬ = 4¬2 + 2¬
−5 2 ¬ − 1 1 ¬−2 1 ¬−5 ¬
, , = 1+ , = 1− , , .
¬ ¬-6 ¬ − 2 ¬−2 ¬−1 ¬−1 3 2
The first two of them are infinitesimals, the next two numbers are finite with infinitesimal
b
parts, and the last two numbers are infinite. In the same way as we have introduced sets, N
b of extended natural and integer numbers, it is possible to consider the set, Q,
and Z, b of
extended rational numbers such that in the fraction qp numbers p and q can be integers
larger than ¬, for instance, the number 3 is not rational since its denominator is larger
5¬
Numerical infinities and infinitesimals 69
Table 7.1. Cardinalities and the number of elements of some infinite sets.
2
the power set of the set of numerals Q2 from (7.7) continuum, c 22¬ +1
7.3. The number of elements of sets having the cardinality of the continuum. Let
us now consider real numbers and again pay attention to the separation of the object of
observation from the tool used to observe the object. Traditionally in Mathematics there
is no such separation and real numbers often are identified with numbers expressible in a
positional numeral system. In the ¬ framework, the real line is an abstract mathematical
object that is observed by numeral systems and different numeral systems (that are our in-
struments of observation) allow us to see different sets of real numbers and the positional
numeral system is just one of the possible lenses that can be used to observe real numbers.
There exist also other lenses. As was already mentioned, the Roman numeral system does
not allow us to see negative numbers and the Pirahã system only detects numbers 1 and 2
across the whole real line. So, once we have fixed a numeral system that we intend to
use to express real numbers, we have decided which real numbers we can observe. The
accuracy of the chosen lens fixes what can be observed and what cannot and, therefore,
it does not make sense to talk about things that are not observable with the chosen lens.
Instead of this, if we need a higher accuracy, it is better to change the present lens with a
stronger one.
Let us consider the real numbers expressible in the positional numeral system, Rbnq ,
with the finite integer radix b, where n ∈ N digits are used to express the integer part, and
q ∈ N digits to write down the fractional part of a number, i.e., Rbnq uses the following
numerals
(an−1 an−2 . . . a1 a0 .a1 a2 . . . aq−1 aq )b . (7.8)
The first observation that can be made concerns the number of numerals expressible in
the form (7.8).
Theorem 7.4. The maximal number of numerals that can be expressed by the numeral
system Rbnq is |Rb¬¬ | = b2¬ .
Proof. In formula (7.8), defining the type of numerals we deal with, there are two se-
quences of digits: the first one, an−1 an−2 . . . a1 a0 , is used to express the integer part of
the number and the second, a1 a2 . . . aq−1 aq , for representing its fractional part. Due to
the definition of sequence and the IUA, each of them can have at maximum ¬ elements.
Thus, it can be at maximum ¬ positions on the left of the dot in (7.8) and, analogously,
¬ positions on the right of the dot. Every position can be filled in by one of the b dig-
its from the alphabet {0, 1, . . . , b − 1}. Thus, we have b¬ combinations to express the
integer part of the number and the same quantity to express its fractional part. As a re-
sult, the positional numeral system using the numerals of the form (7.8) can express b2¬
numbers.
Notice that all the numerals12 of the type (7.8) represent different numbers, the least
and greatest numbers expressible in Rbnq can be explicitly indicated as well as the smallest
positive number expressible in this system.
12 The result of theorem 7.4 does not consider the practical situation of writing down concrete numerals.
Obviously, the number of numerals of the type (7.8) that can be written down in practice is finite and depends
on the chosen numeral system for writing digits. Theorem 7.4 shows us how many numerals of the type (7.8)
exist.
Numerical infinities and infinitesimals 71
Example 7.5. For instance, in the decimal positional system R10,1,¬ the numerals
1. 999
| {z. . . 99}, 2. 000
| {z. . . 00}
¬ digits ¬ digits
represent different numbers and their difference is equal to the smallest positive number
expressible in this system
2. 000
| {z. . . 00} −1. 999
| {z. . . 9} = 0. 000
| {z. . . 01} .
¬ digits ¬ digits ¬ digits
Recall that the traditional point of view on real numbers implies that there exist real num-
bers that can be represented in positional systems by two different infinite sequences
of digits, for instance, in the decimal positional system the records 2.000000 . . . and
1.99999 . . . represent the same number. As always, there is no contradiction between the
traditional and the new points of view. They just use different lenses in their mathematical
microscopes to observe numbers. The instruments used on the traditional view for this
purpose are just too weak to distinguish two different numbers in the records 2.000000 . . .
and 1.99999 . . ..
Analogously the smallest and the largest numbers expressible in Rbnq can be easily
indicated. For the numeral system R10¬¬ they are, respectively,
− 999
| {z. . . 9} . 999
| {z. . . 9}, | {z. . . 9} . 999
999 | {z. . . 9} .
¬ digits ¬ digits ¬ digits ¬ digits
Recall that traditionally it is supposed that any positional numeral system is able to
represent all real numbers (‘the whole real line’) and very often the real line is identified
with numbers written in the positional numeral system. In the ¬-based framework we
emphasize that any numeral system is just an instrument that can be used to observe
certain real numbers. This instrument can be more or less powerful, e.g., the positional
system (7.8) with the base b = 10 can express 102¬ numerals and, therefore, it is more
powerful than the positional system (7.8) with the radix b = 2 that can express 22¬ < 102¬
numerals. Traditionally, this difference is invisible.
In general, two numeral systems can allow us to observe either the same sets of num-
bers, or sets of numbers having an intersection, or two disjoint sets of numbers. Due to
Postulate 2, we are not able to answer the question ‘What is the whole real line?’ because
this is the question asking ‘What is the object of the observation?’, we are able just to
invent more and more powerful numeral systems that will allow us to improve our ob-
servations of numbers by using newly introduced numerals. Thus, questions that can be
formulated and answered require first to fix a numeral system S (our lens of observation)
and then to ask ‘Which real numbers are expressible through S ?’.
The following theorem stresses once again the difference existing between objects of
the observation (e.g., the field of real numbers) and what is viewed through our instru-
ments (e.g., numerals expressible in Rbnq ).
Theorem 7.6. The sets Z, Q1 , Q2 , and Rbnq are not monoids under addition.
Proof. The proof is obvious and thus omitted.
72 Yaroslav D. Sergeyev
7.4. Relations to results of Georg Cantor. In this subsection we revisit in the ¬-based
framework some of results of Cantor related to measuring infinite sets. We start by cal-
culating the number of points in the interval [0, 1) used in Cantor’s diagonal argument
to show that this number is uncountable. To do this we need a definition of the notion
‘point’ and mathematical tools that can be used to indicate a point. Since this concept is
one of the most fundamental, it is very difficult to find an adequate definition for it (re-
mind that in the ¬-based methodology it is better to say ‘description’ since any definition
is done using a certain mathematical language L and, therefore, it allows us to say about
the object only things expressible in L , i.e., each definition describes the object with the
accuracy of L ). If we accept (as is usually done in modern Mathematics) that a point in
[0, 1) is determined by its coordinate x, then x can be used both to indicate the point and
to execute some computations with x (for instance, to evaluate a function f (u) at the point
u = x). Since we can express coordinates only by numerals, different choices of numeral
systems lead to various sets of numerals that can be used for expressing coordinates and,
as a consequence, to different sets of points we can refer to. In each concrete case, the
choice of a numeral system S will define what the word point means for us in this case
and we shall not be able to work with the points whose coordinates are not expressible in
the chosen numeral system S . As a consequence, we are able to calculate the number of
points over [0, 1) if we have already decided which numerals will be used to express the
coordinates of points.
Different numeral systems can be chosen to express coordinates of points in depen-
dence on the precision level we want to obtain and once this choice has been done it
becomes possible to count the number of numerals available to express coordinates. For
example, Pirahã are not able to express any point over the interval [0, 1). If numbers
x ∈ [0, 1) are expressed in the form p−1 , p ∈ N, then the smallest positive number we can
¬
distinguish is 1 and the interval [0, 1) contains the following ¬ points
¬
1 2 ¬−2 ¬−1
0, , , ... , . (7.9)
¬ ¬ ¬ ¬
If we want to count the number of intervals of the form [a − 1, a), a ∈ N, on the ray
x ≥ 0, then, by Postulate 3, the definition of sequence, and Theorem 6.1, not more than ¬
intervals of this type can be distinguished on the ray x ≥ 0. They are
Within each of them we are able to distinguish ¬ points of the kind a + y, where y is
from (7.9), and, therefore, across the entire ray ¬2 points of this type can be observed.
Analogously, the ray x < 0 is represented by the intervals
Hence, this ray also contains ¬2 such points and on the whole line 2¬2 points of this type
can be represented and observed.
Note that the point −¬ is included in this representation and the point ¬ is excluded
from it. Let us slightly modify our numeral system in order to have ¬ representable. For
Numerical infinities and infinitesimals 73
this purpose, intervals of the type (a − 1, a], a ∈ N, should be considered to represent the
ray x > 0 and the separate symbol, 0, should be used to represent zero. Then, on the ray
x > 0 we are able to observe ¬2 points and, analogously, on the ray x < 0 we also are able
to observe ¬2 points. Finally, by adding the symbol used to represent zero we obtain that
on the entire line 2¬2 + 1 points can be observed.
As was already mentioned, different numeral systems allow us to see the interval
[0, 1) and the entire real line in different way. For instance, if only integers are used for
this purpose, then over [0, 1) only the point x = 0 is visible and the real line is represented
by 2¬+1 points
−¬, −¬ + 1, −¬ + 2, . . . − 2, −1, 0, 1, 2, . . . ¬ − 2, ¬ − 1, ¬.
In order to start a comparison with Cantor’s results showing that the number of points
of [0, 1) is uncountable, let us recall that he used the following numerals in his famous
diagonal argument
(.a1 a2 a3 a4 a5 a6 . . .)b (7.10)
belonging to a positional numeral system with a radix b. Notice that traditionally in (7.10)
b = 2 or b = 10 are used and it is supposed that both values can express all real numbers
over the interval [0, 1). Clearly, traditional numeral systems did not allow Cantor to see the
end part of the sequence of digits (7.10). The ¬-based framework offers this possibility
and the following positional numeral system Rb0q
being a particular case of the numeral system from (7.8) can be used for this purpose.
Theorem 7.7. The number of numerals of the type (7.11) for a fixed infinite value of q is
equal to bq .
Corollary 7.8. The maximal number of numerals of the type (7.11) is equal to b¬ .
Corollary 7.9. The entire real line contains 2¬b¬ points of the type a + y, where a is
integer and y is from (7.11) with q = ¬.
Proof. We have already seen above that it is possible to distinguish 2¬ unit intervals
within the real line. Thus, the whole number of points of the type (7.11) on the line is
equal to 2¬b¬ .
In Corollary 7.9, we represented the real line by numerals (7.11) used in each of the
intervals [a, a + 1) where a ∈ Z. If we are not interested in subdividing the line at inter-
vals and want to obtain the number of the points on the line directly by using positional
numerals of the type (7.8) then, as it has already been established in Theorem 7.4, the
number of points expressible by the numerals (7.8) is |Rb | = b2¬ .
74 Yaroslav D. Sergeyev
It is of the essence to stress that the results presented above can be considered as a
more precise analysis of the situation discovered by Cantor. He has proved, by using
his famous diagonal argument, that the number of elements of the set N is less than the
number of real numbers at the interval [0, 1) without calculating the latter. To do this he
expressed real numbers in a positional numeral system (7.10). We have shown that this
number will be different depending on the radix b used in the positional system to express
real numbers. However, the resulting number, b¬ , of points expressible in this numeral
system will be larger than the number of elements of the set of natural numbers, ¬, for
any radix b > 1 and, therefore, the diagonal argument maintains its force.
The analysis made above shows that, as it was for countable sets, the ¬-based method-
ology allows us to realise that sets having the cardinality of continuum can have different
infinite numbers of elements expressible in ¬-based numerals (see Table 7.1). Both nu-
meral systems observe the same objects – infinite sets. However, Cantor through his lens
distinguishes only two dots (countable and continuous) and the Continuum Hypothesis
(metaphorically speaking) asks whether it is possible to distinguish other dots between
them. The ¬ lens instead of just two dots allows us to observe many different dots (i.e.,
infinite sets with different infinite numbers of elements). Moreover, it becomes possible
to observe sets invisible to Cantor’s lens. Some of them have a special importance with
respect to the Continuum Hypothesis. They are studied in the following example.
Example 7.10. Let us consider again numerals (7.11). Due to construction, it is expected
that for infinite values of q the set Rb0q should have the cardinality of the continuum in
the traditional language. Let us consider now q1 = ⌊logb ¬⌋ where ⌊x⌋, as usual, is the
integer part13 of x. Notice that q1 is infinite since ¬ is infinite and b is finite. It follows
then that the number of numerals in Rb0q1 is
i.e., with respect to the traditional language the set Rb0q1 would be countable. Analo-
gously, many different instances of infinite sets that are constructed starting from the con-
tinuum framework and resulting at the end to be countable can be exhibited. For example,
for infinite q2 = 3⌊logb ¬⌋ and q3 = 0.5⌊logb ¬⌋ it follows that
√
b3⌊logb ¬⌋ < b3 logb ¬ = ¬3 < b¬ , b0.5⌊logb ¬⌋ < b0.5 logb ¬ = ¬ < b¬ ,
i.e., the sets of numerals Rb0q2 and Rb0q3 would be also countable from the traditional
point of view.
Thus, thanks to the ¬-based numerals it becomes possible to calculate the exact num-
ber of elements of old (see Table 7.1) and new sets and to exhibit sets that were constructed
as continuum but are indeed countable bridging so the gap between the two groups of sets.
These results show that the difficulty of the Continuum Hypothesis is a consequence of
the weakness of the numeral systems of Cantor’s cardinals and the passage to ¬-based
numerals allows one to avoid troubles. Analogously, it is difficult (impossible) to answer
13 Notice that log ¬ is not integer. It can be shown by the following simple argument. Suppose that there
b
exist integers b and x such that bx = ¬ and b is a finite number. Then ¬ would be divisible only by b and this is
impossible due to the IUA.
Numerical infinities and infinitesimals 75
Y Y
ye
f(x)
0 x 1 X 0 x 1 X
(a) (b)
Figure 7.1. (a) Due to Cantor, the interval (0, 1) and the entire real line Y have the same number of
points. (b) The ¬-based framework allows us to observe three independent mathematical objects:
the set XS1 represented by small circles, the set YS2 represented by stars, and function (7.12)
to the question about the result of V-V in Roman numerals that do not allow one to express
zero whereas this problem vanishes if one works in a positional numeral system where
zero can be expressed.
Clearly, the choice of a mathematical language depends on the practical problem that
one intends to solve and on the accuracy required for such a solution. Such results as
ℵ0 + 1 = ℵ0 , c + ℵ0 = c, ‘many’+1=‘many’
are correct. If one is satisfied with their accuracy, these answers can be successfully used
in practice (even ‘many’ is used by Pirahã nowadays). However, if one needs more precise
results, it is necessary to introduce a more powerful mathematical language (a numeral
system in this case) allowing one to express the required answer in a more accurate way
as the ¬-based numeral system does.
We conclude this section stressing once again that there is no contradiction between
Cantor’s results and the ¬-based framework. Bijections among countable sets have been
already studied in section 4.2. Let us now return to the problem of comparison of infinite
sets and consider Cantor’s result showing that the cardinality of the set of points over the
interval (0, 1) is equal to the cardinality of the set of points over the whole real line.
The proof of this counterintuitive fact is given by establishing a one-to-one correspon-
dence between the elements of the two sets. Due to Cantor, such a mapping may be given,
for example, by the function
y = tan(0.5π(2x − 1)), x ∈ (0, 1), (7.12)
illustrated in Fig. 7.1(a). Cantor shows by using this figure that to any point x ∈ (0, 1) a
point y ∈ (−∞, ∞) can be associated and vice versa. Thus, he concludes that the requested
76 Yaroslav D. Sergeyev
bijection between the sets R and (0, 1) has been established and, therefore, they have the
same cardinality.
In the ¬-based methodology the number of elements of a set is its intrinsic character-
istic (for both finite and infinite sets) that does not depend on any object outside the set.
Thus, in Cantor’s example from Fig. 7.1(a) we have (see Fig. 7.1(b)) three mathematical
objects: (i) a set, XS1 , of points over the interval (0, 1) which we are able to distinguish
using a fixed numeral system S1 ; (ii) a set, YS2 , of points over the vertical real line which
we are able to distinguish using another fixed numeral system S2 ; (iii) the function (7.12)
described using the third fixed numeral system S3 . All these three mathematical objects
are independent of each other. The sets XS1 and YS2 can have the same or a different
number of elements.
It should be stressed that we are not able to evaluate f (x) at any point x. We are
able to do this only at points from XS1 (even if we consider all numeral systems known to
humanity, the number of points which we can express will be limited). Of course, in order
to be able to execute these evaluations it is necessary to conciliate the numeral systems
S1 , S2 , and S3 . The fact that we have made evaluations of f (x) and have obtained the
corresponding values does not at all influence the numbers of elements of the sets XS1 and
YS2 . Moreover, it can happen that the number y = f (x) cannot be expressed in the numeral
system S2 and it is necessary to approximate it by a number ye ∈ S2 . This situation, very
well known to computer scientists, is represented in Fig. 7.1(b).
being one of the strongest sources of the interest to the Riemann zeta function. Many
interesting results have been established for this function (see [24, 36] for a comprehen-
sive discussion). In the context of this paper, the following results will be of our primary
interest.
Traditionally, it is said that ζ(s) diverges on the open half-plane of s such that the real
part ℜ(s) < 1. Then, for s = 0 the following value is attributed to ζ(0)
∞
1 1
ζ(0) = ∑ u0 = 1 + 1 + 1 + . . . = − 2 (8.3)
u=1
It has been shown for the Riemann zeta function that it has trivial zeros at the points
−2, −4, . . . It is also known that any non-trivial zero lies in the complex set {s : 0 < ℜ(s) <
1}, called the critical strip. The complex set {s : ℜ(s) = 1/2} is called the critical line.
The Riemann Hypothesis asserts that any non-trivial zero s has the real part ℜ(s) = 1/2,
i.e., lies on the critical line.
In this section, we study the Riemann zeta function in the framework of the ¬-based
approach using strongly results from sections 6.2 and 6.3 discussing how divergent series
are substituted by sums with a fixed infinite number of addends.
The first observation that can be done consists of the fact that thanks to the intro-
duction of ¬ within ∞ many different infinite numbers can be distinguished. Therefore,
records (8.1) and (8.5) do not describe two single functions and there exist many different
Riemann zeta functions and many different Dirichlet eta functions and to fix a concrete
one it is necessary to fix the number of addends in the corresponding sum (see section 6.2).
In order to define a Riemann zeta function (a Dirichlet eta function), we should choose an
infinite number n expressible in a numeral system using grossone and then write
n
1 n
(−1)u−1
ζ(s, n) = ∑ us , η(s, n) = ∑ us
. (8.6)
u=1 u=1
As a result, for different (infinite or finite) values of n we have different functions. For
example, for n = ¬/2 and n = ¬ we get
¬/2 ¬
1 1
ζ(s, ¬/2) = ∑ us , ζ(s, ¬) = ∑ us
u=1 u=1
(8.1), (8.5) on the one hand and (8.6), on the other hand, describe the same mathematical
objects – sums with an infinite number of addends – but do it with different accuracies.
Notice that there obviously exist questions for which both mathematical languages
used in (8.1) and (8.6) are sufficiently accurate. This can happen when the property we
ask about holds for all (finite and infinite) values of n in (8.6). For instance, it is possible
to answer in the affirmative to the question: ‘Is the inequality ζ(1) > −100 correct?’
because this result holds for all values of n in (8.6) and, therefore, the accuracy of the
record (8.1) allows us to answer in this case.
8.1. Some Euler’s results revisited in the ¬-based framework. Let us consider some
of results of Euler related to the Riemann zeta function (8.1) from the new methodological
positions in order to re-write them (where it is possible) for the form (8.6) with concrete
infinite values of n. First, the relation (8.4) can be re-written as
for n = 2k and as
η(s, n) = ζ(s, 2k + 1) − 21−s ζ(s, k) (8.8)
for n = 2k + 1.
Let us look now at the formula (8.2) that was introduced and studied by Euler (see
[37, 38, 39, 40]) who proved it for integer values of s. This identity has been proved using
the traditional language and, as a consequence, its accuracy is not sufficient to distinguish
within ∞ different infinite numbers. If one uses the ¬-based language, then identities
involving infinite sums and/or products of the type
∞ ∞
∑ ai = ∏ bi (8.9)
u=1 i=1
where both n and k are fixed (and possibly different) infinite numbers. As a consequence,
proofs of such identities should be reconsidered with the accuracy imposed by the new
Numerical infinities and infinitesimals 79
language. It can easily happen that a result, while being correct with the accuracy of one
language is not sufficiently precise when a language with a higher accuracy is used (see
discussion on the accuracy of the numeral system of Pirahã in the bottom of page 8).
The introduction of numerals 3 and 4 allows us to abandon the results 2+1=‘many’,
2+2=‘many’ and to see that 2 + 1 ̸= 2 + 2. Analogously, the passage from the low accu-
racy induced by the symbol ∞ to more accurate ¬-based numerals can lead to situations
where the correctness of (8.9) does not imply the correctness of (8.10).
Let us compare the usage of the traditional language working with ∞ and the new one
using ¬ with respect to the Euler product formula (8.2). To prove it, Euler expands each
of the factors of the right-hand part of (8.2) as follows
1 1 1 1
= 1+ s + 2 s + 3 s +... (8.11)
1 − ps
1 p (p ) (p )
Then he observes that their product is therefore a sum of terms of the form
1
, (8.12)
(pn11 · pn22 · . . . · pnr r )s
where p1 , p2 , . . . pr are distinct primes and n1 , n2 , . . . nr are natural numbers. Euler then
uses the fundamental theorem of arithmetic (every integer can be written as a product of
primes) and concludes that the sum of all items (8.12) is the left-hand part of (8.2).
In the new language this way of doing things is not acceptable because, in order to
start, we should indicate the precise infinite number, n, of items in the sum in the left-hand
part of (8.2). This operation gives us different functions ζ(s, n) from (8.6). Then (8.11)
should be rewritten by indicating the exact infinite number, k, of items in its right hand
part and the result of summing up will be different with respect to (8.2). Namely, we have
1 1 1 1 1 − (p1k )s
1+ + + + . . . + = . (8.13)
ps (p2 )s (p3 )s (pk−1 )s 1 − p1s
The accuracy of the language used by Euler did not allow him either to observe different
infinite values of n and k or to take into account the infinitesimal value (p1k )s . His results
are correct but their accuracy is low.
Another metaphor that can help is the following. Suppose that we have measured two
distances A and B with the accuracy equal to 1 meter and we have found that both of them
are equal to 25 meters. Suppose now that we want to measure them with the accuracy
equal to 1 centimeter. Then, very probably, we shall obtain something like A = 2487
centimeters and B = 2538 centimeters, i.e., A ̸= B. Both answers, A = B and A ̸= B, are
correct but with different accuracies and both of them can be used successfully in different
situations. For instance, if one just wants to go for a walk, then the accuracy of the answer
A = B expressed in meters is sufficient. However, if one needs to connect some devices
with a cable, then a higher accuracy is required and the answer expressed in centimeters
should be used.
We are with the Euler product formula in the same situation. It is correct with the
accuracy of the language using ∞ because this language does not allow one to distinguish
80 Yaroslav D. Sergeyev
different infinite numbers within ∞. At the same time, when a language allows us to
distinguish different infinite and infinitesimal numbers, it follows from (8.13) that for any
infinite (or finite) k we have
1 1 1 1 1
1+ + 2 s + 3 s + . . . + k−1 s ̸= (8.14)
ps (p ) (p ) (p ) 1 − p−s
and, therefore, the following theorem holds (a similar result obtained using a different
reasoning can be found in [114]).
Theorem 8.1. For prime pi and both finite and infinite values of n and k it follows that
n k
1 1
∑ us ∏ 1 − p−s .
̸
= (8.15)
u=1 i=1 i
Let us comment now upon another famous result of Euler related to the Riemann zeta
function – his solution to the Basel problem where he has shown that
1 1 1 π2
1+ 2
+ 2 + 2 +... = . (8.16)
2 3 4 6
Let us first briefly present Euler’s proof and then comment upon it. Note that from the
point of view of modern Mathematics this proof is amenable to criticism from different
points of view. Nowadays there exist many other proofs considered by mathematicians to
be more accurate14 . However, since also in modern proofs the symbol ∞ and the concept
of series are used, the analysis made below can be applied to these proofs, as well.
Euler begins with the standard Taylor expansion of sin(x),
x3 x5 x 7
sin(x) = x − + − +..., (8.17)
3! 5! 7!
which converges for all x. Euler interprets the right-hand side of (8.17) as an ‘infinite
polynomial’ P(x) that, therefore, can be written as product of factors based on its roots.
Since the roots of sin(x) are . . . − 3π, −2π, −π, 0, π, 2π, 3π, . . . then it follows
x2 x2 x2
sin(x) = x(1 − )(1 − )(1 − )... (8.18)
π2 22 π2 32 π2
14 This fact is another manifestation of the continuous mutation of mathematical languages. The views on
accuracy of proofs have changed since Euler’s times and the modern language is considered to be more accurate.
Numerical infinities and infinitesimals 81
Then, as follows from (8.17) and (8.18), Euler has written P(x) in two different ways and
he equates these two records
x3 x5 x 7 x2 x2 x2
sin(x) = x − + − + . . . = x(1 − 2 )(1 − 2 2 )(1 − 2 2 ) . . . , (8.19)
3! 5! 7! π 2 π 3 π
Now Euler equates the coefficients of x3 on both sides of (8.18) and gets first
1 1 1 1 1
− = − 2 − 2 2 − 2 2 − 2 2 ...,
3! π 2 π 3 π 4 π
and then the final beautiful result (8.16).
Let us now consider these results using the ¬-based approach. We take the function
sin(x) introduced using the standard trigonometric reasoning. Note that such a definition
just describes its properties and does not tell us how to calculate sin(x) precisely at all x.
When one uses the trigonometric definition, values of sin(x) only at certain x are known
precisely, e.g., sin( π2 ) = 1, sin(2π) = 0, etc. Notice that in these records we do not use
any approximation of the number π. Similarly to sin(x), it is described by its properties
and a special numeral, π, is introduced to indicate it. Then, those values of sin(x) that are
not linked to geometric ideas are defined through various approximations of both sin(x)
and π.
The ¬-based numerals taken together with the trigonometric definition of sin(x) allow
us to evaluate sin(x) precisely not only at certain finite points but also at certain infinite
points. For instance, it follows that sin(2¬π) = 0, sin(¬π + π2 ) = 1, etc.
Then, the explicit usage of infinite and infinitesimal numbers requires that we move
from (8.17) to the approximation with the polynomial P1 (x, 2k + 1) of the order 2k + 1
where
x3 x 5 x7 x2k+1
sin(x) ≈ P1 (x, 2k + 1) = x − + − + . . . + (−1)k , (8.20)
3! 5! 7! (2k + 1)!
and for different finite or infinite k we get different approximations. Analogously, the idea
used in (8.18) gives us the second kind of approximation where the polynomial P2 (x, 2n +
1) of the order 2n + 1 is used
sin(x) ≈ P2 (x, 1) = x,
2 2 2 (8.21)
sin(x) ≈ P2 (x, 2n + 1) = x(1 − πx 2 )(1 − 22xπ2 ) . . . (1 − n2xπ2 ),
10
8 P (x,13)
2
−2
−4
−6
P1(x,13)
−8
−10
−20 −15 −10 −5 0 5 10 15 20
Figure 8.1. Polynomials P1 (x, 13) and P2 (x, 13) together with sin(x).
However, such a kind of equating should be done for all the coefficients of P1 (x, 2k + 1)
and P2 (x, 2n + 1). Suppose that k in the sum (8.20) is even, by equating the coefficients
of the highest power 2k + 1 of x in two polynomials we get that it should be
1 1
= ,
(2k + 1)! (k!)2 π2k
whence we deduce
π2k 1
= . (8.23)
(2k + 1)! (k!)2
Equating coefficients for other powers of x will give us several different expressions to be
satisfied and it is easy to see that they cannot all simultaneously hold.
Thus, polynomials P1 (x, 2k + 1) and P2 (x, 2n + 1) give us different approximations of
sin(x). As an example we show in Fig. 8.1 polynomials P1 (x, 13) and P2 (x, 13) together
with sin(x). The higher is the order of the polynomials the better are the approximations
but the differences are always present for both finite and infinite values of k and n.
As it was with the product formula, Euler results are correct with the accuracy of the
traditional mathematical language working only with the symbol ∞. Such language does
not allow one to distinguish different infinite values of k and n and, as a consequence,
different infinite polynomials P1 (x, 2k + 1) and P2 (x, 2n + 1).
Since all the traditional computations are executed with finite values of k and n and
only the initial finite part of (8.23) is used, it is always possible to choose sufficiently
high finite values of k and n yielding valid approximations of π with a finite accuracy.
We are not able by using the traditional mathematical tools to observe the behavior of the
polynomials and sin(x) at various points x at infinity. Thus, the fact of the difference of
Numerical infinities and infinitesimals 83
these objects in infinity cannot be detected by finite numeral systems extended only by the
symbol ∞. However, it is perfectly visible when one uses the new numeral system where
different infinite and infinitesimal numbers can be expressed and distinguished. Thus, we
can conclude this subsection by the following theorem that has just been proved.
Theorem 8.2. For both finite and infinite values of n and k it follows that
8.2. Calculating ζ(s, n) for different infinite values of n at s = 0, −1, −2, etc. In this
subsection, we establish some results for functions ζ(s, n) and η(s, n) from (8.6) using the
approach presented in sections 6.2 and 6.3 where it was shown how to move from series
to sums with a fixed infinite number of addends. The analysis made there allows us to
calculate ζ(s, n) and η(s, n) for infinite values of n at s = 0 and s = −1:
ζ(0, n) = 1 + 1 + 1 + . . . + 1 = n,
| {z }
n addends
(n + 1)n
ζ(−1, n) = 1 + 2 + 3 + . . . + n = .
2
For instance, at n = ¬/2 and n = ¬ we obtain
that gives, for example, η(0, ¬) = 0 and η(0, ¬2 − 1) = 1. In section 6.3 we have seen
also how it is possible to calculate η(−1, n) for infinite values of n. For instance, it follows
that η(−1, ¬) = − ¬ 2 . Notice that these results fit well with relation (8.7). For example,
for n = ¬ we have
ζ(0, ¬) − 2ζ(0, ¬/2) = 0 = η(0, ¬),
¬
ζ(−1, ¬) − 4ζ(−1, ¬/2) = − = η(−1, ¬).
2
Let us present now a general method for calculating values of ζ(s, n) and η(s, n) for
infinite or finite n at s being negative finite integers. For this purpose we use the idea of
Euler to multiply the identity
n
f (x) = ∑ xi = 1 + x + x2 + . . . + xn , (8.25)
i=0
84 Yaroslav D. Sergeyev
by x. By using the differential calculus developed in [91] we can apply this idea without
distinction for both finite and infinite values of n and x ̸= 1 that can be finite, infinite or
infinitesimal. By differentiation of (8.25) we obtain
n−1
f ′ (x) = ∑ ixi−1 = 1 + 2x + 3x2 + . . . + nxn−1 , (8.26)
i=0
1 − xn+1
f (x) = (8.27)
1−x
then by differentiation of (8.27) we have that
1 + nxn+1 − (n + 1)xn
f ′ (x) = . (8.28)
(1 − x)2
Thus, we can conclude that for finite and infinite n and for finite, infinite or infinitesimal
x ̸= 1 it follows that
1 + n(−1)n+1 − (n + 1)(−1)n
η(−1, n) = 1 − 2 + 3 − 4 + . . . + n(−1)n−1 = .
4
For example, for n = ¬ we obtain the result that has been already calculated in section 6.3
in a different way
η(−1, ¬) = 1 − 2 + 3 − 4 + . . . + ¬(−1)¬−1 =
1 + ¬(−1)¬+1 − (¬ + 1)(−1)¬ 1 − ¬ − (¬ + 1) ¬
= =− .
4 4 2
If we multiply successively (8.29) by x and use again differentiation of both parts of
the obtained equalities then it becomes possible to obtain the values of η(s, n) for other
finite integer negative points s. Thus, in order to obtain, for instance, formulae for s =
−2, −3, −4, we proceed as follows
(6n4 + 12n3 − 6n2 − 12n + 11)xn+2 − (4n4 + 12n3 + 6n2 − 12n − 11)xn+1 +
(n + 1)4 xn − x3 − 11x2 − 11x − 1)(x − 1)−5 . (8.32)
Taking x = −1 in (8.30), (8.31), and (8.32) we obtain that
1 ( ) −4α ]
n(n + 1)(n + 2)¬−3α + n+2 4 ¬ + . . . + ¬−(n+2)α +
6
[ 1 1
(2n2 + 2n − 1) 1 + (n + 1)¬−α + n(n + 1)¬−2α + (n − 1)n(n + 1)¬−3α +
2 6
(n+1) −4α ] [ 1
4 ¬ + . . . + ¬−(n+1)α − (n + 1)2 1 + n¬−α + (n − 1)n¬−2α +
2
1 ( ) ] )
(n − 2)(n − 1)n¬−3α + n4 ¬−4α + . . . + ¬−nα + ¬−α + 2 .
6
By collecting the terms of grossone we then obtain
1
f (1 + ¬−α , n) = n(n + 1)(2n + 1)+
6
( ( ) (n+1) ( )) −α
4 − (2n + 2n − 1) 4 + (n + 1) 4 ¬
n2 n+2 2 2 n
+ . . . + n2 ¬−(n+2)α . (8.36)
As it can be seen from (8.36), for any finite or infinite value of n there always can be
chosen a number α > 0 such that the contribution of the added infinitesimal ¬−α in f (1 +
¬−α , n) is a sum of infinitesimals (see the second line of (8.36)). Due to the representation
(4.16), (4.17), this contribution can be easily separated from the integer finite or infinite
part represented by the first line of (8.36).
Thus, we have obtained that for finite and infinite values of n it follows
1
ζ(−2, n) = n(n + 1)(2n + 1). (8.37)
6
Analogously, by applying the same procedure again we can obtain the formulae for ζ(s, n)
for other finite integer negative points s. For instance,
1
ζ(−3, n) = n2 (n + 1)2 , (8.38)
4
1
ζ(−4, n) = n(n + 1)(2n + 1)(3n2 + 3n − 1). (8.39)
30
Note that the previous results fit perfectly both well-known formulae for finite values
of n (see [8]) and relation (8.7). For example, using (8.7) and formulae (8.37), (8.38), and
(8.39) with n = ¬ and n = ¬/2 we obtain (cf. (8.33), (8.34), and (8.35)) that
1 1
ζ(−2, ¬) − 8ζ(−2, ¬/2) = ¬(¬ + 1)(2¬ + 1) − ¬(¬ + 2)(¬ + 1) =
6 3
−0.5¬(¬ + 1) = η(−2, ¬),
1
ζ(−3, ¬) − 16ζ(−3, ¬/2) = ¬2 (¬ + 1)2 − ¬2 (0.5¬ + 1)2 =
4
−0.5¬2 (¬ + 3) = η(−3, ¬),
1
ζ(−4, ¬) − 32ζ(−4, ¬/2) = ¬(¬ + 1)(2¬ + 1)(3¬2 + 3¬ − 1)−
30
Numerical infinities and infinitesimals 87
1
¬(¬ + 2)(¬ + 1)(3¬2 + 6¬ − 4) = −0.5¬(¬ + 1)(¬2 + ¬ − 1) = η(−4, ¬).
15
The analysis done in this section shows that the traditional mathematical language
using the symbol ∞ very often does not possess a sufficiently high accuracy when one
deals with problems having their interesting properties at infinity. For instance, it does
not allow us to distinguish within the record f (x) = ∑∞ i=1 ai (x) different functions f n (x) =
∑ni=1 ai (x) emerging for different infinite n. However, functions fn (x) become visible if
one uses more powerful numeral systems allowing one to represent different infinite (and
infinitesimal) numbers. Then, it follows that if an+1 (x) ̸= 0 functions fn (x) and fn+1 (x)
are different and fn (x) ̸= fn+1 (x). When an+1 (x) is an infinitesimal number then the
difference fn (x) − fn+1 (x) is also infinitesimal, i.e., invisible if one uses the traditional
mathematical language but perfectly visible through the ¬-based numeral system.
In particular, we have seen that the Riemann Hypothesis asks a question about the
behavior of a function defined traditionally using the symbol ∞. As we can see using the
¬-based methodology and the respective numeral system distinguishing different infinite
numbers, the Riemann zeta function is not a function but many different functions that
are indistinguishable if one uses the traditional mathematics. The analysis of the Rie-
mann zeta function made in the traditional style does not consider various infinite and
infinitesimal numbers that are crucial not only in the context of the Hypothesis but even
with respect to the definition of the function itself. It has been shown that several classical
results do not represent properties of the the Riemann zeta function but reflect defects of
the numeral system using the symbol ∞. Of course, there remains a possibility that the
question asked in the Hypothesis has the same answer for all infinite n. Unfortunately, this
is impossible due to the analysis made above and the analysis of the partial zeta functions
with a finite number of addends made in [4, 11, 108].
We conclude this subsection by the following remark. It is well known that the Rie-
mann zeta function appears in many different mathematical contexts and there exist many
different equivalent formulations of the Riemann hypothesis. In view of the above analy-
sis, this can be explained again by the accuracy of the traditional mathematical language.
Such phenomena indicate situations where its accuracy is not sufficient. By returning
to the metaphor of the microscope, we can say that out traditional lens is too weak to
distinguish different mathematical objects observed in these situations. Thus, these math-
ematical contexts can be viewed as very promising for obtaining new results by applying
the new numeral system that does not use the symbol ∞ and allows one to distinguish and
to treat numerically various infinite and infinitesimal numbers.
Figure 9.1. What is the probability that the rotating disk stops in such a way that the point A will be
exactly in front of the arrow?
where h is an arc of the circumference containing A and 2πr is its length. However, the
point A can stop in front of the arrow, i.e., this event is not impossible and its probability
should be strictly greater than zero, i.e., P(E) > 0.
In order to consider this example in the ¬-based framework, let us redescribe the
experiment more rigorously by choosing a numeral system that will be used to express
points on the circumference. This choice will fix the number (infinite or finite) of points,
K, that we are able to distinguish on the circumference. This definition of the notion point
allows us to define elementary events in our experiment as follows: the disk has stopped
and the arrow indicates a point. As a consequence, we obtain that the number, N(Ω), of
N(Ω)
all possible elementary events, ei , in our experiment is equal to K where Ω = ∪i=1 ei is
the sample space of our experiment. If our disk is well balanced, all elementary events
1
are equiprobable and, therefore, have the same probability equal to N(Ω) . Thus, we can
calculate P(E) directly by subdividing the number, N(E), of favorable elementary events
by the number, K = N(Ω), of all possible events.
For example, if we use numerals of the type i , i ∈ N, then K = ¬. The number
¬
N(E) depends on our decision about how many numerals we want to use to represent the
point A. If we decide that the point A on the circumference is represented by m numerals,
we obtain
N(E) m m
P(E) = = = > 0.
N(Ω) K ¬
Numerical infinities and infinitesimals 89
where the number m is infinitesimal if m is finite. Note that this representation is inter-
¬
esting also from the point of view of distinguishing the notions ‘point’ and ‘arc’. When m
is finite then we deal with a point, when m is infinite we deal with an arc.
In case we need a higher accuracy, we could choose, for instance, numerals of the
type i¬−2 , 1 ≤ i ≤ ¬2 , for expressing points at the disk. Then it follows K = ¬2 and, as
a result, we obtain P(E) = m¬−2 > 0.
This example with the rotating disk, of course, is a particular instance of the general
situation related to continuous random variables and probabilities of events corresponding
to sets with measure zero. While for a discrete random variable one could say that an event
with probability zero is impossible, this can not be said traditionally in the continuous
case where to events corresponding to sets having measure zero the probability zero is
assigned. As we have shown by the example above, the ¬-based methodology allows us to
model the continuous case using infinitesimals and to avoid appearance of sets of measure
zero. Thus, in both discrete and continuous cases the only event having probability zero
is the impossible event.
Notice also that the obtained probabilities are not absolute, they depend on the ac-
curacy chosen to express points on the circumference. There is again a straight analogy
with Physics where it is not possible to obtain results that have a precision higher than
the accuracy of the instrument used to measure data. With respect to probability, it is not
possible to get a precision of computations that is higher than the precision of numerals
chosen to represent elementary events. Thus, the introduction of the concept of accuracy
and availability of ¬-based infinitesimals allow us to avoid appearance of non-empty sets
having measure zero and to introduce the concept of relativity of probability that depends
on the chosen numeral system.
9.2. Koch snowflake and its quantitative analysis using ¬-based numerals. Fractal
objects have been very well studied during the last few decades (see, e.g., [41, 78] and ref-
erences given therein) and have been applied in various fields (see numerous applications
given in [35, 41, 50, 78, 83, 85]). However, the mathematical analysis of fractals very
often continues to have mainly a qualitative character and tools for a quantitative analysis
of fractals at infinity are not very rich yet. Since fractals are objects defined as a limit of
an infinite process, the computation of their dimension is one of a very few quantitative
characteristics that can be calculated at infinity.
Traditionally, after n iterations of a fractal process it is possible to give numerical
answers to questions regarding fractals (calculation of, e.g., their length, area, volume
or the number of smaller copies of initiators present at the n-th iteration) only for finite
values of n. The same questions very often remain without any answer when we consider
an infinite number of steps because when we speak about limit fractal objects the required
values often either tend to zero and disappear, in practice, or tend to infinity, i.e., become
intractable numerically. Moreover, we cannot distinguish at infinity fractals starting from
similar initiators even though they are different for any fixed finite value of the iteration
number n.
In this subsection, we propose to apply the ¬ methodology to the study of fractals.
In particular, the Koch snowflake being one of the first mathematically described fractals
90 Yaroslav D. Sergeyev
a) b)
c) d)
e) f)
is considered (for more texts on applying the ¬ methodology to traditional and blinking
fractals see [15, 87, 89, 97, 102]; see also papers [68, 69, 70] related to tilings of the
hyperbolic plane and grossone). The snowflake has been introduced by Helge von Koch
in 1904 (see [61]). This fractal is interesting because it is known that in the limit it has
an infinite perimeter but its area is finite. The procedure of its construction is shown in
Fig. 9.2. The initiator (iteration number n = 0) is the triangle shown in Fig. 9.2, a), more
precisely, its three sides. Then each side (segment) is substituted by four smaller segments
as it is shown in Fig. 9.2, b) during the first iteration, in other words, a smaller copy of
the triangle is added to each side. At the iteration n = 2 (see Fig. 9.2, c)) each segment
is substituted again by four smaller segments, an so on. Thus, the Koch snowflake is
the resulting limit object obtained at n → ∞. It is known (see, e.g., [78]) that its fractal
dimension is equal to log 4
log 3 ≈ 1.26186.
Clearly, for finite values of n we can calculate the perimeter Pn and the respective area
An of the snowflake. If iteration numbers n and k are such that n ̸= k then it follows Pn ̸= Pk
and An ̸= Ak . Moreover, the snowflake started from the initiator a) after n iterations will
Numerical infinities and infinitesimals 91
be different from the snowflake started from the configuration b) after n iterations. The
simple illustration for n = 3 can be viewed in Fig. 9.2. In fact, starting from the initiator
a) after three iterations we have the snowflake d) and starting from the initiator b) after
three iterations we have the snowflake e).
Unfortunately, the traditional analysis of fractals does not allow us to have quantitative
answers to the questions stated above when n → ∞. In fact, we know only (see, e.g., [78])
that √
4n 8 3 2
lim Pn = lim n−1 l = ∞, lim An = a0 , a0 = l , (9.1)
n→∞ n→∞ 3 n→∞ 5 4
where a0 is the area of the original triangle from Fig. 9.2, a) expressed in the terms of its
side length l.
By using the ¬-based methodology a more precise quantitative analysis of the Koch
snowflake can be done at infinity. In particular, it becomes possible:
– to show that at infinity the Koch snowflake is not a unique object, namely, different
snowflakes can be distinguished at infinity similarly to different snowflakes that
can be distinguished for different finite values of n;
– to calculate the exact (up to infinitesimals) perimeter Pn of the snowflake (together
with the infinite number of sides and the infinitesimal length of each side) after n
iterations for different infinite values of n;
– to show that for infinite n and k such that k > n it follows that both Pn and Pk are
infinite but Pk > Pn and their difference Pk − Pn can be computed exactly and it
turns out to be infinite;
– to calculate the exact (up to infinitesimals) finite areas An and Ak of the snowflakes
for infinite n and k such that k > n and to show that it follows Ak > An , the difference
Ak − An can be also calculated exactly and it turns out to be infinitesimal;
– to show that the snowflakes constructed starting from different initiators (e.g., from
initiators shown in Fig. 9.2) are different after k iterations where k is an infinite
number, i.e., they have different infinite perimeters and different areas where the
difference can be measured using infinities and infinitesimals.
Let us begin our quantitative analysis. It can be seen from Fig. 9.2 that at each iteration
each side of the snowflake is substituted by 4 new sides having the length of one third of
the segment that has been substituted. Thus, if we indicate as Nn , n ≥ 1, the number of
segments of the snowflake and as Ln their length at the n-th iteration then
1 l
Nn = 4Nn−1 = 3 · 4n , Ln = Ln−1 = n , n > 1, (9.2)
3 3
where l is the length of each side of the original triangle from Fig. 9.2. As a result, the
perimeter Pn of the Koch snowflake is calculated as follows
4 4 4n
Pn = Nn · Ln = Nn−1 · Ln−1 = Pn−1 = n−1 l. (9.3)
3 3 3
Therefore, if we start our computations from the original triangle from Fig. 9.2, after ¬
steps we have the snowflake having the infinite number of segments N¬ = 3 · 4¬ . Each of
92 Yaroslav D. Sergeyev
the segments has the infinitesimal length L¬ = 31¬ l. In order to calculate the perimeter,
P¬ , of the snowflake we should multiply the infinite number N¬ and the infinitesimal
number L¬ . Thus, the perimeter is
1 4¬
P¬ = N¬ · L¬ = 3 · 4¬ · l = l
3¬ 3¬−1
and it is infinite. Analogously, after executing ¬-1 steps we have the infinite perimeter
¬−1
P¬−1 = 43¬−2 l. Since the new numeral systems allows us to execute easily arithmetical
operations with infinite numbers, we can divide the obtained two infinite numbers, P¬ and
P¬−1 , one by another and to obtain as the result the finite number that is in a complete
agreement with (9.3)
4¬
P¬ 3¬−1
l 4
= ¬−1 = .
P¬−1 4 3
¬−2 l
3
The difference of the two perimeters can also be calculated easily
( )
4¬ 4¬−1 4¬−1 4 4¬−1
P¬ − P¬−1 = ¬−1 l − ¬−2 l = ¬−2 l − 1 = ¬−1 l
3 3 3 3 3
and it turns out to be infinite.
In case the infinite number of steps is n = 0.5¬, it follows that the infinite perimeter is
40.5¬
P0.5¬ = 30.5 ¬−1 l and the operation of division of two infinite numbers gives us as the result
also an infinite number that can be calculated precisely:
4¬ ( )0.5¬
P¬ 3¬−1
l 4
= = .
P0.5¬ 40.5¬ 3
30.5¬−1
l
Thus we can distinguish now in a precise manner that the infinite perimeter P¬ is infinitely
longer than the infinite perimeter P0.5¬ .
We can also distinguish at infinity the snowflakes having different initial generators.
As we have already seen, starting from the original triangle after ¬ steps the snowflake has
3 ·4¬ segments and each of them has the infinitesimal length L¬ = 31¬ l and the perimeter of
¬
the snowflake is P¬ = 3¬4−1 l. If we start from the initial configuration shown at Fig. 9.2, c)
and also execute ¬ steps then the resulting snowflake will have N¬+2 = 3 · 4¬+2 segments,
each of them will have the infinitesimal length L¬+2 = 3¬1+2 l and the perimeter of the
¬+2
snowflake will be P¬+2 = 43¬+1 l. Thus, this snowflake will have infinitely many more
segments than the one started from the original triangle. More precisely, this infinite
difference is equal to
N¬+2 − N¬ = 3 · 4¬+2 − 3 · 4¬ = 3 · 4¬ (42 − 1) = 45 · 4¬ .
The lengths of the segments in both snowflakes are infinitesimal and, in spite of the fact
that their difference is also infinitesimal, it can be calculated precisely as follows
( )
1 1 1 1 8l
L¬ − L¬+2 = ¬ l − ¬+2 l = ¬ l 1 − 2 = ¬+2 .
3 3 3 3 3
Numerical infinities and infinitesimals 93
Let us see now what happens with the area of the snowflake at infinity. As it can be
seen from Fig. 9.2, at each iteration n a new triangle is added at each side of the snowflake
built at iteration n − 1 and, therefore, the number of new triangles, Tn , is equal to
Tn = Nn−1 = 3 · 4n−1 .
The area, an , of each triangle added at n-th iteration is 91 of each triangle added during the
iteration n − 1
an−1 a0
an = = n,
9 9
where a0 is the area of the original triangle (see (9.1)). Therefore, the whole new area
added to the snowflake is
( )n ( )n−1
n−1 a0 3 4 a0 4
Tn an = 3 · 4 · n= · · a0 = · (9.4)
9 4 9 3 9
and the complete area, An , of the snowflake at the n-th iteration is
( ( ) ) ( ( ))
n
1 n 4 i−1 1 n−1 4 i
An = a0 + ∑ Tn an = a0 1 + ∑ = a0 1 + ∑ =
i=1 3 i=1 9 3 i=0 9
( ( )n ) ( ( ( )n )) ( ( )n )
1 1 − 49 3 4 a0 4
a0 1+ · = a0 1 + 1− = 8−3 .
3 1− 9 4 5 9 5 9
Traditionally, the limit of the area is considered and the result
8
lim An = a0
n→∞ 5
is obtained. Thanks to ¬-based numerals we are able now to work with infinitesimals
easily and to observe the infinitesimal difference of the areas of the snowflakes at infinity.
For example, after ¬-1 and ¬ iterations the difference between the areas A¬−1 and A¬ is
( ( )¬ ) ( ( )¬−1 ) ( )
a0 4 a0 4 a0 4 ¬−1
A¬ − A¬−1 = 8−3 − 8−3 = > 0.
5 9 5 9 3 9
This number is an infinitesimal and it perfectly corresponds to the general formula (9.4).
where the new methodology helps a lot have been discussed. Numerous examples and
illustrations have been provided.
It has been emphasized that the philosophical triad – researcher, object of investiga-
tion, and tools used to observe the object – existing in such natural sciences as Physics
and Chemistry, exists in Mathematics and Computer Science, too. In natural sciences,
the instrument used to observe the object influences the results of observation. The same
happens in Mathematics and Computer Science, where numeral systems used to express
numbers are among the instruments of observations used by mathematicians. The usage
of powerful numeral systems yields more precise results in Mathematics, in the same way
as the usage of a good microscope gives the possibility to obtain more precise results in
Physics. When a mathematician chooses a numeral system (an instrument), in this mo-
ment he/she chooses both a set of numbers that can be observed through the numerals
available in the chosen numeral system and the accuracy of results that can be obtained
during computations.
The analysis performed in this study shows that the traditional mathematical language
using symbols ∞, ω, ℵ0 , ℵ1 , etc. very often does not possess a sufficiently high accuracy
when one deals with problems having their interesting properties at infinity. This lack
of accuracy can lead to paradoxes and problems that are considered to be very hard in
traditional Mathematics. It has been emphasized that their resistance to solution is often
a consequence of the weakness of numeral systems used traditionally. Numerous theo-
retical and applied problems considered here show that the ¬-based numeral system can
help avoid difficulties and paradoxes on several occasions and it gives rise not only to
powerful computational and theoretical techniques but also to the possibility to work with
mathematical objects that are invisible if one uses traditional tools.
In particular, the Continuum Hypothesis and the Riemann Hypothesis were studied.
It has been shown that the hardness of both problems is not related to their nature but is
induced by the weakness of traditional numeral systems used to formulate and study the
problems. The application of the ¬-based numeral system has allowed us to avoid tradi-
tional difficulties and to observe mathematical objects involved in the two problems in a
new light and with a higher accuracy. The effect of employing the ¬-based methodology
in the study of the above Hypotheses is comparable to the dissolution of computational
problems posed in Roman numerals (e.g., X – X) once a positional system capable of
expressing zero is adopted.
In conclusion, several pure mathematical and numerical applications discussed here
show that the introduction of the ¬-based methodology gives rise to a more accurate
description of a variety of mathematical objects, leads to the construction of new powerful
numerical methods, and provides answers to several problems that look exceptionally
challenging when traditional mathematical tools are applied.
References
[1] R.A. Adams. Single variable calculus. Pearson Education Canada, Ontario, 5
edition, 2003.
Numerical infinities and infinitesimals 95
[17] J.B. Carroll, editor. Language, Thought, and Reality: Selected Writings of Ben-
jamin Lee Whorf. MIT Press, 1956.
[18] J.R. Cash and S. Considine. An mebdf code for stiff initial value problems. ACM
Trans. Math. Software, 18(2):142–155, 1992.
[19] A.L. Cauchy. Le Calcul infinitésimal. Paris, 1823.
[20] M. Cococcioni, M. Pappalardo, and Ya.D. Sergeyev. Lexicographic multi-objective
linear programming using grossone methodology: Theory and algorithm. Applied
Mathematics and Computation, 318:298–311, 2018.
[21] J.S. Cohen. Computer Algebra and Symbolic Computation: Mathematical Meth-
ods. A K Peters, Ltd., Wellesley, MA, 1966.
[22] P.J. Cohen. Set Theory and the Continuum Hypothesis. Benjamin, New York, 1966.
[23] M. Colyvan. An Introduction to the Philosophy of Mathematics. Cambridge Uni-
versity Press, Cambridge, 2012.
[24] J. Brian Conrey. The Riemann hypothesis. Notices Amer. Math. Soc., 50(3):341–
353, 2003.
[25] J.H. Conway and R.K. Guy. The Book of Numbers. Springer-Verlag, New York,
1996.
[26] G. Corliss, C. Faure, A. Griewank, L. Hascoet, and U. Naumann, editors. Automatic
Differentiation of Algorithms: From Simulation to Optimization. Springer-Verlag,
New York, 2002.
[27] L. Corry. Numbers. Oxford University Press, Oxford, 2015.
[28] J. d’Alembert. Différentiel. Encyclopédie, ou dictionnaire raisonné des sciences,
des arts et des métiers, 4, 1754.
[29] L. D’Alotto. Cellular automata using infinite computations. Applied Mathematics
and Computation, 218(16):8077–8082, 2012.
[30] L. D’Alotto. A classification of two-dimensional cellular automata using infinite
computations. Indian Journal of Mathematics, 55:143–158, 2013.
[31] L. D’Alotto. A classification of one-dimensional cellular automata using infinite
computations. Applied Mathematics and Computation, 255:15–24, 2015.
[32] S. De Cosmis and R. De Leone. The use of grossone in mathematical programming
and operations research. Applied Mathematics and Computation, 218(16):8029–
8038, 2012.
[33] R. De Leone. Nonlinear programming and grossone: Quadratic programming
and the role of constraint qualifications. Applied Mathematics and Computation,
318:290–297, 2018.
[34] R. De Leone, G. Fasano, and Ya.D. Sergeyev. Planar methods and grossone for
the conjugate gradient breakdown in nonlinear programming. Computational Op-
timization and Applications, in press.
Numerical infinities and infinitesimals 97
[87] Ya.D. Sergeyev. Blinking fractals and their quantitative analysis using infinite and
infinitesimal numbers. Chaos, Solitons & Fractals, 33(1):50–75, 2007.
[88] Ya.D. Sergeyev. A new applied approach for executing computations with infinite
and infinitesimal quantities. Informatica, 19(4):567–596, 2008.
[89] Ya.D. Sergeyev. Evaluating the exact infinitesimal values of area of Sierpinski’s
carpet and volume of Menger’s sponge. Chaos, Solitons & Fractals, 42(5):3042–
3046, 2009.
[90] Ya.D. Sergeyev. Numerical computations and mathematical modelling with infi-
nite and infinitesimal numbers. Journal of Applied Mathematics and Computing,
29:177–195, 2009.
[91] Ya.D. Sergeyev. Numerical point of view on Calculus for functions assuming finite,
infinite, and infinitesimal values over finite, infinite, and infinitesimal domains.
Nonlinear Analysis Series A: Theory, Methods & Applications, 71(12):e1688–
e1707, 2009.
[92] Ya.D. Sergeyev. Computer system for storing infinite, infinitesimal, and finite quan-
tities and executing arithmetical operations with them. USA patent 7,860,914,
2010.
[93] Ya.D. Sergeyev. Counting systems and the First Hilbert problem. Nonlinear Anal-
ysis Series A: Theory, Methods & Applications, 72(3-4):1701–1708, 2010.
[94] Ya.D. Sergeyev. Lagrange Lecture: Methodology of numerical computations with
infinities and infinitesimals. Rendiconti del Seminario Matematico dell’Università
e del Politecnico di Torino, 68(2):95–113, 2010.
[95] Ya.D. Sergeyev. Higher order numerical differentiation on the Infinity Computer.
Optimization Letters, 5(4):575–585, 2011.
[96] Ya.D. Sergeyev. On accuracy of mathematical languages used to deal with the
Riemann zeta function and the Dirichlet eta function. p-Adic Numbers, Ultrametric
Analysis and Applications, 3(2):129–148, 2011.
[97] Ya.D. Sergeyev. Using blinking fractals for mathematical modelling of processes
of growth in biological systems. Informatica, 22(4):559–576, 2011.
[98] Ya.D. Sergeyev. Solving ordinary differential equations by working with infinites-
imals numerically on the Infinity Computer. Applied Mathematics and Computa-
tion, 219(22):10668–10681, 2013.
[99] Ya.D. Sergeyev. Computations with grossone-based infinities. In Calude C.S.
and Dinneen M.J., editors, Unconventional Computation and Natural Computa-
tion: Proc. of the 14th International Conference UCNC 2015, volume LNCS 9252,
pages 89–106. Springer, New York, 2015.
[100] Ya.D. Sergeyev. The Olympic medals ranks, lexicographic ordering, and numerical
infinities. The Mathematical Intelligencer, 37(2):4–8, 2015.
[101] Ya.D. Sergeyev. Un semplice modo per trattare le grandezze infinite ed infinites-
ime. Matematica nella Società e nella Cultura: Rivista della Unione Matematica
Italiana, 8(1):111–147, 2015.
Numerical infinities and infinitesimals 101
[102] Ya.D. Sergeyev. The exact (up to infinitesimals) infinite perimeter of the Koch
snowflake and its finite area. Communications in Nonlinear Science and Numerical
Simulation, 31(1–3):21–29, 2016.
[103] Ya.D. Sergeyev and A. Garro. Observability of Turing machines: A refinement of
the theory of computation. Informatica, 21(3):425–454, 2010.
[104] Ya.D. Sergeyev and A. Garro. Single-tape and multi-tape Turing machines through
the lens of the Grossone methodology. Journal of Supercomputing, 65(2):645–663,
2013.
[105] Ya.D. Sergeyev, M.S. Mukhametzhanov, F. Mazzia, F. Iavernaro, and P. Amodio.
Numerical methods for solving initial value problems on the Infinity Computer.
International Journal of Unconventional Computing, 12(1):3–23, 2016.
[106] S. Shelah. Cardinal Arithmetic, volume 29 of Oxford Logic Guides. Oxford Uni-
versity Press, 1994.
[107] H.D. Sherali and A.L. Soyster. Preemptive and nonpreemptive multi-objective pro-
gramming: Relationship and counterexamples. Journal of Optimization Theory and
Applications, 39(2):173–186, 1983.
[108] R. Spira. Zeros of sections of the zeta function. I. Mathematics of Computation,
20(96):542–550, 1966.
[109] I.P. Stanimirovic. Compendious lexicographic method for multi-objective opti-
mization. Facta universitatis - series: Mathematics and Informatics, 27(1):55–66,
2012.
[110] M.C. Vita, S. De Bartolo, C. Fallico, and M. Veltri. Usage of infinitesimals in
the Menger’s Sponge model of porosity. Applied Mathematics and Computation,
218(16):8187–8196, 2012.
[111] P. Vopěnka and P. Hájek. The Theory of Semisets. Academia Praha/North Holland
Publishing Company, Praha, 1972.
[112] J. Wallis. Arithmetica infinitorum. 1656.
[113] W. H. Woodin. The Continuum Hypothesis, Part I. Notices of the AMS, 48(6):567–
576, 2001.
[114] A. Zhigljavsky. Computing sums of conditionally convergent and divergent se-
ries using the concept of grossone. Applied Mathematics and Computation,
218(16):8064–8076, 2012.
[115] A. Zhigljavsky and V. Kornikov. Classical areas of mathematics where the concept
of grossone could be useful. In Sergeyev Ya.D., Kvasov D.E., Dell’Accio F., and
Mukhametzhanov M.S., editors, Proc. of the 2nd Intern. Conf. “Numerical Com-
putations: Theory and Algorithms”, volume 1776, page 020004. AIP Publishing,
New York, 2016.
[116] A. Žilinskas. On strong homogeneity of two global optimization algorithms based
on statistical models of multimodal objective functions. Applied Mathematics and
Computation, 218(16):8131–8136, 2012.
102 Yaroslav D. Sergeyev