Professional Documents
Culture Documents
Abstract Recent educational studies in mathematics seek to justify a thesis that there
is a conflict between students intuitions regarding infinity and the standard theory of
infinite numbers. On the contrary, we argue that students intuitions do not match but to
Cantor’s theory, not to any theory of infinity. To this end, we sketch ways of measuring
infinity developed at the turn of the 20th and 21st centuries that provide alternatives
to Cantor’s theory of cardinal and ordinal numbers. Some of them introduce new kinds
of infinite numbers, others simply define new arithmetic for Cantor’s infinite numbers.
With regard to these new approaches, we argue that there are various intuitions of actual
infinity which can find an adequate theory.
We also present pre-Cantorain theories of actual infinity developed within the tradition
of geometrical optics. They corroborate our claim on various intuitions concerning actual
infinity.
2 Gödel’s Fallacy
Kurt Gödel reinforced the belief that there is no alternative to Cantor’s theory of infinite
numbers. Moreover, he claimed that when one adopts Cantor’s infinite numbers, there
is no room for an alternative arithmetic of these numbers. In (Gödel, 1947), he presents
cardinal numbers as extending the system of natural numbers (N, +, ·, 0, 1, <) and seeks
to show that “this extension can be effected in a uniquely determined manner”. To this
end, he discusses (1) definition, (2) equality, (3) total order, and (4) sums and products
of cardinal numbers.
(Ad 1) Gödel assumes that the very definition of infinite number has to be derived from
2
Sometimes it happens that a definition is founded on a theorem. Yet, it is not the case of the definitions
of infinite numebers and their arithmetic.
2
the notion of infinite set. (Ad 2, 3) He claims that “there is hardly any choice left but to
accept Cantor’s definition of equality between numbers, which can easily be extended to
a definition of ‘greater‘ and ‘less’ for infinite numbers”. Here, the equality is based on the
one-to-one correspondence, and the order builds on the subset relation. (Ad 4) Finally,
Gödel concludes: “it becomes possible to extend (again without any arbitrariness) the
arithmetical operations to infinite numbers (including sums and products with any infinite
number of terms or factors) and to prove practically all ordinary rules of computation”.
Arguably, “practically all rules” does not include compatibility of sums and products with
the order: although 2 < 3, it is not the case that 2 + ℵ0 < 3 + ℵ0 or 2 · ℵ0 < 3 · ℵ0 .
In his presentation, Gödel skips explicit reference to ordinal numbers even when he
defines alephs, though he employs ordinal numbers for indexing alephs. In this way, ordinal
numbers that could be considered as a way to reinterpret the arithmetic of cardinal
numbers, were pushed down to seemingly inferior role of indexes. In fact, a modified
arithmetic of ordinal numbers opens a new perspective on Cantor’s paradise.
α = ω η1 · p1 + . . . + ω ηh · ph , β = ω η1 · q1 + . . . + ω ηh · qh ,
Contrary to Cantor’s sums and products of ordinal numbers, normal sums and prod-
ucts are commutative and compatible with the standard order of ordinal numbers, that
is
α +n β = β +n α, α ·n β = β ·n α,
α < β ⇒ α +n γ < β +n γ, α < β ⇒ α ·n γ < β ·n γ.
Thus, the structure (Ord, +n , ·n , 0, 1, <), where Ord stands for the class of ordinal
numbers, is an abelian semigroup.
Hence, e.g. since ω = ω · 1 + 0, and 1 = ω · 0 + 1, we calculate the normal sums of
ω +n 1 and 1 +n ω as follows,
1 +n ω = (ω · 0 + 1) +n (ω · 1 + 0) = ω · (0 + 1) + 1 = ω + 1,
3
For the use of this definition, we assume that some pi or qi could equal 0.
3
ω +n 1 = (ω · 1 + 0) +n (ω · 0 + 1) = ω · (1 + 0) + 1 = ω + 1.
Similarly, we calculate
2 ·n ω = (ω · 0 + 2) ·n (ω · 1 + 0) = ω 2 · 0 + ω · 2 + 0 = ω · 2,
ω ·n 2 = (ω · 1 + 0) ·n (ω · 0 + 2) = ω 2 · 0 + ω · 2 + 0 = ω · 2.
As is well known, in Cantor’s arithmetic the inequalities hold 1 + ω < ω + 1, and
2 · ω < ω · 2.
It can be shown that ordered field of Conway numbers, as developed in (Conway,
2001), includes the structure (Ord, +n , ·n , 0, 1, <); see (Błaszczyk & Fila, 2020). Hence,
each Cantor’s ordinal number is subject to ordered field operations, and all rules of finite
numbers, i.e. rules of an ordered field, can be applied to Cantor infinite numbers. There-
fore, in the field of Conway numbers, next to the number ω there are also numbers such
as −ω, ω/2 or the inverse of ω, that is 1/ω.
3.1 Numerosities
We present a simplified version of the theory of numerosities. It considers subsets of N
only. Still it exemplifies an alternative to the Cantor’s theory. Within Cantor system, every
subset of N is either a finite set or of a set with the cardinality ℵ0 . Theory developed by
Benci and Di Nasso gives the same result regarding finite sets, yet, infinite subsets of N
have a smaller numerosity than the numerosity of N.
We present numerosities as nonstandard natural numbers. By extending the system
of natural numbers via the ultrapower construction to the system of nonstandard natural
numbers, we demonstrate how numerosities measure subsets of N. Specifically, we show
that the numerosity of N is twice as big as the numerosity of the set of even natural
numbers. The extension of the system of natural numbers is placed in a broader context
of extension of the real numbers system. In both cases, we apply the same construction
called the ultrapower.
4
3.2 Ultrafilter on the set N
We start with the definition of an ultrafilter, and present some basic results concerning
ultrafilters.
Definition 1 A family of sets U ⊂ P(N) is an ultrafilter on N if (1) ∅ ∈ / U, (2) if
A, B ∈ U, then A ∩ B ∈ U, (3) if A ∈ U and A ⊂ B, then B ∈ U, (4) for each A ⊂ N,
either A or its complement N \ A belongs to U.
Take the family of sets with finite complements,
{A ⊂ N : N \ A is finite}.
This family is usually called the Fréchet filter on N. Indeed, it obviously satisfies
conditions (1)–(3) listed in the Definition 1. Note, however, that neither the set of odd
numbers nor the set of even numbers has a finite complement, hence, the Fréchet filter
is not an ultrafilter. Still, by applying the Axiom of Choice it can be extended to an
ultrafilter. In what follows, let U be a fix ultrafilter on N which extends the Fréchet filter.
Thus, we know that for every k ∈ N, the family U includes the set
since sets of this kind belong to the Fréchet filter. Moreover, the set N also belongs to U,
since it belongs to the Fréchet filter.
Next, due to the condition (4) of Definition (1), for any subset A of N, either A, or
N \ A belongs U. We apply this fact to prove, e.g. an equivalence (1), as explained below.
Finally, it can be shown that this proposition is true.
Theorem 1 For any subsets A1 , ..., An of N such that Ai ∩Aj = ∅, i 6= j. If ni=1 Ai ∈ U,
S
Hence, the product and sum of hyperreals [(r1 , r2 , ...)] and [(s1 , s2 , ...)] gives [(r1 ·s1 , r2 ·
s2 , ...)], and [(r1 +s1 , r2 +s2 , ...)] respectively. The relation [(r1 , r2 , ...)] <∗ [(s1 , s2 , ...)] holds
4
For details, see (Błaszczyk, 2014), (Błaszczyk, 2016).
5
when, for example, the set {n ∈ N : rn < sn } equals N minus some finite set (though the
definition of order <∗ also includes other cases).
Standard real number, r ∈ R, is represented by the class [(r, r, r, ...)], i.e., the class of a
constant sequence (r, r, r, ...). Note that the sequence representing standard real number,
e.g. 1, can take the same value from some index on, for example
1 = [(1, 1, 1, 1...)] = [(0, 0, 1, 1, ...)].
Owing to the above definitions, we employ the same symbols for real numbers in
the standard and non-standard context; we will also employ the same symbols for sums,
products and order relation in the standard and non-standard context.
It follows from the notion of ultrafilter that the following relation holds
(1) [(rn )] 6= [(sn )] ⇔ {n ∈ N : rn 6= sn } ∈ U.
Due to this fact, we can control, e.g. an inequality such as this one [(rn )] 6= 0. This
fact, in turn, enables to show that the quotient structure is really an ordered field.
In the next section, we consider hyperreal numbers represented by sequences of natural
numbers, that is [(nj )], where (nj ) ⊂ N, for instance
(2) α = [(1, 2, 3, ...)] = [(n)].
According to the definition of product, we have
α2 = [(1, 2, 3, ...)] · [(1, 2, 3, ...)] = [(11 , 22 , 32 , ...)] = [(n2 )].
Then, the hyperreal number α2 is determined by the following equalities
α h i h i
= [(1, 2, 3, ...)] · [( 21 , 12 , 12 , ...)] = 21 , 22 , 32 , ... = n2 .
2
√
Similarly, that is point-wise, we define the hyperreal number α, namely
√ h√ √ √ i h√ i
α= 1, 2, 3, ... = n .
In a similar way, the floor function is defined, namely
j k
[(rj )] = [(brj c)].
Hence, hyperreal numbers such as
α
j√ k
and α ,
2
are represented by sequences of natural numbers, namely
α
hj ki j√ k hj√ ki
n
= 2
, α = n .
2
More specifically,
α
h i hj ki
1 2 3 n
= , ,
2 2 2
, ... = 2
= [(0, 1, 1, 2, 2, 3, 3, ...)].
2
Note that with natural numbers, the following equalities obtains bn2 c+ bn2 c = n or
b n2 c + b n2 c + 1 = n, depending on whether n is even or odd. Similarly, α
2
+ α
2
= α or
α α
2
+ 2
+ 1 = α, depending on whether the set of even numbers belongs or not to the
ultrafilter U. Yet, we skip a discussion needed to justify this claim.
6
3.4 Extending Natural Numbers
In this subsection, we apply the ultrapower construction, as explained above, to natural
numbers (N, +, ·, 0, 1, <). As a result we obtain the nonstandard (and uncountable) model
of Peano arithmetic (N∗ , +, ·, 0, 1, <). Thus, the set N∗ is the quotient class of NN with
respect to the following relation
(nj )≡(mj ) ⇔ {j ∈ N : nj = mj } ∈ U.
New sums and products are defined pointwise, new total order is defined by
Standard natural number, n ∈ N, is represented by the class [(n, n, n, ...)]. Like in the
case of hyperreals, we employ the same symbols for natural numbers, as well as for sums,
products and order, in the standard and non-standard context.
Again, from the fact that the Fréchet filter is the subset of U, it follows that both the
constant sequence (2, 2, 2, ...), and a sequence (nj ) which on a finite set of indexes A takes
0, and for other indexes takes 2, i.e.,
(
0, for j ∈ A,
nj =
2, for j ∈ N \ A,
represent number 2,
[(2, 2, 2, ...)] = 2 = [(nj )].
For the rest of our presentation, we call nonstandard natural numbers numerosities,
and give a special role for the number α, as defined by formula (2): we will show that α
is the numerosity of the set N.
To unify developments of this and the previous sections, we can define nonstandard
natural numbers as a subset of the set of hyperreals as follows
N∗ = {[(nj )] ∈ R∗ | {j ∈ N | nj ∈ N} ∈ U}.
Usually, the symbol X stands for the cardinal number of the set X. Here, yet, it
stands for natural number, since for any n, the set {a ∈ A | a ¬ n} is finite. Thus, we
may interpret the symbol {a ∈ A | a ¬ n} as follows “how many elements of the sets A
are less or equal to n”.
7
Definition 2 The numerosity of the set A is the nonstandard natural number να (A)
represented by the sequence (ϕA (n)), that is
Here are some examples. 1) Let us start with finite sets, e.g. a two–elements set
A = {k, l}, with k < l. We have,
0,
for n < k,
ϕA (n) = 1, for k ¬ n < l,
for l ¬ n.
2,
Since for all but finite number of n we have ϕA (n) = 2, the numerosity of A equals 2,
that is να (A) = 2.
In a similar way, we obtain that numerosity of the set A = {a1 , ..., ak } equals k.
2) Now, we assign a numerosisty to the set of natural numbers N. To this end, observe
that ϕN (n) = n, for every n. Hence, the sequence (ϕN (n)) is (1, 2, 3, ...), and
This fact explains the role of the index α: it is the numerosity of the set N and other
numerosities rely on this basic fact.
3) Now, let us calculate the numerosity of the set of even numbers 2N = {2, 4, 6, ...}.
One can easily figure our the first terms of the sequence (ϕ2N (n)). These are as follows
ϕ2N (1) = 0, ϕ2N (2) = 1, ϕ2N (3) = 1, ϕ2N (4) = 2, ϕ2N (5) = 2, ...
According to Benci and Di Nasso, it justifies the old law The whole is greater than the
part, even when applied to infinite sets. In fact, rule (4) applies to subsets of the set N.
Still, the domain of numerosities can be extended to the subsets of real numbers.
8
Figure 1: Euclid’s Optics, Proposition 2 (left), and its modern counterpart (right)
“It is fair to suppose that the main source of difficulties which accompany the concept
of infinity is the deep contradiction between this concept and our intellectual schemes.
Genuinely built on our practical, real life experience, these schemes are naturally adapted
to finite objects and events. The first solution to that contradiction was to admit, as
Aristotle did, that infinity expresses, in fact, only a pure potentiality, i.e., the non-limited
possibility to increase an interval or to divide it. It was that interpretation of infinity as
a potentiality, which dominated mathematics until the Cantorian revolution” (Fischbein
et al, 1979, p. 3).
Yet, there was a long tradition of geometrical optics that applied actual infinity as a
mathematical tool. It originated from Euclid’s Optics, then, was modified by Islamic and
Medieval tradition, and was developed further in Kepler’s Paralipomena, and Descartes’
Dioptrics. Below we present two samples of how mathematicians tackled with the actual
infinity within that tradition.
9
Figure 2: Descartes’ Dioptrics, p. 21.
A small modification of Euclid’s diagram, results in a modern proof of the fact that
every two line segments are equinumerous, as presented by the right side of Figure 1. But
there is a crucial difference concerning the concept of line segment in the ancient Greek
and modern mathematics. In the 19th century, it was explicitly stated that line segments
consist of points. In the ancient Greek optics, it was an implicit assumption; in the ancient
Greek geometry, line did not consist of points; in the ancient philosophy, this claim was
a moot point.
10
Figure 2 depicts rays of light, ABI, KBL, and PRS, passing from air to glass. In the
Second Discourse, through some rare speculations, Descartes came to the conclusion that
the refraction could be measured by a proportion of some line segments, namely
AH : GI :: KM : N L.
AH : GI :: KM : N L, AH : GI :: P Q : T S, KM : N L :: P Q : T S.
Secondly, Descartes also addressed a problem how to determine a refracted ray when
the angle of incidence is known. In modern optics, the fraction vi /vt is determined by
means of experiments, then sin θt follows from (5). Descartes, managed to determine a
ratio of two infinite velocities, for instance, the velocity of light in air and the velocity
of light in glass, ∞a , ∞g in short. Due to a tricky instrument described in the Ninth
Discourse, he could determine, also by an experiment, a ratio of two segments, say a, b,
which played a similar role to the ratio vi /vt in the modern formula, namely
AH : GI :: a : b, KM : N L :: a : b, P Q : T S :: a : b.
Therefore, when AH is given, and a, b are determined by the experiment, a line seg-
ment, say X, which determines a refracted ray, will follow from the proportion AH : X :: a : b.
To sum up, despite velocities of light in air and in glass were considered to be infinite,
their ratio was measured by a ratio of some finite line segments such as
∞a : ∞g :: a : b.
Summary We argued that although students intuitions may not fit to Cantor’s theory,
there are new theories that match these intuitions. While Cantor’s arithmetic of infinite
numbers hardly mimics the arithmetic of finite numbers, these new theories comply with
the rules of a commutative ordered field, that is, the rules of finite arithmetic, or the Eu-
clid’s law The whole is greater than the part. Thus, it is possible that modern mathematics
provides theories of infinity more intuitive than Cantor’s theory of cardinal and ordinal
numbers.
Our short historical review aimed to buttress the claim that various intuitions of actual
infinity had been mathematised long before Cantor developed his theory.
References
[1] Benci,V., Di Nasso, M. (2019). How to Measure the Infinite. Mathematics with Infinite
and Infinitesimal Numbers. World Scientific. New Jersey.
[2] Błaszczyk, P. (2016). A Purely Algebraic Proof of the Fundamental Theorem of Al-
gebra. Annales Universitatis Paedagogicae Cracoviensis Studia ad Didacticam Math-
ematicae Pertinentia VIII, 6–22.
11
[3] Błaszczyk, P., Fila, M. (2020). Cantor on Infinitesimals. Historical and Modern Per-
spective. Bulletin of the Section of Logic 49(2), 1–32.
[4] Błaszczyk, P., Major, J. (2014). Calculus without the Concept of Limit. Annales Uni-
versitatis Paedagogicae Cracoviensis. Studia ad Didacticam Mathematicae Pertinentia
VI, 19–38
[5] Burton, H.E. (1945). The Optics of Euclid. Journal of the Optical Society of America
35(5), 357–372.
[7] Cantor, G. (1897). Beiträge zur Begründung der transfiniten Mengenlehre. In: (Cantor,
1932), 312–351.
[9] Descartes, R. (1637). Discours de la Méthode pour bien conduire sa raison, et chercher
la vérité dans les sciences, plus la Dioptrique, les Météores, et la Géométrie, qui sont
des essais de cete Méthode. Jan Marire, Leiden
[10] Fischbein, E., Tirosh, D., Hess, P. (1979). The Intuition of Infinity. Educational Stud-
ies in Mathematics 10, 3–40.
[11] Gödel, K. (1947). What is Cantor’s Continuum Problem? The American Mathemat-
ical Monthly, 54(9), 515–525.
[13] Hessenberg, G. (1906). Grunbegriffe der Mengenlehre. Vandenhoeck & Ruprecht. Got-
tingen.
[14] Heiberg, J.L., Menge, H. (1895). Euclidis Opera Omnia. Vol VIII. Taubneri. Lipsiae.
[15] Smith, M. (1999). Ptolemy and the Foundations of Acient Mathematical Optics.
American Philosophical Society, Philadelfia.
[16] Olscamp, P. (2001). René Descartes Discourse on Method, Optics, Geometry, and
Meterology. Hackett Publishing Company, Indainapolis.
12