prev

next

out of 30

View

1Download

0

Embed Size (px)

AXIOMATIZATIONS OF HYPERBOLIC AND ABSOLUTE GEOMETRIES

Victor Pambuccian Department of Integrative Studies Arizona State University West, Phoenix, AZ 85069-7100, USA

pamb@math.west.asu.edu

The two of us, the two of us, without return, in this world, live, exist, wherever we’d go, we’d meet the same faraway point.

Yeghishe Charents, To a chance passerby.

Abstract A survey of finite first-order axiomatizations for hyperbolic and absolute geometries.

1. Hyperbolic Geometry

Elementary Hyperbolic Geometry as conceived by Hilbert

To axiomatize a geometry one needs a language in which to write the axioms, and a logic by means of which to deduce consequences from those axioms. Based on the work of Skolem, Hilbert and Ackermann, Gödel, and Tarski, a consensus had been reached by the end of the first half of the 20th century that, as Skolem had emphasized since 1923, “if we are interested in producing an axiomatic system, we can only use first-order logic” ([21, p. 472]).

The language of first-order logic consists of the logical symbols �

, � , � , � , � , a denumerable list of symbols called individual variables, as well as denumerable lists of � -ary predicate (relation) and function (operation) symbols for all natural numbers � , as well as individual constants (which may be thought of as 0-ary function symbols), together with two quantifiers, � and � which can bind only individual variables, but not sets of individual variables nor predicate or function symbols. Its axioms and rules of deduction are those of classical logic.

Axiomatizations in first-order logic preclude the categoricity of the axiomatized models. That is, one cannot provide an axiom system in first-order logic which admits as its only model a geometry over the field of real numbers, as Hilbert [31] had done (in a very strong logic) in his Grundlagen der Geometrie. By the Löwenheim-Skolem theorem, if such an axiom system admits an infinite model, then it will admit models of any given infinite cardinality.

2

Axiomatizations in first-order logic, which will be the only ones surveyed, produce what is called an elementary version of the geometry to be axiomatized, and in which fewer theorems are true than in the standard versions over the real (or complex) field. It makes, for example, no sense to ask what the perimeter of a given circle is in elementary Euclidean or hyperbolic geometry, since the question cannot be formulated at all within a first-order language.

This does, however, not mean that the axiom systems surveyed here were pre- sented inside a logical formalism by the authors themselves. In fact, those working in the foundations of geometry, unless connected to Tarski’s work, even when they had worked in both logic and the foundations of geometry (such as Hilbert, Bach- mann, and Schütte), avoided any reference to the former in their work on the latter. Some of the varied reasons for this reluctance are: (1) given that the majority of 20th century mathematicians nurtured a strong dislike for and a deep ignorance of sym- bolic logic, it was prudent to stay on territory familiar to the audience addressed; (2) logical formalism is of no help in achieving the crucial foundational aim of proving a representation theorem for the axiom system presented, i. e. for showing that every model of that axiom system is isomorphic to a certain algebraic structure; (3) logical formalism is quite often detrimental to the readability of the axiom system.

The main aim of our survey is the presentation of the axiom systems themselves, and we are primarily concerned with formal aspects of possible axiomatizations of well-established theories for which the representation theorem, arguably one of the most difficult and imaginative part of the foundational enterprise, has been already worked out. It is this emphasis on the manner of narrating a known story which makes the use of the logical formalism indispensable.

We shall survey only finite axiomatizations, i. e. all our axiom systems will consist of finitely many axioms. The infinite ones are interesting for their metamathematical and not their synthetically geometric properties, and were comprehensively surveyed by Schwabhäuser in the second part of [71]. All of the theories discussed in this paper are undecidable, as proved by Ziegler [88], and are consistent, given that they have consistent, complete and decidable extensions. The consistency proof can be carried out inside a weak fragment of arithmetic (as shown by H. Friedman (1999)).

Elementary hyperbolic geometry was born in 1903 when Hilbert [32] provided, using the end-calculus to introduce coordinates, a first-order axiomatization for it by adding to the axioms for plane absolute geometry (the plane axioms contained in groups I (Incidence), II (Betweenness), III (Congruence)) a hyperbolic parallel axiom stating that

HPA � From any point � not lying on a line � there are two rays ��� and ��� through� , not belonging to the same line, which do not intersect � , and such that every ray

through �

contained in the angle formed by ��� and � � does intersect � . Hilbert left many details out. The gaps were filled by Gerretsen (1942) and

Szász [82], [83] (cf. also Hartshorne [26, Ch. 7, 41-43]), after initial attempts by Liebmann (1904), [49] and Schur (1904). Gerretsen, Szász, and Hartshorne

Axiomatizations of hyperbolic and absolute geometries 3

succeeded in showing how a hyperbolic trigonometry could be developed in the absence of continuity, and in providing full details of the coordinatization. Different coordinatizations were proposed by de Kerékjártó (1940/41), Szmielew [85] and Doraczyńska [18] (cf. also [71, II.2]).

Tarski’s language and axiom system. Given that Hilbert’s language is a two- sorted language, with individual variables standing for points and lines, containing point-line incidence, betweenness, segment congruence, and angle congruence as primitive notions, there have been various attempts at simplifying it. The first steps were made by Veblen (1904, 1914) and Mollerup (1904). The former provided in 1904 an axiom system with points as the only individuals and with betweenness as the only primitive notion, arguing that segment and angle congruence may be defined in Cayley’s manner in the projective extension, and thus, in the absence of a precise notion of elementary (first-order) definability, deemed them superfluous. In 1914 he provided an axiom system with points as individual variables and betweenness and equidistance as the only primitive notions. Mollerup (1904) showed that one does not need the concept of angle-congruence, as it can be defined by means of the concept of segment congruence. This was followed by Tarski’s [86] most remarkable simplification of the language and of the axioms, a process started in 1926-1927, when he delivered his first lectures on the subject at the University of Warsaw, by both turning, in the manner of Veblen, to a one-sorted language, with points as the only individual variables — which enables the axiomatization of geometries of arbitrary dimension, without having to add a new type of variable for every dimension, as well as that of dimension-free geometry (in which there is only a lower-dimension axiom, but no axiom bounding the dimension from above) — and two relation symbols, the same used by Veblen in 1914, namely betweenness and equidistance.

We shall denote Tarski’s first-order language by ���������

: there is one sort of individual variables, to be referred to as points, and two relation symbols, a ternary one,

� , with

������ ��� to be read as ‘point

lies between

� and

� ’, and a quaternary

one, �

, with �� ������

to be read as ‘ �

is as distant from

as �

is from � ’, or

equivalently ‘segment ��

is congruent to segment ���

’. For improved readability, we shall use the following abbreviation for the concept of collinearity (we shall use the sign ��� whenever we introduce abbreviations, i. e. defined notions):

� ���� ��� ��� ������ ��� � ���� ����� � �������� � �� (1)

In its most polished form (to be found in [71] (cf. also [87] for the history of the axiom system)), the axioms corresponding to the plane axioms of Hilbert’s groups I, II, III, read as follows (we shall omit to write the universal quantifiers for universal axioms): "! � ! . �� #�$ �� ,

4

"! � � . �� ����� � �� � ��� � ��� � ��� , "! ��� . �� #� ��� � � , "! ��� . � � �� �� � �� ��� ������ � � � � � � �� , "! ��� . ��� � ������ ��� � ������� �� � � � �� � ��� �� � �� � �� ��� � ��� � ��� ���� �� � �� ��� � ��� � ��� ��� , "! ��� . ������ ��� � �� , "! ��� . � � �� ������ �� ��� ������ � �� � ���� �� �� � ���!� � � � ����� � �� , "! ��" . � � �� ��� � � ���� ��� , "! ��# . �$� %� � �&��� ��� � ����