60 METRIC SPACE AXIOMATICS Spring AXIOMATICS THROUGH THE METRIC SPACE AXIOMS JAMES FENNELL, UNIVERSITY COLLEGE CORK In this article we discuss some elementary concepts in axiomatics, which is the study of axioms. The axiomatic system we use is that of a metric space, which will be familiar to many readers. Metric spaces were introduced by the French mathematician Maurice Fréchet in 1906 as a generalization of the real numbers and the “distance function” f (x, y) = |x − y|. The aim is to construct limits, continuity, openness and eventually differential calculus on this generalized base alone; the results will then apply to a wide class of spaces such as R2 without little particular work. Here we are only interested in metric spaces as an example of an axiomatic system; that is, a system which behaviors according to clear, welldefined rules. What do we desire in an axiom set? Intuitively, we might want our axiom sets to be as small as possible. In particular, if one axiom in an axiom set can be logically deduced from the others we might be inclined to remove it, as it serves no purpose and just makes our set less “pure” than it could be. If a specific axiom can’t be deduced from the others then we say that that axiom is independent. If every axiom in a set is independent, we say the whole set is independent. Another, more important, requirement of our set of axioms is that it does imply contradictory statement. Such an axiom set would be effectively useless. An axiom set that isn’t inherently contradictory is called consistent. The traditional definition of a metric space provides suitable grounds to discuss these and other concepts, and to illustrate techniques for proving properties of axiom sets. 1. Independence and equivalence The standard definition of a metric space, as given in [2] for example, is as follows. Definition 1. A metric space M = {S, d} consists of a non-empty set S together with a map d : S × S → R such that for all x, y, z ∈ S A1 d(x, y) = 0 if and only if x = y A2 d(x, y) ≥ 0 A3 d(x, y) = d( y, x) A4 d(x, y) ≤ d(x, z) + d(z, y) 2012 METRIC SPACE AXIOMATICS 61 All of these axioms are intuitive. A1 states that the distance from a point to itself is 0; A2 that the distance from two points can’t be negative; A3 that the distance is the same irrespective of the direction it is measured; and A4 is a generalized statement of the triangle inequality: if you take a detour on your way, the journey can’t be shorter. It can be easily demonstrated that the axiom set {A1 , A2 , A3 , A4 } is not independent. We do this by showing that one of the axioms (in this case A2 ) can be logically deduced from the others. Proposition 2. The axiom set {A1 , A2 , A3 , A4 } is not independent. Proof. If we assume only A1 , A3 and A4 we have d(x, y) ≤ d(x, z) + d(z, y) by A4 d(x, x) ≤ d(x, z) + d(z, x) letting x = y 0 ≤ d(x, z) + d(z, x) by A1 0 ≤ 2d(x, z) by A2 0 ≤ d(x, z) which is precisely A2 . This form of proof is common in the study of axiomatics. As opposed to normal large-scale proofs, these involve tiny, very precise manipulations that make clear exactly which axioms are being employed. Now, the fact that the axiom set in definition 2 is not independent is motivation to find (a presumably smaller) one that is. In [1] Howard Eves provides alternative axioms for the distance function of a metric space which replace those described above. Definition 2. A metric space M = {S, d} consists of a non-empty set S together with a map d : S × S → R such that for all x, y, z ∈ S M1 d(x, y) = 0 if and only if x = y M2 d(x, y) ≤ d( y, z) + d(z, x) The reader will observe that M1 is exactly A1 . M2 is a restatement of the triangle inequality that is different to A4 in a subtle but important way. Our first task is to demonstrate that the axiom set {M1 , M2 } is independent; after that, we will show that the set {A2 , A3 , A4 } can be deduced from it. To prove that a set of axioms is independent we must show that no axiom is implied by the others. This can be achieved by inventing systems in which all of the axioms hold but one. Thus the axiom we are evaluating can’t be implied by the others as if it were implied it would be true in our invented system! We illustrate the technique in the following theorem. 62 METRIC SPACE AXIOMATICS Spring Theorem 1. The axiom set {M1 , M2 } is independent. Proof. First, we will show that M1 ; M2 . Consider the space R, d1 with d1 (x, y) = x − y. Clearly d1 (x, y) = 0 if and only if x = y, and so d1 satisfies M1 . Suppose it satisfies M2 . Then d(x, y) ≤ d( y, z) + d(z, x) x − y ≤ y −z+z− x x−y≤ y−x Setting, for example, x = 2 and y = 1 leads to an absurdity, so d1 does not satisfy M2 . Thus M1 cannot imply M2 . Now consider the space R, d2 with d2 (x, y) = x − y + 1. We have |x − y| + 1 = |x − z + z − y| + 1 ≤ |x − z| + |z − y| + 2 ≤ | y − z| + 1 + (|z − x| + 1) d(x, y) ≤ d( y, z) + d(z, x) so d2 satisfies M2 . But we have d2 (x, x) = 1 6= 0 so it does not satisfy M1 . Thus M2 cannot imply M1 . Now that we have proved that the set {M1 , M2 } is independent and so satisfies our aesthetic demands, we must show that it implies properties A2 , A3 and A4 and then that definition 2 can be used as an alternative to definition 1. Theorem 2. Given the axioms M1 and M2 we have (i) d(x, y) ≥ 0 (ii) d(x, y) = d( y, x) (iii) d(x, y) ≤ d(x, z) + d(z, y) which are precisely axioms A2 , A3 and A4 respectively. Proof of (ii). d(x, y) ≤ d( y, z) + d(z, x) d(x, y) ≤ d( y, x) + d(x, x) d(x, y) ≤ d( y, x) Similarly, we have d( y, x) ≤ d(x, y) 2012 METRIC SPACE AXIOMATICS 63 And so d(x, y) = d( y, x) as required. The proof of (i) is that of proposition 2 and the proof of (iii) is an obvious application of (ii) to axiom M2 . If we go in the other direction – assume axioms {A1 , A2 , A3 , A4 } and deduce axioms {M1 , M2 } – then we will have proved that the two axiom sets are equivalent. In this case either set can be used with the exact same results. The proof is left as a first exercise. The second exercise is to prove that the axioms A4 and M2 are independent. So though both are statements of the triangle inequality they are fundamentally different. 2. Consistency Having considered independence we now turn to the second concept mentioned in the introduction: consistency. Recall that an axiom set is consistent if it does not imply contradictory statements. The standard means for proving an axiom set consistent is the method of models. A model is a specific instance of the general object described by the axiom set. The real numbers and the distance function f (x, y) = |x − y| is thus a model of the metric space axioms. The aim is to furnish a particular model that we know (or hypothesise) to be free of contradictions; if the model satisfies our axioms we will know that the axiom set cannot inherently imply contradictory statements, as those contradictions would appear in the model. There are two kinds of model one can have: ideal and concrete. Ideal models are those that are based on another axiomatic system. The real number model is ideal as the real numbers are themselves governed by axioms. Ideal models seem the natural choice for mathematicians, but they come with a caveat: if the axiomatic system the model is based on hasn’t been proved consistent then the axiom set under consideration isn’t necessarily consistent either. The best we can hope for in such a situation is relative consistency where we prove that if system A is consistent then so is system B (and perhaps vice-versa). The consistency of the real numbers is a famous open question given as one of the 23 problems in David Hilbert’s 1902 lecture. The consistency of other sets, such as the natural numbers, is dependent on that of the real numbers (and vice-versa, in this case). On the other hand, concrete models are taken from the real world. If we accept that the real world contains no contradictions (at the macro-level of human experience at least, and excepting the speeches of politicians) then providing a concrete model will establish the absolute consistency of the axiom system in question. Again, if the set were inconsistent then the contradictions would appear in the real world. We now devise a concrete model of metric space axioms, and thus establish the set’s absolute consistency. 64 METRIC SPACE AXIOMATICS Spring Let I be the set of Irish-associated mathematicians William Rowan Hamilton (WRH) George Boole (GB) and George Stokes (GS). Their date of births are: William Rowan Hamilton George Boole George Stokes 4th of August 1805 2nd of November 1815 13th of August 1818 Now define the “distance function” B to be the number of years between the births of any two members of I, rounded up to the nearest integer. Thus B(WRH, GB) = B(GB, WRH) = 11 B(WRH, GS) = B(GS, WRH) = 14 B(GB, GS) = B(GS, GB) = 3 and B(WRH, WRH) = B(GB, GB) = B(GS, GS) = 0 Now to prove consistency we must show that each possible occurrence M2 is satisfied (M1 being immediate). For example, B(WRH, GS) = 14 ≤ 3 + 11 ≤ B(GS, GB) + B(GB, WRH) The other 5 cases follow similarly, and we have then established Theorem 2.1. The axiom set {M1 , M2 } is consistent. 3. Conclusion We have seen that in the standard definition of a metric space, axiom A2 is redundant (proposition 2), that a small change in the statement of the triangle inequality makes A3 redundant (theorem 2), and that once we have dispensed with A2 and A3 we can go no further (theorem 1). We have also established the consistency of the second axiom set, which by equivalence applies to the first axiom set, too. The fact that a standard definition in the literature is not independent may be surprising, but this occurs for other systems too. The standard definitions of a field, a vector space and an equivalence relation all contain redundant axioms; the reader might find and fix these as an exercise. In practise, what we typically call axioms might be better stylised as properties. A lecturer giving a course on topology will not want to enter an effectively tangential discussion of axiomatics, 2012 UNIVERSAL ALGEBRA 65 and so presents the definition 1 as the properties of the distance function as opposed to taking a strict axiomatic line. Further discussion of axiomatics and the philosophy of mathematics can be found in Howard Eves’ excellent book [1]. References [1] Howard Eves. Foundations and fundamental concepts of mathematics, pages 232–233. Dover Publications, Mineola, NY, 1997. [2] W. A. Sutherland. Introduction to metric and topological spaces, page 21. Clarendon Press, Oxford, 1975. AN INTRODUCTION TO UNIVERSAL ALGEBRA, AND CATEGORIES OF ALGEBRAS OF SOME TYPE F JACK KELLY, TRINITY COLLEGE DUBLIN The goal of this article is to provide a brief introduction to the subject of universal algebra, and also, to prove some basic but important properties of the category of algebras of some type F . Universal algebra, particularly the theory of varieties, provides a generalization of objects such as groups and rings, and only knowledge of these areas should be required to follow the first few sections. Obviously, for the last section, wherein we discuss the category of algebras of some type, some knowledge of category theory is required. Any propositions whose proofs are available in one of the references will not be provided here (as [1] and [2], the books wherein the proofs are available, have been made available online by the authors). The proofs that are present are our own. Hence, it is possible that there are mistakes. Should you find any errors, feel free to get in contact at kellyj1129@gmail.com Any comments and criticisms are greatly appreciated. We would also like to thank Dr. Rupert Levene (UCD) who kindly agreed to supervise a project on this topic as part of the Trinity College Dublin Summer Internship Programme 2011, and whose help and guidance was invaluable. 1. Basic Definitions and Constructions Definition Let A be a set and let ω be a cardinal. An ω-ary operation on A is a map f : Aω → A. For the purposes of this article, ω ∈ N0 . For ω = 0, A0 is