SET THEORY BY FELIX HAUSDORFF TRANSLATED FROM THE GERMAN BY JOHN R. AUMANN, Et aL. SECOND EDITION CHELSEA PUBLISHING COMPANY NEW YORK, N. Y.©1957, Cuetsea PusiisHina ComPANy CopyricuT 1957, sy CHELSEA PuBLisHinc CoMPANY Copyricit ©, 1957 ay CHetsea Pustisuinc CoMPANY CopyricHt ©, 1962 sy CHELsza PuBLisHiING CoMPANY THE PRESENT WORK 1S A TRANSLATION FROM THE ORIGINAL GERMAN, INTO ENGLISH, OF THE THIRD (1937) EDITION or MENGENLEHRE ny Feuix Havsporrr Lrprary of Conaress Catacocur Carp Numper 62-19176 PRINTED IN THE UNITED STATES OF AMERICAEDITOR’S PREFACE Because the present translation of Felix Hausdorff's famous Mengen- lehre is the work of more than one hand, and because during the consid- erable length of time in which this translation was in preparation, other duties, both personal and professional, have arisen to prevent the trans- lators from seeing the work of translation through to completion, it has devolved upon me to make such changes, in terminology as well as lan- guage, as consistency in style would require. A.G. PREFACE TO THE SECOND ENGLISH EDITION Professor R. L. Goodstein has been kind enough to write two appendixes for the present edition: Appendix E, on the contradictions in Naive Set Theory, and Appendix F, on the Axiom of Choice. The Editor wishes to thank Professor Goodstein for pointing out that such supplementation is desirable and for his kindness in acceeding to the Editor’s request that he write the Appendixes. Thanks are also due to Professor Smbat Abian and Mr. Coley C. Mills, Jr. for the correction of a number of errata in the Iirst Edition. A.G.FROM THE PREFACE TO THE SECOND EDITION The present book has as its purpose an exposition of the most impor- tant theorems of the Theory of Sets, along with complete proofs, so that the reader should not find it necessary to go outside this hook for supplementary details while, on the other hand, the book should enable him to undertake a more detailed study of the voluminous literature on the subject. The book does not presuppose any mathematical knowl- edge beyond the differential and integral calculus, but it does require a certain maturity in abstract reasoning ; qualified college seniors and first year graduate students should have no difficulty in making the material their own. More difficult topics, at the end of individual chapters, may be skipped ona first reading. The reader who is interested only in acquir- ing the simplest facts of point-sct theory can, after a quick perusal of the first two chapters, immediately attack the sixth. The mathematician will, I hope, find in this book some things that will be new to him, at least as regards formal presentation and, in particular,.as regards the strengthen- ing of theorems, the simplification of proofs, and the removal of unnec- essary hypotheses. The selection of material must of necessity be somewhat subjective, since the subject is so extensive and is still growing almost from day to day. Unavoidably, various expectations regarding the book, including some of the author’s own, could not be realized. A textbook, after all, can not aim at the completeness of a comprehensive report. In this case, there was the additional restriction that the new edition had to be sub- stantially curtailed in length as compared to the first edition (Grundziige der Mengenichre, Leipzig 1914 | Repr., Chelsea, 1955]) ; this would have necessitated a revision down to the smallest detail, and J found a complete rewriting of the book preferable. I thought it might be easicst to sacrifice, of the topics treated in the first edition, most of the theory of ordered sets, a subject that stands somewhat by itself, as well as the introduction to Lebesgue's theory of integration, which does not lack for exposition else- where, What is more to be regretted is the abandonment, owing to the necessity of saving space, of the topological point of view in point-set theory, which seems to have attracted many people to the first edition of this book; in this new edition, I have restricted myself to the simpler 56 PREFACE To SECOND AND THIRD Epitions theory of metric spaces and have given only a quick survey (§ 40) of topological spaces, which is a rather inadequate substitute. Finally, I have cut down the generality in the other direction as well, by omitting the special theory of Euclidean space (e.g., the Jordan Curve Theorem for Plane Curves) ; that is, roughly speaking, I have omitted all of the material that is based on approximate polygons and polyhedra. Thus, the reader will find a large number of theorems concerning Euclidean space, but only such theorems as hold for Euclidean space considered as a special case of a separable or a complete or a locally connected space, or the like. As against this omission there are some additions, such as a more complete treatment of Borel Sets and Suslin Sets (discovered in 1917) as well as of the Baire functions. Also, continuous mappings and homeo- morphisms are treated in more detail than they were in the first edition. As to a discussion of the so-called paradoxes and of questions regarding the foundations of the subject, I have decided to omit this from the present edition as well. PREFACE TO THE THIRD EDITION Set Theory has continued its steady and active development, and this in itself might have made desirable an actual revision of this book. How- ever, circumstances have prevented my doing this. Accordingly, the first nine chapters are an almost unchanged reprint of the second edition ; but in order to do at least partial justice to the progress achieved by the subject in the meantime, I have added a new — tenth — chapter covering two subjects that seem to be especially worthy of detailed exposition, and I have also touched upon three further topics in short supplements — without proofs, however. Had it not been for lack of space, the number of additional topics could have been considerably extended.Eprror’s PREFACE . From THE PREFACE To THE SECOND EDITION Prerace To THIRp Epition TABLE OF CONTENTS PRELIMINARY REMARKS 2.0... 2000000000 0e cece essence eeeees IIL. Iv. SETS AND THE COMBINING OF SETS §1. Sets ...... §2. Functions . §3. Sum and Intersection §4. Product and Power .. CarpinaL NUMBERS ...........- §5. Comparison of Sets ...... §6. Sum, Product, and Power §7. The Scale of Cardinal Numbers . §8. The Elementary Cardinal Numbers OrveER TyYPEs . §9. Order ... §10. Sum and Product .. 811. The Types & and & . OrpinaL NuMBERs ., §12. The Well-Ordering Theorem . §13. The Comparability of Ordinal Numbers § 14. The Combining of Ordinal Numbers . § 15. The Alefs ...........0c.0c0 eee § 16. The General Concept of Product . SysTEMs oF SETs . § 17. Rings and Fields §18. Borel Systems § 19. Suslin Sets ...8 TABLE OF CONTENTS VI. Porn Sets .. $20. Distance §21. Convergence 119 § 22. Interior Points and Border Points 127 §23. Theo, B, and y Points ....... 130 § 24, Relative and Absolute Concepts . 139 § 25. Separable Spaces 143 § 26. Complete Spaces . 149 § 27. Sets of the First and Second Categories . 159 § 28. Spaces of Sets . 166 § 29. Connectedness . 172 VI. Point Sers AND OrpINAL NUMBERS ................ 188 § 30. Hulls and Kernels 188 § 31. Further Applications of Ordinal Numbers 198 § 32. Borel and Sustin Sets § 33. Existence Proofs ...... § 34. Criteria for Borel Sets . VII. Mappincs or Two Spaces . § 35. Continuous Mappings . § 36, Interval-Images § 37. Images of Suslin Sets . § 38. Homeomorphism §39. Simple Curves § 40. Topological Spaces . IX. Reay Functions §41. Functions and Inverse Image Sets . 264 § 42. Functions of the First Class . . 280 §43. Baire Functions ... . 291 § 44. Sets of Convergence - 305 X. SuPPLEMENT - 312 §45. The Baire Condition § 46. Half-schlicht Mappings . APPENDIXES . 336 BreciocraPHy . 340 FurtHER REFERENCES . - 342PRELIMINARY REMARKS Intervals of real numbers are written with brackets or parentheses according to whether the end points are included or not. Thus if a < 6, (4,5), [4,5), (a,6], (a,b) are sets of numbers x that satisfy the conditions aSzSb aSzea «Sb xcb (— @, + o) is the set of all the real numbers (the entire straight line). The largest and the smallest of a finite number of real numbers 4%, X2,..., _ are called their maximum and their minimum respectively and are denoted by max (41, ¥2,-.., 4a], min (41,42... 40]5 for example, max [2,—3]=2, min [2,—3] =—3, max [2,2] = min [2,2] = 2. The same notation is also used for the largest and smallest numbers, if any, of an infinite set of numbers. Ifa sequence of real numbers 73, +2, ... is bounded from above — that is, if there are numbers v for which v = x, for every n—then there is among these numbers v a smallest one, say 7%. It is variously called the least upper bound (the abbreviation I.u.b is very common), the supremum, and sometimes the upper boundary of the sequence +,; in the German- language editions of this book the term used (following Weierstrass) is obere grenze ; we shall write U1 = sup [41,42,...] = sup a. 910 PrectMInary REMARKS For example, sup [0, 1/2, 2/3, 3/4, ...J = sup [(n—1)/n] = 1. If the sequence has a maximum, this coincides with the supremum, The greatest lower bound (g.l.b), or infimum, or lower boundary (or untere grenze) J of a sequence bounded from below is defined in an exactly similar way. The same terms are also applied to sets of numbers that are not given in the form of a sequence ; for instance, sup f(z). ass For a sequence bounded from above there exist all of the suprema uy = inf [41 x2,...] = ink x, a= SUP [qr Bapsr Enger so] with v; 2 vp =... ; if these-v, are bounded from below and hence have a(g.l.b)z, this is called the upper limit, or limit superior, of the sequence ; in symbols, lim sup +, = lim x. The lower limit, or limit inferior «= lim inf x,= lim xq is defined correspondingly. If the assumptions concerning boundedness do not hold, the symbols =& @ are used. For example, in the case of a sequence , not bounded from above, we write sup x,—= + 0 and lim r,—= + 0; in the case of a sequence x, bounded from above for which the suprema v are not bounded from below, we say that lim x, = — ©. For the convergence of sequences and functions we will, as a rule, use the arrow as symbol, eg., %q—> x means lims,=-+; likewise for proper divergence (x, + 0 and x, — 0). A statement about the natural number » (= 1, 2, 3, ...) holds ulti- mately, or for almost all n (Kowalewski), if it holds from a particular » on (n = no), or for all values of n with at most a finite number of exceptions ; it holds infinitely often if it holds for an infinite number of values of n (eg., for n = 2, 4,6,8,...). When we speak of almost all or of infinitely many terms of a sequence x, we mean the terms +, belonging to almost all or to infinitely many values of n, respectively, regardless of whether the terms themselves are different or not.CHAPTER I SETS AND THE COMBINING OF SETS §1. Sets A set is formed by the grouping together of single objects into a whole. A set is a plurality thought of as a unit. If these or similar statements were set down as definitions, then it could be objected with good reason that they define idem per idem* or even obscurum per obscurius.? How- ever, we can consider them as expository, as references to a primitive concept, familiar to us all, whose resolution into more fundamental con- cepts would perhaps be neither competent nor necessary. We will content ourselves with this construction and will assume as an axiom that an object M inherently determines certain other objects a, b, c, ... in some undefined way, and vice versa. We express this relation by the words: The set M consists of the objects a, b, c,.... A set can consist of a natural number (positive whole number) of objects, or not; it is called finite or infinite accordingly. Examples of finite sets are: the set of the inhabitants of a city, the set of hydrogen atoms in the sun, and the set of natural numbers from 1 to 1000 ; examples of infinite sets are: the set of all natural numbers, the set of all the points of a line, and the set of all the circles in a plane. It is to the undying credit of Georg Cantor (1845-1918) that, in the face of conflict, both internal and external against apparent paradoxes, popular prejudices, and philosophical dicta (infinitum actu non datur*) and even in the face of doubts that had been raised by the very greatest mathematicians, he dared this step into the realm of the infinite. In so doing, he became the creator of a new science, Set Theory — the consideration of finite sets being nothing more than elementary arithmetic and combinatorial analysis — which today forms the basis of all mathematics. In our opinion, it does not detract from the merit of Cantor's ideas that some antinomies that arise from allowing excessively limitless construction of sets still await complete elucidation and removal. ? Something in terms of that same something. * The obscure by the still more obscure. * There is no actual infinite. uu12 I. Sets AND THE CoMBINING oF SETS Following G. Peano, we designate the fundamental relation between an object @ and a set A to which it belongs in the following terms and using the following notation ; aisanelementof A: aed, The negation of this statement reads: ais not an element of A:aé A. Two sets are defined to be equal, in symbols A=B, if and only if each element of one is also an element of the other and vice versa (i.e. if they both contain the same elements). Hence a set is uniquely determined by its elements; we express this by writing the elements of the set, enclosed in braces, as a notation for the set, elements not listed explicitly being indicated by dots. Thus A={a}, A={a,d}, A={a,b,¢ are the sets consisting of the element a, the two elements a, b, and the three elements a, b, ¢, respectively ; A = {a,b,¢,..-} isa set consisting of the elements a, b, c, and (possibly) others. Naturally, it must be specified in some way what the other elements are that have been indicated by dots; some examples are the set of all natural numbers {1, 2, 3,...), the set of all even natural numbers {2, 4, 6, ...}, the set of all squares (1, 4,9,...}, the set of all powers of 2 {1, 2,4, 8. and the set of all prime numbers {2, 3, 5, 7, ...}- A distinction must certainly be made, at least conceptually, between the object a and the set {a} consisting of only this one element (even if the distinction is of no importance from a practical point of view), if for no other reason than that we will admit sets (or systems) whose elements are themselves sets. The set a = {1,2} consists of the two elements 1, 2, while the set {a} consists of the single element a.§1. Sets 13 For reasons of expedience, we also admit the null set, or empty set, denoted by 0, which contains no elements.’ By the definition of equality of sets, there is only one null set. 4 —=0 means that the set 4 has no elements, is empty, “vanishes.” If we were not to admit the null set as a set, then in countless cases where we speak of a set we would have to add “Gf this set exists.” For, the definition of the elements of a set often does not tell us in the least whether such elements exist ; for instance, it is not yet known whether the set of natural numbers for which the equation gt? 4 yh+2— 20+? has a solution in natural numbers x, y, and z is empty or non-empty (that is, whether Fermat’s celebrated Last Theorem is true or false). Thus, the statement 4 = 0 may, in many cases, repre- sent real knowledge but, in other cases, only a triviality; many mathe- matical statements —or, if one is not repelled by artificial devices, all mathematical statements —can be put in the form 40. The intro- duction of the null set, like that of the number zero, is dictated on grounds of expedience ; on the other hand, it sometimes forces one to state explic- itly, as one of the hypotheses of a theorem, that a set does not vanish (just as it is sometimes necessary to state explicitly that a number does not vanish). If A and B are two sets, then the question arises whether the elements of one may not perhaps also belong to the other. If a and b denote ele- ments of A and B respectively, then we first consider the following two pairs of alternatives : EveryaeB, noteveryaeB, everybeA, _noteverybe A. By combining these we obtain four possible cases, the first three of which are denoted in symbols by the accompanying formulas: (1) Every aeB and every bed: A=B; (2) every aeB and not every be A: A B; (4) not every ae B and not every be 4. In case (1), the two sets are equal, by the earlier explanation. In case (2), A contains only elements of B, but not all of them, which character- izes A as the smaller set and B as the larger set; this is expressed in symbols by 4 < B, which is supposed to be reminiscent of the notation a< Bin the case of numbers. In case (3), we have the opposite situation, * Whether the symbol 0 denotes the null set or the number zero will always be apparent from the context.14 I. Sets AnD THE ComBINING oF SETS so that the statement 4 > B amounts to the same as B< A. In general, what we will have will be none of these three cases, but rather case (4); but there is no reason to have a special notation for this case. The relation “smaller than” is transitive ; that is, from 4 < Band Bc C it follows that AO means that B is not empty. The usefulness of this convention becomes apparent, for instance, in counting the subsets of a finite set. The subsets of (1, 2, 3,} are: 0, {4}, {2}, {3}, {4,2}, {1,3}, (2,3), (4, 2, 3}: there are 8 = 2° of them. A set of n elements contains (n) subsets that contain m elements each, where ( ) is the binomial coeffi- n\)___a! m) = mt(a—m)l cient and where(O) = (1) =. The total number of subsets is ()+G)+-+ (2) +()=2 This simple result speaks in favor of the concept of subset as introduced above. If A s B, then we let B-A 1 The equality sign is dispensed with in some texts; in place of our rounded less than sign, other variants of the symbol are in use. * More explicitly: The statement “If a ¢4, then ae B also” is true because the hypothesis o€ A is never satisfied; a statement of this kind is said to be vacuously true. If p and q are statements, then the statement “If p is true, then q is true” (2 implies q) is certainly true when p is false, A false statement implies any state- ment; if 2X 2= 5, then ghosts exist,