Atlanta University Center DigitalCommons@Robert W. Woodruff Library, Atlanta University Center ETD Collection for AUC Robert W. Woodruff Library 1-1-1962 On the equivalence of the axiom of choice, Zorn's Lemma, and the Well-ordering theorem Grace Joy Traylor Atlanta University Follow this and additional works at: http://digitalcommons.auctr.edu/dissertations Part of the Mathematics Commons Recommended Citation Traylor, Grace Joy, "On the equivalence of the axiom of choice, Zorn's Lemma, and the Well-ordering theorem" (1962). ETD Collection for AUC Robert W. Woodruff Library. Paper 1384. This Thesis is brought to you for free and open access by DigitalCommons@Robert W. Woodruff Library, Atlanta University Center. It has been accepted for inclusion in ETD Collection for AUC Robert W. Woodruff Library by an authorized administrator of DigitalCommons@Robert W. Woodruff Library, Atlanta University Center. For more information, please contact [email protected]. ON THE EQUIVALENCE OF THE AXIOM OP CHOICE, ZORN'S LEMMA, AM) THE 1ELL-ORDERING THEOREM A THESIS SUBMITTED TO THE FACULTY OP ATIANTA UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE BY GRACE JOY TRAYLGR DEPARTMENT OF MATHEMATICS ATLANTA UNIVERSITY ATLANTA, GEORGIA JANUARY 1962 PREFACE In dealing with infinite sets there arises a logical question whether it is possible to "order11 the elements of these sets. Such a question has led mathematicians, such as Russell, Godel, and Zermelo, to postulate the "axiom of choice.1* This axiom has other forms, which are more useful in nature and in proofs of theorems; these forms are known as "Zorn's lemma1* and the "well-ordering theorem.11 After some preliminary considerations, these three statements will be shown to be equivalent. The first chapter consists of definitions of an "order relation," a "partially ordered" set, a "totally ordered" set, a "well-ordered" set and "linear order," and a proof of transfinite induction. The second chapter consists of applications of the axiom of choice, Zorn's lemma and the wellordering theorem. The third and final chapter oonsists of proofs showing the equivalence of these three statements. The writer is indebted to Dr. Lonnie Cross for his suggestion of this subjeot and for his helpful criticisms. I also wish to thank my typist, Miss Geraldine Ellis, for doing such a wonderful job of typing. ii TABLE OF CONTENTS Page PREFACE ii GLOSSARY OF SYMBOLS iii Chapter I. II. III. 17. V. ORDER RELATIONS 1 THE AXIOM OF CHOICE 6 ZORN'S LEMMA 8 THE WELL-ORDERING THEOREM 13 THE EQUIVALENCE OF THE AXIOM OF CHOICE, LEMA, AND THE nlELL-ORDERING THEOREM BIBLIOGRAPHY ZORN'S 21 SO iii GLOSSARY OF SYMBOLS SYMBOL NAME OF SYMBOL <=r Set inclusion <^~ Proper inclusion £ is a member of j. Union « Intersection 0 Mull set ^ Such that Implies that There exists If and only if Therefore Identical to Totality of elements in S not in A x does not precede y ^. &.,n ^UJ Not equal to The first number in the brackets refers to the number of tiie book in the biblio graphical Hstj the second number if the page reference Quod erat demonstrandum (L., which was to be demonstrated). Iv CHAPTER I ORDER RELATIONS Before we begin to solve the problem of "ordering" element of an infinite set, we must have in mind a clear meaning of the term "ordering.1* The following explanation gives us a precise definition. Letting S be a set and =^.be a binary relation defined on the set S, we say that ^- is an order relation if the following conditions are satisfied: (1) x-=& x for every x e S. (2) x :siy and y =£ x for x,y e S =S>x « y. (3) x^.y and y=&.z for The above is to be read as follows: In the set S, x^-y is to read as "x precedes y" and is to be read as "y succeeds x." If, however, the set S is a set of real numbers, x,y £ S, then x=^y is to read as "x less than or equal to y," and y-fcx is to be read as "y greater than or equal to x." These two relations are order relations* We have, however, three types of ordered sets: totally-ordered, and well-ordered. type* partially-ordered, We shall consider separately each 2 (1) If it is possible to define an order relation in a set S, then we say that the set is partially ordered by the order relation. (2) If a partially ordered set has the additional property that x^.y, x,ye S, y^x, then S is said to be totally ordered by^. • For example, let us consider the set S = £Buick, Ford, Pontiac, Fiat, !G, Jaguar}, where x=<y, x,ye. S, if x does not occur after y in either the United States made cars or the foreign made cars. partially ordered, but not totally ordered. This set is If we define ^- as above and also say that every foreign made car precedes every United States made car, then relative to this order relation, S becomes totally ordered. In connection with the above definition of an ordered relation ^, we may define the following symbols: (a) xily means y=^x. (b) x <y means x^.y but x^ y. (c) x> y means y-< x, (d) x ?? y means x and y are comparable, that is, one and only one of the three x< y, x-= y, x >- y, holds true. [1%YF$ From the definitions ofs£, >z»-< $ >■ » it is clear that if two elemsnts are comparable, then one and only one of the three relations y, x -=Ly, x y~y holds true* 3 Clearly every two elements of a partially ordered set are not necessarily comparable, but every two elements of a totally ordered set are necessarily comparable* Let S be a set ordered by ^.(totally ordered or partially ordered) and let A be any subset of S. Relative to A we make the following definitions. (a) If for every y €. A^ an a& A * az^y, then a is the first element of A. Similarly, if for every y e A ^ a z fc A > z^c y, then z is the last element of A. (b) If for a t A"* no elements ye A ^ y -< a, then a is a minimal element. Similarly, if for z fe A no elements ye Ad y > z then z is a maximal element of A. (c) If ^ a b g S ^ b dLy for all y <s. A, then b is a lower bound for A. If 3 a g & S * it is the last element of the set of lower bounds for A, then g is the greatest lower bound (g.l.b.) for A; and we write g ? g.l.b. A. Similarly, if 3 a u 6 S ^ u ^.y for all y & A, then u is an upper bound for A* If 3 a 1 e S ^ it is the first element of the set of upper bounds for A, then 1 is the least upper bound (l.u.b.) for Aj and we write lsl.u.b. A. (3) If a set S is a totally ordered relative to and also has the property that every non-null subset has a first element, then S is said to be well-ordered. h An elementary property of well-ordered sets is that we may apply the method of transfinite induction. Before we can actually consider transfinite induction, some terms must be defined. (1) A linear order L consists of a set, also designed by L, and an order relation ^satisfying the following conditions: (2) (a) for every a 6 L, b & L, ^ a-=£ b, either a=£.b or (b) if a^.b, then it is not true that (c) if at£b and bt^-c, then a^c. Two linearly ordered sets L and L' are said to be similar if a one-to-one correspondence a* — f (a) exists between them* if a^b, then f (a)^. f (b), (3) A set of similar linear ordered sets is called an order (10 The order type of a well-ordered set is called an ordinal number, Now we are ready to prove the transfinite induction theorem. Theorem. If S is the set of all ordinal numbers less than a given ordinal number, say$ , and A is a subset of S, and if (a) The first ordinal number 1 is in A, (b) For every ordinal number ^=c>, if all ordinal numbers less than $ are in A then ^ is also in A, then A~ S. Proof. Be If S — A ^r$), then there is a first ordinal S — A. than Q. E. D. By (a), §^l 1. But all ordinal numbers less are in A so that, by (b),§€ A.". S— A=r<£>, CHAPTER II THE AXIOM OF CHOICE In order to understand the axiom of choice, we mast know what is meant by a choice function. non-null set. For instance, we take S and let it be a We letCPbe a mapping (function) of S onto S which associates with every subset A of S an element a -=CP(A) of S. assume, also, that Op (A) e A* for S. We , therefore, is a choice function A choice function, then, is a method of selecting one element from every non-null subset of a given set. The axiom of choice states that every set has a choice function. As an illustration of the use of the axiom of choice, let us consider the following theorem: Theorem. If a set is infinite, then it has a subset equivalent to the set of positive integers, that is, the subset is denuraerable. Proof. Let S be an infinite set and let CP be a choice function for S as proposed by the axiom of choice. LetCp be a unique mapping on the set of positive integers, that is, CP assigns 1 to the element xt selected from S, then 2 to the element 2g selected from S- £x^ } then 3 to the element X3 selected from S - \x^, X£J ) and so forth. Since the range of the choice of a choice function is always the 6 7 subset of a given set, we see that C-P is an onto mapping* Let us suppose ~l a finite set N S — N"=- Then Cf establishes that S H N. This is contrary to the hypothesis that S is infinite. Therefore, for every S — N % $• from this ire conclude that the range of Cfi is equivalent to the set of positive integers, but -foe range of C-? is a subset of S, which means that the theorem is established* Q. E. D. CHAPTER in ZORN'S LEMMA Zorn's lemma is a maximal principle named for Zorn. Unfortunately, the history of it and some closely related maximal principles is very tangled. However, it is certain that Zorn was essentially anticipated by Felix Hausdorff. Zorn's lemma can be stated in two forms* The first form says that "every partially ordered set S contains a maximal (with respect to9) totally ordered subset." The second form says that "if S is a partially ordered set such that every totally ordered subset has an upper bound in S, then S has at least one maximal element," equivalent. These two forms are Let us now prove this equivalence and make an application of this lemma. Consider the following theorem and its proof: Theorem. Proof. The two forms of Zorn's lemma are equivalent. The first form of Zorn's lemma ==>the second form of Zorn's lemma. Let the first form of Zorn's lemma hold true* Let S be partially ordered by -=1 and assume that every totally ordered subset of S has an upper bound in S. By the first form of Zorn's lemma S has a maximal subset, which we shall denote by A* S. By hypothesis A has an upper bound u in We see that for all m in A m^.u. Suppose that u Jfe. A. We claim that u€ A. Let \ti\ denote a set consisting of a 8 9 single element u, and A+- A tt ^ti]f. This set A is a totally ordered subset in S which properly contains A. This is a contradiction, for A is a maximal, and therefore it cannot be contained in any other set. The tipper bound u & S. If there existed a u4" in S a u+>-u, then A**= AU&*9( would be a totally ordered subset in S which contains A. This contradicts the maximal character of A* Conversely, the second form of Zorn's lemma =f> the first form. Let the second form hold true. partially ordered by ^ • Let C be the collection of all totally ordered subsets of S. by-S. . Let S be Now C is partially ordered Let C6 be a totally ordered subset in C relative to £= , and let us consider the set A« •£. S, Ao"=- U A, where A € Co. We claim that A* is totally ordered relative to^- . If x & A» and y 6- Aft then x e A and y e A+for some A and A*1 in C . But C is totally ordered by .£. • Hence A and A are comparable and either x and y are in A or A+. For the sake of argument let x and y be in A+. A* £ C A is totally ordered relative to ^. • and y are comparable and hence A to =£u and A* € C. Because Therefore, x is totally ordered relative Thus every totally ordered subset C of (relative to& ) has an upper bound Ao in C. By the second form of Zorn's lemma, C has at least one maximal element relative to£= . Q. E. D. 10 Before we can make an application of Zorn's lemma we must clear the way by considering certain definitions* (1) A group 6 is a collection of objects on which a binary operation # is defined. This operation is subject to the following laws: (a) Closure* If a and b are in G, then a * b is in G. (b) Associativity* If a,b and c are in G, then (a # b) # c= a # (b # c). (c) Identity* There exists a unique element e in G (called the identity element) such that for all a's in G, a •& e = e -* a = a. (d) Inverse. For every a in G, there exists a unique element af in G called the inverse of a such that a * a' c a' # a — e. (2) An additive group G is a group that indicates that the binary operation is addition. (3) A group G is an abelian group if for all a,b in G, a # b~b •» a. (U) A ring R is an additive abelian group with the additional properties: (a) The group E is closed with respect to a second binary operation designated multiplication. (b) Multiplication is associative, that is, (ab)c ~ a(bc for all elements a,b,c in R. n (c) Multiplication is distributive with respect to addition on both the left and right side, that is, a(b-v c)= ab -v ac (b * c)a •= ba + ca for all elements a,b,c in R. (5) /?} 5U7 A ring R that has a tinit element is a ring with the element e such that ea •*- ae ■= a is true* (6) If N is an additive subgroup of a ring H such that both ra and ar are in N for all elements a in N and all ring elements r in R, then N is an ideal. (7) ^2$737 An ideal N is maximal if it is not contained properly in any other ideal. Now let us consider the following theorem from the theory of rings* Theorem, let R be a ring with a unity element. Then every ideal properly contained in R is contained in a maximal ideal. Proof. Let R be a ring which contains properly N and let C be the collection of all ideals containing N and unequal to R* Assume C is partially ordered byS: and by Zorn's lemma it contains a maximal totally ordered subset Co* *" We see that Q where Q e Co. is an ideal since C We claim that Q* is a maximal ideal. is totally ordered by S • Assert first Q ^ R. Suppose that the unity element 1 were in Q . in some Qt c , Let But if le.§ , the ideal Then 1 would be would become the whole ring E (if an ideal contains the unity element, it then becomes the whole ring), which is a contradiction. We assert 32 second that there exists no ideal Q Suppose such an ideal did exist. such that QCQ c R* The set of ideals ^Co > Q would be totally ordered—which contradicts the maximal character of Co• Hence Q + is a maximal ideal and therefore every ideal properly contained in R is contained in a maximal ideal. Q. E. D. J CHAPTER IV THE 1ELL-CRDERING THEOREM Since Zermelo was the first to give a rigorous proof of the welLordering theorem, we often refer to the theorem as "Zermelo's theorem.n Its proof is presented in two forms. Because of the difficulty in following the steps of the second form of the proof, the first form only will be reproduced here. This proof reveals to us only the existence of a well-ordered set; it does not demonstrate to us how to obtain a well-ordered set. For example, no one has exhibited to us a well-ordering of the set of real numbers* Let us consider the following: Definition. A system of subsets of S is called a tower if it has the following properties: (a) It contains the null set<t> • (b) If it contains a class of subsets of S, then it contains their union. (c) If it contains a proper subset A C_S, then it contains the union of A and the element CP (A), where Cjp (A) is a mapping of S into S — A. Well-ordering Theorem. Eroof. ^j6j*7 Every set can be well-ordered. Let S be an infinite set. Let Q) (A) be a mapping (function) of the elements a e S —A into every proper subset 13 A of S, including the null set. We shall show that the function Cp(A) establishes an onto mapping between S and a class C of subsets of S which is shown to be a well-ordered set with respect to d • Let C be the intersection of all towers. We claim that C itself is a tower. (a) C contains the null set, since it is the inter sections of classes each of which contains the null set. (b) If A* £ C for every uA, then O A4 & C, -i e A. For A*,JU A belongs to every tower, so that VJ A«i, -l £ A, belongs to every tower, whence (c) If A 6 C, then AU^.CP(a)"|j € C. every tower, so that A U$.Q(A)1[ For A belongs to belongs to every tower; hence A L>9j(A) J e C. Therefore, C is a tower. Since C is the intersection of all toners, it follows that any subset of 0 which is a tower must be identical with C. By A 6 C comparable we mean that for every Y e either I C A or A C Y. C We show that if A is comparable then for every Y e C either Y c A or AU$,C0(A)^c Y. show this we show that the sets of Y property form a tower* (a) <$ C. A. C which have this To (b) Given X^fc A. If Yjl C A for every^fe A, then UYjCA,-lfeA. If Y^D AU §#A) 1[ for someJL6 A, then U YjtD AO5.Q(A) \ ,-le A. (c) If Y D AO$G?(A)]f , then YO^Q(Y)^ A U^>(A) j . If Y •=. A, then YOjCfcCY)"} H A O^A) J . If Y is a proper subset of A, then YU^Cp(Y) J C. A. For if YU^CD(Y)}if not a subset of A, it contains some element not in A, but Y is a subset of A, so that this element does not belong to Y. Since Y is a proper subset of A there is an element in A not in Y which belongs to YU^CP(Y)^, since A is conparable and so is a subset of Yb$_C0(Y) J • Hence at least two elements not belonging to Y. This is impossible. The property of comparable sets just obtained allows us to show that the sets A & C which are comparable form a tower, so that all members of G are comparable. (a) 9^ (b) is comparable. Suppose Ajl, J-€A, is comparable. C. If Ajl C Y for all J & A, then U Aj. C Y,<* e A. If Aj. 3 Y for some d. e A then Hence, (c) o U Let Y € A j. D Y, «i fc A. A^,oifeA, is comparable. Suppose A 6. C is comparable. Then for every Y e either Y C A or Ab^QCA) ^C Y, so that A U is comparable. C 16 We now know that for every A e C, B e C, either A C B or BC A. Using HC" as the order relation, the members of C form a linear order. We show that this linear order forms a well-ordered sets Let Aj^,^€ A, be any class of sets in C. A= A Aj,«ie A. B=UB^ c Let Consider all B^ 6 C,^BiB»CA. A, ^6 B. If Then B-=Aj.for someJife A, then B is the first member of this set. If B is not one of the A^ , then B U l^CB)^ is one and must be the first. there is a ^fe A > A^cAa for everyJ-fc A. In any case, Hence C, with « c» as an order relation, is a well-ordered set. Finally we show that there is a one-to-one correspondence between C and S. correspondence. The function a —CP(A) gives the desired For if A G C, B <b C, and A is a proper subset of B, then CP (A) 6 B, so that C?(A) ^=C/>(B). then C? (A) — Cp(B). a "=Cp(A). Hence if A = B, Next, for every a <s S there is an A 6 C* For let A be the union of all sets in C which do not contain a. If a=rCp(A), then a« A(J WCA) ] is a contradiction. Hence a =<-P(A)• This one-to-one correspond ence produces a well-ordering of S. Q. E. D. Let us consider the following problem as an application of the well-ordering theorem* Problem. Find a function f(x) such that (1) f(x*y)-^f(x) + f(y). The function ex obviously has this property, for every constant cj and if only continuous functions are sought, there 17 are no other solutions of this problem. For, put f(l)"= c. then from (1) we get f (2) =r f (1+1) ■= f (1) + f (1) = c + c r 2c, f (3) - £(2+ 1) ^ f (2) + f (1) -=2c V c = 3c, and in general, for every natural number n, (2) Further, f(n)^= nc. f(n*O) = f(n) ■* f(0), and hence f(0) zr 0, Conse quently f fr-n) 4 f (n) s f(0) - 0, so that (2) holds for every integer n. Moreover, analogous considerations for every integer q"=£ 0 shows that f(n)r= f(q-$) q . and hence so that f(r)- re for every rational r. For a continuous function, however, f (x) •=. ex. For discontinuous function this functional equation (1) has solutions other than those just found. To prove this fact, we employ the well-ordering theorem. We first obtain a Hamel basis for the real numbers. By this we mean a set H of real numbers is called a Hamel basis if it has the following properties: (a) For any finite number of basis elements b,, ba, ..., bm, and arbitrary rational numbers r( , r^, 18 ..., r^ (b) not all zero, we never have For every number z "=£ 0 there are among the basis elements a finite number of basis numbers b| , ..., bm, and, corresponding to these, rational numbers r, , ..., r^ z rr all different from zero, a Tk, bv • First we have to show that such a basis exists. We do this by- building up a basis as a well-ordered set by means of transfinite induction. The starting-point is the well-ordering ft: of the set of all real numbers. xas H. Let x, e H if x, i 0. Otherwise, let For a given**. , suppose for every ^ ^-<Athat the decision has been made whether or not x$ £ H. Then xj e H if and only if x is not a linear combination with rational coefficients of any finite number of those x$ , ^^-"J, which are in H. The set His a basis. For suppose that an equation of the sort described under (a) held for some finite number of basis numbers. Then one of these finitely many basis numbers would be the last to have a nonzero coefficient, and could then not be admitted into H, contrary to our assumption. Condition (a) is fulfilled. is seen as follows: not. z=b. That (b) too is fulfilled A given number z^O is either a basis number b or In the first case, (b) is fulfilled trivially by the equation m the second case, an equation of the kind described in (a) z and the basis numbers preceding it in R. On account of property (a) already established for the basis numbers, z has a nonzero coefficient 19 in this equation. The equation can therefore be solved for z, and if we leave out the term with zero coefficients, assertion (b) is obtained. The representation which exists, according to (b), for every number z ■£ 0, is uniquely determined by a. For if there were two distinct representations, subtraction of the two equations would result in a contradiction of (a). Now we define the function f(x) first only for the basis numbers, and this is done in an arbitrary manner. x Further, we let f(0)— 0. If 0 is an arbitrary real number, it can be represented in terms of a finite number of basis numbers with rational coefficients, in the form Here we shall also admit terms whose coefficients are zero; the uniqueness of the representation is then obviously still valid. Now we put f(x)- irt ftt^). Then, if y-^s^ b*. is a representation of y in terms of basis elements, is such a representation of x + y, and, as a matter of fact, due to the uniqueness, precisely the representation. -f(x) Consequently, + f(y). The function f is thus a solution of the functional equation. It is a discontinuous solution if we put, say, f(b,) ~ 0, f(b^) =: 1 for two basis elements b,, b.;j for if f were a continuous solution, the argument 20 presented at the beginning shows that we should have f(b, ): f 0^) = b(: b^ which is not the case. CHAPTER V THE EQUIVALENCE OF THE AXIOM OF CHOICE, ZORN'S TEWk AND THE WELL-CRDERING THEOREM We know that if we prove the equivalence of the axiom of choice and Zorn's lemma and the equivalence of the axiom of choice and the well-ordering theorem, then the equivalence of the wen- ordering theorem and Zorn's lemma automatically follows. axiom of choice and Zorn's lemma. and a lemma. Let us consider the We must begin with a definition This information is as follows: Definition. relative to Suppose S is a non-null partially ordered set and suppose that every totally ordered subset of S has a least upper bound in S. Suppose further that there exists a function ^ which maps S into S ^~ x=S&.ci(x) for an xe S, and finally, suppose there exist no elements x and y •* Let ok x <-< y -< ^.(x). be any fixed element of S. A subset L of S with the properties that: (a) <k€ L. (b) X £ L =S>o|(x) € (c) l. If C s L is a totally ordered subset of S, then l.u.b. C e L. 21 22 will be called an M«t -chain" or simply a chain if the element is understood. If I, and Lxare chains, then so is L,f\ Ljj» the intersection of all chains, i.e., K =• HL. Let K be Then K is the smallest chain, that is, if L is any chain, then K £= L. The element 4. is the first element of K. Proof. Let M be the set of all elements x in S^«i^ x. Clearly M is an «•* -chain. Therefore, K -£s M, and this proves the assertion. Lemma. The set K introduced above is well-ordered with respect to^- and, <^ (l.u.b. K) rz l.u.b. K. Proof. (A) We shall prove the statement: If be K is comparable with every x e K, then so is <\ (b)» Let L consist of all elements x in K ^ x and \ (b) are comparable. We shall show that L is a chain. Clearly L &. K. The element must be in L since 4 is comparable with every element in K («< is the first element of K). Now we shall prove that x £ L=5^ (x) 6. L. arbitrary element in L. Let x be an Then either ^ (b) * x or Now since x e K, we know that x^.<\(x). then o^ (b) :^L c^ (x) and thus «^ (x) e L. Hence if <k (b)=d x, 23 Now we show, even the case x-<<^(b), that <^ (x) €. L. Since b is comparable with every x 6. K and <^ (x) e K, we know that b £ <j (x). Hence either*^ (x) z£ b or b-<<^(x). Now suppose b<<^(x)• Then either xVb or x-<b since b is comparable with every x in K. If x-<b, then x -< b-<«^(x), which contradicts the assertion about <^. • Thus xXb, that is, either x -= b or x > b. If x =b, then «^(x) =«^ (b) and since «^. (b) e have -^ (x) £ L. L, we must If x >b we have b-< x-oj(b), which is the same contradiction as the one before regarding<^ • Now we must consider the caseo^(x)=£-b. since we have b=£ % (b) and therefore, This is simple, °± (x) ^. «^(b)» Since «^, (x) and *\(b), are comparable, <^,(x) is in L. Hence we see that if x fe L, then «^.(x) & L. We have, there fore, proved the two requirements that L is a chain. To prove the third, let C£ Lbe any totally ordered subset of L and let Jf-i Now if"^ since V l.u.b. C. an x in C ^ x ^Gi(b), then we must have *^<^(b), is an upper bound. But if every x in C has the property that x-< %(b), then Y -^ (b) is a bound since «^ (b) is a bound. Since V we have * fc L. is comparable with<=^ (b) in either case, This completes the proof of the statement 2U that L is a chain. Thus K&- L. But L£K by definition. Hence, L = K, which proves the statement (a). We shall next show that: (B) K is totally ordered. Let M be the set of all elements in K which are comparable with every element of K. In particular oi e, M. If b Q M, then «^ (b) e M by statement (A) above, let C&M be totally l.u.b. C. ordered subset of M and let* be Now if y £ C, then y is comparable with every x in K, since y € 11. Consider any fixed x in K. C ^ x=^.y, then, of course, x ^Jf. If not, then y^. x for all y 6 C and x is an upper bound for 0. In either case* & M. Hence K Ss. M. If"^ a y in ThenV^x. Thus we have shown that M is a chain. But M ■£* K by construction. Therefore, K ~ M and statement (A) is proved. Finally we shall shows (C) K is well-ordered. Let N& K be any non-null subset of K. IfeieN, then is the first element of N and our statement (G) is true* Suppose^- & N. Let Q be the set of all elements x in K x =i u for all u s. N. Clearly ck 6 Q» Let P be any subset of Q. p -< l.u.b. P. Then B Then P is totally ordered. We shall show that p e. Q. a u in N 3 p >> u. Let For suppose p £ Q. But every u in N is an upper bound for Q, so that l.u.b. Qi. u -<* p. 25 But p £. Q. Hence l.tub. P*< p, Taihich is a contradiction. Therefore, p c Q. Thus the set Q satisfies conditions (a) and (c) for a chain. chain. If condition (c) is true, then Q is a Suppose for the moment that this is not true. there must be an x in Q ^ *J (x) 4 Q. Then (This is the only possibility since conditions (a) and (c) of an <H -chain are true). Let q "= l.u.b. Q. If x is not equal to q, we assert that °± (x) €: <^ Q. For if (x) &. Q, then there would exist a u in N* u-S^(x). But x £ Q. Hence x ^. u. Equality may be rejected since x— u *==?x e W, and since x e Q we infer x— q. Hence x-< u-<<^(x), which is again a contradiction. We have now q <= Q and % (q) e Q. Since q e Q, we must have q ^.u for all u <=. N and since <J (q) fe Q, we must have °^ (q) y u for some u e. M. u in N. Mow if q 4 N, then q < u for every Then for some u in N. This is a contradiction* Hence q e N and q is the first element of N. Thus K is well-ordered and we have proved proposition (C). This also proves the first part of the lemma. Now we must show that <^ (l.u.b. K)= l.u.b. K. l.u.b. K. that k€. K. Let K Since K is a chain, and totally ordered, we know (Conditions (a) and (c) of an-I -chain). 26 Further, <Y(k) 6- K by condition (b) of axi^L -chain and k :^(k) by definition of the «^ function. which completes our proof. Hence k = <^(k), Q. E. D. ^?;17U-lY67" We are now ready to prove the equivalence of these two theorems* Let us re-state each* Axiom of Choices Every set has a choice function. Zorn's Lemma (first form): Every totally ordered set contains a maximal totally ordered (relative to-S) subset* Theorem* Proof. The axiom of choice =£> Zorn's lemma (first form). Let S be a set partially ordered with respect to and let Cf be a choice function for S. Let us consider C, the set of all totally ordered subsets of S. For each A £■ C, let F(^ be the set of all elements x inS> AU tx\ is still totally ordered set. one A&S, the set F^ We shall show that for at least is the null set. Let y be a mapping of C into G defined by the equations <^ (A) O A <\ (A) if Ffc+<J> if F^ -Ct> We see that we have fulfilled the conditions of our lemma with C playing the role of S relative to-Q , e^ playing the role of ov and any A in G will play the role of oL . Now from the second part of the lemma we know "^ an element A in C For this element A we have F^+ -<fj so that A is a maximal totally ordered subset of S. 27 Conversely, Zorn's lemma => the axiom of choice. let S be a non-null set. Again Now in certain subsets A of S some choice functions can be defined, for example, the finite subsets* Consider the set E of all pairs (A,Cp), where A is a subset of S and Of is a choice function for A* We define the following order relation for E: (A,Q) =^ (U,T) is to mean: A S. U, and for every subset X ^. U we have «f~ (X) —Cp(A P\ X). From our definition of we know that =^. is an order relation. ordered set. Thus E is a partially Now, by Zorn's lemma (first form), let C be a maximal totally ordered (relative to iL ) subset of E. Consider A*^- <-> A, (A,Co) e C, and let (A,, Q? ) and (Aa, C£ ) be any two elements of C. The elements (A,,O> ) and (Aa,Cp^) are comparable, since C is totally ordered. For the sake of argument let (A,,Cp, ) -k (A^CfJ. Let us consider now any X Q- it ♦ A,. r\ TL*4>, then A^rs X^^and cp^k^n x) ^^(a, r\ (Aa r\ xJJ^q^a, r\ x). With the above equation in mind we give the following definition: Let X be any non-null subset of A • (A,Q> ) t C * kf\ X *<b. define <&*~ by Choose any Clearly such A»s exist. Then we If 28 q =cp (ah x). The functionQ? is a choice function for A+. X) =cPP,(Al r\ X) show that Q>* (X) is independent of the choice of A. .'. (A~,Q> ) <s E. Furthermore, (A, q>) ^ (A*", c/ ) for all (A, Q) & C. Hence A \J %AW,Q+ )} is also totally ordered. Since C is a maximal, (A+,Q*) must belong to C and (A^jQ ) most be the unique maximal element of C. We show now that A^= S. be an element of S — A . Suppose that A* - S and let J. Let us consider the set A^- A* Vj 5.x} Let us define a choice function for A For every non-null subset S of A as follows: let ^ O+(av n x), if x =+ <L ^ , if X -S. Clearly CP^ is a choice function for A. since A+ -< A^» But Thus C U ^(A^, C^) would be totally ordered which contradicts the maximal character of C. Thus A = S* Q. E. D. In Chapter IV, in order to prove the wel!L- ordering theorem, we made use of the axiom of choice. the well-ordering theorem. We also showed that the axiom of choice Now we have only to show that the well- ordering theorem j=r>the axiom of choice. Theorem. Proof. If a set is well-ordered, then it has a choice function. If A £ S is a non-null subset of S, then 1 has a first element, a, since S is well-ordered. Let C% (A) — a. The 29 mapping C-P is a choice function for S. Q. E. D. You can see, then that we have proved the following: The axiom of choice, Zorn's lenana, and the well-ordering theorem are equivalent. Q. £. D. B3BU0GRAPHI 1. Casper Gof flnan, Real Functions, New York, Rinehart and Company, 1953. 2. Kenneth S. Miller, Elements of Modern Abstract Algebra, New York, Harper and Brothers Publishers, 1958. 3. Paul R. Halmos, Waive Set Theory, Mew York, D. Van Nostrand, I960. U. E. Kamke: £. Patrick Suppes, Axiomatic Set Theory, New York, D. Van Nostrand Translated by Frederick Bagendhl, Theory of Sets, New York, Dover Publication, Inc., I960. Company, Inc., I960. 30
© Copyright 2026 Paperzz