Download as pdf
Download as pdf
You are on page 1of 8
‘THE SENTENTIAL CALCULUS WITH INPINITELY LONG EXPRESSIONS Cosuthored with Dana Scott Colloquium Mathematicum, vol. 6 (1988), pp. 165-170, THE SENTPNTIAL CALCULUS WITH INFINITELY LONG EXPRESSIONS BY D. SCOTT (PRINCETON, N. J.) aN A, TARSK1 (BERKELE |, CALIF.) ‘The results given in this note are essentially the observations pre viously made by the authors concerning the equational identities in Boo- lean algebras with infinitary operations and in particular the relation between the identities holding in the two-clement algebras and those holding in arbitrary Boolean algebras. These remarks, however, are re- formulated here in terms of the syntax of the sententia! calculus with infinitely long formulas. It is to be noted that the discussion of the sen- tential calculus is part of a comprehensive study concerning the syntax of the predicate logic with infinitely long expressions which has been undertaken and carried out by Mrs, Carol Karp, The results of Mrs. Karp have not: yet been published, but they were presented in the seminar in the foundations of mathematics conducted by L. Henkin and A. Tarski at the University of California at Berkeley in the fall semester of 1056 (?), Let a and f be cardinal numbers. (We shall identify the cardinals with the initial ordinals of their respective number classes). ‘The sentential caleuli considered will have f different sentential variables and will per mit the formation of well-ordered conjunctions and disjunctions in all Tengths less than a, The case where a = f = is simply the ordinary cal- culus. The case where @ = @, retains much analogy with the ordinary case and is examined in detail. The cases where a > w, and B > o present some peculiarites which require the reformulation of the definition of (}) This note ia a summary of a lecture given by the authors at tho Summer Institute of Symbolic Logic at Cornell Univorsity in July 1957; it appeared under tho same titl, though in @ somewhat shorter vorsion, in Summaries of talks presented at the Summer Institute of Symbolic Logie in 1967 ct Cornell University. vol. 1, p. 83-80 (mimoographod). Th» results of this note were obtained and the noto was prepared for publication whilo Tarski was working on @ resoarch project in the foundations. of mathomatics sponsored by the National Seience Foundation. For a Boolean algebraic formulation of the results soe the abstracts Scott [7] and Tarski [10]. Some remarks ‘concerning the predicate logic with infinitely long expressions (which is not discussed in this note) ean be found in Tarski (11) a theorem. The exact statement of the result is given, but the proof will be omitted. ‘As usual there are two aspects of the formal systems to be consi- dered; the syntactic and the semantic, The syntactic details involve the no- tions of a well-formed formula and a theorem, while the semantic consi- derations make use of the substitutions of 1s and F's for the sentential variables in order to define a tautology. For a «, this is generally not the case, unless the notion of theorem is streng- thened. As basic symbols for the calculi we use ->,~, A, V standing for implication, negation, conjunction, and disjunetion. Brackets [ and ] are also used. The cardinal f will be fixed for the discussion, and the symbols Poy Pay +++ Dey ove (Where & B] and ~A; (ill) if Ag, Aas ooey Ags oe #8 @ well-ordered sequence of a-wf{s of type less than a, then A [Ap Ay-.-Ageen] and V[ApAy.. Agee] are a-wffs, ‘Thus, if a = @, we are considering formulas involving finite eonjune- tions and disjunctions of arbitrary lengths; while, if a =, then denu- merable conjunctions and disjunctions are permitted. In turns out to be superfluous to consider certain of the cardinals: namely, the so-called singular numbers, 4. ¢., those ordinals that can be written as an ordinal sum of smaller numbers over a smaller index. The first transfinite example among cardinals is of course @, = S'o,. Notice that any conjunction of the form A[4p4,...4;...J where £<@, should be equivalent in meaning to a conjun jon of type @ of conjunctions of the various types q a8 follows: ALA [Adi DA [Ae dan Tee A ay Aogtreod + Hence, there will be an equivalence between the sentential calculi of a, -wifs and o,,,-Wifs. The details of this equivalence for any singular number wo leave to the reader and assume henceforth that the cardinal a is regular (i. e, non-singular) and infinite. In order to simplify the writing of formulas we adopt the following notation: [AAB] for A[AB), [AVB] for V[AB], AA, for N Ap Agee Agee]; fey VA, for V [Ay Age Aged fey where 7 is the type of the sequence mentioned. DEFINWION 2. The class of atheorems is the least class of a-wffs closed under the following rules: (i) if A,B,C are a-wffs, then [A+ [B> AJ], [4 + [B= 0) >[[4>B) > U4 cil] and [[~B+~4]> [A > B]] are a-theorems; (ii) if Agy Ay, -.-, Agr es #8 a well-ordered sequence of type y 4,] and [A,—> V A,) are a-theorems for each n <7} fey ter (iii) ¢f [A> B] and A are a-theorems, then so is By (iv) ff Ag Ay vsy Ags es i8 @ well-ordered sequence of type y A,] is an aetheorem for all n B] 8 an a-theorem for all <7, then [V Ay-> B] is an a-theorem (2). fey Derrirton 3. Let y Big llé < @) ee ty, is w, a-consiatent (3). Before we can define the notion of a tautology, it must be verified that 7°s and F’s can be substituted for the sentential variables in the proper way. Proceeding by a transfinite induction based on Definition 1, the following lemma should be proved: LEMMA 2. If the function f is a substitution defined on the variables Py taking the values T and P, then there is a unique extension f° of f to the class of all a-wf{s such that (i) f* takes on only the values T and F; (ii) f*(LA -> BY) if and only if either {'(A) =F or [*(B) = T; (ii) f(A) =P if and only af f"(A) =F; liv) (A Ad = 2 af and only if f(A.) = 2 for all F< 5 Sy (v) PUY Ay) = 2 if and only if f(A.) = for some & < y. ey DEFINITION 4. An a-toff A is an a-tautology if and only if f*(A) = T for all substitutions f. TimoREM 1 (COMPLETENESS THEOREM POR COUNTABLE caLctut). If a < oy, then an a-wff is an a-theorem if and only if it is an a-tautology (*). Proof (in outline). It is obvious from the definitions that every a-theorem is an a-tantology. Assume then that 4 is an a-wff that is not an a-theorem. Let § be the least class of a-wifs containing A and closed under the operation of taking subformulas. ‘The essential point of the argument is that 8 is ab most denumerable as a consequence of the hypo- thesis a <,. Thus there exists asequence By, i< @ and a. Of course if f were finite, there would be only a finite number of inequivalent formulas on the basis of the axioms given in Definition 2. Whence, all infinite operations could be eliminated in favor of the finite ones. Thus, assume that p > o. ‘That the axioms given above are inadequate for generating all tautologies is shown by the following counter-example: Let a actually be of greater power than that of the continuum, Let g,, where & < y, be a well-orde ying of all functions from the integers to {0,1}. Then the formula Uf. (en V Pn >, Panvege is an a-tautology that is not an a-theorem (‘). It would seem still to be an open question whether such a counter-example can be given for all a > « without the aid of the continuum hypothesis, Nevertheless, an adequate axiom system can be given for all cardinals a by adding en additional clause to the definition of a-theorems. Durinrri0n 5. The class of a-theorems in the strict sense is the least class of a-tffs closed under the rules of Definition 2 as well as the fol- lowing additional rule: (v) fy o it ean be shown that such a stronger result fails unless possibly a is a strongly ina- coessible cardinal number. Whether the stronger theorem is true in the inaccessible case isan open question (P 250) seemingly involving fundamen- tal set-theoretical problems(*). However, certain stronger results are possi- ble: for example, it should be clear from the proof of Theorem 1 that every at most denumerable and «,,-consistent class of @,-wiis has a substi- tution giving all formulas in the class the value 7, BIBLIOGRAPHY [1] C.C. Chang, On the representation of a-complete Boolean algebras, ‘Teans- actions of thn American Mathematical Socinty 85 (1957). p. 208-218. [2] P. Erdés, and A. Tarski, On familie of mutually exclusive sts, Annals of Mathomatics 44 (1843), p. 315-320. (3] 8. Feferman, Resiew of (5}. Journal of Symbolio Logic 17-(1953). p. 72 [4] L. H. Loomis, On the representation of o-complete Boolean algebras, Bulletin of tho American Mathomatical So ioty 68 (1947), p. 757-760, [5] H. Rasiowa and B. Sikorski, A proof of the completinses theorem of Godt, Pundamanta Mathomatieay 37 (1980), p. 193-200. {0} L. Rieger, On free ny complete Boolean alg:bras, Fundamonta Mathomatican 38 (1951), p. 85:62. [7] D. Scott, A new charactsization of a-repreaottable Boolean algebras, Ab tract, Bulletin of the Amoriean Mathematical Sosioty 61 (1953), p. §22-523. [8] R. Sikorski, On Ue representation of Bool:an alg:bras as fields of sets, Pundamenta, Mathomaticao 33 (1848), p. 247-238, [9] — 4 note on Rieger's paper, Fondanenta Mathomatione 38 (1951).p- 63-55. [10] A. Tarski, Metamathematical proofs of some representation theorems for Boolean Alg:bras, Abstract, Bulloti of tho Auorican Mathomatieal Sasisty 61 (1935), pe 623-524, [11] — Remarks on predicate logic with infinitely long expressions, Colloquium Mathomatioum 6 (1938), p. 171-176. PRINCRTON UNIVERSITY ‘THE UNIVERSITY OF CALIVORNIA AT BERKELEY Repu par ls Rédaction t2 21. 12. 1957 sible numbers (*) ‘Pais problom ia aireotly rolated to thos» probloms about inace formulated at the end of Erdés-Taroki (2).

You might also like