Some proofs of independence in axiomatic set theory (Q3237140)

\textit{K. Gödel} [The consistency of the continuum hypothesis, 3rd ed. Princeton: Princeton University Press (1953; Zbl 0051.28903)] proved the consistency of the axiom \(V = L\) in his version \(G\) of the Zermelo-Fraenkel-v. Neumann-Bernays set theory by defining a function \(F\) with domain the class \(ON\) of all ordinal numbers and showing that the class \(F''\, ON\) of all values of \(F\) formed a universe of sets which, with suitable definitions of ``class'' and ``member'', formed a model of \(G\) which also satisfied \(V = L\). It has been noted by several authors [\textit{C. D. Firestone} and \textit{J. B. Rosser}, The consistency of the hypothesis of accessibility., J. Symb. Logic 14, 79 (1949) (abstract); \textit{J. B. Rosser}, Deux esquisses de logique. Paris: Gauthier-Villars (1955; Zbl 0064.00701); \textit{A. Mostowski}, Fundam. Math. 36, 143--164 (1949; Zbl 0039.00802); \textit{P. Bernays}, J. Symb. Log. 13, 65--79 (1948; Zbl 0030.11502); the reviewer, J. Symb. Log. 17, 225--237 (1952; Zbl 0048.28105)] that a similar construction carried out, not for all ordinal numbers, but for all ordinals less than a certain ordinal \(\beta\), can be used to yield proofs of independence. In fact the set \(F''\, \beta\) defines a model \(M_\beta\) which, for \(\beta = \omega\) satisfies all the axioms of \(G\) except the axiom of infinity; for \(\beta = \omega^\omega\) satisfies all except the replacement axiom; and for \(\beta = \) the first inaccessible ordinal (assuming these exist) satisfies all the axioms of \(G\) but not the postulate that there exist inaccessible ordinals. Thus assuming \(G\) to be consistent we obtain in this way the independence from the other axioms of the axiom of infinity, the axiom of replacement, and the independence from \(G\) of the axiom that there exist inaccessible ordinals. In the existing proofs of these results two things have been shown: 1. \(M_\beta\) satisfies all the remaining axioms, 2. \(M_\beta\) does not satisfy the axiom whose independence is to be proved. The essence of the present paper is that it is not necessary to prove 2. The argument is as follows: from the statement that there exists a model for \(G\) defined by a set we can, in the usual Tarski manner, obtain (within \(G\)) a truth definition for \(G\) and hence a proof of the consistency of \(G\). So, by the Gödel incompleteness theorem (provided \(G\) is consistent) it is not possible to prove from \(G\) that there is any model defined by a set which satisfies all the axioms of \(G\). Hence, having proved from \(G\) that \(M_\beta\) satisfies all the axioms of \(G\) except one it follows without further proof that it cannot be proved to satisfy the remaining axiom, i. e. it follows that this is independent.