Forcing method
A special method for constructing models of axiomatic set theory. It was proposed by P.J. Cohen in 1963 to prove the compatibility of the negation of the continuum hypothesis, , and other set-theoretic assumptions with the axioms of the Zermelo–Fraenkel system \mathsf{ZF} ([1]). The forcing method was subsequently simplified and modernized ([2]–[6]). In particular, it was found that the method was related to the theory of Boolean-valued models ([2], [3]) and Kripke models ([6]).
The central concept of the forcing method is the forcing relation p \Vdash \phi. (This is to be read as “the condition p forces the formula \phi ”.)
The definition of the forcing relation is preceded by the specification of a language \mathcal{L} and a partially ordered set \mathbb{P} of forcing conditions p with an order relation \leq . The language \mathcal{L} may contain variables and constants of different sorts (or types).
The construction of the model of \mathsf{ZF} proposed by Cohen in which \mathsf{CH} is violated proceeds as follows. A set M is called transitive if x \in M \Rightarrow x \subseteq M.
Let M be a denumerable transitive set that is a model of \mathsf{ZF} , and let \lambda \in M be an ordinal number in the sense of von Neumann, i.e., \lambda = \{ \alpha \mid \alpha < \lambda \} . Let A \subseteq \lambda \times \omega_{0} be an arbitrary set (possibly A \notin M ), where \omega_{0} denotes the first infinite ordinal number. If X is a transitive set, then let \operatorname{Def}(X) denote the set of all X -definable subsets, i.e., \operatorname{Def}(X) = \operatorname{Def}(X,\in \!\! |_{X \times X}) . A process similar to that of defining Gödel constructive sets is used to define, via transfinite induction, a set {M_{\alpha}}[A] for any ordinal number \alpha : {M_{\alpha}}[A] \stackrel{\text{df}}{=} \bigcup_{\beta < \alpha} \operatorname{Def}({M_{\beta}}[A]) \cup \{ x \in M \cup \{ A\} \mid x \subseteq {M_{\beta}}[A] \}.
Let M[A] \stackrel{\text{df}}{=} {M_{\alpha_{0}}}[A] , where a_{0} \stackrel{\text{df}}{=} \{ \alpha \in M \mid \alpha \text{ is an ordinal number} \} . The model of \mathsf{ZF} in which \mathsf{CH} is violated is to be found among models of the type M[A] . Let \lambda be an ordinal number such that the statement “ \lambda is the second uncountable ordinal” is true in M .
The set of forcing conditions \mathbb{P} and the relation \leq are defined by the following equivalences: (a) p \in \mathbb{P} \iff p is a function defined on some finite subset of \lambda \times \omega_{0} , with values in \{ 0,1 \} ; (b) p \leq q \iff q is an extension of p . The language \mathcal{L} considered is a so-called ramified language, with many types of variables (each \alpha \leq \alpha_{0} having its own type of variables running through the set {M_{\alpha}}[A] ) and with names (i.e., individual constants) for each set from M[A] . If x \in M , then the name of x is denoted by x^{\vee} . Let a be the name of the set A . The forcing relation p \Vdash \phi is introduced by a definition involving transfinite induction, which has, in particular, the following characteristics:
- (p \Vdash \langle \delta,n \rangle^{\vee} \in a) \iff (p(\langle \delta,n \rangle) = \mathbf{1}) ;
- (p \Vdash \neg \phi) \iff \neg (\exists q \geq p)(q \Vdash \phi) ;
- (p \Vdash (\phi \lor \psi)) \iff ((p \Vdash \phi) \lor (p \Vdash \psi)) ;
- (p \Vdash (\phi \land \psi)) \iff ((p \Vdash \phi) \land (p \Vdash \psi)) ;
- If \alpha is the type of the variable x , then (p \Vdash (\exists x) \phi(x)) \iff (\exists c \in C_{\alpha})(p \Vdash \phi(c)) , where C_{\alpha} is the set of all constants of type \alpha .
A sequence of forcing conditions in \mathbb{P} , p_{0} \leq \ldots \leq p_{n} \leq \ldots, is called complete if for any closed formula \phi of \mathcal{L} , one has \exists n \in \omega_{0}: \qquad (p_{n} \Vdash \phi) \lor (p_{n} \Vdash \neg \phi). The countability of the set of all closed formulas of \mathcal{L} , together with (2), makes it possible to demonstrate the existence of a complete sequence that begins with an arbitrary forcing condition p_{0} \in \mathbb{P} .
A set A contained in \lambda \times \omega_{0} is called generic with respect to the model M if there exists a complete sequence (p_{n})_{n \in \omega_{0}} such that \displaystyle \bigcup_{n < \omega_{0}} p_{n} is the characteristic function \chi_{A} of A . The following two facts about generic sets and the forcing relation are of fundamental importance:
- If A is a generic set, then M[A] \models \phi \iff (\exists p \subseteq \chi_{A})(p \Vdash \phi), where M[A] \models \phi means that formula \phi is true in M[A] .
- The relation p \Vdash \phi(c_{1},\ldots,c_{n}), \qquad (\text{where } c_{1},\ldots,c_{n} \text{ are constants in } \mathcal{L}) regarded as a relation among p,c_{1},\ldots,c_{n} , can be expressed in the model M .
In view of the aforementioned facts, it is sufficient to show that the statement (\forall p)(p \Vdash \neg \neg \phi), \qquad \text{i.e.}, \qquad (\forall p)(\exists q \geq p)(q \Vdash \phi), is true in the model M , i.e., to prove that M[A] \models \phi . The verification of the validity of the axioms of \mathsf{ZF} and \neg \mathsf{CH} in the model M[A] is based upon this proposition. The verification of \neg \mathsf{CH} in M[A] also involves the use of special properties of the set of forcing conditions, which makes it possible to show that:
- If the ordinal numbers \delta_{1},\delta_{2} < \lambda are unequal, then one has (\forall p)(\exists q \geq p)(\exists n < \omega_{0})(q(\langle \delta_{1},n \rangle) \neq q(\langle \delta_{2},n \rangle)), that is, A_{\delta_{1}} \neq A_{\delta_{2}}, \qquad \text{where } A_{\delta} \stackrel{\text{df}}{=} \{ n < \omega_{0} \mid \langle \delta,n \rangle \in A \}.
- M[A] \models (\lambda \text{ is the second uncountable ordinal}) .
There is the following relation between forcing and Boolean-valued models: If one introduces the notation \begin{gather} \| \phi \| \stackrel{\text{df}}{=} \{ p \in \mathbb{P} \mid p \Vdash \neg \neg \phi \}, \\ \mathbb{B} \stackrel{\text{df}}{=} \{ X \subseteq \mathbb{P} \mid (\forall p)((p \in X) \iff (\forall q \geq p)(\exists r \geq q)(r \in X)) \}, \end{gather} then (\mathbb{B},\subseteq) is a complete Boolean algebra, and \| \phi \| \in \mathbb{B} is the Boolean value of the formula \phi . Hence, to specify a partially ordered set (\mathbb{P},\leq) and to define the relation p \Vdash \neg \neg \phi is equivalent to constructing some Boolean-valued model \mathfrak{M} . The analysis of the proof of theorems of the type M[A] \models \phi_{0}, where \phi_{0} is an axiom of \mathsf{ZF} or \neg \mathsf{CH} , leads to the conclusion that the formulas of \mathsf{ZF} expressing the theorem (\forall p)(p \Vdash \neg \neg \phi_{0}), i.e., \| \phi_{0} \| = 1_{\mathbb{B}} , are deducible from the axioms of \mathsf{ZF} . Therefore, \mathfrak{M} is a Boolean-valued model for \mathsf{ZF} + \neg \mathsf{CH} , constructed solely with the tools of \mathsf{ZF} . The assumption of countability of a transitive set that is a model of \mathsf{ZF} and the concept of a generic set are both inessential in the relative consistency proof.
It has been shown that the construction of the Boolean-valued model may be simplified ([2], [3], [5]). In particular, the introduction of the ramified language \mathcal{L} is not necessary. A generic model M[A] may also be constructed as follows ([4]):
A subset X of a partially ordered set (\mathbb{P},\leq) is called dense if (\forall p)(\exists q \geq p)(q \in X). Let \mathbb{P} and the relation \leq be elements of some denumerable transitive set M that is a model of \mathsf{ZF} . A subset G \subseteq \mathbb{P} is called an M -generic filter if
- ((p \in G) \land (q \leq p)) \Rightarrow (q \in G) ;
- ((p \in G) \land (q \in G)) \Rightarrow (\exists r \in G)((p \leq r) \land (q \leq r)) ;
- ((X \in M) \land (X \text{ is dense in } \mathbb{P})) \Rightarrow X \cap G \neq \varnothing .
Let G be an M -generic filter in \mathbb{P} . As M is denumerable, G exists. In general, G \notin M . A relation \in_{G} is then defined by the equivalence x \in_{G} y \iff (\exists p \in G)(\langle x,p \rangle \in y), where x and y are arbitrary elements of the model M .
Let a function F_{G} be defined on M by the equations {F_{G}}(y) \stackrel{\text{df}}{=} \{ {F_{G}}(x) \mid x \in_{G} y \}, and let N_{G} \stackrel{\text{df}}{=} \{ {F_{G}}(x) \mid x \in M \}. If \phi is a closed formula of the language \mathsf{ZF} , supplemented by constants to denote each set from M , then one defines p \Vdash \phi \iff (\forall G \ni p)(G \text{ is an } M\text{-generic filter} \Rightarrow N_{G} \models \phi). It can be now shown that:
(I) N_{G} \models \phi \iff (\exists p \in G)(p \Vdash \phi) .
(II) The relation p \Vdash \phi(c_{1},\ldots,c_{n}) can be defined in the model M for each formula \phi .
By using solely (I) and (II), and basing oneself on the fact that M is a model of \mathsf{ZF} , it is possible to establish that N_{G} is a model of \mathsf{ZF} . If \mathbb{P} is defined by the equivalences (a) and (b) given much earlier on, then one has N_{G} \models \neg \mathsf{CH} , that \displaystyle \bigcup G is the characteristic function of some set A \subseteq \lambda \times \omega_{0} , and M[A] = N_{G} . The relation p \Vdash \phi , which is definable in M , does not satisfy points (3) and (5) of Cohen’s definition of the forcing relation. One has instead (p \Vdash (\exists x) \phi(x)) \iff (\forall p' \geq p)(\exists p'' \geq p')(\exists a \in M)(p'' \Vdash \phi(a)). If it is assumed that \| \phi \| \stackrel{\text{df}}{=} \{ p \in \mathbb{P} \mid p \Vdash \phi \}, then a Boolean-valued model for \mathsf{ZF} + \neg \mathsf{CH} is obtained that is definable in M and has a Boolean algebra \mathbb{B} that is identical with that of Cohen.
Therefore, the forcing method consists, in fact, of constructing a \mathbb{B} -valued model and a homomorphism, which preserves certain infinite unions and intersections of the algebra \mathbb{B} , into the two-element algebra \{ 0,1 \} . For applications of the forcing method in set theory see, for example, [2].
References
[1] | P.J. Cohen, “Set theory and the continuum hypothesis”, Benjamin (1966). |
[2] | T.J. Jech, “Lectures in set theory: with particular emphasis on the method of forcing”, Lect. notes in math., 217, Springer (1971). |
[3] | G. Takeuti, W.M. Zaring, “Axiomatic set theory”, Springer (1973). |
[4] | J.R. Shoenfield, “Mathematical logic”, Addison-Wesley (1967). |
[5] | Yu.I. Manin, “The problem of the continuum”, J. Soviet Math., 5: 4 (1976), pp. 451–502; Itogi Nauk. i Tekhn. Sovrem. Problemy, 5 (1975), pp. 5–73. |
[6] | M.C. Fitting, “Intuitionistic logic, model theory and forcing”, North-Holland (1969). |
Comments
For a modern elementary introduction with applications, see [a1]. A quicker, though incomplete, account is [a3]. [a2] presents the Boolean-valued approach mentioned above.
References
[a1] | K. Kunen, “Set theory, an introduction to independence proofs”, North-Holland (1980). |
[a2] | J.L. Bell, “Boolean-valued models and independence proofs in set theory”, Oxford Univ. Press (1977). |
[a3] | J.P. Burgess, “Forcing”, J. Barwise (ed.), Handbook of mathematical logic, North-Holland (1977), pp. 403–452. |
[a4] | T.J. Jech, “Set theory”, Acad. Press (1978), pp. 523ff. (Translated from German) |
Forcing method. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Forcing_method&oldid=51543