A model of \(\mathsf {ZFA}+ \mathsf {PAC}\) with no outer model of \(\mathsf {ZFAC}\) with the same pure part (Q1756492)

The theory $\mathsf{ZFA}$ is a modified version of Zermelo-Fraenkel set theory ($\mathsf{ZF}$) admitting objects other than sets, the so-called atoms. The domain of a model of $\mathsf{ZFA}$, $\mathcal{P}^{\infty,*}(A)$ (in the paper's notation), is defined via a hierarchy over a given set $A$ of atoms (which is a non-empty set such that no member of $A$ is in the transitive closure of any other member), indexed by ordinals, in a similar manner with the cumulative hierarchy $V_{\alpha}$ of the Von Neumann universe: \[ \begin{aligned} &\mathcal{P}^{0,*}(A)=A;\\ &\mathcal{P}^{\alpha+1,*}(A)=(\mathcal{P}^{\alpha,*}(A)\cup\mathcal{P}(\mathcal{P}^{\alpha,*}(A)))\setminus\{\emptyset\};\\ &\mathcal{P}^{\beta,*}(A)=\bigcup_{\alpha<\beta}\mathcal{P}^{\alpha,*}(A)\qquad (\beta\text{ limit});\\ &\mathcal{P}^{\infty,*}(A)=\bigcup_{\alpha\in\mathrm{Ord}}\mathcal{P}^{\alpha,*}(A), \end{aligned} \] where an arbitrary element of $A$ is let to represent the empty set, and the other members of $A$ represent atoms. Sets in a model of $\mathsf{ZFA}$ whose transitive closures exclude atoms are called pure sets; these sets form an inner model of $\mathsf{ZF}$. The axiom $\mathsf{PAC}$ asserts that this inner model satisfies the axiom of choice ($\mathsf{AC}$). In this paper, the authors construct a model of $\mathsf{ZFA}+\mathsf{PAC}$ such that no outer model of $\mathsf{ZFAC}$ (= $\mathsf{ZFA}+\mathsf{AC}$) has the same pure sets. Their method of producing models of $\mathsf{ZFA}$ is different from the usual Fraenkel-Mostowski method; the domains of their models are subclasses of $\mathcal{P}^{\infty,*}(A)$ (for a given set $A$ of atoms) which are constructed over $A$ using certain elements of $\mathcal{P}^{\infty,*}(A)$ as predicates. In particular, given sets $X$ and $B$ in $\mathcal{P}^{\infty,*}(A)$, the authors let Def$_{B}(X)$ denote the collection of non-empty subsets of $X$ which are definable over $X$ using parameters from $X$ and predicates corresponding to members of $B$. Then they define: \[ \begin{aligned} &U_{0}^{A,B}=A;\\ &U_{\alpha+1}^{A,B}=U_{\alpha}^{A,B}\cup\mathrm{Def}_{B}(U_{\alpha}^{A,B});\\ &U_{\beta}^{A,B}=\bigcup_{\alpha<\beta}U_{\alpha}^{A,B}\qquad (\beta\text{ limit});\\ &U_{\infty}^{A,B}=\bigcup_{\alpha\in\mathrm{Ord}}U_{\alpha}^{A,B}. \end{aligned} \] Given $a\in A$, they let $\mathbf{U}(a, A, B)$ be the model of $\mathsf{ZFA}$ with domain $U_{\infty}^{A,B}$, where $a$ is interpreted as the empty set. The model $\mathbf{U}(a, A, B)$ is, up to isomorphism, the smallest well-founded proper class model of $\mathsf{ZFA}$ with $A\setminus\{a\}$ as its set of atoms and $a$ as its empty set which is closed under intersections with the members of $B$. Every element of $\mathbf{U}(a, A, B)$ is definable in $\mathbf{U}(a, A, B)$ from a finite set of its ordinals, a finite subset of $A$ and finitely many predicates from $B$. Under the above terminology, the main result of this paper is the following: ``In a c.c.c. forcing extension $\mathbf{L}[G]$ of $\mathbf{L}$ there is a model $\mathbf{U}$ of $\mathsf{ZFA}$ of the form $\mathbf{U}(a,A,B)$, for some atom set $A$ in $\mathbf{L}$, some element $a$ of $A$ and some $B$ in $\mathcal{P}^{\infty,*}(A)$, such that the pure part of $\mathbf{U}$ is isomorphic to $\mathbf{L}$ and such that in no outer model of $\mathbf{L}[G]$ is there a model of $\mathsf{ZFAC}$ containing $\mathbf{U}$ whose pure part is isomorphic to $\mathbf{L}$.'' The authors' proof of the above theorem (given in Section 2 of this paper) uses a model-theoretic construction due to \textit{G. Hjorth} [J. Math. Log. 2, No. 1, 113--144 (2002; Zbl 1010.03036)] which produces a sentence in $\mathcal{L}_{\aleph_{1},\aleph_{0}}$ homogeneously characterizing $\aleph_{1}$, and is especially beautiful. In Section 3, the authors discuss this construction, and in Section 4, they explain the necessity of such a construction.