Compactness phenomena in HOD and the Optimality of Magidor’s Covering theorem

Fix A\subseteq \mathbb{P}\cap M a maximal antichain and let p\in \mathbb{P} be arbitrary. Since A is maximal in \mathbb{P}\cap M, there is q\in A that is \mathbb{P}\cap M-compatible with f(p). Let r\in \mathbb{P}\cap M be such that r\leq f(p), q. Since f is a residue we can find p'\in \mathbb{P} such that p'\leq r, p. In particular, p' witnesses the compatibility of p and q, ergo A is a maximal antichain in \mathbb{P} as well.

To establish the second claim we assume that M\prec H_\Theta. Under this assumption, any two conditions in G\cap M are compatible in \mathbb{P}\cap M, and G\cap M is also upwards closed in this latter poset (i.e., G\cap M is a filter). The fact that G\cap M intersects all the maximal antichains of \mathbb{P}\cap M (i.e. \mathbb{P}\cap M-genericity) was already established in the previous argument.

Denote \kappa:=|X|^N and let \pi\colon \kappa \to X be a bijection in N. Consider R\subseteq \kappa\times \kappa the relation defined by

Clearly, \pi\colon (\kappa, R)\to (X,\in) is an isomorphism of structures. Since (X,\in) is transitive, \pi must be the Mostowski collapse map on (\kappa,R). Let R^* \subseteq \kappa be the result of applying Gödel’s pairing function to the relation R. Since we assume that \kappa < \Delta(M,N) it follows that R^*\in M and by absoluteness of Gödel’s pairing and of the collapse we deduce that \pi\in M.

Let A\in \mathcal{P}^N(X), and set B:=\pi^{-1}[A]. Use our assumption to infer that B\in M. Since \pi\in M we conclude that A\in M, concluding that \mathcal{P}^N(X)\subseteq M.

To see the second part we argue as follows: If \Delta(M,N) were not an N-cardinal then |\Delta(M,N)|^N<\Delta(M,N) and by the first part, \mathcal{P}(\Delta(M,N))\subseteq M. However, by the very definition of \Delta(M,N), \mathcal{P}^N(\Delta(M,N))\not\subseteq M. A contradiction.

Suppose otherwise and let j:V\to M be an embedding witnessing (\operatorname{HOD},\mathcal{P}(\kappa))-measurability. Let Y\in \mathcal{P}(\kappa)\setminus \mathcal{P}^{\operatorname{HOD}}(\kappa). Since \operatorname{crit}(j)=\kappa, Y\in M\cap \mathcal{P}(\kappa). Note that Y\notin \mathcal{P}(\kappa)\cap \operatorname{HOD}^M, since by assumption \operatorname{HOD}\cap \mathcal{P}(\kappa)=\operatorname{HOD}^M\cap \mathcal{P}(\kappa). Hence M thinks that j(\kappa)>\operatorname{crit}^{\operatorname{HOD}^M}(M), contradiction.

Suppose for the sake of a contradiction that there is no inner model with a measurable cardinal yet \aleph_\omega is the first cardinal \lambda with \mathcal{P}(\lambda)\nsubseteq \operatorname{HOD}.

Let X\in \mathcal{P}(\aleph_\omega) be an offending set. For each n<\omega, X\cap \aleph_n \in \operatorname{HOD}.

For each n<\omega let \rho_n be the rank of X\cap \aleph_n in the canonical well-order of \operatorname{HOD}. Consider R= \{\rho_n\mid n<\omega\}. By the Dodd–Jensen Covering Theorem (cf. [MitchellCoveringHandbook]) there is S\in \mathcal{K}^{\mathrm{DJ}}\subseteq \operatorname{HOD} such that R\subseteq S with |S|\leq \aleph_1.

Let \alpha:=\mathrm{otp}(S) and \pi \colon S \to \alpha be the unique isomorphism; namely, \pi is just the Mostowski collapse map. Since S\in \operatorname{HOD}, by absoluteness of the Mostowski collapse, \pi\in \operatorname{HOD}. Finally, \pi[R]\in \operatorname{HOD} –in that a subset of \alpha < \aleph_\omega=\Delta(\operatorname{HOD}, V)– so that R\in \operatorname{HOD} as well.

From R we can uniformly define X using ordinal parameters.

Let p= (\vec{\kappa}_p,\vec{c}_p, A_p, h_p, H_p) and q=(\vec{\kappa}_q,\vec{c}_q, A_q, h_q, H_q) be as before. For each i<m let c^*_i:=c^p_i\cup c^p_i. Since j(H_p)(\kappa),j(H_q)(\kappa)\in K there is B\subseteq A_p\cap A_q in \mathcal{U} such that H_p(\beta) is compatible with H_q(\beta). Set A^*:=B, h^*=h_p\cup h_q and H^* be the function with domain A^* such that H^*(\beta):=H_p(\beta)\cup H_q(\beta). Let p^* be the vector obtained after replacing \vec{c}_p, A_p, h_p and H_p by \vec{c}^*, A^*, h^* and H^*, respectively. Define q^* similarly.

Clearly, p^*\leq^* p and q^*\leq^* q.

There is a natural isomorphism \Gamma\colon \mathbb{P}/p^*\to \mathbb{P}/q^* that given r\leq p^* sends it to \Gamma(r), the vector swapping the odd Prikry points of p^*, \langle \kappa^{p}_{2(i+1)}\mid i\leq m(p)\rangle, to the odd Prikry points of q^*, \langle \kappa^q_{2(i+1)}\mid i\leq m(p)\rangle, keeping the other entries unaltered. By construction, \pi[\Gamma[G]]=\pi[G] for all G\subseteq \mathbb{P}/p^* generic.

By Proposition 3.2, it suffices to show that every set of ordinals in \operatorname{HOD}^{V[G]}_x is in V[\pi[G]]. Let a\in \operatorname{HOD}^{V[G]}_x be a subset of some \chi\in \mathrm{Ord}. By definition,

for ordinals \alpha_0,\dots,\alpha_{k-1} and a formula \varphi(y, y_0,\dots, y_{k-1},z) in \mathcal{L}_\in.

Let \dot{x} be a \mathbb{Q}-name such that \dot{x}_G=x that is invariant under any isomorphism \Gamma such that \pi[\Gamma[G]]=\pi[G]. Then, we have

Define

Clearly, \bar{a}\in V[\pi[G]] and a\subseteq \bar{a}. We claim that \bar{a}\subseteq a.

Suppose that \beta\in \bar{a}\setminus a. Then, there is p\in \mathbb{P}/\pi[G] and q\in G such that

By extending both p and q if necessary we may ensure that p\in G and q\in \mathbb{P}/\pi[G] and that both have the same length. Note that in these circumstances we can invoke the previous lemma and find conditions p^*\leq^* p, q^*\leq^* q together with an autormorphism \Gamma sending p^* to q^*, such that \pi[G]=\pi[\Gamma[G]]. Since \Gamma does not change the names involved in the formula we get \beta \in a\cap \bar a. Contradiction.

Let \kappa be a measurable cardinal. Without loss of generality, we may assume that our ground model satisfies ``\kappa is measurable”, V=\mathrm{gHOD} and GCH holds on a large enough interval of cardinals. Let \mathbb{P} be the previous poset and G\subseteq \mathbb{P} a generic filter. Let x be a set of ordinals such that V[x]=V[\pi[G]]. Let \mathrm{Code}(x,\aleph_\omega) be the poset coding x into the power-set-function pattern above \aleph_\omega. Let c \subseteq \mathrm{Code}(x,\aleph_\omega) be generic over V[G]. Since \mathrm{Code}(x,\aleph_\omega) is weakly homogeneous and \mathrm{OD}^{V[G]}_x,

where the first inclusion follows from V=\mathrm{gHOD} and the last inclusion follows from the previous lemma. Thus, \operatorname{HOD}^{V[G][c]}=V[\pi[G]].

In particular, \operatorname{HOD}^{V[G][c]} have the same bounded subsets of \aleph_\omega as V[G] in that this is the case between V[G] and V[\pi[G]]. So, to prove that

it suffices to note that \{\kappa_{2n+1}\mid n<\omega\} is not in V[\pi[G]].

Suppose otherwise – that \{\kappa_{2n+1}\mid n<\omega\}\in V[\pi[G]]. Let \tau be a \mathbb{Q}-name such that \tau_{\pi[G]}=\{\kappa_{2n+1}\mid n<\omega\} and p\in G forcing this.

Since A':=\{\alpha\in A_p\mid \text{A_pis unbounded in}\} is measure one, the Mathias criterion of genericity gives n_*<\omega such that \{\kappa_{n}\mid n\geq n^*\}\subseteq A'. To simplify notations, let us assume that n_*=\ell(p)+1.(Otherwise we just extend p with the Prikry points \{\kappa_n\mid \ell(p)\leq n\leq n_*-2\}.)

Let p_0\leq p{}^\curvearrowright (\kappa_{\ell(p)}, \kappa_{\ell(p)+1}) be in G. Given \alpha\in A^p\cap (\kappa_{\ell(p)-1},\kappa_{\ell(p)+1}) distinct from \kappa_{\ell(p)}, let p_1 be a condition such that p_1\leq p{}^\curvearrowright (\alpha, \kappa_{\ell(p)+1}), \ell(p_0)=\ell(p_1), \vec{c}_{p_1}=\vec{c}_{p_0}, A_{p_1}=A_{p_0}, h_{p_1}=h_{p_0}, H_{p_1}=H_{p_0}.

Note that such a condition p_1 always exists in that adding the odd-points (i.e., \kappa_{\ell(p)} and \alpha) to p does not have any effect on the values of H_p.

Applying Lemma 3.17 we obtain an isomorphism \Gamma\colon \mathbb{P}/p_0\to \mathbb{P}/p_1 such that \pi[{\Gamma[G]}]=\pi[G]. Since p_0,p_1\leq p it follows that p_i\Vdash_{\mathbb{P}} ``\tau=\dot{\vec{\kappa}}_{\mathrm{odd}}" where \dot{\vec{\kappa}}_{\mathrm{odd}} is the standard \mathbb{P}-name for the odd members of the Prikry sequence. As such, p_0\Vdash_{\mathbb{P}}``\check{\kappa}_{\ell(p)}\in {\tau}" and p_1\Vdash_{\mathbb{P}}``\check{\alpha}\in {\tau}". Since p_0\in G, p_1\in \Gamma[G], ergo, \kappa_{\ell(p)}\in \tau_{\pi[G]}=\{\kappa_{2n+1}\mid n<\omega\} and \alpha\in \tau_{\pi[\Gamma[G]]}=\tau_{\pi[G]}. Therefore,

but this is impossible because \alpha\in (\kappa_{\ell(p)-1},\kappa_{\ell(p)+1}) and \alpha\neq \kappa_{\ell(p)}.

It suffices to prove that {\rm cf}^{\operatorname{HOD}}(\kappa)\leq {\rm cf}(\kappa). So we suppose that this is not the case and let us produce a contradiction from this assumption.

Since \kappa\cap \operatorname{cof}(\omega) is a stationary consisting of \operatorname{HOD}-singular cardinals, [GolPov, Theorem 3.5] ensures that {\rm cf}^{\operatorname{HOD}}(\kappa)<\kappa. Set \delta :={\rm cf}^{\operatorname{HOD}}(\kappa).

To complete the proof it would suffice to check that {\rm cf}(\delta)={\rm cf}(\kappa) in that

Let f\colon \delta\to \kappa be a cofinal, increasing, continious function in \operatorname{HOD}.

The existence of this function outright ensures that {\rm cf}(\kappa)\leq {\rm cf}(\delta). For the converse inequality, let c\subseteq \kappa be a club set (in V) with \mathrm{otp}(c)={\rm cf}(\kappa). Define h\colon \kappa\to \delta by h(\eta):=\min\{\alpha <\delta \mid f(\alpha)\geq \eta\}. Clearly, h[c] is a cofinal subset of \delta with cofinality {\rm cf}(\kappa), ergo {\rm cf}(\delta)\leq {\rm cf}(\kappa).

Suppose towards a contradiction that \Delta(\operatorname{HOD}, V)=\kappa. Our proof will combine Theorem 1.4 with ideas from a theorem of Shelah [ShelahHOD]. Note that the assumptions of Theorem 1.4 are met, hence

Working in \operatorname{HOD}, let \vec{\delta}=\langle \delta_i\mid i<{\rm cf}(\kappa)\rangle and \vec{t}=\langle t_i\mid i<\theta \rangle be, respectively, a cofinal, increasing, continuous sequence in \kappa and an injective enumeration of H_\kappa (i.e., of H_\kappa^{\operatorname{HOD}}). From this point onward, we argue similarly to a theorem of Shelah saying that there is an \mathrm{OD}_x well-ordering of \mathcal{P}(\kappa) for some x\subseteq \kappa [ShelahHOD]. As we will see, any set x\subseteq \kappa such that \vec{\delta},\vec{t}\in \operatorname{HOD}_{\{x\}} works to grant this inclusion. In our particular case, both \vec{\delta},\vec{t} are already in \operatorname{HOD} so x=\emptyset will work, and as a result \mathcal{P}(\kappa)\subseteq \operatorname{HOD}.

For each X\subseteq \kappa we define f_X\colon {\rm cf}(\kappa)\to \theta by

Define the following relation on \mathcal{P}(\kappa):

Note that if X\neq Y then f_X(i)\neq f_Y(i) for a tail of i<{\rm cf}(\kappa). Also, notice that \lhd is well-founded: Suppose that \langle X_n\mid n<\omega\rangle were to be a \lhd-decreasing sequence of subsets of \kappa. Since {\rm cf}(\kappa)\geq \omega_1 there is i<{\rm cf}(\kappa) such that f_{X_{n+1}}(i)<f_{X_n}(i) for all n<\omega, which would contradict the well-foundedness of the ordinals.

For each \rho\in \mathrm{Ord} let R_\rho:= \{X\subseteq \kappa\mid \mathrm{rank}_{\lhd}(X)=\rho\}.

For each ordinal \rho, let A_\rho :=\bigcup\{\mathrm{ran}(f_X)\mid X\in R_\rho\}. By the above observation |A_\rho|<\kappa. Let g_\rho\colon A_\rho \to \mathrm{otp}(A_\rho) be the unique order-preserving bijection and, for each X\in R_\rho, g_\rho \circ f_X\colon {\rm cf}^{\operatorname{HOD}}(\kappa) \to \mathrm{otp}(A_\rho). Since this function belongs to H_\kappa it also belongs to H_\kappa^{\operatorname{HOD}}. In particular, there is a unique index i_X<\theta in the enumeration of \vec{t} such that g_\rho\circ f_X= t_{i_X}.

Define a well-order \prec_\rho over each R_\rho by saying that X\prec_\rho Y if and only if i_X<i_Y, and a well-order \prec of \mathcal{P}(\kappa) by stipulating that X\prec Y if and only if \text{ran}_{\lhd}(X)<\mathrm{rank}_{\lhd}(Y) or if \text{ran}_{\lhd}(X)=\rho=\mathrm{rank}_{\lhd}(Y) then X\prec_\rho Y.

The well-order \prec is first-order definable from \vec{t} and \vec{\delta}. Indeed, from \vec{t} and \vec{\delta} one defines the functions f_X, the relation \lhd, the rank levels R_\rho, the sets A_\rho, the order-isomorphisms g_\rho, and finally \prec. Since in our case, \vec{t},\vec{\delta} are in \operatorname{HOD} we conclude that \mathcal{P}(\kappa)\subseteq \operatorname{HOD}: Let X\subseteq \kappa. Let \rho be the order type of the set of \prec-predecessors of X. Then X is the unique subset of \kappa which is the \rho-th element of the \operatorname{OD} well-order \prec. Since X is a set of ordinals, we conclude that X\in \operatorname{HOD}.

Suppose that \kappa is a measurable cardinal and V=g\operatorname{HOD}. Let \mathbb{P} be Prikry forcing relative to a normal measure on \kappa. Since \mathbb{P} is weakly homogeneous, V=\operatorname{HOD}^{V[G]} for any G\subseteq \mathbb{P} generic. However, in V[G], the Prikry sequence converging to \kappa is a countable set in V[G] that cannot be covered by subsets of \kappa in \operatorname{HOD}^{V[G]} of size \aleph_0. Since V[G]_\kappa = V_\kappa it follows that \kappa is the first cardinal where the failure of covering takes place.

Let X\subseteq \kappa^+ be a set with |X|\leq \mu. We construct Y\in M such that X\subseteq Y\subseteq \kappa^+ and |Y|\leq \mu. Through the proof, unless otherwise specified, by cardinality we shall mean cardinality in the sense of the universe V.

Set \sigma:=\sup(X). Since |X|\leq \mu it follows that \sigma < \kappa^+. By cardinal-correctness (Clause (?)) this is the same as saying that \sigma<\kappa^{+M}. Fix a bijection e\colon \kappa' \rightarrow \sigma in M for some \kappa'\leq \kappa. Then, e^{-1}[X]\subseteq \kappa and is a set of cardinality \leq\mu. If we were able to prove the existence of some X^*\in M with e^{-1}[X]\subseteq X^*\subseteq \kappa and |X^*|\leq \mu then we would have that Y:=e[X^*] is the desired covering set from M. So, everything reduces to the problem of covering V-subsets of \kappa of cardinality {\leq}\mu by means of M-subsets of \kappa of cardinality {\leq \mu}. In a slight abuse of notation for the rest of the proof we shall keep denoting this subsets of \kappa of cardinality {\leq}\mu by X.

Let X\subseteq \kappa be a set in V with size {\leq}\mu. If X was bounded in \kappa then, by covering below \kappa (Clause (?)) there would be a covering set in M of size {\leq}\mu.

So, let us assume that X is unbounded in \kappa. Since \kappa has uncountable cofinality in M, we can use the proof of Shelah’s representation theorem (see [AbrahamMagidor, Theorem 2.23] and the comments in [AbrahamMagidor, Theorem 2.33]) inside M to manufacture a scale \langle f_\xi \mid \xi <\kappa^+\rangle in a product (\prod_{i<\mu}\kappa_i^+)^M where \langle \kappa_i\mid i<\mu\rangle is a club on \kappa consisting of M-singular cardinals with \mu<\kappa_0. Moreover, this M-scale can be defined so that for every limit ordinal \xi < \kappa^+ there is a club E_\xi \subseteq\xi with \mathrm{otp}(E_\xi)={\rm cf}(\xi) and

Here, =^* denotes equality modulo bounded. While all of this is done inside M, Magidor argues (see [AbrahamMagidor, Theorem 2.23]) that by shrinking this M-scale to a club subset of \kappa, one obtains a V-scale. More formally, utilizing [AbrahamMagidor, Lemma 2.19] we deduce that (\star)_\theta holds(For the definition of (\star)_\theta see [AbrahamMagidor, Definition 2.8].) for all regular cardinal \theta < \kappa. Hence, \langle f_\xi\mid \xi < \kappa^+\rangle has an exact upper bound h for which the set \{i<\mu\mid {\rm cf}(h(i))<\theta\} is bounded in \mu, for all \theta < \kappa. From this one can find a club \langle \kappa_{i_j}\mid j<\mu\rangle \subseteq \langle \kappa_i\mid i<\mu\rangle for which h(i_j)=\kappa_{i_j}^+ (i.e., \kappa_{i_j}^{+M}, by cardinal correctedness) for all j<\mu (see [AbrahamMagidor, Claim 2.24]). This implies that the restriction of the M-scale to this sub-club, \langle f_\xi \restriction \{i_j\mid j<\mu\}\mid \xi < \kappa^+\rangle, remains a V-scale (i.e., <^*-cofinal) in the product (\prod_{j<\mu} \kappa_{i_j}^+)^V. Crucially, the unrestricted members of this sequence (i.e., the f_\xi’s) all belong to M. We use this scale to cover X.

By Clause (?), stationary many members \kappa_{i_j} of the sequence satisfy

As the subsequent proof will make clear, we can assume that the above cardinal-arithmetic configuration holds for all j<\mu.

For each j<\mu, X\cap \kappa_{i_j} is a subset of \kappa_{i_j} of size \leq \mu. Thus, Clause (?) applies, yielding a cover X_{j}\in M, X\cap \kappa_{i_j}\subseteq X_j\subseteq \kappa_{i_j} with |X_j|\leq \mu.

Let \langle Z^j_\eta \mid \eta <\kappa_{i_j}^+\rangle be an enumeration of all M-subsets of \kappa_{i_j} of cardinality \leq \mu. For this we use Clause (?) and (?) and our choice of the \kappa_{i_j}’s.

Let F\in (\prod_{j<\mu}\kappa_{i_j}^+)^V be the map given by the formula

Since \langle f_\xi\restriction\{i_j\mid j<\mu\}\mid \xi<\kappa^+\rangle is a scale in V, there is an index \xi<\kappa^+ for which F<^* f_\xi\restriction \{i_j\mid j<\mu\}. To simplify the forthcoming argument, we shall assume that F(j)<f_\xi(i_j) for all j<\mu.

For each i<\mu let g_i\colon f_\xi(i)\to \kappa_{i} be an injection in M, which exists by cardinal-correctness below \kappa (i.e., by Clause (?)). Critically, since f_\xi \in M, the sequence of bijections \langle g_i\mid i<\mu\rangle belongs to M, as well.

Look at the set of ordinals

Note that this set may be external to the inner model M.

For each j\in \mu\cap {\rm cf}(\omega) there is some l<j such that g_{i_j}(F(j))<\kappa_{i_l} because \langle \kappa_{i_j}\mid j<\mu\rangle is a club on \kappa. Applying Fodor’s lemma in V we find a stationary set S\subseteq \mu \cap \operatorname{cof}(\omega) and an index j^*<\mu such that g_{i_j}(F(j))<\kappa_{i_{j^*}} for all j\in S. Thus \{g_{i_j}(F(j))\mid j\in S\}\subseteq \kappa_{i_{j^*}} and it has cardinality \leq \mu. While both S and \{g_{i_j}(F(j))\mid j\in S\} may only belong to the universe V we can still invoke the \mu^+-covering property at \kappa_{i_{j^*}} (i.e., Clause (?)) to find a covering set U\subseteq \kappa_{i_{j^*}} in M with |U|\leq \mu.

Now we are ready to cover X. Define Y:=\bigcup\{Z^j_\eta\mid j<\mu\,\wedge\, \eta\in g^{-1}_j(U)\}. Clearly, X\subseteq Y\subseteq \kappa. Also Y\in M, as both U and \langle g_j\mid j<\mu\rangle belong to M.

We are done with the proof of the lemma.

Suppose that \kappa is the first cardinal \lambda for which (\operatorname{HOD}, V) fails to have the \mu^+-cover property at \lambda. We intend to apply Proposition 4.3, which will give an immediate contradiction. For this, we need to verify that all the assumptions hold true. Clearly, (1)–(3) of Proposition 4.3 are guaranteed by our current assumptions, so it suffices to show that {\rm cf}^{\operatorname{HOD}}(\kappa)\leq \mu.

First, notice that {\rm cf}^{\operatorname{HOD}}(\kappa)<\kappa: Indeed, the \mu^+-covering holds at every cardinal less than \kappa, ergo E^\kappa_\omega \subseteq (E^\kappa_{\leq \mu})^{\operatorname{HOD}}, and thus there is a stationary of \operatorname{HOD}-singular cardinals. As a result, the compactness theorem [GolPov, Theorem 3.5] applies, yielding {\rm cf}^{\operatorname{HOD}}(\kappa)<\kappa.

Arguing as in Theorem 1.4, the above implies that {\rm cf}({\rm cf}^{\operatorname{HOD}}(\kappa))={\rm cf}(\kappa)=\mu, so (in V) there is a club c\subseteq {\rm cf}^{\operatorname{HOD}}(\kappa) with \mathrm{otp}(c)=\mu. Since {\rm cf}^{\operatorname{HOD}}(\kappa)<\kappa the \mu^+-covering property at {\rm cf}^{\operatorname{HOD}}(\kappa) yields a covering set d\in \operatorname{HOD} with |d|\leq \mu. This shows that {\rm cf}^{\operatorname{HOD}}(\kappa)\leq \mu.

Let \langle \kappa_\alpha\mid \alpha < {\rm cf}(\kappa)\rangle be a continuous, increasing, cofinal sequence in \kappa. For each \alpha <{\rm cf}(\kappa) let \langle Z^\alpha_\xi\mid \xi < 2^{\kappa_\alpha}\rangle be an injective enumeration of \mathcal{P}(\kappa_\alpha) in V. For each X\subseteq \kappa consider the map F_X\in \prod_{\alpha < {\rm cf}(\kappa)}2^{\kappa_\alpha} defined by F_X(\alpha):=``The unique \xi<2^{\kappa_\alpha} such that X\cap \kappa_\alpha = Z^\alpha_\xi”. Clearly, the map X\mapsto F_X is injective, ergo 2^\kappa\leq |\prod_{\alpha < {\rm cf}(\kappa)} 2^{\kappa_\alpha}|. We shall see that the latter is less than or equal to |(2^\kappa)^{M}|. To show this we use the {\rm cf}(\kappa)^+-cover.

For each X\subseteq\kappa, \mathrm{range}(F_X)\in \mathcal{P}_{{\rm cf}(\kappa)^+}(\kappa) ergo there is Y\in M with |Y|\leq {\rm cf}(\kappa) such that \mathrm{range}(F_X)\subseteq Y\subseteq \kappa. Thus, F_X\in \prod_{\alpha <{\rm cf}(\kappa)} Y and

With this, we obtain the following bounds

because 2^{{\rm cf}(\kappa)} < \kappa \leq |(2^\kappa)^{M}|.

Let us define \restriction_M as follows: Given a condition p=p_{\leftarrow}{}^\smallfrown \langle f, T\rangle in \mathbb{P},

where T\restriction M is defined as \{\langle \nu_0\restriction M,\dots, \nu_{n-1}\restriction M\rangle\mid \langle \nu_0,\dots, \nu_{n-1}\rangle\in T\}.

The first observation is that p\restriction_M is a condition in \mathbb{P}_M. This is a routine checking using the fact that V_\kappa\subseteq M and that {}^\kappa M\subseteq M.

Let us show that \restriction_M is a residue map. Let r\leq_{\mathbb{P}_M} p\restriction_M, r\in M, and let us show that r is \leq_{\mathbb{P}}-compatible with p. First, by elementarity, r\leq_{\mathbb{P}_M} p\restriction_M is the same as r\leq_{\mathbb{P}} p\restriction_M. By definition, there is \vec\nu\in T\restriction M such that

To simplify notations let us assume that \vec\nu =\langle \nu\rangle.

Then, r_\leftarrow decomposes as r_{\leftarrow\leftarrow}{}^\smallfrown r_{\leftarrow\rightarrow} and similarly (p\restriction_M{}^\curvearrowright \langle \nu\rangle )_{\leftarrow} decomposes as (p\restriction_M{}^\curvearrowright \langle \nu\rangle)_{\leftarrow\leftarrow}{}^\smallfrown (p\restriction_M{}^\curvearrowright \langle \nu\rangle)_{\leftarrow\rightarrow}. Notice that (p\restriction_M{}^\curvearrowright \langle \nu\rangle)_{\leftarrow\leftarrow} is just p_{\leftarrow} due to the definition of the map \restriction_M. Set p'_0 to be r_{\leftarrow\leftarrow}.

Now, let us analyze (p\restriction_M{}^\curvearrowright \langle \nu\rangle)_{\leftarrow\rightarrow}. By definition of {}^\curvearrowright \langle \nu\rangle, this is

more explicitly:

• (f\restriction M)\downarrow \langle \nu\rangle is the function with domain \{\nu(\bar \alpha)\mid o(\bar\alpha)\neq 0\} and

  \left((f\restriction M)\downarrow \langle \nu\rangle\right)(\nu(\bar\alpha)):=(f\restriction M)(\bar \alpha)\restriction k_\nu=f(\bar \alpha)\restriction k_\nu,

  where k_\nu:=\min\{l<|f(\bar\alpha)|\mid\forall l\leq i<|f(\bar\alpha)|\, (o(f_i(\bar\alpha))<o(\nu(\bar\alpha))) \}.

• (T\restriction M)\downarrow \langle \nu\rangle is the projected tree

  \{\langle \mu_0\downarrow \nu, \dots, \mu_{n-1}\downarrow\nu\rangle\mid \vec\mu\in (T\restriction M)\},

  which is the same as \{\langle \mu_0\downarrow \nu, \dots, \mu_{n-1}\downarrow\nu\rangle\mid \vec\mu\in T\}.

Let \eta\in T be a \operatorname{dom}(f)-object such that \eta\restriction M=\nu.

We define p'_1 to be \langle f^{r_{\leftarrow\rightarrow}}\cup (f\downarrow \eta), S_1\rangle where S_1 is the intersection of the pullbacks to \operatorname{dom}(f^{r_{\leftarrow\rightarrow}})\cup \operatorname{dom}(f\downarrow\eta) of the trees T^{r_{\leftarrow\rightarrow}} and (T\downarrow\eta).

Finally, let p'_2 to be \langle f^{r_{\rightarrow}}\cup (f{}^\curvearrowright \eta), S_2\rangle where S_2 is defined analogously.

Finally, we set p'=p'_0{}^\smallfrown p'_1{}^\smallfrown p'_2. It follows that this is a condition in \mathbb{P} and that p' witnesses compatibility of r and p.

\pi_{\vec e}\restriction \mathbb{P}_M is clearly order-preserving. Let q\in \mathbb{P}_M be with q\leq p\restriction_M and r\leq_{\mathbb{P}_{\vec e}} \pi_{\vec{e}}(q). Say q=\langle q_i\mid i<\ell(q)\rangle. Since \restriction_M is a residue map, there is p'\in \mathbb{P} such that p'\leq p, q and p'\restriction M=q. In particular, p'_{\leftarrow}=q_{\leftarrow}. Since \vec{e} for some i_0<\ell(p'), \langle p'_i\mid i<i_0\rangle\in \mathbb{P}_{\vec e} there is i^*<\ell(q) such that \langle q_i\mid i<i^*\rangle \in \mathbb{P}_{\vec{e}}. Hence, \pi_{\vec{e}}(q)=\langle q_i\mid i<i^*\rangle. Then, r\leq_{\mathbb{P}_{\vec{e}}} \langle q_i\mid i<i^*\rangle. Letting q'=r{}^\smallfrown \langle q_i\mid i\in [i^*,\ell(q))\rangle we obtain a condition in \mathbb{P}_M, stronger than q such that \pi_{\vec{e}}(q')=r.

This follows from the usual Radin-like analysis establishing that every bounded subset of \kappa in V[G] is a member of V[G_{\vec e}]. Indeed, combining this with (*) above we obtain the desired equality.

Let \pi\colon M\to \bar{M} be the Mostowski collapse map. By elementarity,

Thus,

Observe that \operatorname{crit}(\sigma)=\chi and \sigma(\chi)=\kappa^{++} where \sigma:=\pi^{-1}. This yields \pi(E_\xi)=\{\pi(E)_{\xi,a}\mid a\in [\chi]^{<\omega}\} which equals \{\pi(E_{\xi,a})\mid a\in [\chi]^{<\omega}\}=\{E_{\xi,a}\mid a\in [\chi]^{<\omega}\}=E_\xi\restriction \chi.(To justify \pi(E_{\xi,a})=E_{\xi,a} we have used that \pi(X)=X for every X\in [\kappa]^n and n<\omega.) As a result we have that \pi(\vec{E})=\langle E_\xi\restriction \chi\mid \xi<\omega_1\rangle. Ergo, \bar{M} thinks that ``\pi(\mathbb{P}(\vec{E})) is the Extender Based Radin relative to \vec{F}”. But this definition is absolute between \bar{M} and the universe because \bar{M} is closed under \kappa-sequences of its elements (as so is M and \operatorname{crit}(\sigma)>\kappa). In other words, \pi(\mathbb{P}(\vec{E}))=\mathbb{P}(\vec{F}). But then \pi\colon \mathbb{P}(\vec{E})\cap M \to \mathbb{P}(\vec{F}) establishes the desired isomorphisms: both with respect to \leq_{\mathbb{P}(\vec{E})} and \leq^*_{\mathbb{P}(\vec{E})}.

To justify that \mathbb{P}(\vec{F}) forces 2^\kappa=\kappa^+ we will show that every \mathbb{P}(\vec{F})-name for a subset of \kappa can be ’reduced’ to a name for a subposet of \mathbb{P}(\vec{F}) of size \kappa. Since there are |\mathbb{P}(\vec{F})|^\kappa=\kappa^+ many of such subposets \mathbb{Q} and at most |\mathbb{Q}|^{\kappa\cdot\kappa}=\kappa^+ many \mathbb{Q}-nice names for subsets of \kappa we will conclude that

For the scope of this lemma write \mathbb{P}:=\mathbb{P}(\vec{F}). Let \dot{a} be a \mathbb{P}-name and p\in \mathbb{P} forcing ``\dot{a}\subseteq \kappa". Let N\prec H_\Theta be with |N|=\kappa, {}^{<\kappa} N\subseteq N, N\cap \kappa^+\in \kappa^+ and \{p,\mathbb{P},\dot{a}\}\cup V_\kappa\subseteq N. By properness of \mathbb{P}, there is q\leq^* p that is (\mathbb{P},N)-generic (see [Merimovich, Lemma 4.13]). For each \alpha < \kappa, denote

Clearly, this is dense open and it belongs to N, and thus

Define the (\mathbb{P}\cap N)-name \dot{b}=\{(\check{\alpha},r)\mid \alpha <\kappa,\,r\in \mathbb{P}\cap N,\, r\Vdash_{\mathbb{P}}\check{\alpha}\in \dot{a}\}.

A routine checking shows that q\Vdash_{\mathbb{P}}\dot{b}=\dot{a}.

We have thereby showed that for each \mathbb{P}-name \dot{a} for a subset of \kappa,

is dense. Thus, given any V-generic G\subseteq \mathbb{P},

But as argued above there are at most \kappa^+ such nice names, ergo right-hand-side union has size \kappa^+.

Working in V, let p\in G be a condition and \dot{f} be a \mathbb{P}-name such that p\Vdash_{\mathbb{P}}\dot{f}\colon \omega\rightarrow \mathrm{Ord}. We will show that \dot{f}_G\in V[G_M].

An elementary submodel \mathcal{M}\prec H_\Theta is adequate if |\mathcal{M}|=\kappa, \mathcal{M}^{<\kappa}\subseteq \mathcal{M}, \mathcal{M}\cap \kappa^+\in \kappa^+ and p, \dot{f}, \mathbb{P}\in \mathcal{M}. By work of Merimovich, \mathbb{P} is a \kappa-proper forcing notion [Merimovich, Lemma 4.13]. In particular, the set

is dense below p.

Let p^*\in G belonging to the above set and \mathcal{M} an adequate elementary submodel witnessing it. For each n<\omega, consider the dense open set of conditions deciding \dot{f}(n); more explicitly, we are interested on the set

Clearly, D_n\in \mathcal{M} for all n<\omega. So, by \langle \mathcal{M}, \mathbb{P}\rangle-genericity, for each n<\omega,

Since p^*\in G, in fact D_n\cap G\cap \mathcal{M} is non-empty for all n<\omega.

For each n<\omega, let q\in G\cap \mathcal{M} deciding the value of \dot{f}(n). By elementarity of \mathcal{M}, there is \alpha\in \mathcal{M} such that q\Vdash_{\mathbb{P}}\dot{f}(n)=\alpha. All in all, this shows that

Working in V, let \Phi\colon \mathcal{M}\leftrightarrow \kappa be a bijection and consider f^* the composition \Phi\circ \dot{f}_G\colon \omega\rightarrow \kappa. We claim that f^*\in V[G_M], which yields \dot{f}_G\in V[G_M].

Indeed, since {\rm cf}^{V[G]}(\kappa)=\omega_1 the function f^* is bounded in \kappa; say that \lambda<\kappa is such a bound. Let q\leq p^* be a condition in G forcing \dot{f}^*\colon \check{\omega}\to \check{\lambda}.

By the usual factoring lemma of Radin-like forcings we have that \mathbb{P}/q is isomorphic to a product (\mathbb{P}_{\vec{e}}/q_0)\times (\mathbb{P}/q_1) where the latter has a \leq^*-ordering that is more than (\lambda^+)^V-closed. Combining this with the Prikry property one can show that the set

is dense below q.

Denote by G_{\vec{e}} the \mathbb{P}_{\vec{e}}-generic filter obtained after projecting G via the standard projection between \mathbb{P} and \mathbb{P}_{\vec{e}}. Thus, f^*\in V[G_{\vec{e}}]\subseteq V[G_M].

For a condition p=\langle f, T\rangle \in \mathbb{P}^* define its projection to the normal measure by recursion as follows: For \langle f, T\rangle\in \mathbb{P}^* denote

In general, for a condition p=\langle p_i\mid i\leq n\rangle\in \mathbb{P}, define by recursion

As it turns, \pi_{\{\kappa\}} defines a projection between \mathbb{P} and its image, this latter being the usual Radin forcing defined with respect to the Mitchell increasing sequence of normal measures \langle E_\alpha(\{\kappa\})\mid \alpha <\omega_1\rangle on \operatorname{OB}(\{\kappa\}).

Note that the diagram

commutes and that \pi_{\{\kappa\}}\restriction \mathbb{P}_M is a projection.

In particular, we get the inclusions V[\pi_{\{\kappa\}}[G]]\subseteq V[G_M]\subseteq V[G].

The usual Radin-like analysis show that V[G] and V[\pi_{\{\kappa\}}[G]] compute cofinalities alike, ergo so do V[G_M] and V[G].

This follows from \mathcal{P}(\alpha)^{V[G_M]}=\mathcal{P}(\alpha)^{V[G]} for all \alpha <\kappa. For every \alpha <\kappa there is an extender sequence \vec{e}\in V_\kappa with \mathcal{P}^{V[G_{\vec{e}}]}(\alpha)=\mathcal{P}^{V[G]}(\alpha) but the former is contained in V[G_M] because of (?) on page ?.

This follows from Lemma 4.6 using that (2^{\kappa})^{V[G_M]}<(2^\kappa)^{V[G]}.