Graphs represented by Ext

For each i<\delta let A_i \subseteq \kappa be such that \Gamma_i = A_i/{\mathscr D}_\kappa. If \delta < \kappa, then \Gamma = A/ {\mathscr D}_\kappa is as required where A=\bigcup\limits_{i<\delta} A_i, and if \delta=\kappa, then \Gamma = A/{\mathscr D}_\kappa is as required where A=\bigtriangleup_{i<\kappa} A_i is the diagonal intersection of the sets A_i, i<\kappa.

Let \langle {\mathscr U}_\varepsilon:\varepsilon < 2^\lambda\rangle be a partition of 2^\lambda such that {\mathscr U}_\varepsilon \subseteq [\varepsilon,2^\lambda) has cardinality 2^\lambda. We claim that there are sequences \bar{\mathbb{K}}_\varepsilon, \bar{\mathbf H}_{\varepsilon, \beta} and \bar{\mathbf h}^*_{\varepsilon, \beta}, for \varepsilon \le 2^\lambda and \beta < \mu_2 with the following properties:

1. \oplus_{\varepsilon}

   1. (a) \bar{\mathbb{K}}_\varepsilon = \langle \mathbb{K}_{\varepsilon, \beta}:\beta < \mu_2\rangle is a sequence of abelian groups,

   2. (b) for each \beta < \mu_2 the sequence \langle \mathbb K_{\zeta,\beta}:\zeta \le 2^\lambda \rangle is \subseteq-increasing and continuous,

   3. (c) \mathbb{K}_{\varepsilon,\beta} has cardinality \le 2^\lambda,

   4. (d) \bar{\bf H}_{\varepsilon,\beta} = \langle (\alpha_{\beta,\zeta},h_{\beta,\zeta}):\zeta \in {\mathscr U}_\varepsilon \rangle lists all the pairs (\alpha, h) where \alpha < \mu_1,~\alpha R \beta and h \in {\rm Hom}(\mathbb{G}^1_\alpha, \mathbb{K}_{\varepsilon,\beta}),

   5. (e) \bar{\bf h}^*_{\varepsilon,\beta} = \langle h^*_{\beta,\zeta}:\zeta \in {\mathscr U}_\varepsilon \rangle,

   6. (f) if \varepsilon \le \zeta \in {\mathscr U}_\varepsilon, then h^*_{\beta, \zeta} \in {\rm Hom} (\mathbb{G}^2_{\alpha_{\beta,\zeta}},\mathbb{K}_{\zeta +1,\beta}) extends h_{\beta,\zeta}. The property is conveniently summarized by the subjoined diagram:

      \xymatrix{ && \mathbb{G}^1_{\alpha_{\beta,\zeta}} \ar[r]^{\subseteq}\ar[d]_{h_{\beta,\zeta}}&\mathbb{G}^2_{\alpha_{\beta,\zeta}}\ar[d]^{h^*_{\beta, \zeta}}\\ && \mathbb{K}_{\zeta ,\beta}\ar[r]^{\subseteq}&\mathbb{K}_{\zeta +1,\beta} &&&}

We proceed by a double induction on \zeta and \beta to construct such a sequence. For \zeta = 0 and for all \beta < \mu_2 set

By Fact 3.2(ii), \mathbb{K}_{0, \beta} is free, and so \lambda-free.

For \zeta a limit ordinal and \beta < \mu_2, we set

Now suppose that the groups K_{\varepsilon, \gamma} are defined for all \varepsilon < \zeta+1 and \gamma < \mu_2. We define \mathbb{K}_{\zeta+1, \beta} for \beta<\mu_2 as follows.

Let \bar{\bf H}_{\varepsilon,\beta} = \langle (\alpha_{\beta,\zeta},h_{\beta,\zeta}):\zeta \in {\mathscr U}_\varepsilon \rangle be as in clause (d).

We look at the following diagram

and set

In particular, there is a homomorphism f induced from h_{\beta,\zeta} which commutes the following diagram:

We set h^*_{\beta, \zeta}:=f. By a diagram chasing, h^*_{\beta, \zeta} extends h_{\beta,\zeta}, and recall that g:\mathbb{K}_{\zeta,\beta} \hookrightarrow\mathbb{K}_{\zeta+1,\beta} is an embedding. This completes the inductive construction and hence proves the existence of the sequences claimed to exist above, i.e., the construction of \oplus_{\varepsilon} is now complete. Suppose now that every group \mathbb{G}_\alpha is \lambda-free. By Fact 3.2(iii) it follows that \mathbb{K}_{\zeta,\beta} is \lambda-free as well.

For \beta < \mu_2 let

In view of Lemma 3.1, the above short exact sequence and the hypotheses, each \mathbb{K}_\beta is an \aleph_1-free abelian group of size 2^\lambda. Recall that \lambda-freeness of \mathbb{K}_\alpha follows by \lambda-freeness of \mathbb{G}_\alpha, as desired by \boxtimes^2_{\lambda,\mu_1}(\gamma).

Let us check the item \boxtimes^2_{\lambda,\mu_1}(\beta). First, suppose that \alpha R \beta and let h \in {\rm Hom} (\mathbb{G}^1_\alpha, \mathbb{K}_\beta). Since |\mathbb{G}^1_\alpha| \le \lambda < {\rm cf}(\mu_2), for some \varepsilon < 2^\lambda we have {\rm Rang}(h)\subseteq K_{\varepsilon,\beta}, and hence we can assume that h \in {\rm Hom} (\mathbb{G}^1_\alpha, \mathbb{K}_{\varepsilon,\beta}). By the definition of \bar{\bf H}_{\varepsilon,\beta}, for some \zeta \in {\mathscr U}_\varepsilon we have (\alpha, h) = (\alpha_{\beta,\zeta},h_{\beta,\zeta}). Then by clause (f) of the construction, h^*_{\beta,\zeta} \in {\rm Hom}(\mathbb{G}^2_\alpha,\mathbb{K}_{\zeta +1,\beta}) extends h_{\beta,\zeta}. There is a natural embedding map from {\rm Hom}(\mathbb{G}^2_\alpha,\mathbb{G}_{\zeta +1,\beta}) into {\rm Hom} (\mathbb{G}^2_\alpha, \mathbb{K}_\beta), so without loss of generality, we may and do assume that

Now, we look at the exact sequence

Applying {\rm Hom}(-,\mathbb{K}_\beta) to (+), it induces the following long exact sequence

where the last vanishing {\rm Ext}(\mathbb{G}^2_\alpha, \mathbb{K}_\beta) =0 holds as \mathbb{G}^2_\alpha is free. From this,

Thus, {\rm Ext}(\mathbb{G}_\alpha, \mathbb{K}_\beta)=0 if and only if {\bf f} is surjective. But as we observed above, {\bf f} is onto, and hence we conclude that {\rm Ext}(\mathbb{G}_\alpha, \mathbb{K}_\beta)=0.

Let (\alpha,\beta) \in (\mu_1 \times \mu_2) \setminus R. We need to show {\rm Ext}(\mathbb{G}_\alpha, \mathbb{K}_\beta) \ne 0.

Suppose on the contrary that {\rm Ext}(\mathbb{G}_\alpha, \mathbb{K}_\beta) = 0, and search for a contradiction. Since \mathbb{G}^1_\alpha is a direct summand of \mathbb{K}_{0,\beta}, we have \mathbb{G}^1_\alpha\oplus X= \mathbb{K}_{0,\beta}. Let \rho:\mathbb{G}^1_\alpha\to\mathbb{G}^1_\alpha\oplus X be the natural map, and look at

So by our assumption, there is g^2_\alpha: \mathbb{G}^2_\alpha \rightarrow \mathbb{K}_\beta extending g^1_\alpha:

We now show that there exists u \subseteq 2^\lambda equipped with:

1. (a) |u| = \lambda,

2. (b) 0 ,\alpha,\beta \in u,

3. (c) \varepsilon \in u \Rightarrow {\rm Rang}(h_{\beta, \varepsilon}) \subseteq \sum\limits_{\zeta \in u \cap \varepsilon} {\rm Rang}(h^*_{\beta, \zeta})+ \sum\limits_{\gamma \in u} \mathbb{G}^1_\gamma,

4. (d) {\rm Rang}(g^2_\alpha) \subseteq \sum\limits_{\varepsilon \in u}{\rm Rang}(h^*_{\beta, \varepsilon}) + \sum\limits_{\gamma \in u} \mathbb{G}^1_\gamma.

Indeed, by induction on i <\lambda we define an increasing and continuous sequence \langle u_i:i<\lambda\rangle of subsets of 2^{\lambda} such that for each i <\lambda, the following properties are valid:

1. \bullet |u_i| \leq \lambda,

2. \bullet 0 ,\alpha,\beta \in u_0,

3. \bullet If i is limit ordinal, then u_i=\bigcup_{j<i} u_j,

4. \bullet {\rm Rang}(g^2_\alpha) \subseteq \sum\limits_{\varepsilon \in u_0}{\rm Rang}(h^*_{\beta, \varepsilon}) + \sum\limits_{\gamma \in u_0} \mathbb{G}^1_\gamma,

5. \bullet If \varepsilon \in u _i then

   {\rm Rang}(h_{\beta, \varepsilon}) \subseteq \sum\limits_{\zeta\in u_{i+1}\cap \varepsilon} {\rm Rang}(h^*_{\beta, \zeta})+ \sum\limits_{\gamma \in u_{i+1}} \mathbb{G}^1_\gamma.

Set u:=\bigcup\limits_{i<\lambda}u_i. It is easily seen that u satisfies the required properties.

Let \varepsilon(*) = {\rm otp}(u), where {\rm otp}(u) denotes the order type of the set u and let \langle \gamma_\varepsilon:\varepsilon < \varepsilon(*)\rangle be the increasing enumeration of u.

Set

We now define a construction \bar{\mathbb{L}}:= \langle \mathbb{L}_\varepsilon:\varepsilon \le \varepsilon(*)\rangle by {\mathscr G} over \mathbb{L}_0=\mathbb{G}_\alpha^1. To achieve this, we set

Let A:={\rm coker}( h^*_{\beta, \zeta}). Let put this into the previous diagram and obtain the following:

This yields that:

where f_i are the natural inclusions. Now, we use the {\rm Ker}-{\rm coker} exact sequence:

Recall that {\rm Ker}(=)= {\rm coker}(=)=0, {\rm coker}(f_1)=\frac{{\rm Rang} (h^*_{\beta, \zeta} )}{{\rm Rang} (h_{\beta, \zeta} )} and {\rm coker}(f_2)= \mathbb{G}_{\alpha_{\beta,\epsilon}}. So,

For each \epsilon <\epsilon(\star) we have

In view of Definition 3.3, \bar{\mathbb{L}} is indeed a construction. Recall from (d) that {\rm Rang}(g^2_\alpha)\subset \mathbb{L}_{\varepsilon(*)}. Hence g^2_\alpha:\mathbb{G}^2_\alpha\to \mathbb{L}_{\varepsilon(*)} extends the identity function over \mathbb{G}^1_\alpha. This is in contradiction with \boxtimes^1_{\lambda,\mu_1}(c).

See [GT, Theorem 9.1.17].

Set \dot{\mathbb G}^1: = \mathbb G^1 \oplus \bigoplus\limits_{n<\omega} z_n \mathbb{Z}, where \bigoplus\limits_{n<\omega} z_n \mathbb{Z} is a free abelian group with a base given by \{z_n:n<\omega\}. To define \dot{\mathbb G}^2, let us first set

For any infinite sequence \vec{a}=\langle a_n: n<\omega \rangle\in \prod _\omega2= {}^{\omega}2, we look at the following

i.e., the subgroup of \widehat{\mathbb G^{2, \ast}} generated by \mathbb G^{2, \ast} and the distinguished element

It is routine to see that the properties (a)-(f) are satisfied with \mathbb G^{2, \vec{a}}. We are going to show that \dot{\mathbb G}^2=\mathbb G^{2, \vec{a}} satisfies in the remaining property (g) for a suitable choice of \vec{a}. Suppose on the way of contradiction that for each \vec{a}=\langle a_n: n<\omega \rangle \in {}^{\omega}2 there is a counterexample \bold{\dot{f}}_{\vec a} and \langle \dot{\mathbb{L}}^{\vec a}_\varepsilon: \varepsilon \leq \varepsilon(*) \rangle to clause (g). In particular, \bold{\dot{f}}_{\vec a} \in {\rm Hom}(\mathbb G^{2, \vec{a}}, \dot{\mathbb{L}}_{\varepsilon(*)}), and it extends \bold f \cup {\rm id}_{\dot{\mathbb G}^1}.

For \vec a \in {}^{\omega}2 and m<\omega set

1. Claim (A): Let \vec{a}, \vec{b} and m be such that

   1. \bullet m<\omega,

   2. \bullet a_\ell=b_\ell for all \ell\leq m,

   3. \bullet \bold{\dot{f}}_{\vec a} (y^{\vec a}_\ell)=\bold{\dot{f}}_{\vec b} (y^{\vec b}_\ell) for all \ell \leq n.

   Then \vec{a} = \vec{b}.

To prove the claim, we proceed by induction on \ell < \omega and show that:

1. (i) a_{\ell}=b_{\ell},

2. (ii) \bold{\dot{f}}_{\vec a} (y^{\vec a}_\ell)=\bold{\dot{f}}_{\vec b} (y^{\vec b}_\ell).

The case \ell \leq m follows from the assumption, so we may assume that \ell > m.

Now suppose that \ell =k+1 and the result is true for all natural numbers less or equal to k. Let us evaluate \bold{\dot{f}}_{\vec a} at y^{\vec a}_m from (+), and recall that \bold{\dot{f}}_{\vec a} extends \bold f \cup {\rm id}_{\dot{\mathbb G}^1}. It turns out that:

It immediately follows that for all i<\omega

We apply (\ast) at level i=k and combine it with inductive hypothesis to observe that

Since \dot{\mathbb{L}}_{\varepsilon(*)} is torsion-free, it follows that \bold{\dot{f}}_{\vec a} (y^{\vec a}_\ell)-\bold{\dot{f}}_{\vec b} (y^{\vec b}_\ell)=0, so (ii) is valid. In order to show (i) we apply (\ast) at level i=\ell:

i.e.,

It follows that (\ell+1)|(a_{\ell}-b_{\ell}). Since a_i,b_i\in\{0,1\}, we must have a_{\ell}-b_{\ell}=0. This completes the proof of the desired claim. Let \vec{a} and \vec{b} be distinct in {}^{\omega}2 such that a_0=b_0 and \bold{\dot{f}}_{\vec a} (y^{\vec a}_0)=\bold{\dot{f}}_{\vec b} (y^{\vec b}_0). This contradicts Claim(A), so we are done.

(1): Since S is non-reflecting, given any limit ordinal \delta < \lambda, there exists a club C_\delta \subseteq \delta such that S \cap C_\delta=\emptyset. Furthermore as \diamondsuit_S holds, and in the light of Fact 4.4, we can find a sequence \langle S_\varepsilon:\varepsilon < \lambda \rangle of subsets of S such that the following two properties are satisfied:

• the sets S_\varepsilon are pairwise almost disjoint, i.e., for all \varepsilon < \zeta < \lambda, S_\varepsilon \cap S_\zeta is a bounded subset of \lambda,

• \diamondsuit_{S_\varepsilon} holds for \varepsilon < \lambda.

For \varepsilon < \lambda let

be such that:

• f^\varepsilon_\alpha: \alpha \to \alpha is a function,

• \mathbb{H}^{1, \varepsilon}_\alpha \subseteq \mathbb{H}^{2, \varepsilon}_\alpha are abelian groups,

• the universe of \mathbb{H}^{2, \varepsilon}_\alpha is \alpha,

• \bar{\mathbb{L}}_\alpha=\langle \mathbb{L}^\alpha_{\varepsilon}: \varepsilon \leq \alpha \rangle is a construction such that \mathbb{L}_{\alpha} has universe \alpha,

• each \mathbb{L}_\varepsilon and \mathbb{L}_{\varepsilon+1}/ \mathbb{L}_\varepsilon is \aleph_1-free,

• for any tuple (f, \mathbb{H}^1, \mathbb{H}^2, \bar{\mathbb{L}}), where f: \lambda \to \lambda is a function, \mathbb{H}^1 \subseteq \mathbb{H}^2 are abelian groups where the universe of \mathbb{H}^2 is \lambda and \bar{\mathbb{L}}=\langle \mathbb{L}_{\varepsilon}: \varepsilon \leq \lambda \rangle is a construction such that the universe of \mathbb{L}_\lambda is \lambda and \mathbb{L}_\varepsilon and \mathbb{L}_{\varepsilon+1}/ \mathbb{L}_\varepsilon are \aleph_1-free, then the following set

  \{ \alpha \in S_\varepsilon: (f\restriction \alpha, \mathbb{H}^1 \restriction \alpha, \mathbb{H}^2 \restriction \alpha, \bar{\mathbb{L}} \restriction \alpha)= (f^\varepsilon_\alpha, \mathbb{H}^{1, \varepsilon}_\alpha, \mathbb{H}^{2, \varepsilon}_\alpha, \bar{\mathbb{L}}_\alpha) \}

  is stationary.

Given \varepsilon< \lambda by induction on \alpha < \lambda, we choose the sequences \bar{\mathbb{G}}^\iota_\varepsilon = \langle \mathbb{G}^\iota_{\alpha,\varepsilon}:\alpha < \lambda\rangle, for \iota=1, 2 such that:

1. \circledast

   1. (a) \bar G^\iota_\varepsilon is an increasing and continuous sequence of abelian groups \mathbb{G}^\iota_{\alpha,\varepsilon}=(G^\iota_{\alpha, \varepsilon}, +, 0).

   2. (b) \mathbb{G}^\iota_{0,\varepsilon} = \{0\} and for \alpha < \lambda, the universe of \mathbb{G}^2_{\alpha,\varepsilon}, namely G^2_{\alpha, \varepsilon} is an ordinal \gamma_{\alpha,\varepsilon} < \lambda.

   3. (c) The following holds:

      1. (c1): \mathbb{G}^1_{\alpha,\varepsilon} \subseteq \mathbb{G}^2_{\alpha,\varepsilon} are free.

      2. (c2): If \alpha < \beta then \mathbb{G}^2_{\alpha, \varepsilon} \cap \mathbb{G}^1_{\beta, \varepsilon} = \mathbb{G}^1_{\alpha, \varepsilon}.

      3. (c3): If \alpha < \beta and \alpha \notin S_\varepsilon, then \mathbb{G}^2_\beta/(\mathbb{G}^1_\beta + \mathbb{G}^2_\alpha) is free.

   4. (d) If \alpha < \beta and \alpha \notin S_\varepsilon, then \mathbb{G}^1_{\beta,\varepsilon}/\mathbb{G}^1_{\alpha,\varepsilon} and \mathbb{G}^2_{\beta,\varepsilon}/\mathbb{G}^2_{\alpha,\varepsilon} are free.

   5. (e) if \delta \in S_\varepsilon, then there are no \dot{\bold f} and \langle \dot{\mathbb{L}}_\zeta: \zeta \leq \delta \rangle equipped with the following properties:

      1. (e1) \langle \dot{\mathbb{L}}_\zeta: \zeta \leq \delta \rangle is a construction over \mathbb{G}_{\delta+1, \varepsilon}^1,

      2. (e2) \dot{\bold f} \in {\rm Hom}(\mathbb{G}^2_{\delta+1, \varepsilon}, \dot{\mathbb{L}}_{\delta}),

      3. (e3) \dot{\bold f} extends \bold{f}^\varepsilon_\delta \cup {\rm id}_{\mathbb{G}_{\delta+1, \varepsilon}^1}. This means that \mathbb{G}^2_{\delta, \varepsilon}\subseteq \mathbb{H}^{2, \varepsilon}_{\delta}, and also:

         \xymatrix@=3pc{ & & \dot{\mathbb{L}}_{\delta} \\ \mathbb{G}^1_{\delta+1, \varepsilon} \ar[r]_\subseteq\ar@/^/[rru]^{\subseteq} & \mathbb{G}^2_{\delta+1, \varepsilon} \ar[ru]_{\nexists\dot{\bold f}} & \\ \mathbb{G}^1_{\delta, \varepsilon} \ar [u]^\subseteq \ar[r]_\subseteq & \mathbb{G}^2_{\delta, \varepsilon}\ar[u] ^\subseteq \ar@/_/[ruu]_{\bold{f}^\varepsilon_\delta} & }

      4. (e4) \mathbb{L}^\delta_\zeta \subseteq \dot{\mathbb{L}}_\zeta for \zeta \leq \delta,

      5. (e5) \mathbb{L}^\delta_\zeta = \mathbb{L}^\delta_{\zeta+1} \cap \dot{\mathbb{L}}_\zeta,

      6. (e6) \dot{\mathbb L}_{\zeta+1}/ (\mathbb{L}^\delta_{\zeta+1}+\dot{\mathbb L}_\zeta) is free.

For \alpha=0, set \mathbb{G}^1_{0,\varepsilon} = \mathbb{G}^2_{0,\varepsilon} = \{0\}. For limit ordinal \delta, set \mathbb{G}^\iota_{\delta,\varepsilon} = \bigcup\limits_{\alpha < \delta}\mathbb{G}^\iota_{\alpha,\varepsilon}. Let us show that items (c) and (d) continue to hold. By the induction hypothesis and for all \alpha < \beta < \delta with \alpha \notin S, \mathbb{G}^\iota_{\beta, \varepsilon} / \mathbb{G}^\iota_{\alpha, \varepsilon} and \mathbb{G}^2_\beta/(\mathbb{G}^1_\beta + \mathbb{G}^2_\alpha) are free, and since we have a club C_\delta of \delta which is disjoint to S, it immediately follows that the groups \mathbb{G}^\iota_{\delta, \varepsilon}, \mathbb{G}^\iota_{\delta, \varepsilon}/ \mathbb{G}^\iota_{\alpha, \varepsilon} and \mathbb{G}^2_{\delta, \varepsilon}/(\mathbb{G}^1_{\delta, \varepsilon} + \mathbb{G}^2_{\alpha, \varepsilon}) are free for all \alpha < \delta with \alpha \notin S.

Let \iota=1, 2. Now suppose that \delta < \lambda and we have defined the groups \mathbb{G}^\iota_{\alpha, \varepsilon} and ordinals \alpha \leq \delta. We would like to define the groups \mathbb G^1_{\delta+1, \varepsilon} and \mathbb G^2_{\delta+1, \varepsilon} so that the (a)-(e) continue to hold.

In the case \delta \notin S_\varepsilon, we set

and

It is not difficult to show that items (a)-(d) continue to hold and there is nothing to prove for case (e).

Now suppose that \delta \in S_\varepsilon. We define the groups \mathbb{G}^1_{\delta+1, \varepsilon} and \mathbb{G}^2_{\delta+1, \varepsilon} such that items (a)-(e) above continue to hold, and further we have:

1. (f) \mathbb{G}^2_{\delta+1,\varepsilon}/(\mathbb{G}^1_{\delta,\varepsilon} + \mathbb{G}^2_{\delta,\varepsilon}) is not free,

2. (g) if \gamma \in \delta \backslash S_\varepsilon, then \mathbb{G}^2_{\delta+1,\varepsilon}/(\mathbb{G}^1_{\delta,\varepsilon} + \mathbb{G}^2_{\gamma,\varepsilon}) is free.

As \delta \in S_\varepsilon, we have {\rm cf}(\delta) = \aleph_0, so let \langle \gamma_{\delta,n}:n < \omega\rangle be an increasing sequence of successor ordinals < \delta with limit \delta.

For any n<\omega, \gamma_{\delta, n} \notin S_\varepsilon, it is easily seen that the abelian groups \mathbb G^1_{\gamma_{\delta, n}, \varepsilon} \subseteq \mathbb{G}^2_{\gamma_{\delta, n}, \varepsilon} satisfy the hypotheses of Lemma 4.5, hence by the lemma, we can find the groups \mathbb G^1_{\delta+1, \varepsilon} \subseteq \mathbb G^2_{\delta+1, \varepsilon} such that there for all constructions \mathbb L \supseteq \mathbb G^1_{\delta+1, \varepsilon} as described in (e), there is no \bold f \in {\rm Hom}(\mathbb{G}^2_{\delta+1, \varepsilon}, \mathbb L) such that \bold f \supseteq \bold f^\varepsilon_\delta \cup {\rm id}_{\mathbb G^1_{\delta+1, \varepsilon}}.

This finishes our inductive construction. For \iota=1, 2 and \varepsilon < \lambda, we define:

Set also

• (h1): \mathbb{G}_{\alpha, \varepsilon}=\mathbb{G}^{2}_{\alpha, \varepsilon}/ \mathbb{G}^{1}_{\alpha, \varepsilon},

• (h2): \mathbb{G}_{\varepsilon}=\bigcup\limits_{\alpha < \lambda}\mathbb{G}_{\alpha, \varepsilon},

• (h3): \bar{\mathbb{G}}=\langle \mathbb{G}_{\varepsilon}: \varepsilon < \lambda \rangle.

Let us show that the properties.

(i): This is clear.

(ii): We apply the properties taken from items (c) and (d) of the construction, along with freeness \mathbb{G}^1_\varepsilon \subseteq \mathbb{G}^2_\varepsilon to deduce \mathbb{G}_{\varepsilon} is strongly \lambda-free as witnessed by the sequence

for more details, see [EM02, IV.1.11].

(iii): Suppose by contradiction that there are \varepsilon < \lambda, a construction \bar{\mathbb{L}}=\langle \mathbb{L}_\varepsilon: \varepsilon \leq \lambda \rangle by \{\mathbb{G}_\alpha:\alpha \in \lambda \setminus \varepsilon\} over \mathbb{G}^1_\varepsilon and there is a homomorphism {\bf g} from \mathbb{G}^2_\varepsilon into \mathbb{L}_\lambda which extends {\rm id}_{\mathbb{G}^1_\varepsilon}:

Without loss of generality we can assume that \mathbb{L}_\lambda has size \lambda and that its universe is \lambda. The set

is a club, thus we can find some \delta \in E \cap S_\varepsilon such that:

• (i1): {\bf g} \restriction \delta=f^\varepsilon_\delta,

• (i2): \mathbb{L}_\kappa \restriction \delta=\mathbb{H}^{2, \varepsilon}_\delta,

• (i3): \mathbb{H}^{1, \varepsilon}_\delta=\mathbb{G}^1_{\delta, \varepsilon},

• (i4): \mathbb{G}^2_\varepsilon \cap \delta=\mathbb{G}^2_{\delta, \varepsilon}.

Now note that f={\bf g}\restriction \mathbb{G}^2_{\delta+1, \varepsilon}: \mathbb{G}^2_{\delta+1, \varepsilon} \to \mathbb{L}_\lambda is such that {f}^\varepsilon_\delta\cup {\rm id}_{\mathbb{G}_{\delta+1, \varepsilon}^1} \subseteq{f}, which is in contradiction with clause (e) of the construction.

(2): The proof is similar to the proof of (1), this time, we find the following family

of almost disjoint subsets of \lambda such that \Diamond_{S_\varepsilon} holds for all \varepsilon. By [devlin], such a sequence exists.

This follows from Theorem 3.5 and Theorem 4.7.

See [EM02, Page 444].

(1): By a result of Solovay (see [EM02, Corollary II.4.9]) we can find a partition \langle S_\varepsilon: \varepsilon < \lambda \rangle of \lambda into \lambda many disjoint stationary sets. Let

and recall that its \mathbb Z-adic completion is denoted by \widehat{\mathbb C_\varepsilon}. Clearly, \widehat{\mathbb C_\varepsilon} has cardinality \lambda, so we identify it with \lambda. Let also

Set \chi= \lambda^{+3}. The initial structure for the Black box, corresponding to the stationary set S_\varepsilon is as follows:

where \widehat{C_\varepsilon} denotes the 3-ary relation on \lambda which is the graph of the addition operation on the group \widehat{\mathbb C_\varepsilon}. We take the first bijection \bold g_\varepsilon: \lambda \times \lambda \to \lambda with respect to \lhd, and use it to identify each X \subseteq \lambda with a subset of \widehat{C_\varepsilon} \times \widehat{C_\varepsilon}. Let \{(M^\varepsilon_i, X^\varepsilon_i): i \in I_\varepsilon \} be as in the statement of Theorem 5.2 when m=1 and E_0=S_\varepsilon, and note that for each i\in I_\varepsilon and \varepsilon<\lambda, \bold g_\varepsilon \in M^\varepsilon_i.

Let \iota=1, 2. We proceed as in the previous section and for a given \varepsilon< \lambda, by induction on \alpha < \lambda, we choose the sequences \bar{\mathbb{G}}^\iota_\varepsilon = \langle \mathbb{G}^\iota_{\alpha,\varepsilon}:\alpha < \lambda\rangle equipped with the following five items:

1. (a) \bar{\mathbb{G}}^\iota_\varepsilon is an increasing and continuous sequence of abelian groups \mathbb{G}^\iota_{\alpha,\varepsilon}=(G^\iota_{\alpha, \varepsilon}, +, 0).

2. (b) \mathbb{G}^\iota_{0,\varepsilon} = \{0\} and for \alpha < \lambda, the universe of \mathbb{G}^2_{\alpha,\varepsilon}, namely G^2_{\alpha, \varepsilon} is an ordinal \gamma_{\alpha,\varepsilon} < \lambda.

3. (c) The following three properties hold:

   1. (c1): \mathbb{G}^1_{\alpha,\varepsilon} \subseteq \mathbb{G}^2_{\alpha,\varepsilon} are \aleph_1-free,

   2. (c2): if \alpha < \beta then \mathbb{G}^2_{\alpha, \varepsilon} \cap \mathbb{G}^1_{\beta, \varepsilon} = \mathbb{G}^1_{\alpha, \varepsilon},

   3. (c3): if \alpha < \beta and \alpha \notin S_\varepsilon, then \mathbb{G}^2_\beta/(\mathbb{G}^1_\beta + \mathbb{G}^2_\alpha) is \aleph_1-free.

4. (d) If \alpha < \beta and \alpha \notin S_\varepsilon, then \mathbb{G}^1_{\beta,\varepsilon}/\mathbb{G}^1_{\alpha,\varepsilon} and \mathbb{G}^2_{\beta,\varepsilon}/\mathbb{G}^2_{\alpha,\varepsilon} are \aleph_1-free.

5. (e) If \delta \in S_\varepsilon, then there are no f and \langle \dot{\mathbb{L}}_\zeta: \zeta \leq \delta \rangle equipped with the following properties:

   1. (e1) \langle \dot{\mathbb{L}}_\zeta: \zeta \leq \delta \rangle is a construction over \mathbb{G}_{\delta+1, \varepsilon}^1,

   2. (e2) f \in {\rm Hom}(\mathbb{G}^2_{\delta+1, \varepsilon}, \dot{\mathbb{L}}_{\delta}),

   3. (e3) f extends f^\varepsilon_{\delta, i} \cup {\rm id}_{\mathbb{G}^1_{\delta+1, \varepsilon}} where i \in I_\varepsilon and

      \langle f^\varepsilon_{\delta, i}, \bar{\mathbb{L}}^\varepsilon_{\delta, i}=\langle \mathbb{L}^\varepsilon_{\delta, i, \zeta}: \zeta \leq \delta \rangle \rangle

      is coded by X^\varepsilon_i, under the identification given by \bold g_\varepsilon. Also, f^\varepsilon_{\delta, i} is in {\rm Hom}(\mathbb{G}^2_{\delta, \varepsilon}, \mathbb{L}^\varepsilon_{\delta, i, \delta}) and \bar{\mathbb{L}}^\varepsilon_{\delta, i} is a construction:

      \xymatrix@=3pc{ & \dot{\mathbb{L}}_{\delta} \\ \mathbb{G}^1_{\delta+1, \varepsilon} \ar[r]_\subseteq\ar [ru]^{\subseteq} & \mathbb{G}^2_{\delta+1, \varepsilon} \ar[u]^{\nexists{f}} & \mathbb{G}^2_{\delta, \varepsilon}\ar[lu]_{{f}^\varepsilon_{\delta,i}} \ar[l]^{\subseteq} & }

   4. (e4) \mathbb{L}^\delta_\zeta \subseteq \dot{\mathbb{L}}_\zeta for \zeta \leq \delta,

   5. (e5) \mathbb{L}^\delta_\zeta = \mathbb{L}^\delta_{\zeta+1} \cap \dot{\mathbb{L}}_\zeta,

   6. (e6) \dot{\mathbb L}_{\zeta+1}/ (\mathbb{L}^\delta_{\zeta+1}+\dot{\mathbb L}_\zeta) is free.

For \alpha=0, set \mathbb{G}^1_{0,\varepsilon} = \mathbb{G}^2_{0,\varepsilon} = \{0\}. For the limit ordinal \delta, we set \mathbb{G}^\iota_{\delta,\varepsilon}: = \bigcup\limits_{\alpha < \delta}\mathbb{G}^\iota_{\alpha,\varepsilon}. Let us show that items (c) and (d) continue to hold. By the induction hypothesis and for all \alpha < \beta < \delta with \alpha \notin S, \mathbb{G}^\iota_{\beta, \varepsilon} / \mathbb{G}^\iota_{\alpha, \varepsilon} and \mathbb{G}^2_\beta/(\mathbb{G}^1_\beta + \mathbb{G}^2_\alpha) are \aleph_1-free, and since we have a club C_\delta of \delta which is disjoint to S, it immediately follows from Lemma 3.1 that the groups \mathbb{G}^\iota_{\delta, \varepsilon}, \mathbb{G}^\iota_{\delta, \varepsilon}/ \mathbb{G}^\iota_{\alpha, \varepsilon} and \mathbb{G}^2_{\delta, \varepsilon}/(\mathbb{G}^1_{\delta, \varepsilon} + \mathbb{G}^2_{\alpha, \varepsilon}) are \aleph_1-free for all \alpha < \delta with \alpha \notin S_\varepsilon.

Now suppose that \delta < \lambda and we have defined the groups \mathbb{G}^\iota_{\alpha, \varepsilon} for \iota=1, 2 and ordinals \alpha \leq \delta. We would like to define the groups \mathbb G^1_{\delta+1, \varepsilon} and \mathbb G^2_{\delta+1, \varepsilon} so that the (a)-(e) continue to hold.

If \delta \notin S_\varepsilon, then we set

and

We leave to the reader to check that the items presented from (a) to (d) all are valid, and recall that there is nothing to prove for case (e).

Now suppose that \delta \in S_\varepsilon. We define the groups \mathbb{G}^1_{\delta+1, \varepsilon} and \mathbb{G}^2_{\delta+1, \varepsilon} such that items (a)-(e) above continue to hold, and further we have:

1. (f) \mathbb{G}^2_{\delta+1,\varepsilon}/(\mathbb{G}^1_{\delta,\varepsilon} + \mathbb{G}^2_{\delta,\varepsilon}) is not free,

2. (g) if \gamma \in \delta \backslash S_\varepsilon, then \mathbb{G}^2_{\delta+1,\varepsilon}/(\mathbb{G}^1_{\delta,\varepsilon} + \mathbb{G}^2_{\gamma,\varepsilon}) is \aleph_1-free.

As \delta \in S_\varepsilon, we have {\rm cf}(\delta) = \aleph_0, so let \langle \gamma_{\delta,n}:n < \omega\rangle be an increasing sequence of successor ordinals < \delta with limit \delta.

Let n<\omega be such that \gamma_{\delta, n} \notin S_\varepsilon. It turns out that the abelian groups \mathbb G^1_{\gamma_{\delta, n}, \varepsilon} \subseteq \mathbb{G}^2_{\gamma_{\delta, n}, \varepsilon} are suited well in the hypotheses of Lemma 4.5.

The notation \Sigma_{\delta, \varepsilon} stands for the following set:

Given any i \in \Sigma_{\delta, \varepsilon}, and according to Lemma 4.5, we can find the \aleph_1-free groups \mathbb G^{1, i}_{\delta+1, \varepsilon} \subseteq \mathbb G^{2, i}_{\delta+1, \varepsilon} such that there for all constructions \langle\dot{\mathbb L}_\zeta: \zeta \leq \delta \rangle over \mathbb G^{1,i}_{\delta+1, \varepsilon} as in item (e), if we set \mathbb{L}=\dot{\mathbb{L}}_\delta, then there is no \bold f \in {\rm Hom}(\mathbb{G}^{2, i}_{\delta+1, \varepsilon}, \mathbb L) such that \bold f \supseteq f^\varepsilon_{\delta, i} \cup {\rm id}_{\mathbb G^{1, i}_{\delta+1, \varepsilon}}. For \iota=1,2 we look at

i.e., the pure closure of \langle \mathbb G^\iota_{\delta, \varepsilon} \cup \bigcup\{\mathbb G^{\iota, i}_{\delta+1, \varepsilon}: i \in \Sigma_{\delta, \varepsilon} \} \rangle in \widehat{\mathbb C_\varepsilon}. Let us show that the hypothesis (a)-(e) hold. We first show that the group \frac{\mathbb{G}^2_{\delta+1, \varepsilon}}{\mathbb{G}^2_{\alpha, \varepsilon}} is \aleph_1-free, provided that \alpha is a successor ordinal. Let \mathbb{K} be any countable subgroup of \mathbb{G}^2_{\delta+1, \varepsilon}. We are going to show that \frac{\mathbb{K}+\mathbb{G}^2_{\delta+1, \varepsilon}}{\mathbb{G}^2_{\alpha, \varepsilon}} is free. There is an \omega-sequence \{i_m:m \in \omega\} together with a countable subgroup \mathbb{I}\subset\mathbb{G}^2_{\delta, \varepsilon} such that \mathbb{K} is the subgroup generated by \mathbb{I} together with some countable subset \{w_{n,{i_m}}:n, m \in \omega\} of \bigcup\limits_{i\in \Sigma_{\delta,\varepsilon}}\mathbb G^{\iota, i}_{\delta+1, \varepsilon}. We can assume that for all n, m, y_{\varepsilon, \alpha_{n,{i_m}}}\in \mathbb{I}. Choose an increasing sequence of ordinals \{\alpha_k: k < \omega\} with limit \delta such that \alpha_0=\alpha and for all m\in\omega and all but finitely many of n, \alpha_{n,i_m}\in\{\alpha_k :k < \omega \}. Notice that for any successor ordinal \gamma < \delta

which is free by the induction hypothesis. So for such \gamma, \mathbb{I} \cap \mathbb{G}^2_{\gamma, \varepsilon} is a direct summand of \mathbb{I}. Inductively choose subgroups \mathbb{I}_k so that:

for all k. Hence

In view of Theorem 5.2(b), we are able to choose \{n(m):m < \omega\} so that:

• the collections \{\alpha_{n,i_m} :n(m) < n\} are pairwise disjoint.

• \{\alpha_{n,i_m} :n(m) < n\}\subset \{\alpha_k: k < \omega\}.

We observe that \frac{\mathbb{K}+\mathbb{G}^2_{\delta, \varepsilon}}{\mathbb{G}^2_{\delta, \varepsilon}} is isomorphic to the direct sum of \bigoplus_k \mathbb{I}_k together with the group freely generated by

and

From this, the claim follows. By a similar argument, the group \frac{\mathbb{G}^1_{\delta+1, \varepsilon}}{\mathbb{G}^1_{\alpha, \varepsilon}} is \aleph_1-free, provided that \alpha is a successor ordinal.

In the same vein, we also observe that the group \mathbb G^1_{\delta+1, \varepsilon} is \aleph_1-free and the hypotheses (a)-(e) continue to hold.

The rest of the argument is similar to Theorem 4.7. Let us elaborate the main idea of the proof. For \iota=1, 2 and \varepsilon < \lambda, we define:

Set also

• (h1): \mathbb{G}_{\alpha, \varepsilon}=\mathbb{G}^{2}_{\alpha, \varepsilon}/ \mathbb{G}^{1}_{\alpha, \varepsilon},

• (h2): \mathbb{G}_{\varepsilon}=\bigcup\limits_{\alpha < \lambda}\mathbb{G}_{\alpha, \varepsilon},

• (h3): \bar{\mathbb{G}}=\langle \mathbb{G}_{\varepsilon}: \varepsilon < \lambda \rangle.

Let us show that \boxtimes^1_{\lambda,\mu_1} is satisfied. By items (c) and (d) of the construction, \mathbb{G}^1_\varepsilon \subseteq \mathbb{G}^2_\varepsilon are \aleph_1-free and \mathbb{G}_{\varepsilon} is \aleph_1-free as well. Let us now show that \boxtimes^1_{\lambda,\mu_1}(c) is satisfied as well. Suppose by contradiction that there are \varepsilon < \lambda, a construction \mathbb{L} by \{\mathbb{G}_\alpha:\alpha \in \lambda \setminus \varepsilon\} over \mathbb{G}^1_\varepsilon, witnessed by \langle \mathbb{L}_\varepsilon: \varepsilon \leq \lambda \rangle, and there is a homomorphism {\bf g} from \mathbb{G}^2_\varepsilon into \mathbb{L} which extends {\rm id}_{\mathbb{G}^1_\varepsilon}:

Without loss of generality we can assume that \mathbb{L} has size \lambda and that its universe is \lambda. Let X \subseteq \lambda be codes \langle \bf g, \langle \mathbb{L}_\varepsilon: \varepsilon \leq \lambda \rangle \rangle. The set

is a club, thus we can find some \delta \in E \cap S_\varepsilon and some i with \delta(i)=i such that (M^\varepsilon_i, X^\varepsilon_i) \equiv_{M^\varepsilon_i \cap \lambda} (N_\varepsilon, X). It then follows that M^\varepsilon_i \cap \mathbb{G}^2_{\varepsilon}= M^\varepsilon_i \cap \mathbb{G}^{2, i}_{\delta, \varepsilon}, and since (M^\varepsilon_i, X^\varepsilon_i) \equiv_{M^\varepsilon_i \cap \lambda} (N_\varepsilon, X), we can easily observe that X^\varepsilon_i codes

Now by elementarily and the choice of \bf g,

It follows that i \in \Sigma_{\delta, \varepsilon}, and hence by our construction, there is no \bold f \in {\rm Hom}(\mathbb{G}^{2, i}_{\delta+1, \varepsilon}, \mathbb L) such that \bold f extends f^\varepsilon_{\delta, i} \cup {\rm id}_{\mathbb G^{1, i}_{\delta+1, \varepsilon}}. This property is conveniently summarized by the subjoined diagram:

This is not possible, as \bf g is such an extension. This shows that there is no homomorphism {\bf g} as above and the result follows.

(2): This is similar to the proof of (1). Take a sequence \langle S_\varepsilon: \varepsilon < 2^\lambda \rangle of almost disjoint subsets of \lambda such that each S_\varepsilon is stationary and proceed as before.

This follows using Theorem 3.5 and Theorem 5.4.

Suppose (1) holds as witnessed by the sequence \langle \mathbb G_X ,\mathbb H^X: X \subseteq \mu \rangle and let (\mu, R) be a bipartite graph. Given any \alpha, \beta < \mu set Y_{\alpha}=\{\alpha\} and X_\beta=\{ \alpha < \mu: \alpha R \beta \}. Now define

1. i) \mathbb G_\alpha:=\mathbb G_{Y_\alpha} and

2. ii) \mathbb K_\alpha:= \mathbb H^{X_\alpha}.

This is now straightforward to see

Thus, the family \langle \mathbb G_\alpha, \mathbb K_\alpha: \alpha < \mu \rangle is as required.

Suppose (1) holds. Let \langle X_\alpha: \alpha < \lambda \rangle and \langle Y_\beta: \beta < \lambda \rangle be two enumerations of \mathcal{P}(\mu). Let us define the bipartite graph (\lambda, R) as

By (1), there exists a family \langle \mathbb G_\alpha, \mathbb K_\alpha: \alpha < \lambda \rangle, of abelian groups such that for all \alpha, \beta < \lambda,

For X, Y \subseteq \mu, let \alpha, \beta < \lambda be such that X_\beta=X and Y_\alpha=Y and set

1. \bullet \mathbb G_{Y}:=\mathbb G_\alpha;

2. \bullet \mathbb H^X := \mathbb K_\beta.

Then

The lemma follows.

Let us first show that {\mathbb P}_* is \lambda-distributive. Thus suppose that \mu < \lambda and let f: \mu \to V be a function in the generic extension by {\mathbb P}_*. We will show that f is in V. Let \dot f be a {\mathbb P}_*-name for f and let p \in {\mathbb P}_* be such that

By induction on \xi \leq \mu we define a sequence \langle C_\xi: \xi \leq \mu \rangle of closed subsets of \lambda and an increasing sequence \langle p_\xi: \xi \leq \mu \rangle of conditions in {\mathbb P}_* such that:

1. C_\xi is a closed subset of \alpha_{p_\xi}+1 with \alpha_{p_\xi} \in C, C_\xi \cap \alpha_{p_\xi} is unbounded in \alpha_{p_\xi} and C_\xi \cap E_{p_{\xi}}=\emptyset,

2. C_{\xi+1} \cap \alpha_{p_{\xi}}+1 = C_\xi,

3. if \xi \leq \mu is a limit ordinal, then C_\xi \cap \alpha_{p_\xi}=\bigcup\limits_{\zeta < \xi}C_{p_\zeta},

4. p_0=p,

5. for each \xi < \mu, p_{\xi+1} decides \dot{f}(\xi), say p_{\xi+1} \Vdash“\dot f(\xi)=a_\xi”.

6. If \xi \leq \mu is a limit ordinal and \langle p_\zeta: \zeta < \xi \rangle is defined, then p_\xi is the least upper bound of the sequence \langle p_\zeta: \zeta < \xi \rangle defined as follows:

   1. \alpha_{p_\xi}=\sup\limits_{\zeta < \xi}\alpha_\zeta,

   2. E_{p_\xi}= \big(\bigcup\limits_{\zeta < \xi}E_{p_\zeta} \cup \{ \alpha_{p_\xi} \} \big) \cap S^\lambda_{\aleph_0},

   3. for s \in S and \beta < \alpha_{p_\xi}, G^{p_\xi}_{s, \beta}=G^{p_\zeta}_{s, \beta} for some, and hence all, \zeta < \xi with \beta < \alpha_{p_\zeta},

   4. if (s, t)\notin R and \beta < \alpha_{p_\xi}, then \bold x^{p_\xi}_{s, t, \beta}=\bold x^{p_\zeta}_{s, t, \beta} for some, and hence all, \zeta < \xi with \beta < \alpha_{p_\zeta},

   5. for s \in S, G^{p_\xi}_{s, \alpha_{p_\xi}}=\bigcup\limits_{\zeta < \xi}G^{p_\zeta}_{s, \alpha_{p_\zeta}},

   6. if (s, t)\notin R, then \bold x^{p_\xi}_{s, t, \alpha_{p_\xi}}=\bigcup\limits_{\zeta < \xi}\bold x^{p_\zeta}_{s, t, \alpha_{p_\zeta}}, which is defined in the natural way.

Given \xi \leq \mu, let us show that p_\xi as defined above is a condition, as then it is clear that p_\xi extends all p_\zeta’s , for \zeta < \xi. Items (1) and (5) from Definition 6.4 are clearly satisfied.

In order to see clause (2), it suffices to show that for each s \in S, G^{p_\xi}_{s, \alpha_{p_\xi}} is free. We have C_{\xi} \cap \alpha_{p_\xi} is a club of \alpha_{p_\xi} and

Furthermore, the sequence \langle G^{p_\xi}_{s, \gamma}: \gamma \in C_\xi \rangle is an increasing and continuous sequence of free groups, such that if \gamma < \beta are successive points in C_\xi, then G^{p_\xi}_{s, \beta}/G^{p_\xi}_{s, \gamma} is free. It follows that G^{p_\xi}_{s, \alpha_{p_\xi}} is free as required.

The set E_{p_\xi} does not reflect as witnessed by C_\xi, hence clause (3) is satisfied.

To show that clause (4) holds, let \gamma <\beta < \alpha_{p_\xi} with \gamma \notin E_{p_\xi}. Let \zeta < \xi be such that \beta < \alpha_{p_\zeta}. This implies \gamma \notin E_{p_\zeta}, and consequently

is free.

Let q=p_\mu, and let h=\langle a_\xi: \xi < \mu \rangle. Then h \in V and q \Vdash“\dot {f}=h”. As p was arbitrary, we are done.

Now as \lambda^{<\lambda}=\lambda and {\mathbb P}_* \subseteq {\mathscr H}(\lambda), we have |{\mathbb P}_*|=\lambda and hence it clearly satisfies the \lambda^+-c.c.

This is standard, so we just sketch the proof. Let p \in {\mathbb P}_* and suppose that

Let \chi > \lambda^+ be large enough regular, \lhd be a well-ordering of {\mathscr H}(\chi) and let M \prec ({\mathscr H}(\chi), \in, \lhd) be an elementary submodel of {\mathscr H}(\chi) such that:

1. |M| < \lambda,

2. p, \dot{C}, {\mathbb P}_*, \cdots \in M,

3. M \cap \lambda=\delta for some \delta \in S^\lambda_{\aleph_0}.

By induction on n<\omega we define an increasing sequence \langle p_n: n<\omega \rangle of conditions, together with a sequence \langle C_n: n<\omega \rangle satisfying the following:

• (4) p \leq p_0,

• (5) p_{n+1} decides \dot{C} \cap \alpha_{p_n} to be C_n,

• (6) p_{n+2} \Vdash“\dot{C} \cap (\alpha_{p_n}, \alpha_{p_{n+1}}) \neq \emptyset”,

• (7) \sup\limits_{n<\omega}\alpha_{p_n}=\delta.

We are now ready to define

as follows:

1. (8) for s \in S and \beta < \delta, \mathbb{G}^q_{s, \beta}=\mathbb{G}^{p_n}_{s, \beta}, where n is such that \alpha_{p_n}>\beta,

2. (9) for s\in S, \mathbb{G}^q_\delta=\bigcup\limits_{n<\omega}\mathbb{G}^{p_n}_{s, \alpha_{p_n}},

3. (10) E_q = \bigcup\limits_{n<\omega}E_{p_n} \cup \{\delta\},

4. (11) if (s, t) \notin R and \beta<\delta, \mathbf x^q_{s,t,\beta}=\mathbf x^{p_n}_{s,t,\beta}, for some n with \alpha_{p_n} > \beta,

5. (12) \mathbf x^q_{s,t,\delta} is the exact sequence that is the direct limit of the sequence \mathbf x^{p_n}_{s,t,\alpha_{p_n}}.

It is easily seen that q \in {\mathbb P}_* is well-defined and it forces \delta \in \dot{C} \cap \dot{E}, which completes the proof.

It is clear from Definition 6.4(a)(2) that \mathbb{G}_s is a strongly \lambda-free abelian group of size \lambda. Let us show that it is not free. Recall that we are in \bold V, and let \mathbb{A} be a \lambda-free abelian group of size \lambda which is not free. We may assume that the universe of \mathbb{A} is \lambda. Let also \langle A_\alpha: \alpha < \lambda \rangle be a filtration of \mathbb A. It follows that the set

is stationary. We apply an argument similar to the proof of Lemma 6.8, and we are able to show E_{\mathbb A} remains stationary in \bold V[\bold{G}_*]. In particular, \mathbb A remains non-free in the generic extension \bold V[\bold{G}_*].

For each \gamma < \lambda let D_\gamma be the set of all p \in {\mathbb P}_* such that there exists \beta such that:

• \alpha_p=\beta+1 >\gamma,

• \mathbb{G}^p_{s, \beta+1} includes \mathbb{A}_\beta as a subgroup.

It is easily seen that each set D_\gamma is dense in {\mathbb P}_*. Now for each \gamma< \lambda pick some p_\gamma \in D_\gamma \cap \bold G_* and set \alpha_{p_\gamma}=\beta_\gamma+1 > \gamma.

Now we look at the following commutative diagram:

Taking direct limits of these directed systems, lead us to a natural inclusion map

As \mathbb{A} is non-free, it follows from Fact 6.9 that \mathbb{G}_s is also non-free. This completes the proof.

Suppose towards contradiction that the exact sequence \mathbf x_{s_*,t_*} splits and let p \in {\mathbb P}_* and \let\theaccent\tilde\skewfactor- \mathpalette\oalign{3h\crcr\hidewidth \vbox to.2ex{\hbox{3}\vss}\hidewidth} be such that

As in the proof of Lemma 6.8, we can find an extension q \leq p such that

1. \alpha_q\in E_q \cap S^\lambda_{\aleph_0},

2. q decides \let\theaccent\tilde\skewfactor- \mathpalette\oalign{3h\crcr\hidewidth \vbox to.2ex{\hbox{3}\vss}\hidewidth} \restriction \mathbb G^q_{s_*, \alpha_q}, say

   q \Vdash \text{``} \let\theaccent\tilde\skewfactor- \mathpalette\oalign{$3h$\crcr\hidewidth \vbox to.2ex{\hbox{$3\skew\skewfactor\theaccent{}$}\vss}\hidewidth} \restriction \mathbb G^q_{s_*, \alpha_q} = h_* \text{''}.

We will need the following claim.

Let us continue the proof of Lemma 6.11. By the way we defined r, see the above itemized properties r_1),\ldots,r_6), we have

In order to see this, let r' \leq r be such that r' decides \let\theaccent\tilde\skewfactor- \mathpalette\oalign{3h\crcr\hidewidth \vbox to.2ex{\hbox{3}\vss}\hidewidth} \restriction \mathbb{G}^{r'}_{s_*, \alpha_r}. Then

This contradicts the choice of g'. The argument of Lemma 6.11 is now completed.

The fact that {\mathbb Q}^W_{s, t, \mathbf{x}} is \lambda-closed follows from a combination of Definition 6.13(1) along with Definition 6.4(4). This allows us to take unions (or direct limits) at limit stages and still have a condition. According to \Delta-system lemma, the forcing is \lambda^+-c.c., and claimed.

It suffices to show that any exact sequence

with \mathbb{H} of size \lambda splits. Let \dot{\mathbf x} be a {\mathbb P}_* \ast {\mathbb P}-name for \mathbf x which we may assume that \dot{\mathbf x} \in {\mathscr H}(\lambda^+). Furthermore, we can find some \alpha<\lambda^+ such that \dot{\mathbf x} is a {\mathbb P}_* \ast {\mathbb P}_\alpha-name, and then by the choice of \Phi, we may also assume that \Phi(\alpha)=\dot{\mathbf x}. By definition of forcing notion, we know

and consequently

This completes the proof.

Since the argument is similar to the proof of Lemma 6.11, we just present the sketch of proof. Here, we work in V. Let us first combine Lemma 6.6 along with Lemma 6.15, and note that the set of conditions of the form

such that p, q_\alpha \in V, is dense in {\mathbb P}_* \ast {\mathbb P}_{\lambda^+}. Suppose towards contradiction that the exact sequence \mathbf x_{s,t} splits. This gives us

and \let\theaccent\tilde\skewfactor- \mathpalette\oalign{3h\crcr\hidewidth \vbox to.2ex{\hbox{3}\vss}\hidewidth} satisfying:

As before, we can find an extension

equipped with the following two properties:

1. \alpha_{q}\in E_{q} \cap S^\lambda_{\aleph_0},

2. (q, \langle q_\alpha: \alpha < \lambda^+ \rangle) decides \let\theaccent\tilde\skewfactor- \mathpalette\oalign{3h\crcr\hidewidth \vbox to.2ex{\hbox{3}\vss}\hidewidth} \restriction \mathbb G^{q}_{s, \alpha_{q}}, say

   (q, \langle q_\alpha: \alpha < \lambda^+ \rangle) \Vdash \text{``} \let\theaccent\tilde\skewfactor- \mathpalette\oalign{$3h$\crcr\hidewidth \vbox to.2ex{\hbox{$3\skew\skewfactor\theaccent{}$}\vss}\hidewidth} \restriction \mathbb G^{q}_{s, \alpha_{q}} = h_* \text{''}.

In view of Claim ? we can find an extension r of q such that (r, \langle q_\alpha: \alpha < \lambda^+ \rangle) extends (q, \langle q_\alpha: \alpha < \lambda^+ \rangle) and also, it forces

We get a contradiction and the lemma follows.