More on Weak Diamond

(1) Should be clear.

(2) If \alpha = 0 then (\circledast_{\mathcal A}) holds trivially because there is no such S_1.

If \alpha is a limit ordinal then the condition holds because for every S_1 \in {\rm ID}^-_\alpha (F) there is \beta < \alpha such that S_ 1 \in {\rm ID}^-_\beta (F) and we can use the induction hyp.

Lastly, if \alpha = \beta + 1 this is easy too.

(3) For the first sentence {\rm ID}^-_\alpha (F) is a normal ideal by its definition; this implies {\rm ID}^a(F) is a normal ideal by the second sentence. We still have to prove the second sentence.

By induction on \gamma<\lambda and then by induction on \alpha<\gamma we show that (\forall\gamma<\lambda)(\forall\alpha<\gamma)( {\rm ID}_\alpha(F)\subseteq{\rm ID}_\gamma(F)). If \gamma=1 then this follows immediately from definitions; similarly if \gamma is limit. So suppose now that \gamma=\gamma_0+1 and we proceed by induction on \alpha\leq \gamma_0. There are no problems neither when \alpha=0 nor when \alpha is limit. So suppose that \alpha=\beta+1<\gamma (so \beta<\gamma_0). By the inductive hypothesis we know that {\rm ID}_\beta(F)\subseteq{\rm ID}_{\gamma_0}(F). Let A\in{\rm ID}_{\beta+1}(F). By (2) there are A_\xi\in{\rm ID}^-_{\beta+1} (for \xi<\lambda) such that A=\mathop{\nabla}\limits_{\xi<\lambda}A_\xi. Now look at the definition of {\rm ID}_{\beta+1}^-(F): since {\rm ID}_\beta(F)\subseteq{\rm ID}_{ \gamma_0}(F) we see that A_\xi\in{\rm ID}^-_{\gamma_0+1}(F). Hence A\in {\rm ID}_\gamma.

(4) By induction on \alpha we show that: if \bar{B}=\langle B_\ell:\ell\leq m\rangle, where B_\ell\subseteq B_{\ell+1} (for \ell<m) and B_m\in{\rm ID}_\alpha(F) then \bar{B} has an F–representation.

Case 1: \alpha=0.
Thus the set B_m is not stationary and we may pick up a club E of \lambda disjoint from B_m. Let E=\{\alpha_\zeta:\zeta<\lambda\} be the increasing enumeration. Put T=\{\langle\rangle\}\cup\{\langle i\rangle:i< \lambda\}, \alpha_{\langle\rangle}=1, \ell_{\langle\rangle}=\ell_{ \langle i\rangle}=m, B^\ell_{\langle\rangle}=B_\ell and B^\ell_{\langle i\rangle}=B_\ell\cap\alpha_{i+1}. Now check.

Case 2: \alpha is limit.
It follows from (2) that B_\ell=\mathop{\nabla}\limits_{i<\lambda} B_{\ell,i} for some B_{\ell,i}\in\bigcup\limits_{\beta<\alpha}{\rm ID}_\beta(F). Let B_{\ell,i}' be defined as follows:

if i=(m+1)j+t, \ell<t\leq m then B_{\ell,i}'=\emptyset,

if i=(m+1)j+t, t\leq m, t\leq \ell then B_{\ell,i}'=B_{\ell,i}.

Then for each i,\ell we may find \langle B^{i,\ell}_\eta, f^{i,\ell'}_\eta,\alpha^i_\eta: \eta\in T_i,\ \ell<\ell^{i,1}_\eta, \ \ell'< \ell^{i,2}_\eta\rangle satisfying clauses (a)–(d) and such that \langle B^{\ell,i,k}_{\langle\rangle}: k\leq k^1_\eta\rangle = \langle B'_{\ell,i}: \ell\leq m\rangle (by the induction hypothesis). Put

Checking that \langle B^\ell_\eta,f^{\ell'}_\eta,\alpha_\eta:\eta\in T,\ \ell\leq\ell_\eta,\ \ell'\leq\ell_\eta'\rangle is as required is straightforward.

Case 3: \alpha=\beta+1.
By (2) above and the proof of Case 2 we may assume that B_m\in{\rm ID}^-_\alpha (F). It follows from the definition of {\rm ID}^-_\alpha(F) that there are f_\ell\in{}^{\textstyle\lambda}2 (for \ell\leq m) such that

and hence B^\oplus\stackrel{\rm def}{=}\bigcup\limits_{\ell\leq m} B^\oplus_\ell\in{\rm ID}_\beta(F). Therefore B^*_\ell\stackrel{\rm def}{=} B_\ell\cap B^\oplus\in{\rm ID}_\beta(F). Now apply the inductive hypothesis for \beta and \bar{B}^*=\langle B^*_\ell:\ell\leq m\rangle to get the sequences \langle B^{\ell,*}_\eta,f^{k,*}_\eta: \eta\in T^*,\ \ell\leq \ell^*_\eta,\ k\leq k^*_\eta\rangle satisfying clauses (a)–(d) and such that \langle B^{\ell,*}_{\langle\rangle}:\ell\leq\ell^*_\eta\rangle= \langle B^*_\ell:\ell\leq m\rangle. Put

(5) If f^\ell_\eta is not defined then choose f^* as it.

(6), (7), (8), (9) Easy too.

(a) Check.

(b) The main point is proving {\rm ID^a}(F)\subseteq{\rm ID}_1(F^\oplus).

Suppose that B\in{\rm ID^a}(F). We are going to show that then B\in{\rm ID}^-_1( F^\oplus). So suppose that B'\subseteq B. We want to find g\in {\rm DOM}_\lambda(F^\oplus) such that the set

is in {\rm ID}_0(F^\oplus) (what just means that it is non-stationary). Since B\in{\rm ID^a}(F) we have B'\in{\rm ID^a}(F), so by 2.3(4) we may find \langle B^\ell_\eta,f^k_\eta,\alpha_\eta:\eta\in T,\ \ell\leq\ell_\eta,\ k<k_\eta\rangle such that the clauses (a)–(d) of 2.3(4) are satisfied with \ell_{\langle\rangle}=0, B'=B^0_{\langle\rangle}. Define g as follows. For \beta<\lambda let T_\beta=T\cap{}^{\textstyle \omega{>}}\beta and

Now look at the demands in 2.5(2) – they are exactly what 2.3(4) guarantees us.

1) Let F_1:\bigcup\limits_{\alpha<\lambda}{}^{\textstyle\alpha}({ \mathcal H}(\lambda))\longrightarrow 2. Let h_0 be a one-to-one function from {\mathcal H}(\lambda) to {}^{\textstyle\lambda{>}}2, say h_0(\eta)=\langle \ell_{\eta,i}:i<\ell g\/(h_0(\eta))\rangle. Define a function h_1:{\mathcal H}(\lambda)\longrightarrow{}^{\textstyle\lambda{>}}2 by:

Next, by induction on \ell g\/(\eta), we define a function h^+:\bigcup\limits_{ \alpha<\lambda}{}^{\textstyle\alpha}({\mathcal H}(\lambda))\longrightarrow {}^{\textstyle \lambda{>}}2 as follows:

and if \eta \in \bigcup\limits_{\alpha<\lambda}{}^{\textstyle\alpha}({ \mathcal H}(\lambda)) has length the limit ordinal \delta then h^+( \eta ) = \cup \{h^+(\eta {\restriction} \beta ): \beta < \delta \}

Clearly h^+ is one to one with the right domain and range.

Finally we define a colouring F_2:{}^{\textstyle\lambda{>}}2\longrightarrow 2 by

It is easy to check that F_2 is as required.

Straightforward.

1) It follows from 2.3(3) that {\rm ID}_\gamma(F) increases with \gamma and \beta < \gamma , {\rm ID}_ \beta (F) = {\rm ID}_ {\beta + 1} implyies {\rm ID}_ \beta (F) = {\rm ID}_ \gamma; so the assertion should be clear.

2) By the definition (and 2.12(c)).

3) Assume that {\rm ID}_\alpha(F) is \lambda^+–saturated and \lambda\notin{{\rm WDmId}_\lambda}. By induction on \beta<\lambda^+ we try to choose colourings F_\beta such that

1. (a) {\rm ID}(F) \subseteq {\rm ID} (F_ \alpha )

2. (b) if \beta<\gamma then {\rm ID^a}(F_\beta)\subseteq{\rm ID^a}(F_\gamma),

3. (c) {\rm ID^a}(F_\beta)\neq{\rm ID^a}(F_{\beta+1}).

So we let F_0=F. If \beta is limit then we use 2.9(2) to choose F_\beta so that (\forall\gamma<\beta)(F_\gamma\leq F_\beta). Finally, if \beta=\gamma+1 then we let F_\beta'=(F_\gamma)^\otimes (so {\rm ID^a}(F_\gamma)\subseteq{\rm ID}_1(F_\beta')={\rm ID^a}(F_\beta')\subseteq{{\rm WDmId}_\lambda}). If {\rm ID^a}(F_\beta')\neq{{\rm WDmId}_\lambda} then we choose a set A\in{{\rm WDmId}_\lambda}\setminus {\rm ID^a}(F_\beta') and F_\beta^* witnessing A\in{{\rm WDmId}_\lambda}. We may assume that (\forall\alpha\in\lambda\setminus A)(\forall\eta\in{}^{\textstyle \alpha}2) (F^*_\beta(\eta)=0). Now take a colouring F_\beta such that F_\beta', F^*_\beta\leq F_\beta.

After carrying out the construction choose S^0_\beta\in{\rm ID^a}(F_{\beta+1}) \setminus{\rm ID^a}(F_\beta) (for \beta<\lambda^+) and let S_\beta=S^0_\beta \setminus\mathop{\nabla}\limits_{\gamma<\beta}S^0_\gamma. Then \langle S_\beta:\beta< \lambda^+\rangle is a sequence of pairwise disjoint members of {\mathcal P}(\lambda)\setminus{\rm ID^a}(F_0)\subseteq{\mathcal P}(\lambda)\setminus {\rm ID}_\alpha(F), contradicting our assumptions.

4), 5), 6) Easy too.

Assume not. Then for each \eta\in{}^{\textstyle\lambda}2 the set B_{\alpha^*}[\eta]\stackrel{\rm def}{=}\lambda\setminus S_{\alpha^*[\eta]} is in {\rm ID^b}(F^\otimes). Now,

(see 2.7).

It follows from ?, ? that

But \lambda\setminus S_{\alpha^*[\eta]}[\eta]\in{\rm ID^b}(F^\otimes), so {\mathcal P}(\lambda\setminus S_{\alpha^*[\eta]}[\eta])\subseteq{\mathfrak B}( F^\otimes) so we get a contradiction.

1) We try to choose by induction on \alpha<\lambda^+ sets S_\alpha,S_{\alpha,0},S_{\alpha,1} such that

1. (a) S_\alpha\subseteq\lambda,

2. (b) S_\alpha=S_{\alpha,0}\cup S_{\alpha,1}, S_{\alpha,0}\cap S_{\alpha,1}=\emptyset,

3. (c) if \beta<\alpha and \ell<2 then

   S_\beta\subseteq S_\alpha\mod{\rm ID^b}(F^\otimes)\quad\mbox{ and }\quad S_{\beta,\ell}\subseteq S_{\alpha,\ell}\mod{\rm ID^b}(F^\otimes),

4. (d) the sets S_{\alpha,0},S_{\alpha, 1} witness that S_\alpha \in {\rm ID^c}(F^\otimes) (see 2.2(4)).

At some stage \alpha<\lambda^+ we have to be stuck (as {\rm ID^b}(F^\otimes) is \lambda^+–saturated) and then (S_\alpha,S_{\alpha,0},S_{\alpha,1}) can serve as (S^*_F,S^*_0,S^*_1).

2) By the choice of S^*_F, for some \ell<2 we have

so let Y\subseteq X_\ell be such that Y\notin\{F^\otimes(f)\cap X_\ell: f\in{}^{\textstyle \lambda} 2 \}. Let \rho= 0_Y\cup 1_{X_\ell\setminus Y}. Since without loss of generality \ell=1, we are done.

Case 1. \mu\leq 2^{\|\alpha\|} for some \alpha<\lambda.
Let \rho_i\in {}^{\textstyle\alpha}2 for i<\mu be pairwise distinct. For \eta\in {}^{\textstyle\lambda\!>}2 let h_i(\eta)=\rho_i{}^\frown\!\eta. Define F by:

It is easy to see that F: {}^{\textstyle\lambda\!>}2\longrightarrow\{0,1\} and h_i exemplifies that F_i\leq F.

Case 2. \mu=\lambda.
For \eta\in {}^{\textstyle\lambda\!>}2, i<\mu, we define h_i(\eta ) \in {}^{ i + 1 + \ell g\/(\eta )}2 as follows: for

\gamma < i + 1 + \ell g\/(\eta ) we let

Next, for \nu\in {}^{\textstyle\lambda\!>}2 define:

Now check.

Case 3. Otherwise, for each \alpha<\lambda choose F^{\alpha}:{}^{\textstyle\lambda\!>}2\longrightarrow\{0,1\} such that (\forall i<2^{\|\alpha\|})(F_i\leq F^\alpha) (exists by Case 1). Let F:{}^{\textstyle\lambda\!>}2\longrightarrow\{0,1\} be such that (\forall \alpha<\lambda)(F^\alpha\leq F) (exists by Case 2).

The proposition follows.

Let S^*_F be as in 2.19. Since we are assuming \lambda\notin{\rm ID^c}(F^\otimes) necessarily \lambda\setminus S^*_F\notin{\rm ID^b}(F^\otimes).

[Why? because by 2.20(1)(b) we have {\rm ID^c}(F^\otimes)={\rm ID^b}(F^ \otimes )+S^*_F].

It follows from 2.16 and 2.17 that there are disjoint sets X_0, X_1\subseteq\lambda (even disjoint from S^*_F from 2.19) such that X_0,X_1\notin{\rm ID^b}(F^\otimes), X_0\cup X_1\in{\mathfrak B}(F^\otimes) and for every f\in {}^{\textstyle\lambda}2 we have one of the following:

1. (a) the sequence \langle F(f{\restriction}\delta):\delta\in X_0\cup X_1\rangle is {\rm ID^b}(F^\otimes)–almost constant, or

2. (b) both sequences \langle F(f{\restriction}\delta):\delta\in X_0\rangle and \langle F(f{\restriction}\delta): \delta\in X_1\rangle are not {\rm ID^b}(F^\otimes )–almost constant.

It follows from 2.19(2) that we may assume that there is \eta\in {}^{\textstyle X_1}2 such that for every f\in {}^{\textstyle \lambda}2 the set

is stationary. Define a function \rho \in {}^{\textstyle\lambda }2 as follows:

We may assume that the sets \langle Y_i:i<\theta\rangle are pairwise disjoint (otherwise we use Y'_i=Y_i\setminus\bigcup\limits_{j<i} Y_j). Let \eta\in{}^{\textstyle\lambda}\theta be such that (\forall i<\theta)(\eta {\restriction} Y_i=i). Note that if

then we also have

(use F_i\leq F\leq F^\otimes). Consequently, in this case, we have

If this occurs for every i<\theta then

but for each \delta, for some i<\theta we have F(f{\restriction}\delta)=i, a contradiction.

Let

Note that, by 2.12(h),

1. (*) (\forall Y\in{\rm ID^b}(F^\otimes)^+)(\exists Z\in {\mathcal A}_{F^\otimes})(Z\subseteq Y).

Let X\subseteq\lambda, X\notin{\rm ID^b}(F^\otimes) and let \langle Z_\alpha: \alpha<\alpha^*\rangle be a maximal sequence such that for each \alpha< \alpha^*:

Necessarily \alpha^*<\lambda^+, so without loss of generality \alpha^*\leq \lambda, \min(Z_\alpha)>\alpha and Z_\alpha\cap Z_\beta=\emptyset for \alpha<\beta<\alpha^*. Let \langle Z^0_\alpha, Z^1_\alpha\rangle be a partition of Z_\alpha witnessing Z_\alpha\in {\mathcal A}_{F^\otimes}. Put

Then Z_0\cap Z_1=\emptyset, Z_0\cup Z_1\subseteq X. Note that \bigcup\limits_{\alpha<\alpha^*} Z_\alpha is equal to the diagonal union and, by (*) above, X\setminus\bigcup\limits_{\alpha<\alpha^*} Z_\alpha\in {\rm ID^b}(F^\otimes). Consequently we may assume Z_0\cup Z_1= \bigcup\limits_{ \alpha<\alpha^*} Z_\alpha=X. Next, since

we get

and similarly one shows that \min_{F^\otimes}(Z_1)\supseteq X. Now we use 2.12(h) to finish the proof.

1) By induction on \alpha\leq\theta we choose sets X_\eta \subseteq \lambda for \eta\in {}^{\textstyle\alpha}2 such that:

1. (i) X_{\langle\rangle}\notin{\rm ID^b}(F^\otimes),

2. (ii) if \alpha is limit then X_\eta=\bigcap\limits_{i<\alpha} X_{\eta{\restriction} i},

3. (iii) if \alpha=\beta+1, \eta\in {}^{\textstyle\beta}2 and X_\eta\in {\rm ID^b}(F^\otimes) then X_{\eta{}^\frown\!\langle 0\rangle}=X_\eta, X_{\eta{}^\frown\!\langle 1\rangle}=\emptyset;

   if \alpha=\beta+1, \eta\in {}^{\textstyle\beta}2 and X_\eta\notin {\rm ID^b}(F^\otimes) then (X_{\eta{}^\frown\!\langle 0\rangle},X_{\eta {}^\frown\!\langle 1\rangle}) is a partition of X_\eta such that \min_{ F^\otimes}(X_{\eta{}^\frown\!\langle 0\rangle})=\min_{F^\otimes}(X_{\eta{}^\frown\! \langle 1\rangle})=\min_{F^\otimes}(X_\eta).

It follows from 3.4 that we can carry out the construction.

Clearly \langle X_\eta:\eta\in{}^{\textstyle\theta}2\rangle is a partition of X_{\langle\rangle}, so (as 2^\theta<\lambda and {\rm ID^b}(F^\otimes) is \lambda–complete) we can find a sequence \eta\in{}^{\textstyle\theta}2 such that X_\eta\notin{\rm ID^b}(F^\otimes). Then

(as each of these sets includes X_\eta). Moreover, for each \alpha<\theta and for \ell=0,1 we have

Put Y_\alpha=X_{\eta{\restriction}\alpha{}^\frown\!\langle 1-\eta(\alpha)\rangle}. Then \langle Y_\alpha:\alpha<\theta\rangle is a sequence of pairwise disjoint sets (as X_{\eta{\restriction}\alpha{}^\frown\!\langle 0\rangle}\cap X_{\eta{\restriction}\alpha {}^\frown\!\langle 1\rangle}=\emptyset) and for every \alpha<\theta

Hence \bigcap\limits_{\alpha<\theta}\min_{F^\otimes}(Y_\alpha)\notin{\rm ID^b}( F^\otimes). Let Z_\alpha=Y_\alpha\cap\min_{F^\otimes}(X_\eta). Note that \min_{F^\otimes}(Z_\alpha)=\min_{F^\otimes}(X_\eta) (the “\leq” is clear; if \min_{F^\otimes}(Z_\alpha)<\min_{F^\otimes}(X_\eta) then \min_{F^\otimes}(X_\eta)\setminus\min_{F^\otimes}(Z_\alpha) contradicts the definition of \min_{F^\otimes}(Y_\alpha)). Thus the sequence \langle Z_\alpha:\alpha<\theta\rangle is as required. Moreover

2) Let X\subseteq\lambda, X\notin{\rm ID^b}(F^\otimes). By induction on n we choose sets X_n',X''_n such that X'_n\cap X''_n= \emptyset, X'_n\cup X''_n\supseteq X, and

For n=0 we use 3.4 for X to get X'_0,X''_0. For n+1 we use 3.4 for X''_n to get X'_{n+1},X''_{n+1}.

Finally we let Y_n=X''_n (note that \min_{F^\otimes}(Y_n)=\min_{F^\otimes}(X)).

Assume that S^{\bar{A}}_M\notin{{\rm WDmId}_\lambda}.

We can fix a partition \langle D_i: i < \lambda \rangle of \lambda to sets each of cardinality \lambda.

We may assume that |M|=\lambda. By induction on i<\lambda we choose pairs (B_\eta,N_\eta) and sequences \langle C^\eta_j: j<\lambda\rangle for \eta\in {}^{\textstyle i}2 such that

1. (a) \|B_\eta\|<\lambda, N_\eta\in{\mathfrak K}_\lambda, B_\eta\subseteq |N_\eta|\subseteq \cup \{D_j : j < 1 + \ell g\/ (\eta ) \}

2. (b) \langle C^\eta_j:j<\lambda\rangle is increasing continuous, \bigcup\limits_{j<\lambda} C^\eta_j=|N_\eta|, \|C^\eta_j\|<\lambda,

3. (c) if \nu\vartriangleleft\eta then N_\nu\leq_{\mathfrak K} N_\eta and B_\nu\subseteq B_\eta,

4. (d) if j_1,j_2<i then C^{\eta{\restriction} j_1}_{j_2}\subseteq B_\eta,

5. (e) if the pair (B_\eta,N_\eta) does not have the w. amalgmation property in {\mathfrak K}_\lambda then N_{\eta{}^\frown\!\langle 0\rangle}, N_{\eta {}^\frown\!\langle 1\rangle} witness it (i.e. they cannot be w. amalgmated over B_\eta),

6. (f) if i is limit and \eta\in {}^{\textstyle i}2 then B_\eta= \bigcup\limits_{j<i} B_{\eta{\restriction} j}, \bigcup\limits_{j<i}N_{\eta{\restriction} j} \subseteq N_\eta.

There are no problems with carrying out the construction (remember that \leq_{\mathfrak K} is a nice partial order.

Finally, for \eta\in {}^{\textstyle \lambda}2 we let B_\eta=\bigcup\limits_{i<\lambda}B_{\eta{\restriction} i} and N_\eta=\bigcup\limits_{i<\lambda} N_{\eta{\restriction} i}. Clearly, by 4.1(2)(\gamma), we have N_\eta\in {\mathfrak K} and B_\eta\subseteq |N_\eta| for each \eta\in {}^{\textstyle \lambda}2. Moreover,

and thus B_\eta=|N_\eta|. Since {\mathfrak K} is categorical in \lambda, for each \eta\in {}^{\textstyle \lambda} 2 there is an isomorphism f_\eta:N_\eta \stackrel{\rm onto}{\longrightarrow} M.

Fix \eta\in {}^{\textstyle\lambda}2 for a moment.
Let E_\eta=\{\delta<\lambda:f_\eta[B_{\eta{\restriction}\delta}]=A_\delta=\delta\}. Clearly, E_\eta is a club of \lambda. Note that if \delta\in E_\eta then:

We define a colouring

by letting, for f\in{\rm DOM}_\alpha, \alpha<\lambda,

We have assumed S^{\bar{A}}_M\notin{{\rm WDmId}_\lambda}, so there is \rho\in{}^{ \textstyle \lambda}2 such that for each f\in{\rm DOM}_\lambda the set

is stationary and even \not= \emptyset \mbox{ mod } {\rm ID^b}(F^ \otimes ). Let f\in{\rm DOM}_\lambda be defined by f(i)=(\rho(i),f^{ -1}_\rho(i)) (for i<\lambda). Note that

if \alpha\in E_\rho, \rho(\alpha)=0

then \rho is a witness to F(f{\restriction}\alpha)=1 and hence \alpha\notin S_f.

Since S_f is stationary and even \not= \emptyset \mod {\rm ID^b}(F^ \otimes ) and E_\rho is a club of \lambda we may pick \delta\in S_f\cap E_\rho. Then \rho(\delta)=1 and hence F(f{\restriction}\delta)=1, so let \eta_\delta\in {}^{\textstyle \lambda}2 be a witness for it. It follows from the definition of F that then \eta_\delta(\delta)=0, and \eta_\delta{\restriction}\delta=\rho{\restriction}\delta, and f^{-1}_{\eta_\delta}{\restriction}\delta=f^{-1}_\rho{\restriction}\delta. Hence f_{\eta_\eta } {\restriction} B_{\eta_\delta{\restriction}\delta}= f_\rho{\restriction} B_{\rho{\restriction} \delta}, so both have range A_\delta=\delta (and \delta\in E_{ \eta_\delta}\cap E_\rho\cap S^{\bar{A}}_M). But now we get a contradiction with (\boxtimes).