I was meaning to include a proof of Farah’s lemma in my previous post, but then I realized that the slick proof assumes some background which may worth spelling out, first. Therefore, I am dedicating a short post for a self-contained proof of this lemma.
Recall that a Souslin tree is a poset satisfying the following:
- for every , the set is well-ordered by . The set is often considered as the -level of ;
- has no uncountable antichains. In particular, is at most countable for all ;
- has no uncountable chains. In particular, ;
- is a singleton;
- is well-pruned, that is, the set is empty.
(note that clause (2) entails that is countable, so we can always pass to )
Lemma 1. Forcing with a Souslin tree adds an uncountable chain to .
Proof. Let be -generic. For all , must hit the set . Thus, is an uncountable chain.
Lemma 2. Any Souslin tree is isomorphic to , for some .
Proof. For all , fix an injection . Then, for all , let be the unique function that satisfies for all : It is easy to see that iff . Thus, letting , we get that and are order-isomorphic.
For subsets of , let assert that is infinite, while is finite.
Lemma 3. Suppose that is a Souslin tree (under the relation).
Then there exists an injection such that:
- witnesses that is order-isomorphic ;
- for every , and , we have .
Proof. For every limit nonzero ordinal , let be the increasing enumeration of some fixed cofinal subset of , and let be some bijection.
We now define by recursion on the levels.
- Let ;
- Suppose that we are given , where has already been defined. Then, fix a partition of into infinite sets, and let for all such that ;
- Suppose that is nonzero limit ordinal, and has been defined for all . We shall define for all by recursion on the well-ordering of as dictated by . For , we simply find (e.g., via a recursive construction) an injection such that for all , and then put . Then, for , we pick an injection such that for all , and then, put .
Corollary A. Any Souslin tree is isomorphic to , for some , in which for every , the level consists of mutually disjoint sets.
Lemma 4. Any Souslin tree is an -distributive notion of forcing, that is, if is a sequence of cofinal upward-closed subsets of , then .
Proof. For every , let Then is an antichain, and hence countable. Let Then . Pick . We claim that .
To see this, fix an arbitrary . As is cofinal, let us pick some such that . As is the set of minimal elements of , we may also find such that . So and hence either or . Recalling that and the definition of , we conclude that . Recalling that and is downward closed, we have .
Corolloary B. Forcing with a Souslin tree adds no infinite subsets to .
Proof. Suppose that is a -name for a function from to . For all , the set of all -conditions that decides is cofinal and upward-closed. By Lemma 3, then, we find a condition that decides for all , and hence forces that coincides with a ground model function from to .
Theorem (Farah, 1996). If is a Souslin tree, then .
Proof. By Corollary A, we may assume that , where , and each of the levels of consists of mutually disjoint sets. By Lemma 1, let be, in the extension, some uncountable -chain. By passing to , we may assume that is maximal, that is, for all .
To see that , it now suffices to argue that admits no infinite pseudointersection.
Towards a contradiction, suppose that is a pseudointersection for . Put . By Corollary B, the collection lies in the ground model, and so to reach a contradiction, we argue that .
It is clear that . Next, suppose that . Let be such that . Pick and . Then , and hence , which means that and hence .
Pingback: The S-space problem, and the cardinal invariant p | Assaf Rinot
Pingback: More notions of forcing add a Souslin tree | Assaf Rinot