How statement and proof provenance work
The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.
- Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
- AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
- AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.
These labels describe origin, not correctness: citations and verification chips remain separate evidence.
Monotonicity, density, and decision for forcing
Statement
In ZF, for each fixed membership formula and tuple of names:
- If and , then .
- If is dense below p, then .
- is dense in P.
In addition, if is M-generic, , and is dense below p, then . The same assertion applies to .
Facts & Assumptions
Given: ZF, a nonempty forcing preorder P, and fixed names and a fixed formula. The final genericity assertion uses transitive ZF M containing P and its order.
Forcing relation for all formulas defines conjunction, negation and existential forcing.
Atomic forcing relation defines atomic forcing by common-extension and density conditions.
Dense open sets and generic filters over a model defines density and nonempty, upward closed, internally directed generic filters.
Proof
Every atomic forcing clause persists to stronger conditions: in the subset clause the tested common extensions below q are a subset of those below p, and equality is the conjunction of two such requirements; membership restricts its tested extensions in the same way. For density closure of subset, given and , choose forcing the subset, then apply its clause with that entry and common extension a. For equality, take a densely available condition forcing equality and perform this argument separately for each subset direction. For membership, given , first refine to a condition forcing membership and then refine once more to its equality/coefficient witness. These arguments prove atomic persistence and density closure.
If D is dense below p, the set is dense in P. Indeed a condition compatible with p has a common extension, which can be refined into D; an incompatible condition already belongs to the second set. For , Separation makes . A generic G containing p meets , and directedness prevents it from meeting its incompatible part. Thus it meets .
Induct on formula complexity for persistence and density closure. For conjunction, persistence holds for each conjunct by induction; if conjunction is forced densely, each conjunct is forced densely and hence at p by induction. For negation, no extension of p forcing implies the same at every stronger condition. If negation is forced densely below p, a hypothetical forcing has an extension r forcing its negation; persistence of makes r force , contradicting the negation clause at r itself.
For an existential, let . Forcing the existential means W is dense below p. It is then dense below every stronger condition, proving persistence. If conditions below p forcing the existential are dense below p, any has a refinement a below which W is dense, and hence a further refinement in W. Thus W is dense below p, proving density closure. Together with step 2.1, this completes the induction.
Given any p, either some forces , or no such q exists and p forces by definition. This proves external decision density. When the parameters belong to M, Separation inside M instead forms the set decided by ; F1 does not identify that set with the external one for quantified formulas. No condition forces both and , because p is one of its own extensions. All refinements used above are finitely many existential instantiations; no choice principle is used.
Depends on
Used by
- Forcing is not monotone toward weaker conditions Counterexample
- Perfect-tree splitting of a new-real name Example
- Sealing a named maximal antichain Example
- A closed Easton tail adds no short sequences across its chain-condition head Lemma
- A perfect tree of mutually generic name interpretations Lemma
- Dense forcing name translations preserve forcing Lemma
- Every finite ground aleph is countable in the Feferman–Levy model Lemma
- Master-condition characterizations Lemma
- Real names are captured and coded meagre unions are absorbed Lemma
- Restriction, amalgamation, and the set-sized Prikry property Lemma
- Schema of continuity in the basic Cohen model Lemma
- Search-and-shift prime-ideal construction in the basic Cohen model Lemma
- Set-stage names and the forcing truth lemma for the Easton class product Lemma
- Small sets of ground ordinals are captured at a bounded iteration stage Lemma
- The Good-Tree-Watson symmetric model is closed under omega-sequences from the full extension Lemma
- The Lévy collapse localizes countable ordinal data Lemma
- Truth lemma Lemma
- Uniform head-antichain decisions below a class tail Lemma
- A countably closed forcing adds a normal Suslin tree Theorem
- All sets of reals in Solovay L(R) have LM, BP, and PSP Theorem
- Cohen coordinates are distinct and mutually generic Theorem
- Every ultrafilter on every set is principal in Blass's model Theorem
- Every uncountable Solovay-model set of reals has a perfect subset Theorem
- Forcing theorem Theorem
- Generic factorization and ccc preservation for two-step iterations Theorem
- Nice-name reduction and the ccc counting bound Theorem
- Specializing forcing kills a Suslin tree Theorem
- The Prikry property Theorem
Dependency tree · two levels
9 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.