Math. Log. Quart. 62, No. 4–5, 407–411 (2016) / DOI 10.1002/malq.

201400114

The diamond covering property axiom

Janusz Pawlikowski∗
Department of Mathematics, University of Wrocław, pl. Grunwaldzki 2/4, 50-384 Wrocław, Poland

Received 19 December 2014, accepted 31 August 2015

Published online 11 August 2016

The Covering Property Axiom, which attempts to capture some of the combinatorics of the Sacks model, the
model obtained from V  CH by countable support iteration of length ω2 of the Sacks forcing, seems to miss
a Suslin tree. We add a diamond polish to the axiom to remedy this.

C 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Cantor space C = 2ω , endowed with the Vietoris

Let P be the family of all nonempty perfect subsets of the 
topology. For α < ω1 , by an α-cube we mean a set of the form β<α X (β), where X ∈ P α .
game
One of the versions of the Covering Property Axiom of [1], CPAcube , says that Eve has no winning strategy in
the following game:
Adam and Eve play ω1 rounds.
At each round  < ω1 :
Adam chooses a continuous function a from an ω-cube A into 2ω ,
game
(CPAcube )
Eve responds with an ω-cube E  ⊆ A .

Adam wins if  a [E  ] = 2ω .
game
Let ♦CPAcube say that Eve has no winning strategy in the following game:
Adam and Eve play ω1 rounds.
At each round  < ω1 :
Adam chooses a continuous function a from an ω-cube A into 2 ,
game
Eve responds with an ω-cube E  ⊆ A . (♦CPAcube )
Adam wins if for all t ∈ 2ω1 we have that the set { < ω1 : t ∈ a [E  ]} is
stationary.
game game
Easily, ♦ ⇒ ♦CPAcube ⇒ CPAcube , where, recall, ♦ says that there exists a sequence d <ω1 such that for all
t ∈ 2ω1 we have that the set { < ω1 : t = d } is stationary. Using ♦CPAcube , we shall construct a Suslin tree,
game
game
and then we shall prove that ♦CPAcube holds in the Sacks model.
For our Suslin tree construction we need the following simple fact, which is a straightforward generalization
of a theorem of Galvin (cf. [2, 19.6]).
α α
Lemma 1 Let α < ω 1 and suppose that a set B ⊆ C with the Baire property is nonmeager (comeager) in C .
α α
Then the set {C ∈ P : β<α C(β) ⊆ B} is nonmeager (comeager) in P .
game
Theorem 2 ♦CPAcube implies that there is a Suslin tree.
game
P r o o f . (Cf. [3, Theorem 7.8], [4, Theorem 3.1]). Consider the following ♦CPAcube strategy of Eve. During
the play Eve helps herself by building a certain tree T ; at round  she builds the th level T using all ordinals
from the interval [ω, ω( + 1)).

∗ e-mail: [email protected]


C 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
408 J. Pawlikowski: The diamond covering property axiom

(1) If  = 0, then Eve plays anything legal and puts T0 = [0, ω).
(2) If  is a successor ordinal, then Eve plays anything legal and builds T by just splitting every node of T−1
into infinitely many successor nodes.
(3) If  is a limit ordinal such that  = ω, then Eve plays anything legal and builds T from the initial segment
T< = σ < Tσ as follows. She notes that the set [T< ] of all b ⊆ T< that are maximal branches in T< ,
with the topology generated by the basis {b ∈ [T<] : η ∈ b} (for η ∈ T< , is homeomorphic to the Baire
ω
space ωω . Then she picks any sequence s̄ ∈ [T< ] with ran s̄ dense in [T< ], and puts η <T ω + n if
and only if η ∈ s̄ (n) (for n < ω).
(4) If  is a limit ordinal such that  = ω, and if Adam has just played a : A → 2 , then Eve de-
notes by A the a preimage of the set of all χ ∈ 2 = 2ω that are characteristic functions of maximal
antichains of T< . Next she notes that A is Gδ , since its source is Gδ in 2 and a is continuous.
Then:
 ω
Case a. If A is meager, then Eve uses Lemma 1 to find her E  ⊆ C ω \A , picks any s̄ ∈ [T< ] with ran s̄
dense in [T< ], and buids T as in (3).
 ω
Case b. If A is nonmeager, then Eve notes that the set H = {(x, s̄) ∈ A × [T< ] : ∀n(a (x)s̄(n) ≡ 0)}
 ω
is Gδ , and that all its vertical section (H )x , x ∈ A , are dense in [T< ] . (The condition ∀n(a (x)s̄(n) ≡ 0)
above says that every branch pases through one of the elements ω of the maximal antichain coded by a (x).)
Using the Kuratowski-Ulam theorem Eve finds s̄ ∈ [T< ] with ran s̄ dense in [T< ] and such that the s̄ th
horizontal section
(H )s̄ = {x ∈ C ω : (x, s̄ ) ∈ H }
is nonmeager.1 Eve uses this s̄ to build T as in (3), and she uses Lemma 1 to find her E  ⊆ (H )s̄ .
Suppose that Adam defeats the above strategy with a counter-play a <ω1 , so that for any a ⊆ ω1 , denoting
by χa the characteristic function of a, we have that { < ω1 : χa  ∈ a [E  ]} is stationary. Our wanted Suslin
tree is the tree Eve builds during the play in which Adam uses a <ω1 .
We just need to check that maximal antichains are countable. Given any maximal antichain a of T , by the usual
closure argument, { < ω1 : a ∩ ω is a maximal antichain in T< } is a club. So, we can find  = ω > 0 such
that χa  ∈ a [E  ] and a ∩  is a maximal antichain in T< . Then for all n, we have that χa s̄ (n) ≡ 0. Indeed,
pick x ∈ E  with χa  = a (x), and note that (x, s̄ ) ∈ H .
Now, density of ran s̄ entails that a ∩  is a maximal antichain in T<+1 , and it follows that a ∩  is a maximal
antichain in T . Hence a = a ∩ , and thus a is countable. 
game
Next we show that CPAcube holds in the Sacks model. To do this we consider a more general version of
the axiom.
For α < ω1 , let
α be the family of all continuous injections π : C α → C α that preserve projections, i.e., for
all x, x  ∈ C α and all β < α we have that xβ = x  β if equivalent to π (x)β = π (x  )β. Sets of the form ran π
will be calledα-prisms.
Let
= α<ω1
α . If π ∈
, let |π | be the α for which π ∈
α .
The axiom CPAgame prism of [1] says that Eve has no winning strategy in the following game:

Adam and Eve play ω1 rounds.

At each round  < ω1 :
Adam chooses α < ω1 and a continuous function a from an α -prism A (CPAgame
prism )
into 2ω ,
Eve responds with an α -prism E  ⊆ A .

Adam wins if  a [E  ] = 2ω .

 ω  ω
1 Note that the set of s̄ ∈ ωω with ran s̄ dense in ωω is dense Gδ in ωω .


C 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.mlq-journal.org
Math. Log. Quart. 62, No. 4–5 (2016) / www.mlq-journal.org 409

Let ♦CPAgame
prism say that Eve has no winning strategy in the following game:

Adam and Eve play ω1 rounds.

At each round  < ω1 :
Adam chooses α < ω1 and a continuous function a from an α -prism A (♦CPAgame
prism )
into 2 ,
Eve responds with an α -prism E  ⊆ A .
Adam wins if for all t ∈ 2ω1 we have that the set { < ω1 : t ∈ a [E  ]} is
stationary.
game
Clearly, ♦ ⇒ ♦CPAgame prism ⇒ CPAprism . Also, ♦CPAprism ⇒ ♦CPAcube (cf. [1] for a similar implication without ♦).
game game

It is not hard to see that since  = { < ω1 :  = ω} is a club, we get an equivalent axiom if in ♦CPAgame
prism
the players move only when  ∈ , so that in particular we can identify 2 and C  = (2ω ) . Henceforth we shall
assume that  ∈ .
Recall that the Sacks poset is the set P ordered by inclusion. Recall also that the poset of the ξ th stage of
countable support iteration of P is equivalent to the poset Pξ consisting of ordered by inclusion subcylinders of
C ω2 of the form
[ι̊P] := {z̄ ∈ C ω2 : z̄ ◦ ι ∈ P},
where α < ω1 , ι : α → ξ is increasing, and P is an α-prism; note here that if P  ⊆ P is another α-prism, then
[ι̊P  ]  [ι̊P].
Finally, let G ξ and r ξ , ξ < ω2 , be the canonical Pξ term for the generic filter on Pξ and the canonical Pξ +1
term for the ξ th generic real, respectively.
Theorem 3 ♦CPAgame
prism holds in the Sacks model.

P r o o f . We identify a strategy of Eve with a function that maps sequences of moves of Adam—his partial
plays—to answers of Eve. Let ε̇ be a Pω2-name for such a function.
We may assume that in V we have a sequence d ∈ that serves in V as a ♦-sequence predicting subsets of
ω1 × ω1 , i.e., for all  ∈ , we have d ⊆  ×  and, in V , for any d ⊆ ω1 × ω1 , { ∈  : d ∩ ( × ) = d } is
stationary. We may also assume that there is a function εV ∈ V such that Pω2  ε̇V = εV .2 We may make the
above assumptions because:
 
1. the set ξ < ω2 : Pω2  ε̇V [G ξ ] ∈ V [G ξ ] is ω1 -closed and unbounded in ω2 ;
2. for each ξ < ω2 , Pξ forces that Pω2 /Pξ is equivalent to the countable support iteration of length ω2 of
the Sacks forcing;
3. for each limit ξ < ω2 of cofinality ω1 , Pξ  ♦.

In V , we shall define now a counter-play a ∈ of Adam such that

Pω2  a ∈ defeats ε̇.
To this end fix a large enough regular cardinal κ. Let Adam play at each round  ∈  anything legal, unless he
can pick an elementary embedding
n : (, d ) → (Hκ , ∈)
such that, writing N = ran n , we have

(a)  = ω1 ∩ N ,
(b) n (0) = ( p , t˙ (ξ )ξ <ω1 ), where p ∈ Pω2 , and t˙ (ξ )ξ <ω1 is a sequence of Pω2-names for bits (i.e.,
Pω2  t˙ (ξ ) ∈ 2).

ω2 is a Pω2-generic over V and ε := ε̇/G ω2 , then in V [G ω2 ] the ε output to any coded in V input
2 We mean here that whenever G

aσ σ ∈∩ is coded in V ; moreover (ε(aσ σ ∈∩ )) V = εV (aσV σ ∈∩ ).

www.mlq-journal.org 
C 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
410 J. Pawlikowski: The diamond covering property axiom

Then, letting α be the order type of ω2 ∩ N , and ν : α → ω2 ∩ N the order isomorphism, using fusion
(Axiom A and continuous reading of names argument), Adam finds and plays an α -prism A and a continuous
function a : A → 2 such that [ν̊ A ] is (N , Pω2 )-generic  p that forces that t˙ (σ )σ < = a (r ν (β) β<α ).
Note that if n† : (, d ) → (Hκ , ∈) is any other elementary embedding, then, writing N† = ran n† , we have

(a† )  = ω1 ∩ N† ,
(b† ) n† (0) = ( p† , t˙† (ξ )ξ <ω1 ), where p† ∈ Pω2 , and t˙† (ξ )ξ <ω1 is a sequence of Pω2-names for bits such that
t˙† (σ )σ < = (n† ◦ n−1 )(t˙ (σ ))σ < .

Also, α is the order type of ω2 ∩ N† , and letting ν† : α → ω2 ∩ N† be the order isomorphism, we have that
[ν̊† A ] is (N† , Pω2 )-generic  p† that forces that t˙† (σ )σ < = a (r ν† (β) β<α ).
To see that using the above counterplay Adam is forced to win pick any condition p ∈ Pω2 , a Pω2-name Ċ
for a club in ω1 , and a sequence t˙(ξ )ξ <ω1 of Pω2-names for bits. Expand ω1 ∪ {( p, t˙(ξ )ξ <ω1 ), Ċ} to M ≺ Hκ
of size ω1 , and fix a bijection m : ω1 → M with m(0) = ( p, t˙(ξ )ξ <ω1 ). Let  be the club of those  ∈  that
M ∩ ω1 =  and Ċ ∈ M ≺ Hκ , where M = m[].
By ♦ applied to d = {(β, γ ) : m(β) ∈ m(γ )}, get  ∈  with d = d ∩ ( × ). Then m := m witnesses
that at round  Adam was able to find n (possibly = m ). In particular, letting μ : α → ω2 ∩ M be the or-
der isomorphism, [μ̊ A ] is (M , Pω2 )-generic  p, so it forces that  ∈ Ċ by Ċ ∈ N† and  = N† ∩ ω1 . So
[μ̊ E  ], being  [μ̊ A ], also forces that  ∈ Ċ. It moreover forces that t˙(σ )σ < ∈ a [E  ] because it forces that
t˙(σ )σ < = a (r μ (β) β<α ) and r μ (β) β<α ∈ E  . 

We remark the following: Suppose that V ⊆ V are transitive models of a large enough finite fragment of ZFC
game
such that ω1V = ω1V . Consider the modification of ♦CPAcube which says that for any strategy ε ∈ V of Eve such
that εV ∈ V Adam has a defeating counter-play that is coded in V . (Cf. Footnote 2.)
It is not hard to adapt the proof of Theorem 1 to get from the modified axiom a tree in V that is Suslin in V
and to adapt the proof of Theorem 2 to establish the modified axiom in V provided V  ♦ and V is a generic
extension via countable support product of any length of the Sacks poset.
A similar remark applies to ♦CPAgame prism and to countable support iteration of any length of the Sacks poset.
game
The unmodified axiom ♦CPAcube fails if V = V [rξ ξ <ω2 ] is obtained from V via countable support product
game
of length ω2 of the Sacks posets (here, rξ is the ξ th generic real). In fact (cf. [1]), even a weak version of CPAcube
implies that for any set S ⊆ C of size ω2 there is a continuous function f : C → C with f [S] = C. Now just note
that any continuous function f ∈ V is coded in V [rξ ξ ∈ f ] for some countable  f ∈ V . So, if ζ ∈ ω1 \ f , then
rζ , being generic over V [rξ ξ ∈ω2 \{ζ } ], cannot be in { f (rξ )}ξ ∈ω2 \{ζ } . It follows that no continuous function f ∈ V
can map {rξ }ξ ∈ω2 \ω1 onto C.
In a second remark, consider now (cf. [1, Chapter 6]) a modification of ♦CPAgame prism which says that Eve has
no winning strategy even if Adam wins only when there exists s̄ ∈ C ω1 such that for all t ∈ 2ω1 , we have that
{ ∈  : t ∈ a [E  (s̄)]} is stationary, where E  (s̄) = {z̄ ∈ E  : z̄ = s̄}.
To see that this modified axiom holds in the Sacks model just note that in the closing lines of the proof of
Theorem 2 we have μ (β) = β for β < , so that [μ̊ E  ]  r μ (β) β<α ∈ E  (r β β< ).
Finally, let us make a third remark: Sacks forcing serves as a parameter in CPA, analogous axioms can be
formulated for various other forcings (cf. [1], Overview]). Also, CPA can be expressed in terms of Borel functions
and σ -ideals (cf. [5]). Naturally, this applies to the ♦ versions as well. For instance, let I be the σ -ideal of
countable subsets of C. For α < ω1 , let Iα be the αth Fubini power3 of I, and let Bα be the σ -algebra of Borel

3 Writing I ∗ for the dual filter of I, put X ⊆ C α into I iff X ∩ [T ] = ∅ for some I ∗-splitting continuous tree T ⊂ C <α . Here, a tree T
α
is I ∗-splitting if each node splits into I ∗-many immediate successors, it is continuous if for limit λ < α, τ ∈ T iff τ ∈ C λ and all proper initial
segments of τ are in T , and [T ] is the set of all τ ∈ C α such that all proper initial segments of τ are in T . E.g., I0 = {∅}, since C 0 = {∅},
C <0 = ∅, B0 = {∅, {∅}}, and T = ∅ is the unique tree contained in C <0 , this tree happens to be I ∗-splitting and continuous, and [T ] = {∅}.


C 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.mlq-journal.org
Math. Log. Quart. 62, No. 4–5 (2016) / www.mlq-journal.org 411

subsets of C α . Now, ♦CPAgame

prism can be expressed by saying that Eve has no winning strategy in the following
game:
Adam and Eve play ω1 rounds.
At each round  < ω1 :
Adam chooses α < ω1 , A ∈ Bα \Iα , and a Borel function a : Aα → 2 , (♦CPAgame
prism )
Eve responds with E  ∈ Bα \Iα , E  ⊆ A .
Adam wins if for all t ∈ 2ω1 we have that the set { < ω1 : t ∈ a [E  ]} is stationary.

That the idealized game is equivalent to the original one follows from the Suslin theorem for Iα (cf. [5]), which
claims that for X ∈ Bα we have X ∈ / Iα iff there exists an α-prism P such that P ⊆ X .
When we change in the idealized game the Cantor space to the Baire space ωω , and let I be the σ -ideal
generated by compact subsets ωω , we get an axiom that holds in the Miller model—the model obtained from
V  CH by countable support iteration of length ω2 of the Miller forcing. This version also implies that there
exists a Suslin tree. The proof is virtually the same, except the Kuratowski-Ulam theorem is replaced by its analog
for the Fubini product of the relevant αth Fubini power Iα and the σ -ideal of meager subsets of C.

Acknowledgements The author was partially supported by MNiSW grant N N201 418939.

References
[1] K. Ciesielski and J. Pawlikowski, The Covering Property Axiom CPA, Cambridge Tracts in Mathematics Vol. 164
(Cambridge University Press, 2004).
[2] A. S. Kechris, Classical Descriptive Set Theory, Graduate Texts in Mathematics Vol. 156 (Springer-Verlag, 1995).
[3] K. Kunen, Set Theory, Studies in Logic and the Foundations of Mathematics Vol. 102 (North Holland, 1983).
[4] J. T. Moore, M. Hrušák, and M. Džamonja, Parametrized ♦ principles, Trans. Amer. Math. Soc. 356, 2281–2306 (2004).
[5] J. Zapletal, Forcing Idealized, Cambridge Tracts in Mathematics Vol. 174 (Cambridge University Press, 2008).

www.mlq-journal.org 
C 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim