Professional Documents
Culture Documents
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 1
MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
2
Aasraxcr. We show that it is consistent that the Borel Conjecture and the dual Borel Conjecture hold simulta-
neously.
Ixraootcrrox 3
History. A set X of reals
1
is called strong measure zero (smz), if for all functions f : there are 4
intervals I
n
of measure 1/ f (n) covering X. Obviously, a smz set is a null set (i.e., has Lebesgue measure 5
zero), and it is easy to see that the family of smz sets forms a -ideal and that perfect sets (and therefore 6
uncountable Borel or analytic sets) are not smz.
7
At the beginning of the 20th century, Borel [Bor19, p. 123] conjectured:
Every smz set is countable.
This statement is known as the Borel Conjecture (BC). In the 1970s it was proved that BC is independent, 8
i.e., neither provable nor refutable.
9
Let us very briey comment on the notion of independence: A sentence is called independent of a set 10
T of axioms, if neither nor follows from T. (As a trivial example, (x)(y)x y = y x is independent 11
fromthe group axioms.) The set theoretic (rst order) axiomsystemZFC(Zermelo Fraenkel with the axiom 12
of choice) is considered to be the standard axiomatization of all of mathematics: A mathematical proof is 13
generally accepted as valid i it can be formalized in ZFC. Therefore we just say is independent if 14
is independent of ZFC. Several mathematical statements are independent, the earliest and most prominent 15
example is Hilberts rst problem, the Continuum Hypothesis (CH).
16
BC is independent as well: Sierpi nski [Sie28] showed that CH implies BC (and, since G odel showed 17
the consistency of CH, this gives us the consistency of BC). Using the method of forcing, Laver [Lav76] 18
showed that BC is consistent. 19
Galvin, Mycielski and Solovay [GMS73] proved the following conjecture of Prikry:
X 2
as the set of reals. ( = 0, 1, 2, . . ..) By well-known results both the denition and the theorem also
works for the unit interval [0, 1] or the torus R/Z. Occasionally we also write x is a real for x
.
1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
2 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
etc. See [BJ95, Ch. 8] for several of these results. In this paper, we prove the consistency (and therefore 1
independence) of BC+dBC (i.e., consistently BC and dBC hold simultaneously).
2
The problem. The obvious rst attempt to force BC+dBC is to somehow combine Lavers and Carlsons 3
constructions. However, there are strong obstacles:
4
Lavers construction is a countable support iteration of Laver forcing. The crucial points are:
5
Adding Laver real makes every old uncountable set X non-smz.
6
And this set X remains non-smz after another forcing P, provided that P has the Laver property.
7
So we can start with CH and use a countable support iteration of Laver forcing of length
2
. In the nal 8
model, every set X of reals of size
1
already appeared at some stage <
2
of the iteration; the next Laver 9
real makes X non-smz, and the rest of the iteration (as it is a countable support iteration of proper forcings 10
with the Laver property) has the Laver property, and therefore X is still non-smz in the nal model.
11
Carlsons construction on the other hand adds
2
many Cohen reals in a nite support iteration (or 12
equivalently: nite support product). The crucial points are:
13
A Cohen real makes every old uncountable set X non-sm.
14
And this set X remains non-sm after another forcing P, provided that P has precaliber
1
.
15
So we can start with CH, and use more or less the same argument as above: Assume that X appears at 16
<
2
. Then the next Cohen makes X non-sm. It is enough to show that X remains non-sm at all 17
subsequent stages <
2
. This is guaranteed by the fact that a nite support iteration of Cohen reals of 18
size
1
has precaliber
1
.
19
So it is unclear how to combine the two proofs: A Cohen real makes all old sets smz, and it is easy 20
to see that whenever we add Cohen reals conally often in an iteration of length, say,
2
, all sets of any 21
intermediate extension will be smz, thus violating BC. So we have to avoid Cohen reals,
2
which also 22
implies that we cannot use nite support limits in our iterations. So we have a problem even if we nd a 23
replacement for Cohen forcing in Carlsons proof that makes all old uncountable sets X non-sm and that 24
does not add Cohen reals: Since we cannot use nite support, it seems hopeless to get precaliber
1
, an 25
essential requirement to keep X non-sm.
26
Note that it is the proofs of BC and dBC that are seemingly irreconcilable; this is not clear for the 27
models. Of course Carlsons model, i.e., the Cohen model, cannot satisfy BC, but it is not clear whether 28
maybe already the Laver model could satisfy dBC. (It is even still open whether a single Laver forcing 29
makes every old uncountable set non-sm.) Actually, Bartoszy nski and Shelah [BS03] proved that the Laver 30
model does satisfy the following weaker variant of dBC (note that the continuum has size
2
in the Laver 31
model):
32
Every sm set has size less than the continuum.
33
In any case, it turns out that one can reconcile Lavers and Carlsons proof, by mixing them generi- 34
cally, resulting in the following theorem:
35
Theorem. If ZFC is consistent, then ZFC+BC+dBC is consistent.
36
We give a rather informal overview of the proof in Section 1.
37
Prerequisites. To understand anything of this paper, the reader
38
should have some experience with nite and countable support iteration, proper forcing,
2
-cc, 39
-closed, etc.
40
should know what a quotient forcing is,
41
should have seen some preservation theorem for proper countable support iteration,
42
should have seen some tree forcings (such as Laver forcing).
43
To understand everything, additionally the following is required:
44
The case A preservation theoremfrom[She98], more specically we build on the proof of [Gol93] 45
(or [GK06]).
46
In particular, some familiarity with the property preservation of randoms is recommended. We 47
will use the fact that random and Laver forcing have this property.
1
2
An iteration that forces dBC without adding Cohen reals was given in [BS10], using non-Cohen oracle-cc.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 3
We make some claims about (a rather special case of) ord-transitive models in Section 4.A. The 2
reader can either believe these claims, or check them himself (by some rather straightforward 3
proofs), or look up the proofs (of more general settings) in [She04] or [Kel].
4
From the theory of strong measure zero and strongly meager, we only need the following two results 5
(which are essential for our proofs of BC and dBC, respectively):
6
Pawlikowskis result from [Paw96a] (which we quote as Theorem 1.2), and
7
Theorem 8 of Bartoszy nski and Shelahs [BS10] (which we quote as Lemma 3.2).
8
We do not need any other results of Bartoszy nski and Shelahs paper [BS10]; in particular we do not use the 9
notion of non-Cohen oracle-cc. (Note that the third author claims that our construction is more or less the 10
same as a non-Cohen oracle-cc construction introduced in [She06], and that the extended version presented 11
in [She10] is even closer to our preparatory forcing.)
12
The reader does not have to know the original proofs of Con(BC) and Con(dBC), by Laver and Carlson, 13
respectively.
14
Acknowledgment. We thank Tomek Bartoszy nski for pointing out Pawlikowskis result [Paw96a] to us, 15
and for many interesting discussions on the topic.
16
Notation. Stronger conditions in forcing notions are smaller, i.e., q p means that q is stronger than p.
17
Let P Qbe forcing notions. (As usual, we abuse notation by not distinguishing between the underlying 18
set and the quasiorder on it.)
19
For p
1
, p
2
P we write p
1
P
p
2
for p
1
and p
2
are incompatible. Otherwise we write p
1
j p
2
.
20
q
31
such that
[G] = [G
M
] for all Q-generic lters G. We will usually just identify and
.
32
Analogously, if P M and i : P Q is a function, then i is called an M-complete embedding if it 33
preserves (or at least
2
forces BC+dBC, i.e., that an uncount- 44
able X is neither smz nor sm. We show this under the assumption X V, and then introduce a 45
factorization of R
P that this assumption does not result in loss of generality.
1
3
We will also use so-called ord-transitive models, as dened in Section 4.A.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
4 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
1. Tnr ovravrrw or rnr raoor 2
In this section, we give a rather informal overview of (and background for) the proof. The emphasis is 3
on giving the reader some vague understanding (i.e., a warm fuzzy feeling), at the expense of correctness 4
of the claims (we point out some of the most blatant lies).
5
1.A. The general setup. We assume CH in the ground model. We use a -closed
2
-cc preparatory 6
forcing R, which adds a generic alternating iteration (as dened below)
P = (P
, Q
)
<
2
. Moreover, R 7
forces that
P is ccc. The forcing notion to get BC+dBC is the composition R P
2
.
8
We say that
P is an alternating iteration if
P = (P
, Q
)
<
2
is a forcing iteration of length
2
satisfy- 9
ing the following:
10
At every even step , (P
forces that) Q
forces that) Q
= .
14
At a limit step , we take partial countable support limits. (This means more or less: P
is a 15
subset of the countable support limit of (P
)
<
and contains
_
<
P
set H we can construct (in an absolute way) a closed null set F such that for 26
every X 2
with t + X H.
27
So we can actually show the following: 28
(1.3)
We can construct from the ultralaver real in an absolute way a (code for a) closed
null set F such that X +F is (outer Lebesgue measure) positive for every uncountable
ground model set X.
It is an easy exercise to show that Theorem 1.2 implies the following fact.
29
Fact 1.4. Assume that
P = (P
, Q
: <
2
) is an iteration with direct limit P
2
satisfying the following:
30
For conally many <
2
, Q
2
and even all quotients P
2
/P
2
preserves
1
and satises the
2
-cc.
33
Then P
2
forces BC.
34
Remark 1.5. It is well-known that both Laver reals and randomreals preserve positivity (see Lemma 2.28). 35
As Laver forcing makes every old uncountable set non-smz, we conclude that a countable support iteration 36
of length
2
of Laver reals, or alternatively, a countable support iteration alternating Laver with random 37
reals, forces BC. The latter iteration also forces the failure of dBC, since the random reals increase the 38
covering number of the null ideal, and every set smaller than this cardinal is sm.
1
4
This is a lie, and moreover a stupid (i.e., useless) lie. It is a lie, since we only get something like: for one random over a specic
model, we can nd a system
D such that L
D
preserves randomness. It is a useless lie, since preservation of positivity is not enough
anyway: We need a stronger property that is preserved under proper countable support iterations.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 5
1.C. A preparatory forcing for a single step. Let us rst describe how to generically create a single 2
forcing, e.g., an ultralaver forcing.
3
Let Q be a forcing, M a countable transitive model, P M a subforcing of Q. We say that P is an 4
M-complete subforcing of Q, if every maximal antichain A M of P is also a maximal antichain in Q. In 5
this case every Q-generic lter over V induces a P-generic lter over M.
6
Let M
x
be a (countable) model, and
D
x
(D
x
s
)
s
< a system of ultralters in M
x
. This denes the 7
ultralaver forcing Q
x
= L
D
x in M
x
. Given any system
D of ultralters (in V) such that each D
s
extends 8
D
x
s
, then we can show that Q
x
is an M
x
-complete subforcing of the ultralaver forcing Q L
D
in V. We 9
describe this by
5
(M
x
, Q
x
) canonically embeds into Q.
10
So every Q-generic lter H over V induces a Q
x
-generic lter over M
x
which we call H
x
. A trivial but 11
crucial observation is the following: When we evaluate the ultralaver real for Q in V[H] then we get the 12
same real as when we evaluate it for Q
x
in M
x
[H
x
]. Of course Q
x
is not a complete subforcing of Q, just 13
an M
x
-complete one: While Q
x
is just countable, therefore equivalent to Cohen forcing (from the point 14
of view of V), the real added by the Q
x
-generic H
x
is an ultralaver real (over M
x
as well as over V), and 15
therefore does not add a Cohen real over V. 16
We can dene a preparatory forcing R
L
(for a single ultralaver forcing) consisting of pairs x = (M
x
, Q
x
) 17
as above (M in some H(
) containing (M
x
, Q
x
) and Q, and Mostowski-collapse (N, Q) to y = (M
y
, Q
y
). 30
Then y is in R
L
and stronger than x.
31
1.D. Janus forcing. With Janus forcings we denote a family of forcing notions (as in the case of ul- 32
tralaver forcings). Every Janus forcing J is a subset of H(
1
) and has a countable core (which is the 33
same for every Janus forcing) and some additional stung.
6
The forcing will add a generic real (Janus 34
real) coding a null set Z
. The forcing will not be a complete subforcing of J, but we will require that all 35
maximal antichains involved in the name Z
.
36
The crucial combinatorial content of Janus forcings heavily relies on previous work by Bartoszy nski 37
and Shelah [BS10].
38
Analogously to the case of ultralaver forcing, let R
J
consist of pairs x = (M
x
, Q
x
) such that M
x
is a 39
countable model and Q
x
a Janus forcing in M
x
. Given x R
J
and a Janus forcing Q in V, we say that x 40
canonically embeds into Q if Q
x
is an M
x
-complete subforcing of Q, and we set y x in R
J
if M
y
thinks 41
that x canonically embeds into Q
y
. Again R
J
is a -closed forcing and adds a generic object Q that is 42
(forced to be) a Janus forcing; and for every x in the R
J
-generic lter, x canonically embeds into Q.
43
As in the case of ultralaver forcing, the Janus real Z
is absolute.
44
Other than in the case of ultralaver forcing, every Janus forcing Q
x
in any model M
x
is itself a Janus 45
forcing in V; so (taking the collapse of an elementary submodel as above) we trivially get: 1
(1.6)
For every x R
J
there is a y x such that in M
y
, Q
y
is a countable Janus forcing (so
in particular equivalent to Cohen forcing).
5
Note the linguistic asymmetry here: A symmetric and more verbose variant would say (M
x
, Q
x
) canonically embeds into
(V, Q).
6
Actually, the denition of Janus forcing additionally depends on a real parameter. In our application, we will use ultralaver
forcings as even stages , and use a Janus forcing dened from the ultralaver real in the stage +1. The following claims about Janus
forcings only hold for this situation; in particular the ground-model sets mentioned have to live in the model before the ultralaver
forcing.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
6 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
2
R
P
y
M
y
P
x
M
x
M
x
-complete
M
y
-complete
y < x in G
x G
Frotar 1. G is the generic lter for the preparatory forcing notion R, which adds the
generic iteration
P. The forcing that gives BC+dBC is R P
2
. (Of course, in contrast
to the impression given by the diagram, the set M
x
2
, which is the domain of
P
x
, is
not an interval.)
A crucial property of Janus forcing is that we can make it into random forcing as well: 2
(1.7)
For every x R
J
there is a y x such that in M
y
, Q
y
is forcing equivalent to random
forcing.
So here we see an important property of the preparatory forcing R
J
, which might seem a bit paradoxical 3
at rst: Densely, Q seems to be Cohen as well as random forcing. This two-faced behavior gives Janus 4
forcing its name; one could also describe this behavior as faking (faking to be Cohen and faking to be 5
random).
6
Janus forcing is the forcing notion that replaces the Cohen real in the dBC part of the proof. The crucial
point is:
A countable Janus forcing makes every uncountable ground model set of reals non-
sm.
Well, that is actually not much of a point at all: As Carlson has shown, this is achieved by a Cohen real, 7
and obviously a countable Janus forcing is equivalent to a Cohen real. And Carlson even showed: When 8
adding a Cohen real, this uncountable ground model set remains non-sm even after forcing with another 9
forcing notion, provided this forcing notion has precaliber
1
.
10
So what we actually claim for Janus forcing is the more explicit version of our trivial (after Carlson) 11
claim above. First, let us introduce an obvious notation: 12
(1.8)
Let Z be a null set. We say that Z witnesses that X is not sm if there is no real t with
(X + t) Z = , or equivalently, if X + Z = 2
.
(Clearly X is not sm i there is a witness Z.)
So what we really claim is the following: 13
(1.9)
The canonical null set Z
= 2
= 2
is
preserved by every subsequent -centered forcing.
Of course Z
is interpreted as a code for a null set, not a concrete subset of the reals (otherwise X+Z
= 2
14
could not hold when we add new reals).
15
So the point here is that we can construct the null set Z
)) and
P
x
is in M
x
an alternating 2
iteration. 3
Assume that x Rand that
P is (in V) an alternating iteration. Other than in the case of a single ultralaver 4
forcing, we nowcannot formally assume that
P
x
is a subset of
P, but there is a natural construction that tries 5
to give an M
x
-complete embedding of
P
x
into
P. If this construction works, we say that x is canonically 6
embeddable into
P, and in that case we can treat
P
x
as subset of
P. So if x is canonically embeddable 7
into
P, and H is a P
2
-generic lter over V (which of course induces P
-generic lters H
for all
2
), 8
we can get a canonical P
x
).
9
We dene the order in R as above: For x, y R, we dene y to be stronger than x, if M
x
M
y
and M
y
10
thinks that x canonically embeds into
P
y
.
11
Note that while M
x
thinks that
P
x
is an iteration of length
2
, in V (or in M
y
for y x) the real domain 12
of
P
x
is just countable (since it is a subset of M
x
).
13
As promised, one can show that R is -closed and adds a generic alternating iteration
P, and that
P is 14
ccc. The nal limit P
2
is the direct limit of the P
(and thus does not add any new reals in the last stage). 15
The intermediate stages satisfy CH, while P
2
forces 2
0
=
2
. As might be expected by now, each x in 16
the R-generic lter G canonically embeds into
P. We will call the R-generic lter G. (The situation is 17
illustrated in Figure 1.) So if H is P
2
-generic over V[G], then we get canonical P
x
2
-generic lters H
x
18
for all x G; and the real ultralaver (and Janus) reals calculated in V[G][H] are the same as the ones 19
locally calculated in M
x
[H
x
].
20
Given any x R we can construct (in V) an alternating iteration
P such that x embeds into
P and such 21
that
P has either of the following two properties:
22
All Janus forcings are countable, at all stages not in M
x
we do nothing, and all limits P
are 23
almost nite support over x (basically the limit is nite support, but we more or less add the 24
countably many elements of P
x
).
25
All Janus forcings are equivalent to randomforcing, and all limits P
is -centered: 31
We iterate only countably many forcings, since we do nothing outside M
x
; the single forcings are - 32
centered (in the ultralaver case) and even countable in the Janus case, and the (almost) nite support limits 33
preserve -centeredness.
34
In the second case, we get preservation of positivity (with respect to outer Lebesgue measure): ultralaver 35
as well as random forcings preserve positivity, and preservation is preserved by (almost) countable support 36
(proper) iterations.
8
37
As above, we put the iteration
P into a countable elementary submodel; collapse it, and thus get: 38
(1.10)
For all x R there is a y x such that (M
y
thinks that) P
y
2
is -centered and all
Janus forcings are countable.
39
(1.11)
For all x R there is a y x such that (M
y
thinks that) P
y
2
preserves Lebesgue outer
measure positivity.
Let us again note that densely often we use nite support, but we also use countable support densely 40
often. 41
1.F. Why BC holds. We want to show that BC is forced by R P
2
. Let X be the name of a set of reals of 42
size
1
. Since P
2
has length
2
, we can assume
9
that X is in the ground model V. We want to show BC, 43
so we have to show that X is not smz. The following is illustrated by Figure 2.
1
7
Since we are interested in iterations of length
2
, we cannot use transitive models, that can only see ordinals <
1
. Instead, we
use ord-transitive models.
8
Of course, this is not true, rather we need an iterable property such as preservation of random reals over models, etc. We do not
get this stronger property universally, we can just preserve a specic random; so claim (1.11) is a lie, too.
9
Well, we cant. But we can do something similar, as will be explained in section 6.B.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
8 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
+ 1
2
X
y
+ F pos.
F appears at + 1 Assume
R
P
X + F Z
(3) pick x s.t.:
(4) pick y x s.t:
Z is P
x
-name
(5) contradiction!
so: X
y
+ F pos.
(c) P
y
pres. pos.
(X
y
X,
F, Z abs.)
x
y
R
P
x
P
y
P
(1) (2)
(b) (d)
Frotar 2. The proof of BC.
(1) Fix any ultralaver position . (Well, we x large enough to justify our assumption that X V.) 2
We know that the ultralaver real that is added by Q
2
5
(where the Borel code F is evaluated in the extension). So assume towards a contradiction that 6
X + F is forced to be a subset of a null set (or rather, a Borel code) Z; this already has to happen
10
7
at some stage <
2
. In other words: We assume (towards a contradiction) 8
X + F Z.
(3) Since P
is (forced to be) ccc, we can nd a very absolute (countable) name for Z; and we can 9
nd an x R that already calculates Z correctly.
11
10
(4) Now we construct (in V) a y x in R (with
P
y
forces that X
y
+ F is positive (according to (1.3)).
15
(c) P
y
2
preserves positivity.
16
(d) Therefore P
y
forces that X
y
+ F is positive
12
(and in particular not a subset of Z).
17
(5) This leads to the obvious contradiction: Let G be R-generic over V and contain y, and let H
be 18
P
is P
y
-generic over M
y
, and therefore M
y
[H
y
.
1
10
Each real (and in particular the Borel code Z) in the P
2
-extension already has to appear at some stage <
2
; and the statement
X + F Z is absolute.
11
More formally: We nd an x R and a P
x
-name Z
x
in M
x
such that x forces (in R) that P
-generic).
12
Here we even get positivity of X
y
+ F where F is evaluated in the intermediate extension of stage + 1. However, we get the
contradiction even if we just assume that X
y
+ F is positive where F is evaluated in the P
y
-extension.
13
And again, this is a lie.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 9
(2) According to (1.8), it is enough to show that Z
+ X = 2
in the extension by R P
2
. So assume 2
towards a contradiction that Z
+ X 2
R
P
r Z
+ X.
(3) Again, nd x such that r is an absolute P
x
-name.
5
(4) Now we construct (in V) a y x in R that satises (1.10), and such that X
y
X M
y
M
y
is 6
uncountable in M
y
. In particular, (M
y
thinks that)
7
(a) P
y
2
is -centered.
10
(d) Therefore (1.9) implies that P
y
forces that Z
+ X
y
= 2
, in particular that r Z
+ X
y
.
11
(5) As before, this leads to a contradiction.
12
2. Uiraxixvra roacrxo 13
In this section, we dene the family of ultralaver forcings L
D
, variants of Laver forcing which depend 14
on a system
D of ultralters. We will need the following properties of this forcing notion:
15
Wish List 2.1. (1) L
D
is -centered, hence ccc. (This will follow trivially from the denition.)
16
(2) Ultralaver kills smz: There is a canonical L
D
-name
called the 17
ultralaver real. From this real, we can dene a closed null set F such that X + F is positive for all 18
uncountable X in V (and therefore F witnesses that X is not smz, according to Theorem 1.2). (See 19
Corollary 2.20.)
20
(3) Canonical above in particular means: If M is a countable model, L
D
M M is an ultralaver 21
forcing (with ultralaver real
M
) which M-completely embeds into the ultralaver forcing L
D
(with 22
ultralaver real
), then in the L
D
-extension V[H],
M
[H
M
], where H
M
is the 23
induced lter on L
D
M , similarly for F.
24
(4) Whenever X is uncountable, then L
D
forces that X is not thin. (See Corollary 2.23.)
25
(5) If (M, ) is a countable model of ZFC* and if L
D
M is an ultralaver forcing in M (with ultralaver 26
real
M
), then for any ultralter system
D extending
D
M
, L
D
M is an M-complete subforcing of the 27
ultralaver forcing L
D
(which adds the same ultralaver real). (See Lemma 2.4.)
28
(6) Moreover, given M and L
D
M as above, and a random real r over M, we can choose
D extending 29
D
M
such that L
D
forces that randomness of r is preserved (in a strong way that can be preserved in 30
a countable support iteration). (See Lemma 2.29.)
31
(Advice to the reader: At rst reading, you may want to take these properties as granted, and skip to 32
Section 3.)
33
2.A. Denition of ultralaver. 34
Notation. We use the following fairly standard notation:
35
A tree is a nonempty set p
<
which is closed under initial segments and has no maximal elements.
14
36
The elements (nodes) of a tree are partially ordered by .
37
For each sequence s
<
we write lh(s) for the length of s.
38
For any tree p
<
and any s p we write succ
p
(s) for one of the following two sets: 39
k : s
k p or t p : (k ) t = s
k
and we rely on the context to help the reader decide which set we mean.
40
A branch of p is either of the following:
41
A function f : with f n p for all n .
42
A maximal chain in the partial order (p, ). (As our trees do not have maximal elements, each 43
such chain C determines a branch
_
C in the rst sense, and conversely.)
1
14
Except for the proof of Lemma 2.4, where we also allow trees with maximal elements, and even empty trees.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
10 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
We write [p] for the set of all branches of p.
2
For any tree p
<
and any s p we write p
[s]
for the set t p : t s or t s, and we write [s] for 3
either of the following sets: 4
t p : s t or x [p] : s x.
The stem of a tree p is the shortest s p with succ
p
(s) > 1. (The trees we consider will never be 5
branches, i.e., will always have nite stems.)
6
Denition 2.2. For trees q, p we write q p if q p (q is stronger than p), and we say that q 7
is a pure extension of p (q
0
p) if q p and stem(q) = stem(p).
8
A lter system
D is a family (D
s
)
s
< of lters on . (All our lters will contain the Fr echet lter 9
of conite sets.) We write D
+
s
for the collection of D
s
-positive sets (i.e., sets whose complement is 10
not in D
s
).
11
We dene L
D
as the set of all trees p such that succ
p
(t) D
+
t
for all t p above the stem.
12
The generic lter is determined by the generic branch
= (
i
)
i
=
_
pG
[p] or equivalently,
=
_
pG
stem(p).
14
An ultralter system is a lter system consisting of ultralters. (Since all our lters contain the 15
Fr echet lter, we only consider nonprincipal ultralters.)
16
An ultralaver forcing is a forcing L
D
dened from an ultralter system. The generic real for an 17
ultralaver forcing is also called the ultralaver real.
18
Recall that a forcing notion (P, ) is -centered i P =
_
n
P
n
, where for all n, k and for all 19
p
1
, . . . , p
k
P
n
there is q p
1
, . . . , p
k
.
20
Note that all ultralaver forcings L
D
are -centered (fullling our rst wish 2.1(1)), as nite set of con- 21
ditions with the same stem has a common lower bound. If each D
s
is the Fr echet lter, then L
D
is Laver 22
forcing (often just written L).
23
2.B. M-complete embeddings. Note that for all ultralter systems
D we have: 24
(2.3)
Two conditions in L
D
are compatible if and only if their stems are comparable and
moreover, the longer stem is an element of the condition with the shorter stem.
Lemma 2.4. Let M be countable.
15
In M, let L
D
M be an ultralaver forcing. Let
D be (in V) a lter system 25
extending
16
D
M
. Then L
D
M is an M-complete subforcing of L
D
.
26
Proof. For any tree
17
T, any lter system
E = (E
s
)
s
<, and any s
0
T we dene a sequence (T
E,s
0
)
1
27
of derivatives (where we may abbreviate T
E,s
0
to T
) as follows:
28
T
0
T
[s
0
]
. 29
Given T
, we let T
+1
T
\
_
[s] : s T
, s
0
s, succ
T
(s) E
+
s
, where [s] t : s t.
30
For limit ordinals > 0 we let T
_
<
T
.
31
Then we have 32
(a) Each T
.
33
(b) There is an
0
<
1
such that T
0
= T
0
+1
= T
or T
E,s
0
for T
0
.
34
(c) If s
0
T
E,s
0
, then T
E,s
0
L
E
with stem s
0
. 35
Conversely, if stem(T) = s
0
, and T L
E
, then T
= T.
36
(d) If T contains a tree q L
E
with stem(q) = s
0
, then T
contains q
= q, so in particular s
0
T
.
37
(e) Thus: T contains a condition in L
E
with stem s
0
i s
0
T
E,s
0
.
38
(f) The computation of T
s
0
)
+
. (In particular, this will be true for any
E
3
extending
E, provided that each E
s
0
is an M-ultralter.) 4
Then for each M we have T
E,s
0
= T
,s
0
(and hence T
E,s
0
M). (Proved by induction on .)
5
Now let A = (p
i
: i I) M be a maximal antichain in L
D
M , and assume (in V) that q L
D
. Let 6
s
0
stem(q).
7
We will show that q is compatible with some p
i
. This is clear if there is some i with s
0
p
i
and 8
stem(p
i
) s
0
, by (2.3). (In this case, p
i
q is a condition in L
D
with stem s
0
.)
9
So for the rest of the proof we assume that this is not the case, i.e.: 10
(2.5) There is no i with s
0
p
i
and stem(p
i
) s
0
.
Let J i I : s
0
stem(p
i
). We claim that there is j J with stem(p
j
) q (which as above implies 11
that q and p
j
are compatible).
12
Assume towards a contradiction that this is not the case. Then q is contained in the following tree T:
T (
<
)
[s
0
]
\
_
jJ
[stem(p
j
)] (2.6)
Note that T M. In V we have: 13
(2.7) The tree T contains a condition q with stem s
0
.
so by (e) (applied in V), followed by (g), and again by (e) (now in M) we get: 14
(2.8) The tree T also contains a condition p M with stem s
0
.
Now p has to be compatible with some p
i
. The sequences s
0
= stem(p) and stem(p
i
) have to be comparable, 15
so by (2.3) there are two possibilities:
16
(1) stem(p
i
) stem(p) = s
0
p
i
. We have excluded this case in our assumption (2.5).
17
(2) s
0
= stem(p) stem(p
i
) p. So i J. By construction of T (see (2.6)), we conclude stem(p
i
) T, 18
contradicting stem(p
i
) p T (see 2.8).
19
2.C. Ultralaver kills strong measure zero. The proof of the following lemma is a well-known routine 20
construction that works with many tree forcings. We will delay the proof until Lemma 2.34.
21
Lemma 2.9. If A is a nite set,
an L
D
-name, p L
D
, and p
= .
23
Denition 2.10. Let
be an increasing sequence of natural numbers. We say that X 2
is smz with 24
respect to
, if there exists a sequence (I
k
)
k
of basic intervals of 2
of measure 2
k
(i.e., each I
k
is of 25
the form [s
k
] for some s
k
2
k
) such that X
_
m
_
km
I
k
.
26
Remark 2.11. It is well known and easy to see that the properties
27
For all
there exists exists a sequence (I
k
)
k
of basic intervals of 2
of measure 2
k
such that 28
X
_
k
I
k
.
29
For all
there exists exists a sequence (I
k
)
k
of basic intervals of 2
of measure 2
k
such that 30
X
_
m
_
km
I
k
.
31
are equivalent. Hence, a set X is smz i X is smz with respect to all
.
32
The following lemma is a variant of the corresponding lemma (and proof) for Laver forcing (see for 33
example [Jec03, Lemma 28.20]): Ultralaver makes old uncountable sets non-smz.
34
Lemma 2.12. Let
D be a system of ultralters, and let
be the L
D
-name for the ultralaver real. Then each 35
uncountable set X V is forced to be non-smz (witnessed by the ultralaver real
).
36
More precisely, the following holds:
37
(2.13)
L
D
X V [2
1
(x
k
)
k
2
X
_
m
_
km
[x
k
k
].
We rst give two technical lemmas:
1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
12 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
Lemma 2.14. Let p L
D
with stem s
<
, and let
x be a L
D
-name for a real in 2
i]
xn = n D
s
.
Proof. For each i succ
p
(s), let q
i
0
p
[s
i]
be such that q
i
decides
xi = t
i
(this is possible by Lemma 2.9).
5
Now we dene the real 2
as the D
s
-limit of the t
i
s. In more detail: For each n there is a 6
(unique)
n
2
n
such that i : t
i
n =
n
D
s
; since D
s
is a lter, there is a real 2
with n =
n
for 7
each n. Finally, let q
_
i
q
i
.
8
Lemma 2.16. Let p L
D
with stem s, and let (
x
k
)
k
be a sequence of L
D
-names for reals in 2
. Then 9
there exists a pure extension q
0
p and a family of reals (
)
q, s
2
i]
n =
n D
.
Proof. We apply Lemma 2.14 to each node in p above s (and to
0
p
[]
1
and a
, and let 13
p
2
_
x
k
)
k
is a sequence of names for reals in 2
and p L
D
such that 16
(2.18) p X
_
m
_
km
[
x
k
k
].
Let s
<
be the stem of p.
17
By Lemma 2.16, we can x a pure extension q
0
p and a family (
)
q, s
2
)
q, s
; more specically, we can pick a family of natural numbers (n
)
q, s
such 21
that x
for any .
22
We can now nd r
0
q such that:
23
For all r above s and all i succ
r
() we have i > n
.
24
For all r above s and all i succ
r
() we have r
[
i]
.
25
So for all r above s we have, writing k for , that r
[
i]
forces x
x
k
n
] [
x
k
k
]. We conclude 26
that r forces x
_
ks
[
x
k
k
], contradicting (2.18).
27
Corollary 2.19. Let (t
k
)
k
be a dense subset of 2
.
28
Let
D be a system of ultralters, and let
be the L
D
-name for the ultralaver real. Then the set
H
_
m
_
km
[t
k
k
]
is forced to be a comeager set with the property that
(k) (the function we use will be described in Corollary 3.3). We will assume that
= (
k
)
k
is 36
an increasing sequence of natural numbers with
k+1
B
and k we write X[
k
,
k+1
) for the set x[
k
,
k+1
) : x X. We say that
39
X 2
and B
k
,
k+1
) 40
B
(k).
1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 13
X 2
and B
of
in the following way:
k
=
n
k
where n
k+1
n
k
= B
(k) 2
k
. 7
Then we get: If X is thin with respect to
and B
and B
. Let (X
j
)
j
be an enumeration 9
of Y
i
: i where each Y
i
appears innitely often. So X
_
m
_
jm
X
j
.
10
By induction on j , we nd for all j > 0 some k
j
> k
j1
such that 11
X
j
[
k
j
,
k
j
+1
) B
(k
j
) hence X
j
[0,
k
j
+1
) B
(k
j
) 2
k
j
= n
k
j
+1
n
k
j
.
So we can enumerate X
j
[0,
k
j
+1
) as (s
i
)
n
k
j
i<n
k
j
+1
. Hence X
j
is a subset of
_
n
k
j
i<n
k
j
+1
[s
i
]; and each s
i
has 12
length
k
j
+1
i
, since
k
j
+1
=
n
k
j
+1
and i < n
k
j
+1
. This implies 13
X
_
m
_
jm
X
j
_
m
_
im
[s
i
]
Hence X is smz with respect to
.
14
Lemma 2.12 and Lemma 2.22 yield:
15
Corollary 2.23. Let B
be constructed from B
and
as in Lemma 2.22. 17
Then L
D
forces that for every uncountable X 2
:
18
X is not smz with respect to
.
19
X is not thin with respect to
and B
.
20
2.E. Ultralaver and preservation of Lebesgue positivity. It is well known that both Laver forcing and 21
random forcing preserve Lebesgue positivity; in fact they satisfy a stronger property that is preserved 22
under countable support iterations. (So in particular, a countable support iteration of Laver and random 23
also preserves positivity.)
24
Ultralaver forcing L
D
will in general not preserve positivity. Indeed, if all ultralters D
s
are equal to the 25
same ultralter D
, so every 26
ground model real x 2
Z r i there is a name
n
r.
Let P be a forcing notion, and
Z below p is 5
some code Z
Zm = Z
m. 6
Usually we demand (which allows a simpler proof of the preservation theorem at limit stages) that the 7
sequence (p
0
, p
1
, . . . ) is inconsistent, i.e., p forces that there is an m such that p
m
G. Note that whenever 8
P adds a new -sequence of ordinals, we can nd such an interpretation for any
Z.
9
We now turn to preservation of Lebesgue positivity:
10
Denition 2.27. (1) Aforcing notion P preserves Borel outer measure, if P forces Leb
(A
V
) = Leb(A
V[G
P
]
) 11
for every code for a Borel set A. (Leb
, let P, p,
is an interpretation of
Z, and assume Z
0
r. Then there is an N-generic q p forcing that r is 16
still random over N[G] and moreover,
Z
0
r. (In particular, P has to be proper.)
17
(3) Assume that P is absolutely denable. P strongly preserves randoms over countable models if (2) 18
holds for all countable (transitive
18
) models N of ZFC*.
19
It is easy to see that these properties are increasing in strength. (Of course (3)(2) works only if ZFC* 20
is satised in H(
).)
21
In [KS05] it is shown that (1) implies (3), provided that P is nep (non-elementary proper, i.e., nicely 22
denable and proper with respect to countable models). In particular, every Suslin ccc forcing notion such 23
as random forcing, and also many tree forcing notions including Laver forcing, are nep. However L
D
is not 24
nicely denable in this sense, as its denition uses ultralters as parameters.
25
Lemma 2.28. Both Laver forcing and random forcing strongly preserve randoms over countable models.
26
Proof. For random forcing, this is easy and well known (see, e.g., [BJ95, 6.3.12]).
27
For Laver forcing: By the above, it is enough to show (1). This was done by Woodin (unpublished) and 28
Judah-Shelah [JS90]. A nicer proof (including a variant of (2)) is given by Pawlikowski [Paw96b].
29
Ultralaver will generally not preserve Lebesgue positivity, let alone randomness. However, we get 30
the following local variant of strong preservation of randoms (which will be used in the preservation 31
theorem 4.33). The rest of this section will be devoted to the proof of the following lemma.
32
Lemma 2.29. Assume that M is a countable model,
D
M
an ultralter system in M and r a random real 33
over M. Then there is (in V) an ultralter system
D extending
19
D
M
, such that the following holds: 34
If
35
p L
D
M ,
36
in M,
Z = (
Z
1
, . . . ,
Z
m
) is a sequence of L
D
M-names for codes for null sets,
20
and Z
1
, . . . , Z
m
are 37
interpretations under p, witnessed by a sequence (p
n
)
n
with strictly increasing stems,
38
r is random over M,
39
Z
i
k
i
r for i = 1, . . . , m,
40
then there is a q p in L
D
forcing that
41
r is random over M[G
M
],
42
Z
i
k
i
r for i = 1, . . . , m.
1
18
later we will introduce ord-transitive models, and it is easy to see that it does not make any dierence whether we demand
transitive or not; this can be seen using a transitive collapse
19
This implies, by Lemma 2.4, that the L
D
-generic lter G induces an L
D
M-generic lter over M, which we call G
M
.
20
Recall that nullset(
Z) =
_
n
_
kn
z
h
by
z
h
( y, p
[s]
) : s F, y h(s). (2.31)
(This does not depend on the
D we use, since we set y ( x,
<
) : x y.)
7
Up to forced equality, the name
z
h
is characterized by the fact that p
[s]
forces (in any L
D
) that
z
h
= h(s), 8
for every s in the domain of h.
9
Note that the same object h can be viewed as a front name below p with respect to dierent forcings 10
L
D
1
, L
D
2
, as long as p L
D
1
L
D
2
.
11
Denition 2.32. Let p
<
be a tree. A continuous name below p is either of the following:
12
An -sequence of front names below p.
13
A -increasing function g : p
<
such that lim
n
lh(g(cn)) = for every branch c [p].
14
For each n, the set of minimal elements in s p : lh(g(s)) > n is a front, so each continuous name in the 15
second sense naturally denes a name in the rst sense, and conversely. Being a continuous name below p 16
does not involve the notion of nor does it depend on the lter system
D.
17
If g is a continuous name and
D is any lter system, we can again dene the corresponding L
D
-name
Z
g
(in the sense of forcing); we leave a formal denition of
Z
g
to the reader and content ourselves with this
characterization:
(s p) : p
[s]
L
D
g(s)
Z
g
. (2.33)
Note that a continuous name below p naturally corresponds to a continuous function F : [p]
, and 18
Z
g
is forced (by p) to be the value of F at the generic real
.
19
Lemma 2.34. L
D
has the following pure decision properties:
20
(1) Whenever
y =
z
h
(is forced) for a front name h below p
1
.
22
(2) Whenever
Y =
Z
g
(is forced) for some continuous name g below q.
24
(3) (This is Lemma 2.9.) If A is a nite set,
a name, p L
D
, and p forces
A, then there is A 25
and a pure extension q
0
p such that q
= .
26
Proof. Let p L
D
, s
0
stem(p),
z
h
28
for a front name h). We can nd p
1
0
p such that for all t p
1
above s
0
: If there is q
0
p
[t]
1
such that t is 29
good in q, then t is already good in p
1
.
30
We claim that s
0
is now good (in p
1
). Note that for any bad node s the set t succ
p
1
(s) : t bad is 31
in D
+
s
. Hence, if s
0
is bad, we can inductively construct p
2
0
p
1
such that all nodes of p
2
are bad nodes 32
in p
1
. Now let q p
2
decide
y, s stem(q). Then q
0
p
[s]
1
, so s is good in p
1
, contradiction. This nishes 33
the proof of (1).
34
To prove (2), we rst construct p
1
as in (1) with respect to
y
0
. This gives a front F
1
p
1
deciding
y
0
. 35
Above each node in F
1
we nowrepeat the construction from(1) with respect to
y
1
, yielding p
2
, etc. Finally, 36
q
_
n
p
n
.
37
To prove (3): Similar to (1), we can nd p
1
0
p such that for each t p
1
: If there is a pure extension 38
of p
[t]
1
deciding
, then p
[t]
1
decides
; in this case we again call t good. Since there are only nitely many 39
possibilities for the value of
in V[G] is the 2
evaluation of a continuous name
Z
g
by G.
3
Proof. Assume that p L
D
forces that
Y =
Z
g
. This is a contradiction.
5
We will need the following modication of the concept of continuous names.
6
Denition 2.36. Let p
<
be a tree, b [p] a branch. An almost continuous name below p (with 7
respect to b) is a -increasing function g : p
<
such that lim
n
lh(g(cn)) = for every branch 8
c [p], except possibly for c = b.
9
Note that except possibly for c = b is the only dierence between this denition and the denition of 10
a continuous name. 11
Since for any
D it is forced
22
that the generic real (for L
D
) is not equal to the exceptional branch b, we 12
again get a name
Z
g
of a function in
satisfying: 13
(s p) : p
[s]
L
D
g(s)
Z
g
.
An almost continuous name naturally corresponds to a continuous function F from [p] \ b into
.
14
Note that being an almost continuous name is a very simple combinatorial property of g which does 15
not depend on
D, nor does it involve the notion . Thus, the same function g can be viewed as an almost 16
continuous name for two dierent forcing notions L
D
1
, L
D
2
simultaneously.
17
Lemma 2.37. Let
D be a system of lters (not necessarily ultralters).
18
Assume that p = (p
n
)
n
witnesses that Y
is an interpretation of
Y as Y
Y =
Z
g
.)
25
(5)
Y is 29
continuous below q
[t]
0
. We can do this by Lemma 2.34(2).
30
Now we set 31
q
n
p
n
q
0
= q
[stem(p
n
)]
0
p
n
.
This takes care of (1) and (2). Now we show (4): Any branch c of q
0
not equal to b must contain a node 32
s
k b with s b, so c is a branch in q
[s
k]
0
, below which
Y was continuous.
33
The following lemmas and corollaries are the motivation for considering continuous and almost contin- 34
uous names. 35
Lemma 2.38. Let
D be a system of lters (not necessarily ultralters). Let p L
D
, let b be a branch, 36
and let g : p
<
be an almost continuous name below p with respect to b; write
Z
g
for the associated 37
L
D
-name.
38
Let r 2
be a real, n
0
. Then the following are equivalent:
39
(1) p
L
D
r
_
nn
0
Z
g
(n), i.e.,
Z
g
n
0
r.
40
(2) For all n n
0
and for all s p for which g(s) has length > n we have r g(s)(n).
41
Note that (2) does not mention the notion and does not depend on
D.
1
22
This follows from our assumption that all our lters contain the Fr echet lter.
23
It is easy to see that for every L
D
-name
Y and
Z
g
= (
Z(0),
Z
g
(n).
3
(1) (2): Assume that p does not force r
_
nn
0
Z
g
(n). So there is a condition q p and some 4
n n
0
such that q r
Z
g
(n). By increasing the stem of q, if necessary, we may assume that s stem(q) 5
is not on b (the exceptional branch), and that g(s) has already length > n. Let C
n
g(s)(n) be the n-th 6
entry of g(s). So p
[s]
already forces
Z
g
(n) = C
n
; now q
[s]
p
[s]
, and q
[s]
forces the following statements: 7
r
Z
g
(n),
Z
g
(n) = C
n
. Hence r C
n
, so (2) fails.
8
Corollary 2.39. Let
D
1
and
D
2
be systems of lters, and assume that p is in L
D
1
L
D
2
. Let g : p
<
9
be an almost continuous name of a sequence of clopen sets, and let
Z
g
1
and
Z
g
2
be the associated L
D
1
-name 10
and L
D
2
-name, respectively.
11
Then for any real r and n we have 12
p
L
D
1
Z
g
1
n
r p
L
D
2
Z
g
2
n
r.
(We will use this corollary for the special case that L
D
1
is an ultralaver forcing, and L
D
2
is Laver forcing.)
13
Lemma 2.40. Let
D
1
and
D
2
be systems of lters, and assume that p is in L
D
1
L
D
2
. Let g : p
<
be 14
a continuous name of a sequence of clopen sets, let F p be a front and let h : F be a front name. 15
Again we will write
Z
g
1
,
Z
g
2
for the associated names of codes for null sets, and we will write
n
1
and
n
2
for 16
the associated L
D
1
- and L
D
2
-names, respectively, of natural numbers.
17
Then for any real r we have: 18
p
L
D
1
Z
g
1
n
1
r p
L
D
2
Z
g
2
n
2
r.
Proof. Assume p
L
D
1
Z
g
1
n
1
r. So for each s F we have: p
[s]
L
D
1
Z
g
1
h(s)
r. By Corollary 2.39, we 19
also have p
[s]
L
D
2
Z
g
2
h(s)
r. So also p
[s]
L
D
2
Z
g
2
n
2
r for each s F. Hence p
L
D
2
Z
g
2
n
2
r.
20
Corollary 2.41. Assume q L forces in Laver forcing that
Z
g
k
r for k = 1, 2, . . ., where each g
k
is a 21
continuous name of a code for a null set. Then there is a Laver condition q
0
q such that for all ultralter 22
systems
D we have:
23
If q
L
D
, then q
Z
g
k
r for all k.
24
Proof. By (2.26) we can nd a sequence (
n
k
)
k=1
of L-names such that q
Z
g
k
n
k
r for each k. By 25
Lemma 2.34(2) we can nd q
0
q be such that this sequence is continuous below q
. Since each
n
k
is 26
now a front name below q
,
D
M
M an ultralter system,
D a lter system 28
extending
D
M
, q L
D
. For any V-generic lter G L
D
we write G
M
for the (M-generic, by Lemma 2.4) 29
lter on L
D
M .
30
The following are equivalent:
31
(1) q
L
D
r is random over M[G
M
].
32
(2) For all names
Z r.
33
(3) For all continuous names g M: q
L
D
Z
g
r.
34
Proof. (1)(2) holds because every null set is contained in a set of the form nullset(Z), for some code Z.
35
(2)(3): Every code for a null set in M[G
M
] is equal to
Z
g
[G
M
], for some g M, by Corollary 2.35.
36
Lemma 2.43. Let r be random over a countable model M. Then there is a countable model M
M such 37
that (2
)
M
is countable in M
.
38
Proof. We will need the following forcing notions, all dened in M: 39
M
C
//
B
1
M
C
B
2
M
B
1
P=C
B
2
/B
1
//
M
C
B
2
Let C be the forcing that collapses the continuum to with nite conditions.
1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
18 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
Let B
1
be random forcing (trees T 2
<
of positive measure).
2
Let
B
2
be the C-name of random forcing.
3
Let i : B
1
C
B
2
be the natural complete embedding T (1
C
, T).
4
Let
P be a B
1
-name for the forcing C
B
2
/i[G
B
1
], the quotient of C
B
2
by the complete subforcing 5
i[B
1
].
6
The random real r is B
1
-generic over M. In M[r] we let P
P C
B
2
induces an M-generic lter J C and an M[J]-generic lter K
B
2
[J]; it 8
is easy to check that K interprets the
B
2
-name of the canonical random real as the given random real r.
9
Hence r is random over the countable model M
M[J], and (2
)
M
is countable in M
. 10
M
J
//
r
M[J]
K
M[r]
H
//
M[r][H]
11
Proof of Lemma 2.29. We will rst describe a construction that deals with a single triple ( p,
Z,
Z
) (where 12
p is a sequence of conditions with strictly increasing stems which interprets
Z as
Z
= q
( p,
Z,
Z
Z as
Z
Z and Z
instead of
Z
1
and Z
.
17
Using Lemma 2.37 we may (strengthening the conditions in our interpretation) assume (in M) that the 18
sequence (
Z(k))
k
is almost continuous, witnessed by g : p
<
. By Lemma 2.43, we can nd a model 19
M
M such that (2
)
M
is countable in M
.
20
We now work in M
. Note that g still denes an almost continuous name, which we again call
Z.
21
Each lter in D
M
s
is now countably generated; let A
s
be a pseudo-intersection of D
M
s
which additionally 22
satises A
s
succ
p
(s) for all s p above the stem. Let D
s
be the Fr echet lter on A
s
. Let p
L
D
be the 23
tree with the same stem as p which satises succ
p
(s) = A
s
for all s p
, the forcing L
D
is just Laver forcing L, and that
p
L
p. Using Lemma 2.28 we can nd a condition q p
-generic. (2.44)
q
L
r is random over M
[G
L
] (hence also over M[G
M
]). (2.45)
Moreover, q
L
Z
0
r. (2.46)
Enumerate all continuous L
D
M -names for codes for null sets from M as
Z
g
1
,
Z
g
2
, . . . Applying Corol- 27
lary 2.41 yields a condition q
L
E
, we have q
L
E
28
Z
g
i
r for all i. Corollary 2.39 and Lemma 2.42 now implies: 29
(2.47)
For every lter system
E satisfying q
L
E
, q
forces in L
E
that r is random over M
and that
Z
0
r.
By thinning out q
.
We have now described a construction of q
= q
( p,
Z, Z
).
31
Let ( p
n
,
Z
n
, Z
n
) enumerate all triples ( p,
Z, Z
) M where p interprets
Z as Z
( p
0
,
Z
0
, Z
0
).
35
Given q
n1
and ( p
n
,
Z
n
, Z
n
), we nd k
0
such that
n
k
0
q
0
q
n1
(using (2.48)). Let k
1
be 36
such that stem(p
n
k
1
) has length > k
0
. We replace p
n
by p
(p
n
k
)
kk
1
. (Obviously, p
still interprets 37
Z
n
as Z
n
.) Now let q
n
q
( p
Z
n
, Z
n
).
1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 19
Note that the stem of q
n
is at least as long as the stem of p
n
k
1
, and is therefore not in q
0
q
n1
, so 2
stem(q
i
) and stem(q
j
) are incompatible for all i j. Therefore we can choose for each s an ultralter D
s
3
extending D
M
s
such that stem(q
i
) s implies succ
q
i (s) D
s
.
4
Note that all q
i
are in L
D
. Therefore, we can use (2.47). Also, q
i
p
i
0
.
5
3. Jxxts roacrxo 6
In this section, we dene a family of forcing notions that has two faces (hence the name Janus forcing): 7
Elements of this family may be countable (and therefore equivalent to Cohen), and they may also be 8
essentially random.
9
We will need the following properties of the Janus forcing notions J:
10
Wish List 3.1. Throughout the whole paper we x a function B
= (
m
)
m
. (In 12
our application,
. Whenever X 2
Q, also J
[H] 20
is the same as
Z
M
[H
M
], where H
M
is the induced lter on J
M
.
21
(3) Let M be a countable transitive model and J
M
a Janus forcing in M. Then J
M
is a Janus forcing in 22
V as well (and of course countable in V). (See Fact 3.8.)
23
(4) Whenever M is a countable transitive model, J
M
is a Janus forcing in M, then there is a Janus 24
forcing J such that
25
J is (in V) equivalent to random forcing (actually we just need some forcing that strongly 26
preserves random reals)
27
J
M
is an M-complete subforcing of J.
28
(See Lemma 3.16.)
29
(Advice to the reader: At rst reading, you may want to take these properties as granted, and skip to 30
Section 4.)
31
3.A. Denition of Janus. A Janus forcing J will consist of:
32
A countable core (or: backbone) which is dened in a combinatorial way from a parameter
. 33
(In our application, we will use a Janus forcing immediately after an ultralaver forcing, and
will 34
be a subsequence of the ultralaver real.) This core is of course equivalent to Cohen forcing.
35
Some additional stung J \ (countable
24
or uncountable). We allow great freedomfor this, we 36
just require that the core is a suciently complete subforcing (in a specic combinatorial sense).
37
We will use the following combinatorial theorem from [BS10]:
38
Lemma 3.2 ([BS10, Theorem 8]
25
). For every , > 0 there exists N
,
such that for all suciently 39
large nite sets I there is a nonempty family 7
I
consisting of sets A 2
I
,
A
2
I
such that if X 2
I
, 40
X N
,
then 41
A 7
I
: X + A = 2
I
7
I
1 .
(Recall that X + A x + a : x X, a A.)
42
Rephrasing and specializing to =
1
4
and =
1
2
i
we get:
1
24
Also the trivial case J = is allowed.
25
The theorem in [BS10] actually says for a suciently large I, but the proof shows that this should be read as for all suciently
large I.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
20 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
Corollary 3.3. For every i there exists B
(i) there is a 2
nonempty family 7
I
satisfying the following:
3
7
I
consists of sets A 2
I
with
A
2
I
1
2
i
.
4
For every X 2
I
satisfying X B
= (
i
)
i
of natural numbers; more 6
precisely, the sequence
i+1
i
B
i
,
i+1
) according to Corollary 3.3.
9
Denition 3.5. First we dene the core =
of our forcing: 10
=
_
i
_
j<i
7
j
.
In other words, i = (A
0
, . . . , A
i1
) for some i , A
0
7
0
, . . . , A
i1
7
i1
. We will denote the 11
number i by height().
12
The forcing notion is ordered by reverse inclusion (i.e., end extension): if .
13
Denition 3.6. Let
= (
i
)
i
be as in the assumption above. We say that J is a Janus forcing based on
14
if: 15
(1) (, ) is an incompatibility-preserving subforcing of J.
16
(2) For each i the set : height() = i is predense in J. So in particular, J adds a 17
branch through . The union of this branch is called
= (
0
,
1
,
2
, . . .), where
i
2
L
i
with 18
i
7
i
.
19
(3) Fatness:
26
For all p J and all real numbers > 0 there are arbitrarily large i such that there 20
is a core condition = (A
0
, . . . , A
i1
) (of length i) with
21
A 7
i
:
A j
J
p
7
i
1 .
(4) J is ccc.
22
(5) J is separative.
23
(6) (To simplify some technicalities:) J H(
1
).
24
We now dene
, which will be a canonical J-name for (a code for) a null set. We will use the sequence 25
i
denes a clopen set
i
= x 2
: xL
i
i
of measure at most
1
2
i+1
. The 27
sequence
= (
0
,
1
,
2
, . . .) is (a name for) a code for the null set 28
nullset(
) =
_
n<
_
in
Z
i
.
For simplicity, we will write
instead of nullset(
).
29
For later reference, we record the following trivial fact:
30
Fact 3.8. Let (M
n
)
n
be an increasing sequence of countable models, and let J
n
M
n
be Janus forcings. 31
Assume that J
n
is M
n
-complete in J
n+1
. Then
_
n
J
n
is a Janus forcing, and an M
n
-complete extension of 32
J
n
for all n. 33
This is also true if all M
n
are equal. In other words: If J M is a Janus forcing in M, then J is also a 34
Janus forcing in V.
1
26
Actually, (3) implies (2).
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 21
3.B. Janus and strongly meager. Carlson [Car93] showed that Cohen reals make every uncountable 2
set X of the ground model not strongly meager in the extension (and that not being strongly meager is 3
preserved in a subsequent forcing with precaliber
1
). We show that a countable Janus forcing J does the 4
same (for a subsequent forcing that is even -centered, not just precaliber
1
). This sounds trivial, since 5
any (nontrivial) countable forcing is equivalent to Cohen forcing anyway. However, we show (and will 6
later use) that the canonical null set
= (
i
)
i
and B
and B
, and
.
15
Proof. Let
c :
R witnessing that
R is -centered.
16
Recall that
= 2
. Assume towards 17
a contradiction that (p, r) J
R forces that X +
. Then we can x a (J
R)-name
such that 18
(p, r)
X +
, i.e., (p, r) (x X)
x +
. By denition of
, we get 19
(p, r) (x X) (n ) (i n)
L
i
xL
i
+
i
.
For each x X we can nd (p
x
, r
x
) (p, r) and natural numbers n
x
and m
x
such that p
x
forces 20
that
c(r
x
) = m
x
and 21
(p
x
, r
x
) (i n
x
)
L
i
xL
i
+
i
.
So X =
_
pJ,m,n
X
p,m,n
, where X
p,m,n
is the set of all x with p
x
= p, m
x
= m, n
x
= n. (Note that J is 22
countable, so the union is countable.) As X is not thin, there is some p
, m
, n
such that X
X
p
,m
,n
is 23
not very thin. So we get for all x X
: 24
(3.10) (p
, r
x
) (i n
L
i
xL
i
+
i
.
Since X
L
i
has more than B
(i) elements.
Due to the fact that J is a Janus forcing (see Denition 3.6 (3)), there are arbitrarily large i such that 26
there is a core condition = (A
0
, . . . , A
i1
) with 27
(3.12)
A 7
i
:
A j
J
p
7
i
2
3
.
Fix such an i larger than both i
0
and n
A j
J
p
and A 7
i
: X
L
i
+ A = 2
L
i
.
By (3.12), the relative measure (in 7
i
) of the left one is at least
2
3
; due to (3.11) and the denition of 7
i
30
according to Corollary 3.3, the relative measure of the right one is at least
3
4
; so the two sets in (3.13) are 31
not disjoint, and we can pick an A belonging to both.
32
Clearly,
i
is equal to A. Fix q J witnessing
A j
J
p
. Then 33
(3.14) q
J
X
L
i
+
i
= X
L
i
+ A = 2
L
i
.
Since p
the color
c(r
x
) = m
, we can nd an r
which is (forced by q p
34
to be) a lower bound of the nite set r
x
: x X
, where X
L
i
= X
L
i
.
35
By (3.10), 36
(q, r
L
i
X
L
i
+
i
= X
L
i
+
i
,
contradicting (3.14).
1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
22 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
Corollary 3.15. Let X be uncountable. If L
D
is any ultralaver forcing adding an ultralaver real
, and
2
is dened from
as in Lemma 2.22, and if
Q is any -centered 3
forcing, then L
D
). 7
Then there is a Janus forcing J (based on
) such that
8
J
M
M
J (i.e., J
M
is an M-complete subforcing of J).
9
J is forcing equivalent to random forcing.
10
Proof. Let r
n
: n be an enumeration of J
M
J
M
. Let D
k
: k be an enumeration of all D M 11
which are open dense subsets of J
M
. Recall that =
J
M
was dened in Denition 3.5.
12
We now x n for a while (up to (3.18)). We will construct a nitely splitting tree S
n
<
and a family 13
(
n
s
, p
n
s
,
n
s
)
sS
n satisfying the following (suppressing the superscript n):
14
(a)
s
,
()
= (), s t implies
s
t
, and s
S
n t implies
s
t
. 15
(So in particular the set
t
: t succ
S
n (s) is a (nite) antichain above
s
in .)
16
(b) p
s
J
M
, p
()
= r
n
; if s t then p
t
J
M p
s
(hence p
t
r
n
); s
S
n t implies p
s
J
M p
t
.
17
(c) p
s
J
M
s
.
18
(d)
s
s
, and
t
: t succ
S
n (s) is the set of all succ
s
) which are compatible with p
s
.
19
(e) The set
t
: t succ
S
n (s) is a subset of succ
s
) of relative size at least 1
1
lh(s)+10
.
20
(f) Each s S
n
has at least 2 successors (in S
n
).
21
(g) If k = lh(s), then p
s
D
k
(and therefore also in all D
l
for l < k).
22
Set
()
= () and p
()
= r
n
. Given s,
s
and p
s
, we construct succ
S
n (s) and (
t
, p
t
)
tsucc
S
n (s)
: We apply 23
fatness 3.6(3) to p
s
with =
1
lh(s)+10
. So we get some
s
of height bigger than the height of
s
such 24
that the set B of elements of succ
s
) which are compatible with p
s
has relative size at least 1 . Since 25
p
s
J
M
s
we get that
s
is compatible with (and therefore stronger than)
s
. Enumerate B as
0
, . . . ,
l1
. 26
Set succ
S
n (s) = s
i : i < l and
s
i
=
i
. For t succ
S
(s), choose p
t
J
M
stronger than both
t
and p
s
27
(which is obviously possible since
t
and p
s
are compatible), and moreover p
t
D
lh(t)
. This concludes the 28
construction of the family (
n
s
, p
n
s
,
n
s
)
sS
n .
29
So (S
n
, ) is a nitely splitting nonempty tree of height with no maximal nodes and no isolated 30
branches. [S
n
] is the (compact) set of branches of S
n
. The closed subsets of [S
n
] are exactly the sets of 31
the form [T], where T S
n
is a subtree of S
n
with no maximal nodes. [S
n
] carries a natural (uniform) 32
probability measure
n
, which is characterized by 33
n
((S
n
)
[t]
) =
1
succ
S
n (s)
n
((S
n
)
[s]
)
for all s S
n
and all t succ
S
n (s). (We just write
n
(T) instead of
n
([T]) to increase readability.)
34
We call T S
n
positive if
n
(T) > 0, and we call T pruned if (T
[s]
) > 0 for all s T. (Clearly every 35
tree T contains a pruned tree T
of the same measure, which can be obtained from T by removing all nodes 36
s with (T
[s]
) = 0.)
37
Let T S
n
be a positive, pruned tree and > 0. Then on all but nitely many levels k there is an s T 38
such that 39
(3.17) succ
T
(s) succ
S
n (s) has relative size 1 .
(This follows from Lebesgues density theorem, or can easily be seen directly: Set C
m
=
_
tT, lh(t)=m
(S
n
)
[t]
. 40
Then C
m
is a decreasing sequence of closed sets, each containing [T]. If the claim fails, then
n
(C
m+1
)) 41
n
(C
m
) (1 ) innitely often; so
n
(T) (
_
m
C
m
) = 0.)
42
It is well known that the set of positive, pruned subtrees of S
n
, ordered by inclusion, is forcing equivalent 43
to random forcing (which can be dened as the set of positive, pruned subtrees of 2
<
).
44
We have now constructed S
n
for all n. Dene
J = J
M
_
n
_
(n, T) : T S
n
is a positive pruned tree
_
(3.18)
with the following partial order:
1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 23
The order on J extends the order on J
M
. 2
(n
, T
) (n, T) if n = n
and T
T.
3
(n, T) p if there is a k such that p
n
t
p for all t T of length k. (Note that this will then be true 4
for all bigger k as well.)
5
p (n, T) never holds.
6
The lemma now easily follows from the following properties:
7
(1) The order on J is transitive.
8
(2) J
M
is an incompatibility-preserving subforcing of J. 9
In particular, J satises item (1) of Denition 3.6 of Janus forcing.
10
(3) For all k: the set (n, T
[t]
) : t T, lh(t) = k is a (nite) predense antichain below (n, T).
11
(4) (n, T
[t]
) is stronger than p
n
t
for each t T (witnessed, e.g., by k = lh(t)). Of course, (n, T
[t]
) is 12
stronger than (n, T) as well.
13
(5) Since p
n
t
D
k
for k = lh(t), this implies that each D
k
is predense below each (n, S
n
) and therefore 14
in J. 15
So in particular, J
M
M
J, i.e., J
M
is an M-complete subforcing of J. 16
Also, for each k the set : height() = k is in M (and predense in J
M
). Therefore this set is 17
predense in J as well, i.e., item (2) of the Denition 3.6 of Janus forcing is satised.
18
(6) The condition (n, S
n
) is stronger than r
n
, so (n, S
n
) : n is predense in J and J \ J
M
is dense 19
in J. 20
Below each (n, S
n
), the forcing J is isomorphic to random forcing. 21
Therefore, J itself is forcing equivalent to random forcing. (In fact, the complete Boolean algebra 22
generated by J is isomorphic to the standard random algebra, Borel sets modulo null sets.)
23
(7) The remaining item of the denition of Janus forcing, fatness 3.6(3), is satised. 24
I.e., given (n, T) J and > 0 there is an arbitrarily high
. We claim that
s
is as required:
29
Let B
t
: t succ
S
(s). Note that B = succ
s
) : j p
s
.
30
B has relative size 1
1
lh(s)
1 in succ
s
) (according to property (e) of S
n
).
31
C
t
: t succ
T
(s) is a subset of B of relative size 1 according to our choice of s.
32
So C is of relative size (1 )
2
in succ
s
).
33
Each
t
C is compatible with (n, T), as (n, T
[t]
) p
t
t
(see (4)).
34
It is easy (but not even necessary) to check that J is separative. In any case, we could replace
J
by
J
, 35
thus making J separative without changing
J
M, since J
M
was already separative.
36
4. Aimosr rrxrrr xxo ximosr cotxrxair strroar rrraxrroxs 37
A main tool to construct the forcing for BC+dBC will be partial countable support iterations, more 38
particularly almost nite support and almost countable support iterations. A partial countable support 39
iteration is a forcing iteration (P
, Q
)
<
2
such that for each limit ordinal the forcing notion P
is a subset 40
of the countable support limit of (P
, Q
)
<
which satises some natural properties (see Denition 4.6).
41
Instead of transitive models, we will use ord-transitive models (which are transitive when ordinals are 42
considered as urelements). Why do we do that? We want to approximate the generic iteration
P of length 43
2
with countable models; this can be done more naturally with ord-transitive models (since obviously 44
countable transitive models only see countable ordinals). We call such an ord-transitive model a candi- 45
date (provided it satises some nice properties, see Denition 4.2). A basic point is that forcing extensions 46
work naturally with candidates.
47
In the following, x = (M
x
,
P
x
) will denote a pair such that M
x
is a candidate and
P
x
is (in M
x
) a partial 48
countable support iteration; similarly we write, e.g., y = (M
y
,
P
y
) or x
n
= (M
x
n
,
P
x
n
).
49
We will need the following results to prove BC+dBC. (However, other than in the case of the ultralaver 50
and Janus section, the reader will probably have to read this section to understand the construction in the 51
next section, and not just the wish list.)
1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
24 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
Wish List 4.1. Given x = (M
x
,
P
x
), we can construct by induction on a partial countable support iteration 2
P = (P
, Q
)
<
2
satisfying:
3
There is a canonical M
x
-complete embedding from
P
x
to
P.
4
In this construction, we can use at each stage any desired Q
, as long as P
forces that Q
x
is (evaluated 5
as) an M
x
[H
x
]-complete subforcing of Q
(where H
x
P
x
is the M
x
-generic lter induced by the generic 6
lter H
). 7
Moreover, we can demand either of the following two additional properties of this limit:
27
8
(1) If all Q
2
is -centered.
9
(2) If r is random over M
x
, and all Q
] (see Deni- 10
tion 4.31), then also P
2
locally preserves the randomness of r.
11
Actually, we need the following variant: Assume that we already have P
0
for some
0
M
x
, and that P
x
0
12
canonically embeds into P
0
, and that the respective assumption on Q
0
forces that the quotient P
2
/P
0
satises the respective conclusion.
14
We also need:
28
15
(3) If instead of a single x we have a sequence x
n
such that each P
x
n
canonically (and M
x
n
-completely) 16
embeds into P
x
n+1
, then we can nd a partial countable support iteration
P into which all P
x
n
embed 17
canonically (and we can again use any desired Q
, assuming that Q
x
n
is an M
x
n
[H
x
n
]-complete 18
subforcing of Q
for all n ).
19
(4) (A fact that is easy to prove but awkward to formulate.) If a -system argument produces two 20
x
1
, x
2
as in Lemma 5.7(3), then we can nd a partial countable support iteration
P such that
P
x
i
21
canonically (and M
x
i
-completely) embeds into
P for i = 1, 2.
22
4.A. Ord-transitive models. We will use ord-transitive models, as introduced in [She04] (see also the 23
presentation in [Kel]). We briey summarize the basic denitions and properties (restricted to the rather 24
simple case needed in this paper):
25
Denition 4.2. Fix a suitable nite subset ZFC
).
27
(1) A set M is called candidate, if
28
M is countable,
29
(M, ) is a model of ZFC
,
30
M is ord-absolute: M = Ord i Ord, for all M,
31
M is ord-transitive: if x M \ Ord, then x M,
32
+ 1 M. 33
is a limit ordinal and = + 1 are both absolute between M and V.
34
(2) A candidate M is called nice, if has countable conality and the countable set A is conal 35
in both are absolute between M and V. (So if M has countable conality, then M is 36
conal in .) Moreover, we assume
1
M (which implies
1
M
=
1
) and
2
M (but we do 37
not require
2
M
=
2
).
38
(3) Let P
M
be a forcing notion in a candidate M. (To simplify notation, we can assume without loss 39
of generality that P
M
Ord = (or at least ) and that therefore P
M
M and also A M 40
whenever M thinks that A is a subset of P
M
.) Recall that a subset H
M
of P
M
is M-generic (or: 41
P
M
-generic over M), if A H
M
= 1 for all maximal antichains A in M.
42
(4) Let H
M
be P
M
-generic over M and
a P
M
-name in M. We dene the evaluation
[H
M
]
M
to be x if 43
M thinks that p
P
M
= x for some p H
M
and x M (or equivalently just for x MOrd), and 44
[H
M
]
M
: (
, p)
, p H
M
otherwise. Abusing notation we write
[H
M
] instead of
[H
M
]
M
, 45
and we write M[H
M
] for
[H
M
] :
is a P
M
-name in M.
46
(5) The ord-collapse k (or k
M
) is a recursively dened function with domain M: k(x) = x if x Ord, 47
and k(x) = k(y) : y x M otherwise.
1
27
The -centered version is central for the proof of dBC; the random preserving version for BC.
28
This will give -closure and
2
-cc for the preparatory forcing R.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 25
(6) The ord-transitive closure of a set x is dened inductively on the rank: 2
ordclos(x) = x
_
ordclos(y) : y x \ Ord.
So ordclos(x) is the smallest ord-transitive set containing x as a subset. HCON is the collection of 3
all sets x such that the ord-transitive closure of x is countable. x is in HCON i x is element of 4
some candidate. In particular, all reals and all ordinals are HCON.
5
We write HCON
for the family of all sets x in HCON whose ord-transitive closure (or, in this 6
case equivalently, transitive closure) only contains ordinals < .
7
Fact 4.3. (1) The ord-collapse of a countable elementary submodel of H(
) is a nice candidate.
8
(2) Unions, intersections etc. are generally not absolute for candidates. For example, let x M \ Ord. 9
In M we can construct a set y such that M = y =
1
x. Then y is not an ordinal and therefore a 10
subset of M, and in particular y is countable and y
1
x.
11
(3) Let j : M M
; moreover M 13
is characterized by the pair (M
of ZFC
, f ).
18
(4) All candidates M with M Ord
1
are hereditarily countable, so their number is at most 2
0
. 19
Similarly, the cardinality of HCON
, f ), then 24
H
M
P
M
is P
M
-generic over M i H
j[H
M
] is P
j(P
M
)-generic over M
[H
for all
2
M. Once we know (by induction) that 35
P
M
, we know that a P
-generic lter H
induces a P
M
-generic (over M) 36
lter which we call H
M
. Then M[H
M
], so it does 38
not make any dierence whether we use M[H
M
(that a candidate 40
has to satisfy). The following ne print hopefully absolves us from any liability. (It is entirely irrelevant 41
for the understanding of the paper.)
42
We have to guarantee that each M[H
M
.
47
Of course we can also nd a bigger (still nite) set ZFC
that implies:
10
exists, and each forcing 48
extension of the universe with a forcing of size
4
satises ZFC
. This guarantees that all forcing extensions (by small forcings) of nice candidates will be 1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
26 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
candidates (in particular, satisfy enough of ZFC such that our arguments about Janus or ultralaver forcings 2
work). Also, every ord-collapse of a countable elementary submodel N of H() will be a nice candidate.
3
4.B. Partial countable support iterations. We introduce the notion of partial countable support limit: 4
a subset of the countable support (CS) limit containing the union (i.e., the direct limit) and satisfying some 5
natural requirements.
6
Let us rst describe what we mean by forcing iteration. They have to satisfy the following require- 7
ments: 8
A topless forcing iteration (P
, Q
)
<
is a sequence of forcing notions P
and P
-names Q
9
of quasiorders with a weakest element 1
Q
. 10
Each P
, then p P
. Also, p
P
p() Q
will 12
always be the natural one: q p i q forces (in P
) that q
tot
() p
tot
() for all < , where 13
r
tot
() = r() for all dom(r) and 1
Q
otherwise. P
+1
consists of all p with p P
and 14
p p
tot
() Q
, so it is forcing equivalent to P
.
15
P
.)
16
For any p P
and any q P
is a complete subforcing 18
of P
(which induces P
for 21
). For topped iterations we call the lter on the nal limit sometimes just H instead of H
.
22
We use the following notation for factorization of iterations:
23
For < , in the P
-extension V[H
], we let P
/H
with p H
24
(ordered as in P
/P
for the P
-name of P
/H
.
25
Since P
is a complete subforcing of P
is equivalent to P
(P
/H
).
27
Denition 4.5. Let
P be a (topless) iteration of limit length . We dene three limits of
P:
28
The direct limit is the union of the P
for all 31
< . This is the largest possible limit of the iteration.
32
The full countable support limit P
CS
of
P is the inverse limit if cf() = and the direct limit 33
otherwise. 34
We say that P
is a partial CS limit, if P
is a 35
topped iteration. In particular, this means that P
, q P
.
38
So for a given topless
P there is a well-dened inverse, direct and full CS limit. If cf() > , then they 39
all coincide. If cf() = , then the direct limit and the full CS limit (=inverse limit) dier. Both of them 40
are partial CS limits, but there are many more possibilities for partial CS limits. By denition, all of them 41
will yield iterations.
42
Note that the name CS limit is slightly inappropriate, as the size of supports of condition is not part of 43
the denition. To give a more specic example: Consider a topped iteration
P of length + where P
is 44
the direct limit and P
+
is the full CS limit. Let p be any element of the full CS limit of
P which is not 45
in P
; then p is not in P
+
either. So not every countable subset of + can appear as the support of a 46
condition. 47
Denition 4.6. A forcing iteration
P is called a partial CS iteration, if
48
every limit is a partial CS limit, and
49
every Q
coincides with
.
1
29
The reason for this requirement is briey discussed in Section 7
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 27
The following fact can easily be proved by transnite induction:
2
Fact 4.7. Let
P be a partial CS iteration. Then for all the forcing notion P
is separative.
3
From now on, all iterations we consider will be partial CS iterations. In this paper, we will only be 4
interested in proper partial CS iterations, but properness is not part of the denition of partial CS iteration. 5
(The reader may safely assume that all iterations are proper.)
6
Note that separativity of the Q
-generic. Then p H i p H
: If q
p.
11
(3) For all q, p P
: If q
: P
M
for
P
CS
s naturally:
20
Denition 4.9. For a topped partial CS iteration
P
M
in M of length and a topless one
P in V of length 21
, we dene j : P
M
P
CS
, 23
and build j(p) from the resulting coherent sequence.
24
We do not claim that j : P
M
P
CS
is M-complete.
31
In the following, we will construct partial CS 25
limits P
such that j : P
M
] P
.) We will actually dene two versions: The almost FS and the almost CS limit.
27
Instead of arbitrary systems of embeddings i
is a subset of Q
: P
M
.
35
Let = + 1. By induction hypothesis, i
induces 36
an M-generic lter H
M
i
1
[H
] P
M
[H
M
] is 37
an M[H
M
]-complete subforcing of Q
[H
)
M
. (If
sup( M) < , 39
then i
is a map into P
.) We require that P
contains the 40
range of i
, and that i
(p) = i
.
31
For example, if =
= and if P
M
P
CS
is not complete: In
M, let c
n
be (a P
M
n
-name for) a nontrivial element of Q
M
n
(i.e., 1
Q
M
n
c
n
). Let p
n
be the P
M
n
-condition c
m
H(m) for all m < n
and let q
n
be the P
M
n+1
-condition (c
0
, . . . , c
n1
, c
n
) = p
n
p
n+1
, i.e., n is minimal with c
n
H(n). In M, the set A = q
n
: n
is a maximal antichain in P
M
in P
CS
where
=
_
M
P
M
)
M
has range in P
, which will be 5
a complete subforcing of P
.
6
For now we just record:
7
Lemma 4.12. Assume that we have topped iterations
P
M
(in M) of length and
P (in V) of length
8
sup( M), and that for all M the canonical embedding i
: P
M
works. Let i
: P
M
P
CS
9
be the canonical extension. 10
(1) If P
M
is (in M) a direct limit (which is always the case if has uncountable conality) then i
11
(might not work, but at least) has range in P
in M to predense sets i
[D] 13
P
, then i
has range in
_
<
P
, which is subset of 15
any partial CS limit P
. Incompatibility in P
M
.
17
(2) Fix p
1
, p
2
P
M
. So x a generic lter H P
containing i
(p
1
) and i
(p
2
).
19
In M, we dene the following set D: 20
D q P
M
: (q p
1
q p
2
) or ( < : q
P
M
p
1
) or ( < : q
P
M
p
2
).
Using Fact 4.8(3) it is easy to check that D is dense. Since i
p
1
for some < . But the lter H contains both i
(q) and i
(p
1
), 23
contradicting the assumption that i
preserves incompatibility.
24
4.C. Almost nite support iterations. Recall Denition 4.9 (of the canonical extension) and the setup 25
that was described there: We have to nd a subset P
of P
CS
26
is M-complete.
27
We now dene the almost nite support limit. (The direct limit will in general not do, as it may not 28
contain the range j[P
M
]. The almost nite support limit is the obvious modication of the direct limit, and 29
it is the smallest partial CS limit P
] P
33
work for all M =
M. Let i
of P
CS
: 35
P
q i
(p) P
CS
: p P
M
and ( M) q
P
(p) .
Note that for cf() > , the almost FS limit is equal to the direct limit, as each p P
M
is in fact in P
M
36
for some M, so i
(p) = i
(p) P
.
37
Lemma 4.14. Assume that
P and
P
M
are as above and let P
is a 38
partial CS iteration, and i
works, i.e., i
to P
. (As P
is a complete 39
subforcing of P
is M-complete from P
M
to P
.)
1
32
For example: Let =
1
and
=
1
M. Assume that P
M
1
is (in M) a (or: the unique) partial CS limit of a nontrivial
iteration. Assume that we have a topless iteration
P of length
to be the full CS limit, then we cannot further extend it to any iteration of length
1
such that the canonical embedding
i
1
works: Let p
and q
:
1
is a maximal antichain, and the sequence (p
1
is a decreasing coherent sequence. But in V there is an element p
P
CS
with p
= p
is
clearly incompatible with all elements of j[A] = j(p
[P
M
, and let q i
(p) be a condition in P
. (If
< , i.e., if 4
cf() > , then we can choose p to be the empty condition.) Fix M be such that q P
. Let H
be 5
P
. Now in M[H
M
] the set p
j
: j J, p
j
P
M
/H
M
is predense 6
in P
M
/H
M
p, p
j
in P
M
/H
M
.
8
We can nd q
q deciding j and p
; since certainly q
(p
(p
) q i
(p) (since q
q and p
p), and q
(p
) i
(p
j
) (since 10
p
p
j
).
11
Denition and Claim 4.15. Let
P
M
be a topped partial CS iteration in M of length . We can construct by 12
induction on + 1 an almost nite support iteration
P over
P
M
(or: almost FS iteration) as follows:
13
(1) As induction hypothesis we assume that the canonical embedding i
] makes sense.)
15
(2) Let = + 1. If M, then we can use any Q
is an 16
M[H
M
]-complete subforcing of Q
.)
17
(3) Let M and cf() = . Then P
, Q
)
<
over P
M
.
18
(4) Let M and cf() > . Then P
, Q
)
<
over P
M
(which 19
also happens to be the direct limit).
20
(5) For limit ordinals not in M, P
satisfy the 23
obvious requirement. (Note that we can always nd some suitable Q
itself.)
25
Proof. We have to show (by induction) that the resulting sequence
P is a partial CS iteration, and that
P
M
26
embeds into
P. For successor cases, there is nothing to do. So assume that is a limit. If P
is a direct 27
limit, it is trivially a partial CS limit; if P
: P
M
is 30
M-complete. This was the main claim in Lemma 4.14.
31
The following lemma is natural and easy.
32
Lemma 4.16. Assume that we construct an almost FS iteration
P over
P
M
where each Q
is (forced to be) 33
ccc. Then P
is ccc. For of 35
uncountable conality, we know that we took the direct limit coboundedly often (and all P
(p) for p P
M
and q
_
<
P
. We can thin out the set A such that p are the same and all q 40
are in the same P
45
are nontrivial (e.g., only those with M) and where each Q
is 46
-centered as well. 47
Proof. By induction: The direct limit of countably many -centered forcings is -centered, as is the almost 48
FS limit of -centered forcings (to color q i
over (
P
n
)
n
as follows: Conditions in P
(p) where n , p P
n
, and q P
for some M
n
. Then P
is an M
n
[H
n
to be ccc, then
P is ccc; this is 14
the case if we set Q
.
15
Proof. We can dene P
by induction. If
_
n
M
n
has countable conality, then we use the almost 16
FS limit as in Lemma 4.18. Otherwise we use the direct limit. If M
n
has uncountable conality, 17
then
sup( M) is an element of M
n+1
. In our induction we have already considered
and have 18
dened P
: P
n
is M
n
-complete.
20
Secondly, we can start the almost FS iteration after some
0
(i.e.,
P is already given up to
0
, and we 21
can continue it as an almost FS iteration up to ), and get the same properties that we previously showed 22
for the almost FS iteration, but this time for the quotient P
/P
0
. In more detail:
23
Lemma 4.20. Assume that
P
M
is in M a (topped) partial CS iteration of length , and that
P is in V a 24
topped partial CS iteration of length
0
over
P
M
0
for some
0
M. Then we can extend
P to a 25
(topped) partial CS iteration of length over
P
M
, as in the almost FS iteration (i.e., using the almost FS 26
limit at limit points >
0
with M of countable conality; and the direct limit everywhere else). We 27
can use any Q
for
0
(provided Q
M
is an M[H
M
]-complete subforcing of Q
). If all Q
0
forces that P
/H
0
is ccc (in particular proper); if moreover all Q
0
forces that P
/H
0
is -centered.
30
4.D. Almost countable support iterations. Almost countable support iterations
P (over a given itera- 31
tion
P
M
in a candidate M) will have the following two crucial properties: There is a canonical M-complete 32
embedding of
P
M
into
P, and
P preserves a given random real (similar to the usual countable support 33
iterations).
34
Denition and Claim 4.21. Let
P
M
be a topped partial CS iteration in M of length . We can construct by 35
induction on + 1 the almost countable support iteration
P over
P
M
(or: almost CS iteration):
36
(1) As induction hypothesis, we assume that the canonical embedding i
sup( + 1 : M).
Note that
.
39
(2) Let = + 1. We can choose any desired forcing Q
is 40
an M[H
M
]-complete subforcing of Q
. This denes P
.
41
(3) Let cf() > . Then P
= = ). We dene P
= P
43
as the union of the following two sets:
44
The almost FS limit of (P
, Q
)
<
, see Denition 4.13.
1
33
Or only: M
n
0
for some n
0
.
34
So for successors M, we have
is as in Denition 4.9.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 31
The set P
gen
of M-generic conditions q P
CS
i
1
[H
P
CS
] P
M
is M-generic.
(5) Let cf() = and assume that M but M is conal in , so
= < . We dene P
= P
3
as the union of the following two sets:
4
The direct limit of (P
, Q
)
<
.
5
The set P
gen
of M-generic conditions q P
CS
i
1
[H
P
CS
] P
M
is M-generic.
(Note that the M-generic conditions form an open subset of P
CS
= P
CS
.)
7
(6) Let cf() = and M not conal in (so M). Then P
, Q
)
<
8
(see Denition 4.5).
9
So the claim is that for every choice of Q
1
-name
for a condition in P
. Then there is a P
2
-name
p
2
for a condition in D such that the empty condition of P
2
forces: p
2
p
1
and:
if p
1
is in P
/H
2
, then the condition p
2
is as well.
The following easy fact will also be useful: 14
(4.24)
Let P be a subforcing of Q. We dene Pp r P : r p. Assume that p P
and Pp = Qp.
Then for any P-name
x) i p
Q
(
x).
We now prove by induction on the following statement (which includes that the Denition and 15
Claim 4.21 works up to ). Let ,
be as in (4.22).
16
Lemma 4.25. (a) The topped iteration
P of length is a partial CS iteration.
17
(b) The canonical embedding i
: P
M
: P
M
works.
18
(c) Moreover, assume that
19
M ,
20
p M is a P
M
-name of a P
M
-condition,
21
q P
forces (in P
) that
p[H
M
] is in H
M
.
22
Then there is a q
+
P
(and therefore in P
p[H
M
] is in H
M
.
23
Proof. First let us deal with the trivial cases. It is clear that we always get a partial CS iteration.
24
Assume that =
0
+ 1 M, i.e., =
= . It is clear that i
works. To get q
+
, rst extend q to 25
some q
0
(by induction hypothesis), then dene q
+
extending q
by q
+
(
0
)
p(
0
).
26
If =
0
+ 1 M, there is nothing to do.
27
Assume that cf() > (whether M or not). Then
< . So i
: P
M
works by induction, 28
and similarly (c) follows from the inductive assumption. (Use the inductive assumption for =
; 29
the that we got at that stage is the same as the current , and the q
+
we obtained at that stage will 30
still satisfy all requirements at the current stage.)
31
Assume that cf() = and that M is bounded in . Then the proof is the same as in the 32
previous case.
33
We are left with the cases corresponding to (4) and (5) of Denition 4.21: cf() = and M is conal 34
in . So either M, then
= = , or M, then
whenever q p. If p P
gen
is open in P
CS
and
p M are given, M .
39
Let (D
n
)
n
enumerate all dense sets of P
M
p
n
)
n
satisfying the following in M, for all n > 0:
3
p
0
=
p.
4
p
n
M is a P
M
n
-name of a P
M
-condition in D
n
.
5
P
M
n
p
n
P
M
p
n1
.
6
P
M
n
If
p
n1
n
H
M
n
, then
p
n
n
H
M
n
as well.
7
Using the inductive assumption for the
n
s, we can now nd a sequence (q
n
)
n
of conditions satisfying 8
the following:
9
q
0
= q, q
n
P
n
.
10
q
n
n1
= q
n1
.
11
q
n
P
n
p
n1
n
H
M
n
, so also
p
n
n
H
M
n
.
12
Let q
+
P
CS
p
n
n
H
M
n
, so also q
+
forces this. 14
(Using induction on n.)
15
(2) For all n and all m n: q
+
P
CS
p
m
m
H
M
m
, so also
p
n
m
H
M
m
. 16
(As
p
m
p
n
.)
17
(3) q
+
P
CS
p
n
H
M
. 18
(Recall that P
CS
p
n
) H
i i
n
(
p
m
) H
m
for all large m.)
19
As q
+
P
CS
p
n
D
n
H
M
, we conclude that q
+
P
gen
). In particular, P
gen
20
is dense in P
: Let q i
(p), then q
+
q i
(p).
22
It remains to show that i
, and p P
. Assume 23
towards a contradiction that p forces in P
that i
1
[H
is dense in P
, we can nd some q p in P
gen
. Let 25
P
r P
CS
: r q = r P
: r q,
where the equality holds because P
gen
is open in P
CS
.
26
Let be the canonical name for a P
. Let R be either P
CS
or P
. 27
We write ()
R
for the lter generated by in R, i.e., ()
R
:= r R : (r
) r
r. So 28
(4.26) q
R
H
R
= ()
R
.
We now see that the following hold:
29
q
P
i
1
[H
P
i
1
[()
P
i
1
[()
P
i
1
[()
P
CS
maps A into P
P
CS
, so 33
A i
1
[(Y)
P
] = A i
1
[(Y)
P
CS
i
1
[H
P
CS
, implies q P
gen
, a contradiction.
36
We can also show that the almost CS iteration of proper forcings Q
is 41
proper.
42
Proof. By induction on we prove that for all < the quotient P
/H
/H
n
.
3
(4.30)
If
P is an iteration of length with cf() > , and if all quotients P
/H
(for <
< ) are forced to be proper, then the direct limit P
/H
.
If is a successor, then our inductive claim easily follows from the inductive assumption together 4
with (4.28).
5
Let be a limit of countable conality, say = sup
n
n
. Dene an iteration
P
of length with 6
Q
n
P
n+1
/H
n
. (Each Q
n
is proper, by inductive assumption.) There is a natural forcing equivalence 7
between P
CS
and P
CS
.
8
Let N H(
) contain
P, P
,
P
, M,
P
M
. Let p P
. So below 9
p we can identify P
with P
CS
(of a code
and Q
M
M. Let Q
M
be an M-complete subforcing of Q. 14
We say that Q locally preserves randomness of r over M, if there is a sequence (D
Q
M
k
)
k
in M of open 15
dense subsets of Q
M
such that the following holds: 16
Assume that 17
M thinks that p (p
k
)
k
interprets
Z as Z
(so
Z is a Q
M
-name of a code for a null set and Z
is 18
a code for a null set, both in M),
19
moreover, each p
k
is in D
Q
M
k
(we call such a sequence (p
k
)
k
, or the according interpretation, 20
quick),
21
r is random over M (so there is a minimal c such that Z
c
r).
22
Then there is a q
Q
p
0
forcing that
23
r is random over M[G
M
],
24
Z
c
r.
25
Note that this is trivially satised if r is not random over M.
26
This property is a local variant of strong preservation of randomness (see Denition 2.27). For a 27
variant of this denition, see Section 7.
28
Lemma 4.32. If Q
M
is an ultralaver forcing in M and r a real, then there is an ultralaver forcing 29
Q over
35
Q
M
locally preserving randomness of r over M.
30
If Q
M
is a Janus forcing in M and r a real, then there is a Janus forcing Q over Q
M
(which is in 31
fact equivalent to random forcing) locally preserving randomness of r over M.
32
Proof. Assume that r is random over M (otherwise the claim is vacuous).
33
Ultralaver: This follows directly from Lemma 2.29, setting D
Q
M
n
to be the set of conditions with stem 34
of length at least n.
35
Janus: In this case, the notion of quick interpretations will be trivial, i.e., D
Q
M
k
= Q
M
for all k.
36
According to Lemma 2.43, we can rst nd some countable M
and 37
such that M
thinks that 2
M is countable. In M
the set Q
M
is now a countable Janus forcing; so we can 38
apply Lemma 3.16 to construct a Janus forcing Q
M
over Q
M
which is equivalent to random forcing. In V, 39
let Q be random forcing; since this is a Suslin ccc forcing we know that Q
M
is an M
-complete subforcing 40
of Q. Moreover, as was noted in Lemma 2.28, we know that Q strongly preserves randoms over M
.
41
So assume that (in M) the sequence (p
k
)
k
of Q
M
-conditions interprets
Z as Z
. Since (in M
), Q
M
42
is M-complete in Q
M
, also in M
the Q
M
-name
Z is interpreted as Z
c
r. So 43
by strong preservation of randoms, we can in V nd some q p
0
forcing that r is random over M
[H
M
] 1
35
Q over Q
M
just means that Q
M
is an M-complete subforcing of Q.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
34 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
(and therefore also over the subset M[H
M
]), and that
Z
c
r (where
Z can be evaluated in M
[H
M
] or 2
equivalently in M[H
M
]).
3
We will prove the following preservation theorem:
4
Lemma 4.33. Let
P be an almost CS iteration (of length ) over
P
M
, r random over M, and p P
M
. 5
Assume that each P
forces that Q
].
7
What we will actually need is the following variant:
8
Lemma 4.34. Assume that
P
M
is in M a topped almost CS iteration of length , and we already have some 9
topped partial CS iteration
P over
P
M
0
of length
0
M . Let
r be a P
0
-name of a random real 10
over M[H
M
0
]. Assume that we extend
P to length as an almost CS iteration
36
using forcings Q
which 11
locally preserve the randomness of
k
)
k
. Let p P
M
. Then we can 12
nd a q p in P
forcing that
].
13
Actually, we will only prove the two previous lemmas under the following additional assumption (which 14
is enough for our application, and saves some unpleasant work). This additional assumption is not really 15
necessary; without it, we could use the method of [GK06] for the proof.
16
Assumption 4.35. For each M , (P
M
forces that) Q
M
is either trivial
37
or adds a new - 17
sequence of ordinals. Note that in the latter case we can assume without loss of generality that 18
_
n
D
Q
M
n
= (and, of course, that the D
Q
M
n
are decreasing).
19
Moreover, we assume that already in M there is a set T such that P
forces: Q
is trivial i 20
T. (So whether Q
is trivial or not does not depend on the generic lter below , it is already 21
decided in the ground model.)
22
The result will follow as a special case of the following lemma, which we prove by induction on . 23
(Note that this is a rened version of the proof of Lemma 4.25 and similar to the proof of the preservation 24
theorem in [Gol93, 5.13].)
25
Denition 4.36. Under the assumptions of Lemma 4.34 and Assumption 4.35, let
Z be a P
-name,
0
26
< , and let p = (p
k
)
k
be a sequence of P
-names of conditions in P
/H
. Let Z
be a P
-name.
27
We say that ( p, Z
) is a quick interpretation of
Z if p interprets
Z as Z
(i.e., P
forces that p
k
forces 28
Zk = Z
, i.e., p
k
() D
Q
M
k
for all k.
31
It is easy to see that: 32
(4.37) For every name
).
Lemma 4.38. Under the same assumptions as above, let , ,
). 34
Assume that 35
M (= M ) and
0
(so <
),
36
p M is a P
M
-name of a P
M
-condition,
37
Z M is a P
M
M is a P
M
forces: p = (p
k
)
k
M is a quick sequence in P
M
/H
M
interpreting
Z as Z
(as in Deni- 40
tion 4.36),
41
P
M
forces: if p H
M
, then p
0
p,
42
q P
forces p H
M
,
1
36
Of course our ocial denition of almost CS iteration assumes that we start the construction at 0, so we modify this denition
in the obvious way.
37
More specically, Q
M
= .
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 35
q forces that r is randomover M[H
M
], so in particular there is a P
-name c
0
in V for the minimal c 2
with Z
c
r.
3
Then there is a condition q
+
P
,
5
r is random over M[H
M
],
6
c
r.
7
We actually claim a slightly stronger version, where instead of Z
and
] where q H
. 14
Note that M[H
M
] = M[H
M
]. So r is randomover M[H
M
], and (p
k
())
k
quickly interprets
Z as Z
in Q
M
. 15
Now let q
+
= q, and use the fact that Q
]. Let k
be maximal with 17
p
k
H
M
. Consider
Z as a Q
M
of
Z witnessed by 18
a sequence starting with p
k
(). In M[H
M
], Z
is now a P
M
/H
M
as Z
k.) 20
Using the induction hypothesis, we can rst extend q to a condition q
= M . We may assume
38
that 25
each Q
M
n
is nontrivial.
26
Let (
Z
n
)
n
be a list of all P
M
-names in M for codes for null sets (starting with our given null set
Z =
Z
0
). 27
Let (E
n
)
n
enumerate all open dense sets of P
M
Z
0
n, . . . ,
Z
n
n.
We write p
k
0
for p
k
, and Z
0,0
for Z
; as mentioned above,
Z =
Z
0
.
29
By induction on n we can now nd a sequence p
n
= (p
k
n
)
k
and P
M
n
-names Z
i,n
for i 0, . . . , n 30
satisfying the following:
31
(1) P
M
n
forces that p
0
n
p
k
n1
whenever p
k
n1
P
M
/H
M
n
.
32
(2) P
n
forces that p
0
n
E
n
. (Clearly E
n
P
M
/H
M
n
is a dense set.)
33
(3) p
n
M is a P
M
n
-name for a quick sequence interpreting (
Z
0
, . . . ,
Z
n
) as (Z
0,n
, . . . , Z
n,n
) (in P
M
/H
M
n
), 34
so Z
i,n
is a P
M
n
-name of a code for a null set, for 0 i n.
35
Note that this implies that the sequence (p
k
n1
n
) is (forced to be) a quick sequence interpreting (Z
0,n
, . . . , Z
n1,n
) 36
as (Z
0,n1
, . . . , Z
n1,n1
) .
37
Using the induction hypothesis, we now dene a sequence (q
n
)
n
of conditions q
n
P
n
and a sequence 38
(c
n
)
n
(where c
n
is a P
n
-name) such that (for n > 0) q
n
extends q
n1
and forces the following:
39
p
0
n1
n
H
M
n
.
40
Therefore, p
0
n
p
0
n1
.
41
r is random over M[H
M
n
].
42
Let c
n
be the least c such that Z
n,n
c
r.
43
Z
i,n
c
i
r for i = 0, . . . , n 1.
44
Now let q =
_
n
q
n
P
CS
. Moreover, by (4.39) we 45
get that q forces that
Z
i
= lim
n
Z
i,n
. Since each set C
c,r
:= x : x
c
r is closed, this implies that q forces 1
Z
i
c
i
r, in particular
Z =
Z
0
c
0
r.
38
If from some on all Q
M
= P
M
0
sup( M). Note that
0
and
0
used at stage
0
are the same as the current
and .
4
5. Tnr roacrxo coxsratcrrox 5
In this section we describe a -closed preparatory forcing notion R; the generic lter will dene a 6
generic forcing iteration
P, so elements of R will be approximations to such an iteration. In Section 6 we 7
will show that the forcing R P
2
forces BC and dBC.
8
From now on, we assume CH in the ground model.
9
5.A. Alternating iterations, canonical embeddings and the preparatory forcing R. The preparatory 10
forcing R will consist of pairs (M,
P), where M is a countable model and
P M is an iteration of ultralaver 11
and Janus forcings.
12
Denition 5.1. An alternating iteration
40
is a topped partial CS iteration
P of length
2
satisfying the 13
following:
14
Each P
is proper.
41
15
For even, either both Q
and Q
+1
are (forced by the empty condition to be) trivial,
42
or P
16
forces that Q
, and P
+1
forces that Q
+1
is a 17
Janus forcing based on
(where
is dened from
as in Lemma 2.22).
18
We will call an even index an ultralaver position and an odd one a Janus position.
19
As in any partial CS iteration, each P
2
) is a direct limit.
20
Recall that in Denition 4.10 we have dened the notion
P
M
canonically embeds into
P for nice 21
candidates M and iterations
P V and
P
M
M. Since our iterations now have length
2
, this means that 22
the canonical embedding works up to and including
43
2
.
23
In the following, we will use pairs x = (M
x
,
P
x
) as conditions in a forcing, where
P
x
is an alternating 24
iteration in the nice candidate M
x
. We will adapt our notation accordingly: Instead of writing M,
P
M
, P
M
25
H
M
, H
x
, Q
x
, etc. Instead of
P
x
canonically embeds 26
into
P we will say x canonically embeds into
P (which is a more exact notation anyway, since the test 27
whether the embedding is M
x
-complete uses both M
x
and
P
x
, not just
P
x
).
28
The following rephrases Denition 4.10 of a canonical embedding in our new notation, taking into 29
account that: 30
L
D
x
is an M
x
-complete subforcing of L
D
i
D extends
D
x
31
(see (5) of the ultralaver wish list 2.1).
32
Fact 5.2. x = (M
x
,
P
x
) canonically embeds into
P, if (inductively) for all
2
M
x
2
the following 33
holds: 34
Let = + 1 for alpha even (i.e., an ultralaver position). Then either Q
x
is trivial (and Q
can 35
be trivial or not), or we require that (P
]-ultralter system
D used for Q
36
extends the M
x
[H
x
]-ultralter system
D
x
used for Q
x
.
37
Let = +1 for alpha odd (i.e., a Janus position). Then either Q
x
38
forces that) the Janus forcing Q
x
is an M
x
[H
x
.
39
Let be a limit. Then the canonical extension i
: P
x
is M
x
-complete. (The canonical 40
extension was dened in Denition 4.9.)
41
Fix a suciently large regular cardinal
.
1
40
See Section 7 for possible variants of this denition.
41
This does not seem to be necessary, see Section 7, but it is easy to ensure and might be comforting to some of the readers and/or
authors.
42
For deniteness, let us agree that the trivial forcing is the singleton .
43
This is stronger than to require that the canonical embedding works for every
2
M, even though both P
2
and P
M
2
are
just direct limits; see footnote 32.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 37
Denition 5.3. The preparatory forcing R consists of pairs x = (M
x
,
P
x
) such that M
x
H(
) is a nice 2
candidate containing
2
, and
P
x
is in M
x
an alternating iteration (in particular topped and of length
2
). 3
We dene y to be stronger than x (in symbols: y
R
x), if the following holds: either x = y, or:
4
M
x
M
y
and M
x
is countable in M
y
. 5
M
y
thinks that x canonically embeds into
P
y
.
6
We will sometimes write i
x,y
for the canonical embedding (in M
y
) from P
x
2
to P
y
2
.
7
There are several variants of this denition which result in equivalent forcing notions. We will briey 8
come back to this in Section 7. 9
Note that the order on R is transitive. 10
The following is trivial by elementarity:
11
Fact 5.4. Assume that
P is an alternating iteration (in V), that x R canonically embeds into
P, and that 12
N H(
) contains x and
P. Let y = (M
y
,
P
y
) be the ord-collapse of (N,
P). Then y R and y x.
13
This fact will be used, for example, to get from the following Lemma 5.5 to Corollary 5.6.
14
Lemma 5.5. Given x R, there is an alternating iteration
P such that x canonically embeds into
P.
15
Proof. For the proof, we use either of the partial CS constructions introduced in the previous chapter (i.e., 16
an almost CS iteration or an almost FS iteration over
P
x
). The only thing we have to check is that we can 17
indeed choose Q
that satisfy the denition of an alternating iteration (i.e., as ultralaver or Janus forcings) 18
and such that Q
x
is M
x
-complete in Q
.
19
In the ultralaver case we arbitrarily extend
D
x
to an ultralter system
D, which is justied by ultralaver 20
wish list 2.1 item (5).
21
In the Janus case, we take Q
Q
x
(this works by Janus wish list 3.1 item (3). Alternatively, we could 22
extend Q
x
29
is (forced to be) the union of the Q
x
n
0
= j
y
0
).
34
M
x
[
0
,
2
) [
0
,
1
).
35
M
y
[
0
,
2
) [
1
,
2
).
36
Then there is an alternating iteration
P such that both x and y canonically embed into it.
37
Proof. For (1), use an almost FS iteration. We only use the coordinates in M, use the (countable!) Janus 38
forcings Q
Q
x
will be 39
-centered, by Lemma 4.17.
40
For (2), use the almost FS iteration over the sequence (x
n
)
n
as in Corollary 4.19, and at Janus positions 41
set Q
is M
x
n
-complete in Q
), there is a y R 5
such that 6
y x
n
for all n,
7
M
y
contains b and the sequence x,
8
for all Janus positions , M
y
thinks that the Janus forcing Q
y
,
10
for all limits , M
y
thinks that P
y
)
<
).
11
Proof. Item (4) directly follows from Lemma 5.7(2) and Fact 5.4. Item (1) is a special case of (4), and (2) 12
and (3) are trivial consequences of (1).
13
Another consequence of Lemma 5.7 is:
14
Lemma 5.9. The forcing notion R is
2
-cc.
15
Proof. Recall that we assume that V (and hence V[G]) satises CH.
16
Assume towards a contradiction that (x
i
: i <
2
) is an antichain. Using CH we may without loss of 17
generality assume that for each i
2
the transitive collapse of (M
x
i
,
P
x
i
) is the same. Set L
i
M
x
i
2
. 18
Using the -lemma we nd some uncountable I
2
such that the L
i
for i I forma -system with root L. 19
Set
0
= sup(L) + 3. Moreover, we may assume sup(L
i
) < min(L
j
\
0
) for all i < j.
20
Now take any i, j I, set x x
i
and y x
j
, and use Lemma 5.7(3). Finally, use Fact 5.4 to nd 21
z x
i
, x
j
.
22
5.B. The generic forcing P
2
of the directed system G: Formally, 24
we set 25
P
2
(x, p) : x G and p P
x
2
,
and we set (y, q) (x, p) if y
R
x and q is (in y) stronger than i
x,y
(p) (where i
x,y
: P
x
2
P
y
2
is the 26
canonical embedding). Similarly, we dene for each 27
P
:= (x, p) : x G, M
x
and p P
x
with x G is called P
.
30
Formally, elements of P
2
are dened as pairs (x, p). However, the x does not really contribute any 31
information. In particular:
32
Fact 5.11. (1) Assume that (x, p
x
) and (y, p
y
) are in P
2
, that y x, and that the canonical embedding 33
i
x,y
witnessing y x maps p
x
to p
y
. Then (x, p
x
) =
(y, p
y
).
34
(2) (y, q) is in P
2
stronger than (x, p) i for some (or equivalently: for any) z x, y in G the canoni- 35
cally embedded q is in P
z
2
stronger than the canonically embedded p. The same holds if stronger 36
than is replaced by compatible with or by incompatible with.
37
(3) If (x, p) P
(x, p).
38
In the following, we will therefore often abuse notation and just write p instead of (x, p) for an element 39
of P
.
40
We can dene a natural restriction map from P
2
to P
2
we have (x, p) =
(y, p).
43
Fact 5.12. The following is forced by R:
44
P
for <
2
(witnessed by the natural restriction map).
45
If x G, then P
x
is M
x
-completely embedded into P
for
2
(by the identity map p (x, p)).
1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 39
If cf() > , then P
for < .
2
By denition, P
2
is a subset of V.
3
G will always denote an R-generic lter, while the P
2
-generic lter over V[G] will be denoted by H
2
4
(and the induced P
-generic by H
2
into P
2
(and of P
x
into P
). This way H
induces an M
x
-generic lter H
x
P
x
.
6
So x R forces that P
is approximated by P
x
. In particular we get:
7
Lemma 5.13. Assume that x R,
2
in M
x
, p P
x
, and is a P
x
-name in M
x
. Then M
x
= p
P
x
( ) 8
i x
R
(x, p)
P
M
x
[H
x
] = ( [H
x
]).
9
Proof. is clear. So assume that ( ) is not forced in M
x
. Then some q
P
x
), while P
x
-names or P
-names (e.g.,
).
18
We rst prove a few lemmas that are easy generalizations of the following straightforward observation:
19
Assume that x
R
(
z,
p) P
. In particular, x
z 20
and
p to be z
and p
. Then x
1
. By denition, this 21
means that z
-condition p
can be interpreted as a 22
P
y
-condition p
such that y
R
(
z,
p) =
(y, p
).
23
Since Ris -closed, we can immediately generalize this to countably many (R-names for) P
-conditions:
24
Fact 5.14. Assume that x
R
p
n
P
n
P
y
such that 25
y
R
p
n
=
n
for all n .
26
Recall that more formally we should write: x
R
(
z
n
,
p
n
) P
; and y
R
(
z
n
,
p
n
) =
(y, p
n
).
27
We will need a variant of the previous fact:
28
Lemma 5.15. Assume that P
r
n
is a P
-name for 29
a real (or an HCON object) for every n . Then there is a y x and there are P
y
-names r
n
in M
y
such 30
that y
R
(
P
r
n
= r
n
) for all n.
31
(Of course, we mean:
r
n
is evaluated by G H
, while r
n
is evaluated by H
y
.)
32
Proof. The proof is an obvious consequence of the previous fact, since names of reals in a ccc forcing can 33
be viewed as a countable sequence of conditions.
34
In more detail: For notational simplicity assume all
r
n
are names for elements of 2
. Working in V, we 35
can nd for each n, m names for a maximal antichain
A
n,m
and for a function
f
n,m
:
A
n,m
2 such that 36
x forces that (P
forces that)
r
n
(m) =
f
n,m
(a) for the unique a
A
n,m
H
. Since P
is ccc, each
A
n,m
is 37
countable, and since R is -closed, it is forced that the sequence
= (
A
n,m
,
f
n,m
)
n,m
is in V.
38
In V, we strengthen x to x
1
to decide
to be some
M
x
1
(see 39
Corollary 5.6). Each A
n,m
consists of countably many a such that x
1
forces a P
. So in M
y
, we can use (A
n,m
, f
n,m
)
n,m
to construct P
y
-names r
n
in the obvious way.
42
Now assume that y G and that H
is P
n,m
=
A
n,m
. Since a P
y
, we 43
get a H
y
i a H
n,m
H
y
, and r
n
(m) = f
n,m
(a) =
f
n,m
(a) =
r
n
(m).
44
We will also need the following modication:
45
Lemma 5.16. (Same assumptions as in the previous lemma.) In V[G][H
], set Q
be the union of Q
z
[H
z
] 46
for all z G. In V, assume that x forces that each
r
n
is a name for an element of Q
. Then there is a y x 47
and there is in M
y
a sequence ( r
n
)
n
of P
y
r
n
= r
n
for all n.
1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
40 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
So the dierence to the previous lemma is: We additionally assume that
r
n
is in
_
zG
Q
z
, and we 2
additionally get that r
n
is a name for an element of Q
y
.
3
Proof. Assume x G and work in V[G]. Fix n. P
-name 4
n
M
y
n
of an element of Q
y
n
such that
r
n
(evaluated by H
) is the same as
n
(evaluated by H
y
n
). Since 5
we assume that P
forces y
n
Y
n
. (As R is -closed and Y
n
R V, we must have Y
n
V.)
7
So in V, there is (for each n) an R-name
Y
n
for this countable set. Since R is -closed, we can nd 8
some z
0
x deciding each
Y
n
to be some countable set Y
n
R. In particular, for each y Y
n
we know 9
that z
0
R
y G, i.e., z
0
n
. Let X contain all M
y
such that for some y
_
n
Y
n
, is a P
y
-name for a 11
Q
y
.
12
So X is a set of P
z
-names for Q
z
r
n
X for 13
all n. Using Lemma 5.15, we can additionally assume that there are names P
z
-name r
n
in M
z
such that z 14
forces that
r
n
= r
n
is forced for each n. By Lemma 5.13, we know that M
z
thinks that P
z
forces that r
n
X. 15
Therefore r
n
is a P
z
-name for a Q
z
-element.
16
We now prove by induction on that P
, Q
)
<
19
with limit P
that satises the denition of alternating iteration (up to ), and there is a naturally 20
dened dense embedding j
: P
that maps p to j
-generic lter H
24
(and vice versa).
25
(3) P
is ccc.
27
(5) P
forces CH.
28
Proof. = 0 is trivial (since P
0
and P
0
both are trivial: P
0
is a singleton, and P
0
consists of pairwise 29
compatible elements).
30
So assume that all items hold for all < .
31
Proof of (1).
32
Ultralaver successor case: Let = +1 with an ultralaver position. Let H
be P
denes a P
x
-generic lter 34
over M
x
called H
x
.
35
Denition of Q
(and thus of P
): In M
x
[H
x
is dened as L
D
x for some system 36
of ultralters
D
x
in M
x
[H
x
]. Fix some s
<
. If y x in G, then D
y
s
extends D
x
s
. Let D
s
be the union 37
of all D
x
s
with x G. So D
s
is a proper lter. It is even an ultralter: Let r be a P
-name
r
y
M
y
such that (in V[G][H
]) we 39
have
r
y
[H
y
] = r. So r M
y
[H
y
to be the P
-name for L
D
. (Note that P
forces that Q
literally is 42
the union of the Q
x
[H
x
: Let (x, p) be in P
. If p P
x
, then we set j
(x, p) = j
will extend j
. If 44
p = (p, p()) is in P
x
but not in P
x
, we set j
(x, p) = (r, s) P
where r = j
]. From Q
=
_
xG
Q
x
[H
x
-name in M
y
can be canonically interpreted as a P
z
-name in M
z
, and Q
y
is
(forced to be) a subset of Q
z
.
46
I.e., j
(x, p) = j
(x, p) = j
(x, p).
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 41
The canonical embedding: By induction we know that i
x,
which maps p P
x
to j
2
. So we have to extend the canonical embedding to 3
i
x,
: P
x
to the pair (i
x,
(p), p()). This 4
is the same as j
is just dened as 7
the union of all Q
x
[H
x
] for x G. We will show below that this union satises the ccc; just as in Fact 3.8, 8
it is then easy to see that this union is again a Janus forcing.
9
In particular, Q
consists of hereditarily countable objects (since it is the union of Janus forcings, which 10
by denition consist of hereditarily countable objects). So since P
forces CH, Q
and j
: First we dene j
: P
P
CS
, let j
(x, p) P
be 15
the union of all j
(x, p) (for M
x
). (Note that
1
<
2
implies that j
1
(x, p
1
) is restriction of 16
j
2
(x, p
2
), so this union is indeed an element of P
CS
.)
17
P
[P
], q P
is dense. We will 19
show below that P
for y G (without 22
loss of generality y x), and each i
x,y
is M
x
-complete (see Fact 5.12).
23
Proof of (3).
24
Recall that we assume CH in the ground model.
25
The successor case, = + 1, follows easily from (3)(5) for P
(since P
forces that Q
has size 26
2
0
=
1
=
V
1
).
27
If cf() > , then P
=
_
<
P
are of size
1
.
30
For notational simplicity we assume
P
1
for all < (this is justied by inductive assump- 31
tion (5)). By induction, we can assume that for all < there is a dense P
of size
1
and that every 32
P
p() =
p
(a). Writing
47
p q if p q and q p, the 34
map p (A
p
,
p
)
dom(p)
is 1-1 modulo . Since each A
p
. I.e., we can nd a
-dense subset P
of cardinality 37
1
. Note that a conditions (x, p) P
-equivalent condition (x
, p
). Since 39
=
1
, there are only
1
0
= 2
0
many such candidates x
is countable and p
, 40
there are only 2
0
many pairs (x
, p
).
41
Proof of (4).
42
Ultralaver successor case: Let = + 1 with an ultralaver position. We already know that P
= 43
P
where Q
is ccc.
44
Janus successor case: As above it suces to show that Q
[H
x
] for 45
x G, is (forced to be) ccc.
1
47
Since is separative, p q i p =
and that M
x
are subsets of +
1
together with the interval [
2
,
2
+
1
]; i.e., there are
1
many ordinals that can possibly occur in M
x
, and
therefore there are 2
0
many possible such candidates. Moreover, setting p
(x, p) (similarly
to Fact 5.11).
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
42 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
Assume towards a contradiction that this is not the case, i.e., that we have an uncountable antichain 2
in Q
has size
1
and therefore the uncountable antichain has size
1
. So, 3
working in V, we assume towards a contradiction that 4
(5.18) x
0
R
p
0
P
a
i
: i
1
is a maximal (uncountable) antichain in Q
.
We construct by induction on n a decreasing sequence of conditions such that x
n+1
satises the 5
following:
6
(i) For all i
1
M
x
n
there is (in M
x
n+1
) a P
x
n+1
-name a
i
for a Q
x
n+1
a
i
= a
i
.
Why can we get that? Just use Lemma 5.16.
8
(ii) If is in M
x
n
a P
x
n
, then there is k
()
1
such that 9
x
n+1
R
p
0
P
(i < k
())
a
i
j
P
.
Also, all these k
() are in M
x
n+1
. 10
Why can we get that? First note that x
n
p
0
(i
1
)
a
i
j . Since P
is ccc, x
n
forces that there 11
is some bound
k() to be k
1
_
n
M
x
n
. By Corollary 5.8(4), there is some y such that
14
y x
n
for all n ,
15
(x
n
)
n
and ( a
i
)
i
are in M
y
,
16
(M
y
thinks that) P
y
forces that Q
y
is the union of Q
x
n
.
17
Let G be R-generic (over V) containing y, and let H
be P
i
[H
y
] : i <
. Note that A
is in M
y
[H
y
]. We claim 19
(5.19) A
Q
y
[H
y
] is predense.
Pick any q
0
Q
y
-name of a Q
x
n
-condition, such 20
that q
0
= [H
x
n
]. By (ii) above, x
n+1
and therefore y forces (in R) that for some i < k
() (and therefore 21
some i <
) the condition p
0
forces the following (in P
):
22
The conditions
a
i
and are compatible in Q
. Also,
a
i
= a
i
and both are in Q
y
, and Q
y
23
is an incompatibility-preserving subforcing of Q
. Therefore M
y
[H
y
] thinks that a
i
and 24
are compatible.
25
This proves (5.19).
26
Since Q
y
[H
y
] is M
y
[H
y
]-complete in Q
[H
], and since A
M
y
[H
y
], this implies (as a
i
[H
y
] =
a
i
[G 27
H
) that
a
i
[G H
] : i <
has a 30
dense subset of size
1
(modulo =
a
i
: i
1
is a maximal (uncountable) antichain in P
.
So each
a
i
is an R-name for an HCON object (x, p) in V.
33
To lighten the notation we will abbreviate elements (x, p) P
)-names
(, p) and
be P
/H
. If p is not in R, set 36
(, p) =
a
i
R and
a
i
and p are compatible 37
(in R), and set
(, p)
1
containing all possibilities for
(, p) and similarly
X
r
(, p) 39
consisting of HCON objects for
r(, p).
40
To summarize: For every < and every HCON object p, we can dene (in V) the R-names
(, p) 41
and
X
r
(, p) such that 1
(5.21) x
0
R
P
_
p P
/H
(i
(, p)) (r
X
r
(, p)) r
P
/H
p,
a
i
_
.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 43
Similarly to the Janus successor case, we dene by induction on n a decreasing sequence of con- 2
ditions such that x
n+1
satises the following: For all M
x
n
and p P
x
n
, x
n+1
decides
(, p) and 3
X
r
(, p) to be some X
(, p) and X
r
(, p). For all i
1
M
x
n
, x
n+1
decides
a
i
to be some a
i
P
x
n+1
. 4
Moreover, each such X
and X
r
is in M
x
n+1
, and every r X
r
(, p) is in P
x
n+1
1
_
n
M
x
n
, and set A
i
: i
are in M
y
,
9
(M
y
thinks that) P
y
is predense in P
y
.
Then P
y
is M
y
-completely embedded into P
, and since A
M
y
(and since
a
i
= a
i
for all i
) we get 12
that
a
i
: i
; we will nd 14
i <
i
. Since P
y
, q P
y
and q p.
16
Now let H
be P
]. Since q p, we get p P
/H
. Let
17
be the evaluation by G H
of
be the evaluation of
<
and r
P
y
. 18
So we know that a
/H
witnessed by r
. Find q
forcing r
/H
p, a
. 19
We may nd q
q. Now q
are compatible in P
y
.
20
Proof of (5).
21
This follows from (3) and (4).
22
5.D. The generic alternating iteration
P. In Lemma 5.17 we have seen:
23
Corollary 5.23. Let G be R-generic. Then we can construct
49
(in V[G]) an alternating iteration
P such 24
that the following holds:
25
P is ccc. 26
If x G, then x canonically embeds into
P. (In particular, a P
2
-generic lter H
2
induces a 27
P
x
2
-generic lter over M
x
, called H
x
2
.)
28
Each Q
[H
x
] with x G.
29
P
2
is equivalent to the direct limit P
2
of G: There is a dense embedding j : P
2
P
2
, and for 30
each x G the function p j(x, p) is the canonical embedding.
31
Lemma 5.24. Let x R. Then R forces the following: x G i x canonically embeds into
P
2
.
32
Proof. If x G, then we already know that x canonically embeds into
P.
33
So assume (towards a contradiction) that y forces that x embeds, and that x y (i.e., y x G). Work 34
in V[G] where y G. Both x (by assumption) and y G canonically embed into
P. Let N be an elementary 35
submodel of H
V[G]
(
) containing x, y,
P; let z = (M
z
,
P
z
) be the ord-collapse of (N,
P). Then z V (as R 36
is -closed) and z R, and (by elementarity) z x, y. This shows that x j
R
y, i.e., y cannot force x G, a 37
contradiction. 38
Using ccc, we can nowprove a lemma that is in fact stronger than the lemmas in the previous section 5.C:
39
Lemma 5.25. The following is forced by R: Let N H
V[G]
(
2
, say B = k(A) where A N is a maximal antichain of P
2
. So (by ccc) A is countable, hence A N. 46
So not only A = k
1
(B) but even A = k
1
[B]. Hence k
1
is an M
y
-complete embedding.
1
49
in an absolute way: Given G, we rst dene P
2
to be the direct limit of G, and then inductively construct P
from P
2
.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
44 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
Remark 5.26. Using this lemma we can easily reprove the countable chain condition of P
2
and in fact all 2
other lemmas in section 5.C. We will do this for Fact 5.14 in Corollary 6.4.
3
6. Tnr raoor or BC+dBC 4
We rst prove that no uncountable X in V will be smz or sm in the extension.
50
Then we show how to 5
modify the argument to work for all uncountable sets in V[G H].
6
6.A. BC+dBC for ground model sets. A schematic diagramillustrating the arguments in this section can 7
be found in Figure 2 on page 8.
8
Lemma 6.1. Let X V be an uncountable set of reals. Then R P
2
forces that X is not smz.
9
Proof.
10
(1) Fix any even <
2
(i.e., an ultralaver position) in our iteration. The ultralaver forcing Q
adds a 11
(canonically dened code for a) closed null set
F constructed from the ultralaver real
. (Recall 12
the ultralaver wish list 2.1, item (2) and item (3). In the following, when we consider various 13
forcings Q
, Q
, Q
x
, we treat
F not as an actual name, but rather as a denition which depends on 14
the forcing used.)
15
(2) According to Theorem 1.2, it is enough to show that X +
F is non-null in the R P
2
-extension, or 16
equivalently, in every R P
-name
2
18
forces that X +
F
Z.
19
(3) Using the dense embedding j
2
: P
2
P
2
, we may replace (x, p) by a condition (x, p
) RP
2
. 20
According to Fact 5.14 (recall that we now know that P
2
satises ccc) and Lemma 5.15 we can 21
assume that p
is already a P
x
-condition p
x
and that
-name 22
Z
x
in M
x
. 23
(4) We construct (in V) an iteration
P in the following way:
24
(a) Up to , we take an arbitrary alternating iteration into which x embeds. In particular, P
will 25
be proper and hence forces that X is still uncountable.
26
(b) Let Q
in case M
x
). So according to item (2) of the 27
ultralaver wish list 2.1, we know that Q
forces that X +
F is not null.
28
Therefore we can pick (in V[H
+1
]) some r in X +
F which is random over (the countable 29
model) M
x
[H
x
+1
], where H
x
+1
is induced by H
+1
.
30
(c) In the rest of the construction, we preserve randomness of r over M
x
[H
x
] for each
2
. We 31
can do this using an almost CS iteration over x where at each Janus position we use a random 32
version of Janus forcing and at each ultralaver position we use a suitable ultralaver forcing; 33
this is possible by Lemma 4.32. By Lemma 4.34, this iteration will preserve the randomness 34
of r. 35
(d) So we get
P over x (with canonical embedding i
x
) and q
P
2
i
x
(p
x
) such that q forces (in 36
P
], in particular that r
Z
x
.
37
We now pick a countable N H(
41
be P
] thinks that r
Z
x
(which is absolute) and 42
that r = x + f for some x X
y
X and f F (actually even in F as evaluated in M
y
[H
y
+1
]). So 43
in V[G][H
) forces that 44
r X +
F \
Z, a contradiction to (2).
45
Of course, we need this result not just for ground model sets X, but for R P
2
-names
X = (
x
i
: i
1
) 46
of uncountable sets. It is easy to see that it is enough to deal with R P
X is actually an RP
x
i
) into
P-names in a suciently absolute way. In other words: 3
It is not clear how to end up with some M
y
and
X
y
uncountable in M
y
such that it is guaranteed that
X
y
4
(evaluated in M
y
[H
y
]) will be a subset of
X (evaluated in V[G][H
2
forces that X is not strongly meager.
8
Proof. The proof is parallel to the previous one:
9
(1) Fix any even <
2
(i.e., an ultralaver position) in our iteration. The Janus forcing Q
+1
adds a 10
(canonically dened code for a) null set
Z
= 2
in the R P
2
-extension, or equivalently, 12
in every R P
, i.e., 13
(x X) x + r
Z
-name
2
forces that
r X +
Z
.
15
(3) Again, we can assume that
r is a P
x
-name r
x
in M
x
.
16
(4) We construct (in V) an iteration
P in the following way:
17
(a) Up to , we take an arbitrary alternating iteration into which x embeds. In particular, P
again 18
forces that X is still uncountable. 19
(b1) Let Q
). Then Q
= 2
for all M
x
; see 25
Lemma 4.17. 26
(d) So P
= 2
30
be P
(this time, Z
is evaluated in M
y
[H
y
<
2
in the obvious way: Conditions are pairs x = 33
(M
x
,
P
x
) of nice candidates M
x
(containing
(and not
2
). We call this variant R
.
35
Note that all results of Section 5 about R are still true for R
. In particular, whenever G R
is 36
generic, it will dene a direct limit (which we call P
(called P
); 37
again we will have that x G i x canonically embeds into
P
.
38
There is a natural projection map from R (more exactly: from the dense subset of those x which satisfy 39
M
x
) into R
, mapping x = (M
x
,
P
x
) to x
(M
x
,
P
x
in R
, then let M
z
be a model containg both x and y. In M
z
, we can rst 45
construct an alternating iteration of length
(R/R
), where R/R
, i.e., the (R
) implies: 1
(6.3) R/R
) into P
.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
46 MARTIN GOLDSTERN, JAKOB KELLNER, SAHARON SHELAH, AND WOLFGANG WOHOFSKY
Setup. Fix some
<
2
of uncountable conality.
51
Let G
be R
V[G
]. Set
P
= (P
)
<
, the generic alternating iteration added by R
. Let R
be the quotient 3
R/R
.
4
We claim that R
satises (in V
) all the properties that we proved in Section 5 for R (in V), with the 5
obvious modications. In particular:
6
(A)
is
2
-cc, since it is the quotient of an
2
-cc forcing.
7
(B)
does not add new reals (and more generally, no new HCON objects), since it is the quotient of a 8
-closed forcing.
52
9
(C)
Let G
be R
-generic over V
. Then G
. (Note that P
2
and then P
2
is constructed from G
. 12
(D)
: Let N H
V[G
]
(
) 13
be countable, and let y be the ord-collapse of (N,
P). Then y G
. Moreover: If x G
N, then 14
y x.
15
We can use the last item to prove the R
forces that p P
2
. Then there is a y x and a p
y
P
y
2
such that y 18
forces p
y
=
p.
19
Proof. Let G
contain x. In V[G
. So y 21
and p
y
is as required, where p
y
P
y
2
is the canonical image of p
z
.
22
We nowclaimthat RP
2
forces BC+dBC. We knowthat Ris forcing equivalent to R
. Obviously 23
we have 24
R P
2
= R
,
2
(where P
,
2
is the quotient of P
2
and P
). Note that P
is already determined by R
, so R
is 25
(forced by R
to be) a product R
= P
.
26
But note that this is not the same P
in the P
-extension 27
of V[G
. In 28
other words, we can unfortunately not just argue as follows:
29
Wrong argument. R P
2
is the same as (R
) (R
,
2
); so given an R P
2
-name X of a 30
set of reals of size
1
, we can choose
X =
i
: i
1
is an R P
2
-name for a set of reals of 33
size
1
. So there is a <
2
such that
X is added by R P
(using
2
-cc of R). In the R-extension, 34
P
i
is a system of countably many countable antichains
A
m
i
35
of P
f
m
i
:
A
m
i
0, 1. For the following argument, we prefer to work with the 36
equivalent P
instead of P
A
m
i
,
f
m
i
)
m
is an element 37
of V (since P
with
.
39
From now on, we work in the R
-extension V
. So in V
-names.
42
Note that the
i
are forced (by R
.
1
51
Probably the conality is completely irrelevant, but the picture is clearer this way.
52
It is easy to see that R
is even -closed, by relativizing the proof for R, but we will not need this.
53
Let P
(for
), and P
. The function k
: P
by induction from P
2
yields
the same result as dening
P from P
2
.
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 47
So now we can just repeat the proofs of BC and dBC of Section 6.A, with the following modications 2
(the modications are the same for both proofs): We x the uncountable set
as 3
above, and work in the R
extension V
.
4
(1) Instead of any ultralaver position <
2
, we obviously have to choose an
.
5
(2) No change here (of course we now have an R
-name).
6
(3) Here, we use Corollary 6.4.
7
(4) The iteration obviously has to start with
P
X consists of absolute P
-generic lter G
, and if H is 10
P
2
-generic over V[G
], then
X
y
[H] =
X[H] M
y
X[H]. So y thinks that
X
y
is a name for an 11
uncountable set, and we know that this P
y
2
-name will be evaluated to a subset of
X.
12
(5) No change here.
13
7. A woao ox vxarxxrs or rnr orrrxrrroxs 14
The following is not needed for understanding the paper, we just briey comment on alternative ways 15
some notions could be dened. 16
7.A. Regarding alternating iterations. We call the set of
2
such that Q
). 21
This is equivalent, as conditions coming from elementary submodels are dense in our R, by 22
Fact 5.4. 23
While this denition looks much simpler and therefore nicer (we could replace ord-transitive mod- 24
els by the better understood elementary models), it would not make things easier and just hides 25
the point of the construction: For example, we use models M
x
that are (an ord-collapse of) an 26
elementary submodel of H
V
of V.
27
(2) Require that (M
x
thinks that) the true domain of
P
x
is
2
. 28
This is equivalent for the same reason as (1) (and this requirement is compatible with (1)). 29
This denition would allow to drop the trivial option from the denition, but it would make the 30
dBC argument more cumbersome, as an iteration with uncountably many nontrivial iterands Q
is 31
not -centered. 32
(3) Alternatively, require that (M
x
thinks that) the true domain of
P
x
is countable. 33
Again, equivalence can be seen as in (1), again (3) is compatible with (1) but obviously not with (2). 34
This requirement would not make the denition easier, so there is no reason to adopt it. It would 35
have the slight inconvenience that instead of using ord-collapses as in Fact 5.4, we would have to 36
put another model on top to make the iteration countable. Also, it would have the (purely aesthetic) 37
disadvantage that the generic iteration itself does not satisfy this requirement.
38
(4) Also, we could have dropped the requirement that the iteration is proper. It is never directly used, 39
and densely
P is proper anyway. (E.g., in Lemma 6.1(4a), we would just construct
P up to to 40
be proper or even ccc, so that X remains uncountable.)
41
7.B. Regarding almost CS iterations and separative iterands. Recall that in Denition 4.6 we required 42
that each iterand Q
in a partial CS iteration is separative. This implies the property (actually: the three 43
equivalent properties) from Fact 4.8. Let us call this property suitability for now. Suitability is a prop- 44
erty of the limit P
of
P. Suitability always holds for nite support iterations and for countable support 45
iterations. However, if we do not assume that each Q
Q
).
4
7.C. Regarding preservation of random and quick sequences. Recall Denition 4.31 of local preser- 5
vation of random reals and Lemma 4.32. 6
In some respect the dense sets D
n
are unnecessary. For ultralaver forcing L
D
, the notion of a quick 7
sequence refers to the sets D
n
of conditions with stem of length at least n.
8
We could dene a new partial order on L
D
as follows: 9
q
) is nownot separative any more. Therefore we chose not to take this approach, 12
since losing separativity causes technical inconvenience, as described in 7.B.
13
Rrrrarxcrs 14
[BJ95] Tomek Bartoszy nski and Haim Judah. Set theory. A K Peters Ltd., Wellesley, MA, 1995. On the structure of the real line.
15
[Bor19] E. Borel. Sur la classication des ensembles de mesure nulle. Bull. Soc. Math. France, 47:97125, 1919.
16
[BS03] Tomek Bartoszynski and Saharon Shelah. Strongly meager sets of size continuum. Arch. Math. Logic, 42(8):769779, 17
2003. 18
[BS10] Tomek Bartoszynski and Saharon Shelah. Dual Borel conjecture and Cohen reals. J. Symbolic Logic, 75(4):12931310, 19
2010. 20
[Car93] Timothy J. Carlson. Strong measure zero and strongly meager sets. Proc. Amer. Math. Soc., 118(2):577586, 1993.
21
[GK06] Martin Goldstern and Jakob Kellner. New reals: can live with them, can live without them. MLQ Math. Log. Q., 52(2):115 22
124, 2006.
23
[GMS73] Fred Galvin, Jan Mycielski, and Robert M. Solovay. Strong measure zero sets. Notices of the AMS, pages A280, 1973.
24
[Gol93] Martin Goldstern. Tools for your forcing construction. In Set theory of the reals (Ramat Gan, 1991), volume 6 of Israel 25
Math. Conf. Proc., pages 305360. Bar-Ilan Univ., Ramat Gan, 1993.
26
[Jec03] Thomas Jech. Set theory. Springer Monographs in Mathematics. Springer-Verlag, Berlin, 2003. The third millennium 27
edition, revised and expanded.
28
[JS90] Haim Judah and Saharon Shelah. The Kunen-Miller chart (Lebesgue measure, the Baire property, Laver reals and preser- 29
vation theorems for forcing). J. Symbolic Logic, 55(3):909927, 1990.
30
[Kel] Jakob Kellner. Non-elementary proper forcing. preprint, http://arxiv.org/abs/0910.2132.
31
[KS05] Jakob Kellner and Saharon Shelah. Preserving preservation. J. Symbolic Logic, 70(3):914945, 2005.
32
[Lav76] Richard Laver. On the consistency of Borels conjecture. Acta Math., 137(3-4):151169, 1976.
33
[Paw96a] Janusz Pawlikowski. A characterization of strong measure zero sets. Israel J. Math., 93:171183, 1996.
34
[Paw96b] Janusz Pawlikowski. Lavers forcing and outer measure. In Set theory (Boise, ID, 19921994), volume 192 of Contemp. 35
Math., pages 7176. Amer. Math. Soc., Providence, RI, 1996.
36
[She98] Saharon Shelah. Proper and improper forcing. Perspectives in Mathematical Logic. Springer-Verlag, Berlin, second edi- 37
tion, 1998.
38
[She04] S. Shelah. Properness without elementaricity. J. Appl. Anal., 10(2):169289, 2004.
39
[She06] S. Shelah. Non-Cohen oracle C.C.C. J. Appl. Anal., 12(1):117, 2006.
40
[She10] Saharon Shelah. Large continuum, oracles. Cent. Eur. J. Math., 8(2):213234, 2010.
41
[Sie28] W. Sierpi nski. Sur un ensemble non d enombrable, dont toute image continue est de mesure nulle. Fund. Math., 11:302 42
304, 1928.
43
Ixsrrrtr r ta Drskarrr Mxrnrmxrrk txo Gromrrarr, Trcnxrscnr Uxrvrasrr xr Wrrx, Wrroxra Hxtrrsraxsr 810/104, 1040 Wrrx, 44
Atsrarx 45
E-mail address: martin.goldstern@tuwien.ac.at
46
URL: http://www.tuwien.ac.at/goldstern/
47
Ktar G oori Rrsrxacn Crxrra roa Mxrnrmxrrcxi Loorc, Uxrvrasrr xr Wrrx, W xnarxora Sraxsr 25, 1090 Wrrx, Atsrarx
48
E-mail address: kellner@fsmat.at 49
URL: http://www.logic.univie.ac.at/kellner/
50
Erxsrrrx Ixsrrrtrr or Mxrnrmxrrcs, Eomoxo J. Sxrax Cxmrts, Grvxr Rxm, Tnr Hraarw Uxrvrasrr. or Jratsxirm, Jratsxirm, 51
91904, Isaxri, xxo Drrxarmrxr or Mxrnrmxrrcs, Rtroras Uxrvrasrr., Nrw Batxswrck, NJ 08854, USA
52
E-mail address: shelah@math.huji.ac.il
53
URL: http://shelah.logic.at/
54
Ixsrrrtr r ta Drskarrr Mxrnrmxrrk txo Gromrrarr, Trcnxrscnr Uxrvrasrr xr Wrrx, Wrroxra Hxtrrsraxsr 810/104, 1040 Wrrx, 55
Atsrarx 1
9
6
9
r
e
v
i
s
i
o
n
:
2
0
1
1
-
0
5
-
2
8
m
o
d
i
f
i
e
d
:
2
0
1
1
-
0
5
-
2
9
BOREL CONJECTURE AND DUAL BOREL CONJECTURE 49
E-mail address: wolfgang.wohofsky@gmx.at
2
URL: http://www.wohofsky.eu/math/
2033