5

3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


The distributivity numbers of nite products of T()/n
Saharon Shelah
1
Department of Mathematics, Hebrew University, Givat Ram, 91904 Jerusalem,
ISRAEL
Otmar Spinas
2
Mathematik, ETH-Zentrum, 8092 Z urich, SWITZERLAND
ABSTRACT: Generalizing [ShSp], for every n < we construct a ZFC-model where the
distributivity number of r.o.(T()/n)
n+1
, h(n + 1), is smaller than the one of r.o.(T()/n)
n
.
This answers an old problem of Balcar, Pelant and Simon (see [BaPeSi]). We also show that both
of Laver and Miller forcing collapse the continuum to h(n) for every n < , hence by the rst result,
consistently they collapse it below h(n).
Introduction
For a cardinal let h() be the least cardinal for which r.o.(T()/n)

is not distributive, where

by (T()/n)

we mean the (full) product of T()/n in the forcing sense; so f (T()/n)

if and
only if f : T()/n 0, and the ordering is coordinatewise.
In [ShSp] the consistency of h(2) < h with ZFC has been proved, which provided a (partial) answer to
a question of Balcar, Pelant and Simon in [BaPeSi]. This inequality holds in a model obtained by forcing
with a countable support iteration of Mathias forcing over a model of GCH. The proof is long and dicult.
The following are the key properties of Mathias forcing (M.f.) which are essential to the proof (see [ShSp]
or below for precise denitions):
1
The author is supported by the Basic Research Foundation of the Israel Academy of Sciences; publication
531.
2
The author is supported by the Swiss National Science Foundation
1
5
3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


(1) M.f. factors into a -closed and a -centered forcing.
(2) M.f. is Suslin-proper which means that, rstly, it is simply denable, and, secondly, it permits generic
conditions over every countable model of ZF

.
(3) Every innite subset of a Mathias real is also a Mathias real.
(4) Mathias forcing does not change the conality of any cardinal from above h to below h.
(5) Mathias forcing has the pure decision property and it has the Laver property.
In this paper we present a forcing Q
n
, where 0 < n < , which is an n-dimensional version of M.f.
which satises all the analogues of the ve key properties of M.f. In this paper we only prove these. Once
this has been done the proof of [ShSp] can be generalized in a straightforward way, to prove the following:
Theorem. Suppose V [= ZFC + GCH. If P is a countable support iteration of Q
n
of length
2
and G
is P-generic over V , then V [G] [= h(n + 1) =
1
h(n) =
2
.
Besides the fact that the consistency of h(n+1) < h(n) was an open problem in [BaPeSi], our motivation
for working on it was that in [GoReShSp] it was shown that both of Laver and Miller forcing collapse the
continuum to h. Moreover, using ideas from [GoJoSp] and [GoReShSp] it can be proved that these forcings
do not collapse c below h(). We do not know whether they do collapse it to h(). But in 2 we show that
they collapse it to h(n), for every n < . Combining this with the rst result we conclude that for every
n < , consistently Laver and Miller forcing collapse c strictly below h(n).
The reader should have a copy of [ShSp] at hand. We do not repeat all the denitions from [ShSp] here.
Notions as Ramsey ultralter, Rudin-Keisler ordering, Suslin-proper are explained there and references are
given.
1. The forcing
Denition 1.1. Suppose that D
0
, . . . , D
n1
are ultralters on . The game G(D
0
, . . . , D
n1
) is dened as
follows: In his mth move player I chooses A
0
, . . . , A
n1
) D
0
. . . D
n1
and player II responds playing
k
m
A
mmodn
. Finally player II wins if and only if for every i < n, k
j
: j = imodn D
i
holds.
Lemma 1.2. Suppose D
0
, . . . , D
n1
are Ramsey ultralters which are pairwise not RK-equivalent. Let
m(l) : l < ) be an increasing sequence of integers. There exists a subsequence m(l
j
) : j < ) and sets
Z
i
D
i
, i < n, such that:
(1) l
j+1
l
j
2, for all j < ,
(2) Z
i

j=imodn
[m(l
j
), m(l
j+1
)), for all i < n,
(3) Z
i
[m(l
j
), m(l
j+1
)) has precisely one member, for every i < n and j = imodn.
2
5
3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


Proof: For j < 3, k < dene:
I
j,k
=
(2n1)(3k+j+1)1

s=(2n1)(3k+j)
[m
s
, m
s+1
) ,
J
j
=

k<
I
j,k
.
As the D
i
are Ramsey ultralters, there exist X
i
D
i
such that for every i < n:
(a) X
i
J
j
for some j < 3,
(b) if X
i
J
j
, then X
i
I
j,k
contains precisely one member, for every k < .
Next we want to nd Y
i
D
i
, Y
i
X
i
, such that for every distinct i, i

< n, Z
i
and Z
i
do not meet
any adjacent intervals I
j,k
.
Dene h : X
0
X
1
as follows. Suppose X
0
J
j
. For every k < , h maps the unique element of
X
0
I
j,k
to the unique element of X
1
which belongs to either I
j,k
or to one of the two intervals of the form
I
j

,k
which are adjacent to I
j,k
(note that these are I
2,k1
, I
1,k
if j = 0, or I
0,k
, I
2,k
if j = 1, or I
1,k
, I
0,k+1
if j = 2). As h does not witness that D
0
, D
1
are RK-equivalent, there exist X

i
D
i
, X

i
X
i
(i < 2) such
that h[X

0
] X

1
= . Note that if n = 2, we can let Y
i
= X

i
. Otherwise we repeat this procedure, starting
from X

0
and X
2
, and get X

0
and X

2
. We repeat it again, starting from X

1
and X

2
, and get X

1
and X

2
.
If n = 3 we are done. Otherwise we continue similarly. After nitely many steps we obtain Y
i
as desired.
By denition of I
j,k
it is now easy to add more elements to each Y
i
in order to get Z
i
as in the Lemma.
The worst case is that some Y
i
contains integers s < t such that (s, t) Y
u
= for all u < n. By
construction there is some I
j,k
(s, t). For every u < n 1 pick
x
u
[m((2n 1)(3k +j) + 2u + 1), m((2n 1)(3k +j) + 2u + 2))
and add x
u
to Y
i+u+1modn
. The other cases are similar.
Corollary 1.3. Suppose D
0
, . . . , D
n1
are Ramsey ultralters which are pairwise not RK-equivalent. Then
in the game G(D
0
, . . . , D
n1
) player I does not have a winning strategy.
Proof: Suppose is a strategy for player I. For every m < , i < n let /
m
i
D
i
be the set of all ith
coordinates of moves of player I in an initial segment of length at most 2m + 1 of a play in which player I
follows and player II plays only members of m.
As the D
i
are ppoints and each /
m
i
is nite, there exist X
i
D
i
such that mi < nA /
m
i
(X
i

A). Moreover we may clearly nd a strictly increasing sequence m(l) : l < ) such that m(0) = 0 and for
all l < :
i < nA /
m(l)
i
(X
i
A m(l + 1) X
i
[m(l), m(l + 1)) ,= ).
Applying Lemma 1.2, we obtain a subsequence m(l
j
) : j < ) and sets Z
i
D
i
.
Now let in his jth move player II play k
j
, where k
j
is the unique member of [m(l
j
), m(l
j+1
))X
jmodn

Z
jmodn
if it exists, or otherwise is any member of [m(l
j
), m(l
j+1
)) X
jmodn
(note that this intersection is
3
5
3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


nonempty by denition of m(l
j+1
). Then this play is consistent with , moreover X
i
Z
i
k
j
: j = imodn
for every i < n, and hence it is won by player II. Consequently could not have been a winning strategy for
player I.
Denition 1.3. Let n < be xed. The forcing Q (really Q
n
) is dened as follows: Its members are
(w,

A) []
<
[]

. If k
j
: j < ) is the increasing enumeration of

A we let

A
i
= k
j
: j = imodn for
i < n, and if l
j
: j < m) is the increasing enumeration of w then let w
i
= l
j
: j = imodn, for i < n.
Let (w,

A) (v,

B) if and only if w (max(v) + 1) = v, w
i
v
i


B
i
and

A
i


B
i
, for every i < n.
If p Q, then w
p
, w
p
i
,

A
p
,

A
p
i
have the obvious meaning. We write p
0
q and say p is a pure extension
of q if p q and w
p
= w
q
.
If D
0
, . . . , D
n1
are ultralters on , let Q(D
0
, . . . , D
n1
) denote the subordering of Q containing only
those (w,

A) Q with the property

A
i
D
i
, for every i < n.
Lemma 1.4. The forcing Q is equivalent to (T()/n)
n
Q(

G
0
, . . .

G
n1
), where (

G
0
, . . . ,

G
n1
) is the
canonical name for the generic object added by (T()/n)
n
, which consists of n pairwise not RK-equivalent
Ramsey ultralters.
Proof: Clearly (T()/n)
n
is closed and hence does not add reals. Moreover, members x
0
, . . . , x
n1
)
(T()/n)
n
with the property that if

A =

x
i
: i < n, then x
i
=

A
i
for every i < n are dense. Hence the
map (w,

A) (

A
0
, . . . ,

A
n1
), (w,

A)) is a dense embedding of the respective forcings.
That

G
0
, . . . ,

G
n1
are ((T()/n)
n
forced to be) pairwise not RK-equivalent Ramsey ultralters fol-
lows by an easy genericity argument and again the fact that no new reals are added.
Notation. We will usually abbreviate the decomposition of Qfrom Lemma 1.4. by writing Q = Q

. So
members of Q

are

A,

B []

ordered by

A
i


B
i
for all i < n; Q

is Q(

G
0
, . . . ,

G
n1
). If G is a Qgeneric
lter, by G

we denote its decomposition according to Q = Q

, and we write G

= (G

0
, . . . , G

n1
).
Denition 1.5. Let I Q(D
0
, . . . , D
n1
) be open dense. We dene a rank function rk
I
on []
<
as
follows. Let rk
I
(w) = 0 if and only if (w,

A) I for some

A. Let rk
I
(w) = if and only if is minimal such
that there exists A D
|w|modn
with the property that for every k A, rk
I
(w k) = for some < .
Let rk
I
(w) = if for no ordinal , rk
I
(w) = .
Lemma 1.6. If D
0
, . . . , D
n1
are Ramsey ultralters which are pairwise not RK-equivalent and I
Q(D
0
, . . . , D
n1
) is open dense, then for every w []
<
, rk
I
(w) ,= .
Proof: Suppose we had rk
I
(w) = for some w. We dene a strategy for player I in G(D
0
, . . . , D
n1
)
as follows: () = A
0
, . . . , A
n1
) D
0
. . . D
n1
such that for every k A
|w|modn
, rk
I
(w k) = .
This choice is possible by assumption and as the D
i
are ultralters. In general, suppose that has been
dened for plays of length 2m such that whenever k
0
, . . . , k
m1
are moves of player II which are consistent
with , then k
0
< k
1
. . . < k
m1
and for every k
i0
< . . . < k
i
l1
k
0
, . . . , k
m1
with i
j
= jmodn, j < l,
we have rk
I
(w k
i0
, . . . , k
i
l1
) = . Let S be the set of all k
i0
< . . . < k
i
l1
k
0
, . . . , k
m1
with
i
j
= jmodn, j < l, and l = mmodn. As D
|w|+mmodn
is an ultralter, by induction hypothesis we have that,
letting
A
|w|+mmodn
= k > k
m1
: s S rk
I
(w s k) = ,
4
5
3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


A
|w|+mmodn
D
|w|+mmodn
. For i ,= [w[ +mmodn, choose A
i
D
i
arbitrarily, and dene
k
0
, . . . k
m1
) = A
0
, . . . , A
n1
).
Since by Lemma 1.2. is not a winning strategy for player I, there exist k
0
< . . . < k
m
< . . . which are
moves of player II consistent with , such that, letting

A = k
m
: m < , we have (w,

A) Q(D
0
, . . . , D
n1
).
By construction we have that for every (v,

B) (w,

A), rk
I
(v) = . This contradicts the assumption that
I is dense.
Denition 1.7. Let p Q. A set of the form w
p
k
|w|
< k
|w|+1
< . . . []

is called a branch of p if
and only if max(w
p
) < k
|w|
and k
j
: j = imodn

A
p
i
, for every i < n. A set F []
<
is called a front
in p if for every w F, (w,

A
p
) p and for every branch B of p, B m F for some m < .
Lemma 1.8. Suppose D
0
, . . . , D
n1
are pairwise not RK-equivalent Ramsey ultralters. Suppose p
Q(D
0
, . . . , D
n1
) and I
m
: m < ) is a family of open dense sets in Q(D
0
, . . . , D
n1
). There exists
q Q(D
0
, . . . , D
n1
), q
0
p, such that for every m, w []
<
: (w,

A
q
) I
m
(w,

A
q
) q is a front in
q.
Proof: First we prove it in the case I
m
= I for all m < , by induction on rk
I
(w
p
). We dene a
strategy for player I in G(D
0
, . . . , D
n1
) as follows. Generally we require that
k
0
, . . . , k
r
)
i
k
0
, . . . , k
s
)
i
for every s < r and i < n, where k
0
, . . . , k
r
)
i
is the ith coordinate of k
0
, . . . , k
r
). We also require that
ensures that the moves of II are increasing (see the proof of 1.7). Dene () = A
0
, . . . , A
n1
) such that
for every k A
|w
p
|modn
, rk
I
(w
p
k) < rk
I
(w
p
).
Suppose now that has been dened for plays of length 2m, and let k
0
, . . . , k
m1
be moves of II,
consistent with . The interesting case is that m 1 = 0modn. Let us assume this rst. By denition of
() and the general requirement on we conclude rk
I
(w
p
k
m1
) < rk
I
(w
p
). By induction hypothesis
there exists A
0
, . . . , A
n1
) D
0
. . . D
n1
such that, letting

A =

i<n
A
i
, we have (w
p
,

A) p and
v []
<
: (v,

A) I (v,

A) (w
p
k
m1
,

A)
is a front in (w
p
k
m1
,

A). We shrink

A such that, letting
k
0
, . . . , k
m1
) = A
0
, . . . , A
n1
),
the general requirements on above are satised.
In the case that m1 ,= 0modn, dene k
0
, . . . , k
m1
) arbitrarily, but consistent with the rules and
the general reqirements above.
Let

A = k
i
: i < be moves of player II witnessing that is not a winning strategy. Let q = (w
p
,

A).
Let B = w
p
l
|w
p
|
< l
|w
p
|+1
< . . . be a branch of q. Hence l
|w
p
|
= k
j
for some j = 0modn. Then
w
p
k
j
l
|w
p
|+1
, l
|w
p
|+2
, . . . is a branch of (w
p
k
j
, k
0
, . . . , k
j
)). By denition of there exists m
such that (B m, k
0
, . . . , k
j
)) I. As (B m,

A) (B m, k
0
, . . . , k
j
)) and I is open we are done.
5
5
3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


For the general case where we have innitely many I
m
, we make a diagonalization, using the rst part
of the present proof. Dene a strategy for player I satisfying the same general requirements as in the
rst part as follows. Let () = A
0
, . . . , A
n1
) such that, letting

A =

A
i
: i < n, (w
p
,

A)
0
p and
it satises the conclusion of the Lemma for I
0
. In general, let k
0
, . . . , k
m1
) = A
0
, . . . , A
n1
) such that,
letting

A =

A
i
: i < n, for every v k
i
: i < m and j m, (w
p
v,

A)
0
(w
p
v,

A
p
) and it satises
the conclusion of the Lemma for I
j
(In fact we dont have to consider all such v here, but it does not hurt
doing it). If then

A = k
i
: i < are moves of player II witnessing that is not a winning strategy for I,
similarly as in the rst part it can be veried that q = (w
p
,

A) is as desired.
Corollary 1.9. Let D
0
, . . . , D
n1
be pairwise not RK-equivalent Ramsey ultralters. Suppose

A []

is
such that for every i < n and X D
i
,

A
i

X. Then

A is Q(D
0
, . . . , D
n1
)generic over V .
Proof: Let I Q(D
0
, . . . , D
n1
) be open dense. Let w []
<
. It is easy to see that the set
I
w
= (v,

B) Q(D
0
, . . . , D
n1
) : (w [v mink v
|w|modn
: k > max(w)],

B) I
is open dense. If we apply Lemma 1.8. to p = (, , . . . , ) and the countably many open dense sets I
w
where w []
<
, we obtain q = (,

B). Let a
i
: i < ) be the increasing enumeration of

A. Choose m large
enough so that for each i < n,

A
i
a
j
: j < mn

B
i
. Let w = a
j
: j < mn. By construction, there
exists v

A

B(a
mn1
+1) such that (v,

B) I
w
and wv =

Ak, for some k < . Hence (wv,

B) I,
and so the lter on Q(D
0
, . . . , D
n1
) determined by

A intersects I. As I was arbitrary, we are done.
An immediate consequence of Lemma 1.4. and Corollary 1.9. is the following.
Corollary 1.10. Suppose

A []

is Qgeneric over V , and

B []

is such that

B
i


A
i
for every
i < n. Then

B is Qgeneric over V as well.
Remember that a forcing is called Suslin, if its underlying set is an analytic set of reals and its order
and incompatibility relations are analytic subsets of the plane. A forcing P is called Suslin-proper if it is
Suslin and for every countable transitive model (N, ) of ZF

which contains the real coding P and for every

p P N, there exists a (N, P)generic condition extending p. See [JuSh] for the theory of Suslin-proper
forcing and [ShSp] for its properties which are relevant here.
Corollary 1.11. The forcing Q is Suslin-proper.
Proof: It is trivial to note that Q is Suslin, without parameter in its denition. Let (N, ) be a
countable model of ZFC

, and let p Q N. Without loss of generality, [w

p
[ = 0modn. Let

A []

V
be Qgeneric over N such that p belongs to its generic lter. Hence w
p
i


A
i
w
p
i
(

A
p
i
(max(w
p
) + 1))
for all i < n. But if q = (w
p
,

A), then clearly q
0
p and q is (N, Q)generic, as every

B []

which is
Qgeneric over V and contains q in its generic lter is a subset of

A and hence Q Ngeneric over N by
Corollary 1.10. applied in N.
The following is an immediate consequence of Corollary 1.11.
Corollary 1.12. If p Q and
n
: n < ) are Qnames for members of V , there exist q Q, q
0
p and
X
n
: n < ) such that X
n
V [V ]

and q |
Q
n(
n
X
n
).
6
5
3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


Corollary 1.13. Forcing with Q does not change the conality of any cardinal with cf() h(n) to a
cardinal below .
Proof: Suppose there were a cardinal < h(n) and a Qname

f for a conal function from to .
Working in V and using Corollary 1.12., for every < we may construct a maximal antichain p

: < c)
in Q and X

: < c) such that for all < c, w

p

= , X

[V ]

V and p

|
Q

f() X

.
Then clearly /

A
p

i
: i < n) : < c) is a maximal antichain in (T()/n)
n
. By < h(n),
/

: < ) has a renement, say /. Choose

A
i
: i < n) /. Let

A =

A
i
: i < n. We may assume
that the

A
i
also have the meaning from Denition 1.3. with respect to

A. For each < there exists ()
such that

A
i
: i < n)
(P()/fin)
n

A
p

()
i
: i < n). Then clearly
(,

A) |
Q
range(

f)

()
: < .
But as cf() h(n) and < h(n), we have a contradiction.
Lemma 1.14. Suppose D
0
, . . . , D
n1
are pairwise not RK-equivalent Ramsey ultralters. Then Q(D
0
, . . . , D
n1
)
has the pure decision property (for nite disjunctions), i.e. given a Q(D
0
, . . . , D
n1
)name for a member
of 0, 1 and p Q(D
0
, . . . , D
n1
), there exist q Q(D
0
, . . . , D
n1
) and i 0, 1 such that q
0
p and
q |
Q(D0,...,Dn1)
= i.
Proof: The set I = r Q(D
0
, . . . , D
n1
) : r decides is open dense. By a similar induction on
rk
I
as in the proof of Lemma 1.7. we may nd q Q(D
0
, . . . , D
n1
), q
0
p, such that for every q

q,
if q

decides then (w
q

,

A
q
) decides . Now again by incuction on rk
I
we may assume that for every
k

A
q
|w
q
|modn
, (w
q
k,

A
q
) satises the conclusion of the Lemma, and hence by the construction of q,
(w
q
k,

A
q
) decides . But then clearly a pure extension of q decides , and hence q does.
Lemma 1.15. Lemma 1.14 holds if Q(D
0
, . . . , D
n1
) is replaced by Q.
Proof: Suppose p Q, is a Qname and p |
Q
0, 1. As

A
p
|
Q
p Q(

G
0
, . . . ,

G
n1
), by
Lemma 1.14 there exists a Q

name

A such that

A
p
|
Q
(w
p
,

A) Q

(w
p
,

A) p (w
p
,

A) decides .
As Q

does not add reals there exist

A
1
,

A
2
[]

V such that

A
1


A
p
and

A
1
|
Q

A =

A
2
. Letting

B =

A
1


A
2
we conclude (w
p
,

B) Q, (w
p
,

B)
0
p and (w
p
,

B) decides .
The rest of this section is devoted to the proof that if the forcing Q is iterated with countable supports,
then in the resulting model cov(/) =
1
, where / is the ideal of meagre subsets of the real line, and
cov(/) is the least number of meagre sets needed to cover the real line. Hence for every n < , we obtain
the consistency of cov(/) < h(n).
Denition 1.16. A forcing P is said to have the Laver property if for every Pname

f for a member of

, g

V and p P, if
p |
P
n < (

f(n) < g(n)) ,
7
5
3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


then there exist H : []
<
and q P such that H V , n < ([H(n)[ 2
n
), q p and
q |
P
n < (

f(n) H(n)) .
It is not dicult to see that a forcing with the Laver property does not add Cohen reals. Moreover, by
[Shb, 2.12., p.207] the Laver property is preserved by a countable support iteration of proper forcings. See
also [Go, 6.33., p.349] for a more accessible proof.
Lemma 1.17. The Forcing Q has the Laver property.
Suppose

f is a Qname for a member of

and g

V such that p |
Q
n < (

f(n) < g(n)).
We shall dene q
0
p and H(i) : i < ) such that [H(i)[ 2
i
and q |
Q
i(

f(i) H(i)). We may assume
[w
p
[ = 0modn and min(

A
p
) > max(w
p
).
By Lemma 1.14 choose q
0

0
p and K
0
such that q
0
|
Q

f(0) = K
0
, and let H(0) = K
0
.
Suppose q
i

0
p, H(j) : j i) have been constructed and let a
i
be the set of the rst i +1 members of

A
qi
. Let v
k
: k < k

) list all subsets v of a

i
such that v
l
(a
i
)
l
, for every l < n (see Denition 1.3.). Then
clearly k

2
i+1
. By Lemma 1.14 we may shrink

A
qi
k

times so to obtain

A and K
i+1
k
: k < k

) such
that for every k < k

, (w
qi
v
k
,

A) |
Q

f(i +1) = K
i+1
k
. Without loss of generality, min(

A) > max(a
i
). Let
q
i+1
be dened by w
qi+1
= w
p
and

A
qi+1
= a
i

, where

A

is

A without its rst (i + 1)modn members.
Let H(i + 1) = K
i+1
k
: k < k

. Then q
i+1
|
Q

f(i + 1) H(i + 1). Finally let q be dened by w
q
= w
p
and

A
q
=

a
i
: i < . Then q and H(i) : i < ) is as desired.
As explained above, from Lemma 1.17 and Shelahs preservation theorem it follows that if P is a count-
able support iteration of Q and G is Pgeneric over V , then in V [G] no real is Cohen over V ; equivalently,
the meagre sets in V cover all the reals of V [G]. Now starting with V satisfying CH we obtain the following
theorem.
Theorem 1.18. For every n < , the inequality cov(/) < h(n) is consistent with ZFC.
2. Both of Laver and Miller forcing collapse the continuum below each h(n)
Denition 2.1. Let p
<
be a tree. For any p let succ

(p) = n < : n) p. We say

that p has a stem and denote it stem(p), if there is p such that [succ

(p)[ 2 and for every ,

[succ

(p)[ = 1. Clearly stem(p) is uniquely determined, if it exists. If p has a stem, by p

we denote the set

p : stem(p) . We say that p is a Laver tree if p has a stem and for every p

, succ

(p) is innite.
We say that p is superperfect if for every p there exists p with and [succ

(p)[ = . By L we
8
5
3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


denote the set of all Laver trees, ordered by reverse inclusion. By M we denote the set of all superperfect
trees, ordered by reverse inclusion. L, M is usually called Laver, Miller forcing, respectively.
Theorem 2.2. Suppose that G is Lgeneric or Mgeneric over V . Then in V [G], [c
V
[ = [h(n)[
V
.
Proof: Completely similarly as in [BaPeSi] for the case n = 1, a base tree T for (T()/n)
n
of height
h(n) can be constructed. I.e.
(1) T (T()/n)
n
is dense;
(2) (T,

) is a tree of height h(n);

(3) each level T

, < h(n), is a maximal antichain in (T()/n)

n
;
(4) every member of T has 2

immediate successors.
It follows easily that, rstly, every chain in T of length of countable conality has an upper bound, and
secondly, every member of T has an extension in T

for arbitrarily large < h(n).

Using T, we will dene a Lname for a map from h(n) onto c. For p L and
0
, . . . ,
n1
[p

]
n
,
let

A
p
{i:i<n}
= succ
i
(p) : i < n).
By induction on < c we will construct (p

) Lh(n) c such that the following clauses hold:

(5) if
0
, . . . ,
n1
[p

]
n
, then

A
p
{i:i<}
T

;
(6) if < ,

,
0
, . . . ,
n1
[p

]
n
[p

]
n
, then

A
p
{i:i<n}
,

A
p

{i:i<n}
are incompatible in
(T()/n)
n
;
(7) if p L, < c, then for some < c, every extension of p

is compatible with p and

= .
At stage , by a suitable bookkeeping we are given < c, p L, and have to nd

, p

such that (5),

(6), (7) hold. For p

let B

= succ

(p); for
<
p

, B

= . Let
i
0
, . . . ,
i
n1
: i < ) list
[
<
]
n
such that every member is listed
0
times.
Inductively we dene
i
: i < ) and B

:
<
,
<
2) such that
(8) B

[]

and
i
: i < ) is a strictly increasing sequence of ordinals below h(n);
(9) B

= B

;
(10) for every i < , the map B

i
0
, . . . , B

i
n1
) is one-to-one from
i+1
2 into T
i
;
(11) for every i < k, for every
k+1
2, B

B
i+1

;
Suppose that at stage i of the construction,
j
: j < i) and B

:
j
0
, . . . ,
j
n1
: j < i,
i
2)
have been constructed.
For
i
0
, . . . ,
i
n1
and
i
2, if B

is not yet dened, there is no problem to choose it such

that (8) and (11) hold. Next by the properties of T it is easy to nd
i
and B

, for every
i+1
2 and

i
0
, . . . ,
i
n1
such that (8), (9), (10), (11) hold up to i.
By the remark following the properties of T, letting

= sup
i
: i < , for every
<
and

2,
there exists B

[]

such that
(12) for all i < , B

B
i

;
(13) for all
0
, . . . ,
n1
[
<
]
n
, B

0
, . . . , B

n1
) T

.
For

2 let p

L be dened by
9
5
3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


stem(p

) = stem(p

)
(p

(succ

(p

) = B

).
It is easy to see that every extension of p

is compatible with p

. Moreover, if
0
, . . . ,
n1
[(p

],
then

A
p

{i:i<n}
T

by construction. Hence we have to nd

2 such that, letting p

= p

, (6) holds.
Note that for every
0
, . . . ,
n1
[
<
]
n
and < with

and
0
, . . . ,
n1
[p

]
n
there
exists at most one

2 such that
0
, . . . ,
n1
[(p

]
n
and

A
p

{i:i<n}
,

A
p

{i:i<n}
are compatible in
(T()/n)
n
. In fact, by construction and as T

is an antichain, either

A
p

{i:i<n}
=

A
p

{i:i<n}
or they are
incompatible; and moreover for ,= ,

A
p

{i:i<n}
,

A
p

{i:i<n}
are incompatible. Hence, as
0
[[ < c we may
certainly nd such that, letting p

= p

and

= , (5), (6), (7) hold.

But now it is easy to dene an Lname

f for a function from h(n) to c such that for every < c,
p

|
L

f(

) =

. By (7) we conclude |
L


f : h(n)
V
c
V
is onto.
A similar argument works for Miller forcing.
Combining Theorem 2.2 with Con(h(n + 1) < h(n)) from 1 we obtain the following:
Corollary 2.3. For every n < , it is consistent that both of Laver and Miller forcing collapse the continuum
(strictly) below h(n).
10
5
3
1


r
e
v
i
s
i
o
n
:
1
9
9
8
-
0
1
-
2
2







m
o
d
i
f
i
e
d
:
1
9
9
8
-
0
1
-
2
3


References
[Ba] J.E.Baumgartner, Iterated forcing, in: Surveys in set theory, A.R.D. Mathias (ed.), London Math. Soc.
Lect. Notes Ser. 8, Cambridge Univ. Press, Cambridge (1983), 159
[BaPeSi] B. Balcar, J. Pelant, P.Simon, The space of ultralters on N covered by nowhere dense sets, Fund.
Math. 110 (1980), 1124
[Go] M. Goldstern, Tools for your forcing construction, in: Israel Math. Conf. Proc. 6, H. Judah (ed.)
(1993), 305360
[GoJoSp] M. Goldstern, M. Johnson and O. Spinas, Towers on trees, Proc. AMS 122 (1994), 557564.
[GoReShSp] M. Goldstern, M. Repick` y, S. Shelah and O. Spinas, On tree ideals, Proc. AMS 123 (1995), 15731581.
[JuSh] H. Judah and S. Shelah, Souslin forcing, J. Symb. Logic 53/4 (1988), 11881207.
[Mt] A.R.D. Mathias, Happy families, Ann. Math. Logic 12 (1977), 59111.
[Shb] S. Shelah, Proper forcing, Lecture Notes in Math., vol. 942, Springer
[ShSp] S. Shelah and O. Spinas, The distributivity number of T()/n and its square, Trans. AMS, to appear.
11