Documentos de Académico
Documentos de Profesional
Documentos de Cultura
3
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
1
1
-
0
9
m
o
d
i
f
i
e
d
:
2
0
0
0
-
1
1
-
1
4
MARTINS AXIOM IS CONSISTENT WITH THE EXISTENCE OF NOWHERE
TRIVIAL AUTOMORPHISMS
SAHARON SHELAH AND JURIS STEPR
ANS
Abstract. Martins Axiom does not imply that all automorphisms of P(N)/[N]
<0
are somewhere triv-
ial. An alternate method for obtaining models where every automorphism of P(N)/[N]
<0
is somewhere
trivial is explained.
1. Introduction
In [5] Velickovic constructed a model of Martins Axiom in which there is a non-trivial automorphism
of T(N)/[N]
<
0
. As well as answering a question posed in [4], this put into context another result
of the [5] showing that the conjunction of MA and OCA implies that all automorphisms are trivial.
However, the non-trivial automorphims constructed by Velickovic is trivial on many innite subsets
of the integers. Indeed, it was shown in [3] that this is unavoidable since every automorphism of
T(N)/[N]
<
0
is somewhere trivial in Velickovics model of [5].
Hence, the question arises of whether or not Martins Axiom alone is sucient to imply that, while
there may be non-trivial automorphisms, nevertheless, all automorphisms of T(N)/[N]
<
0
are somewhere
trivial. The main result of this paper is that this is not the case.
The last section presents a simple, alternate method for obtaining models where all automorphisms
are somewhere trivial. It has the advantage that it can produce models where d =
1
whereas the oracle
chain condition method adds Cohen reals and so does not achieve this.
2. Martins Axiom and a nowhere trivial automorphism
If and are ordinals then the notation [, ) will be used to denote the set . The relations
and
will have the usual meaning as relations on subsets of the integers modulo a nite set.
The convention on forcing partial orders will be that larger conditions force more information.
Denition 2.1. If W is a set of ordinals then the indexed family S = (A
, F
, B
)
W
will be said
to be a tower of permutations if
(1) A
N and F
(3) B
and A
(5) if then F
(B)
.
Dene Q(S) to be the set consisting of all quadruples p = (a
p
, f
p
,
p
, B
p
) such that
(1) a
p
N is a nite subset
(2) f
p
m is a permutation of m for each m a
p
and the domain of f
p
is max(a
p
)
(3)
p
W
The research of the rst author was supported by The Israel Science Foundation founded by the Israel Academy of
Sciences and Humanities, and by NSF grant No. NSF-DMS97-04477. Research of the second author for this paper was
partially supported by NSERC of Canada. This is paper number 735 in the rst authors personal listing.
1
7
3
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
1
1
-
0
9
m
o
d
i
f
i
e
d
:
2
0
0
0
-
1
1
-
1
4
2 S. SHELAH AND J. STEPR
ANS
(4) max(a
p
) A
p
and the relation on Q(S) is dened by p q if and only if
(2.1) a
p
a
q
, (max(a
p
) + 1) a
q
= a
p
, f
p
f
q
, B
p
B
q
,
p
q
(2.2) (A
q max(a
q
)) (a
q
a
p
) A
p
and, for each B belonging to B
p
B
q max(a
q
)]
2
)F
q (B [m, n)) = F
pG
a
p
and F
S
[G] to be
pG
f
p
.
Lemma 2.1. For any tower of permutations S the structure (Q(S), ) is a partial order.
Proof. That (Q(S), ) is reexive and antisymmetric is obvious. To prove transitivity suppose that
p q and q r. The condition 2.1 for p r is easily seen to be satised. To see that condition 2.2 for
p r is satised note that
A
r max(a
r
) A
q max(a
r
) A
q max(a
q
) A
p
and that
a
r
a
p
(a
r
a
q
) (a
q
a
p
) A
q max(a
q
) A
p A
p
which shows that (A
r max(a
r
)) (a
r
a
p
) A
p, as required.
To show that conditions 2.3 and 2.4 hold, let B B
p
B
q (B[n, m))
since q r and B
p B
q (B [n, m)) = F
r max(a
r
)]
2
. It follows from p q and q r
that F
q (B[n, m)) = F
r max(a
r
) A
q max(a
q
), that F
q (B
[n, m)) = F
r (B [n, m)) = F
, F
, B
)
W
, an integer n, B
W
B
and
W the following sets are dense in Q(S):
(2.5) p Q(S) : max(a
p
) > n
(2.6) p Q(S) :
p
(2.7) p Q(S) : B B
p
Proof. To prove that the set 2.5 is dense let p Q(S) and n be given. Let k A
p
be such that k > n.
Using Condition 4 of Denition 2.1 it follows that max(a
p
) A
p and, hence, F
p
[max(a
p
), k) is a
permutation of [max(a
p
), k). Letting q = (a
p
k, f
p
F
p
[max(a
p
), k),
p
, B
p
) it follows that q p.
Observe for later reference, that is has actually been shown that
(2.8) (p Q(S))(n N)(q p)max(a
q
) > n and
q
=
p
and B
p
= B
q
.
7
3
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
1
1
-
0
9
m
o
d
i
f
i
e
d
:
2
0
0
0
-
1
1
-
1
4
NOWHERE TRIVIAL AUTOMORPHISMS 3
To prove that the set 2.6 is dense let p Q(S) and W be given. Since it may as well be assumed
that >
p
, it is possible to nd n so large that A
n A
n]
2
and B B
p
F
(B [i, j)) = F
, F
, B
)
W
is a tower of permutations and p Q(S) then
(2.10) p
Q(S)
A
S
[G] max(a
p
) A
p
(2.11) p
Q(S)
(B B
p
B
p)(n, m [A
p max(a
p
)]
2
)F
S
[G](B [n, m)) = F
, as well as a P
-name
for a tower of permutations S
= (A
, F
, B
)
C
, by induction on . Let A
0
[N]
0
and F
0
be
arbitrary subject to the fact that F
0
is a permutation of N such that F
0
[n, m) is a permutation for
each n, m A
0
. Let B
0
= T(N) in the sense of the ground model. Then let S
1
= (A
0
, F
0
, B
0
) and
let P
1
= Q
0
be Cohen forcing. If is a limit then P
and
S
C
S
. If / C then P
+1
= P
where Q
.
If C 0 then P
+1
= P
Q(S
) and A
is dened to be A
S
[G], F
is dened to be F
S
[G] where
G is the canonical name for the generic set on Q(S
). In this case S
+1
= S
(A
, F
, T(N) V
P
).
As usual, if p P
then p
P
p() Q
.
Denition 2.2. Let P
w
, f
such that
= domain(p) C 0
0
if then p
P
p() = ( a
,
f
, B
) for some B
for each
if then max(a
) = k
if and
are in and
< then p
P
B
= B
.
The pair (k, (a
, f
. Let P
, f
) then
if then p + 1
P
+1
A
(A
k)
if then p + 1
P
+1
(B B
)(n, m [A
k]
2
)F
(B [n, m)) =
F
(B [n, m))
Proof. This is an immediate consequence of Lemma 2.3.
7
3
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
1
1
-
0
9
m
o
d
i
f
i
e
d
:
2
0
0
0
-
1
1
-
1
4
4 S. SHELAH AND J. STEPR
ANS
Denition 2.3. If p P
w
is witnessed by (k, (a
, f
) then dene p
+
P
by
p
+
() =
p() if /
(a
, f
, max( ), B
p()
) if .
Lemma 2.5. If p P
w
then p
+
p.
Proof. Proceed by induction on to show that p
+
p . Note that the cases = 0 or a
limit pose no problem. Given that p
+
p let be enumerated, in order, by
1
,
2
, . . .
n
.
If follows directly from Lemma 2.4 and the denition of p
+
that
p
+
P
k (A
k) . . . (A
n
k)
p
+
P
(B B
j
B
i
)(m, m
k)F
j
(B[m, m
)) = F
i
(B[m, m
)) if i j.
In particular, noting that there is some i such that
=
i
,
p
+
P
k A
n
and
p
+
P
(B B
B
n
)(m, m
k)F
(B [m, m
)) = F
n
(B [m, m
)).
Hence, p
+
+ 1 p + 1.
Lemma 2.6. For each the subset P
w
is dense in P
.
Proof. Proceed by induction on noting that the cases 1 and a limit are trivial. Therefore,
suppose that Lemma 2.6 has been established for and that p P
+1
. Without loss of generality it may
be assumed that C. Choose p
p() = ( a,
f,
, B)
and, moreover, p
B = B
i
im
and, for each i m the condition p
) and that p
(i)
P
(i)
B
i
B
p
((i))
for i m. Then use
the induction hypothesis to nd p
P
w
extending p
, f
) witnesses that
p
P
w
implies that
F
= f F
[max(a), k)
and a
() if
(a
, f
, , B
B
p
()
) if =
and note that q P
w
+1
and q p.
Corollary 2.1. For each the subset P
is dense in P
.
Proof. This is immediate from Lemma 2.6 and Lemma 2.5.
Lemma 2.7. Given that p P
w
, f
) and (C )
domain(p) then the following condition p) extends p and is also in P
w
:
p)()
p() if ,=
(k, (j, j)
jk
, 0,
B
p()
) if = .
Proof. Notice that since / domain(p) the restrictions on extension do not apply and it is easy to check
that the condition (k, (j, j)
jk
, max( ),
B
p()
) belongs to Q
. Since p P
w
it follows
that k A
0
and, hence, that p) P
w
such that:
p P
is witnessed by (k, (a
p
, f
p
)
p
)
q P
is witnessed by (k, (a
q
, f
q
)
q
)
max(domain(p)) = max(domain(q)) =
p
q
(a
q
, f
q
) = (a
p
, f
p
)
max(domain(q)) < min(domain(p) 0)
p(0) and (0) are compatible.
Under these conditions p and q are compatible.
Proof. Let the maximum member of
q
. Dene q . p by
(q . p)() =
q() if domain(q) (
q
)
p() if domain(p) (
p
)
(a
q()
, f
q()
,
q()
, (B
p
B
0
) B
q()
) if
q
(a
p()
, f
p()
,
p()
, B
p()
B
q()
) if
p
(a
p( )
, f
p( )
,
q
( ), B
p( )
B
q( )
) if =
and dene p . q by
(p . q)() =
q() if domain(q) (
q
)
p() if domain(p) (
p
)
(a
q()
, f
q()
,
q()
, (B
p
B
0
) B
q()
) if
q
(a
p()
, f
p()
,
p()
, B
p()
B
q()
) if
p
(a
p( )
, f
p( )
,
p
( ), B
p( )
B
q( )
) if =
(The only dierence is to be found in the last lines of the two denitions.) It is easy to check that
p p . q and q q . p and that both p . q and q . p belong to P
w
. Moreover (p . q)
+
= (q . p)
+
.
Hence q and p are compatible.
Lemma 2.9. If the conality of C is not
1
and S = (A
, F
, B
)
C
is a tower of permutations
then Q(S) has property K.
Proof. This is standard. If the conality of C is less than
1
then choose a countable, conal subset
C
of C. Using observation 2.9 of Lemma 2.2 it follows that the set of all p Q(S) such that
p
C
1
Q(S) it is possible to
choose C such that
p
< for each . Using observation 2.9 of Lemma 2.2 it may be assumed,
by extending each condition, that
p
= a and f
p
.
Next, observe that if C is not conal in then the induction hypothesis is easily applied since, in
this case, C has a maximal element below and so Q
ANS
If C has conality dierent from
1
then, once again Lemma 2.9 implies that Q
is in P
, (a
, f
) As well, by
thinning out, it may be assumed that there is k such that k
, f
1
form a -system with root . Using
the fact that is a limit, choose some C such that there is some uncountable
1
such that
if , []
2
and then max(domain(p
)) < min(domain(p
to some p
)
+
.
Now, using the induction hypothesis, nd , []
2
such that and p
)
+
+ 1 is
compatible with p
)
+
+ 1. Choose r P
+1
extending both p
)
+
+ 1 and p
)
+
+ 1.
Let G P
+1
be generic over V and containing r. Observe that P
/P
+1
as interpreted in V [G] is a
partial order like P
/P
+1
is Q((A
, F
it follows that k A
)
+
[, )) and (p
)
+
[, ))
belong to (P
/P
+1
)
)
+
[, )) and (p
)
+
and p
.
Lemma 2.11. Let S = (A
, F
, B
)
W
be a tower of permutations and suppose that A [N]
0
and
is a one-to-one function from A to N. Then Q(S) forces that there are innitely many a A such
that F
S
[G](a) ,= (a).
Proof. Let p Q(S) and suppose that p
Q(S)
(a
A
k)(a) = F
S
[G](a) for some integer k.
From Lemma 2.2 it can be assumed that k max(a
p
). Now choose m A
p(i), F
p ([max(a
p
), m) i, j) g it follows that f(B [max(a
p
), m)) = F
p(B [max(a
p
), m)) for
each B B
p
. Hence p q = (a
p
m, f
p
f,
p
, B
p
) and q
Q(S)
i) ,= F
S
[G](
i).
Lemma 2.9 and Lemma 2.11, in the case case when C has the maximal element, together imply
that there is a trivial automorphism of T(N)/[N]
<
0
and a -centred forcing P such that can be
extended to a trivial automorphism of T(N)/[N]
<
0
in two dierent ways in the generic extension by P.
Hence, these lemmas can be thought of as strengthening the folklore result that certain automorphisms
such as the identity can be extended to generic trivial automorphisms by the natural -centred
forcing; in other words, the forcing which approximates the generic permutation by nite permutations
and promises to respect the identity on certain, nitely many, innite sets.
Lemma 2.12. Let S = (A
, F
, B
)
W
be a tower of permutations such that W has a maximal
element . If A belongs to [N]
0
, but not necessarily to B
[H](j).
Proof. This is a standard use of genericity.
Theorem 2.1. Given any regular, uncountable cardinal , it is consistent relative to the consistency
of set theory that Martins Axiom holds, 2
0
= and there is a nowhere trivial automorphism of T(N)
modulo the nite sets.
Proof. Let C be a closed unbounded set such that C is unbounded. Construct a nite support
iteration P
C
so that Martins Axiom is forced to hold by the iteration on C and such that a tower
of permutations (A
, F
, B
= T(N) V [P
G]
7
3
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
1
1
-
0
9
m
o
d
i
f
i
e
d
:
2
0
0
0
-
1
1
-
1
4
NOWHERE TRIVIAL AUTOMORPHISMS 7
where G is P
(A)
B and A B
(j) = F
(j) for
innitely many j belonging to Z. Therefore F
(Z)] = ([Z]).
It should be noted that Theorem 2.1 would be of interest even if Martins Axiom did not hold in
the model constructed since it would still provide a method for constructing models of set theory with
nowhere trivial automorphisms of T(N)/[N]
<
0
and the continuum arbitrarily large.
3. Ruining automorphisms with Silver reals
Denition 3.1. Suppose that Q and P are partial orders such that Q is completely embedded in P
and that Q is Suslin (see [1]). Then P will be said to be suciently Suslin over Q if for every P-name
g for a function from to 2 there is a dense set of p P such that,
(q, f) Q2
[(p
P)p
q and p
p and p
P
g =
f
is analytic.
Lemma 3.1. Any countable support iteration of a combination of Silver reals and Sacks reals is su-
ciently Suslin over its rst coordinate.
Proof. Let the iteration be obtained from the sequence P
[
+
), )
be a countable elementary submodel containing P
i
enumerate and
let
j
=
i
ij
. Standard fusion arguments will allow the construction of a family p
:n2
<n such
that:
(1) If ()
.
(2) If :
n
2
<n
then p
:n2
<n
and p G
p
then p P.
It follows that, for each q P
0
there is some p
p such that p
(0) q and p
P
g =
f if and only
if, letting S(f, n) = [ :
n
2
<n
and p
P
g(n) =
f( n) and p
S(f,n)
p
n
has a proper lower bound. For Silver and Sacks forcings checking for a lower bound
is easily seen to be
1
1
in the parameter dening the fusion sequence; in fact, it is Borel.
Lemma 3.2. If P is as in lemma 3.1 and is an automorphism of T(N)/[N]
<
0
which is not trivial
on any member of []
0
then it is not possible to extend to an automorphism of T(N)/[N]
<
0
in any
generic extension by S P.
Proof. Assuming the lemma is false, it is possible to nd a condition (s, p) S P and an S P-name
for a set of integers Z such that
(s, p)
SP
(X
G
) = Z
7
3
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
1
1
-
0
9
m
o
d
i
f
i
e
d
:
2
0
0
0
-
1
1
-
1
4
8 S. SHELAH AND J. STEPR
ANS
where X
G
is a name for the generic subset of added by S. Let g =
Z
be the characteristic function
of Z. Using the fact that S P is suciently Suslin, nd (s
, p
) S P such that (s
, p
) (s, p) and
( s, f) S 2
, p
) and q
SP
g =
f
)A has
innite complement in the integers. It follows that if Ris dened to be
(t, f) 2
A
2
: (q (t s
, p
))q
SP
g =
f
then R is analytic.
There are two possibilities. First, suppose that the domain of R is all of 2
A
. In this case it is possible
to nd a continuous function S dened on a comeagre subset of T(A) such that R(t, S(t
1
1)) holds
for all t in the domain of R. Now let : domain(S) T(
(A)) be dened by
(W) = n
(A) : S(W)(n) = 1
and observe that is continuous. Furthermore, (W)
(X
G
)
(A)
(X
G
A)
(W).
Therefore, since (W) and
(W) are sets in the ground model, it follows from the absoluteness of
that (W)
t, p
t, p
)
SP
Z
(A) ,
(t
1
1) which is a
contradiction to the assumption that (s, p)
SP
(X
G
) = Z.
Theorem 3.1. Let V be a model of 2
0
=
1
If P
2
is the countable support iteration of partial orders
as in Lemma 3.1 such that P
+1
= P
2
which is generic over V .
(Hence, it is consistent that every automorphism of T(N)/[N]
<
0
is somewhere trivial and d =
1
.)
Proof. Any automorphism can be reected on a closed unbounded subset of
2
consisting of ordinal of
uncountable conality. If is any such ordinal such that P
+1
= P