Professional Documents
Culture Documents
Krzysztof Ciesielski and Saharon Shelah - Category Analogue of Sup-Measurability Problem
Krzysztof Ciesielski and Saharon Shelah - Category Analogue of Sup-Measurability Problem
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
Category analogue of sup-measurability
problem
Krzysztof Ciesielski
The work of the rst author was partially supported by NSF Cooperative Research
Grant INT-9600548, with its Polish part nanced by Polish Academy of Science PAN.
This work was supported in part by a grant from Basic Research Foundation founded
by the Israel Academy of Sciences and Humanities. Publication 695.
The authors wish to thank Professor Andrzej Ros lanowski for reading earlier versions
of this paper and helping in improving its nal form.
1
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
the original measurable case is the subject of a work in prepartion by
Roslanowski and Shelah.
1 Introduction
Our terminology is standard and follows that from [3], [4], [10], or [12].
In particular, pr: X Y X will stand for the projection onto the rst
coordinate. A subset A of a Polish space X is nowhere meager provided
A U is not meager for every non-empty open subset of X.
The ternary Cantor subset of R will be identied with with its home-
omorphic copy, 2
: s f,
where s 2
<
. Also, since R Q is homeomorphic to 2
E for some
countable set E (the set of all eventually constant functions in 2
) in our
more technical part of the paper we will be able replace R with 2
.
The study of sup-measurable functions
1
comes from the theory of dif-
ferential equations. More precisely it comes from the question: For which
functions F : R
2
R does the Cauchy problem
y
) A are
nowhere meager in 2
onto 2
such that the set pr(A f) does not have the Baire property
in 2
.
Before proving this theorem let us notice that it implies easily the follow-
ing corollary.
Corollary 3 The existence of Baire sup-measurable function F : R
2
R
without the Baire property is independent from the set theory ZFC.
2
It is also true that F : R
2
R is Baire sup-measurable provided F
f
has the Baire
property for every Baire class one function f : R R, and that F : R
2
R is sup-
measurable provided F
f
is measurable for every continuous function f : R R. See for
example [2, lemma 1 and remark 1].
3
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
Proof. As mentioned above under CH there exist Baire sup-measurable
functions without the Baire property. So, it is enough to show that the
property from Theorem 2 implies that there are no such functions.
So, take an arbitrary A R
2
without the Baire property. By (iii) of
Proposition 1 it is enough to show there exists a Borel function f : R R
for which the set pr(A f) does not have the Baire property.
We will rst show this under the additional assumption that the sets A
and R
2
A are nowhere meager in R
2
. But then the set A
0
= A (R Q)
2
and its complement are nowhere meager in (R Q)
2
. Moreover, since R Q
is homeomorphic to 2
E)
2
2
. Then A
0
and its complement are still nowhere
meager in 2
E: x, f(x)) A
0
does not
have the Baire property in 2
. Now, as before, f (2
E) can be considered
as dened on R Q. So if
f : R R is an extension of f (2
E) (under
such identication) to R which is constant on Q, then
f is Borel and the set
pr(A
0
f) does not have the Baire property in R.
Now, if A is an arbitrary subset of R
2
without the Baire property we can
nd non-empty open intervals U and W in R such that A and (U W) A
are nowhere meager in U W. Since U and W are homeomorphic with R
the above case implies the existence of Borel function f
0
: U W such that
pr(A f
0
) does not have the Baire property in U. So any Borel extension
f : R R of f
0
works.
2 Reduction of the proof of Theorem 2 to the
main lemma
The theorem will be proved by the method of iterated forcing, a knowledge
of which is needed from this point on.
The idea of the proof is quite simple. For every nowhere meager subset
A of 2
for which A
c
= (2
, with (2
) A
2
= P
,
Q
): <
2
) such that the
generic extension V [G] of V with respect to P
2
will satisfy 2
= 2
1
=
2
and have the property that
(m) every non-Baire subset A
of 2
= Q
A
for the
sets A of cardinality
1
, whose number is equal to
2
, the length of iteration.
Condition (m) will guarantee that this will give us enough control of all
nowhere meager subsets A
of 2
.
The second problem is that even if at some stage <
2
of our iteration
we will add a homeomorphism f appropriate for a given set A 2
,
that is such that
V [G
,
where G
= G P
def
= <
1
: is a limit ordinal
then
an
1
-oracle is any sequence / = M
: ) where M
is a countable
transitive model of ZFC
, M
is stationary in
1
for every A
1
.
5
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
The existence of an
1
-oracle is equivalent to the diamond principle .
With each
1
-oracle / = M
for A
1
. It is proved
in [12, Claim 1.4] that D
M
is a proper normal lter containing every closed
unbounded subset of .
We will also need the following fact which, for our purposes, can be viewed
as a denition of /-cc property.
Fact 4 Let P be a forcing notion of cardinality
1
, e: P
1
be one-to-
one, and / = M
: ) be an
1
-oracle. If there exists a C D
M
such
that for every C
e
1
(E) is predense in P for every set E M
) A are
nowhere meager in 2
.
The proof of Lemma 5 represents the core of our argument and will be
presented in the next section. In the remainder of this section we will sketch
how Lemma 5 implies Theorem 2. Since this follows the standard path, as
described in [12, Chapter IV], the readers familiar with this treatment may
proceed directly to the next section.
First of all, to dene an appropriate iteration we will treat forcings Q
A
from Lemma 5 as dened on
1
. More precisely, in the iteration we will always
replace Q
A
with its order isomorphic copy
1
,
A
). So, we can treat any -
nite support iteration P
= P
,
Q
): < ) of Q
A
forcing notions as having
an absolute and xed universe, say U
= g (
1
)
2
: g
1
(
1
0) []
<
.
This will allow us to treat the
2
-sequence X
: <
2
) as a sequence of
P
-names of subsets of 2
1
holds and that S 2
.
Proof. By [12, Example 2.2] for any non-empty basic open set W of 2
2
is dened by choosing by induction the sequence
P
,
A
,
/
,
Q
,
f
): <
2
) such that for every <
2
(a) P
= P
,
Q
is a P
forces that
and (
A
)
c
are nowhere meager subsets of 2
,
(c)
/
is a P
forces that
is an
1
-oracle and for every
Q satisfying
/
-cc we have
(i) for every < if P
= P
P
,
then
P
P
,
Q is
/
-cc,
(ii) if = + 1 then
Q pr(
f
), pr(
f
) 2
,
(d)
Q
is a P
forces
is an
/
-cc forcing Q
A
from Lemma 5,
(e)
f
is a P
+1
-name for which P
+1
forces that
is a
Q
,
A
,
/
,
Q
,
f
as X
from the
2
-sequence if it satises (b) and arbitrarily,
still maintaining (b), otherwise. Since steps (d) and (e) are facilitated by
Lemma 5, it is enough to construct
/
. Let /
and P
,
be the
interpretations of
/
and
P
,
, respectively. By the inductive assumption
for every < < forcing P
,
is /
)
P
,
= V
P
there is an
1
-oracle
/
such that if Q is /
-cc then P
,
Q is /
-cc.
So, in V
P
, we have
1
-oracles /
s in a sense that if Q is /
-name
/
for /
= P
.
Since, by (d), P
forces that
Q
is
/
P
,
is
/
-cc
for every < . So, proceeding as in Case 1, in V
P
we can nd
1
-oracles
such that
P
P
,
Q is
/
-cc
for every Q which is
/
-cc. Let / be an
1
-oracle from Lemma 6 used
with S = pr(
f
which
is stronger than all /
-name
/
for /
for
which (c) holds. This nishes the construction of the iteration.
To nish the argument rst note that the interpretations of pr(
f
)
and pr(
f
. This
is the case since, by (e), P
+1
forces that
pr(
f
) and pr(
f
,
and, by (c)(i), that
every
P
+1,
is
/
+1
-cc
8
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
while, by condition (c)(ii), every
/
+1
-cc forcing preserves nowhere meager-
ness of pr(
f
) and pr(
f
of 2
. However, P
2
is ccc. So, if
A is a P
2
-name
for A
: P
A V
P
is nowhere meager in 2
2
guarantees that A
.
3 Proof of Lemma 5
Let / be the family of all sequences
h = h
is a
function from a countable set D
onto R
and that
D
= R
into 2
;
(B) n < and is a permutation of 2
n
;
(C) [D D
[ 1 for every ;
(D) if x D D
then h(x) = h
= n
, h
) is stronger than
p = n, , h), p
p, provided
n n
, h h
, and
. (2)
9
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
Note that the second part of (D) says that for every x D and s 2
n
x [s] if and only if h(x) [(s)]. (3)
Also, if n < we will write [s] 2
n
for x 2
n
: x [s]. Note that in this
notation the part of (2) concerning permutations says that
extends in a
sense that
maps [t] 2
n
onto [(t)] 2
n
for every t 2
n
.
In what follows we will use the following basic property of Q
h
.
() For every q = n, , h) Q
h
and m < there exist an n
m and a
permutation
of 2
n
such that q
= n
, h) Q
h
and q
extends q.
The choice of such n
and
maxm, n such
that x n
,= y n
: x D [t] [t] 2
n
on D
t
by
(x n
) = h(x) n
for every x D
t
.
Then
is a bijection from D
t
onto H
t
and this denition ensures that an
appropriate part of the condition (D) for h and
onto [t] 2
n
) D
t
onto
([(t)] 2
n
) H
t
, then the condition (2) will be satised. Thus such dened
q
= n
, h) belongs to Q
h
and extends q.
Next note that forcing Q
h
has the following property, described in Fact 7,
needed to prove Lemma 5. In what follows we will consider 2
with the
standard distance:
d(r
0
, r
1
) = 2
min{n<: r
0
(n)=r
1
(n)}
for dierent r
0
, r
1
2
.
Fact 7 Let
h = h
: ) / and f =
h: n, , h) H, where H
is a V -generic lter over Q
h
. Then f is a uniformly continuous one-to-one
function from a subset D of 2
into 2
is dense in 2
and f can
be uniquely extended to an autohomeomorphism
f of 2
.
Proof. Clearly f is a one-to-one function from a subset D of 2
into 2
.
To see that it is uniformly continuous choose an > 0. We will nd > 0
10
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
such that r
0
, r
1
D and d(r
0
, r
1
) < imply d(f(r
0
), f(r
1
)) < . For this
note that, by (), the set
S = q = n, , h) Q
h
: 2
n
<
is dense in Q
h
. So take a q = n, , h) H S and put = 2
n
. We claim
that this works.
Indeed, take r
0
, r
1
D such that d(r
0
, r
1
) < . Then there exists a
q
= n
, h
. Therefore, n n
(r
j
) n = (h
(r
j
) n
) n =
(r
j
n
) n = (r
j
n)
by the conditions (D) and (2). Since d(r
0
, r
1
) < = 2
n
implies that
r
0
n = r
1
n we obtain
f(r
0
) n = (r
0
n) = (r
1
n) = f(r
1
) n
that is, d(f(r
0
), f(r
1
)) 2
n
< . So f is uniformly continuous.
Essentially the same argument (with the same values of and ) shows
that f
1
: f[D] D is uniformly continuous. Thus, if
f is the unique
continuous extension of f into cl(D), then
f is a homeomorphism from cl(D)
onto cl(f[D]).
To nish the argument assume that all functions h
of h intersects [t]
is dense in Q
h
. Indeed, if q = n, , h) Q
h
then, by (), strengthening q
if necessary, we can assume that m n. Then, rening t if necessary, we
can also assume that m = n, that is, that t is in the domain of . Now, if
[t] intersects the domain of h, then already q belongs to S
t
. Otherwise take
with D
= and pick x, h
. Then n, , h x, h
(x))) belongs to S
t
and
extends q.
This shows that D [t] ,= for every t 2
<
, that is, D is dense in 2
.
A similar argument shows that for every t 2
<
the set
S
t
= q = n, , h) Q
h
: the range of h intersects [t]
11
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
is dense in Q
h
, which implies that h[D] is dense in 2
. Thus
f is a homeo-
morphism from cl(D) = 2
onto cl(h[D]) = 2
.
Now take A 2
) A are nowhere
meager in 2
and x an
1
-oracle / = M
: ). By Fact 7 in order
to prove Lemma 5 it is enough to nd an
h = h
: ) / such that
Q
A
= Q
h
is /-cc (4)
and Q
h
forces that, in V [H],
the sets pr(f A) and pr(f A) are nowhere meager in 2
. (5)
(In (5) function f is dened as in Fact 7.)
To dene
h we will construct a sequence x
, y
) 2
: <
1
)
aiming at h
= x
+n
, y
+n
): n < , where .
Let s
n
, t
n
): n < be an enumeration of 2
<
2
<
with each pair
s, t) appearing for an odd n and for an even n. Points x
+n
, y
+n
) are
chosen inductively in such a way that
(i) x
+n
, y
+n
) is a Cohen real over M
[x
, y
[x
, y
): < +n)];
(ii) x
+n
, y
+n
) A if n is even, and x
+n
, y
+n
) A
c
otherwise.
(iii) x
+n
, y
+n
) [s
n
] [t
n
].
The choice of x
+n
, y
+n
) is possible since both sets A and A
c
are nowhere
meager, and we consider each time only countably many meager sets. Con-
dition (iii) guarantees that the graph of each of h
will be dense in 2
.
Note that if
0
< <
k1
, where k < , then (by the product
lemma in M
)
x
i
, y
i
): i < k) is an M
)
k
. (6)
For q = n, , h) Q
h
dene
q =
s,t
[s] [t].
12
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
Clearly q is an open subset of 2
, h
)
if q r then q r and q ([s] [t]) ,= for every s, t)
. (7)
Also for let (Q
h
)
n, , h) Q
h
: h
<
h
be a predense subset of (Q
h
)
. Then for
every k < and p = n, , h) (Q
h
)
the set
B
k
p
=
( q)
k
: q extends p and some q
0
E
(8)
is dense in ( p)
k
(2
)
k
.
Proof. By way of contradiction assume that B
k
p
is not dense in ( p)
k
. Then
there exist m < and s
0
, t
0
, . . . , s
k1
, t
k1
2
m
with the property that
P =
i<k
([s
i
] [t
i
]) ( p)
k
is disjoint from B
k
p
. Increasing m and rening
the s
i
s and t
j
s, if necessary, we may assume that m n, all s
i
s and t
j
s are
dierent,
i<k
[s
i
] is disjoint from the domain D of h, and h[D]
i<k
[t
i
] = .
We can also assume that x m ,= y m for every dierent x and y from
D and from h[D]. Now, rening slightly the argument for () we can nd
r = m,
, h) (Q
h
)
(s
i
) = t
i
for every i < k.
(Note that P ( p)
k
.) We will obtain a contradiction with the predensity of
E in (Q
h
)
for
i < k, we would also have ( q)
k
P ,= , contradicting P B
k
p
= . This
nishes the proof of Fact 8.
Now we are ready to prove (4), that is, that Q
h
is /-cc. So, x a bijection
e: Q
h
1
and let
C =
: (Q
h
)
= e
1
() M
.
Then C D
M
. (Just use a suitable nice codding or [12, Claim 1.4(4)].) Take
a C and x an E , E M
, for which e
1
(E) is predense in (Q
h
)
.
By Fact 4 it is enough to show that
e
1
(E) is predense in Q
h
.
13
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
Take p
0
= n, , h
0
) from Q
h
, let h = h
0
<
D
and h
1
= h
0
h,
and notice that the condition p = n, , h) belongs to (Q
h
)
. Assume that
h
1
= x
i
, y
i
): i < k. Since s(h
1
) = x
i
, y
i
): i < k) ( p)
k
, B
k
p
M
(as
dened from (Q
h
)
) and, by Fact 8, B
k
p
is dense in ( p)
k
condition (6)
implies that s(h
1
) B
k
p
. So there are q = n
0
,
0
, g) (Q
h
)
extending p
and some q
0
e
1
(E) for which s(h
1
) q
k
. But then p
= n
0
,
0
, g h
1
)
belongs to Q
h
and extends q. This nishes the proof of (4).
The proof of (5) is similar. We will prove only that pr(f A) = pr(f A
c
)
is nowhere meager in 2
. So there exists an s
2
<
such that pr(f A) is meager in [s
]. Let
a condition p
Q
h
and Q
h
-names
U
m
, for m < , be such
p
Q
h
each
U
m
is an open dense subset of [s
] and pr(f A)
m<
U
m
= .
For each m < , since p
forces that
U
m
is an open dense subset of [s
], for
every t 2
<
extending s
and t 2
<
extending s
there are
s 2
<
and k <
m
s
such that the conditions q and p
m
s,k
are compatible
and t s.
Note that for suciently large we have p
m
s,k
(Q
h
)
and (Q
h
)
is stationary in
1
. (Just use a suitable nice coding, or see [12, Ch.IV,
Claim 1.4(4)]). Thus, using clause (iii) of the choice of x
s, we may nd a
B
0
, an odd j < , and a condition p
0
= n
0
,
0
, h
0
) Q
h
such that
14
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
p
0
p
, s
x
+j
, and
x
+j
belongs to the domain of h
0
.
Then p
0
x
+j
[s
m<
U
m
,
which will nish the proof.
So, assume that this is not the case. Then there exist an i < and
a p
1
= n, , h
1
) Q
h
stronger than p
0
such that p
1
x
+j
/
U
i
. Let
us dene h = h
1
x
: < and h
1
h = a
l
, b
l
): l < m. Notice
that the condition p = n, , h) belongs to (Q
h
)
0
, . . . , z
m1
, z
m1
) (2
)
m
for
which
there exist s 2
<
, k <
i
s
, and q (Q
h
)
such that s z
0
, q extends
p and p
i
s,k
, and z
0
, z
0
, . . . , z
m1
, z
m1
) ( q)
m
.
Claim The set Z belongs to the model M
.
Now note that it is possible to choose a z = z
0
, z
0
, . . . , z
m1
, z
m1
) ( q)
m
such that s z
0
, s
i
z
i
, t
i
z
i
. Then z Z
i<k
([s
i
] [t
i
]).
Since Z is clearly open, this completes the proof of Claim.
Now, by (6) and the Claim above, a
l
, b
l
): l < m) belongs to Z since
a
l
, b
l
): l < m) belongs to ( p
1
)
m
= ( p)
m
. But this means that there exist
q = n
q
,
q
, h
q
) (Q
h
)
and s 2
<
such that:
q p, q [s]
U
i
, and
a
l
, b
l
): l < m) ( q)
m
, and x
+j
= a
0
[s].
15
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
But then p
2
= n
q
,
q
, h
q
a
l
, b
l
): l < m) belongs to Q
h
and extends
both q and p
1
. So, p
2
forces that x
+j
= a
0
[s]
U
i
, contradicting our
assumption that p
1
x
+j
/
U
i
.
This nishes the proof of (5) and of Lemma 5.
References
[1] M. Balcerzak, Some remarks on sup-measurablilty, Real Anal. Exchange
17 (199192), 597607.
[2] M. Balcerzak, K. Ciesielski, On sup-measurable functions problem, Real
Anal. Exchange 23 (199798), 787797.
[3] T. Bartoszy nski, H. Judah, Set Theory, A K Peters, 1995.
[4] K. Ciesielski, Set Theory for the Working Mathematician, London
Math. Soc. Student Texts 39, Cambridge Univ. Press 1997.
[5] E. Grande, Z. Grande, Quelques remarques sur la superposition
F(x, f(x)), Fund. Math. 121 (1984), 199211.
[6] Z. Grande, J. Lipi nski, Un example dune fonction sup-mesurable qui
nest pas mesurable, Colloq. Math. 39 (1978), 7779.
[7] A. B. Kharazishvili, Some questions from the theory of invariant mea-
sures, Bull. Acad. Sci. Georgian SSR 100 (1980) (in Russian).
[8] A. B. Kharazishvili, Sup-measurable and weakly sup-measurable map-
pings in the theory of ordinary dierential equations, J. Appl. Anal.
3(2) (1997), 211223.
[9] A. B. Kharazishvili, Strange Functions in Real Analysis, Pure and Ap-
plied Mathematics 229, Marcel Dekker, 2000.
[10] K. Kunen, Set Theory, North-Holland, 1983.
[11] S. Shelah, Proper Forcing, Lectures Notes in Math. 940, Springer-Verlag
1982.
[12] S. Shelah, Proper and improper forcing, Perspectives in Mathematical
Logic, Springer-Verlag 1998.
16
6
9
5
r
e
v
i
s
i
o
n
:
2
0
0
0
-
0
4
-
1
1
m
o
d
i
f
i
e
d
:
2
0
0
0
-
0
4
-
1
1
[13] J. W.
Sragin, Conditions for measurability of superpositions, Dokl.
Akad. Nauk SSSR 197 (1971), 295298 (in Russian).
17