Abstract
We show that every arrayof elements in a pointwise compact subset of the Baire-1 functions on a Polish space, whose iterated pointwise limit
exists, is converging Ramsey-uniformly. An array (x(i,j)i<j) in a Hausdorff space
is said to converge Ramsey-uniformly to some x in
, if every subsequence of the positive integers has a further subsequence (mi) such that every open neighborhood U of x in
contains all elements x(mi,mj) with i<j except for finitely many i.
It is a well known consequence of Ramsey's
Theorem that every array
(aij)i<j of real numbers with
for some
has the following property: There is a
subsequence (mi) so that for all
there is an
such that
for all n<mi<mj. This result
generalizes easily to Hausdorff spaces which satisfy the first
countability axiom.
The purpose of our note is to show that a corresponding result holds for the space
of functions of the first Baire-class
on a Polish
space
,
given the topology of pointwise convergence.
Let us say that an array
of elements in a Hausdorff space
converges Ramsey-uniformly to some
,
if every subsequence of
has a further
subsequence (mi) such that for every open neighborhood U of x in
there is an
so that
for all n<mi<mj.
With this notation we can state our main result as follows:
A topological space
is Polish, if it is homeomorphic to a complete separable metric space.
A real-valued function is of the first Baire-class on
,
if it
is the pointwise limit of a sequence of continuous functions on
.
It is a fundamental
result of Bourgain, Fremlin and Talagrand [2] that
is
an angelic space, if
is Polish. A Hausdorff space
is angelic, if for every
relatively compact subset A of
each point in the closure of A is the limit of a sequence in A
and if relatively countably compact sets in
are relatively compact. In angelic spaces the notions of
(relative) compactness, (relative) countable compactness and (relative) sequential compactness coincide.
Further basic results about angelic spaces can be found in [7].
Theorem 1 strengthens -- in the case of functions of the first Baire-class on a Polish space -- a result of Boehme and Rosenfeld [1, Theorem 1], which we phrase for our purposes as follows:
Lemma 2 was also obtained independently -- in the
-setting -- by Rosenthal [8, Theorem 3.6].
From Theorem 1 and a result by Odell and Rosenthal [6] we obtain the following Banach space corollary:
The proof of Theorem 1 utilizes Lemma 2 to extract ``nice'' converging subsequences out of the given array (x(i,j)). We use Ramsey theory to produce the subarray for which one obtains Ramsey-uniform convergence.
If M is an infinite subset of
,
will denote the set of
all infinite
subsets of M. We give
the topology, which is inherited
by considering
as a subspace of
endowed with the product topology.
A subset
is called a Ramsey set, if for all
there
is an
such that either
or
.
It is known that analytic (and coanalytic) subsets of
are Ramsey sets [3,9]. For a proof
of this result, some history and more general results see [5].
I would like to thank D. Alspach, E. Odell and H. P. Rosenthal for useful discussions.
We postpone the proof of the lemma and proceed with the proof of the
theorem. Since
is coanalytic,
is a Ramsey set. Let
.
We can thus find
so that
or
.
Lemma 2 shows that the first alternative holds. Moreover,
Lemma 2 asserts that
for
some
.
Suppose now the conclusion of Theorem 1 fails for
.
Then there is an open neighborhood
U of x and a subsequence
with
We now construct a subsequence
inductively as follows:
Let
and
.
Once
have been chosen, we define
n2k+1 and n2k+2 as follows:
If k is odd, we choose an
so that
and let
.
If k is even, we can find an
with
and
then let
.
On the one hand
the sequence
(x(n2k-1,n2k)) is pointwise convergent, on the other hand it contains
two subsequences converging to x and y respectively. This yields a
contradiction.
Proof of Lemma 4:
The proof of Lemma 4 uses techniques similar to those employed in
[10].
Let Y be the set of all real-valued arrays
(a(i,j))i<j, endowed with the topology of pointwise convergence.
We set
and denote by
the canonical map
defined by
.
Since
is a
Borel-measurable map and
is Polish,
is analytic in Y
(see [4, §38]). Consequently
is
analytic in Z.
We define a set
as follows:
Consequently
is analytic in Z. We let
be the projection of Z onto its first coordinate. One can see
easily that the complement of
is
equal to
.
Thus
is analytic in
as the continuous image of an analytic
set in Z (see [4, §38]).
Problem: Does Theorem 1 hold for arbitrary angelic spaces?
Lemma 2 reduces this problem to the apparently open question, whether the set
,
defined at the beginning of the proof, is still a Ramsey set for arbitrary angelic spaces.
Helmut Knaust