6
$\begingroup$

A metrizable space $X$ will be called a generalized Bernstein set if every closed completely metrizable subspace $C$ of $X$ has cardinality $|C|<|X|$.

It is well-known that the real line contains a (generalized) Bernstein set of cardinality $\mathfrak c$.

Moreover, for every cardinal $\kappa$ with $\kappa^\omega=2^\kappa$ there exist a generalized Bernsten set $X$ of cardinality $|X|=2^\kappa=\kappa^\omega$. In particular, there exists a generalized Bernstein set of cardinality $\beth_{\omega+1}$.

Question. What are possible cardinalities of generalized Bernstein sets? For example, is there a generalized Bernstein set $X$ of cardinality $|X|=\mathfrak c^+$?

$\endgroup$
4
  • $\begingroup$ Taras, could you suggest a reference for the result you mention in paragraph 3? $\endgroup$ Jun 26 '16 at 22:12
  • 1
    $\begingroup$ I do not know such a reference, but know a proof: Given a cardinal $\kappa$ with $\kappa^\omega=2^\kappa$, endow $\kappa$ with the discrete $\{0,1\}$-valued metric and consider the complete metric space $X=\kappa^\omega$ of weight $\kappa$. This space contains at most $2^\kappa$ $G_\delta$-subsets, so we can enumerate the family $\mathcal C$ of completely metrizable subsets of cardinality $\kappa^\omega$ as $\mathcal C=\{C_\alpha\}_{\alpha\in 2^\kappa}$. $\endgroup$ Jun 27 '16 at 4:46
  • 1
    $\begingroup$ Then by transfinite induction for every ordinal $\alpha<2^\kappa$ choose two distinct points $x_\alpha,y_\alpha$ in the set $C_\alpha\setminus\{x_\beta,y_\beta\}_{\beta<\alpha}$. The choice of $x_\alpha,y_\alpha$ is always possible as $|C_\alpha|=\kappa^\omega=2^\kappa>|\alpha+\alpha|$. Then $Y=\{y_\alpha\}_{\alpha<2^\kappa}$ is a generalied Bernstein set of cardinality $|Y|=\kappa^\omega$. $\endgroup$ Jun 27 '16 at 4:50
  • 1
    $\begingroup$ To show that $Y$ is a generalized Bernstein set, assume that $Y$ contains a completely metrizable set $C$ of cardinality $|C|=|Y|=\kappa^\omega$. Then $C=C_\alpha$ for some $\alpha<2^\kappa$ and then $x_\alpha\in C\setminus Y$, which means that $C\not\subset Y$. But this contradicts the choice of $C$. $\endgroup$ Jun 27 '16 at 4:54
2
$\begingroup$

After thinking a night on this question and waking up, I realized that the answer is almost trivial: there are restrictions on possible cardinalities of generalized Bernstein set.

Any metrizable space $X$ of density $\kappa$ has cardinality $|X|\le\kappa^\omega$ and contains a discrete (and hence completely metrizable) subspace $D$ of cardinality $|D|=\kappa$. If $\kappa=\kappa^\omega$ (which is the case for the cardinal $\mathfrak c^+$), then $X$ cannot be a generalized Bernstein set.

On the other hand, there is also a less trivial fact involving (even more) generalized Bernstein sets.

Let $\mathfrak c^{+0}=\mathfrak c$ and $\mathfrak c^{+(n+1)}=(\mathfrak c^{+n})^+$ for $n\in\omega$. So, $\mathfrak c^{+n}$ is the $n$th successor cardinal of $\mathfrak c$. Let also $\mathfrak c^{+\omega}=\sup_{n\in\omega}\mathfrak c^{+n}$.

By induction it can be shown that each metrizable space $X$ of cardinality $|X|<\mathfrak c^{+\omega}$ contains a subset $B\subset X$ of cardinality $|B|=|X|$ such that for every uncountable Polish subspace $P\subset X$ both sets $P\cap B$ and $P\setminus B$ are not empty. This implies that $B$ contains no uncountable Polish subspaces, so $B$ can be considered as a generalized Bernstein set (in a somewhat different sense).

$\endgroup$
2
  • 1
    $\begingroup$ Concerning your last paragraph: Bill Weiss has shown that under certain set-theoretic assumptions ("there are no measurable cardinals" will do), every Hausdorff space contains a subset $B$ with the property you describe. (See his article "Partitioning topological spaces" in Mathematics of Ramsey Theory -- unfortunately, I do not know of a free online version.) On the other hand, using a supercompact cardinal Shelah has found a model with a space $X$ admitting no such subspace. (See Sh 460 and search it for "Cantor discontinuum problem" -- it's somewhere near the end.) $\endgroup$
    – Will Brian
    Jun 27 '16 at 16:05
  • $\begingroup$ Thank you, Will. I looked at Sh460 (shelah.logic.at/files/460.pdf) and it gives a consistency of the existence of the Berstein set in any topological space. On the other hand, the paper Sh668 (arxiv.org/pdf/math/9906025v1.pdf) proves the consistency of the existence of a compact Hausdorff space without a Bernstein set. So, what I have asked appeared to be just another bicycle :( $\endgroup$ Jun 27 '16 at 17:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.