Splitting Tychonoff cubes into homeomorphic and homogeneous parts (Q1985651)

To prove that the unit interval \(\mathbb{I}=[0,1]\) and, for every infinite cardinal \(\tau\), the Tychonoff cube \(\mathbb{I}^{\tau}\) can be partitioned into two homogeneous and homeomorphic parts, the authors use a set \(A\subseteq\mathbb{R}\), constructed in [\textit{J. van Mill}, Compos. Math. 46, 3--13 (1982; Zbl 0514.54011)], having the following properties: \(\mathbb{Q}\subseteq A\), \(B=\mathbb{R}\setminus A\) is dense in \(\mathbb{R}\), \(A+\mathbb{Q}=A\) and if \(\phi:\mathbb{R}\to\mathbb{R}\) is defined by \(\phi(x)=x+\pi\) for each \(x\in\mathbb{R}\), then \(\phi(A)=B\). The authors show that the subspaces \(E=(0,1)\cap A\) and \(H=\mathbb{I}\setminus E= \{0,1\}\cup ((0,1)\cap B)\) of \(\mathbb{I}\) are both homogeneous and homeomorphic. Given an infinite cardinal \(\tau\), the subspaces \(E\times\mathbb{I}^{\tau}\) and \(H\times\mathbb{I}^{\tau}\) of \(\mathbb{I}\times\mathbb{I}^{\tau}\) are also homogeneous and homeomorphic; hence \(\mathbb{I}^{\tau}\), being homeomorphic to \(\mathbb{I}\times\mathbb{I}^{\tau}\), admits a partition into two homogeneous and homeomorphic parts. Assuming \(ZFC\), the authors prove that if \(\tau\) is an uncountable cardinal and \(G\) is a dense subspace of \(\mathbb{I}^{\tau}\) such that \(\mathbb{I}^{\tau}\setminus G\) is Lindelöf, then \(G\) is not homeomorphic to a topological group. Next, the authors prove that if \(\tau\) is an uncountable cardinal and \(A\) is a subspace of \(\mathbb{I}^{\tau}\) such that \(A\) is homeomorphic to a topological group, then \(A\) and \(\mathbb{I}^{\tau}\setminus A\) are not homeomorphic. To do this, the authors consider an arbitrary uncountable cardinal \(\tau\) and a subspace \(A\) of \(\mathbb{I}^{\tau}\) such that \(A\) is homeomorphic to a topological group. Assuming that \(A\) and \(B=\mathbb{I}^{\tau}\setminus A\) are homeomorphic, the authors show that both \(A\) and \(B\) are dense pseudocompact subspaces of \(\mathbb{I}^{\tau}\), so \(\mathbb{I}^{\tau}\) is the Čech-Stone compactification of \(A\) and \(B\). The authors give two distinct arguments that this is impossible: (1) by Theorem 4.1 from [\textit{W. W. Comfort} and \textit{K. A. Ross}, Pac. J. Math. 16, 483--496 (1966; Zbl 0214.28502)], the Čech-Stone compactification of a pseudocompact topological group is homeomorphic to a topological group, while \(\mathbb{I}^{\tau}\) is not; (2) since \(A\) and \(B\) are supposed to be homeomorphic, there exists a homeomorphism \(h:\mathbb{I}^{\tau}\to\mathbb{I}^{\tau}\) such that \(h(A)=B\), and this contradicts the Brouwer Fixed-Point Theorem. Finally, the authors notice that it holds in \(ZFC\) that, for every infinite cardinal \(\tau\), the Cantor cube \(\{0,1\}^{\tau}\) contains a subspace \(A\) such that both \(A\) and \(\{0,1\}^{\tau}\setminus A\) are homeomorphic to topological groups. The authors ask the question of whether every non-trivial compact topological group can be split into two homogeneous and homeomorphic parts.