On the complexity of the classification problem for torsion-free abelian groups of rank two. (Q1812471)

The collection of torsion-free abelian groups of rank \(1\leq r\leq n \) can be naturally identified with the set \(S({\mathbb Q}^{n})\) of all non-trivial additive subgroups of \({\mathbb Q}^{n}\). Furthermore, \(S({\mathbb Q} ^{n}) \) is a standard Borel subset of the Polish space \({\mathcal P}({\mathbb Q}^{n}) \) of all the subspaces of \({\mathbb Q}^{n}\), and a standard Borel \(GL_{n}({\mathbb Q} )\)-space under the natural action. The isomorphism relation on \(S({\mathbb Q} ^{n})\) is a countable Borel equivalence relation denoted \(\cong _{n}\). If \(E,F\) are Borel equivalence relations on the standard Borel spaces \(X,Y\) respectively, then \(E\) is {Borel reducible} to \(F\) and we write \(E\leq _{B}F \) if there exists a Borel function \(F:X\rightarrow Y\) such that \(xEy\) if and only if \(f(x)Ff(y)\). The question on the complexity of the classification problem for \(S({\mathbb Q} ^{n})\) can be rephrased as the question whether \((\cong _{1})<_{B}(\cong _{n})\). (Is the classification problem for \(S({\mathbb Q}^{n})\) genuinely more difficult for \(n\geq 2\)?) Denote by \((\cong _{n}^{\ast })\) the restriction of the isomorphism relation to the class of rigid torsion-free abelian groups (in \(S({\mathbb Q}^{n})\)). The main result of this paper is Theorem 1.4. \((\cong _{3}^{\ast })<_{B}(\cong _{2})\). Thus, \(\cong _{2}\) is not a universal countable Borel equivalence relation. This shows that the Hjorth-Kechris conjecture (\((\cong _{n})\sim _{B}E_{\infty }\) for all \(n\geq 2\)) is false. As a consequence, the classification problem for \(S({\mathbb Q}^{3})\) is strictly more complex than that for \(S({\mathbb Q}^{2})\). Everything follows from Theorem 1.6. Let \(\Gamma \) be a countable Kazhdan group and let \(X\) be a standard Borel \(\Gamma \)-space with an invariant probability measure \(\mu\). If \(F:X\rightarrow S(\mathbb{Q}^{2})\) is a Borel function such that \(xE_{\Gamma }^{X}y\) implies \( f(x)\cong _{2}f(y),\) then there exists a \(\Gamma \)-invariant Borel subset \(M\) with \(\mu (M)=1\) such that \(f\) maps \( M \) into a single \(\cong _{2}\)-class.