Representing subalgebras as retracts of finite subdirect powers (Q783295)

For a class \(\mathcal{K}\) of algebras with the same function symbols, \(H \mathcal{K}\) denotes the class of all homomorphic images of algebras in \(\mathcal{K}\), \(S \mathcal{K}\) the class of subalgebras, and \(P_{\text{fs}}\) the class of finite subdirect products of members of \(\mathcal{K}\). \textit{P. M. Neumann} [Bull. Lond. Math. Soc. 2, 91 (1970; Zbl 0215.10304)] proved that for a finite nilpotent group \(\mathbf{A}\), we have \(S \{\mathbf{A}\} \subseteq H P_{\text{fs}} \{\mathbf{A} \}\). The present paper generalizes this result from finite groups to arbitrary algebras. For a general universal algebra, the concept of nilpotency considered here is \textit{supernilpotency with respect to the \(2\)-term higher commutator} (for short, \textit{\(2\)-term-supernilpotency}). A group satisfies this condition if and only if it is nilpotent. The authors prove that for every \(2\)-term-supernilpotent algebra \(\mathbf{A}\), we have \(S \{ \mathbf{A} \} \subseteq H P_{\text{fs}} \{ \mathbf{A} \}\); in fact the homomorphic images can even be taken to be retracts. Hence, if \(\mathcal{K}\) is a class of \(2\)-term-supernilpotent algebras with \(H P_{\text{fs}} \mathcal{K} \subseteq \mathcal{K}\), then \(S \mathcal{K} \subseteq \mathcal{K}\). As another corollary, the authors obtain that for a finite algebra \(\mathbf{A}\) in a variety \(\mathcal{V}\) all of whose members are \(2\)-term-supernilpotent, we have \[ HSP_{\text{fin}} \{\mathbf{A}\} \subseteq H P_{\text{fs}} \{ \mathbf{A} \}. \] A second thread of results concerns what are called \textit{axiomatic formations} of algebras: Suppose that \(\mathcal{K}\) is such that every algebra in \(\mathcal{F} := H P_{\text{fs}} \mathcal{K}\) is \(2\)-term-supernilpotent and \(\mathcal{F}\) can be defined by a set of first-order formulae. Then \(\mathcal{F}\) is a variety. Finally, the authors exhibit a six element nilpotent expanded group with \(S \{\mathbf{A}\} \not\subseteq H P_{\text{fs}} \{\mathbf{A} \}\). An outstanding feature is that the main result makes no assumption on the algebra, such as generating a congruence modular variety or satisfying a Mal'cev condition, other than the algebra being \(2\)-term-supernilpotent.