Varieties with equationally definable factor congruences. II (Q2407971)

Assume \(\mathcal{V}\) is a variety with constants \(0\) and \(1\) satisfying \(\mathcal{V}\models 0=1 \rightarrow x = y\). We say that \(\mathcal{V}\) has \textit{left equationally definable factor congruences (left EDFCs)} if there exists a conjunction of equations which defines the kernel of the projection \(\pi_1:A\times B\to A\) by using the central element \((0,1)\) as a parameter. Right EDFCs are defined analogously with \(\pi_2:A\times B\to B\) and the pair \((0,1)\) (same parameter), while for weak EDFCs there is a conjunction of equations which defines the kernel \(\pi_1\) but using both \((0,1)\) and \((1,0)\) as parameters. Finally, twice EDFCs are the conjunction of left EDFCs and right EDFCs. This paper is a sequel to [Algebra Univers. 70, No. 4, 327--345 (2013; Zbl 1316.03018)], where left EDFCs are studied in detail. We will use the terminology introduced in the first part [loc. cit.]. The authors show that all of the aforementioned notions are different by means of appropriate examples. Similarly to the dichotomy between internal and external direct products, statements regarding definability of factorizations can be presented in an intrinsic or extrinsic manner. The latter is the one preferred in this paper (as in the first part [loc. cit.], where the factorization \(A\times B\) is explicit). As an example, it is shown that for any \(\mathcal{V}\) with 0 and 1, the following are equivalent: {\parindent=0.5 cm \begin{itemize} \item[(1)] \(\mathcal{V}\) has left EDFCs; \item [(2)] if \((0,1)\in S\leq A_1\times A_2\) and \(\sigma:S\to B_1\times B_2\) is a homomorphism such that \(\sigma(0,1)=(0,1)\), then there are homomorphisms \(\sigma_1:\pi_1(S)\to B_1\) and \(\sigma_2:S\to B_2\) such that \(\sigma(a_1,a_2)=(\sigma_1(a_1),\sigma_2(a_1,a_2))\) for each \((a_1,a_2)\in S\). \end{itemize}} If \(\mathcal{V}\) is assumed to have Boolean factor congruences (BFCs), there is a unique factor congruence \(\theta_e^A\) such that \(0\mathrel{\theta_e^A}e\), for each central element \(e\) in an algebra \(A\). The intrinsic version of (2) is then: (2') If \(e\) is central in \(A\), \(e\in S\leq A\) and \(\sigma:S\to B\) is such that \(\sigma(e)\) is central in \(B\), then \(x\mathrel{\theta_e^A}y\) implies \(\sigma(x)\mathrel{\theta_{\sigma(e)}^B}\sigma(y)\) for every \(x,y\in S\). Using extrinsic versions enabled the authors to lift the hypothesis of BFCs in many results of [loc. cit.]. They complete the characterization of left EDFCs and provide many equivalents to weak EDFCs and twice EDFCs, of Mal'cev type and congruence-theoretic ones. They prove that the center can be defined using a set of \(\forall\exists\forall\) formulas for varieties having weak EDFC, and state that this can be improved to \(\forall\exists\) in the case of twice EDFCs. Both bounds on the complexity of definitions are shown to be sharp. They also show that the class of directly indecomposable algebras in varieties with weak DFCs over a finite language can be axiomatized using a \(\forall\exists\forall\) sentence.