On the variety generated by completions of representable relation algebras (Q2297227)

The article considers relation algebras in Tarski's sense, in particular the classes $\mathrm{RA}$ of all relation algebras and $\mathrm{RRA}$ of all representable relation algebras. For any $\mathrm{RA}$ $\mathbf{A}$ let $\mathbf{A}^c$ be the completion of $\mathbf{A}$ in the sense of its Boolean part. It was observed by the reviewer [Math. Nachr. 46, 47--55 (1970; Zbl 0182.32301)] that $\mathbf{A}^c$ is an RA; but \textit{I. Hodkinson} [Ann. Pure Appl. Logic 89, No. 2--3, 117--148 (1997; Zbl 0898.03025)] showed that the completion of an $\mathrm{RRA}$ is not necessarily an $\mathrm{RRA}$. Let $\mathrm{RRA}^c$ be the class of all algebras $\mathbf{A}^c$ with $\mathbf{A}\in\mathrm{RRA}$. Let $V$ be the variety determined by $\mathrm{RRA}^c$. \textit{R. D. Maddux} [Algebra Univers. 79, No. 2, Paper No. 20, 32 p. (2018; Zbl 1444.03177)] showed that $\mathrm{RRA}\subseteq V\subseteq\mathrm{RA}$, and he showed that there are many nonrepresntable members of $V$. \textit{H. Andréka} and \textit{I. Németi} [Algebra Univers. 80, No. 3, Paper No. 30, 14 p. (2019; Zbl 1468.03080)] showed that there are continuum many varieties between $V$ and $\mathrm{RA}$. The main result of the present article is that $\mathrm{RRA}$ is not finitely axiomatizable over $V$. The proof consists of showing that $\mathrm{RRA}$ contains an ultraproduct of members of $V\backslash\mathrm{RRA}$, and this is done using graphs, as follows. For $k<\omega$ and $0<n<\omega$ a graph $G^k_n$ is constructed, and associated with it is an $\mathrm{RRA}$ $\mathbf{A}(G^k_n)$. Then the ultraproduct $\mathbf{A}^k=\prod_{0<n<\omega}\mathbf{A}(G^k_n)/D$ is formed. Then let $\mathbf{C}^k=(\mathbf{A}^k)^c\in V$. It is shown that $\mathbf{C}^k\notin\mathrm{RRA}$. However, $\prod_{k<\omega}\mathbf{C}^k\in\mathrm{RRA}$.