A converse to the Sholander embedding (Q1174172)

A median semilattice is a meet semilattice \(M\) such that (i) every principal lower end \((x]=\{a\in M; y\leq x\}\) is a distributive lattice and (ii) any three elements of \(M\) have an upper bound whenever each pair of them have an upper bound. \textit{M. Sholander} [Proc. Am. Math. Soc. 5, 808--812 (1954; Zbl 0056.26201)] proved that any median semilattice \(M\) may be embedded into a distributive lattice \(L\) in such a way that \(M\) is a lower end in \(L\) and every element of \(L\) is the join of finitely many elements of \(M\). The converse is not true in general: Not every lower set of a distributive lattice need be a median semilattice. Example: Let \(L\) be the power set lattice of \(\{a,b,c\}\) and put \(M:=L\backslash\{a,b,c\}\); then \(M\) is obviously a lower end in \(L\), but the singleton sets have no upper bound in \(M\) while any two of them have. The present paper provides several equivalent conditions for a lower set \(M\) of a distributive lattice \(L\) which (singly) guarantee that \(M\) is indeed a median semilattice under the meet operation inherited from \(L\).