Abstract. We exhibit a counterexample to a recent assertion concerning the subgroup separability of groups in the title. The example also serves as a simplification of work of Gitik and Rips.Denote the generalized free product of groups A and B amalgamating the common subgroup H by A * H B. Let Q be any class of groups, for example the finite groups, the finite p-groups or the nilpotent groups. A group G is said to be residually Q iff it has the following property where g ∈ G : g = 1 if and only if g = 1 in all homomorphic images G of G in Q.Likewise G is called subgroup separable (also called LERF) iff for all g ∈ G and all finitely generated subgroups H of G : g ∈ H if and only if g ∈ H in all finite images G of G.Separation properties such as these are often shared by nilpotent groups and free groups, in part because free groups are residually finitely generated torsion free nilpotent [9]. Furthermore, the proof that each free product F of free groups with cyclic amalgamation is residually finite [2] and potent [1] uses the passage from F down to generalized free products of nilpotent groups. Now in [3] it was shown that each generalized free product F as above is subgroup separable. Consequently one might well expect a similar result for generalized free products of nilpotent groups. The paper [13] attempts to prove this result in the same fashion as [3] by establishing two propositions and then appealing to Lemma 1 of [3]. Unfortunately Proposition 3.5 of [13], whilst true, seems to be insufficient.Counterexamples of the type claimed can readily be obtained by selecting suitable subgroups of the generalized free products constructed in [5]. Consider, for instance, G = A * a B where B = a, c : [a, c] = 1 and
Thus G† is the free product of a finitely generated torsion free nilpotent group of class 2 with a free abelian group amalgamating a 2 , a maximal cyclic subgroup