# A malnormal embedding theorem?

Let $Q$ be a recursively presented group. Is it possible to embed $Q$ into a finitely presented group $G$ such that the image of $Q$ is malnormal in $G$?

Note that a subgroup $H$ of $G$ is malnormal if $H^g\cap H\neq 1$ implies that $g\in H$. That is, $H$ intersects each of its proper conjugates trivially.

This seems far to strong (far too good!) to be true. However, I cannot find anything in the literature that even hints at the existence or non-existence of such an embedding.

(More generally, I am wondering if there is some kinds of embedding which preserves the centraliser of a given element of $Q$. That is, if we fix $x\in Q$ then can $Q$ be embedded in a finitely presented group $G$ such that $C_Q(x)=C_G(x)$ (or even $C_Q(x)\cong C_G(x)$)? Malnormality guarantees this, and is a more interesting question, but I thought I should mention this too...)

As I wrote in Remark 5.23 cited by Ian Agol, the embedding from that paper should answer the first question. It certainly preserves all centralizers, that is basically proved in the paper. That is if $G$ is a finitely generated recursively presented group, $Q>G$ is the group constructed in that paper, then for every $x\ne 1$ in $G$, $C_Q(x)=C_G(x)$. In fact the version of Higman embedding in Olʹshanskii, A. Yu.; Sapir, M. V. The conjugacy problem and Higman embeddings. Mem. Amer. Math. Soc. 170 (2004), no. 804 satisfies this property too. If I had more time, I would list lemmas from that paper which imply that. I am not sure if that embedding answers your first question. It might.