On the uniqueness of factors of amalgamated products (Q4532308)

The authors study some conditions under which an amalgamated product \(A*_CB\) of groups \(A\) and \(B\) (with subgroup \(C\)) is uniquely determined by the groups \(A\) and \(C\). That is, conditions under which \(A*_CB=A*_CD\) implies that \(B\cong D\) (or \(B=D\)). To formulate the main results of the paper let us define the `moves of type 1 and type 2' studied in the paper. Suppose \(B=B_1*_{C_1}B_2\) where \(C_1\leq C\leq B_1\) and \(a^{-1}C_1a\leq C\) for some \(a\in A \). Then \(A*_CB=A*_CD\) holds, where \(D=B_1*_{a^{-1}C_1a}(a^{-1}B_2a)\), and we say that the splitting \(A*_CD\) is obtained from the splitting \(A*_CB\) by a move of type 1. Further, suppose \(B=\langle B_1,t\mid tC_1t^{-1}=C_1'\rangle\) is an HNN-extension with the base group \(B_1\) and associated subgroups \(C_1\) and \(C_1'\), where \(C_1\leq C\leq B_1\) and \(a^{-1}C_1a\leq C\) for some \(a\in A\). Then \(A*_CB=A*_CD\) holds, where \(D=\langle B_1,ta\mid(ta)a^{-1}C_1t(ta)^{-1}=C_1'\rangle\), and we say that the splitting \(A*_CD\) is obtained from the splitting \(A*_CB\) by a move of type 2. Theorem 1. Suppose that \(G=A*_CB=A*_CD\) where \(G\) and \(C\) are finitely generated groups. Then the splitting \(G=A*_CD\) can be obtained from the splitting \(G=A*_CB\) after a finite number of moves of types 1 or 2. Corollary. Let \(G=A*_CB\) where \(G\) and \(C\) are finitely generated and suppose that for any \(C_1\leq C\) and \(a\in A\) with \(aC_1a^{-1}\leq C\) there exists an element \(c\in C\) such that \(ac_1a^{-1}=cc_1c^{-1}\) for all \(c_1\in C\). Then \(G=A*_CD\) implies that there is an isomorphism \(\phi\colon B\to D\) such that \(\phi|_C=\text{Id}_C\). Recall that a subgroup \(C\) of \(A\) is said to be a malnormal subgroup if \(aCa^{-1}\cap C=1\) for all \(a\in A-C\). Theorem 2. Let \(G\) be a finitely generated group and \(C\not=1\) be a malnormal subgroup that does not lie in a proper free factor of \(G\). If \(G\) is a non-trivial amalgamated product over \(C\), then there exists a unique decomposition \(G={*_C}_{i=1}^nG_i\) where \(C\not=G_i\), such that for any splitting \(G=A*_CB\) we have \(A={*_C}_{i\in I_1}G_i\) and \(B={*_C}_{i\in I_2}G_i\), where \(I_1\cup I_2=\{1,\dots,n\}\) and \(I_1\cap I_2=\emptyset\).