Let $H$ be a group with presentation $\langle h_1, \dots, h_n \mid r_1 = \dots = r_m = 1\rangle$. If there are $g_1, \dots, g_n \in G$ which satisfy the relations $r_1, \dots, r_m$, when is $\varphi : H \to G$ generated by $\varphi(h_i) = g_i$ an isomorphism? In particular, is there an effective way to check that the elements $g_i$ don't satisfy any additional relations?

  • $\begingroup$ By Von Dyck's theorem, such a map is surjective onto $\langle g_1,\cdots,g_n\rangle$. If $G\neq\langle g_1,\cdots,g_n\rangle$, then certainly $\varphi$ cannot be an isomorphism. On the other hand, if $G=\langle g_1,\cdots,g_n\rangle$, it reduces to checking the kernel of the map $\varphi$. I do not know any efficient ways to do this, and suspect that it's difficult. $\endgroup$ – KReiser Sep 26 '13 at 4:17

If we could effectively to check that $g_i$ don't satisfy additional relations, then we can decide whether $G\cong H$. But the problem of isomorphism is undecidable.

  • $\begingroup$ In particular, it is unsolvable when $G$ is the trivial group. $\endgroup$ – Derek Holt Sep 26 '13 at 12:48
  • $\begingroup$ @Derek Holt: Yes, of course. $\endgroup$ – Boris Novikov Sep 26 '13 at 13:17
  • $\begingroup$ Of course, if we know that $H$ and $G$ belong to specific classes of groups, then not all is lost! $\endgroup$ – user1729 Sep 27 '13 at 13:18
  • $\begingroup$ @ user1729: For example, if they are Abelian :-) $\endgroup$ – Boris Novikov Sep 27 '13 at 13:30
  • $\begingroup$ @BorisNovikov: Is there an algorithm in the abelian case or do we just know that it is a decidable problem. $\endgroup$ – Michael Albanese Dec 27 '13 at 23:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.