Suppose $H_1,H_2,...,H_n$ are solvable normal subgroups of a group $G$. Prove that there exists a solvable normal subgroup $F$ of $G$ containing all $H_i$.

I am struggling with how to come up with such an $F$. My idea is to let $G$ be solvable of degree $n$ and then since $G$ is solvable of degree $n$, each $H_i \leq G$ is also solvable of degree $n$. Then we have that ${e} = H_{11} \trianglelefteq H_{12} \trianglelefteq...\trianglelefteq H_{1n} = H_1$ and for each $H_i, {e} = H_{i1} \trianglelefteq H_{i2} \trianglelefteq...\trianglelefteq H_{in} = H_i$. We also know that since each of these subgroups $H_i$ is normal, $G/H_i$ is also solvable of degree $n$ and we know that each $H_{ij}/H_{ij-1}$ is abelian. Now I think that there is a way to put them all together and still having all quotients be abelian.

Maybe this is the completely wrong way and I should just consider the union of all $H_i$ then since each element is a subgroup of $G$, we know $F = \cup_{i=1}^{n} H_i \leq G$ so all that remains is to show that $F$ is normal in $G$ and also solvable.

Which of these is correct and how can I finish the proof?

  • $\begingroup$ Hint: Suppose $F$ is the subgroup generated by all the $H_i$'s. Can you prove $F$ is normal? Write an element of $F$ as a finite product of $h$'s, then to conjugate by $g$ insert a $gg^{-1}$ between each $h$. $\endgroup$ – Michael Burr May 18 '18 at 0:35
  • $\begingroup$ @MichaelBurr so you are saying the second method I mentioned? Would such an element look like $\{h_1, h_2, ..., h_m\} \in F$? Then we would know that for each of these $h_i$ we have $\forall g\in G gh_ig^{-1}\in H_j$ for some $H_j$ (whichever $h_i$ came from) so since $H_j \leq F, gh_ig^{-1} \in F$? $\endgroup$ – user562567 May 18 '18 at 0:44
  • $\begingroup$ Is the union of subgroups a subgroup? $\endgroup$ – Michael Burr May 18 '18 at 0:46
  • $\begingroup$ Ooh I thought it was, but I see that its not. So how is such an $F$ constructed? I see you said let $F$ be generated by all the $H_i's$, but can you show what an element would look like? $\endgroup$ – user562567 May 18 '18 at 0:49
  • $\begingroup$ It's all finite words in the $H$'s. $\endgroup$ – Michael Burr May 18 '18 at 0:54


  • Let $F$ be the subgroup of $G$ generated by the $H_i$'s. Then an element of $F$ is $$ h_1h_2\dots h_k, $$ where $h_j\in H_{i_j}$.

  • $F$ is normal because $$ g(h_1h_2\dots h_k)g^{-1}=gh_1g^{-1}gh_2g^{-1}\dots gh_kg^{-1}, $$ and each $H_i$ is normal.

  • $F$ is solvable because if $F'$ is the subgroup generated by $H_1,\dots,H_{n-1}$, then, by the argument above, $F'$ is normal in $F$ and the quotient is a subgroup of $H_n$, and therefore solvable. Then proceed by induction.

  • $\begingroup$ Could you tell me if this is where your induction step is going? I want to say that then $F^{k}$ generated by $H_1...H_{n-k}$ is normal in $F^{k}$ and the quotient is a subgroup of $H_n...H_{n-k}$ and is therefore solvable and then I want to show that the same thing holds for $k+1$ or is that not where you are trying to lead me? $\endgroup$ – user562567 May 18 '18 at 1:16
  • $\begingroup$ TThe induction is on the number of subgroups generating $F$, so it looks like your index $H_{n-k}$ should be $n-k-1$. The quotient would be a subgroup of $H_{n-k}$, not all the larger groups. There is a condition that a group $K$ is solvable iff there is a normal subgroup $N$ so that $N$ and $K/N$ are both solvable. $\endgroup$ – Michael Burr May 18 '18 at 1:22
  • $\begingroup$ So the endgame is to get that $F^{n-1} \triangleleft F^{n-2}$ and we know $F^{n-1}$ is generated by just $H_1$ so it is just $H_1$ so its solvable and $F^{n-2}/F^{n-1}$ is solvable would have been shown in the previous step so now we know $F^{n-2}$ is solvable and then we work our way back up until we reach $F$? $\endgroup$ – user562567 May 18 '18 at 1:37
  • $\begingroup$ I also don't see why it would be a subgroup of just $H_{n-k}$. Could you explain? $\endgroup$ – user562567 May 18 '18 at 1:39
  • $\begingroup$ Since $F^k$ is generated by $H_1,\dots,H_{n-k}$, and you're quotienting out by $F^{k+1}$, which is generated by $H_1,\dots,H_{n-k-1}$, all that you're left with is $H_{n-k}$. Note that $H_{n-k+1},\dots,H_n$ might not even be subsets of $F^k$. Your endgame sounds right (although the induction allows you to hide the endgame details). $\endgroup$ – Michael Burr May 18 '18 at 1:54

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.