11
$\begingroup$

Let $F_n$ be a free group on $n$ generators. Fix a prime $p$. Let $\gamma_k^p(F_n)$ be the mod $p$ lower central series, i.e. the inductively defined series $$\gamma_0^p(F_n) = F_n \quad \text{and} \quad \gamma_{k+1}^p(F_n) = (\gamma_{k}^p(F_n))^p [F_n,\gamma_k^p(F_n)].$$ Observe that the quotients $\gamma_{k}^p(F_n) / \gamma_{k+1}^p(F_n)$ are abelian $p$-groups. Moreover, the quotients $N_n^p := F_n / \gamma_{k}^p(F_n)$ are $p$-groups of nilpotency class $k$. They are universal with this property -- if $G$ is a $p$-group of nilpotency class $k$ and $g_1,\ldots,g_n \in G$, then there is a unique homomorphism $N_n^p \rightarrow G$ taking the generators of $N_n^p$ to the $g_i$.

Question : What are $H_2(N^p_n;\mathbb{Z})$ and $H_2(N^p_n;\mathbb{F}_p)$?

$\endgroup$
6
  • $\begingroup$ Edited it because I realized that I asked about general $H_k$ when I really only care about $H_2$ (as indicated in the title). $\endgroup$
    – Tony
    Mar 5, 2012 at 16:14
  • 1
    $\begingroup$ Did you try the Hopf Formula? Assuming you know the generators/relations. $\endgroup$ Mar 5, 2012 at 16:41
  • $\begingroup$ I tried. The Hopf formula seems to give me the answer for the quotients of the free group by the ordinary lower central series, but I couldn't make it work for the mod p version (for instance, the relations do not live in the commutator subgroup). $\endgroup$
    – Tony
    Mar 5, 2012 at 21:17
  • 2
    $\begingroup$ I don't know the answer to your question, but I don't think that your comments about these groups being universal are quite correct. The quotients in your series are all elementary abelian $p$-groups. You need to define the $p$-nilpotency class of a group using central series with elementary abelian $p$-groups as factors. Your groups are then universal with respect to groups of $p$-nilpotency class $k$. $\endgroup$
    – Derek Holt
    Mar 5, 2012 at 23:37
  • 1
    $\begingroup$ Dear @Tony: It might still be a good idea to add a correction to the question, as not everyone reads the comments. $\endgroup$ Sep 3, 2013 at 11:51

1 Answer 1

2
$\begingroup$

For brevity, I denote by $F$ the free group on $n$ generators, $\lambda_k$ the $k$th terme of the $p$-lower series of $F$, and $N_k$ the relatively free group $F/\lambda_k$.

Also I use the following 5- terme exact sequence (see Generators of sections of free groups): $$0 \longrightarrow H_2(N_k;\mathbb{Z}/p) \longrightarrow \lambda_k/[F,\lambda_k]{\lambda_k}^p \longrightarrow H_1(F;\mathbb{Z}/p) \longrightarrow H_1(N_k;\mathbb{Z}/p) \longrightarrow 0.$$ Now since $N_k$ is minimally $n$-generated $p$-group, it follows from $ H_1(N_k;\mathbb{Z}/p) = N_k/[N_k,N_k]{N_k}^p$ that $ H_1(N_k;\mathbb{Z}/p) \cong (\mathbb{Z}/p)^n$, and also we have $H_1(F;\mathbb{Z}/p) = F/[F,F]F^p \cong (\mathbb{Z}/p)^n$. Now the above exact sequence reduces to $$0 \longrightarrow H_2(N_k;\mathbb{Z}/p) \longrightarrow \lambda_k/[F,\lambda_k]{\lambda_k}^p \longrightarrow 0$$

so we have $H_2(N_k;\mathbb{Z}/p) \cong \lambda_k/\lambda_{k+1}$, and this is completely determined by the rank $r_k$ of the elementary abalian $p$-group $\lambda_k/\lambda_{k+1}$.

The integer $r_k$ can be determined using Lie methods, see for instance Corollary 18 in "The automorphism group of a finite p-group is almost always a p-group" Journal of Algebra, Volume 312 (2007) G. T. Helleloid and U. Martin. In which it is proved that $$r_k = \sum_{i=1}^{k} 1/i (\sum_{j/i} \mu(i/j) n^j)$$ where $\mu(t)$ denotes the Mobius function.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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