Let $G$ be a polycyclic group and assume that every finite quotient of $G$ is nilpotent.

Then $G$ is nilpotent.

First some preliminaries:

  1. Every infinite polycyclic group contains a free abelian normal subgroup.

  2. On a polycyclic group one can use 'Noetherian' induction:

If $G$ is polycyclic and doesn't have a property P, we can assume every proper quotient of $G$ has the property P, while $G$ doesn't. (see the book Polycyclic groups by Daniel Segal)

  1. If $G$ is finitely generated and nilpotent of class $k$, we can define the Hirsch length of $G$ as the sum $\sum_{i =1}^{k}m_i = h(G)$ where $m_i$ is the rank of the free part of $C^i(G)/C^{i+1}(G)$.

  2. If $G$ is polycyclic, then we can define the Hirsch number as the number of infinite factors in a witness series of $G$'s polycyclicity. This number is the same for all such series, and it equals the Hirsch length from above when $G$ is also nilpotent (and hence finitely generated).

Here's where I got to:

If $G$ is finite, by assumption it is nilpotent. So we can assume $G$ is infinite.

Let $A \triangleleft G$ be a free abelian subgroup.

Let $p$ denote a prime number; if $G/A^p$ is trivial, $G$ is abelian and hence nilpotent.

So we can assume that for every $p$, $G/A^p$ has the property that if every finite quotient is nilpotent, then so is $G/A^p$.

But if $T/A^p \triangleleft G/A^p$, and $G/A^p / T/A^p \cong G/T$ is a finite quotient, by assumption $G/T$ is nilpotent. Therefore for every $p$, $G/A^p$ is nilpotent.

Thus, $\forall p$ prime $\exists n_p \in \mathbb{N}$ with $C^{n_p}(G) \subset A^p$.

Now, each such $G/A^p$ is polycyclic, and nilpotent, and hence finitely generated. Thus $h(G/A^p) \leq h(G)$. From here I suspect we can bound the sequence $n_p$ to get that there is $m \in \mathbb{N}$ s.t $C^{m}(G) \subset \cap_{p} A^p$.


  1. Can I bound this sequence $n_p$?

  2. Why is it true that $\cap_p A^p = \{e\}$?


I guess $C^i(G)$ is the lower central series of $G$?

Since $G/A$ is nilpotent, we have $C^m(G) \le A$ for some $m$. Now, for any prime $p$, $A/A^p$ is elementary abelian of order $p^d$, where $A$ is free abelian of rank $d$. So the maximum length of a chain of subgroups of $A/A^p$ is $d$. Hence, since $G/A^p$ is nilpotent, we must have $C^{m+d}(G) \le A^p$. This answers your first question.

Your second question is easy, because elements of $A$ have the form $(x_1,\ldots,x_d)$ for some $x_i \in {\mathbb Z}$, and if $(x_1,\ldots,x_d) \in \cap_{p}A^p$, then $p \vert x_i$ for each $i$ and all primes $p$, so $x_i=0$.


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.