A (possibly infinite) group $G$ is noetherian if it satisfies the following equivalent conditions:

(1) every subgroup of $G$ is finitely generated,

(2) there is no infinite strict ascending chain of subgroups,

(3) any non-empty collection of subgroups has a maximal element under inclusion.

Note that this property is hereditary: if $G$ is noetherian then so are all its subgroups. Also, being noetherian is equivalent to require that the poset of subgroups $(Sub(G),\subseteq)$ is a well-order.

The order type of a well-order $(S,\leq)$ is defined as the supremum of the order type of its linear extensions. By extension, we may define the order type of a noetherian group $G$, denoted by $ot(G)$, as the order type of $(Sub(G),\subseteq)$. We may then ask the following questions:

(1) What ordinals $\alpha$ can be obtained as the order type of a noetherian group?

(2) Is it possible to reduce that question to noetherian simple groups? For instance, if $H$ is a normal subgroup of $G$, is it true that $ot(G) = ot(H) + ot(G/H)$?

  • $\begingroup$ Have you got a communicable intuition what the order-type of a noetherian group may be, or why we can restrict our attention to simple noetherian groups? $\endgroup$ Apr 10, 2014 at 12:19
  • $\begingroup$ Not really, but I would expect the question to be either very easy or very difficult depending on how you interpret it.. As I'm not you I can't tell :) $\endgroup$
    – NisaiVloot
    Apr 10, 2014 at 12:49

1 Answer 1


Here's an argument handling a special case. Consider a fixed poset $P = (S,\leq)$. Say that a (well-founded) forest $F$ is a branching-representation of $P$ if:

  1. the nodes of $F$ are bijectively labelled by the elements of $P$;

  2. for each node $u$ of $F$, its children are ???


We may then assign an ordinal $o(u)$ to each node of $T$ inductively: (i) if $u$ is a leaf then $o(u) = 1$, (ii) if $u$ has children $(u_i)$, then $o(u) = \sum_i o(u_i)$. The order type of $T$ is then defined as $o(T) = o(r)$ where $r$ is the root.

Is the following true?

Conjecture: the order type of $P$ equals the minimum of $o(T)$ over all possible branching-representations $T$ of $P$.

Suppose this is true. Consider a poset $P = (S,\leq)$, and let $P^{\natural} = (S',\leq')$ where $S'$ consists of the finite antichains of $S$ and where the order satisfies $X \leq Y$ iff there is an injection $\phi : X \rightarrow Y$ s.t. $u \leq \phi(u)$ (f.a. $u \in X$). Suppose that $ot(P) = \lambda$ with $\lambda$ a limit ordinal, it should then follow that $ot(P^{\natural}) = \omega^{\lambda}$: starting with a b.p. $T$ of $P$, we would construct a b.p. $T'$ for $P^{\natural}$ by 'iterated substitution'.

This might help to analyze the case of ordered free groups. In this case, we have an ordered alphabet $P = (S,\leq)$, and we consider the free group $P^{\sharp}$ over $\{S \cup S^{-1}\}^*$, equipped by the relation $u \preceq v$ iff $u_i \leq v_{\phi(i)}$ for some injection $\phi$. Now, given an element $B = \{b_1,\ldots,b_k\} \in (P^{\sharp})^{\natural}$, we may define the subgroup $\langle B \rangle$ of $P^{\sharp}$ generated by $\cup_{i = 1}^{k} (b_i \downarrow)$.

It follows that the map $\chi : (P^{\sharp})^{\natural} \rightarrow Sub(P^{\sharp})$ defined by $\chi(B) = \langle B \rangle$ is surjective. Consider the equivalence over $(P^{\sharp})^{\natural}$ s.t. $B \sim B'$ iff $\chi(B) = \chi(B')$. Assuming that the equivalence classes of $\sim$ are finite (which seems intuitively true), I think that it would imply $ot(Sub(P^{\sharp})) = ot((P^{\sharp})^{\natural}) = \omega^{\omega^{\omega^{\lambda}}}$ (possibly up to some lower degree terms).

So if the above is correct, you've got at least the ordinals of the form $\omega^{\omega^{\omega^{\lambda}}}$ in question (1).

  • $\begingroup$ Out of curiosity: are you and HeloLobo different people? $\endgroup$ Apr 15, 2014 at 2:55
  • $\begingroup$ Let's assume that we are the same person. If I answer positively I run the risk of being called a 'liar' by certain people based on the profile colors, and if I answer negatively I run the risk of being called a 'liar' by some other people that may have inside info on the users. Therefore, I'd rather not answer :) By the way, it's still a mystery to me how these colors are attributed, is it explained somewhere on the site? $\endgroup$
    – Super8
    Apr 15, 2014 at 16:38
  • $\begingroup$ The profile colors, as far as I know, are assigned randomly. It's against the spirit of this site to use multiple accounts, as the preference is to even use real names. $\endgroup$
    – Lev Reyzin
    Apr 28, 2014 at 19:43

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.