For a finite abelian $p$-group $G$ we have that $$ G \simeq \mathbf{Z}/(p)^{\lambda_1} \oplus \dotsb \oplus \mathbf{Z}/(p)^{\lambda_r} $$ for some positive integers $\lambda_1 \geq \dotsb \geq \lambda_r$. Note that $G$ is uniquely determined by $p$ and this partition $\lambda = (\lambda_1, \dotsc, \lambda_r)$, so let's call $\lambda$ the type of $G$. For types $\lambda$, $\mu$, and $\nu$, define the Hall number $g_{\mu,\nu}^\lambda(p)$ to be the number of normal subgroups $N \mathrel{\triangleleft} G$ of type $\nu$ such that $G/N$ has type $\mu$. These Hall numbers serve as the structure constants of an associative algebra called the Hall algebra.

It turns out that this algebra is commutative, i.e. $g_{\mu,\nu}^\lambda(p) = g_{\nu,\mu}^\lambda(p)$. The proof of this that I'm looking at, following the more general theory in MacDonald's Symmetric Functions and Hall Polynomials, goes like this: You realize that we're looking at the category of finite-length modules over $\mathbf{Z}_p$, the $p$-adic integers. The Prüfer $p$-group $\mathbf{Z}(p^\infty)$ is the injective hull of $\boldsymbol{k} = \mathbf{Z}/(p)$ in this category, and the functor $\mathrm{Hom}({-},\mathbf{Z}(p^\infty))$, via Matlis duality, gives you a bijection of the short exact sequences in question, so $g_{\mu,\nu}^\lambda(p) = g_{\nu,\mu}^\lambda(p)$.

Proving this can also be approached by developing the theory of characters of finite abelian groups, section 3 in particular. But this is really the same approach in a different language: $\mathbf{Z}(p^\infty)$ plays the role of $S_1$ in this context. But in either approach, we're introducing some heavy stuff just to prove a fact about $p$-groups and partitions. Is there a elementary way to prove that $g_{\mu,\nu}^\lambda(p) = g_{\nu,\mu}^\lambda(p)$ in the case of finite abelian $p$-groups?

  • $\begingroup$ You can replace $\mathbf{Z}\left(p^\infty\right)$ in the duality argument by its "finite approximation" $\mathbf{Z}/p^N\mathbf{Z}$, where $p^N$ is an upper bound on the sizes of your $p$-groups. This makes everything more elementary (it is certainly a lot easier to prove that each group $\mathbf{Z}/p^k\mathbf{Z}$ is isomorphic to its "$N$-dual" group $\operatorname{Hom}\left(\mathbf{Z}/p^k\mathbf{Z},\mathbf{Z}/p^N\mathbf{Z}\right)$ when $k \leq N$). $\endgroup$ Feb 22, 2019 at 3:57
  • $\begingroup$ Then we don't have to bring up the Prüfer group at all. :) But then we have to do some manual labor to show the map $\mathrm{Hom}(G, \mathbf{Z}/p^N\mathbf{Z}) \to \mathrm{Hom}(N, \mathbf{Z}/p^N\mathbf{Z})$ is a surjective. And we've still gotta define $\mathbf{Z}_p$ because these are all $\mathbf{Z}_p$ modules. ... or do you? $\endgroup$ Feb 22, 2019 at 4:30
  • $\begingroup$ You don't need to define $\mathbf{Z}_p$; you can read "finite abelian $p$-group" for "$\mathbf{Z}_p$-module" (since all of your $\mathbf{Z}_p$-modules are finite). $\endgroup$ Feb 22, 2019 at 4:52
  • $\begingroup$ Do you mean $G \cong \mathbf{Z}/(p)^{\lambda_1} \oplus \dotsb \oplus \mathbf{Z}/(p^r)^{\lambda_r}$? $\endgroup$ Feb 22, 2019 at 5:41
  • 1
    $\begingroup$ Yes, that's equivalent to what he is writing. (We always have $\left(n\right)^k = \left(n^k\right)$ as ideals.) $\endgroup$ Feb 22, 2019 at 5:50

1 Answer 1


The commutativity of the Hall algebra is saying that you can 'turn short exact sequences around', so is essentially equivalent to having a duality. You don't necessarily need the Prüfer group, though. You can take the 'usual' injective $\mathbb Q/\mathbb Z$ for abelian groups. Then $\mathrm{Hom}_{\mathbb Z}(\mathbb Z/p^n\mathbb Z,\mathbb Q/\mathbb Z)\cong\mathbb Z/p^n\mathbb Z$ is clear, sending a homomorphism $f$ to the image of the cyclic generator $f(1)$.

In the 'algebraic' setting, rather than the 'number theoretic' setting, the same result holds. Here you are taking finite dimensional $k[t]$-modules on which $t$ acts nilpotently; equivalently finite dimensional $k[[t]]$ modules. In this case one can instead use the usual vector space duality $D=\mathrm{Hom}_k(-,k)$. When the field $k$ is finite, the corresponding Hall algebra is symmetric.


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.