5
$\begingroup$

Let $k$, $m$ be natural numbers, and $C_m:=\mathbb{Z}/ m \mathbb{Z}$ be the cyclic group of order $m$.

Let $N_{k, \, m}$ be the cardinality of the following set: $$\{(a_1, \ldots, a_k) \in (C_m)^k \; | \; \text{the subset } \{a_1, \ldots, a_k\} \text{ generates } C_m \}.$$

Question. Is a closed formula for $N_{k,\, m}$ known? If so, what is a reference?

Clearly $$N_{k, \, m} \geq m^k - (m- \phi(m))^k,$$ with equality when $m=p$ is a prime number, in which case $$N_{k, \, p} = p^k -1.$$ But what about the general situation? This looks to me as a natural problem, so I suspect that the answer is well-known to the people working in the field. However, I am not an expert in Combinatorics, so I may be missing some elementary solution.

Any answer and/or reference to the relevant literature will be appreciated.

$\endgroup$
4
  • 1
    $\begingroup$ In general you can use Möbius inversion on the poset of subgroups of a group to answer questions like this. In fact, I think Philip Hall developed the theory of poset Möbius inversion for precisely this problem. $\endgroup$ Commented Jan 12 at 16:48
  • $\begingroup$ @SamHopkins: very nice, thanks. Could you please elaborate a bit on this (maybe in an answer)? Do you have any references? $\endgroup$ Commented Jan 12 at 16:49
  • 2
    $\begingroup$ The original reference is P. Hall, The Eulerian functions of a group, Quart. J. Math. Oxford Ser. 7 (1936), 134–151 (doi.org/10.1093/qmath/os-7.1.134). See, e.g. arxiv.org/abs/2109.05260 for a modern discussion of Hall's work. I would guess that it appears as an exercise in Chapter 3 of Stanley's "Enumerative Combinatorics," Vol. 1 but for some reason am having trouble locating it there. Maybe see also this prior MO question: mathoverflow.net/questions/410044. $\endgroup$ Commented Jan 12 at 16:56
  • 3
    $\begingroup$ TeX note: \operatorname is really only meant for operators, and the spacing is unlikely to align well with normal inter-word spacing. Instead of \operatorname{the subset} \; \{a_1, \ldots, a_k\} \, \operatorname{generate} \, C_m it is probably better to use \text{the subset } \{a_1, \ldots, a_k\} \text{ generates } C_m. I edited accordingly. (Actually you can use math mode in text in math mode: \text{the subset $\{a_1, \dotsc, a_k\}$ generates $C_m$}. But this seems to be uncommon.) $\endgroup$
    – LSpice
    Commented Jan 12 at 18:08

1 Answer 1

7
$\begingroup$

I believe the answer is $\sum_{d\mid m}d^k\mu(m/d)$ where $\mu$ is the Möbius function. The reason is that $C_m$ has a unique subgroup of order $d$ for each divisor $d$ of $m$, and we can partition the $m^k$ $k$-tuples by the subgroup they generate. So, if $f(n)$ is the number of $k$-tuples generating $C_n$, then we should get $m^k=\sum_{d\mid m} f(d)$ and hence, by Möbius inversion, $f(m)=\sum_{d\mid m}d^k\mu(m/d)$.

Added. As pointed out by @SamHopkins, this problem was essentially solved by P. Hall for general finite groups using the Möbius function for the subgroup lattice, but for cyclic groups nothing that fancy is needed. The general argument is basically the same. You can count $k$-tuples by the subgroups they generate. So if $\Lambda(G)$ is the subgroup lattice of $G$ and $f(H)$ is the number of $k$-tuples of elements of $H$ generating $H$, then you have that $|G|^k = \sum_{H\leq G}f(H)$ and so $f(G) = \sum_{H\leq G}|H|^k\mu(H,G)$ where $\mu$ is the Möbius function of $\Lambda(G)$. For a cyclic group, this boils down to the classical number theoretic Möbius function and inversion formula.

$\endgroup$
3
  • 6
    $\begingroup$ Equivalently, the answer is $m^k\prod_{p|m}(1-p^{-k})$, where $p$ ranges over all primes dividing $m$. This function is known as Jordan's totient function (en.wikipedia.org/wiki/Jordan%27s_totient_function). $\endgroup$ Commented Jan 12 at 19:13
  • 1
    $\begingroup$ @RichardStanley, thanks for that. I didn’t know that name. $\endgroup$ Commented Jan 12 at 23:00
  • 1
    $\begingroup$ Thanks for the answer, I checked the details and everything seems ok. I took the liberty to slightly edit your post, feel free to roll back if you don't like it. $\endgroup$ Commented Jan 18 at 14:22

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .