5
$\begingroup$

For a given positive integer $N$ let $L(N)$ denote the principal congruence subgroup of $\operatorname{SL}(2, \mathbb{Z})$ of level $N$. It is known that $L(N)$ is a finitely generated free group. Let $T(N)$ denote the minimum number of generators. Let $r$ be a number strictly less than $T(N)$.

Question: Is there a congruence subgroup $L$ of $L(N)$ such that the minimum number of generators of $L(N)/L$ is at least $r$?

$\endgroup$
3
  • $\begingroup$ Is there any reason to introduce the variable $r$, rather than asking the same question for $T(N) - 1$? $\endgroup$
    – user44191
    Commented Aug 31, 2021 at 12:10
  • $\begingroup$ I'm not sure whether one or the other is a more reasonable question, but it is an at least apparently different question: the minimum number of generators being at least $r$ is apparently weaker than the minimum number of generators being at least $T(N) - 1$ if $r$ is strictly less than $T(N) - 1$. $\endgroup$
    – LSpice
    Commented Aug 31, 2021 at 13:32
  • $\begingroup$ @user44191 Yes. I would like to understand for what values of r the answer is yes. $\endgroup$ Commented Sep 1, 2021 at 18:03

1 Answer 1

9
$\begingroup$

$\DeclareMathOperator\SL{SL}$Not always. For simplicity, take $N$ a large prime $p$. Following user44191's suggestion in the comments, take $r$ to be $T(N)-1$.

Since any congruence subgroup contains a principal congruence subgroup, we may as well assume $L$ contains a principle congruence subgroup. We can write this as $L(p^e M)$ where $M$ is relatively prime to $M$. Then

$$L(p)/ L(p^eM) = \SL_2( \mathbb Z/M) \times \ker ( \SL_2(\mathbb Z/p^e) \to \SL_2( \mathbb Z/p))$$

Now $\SL_2(\mathbb Z/M)$ is a quotient of $\SL_2(\mathbb Z)$ and so can be generated by $2$ elements. The kernel $\ker ( \SL_2(\mathbb Z/p^e) \to \SL_2( \mathbb Z/p))$ is a $p$-group, so it can be generated by a number of elements equal to its $p$-rank, which is $3$. Thus the product can be generated by $5$ elements. (In fact, if we're careful, I'm pretty sure we can reduce this down to $3$, but this is unnecessary.)

Taking $p$ large enough that the principle congruence subgroup $L(p)$ has more than $5$ generators (i.e. taking $p>3$), we get a counterexample.


In the comments you ask whether, for each congruence subgroup $L$, there are congruence subgroups $L'' \subseteq L' \subseteq L$ such that the number of generators of $L'/ L''$ is at least half the number of generators of $L$.

The answer is no.

To see this, choose our congruence subgroup $L$ to be $L(p)$ for $p$ a large prime as above, and $T(p)$ the number of generators of $L$.

Let $[L:L']$ be the index of $L'$ in $L$. Because $L(p)$ is a free group on $T(p)$ generators, its subgroup $L'$ is a free group on $(T(p)-1) [L: L']+ 1$ generators and thus has generator rank at least $(T(p)-1) [L: L']+ 1$.

On the other hand, because $L/ L''$ can be generated by $5$ elements, it admits a surjection from a free group on $5$ generators, so its subgroup $L'/L''$, of index $[L:L']$, admits a surjection from a free group on $(5-1)[L:L']+1$ generators and thus has generator rank at most $4 [L:L']+1$.

To produce a counterexample, it suffices to have

$$ 4 [L:L']+1 < \frac{ (T(p)-1) [L: L']+ 1}{2} $$

for which, given $[L:L']\geq 1$, it suffices to have $T(p)>10$, something that is easy to achieve for $p$ large enough.

$\endgroup$
12
  • 1
    $\begingroup$ I added a link to @user44191's comment, and did some mild proofreading. In particular, "the number of generators of $T(N) - 1$" seemed to be a typo, since $T(N)$ is already a number of generators, so I removed "the number of generators of". $\endgroup$
    – LSpice
    Commented Aug 31, 2021 at 13:35
  • $\begingroup$ I am not sure I follow the arguments here. Are you saying that for some values of p, e and M, the congruence subgroup L = L(p^e M) ? $\endgroup$ Commented Sep 1, 2021 at 18:10
  • $\begingroup$ @sugatamondal No, I am saying $L$ contains $L(p^e M)$, and then to bound the number of generators mod $L$ it suffices to bound the number of generators mod $L(p^e M)$. $\endgroup$
    – Will Sawin
    Commented Sep 1, 2021 at 19:22
  • $\begingroup$ @WillSawin Right, thanks! $\endgroup$ Commented Sep 2, 2021 at 5:28
  • $\begingroup$ @WillSawin Given a congruence subgroup L does there exist a congruence subgroup L' of L such that L' has a congruence subgroup L" such that the minimum number of generators of L'/L" is at least half the minimum number of generators of L'? $\endgroup$ Commented Sep 2, 2021 at 6:07

You must log in to answer this question.

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