2

Denote $\mathbb{Z}_q$ be the residue ring of module $q$ with $q=p^r$ be a prime power. Similar to the question for $k$-dimensional vector spaces over finite fields, we can also define a $k$-dimensional $\mathbb{Z}_q$-submodule in $\mathbb{Z}_q^d$ as the following:
$$ V = \mathrm{span}\{v_1,\dots,v_k\} $$ where $v_1,\dots,v_k \in \mathbb{Z}_q^d$ are linearly independent, namely, if there exist $a_1,a_2,\dots,a_k \in \mathbb{Z}_q$ such that $a_1v_1+a_2v_2+\dots+a_kv_k = 0$ then $a_1=a_2=\dots=a_k = 0.$

Question: How many $k-$ dimensional $\mathbb{Z}_q$-submodule in $\mathbb{Z}_q^d$? And, how many bases for a $k$-dimensional submodule?

The related question for the case of finite fields was done, you can follow via the link: How to count number of bases and subspaces of a given dimension in a vector space over a finite field?

The Nguyen
  • 194
  • 7
  • 1
    Since $\mathbb{Z}/q\mathbb{Z}$ is not a field, these are not vector spaces. One can ask this question (and it makes sense) replacing "space" by "$\mathbb{Z}/q\mathbb{Z}$-module" in your question. – Mummy the turkey Jan 26 '21 at 09:33
  • Yes, I know the fact that these are not vector spaces. However, I still use the same word "space" for easily understanding the question. – The Nguyen Jan 26 '21 at 09:50
  • 1
    I am not so sure that using terminology which makes the question make no sense is helpful for understanding the question... – Mummy the turkey Jan 26 '21 at 10:01
  • I think that you should be looking at Hall's Polynomials, which count the subgroups of specified type (and whose quotient is of specified type) in an abelian $p$-groups of specified type. I also think using vector space language is unhelpful. – ancient mathematician Jan 26 '21 at 10:29
  • Anything unclear? – reuns Jan 27 '21 at 17:15

2 Answers2

1

Your independence condition means a rank $k$ free submodule, there are also many non-free submodules.

Let $M(d,k,p^r)$ be the set of matrices $A\in \Bbb{Z}/(p^r)^{d\times k}$ such that its reduction $A\bmod p$ (a matrix $ \in \Bbb{Z}/(p)^{d\times k}$) has its $k$ columns linearly independent in the vector space $\Bbb{Z}/(p)^d$.

$$W=\{ span(v_1,\ldots,v_k), \sum_{j=1}^k c_j v_j=0\implies c_1=\ldots =c_k=0\}= M(d,k,p^r)/M(k,k,p^r) $$ The group $M(k,k,p^r)$ acts on $M(d,k,p^r)$ on the right and $M(d,k,p^r)/M(k,k,p^r)$ is the set of orbits, the columns of $A\in M(d,k,p^r)$ represent the $v_j$ and the action of $M(k,k,p^r)$ changes the $v_j$ without changing the submodule it generates.

$$\# W= \frac{\# M(d,k,p^r)}{\# M(k,k,p^r)}=\frac{p^{kd(r-1)}\# M(d,k,p)}{p^{kk(r-1)}\# M(k,k,p)}= \frac{p^{kd(r-1)}\prod_{l=0}^{k-1}(p^d-p^l)}{p^{kk(r-1)}\prod_{l=0}^{k-1}(p^k-p^l)}$$

  • where $p^{kd(r-1)}$ represents the number of matrices having the same reduction modulo $p$

  • $\prod_{l=0}^{k-1}(p^d-p^l)$ means choosing $v_1\bmod p$ in $\Bbb{Z}/(p)^d-\{0\}$, then choosing $v_2\bmod p$ in $\Bbb{Z}/(p)^d-span(v_1)$, choosing $v_3\bmod p$ in $\Bbb{Z}/(p)^d-span(v_1,v_2)$, and so on.

reuns
  • 77,999
  • I think there is something wrong with your answer. Take $k=2$ as an example, the number of two-dimensional submodules, as I counted, is $\frac{(p^{dr}-p^{d(r-1)})(p^{dr}-p^{d(r-1)} - p^r+p^{r-1})}{(p^{2r}-p^{2(r-1)})(p^{2r}-p^{2(r-1)}-p^r+p^{r-1})}$ which different from your answers. – The Nguyen Feb 07 '21 at 09:15
  • Also, my below answer ( I deleted it) is not true because the canonical projection doesn't preserve the linearly independent property. – The Nguyen Feb 07 '21 at 09:18
  • @NguyễnVănThế Why do you say the projection doesn't preserve linearly independent property ? I am counting the free submodules, the reductions $\bmod p$ are free too, and a submodule generated by $d$ elements is free iff its reduction is a $d$-dimensional vector space. – reuns Feb 07 '21 at 14:20
  • I really apologize for my wrong. I just checked it again. It is true that the projection does preserve linearly independent property. Thank you very much. – The Nguyen Feb 07 '21 at 18:41
  • 1
    @NguyễnVănThế The fun step is when showing that no $B\ne 1\in\mathrm{GL}_k(\mathbb{Z}_q)$ is such that $A=AB$ (so that $|\mathrm{GL}_k(\mathbb{Z}_q)|$ distinct matrices give the same free module), this is again because the module generated by the columns of $A$ is free, so that $A(I-B)=0\implies I-B=0$. – reuns Feb 07 '21 at 18:57
  • However, I still confuse why the case $k=2$ is not true when I count the number of $v_1,v_2$ such that they are linearly independent. Indeed, if I follow the projection map, we have the number of $v_1,v_2 \in \mathbb{Z}_q^d$ linearly independent equal to $p^{2d(r-1)}(p^d-1)(p^d-p).$ While if we follow the counting method, as usual, we will get the answer is $(p^{dr}-p^{d(r-1)})(p^{dr}-p^{d(r-1)}-p^r+p^{r-1})$. The latter result is obtained by the following: (see next comment) – The Nguyen Feb 07 '21 at 19:07
  • Choosing $v_1$ in $\mathbb{Z}_q^d \setminus (\mathbb{Z}_q^0)^d,$ therefore, there are $p^{dr}-p^{d(r-1)}$ possibilities.
  • For $v_2,$ choosing $v_2 \in \left(\mathbb{Z}_q^d \setminus (\mathbb{Z}_q^0)^d\right) \setminus \mathrm{span}(v_1),$ so there are $p^{dr}-p^{d(r-1)} - p^r + p^{r-1}$ possibilities for $v_2$. Could you please help me check it? What is wrong?

    – The Nguyen Feb 07 '21 at 19:07
  • Try with $d=2$, $v_1=(1,0)$ then for $v_2$ you need to avoid $(1,p)$ to get a free module. – reuns Feb 07 '21 at 19:38