2
$\begingroup$

Consider the matrix

$$ M=\begin{bmatrix} \alpha_1 & \alpha_2 & \dots & \alpha_m \\ \alpha_1^q & \alpha_2^q & \dots & \alpha_m^q \\ \alpha_1^{q^2} & \alpha_2^{q^2} & \dots & \alpha_m^{q^2} \\ \vdots & \vdots & \ddots & \vdots \\ \alpha_1^{q^{m-1}} & \alpha_2^{q^{m-1}} & \dots & \alpha_m^{q^{m-1}} \\ \end{bmatrix} $$ over $\mathbb F_{q^n}$ for some prime power $q$. Such a matrix is sometimes called a Moore Matrix. It is a classically know result that $M$ is nonsingular if and only if $a_1, \ldots, a_m$ is linearly independent over $\mathbb F_q$ (see, for instance, [Lidl and Niederreiter][1] lemma 3.51).

My question is a strengthening of this result: do the linear dependencies of $\{a_1, \ldots, a_m\}$ over $\mathbb F_q$ completely determine the kernel of $M$ (i.e. the linear dependencies of the columns of $M$ over $\mathbb F_{q^n}$)? More precicely, do we have the equality $$ \mathrm{ker}M = \mathrm{span}_{\mathbb F_{q^n}} \{ v \in {\mathbb (F_q)}^n | Mv = 0 \} $$ Note that this requires identifying $\mathbb F_q$ as a subset of $\mathbb F_{q^n}$ in some way, which may introduce ambiguities/issues, but I am primarily interested in a case where $q$ is prime where this is not a problem. Of course, one inclusion is trivial, but I have been unable to arrive at the other.

For my purposes I am specifically interested in the case of $q=2$, for which I have been unable to find a counterexample with a brute force search in sage. In this case the hypothesis is equivalent to asking whether there exists some basis for $\mathrm{ker}M$ whose entries are all either $0$ or $1$.

Any leads are appreciated.

[1]: Rudolf Lidl and Harald Niederreiter. Introduction to finite fields and their applications. Cambridge university press. 1986.

$\endgroup$
3
  • $\begingroup$ Some remarks: 1) $\mathbf{F}_q$ is canonically identified with a subset of $\mathbf{F}_{q^n}$ - it is the set of elements satisfying $x^q=x$. 2) Your displayed equation makes no sense to me. 3) Note that the rank of $M$ equals the dimension of the space spanned by the $\alpha_i$'s, by considering a sub-matrix corresponding to a maximal independent subset of those. $\endgroup$
    – Uri Bader
    Jun 27, 2017 at 8:56
  • $\begingroup$ Looking again at your displayed equation, probably you just had a typo there. If you meant to write $v\in (\mathbb{F}_q)^m$, then you got it right. In that case, it seems that what you have missed is the observation about the rank of $M$ (see above) and the resulting comparison of dimensions. $\endgroup$
    – Uri Bader
    Jun 27, 2017 at 9:10
  • $\begingroup$ Thanks for pointing out that finite fields embed canonically into each other! And sorry I didn't write $(\mathbb F_q)^n$ very clearly, edited! $\endgroup$ Jun 27, 2017 at 16:24

1 Answer 1

1
$\begingroup$

Consider the vector $(\alpha_1,\ldots,\alpha_m)$ as the $\mathbf{F}_q$-linear transformation $T:(\mathbf{F}_q)^m\to \mathbf{F}_{q^n}$ given by $$ (\beta_1,\ldots,\beta_m) \mapsto \sum \beta_i\alpha_i. $$ Then $\ker(M)$ is the $\mathbf{F}_{q^n}$-span of $\ker(T)$ in $(\mathbf{F}_{q^n})^m$. To see this, note that $\ker(T)<\ker(M)$ and compare dimensions.

$\endgroup$
1
  • $\begingroup$ Oh yeah that works! Perfect, thanks $\endgroup$ Jun 27, 2017 at 16:46

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.