2
$\begingroup$

Let $R = \mathrm{GF}(q)$, $S = \mathrm{GF}(q^n), \ n\geq 2$ be extension of $R$, $h$ be a primitive element of $S$. I want to count or estimate the number $N$ of bases of the following form. Let $$\vec{\beta} = (\beta_0,\beta_1,\ldots,\beta_{n-1})$$ be a basis of the space $_RS$ with the property that there exists $k\in\{1,2,\ldots,n-1\}$ such that $$ \vec{\beta'} = (\beta_0,\ldots,\beta_{k-1},h\beta_k,\ldots,h\beta_{n-1}) $$ is a basis of the space $_RS$.

It is easily to see that $N\geq (q^n-q)\prod_{j=1}^{n-1}(q^n-2q^{j}+q^{j-1})$. This estimation obtained from the case $k = n-1$

It is also easily to see, that the basis of the form $(e,h,\ldots, h^{n-1})$ has the desirable property. My another question is: is there another approches for construction such bases?

$\endgroup$
1
  • $\begingroup$ I should note that bases of this form play important role in construction recursion laws of maximal period sequences with high linear complexity, oriented toward fast software implementation. Such sequences are widely used in stream ciphers. See [article] (mathnet.ru/links/cf8202ececdb879c480aa24b5c9d80e4/mvk115.pdf) for details. $\endgroup$ Commented Apr 18, 2016 at 6:51

1 Answer 1

2
$\begingroup$

Let me estimate the number $N_k$ of such bases for a fixed $k$. Aside remark. The number of bases for $k=k_0$ equals the number of those for $k=n-k_0$, as the bijection $(\beta_0,\dots,\beta_{n-1})\to (\beta_{n-1},\dots,\beta_{k_0},h^{-1}\beta_{k_0-1},\dots,h^{-1}\beta_0)$$ shows.

So let us fix any $k$. The first $k$ elements can be chosen in $\prod_{i=0}^{k-1}(q^n-q^i)$ many ways. Each next element $\beta_i$ should lie outside $V_i=\langle \beta_0,\dots,\beta_{i-1}\rangle$, as well as outside $V_i'=\langle h^{-1}\beta_0,\dots,h^{-1}\beta_{k-1},\beta_k,\dots,\beta_{i-1}\rangle$. If $d_i=\dim(V_i\cap V_i')$, then $\beta_i$ can be chosen in $q^n-2q^i+q^{d_i}$ many ways; notice that $i\geq d_i\geq \max(i-k,2i-n)$. Thus $$ N_k\geq \prod_{i=0}^{k-1}(q^n-q^i)\prod_{i=k}^{n-1}(q^n-2q^i+q^{\max(i-k,2i-n)}). $$

This estimate can be improved by noticing that $d_i\geq d_{i-1}+1$. So, e.g., for $k=1$, if we choose $\beta_1=(h^{-1}+1)\beta_0$, then $V_2=V_2'$, so $d_i=i$ for $i\geq 2$. But it seems appropriate to choose large $k$...

Notice that for $k=n-1$, the estimate gives $$ N_{n-1}\geq (q^n-2q^{n-1}+q^{n-2}).\prod_{i=0}^{n-2}(q^n-q^i), $$ which is $1-\frac1q$ of all bases. Do you need much sharper bounds?

$\endgroup$
2
  • $\begingroup$ Thank you very much! I agree with everything. But I there are some questions. I think that in the last formula for $N_{n-1}$ there should be $\prod\limits_{i=0}^{n-2}(q^n-q^i)$. Also I don't understand the part of yor answer, where $\beta_1 = (h^{-1}+1)\beta_0$. Why in this case $V_1 = V_1'$? $\endgroup$ Commented Apr 18, 2016 at 13:01
  • $\begingroup$ Surely - I've edited the indices. As for he second question: that was another typo, I meant $V_2'=V_2$, since both equal $\langle \beta_0,h^{-1}\beta_0\rangle$... $\endgroup$ Commented Apr 18, 2016 at 15:18

You must log in to answer this question.

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