Fix an algebraic integer $\alpha$ of degree $n$ such that the extension $K=\mathbf{Q}(\alpha)/\mathbf{Q}$ has intermediate fields. (We can assume $K$ is Galois with non-simple Galois group.) This $\alpha$, by its powers, gives rise to a co-ordinatisation for the affine space $K=\mathbf{Q}^n$. Now what is known about subset of algebraic numbers inside $K$ of a fixed degree $m$ less than $n$? Is this a $\mathbf{Q}$-Zariski closed subset of $K$? Or should we consider subsets of elements of degree less than or equal to $m$?

If we fix the Galois group $G$ and the divisor $m$ of $[K:\mathbf Q]$ is the ideal for degree m elements independent of $K$?

  • $\begingroup$ Please, try to use existing tags if suitable ones already exist, and in general please connect individual words that only together convey a specific meaning by a hyphen. Thanks in advance. $\endgroup$ – user9072 Jun 19 '13 at 19:47

Write an arbitrary element of $K$ as $x = x_0 + x_1 \alpha + \dots + x_{n-1} \alpha^{n-1}$ with $x_0, \dots, x_{n-1} \in \mathbb{Q}$. Then you can write $1, x, \dots, x^m$ as $$ x^d = y_{d,0} + y_{d,1} \alpha + \dots + y_{d,n-1} \alpha^{n-1}, $$ where the $y_{d,k}$ are polynomial expressions in $x_0, \dots, x_{n-1}$ (and also in the coefficients of the minimal polynomial of $\alpha$). Then the condition that $x$ have degree $\leq m$ is exactly that the $(m+1) \times n$ matrix with coefficients $y_{d,k}$ has rank strictly less than $m+1$ (then you can also find an element of the nullspace in $\mathbb{Q}$). This is an algebraic condition which can be cut out by the $(m+1) \times (m+1)$ minors. So indeed you get a Zariski closed subset for the elements of degree $\leq m$. I think the condition for degree equal to $m$ will be only locally closed in general. Also, at least from the above it appears that the ideal should depend on $K$. (I guess the case when $m$ is $1$ is an exception.)

  • 4
    $\begingroup$ +1. Just a remark: for any intermediate field $\mathbf{Q} \subset K' \subset K$ of degree $m$ we have that $K'$ is an $m$-dimensional sub-$\mathbf{Q}$-vector space of $K \cong \mathbf{Q}^n$. Hence, under the correspondence from the OP, the elements of $K'$ correspond to an $m$-dimensional linear affine subvariety $V_{K'}$ of $V = \mathbf{A}^n_{\mathbf{Q}}$ (affine $n$-space over $\mathbf{Q}$). We can conclude that each set $V_m$ of elements of degree precisely $m$ is a finite boolean combination of linear affine subspaces of $V$. In particular, $V_m$ will not be closed unless $m=1$. $\endgroup$ – RP_ Jun 19 '13 at 19:14
  • $\begingroup$ good point! that's a much cleaner description of the situation than in my answer :-) $\endgroup$ – Abhinav Kumar Jun 19 '13 at 20:30
  • $\begingroup$ Thanks Abhinav Kumar for the answer and to Ren\'e Pannekoek for throwing more light on it. Originally I had a different, but related question (see recent questions) and I thought I need to get this one out first. $\endgroup$ – P Vanchinathan Jun 20 '13 at 0:28

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.