4
$\begingroup$

Let $K$ be a finitely generated extension of $\mathbb{Q}$ of transcendence degree at least 1. 

Recall that a valuation ring of $K/\mathbb{Q}$ is a sub-$\mathbb{Q}$-algebra $V\subset K$ such that for every $x\in K$ we have $x\in V$ or $x^{-1}\in V$. A place of $K/\mathbb{Q}$ is a $\mathbb{Q}$-algebra homomorphism $p\colon V\to L$, where $V$ is a valuation ring of $K/\mathbb{Q}$, $L$ is a field and $p(x^{-1})=0$ for any $x\in K, x\not\in V$. The image of $p$ is an extension of $\mathbb{Q}$, which is called the residue field of $p$.

If $X/\mathbb{Q}$ is a projective variety with a $K$-rational point and $L$ is the residue field of a place of $K/\mathbb{Q}$, then it is easy to see that $X$ admits an $L$-rational point.

My question: Assume now that we are given a projective variety $Y/\mathbb{Q}$ that does not admit a $K$-rational point. Is there always a place $p$ of $K/\mathbb{Q}$ with residue field $L$, where $\operatorname{trdeg}_{\mathbb{Q}} L < \operatorname{trdeg}_{\mathbb{Q}} K$, such that $Y$ also does not admit an $L$-rational point?

$\endgroup$
4
  • 1
    $\begingroup$ If $K=L(t)$ for $t$ trascendental over $L$, $L$ is a residue field but also $L$ embeds into $K$, so any $L$-rational point can be made a $K$-rational point? $\endgroup$
    – GTA
    Commented Mar 20, 2020 at 21:00
  • $\begingroup$ @GTA Thanks for your comment! If I understood you correctly, this does seem to resolve my question in the special case where K can be written as a purely transcendental extension of an extension of $\mathbb{Q}$. But if we choose $K$ to be e.g. $\mathbb{Q}(t, \sqrt{t^3-t})$, this trick does not seem to apply anymore. Do you maybe have an idea how to handle the general case? $\endgroup$ Commented Mar 21, 2020 at 17:09
  • $\begingroup$ The trick will work as long as there exists a residue field that embeds into $K$. Say that $K$ is the function field of a curve $C$ over a number field $k$. Usually $C$ is geometrically connected, which means that $k$ is algebraically closed in $K$. This implies that any residue field embedding into $K$ must be equal to $k$. If $C$ has a $k$-rational place, then the answer to your question is yes. But in general the $k$-rational places correspond to the $k$-rational points of the normalisation of $C$, and this set can be empty. $\endgroup$ Commented Mar 22, 2020 at 20:02
  • $\begingroup$ I see now that this trick will only work in rather special situations. Do you know whether there is anything else one can do? $\endgroup$ Commented Mar 23, 2020 at 17:55

0

You must log in to answer this question.

Browse other questions tagged .