Let $X$ be as smooth variety over a field $k$ of characteristic $0$.

Consider the following statements:

  • The variety $X$ has no $k((t))$-rational points.
  • No smooth compactification of $X$ has a $k$-rational point.

Are these equivalent? If not, what additional assumptions on $X$ would make them equivalent? I'm particularly interested in the case where $X$ is a homogenous space of a "nice" algebraic group over $k$.

  • $\begingroup$ Assume that $X$ is itself smooth and proper. Then the first condition just says that $X$ doesn't contain any rational curves defined over $k$. That seems much stronger than not having $k$-rational points. $\endgroup$ – Keerthi Madapusi Pera May 3 '12 at 2:09
  • 2
    $\begingroup$ @Keerthi Madapusi Pera: I'm guessing you read $k(t)$ for the proposer's $k((t))$. Even for $k(t)$ it's not quite true because any $k$-rational point is automatically $k(t)$-rational (the map from the $t$-line to $X$ can be constant). $\endgroup$ – Noam D. Elkies May 3 '12 at 4:56
  • $\begingroup$ @Noam--Of course. That was very silly of me. $\endgroup$ – Keerthi Madapusi Pera May 3 '12 at 5:23

Yes, this is true.

One implication is immediate: if $X$ has a $k((t))$ point then by the valuative criterion of properness there is a map $Spec(k[[t]])$ to any compactification of $X$, so the image of the closed point gives a $k$-point of the compactification.

For the converse, if a smooth compactification has a $k$-point then choose a general curve $C$ through that point (which is smooth at that point). Since $C$ is general, it is not contained in the boundary. By completing the local ring of the curve at the smooth point you get a map $Spec(k[[t]])$ to the compactification. Since $C$ is not contained in the boundary, the map retricted to the generic point (which is sent to the generic point of the curve) gives a $k((t))$-point of $X$.

| cite | improve this answer | |
  • $\begingroup$ Nice answer. Where do we use the smoothness of $X$ and the fact that $k$ is of characteristic zero? I think the "obvious" implication you mention using the val. crit. holds without these hypotheses. Moreover, I don't see how you use these hypotheses in the second implication either. Also, why does there always exist a "general" curve on $\overline X$? Looking at the dimension 1 case this doesn't seem to be the case (but of course the dimension 1 case could be handled separately). $\endgroup$ – Harry May 3 '12 at 7:04
  • 2
    $\begingroup$ Charateristic zero is used implicitly in order that there exists a smooth compactification of $X$. Smoothness is used to ensure that there exists a curve passing through any $k$-rational point that is smooth at that point. $\endgroup$ – ulrich May 3 '12 at 7:59
  • $\begingroup$ Great answer, thanks a lot. Could you just explain a bit more why the existence of such a curve follows from smoothness? $\endgroup$ – Wanderer May 3 '12 at 16:28
  • $\begingroup$ Let $\dim(X) = n$. For $x$ a $k$-rational point let $f_1,f_2,\dots,f_n$ be generators of the maximal ideal at $x$. Then we get an etale morphism $F$ from an open neighbourhood $U$ of $X$ to $\mathbb{A}^n$ given by $(f_1,f_2,\dots,f_n)$ such that $F(x) = 0$. Let $L$ be any line in $\mathbb{A}^n$ passing through $0$ and not contained in the image of the boundary. The irreducible component of $F^{-1}(L)$ containing $x$ can be taken to be the curve $C$. $\endgroup$ – ulrich May 4 '12 at 3:13

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.