MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am looking for a reference for the following well-known fact:

Let $k$ be an uncountable field, and let $X$ be a $k$-variety. Let $Z_1, Z_2, \dots \subseteq X$ be proper closed subschemes. Then $\bigcup Z_i(k) \neq X(k)$.


share|cite|improve this question
You also need $k$ to be algebraically closed (otherwise one could well have $X(k) = \emptyset$). – ulrich Aug 26 '11 at 9:12
I'm glad you asked the question. You'd think this would discussed in some standard textbook, but I've never seen it. Most complex algebraic geometers use a sledge hammer (Baire category theorem) but it's certainly more elementary than that. – Donu Arapura Aug 26 '11 at 12:19
It should be possible to give a bare hands proof that given a countable collection of nonzero polynomials $f_i$ in $n$-variables, there exist a point such that $f_i(p)\not=0 $ simultaneously. – Donu Arapura Aug 26 '11 at 12:26
Yes, I left out algebraically closed as a hypothesis. It's frustrating that there (doesn't seem to be) a standard place to quote such a well-known fact. – Sue Sierra Aug 26 '11 at 18:06
As luck would have it, a related question just came up in my research today: does an algebraic torus over $k$ have a dense cyclic subgroup? If $k$ is uncountable and algebraically closed, the above-mentioned fact shows the answer is yes: just choose a generator outside the (countable) union of all proper Zariski closed subgroups. But if $k$ is the (countable) algebraic closure of a finite field, the answer is no, since every element of $k^\times$ has finite order. – Michael Thaddeus Aug 28 '11 at 3:59

Suppose $\dim X>0$ and $k$ is algebraically closed and uncountable. Moreover, if a "variety" is not necessarily irreducible, the $Z_i$ are supposed to have positive codimension in $X$ (otherwise one could take the irreducible components of $X$).

As in MP's answer, one can suppose $X$ is affine. By Noether's Normalization Lemma, there exists a finite surjective morphism $p: X\to \mathbb A^m_k$ with $m=\dim X$. Let $Y_i=p(Z_i)$. This is a closed subset of $\mathbb A^m_k$ of positive codimension. Moreover $\mathbb A^m_k(k)=\cup Y_i(k)$ because $k$ is algebraically closed (which implies that $Y_i(k)=p(Z_i(k))$). As $k$ is uncountable, there exists a hyperplane $H$ in $\mathbb A^m$ not contained in any $Y_i$ (note that $H\subseteq Y_i$ is equivalent to $H=Y_i$). So by induction on $m$ we are reduced to the case $m=1$, and the assertion is obvious.

Without the hypothesis $k$ algebraically closed, one can show similarly that $X\ne \cup_i Z_i$. This is Exercise 2.5.10 in my book. EDIT In fact this statement is trivial because the generic points of $X$ don't belong to any of the $Z_i$'s. But the proof shows that the set of closed points of $X$ is not contained in $\cup_i Z_i$.

share|cite|improve this answer

I do not know a reference, but the following short argument seems to work.

Assume that the dimension of $X$ is at least 1! Argue by induction on the dimension of $X$. Reduce to the case in which the subschemes are irreducible of codimension one. Shrinking $X$ if necessary, reduce also to the case in which $X$ is quasiprojective. Let $L$ be a pencil of integral divisors on $X$. Since the ground-field is uncountable, the pencil $L$ contains uncountably many elements that are integral; let $D \in L$ be an integral element of $L$ that is different from each of the subschemes you want to avoid. By the inductive hypothesis, $D$ contains a point that is not contained in any of the subschemes and you are done.

You can find this stated as a hint in an Exercise V.4.15 (c) in Hartshorne.

share|cite|improve this answer
As ulrich's correctly comments above, the only place that I did not argue (namely, the base case of the induction) is also the place where there is a missing assumption: the field should be algebraically closed! The argument sketched here requires curves over $k$ to have uncountably many points. – M P Aug 26 '11 at 9:23

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.