MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

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

Let $\mathfrak{p}_1, \dotsc, \mathfrak{p}_k$ be relevant homogeneous primes ideals in the graded ring $R := \Bbbk[x_0, \dotsc, x_n]$, where $\Bbbk$ is a field. Prime avoidance (in Eisenbud's terminology) tells us that there exists a nonconstant homogenous polynomial $f \not\in \cup_i \mathfrak{p}_i$. Is there an easy way to see that for some $d \geq 1$, there exist homogeneous $f \in R_d$, $g \in R_{d+1}$, neither of which is contained in $\cup_i \mathfrak{p}_i$?

Note: If $\Bbbk$ is infinite, this is easy: we can use the fact that no vector space is a finite union of proper subspaces to find $\alpha \in R_1 \smallsetminus \cup_i \mathfrak{p}_i$, and then take $f=\alpha$, $g = \alpha^2$.

Motivation: If the $\mathfrak{p}_i$ are the associated primes of a closed subscheme $X$ of $\mathbb{P}^n_{\Bbbk}$ then we can pull back $g/f$ to a nice rational section of $\mathcal{O}_X(1)$. Consequently, we know that $\mathcal{O}_X(1)$ comes from a Cartier divisor. Using the fact that any line bundle can be twisted to a very ample line bundle, this shows that the Cartier divisor group surjects onto $Pic(X)$.

share|cite|improve this question
The problem can be reduced to finding such $f \in R_d$ and $g \in R_m$ with gcd $(m,d)=1$, which seems much easier. If this would not be the case, it would imply that $R \backslash \cup_i p_i \subset \sum_{n\geq 1} R_{dn}$ – Nick S Feb 10 '11 at 23:08
As for the surjection to $Pic(X)$, it is known that it holds for any locally noetherian scheme such that the associted points all live in a same affine open subset (EGA IV, 21.3.5). This contains the case of quasi-projective schemes over a noetherian ring. – Qing Liu Feb 11 '11 at 12:35
up vote 6 down vote accepted

The answer is yes. The following is extracted from a preprint of Gabber-Liu-Lorenzini.

Let $B=\oplus_{n\ge 0}B(n)$ be a graded ring. Let $I=\oplus_{n\ge 0}I(n)$ be a homogeneous ideal of $B$. Let $\mathfrak p_1,\dots,\mathfrak p_r$ be homogeneous prime ideals of $B$ not containing $B(1)$ and not containing $I$. Then there exists an integer $n_0 \geq 1$ such that for all $n\ge n_0$, $I(n)\not\subseteq \cup_{1\le i\le r} \mathfrak p_i$.

Proof. We proceed by induction on $r$. If $r=1$, choose $t\in B(1)\setminus \mathfrak p_1$ and a homogeneous element $\alpha\in I\setminus \mathfrak p_1$, say of degree $n_0$. Then $t^{n-n_0}\alpha\in I(n) \setminus \mathfrak p_1$ for all $n\ge n_0$, as desired.

Let $r\ge 2$ and suppose that the lemma is true for $r-1$. We can suppose that $\mathfrak p_i$ is not contained in $\mathfrak p_r$ for all $i\ne r$, so that $I\mathfrak p_1\cdots\mathfrak p_{r-1} \not\subseteq \mathfrak p_r$. Similarly, we can suppose that $\mathfrak p_r$ is not contained in $\mathfrak p_i$ for all $i\ne r$, so that $I\mathfrak p_r\not\subseteq (\mathfrak p_1\cup \ldots\cup\mathfrak p_{r-1})$. Hence, we can apply the case $r=1$ and the induction hypothesis to obtain that there exists $n_0$ such that for all $n\ge n_0$, there are homogeneous elements $f_n\in I\mathfrak p_1\cdots\mathfrak p_{r-1} \setminus \mathfrak p_r$ and $g_n\in I\mathfrak p_r\setminus (\mathfrak p_1\cup \ldots\cup\mathfrak p_{r-1})$ of degree $n$. It is easy to check that $f_n+g_n\in I(n)\setminus \cup_{1\le i\le r}\mathfrak p_i$, as desired.

share|cite|improve this answer
@Qing: Very nice! May I ask what preprint? – Hailong Dao Feb 11 '11 at 12:55
@Hailong: it is Lemma 5.11 in this very fine paper: – Pete L. Clark Feb 11 '11 at 13:28
Thanks, Pete! This paper looks very appealing to me. – Hailong Dao Feb 11 '11 at 13:51
Thanks Hailong and Pete. The notation here is different because I extracted the above statement from a second (not yet finished) preprint where we need a more technical form of prime avoidance. – Qing Liu Feb 11 '11 at 15:41
This is a really nice paper. – Angelo Feb 11 '11 at 16:46

EDIT: the argument below only works if each $R/\mathfrak{p}_i$ has dimension at least $2$, as pointed out by Angelo.

Suppose $\Bbbk$ is a finite field of size $q$. Let $V_{i,n}= \mathfrak{p}_i \cap R_n$, $d_n= \dim_{\Bbbk} R_n$ and $d_{i,n}=\dim_{\Bbbk}V_{i,n}$. It is enough to show that for $n$ big enough:

$$|V_n| > \sum_1^k |V_{i,n}| $$

Note that for each $i$, $d_n - d_{i,n}$ gives the Hilbert function of $R/\mathfrak{p}_i$, so it eventually becomes a polynomial in $n$ of dimension $={\dim R/\mathfrak{p}_i}-1 \geq 1$ (EDIT: I originally wrote $\dim R/\mathfrak{p}_i$). So for $n\gg 0$, $q^{d_n-d_{i,n}}>k$ for each $i$. Thus

$|V_n|/|V_{i,n}|= q^{d_n-d_{i,n}}> k$

and we are done.

share|cite|improve this answer
It seems to me that the degree of the Hilbert polynomial of $R/\mathfrak p_i$ is its dimension minus 1. – Angelo Feb 11 '11 at 8:31
Opps, you are right! Silly me. – Hailong Dao Feb 11 '11 at 8:56

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.