4
$\begingroup$

Suppose that $\pi:X\to S$ is a smooth projective morphism of relative dimension 1. If $S$ is the spectrum of an algebraically closed field, then it is known that $X$ embeds into $\mathbb{P}^3_S$.

Question: Can a similar statement be made for more general $S$?

I am primarily interested in Dedekind schemes $S$, but would also like to understand the general situation.

Let us assume that $X$ is geometrically integral, to try to account for the algebraically-closedness hypothesis. ([Edit]: In light of Jef's comment, I will strengthen this to "There is a section $S\to X$".) Let us also assume any standard finiteness condition that could be helpful: Noetherian, quasi-compact, etc.

If the statement is not true in this generality, what if we add some cohomological niceness properties: $X/S$ is cohomologically flat, or maybe even $S$ affine.

Question: If we assume the fibers of $\pi$ are curves of genus $0$, can we embed $X$ into $\mathbb{P}^2_S$?

I think this is true over Dedekind schemes, using the ample anti-canonical sheaf $\omega_{X/S}^\vee$, since $\pi_*\omega_{X/S}^\vee$ is locally free of rank 3, but I am not certain.

I would be completely content with an answer for Dedekind schemes $S$, but I would appreciate any help.

$\endgroup$
8
  • 3
    $\begingroup$ This is not even true for $S$ the spectrum of a non-algebraically closed field. If every curve $X/\mathbb{Q}$ would embed in $\mathbb{P}^3$, then every such curve would have a $\Q$-rational effective divisor of degree $3$. But not every curve has this property, e.g. take a genus $1$ curve that corresponds to an element of order $n$ in the Weil--Chatelet group of its Jacobian for $n>>0$ $\endgroup$
    – Jef
    Commented Sep 5, 2022 at 0:59
  • 1
    $\begingroup$ Your second question is true Zariski locally on $S$ (Dedekind or not), and in general $X$ can be embedded in a rank $2$ projective bundle over $S$. (Exactly by the argument you give) $\endgroup$
    – Jef
    Commented Sep 5, 2022 at 1:00
  • 1
    $\begingroup$ Sorry, i was talking nonsense! The degree of the embedding in $\mathbb{P}^3$ could be large. Please ignore my first comment. $\endgroup$
    – Jef
    Commented Sep 5, 2022 at 2:41
  • 8
    $\begingroup$ Consider the case that S is (the spectrum of) a finite field k. If a curve C/k embeds into P^n(k), then C has no more k-rational points than P^n(k). But there are smooth projective curves over k with arbitrarily many k-rational points. Thus for every n there are curves C/k that do not embed into P^n over k. $\endgroup$ Commented Sep 5, 2022 at 2:48
  • 4
    $\begingroup$ For every infinite field $k$ (not necessarily algebraically closed), every smooth projective $k$-variety $X$ of dimension $d$ embeds in $\mathbb{P}^n_k$ for every integer $n\geq 2d+1$. This follows from the usual proof (Whitney embedding theorem, roughly) together with Bertini's theorem for infinite fields. $\endgroup$ Commented Sep 5, 2022 at 11:09

1 Answer 1

4
$\begingroup$

This answer addresses the second question: "If we assume the fibers of $\pi$ are curves of genus $0$, can we embed $X$ into $\mathbb{P}^2_S$?"

The answer to this is also no (providing there is a singular fibre).

Let $\pi: X \to \mathbb{P}^1$ be a conic bundle surface over an algebraically closed field $k$, i.e. $X$ is regular and every fibre is isomorphic to a plane conic (possibly singular).

Consider the relative anticanonical bundle $\omega_{X/\mathbb{P}^1}^{-1}$. This is very ample when restricted to each fibre, and the pushforward to $\mathbb{P}^1$ is a vector bundle $V$ of rank $3$. We obtain an embedding

$$X \to \mathbb{P}(V)$$ which respects $\pi$, where $\mathbb{P}(V)$ denotes the corresponding $\mathbb{P}^2$-projective bundle over $\mathbb{P}^1$.

Now every vector bundle on $\mathbb{P}^1$ splits as a direct sum of line bundles, so we can write

$$V = \mathcal{O}(a_1) \oplus \mathcal{O}(a_2) \oplus \mathcal{O}(a_3).$$

We obtain the trivial projective bundle if and only if $a_1 = a_2 = a_3$. So we just need to give an example where this doesn't hold.

Firstly, if every fibre is smooth, then $X$ is a ruled surface. By the classification of ruled surfaces $X$ is a Hirzebruch surface $\mathbb{F}_n$ for some $n$. But $\mathbb{F}_n$ embeds into $\mathbb{P}^2 \times \mathbb{P}^1$ as $$x_0^n y_0 = x_1^n y_1.$$ So to get a counter-example we need to consider a conic bundle with a singular fibre.

I take $X$ to be a smooth cubic surface in $\mathbb{P}^3$. For any line $L \subset X$, there is an associated conic bundle given by taking the residual intersection of the pencil of planes through $L$. This has exactly 5 singular fibres as $L$ meets exactly 10 other lines of $X$.

Now take $Y$ a conic bundle surface in $\mathbb{P}^2 \times \mathbb{P}^1$. This has bidegree $(2,d)$ for some $d$. Considering the discriminant of the associated quadratic form one sees that this has $3d \neq 5$ singular fibres, as required.

Incidently, one can show that any smooth cubic surface embeds into $\mathbb{P}(V)$ where $V = \mathcal{O}(0) \oplus \mathcal{O}(0) \oplus \mathcal{O}(1).$

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .