I am still on my quest to design a simple example of an ample, globally generated line bundle which is not very ample and also understand the global sections of its smallest, very ample tensor power. Let $\Bbbk$ be some algebraically closed field, Consider $$R = \Bbbk[x,y,z]/(yz-x^2)$$ and the projective parabola $X=\operatorname{Proj}(R)$. Let $I=\langle y,z\rangle\subseteq R$ be the homogeneous ideal generated by $y$ and $z$, and consider the ideal sheaf $\newcommand{\cL}{\mathscr{L}}\cL:=I(1)^\sim$, as in, $\cL(X)=I_1=\Bbbk y \oplus \Bbbk z$. Observe that it is globally generated because $y$ and $z$ do not vanish simultaneously on $X$, otherwise $x$ would have to vanish as well. However, of course, the morphism $X\to\newcommand{\P}{\mathbb P}\P^2$ defined by $[x:y:z]\mapsto [y:z]$ is a finite surjection of degree $2$.

Now, there must be some $k\in\mathbb N$ (probably $k=2$) such that $\cL^{\otimes k}$ is very ample. However, I don't understand how it works, in this very concrete example.

My question is basically: what are the global sections of $\cL^{\otimes k}$ and does $\cL^{\otimes 2}$ contain functions $xz$ and $xy$?

That would be wonderful, but I can't see how to define either of them on both $X_y$ and $X_z$. More precisely, I would think that $$\cL^{\otimes 2}(X_y)=\left\{ \frac{fz^2}{y^n} \:\middle\vert\: f\in R_n \right\}$$ but if $xz=fz^2y^{-n}$ then $x=fzy^{-n}$ and this is hard to do when we only have a relation for $x^2$. I must be mistaken somewhere, and this is driving me crazy.

  • 2
    $\begingroup$ do you have a precise definition of $\tilde{I(1)}$ somewhere ? shouldn't $x = xy/y = xz/z$ belong to $\cL (X)$ ? Your $X$ is isomorphic to $\Bbb P^1( \Bbb k)$, where the ample line bundles are very ample. $\endgroup$ – mercio Nov 9 '13 at 17:27
  • $\begingroup$ I am using Hartshorne's definition, but you are right, the global sections of $I(1)^\sim$ already contain $x$. It was a silly question after all. Still, I am yet to find the example I am looking for. It's fine to know that there are abstract examples, but I would like to see this in coordinates at least once. Sigh. $\endgroup$ – Jesko Hüttenhain Nov 9 '13 at 17:35
  • $\begingroup$ @mercio: Post that as an answer. Would be a shame if the 400 rep would just go to waste. $\endgroup$ – Jesko Hüttenhain Nov 10 '13 at 10:57
  • 1
    $\begingroup$ @JeskoHüttenhain: Hi Jesko, I think the two easiest examples are to take $C$ either an elliptic curve, with $L$ the line bundle of any degree-2 divisor, or to take $C$ a genus-2 (hyperelliptic) curve, with $L$ its canonical bundle. In both cases, $L$ is ample and globally generated, and gives a map to $P^1$ (so it is clearly not very ample). For the elliptic case, $L^{\otimes 2}$ is very ample; for the hyperelliptic case, $L^{\otimes 3}$ is very ample (and $L^{\otimes 2}$ is not). Edit: In coordinates, you can find the genus-2 curve as a smooth curve of type (2,3) on $P^1 \times P^1$. $\endgroup$ – Jake Levinson Jan 30 '14 at 4:18
  • $\begingroup$ @JakeLevinson: Thanks a bunch! $\endgroup$ – Jesko Hüttenhain Jan 30 '14 at 8:44

$X$ is isomorphic to $\Bbb P^1(\Bbb k)$,by the maps induced by $(u:v) \mapsto (uv:u^2:v^2)$ and $(x:y:z) \mapsto (y:x) = (x:z)$.

$x$ is a global section of $\cL$, because $x = xy/y \in \cL(X_y)$ and $x=xz/z \in \cL(X_z)$, and those two are equal on $X_y \cap X_z = X_x$ : $xz/z = xyz/yz = xy/y$, so $x \in \cL(X)$. Actually, the pulled back sheaf of $\cL$ to $\Bbb P^1(\Bbb k)$ is $\mathcal O_{\Bbb P^1(\Bbb k)}(2)$, which is ample and also very ample.


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.