de Jong: Rationally Connected Varieties

Deformation theory

Let $ X$ be a nonsingular projective variety over $ \mathbb{C}$, and let $ C \hookrightarrow X$ be a one-dimensional closed subscheme. We have $ \mathscr{O}_X \supset \mathscr{I}_C$, the ideal sheaf of $ C$, and we assume that $ C$ is a local complete intersection, or what is equivalent, $ \mathscr{I}_C/\mathscr{I}_C^2$ is a locally free sheaf of $ \mathscr{O}_C$-modules of rank $ \dim X-1$. For example, this holds if $ X$ is a nodal curve.

Definition. The normal bundle of the curve $ C$ in $ X$ is

$\displaystyle \mathscr{N}_C X=\mathscr{H}om_{\mathscr{O}_C}(\mathscr{I}_C/\mathscr{I}_C^2,\mathscr{O}_C). $

We have that

$\displaystyle \widetilde{\mathscr{O}}_{{\mathrm{Hilb}}_X,[C]}=\mathbb{C}[[t_1,\dots,t_d]]/(f_1,\dots,f_r), $

where $ d=h^0(C,\mathscr{N}_C X)$, and $ r \leq h^1(C,\mathscr{N}_C(X))$. We say that $ H^0$ is the deformation space, and $ H^1$ are the obstructions to deformation. In particular, the dimension the Zariski tangent space of the Hilbert scheme at $ [C]$ has $ \mathscr{T}_{[C]} {\mathrm{Hilb}}_X=H^0(C,\mathscr{N}_C X)$, and it has dimension at least $ \chi(\mathscr{N}_C X)$, the Euler characteristic.

Example. In the case where $ [f:C \to \mathbb{P}^1] \in \overline{\mathcal{M}_g}(\mathbb{P}^1,d)$, with $ C$ is a smooth genus $ g$ curve, and $ f$ having at worst simple branchings, then there are no obstructions to deformation and

$\displaystyle \mathscr{T}_{[f]} \overline{\mathcal{M}_g}(\mathbb{P}^1,d)=\bigoplus_{P\in{\mathrm{Ram}}(f)} \mathscr{T}_{\mathbb{P}^1}\vert _{f(P)}. $

This is a canonical way of understanding how to move branch points on maps to $ \mathbb{P}^1$.

A map of moduli spaces

Let $ f:X \to \mathbb{P}^1$ be a nonconstant morphism with $ X$ a nonsingular projective variety over $ \mathbb{C}$, and let $ C \subset X$ be a closed subscheme which is a smooth curve of genus $ g$, such that the ramification of $ f\vert _C$ is simple. In a (formal) neighborhood of $ [C \subset X]$, the spaces $ \overline{\mathcal{M}_g}(X,[C])$ and the $ {\mathrm{Hilb}}_X$ are the same. The map

$\displaystyle \overline{\mathcal{M}_g}(X,[C]) \to \overline{\mathcal{M}_g}(\mathbb{P}^1,d) $

where $ d$ is the degree of $ f$ on $ C$, induces on tangent spaces

$\displaystyle H^0(C,\mathscr{N}_C X) \to \textstyle{\bigoplus}_{P\in{\mathrm{Ram}}(f\vert _C)} \mathscr{T}_{\mathbb{P}^1}\vert _{f(P)} $

induced by the right vertical arrow in the diagram

$\displaystyle \xymatrix{
0 \ar[r] & \mathscr{T}_C \ar@{=}[d] \ar[r] & \mathscr{...
...rt _C \ar[r] & (f^*\mathscr{T}_{\mathbb P^1}\vert C)/\mathscr{T}_C \ar[r] & 0

If $ C$ is contained in the smooth locus of $ f$, then the middle vertical map $ \mathscr{T}_X\vert _C \to f^*\mathscr{T}_{\mathbb{P}^1}\vert _C$ is surjective, hence also the right vertical map $ \mathscr{N}_C X \to (f^*\mathscr{T}_{\mathbb{P}^1}\vert C)/\mathscr{T}_C$ is also surjective as maps of sheaves. This will not give a surjection on global sections, however one has:

Corollary. If $ C$ is contained in the smooth locus of $ f$, and $ \mathscr{N}_C X$ is ``sufficiently positive'', then the morphism

$\displaystyle \overline{\mathcal{M}_g}(X,[C]) \to \overline{\mathcal{M}_g}(\mathbb{P}^1,d) $

analytically locally around $ [C \subset X]$ is a projection $ \mathbb{C}^{a+b} \to \mathbb{C}^b$.

One argues that the Hilbert scheme is smooth at the point $ [C \subset X]$ since one can twist by a small number of points and keep that the $ H^1$ vanishes. In particular, the corollary implies that the morphism is surjective.

We are now ready to prove:

Theorem. [G, Harris, Starr] If $ k=\overline{k}$, $ {\mathrm{char}}k=0$, then any rationally connected variety over $ k(C)$ for $ C$ a curve has a rational point.

Proof. Assume that $ (*)$: all fibres of $ X \to \mathbb{P}^1$ are reduced.

Step $ 1$. Take a general complete intersection $ C \subset X$; it will be smooth, irreducible, of say genus $ g$ and degree $ d$. The condition $ (*)$ implies that $ C$ is in the smooth locus and $ C \to \mathbb{P}^1$ (by Bertini) has at worst simple branching.

Step $ 2$. Choose a large integer $ N$ and choose general points $ c_1,\dots,c_N \in C$, and rational curves $ C_i \subset X$ such that:

  1. $ C \cap C_i =\{c_i\}$;
  2. $ C_i \subset f^{-1}(f(C_i))$;
  3. $ \mathscr{N}_{C_i} X_{f(c_i)}$ is very ample;
  4. $ \mathscr{T}_{c_i} C_i \subset \mathscr{T}_{c_i} X_{f(c_i)}$ is in general position.

Now let $ C^{\text{new}}=C \cup C_1 \cup \dots \cup C_N$. The basic property is that $ \mathscr{N}_{C^{\text{new}}} X\vert _{C_i} \supset \mathscr{N}_{C_i} X$. Moreover, $ \mathscr{N}_{C^{\text{new}}} X\vert _C \supset \mathscr{N}_C X$ with colength $ N$ and assumption (iv) gives that this is ``general''. This gives that the sheaf $ \mathscr{N}_{C^{\text{new}}} X$ on $ C^{\text{new}}$ is sufficiently positive.

Now deform this curve to a simply branched curve, and this gives the result; conclude by the corollary. $ \qedsymbol$

Multiple fibres

We must deal with the case when $ (*)$ fails. Suppose we have a family of varieties $ X \to \mathbb{P}^1$ with fibres at $ t_1,\dots,t_r$ irreducible of multiplicity $ m_1,\dots,m_r$. Since the curve must intersect these fibres transversally, this must be preserved in any deformation, meaning that the ramification index at $ t_i$ will be divisible by $ m_i$.

In this case, the problem is: $ \overline{\mathcal{M}_g}(X,\beta) \to \overline{\mathcal{M}_g}(\mathbb{P}^1,d)$ cannot dominate. Instead, we consider consider the subset $ Z_{g,d}^{(t_i,m_i)} \subset \overline{\mathcal{M}_g}(\mathbb{P}^1,d)$ consisting of stable maps $ f:C \to \mathbb{P}^1$, $ C$ of genus $ g$, $ f$ of degree $ d$, such that all ramification indices above $ t_i$ are equal to $ m_i$.

Now we have the additional problems: Which reducible curves are in $ Z_{g,d}^{(t_i,m_i)}$? And perhaps $ \dim Z$ is too small? To resolve both problems, enlarge the genus $ g$ (but not $ d$) by adding loops to $ C$: join two points with a good rational curve. This allows you to break off a component even in this case.


This is work with Jason Starr. What will guarantee the existence of a rational point on a variety over a function field in two variables? Is there a geometric condition which would be like rational connectedness in this case? This is too much to hope for, there are many surfaces $ S$ with a nontrivial Brauer-Severi variety $ X \to S$. Maybe there are geometric restrictions on the fibres $ X \to S$ such that one obtains a rational section.

A good guess for this condition: demand that certain moduli spaces of rational curves on the fibers are themselves rationally connected. For example, Starr and Harris proved that for hypersurfaces $ X$ of degree $ d$ in $ \mathbb{P}^n$ with $ d^2+d+2 \leq n$, the moduli spaces of rational curves of fixed degree on $ X$ are themselves rationally connected.

Back to the main index for Rational and integral points on higher dimensional varieties.