The Stacks project

53.23 Contracting rational bridges

In this section we discuss the next simplest possible case (after the case discussed in Section 53.22) of contracting a scheme to improve positivity properties of its canonical sheaf.

Example 53.23.1 (Contracting a rational bridge). Let $k$ be a field. Let $X$ be a proper scheme over $k$ having dimension $1$ and $H^0(X, \mathcal{O}_ X) = k$. Assume the singularities of $X$ are at-worst-nodal. A rational bridge will be an irreducible component $C \subset X$ (viewed as an integral closed subscheme) with the following properties

  1. $X' \not= \emptyset $ where $X' \subset X$ is the scheme theoretic closure of $X \setminus C$,

  2. the scheme theoretic intersection $C \cap X'$ has degree $2$ over $H^0(C, \mathcal{O}_ C)$, and

  3. $C$ has genus zero.

Set $k' = H^0(C, \mathcal{O}_ C)$ and $k'' = H^0(C \cap X', \mathcal{O}_{C \cap X'})$. Then $k'$ is a field (Varieties, Lemma 33.9.3) and $\dim _{k'}(k'') = 2$. Since there are at least two irreducible components of $X$ passing through each point of $C \cap X'$, we conclude these points are nodes of $X$ and smooth points on both $C$ and $X'$ (Lemma 53.19.17). Hence $k'/k$ is a finite separable extension of fields and $k''/k'$ is either a degree $2$ separable extension of fields or $k'' = k' \times k'$ (Lemma 53.19.7). By Section 53.14 there exists a pushout

\[ \xymatrix{ C \cap X' \ar[r] \ar[d] & X' \ar[d]^ a \\ \mathop{\mathrm{Spec}}(k') \ar[r] & Y } \]

with many good properties (all of which we will use below without further mention). Let $y \in Y$ be the image of $\mathop{\mathrm{Spec}}(k') \to Y$. Then

\[ \mathcal{O}_{Y, y}^\wedge \cong k'[[s, t]]/(st) \quad \text{or}\quad \mathcal{O}_{Y, y}^\wedge \cong \{ f \in k''[[s]] : f(0) \in k'\} \]

depending on whether $C \cap X'$ has $2$ or $1$ points. This follows from Lemma 53.14.1 and the fact that $\mathcal{O}_{X', p} \cong \kappa (p)[[t]]$ for $p \in C \cap X'$ by More on Algebra, Lemma 15.38.4. Thus we see that $y \in Y$ is a node, see Lemmas 53.19.7 and 53.19.4 and in particular the discussion of Case II in the proof of (2) $\Rightarrow $ (1) in Lemma 53.19.4. Thus the singularities of $Y$ are at-worst-nodal.

We can extend the commutative diagram above to a diagram

\[ \xymatrix{ C \cap X' \ar[r] \ar[d] & X' \ar[d]^ a \ar[r] & X \ar[ld]^ c & C \ar[ld] \ar[l] \\ \mathop{\mathrm{Spec}}(k') \ar[r] & Y & \mathop{\mathrm{Spec}}(k') \ar[l] } \]

where the two lower horizontal arrows are the same. Namely, $X$ is the scheme theoretic union of $X'$ and $C$ (thus a pushout by Morphisms, Lemma 29.4.6) and the morphisms $C \to Y$ and $X' \to Y$ agree on $C \cap X'$. Finally, we claim that

\[ c_*\mathcal{O}_ X = \mathcal{O}_ Y \quad \text{and}\quad R^1c_*\mathcal{O}_ X = 0 \]

To see this use the exact sequence

\[ 0 \to \mathcal{O}_ X \to \mathcal{O}_ C \oplus \mathcal{O}_{X'} \to \mathcal{O}_{C \cap X'} \to 0 \]

of Morphisms, Lemma 29.4.6. The long exact sequence of higher direct images is

\[ 0 \to c_*\mathcal{O}_ X \to c_*\mathcal{O}_ C \oplus c_*\mathcal{O}_{X'} \to c_*\mathcal{O}_{C \cap X'} \to R^1c_*\mathcal{O}_ X \to R^1c_*\mathcal{O}_ C \oplus R^1c_*\mathcal{O}_{X'} \]

Since $c|_{X'} = a$ is affine we see that $R^1c_*\mathcal{O}_{X'} = 0$. Since $c|_ C$ factors as $C \to \mathop{\mathrm{Spec}}(k') \to X$ and since $C$ has genus zero, we find that $R^1c_*\mathcal{O}_ C = 0$. Since $\mathcal{O}_{X'} \to \mathcal{O}_{C \cap X'}$ is surjective and since $c|_{X'}$ is affine, we see that $c_*\mathcal{O}_{X'} \to c_*\mathcal{O}_{C \cap X'}$ is surjective. This proves that $R^1c_*\mathcal{O}_ X = 0$. Finally, we have $\mathcal{O}_ Y = c_*\mathcal{O}_ X$ by the exact sequence and the description of the structure sheaf of the pushout in More on Morphisms, Proposition 37.67.3.

All of this means that $Y$ is also a proper scheme over $k$ having dimension $1$ and $H^0(Y, \mathcal{O}_ Y) = k$ whose singularities are at-worst-nodal (Lemma 53.19.17) and that $Y$ has the same genus as $X$. We will say $c : X \to Y$ is the contraction of a rational bridge.

Lemma 53.23.2. Let $k$ be a field. Let $X$ be a proper scheme over $k$ having dimension $1$ and $H^0(X, \mathcal{O}_ X) = k$. Assume the singularities of $X$ are at-worst-nodal. Let $C \subset X$ be a rational bridge (Example 53.23.1). Then $\deg (\omega _ X|_ C) = 0$.

Proof. Let $X' \subset X$ be as in the example. Then we have a short exact sequence

\[ 0 \to \omega _ C \to \omega _ X|_ C \to \mathcal{O}_{C \cap X'} \to 0 \]

See Lemmas 53.4.6, 53.19.16, and 53.19.17. With $k''/k'/k$ as in the example we see that $\deg (\omega _ C) = -2[k' : k]$ as $C$ has genus $0$ (Lemma 53.5.2) and $\deg (C \cap X') = [k'' : k] = 2[k' : k]$. Hence $\deg (\omega _ X|_ C) = 0$. $\square$

Lemma 53.23.3. Let $k$ be a field. Let $X$ be a proper scheme over $k$ having dimension $1$ and $H^0(X, \mathcal{O}_ X) = k$. Assume the singularities of $X$ are at-worst-nodal. Let $C \subset X$ be a rational bridge (Example 53.23.1). For any field extension $K/k$ the base change $C_ K \subset X_ K$ is a finite disjoint union of rational bridges.

Proof. Let $k''/k'/k$ be as in the example. Since $k'/k$ is finite separable, we see that $k' \otimes _ k K = K'_1 \times \ldots \times K'_ n$ is a finite product of finite separable extensions $K'_ i/K$. The corresponding product decomposition $k'' \otimes _ k K = \prod K''_ i$ gives degree $2$ separable algebra extensions $K''_ i/K'_ i$. Set $C_ i = C_{K'_ i}$. Then $C_ K = \coprod C_ i$ and therefore each $C_ i$ has genus $0$ (viewed as a curve over $K'_ i$), because $H^1(C_ K, \mathcal{O}_{C_ K}) = 0$ by flat base change. Finally, we have $X'_ K \cap C_ i = \mathop{\mathrm{Spec}}(K''_ i)$ has degree $2$ over $K'_ i$ as desired. $\square$

Lemma 53.23.4. Let $c : X \to Y$ be the contraction of a rational bridge (Example 53.23.1). Then $c^*\omega _ Y \cong \omega _ X$.

Proof. You can prove this by direct computation, but we prefer to use the characterization of $\omega _ X$ as the coherent $\mathcal{O}_ X$-module which represents the functor $\textit{Coh}(\mathcal{O}_ X) \to \textit{Sets}$, $\mathcal{F} \mapsto \mathop{\mathrm{Hom}}\nolimits _ k(H^1(X, \mathcal{F}), k) = H^1(X, \mathcal{F})^\vee $, see Lemma 53.4.2 or Duality for Schemes, Lemma 48.22.5.

To be precise, denote $\mathcal{C}_ Y$ the category whose objects are invertible $\mathcal{O}_ Y$-modules and whose maps are $\mathcal{O}_ Y$-module homomorphisms. Denote $\mathcal{C}_ X$ the category whose objects are invertible $\mathcal{O}_ X$-modules $\mathcal{L}$ with $\mathcal{L}|_ C \cong \mathcal{O}_ C$ and whose maps are $\mathcal{O}_ Y$-module homomorphisms. We claim that the functor

\[ c^* : \mathcal{C}_ Y \to \mathcal{C}_ X \]

is an equivalence of categories. Namely, by More on Morphisms, Lemma 37.72.8 it is essentially surjective. Then the projection formula (Cohomology, Lemma 20.54.2) shows $c_*c^*\mathcal{N} = \mathcal{N}$ and hence $c^*$ is an equivalence with quasi-inverse given by $c_*$.

We claim $\omega _ X$ is an object of $\mathcal{C}_ X$. Namely, we have a short exact sequence

\[ 0 \to \omega _ C \to \omega _ X|_ C \to \mathcal{O}_{C \cap X'} \to 0 \]

See Lemma 53.4.6. Taking degrees we find $\deg (\omega _ X|_ C) = 0$ (small detail omitted). Thus $\omega _ X|_ C$ is trivial by Lemma 53.10.1 and $\omega _ X$ is an object of $\mathcal{C}_ X$.

Since $R^1c_*\mathcal{O}_ X = 0$ the projection formula shows that $R^1c_*c^*\mathcal{N} = 0$ for $\mathcal{N} \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{C}_ Y)$. Therefore the Leray spectral sequence (Cohomology, Lemma 20.13.6) the diagram

\[ \xymatrix{ \mathcal{C}_ Y \ar[rr]_{c^*} \ar[dr]_{H^1(Y, -)^\vee } & & \mathcal{C}_ X \ar[ld]^{H^1(X, -)^\vee } \\ & \textit{Sets} } \]

of categories and functors is commutative. Since $\omega _ Y \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{C}_ Y)$ represents the south-east arrow and $\omega _ X \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{C}_ X)$ represents the south-east arrow we conclude by the Yoneda lemma (Categories, Lemma 4.3.5). $\square$

Lemma 53.23.5. Let $k$ be a field. Let $X$ be a proper scheme over $k$ having dimension $1$ and $H^0(X, \mathcal{O}_ X) = k$. Assume

  1. the singularities of $X$ are at-worst-nodal,

  2. $X$ does not have a rational tail (Example 53.22.1),

  3. $X$ does not have a rational bridge (Example 53.23.1),

  4. the genus $g$ of $X$ is $\geq 2$.

Then $\omega _ X$ is ample.

Proof. It suffices to show that $\deg (\omega _ X|_ C) > 0$ for every irreducible component $C$ of $X$, see Varieties, Lemma 33.44.15. If $X = C$ is irreducible, this follows from $g \geq 2$ and Lemma 53.8.3. Otherwise, set $k' = H^0(C, \mathcal{O}_ C)$. This is a field and a finite extension of $k$ and $[k' : k]$ divides all numerical invariants below associated to $C$ and coherent sheaves on $C$, see Varieties, Lemma 33.44.10. Let $X' \subset X$ be the closure of $X \setminus C$ as in Lemma 53.4.6. We will use the results of this lemma and of Lemmas 53.19.16 and 53.19.17 without further mention. Then we get a short exact sequence

\[ 0 \to \omega _ C \to \omega _ X|_ C \to \mathcal{O}_{C \cap X'} \to 0 \]

See Lemma 53.4.6. We conclude that

\[ \deg (\omega _ X|_ C) = \deg (C \cap X') + \deg (\omega _ C) = \deg (C \cap X') - 2\chi (C, \mathcal{O}_ C) \]

Hence, if the lemma is false, then

\[ 2[k' : k] \geq \deg (C \cap X') + 2\dim _ k H^1(C, \mathcal{O}_ C) \]

Since $C \cap X'$ is nonempty and by the divisiblity mentioned above, this can happen only if either

  1. $C \cap X'$ is a single $k'$-rational point of $C$ and $H^1(C, \mathcal{O}_ C) = 0$, and

  2. $C \cap X'$ has degree $2$ over $k'$ and $H^1(C, \mathcal{O}_ C) = 0$.

The first possibility means $C$ is a rational tail and the second that $C$ is a rational bridge. Since both are excluded the proof is complete. $\square$

Lemma 53.23.6. Let $k$ be a field. Let $X$ be a proper scheme over $k$ of dimension $1$ with $H^0(X, \mathcal{O}_ X) = k$ having genus $g \geq 2$. Assume the singularities of $X$ are at-worst-nodal and that $X$ has no rational tails. Consider a sequence

\[ X = X_0 \to X_1 \to \ldots \to X_ n = X' \]

of contractions of rational bridges (Example 53.23.1) until none are left. Then $\omega _{X'}$ ample. The morphism $X \to X'$ is independent of choices and formation of this morphism commutes with base field extensions.

Proof. We proceed by contracting rational bridges until there are none left. Then $\omega _{X'}$ is ample by Lemma 53.23.5.

Denote $f : X \to X'$ the composition. By Lemma 53.23.4 and induction we see that $f^*\omega _{X'} = \omega _ X$. We have $f_*\mathcal{O}_ X = \mathcal{O}_{X'}$ because this is true for contraction of a rational bridge. Thus the projection formula says that $f_*f^*\mathcal{L} = \mathcal{L}$ for all invertible $\mathcal{O}_{X'}$-modules $\mathcal{L}$. Hence

\[ \Gamma (X', \omega _{X'}^{\otimes m}) = \Gamma (X, \omega _ X^{\otimes m}) \]

for all $m$. Since $X'$ is the Proj of the direct sum of these by Morphisms, Lemma 29.43.17 we conclude that the morphism $X \to X'$ is completely canonical.

Let $K/k$ be an extension of fields, then $\omega _{X_ K}$ is the pullback of $\omega _ X$ (Lemma 53.4.4) and we have $\Gamma (X, \omega _ X^{\otimes m}) \otimes _ k K$ is equal to $\Gamma (X_ K, \omega _{X_ K}^{\otimes m})$ by Cohomology of Schemes, Lemma 30.5.2. Thus formation of $f : X \to X'$ commutes with base change by $K/k$ by the arguments given above. Some details omitted. $\square$


Comments (4)

Comment #4549 by Jonas Ehrhard on

The intro text here is exactly the same as in the previous section. Both claim to be "the simplest possible case" of cantractions.

Comment #7450 by Christophe Marciot on

In the intro at \emph{we conclude these points are nodes and smooth points on both} and , would it a bit better for comprehension to add a \emph{respectively} after the ?


Post a comment

Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like $\pi$). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.




In order to prevent bots from posting comments, we would like you to prove that you are human. You can do this by filling in the name of the current tag in the following input field. As a reminder, this is tag 0E7M. Beware of the difference between the letter 'O' and the digit '0'.