Chow's lemma

Not to be confused with Chow's theorem.

Chow's lemma, named after Wei-Liang Chow, is one of the foundational results in algebraic geometry. It roughly says that a proper morphism is fairly close to being a projective morphism. More precisely, a version of it states the following:[1]

If X is a scheme that is proper over a noetherian base S, then there exists a projective S-scheme X' and a surjective S-morphism f\colon X' \to X that induces an isomorphism f^{-1}(U) \simeq U for some dense open U\subseteq X.

Proof

The proof here is a standard one (cf. EGA II, 5.6.1).

It is easy to reduce to the case when X is irreducible, as follows. X is noetherian since it is of finite type over a noetherian base. Then it's also topologically noetherian, and consists of a finite number of irreducible components  X_i , which are each proper over S (because they're closed immersions in the scheme X which is proper over S). If, within each of these irreducible components, there exists a dense open U_i \subset X_i, then we can take U := \bigsqcup (U_i \setminus \bigcup \limits_{i \neq j} X_j). It is not hard to see that each of the disjoint pieces are dense in their respective X_i, so the full set U is dense in X. In addition, it's clear that we can similarly find a morphism g which satisfies the density condition.

Having reduced the problem, we now assume X is irreducible. We recall that it must also be noetherian. Thus, we can find a finite open affine cover X = \bigcup_{i=1}^n U_i. U_i are quasi-projective over S; there are open immersions over S, \phi_i \colon U_i \to P_i into some projective S-schemes P_i. Put U = \cap U_i. U is nonempty since X is irreducible. Let

\phi\colon U \to P = P_1 \times_S \cdots \times_S P_n.

be given by \phi_i's restricted to U over S. Let

\psi\colon U \to X \times_S P.

be given by U \hookrightarrow X and \phi over S. \psi is then an immersion; thus, it factors as an open immersion followed by a closed immersion X' \to X \times_S P. Let f\colon X' \to X be the immersion followed by the projection. We claim f induces f^{-1}(U) \simeq U; for that, it is enough to show f^{-1}(U) = \psi(U). But this means that \psi(U) is closed in U \times_S P. \psi factorizes as U \overset{\Gamma_\phi}\to U \times_S P \to X \times_S P. P is separated over S and so the graph morphism \Gamma_\phi is a closed immersion. This proves our contention.

It remains to show X' is projective over S. Let g\colon X' \to P be the closed immersion followed by the projection. Showing that g is a closed immersion shows X' is projective over S. This can be checked locally. Identifying U_i with its image in P_i we suppress \phi_i from our notation.

Let V_i = p_i^{-1}(U_i) where p_i \colon P \to P_i. We claim g^{-1}(V_i) are an open cover of X'. This would follow from f^{-1}(U_i) \subset g^{-1}(V_i) as sets. This in turn follows from f = p_i \circ g on U_i as functions on the underlying topological space. Since X is separated over S and U_i is dense, this is clear from looking at the relevant commutative diagram. Now, X \times_S P \to P is closed since it is a base extension of the proper morphism X \to S. Thus, g(X') is a closed subscheme covered by V_i, and so it is enough to show that for each i the map g\colon g^{-1}(V_i) \to V_i, denoted by h, is a closed immersion.

Fix i. Let Z be the graph of u\colon V_i \overset{p_i}\to U_i \hookrightarrow X. It is a closed subscheme of X \times_S V_i since X is separated over S. Let q_1 \colon X \times_S P \to X,\ q_2 \colon X \times_S P \to P be the projections. We claim that h factors through Z, which would imply h is a closed immersion. But for w\colon U' \to V_i we have:

v = \Gamma_u \circ w \quad\Leftrightarrow\quad q_1 \circ v = u \circ q_2 \circ v \quad\Leftrightarrow\quad q_1 \circ \psi = u \circ q_2 \circ \psi \quad\Leftrightarrow\quad q_1 \circ \psi = u \circ \phi.

The last equality holds and thus there is w that satisfies the first equality. This proves our claim. \square

Additional statements

In the statement of Chow's lemma, if X is reduced, irreducible, or integral, we can assume that the same holds for X'. If both X and X' are irreducible, then f\colon X' \to X is a birational morphism. (cf. EGA II, 5.6).

References

  1. Hartshorne, Ch II. Exercise 4.10