The Stacks project

37.53 Stein factorization

Stein factorization is the statement that a proper morphism $f : X \to S$ with $f_*\mathcal{O}_ X = \mathcal{O}_ S$ has connected fibres.

Lemma 37.53.1. Let $S$ be a scheme. Let $f : X \to S$ be a universally closed and quasi-separated morphism. There exists a factorization

\[ \xymatrix{ X \ar[rr]_{f'} \ar[rd]_ f & & S' \ar[dl]^\pi \\ & S & } \]

with the following properties:

  1. the morphism $f'$ is universally closed, quasi-compact, quasi-separated, and surjective,

  2. the morphism $\pi : S' \to S$ is integral,

  3. we have $f'_*\mathcal{O}_ X = \mathcal{O}_{S'}$,

  4. we have $S' = \underline{\mathop{\mathrm{Spec}}}_ S(f_*\mathcal{O}_ X)$, and

  5. $S'$ is the normalization of $S$ in $X$, see Morphisms, Definition 29.53.3.

Formation of the factorization $f = \pi \circ f'$ commutes with flat base change.

Proof. By Morphisms, Lemma 29.41.8 the morphism $f$ is quasi-compact. Hence the normalization $S'$ of $S$ in $X$ is defined (Morphisms, Definition 29.53.3) and we have the factorization $X \to S' \to S$. By Morphisms, Lemma 29.53.11 we have (2), (4), and (5). The morphism $f'$ is universally closed by Morphisms, Lemma 29.41.7. It is quasi-compact by Schemes, Lemma 26.21.14 and quasi-separated by Schemes, Lemma 26.21.13.

To show the remaining statements we may assume the base scheme $S$ is affine, say $S = \mathop{\mathrm{Spec}}(R)$. Then $S' = \mathop{\mathrm{Spec}}(A)$ with $A = \Gamma (X, \mathcal{O}_ X)$ an integral $R$-algebra. Thus it is clear that $f'_*\mathcal{O}_ X$ is $\mathcal{O}_{S'}$ (because $f'_*\mathcal{O}_ X$ is quasi-coherent, by Schemes, Lemma 26.24.1, and hence equal to $\widetilde{A}$). This proves (3).

Let us show that $f'$ is surjective. As $f'$ is universally closed (see above) the image of $f'$ is a closed subset $V(I) \subset S' = \mathop{\mathrm{Spec}}(A)$. Pick $h \in I$. Then $h|_ X = f^\sharp (h)$ is a global section of the structure sheaf of $X$ which vanishes at every point. As $X$ is quasi-compact this means that $h|_ X$ is a nilpotent section, i.e., $h^ n|X = 0$ for some $n > 0$. But $A = \Gamma (X, \mathcal{O}_ X)$, hence $h^ n = 0$. As every element of $I$ is nilpotent, we conclude that $V(I) = S'$ as desired.

By Cohomology of Schemes, Lemma 30.5.2 we see that formation of $f_*\mathcal{O}_ X$ commutes with flat base change. Formation of the relative spectrum commutes with any base change by Constructions, Lemma 27.4.6. Thus formation of the factorization commutes with flat base change. $\square$

Lemma 37.53.2. In Lemma 37.53.1 assume in addition that $f$ is locally of finite type. Then for $s \in S$ the fibre $\pi ^{-1}(\{ s\} ) = \{ s_1, \ldots , s_ n\} $ is finite and the field extensions $\kappa (s_ i)/\kappa (s)$ are finite.

Proof. Recall that there are no specializations among the points of $\pi ^{-1}(\{ s\} )$, see Algebra, Lemma 10.36.20. As $f'$ is surjective, we find that $|X_ s| \to \pi ^{-1}(\{ s\} )$ is surjective. Observe that $X_ s$ is a quasi-separated scheme of finite type over a field (quasi-compactness was shown in the proof of the referenced lemma). Thus $X_ s$ is Noetherian (Morphisms, Lemma 29.15.6). A topological argument (omitted) now shows that $\pi ^{-1}(\{ s\} )$ is finite. For each $i$ we can pick a finite type point $x_ i \in X_ s$ mapping to $s_ i$ (Morphisms, Lemma 29.16.7). We conclude that $\kappa (s_ i)/\kappa (s)$ is finite: $x_ i$ can be represented by a morphism $\mathop{\mathrm{Spec}}(k_ i) \to X_ s$ of finite type (by our definition of finite type points) and hence $\mathop{\mathrm{Spec}}(k_ i) \to s = \mathop{\mathrm{Spec}}(\kappa (s))$ is of finite type (as a composition of finite type morphisms), hence $k_ i/\kappa (s)$ is finite (Morphisms, Lemma 29.16.1). $\square$

Lemma 37.53.3. Let $f : X \to S$ be a morphism of schemes. Let $s \in S$. Then $X_ s$ is geometrically connected, if and only if for every étale neighbourhood $(U, u) \to (S, s)$ the base change $X_ U \to U$ has connected fibre $X_ u$.

Proof. If $X_ s$ is geometrically connected, then any base change of it is connected. On the other hand, suppose that $X_ s$ is not geometrically connected. Then by Varieties, Lemma 33.7.11 we see that $X_ s \times _{\mathop{\mathrm{Spec}}(\kappa (s))} \mathop{\mathrm{Spec}}(k)$ is disconnected for some finite separable field extension $k/\kappa (s)$. By Lemma 37.35.2 there exists an affine étale neighbourhood $(U, u) \to (S, s)$ such that $\kappa (u)/\kappa (s)$ is identified with $k/\kappa (s)$. In this case $X_ u$ is disconnected. $\square$

Theorem 37.53.4 (Stein factorization; Noetherian case). Let $S$ be a locally Noetherian scheme. Let $f : X \to S$ be a proper morphism. There exists a factorization

\[ \xymatrix{ X \ar[rr]_{f'} \ar[rd]_ f & & S' \ar[dl]^\pi \\ & S & } \]

with the following properties:

  1. the morphism $f'$ is proper with geometrically connected fibres,

  2. the morphism $\pi : S' \to S$ is finite,

  3. we have $f'_*\mathcal{O}_ X = \mathcal{O}_{S'}$,

  4. we have $S' = \underline{\mathop{\mathrm{Spec}}}_ S(f_*\mathcal{O}_ X)$, and

  5. $S'$ is the normalization of $S$ in $X$, see Morphisms, Definition 29.53.3.

Proof. Let $f = \pi \circ f'$ be the factorization of Lemma 37.53.1. Note that besides the conclusions of Lemma 37.53.1 we also have that $f'$ is separated (Schemes, Lemma 26.21.13) and finite type (Morphisms, Lemma 29.15.8). Hence $f'$ is proper. By Cohomology of Schemes, Proposition 30.19.1 we see that $f_*\mathcal{O}_ X$ is a coherent $\mathcal{O}_ S$-module. Hence we see that $\pi $ is finite, i.e., (2) holds.

This proves all but the most interesting assertion, namely that all the fibres of $f'$ are geometrically connected. It is clear from the discussion above that we may replace $S$ by $S'$, and we may therefore assume that $S$ is Noetherian, affine, $f : X \to S$ is proper, and $f_*\mathcal{O}_ X = \mathcal{O}_ S$. Let $s \in S$ be a point of $S$. We have to show that $X_ s$ is geometrically connected. By Lemma 37.53.3 we see that it suffices to show $X_ u$ is connected for every étale neighbourhood $(U, u) \to (S, s)$. We may assume $U$ is affine. Thus $U$ is Noetherian (Morphisms, Lemma 29.15.6), the base change $f_ U : X_ U \to U$ is proper (Morphisms, Lemma 29.41.5), and that also $(f_ U)_*\mathcal{O}_{X_ U} = \mathcal{O}_ U$ (Cohomology of Schemes, Lemma 30.5.2). Hence after replacing $(f : X \to S, s)$ by the base change $(f_ U : X_ U \to U, u)$ it suffices to prove that the fibre $X_ s$ is connected when $f_*\mathcal{O}_ X = \mathcal{O}_ S$. We can deduce this from Derived Categories of Schemes, Lemma 36.32.7 (by looking at idempotents in the structure sheaf of $X_ s$) but we will also give a direct argument below.

Namely, we apply the theorem on formal functions, more precisely Cohomology of Schemes, Lemma 30.20.7. It tells us that

\[ \mathcal{O}^\wedge _{S, s} = (f_*\mathcal{O}_ X)_ s^\wedge = \mathop{\mathrm{lim}}\nolimits _ n H^0(X_ n, \mathcal{O}_{X_ n}) \]

where $X_ n$ is the $n$th infinitesimal neighbourhood of $X_ s$. Since the underlying topological space of $X_ n$ is equal to that of $X_ s$ we see that if $X_ s = T_1 \amalg T_2$ is a disjoint union of nonempty open and closed subschemes, then similarly $X_ n = T_{1, n} \amalg T_{2, n}$ for all $n$. And this in turn means $H^0(X_ n, \mathcal{O}_{X_ n})$ contains a nontrivial idempotent $e_{1, n}$, namely the function which is identically $1$ on $T_{1, n}$ and identically $0$ on $T_{2, n}$. It is clear that $e_{1, n + 1}$ restricts to $e_{1, n}$ on $X_ n$. Hence $e_1 = \mathop{\mathrm{lim}}\nolimits e_{1, n}$ is a nontrivial idempotent of the limit. This contradicts the fact that $\mathcal{O}^\wedge _{S, s}$ is a local ring. Thus the assumption was wrong, i.e., $X_ s$ is connected, and we win. $\square$

Theorem 37.53.5 (Stein factorization; general case). Let $S$ be a scheme. Let $f : X \to S$ be a proper morphism. There exists a factorization

\[ \xymatrix{ X \ar[rr]_{f'} \ar[rd]_ f & & S' \ar[dl]^\pi \\ & S & } \]

with the following properties:

  1. the morphism $f'$ is proper with geometrically connected fibres,

  2. the morphism $\pi : S' \to S$ is integral,

  3. we have $f'_*\mathcal{O}_ X = \mathcal{O}_{S'}$,

  4. we have $S' = \underline{\mathop{\mathrm{Spec}}}_ S(f_*\mathcal{O}_ X)$, and

  5. $S'$ is the normalization of $S$ in $X$, see Morphisms, Definition 29.53.3.

Proof. We may apply Lemma 37.53.1 to get the morphism $f' : X \to S'$. Note that besides the conclusions of Lemma 37.53.1 we also have that $f'$ is separated (Schemes, Lemma 26.21.13) and finite type (Morphisms, Lemma 29.15.8). Hence $f'$ is proper. At this point we have proved all of the statements except for the statement that $f'$ has geometrically connected fibres.

We may assume that $S = \mathop{\mathrm{Spec}}(R)$ is affine. Set $R' = \Gamma (X, \mathcal{O}_ X)$. Then $S' = \mathop{\mathrm{Spec}}(R')$. Thus we may replace $S$ by $S'$ and assume that $S = \mathop{\mathrm{Spec}}(R)$ is affine $R = \Gamma (X, \mathcal{O}_ X)$. Next, let $s \in S$ be a point. Let $U \to S$ be an étale morphism of affine schemes and let $u \in U$ be a point mapping to $s$. Let $X_ U \to U$ be the base change of $X$. By Lemma 37.53.3 it suffices to show that the fibre of $X_ U \to U$ over $u$ is connected. By Cohomology of Schemes, Lemma 30.5.2 we see that $\Gamma (X_ U, \mathcal{O}_{X_ U}) = \Gamma (U, \mathcal{O}_ U)$. Hence we have to show: Given $S = \mathop{\mathrm{Spec}}(R)$ affine, $X \to S$ proper with $\Gamma (X, \mathcal{O}_ X) = R$ and $s \in S$ is a point, the fibre $X_ s$ is connected.

To do this it suffices to show that the only idempotents $e \in H^0(X_ s, \mathcal{O}_{X_ s})$ are $0$ and $1$ (we already know that $X_ s$ is nonempty by Lemma 37.53.1). By Derived Categories of Schemes, Lemma 36.32.7 after replacing $R$ by a principal localization we may assume $e$ is the image of an element of $R$. Since $R \to H^0(X_ s, \mathcal{O}_{X_ s})$ factors through $\kappa (s)$ we conclude. $\square$

Here is an application.

Lemma 37.53.6. Let $f : X \to S$ be a morphism of schemes. Assume

  1. $f$ is proper,

  2. $S$ is integral with generic point $\xi $,

  3. $S$ is normal,

  4. $X$ is reduced,

  5. every generic point of an irreducible component of $X$ maps to $\xi $,

  6. we have $H^0(X_\xi , \mathcal{O}) = \kappa (\xi )$.

Then $f_*\mathcal{O}_ X = \mathcal{O}_ S$ and $f$ has geometrically connected fibres.

Proof. Apply Theorem 37.53.5 to get a factorization $X \to S' \to S$. It is enough to show that $S' = S$. This will follow from Morphisms, Lemma 29.54.8. Namely, $S'$ is reduced because $X$ is reduced (Morphisms, Lemma 29.53.8). The morphism $S' \to S$ is integral by the theorem cited above. Every generic point of $S'$ lies over $\xi $ by Morphisms, Lemma 29.53.9 and assumption (5). On the other hand, since $S'$ is the relative spectrum of $f_*\mathcal{O}_ X$ we see that the scheme theoretic fibre $S'_\xi $ is the spectrum of $H^0(X_\xi , \mathcal{O})$ which is equal to $\kappa (\xi )$ by assumption. Hence $S'$ is an integral scheme with function field equal to the function field of $S$. This finishes the proof. $\square$

Here is another application.

Lemma 37.53.7. Let $X \to S$ be a flat proper morphism of finite presentation. Let $n_{X/S}$ be the function on $S$ counting the numbers of geometric connected components of fibres of $f$ introduced in Lemma 37.28.3. Then $n_{X/S}$ is lower semi-continuous.

Proof. Let $s \in S$. Set $n = n_{X/S}(s)$. Note that $n < \infty $ as the geometric fibre of $X \to S$ at $s$ is a proper scheme over a field, hence Noetherian, hence has a finite number of connected components. We have to find an open neighbourhood $V$ of $s$ such that $n_{X/S}|_ V \geq n$. Let $X \to S' \to S$ be the Stein factorization as in Theorem 37.53.5. By Lemma 37.53.2 there are finitely many points $s'_1, \ldots , s'_ m \in S'$ lying over $s$ and the extensions $\kappa (s'_ i)/\kappa (s)$ are finite. Then Lemma 37.42.1 tells us that after replacing $S$ by an étale neighbourhood of $s$ we may assume $S' = V_1 \amalg \ldots \amalg V_ m$ as a scheme with $s'_ i \in V_ i$ and $\kappa (s'_ i)/\kappa (s)$ purely inseparable. Then the schemes $X_{s_ i'}$ are geometrically connected over $\kappa (s)$, hence $m = n$. The schemes $X_ i = (f')^{-1}(V_ i)$, $i = 1, \ldots , n$ are flat and of finite presentation over $S$. Hence the image of $X_ i \to S$ is open (Morphisms, Lemma 29.25.10). Thus in a neighbourhood of $s$ we see that $n_{X/S}$ is at least $n$. $\square$

Lemma 37.53.8. Let $f : X \to S$ be a morphism of schemes. Assume

  1. $f$ is proper, flat, and of finite presentation, and

  2. the geometric fibres of $f$ are reduced.

Then the function $n_{X/S} : S \to \mathbf{Z}$ counting the numbers of geometric connected components of fibres of $f$ is locally constant.

Proof. By Lemma 37.53.7 the function $n_{X/S}$ is lower semincontinuous. For $s \in S$ consider the $\kappa (s)$-algebra

\[ A = H^0(X_ s, \mathcal{O}_{X_ s}) \]

By Varieties, Lemma 33.9.3 and the fact that $X_ s$ is geometrically reduced $A$ is finite product of finite separable extensions of $\kappa (s)$. Hence $A \otimes _{\kappa (s)} \kappa (\overline{s})$ is a product of $\beta _0(s) = \dim _{\kappa (s)} H^0(E \otimes ^\mathbf {L} \kappa (s))$ copies of $\kappa (\overline{s})$. Thus $X_{\overline{s}}$ has $\beta _0(s) = \dim _{\kappa (s)} A$ connected components. In other words, we have $n_{X/S} = \beta _0$ as functions on $S$. Thus $n_{X/S}$ is upper semi-continuous by Derived Categories of Schemes, Lemma 36.32.1. This finishes the proof. $\square$

A final application.

reference

Lemma 37.53.9. Let $(A, I)$ be a henselian pair. Let $X \to \mathop{\mathrm{Spec}}(A)$ be separated and of finite type. Set $X_0 = X \times _{\mathop{\mathrm{Spec}}(A)} \mathop{\mathrm{Spec}}(A/I)$. Let $Y \subset X_0$ be an open and closed subscheme such that $Y \to \mathop{\mathrm{Spec}}(A/I)$ is proper. Then there exists an open and closed subscheme $W \subset X$ which is proper over $A$ with $W \times _{\mathop{\mathrm{Spec}}(A)} \mathop{\mathrm{Spec}}(A/I) = Y$.

Proof. We will denote $T \mapsto T_0$ the base change by $\mathop{\mathrm{Spec}}(A/I) \to \mathop{\mathrm{Spec}}(A)$. By Chow's lemma (in the form of Limits, Lemma 32.12.1) there exists a surjective proper morphism $\varphi : X' \to X$ such that $X'$ admits an immersion into $\mathbf{P}^ n_ A$. Set $Y' = \varphi ^{-1}(Y)$. This is an open and closed subscheme of $X'_0$. Suppose the lemma holds for $(X', Y')$. Let $W' \subset X'$ be the open and closed subscheme proper over $A$ such that $Y' = W'_0$. By Morphisms, Lemma 29.41.7 $W = \varphi (W') \subset X$ and $Q = \varphi (X' \setminus W') \subset X$ are closed subsets and by Morphisms, Lemma 29.41.9 $W$ is proper over $A$. The image of $W \cap Q$ in $\mathop{\mathrm{Spec}}(A)$ is closed. Since $(A, I)$ is henselian, if $W \cap Q$ is nonempty, then we find that $W \cap Q$ has a point lying over $\mathop{\mathrm{Spec}}(A/I)$. This is impossible as $W'_0 = Y' = \varphi ^{-1}(Y)$. We conclude that $W$ is an open and closed subscheme of $X$ proper over $A$ with $W_0 = Y$. Thus we reduce to the case described in the next paragraph.

Assume there exists an immersion $j : X \to \mathbf{P}^ n_ A$ over $A$. Let $\overline{X}$ be the scheme theoretic image of $j$. Since $j$ is a quasi-compact morphism (Schemes, Lemma 26.21.14) we see that $j : X \to \overline{X}$ is an open immersion (Morphisms, Lemma 29.7.7). Hence the base change $j_0 : X_0 \to \overline{X}_0$ is an open immersion as well. Thus $j_0(Y) \subset \overline{X}_0$ is open. It is also closed by Morphisms, Lemma 29.41.7. Suppose that the lemma holds for $(\overline{X}, j_0(Y))$. Let $\overline{W} \subset \overline{X}$ be the corresponding open and closed subscheme proper over $A$ such that $j_0(Y) = \overline{W}_0$. Then $T = \overline{W} \setminus j(X)$ is closed in $\overline{W}$, hence has closed image in $\mathop{\mathrm{Spec}}(A)$ by properness of $\overline{W}$ over $A$. Since $(A, I)$ is henselian, we find that if $T$ is nonempty, then there is a point of $T$ mapping into $\mathop{\mathrm{Spec}}(A/I)$. This is impossible because $j_0(Y) = \overline{W}_0$ is contained in $j(X)$. Hence $\overline{W}$ is contained in $j(X)$ and we can set $W \subset X$ equal to the unique open and closed subscheme mapping isomorphically to $\overline{W}$ via $j$. Thus we reduce to the case described in the next paragraph.

Assume $X \subset \mathbf{P}^ n_ A$ is a closed subscheme. Then $X \to \mathop{\mathrm{Spec}}(A)$ is a proper morphism. Let $Z = X_0 \setminus Y$. This is an open and closed subscheme of $X_0$ and $X_0 = Y \amalg Z$. Let $X \to X' \to \mathop{\mathrm{Spec}}(A)$ be the Stein factorization as in Theorem 37.53.5. Let $Y' \subset X'_0$ and $Z' \subset X'_0$ be the images of $Y$ and $Z$. Since the fibres of $X \to Z$ are geometrically connected, we see that $Y' \cap Z' = \emptyset $. Hence $X'_0 = Y' \amalg Z'$ as $X \to X'$ is surjective. Since $X' \to \mathop{\mathrm{Spec}}(A)$ is integral, we see that $X'$ is the spectrum of an $A$-algebra integral over $A$. Recall that open and closed subsets of spectra correspond $1$-to-$1$ with idempotents in the corresponding ring, see Algebra, Lemma 10.21.3. Hence by More on Algebra, Lemma 15.11.6 we see that we may write $X' = W' \amalg V'$ with $W'$ and $V'$ open and closed and with $Y' = W'_0$ and $Z' = V'_0$. Let $W$ be the inverse image in $X$ to finish the proof. $\square$


Comments (3)

Comment #8244 by Sasha on

Maybe it is worth adding a couple of words justifying the base change property in the proof of Lemma https://stacks.math.columbia.edu/tag/03GY.


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 03GX. Beware of the difference between the letter 'O' and the digit '0'.