Let $X,Y$ surfaces (so $2$-dimensional proper $k$-schemes) which are regular (so the stalks are regular) and birational and denote by $f: X \dashrightarrow Y$ the corresponding rational birational morphism between $X$ and $Y$.

My aim is to show that for regular surfaces the dimensions of cohomology groups are birational invariants; therefore $$\dim_k H^i(X,O_X)= \dim_k H^i(Y,O_Y)$$ for $i=0,1$.

In order to do it I tried following consideration:

It is well know that if $b: B:=Bl_z(Z) \to Z$ is the blowing up of regular surface $Z$ at $z \in Z$, then the dimensions of cohomology groups are conserved; i.e. invariant.

If $f$ would be a "classical" morphism (so not only rational) then according to a factorization theorem (which one I don't know more; does anybody know it's name?) then $f$ factorize into a finite sequence of successively blowing ups

$$f: X= Y_n \to Y_{n-1} \to \dots \to Y_0=Y$$

where $Y_{k+1}= Bl_{Z_i}(Y_k)$ is the blow up of the previous one.

The problem here is that our $f$ is just a rational map.

My idea was to try to construct following diagram (D)

enter image description here

where $\Gamma := \overline{\Gamma_f} \subset X \times Y$ where $\Gamma_f$ also rational defined graph morphism via the pull back on $\operatorname{id}_Y:Y \to Y$ along $f$. This induces rational projection $\widetilde{pr_X}: \Gamma_f \dashrightarrow X$.

My goal would be that after taking closure $\Gamma := \overline{\Gamma_f} \subset X \times Y$ the rational morphism $\widetilde{pr_X}$ would induce "classical" map $pr_X:\Gamma \to X$ making the diagram (D) commutative. Then - if $pr_X,pr_Y$ are birational and $\Gamma$ regular - I can apply the factorization theorem to $pr_X,pr_Y$ and ontain the desired result.

And exactly this is the point: Which properties does this closure $\Gamma$ inherit? Stays it regular, proper and birational to $X,Y$? Why?

The problem is that I'm not sure what control over $\Gamma_f$ I have after taking the closure in $X \times Y$.

  • 1
    $\begingroup$ Certainly $\Gamma$ is proper and birational to $X$ and $Y$. There is no reason for it to be regular any more, but you can resolve its singularities and then replace it by the resolution. $\endgroup$
    – Bort
    May 28, 2019 at 9:03
  • $\begingroup$ Just to be clear: regular for you means smooth, right? Because usually, in surface theory, the term regular means with vanishing $q(X)$. $\endgroup$ May 28, 2019 at 9:34
  • $\begingroup$ @FrancescoPolizzi: I assume from context that regular here means that all local rings are regular (not anything to do with irregularity). Since $k$ is not assumed to be algebraically closed, this is a bit weaker than smooth. $\endgroup$
    – Bort
    May 28, 2019 at 10:29
  • $\begingroup$ yes in this thread I'm working with definition of regularity as Bort explained above. will update this in my question $\endgroup$
    – user267839
    May 28, 2019 at 11:11
  • 1
    $\begingroup$ $\Gamma$ is proper because it is a closed subset of the proper variety $X \times Y$. It is birational to $X$ because if $U$ is the open subset where $f$ is defined, the map $U \mapsto \Gamma_f$ given by $x \mapsto (x,f(x))$ is an isomorphism. $\endgroup$
    – Bort
    May 28, 2019 at 12:41

1 Answer 1


This is true pretty generally. We recall that a quasi-compact scheme $Y$ of finite Krull dimension has pseudo-rational singularities if

  1. $Y$ is an excellent normal Cohen–Macaulay scheme that admits a dualizing complex; and
  2. for every normal scheme $X$ and every projective birational morphism $f\colon X \to Y$, the trace map $f_*\omega_X \to \omega_Y$ is an isomorphism.

We then have the following:

Proposition. Let $k$ be field, and let $X$ and $Y$ be two proper $k$-schemes of pure dimension $d$ with pseudo-rational singularities. Suppose $X$ and $Y$ are birational over $k$. Then, $$\dim_k H^i(X,\mathcal{O}_X) = \dim_k H^i(Y,\mathcal{O}_Y)$$ for all $i$.

Proof. Let $f\colon X \dashrightarrow Y$ be the given birational map over $k$, and let $U$ be the subset of $X$ inducing a dense open embedding $f\rvert_U\colon U \hookrightarrow Y$. Denote the image of $f\rvert_U$ by $V$. We have two projection morphisms $$X \overset{\mathrm{pr}_1}{\longleftarrow} X \times_k Y \overset{\mathrm{pr}_2}{\longrightarrow} Y.$$ By the universal property of the fiber product, $f\rvert_U$ induces an open embedding $U \hookrightarrow X \times_k Y$. We denote by $\Gamma_f$ the closure of $U$ under this open embedding to $X \times_k Y$; note that $\Gamma_f$ is proper over $k$ since it is a closed subscheme of the proper scheme $X \times_k Y$. The projection morphisms $\mathrm{pr}_1\colon \Gamma_f \to X$ and $\mathrm{pr}_2\colon \Gamma_f \to Y$ are proper and birational, since they are morphisms of proper schemes over $k$, and since for $i \in \{1,2\}$, the morphism $\mathrm{pr}_i$ induces an isomorphism from the image of $U$ in $\Gamma_f$ onto $U$ and $V$, respectively.

We now consider a projective Macaulayfication [Kov, Cor. 4.5] $g\colon \widetilde{\Gamma}_f \to \Gamma_f$ of $\Gamma_f$, i.e., $g$ is a projective birational morphism and $\widetilde{\Gamma}_f$ is a Cohen–Macaulay projective $k$-scheme. Now [Kov, Thm. 8.6] implies the canonical morphisms $$\mathcal{O}_X \longrightarrow \mathbf{R}(\mathrm{pr}_1 \circ g)_* \mathcal{O}_{\widetilde{\Gamma}_f} \qquad\text{and}\qquad \mathcal{O}_Y \longrightarrow \mathbf{R}(\mathrm{pr}_2 \circ g)_* \mathcal{O}_{\widetilde{\Gamma}_f}\tag{1}\label{eq:qis}$$ are quasi-isomorphisms, which induce isomorphisms $$H^i(X,\mathcal{O}_X) \overset{\sim}{\longrightarrow} H^i(\widetilde{\Gamma}_f,\mathcal{O}_{\widetilde{\Gamma}_f}) \overset{\sim}{\longleftarrow} H^i(Y,\mathcal{O}_Y)$$ as required. $\blacksquare$

Note that regular schemes have pseudo-rational singularities [Kov, Lem. 7.7], hence the proposition above implies the special case you are interested in. One can also avoid using [Kov] when both $X$ and $Y$ are regular $k$-schemes of pure dimension $2$ as follows: Fix notation as in the first paragraph of the proof above. Instead of constructing $g \colon \widetilde{\Gamma}_f \to \Gamma_f$ as a projective Macaulayfication, we may define $g$ to be a resolution of singularities, in which case $\widetilde{\Gamma}_f$ is a regular projective $k$-scheme of pure dimension $2$, and $g$ is a projective birational morphism. Now [Stacks, Tag 0C5R] implies $g$ is a composition of blowups at closed points. Moreover, combining [Stacks, Tag 0AGS] and the strategy of [Har77, Prop. V.3.4] implies the quasi-isomorphisms \eqref{eq:qis} hold.


[Har77] Robin Hartshorne. Algebraic geometry. Grad. Texts in Math., Vol. 52. New York-Heidelberg: Springer-Verlag, 1977. DOI: 10.1007/978-1-4757-3849-0. MR: 463157.

[Kov] Sándor J. Kovács. "Rational singularities." May 11, 2018. arXiv:1703.02269v6 [math.AG].

[Stacks] The Stacks project authors. The Stacks project. 2019. https://stacks.math.columbia.edu.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.