Let $X$ be a normal variety with only one singularity at $x$ and $(X,x)$ is a canonical singularity i.e. $(X.x)$ satisfies $(i)$ and $(ii)$.
$(i)$ $(X,x)$ is a $\mathbb{Q}$ Goreinstein singularity.
$(ii)$ For any resolution $f:Y \rightarrow X$, we can write $K_Y=f^*K_X+\sum_{i=1}^r m_iE_i$ with $m_i\geqq 0$

How can I prove that for any resolution $f:Y\rightarrow X$ and $m\in > \mathbb{Z}_{\geqq 0}$, $f_*\mathcal{O}(mK_Y)\cong \mathcal{O}(mK_X)$?

I don't know even what is a morphism.

Motivation. I want to show the unique ness of "canonical resolution"
A resolution of $f:Y\rightarrow X$ is called canonical resolution if
(i) $Y$ has at worst canonical singularity.
(ii) $K_Y$ is $f$ ample.
My strategy is this.
Let $f_i:Y_i\rightarrow X$ be two canonical resolution.
Make two common resolution $g_1:X\rightarrow Y_1$ and $g_2:X\rightarrow Y_2$.
From (ii), we get $Y_i\cong$ Proj($\oplus_{m\geqq 0}{f_i}*\mathcal{O}(mK_{Y_i})$).
So if we can show ${g_i}_*\mathcal{O}(mK_{X})\cong \mathcal{O}(mK_{Y_i})$, we can show $Y_1\cong Y_2\cong $Proj($\oplus_{m\geqq 0}{g_1}_*{f_1}_*\mathcal{O}(mK_{Y_i})$)

  • $\begingroup$ Probably, you intended to write $K_Y=f^*K_X+\sum_{i=1}^r m_iE_i$ and the $a_i$ are $m_i$? $\endgroup$
    – user267839
    Apr 23 at 22:00
  • $\begingroup$ Do you really mean for all $m $ or big enough? Since $K_X$ is $\mathbb{Q}$-Cartier it's multiple for big enough $m$ becomes Cartier, so invertible. But as $X$ is normal, you can use Hartog's argument on the base as $x$ has codim $\ge 2$ (if your $X$ is not a curve) $\endgroup$
    – user267839
    Apr 23 at 22:11
  • $\begingroup$ @user267839 Thanks. You are right. $\endgroup$
    – George
    Apr 23 at 22:52
  • $\begingroup$ @user267839 I mean for all non negative m. Could you see my motivation added in the question. $\endgroup$
    – George
    Apr 23 at 22:53
  • $\begingroup$ @user267839 Hartog's argument might be used for my porpose too. Thanks. $\endgroup$
    – George
    Apr 23 at 22:55

1 Answer 1


This essentially follows from Claim 2.3.1, p. 39 of

J. Kollár, S. Kovács: Singularities of the Minimal Model Program, Cambridge Tracts in Mathematics 200 (2013).

The Authors prove the result for $m=1$, but the same proof works without modifications for every $m \geq 1$. Compare also with Claim 2.3.2 (same page), Proposition 2.17 (p. 48) and Proposition 2.18 (p. 49).

The last result, applied with $S=X, \, p_X=\text{id}_X$, $\Delta=\Delta_Y=\emptyset$, gives precisely the statement you need.


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.