1
$\begingroup$

The proof of Lemma 5-1-5 in this 1987 paper by Kawamata, Matsuda and Matsuki (link on Projecteuclid) seems to say that a variety with terminal singularities is $\mathbb{Q}$-factorial ( I only need the case $ \Delta =0$ in which case, weak log terminal might be the same as just terminal.). Does anyone know how to prove this, or a place where this is proved?

$\endgroup$
2
  • 3
    $\begingroup$ In dimension two all klt singularities are $\mathbb{Q}$-factorial. In dimension three there are singularities which are terminal and not $\mathbb{Q}$-factorial, for instance the cone over $\mathbb{P}^1\times\mathbb{P}^1$. The Lemma you cite only talks about fibrations, I don't know how it would imply such statement. In any case, if you have a klt singularity, you can make it $\mathbb{Q}$-factorial with a small morphism, this is called a small $\mathbb{Q}$-factorialization. $\endgroup$ Feb 5, 2020 at 21:34
  • 1
    $\begingroup$ I believe the paper assumes at the very beginning of section 5-1 that $X$ is $\mathbb Q$-factorial? As @JoaquínMoraga says, this is not true otherwise. $\endgroup$ Feb 6, 2020 at 3:39

1 Answer 1

1
$\begingroup$

I do not know how you get this conclusion from the paper of Kawamata, Matsuda and Matsuki. But this is not true in dimension at least $3$. Of course in dimension 2 every terminal singularity is smooth.

The easiest example of a non-$\mathbb{Q}$-factorial terminal singularity in dimension $3$ is the hypersurface $(xy=zw)\subset \mathbb{A}^4_{x,y,z,w}$. It is terminal as it admits a crepant small resolution. On the other hand, it is not $\mathbb{Q}$-factorial, as there are $2$ Weil divisors $D=(x=z=0)$ and $D'=(y=w=0)$ such that the intersection $D\cap D'$ is exactly one point, the origin. This contradicts the fact that the intersection of $2$ $\mathbb{Q}$-Cartier divisor has codimension at least $2$.

$\endgroup$

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.