2
$\begingroup$

Let $(X,B)$ be a projective klt pair, H is an $\mathbb{R}$-divisor s.t. $K_X+B+H$ is nef. Suppose that we can run $(K_X+B)$-MMP with scaling $H$(that is, the flip exists) and denote $\alpha_i:X \dashrightarrow X_i$ to be the map to the i-step of the MMP. Then my question is how can I show that $\alpha_i\neq\alpha_j$ for any $i\neq j$ ?(That is, the induced birational map $\alpha_i\cdot\alpha_j^{-1}:X_j\dashrightarrow X_i$ is not an isomorphism.

Of course we can assume each step is the flip. Then from the construction of MMP with scaling we obtain a sequence of real numbers $1\geq t_0\geq t_1\geq t_2\geq...\geq 0$ . Can we prove the strict inequality : $t_i> t_{i+1}$?

Thanks for your comments!

$\endgroup$
3
  • 1
    $\begingroup$ Welcome new contributor. At each stage in the Mori program, either the dimension strictly decreases, the Picard rank strictly decreases, or the $K_X+B+H$-negative cone of curves strictly decreases (it loses an extremal ray). $\endgroup$ Commented Dec 7, 2024 at 12:24
  • $\begingroup$ @JasonStarr Thanks for comments professor Starr, but the point is that if each step is a flip the picard number does not change. Now I've understanded this problem and will write an answer. $\endgroup$ Commented Dec 7, 2024 at 13:08
  • $\begingroup$ In the case of flips, a negative external ray “flips” to positive. So the negative cone becomes strictly smaller. Proving existence and termination of flips is hard. Proving that the steps in a Mori program do not cycle to an earlier stage is automatic. $\endgroup$ Commented Dec 7, 2024 at 13:34

1 Answer 1

0
$\begingroup$

I realized how to solve this problem now. As pointed above, we may assume each step of MMP is a flip. For $i<j$, We can choose an exceptional divisor $E$ over $X_i$ such that its center lies in the flipping locus, then by the negativity lemma we have that the discrepancy $a(X_i,B_i,E)<a(X_{i+1},B_{i+1},E)$. By the negativity of MMP we know that $a(X_{i+1},B_{i+1},E)\leq a(X_{j},B_{j},E)$. So we have $a(X_i,B_i,E)<a(X_{j},B_{j},E)$ and hence $X_i$ and $X_i$ cannot be isomorphism.

But it seems like that we do not really need MMP with scaling.

$\endgroup$

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.