ScalingStacks

Subsubsection [04UX]

Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.

Complete original source context ยท Original author HTML

(2.2.5) Let ๐’ณ1\mathscr{X}_{1} and ๐’ณ2\mathscr{X}_{2} be two dโ€‹lโ€‹tdlt-models of XX over ๐’ž\mathscr{C}. We say that ๐’ณ1\mathscr{X}_{1} and ๐’ณ2\mathscr{X}_{2} are crepant birational if there exist a normal proper ๐’ž\mathscr{C}-model ๐’ด\mathscr{Y} of XX and morphisms of ๐’ž\mathscr{C}-models fi:๐’ดโ†’๐’ณif_{i}:\mathscr{Y}\to\mathscr{X}_{i} for i=1,2i=1,2 such that the log pullbacks of (๐’ณ1,s)red(\mathscr{X}_{1,s})_{\mathrm{red}} and (๐’ณ2,s)red(\mathscr{X}_{2,s})_{\mathrm{red}} coincide (see [Ko13, 2.23]). Note that we can always assume that ๐’ด\mathscr{Y} is an sโ€‹nโ€‹csnc-model, by taking a log resolution of (๐’ด,๐’ดs)(\mathscr{Y},\mathscr{Y}_{s}). The following theorem collects two fundamental results from the Minimal Model Program.

Theorem 2.2.6.
  1. (1)

    The CC-scheme XX has a good minimal dโ€‹lโ€‹tdlt-model if and only if KXK_{X} is semi-ample over CC.

  2. (2)

    Any two good minimal dโ€‹lโ€‹tdlt-models of XX are crepant birational.

Proof.

(1) The condition that KXK_{X} is semi-ample over CC is obviously necessary, since for every dโ€‹lโ€‹tdlt-model ๐’ณ\mathscr{X} of XX, the divisor KXK_{X} is โ„š\mathbb{Q}-linearly equivalent to the restriction of K๐’ณ+(๐’ณs)redK_{\mathscr{X}}+(\mathscr{X}_{s})_{\mathrm{red}} to XX. Conversely, assume that KXK_{X} is semi-ample over CC, and let ๐’ด\mathscr{Y} be a proper sโ€‹nโ€‹csnc-model of XX. Then applying [HX13, 2.12] to the dโ€‹lโ€‹tdlt-pair (๐’ด,(๐’ดs)red)(\mathscr{Y},(\mathscr{Y}_{s})_{\mathrm{red}}), we see that XX has a good minimal dโ€‹lโ€‹tdlt-model. Condition (1) of [HX13, 2.12] follows from our assumption, and condition (2) follows from the following observation. Let mm be a positive integer such that mโก(K๐’ด+(๐’ดs)red)m(K_{\mathscr{Y}}+(\mathscr{Y}_{s})_{\mathrm{red}}) is Cartier. Over a sufficiently small open neighbourhood of ss in ๐’ž\mathscr{C}, we have an isomorphism of ๐’ช๐’ž\mathcal{O}_{\mathscr{C}}-algebras

Rโก(๐’ด/๐’ž,mโก(K๐’ด+(๐’ดs)red))โ‰…Rโก(๐’ด/๐’ž,mโก(K๐’ด+(๐’ดs)red)โˆ’๐’ดs),R(\mathscr{Y}/\mathscr{C},m(K_{\mathscr{Y}}+(\mathscr{Y}_{s})_{\mathrm{red}}))\cong R(\mathscr{Y}/\mathscr{C},m(K_{\mathscr{Y}}+(\mathscr{Y}_{s})_{\mathrm{red}})-\mathscr{Y}_{s}),

where Rโก(๐’ด/๐’ž,L):=โจjโ‰ฅ0ฯ€โˆ—โ€‹(๐’ช๐’ดโ€‹(jโ€‹L))R(\mathscr{Y}/\mathscr{C},L):=\bigoplus_{j\geq 0}\pi_{*}(\mathcal{O}_{\mathscr{Y}}(jL)) with ฯ€:๐’ดโ†’๐’ž\pi:\mathscr{Y}\to\mathscr{C} the structural morphism. Thus it suffices to show that

๐’œ=Rโก(๐’ด/๐’ž,mโก(K๐’ด+(๐’ดs)red)โˆ’๐’ดs)\mathcal{A}=R(\mathscr{Y}/\mathscr{C},m(K_{\mathscr{Y}}+(\mathscr{Y}_{s})_{\mathrm{red}})-\mathscr{Y}_{s})

is a finitely generated ๐’ช๐’ž\mathcal{O}_{\mathscr{C}}-algebra. If we denote by MM the maximum of the multiplicities of the components in ๐’ดs\mathscr{Y}_{s} then we may assume that m>Mm>M, so that

(๐’ด,(๐’ดs)redโˆ’1mโ€‹๐’ดs)(\mathscr{Y},(\mathscr{Y}_{s})_{\mathrm{red}}-\frac{1}{m}\mathscr{Y}_{s})

is a kโ€‹lโ€‹tklt pair. Hence, the finite generation of ๐’œ\mathcal{A} follows from [BCHM10].

(2) It is already observed in Definition 15 of [dFKX12] that this follows from the proof of [KM98, 3.52]. โˆŽ

Original mathematics by the credited authors. Source-backed reader collection; mathematical self-containment is not assessed.