Pome

18.4 Ideal projection3458DC26AA54

3458DC26AA54

We return to the general setting fixed at the beginning of the chapter. The multiplier-ideal envelope is the greatest nonpositive potential in a multiplier-ideal class. Taking a supremum over the class could a priori enlarge the multiplier ideals; the increasing-limit form of strong openness rules this out, and the envelope is then a projection onto the class.

Definition 18.4.1.F1F202AEE8FD For $u\in \mathrm{PSH}(X,\theta )$ define

\[ P_\theta [u]_{\mathcal I} :=\left(\sup \{ v\in \mathrm{PSH}(X,\theta ):v\leq 0, \ v\sim _{\mathcal I}u\} \right)^*. \]

As for ordinary envelopes, the members of the defining family are called competitors. The family is nonempty because it contains $u-\sup _ Xu$; bounded perturbations do not change multiplier ideals. Since it is bounded above by zero, its regularized supremum is a $\theta $-plurisubharmonic function. The envelope depends only on the multiplier-ideal class of $u$.

Theorem 18.4.2.7F20FD198864 For every $u\in \mathrm{PSH}(X,\theta )$,

\[ P_\theta [u]_{\mathcal I}\sim _{\mathcal I}u,\qquad P_\theta [P_\theta [u]_{\mathcal I}]_{\mathcal I}=P_\theta [u]_{\mathcal I}. \]

The regularization in definition 18.4.1 does not change the pointwise supremum. Moreover,

\[ \sup _ X P_\theta [u]_{\mathcal I}=0. \]

Proof. Write $q=P_\theta [u]_{\mathcal I}$. By Choquet's lemma, some sequence from the defining family has regularized supremum $q$. Replacing its terms by successive finite maxima, we obtain by lemma 5.37.4 an increasing sequence $v_ j$ that remains in the family and converges to $q$ almost everywhere.

Fix $x\in X$ and $\lambda >0$. On a coordinate ball about $x$, choose one smooth function $r$ such that every $v_ j+r$ and $q+r$ is plurisubharmonic; this is possible because all these functions have complex Hessian bounded below by the same smooth form $-\theta $. The increasing-limit assertion of theorem 5.37.3 gives

\[ \mathcal I(\lambda (q+r))_ x =\mathcal I(\lambda (v_ j+r))_ x =\mathcal I(\lambda u)_ x \]

for all large $j$; the second equality holds because smooth additions do not change multiplier ideals. Hence $q\sim _{\mathcal I}u$.

Since $q\leq 0$, it belongs to its own defining family. The pointwise supremum is therefore at least $q$, and it is at most its regularization $q$; this proves the assertion about regularization. The defining families for $P_\theta [q]_{\mathcal I}$ and $P_\theta [u]_{\mathcal I}$ coincide, which proves idempotence. Finally, if $\sup _ Xq<0$, then $q-\sup _ Xq$ would be a larger competitor, which is impossible. $\square$

Ideal projection commutes with pullback by a modification because both the potential class and the all-weight ideal class descend.

Proposition 18.4.3.9AE7825754C3 Let $\mu :Y\to X$ be a modification of connected compact complex manifolds. For every smooth real $(1,1)$-form $\theta $ on $X$ and every $u\in \mathrm{PSH}(X,\theta )$,

\[ P_{\mu ^*\theta }[u\circ \mu ]_{\mathcal I} =P_\theta [u]_{\mathcal I}\circ \mu . \]

Proof. By proposition 5.43.18, pullback identifies the potentials in the two backgrounds. Surjectivity of $\mu $ identifies their nonpositivity constraints, and lemma 5.43.24 identifies their all-weight multiplier-ideal classes. Hence pullback gives a bijection of the complete competitor families. The ideal envelope is itself a competitor and is their greatest member by theorem 18.4.2. Its pullback is therefore the greatest competitor upstairs, which proves the identity. $\square$

Proposition 18.4.4.2506B7AD6F19 For every $u\in \mathrm{PSH}(X,\theta )$,

\[ u-\sup _ Xu\leq P_\theta [u]\leq P_\theta [u]_{\mathcal I}\leq V_\theta . \]

Moreover,

\[ \begin{gathered} P_\theta [u]\sim _{\mathcal I}u,\qquad P_\theta [P_\theta [u]]_{\mathcal I}=P_\theta [u]_{\mathcal I},\\ P_\theta [P_\theta [u]_{\mathcal I}]=P_\theta [u]_{\mathcal I}. \end{gathered} \]

Proof. Let $h\leq 0$ be a competitor for the ordinary envelope, so that $h\preceq u$. The function $\max (h,u-\sup _ Xu)$ is nonpositive, ordinarily equivalent to $u$, and dominates $h$. Since ordinarily equivalent potentials have the same multiplier ideals, it is a competitor for the ideal envelope. Taking regularized suprema gives $P_\theta [u]\leq P_\theta [u]_{\mathcal I}$; the two outer inequalities follow directly from the defining families.

Projection and monotonicity of local integrability now give, for every $\lambda >0$,

\[ \mathcal I(\lambda u) \subseteq \mathcal I(\lambda P_\theta [u]) \subseteq \mathcal I(\lambda P_\theta [u]_{\mathcal I}) =\mathcal I(\lambda u). \]

Thus the ordinary envelope preserves all multiplier ideals, and it has the same ideal envelope as $u$. For $q=P_\theta [u]_{\mathcal I}$, the inequalities already proved and idempotence give $q\leq P_\theta [q]\leq P_\theta [q]_{\mathcal I}=q$. $\square$

We call $u\in \mathrm{PSH}(X,\theta )$ normalized if $\sup _ Xu=0$. Recall from definition 5.40.4 that $u$ is an ordinary model, or envelope fixed point, if $P_\theta [u]=u$; we call $u$ an ideal model if $P_\theta [u]_{\mathcal I}=u$. Models of either kind are normalized, and by the last identity in proposition 18.4.4 every ideal model is an ordinary model. The map $u\mapsto P_\theta [u]_{\mathcal I}$ is called ideal projection.

Corollary 18.4.5.B853F2E6C70B If $\theta $ is semipositive and big, then $P_\theta [u]_{\mathcal I}$ has finite extended Monge–Ampère mass for every $u\in \mathrm{PSH}(X,\theta )$.

Proof. By the last identity in corollary 18.4.4, $P_\theta [u]_{\mathcal I}$ is an ordinary model, that is, an envelope fixed point. The mass estimate for envelope fixed points in proposition 16.28.7 applies. $\square$

Proposition 18.4.6.EAFF378DAF48 For $u,v\in \mathrm{PSH}(X,\theta )$,

\[ u\preceq _{\mathcal I}v\quad \Longrightarrow \quad P_\theta [u]_{\mathcal I}\leq P_\theta [v]_{\mathcal I}. \]

For $a>0$ one has $P_{a\theta }[a u]_{\mathcal I}=aP_\theta [u]_{\mathcal I}$. For smooth real $(1,1)$-forms $\theta _ i$ and potentials $u_ i\in \mathrm{PSH}(X,\theta _ i)$, $i=1,2$,

\[ P_{\theta _1}[u_1]_{\mathcal I}+P_{\theta _2}[u_2]_{\mathcal I} \leq P_{\theta _1+\theta _2}[u_1+u_2]_{\mathcal I}. \]

Proof. Let $h$ be a competitor for $P_\theta [u]_{\mathcal I}$. Since $h\sim _{\mathcal I}u\preceq _{\mathcal I}v$, the Lelong-number criterion in theorem 5.37.3 shows that $\max (h,v-\sup _ Xv)$ is multiplier-ideal equivalent to $v$. This function is nonpositive and dominates $h$, so it is a competitor for $P_\theta [v]_{\mathcal I}$; taking the supremum over $h$ proves that the envelope preserves the order. Multiplication by $a$ is a bijection between the defining families in the scaling identity. For the sum, $P_{\theta _1}[u_1]_{\mathcal I}+P_{\theta _2}[u_2]_{\mathcal I}$ is nonpositive and $(\theta _1+\theta _2)$-plurisubharmonic, and it is multiplier-ideal equivalent to $u_1+u_2$ by projection and lemma 5.37.4. It is therefore a competitor for the envelope on the right. $\square$

18.4.7 Analytic and divisorial singularities89E4AF40A761

For a potential $a$ with analytic singularities, the multiplier-ideal order controls the ordinary order: by lemma 5.38.11, every quasi-plurisubharmonic $v$ with $v\preceq _{\mathcal I}a$ satisfies $v\preceq a$. The ideal envelope of $a$ is then a competitor for its ordinary envelope, and the two envelopes coincide.

Proposition 18.4.8.EB95710FF4D7 Let $a\in \mathrm{PSH}(X,\theta )$ have analytic singularities. Then

\[ a\sim P_\theta [a]=P_\theta [a]_{\mathcal I}. \]

In particular the common envelope has analytic singularities with the same coefficient and ideal as $a$, up to a bounded remainder.

Proof. Set $q=P_\theta [a]_{\mathcal I}$. Projection gives $q\sim _{\mathcal I}a$, so lemma 5.38.11 gives $q\preceq a$. Since $q\leq 0$, it is a competitor for $P_\theta [a]$. Together with proposition 18.4.4, this yields

\[ a-\sup _ Xa\leq P_\theta [a]\leq q\leq P_\theta [a]. \]

Hence the two envelopes are equal and ordinarily equivalent to $a$. Since they differ from $a$ by a bounded function, they inherit its analytic presentation. $\square$

The same argument applies to the potential of a countable divisorial current of finite mass, with lemma 5.38.5 in place of lemma 5.38.11.

Corollary 18.4.9.B5BE9775FBF5 Let $S$ be a closed positive divisorial current of finite mass on a connected compact complex manifold $Z$, and choose a smooth residual decomposition $S=\beta +\mathrm{dd}^{\mathrm{c}}v$ as in corollary 5.38.6. For every Hermitian form $\Omega $ with $v\in \mathrm{PSH}(Z,\Omega )$,

\[ P_\Omega [v]=P_\Omega [v]_{\mathcal I}\sim v. \]

Proof. Set $q=P_\Omega [v]_{\mathcal I}$. Projection gives $q\sim _{\mathcal I}v$, and lemma 5.38.5 gives $q\preceq v$. Thus $q$ is a competitor for $P_\Omega [v]$. Together with proposition 18.4.4, this yields $v-\sup _ Zv\leq P_\Omega [v]=q\preceq v$. $\square$

Cite this page

Comments

Corrections, questions, and mathematical discussion are welcome. Sign in with GitHub to comment. Comments are public.