Pome

18.32 Partial envelopes on a weighted compact set848269EDF931

848269EDF931

Two constraints act on limits of partial Fubini–Study potentials. The continuous weight is an obstacle on the set where sections are measured, and the selecting potential imposes a multiplier-ideal constraint on the whole manifold. The partial envelope is the largest potential satisfying both. Since an obstacle on a non-pluripolar compact set bounds every competitor globally, the partial envelope lies below a translate of the multiplier-ideal envelope of the selecting potential; conversely, a translate of that envelope lying below the obstacle is a competitor. The partial envelope therefore differs from the multiplier-ideal envelope by a bounded function.

Let $\theta $ be a smooth real $(1,1)$-form on $X$, let $\varphi \in \mathrm{PSH}(X,\theta )$, let $K\subseteq X$ be compact and non-pluripolar, and let $w\in C(K,\mathbb R)$. Recall that theorem 5.37.3 characterizes the order $\preceq _{\mathcal I}$ by Lelong numbers on local modifications. As in section 5.40, an inequality holds quasi-everywhere if it holds outside a pluripolar set. The partial envelope of $\varphi $ with obstacle $w$ on $K$ is

B5A5A1951FC8 18.32.1
\begin{equation} \begin{aligned} E_{\theta ,K}[\varphi ]_{\mathcal I}(w) & =\Bigl (\sup \bigl \{ a\in \mathrm{PSH}(X,\theta ): a\preceq _{\mathcal I}\varphi ,\\ & \hspace{52pt}a\leq w\text{ quasi-everywhere on }K\bigr \} \Bigr )^*. \end{aligned} \end{equation}

Theorem 18.32.2.04BABC8924B6 The function $E=E_{\theta ,K}[\varphi ]_{\mathcal I}(w)$ is $\theta $-plurisubharmonic, and it is the largest member of its defining family. Let $q=P_\theta [\varphi ]_{\mathcal I}$, and let $C$ be a constant as in lemma 5.40.1, so that $\sup _ Xa\leq C$ for every $a\in \mathrm{PSH}(X,\theta )$ with $a\leq w$ quasi-everywhere on $K$. Then

\[ q+\min _ Kw\leq E\leq q+C. \]

In particular, $E\sim q$ and

\[ E\sim _{\mathcal I}\varphi . \]

Proof. By theorem 18.4.2, $q\leq 0$ and $q\sim _{\mathcal I}\varphi $. Hence $q+\min _ Kw$ belongs to the defining family. Lemma 5.40.1 bounds the entire family above by $C$, so its regularized supremum $E$ is $\theta $-plurisubharmonic, and $E\geq q+\min _ Kw$.

For the upper bound, let $a$ be a member of the family. Since $\sup _ Xa\leq C$, proposition 18.4.4, proposition 18.4.6 give

\[ a-C\leq a-\sup _ Xa\leq P_\theta [a]_{\mathcal I} \leq P_\theta [\varphi ]_{\mathcal I}=q. \]

Taking the regularized supremum over $a$ proves $E\leq q+C$.

The two bounds show that $E-q$ is bounded. Hence $E$ and $q$ have the same multiplier ideals, and these are the multiplier ideals of $\varphi $.

It remains to show that $E$ itself satisfies the obstacle condition, which is not automatic after regularization. By Choquet's lemma, $E$ is the regularization of the supremum of a countable subfamily. Replace its members by their successive finite maxima. Finite maxima are again members of the family: the multiplier-ideal order is preserved by theorem 5.37.3, and a finite union of pluripolar exceptional sets is pluripolar. We obtain an increasing sequence $a_ j$ in the family whose limit has regularization $E$. Outside a countable union of pluripolar subsets of $K$, every $a_ j$ is at most $w$, and hence so is $\lim _ ja_ j$. By the local upper-envelope theorem recalled in section 5.40, $E$ differs from $\lim _ ja_ j$ only on a pluripolar set. Thus $E\leq w$ quasi-everywhere on $K$.

We have shown that $E$ belongs to its defining family. It dominates every member by construction, so it is the largest member. In particular, the supremum in (18.32.1) equals $E$ at every point even before regularization. $\square$

Proposition 18.32.3.ABD0D0DB843C The partial envelope depends only on the multiplier-ideal class of $\varphi $, and it is increasing in the obstacle and in $\varphi $ with respect to $\preceq _{\mathcal I}$. For $w,w'\in C(K)$ and $c\in \mathbb R$,

\[ \begin{split} E_{\theta ,K}[\varphi ]_{\mathcal I}(w+c) & =E_{\theta ,K}[\varphi ]_{\mathcal I}(w)+c,\\ \bigl |E_{\theta ,K}[\varphi ]_{\mathcal I}(w) -E_{\theta ,K}[\varphi ]_{\mathcal I}(w')\bigr | & \leq \| w-w'\| _{C(K)}. \end{split} \]

The second line stands for the two corresponding pointwise inequalities, which make sense also at the poles. Let $\theta _0,\theta _1$ be smooth real $(1,1)$-forms on $X$, and let $\varphi _ i\in \mathrm{PSH}(X,\theta _ i)$ and $w_ i\in C(K)$ for $i=0,1$. For $0<t<1$ put $\theta _ t=(1-t)\theta _0+t\theta _1$, and define $\varphi _ t$ and $w_ t$ in the same way. Then

\[ (1-t)E_{\theta _0,K}[\varphi _0]_{\mathcal I}(w_0) +tE_{\theta _1,K}[\varphi _1]_{\mathcal I}(w_1) \leq E_{\theta _ t,K}[\varphi _ t]_{\mathcal I}(w_ t). \]

Finally, for a smooth real function $f$, changing the background and the obstacle gives the identity

\[ E_{\theta +\mathrm{dd}^{\mathrm{c}}f,K}[\varphi -f]_{\mathcal I}(w-f|_ K) =E_{\theta ,K}[\varphi ]_{\mathcal I}(w)-f. \]

Proof. The first two assertions hold because the defining families are equal or nested. Adding a constant $c$ is a bijection between the families for $w$ and $w+c$, which proves the translation rule. Applying monotonicity and translation to $w'-\| w-w'\| _{C(K)}\leq w\leq w'+\| w-w'\| _{C(K)}$ gives the two inequalities in the second line.

For concavity, the left-hand side is $\theta _ t$-plurisubharmonic, and it lies below $w_ t$ outside the union of two pluripolar exceptional sets. By theorem 18.32.2, its two summands have the multiplier-ideal classes of $\varphi _0$ and $\varphi _1$, so the addition and scaling rules of lemma 5.37.4 show that the left-hand side has the multiplier-ideal class of $\varphi _ t$. It is therefore a member of the family defining the right-hand side. Finally, $a\mapsto a-f$ is a bijection between the two families in the last identity, and regularization commutes with subtraction of a continuous function. $\square$

In the next section the partial envelope bounds every limit of partial Fubini–Study potentials. The point is to show that both constraints, the multiplier-ideal order and the obstacle on $K$, survive $L^1$ limits.

Cite this page

Comments

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