Pome

13.10 Bedford–Taylor theory7CD616A8B803

7CD616A8B803 Let $X$ be a complex manifold and $\varphi _1,\ldots ,\varphi _ p\in \mathrm{PSH}(X)$ ($p\in \mathbb {N}$) be locally bounded plurisubharmonic functions on $X$1. In this case, there is a canonical definition of the Monge–Ampère type product (13.1.1).

Definition 13.10.1.F6CEAE1224B7 We define the closed positive $(p,p)$-current (13.1.1) on $X$ as follows: We proceed by induction on $p\geq 0$. When $p=0$, we define (13.1.1) as the $(0,0)$-current $[X]$. When $p>0$, we let

\[ \mathrm{dd}^{\mathrm{c}}\varphi _1\wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}\varphi _ p\coloneqq \mathrm{dd}^{\mathrm{c}}\left( \varphi _1\, \mathrm{dd}^{\mathrm{c}}\varphi _2 \wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}\varphi _ p \right). \]

We call this product the Bedford–Taylor product .

Remark 13.10.2.918938F52375 There is also a slightly more general version of this construction. Given a closed positive current $T$, one can also define the product

\[ \mathrm{dd}^{\mathrm{c}}\varphi _1\wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}\varphi _ p \wedge T \]

in a very similar way.

Proposition 13.10.3.E1938BA48A2F The product $\mathrm{dd}^{\mathrm{c}}\varphi _1\wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}\varphi _ p$ is a closed positive $(p,p)$-current on $X$. Moreover, the product is symmetric in the $\varphi _ i$'s.

Proof. The recursive construction here is the same as the one in the positive Bedford--Taylor recursion. That result proves that each step is a positive current of order zero, so multiplication by the next locally bounded potential is defined; applying $\mathrm{dd}^{\mathrm c}$ makes the next product closed. Symmetry is proved, together with monotone continuity, in the Bedford--Taylor continuity theorem. Its proof uses the common-collar integration-by-parts identity and the Chern--Levine--Nirenberg bound, not the incorrect claim that $\mathrm du\wedge\mathrm d^{\mathrm c}u\wedge T$ is closed. Since all constructions agree in overlapping charts, they define the stated global current.

See [Proposition 3.3, Corollary 3.12, GZ17]. The proof relies crucially on an important estimate, known as the Chern–Levine–Nirenberg inequality. See [CLN69] or [Theorem 3.9, GZ17].

The Bedford–Taylor theory has many satisfactory properties.

Theorem 13.10.4.032A3F745585 Let $(\varphi _ i^ j)_{j\in \mathbb {Z}_{>0}}$ be decreasing sequences (resp. increasing sequences) of locally bounded psh functions on $X$ converging (resp. converging a.e.) to locally bounded psh functions $\varphi _ i$'s, where $i=0,\ldots ,p$. Then

\[ \varphi _0^ j\, \mathrm{dd}^{\mathrm{c}}\varphi _1^ j\wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}\varphi _ p^ j\rightharpoonup \varphi _0\, \mathrm{dd}^{\mathrm{c}}\varphi _1\wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}\varphi _ p \]

as $j\to \infty $. In particular, if $(\varphi _0^ j)_ j$ is the constant sequence $1$, we have

\[ \mathrm{dd}^{\mathrm{c}}\varphi _1^ j\wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}\varphi _ p^ j\rightharpoonup \mathrm{dd}^{\mathrm{c}}\varphi _1\wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}\varphi _ p. \]

Proof. On a relatively compact coordinate ball, the decreasing case, including the factor $\varphi_0^j$, is weighted Bedford--Taylor decreasing continuity. The increasing case is weighted increasing continuity, which allows almost-everywhere convergence to the specified locally bounded representatives. The local bounds needed in these results hold on a slightly larger compact set: in the decreasing case they are bounded below by the limit and above by the first term; in the increasing case they are bounded below by the first term and above by the limit, using almost-everywhere order and uniqueness of plurisubharmonic representatives.

For a compactly supported test form on $X$, use a finite partition of unity subordinate to such balls and sum the local limits. This proves weak convergence of the currents globally. Taking the auxiliary factor constantly equal to $1$ gives the unweighted assertion.

Here the notation $\rightharpoonup $ denotes the weak-* convergence of currents.

We refer to [Theorem 3.18, Theorem 3.23, GZ17] for the proofs.

By contrast, we emphasize that the Bedford–Taylor product is not continuous with respect to the $L^1_{\mathrm{loc}}$-convergence in general. A simple example can be found in [Example 3.25, GZ17].

Lemma 13.10.5.10BB181E99E0 Let $Q\subseteq \mathbb {C}^ n$ and $D\subseteq \mathbb {C}^ m$ be domains, let $p_2\colon Q\times D\to D$ be the second projection, and let $\eta $ be a smooth real $(m,m)$-form on $D$. Let $u_1,\ldots ,u_ n$ be locally bounded psh functions on $Q\times D$, and put

\[ M=\mathrm{dd}^{\mathrm{c}}u_1\wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}u_ n,\qquad M_ z=\mathrm{dd}^{\mathrm{c}}_ x(u_1)_ z\wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}_ x(u_ n)_ z. \]

Here $(u_ r)_ z(x)=u_ r(x,z)$. For every bounded Borel function $h$ with compact support in $Q\times D$,

E7F106829576 13.10.6
\begin{equation} \int _{Q\times D}h\, p_2^*\eta \wedge M =\int _ D\left(\int _ Q h_ z\, M_ z\right)\eta (z), \qquad h_ z(x)=h(x,z). \end{equation}

Proof. By linearity, it suffices to consider $\eta \geq 0$. We first prove (13.10.6) when $h$ is continuous. On a product neighborhood of $\operatorname{Supp}h$, choose smooth psh functions $u_ r^{(l)}\searrow u_ r$, $1\leq r\leq n$. For the smooth approximants, decompose each $\mathrm{dd}^{\mathrm{c}}u_ r^{(l)}$ according to the $x$- and $z$-directions. After wedging with $p_2^*\eta $, every term containing a factor $\mathrm{d}z_\alpha $ or $\mathrm{d}\bar z_\beta $ vanishes; hence only the factors $\mathrm{dd}^{\mathrm{c}}_ x(u_ r^{(l)})_ z$ remain, and (13.10.6) follows from the usual Fubini theorem. As $l\to \infty $, the Bedford–Taylor monotone continuity theorem theorem 13.10.4 gives convergence on $Q\times D$ and on every fibre. Fix $K\Subset Q$ and $L\Subset D$ containing the projections of $\operatorname{Supp}h$. The Chern–Levine–Nirenberg estimate [Theorem 3.9, GZ17], together with the local uniform bounds for $u_ r^{(l)}$, bounds the masses of the fibre products on $K$ uniformly for $z\in L$ and $l$. Since $h_ z=0$ for $z\notin L$, dominated convergence proves (13.10.6).

It remains to pass from continuous to bounded Borel functions. Fix relatively compact open sets $U\Subset Q$ and $V\Subset D$. If $\chi \in C_ c(U)$, then

\[ F_\chi (z)\coloneqq \int _ U\chi \, M_ z, \qquad z\in V, \]

is Borel measurable. Indeed, choose the smooth decreasing approximations on a neighborhood of $\overline U\times \overline V$ and set

\[ M_ z^{(l)}\coloneqq \mathrm{dd}^{\mathrm{c}}_ x(u_1^{(l)})_ z\wedge \cdots \wedge \mathrm{dd}^{\mathrm{c}}_ x(u_ n^{(l)})_ z. \]

The functions $z\mapsto \int _ U\chi \, M_ z^{(l)}$ are continuous on $V$ and converge pointwise to $F_\chi $ by theorem 13.10.4.

Let $\mathcal{A}_{U,V}$ be the vector space of bounded Borel functions $\chi $ on $U$ for which $F_\chi $ is Borel measurable on $V$. It is closed under bounded increasing pointwise limits of non-negative functions by the monotone convergence theorem. Taking compactly supported continuous cut-off functions increasing to $1$ on $U$ shows that $1\in \mathcal{A}_{U,V}$. Thus $\mathcal{A}_{U,V}$ contains the algebra $\mathbb {R}+C_ c(U)$, which generates the Borel $\sigma $-algebra of $U$. The functional monotone class theorem therefore shows that $\mathcal{A}_{U,V}$ contains every bounded Borel function on $U$.

Next, let $\mathcal{B}_{U,V}$ be the vector space of bounded Borel functions $g$ on $U\times V$ such that

\[ z\longmapsto \int _ U g_ z\, M_ z, \qquad g_ z(x)=g(x,z), \]

is Borel measurable on $V$. This space is again closed under bounded increasing pointwise limits of non-negative functions. It contains every product $g(x,z)=\chi (x)\psi (z)$ of bounded Borel functions, by the preceding paragraph. Finite sums of such products form an algebra containing the indicators of all Borel rectangles $A\times B\subseteq U\times V$; hence this algebra generates the Borel $\sigma $-algebra of $U\times V$. A second application of the functional monotone class theorem shows that $\mathcal{B}_{U,V}$ contains every bounded Borel function on $U\times V$.

To pass from $U\times V$ to $Q\times D$, choose exhaustions $U_ j\Subset Q$ and $V_ j\Subset D$ by relatively compact open sets. For a Borel set $E\subseteq Q\times D$, the preceding paragraph, applied to the restriction of $\mathds {1}_ E$ to $U_ j\times V_ j$, shows that

\[ G_ j(z)\coloneqq \mathds {1}_{V_ j}(z)\int _{U_ j}\mathds {1}_ E(x,z)\, M_ z \]

is Borel measurable on $D$. We may take both exhaustions increasing, so $G_ j\nearrow M_ z(E_ z)$ pointwise, where

\[ E_ z\coloneqq \{ x\in Q:(x,z)\in E\} . \]

Thus $z\mapsto M_ z(E_ z)$ is Borel measurable on $D$. We may therefore define

\[ \lambda (E)\coloneqq \int _ D M_ z(E_ z)\, \eta (z). \]

The monotone convergence theorem, first on each fibre and then on $D$, shows that $\lambda $ is a Borel measure on $Q\times D$. The local uniform mass bound above shows that $\lambda $ is locally finite and hence Radon. Approximation by non-negative simple functions gives

\[ \int _{Q\times D}h\, \mathrm{d}\lambda =\int _ D\left(\int _ Qh_ z\, M_ z\right)\eta (z) \]

for every non-negative Borel function $h$. For continuous compactly supported $h$, the first part of the proof shows that $\lambda $ agrees with $p_2^*\eta \wedge M$. The uniqueness of Radon measures then gives the same equality for every bounded Borel function $h$ with compact support. $\square$

Cite this page

Comments

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