Pome

16.8 The $P$ and $\mathcal{I}$-preordersEC95C2958023

EC95C2958023

Let $X$ be a connected compact Kähler manifold of dimension $n$.

Recall that we defined a preorder on $\mathrm{QPSH}(X)$ in definition 16.7.1 to compare the singularity types of quasi-plurisubharmonic functions. The problem with this preorder is that it is too fine. For our purposes, it is useful to consider rougher relations.

16.8.1 The definitions of the preordersF7493F03C4C2

Recall that the $P$-envelope is defined in definition 16.3.5.

Definition 16.8.2.4E959BD934C4 For $\varphi ,\psi \in \mathrm{QPSH}(X)$, we say $\varphi $ is $P$-more singular than $\psi $ and write $\varphi \preceq _ P \psi $ if for some (equivalently, every, due to 16.8.4 below) closed smooth real $(1,1)$-form $\theta $ on $X$ such that $\varphi ,\psi \in \mathrm{PSH}(X,\theta )_{>0}$, we have

D3383E5F872D 16.8.3
\begin{equation} P_{\theta }[\varphi ]\leq P_{\theta }[\psi ]. \end{equation}

If $\varphi \preceq _ P \psi $ and $\psi \preceq _ P \varphi $, we write $\varphi \sim _ P \psi $ and say that $\varphi $ and $\psi $ have the same $P$-singularity type.

Recall that $\mathrm{PSH}(X,\theta )_{>0}$ was introduced in (13.13.1). Note that if $\varphi \preceq \psi $, then $\varphi \preceq _ P \psi $. So the $P$-preorder is coarser than $\preceq $.

The $P$-equivalence relation can also be formulated using the language of relative full mass classes developed in 16.5. In fact, $\varphi ,\psi \in \mathrm{PSH}(X,\theta )_{>0}$ satisfies $\varphi \sim _ P \psi $ if and only if $\varphi ,\psi \in \mathcal{E}(X,\theta ;\phi )$ for a common model potential $\phi \in \mathrm{PSH}(X,\theta )_{>0}$.

The condition (16.8.3) is independent of the choice of $\theta $:

Lemma 16.8.4.A3C93A5CDBF4 Let $\varphi ,\psi \in \mathrm{PSH}(X,\theta )_{>0}$. For any Kähler form $\omega $ on $X$, the following are equivalent:

  1. ED4E919B13E0

    $P_{\theta }[\varphi ]\leq P_{\theta }[\psi ]$;

  2. E3965942668C

    $P_{\theta +\omega }[\varphi ]\leq P_{\theta +\omega }[\psi ]$.

In particular, $\preceq _ P$ defines a preorder on $\mathrm{QPSH}(X)$.

Proof. 1 $\implies $ 2. Assume 1. Observe that

\[ \varphi \preceq P_{\theta }[\varphi ]\leq P_{\theta +\omega }[\varphi ]. \]

Taking the $P_{\theta +\omega }$-envelope, we find

\[ P_{\theta +\omega }[\varphi ] \leq P_{\theta +\omega }\Bigl [P_{\theta }[\varphi ]\Bigr ] \leq P_{\theta +\omega }\Bigl [P_{\theta +\omega }[\varphi ]\Bigr ]=P_{\theta +\omega }[\varphi ], \]

where the equality follows from theorem 16.3.15. In particular,

7E95C2DD4B7D 16.8.5
\begin{equation} P_{\theta +\omega }[\varphi ]=P_{\theta +\omega }\Bigl [P_{\theta }[\varphi ]\Bigr ]. \end{equation}

A similar formula holds for $\psi $. So we see that 2 holds.

2 $\implies $ 1. Assume 2. By (16.8.5), we may assume that $\varphi $ and $\psi $ are both model potentials in $\mathrm{PSH}(X,\theta )_{>0}$.

Observe that $\varphi \lor \psi \preceq P_{\theta +\omega }[\psi ]$. It follows that $P_{\theta +\omega }[\varphi \lor \psi ]\leq P_{\theta +\omega }[\psi ]$. The reverse inequality is trivial, so

\[ P_{\theta +\omega }[\varphi \lor \psi ]= P_{\theta +\omega }[\psi ]. \]

From the direction we have proved, for any $C\geq 1$,

\[ P_{\theta +C\omega }[\varphi \lor \psi ]= P_{\theta +C\omega }[\psi ]. \]

So by proposition 16.3.10,

\[ \int _ X \Bigl (\theta +C\omega +\mathrm{dd}^{\mathrm{c}}(\varphi \lor \psi )\Bigr )^ n=\int _ X\left(\theta +C\omega +\mathrm{dd}^{\mathrm{c}}\psi \right)^ n. \]

Since both sides are polynomials in $C$, the equality extends to $C=0$, namely,

\[ \int _ X \theta _{\varphi \lor \psi }^ n=\int _ X \theta _{\psi }^ n. \]

In particular, $\varphi \lor \psi \leq P_{\theta }[\psi ]=\psi $ by (16.3.16). So 1 follows. $\square$

As a consequence of 16.8.4, we can define the $P$-preorder at the level of currents. Given closed positive $(1,1)$-currents $T=\theta _{\varphi }$, $S=\theta '_{\psi }$, we write $T\preceq _ P S$ (resp. $T\sim _ P S$) if $\varphi \preceq _ P \psi $ (resp. $\varphi \sim _ P \psi $). This definition is independent of the decompositions of $T$ and $S$.

As a first example of $P$-equivalence, we have:

Example 16.8.6.14285897F9E7 Let $\theta $ be a closed smooth real $(1,1)$-form on $X$ and $\varphi \in \mathrm{PSH}(X,\theta )_{>0}$. Then

\[ \varphi \sim _ P P_{\theta }[\varphi ]{}^{\href{#sgpt-a0000048053}{1}}. \]

This follows immediately from theorem 16.3.15.

We give a very useful criterion of the $P$-equivalence in terms of the non-pluripolar masses.

Proposition 16.8.7.C7C64ABB4647 Let $\theta $ be a closed smooth real $(1,1)$-form on $X$ representing a big cohomology class. Let $\varphi ,\psi \in \mathrm{PSH}(X,\theta )$ and $\varphi \preceq _ P \psi $. Then the following are equivalent:

  1. B4083090A0B4

    $\varphi \sim _ P \psi $;

  2. 28D1D9DDF8EF

    for each $j=0,\ldots ,n$, we have

    DF3EF73A9B89 16.8.8
    \begin{equation} \int _ X \theta _{\varphi }^ j\wedge \theta _{V_{\theta }}^{n-j}=\int _ X \theta _{\psi }^ j\wedge \theta _{V_{\theta }}^{n-j}. \end{equation}

Assume furthermore that $\varphi ,\psi \in \mathrm{PSH}(X,\theta )_{>0}$, then these conditions are equivalent to the following:

  1. 2A79C290D213

    We have

    \[ \int _ X \theta _{\varphi }^ n=\int _ X \theta _{\psi }^ n. \]
Recall that $V_{\theta }$ is introduced in (13.13.7).

Proof. We first prove the equivalence between 1 and 1 when $\varphi ,\psi \in \mathrm{PSH}(X,\theta )_{>0}$.

1 $\implies $ 1. Assume that $\varphi \sim _ P \psi $. By 16.8.4, we have

\[ P_{\theta }[\varphi ]=P_{\theta }[\psi ]. \]

So 1 follows from proposition 16.3.10.

1 $\implies $ 1. Since $\varphi \preceq _ P\psi $, we have $P_{\theta }[\varphi ]\leq P_{\theta }[\psi ]$. Moreover, proposition 16.3.10 shows that $P_{\theta }[\varphi ]$ and $P_{\theta }[\psi ]$ have the same positive mass. By theorem 16.3.15, $P_{\theta }[\psi ]$ is a candidate in the characterization of $P_{\theta }\bigl [P_{\theta }[\varphi ]\bigr ]=P_{\theta }[\varphi ]$. Hence $P_{\theta }[\psi ]\leq P_{\theta }[\varphi ]$, and therefore $P_{\theta }[\varphi ]=P_{\theta }[\psi ]$. So 1 follows.

Let us come back to the general case with $\varphi ,\psi \in \mathrm{PSH}(X,\theta )$.

1 $\implies $ 2. Fix $j\in \{ 0,\ldots ,n\} $. We argue (16.8.8).

Take a Kähler form $\omega $ on $X$. By 16.8.4, for each $\epsilon >0$, we have

\[ P_{\theta +\epsilon \omega }[\varphi ]= P_{\theta +\epsilon \omega }[\psi ]. \]

It follows from proposition 16.3.10 that

\[ \begin{aligned} \int _ X \left(\theta +\epsilon \omega +\mathrm{dd}^{\mathrm{c}}\psi \right)^ j\wedge \theta _{V_{\theta }}^{n-j}=& \int _ X \Bigl (\theta +\epsilon \omega +\mathrm{dd}^{\mathrm{c}}P_{\theta +\epsilon \omega }[\psi ] \Bigr )^ j\wedge \theta _{V_{\theta }}^{n-j}\\ =& \int _ X \Bigl (\theta +\epsilon \omega +\mathrm{dd}^{\mathrm{c}}P_{\theta +\epsilon \omega }[\varphi ] \Bigr )^ j\wedge \theta _{V_{\theta }}^{n-j} \\ =& \int _ X \left(\theta +\epsilon \omega +\mathrm{dd}^{\mathrm{c}}\varphi \right)^ j\wedge \theta _{V_{\theta }}^{n-j}. \end{aligned} \]

Since the two extremes are both polynomials in $\epsilon $ by the multi-linearity of the non-pluripolar products proposition 13.11.4 6, we conclude that the same holds when $\epsilon =0$, that is, (16.8.8) holds.

2 $\implies $ 1. Assume (16.8.8) holds for all $j=0,\ldots ,n$. Then the same holds with $P_{\theta }[\psi ]$ in place of $\psi $, thanks to proposition 16.3.10. Therefore, for each $t\in (0,1)$, we have

\[ \int _ X \theta _{t\varphi +(1-t)V_{\theta }}^ n= \int _ X \theta _{tP_{\theta }[\psi ]+(1-t)V_{\theta }}^ n \]

by the binomial expansion. Furthermore, $t\varphi +(1-t)V_{\theta }\preceq tP_{\theta }[\psi ]+(1-t)V_{\theta }$. By the implication 1 $\implies $ 1, we have

\[ t\varphi +(1-t)V_{\theta }\sim _ P t P_{\theta }[\psi ]+(1-t)V_{\theta } \]

for each $t\in (0,1)$.

Fix a Kähler form $\omega $ on $X$. From the implication 1 $\implies $ 1, we have

\[ \int _ X (\theta +\omega )_{t\varphi +(1-t)V_{\theta }}^ n= \int _ X (\theta +\omega )_{tP_{\theta }[\psi ]+(1-t)V_{\theta }}^ n. \]

Since both sides are polynomials in $t$ by the multi-linearity of the non-pluripolar products proposition 13.11.4 6, the same holds when $t=1$. From the implication 1 $\implies $ 1 again, we have $\varphi \sim _ P P_{\theta }[\psi ]\sim _ P \psi $. $\square$

Next we introduce a different preorder.

Proposition 16.8.9.EE590FE41862 Given $\varphi ,\psi \in \mathrm{QPSH}(X)$, the following are equivalent:

  1. ABA41C3D6B24

    For any $k\in \mathbb {Z}_{>0}$, we have

    \[ \mathcal{I}(k\varphi )\subseteq \mathcal{I}(k\psi ); \]
  2. C970FD27DA7D

    for any $\lambda \in \mathbb {R}_{>0}$, we have

    \[ \mathcal{I}(\lambda \varphi )\subseteq \mathcal{I}(\lambda \psi ); \]
  3. EE2809B429C8

    for any modification $\pi \colon Y\rightarrow X$ and any $y\in Y$, we have

    \[ \nu (\pi ^*\varphi ,y)\geq \nu (\pi ^*\psi ,y); \]
  4. F03432967BAB

    for any proper bimeromorphic morphism $\pi \colon Y\rightarrow X$ from a Kähler manifold and any $y\in Y$, we have

    \[ \nu (\pi ^*\varphi ,y)\geq \nu (\pi ^*\psi ,y); \]
  5. 7B8BA50C8017

    for any prime divisor $E$ over $X$, we have

    \[ \nu (\varphi ,E)\geq \nu (\psi ,E). \]

Proof. The proof is almost identical to that of proposition 16.6.2. $\square$

Definition 16.8.10.5C5512F45CC2 Let $\varphi ,\psi \in \mathrm{QPSH}(X)$, we say $\varphi $ is $\mathcal{I}$-more singular than $\psi $ and write $\varphi \preceq _{\mathcal{I}} \psi $ if the equivalent conditions in proposition 16.8.9 are satisfied.

It is clear that $\preceq _{\mathcal{I}}$ is a preorder on $\mathrm{QPSH}(X)$.

Note that $\varphi \preceq _{\mathcal{I}}\psi $ and $\psi \preceq _{\mathcal{I}}\varphi $ both hold if and only if $\varphi \sim _{\mathcal{I}}\psi $ in the sense of definition 16.6.3.

Given closed positive $(1,1)$-currents $T=\theta _{\varphi }$, $S=\theta '_{\psi }$, we write $T\preceq _{\mathcal{I}} S$ (resp. $T\sim _{\mathcal{I}} S$) if $\varphi \preceq _{\mathcal{I}} \psi $ (resp. $\varphi \sim _{\mathcal{I}} \psi $). This definition is independent of the decompositions of $T$ and $S$.

Lemma 16.8.11.4A5CF38738F2 Let $\theta $ be a closed smooth real $(1,1)$-form on $X$ representing a big cohomology class. Let $\varphi ,\psi \in \mathrm{PSH}(X,\theta )_{>0}$. Then

295DE82294BA 16.8.12
\begin{equation} P_{\theta }[\varphi \lor \psi ]=P_{\theta }\Bigl [ P_{\theta }[\varphi ]\lor P_{\theta }[\psi ] \Bigr ]. \end{equation}

Proof. Since $\varphi \lor \psi \preceq P_{\theta }[\varphi ]\lor P_{\theta }[\psi ]$, the $\leq $ direction of (16.8.12) follows. Conversely, it suffices to show that

\[ P_{\theta }[\varphi \lor \psi ]\geq P_{\theta }[\varphi ]\lor P_{\theta }[\psi ], \]

which follows from $\varphi \preceq \varphi \lor \psi $, $\psi \preceq \varphi \lor \psi $, and the monotonicity of the $P$-envelope with respect to $\preceq $. $\square$

Lemma 16.8.13.005A077A0167 Let $\varphi ,\psi \in \mathrm{QPSH}(X)$. Then the following are equivalent:

  1. F123EB79E13C

    $\varphi \preceq _ P \psi $ (resp. $\varphi \preceq _{\mathcal{I}} \psi $);

  2. 1D8449F6F20B

    $\varphi \lor \psi \sim _ P\psi $ (resp. $\varphi \lor \psi \sim _{\mathcal{I}}\psi $).

Proof. Take a closed real smooth $(1,1)$-form $\theta $ on $X$ such that $\varphi ,\psi \in \mathrm{PSH}(X,\theta )_{>0}$. We only prove the $P$ case, the $\mathcal{I}$ case is similar.

2 $\implies $ 1. Assume 2. By 2 and example 16.8.6, $P_{\theta }[\varphi \lor \psi ]=P_{\theta }[\psi ]\sim _ P \psi $. But $\varphi \preceq P_{\theta }[\varphi \lor \psi ]$, so 1 follows.

1 $\implies $ 2. Assume 1. We may assume that $\varphi ,\psi $ are both model by 16.8.11. Then $\varphi \leq \psi $ and 2 follows. $\square$

Corollary 16.8.14.FE600747C81E Let $\varphi ,\psi \in \mathrm{QPSH}(X)$. Assume that $\varphi \preceq _ P \psi $. Then $\varphi \preceq _{\mathcal{I}}\psi $.

Therefore, the preorder $\preceq _{\mathcal{I}}$ is coarser than $\preceq _ P$. Given $\varphi ,\psi \in \mathrm{QPSH}(X)$, we therefore have the implications:
\[ \varphi \preceq \psi \implies \varphi \preceq _ P \psi \implies \varphi \preceq _{\mathcal{I}} \psi . \]

Proof. Thanks to 16.8.13, we have $\varphi \lor \psi \sim _ P \psi $. Take a closed smooth real $(1,1)$-form $\theta $ on $X$ such that $\varphi ,\psi \in \mathrm{PSH}(X,\theta )_{>0}$. Then by proposition 16.6.18, we have

\[ \psi \sim _{\mathcal{I}} P_{\theta }[\psi ]= P_{\theta }\left[\varphi \lor \psi \right]\sim _{\mathcal{I}} \varphi \lor \psi . \]

Our assertion follows from the $\mathcal{I}$-version of 16.8.13. $\square$

Next we give a few extra characterizations of the $P$-envelope.

Corollary 16.8.15.564E8A0E68F9 Let $\theta $ be a closed smooth real $(1,1)$-form on $X$ representing a big cohomology class. Assume that $\varphi \in \mathrm{PSH}(X,\theta )_{>0}$. Then

\[ \begin{aligned} P_{\theta }[\varphi ]=& \sup \left\{ \psi \in \mathrm{PSH}(X,\theta ): \psi \leq 0, \psi \sim _ P \varphi \right\} \\ =& \sup \left\{ \psi \in \mathrm{PSH}(X,\theta ): \psi \leq 0, \psi \preceq _ P \varphi \right\} . \end{aligned} \]
Just for comparison, let us recall a few other characterizations of the $P$-envelope for $\varphi \in \mathrm{PSH}(X,\theta )_{>0}$:
\[ \begin{aligned} P_{\theta }[\varphi ]=& \operatorname *{sup}_{}{}\! {}^{*}\left\{ \psi \in \mathrm{PSH}(X,\theta ):\psi \leq 0,\psi \preceq \varphi \right\} \\ =& \operatorname *{sup}_{}{}\! {}^{*}\left\{ \psi \in \mathrm{PSH}(X,\theta ):\psi \leq 0,\psi \sim \varphi \right\} \\ =& \operatorname *{sup}_{C\in \mathbb {Z}_{>0}}{}\! {}^{*} (\varphi +C)\land V_{\theta }\\ =& \sup \left\{ \psi \in \mathrm{PSH}(X,\theta ): \psi \leq 0,\varphi \preceq \psi , \int _ X \theta _{\varphi }^ n=\int _ X \theta _{\psi }^ n\right\} . \end{aligned} \]

Proof. Note that $\psi \sim _ P \varphi $ implies that $\psi \in \mathrm{PSH}(X,\theta )_{>0}$ by proposition 16.3.10. We observe that

\[ \begin{split} & \sup \left\{ \psi \in \mathrm{PSH}(X,\theta ): \psi \leq 0, \psi \sim _ P \varphi \right\} \\ =& \sup \left\{ \psi \in \mathrm{PSH}(X,\theta ): \psi \leq 0, \varphi \preceq \psi , \psi \sim _ P \varphi \right\} \end{split} \]

since for any candidate $\psi $ for the first supremum, $\psi \lor (\varphi -\sup _ X \varphi )$ is larger than $\psi $, and is a candidate for the second supremum thanks to 16.8.13. So the first equality is a direct consequence of proposition 16.8.7 and theorem 16.3.15.

Next we prove the second equality. We only need to show that for any $\psi \in \mathrm{PSH}(X,\theta )$ with $\psi \leq 0$ and $\psi \preceq _ P \varphi $, we have $\psi \leq P_{\theta }[\varphi ]$.

By 16.8.13 and example 16.8.6, we know that $P_{\theta }[\varphi ]\lor \psi \sim _ P \varphi $ and $P_{\theta }[\varphi ]\lor \psi \leq 0$. It follows from the first equality that $\psi \leq P_{\theta }[\varphi ]$. $\square$

Similarly, we have a new characterization of the $\mathcal{I}$-envelope.

Corollary 16.8.16.69E7C4B20FAE Let $\theta $ be a closed smooth real $(1,1)$-form on $X$ representing a pseudo-effective cohomology class. Assume that $\varphi \in \mathrm{PSH}(X,\theta )$. Then

\[ P_{\theta }[\varphi ]_{\mathcal{I}}=\sup \left\{ \psi \in \mathrm{PSH}(X,\theta ): \psi \leq 0, \psi \preceq _{\mathcal{I}} \varphi \right\} . \]

Proof. It suffices to show that for any $\psi \in \mathrm{PSH}(X,\theta )$ with $\psi \leq 0$ and $\psi \preceq _{\mathcal{I}}\varphi $, we have $\psi \leq P_{\theta }[\varphi ]_{\mathcal{I}}$. By 16.8.13 and proposition 16.6.11, we know that $P_{\theta }[\varphi ]_{\mathcal{I}}\lor \psi \sim _{\mathcal{I}}\varphi $. Therefore, $P_{\theta }[\varphi ]_{\mathcal{I}}\lor \psi $ is a candidate in the definition of $P_{\theta }[\varphi ]_{\mathcal{I}}$ (cf. (16.6.8)), so

\[ \psi \leq P_{\theta }[\varphi ]_{\mathcal{I}}\lor \psi \leq P_{\theta }[\varphi ]_{\mathcal{I}}. \]

We conclude the proof. $\square$

Proposition 16.8.17.2E8F613E7CBE Let $\theta $ be a closed real smooth $(1,1)$-form on $X$ and $\varphi ,\psi \in \mathrm{PSH}(X,\theta )$. Then the following are equivalent:

  1. 42227A6623F5

    $\varphi \preceq _{\mathcal{I}}\psi $;

  2. 6B5F8E56CB73

    $P_{\theta }[\varphi ]_{\mathcal{I}}\leq P_{\theta }[\psi ]_{\mathcal{I}}$.

Proof. 1 $\implies $ 2. This follows immediately from corollary 16.8.16.

2 $\implies $ 1. It follows from proposition 16.6.11 that

\[ \varphi \sim _{\mathcal{I}} P_{\theta }[\varphi ]_{\mathcal{I}}\leq P_{\theta }[\psi ]_{\mathcal{I}} \sim _{\mathcal{I}} \psi . \]

We conclude 1. $\square$

Example 16.8.18.C7A2AD00BFF3 Let us continue our example example 16.3.22, where $X=\mathbb {P}^1$, $\omega $ is the Fubini–Study metric and $\varphi \in \mathrm{PSH}(X,\omega )$ has log-log singularity at $0$. We have shown that $P_{\omega }[\varphi ]=0$ in (16.3.23), so $\varphi \sim _{P} 0$ and hence $\varphi \sim _{\mathcal{I}} 0$. In particular, $P$-equivalence is not equivalent to the equivalence of singularity types.

On the other hand, consider a potential $\psi \in \mathrm{PSH}(X,\omega )$ with log singularity at $0$, as in example 5.20.12. We know that $\nu (\psi ,0)=1$ from the explicit expression (5.20.13). So $\psi \not\sim _{\mathcal{I}} 0$ and hence $\psi \not\sim _{P} 0$.

Moreover, $\psi \preceq _{P} \varphi $ and hence $\psi \preceq _{\mathcal{I}}\varphi $.

We give an example showing that $P$-equivalence is not equivalent to $\mathcal{I}$-equivalence.

Example 16.8.19.325A49CBCF74 Let $X=\mathbb {P}^1$ and $\omega $ be the Fubini–Study metric. Let $K\subseteq \mathbb {P}^1$ be a polar Cantor set carrying an atom free probability measure $\mu $ supported on $K$ (see [Page 31, Car83]). Write $\mu =\omega +\mathrm{dd}^{\mathrm{c}}\varphi $ for some $\omega $-subharmonic function $\varphi $. Since $\mu $ is atom free, we know that all Lelong numbers of $\varphi $ are $0$. On the other hand, $\varphi $ has $0$ non-pluripolar mass since $K$ is pluripolar.

Then observe that $\varphi \sim _{\mathcal{I}} 0$ while $\varphi \not\sim _ P 0$.

For later use, we give the following definition.

Definition 16.8.20.33F50367BB68 Let $L$ be a pseudo-effective line bundle on $X$. An elementary metric on $L$ is a psh metric $h$ on $L$ such that there is a generalized Fubini–Study metric $h'$ on $L$ such that

\[ \mathrm{dd}^{\mathrm{c}}h\sim _{P} \mathrm{dd}^{\mathrm{c}}h'. \]

The set of elementary metrics on $L$ is denoted by $\mathrm{Ele}(L)$.

We also say $\mathrm{dd}^{\mathrm{c}}h$ is elementary. If we have fixed a Hermitian metric $h_0$ on $L$, and if we represent $h$ as $h_0\exp (-\varphi )$, we also say the quasi-psh function $\varphi $ is elementary.

Recall that the generalized Fubini–Study metrics are defined in definition 5.20.15.

16.8.21 Properties of the preordersF4304BE7EF91

Now we state a more natural version of the monotonicity theorem theorem 13.13.11.

Proposition 16.8.22 (Monotonicity theorem).82DCD8E3B1BA Let $\theta _1,\ldots ,\theta _ n$ be closed real smooth $(1,1)$-forms on $X$. Let $\varphi _ i,\psi _ i\in \mathrm{PSH}(X,\theta _ i)$ for $i=1,\ldots ,n$. Assume that $\varphi _ i\preceq _ P \psi _ i$ for each $i$. Then

\[ \int _ X \theta _{1,\varphi _1}\wedge \cdots \wedge \theta _{n,\varphi _ n}\leq \int _ X \theta _{1,\psi _1}\wedge \cdots \wedge \theta _{n,\psi _ n}. \]

In other words, the non-pluripolar mass is a decreasing quantity with respect to the $P$-singularity types of the potentials.

Proof. Fix a Kähler form $\omega $ on $X$. For each $i=1,\ldots ,n$, since $\varphi _ i\preceq _ P\psi _ i$, we have

\[ P_{\theta _ i+\epsilon \omega }[\varphi _ i]\leq P_{\theta _ i+\epsilon \omega }[\psi _ i] \]

for all $\epsilon >0$. Therefore, by proposition 16.3.10 and our previous monotonicity theorem theorem 13.13.11, we have

\[ \int _ X (\theta _1+\epsilon \omega )_{\varphi _1}\wedge \cdots \wedge (\theta _ n+\epsilon \omega )_{\varphi _ n}\leq \int _ X (\theta _1+\epsilon \omega )_{\psi _1}\wedge \cdots \wedge (\theta _ n+\epsilon \omega )_{\psi _ n}. \]

Letting $\epsilon \to 0+$ and applying the multi-linearity of the non-pluripolar products proposition 13.11.4 6, we find the desired inequality. $\square$

The $P$-preorder allows us to derive a natural generalization of 13.13.30.

Theorem 16.8.23.68E00C8C52A1 Let $\varphi ,\psi \in \mathrm{PSH}(X,\theta )_{>0}$, $\varphi \preceq _ P \psi $. Then for any

3BF94E656512 16.8.24
\begin{equation} a\in \left( 1, \left(\frac{\int _ X \theta _{\psi }^ n}{\int _ X \theta _{\psi }^ n-\int _ X \theta _{\varphi }^ n}\right)^{1/n} \right), \end{equation}

the function

\[ \eta \coloneqq P_{\theta }\Bigl (a\varphi +(1-a)\psi \Bigr ) \]

lies in $\mathrm{PSH}(X,\theta )_{>0}$, satisfies $\eta \preceq _ P \varphi $ and

9690BB77C920 16.8.25
\begin{equation} a^{-1}\eta +\left(1-a^{-1}\right)\psi \leq \varphi . \end{equation}

If in addition, $\varphi \sim _ P \psi $, then $\eta \sim _ P \varphi $ as well.

In (16.8.24), the fraction is understood as $\infty $ if either $\int _ X \theta _{\psi }^ n=\int _ X \theta _{\varphi }^ n$ or $n=0$, exactly as in 13.13.30.

Proof. By our assumption and (16.3.8), we have $\varphi \preceq P_{\theta }[\varphi ]\leq P_{\theta }[\psi ]$. Due to proposition 16.3.10, the interval (16.8.24) remains unchanged after replacing $\psi $ by $P_{\theta }[\psi ]$. Fix $a$ in this interval. So by 13.13.30 and the discussions afterwards,

\[ \eta '\coloneqq P_{\theta }\Bigl (a\varphi +(1-a)P_{\theta }[\psi ]\Bigr )\in \mathrm{PSH}(X,\theta )_{>0}. \]

We first check that $\eta $ is a genuine $\theta $-psh function. Choose a closed non-pluripolar set $K\subseteq X$ on which $\varphi $ and $\psi $ are bounded. Then the obstacle $a\varphi +(1-a)\psi $ is bounded above on $K$. Every candidate in the definition of $\eta $ is therefore uniformly bounded above on $X$ by 5.17.4, so $\eta \not\equiv \infty $ and hence $\eta \in \mathrm{PSH}(X,\theta )\cup \{ -\infty \} $. On the other hand, since $\psi \preceq P_{\theta }[\psi ]$, we can find $C>0$ such that $\psi \leq P_{\theta }[\psi ]+C$. Then

\[ \eta '-(a-1)C\leq a\varphi +(1-a)\psi \]

quasi-everywhere, hence $\eta \geq \eta '-(a-1)C$. Thus $\eta '\preceq \eta $, and the monotonicity theorem theorem 13.13.11 gives $\eta \in \mathrm{PSH}(X,\theta )_{>0}$.

By (13.13.20),

\[ \eta \leq a\varphi +(1-a)\psi \]

quasi-everywhere. Equivalently, (16.8.25) holds quasi-everywhere. Since both sides of (16.8.25) are $\theta $-psh, proposition 5.14.6 upgrades the inequality to an everywhere inequality.

It remains to prove the assertion about the $P$-order. Using proposition 16.4.2, we get

\[ a^{-1} P_{\theta }[\eta ]+\left(1-a^{-1}\right)P_{\theta }[\psi ] \leq P_{\theta }[\varphi ]. \]

By the assumption $\varphi \preceq _ P\psi $, we also have $P_{\theta }[\varphi ]\leq P_{\theta }[\psi ]$. Therefore,

\[ a^{-1} P_{\theta }[\eta ]\leq P_{\theta }[\varphi ]-\left(1-a^{-1}\right)P_{\theta }[\psi ]\leq a^{-1} P_{\theta }[\varphi ] \]

quasi-everywhere and hence everywhere by proposition 5.14.6. It follows that $P_{\theta }[\eta ]\leq P_{\theta }[\varphi ]$, namely $\eta \preceq _ P\varphi $.

Finally assume that $\varphi \sim _ P\psi $. Then $\int _ X\theta _{P_{\theta }[\psi ]}^ n=\int _ X\theta _{\varphi }^ n$ by proposition 16.3.10. Applying corollary 13.13.42 gives

\[ \int _ X\theta _{\eta '}^ n=\int _ X\theta _{\varphi }^ n. \]

Since $\eta '\preceq \eta $, the monotonicity theorem theorem 13.13.11 gives

\[ \int _ X\theta _{\eta }^ n\geq \int _ X\theta _{\varphi }^ n. \]

The reverse inequality follows from $\eta \preceq _ P\varphi $ and proposition 16.8.22. Hence $\eta \sim _ P\varphi $ by proposition 16.8.7. $\square$

Next we show that the $P$ and $\mathcal{I}$-preorders are preserved by some natural operations.

Lemma 16.8.26.FB1663F2B6E1 Let $\pi \colon Y\rightarrow X$ be a proper bimeromorphic morphism from a Kähler manifold $Y$. Given two quasi-plurisubharmonic functions $\varphi ,\psi $ on $X$, then the following are equivalent:

  1. 8ED2C9D6D32D

    $\varphi \preceq _{P} \psi $;

  2. 754FFF454EEB

    $\pi ^*\varphi \preceq _{P} \pi ^*\psi $.

The same holds with $\mathcal{I}$ in place of $P$.

Proof. In the $P$-case, this follows from proposition 16.4.1, while in the $\mathcal{I}$-case, this follows from proposition 16.6.10. $\square$

Proposition 16.8.27.C42706880DED Let $\varphi ,\psi ,\varphi ',\psi '\in \mathrm{QPSH}(X)$. Assume that

\[ \varphi \preceq _ P \psi ,\quad \varphi '\preceq _ P \psi '. \]

Then

\[ \varphi +\varphi '\preceq _ P \psi +\psi '. \]

The same holds with $\preceq _{\mathcal{I}}$ in place of $\preceq _ P$.

Proof. Take a Kähler form $\omega $ on $X$ such that $\varphi ,\psi ,\varphi ',\psi '\in \mathrm{PSH}(X,\omega )_{>0}$. The statement for $\preceq _{\mathcal{I}}$ is a simple consequence of proposition 5.16.8. We only need to handle the case of $\preceq _{P}$.

Step 1. We first show that

\[ P_{\omega }[\varphi ]+P_{\omega }[\varphi ']\sim _ P \varphi +\varphi '. \]

In fact, we clearly have

\[ P_{\omega }[\varphi ]+P_{\omega }[\varphi ']\succeq \varphi +\varphi '. \]

So by proposition 16.8.7, it suffices to show that they have the same mass. We compute

\[ \begin{aligned} & \int _ X \Bigl (2\omega +\mathrm{dd}^{\mathrm{c}}P_{\omega }[\varphi ]+\mathrm{dd}^{\mathrm{c}}P_{\omega }[\varphi ']\Bigr )^ n\\ =& \sum _{j=0}^ n \binom {n}{j} \int _ X \Bigl (\omega +\mathrm{dd}^{\mathrm{c}}P_{\omega }[\varphi ]\Bigr )^ j\wedge \Bigl (\omega +\mathrm{dd}^{\mathrm{c}}P_{\omega }[\varphi ']\Bigr )^{n-j}\\ =& \sum _{j=0}^ n \binom {n}{j} \int _ X \omega _{\varphi }^ j\wedge \omega _{\varphi '}^{n-j}\\ =& \int _ X \left(2\omega +\mathrm{dd}^{\mathrm{c}}\varphi +\mathrm{dd}^{\mathrm{c}}\varphi '\right)^ n, \end{aligned} \]

where we applied proposition 16.3.10 to the mixed products on the third line.

Step 2. By Step 1, we may assume that $\varphi ,\psi ,\varphi ',\psi '$ are all model potentials, after replacing them by their $P_{\omega }$-envelopes. So $\varphi \leq \psi $ and $\varphi '\leq \psi '$. Our assertion follows. $\square$

Proposition 16.8.28.1C6DED516857 Let $(\varphi _ i)_{i\in I}$, $(\psi _ i)_{i\in I}$ be uniformly bounded from above non-empty families in $\mathrm{QPSH}(X)$. Assume that there exists a closed smooth real $(1,1)$-form $\theta $ such that $\varphi _ i,\psi _ i\in \mathrm{PSH}(X,\theta )$ and $\varphi _ i\preceq _ P \psi _ i$ for all $i\in I$. Then

\[ \operatorname *{sup}_{i\in I}{}\! {}^{*} \varphi _ i\preceq _ P \operatorname *{sup}_{i\in I}{}\! {}^{*}\psi _ i. \]

The same holds with $\preceq _{\mathcal{I}}$ in place of $\preceq _ P$.

Proof. By adding a Kähler form to $\theta $, we may assume that $\varphi _ i,\psi _ i\in \mathrm{PSH}(X,\theta )_{>0}$ for all $i\in I$. The statement for $\preceq _{\mathcal{I}}$ is a simple consequence of corollary 5.16.9, we only have to consider the statement for $\preceq _ P$.

Step 1. We first handle the case where $I$ is a directed set and $(\varphi _ i)_{i\in I}$ and $(\psi _ i)_{i\in I}$ are increasing nets.

In this case, our assertion follows simply from proposition 16.4.9.

Step 2. We handle the case where $I$ is finite. We may assume that $I=\{ 0,1\} $. It suffices to show that

\[ P_{\theta }[\varphi _0]\lor P_{\theta }[\varphi _1]\sim _ P \varphi _0\lor \varphi _1, \]

which follows from 16.8.11.

Step 3. The general case can be reduced to the two cases handled in Step 1 and Step 2. More precisely, by applying proposition 5.14.3 to the two families and taking the union of the resulting countable subsets, we can find a countable subset $J\subseteq I$ such that

\[ \operatorname *{sup}_{j\in J}{}\! {}^{*} \varphi _ j=\operatorname *{sup}_{i\in I}{}\! {}^{*} \varphi _ i,\quad \operatorname *{sup}_{j\in J}{}\! {}^{*} \psi _ j=\operatorname *{sup}_{i\in I}{}\! {}^{*} \psi _ i. \]

We may replace $I$ by $J$ and assume that $I$ is countable. If $I$ is finite, Step 2 applies directly. Otherwise, write $I=\mathbb {Z}_{>0}$ and set

\[ \tilde{\varphi }_ m=\varphi _1\lor \cdots \lor \varphi _ m,\quad \tilde{\psi }_ m=\psi _1\lor \cdots \lor \psi _ m. \]

By Step 2, $\tilde{\varphi }_ m\preceq _ P\tilde{\psi }_ m$ for every $m$. The two new sequences are increasing and have the same upper envelopes as the original families, so Step 1 applies. $\square$

We now refine 16.3.32 using the $P$-equivalence relation.

Proposition 16.8.29.8C181599D520 Let $\varphi ,\psi ,\varphi ',\psi '\in \mathrm{PSH}(X,\theta )$ for some closed smooth real $(1,1)$-form $\theta $ on $X$. Assume that

\[ \varphi \sim _ P \varphi ',\quad \psi \sim _ P\psi ',\quad \varphi '\land \psi '\in \mathrm{PSH}(X,\theta )_{>0}. \]

Then

\[ \varphi \land \psi \in \mathrm{PSH}(X,\theta )_{>0},\quad \varphi \land \psi \sim _ P \varphi '\land \psi '. \]

Proof. We first observe that $\varphi ,\varphi ',\psi ,\psi '\in \mathrm{PSH}(X,\theta )_{>0}$ by assumption. Let

\[ \phi \coloneqq P_{\theta }[\varphi ]=P_{\theta }[\varphi '],\quad \gamma =P_{\theta }[\psi ]=P_{\theta }[\psi ']. \]

Since $\varphi '\preceq \phi $ and $\psi '\preceq \gamma $, after adding a common constant to $\varphi '$ and $\psi '$, we may assume that $\varphi '\leq \phi $ and $\psi '\leq \gamma $. Then $\varphi '\land \psi '\leq \phi \land \gamma $, so the assumption $\varphi '\land \psi '\in \mathrm{PSH}(X,\theta )_{>0}$ implies $\phi \land \gamma \in \mathrm{PSH}(X,\theta )_{>0}$ by monotonicity. It follows from 16.3.32 that

\[ \int _ X \theta _{\varphi '\land \psi '}^ n=\int _ X \theta _{\phi \land \gamma }^ n. \]

Next, we apply 16.3.32 again to conclude that $\varphi \land \psi \in \mathrm{PSH}(X,\theta )$ and

\[ \int _ X \theta _{\varphi \land \psi }^ n=\int _ X \theta _{\phi \land \gamma }^ n=\int _ X \theta _{\varphi '\land \psi '}^ n>0. \]

Since $\varphi \land \psi \preceq \phi \land \gamma $ and the two potentials have the same positive mass, theorem 16.3.15 gives $P_{\theta }[\varphi \land \psi ]=\phi \land \gamma $. The same argument gives $P_{\theta }[\varphi '\land \psi ']=\phi \land \gamma $, hence $\varphi \land \psi \sim _ P\varphi '\land \psi '$. $\square$

We can formulate a generalization of proposition 16.5.54 using the $P$-partial order.

Corollary 16.8.30.21CC5153F56F Let $\varphi ,\psi \in \mathrm{PSH}(X,\theta )_{>0}$. Assume that $\psi \preceq _ P \varphi $. Then for any $C>0$, $\psi \land (\varphi +C)\in \mathrm{PSH}(X,\theta )_{>0}$ and

42D232E5CD85 16.8.31
\begin{equation} P_{\theta }[\varphi ](\psi )=\operatorname *{sup}_{C>0}{}\! {}^{*} \psi \land (\varphi +C)=\psi . \end{equation}
We refer to definition 16.3.39 for the definition of $P_{\theta }[\varphi ](\psi )$.

Proof. Since $\psi \preceq _ P\varphi $, by proposition 16.8.29, we get

\[ \psi \land (\varphi +C)\in \mathrm{PSH}(X,\theta )_{>0} \]

and $\psi \land (\varphi +C)\sim _ P\psi $ for every $C>0$.

Let

\[ \eta \coloneqq \operatorname *{sup}_{C>0}{}\! {}^{*}\psi \land (\varphi +C). \]

Then $\eta \sim _ P \psi $. By corollary 16.3.14, we have

\[ \theta _\eta ^ n\leq \mathds {1}_{\{ \eta =\psi \} }\, \theta _\psi ^ n. \]

The domination principle theorem 13.13.43 gives $\eta \geq \psi $. The reverse inequality is trivial, so $\eta =\psi $. $\square$

The $P$-equivalence relation characterizes when a subgeodesic exists. See definition 17.10.2 for the notion of subgeodesics.

Theorem 16.8.32.1CA975F04423 Let $\varphi _0,\varphi _1\in \mathrm{PSH}(X,\theta )_{>0}$. Then the following are equivalent:

  1. 8E345DA2D98A

    There is a subgeodesic from $\varphi _0$ to $\varphi _1$;

  2. 5498C9F0BF2D

    $\varphi _0\sim _ P\varphi _1$.

Proof. 2 $\implies $ 1. This follows from proposition 17.10.24.

1 $\implies $ 2. Let $(\varphi _ t)_{t\in (0,1)}$ be a subgeodesic from $\varphi _0$ to $\varphi _1$.

Step 1. We assume that $\varphi _0\geq \varphi _1$.

We first replace the given subgeodesic by the geodesic from $\varphi _0$ to $\varphi _1$. Note that the geodesic is necessarily a subgeodesic as well by 17.10.22.

Observe that $t\mapsto \varphi _ t$ is decreasing due to 17.10.28.

Let $\varphi _ t=\varphi _1$ for all $t>1$. Then by the gluing lemma 17.10.16, we find that $(\varphi _ t)_{t\geq 0}$ is a subgeodesic ray.

Next, we consider the Legendre transform

\[ \Gamma _{\tau }\coloneqq \inf _{t\geq 0}(\varphi _ t-t\tau ),\quad \tau \in \mathbb {R}. \]

It follows from Kiselman's principle proposition 5.14.15 that $\Gamma _{\tau }\in \mathrm{PSH}(X,\theta )\cup \{ -\infty \} $. Note that for $\tau >0$, we clearly have $\Gamma _{\tau }\equiv -\infty $. On the other hand, for $\tau \leq 0$,

\[ \Gamma _{\tau }=\inf _{t\in [0,1]}(\varphi _ t-t\tau )\in \mathrm{PSH}(X,\theta ). \]

By Legendre inversion (cf. corollary 21.11.26), for $t>0$,

\[ \varphi _ t=\sup _{\tau <0} \left(\Gamma _{\tau }+t\tau \right). \]

Fix a Kähler form $\omega $ on $X$. It follows from proposition 16.8.28 that for each $t>0$,

\[ \varphi _ t\sim _ P \operatorname *{sup}_{\tau <0}{}\! {}^{*} P_{\theta +\omega }[\Gamma _{\tau }]. \]

The right-hand side is independent of $t$. Here by adding $\omega $, we no longer have to worry about the possibility where $\Gamma _{\tau }$ has vanishing mass.

Write

\[ \varphi _0=\operatorname *{sup}_{t\in (0,1)}{}\! {}^{*}\varphi _ t. \]

By proposition 16.8.28 again, we find that

\[ \varphi _0\sim _ P \operatorname *{sup}_{\tau <0}{}\! {}^{*} P_{\theta +\omega }[\Gamma _{\tau }] \]

as well. So $\varphi _0\sim _ P \varphi _1$.

Step 2. We prove the general case.

Observe that $(\varphi _ t\lor \varphi _1)_{t\in (0,1)}$ is a subgeodesic from $\varphi _0\lor \varphi _1$ to $\varphi _1$. By Step 1,

\[ \varphi _0\lor \varphi _1\sim _ P \varphi _1. \]

Hence $\varphi _0\preceq _ P \varphi _1$ by 16.8.13. The converse is proved similarly. Hence 2 follows. $\square$

Restating the 1 $\implies $ 2 direction in the theorem in terms of the complexification, we find the following interesting result.

Let $S=\{ z\in \mathbb {C}:0< \operatorname{Re}z <1\} $. We write $p_1\colon X\times S\rightarrow X$ for the natural projection.

Corollary 16.8.33.4EE47D80CC25 Let $\Phi \in \mathrm{PSH}(X\times S,p_1^*\theta )$. Assume that for any $c\in \mathbb {R}$, $x\in X$ and $z\in S$, we have

\[ \Phi (x,z)=\Phi (x,z+\mathrm{i}c). \]

Then $\int _ X (\theta +\mathrm{dd}^{\mathrm{c}}\Phi _ z)^ n$ is independent of $z\in S$, where $\Phi _ z \in \mathrm{PSH}(X,\theta )$ is given by $\Phi _ z(x)=\Phi (x,z)$.

Proof. Fix a Kähler form $\omega $ on $X$. We may regard $\Phi $ as an element of $\mathrm{PSH}(X\times S,p_1^*(\theta +\omega ))$. For any $0<a<b<1$, the restriction $(\Phi _ t)_{t\in [a,b]}$ is then a subgeodesic from $\Phi _ a$ to $\Phi _ b$ in the class $\theta +\omega $. The slices have positive mass in this class, so theorem 16.8.32 gives $\Phi _ a\sim _ P\Phi _ b$. By proposition 16.8.22, this directly implies

\[ \int _ X \theta _{\Phi _ a}^ n=\int _ X \theta _{\Phi _ b}^ n. \]

Since $\Phi $ is invariant under imaginary translations, $\Phi _ z$ depends only on $\operatorname{Re}z$, so the mass is independent of $z\in S$. $\square$

Cite this page

Comments

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