ScalingStacks

7. General θ -psh functions and semipositive singular metrics [01GJ]

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 · Original author HTML

7. General θ\theta-psh functions and semipositive singular metrics

We are now ready to introduce the class of general θ\theta-psh functions and their cousins: semipositive singular metrics. The equicontinuity result in Corollary 6.2 will be used to show Theorem A, asserting that the space of θ\theta-psh functions is compact up to translation.

Throughout this section we let XX be as before a smooth connected projective KK-analytic variety and fix a closed (1,1)(1,1)-form θ∈𝒵1,1​(X)\theta\in\mathcal{Z}^{1,1}(X) whose de Rham class {θ}∈N1​(X)\{\theta\}\in N^{1}(X) is ample. As before we write “psh” as a shorthand for “plurisubharmonic”. Similarly, “usc” will mean “upper semicontinuous”.

Definition 7.1.

Let θ\theta be as above. A θ\theta-psh function φ:X→[−∞,+∞[\varphi:X\to[-\infty,+\infty[ is an usc function such that for each SNC model 𝒳\mathcal{X} of XX on which θ\theta is determined we have

  1. (i)

    φ≤φ∘p𝒳\varphi\leq\varphi\circ p_{\mathcal{X}} on XX;

  2. (ii)

    the restriction φ∘emb𝒳\varphi\circ\emb_{\mathcal{X}} of φ\varphi to Δ𝒳\Delta_{\mathcal{X}} is a uniform limit of restrictions of θ\theta-psh model functions.

We write PSH⁡(X,θ)\PSH(X,\theta) for the set of θ\theta-psh functions on XX.

We say that φ:X→[−∞,+∞[\varphi:X\to[-\infty,+\infty[\, is quasi-psh if φ\varphi is θ\theta-psh for some θ\theta as above. Thanks to Theorem 5.11, the previous definition is consistent with Definition 5.5 when φ\varphi is a model function.

Remark 7.2.

A function on a compact (complex) Kähler manifold XX is quasi-psh if it is locally the sum of a psh function and a smooth function. Given a closed (1,1)(1,1)-form θ\theta, an θ\theta-psh function φ\varphi is a quasi-psh function such that θ+d​dc​φ≥0\theta+dd^{c}\varphi\geq 0 in the sense of currents. When the de Rham class {θ}∈H1,1​(X)\{\theta\}\in H^{1,1}(X) is a Kähler class, we have a global characterization: θ\theta-psh functions are decreasing limits of sequences of smooth θ\theta-psh functions, see [Dem92, Theorem 1.1]. In our current non-Archimedean setting, a local theory of psh functions is still to be developed. For this reason, we work globally and assume that {θ}\{\theta\} is ample.

Recall the definitions from §4 of singular and model metrics on line bundles.

Definition 7.3.

Let LL be an ample line bundle on XX. A singular metric ∥⋅∥\|\cdot\| on LL is semipositive if it is of the form ∥⋅∥=∥⋅∥e−φ\|\cdot\|=\|\cdot\|e^{-\varphi}, where ∥⋅∥\|\cdot\| denotes a model metric and φ\varphi is c1(L,∥⋅∥)c_{1}(L,\|\cdot\|)-psh.

One checks that this definition does not depend on the choice of reference metric ∥⋅∥\|\cdot\|. Below we shall state various properties of θ\theta-psh functions. We leave it to the reader to formulate analogous assertions about semipositive singular metrics on ample line bundles.

7.1. Basic properties

Fix a closed (1,1)(1,1)-form θ∈𝒵1,1​(X)\theta\in\mathcal{Z}^{1,1}(X) as above. From Proposition 5.8 and 5.9 we obtain:

Proposition 7.4.

The set PSH⁡(X,θ)\PSH(X,\theta) is convex. If φ\varphi is θ\theta-psh and c∈𝐑c\in\mathbf{R}, then φ+c\varphi+c is θ\theta-psh. If θ\theta is furthermore semipositive, then max⁡{φ,ψ}\max\{\varphi,\psi\} is θ\theta-psh when φ,ψ∈PSH⁡(X,θ)\varphi,\psi\in\PSH(X,\theta).

Proposition 7.5.

If φ∈PSH⁡(X,θ)\varphi\in\PSH(X,\theta) and 𝒳\mathcal{X} is an SNC model on which θ\theta is determined, then:

  • (i)

    φ∘emb𝒳\varphi\circ\emb_{\mathcal{X}} is continuous on Δ𝒳\Delta_{\mathcal{X}} and convex on each face;

  • (ii)

    φ∘p𝒳\varphi\circ p_{\mathcal{X}} is continuous on XX and φ≤φ∘p𝒳\varphi\leq\varphi\circ p_{\mathcal{X}}.

The next result shows how to reconstruct a θ\theta-psh function from its values on quasimonomial points.

Proposition 7.6.

Let φ∈PSH⁡(X,θ)\varphi\in\PSH(X,\theta). Then, as 𝒳\mathcal{X} runs through the directed set of SNC models on which θ\theta is determined, (φ∘p𝒳)𝒳(\varphi\circ p_{\mathcal{X}})_{\mathcal{X}} forms a decreasing net of continuous functions on XX, converging pointwise to φ\varphi.

Proof.

Let 𝒳′≥𝒳\mathcal{X}^{\prime}\geq\mathcal{X} be two SNC models on which θ\theta is determined. Then p𝒳∘p𝒳′=p𝒳p_{\mathcal{X}}\circ p_{\mathcal{X}^{\prime}}=p_{\mathcal{X}}. By Proposition 7.5 (ii) this implies φ≤φ∘p𝒳′≤φ∘p𝒳∘p𝒳′=φ∘p𝒳\varphi\leq\varphi\circ p_{\mathcal{X}^{\prime}}\leq\varphi\circ p_{\mathcal{X}}\circ p_{\mathcal{X}^{\prime}}=\varphi\circ p_{\mathcal{X}}, with equality on emb𝒳⁡(Δ𝒳)\emb_{\mathcal{X}}(\Delta_{\mathcal{X}}). Set φ~:=lim𝒳φ∘p𝒳\tilde{\varphi}:=\lim_{\mathcal{X}}\varphi\circ p_{\mathcal{X}}. Then φ~≥φ\tilde{\varphi}\geq\varphi. On the other hand, it follows from Corollary 3.2 that p𝒳p_{\mathcal{X}} converges to the identity on XX, so by upper semicontinuity of φ\varphi we have φ≥φ~\varphi\geq\tilde{\varphi}. ∎

7.2. Equicontinuity

The Lipschitz estimates in Theorem 6.1 carry over to general θ\theta-psh functions. As a consequence we have

Corollary 7.7.

For any SNC model on which θ\theta is determined, the family

{φ∘emb𝒳∣φ∈PSH⁡(X,θ)}\{\varphi\circ\emb_{\mathcal{X}}\mid\varphi\in\PSH(X,\theta)\}

is an equicontinuous family of functions on Δ𝒳\Delta_{\mathcal{X}}.

7.3. Compactness

We endow the set PSH⁡(X,θ)\PSH(X,\theta) of all θ\theta-psh functions with the topology of uniform convergence on dual complexes. A basis of open neighborhoods of a fixed θ\theta-psh function φ0\varphi_{0} is then given by {φ∣supΔ𝒳|φ−φ0|≤ε}\{\varphi\mid\sup_{\Delta_{\mathcal{X}}}|\varphi-\varphi_{0}|\leq\varepsilon\} where 𝒳\mathcal{X} ranges over SNC models on which θ\theta is determined and where ε>0\varepsilon>0. Thanks to Proposition 7.6, the natural map

PSH⁡(X,θ)→∏𝒳C0​(Δ𝒳)\PSH(X,\theta)\to\prod_{\mathcal{X}}C^{0}(\Delta_{\mathcal{X}})

is then a homeomorphism onto its image. Note also that 𝒟⁡(X)∩PSH⁡(X,θ)\mathcal{D}(X)\cap\PSH(X,\theta) is dense in PSH⁡(X,θ)\PSH(X,\theta) by definition. The following result implies Theorem A.

Theorem 7.8.

The map PSH⁡(X,θ)→𝐑\PSH(X,\theta)\to\mathbf{R} defined by φ↦supXφ\varphi\mapsto\sup_{X}\varphi is continuous and proper. Hence PSH⁡(X,θ)/𝐑\PSH(X,\theta)/\mathbf{R} is compact. Furthermore, the topology on PSH⁡(X,θ)\PSH(X,\theta) is equivalent to the topology of pointwise convergence on either XqmX^{\mathrm{qm}} or XdivX^{\mathrm{div}}.

Proof.

If 𝒳\mathcal{X} is an SNC model on which θ\theta is determined, then it follows from Proposition 7.6 (ii) that the supremum of any φ∈PSH⁡(X,θ)\varphi\in\PSH(X,\theta) is attained on emb𝒳⁡(Δ𝒳)\emb_{\mathcal{X}}(\Delta_{\mathcal{X}}). This implies the continuity of φ↦supXφ\varphi\mapsto\sup_{X}\varphi.

To prove properness, recall that PSH⁡(X,θ)\PSH(X,\theta) embeds in ∏𝒳C0​(Δ𝒳)\prod_{\mathcal{X}}C^{0}(\Delta_{\mathcal{X}}). By Tychonoff’s theorem, the compactness of

ℱC:={φ∈PSH⁡(X,θ)∣|supXφ|≤C}\mathcal{F}_{C}:=\{\varphi\in\PSH(X,\theta)\mid|\sup_{X}\varphi|\leq C\}

is therefore equivalent to the compactness in C0​(Δ𝒳)C^{0}(\Delta_{\mathcal{X}}) of the closure of the image of ℱC\mathcal{F}_{C} in C0​(Δ𝒳)C^{0}(\Delta_{\mathcal{X}}), for each SNC model 𝒳\mathcal{X} on which θ\theta is determined. But this is a direct consequence of Corollary 7.7 and Ascoli’s theorem.

For the last statement, it is clear that convergence in PSH⁡(X,θ)\PSH(X,\theta) implies pointwise convergence on XqmX^{\mathrm{qm}} which in turn implies pointwise convergence on XdivX^{\mathrm{div}}. Now let (φα)α∈A(\varphi_{\alpha})_{\alpha\in A} be a net of θ\theta-psh functions converging pointwise to φ∈PSH⁡(X,θ)\varphi\in\PSH(X,\theta) on XdivX^{\mathrm{div}}. Fix any SNC model 𝒳\mathcal{X} on which θ\theta is determined. We must show that φα\varphi_{\alpha} converges uniformly to φ\varphi on emb𝒳⁡(Δ𝒳)\emb_{\mathcal{X}}(\Delta_{\mathcal{X}}). But Xdiv∩emb𝒳⁡(Δ𝒳)X^{\mathrm{div}}\cap\emb_{\mathcal{X}}(\Delta_{\mathcal{X}}) is the image under emb𝒳\emb_{\mathcal{X}} of the rational points in Δ𝒳\Delta_{\mathcal{X}} by Corollary 3.13, and is therefore dense in emb𝒳⁡(Δ𝒳)\emb_{\mathcal{X}}(\Delta_{\mathcal{X}}). The uniform convergence on emb𝒳⁡(Δ𝒳)\emb_{\mathcal{X}}(\Delta_{\mathcal{X}}) therefore follows from the equicontinuity statement in Corollary 7.7. ∎

7.4. Upper envelopes

As a consequence of compactness we shall prove the following result, whose complex analogue serves as a basic ingredient of pluripotential theory. While we will not go deeper into pluripotential theory here, we will use the result below in §8.

Theorem 7.9.

Let (φα)α∈A(\varphi_{\alpha})_{\alpha\in A} be an arbitrary set of θ\theta-psh functions on XX and assume that (φα)(\varphi_{\alpha}) is uniformly bounded from above. If we set φ⁡(x):=supα∈Aφα​(x)\varphi(x):=\sup_{\alpha\in A}\varphi_{\alpha}(x) for each x∈Xx\in X, then the usc regularization φ∗\varphi^{*} of φ\varphi is θ\theta-psh and coincides with φ\varphi on Xqm=⋃𝒳emb𝒳⁡(Δ𝒳)X^{\mathrm{qm}}=\bigcup_{\mathcal{X}}\emb_{\mathcal{X}}(\Delta_{\mathcal{X}}).

Recall that the usc regularization of a function uu on a topological space XX is the smallest usc function u∗≥uu^{*}\geq u.

Lemma 7.10.

Let u:X→[−∞,+∞[u:X\to[-\infty,+\infty[ be a function such that for each SNC model 𝒳\mathcal{X} we have

  • (i)

    u∘i𝒳u\circ i_{\mathcal{X}} is continuous on Δ𝒳\Delta_{\mathcal{X}}.

  • (ii)

    u≤u∘p𝒳u\leq u\circ p_{\mathcal{X}}.

Then u∗=inf𝒳u∘p𝒳u^{*}=\inf_{\mathcal{X}}u\circ p_{\mathcal{X}}, hence u∗∘i𝒳=u∘i𝒳u^{*}\circ i_{\mathcal{X}}=u\circ i_{\mathcal{X}} for all 𝒳\mathcal{X}.

Proof.

Condition (i) implies that u∘p𝒳=u∘i𝒳∘p𝒳u\circ p_{\mathcal{X}}=u\circ i_{\mathcal{X}}\circ p_{\mathcal{X}} is continuous for all 𝒳\mathcal{X}, so that v:=inf𝒳u∘p𝒳v:=\inf_{\mathcal{X}}u\circ p_{\mathcal{X}} is usc. It follows that v≥u∗v\geq u^{*}, since v≥uv\geq u by (ii). Conversely, for each x∈Xx\in X we have lim𝒳p​𝒳​(x)=x\lim_{\mathcal{X}}p\mathcal{X}(x)=x, hence

u∗​(x)≥lim sup𝒳u∗∘p𝒳​(x)≥inf𝒳u∘p𝒳​(x)=v⁡(x),u^{*}(x)\geq\limsup_{\mathcal{X}}u^{*}\circ p_{\mathcal{X}}(x)\geq\inf_{\mathcal{X}}u\circ p_{\mathcal{X}}(x)=v(x),

which shows that v=u∗v=u^{*}. Finally, (ii) shows that 𝒳′≥𝒳⇒u∘p𝒳′≤u∘p𝒳\mathcal{X}^{\prime}\geq\mathcal{X}\Rightarrow u\circ p_{\mathcal{X}^{\prime}}\leq u\circ p_{\mathcal{X}}, hence v∘p𝒳=u∘p𝒳v\circ p_{\mathcal{X}}=u\circ p_{\mathcal{X}}, which is equivalent to the last assertion. ∎

Proof of Theorem 7.9.

Upon considering the new family φI=maxα∈I⁡φα\varphi_{I}=\max_{\alpha\in I}\varphi_{\alpha} with II ranging over all finite subsets of AA, we may assume that AA is a directed set and (φα)(\varphi_{\alpha}) is an increasing net. For each SNC model 𝒳\mathcal{X} we have φα≤φα∘p𝒳\varphi_{\alpha}\leq\varphi_{\alpha}\circ p_{\mathcal{X}} for all α\alpha, hence φ≤φ∘p𝒳\varphi\leq\varphi\circ p_{\mathcal{X}}. By Corollary 7.7 φα∘i​𝒳\varphi_{\alpha}\circ i\mathcal{X} converges uniformly to φ∘i𝒳\varphi\circ i_{\mathcal{X}}, which is therefore continuous. Using Lemma 7.10 we conclude that φ∗\varphi^{*} is usc, satisfies φ∗≤φ∗∘p𝒳\varphi^{*}\leq\varphi^{*}\circ p_{\mathcal{X}}, and φ∗∘i𝒳=φ∘i𝒳\varphi^{*}\circ i_{\mathcal{X}}=\varphi\circ i_{\mathcal{X}} is a uniform limit of restrictions to Δ𝒳\Delta_{\mathcal{X}} of θ\theta-psh functions, hence φ∗\varphi^{*} is θ\theta-psh. ∎

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