Positivity and 𝐋 𝟐 \mathbf{L^{2}} Extension

Dror Varolin · 2025

Read Positivity and 𝐋 𝟐 \mathbf{L^{2}} Extension by Dror Varolin (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-01753.txt.

Introduction

Dror Varolin, writing alone in 2025, examines when positivity of a vector bundle guarantees holomorphic extension. The L2 extension problem asks whether a holomorphic form defined on a subvariety extends to the whole space with controlled size. For line bundles, curvature positivity answers the question, but vector bundles come in several grades of positivity.

Varolin shows by example that Griffiths positivity is not enough for top-degree bundle-valued forms. He also gives a new proof that Nakano positivity suffices, following Berndtsson and Lempert, and proves a bundle version of a known Berndtsson result. The distinction sharpens tools used across complex geometry.

Catalog record

CREATORDror Varolin
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE101 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-01753

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Positivity and 𝐋 𝟐 \mathbf{L^{2}} Extension Dror Varolin Abstract . We examine the relationship between positivity of a vector bundle E and the problem of L 2 L^{2} extension of holomorphic E-valued forms of top degree. In particular, we show by example that Griffiths positivity is not enough. Though the sufficiency of Nakano positivity of E has been known for some time, we provide another proof along the lines of Berndtsson and Lempert. For such a proof we establish a vector bundle analogue of a well-known result of Berndtsson. This result is of independent interest and should have many other useful applications. 1. Introduction The L^{2} Extension Theorem for holomorphic sections of vector bundles has been of fundamental importance since its original inception by Ohsawa and Takegoshi [ OT-1987 ] . There are many versions of the result, but perhaps the following version strikes a good balance between generality and relative simplicity. Theorem 1.1 . Let X be an essentially Stein manifold equipped with a Kähler metric g , let Z be a smooth hypersurface cut out by a holomorphic section T\in H^{0}(X,{\mathcal{O}}(L_{Z})) of the line bundle L_{Z}\to X associated to the smooth divisor Z , and let {\mathfrak{g}} be a smooth metric for L_{Z} such that \sup_{X}|T|^{2}e^{-\lambda}\leq 1. Let E\to X be a holomorphic vector bundle with a smooth Hermitian metric {\mathfrak{h}} such that (1) \sqrt{\text{\rm-}1}\Theta({\mathfrak{h}})\geq_{\rm Nak}t\delta\sqrt{\text{\rm-}1}\Theta({\mathfrak{g}})\otimes{\rm Id}_{E}\quad\text{for all }t\in[0,1]. Then for every f\in H^{0}(Z,{\mathcal{O}}_{Z}(K_{Z}\otimes E|_{Z})) satisfying \int_{Z}|f|^{2}_{{\mathfrak{h}}}<+\infty there exists F\in H^{0}(X,{\mathcal{O}}_{X}(K_{X}\otimes L_{Z}\otimes E)) such that F|_{Z}=f\wedge dT\quad\text{and}\quad\int_{X}|F|^{2}_{{\mathfrak{h}}\otimes{\mathfrak{g}}}\leq\frac{\pi(1+\delta)}{\delta}\int_{Z}|f|^{2}_{{\mathfrak{h}}}. Though it is not immediately obvious, Theorem 1.1 follows from [ GZ-2015 , Theorem 2.1] . (If the optimal constant \frac{\pi(1+\delta)}{\delta} is not insisted upon then the result is much older.) However, the theorem of Guan-Zhou, itself an optimal-constant version of a theorem of Ohsawa [ O-2001 , Theorem 4] , is very general, and a large fraction of the content of [ GZ-2015 ] is devoted to extracting and proving corollaries. Theorem 1.1 is not among these corollaries, so in Section 5 we show how it follows. In the past few years a link has been drawn between L^{2} Extension and Griffiths positivity. In fact, in [ DNWZ-2023 ] it is shown that asymptotically universal local L^{2} extension from a point 1 1 1 This is not the terminology of [ DNWZ-2023 ] . implies Griffiths positivity. The converse was proved in [ KP-2021 ] for smooth metrics and more generally for metrics whose determinant is locally bounded. The aforementioned link suggests the following question. Question . If the rank of E is at least two, is Theorem 1.1 true if the curvature hypothesis ( 1 ) is required to hold in the weakest sense, namely in the sense of Griffiths? Unfortunately the answer is negative, as the following theorem shows. Theorem 1 . There exist a compact Kähler manifold X , a compact complex submanifold Z\subset X of codimension 1 , a section T\in H^{0}(X,{\mathcal{O}}(L_{Z})) , a metric {\mathfrak{g}} for L_{Z} , a holomorphic vector bundle E\to X and a metric {\mathfrak{h}} for E\to X such that for some \delta>0 \sup_{X}|T|^{2}_{{\mathfrak{g}}}\leq 1,\quad\sqrt{\text{\rm-}1}\Theta({\mathfrak{h}})\geq_{\rm Grif}\delta\sqrt{\text{\rm-}1}\Theta({\mathfrak{g}})\quad\text{for all }t\in[0,1], and the restriction map H^{0}(X,{\mathcal{O}}(K_{X}\otimes L_{Z}\otimes E))\ni F\mapsto\frac{F|_{Z}}{dT}\in H^{0}(Z,{\mathcal{O}}(K_{Z}\otimes E|_{Z})) fails to be surjective. Theorem 1 is proved in Section 3 . If one nevertheless tries to prove an extension theorem like Theorem 1.1 for a Griffiths-positive vector bundle E\to X , the natural step is to pass to the projectivization {\mathcal{O}}_{E}(1)\to{\mathbb{P}}(E^{*}) and try to invoke the rank-1 case of Theorem 1.1 . In Section 6 we investigate this approach. After the appearance of [ GZ-2015 ] , Berndtsson and Lempert gave a proof of Theorem 1.1 for functions on domains and under curvature hypotheses that are slightly more restrictive than ( 1 ). In Section 5 we extend the proof of Berndtsson and Lempert to the setting of vector bundles over Stein manifolds. To do so, we also need a higher-rank version of Berndtsson’s result on positivity of direct images [ B-2009 , Theorem 1.1] , which we now formulate. Let X be a relatively compact domain in a complex manifold Y of complex dimension n , let B be a complex manifold of dimension m , and denote by p_{1}:Y\times B\to Y and p_{2}:Y\times B\to B the Cartesian projections. Let \tilde{E}\to Y be a holomorphic vector bundle of rank r . Fix a smooth metric \tilde{\mathfrak{h}} for the holomorphic vector bundle p_{1}^{*}\tilde{E}\to Y\times B . We set E:=\tilde{E}|_{X} and {\mathfrak{h}}:=\tilde{\mathfrak{h}}|_{X\times B} . Thus we have a trivial holomorphic family of vector bundles with non-trivial metric: (p_{1}^{*}E,{\mathfrak{h}})\to X\times B\stackrel{{\scriptstyle p_{2}}}{{\to}}B. Though possibly non-trivial, the metric {\mathfrak{h}} very usefully extends smoothly to \partial X\times B . To each t\in B we associate the Bergman space {\mathscr{H}}_{t}:=\left\{f\in H^{0}(X,{\mathcal{O}}(K_{X}\otimes E))\ ;\ \int_{X}\sqrt{\text{\rm-}1}^{n^{2}}\left<f\wedge\bar{f},{\mathfrak{h}}^{(t)}\right><+\infty\right\}, where {\mathfrak{h}}^{(t)} is the metric for E\to X that we canonically identify with the metric {\mathfrak{h}}(\cdot,t) obtained by restricting the metric {\mathfrak{h}} for p_{1}^{*}E\to X\times B to the fiber over X\times\{t\} . It is well-known that {\mathscr{H}}_{t} is a Hilbert space. Since X is a bounded domain and the metrics {\mathfrak{h}}^{(t)} , t\in B , are all smooth up to the boundary of X , the vector subspaces {\mathscr{H}}_{t}\subset H^{0}(X,{\mathcal{O}}(K_{X}\otimes E)) do not depend on t . We can therefore define the trivial vector bundle {\mathscr{H}}\to B whose fiber over t\in B is {\mathscr{H}}_{t} . Note that a smooth (resp. holomorphic) section of {\mathscr{H}}\to B is uniquely determined by a smooth (resp. holomorphic) section of p_{1}^{*}E\to X\times B whose restriction to each fiber X\times\{t\} lies in {\mathscr{H}}_{t} . One defines the metric (f_{1},f_{2}):=\sqrt{\text{\rm-}1}^{n^{2}}\int_{X}\left<f_{1}\wedge\bar{f}_{2},{\mathfrak{h}}^{(t)}\right>,\quad f_{1},f_{2}\in{\mathscr{H}}_{t} for the possibly infinite-rank vector bundle {\mathscr{H}}\to B . Then one has a Chern connection (which is only densely-defined if the rank of {\mathscr{H}} is infinite) that we compute in Section 4 . We then have the following theorem. Theorem 2 . With the above notation, assume that Y is Stein and that X is pseudoconvex. If a. (E,{\mathfrak{h}}^{(t)}) is Nakano-nonnegative on X for each t\in B and b. (E,{\mathfrak{h}}) is k -positive (resp. k -nonnegative) on X\times B then {\mathscr{H}}\to B is k -positive (resp. k -nonnegative). The proof of Theorem 2 occupies Section 4 . Such a ‘deformation of Bergman spaces’ is a special case of a more general story. A short summary is discussed in Section 3 , and a more detailed presentation can be found in [ V-2024 ] . Acknowledgments . I am grateful to Sebastien Boucksom and Mihai Păun for stimulating conversations. This work was done in part during visits to Sorbonne Université and Universität Bayreuth, and I thank those institutions for their hospitality. Finally, I am grateful to the anonymous referee for their effort and their very useful comments, which have made the paper much easier to read. 2. Background and preliminaries We present some background material for the reader’s convenience and to establish notation. 2.1. Positivity of holomorphic Hermitian vector bundles Let E\to X be a holomorphic vector bundle with a smooth Hermitian metric {\mathfrak{h}} . The curvature of the Chern connection is a {\rm Hom}(E,E) -valued (1,1) -form \Theta({\mathfrak{h}}) on X that is {\mathfrak{h}} -skew Hermitian, i.e., {\mathfrak{h}}(\Theta({\mathfrak{h}})v,w)+{\mathfrak{h}}(v,\Theta({\mathfrak{h}})w)=0\quad\text{for all $x\in X$ and all $v,w\in E_{x}$}. One therefore obtains a Hermitian form on T^{1,0}_{X}\otimes E\to X defined on indecomposable tensors by \{\xi\otimes

Read the complete text