This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Uniform L2L^{2}-estimates for flat nontrivial line bundles on compact complex manifolds

Yoshinori Hashimoto1 [email protected] 1,2Department of Mathematics, Osaka Metropolitan University, 3-3-138, Sugimoto, Sumiyoshi-ku, Osaka, 558-8585, Japan. Takayuki Koike2 [email protected]  and  Shin-ichi Matsumura3 [email protected] [email protected] 3Mathematical Institute, Tohoku University, 6-3, Aramaki Aza-Aoba, Aoba-ku, Sendai 980-8578, Japan.
Abstract.

In this paper, we extend the uniform L2L^{2}-estimate of ¯\bar{\partial}-equations for flat nontrivial line bundles, proved for compact Kähler manifolds in [KH22], to compact complex manifolds. In the proof, by tracing the Dolbeault isomorphism in detail, we derive the desired L2L^{2}-estimate directly from Ueda’s lemma.

1. Introduction

This paper studies the uniform estimate of the L2L^{2}-norm u\|u\| of the solution of the ¯\bar{\partial}-equation ¯u=v\bar{\partial}u=v for flat nontrivial line bundles. We begin by recalling the uniform L2L^{2}-estimate on compact Kähler manifolds proved by the first two authors in [KH22]. Let XX be a compact complex manifold, and let 𝒫(X)\mathcal{P}(X) be the set of all unitarily flat (holomorphic) line bundles on XX, which can be regarded as an abelian group under the tensor product and is compact with respect to a group-invariant distance 𝖽\mathsf{d} as defined in e.g. [KH22]*Lemma 2.6 and [Ueda]*§4.5 (see also [KoikeUehara]*§A.3). It is well-known that 𝒫(X)\mathcal{P}(X) is a direct sum of copies of the Picard variety Pic0(X)\mathrm{Pic}^{0}(X) (see e.g. [KH22]*Lemma 2.2). We write 𝕀X\mathbb{I}_{X} for the trivial line bundle on XX, which is the identity in 𝒫(X)\mathcal{P}(X). Then, we have:

Theorem 1.1 ([KH22]*Theorem 1.1).

Let (X,g)(X,g) be a compact Kähler manifold. Then, there exists a constant K>0K>0 such that, for any element F𝒫(X){𝕀X}F\in\mathcal{P}(X)\setminus\{\mathbb{I}_{X}\} and any smooth ¯\bar{\partial}-closed (0,1)(0,1)-form vv with values in FF whose Dolbeault cohomology class [v]H0,1(X,F)[v]\in H^{0,1}(X,F) is trivial, there exists a unique smooth section uu of FF such that ¯u=v\bar{\partial}u=v and

X|u|h2𝑑VgK𝖽(𝕀X,F)X|v|h,g2𝑑Vg\sqrt{\int_{X}|u|_{h}^{2}\,dV_{g}}\leq\frac{K}{\mathsf{d}(\mathbb{I}_{X},F)}\sqrt{\int_{X}|v|_{h,g}^{2}\,dV_{g}}

hold for a flat Hermitian metric hh on FF.

This result can be regarded as an L2L^{2}-version of Ueda’s lemma. Ueda’s lemma (discussed below) is a foundational result in complex dynamics and complex analysis.

Theorem 1.2 (Ueda’s lemma, [Ueda]*Lemma 4).

Let XX be a compact complex manifold and 𝒰:={Uj}\mathcal{U}:=\{U_{j}\} be a finite open covering of XX such that each UjU_{j} is a Stein open set and UjU_{j} trivializes any F𝒫(X)F\in\mathcal{P}(X). Then, there exists a constant K𝒰>0K_{\mathcal{U}}>0 such that, for any F𝒫(X){𝕀X}F\in\mathcal{P}(X)\setminus\{\mathbb{I}_{X}\} and any Čech 0-cochain 𝔣:={(Uj,fj)}Cˇ0(𝒰,𝒪X(F))\mathfrak{f}:=\{(U_{j},f_{j})\}\in\check{C}^{0}(\mathcal{U},\mathcal{O}_{X}(F)), the inequality

maxjsupUj|fj|hK𝒰𝖽(𝕀X,F)maxj,ksupUjk|fjk|h\max_{j}\sup_{U_{j}}|f_{j}|_{h}\leq\frac{K_{\mathcal{U}}}{\mathsf{d}(\mathbb{I}_{X},F)}\max_{j,k}\sup_{U_{jk}}|f_{jk}|_{h}

holds for a flat Hermitian metric hh on FF, where {(Ujk,fjk)}:=δ𝔣Cˇ1(𝒰,𝒪X(F))\{(U_{jk},f_{jk})\}:=\delta\mathfrak{f}\in\check{C}^{1}(\mathcal{U},\mathcal{O}_{X}(F)) is the Čech coboundary of 𝔣\mathfrak{f}.

This can be partially generalized to higher degree cohomology classes, leading to a cohomology vanishing result [KH22]*Theorem 1.2, which seems to be related to the generic vanishing theorem (see e.g. [Lazarsfeld1]*§4.4).

As written in [KH22]*§2.3, Theorem 1.1 implies Theorem 1.2 when XX is a compact Kähler manifold. It is natural to speculate that Theorem 1.1 may hold for a compact complex manifold that is not necessarily Kähler. We affirmatively answer this question, as follows:

Theorem 1.3.

Theorem 1.1 holds for a compact complex manifold (X,g)(X,g) with a Hermitian metric gg.

An important distinction from Theorem 1.1 is that the proof of the above theorem crucially depends on Ueda’s lemma (Theorem 1.2); Indeed, Theorem 1.3 is directly derived from Ueda’s lemma. One of the motivations behind Theorem 1.1 was to find a new proof of Ueda’s lemma in such a way that it is more geometric compared to the original argument in [Ueda], which was rather technical. Here, we establish a direction that is somewhat contrary to the one discussed in [KH22], with the merit of being able to extend [KH22]*Theorem 1.1 to general compact complex manifolds.

Acknowledgements

The authors would like to thank the organizers of the 7th workshop “Complex Geometry and Lie Groups,” which served as an impetus for this collaboration.

The first named author thanks Osamu Fujino, Hisashi Kasuya, and Shinnosuke Okawa for helpful discussions, and is supported by Grant-in-Aid for Scientific Research (C) \sharp23K03120 and Grant-in-Aid for Scientific Research (B) \sharp24K00524. The second named author is supported by Grant-in-Aid for Scientific Research (C) \sharp23K03119. The third named author is supported by Grant-in-Aid for Scientific Research (B) \sharp21H00976 from JSPS.

2. Proof of Theorem 1.3

This section is devoted to the proof of Theorem 1.3. In what follows, we adhere to the notation used in Theorem 1.3 and [KH22].

Proof of Theorem 1.3.

The strategy of the proof is to chase the Dolbeault isomorphism between ¯\bar{\partial}-cohomology and Čech cohomology, using the Hörmander’s L2L^{2}-method, which allows us to apply Ueda’s lemma for the δ\delta-equation.

We work with a fixed Hermitian metric gg on XX, with all norms and volume forms defined with respect to gg. We now recall a foundational result by Hörmander (see e.g. [Demailly_book]*Chapter VIII, Theorem 6.9, [Dem82]*5, 4.1 Théorème, or [Hormbook]*Theorem 4.4.2), which guarantees the following: Let UxU^{*}_{x} be a Stein open neighborhood of a given point xXx\in X, and let h=eφh=e^{-\varphi} be a singular Hermitian metric on the trivial line bundle with φ\varphi psh. Then, for any smooth (0,1)(0,1)-form vxv_{x} on UxU^{*}_{x} such that ¯vx=0\bar{\partial}v_{x}=0 and vxL2(Ux),h<\|v_{x}\|_{L^{2}(U^{*}_{x}),h}<\infty, there exists a smooth function uxu_{x} on UxU^{*}_{x} such that

¯ux=vx on Ux\bar{\partial}u_{x}=v_{x}\text{ on }U_{x}^{*}

with the L2L^{2} estimate

uxL2(Ux),hCx,gvxL2(Ux),h,\|u_{x}\|_{L^{2}(U^{*}_{x}),h}\leq C_{x,g}\|v_{x}\|_{L^{2}(U^{*}_{x}),h},

where the constant Cx,g>0C_{x,g}>0 depends only on UxU^{*}_{x} and gg. Note that the solution of uxu_{x} can be chosen to be smooth if h=eφh=e^{-\varphi} and vxv_{x} are smooth.

We may assume, without loss of generality and by shrinking each UxU^{*}_{x}, that UxU^{*}_{x} trivializes any flat line bundle on XX (see [KH22]*Lemma 2.7). For each xXx\in X, we pick a relatively compact open subset UxUxU^{**}_{x}\Subset U^{*}_{x}. We then cover XX with the open sets {Ux}xX\{U^{**}_{x}\}_{x\in X}, and by the compactness of XX, we may select a finite subcover {Uxj}jI\{U^{**}_{x_{j}}\}_{j\in I}. By setting Uj:=UxjU^{\prime}_{j}:=U^{**}_{x_{j}} and Uj:=UxjU_{j}:=U^{*}_{x_{j}} for each jIj\in I, we obtain two finite covers {Uj}jI\{U^{\prime}_{j}\}_{j\in I} and {Uj}jI\{U_{j}\}_{j\in I} of XX. By construction, we can observe that UjU_{j} are Stein, UjUjU^{\prime}_{j}\Subset U_{j}, and UjU_{j} (and thus UjU^{\prime}_{j} as well) trivializes any flat line bundle on XX.

Let FF be a flat line bundle endowed with a flat Hermitian metric hh, which is unique up to a multiplicative constant (see e.g. [KH22]*Lemma 2.4). Let vv be a smooth ¯\bar{\partial}-closed (0,1)(0,1)-form with values in FF, whose Dolbeault cohomology class [v]H0,1(X,F)[v]\in H^{0,1}(X,F) is trivial. In what follows, all norms are with respect to the fixed flat Hermitian metric hh on FF and the fixed Hermitian metric gg on XX. By applying Hörmander’s estimates and trivializing FF equipped with the metric hh, we find that for each jIj\in I there exists a smooth function uju_{j} on UjU_{j} such that

(1) ¯uj=v|Uj and ujL2(Uj),hCj,g(1)vL2(Uj),h\bar{\partial}u_{j}=v|_{U_{j}}\text{ and }\|u_{j}\|_{L^{2}(U_{j}),h}\leq C^{(1)}_{j,g}\|v\|_{L^{2}(U_{j}),h}

for a constant Cj,g(1)>0C^{(1)}_{j,g}>0, which depends only on UjU_{j} and gg. The key point of this estimate is that it holds uniformly for all flat line bundles (F,h)(F,h) and for all vv representing a trivial Dolbeault class in H0,1(X,F)H^{0,1}(X,F).

We then find that {uiuj}i,jI\{u_{i}-u_{j}\}_{i,j\in I} defines a holomorphic section of FF on Uij:=UiUjU_{ij}:=U_{i}\cap U_{j}. The isomorphism between Čech cohomology and Dolbeault cohomology implies that {uiuj}i,jI\{u_{i}-u_{j}\}_{i,j\in I} represents the Čech cohomology class corresponding to [v][v]. Since [v][v] is trivial, we find that {uiuj}i,jI\{u_{i}-u_{j}\}_{i,j\in I} is a Čech 11-coboundary (after replacing UxU^{*}_{x} with a smaller polydisk if necessary), meaning that there exist local holomorphic sections {fj}jI\{f_{j}\}_{j\in I} of FF such that

fifj=uiujf_{i}-f_{j}=u_{i}-u_{j}

on UijU_{ij} for all i,jIi,j\in I. We take a partition of unity {ρj}jI\{\rho_{j}\}_{j\in I} subordinate to {Uj}jI\{U_{j}\}_{j\in I}. The equation fifj=uiujf_{i}-f_{j}=u_{i}-u_{j} on UijU_{ij} for all i,jIi,j\in I then implies

jIρj(uiuj)=jIρj(fifj)\sum_{j\in I}\rho_{j}(u_{i}-u_{j})=\sum_{j\in I}\rho_{j}(f_{i}-f_{j})

on UiU_{i}. Taking the ¯\bar{\partial}-operator of both sides, we have

¯ui¯(jIρjuj)=¯(jIρjfj)\bar{\partial}u_{i}-\bar{\partial}\left(\sum_{j\in I}\rho_{j}u_{j}\right)=-\bar{\partial}\left(\sum_{j\in I}\rho_{j}f_{j}\right)

on UiU_{i}, as jIρj1\sum_{j\in I}\rho_{j}\equiv 1. Since ¯ui=v|Ui\bar{\partial}u_{i}=v|_{U_{i}}, we get

v|Ui=¯(jIρj(ujfj)|Ui).v|_{U_{i}}=\bar{\partial}\left(\sum_{j\in I}\rho_{j}(u_{j}-f_{j})|_{U_{i}}\right).

Noting that jIρj(ujfj)\sum_{j\in I}\rho_{j}(u_{j}-f_{j}) is defined globally on XX, we have

v=¯(jIρj(ujfj)),v=\bar{\partial}\left(\sum_{j\in I}\rho_{j}(u_{j}-f_{j})\right),

and hence u:=jIρj(ujfj)u:=\sum_{j\in I}\rho_{j}(u_{j}-f_{j}) gives the solution to the equation ¯u=v\bar{\partial}u=v, which is necessarily unique since FF is flat and nontrivial (a well-known result, see e.g. [KH22]*Lemma 2.3 for the proof).

The argument above (except for the uniqueness of solutions) is valid for any (0,q)(0,q)-forms vv with values in F(h)F\otimes\mathcal{I}(h), where hh is a singular Hermitian metric on FF with positive curvature current (see [Mat]*Subsection 5.3 for details). In the following discussion, we essentially use the flatness property.

We apply Ueda’s lemma (Theorem 1.2) to the Čech 0-cochain 𝔣:={(Uj,fj|Uj)}\mathfrak{f}:=\{(U^{\prime}_{j},f_{j}|_{U^{\prime}_{j}})\} corresponding to the smaller cover 𝒰:={Uj}jI\mathcal{U}^{\prime}:=\{U^{\prime}_{j}\}_{j\in I}, to obtain

(2) maxjIsupUj|fj|hK𝒰𝖽(𝕀X,F)maxj,kIsupUjk|fjk|h\max_{j\in I}\sup_{U^{\prime}_{j}}|f_{j}|_{h}\leq\frac{K_{\mathcal{U}^{\prime}}}{\mathsf{d}(\mathbb{I}_{X},F)}\max_{j,k\in I}\sup_{U^{\prime}_{jk}}|f_{jk}|_{h}

uniformly for all F𝒫(X){𝕀X}F\in\mathcal{P}(X)\setminus\{\mathbb{I}_{X}\}, for some constant K𝒰>0K_{\mathcal{U}^{\prime}}>0. Note that we have

(3) fjL2(Uj)Vol(Uj)1/2maxjIsupUj|fj|h\|f_{j}\|_{L^{2}(U^{\prime}_{j})}\leq\mathrm{Vol}(U^{\prime}_{j})^{1/2}\max_{j\in I}\sup_{U^{\prime}_{j}}|f_{j}|_{h}

for each jIj\in I, and

maxj,kIsupUjk|fjk|h=maxj,kIsupUjk|fjfk|h=maxj,kIsupUjk|ujuk|h\max_{j,k\in I}\sup_{U^{\prime}_{jk}}|f_{jk}|_{h}=\max_{j,k\in I}\sup_{U^{\prime}_{jk}}|f_{j}-f_{k}|_{h}=\max_{j,k\in I}\sup_{U^{\prime}_{jk}}|u_{j}-u_{k}|_{h}

since ujuk=fjfku_{j}-u_{k}=f_{j}-f_{k}. Consider a weight φj\varphi_{j} of hh on UjU_{j} such that h=eφjh=e^{-\varphi_{j}}. Since φj\varphi_{j} is pluriharmonic, there exists a holomorphic function wjw_{j} on UjU_{j} such that φj\varphi_{j} is the real part of wjw_{j}, which implies that h=eφj=|ewj|h=e^{-\varphi_{j}}=|e^{-w_{j}}|. Given that (ujuk)ewj(u_{j}-u_{k})e^{-w_{j}} is holomorphic on UjkU_{jk} (which contains UjkU^{\prime}_{jk}), we have

supUjk|ujuk|hCj,g(2)ujukL2(Ujk),h2Cj,g(2)ujL2(Ujk),h\sup_{U^{\prime}_{jk}}|u_{j}-u_{k}|_{h}\leq C^{(2)}_{j,g}\|u_{j}-u_{k}\|_{L^{2}(U_{jk}),h}\leq 2C^{(2)}_{j,g}\|u_{j}\|_{L^{2}(U_{jk}),h}

for some constant Cj,g(2)>0C^{(2)}_{j,g}>0, depending only on UjU^{\prime}_{j}, UjU_{j}, and gg, by the mean value inequality and the Cauchy–Schwarz inequality. Hence

maxj,kIsupUjk|fjk|h=maxj,kIsupUjk|ujuk|h\displaystyle\max_{j,k\in I}\sup_{U^{\prime}_{jk}}|f_{jk}|_{h}=\max_{j,k\in I}\sup_{U^{\prime}_{jk}}|u_{j}-u_{k}|_{h} 2maxj,kICj,g(2)ujL2(Ujk),h\displaystyle\leq 2\max_{j,k\in I}C^{(2)}_{j,g}\|u_{j}\|_{L^{2}(U_{jk}),h}
2maxjICj,g(2)ujL2(Uj),h\displaystyle\leq 2\max_{j\in I}C^{(2)}_{j,g}\|u_{j}\|_{L^{2}(U_{j}),h}
(4) 2maxjICj,g(2)Cj,g(1)vL2(Uj),hCg(3)vL2(X),h\displaystyle\leq 2\max_{j\in I}C^{(2)}_{j,g}C^{(1)}_{j,g}\|v\|_{L^{2}(U_{j}),h}\leq C^{(3)}_{g}\|v\|_{L^{2}(X),h}

by the Hörmander estimate (1) on UjU_{j}, where we set Cg(3):=2maxjCj,g(2)Cj,g(1)>0C^{(3)}_{g}:=2\max_{j}C^{(2)}_{j,g}C^{(1)}_{j,g}>0. With (2) and (3), the estimate (4) gives

(5) fjL2(Uj),hkIfkL2(Uk),hK𝒰Cg(3)𝖽(𝕀X,F)(kIVol(Uk)1/2)vL2(X),h\|f_{j}\|_{L^{2}(U_{j}),h}\leq\sum_{k\in I}\|f_{k}\|_{L^{2}(U^{\prime}_{k}),h}\leq\frac{K_{\mathcal{U}^{\prime}}C^{(3)}_{g}}{\mathsf{d}(\mathbb{I}_{X},F)}\left(\sum_{k\in I}\mathrm{Vol}(U^{\prime}_{k})^{1/2}\right)\|v\|_{L^{2}(X),h}

for any jIj\in I.

We now recall u=jIρj(ujfj)u=\sum_{j\in I}\rho_{j}(u_{j}-f_{j}) and evaluate

uL2(X),hCg(4)jI(ujL2(Uj),h+fjL2(Uj),h)Cg(4)jI(Cj,g(1)vL2(X),h+fjL2(Uj),h)\|u\|_{L^{2}(X),h}\leq C^{(4)}_{g}\sum_{j\in I}\left(\|u_{j}\|_{L^{2}(U_{j}),h}+\|f_{j}\|_{L^{2}(U_{j}),h}\right)\leq C^{(4)}_{g}\sum_{j\in I}\left(C^{(1)}_{j,g}\|v\|_{L^{2}(X),h}+\|f_{j}\|_{L^{2}(U_{j}),h}\right)

again by the Hörmander estimate (1) on UjU_{j}, where Cg(4)>0C^{(4)}_{g}>0 is a constant which depends only on gg. Combined with (5), we thus get

uL2(X),hCg(4)𝖽(𝕀X,F)(jI(Cj,g(1)𝖽(𝕀X,F)+K𝒰Cg(3)kIVol(Uk)1/2))vL2(X),h.\|u\|_{L^{2}(X),h}\leq\frac{C_{g}^{(4)}}{\mathsf{d}(\mathbb{I}_{X},F)}\left(\sum_{j\in I}\left(C^{(1)}_{j,g}\mathsf{d}(\mathbb{I}_{X},F)+K_{\mathcal{U}^{\prime}}C^{(3)}_{g}\sum_{k\in I}\mathrm{Vol}(U^{\prime}_{k})^{1/2}\right)\right)\|v\|_{L^{2}(X),h}.

Noting that supF𝒫(X)𝖽(𝕀X,F)\sup_{F\in\mathcal{P}(X)}\mathsf{d}(\mathbb{I}_{X},F) is finite and II is a finite set, we get the required estimate. ∎

References