betaFeedback

WorksMathematics

On pluriclosed torsion-critical Hermitian metrics

Abstract

In our previous work, we studied the squared L2L^2-norm functionals G\mathcal G of Gauduchon's torsion 11-form and F\mathcal F of the Chern torsion over the space of all Hermitian metrics with unit volume on a compact complex manifold. We introduced torsion-critical metrics as critical points of F\mathcal F. In this paper, We prove that a pluriclosed and astheno-Kähler metric is torsion-critical if and only if it is Kähler. In particular, every compact pluriclosed torsion-critical Hermitian threefold is Kähler. We also give explicit examples and derive geometric consequences under additional curvature assumptions.

Introduction and statement of results

Let (Mn,g)(M^n,g) be a compact Hermitian manifold of complex dimension nn. Denote by ω\omega its Kähler form, ∇\nabla its Chern connection, and TT the torsion of ∇\nabla. The Gauduchon torsion 1-form η\eta, defined as the trace of the torsion tensor TT, is the global (1,0)(1,0)-form η\eta on MnM^n given by

∂ωn−1=−η∧ωn−1.(1)\partial\omega^{n-1}=-\eta\wedge\omega^{n-1}. \tag*{(1)}

For n≥2n\ge2, the condition η=0\eta=0 is equivalent to balancedness, that is, dωn−1=0\mathrm{d}\omega^{n-1}=0. If the torsion TT vanishes, then the metric gg is Kähler, that is, dω=0\mathrm{d}\omega=0. Gauduchon initiated the variational study of the torsion 1-form in [14]. He considered the functional

G(g)=V1−nn∫M∣η∣2 dv,g∈HM,(2)\mathcal{G}(g)=V^{\frac{1-n}{n}}\int_M|\eta|^2\,\mathrm{d}v,\qquad g\in\mathcal{H}_M, \tag*{(2)}

he derived its Euler–Lagrange equation, and proved that its critical metrics on compact complex surfaces are balanced.

In a previous work [29 Proposition 1], we extended this conclusion to all complex dimensions n≥2n\ge2. In the same work, we introduced torsion-critical metrics which are Hermitian metrics on a compact complex manifold MnM^n that are critical points of the functional of L2L^2-norm of the Chern torsion tensor TT:

F(g)=V1−nn∫M∣T∣2 dv,g∈HM.(3)\mathcal{F}(g)=V^{\frac{1-n}{n}}\int_M|T|^2\,\mathrm{d}v,\qquad g\in\mathcal{H}_M. \tag*{(3)}

Throughout this paper, we use their normalization ∣T∣2=∑i,j,k∣Tjki∣2\lvert T\rvert^2=\sum_{i,j,k}\lvert T^i_{jk}\rvert^2. Where HM\mathcal{H}_M is the set of all Hermitian metrics on the complex manifold MnM^n, dv=ωnn!dv=\frac{\omega^n}{n!} the volume form, and V=∫MdvV=\int_Mdv the volume. Note that the factor in front of the integral is simply to make the functional independent of the scaling of metric (by constant multiples). Equivalently, one can drop this VV factor but restrict to the subset of all Hermitian metrics with unit volume.

One may also consider the squared norm of the Bismut torsion. Let TBT^{\mathrm{B}} denote the torsion tensor of the Bismut connection. The Bismut connection is the unique Hermitian connection whose torsion is totally skew-symmetric [3, 15], for flatness questions involving Hermitian connections, see [26, 27]. Its torsion three-form is dcωd^c\omega up to sign, where dc=−1(∂ˉ−∂)d^c=\sqrt{-1}(\bar{\partial}-\partial); the sign depends on the convention for the fundamental form. Hence, with the normalization of the Chern torsion used in this paper,

∣TB∣2=6∣dω∣2=12∣∂ω∣2=6∣T∣2.\lvert T^{\mathrm{B}}\rvert^2=6\lvert d\omega\rvert^2=12\lvert\partial\omega\rvert^2=6\lvert T\rvert^2.

It follows that

V1−nn∫M∣TB∣2 dv=6F(g)V^{\frac{1-n}{n}}\int_M\lvert T^{\mathrm{B}}\rvert^2\,dv=6\mathcal{F}(g)

for every g∈HMg\in\mathcal{H}_M. Thus the two functionals have the same critical points with respect to all Hermitian variations.

The critical metrics of G\mathcal{G} are all balanced, and hence all its critical values are zero. In contrast, F\mathcal{F} admits non-Kähler critical metrics in addition to Kähler critical metrics, at which its value is zero. Zhang and Zheng constructed non-Kähler examples and proved several non-existence results in [29]. Their subsequent work [30] addresses the compact Chern flat case: the complex Lie group arising as the universal cover of a non-Kähler Chern flat torsion-critical manifold must be semisimple. Conversely, canonical metrics on semisimple complex Lie groups yield torsion-critical metrics on compact quotients whenever such quotients exist. Thus, torsion-criticality alone does not imply the Kähler condition. Related variational problems within conformal classes were studied by Angella, Istrati, Otiman, and Tardini [2], while Kawamura [18] considered a torsion variational problem in the almost Hermitian setting. In this paper, we fix an integrable complex structure and consider general variations of compatible Hermitian metrics.

One of the questions raised in [29] concerns torsion-critical metrics satisfying the pluriclosed condition

∂∂ˉω=0.(4)\partial\bar{\partial}\omega=0. \tag*{(4)}

Such metrics are also called strong Kähler with torsion, or SKT. They arise naturally in the theory of Hermitian connections with torsion; see [3, 15, 25] and the survey [10]. Non-Kähler pluriclosed metrics on complex three-dimensional nilmanifolds were studied in [9]. Our three-dimensional result implies that such metrics cannot be torsion-critical. The relation with symplectic forms taming the complex structure was investigated in [7]. Other restrictions on the existence of non-Kähler pluriclosed metrics appear in [6]. The pluriclosed condition is also preserved by the geometric flow introduced by Streets and Tian [24].

We study this question under the additional astheno-Kähler condition

∂∂ˉωn−2=0,n≥3.(5)\partial\bar{\partial}\omega^{n-2}=0,\qquad n\geq3. \tag*{(5)}

Introduced by Jost and Yau [17], this condition has been studied from the perspectives of both construction and rigidity; see, for example, [8, 11]. Related conditions on powers of the fundamental form occur in the study of generalized Gauduchon metrics [12]. The pluriclosed and astheno-Kähler conditions coincide in complex dimension three but are distinct in higher dimensions. In this paper, simultaneous conditions are imposed on the same metric.

Our argument uses balancedness as an intermediate step. Balanced metrics were introduced systematically by Michelsohn [20], and their interaction with other torsion conditions is known to be an important source of Kähler rigidity; see [1, 13]. Note, however, that the existence of a single metric satisfying two conditions is a strictly stronger requirement than the existence of a manifold admitting different metrics that satisfy them separately. In fact, [8, 19] give examples of non-Kähler compact complex manifolds that admit both a balanced and an astheno-Kähler metric.

Our main result gives a partial answer to the pluriclosed problem posed in [29].

Theorem 1.1. Let (Mn,g)(M^n,g) be a compact Hermitian manifold, with n≥3n \ge3. Suppose that gg is both pluriclosed and astheno-Kähler. Then gg is torsion-critical if and only if it is Kähler.

In particular, in the complex three-dimensional case, we obtain the following consequence:

Corollary 1.2. On a compact complex threefold, a Hermitian metric is pluriclosed and torsion-critical if and only if it is Kähler.

In complex dimension two, the identity F=2G\mathcal{F}=2\mathcal{G} under the above normalization and Gauduchon’s theorem imply that every torsion-critical metric is Kähler. In complex dimension one, every Hermitian metric is Kähler. Thus, the characterization in Corollary 1.2 holds in all complex dimensions at most three. In contrast to the Chern-flat results of [30], Theorem 1.1 requires no curvature or homogeneity assumptions.

The proof rests on the identity

F(g)−2G(g)=2V1−nn(n−2)!∫M−1∂∂ˉω∧ωn−2,(6)\mathcal{F}(g)-2\mathcal{G}(g)=\frac{2V^{\frac{1-n}{n}}}{(n-2)!}\int_M\sqrt{-1}\partial\bar{\partial}\omega\wedge\omega^{n-2}, \tag*{(6)}

which holds for every Hermitian metric when n≥3n \ge3. For a smooth curve gt∈HMg_t\in\mathcal{H}_M, write

DFg(γ)=ddt∣t=0F(gt),D\mathcal{F}_g(\gamma)=\left.\frac{d}{dt}\right|_{t=0}\mathcal{F}(g_t),

and define DGg(γ)D\mathcal{G}_g(\gamma) similarly. Here γ\gamma is an arbitrary smooth real (1,1)(1,1)-form. At a pluriclosed metric, differentiating (6) and integrating by parts gives

(DFg−2DGg)(γ)=2V1−nn(n−2)!∫Mγ∧−1∂∂ˉωn−2.(7)\left(D\mathcal{F}_g-2D\mathcal{G}_g\right)(\gamma)=\frac{2V^{\frac{1-n}{n}}}{(n-2)!}\int_M\gamma\wedge\sqrt{-1}\partial\bar{\partial}\omega^{n-2}. \tag*{(7)}

The astheno-Kähler condition therefore yields DFg=2DGgD\mathcal{F}_g=2D\mathcal{G}_g. If gg is torsion-critical, it is consequently a critical point of G\mathcal{G} and hence balanced by [29 Proposition 1]. Equation (6) then implies F(g)=0\mathcal{F}(g)=0, so gg is Kähler. The converse follows because every Kähler metric minimizes the nonnegative functional F\mathcal{F}.

For a pluriclosed metric, the identity F=2G\mathcal{F}=2\mathcal{G} does not by itself imply DF=2DGD\mathcal{F}=2D\mathcal{G}. The astheno-Kähler condition ensures equality of these first variations. Without this additional condition, the pluriclosed torsion-critical problem remains unresolved by our argument in complex dimensions n≥4n \ge4.

The paper is organized as follows. In Section 2, we prove Theorem 1.1 and derive several consequences. In Section 3, we give some examples illustrating the theorem, and discuss the role of the astheno-Kähler condition. In Section 4, we apply the main theorem to obtain geometric consequences under Chern flatness or positive Chern holomorphic bisectional curvature.

Proof of Theorem 1.1

Throughout this section, let (Mn,g)(M^n,g) be a compact Hermitian manifold. We use the same notation gg for its complex bilinear extension to the complexified tangent bundle. Let ∇\nabla be the Chern connection, with torsion defined by

T(X,Y)=∇XY−∇YX−[X,Y].T(X,Y)=\nabla_XY-\nabla_YX-[X,Y].

The Chern torsion has no mixed component: T(X,Y‾)=0T(X,\overline{Y})=0 for X,Y∈T1,0MX,Y\in T^{1,0}M.

In local holomorphic coordinates (z1,…,zn)(z_1,\ldots,z_n), set

∂i=∂∂zi,gijˉ=g(∂i,∂j‾),\partial_i=\frac{\partial}{\partial z_i},\qquad g_{i\bar{j}}=g(\partial_i,\overline{\partial_j}),

and write (gji‾)(\overline{g^{ji}}) for the inverse of (gijˉ)(g_{i\bar j}), so that

∑jgijˉgjk‾=δik.\sum_j g_{i\bar j}\overline{g^{jk}}=\delta_i^k.

We adopt the convention

∇∂k∂i=∑jΓikj∂j,Γikj=∑ℓgjℓ‾∂kgiℓˉ.\nabla_{\partial_k}\partial_i=\sum_j\Gamma_{ik}^j\partial_j,\qquad\Gamma_{ik}^j=\sum_\ell\overline{g^{j\ell}}\partial_k g_{i\bar\ell}.

Consequently, the coefficients in T(∂i,∂k)=∑jTikj∂jT(\partial_i,\partial_k)=\sum_jT_{ik}^j\partial_j satisfy

Tikj=Γkij−Γikj=∑ℓgjℓ‾(∂igkℓˉ−∂kgiℓˉ).(8)T_{ik}^j=\Gamma_{ki}^j-\Gamma_{ik}^j=\sum_\ell\overline{g^{j\ell}}(\partial_i g_{k\bar\ell}-\partial_k g_{i\bar\ell}). \tag*{(8)}

The fundamental form of gg is

ω=−1∑i,jgij‾ dzi∧dzj‾.\omega=\sqrt{-1}\sum_{i,j}\overline{g^{ij}}\,dz_i\wedge d\overline{z_j}.

With the sign convention in (1), Gauduchon’s torsion 1-form is η=∑iηi dzi\eta=\sum_i\eta_i\,dz_i, where

ηi=∑kTikk=∑k(Γikk−Γkik).(9)\eta_i=\sum_kT_{ik}^k=\sum_k(\Gamma_{ik}^k-\Gamma_{ki}^k). \tag*{(9)}

For pointwise computations, choose a local unitary frame {ei}i=1n\{e_i\}_{i=1}^n of T1,0MT^{1,0}M with dual coframe {φi}i=1n\{\varphi_i\}_{i=1}^n. We use the same component notation relative to the chosen frame:

ω=−1∑iφi∧φi‾,T(ei,ek)=∑jTikjej,η=∑i(∑kTkik)φi.(10)\omega=\sqrt{-1}\sum_i\varphi_i\wedge\overline{\varphi_i},\qquad T(e_i,e_k)=\sum_jT_{ik}^je_j,\qquad\eta=\sum_i\left(\sum_kT_{ki}^k\right)\varphi_i. \tag*{(10)}

We use the complex orientation and set dv=ωn/n!dv=\omega^n/n!. Norms of differential forms are induced by gg, and the Hodge star is extended complex linearly. For the Chern torsion, we use the normalization

∣T∣2=∑i,k,j∣Tikj∣2.|T|^2=\sum_{i,k,j}|T_{ik}^j|^2.

Since Tikj=−TkijT_{ik}^j=-T_{ki}^j, the Chern structure equations yield

∂ω=−1∑i<k∑jTikjφi∧φk∧φj‾,∣∂ω∣2=∑i<k∑j∣Tikj∣2=12∣T∣2.(11)\partial\omega=\sqrt{-1}\sum_{i<k}\sum_jT_{ik}^j\varphi_i\wedge\varphi_k\wedge\overline{\varphi_j},\qquad|\partial\omega|^2=\sum_{i<k}\sum_j|T_{ik}^j|^2=\frac{1}{2}|T|^2. \tag*{(11)}

Lemma 2.1. On a Hermitian manifold of complex dimension n≥3n\ge3,

−1∂ω∧∂ω‾∧ωn−3(n−3)!=(∣η∣2−∣∂ω∣2) dv.(12)\frac{\sqrt{-1}\partial\omega\wedge\overline{\partial\omega}\wedge\omega^{n-3}}{(n-3)!}=(|\eta|^2-|\partial\omega|^2)\,dv. \tag*{(12)}

Proof. The argument is pointwise. Put α=∂ω\alpha=\partial\omega. On a Hermitian vector space, the Lefschetz decomposition gives

α=α0+β∧ω,α0∧ωn−2=0,(13)\alpha=\alpha_0+\beta\wedge\omega,\qquad\alpha_0\wedge\omega^{n-2}=0, \tag*{(13)}

where α0\alpha_0 is a primitive (2,1)(2,1)-form and β\beta is a (1,0)(1,0)-form. First take ∂\partial of ωn−1\omega^{n-1}. Since ω\omega has total degree 2, no minus sign appears in the product rule:

∂ωn−1=(n−1)∂ω∧ωn−2.\partial\omega^{n-1}=(n-1)\partial\omega\wedge\omega^{n-2}.

By (1),

(n−1)α∧ωn−2=−η∧ωn−1.(n-1)\alpha\wedge\omega^{n-2}=-\eta\wedge\omega^{n-1}.

Wedging 1-forms with ωn−1\omega^{n-1} is injective. Substitution of (13) therefore gives β=−η/(n−1)\beta=-\eta/(n-1). Since the Lefschetz decomposition is orthogonal and ∣β∧ω∣2=(n−1)∣β∣2|\beta\wedge\omega|^2=(n-1)|\beta|^2, we obtain

∣α∣2=∣α0∣2+∣η∣2n−1.(14)|\alpha|^2=|\alpha_0|^2+\frac{|\eta|^2}{n-1}. \tag*{(14)}

:::`

The Hodge star operator ∗* is the complex linear extension of the real Hodge star operator, and therefore

u∧∗uˉ=∣u∣2dv.u \wedge*\bar{u} = |u|^2dv.

For clarity, with our conventions the Hodge-star formula for a primitive (p,r)(p,r)-form uu of degree k=p+rk=p+r is

∗u=(−1)k(k+1)/2(−1)p−ru∧ωn−k(n−k)!.*u = (-1)^{k(k+1)/2}(\sqrt{-1})^{p-r}\frac{u\wedge\omega^{n-k}}{(n-k)!}.

Applying this to αˉ0\bar{\alpha}_0, and using α0∧∗αˉ0=∣α0∣2dv\alpha_0\wedge*\bar{\alpha}_0=|\alpha_0|^2dv, yields

−1α0∧αˉ0∧ωn−3(n−3)!=−∣α0∣2dv.(15)\frac{\sqrt{-1}\alpha_0\wedge\bar{\alpha}_0\wedge\omega^{n-3}}{(n-3)!}=-|\alpha_0|^2dv. \tag*{(15)}

Also, a direct calculation in a unitary coframe gives

−1η∧ηˉ∧ωn−1(n−1)!=∣η∣2dv.\frac{\sqrt{-1}\eta\wedge\bar{\eta}\wedge\omega^{n-1}}{(n-1)!}=|\eta|^2dv.

Since ω\omega is real, the primitivity of α0\alpha_0 gives

α0∧ωn−2=0,αˉ0∧ωn−2=0.\alpha_0\wedge\omega^{n-2}=0,\qquad\bar{\alpha}_0\wedge\omega^{n-2}=0.

Thus the mixed terms in the expansion of α∧αˉ∧ωn−3\alpha\wedge\bar{\alpha}\wedge\omega^{n-3} vanish, and hence

α∧αˉ∧ωn−3=α0∧αˉ0∧ωn−3+β∧βˉ∧ωn−1.\alpha\wedge\bar{\alpha}\wedge\omega^{n-3}=\alpha_0\wedge\bar{\alpha}_0\wedge\omega^{n-3}+\beta\wedge\bar{\beta}\wedge\omega^{n-1}.

For the third term, take the complex conjugate of (13) and use ωˉ=ω\bar{\omega}=\omega, giving αˉ0∧ωn−2=0\bar{\alpha}_0\wedge\omega^{n-2}=0. Therefore

β∧αˉ0∧ωn−2=0.\beta\wedge\bar{\alpha}_0\wedge\omega^{n-2}=0.

Consequently,

−1α∧αˉ∧ωn−3(n−3)!=−1α0∧αˉ0∧ωn−3(n−3)!+−1β∧βˉ∧ωn−1(n−3)!=(−∣α0∣2+(n−1)(n−2)(n−1)2∣η∣2)dv=(∣η∣2−∣α∣2)dv,\begin{aligned} \frac{\sqrt{-1}\alpha\wedge\bar{\alpha}\wedge\omega^{n-3}}{(n-3)!} &=\frac{\sqrt{-1}\alpha_0\wedge\bar{\alpha}_0\wedge\omega^{n-3}}{(n-3)!}+\frac{\sqrt{-1}\beta\wedge\bar{\beta}\wedge\omega^{n-1}}{(n-3)!}\\ &=\left(-|\alpha_0|^2+\frac{(n-1)(n-2)}{(n-1)^2}|\eta|^2\right)dv\\ &=\left(|\eta|^2-|\alpha|^2\right)dv, \end{aligned}

where the last equality follows from (14). Since αˉ=∂ˉω\bar{\alpha}=\bar{\partial}\omega, this is (12). □

Proposition 2.2. Let (Mn,g)(M^n,g) be a compact Hermitian manifold, with n≥3n\geq3. Then

∫M−1∂∂ˉω∧ωn−2=(n−2)!∫M(12∣T∣2−∣η∣2)dv.(16)\int_M\sqrt{-1}\partial\bar{\partial}\omega\wedge\omega^{n-2}=(n-2)!\int_M\left(\frac{1}{2}|T|^2-|\eta|^2\right)dv. \tag*{(16)}

In particular, (6) holds, and every pluriclosed metric satisfies F(g)=2G(g)F(g)=2G(g).

Proof. The form ∂ˉω\bar{\partial}\omega has odd total degree. The product rule, followed by interchange of the two three-forms, gives

∂(∂ˉω∧ωn−2)=∂∂ˉω∧ωn−2+(n−2)∂ω∧∂ˉω∧ωn−3.\partial(\bar{\partial}\omega\wedge\omega^{n-2})=\partial\bar{\partial}\omega\wedge\omega^{n-2}+(n-2)\partial\omega\wedge\bar{\partial}\omega\wedge\omega^{n-3}.

The form ∂ˉω∧ωn−2\bar{\partial}\omega\wedge\omega^{n-2} has type (n−1,n)(n-1,n), so its ∂ˉ\bar{\partial}-derivative vanishes. Its ∂\partial-derivative therefore equals its exterior derivative, whose integral is zero by Stokes’ theorem. Multiplying by −1\sqrt{-1} and applying Lemma 2.1, we obtain

∫M−1∂∂ˉω∧ωn−2=−(n−2)∫M−1∂ω∧∂ˉω∧ωn−3=(n−2)!∫M(∣∂ω∣2−∣η∣2)dv.\begin{aligned} \int_M\sqrt{-1}\partial\bar{\partial}\omega\wedge\omega^{n-2} &=-(n-2)\int_M\sqrt{-1}\partial\omega\wedge\bar{\partial}\omega\wedge\omega^{n-3}\\ &=(n-2)!\int_M\left(|\partial\omega|^2-|\eta|^2\right)dv. \end{aligned}

Equation (11) proves (16). Multiplication by 2V1−nn/(n−2)!2V^{\frac{1-n}{n}}/(n-2)! proves (6). Its right-hand side vanishes if gg is pluriclosed. □

Remark 2.3. In complex dimension two, (1) reduces to

∂ω=−η∧ω.\partial\omega=-\eta\wedge\omega.

Since the map β↦β∧ω\beta\mapsto\beta\wedge\omega is an isometry on (1,0)(1,0)-forms in this dimension, (11) gives

∣T∣2=2∣∂ω∣2=2∣η∣2.|T|^2=2|\partial\omega|^2=2|\eta|^2.

Thus F=2G\mathcal{F}=2\mathcal{G} on the entire space HM\mathcal{H}_M, without any pluriclosedness assumption.

Let γ\gamma be an arbitrary smooth real (1,1)(1,1)-form and set ωt=ω+tγ\omega_t=\omega+t\gamma. By compactness, ωt\omega_t is positive for sufficiently small ∣t∣|t| and hence determines a smooth family of Hermitian metrics gt∈HMg_t\in\mathcal{H}_M. No differential constraint is imposed on γ\gamma. Since the complex structure is fixed,

ddt∣t=0∂∂‾ωt=∂∂‾γ,ddt∣t=0ωtn−2=(n−2)γ∧ωn−3.\left.\frac{d}{dt}\right|_{t=0}\partial\overline{\partial}\omega_t=\partial\overline{\partial}\gamma,\qquad \left.\frac{d}{dt}\right|_{t=0}\omega_t^{n-2}=(n-2)\gamma\wedge\omega^{n-3}.

Lemma 2.4. For a real (1,1)(1,1)-form γ\gamma and a smooth (n−2,n−2)(n-2,n-2)-form Ω\Omega on MM,

∫M−1∂∂‾γ∧Ω=∫Mγ∧−1∂∂‾Ω.(17)\int_M\sqrt{-1}\partial\overline{\partial}\gamma\wedge\Omega=\int_M\gamma\wedge\sqrt{-1}\partial\overline{\partial}\Omega. \tag*{(17)}

Proof. Two applications of Stokes’ theorem give

0=∫M∂(∂‾γ∧Ω)=∫M∂∂‾γ∧Ω−∫M∂‾γ∧∂Ω,0=\int_M\partial(\overline{\partial}\gamma\wedge\Omega)=\int_M\partial\overline{\partial}\gamma\wedge\Omega-\int_M\overline{\partial}\gamma\wedge\partial\Omega,
0=∫M∂‾(γ∧∂Ω)=∫M∂‾γ∧∂Ω−∫Mγ∧∂‾∂Ω.0=\int_M\overline{\partial}(\gamma\wedge\partial\Omega)=\int_M\overline{\partial}\gamma\wedge\partial\Omega-\int_M\gamma\wedge\overline{\partial}\partial\Omega.

In the second line we used ∂‾∂=−∂∂‾\overline{\partial}\partial=-\partial\overline{\partial}. Combining the two equations and multiplying by −1\sqrt{-1} proves the formula. □

Proposition 2.5. For every g∈HMg\in\mathcal{H}_M of complex dimension n≥3n\geq3, one has

(DFg−2DGg)(γ)=2V1−nn(n−2)!{∫Mγ∧−1∂∂‾ωn−2+(n−2)∫M−1∂∂‾ω∧γ∧ωn−3+1−nn∫M−1∂∂‾ω∧ωn−2V∫Mγ∧ωn−1(n−1)!}.(18)(D\mathcal{F}_g-2D\mathcal{G}_g)(\gamma)=\frac{2V^{\frac{1-n}{n}}}{(n-2)!}\left\{\int_M\gamma\wedge\sqrt{-1}\partial\overline{\partial}\omega^{n-2}+(n-2)\int_M\sqrt{-1}\partial\overline{\partial}\omega\wedge\gamma\wedge\omega^{n-3}+\frac{\frac{1-n}{n}\int_M\sqrt{-1}\partial\overline{\partial}\omega\wedge\omega^{n-2}}{V}\int_M\frac{\gamma\wedge\omega^{n-1}}{(n-1)!}\right\}. \tag*{(18)}

If gg is pluriclosed, this reduces to

(DFg−2DGg)(γ)=2V1−nn(n−2)!∫Mγ∧−1∂∂‾ωn−2.(19)(D\mathcal{F}_g-2D\mathcal{G}_g)(\gamma)=\frac{2V^{\frac{1-n}{n}}}{(n-2)!}\int_M\gamma\wedge\sqrt{-1}\partial\overline{\partial}\omega^{n-2}. \tag*{(19)}

Proof. Let ωt=ω+tγ\omega_t=\omega+t\gamma, and let

Vt=∫Mωtnn!.V_t=\int_M\frac{\omega_t^n}{n!}.

Since the complex structure is fixed, differentiation gives

ddt∣t=0∫M−1∂∂‾ωt∧ωtn−2=∫M−1∂∂‾γ∧ωn−2+(n−2)∫M−1∂∂‾ω∧γ∧ωn−3.\left.\frac{d}{dt}\right|_{t=0}\int_M\sqrt{-1}\partial\overline{\partial}\omega_t\wedge\omega_t^{n-2} =\int_M\sqrt{-1}\partial\overline{\partial}\gamma\wedge\omega^{n-2}+(n-2)\int_M\sqrt{-1}\partial\overline{\partial}\omega\wedge\gamma\wedge\omega^{n-3}.

By Lemma 2.4, the first integral on the right is ∫Mγ∧−1∂∂‾ωn−2\int_M\gamma\wedge\sqrt{-1}\partial\overline{\partial}\omega^{n-2}. Moreover,

ddt∣t=0Vt=∫Mγ∧ωn−1(n−1)!.\left.\frac{d}{dt}\right|_{t=0}V_t=\int_M\frac{\gamma\wedge\omega^{n-1}}{(n-1)!}.

Differentiating (6) and applying the product rule now gives (18).

If gg is pluriclosed, then ∂∂‾ω=0\partial\overline{\partial}\omega=0. The second and third terms in (18) vanish, and (19) follows. □

Proposition 2.6. Let gg be a pluriclosed metric on a compact complex manifold of dimension n≥3n \ge3. The following statements are equivalent:

  1. gg is astheno-Kähler;

  2. DFg=2DGgDF_g = 2DG_g as linear functionals on all smooth real (1,1)(1,1)-forms;

  3. gg is a critical point of F−2G\mathcal{F} - 2\mathcal{G} on HM\mathcal{H}_M.

Proof. The implication (1)⇒\Rightarrow(2) follows from (19), and (2) and (3) are equivalent by definition. To prove (2) ⇒\Rightarrow (1), set

Ψ=−1∂∂ˉωn−2.\Psi= \sqrt{-1}\partial\bar\partial\omega^{n-2}.

Since ω\omega is real and ∂∂ˉ=−∂ˉ∂\partial\bar\partial= -\bar\partial\partial, the form Ψ\Psi is real and of type (n−1,n−1)(n-1,n-1). Its Hodge dual ∗Ψ*\Psi is a smooth real (1,1)(1,1)-form and is an admissible variation direction. Under condition (2), equation (19), applied to γ=∗Ψ\gamma= *\Psi, gives

0=∫M∗Ψ∧Ψ=∫M∣Ψ∣2 dv.0 = \int_M *\Psi\wedge\Psi= \int_M |\Psi|^2\,dv.

Here the middle equality follows because both forms have even degree. Hence Ψ=0\Psi= 0, so gg is astheno-Kähler. □

Remark 2.7. The equivalence concerns equality of the full differentials. A variation that remains pluriclosed satisfies ∂∂ˉγ=0\partial\bar\partial\gamma= 0 to first order. Equality of derivatives only for such directions is weaker and cannot be used to conclude DGg=0DG_g = 0 on HM\mathcal{H}_M. Proposition 2.5 avoids this issue by computing at an arbitrary base metric first.

We use the following theorem, and follow the conventions established above.

Theorem 2.8 (Zhang–Zheng [29 Proposition 1]). On a compact complex manifold of complex dimension n≥2n \ge2, a Hermitian metric is a critical point of G\mathcal{G} on HM\mathcal{H}_M if and only if it is balanced.

The hard part of the theorem is proving that critical metrics are balanced. Conversely, a balanced metric satisfies G(g)=0\mathcal{G}(g) = 0 and hence is an absolute minimum of the nonnegative functional G\mathcal{G}. To apply this theorem, we need DGg(γ)=0D\mathcal{G}_g(\gamma) = 0 for every smooth real (1,1)(1,1)-form γ\gamma. This is why (19) is established for arbitrary Hermitian variations.

Proof of Theorem 1.1. Suppose first that gg is torsion-critical, pluriclosed, and astheno-Kähler. For any smooth real (1,1)(1,1)-form γ\gamma, Proposition 2.5 gives

0=DFg(γ)=2DGg(γ).0 = D\mathcal{F}_g(\gamma) = 2D\mathcal{G}_g(\gamma).

Hence gg is a critical point of G\mathcal{G} on HM\mathcal{H}_M. Theorem 2.8 implies η=0\eta= 0. In particular, G(g)=0\mathcal{G}(g) = 0. By Proposition 2.2 and pluriclosedness,

F(g)=2G(g)=0.\mathcal{F}(g) = 2\mathcal{G}(g) = 0.

The function ∣T∣2|T|^2 is smooth and nonnegative, and V>0V > 0, so TT vanishes identically. Equation (11) gives ∂ω=0\partial\omega= 0; since ω\omega is real, also ∂ˉω=0\bar\partial\omega= 0. Thus dω=0d\omega= 0 and gg is Kähler.

Conversely, a Kähler metric satisfies both differential conditions. It also satisfies F(g)=0\mathcal{F}(g) = 0 and therefore is an absolute minimum of the nonnegative smooth functional F\mathcal{F}. Its differential vanishes in every direction, so it is torsion-critical. □

Proof of Corollary 1.2. When n=3n = 3, one has ωn−2=ω\omega^{n-2} = \omega. Thus pluriclosedness and the astheno-Kähler condition coincide, and Theorem 1.1 applies. □

Corollary 2.9. Every compact pluriclosed torsion-critical Hermitian manifold of complex dimension at most three is Kähler. In complex dimension two, the pluriclosed assumption is unnecessary.

Proof. The case n=3n = 3 follows from Corollary 1.2. For n=2n = 2, Remark 2.3 gives F=2G\mathcal{F} = 2\mathcal{G}, so Theorem 2.8 implies that every torsion-critical metric is balanced, hence Kähler. The case n=1n = 1 is automatic. □

We also recover the following well-known consequence.

Corollary 2.10. Let (Mn,g)(M^n,g) be a compact Hermitian manifold. If gg is both balanced and pluriclosed, then it is Kähler.

Proof. For n≥3n \ge3, balancedness gives G(g)=0\mathcal{G}(g)=0, and Proposition 2.2 gives F(g)=2G(g)=0\mathcal{F}(g)=2\mathcal{G}(g)=0. Thus T=0T=0. Dimensions one and two are immediate as in Corollary 2.9. □

Let ∇B\nabla^{\mathrm{B}} denote the Bismut connection (or Strominger connection), we call gg Strominger torsion parallel (STP) if ∇BT=0\nabla^{\mathrm{B}}T=0, where TT is the Chern torsion; see [29]. This convention also appears in work relating the Bismut connection to pluriclosed geometry; see [31].

Corollary 2.11. Let (Mn,g)(M^n,g) be a compact Hermitian manifold. If gg is pluriclosed, torsion-critical, and STP, then it is Kähler.

Proof. By [29], Proposition 3, every compact torsion-critical STP metric is balanced, Corollary 2.10 implies that it is Kähler. □

Remark 2.12. The argument for Corollary 2.11 uses the STP theorem of [29] rather than Theorem 1.1; it does not assume that every pluriclosed STP metric is astheno-Kähler. Likewise, the non-Kähler compact Chern-flat critical examples in [30] do not contradict the main theorem. Compact Chern-flat metrics are balanced, as recalled in that work using [28], so Corollary 2.10 already excludes pluriclosedness for those non-Kähler metrics.

Examples and the astheno-Kähler condition

As shown in the introduction, the L2L^2-norm functional of the Bismut torsion equals 6F6\mathcal{F} and therefore has the same critical metrics. We now give some examples and discuss the role of the astheno-Kähler condition.

Example 3.1 (A flat complex three-torus). Let M=C3/(Z+−1Z)3M=\mathbb{C}^3/(\mathbb{Z}+\sqrt{-1}\mathbb{Z})^3 and write zj=xj+−1yjz_j=x_j+\sqrt{-1}y_j. The Euclidean metric and its fundamental form are

g0=∑j=13(dxj2+dyj2),ω0=−12∑j=13dzj∧dzˉj.g_0=\sum_{j=1}^{3}(dx_j^2+dy_j^2), \qquad\omega_0=\frac{\sqrt{-1}}{2}\sum_{j=1}^{3} dz_j\wedge d\bar{z}_j.

Both descend to MM, and dω0=0d\omega_0=0. Thus g0g_0 is Kähler and hence pluriclosed. Since F(g0)=0\mathcal{F}(g_0)=0 and F≥0\mathcal{F}\ge0, g0g_0 is torsion-critical. This is an example of Corollary 1.2.

Example 3.2 (Complex projective space). For every n≥4n\ge4, consider CPn\mathbb{CP}^n with the Fubini–Study form, written in an affine chart as

ωFS=−1∂∂ˉlog⁡(1+∑j=1n∣zj∣2).\omega_{\mathrm{FS}}=\sqrt{-1}\partial\bar{\partial}\log\left(1+\sum_{j=1}^{n}|z_j|^2\right).

These local expressions define a global positive closed real (1,1)(1,1)-form. Thus ∂∂ˉωFS=0\partial\bar{\partial}\omega_{\mathrm{FS}}=0 and ∂∂ˉωFSn−2=0\partial\bar{\partial}\omega_{\mathrm{FS}}^{n-2}=0. The Chern torsion vanishes, so the metric is an absolute minimum of F\mathcal{F} on HM\mathcal{H}_M. It therefore satisfies all the hypotheses of Theorem 1.1. This also shows that the theorem does not require flatness: the Fubini–Study metric has positive holomorphic bisectional curvature.

The next example shows why torsion-criticality cannot be omitted. We use the usual description of a primary Kodaira surface as a nilmanifold; see [16, 22] for the nilmanifold setting.

Example 3.3 (A primary Kodaira surface times a flat complex torus). For every n>3n>3, we construct a non-Kähler Hermitian metric on a compact complex nn-manifold that is both pluriclosed and astheno-Kähler. We also verify directly that it is not torsion-critical. To construct it, let G=R4G=\mathbb{R}^{4} with multiplication

(x,y,z,t)(x′,y′,z′,t′)=(x+x′,y+y′,z+z′+xy′,t+t′).(x,y,z,t)(x',y',z',t')=(x+x',y+y',z+z'+xy',t+t').

The integer points form a discrete cocompact subgroup Γ\Gamma. Indeed, left multiplication by integer points reduces x,y,tx,y,t to bounded fundamental intervals and then does the same for zz. On S=Γ\GS=\Gamma\backslash G, the left-invariant forms

e1=dx,e2=dy,e3=dz−x dy,e4=dte^{1}=dx,\qquad e^{2}=dy,\qquad e^{3}=dz-x\,dy,\qquad e^{4}=dt

are globally defined and satisfy

de1=de2=de4=0,de3=−e1∧e2.de^{1}=de^{2}=de^{4}=0,\qquad de^{3}=-e^{1}\wedge e^{2}.

Let φ1=e1+−1e2\varphi^{1}=e^{1}+\sqrt{-1}e^{2} and φ2=e3+−1e4\varphi^{2}=e^{3}+\sqrt{-1}e^{4} be of type (1,0)(1,0). Then

dφ1=0,dφ2=−−12φ1∧φ1‾.(20)d\varphi^{1}=0,\qquad d\varphi^{2}=-\frac{\sqrt{-1}}{2}\varphi^{1}\wedge\overline{\varphi^{1}}. \tag*{(20)}

There are no (0,2)(0,2) terms, so the complex structure is integrable. Take

gS=∑a=14ea⊗ea,α=−12(φ1∧φ1‾+φ2∧φ2‾).g_{S}=\sum_{a=1}^{4}e^{a}\otimes e^{a},\qquad\alpha=\frac{\sqrt{-1}}{2}\left(\varphi^{1}\wedge\overline{\varphi^{1}}+\varphi^{2}\wedge\overline{\varphi^{2}}\right).

Since α=e1∧e2+e3∧e4\alpha=e^{1}\wedge e^{2}+e^{3}\wedge e^{4}, we have

dα=−e1∧e2∧e4≠0.d\alpha=-e^{1}\wedge e^{2}\wedge e^{4}\ne0.

By (20), dφ2d\varphi^{2} is a real (1,1)(1,1)-form. Hence

∂‾φ2=∂φ2‾=−−12φ1∧φ1‾,∂φ2=∂‾φ2‾=0.\overline{\partial}\varphi^{2}=\partial\overline{\varphi^{2}}=-\frac{\sqrt{-1}}{2}\varphi^{1}\wedge\overline{\varphi^{1}},\qquad\partial\varphi^{2}=\overline{\partial}\overline{\varphi^{2}}=0.

Since dφ1=0d\varphi^{1}=0, it follows that

∂‾α=14φ1∧φ1‾∧φ2‾,∂∂‾α=−−18φ1∧φ1‾∧φ1∧φ1‾=0.\overline{\partial}\alpha=\frac{1}{4}\varphi^{1}\wedge\overline{\varphi^{1}}\wedge\overline{\varphi^{2}},\qquad\partial\overline{\partial}\alpha=-\frac{\sqrt{-1}}{8}\varphi^{1}\wedge\overline{\varphi^{1}}\wedge\varphi^{1}\wedge\overline{\varphi^{1}}=0.

Thus α\alpha is pluriclosed.

Let X=Cn−2/(Z+−1Z)n−2X=\mathbb{C}^{n-2}/(\mathbb{Z}+\sqrt{-1}\mathbb{Z})^{n-2} with its flat Kähler form β\beta. On M=S×XM=S\times X, set

ω=πS∗α+πX∗β,\omega=\pi_{S}^{*}\alpha+\pi_{X}^{*}\beta,

where πS\pi_{S} and πX\pi_{X} are the projections onto the two factors. Since these projections are holomorphic,

∂∂‾ω=πS∗(∂∂‾α)+πX∗(∂∂‾β)=0.\partial\overline{\partial}\omega=\pi_{S}^{*}(\partial\overline{\partial}\alpha)+\pi_{X}^{*}(\partial\overline{\partial}\beta)=0.

Moreover,

dω=πS∗(dα)≠0,d\omega=\pi_{S}^{*}(d\alpha)\ne0,

so the product metric is not Kähler.

Since SS has complex dimension two, α3=0\alpha^{3}=0 and d(α2)=0d(\alpha^{2})=0. For k≥2k\ge2, we therefore have

ωk=πX∗(βk)+kπS∗α∧πX∗(βk−1)+(k2)πS∗(α2)∧πX∗(βk−2).\omega^{k}=\pi_{X}^{*}(\beta^{k})+k\pi_{S}^{*}\alpha\wedge\pi_{X}^{*}(\beta^{k-1})+\binom{k}{2}\pi_{S}^{*}(\alpha^{2})\wedge\pi_{X}^{*}(\beta^{k-2}).

Using dβ=0d\beta=0, ∂∂‾α=0\partial\overline{\partial}\alpha=0, and d(α2)=0d(\alpha^{2})=0, we obtain ∂∂‾ωk=0\partial\overline{\partial}\omega^{k}=0. This also holds for k=1k=1. Taking k=n−2k=n-2 shows that the metric is astheno-Kähler for every n≥3n\ge3.

We now verify directly that the product metric is not torsion-critical. For s>0s>0, set

ωs=sπS∗α+πX∗β,\omega_{s}=s\pi_{S}^{*}\alpha+\pi_{X}^{*}\beta,

and let gsg_{s} be the corresponding Hermitian metric. Explicitly, gs=sπS∗gS+πX∗gXg_{s}=s\pi_{S}^{*}g_{S}+\pi_{X}^{*}g_{X}, where gXg_{X} is the flat Kähler metric associated with β\beta. Thus g1g_{1} is the original product metric, with fundamental form ω1=ω\omega_{1}=\omega.

Since SS has complex dimension two,

dvgs=s2 dvg1,V(gs)=s2V(g1).dv_{g_{s}}=s^{2}\,dv_{g_{1}},\qquad V(g_{s})=s^{2}V(g_{1}).

The Chern torsion comes entirely from the surface factor SS. Scaling gSg_S by ss leaves its Chern connection unchanged and multiplies the squared torsion norm by s−1s^{-1}. Thus

∣T(gs)∣gs2=s−1∣T(g1)∣g12.\lvert T(g_s)\rvert_{g_s}^{2}=s^{-1}\lvert T(g_1)\rvert_{g_1}^{2}.

It follows that

∫M∣T(gs)∣gs2 dvgs=s∫M∣T(g1)∣g12 dvg1.\int_M \lvert T(g_s)\rvert_{g_s}^{2}\,dv_{g_s}=s\int_M \lvert T(g_1)\rvert_{g_1}^{2}\,dv_{g_1}.

By the definition of F\mathcal{F}, we obtain

F(gs)=s2−nnF(g1),dds∣s=1F(gs)=2−nnF(g1)<0.\mathcal{F}(g_s)=s^{\frac{2-n}{n}}\mathcal{F}(g_1),\qquad \left.\frac{d}{ds}\right|_{s=1}\mathcal{F}(g_s)=\frac{2-n}{n}\mathcal{F}(g_1)<0.

Here F(g1)>0\mathcal{F}(g_1)>0 because g1g_1 is not Kähler, and n≥3n\geq3. Thus g1g_1 is not torsion-critical. This example shows that the torsion-critical assumption in Theorem 1.1 cannot be omitted.

Remark 3.4. The product calculation works more generally for any compact pluriclosed Hermitian surface and any compact Kähler factor. Example 3.3 shows that pluriclosedness and the astheno-Kähler condition together do not imply Kählerness. It establishes the need for torsion-criticality in Theorem 1.1. It does not establish the necessity of the astheno-Kähler hypothesis among pluriclosed torsion-critical metrics.

We now explain the role of the astheno-Kähler hypothesis in the comparison of first variations.

For an integer k≥1k\geq1, repeated use of the product rule gives

∂ˉ∂ωk=k∂ˉ∂ω∧ωk−1+k(k−1)∂ω∧∂ˉω∧ωk−2,(21)\bar{\partial}\partial\omega^k = k\bar{\partial}\partial\omega\wedge\omega^{k-1} + k(k-1)\partial\omega\wedge\bar{\partial}\omega\wedge\omega^{k-2}, \tag*{(21)}

where the second term is absent for k=1k=1.

If gg is pluriclosed and n≥4n\geq4, equation (21) becomes

∂ˉ∂ωn−2=(n−2)(n−3)∂ω∧∂ˉω∧ωn−4.(22)\bar{\partial}\partial\omega^{n-2} = (n-2)(n-3)\partial\omega\wedge\bar{\partial}\omega\wedge\omega^{n-4}. \tag*{(22)}

Consequently, at a pluriclosed torsion-critical metric,

DGg(γ)=−V1−nn(n−2)!∫Mγ∧−1∂ˉ∂ωn−2(23)DG_g(\gamma) = -\frac{V^{\frac{1-n}{n}}}{(n-2)!}\int_M\gamma\wedge\sqrt{-1}\bar{\partial}\partial\omega^{n-2} \tag*{(23)}

for every smooth real (1,1)(1,1)-form γ\gamma. The right-hand side is the precise term left by the comparison method. Theorem 2.8 cannot be applied unless DGgDG_g vanishes on all such directions.

Set Ψ=−1∂ˉ∂ωn−2\Psi=\sqrt{-1}\bar{\partial}\partial\omega^{n-2}. Taking the real (1,1)(1,1)-form γ=∗Ψ\gamma=* \Psi in 3.4 gives

DGg(∗Ψ)=−V1−nn(n−2)!∫M∣Ψ∣2 dv.DG_g(*\Psi) = -\frac{V^{\frac{1-n}{n}}}{(n-2)!}\int_M\lvert\Psi\rvert^2\,dv.

Thus, if Ψ≠0\Psi\neq0, this variation decreases G\mathcal{G} to first order. This is consistent with DFg=0DF_g=0, which alone does not imply DGg=0DG_g=0.

Consequences under curvature assumptions

The rigidity theorem above works for a fixed Hermitian metric. Without extra assumptions, it cannot tell us the biholomorphism type of the complex manifold. We can get more results if we add conditions on curvature and topology. These results follow from Theorem 1.1, combined with known theorems about flat manifolds and manifolds with positive holomorphic bisectional curvature. Readers may refer to Boothby [4] for the structure of compact Chern-flat Hermitian manifolds.

Corollary 4.1. Let (Mn,g)(M^n,g) be a compact connected Hermitian manifold, with n≥3n\geq3. Suppose that gg is pluriclosed, astheno-Kähler, and torsion-critical. If its Chern curvature vanishes, RC=0R^{\mathrm{C}}=0, then there is a finite unramified holomorphic covering

Cn/Λ⟶M,\mathbb{C}^n/\Lambda\longrightarrow M,

where Λ⊂Cn\Lambda\subset\mathbb{C}^{n} is a lattice of real rank 2n2n. If, in addition, b1(M)=2nb_{1}(M)=2n, then MM itself is biholomorphic to a complex torus. Here b1(M)=dim⁡RHdR1(M;R)b_{1}(M)=\dim_{\mathbb{R}}H^{1}_{\mathrm{dR}}(M;\mathbb{R}).

Proof. By Theorem 1.1, gg is Kähler. Its Chern connection coincides with the Levi-Civita connection. Thus, Rc=0R^{c}=0 means gg is flat. The universal cover inherits the lifted metric. It is complete, simply connected and flat, so it is isometric to R2n\mathbb{R}^{2n}. The lifted complex structure is parallel. We may choose this isometry to be holomorphic, and it identifies the universal cover with Cn\mathbb{C}^{n}.

Let Γ\Gamma be the deck transformation group. Each δ∈Γ\delta\in\Gamma has the form

δ(z)=Aδz+aδ,Aδ∈U(n),aδ∈Cn.\delta(z)=A_{\delta}z+a_{\delta},\qquad A_{\delta}\in U(n),\qquad a_{\delta}\in\mathbb{C}^{n}.

By the first Bieberbach theorem, the translation subgroup of Γ\Gamma is a full lattice Λ\Lambda of finite index; see [5]. Thus

Cn/Λ⟶Cn/Γ=M\mathbb{C}^{n}/\Lambda\longrightarrow\mathbb{C}^{n}/\Gamma=M

is a finite covering. The map is holomorphic and locally biholomorphic, so it is unramified.

Suppose now that b1(M)=2nb_{1}(M)=2n. By Hodge theory, the space of harmonic real 1-forms has dimension 2n2n.

For a real 1-form α\alpha, by the Weitzenböck–Bochner formula in [21],

Δdα=(∇LC)∗∇LCα+Ric⁡⋅α,\Delta_{d}\alpha=(\nabla^{\mathrm{LC}})^{*}\nabla^{\mathrm{LC}}\alpha+\operatorname{Ric}\cdot\alpha,

where Δd=dd∗+d∗d\Delta_{d}=dd^{*}+d^{*}d. If α\alpha is harmonic, then Δdα=0\Delta_{d}\alpha=0. Take the inner product with α\alpha and integrate:

0=∫M⟨(∇LC)∗∇LCα,α⟩dv+∫M⟨Ric⁡⋅α,α⟩dv.0=\int_{M}\left\langle(\nabla^{\mathrm{LC}})^{*}\nabla^{\mathrm{LC}}\alpha,\alpha\right\rangle dv+\int_{M}\left\langle\operatorname{Ric}\cdot\alpha,\alpha\right\rangle dv.

Integration by parts gives

∫M⟨(∇LC)∗∇LCα,α⟩dv=∫M∣∇LCα∣2dv.\int_{M}\left\langle(\nabla^{\mathrm{LC}})^{*}\nabla^{\mathrm{LC}}\alpha,\alpha\right\rangle dv=\int_{M}\left|\nabla^{\mathrm{LC}}\alpha\right|^{2}dv.

Since the metric is flat, the Ricci curvature vanishes, so

0=∫M∣∇LCα∣2dv.0=\int_{M}\left|\nabla^{\mathrm{LC}}\alpha\right|^{2}dv.

The integrand is non-negative and continuous, hence ∇LCα=0\nabla^{\mathrm{LC}}\alpha=0.

Thus every harmonic 1-form is parallel. Evaluation at any point is injective on the space of parallel one-forms: a parallel form that vanishes at one point vanishes everywhere on the connected manifold. Hence the 2n2n linearly independent harmonic 1-forms form a real coframe at every point.

Their lifts to Cn\mathbb{C}^{n} are constant 1-forms invariant under all deck transformations. Thus Aδ∗A_{\delta}^{*} fixes a basis of the real cotangent space, so Aδ=IA_{\delta}=I for every δ∈Γ\delta\in\Gamma. Every deck transformation is therefore a translation. Hence Γ=Λ\Gamma=\Lambda, and M=Cn/ΛM=\mathbb{C}^{n}/\Lambda is a complex torus. □\square

Corollary 4.2. Let (Mn,g)(M^{n},g) be a compact connected Hermitian manifold, with n≥3n\geq3. Suppose that gg is pluriclosed, astheno-Kähler, and torsion-critical. If its Chern holomorphic bisectional curvature is strictly positive, then MM is biholomorphic to CPn\mathbb{CP}^{n}.

Proof. Theorem 1.1 implies that gg is Kähler. In particular, its Chern connection agrees with the connection induced by the Levi-Civita connection, so the Chern holomorphic bisectional curvature is precisely the usual Kähler holomorphic bisectional curvature.

For any point p∈Mp\in M, define the bisectional curvature

Bg(X,Y)=Rg(X,X‾,Y,Y‾)∣X∣g2∣Y∣g2,X,Y≠0,X,Y∈Tp1,0M.B_{g}(X,Y)=\frac{R_{g}(X,\overline{X},Y,\overline{Y})}{|X|_{g}^{2}|Y|_{g}^{2}},\qquad X,Y\neq0,\qquad X,Y\in T_{p}^{1,0}M.

Since the denominator is strictly positive,

Bg(X,Y)>0⟺Rg(X,X‾,Y,Y‾)>0.B_{g}(X,Y)>0\Longleftrightarrow R_{g}(X,\overline{X},Y,\overline{Y})>0.

The Siu–Yau theorem [23] states that a compact Kähler manifold with strictly positive holomorphic bisectional curvature is biholomorphic to complex projective space. Applying that theorem gives M≅CPnM \cong\mathbb{CP}^{n}.

Remark 4.3. Corollary 4.1 gives a finite covering of MM by a complex torus. It does not imply that MM itself is a complex torus. Indeed, the deck transformations of the universal cover may have nontrivial linear parts. The condition b1(M)=2nb_{1}(M)=2n forces all these linear parts to be the identity. Then MM is a complex torus.

In this Chern-flat case, the astheno-Kähler and torsion-critical assumptions can be omitted. Compact Chern-flat Hermitian metrics are balanced; see [30] and the references therein. Thus pluriclosedness and Corollary 2.10 already imply that gg is Kähler.

Acknowledgments. We would like to express our sincere gratitude to Professor Fangyang Zheng for his insightful guidance, valuable discussions, and continued support throughout the course of this research.

Disclosure on AI assistance. The research problem and the principal mathematical ideas were formulated by the authors. AI tools were used only during the final stages of manuscript preparation to improve the exposition. The authors retain full responsibility for the correctness of the final work.

References

Dongmei Zhang. School of Mathematical Sciences, Chongqing Normal University, Chongqing 401331, China

Email address: [email protected]

Yuqin Guo. School of Mathematical Sciences, Chongqing Normal University, Chongqing 401331, China

Email address: [email protected]

References

  1. [1]B. Alexandrov and S. Ivanov, Vanishing theorems on Hermitian manifolds, Differential Geom. Appl. 14 (2001), no. 3, 251–265. (cited on page 2)DOI
  2. [2]D. Angella, N. Istrati, A. Otiman, and N. Tardini, Variational problems in conformal geometry, J. Geom. Anal. 31 (2021), no. 3, 3230–3251. (cited on page 2)DOI
  3. [3]J.-M. Bismut, A local index theorem for non-Kähler manifolds, Math. Ann. 284 (1989), no. 4, 681–699. (cited on page 2)
  4. [4]W. M. Boothby, Hermitian manifolds with zero curvature, Michigan Math. J. 5 (1958), 229–233. (cited on page 10)DOI
  5. [5]L. S. Charlap, Bieberbach Groups and Flat Manifolds, Universitext, Springer-Verlag, New York, 1986. (cited on page 11)DOI
  6. [6]I. Chiose, Obstructions to the existence of Kähler structures on compact complex manifolds, Proc. Amer. Math. Soc. 142 (2014), no. 10, 3561–3568. (cited on page 2)DOI
  7. [7]N. Enrietti, A. Fino, and L. Vezzoni, Tamed symplectic forms and strong Kähler with torsion metrics, J. Symplectic Geom. 10 (2012), no. 2, 203–223. (cited on page 2)
  8. [8]A. Fino, G. Grantcharov, and L. Vezzoni, Astheno-Kähler and balanced structures on fibrations, Int. Math. Res. Not. IMRN 2019 (2019), no. 22, 7093–7117. (cited on pages 2 and 3)DOI
  9. [9]A. Fino, M. Parton, and S. Salamon, Families of strong KT structures in six dimensions, Comment. Math. Helv. 79 (2004), no. 2, 317–340. (cited on page 2)
  10. [10]A. Fino and A. Tomassini, A survey on strong KT structures, Bull. Math. Soc. Sci. Math. Roumanie (N.S.) 52(100) (2009), no. 2, 99–116. (cited on page 2)
  11. [11]A. Fino and A. Tomassini, Non-Kähler metrics, J. Lond. Math. Soc. (2) 83 (2011), no. 2, 290–308. (cited on page 2)
  12. [12]A. Fino and L. Ugarte, On generalized Gauduchon metrics, Proc. Edinb. Math. Soc. (2) 56 (2013), no. 3, 733–753. (cited on page 2)arxiv.org/abs/1103.1033
  13. [13]A. Fino and L. Vezzoni, On the existence of balanced and SKT metrics on nilmanifolds, Proc. Amer. Math. Soc. 144 (2016), no. 6, 2455–2459. (cited on page 2)
  14. [14]P. Gauduchon, La 1-forme de torsion d’une variété hermitienne compacte, Math. Ann. 267 (1984), no. 4, 495–518. (cited on page 1)DOI
  15. [15]P. Gauduchon, Hermitian connections and Dirac operators, Boll. Un. Mat. Ital. B (7) 11 (1997), no. 2, suppl., 257–288. (cited on page 2)
  16. [16]K. Hasegawa, Minimal models of nilmanifolds, Proc. Amer. Math. Soc. 106 (1989), no. 1, 65–71. (cited on page 8)DOI
  17. [17]J. Jost and S.-T. Yau, A nonlinear elliptic system for maps from Hermitian to Riemannian manifolds and rigidity theorems in Hermitian geometry, Acta Math. 170 (1993), no. 2, 221–254. (cited on page 2)
  18. [18]M. Kawamura, On torsion-critical manifolds in almost Hermitian geometry, Bull. Iran. Math. Soc. 51 (2025), article 68. (cited on page 2)DOI
  19. [19]A. Latorre and L. Ugarte, On non-Kähler compact complex manifolds with balanced and astheno-Kähler metrics, C. R. Math. Acad. Sci. Paris 355 (2017), no. 1, 90–93. (cited on page 3)
  20. [20]M. L. Michelsohn, On the existence of special metrics in complex geometry, Acta Math. 149 (1982), 261–295. (cited on page 2)DOI
  21. [21]P. Petersen, Demystifying the Weitzenböck curvature operator, preprint, available at https://www.math.ucla.edu/~petersen/BLWformulas.pdf. (cited on page 11)
  22. [22]S. M. Salamon, Complex structures on nilpotent Lie algebras, J. Pure Appl. Algebra 157 (2001), no. 2–3, 311–333. (cited on page 8)DOI
  23. [23]Y.-T. Siu and S.-T. Yau, Compact Kähler manifolds of positive bisectional curvature, Invent. Math. 59 (1980), no. 2, 189–204. (cited on page 12)DOI
  24. [24]J. Streets and G. Tian, A parabolic flow of pluriclosed metrics, Int. Math. Res. Not. IMRN 2010 (2010), no. 16, 3101–3133. (cited on page 2)arxiv.org/abs/0903.4418
  25. [25]A. Strominger, Superstrings with torsion, Nuclear Phys. B 274 (1986), 253–284. (cited on page 2)DOI
  26. [26]L. Vezzoni, B. Yang, and F. Zheng, Lie groups with flat Gauduchon connections, Math. Z. 293 (2019), no. 1–2, 597–608. (cited on page 2)arxiv.org/abs/1805.04719
  27. [27]Q. Wang, B. Yang, and F. Zheng, On Bismut flat manifolds, Trans. Amer. Math. Soc. 373 (2020), no. 8, 5747–5772. (cited on page 2)arxiv.org/abs/1603.07058
  28. [28]B. Yang and F. Zheng, On curvature tensors of Hermitian manifolds, Comm. Anal. Geom. 26 (2018), no. 5, 1195–1222. (cited on page 8)arxiv.org/abs/1602.01189
  29. [29]D. Zhang and F. Zheng, On a variational theorem of Gauduchon and torsion-critical manifolds, Proc. Amer. Math. Soc. 151 (2023), no. 4, 1749–1762. (cited on pages 1, 2, 3, 7, and 8)arxiv.org/abs/2205.12480
  30. [30]D. Zhang and F. Zheng, Chern flat manifolds that are torsion-critical, Proc. Amer. Math. Soc. 153 (2025), no. 3, 1275–1282. (cited on pages 2, 3, 8, and 12)
  31. [31]Q. Zhao and F. Zheng, Strominger connection and pluriclosed metrics, J. Reine Angew. Math. 796 (2023), 245–267. (cited on page 8)arxiv.org/abs/1904.06604

Paper details

vlnc:W_syjykcpn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

AI tools

View statement
Looking for related papers…
Contents