betaFeedback

WorksMathematics

Mixed Chern curvature on complete flags

Abstract

We prove that the complete flag manifold SU(m)/T m-1, with m >= 3 and its standard complex structure, admits no invariant Hermitian metric with constant mixed Chern curvature when the coefficient of the holomorphic sectional curvature is nonzero. This extends the preceding nonexistence result for a particular Wallach metric on SU(3)/T2 to all invariant Hermitian metrics and to complete type-A flags of arbitrary dimension. Curvature evaluations on holomorphic root spheres contradict a necessary trace identity, covering both Kahler and non-Kahler invariant metrics and excluding both zero and nonzero curvature constants.

Introduction and statement of result

Complex geometry harbours a long-standing open problem concerning classification of compact Hermitian manifolds endowed with constant holomorphic sectional curvature. In the Kähler case, it is a well established fact that every compact (or complete) Kähler manifold of constant holomorphic sectional curvature belongs to the family of complex space forms. In other words, their universal covers are one of the standard models: complex projective space CPn\mathbb{CP}^{n}, complex Euclidean space Cn\mathbb{C}^{n}, or complex hyperbolic space CHn\mathbb{CH}^{n}.

For non-Kähler metrics gg, the curvature tensor RR fails to satisfy the symmetry identities valid in the Kähler framework. Consequently, the holomorphic sectional curvature HH alone cannot recover full information encoded within RR. Nevertheless, the following statement remains widely conjectured to hold for compact manifolds.

Conjecture 1 (Constant Chern holomorphic sectional curvature; original formulation). Let (Mn,J,g)(M^{n}, J, g) be a compact Hermitian manifold with n≥2n \ge2 and H≡cH \equiv c. If c≠0c \ne0, then gg is Kähler; if c=0c = 0, then gg is Chern flat, namely R≡0R \equiv0.

Note that the compactness assumption is necessary for the validity of the conjecture, without it there are counterexamples in the noncompact case. Compact Chern flat manifolds have been classified by Boothby [4] as all the compact quotients of complex Lie groups equipped with left invariant metrics. Moreover, compact Levi-Civita flat Hermitian threefolds have been classified by Khan-Yang-Zheng in [14], but for n≥4n \ge4, the classification remains an open question.

In complex dimension n=2n = 2, combined contributions from Balas-Gauduchon [2, 3] (1985), Sato-Sekigawa [23] (1990), and Apostolov-Davidov-Muskarov [1] (1996) have confirmed Conjecture 1. For n≥3n \ge3, the first substantive result towards this conjecture originates from Davidov-Grantcharov-Muskarov [12]. Their work shows that CP3\mathbb{CP}^{3} is the unique twistor space admitting constant holomorphic sectional curvature. More recently, Chen-Chen-Nie [5] verified Conjecture 1 for locally conformally Kähler manifolds satisfying c≤0c \le0. Tang [24] validated the conjecture under an extra hypothesis that the metric is Chern-Kähler-like, meaning its Chern curvature RcR^{c} obeys all Kähler symmetry rules. Zhou-Zheng [37] obtained another confirmation for compact balanced threefolds with vanishing real bisectional curvature a condition introduced in [33], which is marginally stronger than the condition encoded by HcH^{c}. For Hermitian manifolds, Tang [28] has given a complete solution to the problem of constant real bisectional curvature. Their argument yields Chern-flatness for such manifolds. Further affirmative instances for ∇c\nabla^{c} were obtained by Li-Zheng [15] and Rao-Zheng [22], working under additional assumptions that either (Mn,g)(M^n,g) is a complex nilmanifold or gg satisfies the Bismut-Kähler-like condition, so that the Bismut curvature RbR^b satisfies all Kähler symmetries.

Recent preprints settle the unrestricted threefold case: Qin–Tian [21] established nonzero-constant rigidity, and Chen–Zheng [7] established Chern flatness for c=0c=0, removing the balanced assumption in Chen–Li [6]. Qin–Tian’s version 3 also proves Kähler rigidity for c≤0c\leq0 in Fujiki’s class CC and reports compact balanced metrics with Hg≡0H_g\equiv0 but R≢0R\not\equiv0 for every n≥7n\geq7. If confirmed, these reported counterexamples disprove the unrestricted zero-curvature assertion in those dimensions.

Geometers have also achieved abundant results on the mixed-curvature problem. Mixed curvature was introduced by Chu–Lee–Tam in the Kähler setting [10]. Its special case α=0\alpha=0 leads to the classical Chern holomorphic sectional curvature problem; see [36] for a survey. In the above notation, we have C1,0=Ric⁡C_{1,0}=\operatorname{Ric} and C0,1=HC_{0,1}=H. When the metric gg is Kähler, the mixed curvature encompasses several interesting curvature conditions. Specifically, C1,1C_{1,1} corresponds to the Ric⁡+(X,X‾)\operatorname{Ric}^{+}(X,\overline{X}) introduced by Ni [18], and C1,−1C_{1,-1} corresponds to the orthogonal Ricci curvature Ric⁡⊥(X,X‾)\operatorname{Ric}^{\perp}(X,\overline{X}) introduced by Ni-Zheng [20]. Additionally, if kk is an integer between 11 and nn, then Ck−1,n−kC_{k-1,n-k} is closely related to the kk-Ricci curvature Ric⁡k\operatorname{Ric}_{k} introduced by Ni [19].

For compact Kähler manifolds with Cα,β≥0C_{\alpha,\beta}\geq0 or Cα,β≤0C_{\alpha,\beta}\leq0, there have been important progress made in recent years. For example, Yang [32] proved that a compact Kähler manifold with C0,1>0C_{0,1}>0 must be projective and rationally connected, confirming a conjecture by Yau [34]. Ni [18] generalized this result, showing that the same holds if Ric⁡k>0\operatorname{Ric}_{k}>0 for some 1≤k≤n1\leq k\leq n. Matsumura [16], [17] established structure theorems for projective manifolds with C0,1≥0C_{0,1}\geq0, while Chu-Lee-Tam [10] and others proved that compact Kähler manifolds with Cα,β>0C_{\alpha,\beta}>0, where α>0\alpha>0 and 3α+2β≥03\alpha+2\beta\geq0, must be projective and simply-connected. Zhang-Zhang [35] extended Yang’s result to the quasi-positive case, confirming a conjecture by Yang. Tang [26] demonstrated the projectivity of compact Kähler manifolds with quasi-positive Cα,βC_{\alpha,\beta}. More recently, Chu-Lee-Zhu [11] established a structure theorem for compact Kähler manifolds with Cα,β≥0C_{\alpha,\beta}\geq0. In the non-positive case, Wu-Yau [30] confirmed a conjecture of Yau that projective Kähler manifolds with C0,1<0C_{0,1}<0 must have ample canonical line bundles. Tosatti-Yang [29] showed that the projectivity assumption can be removed in Wu-Yau’s Theorem, and Chu-Lee-Tam [10] proved that compact Kähler manifolds with Cα,β<0C_{\alpha,\beta}<0 and α≥0\alpha\geq0, β≥0\beta\geq0, must have ample canonical bundles.

Most existing results for mixed-curvature invariants focus predominantly on the Kähler setting; comparatively little is understood for non-Kähler geometries. In [25], Tang formulated the following conjecture:

Conjecture 2 (Tang’s conjecture). Let (Mn,J,g)(M^n,J,g) be a compact Hermitian manifold with Cα,β≡cC_{\alpha,\beta}\equiv c and β≠0\beta\neq0. If c≠0c\neq0, then gg is Kähler.

Chen–Zheng [8] verified this conjecture for several Lie–Hermitian classes and compact balanced BTP threefolds. Chen–Tang [9] studied other Chern–Ricci contractions, and Tang–Wang [27] proved nonzero-constant rigidity for compact LCK metrics with constant kkth-mixed curvature. The unrestricted conjecture remains open; zero mixed curvature requires separate treatment. Our theorem excludes both zero and nonzero constants on complete flags independently of either conjecture.

Compact complex homogeneous spaces constitute a natural framework for studying the constant mixed-curvature condition. Their invariant Hermitian metrics are characterized via isotropy representations, and their first Chern class may be computed using homogeneous line bundles.

Even so, mere transitivity does not guarantee that the mixed curvature is independent of tangent directions. On a complete flag manifold, the distinct root directions yield a finite set of geometrically distinguished directions, against which one may verify the constant-curvature condition. We consider the complete type-A flag manifold

Fm=SU⁡(m)/Tm−1,n=dim⁡CFm=m(m−1)2,(1)\mathcal{F}_m=\operatorname{SU}(m)/T^{m-1},\qquad n=\dim_{\mathbb{C}}\mathcal{F}_m=\frac{m(m-1)}{2}, \tag*{(1)}

where Tm−1T^{m-1} is the diagonal maximal torus. We equip FmF_m with the standard integrable complex structure obtained from the space of complete flags in Cm\mathbb{C}^m. Our main result is the following.

Theorem 1.1. Let m≥3m \ge3, and let gg be an SU⁡(m)\operatorname{SU}(m)-invariant Hermitian metric on FmF_m with its standard invariant complex structure. There are no real numbers α,β,c\alpha,\beta,c with β≠0\beta\ne0 such that

Cα,β(X)=cfor every nonzero X∈T1,0Fm.C_{\alpha,\beta}(X)=c \qquad\text{for every nonzero }X\in T^{1,0}F_m.

The theorem concerns all invariant Hermitian metrics, rather than a single normal or canonical metric. In particular, it includes both Kähler and non-Kähler metrics, and excludes constant mixed curvature even when the proposed constant is zero. The restriction m≥3m \ge3 is sharp: F2=CP1F_2=\mathbb{CP}^1 has constant mixed curvature for every invariant Hermitian metric. The assumption β≠0\beta\ne0 is also essential, since invariant Kähler–Einstein metrics on complete flag manifolds have constant first Ricci curvature.

There is an important preceding special case. In the proof of their Theorem 3, Chen–Zheng [8 Section 6] show that the particular non-Kähler Wallach metric on the threefold

SU⁡(3)/T2≃P(TCP2)\operatorname{SU}(3)/T^2 \simeq\mathbb{P}(T\mathbb{CP}^2)

cannot have constant mixed curvature. Theorem 1.1 extends this nonexistence statement to every invariant Hermitian metric on that threefold and to complete type-A flags of arbitrary dimension. No parallel-torsion assumption is used. Our use of invariant flag geometry is compatible with the framework of Grama–Oliveira [13]; their scalar-curvature condition differs from the mixed-curvature condition studied here.

The proof requires only selected directional values of RR, rather than a formula for all of its components. Every root line is tangent to a holomorphically embedded sphere. Since twice a root is not a root, the isotropy representation forces the Chern second fundamental form of this sphere to vanish. The resulting holomorphic sectional curvature and the anticanonical degree yield, respectively,

H(eij)=2xij,Ric⁡(1)(eij,eˉij)=2(j−i)xij,H(e_{ij})=\frac{2}{x_{ij}},\qquad\operatorname{Ric}^{(1)}(e_{ij},\bar e_{ij})=\frac{2(j-i)}{x_{ij}},

where eije_{ij} is a unit root vector and xij>0x_{ij}>0 is the scale of the induced sphere metric. A spherical average then eliminates the parameters α\alpha and cc and leaves a sum of nonnegative terms with at least one strictly positive term.

Section 2 establishes the averaging identity and the necessary directional trace condition. Section 3 describes invariant metrics, balancedness, and holomorphic root spheres. Section 4 computes their anticanonical degrees and first Chern Ricci curvature. Section 5 proves Theorem 1.1, gives the explicit three-dimensional calculation, and explains the boundary cases.

Averaging and the directional trace identity

Let (Mn,J,g)(M^n,J,g) be a Hermitian manifold, and let RR denote the curvature tensor of its Chern connection. For a nonzero vector X∈T1,0MX\in T^{1,0}M, write

H(X)=R(X,Xˉ,X,Xˉ)∣X∣4,Cα,β(X)=αRic⁡(1)(X,Xˉ)∣X∣2+βH(X).(2)H(X)=\frac{R(X,\bar X,X,\bar X)}{|X|^4},\qquad C_{\alpha,\beta}(X)=\alpha\frac{\operatorname{Ric}^{(1)}(X,\bar X)}{|X|^2}+\beta H(X). \tag*{(2)}

where Ric⁡(1)\operatorname{Ric}^{(1)} is the first Chern Ricci tensor and α,β∈R\alpha,\beta\in\mathbb{R}. Throughout the paper, β≠0\beta\ne0. We use the term mixed curvature for this real linear combination without imposing a sign condition on either coefficient. The notation Cα,β≡cC_{\alpha,\beta}\equiv c means that the same real number cc is obtained at every point and in every nonzero (1,0)(1,0) tangent direction. In holomorphic coordinates, write the fundamental form as

ω=−1∑i,jgijˉ dzi∧dzˉj.\omega=\sqrt{-1}\sum_{i,j}g_{i\bar j}\,dz^i\wedge d\bar z^j.

Our Chern curvature convention is

Rijˉkℓˉ=−∂i∂jˉgkℓˉ+∑r,tgrtˉ(∂igktˉ)(∂jˉgrℓˉ).(3)R_{i\bar j k\bar\ell}=-\partial_i\partial_{\bar j}g_{k\bar\ell}+\sum_{r,t}g^{r\bar t}(\partial_i g_{k\bar t})(\partial_{\bar j}g_{r\bar\ell}). \tag*{(3)}

Here (grˉt)(g^{\bar r t}) denotes the inverse Hermitian tensor, so that

∑tgrˉtgktˉ=δkr.\sum_t g^{\bar r t}g_{k\bar t}=\delta_k^r.

With this normalization, the metric

gwwˉ=x(1+∣w∣2)−2g_{w\bar w}=x(1+|w|^2)^{-2}

on CP1\mathbb{CP}^1 has holomorphic sectional curvature 2/x2/x.

Under a unitary frame e1,…,ene_1,\ldots,e_n, put

Pijˉ=Ric⁡ijˉ(1)=∑kRijˉkkˉ,s=∑iPiiˉ,s~=∑i,kRikˉkiˉ.(4)P_{i\bar j}=\operatorname{Ric}^{(1)}_{i\bar j}=\sum_k R_{i\bar j k\bar k},\qquad s=\sum_i P_{i\bar i},\qquad\widetilde{s}=\sum_{i,k}R_{i\bar k k\bar i}. \tag*{(4)}

Both ss and s~\widetilde{s} are real. We do not identify these two contractions on an arbitrary Hermitian manifold. The first Chern Ricci form is

ρ(1)=−1∑i,jPijˉ dzi∧dzˉj=−−1∂∂ˉlog⁡det⁡(gijˉ),[ρ(1)]=2πc1(M).(5)\rho^{(1)}=\sqrt{-1}\sum_{i,j}P_{i\bar j}\,dz^i\wedge d\bar z^j=-\sqrt{-1}\partial\bar\partial\log\det(g_{i\bar j}),\qquad[\rho^{(1)}]=2\pi c_1(M). \tag*{(5)}

It is closed without a Kähler assumption.

For the full Chern torsion components, we use

Tijk=∑ℓgkℓˉ(∂igjℓˉ−∂jgiℓˉ),ηi=∑kTikk.T^k_{ij}=\sum_\ell g^{k\bar\ell}(\partial_i g_{j\bar\ell}-\partial_j g_{i\bar\ell}),\qquad\eta_i=\sum_k T^k_{ik}.

The metric is balanced if dωn−1=0d\omega^{n-1}=0, equivalently if the torsion trace form η\eta vanishes. With the full torsion convention above, the Chern Bianchi identity in a unitary frame reads

Tij,ℓˉk=Rjℓˉikˉ−Riℓˉjkˉ.T^k_{ij,\bar\ell}=R_{j\bar\ell i\bar k}-R_{i\bar\ell j\bar k}.

see [31]. Here a comma denotes Chern covariant differentiation. Contracting this identity gives

∑iηi,iˉ=∑i,k(Rkiˉikˉ−Riiˉkkˉ)=s~−s.(6)\sum_i\eta_{i,\bar i}=\sum_{i,k}(R_{k\bar i i\bar k}-R_{i\bar i k\bar k})=\widetilde{s}-s. \tag*{(6)}

Consequently, every balanced Hermitian metric satisfies s=s~s=\widetilde{s} pointwise. This conclusion requires neither compactness nor homogeneity.

We include the standard averaging argument below for the sake of clarity [9].

Lemma 2.1. If a Hermitian manifold of complex dimension nn satisfies Cα,β≡cC_{\alpha,\beta}\equiv c, then at every point

n(n+1)c=((n+1)α+β)s+βs~.(7)n(n+1)c=((n+1)\alpha+\beta)s+\beta\widetilde{s}. \tag*{(7)}

Proof. Choose a unitary frame at the point and write

X=∑rzrer,∑r∣zr∣2=1.X=\sum_r z_r e_r,\qquad\sum_r|z_r|^2=1.

Let dσd\sigma be probability measure on the unit sphere in Cn\mathbb{C}^n. Unitary invariance gives

∫zizˉj dσ=δijn,∫zizˉjzkzˉℓ dσ=δijδkℓ+δiℓδkjn(n+1).(8)\int z_i\bar z_j\,d\sigma=\frac{\delta_{ij}}{n},\qquad\int z_i\bar z_jz_k\bar z_\ell\,d\sigma=\frac{\delta_{ij}\delta_{k\ell}+\delta_{i\ell}\delta_{kj}}{n(n+1)}. \tag*{(8)}

To make the exposition transparent, one may compute the fourth-moment constant explicitly. The invariance property with respect to diagonal unitary matrices forces the integral to vanish whenever the multisets formed by holomorphic and antiholomorphic indices do not coincide. For n≥2n\ge2, define

A=∫∣z1∣4 dσ,B=∫∣z1z2∣2 dσ.A=\int|z_1|^4\,d\sigma,\qquad B=\int|z_1z_2|^2\,d\sigma.

The random variable (z1+z2)/2(z_1+z_2)/\sqrt{2} is identically distributed to z1z_1. Expanding its fourth power and invoking phase invariance yields A=A/2+BA=A/2+B, so A=2BA=2B. Furthermore,

1=∫(∑r∣zr∣2)2dσ=nA+n(n−1)B.1=\int\left(\sum_r|z_r|^2\right)^2d\sigma=nA+n(n-1)B.

Thus

B=1n(n+1),A=2n(n+1).B=\frac{1}{n(n+1)},\qquad A=\frac{2}{n(n+1)}.

proving (8). Its formula also holds for n=1n=1.

The second moment yields

∫P(X,Xˉ) dσ=sn.\int P(X,\bar{X})\,d\sigma=\frac{s}{n}.

The two fourth-moment contractions yield

∫R(X,Xˉ,X,Xˉ) dσ=s+s~n(n+1).\int R(X,\bar{X},X,\bar{X})\,d\sigma=\frac{s+\tilde{s}}{n(n+1)}.

Since XX is unit and Cα,β(X)=cC_{\alpha,\beta}(X)=c, averaging (2) and multiplying by n(n+1)n(n+1) proves (7). □

Lemma 2.2. Let (Mn,J,g)(M^n,J,g) be a balanced Hermitian manifold. If Cα,β≡cC_{\alpha,\beta}\equiv c with β≠0\beta\ne0, then every unitary basis at every point satisfies

∑r=1nH(er)=2sn+1.(9)\sum_{r=1}^{n}H(e_r)=\frac{2s}{n+1}. \tag*{(9)}

Proof. Balancedness and (6) give s=s~s=\tilde{s} at the chosen point. Equation (7) becomes

nc=αs+2βsn+1.(10)nc=\alpha s+\frac{2\beta s}{n+1}. \tag*{(10)}

On the other hand, evaluating the mixed-curvature condition at each member of any chosen unitary frame and summing gives

nc=α∑rP(er,eˉr)+β∑rH(er)=αs+β∑rH(er).(11)\begin{aligned} nc&=\alpha\sum_r P(e_r,\bar{e}_r)+\beta\sum_r H(e_r)\\ &=\alpha s+\beta\sum_r H(e_r). \tag*{(11)} \end{aligned}

Subtracting (10) from (11) and using β≠0\beta\ne0 proves (9). □

This cancellation proceeds solely via division by β≠0\beta\ne0, and never involves division by cc, α\alpha, or (n+1)α+2β(n+1)\alpha+2\beta. Consequently, all choices for the sign of the curvature constant and all coefficient ratios are encompassed by this cancellation. For an arbitrary Hermitian metric, the sum appearing on the left-hand side of Equation (9) is not necessarily invariant under frame changes; the hypothesis of constant mixed curvature imposes that this expression attains the stated value in any unitary frame.

Invariant geometry and holomorphic root spheres

In this section, we turn to the geometry of the complete type-A flag manifold Fm=SU(m)/Tm−1\mathcal{F}_m=\mathrm{SU}(m)/T^{m-1}. We first describe its isotropy representation and parameterize all SU(m)\mathrm{SU}(m)-invariant Hermitian metrics on Fm\mathcal{F}_m, without imposing the Kähler condition. The root decomposition of the tangent space yields a family of holomorphic root spheres embedded in Fm\mathcal{F}_m. These holomorphic root spheres are the key submanifolds where we carry out explicit curvature computations, which will be combined with the trace identity derived in Section 2 to establish our nonexistence theorem. Realize Fm\mathcal{F}_m as the set of complete flags

0=E0⊂E1⊂⋯⊂Em=Cm,dim⁡CEk=k.0=E_0\subset E_1\subset\cdots\subset E_m=\mathbb{C}^m,\qquad\dim_{\mathbb{C}}E_k=k.

Let ϵ1,…,ϵm\epsilon_1,\ldots,\epsilon_m be the standard basis of Cm\mathbb{C}^m, and let oo denote the standard flag

Eko=span⁡C{ϵ1,…,ϵk}.E_k^o=\operatorname{span}_{\mathbb{C}}\{\epsilon_1,\ldots,\epsilon_k\}.

Its stabilizer in SU(m)\mathrm{SU}(m) is Tm−1T^{m-1}. For i<ji<j, let vijv_{ij} be the tangent vector represented by the matrix unit EjiE_{ji} sending ϵi\epsilon_i to ϵj\epsilon_j, and put

Vij=Cvij.V_{ij}=\mathbb{C}v_{ij}.

Then

To1,0Fm=⨁1≤i<j≤mVij.(12)T^{1,0}_{o}\mathcal{F}_m=\bigoplus_{1\leq i<j\leq m}V_{ij}. \tag*{(12)}

An element diag⁡(t1,…,tm)\operatorname{diag}(t_1,\ldots,t_m) of the torus acts on VijV_{ij} by the character

χij(t)=tjti−1.(13)\chi_{ij}(t)=t_jt_i^{-1}. \tag*{(13)}

The characters are distinct. Indeed, torus invariance gives

g(vij,v‾ab)=χij(t)χab(t)‾g(vij,v‾ab),t∈Tm−1,g(v_{ij},\overline{v}_{ab})=\chi_{ij}(t)\overline{\chi_{ab}(t)}g(v_{ij},\overline{v}_{ab}),\qquad t\in T^{m-1},

so different root lines are orthogonal. Define

xij=g(vij,v‾ij)>0,eij=xij−1/2vij.(14)x_{ij}=g(v_{ij},\overline{v}_{ij})>0,\qquad e_{ij}=x_{ij}^{-1/2}v_{ij}. \tag*{(14)}

Then {eij:i<j}\{e_{ij}:i<j\} is a unitary basis at oo. These positive parameters describe all invariant Hermitian metrics; no condition such as dω=0d\omega=0 is imposed. Only a basis at a point is used, and no global holomorphic unitary frame is required.

Lemma 3.1. Every SU⁡(m)\operatorname{SU}(m)-invariant Hermitian metric on Fm\mathcal{F}_m is balanced.

Proof. Set Fm=SU⁡(m)/Tm−1\mathcal{F}_m=\operatorname{SU}(m)/T^{m-1}. Let o=eTm−1o=eT^{m-1} denote the standard flag, and take gg to be an SU⁡(m)\operatorname{SU}(m)-invariant Hermitian metric whose associated fundamental form is ω\omega. First, δgω\delta_g\omega is an invariant real one-form. Indeed, for every a∈SU⁡(m)a\in\operatorname{SU}(m), the left translation LaL_a preserves gg, the complex orientation, and ω\omega. An orientation-preserving isometry commutes with the Hodge star; hence it also commutes with the codifferential. Therefore

La∗(δgω)=δg(La∗ω)=δgω.L_a^*(\delta_g\omega)=\delta_g(L_a^*\omega)=\delta_g\omega.

Since the isotropy group at oo is Tm−1T^{m-1}, it follows that

(δgω)o∈(To∗Fm)Tm−1.(\delta_g\omega)_o\in(T_o^*\mathcal{F}_m)^{T^{m-1}}.

We next prove that this fixed subspace is zero. Recall that

To1,0Fm=⨁1≤i<j≤mVij,Vij=Cvij,T_o^{1,0}\mathcal{F}_m=\bigoplus_{1\leq i<j\leq m}V_{ij},\qquad V_{ij}=\mathbb{C}v_{ij},

and that t=diag⁡(t1,…,tm)∈Tm−1t=\operatorname{diag}(t_1,\ldots,t_m)\in T^{m-1} acts on VijV_{ij} through the character

χij(t)=tjti−1.\chi_{ij}(t)=t_jt_i^{-1}.

Consequently, the complexified real tangent representation and its dual have the weight decompositions

ToFm⊗RC=⨁i<j(Cχij⊕Cχij−1),T_o\mathcal{F}_m\otimes_{\mathbb{R}}\mathbb{C}=\bigoplus_{i<j}\left(\mathbb{C}_{\chi_{ij}}\oplus\mathbb{C}_{\chi_{ij}^{-1}}\right),

and

To∗Fm⊗RC=⨁i<j(Cχij−1⊕Cχij),T_o^*\mathcal{F}_m\otimes_{\mathbb{R}}\mathbb{C}=\bigoplus_{i<j}\left(\mathbb{C}_{\chi_{ij}^{-1}}\oplus\mathbb{C}_{\chi_{ij}}\right),

where Cχ\mathbb{C}_{\chi} denotes the one-dimensional representation with character χ\chi. Passing to the dual only replaces each character by its inverse.

Every character occurring in the last decomposition is nontrivial. For a fixed pair i<ji<j, consider the one-parameter subgroup of Tm−1T^{m-1} defined by

ti=e−−1θ,tj=e−1θ,tk=1(k≠i,j).t_i=e^{-\sqrt{-1}\theta},\qquad t_j=e^{\sqrt{-1}\theta},\qquad t_k=1\quad(k\neq i,j).

Its determinant equals one, and

χij(t)=e2−1θ.\chi_{ij}(t)=e^{2\sqrt{-1}\theta}.

Thus neither χij\chi_{ij} nor χij−1\chi_{ij}^{-1} is the trivial character. Since a torus representation is the direct sum of its weight spaces, a fixed vector can have a component only in the zero-weight space. There is no such space here, and therefore

(To∗Fm⊗RC)Tm−1={0}.(T_o^*\mathcal{F}_m\otimes_{\mathbb{R}}\mathbb{C})^{T^{m-1}}=\{0\}.

If a nonzero real covector were fixed by Tm−1T^{m-1}, its complexification would be a nonzero fixed vector in the complexified representation. Hence

(To∗Fm)Tm−1={0}.\left(T_o^*\mathcal{F}_m\right)^{T^{m-1}}=\{0\}.

It follows that (δgω)o=0(\delta_g\omega)_o=0. The SU(m)\mathrm{SU}(m)-invariance of δgω\delta_g\omega and the transitivity of the group action then give

δgω=0\delta_g\omega=0

on all of Fm\mathcal{F}_m.

Finally, we relate this equation to balancedness. On a Hermitian manifold of complex dimension nn, endowed with the complex orientation, the pointwise Hermitian linear-algebra identity is

∗gω=ωn−1(n−1)!.*_g\omega=\frac{\omega^{n-1}}{(n-1)!}.

For a real two-form in real dimension 2n2n, the codifferential is δg=−∗gd∗g\delta_g=-*_g d *_g. Consequently,

δgω=−1(n−1)!∗gdωn−1.\delta_g\omega=-\frac{1}{(n-1)!}*_g d\omega^{n-1}.

The Hodge star is invertible, so

δgω=0⟺dωn−1=0.\delta_g\omega=0\Longleftrightarrow d\omega^{n-1}=0.

Thus every SU(m)\mathrm{SU}(m)-invariant Hermitian metric on Fm\mathcal{F}_m is balanced, as claimed in Lemma 3.1. □

For i<ji<j, let Sij≃SU(2)S_{ij}\simeq\mathrm{SU}(2) be the subgroup acting on span⁡C{ϵi,ϵj}\operatorname{span}_{\mathbb{C}}\{\epsilon_i,\epsilon_j\} and fixing its orthogonal complement. Its orbit through oo will be denoted by CijC_{ij}. It is a holomorphic copy of CP1\mathbb{CP}^1, as can be seen directly. For a line LL in that two-dimensional space, set

Ek(L)={Ek0,k<i or k≥j,span⁡C{ϵa:1≤a≤k, a≠i}⊕L,i≤k<j.(15)E_k(L)= \begin{cases} E_k^0, & k<i\text{ or }k\geq j,\\ \operatorname{span}_{\mathbb{C}}\{\epsilon_a:1\leq a\leq k,\ a\ne i\}\oplus L, & i\leq k<j. \end{cases} \tag*{(15)}

These subspaces are nested and have dimension kk. For i≤k<ji\leq k<j, the fixed summand has dimension k−1k-1 and is disjoint from LL. On the chart L(w)=span⁡C(ϵi+wϵj)L(w)=\operatorname{span}_{\mathbb{C}}(\epsilon_i+w\epsilon_j), the subspaces depend holomorphically on ww, and the same holds in the chart at infinity. Projection of Ei(L)E_i(L) to the fixed plane span⁡C{ϵi,ϵj}\operatorname{span}_{\mathbb{C}}\{\epsilon_i,\epsilon_j\} recovers LL and defines a holomorphic inverse on the image. Thus the map is a holomorphic embedding. Its image is Sij⋅oS_{ij}\cdot o, with stabilizer U(1)\mathrm{U}(1), and its derivative at w=0w=0 is vijv_{ij}.

The induced metric is SijS_{ij}-invariant. Since the tangent isotropy representation of SU(2)/U(1)\mathrm{SU}(2)/\mathrm{U}(1) is a complex line, it is a positive multiple of the Fubini–Study metric. Its value at w=0w=0 is fixed by (14), giving

ω∣Cij=−1xij(1+∣w∣2)2 dw∧dw‾.(16)\left.\omega\right|_{C_{ij}}=\frac{\sqrt{-1}x_{ij}}{(1+|w|^2)^2}\,dw\wedge d\overline{w}. \tag*{(16)}

In particular,

∫Cijω=4πxij∫0∞r dr(1+r2)2=2πxij.(17)\int_{C_{ij}}\omega=4\pi x_{ij}\int_0^\infty\frac{r\,dr}{(1+r^2)^2}=2\pi x_{ij}. \tag*{(17)}

Thus the sphere-area parameter and the squared norm of the root vector have the same normalization.

Lemma 3.2. For every invariant Hermitian metric on Fm\mathcal{F}_m, each CijC_{ij} has vanishing Chern second fundamental form.

Proof. Along a holomorphic submanifold, T1,0CijT^{1,0}C_{ij} is a holomorphic subbundle of T1,0Fm∣CijT^{1,0}\mathcal{F}_m|_{C_{ij}}. The Chern connection has (0,1)(0,1) part equal to the holomorphic structure operator. Its (0,1)(0,1) covariant derivative therefore has zero normal projection. The remaining second fundamental form is

II(X,Y)=(∇XY)⊥,X,Y∈T1,0Cij,\mathrm{II}(X,Y)=(\nabla_XY)^\perp,\qquad X,Y\in T^{1,0}C_{ij},

Here the orthogonal projection is defined with respect to the ambient metric. This tensor is complex bilinear. Upon normal projection, the derivative of the coefficient multiplying YY yields a tangent vector.

Holomorphic isometries preserve both the Chern connection and the orthogonal normal projection. The torus fixes oo and keeps CijC_{ij} invariant. It preserves the defining plane and all fixed summands in (15). Thus the tensor evaluated at oo defines a torus-equivariant map.

Vij⊗Vij⟶⨁(a,b)≠(i,j)a<bVab.(18)V_{ij}\otimes V_{ij}\longrightarrow\bigoplus_{\substack{(a,b)\ne(i,j)\\a<b}}V_{ab}. \tag*{(18)}

Write

IIo(vij,vij)=∑(a,b)≠(i,j)a<bAabvab.\mathrm{II}_{o}(v_{ij},v_{ij})=\sum_{\substack{(a,b)\ne(i,j)\\a<b}}A_{ab}v_{ab}.

Equivariance implies

χij(t)2Aab=χab(t)Aab,t∈Tm−1.\chi_{ij}(t)^{2}A_{ab}=\chi_{ab}(t)A_{ab},\qquad t\in T^{m-1}.

The exponent vector of χij2\chi_{ij}^{2} has entries −2-2 at ii and 22 at jj, while that of every root character has entries −1-1 and 11. Both exponent vectors have sum zero. Equality of their characters on the determinant-one torus would make their difference a multiple of (1,…,1)(1,\ldots,1); the zero sum forces that multiple to vanish. The exponent vectors cannot agree, so every AabA_{ab} is zero. Therefore (18) vanishes.

Finally, SijS_{ij} acts transitively on CijC_{ij} by ambient holomorphic isometries. The vanishing at oo transports to every point of the sphere. □

Lemma 3.3. If eije_{ij} is a unit vector in VijV_{ij}, then

H(eij)=2xij.(19)H(e_{ij})=\frac{2}{x_{ij}}. \tag*{(19)}

Proof. Set E=T1,0Fm∣CijE=T^{1,0}F_m|_{C_{ij}}, S=T1,0CijS=T^{1,0}C_{ij}. The orthogonal projection of the ambient Chern connection onto SS is metric-compatible and has (0,1)(0,1) part ∂ˉS\bar{\partial}_{S}. It is therefore the Chern connection of the induced metric. With the convention (3), the subbundle Gauss identity gives

RE(X,X‾,Y,Y‾)=RS(X,X‾,Y,Y‾)+∥II(X,Y)∥g2.(20)R^{E}(X,\overline{X},Y,\overline{Y})=R^{S}(X,\overline{X},Y,\overline{Y})+\lVert\mathrm{II}(X,Y)\rVert_{g}^{2}. \tag*{(20)}

for tangent vectors X,YX,Y. Lemma 3.2 makes the last term zero, so the ambient and intrinsic holomorphic sectional curvatures agree in the tangent direction. This is a Hermitian holomorphic-subbundle identity and does not require a Kähler metric. Set

h(w)=xij(1+∣w∣2)−2.h(w)=x_{ij}(1+|w|^{2})^{-2}.

At w=0w=0,

h(0)=xij,∂wh(0)=0,∂w∂wˉh(0)=−2xij.h(0)=x_{ij},\qquad\partial_{w}h(0)=0,\qquad\partial_{w}\partial_{\bar{w}}h(0)=-2x_{ij}.

Formula (3) yields

Rwwˉwwˉ(0)=−hwwˉ(0)+h(0)−1∣hw(0)∣2=2xij.R_{w\bar{w}w\bar{w}}(0)=-h_{w\bar{w}}(0)+h(0)^{-1}|h_{w}(0)|^{2} =2x_{ij}.

Dividing by h(0)2=xij2h(0)^{2}=x_{ij}^{2} gives (19). □

Anticanonical degrees and the first Chern Ricci curvature

The holomorphic sectional curvature in Lemma 3.3 is determined by the induced sphere metric. The first Ricci contraction involves all tangent directions of the ambient flag manifold. Its value along the same sphere can instead be recovered from the first Chern class and invariance.

Let Ek\mathcal{E}_{k} be the rank-kk tautological holomorphic bundle on Fm\mathcal{F}_{m}, and let Lk=Ek/Ek−1\mathcal{L}_{k}=\mathcal{E}_{k}/\mathcal{E}_{k-1}, 1≤k≤m1\le k\le m, be the successive quotient line bundles. The determinant of the tangent bundle is

det⁡T1,0Fm≅⨂a<bHom⁡(La,Lb)≅⨂k=1mLk⊗(2k−m−1).(21)\det T^{1,0}\mathcal{F}_{m}\cong\bigotimes_{a<b}\operatorname{Hom}(\mathcal{L}_{a},\mathcal{L}_{b})\cong\bigotimes_{k=1}^{m}\mathcal{L}_{k}^{\otimes(2k-m-1)}. \tag*{(21)}

Negative powers denote powers of the dual line bundle. To justify the isomorphisms, use

Fm=SL⁡(m,C)/B,F_m = \operatorname{SL}(m,\mathbb{C})/B,

where BB is the upper-triangular Borel subgroup. The holomorphic tangent bundle is associated to the BB-representation sl(m,C)/b\mathfrak{sl}(m,\mathbb{C})/\mathfrak{b}, whose diagonal weights are tbta−1t_b t_a^{-1} for a<ba<b. The unipotent radical acts unipotently on this representation and therefore has determinant one. Its determinant character is consequently

∏a<btbta=∏k=1mtk(k−1)−(m−k)=∏k=1mtk2k−m−1.\prod_{a<b}\frac{t_b}{t_a}=\prod_{k=1}^{m}t_k^{(k-1)-(m-k)}=\prod_{k=1}^{m}t_k^{2k-m-1}.

The fiber of Lk\mathcal{L}_k at the standard flag has character tkt_k, so the tensor products in (21) have exactly this BB-character. This proves an isomorphism of determinant line bundles, without asserting a holomorphic splitting of the tangent bundle into root line bundles. Using

c1(T)=c1(det⁡T),c1(L⊗Q)=c1(L)+c1(Q),c1(L∗)=−c1(L),c_1(T)=c_1(\det T),\qquad c_1(L\otimes Q)=c_1(L)+c_1(Q),\qquad c_1(L^*)=-c_1(L),

gives

c1(Fm)=∑a<b(c1(Lb)−c1(La))=∑k=1m(2k−m−1)c1(Lk).(22)\begin{aligned} c_1(\mathcal{F}_m)&=\sum_{a<b}\bigl(c_1(\mathcal{L}_b)-c_1(\mathcal{L}_a)\bigr)\\ &=\sum_{k=1}^{m}(2k-m-1)c_1(\mathcal{L}_k). \tag*{(22)} \end{aligned}

Lemma 4.1. For every i<ji<j, the anticanonical degree along CijC_{ij} is

∫Cijc1(Fm)=2(j−i).(23)\int_{C_{ij}}c_1(\mathcal{F}_m)=2(j-i). \tag*{(23)}

Proof. Put W=span⁡C{ϵi,ϵj}W=\operatorname{span}_{\mathbb{C}}\{\epsilon_i,\epsilon_j\} and regard CijC_{ij} as the projective line of lines in WW. The tautological exact sequence is

0⟶O(−1)⟶W⟶Q⟶0.0\longrightarrow\mathcal{O}(-1)\longrightarrow W\longrightarrow Q\longrightarrow0.

Here WW is the trivial rank-two bundle. The line bundle QQ has degree 1, and hence is isomorphic to O(1)\mathcal{O}(1). Construction (15) identifies

Li∣Cij≃O(−1),Lj∣Cij≃Q≃O(1).\mathcal{L}_i|_{C_{ij}}\simeq\mathcal{O}(-1),\qquad\mathcal{L}_j|_{C_{ij}}\simeq Q\simeq\mathcal{O}(1).

For i<k<ji<k<j, the class of the fixed vector ϵk\epsilon_k gives a nowhere-vanishing section of Lk∣Cij\mathcal{L}_k|_{C_{ij}}. For k<ik<i or k>jk>j, the successive quotient is fixed. All these other line bundles are therefore trivial, and

deg⁡(Li∣Cij)=−1,deg⁡(Lj∣Cij)=1,deg⁡(Lk∣Cij)=0(k≠i,j).\deg(\mathcal{L}_i|_{C_{ij})=-1,\qquad\deg(\mathcal{L}_j|_{C_{ij}})=1,\qquad\deg(\mathcal{L}_k|_{C_{ij}})=0\quad(k\ne i,j).}

Substituting into (22) yields

−(2i−m−1)+(2j−m−1)=2(j−i).-(2i-m-1)+(2j-m-1)=2(j-i).

Lemma 4.2. For an invariant Hermitian metric on Fm\mathcal{F}_m and a unit vector eij∈Vije_{ij}\in V_{ij},

P(eij,e‾ij)=2(j−i)xij.(24)P(e_{ij},\overline{e}_{ij})=\frac{2(j-i)}{x_{ij}}. \tag*{(24)}

Consequently,

s=∑i<j2(j−i)xij.(25)s=\sum_{i<j}\frac{2(j-i)}{x_{ij}}. \tag*{(25)}

Proof. The ambient first Chern Ricci form ρ(1)\rho^{(1)} is invariant under SU⁡(m)\operatorname{SU}(m) because the metric and its Chern connection are invariant. Since CijC_{ij} has real dimension two and ω∣Cij\omega|_{C_{ij}} is nowhere zero, write

ρ(1)∣Cij=fijω∣Cij.\rho^{(1)}|_{C_{ij}}=f_{ij}\omega|_{C_{ij}}.

Both forms are SijS_{ij}-invariant, so fijf_{ij} is invariant. Transitivity on the connected sphere makes it a real constant rijr_{ij}. Thus invariance, together with the degree calculation, determines the pointwise ratio. Equations (5), (17), and (23) give

rij2πxij=∫Cijρ(1)=2π∫Cijc1(Fm)=4π(j−i).r_{ij}2\pi x_{ij}=\int_{C_{ij}}\rho^{(1)} =2\pi\int_{C_{ij}}c_{1}(\mathcal{F}_{m}) =4\pi(j-i).

Hence

rij=2(j−i)xij.r_{ij}=\frac{2(j-i)}{x_{ij}}.

For a unit tangent vector,

ω(eij,eˉij)=−1,ρ(1)(eij,eˉij)=−1P(eij,eˉij),\omega(e_{ij},\bar{e}_{ij})=\sqrt{-1},\qquad\rho^{(1)}(e_{ij},\bar{e}_{ij})=\sqrt{-1}P(e_{ij},\bar{e}_{ij}),

which proves (24). The orthogonal root lines form a unitary basis at oo, so taking the trace gives (25). Invariance gives the same scalar at every point. □\square

The form restricted in this argument is the first Chern Ricci form of the ambient tangent bundle, not the Ricci form of the sphere’s own tangent bundle. This distinction accounts for the factor j−ij-i: the intrinsic tangent bundle of CP1\mathbb{CP}^{1} has degree two, whereas the restricted ambient anticanonical bundle has degree 2(j−i)2(j-i).

Proof of the main theorem and boundary cases

Proof of Theorem 1.1. Suppose that an invariant Hermitian metric satisfies Cα,β≡cC_{\alpha,\beta}\equiv c with β≠0\beta\ne0. The metric is balanced by Lemma 3.1, so Lemma 2.2 applies to the unitary root basis {eij:i<j}\{e_{ij}:i<j\}. Substitution of (19) and (25) gives

∑i<j2xij=2n+1∑i<j2(j−i)xij.\sum_{i<j}\frac{2}{x_{ij}}=\frac{2}{n+1}\sum_{i<j}\frac{2(j-i)}{x_{ij}}.

Multiplying by (n+1)/2(n+1)/2 and moving all terms to one side, we obtain

∑i<jn+1−2(j−i)xij=0.(26)\sum_{i<j}\frac{n+1-2(j-i)}{x_{ij}}=0. \tag*{(26)}

For each root, j−i≤m−1j-i\le m-1. Therefore

n+1−2(j−i)≥m(m−1)2+1−2(m−1)=(m−2)(m−3)2≥0.(27)\begin{aligned} n+1-2(j-i)\ge\frac{m(m-1)}{2}+1-2(m-1) \\ &=\frac{(m-2)(m-3)}{2}\ge0. \tag*{(27)} \end{aligned}

Every denominator in (26) is strictly positive. For any simple root, j=i+1j=i+1, its numerator is n+1−2=n−1>0n+1-2=n-1>0, since m≥3m\ge3 gives n≥3n\ge3. Thus the left-hand side of (26) is strictly positive, contradicting its vanishing. This proves the theorem. □\square

The proof finds an obstruction with no dependence on parameters. Comparing the spherical average against the sum over root directions eliminates α\alpha and cc. No special ratio between α\alpha and β\beta requires inspection. When m=3m=3, the highest-root coefficient is zero. The simple-root coefficients remain positive. When m≥4m\ge4, every coefficient in (26) is strictly positive.

Example. On F3=SU(3)/T2\mathcal{F}_{3}=\mathrm{SU}(3)/T^{2}, set

x=x12,y=x23,z=x13.x=x_{12},\qquad y=x_{23},\qquad z=x_{13}.

The directional values are

(H12,H23,H13)=(2/x,2/y,2/z),(H_{12},H_{23},H_{13})=(2/x,2/y,2/z),

and

(P12,P23,P13)=(2/x,2/y,4/z).(P_{12}, P_{23}, P_{13}) = (2/x, 2/y, 4/z).

Since n=3n = 3, the necessary trace identity would be

2x+2y+2z=12(2x+2y+4z).\frac{2}{x} + \frac{2}{y} + \frac{2}{z} = \frac{1}{2}\left(\frac{2}{x} + \frac{2}{y} + \frac{4}{z}\right).

It reduces to

1x+1y=0,\frac{1}{x} + \frac{1}{y} = 0,

which is impossible. Thus the threefold obstruction holds for every positive triple (x,y,z)(x, y, z); it does not depend on the scale in the nonsimple root direction. In particular, the previously studied Wallach metric is one member of the excluded family.

Remark 1 (The Kähler subcone and the coefficient restriction). In the normalization above, invariant Kähler metrics satisfy the additive relations

xik=xij+xjk,i<j<k.x_{ik} = x_{ij} + x_{jk}, \qquad i < j < k.

The proof never uses these relations, so it treats the full Hermitian cone. To see concretely why β≠0\beta\ne0 is needed, choose

xij=2(j−i)λ,λ>0.x_{ij} = \frac{2(j-i)}{\lambda}, \qquad\lambda> 0.

These parameters satisfy the additive relations and define an invariant Kähler metric. Formula (24) gives

P(eij,eˉij)=λP(e_{ij}, \bar{e}_{ij}) = \lambda

in every root direction. Both PP and gg are torus-invariant Hermitian tensors and are diagonal in the root decomposition, so P=λgP = \lambda g. Hence

Cα,0≡αλC_{\alpha,0} \equiv\alpha\lambda

for this metric.

Remark 2 (The one-dimensional case). For m=2m = 2, the flag manifold is CP1\mathbb{CP}^{1} and there is only one scale x12=xx_{12} = x. Its first Ricci curvature equals its holomorphic sectional curvature, both being 2/x2/x. Therefore

Cα,β≡2(α+β)x.C_{\alpha,\beta} \equiv\frac{2(\alpha+\beta)}{x}.

This verifies the sharpness of the dimension restriction, including the case of a zero mixed-curvature constant.

The choice α=0\alpha= 0 in Theorem 1.1 also excludes constant Chern holomorphic sectional curvature on these invariant metrics. The theorem concerns the Chern connection and the standard invariant complex structure. It makes no assertion about non-invariant Hermitian metrics on the same manifold. Its extension to another homogeneous space would require the corresponding tangent-direction curvature evaluations and degree bounds to be established for that space.

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.

References

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

Email address: [email protected]

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

Email address: [email protected]

References

  1. [1]V. Apostolov, J. Davidov, and O. Muškarov, Compact self-dual Hermitian surfaces, Trans. Amer. Math. Soc. 348 (1996), no. 8, 3051–3063. (cited on page 1)DOI
  2. [2]A. Balas, Compact Hermitian manifolds of constant holomorphic sectional curvature, Math. Z. 189 (1985), 193–210. (cited on page 1)DOI
  3. [3]A. Balas and P. Gauduchon, Any Hermitian metric of constant non-positive (Hermitian) holomorphic sectional curvature on a compact complex surface is Kähler, Math. Z. 190 (1985), no. 1, 39–43. (cited on page 1)DOI
  4. [4]W. Boothby, Hermitian manifolds with zero curvature, Michigan Math. J. 5 (1958), no. 2, 229–233. (cited on page 1)DOI
  5. [5]H. Chen, L. Chen, and X. Nie, Chern–Ricci curvatures, holomorphic sectional curvature and Hermitian metrics, Sci. China Math. 64 (2021), no. 4, 763–780. (cited on page 1)
  6. [6]S. Chen and J. Li, Compact balanced threefolds and LCK manifolds with constant holomorphic sectional curvature, preprint, arXiv:2608.05598 (2026). (cited on page 2)arxiv.org/abs/2608.05598
  7. [7]S. Chen and F. Zheng, Hermitian threefolds with constant holomorphic sectional curvature, preprint, arXiv:2609.39282v1, submitted 30 September 2026. (cited on page 2)arxiv.org/abs/2609.39282
  8. [8]S. Chen and F. Zheng, On Hermitian manifolds with constant mixed curvature, Mediterr. J. Math. 22 (2025), article 175. doi:10.1007/s00009-025-02948-6. (cited on pages 2, 3)DOI
  9. [9]W. Chen and K. Tang, Constant kth-mixed curvature, Differential Geom. Appl. 104 (2026), article 102411. doi:10.1016/j.difgeo.2026.102411. (cited on pages 2, 4)DOI
  10. [10]J. Chu, M.-C. Lee, and L.-F. Tam, Kähler manifolds and mixed curvature, Trans. Amer. Math. Soc. 375 (2022), no. 11, 7925–7944. doi:10.1090/tran/8735. (cited on page 2)DOI
  11. [11]J. Chu, M.-C. Lee, and J. Zhu, On Kähler manifolds with non-negative mixed curvature, preprint, arXiv:2408.14043 (2024). (cited on page 2)arxiv.org/abs/2408.14043
  12. [12]J. Davidov, G. Grantcharov, and O. Mushkarov, Curvature properties of the Chern connection of twistor spaces, Rocky Mountain J. Math. 39 (2009), no. 1, 27–48. (cited on page 1)
  13. [13]L. Grama and A. R. Oliveira, Scalar curvatures of invariant almost Hermitian structures on generalized flag manifolds, SIGMA 17 (2021), article 109, 30 pp. doi:10.3842/SIGMA.2021.109. (cited on page 3)DOI
  14. [14]G. Khan, B. Yang, and F. Zheng, The set of all orthogonal complex structures on the flat 6-torus, Adv. Math. 319 (2017), 451–471. (cited on page 1)arxiv.org/abs/1604.05745
  15. [15]Y. Li and F. Zheng, Complex nmanifolds with constant holomorphic sectional curvature, Proc. Amer. Math. Soc. 150 (2022), no. 1, 319–326. (cited on page 1)arxiv.org/abs/2103.09571
  16. [16]S. Matsumura, On the image of MRC fibrations of projective manifolds with semi-positive holomorphic sectional curvature, Pure Appl. Math. Q. 16 (2020), no. 5, 1419–1439. (cited on page 2)DOI
  17. [17]S. Matsumura, On projective manifolds with semi-positive holomorphic sectional curvature, Amer. J. Math. 144 (2022), no. 3, 747–777. (cited on page 2)arxiv.org/abs/1811.04182
  18. [18]L. Ni, The fundamental group, rational connectedness and the positivity of Kähler manifolds, J. Reine Angew. Math. 774 (2021), 267–299. (cited on page 2)DOI
  19. [19]L. Ni, Liouville theorems and a Schwarz lemma for holomorphic mappings between Kähler manifolds, Comm. Pure Appl. Math. 74 (2021), no. 5, 1100–1126. (cited on page 2)
  20. [20]L. Ni and F. Zheng, Comparison and vanishing theorems for Kähler manifolds, Calc. Var. Partial Differential Equations 57 (2018), no. 6, article 151, 31 pp. (cited on page 2)
  21. [21]X. Qin and Y. Tian, Constant Chern holomorphic sectional curvature: rigidity and counterexamples to the flatness conjecture, preprint, arXiv:2609.25762v3, revised 1 October 2026. (cited on page 2)
  22. [22]P. P. Rao and F. Zheng, Pluriclosed manifolds with constant holomorphic sectional curvature, Acta Math. Sin. (Engl. Ser.) 38 (2022), no. 6, 1094–1104. doi:10.1007/s10114-022-1046-1. (cited on page 1)DOI
  23. [23]T. Sato and K. Sekigawa, Hermitian surfaces of constant holomorphic sectional curvature, Math. Z. 205 (1990), 659–668. (cited on page 1)DOI
  24. [24]K. Tang, Holomorphic sectional curvature and Kähler-like metric, Sci. Sin. Math. 51 (2021), no. 12, 2013–2024 (in Chinese). doi:10.1360/SSM-2020-0038. (cited on page 1)DOI
  25. [25]K. Tang, On mixed curvature for Hermitian manifolds, Pacific J. Math. 344 (2026), no. 1, 201–218. doi:10.2140/pjm.2026.344.201. (cited on page 2)DOI
  26. [26]K. Tang, Quasi-positive mixed curvature, vanishing theorems, and rational connectedness, Bull. Lond. Math. Soc. 58 (2026), no. 2, e70294. (cited on page 2)arxiv.org/abs/2405.03895
  27. [27]K. Tang and Z. Wang, Constant kth-mixed curvature on locally conformal Kähler manifolds, preprint, arXiv:2609.05852 (2026). (cited on page 2)
  28. [28]K. Tang Hermitian manifolds with constant holomorphic sectional curvature or real bisectional curvature, preprint, arxiv:2609.14392 (2026). (cited on page 1)arxiv.org/abs/2609.14392
  29. [29]V. Tosatti and X. Yang, An extension of a theorem of Wu–Yau, J. Differential Geom. 107 (2017), no. 3, 573–579. (cited on page 2)DOI
  30. [30]D. Wu and S.-T. Yau, Negative holomorphic curvature and positive canonical bundle, Invent. Math. 204 (2016), no. 2, 595–604. (cited on page 2)DOI
  31. [31]B. Yang and F. Zheng, On curvature tensors of Hermitian manifolds, Comm. Anal. Geom. 26 (2018), no. 5, 1195–1222. doi:10.4310/CAG.2018.v26.n5.a7. (cited on page 4)DOI
  32. [32]X. Yang, RC-positivity, rational connectedness and Yau’s conjecture, Camb. J. Math. 6 (2018), no. 2, 183–212. (cited on page 2)arxiv.org/abs/1708.06713
  33. [33]X. Yang and F. Zheng, On real bisectional curvature for Hermitian manifolds, Trans. Amer. Math. Soc. 371 (2019), 2703–2718. (cited on page 1)
  34. [34]S.-T. Yau, Seminar on differential geometry, Ann. of Math. Stud. 102 (1982), 669–706. (cited on page 2)
  35. [35]S. Zhang and X. Zhang, Compact Kähler manifolds with quasi-positive holomorphic sectional curvature, preprint, arXiv:2311.18779 (2023). (cited on page 2)
  36. [36]F. Zheng, Constant holomorphic sectional curvature conjecture and Fino–Vezzoni conjecture, preprint, arXiv:2511.20035 (2025). (cited on page 2)arxiv.org/abs/2511.20035
  37. [37]W. Zhou and F. Zheng, Hermitian threefolds with vanishing real bisectional curvature, Sci. Sin. Math. 52 (2022), no. 7, 757–764 (in Chinese); English version: arXiv:2103.04296. (cited on page 1)DOI

Paper details

vlnc:W_c8yyvpns
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

Not recorded for this paper

View statement
Looking for related papers…
Contents