betaFeedback

WorksMathematics

Pluricanonical locally conformally almost Kähler metrics, circle-invariant G2G_2 structures, and Laplacian flows

Abstract

We study explicit Hodge evolutions of pluricanonical locally conformally almost Kähler metrics and their relation to locally conformal parallel G2G_2 structures. A compact locally conformal parallel 77-manifold with a parallel non-zero Lee form, carrying a structure-preserving free circle action of constant orbit length orthogonal to the Lee field, has a 66-dimensional quotient which is pluricanonical with a parallel Lee form and a non-integrable almost complex structure. The fundamental forms upstairs and downstairs are Laplacian eigenforms with an explicit eigenvalue ratio. Conversely, for a fixed normalized SU(3)SU(3) completion, we characterize the curvature of a circle-invariant locally conformal parallel lift. The lift equations force a pluricanonical base to have parallel Lee form, even locally, and exclude strict integrable pluricanonical bases. Independently, the pluricanonical eigenform identity gives a shrinking solution of He's full Hodge equation and an immortal d∗dd^*d solution contracting the Lee plane.

Introduction

A locally conformally symplectic (LCS) structure on a manifold MM of real dimension 2n≥42n \ge4 is a pair (F,θ)(F,\theta), where FF is a nondegenerate 2-form and θ\theta is a closed 1-form such that

dF=θ∧F.dF = \theta\wedge F.

The form θ\theta is called the Lee form. If FF is compatible with an almost complex structure JJ, the Riemannian metric g=F(⋅,J⋅)g = F(\cdot,J\cdot) is locally conformally almost Kähler (LCAK). When JJ is integrable, the metric is locally conformally Kähler (LCK). An LCK metric with parallel Lee form is called Vaisman.

Pluricanonical LCAK metrics were introduced in [1] as an extension of the pluricanonical condition considered by Kokarev in LCK geometry [2]. The condition requires DθD\theta to be JJ-anti-invariant and the image of the Nijenhuis tensor to be orthogonal to the Lee field T=θ♯T = \theta^\sharp and the anti-Lee field JTJT (here DD is the Levi-Civita connection). On compact complex manifolds, pluricanonical LCK metrics are Vaisman [3]. In the non-integrable case the Lee form need not be parallel, even on compact manifolds. We call a pluricanonical LCAK structure strict if θ≠0\theta\ne0, and almost Vaisman if Dθ=0D\theta= 0.

One of the identities established for pluricanonical LCAK metrics in [1] is

ΔgF=(n−1)∥θ∥2F(1)\Delta^g F = (n-1)\lVert\theta\rVert^2 F \tag*{(1)}

where Δg=dδg+δgd\Delta^g = d\delta^g + \delta^g d is the Hodge Laplacian and ∥θ∥2=g(θ,θ)\lVert\theta\rVert^2 = g(\theta,\theta) is constant (here δg\delta^g is the codifferential). This places the fundamental 2-form in a distinguished class for geometric evolutions. It also suggests asking whether pluricanonical structures can arise naturally as quotients of seven-dimensional geometries with analogous differential identities. We address both questions, keeping track of the metric used by the Laplacian and of the distinction between integrable and non-integrable almost complex structures. Our geometric starting point is a positive 3-form φ\varphi with

dφ=ϑ∧φ,d(∗φφ)=43ϑ∧∗φφ,dϑ=0,(2)d\varphi= \vartheta\wedge\varphi,\qquad d(*_\varphi\varphi) = \frac{4}{3}\vartheta\wedge*_\varphi\varphi,\qquad d\vartheta= 0, \tag*{(2)}

2020 Mathematics Subject Classification. 53C15, 53C25, 53C44, 53C55. where ϑ\vartheta is the Lee form of φ\varphi and ∗φ\ast\varphi is the Hodge operator induced by φ\varphi. These are the locally conformal parallel equations in our Lee normalization. Ivanov–Parton–Piccinni [4] describe compact structures in a Gauduchon gauge and their nearly Kähler product covers. A structure-preserving circle action supplies an almost Hermitian quotient. The question relevant here is whether that quotient satisfies the full pluricanonical condition, including the Nijenhuis orthogonality requirement. The following theorem gives an affirmative answer.

Theorem (Theorem 2). Let M7M^7 be compact and connected, with a locally conformal parallel G2G_2 structure φ\varphi whose nonzero Lee form ϑ\vartheta is parallel. Suppose that a free circle action preserves φ\varphi and its generator XX has constant length c>0c>0 and satisfies ϑ(X)=0\vartheta(X)=0. Then B=M/S1B=M/S^1, with its quotient metric, carries a strict almost Vaisman structure with a non-integrable almost complex structure, given by

π∗F=ιX/cφ,π∗θ=ϑ.\pi^*F=\iota_{X/c}\varphi,\qquad\pi^*\theta=\vartheta.

Moreover,

ΔgBF=2∥ϑ∥2F,Δgφφ=43∥ϑ∥2φ.\Delta^{g_B}F=2\lVert\vartheta\rVert^2F,\qquad\Delta^{g_\varphi}\varphi=\frac{4}{3}\lVert\vartheta\rVert^2\varphi.

The universal cover of MM is isometric to a product R×N\mathbb{R}\times N, where NN is homothetic to S3×S3S^3\times S^3 with its canonical nearly Kähler structure.

The converse problem has additional differential and topological constraints. First, we fix a normalized SU(3)\mathrm{SU}(3) completion (F,ψ+,ψ−)(F,\psi_+,\psi_-) of a 6-dimensional LCAK structure. Proposition 3 identifies the circle-lift equations as

κ∧F+dψ+=θ∧ψ+,dψ−=43θ∧ψ−,κ∧ψ−=13θ∧F2.(3)\kappa\wedge F+d\psi_+=\theta\wedge\psi_+,\qquad d\psi_-=\frac{4}{3}\theta\wedge\psi_-,\qquad\kappa\wedge\psi_-=\frac{1}{3}\theta\wedge F^2. \tag*{(3)}

Here κ\kappa is the candidate curvature of the unit vertical 1-form. For a circle parameter of period 2π2\pi and fixed radius c>0c>0, a global circle lift exists precisely when κ\kappa is closed and [κ2πc]\left[\frac{\kappa}{2\pi c}\right] is integral. In particular, the topological triviality of the canonical line bundle alone does not solve the lift problem. The circle-bundle differentiation formulas underlying (3) belong to the established SU(3)\mathrm{SU}(3)–G2G_2 framework of Chiossi–Salamon [5]. The lift equations impose the following additional rigidity on a pluricanonical LCAK base.

Theorem (Theorem 5). If a pluricanonical LCAK 6-manifold has a normalized SU(3)\mathrm{SU}(3) completion and a closed curvature form κ\kappa satisfying the circle-lift equations, then

Dθ=0,ιTκ=0,D\theta=0,\qquad\iota_T\kappa=0,

where T=θ♯T=\theta^\sharp.

Its proof uses the constant length and coclosedness of the lifted Lee form, conformal Ricci flatness, and a pointwise Bochner identity (see [6] for instance). The curvature-norm identity in Proposition 4 also shows that an integrable pluricanonical base satisfying these equations with dκ=0d\kappa=0 has θ=κ=0\theta=\kappa=0. Then, we give an explicit compact lift from S2×S3×S1S^2\times S^3\times S^1, obtained by reducing the product of the canonical nearly Kähler S3×S3S^3\times S^3 with a circle, to illustrate the pluricanonical quotient and the integral curvature condition.

Moreover, the eigenform identity (1) leads to explicit solutions of the two flows introduced by He [7]. On a connected manifold, the full Hodge-Laplacian flow shrinks homothetically, whereas the d∗dd^*d flow fixes the horizontal part of the fundamental form and contracts the Lee plane.

Theorem (Theorems 6 and 8). Let (g0,J,F0,θ)(g_0,J,F_0,\theta) be a strict pluricanonical LCAK structure in real dimension 2n≥42n\geq4. Set

r2=∥θ∥g02,λ=(n−1)r2,FV=r−2θ∧Jθ,FH=F0−FV.r^2=\lVert\theta\rVert_{g_0}^2,\qquad\lambda=(n-1)r^2,\qquad F_V=r^{-2}\theta\wedge J\theta,\qquad F_H=F_0-F_V.

With JJ fixed and g(t)=F(t)(⋅,J⋅)g(t)=F(t)(\cdot,J\cdot), the two families

F(t)=(1−λt)F0solve ∂tF=−Δg(t)F,t<λ−1,F(t)=(1-\lambda t)F_0\qquad\text{solve }\partial_tF=-\Delta_{g(t)}F,\quad t<\lambda^{-1},
F(t)=FH+(1+λt)−1FVsolve ∂tF=−δg(t)dF,t>−λ−1,F(t)=F_H+(1+\lambda t)^{-1}F_V\qquad\text{solve }\partial_tF=-\delta^{g(t)}dF,\quad t>-\lambda^{-1},

with initial form F0F_0, and remain pluricanonical. Their Lee forms are θ\theta and (1+λt)−1θ(1+\lambda t)^{-1}\theta, respectively. On a compact manifold, the second family is the unique smooth forward solution in He’s fixed-JJ tamed class with initial form F0F_0, on every common interval of existence.

Thus the first solution degenerates at finite positive time, while the second is immortal and converges to the closed rank-(2n−2)(2n-2) form FHF_H. Similar deformations of Vaisman metrics were considered for instance in [8] (see also the references therein).

The paper is organized as follows. In Section 2, we fix some notations and conventions. Then, in Section 3, we recognize the quotient of a compact locally conformal parallel G2G_2 structure, carrying a structure-preserving free circle action, as a pluricanonical structure. Moreover, we study the circle-lift equations on a pluricanonical manifold to a locally conformal parallel G2G_2 structure. Finally, in section 4, we use the eigenform property of pluricanonical LCAK metrics to solve He’s Laplacian flows.

Acknowledgement: The author is very grateful to Tedi Drăghici for the discussion about Donaldson’s idea [9] to connect hypersymplectic structures on 4-manifolds and G2G_2 structures, and for suggesting the direct first-kind LCS construction used in the proof of Theorem 2.

Declaration of AI assistance: The author acknowledges the use of ChatGPT 6.1 Sol (OpenAI) for checking routine calculations and improving manuscript exposition. All mathematical ideas, proof strategies, and results are the sole work of the author, who independently verified every step of the proofs.

Preliminaries

Let (M2n,F,θ)(M^{2n},F,\theta) be a locally conformally symplectic (LCS for short) manifold of real dimension 2n2n, where FF is a non-degenerate 2-form and θ\theta is a closed 1-form such that dF=θ∧FdF=\theta\wedge F. The 1-form θ\theta is called the Lee form of the LCS structure. We suppose that FF is compatible with an almost complex structure JJ, that is g:=F(⋅,J⋅)g:=F(\cdot,J\cdot) is a Riemannian metric. The induced metric gg is called a locally conformally almost Kähler (LCAK for short) metric. We denote by DD the Levi-Civita connection of gg. We use the convention (Jα)(X)=−α(JX)(J\alpha)(X)=-\alpha(JX) for a 1-form α\alpha, and vol⁡g=Fnn!\operatorname{vol}_g=\frac{F^n}{n!}. For a covariant 2-tensor ϕ\phi, its JJ-invariant and JJ-anti-invariant parts are ϕJ,±(X,Y)=12(ϕ(X,Y)±ϕ(JX,JY))\phi^{J,\pm}(X,Y)=\frac{1}{2}\left(\phi(X,Y)\pm\phi(JX,JY)\right). Our Nijenhuis tensor convention is

4N(X,Y)=[JX,JY]−[X,Y]−J[JX,Y]−J[X,JY].4N(X,Y)=[JX,JY]-[X,Y]-J[JX,Y]-J[X,JY].

The Hodge Laplacian is defined as Δg=dδg+δgd\Delta_g=d\delta^g+\delta^g d, where δg\delta^g is the codifferential, defined as the adjoint of the exterior derivative dd with respect to the metric gg.

Definition 1. [1] An LCAK metric is pluricanonical if DθD\theta is JJ-anti-invariant and the image of the Nijenhuis tensor is gg-orthogonal to T=θ♯T=\theta^\sharp and JTJT. A pluricanonical LCAK structure is strict if its Lee form is not identically zero.

For such a structure, in addition to (1), we also have

δgθ=0,d∥θ∥2=0,dJθ=−∥θ∥2F+θ∧Jθ,δgF=−(n−1)Jθ,ΔgJθ=(n−1)∥θ∥2Jθ.(4)\delta^g\theta=0,\quad d\lVert\theta\rVert^2=0,\quad dJ\theta=-\lVert\theta\rVert^2F+\theta\wedge J\theta,\quad\delta^g F=-(n-1)J\theta,\quad\Delta^gJ\theta=(n-1)\lVert\theta\rVert^2J\theta. \tag*{(4)}

Moreover, we recall that for a given pluricanonical LCAK structure, the tangent bundle splits into

TM=span⁡{T,JT}⊕H,(5)TM=\operatorname{span}\{T,JT\}\oplus H, \tag*{(5)}

where H=span⁡{T,JT}⊥gH=\operatorname{span}\{T,JT\}^{\perp_g}.

Locally conformal parallel G2G_2 structures

Recognition of pluricanonical quotients

3.1. Recognition of pluricanonical quotients. Let φ\varphi be a positive 3-form on a 7-manifold. Namely, if we choose an orientation and any positive volume form ν\nu, then the symmetric bilinear form BB defined by

Bφ(X,Y)ν=ιXφ∧ιYφ∧φB_\varphi(X,Y)\nu=\iota_X\varphi\wedge\iota_Y\varphi\wedge\varphi

is positive definite, where X,YX,Y are vector fields and ιX\iota_X is the contraction by XX. The metric gφg_\varphi induced by φ\varphi is characterized then by

gφ(X,Y)νgφ=16ιXφ∧ιYφ∧φ.g_\varphi(X,Y)\nu_{g_\varphi}=\frac{1}{6}\iota_X\varphi\wedge\iota_Y\varphi\wedge\varphi.

The dual 4-form ψ\psi is defined as

ψ=∗gφφ.\psi=*_{g_\varphi}\varphi.

The 3-form φ\varphi is called torsion-free if Dgφφ=0D^{g_\varphi}\varphi=0 which is equivalent to dφ=dψ=0d\varphi=d\psi=0 (here DgφD^{g_\varphi} is the Levi-Civita connection of the metric gφg_\varphi). Moreover, φ\varphi is called locally conformally parallel if the following equations are satisfied

dφ=ϑ∧φ,dψ=43ϑ∧ψ,dϑ=0,(6)d\varphi=\vartheta\wedge\varphi,\qquad d\psi=\frac{4}{3}\vartheta\wedge\psi,\qquad d\vartheta=0, \tag*{(6)}

where the closed 1-form ϑ\vartheta is the Lee form. For a compact locally conformal parallel G2\mathrm{G}_2 structure, the Gauduchon representative has a parallel Lee form [4]. Locally, if ϑ=dh\vartheta=dh, the form e−hφe^{-h}\varphi is torsion-free and its metric is e−2h/3gφe^{-2h/3}g_\varphi. It turns out that a circle quotient of a locally conformal parallel structure yields a pluricanonical base under natural metric assumptions.

Theorem 2. Let (M,φ)(M,\varphi) be a 7-dimensional compact and connected G2\mathrm{G}_2 manifold satisfying (6) with ϑ≠0\vartheta\ne0 and Dgφϑ=0D^{g_\varphi}\vartheta=0. Suppose a free circle action preserves φ\varphi, its generator XX has constant length c>0c>0, and ϑ(X)=0\vartheta(X)=0. Set V=X/cV=X/c. Then the quotient metric and the form

π∗F=ιVφ\pi^*F=\iota_V\varphi

define an almost Vaisman structure (gB,JB,F,θ)(g_B,J_B,F,\theta) on B=M/S1B=M/S^1, where π∗θ=ϑ\pi^*\theta=\vartheta, and π:M→B\pi:M\to B is the quotient map. Furthermore, we have

ΔgBF=2∥ϑ∥2F,Δgφφ=4∥ϑ∥23φ.(7)\Delta^{g_B}F=2\lVert\vartheta\rVert^2F,\qquad\Delta^{g_\varphi}\varphi=\frac{4\lVert\vartheta\rVert^2}{3}\varphi. \tag*{(7)}

The induced almost complex structure JBJ_B is non-integrable. Moreover, the universal cover of MM has a nearly Kähler factor homothetic to the canonical S3×S3S^3\times S^3.

Proof. Since the circle action preserves φ\varphi, it preserves the metric gφg_\varphi. Let π:M→B=M/S1\pi:M\to B=M/S^1 be the quotient map. The horizontal, invariant 2-form ιVφ\iota_V\varphi descends to FF on BB. Pointwise G2\mathrm{G}_2 algebra gives an almost complex structure JBJ_B compatible with the quotient metric gBg_B, with F(Y,Z)=gB(JBY,Z)F(Y,Z)=g_B(J_BY,Z). The closed form ϑ\vartheta descends also on BB because ϑ(V)=0\vartheta(V)=0. We write π∗θ=ϑ\pi^*\theta=\vartheta. In the following, we suppress pullbacks when no confusion can arise.

Locally, ϑ=dh\vartheta=dh and e−hφe^{-h}\varphi is parallel for e−2h/3gφe^{-2h/3}g_\varphi. The conformal connection formula therefore gives the global tensor identity

DYgφφ=13(Y♭∧ιϑ♯φ−ϑ∧ιYφ).(8)D^{g_\varphi}_Y\varphi=\frac{1}{3}\left(Y^\flat\wedge\iota_{\vartheta^\sharp}\varphi-\vartheta\wedge\iota_Y\varphi\right). \tag*{(8)}

Since Dgφϑ♯=0D^{g_\varphi}\vartheta^\sharp=0, this implies Dϑ♯gφφ=Lϑ♯φ=0D^{g_\varphi}_{\vartheta^\sharp}\varphi=\mathcal{L}_{\vartheta^\sharp}\varphi=0. Moreover, LVϑ=0\mathcal{L}_V\vartheta=0 and LVgφ=0\mathcal{L}_Vg_\varphi=0, so [V,ϑ♯]=0[V,\vartheta^\sharp]=0. Thus ϑ♯\vartheta^\sharp is horizontal and projects to T=θ♯T=\theta^\sharp. For basic horizontal vector fields Y,ZY,Z, the horizontal connection formula gives

gB(Dπ∗YgBT,π∗Z)=g(DYgφϑ♯,Z)=0.g_B(D^{g_B}_{\pi_*Y}T,\pi_*Z)=g(D^{g_\varphi}_Y\vartheta^\sharp,Z)=0.

Consequently DgBθ=0D^{g_B}\theta=0.

Cartan’s formula gives

dF=−ιVdφ=θ∧F.dF=-\iota_Vd\varphi=\theta\wedge F.

Let S=ιϑ♯φS=\iota_{\vartheta^\sharp}\varphi. Then

dS=−∥ϑ∥2φ+ϑ∧S,π∗(JBθ)=ιϑ♯ιVφ=−ιVS.dS=-\lVert\vartheta\rVert^2\varphi+\vartheta\wedge S,\qquad\pi^*(J_B\theta)=\iota_{\vartheta^\sharp}\iota_V\varphi=-\iota_VS.

Since LVS=0\mathcal{L}_VS=0, differentiation yields

dJBθ=−∥ϑ∥2F+θ∧JBθ.dJ_B\theta=-\lVert\vartheta\rVert^2F+\theta\wedge J_B\theta.

Also, Lϑ♯F=0\mathcal{L}_{\vartheta^\sharp}F=0 upstairs, and hence LTF=0\mathcal{L}_TF=0 downstairs. The form dJBϑ\mathrm{d}J_B\vartheta is JBJ_B-invariant. Since dϑ=0\mathrm{d}\vartheta=0, it follows that

4ϑ(N(Y,Z))=−dJBϑ(JBY,Z)−dJBϑ(Y,JBZ)=0,4\vartheta(N(Y,Z))=-\mathrm{d}J_B\vartheta(J_BY,Z)-\mathrm{d}J_B\vartheta(Y,J_BZ)=0,
4JBϑ(N(Y,Z))=dJBϑ(Y,Z)−dJBϑ(JBY,JBZ)=0.4J_B\vartheta(N(Y,Z))=\mathrm{d}J_B\vartheta(Y,Z)-\mathrm{d}J_B\vartheta(J_BY,J_BZ)=0.

Together with DgBϑ=0D^{g_B}\vartheta=0, these identities show that the quotient is strict almost Vaisman. Then, the first equation in (7) follows from (1) in dimension 2n=62n=6.

To get the second equation in (7), we first trace (8). We get δgφφ=−43S\delta^{g_\varphi}\varphi=-\frac{4}{3}S. Since Dgφφ=0D^{g_\varphi}\varphi=0, we obtain

Δgφφ=−43dS+δgφ(ϑ∧φ)=43∥ϑ∥2φ−43ϑ∧S+43ϑ∧S=43∥ϑ∥2φ.\begin{aligned} \Delta^{g_\varphi}\varphi=-\frac{4}{3}\mathrm{d}S+\delta^{g_\varphi}(\vartheta\wedge\varphi) \\ &=\frac{4}{3}\lVert\vartheta\rVert^2\varphi-\frac{4}{3}\vartheta\wedge S+\frac{4}{3}\vartheta\wedge S=\frac{4}{3}\lVert\vartheta\rVert^2\varphi. \end{aligned}

To prove the non-integrability, we use the natural normalized SU(3)SU(3) completion on BB defined by

ψ+=φ−V♭∧F,ψ−=−ιV(∗gφφ).\psi_+=\varphi-V^\flat\wedge F,\qquad\psi_-=-\iota_V(*_{g_\varphi}\varphi).

These forms are horizontal and invariant, so they descend on BB. Pointwise G2G_2 algebra gives

∗gφφ=12F2−V♭∧ψ−,Ψ=ψ++−1ψ−∈Λ3,0(B).*_{g_\varphi}\varphi=\frac{1}{2}F^2-V^\flat\wedge\psi_-,\qquad\Psi=\psi_++\sqrt{-1}\psi_-\in\Lambda^{3,0}(B).

Comparing the vertical and horizontal parts of d(∗gφφ)=43ϑ∧∗gφφ\mathrm{d}(*_{g_\varphi}\varphi)=\frac{4}{3}\vartheta\wedge*_{g_\varphi}\varphi gives

dψ−=43ϑ∧ψ−,dV♭∧ψ−=13ϑ∧F2.\mathrm{d}\psi_-=\frac{4}{3}\vartheta\wedge\psi_-,\qquad\mathrm{d}V^\flat\wedge\psi_-=\frac{1}{3}\vartheta\wedge F^2.

The second identity implies (dV♭)JB,−=13ιTψ−(\mathrm{d}V^\flat)_{J_B,-}=\frac{1}{3}\iota_T\psi_-. Indeed, wedging with ψ−\psi_- is an isomorphism on real JBJ_B-anti-invariant 2-forms, while JBJ_B-invariant 2-forms wedge trivially with ψ±\psi_\pm. The standard SU(3)SU(3) identities

(ιTψ−)∧ψ−=ϑ∧F2,(ιTψ−)∧ψ+=−2∗gBϑ(\iota_T\psi_-)\wedge\psi_-=\vartheta\wedge F^2,\qquad(\iota_T\psi_-)\wedge\psi_+=-2*_{g_B}\vartheta

therefore give

dV♭∧ψ+=−23∗gBϑ.(9)\mathrm{d}V^\flat\wedge\psi_+=-\frac{2}{3}*_{g_B}\vartheta. \tag*{(9)}

Suppose that JBJ_B is integrable. The complex 4-form

Q=dΨ−43ϑ∧ΨQ=\mathrm{d}\Psi-\frac{4}{3}\vartheta\wedge\Psi

is then of type (3,1)(3,1). The equation for dψ−\mathrm{d}\psi_- makes QQ real, so Q=0Q=0. In particular,

dψ+=43ϑ∧ψ+.(10)\mathrm{d}\psi_+=\frac{4}{3}\vartheta\wedge\psi_+. \tag*{(10)}

Since DgBϑ=0D^{g_B}\vartheta=0, we obtain using Equations (9) and (10)

0=−23d(∗gBϑ)=d(dV♭∧ψ+)=dV♭∧dψ+=43ϑ∧dV♭∧ψ+=−89∥ϑ∥2vol⁡gB,\begin{aligned} 0=-\frac{2}{3}\mathrm{d}(*_{g_B}\vartheta)=\mathrm{d}(\mathrm{d}V^\flat\wedge\psi_+) \\ &=\mathrm{d}V^\flat\wedge\mathrm{d}\psi_+=\frac{4}{3}\vartheta\wedge\mathrm{d}V^\flat\wedge\psi_+=-\frac{8}{9}\lVert\vartheta\rVert^2\operatorname{vol}_{g_B}, \end{aligned}

a contradiction. Hence JBJ_B is non-integrable.

It remains only to identify the global nearly Kähler factor. By [4 Proposition 3.3 and Corollary 3.4], the universal cover of MM is a Riemannian product M~=R×N\widetilde{M}=\mathbb{R}\times N, with

gφ=dt2+gN,ϑ=−∥ϑ∥dt,φ=dt∧ωN+ΨN,+,(11)g_\varphi=\mathrm{d}t^2+g_N,\qquad\vartheta=-\lVert\vartheta\rVert\mathrm{d}t,\qquad\varphi=\mathrm{d}t\wedge\omega_N+\Psi_{N,+}, \tag*{(11)}

where NN is a compact, simply connected, and strictly nearly Kähler manifold. Let E=∂tE = \partial_t. Then, the induced forms have the normalization

ωN=ιEφ,dωN=∥ϑ∥ΨN,+,dΨN,+=0,\omega_N = \iota_E\varphi,\qquad d\omega_N = \lVert\vartheta\rVert\Psi_{N,+},\qquad d\Psi_{N,+} = 0,
dΨN,−=−2∥ϑ∥3ωN2,ΨN,−=∗gNΨN,+,ωN∧ΨN,+=0.(12)d\Psi_{N,-} = -\frac{2\lVert\vartheta\rVert}{3}\omega_N^2,\qquad\Psi_{N,-} = \ast_{g_N}\Psi_{N,+},\qquad\omega_N \wedge\Psi_{N,+} = 0. \tag*{(12)}

The lift of VV is tangent to NN because ϑ(V)=0\vartheta(V) = 0. Since ϑ♯=−∥ϑ∥∂t\vartheta^\sharp= -\lVert\vartheta\rVert\partial_t and [ϑ♯,V]=0[\vartheta^\sharp,V] = 0, this lifted field is independent of tt and restricts to a unit Killing field on NN. After a constant rescaling, [10 Theorem 1.1] applies. As NN is simply connected, it is homothetic to S3×S3S^3 \times S^3 with its canonical nearly Kähler structure. □

The circle-lift equations

Let (B,g,J,F,θ)(B,g,J,F,\theta) be a 6-dimensional LCAK manifold. Suppose that a normalized SU(3)SU(3) completion (F,ψ+,ψ−)(F,\psi_+,\psi_-) is fixed, with

F∧ψ±=0,ψ+∧ψ−=23F3,∗gψ+=ψ−.(13)F \wedge\psi_\pm= 0,\qquad\psi_+ \wedge\psi_- = \frac{2}{3}F^3,\qquad\ast_g\psi_+ = \psi_-. \tag*{(13)}

In particular, Ψ=ψ++iψ−\Psi= \psi_+ + i\psi_- is a nowhere vanishing normalized complex volume 3-form. Such an SU(3)SU(3) completion exists when the (smooth) canonical line bundle Λ3,0(B)\Lambda^{3,0}(B) is trivial which is equivalent to c1(B,J)=0∈H2(B,Z)c_1(B,J) = 0 \in H^2(B,\mathbb{Z}). Let π:P⟶B\pi:P \longrightarrow B be some principal circle bundle, whose principal generator XX has period 2π2\pi. Fix a connection AA and a constant c>0c>0, and set η=cA\eta= cA. The connection metric is gP=π∗g+η2g_P = \pi^*g + \eta^2, and XX has length cc (the connection AA is normalized by A(X)=1A(X) = 1). Define the following forms (pullbacks are suppressed below):

φ=η∧F+ψ+,ψ=∗gPφ=12F2−η∧ψ−.(14)\varphi= \eta\wedge F + \psi_+,\qquad\psi= \ast_{g_P}\varphi= \frac{1}{2}F^2 - \eta\wedge\psi_-. \tag*{(14)}

These formulas define a positive G2G_2 structure with metric gPg_P (this can be checked pointwise).

Proposition 3. For the fixed completion (13), the forms (14) satisfy (6), with ϑ=π∗θ\vartheta= \pi^*\theta, if and only if

κ∧F+dψ+=θ∧ψ+,(15)\kappa\wedge F + d\psi_+ = \theta\wedge\psi_+, \tag*{(15)}
dψ−=43θ∧ψ−,(16)d\psi_- = \frac{4}{3}\theta\wedge\psi_-, \tag*{(16)}
κ∧ψ−=13θ∧F2.(17)\kappa\wedge\psi_- = \frac{1}{3}\theta\wedge F^2. \tag*{(17)}

Equivalently, writing T=θ♯T = \theta^\sharp and

D=dψ+−43θ∧ψ+,\mathcal{D} = d\psi_+ - \frac{4}{3}\theta\wedge\psi_+,

one must have

dψ−=43θ∧ψ−,D∈Λ02,2(B),κ=13ιTψ−+κ0J,+,κ0J,+=∗gD,(18)d\psi_- = \frac{4}{3}\theta\wedge\psi_-,\qquad\mathcal{D} \in\Lambda^{2,2}_0(B),\qquad\kappa= \frac{1}{3}\iota_T\psi_- + \kappa_0^{J,+},\qquad\kappa_0^{J,+} = \ast_g\mathcal{D}, \tag*{(18)}

where Λ02,2(B)={D∈Λ2,2(B)∣D∧F=0}\Lambda^{2,2}_0(B) = \{D \in\Lambda^{2,2}(B) \mid D \wedge F = 0\} and κ0J,+\kappa_0^{J,+} is the trace-free JJ-invariant part of κ\kappa. The candidate curvature κ\kappa is unique. A global circle lift with principal radius cc exists precisely when this curvature is closed and

[κ2πc]∈im⁡(H2(B,Z)⟶H2(B,R)).(19)\left[\frac{\kappa}{2\pi c}\right] \in\operatorname{im}\left(H^2(B,\mathbb{Z}) \longrightarrow H^2(B,\mathbb{R})\right). \tag*{(19)}

Proof. Differentiating ψ\psi in (14), we obtain

dψ=θ∧F2−κ∧ψ−+η∧dψ−.d\psi= \theta\wedge F^2 - \kappa\wedge\psi_- + \eta\wedge d\psi_-.

Comparing the horizontal and vertical components to

43θ∧ψ=23θ∧F2+43η∧θ∧ψ−,\frac{4}{3}\theta\wedge\psi= \frac{2}{3}\theta\wedge F^2 + \frac{4}{3}\eta\wedge\theta\wedge\psi_-,

we obtain (16) and (17). In a similar way, to obtain (15), we differentiate φ\varphi in (14). Moreover, the map LF:Λ2(B)⟶Λ4(B)L_F:\Lambda^2(B)\longrightarrow\Lambda^4(B), ζ⟼ζ∧F\zeta\longmapsto\zeta\wedge F, is an isomorphism in dimension six, so (15) determines κ\kappa uniquely.

Furthermore, differentiating F∧ψ+=0F\wedge\psi_+=0 gives that F∧dψ+=0F\wedge d\psi_+=0, and hence κ∧F2=0\kappa\wedge F^2=0. Decompose

κ=κ0J,−+κ0J,+.\kappa=\kappa_0^{J,-}+\kappa_0^{J,+}.

In an orthonormal SU(3)SU(3) coframe one has

(ιTψ−)∧ψ−=θ∧F2,(ιTψ−)∧F=−θ∧ψ+,(ιTψ−)∧ψ+=−2∗gθ.(20)(\iota_T\psi_-)\wedge\psi_-=\theta\wedge F^2,\qquad (\iota_T\psi_-)\wedge F=-\theta\wedge\psi_+,\qquad (\iota_T\psi_-)\wedge\psi_+=-2\ast_g\theta. \tag*{(20)}

Wedging with ψ−\psi_- is an isomorphism from the real JJ-anti-invariant 2-forms to Λ5(B)\Lambda^5(B), whereas κ0J,+∧ψ+=0\kappa_0^{J,+}\wedge\psi_+=0. Thus (17) gives 3κ0J,−=ιTψ−3\kappa_0^{J,-}=\iota_T\psi_-. Substitution into (15) gives D=−κ0J,+∧F=∗gκ0J,+\mathcal{D}=-\kappa_0^{J,+}\wedge F=\ast_g\kappa_0^{J,+}, proving (18). The converse follows from the same computations.

The condition (19) is the Chern class condition given that the curvature of AA is κc\frac{\kappa}{c}.

Proposition 4. If the system (15)–(17) has a closed curvature, then

δgθ=−43∥θ∥2+32∥κ0J,+∥2.(21)\delta^g\theta=-\frac{4}{3}\lVert\theta\rVert^2+\frac{3}{2}\lVert\kappa_0^{J,+}\rVert^2. \tag*{(21)}

On a pluricanonical base this implies

∥κ0J,+∥2=89∥θ∥2,∥κ∥2=109∥θ∥2.\lVert\kappa_0^{J,+}\rVert^2=\frac{8}{9}\lVert\theta\rVert^2,\qquad \lVert\kappa\rVert^2=\frac{10}{9}\lVert\theta\rVert^2.

If the base is integrable and pluricanonical, then θ=0\theta=0 and κ=0\kappa=0. More generally, the same conclusion holds for a compact LCK base.

Proof. By (20), κ∧ψ+=−23∗gθ\kappa\wedge\psi_+=-\frac{2}{3}\ast_g\theta. Recall that dψ+=43θ∧ψ++∗gκ0J,+d\psi_+=\frac{4}{3}\theta\wedge\psi_++\ast_g\kappa_0^{J,+}. Since dκ=0d\kappa=0, we obtain

d(κ∧ψ+)=κ∧dψ+=43κ∧θ∧ψ++κ∧∗gκ0J,+=−89θ∧∗gθ+κ∧∗gκ0J,+=(−89∥θ∥2+∥κ0J,+∥2)vol⁡g.\begin{aligned} d(\kappa\wedge\psi_+)&=\kappa\wedge d\psi_+\\ &=\frac{4}{3}\kappa\wedge\theta\wedge\psi_++\kappa\wedge\ast_g\kappa_0^{J,+}\\ &=-\frac{8}{9}\theta\wedge\ast_g\theta+\kappa\wedge\ast_g\kappa_0^{J,+}\\ &=\left(-\frac{8}{9}\lVert\theta\rVert^2+\lVert\kappa_0^{J,+}\rVert^2\right)\operatorname{vol}_g. \end{aligned}

But d(κ∧ψ+)=−23d∗gθ=23δgθvol⁡gd(\kappa\wedge\psi_+)=-\frac{2}{3}d\ast_g\theta=\frac{2}{3}\delta^g\theta\operatorname{vol}_g, so we get (21).

When JJ is integrable, dΨ−−43θ∧Ψd\Psi_- - \frac{4}{3}\theta\wedge\Psi has type (3,1)(3,1). By (18) it is also the real (2,2)(2,2) form −κ0J,+∧F-\kappa_0^{J,+}\wedge F, so it vanishes and κ0J,+=0\kappa_0^{J,+}=0. Equation (21) then gives θ=0\theta=0 on a pluricanonical base. For a compact integrable base, its integral gives the same conclusion. Finally (18) gives κ=0\kappa=0.

Theorem 5. Suppose a pluricanonical LCAK 6-dimensional manifold (B,g,J,F,θ)(B,g,J,F,\theta) admits an SU(3)SU(3) completion and a closed curvature solving (15)–(17). Then

Dgθ=0,ιTκ=0.D^g\theta=0,\qquad\iota_T\kappa=0.

Proof. Locally write κ=dA\kappa=dA and take the lift η=ds+A\eta=ds+A on a product with an interval. The metric and positive form are given by (14). Write Θ=π∗θ\Theta=\pi^*\theta and T=Θ♯T=\Theta^\sharp. The unit fibres are Killing and geodesic, so δgPΘ=π∗(δgθ)=0\delta^{g_P}\Theta=\pi^*(\delta^g\theta)=0. Also dΘ=0d\Theta=0 and ∥Θ∥=r\lVert\Theta\rVert=r is constant by pluricanonicity.

For dh=Θdh=\Theta, the form e−hφe^{-h}\varphi is torsion-free. Its metric g~=e−2h/3gP\widetilde g=e^{-2h/3}g_P is Ricci-flat [11 Corollaries 2.28 and 4.12]. The conformal Ricci formula in dimension seven therefore gives

Ric⁡gP=−53DPΘ−59Θ⊗Θ+13(δgPΘ)gP+59r2gP.\operatorname{Ric}_{g_P} =-\frac{5}{3}D^P\Theta-\frac{5}{9}\Theta\otimes\Theta +\frac{1}{3}\left(\delta^{g_P}\Theta\right)g_P +\frac{5}{9}r^2g_P.

Since Θ\Theta is of constant length, we have (DPΘ)(T,T)=0(D^{P}\Theta)(T,T)=0, and consequently Ric⁡gP(T,T)=0\operatorname{Ric}_{g_{P}}(T,T)=0. The pointwise Bochner formula for the harmonic, constant-length form Θ\Theta now yields ∥DPΘ∥2=0\lVert D^{P}\Theta\rVert^{2}=0.

For horizontal lifts YY, ZZ and the unit vertical vector VV, the connection metric formulas are

(DPΘ)(Y,Z)=(DBθ)(Y,Z),(DPΘ)(Y,V)=12κ(T,Y).(D^{P}\Theta)(Y,Z)=(D^{B}\theta)(Y,Z), \qquad(D^{P}\Theta)(Y,V)=\frac{1}{2}\kappa(T,Y).

Both conclusions follow.

We give now an explicit compact solution of the lift system. The quaternionic model and its circle symmetries are described in [12 Section 2]. On the other hand, the circle reduction of the cone over the homogeneous nearly Kähler space is studied in [13 Section 3]. On S3×S3=SU(2)×SU(2)S^{3}\times S^{3}=SU(2)\times SU(2), we write a=p−1dpa=p^{-1}dp and b=q−1dqb=q^{-1}dq for the imaginary quaternion-valued Maurer–Cartan forms. Thus dai=−2aj∧akda_{i}=-2a_{j}\wedge a_{k} and dbi=−2bj∧bkdb_{i}=-2b_{j}\wedge b_{k} for cyclic (i,j,k)(i,j,k). The canonical nearly Kähler structure is

gN=∑i=13(ai⊗ai−12(ai⊗bi+bi⊗ai)+bi⊗bi),ωN=32∑i=13ai∧bi,Ω=⋀i=13(ai+e2π−13bi),g_{N}=\sum_{i=1}^{3}\left(a_{i}\otimes a_{i}-\frac{1}{2}(a_{i}\otimes b_{i}+b_{i}\otimes a_{i})+b_{i}\otimes b_{i}\right), \qquad\omega_{N}=\frac{\sqrt{3}}{2}\sum_{i=1}^{3}a_{i}\wedge b_{i}, \qquad\Omega=\bigwedge_{i=1}^{3}\left(a_{i}+e^{\frac{2\pi\sqrt{-1}}{3}}b_{i}\right),
ΨN,+=−Im⁡Ω,ΨN,−=Re⁡Ω.\Psi_{N,+}=-\operatorname{Im}\Omega, \qquad\Psi_{N,-}=\operatorname{Re}\Omega.

If we differentiate the above equations, we get (12) with ∥ϑ∥=2\lVert\vartheta\rVert=2.

We will take a circle quotient of

P=S3×S3×(R/ℓZ),ℓ>0.P=S^{3}\times S^{3}\times(\mathbb{R}/\ell\mathbb{Z}), \qquad\ell>0.

Fix a unit imaginary quaternion II. The quotient map π:P→B\pi:P\to B is

π(p,q,t)=(p−1Ip,q,t)∈B=S2×S3×S1.\pi(p,q,t)=(p^{-1}Ip,q,t)\in B=S^{2}\times S^{3}\times S^{1}.

On PP, we consider the G2\mathrm{G}_{2}-structure φ=dt∧ωN+ΨN,+\varphi=dt\wedge\omega_{N}+\Psi_{N,+} inducing the metric gφ=dt2+gNg_{\varphi}=dt^{2}+g_{N}. Thus, the Lee form ϑ=−2dt\vartheta=-2dt, but the circle action leaves tt unchanged so

θ0=−2dt.\theta_{0}=-2dt.

Moreover, we can find the unit vertical vector and its dual 1-form. The circle action is (p,q,t)⟼(eIsp,q,t)(p,q,t)\longmapsto(e^{Is}p,q,t), s∈R/(2πZ)s\in\mathbb{R}/(2\pi\mathbb{Z}). It preserves the nearly Kähler structure and has a unit-length generator V=(Ip,0,0)V=(Ip,0,0). Hence,

a(V)=p−1Ip=x∈S2,b(V)=0,dt(V)=0.a(V)=p^{-1}Ip=x\in S^{2}, \qquad b(V)=0, \qquad dt(V)=0.

Remark that xx is unchanged under the circle action. Moreover, writing the nearly Kähler metric as gN=∣a−12b∣2+34∣b∣2g_{N}=\lvert a-\frac{1}{2}b\rvert^{2}+\frac{3}{4}\lvert b\rvert^{2}, (here ∣a∣2=∑i=13ai⊗ai\lvert a\rvert^{2}=\sum_{i=1}^{3}a_{i}\otimes a_{i} etc) we obtain that gN(V,V)=1g_{N}(V,V)=1. Furthermore, for any tangent vector YY, we get gN(V,Y)=x⋅(a(Y)−12b(Y))g_{N}(V,Y)=x\cdot(a(Y)-\frac{1}{2}b(Y)) (here ⋅\cdot is the Euclidean product). Define

A0=x⋅a,ξ=x⋅b.A_{0}=x\cdot a, \qquad\xi=x\cdot b.

Then,

η=V♭=x⋅a−12x⋅b=A0−12ξ.\eta=V^{\flat}=x\cdot a-\frac{1}{2}x\cdot b=A_{0}-\frac{1}{2}\xi.

Differentiating x=p−1Ipx=p^{-1}Ip and using that dp=padp=pa, we get that

a=xA0−12x×dx.a=xA_{0}-\frac{1}{2}x\times dx.

Then, we can write

a−12b=xη+(−12x×dx−12(b−xξ)).a-\frac{1}{2}b=x\eta+\left(-\frac{1}{2}x\times dx-\frac{1}{2}(b-x\xi)\right).

We remark that the vector-valued 1-form −12x×dx−12(b−xξ)-\frac{1}{2}x\times dx-\frac{1}{2}(b-x\xi) is perpendicular to xx. Thus,

∣a−12b∣2=∣η∣2+∣−12x×dx−12(b−xξ)∣2.\left\lvert a-\frac{1}{2}b\right\rvert^{2}=\lvert\eta\rvert^{2}+\left\lvert-\frac{1}{2}x\times dx-\frac{1}{2}(b-x\xi)\right\rvert^{2}.

In particular,

gφ=∣η∣2+[dt2+∣−12x×dx−12(b−xξ)∣2+34∣b∣2].g_{\varphi}=|\eta|^2+\left[dt^2+\left|-\frac{1}{2}x\times dx-\frac{1}{2}(b-x\xi)\right|^2+\frac{3}{4}|b|^2\right].

The metric on BB is obtained then by removing the vertical term η\eta so

gB,0=dt2+∣12x×dx−12(b−xξ)∣2+34∣b∣2.g_{B,0}=dt^2+\left|\frac{1}{2}x\times dx-\frac{1}{2}(b-x\xi)\right|^2+\frac{3}{4}|b|^2.

Define

α=ιVωN,β=ιVΨN,+,γ=ιVΨN,−,τ=ωN−η∧α.\alpha=\iota_V\omega_N,\qquad\beta=\iota_V\Psi_{N,+},\qquad\gamma=\iota_V\Psi_{N,-},\qquad\tau=\omega_N-\eta\wedge\alpha.

Then, we get that

α=32ξ,β=−34dξ,τ=−34C,γ=14S−14E−12U.\alpha=\frac{\sqrt{3}}{2}\xi,\qquad\beta=-\frac{\sqrt{3}}{4}d\xi,\qquad\tau=-\frac{\sqrt{3}}{4}C,\qquad\gamma=\frac{1}{4}S-\frac{1}{4}E-\frac{1}{2}U.

where

S=12x⋅(dx×dx),E=∑i=13dxi∧bi,U=∑cyc3xibj∧bk,C=∑i=13(x×dx)i∧bi.S=\frac{1}{2}x\cdot(dx\times dx),\qquad E=\sum_{i=1}^{3}dx_i\wedge b_i,\qquad U=\sum_{\mathrm{cyc}}^{3}x_i b_j\wedge b_k,\qquad C=\sum_{i=1}^{3}(x\times dx)_i\wedge b_i.

Here the cross product of vector-valued 1-forms uses exterior multiplication. The 2-form SS is the unit-sphere area form with ∫S2S=4π\int_{S^2}S=4\pi. The induced SU(3)SU(3) completion is

F0=−dt∧α+β,ψ+,0=dt∧τ−α∧γ,ψ−,0=−α∧τ−dt∧γ.F_0=-dt\wedge\alpha+\beta,\qquad\psi_{+,0}=dt\wedge\tau-\alpha\wedge\gamma,\qquad\psi_{-,0}=-\alpha\wedge\tau-dt\wedge\gamma.

Moreover,

κ0=dη=12S−12dξ.\kappa_0=d\eta=\frac{1}{2}S-\frac{1}{2}d\xi.

These forms verify (15)–(17) directly. Theorem 2 gives pluricanonicity and non-integrability. Moreover, substitution in the homogeneous nearly Kähler structure gives

ωN=η∧α+τ,\omega_N=\eta\wedge\alpha+\tau,
ΨN,+=η∧β−α∧γ,\Psi_{N,+}=\eta\wedge\beta-\alpha\wedge\gamma,
ΨN,−=η∧γ+α∧β.\Psi_{N,-}=\eta\wedge\gamma+\alpha\wedge\beta.

Finally, we remark that [κ02π]=[S4π]∈H2(B;Z)\left[\frac{\kappa_0}{2\pi}\right]=\left[\frac{S}{4\pi}\right]\in H^2(B;\mathbb{Z}), because ∫S2S=4π\int_{S^2}S=4\pi.

Pluricanonical metrics and He's flows

In [7], He considers the evolution equations

∂F∂t=−Δg(t)F,∂F∂t=−δg(t)dF.(22)\frac{\partial F}{\partial t}=-\Delta^{g(t)}F,\qquad\frac{\partial F}{\partial t}=-\delta^{g(t)}dF. \tag*{(22)}

For a fixed almost complex structure JJ, the evolving metric is determined by the JJ-invariant part of FF. Our solutions remain compatible with JJ, so that g(t)=F(t)(⋅,J⋅)g(t)=F(t)(\cdot,J\cdot) throughout. We point out that these are different equations: the eigenform identity for Δg\Delta^g does not imply the same identity for δgd\delta^g d.

We recall the following identities for a pluricanonical LCAK structure proved in [1]

dδgF=(n−1)(∥θ∥2F−θ∧Jθ),δgdF=(n−1)θ∧Jθ.(23)d\delta^gF=(n-1)(\|\theta\|^2F-\theta\wedge J\theta),\qquad\delta^g dF=(n-1)\theta\wedge J\theta. \tag*{(23)}

A consequence of the pluricanonical LCAK identities is the following explicit solution of He’s full Hodge equation.

Theorem 6. Let (M,g0,J,F0,θ)(M,g_0,J,F_0,\theta) be a pluricanonical LCAK manifold of real dimension 2n2n, n≥2n \ge2. Put λ=(n−1)∥θ∥g02\lambda=(n-1)\lVert\theta\rVert_{g_0}^2. Then

F(t)=(1−λt)F0,g(t)=(1−λt)g0(24)F(t)=(1-\lambda t)F_0,\qquad g(t)=(1-\lambda t)g_0 \tag*{(24)}

is a solution of the full Hodge equation in (22) for 1−λt>01-\lambda t>0. The almost complex structure and the Lee form remain fixed, and the evolving structure is pluricanonical. If λ>0\lambda>0, this family degenerates at t=λ−1t=\lambda^{-1}; if λ=0\lambda=0, it is stationary.

Proof. For a spatially constant c(t)>0c(t)>0 (it depends only on tt), the Hodge Laplacian satisfies Δcg0=c−1Δg0\Delta^{cg_0}=c^{-1}\Delta^{g_0} on differential forms. Consequently,

Δcg0(cF0)=Δg0F0=λF0.\Delta^{cg_0}(cF_0)=\Delta^{g_0}F_0=\lambda F_0.

The equation for F(t)=c(t)F0F(t)=c(t)F_0 is therefore c′=−λc'=-\lambda. Moreover, dF(t)=θ∧F(t)dF(t)=\theta\wedge F(t), and constant metric homotheties leave the Levi-Civita connection unchanged. Thus (Dθ)J,+=0(D\theta)^{J,+}=0 remains true. The image of the Nijenhuis tensor remains gg-orthogonal to TT and JTJT, since multiplying the metric by a positive constant preserves orthogonality.

The d∗dd^{*}d equation

We suppose that the manifold is connected and that θ\theta is nowhere zero. Let V=Span⁡{T,JT}V=\operatorname{Span}\{T,JT\}, where T=θ♯g0T=\theta^{\sharp_{g_0}}, and let H=V⊥g0H=V^{\perp_{g_0}}. We write

FV=∥θ∥−2θ∧Jθ,FH=F0−FV,F_V=\lVert\theta\rVert^{-2}\theta\wedge J\theta,\qquad F_H=F_0-F_V,
gV=∥θ∥−2(θ⊗θ+Jθ⊗Jθ),gH=g0−gV.(25)g_V=\lVert\theta\rVert^{-2}(\theta\otimes\theta+J\theta\otimes J\theta),\qquad g_H=g_0-g_V. \tag*{(25)}

The tensors gHg_H and gVg_V are the restrictions of g0g_0 to the orthogonal distributions, extended by zero. The 22-forms FHF_H and FVF_V are their fundamental forms. Since ∥θ∥\lVert\theta\rVert is constant,

FH=−∥θ∥−2dJθ,dFH=0,dFV=θ∧FH.(26)F_H=-\lVert\theta\rVert^{-2}dJ\theta,\qquad dF_H=0,\qquad dF_V=\theta\wedge F_H. \tag*{(26)}

Proposition 7. For any spatially constant a(t),b(t)>0a(t),b(t)>0, set

Fa,b=aFH+bFV,ga,b=agH+bgV,θa,b=baθ.(27)F_{a,b}=aF_H+bF_V,\qquad g_{a,b}=ag_H+bg_V,\qquad\theta_{a,b}=\frac{b}{a}\theta. \tag*{(27)}

Then (ga,b,J,Fa,b,θa,b)(g_{a,b},J,F_{a,b},\theta_{a,b}) is pluricanonical LCAK. It satisfies

∥θa,b∥ga,b2=ba2∥θ∥2,δga,bdFa,b=λb2a2FV,λ=(n−1)∥θ∥2.(28)\lVert\theta_{a,b}\rVert_{g_{a,b}}^2=\frac{b}{a^2}\lVert\theta\rVert^2,\qquad\delta^{g_{a,b}}dF_{a,b}=\lambda\frac{b^2}{a^2}F_V,\qquad\lambda=(n-1)\lVert\theta\rVert^2. \tag*{(28)}

Proof. Both blocks of ga,bg_{a,b} are JJ-invariant, and a,b>0a,b>0 give a positive definite metric. From (26),

dFa,b=bθ∧FH=θa,b∧Fa,b,dθa,b=0.dF_{a,b}=b\theta\wedge F_H=\theta_{a,b}\wedge F_{a,b},\qquad d\theta_{a,b}=0.

The dual of θa,b\theta_{a,b} for ga,bg_{a,b} is a−1Ta^{-1}T, which proves the Lee norm formula and shows that its Lee plane is still VV. Since JJ is fixed, its Nijenhuis tensor is unchanged. Its image lies in HH for the initial pluricanonical structure, and thus remains ga,bg_{a,b}-orthogonal to VV.

The remaining pluricanonical condition follows from

dJθa,b=−b∥θ∥2aFH=−∥θa,b∥ga,b2Fa,b+θa,b∧Jθa,b.dJ\theta_{a,b}=-\frac{b\lVert\theta\rVert^2}{a}F_H=-\lVert\theta_{a,b}\rVert_{g_{a,b}}^2F_{a,b}+\theta_{a,b}\wedge J\theta_{a,b}.

Indeed, the general LCAK identity

dJϑ=−∥ϑ∥2F+ϑ∧Jϑ+2(Dϑ)⋅,⋅J,++2NJϑ♯,dJ\vartheta=-\lVert\vartheta\rVert^2F+\vartheta\wedge J\vartheta+2(D\vartheta)^{J,+}_{\cdot,\cdot}+2N_{J\vartheta^\sharp},

where NX(Y,Z)=g(N(Y,Z),X)N_X(Y,Z)=g(N(Y,Z),X), from [1 Proposition 2], applied to the new metric, gives (Dga,bθa,b)J,+=0(D^{g_{a,b}}\theta_{a,b})^{J,+}=0. Finally, (23) for this pluricanonical structure gives

δga,bdFa,b=(n−1)θa,b∧Jθa,b=λb2a2FV.\delta^{g_{a,b}}dF_{a,b}=(n-1)\theta_{a,b}\wedge J\theta_{a,b}=\lambda\frac{b^2}{a^2}F_V.

Theorem 8. With the notation above, put b(t)=(1+λt)−1b(t)=(1+\lambda t)^{-1}. Then

F(t)=FH+b(t)FV,g(t)=gH+b(t)gV,θ(t)=b(t)θ,J(t)=J(29)F(t)=F_H+b(t)F_V,\qquad g(t)=g_H+b(t)g_V,\qquad\theta(t)=b(t)\theta,\qquad J(t)=J \tag*{(29)}

is a pluricanonical solution of

∂tF=−δg(t)dF.\partial_tF=-\delta^{g(t)}dF.

on (−λ−1,∞)(-\lambda^{-1},\infty). In particular, the solution is immortal in the forward direction. On a compact manifold, it is the unique smooth solution with this initial form in He’s fixed-JJ tamed class, on every common forward interval of existence. It also solves He’s d∗dd^*d equation for compatible pairs with the compatibility correction determined by the JJ-anti-invariant part of −δg(t)dF(t)-\delta^{g(t)}dF(t).

Proof. Proposition 7 proves pluricanonicity at every positive value of bb. The flow equation implies that a˙(t)=0\dot{a}(t)=0, with a(0)=1a(0)=1. Then, for a=1a=1, the evolution equation reduces to b˙=−λb2\dot{b}=-\lambda b^2, b(0)=1b(0)=1. The velocity F˙(t)\dot{F}(t) is JJ-invariant, so the indicated compatibility correction vanishes. On compact manifolds, uniqueness follows from [7 Theorem 2.1], applied successively from any time at which two smooth solutions agree.

References

[1] E. Addison, T. Drăghici, and M. Lejmi, “On pluricanonical locally conformally almost Kähler metrics.” arXiv:2602.12352v2, 2026. To appear in Tohoku Mathematical Journal. (Cited on pages 1, 3, 9, and 10.)

[2] G. Kokarev, “On pseudo-harmonic maps in conformal geometry,” Proc. Lond. Math. Soc. (3), vol. 99, no. 1, pp. 168–194, 2009. (Cited on page 1.)

[3] A. Moroianu and S. Moroianu, “On pluricanonical locally conformally Kähler manifolds,” Int. Math. Res. Not., vol. 2017, no. 14, pp. 4398–4405, 2017. (Cited on page 1.)

[4] S. Ivanov, M. Parton, and P. Piccinni, “Locally conformal parallel G2 and Spin(7) manifolds,” Math. Res. Lett., vol. 13, no. 2, pp. 167–177, 2006. (Cited on pages 2, 4, and 5.)

[5] S. Chiossi and S. Salamon, “The intrinsic torsion of SU(3) and G2 structures,” in Differential geometry, Valencia, 2001, pp. 115–133, World Sci. Publ., River Edge, NJ, 2002. (Cited on page 2.)

[6] I. Agricola and T. Friedrich, “Geometric structures of vectorial type,” J. Geom. Phys., vol. 56, no. 12, pp. 2403–2414, 2006. (Cited on page 2.)

[7] W. He, “Evolution equations for non-degenerate 2-forms,” Int. Math. Res. Not., vol. 2021, no. 6, pp. 4349–4391, 2021. (Cited on pages 2, 9, and 11.)

[8] V. Apostolov, M. Bailey, and G. Dloussky, “From locally conformally Kähler to bi-Hermitian structures on non-Kähler complex surfaces,” Math. Res. Lett., vol. 22, no. 2, pp. 317–336, 2015. (Cited on page 3.)

[9] S. Donaldson, “Adiabatic limits of co-associative Kovalev-Lefschetz fibrations,” in Algebra, geometry, and physics in the 21st century, vol. 324 of Progr. Math., pp. 1–29, Birkhäuser/Springer, Cham, 2017. (Cited on page 3.)

[10] A. Moroianu, P.-A. Nagy, and U. Semmelmann, “Unit Killing vector fields on nearly Kähler manifolds,” International Journal of Mathematics, vol. 16, no. 3, pp. 281–301, 2005. (Cited on page 6.)

[11] S. Karigiannis, “Flows of G2-structures, I,” The Quarterly Journal of Mathematics, vol. 60, pp. 487–522, 2009. (Cited on page 7.)

[12] K. Dixon, “The multi-moment maps of the nearly Kähler S3×S3S^3\times S^3,” Geometriae Dedicata, vol. 200, pp. 351–362, 2019. (Cited on page 8.)

[13] H. Sá Earp and J. R. Stein, “Isolated singularities in G2-structures with torsion,” Geometriae Dedicata, vol. 220, p. 33, 2026. (Cited on page 8.)

Department of Mathematics, Bronx Community College of CUNY, Bronx, NY 10453, USA.

Email address: [email protected]

References

  1. [1]E. Addison, T. Drăghici, and M. Lejmi, “On pluricanonical locally conformally almost Kähler metrics,” arXiv:2602.12352v2, 2026. To appear in Tohoku Mathematical Journal. (Cited on pages 1, 3, 9, and 10.)arxiv.org/abs/2602.12352
  2. [2]G. Kokarev, “On pseudo-harmonic maps in conformal geometry,” Proc. Lond. Math. Soc. (3), vol. 99, no. 1, pp. 168–194, 2009. (Cited on page 1.)DOI
  3. [3]A. Moroianu and S. Moroianu, “On pluricanonical locally conformally Kähler manifolds,” Int. Math. Res. Not., vol. 2017, no. 14, pp. 4398–4405, 2017. (Cited on page 1.)DOI
  4. [4]S. Ivanov, M. Parton, and P. Piccinni, “Locally conformal parallel G₂ and Spin(7) manifolds,” Math. Res. Lett., vol. 13, no. 2, pp. 167–177, 2006. (Cited on pages 2, 4, and 5.)
  5. [5]S. Chiossi and S. Salamon, “The intrinsic torsion of SU(3) and G₂ structures,” in Differential geometry, Valencia, 2001, pp. 115–133, World Sci. Publ., River Edge, NJ, 2002. (Cited on page 2.)DOI
  6. [6]I. Agricola and T. Friedrich, “Geometric structures of vectorial type,” J. Geom. Phys., vol. 56, no. 12, pp. 2403–2414, 2006. (Cited on page 2.)DOI
  7. [7]W. He, “Evolution equations for non-degenerate 2-forms,” Int. Math. Res. Not., vol. 2021, no. 6, pp. 4349–4391, 2021. (Cited on pages 2, 9, and 11.)
  8. [8]V. Apostolov, M. Bailey, and G. Dloussky, “From locally conformally Kähler to bi-Hermitian structures on non-Kähler complex surfaces,” Math. Res. Lett., vol. 22, no. 2, pp. 317–336, 2015. (Cited on page 3.)arxiv.org/abs/1307.3660
  9. [9]S. Donaldson, “Adiabatic limits of co-associative Kovalev-Lefschetz fibrations,” in Algebra, geometry, and physics in the 21st century, vol. 324 of Progr. Math., pp. 1–29, Birkhäuser/Springer, Cham, 2017. (Cited on page 3.)arxiv.org/abs/1603.08391
  10. [10]A. Moroianu, P.-A. Nagy, and U. Semmelmann, “Unit Killing vector fields on nearly Kähler manifolds,” International Journal of Mathematics, vol. 16, no. 3, pp. 281–301, 2005. (Cited on page 6.)DOI
  11. [11]S. Karigiannis, “Flows of G₂-structures, I,” The Quarterly Journal of Mathematics, vol. 60, pp. 487–522, 2009. (Cited on page 7.)DOI
  12. [12]K. Dixon, “The multi-moment maps of the nearly Kähler S³ × S³,” Geometriae Dedicata, vol. 200, pp. 351–362, 2019. (Cited on page 8.)DOI
  13. [13]H. Sá Earp and J. R. Stein, “Isolated singularities in G₂-structures with torsion,” Geometriae Dedicata, vol. 220, p. 33, 2026. (Cited on page 8.)

Paper details

vlnc:W_6sgfxkzj
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 3

downloads · 0

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

  • Mathematics

ChatGPT 6.1 Sol (OpenAI)

View statement
Looking for related papers…
Contents