1 Introduction

It is known that the Ricci flow can be used to give an alternative proof of the uniformization theorem of Riemann surfaces. Likewise, it can be questioned whether a geometric flow is able to be applied for classifying non-Kähler complex surfaces. These flows should preserve Hermitianness and some additional structures such as pluriclosedness, and should be close to the Ricci flow as much as possible. From this point of view, J. Streets and G. Tian introduced a parabolic evolution equation of pluriclosed metrics with a pluriclosed initial metric \(\omega _0\) on a compact Hermitian manifold,

$$\begin{aligned} \left\{ \begin{array}{l}\dfrac{\partial }{\partial t}\omega (t)=\partial \partial _{g(t)}^*\omega (t)+{\bar{\partial }}{\bar{\partial }}_{g(t)}^*\omega (t) -P(g(t)),\\ \\ \omega (0)=\omega _0,\end{array}\right. \end{aligned}$$

which is called the pluriclosed flow, where \(\partial _{g(t)}^*\) and \({\bar{\partial }}_{g(t)}^*\) are decompositions of the \(L^2\)-adjoint operator of the exterior differential operator and P(g(t)) is the Ricci-type curvature of the Chern connection with respect to metrics g(t) (cf. [9]). Their studies motivated us to generalize their results to almost Hermitian geometry. In [4], the author defined two parabolic flows; the almost Hermitian flow (AHF) and the almost Hermitian curvature flow (AHCF) on almost complex manifolds, which coincide with the pluriclosed flow and the Hermitian curvature flow respectively on complex manifolds, and studied the relationship between these parabolic evolution equations on a compact almost Hermitian manifold. In [5], we derived higher order derivative etimates in the presence of a curvature bound. And moreover, we exibited a long-time existence obstruction for solutions to the almost Hermitian curvature flow by showing smoothing estimates for the curvature and torsion.

In the present paper, we investigate the AHCF, which coincides with the AHF (cf. [4, Theorem 1.2]), on a compact almost Hermitian manifold and we show that if the initial metric has the non-positive first Chern–Ricci curvature, then this positivity can be preserved along the AHCF. We mimic the argument in the proof of the positivity preservation properties of the Hermitian curvature flow (cf. [7]). Such an argument was initiated by G. Liu in order to prove that the Kähler–Ricci flow preserves the non-positivity of Ricci curvature if the initial metric has non-positive bisectional curvature (cf. [8]).

Let (MJ) be a compact almost complex manifold and let g be an almost Hermitian metric on M. Let \(\{Z_r\}\) be an arbitrary local (1, 0)-frame around a fixed point \(p\in M\) and let \(\{\zeta ^r\}\) be the associated coframe. Then the associated real (1, 1)-form \(\omega \) with respect to g takes the local expression \(\omega =\sqrt{-1}g_{r{\bar{k}}}\zeta ^r\wedge \zeta ^{{\bar{k}}}\). We will also refer to \(\omega \) as to an almost Hermitian metric. Let \(\Omega \) be the curvature of the Chern connection and \(\Omega \) splits in \(\Omega =H+R+{\bar{H}}\), where \(R\in \Gamma (\Lambda ^{1,1}M\otimes \Lambda ^{1,1}M)\), \(H\in \Gamma (\Lambda ^{2,0}M\otimes \Lambda ^{1,1}M)\) (see Sect. 2 in detail). The almost Hermitian flow (AHF) with an almost Hermitian initial metric \(\omega _0\) on (MJ) is as follows:

$$\begin{aligned} \left\{ \begin{array}{l}\dfrac{\partial }{\partial t}\omega (t)=\partial \partial _{g(t)}^*\omega (t)+{\bar{\partial }}{\bar{\partial }}_{g(t)}^*\omega (t)-P(\omega (t)),\\ \\ \omega (0)=\omega _0,\end{array}\right. \end{aligned}$$

where \(\partial _{g(t)}^*\) and \({\bar{\partial }}_{g(t)}^*\) are the \(L^2\)-adjoint operators with respect to metrics g(t), and \(P(\omega )\) is one of the Ricci-type curvatures of the Chern curvature, which is callaed the first Chern–Ricci curvature and locally given by \( P_{i{\bar{j}}}=g^{k{\bar{l}}}R_{i{\bar{j}}k{\bar{l}}}\).

We have proven the following short-time existence result for the AHF.

Proposition 1.1

(cf. [4, Theorem 1.1]) Given a compact almost Hermitian manifold \((M,\omega _0,J)\), there exists a unique solution to the AHF with initial condition \(\omega _0\) on \([0,\varepsilon )\) for some \(\varepsilon >0\).

We denote by S one of the Ricci-type curvatures of the Chern curvature, which is called the second Chern–Ricci curvature and is locally given by \( S_{i{\bar{j}}}=g^{k{\bar{l}}}R_{k{\bar{l}}i{\bar{j}}}\).

It has been proved that a solution of the almost Hermitian flow with initial condition \(g_0\) is equivalent to a solution of the following parabolic flow on a compact almost complex manifold with an almost Hermitian metric, we call it the almost Hermitian curvature flow (AHCF):

$$\begin{aligned} \left\{ \begin{array}{l}\dfrac{\partial }{\partial t}g(t)=-S(g(t))-Q^7(g(t))-Q^8(g(t))+BT'(g(t))+{\bar{Z}}(T')(g(t)), \\ \\ g(0)=g_0,\end{array}\right. \end{aligned}$$

where \(Q^1, Q^7, Q^8\) are quadratics in the torsion of the Chern connection (cf. [11, pg. 712])

$$\begin{aligned}&Q^1_{i{\bar{j}}}:=T_{ik{\bar{r}}}T_{{\bar{j}}{\bar{k}}r},\quad Q^7_{i{\bar{j}}}:=T_{irk}T_{{\bar{r}}{\bar{k}}{\bar{j}}},\quad Q^8_{i{\bar{j}}}:=T_{irk}T_{{\bar{j}}{\bar{k}}{\bar{r}}},\quad w_i:=T_{ir{\bar{r}}}, \\&BT'_{i{\bar{j}}}:=B_{{\bar{r}}p}^jT_{ir{\bar{p}}}+B_{{\bar{p}}i}^rT_{pr{\bar{j}}} +B_{{\bar{r}}r}^pT_{pi{\bar{j}}}+B_{{\bar{j}}i}^rw_r, \end{aligned}$$

and

$$\begin{aligned} {\bar{Z}}(T')_{i{\bar{j}}}:=-Z_{{\bar{r}}}(T_{ri}^s)g_{s{\bar{j}}}-Z_{{\bar{j}}}(w_i)-g^{p{\bar{q}}}T_{pi}^rZ_{{\bar{j}}}(g_{r{\bar{q}}}). \end{aligned}$$

These components are defined using an arbitrary unitary frame. In all this paper, we assume the Einstein convention omitting the symbol of sum over repeated indexes.

Note that \(P=S+\text {div}^{\nabla }T'-{\nabla }{\bar{w}}+Q^7+Q^8\) for any almost Hermitian metric g (cf. [11, Lemma 3.5]), where \(T'\) is the torsion of the Chern connection \(\nabla \) associated to g, \((\text {div}^{\nabla }T')_{i{\bar{j}}}=g^{k{\bar{l}}}\nabla _{{\bar{l}}}T_{ki{\bar{j}}},\) \((\nabla {\bar{w}})_{i{\bar{j}}}=g^{k{\bar{l}}}\nabla _iT_{{\bar{j}}{\bar{l}}k}.\)

We say that a Hermitian metric g is pluriclosed if its associated real (1, 1) form \(\omega \) satisfies \(\partial {\bar{\partial }}\omega =0\). We often write \(\omega \) as a metric and say that \(\omega \) is pluriclosed as well. In this paper, we will call the following parabolic flow on a compact complex manifold the Hermitian curvature flow (\(\text {HCF}_{Q^1}\)):

$$\begin{aligned} \left\{ \begin{array}{l}\dfrac{\partial }{\partial t}g(t)=-S(g(t))+Q^1(g(t)),\\ \\ g(0)=g_0,\end{array}\right. \end{aligned}$$

where \(g_0\) is a pluriclosed metric. We can obtain the following equivalence between the AHCF and the \(\text {HCF}_{Q^1}\) when the almost complex structure J is integrable.

Proposition 1.2

(cf. [4, Proposition 1.1]) The AHCF coincides with the \(\text {HCF}_{Q^1}\) starting at a pluriclosed metric if J is integrable.

It is known that the \(\text {HCF}_{Q^1}\) coincides with the pluriclosed flow starting at the same initial metric, which is called Streets–Tian identifiability theorem (cf. [9, Proposition 3.3]). The following result is the generalized version of Streets–Tian identifiability theorem.

Proposition 1.3

(cf. [4, Theorem 1.2]) Let \((M, g_0,J)\) be a compact almost Hermitian manifold with the associated real (1, 1)-form \(\omega _0\). Then a solution to the AHCF with initial condition \(g_0\) is equivalent to a solution to the AHF starting at the initial condition \(\omega _0\).

As in [10, Theorem 1.1], we have developed some regurarity results for the AHCF. And also, we obtained the long-time existence obstruction for the AHCF in [5]. The long-time existence obstruction will be used for estimating the evolution equation of the Chern curvature R along the AHCF in order to prove our main theorem.

Proposition 1.4

(cf. [5, Theorem 1.1]) Let \((M^{2n},J,g(t))\) be a solution to the AHCF for a maximal time interval \([0,\tau _{\text {max}})\) on a compact almost Hermitian manifold which starts at the initial almost Hermitian metric \(g_0\). The following statements \(\mathrm{(i)}\), \(\mathrm{(ii)}\) and \(\mathrm{(iii)}\) hold.

  1. (i)

    We choose arbitrary \(0<\tau <\tau _{\text {max}}\). Assume that, for a positive constants \(\alpha \) with \(\alpha /\tau >1\), the following inequalities hold:

    $$\begin{aligned} \sup _{M\times [0,\tau )}|R|_{g(t)}\le \frac{\alpha }{\tau },\quad \sup _{M\times [0,\tau )}|T'|^2_{g(t)}\le \frac{\alpha }{\tau },\quad \sup _{M\times [0,\tau )}|\nabla T'|_{g(t)}\le \frac{\alpha }{\tau }. \end{aligned}$$

    Then, for any \(m\in {\mathbb {N}}\), the following inequalities hold:

    $$\begin{aligned} |\nabla ^mR|_{g(t)}\le \frac{C_{m,n,\alpha }}{\tau \cdot t^{\frac{m}{2}}},\quad |\nabla ^{m+1}T'|_{g(t)}\le \frac{C_{m,n,\alpha }}{\tau \cdot t^{\frac{m}{2}}} \end{aligned}$$

    for any \(t\in (0,\tau ]\), where \(C_{m,n,\alpha }\) is some positive constant depending only on m, n and \(\alpha \).

  2. (ii)

    If \(\tau _{\text {max}}<\infty \), then

    $$\begin{aligned} \limsup _{t\rightarrow \tau _{\text {max}}}\max \Big \{\max _M|R|_{g(t)}, \max _M|T'|^2_{g(t)}, \max _M|\nabla T'|_{g(t)}\Big \}=\infty . \end{aligned}$$
  3. (iii)

    If J is integrable and \(g_0\) is pluriclosed, then g(t) is pluriclosed for all time in the existence interval and g(t) is a solution to the \(\text {HCF}_{Q^1}\). If furthermore \(g_0\) is Kähler, then g(t) is Kähler for all time and g(t) solves the Kähler–Ricci flow.

Remark 1.1

Notice the fact that for any \(C^\infty \)-family of \(C^\infty \)-functions \(\{\rho _t\}_{t\in [0,\tau _{\max })}\) such that \(\limsup _{t\rightarrow \tau _{\max }}\max _M|\rho _t|=\infty \) and for any \(0<\tau<\tau _{\max }<\infty \),

$$\begin{aligned} \sup _{t\in [0,\tau )}\max _M|\rho _t|\le \max _{t\in [0,\tau ]}\max _M|\rho _t|<\infty \end{aligned}$$

holds because \(M\times [0,\tau ]\) is compact. We may apply this for \(\max _M|T'|^2_{g(t)}\), \(\max _M|\nabla T'|_{g(t)}\) and \(\max _M|R|_{g(t)}\) since we have the long-time existence obstruction in Proposition 1.4\(\mathrm{(ii)}\). Hence, we may assume that the quantities \(|T'|^2_{C^0(g(t))}\), \(|\nabla T'|_{C^0(g(t))}\) and \(|R|_{C^0(g(t))}\) are uniformly bounded on \([0,\tau )\), where \(|\cdot |_{C^0(g(t))}=\max _M|\cdot |_{g(t)}\).

Note that we have shown that the uniform equivalence between almost Hermitian metrics and the solution to the AHF (equivalently, the AHCF) in [6]. By applying the strong maximum principle (see Sect. 5), we have the following main result.

Theorem 1.1

Suppose \((M,J,g_0)\) is a compact almost Hermitian manifold with the Griffiths non-positive Chern curvature (see Definition 2.1). Let g(t), \(t\in [0,\tau _{\max })\) be a solution to the AHCF starting from the initial metric \(g(0)=g_0\), where \(\tau _{\max }\) is the finite explosion time of the AHCF. Then there exists \(0<\tau <\tau _{\max }\) such that the first Chern–Ricci curvature P(g(t)) is non-positive on \([0,\tau ]\). Moreover, if the metric \(g_0\) has the first Chern–Ricci curvature which is negative at some point, then there exists \(0<\tau <\tau _{\max }\) such that for any \(t\in (0,\tau ]\), \(P(g(t))<0\) everywhere on M.

Note that if the initial metric \(g_0\) has the Griffiths non-positive Chern curvature and the first Chern–Ricci curvature which is negative at some point, then \(g_0\) has the quasi-negative first Chern–Ricci curvature (see Definition 2.2).

A quasi-Kähler structure is an almost Hermitian structure whose real (1, 1)-form \(\omega \) satisfies \((d\omega )^{(1,2)}={\bar{\partial }}\omega =0\), which is equivalent to the original definition of quasi-Kählerianity: \(D_XJ(Y)+D_{JX}J(JY)=0\) for all vector fields XY, where D is the Levi-Civita connection with respect to the metric \(\omega \). By letting \({\mathcal {K}}\), \({{\mathcal {Q}}}{{\mathcal {K}}}\), and \({\mathcal {H}}\) denote the class of Kähler manifolds, the class of quasi-Kähler manifolds, and the class of Hermitian manifolds respectively, we have that \({\mathcal {K}}={\mathcal {H}}\cap {{\mathcal {Q}}}{{\mathcal {K}}}\) (cf. [2]).

Notice that in [7], M.-C. Lee has proven that the canonical line bundle of a compact Hermitian manifold with nonpositive curvature in the sense of Griffiths and quasi-negative Ricci curvature must be ample. The following result can be easily given by applying this Lee’s result. We introduce another proof by applying Theorem 1.1. The condition “non-quasi-Kähler” means that the almost complex structure J admits no quasi-Kähler metric.

Corollary 1.1

Suppose \((M,J,g_0)\) is a compact non-quasi-Kähler almost Hermitian manifold with Griffiths non-positive Chern curvature. Moreover, if the metric \(g_0\) has the first Chern–Ricci curvature which is negative at some point, then J cannot be integrable.

Proof

By the short-time existence result in [4], there exists a shot-time solution g(t) to the AHCF starting from the metric \(g_0\). By Theorem 1.1, there exists \(\tau >0\) such that \(P(g(\tau ))<0\) on M. Now, let us assume that the almost complex structure J is integrable. Then the manifold becomes Hermitian and the assumption “non-quasi-Kähler” implies that the complex structure J admits no Kähler metric from the fact that \({\mathcal {K}}={\mathcal {H}}\cap {{\mathcal {Q}}}{{\mathcal {K}}}\). Since we obtain that \(c_1(K_M)>0\) from that \(P(g(\tau ))<0\) on M and the assumption that J is integrable, then the manifold M must be Kähler, which is contradictory to that J does not admit any Kähler metrics. Therefore, the almost complex structure J cannot be integrable under these assumptions. \(\square \)

Note that if \(M\subset {\mathbb {R}}^7\), M is almost-Kähler if and only if M is Kähler, i.e., M is non-almost-Kähler if and only if M is non-Kähler (cf. [3]). This tells us that we may change the condition “non-quasi-Kähler” to “non-almost-Kähler”, which means that the almost complex structure J admits no almost-Kähler metric, for a compact almost Hermitian manifold \(M\subset {\mathbb {R}}^7\) in Corollary 1.1.

This paper is organized as follows: in Sect. 2, we recall some basic definitions and computations. In Sect. 3, we show some estimates for torsions and the term \({\bar{Z}}(T')\) and the curvature H. And also we compute the evolution equation of the curvature R and the first Chern–Ricci curvature P. In Sect. 4, we show the preservation of non-positivity of the first Chern–Ricci curvature on a compact almost Hermitian manifold with non-positive bisectional curvature. In the last section, we prove if moreover, the initial metric has the first Chern–Ricci curvature which is negative at some point, then there exists \(\tau >0\) such that the first Chern–Ricci curvature is negative on \((0,\tau ]\) by applying the strong maximal principle. Notice that we assume the Einstein convention omitting the symbol of sum over repeated indexes in all this paper.

2 Preliminaries

2.1 The Nijenhuis Tensor of the Almost Complex Structure

Let M be a 2n-dimensional smooth differentiable manifold. An almost complex structure on M is an endomorphism J of TM, \(J\in \Gamma (\text {End}(TM))\), satisfying \(J^2=-Id_{TM}\), where TM is the real tangent vector bundle of M. The pair (MJ) is called an almost complex manifold. Let (MJ) be an almost complex manifold. We define a bilinear map on \(C^\infty (M)\) for \(X,Y\in \Gamma (TM)\) by

$$\begin{aligned} 4N(X,Y):=[JX,JY]-J[JX,Y]-J[X,JY]-[X,Y], \end{aligned}$$

which is the Nijenhuis tensor of J. The Nijenhuis tensor N satisfies \(N(X,Y)=-N(Y,X)\), \(N(JX,Y)=-JN(X,Y)\), \(N(X,JY)=-JN(X,Y)\), \(N(JX,JY)=-N(X,Y)\). For any (1, 0)-vector fields W and V, \(N(V,W)=-[V,W]^{(0,1)},\) \(N(V,{\bar{W}})=N({\bar{V}},W)=0\) and \(N({\bar{V}},{\bar{W}})=-[{\bar{V}},{\bar{W}}]^{(1,0)}\) since we have \(4N(V,W)=-2([V,W]+\sqrt{-1}J[V,W])\), \(4N({\bar{V}},{\bar{W}})=-2([{\bar{V}},{\bar{W}}]-\sqrt{-1}J[{\bar{V}},{\bar{W}}])\). An almost complex structure J is called integrable if \(N=0\) on M. There are several equivalent conditions for integrability: as we just mentioned that the Nijenhuis tensor \(N_{jk}^i\) of J vanishes, which is equivalent to that there exist holomorphic coordinates compatible with J, and also equivalent to that the space of (1, 0)-vector fields related to J is closed under Lie bracket. Giving a complex structure to a differentiable manifold M is equivalent to giving an integrable almost complex structure to M. Let (MJ) be an almost complex manifold. A Riemannian metric g on M is called J-invariant if J is compatible with g, i.e., for any \(X,Y\in \Gamma (TM)\), \(g(X,Y)=g(JX,JY).\) In this case, the pair (gJ) is called an almost Hermitian structure. The fundamental 2-form \(\omega \) associated to a J-invariant Riemannian metric g, i.e., an almost Hermitian metric, is determined by, for \(X,Y\in \Gamma (TM)\), \(\omega (X,Y)=g(JX,Y).\) Indeed we have, for any \(X,Y\in \Gamma (TM)\),

$$\begin{aligned} \omega (Y,X)=g(JY,X)=g(J^2Y,JX)=-g(JX,Y)=-\omega (X,Y) \end{aligned}$$

and \(\omega \in \Gamma (\bigwedge ^2T^*M)\). We will also refer to the associated real fundamental (1, 1)-form \(\omega \) as an almost Hermitian metric. The form \(\omega \) is related to the volume form \(dV_g\) by \(n!dV_g=\omega ^n\).

The complexified tangent vector bundle is given by \(T^{{\mathbb {C}}}M=TM\otimes _{{\mathbb {R}}}{\mathbb {C}}\) for the real tangent vector bundle TM. By extending J \({\mathbb {C}}\)-linearly and g, \(\omega \) \({\mathbb {C}}\)-bilinearly to \(T^{{\mathbb {C}}}M\), they are also defined on \(T^{{\mathbb {C}}}M\) and we observe that the complexified tangent vector bundle \(T^{{\mathbb {C}}}M\) can be decomposed as

$$\begin{aligned} T^{{\mathbb {C}}}M=T^{1,0}M\oplus T^{0,1}M, \end{aligned}$$

where \(T^{1,0}M\), \(T^{0,1}M\) are the eigenspaces of J corresponding to eigenvalues \(\sqrt{-1}\) and \(-\sqrt{-1}\), respectively:

$$\begin{aligned} T^{1,0}M=\{X-\sqrt{-1}JX \big | X\in TM\},\quad T^{0,1}M=\{X+\sqrt{-1}JX \big | X\in TM\}. \end{aligned}$$

Let \(\Lambda ^rM=\bigoplus _{p+q=r}\Lambda ^{p,q}M\) for \(0\le r\le 2n\) denote the decomposition of complex differential r-forms into (pq)-forms, where \(\Lambda ^{p,q}M=\Lambda ^p(\Lambda ^{1,0}M)\otimes \Lambda ^q(\Lambda ^{0,1}M)\),

$$\begin{aligned} \Lambda ^{1,0}M=\{\eta +\sqrt{-1}J\eta \big | \eta \in \Lambda ^1M\},\quad \Lambda ^{0,1}M=\{\eta -\sqrt{-1}J\eta \big | \eta \in \Lambda ^1M\} \end{aligned}$$

and \(\Lambda ^1M\) denotes the dual of \(T^{{\mathbb {C}}}M\).

Let \(\{Z_r\}\) be a local (1, 0)-frame on (MJ) with an almost Hermitian metric g and let \(\{\zeta ^r\}\) be a local associated coframe with respect to \(\{Z_r\}\), i.e.,

$$\begin{aligned} \zeta ^i(Z_j)=\delta _j^i,\quad i,j=1,\dots ,n. \end{aligned}$$

Since g is almost Hermitian, its components satsfy \(g_{ij}=g_{{\bar{i}}{\bar{j}}}=0\) and \(g_{i{\bar{j}}}=g_{{\bar{j}}i}={\bar{g}}_{{\bar{i}}j}\). By using these local frame \(\{Z_r\}\) and coframe \(\{\zeta ^r\}\), we have

$$\begin{aligned} N(Z_{{\bar{i}}},Z_{{\bar{j}}})=-[Z_{{\bar{i}}},Z_{{\bar{j}}}]^{(1,0)}=:N_{{\bar{i}}{\bar{j}}}^kZ_k,\quad N(Z_i,Z_j)=-[Z_i,Z_j]^{(0,1)}=\overline{N_{{\bar{i}}{\bar{j}}}^k}Z_{{\bar{k}}}, \end{aligned}$$

and

$$\begin{aligned} N=\frac{1}{2}\overline{N_{{\bar{i}}{\bar{j}}}^k}Z_{{\bar{k}}}\otimes (\zeta ^i\wedge \zeta ^j)+\frac{1}{2}N_{{\bar{i}}{\bar{j}}}^kZ_k\otimes (\zeta ^{{\bar{i}}}\wedge \zeta ^{{\bar{j}}}). \end{aligned}$$

Let (MgJ) be an almost Hermitian manifold with \(\dim _{{\mathbb {R}}}M=2n\). An affine connection D on \(T^{{\mathbb {C}}}M\) is called almost Hermitian connection if \(Dg=DJ=0\). For the almost Hermitian connection, we have the following Lemma (cf. [1]).

Lemma 2.1

Let (MJg) be an almost Hermitian manifold with \(\dim _{{\mathbb {R}}}M=2n\). Then for any given vector valued (1, 1)-form \(\Theta =(\Theta ^i)_{1\le i\le n}\), there exists a unique almost Hermitian connection \(\nabla \) on (MJg) such that the (1, 1)-part of the torsion is equal to the given \(\Theta \).

If the (1, 1)-part of the torsion of an almost Hermitian connection vanishes everywhere, then the connction is called the second canonical connection or the Chern connection. We will refer the connection as the Chern connection and denote it by \(\nabla \).

Now let \(\nabla \) be the Chern connection on M. We can write

$$\begin{aligned}&[Z_i,Z_j]=B_{ij}^rZ_r+B_{ij}^{{\bar{r}}}Z_{{\bar{r}}},\quad [Z_i,Z_{{\bar{j}}}]=B_{i{\bar{j}}}^rZ_r+B_{i{\bar{j}}}^{{\bar{r}}}Z_{{\bar{r}}},\\&\quad [Z_{{\bar{i}}},Z_{{\bar{j}}}]=B_{{\bar{i}}{\bar{j}}}^rZ_r+B_{{\bar{i}}{\bar{j}}}^{{\bar{r}}}Z_{{\bar{r}}} \end{aligned}$$

and also we here note that for instance, \([Z_i,Z_{{\bar{j}}}]=[Z_i,Z_{{\bar{j}}}]^{(1,0)}+[Z_i,Z_{{\bar{j}}}]^{(0,1)},\) where

$$\begin{aligned}&[Z_i,Z_{{\bar{j}}}]^{(1,0)}=\frac{1}{2}([Z_i,Z_{{\bar{j}}}]-\sqrt{-1}J[Z_i,Z_{{\bar{j}}}]), \\&[Z_i,Z_{{\bar{j}}}]^{(0,1)}=\frac{1}{2}([Z_i,Z_{{\bar{j}}}]+\sqrt{-1}J[Z_i,Z_{{\bar{j}}}]). \end{aligned}$$

For any p-form \(\psi \), there holds that

$$\begin{aligned} d\psi (X_1,\dots ,X_{p+1})= & {} \sum _{i=1}^{p+1}(-1)^{i+1}X_i(\psi (X_1,\dots ,\widehat{X_i},\dots ,X_{p+1}))\\&+\sum _{i<j}(-1)^{i+j}\psi ([X_i,X_j],X_1,\dots ,\widehat{X_i},\dots ,\widehat{X_j},\dots ,X_{p+1}) \end{aligned}$$

for any vector fields \(X_1,\dots ,X_{p+1}\) on M (cf. [13]). We directly compute that

$$\begin{aligned} d\zeta ^s=-\frac{1}{2}B_{kl}^s\zeta ^k\wedge \zeta ^l-B_{k{\bar{l}}}^s\zeta ^k\wedge \zeta ^{{\bar{l}}}-\frac{1}{2}B_{{\bar{k}}{\bar{l}}}^s\zeta ^{{\bar{k}}}\wedge \zeta ^{{\bar{l}}}. \end{aligned}$$

For any real (1, 1)-form \(\eta =\sqrt{-1}\eta _{i{\bar{j}}}\zeta ^i\wedge \zeta ^{{\bar{j}}}\), we have

$$\begin{aligned}&\partial \eta =\frac{\sqrt{-1}}{2}\Big (Z_{i}(\eta _{j{\bar{k}}})-Z_{j}(\eta _{i{\bar{k}}})-B_{ij}^s\eta _{s{\bar{k}}}-B_{i{\bar{k}}}^{{\bar{s}}}\eta _{j{\bar{s}}}+B_{j{\bar{k}}}^{{\bar{s}}}\eta _{i{\bar{s}}}\Big )\zeta ^i\wedge \zeta ^{j}\wedge \zeta ^{{\bar{k}}}, \\&{\bar{\partial }}\eta =\frac{\sqrt{-1}}{2}\Big (Z_{{\bar{j}}}(\eta _{k{\bar{i}}})-Z_{{\bar{i}}}(\eta _{k{\bar{j}}})-B_{k{\bar{i}}}^s\eta _{s{\bar{j}}}+B_{k{\bar{j}}}^s\eta _{s{\bar{i}}}+B_{{\bar{i}}{\bar{j}}}^{{\bar{s}}}\eta _{k{\bar{s}}}\Big )\zeta ^k\wedge \zeta ^{{\bar{i}}}\wedge \zeta ^{{\bar{j}}}. \end{aligned}$$

From these computations above, we have

$$\begin{aligned} \partial \omega&=\frac{\sqrt{-1}}{2}\Big (Z_{i}(g_{j{\bar{k}}})-Z_{j}(g_{i{\bar{k}}})-B_{ij}^sg_{s{\bar{k}}}-B_{i{\bar{k}}}^{{\bar{s}}}g_{j{\bar{s}}}+B_{j{\bar{k}}}^{{\bar{s}}}g_{i{\bar{s}}}\Big )\zeta ^i\wedge \zeta ^{j}\wedge \zeta ^{{\bar{k}}}\\&=\frac{\sqrt{-1}}{2}T_{ij{\bar{k}}}\zeta ^i\wedge \zeta ^{j}\wedge \zeta ^{{\bar{k}}} \end{aligned}$$

and

$$\begin{aligned} {\bar{\partial }}\omega&=\frac{\sqrt{-1}}{2}\Big (Z_{{\bar{j}}}(g_{k{\bar{i}}})-Z_{{\bar{i}}}(g_{k{\bar{j}}})-B_{k{\bar{i}}}^s g_{s{\bar{j}}}+B_{k{\bar{j}}}^s g_{s{\bar{i}}}+B_{{\bar{i}}{\bar{j}}}^{{\bar{s}}} g_{k{\bar{s}}}\Big )\zeta ^k\wedge \zeta ^{{\bar{i}}}\wedge \zeta ^{{\bar{j}}}\\&=\frac{\sqrt{-1}}{2}T_{{\bar{j}}{\bar{i}}k}\zeta ^k\wedge \zeta ^{{\bar{i}}}\wedge \zeta ^{{\bar{j}}}. \end{aligned}$$

And in this sense, we also obtain for any (0, 1)-form \(\beta \),

$$\begin{aligned} (\partial \beta )_{k{\bar{j}}}=Z_k(\beta _{{\bar{j}}})-B_{k{\bar{j}}}^{{\bar{m}}}\beta _{{\bar{m}}}=\nabla _k\beta _{{\bar{j}}}. \end{aligned}$$

2.2 The Torsion and the Curvature on Almost Complex Manifolds

Since the Chern connection \(\nabla \) preserves J, we have

$$\begin{aligned} \nabla _iZ_j=\Gamma _{ij}^rZ_r,\quad \nabla _iZ_{{\bar{j}}}=\Gamma _{i{\bar{j}}}^{{\bar{r}}}Z_{{\bar{r}}}, \end{aligned}$$

where

$$\begin{aligned} \Gamma _{ij}^r=g^{r{\bar{s}}}Z_i(g_{j{\bar{s}}})-g^{r{\bar{s}}}g_{j{\bar{l}}}B_{i{\bar{s}}}^{{\bar{l}}},\quad \Gamma _{ip}^p=Z_i(\log \det g)-B_{i{\bar{s}}}^{{\bar{s}}}. \end{aligned}$$

We can obtain that

$$\begin{aligned} \Gamma _{i{\bar{j}}}^{{\bar{r}}}=B_{i{\bar{j}}}^{{\bar{r}}} \end{aligned}$$

since the (1, 1)-part of the torsion of the Chern connection vanishes everywhere (cf. [4]). For any (0, 1)-form \(\beta \), we have \(\beta =\beta _{{\bar{j}}}\zeta ^{{\bar{j}}}\),

$$\begin{aligned} \nabla _k\beta _{{\bar{j}}}=Z_k(\beta _{{\bar{j}}})-\Gamma _{k{\bar{j}}}^{{\bar{l}}}\beta _{{\bar{l}}}=Z_k(\beta _{{\bar{j}}})-B_{k{\bar{j}}}^{{\bar{l}}}\beta _{{\bar{l}}}. \end{aligned}$$

Note that the mixed derivatives \(\nabla _iZ_{{\bar{j}}}\) do not depend on g (cf. [11]). Let \(\{\gamma _j^i\}\) be the connection form, which is defined by \(\gamma _j^i=\Gamma _{sj}^i\zeta ^s+\Gamma _{{\bar{s}}j}^i\zeta ^{{\bar{s}}}\). The torsion T of the Chern connection \(\nabla \) is given by \(T^i=d\zeta ^i-\zeta ^p\wedge \gamma ^i_p\), \(T^{{\bar{i}}}=d\zeta ^{{\bar{i}}}-\zeta ^{{\bar{p}}}\wedge \gamma _{{\bar{p}}}^{{\bar{i}}}\), which has no (1, 1)-part and the only non-vanishing components are as follows:

$$\begin{aligned} T_{ij}^s=\Gamma _{ij}^s-\Gamma _{ji}^s-B_{ij}^s\quad T_{ij}^{{\bar{s}}}=-B_{ij}^{{\bar{s}}}. \end{aligned}$$

These tell us that \(T=(T^i)\) splits into \(T=T'+T''\), where \(T'\in \Gamma (\Lambda ^{2,0}M\otimes T^{1,0}M)\), \(T''\in \Gamma (\Lambda ^{0,2}M\otimes T^{1,0}M)\). We also lower the index of torsion and denote it by

$$\begin{aligned} T_{ij{\bar{k}}}=T_{ij}^sg_{s{\bar{k}}}=Z_i(g_{j{\bar{k}}})-Z_j(g_{i{\bar{k}}})+B_{{\bar{k}}i}^{{\bar{q}}}g_{j{\bar{q}}}-B_{{\bar{k}}j}^{{\bar{q}}}g_{i{\bar{q}}}-B_{ij}^sg_{s{\bar{k}}}. \end{aligned}$$

Note that \(T''\) depends only on J and it can be regarded as the Nijenhuis tensor of J, that is, J is integrable if and only if \(T''\) vanishes.

We denote by \(\Omega \) the curvature of the Chern connection \(\nabla \). We can regard \(\Omega \) as a section of \(\Lambda ^2M\otimes \Lambda ^{1,1}M\), \(\Omega \in \Gamma (\Lambda ^2M\otimes \Lambda ^{1,1}M)\) and \(\Omega \) splits in \(\Omega =H+R+{\bar{H}}\), where \(R\in \Gamma (\Lambda ^{1,1}M\otimes \Lambda ^{1,1}M)\), \(H\in \Gamma (\Lambda ^{2,0}M\otimes \Lambda ^{1,1}M)\). The curvature form can be expressed by \(\Omega _j^i=d\gamma _j^i+\gamma _s^i\wedge \gamma _j^s\).

In terms of \(Z_r\)’s, we have

$$\begin{aligned}&R_{i{\bar{j}}k}^{\quad \,\, r}=\Omega _k^r(Z_i,Z_{{\bar{j}}})=Z_i(\Gamma _{{\bar{j}}k}^r)-Z_{{\bar{j}}}(\Gamma _{ik}^r)\\&\quad \quad \quad \quad +\Gamma _{is}^r\Gamma _{{\bar{j}}k}^s-\Gamma _{{\bar{j}}s}^r\Gamma _{ik}^s-B_{i{\bar{j}}}^s\Gamma _{sk}^r+B_{{\bar{j}}i}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^r=-R_{{\bar{j}}ik}^{\quad \,\,r}, \\&H_{ijk}^{\quad \,\,r}=\Omega _k^r(Z_i,Z_j)=Z_i(\Gamma _{jk}^r)-Z_{j}(\Gamma _{ik}^r)+\Gamma _{is}^r\Gamma _{jk}^s-\Gamma _{js}^r\Gamma _{ik}^s-B_{ij}^s\Gamma _{sk}^r\\&\quad \quad \quad \quad -B_{ij}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^r=-H_{jik}^{\quad \,\,r}, \\&H_{{\bar{i}}{\bar{j}}k}^{\quad \,\,r}=\Omega _k^r(Z_{{\bar{i}}},Z_{{\bar{j}}})=Z_{{\bar{i}}}(\Gamma _{{\bar{j}}k}^r)-Z_{{\bar{j}}}(\Gamma _{{\bar{i}}k}^r)+\Gamma _{{\bar{i}}s}^r\Gamma _{{\bar{j}}k}^s-\Gamma _{{\bar{j}}s}^r\Gamma _{{\bar{i}}k}^s-B_{{\bar{i}}{\bar{j}}}^s\Gamma _{sk}^r\\&\quad \quad \quad \quad -B_{{\bar{i}}{\bar{j}}}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^r=-H_{{\bar{j}}{\bar{i}}k}^{\quad \,\,r}. \end{aligned}$$

We can write \(\Omega =(\Omega _j^i)=\Omega ^{(2,0)}+\Omega ^{(1,1)}+\Omega ^{(0,2)}=H+R+{\bar{H}}\), with

$$\begin{aligned} \Omega ^{(2,0)}&=\Big (\frac{1}{2}H_{ijk}^{\quad \,\,r}\zeta ^i\wedge \zeta ^j\Big ),\quad \Omega ^{(1,1)}=\Big (R_{i{\bar{j}}k}^{\quad \,\,r}\zeta ^i\wedge \zeta ^{{\bar{j}}}\Big ),\quad \\&\Omega ^{(0,2)}=\Big (\frac{1}{2}H_{{\bar{i}}{\bar{j}}k}^{\quad \,\,r}\zeta ^{{\bar{i}}}\wedge \zeta ^{{\bar{j}}}\Big ). \end{aligned}$$

Then the Chern–Ricci form is \((\sqrt{-1}\Omega _i^i)\in c_1(M,J)\in H^2(M,{\mathbb {R}})\), where \(c_1(M,J)\) is the first Chern class of (MJ).

We deduce that by using \(\Gamma _{kp}^p=Z_k(\log \det g)-B_{k{\bar{p}}}^{{\bar{p}}}\),

$$\begin{aligned} P_{i{\bar{j}}}= & {} R_{i{\bar{j}}r}^{\quad \,\,r}\\= & {} Z_i(\Gamma _{{\bar{j}}r}^r)-Z_{{\bar{j}}}(\Gamma _{ir}^r)-B_{i{\bar{j}}}^s\Gamma _{sr}^r-B_{i{\bar{j}}}^{{\bar{s}}}\Gamma _{{\bar{s}}r}^r\\= & {} Z_i(\Gamma _{{\bar{j}}r}^r)-Z_{{\bar{j}}}Z_i(\log \det g)+Z_{{\bar{j}}}(B_{i{\bar{r}}}^{{\bar{r}}})-B_{i{\bar{j}}}^sZ_s(\log \det g)+B_{i{\bar{j}}}^sB_{s{\bar{r}}}^{{\bar{r}}}-B_{i{\bar{j}}}^{{\bar{s}}}\Gamma _{{\bar{s}}r}^r\\= & {} Z_i(\Gamma _{{\bar{j}}r}^r)+[Z_i,Z_{{\bar{j}}}](\log \det g)-Z_iZ_{{\bar{j}}}(\log \det g)+Z_{{\bar{j}}}(B_{i{\bar{r}}}^{{\bar{r}}})\\&-[Z_i,Z_{{\bar{j}}}]^{(1,0)}(\log \det g)+B_{i{\bar{j}}}^sB_{s{\bar{r}}}^{{\bar{r}}}-B_{i{\bar{j}}}^{{\bar{s}}}\Gamma _{{\bar{s}}r}^r\\= & {} -(Z_iZ_{{\bar{j}}}-[Z_i,Z_{{\bar{j}}}]^{(0,1)})(\log \det g)+Z_{{\bar{j}}}(B_{i{\bar{r}}}^{{\bar{r}}})+Z_i(B_{{\bar{j}}r}^r)+B_{i{\bar{j}}}^sB_{s{\bar{r}}}^{{\bar{r}}}-B_{i{\bar{j}}}^{{\bar{s}}}B_{{\bar{s}}r}^r\\ R_{ij}= & {} H_{ijr}^{\quad \,\,r}\\= & {} Z_i(\Gamma _{jr}^r)-Z_{j}(\Gamma _{ir}^r)-B_{ij}^s\Gamma _{sr}^r-B_{ij}^{{\bar{s}}}\Gamma _{{\bar{s}}r}^r\\= & {} Z_iZ_j(\log \det g)-Z_i(B_{j{\bar{r}}}^{{\bar{r}}})-Z_jZ_i(\log \det g)+Z_j(B_{i{\bar{r}}}^{{\bar{r}}})\\&-B_{ij}^sZ_s(\log \det g)+B_{ij}^sB_{s{\bar{r}}}^{{\bar{r}}}-B_{ij}^{{\bar{s}}}B_{{\bar{s}}r}^r\\= & {} ([Z_i,Z_j]-[Z_i,Z_j]^{(1,0)})(\log \det g)-Z_i(B_{j{\bar{r}}}^{{\bar{r}}})+Z_j(B_{i{\bar{r}}}^{{\bar{r}}}) +B_{ij}^sB_{s{\bar{r}}}^{{\bar{r}}}-B_{ij}^{{\bar{s}}}B_{{\bar{s}}r}^r\\= & {} [Z_i,Z_j]^{(0,1)}(\log \det g)-Z_i(B_{j{\bar{r}}}^{{\bar{r}}})+Z_j(B_{i{\bar{r}}}^{{\bar{r}}}) +B_{ij}^sB_{s{\bar{r}}}^{{\bar{r}}}-B_{ij}^{{\bar{s}}}B_{{\bar{s}}r}^r \end{aligned}$$

and

$$\begin{aligned} R_{{\bar{i}}{\bar{j}}}= & {} H_{{\bar{i}}{\bar{j}}r}^{\quad \,\,r}\\= & {} Z_{{\bar{i}}}(\Gamma _{{\bar{j}}r}^r)-Z_{{\bar{j}}}(\Gamma _{{\bar{i}}r}^r) -B_{{\bar{i}}{\bar{j}}}^s\Gamma _{sr}^r-B_{{\bar{i}}{\bar{j}}}^{{\bar{s}}}\Gamma _{{\bar{s}}r}^r\\= & {} Z_{{\bar{i}}}(B_{{\bar{j}}r}^r)-Z_{{\bar{j}}}(B_{{\bar{i}}r}^r)-B_{{\bar{i}}{\bar{j}}}^sZ_s(\log \det g)+B_{{\bar{i}}{\bar{j}}}^sB_{s{\bar{r}}}^{{\bar{r}}}-B_{{\bar{i}}{\bar{j}}}^{{\bar{s}}}B_{{\bar{s}}r}^r\\= & {} -[Z_{{\bar{i}}},Z_{{\bar{j}}}]^{(1,0)}(\log \det g)+Z_{{\bar{i}}}(B_{{\bar{j}}r}^r)-Z_{{\bar{j}}}(B_{{\bar{i}}r}^r)+B_{{\bar{i}}{\bar{j}}}^sB_{s{\bar{r}}}^{{\bar{r}}}-B_{{\bar{i}}{\bar{j}}}^{{\bar{s}}}B_{{\bar{s}}r}^r. \end{aligned}$$

The Chern–Ricci form \(\mathop {\mathrm {Ric}}\nolimits (\omega )\) is defined by

$$\begin{aligned} \mathop {\mathrm {Ric}}\nolimits (\omega ):=\frac{\sqrt{-1}}{2}R_{kl}\zeta ^k\wedge \zeta ^l+\sqrt{-1}P_{k{\bar{l}}}\zeta ^k\wedge \zeta ^{{\bar{l}}}+\frac{\sqrt{-1}}{2}R_{{\bar{k}}{\bar{l}}}\zeta ^{{\bar{k}}}\wedge \zeta ^{{\bar{l}}}. \end{aligned}$$

It is a closed real 2-form. If J is integrable, it is a closed real (1, 1)-form. If furthermore, J is integrable and \(d\omega =0\), then the Chern–Ricci form coincides with the Ricci form defined by the Levi-Civita connection of \(\omega \). Assume that \({\tilde{\omega }}=\sqrt{-1}{\tilde{g}}_{i{\bar{j}}}\zeta ^i\wedge \zeta ^{{\bar{j}}}\) is another almost Hermitian metric. Then we have

$$\begin{aligned} \mathop {\mathrm {Ric}}\nolimits ({\tilde{\omega }})-\mathop {\mathrm {Ric}}\nolimits (\omega )=-\frac{1}{2}dJd\log \frac{{\tilde{\omega }}^n}{\omega ^n}, \end{aligned}$$

with \(\mathop {\mathrm {Ric}}\nolimits (\omega )\in 2\pi c_1(M,J)\in H^2(M,{\mathbb {R}})\) (cf. [13]).

Lemma 2.2

(The first Bianchi identity for the Chern curvature) For any \(X,Y,Z\in T^{{\mathbb {C}}}M\),

$$\begin{aligned} \sum \Omega (X,Y)Z=\sum \Big (T(T(X,Y),Z)+\nabla _XT(Y,Z)\Big ), \end{aligned}$$

where the sum is taken over all cyclic permutations.

This identity induces the following formulae:

$$\begin{aligned}&R_{i{\bar{j}}k}^{\quad \,\,l}=R_{k{\bar{j}}i}^{\quad \,\,l}-T_{ik}^{{\bar{r}}}T_{{\bar{r}}{\bar{j}}}^l+\nabla _{{\bar{j}}}T_{ki}^l=R_{k{\bar{j}}i}^{\quad \,\,l}-B_{ik}^{{\bar{r}}}B_{{\bar{r}}{\bar{j}}}^l+\nabla _{{\bar{j}}}T_{ki}^l, \end{aligned}$$
(2.1)
$$\begin{aligned}&H_{ijk}^{\quad \,\,l}=T_{ji}^{{\bar{r}}}T_{{\bar{r}}{\bar{l}}}^{{\bar{k}}}+\nabla _{{\bar{l}}}T_{ji}^{{\bar{k}}}=-B_{ji}^{{\bar{r}}}T_{{\bar{r}}{\bar{l}}}^{{\bar{k}}}-\nabla _{{\bar{l}}}B_{ji}^{{\bar{k}}}, \end{aligned}$$
(2.2)

where used that \(R_{ij{\bar{k}}{\bar{l}}}=R_{{\bar{i}}{\bar{j}}kl}=H_{j{\bar{l}}ik}=H_{{\bar{j}}l{\bar{i}}{\bar{k}}}=H_{{\bar{l}}ijk}=H_{l{\bar{i}}{\bar{j}}{\bar{k}}}=0\).

Lemma 2.3

(The second Bianchi identity for the Chern curvature) For any \(X, Y, Z\in T^{{\mathbb {C}}}M\),

$$\begin{aligned} \sum \nabla _X\Omega (Y,Z)=-\sum \Omega (T(X,Y),Z), \end{aligned}$$

where the sum is taken over all cyclic permutations.

This identity induces the following formulae:

$$\begin{aligned}&\nabla _iR_{r{\bar{s}}k}^{\quad \,\,l} =\nabla _rR_{i{\bar{s}}k}^{\quad \,\,l}+\nabla _{{\bar{s}}}H_{rik}^{\quad \,\,l} +T_{ri}^mR_{m{\bar{s}}k}^{\quad \,\,l}+T_{ri}^{{\bar{m}}}H_{{\bar{m}}{\bar{s}}k}^{\quad \,\,l}, \end{aligned}$$
(2.3)
$$\begin{aligned}&\nabla _{{\bar{i}}}H_{rsk}^{\quad \,\,l}=\nabla _sR_{r{\bar{i}}k}^{\quad \,\,l} +\nabla _rR_{{\bar{i}}sk}^{\quad \,\,l}+T_{rs}^qR_{q{\bar{i}}k}^{\quad \,\,l} +T_{rs}^{{\bar{q}}}H_{{\bar{q}}{\bar{i}}k}^{\quad \,\,l}. \end{aligned}$$
(2.4)

Taking into account the Bianchi identities, we have the following lemma:

Lemma 2.4

([11, Lemma 3.5]) The following formula holds

$$\begin{aligned} P=S+\text {div}^{\nabla }T'-{\nabla }{\bar{w}}+Q^7+Q^8 \end{aligned}$$
(2.5)

for any almost Hermitian metric g , where \(T'\) is the torsion of the Chern connection \(\nabla \) associated to g,

$$\begin{aligned} (\text {div}^{\nabla }T')_{i{\bar{j}}}=g^{k{\bar{l}}}\nabla _{{\bar{l}}}T_{ki{\bar{j}}},\quad (\nabla {\bar{w}})_{i{\bar{j}}}=g^{k{\bar{l}}}\nabla _iT_{{\bar{j}}{\bar{l}}k}. \end{aligned}$$

We define the curvature condition as follows:

Definition 2.1

We say that an almost Hermitian manifold (MJg) has the Griffiths non-positive Chern curvature if there is non-positive function \(\kappa \) such that for any \(p\in M\), \(X,Y\in T_p^{1,0}M\),

$$\begin{aligned} R(X,{\bar{X}},Y,{\bar{Y}})\le \kappa (p) B(X,{\bar{X}},Y,{\bar{Y}}), \end{aligned}$$

where \(B_{i{\bar{j}}k{\bar{l}}}:=g_{i{\bar{j}}}g_{k{\bar{l}}}+g_{i{\bar{l}}}g_{k{\bar{j}}}\).

Definition 2.2

We say that an almost Hermitian manifold (MJg) has the first Chern–Ricci curvature bounded above by a function \(\kappa \) if for any \(p\in M\), \(X\in T^{1,0}_pM\),

$$\begin{aligned} P(X,{\bar{X}})\le \kappa (p) g(X,{\bar{X}}). \end{aligned}$$

If the function \(\kappa \) is non-positive (resp. negative), then we say that g has the non-positive (resp. negative) first Chern–Ricci curvature. If the function \(\kappa \) is non-positive and negative at some point, then we say that g has the quasi-negative first Chern–Ricci curvature.

3 Evolution Equations Along the Almost Hermitian Curvature Flow

Let \((M^{2n},g,J)\) be a compact almost Hermitian manifold. Let \(\nabla \) be the Chern connection on \((M^{2n},g,J)\). Let \(\{Z_r\}\) be a local unitary (1, 0)-frame with respect to g around a fixed point \(p\in M\). Note that unitary frames always exist locally since we can take any frame and apply the Gram-Schmidt process. Then with respect to a local g-unitary frame, we have \(g_{i{\bar{j}}}=\delta _{ij}\), \(Z_k(g_{i{\bar{j}}})=0\) for any \(i,j,k=1,\dots ,n\), and the Christoffel symbols satisfy

$$\begin{aligned} \Gamma _{ij}^k=-\Gamma _{i{\bar{k}}}^{{\bar{j}}},\quad \Gamma _{{\bar{i}}{\bar{j}}}^{{\bar{k}}}=-\Gamma _{{\bar{i}}k}^{j}, \end{aligned}$$

since we have

$$\begin{aligned}&\Gamma _{ij}^k=g(\nabla _iZ_j,Z_{{\bar{k}}})=Z_i(g_{j{\bar{k}}})-g(Z_j,\nabla _iZ_{{\bar{k}}})=-\Gamma _{i{\bar{k}}}^{{\bar{j}}}, \\&\Gamma _{{\bar{i}}{\bar{j}}}^{{\bar{k}}}=g(Z_k,\nabla _{{\bar{i}}}Z_{{\bar{j}}})=Z_{{\bar{i}}}(g_{k{\bar{j}}})-g(\nabla _{{\bar{i}}}Z_k,Z_{{\bar{j}}})=-\Gamma _{{\bar{i}}k}^{j}. \end{aligned}$$

With respect to such a frame, the components of the torsion can be written as

$$\begin{aligned} T_{ij}^k=-B_{i{\bar{k}}}^{{\bar{j}}}+B_{j{\bar{k}}}^{{\bar{i}}}-B_{ij}^k \end{aligned}$$

and the components of w can be written as

$$\begin{aligned} w_j=-B_{jr}^r-B_{j{\bar{r}}}^{{\bar{r}}}+B_{r{\bar{r}}}^{{\bar{j}}}. \end{aligned}$$

And also we have

$$\begin{aligned} R_{i{\bar{j}}k}^{\quad \,\,r}= & {} Z_i(\Gamma _{{\bar{j}}k}^r)-Z_{{\bar{j}}}(\Gamma _{ik}^r)+\Gamma _{is}^r\Gamma _{{\bar{j}}k}^s-\Gamma _{{\bar{j}}s}^r\Gamma _{ik}^s-B_{i{\bar{j}}}^s\Gamma _{sk}^r+B_{{\bar{j}}i}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^r \\= & {} -Z_i(\Gamma _{{\bar{j}}{\bar{r}}}^{{\bar{k}}})+Z_{{\bar{j}}}(\Gamma _{i{\bar{r}}}^{{\bar{k}}})+\Gamma _{i{\bar{r}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{k}}}-\Gamma _{{\bar{j}}{\bar{r}}}^{{\bar{s}}}\Gamma _{i{\bar{s}}}^{{\bar{k}}} +B_{i{\bar{j}}}^s\Gamma _{s{\bar{r}}}^{{\bar{k}}}-B_{{\bar{j}}i}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{r}}}^{{\bar{k}}} \\= & {} -R_{i{\bar{j}}{\bar{r}}}^{\quad \,\,{\bar{k}}},\\ H_{ijk}^{\quad \,\,r}= & {} Z_i(\Gamma _{jk}^r)-Z_{j}(\Gamma _{ik}^r)+\Gamma _{is}^r\Gamma _{jk}^s-\Gamma _{js}^r\Gamma _{ik}^s-B_{ij}^s\Gamma _{sk}^r-B_{ij}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^r\\= & {} -Z_i(\Gamma _{j{\bar{r}}}^{{\bar{k}}})-Z_{j}(\Gamma _{i{\bar{r}}}^{{\bar{k}}}) +\Gamma _{i{\bar{r}}}^{{\bar{s}}}\Gamma _{j{\bar{s}}}^{{\bar{k}}} -\Gamma _{j{\bar{r}}}^{{\bar{s}}}\Gamma _{i{\bar{s}}}^{{\bar{k}}} +B_{ij}^s\Gamma _{s{\bar{r}}}^{{\bar{k}}} +B_{ij}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{r}}}^{{\bar{k}}}\\= & {} -H_{ij{\bar{r}}}^{\quad \,\,{\bar{k}}} \end{aligned}$$

and

$$\begin{aligned} \overline{R_{i{\bar{j}}k}^{\quad \,\,r}}= & {} Z_{{\bar{i}}}(\Gamma _{j{\bar{k}}}^{{\bar{r}}})-Z_{j}(\Gamma _{{\bar{i}}{\bar{k}}}^{{\bar{r}}}) +\Gamma _{{\bar{i}}{\bar{s}}}^{{\bar{r}}}\Gamma _{j{\bar{k}}}^{{\bar{s}}} -\Gamma _{j{\bar{s}}}^{{\bar{r}}}\Gamma _{{\bar{i}}{\bar{k}}}^{{\bar{s}}}-B_{{\bar{i}}j}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{k}}}^{{\bar{r}}} +B_{j{\bar{i}}}^{s}\Gamma _{s{\bar{k}}}^{{\bar{r}}} \\= & {} Z_j(\Gamma _{{\bar{i}}r}^{k})-Z_{{\bar{i}}}(\Gamma _{jr}^{k})+\Gamma _{{\bar{i}}r}^{s}\Gamma _{js}^{k} -\Gamma _{jr}^{s}\Gamma _{{\bar{i}}s}^{k} -B_{j{\bar{i}}}^s\Gamma _{sr}^{k}+B_{{\bar{i}}j}^{{\bar{s}}}\Gamma _{{\bar{s}}r}^{k} \\= & {} R_{j{\bar{i}}r}^{\quad \,\,k},\\ \overline{H_{ijk}^{\quad \,\,r}}= & {} Z_{{\bar{i}}}(\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{r}}}) -Z_{{\bar{j}}}(\Gamma _{{\bar{i}}{\bar{k}}}^{{\bar{r}}}) +\Gamma _{{\bar{i}}{\bar{s}}}^{{\bar{r}}}\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{s}}} -\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{r}}}\Gamma _{{\bar{i}}{\bar{k}}}^{{\bar{s}}} -B_{{\bar{i}}{\bar{j}}}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{k}}}^{{\bar{r}}} -B_{{\bar{i}}{\bar{j}}}^{s}\Gamma _{s{\bar{k}}}^{{\bar{r}}}\\= & {} -Z_{{\bar{i}}}(\Gamma _{{\bar{j}}r}^{k}) +Z_{{\bar{j}}}(\Gamma _{{\bar{i}}r}^{k}) +\Gamma _{{\bar{i}}r}^{s}\Gamma _{{\bar{j}}s}^{k} -\Gamma _{{\bar{j}}r}^{s}\Gamma _{{\bar{i}}s}^{k} -B_{{\bar{j}}{\bar{i}}}^{{\bar{s}}}\Gamma _{{\bar{s}}r}^{k} -B_{{\bar{j}}{\bar{i}}}^{s}\Gamma _{sr}^{k}\\= & {} H_{{\bar{j}}{\bar{i}}r}^{\quad \,\,k}. \end{aligned}$$

Hence we obtain \(R_{i{\bar{j}}k{\bar{r}}}=-R_{i{\bar{j}}{\bar{r}}k}\), \(H_{ijk{\bar{r}}}=-H_{ij{\bar{r}}k}\) and \(\overline{R_{i{\bar{j}}k{\bar{r}}}}=R_{j{\bar{i}}r{\bar{k}}}\), \(\overline{H_{ijk{\bar{r}}}}=H_{{\bar{j}}{\bar{i}}r{\bar{k}}}\) by using a local unitary frame with respect to g.

Let \(B^{\circ }\) be the terms of B’s depending only on J, which means that these terms do not depend on t along with the solution to the AHCF. Note that \(B_{j{\bar{b}}}^{{\bar{q}}}\), \(B_{{\bar{j}}b}^{q}\)’s do not depend on g, which depend only on J since the mixed derivatives \(\nabla _jZ_{{\bar{b}}}\), \(\nabla _{{\bar{j}}}Z_b\) do not depend on g. Since we have \(B_{b{\bar{j}}}^{q}=-B_{{\bar{j}}b}^{q}\), we have that \(B_{b{\bar{j}}}^{q}\), \(B_{{\bar{b}}j}^{{\bar{q}}}\)’s also do not depend on g. Also note that \(B_{ri}^{{\bar{s}}}\), \(B_{{\bar{r}}{\bar{i}}}^{s}\) do not depend on g, depend only on J. These components \(B_{j{\bar{b}}}^{{\bar{q}}}\), \(B_{{\bar{j}}b}^{q}\), \(B_{b{\bar{j}}}^{q}\), \(B_{{\bar{b}}j}^{{\bar{q}}}\), \(B_{ri}^{{\bar{s}}}\) and \(B_{{\bar{r}}{\bar{i}}}^{s}\) are denoted by \(B^\circ \).

Lemma 3.1

(cf. [4, Lemma 3.1], [5, Lemma 3.1]) Let \((M,\omega (t),J)\) be a solution to the AHCF for \(t\in [0,\tau )\) starting at the initial almost Hermitian metric \(\omega _0\). Then one has for all \(l=0,1,2,\dots \), for any fixed time \(t_0\in [0,\tau )\),

$$\begin{aligned} |Z^l(\Gamma (g(t_0)))|_{g(t_0)}\le C_l \end{aligned}$$

for some uniform constant \(C_l>0\).

Proof

Fix an arbitrary chosen time \(t_0\in [0,\tau )\). By using a local unitary frame with respect to \(g(t_0)\), since we have \(g(t_0)_{i{\bar{j}}}=\delta _{ij}\), and \(\Gamma _{ij}^k(g(t_0))=-\Gamma _{i{\bar{k}}}^{{\bar{j}}}(g(t_0))=-B_{i{\bar{k}}}^{{\bar{j}}}=-B^{\circ }\), which means that these coefficients do not depend on \(t_0\), we obtain on M,

$$\begin{aligned} |\Gamma _{ij}^k(g(t_0))|^2_{g(t_0)}= & {} g(t_0)_{k{\bar{l}}}g(t_0)^{i{\bar{r}}}g(t_0)^{j{\bar{s}}}\Gamma _{ij}^k(g(t_0)) \Gamma _{{\bar{r}}{\bar{s}}}^{{\bar{l}}}(g(t_0))\\= & {} \Gamma _{i{\bar{k}}}^{{\bar{j}}}(g(t_0)) \Gamma _{{\bar{i}}k}^{j}(g(t_0))\\= & {} B_{i{\bar{k}}}^{{\bar{j}}}B_{{\bar{i}}k}^{j}\le C_0 \end{aligned}$$

for some uniform constant \(C>0\) since \(B_{j{\bar{b}}}^{{\bar{q}}}\), \(B_{{\bar{j}}b}^{q}\)’s do not depend on g, which depend only on J because the mixed derivatives \(\nabla _jZ_{{\bar{b}}}\) do not depend on g (cf. [11, Lemma 5.2]).

Likewise, using a local unitary frame with respect to \(g(t_0)\), for all \(l=0,1,2,\dots \), we have

$$\begin{aligned} |Z^l(\Gamma _{ij}^{k}(g(t_0))|^2_{g(t_0)}=|Z^l(\Gamma _{i{\bar{k}}}^{{\bar{j}}}(g(t_0))|^2_{g(t_0)}=|Z^l(B_{i{\bar{k}}}^{{\bar{j}}})|^2_{g(t_0)}=|Z^l(B^{\circ })|^2_{g(t_0)}\le C_l \end{aligned}$$

for some uniform constants \(C_l>0\). \(\square \)

We introduce some evolution equations in the following.

Lemma 3.2

(cf. [11, Lemma 5.1, 5.2 and 5.5]) Let g(t) be a smooth family of metric on M compatible with J. We denote by \(h=\frac{\partial }{\partial t}g\) the variation of g. Then one has

$$\begin{aligned} \frac{\partial }{\partial t}\Gamma _{ij}^k=g^{k{\bar{l}}}\nabla _ih_{j{\bar{l}}},\quad \frac{\partial }{\partial t}\Gamma _{ij}^{{\bar{k}}}=0,\quad \frac{\partial }{\partial t}R_{i{\bar{j}}k{\bar{l}}}=R_{i{\bar{j}}k}^rh_{r{\bar{l}}}-\nabla _{{\bar{j}}}\nabla _ih_{k{\bar{l}}} \end{aligned}$$

and

$$\begin{aligned} \frac{\partial }{\partial t}T_{ij{\bar{k}}}=\nabla _ih_{j{\bar{k}}}-\nabla _jh_{i{\bar{k}}}+T_{ij}^mh_{m{\bar{k}}}. \end{aligned}$$

The second formula follows from the fact that the \(\Gamma _{ij}^{{\bar{k}}}\)’s do not depend on t.

We need the following computation for estimating components \({\bar{Z}}(T')_{i{\bar{j}}}\).

In order to avoid a notational quagmire, we adopte the following \(*\)-convention \(A_1*A_2\) between two quantities \(A_1\) and \(A_2\) with respect to a metric g:

  1. (i)

    Summation over pairs of maching upper and lower indices.

  2. (ii)

    Contraction on upper indices with respect to the metric.

  3. (iii)

    Contraction on lower indices with respect to the dual metrics.

Lemma 3.3

Let \((M^{2n},g,J)\) be a compact almost Hermitian manifold. Let \(\{Z_r\}\) be a local (1, 0)-frame with respect to g around a fixed point \(p\in M\). Then one has that

$$\begin{aligned} {\bar{Z}}(T')= & {} {\bar{Z}}(\Gamma )+Z(B^\circ )+B^\circ *\Gamma \nonumber \\&+T'*{\bar{\Gamma }}+\Gamma *{\bar{\Gamma }}+B^\circ *T'+B^\circ *{\bar{T}}'+T'*{\bar{T}}'+\Gamma *{\bar{T}}'\nonumber \\&+B^\circ *B^\circ +{\mathcal {O}}(Z(g))+{\mathcal {O}}({\bar{Z}}(g)). \end{aligned}$$
(3.1)

Moreover, we then have the estimate

$$\begin{aligned} {\bar{Z}}(T')\le C(|\nabla \Gamma |_g+ |\Gamma |^2_g+|T'|^2_g+1)\omega +{\mathcal {O}}(Z(g))+{\mathcal {O}}({\bar{Z}}(g)), \end{aligned}$$

where \(\omega \) is the associated real (1, 1)-form with respect to g.

Proof

Let \(\{Z_r\}\) be an arbitrary local (1, 0)-frame around a fixed point \(p\in M\). Now let D be the Levi-Civita connection with respect to g and let \(\nabla \) be the Chern connection with respect to g. The relation between D and \(\nabla \) is as follows (cf. [12, Lemma 3.1]):

$$\begin{aligned} g(D_YX,Z)=g(\nabla _YX,Z)+\frac{1}{2}(g(T(X,Y),Z)+g(T(Y,Z),X)-g(T(Z,X),Y)) \end{aligned}$$

for any tangent vector fields X, Y and Z. Here notice that the torsion T of the Chern connection \(\nabla \) is also defined as

$$\begin{aligned} T(X,Y)=\nabla _XY-\nabla _YX-[X,Y], \end{aligned}$$

and in this sense, we compute as follows with a local (1, 0)-frame \(\{Z_r\}\) with respect to g:

$$\begin{aligned} T_{ij}:= & {} T(Z_i,Z_j)=\nabla _{Z_i}Z_j-\nabla _{Z_j}Z_i-[Z_i,Z_j]\\= & {} \Gamma _{ij}^kZ_k-\Gamma _{ji}^kZ_k-B_{ij}^kZ_k-B_{ij}^{{\bar{k}}}Z_{{\bar{k}}}=T_{ij}^kZ_k+T_{ij}^{{\bar{k}}}Z_{{\bar{k}}}, \\ T_{i{\bar{j}}}= & {} T(Z_i,Z_{{\bar{j}}})=\nabla _{Z_i}Z_{{\bar{j}}}-\nabla _{Z_{{\bar{j}}}}Z_i-[Z_i,Z_{{\bar{j}}}]\\= & {} (B_{{\bar{j}}i}^k-\Gamma _{{\bar{j}}i}^k)Z_k+ (\Gamma _{i{\bar{j}}}^{{\bar{k}}}-B_{i{\bar{j}}}^{{\bar{k}}})Z_{{\bar{k}}}=0. \end{aligned}$$

Then we have the following:

$$\begin{aligned} g(D_rZ_i-D_iZ_r,Z_{{\bar{s}}})= & {} g(\nabla _rZ_i,Z_{{\bar{s}}})+\frac{1}{2}(g(T_{ir},Z_{{\bar{s}}})+g(T_{r{\bar{s}}},Z_{i})-g(T_{{\bar{s}}i},Z_{r}))\\&-g(\nabla _iZ_r,Z_{{\bar{s}}}) -\frac{1}{2}(g(T_{ri},Z_{{\bar{s}}})+g(T_{i{\bar{s}}},Z_{r})-g(T_{{\bar{s}}r},Z_{i}))\\= & {} \Gamma _{ri}^kg_{k{\bar{s}}}-\Gamma _{ir}^kg_{k{\bar{s}}}+T_{ir}^kg_{k{\bar{s}}}\\= & {} B_{ri}^kg_{k{\bar{s}}}\\= & {} g([Z_r,Z_i],Z_{{\bar{s}}}),\\ g(D_{{\bar{r}}}Z_i-D_iZ_{{\bar{r}}},Z_{{\bar{s}}})= & {} g(\nabla _{{\bar{r}}}Z_i,Z_{{\bar{s}}})+\frac{1}{2}(g(T_{i{\bar{r}}},Z_{{\bar{s}}})+g(T_{{\bar{r}}{\bar{s}}},Z_{i})-g(T_{{\bar{s}}i},Z_{{\bar{r}}}))\\&-g(\nabla _iZ_{{\bar{r}}},Z_{{\bar{s}}}) -\frac{1}{2}(g(T_{{\bar{r}}i},Z_{{\bar{s}}})+g(T_{i{\bar{s}}},Z_{{\bar{r}}})-g(T_{{\bar{s}}{\bar{r}}},Z_{i}))\\= & {} \Gamma _{{\bar{r}}i}^kg_{k{\bar{s}}}\\= & {} B_{{\bar{r}}i}^kg_{k{\bar{s}}}\\= & {} g([Z_{{\bar{r}}},Z_i],Z_{{\bar{s}}}). \end{aligned}$$

We compute

$$\begin{aligned}&Z_{{\bar{r}}}(g([Z_r,Z_i],Z_{{\bar{j}}}))=g(D_{{\bar{r}}}[Z_r,Z_i],Z_{{\bar{j}}})+g([Z_r,Z_i],D_{{\bar{r}}}Z_{{\bar{j}}}), D_{{\bar{r}}}[Z_r,Z_i]\\&\qquad =D_{{\bar{r}}}(D_rZ_i-D_iZ_r)\\&\qquad =[Z_{{\bar{r}}},Z_r]Z_i+D_r[Z_{{\bar{r}}},Z_i]-[Z_{{\bar{r}}},Z_i]Z_r-D_i[Z_{{\bar{r}}},Z_r]+[Z_r,Z_i]Z_{{\bar{r}}} \end{aligned}$$

and

$$\begin{aligned}&Z_{{\bar{r}}}(g([Z_r,Z_i],Z_{{\bar{j}}}))\\&\quad = g([Z_{{\bar{r}}},Z_r]Z_i,Z_{{\bar{j}}})+g(D_r[Z_{{\bar{r}}},Z_i],Z_{{\bar{j}}})\nonumber \\&\qquad -g([Z_{{\bar{r}}},Z_i]Z_r,Z_{{\bar{j}}})-g(D_i[Z_{{\bar{r}}},Z_r],Z_{{\bar{j}}})+g([Z_r,Z_i]Z_{{\bar{r}}},Z_{{\bar{j}}})+g([Z_r,Z_i],D_{{\bar{r}}}Z_{{\bar{j}}})\\&\quad =g([Z_{{\bar{r}}},Z_r]Z_i,Z_{{\bar{j}}})+Z_r(g([Z_{{\bar{r}}},Z_i],Z_{{\bar{j}}}))-g([Z_{{\bar{r}}},Z_i],D_rZ_{{\bar{j}}})-g([Z_{{\bar{r}}},Z_i]Z_r,Z_{{\bar{j}}})\\&\qquad -Z_i(g([Z_{{\bar{r}}},Z_r],Z_{{\bar{j}}}))+g([Z_{{\bar{r}}},Z_r],D_iZ_{{\bar{j}}})+g([Z_r,Z_i]Z_{{\bar{r}}},Z_{{\bar{j}}})+g([Z_r,Z_i],D_{{\bar{r}}}Z_{{\bar{j}}})\\&\quad =g([Z_{{\bar{r}}},Z_r]Z_i,Z_{{\bar{j}}})-g([Z_{{\bar{r}}},Z_i],D_rZ_{{\bar{j}}})-g([Z_{{\bar{r}}},Z_i]Z_r,Z_{{\bar{j}}})+g([Z_{{\bar{r}}},Z_r],D_iZ_{{\bar{j}}})\\&\qquad +g([Z_r,Z_i]Z_{{\bar{r}}},Z_{{\bar{j}}})+g([Z_r,Z_i],D_{{\bar{r}}}Z_{{\bar{j}}})+Z_r(B_{{\bar{r}}i}^s)g_{s{\bar{j}}}-Z_i(B_{{\bar{r}}r}^s)g_{s{\bar{j}}}+{\mathcal {O}}(Z(g)), \end{aligned}$$

where we used that

$$\begin{aligned}&Z_r(g([Z_{{\bar{r}}},Z_i],Z_{{\bar{j}}}))-Z_i(g([Z_{{\bar{r}}},Z_r],Z_{{\bar{j}}}))\\&\quad =Z_r(g(B_{{\bar{r}}i}^sZ_s+B_{{\bar{r}}i}^{{\bar{s}}}Z_{{\bar{s}}},Z_{{\bar{j}}}))-Z_i(g(B_{{\bar{r}}r}^sZ_s+B_{{\bar{r}}r}^{{\bar{s}}}Z_{{\bar{s}}},Z_{{\bar{j}}}))\\&\quad =Z_r(B_{{\bar{r}}i}^s)g_{s{\bar{j}}}-Z_i(B_{{\bar{r}}r}^s)g_{s{\bar{j}}}+{\mathcal {O}}(Z(g)). \end{aligned}$$

We compute by using that \([Z_r,Z_i]=B_{ri}^sZ_s+B_{ri}^{{\bar{s}}}Z_{{\bar{s}}}\),

$$\begin{aligned} Z_{{\bar{r}}}(B_{ri}^s)g_{s{\bar{j}}}= & {} g(\nabla _{{\bar{r}}}(B_{ri}^s)Z_s,Z_{{\bar{j}}})\\= & {} g(\nabla _{{\bar{r}}}(B_{ri}^sZ_s)-B_{ri}^s\nabla _{{\bar{r}}}Z_s,Z_{{\bar{j}}})\\= & {} Z_{{\bar{r}}}(g(B_{ri}^sZ_s,Z_{{\bar{j}}}))-g(B_{ri}^sZ_s,\nabla _{{\bar{r}}}Z_{{\bar{j}}})-B_{ri}^sg(\nabla _{{\bar{r}}}Z_s,Z_{{\bar{j}}})\\= & {} Z_{{\bar{r}}}(g([Z_r,Z_i]-B_{ri}^{{\bar{s}}}Z_{{\bar{s}}},Z_{{\bar{j}}}))-g(B_{ri}^sZ_s,\Gamma _{{\bar{r}}{\bar{j}}}^{{\bar{k}}}Z_{{\bar{k}}})-B_{ri}^sg(\Gamma _{{\bar{r}}s}^kZ_k,Z_{{\bar{j}}})\\= & {} g([Z_{{\bar{r}}},Z_r]Z_i,Z_{{\bar{j}}})-g([Z_{{\bar{r}}},Z_i],D_rZ_{{\bar{j}}})-g([Z_{{\bar{r}}},Z_i]Z_r,Z_{{\bar{j}}})\\&+g([Z_{{\bar{r}}},Z_r],D_iZ_{{\bar{j}}})+g([Z_r,Z_i]Z_{{\bar{r}}},Z_{{\bar{j}}})+g([Z_r,Z_i],D_{{\bar{r}}}Z_{{\bar{j}}})\\&+Z_r(B_{{\bar{r}}i}^s)g_{s{\bar{j}}}-Z_i(B_{{\bar{r}}r}^s)g_{s{\bar{j}}}-B_{ri}^s\Gamma _{{\bar{r}}{\bar{j}}}^{{\bar{k}}}g_{s{\bar{k}}}-B_{ri}^s\Gamma _{{\bar{r}}s}^kg_{k{\bar{j}}}+{\mathcal {O}}(Z(g)). \end{aligned}$$

We also compute

$$\begin{aligned} g([Z_{{\bar{r}}},Z_r]Z_i,Z_{{\bar{j}}})= & {} g(B_{{\bar{r}}r}^kD_kZ_i+B_{{\bar{r}}r}^{{\bar{k}}}D_{{\bar{k}}}Z_i,Z_{{\bar{j}}})\\= & {} B_{{\bar{r}}r}^k(g(\nabla _kZ_i,Z_{{\bar{j}}})+\frac{1}{2}(g(T_{ik},Z_{{\bar{j}}})+g(T_{k{\bar{j}}},Z_{i})-g(T_{{\bar{j}}i},Z_k)))\\&+B_{{\bar{r}}r}^{{\bar{k}}}(g(\nabla _{{\bar{k}}}Z_i,Z_{{\bar{j}}})+\frac{1}{2}(g(T_{i{\bar{k}}},Z_{{\bar{j}}})+g(T_{{\bar{k}}{\bar{j}}},Z_{i})-g(T_{{\bar{j}}i},Z_{{\bar{k}}})))\\= & {} B_{{\bar{r}}r}^k\Gamma _{ki}^sg_{s{\bar{j}}}+\frac{1}{2}B_{{\bar{r}}r}^kT_{ik}^sg_{s{\bar{j}}}+B_{{\bar{r}}r}^{{\bar{k}}}\Gamma _{{\bar{k}}i}^sg_{s{\bar{j}}}+\frac{1}{2}B_{{\bar{r}}r}^{{\bar{k}}}\overline{T_{kj}^s}g_{i{\bar{s}}}\\= & {} B^\circ *\Gamma +B^\circ *T'+B^\circ *B^\circ +B^\circ *{\bar{T}}',\\ g([Z_{{\bar{r}}},Z_i],D_rZ_{{\bar{j}}})= & {} g(B_{{\bar{r}}i}^sZ_s+B_{{\bar{r}}i}^{{\bar{s}}}Z_{{\bar{s}}},D_rZ_{{\bar{j}}})\\= & {} B_{{\bar{r}}i}^s(g(\nabla _rZ_{{\bar{j}}},Z_{s})+\frac{1}{2}(g(T_{{\bar{j}}r},Z_{s})+g(T_{rs},Z_{{\bar{j}}})-g(T_{s{\bar{j}}},Z_r)))\\&+B_{{\bar{r}}i}^{{\bar{s}}}(g(\nabla _{r}Z_{{\bar{j}}},Z_{{\bar{s}}})+\frac{1}{2}(g(T_{{\bar{j}}r},Z_{{\bar{s}}})+g(T_{r{\bar{s}}},Z_{{\bar{j}}})-g(T_{{\bar{s}}{\bar{j}}},Z_{r})))\\= & {} B_{{\bar{r}}i}^s\Gamma _{r{\bar{j}}}^{{\bar{k}}}g_{s{\bar{k}}}+\frac{1}{2}B_{{\bar{r}}i}^sT_{rs}^kg_{k{\bar{j}}}+\frac{1}{2}B_{{\bar{r}}i}^{{\bar{s}}}\overline{T_{js}^k}g_{r{\bar{k}}}\\= & {} B^\circ *B^\circ +B^\circ *T'+B^\circ *{\bar{T}}',\\ g([Z_{{\bar{r}}},Z_i]Z_r,Z_{{\bar{j}}})= & {} g(B_{{\bar{r}}i}^sD_sZ_r+B_{{\bar{r}}i}^{{\bar{s}}}D_{{\bar{s}}}Z_r,Z_{{\bar{j}}})\\= & {} B_{{\bar{r}}i}^s(g(\nabla _sZ_r,Z_{{\bar{j}}})+\frac{1}{2}(g(T_{rs},Z_{{\bar{j}}})+g(T_{s{\bar{j}}},Z_{r})-g(T_{{\bar{j}}r},Z_s)))\\&+B_{{\bar{r}}i}^{{\bar{s}}}(g(\nabla _{{\bar{s}}}Z_r,Z_{{\bar{j}}})+\frac{1}{2}(g(T_{r{\bar{s}}},Z_{{\bar{j}}})+g(T_{{\bar{s}}{\bar{j}}},Z_{r})-g(T_{{\bar{j}}r},Z_{{\bar{s}}})))\\= & {} B_{{\bar{r}}i}^s\Gamma _{sr}^kg_{k{\bar{j}}}+\frac{1}{2}B_{{\bar{r}}i}^sT_{rs}^kg_{k{\bar{j}}}+B_{{\bar{r}}i}^{{\bar{s}}}\Gamma _{{\bar{s}}r}^kg_{k{\bar{j}}}+\frac{1}{2}B_{{\bar{r}}i}^{{\bar{s}}}\overline{T_{sj}^k}g_{r{\bar{k}}}\\= & {} B^\circ *\Gamma +B^\circ *T'+B^\circ *B^\circ +B^\circ *{\bar{T}}',\\ g([Z_{{\bar{r}}},Z_r],D_iZ_{{\bar{j}}})= & {} g(B_{{\bar{r}}r}^sZ_s+B_{{\bar{r}}r}^{{\bar{s}}}Z_{{\bar{s}}},D_iZ_{{\bar{j}}})\\= & {} B_{{\bar{r}}r}^s(g(\nabla _iZ_{{\bar{j}}},Z_{s})+\frac{1}{2}(g(T_{{\bar{j}}i},Z_{s})+g(T_{is},Z_{{\bar{j}}})-g(T_{s{\bar{j}}},Z_i)))\\&+B_{{\bar{r}}r}^{{\bar{s}}}(g(\nabla _{i}Z_{{\bar{j}}},Z_{{\bar{s}}})+\frac{1}{2}(g(T_{{\bar{j}}i},Z_{{\bar{s}}})+g(T_{i{\bar{s}}},Z_{{\bar{j}}})-g(T_{{\bar{s}}{\bar{j}}},Z_{i})))\\= & {} B_{{\bar{r}}r}^s\Gamma _{i{\bar{j}}}^{{\bar{l}}}g_{s{\bar{l}}}+\frac{1}{2}B_{{\bar{r}}r}^sT_{is}^lg_{l{\bar{j}}}+\frac{1}{2}B_{{\bar{r}}r}^{{\bar{s}}}\overline{T_{js}^l}g_{i{\bar{l}}}\\= & {} B^\circ *B^\circ +B^\circ *T'+B^\circ *{\bar{T}}',\\ g([Z_r,Z_i]Z_{{\bar{r}}},Z_{{\bar{j}}})= & {} g(B_{ri}^sD_sZ_{{\bar{r}}}+B_{ri}^{{\bar{s}}}D_{{\bar{s}}}Z_{{\bar{r}}},Z_{{\bar{j}}})\\= & {} B_{ri}^s(g(\nabla _sZ_{{\bar{r}}},Z_{{\bar{j}}})+\frac{1}{2}(g(T_{{\bar{r}}s},Z_{{\bar{j}}})+g(T_{s{\bar{j}}},Z_{{\bar{r}}})-g(T_{{\bar{j}}{\bar{r}}},Z_s)))\\&+B_{{\bar{r}}i}^{{\bar{s}}}(g(\nabla _{{\bar{s}}}Z_{{\bar{r}}},Z_{{\bar{j}}})+\frac{1}{2}(g(T_{{\bar{r}}{\bar{s}}},Z_{{\bar{j}}})+g(T_{{\bar{s}}{\bar{j}}},Z_{{\bar{r}}})-g(T_{{\bar{j}}{\bar{r}}},Z_{{\bar{s}}})))\\= & {} \frac{1}{2}B_{ri}^s\overline{T_{rj}^l}g_{s{\bar{l}}}+\frac{1}{2}B_{ri}^{{\bar{s}}}T_{{\bar{r}}{\bar{s}}}^lg_{l{\bar{j}}}+\frac{1}{2}B_{ri}^{{\bar{s}}}T_{{\bar{s}}{\bar{j}}}^lg_{l{\bar{r}}}+\frac{1}{2}B_{ri}^{{\bar{s}}}T_{{\bar{r}}{\bar{j}}}^lg_{l{\bar{s}}}\\= & {} B*{\bar{T}}'+B^\circ *B^\circ ,\\ g([Z_r,Z_i],D_{{\bar{r}}}Z_{{\bar{j}}})= & {} g(B_{ri}^sZ_s+B_{ri}^{{\bar{s}}}Z_{{\bar{s}}},D_{{\bar{r}}}Z_{{\bar{j}}})\\= & {} B_{ri}^s(g(\nabla _{{\bar{r}}}Z_{{\bar{j}}},Z_{s})+\frac{1}{2}(g(T_{{\bar{j}}{\bar{r}}},Z_{s})+g(T_{{\bar{r}}s},Z_{{\bar{j}}})-g(T_{s{\bar{j}}},Z_{{\bar{r}}})))\\&+B_{ri}^{{\bar{s}}}(g(\nabla _{{\bar{r}}}Z_{{\bar{j}}},Z_{{\bar{s}}})+\frac{1}{2}(g(T_{{\bar{j}}{\bar{r}}},Z_{{\bar{s}}})+g(T_{{\bar{r}}{\bar{s}}},Z_{{\bar{j}}})-g(T_{{\bar{s}}{\bar{j}}},Z_{{\bar{r}}})))\\= & {} B_{ri}^s\Gamma _{{\bar{r}}{\bar{j}}}^{{\bar{l}}}g_{s{\bar{l}}}+\frac{1}{2}B_{ri}^s\overline{T_{jr}^l}g_{s{\bar{l}}}+\frac{1}{2}B_{ri}^{{\bar{s}}}T_{{\bar{j}}{\bar{r}}}^lg_{l{\bar{s}}}+\frac{1}{2}B_{ri}^{{\bar{s}}}T_{{\bar{r}}{\bar{s}}}^lg_{l{\bar{j}}}+\frac{1}{2}B_{ri}^{{\bar{s}}}T_{{\bar{j}}{\bar{s}}}^lg_{l{\bar{r}}}\\= & {} B*{\bar{\Gamma }}+B*{\bar{T}}'+B^\circ *B^\circ , \end{aligned}$$

where we used that \(g_{{\bar{l}}{\bar{s}}}=0\), \(\Gamma _{r{\bar{j}}}^{{\bar{k}}}=B_{r{\bar{j}}}^{{\bar{k}}}\), \(\Gamma _{{\bar{k}}i}^s=\overline{\Gamma _{k{\bar{i}}}^{{\bar{s}}}}=\overline{B_{k{\bar{i}}}^{{\bar{s}}}}=B_{{\bar{k}}i}^s\), \(T_{{\bar{r}}{\bar{s}}}^l=\overline{T_{rs}^{{\bar{l}}}}=-\overline{B_{rs}^{{\bar{l}}}}=-B_{{\bar{r}}{\bar{s}}}^l\). Using these computations, we obtain

$$\begin{aligned}&-Z_{{\bar{r}}}(T_{ri}^s)g_{s{\bar{j}}}=-Z_{{\bar{r}}}(\Gamma _{ri}^s-\Gamma _{ir}^s-B_{ri}^s)g_{s{\bar{j}}}\\&\quad =-Z_{{\bar{r}}}(\Gamma _{ri}^s)g_{s{\bar{j}}}+Z_{{\bar{r}}}(\Gamma _{ir}^s)g_{s{\bar{j}}}+Z_{{\bar{r}}}(B_{ri}^s)g_{s{\bar{j}}}\\&\quad =-Z_{{\bar{r}}}(\Gamma _{ri}^s)g_{s{\bar{j}}}+Z_{{\bar{r}}}(\Gamma _{ir}^s)g_{s{\bar{j}}}+g([Z_{{\bar{r}}},Z_r]Z_i,Z_{{\bar{j}}})\\&\qquad -g([Z_{{\bar{r}}},Z_i],D_rZ_{{\bar{j}}})\\&\qquad -g([Z_{{\bar{r}}},Z_i]Z_r,Z_{{\bar{j}}})+g([Z_{{\bar{r}}},Z_r],D_iZ_{{\bar{j}}})\\&\qquad +g([Z_r,Z_i]Z_{{\bar{r}}},Z_{{\bar{j}}})+g([Z_r,Z_i],D_{{\bar{r}}}Z_{{\bar{j}}})\\&\qquad +Z_r(B_{{\bar{r}}i}^s)g_{s{\bar{j}}}-Z_i(B_{{\bar{r}}r}^s)g_{s{\bar{j}}}-B_{ri}^s\Gamma _{{\bar{r}}{\bar{j}}}^{{\bar{k}}}g_{s{\bar{k}}}\\&\qquad -B_{ri}^s\Gamma _{{\bar{r}}s}^kg_{k{\bar{j}}}+{\mathcal {O}}(Z(g))\\&\quad =-Z_{{\bar{r}}}(\Gamma _{ri}^s)g_{s{\bar{j}}}+Z_{{\bar{r}}}(\Gamma _{ir}^s)g_{s{\bar{j}}}+Z_r(B_{{\bar{r}}i}^s)g_{s{\bar{j}}}\\&\qquad -Z_i(B_{{\bar{r}}r}^s)g_{s{\bar{j}}}-B_{ri}^s\Gamma _{{\bar{r}}{\bar{j}}}^{{\bar{k}}}g_{s{\bar{k}}}-B_{ri}^s\Gamma _{{\bar{r}}s}^kg_{k{\bar{j}}}\\&\qquad +B^\circ *\Gamma +B*{\bar{\Gamma }}+B^\circ *T'+B^\circ *{\bar{T}}'+B*{\bar{T}}'+B^\circ *B^\circ +{\mathcal {O}}(Z(g)). \end{aligned}$$

Similarly, we have

$$\begin{aligned} -Z_{{\bar{j}}}(w_i)= & {} -Z_{{\bar{j}}}(T_{ir{\bar{r}}})\\= & {} -Z_{{\bar{j}}}(T_{ir}^sg_{s{\bar{r}}})\\= & {} -Z_{{\bar{j}}}(T_{ir}^s)g_{s{\bar{r}}}-T_{ir}^sZ_{{\bar{j}}}(g_{s{\bar{r}}})\\= & {} -Z_{{\bar{j}}}(\Gamma _{ir}^s)g_{s{\bar{r}}}+Z_{{\bar{j}}}(\Gamma _{ri}^s)g_{s{\bar{r}}}\\&+Z_{{\bar{j}}}(B_{ir}^s)g_{s{\bar{r}}}-T_{ir}^sZ_{{\bar{j}}}(g_{s{\bar{r}}})\\= & {} -Z_{{\bar{j}}}(\Gamma _{ir}^s)g_{s{\bar{r}}}+Z_{{\bar{j}}}(\Gamma _{ri}^s)g_{s{\bar{r}}}\\&+g([Z_{{\bar{j}}},Z_i]Z_r,Z_{{\bar{r}}})+g(D_i[Z_{{\bar{j}}},Z_r],Z_{{\bar{r}}})\\&-g([Z_{{\bar{j}}},Z_r]Z_i,Z_{{\bar{r}}})-g(D_r[Z_{{\bar{j}}},Z_i],Z_{{\bar{r}}})+g([Z_i,Z_r]Z_{{\bar{j}}},Z_{{\bar{r}}})\\&+g([Z_i,Z_r],D_{{\bar{j}}}Z_{{\bar{r}}})-B_{ir}^s\overline{\Gamma _{jr}^k}g_{s{\bar{k}}}-B_{ir}^sB_{{\bar{j}}s}^kg_{k{\bar{r}}}-T_{ir}^sZ_{{\bar{j}}}(g_{s{\bar{r}}})\\= & {} -Z_{{\bar{j}}}(\Gamma _{ir}^s)g_{s{\bar{r}}}+Z_{{\bar{j}}}(\Gamma _{ri}^s)g_{s{\bar{r}}}\\&-B_{ir}^s\overline{\Gamma _{jr}^k}g_{s{\bar{k}}}-B_{ir}^s\Gamma _{{\bar{j}}s}^kg_{k{\bar{r}}}-T_{ir}^sZ_{{\bar{j}}}(g_{s{\bar{r}}})\\&+g([Z_{{\bar{j}}},Z_i]Z_r,Z_{{\bar{r}}})+Z_i(B_{{\bar{j}}r}^s)g_{s{\bar{r}}}+B_{{\bar{j}}r}^sZ_i(g_{s{\bar{r}}})-g([Z_{{\bar{j}}},Z_r],D_iZ_{{\bar{r}}})\\&-g([Z_{{\bar{j}}},Z_r]Z_i,Z_{{\bar{r}}})\\&-Z_r(B_{{\bar{j}}i}^s)g_{s{\bar{r}}}-B_{{\bar{j}}i}^sZ_r(g_{s{\bar{r}}})+g([Z_{{\bar{j}}},Z_i],D_rZ_{{\bar{r}}})+g([Z_i,Z_r]Z_{{\bar{j}}},Z_{{\bar{r}}})\\&+g([Z_i,Z_r],D_{{\bar{j}}}Z_{{\bar{r}}})\\= & {} -Z_{{\bar{j}}}(\Gamma _{ir}^s)g_{s{\bar{r}}}+Z_{{\bar{j}}}(\Gamma _{ri}^s)g_{s{\bar{r}}}\\&+Z_i(B_{{\bar{j}}r}^s)g_{s{\bar{r}}}-Z_r(B_{{\bar{j}}i}^s)g_{s{\bar{r}}}\\&-B_{ir}^s\overline{\Gamma _{jr}^k}g_{s{\bar{k}}}-B_{ir}^s\Gamma _{{\bar{j}}s}^kg_{k{\bar{r}}}\\&+B^\circ *\Gamma +B*{\bar{\Gamma }}+B^\circ *T'+B^\circ *{\bar{T}}'+B*{\bar{T}}'\\&+B^\circ *B^\circ +{\mathcal {O}}(Z(g))+{\mathcal {O}}({\bar{Z}}(g)). \end{aligned}$$

Combining these with the term \(-g^{p{\bar{q}}}T_{pi}^rZ_{{\bar{j}}}(g_{r{\bar{q}}})=T_{ir}^sZ_{{\bar{j}}}(g_{s{\bar{r}}})\), we have

$$\begin{aligned} {\bar{Z}}(T')_{i{\bar{j}}}= & {} -Z_{{\bar{r}}}(T_{ri}^s)g_{s{\bar{j}}}-Z_{{\bar{j}}}(w_i)-g^{p{\bar{q}}}T_{pi}^rZ_{{\bar{j}}}(g_{r{\bar{q}}})\\= & {} -Z_{{\bar{r}}}(\Gamma _{ri}^s)g_{s{\bar{j}}}+Z_{{\bar{r}}}(\Gamma _{ir}^s)g_{s{\bar{j}}}+Z_r(B_{{\bar{r}}i}^s)g_{s{\bar{j}}}\\&-Z_i(B_{{\bar{r}}r}^s)g_{s{\bar{j}}}-B_{ri}^s\Gamma _{{\bar{r}}{\bar{j}}}^{{\bar{k}}}g_{s{\bar{k}}}-B_{ri}^s\Gamma _{{\bar{r}}s}^kg_{k{\bar{j}}}\\&-Z_{{\bar{j}}}(\Gamma _{ir}^s)g_{s{\bar{r}}}+Z_{{\bar{j}}}(\Gamma _{ri}^s)g_{s{\bar{r}}}+Z_i(B_{{\bar{j}}r}^s)g_{s{\bar{r}}}\\&-Z_r(B_{{\bar{j}}i}^s)g_{s{\bar{r}}} -B_{ir}^s\Gamma _{{\bar{j}}{\bar{r}}}^{{\bar{k}}}g_{s{\bar{k}}}-B_{ir}^s\Gamma _{{\bar{j}}s}^kg_{k{\bar{r}}}\\&+B^\circ *\Gamma +B*{\bar{\Gamma }}+B^\circ *T'+B^\circ *{\bar{T}}'\\&+B*{\bar{T}}'+B^\circ *B^\circ +{\mathcal {O}}(Z(g))+{\mathcal {O}}({\bar{Z}}(g))\\= & {} {\bar{Z}}(\Gamma )+Z(B^\circ )+B*B^\circ +B*{\bar{\Gamma }} \\&+B^\circ *\Gamma +B*{\bar{\Gamma }}+B^\circ *T'+B^\circ *{\bar{T}}'+B*{\bar{T}}'\\&+B^\circ *B^\circ +{\mathcal {O}}(Z(g))+{\mathcal {O}}({\bar{Z}}(g))\\= & {} {\bar{Z}}(\Gamma )+Z(B^\circ )+B^\circ *\Gamma +T'*{\bar{\Gamma }}+\Gamma *{\bar{\Gamma }}\\&+B^\circ *T'+B^\circ *{\bar{T}}'+T'*{\bar{T}}'+\Gamma *{\bar{T}}'\\&+B^\circ *B^\circ +{\mathcal {O}}(Z(g))+{\mathcal {O}}({\bar{Z}}(g)), \end{aligned}$$

where we used that \(B=T'+\Gamma \) and then \(B*B^\circ =T'*B^\circ +\Gamma *B^\circ \), \(B*{\bar{\Gamma }}=T'*{\bar{\Gamma }}+\Gamma *{\bar{\Gamma }}\) \(B*{\bar{T}}'=T'*{\bar{T}}'+\Gamma *{\bar{T}}'\) at the third equality. \(\square \)

Lemma 3.4

Let (Mg(t), J) be a solution to the AHCF for \(t\in [0,\tau )\) starting at the initial almost Hermitian metric \(\omega _0\). Suppose that there exist uniform bounds for \(|T'(g(t))|^2_{C^0(g(t))}\) and \(|\nabla T'(g(t))|_{C^0(g(t))}\). Then we have that for any fixed time \(t_0\in [0,\tau )\),

$$\begin{aligned}&|{\bar{\nabla }}\nabla T'(g(t_0))|_{C^0(g(t_0))}\le C,\quad |{\bar{\nabla }}\nabla {\bar{T}}'(g(t_0))|_{C^0(g(t_0))}\le C, \\&|{\bar{\nabla }}^2 T'(g(t_0))|_{C^0(g(t_0))}\le C,\quad |\nabla ^2 {\bar{T}}'(g(t_0))|_{C^0(g(t_0))}\le C \end{aligned}$$

for some uniform constant \(C>0\).

Proof

We compute

$$\begin{aligned} \nabla _{{\bar{i}}}\nabla _m(T_{jk{\bar{r}}})= & {} \nabla _{{\bar{i}}}\nabla _m(T_{jk}^l)g_{l{\bar{r}}}\\= & {} \nabla _{{\bar{i}}}(Z_m(T_{jk}^l)-\Gamma _{mj}^sT_{sk}^l-\Gamma _{mk}^sT_{js}^l +\Gamma _{ms}^lT_{jk}^s)g_{l{\bar{r}}}\\= & {} \nabla _{{\bar{i}}}(Z_m(T_{jk}^l))+{\bar{\nabla }}\Gamma *T'+\Gamma *{\bar{\nabla }}T' \end{aligned}$$

and

$$\begin{aligned} \nabla _{{\bar{i}}}(Z_m(T_{jk}^l))= & {} Z_{{\bar{i}}}Z_m(T_{jk}^l)-\Gamma _{{\bar{i}}m}^sZ_s(T_{jk}^l)-\Gamma _{{\bar{i}}j}^sZ_m(T_{sk}^l)-\Gamma _{{\bar{i}}k}^sZ_m(T_{js}^l)+\Gamma _{{\bar{i}}s}^lZ_m(T_{jk}^s)\\= & {} Z_mZ_{{\bar{i}}}(T_{jk}^l)+[Z_{{\bar{i}}},Z_m](T_{jk}^l)\\&-\Gamma _{{\bar{i}}m}^sZ_s(T_{jk}^l)-\Gamma _{{\bar{i}}j}^sZ_m(T_{sk}^l)-\Gamma _{{\bar{i}}k}^sZ_m(T_{js}^l)+\Gamma _{{\bar{i}}s}^lZ_m(T_{jk}^s).\\= & {} Z_mZ_{{\bar{i}}}(T_{jk}^l)+B^\circ *Z(T')+B^\circ *{\bar{Z}}(T') \end{aligned}$$

Using that \([Z_r,Z_i]=B_{ri}^sZ_s+B_{ri}^{{\bar{s}}}Z_{{\bar{s}}}\), we compute

$$\begin{aligned} Z_{{\bar{i}}}(B_{jk}^l)g_{l{\bar{r}}}= & {} g(\nabla _{{\bar{i}}}(B_{jk}^l)Z_l,Z_{{\bar{r}}})c\\= & {} g(\nabla _{{\bar{i}}}(B_{jk}^lZ_l)-B_{jk}^l\nabla _{{\bar{i}}}Z_l,Z_{{\bar{r}}})\\= & {} Z_{{\bar{i}}}(g(B_{jk}^lZ_l,Z_{{\bar{r}}}))-g(B_{jk}^lZ_l,\nabla _{{\bar{i}}}Z_{{\bar{r}}})-B_{jk}^lg(\nabla _{{\bar{i}}}Z_l,Z_{{\bar{r}}})\\= & {} Z_{{\bar{i}}}(g([Z_j,Z_k]-B_{jk}^{{\bar{l}}}Z_{{\bar{l}}},Z_{{\bar{r}}}))-g(B_{jk}^lZ_l,\Gamma _{{\bar{i}}{\bar{r}}}^{{\bar{s}}}Z_{{\bar{s}}})-B_{jk}^lg(\Gamma _{{\bar{i}}l}^sZ_s,Z_{{\bar{r}}})\\= & {} g(D_{{\bar{i}}}[Z_j,Z_k],Z_{{\bar{r}}})+g([Z_j,Z_k],D_{{\bar{i}}}Z_{{\bar{r}}})-B_{jk}^l\Gamma _{{\bar{i}}{\bar{r}}}^{{\bar{s}}}g(Z_l,Z_{{\bar{s}}})-B_{jk}^l\Gamma _{{\bar{i}}l}^sg(Z_s,Z_{{\bar{r}}})\\= & {} g([Z_{{\bar{i}}},Z_j]Z_k,Z_{{\bar{r}}})+g([Z_k,Z_{{\bar{i}}}]Z_j,Z_{{\bar{r}}})+g(D_j[Z_{{\bar{i}}},Z_k],Z_{{\bar{r}}})\\&+g(D_k[Z_j,Z_{{\bar{i}}}],Z_{{\bar{r}}})+g([Z_j,Z_k]Z_{{\bar{i}}},Z_{{\bar{r}}})+g([Z_j,Z_k],D_{{\bar{i}}}Z_{{\bar{r}}})\\&-B_{jk}^l\Gamma _{{\bar{i}}{\bar{r}}}^{{\bar{s}}}g_{l{\bar{s}}}-B_{jk}^l\Gamma _{{\bar{i}}l}^sg_{s{\bar{r}}}\\= & {} g([Z_{{\bar{i}}},Z_j]Z_k,Z_{{\bar{r}}})+g([Z_k,Z_{{\bar{i}}}]Z_j,Z_{{\bar{r}}})+Z_j(g([Z_{{\bar{i}}},Z_k],Z_{{\bar{r}}}))\\&-g([Z_{{\bar{i}}},Z_k],D_jZ_{{\bar{r}}})\\&+Z_k(g([Z_j,Z_{{\bar{i}}}],Z_{{\bar{r}}}))-g([Z_j,Z_{{\bar{i}}}],D_kZ_{{\bar{r}}})+g([Z_j,Z_k]Z_{{\bar{i}}},Z_{{\bar{r}}})\\&+g([Z_j,Z_k],D_{{\bar{i}}}Z_{{\bar{r}}})\\&-B_{jk}^l\Gamma _{{\bar{i}}{\bar{r}}}^{{\bar{s}}}g_{l{\bar{s}}}-B_{jk}^l\Gamma _{{\bar{i}}l}^sg_{s{\bar{r}}}, \end{aligned}$$

where we used that

$$\begin{aligned} D_{{\bar{i}}}[Z_j,Z_k]=[Z_{{\bar{i}}},Z_j]Z_k+[Z_k,Z_{{\bar{i}}}]Z_j+D_j[Z_{{\bar{i}}},Z_k]+D_k[Z_j,Z_{{\bar{i}}}]+[Z_j,Z_k]Z_{{\bar{i}}}. \end{aligned}$$

We also compute

$$\begin{aligned} g([Z_{{\bar{i}}},Z_j]Z_k,Z_{{\bar{r}}})= & {} g(B_{{\bar{i}}j}^sD_sZ_k+B_{{\bar{i}}j}^{{\bar{s}}}D_{{\bar{s}}}Z_k,Z_{{\bar{r}}})\\= & {} B_{{\bar{i}}j}^s(g(\nabla _sZ_k,Z_{{\bar{r}}})+\frac{1}{2}(g(T_{ks},Z_{{\bar{r}}})+g(T_{s{\bar{r}}},Z_k)-g(T_{{\bar{r}}k},Z_s)))\\&+B_{{\bar{i}}j}^{{\bar{s}}}(g(\nabla _{{\bar{s}}}Z_k,Z_{{\bar{r}}})+\frac{1}{2}(g(T_{k{\bar{s}}},Z_{{\bar{r}}})+g(T_{{\bar{s}}{\bar{r}}},Z_k)-g(T_{{\bar{r}}k},Z_{{\bar{s}}})))\\= & {} B_{{\bar{i}}j}^s\Gamma _{sk}^lg_{l{\bar{r}}}+\frac{1}{2}B_{{\bar{i}}j}^sT_{ks}^lg_{l{\bar{r}}}+B_{{\bar{i}}j}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^lg_{l{\bar{r}}}+\frac{1}{2}B_{{\bar{i}}j}^{{\bar{s}}}\overline{T_{sr}^l}g_{k{\bar{l}}}\\= & {} B^\circ *\Gamma +B^\circ *T'+B^\circ *B^\circ +B^\circ *{\bar{T}}', \\ g([Z_k,Z_{{\bar{i}}}]Z_j,Z_{{\bar{r}}})= & {} g(B_{k{\bar{i}}}^sD_sZ_j+B_{k{\bar{i}}}^{{\bar{s}}}D_{{\bar{s}}}Z_j,Z_{{\bar{r}}})\\= & {} B_{k{\bar{i}}}^s(g(\nabla _sZ_j,Z_{{\bar{r}}})+\frac{1}{2}(g(T_{js},Z_{{\bar{r}}})+g(T_{s{\bar{r}}},Z_j)-g(T_{{\bar{r}}j},Z_s)))\\&+B_{k{\bar{i}}}^{{\bar{s}}}(g(\nabla _{{\bar{s}}}Z_j,Z_{{\bar{r}}})+\frac{1}{2}(g(T_{j{\bar{s}}},Z_{{\bar{r}}})+g(T_{{\bar{s}}{\bar{r}}},Z_j)-g(T_{{\bar{r}}j},Z_{{\bar{s}}})))\\= & {} B_{k{\bar{i}}}^s\Gamma _{sj}^lg_{l{\bar{r}}}+\frac{1}{2}B_{k{\bar{i}}}^sT_{js}^lg_{l{\bar{r}}}+B_{k{\bar{i}}}^{{\bar{s}}}\Gamma _{{\bar{s}}j}^lg_{l{\bar{r}}}+\frac{1}{2}B_{k{\bar{i}}}^{{\bar{s}}}\overline{T_{sr}^l}g_{j{\bar{l}}}\\= & {} B^\circ *\Gamma +B^\circ *T'+B^\circ *B^\circ +B^\circ *{\bar{T}}', \\ g([Z_{{\bar{i}}},Z_k],D_jZ_{{\bar{r}}})= & {} g(B_{{\bar{i}}k}^sZ_s+B_{{\bar{i}}k}^{{\bar{s}}}Z_{{\bar{s}}},D_jZ_{{\bar{r}}})\\= & {} B_{{\bar{i}}k}^s(g(\nabla _jZ_{{\bar{r}}},Z_s)+\frac{1}{2}(g(T_{{\bar{r}}j},Z_s)+g(T_{js},Z_{{\bar{r}}})-g(T_{s{\bar{r}}},Z_j)))\\&+B_{{\bar{i}}k}^{{\bar{s}}}(g(\nabla _jZ_{{\bar{r}}},Z_{{\bar{s}}})+\frac{1}{2}(g(T_{{\bar{r}}j},Z_{{\bar{s}}})+g(T_{j{\bar{s}}},Z_{{\bar{r}}})-g(T_{{\bar{s}}{\bar{r}}},Z_j)))\\= & {} B_{{\bar{i}}k}^s\Gamma _{j{\bar{r}}}^{{\bar{l}}}g_{s{\bar{l}}} +\frac{1}{2}B_{{\bar{i}}k}^sT_{js}^lg_{l{\bar{r}}} +\frac{1}{2}B_{{\bar{i}}k}^{{\bar{s}}}\overline{T_{rs}^l}g_{j{\bar{l}}}\\= & {} B^\circ *B^\circ +B^\circ *T'+B^\circ *{\bar{T}}',\\ g([Z_j,Z_{{\bar{i}}}],D_kZ_{{\bar{r}}})= & {} g(B_{j{\bar{i}}}^sZ_s+B_{j{\bar{i}}}^{{\bar{s}}}Z_{{\bar{s}}},D_kZ_{{\bar{r}}})\\= & {} B_{j{\bar{i}}}^s(g(\nabla _kZ_{{\bar{r}}},Z_s)+\frac{1}{2}(g(T_{{\bar{r}}k},Z_s)+g(T_{ks},Z_{{\bar{r}}})-g(T_{s{\bar{r}}},Z_k)))\\&+B_{j{\bar{i}}}^{{\bar{s}}}(g(\nabla _kZ_{{\bar{r}}},Z_{{\bar{s}}})+\frac{1}{2}(g(T_{{\bar{r}}k},Z_{{\bar{s}}})+g(T_{k{\bar{s}}},Z_{{\bar{r}}})-g(T_{{\bar{s}}{\bar{r}}},Z_k)))\\= & {} B_{j{\bar{i}}}^s\Gamma _{k{\bar{r}}}^{{\bar{l}}}g_{s{\bar{l}}} +\frac{1}{2}B_{j{\bar{i}}}^sT_{ks}^lg_{l{\bar{r}}} +\frac{1}{2}B_{j{\bar{i}}}^{{\bar{s}}}\overline{T_{rs}^l}g_{k{\bar{l}}}\\= & {} B^\circ *B^\circ +B^\circ *T'+B^\circ *{\bar{T}}', \\ g([Z_j,Z_k]Z_{{\bar{i}}},Z_{{\bar{r}}})= & {} g(B_{jk}^sD_sZ_{{\bar{i}}}+B_{jk}^{{\bar{s}}}D_{{\bar{s}}}Z_{{\bar{i}}},Z_{{\bar{r}}})\\= & {} B_{jk}^s(g(\nabla _sZ_{{\bar{i}}},Z_{{\bar{r}}})+\frac{1}{2}(g(T_{{\bar{i}}s},Z_{{\bar{r}}})+g(T_{s{\bar{r}}},Z_{{\bar{i}}})-g(T_{{\bar{r}}{\bar{i}}},Z_s)))\\&+B_{jk}^{{\bar{s}}}(g(\nabla _{{\bar{s}}}Z_{{\bar{i}}},Z_{{\bar{r}}})+\frac{1}{2}(g(T_{{\bar{i}}{\bar{s}}},Z_{{\bar{r}}})+g(T_{{\bar{s}}{\bar{r}}},Z_{{\bar{i}}})-g(T_{{\bar{r}}{\bar{i}}},Z_{{\bar{s}}})))\\= & {} \frac{1}{2}B_{jk}^s\overline{T_{ir}^l}g_{s{\bar{l}}} +\frac{1}{2}B_{jk}^{{\bar{s}}}T_{{\bar{i}}{\bar{s}}}^lg_{l{\bar{r}}}+\frac{1}{2}B_{jk}^{{\bar{s}}}T_{{\bar{s}}{\bar{r}}}^lg_{l{\bar{i}}}+\frac{1}{2}B_{jk}^{{\bar{s}}}T_{{\bar{i}}{\bar{r}}}^lg_{l{\bar{s}}}\\= & {} B*{\bar{T}}'+B^\circ *B^\circ , \\ g([Z_j,Z_k],D_{{\bar{i}}}Z_{{\bar{r}}})= & {} g(B_{jk}^sZ_s+B_{jk}^{{\bar{s}}}Z_{{\bar{s}}},D_{{\bar{i}}}Z_{{\bar{r}}})\\= & {} B_{jk}^s(g(\nabla _{{\bar{i}}}Z_{{\bar{r}}},Z_s)+\frac{1}{2}(g(T_{{\bar{r}}{\bar{i}}},Z_s)+g(T_{{\bar{i}}s},Z_{{\bar{r}}})-g(T_{s{\bar{r}}},Z_{{\bar{i}}})))\\&+B_{jk}^{{\bar{s}}}(g(\nabla _{{\bar{i}}}Z_{{\bar{r}}},Z_{{\bar{s}}})+\frac{1}{2}(g(T_{{\bar{r}}{\bar{i}}},Z_{{\bar{s}}})+g(T_{{\bar{i}}{\bar{s}}},Z_{{\bar{r}}})-g(T_{{\bar{s}}{\bar{r}}},Z_{{\bar{i}}})))\\= & {} B_{jk}^s\overline{\Gamma _{ir}^l}g_{s{\bar{l}}}+\frac{1}{2}B_{jk}^s\overline{T_{ri}^l}g_{s{\bar{l}}} +\frac{1}{2}B_{jk}^{{\bar{s}}}T_{{\bar{r}}{\bar{i}}}^lg_{l{\bar{s}}} +\frac{1}{2}B_{jk}^{{\bar{s}}}T_{{\bar{i}}{\bar{s}}}^lg_{l{\bar{r}}} +\frac{1}{2}B_{jk}^{{\bar{s}}}T_{{\bar{r}}{\bar{s}}}^lg_{l{\bar{i}}}\\= & {} B*{\bar{\Gamma }}+B*{\bar{T}}'+B^\circ *B^\circ . \end{aligned}$$

And we compute that

$$\begin{aligned}&Z_j(g([Z_{{\bar{i}}},Z_k],Z_{{\bar{r}}}]))=Z_j(B_{{\bar{i}}k}^lg_{l{\bar{r}}})=Z_j(B_{{\bar{i}}k}^l)g_{l{\bar{r}}} +{\mathcal {O}}(Z(g)), \\&Z_k(g([Z_j,Z_{{\bar{i}}}],Z_{{\bar{r}}}]))=Z_k(B_{j{\bar{i}}}^lg_{l{\bar{r}}})=Z_k(B_{j{\bar{i}}}^l)g_{l{\bar{r}}} +{\mathcal {O}}(Z(g)). \end{aligned}$$

Therefore we have

$$\begin{aligned} Z_{{\bar{i}}}(B_{jk}^l)g_{l{\bar{r}}}= & {} B^\circ *\Gamma +B^\circ *T' +B^\circ *B^\circ +B^\circ *{\bar{T}}'+B*{\bar{T}}'+B*{\bar{\Gamma }}\\&+Z_j(B_{{\bar{i}}k}^sg_{s{\bar{r}}})+Z_k(B_{j{\bar{i}}}^sg_{s{\bar{r}}}) -B_{jk}^l\Gamma _{{\bar{i}}{\bar{r}}}^{{\bar{s}}}g_{l{\bar{s}}}-B_{jk}^l\Gamma _{{\bar{i}}l}^{s}g_{s{\bar{r}}}. \end{aligned}$$

Using these computations, we obtain

$$\begin{aligned}&Z_mZ_{{\bar{i}}}(T_{jk}^l)g_{l{\bar{r}}}=Z_m(Z_{{\bar{i}}}(\Gamma _{jk}^l)-Z_{{\bar{i}}}(\Gamma _{kj}^l)-Z_{{\bar{i}}}(B_{jk}^l))g_{l{\bar{r}}}\\&\quad =Z_mZ_{{\bar{i}}}(\Gamma _{jk}^l)g_{l{\bar{r}}}-Z_mZ_{{\bar{i}}}(\Gamma _{kj}^l)g_{l{\bar{r}}}-Z_m(Z_{{\bar{i}}}(B_{jk}^l)g_{l{\bar{r}}})+Z_{{\bar{i}}}(B_{jk}^l)Z_m(g_{l{\bar{r}}})\\&\quad =Z_mZ_{{\bar{i}}}(\Gamma _{jk}^l)g_{l{\bar{r}}}-Z_mZ_{{\bar{i}}}(\Gamma _{kj}^l)g_{l{\bar{r}}}\\&\qquad -Z_m(B^\circ *\Gamma +B^\circ *T'+B^\circ *B^\circ +B^\circ *{\bar{T}}' +B*{\bar{T}}'+B*{\bar{\Gamma }})\\&\qquad -Z_mZ_j(B_{{\bar{i}}k}^l)g_{l{\bar{r}}}-Z_mZ_k(B_{j{\bar{i}}}^l)g_{l{\bar{r}}}+Z_m(B_{jk}^l\Gamma _{{\bar{i}}{\bar{r}}}^{{\bar{s}}})g_{l{\bar{s}}}+Z_m(B_{jk}^l\Gamma _{{\bar{i}}l}^{s})g_{s{\bar{r}}}+{\mathcal {O}}(Z(g)) \\&\quad =Z{\bar{Z}}(\Gamma )+ZZ(B^\circ )+Z(B^\circ *\Gamma +B^\circ *T'+B^\circ *B^\circ +B^\circ *{\bar{T}}' +T'*{\bar{\Gamma }}\\&\qquad +\Gamma *{\bar{\Gamma }}+\Gamma *{\bar{T}}'+T'*{\bar{T}}')+{\mathcal {O}}(Z(g)), \end{aligned}$$

where we used that \(B=T'+\Gamma \). Combining these computations, we obtain

$$\begin{aligned} \nabla _{{\bar{i}}}\nabla _m(T_{jk{\bar{r}}})= & {} Z{\bar{Z}}(\Gamma )+ZZ(B^\circ )+Z(B^\circ *\Gamma +B^\circ *T'+B^\circ *B^\circ \\&+B^\circ *{\bar{T}}' +T'*{\bar{\Gamma }}+\Gamma *{\bar{\Gamma }}+\Gamma *{\bar{T}}'+T'*{\bar{T}}')+{\mathcal {O}}(Z(g))\\&+B^\circ *Z(T')+B^\circ *{\bar{Z}}(T')+{\bar{\nabla }}\Gamma *T'+\Gamma *{\bar{\nabla }}T'. \end{aligned}$$

Now fix an arbitrary chosen time \(t_0\in [0,\tau )\) and using a local \(g(t_0)\)-unitary frame, we have that at \(t_0\), \(|Z_iZ_{{\bar{i}}}(\Gamma _{jk}^l)|\le C\) as in Lemma 3.1, and we also have that \({\mathcal {O}}(Z(g))=0\). By using Lemma 3.3, then we obtain that \(|{\bar{\nabla }}\nabla T'(g(t_0))|_{C^0(g(t_0))}\le C\) for some uniform constant \(C>0\) under our assumptions that \(|T'(g(t))|^2_{C^0(g(t))}\) and \(|\nabla T'(g(t))|_{C^0(g(t))}\) have uniform bounds. Similarly, one can obtain the rest of uniform bounds, since we have

$$\begin{aligned} \nabla _{{\bar{i}}}\nabla _{{\bar{m}}}(T_{jk{\bar{r}}})= & {} Z_{{\bar{i}}}Z_{{\bar{m}}}(T_{jk}^l)g_{l{\bar{r}}}-\Gamma _{{\bar{i}}{\bar{m}}}^{{\bar{s}}}Z_{{\bar{s}}}(T_{jk}^l)-\Gamma _{{\bar{i}}j}^sZ_{{\bar{m}}}(T_{sk}^l)-\Gamma _{{\bar{i}}k}^sZ_{{\bar{m}}}(T_{js}^l)+\Gamma _{{\bar{i}}s}^lZ_{{\bar{m}}}(T_{jk}^s)\\&-\nabla _{{\bar{i}}}(B_{{\bar{m}}j}^sT_{sk}^l+B_{{\bar{m}}k}^sT_{js}^l-B_{{\bar{m}}s}^lT_{jk}^s)g_{l{\bar{r}}}\\= & {} Z_{{\bar{i}}}Z_{{\bar{m}}}(\Gamma _{jk}^l)g_{l{\bar{r}}}-Z_{{\bar{i}}}Z_{{\bar{m}}}(\Gamma _{kj}^l)g_{l{\bar{r}}}-Z_{{\bar{i}}}(Z_{{\bar{m}}}(B_{jk}^l)g_{l{\bar{r}}})+Z_{{\bar{m}}}(B_{jk}^l)Z_{{\bar{i}}}(g_{l{\bar{r}}})\\&+{\bar{\Gamma }}*{\bar{Z}}(T')+B^\circ *{\bar{Z}}(T')+{\bar{\nabla }}B^\circ *T'+B^\circ *{\bar{\nabla }}T'\\= & {} {\bar{Z}}{\bar{Z}}(\Gamma )-Z_{{\bar{i}}}(B^\circ *\Gamma +B^\circ *T'+B^\circ *B^\circ +B^\circ *{\bar{T}}' +B*{\bar{T}}'+B*{\bar{\Gamma }}) \\&-Z_{{\bar{i}}}Z_j(B_{{\bar{m}}k}^lg_{l{\bar{r}}})-Z_{{\bar{i}}}Z_k(B_{j{\bar{m}}}^lg_{l{\bar{r}}})+Z_{{\bar{i}}}(B_{jk}^l\Gamma _{{\bar{m}}{\bar{r}}}^{{\bar{s}}}g_{l{\bar{s}}}+B_{jk}^l\Gamma _{{\bar{m}}l}^sg_{s{\bar{r}}}) \\&+Z_{{\bar{m}}}(B_{jk}^l)Z_{{\bar{i}}}(g_{l{\bar{r}}})+{\bar{\Gamma }}*{\bar{Z}}(T')+B^\circ *{\bar{Z}}(T')+{\bar{\nabla }}B^\circ *T'+B^\circ *{\bar{\nabla }}T'\\= & {} {\bar{Z}}{\bar{Z}}(\Gamma )+{\bar{Z}}(B^\circ *\Gamma +B^\circ *T'+B^\circ *B^\circ +B^\circ *{\bar{T}}'+B*{\bar{T}}'+B*{\bar{\Gamma }})\\&+{\bar{Z}}Z(B^\circ )+{\bar{Z}}(B)*{\bar{\Gamma }}+B*{\bar{Z}}({\bar{\Gamma }}) +{\bar{Z}}(B)*B^\circ +B*{\bar{Z}}(B^\circ )\\&+{\bar{\Gamma }}*{\bar{Z}}(T')+B^\circ *{\bar{Z}}(T')+{\bar{\nabla }}B^\circ *T'+B^\circ *{\bar{\nabla }}T'+{\mathcal {O}}(Z(g)), \\ \nabla _{m}\nabla _{i}(T_{{\bar{j}}{\bar{k}}r})= & {} ZZ({\bar{\Gamma }})+\Gamma *Z({\bar{T}}')+B^\circ *Z({\bar{T}}') +\nabla B^\circ *{\bar{T}}'+B^\circ *\nabla {\bar{T}}'\\&-Z_{m}(\overline{Z_{{\bar{i}}}(B_{jk}^l)g_{l{\bar{r}}}})+Z_i(\overline{B_{jk}^l})Z_{m}(g_{r{\bar{l}}})\\= & {} ZZ(\Gamma )+\Gamma *Z({\bar{T}}')+B^\circ *Z({\bar{T}}') +\nabla B^\circ *{\bar{T}}'+B^\circ *\nabla {\bar{T}}'\\&+Z(B^\circ *{\bar{\Gamma }}+B^\circ *{\bar{T}}'+B^\circ *B^\circ +B^\circ *T'+{\bar{B}}*T'+{\bar{B}}*\Gamma )\\&-Z_{m}Z_{{\bar{j}}}(B_{i{\bar{k}}}^{{\bar{l}}}g_{r{\bar{l}}})-Z_{m}Z_{{\bar{k}}}(B_{{\bar{j}}i}^{{\bar{l}}}g_{r{\bar{l}}})\\&+Z_{m}(\overline{B_{jk}^l}\Gamma _{ir}^sg_{s{\bar{l}}} +\overline{B_{jk}^l}\Gamma _{i{\bar{l}}}^{{\bar{s}}}g_{r{\bar{s}}})+Z_i(\overline{B_{jk}^l})Z_{m}(g_{r{\bar{l}}})\\= & {} ZZ(\Gamma )+\Gamma *Z({\bar{T}}')+B^\circ *Z({\bar{T}}') +\nabla B^\circ *{\bar{T}}'+B^\circ *\nabla {\bar{T}}'\\&+Z(B^\circ *{\bar{\Gamma }}+B^\circ *{\bar{T}}'+B^\circ *B^\circ +B^\circ *T'+{\bar{B}}*T'+{\bar{B}}*\Gamma )\\&+Z{\bar{Z}}(B^\circ )+Z({\bar{B}})*\Gamma +{\bar{B}}*Z(\Gamma ) +Z({\bar{B}})*B^\circ +{\bar{B}}*Z(B^\circ )+{\mathcal {O}}(Z(g)) \end{aligned}$$

and

$$\begin{aligned} \nabla _{{\bar{i}}}\nabla _{m}(T_{{\bar{j}}{\bar{k}}r})= & {} {\bar{Z}}Z({\bar{\Gamma }})+B^\circ *Z({\bar{T}}')+{\bar{\Gamma }}*Z({\bar{T}}') +{\bar{\nabla }}B^\circ *{\bar{T}}'+B^\circ *{\bar{\nabla }}{\bar{T}}'\\&-Z_{{\bar{i}}}(\overline{Z_{{\bar{m}}}(B_{jk}^l)g_{l{\bar{r}}}})+Z_m(\overline{B_{jk}^l})Z_{{\bar{i}}}(g_{r{\bar{l}}})\\= & {} {\bar{Z}}Z({\bar{\Gamma }})+B^\circ *Z({\bar{T}}')+{\bar{\Gamma }}*Z({\bar{T}}') +{\bar{\nabla }}B^\circ *{\bar{T}}'+B^\circ *{\bar{\nabla }}{\bar{T}}'\\&-Z_{{\bar{i}}}(B^\circ *{\bar{\Gamma }}+B^\circ *{\bar{T}}'+B^\circ *B^\circ +B^\circ *T' +{\bar{B}}*T'+{\bar{B}}*\Gamma )\\&-Z_{{\bar{i}}}Z_{{\bar{j}}}(B_{m{\bar{k}}}^{{\bar{l}}}g_{r{\bar{l}}})-Z_{{\bar{i}}}Z_{{\bar{k}}}(B_{{\bar{j}}m}^{{\bar{l}}}g_{r{\bar{l}}})+Z_{{\bar{i}}}(\overline{B_{jk}^l}\Gamma _{mr}^sg_{s{\bar{l}}} +\overline{B_{jk}^l}\Gamma _{m{\bar{l}}}^{{\bar{s}}}g_{r{\bar{s}}})\\&+Z_m(\overline{B_{jk}^l})Z_{{\bar{i}}}(g_{r{\bar{l}}})\\= & {} {\bar{Z}}Z({\bar{\Gamma }})+{\bar{Z}}{\bar{Z}}(B^\circ )+B^\circ *Z({\bar{T}}')+{\bar{\Gamma }}*Z({\bar{T}}') +{\bar{\nabla }}B^\circ *{\bar{T}}'+B^\circ *{\bar{\nabla }}{\bar{T}}'\\&+{\bar{Z}}(B^\circ *{\bar{\Gamma }}+B^\circ *{\bar{T}}'+B^\circ *B^\circ +B^\circ *T' +{\bar{B}}*T'+{\bar{B}}*\Gamma )\\&+{\bar{Z}}({\bar{B}})*\Gamma +{\bar{B}}*{\bar{Z}}(\Gamma )+{\bar{Z}}({\bar{B}})*B^\circ +{\bar{B}}*{\bar{Z}}(B^\circ )+{\mathcal {O}}({\bar{Z}}(g)) \end{aligned}$$

Then, by applying the estimates in Lemmas 3.1, 3.3, we obtain the uniform bound for \(|{\bar{\nabla }}^2T'(g(t_0))|^2_{C^0(g(t_0))}\), \(|{\bar{\nabla }}\nabla {\bar{T}}'(g(t_0))|^2_{C^0(g(t_0))}\) for any fixed time \(t_0\in [0,\tau )\). \(\square \)

Recall that we have

$$\begin{aligned} H= & {} H_{ijk{\bar{r}}}\\= & {} H_{ijk}^{\quad \,\,l}g_{l{\bar{r}}}\\= & {} (Z_i(\Gamma _{jk}^l)-Z_{j}(\Gamma _{ik}^l)+\Gamma _{is}^l\Gamma _{jk}^s-\Gamma _{js}^l\Gamma _{ik}^s-B_{ij}^s\Gamma _{sk}^l-B_{ij}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^l)g_{l{\bar{r}}}\\= & {} Z(\Gamma )+\Gamma *\Gamma +B*\Gamma +B^\circ *B^\circ \\= & {} Z(\Gamma )+\Gamma *\Gamma +T'*\Gamma +B^\circ *B^\circ . \end{aligned}$$

Note that we have for any \(j=0,1,2\dots \),

$$\begin{aligned} {\bar{\nabla }}^jH= & {} {\bar{\nabla }}^j(Z(\Gamma )+\Gamma *\Gamma +T'*\Gamma +B^\circ *B^\circ )\\= & {} {\bar{\nabla }}^j(Z(\Gamma ))+\sum _{l=0}^j{\bar{\nabla }}^l\Gamma *{\bar{\nabla }}^{j-l}\Gamma +\sum _{l=0}^j{\bar{\nabla }}^lT'*{\bar{\nabla }}^{j-l}\Gamma +\sum _{l=0}^j{\bar{\nabla }}^lB^\circ *{\bar{\nabla }}^{j-l}B^\circ . \end{aligned}$$

From the equality above and Lemma 3.1, we have the following estimate.

Lemma 3.5

One can obtain the following estimate for any \(j=0,1,2,\dots \) and for a time \(t_0\in [0,\tau )\),

$$\begin{aligned} |{\bar{\nabla }}^jH(g(t_0))|_{C^0(g(t_0))}\le C\sum _{l=0}^j|{\bar{\nabla }}^{j-l}T'(g(t_0))|_{C^0(g(t_0))}+C. \end{aligned}$$

Especially, from Lemmas 3.4 and 3.5, under the assumption that \(|T'(g(t))|^2_{C^0(g(t))}\), \(|\nabla T'(g(t))|_{C^0(g(t))}\) have uniform bounds, then we obtain for any fixed time \(t_0\in [0,\tau )\),

$$\begin{aligned} |{\bar{\nabla }}^2H(g(t_0))|_{C^0(g(t_0))}\le C,\quad |\nabla ^2{\bar{H}}(g(t_0))|_{C^0(g(t_0))}\le C \end{aligned}$$

for some uniform constant \(C>0\) independent \(t_0\).

From now on, we consider the solution \(g=g(t)\) of the AHCF starting at the initial almost Hermitian metric \(g_0\) on a compact almost Hermitian manifold M satisfying

$$\begin{aligned} \left\{ \begin{array}{l}\dfrac{\partial }{\partial t}g(t)=-S(g(t))-Q^7(g(t))-Q^8(g(t))+BT'(g(t))+{\bar{Z}}(T')(g(t)), \\ \\ g(0)=g_0,\end{array}\right. \end{aligned}$$

Lemma 3.6

For a solution to the AHCF, one has that

(3.2)

Proof

We consider the term \(-S\) in the evolution equation of g. Using Lemma 3.2, the evolution \(\frac{\partial }{\partial t}g=-S\) yields

$$\begin{aligned} \frac{\partial }{\partial t}R_{i{\bar{j}}k{\bar{l}}}= -R_{i{\bar{j}}k}^{\quad r}S_{r{\bar{l}}}+\nabla _{{\bar{j}}}\nabla _iS_{k{\bar{l}}}. \end{aligned}$$

Applying the second Bianchi identity in Lemma 2.2, we have

$$\begin{aligned} \nabla _{{\bar{j}}}\nabla _iS_{k{\bar{l}}}= & {} g^{m{\bar{n}}}\nabla _{{\bar{j}}}\nabla _iR_{m{\bar{n}}k{\bar{l}}}\\= & {} g^{m{\bar{n}}}\nabla _{{\bar{j}}}(\nabla _mR_{i{\bar{n}}k{\bar{l}}}+\nabla _{{\bar{n}}}H_{mik{\bar{l}}}+T_{mi}^sR_{s{\bar{n}}k{\bar{l}}}-B_{mi}^{{\bar{s}}}H_{{\bar{s}}{\bar{n}}k{\bar{l}}})\\= & {} g^{m{\bar{n}}}(\nabla _{{\bar{j}}}\nabla _mR_{i{\bar{n}}k{\bar{l}}}+\nabla _{{\bar{j}}}\nabla _{{\bar{n}}}H_{mik{\bar{l}}}+\nabla _{{\bar{j}}}T_{mi}^sR_{s{\bar{n}}k{\bar{l}}}+T_{mi}^s\nabla _{{\bar{j}}}R_{s{\bar{n}}k{\bar{l}}}\\&-\nabla _{{\bar{j}}}B_{mi}^{{\bar{s}}}H_{{\bar{s}}{\bar{n}}k{\bar{l}}}-B_{mi}^{{\bar{s}}}\nabla _{{\bar{j}}}H_{{\bar{s}}{\bar{n}}k{\bar{l}}}). \end{aligned}$$

We compute

$$\begin{aligned}&\nabla _{{\bar{j}}}\nabla _mR_{i{\bar{n}}k{\bar{l}}} =\nabla _m\nabla _{{\bar{j}}}R_{i{\bar{n}}k{\bar{l}}}+\Gamma _{m{\bar{j}}}^{{\bar{s}}}\nabla _{{\bar{s}}}R_{i{\bar{n}}k{\bar{l}}}+\Gamma _{mi}^s\nabla _{{\bar{j}}}R_{s{\bar{n}}k{\bar{l}}}+\Gamma _{m{\bar{n}}}^{{\bar{s}}}\nabla _{{\bar{j}}}R_{i{\bar{s}}k{\bar{l}}}\\&\quad +\Gamma _{mk}^s\nabla _{{\bar{j}}}R_{i{\bar{n}}s{\bar{l}}}+\Gamma _{m{\bar{l}}}^{{\bar{s}}}\nabla _{{\bar{j}}}R_{i{\bar{n}}k{\bar{s}}}+Z_m(\Gamma _{{\bar{j}}i}^sR_{s{\bar{n}}k{\bar{l}}}+\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}R_{i{\bar{s}}k{\bar{l}}}+\Gamma _{{\bar{j}}k}^sR_{i{\bar{n}}s{\bar{l}}}+\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}R_{i{\bar{n}}k{\bar{s}}})\\&\quad +[Z_{{\bar{j}}},Z_m]R_{i{\bar{n}}k{\bar{l}}} -Z_{{\bar{j}}}(\Gamma _{mi}^sR_{s{\bar{n}}k{\bar{l}}}+\Gamma _{m{\bar{n}}}^{{\bar{s}}}R_{i{\bar{s}}k{\bar{l}}}+\Gamma _{mk}^sR_{i{\bar{n}}s{\bar{l}}}+\Gamma _{m{\bar{l}}}^{{\bar{s}}}R_{i{\bar{n}}k{\bar{s}}})\\&\quad -\Gamma _{{\bar{j}}i}^s\nabla _mR_{s{\bar{n}}k{\bar{l}}}-\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\nabla _mR_{i{\bar{s}}k{\bar{l}}} -\Gamma _{{\bar{j}}k}^s\nabla _mR_{i{\bar{n}}s{\bar{l}}} -\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\nabla _mR_{i{\bar{n}}k{\bar{s}}} -\Gamma _{{\bar{j}}m}^{s}\nabla _sR_{i{\bar{n}}k{\bar{l}}} \end{aligned}$$

and

$$\begin{aligned} \nabla _m\nabla _{{\bar{j}}}R_{i{\bar{n}}k{\bar{l}}}= & {} \nabla _m\overline{\nabla _{j}R_{n{\bar{i}}l{\bar{k}}}}\\= & {} \nabla _m(\overline{\nabla _{n}R_{j{\bar{i}}l{\bar{k}}}+\nabla _{{\bar{i}}}H_{njl{\bar{k}}}+T_{nj}^sR_{s{\bar{i}}l{\bar{k}}}+T_{nj}^{{\bar{s}}}H_{{\bar{s}}{\bar{i}}l{\bar{k}}}})\\= & {} \nabla _m\nabla _{{\bar{n}}}R_{i{\bar{j}}k{\bar{l}}} +\nabla _m\nabla _{i}H_{{\bar{j}}{\bar{n}}k{\bar{l}}}+\nabla _m(T_{{\bar{n}}{\bar{j}}}^{{\bar{s}}}R_{i{\bar{s}}k{\bar{l}}}+T_{{\bar{n}}{\bar{j}}}^{s}H_{isk{\bar{l}}}). \end{aligned}$$

Hence, we obtain

$$\begin{aligned}&\nabla _{{\bar{j}}}\nabla _iS_{k{\bar{l}}}=\Delta R_{i{\bar{j}}k{\bar{l}}}+g^{m{\bar{n}}}(\nabla _m\nabla _{i}H_{{\bar{j}}{\bar{n}}k{\bar{l}}}+\nabla _m(T_{{\bar{n}}{\bar{j}}}^{{\bar{s}}}R_{i{\bar{s}}k{\bar{l}}}+T_{{\bar{n}}{\bar{j}}}^{s}H_{isk{\bar{l}}}))\\&\qquad +g^{m{\bar{n}}}(\Gamma _{m{\bar{j}}}^{{\bar{s}}}\nabla _{{\bar{s}}}R_{i{\bar{n}}k{\bar{l}}}+\Gamma _{mi}^s\nabla _{{\bar{j}}}R_{s{\bar{n}}k{\bar{l}}}+\Gamma _{m{\bar{n}}}^{{\bar{s}}}\nabla _{{\bar{j}}}R_{i{\bar{s}}k{\bar{l}}}+\Gamma _{mk}^s\nabla _{{\bar{j}}}R_{i{\bar{n}}s{\bar{l}}}\\&\qquad +\Gamma _{m{\bar{l}}}^{{\bar{s}}}\nabla _{{\bar{j}}}R_{i{\bar{n}}k{\bar{s}}})+g^{m{\bar{n}}}Z_m(\Gamma _{{\bar{j}}i}^sR_{s{\bar{n}}k{\bar{l}}}+\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}R_{i{\bar{s}}k{\bar{l}}}+\Gamma _{{\bar{j}}k}^sR_{i{\bar{n}}s{\bar{l}}}+\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}R_{i{\bar{n}}k{\bar{s}}})\\&\qquad +g^{m{\bar{n}}}[Z_{{\bar{j}}},Z_m]R_{i{\bar{n}}k{\bar{l}}} -g^{m{\bar{n}}}Z_{{\bar{j}}}(\Gamma _{mi}^sR_{s{\bar{n}}k{\bar{l}}}+\Gamma _{m{\bar{n}}}^{{\bar{s}}}R_{i{\bar{s}}k{\bar{l}}}+\Gamma _{mk}^sR_{i{\bar{n}}s{\bar{l}}}+\Gamma _{m{\bar{l}}}^{{\bar{s}}}R_{i{\bar{n}}k{\bar{s}}})\\&\qquad -g^{m{\bar{n}}}(\Gamma _{{\bar{j}}i}^s\nabla _mR_{s{\bar{n}}k{\bar{l}}}+\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\nabla _mR_{i{\bar{s}}k{\bar{l}}} +\Gamma _{{\bar{j}}k}^s\nabla _mR_{i{\bar{n}}s{\bar{l}}} +\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\nabla _mR_{i{\bar{n}}k{\bar{s}}} \\&\qquad +\Gamma _{{\bar{j}}m}^{s}\nabla _sR_{i{\bar{n}}k{\bar{l}}}) +g^{m{\bar{n}}}(\nabla _{{\bar{j}}}\nabla _{{\bar{n}}}H_{mik{\bar{l}}}+\nabla _{{\bar{j}}}(T_{mi}^sR_{s{\bar{n}}k{\bar{l}}}-B_{mi}^{{\bar{s}}}H_{{\bar{s}}{\bar{n}}k{\bar{l}}}))\\&\quad =\Delta R_{i{\bar{j}}kl} +g^{m{\bar{n}}}(\nabla _m\nabla _{i}H_{{\bar{j}}{\bar{n}}k{\bar{l}}}+\nabla _mT_{{\bar{n}}{\bar{j}}}^{{\bar{s}}}R_{i{\bar{s}}k{\bar{l}}}+T_{{\bar{n}}{\bar{j}}}^{{\bar{s}}}\nabla _mR_{i{\bar{s}}k{\bar{l}}}+\nabla _mT_{{\bar{n}}{\bar{j}}}^{s}H_{isk{\bar{l}}}\\&\qquad +T_{{\bar{n}}{\bar{j}}}^{s}\nabla _mH_{isk{\bar{l}}}+\Gamma _{m{\bar{j}}}^{{\bar{s}}}\nabla _{{\bar{s}}}R_{i{\bar{n}}k{\bar{l}}}+\Gamma _{mi}^s\nabla _{{\bar{j}}}R_{s{\bar{n}}k{\bar{l}}}+\Gamma _{m{\bar{n}}}^{{\bar{s}}}\nabla _{{\bar{j}}}R_{i{\bar{s}}k{\bar{l}}}+\Gamma _{mk}^s\nabla _{{\bar{j}}}R_{i{\bar{n}}s{\bar{l}}}\\&\qquad +\Gamma _{m{\bar{l}}}^{{\bar{s}}}\nabla _{{\bar{j}}}R_{i{\bar{n}}k{\bar{s}}}+Z_m(\Gamma _{{\bar{j}}i}^s)R_{s{\bar{n}}k{\bar{l}}}+\Gamma _{{\bar{j}}i}^s\nabla _mR_{s{\bar{n}}k{\bar{l}}}+\Gamma _{{\bar{j}}i}^s\Gamma _{ms}^pR_{p{\bar{n}}k{\bar{l}}}+\Gamma _{{\bar{j}}i}^s\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{s{\bar{p}}k{\bar{l}}}\\&\qquad +\Gamma _{{\bar{j}}i}^s\Gamma _{mk}^pR_{s{\bar{n}}p{\bar{l}}}+\Gamma _{{\bar{j}}i}^s\Gamma _{m{\bar{l}}}^{{\bar{p}}}R_{s{\bar{n}}k{\bar{p}}}+Z_m(\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}})R_{i{\bar{s}}k{\bar{l}}} +\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\nabla _mR_{i{\bar{s}}k{\bar{l}}}+\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{mi}^pR_{p{\bar{s}}k{\bar{l}}}\\&\qquad +\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}}+\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{mk}^pR_{i{\bar{s}}p{\bar{l}}} +\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{m{\bar{l}}}^{{\bar{p}}}R_{i{\bar{s}}k{\bar{p}}}+Z_m(\Gamma _{{\bar{j}}k}^s)R_{i{\bar{n}}s{\bar{l}}}+\Gamma _{{\bar{j}}k}^s\nabla _mR_{i{\bar{n}}s{\bar{l}}}\\&\qquad +\Gamma _{{\bar{j}}k}^s\Gamma _{mi}^pR_{p{\bar{n}}s{\bar{l}}}+\Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}s{\bar{l}}} +\Gamma _{{\bar{j}}k}^s\Gamma _{ms}^pR_{i{\bar{n}}p{\bar{l}}}+\Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}s{\bar{p}}}+Z_m(\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}})R_{i{\bar{n}}k{\bar{s}}}\\&\qquad +\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\nabla _mR_{i{\bar{n}}k{\bar{s}}} +\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{mi}^pR_{p{\bar{n}}k{\bar{s}}}+\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{s}}} +\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{mk}^pR_{i{\bar{n}}p{\bar{s}}}+\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}}\\&\qquad +B_{{\bar{j}}m}^s\nabla _sR_{i{\bar{n}}k{\bar{l}}}+B_{{\bar{j}}m}^s\Gamma _{si}^pR_{p{\bar{n}}k{\bar{l}}} +B_{{\bar{j}}m}^s\Gamma _{s{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}} +B_{{\bar{j}}m}^s\Gamma _{sk}^pR_{i{\bar{n}}p{\bar{l}}}+B_{{\bar{j}}m}^s\Gamma _{s{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}}\\&\qquad +B_{{\bar{j}}m}^{{\bar{s}}}\nabla _{{\bar{s}}}R_{i{\bar{n}}k{\bar{l}}} +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}i}^pR_{p{\bar{n}}k{\bar{l}}} +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}} +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^{p}R_{i{\bar{n}}p{\bar{l}}} +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}} \\&\qquad -Z_{{\bar{j}}}(\Gamma _{mi}^s)R_{s{\bar{n}}k{\bar{l}}}-\Gamma _{mi}^s\nabla _{{\bar{j}}}R_{s{\bar{n}}k{\bar{l}}} -\Gamma _{mi}^s\Gamma _{{\bar{j}}s}^pR_{p{\bar{n}}k{\bar{l}}} -\Gamma _{mi}^s\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}R_{s{\bar{p}}k{\bar{l}}} -\Gamma _{mi}^s\Gamma _{{\bar{j}}k}^{p}R_{s{\bar{n}}p{\bar{l}}} \\&\qquad -\Gamma _{mi}^s\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}R_{s{\bar{n}}k{\bar{p}}} -Z_{{\bar{j}}}(\Gamma _{m{\bar{n}}}^{{\bar{s}}})R_{i{\bar{s}}k{\bar{l}}} -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\nabla _{{\bar{j}}}R_{i{\bar{s}}k{\bar{l}}} -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}i}^pR_{p{\bar{s}}k{\bar{l}}} -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}} \\&\qquad -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}k}^{p}R_{i{\bar{s}}p{\bar{l}}} -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}R_{i{\bar{s}}k{\bar{p}}} -Z_{{\bar{j}}}(\Gamma _{mk}^{s})R_{i{\bar{n}}s{\bar{l}}} -\Gamma _{mk}^{s}\nabla _{{\bar{j}}}R_{i{\bar{n}}s{\bar{l}}} -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}i}^pR_{p{\bar{n}}s{\bar{l}}} \\&\qquad -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}s{\bar{l}}} -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}s}^{p}R_{i{\bar{n}}p{\bar{l}}} -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}s{\bar{p}}} -Z_{{\bar{j}}}(\Gamma _{m{\bar{l}}}^{{\bar{s}}})R_{i{\bar{n}}k{\bar{s}}} -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\nabla _{{\bar{j}}}R_{i{\bar{n}}k{\bar{s}}}\\&\qquad -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}i}^pR_{p{\bar{n}}k{\bar{s}}} -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{s}}} -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}k}^{p}R_{i{\bar{n}}p{\bar{s}}} -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}} \\&\qquad -\Gamma _{{\bar{j}}i}^s\nabla _mR_{s{\bar{n}}k{\bar{l}}}-\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\nabla _mR_{i{\bar{s}}k{\bar{l}}} -\Gamma _{{\bar{j}}k}^s\nabla _mR_{i{\bar{n}}s{\bar{l}}} -\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\nabla _mR_{i{\bar{n}}k{\bar{s}}} -\Gamma _{{\bar{j}}m}^{s}\nabla _sR_{i{\bar{n}}k{\bar{l}}}\\&\qquad +\nabla _{{\bar{j}}}\nabla _{{\bar{n}}}H_{mik{\bar{l}}}+\nabla _{{\bar{j}}}T_{mi}^sR_{s{\bar{n}}k{\bar{l}}}+T_{mi}^s\nabla _{{\bar{j}}}R_{s{\bar{n}}k{\bar{l}}}-\nabla _{{\bar{j}}}B_{mi}^{{\bar{s}}}H_{{\bar{s}}{\bar{n}}k{\bar{l}}}-B_{mi}^{{\bar{s}}}\nabla _{{\bar{j}}}H_{{\bar{s}}{\bar{n}}k{\bar{l}}}), \end{aligned}$$

where we used that \(R_{{\bar{p}}{\bar{n}}k{\bar{l}}}=R_{ipk{\bar{l}}}=0\),

$$\begin{aligned}{}[Z_{{\bar{j}}},Z_m](R_{i{\bar{n}}k{\bar{l}}})=B_{{\bar{j}}m}^sZ_s(R_{i{\bar{n}}k{\bar{l}}})+B_{{\bar{j}}m}^{{\bar{s}}}Z_{{\bar{s}}}(R_{i{\bar{n}}k{\bar{l}}}). \end{aligned}$$

Note that we have that \(\Gamma _{m{\bar{j}}}^{{\bar{s}}}\nabla _{{\bar{s}}}R_{i{\bar{n}}k{\bar{l}}}+B_{{\bar{j}}m}^{{\bar{s}}}\nabla _{{\bar{s}}}R_{i{\bar{n}}k{\bar{l}}}=0\) since \(B_{{\bar{j}}m}^{{\bar{s}}}=-B_{m{\bar{j}}}^{{\bar{s}}}=-\Gamma _{m{\bar{j}}}^{{\bar{s}}}\).

Then we obtain the following equality:

$$\begin{aligned}&\nabla _{{\bar{j}}}\nabla _iS_{k{\bar{l}}} =\Delta R_{i{\bar{j}}k{\bar{l}}}+T_{{\bar{m}}{\bar{j}}}^{{\bar{s}}}\nabla _mR_{i{\bar{s}}k{\bar{l}}}+T_{mi}^s(\nabla _{{\bar{j}}}R_{s{\bar{m}}k{\bar{l}}}+\nabla _sH_{{\bar{j}}{\bar{m}}k{\bar{l}}}+T_{{\bar{m}}{\bar{j}}}^{{\bar{p}}}R_{s{\bar{p}}k{\bar{l}}}\\&\quad +T^p_{{\bar{m}}{\bar{j}}}H_{spk{\bar{l}}}) +g^{m{\bar{n}}}(\nabla _m\nabla _{i}H_{{\bar{j}}{\bar{n}}k{\bar{l}}}+\nabla _mT_{{\bar{n}}{\bar{j}}}^{{\bar{s}}}R_{i{\bar{s}}k{\bar{l}}}+\nabla _mT_{{\bar{n}}{\bar{j}}}^{s}H_{isk{\bar{l}}}+T_{{\bar{n}}{\bar{j}}}^{s}\nabla _mH_{isk{\bar{l}}}\\&\quad +Z_m(\Gamma _{{\bar{j}}i}^s)R_{s{\bar{n}}k{\bar{l}}}+\Gamma _{{\bar{j}}i}^s\Gamma _{ms}^pR_{p{\bar{n}}k{\bar{l}}}+\Gamma _{{\bar{j}}i}^s\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{s{\bar{p}}k{\bar{l}}}+\Gamma _{{\bar{j}}i}^s\Gamma _{mk}^pR_{s{\bar{n}}p{\bar{l}}}+\Gamma _{{\bar{j}}i}^s\Gamma _{m{\bar{l}}}^{{\bar{p}}}R_{s{\bar{n}}k{\bar{p}}}\\&\quad +Z_m(\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}})R_{i{\bar{s}}k{\bar{l}}} +\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{mi}^pR_{p{\bar{s}}k{\bar{l}}} +\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}}+\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{mk}^pR_{i{\bar{s}}p{\bar{l}}}+\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{m{\bar{l}}}^{{\bar{p}}}R_{i{\bar{s}}k{\bar{p}}}\\&\quad +Z_m(\Gamma _{{\bar{j}}k}^s)R_{i{\bar{n}}s{\bar{l}}}+\Gamma _{{\bar{j}}k}^s\Gamma _{mi}^pR_{p{\bar{n}}s{\bar{l}}} +\Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}s{\bar{l}}}+\Gamma _{{\bar{j}}k}^s\Gamma _{ms}^pR_{i{\bar{n}}p{\bar{l}}}+\Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}s{\bar{p}}}\\&\quad +Z_m(\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}})R_{i{\bar{n}}k{\bar{s}}} +\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{mi}^pR_{p{\bar{n}}k{\bar{s}}}+\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{s}}}+\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{mk}^pR_{i{\bar{n}}p{\bar{s}}}+\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}}\\&\quad +B_{{\bar{j}}m}^s\Gamma _{si}^pR_{p{\bar{n}}k{\bar{l}}} +B_{{\bar{j}}m}^s\Gamma _{s{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}} +B_{{\bar{j}}m}^s\Gamma _{sk}^pR_{i{\bar{n}}p{\bar{l}}}+B_{{\bar{j}}m}^s\Gamma _{s{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}} +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}i}^pR_{p{\bar{n}}k{\bar{l}}}\\&\quad +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}} +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^{p}R_{i{\bar{n}}p{\bar{l}}}+B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}} -Z_{{\bar{j}}}(\Gamma _{mi}^s)R_{s{\bar{n}}k{\bar{l}}}-\Gamma _{mi}^s\Gamma _{{\bar{j}}s}^pR_{p{\bar{n}}k{\bar{l}}}\\&\quad -\Gamma _{mi}^s\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}R_{s{\bar{p}}k{\bar{l}}} -\Gamma _{mi}^s\Gamma _{{\bar{j}}k}^{p}R_{s{\bar{n}}p{\bar{l}}} -\Gamma _{mi}^s\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}R_{s{\bar{n}}k{\bar{p}}}-Z_{{\bar{j}}}(\Gamma _{m{\bar{n}}}^{{\bar{s}}})R_{i{\bar{s}}k{\bar{l}}} -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}i}^pR_{p{\bar{s}}k{\bar{l}}}\\&\quad -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}} -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}k}^{p}R_{i{\bar{s}}p{\bar{l}}}-\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}R_{i{\bar{s}}k{\bar{p}}}-Z_{{\bar{j}}}(\Gamma _{mk}^{s})R_{i{\bar{n}}s{\bar{l}}} -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}i}^pR_{p{\bar{n}}s{\bar{l}}}\\&\quad -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}s{\bar{l}}} -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}s}^{p}R_{i{\bar{n}}p{\bar{l}}} -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}s{\bar{p}}} -Z_{{\bar{j}}}(\Gamma _{m{\bar{l}}}^{{\bar{s}}})R_{i{\bar{n}}k{\bar{s}}}-\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}i}^pR_{p{\bar{n}}k{\bar{s}}}\\&\quad -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{s}}} -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}k}^{p}R_{i{\bar{n}}p{\bar{s}}} -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}} +\nabla _{{\bar{j}}}\nabla _{{\bar{n}}}H_{mik{\bar{l}}}\\&\quad +\nabla _{{\bar{j}}}T_{mi}^sR_{s{\bar{n}}k{\bar{l}}} -\nabla _{{\bar{j}}}B_{mi}^{{\bar{s}}}H_{{\bar{s}}{\bar{n}}k{\bar{l}}}-B_{mi}^{{\bar{s}}}\nabla _{{\bar{j}}}H_{{\bar{s}}{\bar{n}}k{\bar{l}}}), \end{aligned}$$

where we have used that by applying (2.3),

$$\begin{aligned}&\nabla _{{\bar{j}}}R_{s{\bar{m}}k{\bar{l}}}=\overline{\nabla _jR_{m{\bar{s}}l{\bar{k}}}}\\&\quad =\overline{\nabla _mR_{j{\bar{s}}l{\bar{k}}}+\nabla _{{\bar{s}}}H_{mjl{\bar{k}}}+T_{mj}^pR_{p{\bar{s}}l{\bar{k}}}+T_{mj}^{{\bar{p}}}H_{{\bar{p}}{\bar{s}}l{\bar{k}}}}\\&\quad =\nabla _{{\bar{m}}}R_{s{\bar{j}}k{\bar{l}}}+\nabla _sH_{{\bar{j}}{\bar{m}}k{\bar{l}}}+T_{{\bar{m}}{\bar{j}}}^{{\bar{p}}}R_{s{\bar{p}}k{\bar{l}}}+T^p_{{\bar{m}}{\bar{j}}}H_{spk{\bar{l}}}. \end{aligned}$$

We have the following:

$$\begin{aligned}&-(Z_{{\bar{j}}}(\Gamma _{m{\bar{n}}}^{{\bar{s}}})-Z_m(\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}))R_{i{\bar{s}}k{\bar{l}}}=-\overline{(Z_{j}(\Gamma _{{\bar{m}}n}^{s})-Z_{{\bar{m}}}(\Gamma _{jn}^{s}))}R_{i{\bar{s}}k{\bar{l}}}\\&\quad =-\overline{(R_{j{\bar{m}}n}^{\quad \,\, s}-\Gamma _{jp}^s\Gamma _{{\bar{m}}n}^p+\Gamma _{{\bar{m}}p}^s\Gamma _{jn}^p+B_{j{\bar{m}}}^p\Gamma _{pn}^s-B_{{\bar{m}}j}^{{\bar{p}}}\Gamma _{{\bar{p}}n}^s)}R_{i{\bar{s}}k{\bar{l}}}\\&\quad =(-R_{m{\bar{j}}\,\,\,\,{\bar{n}}}^{\quad \,\,{\bar{s}}}+\Gamma _{{\bar{j}}{\bar{p}}}^{{\bar{s}}}\Gamma _{m{\bar{n}}}^{{\bar{p}}}-\Gamma _{m{\bar{p}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}-B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}{\bar{n}}}^{{\bar{s}}}+B_{m{\bar{j}}}^{p}\Gamma _{p{\bar{n}}}^{{\bar{s}}})R_{i{\bar{s}}k{\bar{l}}}, \end{aligned}$$

and similarly,

$$\begin{aligned}&-(Z_{{\bar{j}}}(\Gamma _{m{\bar{l}}}^{{\bar{s}}})-Z_m(\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}))R_{i{\bar{n}}k{\bar{s}}}\\&\quad =(-R_{m{\bar{j}}\,\,\,\,{\bar{l}}}^{\quad \,\,{\bar{s}}}+\Gamma _{{\bar{j}}{\bar{p}}}^{{\bar{s}}}\Gamma _{m{\bar{l}}}^{{\bar{p}}}-\Gamma _{m{\bar{p}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}-B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}{\bar{l}}}^{{\bar{s}}}+B_{m{\bar{j}}}^{p}\Gamma _{p{\bar{l}}}^{{\bar{s}}})R_{i{\bar{n}}k{\bar{s}}},\\&(Z_m(\Gamma _{{\bar{j}}k}^s)-Z_{{\bar{j}}}(\Gamma _{mk}^s))R_{i{\bar{n}}s{\bar{l}}}\\&\quad =(R_{m{\bar{j}}k}^{\quad \,\,s}-\Gamma _{mp}^s\Gamma _{{\bar{j}}k}^p+\Gamma _{{\bar{j}}p}^s\Gamma _{mk}^p+B_{m{\bar{j}}}^p\Gamma _{pk}^s-B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}k}^s)R_{i{\bar{n}}s{\bar{l}}},\\&(Z_m(\Gamma _{{\bar{j}}i}^s)-Z_{{\bar{j}}}(\Gamma _{mi}^s))R_{s{\bar{n}}k{\bar{l}}}\\&\quad =(R_{m{\bar{j}}i}^{\quad \,\,s}-\Gamma _{mp}^s\Gamma _{{\bar{j}}i}^p+\Gamma _{{\bar{j}}p}^s\Gamma _{mi}^p+B_{m{\bar{j}}}^p\Gamma _{pi}^s-B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}i}^s)R_{s{\bar{n}}k{\bar{l}}}\\&\quad =(R_{i{\bar{j}}m}^{\quad \,\,s}-B_{mi}^{{\bar{q}}}B_{{\bar{q}}{\bar{j}}}^s+\nabla _{{\bar{j}}}T_{im}^s-\Gamma _{mp}^s\Gamma _{{\bar{j}}i}^p+\Gamma _{{\bar{j}}p}^s\Gamma _{mi}^p+B_{m{\bar{j}}}^p\Gamma _{pi}^s-B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}i}^s)R_{s{\bar{n}}k{\bar{l}}},\nonumber \end{aligned}$$

where we used that \(R_{m{\bar{j}}i}^{\quad \,\,s}=R_{i{\bar{j}}m}^{\quad \,\,s}-B_{mi}^{{\bar{q}}}B_{{\bar{q}}{\bar{j}}}^s+\nabla _{{\bar{j}}}T_{im}^s.\)

Choosing \(t=t_0\) and a local unitary (1, 0)-frame \(\{Z_r\}\) around a fixed point \(p\in M\) with respect to \(g(t_0)\). Then using the local \(g(t_0)\)-unitary frame, \(g(t_0)^{m{\bar{n}}}=\delta _{mn}\),

$$\begin{aligned} \Gamma _{m{\bar{j}}}^{{\bar{p}}}=-\Gamma _{mp}^j,\quad \Gamma _{{\bar{i}}{\bar{j}}}^{{\bar{s}}}=-\Gamma _{{\bar{i}}s}^j. \end{aligned}$$

By combining these calculations, we compute for the evolution \(\frac{\partial }{\partial t}g=-S\), with a local unitary frame,

$$\begin{aligned}&\frac{\partial }{\partial t}R_{i{\bar{j}}k{\bar{l}}}=-R_{i{\bar{j}}k}^{\quad \,\,r}S_{r{\bar{l}}}+\nabla _{{\bar{j}}}\nabla _iS_{k{\bar{l}}}\\&\quad =-R_{i{\bar{j}}k}^{\quad \,\,r}S_{r{\bar{l}}}+\Delta R_{i{\bar{j}}k{\bar{l}}}+g^{r{\bar{s}}}(T_{{\bar{s}}{\bar{j}}}^{{\bar{q}}}\nabla _rR_{i{\bar{q}}k{\bar{l}}}+T_{ri}^p\nabla _{{\bar{s}}}R_{p{\bar{j}}k{\bar{l}}})\\&\qquad +g^{n{\bar{m}}}T_{mi}^s(\nabla _sH_{{\bar{j}}{\bar{n}}k{\bar{l}}}+T_{{\bar{n}}{\bar{j}}}^{{\bar{p}}}R_{s{\bar{p}}k{\bar{l}}}+T_{{\bar{n}}{\bar{j}}}^{p}H_{spk{\bar{l}}})\\&\qquad +g^{m{\bar{n}}}(R_{i{\bar{j}}m}^{\quad \,\,r}R_{r{\bar{n}}k{\bar{l}}}+R_{m{\bar{j}}k}^{\quad \,\,s}R_{i{\bar{n}}s{\bar{l}}}-R_{m{\bar{j}}r{\bar{l}}}R_{i{\bar{n}}k}^{\quad \,\,r})-g^{m{\bar{n}}}R_{m{\bar{j}}\,\,\,\,{\bar{n}}}^{\quad \,\,{\bar{s}}}R_{i{\bar{s}}k{\bar{l}}}\\&\qquad +g^{m{\bar{n}}}(\nabla _m\nabla _iH_{{\bar{j}}{\bar{n}}k{\bar{l}}}+\nabla _mT_{{\bar{n}}{\bar{j}}}^{{\bar{s}}}R_{i{\bar{s}}k{\bar{l}}}+\nabla _mT_{{\bar{n}}{\bar{j}}}^sH_{isk{\bar{l}}}+T_{{\bar{n}}{\bar{j}}}^s\nabla _mH_{isk{\bar{l}}}\\&\qquad +\Gamma _{{\bar{j}}i}^s\Gamma _{ms}^pR_{p{\bar{n}}k{\bar{l}}} +\Gamma _{{\bar{j}}i}^s\Gamma _{mk}^{p}R_{s{\bar{n}}p{\bar{l}}} +\Gamma _{{\bar{j}}i}^s\Gamma _{m{\bar{l}}}^{{\bar{p}}}R_{s{\bar{n}}k{\bar{p}}} +\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{mk}^{p}R_{i{\bar{s}}p{\bar{l}}}\\&\qquad +\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{m{\bar{l}}}^{{\bar{p}}}R_{i{\bar{s}}k{\bar{p}}} +\Gamma _{{\bar{j}}k}^s\Gamma _{mi}^{p}R_{p{\bar{n}}s{\bar{l}}} +\Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}s{\bar{l}}}\\&\qquad +\Gamma _{{\bar{j}}k}^s\Gamma _{ms}^{p}R_{i{\bar{n}}p{\bar{l}}} +\Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}s{\bar{p}}} +\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{mi}^{p}R_{p{\bar{n}}k{\bar{s}}} +\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{s}}}\\&\qquad +\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{mk}^{p}R_{i{\bar{n}}p{\bar{s}}} +\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{s}}}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}} +B_{{\bar{j}}m}^{s}\Gamma _{si}^pR_{p{\bar{n}}k{\bar{l}}} +B_{{\bar{j}}m}^{s}\Gamma _{sk}^pR_{i{\bar{n}}p{\bar{l}}}\\&\qquad +B_{{\bar{j}}m}^{s}\Gamma _{s{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}} +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}i}^pR_{p{\bar{n}}k{\bar{l}}} +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}}\\&\qquad +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^pR_{i{\bar{n}}p{\bar{l}}} +B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}} \\&\qquad -\Gamma _{mi}^s\Gamma _{{\bar{j}}s}^pR_{p{\bar{n}}k{\bar{l}}} -\Gamma _{mi}^s\Gamma _{{\bar{j}}k}^{p}R_{s{\bar{n}}p{\bar{l}}} -\Gamma _{mi}^s\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}R_{s{\bar{n}}k{\bar{p}}} -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}}\\&\qquad -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}k}^{p}R_{i{\bar{s}}p{\bar{l}}} -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}R_{i{\bar{s}}k{\bar{p}}} -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}i}^{p}R_{p{\bar{n}}s{\bar{l}}} -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}s{\bar{l}}}\\&\qquad -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}s}^{p}R_{i{\bar{n}}p{\bar{l}}} -\Gamma _{mk}^{s}\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}R_{i{\bar{n}}s{\bar{p}}} -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}i}^{p}R_{p{\bar{n}}k{\bar{s}}} -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{s}}}\\&\qquad -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}k}^{p}R_{i{\bar{n}}p{\bar{s}}} -\Gamma _{m{\bar{l}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}}\\&\qquad +\nabla _{{\bar{j}}}\nabla _{{\bar{n}}}H_{mik{\bar{l}}}+\nabla _{{\bar{j}}}T_{mi}^sR_{s{\bar{n}}k{\bar{l}}}-\nabla _{{\bar{j}}}B_{mi}^{{\bar{s}}}H_{{\bar{s}}{\bar{n}}k{\bar{l}}}-B_{mi}^{{\bar{s}}}\nabla _{{\bar{j}}}H_{{\bar{s}}{\bar{n}}k{\bar{l}}})\\&\qquad -g^{m{\bar{n}}}(B_{mi}^{{\bar{q}}}B_{{\bar{q}}{\bar{j}}}^s-\nabla _{{\bar{j}}}T_{im}^s+\Gamma _{mp}^s\Gamma _{{\bar{j}}i}^p-\Gamma _{{\bar{j}}p}^s\Gamma _{mi}^p-B_{m{\bar{j}}}^p\Gamma _{pi}^s+B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}i}^s)R_{s{\bar{n}}k{\bar{l}}}\\&\qquad -g^{m{\bar{n}}}(\Gamma _{mp}^s\Gamma _{{\bar{j}}k}^p-\Gamma _{{\bar{j}}p}^s\Gamma _{mk}^p-B_{m{\bar{j}}}^p\Gamma _{pk}^s+B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}k}^s)R_{i{\bar{n}}s{\bar{l}}}\\&\qquad +g^{m{\bar{n}}}(\Gamma _{{\bar{j}}{\bar{p}}}^{{\bar{s}}}\Gamma _{m{\bar{n}}}^{{\bar{p}}}-\Gamma _{m{\bar{p}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}-B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}{\bar{n}}}^{{\bar{s}}}+B_{m{\bar{j}}}^{p}\Gamma _{p{\bar{n}}}^{{\bar{s}}})R_{i{\bar{s}}k{\bar{l}}}\\&\qquad +g^{m{\bar{n}}}(\Gamma _{{\bar{j}}{\bar{p}}}^{{\bar{s}}}\Gamma _{m{\bar{l}}}^{{\bar{p}}}-\Gamma _{m{\bar{p}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}-B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}{\bar{l}}}^{{\bar{s}}}+B_{m{\bar{j}}}^{p}\Gamma _{p{\bar{l}}}^{{\bar{s}}})R_{i{\bar{n}}k{\bar{s}}}, \end{aligned}$$

where we have used that

$$\begin{aligned}&\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{mi}^{p}R_{p{\bar{s}}k{\bar{l}}} -\Gamma _{mi}^{s}\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}R_{s{\bar{p}}k{\bar{l}}}=0, \\&\Gamma _{{\bar{j}}i}^{{\bar{s}}}\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{s{\bar{p}}k{\bar{l}}} -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}i}^{p}R_{p{\bar{s}}k{\bar{l}}}=0, \end{aligned}$$

and

$$\begin{aligned}&g^{m{\bar{n}}}(\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{s}}}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}}+B_{{\bar{j}}m}^{s}\Gamma _{s{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}})\\&\quad =\Gamma _{{\bar{j}}{\bar{m}}}^{{\bar{s}}}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}}+B_{{\bar{j}}m}^{s}\Gamma _{s{\bar{m}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}}\\&\quad =-\Gamma _{{\bar{j}}s}^{m}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}}+B_{{\bar{j}}m}^{s}\Gamma _{s{\bar{m}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{l}}}=0. \end{aligned}$$

Combining this formula and the terms from \(-Q^7-Q^8+BT'+{\bar{Z}}(T')\) gives the result. \(\square \)

Lemma 3.7

Along the AHCF, we have the following evolution equation for the first Chern–Ricci curvature.

$$\begin{aligned}&\frac{\partial }{\partial t}P_{i{\bar{j}}}=\Delta P_{i{\bar{j}}} +g^{r{\bar{s}}}(T_{{\bar{s}}{\bar{j}}}^{{\bar{q}}}\nabla _rP_{i{\bar{q}}}+T_{ri}^p\nabla _{{\bar{s}}}P_{p{\bar{j}}}+T_{ri}^pT_{{\bar{s}}{\bar{j}}}^{{\bar{q}}}P_{p{\bar{q}}}+R_{i{\bar{j}}r}^{\quad \,\,p}P_{p{\bar{s}}})\nonumber \\&\qquad -g^{r{\bar{s}}}g^{m{\bar{n}}}R_{m{\bar{j}}r{\bar{n}}}P_{i{\bar{s}}}+g^{m{\bar{n}}}T_{mi}^s(\nabla _sR_{{\bar{j}}{\bar{n}}}+T_{{\bar{n}}{\bar{j}}}^pR_{sp})\nonumber \\&\qquad +g^{m{\bar{n}}}(\nabla _m\nabla _iR_{{\bar{j}}{\bar{n}}}+\nabla _mT_{{\bar{n}}{\bar{j}}}^{{\bar{s}}}\cdot P_{i{\bar{s}}}+\nabla _mT_{{\bar{n}}{\bar{j}}}^s\cdot R_{is}+T_{{\bar{n}}{\bar{j}}}^s\nabla _mR_{is}\nonumber \\&\qquad +\Gamma _{{\bar{j}}i}^s\Gamma _{ms}^pP_{p{\bar{n}}}+B_{{\bar{j}}m}^s\Gamma _{si}^pP_{p{\bar{n}}}+B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}i}^pP_{p{\bar{n}}}+B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{n}}}^{{\bar{p}}}P_{i{\bar{p}}}-\Gamma _{mi}^s\Gamma _{{\bar{j}}s}^pP_{p{\bar{n}}} \nonumber \\&\qquad -\Gamma _{m{\bar{n}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{p}}}P_{i{\bar{p}}} +\nabla _{{\bar{j}}}\nabla _{{\bar{n}}}R_{mi}+\nabla _{{\bar{j}}}T_{mi}^s\cdot P_{s{\bar{n}}}-\nabla _{{\bar{j}}}B_{mi}^{{\bar{s}}}\cdot R_{{\bar{s}}{\bar{n}}}-B_{mi}^{{\bar{s}}}\nabla _{{\bar{j}}}R_{{\bar{s}}{\bar{n}}}\nonumber \\&\qquad -g^{m{\bar{n}}}(B_{mi}^{{\bar{q}}}B_{{\bar{q}}{\bar{j}}}^s-\nabla _{{\bar{j}}}T_{im}^s+\Gamma _{mp}^s\Gamma _{{\bar{j}}i}^p-\Gamma _{{\bar{j}}p}^s\Gamma _{mi}^p-B_{m{\bar{j}}}^p\Gamma _{pi}^s+B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}i}^s)P_{s{\bar{n}}}\nonumber \\&\qquad -g^{m{\bar{n}}}g^{k{\bar{l}}}(\Gamma _{mp}^s\Gamma _{{\bar{j}}k}^p-\Gamma _{{\bar{j}}p}^s\Gamma _{mk}^p-B_{m{\bar{j}}}^p\Gamma _{pk}^s+B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}k}^s)R_{i{\bar{n}}s{\bar{l}}}\nonumber \\&\qquad +g^{m{\bar{n}}}(\Gamma _{{\bar{j}}{\bar{p}}}^{{\bar{s}}}\Gamma _{m{\bar{n}}}^{{\bar{p}}}-\Gamma _{m{\bar{p}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{n}}}^{{\bar{p}}}-B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}{\bar{n}}}^{{\bar{s}}}+B_{m{\bar{j}}}^{p}\Gamma _{p{\bar{n}}}^{{\bar{s}}})P_{i{\bar{s}}}\nonumber \\&\qquad +g^{m{\bar{n}}}g^{k{\bar{l}}}(\Gamma _{{\bar{j}}{\bar{p}}}^{{\bar{s}}}\Gamma _{m{\bar{l}}}^{{\bar{p}}}-\Gamma _{m{\bar{p}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{l}}}^{{\bar{p}}}-B_{{\bar{j}}m}^{{\bar{p}}}\Gamma _{{\bar{p}}{\bar{l}}}^{{\bar{s}}}+B_{m{\bar{j}}}^{p}\Gamma _{p{\bar{l}}}^{{\bar{s}}})R_{i{\bar{n}}k{\bar{s}}}\nonumber \\&\qquad +g^{k{\bar{l}}}\nabla _{{\bar{j}}}\nabla _{i}(Q^7+Q^8-BT'-{\bar{Z}}(T'))_{k{\bar{l}}}\nonumber \\&\quad =\Delta P_{i{\bar{j}}} +g^{r{\bar{s}}}(T_{{\bar{s}}{\bar{j}}}^{{\bar{q}}}\nabla _rP_{i{\bar{q}}}+T_{ri}^p\nabla _{{\bar{s}}}P_{p{\bar{j}}}+T_{ri}^pT_{{\bar{s}}{\bar{j}}}^{{\bar{q}}}P_{p{\bar{q}}}+R_{i{\bar{j}}r}^{\quad \,\,p}P_{p{\bar{s}}})+E_{i{\bar{j}}}, \end{aligned}$$
(3.3)

where we put

$$\begin{aligned} E_{i{\bar{j}}}:=g^{k{\bar{l}}}E_{i{\bar{j}}k{\bar{l}}}. \end{aligned}$$
(3.4)

Proof

We compute that

$$\begin{aligned}&\frac{\partial }{\partial t}P_{i{\bar{j}}}(g(t))=\frac{\partial }{\partial t}(g^{k{\bar{l}}}R_{i{\bar{j}}k{\bar{l}}})\\&\quad =(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))^{k{\bar{l}}}R_{i{\bar{j}}k{\bar{l}}}+g^{k{\bar{l}}}\frac{\partial }{\partial t}R_{i{\bar{j}}k{\bar{l}}}. \end{aligned}$$

Choosing \(t=t_0\) and a local unitary (1, 0)-frame \(\{Z_r\}\) around a fixed point \(p\in M\) with respect to \(g(t_0)\). Then using the local \(g(t_0)\)-unitary frame, \(g(t_0)^{m{\bar{n}}}=\delta _{mn}\),

$$\begin{aligned} \Gamma _{m{\bar{j}}}^{{\bar{p}}}=-\Gamma _{mp}^j,\quad \Gamma _{{\bar{i}}{\bar{j}}}^{{\bar{s}}}=-\Gamma _{{\bar{i}}s}^j. \end{aligned}$$

In the evolution equation in Lemma 3.6 (3.2), we see that

$$\begin{aligned}&g^{k{\bar{l}}}g^{r{\bar{s}}}(R_{r{\bar{j}}k}^{\quad \,\,p}R_{i{\bar{s}}p{\bar{l}}}-R_{r{\bar{j}}p{\bar{l}}}R_{i{\bar{s}}k}^{\quad \,\,p})=0,\\&g^{k{\bar{l}}}R_{i{\bar{j}}k}^{\quad \, r}(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{r{\bar{l}}}\\&\quad =g^{k{\bar{l}}}g^{r{\bar{s}}}R_{i{\bar{j}}k{\bar{s}}}(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{r{\bar{l}}}\\&\quad =(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))^{k{\bar{l}}}R_{i{\bar{j}}k{\bar{l}}},\\&\quad \Gamma _{{\bar{j}}i}^s\Gamma _{mk}^pR_{s{\bar{m}}p{\bar{k}}}+\Gamma _{{\bar{j}}i}^s\Gamma _{m{\bar{k}}}^{{\bar{p}}}R_{s{\bar{m}}k{\bar{p}}}=\Gamma _{{\bar{j}}i}^s\Gamma _{mk}^pR_{s{\bar{m}}p{\bar{k}}}-\Gamma _{{\bar{j}}i}^s\Gamma _{mp}^{k}R_{s{\bar{m}}k{\bar{p}}}=0, \\&\quad \Gamma _{{\bar{j}}{\bar{m}}}^{{\bar{s}}}\Gamma _{mk}^pR_{i{\bar{s}}p{\bar{k}}}+\Gamma _{{\bar{j}}{\bar{m}}}^{{\bar{s}}}\Gamma _{m{\bar{k}}}^{{\bar{p}}}R_{i{\bar{s}}k{\bar{p}}}=\Gamma _{{\bar{j}}{\bar{m}}}^{{\bar{s}}}\Gamma _{mk}^pR_{i{\bar{s}}p{\bar{k}}}-\Gamma _{{\bar{j}}{\bar{m}}}^{{\bar{s}}}\Gamma _{mp}^{k}R_{i{\bar{s}}k{\bar{p}}}=0, \\&\quad \Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{k}}}^{{\bar{p}}}R_{i{\bar{m}}s{\bar{p}}}+\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{s}}}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{m}}k{\bar{p}}}=\Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{k}}}^{{\bar{p}}}R_{i{\bar{m}}s{\bar{p}}}-\Gamma _{{\bar{j}}s}^{k}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{m}}k{\bar{p}}}=0, \\&\quad \Gamma _{{\bar{j}}k}^s\Gamma _{mi}^{p}R_{p{\bar{m}}s{\bar{k}}}+\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{s}}}\Gamma _{mi}^{p}R_{p{\bar{m}}k{\bar{s}}}=\Gamma _{{\bar{j}}k}^s\Gamma _{mi}^{p}R_{p{\bar{m}}s{\bar{k}}}-\Gamma _{{\bar{j}}s}^{k}\Gamma _{mi}^{p}R_{p{\bar{m}}k{\bar{s}}}=0, \\&\quad \Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}s{\bar{k}}}+\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{s}}}\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{s}}}=\Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}s{\bar{k}}}-\Gamma _{{\bar{j}}s}^{k}\Gamma _{m{\bar{n}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{s}}}=0, \\&\quad \Gamma _{{\bar{j}}k}^s\Gamma _{ms}^{p}R_{i{\bar{m}}p{\bar{k}}}+\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{s}}}\Gamma _{mk}^{p}R_{i{\bar{m}}p{\bar{s}}}=\Gamma _{{\bar{j}}k}^s\Gamma _{ms}^{p}R_{i{\bar{m}}p{\bar{k}}}-\Gamma _{{\bar{j}}s}^{k}\Gamma _{mk}^{p}R_{i{\bar{m}}p{\bar{s}}}=0, \\&\quad \Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{k}}}^{{\bar{p}}}R_{i{\bar{m}}s{\bar{p}}}+\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{s}}}\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{m}}k{\bar{p}}}=\Gamma _{{\bar{j}}k}^s\Gamma _{m{\bar{k}}}^{{\bar{p}}}R_{i{\bar{m}}s{\bar{p}}}-\Gamma _{{\bar{j}}s}^k\Gamma _{m{\bar{s}}}^{{\bar{p}}}R_{i{\bar{m}}k{\bar{p}}}=0, \\&\quad B_{{\bar{j}}m}^s\Gamma _{sk}^{p}R_{i{\bar{m}}p{\bar{k}}}+B_{{\bar{j}}m}^{s}\Gamma _{s{\bar{k}}}^{{\bar{p}}}R_{i{\bar{m}}k{\bar{p}}}=B_{{\bar{j}}m}^s\Gamma _{sk}^{p}R_{i{\bar{m}}p{\bar{k}}}-B_{{\bar{j}}m}^{s}\Gamma _{sp}^{k}R_{i{\bar{m}}k{\bar{p}}}=0, \\&\quad B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^{p}R_{i{\bar{m}}p{\bar{k}}}+B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}{\bar{k}}}^{{\bar{p}}}R_{i{\bar{m}}k{\bar{p}}}=B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}k}^{p}R_{i{\bar{m}}p{\bar{k}}}-B_{{\bar{j}}m}^{{\bar{s}}}\Gamma _{{\bar{s}}p}^{k}R_{i{\bar{m}}k{\bar{p}}}=0, \\&\quad \Gamma _{mi}^s\Gamma _{{\bar{j}}k}^pR_{s{\bar{m}}p{\bar{k}}}+\Gamma _{mi}^s\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{p}}}R_{s{\bar{m}}k{\bar{p}}}=\Gamma _{mi}^s\Gamma _{{\bar{j}}k}^pR_{s{\bar{m}}p{\bar{k}}}-\Gamma _{mi}^s\Gamma _{{\bar{j}}p}^{k}R_{s{\bar{m}}k{\bar{p}}}=0, \\&\quad \Gamma _{m{\bar{m}}}^{{\bar{s}}}\Gamma _{{\bar{j}}k}^pR_{i{\bar{s}}p{\bar{k}}}+\Gamma _{m{\bar{m}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{p}}}R_{i{\bar{s}}k{\bar{p}}}=\Gamma _{m{\bar{m}}}^{{\bar{s}}}\Gamma _{{\bar{j}}k}^pR_{i{\bar{s}}p{\bar{k}}}-\Gamma _{m{\bar{m}}}^{{\bar{s}}}\Gamma _{{\bar{j}}p}^{k}R_{i{\bar{s}}k{\bar{p}}}=0, \\&\quad \Gamma _{mk}^{s}\Gamma _{{\bar{j}}i}^pR_{p{\bar{n}}s{\bar{k}}}+\Gamma _{m{\bar{k}}}^{{\bar{s}}}\Gamma _{{\bar{j}}i}^{p}R_{p{\bar{m}}k{\bar{s}}}=\Gamma _{mk}^{s}\Gamma _{{\bar{j}}i}^pR_{p{\bar{n}}s{\bar{k}}}-\Gamma _{ms}^{k}\Gamma _{{\bar{j}}i}^{p}R_{p{\bar{m}}k{\bar{s}}}=0, \\&\quad \Gamma _{mk}^{s}\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{p}}}R_{i{\bar{m}}s{\bar{p}}}+\Gamma _{m{\bar{k}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}}=\Gamma _{mk}^{s}\Gamma _{{\bar{j}}{\bar{k}}}^{{\bar{p}}}R_{i{\bar{m}}s{\bar{p}}}-\Gamma _{ms}^{k}\Gamma _{{\bar{j}}{\bar{s}}}^{{\bar{p}}}R_{i{\bar{n}}k{\bar{p}}}=0, \\&\quad \Gamma _{mk}^{s}\Gamma _{{\bar{j}}s}^pR_{i{\bar{m}}p{\bar{k}}}+\Gamma _{m{\bar{k}}}^{{\bar{s}}}\Gamma _{{\bar{j}}k}^{p}R_{i{\bar{m}}p{\bar{s}}}=\Gamma _{mk}^{s}\Gamma _{{\bar{j}}s}^pR_{i{\bar{m}}p{\bar{k}}}-\Gamma _{ms}^{k}\Gamma _{{\bar{j}}k}^{p}R_{i{\bar{m}}p{\bar{s}}}=0, \\&\quad \Gamma _{mk}^{s}\Gamma _{{\bar{j}}{\bar{m}}}^{{\bar{p}}}R_{i{\bar{p}}s{\bar{k}}}+\Gamma _{m{\bar{k}}}^{{\bar{s}}}\Gamma _{{\bar{j}}{\bar{m}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{s}}}=\Gamma _{mk}^{s}\Gamma _{{\bar{j}}{\bar{m}}}^{{\bar{p}}}R_{i{\bar{p}}s{\bar{k}}}-\Gamma _{ms}^{k}\Gamma _{{\bar{j}}{\bar{m}}}^{{\bar{p}}}R_{i{\bar{p}}k{\bar{s}}}=0, \end{aligned}$$

which leads the desired result. \(\square \)

We need the estimate of the term \(\nabla _{{\bar{j}}}\nabla _{i}(Q^7+Q^8-BT'-{\bar{Z}}(T'))_{k{\bar{l}}}\) included in the terms \(E_{i{\bar{j}}k{\bar{l}}}\) and \(E_{i{\bar{j}}}\) for giving a proof of Theorem 1.1. From Lemma 3.3, we have that

$$\begin{aligned} {\bar{Z}}(T')= & {} {\bar{Z}}(\Gamma )+Z(B^\circ )+B^\circ *\Gamma +T'*{\bar{\Gamma }}+\Gamma *{\bar{\Gamma }}+B^\circ *T'+B^\circ *{\bar{T}}'\\&+T'*{\bar{T}}'+\Gamma *{\bar{T}}'+B^\circ *B^\circ +{\mathcal {O}}(Z(g))+{\mathcal {O}}({\bar{Z}}(g)) \end{aligned}$$

From the definitions of \(BT'\), \({\bar{Z}}(T')\) and the equality above, we have

$$\begin{aligned}&(-Q^7-Q^8+BT'+{\bar{Z}}(T'))_{i{\bar{j}}}\\&\quad =-B_{ir}^{{\bar{s}}}B_{{\bar{r}}{\bar{k}}}^lg_{k{\bar{s}}}g_{l{\bar{j}}}-B_{ir}^{{\bar{s}}}B_{{\bar{j}}{\bar{k}}}^lg_{k{\bar{s}}}g_{l{\bar{r}}}+B_{{\bar{r}}p}^jT_{ir{\bar{p}}}+B_{{\bar{p}}i}^rT_{pr{\bar{j}}} +B_{{\bar{r}}r}^pT_{pi{\bar{j}}}+B_{{\bar{j}}i}^rw_r\\&\qquad -Z_{{\bar{r}}}(T_{ri}^s)g_{s{\bar{j}}}-Z_{{\bar{j}}}(w_i)-g^{p{\bar{q}}}T_{pi}^rZ_{{\bar{j}}}(g_{r{\bar{q}}})\\&\quad =B^\circ *B^\circ +B^\circ *T'+B^\circ *{\bar{T}}'+T'*{\bar{\Gamma }}+{\bar{T}}'*\Gamma +T'*{\bar{T}}'+\Gamma *{\bar{\Gamma }}\\&\qquad +\Gamma *B^\circ +{\bar{Z}}(\Gamma )+Z(B^\circ )+{\mathcal {O}}(Z(g))+{\mathcal {O}}({\bar{Z}}(g)) \end{aligned}$$

and then we obtain

$$\begin{aligned}&-\nabla _{{\bar{j}}}\nabla _i(-Q^7-Q^8+BT'+{\bar{Z}}(T'))_{k{\bar{l}}}\\&\quad ={\bar{\nabla }}\nabla B^\circ *B^\circ +\nabla B^\circ *{\bar{\nabla }} B^\circ +{\bar{\nabla }}\nabla B^\circ *T'+\nabla B^\circ *{\bar{\nabla }} T' +B^\circ *{\bar{\nabla }}\nabla T'\\&\qquad +{\bar{\nabla }}\nabla B^\circ *{\bar{T}}' +\nabla B^\circ *{\bar{\nabla }} {\bar{T}}'+{\bar{\nabla }} B^\circ *\nabla {\bar{T}}' +B^\circ *{\bar{\nabla }} \nabla {\bar{T}}'+{\bar{\nabla }} \nabla T'*{\bar{\Gamma }}\\&\qquad +\nabla T'*{\bar{\nabla }} {\bar{\Gamma }}+{\bar{\nabla }} T'*\nabla {\bar{\Gamma }} +T'*{\bar{\nabla }} \nabla {\bar{\Gamma }}+{\bar{\nabla }} \nabla {\bar{T}}'*\Gamma +\nabla {\bar{T}}'*{\bar{\nabla }} \Gamma \\&\qquad +{\bar{\nabla }} {\bar{T}}'*\nabla \Gamma +{\bar{T}}'*{\bar{\nabla }} \nabla \Gamma +{\bar{\nabla }} \nabla T'*{\bar{T}}'+\nabla T'*{\bar{\nabla }} {\bar{T}}' +{\bar{\nabla }} T'*\nabla {\bar{T}}'\\&\qquad +T'*{\bar{\nabla }} \nabla {\bar{T}}' +{\bar{\nabla }} \nabla \Gamma *{\bar{\Gamma }}+\nabla \Gamma *{\bar{\nabla }} {\bar{\Gamma }} +{\bar{\nabla }} \Gamma * \nabla {\bar{\Gamma }}+\Gamma *{\bar{\nabla }} \nabla {\bar{\Gamma }} +{\bar{\nabla }} \nabla \Gamma *B^\circ \\&\qquad +\nabla \Gamma *{\bar{\nabla }} B^\circ +{\bar{\nabla }} \Gamma *\nabla B^\circ +\Gamma *{\bar{\nabla }} \nabla B^\circ \\&\qquad +{\bar{\nabla }} \nabla {\bar{Z}}(\Gamma )+{\bar{\nabla }} \nabla {\bar{Z}}(B^\circ ) +{\bar{\nabla }}\nabla {\mathcal {O}}(Z(g))+{\bar{\nabla }}\nabla {\mathcal {O}}({\bar{Z}}(g)). \end{aligned}$$

Hence, by applying Lemmas 3.1, 3.4, we have the following estimate since we may assume that the quantities \(|T'|^2_{C^0(g(t))}\), \(|\nabla T'|_{C^0(g(t))}\) and \(|R|_{C^0(g(t))}\) are uniformly bounded on \([0,\tau )\) for any \(0<\tau<\tau _{\max }<\infty \) (see Remark 1.1),

$$\begin{aligned}&\Big |\nabla _{{\bar{j}}}\nabla _i(-Q^7-Q^8+BT'+{\bar{Z}}(T'))_{k{\bar{l}}}\Big |_{g(t)}\nonumber \\&\quad =\Big |{\bar{\nabla }}\nabla B^\circ *B^\circ +\nabla B^\circ *{\bar{\nabla }} B^\circ +{\bar{\nabla }}\nabla B^\circ *T'+\nabla B^\circ *{\bar{\nabla }} T' \nonumber \\&\qquad +B^\circ *{\bar{\nabla }}\nabla T'+{\bar{\nabla }}\nabla B^\circ *{\bar{T}}'+\nabla B^\circ *{\bar{\nabla }} {\bar{T}}'+{\bar{\nabla }} B^\circ *\nabla {\bar{T}}' +B^\circ *{\bar{\nabla }} \nabla {\bar{T}}' \nonumber \\&\qquad +{\bar{\nabla }} \nabla T'*{\bar{\Gamma }}+\nabla T'*{\bar{\nabla }} {\bar{\Gamma }}+{\bar{\nabla }} T'*\nabla {\bar{\Gamma }} +T'*{\bar{\nabla }} \nabla {\bar{\Gamma }}+{\bar{\nabla }} \nabla {\bar{T}}'*\Gamma \nonumber \\&\qquad +\nabla {\bar{T}}'*{\bar{\nabla }} \Gamma +{\bar{\nabla }} {\bar{T}}'*\nabla \Gamma +{\bar{T}}'*{\bar{\nabla }} \nabla \Gamma +{\bar{\nabla }} \nabla T'*{\bar{T}}'+\nabla T'*{\bar{\nabla }} {\bar{T}}' \nonumber \\&\qquad +{\bar{\nabla }} T'*\nabla {\bar{T}}' +T'*{\bar{\nabla }} \nabla {\bar{T}}'+{\bar{\nabla }} \nabla \Gamma *{\bar{\Gamma }}+\nabla \Gamma *{\bar{\nabla }} {\bar{\Gamma }}+{\bar{\nabla }} \Gamma * \nabla {\bar{\Gamma }}+\Gamma *{\bar{\nabla }} \nabla {\bar{\Gamma }}\nonumber \\&\qquad +{\bar{\nabla }} \nabla \Gamma *B^\circ +\nabla \Gamma *{\bar{\nabla }} B^\circ +{\bar{\nabla }} \Gamma *\nabla B^\circ +\Gamma *{\bar{\nabla }} \nabla B^\circ +{\bar{\nabla }} \nabla {\bar{Z}}(\Gamma )\nonumber \\&\qquad +{\bar{\nabla }} \nabla {\bar{Z}}(B^\circ )+{\bar{\nabla }}\nabla {\mathcal {O}}(Z(g))+{\bar{\nabla }}\nabla {\mathcal {O}}({\bar{Z}}(g))\Big |_{g(t)}\nonumber \\&\quad \le C'R'+\Big |{\bar{\nabla }}\nabla {\mathcal {O}}(Z(g))+{\bar{\nabla }}\nabla {\mathcal {O}}({\bar{Z}}(g))\Big |_{g(t)} \end{aligned}$$
(3.5)

for some uniform constant \(C'>0\), where \(R'\) is the time dependent tensor field defined by \(R'_{i{\bar{j}}k{\bar{l}}}:=g(t)_{i{\bar{j}}}g(t)_{k{\bar{l}}}\).

4 Preservation of Curvature Conditions

Remark 4.1

As we confirm in Remark 1.1, we may assume that the quantities \(|T'|^2_{C^0(g(t))}\), \(|\nabla T'|_{C^0(g(t))}\) and \(|R|_{C^0(g(t))}\) are uniformly bounded on \([0,\tau )\) for any \(0<\tau<\tau _{\max }<\infty \). We define

$$\begin{aligned} K_0:=\sup _{M\times \{0\}}(|R|_{g(0)}+|T'|^2_{g(0)}+|\nabla T'|_{g(0)}) \end{aligned}$$

and choose sufficiently large \(K\gg K_0\) so that \(\tau =K^{-1}<\tau _{\max }<\infty \) and

$$\begin{aligned} \sup _{M\times [0,\tau ]}(|R|_{g(t)}+|T'|^2_{g(t)}+|\nabla T'|_{g(t)})\le 2K_0. \end{aligned}$$
(4.1)

Proposition 4.1

Suppose that (MJg(t)) is a solution to the almost Hermitian curvature flow such that the initial metric \(g(0)=g_0\) has the Griffiths non-positive Chern curvature. There exist \(\tau >0\) and \(K>0\) such that for all \(t\in [0,\tau ]\), \(R=R(g(t))\) satisfies the following conditions:

  1. (i)

    \(P\le 0;\)

  2. (ii)

    \(|R_{u{\bar{v}}x{\bar{x}}}|^2\le (1+Kt)P_{u{\bar{u}}}P_{v{\bar{v}}}\) for all x, u, \(v\in T^{1,0}M\), \(|x|=1\).

We consider \(R^{\varepsilon }_{i{\bar{j}}k{\bar{l}}}:=R_{i{\bar{j}}k{\bar{l}}}-\varepsilon B_{i{\bar{j}}k{\bar{l}}},\) where \(B_{i{\bar{j}}k{\bar{l}}}=g_{i{\bar{j}}}g_{k{\bar{l}}}+g_{i{\bar{l}}}g_{k{\bar{j}}}\) and \(\varepsilon \) is a sufficiently small real number. The result of Proposition 4.1 follows directly as a consequence of the following Lemma by letting \(\varepsilon \rightarrow 0\).

Lemma 4.1

Under the assumption of Proposition 4.1, there exist \(\tau >0\) and \(K>0\) such that for any sufficiently small \(\varepsilon >0\), and for any \(t\in [0,\tau ]\), the following hold.

  1. (i)

    \(P^{\varepsilon }_{i{\bar{j}}}<-\varepsilon e^{-Kt}g_{i{\bar{j}}}\), where we put \(P^{\varepsilon }_{i{\bar{j}}}:=g^{k{\bar{l}}}R^{\varepsilon }_{i{\bar{j}}k{\bar{l}}};\)

  2. (ii)

    \(|R^{\varepsilon }_{u{\bar{v}}x{\bar{x}}}|^2<(1+Kt)P^{\varepsilon }_{u{\bar{u}}}P^{\varepsilon }_{v{\bar{v}}}\) for all x, u, \(v\in T^{1,0}M\), \(|x|=1\).

We firstly show that \(g(0)=g_0\) satisfies the assumptions in Lemma 4.1.

Lemma 4.2

Under the assumption of Proposition 4.1, \(R^{\varepsilon }(g_0)\) satisfies for any sufficiently small \(\varepsilon >0\),

  1. (i)

    \(P^{\varepsilon }(g_0)_{i{\bar{j}}}<-\varepsilon (g_0)_{i{\bar{j}}};\)

  2. (ii)

    \(|R^{\varepsilon }(g_0)_{u{\bar{v}}x{\bar{x}}}|^2_{g_0}< P(g_0)^{\varepsilon }_{u{\bar{u}}}P(g_0)^{\varepsilon }_{v{\bar{v}}}\) for all x, u, \(v\in T^{1,0}M\), \(|x|=1\).

Proof

\(\mathrm{(i)}\) follows from that \(P(g_0)\le 0\) and \(g_0^{k{\bar{l}}}B(g_0)_{i{\bar{j}}k{\bar{l}}}=(n+1)(g_0)_{i{\bar{j}}}\). Next we show \(\mathrm{(ii)}\). For a fixed \(x\in T^{1,0}M\), since \(R^{\varepsilon }(g_0)_{i{\bar{j}}x{\bar{x}}}\) is almost Hermitian form, we may choose eigenvectors \(\{e_i\}_{i=1}^n\) such that \(R^{\varepsilon }(g_0)_{i{\bar{j}}x{\bar{x}}}=\lambda _i\delta _{ij}\), where \(\lambda _i<0\) since we have assumed that the initial metric \(g_0\) has the Griffiths non-positive Chern curvature. Hence, for \(u=\sum _{i=1}^nu^ie_i\) and \(v=\sum _{i=1}^nv^ie_i\),

$$\begin{aligned} |R^{\varepsilon }(g_0)_{u{\bar{v}}x{\bar{x}}}|_{g_0}^2= & {} \left| \sum _{i=1}^n\lambda _iu^iv^{{\bar{i}}}\right| _{g_0}^2\\\le & {} \left( \sum _{i=1}^n\lambda _i|u^i|_{g_0}^2\right) \left( \sum _{i=1}^n\lambda _i|v^i|^2_{g_0}\right) \\= & {} R^{\varepsilon }(g_0)_{u{\bar{u}}x{\bar{x}}}R^{\varepsilon }(g_0)_{v{\bar{v}}x{\bar{x}}}\\< & {} P^{\varepsilon }(g_0)_{u{\bar{u}}}P^{\varepsilon }(g_0)_{v{\bar{v}}}. \end{aligned}$$

\(\square \)

Proof of Lemma 4.1

As in Remark 4.1, we choose sufficiently large \(K\gg K_0\) so that \(\tau =K^{-1}<\tau _{\max }\) and satisfying (4.1). If conditions \(\mathrm{(i)}\), \(\mathrm{(ii)}\) are true on \([0,\tau ]\), then we are done. Let \(t_0\in (0,\tau ]\) be the first time such that one of conditions \(\mathrm{(i)}\), \(\mathrm{(ii)}\) fails. First, we assume that \(\mathrm{(i)}\) is true on \([0,t_0)\) and fails at \(t=t_0\). Then there exist \(p_0\in M\) and \(X_0\in T_{p_0}^{1,0}M\) with \(|X_0|_{g(t_0)}=1\) such that \(P^{\varepsilon }(g(t_0))_{X_0\bar{X_0}}=-\varepsilon e^{-Kt_0}g(t_0)_{X_0\bar{X_0}}=-\varepsilon e^{-Kt_0}.\) Moreover, for all \(p\in M\), \(t\in [0,t_0]\), \(Y, U, V\in T_p^{1,0}M\) with \(|Y|=1\),

$$\begin{aligned} P^{\varepsilon }_{Y{\bar{Y}}}\le -\varepsilon e^{-Kt}g_{Y{\bar{Y}}}=-\varepsilon e^{-Kt},\quad |R^{\varepsilon }_{U{\bar{V}}Y{\bar{Y}}}|^2<(1+Kt)P^{\varepsilon }_{U{\bar{U}}}P^{\varepsilon }_{V{\bar{V}}}. \end{aligned}$$

Using polarization as in [7, 8] and (4.1) to infer that for sufficiently small \(\varepsilon >0\), any \(e_k, e_l\in T^{1,0}_pM\) with unit 1 and \(e_i, e_j\in T^{1,0}_pM\) for all \(p\in M\),

$$\begin{aligned} |R_{i{\bar{j}}k{\bar{l}}}^{\varepsilon }|^2\le C_nP_{i{\bar{i}}}^{\varepsilon }P_{j{\bar{j}}}^{\varepsilon },\quad |R_{i{\bar{j}}k{\bar{l}}}^{\varepsilon }|^2\le C_nK_0|P_{i{\bar{i}}}^{\varepsilon }|. \end{aligned}$$

We consider

$$\begin{aligned} A_{i{\bar{j}}}:=P_{i{\bar{j}}}+\varepsilon (e^{-Kt}-(n+1))g_{i{\bar{j}}}=P^{\varepsilon }_{i{\bar{j}}}+\varepsilon e^{-Kt}g_{i{\bar{j}}}, \end{aligned}$$

which satisfies \(A(X_0,{\bar{X}}_0)=0\) and \(A(Y,{\bar{Y}})\le 0\) for all \(Y\in T^{1,0}_pM\), \(p\in M\). We may assume that \(|X_0|_{g(t_0)}=1\) by rescaling. We extend \(X_0\) locally to a vector field around \((p_0,t_0)\) such that at \((p_0,t_0)\),

$$\begin{aligned} \nabla _{{\bar{q}}}X^p=0,\quad \nabla _{p}X^q=T_{pl}^qX^l. \end{aligned}$$

Then \(A(X,{\bar{X}})\) locally defines a function and satisfies

$$\begin{aligned} \square A(X,{\bar{X}})\ge 0, \end{aligned}$$

where \(\square :=(\frac{\partial }{\partial t}-\Delta )\). At \((p_0,t_0)\), we have

$$\begin{aligned}&\frac{\partial }{\partial t}A(X,{\bar{X}})=\Big (\frac{\partial }{\partial t}A_{i{\bar{j}}}\Big )X^iX^{{\bar{j}}}+A_{i{\bar{j}}}\Big (\frac{\partial }{\partial t}X^i\cdot X^{{\bar{j}}}+X^i\frac{\partial }{\partial t} X^{{\bar{j}}}\Big )\nonumber \\&\quad =\Big (\frac{\partial }{\partial t}P_{i{\bar{j}}}-\varepsilon (e^{-Kt}-(n+1))(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{i{\bar{j}}}-\varepsilon Ke^{-Kt}g_{i{\bar{j}}}\Big )X^iX^{{\bar{j}}}\nonumber \\&\qquad +A_{i{\bar{j}}}\Big (\frac{\partial }{\partial t}X^iX^{{\bar{j}}}+X^i\frac{\partial }{\partial t}X^{{\bar{j}}}\Big )\nonumber \\&\quad \le \Big (\frac{\partial }{\partial t}P_{i{\bar{j}}}\Big )X^iX^{{\bar{j}}}-\frac{1}{2}\varepsilon Ke^{-Kt}, \end{aligned}$$
(4.2)

where we used (4.1) and the fact that for any \(Y\in T_{p_0}^{1,0}M\), \(A_{X_0{\bar{Y}}}=0\). Choosing a local unitary (1, 0)-frame with respect to \(g(t_0)\) around a point \(p_0\), we have \(g_{i{\bar{j}}}(t_0)=\delta _{i{\bar{j}}}\), \(Z_r(g_{i{\bar{j}}}(t_0))=0\) at \(p_0\). Then we obtain

$$\begin{aligned} \Delta A(X,{\bar{X}})= & {} \frac{1}{2}g^{r{\bar{s}}}(\nabla _r\nabla _{{\bar{s}}}+\nabla _{{\bar{s}}}\nabla _r)(A_{i{\bar{j}}}X^iX^{{\bar{j}}})\nonumber \\= & {} \Delta A_{i{\bar{j}}}\cdot X^iX^{{\bar{j}}}+A_{i{\bar{j}}}X^i\Delta X^{{\bar{j}}}+A_{i{\bar{j}}}X^{{\bar{j}}}\Delta X^i\nonumber \\&+\nabla _rA_{i{\bar{j}}}\nabla _{{\bar{r}}}X^i\cdot X^{{\bar{j}}}+\nabla _{{\bar{r}}}A_{i{\bar{j}}}\nabla _rX^i\cdot X^{{\bar{j}}}\nonumber \\&+\nabla _rA_{i{\bar{j}}}\cdot X^i\nabla _{{\bar{r}}}X^{{\bar{j}}}+\nabla _{{\bar{r}}}A_{i{\bar{j}}}\cdot X^i\nabla _rX^{{\bar{j}}}\nonumber \\&+A_{i{\bar{j}}}\nabla _rX^i\nabla _{{\bar{r}}}X^{{\bar{j}}}+A_{i{\bar{j}}}\nabla _{{\bar{r}}}X^i\nabla _rX^{{\bar{j}}}\nonumber \\= & {} \Delta P_{i{\bar{j}}}\cdot X^iX^{{\bar{j}}}+\nabla _rP_{i{\bar{j}}}\nabla _{{\bar{r}}}X^i\cdot X^{{\bar{j}}}+\nabla _{{\bar{r}}}P_{i{\bar{j}}}\nabla _rX^i\cdot X^{{\bar{j}}}\nonumber \\&+\nabla _rP_{i{\bar{j}}}\cdot X^i\nabla _{{\bar{r}}}X^{{\bar{j}}}+\nabla _{{\bar{r}}}P_{i{\bar{j}}}\cdot X^i\nabla _rX^{{\bar{j}}}\nonumber \\&+A_{i{\bar{j}}}\nabla _rX^i\nabla _{{\bar{r}}}X^{{\bar{j}}}+A_{i{\bar{j}}}\nabla _{{\bar{r}}}X^i\nabla _rX^{{\bar{j}}}, \end{aligned}$$
(4.3)

where we used \(\nabla g=0\) and the fact that for any \(Y\in T_{p_0}^{1,0}M\), \(A_{X_0{\bar{Y}}}=0\) for the terms involve \(A(\Delta X,{\bar{X}})\) or its conjugate.

By combining them, we obtain by using the formula (3.3), and applying (4.2), (4.3),

$$\begin{aligned}&\square A(X,{\bar{X}})\le X^i X^{{\bar{j}}}\square P_{i{\bar{j}}}-\frac{1}{2}\varepsilon Ke^{-Kt}\\&\qquad -g^{r{\bar{s}}}(\nabla _rP_{i{\bar{j}}}\nabla _{{\bar{s}}}X^i\cdot X^{{\bar{j}}}+\nabla _{{\bar{s}}}P_{i{\bar{j}}}\nabla _rX^i\cdot X^{{\bar{j}}} +\nabla _rP_{i{\bar{j}}}X^i\nabla _{{\bar{s}}}X^{{\bar{j}}}+\nabla _{{\bar{s}}}P_{i{\bar{j}}}X^i\nabla _rX^{{\bar{j}}})\\&\qquad -g^{r{\bar{s}}}(P_{i{\bar{j}}}\nabla _rX^i\nabla _{{\bar{s}}}X^{{\bar{j}}}+P_{i{\bar{j}}}\nabla _{{\bar{s}}}X^i\nabla _rX^{{\bar{j}}})\\&\qquad -\varepsilon (e^{-Kt}-(n+1))g^{r{\bar{s}}}(\nabla _rX^i\nabla _{{\bar{s}}}X^{{\bar{j}}}+\nabla _{{\bar{s}}}X^i\nabla _rX^{{\bar{j}}})\\&\quad =X^i X^{{\bar{j}}}g^{r{\bar{s}}}(T_{{\bar{s}}{\bar{j}}}^{{\bar{q}}}\nabla _rP_{i{\bar{q}}}+T_{ri}^p\nabla _{{\bar{s}}}P_{p{\bar{j}}}+T_{ri}^pT_{{\bar{s}}{\bar{j}}}^{{\bar{q}}}P_{p{\bar{q}}}+R_{i{\bar{j}}r}^{\quad \,\,p}P_{p{\bar{s}}})+E_{i{\bar{j}}}X^i X^{{\bar{j}}}-\frac{1}{2}\varepsilon Ke^{-Kt}\\&\qquad -g^{r{\bar{s}}}(\nabla _{{\bar{s}}}P_{i{\bar{j}}}\cdot T_{rk}^iX^kX^{{\bar{j}}} +\nabla _rP_{i{\bar{j}}}\cdot X^iT_{{\bar{s}}{\bar{l}}}^{{\bar{j}}}X^{{\bar{l}}}+P_{i{\bar{j}}}T_{rk}^iX^kT_{{\bar{s}}{\bar{l}}}^{{\bar{j}}}X^{{\bar{l}}})\\&\qquad -\varepsilon (e^{-Kt}-(n+1))g^{r{\bar{s}}}T_{rk}^iX^kT_{{\bar{s}}{\bar{l}}}^{{\bar{j}}}X^{{\bar{l}}}\\&\quad =g^{r{\bar{s}}}R_{i{\bar{j}}r}^{\quad \,\,p}X^i X^{{\bar{j}}}P_{p{\bar{s}}}+E_{i{\bar{j}}}X^i X^{{\bar{j}}}-\frac{1}{2}\varepsilon Ke^{-Kt}-\varepsilon (e^{-Kt}-(n+1))g^{r{\bar{s}}}T_{rk}^iX^kT_{{\bar{s}}{\bar{l}}}^{{\bar{j}}}X^{{\bar{l}}}. \end{aligned}$$

We have the following estimate from the estiamte (3.5),

$$\begin{aligned} \Big |\nabla _{{\bar{j}}}\nabla _i(-Q^7-Q^8+BT'+{\bar{Z}}(T'))_{k{\bar{l}}}\Big |_{g(t)}\le C'R' \end{aligned}$$
(4.4)

for a uniform constant \(C'>0\), where we used that \(Z(g)=0\) with a local unitary frame.

As \(K\gg K_0\) sufficiently large, since we may assume that the quantities \(|T'|^2_{C^0(g(t))}\), \(|\nabla T'|_{C^0(g(t))}\) and \(|R|_{C^0(g(t))}\) are uniformly bounded on \([0,\tau ]\), we have the estimates in Lemma 3.1, 3.4, 3.5, and the estimate (4.4),

$$\begin{aligned} 0\le \square A(X,{\bar{X}})\le R_{X{\bar{X}}k}^{\qquad p}P_p^{\,\,\,\,k}-\frac{1}{4}\varepsilon K\le -\frac{1}{8}\varepsilon K, \end{aligned}$$

which is a contrsdiction.

Next, we suppose that \(\mathrm{(ii)}\) is not true at \(t=t_0\). Then, there exist \(p_0\in M\), \(X_0, U_0, V_0\in T^{1,0}_{p_0}M\) with \(|X_0|_{g(t_0)}=1\) such that \(|R^{\varepsilon }_{U_0\bar{V_0}X_0\bar{X_0}}|^2_{g(t_0)}=(1+Kt_0)P^{\varepsilon }_{U_0\bar{U_0}}P^{\varepsilon }_{V_0\bar{V_0}}\). By rescaling, we may assume that \(|U_0|_{g(t_0)}=|V_0|_{g(t_0)}=1\). For all \((p,t)\in M\times [0,t_0]\), \(X, U, V\in T_p^{1,0}M\) with \(|X|=1\),

$$\begin{aligned} P^{\varepsilon }_{X{\bar{X}}}<-\varepsilon e^{-Kt}g_{X{\bar{X}}}=-\varepsilon e^{-Kt},\quad |R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}|^2\le (1+Kt)P^{\varepsilon }_{U{\bar{U}}}P^{\varepsilon }_{V{\bar{V}}}. \end{aligned}$$
(4.5)

For sufficiently small \(\varepsilon >0\), and for any \(e_k, e_l\in T^{1,0}_pM\) for all \(p\in M\) with unit 1,

$$\begin{aligned} |R^{\varepsilon }_{i{\bar{j}}k{\bar{l}}}|^2\le C_nP^{\varepsilon }_{i{\bar{i}}}P^{\varepsilon }_{j{\bar{j}}},\quad |R^{\varepsilon }_{i{\bar{j}}k{\bar{l}}}|^2\le C_nK_0|P^{\varepsilon }_{i{\bar{i}}}|. \end{aligned}$$
(4.6)

As in the previous case, we extend \(X_0, U_0\) to a local vector field XU around \((p_0,t_0)\). We extend \(X_0\), \(U_0\), \(V_0\) to X, U and V around \((p_0,t_0)\) such that at \((p_0,t_0)\),

$$\begin{aligned}&\nabla _{{\bar{s}}}U^r=0,\quad \nabla _pU^r=T_{pq}^rU^q, \\&\nabla _{{\bar{s}}}V^r=0,\quad \nabla _pV^r=T_{pq}^rV^q, \\&\nabla _{{\bar{s}}}X^r=0,\quad \nabla _pX^r=0, \quad \square X^r=0. \end{aligned}$$

Notice that we have at \((p_0,t_0)\),

$$\begin{aligned} \square U^r= & {} \frac{\partial }{\partial t}U^r-\frac{1}{2}g^{k{\bar{l}}}(\nabla _k\nabla _{{\bar{l}}}+\nabla _{{\bar{l}}}\nabla _k)U^r\\= & {} \frac{\partial }{\partial t}U^r-\frac{1}{2}g^{k{\bar{l}}}\nabla _{{\bar{l}}}(T_{ks}^rU^s)\\= & {} \frac{\partial }{\partial t}U^r-\frac{1}{2}g^{k{\bar{l}}}(\nabla _{{\bar{l}}}T_{ks}^r\cdot U^s+T_{ks}^r\nabla _{{\bar{l}}}U^s)\\= & {} \frac{\partial }{\partial t}U^r-\frac{1}{2}\nabla _{{\bar{k}}}T_{ks}^r\cdot U^s, \end{aligned}$$

and note that we get

$$\begin{aligned} |\square U^r|_{g(t_0)}\le \Big |\frac{\partial }{\partial t}U^r\Big |_{g(t_0)}+\frac{1}{2}|\nabla _{{\bar{k}}}T_{ks}^r|_{g(t_0)}|U^s|_{g(t_0)}\le C \end{aligned}$$

for some uniform constant \(C>0\), where we have used that (4.1) and Lemma 3.4.

In particular, \(|X|_{g(t_0)}=1\). Then the function

$$\begin{aligned} F(x,t)=g_{X{\bar{X}}}^{-2}|R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}|^2-(1+Kt)P^{\varepsilon }_{U{\bar{U}}}P^{\varepsilon }_{V{\bar{V}}} \end{aligned}$$

attains its local maximum at \((p_0,t_0)\) and therefore satisfies

$$\begin{aligned} \square F|_{(p_0,t_0)}\ge 0. \end{aligned}$$

We calculate by making use of (3.3),

$$\begin{aligned} \square P^{\varepsilon }_{U{\bar{U}}}= & {} \square P_{i{\bar{j}}}^{\varepsilon } \cdot U^iU^{{\bar{j}}}+P_{i{\bar{j}}}^{\varepsilon }U^i\square U^{{\bar{j}}}+P_{i{\bar{j}}}^{\varepsilon }\square U^i\cdot U^{{\bar{j}}}\nonumber \\&-g^{r{\bar{s}}}P_{i{\bar{j}}}^{\varepsilon }\nabla _rU^i\nabla _{{\bar{s}}}U^{{\bar{j}}}-g^{r{\bar{s}}}\nabla _rP_{i{\bar{j}}}\cdot U^i\nabla _{{\bar{s}}}U^{{\bar{j}}}-g^{r{\bar{s}}}\nabla _{{\bar{s}}}P_{i{\bar{j}}}\nabla _rU^i\cdot U^{{\bar{j}}}\nonumber \\= & {} \square P_{i{\bar{j}}}\cdot U^iU^{{\bar{j}}}+\varepsilon (n+1)(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{i{\bar{j}}}U^iU^{{\bar{j}}}\nonumber \\&+P_{i{\bar{j}}}^{\varepsilon }U^i\square U^{{\bar{j}}}+P_{i{\bar{j}}}^{\varepsilon }U^{{\bar{j}}}\square U^i+\varepsilon (n+1)g^{r{\bar{s}}}g_{i{\bar{j}}}T^i_{rp}T_{{\bar{s}}{\bar{q}}}^{{\bar{j}}}U^pU^{{\bar{q}}}\nonumber \\&-g^{r{\bar{s}}}P_{i{\bar{j}}}T_{rp}^iT_{{\bar{s}}{\bar{q}}}^{{\bar{j}}}U^pU^{{\bar{q}}}-g^{r{\bar{s}}}\nabla _rP_{i{\bar{j}}}\cdot T_{{\bar{s}}{\bar{q}}}^{{\bar{j}}}U^iU^{{\bar{q}}}-g^{r{\bar{s}}}\nabla _{{\bar{s}}}P_{i{\bar{j}}}\cdot T_{rp}^iU^pU^{{\bar{j}}}\nonumber \\= & {} R_{i{\bar{j}}k}^{\quad \, p}U^iU^{{\bar{j}}}P_p^k+P_{i{\bar{j}}}U^{{\bar{j}}}\square U^i+P_{i{\bar{j}}}U^i\square U^{{\bar{j}}}-2\varepsilon (n+1)U^i\square U^{{\bar{i}}}\nonumber \\&+E_{i{\bar{j}}}U^iU^{{\bar{j}}}+\varepsilon (n+1)g^{r{\bar{s}}}g_{i{\bar{j}}}T_{rk}^iU^kT_{{\bar{s}}{\bar{l}}}^{{\bar{j}}}U^{{\bar{l}}}\nonumber \\&+\varepsilon (n+1)(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{i{\bar{j}}}U^iU^{{\bar{j}}}. \end{aligned}$$
(4.7)

Similarly we have

$$\begin{aligned} \square P^{\varepsilon }_{V{\bar{V}}}= & {} R_{i{\bar{j}}k}^{\quad \,p}V^iV^{{\bar{j}}}P_p^k+P_{i{\bar{j}}}V^{{\bar{j}}}\square V^i+P_{i{\bar{j}}}V^i\square V^{{\bar{j}}}-2\varepsilon (n+1)V^i\square V^{{\bar{i}}}\nonumber \\&+E_{i{\bar{j}}}V^iV^{{\bar{j}}}+\varepsilon (n+1)g^{r{\bar{s}}}g_{i{\bar{j}}}T_{rk}^iV^kT_{{\bar{s}}{\bar{l}}}^{{\bar{j}}}V^{{\bar{l}}}\nonumber \\&+\varepsilon (n+1)(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{i{\bar{j}}}V^iV^{{\bar{j}}}. \end{aligned}$$
(4.8)

By combining these and chossing \(K\gg K_0\) sufficiently large and \(\tau =K^{-1}\), we obtain by applying (4.1), (4.5)–(4.8),

$$\begin{aligned} \square [(1+Kt)P^{\varepsilon }_{U{\bar{U}}}P^{\varepsilon }_{V{\bar{V}}}]\ge & {} (1+Kt)P^{\varepsilon }_{U{\bar{U}}}(R^{\varepsilon \quad \,\, p}_{V{\bar{V}}k}P_p^{\,\,\,\,k})+ (1+Kt)P^{\varepsilon }_{V{\bar{V}}}(R^{\varepsilon \quad \,\, p}_{U{\bar{U}}k}P_p^{\,\,\,\,k})\nonumber \\&-2(1+Kt)\text {Re}(g^{r{\bar{s}}}\nabla _rP_{U{\bar{U}}}^{\varepsilon }\nabla _{{\bar{s}}}P_{V{\bar{V}}}^{\varepsilon })+\frac{1}{2}KP_{U{\bar{U}}}^{\varepsilon }P_{V{\bar{V}}}^{\varepsilon }\nonumber \\\ge & {} -2(1+Kt)\text {Re}(g^{r{\bar{s}}}\nabla _rP_{U{\bar{U}}}^{\varepsilon }\nabla _{{\bar{s}}}P_{V{\bar{V}}}^{\varepsilon })+\frac{1}{4}KP_{U{\bar{U}}}^{\varepsilon }P_{V{\bar{V}}}^{\varepsilon },\nonumber \\ \end{aligned}$$
(4.9)

where we used that the quantities \(|T'|^2_{C^0(g(t))}\), \(|\nabla T'|_{C^0(g(t))}\) and \(|R|_{C^0(g(t))}\) are uniformly bounded on \([0,\tau ]\), and we applied Lemmas 3.1, 3.4, 3.5, the estiamte (4.4).

We have by using (3.2),

$$\begin{aligned} \square R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}= & {} \square R_{i{\bar{j}}k{\bar{l}}} \cdot U^iV^{{\bar{j}}}X^kX^{{\bar{l}}}-\varepsilon \square B_{i{\bar{j}}k{\bar{l}}}\cdot U^iV^{{\bar{j}}}X^kX^{{\bar{l}}}\nonumber \\&+R^{\varepsilon }_{i{\bar{j}}k{\bar{l}}}\square U^i\cdot V^{{\bar{j}}}X^kX^{{\bar{l}}}+R^{\varepsilon }_{i{\bar{j}}k{\bar{l}}}U^i\square V^{{\bar{j}}}\cdot X^kX^{{\bar{l}}}\nonumber \\&+R^{\varepsilon }_{i{\bar{j}}k{\bar{l}}}U^iV^{{\bar{j}}}(\square X^k\cdot X^{{\bar{l}}}+X^k\square X^{{\bar{l}}})\nonumber \\&-g^{r{\bar{s}}}\nabla _rR_{i{\bar{j}}k{\bar{l}}}(U^i\nabla _{{\bar{s}}}V^{{\bar{j}}}\cdot X^kX^{{\bar{l}}}+U^iV^{{\bar{j}}}X^k\nabla _{{\bar{s}}}X^{{\bar{l}}})\nonumber \\&-g^{r{\bar{s}}}\nabla _{{\bar{s}}}R_{i{\bar{j}}k{\bar{l}}}(\nabla _rU^i\cdot V^{{\bar{j}}}X^kX^{{\bar{l}}}+U^iV^{{\bar{j}}}\nabla _rX^k\cdot X^{{\bar{l}}})\nonumber \\&-R^{\varepsilon }_{i{\bar{j}}k{\bar{l}}}(\nabla _rU^i\nabla _{{\bar{s}}}V^{{\bar{j}}}\cdot X^kX^{{\bar{l}}}+U^i\nabla _{{\bar{s}}}V^{{\bar{j}}}\nabla _rX^k\cdot X^{{\bar{l}}}\nonumber \\&+\nabla _rU^i\cdot V^{{\bar{j}}}X^k\nabla _{{\bar{s}}}X^{{\bar{l}}}+U^iV^{{\bar{j}}}\nabla _rX^k\nabla _{{\bar{s}}}X^{{\bar{l}}})\nonumber \\= & {} g^{r{\bar{s}}}(R_{U{\bar{V}}r}^{\quad \,\,\,\,\, p}R_{p{\bar{s}}X{\bar{X}}}+R_{r{\bar{V}}X}^{\quad \,\,\,\,p}R_{U{\bar{s}}p{\bar{X}}}-R_{r{\bar{V}}p{\bar{X}}}R_{U{\bar{s}}X}^{\quad \,\,\,\,p})\nonumber \\&-R_{U{\bar{V}}X}^{\qquad r}(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{r{\bar{X}}}+R_{i{\bar{V}}X{\bar{X}}}\square U^i\nonumber \\&+R_{U{\bar{j}}X{\bar{X}}}\square V^{{\bar{j}}} +\varepsilon ((S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{U{\bar{V}}}g_{X{\bar{X}}}\nonumber \\&+g_{U{\bar{V}}}(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{X{\bar{X}}}\nonumber \\&+(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{U{\bar{X}}}g_{X{\bar{V}}}\nonumber \\&+g_{U{\bar{X}}}(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{X{\bar{V}}})\nonumber \\&+\varepsilon B_{i{\bar{j}}X{\bar{X}}}T_{rU}^iT_{{\bar{s}}{\bar{V}}}^{{\bar{j}}}-\varepsilon (B_{i{\bar{V}}X{\bar{X}}}\square U^i+B_{U{\bar{j}}X{\bar{X}}}\square V^{{\bar{j}}})+E_{U{\bar{V}}X{\bar{X}}}, \nonumber \\ \end{aligned}$$
(4.10)

where we have used that

$$\begin{aligned} R^{\varepsilon }_{i{\bar{j}}k{\bar{l}}}\nabla _rU^i\nabla _{{\bar{s}}}V^{{\bar{j}}}\cdot X^kX^{{\bar{l}}}=T_{rU}^pT_{{\bar{s}}{\bar{V}}}^{{\bar{q}}}R_{p{\bar{q}}X{\bar{X}}}-\varepsilon B_{i{\bar{j}}X{\bar{X}}}T_{rU}^iT_{{\bar{s}}{\bar{V}}}^{{\bar{j}}}. \end{aligned}$$

Similarly, we have

$$\begin{aligned}&\square R^{\varepsilon }_{V{\bar{U}}X{\bar{X}}} =g^{r{\bar{s}}}(R_{V{\bar{U}}r}^{\quad \,\,\,\, p}R_{p{\bar{s}}X{\bar{X}}}+R_{r{\bar{U}}X}^{\quad \,\,\,\,p}R_{V{\bar{s}}p{\bar{X}}}-R_{r{\bar{U}}p{\bar{X}}}R_{V{\bar{s}}X}^{\quad \,\,\,\,p})\nonumber \\&\quad -R_{V{\bar{U}}X}^{\qquad r}(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{r{\bar{X}}}+R_{i{\bar{U}}X{\bar{X}}}\square V^i+R_{V{\bar{j}}X{\bar{X}}}\square U^{{\bar{j}}}\nonumber \\&\quad +\varepsilon ((S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{V{\bar{U}}}g_{X{\bar{X}}}+g_{V{\bar{U}}}(S+Q^7+Q^8-BT'\nonumber \\&\quad -{\bar{Z}}(T'))_{X{\bar{X}}} +(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{V{\bar{X}}}g_{X{\bar{U}}}\nonumber \\&\quad +g_{V{\bar{X}}}(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{X{\bar{U}}}) +\varepsilon B_{i{\bar{j}}X{\bar{X}}}T_{rV}^iT_{{\bar{s}}{\bar{U}}}^{{\bar{j}}}\nonumber \\&\quad -\varepsilon (B_{i{\bar{U}}X{\bar{X}}}\square V^i+B_{V{\bar{j}}X{\bar{X}}}\square U^{{\bar{j}}})+E_{V{\bar{U}}X{\bar{X}}}. \end{aligned}$$
(4.11)

By combining these, we obtain by applying (4.1), (4.4), (4.5), (4.9) and (4.10),

$$\begin{aligned} \square (R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}R^{\varepsilon }_{V{\bar{U}}X{\bar{X}}})\le -|\nabla R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}|^2-|{\bar{\nabla }}R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}|^2+C_nK_0P^{\varepsilon }_{U{\bar{U}}}P^{\varepsilon }_{V{\bar{V}}}, \end{aligned}$$

where we used that the quantities \(|T'|^2_{C^0(g(t))}\), \(|\nabla T'|_{C^0(g(t))}\) and \(|R|_{C^0(g(t))}\) are uniformly bounded on \([0,\tau ]\), and applied the estimates in Lemmas 3.1, 3.4, 3.5, the estimate (4.4).

Therefore, at \((p_0,t_0)\),

$$\begin{aligned} \square F\le & {} 2(1+Kt)\text {Re}(g^{r{\bar{s}}}\nabla _rP^{\varepsilon }_{U{\bar{U}}}\nabla _{{\bar{s}}}P^{\varepsilon }_{V{\bar{V}}})\nonumber \\&-|\nabla R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}|^2-|{\bar{\nabla }}R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}|^2\nonumber \\&-\frac{1}{8}KP^{\varepsilon }_{U{\bar{U}}}P^{\varepsilon }_{V{\bar{V}}}+2(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{X{\bar{X}}}|R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}|^2\nonumber \end{aligned}$$

By using that \(\nabla F=0\) and \(F=0\) at \((p_0,t_0)\), one may conclude that

$$\begin{aligned} 2(1+Kt)\text {Re}(g^{r{\bar{s}}}\nabla _rP^{\varepsilon }_{U{\bar{U}}}\nabla _{{\bar{s}}}P^{\varepsilon }_{V{\bar{V}}})\le |\nabla R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}|^2+|{\bar{\nabla }}R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}|^2. \end{aligned}$$

Using (4.1) and \(F(p_0,t_0)=0\), we deduce that

$$\begin{aligned} 2(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{X{\bar{X}}}|R^{\varepsilon }_{U{\bar{V}}X{\bar{X}}}|^2\le C_nK_0P^{\varepsilon }_{U{\bar{U}}}P^{\varepsilon }_{V{\bar{V}}} \end{aligned}$$

and hence for sufficiently large \(K>{\tilde{C}}_nK_0\) for some \({\tilde{C}}_n\gg 1\), at \((p,t_0)\),

$$\begin{aligned} 0\le \square F<-\frac{1}{16}KP^{\varepsilon }_{U{\bar{U}}}P^{\varepsilon }_{V{\bar{V}}}, \end{aligned}$$

which is a contradiction. \(\square \)

5 Strong Maximum Principle

Theorem 5.1

Suppose (MJg(t)) is a solution to the AHCF on \(t\in [0,\tau _{\max })\) be a solution to the AHCF starting from the initial metric \(g(0)=g_0\), where \(\tau _{\max }\) is the finite explosion time of the AHCF. If the metric \(g_0\) has the Griffiths non-positive Chern curvature and its first Chern–Ricci curvature is negative at some \(p\in M\), then there exists \(0<\tau<\tau _{\max }<\infty \) such that \(P(g(t))<0\) on \((0,\tau ]\). Note that under these assumptions, the metric \(g_0\) has the quasi-negative first Chern–Ricci curvature.

Proof

Let \(\tau \) be the constant obtained in Proposition 4.1. Let \(y\in M\) be a point at which the first Chern–Ricci curvature is negative. Let \(\phi _0\) be a smooth non-negative function such that \(\phi _0(y)>0\), \(\phi _0=0\) outside a neighbour of y and

$$\begin{aligned} P(g_0)+\phi _0g_0\le 0\quad \text {on { M}.} \end{aligned}$$

Let \(\phi (x,t)\) be the solution to the heat equation

$$\begin{aligned}&\Big (\frac{\partial }{\partial t}-\Delta _{g(t)}\Big )\phi (x,t)=0,\quad \hbox { on}\ M\times [0,\tau ]; \\&\phi (x,0)=\phi _0. \end{aligned}$$

By the strong maximum principle, it follows that \(\phi (x,t)>0\) on \(M\times (0,\tau ]\). By rescaling, one may assume that \(\phi (x,t)\le 1\). Let \(k:=c_nK_0\) with \(c_n\gg 1\). For any \(\varepsilon >0\), we consider

$$\begin{aligned} A^{\varepsilon }=A^{\varepsilon }(g(t)):=P(g(t))+e^{-kt}\phi ^2g(t)-\varepsilon e^{\Lambda t}g(t), \end{aligned}$$

where \(\Lambda \) is some sufficiently large constant which will be determined later. We claim that \(A^{\varepsilon }\le 0\) on \(M\times [0,\tau ]\). Then the result follows by letting \(\varepsilon \rightarrow 0\). Note that \(A^{\varepsilon }(g(0))<0\) on M. Suppose not, there is \(t_0\in (0,\tau ]\) such that for all \((p,t)\in M\times [0,t_0]\), \(U\in T^{1,0}_pM\), \(A^{\varepsilon }_{U{\bar{U}}}(p,t)\le 0\). And there exists \(p_0\in M\), \(V\in T_{p_0}^{1,0}M\) so that \(A^{\varepsilon }_{V{\bar{V}}}(p_0,t_0)=0\). By rescaling, we may assume that \(|V|_{g(t_0)}=1\). We extend V around \((p_0,t_0)\) locally such that

$$\begin{aligned} \nabla _{{\bar{j}}}V^i=0,\quad \nabla _rV^i=T_{rk}^iV^k. \end{aligned}$$

Then the function \(A^{\varepsilon }_{V{\bar{V}}}\) attains local maximum at \((p_0,t_0)\) and obeys

$$\begin{aligned} \square A^{\varepsilon }_{V{\bar{V}}}|_{(p_0,t_0)}\ge 0. \end{aligned}$$
(5.1)

Choosing a local unitary (1, 0)-frame with respect to \(g(t_0)\) around \(p_0\) such that \(g_{i{\bar{j}}}(t_0)=\delta _{i{\bar{j}}}\), then we compute

$$\begin{aligned} \square A^{\varepsilon }_{V{\bar{V}}}= & {} \square A^{\varepsilon }_{i{\bar{j}}}\cdot V^iV^{{\bar{j}}}+A^{\varepsilon }_{i{\bar{j}}}(V^i\square V^{{\bar{j}}}+V^{{\bar{j}}}\square V^i)\nonumber \\&-g^{r{\bar{s}}}(A^{\varepsilon }_{i{\bar{j}}}\nabla _rV^i\nabla _{{\bar{s}}}V^{{\bar{j}}}+\nabla _rA^{\varepsilon }_{i{\bar{j}}}\cdot V^i\nabla _{{\bar{s}}}V^{{\bar{j}}}+\nabla _{{\bar{s}}}A^{\varepsilon }_{i{\bar{j}}}\nabla _rV^i\cdot V^{{\bar{j}}})\nonumber \\= & {} \square A^{\varepsilon }_{i{\bar{j}}}\cdot V^iV^{{\bar{j}}}-A^{\varepsilon }_{i{\bar{j}}}T_{rk}^iV^kT_{{\bar{r}}{\bar{l}}}^{{\bar{j}}}V^{{\bar{l}}}-\nabla _rA^{\varepsilon }_{i{\bar{j}}}\cdot T_{{\bar{r}}{\bar{l}}}^{{\bar{j}}}V^{{\bar{l}}}V^i-\nabla _{{\bar{r}}}A^{\varepsilon }_{i{\bar{j}}}\cdot T_{rk}^iV^kV^{{\bar{j}}},\nonumber \\ \end{aligned}$$
(5.2)

where we have used that \(A^{\varepsilon }_{V{\bar{U}}}=0\) for all \(U\in T^{1,0}_{p_0}M\), hence we get that

$$\begin{aligned} A^{\varepsilon }(V,\square {\bar{V}})=A^{\varepsilon }(\square V,{\bar{V}})=0, \end{aligned}$$

which can be seen by considering the first derivation of functions: \(A^{\varepsilon }(V+tU,{\bar{V}}+t{\bar{U}})\) and \(A^{\varepsilon }(V+t\sqrt{-1}U,{\bar{V}}-t\sqrt{-1}{\bar{U}})\) at \(t=0\). From the formula (3.3), we obtain at \((p_0,t_0)\),

$$\begin{aligned} \square A^{\varepsilon }_{i{\bar{j}}}\cdot V^iV^{{\bar{j}}}= & {} P_{i{\bar{j}}}T_{rs}^iV^sT_{{\bar{r}}{\bar{l}}}^{{\bar{j}}}V^{{\bar{l}}}+\nabla _rP_{i{\bar{j}}}\cdot T_{{\bar{r}}{\bar{l}}}^{{\bar{j}}}V^{{\bar{l}}}V^i+\nabla _{{\bar{r}}}P_{i{\bar{j}}}\cdot T_{rs}^iV^sV^{{\bar{j}}}\\&+R_{i{\bar{j}}s}^{\quad \,\,p}V^iV^{{\bar{j}}}P_p^{\,\,\,\,s}+E_{i{\bar{j}}}V^iV^{{\bar{j}}}\\&-ke^{-kt}\phi ^2-2|\nabla \phi |^2e^{-kt}-\phi ^2e^{-kt}(S+Q^7+Q^8-BT' \\&-{\bar{Z}}(T'))_{i{\bar{j}}}V^iV^{{\bar{j}}} -\varepsilon \Lambda e^{\Lambda t}+\varepsilon e^{\Lambda t}(S+Q^7+Q^8-BT'\\&-{\bar{Z}}(T'))_{i{\bar{j}}}V^iV^{{\bar{j}}}. \end{aligned}$$

Combining them with Proposition 4.1 and the fact that \(A^{\varepsilon }_{V{\bar{V}}}(p_0,t_0)=0\), which gives us that \(P_{V{\bar{V}}}|_{(p_0,t_0)}=-e^{-kt}\phi ^2+\varepsilon e^{\Lambda t}\), we have at \((p_0,t_0)\), using (4.1), (5.1) and (5.2),

$$\begin{aligned} 0\le \square A^{\varepsilon }_{V{\bar{V}}}= & {} (-\phi ^2e^{-kt}+\varepsilon e^{\Lambda t})g_{i{\bar{j}}}T_{rs}^iV^sT_{{\bar{r}}{\bar{l}}}^{{\bar{j}}}V^{{\bar{l}}}-4\phi e^{-kt}\text {Re}(\phi _{{\bar{r}}}T_{rp{\bar{q}}}V^pV^{{\bar{q}}})\\&+R_{i{\bar{j}}p{\bar{q}}}V^iV^{{\bar{j}}}P^{{\bar{q}}p}+E_{i{\bar{j}}}V^iV^{{\bar{j}}}-ke^{-kt}\phi ^2-2|\nabla \phi |^2e^{-kt}\\&-\varepsilon \Lambda e^{\Lambda t}+P_{s{\bar{l}}}V^sV^{{\bar{l}}}(S+Q^7+Q^8-BT'-{\bar{Z}}(T'))_{i{\bar{j}}}V^iV^{{\bar{j}}}\\\le & {} (-k+C_nK_0)\phi ^2e^{-kt}+\varepsilon e^{\Lambda t}(-\Lambda +C_nK_0), \end{aligned}$$

which leads a contradiction by choosing k and \(\Lambda \) sufficiently large, since the quantities \(|T'|^2_{C^0(g(t))}\), \(|\nabla T'|_{C^0(g(t))}\) and \(|R|_{C^0(g(t))}\) are uniformly bounded on \([0,\tau ]\), where we applied Lemmas 3.1, 3.4, 3.5, the estimate (4.4). As a result, we have shown that there exist sufficiently large constants \(k>0\), \(\Lambda >0\) such that for all \(\varepsilon >0\), \((x,t)\in M\times [0,\tau ]\),

$$\begin{aligned} P(g(t))\le (-e^{-kt}\phi (x,t)+\varepsilon e^{\Lambda t})g(t). \end{aligned}$$

In particular, by letting \(\varepsilon \rightarrow 0\), we obtain \(P(g(t))<0\) for \(t\in (0,\tau ]\). \(\square \)