On solvmanifolds with complex commutator and constant holomorphic sectional curvature

Xin Huang and Fangyang Zheng

arXiv: 2501.00810v1 · PDF · 1 January 2025
Categories: math.DG
Journal: Intern.J.Math. 36 (2025), no.6, art no 2550006
Licence: CC BY 4.0 The site shows the text.

Abstract

An old open question in non-Kähler geometry predicts that any compact Hermitian manifold with constant holomorphic sectional curvature must be Kähler or Chern flat. The conjecture is known to be true in dimension 2 due to the work by Balas-Gauduchon and Apostolov-Davidov-Muskarov in the 1980s and 1990s, but is still open in dimensions 3 or higher, except in several special cases. The difficulty in this quest for ‘Hermitian space forms’ is largely due to the algebraic complicity or lack of symmetry for the curvature tensor of a general Hermitian metric. In this article, we confirm the conjecture for all solvmanifolds with complex commutator, extending earlier result on nilmanifolds by Li and the second named author.

Items 14

Conjecture

Conjecture 1

SMQB

If the Chern connection of a compact Hermitian manifold has constant holomorphic sectional curvature, then it must be either Kähler (hence a complex space form) or Chern flat.

Theorem

Theorem 1

5601

Let (M^n,g) be a Lie-Hermitian manifold with universal cover (G,J,g). Denote by {\mathfrak g} the Lie algebra of G and {\mathfrak g}'=[{\mathfrak g},{\mathfrak g}] its commutator. Assume that {\mathfrak g} is solvable and J{\mathfrak g}'={\mathfrak g}'. If the Chern connection of g has constant holomorphic sectional curvature, then it must be Chern flat.

Proof of Theorem 1

Let ({\mathfrak g}, J,g) be a solvable Lie algebra equipped with a Hermitian structure. Assume that J{\mathfrak g}'={\mathfrak g}' and the Chern connection of g has constant holomorphic sectional curvature: H=c. Then c=0 by Lemma 3. Our goal is to show that the Chern curvature of g must vanish identically: R=0. Let e be an admissible frame of {\mathfrak g} so (?) and (?) hold. By (4) the only possibly non-zero D components are D^j_{i\alpha}, where 1\leq i,j\leq r and r\!+\!1 \leq \alpha \leq n. In particular, D^{\alpha}_{\ast\ast} = D^{\ast}_{\alpha \ast}=0. So by (2) we know that

R_{\ast \bar{\ast} \alpha \bar{\ast}} = R_{\ast \bar{\ast} \ast \bar{\alpha} }=0, \ \ \ \ \ \ \forall \ r\!+\!1 \leq \alpha \leq n.

Therefore R_{a\bar{b}c\bar{d}}=0 if at least three of the indices belong to \{ r\!+\!1, \ldots , n\}. Also by (2) we have

R_{a\bar{b}c\bar{d}} = \sum_{s=1}^n D^s_{ca} \overline{D^s_{db} } - \sum_{s=1}^n D^d_{sa} \overline{D^c_{sb} },

since for the last two terms in the right hand side of (2) the index s cannot be simultaneously >r and \leq r. If exactly two of the four indices belong to \{ r\!+\!1, \ldots , n\}, then by (Proof) we get

R_{\alpha \bar{\beta}i\bar{j}}=\widehat{R}_{\alpha \bar{\beta}i\bar{j}}=0.

If only one of the four indices belongs to \{ r\!+\!1, \ldots , n\}, then it is

R_{\alpha \bar{j}k\bar{\ell}}= \sum_{s=1}^n \big( D^s_{k\alpha } \overline{D^s_{\ell j} } - D^{\ell}_{s\alpha } \overline{D^k_{sj} }\big).

This equals to zero since for each of the two terms on the right-hand side, the second factor is zero by (4). Similarly, when all four indices are in \{ 1, \ldots , r\}, R_{i\bar{j}k\bar{\ell}}=0 by (4). Therefore, we have shown that R=0, completing the proof of Theorem 1.

Uses Lemma 3, Lemma 4, Lemma 2

Lemma

Lemma 1

PDM2

Let (M^n,g) be a Hermitian manifold. Then we have

H=c \ \ \Longleftrightarrow \ \ \widehat{R}_{i\bar{j}k\bar{\ell}} = \frac{c}{2}\big( g_{i\bar{j}} g_{k\bar{\ell}} + g_{i\bar{\ell }} g_{k\bar{j}} \big) , \ \ \forall \ 1\leq i,j,k,\ell \leq n.

Lemma

Lemma 2

2L6T

Given a Hermitian Lie algebra ({\mathfrak g}, J,g), let e be a unitary frame, and structure constants C and D be given by (?). Then the Chern torsion components T^j_{ik} and Chern curvature components R_{i\bar{j}k\bar{\ell}} under e are given by

\begin{aligned} T^j_{ik} & = & - C^j_{ik} - D^{j}_{i k} + D^{j}_{k i} , \\ R_{i\bar{j}k\bar{\ell}} & = & \sum_{s=1}^n \big( D^s_{ki}\overline{D^s_{\ell j}} - D^{\ell}_{si}\overline{D^k_{s j}} - D^j_{si}\overline{D^k_{ \ell s}} - \overline{D^i_{sj}} D^{\ell}_{k s} \big) . \end{aligned}

Used by Proof, Proof

Lemma

Lemma 3

VHM3

Let ({\mathfrak g},J,g) be a Lie algebra with Hermitian structure such that {\mathfrak g} is solvable and J{\mathfrak g}'={\mathfrak g}'. Assume that the holomorphic sectional curvature H of the Chern connection of g is a constant c. Then c=0

Used by Proof

Proof of Lemma 3

Let e be an admissible frame of {\mathfrak g}. Fix any \alpha in \{ r\!+\!1, \ldots ,n\}. By the fact D^{\ast}_{\alpha \ast}=0 and by formula (2) in Lemma 2, we get

c=R_{\alpha \bar{\alpha} \alpha \bar{\alpha} } = - \sum_{s=1}^n |D^{\alpha }_{s\alpha}|^2 \ \leq \,0.

Similarly, for any 1\leq i\leq r, we have

c=R_{i \bar{i} i \bar{i} } = \sum_{s=1}^n |D^s_{ii}|^2 - \sum_{s=1}^r |D^{i}_{si}|^2 ,

since the summing index s for the last two terms on the right-hand side of formula (2) cannot exceed r, while D^i_{is}=0 by (?). In particular, if we take i=r, then the second term on the right-hand side of (Proof) vanishes by (?), so we get

c=R_{r \bar{r} r \bar{r} } = \sum_{s=1}^n |D^s_{rr}|^2 \ \geq \, 0.

Comparing (Proof) with (Proof), we conclude that c=0.

Uses Lemma 2

Lemma

Lemma 4

5EWR

D^{\alpha}_{j\beta}=D^j_{ik}=D^{\alpha}_{ij}=0, \ \ \ \ \ \ \ \ \forall\ 1\leq i,j,k\leq r, \ \ \ \forall \ r\!+\!1\leq \alpha , \beta \leq n.

Used by Proof

Proof of Lemma 4

By formula (Proof) and the vanishing of the holomorphic sectional curvature, we know that D^{\alpha}_{j\alpha}=0 for any j and any \alpha. If we make any unitary change on \{ e_{r\!+\!1},\ldots , e_n\}, formula (?) and (?) would not be affected and e would still be an admissible frame. So formula (Proof) actually gives us D^{X}_{jX}=0 for any j and any X=\sum_{\alpha =r\!+\!1}^n X_{\alpha}e_{\alpha}. That is, \sum_{\alpha ,\beta=r\!+\!1}^n \overline{X}_{\alpha} X_{\beta} D^{\alpha}_{j\beta }=0. This leads to D^{\alpha}_{j\beta}=0 for any j and any \alpha, \beta.

Analogously, the vanishing of H and (Proof) give us D^{\ast}_{rr}=0. Our next goal is to prove the remaining part of (4), namely, D^j_{ik}=D^{\alpha}_{ij}=0 for any i, j, k and any \alpha. To prove this, let us fix any 1\leq i<k\leq r and consider

\begin{aligned} 0 & = & 4\widehat{R}_{i\bar{i}k\bar{k}} \ = \ R_{i\bar{i}k\bar{k}} + R_{k\bar{k}i\bar{i}} + R_{i\bar{k}k\bar{i}} +R_{k\bar{i}i\bar{k}} \\ & = & \sum_{s=1}^n \{ \big( |D^s_{ki}|^2 - |D^k_{si}|^2\big) + \big(|D^s_{ik}|^2 - |D^i_{sk}|^2\big) +2\mbox{Re} \big( D^s_{ki}\overline{D^s_{ik}} - D^i_{si}\overline{D^k_{sk}} - \overline{D^i_{sk}} D^i_{ks} \big) \} \\ & = & \sum_{s=1}^n |D^s_{ki}+D^s_{ik}|^2 - \sum_{s=1}^r \big( |D^k_{si}|^2 + |D^i_{sk}|^2 + 2\mbox{Re} \{ D^i_{si}\overline{D^k_{sk}} + \overline{D^i_{sk}} D^i_{ks} \} \big). \end{aligned}

In the last equality above, we used the fact that D^{\ast}_{\alpha \ast}=0 for any \alpha >r by (?). Let k=r in (Proof), and by (?) we get

\sum_{s=1}^n |D^s_{ri}+D^s_{ir}|^2 \, = \, \sum_{s=1}^r \big( |D^i_{sr}|^2 + 2\mbox{Re} \{ \overline{D^i_{sr}} D^i_{rs} \} \big) , \ \ \ \ \ \ \forall \ 1\leq i<r.

We claim that the above will lead to the following:

D^s_{ri}+D^s_{ir}=D^i_{jr}=D^i_{rj}=0, \ \ \ \ \ \ \forall \ 1\leq i,j\leq r, \ \ \forall \ 1\leq s\leq n.

To prove the claim, let us perform induction on i. It holds for i=r by (?) and the fact D^{\ast}_{rr}=0. Let i=r\!-\!1 in (Proof). The right-hand side of (Proof) vanishes as s has to be r by (?) while D^{\ast}_{rr}=0. Therefore we get D^s_{r,r\!-\!1}+D^s_{r\!-\!1,r} =0 for any s, hence (Proof) holds for i=r\!-\!1. Next let us take i=r\!-\!2 in (Proof). On the right hand side, s has to be r\!-\!1. But by (Proof) for i=r\!-\!1 we have D^{r\!-\!2}_{r\!-\!1,r} + D^{r\!-\!2}_{r,r\!-\!1}=0. Therefore the right-hand side of (Proof) becomes -| D^{r\!-\!2}_{r\!-\!1,r} |^2, so we get (Proof) for the i=r\!-\!2 case. Repeating this process, the Claim (Proof) is proved. In particular, we have shown that

D^i_{jr}=D^i_{rj}=D^r_{ij} =0, \ \ \ \ \ \ \forall \ 1\leq i,j\leq r.

If we look at the i=r\!-\!1 case of (Proof), then the second term on the right-hand side vanishes, so we have

D^s_{r\!-\!1,r\!-\!1}=0, \ \ \ \ \ \ \forall \ 1\leq s\leq n.

Now let us take k=r\!-\!1 in (Proof), and get

\sum_{s=1}^n |D^s_{r\!-\!1,i}+D^s_{i,r\!-\!1}|^2 \, = \, \sum_{s=1}^r \big( |D^i_{s,r\!-\!1}|^2 + 2\mbox{Re} \{ \overline{D^i_{s,r\!-\!1}} D^i_{r\!-\!1,s} \} \big) , \ \ \ \ \ \ \forall \ 1\leq i<r\!-\!1.

It leads to the following by arguing inductively in i in decreasing order as we have seen before:

D^s_{r\!-\!1,i}+D^s_{i,r\!-\!1}=D^i_{j,r\!-\!1}=D^i_{r\!-\!1,j}=0, \ \ \ \ \ \ \forall \ 1\leq i,j\leq r, \ \ \forall \ 1\leq s\leq n.

In particular, we have shown that

D^i_{j,r\!-\!1}=D^i_{r\!-\!1,j}=D^{r\!-\!1}_{ij} =0, \ \ \ \ \ \ \forall \ 1\leq i,j\leq r.

In other words, for D^j_{ik} with 1\leq i,j,k\leq r, it vanishes if any of the index is r or r\!-\!1. Repeating this process, we conclude that D^j_{ik}=0 for all 1\leq i,j,k\leq r. This completes the proof of the second equality of (4). To prove the third equality of (4), note that the second term on the right-hand side of (Proof) now vanishes, so it gives us D^{\alpha}_{ii}=0. If we perform a unitary change on \{ e_1, \ldots , e_r\}, we still have D^j_{ik}=0 for all 1\leq i,j,k\leq r. So what (Proof) gives us is actually D^{\alpha}_{vv}=0 for any v=v_1e_1+\cdots + v_re_r. In particular we have

D^{\alpha}_{ij}+ D^{\alpha}_{ji} =0 , \ \ \ \ \ \ \ \ \forall \ 1\leq i,j\leq r, \ \ \forall \ r\!+\!1\leq \alpha \leq n.

For any fixed 1\leq i\leq r and r\!+\!1\leq \alpha \leq n, by (?) and the first equality of (4) we have

\begin{aligned} 0 & = & 4\widehat{R}_{i\bar{i}\alpha \bar{\alpha}} \ = \ R_{i\bar{i}\alpha \bar{\alpha}} + R_{\alpha \bar{\alpha}i\bar{i}} + R_{i\bar{\alpha}\alpha \bar{i}} +R_{\alpha \bar{i}i\bar{\alpha}} \\ & = & \sum_{s=1}^r \{ |D^s_{i\alpha}|^2 - |D^i_{s\alpha}|^2- |D^{\alpha}_{si}|^2 - 2\mbox{Re} \big( D^{\alpha}_{si}\overline{D^{\alpha}_{is}} \big) \} \\ & = & \sum_{s=1}^r \{ |D^s_{i\alpha}|^2 - |D^i_{s\alpha}|^2 + |D^{\alpha}_{si}|^2 \}. \end{aligned}

Here in the last equality we used the (Proof) for the last term. If we sum i from 1 to r, then the first two terms on the right-hand side of the last line above will cancel each other, and it gives us D^{\alpha}_{si}=0 for any 1\leq s,i\leq r. This completes the proof of the third equality of (4) hence the lemma.

Uses Proof

Example

Example Example

9PV1

Consider the Heisenberg Lie algebra {\mathfrak h}_{\sigma} = U\oplus {\mathbb R}, where \sigma is a non-degenerate skew-symmetric bilinear form on U\cong {\mathbb R}^{2r}, and the Lie bracket is defined by

[(v_1, s_1), (v_2, s_2)] = (0, \sigma (v_1, v_2)).

Equivalently, it can be described as {\mathfrak h}_{\sigma}={\mathbb R}\{ X_1, \ldots ,X_r, Y_1, \ldots , Y_r, Z\} where the only non-trivial Lie brackets are given by

[X_i, Z] = a_iX_i + b_i Y_i, \ \ \ [Y_i, Z] = -b_iX_i + a_i Y_i, \ \ \ 1\leq i\leq r.

Here we assumed that \lambda_i = -(b_i + \sqrt{-1}a_i) \neq 0 for each i. Consider the Lie algebra {\mathfrak h}= {\mathfrak h}_{\sigma}\oplus {\mathbb R} equipped with an almost complex structure J given by

JW=Z, \ \ \ JZ=-W, \ \ \ JX_i=Y_i, \ \ \ JY_i = - X_i, \ \ \ 1\leq i\leq r,

where W is the unit of the {\mathbb R} factor. Then it is easy to see that J is integrable, and the commutator {\mathfrak h}'=[{\mathfrak h}, {\mathfrak h}]=U is abelian and J-invariant, thus {\mathfrak h} is 2-step solvable of pure type II.

Example

Example Example

F5RU

When r_0=r, or equivalently Y=0, then {\mathfrak g} is a (solvable) complex Lie algebra. In this case it is well-known that any metric compatiblle with the complex structure is Chern flat.

Example

Example Example

L5TS

When C^j_{ik}=0 for all 1\leq i,j,k\leq r, or equivalently when {\mathfrak g}' is abelian, then {\mathfrak g} is 2-step solvable of pure type II. In this case the first line of (?) means (?) and

[C_{\alpha}, C_{\beta}]=0, \ \ \ \ \,^t\!C_{\alpha} w_{\beta\gamma} + \,^t\!C_{\beta} w_{\gamma\alpha}+\,^t\!C_{\gamma} w_{\alpha\beta} =0, \ \ \ \ \ \forall \ r<\alpha, \beta , \gamma \leq n,

where w_{\alpha\beta} is the column vector in {\mathbb C}^r with entries C^i_{\alpha\beta}.

Example

Example Example

3K44

Fix any integers 1\leq r<n. Let V\cong {\mathbb R}^{2n} be a vector space and \{ \varepsilon_1, \ldots , \varepsilon_{2n}\} be a basis of V. Let g=\langle ,\rangle be the metric on V so that \varepsilon is orthonormal, and J the almost complex structure on V so that J\varepsilon_i = \varepsilon_{n+i}=\varepsilon_{i^{\ast}} for 1\leq i\leq n. For any 1\leq i\leq r and any r<\alpha \leq n, let x_{i\alpha}, y_{i\alpha}, u_{i\alpha}, v_{i\alpha} be constants and define Lie brackets

\begin{aligned} && [\varepsilon_i, \varepsilon_{\alpha}] = x_{i\alpha} \varepsilon_i + y_{i\alpha} \varepsilon_{i^{\ast}}, \ \ \ \ \ \ [\varepsilon_{i^{\ast}}, \varepsilon_{\alpha}] = -y_{i\alpha} \varepsilon_i + x_{i\alpha} \varepsilon_{i^{\ast}}, \\ && [\varepsilon_i, \varepsilon_{\alpha^{\ast}}] = v_{i\alpha} \varepsilon_i -u_{i\alpha} \varepsilon_{i^{\ast}}, \ \ \ \ \ [\varepsilon_{i^{\ast}}, \varepsilon_{\alpha^{\ast}}] = -u_{i\alpha} \varepsilon_i + v_{i\alpha} \varepsilon_{i^{\ast}}, \end{aligned}

while all other Lie brackets are zero. Then it is easy to verify that V becomes a Lie algebra {\mathfrak g} and J is integrable. It is also easy to see that {\mathfrak g} is 2-step solvable of pure type II and g is Chern flat.

Example

Example Example

AH19

Consider the 6-dimensional Lie algebra {\mathfrak g} equipped with a Hermitian structure (J,g), where under some unitary frame e the only non-trivial Lie brackets are

[e_1, e_2]=e_1;\ \ [e_i,e_3]=Z_ie_i,\ \ [\overline{e}_i,e_3]=Y_i\overline{e}_i, \ i=1,2.

Assume that (Y_1,Z_1)\neq (0,0) and (Y_2,Z_2)\neq (0,0). Then {\mathfrak g} is solvable, J{\mathfrak g}'={\mathfrak g}', and g is Chern flat. It is clear that {\mathfrak g}' is not abelian, so {\mathfrak g} is not 2-step solvable. It is 3-step solvable due to the low dimension here.

Source and licence

The text on this page is from arXiv:2501.00810v1 by Xin Huang and Fangyang Zheng, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2501/2501.00810.jsonl.