MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $M$ be a Kaehler manifold with positive holomorphic sectional curvature. then the maximum of sectional curvatures at point $p$ is assumed at the holomorphic planes. I read this claim in Klingenberg's paper "On Compact Kaehlerian Manifolds with Positive Holomorphic Curvature" He refers this result to Berger "Pincement Riemannien et pincement holomorphe"

But I don't have access to Berger's paper nor do I read French, so I ask here to see if this can be proved very quickly or has been written in somewhere else.

share|cite|improve this question
the paper of Berger is here: – YangMills Dec 24 '12 at 19:01
the proof that you want is on pages 154-155 there (even if you can't read French, I think you'll have no problem following the proof). Note that some inequalities on page 155 are wrong, and are corrected in the erratum (linked above). – YangMills Dec 24 '12 at 19:09
@yangMills, That's fantastic! Merry Christmas – Ralph Dec 24 '12 at 22:23
up vote 12 down vote accepted

There is a more enlightening proof of this statement than Berger's calculation and, in fact, it proves something a bit more general. First, a definition: Let $(M,g)$ be a Riemannian $n$-manifold with Riemann curvature tensor $R$ and let $E\subset T_xM$ be a $p$-plane with orthonormal basis $e_1,\ldots,e_p$. Define $$ \sigma(E) = \sum_{i,j=1}^p R(e_i,e_j,e_j,e_i), $$ which is easily seen not to depend on the choice of orthonormal basis. The quantity $\sigma(E)$ is the scalar curvature of $E$. For $p=2$, this is twice the sectional curvature of the $2$-plane $E$, while $\sigma(T_xM)$ is simply the scalar curvature of $g$ at $x$. In general, the scalar curvature of $E$ is $p(p{-}1)$ times the average of the sectional curvatures of the $2$-planes that lie in $E$.

Fact: If $(M,g,J)$ is a Kähler manifold, and $E\subset T_xM$ is a $4$-plane that is complex (i.e., stable under $J$), then $\sigma(E)$ is $6$ times the average sectional curvature of the complex $2$-planes that lie in $E$. (See below for a proof.)

Proposition: If $(M,g,J)$ is a Kahler manifold and $E\subset T_xM$ is a $4$-plane that is complex and satisfies $\sigma(E)>0$, then the only local maxima of the sectional curvature function on $\mathrm{Gr}_2(E)$ occur at $2$-planes $L$ that are complex.

Remark: If the holomorphic sectional curvatures of $M$ are positive, then, by the Fact, $\sigma(E)>0$ for all $4$-planes $E$ that are complex, and hence, by the Proposition, the maximum of the sectional curvature on $\mathrm{Gr}_2(T_xM)$ is attained only by complex $2$-planes (since each noncomplex tangent $2$-plane lies in some tangent $4$-plane that is complex). Meanwhile, one can have $\sigma(E)>0$ for all $4$-planes that are complex without having positive holomorphic sectional curvature; just consider a Kähler surface with positive scalar curvature that does not have positive holomorphic sectional curvature. Thus, this gives a stronger result than Klingenberg's claim.

To prove the Fact and the Proposition, it clearly suffices to do it for a Kähler manifold of complex dimension $2$ with positive scalar curvature, in which case $E=T_xM$ for some $x\in M$ that one can suppose fixed for the purposes of this discussion. This is a local argument, so one can choose an orthonormal coframing $\omega_0,\ldots,\omega_3$ such that the Kähler form is given by $\Omega=\omega_0\wedge\omega_1+\omega_2\wedge\omega_3$. Let $\phi_{ij}=-\phi_{ji}$ be the connection forms, so that $d\omega_i=-\phi_{ij}\wedge\omega_j$ and, because of the Kähler condition, $\phi_{02}+\phi_{31}=\phi_{03}+\phi_{12}=0$. Then, by the Bianchi identities for a Kahler metric, one has the following expression for the curvature forms $\Phi_{ij}=d\phi_{ij}+\phi_{ik}\wedge\phi_{kj}\ $: $$ \begin{pmatrix} \Phi_{01}+\Phi_{23}\cr \Phi_{02}+\Phi_{31}\cr \Phi_{03}+\Phi_{12}\cr \Phi_{01}-\Phi_{23}\cr \Phi_{02}-\Phi_{31}\cr \Phi_{03}-\Phi_{12} \end{pmatrix} = \begin{pmatrix} 3s&0&0&r_1&r_2&r_3\cr 0&0&0&0&0&0\cr 0&0&0&0&0&0\cr r_1&0&0&s{+}w_{11}&w_{12}&w_{13}\cr r_2&0&0&w_{12}&s{+}w_{22}&w_{23}\cr r_3&0&0&w_{13}&w_{23}&s{+}w_{33} \end{pmatrix} \begin{pmatrix} \omega_0\wedge\omega_1+\omega_2\wedge\omega_3\cr \omega_0\wedge\omega_2+\omega_3\wedge\omega_1\cr \omega_0\wedge\omega_3+\omega_1\wedge\omega_2\cr \omega_0\wedge\omega_1-\omega_2\wedge\omega_3\cr \omega_0\wedge\omega_2-\omega_3\wedge\omega_1\cr \omega_0\wedge\omega_3-\omega_1\wedge\omega_2 \end{pmatrix} $$ Here, $12s$ is the scalar curvature (which is positive by hypothesis) and $w_{ij}=w_{ji}$ satisfies $w_{11}{+}w_{22}{+}w_{33}=0$. [The reader may recognize this as the classic presentation of the curvature of a Riemannian $4$-manifold due to Singer and Thorpe; in this case, it has been adapted to the case of a Kähler metric.]

Define functions $u_i$ and $v_i$ on the Grassmann bundle $\mathrm{Gr}^+_2(M)$ of oriented tangent $2$-planes as follows: If $L$ is an oriented $2$-plane with oriented orthonormal basis $(X,Y)$, then $$ \begin{pmatrix} u_1(L)\cr u_2(L)\cr u_3(L)\cr v_1(L)\cr v_2(L)\cr v_3(L) \end{pmatrix} = \begin{pmatrix} \omega_0\wedge\omega_1+\omega_2\wedge\omega_3\cr \omega_0\wedge\omega_2+\omega_3\wedge\omega_1\cr \omega_0\wedge\omega_3+\omega_1\wedge\omega_2\cr \omega_0\wedge\omega_1-\omega_2\wedge\omega_3\cr \omega_0\wedge\omega_2-\omega_3\wedge\omega_1\cr \omega_0\wedge\omega_3-\omega_1\wedge\omega_2 \end{pmatrix} (X,Y). $$ One then has ${u_1}^2+{u_2}^2+{u_3}^2={v_1}^2+{v_2}^2+{v_3}^2=1$, and the map $(u,v):\mathrm{Gr}^+_2(M)\to S^2\times S^2$ carries each fiber of $\mathrm{Gr}^+_2(M)\to M$ diffeomorphically onto $S^2\times S^2$. Note that a $2$-plane $L$ is complex if and only if $u_2(L)=u_3(L)=0$, i.e., if and only if $u_1(L)=\pm1$.

Now, computing from the definitions, one finds $$ \sigma(L) = 3s\ {u_1}^2 + 2u_1(r_iv_i) + s + w_{ij}v_iv_j\ . $$ (The proof of the Fact about the average of sectional curvatures of complex $2$-planes in $E$ follow immediately from this formula.)

Now, $\sigma(L)$ does not depend on $u_2$ or $u_3$, and we can thus think of $\sigma$ restricted to a single fiber $\mathrm{Gr}^+_2(T_xM)$ of $\mathrm{Gr}^+_2(M)\to M$ as a function on $[-1,1]\times S^2$ instead of $S^2\times S^2$. Because $s>0$, when one fixes a $v\in S^2$, the quadratic function of $u_1$ above cannot attain a local maximum when $-1 < u_1 <1 $. It follows that the local maxima of $\sigma$ on $\mathrm{Gr}^+_2(T_xM)$ must occur only at those $L$ for which $u_1(L)=\pm1$, i.e., those $L$ that are complex. QED

share|cite|improve this answer
wow!........... – Gil Kalai Dec 25 '12 at 15:25
This is really a holiday gift! – Ralph Dec 26 '12 at 22:24
@Robert, I am confused by "$\sigma(E)$ is 6 times the average sectional curvature of the complex 2-planes that lie in $E$". Suppose $e_1, Je_1, e_2, Je_2$ be a o.n. basis of $E$. Then the $\sigma(E)$ involves all sectional curvatures including the space $L:= span(e_1, e_2)$, not necessary just holomorphic ones $L_1:=span(e_1, Je_1)$ and $L_2:=span(e_2, Je_2)$. Then how to get the average only over complex 2-planes? – Ralph Dec 26 '12 at 22:45
@unknown (google): The complex $2$-planes (oriented positively) that lie in the $4$-plane $E$ are mapped by $(u,v)$ to the subset $\lbrace(1,0,0)\rbrace\times S^2$, and on this set, which can be identified with the $2$-sphere in the $v$-variables, we have $\sigma=4s+2(r_iv_i)+w_{ij}v_iv_j$. (Remember that $\sigma(L)$ is twice the sectional curvature of $L$.) Since $w_{11}+w_{22}+w_{33} =0$, averaging this $\sigma$ over the unit $2$-sphere in $v$-space results in the value $4s$, so it follows that the average scalar curvature of the complex $2$-planes is $2s$. However, $\sigma(E)=12s$. QED – Robert Bryant Dec 27 '12 at 1:19
@unknown (google): Oops! I should have written 'the average sectional curvature of the complex $2$-planes [in $E$] is $2s$'. Sorry about the slip. – Robert Bryant Dec 27 '12 at 1:39

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.