13
$\begingroup$

Let $(M, \omega_M, J_M)$ and $(N, \omega_N, J_N)$ be compact Kähler manifolds. Denote $g_M=\omega_M(\cdot, J_M\cdot)$ and $g_N=\omega_N(\cdot, J_N\cdot)$.

Assume there is a diffeomorphism $\nu:M\to N$ such that $\nu^*(\omega_N)=\omega_M$, there is a diffeomorphism $\phi:M\to N$ such that $\phi^*(J_N)=J_M$ and there is an orientation-preserving diffeomorphism $\chi:M\to N$ such that $\chi^*g_N=g_M$.

Is there a diffeomorphism $\psi:M\to N$ such that $\psi^*(\omega_N)=\omega_M$ and $\psi^*(J_N)=J_M$ (and hence also $\psi^*(g_N)=g_M$)?

$\endgroup$
7
  • $\begingroup$ What do you mean by "isomorphic" Kaehler manifolds? Is isometry is implied in your definition of isomorphism? $\endgroup$
    – Qfwfq
    Commented Sep 3, 2020 at 19:27
  • $\begingroup$ @Qfwfq It probably means an isometry that preserves the complex and the symplectic structure at the same time. $\endgroup$ Commented Sep 3, 2020 at 19:28
  • 2
    $\begingroup$ On a Hermitian manifold, the symplectic and the complex structure determine the metric. $\endgroup$
    – abx
    Commented Sep 3, 2020 at 19:56
  • $\begingroup$ Take $\mathbb CP^1$ with Fubini-Studi metric of area $1$, and take any other Kaher metric on $\mathbb CP^1$ which has area $1$ but has non-constant curvature. Then $\psi$ doesn't exist, because the manfiolds are not isometric and your condition implies that $\psi$ is an isomery. But they are both symplectomorphic and obviously biholomorphic. $\endgroup$
    – aglearner
    Commented Sep 3, 2020 at 21:03
  • $\begingroup$ @alglearner: the metric on the sphere would have to be invariant under rotations of $\pi/2$ at every point (see comments to my wrong answer). Is there such a metric on the $2$-sphere, distinct from FS? $\endgroup$
    – Qfwfq
    Commented Sep 3, 2020 at 21:07

1 Answer 1

15
$\begingroup$

The answer is 'no, not necessarily'.

Consider the following example: Let $M=N=\mathbb{CP}^2$, let $(\omega_0,J_0)$ be the standard Fubini-Study Kähler structure on $M$. Now let $f$ be an arbitrary, but '$C^2$-small' smooth function on $M$, so that $\omega_0 + t\,\mathrm{i}\,\partial\bar\partial f$ is nondegenerate (and hence symplectic) for all $0\le t\le 1$.

Let $\omega_M = \omega_0 + \mathrm{i}\,\partial\bar\partial f$ and let $J_M= J_0$. Let $\omega_N=-\omega_M$ and let $J_N=-J_0$. Note that $g_M=g_N$, so $(M,g_M)$ and $(N,g_N)$ are isometric via the identity map.

When $f$ is chosen sufficiently generically, the isometry group of $g_M$ will consist of only the identity, so suppose this.

Note that $(M,J_M)$ and $(N, J_N)$ are biholomorphic, since $\mathbb{CP}^2$ is biholomorphic to its conjugate complex manifold.

By Weinstein's theorem, since $\omega_M$ and $\omega_0$ are $C^0$-close and cohomologous, there is a symplectomorphism between $(M,\omega_M)$ and $(M,\omega_0)$. Similarly, there is a symplectomorphism between $(N,\omega_N)$ and $(N,-\omega_0)$. Since, as has already been noted, $(M,\omega_0)$ and $(N,-\omega_0)$ are symmplectomorphic, it follows that $(M,\omega_M)$ and $(N,\omega_N)$ are symplectomorphic.

However, when $f$ is chosen sufficiently generically, the only map $\psi:M\to N$ that aligns the metrics $g_M$ and $g_N$ is the identity, which is neither a biholomorphism nor a symplectomorphism.

Remark: Of course, I woke up this morning and was struck by the fact that there is an even simpler example: Let $g$ be any metric on $S^2=\mathbb{CP}^1$ whose isometry group is trivial, let $J$ be one of the two $g$-orthogonal complex structures on $S^2$, and let $\omega_J$ be the associated area $2$-form on $S^2$. Let $(M,\omega_M,J_M)=(S^2,\omega_J,J)$ and let $(N,\omega_N,J_N)=(S^2,-\omega_J,-J)$.

Then $(M,\omega_M,J_M)$ and $(N,\omega_N,J_N)$ are biholomorphic, symplectomorphic, and isometric, but not isomorphic as Kähler manifolds.

Moreover, it is not difficult to choose an unramified rational curve in $\mathbb{CP}^2$ (it may have crossings, but that doesn't matter, for example, the generic irreducible cubic with one node but no cusp would certainly do) so that the metric on the normalized curve (which is an $S^2$ topologically) induced by the Fubini-Study metric on $\mathbb{CP}^2$ has no nontrivial isometries. Thus, one can even construct a pair of examples with the additional property requested by the OP that the metric be induced from the Fubini-Study metric by an immersion into $\mathbb{CP}^2$. (Using $\mathbb{CP}^3$ as a target, one could even arrange the map to be an embedding.)

$\endgroup$
8
  • $\begingroup$ You sir, are a gift to MO. $\endgroup$
    – AmorFati
    Commented Sep 5, 2020 at 4:06
  • $\begingroup$ is there a holomorphic embedding $\mathbb{C}P^2\to \mathbb{C}P^n$ that pulls back the Fubini-Study metric to $g_M$? $\endgroup$
    – user164740
    Commented Sep 5, 2020 at 8:33
  • $\begingroup$ @JoeT: No. In the first place, as constructed above, $g_M$ will generically not be real-analytic in any coordinate system, and hence cannot be the pullback of a real-analytic metric via a real-analytic embedding. More directly, $g_M$ has the same total volume as the Fubini-Study metric on $\mathbb{CP}^2$, so any holomorphic embedding into $\mathbb{CP}^n$ that induced this metric would have to be a linear embedding and hence be isometric to the Fubini-Study metric on $\mathbb{CP}^2$ itself. $\endgroup$ Commented Sep 5, 2020 at 10:11
  • $\begingroup$ @RobertBryant "...there is a symplectomorphism between $(M,J_M)$ and $(M,\omega_0)$." I think that should read $(M,\omega_M)$? Similarly in the next sentence. $\endgroup$ Commented Sep 5, 2020 at 13:07
  • $\begingroup$ @DanielPomerleano: Oh, yes, you are right, those are typos. I'll fix them now. $\endgroup$ Commented Sep 5, 2020 at 13:34

You must log in to answer this question.