I want to understand why Kähler manifolds are formal.
This was first proved by Deligne, Griffiths, Morgan, Sullivan

Let $\mathcal M$ be a minimal differential algebra and $H^*(\mathcal M)$ the cohomology of $\mathcal M$, viewed as a differential algebra with $d =0$.


  1. $\mathcal M$ is formal if there is a map of differential algebras $\psi: \mathcal M \to H^*(\mathcal M)$ inducing the identity on cohomology.
  2. The homotopy type of a differential algebra $\mathcal A$ is a formal consequence of its cohomology if its minimal model is formal.
  3. The real (or complex) homotopy type of a manifold $M$ is a formal consequence of the cohomology $M$ if the de Rham homotopy type of the real (or complex) forms $\mathcal E$ is a formal consequence of its cohomology.

In section 6, the following (main) theorem is proved:

Let $M$ be a compact complex manifold for which the $dd^c$-lemma holds (e.g. $M$ Kähler, or $M$ a Moisezon space). Then the real homotopy type of $M$ is a formal consequence of the cohomology ring $H^*(M; \mathbb R)$

Let $\{\mathcal E^*_M,d\}$ be the de-Rham complex on $M$, $\{\mathcal E^c_M,d\}$ the subcomplex of $d^c$-closed forms and $\{H_{d^c},d\}$ the quotient complex $\mathcal E_M^c/d^c \mathcal E_M$.

Using the $dd^c$-lemma ($\partial \bar\partial$-lemma), it is an easy calculation, that the natural maps $$\{\mathcal E^*_M,d\} \stackrel i\leftarrow \{\mathcal E^c_M,d\} \stackrel p\to \{H_{d^c},d\} $$ are quasi-isomorphisms and that the differential on $H_{d^c}(M)$ vanishes.

In the proof, the theorem follows immediately from the above. I don't see how and I am a little confused that there were no reasons given why the theorem follows. Only "This proves the claim and consequently [part (1)] of the theorem".

The maps seem to be maps of differential algebras, so this is fine.
But as far as I see, the theorem only follows if we can replace $\{H_{d^c},d=0\}$ with $\{H_M,d=0\}$.

Update/ Solution:
I repeat: The theorem only follows if we can replace $\{H_{d^c},d=0\}$ with $\{H_M,d=0\}$.
But the above argument shows, that the cohomologies are isomorphic. More precisely, the isomorphism is induced by $i$ and $p$.


A dg-algebra with vanishing differential is its own cohomology. The morphisms $i_*,\rho_*$ are isomorphisms between the cohomology of $(\mathscr{E}^*_M,d)$ -- which is $H^*(M)$ -- and the cohomology of $(H_{d^c}, d = 0)$ -- which is $H_{d^c}$, because the differential vanishes.

More generally if a dg-algebra $A$ is quasi-isomorphic to any dg-algebra $B$ with vanishing differential, then $B$ is isomorphic to the cohomology of $A$, pretty much by definition; hence $A$ is formal.

  • $\begingroup$ Thank you. I already had everything in mind what you said, for some reason I just failed to make the last (surprisingly easy) step. It's clear now. $\endgroup$ – klirk Dec 7 '18 at 23:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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