Let $\mathcal{C}$ be a nice $k$-linear abelian category (the example I have in mind is the category of coherent sheaves on a smooth projective variety over $\mathbb{C}$). Let $B \in D^{b}(\mathcal{C})$ such that the DG-algebra $\mathrm{RHom}(B,B)$ is formal (that is quasi-isomorphic, as a DG-algebra to its cohomology algebra).

I would like to know if there exists a complex $B^{\bullet}$ representing $B$ such that: $$\mathrm{Hom}^{\bullet}(B^{\bullet}, B^{\bullet}) = \mathrm{Ext}^{\bullet}(B,B),$$ where the sign equal really means equality (and not just quasi-isomorphism of DG algebras)?

Here $\mathrm{Hom}^{\bullet}(B^{\bullet}, B^{\bullet})$ is the DG-algebra defined by: _$\mathrm{Hom}^{k}(B^{\bullet}, B^{\bullet})$ is the vector space of sequences of maps $ f^p : B^{p} \rightarrow B^{p+k}$,

_the multiplication map in $\mathrm{Hom}^{\bullet}(B^{\bullet}, B^{\bullet})$ is the composition of maps,

_the differential is the one inherited from that of $B^{\bullet}$.

Thanks a lot!


The following seems to be a counterexample. Let $\mathcal C$ be the category of $k[x]$-modules and $B=k=k[x]/(x)$.

The derived endomorphism DG-algebra of $B$ can be computed as $\operatorname{RHom}(B,B)=\operatorname{Hom}^\bullet(X^\bullet,X^\bullet)$, where $X^\bullet$ is the complex

\[\cdots\rightarrow 0\rightarrow k[x]\stackrel{x}\longrightarrow k[x]\rightarrow 0\rightarrow \cdots\]

Its cohomology $H^\bullet\operatorname{RHom}(B,B)=\operatorname{Ext}^\bullet(B,B)=\Lambda(t)=k\cdot 1\oplus k\cdot t$ is the exterior algebra on a degree $1$ generator and there is a unique quasi-isomorphism of $k$-DGAs $\Lambda(t)\rightarrow \operatorname{Hom}^\bullet(X^\bullet,X^\bullet)$ sending $t$ to the cochain map $X^\bullet\rightarrow X^\bullet[1]$ which is the identity in $k[x]$ in degree $0$ and trivial otherwise.

If there were a complex $B^\bullet$ with $\operatorname{Hom}^\bullet(B^\bullet,B^\bullet)=\Lambda (t)$ then $\operatorname{Hom}^0(B^\bullet,B^\bullet)$ would be $1$-dimensional as a $k$-vector space. This forces $B^\bullet$ to be globally $1$-dimensional over $k$. Since $B^\bullet$ is quasi-isomorphic to $B$, it must be $B^\bullet=B=k$, and in this case $\operatorname{Hom}^1(B,B)=0\ncong k$.

  • 1
    $\begingroup$ Super! Thank you. $\endgroup$ – Libli Oct 24 '15 at 13:26

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.