This is a somewhat embarrassing question, but still I will ask it. Let $V$ be a vector space over $\mathbb C$ of dimension $d$. Let $X$ be the dg-preimage of $0$ under the natural map $V\to Sym^2(V)$ (i.e. impose equations of the form $f=0$ for all homogeneous polynomials of degree 2 on $V$; we have $d\choose{2}$ equations, so $X$ always has a non-trivial dg-part for $d>1$).

$\mathbf{Question:}$ What does the cotangent complex of $X$ look like? More precisely, I want to know its restriction to the unique $\mathbb{C}$-point of $X$. I would be happy to understand the case $d=2$.


Well, unless I am mistaken, whenever you have a cdg ring given by dg generators and relations, then the cotangent space at zero is simply the cone of the differential $T^*\text{Spec}(S^*{\text{generators}})\to T^*\text{Spec}(S^*{\text{relations}})$ at the corresponding point. In this case, as $d(\text{anything quadratic}) = 0$ at $0\in V$, this should give a complex with zero differential, $V\oplus S^2(V)[-1]$ at the point (here $S^2(V^*)[1]$ means it's in degree $1$, not a shift).

More concretely in this case, you have a Koszul CDG-free resolution for $X$ as follows: $A = \left(\bigwedge^* S^2(V)\otimes S^*[V], d\right)$ with $S^2(V)$ in degree -1, with differential determined on the degree-minus-one generators by $d(v\otimes w) = vw\in S^*(V)$. It's more intuitive for me to understand the dual problem of computing the tangent rather than the cotangent complex, i.e. derivations $A\to \mathbb{C}$. As a graded space, this is dual to the space of generators, so $V^*\oplus S^2(V^*)[-1]$, where the differential $d:V^*\to S^2(V^*)$ is $d(\partial_\xi) = [d, \partial_\chi] = -\partial_\xi\circ d_A$ acts on $v w\in S^2(V)$ by $d(\partial_\xi)(vw[-1]) = \partial_\xi(vw[0]) = v\partial_xi(w) \pm w \partial_\xi(v)$ (here I'm using the crude notation $[0], [-1]$ to distinguish degree 1 and degree 0 elements), and as both $\partial_\xi(v), \partial_\xi(w)\in \mathbb{C}$ are killed by $V$, this just gives us $d=0,$ in the tangent complex hence the desired result.

  • $\begingroup$ Thanks, I think you are right, except that one must formulate more carefully what "given by dg generators and relations" means (some smoothness condition should be there). $\endgroup$ – Alexander Braverman Apr 26 '16 at 9:52

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.