6
$\begingroup$

Following 4.6 and 4.7 of this paper by Seidel and Thomas, a graded algebra $A$ is called intrinsically formal if any two dgas with cohomology $A$ are quasi-isomorphic. There is a sufficient condition that says for an augmented graded algebra $A$, if the Hochschild cohomology $HH^q(A, A[2 − q]) = 0$ for all $q > 2$, then $A$ is intrinsically formal.

Is it possible to verify this condition for the exterior algebra on a vector space $V$? To be clear, it could be either $Sym(V[1])$ or $Sym(V[-1])$. The characteristic and the dimension of $V$ might be important. I'm focusing on char$=0$ or $2$ and dimensions $1$ and $2$.

We may also regard the exterior algebra as a superalgebra ($\mathbb{Z}/2$-graded algebra) so that $Sym(V[1])$ and $Sym(V[-1])$ are the same. Is there a similar sufficient condition for intrinsically formality (we also consider $\mathbb{Z}/2$-graded dgas)?

There is a related question on the computation, but I'm not sure if the answer is enough to verify the condition. There are also related papers 1 2 3, but all of them seem to deal with the exterior algebra as an ungraded algebra.

$\endgroup$
3
  • 1
    $\begingroup$ In what homological degree do you put $V$ when you define the "exterior algebra"? In all contexts where we deal with cohomology and homotopy, it is really much better to say "symmetric algebra on the degree shift of $V$" and specify the degree shift, to avoid ambiguity. $\endgroup$ Commented Mar 4 at 19:12
  • $\begingroup$ I care about both cases. Thanks for comments. I edited the question. $\endgroup$
    – Faniel
    Commented Mar 4 at 22:17
  • $\begingroup$ The question you link contains a counterexample to formality, so the sufficient condition doesn't hold. $\endgroup$ Commented Apr 1 at 9:19

2 Answers 2

4
$\begingroup$

It's not intrinsically formal. Let $A=\Lambda(V)$ be the exterior algebra on an $n$-dimensional vector space $V$, it follows from the Hochschild-Konstant-Rosenberg isomorphism that

$\mathit{HH}^q(A,A[2-q])\cong\mathit{Sym}^q(V^\vee)\otimes\Lambda^2(V),$

which is large for $q>2$.

However, the following is true:

If $V\cong\mathbb{C}^2$, then a non-trivial $A_\infty$-deformation of $A=\Lambda(\mathbb{C}^2)$ cannot be a cyclic $A_\infty$-algebra. Roughly speaking, it means that the $A_\infty$-deformations do not preserve chain level Poincare duality. More general discussions can be found here https://arxiv.org/abs/2106.07692, Section 1.3.

$\endgroup$
1
  • $\begingroup$ Thanks a lot for your answer! The cyclic condition seems interesting. Does the argument about $\mathbb{C}^2$ also works for $\mathbb{Z}/2$-graded case? I looked the reference and it seems that lemma 4.5 is the essential argument, but I'm not sure about the connection between it and what you are saying. Could you explain more on this case or suggest more references? $\endgroup$
    – Faniel
    Commented Mar 25 at 19:36
3
$\begingroup$

Let me focus on the $\mathbb Z$-graded picture in characteristic $0$. (You can collapse the $\mathbb Z$-grading to a $\mathbb Z / 2$-grading if you like, but one should be a bit careful when generalizing this to characteristic $2$.)

The Hochschild cohomology groups $\mathrm{HH}^q (A, A [2-q])$ control higher multiplications $m_q$ which form part of an A$_\infty$ structure on $A$. The vanishing for $q > 2$ implies that there cannot be any cocycles defining higher multiplications which are not coboundaries, i.e. there are no "interesting" higher structures on $A$. This is one way of understanding intrinsic formality.

For the exterior algebra (or indeed any Koszul algebra) a useful trick can be to notice that the exterior algebra on $n$ generators of degree 1, say, is Koszul-dual to the symmetric algebra on $n$ generators of degree 0. But Koszul-dual algebras have "the same deformation theory". In fact, their (bigraded) Hochschild complexes are "isomorphic in the homotopy category of B$_\infty$ algebras", which implies in particular, that their DG Lie structures given by Hochschild differential plus Gerstenhaber bracket are L$_\infty$ quasi-isomorphic. This is a result of Keller [1]. (You can find a short summary of the general theory in §4 of [2].)

The Hochschild cohomology groups $\mathrm{HH}^q (A, A [2-q])$ comprise the Hochschild cohomology of $A$ in total degree 2. On the Koszul-dual side this just matches the usual Hochschild cohomology in degree $2$, i.e. $\mathrm{HH}^2 (A^!, A^!)$, whenever $A^!$ is trivially graded (which is the case for $A^! = \mathrm{Sym} (V^*)$). For the symmetric algebra, the latter is just $\mathrm H^0 (\Lambda^2 \mathcal T_{\mathbb A^n})$ via the HKR isomorphism, as noted in the answer by YHBKJ.

You can therefore understand A$_\infty$ deformations of the exterior algebra $\mathrm{Sym} (V [-1])$ in terms of associative deformations of the polynomial ring $\mathrm{Sym} (V^*)$, which are nothing but deformation quantizations of affine $n$-space. This link between A$_\infty$ deformations and quantizations was studied by Calaque, Felder and Rossi [3].

Under this isomorphism, the condition $q > 2$ just turns into polynomial degree $> 2$. That is, an A$_\infty$ deformation of the exterior algebra using cocycles defining classes in $\mathrm{HH}^q (A, A[2-q])$ for $q > 2$ correspond to deformation quantizations of Poisson structures of degree $\geq 3$.

Now, to come to your examples:

  • If $\dim V = 1$, then $A = \mathrm{Sym} (V [-1]) = \Bbbk [X] / (X^2)$ with $|X| = 1$ and $A^! = \Bbbk [x]$ with $|x| = 0$. We have $\mathrm{HH}^q (A, A[2-q]) \simeq \mathrm{H}^0 (\Lambda^2 \mathcal T_{\mathbb A^1})_q = 0$ for all $q$ (where the subscript $_q$ means "bivector fields with homogeneous degree $q$ coefficients").
  • If $\dim V = 2$, then $A \simeq \Bbbk \langle X, Y \rangle / (X Y + Y X, X^2, Y^2)$ and $\mathrm{HH}^q (A, A[2-q]) \simeq \mathrm{H}^0 (\Lambda^2 \mathcal T_{\mathbb A^2})_q \simeq \Bbbk^{q \choose 2}$, reflecting the fact that on $\mathbb A^2$ with coordinates $u, v$ any bivector is of the form $f \frac{\partial}{\partial u} \wedge \frac{\partial}{\partial v}$, and the space of homogeneous degree $q$ polynomials in $u, v$ has dimension $q \choose 2$.
    Note that in dimension $2$ there are no obstructions since $\mathrm{HH}^3 (A^!, A^!) \simeq \mathrm H^0 (\Lambda^3 \mathcal T_{\mathbb A^2}) = 0$, so you can essentially define higher multiplications $m_k$ on the generators freely, for example there is an A$_\infty$ structure determined by setting, say, $m_3 (X, Y, Y) = X Y$ (which is indeed of degree $2 - 3 = -1$), corresponding to the quantization of the Poisson structure determined by $\{ y, x \} = x y^2$, etc.

In higher dimensions, this will get more and more involved, as you can see from the Koszul-dual side, where deformation quantizations of (polynomial) Poisson structures is a nontrivial problem.

[1] Keller, Bernhard, Derived invariance of higher structures on the Hochschild complex
[2] Barmeier, Severin; Wang, Zhengfang, A$_\infty$ deformations of extended Khovanov arc algebras and Stroppel's Conjecture
[3] Calaque, Damien; Felder, Giovanni; Rossi, Carlo A., Deformation quantization with generators and relations, J. Algebra 337, No. 1, 1-12 (2011). ZBL1235.53095.

$\endgroup$

You must log in to answer this question.