5
$\begingroup$

Let $k$ be a commutative ring, $A$ a $k$-algebra and $HH^n(A, M)$ the $n$-th Hochschild cohomology of $A$ with coefficients in the $A$-bimodule $M$. In the book Cyclic Homology by Loday the following curious few lines appear on page 40 when talking about the form that the coefficients can take:

For $M=A$ the groups $HH^n(A, A)$ have been extensively studied in the literature because they are related to deformation theory. But one should note that they are not functors of $A$. However if $M = A^* = \mbox{Hom}_k(A, k)$, then the groups $HH^n(A, A^*)$ are indeed functors of $A$.

Now I'm quite confused by this assertion that $HH^n(A, A)$ isn't a functor of $A$, because I know that you can interpret the $n$-th cohomology as the derived functor

$$HH^n(A, M) = \mbox{Ext}_{A^e}^n(A, M)$$

where $A^e = A\otimes_k A^{op}$ is the enveloping algebra of $A$. In fact, it's mentioned on the very next page of the book. So clearly each $HH^n(A, M)$ is a functor if I've understood this correctly. So presumably the important part is that $HH^n(A, A)$ is not a functor "of $A$", but I'm not entirely sure what this means. I'd be grateful if someone could help me understand how it is that $HH^n(A, A)$ isn't a functor of $A$.

Edit:

Thanks to the comments beneath the question (thanks for your help guys, I appreciate it) I think I have a better idea of what's going on. So using the example of $\mbox{Hom}_k(A, A)$, the reason why this isn't a functor of $A$ is because given another $k$-algebra $B$, and a map $f:A\to B$, there is no real way of assigning $f$ to a map

$$\mbox{Hom}_k(A, A)\to\mbox{Hom}_k(B, B)$$

However, now that I understand this, it seems to me $\mbox{Hom}_k(A, A^*)$ suffers from the same issue. So how is it that (as mentioned in the paragraph I cite above) $HH^n(A, A^*)$ is a functor of $A$? Am I just missing some obvious way of assigning $f:A\to B$ to $\mbox{Hom}_k(A, A^*)\to\mbox{Hom}_k(B, B^*)$?

$\endgroup$
14
  • 4
    $\begingroup$ $A$ appears in the formula $\textrm{Ext}^n_{A^e} (A, A)$ several times, but not all with the same variance. So it isn't obviously a functor of $A$. (Recall that a functor acts on objects and morphisms. Can you define the action on morphisms here?) $\endgroup$
    – Zhen Lin
    Aug 4 '20 at 11:16
  • 4
    $\begingroup$ Well, for example, $\textrm{Hom} (X, Y)$ is contravariant in $X$ and covariant in $Y$. Thus $X$ appears in the expression $\textrm{Hom} (X, X)$ with mixed variance – and this is essentially the reason why $\textrm{End} (X)$ is not a functor of $X$. Hochschild cohomology is just a souped up version of this. $\endgroup$
    – Zhen Lin
    Aug 4 '20 at 12:05
  • 5
    $\begingroup$ @SeraPhim Maybe it's easier to see in the case of $\mathrm{Hom}(X,X)$ in some category, which is basically the same issue. If you want to make this a functor, then given $f:X\to Y$ you have to define a map in one direction or the other between $\mathrm{Hom}(X,X)$ and $\mathrm{Hom}(Y,Y)$, and in general there is no reasonable way to do this. $\endgroup$ Aug 4 '20 at 16:50
  • 4
    $\begingroup$ @SeraPhim: Given a map $f \colon X \rightarrow Y$, you have an induced pullback map $f^{*} \colon \operatorname{Hom}_k(Y,Y^{*}) \rightarrow \operatorname{Hom}_k(X,X^{*})$ given by $((f^{*}(\phi))(x))(x') = (\phi(f(x)))(f(x'))$. Similarly, you can define a map $f^{*} \colon C^n(B,B^{*}) \rightarrow C^n(A,A^{*})$ which commutes with the differential and induces a map on the Hochschild cohomology. $\endgroup$
    – levap
    Aug 6 '20 at 11:36
  • 2
    $\begingroup$ The point is that dualisation $(-)^*$ is a contravariant functor. Thus $\textrm{Hom} (X, X^*)$ is purely contravariant in $X$. More to the point, functors can be composed, and we have functors $X \mapsto (X, X^*)$ and $(X, Y) \mapsto \textrm{Hom} (X, Y)$, so we have the composite functor $X \mapsto \textrm{Hom} (X, X^*)$. By contrast $X \mapsto (X, X)$ and $(X, Y) \mapsto \textrm{Hom} (X, Y)$ cannot be composed because of incompatible variance. $\endgroup$
    – Zhen Lin
    Aug 6 '20 at 11:59
2
$\begingroup$

$A$ appears in the formula $\textrm{Ext}^n_{A^e} (A, A)$ several times, but not all with the same variance. So it isn't obviously a functor of $A$.

[...] $\textrm{Hom} (X, Y)$ is contravariant in $X$ and covariant in $Y$. Thus $X$ appears in the expression $\textrm{Hom} (X, X)$ with mixed variance – and this is essentially the reason why $\textrm{End} (X)$ is not a functor of $X$. Hochschild cohomology is just a souped up version of this.

Maybe it's easier to see in the case of $\mathrm{Hom}(X,X)$ in some category, which is basically the same issue. If you want to make this a functor, then given $f:X\to Y$ you have to define a map in one direction or the other between $\mathrm{Hom}(X,X)$ and $\mathrm{Hom}(Y,Y)$, and in general there is no reasonable way to do this.

$\endgroup$

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.