9
$\begingroup$

Let $A$ be a finite-dimensional $k$-algebra and $U$ and $V$ two finite-dimensional projective $A$-modules (maybe neither the finiteness nor projectivity has to play a role, but these requirements are satisfied in my problem).

Now consider the $A$-bimodule \begin{align} M := \mathrm{Hom}_A (U,A) \otimes \mathrm{Hom}_A (A,V) \ , \end{align} where $\mathrm{Hom}_A (U,A)$ is a right $A$-module via right multiplication applied to $A$, and $\mathrm{Hom}_A (A,V)$ a left $A$-module by right action on $A$ in the source (again by right multiplication); this ensures $\mathrm{Hom}_A (A,V) =V$ as left $A$-modules.

I would like to compute the Hochschild homology $HH_*(A;M)$.

If you write down the corresponding Hochschild complex, there is an augmentation by $\mathrm{Hom}_A (U,V)$. Does that induce a quasi-isomorphism? Maybe such a result is known to experts?

Is there, more generally, a coefficient theorem I could use?

As usual, thank you for any hints.

$\endgroup$
  • 3
    $\begingroup$ To get $\operatorname{Hom}_A(U,V)$ you do need f.g.p. of $U$, although nothing of $V$. $\endgroup$ – მამუკა ჯიბლაძე Oct 29 '18 at 9:03
  • $\begingroup$ So the homology is $\mathrm{Hom}_A(U,V)$ (concentrated in degree zero) if $U$ is finitely generated projective? Can you provide an argument or a reference? $\endgroup$ – Lukas Woike Oct 29 '18 at 9:23
  • $\begingroup$ No no I don't know that much. I only wanted to say that the canonical map $\operatorname{Hom}_A(U,A)\otimes_AV\to\operatorname{Hom}_A(U,V)$ is an isomorphism for all $V$ if and only if $U$ is fgp, nothing more $\endgroup$ – მამუკა ჯიბლაძე Oct 29 '18 at 9:34
  • $\begingroup$ Ok, I will see what I can do with that. $\endgroup$ – Lukas Woike Oct 29 '18 at 9:40
  • $\begingroup$ (From how the OP is formulated, something tells me that for non-commutative rings $\mathrm{Hom}_A(A,V)$ is not isomorphic to $V$. Is that right?) $\endgroup$ – Qfwfq Oct 29 '18 at 13:40
8
$\begingroup$

(I'm assuming in the following that the base ring $k$ was a field.)

First, we note that for any $A$-bimodule of the form $M \otimes N$, where $M$ is a left $A$-module and $N$ is a right $A$-module, has an isomorphism $$ HH_*(A;M \otimes N) \cong Tor^A_*(N,M). $$ To see this, we note that there is an explicit simplicial isomorphism between the cyclic bar construction computing Hochschild homology and the two-sided bar construction computing Tor: $$ (M \otimes N) \otimes A^{\otimes p} \to N \otimes A^{\otimes p} \otimes M $$ In particular, the last map which moves the last factor of $A$ around and multiplies it on the left is carried simply to its left action on $M$.

Therefore, we find $$ HH_*(A;Hom_A(U,A) \otimes Hom_A(A,V)) \cong Tor^A_*(Hom_A(A,V), Hom_A(U,A)) \cong Tor^A_*(V, Hom_A(U,A)). $$ If $V$ is projective as a (right) $A$-module, then we find that the Tor-groups vanish for $* > 0$ and that the zero'th group is $$ V \otimes_A Hom_A(U,A). $$ The natural map augmentation that you are describing is then the natural map $V \otimes_A Hom_A(U,A) \to Hom_A(U,V)$, and (because $V$ is projective) this map is an isomorphism whenever $U$ is finitely presented. In particular, we don't need $U$ to be projective. (We could instead ask that $U$ is finitely generated projective and $V$ is arbitrary and get this result; this was stated in the comments already by მამუკა ჯიბლაძე.)

$\endgroup$
  • $\begingroup$ Out of curiosity - Tor's would also vanish for $V$ just flat, and then also I think $V\otimes_A\operatorname{Hom}_A(U,A)\to\operatorname{Hom}_A(U,V)$ will be iso for $U$ f.g., since a flat $V$ is a filtered colimit of projectives while $\operatorname{Hom}_A(U,-)$ commutes with filtered colimits if $U$ is f.g. (or does one still need f.p. for that?) $\endgroup$ – მამუკა ჯიბლაძე Oct 29 '18 at 17:37
  • 1
    $\begingroup$ @მამუკაჯიბლაძე You're absolutely correct about flatness on $V$. I think that you do need $U$ finitely presented to commute it across filtered colimits. E.g. if $U = A / (x_0, x_1, x_2, \dots)$ then you may have an element in $V$ which is annihilated by all the $x_i$ but any lift to the directed system is annihilated by only finitely many. This in fact is a characterization of being finitely presented. Life is much easier if you're Noetherian! $\endgroup$ – Tyler Lawson Oct 29 '18 at 21:10
  • $\begingroup$ If I am not mistaken finitely generated suffices for directed colimits but I am not sure which modules are directed (rather than just filtered) colimits of projectives $\endgroup$ – მამუკა ჯიბლაძე Oct 29 '18 at 21:55

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.