# Multiplicative Structures On Free Resolutions

Hello,

this question is related to Differential graded structures on free resolution?.

Given a regular local ring $S$ and $f\in{\mathfrak m}_S\setminus\{0\}$, I am interested in studying $R$-modules through their $S$-free resolutions. More precisely, given an $R$-module $M$, any $S$-free resolution $F^{\ast}$ of $M$ admits homotopies $s_n$ of respective degree $2n-1$ such that $s_0$ equals the differential, $s_1$ is a nullhomotopy for the multiplication with $f$ and for the higher $s_n$ we have

$s_0 s_n + s_1 s_{n-1} + ... + s_n s_0 = 0$ for $n\geq 2$.

If $s_1$ can be chosen in such a way that $s_1^2=0$, we can consider $F^{\ast}$ as a dg-module over the Koszul dg-S-algebra of $S/f$.

Now I have two questions:

(1) What happens if the $s_1$ can not be chosen such that $s_1^2=0$? Can we still write the datum of higher homotopies $s_n$ as a dg-module structure over some dg-resolution of $S/f$? A naive guess would be a free non-commutative dg-algebra generated by elements $s_n$ of degree $2n-1$ such that $\text{d}(s_n) = s_1 s_{n-1} + ... + s_{n-1} s_1$ for $n\geq 2$ and $\text{d}(s_1) = f\cdot 1$. Is this studied anywhere?

(2) Can we do all this somehow functorially in $M$? I'm thinking of something like the canonical functor from R-mod into the derived category, which after identification with the homotopy category of projectives turns "projective resolution" into a functor. However, in the concrete example I'm struggling with extending a morphism between R-modules to a morphism of free resolutions respecting the chosen homotopies. Perhaps this is just some diagram chase, but I don't see it.

Apart from technicalities, my goal is to study $S/f$-modules through dg-modules over an appropriate dg-$S$-algebra up to homotopy. If this is possible: do you know books or articles where it is treated?

Thank you very much!

Hanno

-
Whenever f is not a zero divisor, your "free nonassociative" resolution is indeed weakly equivalent to S/f, and freeness allows it to automatically lift to the endomorphism algebra of a free resolution of any S/f-module. So far as functoriality, whether you use the Koszul algebra or the free nonassociative algebra there are functorial resolutions of M that you want - the obstructions only come if you're trying to work with a specific resolution such as a minimal resolution. In fact, S/f, the Koszul algebra, and the nonassoc. alg. all have equivalent derived categories (Quillen equivalence!). – Tyler Lawson Nov 29 '10 at 16:32
That sounds very intersting, Tyler, thank you! Could you give references? – Hanno Becker Nov 29 '10 at 18:27
Unfortuantely I don't remember references immediately - esp for the equivalence between the Koszul algebra and the nonassociative one, I only know the argument - but am sure someone else probably does. So far as functorial resolutions, a standard technique called the "two-sided bar construction" using the forgetful-free adjoint pair between A-modules and sets (or A-modules and some base algebra over which S and S/f are flat) works. And model categories/Quillen equivalences have quite a number of references now. If there are specifics please let me know and I'll try to track something down. – Tyler Lawson Nov 29 '10 at 19:14
Ok, I will ask more specifically, hopefully it's not too basic: First, what do you mean by the "free nonassociative resolution"? The dg-algebra I thought of is the free associative algebra with generators and relations given above, but this doesn't seem to be what you are talking about. Secondly, what's the precise relation to the general two-sided bar construction you mentioned, und what are good references for the latter? – Hanno Becker Nov 30 '10 at 0:17
For some blasted reason I have written the word "nonassociative" absolutely everywhere, when I instead mean "associative (noncommutative)" - and obliviously have not figured out why until now. Sorry. – Tyler Lawson Dec 4 '10 at 13:43

You can always find a free resolution $F$, of $M$ over $S$, such that $F$ is a dg-module over the Koszul complex. It may not be minimal, but in many cases that's not an issue. This is the path taken by Avramov and Buchweitz in their paper "Homological algebra modulo a regular sequence with special attention to codimension 2". In particular these resolutions can be taken functorially.