# “as close to being semisimple as it can possibly be.”

I had originally asked this question on math stack exchange but I think maybe it's more appropriate to ask it here.

In the paper of Beilinson, Ginzburg and Soergel entitled "Koszul Duality Patterns..." After stating the following proposition, they make the statement: " a Koszul ring is a positively graded ring that is "as close to being semisimple as it can possibly be.""

Proposition: Let $A = \bigoplus_{j\geq 0} A_j$ be a positively graded ring and suppose that $A_0$ is semisimple. The following are equivalent.

1) $A$ is Koszul

2) For any two pure $A$-modules $M$, $N$ of weights $m$, $n$ respectively we have $ext^i_A (M,N) = 0$ unless $i = m-n$.

3) $ext_A ^i (A_0, A_0 \langle n \rangle) = 0$ unless $i=n$.

To clarify some things: A graded module $M$ over a graded ring, is pure of weight $m$ iff $M = M_{-m}$. Also the $ext$'s we are using are in the category of graded modules.

I am very new to all this and I don't even have a vague idea of what they mean. Any thoughts at all would be super helpful.

-
Why did you delete your question on math stack ? – Dietrich Burde Aug 2 '13 at 13:23
Let $M$ and $N$ be pure. In the semisimple case $ext^i_A (M,N) = 0$ for all $i$. In the Koszul case $ext^i_A (M,N) = 0$ for all $i$ except possibly one value. I guess this is what they mean by the phrase in the title. – Dag Oskar Madsen Aug 2 '13 at 13:41
@Dietrich Thanks for your responses. As to why I deleted the question on math stack, when I deleted it there were no comments or answers to the question and I thought it would be easier to just have one version of the question up. – Anette Aug 2 '13 at 14:10
@Dag Thanks for your response! – Anette Aug 2 '13 at 14:10
To make my previous comment more precise, I should add that if $M$ and $N$ are pure of the same weight, then you can have $\mathrm{ext}^0_A (M,N) \neq 0$ also in the semisimple case. – Dag Oskar Madsen Aug 2 '13 at 18:26

In the semisimple case it is really easy to calculate $ext_A^i(M,N)$, $i\ge 1$, with the above assumptions. It is zero. For Koszul rings this is almost true, i.e., $ext^i(M,N)$ is concentrated in degree $i$ (that is, if $S$ is the direct sum of all simples, then the two natural gradings on the algebra $A^{!}=ext^{\bullet}(S,S)$ coincide). It seems reasonable that the authors had this in mind as they wrote: "Morally a Koszul ring is a graded ring that is as close to semisimple as a $\mathbb{Z}$-graded ring possibly can be”.