MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let M be the $sl(n,C)$-representation of the inclusion $sl(n,C)\hookrightarrow gl(n,C)$. Let q be a symbol. $f(q)=1-M q + \wedge^2Mq^2-...+(-1)^n\wedge^nMq^n$

$g(q)=\sum_{i=0}^\infty Sym^iM \; q^i$

I want to prove that $f(q)g(q)=1$ which is equivalent to some isomorphism between many representations.

I am not sure if this choice of M is essential. Will any n-dim representation be enough? The worst method might be computing characters with character formulae...

btw product is tenor product. sum is direct sum.

share|cite|improve this question
Does this really have anything to do with geometric Langlands? – darij grinberg Nov 17 '12 at 1:34
BTW I'm interested in characteristic-free proofs (or refutations...) of this $f\left(q\right)g\left(q\right)=1$ equality (viewed as an equality in the ring of formal power series in $q$ over the Grothendieck ring). – darij grinberg Nov 17 '12 at 1:35
This should be Koszul duality for the symmetric algebra on $M$ in the category of $sl(n)$-modules, or something. – Mariano Suárez-Alvarez Nov 17 '12 at 1:44
Mariano, could you explain this to me? I happen to need some nice examples to understand Koszul duality... – darij grinberg Nov 17 '12 at 1:50
up vote 2 down vote accepted

It's easy to show for any $M$ using your "worst method," though in the definition of $f(q)$ we need to replace $n$ with $m=\dim M$. Let $A\in\mathrm{sl}(n,C)$ and suppose that $M\cdot A$ has eigenvalues $\theta_1,\dots,\theta_m$. Then the trace of $A$ acting on $f(q)$ is $(1-\theta_1 q)\cdots (1-\theta_m q)$, while on $g(q)$ is $1/(1-\theta_1 q)\cdots (1-\theta_m q)$. Since the character of a (finite-dimensional) representation of $\mathrm{sl}(n,C)$ determines the representation (up to equivalence), it follows that $f(q)g(q)=1$ as virtual representations.

share|cite|improve this answer
yes, good, this is elementary enough. – Jeep Wrangler Nov 17 '12 at 22:08

Let $SM$ be the symmetric algebra on $M$. There is an exact sequence $$0\to SM\otimes\Lambda^nM\to SM\otimes\Lambda^{n-1}M\to\cdots SM\otimes\Lambda^2M\to SM\otimes M\to SM\to k$$ which is just the Koszul complex for $SM$. After appropriately shifting the degrees of the modules, this is an exact complex of $\mathbb Z$-graded $sl(n)$-modules. Taking the Euler characteristic of the Hilbert series with coefficients in the Grothendieck ring of finite dimensional $sl(n)$-modules should prove your identity.

share|cite|improve this answer
Very nice! So this gives a simple proof over any field or ring and for modules over any cocommutative Hopf algebra instead of $\mathfrak{sl}\left(n\right)$-modules. Only one question: where is Koszul duality here? I just see a Koszul complex. – darij grinberg Nov 17 '12 at 2:16
Well, Koszul duality is the relationship between a Koszul quadratic algebra $A$ and its Koszul dual $A^!$, which can be described in many ways, one of which is that $A\otimes A^!$ can be made, by grading appropiately, into a resolution of the trivial module for $A$ (or for $A^!$, for that matter) When $A$ is the symmtric algebra on a vector space, then $A^!$ is the exterior algebra, and the corresponding compex is the Koszul complex (which is an $SM$-projective resolution of $k$; with a different grading, it becomes a $\Lambda M$-projective resolution of $k) – Mariano Suárez-Alvarez Nov 17 '12 at 2:23
I didn't understand this complex yet. Can you give any reference about the maps in this complex? – Jeep Wrangler Nov 17 '12 at 22:08

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.