Let $R$ be a noetherian ring (perhaps we have to assume $2 \in R^*$), $1 \leq d \leq n$ and consider the Plücker embedding $\omega : G \hookrightarrow P$, where $G = \text{Grass}_d(R^n)$ and $P = \mathbb{P}(R^{\binom{n}{d}}) = \text{Proj}(R[\{x_{i_1,...,i_d}\}])$. On $X$-valued points points, this maps a quotient $s: \mathcal{O}_X^n \twoheadrightarrow \mathcal{F}$, where $\mathcal{F}$ is locally free of rank $d$, to the $d$th exterior power $x : \mathcal{O}_X^{\binom{n}{d}} \twoheadrightarrow \wedge^d \mathcal{F}$; see EGA I, § 9.

Let's make the convention that every expression of the form $p_{i_1,...,i_d}$ is meant to be alternating in the indices. Now $\omega$ is a closed immersion, which is cut out by the quasi-coherent ideal $I$ corresponding to the graded ideal of $R[\{x_{i_1,...,i_d}\}]$, generated by the Plücker relations

$\sum_{\lambda=1}^{d} (-1)^{\lambda} x_{i_1,...,i_{d-1},k_\lambda} x_{k_0,...,\hat{k_\lambda},...,k_d}$

for all $1 \leq i_1,...,i_{d-1},k_0,...,k_d \leq n$. Thus $\omega$ induces an isomorphism $G \cong V(I) \subseteq P$. This gives another interpretation (and also proof) of the Plücker embedding, see here for a related topic.

Now considering the identity of $G$ as a $G$-valued point of $G$, we get a (universal) locally free sheaf $\mathcal{F}$ of rank $d$ on $G$ (together with an epimorphism $s : \mathcal{O}^n \twoheadrightarrow \mathcal{F}$). Consider this on $V(I)$. Then $\omega_* \mathcal{F}$ is a coherent sheaf on $P$. If we apply Serre's Theorem twice, we see that $\omega_* \mathcal{F}$ fits into an exact sequence of the form

$\mathcal{O}_P(q_1)^{r_1} \to \mathcal{O}_P(q_2)^{r_2} \to \omega_\* \mathcal{F} \to 0.$

This may be viewn as a "twisted presentation". Here the map on the left is given by a $r_2 \times r_1$-matrix of global sections of $\mathcal{O}_P(q_2 - q_1)$, which are just homogenuous polynomials of degree $q_2 - q_1$ in the variables $x_{i_1,...,i_d}$.

Question. How does such a matrix look like explicitly?

Here is what I've done so far: Using the identifications, calculate explicitely $\mathcal{F}$ by means of cocycles on the covering of basic-opens $D_+(x_{i_1,...,i_d})$. This works. Then I wanted to calculate the graded module $\Gamma_*(\mathcal{F})$ and try to see twisted generators and relations. But it's all a big mess. By the way, I need this calculation for a sort of Plücker embedding in another context.

  • $\begingroup$ Sasha's exact sequence below is correct, but there are also more elementary means to find it. $\endgroup$ Jul 20 '11 at 12:38

To get a simpler answer it is better to fix a free $R$-module $V$ of rank $n$ so that $G = Gr(d,V)$ and $P = P(\Lambda^dV)$. First we have a tautological epimorphism $V\otimes O_G \to F$ on $G$ which by adjunction gives an epimorphism $V\otimes O_P \to i_*F$. Denote the kernel by $K$. Twisting the short exact sequence $$ 0 \to K \to V\otimes O_P \to i_*F \to 0 $$ by $O(1)$ and taking cohomology one gets $$ H^0(K) = Ker(V\otimes \Lambda^d V) \to \Sigma^{2,1^{d-1}}V), $$ where $\Sigma$ is the Schur functor. It follows that $H^0(K) = \Lambda^{d+1}V$, so we have $$ \Lambda^{d+1}V\otimes O_P(-1) \to V\otimes O_P \to i_*F \to 0. $$ I guess this is what you need.

  • $\begingroup$ Thanks a lot! Twice you mean $K(1)$, right? You seem to use that $K(1)$ is generated by its global sections and that the first cohomology vanishes, why is that so? Also, I have to learn first what Schur functors are at all, could you please tell me what you have used here about their connection with cohomology? $\endgroup$ Jul 6 '11 at 8:59
  • $\begingroup$ I've read some basics about Schur functors, but still don't understand your proof. Also, could you please indicate how the morphism $\wedge^{d+1} V \otimes O_P(-1) \to V \otimes O_P$ looks like? $\endgroup$ Jul 7 '11 at 19:57
  • $\begingroup$ The map is the composition $\Lambda^{d+1}V\otimes O(-1) \subset V\otimes \Lambda^dV\otimes O(-1) \to V\otimes O$, the first map is the natural embedding and the second is the tautological map on $P$ tensored by $V$. $\endgroup$
    – Sasha
    Jul 7 '11 at 20:19
  • $\begingroup$ Thanks. Could you also explain the rest? We have an exact sequence $0 \to H^0(K(1)) \to H^0(V \otimes O_P(1)) \to H^0((i_* F)(1)) \to H^1(K(1)) \to H^1(V \otimes O_P(1))$. Here we have $H^1(V \otimes O_P(1))=0$ for $1<d<n$, and $H^0(V \otimes O_P(1)) = V \otimes \wedge^d(V)$. But what about $H^0((i_* F)(1)) \to H^1(K(1))$? $\endgroup$ Jul 7 '11 at 20:57
  • $\begingroup$ $H^0(i_*F(1)) = H^0(F(1)) = \Sigma^{2,1^{d-1}}V$ by Theorem of Borel--Bott--Weil. The map $V\otimes \Lambda^dV \to \Sigma^{2,1^{d-1}}V$ is the canonical projection, so $H^1(K(1)) = 0$, $H^0(K(1)) = \Lambda^[d+1}V$. $\endgroup$
    – Sasha
    Jul 7 '11 at 21:33

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.