3
$\begingroup$

I've been recently wondering how to prove the existence of a Hilbert polynomial for finitely generated bigraded modules $M$ over a polynomial ring $R=k[X_0,...,X_n,Y_0,...,Y_m]$ with the usual bigraded structure; concretely, is there a polynomial $P\in \mathbb{Q}[T,S]$ such that $P(a,b)=\text{dim}_k M_{a,b}$ for all sufficiently large $a,b\in \mathbb{Z}$?

I know that the usual proof of this fact for the graded case uses the existence of a finite free resolution of the module. So I've been trying to apply the same technique for the bigraded case: I take a resolution of $M$ of the form $$ 0\rightarrow K \rightarrow L_{n+m+1}\rightarrow ... \rightarrow L_{0}\rightarrow M\rightarrow 0$$ where $L_i$ are finitely generated free modules (now, free in the sense of having a basis of bihomogeneous elements), $K$ is finitely generated and all the maps are compatible with the bigrading and of bidegree (0,0). We can view this resolution as a graded one (take, for instance, $M_r=\sum_{a+b=r} M_{a,b}$ as the $r$-degree piece of $M$), and then Hilbert's Syzygy Theorem asserts that $K$ is also free $\textbf{as graded module}$. I've trying to prove that in fact we can take a basis of $K$ consisting of bihomogeneous elements, but I didn't succeed (in fact I don't even know if what I'm trying to prove is true). So here's the question:

Can we assure that a finite free resolution of this sort exists for bigraded modules (that is, in which every term is freely generated by bihomogeneous elements)? Is there an analogous Hilbert Syzygy Theorem for bigraded modules? And finally (and importantly!), can this be extended to the multigraded case?

Thank you all a lot in advance.

$\endgroup$
6
  • $\begingroup$ The existence of such a Hilbert polynomial in two variables follows from Hirzebruch-Riemann-Roch. $\endgroup$ Commented Dec 16, 2022 at 0:55
  • 1
    $\begingroup$ Actually, you need much less than Hirzebruch-Riemann-Roch. Snapper's method works just fine: for every finite set of invertible sheaves, $(\mathcal{L}_1,\dots,\mathcal{L}_n)$, on a projective scheme $X$, there exists a very ample invertible sheaf $\mathcal{A}$ such that each invertible sheaf $\mathcal{A}\otimes \mathcal{L}_i$ is also very ample. In the K-group of $\mathbb{P}^r$, each $[\mathcal{O}(d)]$ is a linear combination $\sum_{e=0}^r f_e(d)[\mathcal{O}(-e)]$ where $f_e(d)$ is a polynomial of degree $r$ . . . $\endgroup$ Commented Dec 16, 2022 at 3:12
  • $\begingroup$ . . . Thus, in the K-group of $X$, there exist finitely many elements $\mathcal{B}_{e_1,\dots,e_r}=\bigotimes_{i=1}^n(\mathcal{A}\otimes \mathcal{L}_i)^{\otimes(-e_i)}$ with $0\leq e_i\leq r_i$ such that every element $\bigotimes_{i=1}^n \mathcal{L}_i^{\otimes d_i}$ is a linear combination of the elements $\mathcal{B}_{e_1,\dots,e_r}$ with coefficients that are polynomials in $(d_1,\dots,d_r)$. This gives the existence of a "multigraded Hilbert polynomial" on $X$. $\endgroup$ Commented Dec 16, 2022 at 3:15
  • $\begingroup$ Olsson and I considered writing this up for the Hilbert polynomials section of our article on Quot schemes for stacks, but we decided against it since we could not see a low-tech way to extend this to the standard generators of the K-group of the stack $B\textbf{GL}_m$. $\endgroup$ Commented Dec 16, 2022 at 3:17
  • $\begingroup$ @JasonStarr Thank you very much for the explanation and the references. I'll take a careful look at all this concepts because there are some of them I'm totally unfamiliar with. The question I posted arose when I was studying the concept of regularity of sheaves and the possibility of using it to define the Hilbert scheme for products of projective spaces from a 'bigraded point of view', in a similar way we do for the Hilbert scheme in the usual projective space. Anyway, I'd like to introduce all this stuff in a more elementary way, so, do you know if there is a purely algebraic proof... $\endgroup$
    – Carnby
    Commented Dec 16, 2022 at 9:59

1 Answer 1

0
$\begingroup$

There are several questions above. The answer below gives one multigraded finite free resolution of multigraded finitely generated modules over the multigraded polynomial ring that are saturated, i.e., equal to the multigraded module of a coherent sheaf on the associated product of projective schemes. Since the Hilbert polynomial of a finitely generated multigraded module equals the Hilbert polynomial of its saturation, this suffices for the application to multigraded Hilbert polynomials.

One useful result is Beilinson's resolution of the diagonal. Since this is completely functorial, it helps with these types of questions. Let $V$ be a $k$-vector space of finite dimension $n$. Let $r$ be an integer with $0<r<n$. Denote by $$(G\to \text{Spec}\ k,\ q^\dagger:V^\vee\otimes_k \mathcal{O}_G \twoheadrightarrow S^\vee)$$ a universal pair of a $k$-scheme together with a surjective homomorphism of locally free sheaves whose target has rank $r$. The adjoint homomorphism fits into a short exact sequence of locally free sheaves, $$0 \to S \xrightarrow{q} V\otimes_k \mathcal{O}_G \xrightarrow{p} Q \to 0.$$ On the self-product in the category of $k$-schemes, $G\times G$, there is an associated homomorphism of locally free sheaves, $$\text{pr}_1^*p\circ \text{pr}_2^*q:\text{pr}_2^*S \to V\otimes_k \mathcal{O}_{G\times G} \to \text{pr}_1^* Q.$$ By adjointness of Hom and tensor product, this is equivalent to a homomorphism of locally free sheaves, $$\alpha:\mathcal{E} \to \mathcal{O}_{G\times G}, \ \ \mathcal{E}:= \text{pr}_1^*Q^\vee \otimes_{\mathcal{O}_{G\times G}} \text{pr}_2^*S.$$ Beilinson's observation is that the image of $\alpha$ is precisely the ideal sheaf of the diagonal in $G\times G$, which is itself everywhere locally cut out by a regular sequence of length $m=\text{dim}\ G = \text{rank}\ \mathcal{E}=r(n-r)$. Thus, the Koszul complex of $\alpha$ gives a finite, locally free resolution of the structure sheaf of the diagonal, $$\left( K_\ell(\alpha) := \bigwedge^\ell_{\mathcal{O}_{G\times G}}\mathcal{E} \right)_{0\leq \ell \leq m}, \ \ (d_\ell:K_\ell(\alpha) \to K_{\ell-1}(\alpha))_{1\leq \ell \leq m}. $$ Note that the exterior power $\bigwedge^\ell \ \mathcal{E}$ has a direct sum decomposition as a tensor power of Schur functors $\text{pr}_1^*\mathbb{S}_{\lambda}(Q^\vee)\otimes_{\mathcal{O}_{G\times G}}\text{pr}_2^*\mathbb{S}_{\lambda'}(S),$ where $\lambda$ runs over all partitions of $\ell$ into integers no greater than $r$ and with no more than $n-r$ parts.

The same holds for any product of projective spaces, i.e., for an integer $\rho \geq 1$ and an ordered $\rho$-tuple of ordered pairs $(V_i,r_i)$ as above, for the product scheme $X=G_1\times \dots \times G_\rho$ with its coordinate projections $\pi_i:X\to G_i$, on the product $k$-scheme $X\times X$, one locally free resolution of the structure sheaf of the diagonal is the Koszul complex of the homomorphism of locally free sheaves, $$\beta: \mathcal{F} \to \mathcal{O}_{X\times X}, \ \ \mathcal{F}:= \bigoplus_{i=1}^\rho (\pi_i\circ \text{pr}_1,\pi_i\circ \text{pr}_2)^*\mathcal{E}_i ,$$ whose components are the maps $(\pi_i\circ \text{pr}_1,\pi_i\circ \text{pr}_2)^*\alpha_i$. Note that each locally free sheaf $\bigwedge^\ell \mathcal{F}$ has a canonical direct sum decomposition into tensor products of pullbacks of locally free sheaves $\bigwedge^{\ell_i} \mathcal{E}_i$, indexed by all ordered $\rho$-tuples of nonnegative integers $(\ell_1,\dots,\ell_\rho)$, such that $\ell_1+\dots+\ell_\rho$ equals $\ell$ and each $\ell_i\leq m_i$.
One application of this is the following.

Theorem [Beilinson resolution]. Let $\mathcal{G}$ be a coherent sheaf on $X$ such that for every ordered $\rho$-tuple $(\ell_1,\dots,\ell_\rho)$ as above, all higher cohomology vanishes for the tensor product of $\mathcal{G}$ and the pullbacks $\pi_i^*\mathbb{S}_{\lambda_i}(Q^\vee_i)$, where each $\lambda_i$ is a partition of $\ell_i$ into integers no greater than $r_i$ with at most $n_i-r_i$ parts. Then the complex $\text{pr}_2^*(K_\bullet(\beta)\otimes_{\mathcal{O}_{X\times X}}\text{pr}_1^*\mathcal{G})$ is a locally free resolution of $\mathcal{G}$ whose terms are direct sums of tensor products of locally free sheaves $\mathbb{S}_{\lambda'_i}(S_i)$. This resolution is functorial in $\mathcal{G}$.

A particularly important case is when each $G_i$ is just a projective space, i.e., every $r_i$ equals $1$. In this case, the terms of the locally free resolution above are direct sums of tensor products of invertible sheaves $\pi_i^*\mathcal{O}_{\mathbb{P} V_i}(-d_i)$ for $0\leq d_i < n_i$. Since the tensor product of $\mathcal{G}$ with the invertible sheaves $\pi_i^*\mathcal{O}_{\mathbb{P} V_i}(e_i)$ satisfies the hypothesis above for all integers $e_i \gg 0$, this gives the following corollary.

Corollary. On a product of projective spaces, $X=\mathbb{P} V_1 \times \dots \mathbb{P} V_\rho$, for every coherent sheaf $\mathcal{G}$, for all ordered $\rho$-tuple of integers $(e_1,\dots,e_\rho)$ with $e_i\gg 0$, there is a functorial locally free resolution of the tensor product of $\mathcal{G}$ and the invertible sheaves $\pi_i^*\mathcal{O}_{\mathbb{P} V_i}(e_i)$ as above. After tensoring this locally free resolution with the tensor product of the invertible sheaves $\pi_i^*\mathcal{O}_{\mathbb{P} V_i}(-e_i)$, this gives a functorial locally free resolution of $\mathcal{G}$ whose terms are direct sums of tensor products of invertible sheaves $\pi_i^*\mathcal{O}_{\mathbb{P} V_i}(-d_i-e_i)$ with $0\leq d_i < n_i$.

Translating this back into the language of multigraded modules over the multigraded homogeneous coordinate ring of $X$, this gives a functorial finite free resolution by direct sums of shifts of the multigraded homogeneous coordinate ring.

As mentioned in my comments above, one needs much less if one only needs to prove the existence of a multigraded analogue of the Hilbert polynomial. However, the functoriality in the Beilinson resolution is frequently useful.

$\endgroup$
2
  • $\begingroup$ Thank you SO MUCH for this amazing answer. Now I see why my naive approach did not succeed, more machinery is needed here. I’ll take a careful look at all this concepts. I really appreciate your explanation. $\endgroup$
    – Carnby
    Commented Dec 22, 2022 at 7:28
  • $\begingroup$ Your approach probably works. The above gives something stronger than Hilbert's Syzygy Theorem, but also weaker. It only applies to "saturated" sheaves (which is good enough for the properties of the Hilbert polynomial). Also, note that when it works, there is a discrepancy between the length of the projective resolution produced by the Beilinson resolution and the length of the projective resolution in Hilbert's Syzygy Theorem. Via the Auslander-Buchsbaum Formula, this is because the depth of a saturated module is already at least as positive as the dimension of a general torus orbit. $\endgroup$ Commented Dec 22, 2022 at 12:12

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .