3
$\begingroup$

Suppose $X \subset \mathbb{P}^n$ is an irreducible projective scheme. Then its multiplicity $\mu_X$ is defined as the length of the local ring $\mathcal{O}_{X,\eta}$ over itself, where $\eta$ is the generic point of $X$.

The degree $d_X$ of $X$ is defined as $\frac{c}{n!}$, where $c$ is the leading coefficient of the Hilbert polynomial $P_X$ of $X$, and $n$ is the degree of $P_X$.

I would like to show that $d_X = \mu_X \cdot d_{X_{red}}$, where $X_{red} \subset X \subset \mathbb{P}^n$ is the reduced subscheme with the same set of underlying points, but I don't see how to relate the Hilbert polynomial to the multiplicity. So any ideas or hints would be appreciated.

One thought was that maybe even $P_X = \mu_X\cdot P_{X_{red}}$ holds. But on further thought this is inplausible, because $X$ might have embedded components, which do not appear in $Y$, but contribute a part to $P_X$. But doing a (homogeneous) primary decomposition we can decompose $X = X_1 \cup \dots \cup X_r$ scheme theoretically, and $X_1$ corresponds to the maximal component. Because $X_2, \dots X_r$ do not contribute to the leading coefficient of $P_X$, we might assume wlog that $X$ does not have any embedded components. Does this imply $P_X = \mu_X \cdot P_{X_{red}}$?

$\endgroup$
0
$\begingroup$

We will use Proposition I 7.4 from Hartshorne's Algebraic geometry (p. 50), that is:

Let $M$ be a finitely generated graded module over a noetherian graded ring $S$. Then there exists a filtration $0 = M^0 \subset M^1 \subset \dots \subset M^r = M$ by graded submodules, such that for each $i$, $M^i / M^{i+1} \cong (S/\mathfrak{p}_i)(l_i)$, where $\mathfrak{p}_i$ is a graded homogeneous prime ideal of $S$, and $l_i \in \mathbb{Z}$. The filtration is not unique, but for any such filtration we have:

  1. if $\mathfrak{p}$ is a homogeneous prime ideal of $S$, then $\mathfrak{p} \supset \text{Ann }M \Leftrightarrow \mathfrak{p} \supset \mathfrak{p_i}$ for some $i$. In particular, the minimal elements of the set $\{\mathfrak{p}_1, \dots,\mathfrak{p}_r\}$ are just the minimal primes of $M$, i.e. the primes which are minimal containing $\text{Ann }M$.
  2. for each minimal prime of $M$, the number of times which $\mathfrak{p}$ occurs in the set $\{\mathfrak{p_1},\dots,\mathfrak{p}_r\}$ is equal to the length of $M_\mathfrak{p}$ over the local ring $S_\mathfrak{p}$ (and hence is independent of the filtration).

To compare the Hilbert polynomial of $X$ and $X_{red}$, we consider the homomorphism on homogeneous coordinate rings $$ S(X) = k[x_0,\dots,x_n]/I \rightarrow k[x_0,\dots,x_n]/\sqrt{I} = S(X_{red}).$$

Applying the proposition, we see that $P_X(l) = \sum_i \dim(S(X)/\mathfrak{p}_i)_{l_i + l}$ for $l \gg 0$. Only the minimal primes of $S(X)$ contribute to the leading coefficient of $P_X$, because the degree of the Hilbert polynomial equals the dimension of the scheme. The only minimal prime is the nilradical, so we see that the leading coefficient is a multiple of the leading coefficient of $S(X_{red})$, noting that the shift $l \mapsto l + l_i$ does not change the leading coefficient of a polynomial.

According to part 2. of the proposition, the factor of the degree equals $\text{length}_{S_\mathfrak{p}} M_\mathfrak{p}$. But following the proof of the proposition, one can check that this also holds if we make to homogeneous localisation, i.e. take $S_{(\mathfrak{p})}$ and $M_{(\mathfrak{p})}$ instead.

I think this works without mentioning primary decompositions, though the connection would be interesting.

$\endgroup$

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.