4
$\begingroup$

I am looking at exercise 12.21 from Gathmann's notes on algebraic geometry. I am given a homogeneous ideal $$I \unlhd k[x, y, z] $$ with a dimension $0$ projective locus. WLOG, we assume that this has non-vanishing $z$ coordinate, and hence we can define the ideal $$ J = \left\lbrace f(x, y, 1) : f \in I \right\rbrace \unlhd k[x,y].$$ The task is to show that $$ \deg I = \chi_{I} = \dim_{k}k[x, y]/J $$ where $\chi_{I}$ is the Hilbert polynomial of $I$. In other words, I need to show that I can determine the Hilbert polynomial of a dimension $0$ projective set by taking an affine chart. I feel like this should not be a difficult question. Any help would be appreciated.

Edit. $\deg I$ in this case is the degree $0$ Hilbert polynomial. In other words, it is the unique natural number $n$ such that $n=\dim_kk[x,y,z]_d/I_d$ for "almost all" values of $n$.

Thanks.

$\endgroup$
4
$\begingroup$

Let $S=k[x,y,z]$ and let $I$ be a homogeneous ideal such that the projective variety defined by $I$ has dimension $0$. Then $S/I$ has Krull dimension $1$ and so its Hilbert polynomial is a constant, i.e., the Hilbert function is eventually a constant. Suppose that $H_{S/I}(n) = H_{S/I}(n_0), \, \forall n \ge n_0$. Now let $R = k[x,y]$ and let $J$ be the dehomogenization of $I$ with respect to $z$, i.e., take all polynomials in $I$ and set $z=1$. Let $\mathcal{B}$ be a $k$-basis for the vector space $J_{\le n_0}$, i.e., the vector space of all polynomials in $J$ of degree $\le n_0$. Then $\mathcal{B}^h = \left\{z^{n_0} p(x/z,y/z): \, p \in \mathcal{B} \right\}$ is a $k$-basis for the vector space $I_{n_0}$, i.e., $\dim_k J_{\le n_0} = \dim_k I_{n_0}$. Since $\dim_k S_{n_0} = \dim_k R_{\le n_0}$, we have $\dim_k (S/I)_{n_0} = \dim_k R_{\le n_0} / J_{\le n_0}$. In fact, we have that $\dim_k R_{\le n_0} / J_{\le n_0} = \dim_k R_{\le n} / J_{\le n}, \, \forall n \ge n_0$.

We next show that $\dim_k R/J = \dim_k R_{\le n} / J_{\le n}$ for all sufficiently large $n$. First, notice that for any $n$ we have a morphism of $k$-vector spaces \begin{align} R_{\le n} \rightarrow \frac{R}{J}, \, \, \, (\dagger) \end{align} which takes an element $p \in R_{\le n}$ to its class in $R/J$. The kernel of this morphism is clearly $J_{\le n}$. Consequently, we have a monomorphism \begin{align} \frac{R_{\le n}}{J_{\le n}} \hookrightarrow \frac{R}{J}, (\ddagger) \end{align} and so $\dim_k R_{\le n} / J_{\le n} \le \dim_k R/J$. On the other hand, recall that the affine variety defined by $J$ has the same dimension as the projective variety defined by $I$ (the latter is the projective closure of the former), the ring $R/J$ must have Krull dimension zero (why?), and so it must be a finite dimensional $k$-vector space. Let $p_1,\dots,p_s$ be elements of $R$ such that their classes in $R/J$ form a $k$-basis for $R/J$. Let $d$ be the maximal degree among $\deg(p_1),\dots,\deg(p_s)$. Then, for $n \ge d$ the morphism $(\dagger)$ becomes surjective and so the embedding $(\ddagger)$ actually becomes an isomoprhism. Thus \begin{align} H_{S/I}(n_0) = \dim_k R_{\le \max(n_0,d)} / J_{\le \max(n_0,d)}=\dim_k R/J. \end{align}

$\endgroup$

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.