Let's say I have a complex projective variety $X\subseteq\mathbb P^n$ with homogeneous coordinate ring $S=\bigoplus_{d\ge 0} S_d$. The localization by some homogeneous $f\in S$ (of nonzero degree) yields a graded ring $S_f=:R=\bigoplus_{d\in\mathbb Z} R_d$. Now, instead of looking at the spectrum of the degree zero part, I could look at $R_{\ge 0}=\bigoplus_{d\ge 0} R_d$. I have an inclusion of graded $\mathbb C$-algebras $S\hookrightarrow R_{\ge 0}$ and both are positively graded, so I may very well consider the morphism $$Y:=\mathrm{Proj}(R_{\ge 0})\xrightarrow{\quad\textstyle\pi\quad}\mathrm{Proj}(S)=X.$$ Since $\mathrm{Spec}(R_0)=X_f$, the variety $Y$ is projective over $X_f$. However, what is the nature of $\pi$? Is it surjective? I am sure this must have interested someone before, but I cannot find anything in the 'standard' literature on algebraic geometry.

More generally, is there any deeper geometric meaning to taking the positive part of a graded ring?

  • $\begingroup$ You should specify that $f$ is homogeneous of nonzero degree. $\endgroup$ – Jason Starr Oct 4 '13 at 11:32
  • $\begingroup$ To understand the geometric meaning to taking the positive part of a graded ring for defining Proj, you need go back to geometric quotients by algebraic groups. Proj is a quotient of an affine space (without "the origin") by the multiplicative group G_m. Unfortunately, this is not explained in the usual common books. $\endgroup$ – Al-Amrani Oct 14 '13 at 11:51

Up to replacing $S$ by the "Veronese subring" $S_{(e)} := \oplus_d S_{de}$, which does not change Proj, you may as well assume that $f$ has degree $1$. Consider the graded $R_0$-algebra homomorphism $u:R_0[t]\to R_{\geq 0}$ by $u(t) = f$.

First, $u$ is surjective. Indeed, for every homogeneous element $g$ in $R_d$, then $h=gf^{-d}$ is in $R_0$. Thus $ht^d$ is an element in $R_0[t]_d$ such that $u(ht^d)$ equals $g$.

Next, $u$ is injective. To prove this, observe that, since $u$ is a graded $R_0$-algebra homomorphism, also $\text{Ker}(u)$ is a homogeneous ideal. Every element in $R_0[t]_d$ is of the form $ht^d$ for some $h\in R_0$. Also, $u(ht^d)$ equals $hf^d$. Since $f$ is not a zero divisor in $R$ (by construction), if $hf^d$ equals $0$, then $h$ equals $0$, hence $ht^d$ equals $0$. Therefore the only homogeneous element in $\text{Ker}(u)$ is $0$, i.e., $u$ is injective.

Since $u$ is an isomorphism of graded $R_0$-algebras, also $\text{Proj}(R_{\geq 0})$ equals $\text{Proj}(R_0[t])$. Of course $\text{Proj}(R_0[t])$ is just $\text{Spec}(R_0)$. Thus $\text{Proj}(R_{\geq 0})$ is canonically isomorphic to $\text{Spec}(R_0)$.

  • $\begingroup$ Ouch, I was 1 min late. Anyway, perfect answer! $\endgroup$ – Anton Fonarev Oct 4 '13 at 12:06
  • $\begingroup$ Conclusion:the morphism π in the question is nothing else but the canonical open immersion defined by "f different from 0". Recall that Proj is, by construction, covered by such affine open subsets. $\endgroup$ – Al-Amrani Oct 14 '13 at 11:36

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.