the situation is the following: I have a projective scheme $\tilde{P}\rightarrow S=Spec(A)$ with $A$ excellent and $I$-adically complete for some ideal of $A$. A group $Y$ acting on $\tilde{P}$ freely in Zariski topology and $P$ is the quotient by $Y$ and it is proper over $S$. Moreover I know that the fiber $\tilde{P}_0$ of $\tilde{P}$ over $S_0=Spec(A/I)$ is connected. I have to prove that $P$ is irreducible. I read that up to replace $\tilde{P}$ with is normalization it can be assumed that $P$ is normal (this is the first thing I do not understand). Assuming this I read that it is enough to show that $P$ is connected. But why?does normal+connected implies irreducible? I have in mind this example: if we take k-planes, $k>2$, in a $\mathbb{P}^n$, for big n, intersecting only in the origin, this is normal (regular in codimension 1 implies normal right?) and connected but not irreducible. Last problem: I read that since $P$ proper over $S$ and $P_0$ connected then $P$ is connected too.

  • 2
    $\begingroup$ Normal implies regular in codimension 1. The converse is not true: for instance, two planes in $\mathbb{P}^4$ intersecting in a single point are not a normal variety. This follows by taking normalization (two disconnected plains) and then using Zariski's Main Theorem. $\endgroup$ – Francesco Polizzi Mar 21 '11 at 17:07
  • $\begingroup$ @Francesco Polizzi ok sorry! so why normal+connected implies irreducible?I have in mind a solution even if it seems to me there is someting wrong: if $X=Y\cup Z$ with $Y,Z$ irreducible then consider $Y\coprod Z \rightarrow X$. Is this birational?(I think so) then $X$ normal + Zariski's Main thm implies connected fibers but on the intersection ($Z\cap Y\subset X$) fibers are not connected so $X$ must be irreducible.Is it correct? $\endgroup$ – unknown Mar 21 '11 at 17:56
  • 1
    $\begingroup$ @unknown (for normal+connected => irreducible), that sounds right. There is an algebraic approach which doesn't use any real machinary though. All you need to do is notice that a normal local ring must be a domain. See page 64 and Exercise 9.11 in Matsumura's commutative ring theory. $\endgroup$ – Karl Schwede Mar 21 '11 at 18:11
  • $\begingroup$ @Karl Schwede thanks!!If I have understand a prof coul be the following:the normalization is defined as the disjoint union of the normalization of each irreducible component. So $X$ normal+connected implies $X$ is one of these. If I show that locally this is a $Spec(A)$ with $A$ a domain we have integral implies irreducible. By structure thm for integrally closed domain $A=A/p_1\times\dots\times A/p_r$ for $p_i$ minimal primes, but by connectness of $Spec(A)$ there is only one of these and $A$ is integral. So $X$ is integral and that's all. Is it correct? Any idea for the other questions? $\endgroup$ – unknown Mar 21 '11 at 18:36
  • $\begingroup$ For your other questions, can you clarify something for me? You say Moreover I know that the fiber $\tilde P_0$ of $\tilde{P}$ over $S_0 = \text{Spec}(A/I)$ is connected. I have to prove that *this* is irreducible.'' When you say I have to prove that this is irreducible'', do you mean you need to show that $\tilde P_0$ is irreducible or that $\tilde P$ is irreducible? $\endgroup$ – Karl Schwede Mar 21 '11 at 18:57

Just to be clear: "normal" for a ring (assumed Noetherian to be safe) means (essentially by definition) that it is is a product of normal domains.

Here is one justification for this: suppose that a reduced ring $A$ is integrally closed in its total quotient ring (i.e. in the localization of $A$ with respect to its non-zero divisors). If $K(A)$ denotes this total quotient ring, then $K(A) = \prod K_i$ is a product of field $K_i$. Let $e_i$ denote the element of $K(A)$ which is $1$ in the $i$th place and $0$ everywhere else. Then $e_i^2 = e_i = 0$, thus $e_i$ is integral over $A$, thus $e_i \in A$ by the assumption, and so $A$ has a decomposition $A = \prod A_i$ into a product of domains corresponding to the decomposition of $K(A)$ into a product of fields.

In particular, if Spec $A$ is connected, then $A$ is itself a domain, and so Spec $A$ is in fact irreducible.

(This fact about normal rings is one of the basic motivations for their application in geometry: normalization is a process for resolving singularities that are caused by differenet irreducible components crossing, because it resolves different irreducible components into different connected components.)


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.