Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Suppose X is dimension two locally Noetherian scheme. Y is a closed subscheme of X, with codimension 1. Denote X' to be the blow up of X along Y. Prove that the structure morphism f:X'-->X is a finite morphism.

It suffices to show it's quasi-finite according to Zariski's main theorem. But I can't exclude the possibility that an irreducible component of $f^{-1}(Y)$ maps to a closed point of Y.

share|improve this question

2 Answers 2

up vote 13 down vote accepted

I think it's not true :

Let $X=Spec(A)$ with $A=k[x,y,z]/(x^2-y^2-z^2)$ be a quadratic cone. Let $Y$ be a line through the origin of the cone : its ideal is $I=(z,x-y)$. We calculate :

$$X'=Proj_{A}A[t,u]/(zt-(x+y)u,(x-y)t-zu),$$ [EDIT : THE FORMULA HAS BEEN CORRECTED]

where, in the graded $A$-algebra $A+I+I^2+....$ we denoted $t$ and $u$ the degree one generators corresponding to $z$ and $x-y$. Now, quotienting by $x$, $y$, and $z$, we calculate the fiber over the origin of this blow-up It is Proj(k[t,u]), which is a positive-dimensional projective line !

share|improve this answer
    
Here I take $Proj (\sum_{n>=0}I^n)$ as the definition of blowing up along V(I) on Spec A. So it is not always equal to $Proj_{A}A[t,u]/(zt-(x+y)u)$. More precisely, it is equal to $Proj_{A}A[t,u]/(zt-(x+y)u)$ when $Proj_{A}A[t,u]/(zt-(x+y)u)$ is integral. But here this expression is not an integral scheme. So this counterexample does not work under our definition of blowing up! –  Taisong Jing Jan 6 '10 at 21:53
    
We do have the same definition ! And $Proj_A A[t,u](zt−(x+y)u)$ is just the concrete expression of $Proj(\sum_{n\geq 0} I^n)$ in my example ! –  Olivier Benoist Jan 6 '10 at 21:59
    
And after calculation, if you take $Proj (\sum_{n>=0}I^n)$ as the definition of blowing up, then you will see it's indeed quasi-finite in this example. –  Taisong Jing Jan 6 '10 at 22:00
    
No, unfortunately. After localizing on the open set D(u), one would get (k[x,y,z,t]/(z^2-x^2-y^2, zt-(x-y)))_z = (k[y,z,t]/(z^2-y^2-(zt+y)^2))_z = k[z,z^{-1},t], and the fiber of the origin point is indeed finite. As you have seen, the point is that you need to take localization at D(z) if you take the definition as $Proj (\sum_{n>=0}I^n)$, but you don't localize if you take definition as $Proj_A A[t,u](zt−(x+y)u)$. –  Taisong Jing Jan 6 '10 at 22:07
    
No, it's not. It is the union of two lines, the strict transform of $Y$, and another contracted to the origin by the blow-up as you can see from : $$Proj(\sum_{n\geq 0}I^n)=Proj (A/I)[t,u]/((x+y)u).$$ –  Olivier Benoist Jan 6 '10 at 22:17

How do you define blow-up? It should be straightforward to show that an explicit construction has relative dimension 0 over $X$.

Update: In the comments I suggest to take the formal two-dimensional ring $k[[x_1, x_2]]$ and work with it. This assumes that $X$ is smooth. If it's not, then Olivier gave an example of quadratic cone where blowup has relative dimension 1.

share|improve this answer
    
It can be resated algebraically as follows: A is Noetherian ring, I is a height one ideal, prove that $Proj (\sum_{n>=0} I^n/I^{n+1})$ --> $Spec A/I$ is a finite morphism. –  Taisong Jing Jan 6 '10 at 20:46
    
I'm not good in commutative algebra, but you can reduce to $A$ being a formal ring $k[[x_1, x_2]]$ and $I = (f)$, where $f$ is homogenious and then everything should follow. –  Ilya Nikokoshev Jan 6 '10 at 21:08
1  
No, you can't unless Y is a Cartier divisor. in this case X'=X... which is not very interesting ! –  Olivier Benoist Jan 6 '10 at 21:17
    
Yes, just sa Oliver has said, the point is that the Weil divisor is not necessarily a Cartier divisor; and because we are not only considering algebraically varieties, so we cannot even have a good expectation of the normal locus on X. –  Taisong Jing Jan 6 '10 at 21:22

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.