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

Deformation to the normal cone appears in several places including Intersection theory and Verdier specialisation of construtible sheaves or D-modules. I'd like to understand what happens when we iterate the process: deform to the cone of $A_1 \subset X$ and then to the cone of $C_{A_1\cap A_2} A_2 \subset C_{A_1} X$. In particular is it commutative (can we exchange $A_1$ and $A_2$)?

Deformation to the normal cone: Let $A$ be a closed subscheme of $X$ with ideal $I \subset O_X$. One has the space of deformation to the normal cone $D_A X = Spec_X( R_I O_X)$ where $(R_I O_X)\subset O_X[t^{\pm 1}]$ is the Rees algebra of $I$ defined by $$ R_I O_X = \bigoplus_{k\in \mathbb{Z}} I^{-k} t^k $$ with $I^{k} = O_X$ if $k \leq 0$. The projection $t: D_A X \to \mathbb{A}^1$ is flat. For $t \neq 0$ the fiber is $X$ while for $t =0$ the fiber is the normal cone $$ C_A X = Spec \left( \bigoplus I^{k}/I^{k+1} \right) $$ hence the name. In a more geometric fashion, $D_A X$ is the complementary of $\mathbb{P}(C_AX) \subset \mathbb{P}(C_{A}X \oplus 1)$ inside the blow-up $B_{A\times 0} X\times \mathbb{A}^1$. The projection onto $X\times \mathbb{A}^1$ being given by the inclusion $O_X[t] \to R_I O_X$.

The construction is fonctorial. For $f:X\to Y$, $B\subset Y$ and $A = X\times_Y B$, we have $$ f^* \left( \bigoplus I_B^{-k} t^k \right) \to \bigoplus I_A^{-k} t^k $$ inducing a morphism $D(f):D_A X \to X\times_Y D_B Y$ compatible with the projections onto $\mathbb{A}^1$. If $f$ is a closed embedding then so is $D(f)$.

Multiple deformations: Now consider $A_1,A_2 \subset X$ with ideals $I_1,I_2$.

We can form the multi-Rees algebra $$ R_{(I_1,I_2)} O_X := \bigoplus_{k_1,k_2\in \mathbb{Z}} (I_1^{-k_1} \cap I_2^{-k_2}) t_1^{k_1} t_2^{k_2} \subset O_X[t_1^{\pm 1},t_2^{\pm 1}] $$ Its Spec aver $X$ is a space $D_{(A_1,A_2)} X$ together with a morphism $D_{(A_1,A_2)} X \to X\times \mathbb{A}^2$ given by the coordinates $(t_1,t_2)$.

We also have a closed immersion $D_{A_1 \cap A_2} A_2 \subset D_{A_1} X$ with ideal $$ J_2 = \bigoplus_{k\in \mathbb{Z}} (I_2 \cap I_1^{-k}) t^{k} $$ So we can form the Rees algebra $R_{J_2} R_{I_1} O_X$ and its Spec over $D_{A_1} X$, we simply denote by $D_{A_2} D_{A_1} X$. It also has a canonical morphism $$ D_{A_2} D_{A_1} X \to D_{A_1}X \times \mathbb{A}^1 \to X\times \mathbb{A}^1 \times \mathbb{A}^1. $$

Question 1: Is it the same thing to

  1. Deform simultaneously using $D_{(A_1,A_2)} X$

  2. Deform to the normal cone of $A_1$ in $X$ and then to the normal cone of $C_{A_1\cap A_2} A_2$ in $C_{A_1} X$

Question 2: Do we have a canonical isomorphism $$ R_{(I_1,I_2)} O_X = R_{J_2} R_{I_1} O_X? $$

Does is it induce an isomorphism of $(X\times \mathbb{A}^2)$-schemes compatible with the $\mathbb{G}_m^2$-actions coming from the gradings?

Question 3: Do we have a canonical isomorphim $$ D_{(A_1,A_2)} X|_{t_1 = 0} = D_{C_{A_1\cap A_2} A_2} C_{A_1} X $$ i.e. $$ R_{(I_1,I_2)} O_X / (t_1) = R_{Gr_{I_1} I_2} Gr_{I_1} O_X $$

Question 4 What can we say if $A_1$ and $A_2$ are transverse subvarieties? What changes for the deformation spaces?

Note: I know that in this case the canonical morphism $C(i_2) : D_{A_1\cap A_2} A_2 \to A_2 \times_{X} D_{A_1} X$ is an isomorphism.

Question 5 Does any one know any good reference where basic functorial properties of Rees algebras are detailed?

share|improve this question
    
Have you worked out the example where $X$ is $\mathbb{C}^3$, with $A_1=\{ x=y=0 \}$ and $A_2 = \{ y=z=0 \}$? That's the standard example of two blowups not commuting. I've been meaning to do it for you, but it doesn't look like I'll get to it. –  David Speyer Mar 1 '11 at 0:45
    
I don't know the universal property of the deformation space so I tried to deal with the Rees Algebras. From what I've worked out, it seems that the bigradings aren't the same. Perhaps I'm looking at the wrong projection $D_{A_2} D_{A_1} X \to X \times \mathbb{A}^2$. –  YBL Mar 1 '11 at 1:32
1  
In my experience, from playing with the Rees algebras, no two degenerations like these ever nonobviously match. :( –  Allen Knutson Mar 1 '11 at 5:33
    
Added a few questions –  YBL Mar 1 '11 at 18:01
add comment

1 Answer

Ok so the solution is to consider the product of ideals instead of their intersection.

Just like we have $$ B_{\widetilde{I}_1} B_{I_2} X = B_{I_1I_2} X = B_{\widetilde{I}_2} B_{I_1} X $$ with $\widetilde{I}_j$ the total transform of the ideal $I_j \subset O_X$, we have $$ D_{\widetilde{D}_1} D_{A_2} X = D_{(A_1,A_2)} X = D_{\widetilde{D}_2} D_{A_1} X $$ with $\widetilde{D}_j = A_i \times_X D_{A_j} X$ the total pullback of the normal cone deformation space to $A_i$.

With this definition, we have nice functoriality properties w/r to maps and direct products.

share|improve this answer
add comment

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.