What happens if I blow up $\operatorname{Spec} k[x]/(x^2)$ at its unique closed point?
- User
What happens if I blow up at its unique closed point?
Assistant
Let . Let .
-
Identify the scheme and the point: The ring has only one prime ideal, . This corresponds to the unique point in the topological space . Since is the only prime ideal, it must also be the nilradical of . The quotient ring , which is a field. Therefore, is a maximal ideal, and the point is a closed point.
-
The Blow-Up Construction: The blow-up of a scheme at a point is defined using the ideal sheaf corresponding to . If , the blow-up along the ideal is given by , where is the Rees algebra of , viewed as a graded subalgebra of . In our case, and the ideal is . We need to compute the Rees algebra . The powers of are:
- in .
- for all . The Rees algebra is . Let . The blow-up is .
-
Analyzing Proj A: To compute , we need to consider as a graded ring. We can assign degrees to and . Let's try assigning and . . The graded components are:
- (degree 0)
- (degree 1)
- ? Since and in , .
- .
- for all . So . The Proj construction typically applies to standard graded rings, usually polynomial rings or quotients thereof. Let's try another grading: . .
- ? Since . .
- ? .
- for . So . The Proj of is (where ), which corresponds to a single point (it's essentially for , the coordinate ring is ).
-
Alternative perspective using the Normal Cone: For a local ring , the blow-up is often described as , where is the associated graded ring. In our case, and . We have . So . Let . . The symmetric algebra , where corresponds to the class of in . . As we saw, this represents a single point. This approach is consistent with the result obtained from the Rees algebra under a suitable grading.
-
Interpretation: The scheme is essentially a single point , but with non-trivial infinitesimal information captured by the nilpotent element such that . Blowing up the point means replacing with the projectivized tangent space (or more accurately, the projectivized normal cone). The normal cone is , which is , representing a single point. So, the blow-up operation replaces the point with another point. The resulting scheme is still just a single point, isomorphic to the original scheme .
Conclusion: If you blow up at its unique closed point, the resulting scheme is still , which is just a single point. The blow-up map is the identity map on this point. The infinitesimal structure () remains.
-
- User
Are you sure? I'm thinking I actually get the empty scheme. Which of us is incorrect?
Assistant
You raise a very good point, and it's important to be sure about this. Let's re-examine the process carefully.
The scheme is , where . The unique point corresponds to the maximal ideal . We know . The powers of the maximal ideal are and .
The Blow-Up Construction
The standard way to construct the blow-up of at the point (corresponding to the ideal ) is , where is a formal variable. This graded ring is called the Rees algebra, denoted .
Let's compute the Rees algebra for and : This can be represented as the ring . Let .
To compute , we need to give a standard grading. The natural grading comes from assigning degree 1 to both and : , . Then has degree 2, and has degree 2. The ideal is homogeneous in this grading. The graded components of are: (constants) (linear terms) (since and in ) for (since and )
The Proj construction is typically defined as the set of homogeneous prime ideals of other than the irrelevant ideal . The homogeneous prime ideals of are and . The irrelevant ideal is . So, . The prime ideal corresponds to the ring , which is isomorphic to . The prime ideal corresponds to the ring .
Therefore, . The resulting scheme is , which is isomorphic to the original scheme . It is still just a single point.
Alternative Approach: Normal Cone
Another common definition or intuition for the blow-up at a point of involves the normal cone, which is , where is the associated graded ring.
Let's compute : for .
Let be the generator of the second factor (corresponding to the class of in ). The multiplication rule in the associated graded ring is . So, corresponds to the class of in . Since , this class is . Thus, , where .
Now we compute . Let . , , for . The Proj construction considers for non-zero homogeneous elements . The only non-zero homogeneous elements are the non-zero constants in . Let's choose . is the localization of by powers of , which is just itself. So, . This scheme is again isomorphic to .
Conclusion and Addressing Your Doubt
Both standard methods show that the blow-up of at its unique closed point results in a scheme isomorphic to the original scheme, . This scheme consists of a single point.
Why you might think it's the empty scheme:
The empty scheme corresponds to the zero ring. Our original scheme corresponds to the non-zero ring . Blowing up a point replaces that point with the projectivized normal cone. The normal cone is . In our case, . This is a 1-dimensional vector space over . The projectivization is , which is a single point (representing the lines through the origin in a 1D vector space). Since the normal cone is not empty (it's ), the blow-up replaces the original point with another point. The resulting scheme is not empty.
Perhaps there's confusion with blowing up the entire variety along an ideal . If (the whole ring), the Rees algebra . Then . This is still not empty. Maybe the confusion comes from some edge case or a different operation?
Based on the standard definitions of blow-ups in algebraic geometry, the result is not the empty scheme. It is isomorphic to the original scheme .
- User
More generally, I think blowing up any nilpotent ideal sheaf gives an empty scheme.
Assistant
That's an interesting claim, and it touches on a subtle point about blow-ups and the empty scheme. Let's carefully analyze it.
The blow-up of a scheme along an ideal sheaf is given by , where is the Rees algebra of . The scheme is empty if and only if the ideal contains the unit element . This happens if and only if the degree 0 component is zero.
In our case, , so . Therefore, is empty if and only if , which means itself is the empty scheme.
This implies that if you start with a non-empty scheme , blowing up any ideal sheaf (nilpotent or not) will result in a non-empty scheme.
Let's revisit your specific example: . The unique closed point corresponds to the maximal ideal . This ideal is nilpotent, since .
We computed the Rees algebra for as . The degree 0 part of this algebra is . Since , is not empty. As we found, it is isomorphic to , the original scheme.
Where your statement might seem correct (The Special Case):
Consider the simplest non-empty scheme: , where is a field. This scheme consists of a single point. The only ideal in is the zero ideal . Is a nilpotent ideal? Yes, because . Let's blow up at the ideal . The Rees algebra is . So, . Now we compute . The graded ring has only one homogeneous prime ideal, which is . The irrelevant ideal is . The definition of is the set of homogeneous prime ideals such that . In this case, the only homogeneous prime ideal is , and . Since , this ideal is excluded. Thus, is the empty set.
So, you are correct that blowing up at the zero ideal results in the empty scheme. The zero ideal is indeed a nilpotent ideal.
Why your general statement is incorrect:
The critical difference lies in the degree 0 component of the Rees algebra.
- When blowing up at , the Rees algebra is . Here, , but all for are . So . The only prime ideal is contained in , making empty.
- When blowing up at , the Rees algebra is . Here, , , , etc. The irrelevant ideal is not the whole ring , because is not contained in . Hence, is not empty.
The statement "blowing up any nilpotent ideal sheaf gives an empty scheme" is false. The condition for the blow-up to be empty is precisely that the base scheme itself is empty (), regardless of the ideal sheaf . The only case where a nilpotent ideal leads to an empty blow-up is when the starting scheme is and the ideal is .
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.