2 and 3

Here is a question I asked myself yesterday: Suppose that X is an algebraic space which has degree 2 finite etale covering X_2 —> X and a degree 3 finite etale covering X_3 —> X such that both X_2 and X_3 are schemes. Is X a scheme?

I thought there might be a chance that the answer is yes, but just now in the common room, Philip, Anand, Davesh, and me proved that the answer is: no!

Namely, following Hironaka’s example, we made a smooth proper 3-fold X_6 with an action of a cyclic group G of order 6 such that X = X_6/G is not a scheme, but the two intermediate quotients X_2 and X_3 are schemes. Namely, start with a smooth projective 3-fold Y with an action of G and a 6-gon of smooth rational curves C_0, C_1, C_2, C_3, C_4, C_5 which are cyclically permuted by G. In other words, I mean that C_i ∩ C_{i + 1} is exactly one point P_i and the intersection is transversal. Then you do as in the Hironaka example: blow up all these curves but at P_i analytic locally blow up C_i first and C_{i + 1} second. The result is a proper scheme X_6 with an action of G. Over the P_i there are two curves E_i, E’_i such that E_0 + E_1 + … + E_5 is zero in the Chow group (see picture of Philip below). Hence orbit of a point on E_0 cannot be contained in an affine open of X_6, which proves that X_6/G is not a scheme. However, the morphism

X_6 – E_i ∪ E’_i —-> Y – P_i

is a usual blow up, hence the source is a quasi-projective variety and any finite collection of points is contained in an affine open. This quickly gives that X_3 and X_2 are schemes.
To finish, here is a picture X_6 (thanks Philip!):IMG_20150422_153508_334

Lecture notes on etale cohomology

In 2009 I gave a course on etale cohomology here at Columbia University. Notes were taken by Thibaut Pugin, Zachary Maddock and Min Lee (in reverse alphabetical ordering). There were the basis for the chapter on etale cohomology of the Stacks project. Since the Stacks project is all about creating one consistent whole with lemmas proved at the correct level of generality, the chapter no longer has the flavor of lecture notes. Since some of you may prefer the original exposition and since the link to the lecture notes on Pugin’s home page no longer works, here are the original lecture notes by Thibaut Pugin, Zachary Maddock and Min Lee with all the mistakes and imperfections, etc, etc, etc. Enjoy!

Updated once more

Allright, I worked through your comments (as well as the suggestions we get via our email address) once more. Thanks to all. Go back there and leave some more comments.

Just a reminder: Once we reach 5000 pages we’re going to have a party — you’ll be invited via this blog. Currently we’re at 4665 pages. You can help make the party happen!

Intersection Theory

Today I finished the first complete version of a chapter on intersection theory. As usual comments and suggestion are very welcome.

The chapter uses Serre’s Tor formula and moving lemmas to define an intersection product on the Chow groups of nonsingular projective varieties over an algebraically closed ground field and that is all it does. You can read the introduction for a little bit more information.

There are some improvements that can be made to this chapter. The first is that some of the material on Serre’s Tor formula belongs properly in one of the chapters on commutative algebra. Of course, there is a lot more one can say about regular local rings and the Tor formula, leading up to recent work on homological conjectures in commutative algebra. Also, some of the arguments in the moving lemmas use geometric arguments on varieties over algebraically closed fields and we need to write more of the API to easily translate these into scheme theoretic language. Finally, the chapter is missing examples and more references to the literature.

What often happens with new chapters is that a few years down the road, we take a second look and make substantial improvements.

One aspect of the material in the new chapter is that it was not as straightforward to write as the material on constructible sheaves which was like butter. The conclusion must therefore be that intersection theory is not like butter!

Nonetheless: Enjoy!

Updated again

As you may have noticed I do not always immediately respond to the comments on the stacks project. It turns out to be more efficient to wait till there are quite a few of them and then take out a block of time to deal with them. I just finished with this once more and this time it took a bit more time. Thanks especially to Juan Pablo Acosta López.

Chow homology

Last week I started looking at the chapter on chow homology. When I first wrote it in 2009, it was a bit of an experiment. As a consequence, various proofs used three different approaches: one using K-groups, one using blow-up lemmas as in Fulton, and one via the key lemma (see below). Today I rearranged the whole thing to simplify the exposition and stick with the approach using the key lemma.

How much intersection theory does the chapter cover? We work consistently with schemes locally of finite type over a fixed universally catenary, locally Noetherian base scheme. We introduce cycle groups, flat pullback, proper pushforward, and rational equivalence. After proving some basic properties, we introduce the operation c_1(L) ∩ – where L is an invertible sheaf and the Gysin map for an effective Cartier divisor. Having proved the basic properties of these operations, we have enough to introduce chern classes of locally free sheaves and prove their basic properties. The chapter ends with stating the Grothendieck-Riemann-Roch theorem (without proof).

The key lemma is a statement about tame symbols over a Noetherian local domain of dimension 2. I think of it as the statement that the secondary ramifications add up to zero. I’d love it if you could tell me a reference for this lemma in the literature (I assume there is some paper on K-theory that contains this result).

The key lemma implies our key formula. I am sure I discussed this statement with somebody in my office at some point, but I cannot remember who; if it was you, please email me! Anyway, the key formula quickly implies that c_1(L) ∩ – passes through rational equivalence (Section Tag 02TG), the fact that c_1(L) ∩ c_1(N) ∩ – = c_1(N) ∩ c_1(L) ∩ -, and that the Gysin map for an effective Cartier divisor passes through rational equivalence (Section Tag 02TK).


Up to date again

With a very stuffy nose and lot’s of sneezing, today I worked through the comments that were left on the Stacks project. Thanks to all of you!

Here is what is very helpful (especially on those days were my head is all stuffed up): when you find a mistake, please also point out how to fix it (if you know how). Even just making a guess at what went wrong is helpful. Thanks!

Up to date with comments

Hello again! This is just to let you know that I’ve just finished working through the comments on the Stacks project and all (known) mistakes, typos etc are fixed. Thanks for all your help! Please go back there and find more.

Also, please let me know if you or somebody else have helped out in the past and we’ve forgotten to add your or their name to the contributors list. It seems idiotic, but it turns out it is actually kind of hard to keep track of everybody who is helping out (and in some sense we still do not have a really good way to do this). Any suggestions on how to manage this are very welcome too.


So I’ve been playing around with tracing the history of lemmas, etc in the Stacks project over time. I’ve written a bunch of scripts to extract this information out of the git logs. Here a pdf file showing what happened to Nakayama’s lemma over time. Enjoy!

PS: Feel free to improve on the current version of Nakayama’s lemma in the Stacks project and to submit it for inclusion, so we can add more stages to its history!


We clarify the discussion in this post resulting in a generalization of a result of Mike Artin.

Let X be a Noetherian algebraic space. Let T ⊂ X be a closed subspace. Let us denote X/T the formal completion of X along T (Tag 0AIX). Let W —> X/T be a rig-etale morphism of formal algebraic spaces, which means that

  1. W —> X/T is representable by algebraic spaces, i.e., it is an adic morphism of formal algebraic spaces Tag 0AQ2),
  2. W —> X/T is locally of finite type, i.e., it is etale locally on affine formal algebraic pieces given by a continuous ring map A —> B which is topologically of finite type Tag 0ALL),
  3. these ring maps A —> B have a completed cotangent complex whose cohomology groups are annihilated by an ideal of definition of A, for more details see Tag 0ALP and Tag 0AQK.

These three conditions correspond to condition (i) of Definition 1.7 of Artin’s paper “Formal Moduli: II”. The first result is that

Given W —> X/T rig-etale there exists a morphism of algebraic spaces Y —> X which is an isomorphism over X – T and whose completion Y/T —> X/T is isomorphic to W —> X/T.

In fact, Theorem 0ARB tells us we obtain an equivalence of categories.

This theorem does not often produce separated morphisms Y —> X if we start with a random W —> X/T. A typical thing that happens can be seen by starting with X equal to the affine line, T = {0} and W two copies of the completion of X at 0. Then the resulting Y is the affine line with 0 doubled.

Thus to get Artin’s theorem on dilatations we need to impose conditions on W —> X/T guaranteeing that Y —> X is separated or even proper. To do this we will use the notion of a rig-surjective morphism W’ —> W of locally Noetherian formal algebraic spaces W, W’ defined by requiring adic morphisms Spf(R) —> W with R a cdvr to lift to Spf(R’) —> W’ for some extension of cdvr R ⊂ R’ (Tag 0AQP). Let’s say an adic morphism W —> W’ of locally Noetherian formal algebraic spaces is a rig-monomorphism if the diagonal morphism is rig-surjective. In this language the conditions (ii) and (iii) from Definition 1.7 of Artin’s paper have the following interpretations:

  1. If W —> X/T as above is separated and a rig-monomorphism, then Y —> X is separated (Tag 0ARW),
  2. If W —> X/T as above is proper, a rig-monomorphism, and rig-surjective, then Y —> X is proper (Tag 0ARX).

The second statement recovers exactly Artin’s theorem on dilatations.

One typically applies the result to construct modifications Y —> X (Tag 0AD7) by taking the complete local ring A of X at a closed point x and setting W equal to the formal completion of the blow up of A at an ideal I ⊂ A which is locally principal on the puctured spectrum of A. Here the funny situation occurs that we can first read Tag 0ARX backwards over Spec(A) to conclude that W —> X/x has the required properties and then forwards to conclude that Y —> X is proper. In other situations it may not be that easy to verify the assumptions needed for the application of the theorem and it would behoove us to prove a few lemmas that help with this task.

Your help with this and other tasks is always welcome!