Math 216A. A gluing construction of Proj(S)

Similar documents
SEPARATED AND PROPER MORPHISMS

Math 248B. Base change morphisms

SEPARATED AND PROPER MORPHISMS

CHOW S LEMMA. Matthew Emerton

VALUATIVE CRITERIA FOR SEPARATED AND PROPER MORPHISMS

VALUATIVE CRITERIA BRIAN OSSERMAN

Descent on the étale site Wouter Zomervrucht, October 14, 2014

Math 754 Chapter III: Fiber bundles. Classifying spaces. Applications

The Segre Embedding. Daniel Murfet May 16, 2006

The Clifford algebra and the Chevalley map - a computational approach (detailed version 1 ) Darij Grinberg Version 0.6 (3 June 2016). Not proofread!

14 Lecture 14: Basic generallities on adic spaces

where Σ is a finite discrete Gal(K sep /K)-set unramified along U and F s is a finite Gal(k(s) sep /k(s))-subset

3 Lecture 3: Spectral spaces and constructible sets

HARTSHORNE EXERCISES

ALGEBRAIC GROUPS. Disclaimer: There are millions of errors in these notes!

Lecture 6: Etale Fundamental Group

MATH 8253 ALGEBRAIC GEOMETRY WEEK 12

The Proj Construction

Synopsis of material from EGA Chapter II, 4. Proposition (4.1.6). The canonical homomorphism ( ) is surjective [(3.2.4)].

Algebraic Geometry Spring 2009

(C) The rationals and the reals as linearly ordered sets. Contents. 1 The characterizing results

In the index (pages ), reduce all page numbers by 2.

Lecture 3: Flat Morphisms

Course 311: Michaelmas Term 2005 Part III: Topics in Commutative Algebra

Categories and Natural Transformations

COMPLEX VARIETIES AND THE ANALYTIC TOPOLOGY

Problem Set. Problems on Unordered Summation. Math 5323, Fall Februray 15, 2001 ANSWERS

Review of Linear Algebra

Reid 5.2. Describe the irreducible components of V (J) for J = (y 2 x 4, x 2 2x 3 x 2 y + 2xy + y 2 y) in k[x, y, z]. Here k is algebraically closed.

k k would be reducible. But the zero locus of f in A n+1

Math 203A, Solution Set 6.

Rings and groups. Ya. Sysak

Adic Spaces. Torsten Wedhorn. June 19, 2012

CW complexes. Soren Hansen. This note is meant to give a short introduction to CW complexes.

Homogeneous Coordinate Ring

FILTERED RINGS AND MODULES. GRADINGS AND COMPLETIONS.

Synopsis of material from EGA Chapter II, 3

Variations on a Casselman-Osborne theme

Synopsis of material from EGA Chapter II, 5

Math 396. Quotient spaces

2 Coherent D-Modules. 2.1 Good filtrations

CLASS NOTES MATH 527 (SPRING 2011) WEEK 6

3. The Sheaf of Regular Functions

INVERSE LIMITS AND PROFINITE GROUPS

Representation Theory of Hopf Algebroids. Atsushi Yamaguchi

Math 145. Codimension

Summer Algebraic Geometry Seminar

How to glue perverse sheaves

LIMITS AND COLIMITS. m : M X. in a category G of structured sets of some sort call them gadgets the image subset

12. Linear systems Theorem Let X be a scheme over a ring A. (1) If φ: X P n A is an A-morphism then L = φ O P n

7 Lecture 7: Rational domains, Tate rings and analytic points

NONSINGULAR CURVES BRIAN OSSERMAN

1 Categories, Functors, and Natural Transformations. Discrete categories. A category is discrete when every arrow is an identity.

Notas de Aula Grupos Profinitos. Martino Garonzi. Universidade de Brasília. Primeiro semestre 2018

Math 676. A compactness theorem for the idele group. and by the product formula it lies in the kernel (A K )1 of the continuous idelic norm

ALGEBRAIC GEOMETRY CAUCHER BIRKAR

ELEMENTARY SUBALGEBRAS OF RESTRICTED LIE ALGEBRAS

214A HOMEWORK KIM, SUNGJIN

SCHEMES. David Harari. Tsinghua, February-March 2005

MATH 221 NOTES BRENT HO. Date: January 3, 2009.

(dim Z j dim Z j 1 ) 1 j i

Math 396. Bijectivity vs. isomorphism

We simply compute: for v = x i e i, bilinearity of B implies that Q B (v) = B(v, v) is given by xi x j B(e i, e j ) =

FOUNDATIONS OF ALGEBRAIC GEOMETRY CLASS 43

Math 210B. Profinite group cohomology

FOUNDATIONS OF ALGEBRAIC GEOMETRY CLASS 25

HSP SUBCATEGORIES OF EILENBERG-MOORE ALGEBRAS

The homology of real subspace arrangements

1. Valuative Criteria Specialization vs being closed

ALGEBRAIC GEOMETRY COURSE NOTES, LECTURE 2: HILBERT S NULLSTELLENSATZ.

ZORN S LEMMA AND SOME APPLICATIONS

CATEGORIES. 1.1 Introduction

GENERALIZED ABSTRACT NONSENSE: CATEGORY THEORY AND ADJUNCTIONS

Formal power series rings, inverse limits, and I-adic completions of rings

A NOTE ON HENSEL S LEMMA IN SEVERAL VARIABLES

Exercises of the Algebraic Geometry course held by Prof. Ugo Bruzzo. Alex Massarenti

DIVISORS ON NONSINGULAR CURVES

Boolean Algebras, Boolean Rings and Stone s Representation Theorem

Math 215B: Solutions 1

AN INTRODUCTION TO AFFINE SCHEMES

1. Algebraic vector bundles. Affine Varieties

NOTES ON FINITE FIELDS

ALGEBRAIC GEOMETRY COURSE NOTES, LECTURE 9: SCHEMES AND THEIR MODULES.

Finite Dimensional Hilbert Spaces are Complete for Dagger Compact Closed Categories (Extended Abstract)

Honors Algebra 4, MATH 371 Winter 2010 Assignment 3 Due Friday, February 5 at 08:35

NAP Module 4. Homework 2 Friday, July 14, 2017.

Algebraic Geometry Spring 2009

Math 248B. Applications of base change for coherent cohomology

Math 6510 Homework 10

Remarks on the existence of Cartier divisors

GLUING SCHEMES AND A SCHEME WITHOUT CLOSED POINTS

B 1 = {B(x, r) x = (x 1, x 2 ) H, 0 < r < x 2 }. (a) Show that B = B 1 B 2 is a basis for a topology on X.

ne varieties (continued)

This is a closed subset of X Y, by Proposition 6.5(b), since it is equal to the inverse image of the diagonal under the regular map:

Yuriy Drozd. Intriduction to Algebraic Geometry. Kaiserslautern 1998/99

Honors Algebra 4, MATH 371 Winter 2010 Assignment 4 Due Wednesday, February 17 at 08:35

ALGEBRAIC GEOMETRY COURSE NOTES, LECTURE 4: MORE ABOUT VARIETIES AND REGULAR FUNCTIONS.

Chapter 2 Linear Transformations

Analytic continuation in several complex variables

Transcription:

Math 216A. A gluing construction o Proj(S) 1. Some basic deinitions Let S = n 0 S n be an N-graded ring (we ollows French terminology here, even though outside o France it is commonly accepted that N does not include 0). Morphisms between N-graded rings are understood to respect the grading. The irrelevant ideal is S + = n>0 S n ; keep in mind that we allow S 0 to have a nontrivial ideal theory (that is, it need not be a ield). An element S is homogeneous i S d or some d, and then d is unique i 0; we call d the degree o (when 0), and we consider 0 as having arbitrary degree. Note that the equation deg(g) = deg() + deg g is valid even i one o, g, or g vanishes, using the convention that 0 may be considered to have arbitrary degree. For example, S + is exactly the set o elements (including 0) with positive degree. For a general element S, the homogeneous parts o are the projections d o into each S d (so d = 0 or all but initely many d). An ideal I in S is homogeneous i an element = n 0 n o S lies in I i and only i each homogeneous part n lies in I. It is a simple exercise (inducting on degrees) to check that an ideal generated by homogeneous elements is a homogeneous ideal, and that homogeneous ideal I in S is prime i and only i it is a proper ideal and g I I or g I or homogeneous, g S. It is also clear that the kernel o a morphism o N-graded rings is a homogeneous ideal, and that or any homogeneous ideal I o S there is a natural N-grading on S/I. Deinition 1.1. Let S be an N-graded ring. The topological space Proj(S) has underlying set Proj(S) = {p a homogeneous prime such that S + p}, and the closed sets are the loci V (I) = {p Proj(S) I p} or homogeneous ideals I o S (context will prevent conusion with the analogous V (I) notation or aine schemes). It is easy to check that the V (I) s do satisy the axioms to deine the closed sets or a topology on Proj(S) (the empty set is V (S) and Proj(S) = V (0)). A homogeneous prime p ails to contain I i and only i there exists a homogeneous element I that does not lie in p (here we use crucially that I is homogeneous). Thus, a base o open sets or the topology on Proj(S) is given by loci D + () = {p Proj(S) p} = Proj(S) V (S) or homogeneous S. A crucial act is that it is even enough to take with positive degree: Lemma 1.2. A base o open sets or the topology on Proj(S) is given by loci D + () or homogeneous S +. Proo. Consider a homogeneous S and a point p D + (). We need to ind a homogeneous element g S + such that p D + (g) D + (). Since p does not contain S + (by the deinition o Proj(S)), there exists h S + not in p. Thus, the condition p implies h p, and h is homogenous with positive degree (possibly even h = 0) since both and h are homogeneous and deg h > 0. We conclude that p D + (h), and clearly D + (h) D + (). Beware that Proj(S) is generally not quasi-compact! For example, Proj(k[x 1, x 2,... ]) with ininitely many indeterminates o degree 1 is not quasi-compact, as it is covered by opens D + (x i ) and there is evidently no inite subcover. This issue is best understood as ollows: 1

2 Theorem 1.3. For an N-graded ring S, Proj(S) is empty i and only i all elements o S + are nilpotent. More generally, or positive-degree homogeneous elements and { i } i I in S, D + () D + ( i ) i and only i some power o lies in the homogeneous ideal generated by the i s. In particular, a collection o D + ( i ) s with all deg i > 0 covers Proj(S) i and only i every element o S + has some power lying in the homogeneous ideal generated by the i s. The contrast with Spec is o course that Spec(A i ) s cover Spec A i and only i the i s generate the unit ideal. The intererence o S + in the analogous covering criterion or Proj, coupled with the possibility that S + might not be initely generated, is the reason why Proj(S) can ail to be quasi-compact. On the other hand, in most interesting situations the ideal S + is initely generated and hence Proj(S) is quasi-compact. However, we note that some undamental constructions o Mumord in the study o moduli o abelian varieties rest crucially on the use o non-quasi-compact Proj s. Proo. Let I be the homogeneous ideal generated by the i s, so the complement o D + ( i ) is the set o p Proj(S) that contain I. Hence, we need to determine when D + () is disjoint rom the set o such p s, or equivalently when every p Proj(S) that contains I also contains ; we want to show that this condition is exactly the condition that a power o lies in I. Passing to the N-graded S/I, we are reduced to proving that a homogeneous S + lies in p or all p Proj(S) i and only i is nilpotent; keep in mind that has positive degree. One direction is obvious, and conversely we must prove that i S + is homogeneous o degree d > 0 and is not nilpotent, then there exists a homogeneous prime p such that p (and so S + p too, so p Proj(S)). We will make use o an auxiliary construction that will play an important role later. Let S (d) = n 0 S dn (so S (d) = S i d = 1). This is naturally an N-graded ring with vanishing graded pieces in degrees not divisible by d. Consider the localized ring (S (d) ) ; since (S (d) ) = S (d) [T ]/(1 T ), by assigning T degree d we see that (S (d) ) naturally has a Z-grading (with vanishing terms away rom degrees divisible by d). For example, s/ n is assigned degree deg(s) nd or homogeneous elements s S (d). Let (S (d) ) () (S (d) ) denote the direct summand o degree-0 elements in the Z-graded (S (d) ). This is a ring, and i is not nilpotent in S then it is not nilpotent in S (d), so then (S (d) ) 0 and hence the subring (S (d) ) () is nonzero. It then ollows that there exists a prime ideal q in (S (d) ) (). We will use this to construct a homogeneous prime p in S (d) that does not contain (and so in particular does not contain S (d) + since deg > 0); the ideal generated by the homogeneous a S such that a d S (d) lies in p is then readily checked to be a homogeneous prime ideal o S that does not contain (this rests crucially on the act that membership in the homogeneous p may be checked on component-parts). Let p be the contraction o q(s (d) ) under S (d) (S (d) ). The ideal p o S (d) does not contain, since otherwise q(s (d) ) would contain the degree-0 element 1, which is absurd since (q(s (d) ) ) (S (d) ) () = q is a proper ideal. To check that p is homogeneous prime, irst observe that (by contstruction) q(s (d) ) is a homogeneous ideal o the Z-graded (S (d) ), so p is a homogeneous ideal o the N-graded S (d). Hence, to veriy primality it is suicient to work with homogeneous elements. That is, we consider homogeneous a, a S (d) with respective degrees dn and dn and we assume aa p. Our goal is to prove a p or a p. Since aa p, the homogenous image o aa in (S (d) ) is contained in q(s (d) ), so aa = (x/ e ) r with r Z, x S kd, and x/ e q (S (d) ) (). Thus, by comparing degrees we get dn + dn = dr, so n + n = r. Hence, aa / r = (a/ n )(a / n ) (S (d) ) is a product o terms with degree 0. However, a a n = x n e (S(d) ) () (q(s (d) ) ) = q, so by primality o q in (S (d) ) () we conclude that at least o the degree-0 elements a/ n or a / n lies in q! Hence, either a or a in S (d) map into q(s (d) ) upon inverting, so by deinition either a or a lie in p.

3 2. First steps towards a scheme structure For homogeneous S +, we get an open set D + () Proj(S) consisting o those p Proj(S) that do not contain. These are a base o open sets, and we claim that D + () is naturally homeomorphic to Spec S (), where S () S is the degree-0 part o the Z-graded localization o S at the homogeneous. To deine a homeomorphism ϕ : D + () Spec S (), to each p D + () we associate the prime ideal p () = (ps ) S () Spec S () ; this is prime because it is the contraction o the prime ps = p o S under the ring map S () S (note that p is prime since p is a prime o S not containing ). Theorem 2.1. The map ϕ : D + () Spec(S () ) is a homeomorphism. Proo. For any homogeneous ideal a o S, we generalize the above operation on homogeneous prime ideals by deining ϕ(a) = (as ) S (). For any p D + (), we claim (1) ϕ(a) ϕ(p) a p. Once this is proved, it will ollow that ϕ is at least injective. The ( ) implication is obvious, and or the converse it suices to prove that i a a is a homogeneous element then a p. Let n = deg a 0 and let d = deg > 0. It ollows that a d n as S () = ϕ(a) ϕ(p) = ps S (), so there exists a homogeneous x p such that a d / n = x/ m in S with md = deg(x). Thus, or some e 0 we have e ( m a d n x) = 0 in S, and since p we must have m a d n x p. However, x p, so m a d p. Since p is prime, p, and d is positive, we conclude a p as desired. This completes the proo o injectivity or ϕ. Once we prove ϕ is surjective, and hence is bijective, (1) implies ϕ(v (a) D + ()) = V (ϕ(a)). Hence, or any ideal b o S (), the preimage a o bs in S is a homogeneous ideal satisying ϕ(a) = b. We may thereore conclude that every closed set V (b) in Spec S () corresponds (under the bijection ϕ) to a closed set V (a) D + () in D + (). However, all closed sets in D + () (with the subspace topology rom Proj(S)) have such a orm or some a, so we thereby get that ϕ is a homeomorphism. It remains to check that ϕ is surjective. A key observation is that the natural map (S (d) ) () S () is an isomorphism. The basic idea is that a degree-0 element in S () must have the orm x/ n with homogeneous x o degree deg(x) = nd S nd, so x is in S (d) ; the straightorward details are let to the reader (hint: equality o subrings o S ). Via this identiication, any prime ideal o S () may be considered as a prime ideal in (S (d) ) (). However, in the proo o Theorem 1.3 it was proved (check!) that every prime ideal q o (S (d) ) () has the orm ϕ(p) or some homogeneous prime p o S not containing (that is, or some p D + ()). Let us now write ϕ : D + () Spec(S () ) to denote the homeomorphism constructed above, with S + any positive-degree homogeneous element (so ϕ (p) = ps S () ). We shall use this homeomorphism to endow D + () with a structure o aine scheme, using the structure shea on Spec(S () ). In view o the act that the D + () s orm a base o opens in Proj(S), the key issue is to identiy S () as the ring o sections

4 on the open subset D + () Proj(S), and to this end it is useul to note that S () may be described entirely in terms o the subset D + () Proj(S) and the ring S without mentioning : Theorem 2.2. For homogeneous S +, let T be the multiplicative set o homogeneous elements g S such that g p or all p D + () Proj(S) (despite the notation, T only depends on D + () and not on ). The natural map S () (T 1 S) 0 to the degree-0 part o the Z-graded T 1 S induced by S T 1 S is an isomorphism. Proo. Let d = deg > 0. For injectivity, suppose x S is homogeneous o degree nd and the degree-0 element x/ n S maps to 0 in T 1 S. Hence, there exists g T such that gx = 0 in S. Replacing g with g d i necessary, we can assume deg g = md. Thus, (g/ m )(x/ n ) = 0 in S, and hence this equality holds in S (). By the deinition o T, or all p D + () we have g p, so g/ m is not contained in the prime ideal ϕ (p) = (ps ) S () o S () (as, g p). But ϕ is bijective onto Spec S (), so g/ m S () is not contained in any primes. It ollows that g/ m S () is a unit, so the vanishing o (g/ m )(x/ n ) in S () orces x/ n = 0 in S (). This gives exactly the desired injectivity. Now choose g T and x S with deg(x) = deg(g), so x/g (T 1 S) 0 makes sense. We seek a homogeneous a S o some degree nd (or some n 0) such that a/ n S () maps to x/g in T 1 S. We may replace x with g d 1 x and g with g d to get to the case deg g = md or some m 0. Thus, using the deinition o T and the bijectivity o ϕ we see that g/ m S () is not contained in any prime ideals, so it is a unit. In the degree-0 part o the Z-graded T 1 S we have x g = m g g m x g = m g so (g/ m ) 1 (x/ m ) S () maps to x/g (T 1 S) 0. This proves the desired surjectivity. x m, 3. A scheme structure on Proj(S) By Theorem 2.2, whenever, h S + are homogeneous elements such that D + (h) D + () inside o Proj(S) we have (by the deinitions!) T T h inside S, and so we get a canonical map (2) S () = (T 1 S) 0 (T 1 S) 0 = S (h) on degree-0 parts induced by the map T 1 S T 1 h S o Z-graded localizations. We may thereore consider the diagram o topological spaces D + () ϕ h Spec((T 1 S) 0 ) D + (h) ϕ h Spec((T 1 h S) 0) where the let column is the inclusion within Proj(S). One readily checks (upon reviewing the deinitions o the various maps) that this diagram commutes with the right side an open embedding, ultimately because the canonical equality (S () ) h deg / deg h = S (h) = (S (h) ) deg h /h deg inside o S h (check!) and the act that deg h /h deg S (h) (since T T h ) implies that (2) induces an isomorphism (S () ) h deg / S (h). deg h Clearly D + () D + (g) = D + (g), and by taking h = g above we see that this open subset o D + () is carried by ϕ onto the open subset Spec((S () ) g deg / deg g) Spec(S ()).

5 Likewise, as an open subset o D + (g) it is carried by ϕ g onto the open subset Spec((S (g) ) deg g /g deg ) Spec(S (g)). We now have put three scheme structures on D + (g), namely Spec S (g) and the two as basic opens in Spec S () and in Spec S (g). These three structures are identiied by means o the ring isomorphisms (3) (S () ) g deg / deg g S (g) (S (g) ) deg g /g deg that are really equalities as subrings o S g. Consequently, the cocycle condition or gluing is satisied (it comes down to transitivity or equality among three subrings o S (gh) or any three homogeneous, g, h S + ), so we may glue the structure sheaves O Spec(S() ) over the D + () s via (3). That is, we are gluing the Spec S () s (as ringed spaces) along the Spec S (g) s, where the underlying topological space Proj(S) o the gluing was made at the start. The glued structure shea over P = Proj(S) will be denoted O P, and so the ringed space (P, O P ) is covered by open subspaces (D + (), O P D+()) Spec(S () ) or homogeneous S +. Hence, (P, O P ) is a scheme. Deinition 3.1. Let S be an N-graded ring. The scheme Proj(S) is the topological space denoted Proj(S) above, equipped with the unique shea o rings O P whose restriction to D + () is O Spec(S() ) (using ϕ ) or all homogeneous S +, with the overlap-gluing isomorphism O Spec((S() ) g deg / deg g ) = O Spec(S() ) D+() D +(g) O Spec(S(g) ) D+(g) D +() = O Spec((S(g) ) deg g /g deg ) deined by the isomorphism Spec((S () ) g deg / g) Spec((S (g)) deg deg g /gdeg ) arising rom the canonical ring isomorphism in (3) or homogeneous, g S +. By Theorem 1.3, we obtain a useul alternative description: Corollary 3.2. Let { i } be a collection o homogeneous elements in S + such that every element o S + has some power contained in the ideal generated by the i s. The scheme Proj(S) is obtained by gluing the aine schemes Spec(S (i)) along the open aine overlaps Spec(S (i j)) Spec(S (i)) deined by the isomorphisms S (i j) (S (i)) deg i j / deg j. i Remark 3.3. We emphasize that there is content in this construction, namely that the above ring isomorphisms satisy triple overlap compatibility; this is most painlessly seen in terms o a triple equality o subrings o S (i j k ). Example 3.4. Let S = A[X 0,..., X n ] be an N-graded ring by putting A in degree 0 and declaring each X i to be homogeneous o degree 1. It ollows that Proj(S) is covered by the opens D + (X i ) = Spec S (Xi) = Spec A[X 0 /X i,..., X n /X i ] or 0 i n, and the gluing isomorphism is determined by the isomorphism (S (Xi)) Xj/X i (S (Xj)) Xi/X j deined by X k /X i (X k /X j ) (X i /X j ) 1 or k i. These are exactly the standard ormulas that express projective n-space as the gluing o n + 1 copies o aine n-space along certain open overlaps deined by non-vanishing o various coordinate unctions. Inspired by the above example, or any ring A we deine projective n-space over A to be P n A = Proj(A[X 0,..., X n ]) with the usual grading on A[X 0,..., X n ]. This is naturally a scheme over Spec A since each basic open aine D + () is naturally an A-scheme (as A has degree 0 in the N-grading being used) and the open-aine gluing data is one o A-algebras (more generally, Proj(S) is always naturally a scheme over Spec S 0 ). As a particularly degenerate example, we have P 0 A = Spec A[X 0] (X0) = Spec A.

6 Remark 3.5. I we assign A[X 0,..., X n ] an N-graded structure by putting A in degree 0 and assigning X i some positive degree d i, the resulting N-graded rings are generally not isomorphic as N-graded rings or dierent n-tuples d = (d 0,..., d n ), and their A-scheme Proj s (so-called weighted projective n-spaces over A with weights d) are generally not isomorphic to each other.