Intersection Theory and the Homological Conjectures in Commutative Algebra Paul C. Roberts Department of Mathematics, University of Utah, Salt Lake City, UT 84112, USA The subject of this article is a set of conjectures which have been a central topic of investigation in Commutative Algebra for several years. Many of these conjectures are closely related to questions in Intersection Theory, and in some cases they arose directly from attempts to define intersection multiplicities in an algebraic setting. We begin with a discussion of some of these questions in Intersection Theory and their influence in Commutative Algebra. Next we discuss a few of the main conjectures related to this topic. Finally, we show how recent developments in Intersection Theory have made it possible to settle some of these problems. 1. Introduction: Serre's Definition of Intersection Multiplicities We begin this section with some background on the problem of defining intersection multiplicities. The general question is as follows : given a variety V, a point p of V, and subvarieties X and Y of F such that p is an isolated point of the intersection X n Y, we wish to to define the multiplicity of the intersection of X and Y at the point p. We are interested here in an algebraic definition (which does not involve topology, for instance), and in one which is defined locally at the point. As an illustration, we consider the case in which X and Y are curves in the plane. Let A be the local ring at the point p; that is, the ring of rational functions defined in a neighborhood of p. Let the curves X and Y be defined locally near p as the sets of zeros of polynomials f(x,y) and g(x,y). Then the correct definition for the multiplicity is the length of the quotient ring (A/(f,g)), where (f,g) is the ideal generated by / and g (or, in the complex case, the dimension of A/(f,g) as a vector space over (C). This definition agrees with intuition; if the curves are smooth and not tangent at p it gives multiplicity one, and if they are tangent it is greater than one. In addition, it satisfies Bézout's Theorem, which states that if V is the projective plane, and if X and Y are subvarieties of V of degrees m and n respectively, then the total number of intersections, counted with multiplicities, is mn. This theorem is very important in enumerative geometry. Before proceeding, we recall some basic definitions and notation. Let A be a commutative Noetherian local ring with maximal ideal m and residue field A/m = k. Let dim (A) = d denote the Krull dimension of A; this can be defined Proceedings of the International Congress of Mathematicians, Kyoto, Japan, 1990 © The Mathematical Society of Japan, 1991 362 Paul C. Roberts either as the maximum length of a chain of prime ideals or as the minimum number d such that there are elements x\,...Xd G m which generate an ideal primary to the maximal ideal. Such a sequence of elements is called a system of parameters. The ring A is regular if the maximal ideal can be generated by d elements; in the geometric situation, the local ring at a point is regular when the variety is smooth at the point. The ring is Cohen-Macaulay when the system of parameters form a regular sequence; that is, when axt G (xi,...,Xf_i) implies a G (xi,...,Xj_i) for each i. We remark that the above definition of multiplicities for curves works because A is Cohen-Macaulay and (f,g) is a system of parameters in that case. If A is an integral domain, we say that A is equicharacteristic if the characteristic of A is the same as the characteristic of k; otherwise A has mixed characteristic. We now return to the question of defining intersection multiplicities. Let A be a local ring, and let / and J be ideals of A; in the geometric situation, J and J will be the ideals of functions vanishing on subvarieties X and Y. The straightforward generalization of the definition for curves would be to define the multiplicity to be the length of the quotient A/(I + J). However, this will not satisfy Bézout's Theorem (among other things). The idea of Serre was to add extra correction terms. In addition, it is more convenient to generalize from the quotients A/1 and A/J of the ring A to more general finitely generated modules M and N. Definition (Serre [28]). Let A be a regular local ring and let M and N be finitely generated ^4-modules such that M ®A N is a module of finite length. Then the intersection multiplicity of M and N is defined to be X(M,N) = ^(-l)4ength(Tor f (M,iV)). This definition makes sense since all modules over regular local rings have finite projective dimension. Furthermore, it satisfies Bézout's Theorem, and it agrees with the definition given above in the case of curves. Serre stated three additional properties and proved them in the equicharacteristic case. 1. dim(M)+ dim(N) < dim(A). 2. (Vanishing) If dim(M)+ dim(iV) < dim(A), then x(M,N) = 0. 3. (Positivity) If dim(M)+ dim(iV) = dim(A), then x(M,N) > 0. The first of these Serre proved in general; we note that it can be interpreted as a statement on the topology of intersections in smooth varieties. The second and third were proven in the equicharacteristic case, and were conjectured in the general regular case. Intersection Theory and the Homological Conjectures 363 2. The Peskine-Szpiro Intersection Theorem Attempts to prove Serre's conjectures in the ensuing years included various generalizations of them to non-regular rings. The three statements listed above make sense for any local ring as long as M is assumed to have finite projective dimension (and, of course, M ®AN has finite length). In fact, the first statement makes sense even without the assumption of finite projective dimension, but it is false in this generality; whether it holds with this assumption is not known. The next major advance was the proof of the "Intersection Theorem" in a quite general situation by Peskine and Szpiro [16, 17]. This theorem states: Theorem (Intersection Theorem). Let M be a module of finite projective dimension and let N be a module such that M ®A N has finite length as a module. Then dim N < proj dim M. If one replaces projective dimension in this theorem by codimension (i.e. dim(y4)— dim(M)), it becomes the first statement of Serre's conjectures, and, like that statement, it can be considered to describe the topology of the intersection of a variety which is the support of a module of finite projective dimension. The importance of the theorem was demonstrated by the fact that it implies several other conjectures (Peskine-Szpiro [17]): Corollary (Bass's Conjecture). If A has a non-zero finitely generated module of finite injective dimension, then A is Cohen-Macaulay. Corollary (Zero-Divisor Conjecture). If M is an A-module of finite projective dimension and x in A is a zero-divisor on A, then x is a zero-divisor on M. Shortly thereafter, a stronger version of this theorem was stated (PeskineSzpiro [18], Roberts [20]): Theorem (New Intersection Theorem). Let 0 - • Fk -> ... - • F 0 -> 0 be a complex of finitely generated free modules with homology of finite length. Then if Fm is not exact, we have dim(A) < k. This theorem implies the original Intersection Theorem. In the paper cited above, the Intersection Theorem was proven for rings of positive characteristic and for rings whose completion is the completion of a ring of finite type over a ring of characteristic zero. The method, which has since been applied successfully to a wide variety of questions, was to reduce (using the Artin Approximation Theorem) to the case of a ring of positive characteristic p, and then to use iterations of the Frobenius map (the map which sends a to ap) to prove the theorem in the positive characteristic case. The method of reducing using the Artin Approximation Theorem was extended by Höchster [11] to the general equicharacteristic case. These techniques did not extend to the case of mixed characteristic. 364 Paul C. Roberts 3. The Homological Conjectures The conjectures of Serre and the Intersection Conjecture in mixed characteristic, together with a number of others, became known as the Homological Conjectures. These conjectures dealt with intersection multiplicities, modules of finite projective dimension, the existence of Cohen-Macaulay modules, and properties of systems of parameters, together with relations between these topics. A summary of these conjectures in 1975 can be found in Höchster [12]; in this monograph Höchster introduced big Cohen-Macaulay modules and proved their existence (thus implying several other conjectures, including the Intersection Conjecture) in the equicharacteristic case. The proof made use of reduction to positive characteristic. In a somewhat different approach, in Roberts [21] the notion of CohenMacaulay complex is defined and proven to exist in the complex algebraic case via the Grauert-Riemenschneider vanishing theorem. The connection between the technique of reduction to positive characteristic and vanishing theorems in cohomology coming from Hodge theory was further shown by a proof of the Kodaira vanishing theorem by reduction to positive characteristic, conjectured by Szpiro and proven recently by Deligne and Illusie [2]. Among these conjectures, as mentioned above, was the question of whether Serre's vanishing and positivity conjectures held for arbitrary rings, when M was assumed to have finite projective dimension. This conjecture was proved for graded modules over a graded ring by Peskine and Szpiro [18] using computations with Hilbert polynomials in a free resolution, and they showed that several other conjectures followed from this one. These results led to specuation that an appropriate theory of Chern classes and Riemann-Roch theorem over general local rings could be used to prove the multiplicity conjectures, and hence many others, in general. The idea was that a bounded complex of free modules F. would define a sequence of Chern characters chj(F.), and that these would vanish for i less than the codimension of the support of Fm. In codimension one this program was carried out by Foxby [7], using a construction of MacRae [15], leading to proofs of some of the conjectures in low dimension. However, Dutta, Höchster and McLaughlin [4] constructed an example which showed that this generalization of Serre vanishing was false in dimension three. They construct a module M of finite projective dimension and finite length over the ring A = k[[X, Y,Z, W]]/(XY -ZW), which has dimension three, such that if jV = A/(X,Z), then x(M,N) = - 1 . Thus the theory of local Chern characters, even if it existed, could not have the hoped for vanishing properties. In the next section we describe such a theory and then show how in spite of these setbacks, it could solve some of the conjectures. 4. Localized Chern Characters The Chern characters alluded to in the previous section were defined by Baum, Fulton, and MacPherson [1]. There were two major problems in defining such a theory for complexes over a local ring. The first was that there is no cohomology theory. This problem was answered by defining Chern characters as operators on the Chow group. The second problem was that there are no non-trivial vector bundles over a local ring, so that straightforward definitions of Chern characters Intersection Theory and the Homological Conjectures 365 tend to give zero. The solution here was to give a "localized" definition, with values not in the Chow group of the entire ring, but in the Chow group of the support of the complex. We give a very brief summary of the part of the theory we need ; we refer to Fulton [8] for a more complete description. Let V = Spec(>4), and let Y be a closed subset of V. For each integer i, let Zi(Y) denote the rational group of cycles of dimension /'; that is, Z/(7) consists of the free Q-module with generators all reduced and irreducible subschemes of Y of dimension /, or, equivalently, all prime ideals P with dim(A/P) = /. Then the y-th graded piece of the Chow group, denoted Aj(Y), is Z,(7) modulo rational equivalence, where rational equivalence is defined by setting the divisor of a rational function on a reduced and irreducible suscheme of dimension i + 1 to zero. Let Fm be a bounded complex of free modules with support Z. Then there are Chern characters chk(Fm) for each k ^ 0, which map A^(X) to Ai-k(X n Z). The connection with Euler characteristics was given by the local RiemannRoch theorem (Fulton [8], Example 18.3.12). We state a very special case of this theorem here. Let F. be a complex of free modules with homology of finite length, or, equivalently, with support at the closed point of Spec(yl). Assume that A is an integral domain, and let [A] denote the class of Spec(y4) in the highest dimensional graded piece of the Chow group Ad(SpGc(A)). Then there is a class T(A), with [A] as its highest component, which satisfies X(Fm) = ^(äii(F.)(zi(A)). feO In this equation, the left hand side is an integer, and the right hand side is an element of the Chow group of the support of F.. In this case, the support of F. is a point p, so its Chow group reduces to Ao(p), which we identify with Q. Furthermore, if A is regular (or, more generally, a complete intersection), then T(A) = [A]. 5. The Proofs of Serre's Vanishing Conjecture and the Intersection Conjecture Using the theory described in the last section, it is possible to prove Serre's vanishing conjecture without regard to characteristic and to prove the PeskineSzpiro Intersection Theorem in mixed characteristic. We briefly outline these proofs here and attempt to show how the theory of localized Chern characters is used in their solution. Let M and N be modules over a regular local ring A such that M ®AN has finite length and dim(M) + dim(iV) < dim(>4). Let F. and G. be free resolutions of M and N respectively. The problem is to show that #(F. ® G.) = 0, since the homology of F. ® G. is precisely Tor(M,N). By the local Riemann-Roch formula we have x(F.®G.)=ch(F.®G.)([A\). We must now use some properties of local Chern characters. First, they are multiplicative, which means that we can write 366 Paul C. Roberts ch(F. ® G.)([A]) = £ chl(F.)(chi(G.))(M). i+j=d Let X and 7 denote the supports of M and N respectively; these are also the supports of F. and G.. Now suppose that d — j > dim(7). Then the element chj(G.)([A]), which is in the group Ad-j(Y), must be zero, since there can be no subvarieties of 7 of dimension larger than the dimension of 7. Thus all terms for which d—j> dim(7) are zero. Using the commutativity of Chern characters (proven in Roberts [23]), one shows also that terms with d — i> dvax(X) are zero. The hypothesis that dim(X) + dim(7) < d implies that all terms fall into one of these two sets, and therefore the entire sum is zero, as was to be shown. . This theorem also holds if A is a complete intersection. It was proven independently using Adams operations in X-theory by Gillet and Soulé [9,10]. It was proven when the singular locus has dimension at most one in Roberts [23]. We note that this argument gives a vanishing theorem for intersection multiplicities defined by local Chern characters for modules of finite projective dimension in general, and this approach appears to work better than the approach based on Euler characteristics for singular points. We next describe how localized Chern characters are used in the proof of the New Intersection Theorem in mixed characteristic. Let A be a domain of mixed characteristic of dimension d, and let 0 -> Fd-i ->...-> F0 -> 0 be a complex of free modules with homology of finite length which is not exact. We wish to prove that such a complex cannot exist. Let F. denote F. ® A/pA, where p is the characteristic of the residue field of A. Then F. is a complex of free modules over A/pA. The main idea of the proof is to compute chd-i([A/pA]) in two different ways. First, using the properties of localized Chern characters and the fact that F . comes from restriction to a divisor of a complex supported at one point, we deduce that chd-i([A/pA]) = 0. Second, we use that F. is a non-trivial complex of length d — 1 over a ring of positive characteristic of dimension d — 1. By means of an asymptotic Euler characteristic introduced by Dutta [3] and put on a firm foundation by Seibert [27], together with a local cohomology argument, we can then show that it must be positive (details can be found in Roberts [24, 25]). This contradiction proves the theorem. 6. Open Questions We conclude with some remarks on three questions which are still open. 1. Serre's Positivity Conjecture. The method of local Chern characters provides a geometric approach to the question of positivity, but as of yet it has not led to a solution. A result of Tennison [30] states that if one computes x(A/P,A/Q), where dim(A/P) + dim(A/Q) = dim(4), and if the subschemes defined by the prime ideals P and Q are not tangent at the closed point of Spec(R) (this condition can be stated precisely in terms of associated graded rings), then x(R/P,R/Q) is the product of the multiplicities of the A/P and A/Q and is therefore positive. We remark that an example of Roberts [26] shows that positivity is unlikely for two Intersection Theory and the Homological Conjectures 367 modules of finite projective dimension in the non-regular case, even using Chern characters rather than Euler characteristics, although this question is also open. 2. The Improved New Intersection Conjecture (Evans-Griffith [5, 6]). This conjecture states the following: Let F. = 0 -> Fk -» ... - • Fo -> 0 be a complex of free modules with finite length homology except possibly for Ho(F.). Assume that there is a minimal generator of Ho(Fm) which is annihilated by a power of the maximal ideal of A. Then k > dim(A). It seems reasonable to try to prove this theorem using the methods which worked for the "unimproved" version, but this version appears to be more difficult. This conjecture is equivalent to several others, such as Hochster's Monomial Conjecture (see [13]), and is open in dimension three in mixed characteristic. 3. Another Conjecture on Bounded Complexes. The New Intersection Theorem states that the minimum length of a bounded complex of free modules is d. Suppose that F. has length exactly d. In this case, the question is whether the cycles Z\ are integral over the boundaries B\ in degrees larger than zero. This result would imply the Monomial Conjecture, it can be proven in positive characteristic by the theory of tight closure of Höchster and Huneke [14], and it has been proven by Rees [19] for the Koszul complex in any characteristic. References 1. Baum, P., Fulton, W., MacPherson, R.: Riemann-Roch for singular varieties. Pubi. Math. IHES 45 (1975) 101-145 2. Deligne, P., Illusie, L.: Relèvements modulo p2 et décomposition du complexe de de Rham. Invent. math. 89 (1987) 247-270 3. Dutta, S.P.: Frobenius and multiplicities. J. Algebra 85 (1983) 424-448 4. Dutta, S.P., Höchster, M., McLaughlin, J.E. : Modules of finite projective dimension with negative intersection multiplicities. Invent, math. 79 (1985) 253-291 5. Evans, E.G., Griffith, P.: The syzygy problem. Ann. Math. 114 (1981) 323-333 6. Evans, E.G., Griffith, P. : The syzygy problem : a new proof and historical perspective. Commutative Algebra (Durham 1981), London Math Soc. Lecture Note Series 72 (1982) 2-11 7. Foxby, H.-B.: The MacRae invariant. Commutative Algebra (Durham 1981), London Math Soc. Lecture Note Series 72 (1982) 121-128 8. Fulton, W.: Intersection theory. Springer, Berlin Heidelberg New York 1984 9. Gillet, H., Soulé, C: K-théorie et nullité des multiplicités d'intersection. C. R. Acad. Se. Paris Série I, no. 3, 300 (1985) 71-74 10. Gillet, H., Soulé, C: Intersection theory using Adams operations. Invent, math. 90 (1987) 243-277 11. Höchster, M.: The equicharacteristic case of some homological conjectures on local rings. Bull. Amer. Math. Soc. 80 (1974) 683-686 12. Höchster, M. : Topics in the homological theory of modules over commutative rings. Regional Conference Series in Mathematics 24, 1975 13. Höchster, M.: Canonical elements in local cohomology modules and the direct summand conjecture. J. Algebra 84 (1983) 503-553 14. Höchster, M., Huneke, G: Tight closure, invariant theory, and the Briançon-Skoda Theorem. J. Amer. Math. Soc. 1 (1990) 31-116 15. MacRae, R.E.: On an application of the Fitting invariants. J. Algebra 2 (1965) 153-169 368 Paul G Roberts 16. Peskine, G; Szpiro, L.: Sur la topologie des sous-schémas fermés d'un schéma localement noethérien, définis comme support d'un faisceau cohérent localement de dimension projective finie. C. R. Acad. Sci. Paris Sér. A 269 (1969) 49-51 17! Peskine, G, Szpiro, L.: Dimension projective finie et cohomologie locale. Pubi. Math. IHES 42 (1973) 47-119 18. Peskine, G, Szpiro, L.: Syzygies et Multiplicités. G R. Acad. Sci. Paris Sér. A 278 (1974) 1421-1424 19. Rees, D.: Reduction of Modules. Math. Proc. Camb. Philos. Soc. 101 (1987) 431-449 20. Roberts, P. : Two applications of dualizing complexes over local rings. Ann. Sci. Ec. Norm. Sup. 9 (1976) 103-106 21. Roberts, R: Cohen-Macaulay complexes and an analytic proof of the New Intersection Conjecture. J. Algebra 66 (1980) 220-225 22. Roberts, P.: The vanishing of intersection multiplicities of perfect complexes. Bull. Amer. Math. Soc, 13 (1985) 127-130 23. Roberts, R : Local Chern characters and intersection multiplicities. Proc. Sympos. Pure Math. 46 2, Amer., Math. Soc, Providence, R.I. 1987, pp. 389-400 24. Roberts, P.: Le théorème d'intersection. C. R. Acad. Se. Paris Sér. I, no. 7, 304 (1987) 177-180 25. Roberts, P.: Intersection theorems. Commutative Algebra, Proceedings of an MSRI Microprogram. Springer, Berlin Heidelberg New York 1989, pp. 417-436 26. Roberts, P. : Negative intersection multiplicities on singular varieties. To appear in the Proceedings of the Zeuthen Conference, Copenhagen 1989 27. Seibert, G.: Complexes with homology of finite length and Frobenius functors. J. Algebra 125 (1989) 278-287 28. Serre, J.-R: Algèbre locale - multiplicités. (Lecture Notes in Mathematics, vol. 11.) Springer, Berlin Heidelberg New York 1961 29. Szpiro, L.: Sur la théorie des complexes parfaits. Commutative Algebra (Durham 1981), London Math. Soc. Lecture Note Series 72 (1982) 83-90 30. Tennison, B.R.: Intersection multiplicities and tangent cones. Math. Proc. Camb. Philos. Soc. 85 (1979) 33-42
© Copyright 2026 Paperzz