⃝ World Scientific Publishing Companyc DOI:10.1142/S021949881550098X
A remark on bigness of the tangent bundle of a smooth projective variety and D-simplicity of its section rings
Jen-Chieh Hsiao Department of Mathematics National Cheng Kung University
Tainan 70101, Taiwan jhsiao@mail.ncku.edu.tw
Received 3 December 2013 Accepted 8 September 2014
Published 23 April 2015 Communicated by S. P. Smith
We point out a connection between bigness of the tangent bundle of a smooth projective variety X over C and simplicity of the section rings of X as modules over their rings of differential operators. As a consequence, we see that the tangent bundle of a smooth projective toric variety or a (partial) flag variety is big. Some other applications and related questions are discussed.
Keywords: Big vector bundle; tangent bundle; D-simple ring.
Mathematics Subject Classification: 13N10, 14J60
1. Introduction
Let X be a smooth projective variety of dimension n over C and TXbe the tangent bundle of X. A well-known theorem of Mori [17] says that if TX is ample, then X is isomorphic to the projective space Pn. In [3], Campana and Peternell initiated a program to characterize projective manifolds whose tangent bundles are numerically effective (nef). In view of the theory of positivity of vector bundles [13, 14], it is natural to consider the following.
Problem 1.1. Classify projective manifolds whose tangent bundles are big.
Recall that a line bundle L on X is big if and only if the map φm: X!!" PH0(X, L⊗m)
defined by L⊗m is birational onto its image for some m > 0 [14, Definition 2.2.1].
A vector bundle E on X is big if the line bundle O(1) on the projective bundle
P(E) := Proj(Sym E) is big. Except for the case where E is nef (see the dis- cussion in Sec. 4), it is in general not clear how to determine the bigness of a vector bundle. Here, we propose the following criterion for bigness of the tangent bundle TX.
Theorem 1.2. If X admits an ample line bundle whose section ring S is simple as a module over its ring of C-linear differential operators DS, then TX is a big vector bundle.
In fact, it is shown that if TX is not big, then the section ring S associated to any ample line bundle on X has no differential operators of negative weight. In particular, the maximal graded ideal of S is a DS-submodule (S is not DS-simple) and DS itself is not a simple ring.
When dim X = 1, Theorem1.2was proved by Levasseur [15] following the results in [1, 12]. The essential ideas due to Bernˇste˘ın, Gel′fand, Gel′fand [1] show that there exist no differential operators of negative weight on
S = C[x, y, z]/⟨x3+ y3+ z3⟩
by using the geometric properties of the cubic curve Proj S. These ideas generalize readily to the higher-dimensional cases (see Sec.2).
As the first consequence of Theorem1.2, we have the following corollary whose proof can be found in Sec.3.
Corollary 1.3. The tangent bundle on a smooth projective toric variety or on a (partial) flag variety is big.
Another application of Theorem1.2is that if TX is not big and if S is a quasi- Gorenstein section ring on X, then the multiplier ideal J (S) is a DS-module (see Remark3.4). This can be regarded as a characteristic 0 analog of Theorem 2.2(1) in [19]: every test ideal is a D-module.
In the case where TX is nef, the bigness of TX is equivalent to the positivity of the Segre class sn(TX∗) [8, Sec. 3.1]. Using this and the results in [3], we show in Proposition4.1that if TX is nef and big then X is Fano (i.e. −KX is ample). This provides higher-dimensional examples of smooth projective varieties whose section rings are not D-simple (e.g. abelian varieties or other non-Fano manifolds with nef TX), generalizing the classical one-dimensional examples in [1, 15]. On the other hand, Proposition 4.1 gives a partial solution to Problem 1.1 when TX is nef. In fact, following the discussions in Sec.5we expect that the tangent bundle TX on a Fano manifold X is always big. When dim X ≤ 3 and TX is nef, this is verified by computing the Segre class sn(TX∗) using the classification in [3]. See the paragraph after Question4.5for more details.
To end the introduction, we mention that inspired by the proof of Proposi- tion4.1, one can prove a conjecture in [3] that states: if TX is nef and χ(OX) = 1 then X is Fano (see Theorem4.3).
2. The Proof of Theorem 1.2
Except for Proposition2.1, most parts of the following proof of Theorem1.2come from the arguments in [1,12, 15]. We reproduce them here for the convenience of the reader.
Fix an ample line bundle L on X and set Si := H0(X, Li). Consider the asso- ciated section ring
S = SL:= !
i∈Z≥0
Si.
Since X is smooth, it is well-known that S is a N-graded normal domain that is finitely generated over C = S0. The maximal graded ideal of S will be denoted by
m:=!
i∈N
Si.
Choose a representation T/J of S, where T = C[x1, . . . , xs] is a polynomial ring whose variable xihas weight di and J is a homogeneous prime ideal. Thanks to the ampleness of L, the graded ring S satisfies the condition (♯) in [12] which states that there exists l0 such that for every l ≥ l0, S(l) = "i∈NSil is generated by Sl= [S(l)]1 over S0= C.
Let DS be the ring of C-linear differential operators on S in the sense of Grothendieck. By definition, the ring DS has an increasing filtration {Dm}m≥0
by the order of the differential operators, where D0= S and for m > 0 Dm= {δ ∈ HomC(S, S) | [δ, s] ∈ Dm−1 for all s ∈ S}.
Following [1] or [18], we decompose DS as DS= S! #
m≥1
Derm(S), where
Derm(S) := {δ ∈ DS| ord δ ≤ m and δ(1) = 0}
is the set of all derivations of order ≤ m. An element δ ∈ Derm(S) is called homoge- neous of weight l (l ∈ Z) if δ(Si) ⊆ Si+l for all i. In particular, the Euler derivation
I =
$s i=1
dixi
∂
∂xi ∈ Der1(S)
is homogeneous of weight 0. Moreover, Derm(S) can be decomposed as Derm(S) =!
l∈Z
Derml (S),
where Derml (S) is the space of homogeneous derivations of S with weight l and order ≤ m. Notice that if
Derml (S) = 0 for all m ≥ 0 and l < 0, (2.1) then the maximal ideal m is a DS-submodule of S; in particular, the ring S is not DS-simple. Our goal is to show that non-bigness of TX implies the condition (2.1).
Let ˆX = Spec(S) be the cone over X and U := ˆX\{m} be the punctured spectrum. We have the natural projection
π : U → X = Proj S.
Let Derm be the sheaf of derivations on U of order ≤ m and let Derlm be the sheaf of derivations δ on U of order ≤ m that satisfy the condition [I, δ] = Iδ − δI = lδ and denote ∆ml := π∗Derml . Since Derm(S) is the dual of the mth-order differential module Ωm(S) = I/Im+1where I is the diagonal ideal of S ⊗CS, it is reflexive. Moreover, since S is normal, standard facts about reflexive sheaves [10, Proposition 1.6] implies that Γ(U, Derm) = Derm(S). In particular, by comparing the weight of derivations we have
Γ(X, ∆ml ) = Γ(U, Derlm) = Derml (S).
For |l| ≥ l0 in condition (♯), one can check locally that ∆ml ∼= ∆m0 ⊗ Ll (m ≥ 0) where this isomorphism is compatible with the natural embedding ∆ml '→ ∆nl
(m < n) [12, Lemma 6]. In particular, we have the following exact sequence 0 → ∆ml −1 → ∆ml → σm⊗ Ll→ 0 (m ≥ 2, |l| ≥ l0). (2.2) Here σm:= ∆m0/∆m0−1 (m ≥ 2) and we set σ1= ∆10.
By [12, Lemma 6(3)], we have the following short exact sequence that comes from the Euler derivation
0 → OX ·I
−→ σ1→ TX→ 0. (2.3)
Taking symmetric power and using the facts that σm= Symmσ1[12, Lemma 6(3)]
and TX are locally free, we have the exact sequence [12, Lemma 7]
0 → σm−1→ σm→ SymmTX→ 0 (m ≥ 2). (2.4) On the other hand, we need the following.
Proposition 2.1. If TX is not big, then
H0(X, SymmTX⊗ Ll) = 0 for m ≥ 1 and l < 0.
Proof. This can be achieved by a similar argument as in [14, Example 6.1.23]:
Suppose H0(X, SymmTX ⊗ A−1) ̸= 0 for some m and some ample line bundle A on X. Since O(1) := OP(TX)(1) is π-ample, O(1) ⊗ π∗(A⊗k) is ample for k sufficiently large. Denote ξ = c1(O(1)) and α = c1(A). By the projection formula, π∗(O(m) ⊗ π∗A−1) = SymmTX⊗ A−1. So H0(P(TX), O(m) ⊗ π∗A−1) ̸= 0 and hence mξ − π∗α is an effective class. For k sufficiently large,
(km + 1)ξ = [kmξ − π∗(kα)] + [ξ + π∗(kα)]
is a sum of an effective class and an ample class. Therefore, ξ is big.
From now, assume that TX is not big. By Proposition2.1, we have H0(X, TX⊗ Ll) = 0 for l < 0.
The global evaluation of (2.3) tensored with Lltogether with the Kodaira vanishing theorem imply that
Der1l(S) = H0(X, ∆1l) = 0 (l < 0).
Moreover, the global evaluation of (2.4) tensored with Lltogether with the vanish- ing
H0(X, SymmTX⊗ Ll) = 0 (m ≥ 2, l < 0) imply that
H0(X, σm⊗ Ll) = H0(X, σ1⊗ Ll) = H0(X, ∆1l) = 0 (m ≥ 1, l < 0).
Therefore, it follows from the global evaluation of (2.2) ⊗ Llthat Derml (S) = H0(X, ∆ml ) = H0(X, ∆1l) = 0 (m ≥ 1, l ≤ −l0).
Now, for any δ ∈ Derml (S)(m ≥ 1, l < 0), we have δk∈ Derkmkl (S) = 0 for k sufficiently large. This forces δ = 0, and the proof of Theorem1.2is finished.
3. Some Applications
The following fact ([19, Proposition 3.1] or [16, Proposition 3.5]) says that D-simplicity is preserved when considering pure subrings.
Proposition 3.1. Let S and T be arbitrary commutative algebra over a commu- tative ring A. Suppose that S '→ T and that this inclusion splits as a map of S-modules. If T is a simple DA(T )-module, then S is a simple DA(S)-module.
In particular, if S admits a split embedding into a polynomial algebra C[x1, . . . , xt], then S is DS-simple.
We are now ready to prove the main corollary in the introduction which states:
the tangent bundle on a smooth projective toric variety or on a (partial) flag variety is always big.
Proof of Corollary 1.3. For the toric case, this follows from the facts that every section ring of a smooth projective toric variety is a normal semigroup ring [6, Sec. 3.4] and that every positive normal semigroup ring admits a split embedding into a polynomial algebra [2, 6.1.10]. Alternatively, one can check directly that every normal semigroup ring S is DS-simple (see e.g. [11, Theorem 3.7]).
For the flag case, note that the homogeneous coordinate ring of a (partial) flag variety under the Pl¨ucker embedding is isomorphic to the ring of invariants RG of a polynomial algebra R under the action of certain linearly reductive group G [7, Sec. 9.2]. The existence of a Reynolds operator ρ : R → RG [2, Sec. 6.5]
guarantees that the embedding RG→ R is split.
Remark 3.2. The flag case of Corollary 1.3 recovers the bigness of TP(TP2) in [5, Example 2].
Remark 3.3. It is pointed out by Mustat¸˘a that the toric case of Corollary 1.3 can also be recovered by using the Euler exact sequence. On the other hand, the pseudoeffective cone of a projectivized toric vector bundle P(F) on a toric variety X is described using Klyachko filtration [9]. In the case where F is the tangent bundle TX, the pseudoeffective cone of P(TX) is generated by the classes DH, for hypersurfaces H in PF (the π-fiber over 1T), and the classes Di (the preimage under π of the torus invariant prime divisors on X). As Payne mentioned to us, the results in [9] might be helpful to re-establish the bigness of tangent bundles on smooth projective toric varieties.
We mention another consequence of Theorem1.2.
Remark 3.4. For a normal standard graded ring S that is quasi-Gorenstein and has an isolated singularity, the multiplier ideals J (S) of S is a power of the max- imal ideal m [21, Example 3.7]. Therefore, if the section ring S in Theorem1.2 is quasi-Gorenstein, then J (S) is a DS-submodule of S. This can be regarded as a characteristic 0 analog of [19, Theorem 2.2(1)]: every test ideal is a D-module.
4. The Case where TX is nef
Recall that a vector bundle E of rank r + 1 on X is numerically effective (nef) if the line bundle O(1) on P(E) is nef, i.e. c1(O(1)) · C ≥ 0 for all curves C in P(E).
In this case, the vector bundle E is big if and only if the volume of OP(E)(1)
%
P(E)c1(OP(E)(1))n+r =%
X
sn(E∗) > 0,
where sn(E∗) is the nth Segre class of the dual of E [13, Sec. 2.2.C], [8, Sec. 3.1].
The reader is cautioned that in [8], P(E) denotes the projective bundle of lines in E.
The study of projective manifolds with nef tangent bundle was initiated in [3].
In general, if TX is nef, then the Chern classes ci:= ci(X) = ci(TX) ≥ 0.
In this case, we have (−KX)n = cn1 ≥ 0 and the inequality is strict if and only if X is Fano (i.e.−KX is ample) [4, Theorem 1.2]. Moreover, the Schur polynomial [8, Example 12.1.7]
∆λ= det(cλi+j−i)1≤i,j≤n≥ 0
for any partition λ = (λ1, λ2, . . . , λn) of n with n ≥ λ1≥ λ2≥ · · · ≥ λn ≥ 0. Note that by [8, Lemma A.9.2]
sn(TX∗) = ∆(1n), where (1n) = (1, 1, . . . , 1),
so TX is big if and only if ∆(1n) > 0. On the other hand, for a partition µ = (µ1, . . . , µn) of n with n ≥ µ1≥ · · · ≥ µn ≥ 0, denote
cµ= cµ1· cµ2· · · cµn.
It follows from the Pieri’s formula that [7, Sec. 2.2]
cµ=$
λ
Kλµ∆λ, (4.1)
where Kλµ is the Kostka number and the sum runs over all partitions of n. We note that Kλλ= 1 and that
Kλµ̸= 0 if and only if µ # λ (i.e. µ1+ · · · + µi≤ λ1+ · · · + λi for all i).
Proposition 4.1. If TX is nef and big, then X is Fano.
Proof. Form the discussion above, it suffices to show cn1 = 0 ⇒ ∆(1n)= 0.
By (4.1),
cn1 = c(1n)=$
λ
Kλ(1n)∆λ= 0.
Since (1n)# λ for all partition λ, we have Kλ(1n)> 0 and hence ∆λ = 0 for all λ.
In particular, ∆(1n)= 0.
Remark 4.2. Proposition 4.1 provides new examples where DS is not simple, generalizing the classical examples in [1,15].
Inspired by the proof of Proposition 4.1, we prove the following theorem that was conjectured in [3] and was verified for n = dim X ≤ 4 [4].
Theorem 4.3. Let X be a projective manifold with TX nef. If χ(OX) ̸= 0, then X is Fano.
Proof. It suffices to show that cn1 = 0 implies χ(OX) = 0. By Hirzebruch–
Riemann–Roch [8, Corollary 15.2.2], χ(OX) =%
X
tdn(TX).
Since the nth Todd polynomial tdn(TX) of TX is a Q-linear combination of the cλ’s [8, Example 3.2.4], we only have to show that cλ= 0 for all λ ∈ Λn. Again, it follows from (4.1) that
[cn1 = 0] =⇒ [∆λ= 0 for all λ ∈ Λn] =⇒ [cλ= 0 for all λ ∈ Λn].
Example 4.4. We give some explicit examples to illustrate (4.1).
(1) When n = 1, c1= ∆(1).
(2) When n = 2, write λ1= (1, 1) and λ2= (2, 0). The Segre classes
∆(1,1)= c21− c2= cλ1− cλ2 and ∆(2,0)= c2= cλ2.
We have
(cλ1, cλ2) = (∆λ1, ∆λ2) ·&1 0 1 1 '
.
(3) When n = 3, write λ1= (1, 1, 1), λ2= (2, 1, 0) and λ3= (3, 0, 0). We have
(cλ1, cλ2, cλ3) = (∆λ1, ∆λ2, ∆λ3) ·
⎛
⎝
1 0 0 2 1 0 1 1 1
⎞
⎠.
(4) When n = 4, write λ1 = (1, 1, 1, 1), λ2 = (2, 1, 1, 0), λ3 = (2, 2, 0, 0), λ4 = (3, 1, 0, 0) and λ5= (4, 0, 0, 0). We have
(cλ1, cλ2, cλ3, cλ4, cλ5) = (∆λ1, ∆λ2, ∆λ3, ∆λ4, ∆λ5) ·
⎛
⎜⎜
⎜⎜
⎝
1 0 0 0 0 3 1 0 0 0 2 1 1 0 0 3 2 1 1 0 1 1 1 1 1
⎞
⎟⎟
⎟⎟
⎠.
Conversely, we ask the following.
Question 4.5. If X is Fano with nef TX, is it true that TX is big?
Using the classification in [3], Question 4.5 can be verified for n ≤ 3 by com- puting the Segre class ∆(1n) or quoting the known results.
(1) When n = 1, the only Fano curve is P1and TP1 is ample.
(2) When n = 2, the only Fano surfaces with nef TX are P2 and P1× P1. These are toric varieties and hence have big tangent bundles.
(3) When n = 3, the only Fano threefolds with nef TXare P3, P1×P2, P1×P1×P1, Q3, and P(TP2). The first three are toric varieties. The last two have big tangent bundles by [5, Example 1,2].
A stronger version of Question4.5(dropping the nefness assumption) has to do with D-simplicity of the section ring S (see Question5.2).
5. Questions on D-Simplicity
Let Y be an affine algebraic variety over C. When Y is smooth, it is well-known that its ring of differential operators DY is simple and Noetherian. The example in [1] shows that DY may not be simple nor Noetherian. On the other hand, it is shown in [16] that DY is simple if Y is a geometric invariant quotient of classical algebraic group acting on an affine space. It is asked in [16] whether the same result holds for other linear algebraic groups. See also Conjectures 1.1 and 1.2 of [22], in which the conjectures are proved in positive characteristics. Unfortunately, in view of Proposition3.1, Theorem 1.2 is not helpful in finding counterexamples to this problem.
The results in [16] also suggest that DY might have nice properties when Y has only rational singularities. Although this is not true in general, it was proposed in [19] that Gorensteinness should play a role in answering this question. In fact, the main theorem in [19] shows that in positive characteristics a domain is strongly F -regular if and only if it is F -split and D-simple. In view of the correspondence between strong F -regularity and log terminal singularities and the fact that Goren- stein log terminal singularities are rational singularities, it is natural to ask the following question.
Question 5.1. Let Y be an affine variety over C that is Gorenstein. Is it true that Y has at worst rational singularities if and only if OY is a simple DY-module?
By [20, Proposition 6.2(2)], a normal projective variety is Fano and has rational singularities if and only if it admits a Gorenstein section ring with rational singu- larities. Therefore, Theorem 1.2suggests the following question in Fano geometry.
Question 5.2. Does there exist a smooth Fano variety over C whose tangent bun- dle is not big?
A positive answer to Question5.2will provide a counterexample to the necessary part of Question5.1.
Acknowledgment
We thank Thomas Peternell and Sam Payne for the correspondences concerning the bigness of tangent bundles. We also thank Shin-Yao Jow, Ching-Jui Lai, and Mircea Mustat¸˘a for their interests in this work. Special thanks goes to Holger Brenner, who helped to improve the presentation of the results and indicated possible directions for further developments. The author was partially supported by National Science Council in Taiwan under grant 101-2115-M-006-011-MY2.
References
[1] I. N. Bernˇste˘ın, I. M. Gel′fand and S. I. Gel′fand, Differential operators on a cubic cone, Uspehi Mat. Nauk 27(1)(163) (1972) 185–190, MR 0385159 (52 #6024).
[2] W. Bruns and J. Herzog, Cohen–Macaulay Rings, Cambridge Studies in Advanced Mathematics, Vol. 39 (Cambridge University Press, Cambridge, 1993), MR MR1251956 (95h:13020).
[3] F. Campana and T. Peternell, Projective manifolds whose tangent bundles are numer- ically effective, Math. Ann. 289(1) (1991) 169–187, MR 1087244 (91m:14061).
[4] F. Campana and T. Peternell, 4-folds with numerically effective tangent bundles and second Betti numbers greater than one, Manuscripta Math. 79(3–4) (1993) 225–238, MR 1223018 (94e:14052).
[5] L. E. Sol´a Conde and J. A. Wi´sniewski, On manifolds whose tangent bundle is big and 1-ample, Proc. London Math. Soc. (3) 89(2) (2004) 273–290, MR 2078708 (2005c:14018).
[6] W. Fulton, Introduction to Toric Varieties, Annals of Mathematics Studies, Vol. 131 (Princeton University Press, Princeton, NJ, 1993), the William H. Roever Lectures in Geometry, MR MR1234037 (94g:14028).
[7] W. Fulton, Young Tableaux, London Mathematical Society Student Texts, Vol. 35 (Cambridge University Press, Cambridge, 1997), with applications to representation theory and geometry, MR 1464693 (99f:05119).
[8] W. Fulton, Intersection Theory, 2nd edn., Ergebnisse der Mathematik und ihrer Grenzgebiete, 3. Folge, A Series of Modern Surveys in Mathematics [Results in Math- ematics and Related Areas, 3rd Series, A Series of Modern Surveys in Mathematics], Vol. 2 (Springer-Verlag, Berlin, 1998), MR 1644323 (99d:14003).
[9] J. Gonz´alez, M. Hering, S. Payne and H. S¨uß, Cox rings and pseudoeffective cones of projectivized toric vector bundles, Algebra Number Theory 6(5) (2012) 995–1017, MR 2968631.
[10] R. Hartshorne, Stable reflexive sheaves, Math. Ann. 254(2) (1980) 121–176, MR 597077 (82b:14011).
[11] J.-C. Hsiao, D-module structure of local cohomology modules of toric algebras, Trans.
Amer. Math. Soc. 364(5) (2012) 2461–2478, MR 2888215.
[12] Y. Ishibashi, Remarks on a conjecture of Nakai, J. Algebra 95(1) (1985) 31–45, MR 797652 (87b:13004).
[13] R. Lazarsfeld, Positivity in Algebraic Geometry I, Ergebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results in Mathematics and Related Areas, 3rd Series, A Series of Modern Surveys in Math- ematics], Vol. 48 (Springer-Verlag, Berlin, 2004), Classical setting: line bundles and linear series, MR 2095471 (2005k:14001a).
[14] R. Lazarsfeld, Positivity in algebraic geometry II, Ergebnisse der Mathematik und ihrer Grenzgebiete, 3. Folge, A Series of Modern Surveys in Mathematics [Results in Mathematics and Related Areas, 3rd Series, A Series of Modern Surveys in Math- ematics], Vol. 49 (Springer-Verlag, Berlin, 2004), Positivity for vector bundles, and multiplier ideals, MR MR2095472 (2005k:14001b).
[15] T. Levasseur, Op´erateurs diff´erentiels sur les surfaces munies d’une bonne C∗- action, S´eminaire d’Alg`ebre Paul Dubreil et Marie-Paul Malliavin, 39`eme Ann´ee (Paris, 1987/1988), Lecture Notes in Mathematics, Vol. 1404 (Springer, Berlin, 1989), pp. 269–295, MR 1035229 (91g:14026).
[16] T. Levasseur and J. T. Stafford, Rings of differential operators on classical rings of invariants, Mem. Amer. Math. Soc. 81(412) (1989) vi+117, MR 988083 (90i:17018).
[17] S. Mori, Projective manifolds with ample tangent bundles, Ann. of Math. (2) 110(3) (1979) 593–606, MR 554387 (81j:14010).
[18] Y. Nakai, High order derivations. I, Osaka J. Math. 7 (1970) 1–27, MR 0263804 (41
#8404).
[19] K. E. Smith, The D-module structure of F -split rings, Math. Res. Lett. 2(4) (1995) 377–386, MR 1355702 (96j:13024).
[20] K. E. Smith, Globally F-regular varieties: Applications to vanishing theorems for quo- tients of Fano varieties, Michigan Math. J. 48 (2000) 553–572, Dedicated to William Fulton on the occasion of his 60th birthday, MR 1786505 (2001k:13007).
[21] K. E. Smith, The multiplier ideal is a universal test ideal, Comm. Algebra 28(12) (2000) 5915–5929, Special issue in honor of Robin Hartshorne, MR 1808611 (2002d:13008).
[22] K. E. Smith and M. Van den Bergh, Simplicity of rings of differential operators in prime characteristic, Proc. London Math. Soc. (3) 75(1) (1997) 32–62, MR 1444312 (98d:16039).
