betaFeedback

WorksMathematics

A Complete Local Domain without a Small Cohen–Macaulay Module

Abstract

We construct a three-dimensional complete Noetherian normal local domain containing C\mathbb{C}, with residue field C\mathbb{C}, that has no nonzero finitely generated maximal Cohen–Macaulay module. This disproves the domain form of the small Cohen–Macaulay module conjecture.

Introduction

A small Cohen–Macaulay module over a Noetherian local ring (R,m)(R,\mathfrak{m}) is a nonzero finitely generated RR-module MM such that depth⁡RM=dim⁡R\operatorname{depth}_{R} M = \dim R. Such a module is also called a maximal Cohen–Macaulay module. The domain form of the small Cohen–Macaulay module conjecture asks whether every complete Noetherian local domain has one. We prove the following negative answer.

Theorem 1.1. There exists a three-dimensional complete Noetherian normal local domain RR containing C\mathbb{C}, with residue field C\mathbb{C}, such that no nonzero finitely generated RR-module has depth three.

The distinction between modules and algebras matters here. A module-finite extension domain R⊆BR \subseteq B that is Cohen–Macaulay supplies such a module, but the converse need not hold. Our obstruction applies to every finite module; it requires neither a grading nor an algebra structure. Hochster formulated the finite-module existence problem in his early work on Cohen–Macaulay modules and homological conjectures [13, 14]. In dimensions at most two, a complete local domain has a small Cohen–Macaulay module: its finite normalization is Cohen–Macaulay [15 Section 2]. There are also important positive results in the graded setting. The theorem of Hartshorne, Peskine–Szpiro, and Hochster gives graded small Cohen–Macaulay modules for three-dimensional finitely generated nonnegatively graded domains whose degree-zero part is a perfect field of positive characteristic; see [17 Theorem 5.2] for the statement and historical attribution. These results do not cover the characteristic-zero cones considered here. Hochster’s later account explicitly records both his early existence conjecture and his subsequent expectation of counterexamples [15 Section 2, Conjectures 2.1 and 2.2].

Dropping finite generation changes the problem. In the balanced convention, a big Cohen–Macaulay module is a possibly infinitely generated module on which every system of parameters is a regular sequence and whose quotient by the maximal ideal is nonzero. Hochster proved existence in equal characteristic [14]. Hochster–Huneke showed that the absolute integral closure of an excellent local domain of positive characteristic is such an algebra [16]. Perfectoid methods later yielded big Cohen–Macaulay algebras in mixed characteristic, completing the existence theorem for Noetherian local rings [1 Theorem 0.7.1]. These results impose no finite-generation conclusion.

For a normal local domain containing the rational numbers, trace makes the ring a direct summand of every module-finite extension domain. A Cohen–Macaulay extension would therefore force the original ring to be Cohen–Macaulay. Bhatt constructed positive-characteristic complete normal local domains with no finite Cohen–Macaulay extension, using a Witt-vector cohomological obstruction [6]. Neither obstruction excludes arbitrary finite modules. Bhatt, Hochster, and Ma instead study asymptotic substitutes for small Cohen–Macaulay modules. Their August 25, 2026 version describes the existence problem as open even for local rings at maximal ideals of affine three-dimensional domains over algebraically closed fields [7 Sections 1.1 and 1.3].

The geometric input is a smooth integral projective surface XX over C\mathbb{C} with an ample globally generated line bundle OX(H)\mathcal{O}_{X}(H). Write

T(X,H)=⨁j≥0H0(X,OX(jH)),R(X,H)=T(X,H)T(X,H)>0^.T(X,H)=\bigoplus_{j\geq0}H^{0}(X,\mathcal{O}_{X}(jH)),\qquad R(X,H)=\widehat{T(X,H)_{T(X,H)_{>0}}}.

Here the hat denotes completion at the maximal ideal. We write KXK_{X} for the canonical divisor and c2(X)c_{2}(X) for the second Chern number. For the specific pair (X,H)(X,H) used for Theorem 1.1, the uncompleted vertex localization T(X,H)T(X,H)>0T(X,H)_{T(X,H)>0} also has no small Cohen–Macaulay module (Corollary 7.1). It is a three-dimensional local domain essentially of finite type over C\mathbb{C}, so this supplies a counterexample in the affine-local class just described. The deduction uses faithful flatness of completion. Our main intermediate result is the following obstruction.

Theorem 1.2. Let XX be a smooth integral projective surface over C\mathbb{C}, and let HH be an ample globally generated divisor. If R(X,H)R(X,H) admits a small Cohen–Macaulay module, then

KX2−2c2(X)≤6H2.(1)K_X^2 - 2c_2(X) \le6H^2. \tag*{(1)}

A numerical obstruction for Ulrich sheaves on surfaces follows from Bogomolov’s inequality and the Hodge index theorem. Beauville presented this route for surfaces of Néron–Severi rank one [4 pp. 18–20]; Anghel gives the surface inequality without that Néron–Severi rank restriction [3 Proposition 2.1 and Remark 2.3]. For a projective surface, an Ulrich sheaf has trivial vector-bundle pushforward under a finite linear projection to P2\mathbb{P}^2 [10 Proposition 2.1]. The argument here replaces that triviality condition by a comparison of coherent extensions across the exceptional divisor of a completed cone. A module over the completed local ring need not carry a grading. We therefore retain its actual vector bundle on the punctured spectrum, rather than assume it descends to an Ulrich sheaf on XX. It is this comparison that permits arbitrary nongraded modules. The explicit surface used below is the complete-quadrangle Hirzebruch–Kummer construction appearing in [3 Section 7.2]. We give its construction and all the required numerical and positivity checks directly.

There are two further recent comparisons. Anghel’s September 14, 2026 preprint excludes arithmetically Cohen–Macaulay bundles under the sharper inequality KX2−2c2(X)>3H2K_X^2 - 2c_2(X) > 3H^2, and excludes graded maximal Cohen–Macaulay modules over the same complete-quadrangle section rings used here [2 Theorems 1.1 and 1.3, Corollary 4.3]. That statement does not assert nonexistence of arbitrary modules over the vertex completion. Chen’s September 21, 2026 preprint proposes a quantitative local-cohomology bound for every nonzero finite reflexive module over a completed cone when KX≡4HK_X \equiv4H and 15H2/8−χ(OX)>015H^2/8 - \chi(\mathcal{O}_X) > 0, and obtains a proposed completed-local counterexample from the exponent-six cover [8 Main claim 1.1 and Consequence 1.2]. Thus the existential completed-module claim overlaps with that work. Our obstruction applies without the proportionality hypothesis on KXK_X and HH; the exponent-seven polarization used below has KX≡5HK_X\equiv5H.1

Proof overview

Choose a finite morphism f:X→P2f:X \to\mathbb{P}^2 with f∗OP2(1)=OX(H)f^*\mathcal{O}_{\mathbb{P}^2}(1) = \mathcal{O}_X(H), and put h=H2=deg⁡fh = H^2 = \deg f. The completed cone is finite over the regular local ring A=C[[x0,x1,x2]]A = \mathbb{C}[[x_0,x_1,x_2]]. A small Cohen–Macaulay module is free over AA. Its restriction to the punctured cone is a vector bundle, which extends to a coherent sheaf E\mathcal{E} on a regular model W→Spec⁡R(X,H)W \to\operatorname{Spec} R(X,H) that is an isomorphism away from the exceptional surface XX.

For a nonzero torsion-free sheaf BB on XX, its normalized slope is

μ(B)=c1(B)⋅Hrk⁡(B)H2.\mu(B) = \frac{c_1(B)\cdot H}{\operatorname{rk}(B)H^2}.

It is slope-semistable if no nonzero subsheaf of smaller rank has larger slope. Its Harder–Narasimhan filtration has torsion-free semistable factors with decreasing slopes. We choose E\mathcal{E} with torsion-free restriction P=E∣XP=\mathcal{E}|_{X}. First, elementary modifications of E\mathcal{E} along XX arrange that the slopes of the Harder–Narasimhan factors of PP lie in an interval of length at most one. These modifications preserve the bundle away from XX. Second, that bundle comes from an AA-free module. Comparing the pushforward of E\mathcal{E} with a trivial sheaf on the blowup of Spec⁡A\operatorname{Spec} A gives, for F=f∗PF=f_{*}P of rank NN, the inequality 2Nch⁡2(F)≥c1(F)22N\operatorname{ch}_{2}(F)\geq c_{1}(F)^{2}, with Chern characters evaluated on P2\mathbb{P}^{2}. Bogomolov’s inequality for the semistable factors bounds the same normalized Chern expression above. More precisely, if r=rk⁡Pr=\operatorname{rk}P, then N=rhN=rh and the two comparisons give

0≤2ch⁡2(f∗P)N−(c1(f∗P)N)2≤12−KX2−2c2(X)12H2.0\leq\frac{2\operatorname{ch}_{2}(f_{*}P)}{N}-\left(\frac{c_{1}(f_{*}P)}{N}\right)^{2}\leq\frac{1}{2}-\frac{K_{X}^{2}-2c_{2}(X)}{12H^{2}}.

This yields (1). Neither PP nor f∗Pf_{*}P is assumed semistable or Ulrich.

Section 2 constructs the resolution and extends an arbitrary module. Section 3 proves the slope modification lemma. Section 4 establishes the Chern-character inequality, and Section 5 combines it with surface theory. Finally, Section 6 constructs a surface with KX2−2c2(X)>6H2K_{X}^{2}-2c_{2}(X)>6H^{2}.

From a completed cone to its exceptional surface

Throughout this section, XX and HH satisfy the hypotheses of Theorem 1.2. We construct a regular model of the completed cone and explain how a small Cohen–Macaulay module produces a sheaf on it. We also prove that the completed ring is normal.

The finite map and the completed ring

Global generation supplies sections s0,s1,s2s_{0},s_{1},s_{2} of OX(H)\mathcal{O}_{X}(H) with no common zero. Indeed, choose s0≠0s_{0}\ne0, then s1s_{1} avoiding the components of its zero divisor, and then s2s_{2} nonvanishing at their finitely many common zeros. The resulting morphism

f:X⟶D:=P2,f∗OD(1)=OX(H),f:X\longrightarrow D:=\mathbb{P}^{2},\qquad f^{*}\mathcal{O}_{D}(1)=\mathcal{O}_{X}(H),

is finite and surjective. A positive-dimensional fiber would contain a curve on which HH is both ample and trivial, which is impossible. Properness then gives finiteness, and the intersection formula gives deg⁡f=h:=H2\deg f=h:=H^{2}.

Put

S=C[x0,x1,x2],T=T(X,H),A=C[[x0,x1,x2]],R=T⊗SA.S=\mathbb{C}[x_{0},x_{1},x_{2}],\qquad T=T(X,H),\qquad A=\mathbb{C}[[x_{0},x_{1},x_{2}]],\qquad R=T\otimes_{S}A.

where xix_{i} acts as sis_{i}. Surjectivity of ff makes S→TS\to T injective. The ring TT is a domain, as seen by trivializing HH rationally and embedding TT in C(X)[t]\mathbb{C}(X)[t]. It is a finite graded SS-module: by the projection formula, its graded pieces are the nonnegative twisted sections of the coherent sheaf f∗OXf_{*}\mathcal{O}_{X} on DD, and Serre vanishing implies finite generation. Explicitly, a surjection OD(−a)p→f∗OX\mathcal{O}_{D}(-a)^{p}\to f_{*}\mathcal{O}_{X} and the vanishing of the first cohomology of its kernel in large twists generate all sufficiently large degrees; the remaining finitely many degrees complete a generating set.

Let n=T>0\mathfrak{n}=T_{>0} and a=(x0,x1,x2)T\mathfrak{a}=(x_{0},x_{1},x_{2})T. The a\mathfrak{a}-adic filtration is cofinal with the degree filtration: if homogeneous SS-module generators have degrees at most a′a', then

aj⊆T≥j,T≥j+a′⊆aj.\mathfrak{a}^{j}\subseteq T_{\geq j},\qquad T_{\geq j+a'}\subseteq\mathfrak{a}^{j}.

Finiteness over SS and completion of finite modules [19 Tags 00MA and 00MB] therefore identify

R=T^ a=∏j≥0Tj=Tn^,(2)R=\widehat{T}^{\,\mathfrak a}=\prod_{j\geq0}T_j=\widehat{T_{\mathfrak n}}, \tag*{(2)}

with multiplication by the graded Cauchy product. The product of the first nonzero homogeneous terms proves that RR is a domain. An element with nonzero constant term is a unit. Moreover, T/aT/\mathfrak a is a finite-dimensional positively graded algebra with degree-zero part C\mathbb{C}, so a=n\sqrt{\mathfrak a}=\mathfrak n. This justifies the last equality in (2) and shows that RR is a complete Noetherian local ring with residue field C\mathbb{C}. Flatness of S→AS\to A gives an injection A→RA\to R; this finite integral extension has dimension three.

The regular model and normality

Let V0=Tot⁡(OD(−1))V_0=\operatorname{Tot}(\mathcal{O}_D(-1)), the blowup of Spec⁡S\operatorname{Spec}S at the origin, and let

W0=X×DV0=Tot⁡(OX(−H)).W_0=X\times_D V_0=\operatorname{Tot}(\mathcal{O}_X(-H)).

The map W0→V0W_0\to V_0 is finite. Evaluation of homogeneous sections gives W0→Spec⁡TW_0\to\operatorname{Spec}T. It is an isomorphism off the zero section. To see this without any assumption that TT is generated in degree one, set Xi={si≠0}X_i=\{s_i\ne0\}. The set XiX_i is affine, and clearing poles along the zero divisor of sis_i gives

T[1/si]=Γ(Xi,OXi)[si,si−1].T[1/s_i]=\Gamma(X_i,\mathcal{O}_{X_i})[s_i,s_i^{-1}].

Both the punctured line bundle over XiX_i and the corresponding open subset of Spec⁡T\operatorname{Spec}T have this coordinate ring.

Base change along S→AS\to A to obtain g:W→Vg:W\to V and p:W→Spec⁡Rp:W\to\operatorname{Spec}R. The zero sections remain X⊂WX\subset W and D⊂VD\subset V. Their ideal sheaves J\mathcal{J} and I\mathcal{I} satisfy

J=g∗I,I∣D=OD(1),J∣X=OX(H).(3)\mathcal{J}=g^*\mathcal{I},\qquad\mathcal{I}|_D=\mathcal{O}_D(1),\qquad\mathcal{J}|_X=\mathcal{O}_X(H). \tag*{(3)}

The zero sections are Cartier divisors by flatness. The relevant maps fit into the diagram

X→W→pSpec⁡Rf↓g↓↓D→V→Spec⁡A\begin{CD} X @>>> W @>{p}>> \operatorname{Spec}R \\ @V{f}VV @V{g}VV @VVV \\ D @>>> V @>>> \operatorname{Spec}A \end{CD}

The maps gg and Spec⁡R→Spec⁡A\operatorname{Spec}R\to\operatorname{Spec}A are finite, and VV and WW are proper over Spec⁡A\operatorname{Spec}A. Since Spec⁡R\operatorname{Spec}R is separated over Spec⁡A\operatorname{Spec}A, the map pp is proper as well. The maps pp and V→Spec⁡AV\to\operatorname{Spec}A identify W∖XW\setminus X and V∖DV\setminus D with the punctured spectra of RR and AA, respectively.

Both VV and WW are regular. At a point of the closed fiber, a local Cartier equation is a nonzerodivisor and its quotient local ring is regular, since DD and XX are smooth. Lifting generators of the quotient maximal ideal shows that the embedding dimension is at most one more than the quotient dimension; the nonzerodivisor raises dimension by one. The local ring is therefore regular. Properness over the local base ensures that the closure of every point meets the closed fiber. Regularity at all other points follows by localization. A regular Noetherian scheme has disjoint irreducible components; here every component meets the integral closed fiber, so each of VV and WW is integral.

This model also proves normality of RR. The line bundle projection W0→XW_0 \to X identifies its global functions with TT. Flat base change [19 Tag 02KH] then gives

Γ(W,OW)=Γ(W0,OW0)⊗SA=R.(4)\Gamma(W,\mathcal{O}_W)=\Gamma(W_0,\mathcal{O}_{W_0})\otimes_S A=R. \tag*{(4)}

For this equality, take a finite affine cover of the separated scheme W0W_0 and tensor the kernel computing global sections with the flat SS-algebra AA. The global functions on an integral normal scheme form an integrally closed domain: a fraction integral over the global ring is integral over every local ring, hence regular everywhere. Since WW is regular and integral, (4) proves that RR is normal.

Extending an arbitrary module

Suppose that MM is a small Cohen–Macaulay RR-module. The parameters x0,x1,x2x_0,x_1,x_2 form an MM-regular sequence, so MM has depth three over AA. The Auslander–Buchsbaum formula over the regular local ring AA [19 Tags 00N6, 00O7 and 090V] shows that

M≃ANas an A-module,N>0.(5)M \simeq A^N \quad\text{as an } A\text{-module}, \qquad N>0. \tag*{(5)}

The localization R⊗AFrac⁡(A)R\otimes_A\operatorname{Frac}(A) is a finite-dimensional domain over a field, hence a field. Thus MM has positive generic rank over RR and full support.

The restriction of MM to the punctured spectrum is locally free. Indeed, Cohen–Macaulay modules of full support localize to Cohen–Macaulay modules. More explicitly, for a prime q\mathfrak{q} the depth localization inequality [19 Tag 0FCC] gives

depth⁡RqMq≥3−dim⁡(R/q)≥dim⁡Rq.\operatorname{depth}_{R_{\mathfrak{q}}}M_{\mathfrak{q}} \ge3-\dim(R/\mathfrak{q}) \ge\dim R_{\mathfrak{q}}.

The second inequality follows by concatenating chains of primes below and above q\mathfrak{q}. On the puncture RqR_{\mathfrak{q}} is regular by the model WW, and Auslander–Buchsbaum gives freeness.

On WW, take

E=(p∗M~)∗∗,P=E∣X,\mathcal{E}=(p^*\widetilde{M})^{**}, \qquad P=\mathcal{E}|_X,

where duals are over OW\mathcal{O}_W. Then E\mathcal{E} is coherent and torsion-free, and agrees with MM away from XX. The restriction PP is torsion-free of positive rank on XX. Here is the local justification. On a regular integral scheme a coherent dual has depth at least min⁡(2,dim⁡OW,w)\min(2,\dim\mathcal{O}_{W,w}): dualizing a finite presentation expresses it as a kernel of a map between finite free modules, whose image is torsion-free, and the depth lemma applies. At any nongeneric point of XX, quotienting E\mathcal{E} by the Cartier equation therefore leaves depth at least one. Thus PP has no associated points except the generic point of XX. At that generic point E\mathcal{E} is free over a discrete valuation ring, so PP has the same positive rank as E\mathcal{E}.

Finally, g∗Eg_*\mathcal{E} is torsion-free and, by (5), agrees with OVN\mathcal{O}_V^N off DD. We have obtained the precise input for the next two sections: a sheaf that can be modified along XX, whose pushforward remains trivial away from DD.

Balancing slopes by modifications

We now modify the extension along XX so that its restriction has Harder–Narasimhan slopes in an interval of length at most one. Recall the normalization μ(B)=c1(B)⋅H/(rk⁡(B)h)\mu(B)=c_1(B)\cdot H/(\operatorname{rk}(B)h), where h=H2h=H^2. Tensoring with OX(H)\mathcal{O}_X(H) increases slope by one.

Elementary modification to constrain the slopes of exceptional-divisor restrictions has an antecedent in Chen–Sun’s optimal extensions of reflexive analytic sheaves [9] (Theorem 1.4(I), Proposition 2.6 and Remark 1.5). Their analytic setting and reflexivity hypotheses differ from the Noetherian Cartier-divisor setting below; we give the needed algebraic modification and termination argument.

We recall briefly why the filtration exists in this setting. A torsion-free sheaf embeds in a sum of sufficiently positive line bundles, so its subsheaf slopes are bounded above by taking determinants. Their denominators are bounded because their ranks are bounded. Choose a subsheaf of maximum slope and, among these, of maximum rank; saturate it. Saturation does not decrease degree, since a torsion sheaf has effective divisorial first Chern class. The chosen subsheaf is semistable, and every subsheaf of its torsion-free quotient has strictly smaller slope. Induction on rank gives the filtration. This is the usual construction of the Harder–Narasimhan filtration [18] (Section 1.6, Theorem 1.6.7 and Definition/Corollary 1.6.9).

Lemma 3.1. Let WW be an integral Noetherian scheme, let i:X↪Wi:X\hookrightarrow W be an effective Cartier divisor that is a smooth integral projective surface over C\mathbb{C}, and suppose its ideal J\mathcal{J} restricts to an ample line bundle OX(H)\mathcal{O}_X(H). Let E\mathcal{E} be a coherent torsion-free sheaf on WW such that P=E∣XP=\mathcal{E}|_X is torsion-free of positive rank. There is a coherent torsion-free sheaf E′\mathcal{E}' agreeing with E\mathcal{E} off XX such that P′=E′∣XP'=\mathcal{E}'|_X is torsion-free and the normalized slopes of its Harder–Narasimhan factors lie in an interval of length at most one.

Proof. Suppose the current largest and smallest factor slopes differ by more than one. Let GG be the last semistable quotient of PP, and let B=ker⁡(P→G)B=\ker(P\to G). Replace E\mathcal{E} by E′=ker⁡(E→i∗G)\mathcal{E}'=\ker(\mathcal{E}\to i_*G). This is torsion-free and unchanged off XX. The Cartier resolution gives

0⟶G(H)⟶E′∣X⟶B⟶0.(6)0\longrightarrow G(H)\longrightarrow\mathcal{E}'|_X\longrightarrow B\longrightarrow0. \tag*{(6)}

Indeed, Tor⁡1W(E,OX)=0\operatorname{Tor}^{W}_{1}(\mathcal{E},\mathcal{O}_X)=0 because the Cartier equation acts injectively on E\mathcal{E}, whereas

Tor⁡1W(i∗G,OX)=G⊗J∣X=G(H).\operatorname{Tor}^{W}_{1}(i_*G,\mathcal{O}_X)=G\otimes\mathcal{J}|_X=G(H).

Thus the new restriction remains torsion-free of the same rank rr.

The maximum slope of an extension of torsion-free sheaves is at most the larger of their maximum slopes: intersect a subsheaf with the kernel and take its image in the quotient. Since μ(G)+1<μmax⁡(P)\mu(G)+1<\mu_{\max}(P), (6) shows that the maximum slope does not increase. On the other hand, additivity of rank and first Chern class gives

μ(E′∣X)=μ(P)+rk⁡Gr≥μ(P)+1r.\mu(\mathcal{E}'|_X)=\mu(P)+\frac{\operatorname{rk}G}{r}\geq\mu(P)+\frac{1}{r}.

The average slope is always bounded above by the maximum slope, hence by the initial maximum. Consequently this procedure terminates after finitely many steps. At termination the slope interval has length at most one. □

Apply Lemma 3.1 to the sheaf constructed in Section 2, and again call the resulting sheaves E\mathcal{E} and PP. The rank and the bundle on the puncture are unchanged. In particular, g∗Eg_*\mathcal{E} still agrees with OVN\mathcal{O}_V^N off DD. The restriction PP need not be semistable: it is the width of its factor slopes, rather than semistability, that we have arranged. This agreement on the puncture imposes the lower bound proved next.

A Chern-character inequality from a divisor

The pushforward of our extension agrees with a trivial bundle away from DD. We show that this agreement forces a nonnegative expression in the Chern characters of its restriction to DD. For a coherent sheaf FF of positive rank NN on D=P2D = \mathbb{P}^{2}, identify c1(F)c_{1}(F) with its integer coefficient in the line class, and let ch⁡2(F)\operatorname{ch}_{2}(F) mean the degree of the second Chern character. Define

v(F)=2ch⁡2(F)N−(c1(F)N)2v(F) = \frac{2\operatorname{ch}_{2}(F)}{N} - \left(\frac{c_{1}(F)}{N}\right)^{2}

For F=⨁i=1NOD(ai)F = \bigoplus_{i=1}^{N}\mathcal{O}_{D}(a_{i}), this is the ordinary variance of the integers aia_{i}. It need not be nonnegative for arbitrary sheaves. The expression is unchanged by twisting FF by OD(b)\mathcal{O}_{D}(b).

Lemma 4.1. Let VV be an integral Noetherian scheme with an effective Cartier divisor D≃P2D \simeq\mathbb{P}^{2}, whose ideal I\mathcal{I} satisfies I∣D≃OD(1)\mathcal{I}|_{D} \simeq\mathcal{O}_{D}(1). Suppose L′L' is a coherent torsion-free sheaf on VV and L′∣V∖D≃OV∖DNL'|_{V\setminus D} \simeq\mathcal{O}_{V\setminus D}^{N}, where N>0N > 0. Then F=L′∣DF = L'|_{D} has rank NN and v(F)≥0v(F) \geq0.

Proof. At the generic point of DD, the local ring of VV has maximal ideal generated by the Cartier equation, so it is a discrete valuation ring. The torsion-free sheaf L′L' is free of rank NN there; hence FF has rank NN and v(F)v(F) is defined. Write L=OVNL = \mathcal{O}_{V}^{N} and identify LL and L′L' after inverting the Cartier equation. Clearing denominators on a finite affine cover gives

IbL⊆L′⊆I−bL\mathcal{I}^{b}L \subseteq L' \subseteq\mathcal{I}^{-b}L

for some b≥0b \geq0. Replacing L′L' by L′⊗IbL' \otimes\mathcal{I}^{b} replaces FF by F(b)F(b) and does not change v(F)v(F). We may therefore assume I2bL⊆L′⊆L\mathcal{I}^{2b}L \subseteq L' \subseteq L.

The quotient Q=L/L′Q = L/L' is killed by a power of I\mathcal{I}. Its finite filtration has factors

Qj=IjQ/Ij+1Q(j≥0)Q_{j} = \mathcal{I}^{j}Q/\mathcal{I}^{j+1}Q \qquad(j \geq0)

on DD. Each QjQ_{j} is a quotient of OD(j)N\mathcal{O}_{D}(j)^{N}. If nj=rk⁡Qjn_{j} = \operatorname{rk} Q_{j}, then

0≤nj≤N,c1(Qj)≥jnj.(7)0 \leq n_{j} \leq N,\qquad c_{1}(Q_{j}) \geq jn_{j}. \tag*{(7)}

For the second assertion, Qj(−j)Q_{j}(-j) is globally generated. The determinant of its torsion-free quotient has a nonzero section, and its torsion contributes a nonnegative divisorial first Chern class.

Derived restriction to DD gives the following identity in the Grothendieck group of coherent sheaves on DD:

[F]=[ODN]+∑j≥0([Qj(1)]−[Qj]).(8)[F] = [\mathcal{O}_{D}^{N}] + \sum_{j\geq0} \bigl([Q_{j}(1)] - [Q_{j}]\bigr). \tag*{(8)}

Indeed, restriction minus first Tor is additive, with no higher Tor against a Cartier divisor. The first Tor vanishes for LL, L′L', while for a sheaf QjQ_{j} supported on DD it equals Qj(1)Q_{j}(1). Taking Chern characters in (8) gives rank NN and

c1(F)=∑jnj,2ch⁡2(F)=∑j(2c1(Qj)+nj)≥∑j(2j+1)nj.c_{1}(F) = \sum_{j} n_{j},\qquad2\operatorname{ch}_{2}(F) = \sum_{j}\bigl(2c_{1}(Q_{j}) + n_{j}\bigr) \geq\sum_{j}(2j+1)n_{j}.

The remaining assertion is the elementary inequality

N∑j(2j+1)nj−(∑jnj)2=∑jnj(N−nj)+2∑j>ℓnj(N−nℓ)≥0,N\sum_{j}(2j+1)n_{j} - \left(\sum_{j}n_{j}\right)^{2} = \sum_{j}n_{j}(N-n_{j}) + 2\sum_{j>\ell}n_{j}(N-n_{\ell}) \geq0,

which follows from (7). Dividing by N2N^{2} proves the lemma.

For L′=g∗EL' = g_* \mathcal{E}, the equality J=g∗I\mathcal{J} = g^* \mathcal{I} in (3) and finite pushforward give

L′∣D=f∗P.L'|_D = f_*P.

Writing r=rk⁡Pr = \operatorname{rk} P, this sheaf has rank rhrh by finite pushforward along ff, and rank NN by generic restriction along DD in Lemma 4.1. Thus the two ranks agree: N=rhN = rh. The hypotheses of Lemma 4.1 hold by (3) and the construction of E\mathcal{E}. Consequently

v(f∗P)≥0.(9)v(f_*P) \ge0. \tag*{(9)}

It remains to bound this expression above using the narrow slope interval of PP.

The numerical obstruction

We first compute the contribution of one semistable factor. We use surface Riemann–Roch and the Hodge index theorem in their standard forms; see [11], Appendix A, Theorem 4.1 and Example 4.1.2; Chapter V, Theorem 1.9. On each smooth projective surface, finite locally free resolutions extend Riemann–Roch additively to coherent sheaves [11], Appendix A, Section 5. This covers the possibly non-locally-free sheaves GG and f∗Gf_*G below; we compare their Euler characteristics, without requiring the finite morphism ff to be smooth. All intersections in this section are on XX, except for Chern characters explicitly taken on D=P2D = \mathbb{P}^2.

Lemma 5.1. Let f:X→P2f : X \to\mathbb{P}^2 be finite, with XX a smooth integral projective complex surface and H=f∗OP2(1)H = f^*\mathcal{O}_{\mathbb{P}^2}(1) ample. Put h=H2h = H^2. For every nonzero torsion-free HH-slope-semistable sheaf GG on XX,

v(f∗G)≤14−KX2−2c2(X)12h.v(f_*G) \le\frac{1}{4} - \frac{K_X^2 - 2c_2(X)}{12h}.

Moreover,

c1(f∗G)rk⁡(f∗G)=μ(G)−KX⋅H2h−32.\frac{c_1(f_*G)}{\operatorname{rk}(f_*G)} = \mu(G) - \frac{K_X \cdot H}{2h} - \frac{3}{2}.

Proof. Write u=rk⁡Gu = \operatorname{rk} G, k=KX⋅Hk = K_X \cdot H, and

m=c1(f∗G)uh,q=ch⁡2(f∗G)uh.m = \frac{c_1(f_*G)}{uh}, \qquad q = \frac{\operatorname{ch}_2(f_*G)}{uh}.

The rank of f∗Gf_*G is uhuh. Finite pushforward and the projection formula identify the twist Euler characteristics on XX and P2\mathbb{P}^2. Surface Riemann–Roch, divided by uhuh, therefore gives

t22+(32+m)t+1+32m+q=t22+(μ(G)−k2h)t+χ(OX)h+ch⁡2(G)uh−c1(G)⋅KX2uh.\frac{t^2}{2} + \left(\frac{3}{2} + m\right)t + 1 + \frac{3}{2}m + q = \frac{t^2}{2} + \left(\mu(G) - \frac{k}{2h}\right)t + \frac{\chi(\mathcal{O}_X)}{h} + \frac{\operatorname{ch}_2(G)}{uh} - \frac{c_1(G) \cdot K_X}{2uh}.

Comparison of the linear coefficients proves the formula for mm. Comparison of constants, followed by completing the square, yields

2q−m2=14+2χ(OX)h+2ch⁡2(G)/u−c1(G)⋅KX/uh−(μ(G)−k2h)2.(10)2q - m^2 = \frac{1}{4} + \frac{2\chi(\mathcal{O}_X)}{h} + \frac{2\operatorname{ch}_2(G)/u - c_1(G) \cdot K_X/u}{h} - \left(\mu(G) - \frac{k}{2h}\right)^2. \tag*{(10)}

Bogomolov’s inequality for a torsion-free slope-semistable sheaf on a smooth projective complex surface states that

2uch⁡2(G)≤c1(G)22u\operatorname{ch}_2(G) \le c_1(G)^2

[18], Theorem 3.4.1. Its hypotheses apply to GG and the ample polarization HH; normalizing slope by hh does not change semistability. Set C=c1(G)/u−KX/2C=c_1(G)/u-K_X/2. The last two terms of (10) are at most

C2−KX2/4h−(C⋅H)2h2≤−KX24h,\frac{C^2-K_X^2/4}{h}-\frac{(C\mathbin{\cdot}H)^2}{h^2}\leq-\frac{K_X^2}{4h},

by the Hodge index theorem for the rational divisor class CC. Using Noether’s formula 12χ(OX)=KX2+c2(X)12\chi(\mathcal{O}_X)=K_X^2+c_2(X) now gives the result.

Proof of Theorem 1.2. Starting with a small Cohen–Macaulay module, Sections 2 and 3 construct a torsion-free sheaf PP on XX whose semistable factors GiG_i have normalized slopes in an interval of length at most one. Write ui=rk⁡Giu_i=\operatorname{rk}G_i, r=rk⁡Pr=\operatorname{rk}P, and wi=ui/rw_i=u_i/r. Thus wi>0w_i>0 and ∑iwi=1\sum_i w_i=1. Finite pushforward is exact, so F=f∗PF=f_*P has rank rh=Nrh=N, and its normalized Chern characters are the weighted averages of those of f∗Gif_*G_i. If mi=c1(f∗Gi)/(uih)m_i=c_1(f_*G_i)/(u_i h) and m‾=∑iwimi\overline m=\sum_iw_i m_i, then

v(F)=∑iwiv(f∗Gi)+∑iwi(mi−m‾)2.(11)v(F)=\sum_i w_i v(f_*G_i)+\sum_i w_i(m_i-\overline m)^2. \tag*{(11)}

Lemma 5.1 shows that the mim_i differ from μ(Gi)\mu(G_i) by a common constant. They therefore lie in an interval of length at most one. Their weighted variance is at most 1/41/4: the mean minimizes the mean square distance, and every point is within 1/21/2 of the interval midpoint. Combining (9), (11), and Lemma 5.1 gives

0≤v(F)≤12−KX2−2c2(X)12h.0\leq v(F)\leq\frac{1}{2}-\frac{K_X^2-2c_2(X)}{12h}.

This is precisely (1).

An explicit surface and its polarization

We construct a family of smooth projective surfaces from which we will choose one violating the inequality established above. The construction uses the six lines joining four points of the plane, followed by a cover obtained by taking roots of their ratios. The same line arrangement provides the sections needed for an ample globally generated divisor on the cover. This choice of arrangement and polarization is the complete-quadrangle construction in [3], Section 7.2; all properties needed here are verified below.

Proposition 6.1. For each prime n≥3n\geq3, there is a smooth integral projective complex surface XX with an ample globally generated divisor HH such that

H2=5n3,KX2−2c2(X)=n3(n2−10).H^2=5n^3,\qquad K_X^2-2c_2(X)=n^3(n^2-10).

We prove the proposition by constructing the cover, checking its local coordinates, and then finding sections of the required polarization. The covering construction goes back to Hirzebruch’s work on line arrangements [12].

The branch divisor

Choose four points p1,…,p4∈PC2p_1,\ldots,p_4\in\mathbb{P}^2_{\mathbb{C}}, no three collinear, and let BB be their blowup. Write ℓ\ell for the pullback of the line class and e1,…,e4e_1,\ldots,e_4 for the exceptional curves. Thus

ℓ2=1,ℓ⋅ei=0,ei⋅ej=−δij,KB=−3ℓ+∑i=14ei.\ell^2=1,\qquad\ell\mathbin{\cdot}e_i=0,\qquad e_i\mathbin{\cdot}e_j=-\delta_{ij},\qquad K_B=-3\ell+\sum_{i=1}^{4}e_i.

Number the six joining lines from 1 to 6, and choose defining linear forms l1,…,l6l_1,\ldots,l_6. Denote the strict transform of line aa by LaL_a and its total transform by TaT_a. If its endpoints are pi,pjp_i,p_j, then

Ta=La+ei+ej∼ℓ.T_a = L_a + e_i + e_j \sim\ell.

Two joining lines are called opposite if their endpoint pairs are disjoint. There are three opposite pairs.

Let

Δ=∑a=16La+∑i=14ei.\Delta= \sum_{a=1}^{6} L_a + \sum_{i=1}^{4} e_i.

This reduced divisor has simple normal crossings. Each exceptional curve meets the three incident strict transforms in three distinct points. Strict transforms of lines with a common endpoint are disjoint, whereas each opposite pair meets transversely in one point away from the exceptional curves. The three latter points are distinct: an equality between two would force two lines with a common endpoint to meet away from that endpoint. Consequently Δ\Delta has ten rational components, each meeting the others in three points, and fifteen nodes in total. Figure 1 records these incidences; in the figure LijL_{ij} denotes the strict transform of the line joining pip_i and pjp_j. Since each point belongs to three joining lines,

The branch-incidence graph on the blown-up plane

Figure 1. The branch-incidence graph on the blown-up plane. Vertices represent the ten rational components, not points of the surface; edges represent its fifteen branch nodes. Twelve edges join eie_i to an incident LijL_{ij}, and three join opposite line transforms. Every vertex has degree three. Crossings of drawn graph edges do not represent additional intersections.

Δ∼6ℓ−2∑i=14ei=−2KB,KB2=5.(12)\Delta\sim6\ell- 2\sum_{i=1}^{4} e_i = -2K_B,\qquad K_B^2 = 5. \tag*{(12)}

The cover and its local coordinates

Fix a prime n≥3n \ge3, to be chosen at the end. In an algebraic closure of K=C(B)K = \mathbb{C}(B) choose elements yay_a satisfying

yan=la/l6(1≤a≤5),y_a^n = l_a/l_6 \qquad(1 \le a \le5),

and let π:X→B\pi: X \to B be the normalization of BB in K′=K(y1,…,y5)K' = K(y_1,\ldots,y_5). The base BB is normal and Noetherian, and this finite characteristic-zero field extension is separable. Its normalization is therefore finite [19 Tag 032L], so XX is an integral projective surface.

Lemma 6.2. The surface XX is smooth and deg⁡π=d=n5\deg\pi= d = n^5. The map is unramified outside Δ\Delta and has ramification index nn along every component of Δ\Delta. More precisely, near a point at which exactly s∈{0,1,2}s \in\{0,1,2\} components of Δ\Delta meet, it is, after an étale base change, the disjoint union of n5−sn^{5-s} copies of

(t1,t2)⟼{(t1,t2),s=0,(t1n,t2),s=1,(t1n,t2n),s=2.(t_1,t_2) \longmapsto \begin{cases} (t_1,t_2), & s=0,\\ (t_1^n,t_2), & s=1,\\ (t_1^n,t_2^n), & s=2. \end{cases}

Here the displayed maps are understood in étale coordinates on the base and the source.

Proof. First, suppose that integers c1,…,c5c_1,\ldots,c_5 satisfy

∏a=15(la/l6)ca∈K∗n.\prod_{a=1}^{5}(l_a/l_6)^{c_a} \in K^{*n}.

Taking valuations along LaL_a shows that nn divides cac_a for every aa. Thus the five radicands are independent in K∗/K∗nK^*/K^{*n}. Since KK contains the nnth roots of unity, the extension K′/KK'/K is Galois and its group embeds in (Z/n)5(\mathbb{Z}/n)^5 by its action on the roots. If its image were a proper subgroup, some nonzero linear functional on (Z/n)5(\mathbb{Z}/n)^5 would annihilate it. The corresponding product of the yay_a would then lie in KK, contradicting the independence just proved. This gives [K′:K]=n5[K':K]=n^5.

We next record the valuations modulo nn along the branch curves. Use the vector space

Un=Fn6/⟨(1,1,1,1,1,1)⟩U_n = \mathbb{F}_n^6/\langle(1,1,1,1,1,1)\rangle

and denote the images of its six standard vectors by ε1,…,ε6\varepsilon_1,\ldots,\varepsilon_6. Identifying ε1,…,ε5\varepsilon_1,\ldots,\varepsilon_5 with the standard basis of Fn5\mathbb{F}_n^5, the valuation vector along LaL_a is εa\varepsilon_a; in particular ε6=−∑a=15εa\varepsilon_6=-\sum_{a=1}^5\varepsilon_a. If IiI_i is the set of the three lines incident with pip_i, the valuation vector along eie_i is

wi=∑a∈Iiεa.w_i=\sum_{a\in I_i}\varepsilon_a.

All these vectors are nonzero. At a node formed by two strict transforms, the vectors are two distinct εa\varepsilon_a, hence are independent. At a node La∩eiL_a\cap e_i, one has a∈Iia\in I_i, and εa,wi\varepsilon_a,w_i are independent as well. Indeed, if λεa+μwi=0\lambda\varepsilon_a+\mu w_i=0 in UnU_n, a coordinate outside IiI_i forces the corresponding constant vector to be zero. A coordinate in Ii∖{a}I_i\setminus\{a\} then gives μ=0\mu=0, followed by λ=0\lambda=0.

It remains to check that this independence gives the asserted smooth normalization, including at the nodes. At a closed point of BB, choose regular parameters u1,u2u_1,u_2 such that the local branches of Δ\Delta are u1=0,…,us=0u_1=0,\ldots,u_s=0. Shrink to an affine neighborhood on which these parameters define an étale map to A2\mathbb{A}^2. Each radicand is a monomial in u1,…,usu_1,\ldots,u_s, with possibly negative exponents, times a unit. Adjoining nnth roots of these units is an étale base change. On an integral neighborhood of a point of that base change, divide the yay_a by the resulting roots of the units.

The ss columns of exponents modulo nn are independent by the preceding calculation. Multiplicative changes of the five root generators by a matrix in GL⁡5(Fn)\operatorname{GL}_5(\mathbb{F}_n) therefore reduce their equations to

Yjn=uj(1≤j≤s),Yjn=1(s<j≤5).Y_j^n=u_j\quad(1\leq j\leq s),\qquad Y_j^n=1\quad(s<j\leq5).

To make this reduction in the function field, lift the matrix entries to integers and divide by suitable monomials in the uju_j to remove exponents divisible by nn. The generator change is reversible because its matrix is invertible modulo nn. It consequently describes the whole generic algebra after base change, not merely a subextension. The first ss equations generate a field of degree nsn^s, by the valuations along uj=0u_j = 0, and the remaining equations split completely. The generic algebra is therefore a product of n5−sn^{5-s} copies of that field.

On this neighborhood, with coordinate ring CC, the normalization in each factor is

C[t1,…,ts]/(t1n−u1,…,tsn−us),C[t_1,\ldots,t_s]/(t_1^n-u_1,\ldots,t_s^n-u_s),

so the full normalization is a product of n5−sn^{5-s} such algebras. Indeed, this algebra is finite and free over CC, has the indicated generic field, and is smooth: it is the base change of the étale coordinate map to A2\mathbb{A}^2 along the corresponding coordinate power map. In particular it is normal, so it is the integral closure of CC in that field. Normalization commutes with this étale base change: the pullback of XX is normal and has the same generic algebra, and is pure of dimension two, since it is étale over the normal integral surface XX. Each component therefore dominates the new integral base: its finite image is closed and has dimension two. This proves the local description and hence the lemma. □

The Chern numbers

The local coordinates in Lemma 6.2 give both the canonical divisor formula and the degrees of the coverings of the branch strata. Thus no further description of XX is needed to compute its Chern numbers. These calculations recover Hirzebruch’s formulas [12] (Section 2.2, (5) and (7)) for this six-line arrangement.

The Jacobian of the coordinate power map gives

KX∼Qπ∗(KB+(1−1/n)Δ)∼Q(1−2/n)π∗(−KB).K_X \sim_{\mathbb{Q}} \pi^*\left(K_B+(1-1/n)\Delta\right) \sim_{\mathbb{Q}} (1-2/n)\pi^*(-K_B).

Using (12) and the degree of π\pi, we obtain

KX2=5d(1−2/n)2.(13)K_X^2=5d(1-2/n)^2. \tag*{(13)}

For the second Chern number, let ece_c denote compactly supported topological Euler characteristic. The surface BB has Euler characteristic 3+4=73+4=7. Removing the three nodes from each rational component of Δ\Delta gives the one-branch stratum Δ∘\Delta^\circ, whose Euler characteristic is 10(2−3)=−1010(2-3)=-10. The node set has fifteen points. Additivity therefore gives

ec(B∖Δ)=7−(−10)−15=2.e_c(B\setminus\Delta)=7-(-10)-15=2.

Over the stratum with ss branches, the reduced inverse image is a finite étale cover of degree d/nsd/n^s. By additivity and multiplicativity of ece_c, followed by Chern–Gauss–Bonnet [5] (Section 1.3 and Proposition 1.6),

c2(X)=ec(X)=d(2−10n+15n2).(14)c_2(X)=e_c(X)=d\left(2-\frac{10}{n}+\frac{15}{n^2}\right). \tag*{(14)}

Subtracting twice (14) from (13) yields

KX2−2c2(X)=d(1−10n2).(15)K_X^2-2c_2(X)=d\left(1-\frac{10}{n^2}\right). \tag*{(15)}

An ample globally generated divisor

To compare (15) with the obstruction, we seek an integral divisor HH satisfying nH∼π∗(−KB)nH \sim\pi^*(-K_B). Such a divisor has square 5d/n25d/n^2, as required by Proposition 6.1. We first construct HH and sections that generate OX(H)\mathcal{O}_X(H), and then prove ampleness separately. The ramification makes the pullbacks of all branch components divisible by nn as integral divisors. Put

Ra=1nπ∗La,Fi=1nπ∗ei,E=∑i=14Fi,J=1nπ∗T6.R_a=\frac{1}{n}\pi^*L_a,\qquad F_i=\frac{1}{n}\pi^*e_i,\qquad E=\sum_{i=1}^{4}F_i,\qquad J=\frac{1}{n}\pi^*T_6.

These are effective Cartier divisors because XX is smooth. If line aa joins pi,pjp_i,p_j, write

Ja=1nπ∗Ta=Ra+Fi+Fj.J_a=\frac{1}{n}\pi^*T_a=R_a+F_i+F_j.

For 1≤a≤51\leq a\leq5, the root ratios give

div⁡(ya)=Ja−J.\operatorname{div}(y_a)=J_a-J.

Since J6=JJ_6=J, all six JaJ_a are linearly equivalent to JJ. Let zaz_a be the resulting sections of OX(J)\mathcal{O}_X(J) with zero divisors JaJ_a.

These six sections have no common zero. Away from the exceptional curves, no point of BB lies on all six joining lines. On an exceptional curve eie_i, the total transform of any line not incident with pip_i is disjoint from eie_i. Thus the six total transforms have empty common intersection on BB, and the same holds for their reduced pullbacks on XX.

For an opposite pair a,ba,b, its endpoints are the four distinct points, and hence

Ja+Jb=Ra+Rb+E.J_a+J_b=R_a+R_b+E.

Dividing zazbz_a z_b by the canonical section of OX(E)\mathcal{O}_X(E) gives a section of OX(2J−E)\mathcal{O}_X(2J-E) with zero divisor Ra+RbR_a+R_b. These three sections have no common zero. In fact the unions La∪LbL_a\cup L_b belonging to different opposite pairs are disjoint on BB: a line in one pair shares an endpoint with every line in either other pair, so their strict transforms are disjoint. We have therefore proved that both JJ and 2J−E2J-E are globally generated. Their sum

H=3J−E(16)H=3J-E \tag*{(16)}

is globally generated as well.

To check ampleness, first note that 2ℓ−∑iei2\ell-\sum_i e_i is globally generated on BB by the three opposite-pair conics: after removing their exceptional components, their zero divisors are precisely the three disjoint unions just considered. Thus this divisor is nef, as is ℓ\ell. The class ℓ\ell has positive intersection with every nonexceptional curve, and (2ℓ−∑iei)⋅ej=1(2\ell-\sum_i e_i)\cdot e_j=1 for every exceptional curve. Consequently

−KB=ℓ+(2ℓ−∑iei)-K_B=\ell+\left(2\ell-\sum_i e_i\right)

has positive intersection with every integral curve. Its square is 55, so the surface Nakai–Moishezon criterion [11 Chapter V, Theorem 1.10] proves that −KB-K_B is ample. Since

nH∼π∗(3ℓ−∑iei)=π∗(−KB),nH\sim\pi^*\left(3\ell-\sum_i e_i\right)=\pi^*(-K_B),

finite pullback and passage to a positive tensor root [19 Tags 0892 and 01PT] show that HH is ample. The intersection formula now gives

H2=5dn2.(17)H^2 = \frac{5d}{n^2}. \tag*{(17)}

Equations (15) and (17), with d=n5d=n^5, prove Proposition 6.1.

The counterexample

Proof of Theorem 1.1. Take the surface and polarization of Proposition 6.1 with n=7n=7. Their invariants are

H2=5⋅73=1715,KX2−2c2(X)=39⋅73=13377>10290=6H2.H^2 = 5 \cdot7^3 = 1715,\qquad K_X^2 - 2c_2(X) = 39 \cdot7^3 = 13377 > 10290 = 6H^2.

The ring R(X,H)R(X,H) is a complete Noetherian normal local domain of dimension three, with residue field C\mathbb{C}, by Section 2. If it had a nonzero finite module of depth three, Theorem 1.2 would give the opposite numerical inequality. Thus it has no such module.

Corollary 7.1. Let XX and HH be the surface and polarization constructed in Proposition 6.1 for n=7n=7, and put

T=T(X,H),n=T>0.T=T(X,H),\qquad\mathfrak{n}=T_{>0}.

Then TT is a finitely generated graded three-dimensional C\mathbb{C}-domain, n\mathfrak{n} is maximal with T/n=CT/\mathfrak{n}=\mathbb{C}, and the three-dimensional local domain TnT_{\mathfrak{n}} has no nonzero finitely generated maximal Cohen–Macaulay module.

Proof. Section 2 shows that TT is a domain finite over the polynomial subring S=C[x0,x1,x2]S=\mathbb{C}[x_0,x_1,x_2]. Thus TT is a finitely generated C\mathbb{C}-algebra and has dimension three by invariance of dimension under integral extensions [19 Tag 00OK]. Since T0=CT_0=\mathbb{C}, n\mathfrak{n} is maximal, and R0:=TnR_0:=T_{\mathfrak{n}} is a Noetherian local domain essentially of finite type over C\mathbb{C}, with residue field C\mathbb{C}. Equation (2) identifies its maximal-ideal completion with the ring R=R(X,H)R=R(X,H) used above. Completion preserves the dimension of a Noetherian local ring [19 Tag 07NV], so dim⁡R0=dim⁡R=3\dim R_0=\dim R=3.

Suppose that MM is a nonzero finite R0R_0-module of depth three, and choose an MM-regular sequence z1,z2,z3z_1,z_2,z_3 in the maximal ideal of R0R_0. The completion map R0→RR_0\to R is faithfully flat, and, since MM is finite, its completion satisfies M^≅M⊗R0R\widehat{M}\cong M\otimes_{R_0}R [19 Tags 00MC and 00MA]. This is a finite nonzero RR-module. Put M0=MM_0=M and Mi=M/(z1,…,zi)MM_i=M/(z_1,\ldots,z_i)M for 1≤i≤31\leq i\leq3. Nakayama’s lemma gives Mi≠0M_i\ne0 successively. For 0≤i<30\leq i<3, flatness preserves the exact sequence

0⟶Mi→zi+1Mi⟶Mi+1⟶00\longrightarrow M_i \xrightarrow{z_{i+1}} M_i\longrightarrow M_{i+1}\longrightarrow0

after tensoring with RR. The canonical identifications

Mi⊗R0R≅M^/(z1,…,zi)M^(1≤i≤3)M_i\otimes_{R_0}R\cong\widehat{M}/(z_1,\ldots,z_i)\widehat{M}\qquad(1\leq i\leq3)

and faithful flatness show that these completed quotients remain nonzero. The images of the ziz_i lie in the maximal ideal of RR, so they form an M^\widehat{M}-regular sequence of length three. Hence depth⁡RM^≥3\operatorname{depth}_R\widehat{M}\geq3. For a nonzero finite module over a Noetherian local ring, depth is at most the dimension of its support [19 Tag 00LK], which here is at most dim⁡R=3\dim R=3. Thus M^\widehat{M} has depth three, contradicting the completed-ring conclusion just proved for R=R(X,H)R=R(X,H).

References

  1. [1]Yves André. La conjecture du facteur direct. Publications Mathématiques de l’IHÉS, 127:71–93, 2018.arxiv.org/abs/1609.00345
  2. [2]Cristian Anghel. Polarized varieties without arithmetically Cohen–Macaulay bundles and section rings without graded maximal Cohen–Macaulay modules in characteristic zero, 2026. arXiv:2609.15589v1, September 14, 2026.arxiv.org/abs/2609.15589
  3. [3]Cristian Anghel. Smooth counterexamples to the Eisenbud–Schreyer–Weyman Ulrich existence problem, 2026. arXiv:2609.02718v1, September 2, 2026.arxiv.org/abs/2609.02718
  4. [4]Arnaud Beauville. Ulrich bundles: to be or not to be. Lecture slides, Beijing, October 2017. Author’s slides.math.univ-cotedazur.fr/u/beauvill/conf/Ulrich.pdf
  5. [5]Kai Behrend. Donaldson–Thomas type invariants via microlocal geometry. Annals of Mathematics, 170(3):1307–1338, 2009.DOI
  6. [6]Bhargav Bhatt. On the non-existence of small Cohen–Macaulay algebras. Journal of Algebra, 411:1–11, 2014.DOI
  7. [7]Bhargav Bhatt, Mel Hochster, and Linquan Ma. Lim Cohen–Macaulay sequences of modules. Journal für die reine und angewandte Mathematik, 835:111–169, 2026. Numbered references follow arXiv:2410.18372v2, August 25, 2026.arxiv.org/abs/2410.18372
  8. [8]Liang Chen. A local cohomology obstruction to small Cohen–Macaulay modules, 2026. arXiv:2609.24142v1, September 21, 2026.arxiv.org/abs/2609.24142
  9. [9]Xuemiao Chen and Song Sun. Algebraic tangent cones of reflexive sheaves. International Mathematics Research Notices, 2020(24):10042–10063, 2020.
  10. [10]David Eisenbud and Frank-Olaf Schreyer. Resultants and Chow forms via exterior syzygies. Journal of the American Mathematical Society, 16(3):537–579, 2003. With an appendix by Jerzy Weyman.
  11. [11]Robin Hartshorne. Algebraic Geometry, volume 52 of Graduate Texts in Mathematics. Springer-Verlag, New York, 1977.
  12. [12]Friedrich Hirzebruch. Arrangements of lines and algebraic surfaces. In Michael Artin and John Tate, editors, Arithmetic and Geometry, Vol. II, volume 36 of Progress in Mathematics, pages 113–140. Birkhäuser, Boston, 1983.
  13. [13]Melvin Hochster. Cohen–Macaulay modules. In James W. Brewer and Edgar A. Rutter, editors, Conference on Commutative Algebra (Lawrence, 1972), volume 311 of Lecture Notes in Mathematics, pages 120–152. Springer, 1973.
  14. [14]Melvin Hochster. Topics in the Homological Theory of Modules over Commutative Rings, volume 24 of CBMS Regional Conference Series in Mathematics. American Mathematical Society, Providence, RI, 1975.DOI
  15. [15]Melvin Hochster. Homological conjectures and lim Cohen–Macaulay sequences. In Homological and Computational Methods in Commutative Algebra, volume 20 of Springer INdAM Series, pages 173–197. Springer, 2017.DOI
  16. [16]Melvin Hochster and Craig Huneke. Infinite integral extensions and big Cohen–Macaulay algebras. Annals of Mathematics, 135(1):53–89, 1992.DOI
  17. [17]Melvin Hochster and Yongwei Yao. Frobenius splitting, strong F-regularity, and small Cohen–Macaulay modules. Transactions of the American Mathematical Society, 376(9):6729–6765, 2023.DOI
  18. [18]Daniel Huybrechts and Manfred Lehn. The Geometry of Moduli Spaces of Sheaves, volume E31 of Aspects of Mathematics. Friedr. Vieweg & Sohn, Braunschweig, 1997.
  19. [19]The Stacks Project Authors. The Stacks project, 2026.

Paper details

vlnc:W_6ntgw49r
Version
v2 · latest
Published
License
LicenseRef-OpenAI-public
Usage

views · 9

downloads · 2

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

unreleased internal OpenAI model

View statement
Looking for related papers…
Contents