betaFeedback

WorksMathematics

Intersection Product Formulas and Integral Dependence

Abstract

We prove the intersection product formulas of Gaffney and Gassler for polar multiplicities and Segre numbers over arbitrary Noetherian local rings. For equidimensional and universally catenary rings, we give a geometric refinement of the Principle of Specialization of Integral Dependence, in which the maximal dimension of the images of exceptional components mapping into the hypersurface determines a precise Segre number that strictly increases.

Introduction

The study of integral dependence through numerical invariants goes back to Rees’ multiplicity theorem [18]. In an equidimensional and universally catenary Noetherian local ring (R,m)(R,\mathfrak{m}), this theorem states that two m\mathfrak{m}-primary ideals I⊆JI \subseteq J have the same integral closure if and only if e(I,R)=e(J,R)e(I,R)=e(J,R). For ideals of arbitrary height, the corresponding criterion is expressed in terms of Segre numbers. In the complex analytic setting, Gaffney and Gassler [11] proved this extension of Rees’ theorem by establishing a Principle of Specialization of Integral Dependence. Their work also establishes the lexicographic upper semicontinuity of Segre numbers and gives cycle descriptions and intersection product formulas for Segre numbers and polar multiplicities.

Several of Gaffney–Gassler’s results have since been extended to the algebraic setting. Achilles and Rams [2] identified the Segre numbers of Gaffney and Gassler with the multiplicity sequence introduced by Achilles and Manaresi [1]. Polini, Trung, Ulrich, and Validashti [17] proved that equality of these sequences characterizes integral dependence of arbitrary ideals in an equidimensional and universally catenary Noetherian local ring. Algebraic cycle descriptions and semicontinuity results were obtained by Cid-Ruiz, Polini, and Ulrich [6]. One of the main goals of this paper is to establish the intersection product formulas of Gaffney and Gassler over arbitrary Noetherian local rings. We also use the geometry of the blow-up to give a different proof of the specialization criterion in [17 Theorem 3.3] and obtain a more precise statement that identifies a precise Segre number that must increase.

We now describe our results more precisely. Let (R,m,κ)(R,\mathfrak{m},\kappa) be a Noetherian local ring of dimension dd and I⊂RI \subset R be a proper ideal. Let X=Spec⁡(R)X=\operatorname{Spec}(R) and consider the punctured spectrum U=X∖{m}U=X\setminus\{\mathfrak{m}\}. The polar multiplicities mi(I,R)m_i(I,R) and the Segre numbers ci(I,R)c_i(I,R) are defined by the bigraded Hilbert polynomials associated to the standard bigraded κ\kappa-algebras

gr⁡m(K(I))andgr⁡m(gr⁡I(R)),\operatorname{gr}_{\mathfrak{m}}(\mathcal{K}(I)) \qquad\text{and} \qquad\operatorname{gr}_{\mathfrak{m}}(\operatorname{gr}_{I}(R)),
Equation not extracted

respectively (see Definition 2.2). We also consider the polar-Segre multiplicities νi(I,R)=mi(I,R)+ci(I,R)\nu_i(I,R)=m_i(I,R)+c_i(I,R).

To express these numbers as intersection products, we work on the joint blow-up

b: B=BmI(X)⟶X.b:\ B=\mathcal{B}_{\mathfrak{m}I}(X)\longrightarrow X.

On BB, both IOBI\mathcal{O}_B and mOB\mathfrak{m}\mathcal{O}_B are invertible ideals, and so they determine effective Cartier divisors EIE_I and EmE_{\mathfrak{m}} given by IOB=OB(−EI)I\mathcal{O}_B=\mathcal{O}_B(-E_I) and mOB=OB(−Em)\mathfrak{m}\mathcal{O}_B=\mathcal{O}_B(-E_{\mathfrak{m}}). We write [EI]=[EI]b−1(U)+[EI]b−1({m})[E_I]=[E_I]_{b^{-1}(U)}+[E_I]_{b^{-1}(\{\mathfrak{m}\})}, where the two summands consist of the components meeting b−1(U)b^{-1}(U) and those mapping to the closed point, respectively, with their multiplicities. The following theorem expresses the multiplicities in terms of these divisors and cycles.

Theorem A (Gaffney-Gassler intersection formulas; Theorem 3.5). With the notation above, for all 1≤i≤d−11\leq i\leq d-1, the following formulas hold:

mi(I,R)=−∫B(−EI)i(−Em)d−i⋅[B],m_i(I,R)=-\int_B(-E_I)^i(-E_{\mathfrak{m}})^{d-i}\cdot[B],
ci(I,R)=−∫B(−EI)i−1(−Em)d−i⋅[EI]b−1(U),c_i(I,R)=-\int_B(-E_I)^{i-1}(-E_{\mathfrak{m}})^{d-i}\cdot[E_I]_{b^{-1}(U)},
νi(I,R)=∫B(−EI)i−1(−Em)d−i⋅[EI]b−1({m}).\nu_i(I,R)=\int_B(-E_I)^{i-1}(-E_{\mathfrak{m}})^{d-i}\cdot[E_I]_{b^{-1}(\{\mathfrak{m}\})}.

Moreover, when κ\kappa is infinite, for all 1≤i≤d−21\leq i\leq d-2, we have the following cutting formulas:

(i) If dim⁡(R/I)<d\dim(R/I)<d and f∈If\in I is a general element, then

mi(I,R/fR)=mi+1(I,R),ci(I,R/fR)=ci+1(I,R),νi(I,R/fR)=νi+1(I,R).m_i(I,R/fR)=m_{i+1}(I,R),\qquad c_i(I,R/fR)=c_{i+1}(I,R),\qquad\nu_i(I,R/fR)=\nu_{i+1}(I,R).

(ii) If x∈mx\in\mathfrak{m} is a general element, then

mi(I,R/xR)=mi(I,R),ci(I,R/xR)=ci(I,R),νi(I,R/xR)=νi(I,R).m_i(I,R/xR)=m_i(I,R),\qquad c_i(I,R/xR)=c_i(I,R),\qquad\nu_i(I,R/xR)=\nu_i(I,R).

Here the integrals denote local intersection degrees with support over the closed point (see § 2.2). The proof of Theorem A uses Thorup’s rational equivalence theory [24], which allows us to work over an arbitrary Noetherian local ring.

We next turn to specialization, the main step in the integral-dependence criteria established in [11] and [17]. Geometrically, the issue is whether a component of the exceptional divisor Proj⁡(gr⁡I(R))\operatorname{Proj}(\operatorname{gr}_I(R)) maps into a given hypersurface V(t)⊂XV(t)\subset X. Our next result shows that such components force a strict increase in a precise Segre number: the index of this number is determined by the largest dimension of their images in XX.

Theorem B (Theorem 4.3). Assume that RR is equidimensional and universally catenary and that ht⁡(I)≥1\operatorname{ht}(I)\geq1. Let t∈mt\in\mathfrak{m} be an element such that ht⁡(t)=1\operatorname{ht}(t)=1 and ht⁡(I,t)≥2\operatorname{ht}(I,t)\geq2. Assume there is a minimal prime of gr⁡I(R)\operatorname{gr}_I(R) that contains tt. Let ss be the maximal dimension

s:=max⁡{dim⁡(R/Q∩R)∣Q minimal prime of gr⁡I(R) with t∈Q}s:=\max\left\{\dim(R/Q\cap R)\mid Q\text{ minimal prime of }\operatorname{gr}_I(R)\text{ with }t\in Q\right\}

of the contractions to RR of the minimal primes of gr⁡I(R)\operatorname{gr}_I(R) containing tt. Let j:=d−1−sj:=d-1-s. Then we have the strict inequality

cj(I,R/tR)>cj(I,R).c_j(I,R/tR)>c_j(I,R).

In an analytic setting, Teissier’s Principle of Specialization of Integral Dependence (PSID) (see [22] and [23 Appendice I]) states that for a family of zero-dimensional ideals with constant Hilbert-Samuel multiplicity, a section is integrally dependent on the total family if and only if it is integrally dependent on the fibers corresponding to a Zariski-open dense subset of the base. Using Segre numbers, Gaffney and Gassler [11] (also in the analytic setting) extended the PSID to families of ideals that are not necessarily zero-dimensional. As a direct consequence of Theorem B, we obtain in Corollary 4.4 an algebraic version of the PSID, showing that, under certain hypotheses, it suffices to check integral dependence on the open set D(t)⊂Spec⁡(R)D(t) \subset\operatorname{Spec}(R).

The basic outline of this paper is as follows.

Outline. In § 2, we recall the definitions and preliminary results used throughout the paper. The proof of Theorem A, together with the cutting formulas and cycle descriptions, is given in § 3. We conclude that section by studying integral dependence under general hypersurface sections (see § 3.1). Finally, in § 4, we prove Theorem B.

Basic tools

In this section, we recall the definitions and results used throughout the paper. We also prove some basic tools that will be necessary. The following data is used throughout.

Setup 2.1. Let (R,m,κ)(R,\mathfrak{m},\kappa) be a dd-dimensional Noetherian local ring and I⊂RI \subset R be a proper ideal. Let X:=Spec⁡(R)X := \operatorname{Spec}(R).

We denote by Zi(X)Z_i(X) the free Abelian group of cycles on XX of Krull dimension ii (in § 2.2 below, we also consider the notion of rational equivalence developed by Thorup [24]). For r≥0r \ge0, set Z≤r(X):=⨁i=0rZi(X)Z_{\le r}(X) := \bigoplus_{i=0}^{r} Z_i(X). We define the multiplicity homomorphism er:Z≤r(X)→Ze_r : Z_{\le r}(X) \to\mathbb{Z} by

er(∑pnp[V(p)]):=∑dim⁡(R/p)=rnpe(R/p),e_r\left(\sum_{\mathfrak{p}} n_{\mathfrak{p}}[V(\mathfrak{p})]\right) := \sum_{\dim(R/\mathfrak{p})=r} n_{\mathfrak{p}}e(R/\mathfrak{p}),

where e(R/p)e(R/\mathfrak{p}) denotes Hilbert–Samuel multiplicity.

Polar multiplicities and Segre numbers

First, we recall the notions of polar multiplicities and Segre numbers.

Definition 2.2. Consider the standard bigraded κ\kappa-algebras A:=gr⁡m(R(I))A := \operatorname{gr}_{\mathfrak{m}}(\mathcal{R}(I)) and G:=gr⁡m(gr⁡I(R))G := \operatorname{gr}_{\mathfrak{m}}(\operatorname{gr}_{I}(R)). Their bigraded components are

Av,n=mnIvmn+1IvandGv,n=mnIv+Iv+1mn+1Iv+Iv+1.A_{v,n} = \frac{\mathfrak{m}^{n}I^{v}}{\mathfrak{m}^{n+1}I^{v}} \qquad\text{and}\qquad G_{v,n} = \frac{\mathfrak{m}^{n}I^{v}+I^{v+1}}{\mathfrak{m}^{n+1}I^{v}+I^{v+1}}.

For v,n≫0v,n \gg0, the Hilbert function of AA is a polynomial with homogeneous part of degree d−1d-1 given by

∑i=0d−1mi(I,R)vind−1−ii!(d−1−i)!.\sum_{i=0}^{d-1} m_i(I,R)\frac{v^i n^{d-1-i}}{i!(d-1-i)!}.

The second sum transform of the Hilbert function

HG(v,n):=∑a=0v∑b=0ndim⁡κ(Ga,b)H_G(v,n) := \sum_{a=0}^{v}\sum_{b=0}^{n}\dim_{\kappa}(G_{a,b})

is eventually a polynomial with homogeneous part of degree dd given by

∑i=0dci(I,R)νinnd−ii!(d−i)!.\sum_{i=0}^{d} c_i(I,R)\frac{\nu_i^n n^{d-i}}{i!(d-i)!}.

We call mi(I,R)m_i(I,R) the polar multiplicities and ci(I,R)c_i(I,R) the Segre numbers of II. We put md(I,R)=0m_d(I,R)=0 and νi(I,R)=mi(I,R)+ci(I,R)\nu_i(I,R)=m_i(I,R)+c_i(I,R) for 1≤i≤d1\le i\le d.

When d≥2d\ge2, the coefficients ci(I,R)c_i(I,R) with 1≤i≤d−11\le i\le d-1 can be read from the ordinary bigraded Hilbert polynomial of GG. The two end coefficients are given by

c0(I,R)=e(R)−m0(I,R)andcd(I,R)=j(I)=ed−1(Hm0(gr⁡I(R))).c_0(I,R)=e(R)-m_0(I,R)\qquad\text{and}\qquad c_d(I,R)=j(I)=e_{d-1}\left(H^0_{\mathfrak m}(\operatorname{gr}_I(R))\right).

We have that: if ht⁡(I)≥1\operatorname{ht}(I)\ge1, then c0(I,R)=0c_0(I,R)=0; if II is m\mathfrak m-primary, then ci(I,R)=0c_i(I,R)=0 for i<di<d and cd(I,R)=e(I,R)c_d(I,R)=e(I,R).

Thorup's rational equivalence theory

We recall the basics of the rational equivalence theory developed by Thorup [24] over Noetherian schemes, which will be used in this paper (a slightly less general treatment is made in [21 Tag 02QK] over a universally catenary scheme). See also the recent developments [4, 5, 7].

A grading on a Noetherian base scheme assigns integers to its integral closed subschemes and induces a grading on schemes essentially of finite type over the base. In Thorup's theory [24] one can always use the canonical grading given by

δcan(V):=−dim⁡OX,ηV.\delta_{\mathrm{can}}(V):=-\dim\mathcal{O}_{X,\eta_V}.

where ηV\eta_V is the generic point of an integral closed subscheme V⊂XV\subset X. Since X=Spec⁡(R)X=\operatorname{Spec}(R) has finite Krull dimension, we use instead the topological grading

δtop(V):=dim⁡(V)\delta_{\mathrm{top}}(V):=\dim(V)

for an integral closed subscheme V⊂XV\subset X.

Let π:Y→X\pi:Y\to X be a morphism essentially of finite type. The grading on YY induced by the topological grading on XX is given by

δX,top(W):=trdeg⁡κ(V)(κ(W))+dim⁡(V),\delta_{X,\mathrm{top}}(W):=\operatorname{trdeg}_{\kappa(V)}(\kappa(W))+\dim(V),

where W⊂YW\subset Y is an integral closed subscheme and V=π(W)⊂XV=\pi(W)\subset X. The dimension formula gives

dim⁡(W)≤δX,top(W)\dim(W)\le\delta_{X,\mathrm{top}}(W)

(see, e.g., [15 Theorem 15.5]). Moreover, Thorup [24 Proposition 4.3] showed that if π\pi is proper, then

δX,top(W)=dim⁡(W).\delta_{X,\mathrm{top}}(W)=\dim(W).

Given a grading on XX and a scheme YY essentially of finite type over XX, by utilizing the induced grading on $Y, Thorup builds a rational equivalence theory for YY. This leads to the following very convenient convention.

Convention 2.3. Let YY be a proper scheme over X=Spec⁡(R)X=\operatorname{Spec}(R). Throughout this paper, we consider Thorup's rational equivalence theory on YY with the induced grading δX,top\delta_{X,\mathrm{top}}. Most importantly, the topological grading of YY coincides with the induced grading δX,top\delta_{X,\mathrm{top}} on $Y.

Therefore, for every proper XX-scheme YY, we may use the ordinary dimensional cycle group Zr(Y)Z_r(Y). We write Ar(Y/X)A_r(Y/X) for the quotient by Thorup rational equivalence and [α]r[\alpha]_r for the rr-dimensional part of a cycle α\alpha. If f:Z→Yf: Z \to Y is a proper morphism of proper XX-schemes, we get a functorial homomorphism

f∗:Ar(Z/X)⟶Ar(Y/X)f_*: A_r(Z/X) \longrightarrow A_r(Y/X)

by [24 Proposition 6.5]. Moreover, Cartier-divisor intersections and the projection formula are available for all the proper schemes used throughout this paper.

We use the following convention for local intersection degrees. Let π:Y→X\pi: Y \to X be a proper morphism and set Y0:=Y×XSpec⁡(κ)Y_0 := Y \times_X \operatorname{Spec}(\kappa). Since Y0Y_0 is proper over Spec⁡(κ)\operatorname{Spec}(\kappa), its structure morphism π0:Y0→Spec⁡(κ)\pi_0: Y_0 \to\operatorname{Spec}(\kappa) induces the usual degree homomorphism

deg⁡κ:A0(Y0)→(π0)∗A0(Spec⁡(κ))≅Z\deg_{\kappa}: A_0(Y_0) \xrightarrow{(\pi_0)_*} A_0(\operatorname{Spec}(\kappa)) \cong\mathbb{Z}

(see [10 Definition 1.4]). For a class β∈A0(Y0)\beta\in A_0(Y_0) represented by a zero-dimensional cycle ∑VnV[V]\sum_V n_V[V], we write

∫Yβ:=deg⁡κ(β)=∑VnV[R(V):κ].\int_Y \beta:= \deg_{\kappa}(\beta) = \sum_V n_V[\mathop{\rm R}(V):\kappa].

This integer is independent of the chosen representative, since proper pushforward respects rational equivalence on Y0Y_0.

Passage to an infinite residue field

Let S:=R[z]mR[z]S := \mathbb{R}[z]_{\mathfrak{m}\mathbb{R}[z]}, where zz is a new indeterminate. The extension R→S\mathbb{R} \to S is faithfully flat and local, dim⁡(S)=d\dim(S) = d, its maximal ideal is mS\mathfrak{m}S, and its residue field is κ(z)\kappa(z). For every ideal I⊂RI \subset\mathbb{R}, flat base change gives

(R/I)⊗RS≅(R/I)[z]m(R/I)[z]andBℓIS(Spec⁡(S))≅BℓI(X)×XSpec⁡(S).(\mathbb{R}/I) \otimes_{\mathbb{R}} S \cong(\mathbb{R}/I)[z]_{\mathfrak{m}(\mathbb{R}/I)[z]} \qquad\text{and}\qquad \mathcal{B}\ell_{IS}(\operatorname{Spec}(S)) \cong\mathcal{B}\ell_I(X) \times_X \operatorname{Spec}(S).

The bigraded components in Definition 2.2 are extended from κ\kappa to κ(z)\kappa(z), so their dimensions and the resulting multiplicities are unchanged. For a proper scheme YY over X=Spec⁡(R)X = \operatorname{Spec}(\mathbb{R}), on the closed fiber Y0=Y×XSpec⁡(κ)Y_0 = Y \times_X \operatorname{Spec}(\kappa), the base change is the field extension κ⊂κ(z)\kappa\subset\kappa(z), and so it preserves degrees and commutes with proper pushforward and Cartier intersections. Therefore, passing to an infinite residue field is a mild operation that keeps the intersection-theoretic data.

General sections

We first give the avoidance argument used to choose general sections.

Lemma 2.4. Assume that the residue field κ\kappa is infinite. Let π:Y→X=Spec⁡(R)\pi: Y \to X = \operatorname{Spec}(\mathbb{R}) be a proper morphism and L\mathcal{L} be a globally generated line bundle on YY. Choose points η1,…,ηs∈Y\eta_1,\ldots,\eta_s \in Y. Then, for a general section σ∈H0(Y,L)\sigma\in H^0(Y,\mathcal{L}), the zero scheme Z(σ)⊂YZ(\sigma) \subset Y does not contain any of the points η1,…,ηs\eta_1,\ldots,\eta_s.

Proof. Since π\pi is proper and R\mathbb{R} is Noetherian, M:=H0(Y,L)M := H^0(Y,\mathcal{L}) is a finitely generated R\mathbb{R}-module. For 1≤i≤s1 \le i \le s, let

Mi:=Ker⁡(M⟶Lηi⊗OY,ηiκ(ηi)).M_i := \operatorname{Ker}\left(M \longrightarrow\mathcal{L}_{\eta_i} \otimes_{\mathcal{O}_{Y,\eta_i}} \kappa(\eta_i)\right).

The global generation of L\mathcal{L} implies that each MiM_i is a proper submodule of MM. Let

V:=M/mMandVi:=(Mi+mM)/mM⊂V.V := M/\mathfrak{m}M \qquad\text{and}\qquad V_i := (M_i+\mathfrak{m}M)/\mathfrak{m}M \subset V.

If Vi=VV_i = V, then Nakayama’s lemma applied to M/MiM/M_i would give M=MiM = M_i, a contradiction. Thus each ViV_i is a proper κ\kappa-linear subspace of VV. Since κ\kappa is infinite, V∖⋃iViV \setminus\bigcup_i V_i is a nonempty Zariski-open subset. Every section σ∈M\sigma\in M whose image lies in this open subset has nonzero value in the fiber of L\mathcal{L} at each ηi\eta_i, and so ηi∉Z(σ)\eta_i \notin Z(\sigma) for all ii.

Corollary 2.5. Assume that the residue field κ\kappa is infinite. Let π:Y→X=Spec⁡(R)\pi:Y \to X=\operatorname{Spec}(R) be a proper morphism with YY integral and L\mathcal{L} be a globally generated line bundle on YY. Let T⊂XT \subset X be a closed subset such that π(Y)⊄T\pi(Y) \not\subset T. Then, for a general section σ∈H0(Y,L)\sigma\in H^0(Y,\mathcal{L}), no irreducible component of π(Z(σ))\pi(Z(\sigma)) is contained in TT.

Proof. Let W:=π−1(T)⊊YW:=\pi^{-1}(T)\subsetneq Y. By Lemma 2.4, the zero scheme Z(σ)Z(\sigma) of a general section σ∈H0(Y,L)\sigma\in H^0(Y,\mathcal{L}) avoids the generic point of YY and the generic points of all irreducible components of WW. Since YY is integral, every irreducible component of Z(σ)Z(\sigma) has codimension one in YY by the principal ideal theorem. If such a component were contained in WW, it would therefore be an irreducible component of WW, a contradiction. So the proof of the corollary is complete.

The following lemma describes specialization of a biprojective Rees algebra by a general element.

Lemma 2.6. Assume that the residue field κ\kappa is infinite. Let I,J⊆RI,J\subseteq R be ideals, and consider the multi-Rees algebra RR(I,J)=R[IT1,JT2]\mathcal{R}_R(I,J)=R[IT_1,JT_2]. For a general element x∈Ix\in I, put R‾=R/(x)\overline{R}=R/(x), I‾=IR‾\overline{I}=I\overline{R}, and J‾=JR‾\overline{J}=J\overline{R}. The natural surjection RR(I,J)/(xT1)⟶RR‾(I‾,J‾)\mathcal{R}_R(I,J)/(xT_1)\longrightarrow\mathcal{R}_{\overline{R}}(\overline{I},\overline{J}) induces a natural isomorphism

BiProj⁡(RR(I,J)/(xT1))≅BiProj⁡(RR‾(I‾,J‾)).\operatorname{BiProj}\bigl(\mathcal{R}_R(I,J)/(xT_1)\bigr)\cong\operatorname{BiProj}\bigl(\mathcal{R}_{\overline{R}}(\overline{I},\overline{J})\bigr).

Proof. This follows from Rees’ lemma [19 Lemma 1.2]; see also Trung–Verma [TV07].

Intersection product formulas

In this section, by building on the work of Thorup [24] (see § 2.2), we prove the intersection product formulas stated in Theorem A. Throughout this section, we use the following setup.

Setup 3.1. Let (R,m,κ)(R,\mathfrak{m},\kappa) be a dd-dimensional Noetherian local ring and I⊂RI\subset R be a proper ideal. Let X:=Spec⁡(R)X:=\operatorname{Spec}(R) and consider the punctured spectrum U:=X∖{m}U:=X\setminus\{\mathfrak{m}\}. Let bI:BI:=Bl⁡I(X)→Xb_I:B_I:=\operatorname{Bl}_I(X)\to X be the blow-up of X=Spec⁡(R)X=\operatorname{Spec}(R) along II. Likewise, the blow-up of XX at the origin is bm:Bm:=Bl⁡m(X)→Xb_{\mathfrak{m}}:B_{\mathfrak{m}}:=\operatorname{Bl}_{\mathfrak{m}}(X)\to X. Finally, we consider the joint blow-up b:B:=Bl⁡mI(X)→Xb:B:=\operatorname{Bl}_{\mathfrak{m}I}(X)\to X. Algebraically, we have the descriptions

BI=Proj⁡(R(I)),Bm=Proj⁡(R(m))andB=BiProj⁡(R(I,m))≅Proj⁡(R(mI)).B_I=\operatorname{Proj}(\mathcal{R}(I)),\qquad B_{\mathfrak{m}}=\operatorname{Proj}(\mathcal{R}(\mathfrak{m}))\qquad\text{and}\qquad B=\operatorname{BiProj}(\mathcal{R}(I,\mathfrak{m}))\cong\operatorname{Proj}(\mathcal{R}(\mathfrak{m}I)).

Here R(I):=R[IT]⊂R[T]\mathcal{R}(I):=R[IT]\subset R[T], R(m):=R[mT]⊂R[T]\mathcal{R}(\mathfrak{m}):=R[\mathfrak{m}T]\subset R[T] and R(mI):=R[mIT]⊂R[T]\mathcal{R}(\mathfrak{m}I):=R[\mathfrak{m}IT]\subset R[T] are Rees algebras, and R(I,m):=R[IT1,mT2]⊂R[T1,T2]\mathcal{R}(I,\mathfrak{m}):=R[IT_1,\mathfrak{m}T_2]\subset R[T_1,T_2] is a multi-Rees algebra. We have the following commutative diagram

B→b~mBIb~I↓↓bIBm→bmX\begin{CD} B @>{\widetilde{b}_{\mathfrak{m}}}>> B_I \\ @V{\widetilde{b}_I}VV @VV{b_I}V \\ B_{\mathfrak{m}} @>{b_{\mathfrak{m}}}>> X \end{CD}

with b=bI∘b~m=bm∘b~Ib=b_I\circ\widetilde{b}_{\mathfrak{m}}=b_{\mathfrak{m}}\circ\widetilde{b}_I. Let DI⊂BID_I\subset B_I and Dm⊂BmD_{\mathfrak{m}}\subset B_{\mathfrak{m}} be the exceptional divisors of bIb_I and bmb_{\mathfrak{m}}, respectively. On the joint blow-up BB, consider the effective Cartier divisors

EI:=V(IOB),Em:=V(mOB)andE:=V(mIOB).E_I:=V(I\mathcal{O}_B),\qquad E_{\mathfrak{m}}:=V(\mathfrak{m}\mathcal{O}_B)\qquad\text{and}\qquad E:=V(\mathfrak{m}I\mathcal{O}_B).

Then EE is the exceptional divisor of BB, and EI=b~I∗(DI)E_{\mathrm{I}}=\widetilde{b}^{*}_{\mathrm{I}}(D_{\mathrm{I}}) and Em=b~m∗(Dm)E_{\mathrm{m}}=\widetilde{b}^{*}_{\mathrm{m}}(D_{\mathrm{m}}). Let

hI:=c1(OB(1,0)),hm:=c1(OB(0,1)),h:=c1(OB(1,1)).h_{\mathrm{I}}:=c_{1}\bigl(\mathcal{O}_{B}(1,0)\bigr),\qquad h_{\mathrm{m}}:=c_{1}\bigl(\mathcal{O}_{B}(0,1)\bigr),\qquad h:=c_{1}\bigl(\mathcal{O}_{B}(1,1)\bigr).

We start with the following basic lemma.

Lemma 3.2. On B=BℓmI(X)B=B\ell_{\mathrm{mI}}(X), the following statements hold:

(i) OB(1,0)=OB(−EI)\mathcal{O}_{B}(1,0)=\mathcal{O}_{B}(-E_{\mathrm{I}}), OB(0,1)=OB(−Em)\mathcal{O}_{B}(0,1)=\mathcal{O}_{B}(-E_{\mathrm{m}}) and OB(1,1)=OB(−E)\mathcal{O}_{B}(1,1)=\mathcal{O}_{B}(-E).

(ii) E=EI+EmE=E_{\mathrm{I}}+E_{\mathrm{m}}.

(iii) h=hI+hmh=h_{\mathrm{I}}+h_{\mathrm{m}}.

Proof. Since B=BiProj⁡(R(I,m))B=\operatorname{BiProj}(\mathcal{R}(I,\mathfrak{m})), the invertible sheaves OB(1,0)\mathcal{O}_{B}(1,0), OB(0,1)\mathcal{O}_{B}(0,1), and OB(1,1)\mathcal{O}_{B}(1,1) correspond to the invertible ideals IOBI\mathcal{O}_{B}, mOB\mathfrak{m}\mathcal{O}_{B}, and (mI)OB(\mathfrak{m}I)\mathcal{O}_{B}, respectively. This gives

OB(1,0)=OB(−EI),OB(0,1)=OB(−Em),OB(1,1)=OB(−E).\mathcal{O}_{B}(1,0)=\mathcal{O}_{B}(-E_{\mathrm{I}}),\qquad\mathcal{O}_{B}(0,1)=\mathcal{O}_{B}(-E_{\mathrm{m}}),\qquad\mathcal{O}_{B}(1,1)=\mathcal{O}_{B}(-E).

Since (mI)OB=mOB⋅IOB(\mathfrak{m}I)\mathcal{O}_{B}=\mathfrak{m}\mathcal{O}_{B}\cdot I\mathcal{O}_{B}, we get E=EI+EmE=E_{\mathrm{I}}+E_{\mathrm{m}}. Taking first Chern classes gives h=hI+hmh=h_{\mathrm{I}}+h_{\mathrm{m}}. □

Remark 3.3. We recall some useful notation from [12, 13]. Let YY be a scheme, WW a subscheme of YY, and S∈Z∗(Y)S\in Z_{*}(Y) a cycle on YY. We define the WW-part SWS^{W} as follows: if SS is the fundamental cycle of an integral closed subscheme S⊂YS\subset Y, then SW=SS^{W}=S if WW contains the generic point of SS, and SW=0S^{W}=0 otherwise; then extend this definition by linearity.

The next proposition will play an important role in our approach.

Proposition 3.4. On B=BℓmI(X)B=B\ell_{\mathrm{mI}}(X), the following equalities of cycles hold:

[Em]=[BiProj⁡(gr⁡m(R(I)))]and[EI]b−1(U)=[BiProj⁡(R(m⋅gr⁡I(R)))].[E_{\mathrm{m}}]=\left[\operatorname{BiProj}\bigl(\operatorname{gr}_{\mathrm{m}}(\mathcal{R}(I))\bigr)\right]\qquad\text{and}\qquad[E_{\mathrm{I}}]^{b^{-1}(U)}=\left[\operatorname{BiProj}\bigl(\mathcal{R}(\mathfrak{m}\cdot\operatorname{gr}_{I}(R))\bigr)\right].

Proof. For the first equality, since Em=V(mOB)E_{\mathrm{m}}=V(\mathfrak{m}\mathcal{O}_{B}), we have gr⁡m(R(I))≅R(I,m)/mR(I,m)\operatorname{gr}_{\mathfrak{m}}(\mathcal{R}(I))\cong\mathcal{R}(I,\mathfrak{m})/\mathfrak{m}\mathcal{R}(I,\mathfrak{m}), and therefore Em≅BiProj⁡(gr⁡m(R(I)))E_{\mathrm{m}}\cong\operatorname{BiProj}(\operatorname{gr}_{\mathfrak{m}}(\mathcal{R}(I))).

For the second equality, since EI=V(IOB)E_{\mathrm{I}}=V(I\mathcal{O}_{B}), we have EI≅BiProj⁡(S)⊂BE_{\mathrm{I}}\cong\operatorname{BiProj}(S)\subset B with S:=R(I,m)/IR(I,m)S:=\mathcal{R}(I,\mathfrak{m})/I\mathcal{R}(I,\mathfrak{m}). Consider the specialization map

s:S≅R(m⋅R(I))I⋅R(m⋅R(I))⟶R(m⋅gr⁡I(R))≅R(m⋅R(I)I⋅R(I)).\mathfrak{s}: S\cong\frac{\mathcal{R}(\mathfrak{m}\cdot\mathcal{R}(I))}{I\cdot\mathcal{R}(\mathfrak{m}\cdot\mathcal{R}(I))}\longrightarrow\mathcal{R}(\mathfrak{m}\cdot\operatorname{gr}_{I}(R))\cong\mathcal{R}\left(\mathfrak{m}\cdot\frac{\mathcal{R}(I)}{I\cdot\mathcal{R}(I)}\right).

For any prime p∈U\mathfrak{p}\in U on the punctured spectrum, we have that the map s⊗RRp\mathfrak{s}\otimes_{R}R_{\mathfrak{p}} is an isomorphism. Hence we obtain

ml⋅Ker⁡(s)=0for some l≥1.\mathfrak{m}^{l}\cdot\operatorname{Ker}(\mathfrak{s})=0\qquad\text{for some }l\ge1.

This yields the following set-theoretical decomposition

Supp⁡EI(S/m⋅S)∪Supp⁡EI(S/Ker⁡(s))=EI.\operatorname{Supp}_{E_{\mathrm{I}}}\left(S/\mathfrak{m}\cdot S\right)\cup\operatorname{Supp}_{E_{\mathrm{I}}}\left(S/\operatorname{Ker}(\mathfrak{s})\right)=E_{\mathrm{I}}.

Since the biprojective scheme BiProj⁡(R(m⋅gr⁡I(R)))⊂B\operatorname{BiProj}(\mathcal{R}(\mathfrak{m}\cdot\operatorname{gr}_{I}(R)))\subset B is isomorphic to the blow-up BℓZ(Y)B\ell_{Z}(Y) of Y:=Proj⁡(gr⁡I(R))Y:=\operatorname{Proj}(\operatorname{gr}_{I}(R)) along Z:=V(m⋅OY)Z:=V(\mathfrak{m}\cdot\mathcal{O}_{Y}), its generic points are contained in b−1(U)b^{-1}(U). On the other hand, by the decomposition (1), the generic points of EIE_{\mathrm{I}} that are contained in b−1(U)b^{-1}(U) should also be contained in V(Ker⁡(s))≅BiProj⁡(R(m⋅gr⁡I(R)))V(\operatorname{Ker}(\mathfrak{s}))\cong\operatorname{BiProj}(\mathcal{R}(\mathfrak{m}\cdot\operatorname{gr}_{I}(R))). Therefore, the asserted equality of cycles

[EI]b−1(U)=[BiProj⁡(R(m⋅gr⁡I(R)))][E_{\mathrm{I}}]^{b^{-1}(U)}=\left[\operatorname{BiProj}\bigl(\mathcal{R}(\mathfrak{m}\cdot\operatorname{gr}_{I}(R))\bigr)\right]

follows. □

We can now express polar multiplicities and Segre numbers as intersection products of divisors on the joint blow-up B=BmI(X)\mathrm{B}=\mathcal{B}_{\mathfrak{m}I}(X). The following theorem extends several results of Gaffney and Gassler [11] to arbitrary Noetherian local rings.

Theorem 3.5 (Gaffney-Gassler intersection formulas). Assume Setup 3.1. Then we have the following intersection product formulas and cutting results.

(i) (Polar multiplicities):

mi(I,R)=−∫B(−EI)i(−Em)d−i−1⋅[B]for 0≤i≤d−1.m_i(I,R)=-\int_B(-E_I)^i(-E_{\mathfrak m})^{d-i-1}\cdot[B]\qquad\text{for }0\leq i\leq d-1.

(ii) (Segre numbers):

ci(I,R)=−∫B(−EI)i−1(−Em)d−i⋅[EI]b−1(U)for 1≤i≤d−1.c_i(I,R)=-\int_B(-E_I)^{i-1}(-E_{\mathfrak m})^{d-i}\cdot[E_I]^{b-1}(U)\qquad\text{for }1\leq i\leq d-1.

(iii) (Polar-Segre multiplicities):

νi(I,R)=∫B(−EI)i−1(−Em)d−i⋅[EI]b−1({m})for 1≤i≤d−1.\nu_i(I,R)=\int_B(-E_I)^{i-1}(-E_{\mathfrak m})^{d-i}\cdot[E_I]^{b-1}(\{\mathfrak m\})\qquad\text{for }1\leq i\leq d-1.

(iv) (Cutting with a general element of II): Assume that κ\kappa is infinite and dim⁡(R/I)<d\dim(R/I)<d. For a general element f∈If\in I, we have

mi(I,R/fR)=mi+1(I,R),ci(I,R/fR)=ci+1(I,R),νi(I,R/fR)=νi+1(I,R)m_i(I,R/fR)=m_{i+1}(I,R),\qquad c_i(I,R/fR)=c_{i+1}(I,R),\qquad\nu_i(I,R/fR)=\nu_{i+1}(I,R)

for $1\leq i\leq d-2.

(v) (Cutting with a general element of m\mathfrak m): Assume that κ\kappa is infinite. For a general element x∈mx\in\mathfrak m, we have

mi(I,R/xR)=mi(I,R),ci(I,R/xR)=ci(I,R),νi(I,R/xR)=νi(I,R)m_i(I,R/xR)=m_i(I,R),\qquad c_i(I,R/xR)=c_i(I,R),\qquad\nu_i(I,R/xR)=\nu_i(I,R)

for $1\leq i\leq d-2.

Proof. (i) Let Y:=BiProj⁡(gr⁡m(R(I)))⊂BY:=\operatorname{BiProj}(\operatorname{gr}_{\mathfrak m}(\mathcal{R}(I)))\subset B. By [12 §4], Lemma 3.2 and Proposition 3.4, we obtain

−∫B(−EI)i(−Em)d−i−1⋅[B]=∫BhIihmd−i−1⋅[Y]=∫Yc1(OY(1,0))ic1(OY(0,1))d−i−1⋅[Y]=mi(I,R),\begin{aligned} -\int_B(-E_I)^i(-E_{\mathfrak m})^{d-i-1}\cdot[B] =\int_B h_I^i h_{\mathfrak m}^{d-i-1}\cdot[Y] \\ &=\int_Y c_1(\mathcal{O}_Y(1,0))^i c_1(\mathcal{O}_Y(0,1))^{d-i-1}\cdot[Y] \\ &=m_i(I,R), \end{aligned}

where the last equality follows by Definition 2.2.

(ii) Let W:=BiProj⁡(R(m⋅gr⁡I(R)))⊂BW:=\operatorname{BiProj}(\mathcal{R}(\mathfrak m\cdot\operatorname{gr}_I(R)))\subset B and Z:=BiProj⁡(gr⁡m(gr⁡I(R)))⊂BZ:=\operatorname{BiProj}(\operatorname{gr}_{\mathfrak m}(\operatorname{gr}_I(R)))\subset B. From [12 §4], Lemma 3.2 and Proposition 3.4, we get

−∫B(−EI)i−1(−Em)d−i⋅[EI]b−1(U)=−∫B(−EI)i−1(−Em)d−i⋅[W]=∫B(−EI)i−1(−Em)d−i−1⋅[Z]=∫BhIi−1hmd−i−1⋅[Z]=∫Zc1(OZ(1,0))i−1c1(OZ(0,1))d−i−1⋅[Z]=ci(I,R),\begin{aligned} -\int_B(-E_I)^{i-1}(-E_{\mathfrak m})^{d-i}\cdot[E_I]^{b-1}(U) =-\int_B(-E_I)^{i-1}(-E_{\mathfrak m})^{d-i}\cdot[W] \\ &=\int_B(-E_I)^{i-1}(-E_{\mathfrak m})^{d-i-1}\cdot[Z] \\ &=\int_B h_I^{i-1}h_{\mathfrak m}^{d-i-1}\cdot[Z] \\ &= \int_Z c_1\left(\mathcal{O}_Z(1,0)\right)^{i-1} c_1\left(\mathcal{O}_Z(0,1)\right)^{d-i-1} \cdot[Z] \\ &= c_i(I,R), \end{aligned}

where the last equality follows by Definition 2.2.

(iii) For 1≤i≤d−11 \le i \le d-1, parts (i) and (ii) give

νi(I,R)=mi(I,R)+ci(I,R)=∫B(−EI)i−1(−Em)d−i⋅([EI]−[EI]b−1(U))=∫B(−EI)i−1(−Em)d−i⋅[EI]b−1({m}).\begin{aligned} \nu_i(I,R) = m_i(I,R) + c_i(I,R) \\ &= \int_B (-E_I)^{i-1}(-E_m)^{d-i} \cdot\left([E_I]-[E_I]^{b^{-1}(U)}\right) \\ &= \int_B (-E_I)^{i-1}(-E_m)^{d-i} \cdot[E_I]^{b^{-1}(\{m\})}. \end{aligned}

(iv) Let R‾:=R/fR\overline{R}:=R/fR, which has dimension d−1d-1. We have that OB(1,0)≅OB(−EI)≅OB\mathcal{O}_B(1,0)\cong\mathcal{O}_B(-E_I)\cong\mathcal{O}_B is globally generated by sections induced from elements of II. Let σf∈H0(B,OB(1,0))\sigma_f\in H^0(B,\mathcal{O}_B(1,0)) be the section induced by f∈If\in I. Let H:=Z(σf)⊂BH:=Z(\sigma_f)\subset B be the zero scheme determined by σf\sigma_f. By Lemma 2.6, we have the isomorphism

H≅BiProj⁡(T)H \cong\operatorname{BiProj}(T)

with T:=RR‾(I‾,m‾)T:=\mathcal{R}_{\overline{R}}(\overline{I},\overline{\mathfrak{m}}).

The avoidance arguments of Lemma 2.4 and Corollary 2.5 imply that HH is an effective Cartier divisor on BB, that restricts to an effective Cartier divisor on EIE_I, and that it commutes with the decomposition of [EI][E_I] into its closed-fiber and punctured-spectrum parts:

H⋅[EI]d−1b−1({m})=(H⋅[EI]d−1)b−1({m})=([BiProj⁡(T/IT)]d−2)b−1({m})∈Zd−2(B)H\cdot[E_I]_{d-1}^{b^{-1}(\{m\})} = \left(H\cdot[E_I]_{d-1}\right)^{b^{-1}(\{m\})} = \left([\operatorname{BiProj}(T/IT)]_{d-2}\right)^{b^{-1}(\{m\})} \in Z_{d-2}(B)

and

H⋅[EI]d−1b−1(U)=(H⋅[EI]d−1)b−1(U)=([BiProj⁡(T/IT)]d−2)b−1(U)∈Zd−2(B).H\cdot[E_I]_{d-1}^{b^{-1}(U)} = \left(H\cdot[E_I]_{d-1}\right)^{b^{-1}(U)} = \left([\operatorname{BiProj}(T/IT)]_{d-2}\right)^{b^{-1}(U)} \in Z_{d-2}(B).

Therefore, part (i) yields

mi+1(I,R)=−∫B(−EI)i+1(−Em)d−i−1⋅[B]=−∫B(−EI)i(−Em)d−i−1⋅[BiProj⁡(T)]=mi(I,R/fR).\begin{aligned} m_{i+1}(I,R)=-\int_B(-E_I)^{i+1}(-E_m)^{d-i-1}\cdot[B] \\ &=-\int_B(-E_I)^i(-E_m)^{d-i-1}\cdot[\operatorname{BiProj}(T)] \\ &=m_i(I,R/fR). \end{aligned}

From part (ii), we obtain

ci+1(I,R)=−∫B(−EI)i(−Em)d−i−1⋅[EI]b−1(U)=−∫B(−EI)i−1(−Em)d−i−1⋅[BiProj⁡(T/IT)]b−1(U)=ci(I,R/fR).\begin{aligned} c_{i+1}(I,R)=-\int_B(-E_I)^i(-E_m)^{d-i-1}\cdot[E_I]^{b^{-1}(U)} \\ &=-\int_B(-E_I)^{i-1}(-E_m)^{d-i-1}\cdot[\operatorname{BiProj}(T/IT)]^{b^{-1}(U)} \\ &=c_i(I,R/fR). \end{aligned}

Since νi(I,R)=mi(I,R)+ci(I,R)\nu_i(I,R)=m_i(I,R)+c_i(I,R), the proof of this part is now complete.

(v) The proof of this part follows similarly to the one of part (iv). □

We recall the polar schemes and Segre cycles (see [11, 6]). Assume that κ\kappa is infinite, and let K1,…,KdK_1,\ldots,K_d be general hyperplanes on BIB_I corresponding to general elements g1,…,gd∈Ig_1,\ldots,g_d\in I. Let

Pi(I,X):=bI(K1∩⋯∩Ki)andΛi(I,X):=(bI)∗([DI∩K1∩⋯∩Ki−1]d−i),P_i(I,X):=b_I(K_1\cap\cdots\cap K_i) \qquad\text{and}\qquad \Lambda_i(I,X):=(b_I)_*\left([D_I\cap K_1\cap\cdots\cap K_{i-1}]_{d-i}\right),

where the first image is scheme-theoretic. We now recover their multiplicity formulas for 1≤i≤d−11 \le i \le d-1.

Corollary 3.6. With notation as above, for all 1≤i≤d−11 \le i \le d-1, the following equalities hold:

mi(I,R)=ed−i([Pi(I,X)]d−i)andci(I,R)=ed−i(Λi(I,X)).m_i(I,R)=e_{d-i}\left([P_i(I,X)]_{d-i}\right) \qquad\text{and}\qquad c_i(I,R)=e_{d-i}\left(\Lambda_i(I,X)\right).

Proof. Let r:=d−i>0r:=d-i>0 and Hj:=(b~m)∗(Kj)H_j:=(\widetilde b_m)^*(K_j). By Lemma 2.4, we can assume that the following successive cuttings are all Cartier and yield the following effective cycles

Γi:=H1⋯Hi⋅[B]d∈Zr(B)andΣi:=H1⋯Hi−1⋅[EI]d−1bI−1(U)∈Zr(B).\Gamma_i:=H_1\cdots H_i\cdot[B]_d\in Z_r(B) \qquad\text{and}\qquad\Sigma_i:=H_1\cdots H_{i-1}\cdot[E_I]_{d-1}^{b_I^{-1}(U)}\in Z_r(B).

Successively applying the avoidance result of Corollary 2.5 implies that no component of Γi\Gamma_i is contained in b−1(V(I))b^{-1}(V(I)) and that no component of Σi\Sigma_i is contained in b−1({m})b^{-1}(\{\mathfrak m\}).

The projection formula (at the level of cycles; see [10 Proposition 1.4], [21 Tag 02RT]) yields the equalities

b∗(Σi)=(bI)∗((b~m)∗(Σi))=(bI)∗(K1⋯Ki−1⋅[DI]d−1bI−1(U))=Λi(I,X),b_*(\Sigma_i)=(b_I)_*\left((\widetilde b_m)_*(\Sigma_i)\right)=(b_I)_*\left(K_1\cdots K_{i-1}\cdot[D_I]_{d-1}^{b_I^{-1}(U)}\right)=\Lambda_i(I,X),

where the last equality follows because r=d−i>0r=d-i>0 and so bIb_I pushes to zero a component of [DI][D_I] that is contained in bI−1({m})b_I^{-1}(\{\mathfrak m\}). The natural isomorphisms b−1(X∖V(I))→∼bI−1(X∖V(I))→∼X∖V(I)b^{-1}(X\setminus V(I))\xrightarrow{\sim}b_I^{-1}(X\setminus V(I))\xrightarrow{\sim}X\setminus V(I) yield the isomorphism

(H1∩⋯∩Hi)∣b−1(X∖V(I))→∼(K1∩⋯∩Ki)∣bI−1(X∖V(I)).(H_1\cap\cdots\cap H_i)|_{b^{-1}(X\setminus V(I))}\xrightarrow{\sim}(K_1\cap\cdots\cap K_i)|_{b_I^{-1}(X\setminus V(I))}.

Since no component of H1⋯Hi⋅[B]dH_1\cdots H_i\cdot[B]_d is contained in b−1(V(I))b^{-1}(V(I)) and no component of K1⋯Ki⋅[BI]dK_1\cdots K_i\cdot[B_I]_d is contained in bI−1(V(I))b_I^{-1}(V(I)), we obtain

b∗(Γi)=[Pi(I,X)]r.b_*(\Gamma_i)=[P_i(I,X)]_r.

Finally, Theorem 3.5 and Lemma 3.7 give

mi(I,R)=−∫B(−Em)r⋅Γi=er(b∗(Γi))=er([Pi(I,X)]r)m_i(I,R)=-\int_B(-E_{\mathfrak m})^r\cdot\Gamma_i=e_r\left(b_*(\Gamma_i)\right)=e_r\left([P_i(I,X)]_r\right)

and

ci(I,R)=−∫B(−Em)r⋅Σi=er(b∗(Σi))=er(Λi(I,X)).c_i(I,R)=-\int_B(-E_{\mathfrak m})^r\cdot\Sigma_i=e_r\left(b_*(\Sigma_i)\right)=e_r\left(\Lambda_i(I,X)\right).

So the proof of the corollary is complete. □

The following technical lemma justifies the multiplicity computation used above.

Lemma 3.7. Let r>0r>0 and γ∈Zr(B)\gamma\in Z_r(B) be a cycle with no component contained in b−1({m})b^{-1}(\{\mathfrak m\}). Then

er(b∗(γ))=−∫B(−Em)r⋅γ.e_r\left(b_*(\gamma)\right)=-\int_B(-E_{\mathfrak m})^r\cdot\gamma.

Proof. By linearity, we may assume γ=[W]\gamma=[W], where W⊂BW\subset B is an integral closed subscheme of dimension rr such that W⊄b−1({m})W\not\subset b^{-1}(\{\mathfrak m\}). Let V:=b(W)V:=b(W) and consider the blow-up

π:V~:=Bl⁡mOV(V)⟶V\pi:\widetilde V:=\operatorname{Bl}_{\mathfrak m\mathcal O_V}(V)\longrightarrow V

with exceptional divisor FF determined by mOV~=OV~(−F)\mathfrak m\mathcal O_{\widetilde V}=\mathcal O_{\widetilde V}(-F). Since mOW=OW(−Em∣W)\mathfrak m\mathcal O_W=\mathcal O_W(-E_{\mathfrak m}|_W) is invertible, the universal property of the blow-up gives a proper factorization b∣W:W→fV~→πVb|_W:W\xrightarrow{f}\widetilde V\xrightarrow{\pi}V such that f∗(F)=Em∣Wf^*(F)=E_{\mathfrak m}|_W.

The projection formula and [4 Theorem 3.1] yield

−∫B(−Em)r⋅[W]=−∫V~(−F)r⋅f∗([W])=er(b∗([W])).-\int_{B}(-E_{m})^{r}\cdot[W]=-\int_{\widetilde{V}}(-F)^{r}\cdot f_{*}([W])=e_{r}(b_{*}([W])).

thus completing the proof of the lemma. □

Integral dependence under general hypersurface sections

By [17 Theorem 4.2], Segre numbers characterize integral dependence in equidimensional and universally catenary Noetherian local rings. By Theorem 3.5(v), when the residue field is infinite, a general element x∈mx \in\mathfrak{m} preserves the Segre numbers ci(I,R)c_{i}(I,R) for 1≤i≤d−21 \leq i \leq d-2. Together, these facts motivate the study of integral dependence under general hypersurface sections.

For ideals I⊆JI \subseteq J in RR, we consider the non-integral locus

N(I,J):={p∈Spec⁡(R)∣Ip≠Jp}⊂X.N(I,J):=\{\mathfrak{p}\in\operatorname{Spec}(R)\mid I_{\mathfrak{p}}\ne J_{\mathfrak{p}}\}\subset X.

The following theorem, valid over an arbitrary Noetherian local ring with infinite residue field, shows that if J‾(R/xR)\overline{J}(R/xR) is integral over I‾(R/xR)\overline{I}(R/xR) for a general x∈mx\in\mathfrak{m}, then dim⁡(N(I,J))≤1\dim(N(I,J))\leq1.

Theorem 3.8. Assume Setup 3.1 with κ\kappa infinite and I⊆JI\subseteq J two ideals in RR. Suppose there exists a prime p∈X=Spec⁡(R)\mathfrak{p}\in X=\operatorname{Spec}(R) such that dim⁡(R/p)≥2\dim(R/\mathfrak{p})\geq2 and JpJ_{\mathfrak{p}} is not integral over IpI_{\mathfrak{p}}. Then, for a general element x∈mx\in\mathfrak{m}, the ideal J‾=J‾R\overline{J}=\overline{J}R is not integral over I‾=I‾R\overline{I}=\overline{I}R in the quotient ring R‾=R/xR\overline{R}=R/xR.

Proof. Let q:Y=Proj⁡(⨁n≥0Jn)⟶Spec⁡(R)q:Y=\operatorname{Proj}\left(\bigoplus_{n\geq0}J^{n}\right)\longrightarrow\operatorname{Spec}(R) be the blow-up along the ideal JJ and EE denote the exceptional divisor of YY. Let E1,…,EkE_{1},\ldots,E_{k} be the irreducible components of EE. Let pi∈X=Spec⁡(R)\mathfrak{p}_{i}\in X=\operatorname{Spec}(R) denote the prime ideal corresponding to the generic point of the image q(Ei)q(E_{i}).

Since IOY⊆JOYI\mathcal{O}_{Y}\subseteq J\mathcal{O}_{Y} and JOYJ\mathcal{O}_{Y} is an invertible sheaf, we can globally twist by its inverse to find a uniquely determined ideal sheaf b⊆OY\mathfrak{b}\subseteq\mathcal{O}_{Y} such that

IOY=b⋅JOY.I\mathcal{O}_{Y}=\mathfrak{b}\cdot J\mathcal{O}_{Y}.

The ideal sheaf b\mathfrak{b} describes the exact obstruction to the integrality of JJ over II (see, e.g., [20 Corollary 1.2.5, §8.2]).

Let W=V(b)⊂YW=V(\mathfrak{b})\subset Y be the closed subscheme determined by b\mathfrak{b}. Since q(W)⊂Xq(W)\subset X is closed and p∈q(W)\mathfrak{p}\in q(W), we obtain the inequality

dim⁡(q(W))≥dim⁡(V(p))=dim⁡(R/p)≥2.\dim(q(W))\geq\dim(V(\mathfrak{p}))=\dim(R/\mathfrak{p})\geq2.

Let S:={pi∣1≤i≤k and pi⊊m}\mathcal{S}:=\{\mathfrak{p}_{i}\mid1\leq i\leq k\text{ and }\mathfrak{p}_{i}\subsetneq\mathfrak{m}\} be the set of primes strictly contained in m\mathfrak{m} given as the image of the generic points of the irreducible components EiE_{i}. By prime avoidance, we obtain that

dim⁡(V(x)∩q(W))=dim⁡(q(W))−1≥1\dim(V(x)\cap q(W))=\dim(q(W))-1\geq1

and that

x∉pifor any pi∈S.x\notin\mathfrak{p}_{i}\qquad\text{for any }\mathfrak{p}_{i}\in\mathcal{S}.

Thus we can fix a prime ideal q⊊m\mathfrak{q}\subsetneq\mathfrak{m} such that q∈V(x)∩q(W)\mathfrak{q}\in V(x)\cap q(W). Since q∈q(W)\mathfrak{q}\in q(W), we can choose a point w∈Ww\in W in the blow-up YY such that q(w)=qq(w)=\mathfrak{q}. Let A:=OY,wA:=\mathcal{O}_{Y,w} be the local ring at ww and mw⊂OY,w\mathfrak{m}_{w}\subset\mathcal{O}_{Y,w} be its maximal ideal. Since x∈q=q(w)x\in\mathfrak{q}=q(w), the pullback of xx into AA satisfies x∈mwx\in\mathfrak{m}_{w}.

By contradiction, assume that J‾=J‾R\overline{J}=\overline{J}R is integral over I‾=I‾R\overline{I}=\overline{I}R in the quotient ring R‾=R/xR\overline{R}=R/xR. Considering the natural map R/xR⟶A/xAR/xR\longrightarrow A/xA implies that J(A/xA)J(A/xA) is integral over I(A/xA)I(A/xA). Since JOY\mathcal{J}\mathcal{O}_Y is an invertible sheaf, locally at AA it is generated by a single regular element f∈Af \in A, so JA=(f)\mathcal{J}A=(f). The global equality IOY=b⋅JOYI\mathcal{O}_Y=\mathfrak{b}\cdot\mathcal{J}\mathcal{O}_Y at the point ww turns into IA=bwfIA=\mathfrak{b}_w f. Moreover, since ww was chosen to lie in W=V(b)W=V(\mathfrak{b}), we obtain bw⊆mw\mathfrak{b}_w\subseteq\mathfrak{m}_w. Since J(A/xA)\mathcal{J}(A/xA) is assumed to be integral over I(A/xA)I(A/xA), we have

fnA⊆(bwf)fn−1A+xA=bwfn+xA.\mathfrak{f}^n A\subseteq(\mathfrak{b}_w f)\mathfrak{f}^{n-1}A+xA=\mathfrak{b}_w f^n+xA.

for some n≥1n\geq1. This implies fn=bfn+axf^n=bf^n+ax for some b∈bwb\in\mathfrak{b}_w and a∈Aa\in A. Since b∈mwb\in\mathfrak{m}_w, we obtain

fn=(1−b)−1ax∈xA.f^n=(1-b)^{-1}ax\in xA.

We now divide the proof into two cases.

CASE 1: Assume that w∉Ew\notin E. Then JA=A\mathcal{J}A=A and so A=(fn)=xAA=(f^n)=xA. This leads to the contradiction that xx is a unit in AA and x∈mwx\in\mathfrak{m}_w.

CASE 2: Assume that w∈Ew\in E. Let Q∈Spec⁡(A)\mathfrak{Q}\in\operatorname{Spec}(A) be a minimal prime ideal lying over xAxA. The containment fnA⊂xAf^nA\subset xA implies that Q\mathfrak{Q} represents the generic point of some irreducible component EjE_j. Moving back to the base ring RR, this means xx belongs to the prime ideal pj\mathfrak{p}_j. However, since ww belongs to EjE_j, we should have pj⊆q(w)=q\mathfrak{p}_j\subseteq\mathfrak{q}(w)=\mathfrak{q}. Finally, this leads to the contradiction

x∈pj⊆q(w)=q⊊m;x\in\mathfrak{p}_j\subseteq\mathfrak{q}(w)=\mathfrak{q}\subsetneq\mathfrak{m};

indeed, we would that xx belongs to the prime pj\mathfrak{p}_j that lies in the set S\mathcal{S}.

This concludes the proof of the theorem since we got a contradiction in both cases. □

The next example shows that Theorem 3.8 is sharp.

Example 3.9. The assumption dim⁡(R/p)≥2\dim(R/\mathfrak{p})\geq2 in Theorem 3.8 is necessary. Let kk be an infinite field, R:=k[x,y](x,y)R:=k[x,y]_{(x,y)}, m:=(x,y)\mathfrak{m}:=(x,y), I=(x2,xy)⊂J:=(x,y)2I=(x^2,xy)\subset J:=(x,y)^2 and p:=(x)\mathfrak{p}:=(x). Then

dim⁡(R/p)=1,Ip=xRpandJp=Rp.\dim(R/\mathfrak{p})=1,\qquad I_{\mathfrak{p}}=xR_{\mathfrak{p}}\qquad\text{and}\qquad J_{\mathfrak{p}}=R_{\mathfrak{p}}.

Thus JpJ_{\mathfrak{p}} is not integral over IpI_{\mathfrak{p}}. On the other hand, for a general element ℓ∈m\ell\in\mathfrak{m}, we have the equality

IR‾‾=x2R‾=JR‾‾\overline{I\overline{R}}=x^2\overline{R}=\overline{J\overline{R}}

in R‾=R/ℓR\overline{R}=R/\ell R.

Integral dependence and Specialization

In this section, we give a different proof of the following result that was proved in [17 Theorem 3.3]. The statement of this result is a Principle of Specialization of Integral Dependence (PSID) type of argument [11]. A direct corollary of this result is a proof of the fact that Segre numbers detect integral dependence (see [17 Theorem 4.2]).

Throughout this section, we fix the following data.

Setup 4.1. Let (R,m)(R,\mathfrak{m}) be an equidimensional and universally catenary Noetherian local ring of dimension d≥2d\geq2. Let I⊂RI\subset R be an ideal with height ht⁡(I)≥1\operatorname{ht}(I)\geq1. Let t∈mt\in\mathfrak{m} be an element such that ht⁡(t)=1\operatorname{ht}(t)=1 and ht⁡(I,t)≥2\operatorname{ht}(I,t)\geq2.

Theorem 4.2 ([17 Theorem 3.3]). Assume Setup 4.1 and that

ci(I,R/tR)≤ci(I,R)for all 1≤i≤d−1.c_i(I,R/tR)\leq c_i(I,R)\qquad\text{for all }1\leq i\leq d-1.

Then tt, viewed in degree zero in gr⁡I(R)0=R/I\operatorname{gr}_{I}(R)_0 = R/I, belongs to no minimal prime of gr⁡I(R)\operatorname{gr}_{I}(R).

In fact, we show the following more explicit extension of the above theorem. It shows that Segre numbers are sensitive to the dimension of the projections of the bad components (i.e., those lying over V(t)V(t)) of the exceptional divisor.

Theorem 4.3. Assume Setup 4.1 and that there is a minimal prime of gr⁡I(R)\operatorname{gr}_{I}(R) that contains tt. Let ss be the maximal dimension

s:=max⁡{dim⁡(R/Q∩R)∣Q minimal prime of gr⁡I(R) with t∈Q}s := \max\left\{ \dim(R/Q \cap R) \mid Q \text{ minimal prime of } \operatorname{gr}_{I}(R) \text{ with } t \in Q \right\}

of the contractions to RR of the minimal primes of gr⁡I(R)\operatorname{gr}_{I}(R) containing tt. Let j:=d−1−sj := d-1-s. Then we have the strict inequality

cj(I,R/tR)>cj(I,R).c_j(I,R/tR) > c_j(I,R).

A direct consequence of the above theorem is the following algebraic version of the Principle of Specialization of Integral Dependence (PSID).

Corollary 4.4. Assume Setup 4.1. If ci(I,R)=ci(I,R/tR)c_i(I,R) = c_i(I,R/tR) for all 1≤i≤k1 \leq i \leq k, then

for all p∈Ass⁡R(R/I‾)∩V(t),we have ht⁡(p)≥k+2.\text{for all } \mathfrak{p} \in\operatorname{Ass}_{R}(R/\overline{I}) \cap V(t), \quad\text{we have } \operatorname{ht}(\mathfrak{p}) \geq k+2.

In particular, if ci(I,R)=ci(I,R/tR)c_i(I,R) = c_i(I,R/tR) for all 1≤i≤d−11 \leq i \leq d-1, then

f∈IRt‾if and only iff∈I‾f \in\overline{I R_t} \quad\text{if and only if} \quad f \in\overline{I}

for all f∈Rf \in R.

Proof. By [16 3.9 and 4.1], the associated primes of R/I‾R/\overline{I} come from contractions of minimal primes of gr⁡I(R)\operatorname{gr}_{I}(R) (see also [14 §9.6]). So the claims follow from Theorem 4.3. □

We now proceed to prove Theorem 4.3. Let

X:=Spec⁡(R),Z:=Spec⁡(R/I)andT:=Spec⁡(R/tR).X := \operatorname{Spec}(R), \qquad Z := \operatorname{Spec}(R/I) \qquad\text{and} \qquad T := \operatorname{Spec}(R/tR).

Let

π:B:=BlZ(X)=Proj⁡(R(I))⟶X\pi: B := \mathop{\mathrm{Bl}}\nolimits_Z(X) = \operatorname{Proj}(R(\mathcal{I})) \longrightarrow X

be the blow-up of X=Spec⁡(R)X = \operatorname{Spec}(R) along ZZ, and E:=EZX=Proj⁡(gr⁡I(R))E := E_{ZX} = \operatorname{Proj}(\operatorname{gr}_{I}(R)) be the exceptional divisor. Throughout this section we shall use the fact that gr⁡I(R)\operatorname{gr}_{I}(R) is also equidimensional of dimension dd. By utilizing the faithfully flat extension R→R[z]mR[z]R \to R[z]_{\mathfrak{m}R[z]}, we may assume that the residue field of RR is not an algebraic extension of a finite field (see § 2.3). Let H=H1,…,HdH = H_1,\ldots,H_d be a sequence of general hyperplanes on the blow-up B=BlZ(X)B = \mathop{\mathrm{Bl}}\nolimits_Z(X), and denote by g=g1,…,gdg = g_1,\ldots,g_d the associated sequence of elements in II.

By [6 Theorem C], the ii-th polar scheme is given by

Pi(I,X)=π(H1∩⋯∩Hi)=Spec⁡(R/(g1,…,gi):RI∞)P_i(I,X) = \pi(H_1 \cap\cdots\cap H_i) = \operatorname{Spec}\left(R/(g_1,\ldots,g_i) :_R I^\infty\right)

and the ii-th Segre cycle is given by

Λi(I,X)=π∗([E∩H1∩⋯∩Hi−1]d−i)∈Zd−i(X).\Lambda_i(I,X) = \pi_*\left([E \cap H_1 \cap\cdots\cap H_{i-1}]_{d-i}\right) \in Z_{d-i}(X).

Moreover, we have that

ci(I,R)=ed−i(Λi(I,X))andΛi(I,X)=Λ1(I,Pi−1(I,X)).c_i(I,R) = e_{d-i}\left(\Lambda_i(I,X)\right) \quad\text{and} \quad\Lambda_i(I,X) = \Lambda_1\left(I,P_{i-1}(I,X)\right).

We start with a simple specialization argument with respect to TT.

Lemma 4.5. Assume that no irreducible components of Pi−1(I,X)\mathrm{P}_{i-1}(\mathrm{I},X) is contained in TT. Then we have the equality of cycles

Λi(I,T)=Λ1(I,Spec⁡(S/tS)),\Lambda_i(\mathrm{I},T)=\Lambda_1(\mathrm{I},\operatorname{Spec}(S/tS)),

where S=R/(g1,…,gi−1):RI∞S=R/(g_1,\ldots,g_{i-1}):_R\mathrm{I}^{\infty}.

Proof. Notice that Λi(I,T)=Λ1(I,Pi−1(I,T))\Lambda_i(\mathrm{I},T)=\Lambda_1(\mathrm{I},\mathrm{P}_{i-1}(\mathrm{I},T)) and Λ1(I,Spec⁡(S/tS))=Λ1(I,Spec⁡(S/tS:SI∞))\Lambda_1(\mathrm{I},\operatorname{Spec}(S/tS))=\Lambda_1(\mathrm{I},\operatorname{Spec}(S/tS:_S\mathrm{I}^{\infty})). Since we have the equalities

Pi−1(I,T)=Spec⁡(R/(t,g1,…,gi−1):RI∞)\mathrm{P}_{i-1}(\mathrm{I},T)=\operatorname{Spec}(R/(t,g_1,\ldots,g_{i-1}):_R\mathrm{I}^{\infty})
=Spec⁡(R/((g1,…,gi−1):RI∞,t):RI∞)=Spec⁡(S/tS:SI∞),=\operatorname{Spec}(R/((g_1,\ldots,g_{i-1}):_R\mathrm{I}^{\infty},t):_R\mathrm{I}^{\infty})=\operatorname{Spec}(S/tS:_S\mathrm{I}^{\infty}),

the result of the lemma follows.

One of our main technical tools is the following explicit excess type of argument in the two dimensional case. After cutting with suitably general hyperplanes on the blow-up, the general case will be deduced locally to this one.

Proposition 4.6 (Surface excess). With notation as above, we assume that dim⁡(R)=2\dim(R)=2, (0:RI∞)=0(0:_R\mathrm{I}^{\infty})=0, ht⁡(t)=1\operatorname{ht}(t)=1 and (I,t)(\mathrm{I},t) is m\mathfrak{m}-primary. Then

e(I,R/tR)≥∑q∈V(I)ht⁡(q)=1e(I,Rq)e(t,R/q).e(\mathrm{I},R/tR)\geq\sum_{\substack{\mathfrak{q}\in V(\mathrm{I})\\ \operatorname{ht}(\mathfrak{q})=1}}e(\mathrm{I},R_{\mathfrak{q}})e(t,R/\mathfrak{q}).

If E=EZXE=E_{Z}X has an irreducible component mapping to the closed point m\mathfrak{m}, then the inequality is strict.

Proof. We may assume that ht⁡(I)=1\operatorname{ht}(\mathrm{I})=1 because the statement of the proposition holds trivially when I\mathrm{I} is m\mathfrak{m}-primary.

Choose a general element g∈Ig\in\mathrm{I}, and let H⊂BH\subset B be the corresponding hyperplane on the blow-up. We choose gg so that: gg is a nonzerodivisor on RR, (g,t)(g,t) is m\mathfrak{m}-primary, gg is a reduction of I(R/tR)\mathrm{I}(R/tR), and H⊂BH\subset B contains no irreducible component of EE.

The first Segre cycle is given explicitly by

Λ1(I,X)=π∗([E])=∑q∈V(I)ht⁡(q)=1e(I,Rq)[V(q)]\Lambda_1(\mathrm{I},X)=\pi_*([E])=\sum_{\substack{\mathfrak{q}\in V(\mathrm{I})\\ \operatorname{ht}(\mathfrak{q})=1}}e(\mathrm{I},R_{\mathfrak{q}})[V(\mathfrak{q})]

(see [6 Theorem C(ii)], [17 Proposition 2.4(a)] and [1 Proposition 2.3(iii)]). Let D=V(g)D=V(g) be the effective Cartier divisor on XX corresponding to gg. We have the equality of Cartier divisors

π∗(D)=E+H\pi^*(D)=E+H

on BB (see [6 Proposition 5.8], [10 Lemma 2.4]). Then the projection formula yields the equality of cycles

[D]=π∗([π∗(D)])=π∗([E])+π∗([H])∈Z1(X)[D]=\pi_*([\pi^*(D)])=\pi_*([E])+\pi_*([H])\in Z_1(X)

on $X (see [10 Proposition 1.4], [21 Tag 02RT]). Since (g,t)(g,t) is m\mathfrak{m}-primary, no component of DD is contained in TT. By [9 Corollary 1.2.14], we get

e((g,t),R)=e(t,R/gR)=e(t,[D])e((g,t),R)=e(t,R/gR)=e(t,[D])

and

e(g,R/tR)≥e((g,t),R).e(g,R/tR) \ge e((g,t),R).

Thus, since gg is a reduction of I(R/tR)I(R/tR), we obtain

e(I,R/tR)=e(g,R/tR)≥e(t,[D]).e(I,R/tR) = e(g,R/tR) \ge e(t,[D]).

On the other hand, we have

e(t,π∗([E]))=∑q∈V(I), ht⁡(q)=1e(I,Rq)e(t,R/q).e(t,\pi_*([E])) = \sum_{\substack{q\in V(I),\ \operatorname{ht}(q)=1}} e(I,R_q)e(t,R/q).

By combining the above equations, we obtain

e(I,R/tR)≥e(t,[D])=e(t,π∗([E]))+e(t,π∗([H]))=∑q∈V(I), ht⁡(q)=1e(I,Rq)e(t,R/q)+e(t,π∗([H])).(1)\begin{aligned} e(I,R/tR) &\ge e(t,[D]) \\ &= e(t,\pi_*([E]))+e(t,\pi_*([H])) \\ &= \sum_{\substack{q\in V(I),\ \operatorname{ht}(q)=1}} e(I,R_q)e(t,R/q) + e(t,\pi_*([H])). \tag*{(1)} \end{aligned}

This proves the weak inequality.

Now assume that EE has an irreducible component FF over the closed point m\mathfrak{m}. We show that this assumption implies the nonvanishing H≠∅H \ne\varnothing. Since OF(1)=OB(1)⊗OBOF\mathcal{O}_F(1)=\mathcal{O}_B(1)\otimes_{\mathcal{O}_B}\mathcal{O}_F is ample over Spec⁡(R/m)\operatorname{Spec}(R/\mathfrak{m}), it has positive degree. By assumption HH restricts to a general section of OF(1)\mathcal{O}_F(1), and so we get F∩H≠∅F\cap H \ne\varnothing. In particular, we obtain H≠∅H \ne\varnothing.

Let C⊂HC\subset H be an irreducible component of HH. It necessarily follows that CC is not contained E=B×XSpec⁡(R/I)E=B\times_X\operatorname{Spec}(R/I) because by assumption HH contains no irreducible component of EE. This implies that CC is not contained in the special fiber

B×XSpec⁡(R/m)⊆B×XSpec⁡(R/I)=E.B\times_X\operatorname{Spec}(R/\mathfrak{m})\subseteq B\times_X\operatorname{Spec}(R/I)=E.

Therefore, π∗([C])≠0\pi_*([C])\ne0, and so π∗([H])≠0\pi_*([H])\ne0. Since no component of π∗([H])\pi_*([H]) is contained in TT, we get e(t,π∗([H]))>0e(t,\pi_*([H]))>0. Finally, by (1), the strict inequality follows.

The above proposition already yields the result of Theorem 4.3 in the following case.

Corollary 4.7. With notation as above, we assume (0:RI∞)=0(0:_R I^\infty)=0, ht⁡(t)=1\operatorname{ht}(t)=1 and ht⁡(I,t)≥2\operatorname{ht}(I,t)\ge2. Suppose that EE has an irreducible component FF such that

π(F)⊂T,p0∈Spec⁡(R) with V(p0)=π(F),ht⁡(p0)=2.\pi(F)\subset T,\qquad\mathfrak{p}_0\in\operatorname{Spec}(R)\text{ with }V(\mathfrak{p}_0)=\pi(F),\qquad\operatorname{ht}(\mathfrak{p}_0)=2.

Then c1(I,R/tR)>c1(I,R)c_1(I,R/tR)>c_1(I,R).

Proof. We consider the following Segre cycles

Λ1(I,X)=∑q∈V(I), ht⁡(q)=1e(I,Rq)[V(q)]\Lambda_1(I,X)=\sum_{\substack{q\in V(I),\ \operatorname{ht}(q)=1}}e(I,R_q)[V(q)]

and

Λ1(I,T)=∑p∈V(I,t), ht⁡(p)=2e(I,Rp/tRp)[V(p)].\Lambda_1(I,T)=\sum_{\substack{p\in V(I,t),\ \operatorname{ht}(p)=2}}e(I,R_p/tR_p)[V(p)].

Recall that c1(I,R)=e(Λ1(I,X))c_1(I,R)=e(\Lambda_1(I,X)) and c1(I,R/tR)=e(Λ1(I,T))c_1(I,R/tR)=e(\Lambda_1(I,T)). Since ht⁡(I,t)≥2\operatorname{ht}(I,t)\ge2, no component of Λ1(I,X)\Lambda_1(I,X) is contained in TT. Hence we can consider the cycle

T⋅Λ1(I,X)=∑q∈V(I), ht⁡(q)=1e(I,Rq)∑p∈V(I,t), ht⁡(p)=2e(t,Rp/qRp)[V(p)]=∑p∈V(I,t), ht⁡(p)=2(∑q∈V(I), q⊆p, ht⁡(q)=1e(I,Rq)e(t,Rp/qRp))[V(p)].\begin{aligned} T\cdot\Lambda_1(I,X)=\sum_{\substack{q\in V(I),\ \operatorname{ht}(q)=1}}e(I,R_q)\sum_{\substack{p\in V(I,t),\ \operatorname{ht}(p)=2}}e(t,R_p/qR_p)[V(p)] \\ &=\sum_{\substack{p\in V(I,t),\ \operatorname{ht}(p)=2}}\left(\sum_{\substack{q\in V(I),\ q\subseteq p,\ \operatorname{ht}(q)=1}}e(I,R_q)e(t,R_p/qR_p)\right)[V(p)]. \end{aligned}

Therefore, by applying Proposition 4.6 to each localization RpR_p with p∈V(I,t)p\in V(I,t) and ht⁡(p)=2\operatorname{ht}(p)=2, we obtain

Coeff⁡[V(p)](Λ1(I,T))=e(I,Rp/tRp)≥∑q∈V(I), q⊆p, ht⁡(q)=1e(I,Rq)e(t,Rp/qRp)=Coeff⁡[V(p)](T⋅Λ1(I,X)).\begin{aligned} \operatorname{Coeff}_{[V(p)]}(\Lambda_1(I,T))=e(I,R_p/tR_p) \\ &\ge\sum_{\substack{q\in V(I),\ q\subseteq p,\ \operatorname{ht}(q)=1}}e(I,R_q)e(t,R_p/qR_p) \\ &=\operatorname{Coeff}_{[V(p)]}(T\cdot\Lambda_1(I,X)). \end{aligned}

By assumption, the exceptional divisor Ep0=Proj⁡(gr⁡I(Rp0))E_{p_0}=\operatorname{Proj}(\operatorname{gr}_I(R_{p_0})) has an irreducible component lying over the closed point of Spec⁡(Rp0)\operatorname{Spec}(R_{p_0}), and so Proposition 4.6 yields the strict inequality

Coeff⁡[V(p0)](Λ1(I,T))>Coeff⁡[V(p0)](T⋅Λ1(I,X)).\operatorname{Coeff}_{[V(p_0)]}(\Lambda_1(I,T))>\operatorname{Coeff}_{[V(p_0)]}(T\cdot\Lambda_1(I,X)).

Thus e(Λ1(I,T))>e(T⋅Λ1(I,X))e(\Lambda_1(I,T))>e(T\cdot\Lambda_1(I,X)). By combining this with [9 Corollary 1.2.12], we get

c1(I,R/tR)=e(Λ1(I,T))>e(T⋅Λ1(I,X))≥e(Λ1(I,X))=c1(I,R).c_1(I,R/tR)=e(\Lambda_1(I,T))>e(T\cdot\Lambda_1(I,X))\ge e(\Lambda_1(I,X))=c_1(I,R).

So the proof of the corollary is complete. □

Finally, we are ready for the proof of Theorem 4.3.

Proof of Theorem 4.3. Let QQ be a minimal prime of gr⁡I(R)\operatorname{gr}_I(R) containing tt such that the contraction p0:=Q∩R∈Spec⁡(R)p_0:=Q\cap R\in\operatorname{Spec}(R) has the maximal possible dimension dim⁡(R/p0)=s\dim(R/p_0)=s. Note that QQ corresponds to an irreducible component FF of E=Proj⁡(gr⁡I(R))E=\operatorname{Proj}(\operatorname{gr}_I(R)) because dim⁡(gr⁡I(R)/gr⁡I(R)+)=dim⁡(R/I)<dim⁡(gr⁡I(R))=d\dim(\operatorname{gr}_I(R)/\operatorname{gr}_I(R)_+)=\dim(R/I)<\dim(\operatorname{gr}_I(R))=d. Let j:=d−1−sj:=d-1-s. Since ht⁡(I,t)≥2\operatorname{ht}(I,t)\ge2, we have 0≤s≤d−20\le s\le d-2 and so 1≤j≤d−11\le j\le d-1.

We analyze the jj-th Segre cycle

Λj(I,X)=Λ1(I,Y),\Lambda_j(I,X)=\Lambda_1(I,Y),

where Y:=Pj−1(I,X)=π(H1∩⋯∩Hj−1)=Spec⁡(S)Y:=P_{j-1}(I,X)=\pi\left(H_1\cap\cdots\cap H_{j-1}\right)=\operatorname{Spec}(S) and S:=R/(g1,…,gj−1):RI∞S:=R/(g_1,\ldots,g_{j-1}):_R I^\infty.

We now check the necessary conditions to apply Corollary 4.7 to SS:

(a) Let κ(p0):=Rp0/p0Rp0\kappa(p_0):=R_{p_0}/p_0R_{p_0}. The fact that gr⁡I(R)\operatorname{gr}_I(R) has a minimal prime contracting to p0p_0 is equivalent to saying that Ip0I_{p_0} has maximal analytic spread ℓ(Ip0):=dim⁡(gr⁡I(R)⊗Rκ(p0))=dim⁡(Rp0)=d−s\ell(I_{p_0}):=\dim(\operatorname{gr}_I(R)\otimes_R\kappa(p_0))=\dim(R_{p_0})=d-s. Since we may assume that g1,…,gdg_1,\ldots,g_d are general elements in Ip0I_{p_0} (see [17 Lemma 2.6]), it follows that

dim⁡(gr⁡I(S)⊗Rκ(p0))=d−s−j+1=2anddim⁡(Sp0)=d−s−j+1=2\dim(\operatorname{gr}_I(S)\otimes_R\kappa(p_0))=d-s-j+1=2 \quad\text{and}\quad \dim(S_{p_0})=d-s-j+1=2

(see [20 §8.5] and [6 Remark 5.11]). This implies that p0⊃(g1,…,gj−1):RI∞p_0\supset(g_1,\ldots,g_{j-1}):_R I^\infty, that ht⁡(p0S)=2\operatorname{ht}(p_0S)=2 and that there is a minimal prime of gr⁡I(S)\operatorname{gr}_I(S) lying over p0Sp_0S. We have that SS is equidimensional of dimension dim⁡(S)=d−j+1=s+2\dim(S)=d-j+1=s+2 (see [6 Remark 5.11]). Recall also that t∈p0t\in p_0. (b) By applying Corollary 2.5 successively to the irreducible components of the blow-up BB, we obtain that no component of Y=π(H1∩⋯∩Hj−1)Y = \pi(H_1 \cap\cdots\cap H_{j-1}) is contained in TT, and so ht⁡(tS)≥1\operatorname{ht}(tS) \ge1.

(c) The components of Λj(I,X)=π∗([E∩H1∩⋯∩Hj−1]d−j)\Lambda_j(I,X) = \pi_*([E \cap H_1 \cap\cdots\cap H_{j-1}]_{d-j}) have dimension d−j=s+1d-j = s+1. Let C⊂EC \subset E be an irreducible component of EE. If π(C)\pi(C) is not contained in TT, then by applying Corollary 2.5 successively to CC, we obtain that no component of π(C∩H1∩⋯∩Hj−1)\pi(C \cap H_1 \cap\cdots\cap H_{j-1}) is contained in TT. If π(C)\pi(C) is contained in TT, then dim⁡(π(C∩H1∩⋯∩Hj−1))≤dim⁡(π(C))≤s\dim(\pi(C \cap H_1 \cap\cdots\cap H_{j-1})) \le\dim(\pi(C)) \le s. By definition of pushforward, this implies that no component of Λ1(I,Y)=Λj(I,X)\Lambda_1(I,Y) = \Lambda_j(I,X) is contained in TT. Since

Λ1(I,Y)=∑q∈V(IS)ht⁡(q)=1e(I,Sq)[V(q)],\Lambda_1(I,Y) = \sum_{\substack{q \in V(IS)\\ \operatorname{ht}(q)=1}} e(I,S_q)[V(q)],

it follows that ht⁡(IS,tS)≥2\operatorname{ht}(IS,tS) \ge2.

From Corollary 4.7, we obtain

c1(I,S/tS)>c1(I,S).c_1(I,S/tS) > c_1(I,S).

Finally, the equalities of cycles Λ1(I,Spec⁡(S/tS))=Λj(I,T)\Lambda_1(I,\operatorname{Spec}(S/tS)) = \Lambda_j(I,T) (see Lemma 4.5) and Λ1(I,Y)=Λj(I,X)\Lambda_1(I,Y) = \Lambda_j(I,X) yield the strict inequality

cj(I,R/tR)=c1(I,S/tS)>c1(I,S)=cj(I,R).c_j(I,R/tR) = c_1(I,S/tS) > c_1(I,S) = c_j(I,R).

So the proof of the theorem is complete. □

Acknowledgments

YCR received support from NSF grant DMS-2502321 and Simons Foundation Travel Support for Mathematicians Award MPS-TSM-00013551. SDC received support from NSF grant DMS-2348849.

References

(Cid-Ruiz) DEPARTMENT OF MATHEMATICS, NORTH CAROLINA STATE UNIVERSITY, RALEIGH, NC 27695, USA

Email address: [email protected]

(Cutkosky) DEPARTMENT OF MATHEMATICS, UNIVERSITY OF MISSOURI, COLUMBIA, MO 65211, USA

Email address: [email protected]

References

  1. [1]Rüdiger Achilles and Mirella Manaresi, Multiplicities of a bigraded ring and intersection theory, Math. Ann. 309 (1997), no. 4, 573–591.DOI
  2. [2]Rüdiger Achilles and Sławomir Rams, Intersection numbers, Segre numbers and generalized Samuel multiplicities, Arch. Math. (Basel) 77 (2001), no. 5, 391–398.DOI
  3. [4]Steven Dale Cutkosky and Jonathan Montaño, Multiplicities and degree functions in local rings via intersection products, J. Lond. Math. Soc. 112 (2025), no. 5, e70338.
  4. [5]———, Degree functions of graded families of ideals, Collect. Math. (2026), published online.arxiv.org/abs/2506.05248
  5. [6]Yairon Cid-Ruiz, Claudia Polini, and Bernd Ulrich, Multidegrees, families, and integral dependence, preprint, 2025.
  6. [7]Steven Dale Cutkosky, Multiplicities of graded families of ideals on Noetherian local rings, arXiv:2603.06844, 2026.arxiv.org/abs/2603.06844
  7. [9]Hubert Flenner, Liam O’Carroll, and Wolfgang Vogel, Joins and intersections, Springer Monographs in Mathematics, Springer, Berlin, 1999.
  8. [10]William Fulton, Intersection Theory, second ed., Ergebnisse der Mathematik und ihrer Grenzgebiete (3), vol. 2, Springer-Verlag, Berlin, 1998.
  9. [11]Terence Gaffney and Robert Gassler, Segre numbers and hypersurface singularities, J. Algebraic Geom. 8 (1999), no. 4, 695–736.
  10. [12]Steven L. Kleiman and Anders Thorup, A geometric theory of the Buchsbaum–Rim multiplicity, J. Algebra 167 (1994), no. 1, 168–231.
  11. [13]———, Mixed Buchsbaum–Rim multiplicities, Amer. J. Math. 118 (1996), no. 3, 529–569.
  12. [14]Robert Lazarsfeld, Positivity in algebraic geometry II: Positivity for vector bundles, and multiplier ideals, Ergebnisse der Mathematik und ihrer Grenzgebiete (3), vol. 49, Springer-Verlag, Berlin, 2004.
  13. [15]Hideyuki Matsumura, Commutative ring theory, Cambridge Studies in Advanced Mathematics, vol. 8, Cambridge University Press, Cambridge, 1986, translated from the Japanese by M. Reid.DOI
  14. [16]Stephen McAdam, Asymptotic prime divisors, Lecture Notes in Math., vol. 1023, Springer-Verlag, Berlin, 1983.DOI
  15. [17]Claudia Polini, Ngô Viêt Trung, Bernd Ulrich, and Javid Validashti, Multiplicity sequence and integral dependence, Math. Ann. 378 (2020), 951–969.DOI
  16. [18]David Rees, a-transforms of local rings and a theorem on multiplicities of ideals, Proc. Cambridge Philos. Soc. 57 (1961), no. 1, 8–17.DOI
  17. [19]David Rees, Generalizations of reductions and mixed multiplicities, J. London Math. Soc. (2) 29 (1984), no. 3, 397–414.DOI
  18. [20]Irena Swanson and Craig Huneke, Integral closure of ideals, rings, and modules, London Mathematical Society Lecture Note Series, vol. 336, Cambridge Univ. Press, Cambridge, 2006.
  19. [21]The Stacks project authors, The Stacks project, https://stacks.math.columbia.edu.
  20. [22]Bernard Teissier, Cycles évanescents, sections planes et conditions de Whitney, Singularités à Cargèse, Astérisque, nos. 7–8, Soc. Math. France, Paris, 1973, pp. 285–362.
  21. [23]———, Résolution simultanée II: Résolution simultanée et cycles évanescents, Séminaire sur les singularités des surfaces (M. Demazure, H. C. Pinkham, and B. Teissier, eds.), Lecture Notes in Math., vol. 777, Springer-Verlag, Berlin, 1980, pp. 82–146.DOI
  22. [24]Anders Thorup, Rational equivalence theory on arbitrary Noetherian schemes, Enumerative Geometry (Sitges, 1987), Lecture Notes in Math., vol. 1436, Springer, Berlin, 1990, pp. 256–297.DOI

Paper details

vlnc:W_5pxrchxq
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 0

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

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents