---
title: Joint Reduction Number Zero Theorem
url: https://www.emergentmind.com/topics/joint-reduction-number-zero-theorem
type: topic
---

# Joint Reduction Number Zero Theorem

The “Joint-Reduction-Number-Zero Theorem” is not a single theorem with a universally fixed statement, but rather a family of closely related results about joint reductions, mixed multiplicities, and multigraded filtrations. In the literature represented by "A note on joint reductions and mixed multiplicities" [1110.6239], "An extension of Rees theorem and two interpretations of a vector in the joint reduction lattice" [1405.1550], "Local Cohomology of Multi-Rees Algebras with Applications to Joint Reductions and Complete Ideals" [1407.1493], "Local cohomology of multi-Rees algebras, Joint reduction numbers and product of complete ideals" [1601.05615], and "Joint reductions and mixed Buchsbaum-Rim multiplicities of modules and a joint-reduction-number-zero theorem" [2508.07437], the phrase denotes the phenomenon that asymptotic data attached to products or filtrations of ideals—or modules—are already controlled at the first relevant level by a suitable joint reduction. In its most classical form, this means that mixed multiplicities are equal to Hilbert–Samuel multiplicities of parameter ideals generated by joint reductions; in dimension-two and dimension-three Rees-type theorems, it means that a joint reduction identity holds for all positive exponents; in module-theoretic extensions, it means that the product of two integrally closed modules is already generated by the corresponding joint reduction at level zero.

## 1. Basic definitions and the meaning of “number zero”

For ideals \(I_1,\dots,I_s\) in a Noetherian local ring \((A,\mathfrak m)\) and a finitely generated \(A\)-module \(M\), a joint reduction of \((I_1,\dots,I_s)\) of type \((k_1,\dots,k_s)\) is a finite set \(\mathfrak A\) consisting of \(k_i\) elements from \(I_i\) such that, for all large \(n_1,\dots,n_s\),
\[
I_1^{n_1+1}\cdots I_s^{n_s+1}M
=
\sum_{i=1}^s (J_i)\,I_1^{n_1+1}\cdots I_i^{n_i}\cdots I_s^{n_s+1}M,
\]
where \(J_i=(\mathfrak A)\cap I_i\). In the mixed setting one also considers joint reductions of \((I_1,\dots,I_s,J)\) of type \((k_1,\dots,k_s,k_0+1)\), where \(J\) is \(\mathfrak m\)-primary. For multigraded filtrations \(\mathcal F=\{F(\mathbf n)\}\), the analogous condition is
\[
F(\mathbf n)=\sum_i a_i F(\mathbf n-\mathbf e_i)
\quad\text{for all }\mathbf n\gg 0.
\]
In dimension three, for the integral closure filtration \(\{\overline{I^rJ^sK^t}\}\), a good joint reduction \((a,b,c)\) satisfies
\[
\overline{I^rJ^sK^t}
=
a\,\overline{I^{r-1}J^sK^t}
+b\,\overline{I^rJ^{s-1}K^t}
+c\,\overline{I^rJ^sK^{t-1}}
\quad\text{for }r,s,t\gg 0.
\]

The phrase “joint reduction number zero” refers to the strongest possible form of this stabilization. In the dimension-three normal setting, the normal joint reduction number of \(I,J,K\) is zero with respect to \((a,b,c)\) if the displayed equality already holds for all \(r,s,t>0\). In the multigraded formulation of admissible filtrations, the joint reduction number of type \(\mathbf q\) is zero if the defining equality holds at \(n=0\), i.e. from the smallest relevant multidegrees onward. In the ordinary two-ideal setting, \(r(I\mid J)=0\) means that for some joint reduction \((a,b)\),
\[
IJ=aJ+bI.
\]
Thus “number zero” always means immediate stabilization: no additional delay, no higher correction term, and no need to pass further into the filtration before the joint reduction controls it [1110.6239; 1407.1493; 1601.05615].

## 2. Mixed multiplicities as multiplicities of joint reductions

A central source of the terminology is the reinterpretation of mixed multiplicities as Hilbert–Samuel multiplicities of parameter ideals generated by joint reductions. Let
\[
I=I_1\cdots I_s,\qquad q=\dim M/(0_M:I^\infty).
\]
For all large \(n_0,n_1,\dots,n_s\),
\[
\ell_A\!\left(
\frac{J^{n_0}I_1^{n_1}\cdots I_s^{n_s}M}
{J^{n_0+1}I_1^{n_1}\cdots I_s^{n_s}M}
\right)
\]
is a polynomial of total degree \(q-1\), and its top-degree coefficients define the mixed multiplicities
\[
e_A\big(J^{[k_0+1]}\mid I_1^{[k_1]}\mid\cdots\mid I_s^{[k_s]},M\big).
\]
The main theorem of [1110.6239] states that if \(M\) has dimension \(d>0\), \(J\) is \(\mathfrak m\)-primary, \(I=I_1\cdots I_s\), and
\[
\operatorname{ht}\frac{I+\operatorname{Ann}M}{\operatorname{Ann}M}=h>0,
\qquad
k_0+\cdots+k_s=d-1,
\qquad
k_1+\cdots+k_s<h,
\]
then for any joint reduction \(\mathfrak x=\{x_1,\dots,x_d\}\) of \((I_1,\dots,I_s,J)\) of type \((k_1,\dots,k_s,k_0+1)\) that is a system of parameters for \(M\),
\[
e_A\big(J^{[k_0+1]}\mid I_1^{[k_1]}\mid\cdots\mid I_s^{[k_s]},M\big)
=
e_A(\mathfrak x,M).
\]
This is the archetypal joint-reduction-number-zero statement in multiplicity theory: an invariant originally defined from the asymptotic leading term of a multivariable Hilbert polynomial is determined by one parameter ideal generated by a joint reduction.

The same paper shows that Rees’s superficial sequences provide canonical realizations of such joint reductions. If \(x_1,\dots,x_d\) is a Rees’s superficial sequence of \((I_1,\dots,I_s,J)\) of type \((k_1,\dots,k_s,k_0+1)\) that is a system of parameters, then
\[
e_A\big(J^{[k_0+1]}\mid I_1^{[k_1]}\mid\cdots\mid I_s^{[k_s]},M\big)
=
e_A(x_1,\dots,x_d,M).
\]
Corollary 3.6 of [1110.6239] recovers Rees’s original theorem for \(\mathfrak m\)-primary ideals. The paper also makes clear that the height restriction is essential: when \(k_1+\cdots+k_s\ge h\), the theorem can fail because mixed multiplicities may vanish and therefore cannot equal the multiplicity of any system of parameters.

## 3. Dimension-two ordinary powers and the joint reduction lattice

In dimension two, the theorem acquires a sharper and more literal form. For \(\mathfrak m\)-primary ideals \(I,J\) in a two-dimensional Cohen–Macaulay local ring, a joint reduction \((a,b)\) satisfies
\[
I^{r+1}J^{s+1}=aI^rJ^{s+1}+bI^{r+1}J^s
\quad\text{for some and hence for all }r,s\gg0.
\]
The ideals \(I\) and \(J\) have joint reduction number zero, written \(r(I\mid J)=0\), if there exists such a joint reduction with
\[
IJ=aJ+bI.
\]
Equivalently,
\[
\mathcal A(I\mid J)=\mathbb N^2,
\]
where \(\mathcal A(I\mid J)\) is the joint reduction lattice.

The decisive extension from normal powers to ordinary powers was obtained in [1405.1550]. There the first modified homology module
\[
M^1_{r,s}(a^k,b^k)
=
\frac{I^{r+k}J^{s+k}}{a^kI^rJ^{s+k}+b^kI^{r+k}J^s}
\]
measures the failure of the joint reduction identity at the vector \((r,s)\). For \((r,s)=(0,0)\),
\[
M^1_{0,0}(a^k,b^k)=0
\quad\Longleftrightarrow\quad
I^kJ^k=a^kJ^k+b^kI^k.
\]
Theorem 3.25 proves that, for a two-dimensional Cohen–Macaulay local ring,
\[
r(I^k\mid J^k)=0\text{ for all }k\gg0
\]
is equivalent to
\[
e(1,0)=e_1(I),\qquad e(0,1)=e_1(J),\qquad e_2(IJ)=e_2(I)+e_2(J),
\]
and also equivalent to the vanishing of \(M^1_{0,0}(a^k,b^k)\) for all \(k\gg0\) for a joint reduction \((a,b)\). Under the additional assumptions
\[
\depth G(I)\ge1,\qquad \depth G(J)\ge1,
\]
Theorem 3.28 sharpens this to the actual ideals \(I\) and \(J\):
\[
r(I\mid J)=0
\]
if and only if the same three Hilbert-coefficient equalities hold, and if and only if there exists a joint reduction \((a,b)\) such that
\[
M^1_{0,0}(a^k,b^k)=0\quad\text{for all }k\ge1.
\]

This framework also yields a lattice-theoretic generalization. For a fixed vector \((r_0,s_0)\), vanishing of \(M^1_{r_0,s_0}(a^k,b^k)\) for all \(k\) characterizes membership of \((r_0,s_0)\) in the joint reduction lattice, provided certain Rees-superficial type colon conditions hold. The joint-reduction-number-zero theorem is therefore the special case \((r_0,s_0)=(0,0)\) of a broader lattice-theoretic structure [1405.1550].

## 4. Normal filtrations, local cohomology, and the dimension-three Rees-type theorem

For the integral closure filtration in dimension three, the theorem takes a cohomological form. Let \((R,\mathfrak m)\) be a three-dimensional analytically unramified Cohen–Macaulay local ring, and let \(I,J,K\) be \(\mathfrak m\)-primary ideals. For a good joint reduction \((a,b,c)\) of the filtration \(\{\overline{I^rJ^sK^t}\}\), one studies the extended multi-Rees algebra
\[
R'=\bigoplus_{(r,s,t)\in\mathbb Z^3}\overline{I^rJ^sK^t}\,t_1^rt_2^st_3^t
\]
and the ideal
\[
\mathcal Q=(at_1,bt_2,ct_3)\subset R'.
\]
The key local cohomology component is
\[
[H^3_{\mathcal Q}(R')]_{(0,0,0)}.
\]

Theorem 5.3 establishes the formula
\[
\lambda_R\bigl([H^3_{\mathcal Q}(R')]_{(0,0,0)}\bigr)
=
\overline{e}_3(IJK)
-\bigl(\overline{e}_3(IJ)+\overline{e}_3(IK)+\overline{e}_3(JK)\bigr)
+\overline{e}_3(I)+\overline{e}_3(J)+\overline{e}_3(K),
\]
and also shows that this length equals the eventual length of the defect quotient
\[
\frac{\overline{I^rJ^sK^t}}
{a\,\overline{I^{r-1}J^sK^t}
+b\,\overline{I^rJ^{s-1}K^t}
+c\,\overline{I^rJ^sK^{t-1}}}
\quad\text{for }r,s,t\gg0.
\]
Theorem 5.4 then gives the dimension-three joint-reduction-number-zero theorem: the following are equivalent:
\[
[H^3_{\mathcal Q}(R')]_{(0,0,0)}=0;
\]
the normal joint reduction number of \(I,J,K\) is zero with respect to some good joint reduction;
the normal joint reduction number is zero with respect to any good joint reduction; and
\[
\overline{e}_3(IJK)
-\bigl(\overline{e}_3(IJ)+\overline{e}_3(IK)+\overline{e}_3(JK)\bigr)
+\overline{e}_3(I)+\overline{e}_3(J)+\overline{e}_3(K)=0.
\]

The paper explicitly presents this as a generalization, in dimension \(3\), of a theorem of David Rees about joint reductions of the bigraded filtration \(\{\overline{I^rJ^s}\}\). A special case, Theorem 5.5, shows that if
\[
\lambda(R/\overline{I^rJ^sK^t})=P_{I,J,K}(r,s,t)
\quad\text{for all }r+s+t>0,
\]
then normal joint reduction number zero is equivalent to \(\overline e_3(IJK)=0\). The same circle of ideas yields applications to monomial ideals in \(k[x,y,z]\): if \(I,J,K\) are \(\mathfrak m\)-primary monomial ideals and \(I^rJ^sK^t\) is complete whenever \(r+s+t<2\), then all products \(I^rJ^sK^t\) are complete [1407.1493].

## 5. Admissible filtrations, Hyry’s condition, and bounded joint reduction numbers

A further generalization replaces specific ideal powers by arbitrary \(\mathbf I\)-admissible filtrations \(\mathcal F=\{F(\mathbf n)\}_{\mathbf n\in\mathbb Z^s}\). Here the central hypothesis is a high-degree local cohomology vanishing condition for the multi-Rees algebra, called Hyry’s condition:
\[
[H^i_{R_{++}}(M)]_{\mathbf n}=0
\quad\text{for all }i>0\text{ and all }\mathbf n\ge\mathbf m.
\]
If \(\mathcal R(\mathcal F)\) satisfies \(H_{\mathcal R(\mathbf I)}(\mathcal R(\mathcal F),\mathbf0)\), then Theorem 3.11 of [1601.05615] produces a joint reduction \(\mathcal A_{\mathbf q}(\mathcal F)\) of type \(\mathbf q\), where \(|\mathbf q|=d\), such that
\[
F(\mathbf n)=\sum_{i=1}^s\sum_{j=1}^{q_i}a_{ij}F(\mathbf n-\mathbf e_i)
\quad\text{for all }\mathbf n\ge\mathbf q,
\]
and
\[
\mathrm{jr}_{\mathbf q}(\mathcal F)\le \max\{q_i\mid q_i\ge1\}-1.
\]
In particular, if all nonzero \(q_i\) are equal to \(1\), then \(\mathrm{jr}_{\mathbf q}(\mathcal F)=0\).

Example 3.12 gives a concrete case. For
\[
R=k[[X,Y]],\qquad I=(X,Y^2),\qquad J=(X^2,Y),
\]
and the bigraded filtration \(\mathcal F=\{I^rJ^s\}\), the multi-Rees algebra satisfies Hyry’s condition, \(A=\{X,Y\}\) is a joint reduction of type \(\mathbf e=(1,1)\), and
\[
X I^rJ^{s+1}+Y I^{r+1}J^s=I^{r+1}J^{s+1}
\quad\text{for all }r,s\ge0.
\]
Hence
\[
\mathrm{jr}_{\mathbf e}(\mathcal F)=0.
\]

This joint-reduction-number-zero phenomenon is then used to prove completeness theorems for products of complete ideals. If \((R,\mathfrak m)\) is analytically unramified of dimension \(d\ge2\), \(\mathcal R(\mathbf I)\) satisfies Hyry’s condition, and \(I^{\mathbf n}\) is complete for all \(1\le|\mathbf n|\le d-1\), then \(I^{\mathbf n}\) is complete for all \(|\mathbf n|\ge1\). For monomial ideals in \(k[X_1,\dots,X_d]\), the same conclusion holds because \(\mathcal R(\mathbf I)\) is a normal Cohen–Macaulay semigroup ring [1601.05615].

## 6. Modules and mixed Buchsbaum–Rim multiplicities

The theorem also has a module-theoretic extension. Let \((R,\mathfrak m,k)\) be a Noetherian local ring of positive dimension with infinite residue field, and for \(1\le k\le q\) let \(M_k\subseteq F_k\) be a finite-colength submodule of a free module \(F_k\) of rank \(r_k\). A joint reduction of \((M_1,\dots,M_q)\) is a collection \((B_1,\dots,B_q)\) where each \(B_k\subseteq M_k\) is generated by exactly \(r_k\) elements and, for some \(n\),
\[
S_{n+1}(M_1)\cdots S_{n+1}(M_q)
=
\sum_{k=1}^q
S_{n+1}(M_1)\cdots B_k S_n(M_k)\cdots S_{n+1}(M_q)
\]
inside the symmetric algebra \(S(M)\). The smallest such \(n\) is the joint reduction number with respect to \((B_1,\dots,B_q)\).

The paper [2508.07437] proves that this definition is equivalent to a valuative condition and to a determinantal condition: \((B_1,\dots,B_q)\) is a joint reduction of modules if and only if \((\det(B_1),\dots,\det(B_q))\) is a joint reduction, in Rees’s sense, of the maximal-minor ideals \(I(M_1),\dots,I(M_q)\). It also introduces the mixed Buchsbaum–Rim multiplicity
\[
br(M_1|\cdots|M_d),
\]
defined as the top coefficient of a joint Buchsbaum–Rim polynomial, and shows that
\[
br(M_1|\cdots|M_d)=\chi(K_\bullet(\phi_1,\dots,\phi_d))
\]
for the tensor product of \(2\)-term complexes attached to a joint reduction, and moreover
\[
br(M_1|\cdots|M_d)=e(I_1|\cdots|I_d)
\]
for the maximal-minor ideals \(I_k=I(M_k)\).

In the two-dimensional regular local case the paper proves an actual joint-reduction-number-zero theorem for modules. If \((R,\mathfrak m)\) is a two-dimensional regular local ring with infinite residue field, and
\[
M_1\subseteq F_1,\qquad M_2\subseteq F_2
\]
are integrally closed, torsion-free \(R\)-modules of finite colength, then for any joint reduction \((B_1,B_2)\),
\[
M_1M_2=B_1M_2+M_1B_2
\qquad\text{inside }S(M_1\oplus M_2).
\]
Equivalently, the joint reduction number of \((M_1,M_2)\) with respect to \((B_1,B_2)\) is \(0\). The paper gives two proofs: one by quadratic transforms and the structure theory of integrally closed modules, and one via Hoskin–Deligne type length formulas and the identification of mixed Buchsbaum–Rim multiplicity with mixed multiplicity of maximal minors [2508.07437].

## 7. Related viewpoints, limitations, and common misconceptions

A common misconception is that the phrase names a single theorem with a fixed formulation. The cited literature shows instead that it is an umbrella for several precise theorems, depending on whether one works with mixed multiplicities, ordinary powers, integral closures, admissible filtrations, or modules. In [1110.6239] the paper does not introduce or prove a theorem explicitly titled “Joint-Reduction-Number-Zero Theorem”; in [1601.05615] there likewise is not a theorem explicitly named that way, although Theorem 3.11 and Example 3.12 play that role.

A second misconception is that the equality between mixed multiplicity and multiplicity of a joint reduction should hold without restrictions. This is false in the setting of [1110.6239]: the condition
\[
k_1+\cdots+k_s<h
\]
is essential, and Remark 3.5 gives an equimultiple-ideal example showing that Theorem 3.1 fails without it. The later paper [1912.06947] strengthens [1110.6239] by removing the hypothesis that the joint reduction itself be assumed a system of parameters, but it still requires the dimension inequality
\[
\dim M/IM<\dim M-|k|,
\]
and its Remark 3.4 shows that the conclusion can fail when this strict inequality is not satisfied.

A third misconception is that local cohomology hypotheses are necessary whenever joint reduction number zero leads to completeness of products. The examples in [1601.05615] show the contrary: Hyry’s condition is sufficient but not necessary. Example 3.16 and Example 3.17 exhibit situations where powers remain complete or integrally closed even though the expected local cohomology vanishing fails.

There is also a useful one-ideal analogue. Under suitable assumptions, [1210.0067] proves that
\[
J^{n+1}:I^n
\]
is independent of the minimal reduction \(J\) if and only if
\[
r(I)\le \ell(I)-\operatorname{ht}I+n
\]
when either \(R\) is one-dimensional or \(\operatorname{gr}_I(R)\) is Cohen–Macaulay. This is not a joint-reduction theorem in the multigraded sense, but it has the same structural pattern: stabilization of colon ideals detects minimal or near-minimal reduction number.

Finally, the graded asymptotic results of [1804.03382] show that in a standard graded algebra the functions
\[
r(I_1^{a_1}\cdots I_m^{a_m}M)
\quad\text{and}\quad
r(M/I_1^{a_1}\cdots I_m^{a_m}M)
\]
are eventually the maximum of finitely many linear functions in \((a_1,\dots,a_m)\), with coefficients coming from generator degrees. This suggests that a global “number zero” statement for all large multidegrees is generally impossible in that graded setting, except in trivial or degenerate cases. A plausible implication is that joint-reduction-number-zero theorems are inherently sensitive to the ambient filtration, the dimension, and the cohomological or integrality properties built into the hypotheses, rather than being universal across all asymptotic reduction problems.

Source: https://www.emergentmind.com/topics/joint-reduction-number-zero-theorem