Normal Reduction Number in Local Rings

• Normal Reduction Number is an invariant defining the stage at which the integral closures of ideal powers stabilize under a minimal reduction in a Noetherian local ring.
• It plays a critical role in connecting ideal theory, local cohomology, and resolution geometry, with its stabilization detected via cohomological sequences.
• Various bounds relate the normal reduction number to geometric invariants, such as the geometric and arithmetic genus, thereby informing analyses of singularity types.

The normal reduction number is an invariant of the normal filtration $\{\overline{I^n}\}_{n\ge 0}$ of an $\mathfrak m$-primary ideal $I$ in a Noetherian local ring. It records the first stage at which the integral closures of the powers of $I$ are generated by a minimal reduction in a stable way. In the literature on normal surface singularities, closely related notations coexist: $\overline{r}_J(I)$ for a fixed minimal reduction $J$, $nr(I)$ for the first equality $\overline{I^{n+1}}=Q\overline{I^n}$, and $\overline r(I)$ for the eventual stabilization index $\overline{I^{N+1}}=Q\overline{I^N}$ for all $\mathfrak m$0 (Goel et al., 2019, Okuma et al., 2019, Okuma, 2019). In that setting, the invariant functions as a stabilization bound for powers of integrally closed ideals and links ideal theory, local cohomology, and the geometry of resolutions (Nagy et al., 2021).

1. Definitions and basic variants

For an $\mathfrak m$1-primary ideal $\mathfrak m$2, the normal filtration is the sequence $\mathfrak m$3, where $\mathfrak m$4 denotes the integral closure of $\mathfrak m$5. If $\mathfrak m$6 is a minimal reduction generated by $\mathfrak m$7 elements and satisfying

$\mathfrak m$8

then the smallest such $\mathfrak m$9 is the normal reduction number of $I$0 with respect to $I$1, denoted $I$2 (Goel et al., 2019).

For integrally closed $I$3-primary ideals in normal surface singularities, two finer indices are standard: $I$4

$I$5

where $I$6 is a minimal reduction (Okuma et al., 2019, Okuma, 2019).

Invariant	Definition	Context
$I$7	least $I$8 with $I$9 for all $I$0	normal filtration (Goel et al., 2019)
$I$1	least $I$2 with $I$3	first equality (Okuma et al., 2019)
$I$4	least $I$5 with $I$6 for all $I$7	stable equality (Okuma et al., 2019)
$I$8	$I$9 over integrally closed $\overline{r}_J(I)$0-primary ideals	normal surface singularity (Nagy et al., 2021)

A basic structural point is that, unlike the classic reduction number, $\overline{r}_J(I)$1 and $\overline{r}_J(I)$2 do not depend on the chosen minimal reduction $\overline{r}_J(I)$3 (Okuma, 2019). In the notation of normal surface singularities, one also writes

$\overline{r}_J(I)$4

where $\overline{r}_J(I)$5 denotes the same stabilization index and $\overline{r}_J(I)$6 is described as the universal optimal bound from which powers of certain ideals have stabilization properties (Nagy et al., 2021).

2. Cohomological interpretation on resolutions

For normal surface singularities, the invariant admits a precise cohomological description. By Lipman’s correspondence, an integrally closed $\overline{r}_J(I)$7-primary ideal $\overline{r}_J(I)$8 is represented on a resolution $\overline{r}_J(I)$9 by an anti-nef cycle $J$0, and one studies

$J$1

The sequence $J$2 is non-increasing and stabilizes for large $J$3 (Nagy et al., 2021, Okuma, 2019).

In this language,

$J$4

$J$5

so $J$6 detects stabilization of first differences, while $J$7 detects the first constant value of the cohomology sequence (Okuma, 2019). Equivalently, in the notation of line bundles $J$8,

$J$9

which identifies the normal reduction number with the stabilization step of the $nr(I)$0-sequence (Nagy et al., 2021).

The same cohomological framework extends to base point free line bundles and Abel maps. For a line bundle $nr(I)$1 on a cycle $nr(I)$2, the sequence $nr(I)$3 is non-increasing and eventually constant; its stabilization number $nr(I)$4 is the minimal $nr(I)$5 such that $nr(I)$6 thereafter. For Abel maps $nr(I)$7, the sequence $nr(I)$8 is non-decreasing and stabilizes, with stabilization step $nr(I)$9 (Nagy et al., 2021). This places the normal reduction number inside a broader stabilization theory on resolutions.

3. Local cohomology, Hilbert coefficients, and bounds

In analytically unramified Cohen–Macaulay local rings, the normal reduction number is controlled by graded pieces of local cohomology of the extended Rees algebra of the normal filtration. If

$\overline{I^{n+1}}=Q\overline{I^n}$0

then the normal filtration satisfies the condition $\overline{I^{n+1}}=Q\overline{I^n}$1: $\overline{I^{n+1}}=Q\overline{I^n}$2 If, in addition, the same vanishing holds for all $\overline{I^{n+1}}=Q\overline{I^n}$3, then $\overline{I^{n+1}}=Q\overline{I^n}$4 (Goel et al., 2019).

This local-cohomological control yields necessary and sufficient conditions for the vanishing of normal Hilbert coefficients. In an analytically unramified Cohen–Macaulay local ring of dimension $\overline{I^{n+1}}=Q\overline{I^n}$5, with $\overline{I^{n+1}}=Q\overline{I^n}$6 a parameter ideal and under the stated vanishing and length conditions, one has

$\overline{I^{n+1}}=Q\overline{I^n}$7

and in this case the associated graded ring $\overline{I^{n+1}}=Q\overline{I^n}$8 of the normal filtration is Cohen–Macaulay (Goel et al., 2019). The case $\overline{I^{n+1}}=Q\overline{I^n}$9 yields the formulation

$\overline r(I)$0

under the corresponding hypotheses, which is presented as a generalization of Itoh’s conjectural picture (Goel et al., 2019).

These results show that the normal reduction number is not merely a stabilization index for powers; it is also a threshold governing when the normal filtration satisfies intersection conditions such as $\overline r(I)$1, when local cohomology vanishes in prescribed bidegrees, and when the normal Hilbert polynomial exhibits coefficient vanishing (Goel et al., 2019).

4. Bounds and non-invariance for surface singularities

Several global upper bounds are known for the ring invariant $\overline r(I)$2 or $\overline r(I)$3. For two-dimensional normal local rings,

$\overline r(I)$4

where $\overline r(I)$5 is the geometric genus (Okuma et al., 2019). A stronger numerical constraint is

$\overline r(I)$6

which bounds the normal reduction number from above in terms of geometric genus (Okuma et al., 2018).

More recently, for an excellent two-dimensional normal local ring containing an algebraically closed field,

$\overline r(I)$7

where $\overline r(I)$8 is the arithmetic genus. For almost cone singularities one has the sharper bound

$\overline r(I)$9

and more precisely

$\overline{I^{N+1}}=Q\overline{I^N}$0

with $\overline{I^{N+1}}=Q\overline{I^N}$1, $\overline{I^{N+1}}=Q\overline{I^N}$2 the gonality of the central curve, and $\overline{I^{N+1}}=Q\overline{I^N}$3 (Okuma et al., 15 Dec 2025).

When the link is a rational homology sphere, the singularity invariant satisfies the topological estimate

$\overline{I^{N+1}}=Q\overline{I^N}$4

and the same bound applies to the stabilization indices $\overline{I^{N+1}}=Q\overline{I^N}$5 and $\overline{I^{N+1}}=Q\overline{I^N}$6 attached to line bundles and Abel maps (Nagy et al., 2021). For cone-like singularities with exceptional curve $\overline{I^{N+1}}=Q\overline{I^N}$7 of genus $\overline{I^{N+1}}=Q\overline{I^N}$8 and $\overline{I^{N+1}}=Q\overline{I^N}$9, one has

$\mathfrak m$00

together with a refinement in terms of $\mathfrak m$01 when the representing cycle meets $\mathfrak m$02 negatively (Okuma et al., 2019).

Two recurrent misconceptions are explicitly ruled out by the literature. First, $\mathfrak m$03 is not a combinatorial invariant in general: singularities with the same resolution graph can have different normal reduction numbers (Okuma et al., 15 Dec 2025). Second, the value $\mathfrak m$04 does not characterize elliptic singularities: although $\mathfrak m$05 for elliptic singularities, there are non-elliptic singularities with $\mathfrak m$06 as well (Nagy et al., 2021).

5. Computations and model classes

The invariant is explicitly