Eisenbud–Goto Conjecture

• Eisenbud–Goto Conjecture is a bound relating the Castelnuovo–Mumford regularity of a projective variety to its degree and codimension, capturing key geometric invariants.
• The conjecture has guided research on syzygies and Hilbert functions, with proofs for curves, surfaces, and toric varieties, while also inspiring explicit counterexamples.
• Recent studies identify conditions like 2-very ampleness and mild singularities where the conjecture holds, contrasting with failures in cases with severe singularity profiles.

The Eisenbud–Goto Conjecture is a central prediction in modern algebraic geometry concerning the interplay between projective embedding invariants—degree, codimension, and Castelnuovo–Mumford regularity—of nondegenerate projective varieties. Over four decades, it has shaped research on syzygies, Hilbert functions, and the geometry of projective schemes, inspiring a vast literature spanning positive results, explicit counterexamples, and a broad web of connections to singularity theory, combinatorics, and computational methods.

1. Statement and Invariants

Given a nondegenerate projective variety $X \subset \mathbb{P}^r$ of dimension $n$ and codimension $e = r-n$ over an algebraically closed field of characteristic zero, the Eisenbud–Goto Conjecture asserts the bound

$$\operatorname{reg} X \leq \operatorname{deg} X - \operatorname{codim} X + 1,$$

where $\operatorname{reg} X$ is the Castelnuovo–Mumford regularity of $X$, i.e., the minimal $m$ such that the ideal sheaf $\mathcal{I}_X$ is $m$-regular: $$H^i(\mathbb{P}^r, \mathcal{I}_X(m-i)) = 0 \quad\text{for all } i \geq 1.$$ Here $\operatorname{deg} X$ is the degree of $X$ (intersection number with a general linear subspace of dimension $n$), and codimension $e = r-n$. The conjecture originally arose from a desire to relate the complexity of syzygies (as captured by regularity) to the geometric invariants of $X$ (Han, 22 Jan 2026).

2. Known Cases and Historical Developments

Early evidence for the conjecture came from several key results:

• Curves: The bound holds for projective curves by the work of Gruson, Lazarsfeld, and Peskine (1983).
• Smooth Surfaces: Pinkham and Lazarsfeld (1985–87) established the conjecture for smooth projective surfaces.
• Higher-Dimensional Varieties: Most smooth threefolds in high codimension (Ran, 1990) and weak regularity bounds for smooth fourfolds (Kwak, 1998–99).
• Cohen–Macaulay and Toric Cases: Arithmetically Cohen–Macaulay varieties and projective toric varieties of codimension two (Peeva–Sturmfels, 1998) automatically satisfy the conjecture (Cranford et al., 2021).

Further results address singular varieties: normal surfaces with rational, Gorenstein elliptic, or log canonical singularities (Niu, 2013); normal threefolds with rational singularities or isolated $\mathbb{Q}$-Gorenstein singularities (Han, 22 Jan 2026); and seminormal simplicial affine semigroup rings (Nitsche, 2011).

Despite this evidence, the conjecture was invalidated in full generality only after the explicit construction of counterexamples.

3. Counterexamples and Mechanisms of Failure

Significant counterexamples were first constructed by McCullough and Peeva (2018), employing Rees-like algebras and step-by-step homogenization to produce singular projective varieties with regularity far exceeding the Eisenbud–Goto bound (Mantero et al., 2019). These examples demonstrate that, in general, no polynomial bound of the form $\operatorname{reg} X = O((\operatorname{deg} X)^k)$ universally holds.

Recent advances have shown counterexamples in all dimensions $n\geq 3$ and all codimensions $e\geq 2$, using unprojection techniques and three-generated ideals of large regularity (Choe, 2022). Notably, the counterexamples constructed by Han and Kwak provided the first known failures in dimension $2$—surfaces in $\mathbb{P}^4$—using explicit binomial rational maps (Han et al., 2022).

A summary table of counterexamples appears below.

Reference	Dimension/Codim	Construction method	Regularity Behavior
(Mantero et al., 2019)	$\geq3$, $\geq2$	Rees-like algebras	Reg $\gg$ Degree
(Choe, 2022)	Arbitrary	Unprojection	Polynomial in Degree
(Han et al., 2022)	$2$, $2$	Binomial rational map	Reg $>$ Deg $-$ Codim $+1$

In all known cases, violations are linked to the presence of severe singularities, especially at unprojection points or loci with high tangent-space dimension—suggesting that singularity type is a crucial obstruction.

4. Classes Where the Conjecture Holds

4.1. Projectively Normal 2-Very Ample Varieties with Mild Singularities

Recent progress has unified several strands by showing that the Eisenbud–Goto Conjecture holds for nondegenerate, projectively normal varieties $X \subset \mathbb{P}^r$ that are $2$-very ample and admit only factorial rational singularities with tangent space dimensions bounded by $n+1$ at every point (Han, 22 Jan 2026). Specifically:

• 2-Very Ample: For every length-$3$ $0$-dimensional subscheme $\xi \subset X$, the evaluation map $$H^0(X, \mathcal{O}_X(1)) \to H^0(\xi, \mathcal{O}_X(1)|_\xi)$$ is surjective. Geometrically, there are no trisecant lines.
• Projectively Normal: $$H^0(\mathbb{P}^r, \mathcal{O}(j)) \to H^0(X, \mathcal{O}_X(j))$$ is surjective for all $j \geq 0$.
• Mild singularities: $X$ is normal, Cohen–Macaulay, factorial in codimension one, with rational singularities, and $\dim T_p X \leq n+1$ for all $p \in X$.

Under these hypotheses, every such $X$ satisfies $$\operatorname{reg} X \leq \operatorname{deg} X - \operatorname{codim} X + 1$$. The result recovers all previously established positive cases—smooth, toric codimension $2$, normal $\mathbb{Q}$-Gorenstein isolated singularities, and broad classes of normal varieties with mild singularities (Han, 22 Jan 2026).

4.2. Toric and Semigroup Ring Cases

For projective toric varieties, the conjecture is entirely classified in codimension $2$ (Cranford et al., 2021). In this setting, equality in the bound occurs only for explicit families (including complete intersections and certain parameterized combinatorial classes). For seminormal simplicial semigroup rings, combinatorial decompositions explicitly yield the bound (Nitsche, 2011), and algorithmic techniques verify large new cases (Boehm et al., 2012).

For monomial curves (dimension $1$), Nitsche provided a combinatorial proof of the conjecture exploiting the structure of two-variable monomial ideals and reductions to the behavior of minimal generating sets (Nitsche, 2011).

For square-free monomial ideals arising from Stanley–Reisner rings, an Eisenbud–Goto type upper bound relating regularity, degree, and codimension is established via topological invariants such as the Leray number (Jung et al., 2023).

5. Techniques and Proof Strategies

5.1 Geometric Projections and Vanishing Theorems

The proof for mild singularities (Han, 22 Jan 2026) is representative. After selecting $e-1$ general points and projecting $X$ to $\mathbb{P}^{n+1}$, one analyzes the birational model and its associated divisors. Key ingredients include:

• Singular birational double–point formulas for factorial points,
• Zariski–Fujita theorem to control basepoint freeness and ampleness,
• Kodaira-type vanishing on singular varieties,
• Long-exact cohomology sequences to deduce vanishing for syzygy sheaves.

This method separates the regularity of the structure sheaf $\mathcal{O}_X$ from that of the ideal sheaf and relies crucially on controlling secant lines (via $2$-very ampleness) and mildness of singularities.

5.2 Combinatorial and Algorithmic Methods

For toric and semigroup ring cases, decompositions into direct sums of monomial ideals reduce the regularity calculation to the analysis of squarefree ideals with combinatorial invariants tied to the geometry of the associated polytopes or semigroups (Nitsche, 2011, Nitsche, 2011, Boehm et al., 2012). Various bounds (e.g., the Hoa–Trung bound, reduction numbers) and criteria (seminormality by "box" containment) provide the key estimates.

Algorithmic implementations (e.g., the MonomialAlgebras package for Macaulay2) allow for systematic checks across large families (Boehm et al., 2012).

6. Singularities and the Boundary of the Conjecture

Current evidence indicates that the essential obstructions to the Eisenbud–Goto bound are:

• Presence of singularities more severe than factorial rational (e.g., higher Loewy length, failure of Cohen–Macaulay or rational conditions),
• Failure of secant conditions ($k$-very ampleness).

All known counterexamples exhibit highly singular loci, often at unprojection points, and careful analysis suggests that the regularity of the tangent cone plays a decisive role (Han et al., 2022, Choe, 2022). There is ongoing investigation into whether a refined inequality involving an explicit invariant measuring singularity complexity could bound the regularity, perhaps extending the conjecture in more general singular settings.

7. Open Questions and Future Directions

Major open problems include:

• Extension of the positive results to varieties with non-factorial or non-rational singularities, or allowing secant behavior beyond $2$-very ampleness,
• Determination of whether all counterexamples necessarily fail $k$-very ampleness or exhibit large tangent space dimension,
• Possible existence of prime codimension-$2$ counterexamples defined by two generators,
• Classification of regularity bounds for semicombinatorial or non-simplicial toric and related classes,
• Development of invariants $\sigma(X)$ such that $$\operatorname{reg} X \leq \operatorname{deg} X - \operatorname{codim} X + 1 + \sigma(X)$$, where $\sigma(X)$ measures singularity severity (Choe, 2022).

The interplay of algebraic, combinatorial, and geometric conditions remains an active area of research with implications for computational algebraic geometry, singularity theory, and the classification of projective schemes.

---

For a comprehensive account with details of proofs, counterexamples, and purely combinatorial results, see (Han, 22 Jan 2026, Han et al., 2022, Choe, 2022, Cranford et al., 2021, Nitsche, 2011), and (Niu, 2013).