Syzygy Theoretic Approach to Horrocks-type Criteria for Vector Bundles

Technical Summary: Syzygy Theoretic Approach to Horrocks-type Criteria for Vector Bundles

Problem Statement
The paper addresses the structural classification of vector bundles on projective space $\mathbb{P}^n$, specifically focusing on those with non-vanishing intermediate cohomologies. While Horrocks' celebrated theorem establishes that vector bundles with vanishing intermediate cohomologies (ACM bundles) split into direct sums of line bundles, and Chang-Goto theorems characterize Buchsbaum bundles as direct sums of twisted sheaves of differential forms, the structure of quasi-Buchsbaum bundles remains less understood. A quasi-Buchsbaum bundle $E$ is defined by the condition $\mathfrak{m}H^i_*(E) = 0$ for $1 \le i \le n-1$, where $\mathfrak{m}$ is the maximal ideal of the polynomial ring $S$. The paper seeks to clarify the structure of these bundles, particularly those with "simple" intermediate cohomologies (e.g., $\dim_k H^i_*(E) = 1$ for specific $i$), and to extend Horrocks-type criteria to this broader class.

Methodology
The author employs a syzygy theoretic approach combined with the theory of standard systems of parameters and spectral sequences. The core methodology involves:

1. Cohomological Analysis: Investigating the behavior of intermediate cohomology modules $H^i_*(E)$ under the action of linear forms (system of parameters).
2. Spectral Sequences: Utilizing double complexes formed by the tensor product of a Koszul complex (associated with the system of parameters) and a Čech resolution of the vector bundle. This yields spectral sequences that relate Koszul cohomologies to local cohomologies ($H^i_{\mathfrak{m}}$).
3. Skew-Symmetric Matrices: Constructing skew-symmetric matrices from the maps induced by the wedge product of linear forms on cohomology groups. Specifically, for a bundle $E$ on $\mathbb{P}^3$ with $\dim_k H^1_*(E) = \dim_k H^2_*(E) = 1$, a map $\Phi: H^2_*(E(-3)) \to H^1_*(E(-1))^{\oplus 6}$ is constructed, defining a $4 \times 4$ skew-symmetric matrix $A$.
4. Free Resolutions: Constructing explicit free resolutions for vector bundles arising from non-vanishing cohomologies, often using mapping cones of exact sequences involving sheaves of differential forms (e.g., $\Omega^p_{\mathbb{P}^n}$).

Key Contributions and Results

• Characterization of Null-Correlation Bundles:
  The paper provides a new characterization of null-correlation bundles on $\mathbb{P}^n$ (where $n$ is odd). These bundles are shown to be indecomposable quasi-Buchsbaum bundles with $\dim_k H^1_*(E) = \dim_k H^{n-1}_*(E) = 1$. The structure is linked to a skew-symmetric matrix of full rank ($n+1$).

• Classification on $\mathbb{P}^3$:
  For indecomposable quasi-Buchsbaum bundles on $\mathbb{P}^3$ with $\dim_k H^1_*(E) = \dim_k H^2_*(E) = 1$, the paper establishes a dichotomy based on the rank of the associated skew-symmetric matrix $A$:

  • Nonstandard-Buchsbaum: If $\text{rank}(A) = 4$, the bundle is isomorphic to a null-correlation bundle (twisted).
  • Pseudo-Buchsbaum: If $\text{rank}(A) = 2$, the bundle is classified into three distinct types based on the structure of the matrix and its behavior under hyperplane sections. These bundles are isomorphic to the cokernel of a map $\psi: \mathcal{O}_{\mathbb{P}^3} \to \Omega^1_{\mathbb{P}^3}(2) \oplus \mathcal{O}_{\mathbb{P}^3}(1)^{\oplus 2}$.

• Standard Systems of Parameters:
  The paper introduces a refined notion of "standard ideals" for quasi-Buchsbaum modules. It demonstrates that the property of a bundle being Buchsbaum, pseudo-Buchsbaum, or nonstandard-Buchsbaum corresponds to whether the maximal ideal $\mathfrak{m}$ (or specific sub-ideals generated by linear forms) acts as a standard system of parameters. Specifically, a bundle is Buchsbaum if and only if $\mathfrak{m}$ is a standard ideal; otherwise, the existence of smaller standard ideals determines the specific type of quasi-Buchsbaum structure.

• Generalization to $\mathbb{P}^n$:
  The approach is extended to $\mathbb{P}^n$ for bundles with non-vanishing cohomologies only in degrees $j_1$ and $j_2$. The paper constructs a vector bundle $N_{12}$ (dual to the cokernel of an injective map $\mathcal{O}_{\mathbb{P}^n}(-1) \to \Omega^{n-2}_{\mathbb{P}^n}(n-2)$) and posits that the existence of a "nonstandard element" in the cohomology implies the bundle contains $N_{12}$ as a direct summand.

• Multiprojective Spaces:
  The paper briefly addresses vector bundles on multiprojective spaces ($\mathbb{P}^m \times \mathbb{P}^n$). It proves that no vector bundle exists on such spaces with vanishing intermediate cohomologies for all twists, except for trivial cases, by utilizing Castelnuovo-Mumford regularity and Serre duality. It also provides a cohomological criterion for ACM bundles on smooth quadric surfaces ($\mathbb{P}^1 \times \mathbb{P}^1$).

Significance and Claims
The author claims that this work provides "new sights" into the investigation of quasi-Buchsbaum bundles by shifting the focus from purely geometric restriction theorems to an algebraic syzygy-theoretic framework. The primary significance lies in:

1. Unifying Structure: Connecting the geometric concept of null-correlation bundles with the algebraic properties of skew-symmetric matrices and standard systems of parameters.
2. Classification: Offering a complete classification for a specific, non-trivial class of indecomposable quasi-Buchsbaum bundles on $\mathbb{P}^3$ (those with simple intermediate cohomologies), distinguishing them via the rank of associated matrices.
3. Methodological Extension: Demonstrating that the behavior of system of parameters (specifically the distinction between standard and nonstandard elements) determines the hyperplane section properties and the splitting behavior of the bundles.

The paper concludes by noting that while the results are established for specific cases (like $\mathbb{P}^3$ and specific cohomological dimensions), the syzygy theoretic method offers a pathway to extend these classifications to higher dimensions and different cohomological patterns, as suggested by the open questions regarding $j_1, j_2$ in $\mathbb{P}^n$.