[Paper Review] Introduction to a theory of b-functions

Morihiko Saito|ArXiv.org|Oct 26, 2006

Commutative Algebra and Its Applications36 references3 citations

TL;DR

This paper provides a comprehensive introduction to the theory of b-functions (Bernstein-Sato polynomials) in the context of D-modules and holonomic systems, establishing connections to singularities, multiplier ideals, and hyperplane arrangements. It derives explicit formulas for b-functions of monomial ideals and central hyperplane arrangements, including cases with non-isolated singularities, and proves that the b-function for a generic central arrangement of degree d is $ b_f(s) = (s+1)^{n-1} \prod_{j=n}^{2d-2} (s + \frac{j}{d}) $, with precise conditions for roots and jumping numbers.

ABSTRACT

We give an introduction to a theory of b-functions, i.e. Bernstein-Sato polynomials. After reviewing some facts from D-modules, we introduce b-functions including the one for arbitrary ideals of the structure sheaf. We explain the relation with singularities, multiplier ideals, etc., and calculate the b-functions of monomial ideals and also of hyperplane arrangements in certain cases.

Motivation & Objective

• To develop a foundational theory of b-functions (Bernstein-Sato polynomials) within the framework of D-modules and holonomic systems.
• To clarify the relationship between b-functions, singularities, multiplier ideals, and vanishing cycles in complex algebraic geometry.
• To compute b-functions explicitly for monomial ideals and central hyperplane arrangements, especially in cases with non-isolated singularities.
• To determine the set of roots of the b-function and relate them to jumping numbers and cohomological invariants such as $ \chi(U) $ and $ h^i(F_0, \mathbb{C})_\lambda $.
• To establish criteria for when $ \alpha \in R_f $, the set of roots of the b-function, using combinatorial and cohomological conditions.

Proposed method

• Utilizes the theory of filtered D-modules, with a focus on good filtrations and characteristic varieties to define holonomic modules.
• Applies the de Rham functor to relate D-modules to perverse sheaves, leveraging Kashiwara’s theorem on holonomic D-modules.
• Employs the Bernstein-Sato functional equation $ b_f(s) f^s \in \mathcal{D}_X[s] f^s $ to define the b-function and analyze its roots.
• Uses the structure of the logarithmic de Rham complex and the Gauss-Manin connection to compute cohomology groups $ H^{n-1}(\mathcal{A}^\bullet_\alpha) $.
• Applies Walther’s method and duality to compute b-functions for generic central arrangements, reducing the problem to combinatorial conditions on edges and multiplicities.
• Introduces the subspace $ V(I) \subset H^{n-1}(\mathcal{A}^\bullet_\alpha) $ generated by wedge products of forms indexed by subsets $ I $, to test non-vanishing and determine root inclusion.

Experimental results

Research questions

• RQ1 What is the structure of the b-function for monomial ideals, and how does it relate to the singularities of the defining variety?
• RQ2 How do the roots of the b-function relate to jumping numbers and multiplier ideals in the context of hyperplane arrangements?
• RQ3 What conditions determine whether a rational number $ \alpha = k/d $ is a root of the b-function for a central hyperplane arrangement?
• RQ4 How does the cohomology of the Milnor fiber and the Euler characteristic $ \chi(U) $ influence the set of roots $ R_f $?
• RQ5 Under what conditions does the b-function of a central arrangement have a specific factorization, such as $ (s+1)^{n-1} \prod_{j=n}^{2d-2} (s + \frac{j}{d}) $?

Key findings

• For a generic central hyperplane arrangement of degree $ d $, the b-function is $ b_f(s) = (s+1)^{n-1} \prod_{j=n}^{2d-2} (s + \frac{j}{d}) $, with the product ranging from $ j=n $ to $ 2d-2 $.
• The root $ \alpha = 8/5 $ is not in $ R_f $ for the arrangement $ (x^2-1)(y^2-1)=0 $ in $ \mathbb{C}^2 $, despite $ \chi(U)=1 $, due to cohomological conditions and the criterion in (d).
• For $ (x^2-1)(y^2-1)(x+y)=0 $, $ 10/6 \notin R_f $, and this is confirmed by the non-vanishing of $ V(I) $ and the condition in (f), where $ I^c $ corresponds to $ (x+1)(y+1)=0 $.
• In the case $ (x^2-y^2)(x^2-1)(y^2-1)=0 $ with $ d=7 $, $ 12/7 \notin R_f $, and this is explained by the vanishing of $ V(I) $ for $ I^c $ corresponding to $ (x+1)(y+1)=0 $, satisfying condition (f).
• For $ n=3 $, $ \text{mult}_z Z \leq 3 $, $ d \leq 7 $, and $ \nu_3 \neq 0 $, the b-function is $ b_f(s) = (s+1) \prod_{i=2}^4 (s + \frac{i}{3}) \prod_{j=3}^r (s + \frac{j}{d}) $, with $ r=2d-2 $ if $ \nu_3 < d-3 $, and $ r=2d-3 $ otherwise.
• The root $ \alpha = 5/7 $ is in $ R_f $ for $ d=7 $, but $ 12/7 \notin R_f $, and this is consistent with the criterion (e) and (f), where $ \alpha \geq \alpha'_f $ and $ V(I) \neq 0 $, but $ V(I) = H^{n-1}(\mathcal{A}^\bullet_\alpha) $, so $ \alpha+1 \notin R_f $.