🔢 mathematics

Characteristic cycles of constructible sheaves and microlocalization

This paper investigates whether the microlocalization construction used to define characteristic cycles for constructible sheaves in a transcendental setting can be adapted to the algebraic setting of smooth schemes over fields of positive characteristic, proving partial positive results for this extension.

Original authors: Takeshi Saito

Published 2026-07-21

📖 3 min read🧠 Deep dive

CC BY 4.0

Original authors: Takeshi Saito

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to map the hidden "rough spots" on a smooth, invisible landscape. In the world of modern mathematics, specifically a field called algebraic geometry, scientists study shapes defined by equations. Sometimes, these shapes look perfectly smooth, but they hide secret jagged edges or "singularities" that only reveal themselves when you look at them through a special lens. To find these hidden spots, mathematicians use a tool called a "constructible sheaf," which is like a layer of data wrapped around the shape, telling us how the shape behaves at every single point.

To understand where the trouble lies, mathematicians created two special maps. The first is the "singular support," which is like a radar screen showing exactly where the data gets messy or breaks down. The second is the "characteristic cycle," which is a more detailed blueprint that not only shows where the mess is but also counts how "loud" or intense the mess is at each location. In the world of complex numbers and calculus (the "transcendental" world), there is a clever trick called "microlocalization" that acts like a high-powered microscope, allowing researchers to build these maps directly from the data. However, in the world of algebraic geometry over fields with positive characteristic (a specific type of number system used in cryptography and coding theory), this microscope was missing. For a long time, mathematicians weren't sure if they could build this microscope in the algebraic world without breaking the rules of the game.

This paper, written by Takeshi Saito, asks a big question: Can we use this "microlocalization" trick to build our maps in the algebraic world, just like we do in the calculus world? The author tries to construct a new object called the "microlocalization" of a sheaf, which is essentially a Fourier transform (a mathematical way of shifting perspectives, similar to how a prism splits light into colors) of a specialized version of the data. The goal is to see if the "singular support" and "characteristic cycle" derived from this new object match the ones we already know and trust.

The paper finds that the answer is a cautious "yes, but with conditions." The author proves that for simple shapes, specifically curves (one-dimensional lines), the new microlocalization method works perfectly. It produces a map that is identical to the trusted, old-fashioned map. However, for more complex, multi-dimensional shapes, the author cannot prove it works in every single case. Instead, they show that if the new map doesn't get too big (specifically, if its dimension doesn't exceed the dimension of the shape itself), then it will match the old map perfectly. The paper also includes a specific, tricky example involving a "wildly ramified" sheaf (a type of data with very chaotic behavior) to show that the method can be calculated directly, even in difficult scenarios. While the author initially thought the method might fail completely, this research suggests it is a powerful tool that works reliably for simple cases and likely works for complex ones, provided the data doesn't get too unruly.

Technical Summary: Characteristic Cycles of Constructible Sheaves and Microlocalization

Problem Statement
For a constructible sheaf $F$ on a smooth scheme $X$ over a field $k$ of characteristic $p > 0$ , the singular support $SSF$ and the characteristic cycle $CCF$ are well-defined on the cotangent bundle $T^*X$ (based on [2], [16]). However, unlike the theory of $D$ -modules where singular supports and characteristic cycles arise from modules over the ring of microdifferential operators, there is no direct construction of an object on $T^*X$ that generates these invariants in the algebraic setting.

In the transcendental setting, Kashiwara and Schapira [11] constructed the singular support and characteristic cycle using the microlocalization $\mu\text{Hom}(F, F)$ , defined on the cotangent bundle. This paper investigates whether a similar construction using microlocalization works in the algebraic setting of constructible sheaves over fields of positive characteristic. Specifically, the author asks if the support of the microlocalization, denoted $SS^\mu F$ , coincides with the singular support $SSF$, and if the cohomology class $CC^\mu F$ derived from this microlocalization equals the cycle class of the characteristic cycle $CCF$.

Methodology
The paper employs a construction based on the Fourier transform of the specialization functor, following the framework of Verdier [19].

Microlocalization Definition: For an object $F \in D^b_{ctf}(X, \Lambda)$ , the microlocalization $\mu\text{Hom}(F, F)$ is defined on $T^*X$ as the Fourier transform of the specialization $\nu\text{Hom}(F, F)$ on the tangent bundle $TX$. The specialization is obtained by applying the nearby cycles functor to the deformation to the normal bundle of the diagonal immersion $\delta: X \to X \times X$ .
Definitions of Invariants:
- $SS^\mu F$ is defined as the support of $\mu\text{Hom}(F, F)$ in $T^*X$ (Definition 2.1.1).
- $CC^\mu F$ is defined as a cohomology class in $H^0_{SS^\mu F}(T^*X, e^\vee_* K_X)$ , constructed via the adjoint of the evaluation morphism and the microlocalization functor (Definition 2.1.4).
Technical Tools: The analysis relies heavily on the properties of Fourier transforms (Katz–Laumon [12]), the behavior of specialization and nearby cycles under proper push-forward and smooth pull-back, and the compatibility of these functors with the diagonal immersion. The author utilizes the Grothendieck–Ogg–Shafarevich formula and the Milnor formula to compare the newly defined invariants with the established characteristic cycle.

Key Contributions and Results
The paper provides partial positive answers to the question of whether the microlocalization construction recovers the standard singular support and characteristic cycle.

Dimension 1 (Curves): In Proposition 2.5.1, the author proves that if $\dim X = 1$ , then $CC^\mu F$ equals the cycle class of $CCF$. The proof involves explicit computations for tamely ramified sheaves on curves (Corollary 2.2.5) and utilizes a global argument based on the Katz–Gabber theorem [13] to reduce the general case to the tamely ramified case. The equality is verified by comparing the index formula for $CC^\mu$ with the Grothendieck–Ogg–Shafarevich formula.
General Dimension (Conditional): For general $X$ , Proposition 2.5.2 establishes a conditional result: if the inclusion $SS^\mu F \subset SSF$ holds, then $CC^\mu F$ equals the cycle class of $CCF$. The proof reduces the problem to the case of a projective scheme using a Lefschetz pencil and the compatibility of $CC^\mu$ with proper push-forward. It then compares the coefficients of the irreducible components of dimension $\dim X$ using the Milnor formula.
Wild Ramification Example: Section 2.6 provides an explicit, independent computation of $CC^\mu F$ for a wildly ramified Artin–Schreier sheaf on a curve. The result, $CC^\mu F = -\text{cl}([T^*_X X] + 2 \cdot [T^*_0 X])$ , demonstrates that the construction works even in the presence of wild ramification, correcting the author's earlier assumption (made with Ahmed Abbes in 2004) that such a construction would fail.

Significance and Claims
The paper claims to establish a bridge between the transcendental microlocal analysis of Kashiwara–Schapira and the algebraic theory of constructible sheaves in positive characteristic. While the full equivalence $SS^\mu F = SSF$ remains an open question (Question 2.1.9.1), the paper demonstrates that the microlocalization construction yields the correct characteristic cycle in dimension 1 and under the assumption that the microlocal support is contained within the singular support.

The author notes that the construction was previously thought to be non-viable due to the success of alternative approaches using blow-ups (as in [1]). However, the explicit computation in Section 2.6 reveals that the microlocalization approach is viable and yields non-trivial results even for wildly ramified sheaves. The work suggests that the microlocalization functor provides a robust framework for defining characteristic cycles algebraically, potentially offering a unified perspective with the transcendental theory. The paper also notes that similar questions have been considered independently by D. Cisinski, A. Khan, and E. Yang.

More like this