Tannakian duality and Gauss-Manin connections for a family of curves

This paper establishes a short exact sequence relating the differential fundamental groupoids of a smooth family of varieties to its base and relative components, proving that for families of curves of genus at least 1, the resulting maps are isomorphisms that interpret Gauss-Manin connections via group cohomology and demonstrate that the total space becomes a de Rham $K(\pi, 1)$ after shrinking.

The Gist

Imagine you are an explorer trying to understand a vast, shifting landscape. In mathematics, this landscape is a family of curves (think of a bundle of rubber bands that can stretch, twist, and change shape) moving over a base path.

This paper, written by Phung Ho Hai, Vo Quoc Bao, and Tran Phan Quoc Bao, is about building a "universal translator" between two very different languages used to describe this landscape:

1. The Language of Shape (Topology): How the curves are connected and loop around each other.
2. The Language of Flow (Calculus): How things change smoothly as you move along the path.

Here is the story of their discovery, broken down into simple concepts.

1. The Setup: A Moving Train of Curves

Imagine a train (the "base" $S$) traveling along a track. On this train, there are many carriages, and inside each carriage is a complex, looping shape (a "curve" $X$). As the train moves, the shapes inside the carriages might wiggle or change form, but they remain connected.

Mathematicians want to know: If I know how the shapes change inside one carriage, can I predict how they change for the whole train?

2. The Two Languages

To answer this, mathematicians have two main tools:

• The "Group" Language (Fundamental Groups): Think of this as a map of all the possible loops you can draw on the shapes. If you can walk in a circle and return to your start without getting stuck, that's a "loop." The collection of all these loops forms a "group." It tells you about the connectivity of the shape.
• The "Flow" Language (Connections & Cohomology): This is about Gauss-Manin connections. Imagine water flowing over the surface of the shapes. As the train moves, the water level rises and falls. The "Gauss-Manin connection" is the rulebook that tells you exactly how the water level changes as you move from one carriage to the next.

The Big Question: Can we translate the "Flow" rules (how the water changes) directly into the "Group" language (the loops)?

3. The "Inflation" Trick

The authors realized that the shapes on the train are actually just "inflated" versions of the shapes on the track.

• Imagine you have a flat map of the track ($S$).
• You blow it up into a 3D balloon ($X$).
• The "inflation" is the process of turning the flat map into the balloon.

The paper proves that if you take the rules for the flat map and "inflate" them, you get the rules for the balloon. This allows them to link the absolute group (the whole balloon) with the relative group (the balloon compared to the track).

4. The "Fundamental Exact Sequence"

In the world of loops, there is a famous rule called the "Homotopy Exact Sequence." It's like a family tree:

• Child: The loops inside a single carriage (the fiber).
• Parent: The loops of the whole train (the total space).
• Grandparent: The loops of the track itself (the base).

The paper proves that for these specific types of curves (genus $\ge 1$, meaning they have at least one hole, like a donut), this family tree is perfectly balanced. The loops of the carriage, the train, and the track fit together in a perfect, unbroken chain.

5. The Main Discovery: The Universal Translator

The authors' biggest breakthrough is proving that the two languages are actually the same thing for these curves.

They constructed a "dictionary" (mathematically, an isomorphism) that translates:

• From: The "Group Cohomology" (counting the loops and how they interact).
• To: The "De Rham Cohomology" (measuring the flow of water and the Gauss-Manin connection).

The Analogy:
Imagine you have a song.

• Language A describes the song by listing every note and how they harmonize (Group Cohomology).
• Language B describes the song by measuring the sound waves and their frequency (De Rham Cohomology).

The paper proves that for these specific curves, if you know the notes, you automatically know the sound waves, and vice versa. There is no information lost in translation.

6. The "K(π, 1)" Surprise

The paper concludes with a surprising result: After "shrinking" the train (looking at a smaller section of the track), the entire surface becomes a $K(\pi, 1)$ space.

What does that mean?
In simple terms, a $K(\pi, 1)$ space is a shape where everything is determined by its loops.

• If you know the loops (the fundamental group), you know everything about the shape's geometry and calculus.
• It's like saying: "If I tell you how the rubber bands are tied, I have told you everything about the shape of the balloon, including how the wind blows over it."

Summary

This paper is a triumph of Tannakian Duality. It's a fancy way of saying: "We found a perfect mirror between the world of algebraic loops and the world of differential flows."

For families of curves with holes (genus $\ge 1$), the authors showed that the complex rules governing how these shapes change (Gauss-Manin connections) are not mysterious external forces. Instead, they are simply the natural consequence of the shapes' internal loop structures. They proved that the "flow" is just the "loops" speaking in a different accent.

Technical Summary

Here is a detailed technical summary of the paper "Tannakian Duality and Gauss-Manin Connections for a Family of Curves" by Phùng Hô Hai, Võ Quốc Bảo, and Trân Phan Quốc Bảo.

1. Problem Statement and Motivation

The paper addresses a fundamental question in arithmetic geometry and the theory of differential equations: Can the Gauss-Manin connection on the relative de Rham cohomology of a family of curves be described purely in terms of the cohomology of fundamental group schemes?

Specifically, the authors investigate two sub-questions raised by Hélène Esnault:

1. Is the natural map from the group cohomology of the differential fundamental group scheme (with coefficients in a representation) to the de Rham cohomology of the corresponding connection an isomorphism?
2. Is there a well-defined notion of a "relative differential fundamental group" for a relative scheme $X/S$, and how does it relate to the absolute differential fundamental groups of $X$ and $S$?

While these questions are generally open or have negative answers in full generality, the authors focus on a specific, yet rich, setting:

• $S$ is a smooth affine curve over a field $k$ of characteristic 0 (spectrum of a Dedekind domain $R$).
• $f: X \to S$ is a smooth, proper morphism of relative dimension 1 (a family of curves) with genus $g \geq 1$.
• The family admits a section $\eta: S \to X$.

2. Methodology

The authors employ Tannakian Duality as the primary framework to translate geometric problems into representation-theoretic ones.

• Tannakian Categories: They utilize the equivalence between the category of coherent sheaves with flat connections ($MIC$) and the representation categories of affine group schemes (fundamental groups) and groupoid schemes (fundamental groupoids).
• Inflation Functor: A key technical tool is the "inflation" of an absolute connection on $X/k$ to a relative connection on $X/S$. This allows the authors to define a "geometric relative fundamental group" $\pi_{geom}(X/S)$ as the Tannakian dual of the subcategory of relative connections generated by these inflated objects.
• Exact Sequences: The authors construct and prove the exactness of a "fundamental exact sequence" of group schemes:
  $$1 \to \pi_{geom}(X/S) \to \Pi(X/k) \to \Pi(S/k) \to 1$$
  where $\Pi(X/k)$ is the absolute differential fundamental groupoid and $\Pi(S/k)$ is that of the base. This sequence is the differential analogue of Grothendieck's homotopy exact sequence for étale fundamental groups.
• Comparison Maps: They construct natural maps between group cohomology (of $\pi_{geom}(X/S)$) and relative de Rham cohomology ($H^i_{dR}(X/S, V)$).
• Fiber-wise Analysis: To prove isomorphisms, the authors analyze the behavior of these maps on the generic fiber (over the function field $K$) and the closed fibers (over residue fields). They utilize:
  • Universal extension theorems.
  • Poincaré duality for de Rham cohomology.
  • The existence of infinitely many non-isomorphic simple connections.
  • The Lyndon-Hochschild-Serre (LHS) spectral sequence.

3. Key Contributions and Results

A. The Fundamental Exact Sequence

The paper establishes the exactness of the sequence relating the relative and absolute fundamental groupoids:
$$1 \to \pi_{geom}(X/S) \to \Pi(X/k) \xrightarrow{f^*} \Pi(S/k) \to 1$$
Here, $\pi_{geom}(X/S)$ is defined as the Tannakian dual of the category of relative connections of "geometric origin" (sub-quotients of inflated absolute connections). This generalizes previous results by Esnault-Hübner and dos Santos.

B. Isomorphism of Cohomologies (The Main Theorem)

Theorem 1: For a smooth projective relative curve $f: X \to S$ of genus $g \geq 1$ with a section, and for any flat connection $(V, \nabla)$ on $X/k$, the natural map:
$$\nu_i: H^i(\pi_{geom}(X/S), V) \xrightarrow{\sim} H^i_{dR}(X/S, (V, \nabla/S))$$
is an isomorphism for all $i \geq 0$.

• Interpretation: This result confirms that the relative de Rham cohomology is entirely captured by the group cohomology of the relative fundamental group scheme.
• Gauss-Manin Connection: The paper proves that the Gauss-Manin connection on $H^i_{dR}(X/S, V)$ corresponds precisely to the action of the base groupoid $\Pi(S/k)$ on the group cohomology $H^i(\pi_{geom}(X/S), V)$, induced via the Lyndon-Hochschild-Serre spectral sequence associated with the exact sequence above.

C. De Rham $K(\pi, 1)$ Property

Proposition 2: Under the same assumptions, there exists a Zariski open subset $U \subseteq S$ such that the surface $X_U$ (viewed as a scheme over $k$) is a de Rham $K(\pi, 1)$.

• This means that for any flat connection on $X_U$, the map from the group cohomology of the absolute fundamental group $\pi(X_U/k)$ to the absolute de Rham cohomology $H^*_{dR}(X_U/k)$ is an isomorphism.
• This implies that the higher cohomology of the fundamental group scheme vanishes in degrees where the de Rham cohomology vanishes (specifically, for curves of genus $\geq 1$, cohomology vanishes for $i > 2$).

4. Technical Highlights of the Proof

1. Generic Fiber ($K$): The authors show that over the function field $K$, the relative fundamental group $\pi_{geom}(X/S)_K$ behaves like the absolute fundamental group of the generic fiber curve $X_K$. They use the fact that for curves of genus $g \geq 1$, the de Rham cohomology and group cohomology are isomorphic (a result from [BHT25]). They handle the case $g=1$ (abelian fundamental group) and $g \geq 2$ (using Poincaré duality and the existence of non-trivial extensions) separately.
2. Closed Fibers ($s$): They prove that the isomorphism holds for the special fibers by using a "shifting dimension" argument and the fact that the induction functor is exact.
3. Global Isomorphism: By establishing injectivity and bijectivity on both generic and closed fibers, and utilizing the flatness of the modules involved, they deduce the global isomorphism over the Dedekind domain $R$.
4. Universal Extension: The proof for $i=1$ relies heavily on the Universal Extension Theorem, which constructs specific extensions of connections to realize cohomology classes.

5. Significance

• Unification of Perspectives: The paper successfully unifies the geometric perspective (Gauss-Manin connections, de Rham cohomology) with the arithmetic/group-theoretic perspective (fundamental group schemes, group cohomology). It provides a concrete "group-theoretic" interpretation of the Gauss-Manin connection.
• Generalization of Known Results: It extends the known $K(\pi, 1)$ property of curves over fields to families of curves over affine bases, showing that the property is stable under "shrinking" the base.
• Foundation for Further Study: The construction of the "fundamental exact sequence" for differential groupoids and the precise identification of the geometric relative fundamental group provide a robust framework for studying families of varieties in positive characteristic or more general base schemes in future work.
• Resolution of Esnault's Question: It provides an affirmative answer to the specific questions posed by Hélène Esnault in the context of families of curves with sections, demonstrating that the cohomology of the fundamental group scheme is a complete invariant for the relative de Rham cohomology in this setting.

In summary, the paper demonstrates that for families of curves of genus $\geq 1$, the complex geometry of the Gauss-Manin connection is entirely encoded in the representation theory of a specific relative fundamental group scheme, effectively establishing a "Tannakian" description of the Gauss-Manin connection.