Descent Theory and Monads

Table of Contents

• #1. Motivation: the Descent Problem|1. Motivation: the Descent Problem
  • #1.1 Gluing in Topology|1.1 Gluing in Topology
  • #1.2 Vector Bundles and the Gluing Condition|1.2 Vector Bundles and the Gluing Condition
  • #1.3 Sheaves as the Prototypical Descent|1.3 Sheaves as the Prototypical Descent
• #2. Grothendieck Descent|2. Grothendieck Descent
  • #2.1 Fibered Categories|2.1 Fibered Categories
  • #2.2 Descent Data and the Cocycle Condition|2.2 Descent Data and the Cocycle Condition
  • #2.3 The Descent Category|2.3 The Descent Category
  • #2.4 Effective Descent Morphisms|2.4 Effective Descent Morphisms
• #3. The Monad Connection|3. The Monad Connection
  • #3.1 The Descent Monad of a Morphism|3.1 The Descent Monad of a Morphism
  • #3.2 Eilenberg-Moore Algebras as Descent Data|3.2 Eilenberg-Moore Algebras as Descent Data
  • #3.3 Beck’s Monadicity Theorem|3.3 Beck’s Monadicity Theorem
  • #3.4 The Benabou-Roubaud Theorem|3.4 The Benabou-Roubaud Theorem
  • #3.5 Faithfully Flat Descent as a Special Case|3.5 Faithfully Flat Descent as a Special Case
• #4. The Beck-Chevalley Condition|4. The Beck-Chevalley Condition
  • #4.1 Base Change and Mates|4.1 Base Change and Mates
  • #4.2 The Condition and its Significance|4.2 The Condition and its Significance
  • #4.3 The Six-Functor Formalism|4.3 The Six-Functor Formalism
• #5. Cohomological Descent|5. Cohomological Descent
  • #5.1 The Cech Nerve of a Cover|5.1 The Cech Nerve of a Cover
  • #5.2 Hypercoverings|5.2 Hypercoverings
  • #5.3 Cohomological Descent|5.3 Cohomological Descent
  • #5.4 The Bar Construction for Monads|5.4 The Bar Construction for Monads
• #6. Galois Theory as Descent|6. Galois Theory as Descent
  • #6.1 Galois Descent for Vector Spaces|6.1 Galois Descent for Vector Spaces
  • #6.2 The Fundamental Theorem via Descent|6.2 The Fundamental Theorem via Descent
  • #6.3 Tannakian Formalism|6.3 Tannakian Formalism
• #7. The Higher-Categorical Perspective|7. The Higher-Categorical Perspective
  • #7.1 Infinity-Categorical Descent|7.1 Infinity-Categorical Descent
  • #7.2 The Barr-Beck-Lurie Theorem|7.2 The Barr-Beck-Lurie Theorem
  • #7.3 Infinity-Stacks and Sheaves|7.3 Infinity-Stacks and Sheaves
• #References|References

1. Motivation: the Descent Problem 🔭

The word descent in mathematics refers to a family of techniques for reconstructing a global object from local data that is already known to cohere. The paradigm is as follows: given a space \(X\) and a cover \(\{U_i \to X\}\), suppose we know the object we want on each \(U_i\). When can we descend this data to a single object on \(X\)?

1.1 Gluing in Topology

The simplest instance is gluing continuous functions. Suppose \(X = U_1 \cup U_2\) is a topological space covered by two open sets, and we have continuous functions \(f_i : U_i \to Y\). We can glue them into a single \(f : X \to Y\) if and only if \(f_1|_{U_1 \cap U_2} = f_2|_{U_1 \cap U_2}\).

This is a descent statement: a function on the cover descends to the base if and only if it satisfies a compatibility condition on overlaps. The key data is: 1. Objects \(f_i\) over each piece \(U_i\). 2. An isomorphism datum (here, equality) over each double overlap \(U_i \cap U_j\). 3. A cocycle condition over triple overlaps \(U_i \cap U_j \cap U_k\) ensuring consistency.

1.2 Vector Bundles and the Gluing Condition

The classical geometric case is vector bundles. Let \(X\) be a topological space with open cover \(\{U_i\}\), and let \(V_i\) be a vector bundle on each \(U_i\). A gluing datum consists of bundle isomorphisms

\[\varphi_{ij} : V_i|_{U_i \cap U_j} \xrightarrow{\sim} V_j|_{U_i \cap U_j}\]

satisfying: - Unit condition: \(\varphi_{ii} = \mathrm{id}_{V_i}\). - Cocycle condition: \(\varphi_{jk} \circ \varphi_{ij} = \varphi_{ik}\) on \(U_i \cap U_j \cap U_k\).

A standard theorem in topology asserts that vector bundles on \(X\) are in bijection (up to isomorphism) with such gluing data.

1.3 Sheaves as the Prototypical Descent

A sheaf on a topological space \(X\) is the cleanest formalization of the descent idea. Recall:

Definition (Sheaf). A presheaf \(\mathcal{F} : \mathrm{Open}(X)^{\mathrm{op}} \to \mathbf{Set}\) is a sheaf if for every open cover \(U = \bigcup_i U_i\), the diagram

\[\mathcal{F}(U) \xrightarrow{\ \prod \rho_i\ } \prod_i \mathcal{F}(U_i) \underset{p_2^*}{\overset{p_1^*}{\rightrightarrows}} \prod_{i,j} \mathcal{F}(U_i \cap U_j)\]

is an equalizer, where \(p_1^*, p_2^*\) are restriction to the two factors.

This says: sections over \(U\) are precisely the compatible families of sections over the \(U_i\). The right-hand equalizer is the beginning of a simplicial diagram — the Cech nerve — which will reappear in Section 5.

2. Grothendieck Descent 🏗️

We now develop the abstract framework for descent along an arbitrary morphism \(f : Y \to X\) in a category \(\mathcal{C}\).

2.1 Fibered Categories

Definition (Fibered Category). Let \(p : \mathcal{E} \to \mathcal{B}\) be a functor. A morphism \(\varphi : e' \to e\) in \(\mathcal{E}\) is cartesian (over \(u : b' \to b = p(e') \to p(e)\)) if for every \(e'' \in \mathcal{E}\) and every morphism \(\psi : e'' \to e\) with \(p(\psi) = u \circ v\) for some \(v : p(e'') \to b'\), there exists a unique \(\chi : e'' \to e'\) with \(p(\chi) = v\) and \(\varphi \circ \chi = \psi\).

In the notation of a diagram, \(\varphi\) is cartesian if the following square is a pullback in \(\mathcal{E}\) “over” the given square in \(\mathcal{B}\):

\usepackage{tikz-cd}
\begin{document}
\begin{tikzcd}
e'' \arrow[r, dashed, "\chi"] \arrow[rr, bend left, "\psi"] & e' \arrow[r, "\varphi"] & e \\
p(e'') \arrow[r, "v"] & b' \arrow[r, "u"] & b
\end{tikzcd}
\end{document}

Definition (Fibration). A functor \(p : \mathcal{E} \to \mathcal{B}\) is a Grothendieck fibration (or fibered category) if for every morphism \(u : b' \to b\) in \(\mathcal{B}\) and every \(e \in \mathcal{E}\) with \(p(e) = b\), there exists a cartesian morphism \(\varphi : e' \to e\) with \(p(\varphi) = u\).

For each \(b \in \mathcal{B}\), the fiber is \(\mathcal{E}_b = p^{-1}(b)\). Given \(u : b' \to b\), the cartesian lifting gives a base-change functor \(u^* : \mathcal{E}_b \to \mathcal{E}_{b'}\), well-defined up to unique isomorphism.

The category of pseudofunctors \(\mathcal{B}^{\mathrm{op}} \to \mathbf{Cat}\) is equivalent to the category of fibrations over \(\mathcal{B}\) (the Grothendieck construction). Thus, a fibration encodes a “family of categories parameterized by \(\mathcal{B}\).”

2.2 Descent Data and the Cocycle Condition

Fix a fibration \(p : \mathcal{E} \to \mathcal{B}\) and a morphism \(f : Y \to X\) in \(\mathcal{B}\). The fiber products \(Y \times_X Y\) and \(Y \times_X Y \times_X Y\) (assumed to exist) give the Cech nerve diagram:

\usepackage{tikz-cd}
\begin{document}
\begin{tikzcd}
Y \times_X Y \times_X Y \arrow[r, shift left=2, "p_{12}"] \arrow[r, "p_{23}"] \arrow[r, shift right=2, "p_{13}"'] & Y \times_X Y \arrow[r, shift left, "\pi_1"] \arrow[r, shift right, "\pi_2"'] & Y \arrow[r, "f"] & X
\end{tikzcd}
\end{document}

Definition (Descent Datum). A descent datum for \(f : Y \to X\) in the fibration \(p : \mathcal{E} \to \mathcal{B}\) is a pair \((E, \varphi)\) where: - \(E \in \mathcal{E}_Y\) is an object over \(Y\). - \(\varphi : \pi_1^* E \xrightarrow{\sim} \pi_2^* E\) is an isomorphism in \(\mathcal{E}_{Y \times_X Y}\), called the descent isomorphism. - The cocycle condition holds: \(p_{23}^*\varphi \circ p_{12}^*\varphi = p_{13}^*\varphi\) as isomorphisms in \(\mathcal{E}_{Y \times_X Y \times_X Y}\).

The cocycle condition can be drawn:

\usepackage{tikz-cd}
\begin{document}
\begin{tikzcd}
p_{12}^*\pi_1^* E \arrow[rr, "p_{12}^*\varphi"] \arrow[dr, "p_{13}^*\varphi"'] & & p_{12}^*\pi_2^* E \arrow[dl, "p_{23}^*\varphi"] \\
& p_{13}^*\pi_2^* E &
\end{tikzcd}
\end{document}

A morphism of descent data \((E, \varphi) \to (E', \varphi')\) is a morphism \(h : E \to E'\) in \(\mathcal{E}_Y\) such that \(\varphi' \circ \pi_1^* h = \pi_2^* h \circ \varphi\).

2.3 The Descent Category

Definition (Descent Category). The descent category \(\mathrm{Des}(f)\) (or \(\mathrm{Des}(p, f)\) when the fibration must be specified) is the category whose objects are descent data \((E, \varphi)\) and whose morphisms are morphisms of descent data.

There is a canonical functor — the comparison functor — from the fiber over \(X\) to the descent category:

\[\Phi_f : \mathcal{E}_X \to \mathrm{Des}(f), \qquad F \mapsto (f^* F, \, \mathrm{can})\]

where \(\mathrm{can} : \pi_1^* f^* F \xrightarrow{\sim} \pi_2^* f^* F\) is the canonical isomorphism \((\pi_1 \circ f)^* F \cong (\pi_2 \circ f)^* F\) coming from \(f \circ \pi_1 = f \circ \pi_2\) (both projections to \(X\) are equal to \(f \circ \pi_i\)).

2.4 Effective Descent Morphisms

Definition (Effective Descent Morphism). A morphism \(f : Y \to X\) is an effective descent morphism for the fibration \(p\) if the comparison functor \(\Phi_f : \mathcal{E}_X \to \mathrm{Des}(f)\) is an equivalence of categories.

Intuitively: \(f\) is an effective descent morphism if every descent datum along \(f\) actually comes from an object on \(X\). The data “glues.”

3. The Monad Connection 🔗

The profound insight, due to Beck (unpublished, 1960s) and formalized by Bénabou–Roubaud (1970), is that descent theory is secretly monadic. Every morphism gives rise to a monad, and descent data are precisely algebras over this monad.

3.1 The Descent Monad of a Morphism

Fix a category \(\mathcal{C}\) with pullbacks and a morphism \(f : Y \to X\). The pullback functor

\[f^* : \mathcal{C}/X \to \mathcal{C}/Y\]

(sending \([E \to X]\) to \([E \times_X Y \to Y]\)) has a right adjoint given by composition with \(f\):

\[f_* : \mathcal{C}/Y \to \mathcal{C}/X, \qquad [V \to Y] \mapsto [V \xrightarrow{} Y \xrightarrow{f} X].\]

This gives an adjunction \(f^* \dashv f_*\) between slice categories:

\usepackage{tikz-cd}
\begin{document}
\begin{tikzcd}
\mathcal{C}/X \arrow[r, bend left, "f^*"] \arrow[r, phantom, "\perp"] & \mathcal{C}/Y \arrow[l, bend left, "f_*"]
\end{tikzcd}
\end{document}

Definition (Descent Monad). The descent monad \(\mathbb{T}_f = (T_f, \eta, \mu)\) on \(\mathcal{C}/X\) is the monad induced by the adjunction \(f^* \dashv f_*\):

\[T_f = f_* \circ f^*, \qquad \eta : \mathrm{Id} \Rightarrow T_f, \qquad \mu : T_f^2 \Rightarrow T_f.\]

Explicitly, for \([E \to X] \in \mathcal{C}/X\):

\[T_f(E) = E \times_X Y \xrightarrow{f \circ \pi_Y} X\]

(the Weil restriction of the base change). The unit \(\eta_E : E \to T_f(E)\) comes from the universal property of the pullback, and the multiplication \(\mu_E : T_f^2(E) \to T_f(E)\) encodes the composition of the two pullbacks.

3.2 Eilenberg-Moore Algebras as Descent Data

Let us unpack what a \(T_f\)-algebra is. See concepts/category-theory/foundations/04-adjoint-functor-theorems-monads|Monads for the general theory.

Claim. The category \(\mathrm{Alg}(T_f)\) of Eilenberg–Moore algebras for \(\mathbb{T}_f\) is equivalent to the descent category \(\mathrm{Des}(f)\).

Proof sketch. An Eilenberg–Moore algebra is a pair \((E \to X, \, \alpha : T_f(E) \to E)\) satisfying unit and associativity axioms. Unwrapping:

\[T_f(E) = f_*(f^*(E)) = [E \times_X Y \xrightarrow{f \circ \pi_Y} X].\]

The structure map \(\alpha\) is a morphism over \(X\), i.e., a map \(E \times_X Y \to E\) over \(X\). By adjunction \((f^* \dashv f_*)\), giving such a map is equivalent to giving a map \(f^*(E) \to f^*(E)\) over \(Y\), i.e., an automorphism \(\varphi : \pi_1^* E \cong f^* E \to f^* E \cong \pi_2^* E\) over \(Y \times_X Y\) (where the identification uses \(\pi_i^* \cong f^*\) after unfolding pullbacks). The unit and associativity axioms for \(\alpha\) translate precisely to: - \(\varphi|_{\Delta} = \mathrm{id}\) (unit, where \(\Delta : Y \to Y \times_X Y\) is the diagonal). - \(p_{23}^* \varphi \circ p_{12}^* \varphi = p_{13}^* \varphi\) (cocycle condition).

These are exactly the axioms for a descent datum \((f^* E, \varphi) \in \mathrm{Des}(f)\). \(\square\)

3.3 Beck’s Monadicity Theorem

The comparison functor \(K : \mathcal{C}/X \to \mathrm{Alg}(T_f)\), given by \(K(E) = (E, \eta_{T_f(E)} \circ \alpha_E)\)… wait, more precisely, the canonical comparison functor is

\[K : \mathcal{C}/Y \to \mathrm{Coalg}(G_f), \qquad V \mapsto (V, \delta_V : V \to G_f(V))\]

using the counit of the adjunction. The question of when \(K\) is an equivalence is answered by Beck’s theorem.

Theorem (Beck’s Monadicity Theorem). Let \(U : \mathcal{D} \to \mathcal{C}\) be a functor with a left adjoint \(F : \mathcal{C} \to \mathcal{D}\). Then \(U\) is monadic — i.e., the comparison functor \(K : \mathcal{D} \to \mathrm{Alg}(U F)\) is an equivalence — if and only if: 1. \(U\) reflects isomorphisms: if \(U(h)\) is an isomorphism then so is \(h\). 2. \(\mathcal{D}\) has coequalizers of \(U\)-split pairs, and \(U\) preserves them.

A parallel \(f \underset{g}{\overset{h}{\rightrightarrows}} e\) in \(\mathcal{D}\) is \(U\)-split if \(U(f), U(g)\) have a split coequalizer in \(\mathcal{C}\), i.e., there exist \(t : Ue \to Uf\) and \(s : \mathrm{coeq}(Uf, Ug) \to Ue\) splitting the coequalizer.

Corollary (Descent via Monadicity). The morphism \(f : Y \to X\) is an effective descent morphism for the fibration of objects in \(\mathcal{C}\) if and only if the pullback functor \(f^* : \mathcal{C}/X \to \mathcal{C}/Y\) is comonadic (equivalently, \(f_* : \mathcal{C}/Y \to \mathcal{C}/X\) is monadic). This is the fundamental equivalence: descent = monadicity of \(f^*\).

3.4 The Benabou-Roubaud Theorem

The Bénabou–Roubaud theorem (1970) makes the relationship between fibered descent and monadic descent precise in the setting of bifibrations.

Definition (Bifibration). A functor \(p : \mathcal{E} \to \mathcal{B}\) is a bifibration if it is both a Grothendieck fibration and a Grothendieck opfibration. That is, every morphism in \(\mathcal{B}\) has both a cartesian lifting (giving the base-change functor \(u^*\)) and a opcartesian lifting (giving the pushforward functor \(u_*\)), yielding adjunctions \(u_! \dashv u^* \dashv u_*\) on each fiber.

Theorem (Bénabou–Roubaud). Let \(p : \mathcal{E} \to \mathcal{B}\) be a bifibration satisfying the Beck-Chevalley condition (Section 4) for a morphism \(a : A_1 \to A_0\) in \(\mathcal{B}\). Then there is an isomorphism of categories

\[\mathrm{Alg}(\mathbb{T}^a) \cong \mathrm{Des}(p, a)\]

where \(\mathbb{T}^a = a^* a_*\) is the monad on \(\mathcal{E}_{A_0}\) induced by the adjunction \(a_! \dashv a^*\) (or \(a^* \dashv a_*\)), and \(\mathrm{Des}(p, a)\) is the descent category of \(a\) in the fibration \(p\).

Combining this with Beck’s monadicity theorem:

Corollary. Under the hypotheses of the Bénabou–Roubaud theorem, \(a : A_1 \to A_0\) is an effective descent morphism if and only if the base-change functor \(a^* : \mathcal{E}_{A_0} \to \mathcal{E}_{A_1}\) is monadic.

3.5 Faithfully Flat Descent as a Special Case

The most important application in algebraic geometry is faithfully flat (or fpqc) descent.

Setup. Let \(A \to B\) be a faithfully flat ring homomorphism (flat plus \(\mathrm{Spec}(B) \to \mathrm{Spec}(A)\) surjective). Consider the fibration \(p : \mathrm{QCoh} \to \mathbf{Sch}^{\mathrm{op}}\) assigning to a scheme its category of quasi-coherent sheaves.

The monad \(\mathbb{T}_{f}\) on \(A\text{-}\mathbf{Mod}\) induced by \(f^*(M) = M \otimes_A B\) is \(T_f(M) = M \otimes_A B\), and a \(T_f\)-algebra structure on \(M \otimes_A B\) is a \(B \otimes_A B\)-module isomorphism

\[\varphi : M \otimes_A B \otimes_{B} (B \otimes_A B) \xrightarrow{\sim} (B \otimes_A B) \otimes_B M \otimes_A B,\]

which, after simplification, is a \(B \otimes_A B\)-module isomorphism \(\varphi : M \otimes_A B \otimes_A B \xrightarrow{\sim} B \otimes_A M \otimes_A B\) satisfying the cocycle condition over \(B \otimes_A B \otimes_A B\). This is precisely the classical descent datum for modules.

Theorem (Grothendieck Faithfully Flat Descent). For a faithfully flat ring map \(A \to B\), the functor

\[A\text{-}\mathbf{Mod} \to \mathrm{Des}(A \to B)\]

is an equivalence. An \(A\)-module \(M\) is completely determined by its base change \(M \otimes_A B\) together with descent data.

Why does monadicity apply? Since \(A \to B\) is faithfully flat: 1. \(f^*(M) = - \otimes_A B\) is conservative (reflects isomorphisms): if \(M \otimes_A B \cong N \otimes_A B\) as \(B\)-modules with compatible descent data, then \(M \cong N\). 2. Coequalizers of \(f^*\)-split pairs exist and are preserved, because \(- \otimes_A B\) is exact (flat) and reflects exactness (faithfully flat). \(\square\)

4. The Beck-Chevalley Condition 🔀

4.1 Base Change and Mates

In a 2-category (or the 2-category \(\mathbf{Cat}\) of categories), given a commutative square of functors and a pair of adjunctions, one can form a mate — a canonical natural transformation going in the opposite direction.

Definition (Mate). Suppose \(f \dashv f^*\) and \(g \dashv g^*\) are adjunctions (with \(f, g\) the left adjoints). Given a natural transformation \(\alpha : g^* \circ h \Rightarrow k \circ f^*\), the mate of \(\alpha\) is the natural transformation

\[\bar{\alpha} : h \circ f \Rightarrow g \circ k\]

constructed as \(\bar{\alpha} = (g \circ \varepsilon_k) \circ (g \circ k \circ \alpha \circ f) \circ (\eta_h \circ f)\) (using the units and counits of the adjunctions).

In the context of a fibration, given a commutative square in the base:

\usepackage{tikz-cd}
\begin{document}
\begin{tikzcd}
B' \arrow[r, "h"] \arrow[d, "k"'] & A' \arrow[d, "f"] \\
B \arrow[r, "g"] & A
\end{tikzcd}
\end{document}

the commutativity \(f \circ h = g \circ k\) induces, via the pullback functors, a canonical base-change natural transformation

\[\chi : k^* g^* \Rightarrow h^* f^*.\]

(Since \(g \circ k = f \circ h\) and pullback is functorial, both sides pull the same composite back to \(B'\), and there is a canonical comparison.)

4.2 The Condition and its Significance

Definition (Beck-Chevalley Condition). A commutative square in the base (as above) satisfies the Beck-Chevalley condition if the canonical natural transformation \(\chi : k^* g^* \Rightarrow h^* f^*\) is an isomorphism (equivalently, its mate \(\bar{\chi} : h_! k^* \Rightarrow f^* g_!\) is an isomorphism, where \(h_!, g_!\) are left adjoints to \(h^*, g^*\) if they exist).

The condition is crucial for two reasons:

1. Monad coherence. It ensures that the “fiber-wise” monads \(\mathbb{T}^f_b = b^* b_*\) on fibers \(\mathcal{E}_b\) assemble into a coherent pseudofunctor. Without Beck-Chevalley, the composition \(b^* \circ b_*\) does not commute with base change in the way needed for the Bénabou–Roubaud theorem.

2. Functoriality of indexed adjoints. If \(p : \mathcal{E} \to \mathcal{B}\) is a fibration and every \(f : b' \to b\) admits both \(f_!\) and \(f_*\), then the families \((f_!)_{f \in \mathcal{B}}\) and \((f_*)_{f \in \mathcal{B}}\) assemble into indexed functors if and only if the Beck-Chevalley condition holds for all pullback squares.

4.3 The Six-Functor Formalism

When a morphism \(f : Y \to X\) admits all three adjoints — \(f_! \dashv f^* \dashv f_*\) — one is in the realm of the six-functor formalism (Grothendieck, later formalized by Scholze):

\[f_!, \, f^*, \, f_*, \, f^!, \, \otimes, \, \mathcal{H}om.\]

The Beck-Chevalley condition then becomes a key compatibility: for a Cartesian square

\usepackage{tikz-cd}
\begin{document}
\begin{tikzcd}
Y' \arrow[r, "h"] \arrow[d, "k"'] & Y \arrow[d, "f"] \\
X' \arrow[r, "g"] & X
\end{tikzcd}
\end{document}

the proper base change theorem states \(g^* f_! \cong k_! h^*\), and the smooth base change theorem states \(g^* f_* \cong k_* h^*\) (under appropriate hypotheses). These are the Beck-Chevalley conditions for \(f_!\) and \(f_*\) respectively.

5. Cohomological Descent 🌐

5.1 The Cech Nerve of a Cover

Given a morphism \(f : Y \to X\), its Cech nerve \(\check{C}(f)_\bullet\) is the simplicial object (see concepts/category-theory/foundations/03-limits-colimits|Limits and Colimits for simplicial machinery):

Definition (Cech Nerve). The Cech nerve \(\check{C}(f)_\bullet\) is the simplicial object in \(\mathcal{C}\) with

\[\check{C}(f)_n = Y \times_X Y \times_X \cdots \times_X Y \quad (n+1 \text{ factors})\]

with face maps \(d_i : \check{C}(f)_n \to \check{C}(f)_{n-1}\) given by omitting the \(i\)-th factor (via the projection), and degeneracy maps \(s_i : \check{C}(f)_n \to \check{C}(f)_{n+1}\) given by repeating the \(i\)-th factor (via the diagonal).

Explicitly: \(\check{C}(f)_0 = Y\), \(\check{C}(f)_1 = Y \times_X Y\), \(\check{C}(f)_2 = Y \times_X Y \times_X Y\), and so on. The augmented Cech nerve includes the map \(f : Y \to X\) as an augmentation \(\check{C}(f)_\bullet \to X\).

5.2 Hypercoverings

The Cech nerve handles “one-level” covers, but for cohomology one sometimes needs hypercoverings, which are more flexible.

Definition (Hypercovering). Let \((\mathcal{C}, \tau)\) be a site (a category with a Grothendieck topology). A hypercovering of \(X \in \mathcal{C}\) is an augmented simplicial object \(U_\bullet \to X\) in \(\mathcal{C}\) such that for each \(n \geq 0\), the canonical map

\[U_n \to (\mathrm{cosk}_{n-1} U_\bullet)_n\]

is a covering morphism (in the topology \(\tau\)), where \(\mathrm{cosk}_{n-1}\) denotes the \((n-1)\)-coskeleton.

Every Cech nerve of a cover is a hypercovering, but hypercoverings are strictly more general: they allow the \(n\)-th level to be a cover of the matching object (the coskeleton) rather than the naive iterated fiber product. Hypercoverings arise naturally when the topology is not generated by single covers (e.g., the etale topology in characteristic \(p\)).

5.3 Cohomological Descent

Now suppose we are in an abelian situation: \(\mathcal{F}\) is a sheaf of abelian groups (or a complex of sheaves) on a site \((\mathcal{C}, \tau)\).

Definition (Cohomological Descent). A morphism \(f : Y \to X\) (or an augmented simplicial object \(U_\bullet \to X\)) is of cohomological descent for a sheaf \(\mathcal{F}\) if the canonical map

\[R\Gamma(X, \mathcal{F}) \xrightarrow{\sim} R\Gamma(\check{C}(f)_\bullet, f^*\mathcal{F}) := R\Gamma(Y, f^*\mathcal{F}) \times^h R\Gamma(Y \times_X Y, \pi^* f^*\mathcal{F}) \times^h \cdots\]

is a quasi-isomorphism. More precisely, there is a descent spectral sequence:

\[E_1^{p,q} = H^q(\check{C}(f)_p, f^*\mathcal{F}) \Rightarrow H^{p+q}(X, \mathcal{F}).\]

For a hypercovering \(U_\bullet \to X\), the correct statement uses the totalization of the cosimplicial cochain complex \(R\Gamma(U_\bullet, \mathcal{F}|_{U_\bullet})\):

\[R\Gamma(X, \mathcal{F}) \xrightarrow{\sim} R\lim_{\Delta} R\Gamma(U_n, \mathcal{F}|_{U_n}).\]

Theorem (Cohomological Descent via Hypercovering). If \(U_\bullet \to X\) is a hypercovering in a site where \(\mathcal{F}\) is a sheaf, then the natural map

\[R\Gamma(X, \mathcal{F}) \xrightarrow{\sim} \mathrm{Tot}\big(R\Gamma(U_\bullet, \mathcal{F}|_{U_\bullet})\big)\]

is a quasi-isomorphism (Conrad, SGA 4).

5.4 The Bar Construction for Monads

The bar construction for the monad \(\mathbb{T}_f\) on \(\mathcal{C}/X\) is the simplicial object

\[B(\mathbb{T}_f, E)_\bullet : \quad \cdots \underset{}{\overset{}{\overset{}{\underset{}{\rightrightarrows\atop\leftarrow}}}} T_f^2(E) \underset{}{\overset{}{\rightarrow\atop\leftarrow}} T_f(E) \to E\]

with face maps \(d_i = T_f^{n-i} \mu T_f^{i-1}\) (applying \(\mu\) in the \(i\)-th position) and degeneracy maps \(s_j = T_f^{n-j} \eta T_f^j\) (applying \(\eta\)).

The normalized bar complex computes the derived functor of the coinvariants:

Claim. The bar construction \(B(\mathbb{T}_f, \mathbb{T}_f, E)_\bullet\) (the two-sided bar construction) is contractible in \(\mathcal{C}/X\) (it has a contracting homotopy given by the extra degeneracy \(T_f(E) \to T_f^2(E)\) from the unit \(\eta\)). This contractibility is the categorical content of the statement that \(T_f\)-algebras form a “resolution” of objects in \(\mathcal{C}/X\).

Under the identification \(T_f(E) = E \times_X Y\), we have \(T_f^n(E) = E \times_X Y^{\times_X n}\), and the bar complex becomes the Cech nerve augmented over \(E\). The bar construction for the monad \(\mathbb{T}_f\) is exactly the Cech nerve of \(f\), and its contractibility is the content of cohomological descent. This is the deepest link between the monad-theoretic and sheaf-theoretic perspectives.

6. Galois Theory as Descent 🔑

6.1 Galois Descent for Vector Spaces

The paradigm example of descent in algebra is Galois descent. Let \(K/k\) be a finite Galois extension with Galois group \(G = \mathrm{Gal}(K/k)\).

Definition (Semilinear Action). A semilinear \(G\)-action on a \(K\)-vector space \(V\) is a group action of \(G\) on the underlying set of \(V\) such that for all \(\sigma \in G\), \(v \in V\), \(\lambda \in K\):

\[\sigma(\lambda \cdot v) = \sigma(\lambda) \cdot \sigma(v),\]

where \(G\) acts on \(K\) via the Galois action.

Definition (Galois Descent Datum). A descent datum for a \(K\)-vector space \(V\) relative to \(K/k\) is a collection of semilinear isomorphisms \(\varphi_\sigma : V \to V\) (one for each \(\sigma \in G\)) satisfying the cocycle condition \(\varphi_{\sigma\tau} = \varphi_\sigma \circ \sigma_*(\varphi_\tau)\), where \(\sigma_*(\varphi_\tau)\) denotes \(\varphi_\tau\) twisted by \(\sigma\).

Equivalently, writing \(\varphi_\sigma(v) = \sigma \cdot v\), a descent datum is a semilinear \(G\)-action on \(V\).

Theorem (Galois Descent). The functor

\[k\text{-}\mathbf{Vect} \to \{K\text{-}\mathbf{Vect} \text{ with semilinear } G\text{-action}\}, \qquad W \mapsto (W \otimes_k K, \text{ natural action})\]

is an equivalence of categories. The inverse is \(V \mapsto V^G = \{v \in V : \sigma(v) = v \, \forall \sigma \in G\}\).

Proof sketch. This is a special case of faithfully flat descent. The extension \(k \hookrightarrow K\) is faithfully flat (it is a finite separable extension). A \(K\)-module with \(B \otimes_A B\)-descent data, where \(B = K\) and \(A = k\), amounts to: for each \(\sigma \in G\), an isomorphism \(\varphi_\sigma : V \otimes_{k,\sigma} K \to V\) satisfying the cocycle condition. Since \(K \otimes_k K \cong \prod_{\sigma \in G} K\) (by the Chinese Remainder Theorem, as \(K/k\) is Galois), descent data decompose into a semilinear \(G\)-action. \(\square\)

6.2 The Fundamental Theorem via Descent

Grothendieck’s formulation of Galois theory reinterprets the fundamental theorem as a descent statement.

Setup. Let \(X = \mathrm{Spec}(k)\) for a field \(k\) with separable closure \(k^s\) and absolute Galois group \(\pi_1(X) = \mathrm{Gal}(k^s/k)\). Consider the fibration \(\mathbf{FEt} \to \mathbf{Sch}\) assigning to a scheme its category of finite etale covers.

Theorem (Grothendieck Galois Theory). There is an equivalence of categories

\[\mathbf{FEt}(X) \xrightarrow{\sim} \pi_1(X, \bar{x})\text{-}\mathbf{FinSets}\]

between finite etale covers of \(X = \mathrm{Spec}(k)\) and finite sets with continuous \(\mathrm{Gal}(k^s/k)\)-action (permutation representations).

The comparison functor sends \(Y \to X\) to the fiber \(Y_{\bar{x}} = Y \times_X \mathrm{Spec}(k^s)\). This fiber functor \(\omega : \mathbf{FEt}(X) \to \mathbf{FinSets}\) is a fiber functor in the sense of Tannakian formalism, and the Galois group is recovered as \(\mathrm{Aut}(\omega)\) — the automorphism group of \(\omega\) as a functor.

The descent perspective: a finite etale \(X\)-scheme \(Y\) descends from \(\mathrm{Spec}(k^s)\) back to \(\mathrm{Spec}(k)\) precisely when the corresponding finite set is equipped with a continuous \(\mathrm{Gal}(k^s/k)\)-action satisfying the Galois descent condition.

6.3 Tannakian Formalism

Tannakian formalism is a vast generalization of Galois descent, where instead of sets with group action one considers linear representations, and the Galois group is replaced by an algebraic group (or group scheme).

Definition (Tannakian Category). A Tannakian category over a field \(k\) is a \(k\)-linear abelian rigid symmetric monoidal category \((\mathcal{T}, \otimes, \mathbf{1})\) such that \(\mathrm{End}(\mathbf{1}) \cong k\), together with an exact faithful tensor functor (a fiber functor) \(\omega : \mathcal{T} \to k\text{-}\mathbf{Vect}\).

Theorem (Tannaka Reconstruction). Let \((\mathcal{T}, \omega)\) be a neutral Tannakian category over \(k\). Define

\[G = \underline{\mathrm{Aut}}^\otimes(\omega) : \mathbf{Alg}_k \to \mathbf{Grp}, \qquad R \mapsto \mathrm{Aut}^\otimes(\omega_R)\]

where \(\omega_R = \omega \otimes_k R\). Then \(G\) is an affine group scheme over \(k\), and there is an equivalence of tensor categories

\[\mathcal{T} \xrightarrow{\sim} \mathrm{Rep}_k(G).\]

Descent perspective. The category \(\mathrm{Rep}_k(G)\) is the descent of the trivial representation over \(k^s\) (where the representation theory is trivial — every representation is a free module) back to \(k\). The fiber functor \(\omega\) plays the role of the “base change to \(k^s\),” and the Galois action is replaced by the \(G\)-action. Tannakian reconstruction is Galois descent for tensor categories.

7. The Higher-Categorical Perspective 🔬

7.1 Infinity-Categorical Descent

In Lurie’s \(\infty\)-categorical framework (Higher Topos Theory, HTT), descent is recast in terms of \(\infty\)-sheaves on \(\infty\)-sites.

Definition (Descent in an Infinity-Category). Let \(\mathcal{X}\) be an \(\infty\)-category and \(U_\bullet \to X\) an augmented simplicial object (thought of as a “cover”). Then descent holds for an object \(\mathcal{F} \in \mathcal{X}\) along this cover if the canonical map

\[\mathcal{F}(X) \to \lim_{\Delta} \mathcal{F}(U_\bullet)\]

is an equivalence in \(\mathcal{X}\) (where the limit is the \(\infty\)-categorical limit over the simplex category \(\Delta^{\mathrm{op}}\)).

Definition (Infinity-Sheaf / Infinity-Stack). An \(\infty\)-presheaf \(\mathcal{F} : \mathcal{C}^{\mathrm{op}} \to \mathcal{S}\) (where \(\mathcal{S}\) is the \(\infty\)-category of spaces) is an \(\infty\)-sheaf on a site \((\mathcal{C}, \tau)\) if for every cover \(U_\bullet \to X\), the descent condition holds.

The \(\infty\)-categorical formulation is strictly stronger than the classical one: it requires descent for all homotopy limits, not just the \(\pi_0\) truncation. This is why the \(\infty\)-categorical theory handles stacks, higher stacks, and derived stacks uniformly.

7.2 The Barr-Beck-Lurie Theorem

The \(\infty\)-categorical generalization of Beck’s monadicity theorem is due to Lurie (HTT, §4.7 and HA, §4.7).

Definition (Monadic Functor, Infinity-Categorical). A functor \(G : \mathcal{D} \to \mathcal{C}\) between \(\infty\)-categories is monadic if: 1. \(G\) has a left adjoint \(F : \mathcal{C} \to \mathcal{D}\) (with respect to concepts/category-theory/foundations/02-adjoints-representables|Adjoints in the \(\infty\)-categorical sense). 2. The comparison functor \(K : \mathcal{D} \to \mathrm{Alg}_{\mathbb{T}}(\mathcal{C})\) to the \(\infty\)-category of algebras over the induced monad \(\mathbb{T} = G \circ F\) is an equivalence.

Theorem (Barr-Beck-Lurie). Let \(G : \mathcal{D} \to \mathcal{C}\) be a functor of \(\infty\)-categories admitting a left adjoint \(F\). Then \(G\) is monadic if and only if: 1. \(G\) is conservative: \(G(f)\) is an equivalence implies \(f\) is an equivalence. 2. \(\mathcal{D}\) admits geometric realizations (i.e., \(|\Delta^{\mathrm{op}}|\)-shaped colimits) of \(G\)-split simplicial objects, and \(G\) preserves these geometric realizations.

Condition (2) is the \(\infty\)-categorical replacement of “has and preserves coequalizers of \(U\)-split pairs” — in the \(\infty\)-world, coequalizers are replaced by geometric realizations (colimits of simplicial diagrams), because the higher coherences built into an \(\infty\)-algebra are precisely captured by a resolution by a simplicial object.

Corollary (Infinity-Categorical Descent). For a map \(f : Y \to X\) in an \(\infty\)-topos \(\mathcal{X}\), the pullback functor \(f^* : \mathcal{X}_{/X} \to \mathcal{X}_{/Y}\) is comonadic if and only if \(f\) is an effective descent morphism in the \(\infty\)-categorical sense.

7.3 Infinity-Stacks and Sheaves

An \(\infty\)-topos (Lurie, HTT Chapter 6) is an \(\infty\)-category \(\mathcal{X}\) satisfying: 1. \(\mathcal{X}\) has all small colimits and small limits. 2. Colimits are universal (stable under base change). 3. Descent: for every simplicial object \(U_\bullet\) in \(\mathcal{X}\) with colimit \(|U_\bullet| = X\), the canonical functor \(\mathcal{X}_{/X} \to \lim_\Delta \mathcal{X}_{/U_\bullet}\) is an equivalence.

Condition (3) is literally the \(\infty\)-categorical descent condition built into the axioms of an \(\infty\)-topos. The descent axiom ensures that \(\mathcal{X}\) “looks like” the \(\infty\)-category of \(\infty\)-sheaves on some \(\infty\)-site.

Theorem (Lurie, HTT 6.1.3). An \(\infty\)-category \(\mathcal{X}\) is an \(\infty\)-topos if and only if it is equivalent to the \(\infty\)-category of \(\infty\)-sheaves \(\mathrm{Sh}_\infty(\mathcal{C}, \tau)\) on some \(\infty\)-site \((\mathcal{C}, \tau)\).

References