Arithmetic Geometry and Etale Cohomology

⚡ Key Takeaways

Étale cohomology is a cohomology theory for schemes that mimics singular cohomology of topological spaces by replacing open coverings with étale ("finite, unramified, flat") maps. It supplies the ℓ-adic cohomology groups $H^i_{\text{ét}}(X, \mathbb{Q}_\ell)$ that serve as the long-sought Weil cohomology over arbitrary fields, providing the framework in which the Weil conjectures were proved (Grothendieck, Deligne) and counting points on varieties over finite fields. Developed by Grothendieck, Artin, and Verdier in the SGA seminars (1962–1973).

💡 The Core Insight

Cohomology is governed not by the points of a space but by the lattice of coverings one allows. The Zariski topology of a scheme has too few opens to see anything; ordinary topology is unavailable over $\mathbb{F}_q$ . Grothendieck's resolution is to abandon topological spaces altogether and define cohomology relative to a site — a category of "covers" satisfying gluing axioms. The étale site uses étale morphisms (the algebraic analogue of local homeomorphisms) as its covers. With finite or $\ell$ -adic coefficients ( $\ell \neq \text{char}$ ), the resulting cohomology is finite-dimensional, satisfies all Weil-cohomology axioms, and carries a canonical Galois action linking geometry to arithmetic.

DefinitionÉtale Morphism

A morphism of schemes $f: X \to Y$ is étale if it is flat and unramified; equivalently, $f$ is smooth of relative dimension $0$ ; equivalently, $f$ is locally of finite presentation and induces isomorphisms on completed local rings $\widehat{\mathcal{O}}_{Y, f(x)} \to \widehat{\mathcal{O}}_{X,x}$ (when residue fields agree). Concretely, étale maps are "open immersions up to finite covering": locally

1

\quad X = \operatorname{Spec}\!\Big( B[T]/(g(T)) \Big)_{g'(T)}, \qquad B = \mathcal{O}_Y(U),

with $g$ a monic polynomial whose derivative $g'$ is invertible on $X$ . Derivation hint: invertibility of $g'$ encodes the Jacobian criterion — the differential is an isomorphism, the algebraic shadow of the inverse function theorem.

DefinitionThe Étale Site and Étale Sheaves

The small étale site $X_{\text{ét}}$ of a scheme $X$ has as objects the étale $X$ -schemes $U \to X$ , morphisms the $X$ -morphisms (automatically étale), and coverings the families $\{U_i \to U\}$ that are jointly surjective. A sheaf $\mathcal{F}$ on $X_{\text{ét}}$ is a contravariant functor to abelian groups such that for every covering

2

\quad \mathcal{F}(U) \;=\; \operatorname{Eq}\!\Big( \prod_i \mathcal{F}(U_i) \;\rightrightarrows\; \prod_{i,j} \mathcal{F}(U_i \times_U U_j) \Big),

i.e. sections glue along the fibre products $U_i \times_U U_j$ playing the role of intersections. Derivation hint: replace "open subsets" and "intersections" in the sheaf axiom by "étale maps" and "fibre products"; the equaliser condition is exactly the descent datum.

DefinitionÉtale Cohomology

Étale sheaves form an abelian category $\mathbf{Sh}(X_{\text{ét}})$ with enough injectives. The global sections functor $\Gamma(X, -)$ is left exact, and étale cohomology is its right derived functor:

3

\quad H^i_{\text{ét}}(X, \mathcal{F}) \;:=\; R^i\Gamma(X, \mathcal{F}) \;=\; H^i\big( \Gamma(X, \mathcal{I}^\bullet) \big),

where $\mathcal{F} \to \mathcal{I}^\bullet$ is an injective resolution. Derivation hint: the standard derived-functor recipe — resolve by injectives, apply $\Gamma$ , take cohomology of the resulting complex; independence of resolution is the usual homological lemma.

Definitionℓ-adic Cohomology

Naïve étale cohomology with $\mathbb{Z}_\ell$ coefficients is ill-behaved (the constant sheaf $\underline{\mathbb{Z}_\ell}$ does not capture limits correctly). One instead defines, for $\ell \neq \operatorname{char}$ ,

4

\quad H^i_{\text{ét}}(X, \mathbb{Z}_\ell) := \varprojlim_n H^i_{\text{ét}}(X, \mathbb{Z}/\ell^n\mathbb{Z}), \qquad H^i_{\text{ét}}(X, \mathbb{Q}_\ell) := H^i_{\text{ét}}(X, \mathbb{Z}_\ell)\otimes_{\mathbb{Z}_\ell} \mathbb{Q}_\ell.

Derivation hint: torsion coefficients are well-behaved; pass to the inverse limit over the $\ell$ -power torsion sheaves $\mu_{\ell^n}$ , then invert $\ell$ to land in a characteristic-zero coefficient field.

TheoremComparison, Finiteness, and Weil-Cohomology Axioms

Let $X$ be a smooth proper variety of dimension $d$ over an algebraically closed field $k$ , and $\ell \neq \operatorname{char}(k)$ . Then: 1. (Finiteness) $H^i_{\text{ét}}(X,\mathbb{Q}_\ell)$ is a finite-dimensional $\mathbb{Q}_\ell$ -vector space, zero for $i \notin [0,2d]$ . 2. (Comparison) If $k = \mathbb{C}$ , there is a canonical isomorphism with singular cohomology:

5

\quad H^i_{\text{ét}}(X, \mathbb{Z}/n\mathbb{Z}) \;\cong\; H^i_{\text{sing}}\big( X(\mathbb{C}), \mathbb{Z}/n\mathbb{Z}\big),

hence $\dim_{\mathbb{Q}_\ell} H^i_{\text{ét}}(X,\mathbb{Q}_\ell) = b_i(X)$ , the topological Betti numbers. 3. (Poincaré Duality) There is a perfect pairing

6

\quad H^i_{\text{ét}}(X,\mathbb{Q}_\ell) \times H^{2d-i}_{\text{ét}}(X,\mathbb{Q}_\ell(d)) \longrightarrow H^{2d}_{\text{ét}}(X,\mathbb{Q}_\ell(d)) \cong \mathbb{Q}_\ell.

4. (Lefschetz Trace Formula) For $X/\mathbb{F}_q$ smooth proper and $F$ the geometric Frobenius acting on $\overline{X} = X \times_{\mathbb{F}_q}\overline{\mathbb{F}_q}$ ,

7

\quad \#X(\mathbb{F}_q) \;=\; \sum_{i=0}^{2d} (-1)^i \operatorname{Tr}\!\big( F \mid H^i_{\text{ét}}(\overline{X},\mathbb{Q}_\ell)\big).

Derivation hint: (5) follows from rigidity + Artin's comparison via henselian local rings; (6) from the trace map and cup product on the cohomology of $\mu_{\ell^n}$ ; (7) from the étale Lefschetz fixed-point theorem applied to the Frobenius correspondence, whose fixed points are exactly the $\mathbb{F}_q$ -points.

ProofFrom the Trace Formula to the Zeta Function

We show how [7] forces the rationality and functional equation of the Hasse–Weil zeta function, the first two Weil conjectures. 1. Definition of the zeta function. For $X/\mathbb{F}_q$ define

8

\quad Z(X,t) \;=\; \exp\!\Big( \sum_{m\ge 1} \#X(\mathbb{F}_{q^m}) \,\frac{t^m}{m}\Big).

Derivation hint: this is the generating function whose logarithmic derivative recovers the point-counts $\#X(\mathbb{F}_{q^m})$ . 2. Insert the trace formula. Replacing $q$ by $q^m$ in [7] gives $\#X(\mathbb{F}_{q^m}) = \sum_i (-1)^i \operatorname{Tr}(F^m \mid H^i)$ . Substitute into [8] and exchange the sums. 3. Use the trace–determinant identity. For an endomorphism $F$ of a finite-dimensional space,

9

\quad \exp\!\Big( \sum_{m\ge1} \operatorname{Tr}(F^m)\,\frac{t^m}{m}\Big) \;=\; \frac{1}{\det(1 - tF)}.

Derivation hint: diagonalise $F$ with eigenvalues $\alpha_j$ ; then $\sum_m \alpha_j^m t^m/m = -\log(1-\alpha_j t)$ and exponentiate the sum over $j$ . 4. Assemble the alternating product. The $(-1)^i$ signs become exponents:

10

\quad Z(X,t) \;=\; \prod_{i=0}^{2d} \det\!\big(1 - tF \mid H^i_{\text{ét}}(\overline{X},\mathbb{Q}_\ell)\big)^{(-1)^{i+1}} \;=\; \frac{P_1(t)\,P_3(t)\cdots P_{2d-1}(t)}{P_0(t)\,P_2(t)\cdots P_{2d}(t)}.

Derivation hint: odd-degree cohomology contributes to the numerator, even-degree to the denominator, by the sign in [7] combined with [9]. 5. Rationality. Each $P_i(t) = \det(1 - tF \mid H^i)$ is a polynomial with $\mathbb{Q}_\ell$ -coefficients (in fact $\mathbb{Z}$ -coefficients, independent of $\ell$ , by Deligne). Hence $Z(X,t) \in \mathbb{Q}(t)$ — the Weil rationality conjecture. 6. Functional equation. Poincaré duality [6] relates $H^i$ to $H^{2d-i}$ , exchanging eigenvalues $\alpha \leftrightarrow q^d/\alpha$ , producing

11

\quad Z\!\Big(X, \frac{1}{q^d t}\Big) \;=\; \pm\, q^{d\chi/2}\, t^{\chi}\, Z(X,t), \qquad \chi = \sum_i (-1)^i b_i.

Derivation hint: the duality pairing identifies the characteristic polynomial of $F$ on $H^i$ with the "reversed" polynomial on $H^{2d-i}$ twisted by $q^d$ ; the Euler characteristic $\chi$ tracks the total degree shift. The deepest statement — that the eigenvalues $\alpha_{i,j}$ of $F$ on $H^i$ satisfy $|\alpha_{i,j}| = q^{i/2}$ (the Riemann Hypothesis) — is Deligne's theorem (Weil I, 1974) and does not follow formally; it requires monodromy and a clever use of Lefschetz pencils.

∎

mathematician · Alexander Grothendieck | 1928–2014

Grothendieck reconceived algebraic geometry around the language of schemes, functors, and sites, and conceived étale cohomology precisely to attack the Weil conjectures, which he framed as the existence of a "Weil cohomology theory." In the seminars SGA 1, 4, 4½, 5, 7 (with Artin, Verdier, Deligne, Illusie, Raynaud, and others) he built the entire machinery: étale topology, topoi, the six functor formalism, duality, and the trace formula. He proved rationality and the functional equation and reduced the Riemann Hypothesis to his standard conjectures, which remain open. The final RH step was completed by his student Pierre Deligne in 1974 by a different route, a result for which Deligne received the Fields Medal (1978). Grothendieck's vision — that cohomology should be derived-functor cohomology on an abstract site — reshaped twentieth-century mathematics.

🧠 Intuition

Think of why ordinary cohomology fails over $\mathbb{F}_q$ and how étale fixes it.

Topological coverings. The circle $S^1$ has a connected double cover $z \mapsto z^2$ . This degree-2 unramified map is the prototype of an étale morphism. Its existence is detected by $H^1(S^1,\mathbb{Z}/2)\neq 0$ . Algebraically, $\mathbb{G}_m = \operatorname{Spec}\, k[t,t^{-1}]$ has the étale cover $t\mapsto t^2$ — invisible to the Zariski topology, which sees only the single generic point's coarse opens.
Why ℓ ≠ p. Over $\mathbb{F}_p$ , the $p$ -power covers (Artin–Schreier $x^p - x = a$ ) are not étale (their derivative is $-1$ but the wild ramification corrupts $p$ -torsion sheaves). So the constant sheaf $\mathbb{Z}/p$ behaves pathologically; only coefficients prime to the characteristic give a clean Weil cohomology.
Frobenius is the monodromy. Over $\mathbb{C}$ , a family of varieties degenerating around a puncture acquires a monodromy operator. Over $\mathbb{F}_q$ , the absolute Galois group $\operatorname{Gal}(\overline{\mathbb{F}_q}/\mathbb{F}_q) \cong \widehat{\mathbb{Z}}$ is topologically a loop around the "point" $\operatorname{Spec}\mathbb{F}_q$ , and its generator — geometric Frobenius — plays the role of monodromy. Counting fixed points of this monodromy is counting $\mathbb{F}_q$ -points.

Figure

Comparing the Zariski and étale "neighbourhoods" of a point, and the Frobenius loop encoding Galois action

ExampleThe Projective Line over a Finite Field

Given. $X = \mathbb{P}^1_{\mathbb{F}_q}$ , dimension $d=1$ . We compute its zeta function from cohomology.

Step 1 — Cohomology groups. By the comparison theorem [5], $H^i_{\text{ét}}(\overline{\mathbb{P}^1},\mathbb{Q}_\ell)$ matches the singular cohomology of $\mathbb{CP}^1 = S^2$ :

12

\quad H^0 = \mathbb{Q}_\ell, \quad H^1 = 0, \quad H^2 = \mathbb{Q}_\ell(-1), \quad b_0=1,\,b_1=0,\,b_2=1.

Derivation hint: $S^2$ has Betti numbers $1,0,1$ ; the Tate twist $(-1)$ records the Frobenius weight on the fundamental class.

Step 2 — Frobenius eigenvalues. On $H^0$ , geometric Frobenius acts trivially (eigenvalue $1$ ). On $H^2 = \mathbb{Q}_\ell(-1)$ , it acts by multiplication by $q$ (the Tate twist). So $\operatorname{Tr}(F^m\mid H^0)=1$ , $\operatorname{Tr}(F^m\mid H^2)=q^m$ .

Step 3 — Point count via [7].

13

\quad \#\mathbb{P}^1(\mathbb{F}_{q^m}) = \operatorname{Tr}(F^m\mid H^0) + \operatorname{Tr}(F^m\mid H^2) = 1 + q^m.

Derivation hint: only the even-degree groups contribute (with $+$ signs since $H^1=0$ ); add their Frobenius traces.

Step 4 — Zeta function via [10].

14

\quad Z(\mathbb{P}^1,t) = \frac{1}{(1-t)(1-qt)}.

Answer. $\boxed{\#\mathbb{P}^1(\mathbb{F}_{q^m}) = 1 + q^m, \qquad Z(\mathbb{P}^1,t) = \dfrac{1}{(1-t)(1-qt)}}$ — eigenvalues $1$ and $q$ have absolute values $q^0$ and $q^1$ , confirming RH for weights $0$ and $2$ .

ExampleAn Elliptic Curve over 𝔽_q

Given. $E/\mathbb{F}_q$ a smooth proper genus-1 curve, $d=1$ . We extract its point count.

Step 1 — Cohomology. Like the torus $T^2$ : $b_0 = 1,\ b_1 = 2,\ b_2 = 1$ . Thus $H^1_{\text{ét}}(\overline{E},\mathbb{Q}_\ell)$ is 2-dimensional, dual to the $\ell$ -adic Tate module $T_\ell E \otimes \mathbb{Q}_\ell$ .

Step 2 — Frobenius on $H^1$ . Let $\alpha,\beta$ be the eigenvalues of $F$ on $H^1$ . The RH (Hasse's theorem, the elliptic case proved 1933) gives

15

\quad |\alpha| = |\beta| = q^{1/2}, \qquad \alpha\beta = q, \qquad \alpha+\beta = a_q,

where $a_q := q + 1 - \#E(\mathbb{F}_q)$ is the trace of Frobenius. Derivation hint: $H^0$ and $H^2$ contribute $1$ and $q$ ; matching to the full point count below isolates $\alpha+\beta$ , and Poincaré duality forces $\alpha\beta=q$ .

Step 3 — Point count. From [7] with the $-$ sign on the odd group:

16

\quad \#E(\mathbb{F}_q) = 1 - (\alpha+\beta) + q = q + 1 - a_q.

Step 4 — Numbers. Take $E: y^2 = x^3 + x + 1$ over $\mathbb{F}_5$ . Direct enumeration gives $9$ affine points plus the point at infinity, so $\#E(\mathbb{F}_5)=9$ . Then $a_5 = 5 + 1 - 9 = -3$ , and the Hasse bound $|a_5| \le 2\sqrt{5}\approx 4.47$ is satisfied. The characteristic polynomial of Frobenius is $t^2 + 3t + 5$ , eigenvalues $\frac{-3 \pm \sqrt{-11}}{2}$ with $|\alpha|=\sqrt{5}$ .

Answer. $\boxed{\#E(\mathbb{F}_5) = 9,\quad a_5 = -3,\quad Z(E,t) = \dfrac{1 + 3t + 5t^2}{(1-t)(1-5t)}}$ .

⚠️ Common mistake

Étale cohomology is not the cohomology of the étale topology with any coefficients.

With $p$ -torsion coefficients in characteristic $p$ ( $\ell = p$ ), the theory is pathological: $H^i_{\text{ét}}(X,\mathbb{Z}/p)$ does not give Betti numbers, fails finiteness in the naïve sense, and the correct $p$ -adic theory is crystalline cohomology, an entirely different construction. Always take $\ell \neq \operatorname{char}$ .
$H^i_{\text{ét}}(X,\underline{\mathbb{Z}_\ell})$ with the constant sheaf is wrong. One must define $\mathbb{Z}_\ell$ -cohomology as the inverse limit over $\mathbb{Z}/\ell^n$ as in [4]; the constant sheaf $\underline{\mathbb{Z}_\ell}$ has trivial higher cohomology on a normal scheme.
Étale cohomology of $\operatorname{Spec} k$ is Galois cohomology. $H^i_{\text{ét}}(\operatorname{Spec} k, \mathcal{F}) = H^i(\operatorname{Gal}(\overline{k}/k), \mathcal{F})$ . So "cohomology of a point" is generally nonzero — a feature, not a bug; it is where arithmetic lives.

CounterexampleWhy the Zariski Topology Fails

For an irreducible variety $X$ over an algebraically closed field, the constant Zariski sheaf $\underline{\mathbb{Z}/n}$ is flasque (because every nonempty open is connected and dense, forcing all restriction maps to be isomorphisms). Hence

17

\quad H^i_{\text{Zar}}(X,\underline{\mathbb{Z}/n}) = 0 \quad \text{for all } i > 0,

even when $X = \mathbb{P}^1$ , whose "topological" $H^2$ should be nonzero. The étale topology breaks this rigidity: the connected double cover of $\mathbb{G}_m$ exhibits a nontrivial $H^1_{\text{ét}}(\mathbb{G}_m, \mu_2) \neq 0$ , recovering the expected $\mathbb{Z}/2$ . Derivation hint: flasque sheaves are acyclic; irreducibility makes the constant sheaf flasque in Zariski but not in étale, where finite connected covers furnish genuine "loops."

connection

Étale cohomology sits at the crossroads of several theories:

Galois Cohomology. For $X = \operatorname{Spec} k$ , étale cohomology is Galois cohomology. The Hochschild–Serre spectral sequence $H^p(\operatorname{Gal}(\overline{k}/k), H^q_{\text{ét}}(\overline{X})) \Rightarrow H^{p+q}_{\text{ét}}(X)$ bridges the geometric ( $\overline{X}$ ) and arithmetic ( $k$ ) layers.
Singular Cohomology. Via the comparison theorem [5], étale cohomology equals topological cohomology over $\mathbb{C}$ , so it imports Betti numbers, Lefschetz theory, and the Hodge–de Rham picture.
Crystalline Cohomology. The " $p$ -adic companion" handling $\ell = p$ ; together with étale cohomology it forms the system of mixed motives and feeds into $p$ -adic Hodge theory.
Weil Conjectures. Étale cohomology was created to prove them; rationality and the functional equation follow formally [10]–[11], the Riemann Hypothesis from Deligne's Weil I.
Class Field Theory. Étale (= Galois) cohomology of number rings, via the Brauer group $\operatorname{Br}(k) = H^2_{\text{ét}}(\operatorname{Spec} k, \mathbb{G}_m)$ , encodes local–global reciprocity.
L-functions and Motives. $\ell$ -adic representations $\operatorname{Gal} \to \operatorname{GL}(H^i_{\text{ét}})$ define motivic $L$ -functions, the home of the Langlands program and BSD.

🔬 Where this appears

⏳ History

- 1949 — Weil conjectures stated. André Weil, computing zeta functions of diagonal hypersurfaces over

\mathbb{F}_q

, conjectures rationality, a functional equation, and the Riemann Hypothesis, and predicts a cohomological explanation modelled on Lefschetz fixed-point theory.

- 1949–1955 — Search for a Weil cohomology. Serre's work on coherent sheaf cohomology (FAC, 1955) shows Zariski cohomology cannot suffice; the need for a topology with "enough covers" becomes clear.

- 1958 — Grothendieck's vision. At the Edinburgh ICM, Grothendieck announces his program of schemes and the search for a Weil cohomology, sketching the idea of a topology built from étale maps.

- 1960–1964 — The étale topology. Grothendieck and Michael Artin develop the étale site and basic properties; Artin's Harvard seminar notes (1962) lay foundations.

- 1963–1973 — SGA 4, 5, 7. The seminars Théorie des Topos et Cohomologie Étale des Schémas (SGA 4, with Artin & Verdier), SGA 5 (trace formula,

L

-functions), and SGA 7 (monodromy, vanishing cycles) build the complete theory; proper and smooth base change, duality, and the trace formula are established.

- 1964 — Grothendieck proves rationality and the functional equation of the zeta function using the trace formula [7] and Poincaré duality.

- 1968–1970 — Standard conjectures. Grothendieck reduces the Riemann Hypothesis to his standard conjectures on algebraic cycles; these remain open.

- 1974 — Deligne's Weil I. Pierre Deligne proves the Riemann Hypothesis

|\alpha_{i,j}| = q^{i/2}

via Lefschetz pencils, monodromy, and a clever positivity/Rankin argument, not the standard conjectures.

- 1980 — Deligne's Weil II. Generalises RH to a sweeping theory of weights for arbitrary sheaves, the foundation of the modern theory of perverse sheaves and the decomposition theorem.

open-problem

Standard Conjectures. Grothendieck's standard conjectures (Lefschetz type, Hodge type) on algebraic cycles, which would yield a purely cohomological proof of RH and the semisimplicity of Frobenius, remain open in general. They imply the existence of a Tannakian category of pure motives.
Tate Conjecture. Whether the subspace of $H^{2i}_{\text{ét}}(\overline{X},\mathbb{Q}_\ell(i))$ fixed by Frobenius is spanned by algebraic cycle classes is open beyond special cases (divisors on abelian varieties, K3 surfaces in many cases). It is the arithmetic counterpart of the Hodge conjecture.
**Integral and $

Arithmetic Geometry and Etale Cohomology

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits