Class Field Theory — Overview and Main Theorems

⚡ Key Takeaways

Class field theory classifies all abelian extensions of a local or global field $K$ in terms of arithmetic data intrinsic to $K$ itself (its ideal class group, idele class group, or multiplicative group). Key result in words: there is a canonical isomorphism — the reciprocity map — between the Galois group of the maximal abelian extension and a quotient of the idele class group, and every finite abelian extension corresponds to an open finite-index subgroup. Where it applies: number fields, function fields over finite fields, and local fields ( $p$ -adic fields and Laurent-series fields). Who/when: Kronecker, Weber, Hilbert (1890s–1900s); Takagi (1920); Artin (1927, reciprocity); Chevalley (1930s–40s, ideles); Tate, Iwasawa (cohomological reformulation, 1950s).

💡 The Core Insight

The arithmetic of $K$ — how primes split, ramify, and remain inert in extensions — is governed by the multiplicative structure of $K$ packaged adelically. Abelian Galois groups are recovered from $K$ without ever naming the extension: $\mathrm{Gal}(L/K)^{\mathrm{ab}}$ is a quotient of $C_K = \mathbb{A}_K^\times / K^\times$ . The reciprocity map turns the global problem of constructing fields into a local-to-global problem about characters of idele class groups, where the product formula and local reciprocity assemble into a single coherent law.

DefinitionLocal field, global field, ideles

A local field is a field complete with respect to a discrete (or archimedean) absolute value with finite (or $\mathbb{R},\mathbb{C}$ ) residue field: e.g. $\mathbb{Q}_p$ , finite extensions thereof, $\mathbb{F}_q((t))$ , $\mathbb{R}$ , $\mathbb{C}$ . A global field is a finite extension of $\mathbb{Q}$ (a number field) or a finite extension of $\mathbb{F}_q(t)$ (a function field). For a global field $K$ with set of places $M_K$ , the adele ring and idele group are the restricted products

1

\quad \mathbb{A}_K = {\prod_{v\in M_K}}' \, (K_v, \mathcal{O}_v), \qquad \mathbb{A}_K^\times = {\prod_{v}}' \, (K_v^\times, \mathcal{O}_v^\times).

Derivation hint: restricted product means almost all components lie in the compact subring $\mathcal{O}_v$ (resp. unit group $\mathcal{O}_v^\times$ ); topologize so that $\prod_{v\in S}K_v \times \prod_{v\notin S}\mathcal{O}_v$ are open for finite $S$ . The idele class group is

2

\quad C_K = \mathbb{A}_K^\times / K^\times,

where $K^\times$ embeds diagonally. Derivation hint: the product formula $\prod_v |x|_v = 1$ for $x\in K^\times$ shows the diagonal embedding lands in the norm-one ideles and is discrete cocompact in that subgroup.

DefinitionFrobenius and the Artin symbol

Let $L/K$ be a finite abelian extension of global fields and $\mathfrak{p}$ a prime of $K$ unramified in $L$ . There is a unique element $\mathrm{Frob}_\mathfrak{p}\in\mathrm{Gal}(L/K)$ — the Artin symbol $\left(\frac{L/K}{\mathfrak{p}}\right)$ — characterized by

3

\quad \mathrm{Frob}_\mathfrak{p}(x) \equiv x^{N\mathfrak{p}} \pmod{\mathfrak{P}}, \qquad \forall x\in\mathcal{O}_L,

where $\mathfrak{P}\mid\mathfrak{p}$ and $N\mathfrak{p} = |\mathcal{O}_K/\mathfrak{p}|$ . Derivation hint: the decomposition group $D_\mathfrak{P}\cong\mathrm{Gal}(\kappa(\mathfrak{P})/\kappa(\mathfrak{p}))$ is cyclic generated by $x\mapsto x^{N\mathfrak{p}}$ ; abelianness makes it independent of the choice of $\mathfrak{P}$ .

TheoremArtin Reciprocity (global, idelic form)

Let $L/K$ be a finite abelian extension of global fields. There is a canonical surjective homomorphism, the Artin (reciprocity) map,

4

\quad \theta_{L/K}:\ C_K = \mathbb{A}_K^\times/K^\times \ \twoheadrightarrow\ \mathrm{Gal}(L/K),

with kernel exactly $N_{L/K}(C_L) = N_{L/K}(\mathbb{A}_L^\times)\,K^\times / K^\times$ , inducing

5

\quad C_K / N_{L/K}(C_L) \ \xrightarrow{\ \sim\ }\ \mathrm{Gal}(L/K).

Derivation hint: $\theta_{L/K}$ is built as the product of local reciprocity maps $\theta_v: K_v^\times\to\mathrm{Gal}(L_w/K_v)$ ; the key non-trivial assertion (Artin's reciprocity law) is that $\theta_{L/K}$ kills the diagonal $K^\times$ . Moreover, on unramified primes the map sends a uniformizer to Frobenius:

6

\quad \theta_{L/K}\big(\dots,1,\pi_\mathfrak{p},1,\dots\big) = \mathrm{Frob}_\mathfrak{p} \in \mathrm{Gal}(L/K).

Derivation hint: this is local CFT at $\mathfrak{p}$ — the local Artin map sends a uniformizer to arithmetic Frobenius in the unramified case; gluing over all places yields the global statement.

TheoremExistence and Classification (Takagi)

The map $L \mapsto N_{L/K}(C_L)$ is an inclusion-reversing bijection between finite abelian extensions $L/K$ (inside a fixed separable closure) and open finite-index subgroups $N\subset C_K$ :

7

\quad \{\,L/K \text{ finite abelian}\,\}\ \xleftrightarrow{\ 1:1\ }\ \{\,N\subset C_K\ \text{open, finite index}\,\}.

Derivation hint: surjectivity (every such $N$ is a norm group) is the "Existence Theorem," proved via Kummer theory plus a counting/index argument over the class formation axioms; injectivity follows from reciprocity [5]. Consequently $\mathrm{Gal}(K^{\mathrm{ab}}/K)$ is the profinite completion of $C_K$ :

8

\quad \mathrm{Gal}(K^{\mathrm{ab}}/K)\ \cong\ \widehat{C_K} \ =\ \varprojlim_{N} C_K/N.

Derivation hint: take the inverse limit of [5] over all finite abelian $L$ ; for number fields the connected component of $1$ in $C_K$ is killed, giving the profinite completion.

TheoremLocal Class Field Theory

Let $K$ be a non-archimedean local field. There is a canonical injective homomorphism with dense image (an isomorphism after profinite completion),

9

\quad \theta_K:\ K^\times \ \hookrightarrow\ \mathrm{Gal}(K^{\mathrm{ab}}/K), \qquad \widehat{K^\times}\xrightarrow{\sim}\mathrm{Gal}(K^{\mathrm{ab}}/K),

under which finite abelian $L/K$ correspond to finite-index open subgroups $N_{L/K}(L^\times)\subset K^\times$ , and

10

\quad K^\times / N_{L/K}(L^\times)\ \xrightarrow{\ \sim\ }\ \mathrm{Gal}(L/K).

Derivation hint: $\theta_K$ sends $\mathcal{O}_K^\times$ onto the inertia subgroup and a uniformizer to a Frobenius lift; constructed via Lubin–Tate formal groups or via Galois cohomology with the invariant map $\mathrm{inv}_K:\mathrm{Br}(K)\xrightarrow{\sim}\mathbb{Q}/\mathbb{Z}$ .

ProofArchitecture of the proof via class formations

The modern proof unifies local and global CFT through the axiomatic notion of a class formation (Artin–Tate).

Step 1 — Set-up. Let $G = \mathrm{Gal}(\bar K/K)$ act on a module $A$ (for global CFT, $A =$ idele classes of $\bar K$ ; for local, $A = (\bar K)^\times$ ). The pair $(G,A)$ is a class formation if for every finite extension $E/K$ (fixed field $G_E$ ):

11

\quad H^1(G_{E},A)=0,\qquad H^2(G_E,A)\ \xrightarrow{\ \mathrm{inv}_E\ }\ \tfrac{1}{[E:K]}\mathbb{Z}/\mathbb{Z}\subset\mathbb{Q}/\mathbb{Z}.

Derivation hint: $H^1=0$ is "Hilbert 90" in the relevant guise; the $H^2$ statement packages the fundamental class.

Step 2 — Tate's theorem. For a cyclic (more generally any finite) $E/F$ inside the formation, cup product with the fundamental class $u_{E/F}\in H^2(\mathrm{Gal}(E/F),A_E)$ gives isomorphisms on all Tate cohomology:

12

\quad \widehat{H}^{\,n}(\mathrm{Gal}(E/F),\mathbb{Z})\ \xrightarrow[\sim]{\ \cup\, u_{E/F}\ }\ \widehat{H}^{\,n+2}(\mathrm{Gal}(E/F),A_E).

Derivation hint: Tate–Nakayama; the hypotheses $\widehat H^1=0$ and $\widehat H^2$ cyclic of order $[E:F]$ verify the conditions of Tate's cohomology isomorphism.

Step 3 — Reciprocity for $n=-2$ . Setting $n=-2$ in [12] and using $\widehat{H}^{-2}(G,\mathbb{Z})=G^{\mathrm{ab}}$ , $\widehat{H}^{0}(G,A_E)=A_F/N_{E/F}A_E$ gives

13

\quad \mathrm{Gal}(E/F)^{\mathrm{ab}}\ \xrightarrow{\ \sim\ }\ A_F / N_{E/F} A_E,

the inverse of which is the reciprocity isomorphism. Derivation hint: $\widehat H^{-2}(G,\mathbb Z)\cong H_1(G,\mathbb Z)\cong G^{\mathrm{ab}}$ and $\widehat H^0$ is invariants modulo norms; cup with $u$ identifies them.

Step 4 — Existence theorem. Show every open finite-index subgroup $N\subset C_K$ is a norm group. Reduce (via $C_K/C_K^n$ ) to the case where $K$ contains $\mu_n$ , then use Kummer theory: $K^\times/(K^\times)^n$ pairs perfectly with $\mathrm{Gal}(K(\sqrt[n]{K^\times})/K)$ . A computation of indices using [13] forces equality $[C_K:N]=[L:K]$ for $L$ the fixed field, proving $N=N_{L/K}C_L$ . Derivation hint: the "First and Second Inequalities" $[C_K:N_{L/K}C_L]\le[L:K]$ and $\ge[L:K]$ together pin the index; the second inequality is analytic (uses $L$ -functions / nonvanishing) or cohomological (Herbrand quotient).

Step 5 — Local-global compatibility. The global Artin map is the restricted product of local maps; that $\theta_{L/K}$ annihilates $K^\times$ is equivalent to the product-of-local-symbols identity $\prod_v (a, L/K)_v = 1$ for $a\in K^\times$ , the precise content of Artin reciprocity. $\blacksquare$

∎

mathematician · Emil Artin | 1898–1962

Artin proved the general reciprocity law in 1927, realizing Hilbert's vision (his 9th problem) of a "most general reciprocity law" governing power residues. His insight was to define the Frobenius-to-class-group correspondence and prove it is a homomorphism killing principal ideles — turning Takagi's existence-theoretic class field theory (1920), which classified abelian extensions abstractly, into an explicit, canonical isomorphism. Artin also introduced Artin $L$ -functions, conjectured their holomorphy, and (with Tate) recast the whole theory cohomologically. His reciprocity law contains quadratic, cubic, and Eisenstein reciprocity as special cases and is the abelian seed of the Langlands program.

🧠 Intuition

Think of CFT as a dictionary between two worlds:

Field-theoretic world: abelian extensions $L/K$ , Galois groups, how primes $\mathfrak{p}$ split.
Arithmetic world: congruence/ray class groups, ideles, units modulo conductors.

The dictionary entry is: " $\mathfrak{p}$ splits completely in $L$ " $\iff$ " $\mathfrak{p}$ lies in the subgroup of $C_K$ corresponding to $L$ ." Splitting is thus governed by a congruence condition on $\mathfrak{p}$ . The simplest instance: in $\mathbb{Q}(\zeta_m)/\mathbb{Q}$ , a prime $p\nmid m$ splits completely iff $p\equiv 1\pmod m$ — a pure congruence. CFT says every abelian splitting law is, ultimately, a congruence law of this type (possibly with archimedean/sign conditions encoded by infinite places). The ideles exist precisely to package "congruence at all primes simultaneously, including sign conditions at infinity" into one group whose characters are exactly the abelian extensions.

Figure

The Artin reciprocity map: ideles to Galois group, with the Frobenius arrow

ExampleCyclotomic CFT over

\mathbb{Q}

: the Kronecker–Weber phenomenon

Given: $K=\mathbb{Q}$ , $L=\mathbb{Q}(\zeta_m)$ with $m=15$ .

Goal: Determine $\mathrm{Gal}(L/K)$ as a quotient of $C_\mathbb{Q}$ and find how $p=2$ behaves.

Steps. 1. $\mathrm{Gal}(\mathbb{Q}(\zeta_{15})/\mathbb{Q}) \cong (\mathbb{Z}/15)^\times$ , sending $\sigma_a:\zeta\mapsto\zeta^a$ to $a$ . Order $= \varphi(15)=8$ . 2. The Artin map sends an unramified prime $p\nmid 15$ to $\mathrm{Frob}_p = \sigma_p$ , i.e. to $p\bmod 15$ . 3. For $p=2$ : $\mathrm{Frob}_2 = \sigma_2 \leftrightarrow 2\in(\mathbb{Z}/15)^\times$ . The order of $2$ mod $15$ : $2,4,8,16\equiv1$ , so order $4$ . 4. The residue degree $f$ = order of $\mathrm{Frob}_p$ = $4$ ; number of primes above $2$ is $g=\varphi(15)/f = 8/4 = 2$ . 5. The norm subgroup is $N_{L/\mathbb{Q}}(C_L)$ corresponding under $C_\mathbb{Q}/N \cong (\mathbb{Z}/15)^\times$ to congruences mod $15$ .

Answer.

\boxed{\ 2 \text{ factors in } \mathbb{Q}(\zeta_{15}) \text{ as } 2\mathcal{O}_L=\mathfrak{P}_1\mathfrak{P}_2,\quad f=4,\ g=2,\ e=1.\ }

ExampleHilbert class field of

\mathbb{Q}(\sqrt{-23})

Given: $K=\mathbb{Q}(\sqrt{-23})$ , class number $h_K = 3$ .

Goal: Identify the Hilbert class field $H$ (maximal unramified abelian extension) and determine splitting of a prime via the class group.

Steps. 1. By CFT with conductor $1$ (no ramification, no sign conditions), $\mathrm{Gal}(H/K)\cong \mathrm{Cl}(K)\cong \mathbb{Z}/3$ . So $[H:K]=3$ , $[H:\mathbb{Q}]=6$ . 2. $H = K(\alpha)$ where $\alpha$ is a root of $x^3 - x - 1$ (whose discriminant is $-23$ ); thus $H$ is the splitting field of $x^3-x-1$ over $\mathbb{Q}$ . 3. A rational prime $p$ (unramified, $p\ne23$ ) splits completely in $H$ iff $p$ splits in $K$ and the prime above it is principal — equivalently iff $\mathrm{Frob}_\mathfrak{p}=1$ in $\mathrm{Cl}(K)$ . 4. Concretely $p$ splits completely in $H$ iff $x^3-x-1$ has three roots mod $p$ iff $p = x^2+23y^2$ is solvable. 5. Test $p=59$ : $-23$ is a QR mod $59$ ( $59$ splits in $K$ ), and indeed $59 = 6^2+23\cdot1^2 = 36+23$ . So $59$ is represented; $\mathrm{Frob}=1$ .

Answer.

\boxed{\ H=\text{splitting field of } x^3-x-1;\quad p \text{ splits completely} \iff p=x^2+23y^2.\ }

⚠️ Common mistake

CFT does not describe non-abelian extensions. The reciprocity isomorphism [4] only sees $\mathrm{Gal}(L/K)$ when $L/K$ is abelian; for general $L$ it sees only $\mathrm{Gal}(L/K)^{\mathrm{ab}}$ . There is no classical CFT for, say, the splitting field of $x^5 - x - 1$ (Galois group $S_5$ ).

The Artin map is not "the obvious quotient." The hard theorem is that $\theta_{L/K}$ kills the diagonal $K^\times\subset\mathbb{A}_K^\times$ . Building local maps is comparatively easy; reciprocity is the global gluing.

"Existence theorem" $\ne$ explicit construction. CFT proves the abelian extension corresponding to a given norm group exists, but does not in general produce generators. Explicitly generating $K^{\mathrm{ab}}$ is Hilbert's 12th problem, solved only for $K=\mathbb{Q}$ (roots of unity, Kronecker–Weber) and imaginary quadratic $K$ (singular moduli, torsion of CM elliptic curves — Kronecker's Jugendtraum).

CounterexampleThe Hilbert class field need not be generated by units of

K

For $K=\mathbb{Q}(\sqrt{-23})$ , one might hope $H$ is $K(\sqrt[3]{u})$ for some explicit unit or simple radical in $K$ . But $H$ is unramified over $K$ , so it cannot be a Kummer extension $K(\sqrt[3]{a})$ with $a$ a non-unit (that would ramify), and $\mathcal{O}_K^\times=\{\pm1\}$ offers no nontrivial radicals. The actual generator comes from the transcendental $j$ -function evaluated at a CM point: $H=K(j(\tfrac{1+\sqrt{-23}}{2}))$ . This shows the Existence Theorem is genuinely non-constructive within elementary algebra — explicit class field generation requires analytic (modular) input, illustrating exactly why Hilbert's 12th problem is hard.

connection

Quadratic Reciprocity: the law $\left(\frac{p}{q}\right)\left(\frac{q}{p}\right)=(-1)^{\frac{p-1}{2}\frac{q-1}{2}}$ is the $n=2$ shadow of Artin reciprocity, recovered from $\theta$ for $\mathbb{Q}(\sqrt{q^*})/\mathbb{Q}$ and $\prod_v(a,b)_v=1$ .
Galois Cohomology and the Brauer Group: local CFT is encoded by $\mathrm{inv}_K:\mathrm{Br}(K)\xrightarrow{\sim}\mathbb{Q}/\mathbb{Z}$ ; the global reciprocity sequence $0\to\mathrm{Br}(K)\to\bigoplus_v\mathrm{Br}(K_v)\xrightarrow{\sum\mathrm{inv}_v}\mathbb{Q}/\mathbb{Z}\to0$ is equivalent to Artin reciprocity.
Iwasawa Theory: studies how class groups grow in $\mathbb{Z}_p$ -towers, refining CFT into $p$ -adic $L$ -function statements (Main Conjecture).
Langlands Program: abelian CFT $\cong$ the $\mathrm{GL}_1$ case; Langlands reciprocity generalizes $\theta$ to a correspondence between $n$ -dimensional Galois representations and automorphic forms on $\mathrm{GL}_n$ .
Complex Multiplication: generates abelian extensions of imaginary quadratic fields explicitly, solving Hilbert 12 in that case.

🔬 Where this appears

⏳ History

- 1853–1880, Kronecker & Weber: Kronecker conjectures that every abelian extension of

\mathbb{Q}

lies in a cyclotomic field; Jugendtraum for imaginary quadratic fields via elliptic functions. Weber gives early definitions of class fields and ray class groups.

- 1896–1897, Hilbert: Zahlbericht synthesizes the theory; Hilbert defines the Hilbert class field and frames the program; his 9th and 12th problems (1900) set the agenda.

- 1903, Furtwängler: proves the principal ideal theorem in special cases (capitulation).

- 1907–1920, Takagi: during WWI isolation in Japan, Takagi proves the general existence theorem and classifies all abelian extensions by norm groups, establishing class field theory's main structural results.

- 1923–1927, Artin: formulates and proves the general reciprocity law, giving the canonical Frobenius-based isomorphism; the keystone that makes the theory explicit.

- 1930, Furtwängler & Artin: principal ideal theorem proved in full (via Artin's reduction to a group-theoretic statement, completed by Furtwängler).

- 1936–1940, Chevalley: introduces ideles and gives a purely arithmetic (analysis-free) proof, removing reliance on

L

-function nonvanishing for the second inequality.

- 1950s, Hochschild–Nakayama–Tate–Artin: cohomological reformulation via class formations, the fundamental class, and Tate's theorem; Tate's 1950 thesis recasts

L

-functions idelically.

- 1965–1968, Lubin–Tate: explicit local class field theory via formal groups, giving generators of local abelian extensions; Iwasawa develops

\mathbb{Z}_p

-extensions, opening Iwasawa theory.

- 1967–, Langlands: non-abelian generalization conjectured, recasting CFT as the

\mathrm{GL}_1

case of a vast reciprocity.

open-problem

Hilbert's 12th problem (explicit class fields): Beyond $\mathbb{Q}$ (Kronecker–Weber, roots of unity) and imaginary quadratic fields (CM, singular moduli), there is no known explicit description of generators of $K^{\mathrm{ab}}$ for a general number field. Stark's conjectures predict that special values $L'(0,\chi)$ of Artin $L$ -functions yield "Stark units" generating abelian extensions — proven only in special cases (rank-one abelian over $\mathbb{Q}$ , some real quadratic data).

Non-abelian reciprocity: CFT stops at $\mathrm{GL}_1$ . The Langlands program conjectures a correspondence between $n$ -dimensional Galois representations and automorphic representations of $\mathrm{GL}_n$ , generalizing $\theta$ . Vast portions remain open; even Artin's holomorphy conjecture (that non-trivial Artin $L$ -functions are entire) is unproven in general, following only in cases where modularity / functoriality is established.

Function-field vs number-field asymmetry: function-field CFT is more "geometric" (covers of curves), and the analogous Langlands correspondence is fully proven (Drinfeld for $\mathrm{GL}_2$ , L. Lafforgue for $\mathrm{GL}_n$ ), whereas the number-field case remains largely conjectural — a striking unresolved analogy.

Class Field Theory — Overview and Main Theorems

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also