Class Field Theory and Artin L-Functions Thomas Poguntke 24. Januar 2012 Class field theory Let L | K be a normal extension of algebraic number fields. Definition (a) Recall that the idele group of K is defined as Y Kp× | ap ∈ Up for almost all p}, IK := {(ap )p ∈ p where the product is over all places p of K, and Up denotes the units in Kp× . We can embed K × into IK via the map x 7→ (x)p to form the subgroup of principal ideles. The quotient CK := IK /K × is called the idele class group of K. (b) Let G = GL|K denote the Galois group of L | K. Then IL and CL are G-modules in the following manner: for a = (aP ) ∈ IL and σ ∈ GL|K , define σa ∈ IL to be (σa)P := σ(aσ−1 P ) at the place P, where σ : Kσ−1 P → KP is the component-wise Kp × isomorphism, induced by σ. This Q action obviously factors over K . Therefore, the norm NL|K := σ∈G σ yields a subgroup NL|K CL ≤ CK . Artin reciprocity law There is a canonical exact sequence ( ,L|K) 1 → NL|K CL → CK −−−−→ Gab L|K → 1. In particular, for abelian extensions, we will identify GL|K = CK /NL|K CL . Proposition 1 Let a ∈ CK , and a = (ap ) ∈ IK be any representative. Then Y (a, L | K) = (ap , LP | Kp ), p ∼ where ( , LP | Kp ) : Kp× − → GLP |Kp = GP ≤ GL|K are the local norm residue symbols (GP the decomposition group), for any P | p. 1 Remark Artin reciprocity is one of the main results of class field theory, and has many consequences. For example, it contains the quadratic reciprocity law as a (very) special case, which is where the name comes from. √ In fact, for a, b ∈ K × , let (a, b)p be the p-adic Hilbert symbol. Let L := K( b). Obviously, a = 1 ∈ CK , so proposition 1 states that Y Y (a, LP | Kp ) = (a, b)p , 1 = (a, L | K) = p p which is Hilbert’s reciprocity law. If then K = Q, and a = p 6= q = b are odd primes, we have Y p−1 q−1 p q 1= (p, q)p = (−1) 2 2 , q p p the quadratic reciprocity law. For more details on this, see Bac:III. Definition A subgroup N ≤ CK is called a norm group, if there is a normal extension L | K, such that N = NL|K CL . Proposition 2 The map L 7→ NL := NL|K CL is an inclusion-reversing bijection between all abelian extensions L | K and the norm groups of CK . The field L is called the class field of NL . Existence theorem The norm groups of CK are precisely the closed subgroups of finite index. Example Q Q 1 1 1 1 K × /K × . Then, by definition, IK ≤ IK is := IK := p|∞ Kp× × p-∞ Up , and CK Let IK 1 open, hence so is CK ≤ CK , and therefore closed. To compute the index, we observe . × 1 1 × 1 CK /CK = IK /K IK K × /K × = IK /IK K = Cl(K), Q the ideal class group of K. The last isomorphism is given by [a] 7→ [ p-∞ pνp (ap ) ]. 1 This means we have [CK : CK ] = hK < ∞. Thus, by the existence theorem, we obtain 1 a normal (abelian) extension L | K, such that CK = NL|K CL , and so 1 = Cl(K). GL|K = CK /NL|K CL = CK /CK We call L the Hilbert class field of K. There are some nice results for L: Proposition 3 The Hilbert class field of K is the maximal unramified abelian extension of K. 2 Principal ideal theorem In the Hilbert class field L of K, any ideal of K becomes a principal ideal, i.e. the natural map Cl(K) → Cl(L) is trivial. The last result from class field theory we want to mention, is the following: Proposition 4 Let L | K be abelian, n := [L : K], p an unramified prime ideal of K. Furthermore, let π ∈ Kp× be a prime element, π̃ ∈ CK the class of the idele (. . . , 1, π, 1, . . .) ∈ IK , as well as f the order of π̃ in CK /NL|K CL . Then p splits into r = nf distinct prime ideals of L: p = P1 · · · Pr , each of inertia degree f . Proof: By Artin reciprocity, CK /NL|K CL = GL|K , so π̃ mod NL|K CL has the same order as the element (π̃, L | K) = (π, LP | Kp ) ∈ GP (see prop. 1). From local class field theory, we know that (π, LP | Kp ) = ϕP , the Frobenius morphism, hence f = ord GP = f (P | p), the inertia degree. Then the fundamental equation n = rf yields our result. Artin L-functions Let L | K be a normal extension of algebraic number fields, and (V, ρ : G → GL(V )) a representation of the Galois group G = GL|K . Definition & lemma 1 Let p be a place of K, GP and IP the decomposition and inertia groups of P | p, respectively. Then GP /IP ∼ = Gk(P)|k(p) is generated by the Frobenius morphism ϕP (where k(p) denotes the residue field of p). We can understand ϕP as an endomorphism of the fix module V IP V IP via ρ (caution! we will call this ϕP = ρ(ϕP ) again!). Then the characteristic polynomial det(1 − ϕP t; V IP ) is well-defined. Proof: If P, P0 | p, there is σ ∈ G such that P0 = σP, whence ϕP = σ −1 ϕP0 σ. Definition The Artin L-function of ρ is defined as Y L(L | K, χ, s) := p 1 , det(1 − ϕP N (p)−s ; V IP ) where χ : σ 7→ tr(ρσ) is the character of ρ. 3 Remark L(L | K, χ, s) is an analytic function on Re(s) > 1. Example Let K = Q, L = Q(µm ) (with µm the m-th roots of unity). If χ : GL|K → GL1 C is a 1-dimensional representation, we can use Artin’s reciprocity law θ : (Z/mZ)× ∼ = GL|K × × to understand χ as a Dirichlet character (Z/mZ) → C . Note that θ(p mod m) = ϕp for any prime p - m (this is also enough to establish the isomorphism; we will really need Artin only later in our general setting). We then have Q 1 L(L | K, χ, s) = det(1−ϕp N (p)−s ; C) p Q 1 = 1−χ(ϕp )p−s p-m Q 1 = 1−χ(p)p−s p = L(χ, s), the Dirichlet L-function. Note that we can ignore the primes p | m, since χ(p) = 0. Also remember that we write ϕp = χ(ϕp ). Proposition 5 (i) For the trivial character, L(L | K, 1, s) = ζK (s) is the Dedekind zeta function. (ii) If χ, χ0 are two characters, we have L(L | K, χ+χ0 , s) = L(L | K, χ, s)L(L | K, χ0 , s). (iii) Let L0 | L | K be normal over K, then L(L0 | K, χ, s) = L(L | K, χ, s). (iv) For any subextension L | M | K, if χ is a character of GL|M , we have L(L | M, χ, s) = L(L | K, χ∗ , s), with χ∗ the induced character. Proof: Q (i) is obvious from the definition: L(L | K, 1, s) = p 1−N1(p)−s . (ii) Let (V, ρ) and (V 0 , ρ0 ) be representations of GL|K with characters χ and χ0 , respectively. Then we know that (V ⊕ V 0 , ρ ⊕ ρ0 ) corresponds to χ + χ0 . Therefore, from det(1 − ϕP t; (V ⊕ V 0 )IP ) = det(1 − ϕP t; V IP ) det(1 − ϕP t; V 0IP ) follows our result. (iii) Let P0 | P | p. The operation of GL0 |K on V is induced by the projection GL0 |K GL|K , which in turn also yields epimorphisms GP0 GP , IP0 IP and GP0 /IP0 GP /IP , such that ϕP0 7→ ϕP . Hence ϕP0 = ϕP : V IP → V IP . (iv) omitted for brevity. See Neu:VII.10.4. 4 Corollary We have the following relation between Dedekind zeta functions: Y ζL (s) = ζK (s) · L(L | K, χ, s)χ(1) , χ6=1 where the product is over all (non-trivial) irreducible characters of GL|K . Proof: Since 1∗ = P χ χ(1)χ, proposition 5 tells us (for any normal L0 | L): ζL (s) (i) = L(L0 | L, 1, s) (iv) L(L0 | K, 1∗ , s) = (iii) = (ii) = (i) = L(L | K, 1∗ , s) Q χ(1) χ L(L | K, χ, s) Q ζK (s) · χ6=1 L(L | K, χ, s)χ(1) . The Artin conjecture The Artin L-function L(L | K, χ, s) is analytic on the whole complex plane, for all irreducible χ 6= 1. In particular, by the corollary, ζζKL (s) is holomorphic on C. (s) We will now prove this for 1-dimensional representations, i.e. for L | K abelian. Theorem Let L | K be abelian, χ 6= 1 an irreducible character of GL|K . Then there is a Hecke character χ̃, such that L(L | K, χ, s) = L(χ̃, s), the Hecke L-function (which is analytic on C). Proof: By Artin reciprocity, CK /NL|K CL = GL|K , so χ induces a character χ̃ : CK → C× . For each P | p, we have χ̃p |Up 6= 1 ⇔ χ(IP ) 6= 1 ⇔ CIP = 0, so the Euler factors of L(L | K, χ, s) at those p are trivial, as required. For the other places, CIP = C, hence det(1 − ϕP N (p)−s ; V IP ) = 1 − ϕP N (p)−s . Now, by proposition 1, if πp is a local uniformizer, ϕP = χ(ϕP ) = χ(πp , LP | Kp ) = χ̃(. . . , 1, πp , 1, . . .) = χ̃p (πp ). Taking all this together, we see that indeed: Y Y 1 1 = = L(χ̃, s). L(L | K, χ, s) = −s −s 1 − ϕ N (p) 1 − χ̃ (π )N (p) P p p p p 5 Corollary Let L | K be abelian. The Artin L-function L(L | K, χ, s) has a meromorphic continuation to all of C, for all characters χ of GL|K . Proof: By Brauer’s theorem on induced characters, we can write χ = r P ni χi∗ , with ni ∈ Z, i=1 and χi irreducible 1-dimensional characters of GL|Ki . Thus, by proposition 5, L(L | K, χ, s) = r Y ni L(L | K, χi∗ , s) = i=1 r Y ni L(L | Ki , χi , s) i=1 = r Y L(χ̃i , s)ni . i=1 Literature [Bac] My bachelor thesis. Reciprocity, and the Brauer-Hasse-Noether and Hasse-Minkowski theorems for algebraic number fields. [Neu] Neukirch, Jürgen. Algebraic Number Theory. 6