Bourbaki

§ 11. Abelian extensions

Throughout this paragraph K denotes a field.

1. Abelian extensions and the abelian closure

Definition 1

— An extension E of K is said to be abelian if it is Galois and its Galois group is commutative.

Since every subgroup of a commutative group is normal, Cor. 4 of V, p. 68 shows that every subextension of an abelian extension is abelian.

Proposition 1

— Let E be a Galois extension of K and Γ its Galois group. Let A be the derived group of Γ (I, p. 10, Def. 4) and F the field of invariants of A. For a subextension L to E to be abelian it is necessary and sufficient that it should be contained in F.

Firstly we note that F is also the field of invariants of the closure Δˉ\bar{\Delta} of A in Γ, and that Δˉ\bar{\Delta} is a closed normal subgroup of Γ. By V, p. 68, Cor. 4, F is thus a Galois extension of K. Further, the Galois group of F over K is isomorphic to Γ/Δˉ\Gamma / \bar{\Delta} and hence is commutative. Hence every subextension of F is abelian. Conversely let L be an abelian extension of K contained in E, and let Π be the Galois group of E over L. Since L is Galois, Π is a normal subgroup of Γ and the Galois group of L over K is isomorphic to Γ/Π\Gamma / \Pi (V, p. 68, Cor. 4). Therefore Γ/Π\Gamma / \Pi is commutative and Π contains A, whence L⊂FL \subset F.

Corollary

— Let E be an extension of K and let (Ei)i∈I(E_i)_{i \in I} be a family of subextensions of E such that E=K(⋃i∈IEi)E = K \left( \bigcup_{i \in I} E_i \right). Suppose that each extension EiE_i is abelian, then the same is true of E.

In the first place E is a Galois extension of K (V, p. 57, Prop. 1). If the field F is defined as in Prop. 1, then we have Ei⊂FE_i \subset F for all i∈Ii \in I, whence E=FE = F.

An extension E of K is said to be an abelian closure of K if it is an abelian extension of K and every abelian extension of K is isomorphic to a subextension of E. Prop. 1 implies the existence of an abelian closure of K: for let KsK_s be a separable closure of K, with Galois group Γ\Gamma and let (Γ‾,Γ‾)(\overline{\Gamma}, \overline{\Gamma}) be the closure of the derived group of Γ\Gamma; denote by KabK_{ab} the field of invariants of (Γ‾,Γ‾)(\overline{\Gamma}, \overline{\Gamma}); since every separable algebraic extension of K is isomorphic to a subextension of KsK_s (V, p. 45, Cor.). Prop. 1 shows that KabK_{ab} is an abelian closure of K. The Galois group of KabK_{ab} over K is canonically isomorphic to Γ/(Γ‾,Γ‾)\Gamma / (\overline{\Gamma}, \overline{\Gamma}). Let us now show the uniqueness of abelian closures: let E, E' be two abelian closures of K; by definition there exist K-homomorphisms u:E→E′u : E \to E' and v:E′→Ev : E' \to E and Prop. 1 (V, p. 52) implies that v(u(E))=Ev(u(E)) = E and u(v(E′))=E′u(v(E')) = E', hence uu is a K-isomorphism of E onto E'. Any other K-isomorphism of E onto E' is of the form u1=σ0∘uu_1 = \sigma_0 \circ u with σ0∈Gal(E′/K)\sigma_0 \in \mathrm{Gal}(E'/K); since Gal(E′/K)\mathrm{Gal}(E'/K) is commutative, the isomorphism σ↦u∘σ∘u−1\sigma \mapsto u \circ \sigma \circ u^{-1} of Gal(E/K)\mathrm{Gal}(E/K) onto Gal(E′/K)\mathrm{Gal}(E'/K) is independent of uu; it is called the canonical isomorphism of Gal(E/K)\mathrm{Gal}(E/K) onto Gal(E′/K)\mathrm{Gal}(E'/K).

2. Roots of unity

Definition 2

An element ζ\zeta of K is said to be a root of unity if there exists an integer n>0n > 0 such that ζn=1\zeta^n = 1; for every integer n>0n > 0 such that ζn=1\zeta^n = 1, ζ\zeta is called an n-th root of unity.

It amounts to the same to say that the roots of unity are the elements of finite order of the multiplicative group K∗K^* of non-zero elements of K (I, p. 51). The roots of unity form a subgroup μ∞(K)\mu_\infty(K) of K∗K^*, the n-th roots form a subgroup μn(K)\mu_n(K) of μ∞(K)\mu_\infty(K). We have μ∞(K)=⋃n≥1μn(K)\mu_\infty(K) = \bigcup_{n \geq 1} \mu_n(K) and μn(K)⊂μm(K)\mu_n(K) \subset \mu_m(K) if nn divides mm. For every root of unity ζ\zeta there exists a least integer n≥1n \geq 1 such that ζ\zeta belongs to μn(K)\mu_n(K), namely the order of ζ\zeta in the group K∗K^*.

The group μn(K)\mu_n(K), being the set of roots of the polynomial Xn−1X^n - 1, is of finite order ≤n\leq n. Let pp be the characteristic of K. When p=0p = 0, or when p≠0p \neq 0 and nn is not divisible by pp, the derivative nXn−1nX^{n-1} of Xn−1X^n - 1 is relatively prime to Xn−1X^n - 1 and so the polynomial Xn−1X^n - 1 is then separable; if moreover K is algebraically closed, μn(K)\mu_n(K) is thus a group of nn elements.

Suppose that K is of non-zero characteristic pp and let r≥0r \geq 0 be an integer; since the mapping x↦xprx \mapsto x^{p^r} of K into K is injective, we have μnpr(K)=μn(K)\mu_{np^r}(K) = \mu_n(K) for every integer n≥1n \geq 1.

We remark that a field may contain no n-th root of unity other than 1: this is the case for example with the prime fields Q\mathbf{Q} and Fp\mathbf{F}_p for any odd integer nn.

Theorem 1

Let pp be the characteristic exponent of K and let n>0n > 0 be an integer. Then the group μn(K)\mu_n(K) of n-th roots of unity in K is cyclic and its order divides nn; when K is algebraically closed and nn is prime to pp, the group μn(K)\mu_n(K) is cyclic of order nn.

It is enough to prove the first assertion of the theorem which is a consequence of the following more precise lemma:

Lemma 1

Let G be a finite subgroup of K∗K^* of order mm; then G is cyclic and we have G=μm(K)G = \mu_m(K).

Consider G as Z-module; we have mx=0mx = 0 for all x∈Gx \in G, hence the annihilator of G is an ideal of the form rZr\mathbf{Z}, where the integer r≥1r \geq 1 divides mm. We thus have G⊂μr(K)G \subset \mu_r(K). By Lemma 4 (V, p. 74) applied with A=ZA = \mathbf{Z} and M=GM = G there exists an element xx of G of order rr; let G′G' be the cyclic subgroup of G generated by xx. We have G′⊂μr(K)G' \subset \mu_r(K), Card (G′)=r(G') = r and Card μr(K)≤r\mu_r(K) \leq r; therefore we have

G' = μ_r(K) ⊃ G and G is cyclic of order r, hence equal to μ_r(K). Since G has order m, we have m = r and the lemma follows.

Proposition 2

— Let K be algebraically closed and let p be its characteristic exponent. There exists an isomorphism of μ_∞(K) onto the group Q/Z[1/p].

We have denoted by Z[1/p] the subring of Q generated by 1/p, that is, the set of rational numbers of the form a/p^n with a ∈ Z and n ≥ 1; thus we have Z[1/p] = Z if K is of characteristic 0.

Let (v_n), be the strictly increasing sequence consisting of all the integers which are not divisible by p, if p ≠ 1; put X = v_1 v_2 ... v_n and denote by H_n the group of λ_n-th roots of unity; we have H_{n+1} ⊃ H_n and μ_∞(K) = U H_n. Since H_n is cyclic of order X, (Th. 1), there exists a sequence (a_n), of roots of unity such that a_n generates H_n and a_n = α_n^{v_n + 1}.

Further let β_n be the residue class mod Z[1/p] of 1/λ_n and let H'_n be the cyclic subgroup of Q/Z[1/p] generated by β_n. It is clear that β_n = v_{n+1} β_{n+1} and H'_n is of order λ_n because X is not divisible by p if p ≠ 1. Thus there exists for all n ≥ 1 an isomorphism φ_n : H_n → H'_n such that φ_n(α_n) = β_n and the relations α_n = α_{n+1}^{v_n + 1}, β_n = v_{n+1} β_{n+1}, show that φ_n+, extends φ_n. Finally there exists a unique isomorphism φ of μ_∞(K) onto Q/Z[1/p] extending the isomorphisms φ_n, that is, φ(α_n) = β_n for all n ≥ 1.

Remark 1

When K is an algebraically closed field of characteristic 0, the group μ_∞(K) is thus isomorphic (not canonically) to Q/Z. When K is the field C of complex numbers, we can write down an explicit such isomorphism: in effect the mapping x ↦ e^{2πix} is a homomorphism of Q into C with kernel Z and image μ_∞(C); thus by passage to quotients it defines an isomorphism of Q/Z onto μ_∞(C).

Remark 2

The following result may be shown (cf. V, p. 165, Ex. 21): let G and H be two commutative groups all of whose elements are of finite order. Suppose that for every integer n ≥ 1, the equation nx = 0 has the same number of solutions, assumed finite in G and in H. Then the groups G and H are isomorphic. This provides a new proof of Prop. 2.

Remark 3

For every prime number l put μ_{l∞}(K) = U μ_{ln}(K). When l is the characteristic p of K, we have μ_{p∞}(K) = {1}. From I, p. 80, Th. 4, we deduce that μ_∞(K) is a direct sum of the subgroups μ_{l∞}(K), where l runs over the set of all prime numbers distinct from p. For a given prime number l only two cases are possible: either μ_{l∞}(K) is finite and then μ_{l∞}(K) is isomorphic to Z/l^nZ for a suitable n (Th. 1), or μ_{l∞}(K) is infinite and then μ_{l∞}(K) is isomorphic to Z[l^{-1}]/Z (cf. Remark 2).

3. Primitive roots of unity

Let n ≥ 1 be an integer. By the Euler indicator of n, written φ(n), we understand the number of invertible elements of the ring Z/nZ of integers modulo n. By the next proposition φ(n)\varphi(n) is also the number of integers kk prime to nn and such that 0≤k<n0 \leq k < n.

Proposition 3

Let kk and n≥1n \geq 1 be two integers. Then the following assertions are equivalent*:

a) the residue class of kk mod nn is invertible in the ring Z/nZ\mathbf{Z}/n\mathbf{Z}; b) the residue class of kk mod nn generates the cyclic group Z/nZ\mathbf{Z}/n\mathbf{Z}; c) the integers kk and nn are relatively prime (I, p. 112).

Each of the conditions a) and b) means that there exists an integer xx such that kx≡1(modn)kx \equiv 1 \pmod{n}, that is, there exist two integers xx and yy such that kx+ny=1kx + ny = 1. This latter condition just means that kk and nn are relatively prime.

Corollary 1

Let GG be a cyclic group of order nn and let dd be a divisor of nn. Then the number of elements of order dd in GG is equal to φ(d)\varphi(d). In particular, φ(n)\varphi(n) is the number of generators of GG.

Since GG is isomorphic to Z/nZ\mathbf{Z}/n\mathbf{Z}, the number of generators of GG is equal to φ(n)\varphi(n) by Prop. 3. Let gg be a generator of GG; then the elements hh of GG such that hd=1h^d = 1 constitute the subgroup HH of GG generated by gn/dg^{n/d}; this group is cyclic of order dd and the elements of order dd of GG are the generators of HH, hence their number is φ(d)\varphi(d).

Corollary 2

For every integer n≥1n \geq 1, we have*

∑d∣nφ(d)=n, \sum_{d \mid n} \varphi(d) = n,
where the integer dd runs over the set of divisors > 0 of nn ^1.

With the notation of Cor. 1, every element of GG has a finite order which is a divisor dd of nn and for a fixed dd there are φ(d)\varphi(d) such elements.

The calculation of φ(n)\varphi(n) is based on the two formulae:

(2) φ(mn)=φ(m)φ(n)\varphi(mn) = \varphi(m)\varphi(n) if mm and nn are relatively prime , (3) φ(pa)=pa−1(p−1)\varphi(p^a) = p^{a-1}(p-1) (pp prime, a≥1a \geq 1).

The first follows at once from the fact that the rings Z/mnZ\mathbf{Z}/mn\mathbf{Z} and (Z/mZ)×(Z/nZ)(\mathbf{Z}/m\mathbf{Z}) \times (\mathbf{Z}/n\mathbf{Z}) are isomorphic (I, p. 112), and that (A×B)∗=A∗×B∗(A \times B)^* = A^* \times B^* for two rings AA and BB. To prove (3) we note that the positive divisors of pap^a are 1,p,p2,…,pa1, p, p^2, \ldots, p^a; hence the integer kk has no common divisor with pap^a other than 11 if and only if it is not divisible by pp; since there are pa−1p^a - 1 multiples of pp between 00 and pa−1p^a - 1, we find indeed (3).

The formulae (2) and (3) show at once that

φ(n)=n∏p(1−1/p), \varphi(n) = n \prod_p (1 - 1/p),
where pp runs over the set of prime divisors of nn.

^1 The relation d∣nd \mid n between integers > 0 means « d divides n » (VI, p. 5).

An n-th root of unity is said to be primitive if it is of order nn; if there exists such a root ζ\zeta, the group μn(K)\mu_n(K) is of order nn and is generated by ζ\zeta. \* For example, the primitive n-th roots of unity in C\mathbf{C} are the numbers e2πik/ne^{2\pi i k / n} with 0≤k<n0 \leq k < n and kk prime to nn. \* Cor. 1 of Prop. 3 now implies the following result.

Proposition 4

— Let n≥1n \geq 1 be an integer; suppose that there exist n n-th roots of unity in K (this is true for example if K is separably closed and n.1,≠0n . 1, \neq 0). Then the number of primitive n-th roots of unity in K is equal to φ(n)\varphi(n).

4. Cyclotomic extensions

Let pp be the characteristic exponent of K and let n≥1n \geq 1 be an integer prime to pp; by a cyclotomic extension of level nn over K we understand any splitting extension E of the polynomial Xn−1X^n - 1 over K (V, p. 21). Since this polynomial is separable, E is a Galois extension of K, of finite degree (V, p. 57). There exists a primitive n-th root of unity in E ; if ζ\zeta is such a root, then every n-th root of unity is a power of ζ\zeta, hence E=K(ζ)E = K(\zeta).

For the rest of this No. we shall choose a separable closure KsK_s of K. For every n≥1n \geq 1 prime to pp the group μn(Ks)\mu_n(K_s) is cyclic of order nn and the field

Rn(K)=K(μn(Ks)) R_n(K) = K(\mu_n(K_s))

is a cyclotomic extension of level nn of K. We can consider μn(Ks)\mu_n(K_s) as a free module of rank 1 over the ring Z/nZ\mathbf{Z}/n\mathbf{Z}, and every element aa of Gal(Ks/K)\mathrm{Gal}(K_s/K) induces an automorphism of μn(Ks)\mu_n(K_s); there exists therefore a homomorphism χn:Gal(Ks/K)→(Z/nZ)∗\chi_n : \mathrm{Gal}(K_s/K) \to (\mathbf{Z}/n\mathbf{Z})^* characterized by the formula u(ζ)=ζju(\zeta) = \zeta^j for every n-th root of unity ζ\zeta in KsK_s, every uu in Gal(Ks/K)\mathrm{Gal}(K_s/K) and every integer jj in the residue class χn(u)\chi_n(u) mod nn. Since Rn(K)=K(μn(Ks))R_n(K) = K(\mu_n(K_s)), the kernel of χn\chi_n is the subgroup Gal(Ks/Rn(K))\mathrm{Gal}(K_s/R_n(K)) of Gal(Ks/K)\mathrm{Gal}(K_s/K); hence we have χn=φn∘ψn\chi_n = \varphi_n \circ \psi_n where ψn\psi_n is the restriction homomorphism of Gal(Ks/K)\mathrm{Gal}(K_s/K) over Gal(Rn(K)/K)\mathrm{Gal}(R_n(K)/K) and φn\varphi_n is an injective homomorphism of Gal(Rn(K)/K)\mathrm{Gal}(R_n(K)/K) in (Z/nZ)∗(\mathbf{Z}/n\mathbf{Z})^*. In particular we have the following result :

Proposition 5

— For every integer n≥1n \geq 1 prime to pp the extension Rn(K)R_n(K) of K is abelian of finite degree, its Galois group is isomorphic to a subgroup of (Z/nZ)∗(\mathbf{Z}/n\mathbf{Z})^* and its degree divides the order φ(n)\varphi(n) of (Z/nZ)∗(\mathbf{Z}/n\mathbf{Z})^*.

Let Q be an algebraic closure of the field Q of rational numbers, and let n be an integer. Then the cyclotomic polynomial Φn\Phi_n of level nn is defined by

Φn(X)=∏ζ∈Sn(X−ζ), \Phi_n(X) = \prod_{\zeta \in S_n} (X - \zeta),

where SnS_n is the set of primitive n-th roots of unity in QQ. The polynomial Φn\Phi_n is of degree φ(n)\varphi(n) (Prop. 4). It is clear that Φn(X)\Phi_n(X) is invariant under all automorphisms of Qˉ\bar{\mathbf{Q}}, and so belongs to Q[X]\mathbf{Q}[X]. Since every element ζ\zeta of SnS_n is a root of the polynomial Xn−1X^n - 1, the polynomial Φn(X)\Phi_n(X) divides Xn−1X^n - 1, and the following lemma shows that Φn(X)\Phi_n(X) is a monic polynomial with integer coefficients.

Lemma 2

Let f, g and h be monic polynomials in Q[X]\mathbf{Q}[X] such that f=ghf = gh. If f has integer coefficients, the same is true of g and h.

Let aa (resp. bb) be the least of the integers α≥3\alpha \geq 3 (resp. β≥1\beta \geq 1) such that αg\alpha g (resp. βh\beta h) has integer coefficients; we put g′=agg' = ag and h′=bhh' = bh and show by reductio ad absurdum that a=b=1a = b = 1. If this were not so, there would be a prime divisor pp of abab. If u∈Z[X]u \in \mathbf{Z}[X], let us write uˉ\bar{u} for the polynomial with coefficients in the field Fp\mathbf{F}_p obtained by reduction mod pp of the coefficients of uu. We have g′h′=abfg'h' = abf, whence gˉ′hˉ′=0\bar{g}'\bar{h}' = 0; since the ring Fp[X]\mathbf{F}_p[X] is an integral domain (IV, p. 9, Prop. 8) we thus have gˉ′=0\bar{g}' = 0 or hˉ′=0\bar{h}' = 0. In other words, pp divides all the coefficients of g′g' or all those if h′h' and this contradicts the hypothesis.

We have the relation

Xn−1=∏d∣nΦd(X). X^n - 1 = \prod_{d|n} \Phi_d(X) .

For we have Xn−1=∏ζ∈μn(Q)(X−ζ)X^n - 1 = \prod_{\zeta \in \mu_n(\mathbf{Q})} (X - \zeta) and the sets SdS_d for dd dividing nn form a partition of μn(Q)\mu_n(\mathbf{Q}).

The formula (6) determines Φn(X)\Phi_n(X) when we know Φd(X)\Phi_d(X) for all the divisors d<nd < n of nn; since Φ1(X)=X−1\Phi_1(X) = X - 1, we thus have a recursive procedure for calculating Φn\Phi_n. For example for prime pp we have

Xp−1=(X−1)Φp(X), X^p - 1 = (X - 1)\Phi_p(X) ,

whence

Φp(X)=Xp−1+Xp−2+⋯+X+1, \Phi_p(X) = X^{p-1} + X^{p-2} + \cdots + X + 1 ,

and

Φpr+1(X)=Φp(Xpr)forr≥0. \Phi_{p^r+1}(X) = \Phi_p(X^{p^r}) \quad \text{for} \quad r \geq 0 .

Let us list the values of the polynomials Φn(X)\Phi_n(X) for 1≤n≤121 \leq n \leq 12:

Φ1(X)=X−1Φ2(X)=X+1Φ3(X)=X2+X+1Φ4(X)=X2+1Φ5(X)=X4+X3+X2+X+1 \begin{align*} \Phi_1(X) &= X - 1 \\ \Phi_2(X) &= X + 1 \\ \Phi_3(X) &= X^2 + X + 1 \\ \Phi_4(X) &= X^2 + 1 \\ \Phi_5(X) &= X^4 + X^3 + X^2 + X + 1 \\ \end{align*}
Φ6(X)=X2−X+1 \Phi_6(X) = X^2 - X + 1
Φ7(X)=X6+X5+X4+X3+X2+X+1 \Phi_7(X) = X^6 + X^5 + X^4 + X^3 + X^2 + X + 1
Φ8(X)=X4+1 \Phi_8(X) = X^4 + 1
Φ9(X)=X6+X3+1 \Phi_9(X) = X^6 + X^3 + 1
Φ10(X)=X4−X3+X2−X+1 \Phi_{10}(X) = X^4 - X^3 + X^2 - X + 1
Φ11(X)=X10+X9+X8+X7+X6+X5+X4+X3+X2+X+1 \Phi_{11}(X) = X^{10} + X^9 + X^8 + X^7 + X^6 + X^5 + X^4 + X^3 + X^2 + X + 1
Φ12(X)=X4−X2+1 \Phi_{12}(X) = X^4 - X^2 + 1

The values of Φ1,Φ2,Φ3,Φ4,Φ5,Φ7,Φ8,Φ9\Phi_1, \Phi_2, \Phi_3, \Phi_4, \Phi_5, \Phi_7, \Phi_8, \Phi_9 and Φ11\Phi_{11} follow directly from (7); now we have Φ1Φ2Φ3Φ6=X6−1\Phi_1 \Phi_2 \Phi_3 \Phi_6 = X^6 - 1 and Φ1Φ2Φ3Φ4Φ6Φ12=X12−1\Phi_1 \Phi_2 \Phi_3 \Phi_4 \Phi_6 \Phi_{12} = X^{12} - 1, whence Φ4Φ12=X12−1X6−1−X6+1\Phi_4 \Phi_{12} = \frac{X^{12} - 1}{X^6 - 1} - X^6 + 1 and finally Φ12=X6+1X2+1=X4−X2+1\Phi_{12} = \frac{X^6 + 1}{X^2 + 1} = X^4 - X^2 + 1. The cases n=6n = 6 and n=10n = 10 may be treated similarly.

Remark

\* For every integer n>0n > 0 a function μ(n)\mu(n) is defined as follows: if nn is divisible by the square of a prime number, we put μ(n)=0\mu(n) = 0, otherwise μ(n)=(−1)h\mu(n) = (-1)^h if nn is the product of hh distinct prime numbers (« Mobius function »). It may be shown that

Φn(X)=∏d∣n(Xn/d−1)μ(d), \Phi_n(X) = \prod_{d|n} (X^{n/d} - 1)^{\mu(d)},

or more explicitly

Φn(X)=∏p1<…<ph(Xn/p1⋯ph−1)(−1)h \Phi_n(X) = \prod_{p_1 < \ldots < p_h} (X^{n/p_1 \cdots p_h} - 1)^{(-1)^h}

where (p1,…,ph)(p_1, \ldots, p_h) runs over the set of all strictly increasing sequences of prime divisors of nn (cf. Lie, II, p. 207, Ex. 1).

5. Irreducibility of cyclotomic polynomials

Let pp be the characteristic exponent of KK and let nn be an integer prime top. We denote by Φn∈K[X]\Phi_n \in K[X] the image of the polynomial with integer coefficients Φn\Phi_n under the unique homomorphism of Z[X]\mathbf{Z}[X] into K[X]K[X] which maps XX to XX.

Lemma 3

The roots of Φn\Phi_n in KsK_s are the primitive nn-th roots of unity.

Denote by SS, the set of roots of Φn\Phi_n in KsK_s. By Formula (6), the set μn(Ks)\mu_n(K_s) is the union of the SdS_d for dd dividing nn. Every primitive nn-th root of unity thus belongs to SnS_n and the lemma now follows from Prop. 4 (V, p. 81).

Proposition 6

— Let pp be the characteristic exponent of KK and let n≥1n \geq 1 be an integer prime to pp. For the polynomial Φn(X)\Phi_n(X) to be irreducible in K[X]K[X] it is necessary and sufficient that the homomorphism χn:Gal(Ks/K)→(Z/nZ)∗\chi_n : \mathrm{Gal}(K_s/K) \to (\mathbf{Z}/n\mathbf{Z})^* should be surjective.

By Lemma 3 we have Rn(K)=K(ζ)R_n(K) = K(\zeta) for each root ζ\zeta of Φn(X)\Phi_n(X) and hence Φn(X)\Phi_n(X) is irreducible in K[X]K[X] if and only if the degree φ(n)\varphi(n) of Φn(X)\Phi_n(X) is equal to [Rn(K):K][R_n(K):K]. Further, the Galois group of Rn(K)R_n(K) over KK is of order [Rn(K):K][R_n(K):K] and it is isomorphic to the subgroup of (Z/nZ)∗(\mathbf{Z}/n\mathbf{Z})^* which is the image of χn\chi_n. Now Prop. 6 follows from the fact that (Z/nZ)∗(\mathbf{Z}/n\mathbf{Z})^* is of order φ(n)\varphi(n).

Theorem 2 (Gauss)

— Let QQ be an algebraic closure of QQ and let n≥1n \geq 1 be an integer.

a) The cyclotomic polynomial Φn(X)\Phi_n(X) is irreducible in Q[X]Q[X].

b) The degree of Rn(Q)R_n(Q) over QQ is φ(n)\varphi(n).

c) The homomorphism χn\chi_n of Gal(Qˉ/Q)\mathrm{Gal}(\bar{Q}/Q) into (Z/nZ)∗(\mathbf{Z}/n\mathbf{Z})^* is surjective and defines by passage to quotients an isomorphism of Gal(Rn(Q)/Q)\mathrm{Gal}(R_n(Q)/Q) onto (Z/nZ)∗(\mathbf{Z}/n\mathbf{Z})^*.

Taking account of Prop. 6, we need only prove c). Every integer rr prime to nn is a product of prime numbers p1,…,psp_1, \ldots, p_s not dividing nn; so it is enough to show that for every prime number pp not dividing nn, the mapping x↦xpx \mapsto x^p of μn(Q)\mu_n(Q) into itself extends to an automorphism of Rn(Q)R_n(Q). It suffices to prove that if ζ\zeta is a primitive nn-th root of unity, the minimal polynomial ff of ζ\zeta over QQ is equal to the minimal polynomial gg of ζp\zeta^p over QQ.

We shall assume that f≠gf \neq g and argue by contradiction. The polynomials ff and gg are monic irreducible in Q[X]Q[X] and divide Xn−1X^n - 1, so there exists u∈Q[X]u \in Q[X] such that Xn−1=fguX^n - 1 = fg u (IV, p. 13, Prop. 13). Lemma 2 (V, p. 82) shows that f,gf, g and uu have integer coefficients. Let us denote by vˉ\bar{v} the polynomial with coefficients in Fp\mathbf{F}_p obtained from a polynomial v∈Z[X]v \in \mathbf{Z}[X] by reduction mod pp. We thus have Xn−1=fˉgˉuˉX^n - 1 = \bar{f} \bar{g} \bar{u} in Fp[X]\mathbf{F}_p[X].

Further we have g(ζp)=0g(\zeta^p) = 0 and so g(Xp)g(X^p) is a multiple of f(X)f(X) in Q[X]\mathbf{Q}[X]. By Lemma 2 there exists h∈Z[X]h \in \mathbf{Z}[X] such that g(Xp)=f(X)⋅h(X)g(X^p) = f(X) \cdot h(X). Now we have v(Xp)=v(X)pv(X^p) = v(X)^p for every polynomial v∈Fp[X]v \in \mathbf{F}_p[X]. By reduction mod pp we thus obtain gˉp=fˉhˉ\bar{g}^p = \bar{f} \bar{h}. If vv is an irreducible polynomial in Fp[X]\mathbf{F}_p[X] dividing fˉ\bar{f}, it must then divide gˉ\bar{g}. Since fˉgˉ\bar{f} \bar{g} divides Xn−1X^n - 1, we conclude that v2v^2 divides Xn−1X^n - 1 in Fp[X]\mathbf{F}_p[X]. This is absurd because the polynomial Xn−1X^n - 1 is separable in Fp[X]\mathbf{F}_p[X].

It may be shown that for every abelian extension EE of finite degree over QQ there exists an integer n≥1n \geq 1 such that EE is isomorphic to a subextension of Rn(Q)R_n(Q). \* In other words, the field Q(μ∞(C))Q(\mu_{\infty}(C)) is an abelian closure of QQ. *(« Kronecker-Weber theorem »).

6. Cyclic extensions

Definition 3

— *An extension E of K is said to be cyclic if it is Galois and its Galois group is cyclic.

*Examples. — 1) Every Galois extension of prime degree is cyclic, because every finite group G of prime order pp is cyclic (for every element x≠1x \neq 1 of G is of order pp, hence generates G).

2) Let F(X)=X2+aX+bF(X) = X^2 + aX + b be an irreducible polynomial in K[X]K[X]. The only case where F(X)F(X) is not separable is that where KK is of characteristic 2 and a=0a = 0. We shall leave this case aside; let E be an extension of K generated by a root xx of F(X)F(X). We have [E:K]=2[E : K] = 2 and F(X)=(X−x)(X+a+x)F(X) = (X - x)(X + a + x), hence E is a Galois extension of K. The Galois group of E over K is of order 2, hence cyclic.

3) Let F\mathbf{F} be a field and σ\sigma an automorphism of finite order. The field E of invariants of σ\sigma is also the field of invariants of the cyclic group of order nn generated by σ\sigma, and hence (V, p. 66, Th. 3) F is a cyclic extension of degree nn of E.

We know (I, p. 50) that every subgroup and every quotient group of a cyclic group is cyclic. Therefore (V, p. 68, Cor. 4), if E is a cyclic extension of a field K, of degree nn, then every subextension F of E is cyclic over K, and E is cyclic over F. For every divisor d of nn there exists a unique subfield F of degree d over K contained in E : for in a cyclic group of order nn there exists a unique subgroup of index d.

Theorem 3 (Hilbert)

— *Let E be a cyclic extension of K and let a be a generator of the Galois group Γ\Gamma of E over K.

a) For an element x∈Ex \in E to satisfy NE/K(x)=1N_{E/K}(x) = 1 it is necessary and sufficient that y∈E∗y \in E^* exist such that x=y/σ(y)x = y / \sigma(y); every y1∈E∗y_1 \in E^* with x=y1/σ(y1)x = y_1 / \sigma(y_1) is then of the form λy\lambda y where λ∈K∗\lambda \in K^*.

b) For an element x∈Ex \in E to satisfy TrE/K(x)=0\mathrm{Tr}_{E/K}(x) = 0 it is necessary and sufficient that z∈Ez \in E exist with x=z−σ(z)x = z - \sigma(z); then every z1∈Ez_1 \in E with x=z1−σ(z1)x = z_1 - \sigma(z_1) is of the form z+μz + \mu, where μ∈K\mu \in K.

Let us first prove a lemma.*

Lemma 4

Let Γ\Gamma be a cyclic group of order nn, σ\sigma a generator of Γ\Gamma and M a commutative group on which Γ\Gamma operates according to the rule γ.(m+m′)=γ.m+γ.m′\gamma . (m + m') = \gamma . m + \gamma . m' for all γ∈Γ\gamma \in \Gamma and m,m′∈Mm, m' \in M. Let Z be the set of all mappings of Γ\Gamma into M satisfying the relation

f(ττ′)=f(τ)+τ.f(τ′)for τ,τ′ in Γ f(\tau \tau') = f(\tau) + \tau . f(\tau') \quad \text{for } \tau, \tau' \text{ in } \Gamma

Put u(f)=f(a)u(f) = f(a) for f∈Zf \in Z and t(m)=∑τ∈Γτ.mt(m) = \sum_{\tau \in \Gamma} \tau . m for m∈Mm \in M. Then the sequence

0→Z→uM→tM 0 \to Z \xrightarrow{u} M \xrightarrow{t} M

is exact.

Let f∈Zf \in \mathbf{Z}; taking τ=τ′=1\tau = \tau' = 1 in (10) we find that f(1)=0f(1) = 0. Further, by induction on m⩾0m \geqslant 0 we deduce the relation

f(σm)=f(σ)+σ⋅f(σ)+⋯+σm−2⋅f(σ)+σm−1⋅f(σ) f(\sigma^m) = f(\sigma) + \sigma \cdot f(\sigma) + \cdots + \sigma^{m-2} \cdot f(\sigma) + \sigma^{m-1} \cdot f(\sigma)

We have σn=1\sigma^n = 1, whence f(σn)=0f(\sigma^n) = 0; the preceding relation with m=nm = n is equivalent to the equality t(u(f))=0t(u(f)) = 0, whence Im⁡ u⊂Ker⁡ t\operatorname{Im}\ u \subset \operatorname{Ker}\ t. Moreover, from (11) it follows that Ker⁡ u=0\operatorname{Ker}\ u = 0.

Let m∈Mm \in M be such that t(m)=0t(m) = 0, that is, m+σ⋅m+⋯+σn−1⋅m=0m + \sigma \cdot m + \cdots + \sigma^{n-1} \cdot m = 0. Let us define the mapping ff of Γ\Gamma into MM by

f(σj)=m+σ⋅m+⋯+σj−2⋅m+σj−1⋅m f(\sigma^j) = m + \sigma \cdot m + \cdots + \sigma^{j-2} \cdot m + \sigma^{j-1} \cdot m

for 0⩽j⩽n−10 \leqslant j \leqslant n - 1. It may be left to the reader to establish the relation (10). We clearly have m=f(a)m = f(a), whence Im⁡ u⊃Ker⁡ t\operatorname{Im}\ u \supset \operatorname{Ker}\ t.

With the lemma now proved, let us take y∈E∗y \in E^* and x=y/σ(y)x = y/\sigma(y); we have NE/K(σ(y))=NE/K(y)N_{E/K}(\sigma(y)) = N_{E/K}(y), whence NE/K(x)=1N_{E/K}(x) = 1. Conversely, let xx in E∗E^* be such that NE/K(x)=1N_{E/K}(x) = 1; by Lemma 4, applied to M=E∗M = E^*, there exists a family of elements (cτ)τ∈Γ(c_\tau)_{\tau \in \Gamma} of E∗E^* satisfying the relation cττ′=cτ⋅τ(cτ′)c_{\tau \tau'} = c_\tau \cdot \tau(c_{\tau'}) for τ,τ′\tau, \tau' in Γ\Gamma and cσ=xc_\sigma = x. By Cor. 1 of Prop. 9 (V, p. 65) there exists y∈E∗y \in E^* with cτ=y/τ(y)c_\tau = y/\tau(y) for all τ∈Γ\tau \in \Gamma, whence in particular x=cσ=y/σ(y)x = c_\sigma = y/\sigma(y). If y1∈E∗y_1 \in E^* satisfies x=y1/σ(y1)x = y_1/\sigma(y_1), then we have

σ(y1y−1)=y1y−1, \sigma(y_1 y^{-1}) = y_1 y^{-1},

hence y1y−1y_1 y^{-1} belongs to K∗K^* because aa generates the Galois group of EE over KK. This proves a).

The assertion b) follows in similar fashion from Cor. 2 of Prop. 9 (V, p. 65).

7. Duality of Z/nZ\mathbf{Z}/n\mathbf{Z}-modules

In this No. we denote by nn an integer >0> 0 and by TT a cyclic group of order nn. A group GG is said to be annihilated by nn if gn=1g^n = 1 for all g∈Gg \in G; if moreover GG is commutative, the group structure of GG is underlying a unique Z/nZ\mathbf{Z}/n\mathbf{Z}-module structure.

For every group GG we denote by Hom⁡(G,T)\operatorname{Hom}(G, T) the group of homomorphisms of GG into TT; it is a commutative group annihilated by nn.

Proposition 7

— Let GG be a commutative group annihilated by nn and HH a subgroup of GG. Then the restriction homomorphism Hom⁡(G,T)→Hom⁡(H,T)\operatorname{Hom}(G, T) \to \operatorname{Hom}(H, T) is surjective.

For let f:H→Tf : H \to T be a homomorphism; we shall prove that there exists a homomorphism of GG into TT extending ff. Suppose first that GG is cyclic, generated by an element gg of order rr dividing nn; denote by tt a generator of TT. There exists a divisor s of r such that H is generated by gsg^s (I, p. 50, Prop. 19), and for every x∈Zx \in \mathbf{Z} we have f(gsx)=taxf(g^{sx}) = t^{ax}, where a is an integer such that n divides ar/sar/s. Then a/s=(ar/ns)(n/r)a/s = (ar/ns)(n/r) is an integer and the homomorphism gx↦t(a/s)xg^x \mapsto t^{(a/s)x}, x∈Zx \in \mathbf{Z}, of G into T extends f. In the general case consider the set of pairs (H′,f′)(H', f') where H′H' is a subgroup of G containing H and f′f' is a homomorphism of H′H' into T extending ff, and let us order this set by the relation (H′,f′)≤(H′′,f′′)(H', f') \leq (H'', f'') if H′⊂H′′H' \subset H'' and the restriction of f′′f'' to H′H' is f′f'. By Set Theory, III, p. 154, Def. 3 and Th. 2, this set has a maximal element (H1,f1)(H_1, f_1) and it suffices to prove that H1=GH_1 = G; if this is not the case, there exists g∈G,g∉H1g \in G, g \notin H_1 and it is enough to prove that f1f_1 may be extended to a homomorphism T of the subgroup of G generated by H1H_1 and g. Now if C denotes the cyclic group generated by g, the restriction of f1f_1 to C∩H1C \cap H_1 extends to a homomorphism f2f_2 of C into T and the homomorphism xy↦f1(x)f2(y),x∈H1,y∈C,xy \mapsto f_1(x) f_2(y), x \in H_1, y \in C, of H1CH_1C into T is the desired mapping.

Corollary 1

— If G is a commutative group annihilated by n, and G≠{1}G \neq \{1\}, then Hom(G,T)≠{1}\mathrm{Hom}(G, T) \neq \{1\}.

For it suffices to note that if H is a cyclic subgroup of G distinct from {1}\{1\}, then Hom(H,T)≠{1}\mathrm{Hom}(H, T) \neq \{1\}, and to apply Prop. 7.

Corollary 2

— If G is a finite commutative group annihilated by n, then the groups G and Hom(G,T)\mathrm{Hom}(G, T) have the same order.

If G is cyclic of order r, with generator g, then the mapping f↦f(g)f \mapsto f(g) is a bijection of Hom(G,T)\mathrm{Hom}(G, T) onto the set of elements t of T such that tr=1t^r = 1, hence the assertion follows in this case. On the other hand, if H is a cyclic subgroup of G, we have Card(G)=Card(H)⋅Card(G/H)\mathrm{Card}(G) = \mathrm{Card}(H) \cdot \mathrm{Card}(G/H); further, we have an exact sequence

{1}→Hom(G/H,T)→Hom(G,T)→Hom(H,T)→{1} \{1\} \to \mathrm{Hom}(G/H, T) \to \mathrm{Hom}(G, T) \to \mathrm{Hom}(H, T) \to \{1\}

(II, p. 227, Th. 1 and Prop. 7 above), hence

Card(Hom(G,T))=Card(Hom(H,T))⋅Card(Hom(G/H,T)). \mathrm{Card}(\mathrm{Hom}(G, T)) = \mathrm{Card}(\mathrm{Hom}(H, T)) \cdot \mathrm{Card}(\mathrm{Hom}(G/H, T)).

Since Card(Hom(H,T))=Card(H)\mathrm{Card}(\mathrm{Hom}(H, T)) = \mathrm{Card}(H), it comes to the same to prove the corollary for G or for G/HG/H. Now the result follows by induction on Card(G)\mathrm{Card}(G).

Now let G and H be two groups and f:G×H→Tf : G \times H \to T a bimultiplicative mapping, that is, such that for any g,g′∈G,h,h′∈Hg, g' \in G, h, h' \in H we have

f(gg′,h)=f(g,h)f(g′,h),f(g,hh′)=f(g,h)f(g,h′) f(gg', h) = f(g, h) f(g', h), \quad f(g, hh') = f(g, h) f(g, h')

We define group homomorphisms

sf:G→Hom(H,T),df:H→Hom(G,T), s_f : G \to \mathrm{Hom}(H, T), \quad d_f : H \to \mathrm{Hom}(G, T),

by sf(g)(h)=df(h)(g)=f(g,h)s_f(g)(h) = d_f(h)(g) = f(g, h) (cf. II, p. 268, Cor. to Prop. 1, when G and H are commutative).

Proposition 8

— Suppose that G and H are commutative and annihilated by n. If sf is bijective and H is finite, then d_f is bijective and we have Card(G) = Card(H).

If s_f is bijective and H is finite, we have by Cor. 2 to Prop. 7 the relation Card(G) = Card(Hom(H, T)) = Card(H), hence Card(G) is finite and by another application of the Corollary, Card(Hom(G, T)) = Card(H). So it is enough to prove that d_f is injective. Now if h \in \mathrm{Ker}(d_f), we have f(g, h) = 1 for all g \in G, hence since s_f is bijective, \varphi(h) = 1 for all \varphi \in \mathrm{Hom}(H, T); by Prop. 7 this implies \mathrm{Hom}(\mathrm{Ker}(d_f), T) = \{1\}, hence \mathrm{Ker}(d_f) = \{1\} by Cor. 1 to Prop. 7.

8. Kummer theory

In this No. we denote by nn an integer > 0, and we shall suppose that μn(K)\mu_n(K) has nn elements; by V, p. 78 this means also that nn is prime to the characteristic exponent of K and that all the n-th roots of unity in an algebraic closure Ω\Omega of K lie in K.

We shall say that an extension L of K is abelian of exponent dividing n if it is abelian (V, p. 77, Def. 1) and its Galois group Gal(L/K) is annihilated by nn (V, p. 86).

Let A be a subset of K∗K^*; we denote by K(A1/n)K(A^{1/n}) the subextension of Ω\Omega generated by all 0∈Ω0 \in \Omega such that θn∈A\theta^n \in A.

Lemma 5

— *K(A1/n)K(A^{1/n}) is an abelian extension of K of exponent dividing n.

Since the polynomials Xn−a,a∈AX^n - a, a \in A are separable over K, L=K(A1/n)L = K(A^{1/n}) is a separable, hence a Galois extension of K. Let σ∈Gal(L/K)\sigma \in \mathrm{Gal}(L/K) and let 0∈Ω0 \in \Omega be such that θn∈A\theta^n \in A. We have σ(θ)n=θn\sigma(\theta)^n = \theta^n; hence there exists ζ∈μn(Ω)=μn(K)\zeta \in \mu_n(\Omega) = \mu_n(K) such that σ(θ)=ζθ\sigma(\theta) = \zeta \theta; this implies that σn(θ)=ζnθ=0\sigma^n(\theta) = \zeta^n \theta = 0, whence σn=1\sigma^n = 1. If σ′\sigma' is another element of Gal(L/K)\mathrm{Gal}(L/K) there exists ζ′∈μn(K)\zeta' \in \mu_n(K) such that σ′(θ)=ζ′θ\sigma'(\theta) = \zeta' \theta, whence σ′σ(θ)=ζζ′θ=σσ′(θ)\sigma'\sigma(\theta) = \zeta \zeta' \theta = \sigma\sigma'(\theta) and so σ′σ=σσ′\sigma'\sigma = \sigma\sigma'.

Lemma 6

— *Let L be a Galois extension of K. There exists a unique mapping (σ,a)↦⟨σ,a⟩(\sigma, a) \mapsto \langle \sigma, a \rangle of Gal(L/K)×((Ln∩K∗)/K∗n)\mathrm{Gal}(L/K) \times ((L^n \cap K^*)/K^{*n}) into μn(K)\mu_n(K) such that for every σ∈Gal(L/K)\sigma \in \mathrm{Gal}(L/K) and every element 0∈L∗0 \in L^* such that θn∈K\theta^n \in K we have, on denoting by θn‾\overline{\theta^n} the residue class of θn\theta^n mod K∗nK^{*n}:

⟨σ,θn‾⟩=σ(θ)/θ \langle \sigma, \overline{\theta^n} \rangle = \sigma(\theta)/\theta

This mapping is bimultiplicative.

In effect the right-hand side of (13) is an n-th root of unity which only depends on the residue class mod K∗nK^{*n} of θn\theta^n; this proves the first assertion. The second assertion is verified without difficulty.

For every Galois extension L of K let us write

kL:(Ln∩K∗)/K∗n→Hom(Gal(L/K),μn(K)), k_L : (L^n \cap K^*)/K^{*n} \to \mathrm{Hom}(\mathrm{Gal}(L/K), \mu_n(K)) ,
kL′:Gal(L/K)→Hom((Ln∩K∗)/K∗n,μn(K)), k'_L : \mathrm{Gal}(L/K) \to \mathrm{Hom}((L^n \cap K^*)/K^{*n}, \mu_n(K)) ,

for the homomorphisms derived from the above bimultiplicative mapping (V, p. 87).

Proposition 9

For every Galois extension of finite degree L of K, the homomorphism kLk_L is bijective.

Let 0∈L∗0 \in L^* be such that 0n∈K0^n \in K and the residue class of 0nmod  K∗n0^n \mod K^{*n} belongs to the kernel of kLk_L. For every σ∈Gal(L/K)\sigma \in \mathrm{Gal}(L/K) we have by definition σ(θ)=0\sigma(\theta) = 0; hence 0∈K∗0 \in K^* and θn∈K∗n\theta^n \in K^{*n}. This proves the injectivity of kLk_L. Now let f:Gal(L/K)→μn(K)f : \mathrm{Gal}(L/K) \to \mu_n(K) be a homomorphism; for all σ,τ∈Gal(L/K)\sigma, \tau \in \mathrm{Gal}(L/K) we have

f(στ)=f(σ)f(τ)=f(σ)⋅σf(τ),f(σ)n=1. f(\sigma \tau) = f(\sigma) f(\tau) = f(\sigma) \cdot \sigma f(\tau) , \quad f(\sigma)^n = 1 .

By V, p. 65, Cor. 1 there exists 0∈L∗0 \in L^* such that f(σ)=σ(θ)/θf(\sigma) = \sigma(\theta)/\theta for all σ∈Gal(L/K)\sigma \in \mathrm{Gal}(L/K); since f(σ)n=1f(\sigma)^n = 1, we have σ(θn)=0n\sigma(\theta^n) = 0^n for all σ∈Gal(L/K)\sigma \in \mathrm{Gal}(L/K), hence 0n∈K∗0^n \in K^*; if aa is the residue class of θnmod  K∗n\theta^n \mod K^{*n}, we have by definition f(σ)=⟨σ,a⟩f(\sigma) = \langle \sigma, a \rangle for σ∈Gal(L/K)\sigma \in \mathrm{Gal}(L/K), hence f=kL(a)f = k_L(a).

Corollary

If L is a Galois extension of K, the homomorphism kLk_L is injective and its image is the group Homc(Gal(L/K),μn(K))\mathrm{Hom}_c(\mathrm{Gal}(L/K), \mu_n(K)) of continuous homomorphisms of the topological group Gal(L/K)\mathrm{Gal}(L/K) into the discrete group μn(K)\mu_n(K).

This follows at once from what has been said, using the fact that L is an increasing directed union of Galois extensions LiL_i of finite degree and that the homomorphism of Gal(L/K)\mathrm{Gal}(L/K) into μn(K)\mu_n(K) is continuous if and only if it can be factored through one of the quotients Gal(Li/K)\mathrm{Gal}(L_i/K) of Gal(L/K)\mathrm{Gal}(L/K).

Theorem 4

a) The mapping H↦K(H1/n)H \mapsto K(H^{1/n}) is an inclusion preserving bijection of the set of subgroups of K∗K^* containing K∗nK^{*n} onto the set of abelian subextensions of exponent dividing n of Ω\Omega. The inverse mapping is L↦L′′∩K∗L \mapsto L'' \cap K^*.

b) For every subgroup H of K∗K^* containing K∗nK^{*n} the homomorphism

k′:Gal(K(H1/n)/K)→Hom(H/K∗n,μn(K)) k' : \mathrm{Gal}(K(H^{1/n})/K) \to \mathrm{Hom}(H/K^{*n}, \mu_n(K))

is bijective, and it is a homeomorphism when the group Hom(H/K∗n,μn(K))\mathrm{Hom}(H/K^{*n}, \mu_n(K)) is equipped with the topology of simple convergence.

c) Let H be a subgroup of K∗K^* containing K∗"nK^{*"n}. For every a∈H/K∗na \in H/K^{*n} let θa\theta_a be an element of Ω\Omega such that θan\theta_a^n is a representative of a in H. Then the θa,a∈H/K∗n\theta_a, a \in H/K^{*n}, form a basis of the vector K-space K(H1/n)K(H^{1/n}). In particular,

[K(H1/n):K]=(H:K∗"n). [K(H^{1/n}):K] = (H:K^{*"n}) .

A) For every abelian extension L of K of exponent dividing n let us put HL=L′′∩K∗H_L = L'' \cap K^*. If [L:K][L:K] is finite, the homomorphism kL′k'_L of Gal(L/K)\mathrm{Gal}(L/K) into Hom(HL,μn(K))\mathrm{Hom}(H_L, \mu_n(K)) is bijective by Prop. 9 and V, p. 88, Prop. 8.

Since every abelian extension of K of exponent dividing nn is an increasing directed union of abelian subextensions of finite degree of exponent dividing nn, we deduce by passage to inverse limits that kL′k'_L is a homeomorphism of topological groups for every abelian extension L of K dividing n.

B) Let L be an abelian extension of finite degree, of exponent dividing nn, of K, and let L′=K(HL1/n)L' = K(H_L^{1/n}); this is a subextension of L ; moreover HL′H_{L'} contains HLH_L and so is equal to it. Since the homomorphisms kL′k'_L and kL′′k'_{L'} are bijective by A) and HL=HL′H_L = H_{L'}, the groups Gal(L/K)(L/K) and Gal(L′/K)(L'/K) have the same order and so are equal. This shows that L′=LL' = L, and hence that L=K(HL1/n)L = K(H_L^{1/n}). If L is an abelian extension of exponent dividing nn of K, we have K(HL1/n)=LK(H_L^{1/n}) = L, because K(HL1/n)K(H_L^{1/n}) is a subextension of L which contains every subextension of finite degree of L.

C) Let H be a subgroup of K∗K^* containing K∗nK^{*n}; put L=K(H1/n)L = K(H^{1/n}), then this is an abelian extension of K of exponent dividing nn (V, p. 88, Lemma 5). We have H⊂HLH \subset H_L, whence we obtain an exact sequence of commutative groups annihilated by nn

{1}→H/K∗n→HL/K∗n→HL/H→{1}. \{1\} \to H/K^{*n} \to H_L/K^{*n} \to H_L/H \to \{1\}.

From this we obtain an exact sequence

{1}→Hom⁡(HL/H,μn(K))→Hom⁡(HL/K∗n,μn(K))→ιHom⁡(H/K∗n,μn(K)), \{1\} \to \operatorname{Hom}(H_L/H, \mu_n(K)) \to \operatorname{Hom}(H_L/K^{*n}, \mu_n(K)) \xrightarrow{\iota} \operatorname{Hom}(H/K^{*n}, \mu_n(K)),

where ι\iota is the restriction homomorphism.

If we identify Hom⁡(HL/K∗n,μn(K))\operatorname{Hom}(H_L/K^{*n}, \mu_n(K)) with Gal(L/K)(L/K) by means of the isomorphism kL′k'_L, the kernel of ι\iota is identified with the set of σ∈Gal⁡(L/K)\sigma \in \operatorname{Gal}(L/K) such that σ(θ)=θ\sigma(\theta) = \theta for all θ∈H1/n\theta \in H^{1/n}. It follows that ι\iota is injective, hence Hom⁡(HL/H,μn(K))\operatorname{Hom}(H_L/H, \mu_n(K)) is reduced to (1) ; by Cor. 1 of V, p. 87 we thus have H=HLH = H_L. This completes the proof of a) and b).

D) Let us prove c). If a,b∈Ha, b \in H, we have θaθb/θab∈K\theta_a \theta_b / \theta_{ab} \in K. It follows that the vector subspace of K(H1/n)K(H^{1/n}) generated by the θa\theta_a is stable under multiplication and so coincides with K(H1/n)K(H^{1/n}). It only remains to show that the θa\theta_a are linearly independent ; to do this we may evidently assume that H/K∗nH/K^{*n} is finite ; then [K(H1/n):K]=(Gal⁡(K(H1/n)/K):{1})=(H:K∗n)[K(H^{1/n}) : K] = (\operatorname{Gal}(K(H^{1/n})/K) : \{1\}) = (H : K^{*n}) by b) and Cor. 2 of V, p. 87 ; since the number of θa\theta_a is equal to (H:K∗n)(H : K^{*n}) and they generate the vector K-space K(H1/n)K(H^{1/n}), they are linearly independent.

Example 1

There exists a largest abelian extension of exponent dividing nn of K, contained in Ω\Omega ; it is obtained by adjoining to K the n-th roots of all its elements ; its Galois group may be identified with Hom⁡(K∗/K∗n,μn(K))\operatorname{Hom}(K^*/K^{*n}, \mu_n(K)), hence also with Hom⁡(K∗,μn(K))\operatorname{Hom}(K^*, \mu_n(K)).

Example 2

\* Let us take K=QK = \mathbf{Q} and n=2n = 2. Then Q∗/Q∗2\mathbf{Q}^*/\mathbf{Q}^{*2} is a vector F2\mathbf{F}_2-space having as a basis the union of {−1}\{-1\} and the set of all prime numbers. The largest abelian extension of exponent 2 of Q contained in C\mathbf{C} is thus the subfield Q(i,2,3,5,…)\mathbf{Q}(i, \sqrt{2}, \sqrt{3}, \sqrt{5}, \ldots) of C\mathbf{C}. Its Galois group consists of all the automorphisms obtained by multiplying each of the elements i, 2, 3, 5i,\ \sqrt{2},\ \sqrt{3},\ \sqrt{5} etc. by ±1\pm 1. \

Example 3

Let L be a cyclic extension of K of degree nn; then the group (Ln∩K∗)/K∗n(L^n \cap K^*)/K^{*n} is cyclic of order nn. If a∈K∗a \in K^* is such that the residue class of aa mod K∗nK^{*n} is a generator of this group, then L is K-isomorphic to K[X]/(Xn−a)K[X]/(X^n - a) and the group Gal(L/K)\mathrm{Gal}(L/K) consists of the nn automorphisms mapping XX to ζX\zeta X, ζ∈μn(K)\zeta \in \mu_n(K).

Example 4

Conversely let a∈K∗a \in K^* and let rr be the least integer >0> 0 such that a′∈K∗na' \in K^{*n}; then the subfield LL of Ω\Omega generated by the roots of the polynomial Xn−aX^n - a is a cyclic extension of KK of degree rr. In particular, Xn−aX^n - a is irreducible if and only if r=nr = n.

Remark

Let a∈K∗a \in K^* and let r be the least integer >0> 0 such that a′∈Kna' \in K^n. Let B be the set of roots in K of the polynomial Xn/r−aX^{n/r} - a; then we have

Xn−a=∏b∈B(X′−b), X^n - a = \prod_{b \in B} (X' - b),

by substitution of X′X' for T in the relation Tn/r−a=Π(T−b)T^{n/r} - a = \Pi (T - b). By Example 4 each of the polynomials X′−bX' - b is irreducible, so that (14) is the decomposition of Xn−aX^n - a into irreducible polynomials in K[X]K[X].

9. Artin-Schreier theory

In this No. we shall use pp to denote a prime number and assume that K is of characteristic pp. We denote by Ω\Omega an algebraic closure of K and by φ\varphi the endomorphism of the additive group of Ω\Omega defined by

φ(x)=xp−x. \varphi(x) = x^p - x.

By V, p. 93 the kernel of φ\varphi is the prime subfield Fp\mathbf{F}_p of KK. For every subset AA of K we denote by K(φ−1(A))K(\varphi^{-1}(A)) the subextension of Ω\Omega generated by all x∈Ωx \in \Omega such that φ(x)∈A\varphi(x) \in A.

Lemma 7

K(φ−1(A))K(\varphi^{-1}(A)) is an abelian extension of KK of exponent dividing pp.

Since the polynomials φ−a=Xp−X−a, a∈A\varphi - a = X^p - X - a,\ a \in A are separable over KK, the extension L=K(φ−1(A))L = K(\varphi^{-1}(A)) is Galois. Let a∈Gal(L/K)a \in \mathrm{Gal}(L/K) and x∈φ−1(A)x \in \varphi^{-1}(A); we have φ(σ(x))=φ(x)\varphi(\sigma(x)) = \varphi(x), hence σ(x)−x∈Fp\sigma(x) - x \in \mathbf{F}_p, that is, σ(x)=x+i, i∈Fp\sigma(x) = x + i,\ i \in \mathbf{F}_p. This implies that σp(x)=x+pi=x\sigma^p(x) = x + pi = x, hence up=1u^p = 1; similarly if a′∈Gal(L/K)a' \in \mathrm{Gal}(L/K) and σ′(x)=x+j\sigma'(x) = x + j, then we have σ∘σ′(x)=x+i+j=a′∘σ(x)\sigma \circ \sigma'(x) = x + i + j = a' \circ \sigma(x), so σ∘σ′=σ′∘σ\sigma \circ \sigma' = \sigma' \circ \sigma.

Lemma 8

Let L be a Galois extension of K. There exists a unique mapping (σ,a)↦[a,a′)(\sigma, a) \mapsto [a, a') of Gal(L/K)×((φ(L)∩K)/φ(K))\mathrm{Gal}(L/K) \times ((\varphi(L) \cap K)/\varphi(K)) into Fp\mathbf{F}_p such that for all σ∈Gal(L/K)\sigma \in \mathrm{Gal}(L/K) and every element x∈Lx \in L such that φ(x)∈K\varphi(x) \in K we have, on denoting by φ(x)‾\overline{\varphi(x)} the residue class of φ(x)\varphi(x) mod φ(K)\varphi(K),

[σ,φ(x)‾]=σ(x)−x. [\sigma, \overline{\varphi(x)}] = \sigma(x) - x.

This mapping is Z-bilinear (for σ,τ∈Gal(L/K)\sigma, \tau \in \mathrm{Gal}(L/K), a,b∈(P(L)∩K)/P(K)a, b \in (\mathscr{P}(L) \cap K)/\mathscr{P}(K), we have [στ,a)=[σ,a)+[τ,a)[\sigma \tau, a) = [\sigma, a) + [\tau, a), [σ,a+b)=[σ,a)+[a,b)[\sigma, a + b) = [\sigma, a) + [a, b)).

For the right-hand side of (15) is an element of Fp\mathbf{F}_p which depends only on the residue class of P(x)mod  P(K)\mathscr{P}(x) \mod \mathscr{P}(K); this proves the first assertion; the second may be verified without difficulty.

For every Galois extension LL of KK let us write

aL:(P(L)∩K)/P(K)→Hom(Gal(L/K),Fp) a_L : (\mathscr{P}(L) \cap K)/\mathscr{P}(K) \to \mathrm{Hom}(\mathrm{Gal}(L/K), \mathbf{F}_p)
aL′:Gal(L/K)→Hom((P(L)∩K)/P(K),Fp) a_L' : \mathrm{Gal}(L/K) \to \mathrm{Hom}((\mathscr{P}(L) \cap K)/\mathscr{P}(K), \mathbf{F}_p)

for the homomorphisms obtained from the above Z-bilinear mapping (V, p. 87).

Proposition 10

For every Galois extension LL of finite degree of KK, the homomorphism aLa_L is bijective.

Let x∈Lx \in L be such that P(x)∈K\mathscr{P}(x) \in K and the residue class of P(x)mod  P(K)\mathscr{P}(x) \mod \mathscr{P}(K) lies in the kernel of aLa_L. For every σ∈Gal(L/K)\sigma \in \mathrm{Gal}(L/K) we have by definition σ(x)=x\sigma(x) = x; hence x∈Kx \in K and P(x)∈P(K)\mathscr{P}(x) \in \mathscr{P}(K). This proves the injectivity of aLa_L. Now let f:Gal(L/K)→Fpf : \mathrm{Gal}(L/K) \to \mathbf{F}_p be a homomorphism; for all σ,τ∈Gal(L/K)\sigma, \tau \in \mathrm{Gal}(L/K) we have

f(στ)=f(σ)+σ(f(τ)),f(σ)∈Fp. f(\sigma \tau) = f(\sigma) + \sigma(f(\tau)), \quad f(\sigma) \in \mathbf{F}_p.

By V, p. 65, Cor. 2 there exists x∈Lx \in L such that f(u)=σ(x)−xf(u) = \sigma(x) - x for all o∈Gal(L/K)o \in \mathrm{Gal}(L/K). Since f(o)∈Fpf(o) \in \mathbf{F}_p we have P(σ(x))=P(x)\mathscr{P}(\sigma(x)) = \mathscr{P}(x), hence σ(P(x))=P(x)\sigma(\mathscr{P}(x)) = \mathscr{P}(x) for all a∈Gal(L/K)a \in \mathrm{Gal}(L/K) and P(x)∈K\mathscr{P}(x) \in K. If aa is the residue class of P(x)mod  P(K)\mathscr{P}(x) \mod \mathscr{P}(K), we have f(σ)=[σ,a)f(\sigma) = [\sigma, a), so f=aL(a)f = a_L(a).

Corollary

If LL is a Galois extension of KK, the homomorphism aLa_L is injective and its image is the group Homc(Gal(L/K),Fp)\mathrm{Hom}_c(\mathrm{Gal}(L/K), \mathbf{F}_p) of continuous homomorphisms of the topological group Gal(L/K)\mathrm{Gal}(L/K) into the discrete group Fp\mathbf{F}_p.

This is proved in the same way as the Cor. of Prop. 9, V, p. 89.

Theorem 5

a) The mapping A↦K(P−1(A))A \mapsto K(\mathscr{P}^{-1}(A)) is a bijection of the set of subgroups of KK containing P(K)\mathscr{P}(K) onto the set of abelian subextensions of exponent dividing pp in Ω\Omega. The inverse mapping is L↦P(L)∩KL \mapsto \mathscr{P}(L) \cap K. b) For every subgroup AA of KK containing P(K)\mathscr{P}(K), the homomorphism

a′:Gal(K(P−1(A))/K)→Hom(A/P(K),Fp) a' : \mathrm{Gal}(K(\mathscr{P}^{-1}(A))/K) \to \mathrm{Hom}(A/\mathscr{P}(K), \mathbf{F}_p)

is bijective, and it is a homeomorphism when Hom(A/P(K),Fp)\mathrm{Hom}(A/\mathscr{P}(K), \mathbf{F}_p) is equipped with the topology of simple convergence. c) Let AA be a subgroup of KK containing P(K)\mathscr{P}(K) and let BB be a basis of the vector Fp\mathbf{F}_p-space A/P(K)A/\mathscr{P}(K). For each a∈Ba \in B let xax_a be an element of Ω\Omega such that P(xa)\mathscr{P}(x_a) is a representative of aa in AA. Then the monomials xα=∏a∈Bxaα(a)x^\alpha = \prod_{a \in B} x_a^{\alpha(a)} with a = (\alpha(a)) in N^{(B)} such that 0 \leq \alpha(a) < p for each a \in B form a basis of the vector K-space K(\mathfrak{p}^{-1}(A)). In particular we have [K(\mathfrak{p}^{-1}(A)):K] = (A:\mathfrak{p}(K)).

Th. 5 is proved in the same way as Th. 4 (V, p. 89) mutatis mutandis.

Example 1

There exists a greatest abelian extension of K of exponent dividing p, contained in \Omega; it is K(B^1(K)), its Galois group may be identified with Horn(K/\mathfrak{p}(K),\mathbf{F}_p).

Example 2

Let L be a cyclic extension of K of degree p; then the group (\mathfrak{p}(L) \cap K)/\mathfrak{p}(K) is cyclic of order p. If a \in K is such that the residue class of a mod \mathfrak{p}(K) is a generator of this group, then L is K-isomorphic to K[X]/(X^p - X - a) and the Galois group Gal(L/K) consists of the p automorphisms mapping X to X + i, i \in \mathbf{F}_p.

Example 3

Conversely, if a \in K - \mathfrak{p}(K), then the polynomial X^p - X - a is irreducible and the subfield L of \Omega generated by its roots is a cyclic extension of K of degree p. If a \in \mathfrak{p}(K), then X^p - X - a = \prod_{a \in \mathfrak{p}^{-1}(a)} (X - \alpha).

Exercises

See the exercises for § 11.

Transcribed from the printed edition, 2003, Springer. Not translated.