Bourbaki

§ 1. Prime fields. Characteristic

1. Prime fields

The field of fractions of the ring Z of rational integers is called the field of rational numbers and is denoted by Q (I, p. 117). For every prime number p the quotient ring Z/(p)\mathbf{Z}/(p) is a finite 1^1 field of p elements, denoted by Fp\mathbf{F}_p in the sequel. The field Q is infinite since it contains Z, and is therefore not isomorphic to any field Fp\mathbf{F}_p. If p and p′p' are distinct prime numbers, the fields Fp\mathbf{F}_p and Fp′\mathbf{F}_{p'} have distinct cardinals and so are not isomorphic.

Definition 1

A field is said to be prime if it is isomorphic either to Q or to one of the fields Fp\mathbf{F}_p.

Every subfield of Q contains the ring Z and hence the field of fractions Q of Z; every subring of Fp\mathbf{F}_p is necessarily equal to Fp\mathbf{F}_p. Therefore every subfield of a prime field is necessarily equal to it (cf. Cor. 2 of Th. 1 below). Let P be a prime field and A a ring; if f and f′f' are two homomorphisms of P into A, then the set of x∈Px \in P such that f(x)=f′(x)f(x) = f'(x) is a subfield of P, hence by what has been said we must have f=f′f = f'. In particular, the only endomorphism of a prime field is the identity mapping.

1^1 By abuse of language one says that a ring or field is finite if its underlying set is finite.

Theorem 1

— Let AA be a ring ; suppose that there exists a subfield of AA. Then AA has a unique subfield PP which is a prime field. Moreover, PP is contained in the centre of AA and in every subfield of AA.

Let KK be a subfield of AA, CC the centre of AA, and put K′=K∩CK' = K \cap C; then K′K' is a subfield of AA. Let ff be the unique homomorphism of Z\mathbf{Z} into AA and pp its kernel. Every subring of AA, in particular K′K', contains f(Z)f(\mathbf{Z}); hence the ideal pp is prime (I, p. 116-117). If p=(0)p = (0), the homomorphism ff of Z\mathbf{Z} into K′K' is injective; hence it extends (I, p. 116) to an isomorphism fˉ\bar{f} of Q\mathbf{Q} onto a subfield PP of K′K'. If p≠(0)p \neq (0), there is a strictly positive integer pp such that p=(p)p = (p) (I, p. 111); if we had p=abp = ab with a>1a > 1 and b>1b > 1, this would mean a∉p,b∉pa \notin p, b \notin p and ab∈pab \in p in contradiction with the fact that pp is prime. The number pp is therefore prime and by passage to quotients ff defines an isomorphism of Fp=Z/pF_p = \mathbf{Z}/p onto a subfield PP of K′K'. In both cases PP is a subfield of AA contained in the centre CC of AA, and it is a prime field. Let LL be a subfield of AA; then P∩LP \cap L is a subfield of PP, and since PP is prime, we have P∩L=PP \cap L = P, whence P⊂LP \subset L. If P′P' is a subfield of AA and is a prime field, then by what has been said, P⊂P′P \subset P', whence P=P′P = P' because P′P' is a prime field.

Corollary 1

— Let KK be a field. There exists a unique subfield of KK which is a prime field, and this is the least subfield of KK.

Corollary 2

— For a field to be prime it is necessary and sufficient that it should contain no subfield other than itself.

2. Characteristic of a ring and of a field

We shall define the characteristic of a ring AA only when AA has a subfield. When this is so, let ff be the unique ring homomorphism of Z\mathbf{Z} into AA, and let nn be the unique positive integer generating the ideal of Z\mathbf{Z} which is the kernel of ff (I, p. 111); then the integer nn is called the characteristic of AA.

Let AA be a ring for which the characteristic is defined; then AA does not reduce to 0. By Th. 1 there exists a unique subfield PP of AA which is a prime field; we shall call it the prime subfield of AA. By the proof of Th. 1 there are the following two possibilities: a) the characteristic of AA is 0, PP is isomorphic to QQ, b) the characteristic of AA is a prime number pp, PP is isomorphic to Fp\mathbf{F}_p.

If the characteristic of AA is zero, there exists a unique ring homomorphism of QQ into AA; its image is the prime subfield of AA, contained in the centre of AA. Therefore there exists a unique QQ-algebra structure of AA compatible with the ring structure. When the characteristic of AA is a prime number pp, we have the corresponding properties on replacing the field QQ by the field Fp\mathbf{F}_p.

Proposition 1

— Let AA be a ring not reduced to 0. a) For AA to be of characteristic 0 it is necessary and sufficient that the mapping x↦n⋅xx \mapsto n \cdot x of AA into itself should be bijective, for every integer n≠n \neq

b) Let pp be a prime number. For AA to be of characteristic pp it is necessary and sufficient that p⋅x=0p \cdot x = 0 for all x∈Ax \in A.

Let ff be the unique homomorphism of Z\mathbf{Z} into AA; we have n⋅x=f(n)xn \cdot x = f(n)x for every integer nn and every xx in AA. For AA to be of characteristic 0, it is necessary and sufficient that ff should extend to a homomorphism of Q\mathbf{Q} into AA, that is, f(n)f(n) should be invertible in AA for every n≠0n \neq 0 (I, p. 113); this proves aa. Similarly for AA to be of characteristic pp it is necessary and sufficient that ff should annihilate pZp\mathbf{Z}, that is, f(p)=0f(p) = 0, or also that p⋅x=0p \cdot x = 0 for all x∈Ax \in A; this proves bb.

Let us take for AA a not necessarily commutative field. The centre of AA is a (commutative) field; therefore the characteristic and the prime subfield of AA are defined.

Remark 1

Let AA and A′A' be two rings not reduced to 0. Suppose that the characteristic of AA is defined and that there is a homomorphism uu of AA into A′A'. The image under uu of the prime subfield of AA is a subfield P′P' of A′A', isomorphic to PP, and hence prime. It follows that the characteristic of A′A' is defined and is equal to that of AA. If AA and A′A' are of characteristic 0 (resp. p≠0p \neq 0), the mapping uu is a homomorphism of algebras over Q\mathbf{Q} (resp. Fp\mathbf{F}_p).

Remark 2

Remark 1 shows that if AA is a ring of characteristic 0 (resp. p≠0p \neq 0), the same holds of any ring A′A' containing AA as subring, or of any quotient of AA by a two-sided ideal a≠Aa \neq A. In particular, if KK is a field, every subfield of KK and every extension field of KK have the same characteristic as KK.

Remark 3

Let AA be an algebra not reduced to 0 over a field KK. Since the mapping λ↦λ⋅1\lambda \mapsto \lambda \cdot 1 of KK into AA is a ring homomorphism, Remark 1 shows that the characteristic of AA is defined and equal to that of KK.

Remark 4

Since the field Q\mathbf{Q} is infinite, every ring of characteristic 0 is infinite; it follows that every finite field has non-zero characteristic.

Remark 5

Let AA be a ring not reduced to 0, whose additive group is a torsion-free Z\mathbf{Z}-module, and put B=Q⊗ZAB = \mathbf{Q} \otimes_{\mathbf{Z}} A. The mapping x↦1⊗xx \mapsto 1 \otimes x of AA into BB is injective (II, p. 314), hence AA is isomorphic to a subring of a ring of characteristic 0.

3. Commutative rings of characteristic p

In this No. and the following one pp denotes a prime number.

Theorem 2

— Let AA be a commutative ring of characteristic pp. The mapping a↦apa \mapsto a^p is an endomorphism of the ring AA, that is, we have the relations

(1)

(a+b)p=ap+bp(a + b)^p = a^p + b^p
(2)
(ab)p=apbp(ab)^p = a^p b^p

for a,ba, b in AA.

Formula (2) follows from the commutativity of AA. To prove (1) we use the binomial formula (a+b)p=ap+bp+∑i=1p−1(pi)⋅aibp−i(a + b)^p = a^p + b^p + \sum_{i=1}^{p-1} \binom{p}{i} \cdot a^i b^{p-i}; since p⋅x=0p \cdot x = 0 for all x∈Ax \in A, it suffices to prove the following lemma:

Lemma 1

Let pp be a prime number and ii an integer in the range from 1 to p−1p - 1, then the binomial coefficient (pi)\binom{p}{i} is an integer divisible by pp.

We argue by induction on ii, the case i=1i = 1 being immediate from the formula (p1)=p\binom{p}{1} = p. Suppose that 2≤i≤p−12 \leq i \leq p - 1 and that (ii−1)\binom{i}{i-1} is divisible by pp. Then the integer i(pi)=(p−i+1)(pi−1)i \binom{p}{i} = (p - i + 1) \binom{p}{i-1} belongs to the prime ideal pZp\mathbf{Z} of Z\mathbf{Z}; since i∉pZi \notin p\mathbf{Z}, we have (pi)∈pZ\binom{p}{i} \in p\mathbf{Z} and the lemma follows.

Let AA be a commutative ring of characteristic pp and ff an integer ≥0\geq 0. From Th. 2 we deduce by induction on ff that the mapping a↦apfa \mapsto a^{p^f} is an endomorphism of the ring AA. In particular we have the relation

(a1+⋯+an)pf=a1pf+⋯+anpf (a_1 + \cdots + a_n)^{p^f} = a_1^{p^f} + \cdots + a_n^{p^f}

for any a1,…,ana_1, \ldots, a_n in AA. The mapping a↦apa \mapsto a^p is sometimes called the Frobenius endomorphism of AA. Taking A=FpA = \mathbf{F}_p and ai=1a_i = 1, we obtain from (3) the relation:

npf≡nmod  p(n∈Z,f∈N). n^{p^f} \equiv n \mod p \quad (n \in \mathbf{Z}, f \in \mathbf{N}) .

For each subset SS of AA we denote by SpfS^{p^f} the set of elements of AA of the form xpfx^{p^f} with x∈Sx \in S^1. In particular, if KK is a subring of AA, the set KpfK^{p^f} is a subring of AA. If KK is a subring of AA and SS a subset of AA we denote by K[S]K[S] the subring of AA generated by K∪SK \cup S; when AA is a field, we denote by K(S)K(S) the field of fractions of K[S]K[S], that is, the subfield of AA generated by K∪SK \cup S.

Proposition 2

— *Let AA be a commutative ring of characteristic pp, KK a subring of AA, SS a subset of AA and ff a positive integer.

a) We have K[S]pf=Kpf[Spf]K[S]^{p^f} = K^{p^f}[S^{p^f}], and if AA is a field, K(S)pf=Kpf(Spf)K(S)^{p^f} = K^{p^f}(S^{p^f}).

b) If the KK-module K[S]K[S] is generated by the family (ai)i∈I(a_i)_{i \in I} of elements of AA, then the KK-module K[Spf]K[S^{p^f}] is generated by the family (aipf)i∈I′(a_i^{p^f})_{i \in I'}

Since K[S]K[S] is the subring of AA generated by K∪SK \cup S, its image KISlpfKISlp^f under the endomorphism π:a↦apf\pi : a \mapsto a^{p^f} of the ring AA is the subring of AA generated by the image Kpf∪SpfK^{p^f} \cup S^{p^f} of K∪SK \cup S under π\pi, whence K[S]pf=Kpf[Spf]K[S]^{p^f} = K^{p^f}[S^{p^f}]. The case of fields is treated similarly; this proves a).

It is clear that the family (aipf)i∈I(a_i^{p^f})_{i \in I} generates the KpfK^{p^f}-module K[S]pfK[S]^{p^f}. The KK-module K[Spf]K[S^{p^f}] is generated by products of the form x1pf⋯xnpf=(x,…xn)pfx_1^{p^f} \cdots x_n^{p^f} = (x, \ldots x_n)^{p^f} with x1,…,xnx_1, \ldots, x_n arbitrary in SS, hence also by the set K[S]pfK[S]^{p^f}. Assertion b) follows directly from this.

^1 Of course the set SpfS^{p^f} should not be confused with the set product of pfp^f sets equal to SS, nor with the set of products of pfp^f elements belonging to SS.

4. Perfect rings of characteristic p

Definition 2

— A ring AA of characteristic p≠0p \neq 0 is said to be perfect if it is commutative and the mapping a↦apa \mapsto a^p is bijective.

If the ring AA is perfect of characteristic pp, the mapping a↦apfa \mapsto a^{p^f} is an automorphism of the ring AA for every integer f≥0f \geq 0; the inverse automorphism is denoted by a↦a1/pfa \mapsto a^{1/p^f} or a↦ap−fa \mapsto a^{p^{-f}} and the image of a subset SS of AA under this automorphism is written S1/pfS^{1/p^f} or Sp−fS^{p^{-f}}. It is clear that (ape)pf=ape+f(a^{p^e})^{p^f} = a^{p^{e+f}} for all a∈Aa \in A and any integers ee and ff (of whatever sign).

Let AA be a commutative ring of characteristic pp. For every integer f≥0f \geq 0 let us write nf\mathfrak{n}_f for the kernel of the endomorphism a↦apa \mapsto a^p of the ring AA. Then (nf)f≥0(\mathfrak{n}_f)_{f \geq 0} is an increasing sequence of ideals of AA; since every positive integer is majorized by a power of pp, the ideal n=⋃f≥0nf\mathfrak{n} = \bigcup_{f \geq 0} \mathfrak{n}_f consists of all the nilpotent elements of AA. In particular, if AA is perfect, every nilpotent element of AA is zero.

Definition 3

— Let AA be a commutative ring of characteristic p≠0p \neq 0. By a perfect closure of AA we understand a pair (A,u)(A, u) where AA is a perfect ring of characteristic pp and uu is a homomorphism of AA into AA satisfying the following universal property:

(PC) If B is a perfect ring of characteristic p and v is a homomorphism of A into B, then there exists a unique homomorphism h of A into B such that v=h∘uv = h \circ u.

The universal property (PC) implies at once the uniqueness of the perfect closure, in the following sense : if (A,u)(A, u) and (A′,u′)(A', u') are two perfect closures of AA, then there exists a unique isomorphism hh of AA onto A′A' such that u′=h∘uu' = h \circ u (cf. E, IV, p. 23). We shall now establish the existence of the perfect closure :

Theorem 3

— Let AA be a commutative ring of characteristic p≠0p \neq 0. There exists a perfect closure (A,u)(A, u) of AA. Moreover, the kernel of uu is the set of all nilpotent elements of AA and for each x∈Ax \in A there exists an integer n≥0n \geq 0 such that xpn∈u(A)x^{p^n} \in u(A).

For each integer n≥0n \geq 0, put An=AA_n = A; when m≥nm \geq n we define a homomorphism πm,n\pi_{m,n} of AnA_n into AmA_m by πm,n(a)=apm−n\pi_{m,n}(a) = a^{p^{m-n}}. We thus obtain a direct system of rings (An,πm,n)(A_n, \pi_{m,n}) (I, p. 120); let AA be the direct limit of this system and unu_n the canonical homomorphism of An=AA_n = A into AA; we also put u=u0u = u_0. By construction of the direct limit the kernel n\mathfrak{n} of uu is the union of the kernels of the homomorphisms π_{n,0}: a \mapsto a^{p^n} of AA into AA, thus it consists of all the nilpotent elements of AA. The ring AA is commutative of characteristic pp by Remark 1 of V, p. 3.

The ring AA is the union of the ascending sequence (u,(A))n(u, (A))_n of subrings. We have un(A)pn=Au_n(A)^{p^n} = A; hence for each x∈A^x \in \hat{A} there exists an integer n≥0n \geq 0 such that xpn∈u(A)x^{p^n} \in u(A). We also have un(A)=un+1(A)pu_n(A) = u_{n+1}(A)^p, whence A^p=A\hat{A}^p = A. Let x∈Ax \in A be such that xp=0x^p = 0; choose an integer n≥1n \geq 1 and an element a∈Aa \in A such that x=un(a)x = u_n(a). Then we have un−1(a)=un(a)p=0u_{n-1}(a) = u_n(a)^p = 0; by definition of the direct limit there exists an integer mm such that πm1,n−1(a)=0\pi_m^{1,n-1}(a) = 0, that is, apm−n=0a^{p^{m-n}} = 0. We thus have πm,n(a)=0\pi_{m,n}(a) = 0, whence un(a)=0u_n(a) = 0, that is, x=0x = 0. Therefore the ring AA is perfect of characteristic pp.

Let vv be a homomorphism of AA into a perfect ring BB of characteristic pp. For every integer n≥0n \geq 0, the mapping b↦bpnb \mapsto b^{p^n} is an automorphism of BB and so there exists a homomorphism vnv_n of A,=AA, = A into BB characterized by v(a)=vn(a)pnv(a) = v_n(a)^{p^n}. We then have vm∘πm,n=vnv_m \circ \pi_{m,n} = v_n for m≥n≥0m \geq n \geq 0; by definition of the direct limit there exists a homomorphism hh of AA into BB such that vn=h∘uv_n = h \circ u, for all n≥0n \geq 0; in particular we have v=v0=h∘u0=h∘uv = v_0 = h \circ u_0 = h \circ u. Finally let h′h' be a homomorphism of AA into BB such that h′∘u=vh' \circ u = v. Let x∈Ax \in A; as we have seen, there exist an integer n≥0n \geq 0 and an element a∈Aa \in A such that xpn=u(a)x^{p^n} = u(a). Then we have

h(x)pn=h(u(a))=v(a)=h′(u(a))=h′(x)pn, h(x)^{p^n} = h(u(a)) = v(a) = h'(u(a)) = h'(x)^{p^n},

and since BB is perfect, we find h(x)=h′(x)h(x) = h'(x). Thus we have h′=hh' = h, and this completes the proof that (A,u)(A, u) is a perfect closure of AA.

Proposition 3

— *Let BB be a perfect ring of characteristic pp and AA a subring of BB. Write Ap−∞=⋃f≥0Ap−f\mathbf{A}^{p^{-\infty}} = \bigcup_{f \geq 0} \mathbf{A}^{p^{-f}} and denote by jj the canonical injection of AA in Ap−∞\mathbf{A}^{p^{-\infty}}. Then Ap−∞\mathbf{A}^{p^{-\infty}} is the smallest perfect subring of BB containing AA and (Ap−∞,j)(\mathbf{A}^{p^{-\infty}}, j) is a perfect closure of AA.

For each integer f∈Zf \in \mathbf{Z} denote by πf\pi_f the automorphism b↦bpfb \mapsto b^{p^f} of BB. The sequence of subrings π−f(A)\pi_{-f}(A) of BB (for f≥0f \geq 0) is ascending and its union Ap−∞\mathbf{A}^{p^{-\infty}} is thus a subring of BB. We have π1(Ap−∞)=⋃f≥0π−(f−1)(A)=Ap−m\pi_1(\mathbf{A}^{p^{-\infty}}) = \bigcup_{f \geq 0} \pi_{-(f-1)}(A) = \mathbf{A}^{p^{-m}}, hence Ap−∞\mathbf{A}^{p^{-\infty}} is a perfect subring of BB. Finally let B0B_0 be a perfect subring of BB containing AA; for every integer f≥0f \geq 0 we have π−f(A)⊂π−f(B0)=B0\pi_{-f}(A) \subset \pi_{-f}(B_0) = B_0, whence Ap−∞⊂B0\mathbf{A}^{p^{-\infty}} \subset B_0.

If vv is a homomorphism of AA into a perfect ring B′B' of characteristic pp, then for every integer f≥0f \geq 0 we can define a homomorphism hfh_f of π−f(A)\pi_{-f}(A) into B′B' by hf(π−f(a))=v(a)p−fh_f(\pi_{-f}(a)) = v(a)^{p^{-f}} for all a∈Aa \in A. We see at once that hf+1h_{f+1} agrees with hfh_f on π−f(A)\pi_{-f}(A); thus there exists a homomorphism hh of Ap−∞\mathbf{A}^{p^{-\infty}} into B′B' which induces hfh_f on π−f(A)\pi_{-f}(A) for all f≥0f \geq 0 and in particular, hh extends h0=vh_0 = v. If h′h' is another extension of vv to a homomorphism of Ap−∞A^{p^{-\infty}} into B, the equality h′=hh' = h may be established as in the proof of Th. 3.

5. Characteristic exponent of a field. Perfect fields

Let KK be a field. By the characteristic exponent of KK we understand the integer equal to 1 if KK is of characteristic 0, and equal to the characteristic of KK when this is non-zero.

Proposition 4

— Let KK be a field of characteristic exponent q. For every integer f≥0f \geq 0 the mapping x↦xqfx \mapsto x^{q^f} is an isomorphism of K on one of its subfields (denoted by KqfK^{q^f}). This follows from Th. 2 when q ≠ 1 and is trivial when q = 1. Likewise one can extend Prop. 2 to the case where AA is a field of characteristic exponent q, the case q = 1 being trivial.

Definition 4

— A field K of characteristic exponent q is said to be perfect if we have Kq=KK^q = K. When Kq≠KK^q \neq K, K is called imperfect. By this definition a field is perfect if it is of characteristic 0, or if it is a perfect ring of characteristic p≠0p \neq 0 in the sense of Def. 2. If KK is a field of characteristic p≠0p \neq 0 and (K,u)(K, u) is a perfect closure of KK, then K is a field by Prop. 3 (V, p. 6) and uu is an isomorphism of K onto a subfield of K^\hat{K}. Frequently one identifies KK with its image under uu in K^\hat{K}, so that we have K^=Kp−∞\hat{K} = K^{p^{-\infty}} (Prop. 3). Let K be a field of characteristic 0 ; the characteristic exponent of KK is then 1. By convention the notation xq−fx^{q^{-f}} and Sq−fS^{q^{-f}} is taken to mean x and S respectively (for an element xx of KK and a subset S of KK). In particular we put Kq−∞=KK^{q^{-\infty}} = K and we agree to take the perfect closure of KK to be KK.

Proposition 5

— If K is a field of characteristic 0, or is finite, \ or algebraically closed \*, it is perfect. In particular every prime field is perfect.* Suppose that KK has characteristic p≠0p \neq 0. If KK is finite, then the subfield KpK^p of KK has the same cardinal as KK, whence Kp=KK^p = K. \* If KK is algebraically closed, the polynomial Xp−aX^p - a has a root xx in KK for each a∈Ka \in K (V, p. 20, Def. 1) whence xp=ax^p = a and so Kp=KK^p = K. \* Finally a prime field is of characteristic 0 or finite. Let K0K_0 be a field of characteristic p≠0p \neq 0 and K=K0(X)K = K_0(X) the field of rational fractions in an indeterminate XX over K0K_0. Then KK is imperfect, for there exists no element u(X)/v(X)u(X)/v(X) of KK (u, v polynomials in K0[X]K_0[X]) such that (u(X)/v(X))p=X(u(X)/v(X))^p = X. This may be seen by writing this relation in the form u(X)p=Xv(X)pu(X)^p = Xv(X)^p and comparing the degrees of the two sides.

6. Characterization of polynomials with zero differential

Proposition 6

— *Let K be a commutative ring, A the polynomial algebra K[Xi]i∈IK[X_i]_{i \in I}, and S the set of elements F of A such that dF=0dF = 0.

a) If K is a ring of characteristic 0, then S = K. b) If K is a ring of characteristic p≠0p \neq 0, then S=K[Xip]i∈IS = K[X_i^p]_{i \in I}; if moreover K is perfect, then K=APK = A^P.

The mapping F↦dFF \mapsto dF of A into the module ΩK(A)\Omega_K(A) of K-differentials of A is K-linear and satisfies the relation

d(FF′)=F⋅dF′+F′⋅dF d(FF') = F \cdot dF' + F' \cdot dF

(III, p. 569). Therefore S is a subalgebra.

When K is of characteristic p≠0p \neq 0, put T=K[Xip]i∈IT = K[X_i^p]_{i \in I}; we thus have T=APT = A^P if K is perfect (VV, p. 4, Prop. 2); moreover, we have d(Xip)=pXip−1⋅dXi=0d(X_i^p) = pX_i^{p-1} \cdot dX_i = 0 for all i∈Ii \in I, hence the subalgebra S of A contains T. If K is of characteristic 0, we put T=KT = K, and still find that T⊂ST \subset S. It remains to show that S is contained in T.

For every finite subset J of I let AJA_J be the subalgebra of A generated by the family (Xj)j∈J(X_j)_{j \in J}. We have A∅=KA_\emptyset = K and A=⋃J∈IAJA = \bigcup_{J \in I} A_J; so it suffices to prove the relation S∩AJ⊂TS \cap A_J \subset T, which we shall accomplish by induction on the cardinal of J. Thus let J be a finite subset of I such that S∩AJ⊂TS \cap A_J \subset T, let i be an element of I−JI - J and J′=J∪{i}J' = J \cup \{ i \}. Every element F of A may be written in just one way in the form

F=∑n=0∞Fn⋅Xin, F = \sum_{n=0}^\infty F_n \cdot X_i^n,

with Fn∈AJF_n \in A_J for all n≥0n \geq 0, and then

dF=∑n=0∞Xin⋅dFn+∑n=0∞nXin−1Fn⋅dXi. dF = \sum_{n=0}^\infty X_i^n \cdot dF_n + \sum_{n=0}^\infty nX_i^{n-1}F_n \cdot dX_i.

Suppose that F belongs to S; the family (dXr)r∈J(dX_r)_{r \in J} is a basis of the A-module ΩK(A)\Omega_K(A) (III, p. 570) and dFndF_n is a linear combination of the differentials dXjdX_j for j∈Jj \in J because Fn∈AJ=K[Xj]j∈JF_n \in A_J = K[X_j]_{j \in J}. By (6) we then have dFn=0dF_n = 0 and nFn=0nF_n = 0 for each integer n≥0n \geq 0. By the induction hypothesis Fn∈TF_n \in T for all nn, since dFn=0dF_n = 0.

a) If K is of characteristic 0, we have nFn=0nF_n = 0 for all n≥1n \geq 1, whence Fn=0F_n = 0 by Prop. 1 (VV, p. 2); so we have F=F0F = F_0, whence F∈TF \in T.

b) If K is of characteristic p≠0p \neq 0, then A is an algebra over the field Fp\mathbf{F}_p and the relation nFn=0nF_n = 0 implies Fn=0F_n = 0 for every integer nn not divisible by pp. So we have F=∑m=0mFmpXimpF = \sum_{m=0}^m F_{mp} X_i^{mp}, whence F∈TF \in T

Remark

— We still have S=KS = K when the additive group of K is torsion-free; this follows from the above proof or from Remark 5 of V, p. 3.

Corollary

— Let K be a field and F(X) a polynomial with coefficients in K, whose derivative F'(X) is zero. a) If K is of characteristic 0, then F ∈ K. b) If K is of characteristic p ≠ 0, there exists a polynomial G(X) such that F(X) = G(X^p). For we have dF = F . dX = 0.

Exercises

See the exercises for § 1.

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