Bourbaki

Theorem 2

Tag 0CQ6, permanent. Cite it as [Bourbaki, Tag 0CQ6].

Commutative Algebra, chapter VIII, § 5. Anneaux locaux réguliers

Let AA be a complete regular local noetherian ring whose residue field is of characteristic ≠0\neq 0, and let VV be a Cohen subring of AA. Put n=dim⁡(A)n = \dim(A).

a) Suppose that VV is a field. Then the VV-algebra AA is isomorphic to the algebra of formal series V[[X1,...,Xn]]V[[X_1, ..., X_n]].

b) Suppose that VV is a complete discrete valuation ring and that one has mV⊄mA2m_V \not\subset m_A^2. Then the VV-algebra AA is isomorphic to the algebra of formal series V[[X1,...,Xn−1]]V[[X_1, ..., X_{n-1}]].

c) Suppose that VV is a complete discrete valuation ring and that one has mV⊂mA2m_V \subset m_A^2. Then there exists a special polynomial PP in V[[X1,...,Xn−1]][T]V[[X_1, ..., X_{n-1}]] [T] and a VV-isomorphism of AA onto V[[X1,...,Xn−1]][T]/(P)V[[X_1, ..., X_{n-1}]] [T]/(P).

Assertion a) follows at once from Cor. 3 of Th. 1 of No. 2.

Let us prove b). Let (x1,...,xm)(x_1, ..., x_m) be a sequence of elements of mAm_A, and let φ0\varphi_0 be the homomorphism of V[X1,...,Xm]V[X_1, ..., X_m] into AA which coincides with the identity on VV and sends XiX_i to xix_i for 1≤i≤m1 \leq i \leq m. If aa is the ideal of V[X1,...,Xm]V[X_1, ..., X_m] generated by X1,...,XmX_1, ..., X_m, one has φ0(a)⊂mA\varphi_0(a) \subset m_A, hence φ0\varphi_0 extends by continuity to a homomorphism φ\varphi of Vm=V[[X1,...,Xm]]V_m = V[[X_1, ..., X_m]] into AA. Let π\pi be a uniformizer of VV. By Corollary 2 of Theorem 1 of No. 2, φ\varphi is an isomorphism of VmV_m onto AA if and only if (π,x1,...,xm)(\pi, x_1, ..., x_m) is a system of coordinates of AA. But the canonical mapping of mV/mV2m_V/m_V^2 into mA/mA2m_A/m_A^2 is injective, since one has mV⊄mA2m_V \not\subset m_A^2 and mV/mV2m_V/m_V^2 is of rank 1 over κV\kappa_V. Hence mA/(mV+mA2)m_A/(m_V + m_A^2) is of rank n−1n - 1 and (π,x1,...,xm)(\pi, x_1, ..., x_m) is a system of coordinates of AA if and only if the classes of the xix_i form a basis of mA/(mV+mA2)m_A/(m_V + m_A^2) over κV\kappa_V. Hence b).

Let us prove c). Let (y1,...,yn)(y_1, ..., y_n) be a system of coordinates of AA. As above, put Vn=V[[Y1,...,Yn]]V_n = V[[Y_1, ..., Y_n]] and consider the homomorphism φ\varphi of VnV_n into AA which coincides with the identity on VV and sends YiY_i to yiy_i for 1≤i≤n1 \leq i \leq n. Then gr(φ)\mathrm{gr}(\varphi) is surjective, hence φ\varphi is surjective (III, § 2, no 8, Corollary 2 of Theorem 1). The kernel pp of φ\varphi is a prime ideal of VnV_n; since one has

dim⁡(Vn)=n+1=dim⁡(Vn/p)+1, \dim(V_n) = n + 1 = \dim(V_n/p) + 1,

the prime ideal pp is of height 1 (§ 1, no 3, Proposition 8). But the ring VnV_n is factorial by Proposition 8 of VII, § 3, no 9; consequently, the ideal pp is principal (VII, § 3, no 2, Theorem 1).

Let RR be a generator of the ideal pp of VnV_n. By Lemma 3 of VII, § 3, No. 7, there exist integers u(1),...,u(n−1)u(1), ..., u(n-1) at least equal to 1, and an isomorphism σ\sigma of V[[X1,...,Xn−1,T]]V[[X_1, ..., X_{n-1}, T]] onto VnV_n such that

σ(Xi)=Yi+Ynu(i)for 1≤i<nσ(T)=Yn \begin{align*} \sigma(X_i) &= Y_i + Y_n^{u(i)} \quad \text{for } 1 \leq i < n \\ \sigma(T) &= Y_n \end{align*}

and such that σ−1(R)=Q\sigma^{-1}(R) = Q satisfies Q(0,...,0,T)≠0Q(0, ..., 0, T) \neq 0. Moreover, by the preparation theorem (VII, § 3, No. 8, Prop. 6), there exists a polynomial P in V[[X_1, ..., X_{n-1}]] [T] of the form

P=Td+∑i=1dai(X1,...,Xn−1)Td−i, P = T^d + \sum_{i=1}^d a_i(X_1, ..., X_{n-1}) T^{d-i},

generating the same ideal as Q in V[[X_1, ..., X_{n-1}, T]], and such that ai(0,...,0)∈mVa_i(0, ..., 0) \in m_V for 1≤i≤d1 \leq i \leq d. It is deduced that A is V-isomorphic to V[[X_1, ..., X_{n-1}, T]]/(P). But V[[X_1, ..., X_{n-1}, T]] is the direct sum of the ideal (P) and of the V[[X_1, ..., X_{n-1}]]-module with basis 1, T, ..., Td−1T^{d-1} (VII, § 3, No. 8, Prop. 5); hence A is V-isomorphic to V[[X_1, ..., X_{n-1}]] [T]/(P).

Put C = V[[X_1, ..., X_{n-1}]] and denote by α the class of ad(X1,...,Xn−1)a_d(X_1, ..., X_{n-1}) modulo mC2m_C^2. We have κV=κC=κA\kappa_V = \kappa_C = \kappa_A. By Prop. 4 of No. 4, the kernel of the canonical homomorphism from mC/mC2m_C/m_C^2 into mA/mA2m_A/m_A^2 is equal to κCα\kappa_C \alpha. Since the image of mV/mV2m_V/m_V^2 in mA/mA2m_A/m_A^2 is zero and mV/mV2m_V/m_V^2 is of rank 1 over κC\kappa_C, it follows that ada_d belongs to mV+mC2m_V + m_C^2, but not to mC2m_C^2. Consequently, the polynomial P is special, as was to be proved.

\* Remark. — Let k be a field, A a complete regular local noetherian k-algebra. When κA\kappa_A is not a separable extension of k, it is not true in general that A is isomorphic as a k-algebra to κA[[T1,...,Tn]]\kappa_A[[T_1, ..., T_n]] where n=dim⁡(A)n = \dim(A) (p. 98, Exercise 29). \*

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Nothing.

Translated with machine help, not yet reviewed by a person.