Discrete valuation rings and modules of differentials (Q1270403)

Let \(X\) be a discrete valuation ring, and let \(G\) be a local subring of \(X\) such that the inclusion map \( G \hookrightarrow X \) is a local homomorphism. The author gives a structure theorem for the \(X\)-module \( H_0(G, X, X)\), the module of Kähler differentials of \(X\) over \(G\), under the following hypotheses: Let \(S\) be the residue field of \( G\), and let \(C\) be the residue field of \( X\); then \( H_0(S, C, C)\) and \(H_1(S, C, C) \) are finite-dimensional \(C\)-vector spaces. The first condition implies that \( H_0(G, X, X) \) is a quasifinite \(X\)-module. In the first introductory part the author gives a structure theorem for quasifinite \(X\)-modules. Let \(K\) be the field of quotients of \(X\), and let \( \xi \) be a uniformizing parameter for \(X\). Let \(W\) be a quasifinite \(X\)-module; then we have a decomposition \[ W = K^{a'} \oplus (K/X)^b \oplus \sum_{k \geq 1} (X / \xi^k X)^{d_k} \oplus V \] where \(V\) is a separated torsion-free \(X\)-module, and there is an exact sequence of \(X\)-modules \[ 0 \to X^c \to V \to K^{a''} \to 0. \] Here \( a' \), \( a'' \) and \(b\) are cardinal numbers, \(c\) and \( d_k\) are integers and only finitely many of the integers \( d_k\) are different from zero; they are determined uniquely by \(W\). Let \( a = a' + a'' \), \( d = \sum d_k\). Then one has \( a + c = \dim_K(W \otimes_X K) \), \( c + d = \dim_C(W \otimes_X C) \), \( b - c = \dim_C(\text{Ann}_W(\xi)) - \dim_C(W \otimes_X C) \). Let \( L:= \widehat{X} \otimes_X K \) where \( \widehat{X} \) is the completion of \(X\), and let \( T(W) = \Hom_{\widehat{X}}(\Hom_X(W, L/\widehat{X}), L)\). Then one has \( \dim_L(T(W)) = a + b\). Now let \( W:= H_0(G, X, X) \), the module of Kähler differentials of \(X\) over \(G\), and let \( a, b, c, d \) be the numbers defined by \(W\). In the second part the author determines these numbers. He shows that \[ a + c = \dim_K(H_0(E, K, K))\text{ and } c + d = \dim_C(H_0(S, C, C)) + \lambda \] with \( \lambda \in \{ 0,1 \} \) , and that \(b\) is finite if \( \dim_C(H_1(S, C, C)) < \infty \). Therefore one knows \(a\) and \(d\) if one knows \(c\). Put \( \mu:= \dim_C(H_0(S, C, C)) \) and let \( \tau \) be the degree of transcendency of \(C\) over \(S\). One has \( \nu = 0 \) and \( \mu = \tau \) if \(C\) has characteristic \(0\); in the general case one has \( \mu \leq \nu + \tau \). The extension \(C\) of \(S\) is called a Cartier extension if \( \nu\) and \( \tau \) are finite and we have \( \mu = \nu + \tau \). For the rest of the paper the author assumes that \(C\) is a Cartier extension of \(S\), hence, in particular, that \(b\) is finite. The five main results of this paper now are: (1) If \(G\) and \(X\) are complete, then one has \( b - c = \dim(G) - 1 - \tau - \dim_K(H_1(E, K, K))\). (2) If \(X\) has characteristic zero, then one has \( b - c = \dim(G) - 1 - \tau + \Delta_P\); here the prime ideal \(P\) of \(\widehat G\) is the kernel of the homomorphism \( \widehat{G} \to \widehat{X} \) and \( \Delta_P:= \dim(\widehat{G}_P) + \dim(\widehat{G} / P) - \dim(\widehat G)\). (3) If \(X\) has characteristic zero, then \(c\) is equal to the degree of transcendency of \(C\) over \( \overline{C} \); here \( \overline{C}\) is the residue ring of the discrete valuation ring \( E \cap X \). (4) If \(X\) has positive characteristic \(p\) then one has \( c = \dim_C(H_0(\widetilde{C}, C, C)) + \mu \) where \( \mu \in \{ 0,1 \} \); here \( \widetilde{C} \) is the residue class ring of the discrete valuation ring \( E [K^p] \cap X \). (5) If \(X\) has positive characteristic, then one has \( b - c = \dim(G) - 1 - \tau - \dim_L(\overline{H}_1(E, K, L)) + \Delta_P \); here \( \overline{H}_1(E, K, L) \) is the image of the canonical homomorphism \( H_1(E, K, L) \to H_1(\widehat{G}_P, L, L)\) [note that \( H_1(\widehat{G}_P, L, L) \) is a finitely generated \( \widehat{X}\)-module]. Result 4 is Satz 8 and result 3 is Satz 12 in a paper by \textit{R. Berger} and \textit{E. Kunz} [Math. Z. 77, 314-338 (1961; Zbl 0115.03501)]. Therefore the important new result of of this paper is the determination of \( b\); this implies that now all the invariants \( a, b, c, d \) are known. A reader of this paper should be acquainted with the author's book: [\textit{M. André}, ``Homologie des algèbres commutatives'' (Berlin 1974; Zbl 0284.18009).