On the extensions of group schemes deforming G\(_a\) to G\(_m\) (Q632474)

Let \(A\) be a discrete valuation ring with maximal ideal \(\mathfrak m\), field of fractions \(K\) of characteristic zero, and residue field \(k\) of characteristic \(p>0\). In the article under review, the author studies certain extensions of flat, commutative group schemes of finite type over \(A\) whose generic fibre is a torus, and whose special fibre is a vector group. Group schemes with this property serve to unify Kummer theory over \(K\) and Artin--Schreier--Witt theory over \(k\). One dimensional group schemes over \(A\) whose generic fibre is the multiplicative group \(\mathbb G_{m,K}\) and whose special fibre is the additive group \(\mathbb G_{a,k}\) can be classified. They are all of the form \(\mathcal G^{(\lambda)}\) for some nonzero \(\lambda\in \mathfrak m\), where, as a functor, \(\mathcal G^{(\lambda)}\) associates with an \(A\)-algebra \(B\) the group \[ \mathcal G^{(\lambda)}(B) = \{b\in B \,\,|\,\, 1+\lambda b \in B^\ast\} \] whose multiplication is given by \(b\cdot b' = b+b'+\lambda bb'\). This functor is represented by the affine scheme \[ \mathcal G^{(\lambda)} = \mathrm{spec}A[X,(1+\lambda X)^{-1}] \] which is finitely generated and flat over \(\mathrm{spec}(A)\). For \(\lambda_1,\lambda_2, \dots \in \mathfrak m\setminus\{0\}\), \textit{T. Sekiguchi} and \textit{N. Suwa} [Jap. J. Math., New Ser. 29, No. 2, 221--284 (2003; Zbl 1075.14045)] have classified the commutative extensions \[ 0 \to \mathcal G^{(\lambda_2)} \to \mathcal E^{(\lambda_1,\lambda_2)} \to \mathcal G^{(\lambda_1)} \to 0 \] and also, recursively, the extensions \[ 0 \to \mathcal G^{(\lambda_n)} \to \mathcal E^{(\lambda_1,\lambda_2, \dots, \lambda_n)} \to \mathcal E^{(\lambda_1,\lambda_2, \dots, \lambda_{n-1})} \to 0 \] for any \((n-1)\)-fold extension on the right. In other words, the groups \[ \mathrm{Ext}^1_A(\mathcal E^{(\lambda_1, \dots, \lambda_{n-1})},G^{(\lambda_n)}) \] have been fully described, where \(\mathrm{Ext}^1_A\) means extensions in the category of commutative group schemes over \(A\). Necessarily, the generic fibre of any such iterated extension is a split torus, and its special fibre is a vector group. While extension of groups of the form \(\mathcal E^{(\lambda_1, \dots, \lambda_{n-1})}\) by \(G^{(\lambda_n)}\) are not particularly hard to classify, the reverse situation, that is, extensions of \(G^{(\lambda_n)}\) by \(\mathcal E^{(\lambda_1, \dots, \lambda_{n-1})}\) are somewhat more involved. The main result of the article consists of explicit descriptions of the groups \[ \mathrm{Ext}^1_A(G^{(\lambda_0)}, \mathcal E^{(\lambda_1,\lambda_2)}) \qquad \mathrm{and}\qquad \mathrm{Ext}^1_A(G^{(\lambda_0)}, \mathcal E^{(\lambda_1,\lambda_2, \lambda_3)}) \] the description of the left hand one being attributed to D. Horikawa. The adjective \textit{explicit} really means what it means: For the right hand group, say, a homomorphism \(\Theta\) between certain modules over Witt rings depending on \(\lambda_0\) and \(\mathcal E^{(\lambda_1,\lambda_2, \lambda_3)})\) is written down cash. Then an isomorphism \[ \ker(\Theta) \longrightarrow \mathrm{Ext}^1_A(G^{(\lambda_0)}, \mathcal E^{(\lambda_1,\lambda_2, \lambda_3)}) \] is given, by assigning to every element of \(\ker(\Theta)\) an extension class. As an affine scheme this extension is given as a localisation of \(\mathrm{spec}A[X_0,\dots,X_3]\), with denominators being expressed by certain (absolute) power series.