A shape lemma for ideals of differential operators (Q6982192)

Let \(C\) be a field of characteristic zero and \(K=C(x, \mathbf{y}) = C(x, y_{1},\dots, y_{n})\) the field of rational functions in \(x\) and \(y_{1},\dots, y_{n}\) (the set of the last \(n\) indeterminates is abbreviated as \(\mathbf{y}\)) treated as a differential field with partial derivations \(D_{x}=\partial/\partial x\), \(D_{y_{i}}=\partial/\partial y_{i}\) (\(1\leq i\leq n\)). Let \(K[D_{x}, \mathbf{D}_{y}] = K[D_{x}, D_{y_{1}},\dots, D_{y_{n}}]\) denote the ring of linear differential operators with rational function coefficients, \(K[D_{x}, \mathbf{D}_{y}] = \{\sum_{i, j_{1},\dots,j_{n}=0}^{d}a_{i, j_{1},\dots,j_{n}}(x,\mathbf{y})D_{x}^{i}D_{y_{1}}^{j_{1}}\dots D_{y_{n}}^{j_{n}}\,|\,d\in\mathbb{N}, a_{i, j_{1},\dots,j_{n}}\in K\}\) (in this ring \(D_{x}\cdot x = x\cdot D_{x} + 1\) and \(D_{y_{i}}\cdot y_{i} = y_{i}\cdot D_{y_{i}} + 1\) for \(i=1,\dots, n\)). If \(I\) is an ideal of dimension zero in \(K[D_{x}, \mathbf{D}_{y}]\) such that \(r = \dim_{K}K[D_{x}, \mathbf{D}_{y}]/I\) is finite, then \(I\) is said to be in \textit{normal position} (w.r.t. \(D_{x}\)) if for every choice of \(C\)-linearly independent solutions \(f_{1},\dots, f_{r}\) (elements of the solution space of \(I\)), \(\mathrm{Wr}_{x}(f_{1},\dots, f_{r})\neq 0\). (\(\mathrm{Wr}_{x}(f_{1},\dots, f_{r})\) denotes the Wronskian of \(f_{1},\dots, f_{r}\) w.r.t. \(x\).)\N\NThe main result of the paper under review is the following generalization of the shape lemma for zero-dimensional polynomial ideals (see Theorem 3.7.25 in [\textit{M. Kreuzer} and \textit{L. Robbiano}, Computational commutative algebra. I. Berlin: Springer (2000; Zbl 0956.13008)]). Let \(I\subseteq K[D_{x}, \mathbf{D}_{y}]\) be an ideal of dimension zero. Let \(P\) be a generator of \(I\bigcap K[D_{x}]\). Then the following conditions are equivalent: (1) I is in normal position w.r.t. \(D_{x}\); (2) \(\mathrm{ord}(P) = \dim_{K}K[D_{x}, \mathbf{D}_{y}]/I\); (3) \(K[D_{x}]/\langle P\rangle\) and \(K[D_{x}, \mathbf{D}_{y}]/I\) are isomorphic as \(K[D_{x}]\)-modules; (4) There are \(Q_{1},\dots, Q_{n}\in K[D_{x}]\) with \(\mathrm{ord}(Q_{i}) <\mathrm{ord}(P)\) (\(1\leq i\leq n\)) such that \(\{D_{y_{1}} - Q_{1},\dots, D_{y_{n}} - Q_{n}, P\}\) is a basis of \(I\). \, Another serious result obtained in the paper concerns \(D\)-radical ideals in \(K[D_{x}, \mathbf{D}_{y}]\). (A zero-dimensional ideal \(I\subseteq K[D_{x}, \mathbf{D}_{y}]\) is called \textit{\(D\)-radical} if its solution space \(V(I)\) is linearly disjoint (over \(C\)) with \(K\), that is, any \(C\)-basis of \(V(I)\) is \(K\)-linearly independent). The authors prove that if \(I\subseteq K[D_{x}, \mathbf{D}_{y}]\) is a zero-dimensional \(D\)-radical ideal, then there exist constants \(c_{1},\dots, c_{n}\in C\) such that the ideal \(J\) obtained from \(I\) by applying the linear change of variables \(\mathbf{y} \leftarrow \mathbf{y} +\mathbf{c}x\) (where \(\mathbf{c} = (c_{1},\dots, c_{n})\)) is in normal position w.r.t. \(D_{x}\).