Simple holonomic modules over the second Weyl algebra \(A_2\) (Q1971632)

Let \(K\) be an algebraically closed uncountable field of characteristic zero. The first Weyl algebra \(A_1=A_1(K)\) is generated over \(K\) by \(X\) and \(\delta\) subject to the commutation rule \(\delta X-X\delta=1\). The \(n\)th Weyl algebra \(A_n\) is the tensor product over \(K\) of \(n\) copies of \(A_1\). A famous result of Bernstein shows that if \(M\) is a nonzero module over \(A_n\) then \(\text{GKdim}(M)\geq n\). Modules with \(\text{GKdim}(M)=n\) are said to be holonomic modules. For the first Weyl algebra all simple modules are holonomic modules. Holonomic modules over \(A_n\) are an important class of modules which have many desirable properties. For example, any finitely generated holonomic module has finite length. It was an open question for many years as to whether or not all simple \(A_n\)-modules were holonomic. J. T. Stafford was the first to construct a counterexample of a nonholonomic simple module. Subsequently, Bernstein and Lunts have shown that ``almost all'' elements \(x\) in \(A_n\) generate maximal left ideals, and as a corollary, the factor modules \(A_n/A_nx\) are nonholonomic simple modules. \textit{J. Dixmier}, in his book ``Enveloping algebras'' (1977; Zbl 0339.17007), was pessimistic that it would ever be possible to classify the simple modules over \(A_1\). However, in 1981 Block managed to do this. In the last few years, the first author has given an alternative proof of the classification for \(A_1\) using techniques developed to study generalized Weyl algebras. This, finally, brings us to the content of the paper under review. It is surely beyond the reach of present techniques to hope to describe all simple modules over the second Weyl algebra, \(A_2\); however, the authors do classify all of the simple holonomic modules over \(A_2\), by using generalized Weyl algebra techniques. One obvious source of holonomic \(A_2\)-modules is to take the tensor product of two holonomic \(A_1\)-modules; these are regarded as the trivial holonomic \(A_2\)-modules. Thus, the task is to describe the nontrivial holonomic modules. The first Weyl algebra is a simple Noetherian domain, and, consequently, has a division ring of fractions \(k\). One can consider \(A_2=A_1\otimes A_1\) as a subalgebra of \(k\otimes A_1\cong A_1(k)\). As such, \(A_1(k)\) is an Ore localization of \(A_2\). If \(M\) is a left \(A_2\)-module, then denote by \(\widetilde M\) the module \(A_1(k)\otimes M\). If \(M\) is a simple \(A_1\)-module then there are three obvious possibilities for \(\widetilde M\): (i) \(\widetilde M=0\), (ii) \(\widetilde M\) is nonzero, but finite-dimensional as a left \(k\)-module, (iii) \(\widetilde M\) is infinite-dimensional as a left \(k\)-module. The authors show that case (i) occurs exactly when \(M\) is a trivial holonomic module, and case (ii) occurs exactly when \(M\) is a nontrivial holonomic module. The trivial holonomic modules might be considered as known since they are tensor products of two simple holonomic \(A_1\)-modules, each of which is described by Block. Thus, we need to look at case (ii). For any simple \(A_2\)-module \(M\), if \(\widetilde M\) is nonzero then it is a simple \(A_1(k)\)-module. If we are in case (ii) so that \(M\) is such that \(\widetilde M\) is a nonzero simple \(A_1(k)\)-module that is finite-dimensional over \(k\), then it is easy to see that \(M\) is the socle of \(\widetilde M\) regarded as an \(A_2\)-module. The main result in the paper is that this actually provides a bijection between nontrivial holonomic modules and simple \(A_1(k)\)-modules which are finite-dimensional over \(k\). The hard part of the proof is to show that such an \(A_1(k)\)-module has got a nonzero socle when regarded as an \(A_2\)-module. Finally, the \(k\)-finite-dimensional simple \(A_1(k)\)-modules are classified up to pairs of irreducible elements. The results of this paper are obtained in a wider context of tensor products of certain generalized Weyl algebras, the Weyl algebras being the most important examples.