Maps from the enveloping algebra of the positive Witt algebra to regular algebras (Q312770)

It is well known that for a finite dimensional Lie algebra \(\mathcal{G}\), the universal enveloping algebra \(U(\mathcal{G})\) is noetherian. It is an open question of whether the converse holds as well: namely, whether \(\mathcal{G}\) being infinite dimensional implies that \(U(\mathcal{G})\) is not noetherian. This question can be specified to concrete infinite dimensional Lie algebras, asking whether their universal enveloping algebras are noetherian. As an interesting case study, Dean and Small proposed the Witt (or centerless Virasoro) algebra \(W\), defined by basis \(\{e_n\}_{n\in \mathbb{Z}}\) and Lie brackets \([e_n,e_m]=(m-n)e_{n+m}\). Answering this question, the authors showed in [\textit{S. J. Sierra} and \textit{C. Walton}, Adv. Math. 262, 239--260 (2014; Zbl 1297.14006)] that \(U(W)\) is indeed non-noetherian. Moreover, they proved that \(U(\mathcal{G})\) is non-noetherian whenever \(\mathcal{G}\) is an infinite dimensional \(\mathbb{Z}\)-graded simple Lie algebra of polynomial growth. The methods used in [loc. cit.] are geometric in flavor. Specifically, a ring homomorphism \(\rho: U(W_+)\rightarrow k(X)[t;\tau^*]\) is constructed. Here \(W_+\) is the positive part of \(W\) (i.e.,~the \textit{positive} Witt algebra), \(k(X)\) is the ring of rational functions on \(X\subseteq \mathbb{P}^3_k\) (which is, in turn, a projective cone over a smooth conic in the projective plane), and \(k(X)[t;\tau^*]\) is a twisted polynomial extension of \(k(X)\) by an automorphism induced from a concretely determined automorphism of \(X\). Using the classification of noncommutative, birationally commutative projective surfaces, it is possible then to show that \(\mathrm{Im } \rho\) is neither left nor right noetherian. However, the former work did not result in a \textit{concrete} example of a non-finitely generated left (or right) ideal of \(U(W)\) or \(U(W_+)\). Nevertheless, the current paper suggests a more elementary approach to the study of representations of \(U(W_+)\) onto infinite dimensional algebras, which provides a concrete example of a non-finitely generated ideal. In details, the authors construct a family of ring homomorphisms \(\lambda_a:U(W_+)\rightarrow R\) for all \(a\in k\), where \(R=k\left<x,y\right>/\left<xy-yx-y^2\right>\) is the so-called \textit{Jordan plane}; this is an Artin-Schelter regular algebra of global dimension \(2\). They also construct a homomorphism \(\phi:U(W_+)\rightarrow R[t]\) (the latter being the polynomial ring over \(R\)) whose definition is similar to that of \(\lambda_a\). In fact, these homomorphisms are closely related, and the authors achieve a full understanding of their kernels and images. It turns out that while \(\mathrm{Im } \lambda_a\) is noetherian for all \(a\in k\), it is the case that \(\mathrm{Ker } \lambda_0=\mathrm{Ker } \lambda_1\) (which is generated as a two-sided ideal by the single element \(e_1e_3-e_2^2-e_4\)) is not finitely generated, neither as a left or right ideal of \(U(W_+)\). This fact is proven by observing that \(\lambda_0\) splits through \(\phi\), and a careful analysis of \(\mathrm{Im } \phi\) shows that \(\phi(\mathrm{Ker } \lambda_0)\triangleleft \mathrm{Im } \phi\) is not finitely generated. Moreover, the authors show that \(\mathrm{Ker } \phi=\cap_{a\in k} \ker \lambda_a\) is equal to \(\mathrm{Ker } \rho\); applying now the former work from [loc. cit.], it follows that \(\mathrm{Im } \phi\) is not noetherian (since \(\mathrm{Im } \phi\cong \mathrm{Im } \rho\)). It is computed that \(\mathrm{Ker } \phi\) is generated by the single element \(e_1e_5-4e_2e_4+3e_3^2+2e_6\). The maps \(\lambda_a\) can be extended to maps \(\hat{\lambda}_a\), defined from \(U(W)\) to a localization of the Jordan plane \(R\), thus enabling to prove similar results on non-finitely generated ideals in \(U(W)\). An interesting, intimately related question due to Small remains open: Is every \textit{two-sided} ideal of \(U(W_+)\) finitely generated as a \textit{two-sided} ideal? The authors prove this property holds in \(\mathrm{Im } \phi\), suggesting that a potential negative answer, if exists, should arise from different ring homomorphisms.