Enveloping fields of Witt type algebras (Q1414684)

After some generalities on filtrations (for associative and Lie algebras) and their associated gradings, the author defines algebras of Witt and of Virasoro types, following [\textit{R. W. T. Yu}, Commun. Algebra 25, 1471--1484 (1997; Zbl 0878.17027)]. Given an Abelian group \(\Gamma\) and a function \(f\) on \(\Gamma\) with values in \(k\) such that \(f(0)= 0\), a Lie algebra of Witt type on \(\Gamma\) over a field \(k\), \({\mathfrak w}(f)\), is given by a \(k\)-basis indexed by \(\Gamma\) with multiplication \([e_\alpha, e_\beta]= (f(\beta)- f(\alpha)) e_{\alpha+\beta}\). Of particular interest is the case where \(\Gamma= \mathbb{Z}^d\) and \(f\) is an additive function; then \({\mathfrak w}(f)\) has a natural \(\Gamma\)-grading, called the weight such that wt\((e_\alpha)= \alpha\) for \(\alpha\in \Gamma\). This algebra \({\mathfrak w}(f)\) is simple precisely when \(f\) is injective [R. W. T. Yu (loc. cit.), Theorem 3.7] and in that case \({\mathfrak w}(f)\) admits a unique nontrivial central extension up to isomorphism [R. W. T. Yu (loc. cit.), Theorem 5.1], called the Virasoro algebra associated to \(f\) and denoted by \(\text{Vir}(f)\). As a vector space we have \(\text{Vir}(f)= \bigoplus_{\alpha\in\Gamma} k.e_\alpha\oplus k.z\), while the bracket is given by \([e_\alpha, e_\beta]= (f(\beta)- f(\alpha)) e_{\alpha+\beta}+\delta_{\alpha+\beta,0} (f(\alpha)^3- \delta(\alpha))z\), where \(z\) is central. This yields the exact sequence of \(\Gamma\)-graded Lie algebras \(0\to k.z\to \text{Vir}(f)\to{\mathfrak w}(f)\to 0\) (where wt\((z)= 0\)). When \(\Gamma= \mathbb{Z}\) and \(f\) is the inclusion map in the complex numbers, then \({\mathfrak w}(f)\) is the Witt algebra \(\text{Der}_C C[t,t^{-1}]\) and its extension is the classical Virasoro algebra. If \({\mathcal E}\) is a subset of \(\Gamma\) and \({\mathfrak w}_{\mathcal E}(f)\) the corresponding graded subspace of \({\mathfrak w}(f)\), then it is a subalgebra of \({\mathfrak w}(f)\) if and only if the sum of any two distinct elements of \({\mathcal E}\) is again in \({\mathcal E}\). The author describes the four possible cases, where \({\mathcal E}\) has 1, 2, 3 or infinitely many elements, and proves the following Theorem. \(\text{Dim}^3{\mathcal U}({\mathfrak w}(f))= \text{Dim}^3{\mathcal U}(\text{Vir}(f))= d\), where \({\mathcal U}\) denotes the enveloping algebra and both these enveloping algebras possess a skew field of fractions. Now assume \(\Gamma\) to be ordered and choose a graded subalgebra \({\mathfrak g}\) of \({\mathfrak w}(f)\) or \({\mathfrak g}= \text{Vir}(f)\). Put \({\mathcal U}={\mathcal U}({\mathfrak g})\) and \(K= K({\mathfrak g})= \operatorname {Frac}{\mathcal U}({\mathfrak g})\), equipped with the canonical filtrations. The associated graded algebras are denoted \({\mathcal P}= \text{Gr}({\mathcal U})\) and \({\mathcal Q}= \text{Gr}(K)\), while \(\{\dots\}\) denotes the associated Poisson bracket. The associative algebra \({\mathcal P}\) has a natural identification with a polynomial algebra in an infinity of variables \(E_\alpha= \text{gr}(e_\alpha)\) (where \(\alpha\) ranges over a subset of \(\Gamma\)), and an extra variable \(\mathbb{Z}= \text{gr}(z)\) for the Virasoro algebra. The algebra \({\mathcal Q}\) is then identified with the algebra of rational fractions in the same variables, with homogeneous denominators. We put \(\partial_\alpha (\varphi)= \partial \varphi/ \partial E_\alpha\) where \(\alpha\in \Gamma\) and \(\varphi\in {\mathcal Q}\). For any \(\varphi\in {\mathcal Q}\) we put \(\text{MV}(\varphi)= \sup \{\alpha\in \Gamma\mid \partial_\alpha (\varphi)\neq 0\}\in \Gamma\cup \{-\infty\}\). Thus for \({\mathfrak g}= \text{Vir}(f)\) we have \(\text{MV} (\varphi)= -\infty\) if and only if \(\varphi\in k[\mathbb{Z},\mathbb{Z}^{-1}]\). For \(\gamma\in \Gamma\) we put \({\mathcal Q}(\gamma)= \{\varphi \in{\mathcal Q}\mid \text{MV}(\varphi)\leq \gamma\}\) and \({\mathcal Q}^-(\gamma)= \{\varphi\in {\mathcal Q}\mid \text{MV}(\varphi)< \gamma\}\). Now the author proves the Proposition. (1) For any \(\gamma\in \Gamma\), \({\mathcal Q}(\gamma)\) is a subalgebra of \({\mathcal Q}\) admitting inversion of elements invertible in \({\mathcal Q}\). (2) For \(\alpha,\beta\in \Gamma\), if \(\alpha,\beta\geq 0\), then \(\{{\mathcal Q}(\alpha),{\mathcal Q}(\beta)\}\subseteq {\mathcal Q}(\alpha+\beta)\). (3) Let \(\varphi,\psi\in{\mathcal Q}\) be such that \(\text{MV}(\varphi), \text{MV}(\psi)> 0\); then \(\text{MV} (\{\varphi,\psi\})= \text{MV}(\varphi)+ \text{MV}(\psi)\) if and only if \(\text{MV}(\varphi)\neq \text{MV}(\psi)\). (4) For any \(\varphi\in {\mathcal Q}(\gamma) \setminus {\mathcal Q}^-(\gamma)\) the family \(\{\varphi^m\}_{m\geq 0}\) is free over \({\mathcal Q}^-(\gamma)\). In particular, if \(\psi\in{\mathcal Q}^-(\gamma)\), then \(\text{MV} (\varphi\psi)= \text{MV}(\varphi)\). Further he shows that for a finite subset \(X\) of \({\mathcal Q}\) and \(N\geq 1\), if \({\mathcal Q}_X(N)\) denotes the subspace generated by the Poisson monomials of length at most \(N\) in the elements of \(X\), and if \({\mathcal Q}_X(N)\) contains elements \(x_0,\dots, x_d\) such that \(0< \text{MV}(x_0)< \text{MV}(x_1)< \cdots< \text{MV}(x_d)\) and the family \(\{\text{MV}(x_1),\dots, \text{MV}(x_d)\}\) is free over \(\mathbb{Z}\) in \(\Gamma\), then \(\text{Dim}^3 (\gamma_X)\geq d\). This leads to the Theorem: If \(f: \mathbb{Z}^d\to k\) is an injection, then \[ \text{TrDeg}^3 K({\mathfrak w}(f))= \text{TrDeg}^3 K(\text{Vir}(f))= d. \] Next he determines the centres, showing that the centres of \(K\) and \({\mathcal U}\) for \({\mathfrak w}(f)\) are just \(k\), while for \(\text{Vir}(f)\) they are \(k[z]\) and \(k(z)\), respectively. Now let \(f\), \(g\) be two injections of \(\Gamma\) in \(k\). Then there is equivalence between isomorphisms of the Witt algebras, the enveloping algebras of the Witt algebras, the enveloping algebras of the Virasoro algebras relating to \(f\) and \(g\), and equality of the orbits of \(f(\Gamma)\) and \(g(\Gamma)\) under the action of \(k^*\). There is a similar but more complicated comparison between their skew fields of fractions, illustrated by some examples. When \(\Gamma\) is totally ordered, \(\Gamma_+, {\mathfrak w}_+= {\mathfrak w}_+(f)\) denote the strictly positive part and \(= \text{Gr}(K({\mathfrak w}_+))\) with the Poisson bracket \(\{\dots\}\) and \(\text{ad}(x): y\mapsto \{x,y\}\) \((x,y\in{\mathcal Q})\) the adjoint mapping. The elements of \({\mathcal Q}\) may be regarded as rational fractions in the variable \(E_\gamma\), where \(\gamma\in \Gamma\) and \(\partial_\gamma\) denotes the partial derivation \(\partial/\partial E_\gamma\). Further, for any \(\varphi\in{\mathcal Q}\), \(\text{MV}(\varphi)= \sup\{\gamma\in \Gamma_+\mid \partial_\gamma(\varphi)\neq 0\}\in \Gamma_+\cup \{-\infty\}\) and for \(\varphi,\psi\in{\mathcal Q}\) we put \(\{\varphi, \psi\}_{\alpha,\beta}= \partial_\alpha \varphi. \partial_\beta\psi- \partial_\beta \varphi. \partial_\alpha \psi\). The author also determines the centralizers of elements of \({\mathcal Q}\) relative to Poisson bracket and in the enveloping algebra. In particular, for any Lie subalgebra \({\mathfrak g}\) of \({\mathfrak w}_+(f)\) a non-constant element of \({\mathcal U}({\mathfrak g})\) has a commutative centralizer in \({\mathcal U}({\mathfrak g})\), and the same is shown for a subalgebra of \(K({\mathfrak w}_+(f))\). Further, maximal commutative subalgebras are determined; in the case of \({\mathcal U}({\mathfrak w}_+)\) they are shown to have transcendence degree one.