Singular projective bases and the affine Bol operator (Q1883398)

Let \({\mathcal V}= \text{Vect}(S^1)= \text{Span}_{\mathbb C}\{e_n=z^{n+1}\frac{d}{dz}\,| \,n\in{\mathbb Z}\}\) be the complex Lie algebra of polynomial vector fields on the circle \(S^1\). The projective subalgebra is denoted by \({\mathfrak a}= \text{Span}_{\mathbb C}\{e_0,e_{-1},e_1\}\), a copy of \(sl_2({\mathbb C})\) in \({\mathcal V}\). The article treats \({\mathcal V}\)-modules of linear differential operators between spaces of tensor densities \(A(\gamma)=\bigoplus_{0\leq \text{Re}(a)<1}A(a,\gamma)\) where \(A(a,\gamma)= \text{Span}_{\mathbb C}\{z^{\lambda-\gamma}dz^{\gamma}\mid \gamma\in a+{\mathbb Z}\}\) for \(a,\gamma\in{\mathbb C}\): for \(p\in{\mathbb C}\), let \(E(\gamma,p)\) be the \({\mathcal V}\)-module of differential operators from \(A(\gamma)\) to \(A(\gamma+p)\). The \({\mathcal V}\)-action is given by the bracket of the associative algebra of endomorphisms of the space of tensor densities. \(E^k(\gamma,p):=\text{Span}_{\mathbb C}\{(z^{-1}dz)^pz^{\mu}(e_0)^i\mid \mu\in{\mathbb C},0\leq i\leq k\}\) gives the order filtration of \(E(\gamma,p)\), which is preserved by the \({\mathcal V}\)-action. The subquotients \(E^k(\gamma,p)/E^{k-1}(\gamma,p)\) are \({\mathcal V}\)-isomorphic to \(A(p-k)\). One starting point of the present article is the question whether \(E^{k-1}(\gamma,p)\) admits a (unique) \({\mathfrak a}\)-invariant complement in \(E^k(\gamma,p)\). This is precisely the case whenever the eigenvalue of the \({\mathfrak a}\)-Casimir \(Q\) on \(E^k(\gamma,p)/E^{k-1}(\gamma,p)\), which is \((p-k)(p-k-1)\), is distinct from the \(Q\)-eigenvalue on the lower order subquotients. The condition on \(k\) reads thus \(k\not= p, p+\frac{1}{2}, \ldots,2p-1\). Under this condition, there is a unique \(Q\)-eigenvector \(\epsilon_{\mu}^k(\gamma,p)\) of weight \(\mu\) and symbol \(z^{\mu-p}dz^pe_0^k\). The authors distinguish the cases \(p\notin 1+{\mathbb N}/2\) as \textit{regular}, and otherwise as singular. In the regular case treated by \textit{P. B. Cohen}, \textit{Yu. Manin} and \textit{D. Zagier} [Automorphic pseudodifferential operators, Prog. Nonlinear Differ. Equ. Appl. 26, 17--47 (1997; Zbl 1055.11514)]; \textit{P. B. A. Lecomte} and \textit{V. Yu. Ovsienko} [Lett. Math. Phys. 49, No. 3, 173--196 (1999; Zbl 0989.17015)]; \textit{H. Gargoubi} [Bull. Soc. R. Sci. Liège 69, No. 1, 21--47 (2000; Zbl 0986.17008)] and \textit{C. H. Conley} [Int. Math. Res. Not. 2001, No. 12, 609--628 (2001; Zbl 1056.17020)], all \(E^k(\gamma,p)\) possess \({\mathfrak a}\)-complements, and \(E(\gamma,p)\) is \({\mathfrak a}\)-isomorphic to \(\bigoplus_{k=0}^{\infty}A(p-k)\) by the so-called projective symbol map \(PS(\gamma,p)\). Bases can be written down, on which the action is given explicitly. Conjugating by the map \(PS\), one gets a \({\mathcal V}\)-representation on \(\bigoplus_{k=0}^{\infty}A(p-k)\) written in terms of infinite matrices. These turn out to be lower triangular. Their elements are interpreted as \({\mathfrak a}\)-relative \(1\)-cochains of \({\mathcal V}\) with values in \(E(\gamma,p)\), and can be given explicitly: \(\pi_{ij}=b_{ij}\beta_{i-j}(j)\) for the off-diagonal elements with explicit formulae for \(b_{ij}\) and \(\beta_p(\gamma)\). The main goal is now to have a similar treatment of the singular case. The authors define a splitting in this case giving rise to a linear isomorphism \[ PS(\gamma,p):E(\gamma,p)\to\bigoplus_{k=0}^{\infty}A(p-k), \] which is not an \({\mathfrak a}\)-morphism any more. They examine the resulting matrix. The function \(b:(\gamma,p)\mapsto b_{p-k,p-l}(\gamma,p)\) parametrizing the off-diagonal elements in the regular case shows up a simple pole in \(p\) for certain elements in the singular domain. In order to define still a meaningful basis, one passes to a linear combination which possesses a removable singularity in \(p\). The resulting well-defined basis is called singular projective basis. The main theorem of the article computes explicitly the \({\mathcal V}\)-action in terms of this basis. Some off-diagonal matrix elements denoted \(\alpha_p(\gamma)\) are only \({\mathfrak b}\)-relative cochains where \({\mathfrak b}= \text{Span}_{\mathbb C}\{e_{-1},e_0\}\) is the affine subalgebra of \({\mathfrak a}\). The authors define a projective symbol map \(\overline{PS}\) which is here a \({\mathfrak b}\)-isomorphism. The proof of the main theorem uses many properties of the different matrix elements. Therefore, the authors define first operators abstractly, showing the wanted properties, and then identify them with the matrix elements. The abstract part uses the affine Bol operator \(\text{Bol}_p(\gamma)=dz^p (\frac{d}{dz} )^p\) and its property of being the unique \({\mathfrak b}\)-invariant operator in \(E(\gamma,p)\) up to scalars (if it exists at all). One main point is to regard \(\text{Bol}_p(\gamma)\) as a \(0\)-cochain and to compute its coboundary \(\partial \text{Bol}_p(\gamma)=-p\gamma_p\alpha_p+b_{2,0}\beta_p\) where \(\gamma_p=\gamma+\frac{(p-1)}{2}\). The authors discuss in detail overlap and differences with Cohen-Manin-Zagier's article (loc. cit.) and Gargoubi's work (loc. cit.). Finally, they use transvectants in order to generalize Cohen-Manin-Zagier's (regular) composition formula which gives the composition of the \(\varepsilon_{\mu}^k(\gamma,p)\) in the regular case to the singular case.