Prolongations of Dirac structures related to Weil bundles (Q2263088)

The paper situates itself in a stream of similar works discussing prolongations and lifts of certain geometric structures from a basic manifold to its Weil bundle. The research in this direction was started perhaps by Morimoto in the 70s with a series of papers about tangent and higher tangent lifts of tensor fields. After an important contribution of Kolář and his collaborators it become apparent that the functors \(T\) and \(T^k\) are just special cases of a general concept of a \textit{Weil functor} \(T^{\mathbb{A}}\) (here \(\mathbb{A}\) denotes a \textit{Weil algebra}). The latter can be characterized as product preserving bundle functors on the category of smooth manifolds [\textit{I. Kolář} et al., Natural operations in differential geometry. Berlin: Springer-Verlag (1993; Zbl 0782.53013)]. Since then several researchers (including V. Shurygin and the authors) began to study natural lifts (induced by Weil functors \(T^{\mathbb{A}}\)) of various geometric structures. In general, the lifting procedure for an arbitrary tensor field on a manifold \(M\) to an object of the same type on the Weil bundle \(T^{\mathbb{A}}M\) consists of two steps: an application of the Weil functor \(T^{\mathbb{A}}\) and then an application of a certain ``shifting isomorphism'' which, informally speaking, puts functors in the proper order. For example if \(X:M\rightarrow T M\) is a vector field, then its \(\mathbb{A}\)-lift is just the composition \[ X^{\mathbb{A}}=\kappa_{\mathbb{A}}\circ T^{\mathbb{A}}X:T^{\mathbb{A}}M\longrightarrow T^{\mathbb{A}}T M \longrightarrow T T^{\mathbb{A}}M\;, \] where \(\kappa_{\mathbb{A}}:T^{\mathbb{A}} T M\rightarrow T T^{\mathbb{A}} M\) is the canonical flip. The same idea works for every covariant tensor field. In order to deal with contravariant tensor fields one needs to introduce a contravariant analog of \(\kappa_{\mathbb{A}}\), i.e., the isomorphism \(\varepsilon_{\mathbb{A}}:T^{\mathbb{A}}T^\ast M\rightarrow T^\ast T^{\mathbb{A}} M\). The structure of a Weil algebra on its own does not allow to define such a map in a canonical way -- we need to equip \({\mathbb{A}}\) with an additional structure, i.e., a non-degenerate \({\mathbb{R}}\)-linear map \(p:{\mathbb{A}}\rightarrow\mathbb{R}\) (such pair \(({\mathbb{A}},p)\) is called a \textit{Frobenius-Weil algebra}). Now, \(\varepsilon_{\mathbb{A}}\) can be defined to be the unique map such that the pair \((\kappa_{\mathbb{A}},\varepsilon_{\mathbb{A}})\) intertwines the pairings \(p\circ T^{\mathbb{A}}\langle\cdot,\cdot\rangle_M:T^{\mathbb{A}}TM\times_{T^{\mathbb{A}}M}T^{\mathbb{A}}T^\ast M\rightarrow T^{\mathbb{A}}{\mathbb{R}}={\mathbb{A}}\rightarrow{\mathbb{R}}\) and \(\langle\cdot,\cdot\rangle_{T^{\mathbb{A}}M}:TT^{\mathbb{A}}M\times_{T^{\mathbb{A}}M}T^\ast T^{\mathbb{A}}M\rightarrow {\mathbb{R}}\). Given a differential form \(\psi:M\rightarrow T^\ast M\) we can define its \(p\)-lift as the composition \[ \psi^{(p)}=\varepsilon_{\mathbb{A}}\circ T^{\mathbb{A}}\psi :T^{\mathbb{A}}M\longrightarrow T^{\mathbb{A}}T^\ast M\longrightarrow T^\ast T^{\mathbb{A}} M\;. \] The same idea works for a general contravariant tensor. The paper briefly recalls the above constructions (with some minor modifications) and also provides a construction of a \({\mathbb{A}}\)-lift of a Lie algebroid and a (singular) foliation. The new result is an application of the general lifting philosophy sketched above to a particular case of an (almost) Dirac structure. Recall that given a manifold \(M\), an \textit{almost Dirac structure} on \(M\) is a subbundle \(L\subset T M\oplus_MT^\ast M=:\mathcal{P}M\) which is maximally isotropic with respect to the pairing \(\langle\cdot,\cdot\rangle_{+}:\mathcal{P} M\times\mathcal{P} M\rightarrow \mathbb{R}\) given by \[ \langle X\oplus\alpha,Y\oplus\beta\rangle_{\pm}:=\frac 12 \left(\langle Y,\alpha\rangle_M\pm\langle X,\beta\rangle_M\right)\;, \] where \(\langle\cdot,\cdot\rangle_M:TM\times T^\ast M\rightarrow \mathbb{R}\) denotes the canonical pairing. We call \(L\) a \textit{Dirac structure} if additionally the space of sections \(\Gamma(L)\) is closed under the following skew-symmetric bracket \[ [X\oplus\alpha,Y\oplus\beta]:=[X,Y]\oplus\left(\mathcal{L}_X\alpha-\mathcal{L}_Y\beta- \text{d}\langle X\oplus\alpha,Y\oplus\beta\rangle_{-}\right)\;. \] The importance of Dirac structures lies in the fact that they provide a common generalization of both symplectic and Poisson structures. Namely, given a symplectic manifold \((M,\omega)\) (resp. a Poisson manifold \((M,\Lambda)\)) the graph of the natural map \(\widetilde\omega:T M\rightarrow T^\ast M\) (resp. of \(\widetilde\Lambda:T^\ast M\rightarrow T M\)), being a subbundle of \(T M\oplus T^\ast M\), is naturally a Dirac manifold. Now given a certain Dirac structure \(\iota:L\subset T M\oplus T^\ast M\) it is natural to define its \textit{\(p\)-prolongation} as the composition \[ (\kappa_{\mathbb{A}}\oplus\varepsilon_{\mathbb{A}})\circ T^{\mathbb{A}}\iota :T^{\mathbb{A}}L\subset T^{\mathbb{A}}T M\oplus T^{\mathbb{A}}T^\ast M\longrightarrow T T^{\mathbb{A}} M\oplus T^\ast T^{\mathbb{A}} M\;. \] The main result of the paper states that if \(L\subset\mathcal{P}M\) is an (almost) Dirac structure then it is \(p\)-prolongation \(T^{\mathbb{A}}L\subset \mathcal{P}T^{\mathbb{A}}M\). This result is supplemented with several other ones which discuss the compatibility of the \(p\)-prolongation of a Dirac structure with other prolongations known from the literature. For example if \(E\) is a Lie algebroid naturally associated with a given Dirac structure \(L\subset \mathcal{P}M\), then the Lie algebroid associated with the \(p\)-prolongation \(T^{\mathbb{A}}L\) is precisely the \(\mathbb{A}\)-lift of the initial algebroid \(E\). Finally let me comment on the technical side of the paper. Most of the proofs are done by calculations in the natural coordinates induced by a particular choice of a basis \(\{e_\alpha\}\) of \(\mathbb{A}\) regarded as a linear space over \(\mathbb{R}\). In fact, the authors do not use \(\mathbb{A}\)-lifts of covariant objects directly, but rather work with certain \(\alpha\)-lifts which can be obtained form \(\mathbb{A}\)-lifts using the fixed basis \(\{e_\alpha\}\). This seems to me to be an unnecessary complication. Probably, using \(\mathbb{A}\)-lifts (and their \(\mathbb{A}\)-linearity) directly the calculations could be significantly simplified.