A unified approach to computation of integrable structures (Q384073): Difference between revisions

The authors of this interesting paper develop the idea for a unified computational approach to integrable structures including recursion, Hamiltonian, and symplectic operators. This approach is based on a geometrical theory of partial differential equations (PDE). The authors introduce the jet space \(J^{\infty }(n,m)\) with coordinates \(x^i\), \(u_{\sigma }^{j}\) (\(i=1,\dots ,n\); \(j=1,\dots ,m\); \(\sigma = i_1,i_2,\dots ,i_{|\sigma |}\)), where \(0\leq i_k\leq n\), \(i_1\leq \cdots \leq i_{|\sigma |}\). They introduce also the projection \(\pi :J^{\infty }(n,m) \to \mathbb{R}^n\) and the vector fields \[ D_i=\frac{\partial }{\partial x^i}+ \sum_{j,\sigma }u_{\sigma i}^{j}\frac{\partial } {\partial u_{\sigma }^{j}}, \quad i=1,\dots ,n. \] It is known that one may calculate by this differential operator the total derivative, and may define an \(N\)-dimensional distribution on \(J^{\infty }(n,m)\) called Cartan (or higher contact) distribution \(\mathcal{C}\). Dually, the Cartan distribution is the annihilator of the system called Cartan or ``of higher contact forms'', \[ \omega_{\sigma }^{j}= du_{\sigma }^{j} - \sum_{i}u_{\sigma i}^{j}dx^i , \quad j=1,\dots ,m, \;\;|\sigma |\geq 0. \] Given a PDE written in general form, \[ F^l(\dots , x^i,\dots ,\partial ^{|\sigma |}u^j/\partial x^{\sigma },\dots )=0 \;\;(l=1,\dots , r). \tag{\(*\)} \] Define the prolongations \(D_{\sigma }F^l=0\) (\(l=1,\dots ,r\); \(|\sigma |\geq 0\)), where \(D_{\sigma }=D_{i_1}\circ \cdots \circ D_{i_k}\) for \(\sigma = i_1\dots i_k\), and by the zeros of this define the hypersurface \(\mathcal{E}\subset J^{\infty }(n,m)\), that is a geometrical image of \((\ast )\). The authors establish here a unified computational method for integrable structures and motivate it by the existence of known difficulties in the standard methods. For instance one often reduces a nonlinear PDE to finding a Lax pair. In spite of the efficiency of the inverse scattering method and also some known standard methods, all these methods possess some drawbacks which were discussed by many authors. The geometric approach developed here is based on two constructions associated to the considered equation, namely the tangent and cotangent coverings which serve, in the geometry of PDEs, as counterparts to the tangent and cotangent bundles of finite-dimensional manifolds. These coverings are actually differential equations. It is shown that all the operators corresponding to integrable systems can be identified by higher or generalized symmetries and cosymmetries of the considered equations. Thus the main problem reduces to solving two linear equations \(\ell_{\mathcal{E}}(\phi )=0\) and \(\ell_{\mathcal{E}}^{\ast }(\psi )=0\), where \(\ell_{\mathcal{E}} \) is the linearization operator of the considered PDE, and \({\ell}_{\mathcal{E}}^{\ast }\) is its formally adjoint lifted to the tangent or cotangent coverings. The solutions that possess certain additional properties deliver the needed operators. The resulting equation is denoted by \({\mathcal{I}}({\mathcal{E}})\subset J^{\infty }(n,m)\). The known PDEs as Korteweg de Vries, Kadomtsev-Petviashvily, and Plebanski equations are considered as tutorial examples. An interesting example is the Plebanski equation \[ u_{xz}+u_{ty}+u_{tt}u_{xx}-u_{tx}^2=0. \] Its linearization is self-adjoint. Extend the (co)tangent covering with a nonlocal variable \(R\), i.e. \(R_t=u_{tx}p_t-p_z-u_{tt}p_x\) and \(R_x=p_y+u_{xx}p_t-u_{tx}p_x\). Then it turns out that the extension of the equation \(\tilde{\ell}_{\mathcal{E}}(\phi )=0\) (\(\tilde{\ell}_{\mathcal{E}}\) is the lifting of the linearization operator to corresponding covering) admits the solution \(\Phi_2=R\) besides the trivial one \(\Phi_1=p\).