Laplace transformation in higher dimensions (Q1922426)

The theory of Laplace transformations is a very well developed chapter of differential geometry of conjugate nets on surfaces in a three-dimensional projective space \(\mathbb{P}^3\). For the first time the geometric theory of Laplace transformations of two-dimensional surfaces in \(\mathbb{P}^3\) was constructed by Darboux in 1872. In the 1940's--1950's this theory was extended to the \(m\)-conjugate systems \((m > 2)\) in \(\mathbb{P}^n\). Chern (1944 and 1947) generalized Laplace transformations to \(m\)-conjugate systems. He considered Laplace transformations of \(m\)-conjugate systems that are Cartan varieties: for them, the first osculating space is \(2m\)-dimensional, that is, he considered the case when \(n \geq 2m\). The authors consider Cartan varieties \(V^m\) in Euclidean space of dimension \(2m\), that is, they consider Cartan varieties belonging to their \((2m)\)-dimensional osculating subspace. The functions giving the parametrization \(X(x^1,\dots,x^m)\) of \(V^m\) satisfy a second-order overdetermined system of partial differential equations of the form \[ X_{ij} = \Gamma^i_{ij} X_i + \Gamma^j_{ij} X_j,\quad 1 \leq i \neq j \leq m, \tag{1} \] where \(X_i\) and \(X_{jk}\) denote the derivatives of \(X\) with respect to \(x^i\) and the second derivatives of \(X\) with respect to \(x_i\) and \(x_j\), respectively, \(\Gamma^k_{ij}\) is the Christoffel symbol, and there is no summation on repeated indices. Following Chern's paper, the authors obtain \(m(m-1)\) Laplace transformations for a given Cartan variety. To each of these transformations there corresponds an overdetermined system of the same form (1). Next, the authors obtain a transformation for an overdetermined system of the form \[ y_{,lk} + a^l_{lk} y_{,l} + a^k_{lk} y_{,k} + c_{lk}y + h_{lk} = 0, \quad 1 \leq k\neq l\leq m, \tag{2} \] where \(y_{,k}\) and \(y_{,lk}\) denote the derivatives of \(y\) with respect to \(x^k\) and the second derivatives of \(y\) with respect to \(x^l\), and \(x^k\), respectively. This transformation generalizes the classical Laplace transformation for linear second-order differential hyperbolic equations \(z_{uv}+ az_u+ bz_v+ cz+ l= 0\) in the plane. The authors apply the obtained transformation of (2) to the problem of solving (2) for smooth Cauchy data given by prescribing the values of \(y\) along \(m\) curves passing through a given point \(x^0\). They show that the cases where a Laplace transformation of \(V^m\) degenerates to a curve (in such cases some special invariants vanish; the authors call them Laplace invariants but in the literature they are called Darboux invariants) are exactly those for which the integration of the original system of PDE reduces to that of a system of the same type involving \(m - 1\) independent variables (reduction theorem). This produces a method of integration for such systems. All this is illustrated by examples. The authors note that the conserved quantities for semi-Hamiltonian, strongly hyperbolic systems of hydrodynamic type [see \textit{S. P. Tsarev}, Math. USSR, Izv. 37, No. 2, 397-419 (1991); translation from Izv. Akad. Nauk SSSR, Ser. Mat. 54, No. 5, 1048-1068 (1990; Zbl 0796.76014)] are precisely governed by overdetermined systems (2) to which the higher-dimensional Laplace transformation is applicable.