On the classification of elliptic systems of second order partial differential equations (Q1382866)

For systems of \(n\) linear second-order partial differential equations with constant coefficients \(a_{ij}\) \((i,j= 1,\dots, n)\), with \(n\) unknown functions in \(n\) independent variables a classification problem arises because the correctness of the Dirichlet problem for elliptic systems in the sense of Petrovskii can fail. Here transformations of the system are taken that means multiplying the characteristic matrix \(\Lambda(\Xi)\), \(\Xi= (\xi_1,\dots,\xi_n)\), with the elements \(a_{ij}(\Xi)\), from the left by a constant non-degenerate matrix. Consider only the reduced form, having the minimal number of negative eigenvalues and then put into the same class all matrices \(\Lambda(\Xi)\) which can be deformed one into another by changing continuously the coefficients \(a_{ij}\) so that the determinant of the matrix remains vanishing only for \(\xi_1=\cdots= \xi_n= 0\). Further, consider symmetric systems, so that all eigenvalues are real. Let \(l_-\) and \(l_+\) denote the number of negative and positive eigenvalues and \(l:=\min(l_-, l_+)\). Let \(p:= [n/2]-l\) and call it the ellipticity characteristic of the system. There is the question, if there is a relation between \(p\) and the correctness of the Dirichlet problem. To this end then the author constructs and investigates some concrete examples of systems with given \(p\).