Efficient expansion of subspaces in the Jacobi-Davidson method for standard and generalized eigenproblems (Q1279552)

The treatment of approximations leading to the Jacobi-Davidson correction equation (for matrices) of the form \[ (I- u^{(k)}_j u^{(k)^*}_j)(A- \theta^{(k)}_j I)(I- u^{(k)}_j u^{(k)^*}_j) t^{(k)}= -r^{(k)}_j \] is described. The correction equation is efficient in a subspace orthogonal to the current eigenvector approximation. The use of approximations of the matrix \((A-\theta_j^{(k)}I)\) as a preconditioner for the subspace orthogonal to \(u^{(k)}_j\), the choice of the pair \((\theta^{(k)}_j, u^{(k)}_j)\), the choice of the preconditioner, the generalized eigenproblem, preconditioning for standard eigenproblems, harmonic Ritz values for standard eigenproblems, locally minimal residuals, harmonic Petrov values for generalized eigenproblems, and ill-conditioned preconditioners are discussed.