Numerical solutions for large sparse quadratic eigenvalue problems (Q1899377): Difference between revisions

An algorithm is given for computing few of the smallest positive eigenvalues of large sparse quadratic eigenvalue problems. The authors state that the proposed algorithm sometimes requires more computational time than other presently known methods, but it guarantees convergence to the smallest positive eigenvalue, even in the case when the reciprocal of this eigenvalue is surrounded by the reciprocals of many complex eigenvalues. The convergence can be accelerated by Newton iterations having quadratic convergence. The algorithm has a reasonable stopping criterion. The proposed algorithm can be regarded as a global convergence method for computing a desired eigenvalue.