Application of non-linear optimization methods to the estimation of multivariate curve resolution solutions and of their feasible band boundaries in the investigation of two chemical and environmental simulated data sets.

Although alternating least squares algorithms have revealed extremely useful and flexible to solve multivariate curve resolution problems, other approaches based on non-linear optimization algorithms using non-linear constraints are possible. Once the subspaces defined by PCA solutions are identified, appropriate rotation and perturbation of these solutions can produce solutions fulfilling the constraints obeyed by the physical nature of the investigated systems. In order to perform such a rotation, an optimization algorithm based in the fulfillment of constraints and some examples of application in chemistry and environmental chemistry are given. It is shown that the solutions obtained either by alternating least squares or by the new proposed algorithm are rather similar and that they are both within the boundaries of the band of feasible solutions obtained by an algorithm previously developed to estimate them.