Abstract
In this work, a combinatorial approach to the synthesis of magnetic multilayers is explored. Combinatorial libraries of Co/Pd multilayer thin films were prepared using off-axis magnetron sputtering to enable thickness gradients across the wafer-magnetic properties of the multilayers are controlled by the thicknesses of Co and Pd layers in the repeated bi-layer stack. Polar magneto-optical Kerr effect (MOKE) was used to map magnetic properties of the combinatorial libraries. Multivariate regression analysis and back-propagation neural network (neural networks are known to better handle nonlinear approximations) were used to analyze the combinatorial data and to enable predictive capabilities. In the multivariate analysis, the relationship between the descriptive and output variables was approximated by a second order polynomial of Co and Pd thickness.The neural network model was utilized inversely to design a multilayer with pre-determined magnetic properties.