Nondestructive initial-profile-free 3D elemental mapping in multilayer thin film structures based on EDX and a quadratic programming problem.

We have demonstrated a new data analysis method that enables nondestructive depth profiling of a multilayer thin-film sample from energy-dispersive X-ray spectroscopy (EDX) data without the assumption of initial profiles. This method is based on a quadratic programming problem and allows for three-dimensional elemental mapping in the sample without destroying it, by performing depth profiling for all the pixels in the EDX two-dimensional mapping data. In this paper, first nondestructive depth profiling of two samples with different multilayer structures was performed using the proposed method. The results were compared with those obtained by cross-sectional observation to validate the accuracy and usefulness of the proposed method. Next, an example of the three-dimensional elemental mapping based on the proposed method was demonstrated. This method allows us to nondestructively obtain three-dimensional elemental distribution within a sample over a wide area on the order of mm, which is impossible to obtain using other analytical methods. The way to determine the hyperparameters, which significantly affects the calculation results, is fully described in this paper.