Non-Destructive 3D Elemental Characterization of Multilayer Materials by ANN-Assisted Ion Beam Analysis

Patterned and multilayer materials used in advanced technologies exhibit complex three-dimensional compositional architectures in which buried interfaces and elemental gradients critically influence performance. However, most non-destructive analytical techniques remain largely surface-sensitive, limiting access to subsurface information in opaque systems. In this work, we present a novel framework for non-destructive three-dimensional elemental characterization based on the integration of artificial neural networks with ion beam analysis techniques, namely, Particle-Induced X-ray Emission (PIXE) and Elastic Backscattering Spectrometry (EBS). The proposed approach enables the reconstruction of depth-resolved 3D elemental distributions by combining complementary spectral information with data-driven analysis. The methodology is demonstrated on a GaSb thermophotovoltaic device featuring multilayer metallic contacts, where the elemental distribution beneath thick gold layers is revealed for the first time. The neural network approach overcomes limitations associated with low counting statistics in pixel-resolved spectra, enhancing sensitivity and enabling reliable classification of compositional features. The fusion of PIXE-derived lateral information with EBS-based depth profiling enables full three-dimensional visualization and quantitative and qualitative mapping of elemental distributions. Beyond the specific case study presented, this approach provides a general and scalable strategy for 3D compositional analysis of complex materials, including systems containing both heavy and light elements. The results highlight the potential of combining advanced data-driven methods with ion beam techniques to expand the capabilities of non-destructive characterization, with broad applicability in energy, electronics, and functional materials.