Angle-resolved scatterometry combined with deep learning assisted in-situ monitoring of nanowire doping in molecular beam epitaxy process.

The sub-atomic precision of molecular beam epitaxy (MBE) allows for highly flexible elemental doping in nanowires (NWs). Optimizing doping quality for specific elements requires a comprehensive understanding of the relationship between process parameters and doping concentrations. This necessitates in-situ monitoring of the doping process to define the corresponding process window. However, the reflection high-energy electron diffraction (RHEED) technique, commonly used during MBE growth, has limited sensitivity to atomic arrangement changes caused by doping and is primarily capable of monitoring the structural quality of the sample. To address this limitation, we propose a nanowire doping concentration measurement method based on angle-resolved scatterometry (ARS). This method captures scattering information across the full angular range of NWs, allowing for high-resolution measurement of doping concentration. Using GaN NWs and AlN films doped with Si as a case study, we measured the Si concentration at different doping temperatures. The results demonstrate that the proposed method achieves a doping concentration resolution of 0.01% and 0.06% within the investigated temperature range. Furthermore, we employed deep learning to establish the relationship between angle-resolved reflectivity and nominal doping concentration. The predictive results indicate that the measurement error is maintained below 0.027%. We also validated the robustness of the method across multiple measurement wavelengths and explored the feasibility of using reduced angle reflectance for neural network training. This work paves the way for in-situ monitoring of nanowire doping processes through ARS, significantly enhancing doping control precision in MBE growth.