Measurement of Nb L αβ characteristic x-ray production cross-sections by 5.5–9 keV positron impact and advancements in the numerical-neural network method for solving the thick-target inverse problem

Abstract Cross-section data for inner-shell ionization by low-energy positron or electron impact are crucial for understanding interaction mechanisms between incident projectiles and atoms near the threshold energy. However, such experimental data are scarce, and results from different experimental methods also show significant discrepancies, which highlights the critical need for the development of advanced measurement techniques. In this study, the yields of Nb L αβ characteristic x-ray were measured from a thick Nb target bombarded by 5.5–9 keV positrons. Additionally, we improved the Numerical-Neural Network method proposed by Wu et al from our group. Using this improved method, we inverted the thick-target yield data to obtain the Nb L αβ characteristic x-ray production cross-sections induced by 5.5–9 keV positrons. In parallel, we processed the experimental Al K-shell characteristic x-ray yields from a thick Al target bombarded by 5–27 keV electrons, which were published by Li et al from Sichuan University. From these data, we derived the Al K-shell ionization cross-sections. The derived cross-sections for both elements were subsequently compared with results from the MC-Neural Network method and theoretical distorted-wave born approximation (DWBA) predictions. The results show that the experimental cross-sections from the two neural network methods are in good agreement, with deviations within 5%. For Nb L αβ characteristic x-ray production cross-sections (5.5–9 keV positron impact), the DWBA theoretical predictions are 20%–30% lower than the experimental values from our improved Numerical-Neural Network method, and this experimental cross-section data was obtained for the first time. The Al K-shell ionization cross-sections (5–27 keV electron impact), derived using our improved numerical-neural network method, agree well with published ratio-method results, while corresponding DWBA predictions are 7%–14% lower than our values.