Physics-inspired neural networks for tailored ripple-free optical needle beams

In this work, we present a physics-inspired neural network (PINN) inverse design framework for generating ripple-free optical needle beams with tunable characteristics, including position, length, and spot size. In contrast to conventional Bessel-like beam designs, where needle length and spot size are inherently coupled and significant intensity ripples and central singularities degrade beam quality and transmission efficiency, our approach enables independent control of beam parameters, allowing tailored profiles for specific applications. The inverse design process eliminates the central singularity present in traditional methods, resulting in a simulated one-order-of-magnitude improvement in power transmission efficiency. Experimental implementation using amplitude-based spatial light modulators (SLMs) confirms the generation of clean, ripple-free beams, without the energy loss seen in Bessel-like designs. Additionally, the generated beams exhibit robust self-healing under partial obstruction, highlighting their resilience in complex environments. These results establish the proposed PINN-based inverse design framework as a powerful and flexible tool for advanced structured light engineering.