Optimizing the microstructure of femtosecond laser induced forward transfer tungsten by artificial neural network

Abstract This work demonstrates the successful transfer of refractory tungsten metal via femtosecond laser-induced forward transfer (fs-LIFT) through key process parameters—laser power, repetition frequency, defocus distance, and pulse number. Experimental analysis reveals that laser power, frequency, and pulse number collectively govern morphological evolution by modulating deposited energy, driving transitions from ineffective transfer to effective transfer and eventually substrate ablation as energy increases. An artificial neural network was introduced to classify and predict transfer morphologies under varying laser parameters. Among the models evaluated, the particle swarm optimization-enhanced backpropagation neural network achieved superior performance, it effectively captures the complex nonlinear relationships between laser parameters and morphological outcomes, enabling accurate identification of the process window for high-quality transfer. This approach establishes a reliable and efficient strategy for the controllable fabrication of tungsten microstructures using fs-LIFT.