Modelling laser machining of nickel with spatially shaped three pulse sequences using deep learning

Area of Science:

• Materials Science
• Laser Physics
• Computational Modeling

Background:

• Femtosecond laser machining involves high peak intensities, making microscale prediction challenging.
• Multiple exposures in laser machining require accounting for prior surface modifications.
• Developing predictive models for complex laser-matter interactions is crucial for advanced manufacturing.

Purpose of the Study:

• To develop a neural network model for predicting surface quality in multi-exposure femtosecond laser machining.
• To accurately model photon-atom interactions and cumulative surface effects.
• To enable precise fabrication of microscale structures with controlled surface morphology.

Main Methods:

• Utilized a neural network approach for automated model creation without requiring prior physical process knowledge.
• Applied the model to a 5µm electroless nickel layer on copper.
• Employed spatially shaped laser pulses via a spatial light modulator for sequential, overlapping exposures with varying intensities.

Main Results:

• The neural network accurately predicted the surface profile after three sequential, overlapping exposures.
• Successfully reproduced sub-diffraction limit machining effects achievable with multiple exposures.
• Demonstrated the model's ability to capture the smoothing of edge-burr from previous exposures.

Conclusions:

• Neural network modeling offers a powerful tool for predicting surface quality in complex multi-exposure femtosecond laser machining.
• This approach overcomes limitations of traditional physics-based models for microscale fabrication.
• The developed method facilitates precise control over surface topography in advanced laser processing applications.