Design of the pulse compression fiber taper based on deep learning

Area of Science:

• Photonics
• Optical Engineering
• Nonlinear Optics

Background:

• Generating ultrashort femtosecond pulses is crucial for various scientific applications.
• Existing methods for pulse compression often require complex setups or extensive design time.
• Photonic crystal fibers offer unique properties for nonlinear optical effects like pulse compression.

Purpose of the Study:

• To propose a design method for pulse compression fiber tapers.
• To enable rapid development of tapers optimized for specific laser outputs.
• To achieve continuous self-similar pulse compression.

Main Methods:

• Utilizing a deep neural network to predict fiber cross-sectional characteristics.
• Employing a global search algorithm based on soliton conditions.
• Iteratively calculating structural parameters for cascaded fiber segments.

Main Results:

• Successfully compressed a 1.05 picosecond (ps) pulse to 49.1 femtoseconds (fs) in a 0.276-meter fiber.
• Achieved a high pulse compression factor of 21.55.
• Demonstrated a fast and feasible solution for femtosecond pulse generation.

Conclusions:

• The proposed deep neural network-guided design method accelerates the development of pulse compression fiber tapers.
• This approach is effective for generating femtosecond pulses with high compression ratios.
• The method holds significant reference value for designing irregular optical waveguides.