Hybrid algorithm for predicting the temperature-variation-induced wavelength drift of DBR semiconductor lasers

Area of Science:

• Optoelectronics
• Laser Physics
• Semiconductor Devices

Background:

• Distributed Bragg reflector (DBR) semiconductor lasers are crucial for optical communication.
• Wavelength drift due to ambient temperature variations affects laser stability and performance.
• Accurate prediction of wavelength drift is essential for precise wavelength control.

Purpose of the Study:

• To propose a hybrid algorithm for predicting wavelength drift in DBR semiconductor lasers.
• To enhance the prediction accuracy and generalization performance of existing models.
• To develop a model that can be integrated with tuning mechanisms for stable laser operation.

Main Methods:

• A hybrid algorithm combining a genetic algorithm (GA) for global search and an extreme learning machine (ELM) for supermapping.
• Utilizing tenfold cross-validation to optimize ELM parameters, including activation function and hidden layer nodes.
• Testing the proposed GA-ELM model with multiple datasets to evaluate its predictive capabilities.

Main Results:

• The GA-ELM model accurately and rapidly predicts wavelength drift.
• Achieved an average root-mean-square (rms) error of 4.09×10^-4 nm.
• Demonstrated an average mean absolute percentage error (MAPE) of 0.21%.

Conclusions:

• The proposed GA-ELM hybrid algorithm effectively predicts temperature-induced wavelength drift in DBR semiconductor lasers.
• The model overcomes the randomness of ELM and improves generalization performance.
• Future work aims to integrate this model with temperature and current tuning for rapid, stable wavelength control without extra devices.