Related Experiment Video
Updated: Jun 12, 2026
09:38
Characterizing Far-infrared Laser Emissions and the Measurement of Their Frequencies
Published on: December 18, 2015
Machine Learning Models for Predicting Key Performance Characteristics of High-Temperature THz Quantum Cascade Lasers
Mihailo Stojković1, Novak Stanojević1, Aleksandar Milićević2
1School of Electrical Engineering, University of Belgrade, Bulevar kralja Aleksandra 73, 11120 Belgrade, Serbia.
Nanomaterials (Basel, Switzerland)
|June 11, 2026
Summary
Machine learning models, including Artificial Neural Networks (ANN), accurately predict quantum cascade laser (QCL) performance. This approach accelerates the optimization of QCL designs by enabling rapid evaluation of millions of configurations.
Area of Science:
- Semiconductor device physics
- Computational modeling
- Machine learning applications
Background:
- Quantum cascade lasers (QCLs) are crucial optoelectronic devices.
- Predicting QCL performance characteristics (material gain, current density, emission frequency) typically requires computationally intensive simulations.
- Optimizing QCL design involves navigating a high-dimensional configuration space.
Purpose of the Study:
- To develop a machine learning (ML)-based surrogate modeling framework for predicting QCL performance characteristics.
- To enable faster evaluation and optimization of QCL designs compared to traditional simulation methods.
- To demonstrate the effectiveness of ML models in predicting device performance and identifying optimal designs.
Main Methods:
- Applied Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Artificial Neural Networks (ANN) for predictive modeling.
- Generated training data using a numerical simulator based on the density-matrix transport model.
- Evaluated model performance using standardized Root Mean Square Error (RMSE) values.
Main Results:
- ANN models demonstrated superior predictive performance with the lowest RMSE values (0.04 to 0.22) across all investigated characteristics.
- The ANN framework enabled the evaluation of approximately 44 million configurations in about five minutes, achieving a ~90,000x speedup.
- Identified QCL designs with improved material gain and facilitated efficient parameter optimization.
Conclusions:
- Combining physics simulations with ML, specifically ANNs, provides a reliable and highly efficient method for predicting QCL performance.
- The ML-based surrogate modeling framework significantly accelerates the design and optimization process for QCLs.
- This approach facilitates the exploration of vast design spaces, leading to improved device performance and faster innovation in QCL technology.
