Supervised machine learning prediction and investigation of nonlinear optical rectification in Ge/Si0.15Ge0.85 asymmetric coupled triangle quantum wells
Nonlinear optical rectification (NOR) is a fundamental process in the design and engineering of photonic devices. Predicting the NOR coefficient in the Ge/Si0.15Ge 0.85 asymmetric double triangle quantum wells (ADTQWs) structures contributes significantly to improving the performance and efficiency...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | Results in Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590123025026581 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Nonlinear optical rectification (NOR) is a fundamental process in the design and engineering of photonic devices. Predicting the NOR coefficient in the Ge/Si0.15Ge 0.85 asymmetric double triangle quantum wells (ADTQWs) structures contributes significantly to improving the performance and efficiency of such devices. In this study, the NOR coefficient in the Ge/Si0.15Ge 0.85 ADTQW is calculated using the compact matrix density (CMD) approach after solving the Schrodinger equation under the effective mass approximation (EMA). The calculations are performed for various values of the right-well width LR. The theoretical data set results are subsequently used to train three machine learning (ML) models, such as artificial neural network (ANN-ML), convolutional neural network (CNN-ML), and Decision Tree (DT-ML), to predict the NOR coefficient based on the structural parameters. The dataset is split into 80 % for training and 20 % for internal model evaluation. The performances of these models are evaluated using standard evaluation metrics. Among the three ML models, the DT model yields the most accurate predictions, with RMSE values between 0.0038 and 0.0053 and MAE values between 0.0020 and 0.0027 across all considered LR values. These low error values demonstrate a strong agreement with the theoretical calculation, validating the reliability of the proposed ML-based approach. This methodology provides an effective and efficient tool for the design and optimization of ADTQW in future photonic engineering applications. |
|---|---|
| ISSN: | 2590-1230 |