Insights into the Machine Learning Predictions of the Optical Response of Plasmon@Semiconductor Core-Shell Nanocylinders
The application domain of deep learning (DL) has been extended into the realm of nanomaterials, photochemistry, and optoelectronics research. Here, we used the combination of a computer vision technique, namely convolutional neural network (CNN), with multilayer perceptron (MLP) to obtain the far-fi...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-03-01
|
| Series: | Photochem |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2673-7256/3/1/10 |
| _version_ | 1797609483997806592 |
|---|---|
| author | Ehsan Vahidzadeh Karthik Shankar |
| author_facet | Ehsan Vahidzadeh Karthik Shankar |
| author_sort | Ehsan Vahidzadeh |
| collection | DOAJ |
| description | The application domain of deep learning (DL) has been extended into the realm of nanomaterials, photochemistry, and optoelectronics research. Here, we used the combination of a computer vision technique, namely convolutional neural network (CNN), with multilayer perceptron (MLP) to obtain the far-field optical response at normal incidence (along cylinder axis) of concentric cylindrical plasmonic metastructures such as nanorods and nanotubes. Nanotubes of Si, Ge, and TiO<sub>2</sub> coated on either their inner wall or both their inner and outer walls with a plasmonic noble metal (Au or Ag) were thus modeled. A combination of a CNN and MLP was designed to accept the cross-sectional images of cylindrical plasmonic core-shell nanomaterials as input and rapidly generate their optical response. In addition, we addressed an issue related to DL methods, namely explainability. We probed deeper into these networks’ architecture to explain how the optimized network could predict the final results. Our results suggest that the DL network learns the underlying physics governing the optical response of plasmonic core-shell nanocylinders, which in turn builds trust in the use of DL methods in materials science and optoelectronics. |
| first_indexed | 2024-03-11T06:01:02Z |
| format | Article |
| id | doaj.art-3cd2719264fc4129b2e8183ce00c5b78 |
| institution | Directory Open Access Journal |
| issn | 2673-7256 |
| language | English |
| last_indexed | 2024-03-11T06:01:02Z |
| publishDate | 2023-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Photochem |
| spelling | doaj.art-3cd2719264fc4129b2e8183ce00c5b782023-11-17T13:18:02ZengMDPI AGPhotochem2673-72562023-03-013115517010.3390/photochem3010010Insights into the Machine Learning Predictions of the Optical Response of Plasmon@Semiconductor Core-Shell NanocylindersEhsan Vahidzadeh0Karthik Shankar1Department of Electrical and Computer Engineering, University of Alberta, 9211-116 St., Edmonton, AB T6G 1H9, CanadaDepartment of Electrical and Computer Engineering, University of Alberta, 9211-116 St., Edmonton, AB T6G 1H9, CanadaThe application domain of deep learning (DL) has been extended into the realm of nanomaterials, photochemistry, and optoelectronics research. Here, we used the combination of a computer vision technique, namely convolutional neural network (CNN), with multilayer perceptron (MLP) to obtain the far-field optical response at normal incidence (along cylinder axis) of concentric cylindrical plasmonic metastructures such as nanorods and nanotubes. Nanotubes of Si, Ge, and TiO<sub>2</sub> coated on either their inner wall or both their inner and outer walls with a plasmonic noble metal (Au or Ag) were thus modeled. A combination of a CNN and MLP was designed to accept the cross-sectional images of cylindrical plasmonic core-shell nanomaterials as input and rapidly generate their optical response. In addition, we addressed an issue related to DL methods, namely explainability. We probed deeper into these networks’ architecture to explain how the optimized network could predict the final results. Our results suggest that the DL network learns the underlying physics governing the optical response of plasmonic core-shell nanocylinders, which in turn builds trust in the use of DL methods in materials science and optoelectronics.https://www.mdpi.com/2673-7256/3/1/10energysensingphotocatalysisin-silico designclassificationoptimization |
| spellingShingle | Ehsan Vahidzadeh Karthik Shankar Insights into the Machine Learning Predictions of the Optical Response of Plasmon@Semiconductor Core-Shell Nanocylinders Photochem energy sensing photocatalysis in-silico design classification optimization |
| title | Insights into the Machine Learning Predictions of the Optical Response of Plasmon@Semiconductor Core-Shell Nanocylinders |
| title_full | Insights into the Machine Learning Predictions of the Optical Response of Plasmon@Semiconductor Core-Shell Nanocylinders |
| title_fullStr | Insights into the Machine Learning Predictions of the Optical Response of Plasmon@Semiconductor Core-Shell Nanocylinders |
| title_full_unstemmed | Insights into the Machine Learning Predictions of the Optical Response of Plasmon@Semiconductor Core-Shell Nanocylinders |
| title_short | Insights into the Machine Learning Predictions of the Optical Response of Plasmon@Semiconductor Core-Shell Nanocylinders |
| title_sort | insights into the machine learning predictions of the optical response of plasmon semiconductor core shell nanocylinders |
| topic | energy sensing photocatalysis in-silico design classification optimization |
| url | https://www.mdpi.com/2673-7256/3/1/10 |
| work_keys_str_mv | AT ehsanvahidzadeh insightsintothemachinelearningpredictionsoftheopticalresponseofplasmonsemiconductorcoreshellnanocylinders AT karthikshankar insightsintothemachinelearningpredictionsoftheopticalresponseofplasmonsemiconductorcoreshellnanocylinders |