Deep learning for classifying quantum emission signals in WS₂ monolayers using wavelet transform

Area of Science:

• Quantum Materials Science
• Machine Learning Applications
• Spectroscopy

Background:

• Characterizing quantum emission signals from materials like WS₂ monolayer nanobubbles is crucial but challenging.
• Assessing spectral distinguishability is key for quantum information applications.

Purpose of the Study:

• To develop and evaluate deep learning models for classifying quantum emission signals from WS₂ monolayer nanobubbles.
• To assess the performance of different convolutional neural network architectures for this classification task.

Main Methods:

• Quantum emission signals were preprocessed and transformed into RGB images using Continuous Wavelet Transform (CWT).
• Three CNN architectures (ResNet50, VGG16, Xception) were trained and evaluated using fivefold cross-validation.
• Classification accuracy was assessed across different spectral band combinations.

Main Results:

• All evaluated models achieved high classification accuracy, with VGG16 reaching 99.4% mean accuracy.
• Perfect accuracy was observed for spectrally distant bands, while adjacent bands presented a greater challenge (96.5% for VGG16).
• Xception demonstrated high computational efficiency, converging in as few as 2 epochs.

Conclusions:

• Deep learning combined with CWT offers a robust framework for quantum emission signal classification.
• This approach has significant implications for quantum photonics, cryptography, and sensing.
• The study addresses data scarcity in quantum systems via transfer learning, paving the way for future quantum technology development.