Nitroaromatic explosives' detection and quantification using an attention-based transformer on surface-enhanced Raman...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Artificial Intelligence

Background:

• Nitroaromatic explosives detection is crucial for security.
• Surface-Enhanced Raman Spectroscopy (SERS) offers high sensitivity for explosives detection.
• Current SERS analysis often involves time-consuming preprocessing of spectral data.

Purpose of the Study:

• To develop a novel method for detecting and quantifying nitroaromatic explosives using raw SERS maps.
• To overcome limitations of traditional SERS data preprocessing.
• To enhance the accuracy and efficiency of explosives detection.

Main Methods:

• An attention-based vision transformer neural network was developed.
• The model takes raw SERS maps as direct input, eliminating preprocessing.
• New SERS datasets for 2,4-dinitrophenols (DNP) and picric acid (PA) were created, alongside a 4-nitrobenzenethiol (4-NBT) dataset.

Main Results:

• The proposed AI approach demonstrated competitive or superior performance in detection and quantification accuracy compared to existing methods.
• The model successfully detected analytes down to 1 nM concentrations.
• Attention maps generated by the model identified regions with high signal-to-noise ratios in SERS maps.

Conclusions:

• Utilizing raw SERS maps with an attention-based vision transformer is a promising alternative for explosives detection and quantification.
• This method reduces computational cost and avoids information loss associated with spectral averaging.
• The approach offers a more efficient and potentially more accurate pathway for explosives analysis.