Classification of Samples via Neural-Network Augmented Two-Dimensional Infrared Spectroscopy.

Area of Science:

• Spectroscopy
• Artificial Intelligence
• Chemometrics

Background:

• Artificial neural networks (ANNs) are established tools for spectroscopic analysis.
• Two-dimensional infrared (2D IR) spectroscopy offers unique sample information.
• The integration of ANNs with 2D IR for sample classification remains largely unexplored.

Purpose of the Study:

• To investigate the application of ANNs for end-to-end classification of samples using 2D IR spectra.
• To develop and test ANN models for distinguishing samples based on spectral data.

Main Methods:

• Utilized artificial neural networks (ANNs) for analyzing two-dimensional infrared (2D IR) spectral data.
• Developed a binary classification model based on solvent type.
• Investigated classification using single spectral slices (pump-delay and waiting-time combinations).

Main Results:

• Successfully demonstrated binary classification of samples by solvent using 2D IR spectra and ANNs.
• Achieved accurate classification even with spectrally similar samples using specific spectral slices.
• Confirmed the feasibility of ANN-augmented 2D IR for sample analysis.

Conclusions:

• ANNs are effective for classifying samples based on 2D IR spectra.
• This approach enables high-throughput screening applications.
• The methodology holds significant potential for rapid and accurate chemical analysis.