Computational chromatography: A machine learning strategy for demixing individual chemical components in complex...

Area of Science:

• Analytical Chemistry
• Environmental Science
• Spectroscopy

Background:

• Surface-enhanced Raman spectroscopy (SERS) offers sensitive chemical detection for environmental and biological contaminants.
• Polycyclic aromatic hydrocarbons (PAHs) are priority pollutants found globally, posing significant human health risks.
• PAHs often exist in complex mixtures, challenging individual compound identification with traditional methods.

Purpose of the Study:

• To investigate the potential of combining SERS with machine learning (ML) for identifying individual PAHs within multicomponent mixtures.
• To develop and evaluate an unsupervised ML algorithm for spectral deconvolution of complex PAH samples.

Main Methods:

• Development of an unsupervised ML algorithm named Characteristic Peak Extraction for dimensionality reduction.
• Analysis of SERS spectra from two- and four-component PAH mixtures with varying concentration ratios.
• Extraction of individual component spectra from mixtures for identification against a SERS spectral library.

Main Results:

• The Characteristic Peak Extraction algorithm successfully extracted individual PAH spectra from unknown mixtures.
• The method demonstrated the ability to identify specific PAHs even when present in mixtures with diverse concentration ratios.
• Successful deconvolution of two- and four-component PAH mixtures was achieved.

Conclusions:

• Combining SERS with ML provides a powerful strategy for identifying individual chemical components in complex mixtures.
• This approach represents a significant advancement towards computational demixing of unknown contaminants in environmental samples.
• The developed ML method shows promise for streamlined, on-site detection of priority pollutants like PAHs.