Excitation Emission Matrix Fluorescence Spectroscopy for Combustion Generated Particulate Matter Source...

Area of Science:

• Environmental Science
• Analytical Chemistry
• Public Health

Background:

• Particulate matter (PM) inhalation poses significant health risks, including cardio-pulmonary diseases and respiratory issues.
• Accurate identification of PM sources is vital for developing effective mitigation strategies but remains challenging with current methods.

Purpose of the Study:

• To develop and validate a novel method for identifying particulate matter sources using excitation emission matrix (EEM) fluorescence spectroscopy and machine learning.
• To assess the accuracy and detection limits of this method for common combustion sources.

Main Methods:

• Collected PM2.5 from wood burning, diesel exhaust, and cigarettes on filters.
• Extracted PM into cyclohexane and analyzed using EEM fluorescence spectroscopy.
• Trained a convolutional neural network (CNN) with spectral data for source identification.

Main Results:

• The CNN achieved an 89% accuracy in identifying the three laboratory sources in mixed samples.
• Detection limits were determined for cigarette (0.7 μg/m³), wood (0.9 μg/m³), and diesel (2.6 μg/m³) PM over 24 hours.
• Initial application to field samples showed promise but indicated a need for larger training datasets.

Conclusions:

• EEM fluorescence spectroscopy combined with CNN offers a promising approach for particulate matter source apportionment.
• The method's accuracy and sensitivity support its potential for informing environmental and health policies.
• Further development with diverse field samples is recommended for broader applicability.