Investigation of Statistical Analysis of Mass Spectra as a Tool for Characterization of Secondary Organic Aerosols

Secondary organic aerosols (SOA), which are formed from the oxidation of volatile organic compounds (VOCs) are hypothesized to have a significant role in climate processes and atmospheric phenomena due to their relative abundance in the atmosphere. VOCs are released into the atmosphere by natural and anthropogenic activities and can react with ozone (O3), the most abundant naturally occurring atmospheric oxidant, to yield hundreds of chemically distinct SOA products. It is of interest to improve the current understanding of the chemistry of SOA to better understand their impact on the atmosphere. Mass spectrometry has been an analytical method of focus for the study of SOA in the Petrucci Group. However, differentiating between the spectral peaks of SOA remains a significant challenge since their mass spectra are densely populated and contain many peaks in close proximity to each other. The ability to differentiate between SOA products is critical to the study of their chemistry. Statistical analysis of mass spectra data is proposed as a method to differentiate between SOA which originate from different chemical systems. The approach described herein utilizes the root mean square (RMS) error value as a measure of average difference between two spectra. A two-tailed student’s t-test is then used to determine whether a statistically significant difference exists between two chemical systems by comparing their average difference, as defined by the RMS value, to that of two chemical systems hypothesized to be the same. This approach was first applied to simulated mass spectra to optimize its parameters by investigating the effect of inherent variability and population density on the RMS value. The optimized approach was then applied to the spectra of SOA generated from α-pinene, a biogenic VOC of interest due to its high abundance in the atmosphere, by oxidation with atmospheric O3 and “dry” (in the absence of molecular oxygen) O3. A statistically significant difference between the spectra of SOA formed from atmospheric O3 and dry O3 was determined by this approach. While the results reported here show promise, continued optimization of the proposed approach is suggested to increase the chances of success of statistical analysis across a variety of chemical systems.