1 Insights on Atmospheric Oxidation Processes by Performing Factor Analyses on 1 Subranges of Mass Spectra 2

14 With the recent developments in mass spectrometry, combined with the strengths of factor analysis 15 techniques, our understanding of atmospheric oxidation chemistry has improved significantly. The 16 typical approach for using techniques like positive matrix factorization (PMF) is to input all measured 17 data for the factorization in order to separate contributions from different sources and/or processes to 18 the total measured signal. However, while this is a valid approach for assigning the total signal to 19 factors, we have identified several cases where useful information can be lost if solely using this 20 approach. For example, gaseous molecules emitted from the same source can show different temporal 21 behaviors due differing loss terms, like condensation at different rates due to different molecular 22 masses. This conflicts with one of PMF’s basic assumptions of constant factor profiles. In addition, 23 some ranges of a mass spectrum may contain useful information, despite contributing only minimal 24 fraction to the total signal, in which case they are unlikely to have a significant impact on the 25 factorization result. Finally, certain mass ranges may contain molecules formed via pathways not 26 available to molecules in other mass ranges, e.g. dimeric species versus monomeric species. In this 27 study, we attempted to address these challenges by dividing mass spectra into sub-ranges and 28 applying the newly developed binPMF method to these ranges separately. We utilized a dataset from 29 a chemical ionization atmospheric pressure interface time-of-flight (CI-APi-TOF) mass spectrometer 30 as an example. We compare the results from these three different ranges, each corresponding to 31 molecules of different volatilities, with binPMF results from the combined range. Separate analysis 32 showed clear benefits in dividing factors for molecules of different volatilities more accurately, in 33 https://doi.org/10.5194/acp-2019-838 Preprint. Discussion started: 28 November 2019 c © Author(s) 2019. CC BY 4.0 License.