Date of Award

Winter 12-15-2015

Author's School

School of Engineering & Applied Science

Author's Department

Energy, Environmental & Chemical Engineering

Degree Name

Doctor of Philosophy (PhD)

Degree Type

Dissertation

Abstract

Atmospheric organic aerosols are composed of thousands of individual compounds, interacting with climate through changes in aerosol optical properties and cloud interactions, and can be detrimental to human health. Aerosol mass spectrometry (MS) and gas chromatography (GC)-separated MS measurements have been utilized to better characterize the chemical composition of this material that comes from a variety of sources and experiences continuous oxidation while in the atmosphere. This dissertation describes the development of a novel rapid data analysis method for grouping of major components within chromatography-separated measurements and first application using thermal desorption aerosol gas chromatograph (TAG) – MS data. Chromatograms are binned and inserted directly into a positive matrix factorization (PMF) analysis to determine major contributing components, eliminating the need for manual compound integrations of hundreds of resolved molecules, and incorporating the entirety of the eluting MS signal, including Unresolved Complex Mixtures (UCM) and decomposition products that are often ignored in traditional GC-MS analysis.

Binned GC-MS data has three dimensions: (1) mass spectra index m/z, (2) bin number, and (3) sample number. PMF output is composed of two dimensions; factor profiles and factor time series. The specific arrangement of the input data (three dimensions of variation structured as a two dimensional matrix) in a two dimensional PMF analysis affects the structure of the PMF profiles and time series output. If mass spectra index is in the profile dimension, and bin number and sample number are in the time series dimension, PMF groups components into factors with similar mass spectra, such as major contributing individual compounds, UCM with similar functional composition, and homologous compound series. This type of PMF analysis is described as the binning method for chromatogram deconvolution, and is presented in Chapter 2. If the sample number is in the time series dimension, and the bin number and mass spectra index, arranged as mass spectra resolved retention time/chromatogram (bin number), are in the profile dimension, PMF groups components with similar time series trends. This type of PMF analysis is described as binning method for source apportionment, and is described in Chapter 3.

The binning methods are compared to traditional compound integration methods using previously-collected hourly ambient samples from Riverside, CA during the 2005 Study of Organic Aerosols at Riverside (SOAR) field campaign, as discussed in Chapters 2-3. Further application of the binning method for source apportionment is performed on newly acquired hourly TAG data from East St. Louis, IL, operated as part of the 2013 St. Louis Air Quality Regional Study (SLAQRS). Major sources of biogenic secondary organic aerosol (SOA), anthropogenic primary organic aerosol (POA) were identified, as described in detail in Chapter 4. Finally, our PMF separation method was tested for reliability using primary and secondary sources in a controlled laboratory system. As shown in Chapter 5, we find that for application of PMF on receptor measurements, high signal intensity and unique measurement profiles, like those found in TAG chromatograms, are keys to successful source apportionment. The binning method with component separation by PMF may be a valuable analysis technique for other complex data sets that incorporate measurements (e.g., mass spectrometry, spectroscopy, etc.) with additional separations (e.g., volatility, hygroscopicity, electrical mobility, etc.).

Language

English (en)

Chair

Brent Williams

Committee Members

Pratim Biswas, Rajan Chakrabarty, David Fike, Jose Jimenez, Jay Turner

Comments

Permanent URL: http://doi.dx.org/10.7936/K7W37TKJ

Included in

Engineering Commons

Share

COinS