Analysis of functional groups in atmospheric aerosols by infraredspectroscopy: sparse methods for statistical selection of relevant absorption bands

摘要

Various vibrational modes present in molecular mixtures of laboratory andatmospheric aerosols give rise to complex Fourier transform infrared (FT-IR)absorption spectra. Such spectra can be chemically informative, but they oftenrequire sophisticated algorithms for quantitative characterization of aerosolcomposition. Naïve statistical calibration models developed forquantification employ the full suite of wavenumbers available from a set ofspectra, leading to loss of mechanistic interpretation between chemicalcomposition and the resulting changes in absorption patterns that underpintheir predictive capability. Using sparse representations of the same set ofspectra, alternative calibration models can be built in which only a selectgroup of absorption bands are used to make quantitative prediction of variousaerosol properties. Such models are desirable as they allow us to relate predicted properties to their underlying molecularstructure. In this work, we present an evaluation of four algorithms forachieving sparsity in FT-IR spectroscopy calibration models. Sparsecalibration models exclude unnecessary wavenumbers from infrared spectraduring the model building process, permitting identification and evaluationof the most relevant vibrational modes of molecules in complex aerosolmixtures required to make quantitative predictions of various measures ofaerosol composition. We study two types of models: one which predicts alcoholCOH, carboxylic COH, alkane CH, and carbonyl CO functional group (FG)abundances in ambient samples based on laboratory calibration standards andanother which predicts thermal optical reflectance (TOR) organic carbon (OC)and elemental carbon (EC) mass in new ambient samples by direct calibrationof infrared spectra to a set of ambient samples reserved for calibration. Wedescribe the development and selection of each calibration model andevaluate the effect of sparsity on prediction performance. Finally, weascribe interpretation to absorption bands used in quantitative prediction ofFGs and TOR OC and EC concentrations.