Current chemical theory is incomplete to accurately predict product formation and structure from radical reactions under non-ideal conditions such as complex, multi-component environmental matrices. Radical reactions occur ubiquitously in the atmosphere and drive the complex, multiphase chemistry of volatile organic compound (VOC) oxidation to secondary organic aerosol (SOA) formation. SOA is a significant constituent of fine particulate matter (PM2.5) that is associated with adverse health outcomes. Oxidation pathways that yield SOA are studied extensively. However, a complete understanding of the mechanistic steps in the evolution of VOCs to particle mass remains elusive. For example, the full mechanistic pathways for side reactions are relatively neglected and identification of individual SOA contributors is often incomplete. Techniques to chemically characterize SOA often involve mass spectra deconvolution, as ambient and laboratory samples predominantly contain unidentified peaks. Understudied side products of VOC oxidation include water-soluble organic carbon (WSOC). Measurements of WSOC via wet chemical oxidation (WCO) are subject to interference in the presence of salts, which are abundant in nearly all ambient aquatic samples. Further, ambient PM2.5 mass measured by the federal reference method/federal equivalence method (FRM/FEM) are prone to negative sampling artifacts through loss of semi-volatile species such as WSOC and ammonium nitrate. This dissertation uses elementary step chemical reactions, kinetic modeling and thermodynamic calculations to elucidate key uncertainties in atmospheric measurements of PM2.5 and WSOC. Specifically, I 1) manually curate thousands of elementary-step radical reaction mechanisms and apply deep learning methods to predict the stepwise mechanistic oxidation of VOCs to SOA; 2) develop plausible mechanisms and a kinetic model that describes chloride interferences in WCO methods for analysis of formic and acetic acid; and 3) thermodynamically estimate and analyze ammonium nitrate volatilization from PM2.5 Teflon filters, focusing on California monitoring locations that do not attain the PM2.5 National Ambient Air Quality Standard (NAAQS) such as the San Joaquin Valley. The radical reaction database predicts single-step radical reactions and fosters collaboration by enabling users to add reactions. Plausible mechanisms postulate the formation of stable chlorinated byproducts during WCO of acetic acid and not formic acid, and the kinetic model reproduces empirical measurements in close agreement. Independent experiments with other halogenated salts support the postulated chloride elementary-step mechanisms. This work suggests up to 20% of PM2.5 is lost during ambient sampling in regulatory PM2.5 networks and highlights the application of fundamental chemistry to plausibly explain uncertainties in measurements of WSOC and PM2.5 in order to better understand environmental burdens of pollution.